{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 20027, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.327787021630616e-08, "loss": 2.6513, "step": 1 }, { "epoch": 0.0, "learning_rate": 6.655574043261232e-08, "loss": 2.5868, "step": 2 }, { "epoch": 0.0, "learning_rate": 9.983361064891847e-08, "loss": 3.2392, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.3311148086522465e-07, "loss": 2.7944, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.6638935108153078e-07, "loss": 2.4684, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.9966722129783695e-07, "loss": 2.4611, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.329450915141431e-07, "loss": 3.0208, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.662229617304493e-07, "loss": 2.934, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.9950083194675546e-07, "loss": 3.2858, "step": 9 }, { "epoch": 0.0, "learning_rate": 3.3277870216306157e-07, "loss": 2.7369, "step": 10 }, { "epoch": 0.0, "learning_rate": 3.660565723793678e-07, "loss": 2.4243, "step": 11 }, { "epoch": 0.0, "learning_rate": 3.993344425956739e-07, "loss": 2.9602, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.326123128119801e-07, "loss": 3.1674, "step": 13 }, { "epoch": 0.0, "learning_rate": 4.658901830282862e-07, "loss": 2.7131, "step": 14 }, { "epoch": 0.0, "learning_rate": 4.991680532445924e-07, "loss": 2.2809, "step": 15 }, { "epoch": 0.0, "learning_rate": 5.324459234608986e-07, "loss": 2.5953, "step": 16 }, { "epoch": 0.0, "learning_rate": 5.657237936772047e-07, "loss": 3.0426, "step": 17 }, { "epoch": 0.0, "learning_rate": 5.990016638935109e-07, "loss": 2.83, "step": 18 }, { "epoch": 0.0, "learning_rate": 6.32279534109817e-07, "loss": 2.4139, "step": 19 }, { "epoch": 0.0, "learning_rate": 6.655574043261231e-07, "loss": 2.6351, "step": 20 }, { "epoch": 0.0, "learning_rate": 6.988352745424292e-07, "loss": 3.0026, "step": 21 }, { "epoch": 0.0, "learning_rate": 7.321131447587356e-07, "loss": 2.8643, "step": 22 }, { "epoch": 0.0, "learning_rate": 7.653910149750417e-07, "loss": 2.756, "step": 23 }, { "epoch": 0.0, "learning_rate": 7.986688851913478e-07, "loss": 2.4995, "step": 24 }, { "epoch": 0.0, "learning_rate": 8.319467554076539e-07, "loss": 2.6035, "step": 25 }, { "epoch": 0.0, "learning_rate": 8.652246256239602e-07, "loss": 2.948, "step": 26 }, { "epoch": 0.0, "learning_rate": 8.985024958402663e-07, "loss": 2.7223, "step": 27 }, { "epoch": 0.0, "learning_rate": 9.317803660565724e-07, "loss": 2.6113, "step": 28 }, { "epoch": 0.0, "learning_rate": 9.650582362728786e-07, "loss": 2.9182, "step": 29 }, { "epoch": 0.0, "learning_rate": 9.983361064891848e-07, "loss": 2.8912, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.031613976705491e-06, "loss": 2.4298, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.0648918469217972e-06, "loss": 2.4442, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.0981697171381033e-06, "loss": 3.1673, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.1314475873544094e-06, "loss": 2.3198, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.1647254575707155e-06, "loss": 2.9716, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.1980033277870218e-06, "loss": 3.0616, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.231281198003328e-06, "loss": 2.7379, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.264559068219634e-06, "loss": 2.6423, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.2978369384359402e-06, "loss": 3.33, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.3311148086522463e-06, "loss": 2.7595, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.3643926788685524e-06, "loss": 2.2465, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.3976705490848585e-06, "loss": 2.5682, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.430948419301165e-06, "loss": 2.503, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.4642262895174711e-06, "loss": 3.0273, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.4975041597337772e-06, "loss": 2.6278, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.5307820299500833e-06, "loss": 2.675, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.5640599001663895e-06, "loss": 2.1949, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.5973377703826956e-06, "loss": 2.6989, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.6306156405990017e-06, "loss": 2.9638, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.6638935108153078e-06, "loss": 2.4935, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.6971713810316139e-06, "loss": 3.1223, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.7304492512479204e-06, "loss": 2.3969, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.7637271214642265e-06, "loss": 2.1049, "step": 53 }, { "epoch": 0.0, "learning_rate": 1.7970049916805326e-06, "loss": 2.1866, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.8302828618968388e-06, "loss": 2.8875, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.8635607321131449e-06, "loss": 2.7563, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.896838602329451e-06, "loss": 2.3931, "step": 57 }, { "epoch": 0.0, "learning_rate": 1.9301164725457573e-06, "loss": 2.2624, "step": 58 }, { "epoch": 0.0, "learning_rate": 1.9633943427620634e-06, "loss": 2.5515, "step": 59 }, { "epoch": 0.0, "learning_rate": 1.9966722129783695e-06, "loss": 2.9608, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.0299500831946756e-06, "loss": 2.8242, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.063227953410982e-06, "loss": 2.9479, "step": 62 }, { "epoch": 0.0, "learning_rate": 2.0965058236272883e-06, "loss": 2.6543, "step": 63 }, { "epoch": 0.0, "learning_rate": 2.1297836938435944e-06, "loss": 1.8691, "step": 64 }, { "epoch": 0.0, "learning_rate": 2.1630615640599005e-06, "loss": 2.1696, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.1963394342762066e-06, "loss": 2.2501, "step": 66 }, { "epoch": 0.0, "learning_rate": 2.2296173044925127e-06, "loss": 2.6542, "step": 67 }, { "epoch": 0.0, "learning_rate": 2.262895174708819e-06, "loss": 2.637, "step": 68 }, { "epoch": 0.0, "learning_rate": 2.296173044925125e-06, "loss": 2.5365, "step": 69 }, { "epoch": 0.0, "learning_rate": 2.329450915141431e-06, "loss": 2.3917, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.3627287853577376e-06, "loss": 2.2686, "step": 71 }, { "epoch": 0.0, "learning_rate": 2.3960066555740437e-06, "loss": 2.5296, "step": 72 }, { "epoch": 0.0, "learning_rate": 2.4292845257903498e-06, "loss": 2.2496, "step": 73 }, { "epoch": 0.0, "learning_rate": 2.462562396006656e-06, "loss": 2.4679, "step": 74 }, { "epoch": 0.0, "learning_rate": 2.495840266222962e-06, "loss": 2.8633, "step": 75 }, { "epoch": 0.0, "learning_rate": 2.529118136439268e-06, "loss": 2.5314, "step": 76 }, { "epoch": 0.0, "learning_rate": 2.5623960066555742e-06, "loss": 2.4655, "step": 77 }, { "epoch": 0.0, "learning_rate": 2.5956738768718803e-06, "loss": 2.3166, "step": 78 }, { "epoch": 0.0, "learning_rate": 2.6289517470881864e-06, "loss": 2.2719, "step": 79 }, { "epoch": 0.0, "learning_rate": 2.6622296173044925e-06, "loss": 2.3365, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.6955074875207986e-06, "loss": 2.3752, "step": 81 }, { "epoch": 0.0, "learning_rate": 2.7287853577371048e-06, "loss": 2.1382, "step": 82 }, { "epoch": 0.0, "learning_rate": 2.762063227953411e-06, "loss": 2.065, "step": 83 }, { "epoch": 0.0, "learning_rate": 2.795341098169717e-06, "loss": 1.9709, "step": 84 }, { "epoch": 0.0, "learning_rate": 2.828618968386024e-06, "loss": 2.1425, "step": 85 }, { "epoch": 0.0, "learning_rate": 2.86189683860233e-06, "loss": 2.1951, "step": 86 }, { "epoch": 0.0, "learning_rate": 2.895174708818636e-06, "loss": 2.4851, "step": 87 }, { "epoch": 0.0, "learning_rate": 2.9284525790349423e-06, "loss": 2.1529, "step": 88 }, { "epoch": 0.0, "learning_rate": 2.9617304492512484e-06, "loss": 1.7897, "step": 89 }, { "epoch": 0.0, "learning_rate": 2.9950083194675545e-06, "loss": 1.9884, "step": 90 }, { "epoch": 0.0, "learning_rate": 3.0282861896838606e-06, "loss": 1.7039, "step": 91 }, { "epoch": 0.0, "learning_rate": 3.0615640599001667e-06, "loss": 2.3615, "step": 92 }, { "epoch": 0.0, "learning_rate": 3.094841930116473e-06, "loss": 1.758, "step": 93 }, { "epoch": 0.0, "learning_rate": 3.128119800332779e-06, "loss": 1.8607, "step": 94 }, { "epoch": 0.0, "learning_rate": 3.161397670549085e-06, "loss": 1.6987, "step": 95 }, { "epoch": 0.0, "learning_rate": 3.194675540765391e-06, "loss": 2.0018, "step": 96 }, { "epoch": 0.0, "learning_rate": 3.2279534109816972e-06, "loss": 2.068, "step": 97 }, { "epoch": 0.0, "learning_rate": 3.2612312811980034e-06, "loss": 1.7838, "step": 98 }, { "epoch": 0.0, "learning_rate": 3.2945091514143095e-06, "loss": 2.3186, "step": 99 }, { "epoch": 0.0, "learning_rate": 3.3277870216306156e-06, "loss": 1.3967, "step": 100 }, { "epoch": 0.01, "learning_rate": 3.3610648918469217e-06, "loss": 1.739, "step": 101 }, { "epoch": 0.01, "learning_rate": 3.3943427620632278e-06, "loss": 1.7019, "step": 102 }, { "epoch": 0.01, "learning_rate": 3.4276206322795343e-06, "loss": 1.771, "step": 103 }, { "epoch": 0.01, "learning_rate": 3.460898502495841e-06, "loss": 1.6296, "step": 104 }, { "epoch": 0.01, "learning_rate": 3.494176372712147e-06, "loss": 1.4767, "step": 105 }, { "epoch": 0.01, "learning_rate": 3.527454242928453e-06, "loss": 1.7105, "step": 106 }, { "epoch": 0.01, "learning_rate": 3.560732113144759e-06, "loss": 1.2488, "step": 107 }, { "epoch": 0.01, "learning_rate": 3.5940099833610653e-06, "loss": 1.3368, "step": 108 }, { "epoch": 0.01, "learning_rate": 3.6272878535773714e-06, "loss": 1.3186, "step": 109 }, { "epoch": 0.01, "learning_rate": 3.6605657237936775e-06, "loss": 1.3643, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.6938435940099836e-06, "loss": 0.8143, "step": 111 }, { "epoch": 0.01, "learning_rate": 3.7271214642262897e-06, "loss": 0.9529, "step": 112 }, { "epoch": 0.01, "learning_rate": 3.760399334442596e-06, "loss": 1.4881, "step": 113 }, { "epoch": 0.01, "learning_rate": 3.793677204658902e-06, "loss": 1.176, "step": 114 }, { "epoch": 0.01, "learning_rate": 3.826955074875208e-06, "loss": 1.1105, "step": 115 }, { "epoch": 0.01, "learning_rate": 3.860232945091515e-06, "loss": 1.3743, "step": 116 }, { "epoch": 0.01, "learning_rate": 3.89351081530782e-06, "loss": 1.0162, "step": 117 }, { "epoch": 0.01, "learning_rate": 3.926788685524127e-06, "loss": 1.1235, "step": 118 }, { "epoch": 0.01, "learning_rate": 3.9600665557404325e-06, "loss": 1.4356, "step": 119 }, { "epoch": 0.01, "learning_rate": 3.993344425956739e-06, "loss": 1.1467, "step": 120 }, { "epoch": 0.01, "learning_rate": 4.026622296173045e-06, "loss": 0.8799, "step": 121 }, { "epoch": 0.01, "learning_rate": 4.059900166389351e-06, "loss": 0.9855, "step": 122 }, { "epoch": 0.01, "learning_rate": 4.093178036605658e-06, "loss": 0.9445, "step": 123 }, { "epoch": 0.01, "learning_rate": 4.126455906821964e-06, "loss": 1.1419, "step": 124 }, { "epoch": 0.01, "learning_rate": 4.15973377703827e-06, "loss": 1.193, "step": 125 }, { "epoch": 0.01, "learning_rate": 4.1930116472545765e-06, "loss": 0.9814, "step": 126 }, { "epoch": 0.01, "learning_rate": 4.226289517470882e-06, "loss": 1.0419, "step": 127 }, { "epoch": 0.01, "learning_rate": 4.259567387687189e-06, "loss": 1.0051, "step": 128 }, { "epoch": 0.01, "learning_rate": 4.2928452579034944e-06, "loss": 0.9476, "step": 129 }, { "epoch": 0.01, "learning_rate": 4.326123128119801e-06, "loss": 0.9929, "step": 130 }, { "epoch": 0.01, "learning_rate": 4.359400998336107e-06, "loss": 1.1777, "step": 131 }, { "epoch": 0.01, "learning_rate": 4.392678868552413e-06, "loss": 0.7263, "step": 132 }, { "epoch": 0.01, "learning_rate": 4.425956738768719e-06, "loss": 1.0746, "step": 133 }, { "epoch": 0.01, "learning_rate": 4.459234608985025e-06, "loss": 0.7447, "step": 134 }, { "epoch": 0.01, "learning_rate": 4.492512479201331e-06, "loss": 0.8961, "step": 135 }, { "epoch": 0.01, "learning_rate": 4.525790349417638e-06, "loss": 0.9752, "step": 136 }, { "epoch": 0.01, "learning_rate": 4.559068219633943e-06, "loss": 0.7344, "step": 137 }, { "epoch": 0.01, "learning_rate": 4.59234608985025e-06, "loss": 0.5635, "step": 138 }, { "epoch": 0.01, "learning_rate": 4.6256239600665555e-06, "loss": 0.7749, "step": 139 }, { "epoch": 0.01, "learning_rate": 4.658901830282862e-06, "loss": 1.0281, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.692179700499169e-06, "loss": 1.1008, "step": 141 }, { "epoch": 0.01, "learning_rate": 4.725457570715475e-06, "loss": 0.7665, "step": 142 }, { "epoch": 0.01, "learning_rate": 4.758735440931781e-06, "loss": 0.9111, "step": 143 }, { "epoch": 0.01, "learning_rate": 4.792013311148087e-06, "loss": 0.4934, "step": 144 }, { "epoch": 0.01, "learning_rate": 4.825291181364393e-06, "loss": 0.7395, "step": 145 }, { "epoch": 0.01, "learning_rate": 4.8585690515806996e-06, "loss": 0.4529, "step": 146 }, { "epoch": 0.01, "learning_rate": 4.891846921797005e-06, "loss": 0.7372, "step": 147 }, { "epoch": 0.01, "learning_rate": 4.925124792013312e-06, "loss": 0.6694, "step": 148 }, { "epoch": 0.01, "learning_rate": 4.9584026622296175e-06, "loss": 0.7453, "step": 149 }, { "epoch": 0.01, "learning_rate": 4.991680532445924e-06, "loss": 0.842, "step": 150 }, { "epoch": 0.01, "learning_rate": 5.02495840266223e-06, "loss": 0.8049, "step": 151 }, { "epoch": 0.01, "learning_rate": 5.058236272878536e-06, "loss": 0.5646, "step": 152 }, { "epoch": 0.01, "learning_rate": 5.091514143094842e-06, "loss": 0.7862, "step": 153 }, { "epoch": 0.01, "learning_rate": 5.1247920133111484e-06, "loss": 0.7846, "step": 154 }, { "epoch": 0.01, "learning_rate": 5.158069883527454e-06, "loss": 0.8565, "step": 155 }, { "epoch": 0.01, "learning_rate": 5.191347753743761e-06, "loss": 0.7834, "step": 156 }, { "epoch": 0.01, "learning_rate": 5.224625623960067e-06, "loss": 0.9293, "step": 157 }, { "epoch": 0.01, "learning_rate": 5.257903494176373e-06, "loss": 0.3893, "step": 158 }, { "epoch": 0.01, "learning_rate": 5.291181364392679e-06, "loss": 0.4728, "step": 159 }, { "epoch": 0.01, "learning_rate": 5.324459234608985e-06, "loss": 0.8096, "step": 160 }, { "epoch": 0.01, "learning_rate": 5.357737104825292e-06, "loss": 0.7969, "step": 161 }, { "epoch": 0.01, "learning_rate": 5.391014975041597e-06, "loss": 0.8886, "step": 162 }, { "epoch": 0.01, "learning_rate": 5.424292845257904e-06, "loss": 1.0002, "step": 163 }, { "epoch": 0.01, "learning_rate": 5.4575707154742095e-06, "loss": 0.7217, "step": 164 }, { "epoch": 0.01, "learning_rate": 5.490848585690516e-06, "loss": 0.6248, "step": 165 }, { "epoch": 0.01, "learning_rate": 5.524126455906822e-06, "loss": 0.8358, "step": 166 }, { "epoch": 0.01, "learning_rate": 5.557404326123128e-06, "loss": 0.684, "step": 167 }, { "epoch": 0.01, "learning_rate": 5.590682196339434e-06, "loss": 0.4001, "step": 168 }, { "epoch": 0.01, "learning_rate": 5.6239600665557405e-06, "loss": 0.6369, "step": 169 }, { "epoch": 0.01, "learning_rate": 5.657237936772048e-06, "loss": 0.7153, "step": 170 }, { "epoch": 0.01, "learning_rate": 5.6905158069883535e-06, "loss": 0.6422, "step": 171 }, { "epoch": 0.01, "learning_rate": 5.72379367720466e-06, "loss": 0.4031, "step": 172 }, { "epoch": 0.01, "learning_rate": 5.757071547420966e-06, "loss": 0.5601, "step": 173 }, { "epoch": 0.01, "learning_rate": 5.790349417637272e-06, "loss": 0.7639, "step": 174 }, { "epoch": 0.01, "learning_rate": 5.823627287853578e-06, "loss": 0.5503, "step": 175 }, { "epoch": 0.01, "learning_rate": 5.8569051580698845e-06, "loss": 0.6289, "step": 176 }, { "epoch": 0.01, "learning_rate": 5.89018302828619e-06, "loss": 0.5362, "step": 177 }, { "epoch": 0.01, "learning_rate": 5.923460898502497e-06, "loss": 0.3322, "step": 178 }, { "epoch": 0.01, "learning_rate": 5.956738768718802e-06, "loss": 0.6233, "step": 179 }, { "epoch": 0.01, "learning_rate": 5.990016638935109e-06, "loss": 0.8015, "step": 180 }, { "epoch": 0.01, "learning_rate": 6.023294509151415e-06, "loss": 0.5389, "step": 181 }, { "epoch": 0.01, "learning_rate": 6.056572379367721e-06, "loss": 0.9165, "step": 182 }, { "epoch": 0.01, "learning_rate": 6.089850249584027e-06, "loss": 0.8566, "step": 183 }, { "epoch": 0.01, "learning_rate": 6.123128119800333e-06, "loss": 0.7445, "step": 184 }, { "epoch": 0.01, "learning_rate": 6.15640599001664e-06, "loss": 0.5873, "step": 185 }, { "epoch": 0.01, "learning_rate": 6.189683860232946e-06, "loss": 0.5844, "step": 186 }, { "epoch": 0.01, "learning_rate": 6.222961730449252e-06, "loss": 0.6451, "step": 187 }, { "epoch": 0.01, "learning_rate": 6.256239600665558e-06, "loss": 0.7752, "step": 188 }, { "epoch": 0.01, "learning_rate": 6.289517470881864e-06, "loss": 1.1211, "step": 189 }, { "epoch": 0.01, "learning_rate": 6.32279534109817e-06, "loss": 0.4488, "step": 190 }, { "epoch": 0.01, "learning_rate": 6.356073211314477e-06, "loss": 0.6048, "step": 191 }, { "epoch": 0.01, "learning_rate": 6.389351081530782e-06, "loss": 0.6592, "step": 192 }, { "epoch": 0.01, "learning_rate": 6.422628951747089e-06, "loss": 0.4964, "step": 193 }, { "epoch": 0.01, "learning_rate": 6.4559068219633945e-06, "loss": 0.6109, "step": 194 }, { "epoch": 0.01, "learning_rate": 6.489184692179701e-06, "loss": 0.9144, "step": 195 }, { "epoch": 0.01, "learning_rate": 6.522462562396007e-06, "loss": 0.426, "step": 196 }, { "epoch": 0.01, "learning_rate": 6.555740432612313e-06, "loss": 0.7795, "step": 197 }, { "epoch": 0.01, "learning_rate": 6.589018302828619e-06, "loss": 0.2639, "step": 198 }, { "epoch": 0.01, "learning_rate": 6.6222961730449255e-06, "loss": 0.4791, "step": 199 }, { "epoch": 0.01, "learning_rate": 6.655574043261231e-06, "loss": 0.5504, "step": 200 }, { "epoch": 0.01, "learning_rate": 6.688851913477538e-06, "loss": 0.7711, "step": 201 }, { "epoch": 0.01, "learning_rate": 6.722129783693843e-06, "loss": 0.6829, "step": 202 }, { "epoch": 0.01, "learning_rate": 6.75540765391015e-06, "loss": 0.8088, "step": 203 }, { "epoch": 0.01, "learning_rate": 6.7886855241264556e-06, "loss": 0.6481, "step": 204 }, { "epoch": 0.01, "learning_rate": 6.821963394342762e-06, "loss": 0.6827, "step": 205 }, { "epoch": 0.01, "learning_rate": 6.855241264559069e-06, "loss": 0.5265, "step": 206 }, { "epoch": 0.01, "learning_rate": 6.888519134775375e-06, "loss": 0.635, "step": 207 }, { "epoch": 0.01, "learning_rate": 6.921797004991682e-06, "loss": 0.4105, "step": 208 }, { "epoch": 0.01, "learning_rate": 6.955074875207987e-06, "loss": 0.5779, "step": 209 }, { "epoch": 0.01, "learning_rate": 6.988352745424294e-06, "loss": 0.7686, "step": 210 }, { "epoch": 0.01, "learning_rate": 7.0216306156406e-06, "loss": 1.0449, "step": 211 }, { "epoch": 0.01, "learning_rate": 7.054908485856906e-06, "loss": 0.4115, "step": 212 }, { "epoch": 0.01, "learning_rate": 7.088186356073213e-06, "loss": 0.5073, "step": 213 }, { "epoch": 0.01, "learning_rate": 7.121464226289518e-06, "loss": 0.4769, "step": 214 }, { "epoch": 0.01, "learning_rate": 7.154742096505825e-06, "loss": 0.7284, "step": 215 }, { "epoch": 0.01, "learning_rate": 7.1880199667221306e-06, "loss": 0.6138, "step": 216 }, { "epoch": 0.01, "learning_rate": 7.221297836938437e-06, "loss": 0.4504, "step": 217 }, { "epoch": 0.01, "learning_rate": 7.254575707154743e-06, "loss": 0.5607, "step": 218 }, { "epoch": 0.01, "learning_rate": 7.287853577371049e-06, "loss": 0.5981, "step": 219 }, { "epoch": 0.01, "learning_rate": 7.321131447587355e-06, "loss": 0.3362, "step": 220 }, { "epoch": 0.01, "learning_rate": 7.3544093178036615e-06, "loss": 0.5726, "step": 221 }, { "epoch": 0.01, "learning_rate": 7.387687188019967e-06, "loss": 0.5076, "step": 222 }, { "epoch": 0.01, "learning_rate": 7.420965058236274e-06, "loss": 0.5097, "step": 223 }, { "epoch": 0.01, "learning_rate": 7.4542429284525794e-06, "loss": 0.7553, "step": 224 }, { "epoch": 0.01, "learning_rate": 7.487520798668886e-06, "loss": 0.5035, "step": 225 }, { "epoch": 0.01, "learning_rate": 7.520798668885192e-06, "loss": 0.8002, "step": 226 }, { "epoch": 0.01, "learning_rate": 7.554076539101498e-06, "loss": 0.7861, "step": 227 }, { "epoch": 0.01, "learning_rate": 7.587354409317804e-06, "loss": 0.5058, "step": 228 }, { "epoch": 0.01, "learning_rate": 7.62063227953411e-06, "loss": 0.626, "step": 229 }, { "epoch": 0.01, "learning_rate": 7.653910149750416e-06, "loss": 0.4824, "step": 230 }, { "epoch": 0.01, "learning_rate": 7.687188019966722e-06, "loss": 0.4299, "step": 231 }, { "epoch": 0.01, "learning_rate": 7.72046589018303e-06, "loss": 0.423, "step": 232 }, { "epoch": 0.01, "learning_rate": 7.753743760399335e-06, "loss": 0.6137, "step": 233 }, { "epoch": 0.01, "learning_rate": 7.78702163061564e-06, "loss": 0.4331, "step": 234 }, { "epoch": 0.01, "learning_rate": 7.820299500831948e-06, "loss": 0.5104, "step": 235 }, { "epoch": 0.01, "learning_rate": 7.853577371048254e-06, "loss": 0.6055, "step": 236 }, { "epoch": 0.01, "learning_rate": 7.88685524126456e-06, "loss": 0.4642, "step": 237 }, { "epoch": 0.01, "learning_rate": 7.920133111480865e-06, "loss": 0.6934, "step": 238 }, { "epoch": 0.01, "learning_rate": 7.953410981697172e-06, "loss": 0.5139, "step": 239 }, { "epoch": 0.01, "learning_rate": 7.986688851913478e-06, "loss": 0.5736, "step": 240 }, { "epoch": 0.01, "learning_rate": 8.019966722129784e-06, "loss": 0.5064, "step": 241 }, { "epoch": 0.01, "learning_rate": 8.05324459234609e-06, "loss": 0.6047, "step": 242 }, { "epoch": 0.01, "learning_rate": 8.086522462562397e-06, "loss": 0.4219, "step": 243 }, { "epoch": 0.01, "learning_rate": 8.119800332778702e-06, "loss": 0.3339, "step": 244 }, { "epoch": 0.01, "learning_rate": 8.15307820299501e-06, "loss": 0.595, "step": 245 }, { "epoch": 0.01, "learning_rate": 8.186356073211316e-06, "loss": 0.6483, "step": 246 }, { "epoch": 0.01, "learning_rate": 8.219633943427621e-06, "loss": 0.6809, "step": 247 }, { "epoch": 0.01, "learning_rate": 8.252911813643929e-06, "loss": 0.5092, "step": 248 }, { "epoch": 0.01, "learning_rate": 8.286189683860234e-06, "loss": 0.4476, "step": 249 }, { "epoch": 0.01, "learning_rate": 8.31946755407654e-06, "loss": 0.3591, "step": 250 }, { "epoch": 0.01, "learning_rate": 8.352745424292846e-06, "loss": 0.4339, "step": 251 }, { "epoch": 0.01, "learning_rate": 8.386023294509153e-06, "loss": 0.6802, "step": 252 }, { "epoch": 0.01, "learning_rate": 8.419301164725459e-06, "loss": 0.4698, "step": 253 }, { "epoch": 0.01, "learning_rate": 8.452579034941764e-06, "loss": 0.6469, "step": 254 }, { "epoch": 0.01, "learning_rate": 8.48585690515807e-06, "loss": 0.7443, "step": 255 }, { "epoch": 0.01, "learning_rate": 8.519134775374377e-06, "loss": 0.6655, "step": 256 }, { "epoch": 0.01, "learning_rate": 8.552412645590683e-06, "loss": 0.2662, "step": 257 }, { "epoch": 0.01, "learning_rate": 8.585690515806989e-06, "loss": 0.4917, "step": 258 }, { "epoch": 0.01, "learning_rate": 8.618968386023295e-06, "loss": 0.7027, "step": 259 }, { "epoch": 0.01, "learning_rate": 8.652246256239602e-06, "loss": 0.3742, "step": 260 }, { "epoch": 0.01, "learning_rate": 8.685524126455908e-06, "loss": 0.6099, "step": 261 }, { "epoch": 0.01, "learning_rate": 8.718801996672213e-06, "loss": 0.5076, "step": 262 }, { "epoch": 0.01, "learning_rate": 8.75207986688852e-06, "loss": 0.4859, "step": 263 }, { "epoch": 0.01, "learning_rate": 8.785357737104826e-06, "loss": 0.6672, "step": 264 }, { "epoch": 0.01, "learning_rate": 8.818635607321132e-06, "loss": 0.4585, "step": 265 }, { "epoch": 0.01, "learning_rate": 8.851913477537438e-06, "loss": 0.8009, "step": 266 }, { "epoch": 0.01, "learning_rate": 8.885191347753745e-06, "loss": 0.3948, "step": 267 }, { "epoch": 0.01, "learning_rate": 8.91846921797005e-06, "loss": 0.4391, "step": 268 }, { "epoch": 0.01, "learning_rate": 8.951747088186356e-06, "loss": 0.2514, "step": 269 }, { "epoch": 0.01, "learning_rate": 8.985024958402662e-06, "loss": 0.5028, "step": 270 }, { "epoch": 0.01, "learning_rate": 9.01830282861897e-06, "loss": 0.8387, "step": 271 }, { "epoch": 0.01, "learning_rate": 9.051580698835275e-06, "loss": 0.2806, "step": 272 }, { "epoch": 0.01, "learning_rate": 9.084858569051581e-06, "loss": 0.5524, "step": 273 }, { "epoch": 0.01, "learning_rate": 9.118136439267887e-06, "loss": 0.5017, "step": 274 }, { "epoch": 0.01, "learning_rate": 9.151414309484194e-06, "loss": 0.6453, "step": 275 }, { "epoch": 0.01, "learning_rate": 9.1846921797005e-06, "loss": 0.7362, "step": 276 }, { "epoch": 0.01, "learning_rate": 9.217970049916805e-06, "loss": 0.7521, "step": 277 }, { "epoch": 0.01, "learning_rate": 9.251247920133111e-06, "loss": 0.7763, "step": 278 }, { "epoch": 0.01, "learning_rate": 9.284525790349418e-06, "loss": 0.4485, "step": 279 }, { "epoch": 0.01, "learning_rate": 9.317803660565724e-06, "loss": 0.5487, "step": 280 }, { "epoch": 0.01, "learning_rate": 9.35108153078203e-06, "loss": 0.6473, "step": 281 }, { "epoch": 0.01, "learning_rate": 9.384359400998337e-06, "loss": 0.8381, "step": 282 }, { "epoch": 0.01, "learning_rate": 9.417637271214643e-06, "loss": 0.5193, "step": 283 }, { "epoch": 0.01, "learning_rate": 9.45091514143095e-06, "loss": 0.4581, "step": 284 }, { "epoch": 0.01, "learning_rate": 9.484193011647256e-06, "loss": 0.6971, "step": 285 }, { "epoch": 0.01, "learning_rate": 9.517470881863562e-06, "loss": 0.4512, "step": 286 }, { "epoch": 0.01, "learning_rate": 9.550748752079867e-06, "loss": 0.5953, "step": 287 }, { "epoch": 0.01, "learning_rate": 9.584026622296175e-06, "loss": 0.5508, "step": 288 }, { "epoch": 0.01, "learning_rate": 9.61730449251248e-06, "loss": 0.5791, "step": 289 }, { "epoch": 0.01, "learning_rate": 9.650582362728786e-06, "loss": 0.595, "step": 290 }, { "epoch": 0.01, "learning_rate": 9.683860232945093e-06, "loss": 0.544, "step": 291 }, { "epoch": 0.01, "learning_rate": 9.717138103161399e-06, "loss": 0.6784, "step": 292 }, { "epoch": 0.01, "learning_rate": 9.750415973377705e-06, "loss": 0.316, "step": 293 }, { "epoch": 0.01, "learning_rate": 9.78369384359401e-06, "loss": 0.3656, "step": 294 }, { "epoch": 0.01, "learning_rate": 9.816971713810318e-06, "loss": 0.6058, "step": 295 }, { "epoch": 0.01, "learning_rate": 9.850249584026624e-06, "loss": 0.3951, "step": 296 }, { "epoch": 0.01, "learning_rate": 9.88352745424293e-06, "loss": 0.3001, "step": 297 }, { "epoch": 0.01, "learning_rate": 9.916805324459235e-06, "loss": 0.6839, "step": 298 }, { "epoch": 0.01, "learning_rate": 9.950083194675542e-06, "loss": 0.7014, "step": 299 }, { "epoch": 0.01, "learning_rate": 9.983361064891848e-06, "loss": 0.4253, "step": 300 }, { "epoch": 0.02, "learning_rate": 1.0016638935108154e-05, "loss": 0.6126, "step": 301 }, { "epoch": 0.02, "learning_rate": 1.004991680532446e-05, "loss": 0.5726, "step": 302 }, { "epoch": 0.02, "learning_rate": 1.0083194675540767e-05, "loss": 0.3111, "step": 303 }, { "epoch": 0.02, "learning_rate": 1.0116472545757072e-05, "loss": 0.4644, "step": 304 }, { "epoch": 0.02, "learning_rate": 1.0149750415973378e-05, "loss": 0.42, "step": 305 }, { "epoch": 0.02, "learning_rate": 1.0183028286189684e-05, "loss": 0.4199, "step": 306 }, { "epoch": 0.02, "learning_rate": 1.0216306156405991e-05, "loss": 0.7004, "step": 307 }, { "epoch": 0.02, "learning_rate": 1.0249584026622297e-05, "loss": 0.6928, "step": 308 }, { "epoch": 0.02, "learning_rate": 1.0282861896838603e-05, "loss": 0.5475, "step": 309 }, { "epoch": 0.02, "learning_rate": 1.0316139767054908e-05, "loss": 0.7392, "step": 310 }, { "epoch": 0.02, "learning_rate": 1.0349417637271216e-05, "loss": 0.5701, "step": 311 }, { "epoch": 0.02, "learning_rate": 1.0382695507487521e-05, "loss": 0.4041, "step": 312 }, { "epoch": 0.02, "learning_rate": 1.0415973377703827e-05, "loss": 0.3404, "step": 313 }, { "epoch": 0.02, "learning_rate": 1.0449251247920134e-05, "loss": 0.4034, "step": 314 }, { "epoch": 0.02, "learning_rate": 1.048252911813644e-05, "loss": 0.2916, "step": 315 }, { "epoch": 0.02, "learning_rate": 1.0515806988352746e-05, "loss": 0.4689, "step": 316 }, { "epoch": 0.02, "learning_rate": 1.0549084858569051e-05, "loss": 0.506, "step": 317 }, { "epoch": 0.02, "learning_rate": 1.0582362728785359e-05, "loss": 0.4419, "step": 318 }, { "epoch": 0.02, "learning_rate": 1.0615640599001664e-05, "loss": 0.2989, "step": 319 }, { "epoch": 0.02, "learning_rate": 1.064891846921797e-05, "loss": 0.5396, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.0682196339434276e-05, "loss": 0.6095, "step": 321 }, { "epoch": 0.02, "learning_rate": 1.0715474209650583e-05, "loss": 0.5873, "step": 322 }, { "epoch": 0.02, "learning_rate": 1.0748752079866889e-05, "loss": 0.7436, "step": 323 }, { "epoch": 0.02, "learning_rate": 1.0782029950083195e-05, "loss": 0.5464, "step": 324 }, { "epoch": 0.02, "learning_rate": 1.08153078202995e-05, "loss": 0.4178, "step": 325 }, { "epoch": 0.02, "learning_rate": 1.0848585690515808e-05, "loss": 0.5511, "step": 326 }, { "epoch": 0.02, "learning_rate": 1.0881863560732113e-05, "loss": 0.4452, "step": 327 }, { "epoch": 0.02, "learning_rate": 1.0915141430948419e-05, "loss": 0.6789, "step": 328 }, { "epoch": 0.02, "learning_rate": 1.0948419301164725e-05, "loss": 0.5712, "step": 329 }, { "epoch": 0.02, "learning_rate": 1.0981697171381032e-05, "loss": 0.2847, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.1014975041597338e-05, "loss": 0.6939, "step": 331 }, { "epoch": 0.02, "learning_rate": 1.1048252911813643e-05, "loss": 0.2657, "step": 332 }, { "epoch": 0.02, "learning_rate": 1.108153078202995e-05, "loss": 0.5174, "step": 333 }, { "epoch": 0.02, "learning_rate": 1.1114808652246257e-05, "loss": 0.5081, "step": 334 }, { "epoch": 0.02, "learning_rate": 1.1148086522462562e-05, "loss": 0.7196, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.1181364392678868e-05, "loss": 0.3718, "step": 336 }, { "epoch": 0.02, "learning_rate": 1.1214642262895175e-05, "loss": 0.3079, "step": 337 }, { "epoch": 0.02, "learning_rate": 1.1247920133111481e-05, "loss": 0.482, "step": 338 }, { "epoch": 0.02, "learning_rate": 1.1281198003327788e-05, "loss": 0.6049, "step": 339 }, { "epoch": 0.02, "learning_rate": 1.1314475873544096e-05, "loss": 0.5172, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.1347753743760401e-05, "loss": 0.3901, "step": 341 }, { "epoch": 0.02, "learning_rate": 1.1381031613976707e-05, "loss": 0.4141, "step": 342 }, { "epoch": 0.02, "learning_rate": 1.1414309484193013e-05, "loss": 0.6081, "step": 343 }, { "epoch": 0.02, "learning_rate": 1.144758735440932e-05, "loss": 1.4999, "step": 344 }, { "epoch": 0.02, "learning_rate": 1.1480865224625626e-05, "loss": 0.4024, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.1514143094841932e-05, "loss": 0.1598, "step": 346 }, { "epoch": 0.02, "learning_rate": 1.1547420965058237e-05, "loss": 0.7386, "step": 347 }, { "epoch": 0.02, "learning_rate": 1.1580698835274545e-05, "loss": 0.3675, "step": 348 }, { "epoch": 0.02, "learning_rate": 1.161397670549085e-05, "loss": 0.5723, "step": 349 }, { "epoch": 0.02, "learning_rate": 1.1647254575707156e-05, "loss": 0.6227, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.1680532445923463e-05, "loss": 0.4523, "step": 351 }, { "epoch": 0.02, "learning_rate": 1.1713810316139769e-05, "loss": 0.6071, "step": 352 }, { "epoch": 0.02, "learning_rate": 1.1747088186356075e-05, "loss": 0.614, "step": 353 }, { "epoch": 0.02, "learning_rate": 1.178036605657238e-05, "loss": 0.4308, "step": 354 }, { "epoch": 0.02, "learning_rate": 1.1813643926788688e-05, "loss": 0.3101, "step": 355 }, { "epoch": 0.02, "learning_rate": 1.1846921797004993e-05, "loss": 0.336, "step": 356 }, { "epoch": 0.02, "learning_rate": 1.18801996672213e-05, "loss": 0.5609, "step": 357 }, { "epoch": 0.02, "learning_rate": 1.1913477537437605e-05, "loss": 0.4305, "step": 358 }, { "epoch": 0.02, "learning_rate": 1.1946755407653912e-05, "loss": 0.4518, "step": 359 }, { "epoch": 0.02, "learning_rate": 1.1980033277870218e-05, "loss": 0.5254, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.2013311148086524e-05, "loss": 0.6674, "step": 361 }, { "epoch": 0.02, "learning_rate": 1.204658901830283e-05, "loss": 0.3459, "step": 362 }, { "epoch": 0.02, "learning_rate": 1.2079866888519137e-05, "loss": 0.6832, "step": 363 }, { "epoch": 0.02, "learning_rate": 1.2113144758735442e-05, "loss": 0.4125, "step": 364 }, { "epoch": 0.02, "learning_rate": 1.2146422628951748e-05, "loss": 0.4495, "step": 365 }, { "epoch": 0.02, "learning_rate": 1.2179700499168054e-05, "loss": 0.6825, "step": 366 }, { "epoch": 0.02, "learning_rate": 1.2212978369384361e-05, "loss": 0.5877, "step": 367 }, { "epoch": 0.02, "learning_rate": 1.2246256239600667e-05, "loss": 0.7622, "step": 368 }, { "epoch": 0.02, "learning_rate": 1.2279534109816972e-05, "loss": 0.6971, "step": 369 }, { "epoch": 0.02, "learning_rate": 1.231281198003328e-05, "loss": 0.5152, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.2346089850249586e-05, "loss": 0.5374, "step": 371 }, { "epoch": 0.02, "learning_rate": 1.2379367720465891e-05, "loss": 0.7265, "step": 372 }, { "epoch": 0.02, "learning_rate": 1.2412645590682197e-05, "loss": 0.5669, "step": 373 }, { "epoch": 0.02, "learning_rate": 1.2445923460898504e-05, "loss": 0.5306, "step": 374 }, { "epoch": 0.02, "learning_rate": 1.247920133111481e-05, "loss": 0.4673, "step": 375 }, { "epoch": 0.02, "learning_rate": 1.2512479201331116e-05, "loss": 0.4394, "step": 376 }, { "epoch": 0.02, "learning_rate": 1.2545757071547421e-05, "loss": 0.6228, "step": 377 }, { "epoch": 0.02, "learning_rate": 1.2579034941763729e-05, "loss": 0.6992, "step": 378 }, { "epoch": 0.02, "learning_rate": 1.2612312811980034e-05, "loss": 0.4818, "step": 379 }, { "epoch": 0.02, "learning_rate": 1.264559068219634e-05, "loss": 0.364, "step": 380 }, { "epoch": 0.02, "learning_rate": 1.2678868552412646e-05, "loss": 0.6264, "step": 381 }, { "epoch": 0.02, "learning_rate": 1.2712146422628953e-05, "loss": 0.3691, "step": 382 }, { "epoch": 0.02, "learning_rate": 1.2745424292845259e-05, "loss": 0.3196, "step": 383 }, { "epoch": 0.02, "learning_rate": 1.2778702163061565e-05, "loss": 0.6807, "step": 384 }, { "epoch": 0.02, "learning_rate": 1.281198003327787e-05, "loss": 0.5799, "step": 385 }, { "epoch": 0.02, "learning_rate": 1.2845257903494178e-05, "loss": 0.3613, "step": 386 }, { "epoch": 0.02, "learning_rate": 1.2878535773710483e-05, "loss": 0.4877, "step": 387 }, { "epoch": 0.02, "learning_rate": 1.2911813643926789e-05, "loss": 0.3663, "step": 388 }, { "epoch": 0.02, "learning_rate": 1.2945091514143096e-05, "loss": 0.5379, "step": 389 }, { "epoch": 0.02, "learning_rate": 1.2978369384359402e-05, "loss": 0.3925, "step": 390 }, { "epoch": 0.02, "learning_rate": 1.3011647254575708e-05, "loss": 0.4593, "step": 391 }, { "epoch": 0.02, "learning_rate": 1.3044925124792013e-05, "loss": 0.5322, "step": 392 }, { "epoch": 0.02, "learning_rate": 1.307820299500832e-05, "loss": 0.5117, "step": 393 }, { "epoch": 0.02, "learning_rate": 1.3111480865224626e-05, "loss": 0.4225, "step": 394 }, { "epoch": 0.02, "learning_rate": 1.3144758735440932e-05, "loss": 0.3842, "step": 395 }, { "epoch": 0.02, "learning_rate": 1.3178036605657238e-05, "loss": 0.5255, "step": 396 }, { "epoch": 0.02, "learning_rate": 1.3211314475873545e-05, "loss": 0.2338, "step": 397 }, { "epoch": 0.02, "learning_rate": 1.3244592346089851e-05, "loss": 0.5525, "step": 398 }, { "epoch": 0.02, "learning_rate": 1.3277870216306157e-05, "loss": 0.2334, "step": 399 }, { "epoch": 0.02, "learning_rate": 1.3311148086522462e-05, "loss": 0.5666, "step": 400 }, { "epoch": 0.02, "learning_rate": 1.334442595673877e-05, "loss": 0.5399, "step": 401 }, { "epoch": 0.02, "learning_rate": 1.3377703826955075e-05, "loss": 0.4409, "step": 402 }, { "epoch": 0.02, "learning_rate": 1.3410981697171381e-05, "loss": 0.4151, "step": 403 }, { "epoch": 0.02, "learning_rate": 1.3444259567387687e-05, "loss": 0.2385, "step": 404 }, { "epoch": 0.02, "learning_rate": 1.3477537437603994e-05, "loss": 0.6024, "step": 405 }, { "epoch": 0.02, "learning_rate": 1.35108153078203e-05, "loss": 0.4974, "step": 406 }, { "epoch": 0.02, "learning_rate": 1.3544093178036605e-05, "loss": 0.3184, "step": 407 }, { "epoch": 0.02, "learning_rate": 1.3577371048252911e-05, "loss": 0.5087, "step": 408 }, { "epoch": 0.02, "learning_rate": 1.3610648918469219e-05, "loss": 0.4655, "step": 409 }, { "epoch": 0.02, "learning_rate": 1.3643926788685524e-05, "loss": 0.3615, "step": 410 }, { "epoch": 0.02, "learning_rate": 1.367720465890183e-05, "loss": 0.3989, "step": 411 }, { "epoch": 0.02, "learning_rate": 1.3710482529118137e-05, "loss": 0.3649, "step": 412 }, { "epoch": 0.02, "learning_rate": 1.3743760399334443e-05, "loss": 0.9153, "step": 413 }, { "epoch": 0.02, "learning_rate": 1.377703826955075e-05, "loss": 0.2516, "step": 414 }, { "epoch": 0.02, "learning_rate": 1.3810316139767058e-05, "loss": 0.2178, "step": 415 }, { "epoch": 0.02, "learning_rate": 1.3843594009983363e-05, "loss": 0.4271, "step": 416 }, { "epoch": 0.02, "learning_rate": 1.3876871880199669e-05, "loss": 0.4907, "step": 417 }, { "epoch": 0.02, "learning_rate": 1.3910149750415975e-05, "loss": 0.7105, "step": 418 }, { "epoch": 0.02, "learning_rate": 1.3943427620632282e-05, "loss": 0.4662, "step": 419 }, { "epoch": 0.02, "learning_rate": 1.3976705490848588e-05, "loss": 0.4296, "step": 420 }, { "epoch": 0.02, "learning_rate": 1.4009983361064894e-05, "loss": 0.6427, "step": 421 }, { "epoch": 0.02, "learning_rate": 1.40432612312812e-05, "loss": 0.4919, "step": 422 }, { "epoch": 0.02, "learning_rate": 1.4076539101497507e-05, "loss": 0.4349, "step": 423 }, { "epoch": 0.02, "learning_rate": 1.4109816971713812e-05, "loss": 0.4805, "step": 424 }, { "epoch": 0.02, "learning_rate": 1.4143094841930118e-05, "loss": 0.502, "step": 425 }, { "epoch": 0.02, "learning_rate": 1.4176372712146425e-05, "loss": 0.495, "step": 426 }, { "epoch": 0.02, "learning_rate": 1.4209650582362731e-05, "loss": 0.4689, "step": 427 }, { "epoch": 0.02, "learning_rate": 1.4242928452579037e-05, "loss": 0.3954, "step": 428 }, { "epoch": 0.02, "learning_rate": 1.4276206322795342e-05, "loss": 0.7996, "step": 429 }, { "epoch": 0.02, "learning_rate": 1.430948419301165e-05, "loss": 0.6805, "step": 430 }, { "epoch": 0.02, "learning_rate": 1.4342762063227955e-05, "loss": 0.726, "step": 431 }, { "epoch": 0.02, "learning_rate": 1.4376039933444261e-05, "loss": 0.375, "step": 432 }, { "epoch": 0.02, "learning_rate": 1.4409317803660567e-05, "loss": 0.4049, "step": 433 }, { "epoch": 0.02, "learning_rate": 1.4442595673876874e-05, "loss": 0.4526, "step": 434 }, { "epoch": 0.02, "learning_rate": 1.447587354409318e-05, "loss": 0.3692, "step": 435 }, { "epoch": 0.02, "learning_rate": 1.4509151414309486e-05, "loss": 0.5856, "step": 436 }, { "epoch": 0.02, "learning_rate": 1.4542429284525791e-05, "loss": 0.3092, "step": 437 }, { "epoch": 0.02, "learning_rate": 1.4575707154742099e-05, "loss": 1.0773, "step": 438 }, { "epoch": 0.02, "learning_rate": 1.4608985024958404e-05, "loss": 0.3672, "step": 439 }, { "epoch": 0.02, "learning_rate": 1.464226289517471e-05, "loss": 0.589, "step": 440 }, { "epoch": 0.02, "learning_rate": 1.4675540765391016e-05, "loss": 0.396, "step": 441 }, { "epoch": 0.02, "learning_rate": 1.4708818635607323e-05, "loss": 0.6347, "step": 442 }, { "epoch": 0.02, "learning_rate": 1.4742096505823629e-05, "loss": 0.2514, "step": 443 }, { "epoch": 0.02, "learning_rate": 1.4775374376039934e-05, "loss": 0.1864, "step": 444 }, { "epoch": 0.02, "learning_rate": 1.480865224625624e-05, "loss": 0.4898, "step": 445 }, { "epoch": 0.02, "learning_rate": 1.4841930116472548e-05, "loss": 0.5407, "step": 446 }, { "epoch": 0.02, "learning_rate": 1.4875207986688853e-05, "loss": 0.4383, "step": 447 }, { "epoch": 0.02, "learning_rate": 1.4908485856905159e-05, "loss": 0.6979, "step": 448 }, { "epoch": 0.02, "learning_rate": 1.4941763727121466e-05, "loss": 0.4308, "step": 449 }, { "epoch": 0.02, "learning_rate": 1.4975041597337772e-05, "loss": 0.4351, "step": 450 }, { "epoch": 0.02, "learning_rate": 1.5008319467554078e-05, "loss": 0.5039, "step": 451 }, { "epoch": 0.02, "learning_rate": 1.5041597337770383e-05, "loss": 0.3532, "step": 452 }, { "epoch": 0.02, "learning_rate": 1.507487520798669e-05, "loss": 0.4209, "step": 453 }, { "epoch": 0.02, "learning_rate": 1.5108153078202996e-05, "loss": 0.2709, "step": 454 }, { "epoch": 0.02, "learning_rate": 1.5141430948419302e-05, "loss": 0.5428, "step": 455 }, { "epoch": 0.02, "learning_rate": 1.5174708818635608e-05, "loss": 0.4242, "step": 456 }, { "epoch": 0.02, "learning_rate": 1.5207986688851915e-05, "loss": 0.5926, "step": 457 }, { "epoch": 0.02, "learning_rate": 1.524126455906822e-05, "loss": 0.498, "step": 458 }, { "epoch": 0.02, "learning_rate": 1.5274542429284527e-05, "loss": 0.4586, "step": 459 }, { "epoch": 0.02, "learning_rate": 1.5307820299500832e-05, "loss": 0.1813, "step": 460 }, { "epoch": 0.02, "learning_rate": 1.5341098169717138e-05, "loss": 0.5455, "step": 461 }, { "epoch": 0.02, "learning_rate": 1.5374376039933444e-05, "loss": 0.4782, "step": 462 }, { "epoch": 0.02, "learning_rate": 1.5407653910149753e-05, "loss": 0.8203, "step": 463 }, { "epoch": 0.02, "learning_rate": 1.544093178036606e-05, "loss": 0.3781, "step": 464 }, { "epoch": 0.02, "learning_rate": 1.5474209650582364e-05, "loss": 0.4537, "step": 465 }, { "epoch": 0.02, "learning_rate": 1.550748752079867e-05, "loss": 0.9838, "step": 466 }, { "epoch": 0.02, "learning_rate": 1.5540765391014975e-05, "loss": 0.6676, "step": 467 }, { "epoch": 0.02, "learning_rate": 1.557404326123128e-05, "loss": 0.3435, "step": 468 }, { "epoch": 0.02, "learning_rate": 1.5607321131447587e-05, "loss": 0.3598, "step": 469 }, { "epoch": 0.02, "learning_rate": 1.5640599001663896e-05, "loss": 0.6014, "step": 470 }, { "epoch": 0.02, "learning_rate": 1.56738768718802e-05, "loss": 0.3356, "step": 471 }, { "epoch": 0.02, "learning_rate": 1.5707154742096507e-05, "loss": 0.2961, "step": 472 }, { "epoch": 0.02, "learning_rate": 1.5740432612312813e-05, "loss": 0.2841, "step": 473 }, { "epoch": 0.02, "learning_rate": 1.577371048252912e-05, "loss": 0.3726, "step": 474 }, { "epoch": 0.02, "learning_rate": 1.5806988352745424e-05, "loss": 0.5777, "step": 475 }, { "epoch": 0.02, "learning_rate": 1.584026622296173e-05, "loss": 0.5114, "step": 476 }, { "epoch": 0.02, "learning_rate": 1.5873544093178036e-05, "loss": 0.76, "step": 477 }, { "epoch": 0.02, "learning_rate": 1.5906821963394345e-05, "loss": 0.4386, "step": 478 }, { "epoch": 0.02, "learning_rate": 1.594009983361065e-05, "loss": 0.3875, "step": 479 }, { "epoch": 0.02, "learning_rate": 1.5973377703826956e-05, "loss": 0.3648, "step": 480 }, { "epoch": 0.02, "learning_rate": 1.6006655574043262e-05, "loss": 0.5767, "step": 481 }, { "epoch": 0.02, "learning_rate": 1.6039933444259567e-05, "loss": 0.5889, "step": 482 }, { "epoch": 0.02, "learning_rate": 1.6073211314475873e-05, "loss": 0.3693, "step": 483 }, { "epoch": 0.02, "learning_rate": 1.610648918469218e-05, "loss": 0.5895, "step": 484 }, { "epoch": 0.02, "learning_rate": 1.6139767054908485e-05, "loss": 0.5236, "step": 485 }, { "epoch": 0.02, "learning_rate": 1.6173044925124794e-05, "loss": 0.3914, "step": 486 }, { "epoch": 0.02, "learning_rate": 1.62063227953411e-05, "loss": 0.7155, "step": 487 }, { "epoch": 0.02, "learning_rate": 1.6239600665557405e-05, "loss": 0.228, "step": 488 }, { "epoch": 0.02, "learning_rate": 1.6272878535773714e-05, "loss": 0.3211, "step": 489 }, { "epoch": 0.02, "learning_rate": 1.630615640599002e-05, "loss": 0.2929, "step": 490 }, { "epoch": 0.02, "learning_rate": 1.6339434276206325e-05, "loss": 0.6178, "step": 491 }, { "epoch": 0.02, "learning_rate": 1.637271214642263e-05, "loss": 0.4411, "step": 492 }, { "epoch": 0.02, "learning_rate": 1.6405990016638937e-05, "loss": 0.656, "step": 493 }, { "epoch": 0.02, "learning_rate": 1.6439267886855242e-05, "loss": 0.3893, "step": 494 }, { "epoch": 0.02, "learning_rate": 1.6472545757071548e-05, "loss": 0.4343, "step": 495 }, { "epoch": 0.02, "learning_rate": 1.6505823627287857e-05, "loss": 0.3972, "step": 496 }, { "epoch": 0.02, "learning_rate": 1.6539101497504163e-05, "loss": 0.4227, "step": 497 }, { "epoch": 0.02, "learning_rate": 1.657237936772047e-05, "loss": 0.6724, "step": 498 }, { "epoch": 0.02, "learning_rate": 1.6605657237936774e-05, "loss": 0.4703, "step": 499 }, { "epoch": 0.02, "learning_rate": 1.663893510815308e-05, "loss": 0.3585, "step": 500 }, { "epoch": 0.03, "learning_rate": 1.6672212978369386e-05, "loss": 0.7284, "step": 501 }, { "epoch": 0.03, "learning_rate": 1.670549084858569e-05, "loss": 0.6862, "step": 502 }, { "epoch": 0.03, "learning_rate": 1.6738768718802e-05, "loss": 0.6056, "step": 503 }, { "epoch": 0.03, "learning_rate": 1.6772046589018306e-05, "loss": 0.5791, "step": 504 }, { "epoch": 0.03, "learning_rate": 1.6805324459234612e-05, "loss": 0.4728, "step": 505 }, { "epoch": 0.03, "learning_rate": 1.6838602329450917e-05, "loss": 0.8628, "step": 506 }, { "epoch": 0.03, "learning_rate": 1.6871880199667223e-05, "loss": 0.4695, "step": 507 }, { "epoch": 0.03, "learning_rate": 1.690515806988353e-05, "loss": 0.433, "step": 508 }, { "epoch": 0.03, "learning_rate": 1.6938435940099835e-05, "loss": 0.4315, "step": 509 }, { "epoch": 0.03, "learning_rate": 1.697171381031614e-05, "loss": 0.5845, "step": 510 }, { "epoch": 0.03, "learning_rate": 1.700499168053245e-05, "loss": 0.5543, "step": 511 }, { "epoch": 0.03, "learning_rate": 1.7038269550748755e-05, "loss": 0.2966, "step": 512 }, { "epoch": 0.03, "learning_rate": 1.707154742096506e-05, "loss": 0.5259, "step": 513 }, { "epoch": 0.03, "learning_rate": 1.7104825291181366e-05, "loss": 0.4496, "step": 514 }, { "epoch": 0.03, "learning_rate": 1.7138103161397672e-05, "loss": 0.2697, "step": 515 }, { "epoch": 0.03, "learning_rate": 1.7171381031613978e-05, "loss": 0.5608, "step": 516 }, { "epoch": 0.03, "learning_rate": 1.7204658901830283e-05, "loss": 0.291, "step": 517 }, { "epoch": 0.03, "learning_rate": 1.723793677204659e-05, "loss": 0.4995, "step": 518 }, { "epoch": 0.03, "learning_rate": 1.7271214642262898e-05, "loss": 0.4328, "step": 519 }, { "epoch": 0.03, "learning_rate": 1.7304492512479204e-05, "loss": 0.7457, "step": 520 }, { "epoch": 0.03, "learning_rate": 1.733777038269551e-05, "loss": 0.6318, "step": 521 }, { "epoch": 0.03, "learning_rate": 1.7371048252911815e-05, "loss": 0.4947, "step": 522 }, { "epoch": 0.03, "learning_rate": 1.740432612312812e-05, "loss": 0.2575, "step": 523 }, { "epoch": 0.03, "learning_rate": 1.7437603993344427e-05, "loss": 0.7949, "step": 524 }, { "epoch": 0.03, "learning_rate": 1.7470881863560732e-05, "loss": 0.5901, "step": 525 }, { "epoch": 0.03, "learning_rate": 1.750415973377704e-05, "loss": 0.5153, "step": 526 }, { "epoch": 0.03, "learning_rate": 1.7537437603993347e-05, "loss": 0.4242, "step": 527 }, { "epoch": 0.03, "learning_rate": 1.7570715474209653e-05, "loss": 0.4874, "step": 528 }, { "epoch": 0.03, "learning_rate": 1.760399334442596e-05, "loss": 0.4573, "step": 529 }, { "epoch": 0.03, "learning_rate": 1.7637271214642264e-05, "loss": 0.4113, "step": 530 }, { "epoch": 0.03, "learning_rate": 1.767054908485857e-05, "loss": 0.5502, "step": 531 }, { "epoch": 0.03, "learning_rate": 1.7703826955074875e-05, "loss": 1.3908, "step": 532 }, { "epoch": 0.03, "learning_rate": 1.773710482529118e-05, "loss": 0.3048, "step": 533 }, { "epoch": 0.03, "learning_rate": 1.777038269550749e-05, "loss": 0.4858, "step": 534 }, { "epoch": 0.03, "learning_rate": 1.7803660565723796e-05, "loss": 0.3168, "step": 535 }, { "epoch": 0.03, "learning_rate": 1.78369384359401e-05, "loss": 0.4321, "step": 536 }, { "epoch": 0.03, "learning_rate": 1.7870216306156407e-05, "loss": 0.7047, "step": 537 }, { "epoch": 0.03, "learning_rate": 1.7903494176372713e-05, "loss": 0.2759, "step": 538 }, { "epoch": 0.03, "learning_rate": 1.793677204658902e-05, "loss": 0.6094, "step": 539 }, { "epoch": 0.03, "learning_rate": 1.7970049916805324e-05, "loss": 0.5547, "step": 540 }, { "epoch": 0.03, "learning_rate": 1.800332778702163e-05, "loss": 0.2178, "step": 541 }, { "epoch": 0.03, "learning_rate": 1.803660565723794e-05, "loss": 0.1301, "step": 542 }, { "epoch": 0.03, "learning_rate": 1.8069883527454245e-05, "loss": 0.4968, "step": 543 }, { "epoch": 0.03, "learning_rate": 1.810316139767055e-05, "loss": 0.4965, "step": 544 }, { "epoch": 0.03, "learning_rate": 1.8136439267886856e-05, "loss": 0.552, "step": 545 }, { "epoch": 0.03, "learning_rate": 1.8169717138103162e-05, "loss": 0.4825, "step": 546 }, { "epoch": 0.03, "learning_rate": 1.8202995008319468e-05, "loss": 0.2926, "step": 547 }, { "epoch": 0.03, "learning_rate": 1.8236272878535773e-05, "loss": 0.2452, "step": 548 }, { "epoch": 0.03, "learning_rate": 1.8269550748752082e-05, "loss": 0.5817, "step": 549 }, { "epoch": 0.03, "learning_rate": 1.8302828618968388e-05, "loss": 0.4212, "step": 550 }, { "epoch": 0.03, "learning_rate": 1.8336106489184694e-05, "loss": 0.684, "step": 551 }, { "epoch": 0.03, "learning_rate": 1.8369384359401e-05, "loss": 0.5815, "step": 552 }, { "epoch": 0.03, "learning_rate": 1.8402662229617305e-05, "loss": 0.7249, "step": 553 }, { "epoch": 0.03, "learning_rate": 1.843594009983361e-05, "loss": 0.6557, "step": 554 }, { "epoch": 0.03, "learning_rate": 1.8469217970049916e-05, "loss": 0.5859, "step": 555 }, { "epoch": 0.03, "learning_rate": 1.8502495840266222e-05, "loss": 0.3983, "step": 556 }, { "epoch": 0.03, "learning_rate": 1.853577371048253e-05, "loss": 0.3739, "step": 557 }, { "epoch": 0.03, "learning_rate": 1.8569051580698837e-05, "loss": 0.8405, "step": 558 }, { "epoch": 0.03, "learning_rate": 1.8602329450915143e-05, "loss": 0.4886, "step": 559 }, { "epoch": 0.03, "learning_rate": 1.8635607321131448e-05, "loss": 0.464, "step": 560 }, { "epoch": 0.03, "learning_rate": 1.8668885191347754e-05, "loss": 0.2907, "step": 561 }, { "epoch": 0.03, "learning_rate": 1.870216306156406e-05, "loss": 0.3252, "step": 562 }, { "epoch": 0.03, "learning_rate": 1.8735440931780365e-05, "loss": 0.5496, "step": 563 }, { "epoch": 0.03, "learning_rate": 1.8768718801996674e-05, "loss": 0.4042, "step": 564 }, { "epoch": 0.03, "learning_rate": 1.880199667221298e-05, "loss": 0.2494, "step": 565 }, { "epoch": 0.03, "learning_rate": 1.8835274542429286e-05, "loss": 0.4083, "step": 566 }, { "epoch": 0.03, "learning_rate": 1.8868552412645595e-05, "loss": 0.6349, "step": 567 }, { "epoch": 0.03, "learning_rate": 1.89018302828619e-05, "loss": 0.4271, "step": 568 }, { "epoch": 0.03, "learning_rate": 1.8935108153078206e-05, "loss": 0.395, "step": 569 }, { "epoch": 0.03, "learning_rate": 1.8968386023294512e-05, "loss": 0.4191, "step": 570 }, { "epoch": 0.03, "learning_rate": 1.9001663893510818e-05, "loss": 0.3878, "step": 571 }, { "epoch": 0.03, "learning_rate": 1.9034941763727123e-05, "loss": 0.271, "step": 572 }, { "epoch": 0.03, "learning_rate": 1.906821963394343e-05, "loss": 0.4081, "step": 573 }, { "epoch": 0.03, "learning_rate": 1.9101497504159735e-05, "loss": 0.4884, "step": 574 }, { "epoch": 0.03, "learning_rate": 1.9134775374376044e-05, "loss": 0.3698, "step": 575 }, { "epoch": 0.03, "learning_rate": 1.916805324459235e-05, "loss": 0.3393, "step": 576 }, { "epoch": 0.03, "learning_rate": 1.9201331114808655e-05, "loss": 0.455, "step": 577 }, { "epoch": 0.03, "learning_rate": 1.923460898502496e-05, "loss": 0.612, "step": 578 }, { "epoch": 0.03, "learning_rate": 1.9267886855241266e-05, "loss": 0.2365, "step": 579 }, { "epoch": 0.03, "learning_rate": 1.9301164725457572e-05, "loss": 0.6065, "step": 580 }, { "epoch": 0.03, "learning_rate": 1.9334442595673878e-05, "loss": 0.4121, "step": 581 }, { "epoch": 0.03, "learning_rate": 1.9367720465890187e-05, "loss": 0.4967, "step": 582 }, { "epoch": 0.03, "learning_rate": 1.9400998336106493e-05, "loss": 0.4365, "step": 583 }, { "epoch": 0.03, "learning_rate": 1.9434276206322798e-05, "loss": 0.5468, "step": 584 }, { "epoch": 0.03, "learning_rate": 1.9467554076539104e-05, "loss": 0.5475, "step": 585 }, { "epoch": 0.03, "learning_rate": 1.950083194675541e-05, "loss": 0.4077, "step": 586 }, { "epoch": 0.03, "learning_rate": 1.9534109816971715e-05, "loss": 0.4882, "step": 587 }, { "epoch": 0.03, "learning_rate": 1.956738768718802e-05, "loss": 0.3287, "step": 588 }, { "epoch": 0.03, "learning_rate": 1.9600665557404327e-05, "loss": 0.3624, "step": 589 }, { "epoch": 0.03, "learning_rate": 1.9633943427620636e-05, "loss": 0.3075, "step": 590 }, { "epoch": 0.03, "learning_rate": 1.966722129783694e-05, "loss": 0.6285, "step": 591 }, { "epoch": 0.03, "learning_rate": 1.9700499168053247e-05, "loss": 0.3344, "step": 592 }, { "epoch": 0.03, "learning_rate": 1.9733777038269553e-05, "loss": 0.6168, "step": 593 }, { "epoch": 0.03, "learning_rate": 1.976705490848586e-05, "loss": 0.2461, "step": 594 }, { "epoch": 0.03, "learning_rate": 1.9800332778702164e-05, "loss": 0.6746, "step": 595 }, { "epoch": 0.03, "learning_rate": 1.983361064891847e-05, "loss": 0.584, "step": 596 }, { "epoch": 0.03, "learning_rate": 1.9866888519134776e-05, "loss": 0.5342, "step": 597 }, { "epoch": 0.03, "learning_rate": 1.9900166389351085e-05, "loss": 0.5683, "step": 598 }, { "epoch": 0.03, "learning_rate": 1.993344425956739e-05, "loss": 0.4695, "step": 599 }, { "epoch": 0.03, "learning_rate": 1.9966722129783696e-05, "loss": 0.7433, "step": 600 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 0.4853, "step": 601 }, { "epoch": 0.03, "learning_rate": 1.999999986923155e-05, "loss": 0.4207, "step": 602 }, { "epoch": 0.03, "learning_rate": 1.99999994769262e-05, "loss": 0.304, "step": 603 }, { "epoch": 0.03, "learning_rate": 1.999999882308396e-05, "loss": 0.3466, "step": 604 }, { "epoch": 0.03, "learning_rate": 1.999999790770485e-05, "loss": 0.5195, "step": 605 }, { "epoch": 0.03, "learning_rate": 1.9999996730788894e-05, "loss": 0.38, "step": 606 }, { "epoch": 0.03, "learning_rate": 1.999999529233612e-05, "loss": 0.2201, "step": 607 }, { "epoch": 0.03, "learning_rate": 1.999999359234656e-05, "loss": 0.184, "step": 608 }, { "epoch": 0.03, "learning_rate": 1.9999991630820276e-05, "loss": 0.395, "step": 609 }, { "epoch": 0.03, "learning_rate": 1.9999989407757302e-05, "loss": 0.3715, "step": 610 }, { "epoch": 0.03, "learning_rate": 1.9999986923157706e-05, "loss": 0.4645, "step": 611 }, { "epoch": 0.03, "learning_rate": 1.999998417702155e-05, "loss": 0.5164, "step": 612 }, { "epoch": 0.03, "learning_rate": 1.99999811693489e-05, "loss": 0.4697, "step": 613 }, { "epoch": 0.03, "learning_rate": 1.9999977900139845e-05, "loss": 0.5351, "step": 614 }, { "epoch": 0.03, "learning_rate": 1.9999974369394464e-05, "loss": 0.4327, "step": 615 }, { "epoch": 0.03, "learning_rate": 1.9999970577112854e-05, "loss": 0.5851, "step": 616 }, { "epoch": 0.03, "learning_rate": 1.9999966523295103e-05, "loss": 0.4645, "step": 617 }, { "epoch": 0.03, "learning_rate": 1.999996220794133e-05, "loss": 0.4397, "step": 618 }, { "epoch": 0.03, "learning_rate": 1.999995763105165e-05, "loss": 0.3326, "step": 619 }, { "epoch": 0.03, "learning_rate": 1.9999952792626165e-05, "loss": 0.5795, "step": 620 }, { "epoch": 0.03, "learning_rate": 1.9999947692665018e-05, "loss": 0.6034, "step": 621 }, { "epoch": 0.03, "learning_rate": 1.9999942331168337e-05, "loss": 0.4503, "step": 622 }, { "epoch": 0.03, "learning_rate": 1.9999936708136257e-05, "loss": 0.2032, "step": 623 }, { "epoch": 0.03, "learning_rate": 1.9999930823568935e-05, "loss": 0.5236, "step": 624 }, { "epoch": 0.03, "learning_rate": 1.9999924677466516e-05, "loss": 0.5612, "step": 625 }, { "epoch": 0.03, "learning_rate": 1.999991826982917e-05, "loss": 0.2306, "step": 626 }, { "epoch": 0.03, "learning_rate": 1.9999911600657055e-05, "loss": 0.593, "step": 627 }, { "epoch": 0.03, "learning_rate": 1.999990466995035e-05, "loss": 0.3276, "step": 628 }, { "epoch": 0.03, "learning_rate": 1.9999897477709238e-05, "loss": 0.5337, "step": 629 }, { "epoch": 0.03, "learning_rate": 1.9999890023933903e-05, "loss": 0.6621, "step": 630 }, { "epoch": 0.03, "learning_rate": 1.999988230862454e-05, "loss": 0.1651, "step": 631 }, { "epoch": 0.03, "learning_rate": 1.9999874331781356e-05, "loss": 0.6996, "step": 632 }, { "epoch": 0.03, "learning_rate": 1.9999866093404554e-05, "loss": 0.4213, "step": 633 }, { "epoch": 0.03, "learning_rate": 1.9999857593494357e-05, "loss": 0.4602, "step": 634 }, { "epoch": 0.03, "learning_rate": 1.9999848832050977e-05, "loss": 0.2467, "step": 635 }, { "epoch": 0.03, "learning_rate": 1.9999839809074648e-05, "loss": 0.5303, "step": 636 }, { "epoch": 0.03, "learning_rate": 1.999983052456561e-05, "loss": 0.5374, "step": 637 }, { "epoch": 0.03, "learning_rate": 1.9999820978524098e-05, "loss": 0.5008, "step": 638 }, { "epoch": 0.03, "learning_rate": 1.9999811170950366e-05, "loss": 0.6643, "step": 639 }, { "epoch": 0.03, "learning_rate": 1.9999801101844674e-05, "loss": 0.3886, "step": 640 }, { "epoch": 0.03, "learning_rate": 1.999979077120728e-05, "loss": 0.6693, "step": 641 }, { "epoch": 0.03, "learning_rate": 1.999978017903845e-05, "loss": 0.365, "step": 642 }, { "epoch": 0.03, "learning_rate": 1.9999769325338473e-05, "loss": 0.3881, "step": 643 }, { "epoch": 0.03, "learning_rate": 1.9999758210107627e-05, "loss": 0.469, "step": 644 }, { "epoch": 0.03, "learning_rate": 1.99997468333462e-05, "loss": 0.3509, "step": 645 }, { "epoch": 0.03, "learning_rate": 1.9999735195054493e-05, "loss": 0.5963, "step": 646 }, { "epoch": 0.03, "learning_rate": 1.999972329523281e-05, "loss": 0.4112, "step": 647 }, { "epoch": 0.03, "learning_rate": 1.999971113388146e-05, "loss": 0.5341, "step": 648 }, { "epoch": 0.03, "learning_rate": 1.9999698711000763e-05, "loss": 0.6565, "step": 649 }, { "epoch": 0.03, "learning_rate": 1.9999686026591047e-05, "loss": 0.3617, "step": 650 }, { "epoch": 0.03, "learning_rate": 1.9999673080652635e-05, "loss": 0.2726, "step": 651 }, { "epoch": 0.03, "learning_rate": 1.999965987318587e-05, "loss": 0.2264, "step": 652 }, { "epoch": 0.03, "learning_rate": 1.9999646404191102e-05, "loss": 0.4182, "step": 653 }, { "epoch": 0.03, "learning_rate": 1.999963267366868e-05, "loss": 0.5811, "step": 654 }, { "epoch": 0.03, "learning_rate": 1.9999618681618956e-05, "loss": 0.6982, "step": 655 }, { "epoch": 0.03, "learning_rate": 1.999960442804231e-05, "loss": 0.2564, "step": 656 }, { "epoch": 0.03, "learning_rate": 1.9999589912939104e-05, "loss": 0.4679, "step": 657 }, { "epoch": 0.03, "learning_rate": 1.9999575136309725e-05, "loss": 0.625, "step": 658 }, { "epoch": 0.03, "learning_rate": 1.999956009815455e-05, "loss": 0.465, "step": 659 }, { "epoch": 0.03, "learning_rate": 1.999954479847398e-05, "loss": 0.2883, "step": 660 }, { "epoch": 0.03, "learning_rate": 1.9999529237268413e-05, "loss": 0.3981, "step": 661 }, { "epoch": 0.03, "learning_rate": 1.9999513414538252e-05, "loss": 0.3805, "step": 662 }, { "epoch": 0.03, "learning_rate": 1.999949733028392e-05, "loss": 0.3562, "step": 663 }, { "epoch": 0.03, "learning_rate": 1.999948098450583e-05, "loss": 0.3654, "step": 664 }, { "epoch": 0.03, "learning_rate": 1.9999464377204414e-05, "loss": 0.4827, "step": 665 }, { "epoch": 0.03, "learning_rate": 1.99994475083801e-05, "loss": 0.7515, "step": 666 }, { "epoch": 0.03, "learning_rate": 1.9999430378033334e-05, "loss": 0.2807, "step": 667 }, { "epoch": 0.03, "learning_rate": 1.9999412986164566e-05, "loss": 0.251, "step": 668 }, { "epoch": 0.03, "learning_rate": 1.999939533277425e-05, "loss": 0.7349, "step": 669 }, { "epoch": 0.03, "learning_rate": 1.999937741786284e-05, "loss": 0.5422, "step": 670 }, { "epoch": 0.03, "learning_rate": 1.9999359241430816e-05, "loss": 0.2296, "step": 671 }, { "epoch": 0.03, "learning_rate": 1.9999340803478645e-05, "loss": 0.3482, "step": 672 }, { "epoch": 0.03, "learning_rate": 1.9999322104006813e-05, "loss": 0.496, "step": 673 }, { "epoch": 0.03, "learning_rate": 1.9999303143015806e-05, "loss": 0.3238, "step": 674 }, { "epoch": 0.03, "learning_rate": 1.9999283920506124e-05, "loss": 0.5343, "step": 675 }, { "epoch": 0.03, "learning_rate": 1.9999264436478268e-05, "loss": 0.4335, "step": 676 }, { "epoch": 0.03, "learning_rate": 1.9999244690932748e-05, "loss": 0.4228, "step": 677 }, { "epoch": 0.03, "learning_rate": 1.9999224683870077e-05, "loss": 0.6312, "step": 678 }, { "epoch": 0.03, "learning_rate": 1.999920441529078e-05, "loss": 0.6583, "step": 679 }, { "epoch": 0.03, "learning_rate": 1.9999183885195393e-05, "loss": 0.5711, "step": 680 }, { "epoch": 0.03, "learning_rate": 1.9999163093584447e-05, "loss": 0.4794, "step": 681 }, { "epoch": 0.03, "learning_rate": 1.9999142040458483e-05, "loss": 0.504, "step": 682 }, { "epoch": 0.03, "learning_rate": 1.999912072581806e-05, "loss": 0.2419, "step": 683 }, { "epoch": 0.03, "learning_rate": 1.999909914966373e-05, "loss": 0.6658, "step": 684 }, { "epoch": 0.03, "learning_rate": 1.9999077311996055e-05, "loss": 0.479, "step": 685 }, { "epoch": 0.03, "learning_rate": 1.999905521281561e-05, "loss": 0.519, "step": 686 }, { "epoch": 0.03, "learning_rate": 1.999903285212298e-05, "loss": 0.8845, "step": 687 }, { "epoch": 0.03, "learning_rate": 1.9999010229918735e-05, "loss": 0.5279, "step": 688 }, { "epoch": 0.03, "learning_rate": 1.9998987346203474e-05, "loss": 0.2405, "step": 689 }, { "epoch": 0.03, "learning_rate": 1.9998964200977795e-05, "loss": 0.4193, "step": 690 }, { "epoch": 0.03, "learning_rate": 1.9998940794242302e-05, "loss": 0.292, "step": 691 }, { "epoch": 0.03, "learning_rate": 1.9998917125997614e-05, "loss": 0.3406, "step": 692 }, { "epoch": 0.03, "learning_rate": 1.999889319624434e-05, "loss": 0.448, "step": 693 }, { "epoch": 0.03, "learning_rate": 1.9998869004983116e-05, "loss": 0.4801, "step": 694 }, { "epoch": 0.03, "learning_rate": 1.9998844552214564e-05, "loss": 0.4852, "step": 695 }, { "epoch": 0.03, "learning_rate": 1.999881983793933e-05, "loss": 0.2727, "step": 696 }, { "epoch": 0.03, "learning_rate": 1.999879486215806e-05, "loss": 0.3512, "step": 697 }, { "epoch": 0.03, "learning_rate": 1.999876962487141e-05, "loss": 0.5248, "step": 698 }, { "epoch": 0.03, "learning_rate": 1.9998744126080033e-05, "loss": 0.3523, "step": 699 }, { "epoch": 0.03, "learning_rate": 1.9998718365784597e-05, "loss": 0.3981, "step": 700 }, { "epoch": 0.04, "learning_rate": 1.9998692343985782e-05, "loss": 0.4049, "step": 701 }, { "epoch": 0.04, "learning_rate": 1.9998666060684265e-05, "loss": 0.4329, "step": 702 }, { "epoch": 0.04, "learning_rate": 1.999863951588073e-05, "loss": 0.6958, "step": 703 }, { "epoch": 0.04, "learning_rate": 1.9998612709575876e-05, "loss": 0.6394, "step": 704 }, { "epoch": 0.04, "learning_rate": 1.9998585641770404e-05, "loss": 0.5289, "step": 705 }, { "epoch": 0.04, "learning_rate": 1.9998558312465017e-05, "loss": 0.5215, "step": 706 }, { "epoch": 0.04, "learning_rate": 1.9998530721660433e-05, "loss": 0.3894, "step": 707 }, { "epoch": 0.04, "learning_rate": 1.9998502869357375e-05, "loss": 0.3503, "step": 708 }, { "epoch": 0.04, "learning_rate": 1.999847475555657e-05, "loss": 0.5273, "step": 709 }, { "epoch": 0.04, "learning_rate": 1.999844638025876e-05, "loss": 0.4308, "step": 710 }, { "epoch": 0.04, "learning_rate": 1.999841774346467e-05, "loss": 0.3864, "step": 711 }, { "epoch": 0.04, "learning_rate": 1.9998388845175064e-05, "loss": 0.5435, "step": 712 }, { "epoch": 0.04, "learning_rate": 1.9998359685390695e-05, "loss": 0.7037, "step": 713 }, { "epoch": 0.04, "learning_rate": 1.9998330264112323e-05, "loss": 0.506, "step": 714 }, { "epoch": 0.04, "learning_rate": 1.9998300581340716e-05, "loss": 0.6919, "step": 715 }, { "epoch": 0.04, "learning_rate": 1.9998270637076654e-05, "loss": 0.4974, "step": 716 }, { "epoch": 0.04, "learning_rate": 1.999824043132092e-05, "loss": 0.4094, "step": 717 }, { "epoch": 0.04, "learning_rate": 1.9998209964074302e-05, "loss": 0.4857, "step": 718 }, { "epoch": 0.04, "learning_rate": 1.9998179235337595e-05, "loss": 0.3314, "step": 719 }, { "epoch": 0.04, "learning_rate": 1.999814824511161e-05, "loss": 0.5334, "step": 720 }, { "epoch": 0.04, "learning_rate": 1.999811699339715e-05, "loss": 0.2732, "step": 721 }, { "epoch": 0.04, "learning_rate": 1.9998085480195036e-05, "loss": 0.5558, "step": 722 }, { "epoch": 0.04, "learning_rate": 1.9998053705506094e-05, "loss": 0.4238, "step": 723 }, { "epoch": 0.04, "learning_rate": 1.999802166933115e-05, "loss": 0.161, "step": 724 }, { "epoch": 0.04, "learning_rate": 1.9997989371671044e-05, "loss": 0.4685, "step": 725 }, { "epoch": 0.04, "learning_rate": 1.999795681252662e-05, "loss": 0.6182, "step": 726 }, { "epoch": 0.04, "learning_rate": 1.9997923991898732e-05, "loss": 0.4313, "step": 727 }, { "epoch": 0.04, "learning_rate": 1.9997890909788236e-05, "loss": 0.5525, "step": 728 }, { "epoch": 0.04, "learning_rate": 1.9997857566196002e-05, "loss": 0.5106, "step": 729 }, { "epoch": 0.04, "learning_rate": 1.9997823961122895e-05, "loss": 0.6248, "step": 730 }, { "epoch": 0.04, "learning_rate": 1.9997790094569795e-05, "loss": 0.5445, "step": 731 }, { "epoch": 0.04, "learning_rate": 1.999775596653759e-05, "loss": 0.5022, "step": 732 }, { "epoch": 0.04, "learning_rate": 1.999772157702718e-05, "loss": 0.3424, "step": 733 }, { "epoch": 0.04, "learning_rate": 1.999768692603945e-05, "loss": 0.2517, "step": 734 }, { "epoch": 0.04, "learning_rate": 1.9997652013575314e-05, "loss": 0.2943, "step": 735 }, { "epoch": 0.04, "learning_rate": 1.9997616839635683e-05, "loss": 0.4267, "step": 736 }, { "epoch": 0.04, "learning_rate": 1.999758140422148e-05, "loss": 0.557, "step": 737 }, { "epoch": 0.04, "learning_rate": 1.999754570733363e-05, "loss": 0.5623, "step": 738 }, { "epoch": 0.04, "learning_rate": 1.9997509748973065e-05, "loss": 0.6667, "step": 739 }, { "epoch": 0.04, "learning_rate": 1.999747352914073e-05, "loss": 0.3244, "step": 740 }, { "epoch": 0.04, "learning_rate": 1.9997437047837565e-05, "loss": 0.5595, "step": 741 }, { "epoch": 0.04, "learning_rate": 1.999740030506453e-05, "loss": 0.4064, "step": 742 }, { "epoch": 0.04, "learning_rate": 1.9997363300822584e-05, "loss": 0.4652, "step": 743 }, { "epoch": 0.04, "learning_rate": 1.9997326035112694e-05, "loss": 0.3453, "step": 744 }, { "epoch": 0.04, "learning_rate": 1.999728850793584e-05, "loss": 0.7016, "step": 745 }, { "epoch": 0.04, "learning_rate": 1.9997250719292995e-05, "loss": 0.6548, "step": 746 }, { "epoch": 0.04, "learning_rate": 1.9997212669185154e-05, "loss": 0.7489, "step": 747 }, { "epoch": 0.04, "learning_rate": 1.9997174357613307e-05, "loss": 0.3452, "step": 748 }, { "epoch": 0.04, "learning_rate": 1.999713578457846e-05, "loss": 0.4551, "step": 749 }, { "epoch": 0.04, "learning_rate": 1.999709695008162e-05, "loss": 0.6133, "step": 750 }, { "epoch": 0.04, "learning_rate": 1.9997057854123807e-05, "loss": 0.7607, "step": 751 }, { "epoch": 0.04, "learning_rate": 1.9997018496706034e-05, "loss": 0.4536, "step": 752 }, { "epoch": 0.04, "learning_rate": 1.999697887782934e-05, "loss": 0.4176, "step": 753 }, { "epoch": 0.04, "learning_rate": 1.9996938997494756e-05, "loss": 0.4983, "step": 754 }, { "epoch": 0.04, "learning_rate": 1.9996898855703325e-05, "loss": 0.3832, "step": 755 }, { "epoch": 0.04, "learning_rate": 1.9996858452456096e-05, "loss": 0.508, "step": 756 }, { "epoch": 0.04, "learning_rate": 1.9996817787754134e-05, "loss": 0.3138, "step": 757 }, { "epoch": 0.04, "learning_rate": 1.999677686159849e-05, "loss": 0.194, "step": 758 }, { "epoch": 0.04, "learning_rate": 1.9996735673990245e-05, "loss": 0.4317, "step": 759 }, { "epoch": 0.04, "learning_rate": 1.9996694224930468e-05, "loss": 1.1459, "step": 760 }, { "epoch": 0.04, "learning_rate": 1.999665251442025e-05, "loss": 0.3243, "step": 761 }, { "epoch": 0.04, "learning_rate": 1.9996610542460677e-05, "loss": 0.6271, "step": 762 }, { "epoch": 0.04, "learning_rate": 1.999656830905285e-05, "loss": 0.4297, "step": 763 }, { "epoch": 0.04, "learning_rate": 1.999652581419787e-05, "loss": 0.5777, "step": 764 }, { "epoch": 0.04, "learning_rate": 1.999648305789685e-05, "loss": 0.462, "step": 765 }, { "epoch": 0.04, "learning_rate": 1.999644004015091e-05, "loss": 0.3409, "step": 766 }, { "epoch": 0.04, "learning_rate": 1.9996396760961175e-05, "loss": 0.5232, "step": 767 }, { "epoch": 0.04, "learning_rate": 1.9996353220328776e-05, "loss": 0.5946, "step": 768 }, { "epoch": 0.04, "learning_rate": 1.9996309418254852e-05, "loss": 0.5832, "step": 769 }, { "epoch": 0.04, "learning_rate": 1.9996265354740544e-05, "loss": 0.682, "step": 770 }, { "epoch": 0.04, "learning_rate": 1.999622102978701e-05, "loss": 0.2134, "step": 771 }, { "epoch": 0.04, "learning_rate": 1.9996176443395406e-05, "loss": 0.3737, "step": 772 }, { "epoch": 0.04, "learning_rate": 1.9996131595566905e-05, "loss": 0.7111, "step": 773 }, { "epoch": 0.04, "learning_rate": 1.999608648630267e-05, "loss": 0.5174, "step": 774 }, { "epoch": 0.04, "learning_rate": 1.999604111560389e-05, "loss": 0.4216, "step": 775 }, { "epoch": 0.04, "learning_rate": 1.999599548347174e-05, "loss": 0.5903, "step": 776 }, { "epoch": 0.04, "learning_rate": 1.9995949589907426e-05, "loss": 0.498, "step": 777 }, { "epoch": 0.04, "learning_rate": 1.999590343491214e-05, "loss": 0.65, "step": 778 }, { "epoch": 0.04, "learning_rate": 1.9995857018487092e-05, "loss": 0.3977, "step": 779 }, { "epoch": 0.04, "learning_rate": 1.99958103406335e-05, "loss": 0.4104, "step": 780 }, { "epoch": 0.04, "learning_rate": 1.9995763401352577e-05, "loss": 0.334, "step": 781 }, { "epoch": 0.04, "learning_rate": 1.9995716200645554e-05, "loss": 0.72, "step": 782 }, { "epoch": 0.04, "learning_rate": 1.9995668738513666e-05, "loss": 0.3853, "step": 783 }, { "epoch": 0.04, "learning_rate": 1.9995621014958158e-05, "loss": 0.6285, "step": 784 }, { "epoch": 0.04, "learning_rate": 1.9995573029980272e-05, "loss": 0.3558, "step": 785 }, { "epoch": 0.04, "learning_rate": 1.9995524783581263e-05, "loss": 0.6046, "step": 786 }, { "epoch": 0.04, "learning_rate": 1.9995476275762397e-05, "loss": 0.4308, "step": 787 }, { "epoch": 0.04, "learning_rate": 1.999542750652494e-05, "loss": 0.3247, "step": 788 }, { "epoch": 0.04, "learning_rate": 1.999537847587017e-05, "loss": 0.3782, "step": 789 }, { "epoch": 0.04, "learning_rate": 1.9995329183799367e-05, "loss": 0.3215, "step": 790 }, { "epoch": 0.04, "learning_rate": 1.999527963031382e-05, "loss": 0.6596, "step": 791 }, { "epoch": 0.04, "learning_rate": 1.9995229815414824e-05, "loss": 0.3756, "step": 792 }, { "epoch": 0.04, "learning_rate": 1.9995179739103684e-05, "loss": 0.393, "step": 793 }, { "epoch": 0.04, "learning_rate": 1.999512940138171e-05, "loss": 0.5536, "step": 794 }, { "epoch": 0.04, "learning_rate": 1.999507880225022e-05, "loss": 0.5251, "step": 795 }, { "epoch": 0.04, "learning_rate": 1.9995027941710534e-05, "loss": 0.6082, "step": 796 }, { "epoch": 0.04, "learning_rate": 1.9994976819763984e-05, "loss": 0.6703, "step": 797 }, { "epoch": 0.04, "learning_rate": 1.9994925436411905e-05, "loss": 0.4758, "step": 798 }, { "epoch": 0.04, "learning_rate": 1.999487379165564e-05, "loss": 0.4343, "step": 799 }, { "epoch": 0.04, "learning_rate": 1.999482188549655e-05, "loss": 0.5488, "step": 800 }, { "epoch": 0.04, "learning_rate": 1.9994769717935974e-05, "loss": 0.3557, "step": 801 }, { "epoch": 0.04, "learning_rate": 1.9994717288975292e-05, "loss": 0.2977, "step": 802 }, { "epoch": 0.04, "learning_rate": 1.9994664598615868e-05, "loss": 0.3942, "step": 803 }, { "epoch": 0.04, "learning_rate": 1.9994611646859085e-05, "loss": 0.6165, "step": 804 }, { "epoch": 0.04, "learning_rate": 1.9994558433706322e-05, "loss": 0.353, "step": 805 }, { "epoch": 0.04, "learning_rate": 1.999450495915898e-05, "loss": 0.6796, "step": 806 }, { "epoch": 0.04, "learning_rate": 1.9994451223218446e-05, "loss": 0.2839, "step": 807 }, { "epoch": 0.04, "learning_rate": 1.9994397225886128e-05, "loss": 0.3805, "step": 808 }, { "epoch": 0.04, "learning_rate": 1.9994342967163443e-05, "loss": 0.2475, "step": 809 }, { "epoch": 0.04, "learning_rate": 1.999428844705181e-05, "loss": 0.5086, "step": 810 }, { "epoch": 0.04, "learning_rate": 1.999423366555265e-05, "loss": 0.7062, "step": 811 }, { "epoch": 0.04, "learning_rate": 1.9994178622667403e-05, "loss": 0.3228, "step": 812 }, { "epoch": 0.04, "learning_rate": 1.99941233183975e-05, "loss": 0.3185, "step": 813 }, { "epoch": 0.04, "learning_rate": 1.9994067752744396e-05, "loss": 0.5152, "step": 814 }, { "epoch": 0.04, "learning_rate": 1.9994011925709537e-05, "loss": 0.4825, "step": 815 }, { "epoch": 0.04, "learning_rate": 1.9993955837294385e-05, "loss": 0.4946, "step": 816 }, { "epoch": 0.04, "learning_rate": 1.9993899487500406e-05, "loss": 0.6165, "step": 817 }, { "epoch": 0.04, "learning_rate": 1.999384287632908e-05, "loss": 0.7218, "step": 818 }, { "epoch": 0.04, "learning_rate": 1.999378600378188e-05, "loss": 0.3726, "step": 819 }, { "epoch": 0.04, "learning_rate": 1.9993728869860297e-05, "loss": 0.666, "step": 820 }, { "epoch": 0.04, "learning_rate": 1.9993671474565827e-05, "loss": 0.4997, "step": 821 }, { "epoch": 0.04, "learning_rate": 1.9993613817899967e-05, "loss": 0.4606, "step": 822 }, { "epoch": 0.04, "learning_rate": 1.999355589986423e-05, "loss": 0.223, "step": 823 }, { "epoch": 0.04, "learning_rate": 1.9993497720460126e-05, "loss": 0.3202, "step": 824 }, { "epoch": 0.04, "learning_rate": 1.9993439279689175e-05, "loss": 0.2693, "step": 825 }, { "epoch": 0.04, "learning_rate": 1.999338057755291e-05, "loss": 0.2377, "step": 826 }, { "epoch": 0.04, "learning_rate": 1.9993321614052867e-05, "loss": 0.4756, "step": 827 }, { "epoch": 0.04, "learning_rate": 1.9993262389190585e-05, "loss": 0.3969, "step": 828 }, { "epoch": 0.04, "learning_rate": 1.9993202902967614e-05, "loss": 0.4466, "step": 829 }, { "epoch": 0.04, "learning_rate": 1.9993143155385512e-05, "loss": 0.5998, "step": 830 }, { "epoch": 0.04, "learning_rate": 1.9993083146445834e-05, "loss": 0.2936, "step": 831 }, { "epoch": 0.04, "learning_rate": 1.9993022876150162e-05, "loss": 0.3087, "step": 832 }, { "epoch": 0.04, "learning_rate": 1.999296234450006e-05, "loss": 0.6827, "step": 833 }, { "epoch": 0.04, "learning_rate": 1.9992901551497117e-05, "loss": 0.4767, "step": 834 }, { "epoch": 0.04, "learning_rate": 1.9992840497142924e-05, "loss": 0.309, "step": 835 }, { "epoch": 0.04, "learning_rate": 1.9992779181439074e-05, "loss": 0.3712, "step": 836 }, { "epoch": 0.04, "learning_rate": 1.9992717604387175e-05, "loss": 0.5246, "step": 837 }, { "epoch": 0.04, "learning_rate": 1.9992655765988834e-05, "loss": 0.123, "step": 838 }, { "epoch": 0.04, "learning_rate": 1.999259366624567e-05, "loss": 0.1321, "step": 839 }, { "epoch": 0.04, "learning_rate": 1.9992531305159307e-05, "loss": 0.7116, "step": 840 }, { "epoch": 0.04, "learning_rate": 1.9992468682731372e-05, "loss": 0.365, "step": 841 }, { "epoch": 0.04, "learning_rate": 1.999240579896351e-05, "loss": 0.4311, "step": 842 }, { "epoch": 0.04, "learning_rate": 1.9992342653857363e-05, "loss": 0.3585, "step": 843 }, { "epoch": 0.04, "learning_rate": 1.9992279247414578e-05, "loss": 0.2787, "step": 844 }, { "epoch": 0.04, "learning_rate": 1.999221557963682e-05, "loss": 0.3588, "step": 845 }, { "epoch": 0.04, "learning_rate": 1.999215165052575e-05, "loss": 0.4743, "step": 846 }, { "epoch": 0.04, "learning_rate": 1.9992087460083037e-05, "loss": 0.8619, "step": 847 }, { "epoch": 0.04, "learning_rate": 1.999202300831037e-05, "loss": 0.5117, "step": 848 }, { "epoch": 0.04, "learning_rate": 1.9991958295209423e-05, "loss": 0.4601, "step": 849 }, { "epoch": 0.04, "learning_rate": 1.99918933207819e-05, "loss": 0.3668, "step": 850 }, { "epoch": 0.04, "learning_rate": 1.9991828085029488e-05, "loss": 0.4172, "step": 851 }, { "epoch": 0.04, "learning_rate": 1.9991762587953904e-05, "loss": 0.42, "step": 852 }, { "epoch": 0.04, "learning_rate": 1.999169682955685e-05, "loss": 0.545, "step": 853 }, { "epoch": 0.04, "learning_rate": 1.9991630809840058e-05, "loss": 0.5224, "step": 854 }, { "epoch": 0.04, "learning_rate": 1.999156452880525e-05, "loss": 0.3712, "step": 855 }, { "epoch": 0.04, "learning_rate": 1.9991497986454152e-05, "loss": 0.4913, "step": 856 }, { "epoch": 0.04, "learning_rate": 1.9991431182788515e-05, "loss": 0.3428, "step": 857 }, { "epoch": 0.04, "learning_rate": 1.999136411781008e-05, "loss": 0.4191, "step": 858 }, { "epoch": 0.04, "learning_rate": 1.9991296791520605e-05, "loss": 0.4028, "step": 859 }, { "epoch": 0.04, "learning_rate": 1.9991229203921853e-05, "loss": 0.2626, "step": 860 }, { "epoch": 0.04, "learning_rate": 1.999116135501558e-05, "loss": 0.5927, "step": 861 }, { "epoch": 0.04, "learning_rate": 1.9991093244803567e-05, "loss": 0.2919, "step": 862 }, { "epoch": 0.04, "learning_rate": 1.9991024873287602e-05, "loss": 0.513, "step": 863 }, { "epoch": 0.04, "learning_rate": 1.9990956240469464e-05, "loss": 0.5081, "step": 864 }, { "epoch": 0.04, "learning_rate": 1.9990887346350952e-05, "loss": 0.5821, "step": 865 }, { "epoch": 0.04, "learning_rate": 1.9990818190933865e-05, "loss": 0.3195, "step": 866 }, { "epoch": 0.04, "learning_rate": 1.9990748774220016e-05, "loss": 0.1755, "step": 867 }, { "epoch": 0.04, "learning_rate": 1.9990679096211217e-05, "loss": 0.3703, "step": 868 }, { "epoch": 0.04, "learning_rate": 1.9990609156909295e-05, "loss": 0.5779, "step": 869 }, { "epoch": 0.04, "learning_rate": 1.9990538956316073e-05, "loss": 0.4701, "step": 870 }, { "epoch": 0.04, "learning_rate": 1.9990468494433393e-05, "loss": 0.4756, "step": 871 }, { "epoch": 0.04, "learning_rate": 1.999039777126309e-05, "loss": 0.354, "step": 872 }, { "epoch": 0.04, "learning_rate": 1.9990326786807023e-05, "loss": 0.5444, "step": 873 }, { "epoch": 0.04, "learning_rate": 1.999025554106704e-05, "loss": 0.3025, "step": 874 }, { "epoch": 0.04, "learning_rate": 1.999018403404501e-05, "loss": 0.4335, "step": 875 }, { "epoch": 0.04, "learning_rate": 1.99901122657428e-05, "loss": 0.4621, "step": 876 }, { "epoch": 0.04, "learning_rate": 1.999004023616229e-05, "loss": 0.497, "step": 877 }, { "epoch": 0.04, "learning_rate": 1.998996794530536e-05, "loss": 0.4206, "step": 878 }, { "epoch": 0.04, "learning_rate": 1.9989895393173906e-05, "loss": 0.4808, "step": 879 }, { "epoch": 0.04, "learning_rate": 1.998982257976982e-05, "loss": 0.5142, "step": 880 }, { "epoch": 0.04, "learning_rate": 1.998974950509501e-05, "loss": 0.4406, "step": 881 }, { "epoch": 0.04, "learning_rate": 1.9989676169151386e-05, "loss": 0.3948, "step": 882 }, { "epoch": 0.04, "learning_rate": 1.9989602571940866e-05, "loss": 0.4524, "step": 883 }, { "epoch": 0.04, "learning_rate": 1.9989528713465374e-05, "loss": 0.5723, "step": 884 }, { "epoch": 0.04, "learning_rate": 1.9989454593726843e-05, "loss": 0.2659, "step": 885 }, { "epoch": 0.04, "learning_rate": 1.998938021272721e-05, "loss": 0.5868, "step": 886 }, { "epoch": 0.04, "learning_rate": 1.9989305570468422e-05, "loss": 0.7007, "step": 887 }, { "epoch": 0.04, "learning_rate": 1.998923066695243e-05, "loss": 0.5063, "step": 888 }, { "epoch": 0.04, "learning_rate": 1.9989155502181194e-05, "loss": 0.3909, "step": 889 }, { "epoch": 0.04, "learning_rate": 1.9989080076156678e-05, "loss": 0.8235, "step": 890 }, { "epoch": 0.04, "learning_rate": 1.9989004388880858e-05, "loss": 0.2523, "step": 891 }, { "epoch": 0.04, "learning_rate": 1.998892844035571e-05, "loss": 0.2642, "step": 892 }, { "epoch": 0.04, "learning_rate": 1.9988852230583222e-05, "loss": 0.2485, "step": 893 }, { "epoch": 0.04, "learning_rate": 1.998877575956539e-05, "loss": 0.6707, "step": 894 }, { "epoch": 0.04, "learning_rate": 1.9988699027304206e-05, "loss": 0.4133, "step": 895 }, { "epoch": 0.04, "learning_rate": 1.9988622033801684e-05, "loss": 0.5789, "step": 896 }, { "epoch": 0.04, "learning_rate": 1.9988544779059834e-05, "loss": 0.3578, "step": 897 }, { "epoch": 0.04, "learning_rate": 1.998846726308068e-05, "loss": 0.5302, "step": 898 }, { "epoch": 0.04, "learning_rate": 1.9988389485866247e-05, "loss": 0.3505, "step": 899 }, { "epoch": 0.04, "learning_rate": 1.9988311447418568e-05, "loss": 0.4337, "step": 900 }, { "epoch": 0.04, "learning_rate": 1.9988233147739686e-05, "loss": 0.471, "step": 901 }, { "epoch": 0.05, "learning_rate": 1.998815458683165e-05, "loss": 0.6498, "step": 902 }, { "epoch": 0.05, "learning_rate": 1.998807576469651e-05, "loss": 0.247, "step": 903 }, { "epoch": 0.05, "learning_rate": 1.9987996681336333e-05, "loss": 0.6298, "step": 904 }, { "epoch": 0.05, "learning_rate": 1.9987917336753187e-05, "loss": 0.3025, "step": 905 }, { "epoch": 0.05, "learning_rate": 1.998783773094914e-05, "loss": 0.5415, "step": 906 }, { "epoch": 0.05, "learning_rate": 1.9987757863926282e-05, "loss": 0.3206, "step": 907 }, { "epoch": 0.05, "learning_rate": 1.9987677735686698e-05, "loss": 0.4473, "step": 908 }, { "epoch": 0.05, "learning_rate": 1.9987597346232483e-05, "loss": 0.3215, "step": 909 }, { "epoch": 0.05, "learning_rate": 1.9987516695565738e-05, "loss": 0.3951, "step": 910 }, { "epoch": 0.05, "learning_rate": 1.998743578368858e-05, "loss": 0.1586, "step": 911 }, { "epoch": 0.05, "learning_rate": 1.998735461060312e-05, "loss": 0.3063, "step": 912 }, { "epoch": 0.05, "learning_rate": 1.998727317631148e-05, "loss": 0.4744, "step": 913 }, { "epoch": 0.05, "learning_rate": 1.9987191480815792e-05, "loss": 0.2256, "step": 914 }, { "epoch": 0.05, "learning_rate": 1.998710952411819e-05, "loss": 0.6806, "step": 915 }, { "epoch": 0.05, "learning_rate": 1.998702730622082e-05, "loss": 0.7208, "step": 916 }, { "epoch": 0.05, "learning_rate": 1.998694482712583e-05, "loss": 0.472, "step": 917 }, { "epoch": 0.05, "learning_rate": 1.9986862086835378e-05, "loss": 0.404, "step": 918 }, { "epoch": 0.05, "learning_rate": 1.998677908535163e-05, "loss": 0.3698, "step": 919 }, { "epoch": 0.05, "learning_rate": 1.9986695822676758e-05, "loss": 0.3903, "step": 920 }, { "epoch": 0.05, "learning_rate": 1.9986612298812935e-05, "loss": 0.5167, "step": 921 }, { "epoch": 0.05, "learning_rate": 1.9986528513762348e-05, "loss": 0.357, "step": 922 }, { "epoch": 0.05, "learning_rate": 1.9986444467527187e-05, "loss": 0.5444, "step": 923 }, { "epoch": 0.05, "learning_rate": 1.998636016010965e-05, "loss": 0.3475, "step": 924 }, { "epoch": 0.05, "learning_rate": 1.9986275591511944e-05, "loss": 0.4741, "step": 925 }, { "epoch": 0.05, "learning_rate": 1.9986190761736277e-05, "loss": 0.4286, "step": 926 }, { "epoch": 0.05, "learning_rate": 1.9986105670784875e-05, "loss": 0.4293, "step": 927 }, { "epoch": 0.05, "learning_rate": 1.9986020318659954e-05, "loss": 0.5705, "step": 928 }, { "epoch": 0.05, "learning_rate": 1.9985934705363754e-05, "loss": 0.3772, "step": 929 }, { "epoch": 0.05, "learning_rate": 1.998584883089851e-05, "loss": 0.4295, "step": 930 }, { "epoch": 0.05, "learning_rate": 1.998576269526647e-05, "loss": 0.4914, "step": 931 }, { "epoch": 0.05, "learning_rate": 1.9985676298469886e-05, "loss": 0.3548, "step": 932 }, { "epoch": 0.05, "learning_rate": 1.9985589640511014e-05, "loss": 0.3557, "step": 933 }, { "epoch": 0.05, "learning_rate": 1.9985502721392126e-05, "loss": 0.646, "step": 934 }, { "epoch": 0.05, "learning_rate": 1.998541554111549e-05, "loss": 0.4684, "step": 935 }, { "epoch": 0.05, "learning_rate": 1.998532809968339e-05, "loss": 0.4244, "step": 936 }, { "epoch": 0.05, "learning_rate": 1.9985240397098118e-05, "loss": 0.4209, "step": 937 }, { "epoch": 0.05, "learning_rate": 1.9985152433361956e-05, "loss": 0.4789, "step": 938 }, { "epoch": 0.05, "learning_rate": 1.998506420847721e-05, "loss": 0.7116, "step": 939 }, { "epoch": 0.05, "learning_rate": 1.9984975722446192e-05, "loss": 0.1824, "step": 940 }, { "epoch": 0.05, "learning_rate": 1.998488697527121e-05, "loss": 0.3872, "step": 941 }, { "epoch": 0.05, "learning_rate": 1.998479796695458e-05, "loss": 0.7147, "step": 942 }, { "epoch": 0.05, "learning_rate": 1.998470869749865e-05, "loss": 0.3492, "step": 943 }, { "epoch": 0.05, "learning_rate": 1.9984619166905732e-05, "loss": 0.4238, "step": 944 }, { "epoch": 0.05, "learning_rate": 1.998452937517818e-05, "loss": 0.6941, "step": 945 }, { "epoch": 0.05, "learning_rate": 1.9984439322318342e-05, "loss": 0.3621, "step": 946 }, { "epoch": 0.05, "learning_rate": 1.998434900832857e-05, "loss": 0.5748, "step": 947 }, { "epoch": 0.05, "learning_rate": 1.9984258433211223e-05, "loss": 0.3579, "step": 948 }, { "epoch": 0.05, "learning_rate": 1.9984167596968676e-05, "loss": 0.6284, "step": 949 }, { "epoch": 0.05, "learning_rate": 1.9984076499603304e-05, "loss": 0.4106, "step": 950 }, { "epoch": 0.05, "learning_rate": 1.9983985141117484e-05, "loss": 0.4634, "step": 951 }, { "epoch": 0.05, "learning_rate": 1.998389352151361e-05, "loss": 0.4575, "step": 952 }, { "epoch": 0.05, "learning_rate": 1.998380164079408e-05, "loss": 0.4762, "step": 953 }, { "epoch": 0.05, "learning_rate": 1.9983709498961296e-05, "loss": 0.4596, "step": 954 }, { "epoch": 0.05, "learning_rate": 1.9983617096017662e-05, "loss": 0.5349, "step": 955 }, { "epoch": 0.05, "learning_rate": 1.9983524431965602e-05, "loss": 0.7165, "step": 956 }, { "epoch": 0.05, "learning_rate": 1.9983431506807535e-05, "loss": 0.5047, "step": 957 }, { "epoch": 0.05, "learning_rate": 1.9983338320545893e-05, "loss": 0.4727, "step": 958 }, { "epoch": 0.05, "learning_rate": 1.9983244873183113e-05, "loss": 0.2672, "step": 959 }, { "epoch": 0.05, "learning_rate": 1.9983151164721638e-05, "loss": 0.4627, "step": 960 }, { "epoch": 0.05, "learning_rate": 1.9983057195163924e-05, "loss": 0.324, "step": 961 }, { "epoch": 0.05, "learning_rate": 1.9982962964512423e-05, "loss": 0.3594, "step": 962 }, { "epoch": 0.05, "learning_rate": 1.99828684727696e-05, "loss": 0.5949, "step": 963 }, { "epoch": 0.05, "learning_rate": 1.9982773719937928e-05, "loss": 0.3841, "step": 964 }, { "epoch": 0.05, "learning_rate": 1.9982678706019882e-05, "loss": 0.3775, "step": 965 }, { "epoch": 0.05, "learning_rate": 1.998258343101795e-05, "loss": 0.2114, "step": 966 }, { "epoch": 0.05, "learning_rate": 1.9982487894934627e-05, "loss": 0.4971, "step": 967 }, { "epoch": 0.05, "learning_rate": 1.9982392097772405e-05, "loss": 0.2869, "step": 968 }, { "epoch": 0.05, "learning_rate": 1.998229603953379e-05, "loss": 0.442, "step": 969 }, { "epoch": 0.05, "learning_rate": 1.99821997202213e-05, "loss": 0.2439, "step": 970 }, { "epoch": 0.05, "learning_rate": 1.998210313983745e-05, "loss": 0.9311, "step": 971 }, { "epoch": 0.05, "learning_rate": 1.9982006298384763e-05, "loss": 0.7073, "step": 972 }, { "epoch": 0.05, "learning_rate": 1.998190919586578e-05, "loss": 0.1894, "step": 973 }, { "epoch": 0.05, "learning_rate": 1.9981811832283036e-05, "loss": 0.4441, "step": 974 }, { "epoch": 0.05, "learning_rate": 1.9981714207639075e-05, "loss": 0.4474, "step": 975 }, { "epoch": 0.05, "learning_rate": 1.9981616321936453e-05, "loss": 0.2455, "step": 976 }, { "epoch": 0.05, "learning_rate": 1.998151817517773e-05, "loss": 0.4913, "step": 977 }, { "epoch": 0.05, "learning_rate": 1.9981419767365474e-05, "loss": 0.438, "step": 978 }, { "epoch": 0.05, "learning_rate": 1.9981321098502256e-05, "loss": 0.4414, "step": 979 }, { "epoch": 0.05, "learning_rate": 1.9981222168590657e-05, "loss": 0.3778, "step": 980 }, { "epoch": 0.05, "learning_rate": 1.9981122977633265e-05, "loss": 0.3688, "step": 981 }, { "epoch": 0.05, "learning_rate": 1.9981023525632675e-05, "loss": 0.5525, "step": 982 }, { "epoch": 0.05, "learning_rate": 1.998092381259149e-05, "loss": 0.4054, "step": 983 }, { "epoch": 0.05, "learning_rate": 1.9980823838512314e-05, "loss": 0.2627, "step": 984 }, { "epoch": 0.05, "learning_rate": 1.998072360339776e-05, "loss": 0.4916, "step": 985 }, { "epoch": 0.05, "learning_rate": 1.9980623107250457e-05, "loss": 0.5063, "step": 986 }, { "epoch": 0.05, "learning_rate": 1.9980522350073025e-05, "loss": 0.5045, "step": 987 }, { "epoch": 0.05, "learning_rate": 1.9980421331868105e-05, "loss": 0.4592, "step": 988 }, { "epoch": 0.05, "learning_rate": 1.9980320052638338e-05, "loss": 0.5824, "step": 989 }, { "epoch": 0.05, "learning_rate": 1.9980218512386372e-05, "loss": 0.4045, "step": 990 }, { "epoch": 0.05, "learning_rate": 1.998011671111486e-05, "loss": 0.2375, "step": 991 }, { "epoch": 0.05, "learning_rate": 1.9980014648826465e-05, "loss": 0.5545, "step": 992 }, { "epoch": 0.05, "learning_rate": 1.9979912325523863e-05, "loss": 0.4198, "step": 993 }, { "epoch": 0.05, "learning_rate": 1.9979809741209726e-05, "loss": 0.3729, "step": 994 }, { "epoch": 0.05, "learning_rate": 1.9979706895886736e-05, "loss": 0.3174, "step": 995 }, { "epoch": 0.05, "learning_rate": 1.997960378955758e-05, "loss": 0.393, "step": 996 }, { "epoch": 0.05, "learning_rate": 1.997950042222496e-05, "loss": 0.5604, "step": 997 }, { "epoch": 0.05, "learning_rate": 1.9979396793891577e-05, "loss": 0.3713, "step": 998 }, { "epoch": 0.05, "learning_rate": 1.997929290456014e-05, "loss": 0.4979, "step": 999 }, { "epoch": 0.05, "learning_rate": 1.997918875423337e-05, "loss": 0.2733, "step": 1000 }, { "epoch": 0.05, "learning_rate": 1.9979084342913987e-05, "loss": 0.4051, "step": 1001 }, { "epoch": 0.05, "learning_rate": 1.9978979670604725e-05, "loss": 0.1703, "step": 1002 }, { "epoch": 0.05, "learning_rate": 1.9978874737308318e-05, "loss": 0.4956, "step": 1003 }, { "epoch": 0.05, "learning_rate": 1.9978769543027514e-05, "loss": 0.3422, "step": 1004 }, { "epoch": 0.05, "learning_rate": 1.997866408776506e-05, "loss": 0.6098, "step": 1005 }, { "epoch": 0.05, "learning_rate": 1.9978558371523718e-05, "loss": 0.3887, "step": 1006 }, { "epoch": 0.05, "learning_rate": 1.997845239430625e-05, "loss": 0.4343, "step": 1007 }, { "epoch": 0.05, "learning_rate": 1.997834615611543e-05, "loss": 0.6732, "step": 1008 }, { "epoch": 0.05, "learning_rate": 1.9978239656954036e-05, "loss": 0.5723, "step": 1009 }, { "epoch": 0.05, "learning_rate": 1.9978132896824853e-05, "loss": 0.1781, "step": 1010 }, { "epoch": 0.05, "learning_rate": 1.9978025875730673e-05, "loss": 0.373, "step": 1011 }, { "epoch": 0.05, "learning_rate": 1.9977918593674296e-05, "loss": 0.5584, "step": 1012 }, { "epoch": 0.05, "learning_rate": 1.9977811050658524e-05, "loss": 0.3604, "step": 1013 }, { "epoch": 0.05, "learning_rate": 1.9977703246686174e-05, "loss": 0.5495, "step": 1014 }, { "epoch": 0.05, "learning_rate": 1.9977595181760067e-05, "loss": 0.4594, "step": 1015 }, { "epoch": 0.05, "learning_rate": 1.997748685588302e-05, "loss": 0.5114, "step": 1016 }, { "epoch": 0.05, "learning_rate": 1.9977378269057877e-05, "loss": 0.3768, "step": 1017 }, { "epoch": 0.05, "learning_rate": 1.9977269421287474e-05, "loss": 0.3717, "step": 1018 }, { "epoch": 0.05, "learning_rate": 1.9977160312574656e-05, "loss": 0.5981, "step": 1019 }, { "epoch": 0.05, "learning_rate": 1.9977050942922276e-05, "loss": 0.4226, "step": 1020 }, { "epoch": 0.05, "learning_rate": 1.9976941312333198e-05, "loss": 0.3856, "step": 1021 }, { "epoch": 0.05, "learning_rate": 1.9976831420810288e-05, "loss": 0.4696, "step": 1022 }, { "epoch": 0.05, "learning_rate": 1.9976721268356418e-05, "loss": 0.3168, "step": 1023 }, { "epoch": 0.05, "learning_rate": 1.997661085497447e-05, "loss": 0.4891, "step": 1024 }, { "epoch": 0.05, "learning_rate": 1.9976500180667334e-05, "loss": 0.2841, "step": 1025 }, { "epoch": 0.05, "learning_rate": 1.99763892454379e-05, "loss": 0.3457, "step": 1026 }, { "epoch": 0.05, "learning_rate": 1.9976278049289075e-05, "loss": 0.368, "step": 1027 }, { "epoch": 0.05, "learning_rate": 1.9976166592223763e-05, "loss": 0.5976, "step": 1028 }, { "epoch": 0.05, "learning_rate": 1.9976054874244878e-05, "loss": 0.3403, "step": 1029 }, { "epoch": 0.05, "learning_rate": 1.9975942895355348e-05, "loss": 0.271, "step": 1030 }, { "epoch": 0.05, "learning_rate": 1.9975830655558095e-05, "loss": 0.4727, "step": 1031 }, { "epoch": 0.05, "learning_rate": 1.9975718154856056e-05, "loss": 0.2609, "step": 1032 }, { "epoch": 0.05, "learning_rate": 1.997560539325218e-05, "loss": 0.5581, "step": 1033 }, { "epoch": 0.05, "learning_rate": 1.9975492370749404e-05, "loss": 0.4584, "step": 1034 }, { "epoch": 0.05, "learning_rate": 1.9975379087350694e-05, "loss": 0.4167, "step": 1035 }, { "epoch": 0.05, "learning_rate": 1.9975265543059012e-05, "loss": 0.4573, "step": 1036 }, { "epoch": 0.05, "learning_rate": 1.997515173787732e-05, "loss": 0.5426, "step": 1037 }, { "epoch": 0.05, "learning_rate": 1.9975037671808602e-05, "loss": 0.2437, "step": 1038 }, { "epoch": 0.05, "learning_rate": 1.9974923344855838e-05, "loss": 0.3906, "step": 1039 }, { "epoch": 0.05, "learning_rate": 1.9974808757022018e-05, "loss": 0.1691, "step": 1040 }, { "epoch": 0.05, "learning_rate": 1.997469390831014e-05, "loss": 0.2387, "step": 1041 }, { "epoch": 0.05, "learning_rate": 1.997457879872321e-05, "loss": 0.5516, "step": 1042 }, { "epoch": 0.05, "learning_rate": 1.997446342826423e-05, "loss": 0.3194, "step": 1043 }, { "epoch": 0.05, "learning_rate": 1.997434779693623e-05, "loss": 0.3611, "step": 1044 }, { "epoch": 0.05, "learning_rate": 1.9974231904742224e-05, "loss": 0.5858, "step": 1045 }, { "epoch": 0.05, "learning_rate": 1.9974115751685247e-05, "loss": 0.3131, "step": 1046 }, { "epoch": 0.05, "learning_rate": 1.9973999337768336e-05, "loss": 0.4078, "step": 1047 }, { "epoch": 0.05, "learning_rate": 1.9973882662994534e-05, "loss": 0.4388, "step": 1048 }, { "epoch": 0.05, "learning_rate": 1.9973765727366897e-05, "loss": 0.5168, "step": 1049 }, { "epoch": 0.05, "learning_rate": 1.997364853088848e-05, "loss": 0.3318, "step": 1050 }, { "epoch": 0.05, "learning_rate": 1.9973531073562348e-05, "loss": 0.3612, "step": 1051 }, { "epoch": 0.05, "learning_rate": 1.9973413355391575e-05, "loss": 0.5181, "step": 1052 }, { "epoch": 0.05, "learning_rate": 1.997329537637924e-05, "loss": 0.3398, "step": 1053 }, { "epoch": 0.05, "learning_rate": 1.997317713652842e-05, "loss": 0.2207, "step": 1054 }, { "epoch": 0.05, "learning_rate": 1.997305863584222e-05, "loss": 0.4638, "step": 1055 }, { "epoch": 0.05, "learning_rate": 1.9972939874323735e-05, "loss": 0.2483, "step": 1056 }, { "epoch": 0.05, "learning_rate": 1.997282085197607e-05, "loss": 0.6302, "step": 1057 }, { "epoch": 0.05, "learning_rate": 1.997270156880233e-05, "loss": 0.3015, "step": 1058 }, { "epoch": 0.05, "learning_rate": 1.9972582024805655e-05, "loss": 0.4216, "step": 1059 }, { "epoch": 0.05, "learning_rate": 1.997246221998915e-05, "loss": 0.2963, "step": 1060 }, { "epoch": 0.05, "learning_rate": 1.997234215435596e-05, "loss": 0.4794, "step": 1061 }, { "epoch": 0.05, "learning_rate": 1.997222182790922e-05, "loss": 0.321, "step": 1062 }, { "epoch": 0.05, "learning_rate": 1.9972101240652083e-05, "loss": 0.74, "step": 1063 }, { "epoch": 0.05, "learning_rate": 1.9971980392587696e-05, "loss": 0.2671, "step": 1064 }, { "epoch": 0.05, "learning_rate": 1.9971859283719224e-05, "loss": 0.3656, "step": 1065 }, { "epoch": 0.05, "learning_rate": 1.997173791404983e-05, "loss": 0.5251, "step": 1066 }, { "epoch": 0.05, "learning_rate": 1.9971616283582698e-05, "loss": 0.4789, "step": 1067 }, { "epoch": 0.05, "learning_rate": 1.9971494392321e-05, "loss": 0.6469, "step": 1068 }, { "epoch": 0.05, "learning_rate": 1.9971372240267922e-05, "loss": 0.5758, "step": 1069 }, { "epoch": 0.05, "learning_rate": 1.997124982742667e-05, "loss": 0.3542, "step": 1070 }, { "epoch": 0.05, "learning_rate": 1.9971127153800434e-05, "loss": 0.2882, "step": 1071 }, { "epoch": 0.05, "learning_rate": 1.997100421939243e-05, "loss": 0.262, "step": 1072 }, { "epoch": 0.05, "learning_rate": 1.9970881024205867e-05, "loss": 0.5768, "step": 1073 }, { "epoch": 0.05, "learning_rate": 1.9970757568243973e-05, "loss": 0.4428, "step": 1074 }, { "epoch": 0.05, "learning_rate": 1.9970633851509972e-05, "loss": 0.641, "step": 1075 }, { "epoch": 0.05, "learning_rate": 1.9970509874007104e-05, "loss": 0.4533, "step": 1076 }, { "epoch": 0.05, "learning_rate": 1.997038563573861e-05, "loss": 0.2736, "step": 1077 }, { "epoch": 0.05, "learning_rate": 1.9970261136707737e-05, "loss": 0.3617, "step": 1078 }, { "epoch": 0.05, "learning_rate": 1.9970136376917744e-05, "loss": 0.4829, "step": 1079 }, { "epoch": 0.05, "learning_rate": 1.997001135637189e-05, "loss": 0.411, "step": 1080 }, { "epoch": 0.05, "learning_rate": 1.996988607507345e-05, "loss": 0.2467, "step": 1081 }, { "epoch": 0.05, "learning_rate": 1.9969760533025698e-05, "loss": 0.4156, "step": 1082 }, { "epoch": 0.05, "learning_rate": 1.9969634730231914e-05, "loss": 0.6527, "step": 1083 }, { "epoch": 0.05, "learning_rate": 1.9969508666695395e-05, "loss": 0.3293, "step": 1084 }, { "epoch": 0.05, "learning_rate": 1.996938234241943e-05, "loss": 0.3861, "step": 1085 }, { "epoch": 0.05, "learning_rate": 1.996925575740733e-05, "loss": 0.3825, "step": 1086 }, { "epoch": 0.05, "learning_rate": 1.9969128911662406e-05, "loss": 0.5382, "step": 1087 }, { "epoch": 0.05, "learning_rate": 1.9969001805187972e-05, "loss": 0.3841, "step": 1088 }, { "epoch": 0.05, "learning_rate": 1.996887443798735e-05, "loss": 0.5577, "step": 1089 }, { "epoch": 0.05, "learning_rate": 1.9968746810063872e-05, "loss": 0.4203, "step": 1090 }, { "epoch": 0.05, "learning_rate": 1.996861892142088e-05, "loss": 0.6466, "step": 1091 }, { "epoch": 0.05, "learning_rate": 1.996849077206172e-05, "loss": 0.4252, "step": 1092 }, { "epoch": 0.05, "learning_rate": 1.9968362361989735e-05, "loss": 0.3152, "step": 1093 }, { "epoch": 0.05, "learning_rate": 1.996823369120829e-05, "loss": 0.1677, "step": 1094 }, { "epoch": 0.05, "learning_rate": 1.996810475972075e-05, "loss": 0.3976, "step": 1095 }, { "epoch": 0.05, "learning_rate": 1.996797556753049e-05, "loss": 0.4403, "step": 1096 }, { "epoch": 0.05, "learning_rate": 1.996784611464088e-05, "loss": 0.3872, "step": 1097 }, { "epoch": 0.05, "learning_rate": 1.996771640105531e-05, "loss": 0.4452, "step": 1098 }, { "epoch": 0.05, "learning_rate": 1.996758642677717e-05, "loss": 0.5558, "step": 1099 }, { "epoch": 0.05, "learning_rate": 1.996745619180987e-05, "loss": 0.2429, "step": 1100 }, { "epoch": 0.05, "learning_rate": 1.9967325696156802e-05, "loss": 0.4294, "step": 1101 }, { "epoch": 0.06, "learning_rate": 1.996719493982139e-05, "loss": 0.3892, "step": 1102 }, { "epoch": 0.06, "learning_rate": 1.9967063922807047e-05, "loss": 0.2857, "step": 1103 }, { "epoch": 0.06, "learning_rate": 1.99669326451172e-05, "loss": 0.2788, "step": 1104 }, { "epoch": 0.06, "learning_rate": 1.9966801106755285e-05, "loss": 0.4748, "step": 1105 }, { "epoch": 0.06, "learning_rate": 1.996666930772474e-05, "loss": 0.5051, "step": 1106 }, { "epoch": 0.06, "learning_rate": 1.9966537248029015e-05, "loss": 0.4479, "step": 1107 }, { "epoch": 0.06, "learning_rate": 1.9966404927671565e-05, "loss": 0.5377, "step": 1108 }, { "epoch": 0.06, "learning_rate": 1.9966272346655846e-05, "loss": 0.4133, "step": 1109 }, { "epoch": 0.06, "learning_rate": 1.9966139504985324e-05, "loss": 0.2397, "step": 1110 }, { "epoch": 0.06, "learning_rate": 1.996600640266348e-05, "loss": 0.5077, "step": 1111 }, { "epoch": 0.06, "learning_rate": 1.996587303969379e-05, "loss": 0.2786, "step": 1112 }, { "epoch": 0.06, "learning_rate": 1.9965739416079742e-05, "loss": 0.486, "step": 1113 }, { "epoch": 0.06, "learning_rate": 1.9965605531824833e-05, "loss": 0.5081, "step": 1114 }, { "epoch": 0.06, "learning_rate": 1.9965471386932566e-05, "loss": 0.3595, "step": 1115 }, { "epoch": 0.06, "learning_rate": 1.9965336981406446e-05, "loss": 0.3332, "step": 1116 }, { "epoch": 0.06, "learning_rate": 1.996520231524999e-05, "loss": 0.2279, "step": 1117 }, { "epoch": 0.06, "learning_rate": 1.996506738846672e-05, "loss": 0.4744, "step": 1118 }, { "epoch": 0.06, "learning_rate": 1.9964932201060163e-05, "loss": 0.3642, "step": 1119 }, { "epoch": 0.06, "learning_rate": 1.9964796753033857e-05, "loss": 0.5922, "step": 1120 }, { "epoch": 0.06, "learning_rate": 1.9964661044391342e-05, "loss": 0.4674, "step": 1121 }, { "epoch": 0.06, "learning_rate": 1.9964525075136173e-05, "loss": 0.2449, "step": 1122 }, { "epoch": 0.06, "learning_rate": 1.9964388845271897e-05, "loss": 0.5116, "step": 1123 }, { "epoch": 0.06, "learning_rate": 1.9964252354802084e-05, "loss": 0.565, "step": 1124 }, { "epoch": 0.06, "learning_rate": 1.99641156037303e-05, "loss": 0.4579, "step": 1125 }, { "epoch": 0.06, "learning_rate": 1.9963978592060123e-05, "loss": 0.5709, "step": 1126 }, { "epoch": 0.06, "learning_rate": 1.996384131979514e-05, "loss": 0.3698, "step": 1127 }, { "epoch": 0.06, "learning_rate": 1.9963703786938936e-05, "loss": 0.3608, "step": 1128 }, { "epoch": 0.06, "learning_rate": 1.9963565993495106e-05, "loss": 0.3418, "step": 1129 }, { "epoch": 0.06, "learning_rate": 1.996342793946726e-05, "loss": 0.501, "step": 1130 }, { "epoch": 0.06, "learning_rate": 1.9963289624859006e-05, "loss": 0.4207, "step": 1131 }, { "epoch": 0.06, "learning_rate": 1.996315104967396e-05, "loss": 0.2919, "step": 1132 }, { "epoch": 0.06, "learning_rate": 1.996301221391575e-05, "loss": 0.3397, "step": 1133 }, { "epoch": 0.06, "learning_rate": 1.9962873117588005e-05, "loss": 0.2916, "step": 1134 }, { "epoch": 0.06, "learning_rate": 1.996273376069436e-05, "loss": 0.4944, "step": 1135 }, { "epoch": 0.06, "learning_rate": 1.9962594143238463e-05, "loss": 0.3933, "step": 1136 }, { "epoch": 0.06, "learning_rate": 1.9962454265223968e-05, "loss": 0.2349, "step": 1137 }, { "epoch": 0.06, "learning_rate": 1.9962314126654524e-05, "loss": 0.3802, "step": 1138 }, { "epoch": 0.06, "learning_rate": 1.996217372753381e-05, "loss": 0.3683, "step": 1139 }, { "epoch": 0.06, "learning_rate": 1.9962033067865482e-05, "loss": 0.1853, "step": 1140 }, { "epoch": 0.06, "learning_rate": 1.9961892147653232e-05, "loss": 0.4417, "step": 1141 }, { "epoch": 0.06, "learning_rate": 1.996175096690074e-05, "loss": 0.5244, "step": 1142 }, { "epoch": 0.06, "learning_rate": 1.99616095256117e-05, "loss": 0.4826, "step": 1143 }, { "epoch": 0.06, "learning_rate": 1.996146782378981e-05, "loss": 0.3036, "step": 1144 }, { "epoch": 0.06, "learning_rate": 1.9961325861438775e-05, "loss": 0.4878, "step": 1145 }, { "epoch": 0.06, "learning_rate": 1.9961183638562308e-05, "loss": 0.5203, "step": 1146 }, { "epoch": 0.06, "learning_rate": 1.996104115516413e-05, "loss": 0.3339, "step": 1147 }, { "epoch": 0.06, "learning_rate": 1.9960898411247972e-05, "loss": 0.1807, "step": 1148 }, { "epoch": 0.06, "learning_rate": 1.996075540681756e-05, "loss": 0.224, "step": 1149 }, { "epoch": 0.06, "learning_rate": 1.9960612141876632e-05, "loss": 0.2864, "step": 1150 }, { "epoch": 0.06, "learning_rate": 1.9960468616428944e-05, "loss": 0.3778, "step": 1151 }, { "epoch": 0.06, "learning_rate": 1.9960324830478242e-05, "loss": 0.3504, "step": 1152 }, { "epoch": 0.06, "learning_rate": 1.996018078402829e-05, "loss": 0.3322, "step": 1153 }, { "epoch": 0.06, "learning_rate": 1.9960036477082858e-05, "loss": 0.5547, "step": 1154 }, { "epoch": 0.06, "learning_rate": 1.9959891909645718e-05, "loss": 0.3527, "step": 1155 }, { "epoch": 0.06, "learning_rate": 1.9959747081720648e-05, "loss": 0.2915, "step": 1156 }, { "epoch": 0.06, "learning_rate": 1.9959601993311438e-05, "loss": 0.3813, "step": 1157 }, { "epoch": 0.06, "learning_rate": 1.9959456644421884e-05, "loss": 0.2361, "step": 1158 }, { "epoch": 0.06, "learning_rate": 1.9959311035055787e-05, "loss": 0.4032, "step": 1159 }, { "epoch": 0.06, "learning_rate": 1.9959165165216952e-05, "loss": 0.267, "step": 1160 }, { "epoch": 0.06, "learning_rate": 1.9959019034909194e-05, "loss": 0.3595, "step": 1161 }, { "epoch": 0.06, "learning_rate": 1.9958872644136342e-05, "loss": 0.2201, "step": 1162 }, { "epoch": 0.06, "learning_rate": 1.9958725992902213e-05, "loss": 0.3051, "step": 1163 }, { "epoch": 0.06, "learning_rate": 1.9958579081210658e-05, "loss": 0.4282, "step": 1164 }, { "epoch": 0.06, "learning_rate": 1.9958431909065507e-05, "loss": 0.6165, "step": 1165 }, { "epoch": 0.06, "learning_rate": 1.995828447647061e-05, "loss": 0.444, "step": 1166 }, { "epoch": 0.06, "learning_rate": 1.9958136783429826e-05, "loss": 0.4843, "step": 1167 }, { "epoch": 0.06, "learning_rate": 1.995798882994702e-05, "loss": 0.4783, "step": 1168 }, { "epoch": 0.06, "learning_rate": 1.9957840616026056e-05, "loss": 0.336, "step": 1169 }, { "epoch": 0.06, "learning_rate": 1.9957692141670816e-05, "loss": 0.4361, "step": 1170 }, { "epoch": 0.06, "learning_rate": 1.995754340688518e-05, "loss": 0.4551, "step": 1171 }, { "epoch": 0.06, "learning_rate": 1.9957394411673035e-05, "loss": 0.3109, "step": 1172 }, { "epoch": 0.06, "learning_rate": 1.9957245156038284e-05, "loss": 0.5015, "step": 1173 }, { "epoch": 0.06, "learning_rate": 1.9957095639984826e-05, "loss": 0.5009, "step": 1174 }, { "epoch": 0.06, "learning_rate": 1.9956945863516573e-05, "loss": 0.4608, "step": 1175 }, { "epoch": 0.06, "learning_rate": 1.9956795826637442e-05, "loss": 0.4784, "step": 1176 }, { "epoch": 0.06, "learning_rate": 1.9956645529351358e-05, "loss": 0.5882, "step": 1177 }, { "epoch": 0.06, "learning_rate": 1.995649497166225e-05, "loss": 0.6226, "step": 1178 }, { "epoch": 0.06, "learning_rate": 1.9956344153574057e-05, "loss": 0.6732, "step": 1179 }, { "epoch": 0.06, "learning_rate": 1.9956193075090726e-05, "loss": 0.547, "step": 1180 }, { "epoch": 0.06, "learning_rate": 1.99560417362162e-05, "loss": 0.7632, "step": 1181 }, { "epoch": 0.06, "learning_rate": 1.995589013695445e-05, "loss": 0.5105, "step": 1182 }, { "epoch": 0.06, "learning_rate": 1.995573827730943e-05, "loss": 0.5335, "step": 1183 }, { "epoch": 0.06, "learning_rate": 1.995558615728511e-05, "loss": 0.5472, "step": 1184 }, { "epoch": 0.06, "learning_rate": 1.995543377688548e-05, "loss": 0.7436, "step": 1185 }, { "epoch": 0.06, "learning_rate": 1.9955281136114515e-05, "loss": 0.3148, "step": 1186 }, { "epoch": 0.06, "learning_rate": 1.995512823497621e-05, "loss": 0.6644, "step": 1187 }, { "epoch": 0.06, "learning_rate": 1.9954975073474568e-05, "loss": 0.3571, "step": 1188 }, { "epoch": 0.06, "learning_rate": 1.9954821651613595e-05, "loss": 0.3138, "step": 1189 }, { "epoch": 0.06, "learning_rate": 1.9954667969397296e-05, "loss": 0.554, "step": 1190 }, { "epoch": 0.06, "learning_rate": 1.9954514026829692e-05, "loss": 0.553, "step": 1191 }, { "epoch": 0.06, "learning_rate": 1.9954359823914814e-05, "loss": 0.6627, "step": 1192 }, { "epoch": 0.06, "learning_rate": 1.9954205360656695e-05, "loss": 0.5708, "step": 1193 }, { "epoch": 0.06, "learning_rate": 1.995405063705937e-05, "loss": 0.4492, "step": 1194 }, { "epoch": 0.06, "learning_rate": 1.9953895653126887e-05, "loss": 0.5621, "step": 1195 }, { "epoch": 0.06, "learning_rate": 1.9953740408863303e-05, "loss": 0.2683, "step": 1196 }, { "epoch": 0.06, "learning_rate": 1.9953584904272674e-05, "loss": 0.3099, "step": 1197 }, { "epoch": 0.06, "learning_rate": 1.995342913935907e-05, "loss": 0.4823, "step": 1198 }, { "epoch": 0.06, "learning_rate": 1.995327311412656e-05, "loss": 0.4908, "step": 1199 }, { "epoch": 0.06, "learning_rate": 1.9953116828579234e-05, "loss": 0.3346, "step": 1200 }, { "epoch": 0.06, "learning_rate": 1.995296028272117e-05, "loss": 0.2815, "step": 1201 }, { "epoch": 0.06, "learning_rate": 1.9952803476556465e-05, "loss": 0.4638, "step": 1202 }, { "epoch": 0.06, "learning_rate": 1.995264641008922e-05, "loss": 0.4525, "step": 1203 }, { "epoch": 0.06, "learning_rate": 1.9952489083323548e-05, "loss": 0.3863, "step": 1204 }, { "epoch": 0.06, "learning_rate": 1.9952331496263556e-05, "loss": 0.3803, "step": 1205 }, { "epoch": 0.06, "learning_rate": 1.9952173648913372e-05, "loss": 0.3795, "step": 1206 }, { "epoch": 0.06, "learning_rate": 1.9952015541277122e-05, "loss": 0.1134, "step": 1207 }, { "epoch": 0.06, "learning_rate": 1.9951857173358937e-05, "loss": 0.3004, "step": 1208 }, { "epoch": 0.06, "learning_rate": 1.9951698545162963e-05, "loss": 0.3809, "step": 1209 }, { "epoch": 0.06, "learning_rate": 1.9951539656693345e-05, "loss": 0.3419, "step": 1210 }, { "epoch": 0.06, "learning_rate": 1.995138050795425e-05, "loss": 0.5506, "step": 1211 }, { "epoch": 0.06, "learning_rate": 1.9951221098949824e-05, "loss": 0.3519, "step": 1212 }, { "epoch": 0.06, "learning_rate": 1.9951061429684245e-05, "loss": 0.323, "step": 1213 }, { "epoch": 0.06, "learning_rate": 1.9950901500161688e-05, "loss": 0.4558, "step": 1214 }, { "epoch": 0.06, "learning_rate": 1.9950741310386337e-05, "loss": 0.1619, "step": 1215 }, { "epoch": 0.06, "learning_rate": 1.995058086036238e-05, "loss": 0.3659, "step": 1216 }, { "epoch": 0.06, "learning_rate": 1.9950420150094014e-05, "loss": 0.4306, "step": 1217 }, { "epoch": 0.06, "learning_rate": 1.995025917958544e-05, "loss": 0.439, "step": 1218 }, { "epoch": 0.06, "learning_rate": 1.9950097948840872e-05, "loss": 0.4805, "step": 1219 }, { "epoch": 0.06, "learning_rate": 1.9949936457864522e-05, "loss": 0.3087, "step": 1220 }, { "epoch": 0.06, "learning_rate": 1.994977470666062e-05, "loss": 0.736, "step": 1221 }, { "epoch": 0.06, "learning_rate": 1.994961269523339e-05, "loss": 0.4274, "step": 1222 }, { "epoch": 0.06, "learning_rate": 1.9949450423587068e-05, "loss": 0.4449, "step": 1223 }, { "epoch": 0.06, "learning_rate": 1.9949287891725906e-05, "loss": 0.1748, "step": 1224 }, { "epoch": 0.06, "learning_rate": 1.9949125099654148e-05, "loss": 0.4602, "step": 1225 }, { "epoch": 0.06, "learning_rate": 1.9948962047376055e-05, "loss": 0.4654, "step": 1226 }, { "epoch": 0.06, "learning_rate": 1.994879873489589e-05, "loss": 0.4995, "step": 1227 }, { "epoch": 0.06, "learning_rate": 1.9948635162217926e-05, "loss": 0.4051, "step": 1228 }, { "epoch": 0.06, "learning_rate": 1.994847132934644e-05, "loss": 0.4563, "step": 1229 }, { "epoch": 0.06, "learning_rate": 1.9948307236285715e-05, "loss": 0.3782, "step": 1230 }, { "epoch": 0.06, "learning_rate": 1.9948142883040048e-05, "loss": 0.254, "step": 1231 }, { "epoch": 0.06, "learning_rate": 1.9947978269613727e-05, "loss": 0.2728, "step": 1232 }, { "epoch": 0.06, "learning_rate": 1.9947813396011065e-05, "loss": 0.344, "step": 1233 }, { "epoch": 0.06, "learning_rate": 1.9947648262236376e-05, "loss": 0.3196, "step": 1234 }, { "epoch": 0.06, "learning_rate": 1.9947482868293978e-05, "loss": 0.4254, "step": 1235 }, { "epoch": 0.06, "learning_rate": 1.9947317214188192e-05, "loss": 0.596, "step": 1236 }, { "epoch": 0.06, "learning_rate": 1.9947151299923352e-05, "loss": 0.5281, "step": 1237 }, { "epoch": 0.06, "learning_rate": 1.9946985125503797e-05, "loss": 0.2786, "step": 1238 }, { "epoch": 0.06, "learning_rate": 1.9946818690933875e-05, "loss": 0.3192, "step": 1239 }, { "epoch": 0.06, "learning_rate": 1.994665199621794e-05, "loss": 0.4967, "step": 1240 }, { "epoch": 0.06, "learning_rate": 1.994648504136035e-05, "loss": 0.5033, "step": 1241 }, { "epoch": 0.06, "learning_rate": 1.9946317826365468e-05, "loss": 0.4907, "step": 1242 }, { "epoch": 0.06, "learning_rate": 1.994615035123767e-05, "loss": 0.4102, "step": 1243 }, { "epoch": 0.06, "learning_rate": 1.9945982615981342e-05, "loss": 0.4464, "step": 1244 }, { "epoch": 0.06, "learning_rate": 1.9945814620600864e-05, "loss": 0.4138, "step": 1245 }, { "epoch": 0.06, "learning_rate": 1.994564636510063e-05, "loss": 0.4339, "step": 1246 }, { "epoch": 0.06, "learning_rate": 1.994547784948504e-05, "loss": 0.4224, "step": 1247 }, { "epoch": 0.06, "learning_rate": 1.994530907375851e-05, "loss": 0.393, "step": 1248 }, { "epoch": 0.06, "learning_rate": 1.994514003792544e-05, "loss": 0.5143, "step": 1249 }, { "epoch": 0.06, "learning_rate": 1.9944970741990258e-05, "loss": 0.3508, "step": 1250 }, { "epoch": 0.06, "learning_rate": 1.9944801185957395e-05, "loss": 0.3418, "step": 1251 }, { "epoch": 0.06, "learning_rate": 1.9944631369831282e-05, "loss": 0.344, "step": 1252 }, { "epoch": 0.06, "learning_rate": 1.9944461293616357e-05, "loss": 0.4576, "step": 1253 }, { "epoch": 0.06, "learning_rate": 1.9944290957317075e-05, "loss": 0.4154, "step": 1254 }, { "epoch": 0.06, "learning_rate": 1.994412036093789e-05, "loss": 0.4276, "step": 1255 }, { "epoch": 0.06, "learning_rate": 1.9943949504483257e-05, "loss": 0.3414, "step": 1256 }, { "epoch": 0.06, "learning_rate": 1.994377838795765e-05, "loss": 0.206, "step": 1257 }, { "epoch": 0.06, "learning_rate": 1.9943607011365548e-05, "loss": 0.3534, "step": 1258 }, { "epoch": 0.06, "learning_rate": 1.9943435374711423e-05, "loss": 0.3339, "step": 1259 }, { "epoch": 0.06, "learning_rate": 1.994326347799977e-05, "loss": 0.4981, "step": 1260 }, { "epoch": 0.06, "learning_rate": 1.9943091321235088e-05, "loss": 0.2868, "step": 1261 }, { "epoch": 0.06, "learning_rate": 1.994291890442187e-05, "loss": 0.8621, "step": 1262 }, { "epoch": 0.06, "learning_rate": 1.9942746227564634e-05, "loss": 0.3357, "step": 1263 }, { "epoch": 0.06, "learning_rate": 1.994257329066789e-05, "loss": 0.3721, "step": 1264 }, { "epoch": 0.06, "learning_rate": 1.9942400093736167e-05, "loss": 0.4944, "step": 1265 }, { "epoch": 0.06, "learning_rate": 1.994222663677399e-05, "loss": 0.3922, "step": 1266 }, { "epoch": 0.06, "learning_rate": 1.99420529197859e-05, "loss": 0.5947, "step": 1267 }, { "epoch": 0.06, "learning_rate": 1.994187894277644e-05, "loss": 0.2203, "step": 1268 }, { "epoch": 0.06, "learning_rate": 1.9941704705750152e-05, "loss": 0.3757, "step": 1269 }, { "epoch": 0.06, "learning_rate": 1.99415302087116e-05, "loss": 0.2361, "step": 1270 }, { "epoch": 0.06, "learning_rate": 1.9941355451665346e-05, "loss": 0.5367, "step": 1271 }, { "epoch": 0.06, "learning_rate": 1.9941180434615963e-05, "loss": 0.5696, "step": 1272 }, { "epoch": 0.06, "learning_rate": 1.9941005157568028e-05, "loss": 0.4098, "step": 1273 }, { "epoch": 0.06, "learning_rate": 1.994082962052612e-05, "loss": 0.4826, "step": 1274 }, { "epoch": 0.06, "learning_rate": 1.9940653823494832e-05, "loss": 0.4288, "step": 1275 }, { "epoch": 0.06, "learning_rate": 1.9940477766478766e-05, "loss": 0.4694, "step": 1276 }, { "epoch": 0.06, "learning_rate": 1.9940301449482525e-05, "loss": 0.341, "step": 1277 }, { "epoch": 0.06, "learning_rate": 1.9940124872510717e-05, "loss": 0.4914, "step": 1278 }, { "epoch": 0.06, "learning_rate": 1.993994803556796e-05, "loss": 0.5407, "step": 1279 }, { "epoch": 0.06, "learning_rate": 1.9939770938658886e-05, "loss": 0.2999, "step": 1280 }, { "epoch": 0.06, "learning_rate": 1.993959358178812e-05, "loss": 0.3509, "step": 1281 }, { "epoch": 0.06, "learning_rate": 1.99394159649603e-05, "loss": 0.3602, "step": 1282 }, { "epoch": 0.06, "learning_rate": 1.9939238088180075e-05, "loss": 0.3765, "step": 1283 }, { "epoch": 0.06, "learning_rate": 1.9939059951452097e-05, "loss": 0.559, "step": 1284 }, { "epoch": 0.06, "learning_rate": 1.9938881554781025e-05, "loss": 0.2759, "step": 1285 }, { "epoch": 0.06, "learning_rate": 1.993870289817152e-05, "loss": 0.3801, "step": 1286 }, { "epoch": 0.06, "learning_rate": 1.9938523981628263e-05, "loss": 0.2024, "step": 1287 }, { "epoch": 0.06, "learning_rate": 1.9938344805155925e-05, "loss": 0.2379, "step": 1288 }, { "epoch": 0.06, "learning_rate": 1.9938165368759194e-05, "loss": 0.2352, "step": 1289 }, { "epoch": 0.06, "learning_rate": 1.9937985672442767e-05, "loss": 0.4441, "step": 1290 }, { "epoch": 0.06, "learning_rate": 1.993780571621134e-05, "loss": 0.4237, "step": 1291 }, { "epoch": 0.06, "learning_rate": 1.9937625500069623e-05, "loss": 0.4193, "step": 1292 }, { "epoch": 0.06, "learning_rate": 1.9937445024022325e-05, "loss": 0.1993, "step": 1293 }, { "epoch": 0.06, "learning_rate": 1.9937264288074165e-05, "loss": 0.4009, "step": 1294 }, { "epoch": 0.06, "learning_rate": 1.9937083292229876e-05, "loss": 0.6029, "step": 1295 }, { "epoch": 0.06, "learning_rate": 1.993690203649419e-05, "loss": 0.5434, "step": 1296 }, { "epoch": 0.06, "learning_rate": 1.9936720520871845e-05, "loss": 0.2127, "step": 1297 }, { "epoch": 0.06, "learning_rate": 1.993653874536759e-05, "loss": 0.4631, "step": 1298 }, { "epoch": 0.06, "learning_rate": 1.9936356709986178e-05, "loss": 0.3981, "step": 1299 }, { "epoch": 0.06, "learning_rate": 1.9936174414732372e-05, "loss": 0.3455, "step": 1300 }, { "epoch": 0.06, "learning_rate": 1.9935991859610935e-05, "loss": 0.38, "step": 1301 }, { "epoch": 0.07, "learning_rate": 1.9935809044626647e-05, "loss": 0.3103, "step": 1302 }, { "epoch": 0.07, "learning_rate": 1.993562596978429e-05, "loss": 0.4513, "step": 1303 }, { "epoch": 0.07, "learning_rate": 1.9935442635088645e-05, "loss": 0.581, "step": 1304 }, { "epoch": 0.07, "learning_rate": 1.993525904054451e-05, "loss": 0.5678, "step": 1305 }, { "epoch": 0.07, "learning_rate": 1.993507518615669e-05, "loss": 0.3123, "step": 1306 }, { "epoch": 0.07, "learning_rate": 1.993489107192999e-05, "loss": 0.3219, "step": 1307 }, { "epoch": 0.07, "learning_rate": 1.993470669786923e-05, "loss": 0.7052, "step": 1308 }, { "epoch": 0.07, "learning_rate": 1.9934522063979225e-05, "loss": 0.5591, "step": 1309 }, { "epoch": 0.07, "learning_rate": 1.9934337170264804e-05, "loss": 0.5031, "step": 1310 }, { "epoch": 0.07, "learning_rate": 1.9934152016730813e-05, "loss": 0.4137, "step": 1311 }, { "epoch": 0.07, "learning_rate": 1.9933966603382083e-05, "loss": 0.605, "step": 1312 }, { "epoch": 0.07, "learning_rate": 1.993378093022347e-05, "loss": 0.6213, "step": 1313 }, { "epoch": 0.07, "learning_rate": 1.9933594997259823e-05, "loss": 0.257, "step": 1314 }, { "epoch": 0.07, "learning_rate": 1.9933408804496016e-05, "loss": 0.2132, "step": 1315 }, { "epoch": 0.07, "learning_rate": 1.993322235193691e-05, "loss": 0.3854, "step": 1316 }, { "epoch": 0.07, "learning_rate": 1.993303563958738e-05, "loss": 0.3281, "step": 1317 }, { "epoch": 0.07, "learning_rate": 1.9932848667452318e-05, "loss": 0.4988, "step": 1318 }, { "epoch": 0.07, "learning_rate": 1.9932661435536603e-05, "loss": 0.274, "step": 1319 }, { "epoch": 0.07, "learning_rate": 1.9932473943845143e-05, "loss": 0.4772, "step": 1320 }, { "epoch": 0.07, "learning_rate": 1.9932286192382836e-05, "loss": 0.3917, "step": 1321 }, { "epoch": 0.07, "learning_rate": 1.993209818115459e-05, "loss": 0.4762, "step": 1322 }, { "epoch": 0.07, "learning_rate": 1.9931909910165324e-05, "loss": 0.6657, "step": 1323 }, { "epoch": 0.07, "learning_rate": 1.9931721379419967e-05, "loss": 0.4903, "step": 1324 }, { "epoch": 0.07, "learning_rate": 1.9931532588923443e-05, "loss": 0.2775, "step": 1325 }, { "epoch": 0.07, "learning_rate": 1.993134353868069e-05, "loss": 0.1873, "step": 1326 }, { "epoch": 0.07, "learning_rate": 1.9931154228696657e-05, "loss": 0.4831, "step": 1327 }, { "epoch": 0.07, "learning_rate": 1.993096465897629e-05, "loss": 0.6201, "step": 1328 }, { "epoch": 0.07, "learning_rate": 1.9930774829524554e-05, "loss": 0.4513, "step": 1329 }, { "epoch": 0.07, "learning_rate": 1.9930584740346405e-05, "loss": 0.5856, "step": 1330 }, { "epoch": 0.07, "learning_rate": 1.9930394391446822e-05, "loss": 0.6292, "step": 1331 }, { "epoch": 0.07, "learning_rate": 1.993020378283078e-05, "loss": 0.3994, "step": 1332 }, { "epoch": 0.07, "learning_rate": 1.9930012914503262e-05, "loss": 0.4681, "step": 1333 }, { "epoch": 0.07, "learning_rate": 1.9929821786469262e-05, "loss": 0.3542, "step": 1334 }, { "epoch": 0.07, "learning_rate": 1.9929630398733782e-05, "loss": 0.367, "step": 1335 }, { "epoch": 0.07, "learning_rate": 1.9929438751301822e-05, "loss": 0.3782, "step": 1336 }, { "epoch": 0.07, "learning_rate": 1.9929246844178397e-05, "loss": 0.3718, "step": 1337 }, { "epoch": 0.07, "learning_rate": 1.9929054677368524e-05, "loss": 0.4708, "step": 1338 }, { "epoch": 0.07, "learning_rate": 1.9928862250877235e-05, "loss": 0.1931, "step": 1339 }, { "epoch": 0.07, "learning_rate": 1.9928669564709557e-05, "loss": 0.3378, "step": 1340 }, { "epoch": 0.07, "learning_rate": 1.992847661887053e-05, "loss": 0.4757, "step": 1341 }, { "epoch": 0.07, "learning_rate": 1.9928283413365198e-05, "loss": 0.4483, "step": 1342 }, { "epoch": 0.07, "learning_rate": 1.9928089948198622e-05, "loss": 0.2741, "step": 1343 }, { "epoch": 0.07, "learning_rate": 1.9927896223375855e-05, "loss": 0.3766, "step": 1344 }, { "epoch": 0.07, "learning_rate": 1.9927702238901964e-05, "loss": 0.3156, "step": 1345 }, { "epoch": 0.07, "learning_rate": 1.9927507994782023e-05, "loss": 0.5795, "step": 1346 }, { "epoch": 0.07, "learning_rate": 1.9927313491021117e-05, "loss": 0.4905, "step": 1347 }, { "epoch": 0.07, "learning_rate": 1.9927118727624328e-05, "loss": 0.4006, "step": 1348 }, { "epoch": 0.07, "learning_rate": 1.992692370459675e-05, "loss": 0.2245, "step": 1349 }, { "epoch": 0.07, "learning_rate": 1.9926728421943483e-05, "loss": 0.5657, "step": 1350 }, { "epoch": 0.07, "learning_rate": 1.992653287966964e-05, "loss": 0.4978, "step": 1351 }, { "epoch": 0.07, "learning_rate": 1.992633707778033e-05, "loss": 0.388, "step": 1352 }, { "epoch": 0.07, "learning_rate": 1.9926141016280672e-05, "loss": 0.3126, "step": 1353 }, { "epoch": 0.07, "learning_rate": 1.9925944695175795e-05, "loss": 0.3387, "step": 1354 }, { "epoch": 0.07, "learning_rate": 1.992574811447084e-05, "loss": 0.623, "step": 1355 }, { "epoch": 0.07, "learning_rate": 1.9925551274170942e-05, "loss": 0.4391, "step": 1356 }, { "epoch": 0.07, "learning_rate": 1.992535417428125e-05, "loss": 0.4599, "step": 1357 }, { "epoch": 0.07, "learning_rate": 1.9925156814806918e-05, "loss": 0.365, "step": 1358 }, { "epoch": 0.07, "learning_rate": 1.992495919575311e-05, "loss": 0.479, "step": 1359 }, { "epoch": 0.07, "learning_rate": 1.9924761317124998e-05, "loss": 0.3966, "step": 1360 }, { "epoch": 0.07, "learning_rate": 1.9924563178927746e-05, "loss": 0.4476, "step": 1361 }, { "epoch": 0.07, "learning_rate": 1.9924364781166544e-05, "loss": 0.3859, "step": 1362 }, { "epoch": 0.07, "learning_rate": 1.9924166123846583e-05, "loss": 0.3858, "step": 1363 }, { "epoch": 0.07, "learning_rate": 1.9923967206973057e-05, "loss": 0.2595, "step": 1364 }, { "epoch": 0.07, "learning_rate": 1.9923768030551163e-05, "loss": 0.2837, "step": 1365 }, { "epoch": 0.07, "learning_rate": 1.9923568594586117e-05, "loss": 0.4289, "step": 1366 }, { "epoch": 0.07, "learning_rate": 1.9923368899083127e-05, "loss": 0.5129, "step": 1367 }, { "epoch": 0.07, "learning_rate": 1.992316894404742e-05, "loss": 0.3691, "step": 1368 }, { "epoch": 0.07, "learning_rate": 1.9922968729484234e-05, "loss": 0.0923, "step": 1369 }, { "epoch": 0.07, "learning_rate": 1.992276825539879e-05, "loss": 0.4685, "step": 1370 }, { "epoch": 0.07, "learning_rate": 1.9922567521796346e-05, "loss": 0.2795, "step": 1371 }, { "epoch": 0.07, "learning_rate": 1.992236652868214e-05, "loss": 0.4193, "step": 1372 }, { "epoch": 0.07, "learning_rate": 1.9922165276061432e-05, "loss": 0.4218, "step": 1373 }, { "epoch": 0.07, "learning_rate": 1.992196376393949e-05, "loss": 0.2421, "step": 1374 }, { "epoch": 0.07, "learning_rate": 1.992176199232158e-05, "loss": 0.5113, "step": 1375 }, { "epoch": 0.07, "learning_rate": 1.9921559961212982e-05, "loss": 0.4198, "step": 1376 }, { "epoch": 0.07, "learning_rate": 1.9921357670618977e-05, "loss": 0.5676, "step": 1377 }, { "epoch": 0.07, "learning_rate": 1.992115512054486e-05, "loss": 0.1755, "step": 1378 }, { "epoch": 0.07, "learning_rate": 1.992095231099592e-05, "loss": 0.4608, "step": 1379 }, { "epoch": 0.07, "learning_rate": 1.9920749241977467e-05, "loss": 0.5646, "step": 1380 }, { "epoch": 0.07, "learning_rate": 1.992054591349481e-05, "loss": 0.4778, "step": 1381 }, { "epoch": 0.07, "learning_rate": 1.9920342325553272e-05, "loss": 0.2457, "step": 1382 }, { "epoch": 0.07, "learning_rate": 1.9920138478158173e-05, "loss": 0.3287, "step": 1383 }, { "epoch": 0.07, "learning_rate": 1.9919934371314845e-05, "loss": 0.1751, "step": 1384 }, { "epoch": 0.07, "learning_rate": 1.9919730005028625e-05, "loss": 0.3876, "step": 1385 }, { "epoch": 0.07, "learning_rate": 1.991952537930486e-05, "loss": 0.6141, "step": 1386 }, { "epoch": 0.07, "learning_rate": 1.99193204941489e-05, "loss": 0.292, "step": 1387 }, { "epoch": 0.07, "learning_rate": 1.9919115349566108e-05, "loss": 0.3791, "step": 1388 }, { "epoch": 0.07, "learning_rate": 1.9918909945561844e-05, "loss": 0.3607, "step": 1389 }, { "epoch": 0.07, "learning_rate": 1.9918704282141478e-05, "loss": 0.4811, "step": 1390 }, { "epoch": 0.07, "learning_rate": 1.9918498359310397e-05, "loss": 0.3105, "step": 1391 }, { "epoch": 0.07, "learning_rate": 1.9918292177073983e-05, "loss": 0.4926, "step": 1392 }, { "epoch": 0.07, "learning_rate": 1.9918085735437628e-05, "loss": 0.5833, "step": 1393 }, { "epoch": 0.07, "learning_rate": 1.9917879034406726e-05, "loss": 0.5362, "step": 1394 }, { "epoch": 0.07, "learning_rate": 1.9917672073986692e-05, "loss": 0.3325, "step": 1395 }, { "epoch": 0.07, "learning_rate": 1.9917464854182934e-05, "loss": 0.4086, "step": 1396 }, { "epoch": 0.07, "learning_rate": 1.991725737500087e-05, "loss": 0.3667, "step": 1397 }, { "epoch": 0.07, "learning_rate": 1.991704963644593e-05, "loss": 0.4473, "step": 1398 }, { "epoch": 0.07, "learning_rate": 1.9916841638523547e-05, "loss": 0.2162, "step": 1399 }, { "epoch": 0.07, "learning_rate": 1.991663338123916e-05, "loss": 0.5148, "step": 1400 }, { "epoch": 0.07, "learning_rate": 1.9916424864598212e-05, "loss": 0.2741, "step": 1401 }, { "epoch": 0.07, "learning_rate": 1.9916216088606163e-05, "loss": 0.3416, "step": 1402 }, { "epoch": 0.07, "learning_rate": 1.991600705326847e-05, "loss": 0.5702, "step": 1403 }, { "epoch": 0.07, "learning_rate": 1.9915797758590597e-05, "loss": 0.227, "step": 1404 }, { "epoch": 0.07, "learning_rate": 1.9915588204578023e-05, "loss": 0.3113, "step": 1405 }, { "epoch": 0.07, "learning_rate": 1.991537839123623e-05, "loss": 0.7238, "step": 1406 }, { "epoch": 0.07, "learning_rate": 1.99151683185707e-05, "loss": 0.6633, "step": 1407 }, { "epoch": 0.07, "learning_rate": 1.991495798658693e-05, "loss": 0.3999, "step": 1408 }, { "epoch": 0.07, "learning_rate": 1.9914747395290415e-05, "loss": 0.3101, "step": 1409 }, { "epoch": 0.07, "learning_rate": 1.991453654468667e-05, "loss": 0.5399, "step": 1410 }, { "epoch": 0.07, "learning_rate": 1.991432543478121e-05, "loss": 0.5106, "step": 1411 }, { "epoch": 0.07, "learning_rate": 1.9914114065579552e-05, "loss": 0.4716, "step": 1412 }, { "epoch": 0.07, "learning_rate": 1.9913902437087224e-05, "loss": 0.4739, "step": 1413 }, { "epoch": 0.07, "learning_rate": 1.9913690549309765e-05, "loss": 0.4243, "step": 1414 }, { "epoch": 0.07, "learning_rate": 1.9913478402252716e-05, "loss": 0.2618, "step": 1415 }, { "epoch": 0.07, "learning_rate": 1.991326599592162e-05, "loss": 0.3015, "step": 1416 }, { "epoch": 0.07, "learning_rate": 1.9913053330322035e-05, "loss": 0.2909, "step": 1417 }, { "epoch": 0.07, "learning_rate": 1.9912840405459526e-05, "loss": 0.3631, "step": 1418 }, { "epoch": 0.07, "learning_rate": 1.9912627221339664e-05, "loss": 0.4153, "step": 1419 }, { "epoch": 0.07, "learning_rate": 1.9912413777968017e-05, "loss": 0.4237, "step": 1420 }, { "epoch": 0.07, "learning_rate": 1.9912200075350166e-05, "loss": 0.4526, "step": 1421 }, { "epoch": 0.07, "learning_rate": 1.9911986113491712e-05, "loss": 0.3189, "step": 1422 }, { "epoch": 0.07, "learning_rate": 1.9911771892398238e-05, "loss": 0.4414, "step": 1423 }, { "epoch": 0.07, "learning_rate": 1.9911557412075356e-05, "loss": 0.3681, "step": 1424 }, { "epoch": 0.07, "learning_rate": 1.9911342672528668e-05, "loss": 0.4946, "step": 1425 }, { "epoch": 0.07, "learning_rate": 1.9911127673763796e-05, "loss": 0.3272, "step": 1426 }, { "epoch": 0.07, "learning_rate": 1.9910912415786362e-05, "loss": 0.4354, "step": 1427 }, { "epoch": 0.07, "learning_rate": 1.9910696898601995e-05, "loss": 0.2603, "step": 1428 }, { "epoch": 0.07, "learning_rate": 1.9910481122216327e-05, "loss": 0.5619, "step": 1429 }, { "epoch": 0.07, "learning_rate": 1.991026508663501e-05, "loss": 0.5144, "step": 1430 }, { "epoch": 0.07, "learning_rate": 1.9910048791863688e-05, "loss": 0.3125, "step": 1431 }, { "epoch": 0.07, "learning_rate": 1.990983223790802e-05, "loss": 0.6642, "step": 1432 }, { "epoch": 0.07, "learning_rate": 1.9909615424773668e-05, "loss": 0.5307, "step": 1433 }, { "epoch": 0.07, "learning_rate": 1.9909398352466307e-05, "loss": 0.6804, "step": 1434 }, { "epoch": 0.07, "learning_rate": 1.990918102099161e-05, "loss": 0.342, "step": 1435 }, { "epoch": 0.07, "learning_rate": 1.9908963430355263e-05, "loss": 0.4095, "step": 1436 }, { "epoch": 0.07, "learning_rate": 1.9908745580562953e-05, "loss": 0.3094, "step": 1437 }, { "epoch": 0.07, "learning_rate": 1.990852747162038e-05, "loss": 0.5878, "step": 1438 }, { "epoch": 0.07, "learning_rate": 1.9908309103533252e-05, "loss": 0.342, "step": 1439 }, { "epoch": 0.07, "learning_rate": 1.9908090476307276e-05, "loss": 0.3619, "step": 1440 }, { "epoch": 0.07, "learning_rate": 1.990787158994817e-05, "loss": 0.32, "step": 1441 }, { "epoch": 0.07, "learning_rate": 1.9907652444461658e-05, "loss": 0.3227, "step": 1442 }, { "epoch": 0.07, "learning_rate": 1.9907433039853475e-05, "loss": 0.4058, "step": 1443 }, { "epoch": 0.07, "learning_rate": 1.9907213376129358e-05, "loss": 0.3969, "step": 1444 }, { "epoch": 0.07, "learning_rate": 1.990699345329505e-05, "loss": 0.1794, "step": 1445 }, { "epoch": 0.07, "learning_rate": 1.9906773271356305e-05, "loss": 0.3984, "step": 1446 }, { "epoch": 0.07, "learning_rate": 1.990655283031888e-05, "loss": 0.4727, "step": 1447 }, { "epoch": 0.07, "learning_rate": 1.990633213018854e-05, "loss": 0.4474, "step": 1448 }, { "epoch": 0.07, "learning_rate": 1.990611117097106e-05, "loss": 0.4775, "step": 1449 }, { "epoch": 0.07, "learning_rate": 1.9905889952672215e-05, "loss": 0.4818, "step": 1450 }, { "epoch": 0.07, "learning_rate": 1.9905668475297794e-05, "loss": 0.6457, "step": 1451 }, { "epoch": 0.07, "learning_rate": 1.990544673885359e-05, "loss": 0.4017, "step": 1452 }, { "epoch": 0.07, "learning_rate": 1.9905224743345397e-05, "loss": 0.7065, "step": 1453 }, { "epoch": 0.07, "learning_rate": 1.9905002488779026e-05, "loss": 0.2823, "step": 1454 }, { "epoch": 0.07, "learning_rate": 1.990477997516029e-05, "loss": 0.4, "step": 1455 }, { "epoch": 0.07, "learning_rate": 1.9904557202495e-05, "loss": 0.383, "step": 1456 }, { "epoch": 0.07, "learning_rate": 1.9904334170788996e-05, "loss": 0.3979, "step": 1457 }, { "epoch": 0.07, "learning_rate": 1.99041108800481e-05, "loss": 0.4576, "step": 1458 }, { "epoch": 0.07, "learning_rate": 1.990388733027816e-05, "loss": 0.3154, "step": 1459 }, { "epoch": 0.07, "learning_rate": 1.9903663521485015e-05, "loss": 0.4946, "step": 1460 }, { "epoch": 0.07, "learning_rate": 1.9903439453674523e-05, "loss": 0.4167, "step": 1461 }, { "epoch": 0.07, "learning_rate": 1.9903215126852546e-05, "loss": 0.4412, "step": 1462 }, { "epoch": 0.07, "learning_rate": 1.9902990541024946e-05, "loss": 0.5085, "step": 1463 }, { "epoch": 0.07, "learning_rate": 1.9902765696197595e-05, "loss": 0.2156, "step": 1464 }, { "epoch": 0.07, "learning_rate": 1.9902540592376382e-05, "loss": 0.2483, "step": 1465 }, { "epoch": 0.07, "learning_rate": 1.990231522956719e-05, "loss": 0.3607, "step": 1466 }, { "epoch": 0.07, "learning_rate": 1.990208960777591e-05, "loss": 0.2718, "step": 1467 }, { "epoch": 0.07, "learning_rate": 1.9901863727008452e-05, "loss": 0.4706, "step": 1468 }, { "epoch": 0.07, "learning_rate": 1.990163758727071e-05, "loss": 0.367, "step": 1469 }, { "epoch": 0.07, "learning_rate": 1.9901411188568608e-05, "loss": 0.3672, "step": 1470 }, { "epoch": 0.07, "learning_rate": 1.9901184530908066e-05, "loss": 0.3281, "step": 1471 }, { "epoch": 0.07, "learning_rate": 1.9900957614295013e-05, "loss": 0.44, "step": 1472 }, { "epoch": 0.07, "learning_rate": 1.990073043873538e-05, "loss": 0.2504, "step": 1473 }, { "epoch": 0.07, "learning_rate": 1.990050300423511e-05, "loss": 0.5569, "step": 1474 }, { "epoch": 0.07, "learning_rate": 1.9900275310800147e-05, "loss": 0.3533, "step": 1475 }, { "epoch": 0.07, "learning_rate": 1.9900047358436453e-05, "loss": 0.4392, "step": 1476 }, { "epoch": 0.07, "learning_rate": 1.9899819147149986e-05, "loss": 0.4616, "step": 1477 }, { "epoch": 0.07, "learning_rate": 1.989959067694672e-05, "loss": 0.4922, "step": 1478 }, { "epoch": 0.07, "learning_rate": 1.9899361947832624e-05, "loss": 0.555, "step": 1479 }, { "epoch": 0.07, "learning_rate": 1.989913295981368e-05, "loss": 0.4593, "step": 1480 }, { "epoch": 0.07, "learning_rate": 1.989890371289588e-05, "loss": 0.5085, "step": 1481 }, { "epoch": 0.07, "learning_rate": 1.989867420708522e-05, "loss": 0.252, "step": 1482 }, { "epoch": 0.07, "learning_rate": 1.98984444423877e-05, "loss": 0.3709, "step": 1483 }, { "epoch": 0.07, "learning_rate": 1.989821441880933e-05, "loss": 0.0887, "step": 1484 }, { "epoch": 0.07, "learning_rate": 1.9897984136356124e-05, "loss": 0.4021, "step": 1485 }, { "epoch": 0.07, "learning_rate": 1.989775359503411e-05, "loss": 0.4864, "step": 1486 }, { "epoch": 0.07, "learning_rate": 1.9897522794849315e-05, "loss": 0.3776, "step": 1487 }, { "epoch": 0.07, "learning_rate": 1.9897291735807772e-05, "loss": 0.6875, "step": 1488 }, { "epoch": 0.07, "learning_rate": 1.989706041791553e-05, "loss": 0.4565, "step": 1489 }, { "epoch": 0.07, "learning_rate": 1.9896828841178634e-05, "loss": 0.4884, "step": 1490 }, { "epoch": 0.07, "learning_rate": 1.989659700560314e-05, "loss": 0.4478, "step": 1491 }, { "epoch": 0.07, "learning_rate": 1.9896364911195117e-05, "loss": 0.4399, "step": 1492 }, { "epoch": 0.07, "learning_rate": 1.989613255796063e-05, "loss": 0.3579, "step": 1493 }, { "epoch": 0.07, "learning_rate": 1.989589994590576e-05, "loss": 0.188, "step": 1494 }, { "epoch": 0.07, "learning_rate": 1.9895667075036585e-05, "loss": 0.4861, "step": 1495 }, { "epoch": 0.07, "learning_rate": 1.9895433945359198e-05, "loss": 0.4438, "step": 1496 }, { "epoch": 0.07, "learning_rate": 1.98952005568797e-05, "loss": 0.2149, "step": 1497 }, { "epoch": 0.07, "learning_rate": 1.9894966909604188e-05, "loss": 0.3827, "step": 1498 }, { "epoch": 0.07, "learning_rate": 1.989473300353878e-05, "loss": 0.4408, "step": 1499 }, { "epoch": 0.07, "learning_rate": 1.989449883868959e-05, "loss": 0.5832, "step": 1500 }, { "epoch": 0.07, "learning_rate": 1.9894264415062738e-05, "loss": 0.4508, "step": 1501 }, { "epoch": 0.07, "learning_rate": 1.9894029732664366e-05, "loss": 0.5892, "step": 1502 }, { "epoch": 0.08, "learning_rate": 1.98937947915006e-05, "loss": 0.5483, "step": 1503 }, { "epoch": 0.08, "learning_rate": 1.9893559591577595e-05, "loss": 0.6316, "step": 1504 }, { "epoch": 0.08, "learning_rate": 1.989332413290149e-05, "loss": 0.2975, "step": 1505 }, { "epoch": 0.08, "learning_rate": 1.9893088415478454e-05, "loss": 0.2391, "step": 1506 }, { "epoch": 0.08, "learning_rate": 1.9892852439314647e-05, "loss": 0.2249, "step": 1507 }, { "epoch": 0.08, "learning_rate": 1.9892616204416244e-05, "loss": 0.2323, "step": 1508 }, { "epoch": 0.08, "learning_rate": 1.989237971078942e-05, "loss": 0.4067, "step": 1509 }, { "epoch": 0.08, "learning_rate": 1.9892142958440362e-05, "loss": 0.4472, "step": 1510 }, { "epoch": 0.08, "learning_rate": 1.989190594737526e-05, "loss": 0.4801, "step": 1511 }, { "epoch": 0.08, "learning_rate": 1.9891668677600317e-05, "loss": 0.5063, "step": 1512 }, { "epoch": 0.08, "learning_rate": 1.989143114912173e-05, "loss": 0.7788, "step": 1513 }, { "epoch": 0.08, "learning_rate": 1.989119336194572e-05, "loss": 0.3658, "step": 1514 }, { "epoch": 0.08, "learning_rate": 1.9890955316078503e-05, "loss": 0.3632, "step": 1515 }, { "epoch": 0.08, "learning_rate": 1.9890717011526305e-05, "loss": 0.2349, "step": 1516 }, { "epoch": 0.08, "learning_rate": 1.9890478448295357e-05, "loss": 0.714, "step": 1517 }, { "epoch": 0.08, "learning_rate": 1.9890239626391898e-05, "loss": 0.4413, "step": 1518 }, { "epoch": 0.08, "learning_rate": 1.9890000545822178e-05, "loss": 0.4573, "step": 1519 }, { "epoch": 0.08, "learning_rate": 1.9889761206592448e-05, "loss": 0.4397, "step": 1520 }, { "epoch": 0.08, "learning_rate": 1.9889521608708966e-05, "loss": 0.5662, "step": 1521 }, { "epoch": 0.08, "learning_rate": 1.9889281752177997e-05, "loss": 0.3389, "step": 1522 }, { "epoch": 0.08, "learning_rate": 1.988904163700582e-05, "loss": 0.3879, "step": 1523 }, { "epoch": 0.08, "learning_rate": 1.9888801263198708e-05, "loss": 0.2171, "step": 1524 }, { "epoch": 0.08, "learning_rate": 1.988856063076295e-05, "loss": 0.3092, "step": 1525 }, { "epoch": 0.08, "learning_rate": 1.9888319739704843e-05, "loss": 0.5596, "step": 1526 }, { "epoch": 0.08, "learning_rate": 1.9888078590030686e-05, "loss": 0.7022, "step": 1527 }, { "epoch": 0.08, "learning_rate": 1.9887837181746783e-05, "loss": 0.3013, "step": 1528 }, { "epoch": 0.08, "learning_rate": 1.988759551485945e-05, "loss": 0.4868, "step": 1529 }, { "epoch": 0.08, "learning_rate": 1.9887353589375005e-05, "loss": 0.3607, "step": 1530 }, { "epoch": 0.08, "learning_rate": 1.9887111405299778e-05, "loss": 0.4217, "step": 1531 }, { "epoch": 0.08, "learning_rate": 1.98868689626401e-05, "loss": 0.3891, "step": 1532 }, { "epoch": 0.08, "learning_rate": 1.9886626261402314e-05, "loss": 0.2665, "step": 1533 }, { "epoch": 0.08, "learning_rate": 1.9886383301592772e-05, "loss": 0.3141, "step": 1534 }, { "epoch": 0.08, "learning_rate": 1.9886140083217818e-05, "loss": 0.4107, "step": 1535 }, { "epoch": 0.08, "learning_rate": 1.988589660628382e-05, "loss": 0.5876, "step": 1536 }, { "epoch": 0.08, "learning_rate": 1.9885652870797148e-05, "loss": 0.4266, "step": 1537 }, { "epoch": 0.08, "learning_rate": 1.988540887676417e-05, "loss": 0.3808, "step": 1538 }, { "epoch": 0.08, "learning_rate": 1.988516462419127e-05, "loss": 0.2886, "step": 1539 }, { "epoch": 0.08, "learning_rate": 1.988492011308484e-05, "loss": 0.3093, "step": 1540 }, { "epoch": 0.08, "learning_rate": 1.9884675343451265e-05, "loss": 0.4234, "step": 1541 }, { "epoch": 0.08, "learning_rate": 1.988443031529696e-05, "loss": 0.606, "step": 1542 }, { "epoch": 0.08, "learning_rate": 1.9884185028628322e-05, "loss": 0.7057, "step": 1543 }, { "epoch": 0.08, "learning_rate": 1.988393948345177e-05, "loss": 0.2154, "step": 1544 }, { "epoch": 0.08, "learning_rate": 1.9883693679773727e-05, "loss": 0.3443, "step": 1545 }, { "epoch": 0.08, "learning_rate": 1.9883447617600625e-05, "loss": 0.4591, "step": 1546 }, { "epoch": 0.08, "learning_rate": 1.988320129693889e-05, "loss": 0.395, "step": 1547 }, { "epoch": 0.08, "learning_rate": 1.9882954717794974e-05, "loss": 0.3393, "step": 1548 }, { "epoch": 0.08, "learning_rate": 1.988270788017532e-05, "loss": 0.4507, "step": 1549 }, { "epoch": 0.08, "learning_rate": 1.9882460784086387e-05, "loss": 0.5668, "step": 1550 }, { "epoch": 0.08, "learning_rate": 1.9882213429534632e-05, "loss": 0.2267, "step": 1551 }, { "epoch": 0.08, "learning_rate": 1.9881965816526532e-05, "loss": 0.4769, "step": 1552 }, { "epoch": 0.08, "learning_rate": 1.988171794506856e-05, "loss": 0.2606, "step": 1553 }, { "epoch": 0.08, "learning_rate": 1.9881469815167193e-05, "loss": 0.583, "step": 1554 }, { "epoch": 0.08, "learning_rate": 1.988122142682893e-05, "loss": 0.3226, "step": 1555 }, { "epoch": 0.08, "learning_rate": 1.988097278006026e-05, "loss": 0.4427, "step": 1556 }, { "epoch": 0.08, "learning_rate": 1.988072387486769e-05, "loss": 0.5354, "step": 1557 }, { "epoch": 0.08, "learning_rate": 1.9880474711257726e-05, "loss": 0.3848, "step": 1558 }, { "epoch": 0.08, "learning_rate": 1.988022528923689e-05, "loss": 0.2765, "step": 1559 }, { "epoch": 0.08, "learning_rate": 1.98799756088117e-05, "loss": 0.4067, "step": 1560 }, { "epoch": 0.08, "learning_rate": 1.987972566998869e-05, "loss": 0.3991, "step": 1561 }, { "epoch": 0.08, "learning_rate": 1.9879475472774394e-05, "loss": 0.7372, "step": 1562 }, { "epoch": 0.08, "learning_rate": 1.9879225017175358e-05, "loss": 0.5334, "step": 1563 }, { "epoch": 0.08, "learning_rate": 1.987897430319813e-05, "loss": 0.2532, "step": 1564 }, { "epoch": 0.08, "learning_rate": 1.987872333084927e-05, "loss": 0.3543, "step": 1565 }, { "epoch": 0.08, "learning_rate": 1.9878472100135337e-05, "loss": 0.3722, "step": 1566 }, { "epoch": 0.08, "learning_rate": 1.9878220611062907e-05, "loss": 0.6564, "step": 1567 }, { "epoch": 0.08, "learning_rate": 1.9877968863638558e-05, "loss": 0.2361, "step": 1568 }, { "epoch": 0.08, "learning_rate": 1.9877716857868865e-05, "loss": 0.4435, "step": 1569 }, { "epoch": 0.08, "learning_rate": 1.987746459376043e-05, "loss": 0.525, "step": 1570 }, { "epoch": 0.08, "learning_rate": 1.987721207131984e-05, "loss": 0.2154, "step": 1571 }, { "epoch": 0.08, "learning_rate": 1.987695929055371e-05, "loss": 0.5532, "step": 1572 }, { "epoch": 0.08, "learning_rate": 1.9876706251468648e-05, "loss": 0.2951, "step": 1573 }, { "epoch": 0.08, "learning_rate": 1.9876452954071265e-05, "loss": 0.4084, "step": 1574 }, { "epoch": 0.08, "learning_rate": 1.9876199398368193e-05, "loss": 0.4203, "step": 1575 }, { "epoch": 0.08, "learning_rate": 1.9875945584366062e-05, "loss": 0.5843, "step": 1576 }, { "epoch": 0.08, "learning_rate": 1.9875691512071507e-05, "loss": 0.4056, "step": 1577 }, { "epoch": 0.08, "learning_rate": 1.987543718149118e-05, "loss": 0.6503, "step": 1578 }, { "epoch": 0.08, "learning_rate": 1.9875182592631724e-05, "loss": 0.2424, "step": 1579 }, { "epoch": 0.08, "learning_rate": 1.9874927745499805e-05, "loss": 0.4355, "step": 1580 }, { "epoch": 0.08, "learning_rate": 1.9874672640102078e-05, "loss": 0.4087, "step": 1581 }, { "epoch": 0.08, "learning_rate": 1.9874417276445225e-05, "loss": 0.6977, "step": 1582 }, { "epoch": 0.08, "learning_rate": 1.9874161654535925e-05, "loss": 0.4348, "step": 1583 }, { "epoch": 0.08, "learning_rate": 1.9873905774380854e-05, "loss": 0.5199, "step": 1584 }, { "epoch": 0.08, "learning_rate": 1.9873649635986712e-05, "loss": 0.4768, "step": 1585 }, { "epoch": 0.08, "learning_rate": 1.98733932393602e-05, "loss": 0.3352, "step": 1586 }, { "epoch": 0.08, "learning_rate": 1.9873136584508014e-05, "loss": 0.2342, "step": 1587 }, { "epoch": 0.08, "learning_rate": 1.9872879671436875e-05, "loss": 0.4586, "step": 1588 }, { "epoch": 0.08, "learning_rate": 1.9872622500153498e-05, "loss": 0.4967, "step": 1589 }, { "epoch": 0.08, "learning_rate": 1.9872365070664612e-05, "loss": 0.3724, "step": 1590 }, { "epoch": 0.08, "learning_rate": 1.987210738297695e-05, "loss": 0.3775, "step": 1591 }, { "epoch": 0.08, "learning_rate": 1.9871849437097243e-05, "loss": 0.5184, "step": 1592 }, { "epoch": 0.08, "learning_rate": 1.987159123303225e-05, "loss": 0.4603, "step": 1593 }, { "epoch": 0.08, "learning_rate": 1.9871332770788715e-05, "loss": 0.3068, "step": 1594 }, { "epoch": 0.08, "learning_rate": 1.98710740503734e-05, "loss": 0.6344, "step": 1595 }, { "epoch": 0.08, "learning_rate": 1.9870815071793073e-05, "loss": 0.4039, "step": 1596 }, { "epoch": 0.08, "learning_rate": 1.9870555835054505e-05, "loss": 0.4081, "step": 1597 }, { "epoch": 0.08, "learning_rate": 1.987029634016448e-05, "loss": 0.5721, "step": 1598 }, { "epoch": 0.08, "learning_rate": 1.9870036587129777e-05, "loss": 0.6357, "step": 1599 }, { "epoch": 0.08, "learning_rate": 1.9869776575957198e-05, "loss": 0.4086, "step": 1600 }, { "epoch": 0.08, "learning_rate": 1.986951630665354e-05, "loss": 0.5652, "step": 1601 }, { "epoch": 0.08, "learning_rate": 1.986925577922561e-05, "loss": 0.2624, "step": 1602 }, { "epoch": 0.08, "learning_rate": 1.9868994993680218e-05, "loss": 0.5061, "step": 1603 }, { "epoch": 0.08, "learning_rate": 1.9868733950024194e-05, "loss": 0.4197, "step": 1604 }, { "epoch": 0.08, "learning_rate": 1.9868472648264355e-05, "loss": 0.5624, "step": 1605 }, { "epoch": 0.08, "learning_rate": 1.986821108840754e-05, "loss": 0.4658, "step": 1606 }, { "epoch": 0.08, "learning_rate": 1.986794927046059e-05, "loss": 0.5339, "step": 1607 }, { "epoch": 0.08, "learning_rate": 1.986768719443035e-05, "loss": 0.5625, "step": 1608 }, { "epoch": 0.08, "learning_rate": 1.9867424860323675e-05, "loss": 0.5686, "step": 1609 }, { "epoch": 0.08, "learning_rate": 1.986716226814743e-05, "loss": 0.6415, "step": 1610 }, { "epoch": 0.08, "learning_rate": 1.9866899417908477e-05, "loss": 0.4479, "step": 1611 }, { "epoch": 0.08, "learning_rate": 1.9866636309613692e-05, "loss": 0.2554, "step": 1612 }, { "epoch": 0.08, "learning_rate": 1.986637294326996e-05, "loss": 0.4141, "step": 1613 }, { "epoch": 0.08, "learning_rate": 1.9866109318884165e-05, "loss": 0.5515, "step": 1614 }, { "epoch": 0.08, "learning_rate": 1.9865845436463205e-05, "loss": 0.4358, "step": 1615 }, { "epoch": 0.08, "learning_rate": 1.986558129601398e-05, "loss": 0.6794, "step": 1616 }, { "epoch": 0.08, "learning_rate": 1.9865316897543394e-05, "loss": 0.4473, "step": 1617 }, { "epoch": 0.08, "learning_rate": 1.986505224105837e-05, "loss": 0.3755, "step": 1618 }, { "epoch": 0.08, "learning_rate": 1.9864787326565822e-05, "loss": 0.5969, "step": 1619 }, { "epoch": 0.08, "learning_rate": 1.986452215407268e-05, "loss": 0.2932, "step": 1620 }, { "epoch": 0.08, "learning_rate": 1.9864256723585886e-05, "loss": 0.5075, "step": 1621 }, { "epoch": 0.08, "learning_rate": 1.9863991035112376e-05, "loss": 0.4984, "step": 1622 }, { "epoch": 0.08, "learning_rate": 1.9863725088659097e-05, "loss": 0.2789, "step": 1623 }, { "epoch": 0.08, "learning_rate": 1.9863458884233014e-05, "loss": 0.3286, "step": 1624 }, { "epoch": 0.08, "learning_rate": 1.9863192421841078e-05, "loss": 0.3266, "step": 1625 }, { "epoch": 0.08, "learning_rate": 1.9862925701490262e-05, "loss": 0.4765, "step": 1626 }, { "epoch": 0.08, "learning_rate": 1.9862658723187546e-05, "loss": 0.4918, "step": 1627 }, { "epoch": 0.08, "learning_rate": 1.9862391486939905e-05, "loss": 0.4751, "step": 1628 }, { "epoch": 0.08, "learning_rate": 1.9862123992754334e-05, "loss": 0.3064, "step": 1629 }, { "epoch": 0.08, "learning_rate": 1.9861856240637825e-05, "loss": 0.6052, "step": 1630 }, { "epoch": 0.08, "learning_rate": 1.9861588230597385e-05, "loss": 0.434, "step": 1631 }, { "epoch": 0.08, "learning_rate": 1.9861319962640023e-05, "loss": 0.2585, "step": 1632 }, { "epoch": 0.08, "learning_rate": 1.986105143677275e-05, "loss": 0.373, "step": 1633 }, { "epoch": 0.08, "learning_rate": 1.9860782653002594e-05, "loss": 0.4852, "step": 1634 }, { "epoch": 0.08, "learning_rate": 1.9860513611336583e-05, "loss": 0.4984, "step": 1635 }, { "epoch": 0.08, "learning_rate": 1.9860244311781752e-05, "loss": 0.3729, "step": 1636 }, { "epoch": 0.08, "learning_rate": 1.985997475434515e-05, "loss": 0.3557, "step": 1637 }, { "epoch": 0.08, "learning_rate": 1.9859704939033814e-05, "loss": 0.3689, "step": 1638 }, { "epoch": 0.08, "learning_rate": 1.9859434865854817e-05, "loss": 0.2655, "step": 1639 }, { "epoch": 0.08, "learning_rate": 1.985916453481521e-05, "loss": 0.5276, "step": 1640 }, { "epoch": 0.08, "learning_rate": 1.985889394592207e-05, "loss": 0.37, "step": 1641 }, { "epoch": 0.08, "learning_rate": 1.9858623099182472e-05, "loss": 0.3327, "step": 1642 }, { "epoch": 0.08, "learning_rate": 1.98583519946035e-05, "loss": 0.4284, "step": 1643 }, { "epoch": 0.08, "learning_rate": 1.9858080632192243e-05, "loss": 0.5086, "step": 1644 }, { "epoch": 0.08, "learning_rate": 1.98578090119558e-05, "loss": 0.3706, "step": 1645 }, { "epoch": 0.08, "learning_rate": 1.985753713390127e-05, "loss": 0.4007, "step": 1646 }, { "epoch": 0.08, "learning_rate": 1.985726499803577e-05, "loss": 0.2979, "step": 1647 }, { "epoch": 0.08, "learning_rate": 1.985699260436641e-05, "loss": 0.2813, "step": 1648 }, { "epoch": 0.08, "learning_rate": 1.9856719952900324e-05, "loss": 0.4799, "step": 1649 }, { "epoch": 0.08, "learning_rate": 1.9856447043644637e-05, "loss": 0.4033, "step": 1650 }, { "epoch": 0.08, "learning_rate": 1.9856173876606487e-05, "loss": 0.4732, "step": 1651 }, { "epoch": 0.08, "learning_rate": 1.9855900451793016e-05, "loss": 0.2049, "step": 1652 }, { "epoch": 0.08, "learning_rate": 1.9855626769211383e-05, "loss": 0.4261, "step": 1653 }, { "epoch": 0.08, "learning_rate": 1.9855352828868734e-05, "loss": 0.2633, "step": 1654 }, { "epoch": 0.08, "learning_rate": 1.9855078630772245e-05, "loss": 0.3487, "step": 1655 }, { "epoch": 0.08, "learning_rate": 1.9854804174929078e-05, "loss": 0.2488, "step": 1656 }, { "epoch": 0.08, "learning_rate": 1.985452946134642e-05, "loss": 0.4542, "step": 1657 }, { "epoch": 0.08, "learning_rate": 1.985425449003145e-05, "loss": 0.5452, "step": 1658 }, { "epoch": 0.08, "learning_rate": 1.9853979260991355e-05, "loss": 0.4009, "step": 1659 }, { "epoch": 0.08, "learning_rate": 1.9853703774233343e-05, "loss": 0.5777, "step": 1660 }, { "epoch": 0.08, "learning_rate": 1.9853428029764618e-05, "loss": 0.6303, "step": 1661 }, { "epoch": 0.08, "learning_rate": 1.9853152027592383e-05, "loss": 0.5612, "step": 1662 }, { "epoch": 0.08, "learning_rate": 1.9852875767723866e-05, "loss": 0.2369, "step": 1663 }, { "epoch": 0.08, "learning_rate": 1.9852599250166286e-05, "loss": 0.5554, "step": 1664 }, { "epoch": 0.08, "learning_rate": 1.9852322474926877e-05, "loss": 0.3231, "step": 1665 }, { "epoch": 0.08, "learning_rate": 1.9852045442012877e-05, "loss": 0.568, "step": 1666 }, { "epoch": 0.08, "learning_rate": 1.9851768151431535e-05, "loss": 0.284, "step": 1667 }, { "epoch": 0.08, "learning_rate": 1.98514906031901e-05, "loss": 0.2101, "step": 1668 }, { "epoch": 0.08, "learning_rate": 1.985121279729583e-05, "loss": 0.2911, "step": 1669 }, { "epoch": 0.08, "learning_rate": 1.9850934733755992e-05, "loss": 0.4846, "step": 1670 }, { "epoch": 0.08, "learning_rate": 1.9850656412577862e-05, "loss": 0.2916, "step": 1671 }, { "epoch": 0.08, "learning_rate": 1.9850377833768712e-05, "loss": 0.4499, "step": 1672 }, { "epoch": 0.08, "learning_rate": 1.9850098997335833e-05, "loss": 0.4153, "step": 1673 }, { "epoch": 0.08, "learning_rate": 1.9849819903286516e-05, "loss": 0.3071, "step": 1674 }, { "epoch": 0.08, "learning_rate": 1.984954055162806e-05, "loss": 0.3647, "step": 1675 }, { "epoch": 0.08, "learning_rate": 1.984926094236777e-05, "loss": 0.2479, "step": 1676 }, { "epoch": 0.08, "learning_rate": 1.9848981075512964e-05, "loss": 0.6195, "step": 1677 }, { "epoch": 0.08, "learning_rate": 1.984870095107095e-05, "loss": 0.582, "step": 1678 }, { "epoch": 0.08, "learning_rate": 1.9848420569049073e-05, "loss": 0.3789, "step": 1679 }, { "epoch": 0.08, "learning_rate": 1.9848139929454648e-05, "loss": 0.5349, "step": 1680 }, { "epoch": 0.08, "learning_rate": 1.9847859032295023e-05, "loss": 0.4102, "step": 1681 }, { "epoch": 0.08, "learning_rate": 1.9847577877577545e-05, "loss": 0.451, "step": 1682 }, { "epoch": 0.08, "learning_rate": 1.984729646530956e-05, "loss": 0.4087, "step": 1683 }, { "epoch": 0.08, "learning_rate": 1.9847014795498443e-05, "loss": 0.5784, "step": 1684 }, { "epoch": 0.08, "learning_rate": 1.9846732868151544e-05, "loss": 0.386, "step": 1685 }, { "epoch": 0.08, "learning_rate": 1.9846450683276248e-05, "loss": 0.3617, "step": 1686 }, { "epoch": 0.08, "learning_rate": 1.984616824087993e-05, "loss": 0.3529, "step": 1687 }, { "epoch": 0.08, "learning_rate": 1.9845885540969974e-05, "loss": 0.4306, "step": 1688 }, { "epoch": 0.08, "learning_rate": 1.984560258355378e-05, "loss": 0.4601, "step": 1689 }, { "epoch": 0.08, "learning_rate": 1.9845319368638747e-05, "loss": 0.322, "step": 1690 }, { "epoch": 0.08, "learning_rate": 1.984503589623228e-05, "loss": 0.2735, "step": 1691 }, { "epoch": 0.08, "learning_rate": 1.9844752166341793e-05, "loss": 0.1655, "step": 1692 }, { "epoch": 0.08, "learning_rate": 1.984446817897471e-05, "loss": 0.3457, "step": 1693 }, { "epoch": 0.08, "learning_rate": 1.9844183934138454e-05, "loss": 0.2905, "step": 1694 }, { "epoch": 0.08, "learning_rate": 1.984389943184046e-05, "loss": 0.672, "step": 1695 }, { "epoch": 0.08, "learning_rate": 1.984361467208817e-05, "loss": 0.3236, "step": 1696 }, { "epoch": 0.08, "learning_rate": 1.9843329654889032e-05, "loss": 0.2266, "step": 1697 }, { "epoch": 0.08, "learning_rate": 1.98430443802505e-05, "loss": 0.2912, "step": 1698 }, { "epoch": 0.08, "learning_rate": 1.9842758848180034e-05, "loss": 0.2688, "step": 1699 }, { "epoch": 0.08, "learning_rate": 1.98424730586851e-05, "loss": 0.1665, "step": 1700 }, { "epoch": 0.08, "learning_rate": 1.9842187011773174e-05, "loss": 0.5243, "step": 1701 }, { "epoch": 0.08, "learning_rate": 1.984190070745174e-05, "loss": 0.3063, "step": 1702 }, { "epoch": 0.09, "learning_rate": 1.9841614145728283e-05, "loss": 0.2472, "step": 1703 }, { "epoch": 0.09, "learning_rate": 1.9841327326610297e-05, "loss": 0.4486, "step": 1704 }, { "epoch": 0.09, "learning_rate": 1.9841040250105287e-05, "loss": 0.4858, "step": 1705 }, { "epoch": 0.09, "learning_rate": 1.9840752916220757e-05, "loss": 0.513, "step": 1706 }, { "epoch": 0.09, "learning_rate": 1.9840465324964226e-05, "loss": 0.5135, "step": 1707 }, { "epoch": 0.09, "learning_rate": 1.984017747634321e-05, "loss": 0.3585, "step": 1708 }, { "epoch": 0.09, "learning_rate": 1.983988937036524e-05, "loss": 0.2825, "step": 1709 }, { "epoch": 0.09, "learning_rate": 1.9839601007037853e-05, "loss": 0.4869, "step": 1710 }, { "epoch": 0.09, "learning_rate": 1.983931238636859e-05, "loss": 0.4193, "step": 1711 }, { "epoch": 0.09, "learning_rate": 1.9839023508364998e-05, "loss": 0.3961, "step": 1712 }, { "epoch": 0.09, "learning_rate": 1.983873437303463e-05, "loss": 0.629, "step": 1713 }, { "epoch": 0.09, "learning_rate": 1.9838444980385058e-05, "loss": 0.2722, "step": 1714 }, { "epoch": 0.09, "learning_rate": 1.9838155330423835e-05, "loss": 0.3973, "step": 1715 }, { "epoch": 0.09, "learning_rate": 1.983786542315855e-05, "loss": 0.4982, "step": 1716 }, { "epoch": 0.09, "learning_rate": 1.983757525859678e-05, "loss": 0.3089, "step": 1717 }, { "epoch": 0.09, "learning_rate": 1.9837284836746115e-05, "loss": 0.353, "step": 1718 }, { "epoch": 0.09, "learning_rate": 1.9836994157614146e-05, "loss": 0.5278, "step": 1719 }, { "epoch": 0.09, "learning_rate": 1.9836703221208483e-05, "loss": 0.4365, "step": 1720 }, { "epoch": 0.09, "learning_rate": 1.9836412027536728e-05, "loss": 0.509, "step": 1721 }, { "epoch": 0.09, "learning_rate": 1.98361205766065e-05, "loss": 0.536, "step": 1722 }, { "epoch": 0.09, "learning_rate": 1.9835828868425426e-05, "loss": 0.4716, "step": 1723 }, { "epoch": 0.09, "learning_rate": 1.9835536903001126e-05, "loss": 0.5222, "step": 1724 }, { "epoch": 0.09, "learning_rate": 1.983524468034124e-05, "loss": 0.2122, "step": 1725 }, { "epoch": 0.09, "learning_rate": 1.983495220045341e-05, "loss": 0.4304, "step": 1726 }, { "epoch": 0.09, "learning_rate": 1.983465946334529e-05, "loss": 0.3627, "step": 1727 }, { "epoch": 0.09, "learning_rate": 1.9834366469024533e-05, "loss": 0.4055, "step": 1728 }, { "epoch": 0.09, "learning_rate": 1.98340732174988e-05, "loss": 0.4028, "step": 1729 }, { "epoch": 0.09, "learning_rate": 1.9833779708775763e-05, "loss": 0.2717, "step": 1730 }, { "epoch": 0.09, "learning_rate": 1.98334859428631e-05, "loss": 0.3911, "step": 1731 }, { "epoch": 0.09, "learning_rate": 1.9833191919768487e-05, "loss": 0.245, "step": 1732 }, { "epoch": 0.09, "learning_rate": 1.9832897639499622e-05, "loss": 0.3599, "step": 1733 }, { "epoch": 0.09, "learning_rate": 1.9832603102064192e-05, "loss": 0.3344, "step": 1734 }, { "epoch": 0.09, "learning_rate": 1.9832308307469916e-05, "loss": 0.4684, "step": 1735 }, { "epoch": 0.09, "learning_rate": 1.983201325572449e-05, "loss": 0.4631, "step": 1736 }, { "epoch": 0.09, "learning_rate": 1.9831717946835633e-05, "loss": 0.3353, "step": 1737 }, { "epoch": 0.09, "learning_rate": 1.983142238081107e-05, "loss": 0.595, "step": 1738 }, { "epoch": 0.09, "learning_rate": 1.9831126557658535e-05, "loss": 0.6173, "step": 1739 }, { "epoch": 0.09, "learning_rate": 1.983083047738576e-05, "loss": 0.4116, "step": 1740 }, { "epoch": 0.09, "learning_rate": 1.9830534140000488e-05, "loss": 0.4727, "step": 1741 }, { "epoch": 0.09, "learning_rate": 1.9830237545510475e-05, "loss": 0.5239, "step": 1742 }, { "epoch": 0.09, "learning_rate": 1.982994069392347e-05, "loss": 0.2125, "step": 1743 }, { "epoch": 0.09, "learning_rate": 1.9829643585247242e-05, "loss": 0.7378, "step": 1744 }, { "epoch": 0.09, "learning_rate": 1.9829346219489563e-05, "loss": 0.3322, "step": 1745 }, { "epoch": 0.09, "learning_rate": 1.9829048596658205e-05, "loss": 0.3919, "step": 1746 }, { "epoch": 0.09, "learning_rate": 1.9828750716760956e-05, "loss": 0.6468, "step": 1747 }, { "epoch": 0.09, "learning_rate": 1.9828452579805606e-05, "loss": 0.3389, "step": 1748 }, { "epoch": 0.09, "learning_rate": 1.9828154185799952e-05, "loss": 0.2971, "step": 1749 }, { "epoch": 0.09, "learning_rate": 1.9827855534751795e-05, "loss": 0.4677, "step": 1750 }, { "epoch": 0.09, "learning_rate": 1.982755662666895e-05, "loss": 0.4309, "step": 1751 }, { "epoch": 0.09, "learning_rate": 1.9827257461559235e-05, "loss": 0.3035, "step": 1752 }, { "epoch": 0.09, "learning_rate": 1.982695803943047e-05, "loss": 0.5668, "step": 1753 }, { "epoch": 0.09, "learning_rate": 1.982665836029049e-05, "loss": 0.4448, "step": 1754 }, { "epoch": 0.09, "learning_rate": 1.9826358424147133e-05, "loss": 0.5168, "step": 1755 }, { "epoch": 0.09, "learning_rate": 1.982605823100824e-05, "loss": 0.4548, "step": 1756 }, { "epoch": 0.09, "learning_rate": 1.982575778088166e-05, "loss": 0.3874, "step": 1757 }, { "epoch": 0.09, "learning_rate": 1.982545707377526e-05, "loss": 0.5124, "step": 1758 }, { "epoch": 0.09, "learning_rate": 1.98251561096969e-05, "loss": 0.329, "step": 1759 }, { "epoch": 0.09, "learning_rate": 1.9824854888654447e-05, "loss": 0.4065, "step": 1760 }, { "epoch": 0.09, "learning_rate": 1.9824553410655784e-05, "loss": 0.3692, "step": 1761 }, { "epoch": 0.09, "learning_rate": 1.9824251675708792e-05, "loss": 0.4827, "step": 1762 }, { "epoch": 0.09, "learning_rate": 1.9823949683821366e-05, "loss": 0.4521, "step": 1763 }, { "epoch": 0.09, "learning_rate": 1.9823647435001407e-05, "loss": 0.5081, "step": 1764 }, { "epoch": 0.09, "learning_rate": 1.982334492925681e-05, "loss": 0.3906, "step": 1765 }, { "epoch": 0.09, "learning_rate": 1.9823042166595495e-05, "loss": 0.4185, "step": 1766 }, { "epoch": 0.09, "learning_rate": 1.982273914702538e-05, "loss": 0.2881, "step": 1767 }, { "epoch": 0.09, "learning_rate": 1.9822435870554386e-05, "loss": 0.3318, "step": 1768 }, { "epoch": 0.09, "learning_rate": 1.982213233719045e-05, "loss": 0.3749, "step": 1769 }, { "epoch": 0.09, "learning_rate": 1.9821828546941504e-05, "loss": 0.6555, "step": 1770 }, { "epoch": 0.09, "learning_rate": 1.9821524499815496e-05, "loss": 0.4534, "step": 1771 }, { "epoch": 0.09, "learning_rate": 1.982122019582038e-05, "loss": 0.3606, "step": 1772 }, { "epoch": 0.09, "learning_rate": 1.9820915634964117e-05, "loss": 0.4304, "step": 1773 }, { "epoch": 0.09, "learning_rate": 1.9820610817254666e-05, "loss": 0.3984, "step": 1774 }, { "epoch": 0.09, "learning_rate": 1.9820305742699997e-05, "loss": 0.3081, "step": 1775 }, { "epoch": 0.09, "learning_rate": 1.9820000411308096e-05, "loss": 0.4201, "step": 1776 }, { "epoch": 0.09, "learning_rate": 1.981969482308695e-05, "loss": 0.4871, "step": 1777 }, { "epoch": 0.09, "learning_rate": 1.9819388978044545e-05, "loss": 0.2626, "step": 1778 }, { "epoch": 0.09, "learning_rate": 1.981908287618888e-05, "loss": 0.253, "step": 1779 }, { "epoch": 0.09, "learning_rate": 1.9818776517527968e-05, "loss": 0.4793, "step": 1780 }, { "epoch": 0.09, "learning_rate": 1.9818469902069816e-05, "loss": 0.5025, "step": 1781 }, { "epoch": 0.09, "learning_rate": 1.981816302982244e-05, "loss": 0.4377, "step": 1782 }, { "epoch": 0.09, "learning_rate": 1.981785590079387e-05, "loss": 0.5533, "step": 1783 }, { "epoch": 0.09, "learning_rate": 1.9817548514992138e-05, "loss": 0.2503, "step": 1784 }, { "epoch": 0.09, "learning_rate": 1.9817240872425284e-05, "loss": 0.4172, "step": 1785 }, { "epoch": 0.09, "learning_rate": 1.9816932973101354e-05, "loss": 0.376, "step": 1786 }, { "epoch": 0.09, "learning_rate": 1.98166248170284e-05, "loss": 0.2623, "step": 1787 }, { "epoch": 0.09, "learning_rate": 1.9816316404214482e-05, "loss": 0.3123, "step": 1788 }, { "epoch": 0.09, "learning_rate": 1.981600773466766e-05, "loss": 0.4628, "step": 1789 }, { "epoch": 0.09, "learning_rate": 1.981569880839602e-05, "loss": 0.3218, "step": 1790 }, { "epoch": 0.09, "learning_rate": 1.981538962540763e-05, "loss": 0.465, "step": 1791 }, { "epoch": 0.09, "learning_rate": 1.981508018571058e-05, "loss": 0.2862, "step": 1792 }, { "epoch": 0.09, "learning_rate": 1.9814770489312963e-05, "loss": 0.2891, "step": 1793 }, { "epoch": 0.09, "learning_rate": 1.9814460536222883e-05, "loss": 0.2558, "step": 1794 }, { "epoch": 0.09, "learning_rate": 1.9814150326448437e-05, "loss": 0.6753, "step": 1795 }, { "epoch": 0.09, "learning_rate": 1.9813839859997745e-05, "loss": 0.478, "step": 1796 }, { "epoch": 0.09, "learning_rate": 1.9813529136878927e-05, "loss": 0.3596, "step": 1797 }, { "epoch": 0.09, "learning_rate": 1.9813218157100107e-05, "loss": 0.3429, "step": 1798 }, { "epoch": 0.09, "learning_rate": 1.981290692066942e-05, "loss": 0.3639, "step": 1799 }, { "epoch": 0.09, "learning_rate": 1.9812595427595002e-05, "loss": 0.4043, "step": 1800 }, { "epoch": 0.09, "learning_rate": 1.9812283677885005e-05, "loss": 0.5943, "step": 1801 }, { "epoch": 0.09, "learning_rate": 1.9811971671547582e-05, "loss": 0.4161, "step": 1802 }, { "epoch": 0.09, "learning_rate": 1.981165940859089e-05, "loss": 0.2907, "step": 1803 }, { "epoch": 0.09, "learning_rate": 1.9811346889023095e-05, "loss": 0.3711, "step": 1804 }, { "epoch": 0.09, "learning_rate": 1.981103411285237e-05, "loss": 0.5168, "step": 1805 }, { "epoch": 0.09, "learning_rate": 1.9810721080086905e-05, "loss": 0.6115, "step": 1806 }, { "epoch": 0.09, "learning_rate": 1.9810407790734877e-05, "loss": 0.3937, "step": 1807 }, { "epoch": 0.09, "learning_rate": 1.9810094244804484e-05, "loss": 0.3806, "step": 1808 }, { "epoch": 0.09, "learning_rate": 1.9809780442303922e-05, "loss": 0.4229, "step": 1809 }, { "epoch": 0.09, "learning_rate": 1.9809466383241403e-05, "loss": 0.4171, "step": 1810 }, { "epoch": 0.09, "learning_rate": 1.9809152067625137e-05, "loss": 0.5046, "step": 1811 }, { "epoch": 0.09, "learning_rate": 1.9808837495463347e-05, "loss": 0.2873, "step": 1812 }, { "epoch": 0.09, "learning_rate": 1.980852266676426e-05, "loss": 0.3204, "step": 1813 }, { "epoch": 0.09, "learning_rate": 1.980820758153611e-05, "loss": 0.6286, "step": 1814 }, { "epoch": 0.09, "learning_rate": 1.9807892239787135e-05, "loss": 0.6027, "step": 1815 }, { "epoch": 0.09, "learning_rate": 1.980757664152559e-05, "loss": 0.3073, "step": 1816 }, { "epoch": 0.09, "learning_rate": 1.980726078675972e-05, "loss": 0.5969, "step": 1817 }, { "epoch": 0.09, "learning_rate": 1.9806944675497786e-05, "loss": 0.3974, "step": 1818 }, { "epoch": 0.09, "learning_rate": 1.9806628307748064e-05, "loss": 1.1223, "step": 1819 }, { "epoch": 0.09, "learning_rate": 1.980631168351882e-05, "loss": 0.5713, "step": 1820 }, { "epoch": 0.09, "learning_rate": 1.980599480281834e-05, "loss": 0.332, "step": 1821 }, { "epoch": 0.09, "learning_rate": 1.980567766565491e-05, "loss": 0.3058, "step": 1822 }, { "epoch": 0.09, "learning_rate": 1.980536027203682e-05, "loss": 0.2405, "step": 1823 }, { "epoch": 0.09, "learning_rate": 1.9805042621972376e-05, "loss": 0.3978, "step": 1824 }, { "epoch": 0.09, "learning_rate": 1.9804724715469885e-05, "loss": 0.4665, "step": 1825 }, { "epoch": 0.09, "learning_rate": 1.9804406552537663e-05, "loss": 0.484, "step": 1826 }, { "epoch": 0.09, "learning_rate": 1.9804088133184028e-05, "loss": 0.4526, "step": 1827 }, { "epoch": 0.09, "learning_rate": 1.9803769457417304e-05, "loss": 0.3439, "step": 1828 }, { "epoch": 0.09, "learning_rate": 1.9803450525245836e-05, "loss": 0.4873, "step": 1829 }, { "epoch": 0.09, "learning_rate": 1.980313133667796e-05, "loss": 0.597, "step": 1830 }, { "epoch": 0.09, "learning_rate": 1.9802811891722022e-05, "loss": 0.3467, "step": 1831 }, { "epoch": 0.09, "learning_rate": 1.980249219038638e-05, "loss": 0.3761, "step": 1832 }, { "epoch": 0.09, "learning_rate": 1.980217223267939e-05, "loss": 0.2847, "step": 1833 }, { "epoch": 0.09, "learning_rate": 1.980185201860943e-05, "loss": 0.2204, "step": 1834 }, { "epoch": 0.09, "learning_rate": 1.980153154818486e-05, "loss": 0.2582, "step": 1835 }, { "epoch": 0.09, "learning_rate": 1.9801210821414077e-05, "loss": 0.5582, "step": 1836 }, { "epoch": 0.09, "learning_rate": 1.980088983830546e-05, "loss": 0.5864, "step": 1837 }, { "epoch": 0.09, "learning_rate": 1.9800568598867405e-05, "loss": 0.4038, "step": 1838 }, { "epoch": 0.09, "learning_rate": 1.980024710310832e-05, "loss": 0.3134, "step": 1839 }, { "epoch": 0.09, "learning_rate": 1.97999253510366e-05, "loss": 0.3614, "step": 1840 }, { "epoch": 0.09, "learning_rate": 1.979960334266067e-05, "loss": 0.4254, "step": 1841 }, { "epoch": 0.09, "learning_rate": 1.9799281077988956e-05, "loss": 0.4515, "step": 1842 }, { "epoch": 0.09, "learning_rate": 1.9798958557029876e-05, "loss": 0.3021, "step": 1843 }, { "epoch": 0.09, "learning_rate": 1.979863577979187e-05, "loss": 0.2674, "step": 1844 }, { "epoch": 0.09, "learning_rate": 1.9798312746283375e-05, "loss": 0.2662, "step": 1845 }, { "epoch": 0.09, "learning_rate": 1.9797989456512848e-05, "loss": 0.4697, "step": 1846 }, { "epoch": 0.09, "learning_rate": 1.979766591048874e-05, "loss": 0.4366, "step": 1847 }, { "epoch": 0.09, "learning_rate": 1.9797342108219507e-05, "loss": 0.5365, "step": 1848 }, { "epoch": 0.09, "learning_rate": 1.979701804971363e-05, "loss": 0.4071, "step": 1849 }, { "epoch": 0.09, "learning_rate": 1.9796693734979576e-05, "loss": 0.4604, "step": 1850 }, { "epoch": 0.09, "learning_rate": 1.9796369164025827e-05, "loss": 0.3347, "step": 1851 }, { "epoch": 0.09, "learning_rate": 1.9796044336860875e-05, "loss": 0.4737, "step": 1852 }, { "epoch": 0.09, "learning_rate": 1.9795719253493213e-05, "loss": 0.2582, "step": 1853 }, { "epoch": 0.09, "learning_rate": 1.9795393913931343e-05, "loss": 0.3743, "step": 1854 }, { "epoch": 0.09, "learning_rate": 1.9795068318183778e-05, "loss": 0.4488, "step": 1855 }, { "epoch": 0.09, "learning_rate": 1.979474246625903e-05, "loss": 0.2885, "step": 1856 }, { "epoch": 0.09, "learning_rate": 1.979441635816562e-05, "loss": 0.4734, "step": 1857 }, { "epoch": 0.09, "learning_rate": 1.979408999391208e-05, "loss": 0.3354, "step": 1858 }, { "epoch": 0.09, "learning_rate": 1.9793763373506943e-05, "loss": 0.4916, "step": 1859 }, { "epoch": 0.09, "learning_rate": 1.979343649695875e-05, "loss": 0.3885, "step": 1860 }, { "epoch": 0.09, "learning_rate": 1.9793109364276056e-05, "loss": 0.4498, "step": 1861 }, { "epoch": 0.09, "learning_rate": 1.979278197546741e-05, "loss": 0.3872, "step": 1862 }, { "epoch": 0.09, "learning_rate": 1.9792454330541382e-05, "loss": 0.432, "step": 1863 }, { "epoch": 0.09, "learning_rate": 1.9792126429506536e-05, "loss": 0.4896, "step": 1864 }, { "epoch": 0.09, "learning_rate": 1.9791798272371444e-05, "loss": 0.4047, "step": 1865 }, { "epoch": 0.09, "learning_rate": 1.97914698591447e-05, "loss": 0.4131, "step": 1866 }, { "epoch": 0.09, "learning_rate": 1.9791141189834883e-05, "loss": 0.4434, "step": 1867 }, { "epoch": 0.09, "learning_rate": 1.979081226445059e-05, "loss": 0.4053, "step": 1868 }, { "epoch": 0.09, "learning_rate": 1.979048308300043e-05, "loss": 0.591, "step": 1869 }, { "epoch": 0.09, "learning_rate": 1.9790153645493008e-05, "loss": 0.4851, "step": 1870 }, { "epoch": 0.09, "learning_rate": 1.978982395193694e-05, "loss": 0.3731, "step": 1871 }, { "epoch": 0.09, "learning_rate": 1.9789494002340844e-05, "loss": 0.3278, "step": 1872 }, { "epoch": 0.09, "learning_rate": 1.9789163796713358e-05, "loss": 0.23, "step": 1873 }, { "epoch": 0.09, "learning_rate": 1.9788833335063113e-05, "loss": 0.3864, "step": 1874 }, { "epoch": 0.09, "learning_rate": 1.9788502617398752e-05, "loss": 0.4924, "step": 1875 }, { "epoch": 0.09, "learning_rate": 1.978817164372893e-05, "loss": 0.4955, "step": 1876 }, { "epoch": 0.09, "learning_rate": 1.9787840414062295e-05, "loss": 0.3892, "step": 1877 }, { "epoch": 0.09, "learning_rate": 1.9787508928407516e-05, "loss": 0.5182, "step": 1878 }, { "epoch": 0.09, "learning_rate": 1.978717718677326e-05, "loss": 0.4046, "step": 1879 }, { "epoch": 0.09, "learning_rate": 1.9786845189168205e-05, "loss": 0.3324, "step": 1880 }, { "epoch": 0.09, "learning_rate": 1.978651293560103e-05, "loss": 0.3725, "step": 1881 }, { "epoch": 0.09, "learning_rate": 1.9786180426080428e-05, "loss": 0.5547, "step": 1882 }, { "epoch": 0.09, "learning_rate": 1.9785847660615095e-05, "loss": 0.4817, "step": 1883 }, { "epoch": 0.09, "learning_rate": 1.9785514639213733e-05, "loss": 0.4462, "step": 1884 }, { "epoch": 0.09, "learning_rate": 1.978518136188505e-05, "loss": 0.7382, "step": 1885 }, { "epoch": 0.09, "learning_rate": 1.978484782863777e-05, "loss": 0.3444, "step": 1886 }, { "epoch": 0.09, "learning_rate": 1.9784514039480607e-05, "loss": 0.449, "step": 1887 }, { "epoch": 0.09, "learning_rate": 1.9784179994422296e-05, "loss": 0.358, "step": 1888 }, { "epoch": 0.09, "learning_rate": 1.9783845693471574e-05, "loss": 0.5479, "step": 1889 }, { "epoch": 0.09, "learning_rate": 1.978351113663718e-05, "loss": 0.2057, "step": 1890 }, { "epoch": 0.09, "learning_rate": 1.9783176323927867e-05, "loss": 0.5155, "step": 1891 }, { "epoch": 0.09, "learning_rate": 1.9782841255352392e-05, "loss": 0.3006, "step": 1892 }, { "epoch": 0.09, "learning_rate": 1.9782505930919512e-05, "loss": 0.3322, "step": 1893 }, { "epoch": 0.09, "learning_rate": 1.978217035063801e-05, "loss": 0.3193, "step": 1894 }, { "epoch": 0.09, "learning_rate": 1.978183451451665e-05, "loss": 0.4016, "step": 1895 }, { "epoch": 0.09, "learning_rate": 1.978149842256422e-05, "loss": 0.3812, "step": 1896 }, { "epoch": 0.09, "learning_rate": 1.978116207478951e-05, "loss": 0.6432, "step": 1897 }, { "epoch": 0.09, "learning_rate": 1.9780825471201322e-05, "loss": 0.3534, "step": 1898 }, { "epoch": 0.09, "learning_rate": 1.9780488611808447e-05, "loss": 0.4747, "step": 1899 }, { "epoch": 0.09, "learning_rate": 1.978015149661971e-05, "loss": 0.3507, "step": 1900 }, { "epoch": 0.09, "learning_rate": 1.9779814125643916e-05, "loss": 0.2771, "step": 1901 }, { "epoch": 0.09, "learning_rate": 1.977947649888989e-05, "loss": 0.4494, "step": 1902 }, { "epoch": 0.1, "learning_rate": 1.977913861636647e-05, "loss": 0.244, "step": 1903 }, { "epoch": 0.1, "learning_rate": 1.9778800478082486e-05, "loss": 0.4481, "step": 1904 }, { "epoch": 0.1, "learning_rate": 1.9778462084046783e-05, "loss": 0.4534, "step": 1905 }, { "epoch": 0.1, "learning_rate": 1.9778123434268214e-05, "loss": 0.4979, "step": 1906 }, { "epoch": 0.1, "learning_rate": 1.977778452875563e-05, "loss": 0.2481, "step": 1907 }, { "epoch": 0.1, "learning_rate": 1.9777445367517904e-05, "loss": 0.4345, "step": 1908 }, { "epoch": 0.1, "learning_rate": 1.977710595056389e-05, "loss": 0.1908, "step": 1909 }, { "epoch": 0.1, "learning_rate": 1.9776766277902484e-05, "loss": 0.265, "step": 1910 }, { "epoch": 0.1, "learning_rate": 1.9776426349542563e-05, "loss": 0.3395, "step": 1911 }, { "epoch": 0.1, "learning_rate": 1.9776086165493013e-05, "loss": 0.5102, "step": 1912 }, { "epoch": 0.1, "learning_rate": 1.977574572576273e-05, "loss": 0.34, "step": 1913 }, { "epoch": 0.1, "learning_rate": 1.9775405030360626e-05, "loss": 0.3743, "step": 1914 }, { "epoch": 0.1, "learning_rate": 1.9775064079295604e-05, "loss": 0.1775, "step": 1915 }, { "epoch": 0.1, "learning_rate": 1.9774722872576584e-05, "loss": 0.4987, "step": 1916 }, { "epoch": 0.1, "learning_rate": 1.9774381410212493e-05, "loss": 0.2489, "step": 1917 }, { "epoch": 0.1, "learning_rate": 1.9774039692212256e-05, "loss": 0.6204, "step": 1918 }, { "epoch": 0.1, "learning_rate": 1.977369771858481e-05, "loss": 0.4818, "step": 1919 }, { "epoch": 0.1, "learning_rate": 1.9773355489339105e-05, "loss": 0.1932, "step": 1920 }, { "epoch": 0.1, "learning_rate": 1.9773013004484088e-05, "loss": 0.3613, "step": 1921 }, { "epoch": 0.1, "learning_rate": 1.9772670264028717e-05, "loss": 0.4375, "step": 1922 }, { "epoch": 0.1, "learning_rate": 1.9772327267981954e-05, "loss": 0.2722, "step": 1923 }, { "epoch": 0.1, "learning_rate": 1.977198401635277e-05, "loss": 0.1545, "step": 1924 }, { "epoch": 0.1, "learning_rate": 1.9771640509150144e-05, "loss": 0.3142, "step": 1925 }, { "epoch": 0.1, "learning_rate": 1.977129674638306e-05, "loss": 0.2562, "step": 1926 }, { "epoch": 0.1, "learning_rate": 1.9770952728060505e-05, "loss": 0.5785, "step": 1927 }, { "epoch": 0.1, "learning_rate": 1.977060845419148e-05, "loss": 0.1772, "step": 1928 }, { "epoch": 0.1, "learning_rate": 1.9770263924784986e-05, "loss": 0.2634, "step": 1929 }, { "epoch": 0.1, "learning_rate": 1.9769919139850043e-05, "loss": 0.3287, "step": 1930 }, { "epoch": 0.1, "learning_rate": 1.9769574099395654e-05, "loss": 0.3321, "step": 1931 }, { "epoch": 0.1, "learning_rate": 1.9769228803430853e-05, "loss": 0.4441, "step": 1932 }, { "epoch": 0.1, "learning_rate": 1.9768883251964668e-05, "loss": 0.49, "step": 1933 }, { "epoch": 0.1, "learning_rate": 1.9768537445006137e-05, "loss": 0.6285, "step": 1934 }, { "epoch": 0.1, "learning_rate": 1.97681913825643e-05, "loss": 0.2205, "step": 1935 }, { "epoch": 0.1, "learning_rate": 1.9767845064648218e-05, "loss": 0.3432, "step": 1936 }, { "epoch": 0.1, "learning_rate": 1.9767498491266934e-05, "loss": 0.3697, "step": 1937 }, { "epoch": 0.1, "learning_rate": 1.9767151662429525e-05, "loss": 0.2416, "step": 1938 }, { "epoch": 0.1, "learning_rate": 1.9766804578145055e-05, "loss": 0.2928, "step": 1939 }, { "epoch": 0.1, "learning_rate": 1.9766457238422607e-05, "loss": 0.3996, "step": 1940 }, { "epoch": 0.1, "learning_rate": 1.976610964327126e-05, "loss": 0.5593, "step": 1941 }, { "epoch": 0.1, "learning_rate": 1.9765761792700107e-05, "loss": 0.3353, "step": 1942 }, { "epoch": 0.1, "learning_rate": 1.976541368671824e-05, "loss": 0.672, "step": 1943 }, { "epoch": 0.1, "learning_rate": 1.9765065325334776e-05, "loss": 0.4251, "step": 1944 }, { "epoch": 0.1, "learning_rate": 1.9764716708558813e-05, "loss": 0.3978, "step": 1945 }, { "epoch": 0.1, "learning_rate": 1.9764367836399475e-05, "loss": 0.3004, "step": 1946 }, { "epoch": 0.1, "learning_rate": 1.976401870886589e-05, "loss": 0.2664, "step": 1947 }, { "epoch": 0.1, "learning_rate": 1.976366932596718e-05, "loss": 0.4539, "step": 1948 }, { "epoch": 0.1, "learning_rate": 1.9763319687712487e-05, "loss": 0.3402, "step": 1949 }, { "epoch": 0.1, "learning_rate": 1.9762969794110954e-05, "loss": 0.2459, "step": 1950 }, { "epoch": 0.1, "learning_rate": 1.9762619645171733e-05, "loss": 0.3578, "step": 1951 }, { "epoch": 0.1, "learning_rate": 1.9762269240903987e-05, "loss": 0.4471, "step": 1952 }, { "epoch": 0.1, "learning_rate": 1.976191858131687e-05, "loss": 0.1162, "step": 1953 }, { "epoch": 0.1, "learning_rate": 1.976156766641956e-05, "loss": 0.5374, "step": 1954 }, { "epoch": 0.1, "learning_rate": 1.9761216496221235e-05, "loss": 0.5284, "step": 1955 }, { "epoch": 0.1, "learning_rate": 1.9760865070731073e-05, "loss": 0.3888, "step": 1956 }, { "epoch": 0.1, "learning_rate": 1.9760513389958272e-05, "loss": 0.4256, "step": 1957 }, { "epoch": 0.1, "learning_rate": 1.9760161453912028e-05, "loss": 0.4719, "step": 1958 }, { "epoch": 0.1, "learning_rate": 1.9759809262601543e-05, "loss": 0.2938, "step": 1959 }, { "epoch": 0.1, "learning_rate": 1.9759456816036033e-05, "loss": 0.505, "step": 1960 }, { "epoch": 0.1, "learning_rate": 1.975910411422471e-05, "loss": 0.5564, "step": 1961 }, { "epoch": 0.1, "learning_rate": 1.97587511571768e-05, "loss": 0.4761, "step": 1962 }, { "epoch": 0.1, "learning_rate": 1.975839794490154e-05, "loss": 0.4474, "step": 1963 }, { "epoch": 0.1, "learning_rate": 1.975804447740816e-05, "loss": 0.6672, "step": 1964 }, { "epoch": 0.1, "learning_rate": 1.9757690754705904e-05, "loss": 0.3713, "step": 1965 }, { "epoch": 0.1, "learning_rate": 1.975733677680403e-05, "loss": 0.3731, "step": 1966 }, { "epoch": 0.1, "learning_rate": 1.9756982543711792e-05, "loss": 0.4654, "step": 1967 }, { "epoch": 0.1, "learning_rate": 1.9756628055438455e-05, "loss": 0.4863, "step": 1968 }, { "epoch": 0.1, "learning_rate": 1.975627331199329e-05, "loss": 0.374, "step": 1969 }, { "epoch": 0.1, "learning_rate": 1.9755918313385573e-05, "loss": 0.5402, "step": 1970 }, { "epoch": 0.1, "learning_rate": 1.9755563059624597e-05, "loss": 0.4197, "step": 1971 }, { "epoch": 0.1, "learning_rate": 1.975520755071964e-05, "loss": 0.2784, "step": 1972 }, { "epoch": 0.1, "learning_rate": 1.975485178668001e-05, "loss": 0.4802, "step": 1973 }, { "epoch": 0.1, "learning_rate": 1.9754495767515004e-05, "loss": 0.0494, "step": 1974 }, { "epoch": 0.1, "learning_rate": 1.975413949323394e-05, "loss": 0.4411, "step": 1975 }, { "epoch": 0.1, "learning_rate": 1.9753782963846134e-05, "loss": 0.1656, "step": 1976 }, { "epoch": 0.1, "learning_rate": 1.9753426179360908e-05, "loss": 0.3063, "step": 1977 }, { "epoch": 0.1, "learning_rate": 1.9753069139787596e-05, "loss": 0.5128, "step": 1978 }, { "epoch": 0.1, "learning_rate": 1.9752711845135537e-05, "loss": 0.4431, "step": 1979 }, { "epoch": 0.1, "learning_rate": 1.9752354295414072e-05, "loss": 0.4868, "step": 1980 }, { "epoch": 0.1, "learning_rate": 1.975199649063255e-05, "loss": 0.235, "step": 1981 }, { "epoch": 0.1, "learning_rate": 1.9751638430800337e-05, "loss": 0.2581, "step": 1982 }, { "epoch": 0.1, "learning_rate": 1.9751280115926793e-05, "loss": 0.5033, "step": 1983 }, { "epoch": 0.1, "learning_rate": 1.9750921546021285e-05, "loss": 0.229, "step": 1984 }, { "epoch": 0.1, "learning_rate": 1.97505627210932e-05, "loss": 0.4491, "step": 1985 }, { "epoch": 0.1, "learning_rate": 1.9750203641151917e-05, "loss": 0.2447, "step": 1986 }, { "epoch": 0.1, "learning_rate": 1.9749844306206827e-05, "loss": 0.2423, "step": 1987 }, { "epoch": 0.1, "learning_rate": 1.974948471626733e-05, "loss": 0.4827, "step": 1988 }, { "epoch": 0.1, "learning_rate": 1.974912487134283e-05, "loss": 0.3584, "step": 1989 }, { "epoch": 0.1, "learning_rate": 1.9748764771442736e-05, "loss": 0.3357, "step": 1990 }, { "epoch": 0.1, "learning_rate": 1.974840441657647e-05, "loss": 0.4115, "step": 1991 }, { "epoch": 0.1, "learning_rate": 1.9748043806753457e-05, "loss": 0.4009, "step": 1992 }, { "epoch": 0.1, "learning_rate": 1.9747682941983123e-05, "loss": 0.3777, "step": 1993 }, { "epoch": 0.1, "learning_rate": 1.974732182227491e-05, "loss": 0.1332, "step": 1994 }, { "epoch": 0.1, "learning_rate": 1.974696044763826e-05, "loss": 0.4451, "step": 1995 }, { "epoch": 0.1, "learning_rate": 1.9746598818082624e-05, "loss": 0.356, "step": 1996 }, { "epoch": 0.1, "learning_rate": 1.9746236933617468e-05, "loss": 0.5591, "step": 1997 }, { "epoch": 0.1, "learning_rate": 1.9745874794252243e-05, "loss": 0.4245, "step": 1998 }, { "epoch": 0.1, "learning_rate": 1.974551239999643e-05, "loss": 0.5023, "step": 1999 }, { "epoch": 0.1, "learning_rate": 1.9745149750859508e-05, "loss": 0.3194, "step": 2000 }, { "epoch": 0.1, "learning_rate": 1.9744786846850953e-05, "loss": 0.4903, "step": 2001 }, { "epoch": 0.1, "learning_rate": 1.9744423687980265e-05, "loss": 0.425, "step": 2002 }, { "epoch": 0.1, "learning_rate": 1.9744060274256936e-05, "loss": 0.438, "step": 2003 }, { "epoch": 0.1, "learning_rate": 1.9743696605690473e-05, "loss": 0.4202, "step": 2004 }, { "epoch": 0.1, "learning_rate": 1.974333268229039e-05, "loss": 0.2368, "step": 2005 }, { "epoch": 0.1, "learning_rate": 1.9742968504066203e-05, "loss": 0.3802, "step": 2006 }, { "epoch": 0.1, "learning_rate": 1.974260407102743e-05, "loss": 0.388, "step": 2007 }, { "epoch": 0.1, "learning_rate": 1.9742239383183613e-05, "loss": 0.4432, "step": 2008 }, { "epoch": 0.1, "learning_rate": 1.9741874440544282e-05, "loss": 0.4263, "step": 2009 }, { "epoch": 0.1, "learning_rate": 1.9741509243118986e-05, "loss": 0.5603, "step": 2010 }, { "epoch": 0.1, "learning_rate": 1.9741143790917273e-05, "loss": 0.5505, "step": 2011 }, { "epoch": 0.1, "learning_rate": 1.9740778083948706e-05, "loss": 0.3835, "step": 2012 }, { "epoch": 0.1, "learning_rate": 1.9740412122222844e-05, "loss": 0.4451, "step": 2013 }, { "epoch": 0.1, "learning_rate": 1.9740045905749265e-05, "loss": 0.5315, "step": 2014 }, { "epoch": 0.1, "learning_rate": 1.9739679434537534e-05, "loss": 0.3021, "step": 2015 }, { "epoch": 0.1, "learning_rate": 1.9739312708597252e-05, "loss": 0.6214, "step": 2016 }, { "epoch": 0.1, "learning_rate": 1.9738945727937995e-05, "loss": 0.278, "step": 2017 }, { "epoch": 0.1, "learning_rate": 1.9738578492569375e-05, "loss": 0.2943, "step": 2018 }, { "epoch": 0.1, "learning_rate": 1.9738211002500987e-05, "loss": 0.3312, "step": 2019 }, { "epoch": 0.1, "learning_rate": 1.9737843257742446e-05, "loss": 0.5409, "step": 2020 }, { "epoch": 0.1, "learning_rate": 1.973747525830337e-05, "loss": 0.3857, "step": 2021 }, { "epoch": 0.1, "learning_rate": 1.973710700419338e-05, "loss": 0.4269, "step": 2022 }, { "epoch": 0.1, "learning_rate": 1.9736738495422108e-05, "loss": 0.2439, "step": 2023 }, { "epoch": 0.1, "learning_rate": 1.9736369731999197e-05, "loss": 0.302, "step": 2024 }, { "epoch": 0.1, "learning_rate": 1.9736000713934285e-05, "loss": 0.2465, "step": 2025 }, { "epoch": 0.1, "learning_rate": 1.973563144123703e-05, "loss": 0.3478, "step": 2026 }, { "epoch": 0.1, "learning_rate": 1.9735261913917082e-05, "loss": 0.3912, "step": 2027 }, { "epoch": 0.1, "learning_rate": 1.9734892131984113e-05, "loss": 0.2775, "step": 2028 }, { "epoch": 0.1, "learning_rate": 1.9734522095447788e-05, "loss": 0.1913, "step": 2029 }, { "epoch": 0.1, "learning_rate": 1.9734151804317788e-05, "loss": 0.589, "step": 2030 }, { "epoch": 0.1, "learning_rate": 1.97337812586038e-05, "loss": 0.2267, "step": 2031 }, { "epoch": 0.1, "learning_rate": 1.9733410458315508e-05, "loss": 0.3751, "step": 2032 }, { "epoch": 0.1, "learning_rate": 1.9733039403462615e-05, "loss": 0.5305, "step": 2033 }, { "epoch": 0.1, "learning_rate": 1.9732668094054824e-05, "loss": 0.5717, "step": 2034 }, { "epoch": 0.1, "learning_rate": 1.973229653010185e-05, "loss": 0.5425, "step": 2035 }, { "epoch": 0.1, "learning_rate": 1.9731924711613405e-05, "loss": 0.2655, "step": 2036 }, { "epoch": 0.1, "learning_rate": 1.9731552638599214e-05, "loss": 0.276, "step": 2037 }, { "epoch": 0.1, "learning_rate": 1.973118031106901e-05, "loss": 0.3165, "step": 2038 }, { "epoch": 0.1, "learning_rate": 1.9730807729032537e-05, "loss": 0.6139, "step": 2039 }, { "epoch": 0.1, "learning_rate": 1.9730434892499524e-05, "loss": 0.4518, "step": 2040 }, { "epoch": 0.1, "learning_rate": 1.9730061801479735e-05, "loss": 0.3344, "step": 2041 }, { "epoch": 0.1, "learning_rate": 1.9729688455982924e-05, "loss": 0.4755, "step": 2042 }, { "epoch": 0.1, "learning_rate": 1.9729314856018855e-05, "loss": 0.2159, "step": 2043 }, { "epoch": 0.1, "learning_rate": 1.9728941001597296e-05, "loss": 0.383, "step": 2044 }, { "epoch": 0.1, "learning_rate": 1.972856689272803e-05, "loss": 0.4569, "step": 2045 }, { "epoch": 0.1, "learning_rate": 1.972819252942084e-05, "loss": 0.2853, "step": 2046 }, { "epoch": 0.1, "learning_rate": 1.9727817911685516e-05, "loss": 0.3232, "step": 2047 }, { "epoch": 0.1, "learning_rate": 1.9727443039531856e-05, "loss": 0.2695, "step": 2048 }, { "epoch": 0.1, "learning_rate": 1.9727067912969664e-05, "loss": 0.3533, "step": 2049 }, { "epoch": 0.1, "learning_rate": 1.972669253200875e-05, "loss": 0.2812, "step": 2050 }, { "epoch": 0.1, "learning_rate": 1.9726316896658932e-05, "loss": 0.4642, "step": 2051 }, { "epoch": 0.1, "learning_rate": 1.9725941006930033e-05, "loss": 0.4108, "step": 2052 }, { "epoch": 0.1, "learning_rate": 1.9725564862831888e-05, "loss": 0.4982, "step": 2053 }, { "epoch": 0.1, "learning_rate": 1.972518846437433e-05, "loss": 0.4668, "step": 2054 }, { "epoch": 0.1, "learning_rate": 1.972481181156721e-05, "loss": 0.4008, "step": 2055 }, { "epoch": 0.1, "learning_rate": 1.9724434904420375e-05, "loss": 0.2882, "step": 2056 }, { "epoch": 0.1, "learning_rate": 1.9724057742943676e-05, "loss": 0.7283, "step": 2057 }, { "epoch": 0.1, "learning_rate": 1.9723680327146985e-05, "loss": 0.4076, "step": 2058 }, { "epoch": 0.1, "learning_rate": 1.972330265704017e-05, "loss": 0.4276, "step": 2059 }, { "epoch": 0.1, "learning_rate": 1.9722924732633114e-05, "loss": 0.202, "step": 2060 }, { "epoch": 0.1, "learning_rate": 1.972254655393569e-05, "loss": 0.1239, "step": 2061 }, { "epoch": 0.1, "learning_rate": 1.97221681209578e-05, "loss": 0.5457, "step": 2062 }, { "epoch": 0.1, "learning_rate": 1.9721789433709334e-05, "loss": 0.3115, "step": 2063 }, { "epoch": 0.1, "learning_rate": 1.9721410492200198e-05, "loss": 0.3719, "step": 2064 }, { "epoch": 0.1, "learning_rate": 1.9721031296440306e-05, "loss": 0.3834, "step": 2065 }, { "epoch": 0.1, "learning_rate": 1.972065184643957e-05, "loss": 0.2766, "step": 2066 }, { "epoch": 0.1, "learning_rate": 1.972027214220792e-05, "loss": 0.6491, "step": 2067 }, { "epoch": 0.1, "learning_rate": 1.971989218375528e-05, "loss": 0.3103, "step": 2068 }, { "epoch": 0.1, "learning_rate": 1.971951197109159e-05, "loss": 0.1721, "step": 2069 }, { "epoch": 0.1, "learning_rate": 1.9719131504226798e-05, "loss": 0.4772, "step": 2070 }, { "epoch": 0.1, "learning_rate": 1.971875078317085e-05, "loss": 0.3853, "step": 2071 }, { "epoch": 0.1, "learning_rate": 1.9718369807933702e-05, "loss": 0.2624, "step": 2072 }, { "epoch": 0.1, "learning_rate": 1.9717988578525323e-05, "loss": 0.2771, "step": 2073 }, { "epoch": 0.1, "learning_rate": 1.9717607094955685e-05, "loss": 0.5678, "step": 2074 }, { "epoch": 0.1, "learning_rate": 1.9717225357234756e-05, "loss": 0.6197, "step": 2075 }, { "epoch": 0.1, "learning_rate": 1.9716843365372525e-05, "loss": 0.1893, "step": 2076 }, { "epoch": 0.1, "learning_rate": 1.9716461119378984e-05, "loss": 0.349, "step": 2077 }, { "epoch": 0.1, "learning_rate": 1.971607861926413e-05, "loss": 0.4064, "step": 2078 }, { "epoch": 0.1, "learning_rate": 1.9715695865037965e-05, "loss": 0.0532, "step": 2079 }, { "epoch": 0.1, "learning_rate": 1.97153128567105e-05, "loss": 0.6058, "step": 2080 }, { "epoch": 0.1, "learning_rate": 1.971492959429175e-05, "loss": 0.2842, "step": 2081 }, { "epoch": 0.1, "learning_rate": 1.9714546077791738e-05, "loss": 0.2902, "step": 2082 }, { "epoch": 0.1, "learning_rate": 1.97141623072205e-05, "loss": 0.4398, "step": 2083 }, { "epoch": 0.1, "learning_rate": 1.9713778282588074e-05, "loss": 0.3688, "step": 2084 }, { "epoch": 0.1, "learning_rate": 1.9713394003904497e-05, "loss": 0.2188, "step": 2085 }, { "epoch": 0.1, "learning_rate": 1.971300947117982e-05, "loss": 0.3209, "step": 2086 }, { "epoch": 0.1, "learning_rate": 1.9712624684424106e-05, "loss": 0.3057, "step": 2087 }, { "epoch": 0.1, "learning_rate": 1.971223964364741e-05, "loss": 0.4641, "step": 2088 }, { "epoch": 0.1, "learning_rate": 1.971185434885981e-05, "loss": 0.3345, "step": 2089 }, { "epoch": 0.1, "learning_rate": 1.971146880007138e-05, "loss": 0.4053, "step": 2090 }, { "epoch": 0.1, "learning_rate": 1.9711082997292204e-05, "loss": 0.2894, "step": 2091 }, { "epoch": 0.1, "learning_rate": 1.9710696940532367e-05, "loss": 0.2405, "step": 2092 }, { "epoch": 0.1, "learning_rate": 1.971031062980197e-05, "loss": 0.381, "step": 2093 }, { "epoch": 0.1, "learning_rate": 1.9709924065111123e-05, "loss": 0.4763, "step": 2094 }, { "epoch": 0.1, "learning_rate": 1.9709537246469924e-05, "loss": 0.5621, "step": 2095 }, { "epoch": 0.1, "learning_rate": 1.97091501738885e-05, "loss": 0.3697, "step": 2096 }, { "epoch": 0.1, "learning_rate": 1.9708762847376967e-05, "loss": 0.3464, "step": 2097 }, { "epoch": 0.1, "learning_rate": 1.970837526694546e-05, "loss": 0.418, "step": 2098 }, { "epoch": 0.1, "learning_rate": 1.9707987432604113e-05, "loss": 0.3758, "step": 2099 }, { "epoch": 0.1, "learning_rate": 1.970759934436307e-05, "loss": 0.4615, "step": 2100 }, { "epoch": 0.1, "learning_rate": 1.9707211002232477e-05, "loss": 0.3133, "step": 2101 }, { "epoch": 0.1, "learning_rate": 1.97068224062225e-05, "loss": 0.4817, "step": 2102 }, { "epoch": 0.11, "learning_rate": 1.9706433556343293e-05, "loss": 0.5773, "step": 2103 }, { "epoch": 0.11, "learning_rate": 1.970604445260503e-05, "loss": 0.4705, "step": 2104 }, { "epoch": 0.11, "learning_rate": 1.970565509501789e-05, "loss": 0.6387, "step": 2105 }, { "epoch": 0.11, "learning_rate": 1.970526548359205e-05, "loss": 0.3925, "step": 2106 }, { "epoch": 0.11, "learning_rate": 1.97048756183377e-05, "loss": 0.5731, "step": 2107 }, { "epoch": 0.11, "learning_rate": 1.9704485499265048e-05, "loss": 0.3424, "step": 2108 }, { "epoch": 0.11, "learning_rate": 1.970409512638428e-05, "loss": 0.4762, "step": 2109 }, { "epoch": 0.11, "learning_rate": 1.9703704499705617e-05, "loss": 0.4078, "step": 2110 }, { "epoch": 0.11, "learning_rate": 1.9703313619239276e-05, "loss": 0.3957, "step": 2111 }, { "epoch": 0.11, "learning_rate": 1.970292248499547e-05, "loss": 0.3028, "step": 2112 }, { "epoch": 0.11, "learning_rate": 1.9702531096984437e-05, "loss": 0.1626, "step": 2113 }, { "epoch": 0.11, "learning_rate": 1.9702139455216413e-05, "loss": 0.3824, "step": 2114 }, { "epoch": 0.11, "learning_rate": 1.970174755970164e-05, "loss": 0.403, "step": 2115 }, { "epoch": 0.11, "learning_rate": 1.9701355410450365e-05, "loss": 0.4439, "step": 2116 }, { "epoch": 0.11, "learning_rate": 1.9700963007472845e-05, "loss": 0.2991, "step": 2117 }, { "epoch": 0.11, "learning_rate": 1.9700570350779346e-05, "loss": 0.2394, "step": 2118 }, { "epoch": 0.11, "learning_rate": 1.970017744038013e-05, "loss": 0.4255, "step": 2119 }, { "epoch": 0.11, "learning_rate": 1.969978427628548e-05, "loss": 0.571, "step": 2120 }, { "epoch": 0.11, "learning_rate": 1.969939085850568e-05, "loss": 0.3711, "step": 2121 }, { "epoch": 0.11, "learning_rate": 1.969899718705101e-05, "loss": 0.5672, "step": 2122 }, { "epoch": 0.11, "learning_rate": 1.969860326193178e-05, "loss": 0.3199, "step": 2123 }, { "epoch": 0.11, "learning_rate": 1.9698209083158278e-05, "loss": 0.4587, "step": 2124 }, { "epoch": 0.11, "learning_rate": 1.9697814650740822e-05, "loss": 0.5286, "step": 2125 }, { "epoch": 0.11, "learning_rate": 1.9697419964689728e-05, "loss": 0.2775, "step": 2126 }, { "epoch": 0.11, "learning_rate": 1.9697025025015315e-05, "loss": 0.4987, "step": 2127 }, { "epoch": 0.11, "learning_rate": 1.969662983172791e-05, "loss": 0.4778, "step": 2128 }, { "epoch": 0.11, "learning_rate": 1.9696234384837855e-05, "loss": 0.3434, "step": 2129 }, { "epoch": 0.11, "learning_rate": 1.969583868435549e-05, "loss": 0.4794, "step": 2130 }, { "epoch": 0.11, "learning_rate": 1.9695442730291163e-05, "loss": 0.4307, "step": 2131 }, { "epoch": 0.11, "learning_rate": 1.969504652265523e-05, "loss": 0.3653, "step": 2132 }, { "epoch": 0.11, "learning_rate": 1.969465006145805e-05, "loss": 0.5138, "step": 2133 }, { "epoch": 0.11, "learning_rate": 1.9694253346709996e-05, "loss": 0.5551, "step": 2134 }, { "epoch": 0.11, "learning_rate": 1.9693856378421447e-05, "loss": 0.3544, "step": 2135 }, { "epoch": 0.11, "learning_rate": 1.969345915660278e-05, "loss": 0.2988, "step": 2136 }, { "epoch": 0.11, "learning_rate": 1.9693061681264384e-05, "loss": 0.3221, "step": 2137 }, { "epoch": 0.11, "learning_rate": 1.9692663952416654e-05, "loss": 0.4793, "step": 2138 }, { "epoch": 0.11, "learning_rate": 1.9692265970069994e-05, "loss": 0.2705, "step": 2139 }, { "epoch": 0.11, "learning_rate": 1.9691867734234817e-05, "loss": 0.4829, "step": 2140 }, { "epoch": 0.11, "learning_rate": 1.9691469244921526e-05, "loss": 0.4747, "step": 2141 }, { "epoch": 0.11, "learning_rate": 1.9691070502140555e-05, "loss": 0.6272, "step": 2142 }, { "epoch": 0.11, "learning_rate": 1.9690671505902325e-05, "loss": 0.4742, "step": 2143 }, { "epoch": 0.11, "learning_rate": 1.969027225621728e-05, "loss": 0.2924, "step": 2144 }, { "epoch": 0.11, "learning_rate": 1.9689872753095848e-05, "loss": 0.2306, "step": 2145 }, { "epoch": 0.11, "learning_rate": 1.968947299654849e-05, "loss": 0.5166, "step": 2146 }, { "epoch": 0.11, "learning_rate": 1.9689072986585658e-05, "loss": 0.366, "step": 2147 }, { "epoch": 0.11, "learning_rate": 1.968867272321781e-05, "loss": 0.3473, "step": 2148 }, { "epoch": 0.11, "learning_rate": 1.968827220645542e-05, "loss": 0.4352, "step": 2149 }, { "epoch": 0.11, "learning_rate": 1.9687871436308957e-05, "loss": 0.2368, "step": 2150 }, { "epoch": 0.11, "learning_rate": 1.9687470412788905e-05, "loss": 0.3849, "step": 2151 }, { "epoch": 0.11, "learning_rate": 1.9687069135905754e-05, "loss": 0.396, "step": 2152 }, { "epoch": 0.11, "learning_rate": 1.9686667605669996e-05, "loss": 0.2253, "step": 2153 }, { "epoch": 0.11, "learning_rate": 1.9686265822092136e-05, "loss": 0.4463, "step": 2154 }, { "epoch": 0.11, "learning_rate": 1.968586378518268e-05, "loss": 0.3095, "step": 2155 }, { "epoch": 0.11, "learning_rate": 1.9685461494952142e-05, "loss": 0.2199, "step": 2156 }, { "epoch": 0.11, "learning_rate": 1.9685058951411046e-05, "loss": 0.4598, "step": 2157 }, { "epoch": 0.11, "learning_rate": 1.9684656154569914e-05, "loss": 0.4144, "step": 2158 }, { "epoch": 0.11, "learning_rate": 1.968425310443929e-05, "loss": 0.3954, "step": 2159 }, { "epoch": 0.11, "learning_rate": 1.9683849801029707e-05, "loss": 0.4946, "step": 2160 }, { "epoch": 0.11, "learning_rate": 1.9683446244351717e-05, "loss": 0.2732, "step": 2161 }, { "epoch": 0.11, "learning_rate": 1.9683042434415873e-05, "loss": 0.3222, "step": 2162 }, { "epoch": 0.11, "learning_rate": 1.968263837123274e-05, "loss": 0.4272, "step": 2163 }, { "epoch": 0.11, "learning_rate": 1.9682234054812877e-05, "loss": 0.3243, "step": 2164 }, { "epoch": 0.11, "learning_rate": 1.9681829485166867e-05, "loss": 0.2624, "step": 2165 }, { "epoch": 0.11, "learning_rate": 1.9681424662305283e-05, "loss": 0.3195, "step": 2166 }, { "epoch": 0.11, "learning_rate": 1.968101958623872e-05, "loss": 0.3312, "step": 2167 }, { "epoch": 0.11, "learning_rate": 1.9680614256977776e-05, "loss": 0.495, "step": 2168 }, { "epoch": 0.11, "learning_rate": 1.9680208674533038e-05, "loss": 0.5577, "step": 2169 }, { "epoch": 0.11, "learning_rate": 1.9679802838915125e-05, "loss": 0.3733, "step": 2170 }, { "epoch": 0.11, "learning_rate": 1.9679396750134644e-05, "loss": 0.2339, "step": 2171 }, { "epoch": 0.11, "learning_rate": 1.967899040820222e-05, "loss": 0.7535, "step": 2172 }, { "epoch": 0.11, "learning_rate": 1.9678583813128477e-05, "loss": 0.4032, "step": 2173 }, { "epoch": 0.11, "learning_rate": 1.9678176964924055e-05, "loss": 0.5066, "step": 2174 }, { "epoch": 0.11, "learning_rate": 1.967776986359959e-05, "loss": 0.2636, "step": 2175 }, { "epoch": 0.11, "learning_rate": 1.9677362509165726e-05, "loss": 0.1812, "step": 2176 }, { "epoch": 0.11, "learning_rate": 1.9676954901633124e-05, "loss": 0.5297, "step": 2177 }, { "epoch": 0.11, "learning_rate": 1.9676547041012443e-05, "loss": 0.4236, "step": 2178 }, { "epoch": 0.11, "learning_rate": 1.9676138927314344e-05, "loss": 0.3908, "step": 2179 }, { "epoch": 0.11, "learning_rate": 1.9675730560549505e-05, "loss": 0.4399, "step": 2180 }, { "epoch": 0.11, "learning_rate": 1.9675321940728607e-05, "loss": 0.3357, "step": 2181 }, { "epoch": 0.11, "learning_rate": 1.9674913067862334e-05, "loss": 0.3384, "step": 2182 }, { "epoch": 0.11, "learning_rate": 1.9674503941961384e-05, "loss": 0.2515, "step": 2183 }, { "epoch": 0.11, "learning_rate": 1.9674094563036456e-05, "loss": 0.3546, "step": 2184 }, { "epoch": 0.11, "learning_rate": 1.9673684931098257e-05, "loss": 0.4178, "step": 2185 }, { "epoch": 0.11, "learning_rate": 1.9673275046157494e-05, "loss": 0.4132, "step": 2186 }, { "epoch": 0.11, "learning_rate": 1.9672864908224892e-05, "loss": 0.4872, "step": 2187 }, { "epoch": 0.11, "learning_rate": 1.9672454517311182e-05, "loss": 0.4222, "step": 2188 }, { "epoch": 0.11, "learning_rate": 1.967204387342709e-05, "loss": 0.6203, "step": 2189 }, { "epoch": 0.11, "learning_rate": 1.9671632976583358e-05, "loss": 0.3138, "step": 2190 }, { "epoch": 0.11, "learning_rate": 1.9671221826790733e-05, "loss": 0.4322, "step": 2191 }, { "epoch": 0.11, "learning_rate": 1.967081042405997e-05, "loss": 0.2101, "step": 2192 }, { "epoch": 0.11, "learning_rate": 1.9670398768401827e-05, "loss": 0.2966, "step": 2193 }, { "epoch": 0.11, "learning_rate": 1.9669986859827066e-05, "loss": 0.3674, "step": 2194 }, { "epoch": 0.11, "learning_rate": 1.966957469834647e-05, "loss": 0.2933, "step": 2195 }, { "epoch": 0.11, "learning_rate": 1.966916228397081e-05, "loss": 0.616, "step": 2196 }, { "epoch": 0.11, "learning_rate": 1.9668749616710873e-05, "loss": 0.2843, "step": 2197 }, { "epoch": 0.11, "learning_rate": 1.9668336696577454e-05, "loss": 0.5749, "step": 2198 }, { "epoch": 0.11, "learning_rate": 1.9667923523581353e-05, "loss": 0.2707, "step": 2199 }, { "epoch": 0.11, "learning_rate": 1.9667510097733374e-05, "loss": 0.4571, "step": 2200 }, { "epoch": 0.11, "learning_rate": 1.9667096419044332e-05, "loss": 0.3185, "step": 2201 }, { "epoch": 0.11, "learning_rate": 1.9666682487525042e-05, "loss": 0.6208, "step": 2202 }, { "epoch": 0.11, "learning_rate": 1.9666268303186337e-05, "loss": 0.3034, "step": 2203 }, { "epoch": 0.11, "learning_rate": 1.9665853866039044e-05, "loss": 0.3225, "step": 2204 }, { "epoch": 0.11, "learning_rate": 1.9665439176094002e-05, "loss": 0.6344, "step": 2205 }, { "epoch": 0.11, "learning_rate": 1.9665024233362054e-05, "loss": 0.3989, "step": 2206 }, { "epoch": 0.11, "learning_rate": 1.9664609037854064e-05, "loss": 0.22, "step": 2207 }, { "epoch": 0.11, "learning_rate": 1.9664193589580876e-05, "loss": 0.4101, "step": 2208 }, { "epoch": 0.11, "learning_rate": 1.9663777888553367e-05, "loss": 0.464, "step": 2209 }, { "epoch": 0.11, "learning_rate": 1.9663361934782405e-05, "loss": 0.2637, "step": 2210 }, { "epoch": 0.11, "learning_rate": 1.9662945728278865e-05, "loss": 0.6211, "step": 2211 }, { "epoch": 0.11, "learning_rate": 1.9662529269053635e-05, "loss": 0.358, "step": 2212 }, { "epoch": 0.11, "learning_rate": 1.966211255711761e-05, "loss": 0.4671, "step": 2213 }, { "epoch": 0.11, "learning_rate": 1.9661695592481686e-05, "loss": 0.3528, "step": 2214 }, { "epoch": 0.11, "learning_rate": 1.9661278375156768e-05, "loss": 0.3923, "step": 2215 }, { "epoch": 0.11, "learning_rate": 1.966086090515377e-05, "loss": 0.546, "step": 2216 }, { "epoch": 0.11, "learning_rate": 1.9660443182483602e-05, "loss": 0.7421, "step": 2217 }, { "epoch": 0.11, "learning_rate": 1.9660025207157205e-05, "loss": 0.3139, "step": 2218 }, { "epoch": 0.11, "learning_rate": 1.9659606979185493e-05, "loss": 0.3346, "step": 2219 }, { "epoch": 0.11, "learning_rate": 1.9659188498579416e-05, "loss": 0.5243, "step": 2220 }, { "epoch": 0.11, "learning_rate": 1.9658769765349915e-05, "loss": 0.5406, "step": 2221 }, { "epoch": 0.11, "learning_rate": 1.965835077950794e-05, "loss": 0.3625, "step": 2222 }, { "epoch": 0.11, "learning_rate": 1.9657931541064452e-05, "loss": 0.6259, "step": 2223 }, { "epoch": 0.11, "learning_rate": 1.9657512050030413e-05, "loss": 0.2723, "step": 2224 }, { "epoch": 0.11, "learning_rate": 1.9657092306416798e-05, "loss": 0.3906, "step": 2225 }, { "epoch": 0.11, "learning_rate": 1.965667231023458e-05, "loss": 0.5053, "step": 2226 }, { "epoch": 0.11, "learning_rate": 1.9656252061494743e-05, "loss": 0.3389, "step": 2227 }, { "epoch": 0.11, "learning_rate": 1.9655831560208286e-05, "loss": 0.6044, "step": 2228 }, { "epoch": 0.11, "learning_rate": 1.96554108063862e-05, "loss": 0.3379, "step": 2229 }, { "epoch": 0.11, "learning_rate": 1.965498980003949e-05, "loss": 0.3532, "step": 2230 }, { "epoch": 0.11, "learning_rate": 1.9654568541179168e-05, "loss": 0.2332, "step": 2231 }, { "epoch": 0.11, "learning_rate": 1.9654147029816248e-05, "loss": 0.5485, "step": 2232 }, { "epoch": 0.11, "learning_rate": 1.9653725265961762e-05, "loss": 0.7758, "step": 2233 }, { "epoch": 0.11, "learning_rate": 1.965330324962673e-05, "loss": 0.4242, "step": 2234 }, { "epoch": 0.11, "learning_rate": 1.96528809808222e-05, "loss": 0.1609, "step": 2235 }, { "epoch": 0.11, "learning_rate": 1.965245845955921e-05, "loss": 0.411, "step": 2236 }, { "epoch": 0.11, "learning_rate": 1.9652035685848808e-05, "loss": 0.3051, "step": 2237 }, { "epoch": 0.11, "learning_rate": 1.9651612659702056e-05, "loss": 0.6305, "step": 2238 }, { "epoch": 0.11, "learning_rate": 1.965118938113002e-05, "loss": 0.3259, "step": 2239 }, { "epoch": 0.11, "learning_rate": 1.9650765850143766e-05, "loss": 0.38, "step": 2240 }, { "epoch": 0.11, "learning_rate": 1.9650342066754367e-05, "loss": 0.4084, "step": 2241 }, { "epoch": 0.11, "learning_rate": 1.9649918030972914e-05, "loss": 0.5582, "step": 2242 }, { "epoch": 0.11, "learning_rate": 1.9649493742810493e-05, "loss": 0.3257, "step": 2243 }, { "epoch": 0.11, "learning_rate": 1.9649069202278203e-05, "loss": 0.5523, "step": 2244 }, { "epoch": 0.11, "learning_rate": 1.9648644409387145e-05, "loss": 0.4505, "step": 2245 }, { "epoch": 0.11, "learning_rate": 1.9648219364148427e-05, "loss": 0.3301, "step": 2246 }, { "epoch": 0.11, "learning_rate": 1.9647794066573173e-05, "loss": 0.4005, "step": 2247 }, { "epoch": 0.11, "learning_rate": 1.96473685166725e-05, "loss": 0.4359, "step": 2248 }, { "epoch": 0.11, "learning_rate": 1.964694271445754e-05, "loss": 0.359, "step": 2249 }, { "epoch": 0.11, "learning_rate": 1.964651665993943e-05, "loss": 0.407, "step": 2250 }, { "epoch": 0.11, "learning_rate": 1.9646090353129307e-05, "loss": 0.3936, "step": 2251 }, { "epoch": 0.11, "learning_rate": 1.9645663794038327e-05, "loss": 0.469, "step": 2252 }, { "epoch": 0.11, "learning_rate": 1.9645236982677645e-05, "loss": 0.3546, "step": 2253 }, { "epoch": 0.11, "learning_rate": 1.9644809919058422e-05, "loss": 0.3504, "step": 2254 }, { "epoch": 0.11, "learning_rate": 1.964438260319183e-05, "loss": 0.4176, "step": 2255 }, { "epoch": 0.11, "learning_rate": 1.964395503508904e-05, "loss": 0.6145, "step": 2256 }, { "epoch": 0.11, "learning_rate": 1.9643527214761236e-05, "loss": 0.3852, "step": 2257 }, { "epoch": 0.11, "learning_rate": 1.964309914221961e-05, "loss": 0.33, "step": 2258 }, { "epoch": 0.11, "learning_rate": 1.964267081747536e-05, "loss": 0.4322, "step": 2259 }, { "epoch": 0.11, "learning_rate": 1.9642242240539676e-05, "loss": 0.5215, "step": 2260 }, { "epoch": 0.11, "learning_rate": 1.9641813411423782e-05, "loss": 0.4243, "step": 2261 }, { "epoch": 0.11, "learning_rate": 1.9641384330138885e-05, "loss": 0.4513, "step": 2262 }, { "epoch": 0.11, "learning_rate": 1.964095499669621e-05, "loss": 0.3839, "step": 2263 }, { "epoch": 0.11, "learning_rate": 1.9640525411106985e-05, "loss": 0.2432, "step": 2264 }, { "epoch": 0.11, "learning_rate": 1.964009557338244e-05, "loss": 0.3752, "step": 2265 }, { "epoch": 0.11, "learning_rate": 1.9639665483533825e-05, "loss": 0.2782, "step": 2266 }, { "epoch": 0.11, "learning_rate": 1.9639235141572384e-05, "loss": 0.3339, "step": 2267 }, { "epoch": 0.11, "learning_rate": 1.9638804547509375e-05, "loss": 0.2785, "step": 2268 }, { "epoch": 0.11, "learning_rate": 1.9638373701356055e-05, "loss": 0.3256, "step": 2269 }, { "epoch": 0.11, "learning_rate": 1.9637942603123697e-05, "loss": 0.7429, "step": 2270 }, { "epoch": 0.11, "learning_rate": 1.9637511252823573e-05, "loss": 0.3519, "step": 2271 }, { "epoch": 0.11, "learning_rate": 1.9637079650466963e-05, "loss": 0.5591, "step": 2272 }, { "epoch": 0.11, "learning_rate": 1.9636647796065162e-05, "loss": 0.2569, "step": 2273 }, { "epoch": 0.11, "learning_rate": 1.9636215689629453e-05, "loss": 0.5634, "step": 2274 }, { "epoch": 0.11, "learning_rate": 1.963578333117115e-05, "loss": 0.222, "step": 2275 }, { "epoch": 0.11, "learning_rate": 1.963535072070155e-05, "loss": 0.4142, "step": 2276 }, { "epoch": 0.11, "learning_rate": 1.9634917858231974e-05, "loss": 0.6576, "step": 2277 }, { "epoch": 0.11, "learning_rate": 1.9634484743773742e-05, "loss": 0.364, "step": 2278 }, { "epoch": 0.11, "learning_rate": 1.9634051377338177e-05, "loss": 0.2315, "step": 2279 }, { "epoch": 0.11, "learning_rate": 1.963361775893662e-05, "loss": 0.3011, "step": 2280 }, { "epoch": 0.11, "learning_rate": 1.9633183888580405e-05, "loss": 0.4033, "step": 2281 }, { "epoch": 0.11, "learning_rate": 1.9632749766280885e-05, "loss": 0.4081, "step": 2282 }, { "epoch": 0.11, "learning_rate": 1.963231539204941e-05, "loss": 0.4966, "step": 2283 }, { "epoch": 0.11, "learning_rate": 1.9631880765897345e-05, "loss": 0.4145, "step": 2284 }, { "epoch": 0.11, "learning_rate": 1.963144588783605e-05, "loss": 0.6776, "step": 2285 }, { "epoch": 0.11, "learning_rate": 1.9631010757876905e-05, "loss": 0.4566, "step": 2286 }, { "epoch": 0.11, "learning_rate": 1.9630575376031288e-05, "loss": 0.1114, "step": 2287 }, { "epoch": 0.11, "learning_rate": 1.9630139742310585e-05, "loss": 0.3167, "step": 2288 }, { "epoch": 0.11, "learning_rate": 1.9629703856726193e-05, "loss": 0.54, "step": 2289 }, { "epoch": 0.11, "learning_rate": 1.9629267719289508e-05, "loss": 0.372, "step": 2290 }, { "epoch": 0.11, "learning_rate": 1.9628831330011936e-05, "loss": 0.5849, "step": 2291 }, { "epoch": 0.11, "learning_rate": 1.962839468890489e-05, "loss": 0.3029, "step": 2292 }, { "epoch": 0.11, "learning_rate": 1.9627957795979796e-05, "loss": 0.4581, "step": 2293 }, { "epoch": 0.11, "learning_rate": 1.9627520651248074e-05, "loss": 0.3218, "step": 2294 }, { "epoch": 0.11, "learning_rate": 1.962708325472116e-05, "loss": 0.191, "step": 2295 }, { "epoch": 0.11, "learning_rate": 1.9626645606410493e-05, "loss": 0.5017, "step": 2296 }, { "epoch": 0.11, "learning_rate": 1.962620770632752e-05, "loss": 0.4021, "step": 2297 }, { "epoch": 0.11, "learning_rate": 1.962576955448369e-05, "loss": 0.4365, "step": 2298 }, { "epoch": 0.11, "learning_rate": 1.9625331150890468e-05, "loss": 0.5112, "step": 2299 }, { "epoch": 0.11, "learning_rate": 1.9624892495559313e-05, "loss": 0.4241, "step": 2300 }, { "epoch": 0.11, "learning_rate": 1.9624453588501703e-05, "loss": 0.312, "step": 2301 }, { "epoch": 0.11, "learning_rate": 1.9624014429729117e-05, "loss": 0.4048, "step": 2302 }, { "epoch": 0.11, "learning_rate": 1.9623575019253036e-05, "loss": 0.3666, "step": 2303 }, { "epoch": 0.12, "learning_rate": 1.9623135357084954e-05, "loss": 0.3311, "step": 2304 }, { "epoch": 0.12, "learning_rate": 1.9622695443236374e-05, "loss": 0.3282, "step": 2305 }, { "epoch": 0.12, "learning_rate": 1.9622255277718797e-05, "loss": 0.1265, "step": 2306 }, { "epoch": 0.12, "learning_rate": 1.9621814860543737e-05, "loss": 0.4198, "step": 2307 }, { "epoch": 0.12, "learning_rate": 1.962137419172271e-05, "loss": 0.6251, "step": 2308 }, { "epoch": 0.12, "learning_rate": 1.9620933271267242e-05, "loss": 0.368, "step": 2309 }, { "epoch": 0.12, "learning_rate": 1.9620492099188867e-05, "loss": 0.3514, "step": 2310 }, { "epoch": 0.12, "learning_rate": 1.962005067549912e-05, "loss": 0.4961, "step": 2311 }, { "epoch": 0.12, "learning_rate": 1.9619609000209548e-05, "loss": 0.3897, "step": 2312 }, { "epoch": 0.12, "learning_rate": 1.9619167073331704e-05, "loss": 0.6448, "step": 2313 }, { "epoch": 0.12, "learning_rate": 1.9618724894877144e-05, "loss": 0.3899, "step": 2314 }, { "epoch": 0.12, "learning_rate": 1.9618282464857433e-05, "loss": 0.3747, "step": 2315 }, { "epoch": 0.12, "learning_rate": 1.9617839783284137e-05, "loss": 0.5658, "step": 2316 }, { "epoch": 0.12, "learning_rate": 1.9617396850168845e-05, "loss": 0.59, "step": 2317 }, { "epoch": 0.12, "learning_rate": 1.961695366552313e-05, "loss": 0.5664, "step": 2318 }, { "epoch": 0.12, "learning_rate": 1.961651022935859e-05, "loss": 0.4114, "step": 2319 }, { "epoch": 0.12, "learning_rate": 1.961606654168682e-05, "loss": 0.4089, "step": 2320 }, { "epoch": 0.12, "learning_rate": 1.9615622602519427e-05, "loss": 0.3318, "step": 2321 }, { "epoch": 0.12, "learning_rate": 1.9615178411868014e-05, "loss": 0.1994, "step": 2322 }, { "epoch": 0.12, "learning_rate": 1.9614733969744207e-05, "loss": 0.4061, "step": 2323 }, { "epoch": 0.12, "learning_rate": 1.9614289276159625e-05, "loss": 0.2204, "step": 2324 }, { "epoch": 0.12, "learning_rate": 1.9613844331125896e-05, "loss": 0.3563, "step": 2325 }, { "epoch": 0.12, "learning_rate": 1.9613399134654666e-05, "loss": 0.4938, "step": 2326 }, { "epoch": 0.12, "learning_rate": 1.961295368675757e-05, "loss": 0.4335, "step": 2327 }, { "epoch": 0.12, "learning_rate": 1.9612507987446265e-05, "loss": 0.2455, "step": 2328 }, { "epoch": 0.12, "learning_rate": 1.9612062036732394e-05, "loss": 0.3022, "step": 2329 }, { "epoch": 0.12, "learning_rate": 1.961161583462764e-05, "loss": 0.4365, "step": 2330 }, { "epoch": 0.12, "learning_rate": 1.961116938114366e-05, "loss": 0.6826, "step": 2331 }, { "epoch": 0.12, "learning_rate": 1.961072267629213e-05, "loss": 0.5472, "step": 2332 }, { "epoch": 0.12, "learning_rate": 1.961027572008474e-05, "loss": 0.2833, "step": 2333 }, { "epoch": 0.12, "learning_rate": 1.9609828512533175e-05, "loss": 0.6508, "step": 2334 }, { "epoch": 0.12, "learning_rate": 1.960938105364913e-05, "loss": 0.3732, "step": 2335 }, { "epoch": 0.12, "learning_rate": 1.960893334344431e-05, "loss": 0.3037, "step": 2336 }, { "epoch": 0.12, "learning_rate": 1.9608485381930425e-05, "loss": 0.5498, "step": 2337 }, { "epoch": 0.12, "learning_rate": 1.960803716911919e-05, "loss": 0.6281, "step": 2338 }, { "epoch": 0.12, "learning_rate": 1.9607588705022326e-05, "loss": 0.3346, "step": 2339 }, { "epoch": 0.12, "learning_rate": 1.960713998965156e-05, "loss": 0.454, "step": 2340 }, { "epoch": 0.12, "learning_rate": 1.9606691023018637e-05, "loss": 0.3058, "step": 2341 }, { "epoch": 0.12, "learning_rate": 1.960624180513529e-05, "loss": 0.3721, "step": 2342 }, { "epoch": 0.12, "learning_rate": 1.960579233601327e-05, "loss": 0.3381, "step": 2343 }, { "epoch": 0.12, "learning_rate": 1.960534261566433e-05, "loss": 0.5188, "step": 2344 }, { "epoch": 0.12, "learning_rate": 1.9604892644100242e-05, "loss": 0.4136, "step": 2345 }, { "epoch": 0.12, "learning_rate": 1.960444242133276e-05, "loss": 0.3423, "step": 2346 }, { "epoch": 0.12, "learning_rate": 1.960399194737367e-05, "loss": 0.2761, "step": 2347 }, { "epoch": 0.12, "learning_rate": 1.960354122223475e-05, "loss": 0.4187, "step": 2348 }, { "epoch": 0.12, "learning_rate": 1.9603090245927784e-05, "loss": 0.3508, "step": 2349 }, { "epoch": 0.12, "learning_rate": 1.9602639018464574e-05, "loss": 0.2989, "step": 2350 }, { "epoch": 0.12, "learning_rate": 1.9602187539856918e-05, "loss": 0.574, "step": 2351 }, { "epoch": 0.12, "learning_rate": 1.9601735810116617e-05, "loss": 0.3792, "step": 2352 }, { "epoch": 0.12, "learning_rate": 1.9601283829255496e-05, "loss": 0.4459, "step": 2353 }, { "epoch": 0.12, "learning_rate": 1.9600831597285376e-05, "loss": 0.3409, "step": 2354 }, { "epoch": 0.12, "learning_rate": 1.9600379114218075e-05, "loss": 0.3703, "step": 2355 }, { "epoch": 0.12, "learning_rate": 1.9599926380065435e-05, "loss": 0.5774, "step": 2356 }, { "epoch": 0.12, "learning_rate": 1.9599473394839293e-05, "loss": 0.4174, "step": 2357 }, { "epoch": 0.12, "learning_rate": 1.9599020158551498e-05, "loss": 0.1853, "step": 2358 }, { "epoch": 0.12, "learning_rate": 1.9598566671213904e-05, "loss": 0.2628, "step": 2359 }, { "epoch": 0.12, "learning_rate": 1.959811293283837e-05, "loss": 0.4272, "step": 2360 }, { "epoch": 0.12, "learning_rate": 1.959765894343676e-05, "loss": 0.5799, "step": 2361 }, { "epoch": 0.12, "learning_rate": 1.9597204703020955e-05, "loss": 0.4787, "step": 2362 }, { "epoch": 0.12, "learning_rate": 1.9596750211602832e-05, "loss": 0.31, "step": 2363 }, { "epoch": 0.12, "learning_rate": 1.9596295469194274e-05, "loss": 0.2437, "step": 2364 }, { "epoch": 0.12, "learning_rate": 1.959584047580718e-05, "loss": 0.3077, "step": 2365 }, { "epoch": 0.12, "learning_rate": 1.9595385231453445e-05, "loss": 0.5837, "step": 2366 }, { "epoch": 0.12, "learning_rate": 1.9594929736144978e-05, "loss": 0.2653, "step": 2367 }, { "epoch": 0.12, "learning_rate": 1.9594473989893686e-05, "loss": 0.4856, "step": 2368 }, { "epoch": 0.12, "learning_rate": 1.9594017992711497e-05, "loss": 0.6401, "step": 2369 }, { "epoch": 0.12, "learning_rate": 1.9593561744610335e-05, "loss": 0.4792, "step": 2370 }, { "epoch": 0.12, "learning_rate": 1.9593105245602127e-05, "loss": 0.6239, "step": 2371 }, { "epoch": 0.12, "learning_rate": 1.959264849569882e-05, "loss": 0.548, "step": 2372 }, { "epoch": 0.12, "learning_rate": 1.9592191494912353e-05, "loss": 0.342, "step": 2373 }, { "epoch": 0.12, "learning_rate": 1.959173424325468e-05, "loss": 0.5199, "step": 2374 }, { "epoch": 0.12, "learning_rate": 1.959127674073776e-05, "loss": 0.6685, "step": 2375 }, { "epoch": 0.12, "learning_rate": 1.9590818987373565e-05, "loss": 0.3525, "step": 2376 }, { "epoch": 0.12, "learning_rate": 1.9590360983174058e-05, "loss": 0.5012, "step": 2377 }, { "epoch": 0.12, "learning_rate": 1.9589902728151218e-05, "loss": 0.7393, "step": 2378 }, { "epoch": 0.12, "learning_rate": 1.9589444222317034e-05, "loss": 0.2628, "step": 2379 }, { "epoch": 0.12, "learning_rate": 1.95889854656835e-05, "loss": 0.5453, "step": 2380 }, { "epoch": 0.12, "learning_rate": 1.9588526458262607e-05, "loss": 0.5105, "step": 2381 }, { "epoch": 0.12, "learning_rate": 1.9588067200066363e-05, "loss": 0.4972, "step": 2382 }, { "epoch": 0.12, "learning_rate": 1.958760769110678e-05, "loss": 0.4791, "step": 2383 }, { "epoch": 0.12, "learning_rate": 1.9587147931395876e-05, "loss": 0.4257, "step": 2384 }, { "epoch": 0.12, "learning_rate": 1.9586687920945674e-05, "loss": 0.3198, "step": 2385 }, { "epoch": 0.12, "learning_rate": 1.9586227659768206e-05, "loss": 0.5253, "step": 2386 }, { "epoch": 0.12, "learning_rate": 1.958576714787551e-05, "loss": 0.4599, "step": 2387 }, { "epoch": 0.12, "learning_rate": 1.9585306385279628e-05, "loss": 0.3972, "step": 2388 }, { "epoch": 0.12, "learning_rate": 1.9584845371992614e-05, "loss": 0.5051, "step": 2389 }, { "epoch": 0.12, "learning_rate": 1.958438410802652e-05, "loss": 0.3943, "step": 2390 }, { "epoch": 0.12, "learning_rate": 1.9583922593393418e-05, "loss": 0.234, "step": 2391 }, { "epoch": 0.12, "learning_rate": 1.958346082810537e-05, "loss": 0.2461, "step": 2392 }, { "epoch": 0.12, "learning_rate": 1.9582998812174455e-05, "loss": 0.3211, "step": 2393 }, { "epoch": 0.12, "learning_rate": 1.958253654561276e-05, "loss": 0.474, "step": 2394 }, { "epoch": 0.12, "learning_rate": 1.958207402843237e-05, "loss": 0.285, "step": 2395 }, { "epoch": 0.12, "learning_rate": 1.9581611260645386e-05, "loss": 0.3742, "step": 2396 }, { "epoch": 0.12, "learning_rate": 1.9581148242263907e-05, "loss": 0.3888, "step": 2397 }, { "epoch": 0.12, "learning_rate": 1.9580684973300047e-05, "loss": 0.2923, "step": 2398 }, { "epoch": 0.12, "learning_rate": 1.958022145376592e-05, "loss": 0.4663, "step": 2399 }, { "epoch": 0.12, "learning_rate": 1.957975768367365e-05, "loss": 0.3128, "step": 2400 }, { "epoch": 0.12, "learning_rate": 1.9579293663035363e-05, "loss": 0.3057, "step": 2401 }, { "epoch": 0.12, "learning_rate": 1.9578829391863194e-05, "loss": 0.5796, "step": 2402 }, { "epoch": 0.12, "learning_rate": 1.957836487016929e-05, "loss": 0.5577, "step": 2403 }, { "epoch": 0.12, "learning_rate": 1.95779000979658e-05, "loss": 0.4145, "step": 2404 }, { "epoch": 0.12, "learning_rate": 1.9577435075264877e-05, "loss": 0.3518, "step": 2405 }, { "epoch": 0.12, "learning_rate": 1.9576969802078684e-05, "loss": 0.4171, "step": 2406 }, { "epoch": 0.12, "learning_rate": 1.957650427841939e-05, "loss": 0.4026, "step": 2407 }, { "epoch": 0.12, "learning_rate": 1.9576038504299165e-05, "loss": 0.3484, "step": 2408 }, { "epoch": 0.12, "learning_rate": 1.9575572479730204e-05, "loss": 0.3735, "step": 2409 }, { "epoch": 0.12, "learning_rate": 1.957510620472468e-05, "loss": 0.6213, "step": 2410 }, { "epoch": 0.12, "learning_rate": 1.9574639679294796e-05, "loss": 0.5361, "step": 2411 }, { "epoch": 0.12, "learning_rate": 1.957417290345275e-05, "loss": 0.3646, "step": 2412 }, { "epoch": 0.12, "learning_rate": 1.9573705877210755e-05, "loss": 0.4572, "step": 2413 }, { "epoch": 0.12, "learning_rate": 1.957323860058102e-05, "loss": 0.3847, "step": 2414 }, { "epoch": 0.12, "learning_rate": 1.9572771073575766e-05, "loss": 0.2036, "step": 2415 }, { "epoch": 0.12, "learning_rate": 1.957230329620723e-05, "loss": 0.6875, "step": 2416 }, { "epoch": 0.12, "learning_rate": 1.9571835268487632e-05, "loss": 0.3451, "step": 2417 }, { "epoch": 0.12, "learning_rate": 1.9571366990429222e-05, "loss": 0.3928, "step": 2418 }, { "epoch": 0.12, "learning_rate": 1.9570898462044242e-05, "loss": 0.7913, "step": 2419 }, { "epoch": 0.12, "learning_rate": 1.957042968334495e-05, "loss": 0.5542, "step": 2420 }, { "epoch": 0.12, "learning_rate": 1.9569960654343605e-05, "loss": 0.3578, "step": 2421 }, { "epoch": 0.12, "learning_rate": 1.9569491375052476e-05, "loss": 0.3958, "step": 2422 }, { "epoch": 0.12, "learning_rate": 1.9569021845483828e-05, "loss": 0.3316, "step": 2423 }, { "epoch": 0.12, "learning_rate": 1.956855206564995e-05, "loss": 0.2187, "step": 2424 }, { "epoch": 0.12, "learning_rate": 1.9568082035563126e-05, "loss": 0.2977, "step": 2425 }, { "epoch": 0.12, "learning_rate": 1.956761175523565e-05, "loss": 0.3988, "step": 2426 }, { "epoch": 0.12, "learning_rate": 1.9567141224679816e-05, "loss": 0.2812, "step": 2427 }, { "epoch": 0.12, "learning_rate": 1.9566670443907933e-05, "loss": 0.4458, "step": 2428 }, { "epoch": 0.12, "learning_rate": 1.956619941293232e-05, "loss": 0.2439, "step": 2429 }, { "epoch": 0.12, "learning_rate": 1.9565728131765287e-05, "loss": 0.3658, "step": 2430 }, { "epoch": 0.12, "learning_rate": 1.9565256600419166e-05, "loss": 0.1188, "step": 2431 }, { "epoch": 0.12, "learning_rate": 1.9564784818906282e-05, "loss": 0.2974, "step": 2432 }, { "epoch": 0.12, "learning_rate": 1.9564312787238985e-05, "loss": 0.2704, "step": 2433 }, { "epoch": 0.12, "learning_rate": 1.956384050542961e-05, "loss": 0.7696, "step": 2434 }, { "epoch": 0.12, "learning_rate": 1.9563367973490514e-05, "loss": 0.3246, "step": 2435 }, { "epoch": 0.12, "learning_rate": 1.9562895191434052e-05, "loss": 0.2756, "step": 2436 }, { "epoch": 0.12, "learning_rate": 1.9562422159272597e-05, "loss": 0.3195, "step": 2437 }, { "epoch": 0.12, "learning_rate": 1.956194887701851e-05, "loss": 0.7536, "step": 2438 }, { "epoch": 0.12, "learning_rate": 1.956147534468418e-05, "loss": 0.3705, "step": 2439 }, { "epoch": 0.12, "learning_rate": 1.956100156228198e-05, "loss": 0.7796, "step": 2440 }, { "epoch": 0.12, "learning_rate": 1.9560527529824303e-05, "loss": 0.4835, "step": 2441 }, { "epoch": 0.12, "learning_rate": 1.956005324732356e-05, "loss": 0.413, "step": 2442 }, { "epoch": 0.12, "learning_rate": 1.955957871479214e-05, "loss": 0.2674, "step": 2443 }, { "epoch": 0.12, "learning_rate": 1.955910393224246e-05, "loss": 0.5007, "step": 2444 }, { "epoch": 0.12, "learning_rate": 1.9558628899686937e-05, "loss": 0.4995, "step": 2445 }, { "epoch": 0.12, "learning_rate": 1.9558153617137997e-05, "loss": 0.1343, "step": 2446 }, { "epoch": 0.12, "learning_rate": 1.955767808460807e-05, "loss": 0.586, "step": 2447 }, { "epoch": 0.12, "learning_rate": 1.9557202302109583e-05, "loss": 0.3955, "step": 2448 }, { "epoch": 0.12, "learning_rate": 1.9556726269654992e-05, "loss": 0.3891, "step": 2449 }, { "epoch": 0.12, "learning_rate": 1.9556249987256743e-05, "loss": 0.3416, "step": 2450 }, { "epoch": 0.12, "learning_rate": 1.955577345492729e-05, "loss": 0.5036, "step": 2451 }, { "epoch": 0.12, "learning_rate": 1.9555296672679104e-05, "loss": 0.4631, "step": 2452 }, { "epoch": 0.12, "learning_rate": 1.955481964052464e-05, "loss": 0.5101, "step": 2453 }, { "epoch": 0.12, "learning_rate": 1.9554342358476386e-05, "loss": 0.3766, "step": 2454 }, { "epoch": 0.12, "learning_rate": 1.9553864826546826e-05, "loss": 0.3303, "step": 2455 }, { "epoch": 0.12, "learning_rate": 1.955338704474844e-05, "loss": 0.6406, "step": 2456 }, { "epoch": 0.12, "learning_rate": 1.9552909013093732e-05, "loss": 0.4478, "step": 2457 }, { "epoch": 0.12, "learning_rate": 1.95524307315952e-05, "loss": 0.3332, "step": 2458 }, { "epoch": 0.12, "learning_rate": 1.9551952200265352e-05, "loss": 0.3914, "step": 2459 }, { "epoch": 0.12, "learning_rate": 1.9551473419116704e-05, "loss": 0.3859, "step": 2460 }, { "epoch": 0.12, "learning_rate": 1.9550994388161783e-05, "loss": 0.4553, "step": 2461 }, { "epoch": 0.12, "learning_rate": 1.9550515107413108e-05, "loss": 0.5002, "step": 2462 }, { "epoch": 0.12, "learning_rate": 1.9550035576883224e-05, "loss": 0.3295, "step": 2463 }, { "epoch": 0.12, "learning_rate": 1.9549555796584665e-05, "loss": 0.4068, "step": 2464 }, { "epoch": 0.12, "learning_rate": 1.9549075766529984e-05, "loss": 0.5178, "step": 2465 }, { "epoch": 0.12, "learning_rate": 1.9548595486731733e-05, "loss": 0.5563, "step": 2466 }, { "epoch": 0.12, "learning_rate": 1.9548114957202472e-05, "loss": 0.5758, "step": 2467 }, { "epoch": 0.12, "learning_rate": 1.954763417795477e-05, "loss": 0.4011, "step": 2468 }, { "epoch": 0.12, "learning_rate": 1.9547153149001203e-05, "loss": 0.3647, "step": 2469 }, { "epoch": 0.12, "learning_rate": 1.9546671870354347e-05, "loss": 0.4569, "step": 2470 }, { "epoch": 0.12, "learning_rate": 1.9546190342026793e-05, "loss": 0.341, "step": 2471 }, { "epoch": 0.12, "learning_rate": 1.9545708564031135e-05, "loss": 0.2389, "step": 2472 }, { "epoch": 0.12, "learning_rate": 1.954522653637997e-05, "loss": 0.33, "step": 2473 }, { "epoch": 0.12, "learning_rate": 1.954474425908591e-05, "loss": 0.5392, "step": 2474 }, { "epoch": 0.12, "learning_rate": 1.9544261732161563e-05, "loss": 0.4733, "step": 2475 }, { "epoch": 0.12, "learning_rate": 1.9543778955619553e-05, "loss": 0.4633, "step": 2476 }, { "epoch": 0.12, "learning_rate": 1.9543295929472505e-05, "loss": 0.3949, "step": 2477 }, { "epoch": 0.12, "learning_rate": 1.954281265373305e-05, "loss": 0.4981, "step": 2478 }, { "epoch": 0.12, "learning_rate": 1.9542329128413827e-05, "loss": 0.528, "step": 2479 }, { "epoch": 0.12, "learning_rate": 1.9541845353527486e-05, "loss": 0.3053, "step": 2480 }, { "epoch": 0.12, "learning_rate": 1.9541361329086678e-05, "loss": 0.2355, "step": 2481 }, { "epoch": 0.12, "learning_rate": 1.9540877055104062e-05, "loss": 0.3362, "step": 2482 }, { "epoch": 0.12, "learning_rate": 1.95403925315923e-05, "loss": 0.4113, "step": 2483 }, { "epoch": 0.12, "learning_rate": 1.953990775856407e-05, "loss": 0.2208, "step": 2484 }, { "epoch": 0.12, "learning_rate": 1.9539422736032046e-05, "loss": 0.278, "step": 2485 }, { "epoch": 0.12, "learning_rate": 1.9538937464008912e-05, "loss": 0.3669, "step": 2486 }, { "epoch": 0.12, "learning_rate": 1.953845194250737e-05, "loss": 0.3176, "step": 2487 }, { "epoch": 0.12, "learning_rate": 1.9537966171540105e-05, "loss": 0.4686, "step": 2488 }, { "epoch": 0.12, "learning_rate": 1.9537480151119825e-05, "loss": 0.4569, "step": 2489 }, { "epoch": 0.12, "learning_rate": 1.953699388125925e-05, "loss": 0.5094, "step": 2490 }, { "epoch": 0.12, "learning_rate": 1.9536507361971088e-05, "loss": 0.4977, "step": 2491 }, { "epoch": 0.12, "learning_rate": 1.9536020593268065e-05, "loss": 0.4593, "step": 2492 }, { "epoch": 0.12, "learning_rate": 1.9535533575162913e-05, "loss": 0.2312, "step": 2493 }, { "epoch": 0.12, "learning_rate": 1.953504630766837e-05, "loss": 0.3266, "step": 2494 }, { "epoch": 0.12, "learning_rate": 1.9534558790797186e-05, "loss": 0.5219, "step": 2495 }, { "epoch": 0.12, "learning_rate": 1.95340710245621e-05, "loss": 0.4746, "step": 2496 }, { "epoch": 0.12, "learning_rate": 1.953358300897587e-05, "loss": 0.4141, "step": 2497 }, { "epoch": 0.12, "learning_rate": 1.953309474405127e-05, "loss": 0.2244, "step": 2498 }, { "epoch": 0.12, "learning_rate": 1.9532606229801058e-05, "loss": 0.3881, "step": 2499 }, { "epoch": 0.12, "learning_rate": 1.9532117466238016e-05, "loss": 0.3289, "step": 2500 }, { "epoch": 0.12, "learning_rate": 1.9531628453374928e-05, "loss": 0.5023, "step": 2501 }, { "epoch": 0.12, "learning_rate": 1.9531139191224583e-05, "loss": 0.3507, "step": 2502 }, { "epoch": 0.12, "learning_rate": 1.9530649679799773e-05, "loss": 0.3149, "step": 2503 }, { "epoch": 0.13, "learning_rate": 1.9530159919113306e-05, "loss": 0.3215, "step": 2504 }, { "epoch": 0.13, "learning_rate": 1.9529669909177985e-05, "loss": 0.3975, "step": 2505 }, { "epoch": 0.13, "learning_rate": 1.9529179650006635e-05, "loss": 0.3031, "step": 2506 }, { "epoch": 0.13, "learning_rate": 1.952868914161207e-05, "loss": 0.1133, "step": 2507 }, { "epoch": 0.13, "learning_rate": 1.9528198384007118e-05, "loss": 0.5509, "step": 2508 }, { "epoch": 0.13, "learning_rate": 1.952770737720462e-05, "loss": 0.3001, "step": 2509 }, { "epoch": 0.13, "learning_rate": 1.9527216121217413e-05, "loss": 0.3094, "step": 2510 }, { "epoch": 0.13, "learning_rate": 1.9526724616058348e-05, "loss": 0.3777, "step": 2511 }, { "epoch": 0.13, "learning_rate": 1.952623286174028e-05, "loss": 0.1988, "step": 2512 }, { "epoch": 0.13, "learning_rate": 1.9525740858276064e-05, "loss": 0.4525, "step": 2513 }, { "epoch": 0.13, "learning_rate": 1.9525248605678575e-05, "loss": 0.4214, "step": 2514 }, { "epoch": 0.13, "learning_rate": 1.9524756103960683e-05, "loss": 0.5013, "step": 2515 }, { "epoch": 0.13, "learning_rate": 1.9524263353135274e-05, "loss": 0.3515, "step": 2516 }, { "epoch": 0.13, "learning_rate": 1.9523770353215228e-05, "loss": 0.2418, "step": 2517 }, { "epoch": 0.13, "learning_rate": 1.9523277104213445e-05, "loss": 0.4633, "step": 2518 }, { "epoch": 0.13, "learning_rate": 1.952278360614282e-05, "loss": 0.4817, "step": 2519 }, { "epoch": 0.13, "learning_rate": 1.9522289859016262e-05, "loss": 0.5865, "step": 2520 }, { "epoch": 0.13, "learning_rate": 1.9521795862846686e-05, "loss": 0.3296, "step": 2521 }, { "epoch": 0.13, "learning_rate": 1.952130161764701e-05, "loss": 0.2937, "step": 2522 }, { "epoch": 0.13, "learning_rate": 1.9520807123430163e-05, "loss": 0.3221, "step": 2523 }, { "epoch": 0.13, "learning_rate": 1.9520312380209075e-05, "loss": 0.3352, "step": 2524 }, { "epoch": 0.13, "learning_rate": 1.9519817387996688e-05, "loss": 0.4271, "step": 2525 }, { "epoch": 0.13, "learning_rate": 1.9519322146805943e-05, "loss": 0.4789, "step": 2526 }, { "epoch": 0.13, "learning_rate": 1.951882665664979e-05, "loss": 0.5089, "step": 2527 }, { "epoch": 0.13, "learning_rate": 1.95183309175412e-05, "loss": 0.2228, "step": 2528 }, { "epoch": 0.13, "learning_rate": 1.9517834929493134e-05, "loss": 0.2878, "step": 2529 }, { "epoch": 0.13, "learning_rate": 1.9517338692518553e-05, "loss": 0.3226, "step": 2530 }, { "epoch": 0.13, "learning_rate": 1.951684220663045e-05, "loss": 0.3806, "step": 2531 }, { "epoch": 0.13, "learning_rate": 1.9516345471841805e-05, "loss": 0.3684, "step": 2532 }, { "epoch": 0.13, "learning_rate": 1.9515848488165608e-05, "loss": 0.3344, "step": 2533 }, { "epoch": 0.13, "learning_rate": 1.9515351255614854e-05, "loss": 0.3247, "step": 2534 }, { "epoch": 0.13, "learning_rate": 1.951485377420255e-05, "loss": 0.2482, "step": 2535 }, { "epoch": 0.13, "learning_rate": 1.951435604394171e-05, "loss": 0.2461, "step": 2536 }, { "epoch": 0.13, "learning_rate": 1.951385806484535e-05, "loss": 0.3653, "step": 2537 }, { "epoch": 0.13, "learning_rate": 1.9513359836926492e-05, "loss": 0.5169, "step": 2538 }, { "epoch": 0.13, "learning_rate": 1.951286136019817e-05, "loss": 0.4428, "step": 2539 }, { "epoch": 0.13, "learning_rate": 1.9512362634673418e-05, "loss": 0.3591, "step": 2540 }, { "epoch": 0.13, "learning_rate": 1.9511863660365277e-05, "loss": 0.4303, "step": 2541 }, { "epoch": 0.13, "learning_rate": 1.9511364437286803e-05, "loss": 0.3873, "step": 2542 }, { "epoch": 0.13, "learning_rate": 1.9510864965451048e-05, "loss": 0.2983, "step": 2543 }, { "epoch": 0.13, "learning_rate": 1.951036524487108e-05, "loss": 0.7015, "step": 2544 }, { "epoch": 0.13, "learning_rate": 1.9509865275559962e-05, "loss": 0.572, "step": 2545 }, { "epoch": 0.13, "learning_rate": 1.9509365057530777e-05, "loss": 0.5254, "step": 2546 }, { "epoch": 0.13, "learning_rate": 1.95088645907966e-05, "loss": 0.3111, "step": 2547 }, { "epoch": 0.13, "learning_rate": 1.9508363875370526e-05, "loss": 0.3053, "step": 2548 }, { "epoch": 0.13, "learning_rate": 1.950786291126565e-05, "loss": 0.5724, "step": 2549 }, { "epoch": 0.13, "learning_rate": 1.950736169849507e-05, "loss": 0.3807, "step": 2550 }, { "epoch": 0.13, "learning_rate": 1.95068602370719e-05, "loss": 0.2833, "step": 2551 }, { "epoch": 0.13, "learning_rate": 1.9506358527009248e-05, "loss": 0.5965, "step": 2552 }, { "epoch": 0.13, "learning_rate": 1.9505856568320244e-05, "loss": 0.4035, "step": 2553 }, { "epoch": 0.13, "learning_rate": 1.950535436101801e-05, "loss": 0.2811, "step": 2554 }, { "epoch": 0.13, "learning_rate": 1.950485190511568e-05, "loss": 0.5457, "step": 2555 }, { "epoch": 0.13, "learning_rate": 1.95043492006264e-05, "loss": 0.3756, "step": 2556 }, { "epoch": 0.13, "learning_rate": 1.950384624756332e-05, "loss": 0.3814, "step": 2557 }, { "epoch": 0.13, "learning_rate": 1.950334304593958e-05, "loss": 0.2217, "step": 2558 }, { "epoch": 0.13, "learning_rate": 1.9502839595768358e-05, "loss": 0.2122, "step": 2559 }, { "epoch": 0.13, "learning_rate": 1.9502335897062806e-05, "loss": 0.4093, "step": 2560 }, { "epoch": 0.13, "learning_rate": 1.9501831949836108e-05, "loss": 0.6813, "step": 2561 }, { "epoch": 0.13, "learning_rate": 1.950132775410144e-05, "loss": 0.192, "step": 2562 }, { "epoch": 0.13, "learning_rate": 1.950082330987199e-05, "loss": 0.4962, "step": 2563 }, { "epoch": 0.13, "learning_rate": 1.9500318617160948e-05, "loss": 0.5564, "step": 2564 }, { "epoch": 0.13, "learning_rate": 1.949981367598152e-05, "loss": 0.5031, "step": 2565 }, { "epoch": 0.13, "learning_rate": 1.9499308486346904e-05, "loss": 0.4022, "step": 2566 }, { "epoch": 0.13, "learning_rate": 1.9498803048270316e-05, "loss": 0.3773, "step": 2567 }, { "epoch": 0.13, "learning_rate": 1.9498297361764974e-05, "loss": 0.194, "step": 2568 }, { "epoch": 0.13, "learning_rate": 1.949779142684411e-05, "loss": 0.3813, "step": 2569 }, { "epoch": 0.13, "learning_rate": 1.949728524352095e-05, "loss": 0.3151, "step": 2570 }, { "epoch": 0.13, "learning_rate": 1.9496778811808732e-05, "loss": 0.2547, "step": 2571 }, { "epoch": 0.13, "learning_rate": 1.9496272131720703e-05, "loss": 0.3725, "step": 2572 }, { "epoch": 0.13, "learning_rate": 1.9495765203270114e-05, "loss": 0.2398, "step": 2573 }, { "epoch": 0.13, "learning_rate": 1.949525802647022e-05, "loss": 0.3758, "step": 2574 }, { "epoch": 0.13, "learning_rate": 1.9494750601334295e-05, "loss": 0.2289, "step": 2575 }, { "epoch": 0.13, "learning_rate": 1.9494242927875603e-05, "loss": 0.1362, "step": 2576 }, { "epoch": 0.13, "learning_rate": 1.949373500610742e-05, "loss": 0.4776, "step": 2577 }, { "epoch": 0.13, "learning_rate": 1.9493226836043036e-05, "loss": 0.5229, "step": 2578 }, { "epoch": 0.13, "learning_rate": 1.9492718417695734e-05, "loss": 0.6116, "step": 2579 }, { "epoch": 0.13, "learning_rate": 1.9492209751078816e-05, "loss": 0.4321, "step": 2580 }, { "epoch": 0.13, "learning_rate": 1.9491700836205588e-05, "loss": 0.3051, "step": 2581 }, { "epoch": 0.13, "learning_rate": 1.9491191673089354e-05, "loss": 0.5507, "step": 2582 }, { "epoch": 0.13, "learning_rate": 1.949068226174343e-05, "loss": 0.1131, "step": 2583 }, { "epoch": 0.13, "learning_rate": 1.9490172602181145e-05, "loss": 0.5643, "step": 2584 }, { "epoch": 0.13, "learning_rate": 1.9489662694415825e-05, "loss": 0.3838, "step": 2585 }, { "epoch": 0.13, "learning_rate": 1.9489152538460808e-05, "loss": 0.3616, "step": 2586 }, { "epoch": 0.13, "learning_rate": 1.9488642134329434e-05, "loss": 0.2662, "step": 2587 }, { "epoch": 0.13, "learning_rate": 1.948813148203505e-05, "loss": 0.3425, "step": 2588 }, { "epoch": 0.13, "learning_rate": 1.9487620581591017e-05, "loss": 0.5418, "step": 2589 }, { "epoch": 0.13, "learning_rate": 1.9487109433010696e-05, "loss": 0.4632, "step": 2590 }, { "epoch": 0.13, "learning_rate": 1.948659803630745e-05, "loss": 0.3744, "step": 2591 }, { "epoch": 0.13, "learning_rate": 1.9486086391494662e-05, "loss": 0.1568, "step": 2592 }, { "epoch": 0.13, "learning_rate": 1.94855744985857e-05, "loss": 0.3082, "step": 2593 }, { "epoch": 0.13, "learning_rate": 1.948506235759397e-05, "loss": 0.5652, "step": 2594 }, { "epoch": 0.13, "learning_rate": 1.9484549968532854e-05, "loss": 0.4754, "step": 2595 }, { "epoch": 0.13, "learning_rate": 1.948403733141576e-05, "loss": 0.3575, "step": 2596 }, { "epoch": 0.13, "learning_rate": 1.9483524446256082e-05, "loss": 0.4117, "step": 2597 }, { "epoch": 0.13, "learning_rate": 1.948301131306725e-05, "loss": 0.6189, "step": 2598 }, { "epoch": 0.13, "learning_rate": 1.9482497931862677e-05, "loss": 0.4066, "step": 2599 }, { "epoch": 0.13, "learning_rate": 1.948198430265579e-05, "loss": 0.3411, "step": 2600 }, { "epoch": 0.13, "learning_rate": 1.9481470425460026e-05, "loss": 0.3615, "step": 2601 }, { "epoch": 0.13, "learning_rate": 1.9480956300288814e-05, "loss": 0.5184, "step": 2602 }, { "epoch": 0.13, "learning_rate": 1.9480441927155613e-05, "loss": 0.5739, "step": 2603 }, { "epoch": 0.13, "learning_rate": 1.947992730607387e-05, "loss": 0.5246, "step": 2604 }, { "epoch": 0.13, "learning_rate": 1.9479412437057045e-05, "loss": 0.3682, "step": 2605 }, { "epoch": 0.13, "learning_rate": 1.9478897320118605e-05, "loss": 0.3249, "step": 2606 }, { "epoch": 0.13, "learning_rate": 1.9478381955272016e-05, "loss": 0.3737, "step": 2607 }, { "epoch": 0.13, "learning_rate": 1.9477866342530768e-05, "loss": 0.3443, "step": 2608 }, { "epoch": 0.13, "learning_rate": 1.9477350481908334e-05, "loss": 0.1317, "step": 2609 }, { "epoch": 0.13, "learning_rate": 1.947683437341821e-05, "loss": 0.3286, "step": 2610 }, { "epoch": 0.13, "learning_rate": 1.9476318017073902e-05, "loss": 0.55, "step": 2611 }, { "epoch": 0.13, "learning_rate": 1.9475801412888904e-05, "loss": 0.3234, "step": 2612 }, { "epoch": 0.13, "learning_rate": 1.947528456087673e-05, "loss": 0.3684, "step": 2613 }, { "epoch": 0.13, "learning_rate": 1.9474767461050897e-05, "loss": 0.2631, "step": 2614 }, { "epoch": 0.13, "learning_rate": 1.9474250113424936e-05, "loss": 0.4884, "step": 2615 }, { "epoch": 0.13, "learning_rate": 1.947373251801237e-05, "loss": 0.2373, "step": 2616 }, { "epoch": 0.13, "learning_rate": 1.947321467482674e-05, "loss": 0.1145, "step": 2617 }, { "epoch": 0.13, "learning_rate": 1.9472696583881586e-05, "loss": 0.3917, "step": 2618 }, { "epoch": 0.13, "learning_rate": 1.947217824519046e-05, "loss": 0.687, "step": 2619 }, { "epoch": 0.13, "learning_rate": 1.9471659658766918e-05, "loss": 0.3178, "step": 2620 }, { "epoch": 0.13, "learning_rate": 1.9471140824624525e-05, "loss": 0.4965, "step": 2621 }, { "epoch": 0.13, "learning_rate": 1.9470621742776846e-05, "loss": 0.3915, "step": 2622 }, { "epoch": 0.13, "learning_rate": 1.9470102413237462e-05, "loss": 0.4643, "step": 2623 }, { "epoch": 0.13, "learning_rate": 1.9469582836019955e-05, "loss": 0.1943, "step": 2624 }, { "epoch": 0.13, "learning_rate": 1.946906301113791e-05, "loss": 0.4135, "step": 2625 }, { "epoch": 0.13, "learning_rate": 1.946854293860492e-05, "loss": 0.4471, "step": 2626 }, { "epoch": 0.13, "learning_rate": 1.9468022618434595e-05, "loss": 0.7916, "step": 2627 }, { "epoch": 0.13, "learning_rate": 1.9467502050640537e-05, "loss": 0.4555, "step": 2628 }, { "epoch": 0.13, "learning_rate": 1.9466981235236364e-05, "loss": 0.4274, "step": 2629 }, { "epoch": 0.13, "learning_rate": 1.9466460172235698e-05, "loss": 0.4042, "step": 2630 }, { "epoch": 0.13, "learning_rate": 1.9465938861652163e-05, "loss": 0.4279, "step": 2631 }, { "epoch": 0.13, "learning_rate": 1.9465417303499398e-05, "loss": 0.4127, "step": 2632 }, { "epoch": 0.13, "learning_rate": 1.9464895497791038e-05, "loss": 0.2478, "step": 2633 }, { "epoch": 0.13, "learning_rate": 1.9464373444540735e-05, "loss": 0.4241, "step": 2634 }, { "epoch": 0.13, "learning_rate": 1.946385114376214e-05, "loss": 0.512, "step": 2635 }, { "epoch": 0.13, "learning_rate": 1.9463328595468912e-05, "loss": 0.3643, "step": 2636 }, { "epoch": 0.13, "learning_rate": 1.946280579967472e-05, "loss": 0.4223, "step": 2637 }, { "epoch": 0.13, "learning_rate": 1.946228275639324e-05, "loss": 0.3138, "step": 2638 }, { "epoch": 0.13, "learning_rate": 1.9461759465638143e-05, "loss": 0.3944, "step": 2639 }, { "epoch": 0.13, "learning_rate": 1.9461235927423125e-05, "loss": 0.1754, "step": 2640 }, { "epoch": 0.13, "learning_rate": 1.9460712141761868e-05, "loss": 0.3695, "step": 2641 }, { "epoch": 0.13, "learning_rate": 1.946018810866808e-05, "loss": 0.739, "step": 2642 }, { "epoch": 0.13, "learning_rate": 1.945966382815546e-05, "loss": 0.4354, "step": 2643 }, { "epoch": 0.13, "learning_rate": 1.9459139300237724e-05, "loss": 0.5775, "step": 2644 }, { "epoch": 0.13, "learning_rate": 1.9458614524928586e-05, "loss": 0.6581, "step": 2645 }, { "epoch": 0.13, "learning_rate": 1.9458089502241778e-05, "loss": 0.4184, "step": 2646 }, { "epoch": 0.13, "learning_rate": 1.9457564232191023e-05, "loss": 0.3856, "step": 2647 }, { "epoch": 0.13, "learning_rate": 1.9457038714790065e-05, "loss": 0.2571, "step": 2648 }, { "epoch": 0.13, "learning_rate": 1.9456512950052647e-05, "loss": 0.333, "step": 2649 }, { "epoch": 0.13, "learning_rate": 1.945598693799252e-05, "loss": 0.3254, "step": 2650 }, { "epoch": 0.13, "learning_rate": 1.9455460678623435e-05, "loss": 0.3825, "step": 2651 }, { "epoch": 0.13, "learning_rate": 1.9454934171959163e-05, "loss": 0.3527, "step": 2652 }, { "epoch": 0.13, "learning_rate": 1.945440741801347e-05, "loss": 0.3209, "step": 2653 }, { "epoch": 0.13, "learning_rate": 1.945388041680013e-05, "loss": 0.2846, "step": 2654 }, { "epoch": 0.13, "learning_rate": 1.9453353168332942e-05, "loss": 0.3717, "step": 2655 }, { "epoch": 0.13, "learning_rate": 1.9452825672625675e-05, "loss": 0.2976, "step": 2656 }, { "epoch": 0.13, "learning_rate": 1.9452297929692133e-05, "loss": 0.5618, "step": 2657 }, { "epoch": 0.13, "learning_rate": 1.945176993954612e-05, "loss": 0.4908, "step": 2658 }, { "epoch": 0.13, "learning_rate": 1.9451241702201447e-05, "loss": 0.3832, "step": 2659 }, { "epoch": 0.13, "learning_rate": 1.9450713217671925e-05, "loss": 0.4731, "step": 2660 }, { "epoch": 0.13, "learning_rate": 1.9450184485971376e-05, "loss": 0.5098, "step": 2661 }, { "epoch": 0.13, "learning_rate": 1.9449655507113626e-05, "loss": 0.5085, "step": 2662 }, { "epoch": 0.13, "learning_rate": 1.944912628111252e-05, "loss": 0.2451, "step": 2663 }, { "epoch": 0.13, "learning_rate": 1.9448596807981887e-05, "loss": 0.6893, "step": 2664 }, { "epoch": 0.13, "learning_rate": 1.9448067087735582e-05, "loss": 0.2639, "step": 2665 }, { "epoch": 0.13, "learning_rate": 1.9447537120387455e-05, "loss": 0.3954, "step": 2666 }, { "epoch": 0.13, "learning_rate": 1.9447006905951368e-05, "loss": 0.55, "step": 2667 }, { "epoch": 0.13, "learning_rate": 1.9446476444441195e-05, "loss": 0.1201, "step": 2668 }, { "epoch": 0.13, "learning_rate": 1.9445945735870796e-05, "loss": 0.3957, "step": 2669 }, { "epoch": 0.13, "learning_rate": 1.9445414780254062e-05, "loss": 0.2751, "step": 2670 }, { "epoch": 0.13, "learning_rate": 1.9444883577604876e-05, "loss": 0.506, "step": 2671 }, { "epoch": 0.13, "learning_rate": 1.944435212793713e-05, "loss": 0.4907, "step": 2672 }, { "epoch": 0.13, "learning_rate": 1.9443820431264724e-05, "loss": 0.5293, "step": 2673 }, { "epoch": 0.13, "learning_rate": 1.944328848760156e-05, "loss": 0.5268, "step": 2674 }, { "epoch": 0.13, "learning_rate": 1.944275629696156e-05, "loss": 0.5701, "step": 2675 }, { "epoch": 0.13, "learning_rate": 1.9442223859358633e-05, "loss": 0.4034, "step": 2676 }, { "epoch": 0.13, "learning_rate": 1.9441691174806712e-05, "loss": 0.345, "step": 2677 }, { "epoch": 0.13, "learning_rate": 1.9441158243319722e-05, "loss": 0.4234, "step": 2678 }, { "epoch": 0.13, "learning_rate": 1.9440625064911604e-05, "loss": 0.3549, "step": 2679 }, { "epoch": 0.13, "learning_rate": 1.9440091639596302e-05, "loss": 0.3333, "step": 2680 }, { "epoch": 0.13, "learning_rate": 1.943955796738777e-05, "loss": 0.406, "step": 2681 }, { "epoch": 0.13, "learning_rate": 1.943902404829996e-05, "loss": 0.3688, "step": 2682 }, { "epoch": 0.13, "learning_rate": 1.943848988234684e-05, "loss": 0.6122, "step": 2683 }, { "epoch": 0.13, "learning_rate": 1.9437955469542377e-05, "loss": 0.2588, "step": 2684 }, { "epoch": 0.13, "learning_rate": 1.943742080990055e-05, "loss": 0.404, "step": 2685 }, { "epoch": 0.13, "learning_rate": 1.943688590343535e-05, "loss": 0.3955, "step": 2686 }, { "epoch": 0.13, "learning_rate": 1.9436350750160753e-05, "loss": 0.4649, "step": 2687 }, { "epoch": 0.13, "learning_rate": 1.943581535009076e-05, "loss": 0.4993, "step": 2688 }, { "epoch": 0.13, "learning_rate": 1.9435279703239376e-05, "loss": 0.3926, "step": 2689 }, { "epoch": 0.13, "learning_rate": 1.9434743809620613e-05, "loss": 0.4615, "step": 2690 }, { "epoch": 0.13, "learning_rate": 1.9434207669248484e-05, "loss": 0.4099, "step": 2691 }, { "epoch": 0.13, "learning_rate": 1.9433671282137006e-05, "loss": 0.5258, "step": 2692 }, { "epoch": 0.13, "learning_rate": 1.9433134648300212e-05, "loss": 0.3647, "step": 2693 }, { "epoch": 0.13, "learning_rate": 1.9432597767752137e-05, "loss": 0.3721, "step": 2694 }, { "epoch": 0.13, "learning_rate": 1.9432060640506823e-05, "loss": 0.2765, "step": 2695 }, { "epoch": 0.13, "learning_rate": 1.9431523266578316e-05, "loss": 0.2959, "step": 2696 }, { "epoch": 0.13, "learning_rate": 1.943098564598067e-05, "loss": 0.3235, "step": 2697 }, { "epoch": 0.13, "learning_rate": 1.9430447778727948e-05, "loss": 0.308, "step": 2698 }, { "epoch": 0.13, "learning_rate": 1.9429909664834215e-05, "loss": 0.2472, "step": 2699 }, { "epoch": 0.13, "learning_rate": 1.9429371304313545e-05, "loss": 0.4188, "step": 2700 }, { "epoch": 0.13, "learning_rate": 1.942883269718002e-05, "loss": 0.447, "step": 2701 }, { "epoch": 0.13, "learning_rate": 1.9428293843447727e-05, "loss": 0.3833, "step": 2702 }, { "epoch": 0.13, "learning_rate": 1.9427754743130754e-05, "loss": 0.5419, "step": 2703 }, { "epoch": 0.14, "learning_rate": 1.9427215396243206e-05, "loss": 0.3715, "step": 2704 }, { "epoch": 0.14, "learning_rate": 1.9426675802799186e-05, "loss": 0.5194, "step": 2705 }, { "epoch": 0.14, "learning_rate": 1.942613596281281e-05, "loss": 0.17, "step": 2706 }, { "epoch": 0.14, "learning_rate": 1.942559587629819e-05, "loss": 0.613, "step": 2707 }, { "epoch": 0.14, "learning_rate": 1.9425055543269455e-05, "loss": 0.3314, "step": 2708 }, { "epoch": 0.14, "learning_rate": 1.942451496374074e-05, "loss": 0.355, "step": 2709 }, { "epoch": 0.14, "learning_rate": 1.942397413772618e-05, "loss": 0.3766, "step": 2710 }, { "epoch": 0.14, "learning_rate": 1.9423433065239916e-05, "loss": 0.2501, "step": 2711 }, { "epoch": 0.14, "learning_rate": 1.9422891746296107e-05, "loss": 0.4087, "step": 2712 }, { "epoch": 0.14, "learning_rate": 1.9422350180908903e-05, "loss": 0.3166, "step": 2713 }, { "epoch": 0.14, "learning_rate": 1.942180836909247e-05, "loss": 0.3599, "step": 2714 }, { "epoch": 0.14, "learning_rate": 1.9421266310860983e-05, "loss": 0.4198, "step": 2715 }, { "epoch": 0.14, "learning_rate": 1.9420724006228614e-05, "loss": 0.4931, "step": 2716 }, { "epoch": 0.14, "learning_rate": 1.9420181455209545e-05, "loss": 0.6221, "step": 2717 }, { "epoch": 0.14, "learning_rate": 1.941963865781797e-05, "loss": 0.6129, "step": 2718 }, { "epoch": 0.14, "learning_rate": 1.9419095614068087e-05, "loss": 0.3594, "step": 2719 }, { "epoch": 0.14, "learning_rate": 1.941855232397409e-05, "loss": 0.3033, "step": 2720 }, { "epoch": 0.14, "learning_rate": 1.9418008787550195e-05, "loss": 0.539, "step": 2721 }, { "epoch": 0.14, "learning_rate": 1.9417465004810614e-05, "loss": 0.5296, "step": 2722 }, { "epoch": 0.14, "learning_rate": 1.9416920975769577e-05, "loss": 0.5293, "step": 2723 }, { "epoch": 0.14, "learning_rate": 1.94163767004413e-05, "loss": 0.2876, "step": 2724 }, { "epoch": 0.14, "learning_rate": 1.941583217884002e-05, "loss": 0.3894, "step": 2725 }, { "epoch": 0.14, "learning_rate": 1.9415287410979988e-05, "loss": 0.3449, "step": 2726 }, { "epoch": 0.14, "learning_rate": 1.9414742396875445e-05, "loss": 0.4855, "step": 2727 }, { "epoch": 0.14, "learning_rate": 1.9414197136540644e-05, "loss": 0.3288, "step": 2728 }, { "epoch": 0.14, "learning_rate": 1.9413651629989848e-05, "loss": 0.3481, "step": 2729 }, { "epoch": 0.14, "learning_rate": 1.941310587723732e-05, "loss": 0.4313, "step": 2730 }, { "epoch": 0.14, "learning_rate": 1.9412559878297342e-05, "loss": 0.3337, "step": 2731 }, { "epoch": 0.14, "learning_rate": 1.9412013633184184e-05, "loss": 0.1917, "step": 2732 }, { "epoch": 0.14, "learning_rate": 1.9411467141912142e-05, "loss": 0.3188, "step": 2733 }, { "epoch": 0.14, "learning_rate": 1.94109204044955e-05, "loss": 0.5155, "step": 2734 }, { "epoch": 0.14, "learning_rate": 1.9410373420948564e-05, "loss": 0.4966, "step": 2735 }, { "epoch": 0.14, "learning_rate": 1.9409826191285633e-05, "loss": 0.6885, "step": 2736 }, { "epoch": 0.14, "learning_rate": 1.9409278715521023e-05, "loss": 0.5268, "step": 2737 }, { "epoch": 0.14, "learning_rate": 1.9408730993669056e-05, "loss": 0.3824, "step": 2738 }, { "epoch": 0.14, "learning_rate": 1.940818302574405e-05, "loss": 0.2384, "step": 2739 }, { "epoch": 0.14, "learning_rate": 1.940763481176034e-05, "loss": 0.1449, "step": 2740 }, { "epoch": 0.14, "learning_rate": 1.9407086351732265e-05, "loss": 0.4212, "step": 2741 }, { "epoch": 0.14, "learning_rate": 1.9406537645674163e-05, "loss": 0.5424, "step": 2742 }, { "epoch": 0.14, "learning_rate": 1.9405988693600394e-05, "loss": 0.2245, "step": 2743 }, { "epoch": 0.14, "learning_rate": 1.9405439495525308e-05, "loss": 0.4417, "step": 2744 }, { "epoch": 0.14, "learning_rate": 1.940489005146327e-05, "loss": 0.4489, "step": 2745 }, { "epoch": 0.14, "learning_rate": 1.9404340361428653e-05, "loss": 0.2591, "step": 2746 }, { "epoch": 0.14, "learning_rate": 1.940379042543583e-05, "loss": 0.2914, "step": 2747 }, { "epoch": 0.14, "learning_rate": 1.9403240243499187e-05, "loss": 0.5626, "step": 2748 }, { "epoch": 0.14, "learning_rate": 1.940268981563311e-05, "loss": 0.3424, "step": 2749 }, { "epoch": 0.14, "learning_rate": 1.9402139141851996e-05, "loss": 0.3973, "step": 2750 }, { "epoch": 0.14, "learning_rate": 1.9401588222170248e-05, "loss": 0.4491, "step": 2751 }, { "epoch": 0.14, "learning_rate": 1.9401037056602276e-05, "loss": 0.3453, "step": 2752 }, { "epoch": 0.14, "learning_rate": 1.9400485645162486e-05, "loss": 0.5139, "step": 2753 }, { "epoch": 0.14, "learning_rate": 1.9399933987865315e-05, "loss": 0.2738, "step": 2754 }, { "epoch": 0.14, "learning_rate": 1.9399382084725172e-05, "loss": 0.2381, "step": 2755 }, { "epoch": 0.14, "learning_rate": 1.9398829935756512e-05, "loss": 0.4593, "step": 2756 }, { "epoch": 0.14, "learning_rate": 1.939827754097376e-05, "loss": 0.4372, "step": 2757 }, { "epoch": 0.14, "learning_rate": 1.939772490039137e-05, "loss": 0.2778, "step": 2758 }, { "epoch": 0.14, "learning_rate": 1.9397172014023794e-05, "loss": 0.3712, "step": 2759 }, { "epoch": 0.14, "learning_rate": 1.9396618881885493e-05, "loss": 0.3902, "step": 2760 }, { "epoch": 0.14, "learning_rate": 1.939606550399093e-05, "loss": 0.4689, "step": 2761 }, { "epoch": 0.14, "learning_rate": 1.9395511880354577e-05, "loss": 0.4995, "step": 2762 }, { "epoch": 0.14, "learning_rate": 1.939495801099092e-05, "loss": 0.4515, "step": 2763 }, { "epoch": 0.14, "learning_rate": 1.9394403895914443e-05, "loss": 0.5019, "step": 2764 }, { "epoch": 0.14, "learning_rate": 1.939384953513964e-05, "loss": 0.1294, "step": 2765 }, { "epoch": 0.14, "learning_rate": 1.9393294928681e-05, "loss": 0.486, "step": 2766 }, { "epoch": 0.14, "learning_rate": 1.939274007655304e-05, "loss": 0.2854, "step": 2767 }, { "epoch": 0.14, "learning_rate": 1.939218497877026e-05, "loss": 0.3126, "step": 2768 }, { "epoch": 0.14, "learning_rate": 1.9391629635347184e-05, "loss": 0.3777, "step": 2769 }, { "epoch": 0.14, "learning_rate": 1.9391074046298337e-05, "loss": 0.413, "step": 2770 }, { "epoch": 0.14, "learning_rate": 1.939051821163825e-05, "loss": 0.3742, "step": 2771 }, { "epoch": 0.14, "learning_rate": 1.9389962131381456e-05, "loss": 0.4847, "step": 2772 }, { "epoch": 0.14, "learning_rate": 1.9389405805542503e-05, "loss": 0.3384, "step": 2773 }, { "epoch": 0.14, "learning_rate": 1.9388849234135936e-05, "loss": 0.3636, "step": 2774 }, { "epoch": 0.14, "learning_rate": 1.9388292417176317e-05, "loss": 0.5055, "step": 2775 }, { "epoch": 0.14, "learning_rate": 1.9387735354678207e-05, "loss": 0.3035, "step": 2776 }, { "epoch": 0.14, "learning_rate": 1.9387178046656173e-05, "loss": 0.4544, "step": 2777 }, { "epoch": 0.14, "learning_rate": 1.9386620493124794e-05, "loss": 0.5701, "step": 2778 }, { "epoch": 0.14, "learning_rate": 1.9386062694098648e-05, "loss": 0.2903, "step": 2779 }, { "epoch": 0.14, "learning_rate": 1.9385504649592325e-05, "loss": 0.4369, "step": 2780 }, { "epoch": 0.14, "learning_rate": 1.9384946359620422e-05, "loss": 0.5262, "step": 2781 }, { "epoch": 0.14, "learning_rate": 1.938438782419754e-05, "loss": 0.5154, "step": 2782 }, { "epoch": 0.14, "learning_rate": 1.9383829043338287e-05, "loss": 0.3242, "step": 2783 }, { "epoch": 0.14, "learning_rate": 1.9383270017057274e-05, "loss": 0.4884, "step": 2784 }, { "epoch": 0.14, "learning_rate": 1.9382710745369123e-05, "loss": 0.2588, "step": 2785 }, { "epoch": 0.14, "learning_rate": 1.9382151228288462e-05, "loss": 0.5967, "step": 2786 }, { "epoch": 0.14, "learning_rate": 1.9381591465829923e-05, "loss": 0.3974, "step": 2787 }, { "epoch": 0.14, "learning_rate": 1.938103145800815e-05, "loss": 0.4866, "step": 2788 }, { "epoch": 0.14, "learning_rate": 1.9380471204837784e-05, "loss": 0.4502, "step": 2789 }, { "epoch": 0.14, "learning_rate": 1.937991070633348e-05, "loss": 0.2805, "step": 2790 }, { "epoch": 0.14, "learning_rate": 1.9379349962509898e-05, "loss": 0.1718, "step": 2791 }, { "epoch": 0.14, "learning_rate": 1.93787889733817e-05, "loss": 0.3911, "step": 2792 }, { "epoch": 0.14, "learning_rate": 1.937822773896356e-05, "loss": 0.2587, "step": 2793 }, { "epoch": 0.14, "learning_rate": 1.937766625927016e-05, "loss": 0.6431, "step": 2794 }, { "epoch": 0.14, "learning_rate": 1.937710453431618e-05, "loss": 0.5538, "step": 2795 }, { "epoch": 0.14, "learning_rate": 1.9376542564116314e-05, "loss": 0.4777, "step": 2796 }, { "epoch": 0.14, "learning_rate": 1.937598034868526e-05, "loss": 0.4736, "step": 2797 }, { "epoch": 0.14, "learning_rate": 1.9375417888037715e-05, "loss": 0.2193, "step": 2798 }, { "epoch": 0.14, "learning_rate": 1.9374855182188398e-05, "loss": 0.3547, "step": 2799 }, { "epoch": 0.14, "learning_rate": 1.937429223115202e-05, "loss": 0.4087, "step": 2800 }, { "epoch": 0.14, "learning_rate": 1.937372903494331e-05, "loss": 0.4976, "step": 2801 }, { "epoch": 0.14, "learning_rate": 1.9373165593576993e-05, "loss": 0.3867, "step": 2802 }, { "epoch": 0.14, "learning_rate": 1.937260190706781e-05, "loss": 0.2722, "step": 2803 }, { "epoch": 0.14, "learning_rate": 1.9372037975430492e-05, "loss": 0.3521, "step": 2804 }, { "epoch": 0.14, "learning_rate": 1.9371473798679802e-05, "loss": 0.2771, "step": 2805 }, { "epoch": 0.14, "learning_rate": 1.937090937683049e-05, "loss": 0.409, "step": 2806 }, { "epoch": 0.14, "learning_rate": 1.9370344709897314e-05, "loss": 0.5363, "step": 2807 }, { "epoch": 0.14, "learning_rate": 1.9369779797895046e-05, "loss": 0.4191, "step": 2808 }, { "epoch": 0.14, "learning_rate": 1.9369214640838462e-05, "loss": 0.4377, "step": 2809 }, { "epoch": 0.14, "learning_rate": 1.9368649238742335e-05, "loss": 0.3918, "step": 2810 }, { "epoch": 0.14, "learning_rate": 1.936808359162146e-05, "loss": 0.2773, "step": 2811 }, { "epoch": 0.14, "learning_rate": 1.936751769949063e-05, "loss": 0.2661, "step": 2812 }, { "epoch": 0.14, "learning_rate": 1.9366951562364644e-05, "loss": 0.472, "step": 2813 }, { "epoch": 0.14, "learning_rate": 1.936638518025831e-05, "loss": 0.3507, "step": 2814 }, { "epoch": 0.14, "learning_rate": 1.9365818553186436e-05, "loss": 0.2007, "step": 2815 }, { "epoch": 0.14, "learning_rate": 1.9365251681163847e-05, "loss": 0.3855, "step": 2816 }, { "epoch": 0.14, "learning_rate": 1.9364684564205364e-05, "loss": 0.3912, "step": 2817 }, { "epoch": 0.14, "learning_rate": 1.9364117202325823e-05, "loss": 0.5485, "step": 2818 }, { "epoch": 0.14, "learning_rate": 1.9363549595540062e-05, "loss": 0.4845, "step": 2819 }, { "epoch": 0.14, "learning_rate": 1.9362981743862924e-05, "loss": 0.4071, "step": 2820 }, { "epoch": 0.14, "learning_rate": 1.9362413647309265e-05, "loss": 0.3926, "step": 2821 }, { "epoch": 0.14, "learning_rate": 1.9361845305893936e-05, "loss": 0.3595, "step": 2822 }, { "epoch": 0.14, "learning_rate": 1.9361276719631804e-05, "loss": 0.3818, "step": 2823 }, { "epoch": 0.14, "learning_rate": 1.9360707888537743e-05, "loss": 0.2166, "step": 2824 }, { "epoch": 0.14, "learning_rate": 1.936013881262663e-05, "loss": 0.3839, "step": 2825 }, { "epoch": 0.14, "learning_rate": 1.9359569491913343e-05, "loss": 0.5006, "step": 2826 }, { "epoch": 0.14, "learning_rate": 1.9358999926412775e-05, "loss": 0.3348, "step": 2827 }, { "epoch": 0.14, "learning_rate": 1.935843011613982e-05, "loss": 0.3374, "step": 2828 }, { "epoch": 0.14, "learning_rate": 1.9357860061109382e-05, "loss": 0.5478, "step": 2829 }, { "epoch": 0.14, "learning_rate": 1.9357289761336376e-05, "loss": 0.4671, "step": 2830 }, { "epoch": 0.14, "learning_rate": 1.935671921683571e-05, "loss": 0.5105, "step": 2831 }, { "epoch": 0.14, "learning_rate": 1.9356148427622307e-05, "loss": 0.3486, "step": 2832 }, { "epoch": 0.14, "learning_rate": 1.9355577393711097e-05, "loss": 0.3916, "step": 2833 }, { "epoch": 0.14, "learning_rate": 1.9355006115117012e-05, "loss": 0.2839, "step": 2834 }, { "epoch": 0.14, "learning_rate": 1.9354434591854997e-05, "loss": 0.0877, "step": 2835 }, { "epoch": 0.14, "learning_rate": 1.9353862823939997e-05, "loss": 0.5268, "step": 2836 }, { "epoch": 0.14, "learning_rate": 1.9353290811386968e-05, "loss": 0.527, "step": 2837 }, { "epoch": 0.14, "learning_rate": 1.9352718554210863e-05, "loss": 0.3016, "step": 2838 }, { "epoch": 0.14, "learning_rate": 1.935214605242666e-05, "loss": 0.237, "step": 2839 }, { "epoch": 0.14, "learning_rate": 1.9351573306049323e-05, "loss": 0.3318, "step": 2840 }, { "epoch": 0.14, "learning_rate": 1.9351000315093835e-05, "loss": 0.346, "step": 2841 }, { "epoch": 0.14, "learning_rate": 1.9350427079575182e-05, "loss": 0.5119, "step": 2842 }, { "epoch": 0.14, "learning_rate": 1.9349853599508352e-05, "loss": 0.0885, "step": 2843 }, { "epoch": 0.14, "learning_rate": 1.934927987490835e-05, "loss": 0.4398, "step": 2844 }, { "epoch": 0.14, "learning_rate": 1.9348705905790183e-05, "loss": 0.5711, "step": 2845 }, { "epoch": 0.14, "learning_rate": 1.9348131692168853e-05, "loss": 0.5076, "step": 2846 }, { "epoch": 0.14, "learning_rate": 1.934755723405938e-05, "loss": 0.2957, "step": 2847 }, { "epoch": 0.14, "learning_rate": 1.9346982531476793e-05, "loss": 0.3917, "step": 2848 }, { "epoch": 0.14, "learning_rate": 1.934640758443612e-05, "loss": 0.2534, "step": 2849 }, { "epoch": 0.14, "learning_rate": 1.93458323929524e-05, "loss": 0.3966, "step": 2850 }, { "epoch": 0.14, "learning_rate": 1.9345256957040673e-05, "loss": 0.5793, "step": 2851 }, { "epoch": 0.14, "learning_rate": 1.934468127671599e-05, "loss": 0.2498, "step": 2852 }, { "epoch": 0.14, "learning_rate": 1.934410535199341e-05, "loss": 0.6982, "step": 2853 }, { "epoch": 0.14, "learning_rate": 1.934352918288799e-05, "loss": 0.5376, "step": 2854 }, { "epoch": 0.14, "learning_rate": 1.93429527694148e-05, "loss": 0.4402, "step": 2855 }, { "epoch": 0.14, "learning_rate": 1.9342376111588924e-05, "loss": 0.5831, "step": 2856 }, { "epoch": 0.14, "learning_rate": 1.9341799209425436e-05, "loss": 0.5456, "step": 2857 }, { "epoch": 0.14, "learning_rate": 1.934122206293942e-05, "loss": 0.3489, "step": 2858 }, { "epoch": 0.14, "learning_rate": 1.934064467214598e-05, "loss": 0.5249, "step": 2859 }, { "epoch": 0.14, "learning_rate": 1.9340067037060213e-05, "loss": 0.6192, "step": 2860 }, { "epoch": 0.14, "learning_rate": 1.9339489157697223e-05, "loss": 0.3916, "step": 2861 }, { "epoch": 0.14, "learning_rate": 1.9338911034072126e-05, "loss": 0.2535, "step": 2862 }, { "epoch": 0.14, "learning_rate": 1.9338332666200048e-05, "loss": 0.4681, "step": 2863 }, { "epoch": 0.14, "learning_rate": 1.933775405409611e-05, "loss": 0.4156, "step": 2864 }, { "epoch": 0.14, "learning_rate": 1.9337175197775437e-05, "loss": 0.479, "step": 2865 }, { "epoch": 0.14, "learning_rate": 1.9336596097253182e-05, "loss": 0.5474, "step": 2866 }, { "epoch": 0.14, "learning_rate": 1.9336016752544484e-05, "loss": 0.6079, "step": 2867 }, { "epoch": 0.14, "learning_rate": 1.9335437163664495e-05, "loss": 0.3592, "step": 2868 }, { "epoch": 0.14, "learning_rate": 1.9334857330628373e-05, "loss": 0.238, "step": 2869 }, { "epoch": 0.14, "learning_rate": 1.9334277253451288e-05, "loss": 0.3235, "step": 2870 }, { "epoch": 0.14, "learning_rate": 1.9333696932148405e-05, "loss": 0.3795, "step": 2871 }, { "epoch": 0.14, "learning_rate": 1.9333116366734902e-05, "loss": 0.2968, "step": 2872 }, { "epoch": 0.14, "learning_rate": 1.9332535557225968e-05, "loss": 0.2613, "step": 2873 }, { "epoch": 0.14, "learning_rate": 1.933195450363679e-05, "loss": 0.4143, "step": 2874 }, { "epoch": 0.14, "learning_rate": 1.933137320598256e-05, "loss": 0.6254, "step": 2875 }, { "epoch": 0.14, "learning_rate": 1.9330791664278487e-05, "loss": 0.2405, "step": 2876 }, { "epoch": 0.14, "learning_rate": 1.9330209878539782e-05, "loss": 0.5405, "step": 2877 }, { "epoch": 0.14, "learning_rate": 1.932962784878166e-05, "loss": 0.3712, "step": 2878 }, { "epoch": 0.14, "learning_rate": 1.9329045575019335e-05, "loss": 0.4902, "step": 2879 }, { "epoch": 0.14, "learning_rate": 1.9328463057268045e-05, "loss": 0.4343, "step": 2880 }, { "epoch": 0.14, "learning_rate": 1.9327880295543023e-05, "loss": 0.3864, "step": 2881 }, { "epoch": 0.14, "learning_rate": 1.9327297289859505e-05, "loss": 0.3652, "step": 2882 }, { "epoch": 0.14, "learning_rate": 1.9326714040232748e-05, "loss": 0.2456, "step": 2883 }, { "epoch": 0.14, "learning_rate": 1.9326130546677998e-05, "loss": 0.2689, "step": 2884 }, { "epoch": 0.14, "learning_rate": 1.932554680921052e-05, "loss": 0.3341, "step": 2885 }, { "epoch": 0.14, "learning_rate": 1.932496282784558e-05, "loss": 0.3859, "step": 2886 }, { "epoch": 0.14, "learning_rate": 1.9324378602598447e-05, "loss": 0.4955, "step": 2887 }, { "epoch": 0.14, "learning_rate": 1.9323794133484406e-05, "loss": 0.4441, "step": 2888 }, { "epoch": 0.14, "learning_rate": 1.9323209420518746e-05, "loss": 0.6116, "step": 2889 }, { "epoch": 0.14, "learning_rate": 1.932262446371675e-05, "loss": 0.6723, "step": 2890 }, { "epoch": 0.14, "learning_rate": 1.9322039263093725e-05, "loss": 0.2464, "step": 2891 }, { "epoch": 0.14, "learning_rate": 1.932145381866497e-05, "loss": 0.355, "step": 2892 }, { "epoch": 0.14, "learning_rate": 1.9320868130445797e-05, "loss": 0.4723, "step": 2893 }, { "epoch": 0.14, "learning_rate": 1.932028219845153e-05, "loss": 0.3679, "step": 2894 }, { "epoch": 0.14, "learning_rate": 1.931969602269749e-05, "loss": 0.4103, "step": 2895 }, { "epoch": 0.14, "learning_rate": 1.9319109603199003e-05, "loss": 0.4595, "step": 2896 }, { "epoch": 0.14, "learning_rate": 1.9318522939971416e-05, "loss": 0.4038, "step": 2897 }, { "epoch": 0.14, "learning_rate": 1.9317936033030062e-05, "loss": 0.3884, "step": 2898 }, { "epoch": 0.14, "learning_rate": 1.9317348882390298e-05, "loss": 0.2208, "step": 2899 }, { "epoch": 0.14, "learning_rate": 1.9316761488067473e-05, "loss": 0.376, "step": 2900 }, { "epoch": 0.14, "learning_rate": 1.931617385007696e-05, "loss": 0.2954, "step": 2901 }, { "epoch": 0.14, "learning_rate": 1.9315585968434118e-05, "loss": 0.4482, "step": 2902 }, { "epoch": 0.14, "learning_rate": 1.9314997843154328e-05, "loss": 0.6389, "step": 2903 }, { "epoch": 0.15, "learning_rate": 1.931440947425297e-05, "loss": 0.6698, "step": 2904 }, { "epoch": 0.15, "learning_rate": 1.931382086174543e-05, "loss": 0.2165, "step": 2905 }, { "epoch": 0.15, "learning_rate": 1.9313232005647105e-05, "loss": 0.3838, "step": 2906 }, { "epoch": 0.15, "learning_rate": 1.9312642905973396e-05, "loss": 0.3432, "step": 2907 }, { "epoch": 0.15, "learning_rate": 1.931205356273971e-05, "loss": 0.4434, "step": 2908 }, { "epoch": 0.15, "learning_rate": 1.931146397596146e-05, "loss": 0.3699, "step": 2909 }, { "epoch": 0.15, "learning_rate": 1.9310874145654065e-05, "loss": 0.4664, "step": 2910 }, { "epoch": 0.15, "learning_rate": 1.9310284071832948e-05, "loss": 0.3864, "step": 2911 }, { "epoch": 0.15, "learning_rate": 1.9309693754513548e-05, "loss": 0.7305, "step": 2912 }, { "epoch": 0.15, "learning_rate": 1.93091031937113e-05, "loss": 0.3571, "step": 2913 }, { "epoch": 0.15, "learning_rate": 1.9308512389441652e-05, "loss": 0.1919, "step": 2914 }, { "epoch": 0.15, "learning_rate": 1.9307921341720056e-05, "loss": 0.4424, "step": 2915 }, { "epoch": 0.15, "learning_rate": 1.9307330050561968e-05, "loss": 0.3562, "step": 2916 }, { "epoch": 0.15, "learning_rate": 1.930673851598285e-05, "loss": 0.3576, "step": 2917 }, { "epoch": 0.15, "learning_rate": 1.9306146737998178e-05, "loss": 0.2765, "step": 2918 }, { "epoch": 0.15, "learning_rate": 1.9305554716623427e-05, "loss": 0.4716, "step": 2919 }, { "epoch": 0.15, "learning_rate": 1.930496245187408e-05, "loss": 0.3174, "step": 2920 }, { "epoch": 0.15, "learning_rate": 1.9304369943765627e-05, "loss": 0.587, "step": 2921 }, { "epoch": 0.15, "learning_rate": 1.930377719231357e-05, "loss": 0.2288, "step": 2922 }, { "epoch": 0.15, "learning_rate": 1.93031841975334e-05, "loss": 0.5134, "step": 2923 }, { "epoch": 0.15, "learning_rate": 1.9302590959440633e-05, "loss": 0.3625, "step": 2924 }, { "epoch": 0.15, "learning_rate": 1.9301997478050783e-05, "loss": 0.3934, "step": 2925 }, { "epoch": 0.15, "learning_rate": 1.9301403753379377e-05, "loss": 0.4171, "step": 2926 }, { "epoch": 0.15, "learning_rate": 1.930080978544193e-05, "loss": 0.3318, "step": 2927 }, { "epoch": 0.15, "learning_rate": 1.9300215574253992e-05, "loss": 0.1899, "step": 2928 }, { "epoch": 0.15, "learning_rate": 1.9299621119831098e-05, "loss": 0.4983, "step": 2929 }, { "epoch": 0.15, "learning_rate": 1.929902642218879e-05, "loss": 0.4109, "step": 2930 }, { "epoch": 0.15, "learning_rate": 1.9298431481342626e-05, "loss": 0.3903, "step": 2931 }, { "epoch": 0.15, "learning_rate": 1.9297836297308167e-05, "loss": 0.4233, "step": 2932 }, { "epoch": 0.15, "learning_rate": 1.9297240870100976e-05, "loss": 0.4516, "step": 2933 }, { "epoch": 0.15, "learning_rate": 1.9296645199736627e-05, "loss": 0.4404, "step": 2934 }, { "epoch": 0.15, "learning_rate": 1.9296049286230702e-05, "loss": 0.4364, "step": 2935 }, { "epoch": 0.15, "learning_rate": 1.929545312959878e-05, "loss": 0.4893, "step": 2936 }, { "epoch": 0.15, "learning_rate": 1.9294856729856456e-05, "loss": 0.4877, "step": 2937 }, { "epoch": 0.15, "learning_rate": 1.929426008701933e-05, "loss": 0.3037, "step": 2938 }, { "epoch": 0.15, "learning_rate": 1.9293663201103006e-05, "loss": 0.9921, "step": 2939 }, { "epoch": 0.15, "learning_rate": 1.929306607212309e-05, "loss": 0.2777, "step": 2940 }, { "epoch": 0.15, "learning_rate": 1.9292468700095206e-05, "loss": 0.6189, "step": 2941 }, { "epoch": 0.15, "learning_rate": 1.929187108503497e-05, "loss": 0.5274, "step": 2942 }, { "epoch": 0.15, "learning_rate": 1.9291273226958022e-05, "loss": 0.4255, "step": 2943 }, { "epoch": 0.15, "learning_rate": 1.929067512587999e-05, "loss": 0.1866, "step": 2944 }, { "epoch": 0.15, "learning_rate": 1.9290076781816517e-05, "loss": 0.385, "step": 2945 }, { "epoch": 0.15, "learning_rate": 1.9289478194783255e-05, "loss": 0.2402, "step": 2946 }, { "epoch": 0.15, "learning_rate": 1.9288879364795857e-05, "loss": 0.3531, "step": 2947 }, { "epoch": 0.15, "learning_rate": 1.9288280291869984e-05, "loss": 0.3932, "step": 2948 }, { "epoch": 0.15, "learning_rate": 1.9287680976021307e-05, "loss": 0.2966, "step": 2949 }, { "epoch": 0.15, "learning_rate": 1.9287081417265498e-05, "loss": 0.4993, "step": 2950 }, { "epoch": 0.15, "learning_rate": 1.928648161561824e-05, "loss": 0.2289, "step": 2951 }, { "epoch": 0.15, "learning_rate": 1.9285881571095216e-05, "loss": 0.581, "step": 2952 }, { "epoch": 0.15, "learning_rate": 1.9285281283712125e-05, "loss": 0.4349, "step": 2953 }, { "epoch": 0.15, "learning_rate": 1.928468075348466e-05, "loss": 0.702, "step": 2954 }, { "epoch": 0.15, "learning_rate": 1.9284079980428534e-05, "loss": 0.297, "step": 2955 }, { "epoch": 0.15, "learning_rate": 1.9283478964559456e-05, "loss": 0.4525, "step": 2956 }, { "epoch": 0.15, "learning_rate": 1.928287770589314e-05, "loss": 0.2372, "step": 2957 }, { "epoch": 0.15, "learning_rate": 1.928227620444532e-05, "loss": 0.6367, "step": 2958 }, { "epoch": 0.15, "learning_rate": 1.9281674460231724e-05, "loss": 0.2689, "step": 2959 }, { "epoch": 0.15, "learning_rate": 1.9281072473268086e-05, "loss": 0.5243, "step": 2960 }, { "epoch": 0.15, "learning_rate": 1.928047024357016e-05, "loss": 0.3186, "step": 2961 }, { "epoch": 0.15, "learning_rate": 1.9279867771153684e-05, "loss": 0.3589, "step": 2962 }, { "epoch": 0.15, "learning_rate": 1.9279265056034422e-05, "loss": 0.4182, "step": 2963 }, { "epoch": 0.15, "learning_rate": 1.927866209822814e-05, "loss": 0.2705, "step": 2964 }, { "epoch": 0.15, "learning_rate": 1.9278058897750596e-05, "loss": 0.4071, "step": 2965 }, { "epoch": 0.15, "learning_rate": 1.927745545461758e-05, "loss": 0.3627, "step": 2966 }, { "epoch": 0.15, "learning_rate": 1.9276851768844865e-05, "loss": 0.4911, "step": 2967 }, { "epoch": 0.15, "learning_rate": 1.9276247840448244e-05, "loss": 0.5922, "step": 2968 }, { "epoch": 0.15, "learning_rate": 1.9275643669443508e-05, "loss": 0.2524, "step": 2969 }, { "epoch": 0.15, "learning_rate": 1.9275039255846465e-05, "loss": 0.2739, "step": 2970 }, { "epoch": 0.15, "learning_rate": 1.9274434599672915e-05, "loss": 0.4436, "step": 2971 }, { "epoch": 0.15, "learning_rate": 1.9273829700938677e-05, "loss": 0.3984, "step": 2972 }, { "epoch": 0.15, "learning_rate": 1.927322455965957e-05, "loss": 0.3665, "step": 2973 }, { "epoch": 0.15, "learning_rate": 1.927261917585142e-05, "loss": 0.3646, "step": 2974 }, { "epoch": 0.15, "learning_rate": 1.927201354953006e-05, "loss": 0.3451, "step": 2975 }, { "epoch": 0.15, "learning_rate": 1.927140768071133e-05, "loss": 0.4717, "step": 2976 }, { "epoch": 0.15, "learning_rate": 1.9270801569411073e-05, "loss": 0.552, "step": 2977 }, { "epoch": 0.15, "learning_rate": 1.9270195215645146e-05, "loss": 0.4147, "step": 2978 }, { "epoch": 0.15, "learning_rate": 1.9269588619429404e-05, "loss": 0.4006, "step": 2979 }, { "epoch": 0.15, "learning_rate": 1.9268981780779713e-05, "loss": 0.4988, "step": 2980 }, { "epoch": 0.15, "learning_rate": 1.9268374699711945e-05, "loss": 0.3147, "step": 2981 }, { "epoch": 0.15, "learning_rate": 1.9267767376241974e-05, "loss": 0.1701, "step": 2982 }, { "epoch": 0.15, "learning_rate": 1.9267159810385686e-05, "loss": 0.338, "step": 2983 }, { "epoch": 0.15, "learning_rate": 1.926655200215897e-05, "loss": 0.5406, "step": 2984 }, { "epoch": 0.15, "learning_rate": 1.9265943951577727e-05, "loss": 0.5363, "step": 2985 }, { "epoch": 0.15, "learning_rate": 1.9265335658657857e-05, "loss": 0.4192, "step": 2986 }, { "epoch": 0.15, "learning_rate": 1.9264727123415264e-05, "loss": 0.2687, "step": 2987 }, { "epoch": 0.15, "learning_rate": 1.926411834586587e-05, "loss": 0.1755, "step": 2988 }, { "epoch": 0.15, "learning_rate": 1.926350932602559e-05, "loss": 0.2872, "step": 2989 }, { "epoch": 0.15, "learning_rate": 1.9262900063910362e-05, "loss": 0.2744, "step": 2990 }, { "epoch": 0.15, "learning_rate": 1.9262290559536114e-05, "loss": 0.4003, "step": 2991 }, { "epoch": 0.15, "learning_rate": 1.9261680812918788e-05, "loss": 0.4998, "step": 2992 }, { "epoch": 0.15, "learning_rate": 1.926107082407433e-05, "loss": 0.5343, "step": 2993 }, { "epoch": 0.15, "learning_rate": 1.9260460593018693e-05, "loss": 0.3665, "step": 2994 }, { "epoch": 0.15, "learning_rate": 1.925985011976784e-05, "loss": 0.3793, "step": 2995 }, { "epoch": 0.15, "learning_rate": 1.9259239404337737e-05, "loss": 0.3953, "step": 2996 }, { "epoch": 0.15, "learning_rate": 1.9258628446744352e-05, "loss": 0.3769, "step": 2997 }, { "epoch": 0.15, "learning_rate": 1.9258017247003668e-05, "loss": 0.3707, "step": 2998 }, { "epoch": 0.15, "learning_rate": 1.9257405805131667e-05, "loss": 0.3639, "step": 2999 }, { "epoch": 0.15, "learning_rate": 1.9256794121144345e-05, "loss": 0.196, "step": 3000 }, { "epoch": 0.15, "learning_rate": 1.9256182195057696e-05, "loss": 0.6903, "step": 3001 }, { "epoch": 0.15, "learning_rate": 1.9255570026887723e-05, "loss": 0.4459, "step": 3002 }, { "epoch": 0.15, "learning_rate": 1.925495761665044e-05, "loss": 0.4057, "step": 3003 }, { "epoch": 0.15, "learning_rate": 1.925434496436186e-05, "loss": 0.3158, "step": 3004 }, { "epoch": 0.15, "learning_rate": 1.9253732070038013e-05, "loss": 0.3118, "step": 3005 }, { "epoch": 0.15, "learning_rate": 1.9253118933694924e-05, "loss": 0.2684, "step": 3006 }, { "epoch": 0.15, "learning_rate": 1.9252505555348627e-05, "loss": 0.3431, "step": 3007 }, { "epoch": 0.15, "learning_rate": 1.9251891935015164e-05, "loss": 0.5339, "step": 3008 }, { "epoch": 0.15, "learning_rate": 1.9251278072710585e-05, "loss": 0.3301, "step": 3009 }, { "epoch": 0.15, "learning_rate": 1.925066396845095e-05, "loss": 0.6724, "step": 3010 }, { "epoch": 0.15, "learning_rate": 1.9250049622252312e-05, "loss": 0.4876, "step": 3011 }, { "epoch": 0.15, "learning_rate": 1.924943503413074e-05, "loss": 0.539, "step": 3012 }, { "epoch": 0.15, "learning_rate": 1.9248820204102308e-05, "loss": 0.3949, "step": 3013 }, { "epoch": 0.15, "learning_rate": 1.9248205132183103e-05, "loss": 0.3146, "step": 3014 }, { "epoch": 0.15, "learning_rate": 1.9247589818389202e-05, "loss": 0.4915, "step": 3015 }, { "epoch": 0.15, "learning_rate": 1.92469742627367e-05, "loss": 0.4949, "step": 3016 }, { "epoch": 0.15, "learning_rate": 1.92463584652417e-05, "loss": 0.4528, "step": 3017 }, { "epoch": 0.15, "learning_rate": 1.9245742425920305e-05, "loss": 0.4955, "step": 3018 }, { "epoch": 0.15, "learning_rate": 1.9245126144788625e-05, "loss": 0.3571, "step": 3019 }, { "epoch": 0.15, "learning_rate": 1.924450962186278e-05, "loss": 0.3428, "step": 3020 }, { "epoch": 0.15, "learning_rate": 1.9243892857158896e-05, "loss": 0.4614, "step": 3021 }, { "epoch": 0.15, "learning_rate": 1.9243275850693103e-05, "loss": 0.2422, "step": 3022 }, { "epoch": 0.15, "learning_rate": 1.924265860248153e-05, "loss": 0.1964, "step": 3023 }, { "epoch": 0.15, "learning_rate": 1.9242041112540333e-05, "loss": 0.4212, "step": 3024 }, { "epoch": 0.15, "learning_rate": 1.9241423380885652e-05, "loss": 0.3958, "step": 3025 }, { "epoch": 0.15, "learning_rate": 1.924080540753365e-05, "loss": 0.5044, "step": 3026 }, { "epoch": 0.15, "learning_rate": 1.9240187192500482e-05, "loss": 0.497, "step": 3027 }, { "epoch": 0.15, "learning_rate": 1.9239568735802322e-05, "loss": 0.4445, "step": 3028 }, { "epoch": 0.15, "learning_rate": 1.9238950037455346e-05, "loss": 0.3721, "step": 3029 }, { "epoch": 0.15, "learning_rate": 1.923833109747573e-05, "loss": 0.45, "step": 3030 }, { "epoch": 0.15, "learning_rate": 1.9237711915879664e-05, "loss": 0.2054, "step": 3031 }, { "epoch": 0.15, "learning_rate": 1.923709249268334e-05, "loss": 0.2956, "step": 3032 }, { "epoch": 0.15, "learning_rate": 1.9236472827902967e-05, "loss": 0.4825, "step": 3033 }, { "epoch": 0.15, "learning_rate": 1.923585292155474e-05, "loss": 0.373, "step": 3034 }, { "epoch": 0.15, "learning_rate": 1.923523277365488e-05, "loss": 0.3731, "step": 3035 }, { "epoch": 0.15, "learning_rate": 1.9234612384219596e-05, "loss": 0.391, "step": 3036 }, { "epoch": 0.15, "learning_rate": 1.9233991753265124e-05, "loss": 0.2422, "step": 3037 }, { "epoch": 0.15, "learning_rate": 1.9233370880807694e-05, "loss": 0.603, "step": 3038 }, { "epoch": 0.15, "learning_rate": 1.923274976686354e-05, "loss": 0.3685, "step": 3039 }, { "epoch": 0.15, "learning_rate": 1.923212841144891e-05, "loss": 0.2098, "step": 3040 }, { "epoch": 0.15, "learning_rate": 1.9231506814580052e-05, "loss": 0.461, "step": 3041 }, { "epoch": 0.15, "learning_rate": 1.9230884976273226e-05, "loss": 0.2891, "step": 3042 }, { "epoch": 0.15, "learning_rate": 1.923026289654469e-05, "loss": 0.4211, "step": 3043 }, { "epoch": 0.15, "learning_rate": 1.922964057541072e-05, "loss": 0.5532, "step": 3044 }, { "epoch": 0.15, "learning_rate": 1.922901801288759e-05, "loss": 0.3867, "step": 3045 }, { "epoch": 0.15, "learning_rate": 1.922839520899158e-05, "loss": 0.5828, "step": 3046 }, { "epoch": 0.15, "learning_rate": 1.9227772163738986e-05, "loss": 0.3638, "step": 3047 }, { "epoch": 0.15, "learning_rate": 1.9227148877146092e-05, "loss": 0.374, "step": 3048 }, { "epoch": 0.15, "learning_rate": 1.9226525349229202e-05, "loss": 0.3245, "step": 3049 }, { "epoch": 0.15, "learning_rate": 1.9225901580004633e-05, "loss": 0.1534, "step": 3050 }, { "epoch": 0.15, "learning_rate": 1.9225277569488684e-05, "loss": 0.3865, "step": 3051 }, { "epoch": 0.15, "learning_rate": 1.922465331769769e-05, "loss": 0.5145, "step": 3052 }, { "epoch": 0.15, "learning_rate": 1.9224028824647968e-05, "loss": 0.2915, "step": 3053 }, { "epoch": 0.15, "learning_rate": 1.922340409035585e-05, "loss": 0.4655, "step": 3054 }, { "epoch": 0.15, "learning_rate": 1.9222779114837685e-05, "loss": 0.413, "step": 3055 }, { "epoch": 0.15, "learning_rate": 1.9222153898109808e-05, "loss": 0.3248, "step": 3056 }, { "epoch": 0.15, "learning_rate": 1.9221528440188574e-05, "loss": 0.2464, "step": 3057 }, { "epoch": 0.15, "learning_rate": 1.9220902741090345e-05, "loss": 0.4163, "step": 3058 }, { "epoch": 0.15, "learning_rate": 1.9220276800831478e-05, "loss": 0.3144, "step": 3059 }, { "epoch": 0.15, "learning_rate": 1.9219650619428352e-05, "loss": 0.4233, "step": 3060 }, { "epoch": 0.15, "learning_rate": 1.9219024196897334e-05, "loss": 0.3722, "step": 3061 }, { "epoch": 0.15, "learning_rate": 1.921839753325482e-05, "loss": 0.3673, "step": 3062 }, { "epoch": 0.15, "learning_rate": 1.921777062851719e-05, "loss": 0.2806, "step": 3063 }, { "epoch": 0.15, "learning_rate": 1.9217143482700838e-05, "loss": 0.3725, "step": 3064 }, { "epoch": 0.15, "learning_rate": 1.9216516095822176e-05, "loss": 0.3154, "step": 3065 }, { "epoch": 0.15, "learning_rate": 1.9215888467897602e-05, "loss": 0.2718, "step": 3066 }, { "epoch": 0.15, "learning_rate": 1.9215260598943535e-05, "loss": 0.4743, "step": 3067 }, { "epoch": 0.15, "learning_rate": 1.9214632488976404e-05, "loss": 0.3946, "step": 3068 }, { "epoch": 0.15, "learning_rate": 1.9214004138012625e-05, "loss": 0.1568, "step": 3069 }, { "epoch": 0.15, "learning_rate": 1.9213375546068636e-05, "loss": 0.4994, "step": 3070 }, { "epoch": 0.15, "learning_rate": 1.9212746713160876e-05, "loss": 0.4393, "step": 3071 }, { "epoch": 0.15, "learning_rate": 1.9212117639305793e-05, "loss": 0.2723, "step": 3072 }, { "epoch": 0.15, "learning_rate": 1.921148832451984e-05, "loss": 0.4753, "step": 3073 }, { "epoch": 0.15, "learning_rate": 1.9210858768819474e-05, "loss": 0.3859, "step": 3074 }, { "epoch": 0.15, "learning_rate": 1.9210228972221162e-05, "loss": 0.1863, "step": 3075 }, { "epoch": 0.15, "learning_rate": 1.9209598934741374e-05, "loss": 0.3894, "step": 3076 }, { "epoch": 0.15, "learning_rate": 1.9208968656396586e-05, "loss": 0.5188, "step": 3077 }, { "epoch": 0.15, "learning_rate": 1.920833813720329e-05, "loss": 0.2664, "step": 3078 }, { "epoch": 0.15, "learning_rate": 1.9207707377177966e-05, "loss": 0.4171, "step": 3079 }, { "epoch": 0.15, "learning_rate": 1.9207076376337113e-05, "loss": 0.2335, "step": 3080 }, { "epoch": 0.15, "learning_rate": 1.9206445134697242e-05, "loss": 0.5328, "step": 3081 }, { "epoch": 0.15, "learning_rate": 1.9205813652274855e-05, "loss": 0.5551, "step": 3082 }, { "epoch": 0.15, "learning_rate": 1.920518192908647e-05, "loss": 0.5196, "step": 3083 }, { "epoch": 0.15, "learning_rate": 1.9204549965148608e-05, "loss": 0.3797, "step": 3084 }, { "epoch": 0.15, "learning_rate": 1.9203917760477795e-05, "loss": 0.4709, "step": 3085 }, { "epoch": 0.15, "learning_rate": 1.920328531509057e-05, "loss": 0.4832, "step": 3086 }, { "epoch": 0.15, "learning_rate": 1.9202652629003472e-05, "loss": 0.2272, "step": 3087 }, { "epoch": 0.15, "learning_rate": 1.9202019702233047e-05, "loss": 0.4979, "step": 3088 }, { "epoch": 0.15, "learning_rate": 1.920138653479585e-05, "loss": 0.6156, "step": 3089 }, { "epoch": 0.15, "learning_rate": 1.9200753126708437e-05, "loss": 0.4789, "step": 3090 }, { "epoch": 0.15, "learning_rate": 1.920011947798738e-05, "loss": 0.4987, "step": 3091 }, { "epoch": 0.15, "learning_rate": 1.919948558864925e-05, "loss": 0.2282, "step": 3092 }, { "epoch": 0.15, "learning_rate": 1.9198851458710617e-05, "loss": 0.3542, "step": 3093 }, { "epoch": 0.15, "learning_rate": 1.9198217088188078e-05, "loss": 0.5234, "step": 3094 }, { "epoch": 0.15, "learning_rate": 1.9197582477098214e-05, "loss": 0.4155, "step": 3095 }, { "epoch": 0.15, "learning_rate": 1.9196947625457634e-05, "loss": 0.3236, "step": 3096 }, { "epoch": 0.15, "learning_rate": 1.919631253328293e-05, "loss": 0.5775, "step": 3097 }, { "epoch": 0.15, "learning_rate": 1.9195677200590714e-05, "loss": 0.3515, "step": 3098 }, { "epoch": 0.15, "learning_rate": 1.919504162739761e-05, "loss": 0.5448, "step": 3099 }, { "epoch": 0.15, "learning_rate": 1.9194405813720233e-05, "loss": 0.2997, "step": 3100 }, { "epoch": 0.15, "learning_rate": 1.9193769759575214e-05, "loss": 0.2899, "step": 3101 }, { "epoch": 0.15, "learning_rate": 1.919313346497919e-05, "loss": 0.575, "step": 3102 }, { "epoch": 0.15, "learning_rate": 1.91924969299488e-05, "loss": 0.2959, "step": 3103 }, { "epoch": 0.15, "learning_rate": 1.9191860154500696e-05, "loss": 0.3145, "step": 3104 }, { "epoch": 0.16, "learning_rate": 1.9191223138651524e-05, "loss": 0.5169, "step": 3105 }, { "epoch": 0.16, "learning_rate": 1.9190585882417953e-05, "loss": 0.5584, "step": 3106 }, { "epoch": 0.16, "learning_rate": 1.9189948385816645e-05, "loss": 0.3825, "step": 3107 }, { "epoch": 0.16, "learning_rate": 1.9189310648864273e-05, "loss": 0.4962, "step": 3108 }, { "epoch": 0.16, "learning_rate": 1.9188672671577517e-05, "loss": 0.2627, "step": 3109 }, { "epoch": 0.16, "learning_rate": 1.918803445397306e-05, "loss": 0.3977, "step": 3110 }, { "epoch": 0.16, "learning_rate": 1.91873959960676e-05, "loss": 0.5152, "step": 3111 }, { "epoch": 0.16, "learning_rate": 1.9186757297877828e-05, "loss": 0.5031, "step": 3112 }, { "epoch": 0.16, "learning_rate": 1.9186118359420452e-05, "loss": 0.3918, "step": 3113 }, { "epoch": 0.16, "learning_rate": 1.9185479180712183e-05, "loss": 0.3985, "step": 3114 }, { "epoch": 0.16, "learning_rate": 1.9184839761769736e-05, "loss": 0.3246, "step": 3115 }, { "epoch": 0.16, "learning_rate": 1.9184200102609837e-05, "loss": 0.5437, "step": 3116 }, { "epoch": 0.16, "learning_rate": 1.918356020324921e-05, "loss": 0.4537, "step": 3117 }, { "epoch": 0.16, "learning_rate": 1.9182920063704596e-05, "loss": 0.4197, "step": 3118 }, { "epoch": 0.16, "learning_rate": 1.9182279683992736e-05, "loss": 0.4015, "step": 3119 }, { "epoch": 0.16, "learning_rate": 1.9181639064130375e-05, "loss": 0.4882, "step": 3120 }, { "epoch": 0.16, "learning_rate": 1.9180998204134272e-05, "loss": 0.4939, "step": 3121 }, { "epoch": 0.16, "learning_rate": 1.9180357104021185e-05, "loss": 0.183, "step": 3122 }, { "epoch": 0.16, "learning_rate": 1.9179715763807884e-05, "loss": 0.5801, "step": 3123 }, { "epoch": 0.16, "learning_rate": 1.9179074183511138e-05, "loss": 0.7266, "step": 3124 }, { "epoch": 0.16, "learning_rate": 1.917843236314773e-05, "loss": 0.3466, "step": 3125 }, { "epoch": 0.16, "learning_rate": 1.9177790302734448e-05, "loss": 0.54, "step": 3126 }, { "epoch": 0.16, "learning_rate": 1.9177148002288074e-05, "loss": 0.4414, "step": 3127 }, { "epoch": 0.16, "learning_rate": 1.917650546182542e-05, "loss": 0.3756, "step": 3128 }, { "epoch": 0.16, "learning_rate": 1.9175862681363284e-05, "loss": 0.2385, "step": 3129 }, { "epoch": 0.16, "learning_rate": 1.9175219660918478e-05, "loss": 0.3322, "step": 3130 }, { "epoch": 0.16, "learning_rate": 1.917457640050782e-05, "loss": 0.1996, "step": 3131 }, { "epoch": 0.16, "learning_rate": 1.9173932900148127e-05, "loss": 0.2873, "step": 3132 }, { "epoch": 0.16, "learning_rate": 1.917328915985624e-05, "loss": 0.3533, "step": 3133 }, { "epoch": 0.16, "learning_rate": 1.917264517964899e-05, "loss": 0.4808, "step": 3134 }, { "epoch": 0.16, "learning_rate": 1.9172000959543218e-05, "loss": 0.3482, "step": 3135 }, { "epoch": 0.16, "learning_rate": 1.9171356499555774e-05, "loss": 0.4371, "step": 3136 }, { "epoch": 0.16, "learning_rate": 1.9170711799703514e-05, "loss": 0.2477, "step": 3137 }, { "epoch": 0.16, "learning_rate": 1.9170066860003297e-05, "loss": 0.3736, "step": 3138 }, { "epoch": 0.16, "learning_rate": 1.9169421680471993e-05, "loss": 0.4837, "step": 3139 }, { "epoch": 0.16, "learning_rate": 1.916877626112647e-05, "loss": 0.3705, "step": 3140 }, { "epoch": 0.16, "learning_rate": 1.9168130601983618e-05, "loss": 0.3759, "step": 3141 }, { "epoch": 0.16, "learning_rate": 1.9167484703060316e-05, "loss": 0.2339, "step": 3142 }, { "epoch": 0.16, "learning_rate": 1.9166838564373463e-05, "loss": 0.6469, "step": 3143 }, { "epoch": 0.16, "learning_rate": 1.916619218593995e-05, "loss": 0.5011, "step": 3144 }, { "epoch": 0.16, "learning_rate": 1.9165545567776686e-05, "loss": 0.4145, "step": 3145 }, { "epoch": 0.16, "learning_rate": 1.9164898709900584e-05, "loss": 0.5423, "step": 3146 }, { "epoch": 0.16, "learning_rate": 1.916425161232856e-05, "loss": 0.4016, "step": 3147 }, { "epoch": 0.16, "learning_rate": 1.9163604275077536e-05, "loss": 0.4205, "step": 3148 }, { "epoch": 0.16, "learning_rate": 1.9162956698164446e-05, "loss": 0.5222, "step": 3149 }, { "epoch": 0.16, "learning_rate": 1.9162308881606223e-05, "loss": 0.4722, "step": 3150 }, { "epoch": 0.16, "learning_rate": 1.9161660825419814e-05, "loss": 0.2757, "step": 3151 }, { "epoch": 0.16, "learning_rate": 1.9161012529622164e-05, "loss": 0.3678, "step": 3152 }, { "epoch": 0.16, "learning_rate": 1.9160363994230234e-05, "loss": 0.227, "step": 3153 }, { "epoch": 0.16, "learning_rate": 1.9159715219260977e-05, "loss": 0.4127, "step": 3154 }, { "epoch": 0.16, "learning_rate": 1.9159066204731368e-05, "loss": 0.3341, "step": 3155 }, { "epoch": 0.16, "learning_rate": 1.915841695065838e-05, "loss": 0.544, "step": 3156 }, { "epoch": 0.16, "learning_rate": 1.9157767457058986e-05, "loss": 0.1531, "step": 3157 }, { "epoch": 0.16, "learning_rate": 1.9157117723950183e-05, "loss": 0.4603, "step": 3158 }, { "epoch": 0.16, "learning_rate": 1.915646775134896e-05, "loss": 0.3167, "step": 3159 }, { "epoch": 0.16, "learning_rate": 1.9155817539272315e-05, "loss": 0.3607, "step": 3160 }, { "epoch": 0.16, "learning_rate": 1.9155167087737254e-05, "loss": 0.4713, "step": 3161 }, { "epoch": 0.16, "learning_rate": 1.9154516396760788e-05, "loss": 0.4041, "step": 3162 }, { "epoch": 0.16, "learning_rate": 1.9153865466359936e-05, "loss": 0.5167, "step": 3163 }, { "epoch": 0.16, "learning_rate": 1.9153214296551723e-05, "loss": 0.241, "step": 3164 }, { "epoch": 0.16, "learning_rate": 1.915256288735318e-05, "loss": 0.6482, "step": 3165 }, { "epoch": 0.16, "learning_rate": 1.915191123878134e-05, "loss": 0.4186, "step": 3166 }, { "epoch": 0.16, "learning_rate": 1.915125935085325e-05, "loss": 0.3182, "step": 3167 }, { "epoch": 0.16, "learning_rate": 1.9150607223585956e-05, "loss": 0.3606, "step": 3168 }, { "epoch": 0.16, "learning_rate": 1.9149954856996517e-05, "loss": 0.4838, "step": 3169 }, { "epoch": 0.16, "learning_rate": 1.914930225110199e-05, "loss": 0.4692, "step": 3170 }, { "epoch": 0.16, "learning_rate": 1.914864940591945e-05, "loss": 0.1395, "step": 3171 }, { "epoch": 0.16, "learning_rate": 1.9147996321465967e-05, "loss": 0.2255, "step": 3172 }, { "epoch": 0.16, "learning_rate": 1.914734299775862e-05, "loss": 0.22, "step": 3173 }, { "epoch": 0.16, "learning_rate": 1.9146689434814498e-05, "loss": 0.499, "step": 3174 }, { "epoch": 0.16, "learning_rate": 1.9146035632650696e-05, "loss": 0.5839, "step": 3175 }, { "epoch": 0.16, "learning_rate": 1.9145381591284308e-05, "loss": 0.371, "step": 3176 }, { "epoch": 0.16, "learning_rate": 1.9144727310732445e-05, "loss": 0.3761, "step": 3177 }, { "epoch": 0.16, "learning_rate": 1.9144072791012216e-05, "loss": 0.4355, "step": 3178 }, { "epoch": 0.16, "learning_rate": 1.914341803214074e-05, "loss": 0.567, "step": 3179 }, { "epoch": 0.16, "learning_rate": 1.9142763034135138e-05, "loss": 0.6102, "step": 3180 }, { "epoch": 0.16, "learning_rate": 1.9142107797012548e-05, "loss": 0.4316, "step": 3181 }, { "epoch": 0.16, "learning_rate": 1.9141452320790098e-05, "loss": 0.358, "step": 3182 }, { "epoch": 0.16, "learning_rate": 1.914079660548494e-05, "loss": 0.1405, "step": 3183 }, { "epoch": 0.16, "learning_rate": 1.914014065111422e-05, "loss": 0.2744, "step": 3184 }, { "epoch": 0.16, "learning_rate": 1.913948445769509e-05, "loss": 0.4361, "step": 3185 }, { "epoch": 0.16, "learning_rate": 1.9138828025244716e-05, "loss": 0.3745, "step": 3186 }, { "epoch": 0.16, "learning_rate": 1.913817135378026e-05, "loss": 0.3432, "step": 3187 }, { "epoch": 0.16, "learning_rate": 1.9137514443318907e-05, "loss": 0.4925, "step": 3188 }, { "epoch": 0.16, "learning_rate": 1.913685729387783e-05, "loss": 0.4888, "step": 3189 }, { "epoch": 0.16, "learning_rate": 1.9136199905474218e-05, "loss": 0.5925, "step": 3190 }, { "epoch": 0.16, "learning_rate": 1.9135542278125262e-05, "loss": 0.2899, "step": 3191 }, { "epoch": 0.16, "learning_rate": 1.9134884411848166e-05, "loss": 0.2144, "step": 3192 }, { "epoch": 0.16, "learning_rate": 1.913422630666013e-05, "loss": 0.4521, "step": 3193 }, { "epoch": 0.16, "learning_rate": 1.913356796257837e-05, "loss": 0.4472, "step": 3194 }, { "epoch": 0.16, "learning_rate": 1.9132909379620098e-05, "loss": 0.3243, "step": 3195 }, { "epoch": 0.16, "learning_rate": 1.913225055780255e-05, "loss": 0.3408, "step": 3196 }, { "epoch": 0.16, "learning_rate": 1.9131591497142946e-05, "loss": 0.3652, "step": 3197 }, { "epoch": 0.16, "learning_rate": 1.9130932197658527e-05, "loss": 0.0823, "step": 3198 }, { "epoch": 0.16, "learning_rate": 1.913027265936654e-05, "loss": 0.4055, "step": 3199 }, { "epoch": 0.16, "learning_rate": 1.9129612882284222e-05, "loss": 0.341, "step": 3200 }, { "epoch": 0.16, "learning_rate": 1.9128952866428844e-05, "loss": 0.3602, "step": 3201 }, { "epoch": 0.16, "learning_rate": 1.9128292611817657e-05, "loss": 0.4062, "step": 3202 }, { "epoch": 0.16, "learning_rate": 1.9127632118467936e-05, "loss": 0.1638, "step": 3203 }, { "epoch": 0.16, "learning_rate": 1.9126971386396948e-05, "loss": 0.4165, "step": 3204 }, { "epoch": 0.16, "learning_rate": 1.9126310415621982e-05, "loss": 0.3687, "step": 3205 }, { "epoch": 0.16, "learning_rate": 1.9125649206160314e-05, "loss": 0.3477, "step": 3206 }, { "epoch": 0.16, "learning_rate": 1.9124987758029248e-05, "loss": 0.6236, "step": 3207 }, { "epoch": 0.16, "learning_rate": 1.9124326071246076e-05, "loss": 0.3195, "step": 3208 }, { "epoch": 0.16, "learning_rate": 1.912366414582811e-05, "loss": 0.5276, "step": 3209 }, { "epoch": 0.16, "learning_rate": 1.9123001981792656e-05, "loss": 0.3708, "step": 3210 }, { "epoch": 0.16, "learning_rate": 1.9122339579157034e-05, "loss": 0.6214, "step": 3211 }, { "epoch": 0.16, "learning_rate": 1.912167693793857e-05, "loss": 0.2875, "step": 3212 }, { "epoch": 0.16, "learning_rate": 1.912101405815459e-05, "loss": 0.536, "step": 3213 }, { "epoch": 0.16, "learning_rate": 1.9120350939822433e-05, "loss": 0.5162, "step": 3214 }, { "epoch": 0.16, "learning_rate": 1.911968758295945e-05, "loss": 0.4899, "step": 3215 }, { "epoch": 0.16, "learning_rate": 1.9119023987582975e-05, "loss": 0.2114, "step": 3216 }, { "epoch": 0.16, "learning_rate": 1.9118360153710375e-05, "loss": 0.4372, "step": 3217 }, { "epoch": 0.16, "learning_rate": 1.9117696081359006e-05, "loss": 0.4816, "step": 3218 }, { "epoch": 0.16, "learning_rate": 1.9117031770546243e-05, "loss": 0.2863, "step": 3219 }, { "epoch": 0.16, "learning_rate": 1.911636722128945e-05, "loss": 0.5272, "step": 3220 }, { "epoch": 0.16, "learning_rate": 1.9115702433606014e-05, "loss": 0.3747, "step": 3221 }, { "epoch": 0.16, "learning_rate": 1.9115037407513318e-05, "loss": 0.2358, "step": 3222 }, { "epoch": 0.16, "learning_rate": 1.911437214302876e-05, "loss": 0.5788, "step": 3223 }, { "epoch": 0.16, "learning_rate": 1.9113706640169737e-05, "loss": 0.4681, "step": 3224 }, { "epoch": 0.16, "learning_rate": 1.9113040898953655e-05, "loss": 0.3541, "step": 3225 }, { "epoch": 0.16, "learning_rate": 1.911237491939792e-05, "loss": 0.4749, "step": 3226 }, { "epoch": 0.16, "learning_rate": 1.9111708701519955e-05, "loss": 0.408, "step": 3227 }, { "epoch": 0.16, "learning_rate": 1.9111042245337183e-05, "loss": 0.2755, "step": 3228 }, { "epoch": 0.16, "learning_rate": 1.9110375550867038e-05, "loss": 0.4041, "step": 3229 }, { "epoch": 0.16, "learning_rate": 1.9109708618126948e-05, "loss": 0.2953, "step": 3230 }, { "epoch": 0.16, "learning_rate": 1.9109041447134363e-05, "loss": 0.4904, "step": 3231 }, { "epoch": 0.16, "learning_rate": 1.910837403790673e-05, "loss": 0.3125, "step": 3232 }, { "epoch": 0.16, "learning_rate": 1.9107706390461504e-05, "loss": 0.2464, "step": 3233 }, { "epoch": 0.16, "learning_rate": 1.9107038504816145e-05, "loss": 0.6422, "step": 3234 }, { "epoch": 0.16, "learning_rate": 1.9106370380988123e-05, "loss": 0.3671, "step": 3235 }, { "epoch": 0.16, "learning_rate": 1.910570201899491e-05, "loss": 0.3107, "step": 3236 }, { "epoch": 0.16, "learning_rate": 1.9105033418853992e-05, "loss": 0.4397, "step": 3237 }, { "epoch": 0.16, "learning_rate": 1.9104364580582844e-05, "loss": 0.1784, "step": 3238 }, { "epoch": 0.16, "learning_rate": 1.910369550419897e-05, "loss": 0.4455, "step": 3239 }, { "epoch": 0.16, "learning_rate": 1.9103026189719857e-05, "loss": 0.663, "step": 3240 }, { "epoch": 0.16, "learning_rate": 1.9102356637163023e-05, "loss": 0.5349, "step": 3241 }, { "epoch": 0.16, "learning_rate": 1.910168684654597e-05, "loss": 0.2606, "step": 3242 }, { "epoch": 0.16, "learning_rate": 1.9101016817886218e-05, "loss": 0.2138, "step": 3243 }, { "epoch": 0.16, "learning_rate": 1.9100346551201293e-05, "loss": 0.313, "step": 3244 }, { "epoch": 0.16, "learning_rate": 1.9099676046508723e-05, "loss": 0.3778, "step": 3245 }, { "epoch": 0.16, "learning_rate": 1.9099005303826044e-05, "loss": 0.2414, "step": 3246 }, { "epoch": 0.16, "learning_rate": 1.90983343231708e-05, "loss": 0.2193, "step": 3247 }, { "epoch": 0.16, "learning_rate": 1.9097663104560536e-05, "loss": 0.5276, "step": 3248 }, { "epoch": 0.16, "learning_rate": 1.909699164801281e-05, "loss": 0.4158, "step": 3249 }, { "epoch": 0.16, "learning_rate": 1.9096319953545186e-05, "loss": 0.3167, "step": 3250 }, { "epoch": 0.16, "learning_rate": 1.9095648021175223e-05, "loss": 0.4382, "step": 3251 }, { "epoch": 0.16, "learning_rate": 1.9094975850920504e-05, "loss": 0.2864, "step": 3252 }, { "epoch": 0.16, "learning_rate": 1.90943034427986e-05, "loss": 0.3421, "step": 3253 }, { "epoch": 0.16, "learning_rate": 1.9093630796827104e-05, "loss": 0.332, "step": 3254 }, { "epoch": 0.16, "learning_rate": 1.90929579130236e-05, "loss": 0.2067, "step": 3255 }, { "epoch": 0.16, "learning_rate": 1.9092284791405695e-05, "loss": 0.3923, "step": 3256 }, { "epoch": 0.16, "learning_rate": 1.909161143199099e-05, "loss": 0.2989, "step": 3257 }, { "epoch": 0.16, "learning_rate": 1.9090937834797094e-05, "loss": 0.4061, "step": 3258 }, { "epoch": 0.16, "learning_rate": 1.9090263999841626e-05, "loss": 0.493, "step": 3259 }, { "epoch": 0.16, "learning_rate": 1.908958992714221e-05, "loss": 0.2864, "step": 3260 }, { "epoch": 0.16, "learning_rate": 1.9088915616716477e-05, "loss": 0.6455, "step": 3261 }, { "epoch": 0.16, "learning_rate": 1.908824106858206e-05, "loss": 0.3716, "step": 3262 }, { "epoch": 0.16, "learning_rate": 1.9087566282756596e-05, "loss": 0.2218, "step": 3263 }, { "epoch": 0.16, "learning_rate": 1.908689125925774e-05, "loss": 0.2882, "step": 3264 }, { "epoch": 0.16, "learning_rate": 1.9086215998103152e-05, "loss": 0.3376, "step": 3265 }, { "epoch": 0.16, "learning_rate": 1.908554049931048e-05, "loss": 0.321, "step": 3266 }, { "epoch": 0.16, "learning_rate": 1.9084864762897397e-05, "loss": 0.4084, "step": 3267 }, { "epoch": 0.16, "learning_rate": 1.9084188788881574e-05, "loss": 0.212, "step": 3268 }, { "epoch": 0.16, "learning_rate": 1.9083512577280693e-05, "loss": 0.4293, "step": 3269 }, { "epoch": 0.16, "learning_rate": 1.9082836128112437e-05, "loss": 0.2557, "step": 3270 }, { "epoch": 0.16, "learning_rate": 1.9082159441394496e-05, "loss": 0.5933, "step": 3271 }, { "epoch": 0.16, "learning_rate": 1.9081482517144573e-05, "loss": 0.4155, "step": 3272 }, { "epoch": 0.16, "learning_rate": 1.908080535538037e-05, "loss": 0.4361, "step": 3273 }, { "epoch": 0.16, "learning_rate": 1.9080127956119594e-05, "loss": 0.429, "step": 3274 }, { "epoch": 0.16, "learning_rate": 1.9079450319379965e-05, "loss": 0.3955, "step": 3275 }, { "epoch": 0.16, "learning_rate": 1.9078772445179207e-05, "loss": 0.199, "step": 3276 }, { "epoch": 0.16, "learning_rate": 1.9078094333535045e-05, "loss": 0.4122, "step": 3277 }, { "epoch": 0.16, "learning_rate": 1.9077415984465218e-05, "loss": 0.542, "step": 3278 }, { "epoch": 0.16, "learning_rate": 1.9076737397987464e-05, "loss": 0.2218, "step": 3279 }, { "epoch": 0.16, "learning_rate": 1.9076058574119528e-05, "loss": 0.4978, "step": 3280 }, { "epoch": 0.16, "learning_rate": 1.907537951287917e-05, "loss": 0.6917, "step": 3281 }, { "epoch": 0.16, "learning_rate": 1.9074700214284152e-05, "loss": 0.386, "step": 3282 }, { "epoch": 0.16, "learning_rate": 1.907402067835223e-05, "loss": 0.5676, "step": 3283 }, { "epoch": 0.16, "learning_rate": 1.9073340905101184e-05, "loss": 0.304, "step": 3284 }, { "epoch": 0.16, "learning_rate": 1.907266089454879e-05, "loss": 0.5953, "step": 3285 }, { "epoch": 0.16, "learning_rate": 1.907198064671283e-05, "loss": 0.0821, "step": 3286 }, { "epoch": 0.16, "learning_rate": 1.9071300161611106e-05, "loss": 0.5343, "step": 3287 }, { "epoch": 0.16, "learning_rate": 1.9070619439261402e-05, "loss": 0.4909, "step": 3288 }, { "epoch": 0.16, "learning_rate": 1.906993847968153e-05, "loss": 0.2476, "step": 3289 }, { "epoch": 0.16, "learning_rate": 1.9069257282889295e-05, "loss": 0.1671, "step": 3290 }, { "epoch": 0.16, "learning_rate": 1.9068575848902516e-05, "loss": 0.5549, "step": 3291 }, { "epoch": 0.16, "learning_rate": 1.906789417773901e-05, "loss": 0.2403, "step": 3292 }, { "epoch": 0.16, "learning_rate": 1.906721226941661e-05, "loss": 0.393, "step": 3293 }, { "epoch": 0.16, "learning_rate": 1.906653012395315e-05, "loss": 0.4381, "step": 3294 }, { "epoch": 0.16, "learning_rate": 1.906584774136647e-05, "loss": 0.3481, "step": 3295 }, { "epoch": 0.16, "learning_rate": 1.9065165121674415e-05, "loss": 0.3213, "step": 3296 }, { "epoch": 0.16, "learning_rate": 1.906448226489484e-05, "loss": 0.3001, "step": 3297 }, { "epoch": 0.16, "learning_rate": 1.9063799171045604e-05, "loss": 0.3575, "step": 3298 }, { "epoch": 0.16, "learning_rate": 1.9063115840144572e-05, "loss": 0.4141, "step": 3299 }, { "epoch": 0.16, "learning_rate": 1.9062432272209613e-05, "loss": 0.4787, "step": 3300 }, { "epoch": 0.16, "learning_rate": 1.9061748467258612e-05, "loss": 0.4801, "step": 3301 }, { "epoch": 0.16, "learning_rate": 1.9061064425309447e-05, "loss": 0.3911, "step": 3302 }, { "epoch": 0.16, "learning_rate": 1.906038014638001e-05, "loss": 0.302, "step": 3303 }, { "epoch": 0.16, "learning_rate": 1.9059695630488194e-05, "loss": 0.5743, "step": 3304 }, { "epoch": 0.17, "learning_rate": 1.905901087765191e-05, "loss": 0.537, "step": 3305 }, { "epoch": 0.17, "learning_rate": 1.9058325887889058e-05, "loss": 0.3004, "step": 3306 }, { "epoch": 0.17, "learning_rate": 1.905764066121756e-05, "loss": 0.2944, "step": 3307 }, { "epoch": 0.17, "learning_rate": 1.905695519765533e-05, "loss": 0.4128, "step": 3308 }, { "epoch": 0.17, "learning_rate": 1.9056269497220298e-05, "loss": 0.2086, "step": 3309 }, { "epoch": 0.17, "learning_rate": 1.9055583559930402e-05, "loss": 0.613, "step": 3310 }, { "epoch": 0.17, "learning_rate": 1.905489738580358e-05, "loss": 0.2655, "step": 3311 }, { "epoch": 0.17, "learning_rate": 1.9054210974857777e-05, "loss": 0.3812, "step": 3312 }, { "epoch": 0.17, "learning_rate": 1.9053524327110943e-05, "loss": 0.6827, "step": 3313 }, { "epoch": 0.17, "learning_rate": 1.9052837442581037e-05, "loss": 0.4065, "step": 3314 }, { "epoch": 0.17, "learning_rate": 1.9052150321286028e-05, "loss": 0.3467, "step": 3315 }, { "epoch": 0.17, "learning_rate": 1.905146296324388e-05, "loss": 0.5429, "step": 3316 }, { "epoch": 0.17, "learning_rate": 1.9050775368472575e-05, "loss": 0.5989, "step": 3317 }, { "epoch": 0.17, "learning_rate": 1.9050087536990096e-05, "loss": 0.3533, "step": 3318 }, { "epoch": 0.17, "learning_rate": 1.904939946881443e-05, "loss": 0.4623, "step": 3319 }, { "epoch": 0.17, "learning_rate": 1.9048711163963572e-05, "loss": 0.3752, "step": 3320 }, { "epoch": 0.17, "learning_rate": 1.9048022622455526e-05, "loss": 0.4713, "step": 3321 }, { "epoch": 0.17, "learning_rate": 1.90473338443083e-05, "loss": 0.3619, "step": 3322 }, { "epoch": 0.17, "learning_rate": 1.9046644829539906e-05, "loss": 0.3414, "step": 3323 }, { "epoch": 0.17, "learning_rate": 1.9045955578168366e-05, "loss": 0.3491, "step": 3324 }, { "epoch": 0.17, "learning_rate": 1.9045266090211708e-05, "loss": 0.3085, "step": 3325 }, { "epoch": 0.17, "learning_rate": 1.904457636568796e-05, "loss": 0.5589, "step": 3326 }, { "epoch": 0.17, "learning_rate": 1.9043886404615165e-05, "loss": 0.246, "step": 3327 }, { "epoch": 0.17, "learning_rate": 1.9043196207011362e-05, "loss": 0.4475, "step": 3328 }, { "epoch": 0.17, "learning_rate": 1.904250577289461e-05, "loss": 0.4353, "step": 3329 }, { "epoch": 0.17, "learning_rate": 1.9041815102282964e-05, "loss": 0.5263, "step": 3330 }, { "epoch": 0.17, "learning_rate": 1.9041124195194486e-05, "loss": 0.4165, "step": 3331 }, { "epoch": 0.17, "learning_rate": 1.9040433051647246e-05, "loss": 0.3731, "step": 3332 }, { "epoch": 0.17, "learning_rate": 1.903974167165932e-05, "loss": 0.3119, "step": 3333 }, { "epoch": 0.17, "learning_rate": 1.9039050055248793e-05, "loss": 0.2352, "step": 3334 }, { "epoch": 0.17, "learning_rate": 1.9038358202433748e-05, "loss": 0.5007, "step": 3335 }, { "epoch": 0.17, "learning_rate": 1.9037666113232282e-05, "loss": 0.1352, "step": 3336 }, { "epoch": 0.17, "learning_rate": 1.9036973787662497e-05, "loss": 0.2356, "step": 3337 }, { "epoch": 0.17, "learning_rate": 1.9036281225742504e-05, "loss": 0.4216, "step": 3338 }, { "epoch": 0.17, "learning_rate": 1.9035588427490405e-05, "loss": 0.4831, "step": 3339 }, { "epoch": 0.17, "learning_rate": 1.9034895392924326e-05, "loss": 0.37, "step": 3340 }, { "epoch": 0.17, "learning_rate": 1.903420212206239e-05, "loss": 0.0812, "step": 3341 }, { "epoch": 0.17, "learning_rate": 1.9033508614922733e-05, "loss": 0.4428, "step": 3342 }, { "epoch": 0.17, "learning_rate": 1.9032814871523492e-05, "loss": 0.6124, "step": 3343 }, { "epoch": 0.17, "learning_rate": 1.9032120891882805e-05, "loss": 0.3993, "step": 3344 }, { "epoch": 0.17, "learning_rate": 1.9031426676018825e-05, "loss": 0.4365, "step": 3345 }, { "epoch": 0.17, "learning_rate": 1.903073222394971e-05, "loss": 0.4355, "step": 3346 }, { "epoch": 0.17, "learning_rate": 1.9030037535693628e-05, "loss": 0.4788, "step": 3347 }, { "epoch": 0.17, "learning_rate": 1.9029342611268732e-05, "loss": 0.3919, "step": 3348 }, { "epoch": 0.17, "learning_rate": 1.9028647450693212e-05, "loss": 0.4507, "step": 3349 }, { "epoch": 0.17, "learning_rate": 1.9027952053985245e-05, "loss": 0.3368, "step": 3350 }, { "epoch": 0.17, "learning_rate": 1.9027256421163012e-05, "loss": 0.2878, "step": 3351 }, { "epoch": 0.17, "learning_rate": 1.902656055224471e-05, "loss": 0.4356, "step": 3352 }, { "epoch": 0.17, "learning_rate": 1.9025864447248543e-05, "loss": 0.2856, "step": 3353 }, { "epoch": 0.17, "learning_rate": 1.902516810619271e-05, "loss": 0.2883, "step": 3354 }, { "epoch": 0.17, "learning_rate": 1.9024471529095432e-05, "loss": 0.4014, "step": 3355 }, { "epoch": 0.17, "learning_rate": 1.902377471597492e-05, "loss": 0.2097, "step": 3356 }, { "epoch": 0.17, "learning_rate": 1.9023077666849393e-05, "loss": 0.5264, "step": 3357 }, { "epoch": 0.17, "learning_rate": 1.9022380381737093e-05, "loss": 0.3765, "step": 3358 }, { "epoch": 0.17, "learning_rate": 1.9021682860656248e-05, "loss": 0.3256, "step": 3359 }, { "epoch": 0.17, "learning_rate": 1.9020985103625108e-05, "loss": 0.3566, "step": 3360 }, { "epoch": 0.17, "learning_rate": 1.9020287110661916e-05, "loss": 0.3546, "step": 3361 }, { "epoch": 0.17, "learning_rate": 1.9019588881784927e-05, "loss": 0.5603, "step": 3362 }, { "epoch": 0.17, "learning_rate": 1.9018890417012407e-05, "loss": 0.2472, "step": 3363 }, { "epoch": 0.17, "learning_rate": 1.901819171636262e-05, "loss": 0.4974, "step": 3364 }, { "epoch": 0.17, "learning_rate": 1.901749277985384e-05, "loss": 0.2839, "step": 3365 }, { "epoch": 0.17, "learning_rate": 1.9016793607504347e-05, "loss": 0.5413, "step": 3366 }, { "epoch": 0.17, "learning_rate": 1.9016094199332425e-05, "loss": 0.4624, "step": 3367 }, { "epoch": 0.17, "learning_rate": 1.9015394555356374e-05, "loss": 0.4773, "step": 3368 }, { "epoch": 0.17, "learning_rate": 1.901469467559448e-05, "loss": 0.5287, "step": 3369 }, { "epoch": 0.17, "learning_rate": 1.901399456006506e-05, "loss": 0.4663, "step": 3370 }, { "epoch": 0.17, "learning_rate": 1.9013294208786415e-05, "loss": 0.3386, "step": 3371 }, { "epoch": 0.17, "learning_rate": 1.9012593621776867e-05, "loss": 0.6376, "step": 3372 }, { "epoch": 0.17, "learning_rate": 1.9011892799054736e-05, "loss": 0.5147, "step": 3373 }, { "epoch": 0.17, "learning_rate": 1.9011191740638352e-05, "loss": 0.397, "step": 3374 }, { "epoch": 0.17, "learning_rate": 1.901049044654605e-05, "loss": 0.3696, "step": 3375 }, { "epoch": 0.17, "learning_rate": 1.9009788916796173e-05, "loss": 0.613, "step": 3376 }, { "epoch": 0.17, "learning_rate": 1.9009087151407067e-05, "loss": 0.3483, "step": 3377 }, { "epoch": 0.17, "learning_rate": 1.9008385150397088e-05, "loss": 0.4945, "step": 3378 }, { "epoch": 0.17, "learning_rate": 1.900768291378459e-05, "loss": 0.3505, "step": 3379 }, { "epoch": 0.17, "learning_rate": 1.9006980441587947e-05, "loss": 0.3215, "step": 3380 }, { "epoch": 0.17, "learning_rate": 1.900627773382553e-05, "loss": 0.4251, "step": 3381 }, { "epoch": 0.17, "learning_rate": 1.900557479051571e-05, "loss": 0.492, "step": 3382 }, { "epoch": 0.17, "learning_rate": 1.900487161167688e-05, "loss": 0.1368, "step": 3383 }, { "epoch": 0.17, "learning_rate": 1.9004168197327424e-05, "loss": 0.3557, "step": 3384 }, { "epoch": 0.17, "learning_rate": 1.9003464547485743e-05, "loss": 0.3476, "step": 3385 }, { "epoch": 0.17, "learning_rate": 1.9002760662170244e-05, "loss": 0.4371, "step": 3386 }, { "epoch": 0.17, "learning_rate": 1.9002056541399325e-05, "loss": 0.2695, "step": 3387 }, { "epoch": 0.17, "learning_rate": 1.9001352185191407e-05, "loss": 0.3634, "step": 3388 }, { "epoch": 0.17, "learning_rate": 1.900064759356492e-05, "loss": 0.4367, "step": 3389 }, { "epoch": 0.17, "learning_rate": 1.8999942766538277e-05, "loss": 0.2754, "step": 3390 }, { "epoch": 0.17, "learning_rate": 1.899923770412992e-05, "loss": 0.4099, "step": 3391 }, { "epoch": 0.17, "learning_rate": 1.899853240635829e-05, "loss": 0.4913, "step": 3392 }, { "epoch": 0.17, "learning_rate": 1.899782687324183e-05, "loss": 0.3369, "step": 3393 }, { "epoch": 0.17, "learning_rate": 1.899712110479899e-05, "loss": 0.4886, "step": 3394 }, { "epoch": 0.17, "learning_rate": 1.8996415101048236e-05, "loss": 0.348, "step": 3395 }, { "epoch": 0.17, "learning_rate": 1.899570886200803e-05, "loss": 0.2966, "step": 3396 }, { "epoch": 0.17, "learning_rate": 1.8995002387696835e-05, "loss": 0.3015, "step": 3397 }, { "epoch": 0.17, "learning_rate": 1.899429567813314e-05, "loss": 0.3492, "step": 3398 }, { "epoch": 0.17, "learning_rate": 1.8993588733335414e-05, "loss": 0.2062, "step": 3399 }, { "epoch": 0.17, "learning_rate": 1.899288155332216e-05, "loss": 0.6153, "step": 3400 }, { "epoch": 0.17, "learning_rate": 1.8992174138111866e-05, "loss": 0.2925, "step": 3401 }, { "epoch": 0.17, "learning_rate": 1.8991466487723034e-05, "loss": 0.3017, "step": 3402 }, { "epoch": 0.17, "learning_rate": 1.8990758602174177e-05, "loss": 0.4132, "step": 3403 }, { "epoch": 0.17, "learning_rate": 1.89900504814838e-05, "loss": 0.3707, "step": 3404 }, { "epoch": 0.17, "learning_rate": 1.898934212567043e-05, "loss": 0.3022, "step": 3405 }, { "epoch": 0.17, "learning_rate": 1.898863353475259e-05, "loss": 0.1885, "step": 3406 }, { "epoch": 0.17, "learning_rate": 1.8987924708748812e-05, "loss": 0.3993, "step": 3407 }, { "epoch": 0.17, "learning_rate": 1.8987215647677635e-05, "loss": 0.522, "step": 3408 }, { "epoch": 0.17, "learning_rate": 1.8986506351557604e-05, "loss": 0.6372, "step": 3409 }, { "epoch": 0.17, "learning_rate": 1.898579682040727e-05, "loss": 0.2583, "step": 3410 }, { "epoch": 0.17, "learning_rate": 1.898508705424519e-05, "loss": 0.4345, "step": 3411 }, { "epoch": 0.17, "learning_rate": 1.8984377053089926e-05, "loss": 0.3498, "step": 3412 }, { "epoch": 0.17, "learning_rate": 1.8983666816960048e-05, "loss": 0.6058, "step": 3413 }, { "epoch": 0.17, "learning_rate": 1.898295634587413e-05, "loss": 0.5404, "step": 3414 }, { "epoch": 0.17, "learning_rate": 1.8982245639850758e-05, "loss": 0.5465, "step": 3415 }, { "epoch": 0.17, "learning_rate": 1.898153469890851e-05, "loss": 0.6197, "step": 3416 }, { "epoch": 0.17, "learning_rate": 1.898082352306599e-05, "loss": 0.382, "step": 3417 }, { "epoch": 0.17, "learning_rate": 1.898011211234179e-05, "loss": 0.5155, "step": 3418 }, { "epoch": 0.17, "learning_rate": 1.8979400466754524e-05, "loss": 0.2347, "step": 3419 }, { "epoch": 0.17, "learning_rate": 1.8978688586322795e-05, "loss": 0.3462, "step": 3420 }, { "epoch": 0.17, "learning_rate": 1.897797647106523e-05, "loss": 0.2445, "step": 3421 }, { "epoch": 0.17, "learning_rate": 1.897726412100045e-05, "loss": 0.3355, "step": 3422 }, { "epoch": 0.17, "learning_rate": 1.897655153614708e-05, "loss": 0.2673, "step": 3423 }, { "epoch": 0.17, "learning_rate": 1.8975838716523764e-05, "loss": 0.3642, "step": 3424 }, { "epoch": 0.17, "learning_rate": 1.8975125662149144e-05, "loss": 0.3128, "step": 3425 }, { "epoch": 0.17, "learning_rate": 1.8974412373041867e-05, "loss": 0.3367, "step": 3426 }, { "epoch": 0.17, "learning_rate": 1.8973698849220588e-05, "loss": 0.4652, "step": 3427 }, { "epoch": 0.17, "learning_rate": 1.897298509070397e-05, "loss": 0.1753, "step": 3428 }, { "epoch": 0.17, "learning_rate": 1.897227109751068e-05, "loss": 0.4179, "step": 3429 }, { "epoch": 0.17, "learning_rate": 1.897155686965939e-05, "loss": 0.608, "step": 3430 }, { "epoch": 0.17, "learning_rate": 1.8970842407168778e-05, "loss": 0.6617, "step": 3431 }, { "epoch": 0.17, "learning_rate": 1.897012771005754e-05, "loss": 0.3056, "step": 3432 }, { "epoch": 0.17, "learning_rate": 1.8969412778344353e-05, "loss": 0.417, "step": 3433 }, { "epoch": 0.17, "learning_rate": 1.8968697612047924e-05, "loss": 0.2086, "step": 3434 }, { "epoch": 0.17, "learning_rate": 1.8967982211186957e-05, "loss": 0.5192, "step": 3435 }, { "epoch": 0.17, "learning_rate": 1.8967266575780164e-05, "loss": 0.3331, "step": 3436 }, { "epoch": 0.17, "learning_rate": 1.8966550705846255e-05, "loss": 0.2576, "step": 3437 }, { "epoch": 0.17, "learning_rate": 1.8965834601403957e-05, "loss": 0.5936, "step": 3438 }, { "epoch": 0.17, "learning_rate": 1.8965118262472e-05, "loss": 0.5265, "step": 3439 }, { "epoch": 0.17, "learning_rate": 1.8964401689069113e-05, "loss": 0.4496, "step": 3440 }, { "epoch": 0.17, "learning_rate": 1.8963684881214043e-05, "loss": 0.4715, "step": 3441 }, { "epoch": 0.17, "learning_rate": 1.8962967838925535e-05, "loss": 0.4143, "step": 3442 }, { "epoch": 0.17, "learning_rate": 1.8962250562222343e-05, "loss": 0.3144, "step": 3443 }, { "epoch": 0.17, "learning_rate": 1.8961533051123227e-05, "loss": 0.6256, "step": 3444 }, { "epoch": 0.17, "learning_rate": 1.896081530564695e-05, "loss": 0.3676, "step": 3445 }, { "epoch": 0.17, "learning_rate": 1.896009732581229e-05, "loss": 0.3783, "step": 3446 }, { "epoch": 0.17, "learning_rate": 1.8959379111638013e-05, "loss": 0.3617, "step": 3447 }, { "epoch": 0.17, "learning_rate": 1.8958660663142914e-05, "loss": 0.5298, "step": 3448 }, { "epoch": 0.17, "learning_rate": 1.895794198034578e-05, "loss": 0.41, "step": 3449 }, { "epoch": 0.17, "learning_rate": 1.8957223063265404e-05, "loss": 0.5944, "step": 3450 }, { "epoch": 0.17, "learning_rate": 1.8956503911920593e-05, "loss": 0.3788, "step": 3451 }, { "epoch": 0.17, "learning_rate": 1.895578452633015e-05, "loss": 0.2646, "step": 3452 }, { "epoch": 0.17, "learning_rate": 1.8955064906512894e-05, "loss": 0.3537, "step": 3453 }, { "epoch": 0.17, "learning_rate": 1.8954345052487646e-05, "loss": 0.2001, "step": 3454 }, { "epoch": 0.17, "learning_rate": 1.895362496427323e-05, "loss": 0.3509, "step": 3455 }, { "epoch": 0.17, "learning_rate": 1.895290464188848e-05, "loss": 0.3719, "step": 3456 }, { "epoch": 0.17, "learning_rate": 1.8952184085352235e-05, "loss": 0.2437, "step": 3457 }, { "epoch": 0.17, "learning_rate": 1.895146329468334e-05, "loss": 0.5554, "step": 3458 }, { "epoch": 0.17, "learning_rate": 1.8950742269900648e-05, "loss": 0.4683, "step": 3459 }, { "epoch": 0.17, "learning_rate": 1.8950021011023013e-05, "loss": 0.3915, "step": 3460 }, { "epoch": 0.17, "learning_rate": 1.8949299518069304e-05, "loss": 0.3809, "step": 3461 }, { "epoch": 0.17, "learning_rate": 1.8948577791058388e-05, "loss": 0.5681, "step": 3462 }, { "epoch": 0.17, "learning_rate": 1.8947855830009136e-05, "loss": 0.2732, "step": 3463 }, { "epoch": 0.17, "learning_rate": 1.8947133634940438e-05, "loss": 0.2931, "step": 3464 }, { "epoch": 0.17, "learning_rate": 1.8946411205871177e-05, "loss": 0.3535, "step": 3465 }, { "epoch": 0.17, "learning_rate": 1.894568854282025e-05, "loss": 0.1549, "step": 3466 }, { "epoch": 0.17, "learning_rate": 1.8944965645806555e-05, "loss": 0.4522, "step": 3467 }, { "epoch": 0.17, "learning_rate": 1.8944242514849e-05, "loss": 0.3938, "step": 3468 }, { "epoch": 0.17, "learning_rate": 1.8943519149966497e-05, "loss": 0.3528, "step": 3469 }, { "epoch": 0.17, "learning_rate": 1.8942795551177963e-05, "loss": 0.3979, "step": 3470 }, { "epoch": 0.17, "learning_rate": 1.8942071718502325e-05, "loss": 0.3462, "step": 3471 }, { "epoch": 0.17, "learning_rate": 1.894134765195851e-05, "loss": 0.2552, "step": 3472 }, { "epoch": 0.17, "learning_rate": 1.8940623351565462e-05, "loss": 0.4996, "step": 3473 }, { "epoch": 0.17, "learning_rate": 1.893989881734212e-05, "loss": 0.3223, "step": 3474 }, { "epoch": 0.17, "learning_rate": 1.8939174049307433e-05, "loss": 0.1022, "step": 3475 }, { "epoch": 0.17, "learning_rate": 1.893844904748036e-05, "loss": 0.224, "step": 3476 }, { "epoch": 0.17, "learning_rate": 1.8937723811879856e-05, "loss": 0.43, "step": 3477 }, { "epoch": 0.17, "learning_rate": 1.8936998342524894e-05, "loss": 0.3886, "step": 3478 }, { "epoch": 0.17, "learning_rate": 1.8936272639434446e-05, "loss": 0.4408, "step": 3479 }, { "epoch": 0.17, "learning_rate": 1.893554670262749e-05, "loss": 0.3529, "step": 3480 }, { "epoch": 0.17, "learning_rate": 1.893482053212301e-05, "loss": 0.4457, "step": 3481 }, { "epoch": 0.17, "learning_rate": 1.8934094127940006e-05, "loss": 0.3161, "step": 3482 }, { "epoch": 0.17, "learning_rate": 1.8933367490097472e-05, "loss": 0.604, "step": 3483 }, { "epoch": 0.17, "learning_rate": 1.8932640618614408e-05, "loss": 0.2516, "step": 3484 }, { "epoch": 0.17, "learning_rate": 1.8931913513509834e-05, "loss": 0.5947, "step": 3485 }, { "epoch": 0.17, "learning_rate": 1.893118617480276e-05, "loss": 0.4326, "step": 3486 }, { "epoch": 0.17, "learning_rate": 1.8930458602512204e-05, "loss": 0.5411, "step": 3487 }, { "epoch": 0.17, "learning_rate": 1.8929730796657202e-05, "loss": 0.3645, "step": 3488 }, { "epoch": 0.17, "learning_rate": 1.892900275725679e-05, "loss": 0.4238, "step": 3489 }, { "epoch": 0.17, "learning_rate": 1.8928274484330002e-05, "loss": 0.3027, "step": 3490 }, { "epoch": 0.17, "learning_rate": 1.8927545977895893e-05, "loss": 0.4976, "step": 3491 }, { "epoch": 0.17, "learning_rate": 1.8926817237973506e-05, "loss": 0.4334, "step": 3492 }, { "epoch": 0.17, "learning_rate": 1.892608826458191e-05, "loss": 0.5425, "step": 3493 }, { "epoch": 0.17, "learning_rate": 1.8925359057740167e-05, "loss": 0.2947, "step": 3494 }, { "epoch": 0.17, "learning_rate": 1.8924629617467348e-05, "loss": 0.6104, "step": 3495 }, { "epoch": 0.17, "learning_rate": 1.892389994378253e-05, "loss": 0.2157, "step": 3496 }, { "epoch": 0.17, "learning_rate": 1.8923170036704794e-05, "loss": 0.6143, "step": 3497 }, { "epoch": 0.17, "learning_rate": 1.8922439896253236e-05, "loss": 0.2383, "step": 3498 }, { "epoch": 0.17, "learning_rate": 1.892170952244695e-05, "loss": 0.4869, "step": 3499 }, { "epoch": 0.17, "learning_rate": 1.8920978915305037e-05, "loss": 0.4056, "step": 3500 }, { "epoch": 0.17, "learning_rate": 1.89202480748466e-05, "loss": 0.489, "step": 3501 }, { "epoch": 0.17, "learning_rate": 1.8919517001090762e-05, "loss": 0.5486, "step": 3502 }, { "epoch": 0.17, "learning_rate": 1.8918785694056637e-05, "loss": 0.5965, "step": 3503 }, { "epoch": 0.17, "learning_rate": 1.8918054153763356e-05, "loss": 0.2417, "step": 3504 }, { "epoch": 0.18, "learning_rate": 1.8917322380230047e-05, "loss": 0.474, "step": 3505 }, { "epoch": 0.18, "learning_rate": 1.8916590373475855e-05, "loss": 0.4505, "step": 3506 }, { "epoch": 0.18, "learning_rate": 1.891585813351992e-05, "loss": 0.3498, "step": 3507 }, { "epoch": 0.18, "learning_rate": 1.8915125660381387e-05, "loss": 0.1691, "step": 3508 }, { "epoch": 0.18, "learning_rate": 1.8914392954079424e-05, "loss": 0.5159, "step": 3509 }, { "epoch": 0.18, "learning_rate": 1.8913660014633187e-05, "loss": 0.3462, "step": 3510 }, { "epoch": 0.18, "learning_rate": 1.8912926842061847e-05, "loss": 0.4215, "step": 3511 }, { "epoch": 0.18, "learning_rate": 1.891219343638458e-05, "loss": 0.4159, "step": 3512 }, { "epoch": 0.18, "learning_rate": 1.8911459797620567e-05, "loss": 0.2862, "step": 3513 }, { "epoch": 0.18, "learning_rate": 1.8910725925788997e-05, "loss": 0.4602, "step": 3514 }, { "epoch": 0.18, "learning_rate": 1.8909991820909058e-05, "loss": 0.2959, "step": 3515 }, { "epoch": 0.18, "learning_rate": 1.8909257482999953e-05, "loss": 0.3826, "step": 3516 }, { "epoch": 0.18, "learning_rate": 1.890852291208089e-05, "loss": 0.3733, "step": 3517 }, { "epoch": 0.18, "learning_rate": 1.8907788108171074e-05, "loss": 0.358, "step": 3518 }, { "epoch": 0.18, "learning_rate": 1.890705307128973e-05, "loss": 0.5002, "step": 3519 }, { "epoch": 0.18, "learning_rate": 1.890631780145608e-05, "loss": 0.5097, "step": 3520 }, { "epoch": 0.18, "learning_rate": 1.890558229868935e-05, "loss": 0.2693, "step": 3521 }, { "epoch": 0.18, "learning_rate": 1.890484656300878e-05, "loss": 0.3868, "step": 3522 }, { "epoch": 0.18, "learning_rate": 1.890411059443362e-05, "loss": 0.2674, "step": 3523 }, { "epoch": 0.18, "learning_rate": 1.89033743929831e-05, "loss": 0.2794, "step": 3524 }, { "epoch": 0.18, "learning_rate": 1.8902637958676483e-05, "loss": 0.5074, "step": 3525 }, { "epoch": 0.18, "learning_rate": 1.8901901291533035e-05, "loss": 0.3543, "step": 3526 }, { "epoch": 0.18, "learning_rate": 1.8901164391572018e-05, "loss": 0.348, "step": 3527 }, { "epoch": 0.18, "learning_rate": 1.8900427258812703e-05, "loss": 0.5583, "step": 3528 }, { "epoch": 0.18, "learning_rate": 1.889968989327437e-05, "loss": 0.3277, "step": 3529 }, { "epoch": 0.18, "learning_rate": 1.889895229497631e-05, "loss": 0.4335, "step": 3530 }, { "epoch": 0.18, "learning_rate": 1.8898214463937804e-05, "loss": 0.3711, "step": 3531 }, { "epoch": 0.18, "learning_rate": 1.8897476400178152e-05, "loss": 0.1776, "step": 3532 }, { "epoch": 0.18, "learning_rate": 1.889673810371666e-05, "loss": 0.2896, "step": 3533 }, { "epoch": 0.18, "learning_rate": 1.8895999574572634e-05, "loss": 0.3996, "step": 3534 }, { "epoch": 0.18, "learning_rate": 1.889526081276539e-05, "loss": 0.4668, "step": 3535 }, { "epoch": 0.18, "learning_rate": 1.8894521818314255e-05, "loss": 0.7535, "step": 3536 }, { "epoch": 0.18, "learning_rate": 1.8893782591238548e-05, "loss": 0.3092, "step": 3537 }, { "epoch": 0.18, "learning_rate": 1.8893043131557608e-05, "loss": 0.5767, "step": 3538 }, { "epoch": 0.18, "learning_rate": 1.889230343929077e-05, "loss": 0.1961, "step": 3539 }, { "epoch": 0.18, "learning_rate": 1.8891563514457383e-05, "loss": 0.4076, "step": 3540 }, { "epoch": 0.18, "learning_rate": 1.88908233570768e-05, "loss": 0.2183, "step": 3541 }, { "epoch": 0.18, "learning_rate": 1.8890082967168374e-05, "loss": 0.2981, "step": 3542 }, { "epoch": 0.18, "learning_rate": 1.8889342344751473e-05, "loss": 0.5225, "step": 3543 }, { "epoch": 0.18, "learning_rate": 1.888860148984547e-05, "loss": 0.4358, "step": 3544 }, { "epoch": 0.18, "learning_rate": 1.888786040246973e-05, "loss": 0.6774, "step": 3545 }, { "epoch": 0.18, "learning_rate": 1.888711908264365e-05, "loss": 0.355, "step": 3546 }, { "epoch": 0.18, "learning_rate": 1.8886377530386607e-05, "loss": 0.2161, "step": 3547 }, { "epoch": 0.18, "learning_rate": 1.8885635745717995e-05, "loss": 0.3182, "step": 3548 }, { "epoch": 0.18, "learning_rate": 1.8884893728657223e-05, "loss": 0.4478, "step": 3549 }, { "epoch": 0.18, "learning_rate": 1.888415147922369e-05, "loss": 0.2708, "step": 3550 }, { "epoch": 0.18, "learning_rate": 1.888340899743681e-05, "loss": 0.3586, "step": 3551 }, { "epoch": 0.18, "learning_rate": 1.8882666283316005e-05, "loss": 0.415, "step": 3552 }, { "epoch": 0.18, "learning_rate": 1.88819233368807e-05, "loss": 0.3462, "step": 3553 }, { "epoch": 0.18, "learning_rate": 1.888118015815032e-05, "loss": 0.2766, "step": 3554 }, { "epoch": 0.18, "learning_rate": 1.8880436747144305e-05, "loss": 0.394, "step": 3555 }, { "epoch": 0.18, "learning_rate": 1.88796931038821e-05, "loss": 0.6271, "step": 3556 }, { "epoch": 0.18, "learning_rate": 1.8878949228383153e-05, "loss": 0.166, "step": 3557 }, { "epoch": 0.18, "learning_rate": 1.8878205120666916e-05, "loss": 0.4592, "step": 3558 }, { "epoch": 0.18, "learning_rate": 1.8877460780752854e-05, "loss": 0.3883, "step": 3559 }, { "epoch": 0.18, "learning_rate": 1.887671620866043e-05, "loss": 0.3712, "step": 3560 }, { "epoch": 0.18, "learning_rate": 1.8875971404409122e-05, "loss": 0.3875, "step": 3561 }, { "epoch": 0.18, "learning_rate": 1.887522636801841e-05, "loss": 0.4307, "step": 3562 }, { "epoch": 0.18, "learning_rate": 1.8874481099507772e-05, "loss": 0.2507, "step": 3563 }, { "epoch": 0.18, "learning_rate": 1.8873735598896706e-05, "loss": 0.3, "step": 3564 }, { "epoch": 0.18, "learning_rate": 1.8872989866204708e-05, "loss": 0.3981, "step": 3565 }, { "epoch": 0.18, "learning_rate": 1.887224390145128e-05, "loss": 0.1666, "step": 3566 }, { "epoch": 0.18, "learning_rate": 1.8871497704655937e-05, "loss": 0.3777, "step": 3567 }, { "epoch": 0.18, "learning_rate": 1.887075127583819e-05, "loss": 0.3697, "step": 3568 }, { "epoch": 0.18, "learning_rate": 1.887000461501756e-05, "loss": 0.085, "step": 3569 }, { "epoch": 0.18, "learning_rate": 1.8869257722213577e-05, "loss": 0.5743, "step": 3570 }, { "epoch": 0.18, "learning_rate": 1.8868510597445777e-05, "loss": 0.3001, "step": 3571 }, { "epoch": 0.18, "learning_rate": 1.8867763240733697e-05, "loss": 0.4955, "step": 3572 }, { "epoch": 0.18, "learning_rate": 1.8867015652096885e-05, "loss": 0.454, "step": 3573 }, { "epoch": 0.18, "learning_rate": 1.8866267831554893e-05, "loss": 0.3897, "step": 3574 }, { "epoch": 0.18, "learning_rate": 1.8865519779127274e-05, "loss": 0.525, "step": 3575 }, { "epoch": 0.18, "learning_rate": 1.8864771494833602e-05, "loss": 0.3406, "step": 3576 }, { "epoch": 0.18, "learning_rate": 1.886402297869344e-05, "loss": 0.2951, "step": 3577 }, { "epoch": 0.18, "learning_rate": 1.8863274230726366e-05, "loss": 0.7244, "step": 3578 }, { "epoch": 0.18, "learning_rate": 1.8862525250951962e-05, "loss": 0.4672, "step": 3579 }, { "epoch": 0.18, "learning_rate": 1.886177603938982e-05, "loss": 0.4185, "step": 3580 }, { "epoch": 0.18, "learning_rate": 1.8861026596059534e-05, "loss": 0.434, "step": 3581 }, { "epoch": 0.18, "learning_rate": 1.8860276920980703e-05, "loss": 0.3392, "step": 3582 }, { "epoch": 0.18, "learning_rate": 1.8859527014172933e-05, "loss": 0.4089, "step": 3583 }, { "epoch": 0.18, "learning_rate": 1.8858776875655838e-05, "loss": 0.4474, "step": 3584 }, { "epoch": 0.18, "learning_rate": 1.8858026505449035e-05, "loss": 0.3257, "step": 3585 }, { "epoch": 0.18, "learning_rate": 1.8857275903572154e-05, "loss": 0.2045, "step": 3586 }, { "epoch": 0.18, "learning_rate": 1.885652507004482e-05, "loss": 0.3348, "step": 3587 }, { "epoch": 0.18, "learning_rate": 1.8855774004886674e-05, "loss": 0.5319, "step": 3588 }, { "epoch": 0.18, "learning_rate": 1.8855022708117357e-05, "loss": 0.4849, "step": 3589 }, { "epoch": 0.18, "learning_rate": 1.885427117975652e-05, "loss": 0.5087, "step": 3590 }, { "epoch": 0.18, "learning_rate": 1.8853519419823817e-05, "loss": 0.6538, "step": 3591 }, { "epoch": 0.18, "learning_rate": 1.885276742833891e-05, "loss": 0.2635, "step": 3592 }, { "epoch": 0.18, "learning_rate": 1.8852015205321463e-05, "loss": 0.3346, "step": 3593 }, { "epoch": 0.18, "learning_rate": 1.8851262750791156e-05, "loss": 0.3486, "step": 3594 }, { "epoch": 0.18, "learning_rate": 1.8850510064767665e-05, "loss": 0.4887, "step": 3595 }, { "epoch": 0.18, "learning_rate": 1.8849757147270674e-05, "loss": 0.8072, "step": 3596 }, { "epoch": 0.18, "learning_rate": 1.8849003998319876e-05, "loss": 0.2608, "step": 3597 }, { "epoch": 0.18, "learning_rate": 1.8848250617934967e-05, "loss": 0.3759, "step": 3598 }, { "epoch": 0.18, "learning_rate": 1.8847497006135653e-05, "loss": 0.258, "step": 3599 }, { "epoch": 0.18, "learning_rate": 1.8846743162941642e-05, "loss": 0.3807, "step": 3600 }, { "epoch": 0.18, "learning_rate": 1.8845989088372657e-05, "loss": 0.3156, "step": 3601 }, { "epoch": 0.18, "learning_rate": 1.8845234782448406e-05, "loss": 0.3068, "step": 3602 }, { "epoch": 0.18, "learning_rate": 1.884448024518863e-05, "loss": 0.3912, "step": 3603 }, { "epoch": 0.18, "learning_rate": 1.884372547661305e-05, "loss": 0.4443, "step": 3604 }, { "epoch": 0.18, "learning_rate": 1.8842970476741422e-05, "loss": 0.5358, "step": 3605 }, { "epoch": 0.18, "learning_rate": 1.8842215245593478e-05, "loss": 0.5806, "step": 3606 }, { "epoch": 0.18, "learning_rate": 1.8841459783188978e-05, "loss": 0.4686, "step": 3607 }, { "epoch": 0.18, "learning_rate": 1.8840704089547674e-05, "loss": 0.1194, "step": 3608 }, { "epoch": 0.18, "learning_rate": 1.8839948164689338e-05, "loss": 0.2482, "step": 3609 }, { "epoch": 0.18, "learning_rate": 1.8839192008633735e-05, "loss": 0.4317, "step": 3610 }, { "epoch": 0.18, "learning_rate": 1.883843562140064e-05, "loss": 0.5875, "step": 3611 }, { "epoch": 0.18, "learning_rate": 1.883767900300984e-05, "loss": 0.4571, "step": 3612 }, { "epoch": 0.18, "learning_rate": 1.883692215348112e-05, "loss": 0.3073, "step": 3613 }, { "epoch": 0.18, "learning_rate": 1.8836165072834275e-05, "loss": 0.329, "step": 3614 }, { "epoch": 0.18, "learning_rate": 1.8835407761089107e-05, "loss": 0.4796, "step": 3615 }, { "epoch": 0.18, "learning_rate": 1.883465021826542e-05, "loss": 0.2648, "step": 3616 }, { "epoch": 0.18, "learning_rate": 1.883389244438303e-05, "loss": 0.2224, "step": 3617 }, { "epoch": 0.18, "learning_rate": 1.8833134439461748e-05, "loss": 0.477, "step": 3618 }, { "epoch": 0.18, "learning_rate": 1.883237620352141e-05, "loss": 0.1797, "step": 3619 }, { "epoch": 0.18, "learning_rate": 1.8831617736581837e-05, "loss": 0.487, "step": 3620 }, { "epoch": 0.18, "learning_rate": 1.883085903866287e-05, "loss": 0.3228, "step": 3621 }, { "epoch": 0.18, "learning_rate": 1.8830100109784358e-05, "loss": 0.3835, "step": 3622 }, { "epoch": 0.18, "learning_rate": 1.8829340949966135e-05, "loss": 0.2663, "step": 3623 }, { "epoch": 0.18, "learning_rate": 1.8828581559228065e-05, "loss": 0.3603, "step": 3624 }, { "epoch": 0.18, "learning_rate": 1.882782193759001e-05, "loss": 0.4451, "step": 3625 }, { "epoch": 0.18, "learning_rate": 1.8827062085071833e-05, "loss": 0.3979, "step": 3626 }, { "epoch": 0.18, "learning_rate": 1.8826302001693407e-05, "loss": 0.51, "step": 3627 }, { "epoch": 0.18, "learning_rate": 1.8825541687474617e-05, "loss": 0.2996, "step": 3628 }, { "epoch": 0.18, "learning_rate": 1.8824781142435344e-05, "loss": 0.5116, "step": 3629 }, { "epoch": 0.18, "learning_rate": 1.8824020366595473e-05, "loss": 0.6946, "step": 3630 }, { "epoch": 0.18, "learning_rate": 1.882325935997491e-05, "loss": 0.3928, "step": 3631 }, { "epoch": 0.18, "learning_rate": 1.8822498122593554e-05, "loss": 0.5176, "step": 3632 }, { "epoch": 0.18, "learning_rate": 1.882173665447131e-05, "loss": 0.4084, "step": 3633 }, { "epoch": 0.18, "learning_rate": 1.8820974955628108e-05, "loss": 0.6983, "step": 3634 }, { "epoch": 0.18, "learning_rate": 1.8820213026083853e-05, "loss": 0.6077, "step": 3635 }, { "epoch": 0.18, "learning_rate": 1.8819450865858483e-05, "loss": 0.3915, "step": 3636 }, { "epoch": 0.18, "learning_rate": 1.8818688474971923e-05, "loss": 0.3038, "step": 3637 }, { "epoch": 0.18, "learning_rate": 1.8817925853444116e-05, "loss": 0.3584, "step": 3638 }, { "epoch": 0.18, "learning_rate": 1.881716300129501e-05, "loss": 0.4954, "step": 3639 }, { "epoch": 0.18, "learning_rate": 1.8816399918544553e-05, "loss": 0.4766, "step": 3640 }, { "epoch": 0.18, "learning_rate": 1.8815636605212707e-05, "loss": 0.3383, "step": 3641 }, { "epoch": 0.18, "learning_rate": 1.881487306131943e-05, "loss": 0.2486, "step": 3642 }, { "epoch": 0.18, "learning_rate": 1.881410928688469e-05, "loss": 0.3753, "step": 3643 }, { "epoch": 0.18, "learning_rate": 1.881334528192847e-05, "loss": 0.5059, "step": 3644 }, { "epoch": 0.18, "learning_rate": 1.8812581046470746e-05, "loss": 0.3315, "step": 3645 }, { "epoch": 0.18, "learning_rate": 1.881181658053151e-05, "loss": 0.2606, "step": 3646 }, { "epoch": 0.18, "learning_rate": 1.881105188413075e-05, "loss": 0.3011, "step": 3647 }, { "epoch": 0.18, "learning_rate": 1.8810286957288465e-05, "loss": 0.3014, "step": 3648 }, { "epoch": 0.18, "learning_rate": 1.880952180002467e-05, "loss": 0.517, "step": 3649 }, { "epoch": 0.18, "learning_rate": 1.8808756412359367e-05, "loss": 0.4035, "step": 3650 }, { "epoch": 0.18, "learning_rate": 1.880799079431258e-05, "loss": 0.2488, "step": 3651 }, { "epoch": 0.18, "learning_rate": 1.880722494590433e-05, "loss": 0.4197, "step": 3652 }, { "epoch": 0.18, "learning_rate": 1.880645886715464e-05, "loss": 0.4177, "step": 3653 }, { "epoch": 0.18, "learning_rate": 1.880569255808356e-05, "loss": 0.2724, "step": 3654 }, { "epoch": 0.18, "learning_rate": 1.8804926018711125e-05, "loss": 0.3711, "step": 3655 }, { "epoch": 0.18, "learning_rate": 1.880415924905738e-05, "loss": 0.4077, "step": 3656 }, { "epoch": 0.18, "learning_rate": 1.8803392249142383e-05, "loss": 0.3804, "step": 3657 }, { "epoch": 0.18, "learning_rate": 1.880262501898619e-05, "loss": 0.2665, "step": 3658 }, { "epoch": 0.18, "learning_rate": 1.880185755860887e-05, "loss": 0.4258, "step": 3659 }, { "epoch": 0.18, "learning_rate": 1.8801089868030493e-05, "loss": 0.4984, "step": 3660 }, { "epoch": 0.18, "learning_rate": 1.880032194727114e-05, "loss": 0.2459, "step": 3661 }, { "epoch": 0.18, "learning_rate": 1.879955379635089e-05, "loss": 0.6689, "step": 3662 }, { "epoch": 0.18, "learning_rate": 1.8798785415289838e-05, "loss": 0.9771, "step": 3663 }, { "epoch": 0.18, "learning_rate": 1.8798016804108074e-05, "loss": 0.4436, "step": 3664 }, { "epoch": 0.18, "learning_rate": 1.8797247962825708e-05, "loss": 0.331, "step": 3665 }, { "epoch": 0.18, "learning_rate": 1.8796478891462843e-05, "loss": 0.696, "step": 3666 }, { "epoch": 0.18, "learning_rate": 1.8795709590039592e-05, "loss": 0.6155, "step": 3667 }, { "epoch": 0.18, "learning_rate": 1.879494005857608e-05, "loss": 0.319, "step": 3668 }, { "epoch": 0.18, "learning_rate": 1.879417029709243e-05, "loss": 0.3415, "step": 3669 }, { "epoch": 0.18, "learning_rate": 1.879340030560877e-05, "loss": 0.3983, "step": 3670 }, { "epoch": 0.18, "learning_rate": 1.8792630084145248e-05, "loss": 0.3058, "step": 3671 }, { "epoch": 0.18, "learning_rate": 1.8791859632721997e-05, "loss": 0.4849, "step": 3672 }, { "epoch": 0.18, "learning_rate": 1.8791088951359174e-05, "loss": 0.4484, "step": 3673 }, { "epoch": 0.18, "learning_rate": 1.8790318040076935e-05, "loss": 0.2623, "step": 3674 }, { "epoch": 0.18, "learning_rate": 1.8789546898895442e-05, "loss": 0.4426, "step": 3675 }, { "epoch": 0.18, "learning_rate": 1.878877552783486e-05, "loss": 0.4749, "step": 3676 }, { "epoch": 0.18, "learning_rate": 1.878800392691537e-05, "loss": 0.2409, "step": 3677 }, { "epoch": 0.18, "learning_rate": 1.878723209615714e-05, "loss": 0.3918, "step": 3678 }, { "epoch": 0.18, "learning_rate": 1.878646003558037e-05, "loss": 0.5512, "step": 3679 }, { "epoch": 0.18, "learning_rate": 1.878568774520524e-05, "loss": 0.5058, "step": 3680 }, { "epoch": 0.18, "learning_rate": 1.878491522505196e-05, "loss": 0.5631, "step": 3681 }, { "epoch": 0.18, "learning_rate": 1.8784142475140723e-05, "loss": 0.383, "step": 3682 }, { "epoch": 0.18, "learning_rate": 1.878336949549175e-05, "loss": 0.1432, "step": 3683 }, { "epoch": 0.18, "learning_rate": 1.8782596286125247e-05, "loss": 0.1769, "step": 3684 }, { "epoch": 0.18, "learning_rate": 1.8781822847061445e-05, "loss": 0.3285, "step": 3685 }, { "epoch": 0.18, "learning_rate": 1.8781049178320567e-05, "loss": 0.3242, "step": 3686 }, { "epoch": 0.18, "learning_rate": 1.878027527992285e-05, "loss": 0.2376, "step": 3687 }, { "epoch": 0.18, "learning_rate": 1.8779501151888533e-05, "loss": 0.267, "step": 3688 }, { "epoch": 0.18, "learning_rate": 1.8778726794237863e-05, "loss": 0.2832, "step": 3689 }, { "epoch": 0.18, "learning_rate": 1.877795220699109e-05, "loss": 0.4898, "step": 3690 }, { "epoch": 0.18, "learning_rate": 1.8777177390168476e-05, "loss": 0.3296, "step": 3691 }, { "epoch": 0.18, "learning_rate": 1.8776402343790278e-05, "loss": 0.4476, "step": 3692 }, { "epoch": 0.18, "learning_rate": 1.877562706787678e-05, "loss": 0.4519, "step": 3693 }, { "epoch": 0.18, "learning_rate": 1.8774851562448245e-05, "loss": 0.3722, "step": 3694 }, { "epoch": 0.18, "learning_rate": 1.877407582752496e-05, "loss": 0.3071, "step": 3695 }, { "epoch": 0.18, "learning_rate": 1.8773299863127213e-05, "loss": 0.5407, "step": 3696 }, { "epoch": 0.18, "learning_rate": 1.87725236692753e-05, "loss": 0.2916, "step": 3697 }, { "epoch": 0.18, "learning_rate": 1.8771747245989525e-05, "loss": 0.619, "step": 3698 }, { "epoch": 0.18, "learning_rate": 1.8770970593290186e-05, "loss": 0.3876, "step": 3699 }, { "epoch": 0.18, "learning_rate": 1.8770193711197597e-05, "loss": 0.2283, "step": 3700 }, { "epoch": 0.18, "learning_rate": 1.876941659973208e-05, "loss": 0.4061, "step": 3701 }, { "epoch": 0.18, "learning_rate": 1.8768639258913962e-05, "loss": 0.3916, "step": 3702 }, { "epoch": 0.18, "learning_rate": 1.8767861688763564e-05, "loss": 0.5012, "step": 3703 }, { "epoch": 0.18, "learning_rate": 1.876708388930123e-05, "loss": 0.3242, "step": 3704 }, { "epoch": 0.19, "learning_rate": 1.8766305860547298e-05, "loss": 0.805, "step": 3705 }, { "epoch": 0.19, "learning_rate": 1.8765527602522117e-05, "loss": 0.2638, "step": 3706 }, { "epoch": 0.19, "learning_rate": 1.8764749115246043e-05, "loss": 0.5821, "step": 3707 }, { "epoch": 0.19, "learning_rate": 1.876397039873944e-05, "loss": 0.4936, "step": 3708 }, { "epoch": 0.19, "learning_rate": 1.8763191453022665e-05, "loss": 0.584, "step": 3709 }, { "epoch": 0.19, "learning_rate": 1.8762412278116097e-05, "loss": 0.2703, "step": 3710 }, { "epoch": 0.19, "learning_rate": 1.876163287404011e-05, "loss": 0.4364, "step": 3711 }, { "epoch": 0.19, "learning_rate": 1.8760853240815092e-05, "loss": 0.5148, "step": 3712 }, { "epoch": 0.19, "learning_rate": 1.8760073378461436e-05, "loss": 0.4162, "step": 3713 }, { "epoch": 0.19, "learning_rate": 1.875929328699953e-05, "loss": 0.2482, "step": 3714 }, { "epoch": 0.19, "learning_rate": 1.8758512966449783e-05, "loss": 0.4242, "step": 3715 }, { "epoch": 0.19, "learning_rate": 1.87577324168326e-05, "loss": 0.3678, "step": 3716 }, { "epoch": 0.19, "learning_rate": 1.8756951638168398e-05, "loss": 0.36, "step": 3717 }, { "epoch": 0.19, "learning_rate": 1.8756170630477593e-05, "loss": 0.6006, "step": 3718 }, { "epoch": 0.19, "learning_rate": 1.8755389393780613e-05, "loss": 0.333, "step": 3719 }, { "epoch": 0.19, "learning_rate": 1.875460792809789e-05, "loss": 0.3487, "step": 3720 }, { "epoch": 0.19, "learning_rate": 1.8753826233449867e-05, "loss": 0.4619, "step": 3721 }, { "epoch": 0.19, "learning_rate": 1.8753044309856983e-05, "loss": 0.6906, "step": 3722 }, { "epoch": 0.19, "learning_rate": 1.875226215733969e-05, "loss": 0.4875, "step": 3723 }, { "epoch": 0.19, "learning_rate": 1.875147977591844e-05, "loss": 0.4305, "step": 3724 }, { "epoch": 0.19, "learning_rate": 1.8750697165613706e-05, "loss": 0.3192, "step": 3725 }, { "epoch": 0.19, "learning_rate": 1.8749914326445945e-05, "loss": 0.4605, "step": 3726 }, { "epoch": 0.19, "learning_rate": 1.8749131258435635e-05, "loss": 0.47, "step": 3727 }, { "epoch": 0.19, "learning_rate": 1.8748347961603257e-05, "loss": 0.4478, "step": 3728 }, { "epoch": 0.19, "learning_rate": 1.8747564435969296e-05, "loss": 0.3406, "step": 3729 }, { "epoch": 0.19, "learning_rate": 1.8746780681554243e-05, "loss": 0.5645, "step": 3730 }, { "epoch": 0.19, "learning_rate": 1.8745996698378596e-05, "loss": 0.4665, "step": 3731 }, { "epoch": 0.19, "learning_rate": 1.8745212486462864e-05, "loss": 0.4115, "step": 3732 }, { "epoch": 0.19, "learning_rate": 1.8744428045827555e-05, "loss": 0.3246, "step": 3733 }, { "epoch": 0.19, "learning_rate": 1.8743643376493178e-05, "loss": 0.3672, "step": 3734 }, { "epoch": 0.19, "learning_rate": 1.8742858478480267e-05, "loss": 0.2809, "step": 3735 }, { "epoch": 0.19, "learning_rate": 1.874207335180934e-05, "loss": 0.6083, "step": 3736 }, { "epoch": 0.19, "learning_rate": 1.874128799650093e-05, "loss": 0.3224, "step": 3737 }, { "epoch": 0.19, "learning_rate": 1.874050241257559e-05, "loss": 0.2604, "step": 3738 }, { "epoch": 0.19, "learning_rate": 1.873971660005385e-05, "loss": 0.4209, "step": 3739 }, { "epoch": 0.19, "learning_rate": 1.8738930558956276e-05, "loss": 0.5016, "step": 3740 }, { "epoch": 0.19, "learning_rate": 1.873814428930342e-05, "loss": 0.4877, "step": 3741 }, { "epoch": 0.19, "learning_rate": 1.8737357791115837e-05, "loss": 0.4569, "step": 3742 }, { "epoch": 0.19, "learning_rate": 1.873657106441411e-05, "loss": 0.2856, "step": 3743 }, { "epoch": 0.19, "learning_rate": 1.8735784109218807e-05, "loss": 0.4407, "step": 3744 }, { "epoch": 0.19, "learning_rate": 1.8734996925550518e-05, "loss": 0.1588, "step": 3745 }, { "epoch": 0.19, "learning_rate": 1.8734209513429823e-05, "loss": 0.7126, "step": 3746 }, { "epoch": 0.19, "learning_rate": 1.873342187287731e-05, "loss": 0.1767, "step": 3747 }, { "epoch": 0.19, "learning_rate": 1.8732634003913598e-05, "loss": 0.5019, "step": 3748 }, { "epoch": 0.19, "learning_rate": 1.8731845906559277e-05, "loss": 0.5306, "step": 3749 }, { "epoch": 0.19, "learning_rate": 1.8731057580834958e-05, "loss": 0.3187, "step": 3750 }, { "epoch": 0.19, "learning_rate": 1.8730269026761272e-05, "loss": 0.4625, "step": 3751 }, { "epoch": 0.19, "learning_rate": 1.872948024435883e-05, "loss": 0.2671, "step": 3752 }, { "epoch": 0.19, "learning_rate": 1.8728691233648266e-05, "loss": 0.368, "step": 3753 }, { "epoch": 0.19, "learning_rate": 1.8727901994650214e-05, "loss": 0.2413, "step": 3754 }, { "epoch": 0.19, "learning_rate": 1.8727112527385317e-05, "loss": 0.2256, "step": 3755 }, { "epoch": 0.19, "learning_rate": 1.8726322831874223e-05, "loss": 0.1849, "step": 3756 }, { "epoch": 0.19, "learning_rate": 1.8725532908137585e-05, "loss": 0.3241, "step": 3757 }, { "epoch": 0.19, "learning_rate": 1.8724742756196065e-05, "loss": 0.2821, "step": 3758 }, { "epoch": 0.19, "learning_rate": 1.8723952376070324e-05, "loss": 0.3681, "step": 3759 }, { "epoch": 0.19, "learning_rate": 1.8723161767781032e-05, "loss": 0.3646, "step": 3760 }, { "epoch": 0.19, "learning_rate": 1.8722370931348874e-05, "loss": 0.3193, "step": 3761 }, { "epoch": 0.19, "learning_rate": 1.872157986679452e-05, "loss": 0.5272, "step": 3762 }, { "epoch": 0.19, "learning_rate": 1.8720788574138677e-05, "loss": 0.4749, "step": 3763 }, { "epoch": 0.19, "learning_rate": 1.8719997053402024e-05, "loss": 0.5332, "step": 3764 }, { "epoch": 0.19, "learning_rate": 1.8719205304605275e-05, "loss": 0.2653, "step": 3765 }, { "epoch": 0.19, "learning_rate": 1.871841332776913e-05, "loss": 0.0901, "step": 3766 }, { "epoch": 0.19, "learning_rate": 1.8717621122914302e-05, "loss": 0.3984, "step": 3767 }, { "epoch": 0.19, "learning_rate": 1.871682869006151e-05, "loss": 0.5904, "step": 3768 }, { "epoch": 0.19, "learning_rate": 1.8716036029231483e-05, "loss": 0.3947, "step": 3769 }, { "epoch": 0.19, "learning_rate": 1.8715243140444948e-05, "loss": 0.3179, "step": 3770 }, { "epoch": 0.19, "learning_rate": 1.8714450023722647e-05, "loss": 0.3088, "step": 3771 }, { "epoch": 0.19, "learning_rate": 1.8713656679085318e-05, "loss": 0.3986, "step": 3772 }, { "epoch": 0.19, "learning_rate": 1.8712863106553713e-05, "loss": 0.503, "step": 3773 }, { "epoch": 0.19, "learning_rate": 1.871206930614858e-05, "loss": 0.0857, "step": 3774 }, { "epoch": 0.19, "learning_rate": 1.871127527789069e-05, "loss": 0.4555, "step": 3775 }, { "epoch": 0.19, "learning_rate": 1.8710481021800803e-05, "loss": 0.3175, "step": 3776 }, { "epoch": 0.19, "learning_rate": 1.8709686537899694e-05, "loss": 0.6257, "step": 3777 }, { "epoch": 0.19, "learning_rate": 1.870889182620814e-05, "loss": 0.3062, "step": 3778 }, { "epoch": 0.19, "learning_rate": 1.8708096886746928e-05, "loss": 0.3622, "step": 3779 }, { "epoch": 0.19, "learning_rate": 1.8707301719536846e-05, "loss": 0.3873, "step": 3780 }, { "epoch": 0.19, "learning_rate": 1.8706506324598693e-05, "loss": 0.2896, "step": 3781 }, { "epoch": 0.19, "learning_rate": 1.870571070195327e-05, "loss": 0.6239, "step": 3782 }, { "epoch": 0.19, "learning_rate": 1.870491485162139e-05, "loss": 0.3771, "step": 3783 }, { "epoch": 0.19, "learning_rate": 1.870411877362386e-05, "loss": 0.222, "step": 3784 }, { "epoch": 0.19, "learning_rate": 1.8703322467981503e-05, "loss": 0.5714, "step": 3785 }, { "epoch": 0.19, "learning_rate": 1.8702525934715146e-05, "loss": 0.3052, "step": 3786 }, { "epoch": 0.19, "learning_rate": 1.870172917384562e-05, "loss": 0.5537, "step": 3787 }, { "epoch": 0.19, "learning_rate": 1.870093218539377e-05, "loss": 0.6389, "step": 3788 }, { "epoch": 0.19, "learning_rate": 1.870013496938043e-05, "loss": 0.8088, "step": 3789 }, { "epoch": 0.19, "learning_rate": 1.869933752582646e-05, "loss": 0.2525, "step": 3790 }, { "epoch": 0.19, "learning_rate": 1.8698539854752708e-05, "loss": 0.5612, "step": 3791 }, { "epoch": 0.19, "learning_rate": 1.869774195618004e-05, "loss": 0.3492, "step": 3792 }, { "epoch": 0.19, "learning_rate": 1.8696943830129322e-05, "loss": 0.2678, "step": 3793 }, { "epoch": 0.19, "learning_rate": 1.8696145476621426e-05, "loss": 0.4124, "step": 3794 }, { "epoch": 0.19, "learning_rate": 1.8695346895677243e-05, "loss": 0.4024, "step": 3795 }, { "epoch": 0.19, "learning_rate": 1.8694548087317645e-05, "loss": 0.331, "step": 3796 }, { "epoch": 0.19, "learning_rate": 1.869374905156353e-05, "loss": 0.444, "step": 3797 }, { "epoch": 0.19, "learning_rate": 1.8692949788435797e-05, "loss": 0.3414, "step": 3798 }, { "epoch": 0.19, "learning_rate": 1.8692150297955346e-05, "loss": 0.4351, "step": 3799 }, { "epoch": 0.19, "learning_rate": 1.869135058014309e-05, "loss": 0.5739, "step": 3800 }, { "epoch": 0.19, "learning_rate": 1.8690550635019943e-05, "loss": 0.5167, "step": 3801 }, { "epoch": 0.19, "learning_rate": 1.868975046260683e-05, "loss": 0.4567, "step": 3802 }, { "epoch": 0.19, "learning_rate": 1.8688950062924668e-05, "loss": 0.2546, "step": 3803 }, { "epoch": 0.19, "learning_rate": 1.8688149435994403e-05, "loss": 0.3966, "step": 3804 }, { "epoch": 0.19, "learning_rate": 1.8687348581836965e-05, "loss": 0.3742, "step": 3805 }, { "epoch": 0.19, "learning_rate": 1.868654750047331e-05, "loss": 0.5814, "step": 3806 }, { "epoch": 0.19, "learning_rate": 1.8685746191924376e-05, "loss": 0.2565, "step": 3807 }, { "epoch": 0.19, "learning_rate": 1.868494465621113e-05, "loss": 0.5056, "step": 3808 }, { "epoch": 0.19, "learning_rate": 1.868414289335453e-05, "loss": 0.4775, "step": 3809 }, { "epoch": 0.19, "learning_rate": 1.868334090337555e-05, "loss": 0.5042, "step": 3810 }, { "epoch": 0.19, "learning_rate": 1.8682538686295157e-05, "loss": 0.4024, "step": 3811 }, { "epoch": 0.19, "learning_rate": 1.8681736242134335e-05, "loss": 0.1897, "step": 3812 }, { "epoch": 0.19, "learning_rate": 1.868093357091408e-05, "loss": 0.1833, "step": 3813 }, { "epoch": 0.19, "learning_rate": 1.8680130672655375e-05, "loss": 0.3294, "step": 3814 }, { "epoch": 0.19, "learning_rate": 1.867932754737922e-05, "loss": 0.4538, "step": 3815 }, { "epoch": 0.19, "learning_rate": 1.867852419510662e-05, "loss": 0.606, "step": 3816 }, { "epoch": 0.19, "learning_rate": 1.8677720615858588e-05, "loss": 0.2644, "step": 3817 }, { "epoch": 0.19, "learning_rate": 1.8676916809656136e-05, "loss": 0.3814, "step": 3818 }, { "epoch": 0.19, "learning_rate": 1.8676112776520292e-05, "loss": 0.4414, "step": 3819 }, { "epoch": 0.19, "learning_rate": 1.8675308516472084e-05, "loss": 0.1221, "step": 3820 }, { "epoch": 0.19, "learning_rate": 1.8674504029532542e-05, "loss": 0.2427, "step": 3821 }, { "epoch": 0.19, "learning_rate": 1.867369931572271e-05, "loss": 0.6455, "step": 3822 }, { "epoch": 0.19, "learning_rate": 1.8672894375063633e-05, "loss": 0.2771, "step": 3823 }, { "epoch": 0.19, "learning_rate": 1.867208920757636e-05, "loss": 0.236, "step": 3824 }, { "epoch": 0.19, "learning_rate": 1.8671283813281956e-05, "loss": 0.3995, "step": 3825 }, { "epoch": 0.19, "learning_rate": 1.867047819220148e-05, "loss": 0.4129, "step": 3826 }, { "epoch": 0.19, "learning_rate": 1.8669672344356002e-05, "loss": 0.4291, "step": 3827 }, { "epoch": 0.19, "learning_rate": 1.8668866269766597e-05, "loss": 0.5022, "step": 3828 }, { "epoch": 0.19, "learning_rate": 1.8668059968454354e-05, "loss": 0.493, "step": 3829 }, { "epoch": 0.19, "learning_rate": 1.8667253440440353e-05, "loss": 0.351, "step": 3830 }, { "epoch": 0.19, "learning_rate": 1.866644668574569e-05, "loss": 0.3105, "step": 3831 }, { "epoch": 0.19, "learning_rate": 1.8665639704391467e-05, "loss": 0.3771, "step": 3832 }, { "epoch": 0.19, "learning_rate": 1.8664832496398785e-05, "loss": 0.3004, "step": 3833 }, { "epoch": 0.19, "learning_rate": 1.866402506178876e-05, "loss": 0.1719, "step": 3834 }, { "epoch": 0.19, "learning_rate": 1.8663217400582505e-05, "loss": 0.2845, "step": 3835 }, { "epoch": 0.19, "learning_rate": 1.8662409512801147e-05, "loss": 0.4687, "step": 3836 }, { "epoch": 0.19, "learning_rate": 1.8661601398465813e-05, "loss": 0.3857, "step": 3837 }, { "epoch": 0.19, "learning_rate": 1.866079305759764e-05, "loss": 0.8354, "step": 3838 }, { "epoch": 0.19, "learning_rate": 1.865998449021777e-05, "loss": 0.395, "step": 3839 }, { "epoch": 0.19, "learning_rate": 1.8659175696347345e-05, "loss": 0.3041, "step": 3840 }, { "epoch": 0.19, "learning_rate": 1.8658366676007523e-05, "loss": 0.2507, "step": 3841 }, { "epoch": 0.19, "learning_rate": 1.865755742921946e-05, "loss": 0.1557, "step": 3842 }, { "epoch": 0.19, "learning_rate": 1.865674795600432e-05, "loss": 0.2886, "step": 3843 }, { "epoch": 0.19, "learning_rate": 1.8655938256383283e-05, "loss": 0.3037, "step": 3844 }, { "epoch": 0.19, "learning_rate": 1.8655128330377514e-05, "loss": 0.4644, "step": 3845 }, { "epoch": 0.19, "learning_rate": 1.8654318178008197e-05, "loss": 0.4716, "step": 3846 }, { "epoch": 0.19, "learning_rate": 1.8653507799296524e-05, "loss": 0.4773, "step": 3847 }, { "epoch": 0.19, "learning_rate": 1.865269719426369e-05, "loss": 0.4272, "step": 3848 }, { "epoch": 0.19, "learning_rate": 1.8651886362930892e-05, "loss": 0.3577, "step": 3849 }, { "epoch": 0.19, "learning_rate": 1.8651075305319338e-05, "loss": 0.7253, "step": 3850 }, { "epoch": 0.19, "learning_rate": 1.8650264021450244e-05, "loss": 0.401, "step": 3851 }, { "epoch": 0.19, "learning_rate": 1.8649452511344822e-05, "loss": 0.2879, "step": 3852 }, { "epoch": 0.19, "learning_rate": 1.8648640775024297e-05, "loss": 0.6268, "step": 3853 }, { "epoch": 0.19, "learning_rate": 1.86478288125099e-05, "loss": 0.6391, "step": 3854 }, { "epoch": 0.19, "learning_rate": 1.8647016623822872e-05, "loss": 0.5857, "step": 3855 }, { "epoch": 0.19, "learning_rate": 1.8646204208984447e-05, "loss": 0.4532, "step": 3856 }, { "epoch": 0.19, "learning_rate": 1.864539156801587e-05, "loss": 0.5056, "step": 3857 }, { "epoch": 0.19, "learning_rate": 1.8644578700938406e-05, "loss": 0.1548, "step": 3858 }, { "epoch": 0.19, "learning_rate": 1.864376560777331e-05, "loss": 0.4989, "step": 3859 }, { "epoch": 0.19, "learning_rate": 1.864295228854184e-05, "loss": 0.5225, "step": 3860 }, { "epoch": 0.19, "learning_rate": 1.8642138743265278e-05, "loss": 0.4936, "step": 3861 }, { "epoch": 0.19, "learning_rate": 1.8641324971964892e-05, "loss": 0.4252, "step": 3862 }, { "epoch": 0.19, "learning_rate": 1.8640510974661974e-05, "loss": 0.2704, "step": 3863 }, { "epoch": 0.19, "learning_rate": 1.8639696751377804e-05, "loss": 0.524, "step": 3864 }, { "epoch": 0.19, "learning_rate": 1.8638882302133687e-05, "loss": 0.2576, "step": 3865 }, { "epoch": 0.19, "learning_rate": 1.8638067626950914e-05, "loss": 0.4325, "step": 3866 }, { "epoch": 0.19, "learning_rate": 1.8637252725850797e-05, "loss": 0.3964, "step": 3867 }, { "epoch": 0.19, "learning_rate": 1.863643759885465e-05, "loss": 0.3667, "step": 3868 }, { "epoch": 0.19, "learning_rate": 1.8635622245983787e-05, "loss": 0.2864, "step": 3869 }, { "epoch": 0.19, "learning_rate": 1.863480666725953e-05, "loss": 0.2666, "step": 3870 }, { "epoch": 0.19, "learning_rate": 1.8633990862703224e-05, "loss": 0.2372, "step": 3871 }, { "epoch": 0.19, "learning_rate": 1.863317483233619e-05, "loss": 0.4421, "step": 3872 }, { "epoch": 0.19, "learning_rate": 1.8632358576179776e-05, "loss": 0.1853, "step": 3873 }, { "epoch": 0.19, "learning_rate": 1.8631542094255327e-05, "loss": 0.389, "step": 3874 }, { "epoch": 0.19, "learning_rate": 1.8630725386584207e-05, "loss": 0.3584, "step": 3875 }, { "epoch": 0.19, "learning_rate": 1.8629908453187765e-05, "loss": 0.4796, "step": 3876 }, { "epoch": 0.19, "learning_rate": 1.862909129408737e-05, "loss": 0.3393, "step": 3877 }, { "epoch": 0.19, "learning_rate": 1.8628273909304392e-05, "loss": 0.4885, "step": 3878 }, { "epoch": 0.19, "learning_rate": 1.8627456298860217e-05, "loss": 0.3612, "step": 3879 }, { "epoch": 0.19, "learning_rate": 1.8626638462776216e-05, "loss": 0.479, "step": 3880 }, { "epoch": 0.19, "learning_rate": 1.862582040107379e-05, "loss": 0.5547, "step": 3881 }, { "epoch": 0.19, "learning_rate": 1.8625002113774324e-05, "loss": 0.4342, "step": 3882 }, { "epoch": 0.19, "learning_rate": 1.8624183600899227e-05, "loss": 0.3929, "step": 3883 }, { "epoch": 0.19, "learning_rate": 1.8623364862469903e-05, "loss": 0.3666, "step": 3884 }, { "epoch": 0.19, "learning_rate": 1.862254589850777e-05, "loss": 0.3415, "step": 3885 }, { "epoch": 0.19, "learning_rate": 1.8621726709034237e-05, "loss": 0.2873, "step": 3886 }, { "epoch": 0.19, "learning_rate": 1.862090729407074e-05, "loss": 0.42, "step": 3887 }, { "epoch": 0.19, "learning_rate": 1.8620087653638696e-05, "loss": 0.2211, "step": 3888 }, { "epoch": 0.19, "learning_rate": 1.8619267787759556e-05, "loss": 0.2489, "step": 3889 }, { "epoch": 0.19, "learning_rate": 1.8618447696454755e-05, "loss": 0.4964, "step": 3890 }, { "epoch": 0.19, "learning_rate": 1.861762737974574e-05, "loss": 0.476, "step": 3891 }, { "epoch": 0.19, "learning_rate": 1.8616806837653972e-05, "loss": 0.2375, "step": 3892 }, { "epoch": 0.19, "learning_rate": 1.8615986070200906e-05, "loss": 0.3071, "step": 3893 }, { "epoch": 0.19, "learning_rate": 1.8615165077408005e-05, "loss": 0.3638, "step": 3894 }, { "epoch": 0.19, "learning_rate": 1.861434385929675e-05, "loss": 0.7633, "step": 3895 }, { "epoch": 0.19, "learning_rate": 1.861352241588861e-05, "loss": 0.419, "step": 3896 }, { "epoch": 0.19, "learning_rate": 1.861270074720508e-05, "loss": 0.4334, "step": 3897 }, { "epoch": 0.19, "learning_rate": 1.861187885326764e-05, "loss": 0.181, "step": 3898 }, { "epoch": 0.19, "learning_rate": 1.8611056734097785e-05, "loss": 0.4981, "step": 3899 }, { "epoch": 0.19, "learning_rate": 1.861023438971702e-05, "loss": 0.5077, "step": 3900 }, { "epoch": 0.19, "learning_rate": 1.8609411820146853e-05, "loss": 0.4685, "step": 3901 }, { "epoch": 0.19, "learning_rate": 1.8608589025408793e-05, "loss": 0.3385, "step": 3902 }, { "epoch": 0.19, "learning_rate": 1.8607766005524368e-05, "loss": 0.3915, "step": 3903 }, { "epoch": 0.19, "learning_rate": 1.8606942760515096e-05, "loss": 0.3831, "step": 3904 }, { "epoch": 0.19, "learning_rate": 1.8606119290402508e-05, "loss": 0.4558, "step": 3905 }, { "epoch": 0.2, "learning_rate": 1.860529559520814e-05, "loss": 0.3601, "step": 3906 }, { "epoch": 0.2, "learning_rate": 1.860447167495354e-05, "loss": 0.3625, "step": 3907 }, { "epoch": 0.2, "learning_rate": 1.860364752966025e-05, "loss": 0.4687, "step": 3908 }, { "epoch": 0.2, "learning_rate": 1.860282315934983e-05, "loss": 0.4272, "step": 3909 }, { "epoch": 0.2, "learning_rate": 1.8601998564043838e-05, "loss": 0.3553, "step": 3910 }, { "epoch": 0.2, "learning_rate": 1.8601173743763836e-05, "loss": 0.6796, "step": 3911 }, { "epoch": 0.2, "learning_rate": 1.8600348698531405e-05, "loss": 0.4537, "step": 3912 }, { "epoch": 0.2, "learning_rate": 1.8599523428368116e-05, "loss": 0.2761, "step": 3913 }, { "epoch": 0.2, "learning_rate": 1.8598697933295556e-05, "loss": 0.5514, "step": 3914 }, { "epoch": 0.2, "learning_rate": 1.8597872213335314e-05, "loss": 0.2305, "step": 3915 }, { "epoch": 0.2, "learning_rate": 1.859704626850898e-05, "loss": 0.4298, "step": 3916 }, { "epoch": 0.2, "learning_rate": 1.859622009883817e-05, "loss": 0.3676, "step": 3917 }, { "epoch": 0.2, "learning_rate": 1.859539370434448e-05, "loss": 0.5724, "step": 3918 }, { "epoch": 0.2, "learning_rate": 1.8594567085049524e-05, "loss": 0.4568, "step": 3919 }, { "epoch": 0.2, "learning_rate": 1.859374024097492e-05, "loss": 0.2983, "step": 3920 }, { "epoch": 0.2, "learning_rate": 1.85929131721423e-05, "loss": 0.3416, "step": 3921 }, { "epoch": 0.2, "learning_rate": 1.859208587857329e-05, "loss": 0.2593, "step": 3922 }, { "epoch": 0.2, "learning_rate": 1.8591258360289524e-05, "loss": 0.2996, "step": 3923 }, { "epoch": 0.2, "learning_rate": 1.859043061731265e-05, "loss": 0.3775, "step": 3924 }, { "epoch": 0.2, "learning_rate": 1.8589602649664317e-05, "loss": 0.2296, "step": 3925 }, { "epoch": 0.2, "learning_rate": 1.8588774457366176e-05, "loss": 0.3314, "step": 3926 }, { "epoch": 0.2, "learning_rate": 1.858794604043989e-05, "loss": 0.4457, "step": 3927 }, { "epoch": 0.2, "learning_rate": 1.858711739890712e-05, "loss": 0.2906, "step": 3928 }, { "epoch": 0.2, "learning_rate": 1.8586288532789544e-05, "loss": 0.3023, "step": 3929 }, { "epoch": 0.2, "learning_rate": 1.8585459442108835e-05, "loss": 0.1948, "step": 3930 }, { "epoch": 0.2, "learning_rate": 1.858463012688668e-05, "loss": 0.3375, "step": 3931 }, { "epoch": 0.2, "learning_rate": 1.8583800587144767e-05, "loss": 0.3969, "step": 3932 }, { "epoch": 0.2, "learning_rate": 1.8582970822904793e-05, "loss": 0.3519, "step": 3933 }, { "epoch": 0.2, "learning_rate": 1.858214083418846e-05, "loss": 0.469, "step": 3934 }, { "epoch": 0.2, "learning_rate": 1.8581310621017474e-05, "loss": 0.3184, "step": 3935 }, { "epoch": 0.2, "learning_rate": 1.858048018341355e-05, "loss": 0.4126, "step": 3936 }, { "epoch": 0.2, "learning_rate": 1.85796495213984e-05, "loss": 0.4908, "step": 3937 }, { "epoch": 0.2, "learning_rate": 1.8578818634993755e-05, "loss": 0.4416, "step": 3938 }, { "epoch": 0.2, "learning_rate": 1.8577987524221347e-05, "loss": 0.698, "step": 3939 }, { "epoch": 0.2, "learning_rate": 1.8577156189102913e-05, "loss": 0.3638, "step": 3940 }, { "epoch": 0.2, "learning_rate": 1.8576324629660193e-05, "loss": 0.2837, "step": 3941 }, { "epoch": 0.2, "learning_rate": 1.8575492845914928e-05, "loss": 0.1982, "step": 3942 }, { "epoch": 0.2, "learning_rate": 1.8574660837888885e-05, "loss": 0.306, "step": 3943 }, { "epoch": 0.2, "learning_rate": 1.857382860560382e-05, "loss": 0.2847, "step": 3944 }, { "epoch": 0.2, "learning_rate": 1.8572996149081494e-05, "loss": 0.5264, "step": 3945 }, { "epoch": 0.2, "learning_rate": 1.8572163468343683e-05, "loss": 0.3119, "step": 3946 }, { "epoch": 0.2, "learning_rate": 1.8571330563412166e-05, "loss": 0.2566, "step": 3947 }, { "epoch": 0.2, "learning_rate": 1.857049743430872e-05, "loss": 0.5832, "step": 3948 }, { "epoch": 0.2, "learning_rate": 1.8569664081055142e-05, "loss": 0.4049, "step": 3949 }, { "epoch": 0.2, "learning_rate": 1.8568830503673224e-05, "loss": 0.342, "step": 3950 }, { "epoch": 0.2, "learning_rate": 1.856799670218477e-05, "loss": 0.2264, "step": 3951 }, { "epoch": 0.2, "learning_rate": 1.8567162676611582e-05, "loss": 0.3185, "step": 3952 }, { "epoch": 0.2, "learning_rate": 1.8566328426975474e-05, "loss": 0.4505, "step": 3953 }, { "epoch": 0.2, "learning_rate": 1.8565493953298266e-05, "loss": 0.4359, "step": 3954 }, { "epoch": 0.2, "learning_rate": 1.8564659255601783e-05, "loss": 0.2403, "step": 3955 }, { "epoch": 0.2, "learning_rate": 1.8563824333907858e-05, "loss": 0.3583, "step": 3956 }, { "epoch": 0.2, "learning_rate": 1.8562989188238322e-05, "loss": 0.306, "step": 3957 }, { "epoch": 0.2, "learning_rate": 1.8562153818615014e-05, "loss": 0.7085, "step": 3958 }, { "epoch": 0.2, "learning_rate": 1.8561318225059795e-05, "loss": 0.4033, "step": 3959 }, { "epoch": 0.2, "learning_rate": 1.8560482407594508e-05, "loss": 0.6495, "step": 3960 }, { "epoch": 0.2, "learning_rate": 1.8559646366241014e-05, "loss": 0.3052, "step": 3961 }, { "epoch": 0.2, "learning_rate": 1.8558810101021184e-05, "loss": 0.5032, "step": 3962 }, { "epoch": 0.2, "learning_rate": 1.855797361195688e-05, "loss": 0.4078, "step": 3963 }, { "epoch": 0.2, "learning_rate": 1.855713689906999e-05, "loss": 0.2374, "step": 3964 }, { "epoch": 0.2, "learning_rate": 1.855629996238239e-05, "loss": 0.603, "step": 3965 }, { "epoch": 0.2, "learning_rate": 1.8555462801915976e-05, "loss": 0.5498, "step": 3966 }, { "epoch": 0.2, "learning_rate": 1.8554625417692636e-05, "loss": 0.2887, "step": 3967 }, { "epoch": 0.2, "learning_rate": 1.8553787809734268e-05, "loss": 0.4622, "step": 3968 }, { "epoch": 0.2, "learning_rate": 1.855294997806279e-05, "loss": 0.3172, "step": 3969 }, { "epoch": 0.2, "learning_rate": 1.8552111922700103e-05, "loss": 0.7291, "step": 3970 }, { "epoch": 0.2, "learning_rate": 1.8551273643668132e-05, "loss": 0.5932, "step": 3971 }, { "epoch": 0.2, "learning_rate": 1.8550435140988797e-05, "loss": 0.4222, "step": 3972 }, { "epoch": 0.2, "learning_rate": 1.8549596414684033e-05, "loss": 0.4935, "step": 3973 }, { "epoch": 0.2, "learning_rate": 1.854875746477577e-05, "loss": 0.2391, "step": 3974 }, { "epoch": 0.2, "learning_rate": 1.8547918291285952e-05, "loss": 0.4007, "step": 3975 }, { "epoch": 0.2, "learning_rate": 1.854707889423653e-05, "loss": 0.1954, "step": 3976 }, { "epoch": 0.2, "learning_rate": 1.8546239273649452e-05, "loss": 0.2324, "step": 3977 }, { "epoch": 0.2, "learning_rate": 1.854539942954668e-05, "loss": 0.4853, "step": 3978 }, { "epoch": 0.2, "learning_rate": 1.854455936195018e-05, "loss": 0.3683, "step": 3979 }, { "epoch": 0.2, "learning_rate": 1.8543719070881918e-05, "loss": 0.6469, "step": 3980 }, { "epoch": 0.2, "learning_rate": 1.8542878556363876e-05, "loss": 0.4838, "step": 3981 }, { "epoch": 0.2, "learning_rate": 1.854203781841803e-05, "loss": 0.3424, "step": 3982 }, { "epoch": 0.2, "learning_rate": 1.8541196857066378e-05, "loss": 0.296, "step": 3983 }, { "epoch": 0.2, "learning_rate": 1.854035567233091e-05, "loss": 0.3602, "step": 3984 }, { "epoch": 0.2, "learning_rate": 1.853951426423362e-05, "loss": 0.4446, "step": 3985 }, { "epoch": 0.2, "learning_rate": 1.8538672632796525e-05, "loss": 0.4718, "step": 3986 }, { "epoch": 0.2, "learning_rate": 1.853783077804163e-05, "loss": 0.2715, "step": 3987 }, { "epoch": 0.2, "learning_rate": 1.853698869999095e-05, "loss": 0.2986, "step": 3988 }, { "epoch": 0.2, "learning_rate": 1.8536146398666512e-05, "loss": 0.1208, "step": 3989 }, { "epoch": 0.2, "learning_rate": 1.8535303874090348e-05, "loss": 0.4001, "step": 3990 }, { "epoch": 0.2, "learning_rate": 1.853446112628449e-05, "loss": 0.4195, "step": 3991 }, { "epoch": 0.2, "learning_rate": 1.853361815527098e-05, "loss": 0.6228, "step": 3992 }, { "epoch": 0.2, "learning_rate": 1.853277496107186e-05, "loss": 0.2323, "step": 3993 }, { "epoch": 0.2, "learning_rate": 1.853193154370919e-05, "loss": 0.2988, "step": 3994 }, { "epoch": 0.2, "learning_rate": 1.853108790320503e-05, "loss": 0.353, "step": 3995 }, { "epoch": 0.2, "learning_rate": 1.853024403958143e-05, "loss": 0.6789, "step": 3996 }, { "epoch": 0.2, "learning_rate": 1.8529399952860478e-05, "loss": 0.3912, "step": 3997 }, { "epoch": 0.2, "learning_rate": 1.852855564306424e-05, "loss": 0.2895, "step": 3998 }, { "epoch": 0.2, "learning_rate": 1.8527711110214794e-05, "loss": 0.6344, "step": 3999 }, { "epoch": 0.2, "learning_rate": 1.852686635433424e-05, "loss": 0.4307, "step": 4000 }, { "epoch": 0.2, "learning_rate": 1.852602137544466e-05, "loss": 0.3951, "step": 4001 }, { "epoch": 0.2, "learning_rate": 1.8525176173568162e-05, "loss": 0.3777, "step": 4002 }, { "epoch": 0.2, "learning_rate": 1.8524330748726842e-05, "loss": 0.3719, "step": 4003 }, { "epoch": 0.2, "learning_rate": 1.852348510094282e-05, "loss": 0.3572, "step": 4004 }, { "epoch": 0.2, "learning_rate": 1.852263923023821e-05, "loss": 0.3344, "step": 4005 }, { "epoch": 0.2, "learning_rate": 1.852179313663513e-05, "loss": 0.4501, "step": 4006 }, { "epoch": 0.2, "learning_rate": 1.8520946820155714e-05, "loss": 0.5025, "step": 4007 }, { "epoch": 0.2, "learning_rate": 1.8520100280822094e-05, "loss": 0.5107, "step": 4008 }, { "epoch": 0.2, "learning_rate": 1.851925351865641e-05, "loss": 0.434, "step": 4009 }, { "epoch": 0.2, "learning_rate": 1.8518406533680808e-05, "loss": 0.469, "step": 4010 }, { "epoch": 0.2, "learning_rate": 1.8517559325917443e-05, "loss": 0.1819, "step": 4011 }, { "epoch": 0.2, "learning_rate": 1.8516711895388468e-05, "loss": 0.2701, "step": 4012 }, { "epoch": 0.2, "learning_rate": 1.851586424211605e-05, "loss": 0.2471, "step": 4013 }, { "epoch": 0.2, "learning_rate": 1.8515016366122353e-05, "loss": 0.4502, "step": 4014 }, { "epoch": 0.2, "learning_rate": 1.8514168267429554e-05, "loss": 0.5449, "step": 4015 }, { "epoch": 0.2, "learning_rate": 1.851331994605984e-05, "loss": 0.6887, "step": 4016 }, { "epoch": 0.2, "learning_rate": 1.8512471402035396e-05, "loss": 0.3823, "step": 4017 }, { "epoch": 0.2, "learning_rate": 1.8511622635378405e-05, "loss": 0.2097, "step": 4018 }, { "epoch": 0.2, "learning_rate": 1.8510773646111076e-05, "loss": 0.4941, "step": 4019 }, { "epoch": 0.2, "learning_rate": 1.8509924434255607e-05, "loss": 0.5629, "step": 4020 }, { "epoch": 0.2, "learning_rate": 1.8509074999834212e-05, "loss": 0.4484, "step": 4021 }, { "epoch": 0.2, "learning_rate": 1.8508225342869107e-05, "loss": 0.1587, "step": 4022 }, { "epoch": 0.2, "learning_rate": 1.850737546338251e-05, "loss": 0.5028, "step": 4023 }, { "epoch": 0.2, "learning_rate": 1.850652536139665e-05, "loss": 0.3784, "step": 4024 }, { "epoch": 0.2, "learning_rate": 1.850567503693376e-05, "loss": 0.4025, "step": 4025 }, { "epoch": 0.2, "learning_rate": 1.8504824490016083e-05, "loss": 0.5948, "step": 4026 }, { "epoch": 0.2, "learning_rate": 1.8503973720665857e-05, "loss": 0.3217, "step": 4027 }, { "epoch": 0.2, "learning_rate": 1.8503122728905342e-05, "loss": 0.4034, "step": 4028 }, { "epoch": 0.2, "learning_rate": 1.8502271514756784e-05, "loss": 0.3348, "step": 4029 }, { "epoch": 0.2, "learning_rate": 1.8501420078242453e-05, "loss": 0.422, "step": 4030 }, { "epoch": 0.2, "learning_rate": 1.8500568419384613e-05, "loss": 0.3564, "step": 4031 }, { "epoch": 0.2, "learning_rate": 1.849971653820554e-05, "loss": 0.5898, "step": 4032 }, { "epoch": 0.2, "learning_rate": 1.8498864434727512e-05, "loss": 0.2366, "step": 4033 }, { "epoch": 0.2, "learning_rate": 1.849801210897282e-05, "loss": 0.2848, "step": 4034 }, { "epoch": 0.2, "learning_rate": 1.8497159560963752e-05, "loss": 0.2596, "step": 4035 }, { "epoch": 0.2, "learning_rate": 1.84963067907226e-05, "loss": 0.4507, "step": 4036 }, { "epoch": 0.2, "learning_rate": 1.8495453798271677e-05, "loss": 0.4377, "step": 4037 }, { "epoch": 0.2, "learning_rate": 1.8494600583633284e-05, "loss": 0.1864, "step": 4038 }, { "epoch": 0.2, "learning_rate": 1.849374714682974e-05, "loss": 0.4305, "step": 4039 }, { "epoch": 0.2, "learning_rate": 1.8492893487883362e-05, "loss": 0.6216, "step": 4040 }, { "epoch": 0.2, "learning_rate": 1.8492039606816482e-05, "loss": 0.3457, "step": 4041 }, { "epoch": 0.2, "learning_rate": 1.8491185503651427e-05, "loss": 0.4704, "step": 4042 }, { "epoch": 0.2, "learning_rate": 1.8490331178410536e-05, "loss": 0.2678, "step": 4043 }, { "epoch": 0.2, "learning_rate": 1.8489476631116156e-05, "loss": 0.3423, "step": 4044 }, { "epoch": 0.2, "learning_rate": 1.848862186179063e-05, "loss": 0.3768, "step": 4045 }, { "epoch": 0.2, "learning_rate": 1.8487766870456322e-05, "loss": 0.2401, "step": 4046 }, { "epoch": 0.2, "learning_rate": 1.8486911657135585e-05, "loss": 0.2341, "step": 4047 }, { "epoch": 0.2, "learning_rate": 1.848605622185079e-05, "loss": 0.544, "step": 4048 }, { "epoch": 0.2, "learning_rate": 1.848520056462431e-05, "loss": 0.4974, "step": 4049 }, { "epoch": 0.2, "learning_rate": 1.8484344685478518e-05, "loss": 0.4304, "step": 4050 }, { "epoch": 0.2, "learning_rate": 1.848348858443581e-05, "loss": 0.2905, "step": 4051 }, { "epoch": 0.2, "learning_rate": 1.848263226151857e-05, "loss": 0.4215, "step": 4052 }, { "epoch": 0.2, "learning_rate": 1.8481775716749188e-05, "loss": 0.3384, "step": 4053 }, { "epoch": 0.2, "learning_rate": 1.8480918950150075e-05, "loss": 0.3788, "step": 4054 }, { "epoch": 0.2, "learning_rate": 1.848006196174363e-05, "loss": 0.2515, "step": 4055 }, { "epoch": 0.2, "learning_rate": 1.8479204751552275e-05, "loss": 0.349, "step": 4056 }, { "epoch": 0.2, "learning_rate": 1.8478347319598427e-05, "loss": 0.2834, "step": 4057 }, { "epoch": 0.2, "learning_rate": 1.8477489665904506e-05, "loss": 0.653, "step": 4058 }, { "epoch": 0.2, "learning_rate": 1.847663179049295e-05, "loss": 0.2512, "step": 4059 }, { "epoch": 0.2, "learning_rate": 1.8475773693386188e-05, "loss": 0.4847, "step": 4060 }, { "epoch": 0.2, "learning_rate": 1.847491537460667e-05, "loss": 0.2216, "step": 4061 }, { "epoch": 0.2, "learning_rate": 1.847405683417684e-05, "loss": 0.3449, "step": 4062 }, { "epoch": 0.2, "learning_rate": 1.847319807211915e-05, "loss": 0.5364, "step": 4063 }, { "epoch": 0.2, "learning_rate": 1.8472339088456067e-05, "loss": 0.587, "step": 4064 }, { "epoch": 0.2, "learning_rate": 1.8471479883210046e-05, "loss": 0.5056, "step": 4065 }, { "epoch": 0.2, "learning_rate": 1.847062045640357e-05, "loss": 0.5741, "step": 4066 }, { "epoch": 0.2, "learning_rate": 1.8469760808059102e-05, "loss": 0.3618, "step": 4067 }, { "epoch": 0.2, "learning_rate": 1.846890093819914e-05, "loss": 0.3567, "step": 4068 }, { "epoch": 0.2, "learning_rate": 1.8468040846846166e-05, "loss": 0.419, "step": 4069 }, { "epoch": 0.2, "learning_rate": 1.8467180534022672e-05, "loss": 0.3156, "step": 4070 }, { "epoch": 0.2, "learning_rate": 1.8466319999751167e-05, "loss": 0.4477, "step": 4071 }, { "epoch": 0.2, "learning_rate": 1.8465459244054145e-05, "loss": 0.3933, "step": 4072 }, { "epoch": 0.2, "learning_rate": 1.8464598266954126e-05, "loss": 0.3857, "step": 4073 }, { "epoch": 0.2, "learning_rate": 1.8463737068473627e-05, "loss": 0.4535, "step": 4074 }, { "epoch": 0.2, "learning_rate": 1.846287564863517e-05, "loss": 0.5086, "step": 4075 }, { "epoch": 0.2, "learning_rate": 1.8462014007461285e-05, "loss": 0.4937, "step": 4076 }, { "epoch": 0.2, "learning_rate": 1.8461152144974508e-05, "loss": 0.3329, "step": 4077 }, { "epoch": 0.2, "learning_rate": 1.8460290061197374e-05, "loss": 0.5634, "step": 4078 }, { "epoch": 0.2, "learning_rate": 1.845942775615244e-05, "loss": 0.7004, "step": 4079 }, { "epoch": 0.2, "learning_rate": 1.8458565229862252e-05, "loss": 0.2188, "step": 4080 }, { "epoch": 0.2, "learning_rate": 1.8457702482349366e-05, "loss": 0.4184, "step": 4081 }, { "epoch": 0.2, "learning_rate": 1.8456839513636354e-05, "loss": 0.3551, "step": 4082 }, { "epoch": 0.2, "learning_rate": 1.8455976323745776e-05, "loss": 0.5618, "step": 4083 }, { "epoch": 0.2, "learning_rate": 1.845511291270021e-05, "loss": 0.4077, "step": 4084 }, { "epoch": 0.2, "learning_rate": 1.845424928052225e-05, "loss": 0.5568, "step": 4085 }, { "epoch": 0.2, "learning_rate": 1.8453385427234468e-05, "loss": 0.4511, "step": 4086 }, { "epoch": 0.2, "learning_rate": 1.845252135285946e-05, "loss": 0.3638, "step": 4087 }, { "epoch": 0.2, "learning_rate": 1.845165705741983e-05, "loss": 0.3268, "step": 4088 }, { "epoch": 0.2, "learning_rate": 1.845079254093818e-05, "loss": 0.2955, "step": 4089 }, { "epoch": 0.2, "learning_rate": 1.8449927803437115e-05, "loss": 0.3087, "step": 4090 }, { "epoch": 0.2, "learning_rate": 1.844906284493926e-05, "loss": 0.3158, "step": 4091 }, { "epoch": 0.2, "learning_rate": 1.844819766546723e-05, "loss": 0.365, "step": 4092 }, { "epoch": 0.2, "learning_rate": 1.8447332265043653e-05, "loss": 0.24, "step": 4093 }, { "epoch": 0.2, "learning_rate": 1.844646664369117e-05, "loss": 0.2738, "step": 4094 }, { "epoch": 0.2, "learning_rate": 1.8445600801432412e-05, "loss": 0.233, "step": 4095 }, { "epoch": 0.2, "learning_rate": 1.844473473829003e-05, "loss": 0.3626, "step": 4096 }, { "epoch": 0.2, "learning_rate": 1.8443868454286667e-05, "loss": 0.3802, "step": 4097 }, { "epoch": 0.2, "learning_rate": 1.8443001949444986e-05, "loss": 0.4873, "step": 4098 }, { "epoch": 0.2, "learning_rate": 1.8442135223787646e-05, "loss": 0.4158, "step": 4099 }, { "epoch": 0.2, "learning_rate": 1.844126827733732e-05, "loss": 0.3163, "step": 4100 }, { "epoch": 0.2, "learning_rate": 1.844040111011668e-05, "loss": 0.3881, "step": 4101 }, { "epoch": 0.2, "learning_rate": 1.8439533722148394e-05, "loss": 0.3765, "step": 4102 }, { "epoch": 0.2, "learning_rate": 1.8438666113455165e-05, "loss": 0.6073, "step": 4103 }, { "epoch": 0.2, "learning_rate": 1.8437798284059674e-05, "loss": 0.3093, "step": 4104 }, { "epoch": 0.2, "learning_rate": 1.843693023398462e-05, "loss": 0.4065, "step": 4105 }, { "epoch": 0.21, "learning_rate": 1.8436061963252706e-05, "loss": 0.317, "step": 4106 }, { "epoch": 0.21, "learning_rate": 1.8435193471886643e-05, "loss": 0.4628, "step": 4107 }, { "epoch": 0.21, "learning_rate": 1.8434324759909144e-05, "loss": 0.4656, "step": 4108 }, { "epoch": 0.21, "learning_rate": 1.8433455827342924e-05, "loss": 0.4063, "step": 4109 }, { "epoch": 0.21, "learning_rate": 1.8432586674210716e-05, "loss": 0.3671, "step": 4110 }, { "epoch": 0.21, "learning_rate": 1.8431717300535247e-05, "loss": 0.3961, "step": 4111 }, { "epoch": 0.21, "learning_rate": 1.8430847706339255e-05, "loss": 0.534, "step": 4112 }, { "epoch": 0.21, "learning_rate": 1.8429977891645483e-05, "loss": 0.3852, "step": 4113 }, { "epoch": 0.21, "learning_rate": 1.8429107856476682e-05, "loss": 0.3624, "step": 4114 }, { "epoch": 0.21, "learning_rate": 1.8428237600855603e-05, "loss": 0.4304, "step": 4115 }, { "epoch": 0.21, "learning_rate": 1.8427367124805008e-05, "loss": 0.3973, "step": 4116 }, { "epoch": 0.21, "learning_rate": 1.8426496428347668e-05, "loss": 0.3845, "step": 4117 }, { "epoch": 0.21, "learning_rate": 1.8425625511506348e-05, "loss": 0.4869, "step": 4118 }, { "epoch": 0.21, "learning_rate": 1.842475437430383e-05, "loss": 0.3928, "step": 4119 }, { "epoch": 0.21, "learning_rate": 1.8423883016762896e-05, "loss": 0.0922, "step": 4120 }, { "epoch": 0.21, "learning_rate": 1.8423011438906334e-05, "loss": 0.4992, "step": 4121 }, { "epoch": 0.21, "learning_rate": 1.842213964075694e-05, "loss": 0.5897, "step": 4122 }, { "epoch": 0.21, "learning_rate": 1.842126762233751e-05, "loss": 0.3529, "step": 4123 }, { "epoch": 0.21, "learning_rate": 1.8420395383670858e-05, "loss": 0.5514, "step": 4124 }, { "epoch": 0.21, "learning_rate": 1.8419522924779796e-05, "loss": 0.5796, "step": 4125 }, { "epoch": 0.21, "learning_rate": 1.841865024568714e-05, "loss": 0.3969, "step": 4126 }, { "epoch": 0.21, "learning_rate": 1.8417777346415712e-05, "loss": 0.2506, "step": 4127 }, { "epoch": 0.21, "learning_rate": 1.841690422698834e-05, "loss": 0.4857, "step": 4128 }, { "epoch": 0.21, "learning_rate": 1.8416030887427867e-05, "loss": 0.5065, "step": 4129 }, { "epoch": 0.21, "learning_rate": 1.8415157327757124e-05, "loss": 0.3504, "step": 4130 }, { "epoch": 0.21, "learning_rate": 1.8414283547998968e-05, "loss": 0.587, "step": 4131 }, { "epoch": 0.21, "learning_rate": 1.8413409548176244e-05, "loss": 0.3528, "step": 4132 }, { "epoch": 0.21, "learning_rate": 1.8412535328311813e-05, "loss": 0.2063, "step": 4133 }, { "epoch": 0.21, "learning_rate": 1.841166088842854e-05, "loss": 0.3292, "step": 4134 }, { "epoch": 0.21, "learning_rate": 1.841078622854929e-05, "loss": 0.4389, "step": 4135 }, { "epoch": 0.21, "learning_rate": 1.8409911348696947e-05, "loss": 0.3776, "step": 4136 }, { "epoch": 0.21, "learning_rate": 1.8409036248894385e-05, "loss": 0.6338, "step": 4137 }, { "epoch": 0.21, "learning_rate": 1.8408160929164496e-05, "loss": 0.2175, "step": 4138 }, { "epoch": 0.21, "learning_rate": 1.840728538953017e-05, "loss": 0.5739, "step": 4139 }, { "epoch": 0.21, "learning_rate": 1.8406409630014302e-05, "loss": 0.446, "step": 4140 }, { "epoch": 0.21, "learning_rate": 1.8405533650639806e-05, "loss": 0.5441, "step": 4141 }, { "epoch": 0.21, "learning_rate": 1.8404657451429583e-05, "loss": 0.4336, "step": 4142 }, { "epoch": 0.21, "learning_rate": 1.8403781032406555e-05, "loss": 0.4047, "step": 4143 }, { "epoch": 0.21, "learning_rate": 1.8402904393593643e-05, "loss": 0.5447, "step": 4144 }, { "epoch": 0.21, "learning_rate": 1.8402027535013767e-05, "loss": 0.7107, "step": 4145 }, { "epoch": 0.21, "learning_rate": 1.8401150456689868e-05, "loss": 0.4046, "step": 4146 }, { "epoch": 0.21, "learning_rate": 1.840027315864488e-05, "loss": 0.2232, "step": 4147 }, { "epoch": 0.21, "learning_rate": 1.8399395640901755e-05, "loss": 0.3305, "step": 4148 }, { "epoch": 0.21, "learning_rate": 1.8398517903483438e-05, "loss": 0.2769, "step": 4149 }, { "epoch": 0.21, "learning_rate": 1.839763994641288e-05, "loss": 0.2099, "step": 4150 }, { "epoch": 0.21, "learning_rate": 1.8396761769713052e-05, "loss": 0.4083, "step": 4151 }, { "epoch": 0.21, "learning_rate": 1.8395883373406914e-05, "loss": 0.4122, "step": 4152 }, { "epoch": 0.21, "learning_rate": 1.839500475751745e-05, "loss": 0.2189, "step": 4153 }, { "epoch": 0.21, "learning_rate": 1.8394125922067628e-05, "loss": 0.4604, "step": 4154 }, { "epoch": 0.21, "learning_rate": 1.8393246867080433e-05, "loss": 0.4032, "step": 4155 }, { "epoch": 0.21, "learning_rate": 1.8392367592578866e-05, "loss": 0.3673, "step": 4156 }, { "epoch": 0.21, "learning_rate": 1.8391488098585913e-05, "loss": 0.493, "step": 4157 }, { "epoch": 0.21, "learning_rate": 1.8390608385124578e-05, "loss": 0.4185, "step": 4158 }, { "epoch": 0.21, "learning_rate": 1.8389728452217874e-05, "loss": 0.2986, "step": 4159 }, { "epoch": 0.21, "learning_rate": 1.838884829988881e-05, "loss": 0.412, "step": 4160 }, { "epoch": 0.21, "learning_rate": 1.8387967928160405e-05, "loss": 0.4056, "step": 4161 }, { "epoch": 0.21, "learning_rate": 1.8387087337055684e-05, "loss": 0.3411, "step": 4162 }, { "epoch": 0.21, "learning_rate": 1.8386206526597683e-05, "loss": 0.3576, "step": 4163 }, { "epoch": 0.21, "learning_rate": 1.8385325496809428e-05, "loss": 0.2859, "step": 4164 }, { "epoch": 0.21, "learning_rate": 1.8384444247713972e-05, "loss": 0.4193, "step": 4165 }, { "epoch": 0.21, "learning_rate": 1.838356277933436e-05, "loss": 0.5712, "step": 4166 }, { "epoch": 0.21, "learning_rate": 1.838268109169364e-05, "loss": 0.3076, "step": 4167 }, { "epoch": 0.21, "learning_rate": 1.838179918481488e-05, "loss": 0.4933, "step": 4168 }, { "epoch": 0.21, "learning_rate": 1.8380917058721137e-05, "loss": 0.3049, "step": 4169 }, { "epoch": 0.21, "learning_rate": 1.8380034713435485e-05, "loss": 0.3945, "step": 4170 }, { "epoch": 0.21, "learning_rate": 1.8379152148981e-05, "loss": 0.2542, "step": 4171 }, { "epoch": 0.21, "learning_rate": 1.837826936538077e-05, "loss": 0.1648, "step": 4172 }, { "epoch": 0.21, "learning_rate": 1.8377386362657874e-05, "loss": 0.4041, "step": 4173 }, { "epoch": 0.21, "learning_rate": 1.837650314083541e-05, "loss": 0.5758, "step": 4174 }, { "epoch": 0.21, "learning_rate": 1.8375619699936483e-05, "loss": 0.3558, "step": 4175 }, { "epoch": 0.21, "learning_rate": 1.8374736039984187e-05, "loss": 0.2696, "step": 4176 }, { "epoch": 0.21, "learning_rate": 1.837385216100164e-05, "loss": 0.3173, "step": 4177 }, { "epoch": 0.21, "learning_rate": 1.837296806301196e-05, "loss": 0.5316, "step": 4178 }, { "epoch": 0.21, "learning_rate": 1.837208374603826e-05, "loss": 0.2724, "step": 4179 }, { "epoch": 0.21, "learning_rate": 1.8371199210103685e-05, "loss": 0.6912, "step": 4180 }, { "epoch": 0.21, "learning_rate": 1.8370314455231357e-05, "loss": 0.395, "step": 4181 }, { "epoch": 0.21, "learning_rate": 1.8369429481444413e-05, "loss": 0.1821, "step": 4182 }, { "epoch": 0.21, "learning_rate": 1.8368544288766003e-05, "loss": 0.5722, "step": 4183 }, { "epoch": 0.21, "learning_rate": 1.836765887721928e-05, "loss": 0.3745, "step": 4184 }, { "epoch": 0.21, "learning_rate": 1.8366773246827398e-05, "loss": 0.4019, "step": 4185 }, { "epoch": 0.21, "learning_rate": 1.8365887397613523e-05, "loss": 0.4572, "step": 4186 }, { "epoch": 0.21, "learning_rate": 1.8365001329600815e-05, "loss": 0.4687, "step": 4187 }, { "epoch": 0.21, "learning_rate": 1.8364115042812457e-05, "loss": 0.3348, "step": 4188 }, { "epoch": 0.21, "learning_rate": 1.8363228537271623e-05, "loss": 0.4764, "step": 4189 }, { "epoch": 0.21, "learning_rate": 1.8362341813001502e-05, "loss": 0.3081, "step": 4190 }, { "epoch": 0.21, "learning_rate": 1.8361454870025285e-05, "loss": 0.401, "step": 4191 }, { "epoch": 0.21, "learning_rate": 1.8360567708366167e-05, "loss": 0.3296, "step": 4192 }, { "epoch": 0.21, "learning_rate": 1.8359680328047348e-05, "loss": 0.5337, "step": 4193 }, { "epoch": 0.21, "learning_rate": 1.835879272909204e-05, "loss": 0.3207, "step": 4194 }, { "epoch": 0.21, "learning_rate": 1.8357904911523458e-05, "loss": 0.592, "step": 4195 }, { "epoch": 0.21, "learning_rate": 1.835701687536482e-05, "loss": 0.724, "step": 4196 }, { "epoch": 0.21, "learning_rate": 1.8356128620639352e-05, "loss": 0.2072, "step": 4197 }, { "epoch": 0.21, "learning_rate": 1.8355240147370285e-05, "loss": 0.1559, "step": 4198 }, { "epoch": 0.21, "learning_rate": 1.8354351455580853e-05, "loss": 0.5195, "step": 4199 }, { "epoch": 0.21, "learning_rate": 1.83534625452943e-05, "loss": 0.4107, "step": 4200 }, { "epoch": 0.21, "learning_rate": 1.8352573416533877e-05, "loss": 0.4044, "step": 4201 }, { "epoch": 0.21, "learning_rate": 1.8351684069322835e-05, "loss": 0.4431, "step": 4202 }, { "epoch": 0.21, "learning_rate": 1.8350794503684436e-05, "loss": 0.3899, "step": 4203 }, { "epoch": 0.21, "learning_rate": 1.8349904719641946e-05, "loss": 0.5111, "step": 4204 }, { "epoch": 0.21, "learning_rate": 1.834901471721863e-05, "loss": 0.4845, "step": 4205 }, { "epoch": 0.21, "learning_rate": 1.8348124496437774e-05, "loss": 0.3131, "step": 4206 }, { "epoch": 0.21, "learning_rate": 1.8347234057322652e-05, "loss": 0.3874, "step": 4207 }, { "epoch": 0.21, "learning_rate": 1.834634339989656e-05, "loss": 0.3282, "step": 4208 }, { "epoch": 0.21, "learning_rate": 1.8345452524182783e-05, "loss": 0.5293, "step": 4209 }, { "epoch": 0.21, "learning_rate": 1.834456143020463e-05, "loss": 0.4802, "step": 4210 }, { "epoch": 0.21, "learning_rate": 1.83436701179854e-05, "loss": 0.6031, "step": 4211 }, { "epoch": 0.21, "learning_rate": 1.834277858754841e-05, "loss": 0.3181, "step": 4212 }, { "epoch": 0.21, "learning_rate": 1.834188683891697e-05, "loss": 0.4039, "step": 4213 }, { "epoch": 0.21, "learning_rate": 1.8340994872114402e-05, "loss": 0.4642, "step": 4214 }, { "epoch": 0.21, "learning_rate": 1.8340102687164047e-05, "loss": 0.4976, "step": 4215 }, { "epoch": 0.21, "learning_rate": 1.8339210284089222e-05, "loss": 0.3063, "step": 4216 }, { "epoch": 0.21, "learning_rate": 1.8338317662913277e-05, "loss": 0.28, "step": 4217 }, { "epoch": 0.21, "learning_rate": 1.8337424823659555e-05, "loss": 0.2982, "step": 4218 }, { "epoch": 0.21, "learning_rate": 1.833653176635141e-05, "loss": 0.1243, "step": 4219 }, { "epoch": 0.21, "learning_rate": 1.833563849101219e-05, "loss": 0.3656, "step": 4220 }, { "epoch": 0.21, "learning_rate": 1.8334744997665264e-05, "loss": 0.4817, "step": 4221 }, { "epoch": 0.21, "learning_rate": 1.8333851286334002e-05, "loss": 0.3096, "step": 4222 }, { "epoch": 0.21, "learning_rate": 1.8332957357041773e-05, "loss": 0.4313, "step": 4223 }, { "epoch": 0.21, "learning_rate": 1.8332063209811957e-05, "loss": 0.5725, "step": 4224 }, { "epoch": 0.21, "learning_rate": 1.8331168844667947e-05, "loss": 0.6279, "step": 4225 }, { "epoch": 0.21, "learning_rate": 1.833027426163312e-05, "loss": 0.2741, "step": 4226 }, { "epoch": 0.21, "learning_rate": 1.8329379460730883e-05, "loss": 0.3454, "step": 4227 }, { "epoch": 0.21, "learning_rate": 1.8328484441984637e-05, "loss": 0.3335, "step": 4228 }, { "epoch": 0.21, "learning_rate": 1.832758920541779e-05, "loss": 0.5122, "step": 4229 }, { "epoch": 0.21, "learning_rate": 1.8326693751053753e-05, "loss": 0.1895, "step": 4230 }, { "epoch": 0.21, "learning_rate": 1.8325798078915947e-05, "loss": 0.351, "step": 4231 }, { "epoch": 0.21, "learning_rate": 1.8324902189027796e-05, "loss": 0.2185, "step": 4232 }, { "epoch": 0.21, "learning_rate": 1.8324006081412733e-05, "loss": 0.3659, "step": 4233 }, { "epoch": 0.21, "learning_rate": 1.8323109756094194e-05, "loss": 0.4104, "step": 4234 }, { "epoch": 0.21, "learning_rate": 1.8322213213095618e-05, "loss": 0.4678, "step": 4235 }, { "epoch": 0.21, "learning_rate": 1.8321316452440457e-05, "loss": 0.5423, "step": 4236 }, { "epoch": 0.21, "learning_rate": 1.8320419474152164e-05, "loss": 0.5151, "step": 4237 }, { "epoch": 0.21, "learning_rate": 1.83195222782542e-05, "loss": 0.4584, "step": 4238 }, { "epoch": 0.21, "learning_rate": 1.831862486477002e-05, "loss": 0.5365, "step": 4239 }, { "epoch": 0.21, "learning_rate": 1.831772723372311e-05, "loss": 0.4816, "step": 4240 }, { "epoch": 0.21, "learning_rate": 1.8316829385136933e-05, "loss": 0.5248, "step": 4241 }, { "epoch": 0.21, "learning_rate": 1.831593131903498e-05, "loss": 0.3955, "step": 4242 }, { "epoch": 0.21, "learning_rate": 1.8315033035440733e-05, "loss": 0.502, "step": 4243 }, { "epoch": 0.21, "learning_rate": 1.831413453437769e-05, "loss": 0.4539, "step": 4244 }, { "epoch": 0.21, "learning_rate": 1.8313235815869347e-05, "loss": 0.4188, "step": 4245 }, { "epoch": 0.21, "learning_rate": 1.8312336879939208e-05, "loss": 0.4471, "step": 4246 }, { "epoch": 0.21, "learning_rate": 1.8311437726610785e-05, "loss": 0.7105, "step": 4247 }, { "epoch": 0.21, "learning_rate": 1.8310538355907596e-05, "loss": 0.1732, "step": 4248 }, { "epoch": 0.21, "learning_rate": 1.830963876785316e-05, "loss": 0.5034, "step": 4249 }, { "epoch": 0.21, "learning_rate": 1.8308738962471006e-05, "loss": 0.4939, "step": 4250 }, { "epoch": 0.21, "learning_rate": 1.830783893978467e-05, "loss": 0.3817, "step": 4251 }, { "epoch": 0.21, "learning_rate": 1.8306938699817682e-05, "loss": 0.3747, "step": 4252 }, { "epoch": 0.21, "learning_rate": 1.8306038242593595e-05, "loss": 0.2099, "step": 4253 }, { "epoch": 0.21, "learning_rate": 1.830513756813596e-05, "loss": 0.5346, "step": 4254 }, { "epoch": 0.21, "learning_rate": 1.8304236676468323e-05, "loss": 0.3169, "step": 4255 }, { "epoch": 0.21, "learning_rate": 1.8303335567614254e-05, "loss": 0.3081, "step": 4256 }, { "epoch": 0.21, "learning_rate": 1.8302434241597323e-05, "loss": 0.5396, "step": 4257 }, { "epoch": 0.21, "learning_rate": 1.8301532698441096e-05, "loss": 0.2298, "step": 4258 }, { "epoch": 0.21, "learning_rate": 1.8300630938169157e-05, "loss": 0.3339, "step": 4259 }, { "epoch": 0.21, "learning_rate": 1.8299728960805085e-05, "loss": 0.2152, "step": 4260 }, { "epoch": 0.21, "learning_rate": 1.8298826766372473e-05, "loss": 0.2425, "step": 4261 }, { "epoch": 0.21, "learning_rate": 1.8297924354894915e-05, "loss": 0.3878, "step": 4262 }, { "epoch": 0.21, "learning_rate": 1.8297021726396016e-05, "loss": 0.3844, "step": 4263 }, { "epoch": 0.21, "learning_rate": 1.829611888089938e-05, "loss": 0.4831, "step": 4264 }, { "epoch": 0.21, "learning_rate": 1.8295215818428623e-05, "loss": 0.3753, "step": 4265 }, { "epoch": 0.21, "learning_rate": 1.8294312539007358e-05, "loss": 0.4984, "step": 4266 }, { "epoch": 0.21, "learning_rate": 1.8293409042659213e-05, "loss": 0.4715, "step": 4267 }, { "epoch": 0.21, "learning_rate": 1.829250532940782e-05, "loss": 0.577, "step": 4268 }, { "epoch": 0.21, "learning_rate": 1.8291601399276805e-05, "loss": 0.3609, "step": 4269 }, { "epoch": 0.21, "learning_rate": 1.829069725228982e-05, "loss": 0.2161, "step": 4270 }, { "epoch": 0.21, "learning_rate": 1.8289792888470504e-05, "loss": 0.3745, "step": 4271 }, { "epoch": 0.21, "learning_rate": 1.8288888307842514e-05, "loss": 0.4583, "step": 4272 }, { "epoch": 0.21, "learning_rate": 1.8287983510429506e-05, "loss": 0.4565, "step": 4273 }, { "epoch": 0.21, "learning_rate": 1.828707849625515e-05, "loss": 0.1949, "step": 4274 }, { "epoch": 0.21, "learning_rate": 1.8286173265343107e-05, "loss": 0.1956, "step": 4275 }, { "epoch": 0.21, "learning_rate": 1.8285267817717055e-05, "loss": 0.6545, "step": 4276 }, { "epoch": 0.21, "learning_rate": 1.8284362153400672e-05, "loss": 0.3754, "step": 4277 }, { "epoch": 0.21, "learning_rate": 1.828345627241765e-05, "loss": 0.2577, "step": 4278 }, { "epoch": 0.21, "learning_rate": 1.828255017479168e-05, "loss": 0.3059, "step": 4279 }, { "epoch": 0.21, "learning_rate": 1.8281643860546456e-05, "loss": 0.1502, "step": 4280 }, { "epoch": 0.21, "learning_rate": 1.828073732970568e-05, "loss": 0.497, "step": 4281 }, { "epoch": 0.21, "learning_rate": 1.8279830582293073e-05, "loss": 0.4935, "step": 4282 }, { "epoch": 0.21, "learning_rate": 1.8278923618332335e-05, "loss": 0.4083, "step": 4283 }, { "epoch": 0.21, "learning_rate": 1.8278016437847198e-05, "loss": 0.2346, "step": 4284 }, { "epoch": 0.21, "learning_rate": 1.8277109040861384e-05, "loss": 0.3911, "step": 4285 }, { "epoch": 0.21, "learning_rate": 1.827620142739862e-05, "loss": 0.4441, "step": 4286 }, { "epoch": 0.21, "learning_rate": 1.827529359748265e-05, "loss": 0.3665, "step": 4287 }, { "epoch": 0.21, "learning_rate": 1.827438555113721e-05, "loss": 0.4439, "step": 4288 }, { "epoch": 0.21, "learning_rate": 1.8273477288386063e-05, "loss": 0.232, "step": 4289 }, { "epoch": 0.21, "learning_rate": 1.8272568809252948e-05, "loss": 0.2899, "step": 4290 }, { "epoch": 0.21, "learning_rate": 1.8271660113761628e-05, "loss": 0.304, "step": 4291 }, { "epoch": 0.21, "learning_rate": 1.8270751201935877e-05, "loss": 0.3565, "step": 4292 }, { "epoch": 0.21, "learning_rate": 1.8269842073799457e-05, "loss": 0.4957, "step": 4293 }, { "epoch": 0.21, "learning_rate": 1.8268932729376156e-05, "loss": 0.3505, "step": 4294 }, { "epoch": 0.21, "learning_rate": 1.826802316868974e-05, "loss": 0.3786, "step": 4295 }, { "epoch": 0.21, "learning_rate": 1.8267113391764013e-05, "loss": 0.4729, "step": 4296 }, { "epoch": 0.21, "learning_rate": 1.8266203398622764e-05, "loss": 0.1935, "step": 4297 }, { "epoch": 0.21, "learning_rate": 1.8265293189289787e-05, "loss": 0.367, "step": 4298 }, { "epoch": 0.21, "learning_rate": 1.8264382763788895e-05, "loss": 0.1698, "step": 4299 }, { "epoch": 0.21, "learning_rate": 1.8263472122143898e-05, "loss": 0.3888, "step": 4300 }, { "epoch": 0.21, "learning_rate": 1.826256126437861e-05, "loss": 0.4487, "step": 4301 }, { "epoch": 0.21, "learning_rate": 1.8261650190516854e-05, "loss": 0.3202, "step": 4302 }, { "epoch": 0.21, "learning_rate": 1.826073890058246e-05, "loss": 0.3182, "step": 4303 }, { "epoch": 0.21, "learning_rate": 1.8259827394599255e-05, "loss": 0.3683, "step": 4304 }, { "epoch": 0.21, "learning_rate": 1.8258915672591086e-05, "loss": 0.2081, "step": 4305 }, { "epoch": 0.22, "learning_rate": 1.8258003734581794e-05, "loss": 0.4578, "step": 4306 }, { "epoch": 0.22, "learning_rate": 1.825709158059523e-05, "loss": 0.4568, "step": 4307 }, { "epoch": 0.22, "learning_rate": 1.8256179210655253e-05, "loss": 0.3014, "step": 4308 }, { "epoch": 0.22, "learning_rate": 1.8255266624785717e-05, "loss": 0.358, "step": 4309 }, { "epoch": 0.22, "learning_rate": 1.8254353823010497e-05, "loss": 0.4967, "step": 4310 }, { "epoch": 0.22, "learning_rate": 1.8253440805353468e-05, "loss": 0.2467, "step": 4311 }, { "epoch": 0.22, "learning_rate": 1.82525275718385e-05, "loss": 0.4054, "step": 4312 }, { "epoch": 0.22, "learning_rate": 1.8251614122489483e-05, "loss": 0.3855, "step": 4313 }, { "epoch": 0.22, "learning_rate": 1.825070045733031e-05, "loss": 0.435, "step": 4314 }, { "epoch": 0.22, "learning_rate": 1.824978657638487e-05, "loss": 0.5358, "step": 4315 }, { "epoch": 0.22, "learning_rate": 1.824887247967707e-05, "loss": 0.5159, "step": 4316 }, { "epoch": 0.22, "learning_rate": 1.8247958167230813e-05, "loss": 0.5556, "step": 4317 }, { "epoch": 0.22, "learning_rate": 1.824704363907001e-05, "loss": 0.5438, "step": 4318 }, { "epoch": 0.22, "learning_rate": 1.8246128895218583e-05, "loss": 0.1294, "step": 4319 }, { "epoch": 0.22, "learning_rate": 1.824521393570046e-05, "loss": 0.451, "step": 4320 }, { "epoch": 0.22, "learning_rate": 1.8244298760539562e-05, "loss": 0.2044, "step": 4321 }, { "epoch": 0.22, "learning_rate": 1.8243383369759828e-05, "loss": 0.6013, "step": 4322 }, { "epoch": 0.22, "learning_rate": 1.8242467763385198e-05, "loss": 0.4016, "step": 4323 }, { "epoch": 0.22, "learning_rate": 1.8241551941439622e-05, "loss": 0.2896, "step": 4324 }, { "epoch": 0.22, "learning_rate": 1.824063590394705e-05, "loss": 0.3506, "step": 4325 }, { "epoch": 0.22, "learning_rate": 1.8239719650931436e-05, "loss": 0.4991, "step": 4326 }, { "epoch": 0.22, "learning_rate": 1.823880318241675e-05, "loss": 0.409, "step": 4327 }, { "epoch": 0.22, "learning_rate": 1.8237886498426955e-05, "loss": 0.5567, "step": 4328 }, { "epoch": 0.22, "learning_rate": 1.823696959898603e-05, "loss": 0.3427, "step": 4329 }, { "epoch": 0.22, "learning_rate": 1.8236052484117953e-05, "loss": 0.3925, "step": 4330 }, { "epoch": 0.22, "learning_rate": 1.823513515384671e-05, "loss": 0.5964, "step": 4331 }, { "epoch": 0.22, "learning_rate": 1.8234217608196292e-05, "loss": 0.1378, "step": 4332 }, { "epoch": 0.22, "learning_rate": 1.8233299847190703e-05, "loss": 0.2485, "step": 4333 }, { "epoch": 0.22, "learning_rate": 1.8232381870853936e-05, "loss": 0.2992, "step": 4334 }, { "epoch": 0.22, "learning_rate": 1.823146367921e-05, "loss": 0.4352, "step": 4335 }, { "epoch": 0.22, "learning_rate": 1.823054527228292e-05, "loss": 0.3156, "step": 4336 }, { "epoch": 0.22, "learning_rate": 1.82296266500967e-05, "loss": 0.3126, "step": 4337 }, { "epoch": 0.22, "learning_rate": 1.8228707812675383e-05, "loss": 0.2999, "step": 4338 }, { "epoch": 0.22, "learning_rate": 1.8227788760042984e-05, "loss": 0.1665, "step": 4339 }, { "epoch": 0.22, "learning_rate": 1.822686949222355e-05, "loss": 0.415, "step": 4340 }, { "epoch": 0.22, "learning_rate": 1.8225950009241117e-05, "loss": 0.5091, "step": 4341 }, { "epoch": 0.22, "learning_rate": 1.822503031111974e-05, "loss": 0.3712, "step": 4342 }, { "epoch": 0.22, "learning_rate": 1.8224110397883462e-05, "loss": 0.4, "step": 4343 }, { "epoch": 0.22, "learning_rate": 1.8223190269556355e-05, "loss": 0.379, "step": 4344 }, { "epoch": 0.22, "learning_rate": 1.8222269926162472e-05, "loss": 0.3558, "step": 4345 }, { "epoch": 0.22, "learning_rate": 1.822134936772589e-05, "loss": 0.5974, "step": 4346 }, { "epoch": 0.22, "learning_rate": 1.8220428594270685e-05, "loss": 0.4279, "step": 4347 }, { "epoch": 0.22, "learning_rate": 1.8219507605820935e-05, "loss": 0.4319, "step": 4348 }, { "epoch": 0.22, "learning_rate": 1.821858640240073e-05, "loss": 0.4521, "step": 4349 }, { "epoch": 0.22, "learning_rate": 1.821766498403416e-05, "loss": 0.1955, "step": 4350 }, { "epoch": 0.22, "learning_rate": 1.8216743350745328e-05, "loss": 0.2897, "step": 4351 }, { "epoch": 0.22, "learning_rate": 1.8215821502558332e-05, "loss": 0.4613, "step": 4352 }, { "epoch": 0.22, "learning_rate": 1.821489943949729e-05, "loss": 0.4739, "step": 4353 }, { "epoch": 0.22, "learning_rate": 1.821397716158631e-05, "loss": 0.2737, "step": 4354 }, { "epoch": 0.22, "learning_rate": 1.8213054668849515e-05, "loss": 0.2901, "step": 4355 }, { "epoch": 0.22, "learning_rate": 1.8212131961311034e-05, "loss": 0.1595, "step": 4356 }, { "epoch": 0.22, "learning_rate": 1.8211209038995e-05, "loss": 0.4365, "step": 4357 }, { "epoch": 0.22, "learning_rate": 1.8210285901925544e-05, "loss": 0.3722, "step": 4358 }, { "epoch": 0.22, "learning_rate": 1.8209362550126813e-05, "loss": 0.6374, "step": 4359 }, { "epoch": 0.22, "learning_rate": 1.8208438983622964e-05, "loss": 0.2979, "step": 4360 }, { "epoch": 0.22, "learning_rate": 1.8207515202438136e-05, "loss": 0.3656, "step": 4361 }, { "epoch": 0.22, "learning_rate": 1.8206591206596504e-05, "loss": 0.4812, "step": 4362 }, { "epoch": 0.22, "learning_rate": 1.8205666996122223e-05, "loss": 0.247, "step": 4363 }, { "epoch": 0.22, "learning_rate": 1.8204742571039472e-05, "loss": 0.4099, "step": 4364 }, { "epoch": 0.22, "learning_rate": 1.8203817931372424e-05, "loss": 0.2179, "step": 4365 }, { "epoch": 0.22, "learning_rate": 1.8202893077145266e-05, "loss": 0.2845, "step": 4366 }, { "epoch": 0.22, "learning_rate": 1.820196800838218e-05, "loss": 0.5651, "step": 4367 }, { "epoch": 0.22, "learning_rate": 1.8201042725107365e-05, "loss": 0.3368, "step": 4368 }, { "epoch": 0.22, "learning_rate": 1.8200117227345017e-05, "loss": 0.5738, "step": 4369 }, { "epoch": 0.22, "learning_rate": 1.819919151511935e-05, "loss": 0.4323, "step": 4370 }, { "epoch": 0.22, "learning_rate": 1.8198265588454557e-05, "loss": 0.3441, "step": 4371 }, { "epoch": 0.22, "learning_rate": 1.8197339447374874e-05, "loss": 0.3765, "step": 4372 }, { "epoch": 0.22, "learning_rate": 1.819641309190451e-05, "loss": 0.4345, "step": 4373 }, { "epoch": 0.22, "learning_rate": 1.8195486522067697e-05, "loss": 0.4859, "step": 4374 }, { "epoch": 0.22, "learning_rate": 1.819455973788867e-05, "loss": 0.2406, "step": 4375 }, { "epoch": 0.22, "learning_rate": 1.8193632739391667e-05, "loss": 0.2981, "step": 4376 }, { "epoch": 0.22, "learning_rate": 1.819270552660093e-05, "loss": 0.4651, "step": 4377 }, { "epoch": 0.22, "learning_rate": 1.819177809954071e-05, "loss": 0.2886, "step": 4378 }, { "epoch": 0.22, "learning_rate": 1.8190850458235262e-05, "loss": 0.1723, "step": 4379 }, { "epoch": 0.22, "learning_rate": 1.8189922602708852e-05, "loss": 0.2598, "step": 4380 }, { "epoch": 0.22, "learning_rate": 1.818899453298574e-05, "loss": 0.4692, "step": 4381 }, { "epoch": 0.22, "learning_rate": 1.81880662490902e-05, "loss": 0.2318, "step": 4382 }, { "epoch": 0.22, "learning_rate": 1.818713775104652e-05, "loss": 0.1991, "step": 4383 }, { "epoch": 0.22, "learning_rate": 1.8186209038878968e-05, "loss": 0.3901, "step": 4384 }, { "epoch": 0.22, "learning_rate": 1.818528011261184e-05, "loss": 0.5861, "step": 4385 }, { "epoch": 0.22, "learning_rate": 1.8184350972269436e-05, "loss": 0.1991, "step": 4386 }, { "epoch": 0.22, "learning_rate": 1.818342161787605e-05, "loss": 0.3528, "step": 4387 }, { "epoch": 0.22, "learning_rate": 1.818249204945599e-05, "loss": 0.263, "step": 4388 }, { "epoch": 0.22, "learning_rate": 1.8181562267033566e-05, "loss": 0.3348, "step": 4389 }, { "epoch": 0.22, "learning_rate": 1.8180632270633096e-05, "loss": 0.3572, "step": 4390 }, { "epoch": 0.22, "learning_rate": 1.8179702060278907e-05, "loss": 0.4773, "step": 4391 }, { "epoch": 0.22, "learning_rate": 1.8178771635995322e-05, "loss": 0.5353, "step": 4392 }, { "epoch": 0.22, "learning_rate": 1.817784099780668e-05, "loss": 0.4554, "step": 4393 }, { "epoch": 0.22, "learning_rate": 1.8176910145737317e-05, "loss": 0.3426, "step": 4394 }, { "epoch": 0.22, "learning_rate": 1.8175979079811577e-05, "loss": 0.4468, "step": 4395 }, { "epoch": 0.22, "learning_rate": 1.8175047800053813e-05, "loss": 0.4498, "step": 4396 }, { "epoch": 0.22, "learning_rate": 1.8174116306488382e-05, "loss": 0.4806, "step": 4397 }, { "epoch": 0.22, "learning_rate": 1.817318459913964e-05, "loss": 0.4711, "step": 4398 }, { "epoch": 0.22, "learning_rate": 1.817225267803197e-05, "loss": 0.6372, "step": 4399 }, { "epoch": 0.22, "learning_rate": 1.8171320543189725e-05, "loss": 0.2425, "step": 4400 }, { "epoch": 0.22, "learning_rate": 1.81703881946373e-05, "loss": 0.2988, "step": 4401 }, { "epoch": 0.22, "learning_rate": 1.816945563239907e-05, "loss": 0.2857, "step": 4402 }, { "epoch": 0.22, "learning_rate": 1.8168522856499432e-05, "loss": 0.1807, "step": 4403 }, { "epoch": 0.22, "learning_rate": 1.8167589866962774e-05, "loss": 0.4228, "step": 4404 }, { "epoch": 0.22, "learning_rate": 1.81666566638135e-05, "loss": 0.2272, "step": 4405 }, { "epoch": 0.22, "learning_rate": 1.816572324707602e-05, "loss": 0.321, "step": 4406 }, { "epoch": 0.22, "learning_rate": 1.8164789616774744e-05, "loss": 0.1982, "step": 4407 }, { "epoch": 0.22, "learning_rate": 1.8163855772934086e-05, "loss": 0.529, "step": 4408 }, { "epoch": 0.22, "learning_rate": 1.8162921715578475e-05, "loss": 0.242, "step": 4409 }, { "epoch": 0.22, "learning_rate": 1.816198744473234e-05, "loss": 0.1976, "step": 4410 }, { "epoch": 0.22, "learning_rate": 1.8161052960420108e-05, "loss": 0.2793, "step": 4411 }, { "epoch": 0.22, "learning_rate": 1.816011826266623e-05, "loss": 0.4082, "step": 4412 }, { "epoch": 0.22, "learning_rate": 1.8159183351495143e-05, "loss": 0.4573, "step": 4413 }, { "epoch": 0.22, "learning_rate": 1.8158248226931304e-05, "loss": 0.26, "step": 4414 }, { "epoch": 0.22, "learning_rate": 1.815731288899917e-05, "loss": 0.2883, "step": 4415 }, { "epoch": 0.22, "learning_rate": 1.8156377337723198e-05, "loss": 0.3468, "step": 4416 }, { "epoch": 0.22, "learning_rate": 1.8155441573127864e-05, "loss": 0.4546, "step": 4417 }, { "epoch": 0.22, "learning_rate": 1.8154505595237636e-05, "loss": 0.2126, "step": 4418 }, { "epoch": 0.22, "learning_rate": 1.8153569404076993e-05, "loss": 0.335, "step": 4419 }, { "epoch": 0.22, "learning_rate": 1.815263299967042e-05, "loss": 0.6438, "step": 4420 }, { "epoch": 0.22, "learning_rate": 1.8151696382042413e-05, "loss": 0.1927, "step": 4421 }, { "epoch": 0.22, "learning_rate": 1.815075955121746e-05, "loss": 0.3004, "step": 4422 }, { "epoch": 0.22, "learning_rate": 1.8149822507220067e-05, "loss": 0.3963, "step": 4423 }, { "epoch": 0.22, "learning_rate": 1.8148885250074745e-05, "loss": 0.4981, "step": 4424 }, { "epoch": 0.22, "learning_rate": 1.8147947779805997e-05, "loss": 0.1063, "step": 4425 }, { "epoch": 0.22, "learning_rate": 1.814701009643835e-05, "loss": 0.5028, "step": 4426 }, { "epoch": 0.22, "learning_rate": 1.8146072199996323e-05, "loss": 0.3844, "step": 4427 }, { "epoch": 0.22, "learning_rate": 1.8145134090504447e-05, "loss": 0.4486, "step": 4428 }, { "epoch": 0.22, "learning_rate": 1.8144195767987258e-05, "loss": 0.5169, "step": 4429 }, { "epoch": 0.22, "learning_rate": 1.81432572324693e-05, "loss": 0.3819, "step": 4430 }, { "epoch": 0.22, "learning_rate": 1.8142318483975107e-05, "loss": 0.5901, "step": 4431 }, { "epoch": 0.22, "learning_rate": 1.814137952252924e-05, "loss": 0.6924, "step": 4432 }, { "epoch": 0.22, "learning_rate": 1.8140440348156257e-05, "loss": 0.4225, "step": 4433 }, { "epoch": 0.22, "learning_rate": 1.813950096088072e-05, "loss": 0.5168, "step": 4434 }, { "epoch": 0.22, "learning_rate": 1.8138561360727195e-05, "loss": 0.3532, "step": 4435 }, { "epoch": 0.22, "learning_rate": 1.8137621547720257e-05, "loss": 0.2146, "step": 4436 }, { "epoch": 0.22, "learning_rate": 1.8136681521884483e-05, "loss": 0.4101, "step": 4437 }, { "epoch": 0.22, "learning_rate": 1.8135741283244465e-05, "loss": 0.5647, "step": 4438 }, { "epoch": 0.22, "learning_rate": 1.8134800831824787e-05, "loss": 0.2527, "step": 4439 }, { "epoch": 0.22, "learning_rate": 1.813386016765005e-05, "loss": 0.4042, "step": 4440 }, { "epoch": 0.22, "learning_rate": 1.8132919290744854e-05, "loss": 0.2154, "step": 4441 }, { "epoch": 0.22, "learning_rate": 1.8131978201133806e-05, "loss": 0.3443, "step": 4442 }, { "epoch": 0.22, "learning_rate": 1.8131036898841516e-05, "loss": 0.4536, "step": 4443 }, { "epoch": 0.22, "learning_rate": 1.813009538389261e-05, "loss": 0.3081, "step": 4444 }, { "epoch": 0.22, "learning_rate": 1.8129153656311703e-05, "loss": 0.3826, "step": 4445 }, { "epoch": 0.22, "learning_rate": 1.8128211716123433e-05, "loss": 0.5144, "step": 4446 }, { "epoch": 0.22, "learning_rate": 1.812726956335243e-05, "loss": 0.3995, "step": 4447 }, { "epoch": 0.22, "learning_rate": 1.812632719802334e-05, "loss": 0.4006, "step": 4448 }, { "epoch": 0.22, "learning_rate": 1.81253846201608e-05, "loss": 0.3985, "step": 4449 }, { "epoch": 0.22, "learning_rate": 1.812444182978947e-05, "loss": 0.4349, "step": 4450 }, { "epoch": 0.22, "learning_rate": 1.812349882693401e-05, "loss": 0.2373, "step": 4451 }, { "epoch": 0.22, "learning_rate": 1.812255561161907e-05, "loss": 0.2369, "step": 4452 }, { "epoch": 0.22, "learning_rate": 1.812161218386933e-05, "loss": 0.3817, "step": 4453 }, { "epoch": 0.22, "learning_rate": 1.8120668543709463e-05, "loss": 0.4998, "step": 4454 }, { "epoch": 0.22, "learning_rate": 1.811972469116414e-05, "loss": 0.4749, "step": 4455 }, { "epoch": 0.22, "learning_rate": 1.811878062625806e-05, "loss": 0.5623, "step": 4456 }, { "epoch": 0.22, "learning_rate": 1.8117836349015904e-05, "loss": 0.2235, "step": 4457 }, { "epoch": 0.22, "learning_rate": 1.8116891859462368e-05, "loss": 0.2148, "step": 4458 }, { "epoch": 0.22, "learning_rate": 1.8115947157622158e-05, "loss": 0.3007, "step": 4459 }, { "epoch": 0.22, "learning_rate": 1.811500224351998e-05, "loss": 0.3477, "step": 4460 }, { "epoch": 0.22, "learning_rate": 1.811405711718055e-05, "loss": 0.3289, "step": 4461 }, { "epoch": 0.22, "learning_rate": 1.8113111778628584e-05, "loss": 0.3767, "step": 4462 }, { "epoch": 0.22, "learning_rate": 1.81121662278888e-05, "loss": 0.4977, "step": 4463 }, { "epoch": 0.22, "learning_rate": 1.8111220464985938e-05, "loss": 0.3951, "step": 4464 }, { "epoch": 0.22, "learning_rate": 1.8110274489944725e-05, "loss": 0.3621, "step": 4465 }, { "epoch": 0.22, "learning_rate": 1.8109328302789907e-05, "loss": 0.1587, "step": 4466 }, { "epoch": 0.22, "learning_rate": 1.810838190354623e-05, "loss": 0.3718, "step": 4467 }, { "epoch": 0.22, "learning_rate": 1.8107435292238443e-05, "loss": 0.2903, "step": 4468 }, { "epoch": 0.22, "learning_rate": 1.8106488468891304e-05, "loss": 0.1677, "step": 4469 }, { "epoch": 0.22, "learning_rate": 1.810554143352958e-05, "loss": 0.3361, "step": 4470 }, { "epoch": 0.22, "learning_rate": 1.810459418617803e-05, "loss": 0.3912, "step": 4471 }, { "epoch": 0.22, "learning_rate": 1.810364672686144e-05, "loss": 0.3275, "step": 4472 }, { "epoch": 0.22, "learning_rate": 1.810269905560458e-05, "loss": 0.4003, "step": 4473 }, { "epoch": 0.22, "learning_rate": 1.810175117243224e-05, "loss": 0.5063, "step": 4474 }, { "epoch": 0.22, "learning_rate": 1.8100803077369212e-05, "loss": 0.1459, "step": 4475 }, { "epoch": 0.22, "learning_rate": 1.8099854770440285e-05, "loss": 0.3918, "step": 4476 }, { "epoch": 0.22, "learning_rate": 1.809890625167027e-05, "loss": 0.489, "step": 4477 }, { "epoch": 0.22, "learning_rate": 1.8097957521083967e-05, "loss": 0.2457, "step": 4478 }, { "epoch": 0.22, "learning_rate": 1.8097008578706196e-05, "loss": 0.3227, "step": 4479 }, { "epoch": 0.22, "learning_rate": 1.8096059424561762e-05, "loss": 0.3824, "step": 4480 }, { "epoch": 0.22, "learning_rate": 1.8095110058675503e-05, "loss": 0.4061, "step": 4481 }, { "epoch": 0.22, "learning_rate": 1.809416048107224e-05, "loss": 0.213, "step": 4482 }, { "epoch": 0.22, "learning_rate": 1.8093210691776818e-05, "loss": 0.255, "step": 4483 }, { "epoch": 0.22, "learning_rate": 1.8092260690814065e-05, "loss": 0.4113, "step": 4484 }, { "epoch": 0.22, "learning_rate": 1.8091310478208833e-05, "loss": 0.2926, "step": 4485 }, { "epoch": 0.22, "learning_rate": 1.8090360053985976e-05, "loss": 0.3424, "step": 4486 }, { "epoch": 0.22, "learning_rate": 1.8089409418170345e-05, "loss": 0.3699, "step": 4487 }, { "epoch": 0.22, "learning_rate": 1.808845857078681e-05, "loss": 0.6413, "step": 4488 }, { "epoch": 0.22, "learning_rate": 1.808750751186023e-05, "loss": 0.3619, "step": 4489 }, { "epoch": 0.22, "learning_rate": 1.8086556241415486e-05, "loss": 0.2262, "step": 4490 }, { "epoch": 0.22, "learning_rate": 1.8085604759477455e-05, "loss": 0.4831, "step": 4491 }, { "epoch": 0.22, "learning_rate": 1.8084653066071024e-05, "loss": 0.3663, "step": 4492 }, { "epoch": 0.22, "learning_rate": 1.8083701161221077e-05, "loss": 0.6879, "step": 4493 }, { "epoch": 0.22, "learning_rate": 1.8082749044952516e-05, "loss": 0.4953, "step": 4494 }, { "epoch": 0.22, "learning_rate": 1.808179671729024e-05, "loss": 0.3158, "step": 4495 }, { "epoch": 0.22, "learning_rate": 1.8080844178259153e-05, "loss": 0.3729, "step": 4496 }, { "epoch": 0.22, "learning_rate": 1.8079891427884177e-05, "loss": 0.6602, "step": 4497 }, { "epoch": 0.22, "learning_rate": 1.807893846619022e-05, "loss": 0.3932, "step": 4498 }, { "epoch": 0.22, "learning_rate": 1.8077985293202207e-05, "loss": 0.3146, "step": 4499 }, { "epoch": 0.22, "learning_rate": 1.807703190894507e-05, "loss": 0.3394, "step": 4500 }, { "epoch": 0.22, "learning_rate": 1.8076078313443744e-05, "loss": 0.3777, "step": 4501 }, { "epoch": 0.22, "learning_rate": 1.8075124506723168e-05, "loss": 0.4904, "step": 4502 }, { "epoch": 0.22, "learning_rate": 1.8074170488808286e-05, "loss": 0.4495, "step": 4503 }, { "epoch": 0.22, "learning_rate": 1.8073216259724048e-05, "loss": 0.3067, "step": 4504 }, { "epoch": 0.22, "learning_rate": 1.8072261819495415e-05, "loss": 0.3704, "step": 4505 }, { "epoch": 0.22, "learning_rate": 1.8071307168147346e-05, "loss": 0.2457, "step": 4506 }, { "epoch": 0.23, "learning_rate": 1.807035230570481e-05, "loss": 0.3924, "step": 4507 }, { "epoch": 0.23, "learning_rate": 1.8069397232192783e-05, "loss": 0.3612, "step": 4508 }, { "epoch": 0.23, "learning_rate": 1.8068441947636236e-05, "loss": 0.1198, "step": 4509 }, { "epoch": 0.23, "learning_rate": 1.806748645206016e-05, "loss": 0.465, "step": 4510 }, { "epoch": 0.23, "learning_rate": 1.806653074548954e-05, "loss": 0.353, "step": 4511 }, { "epoch": 0.23, "learning_rate": 1.8065574827949374e-05, "loss": 0.5209, "step": 4512 }, { "epoch": 0.23, "learning_rate": 1.8064618699464662e-05, "loss": 0.3637, "step": 4513 }, { "epoch": 0.23, "learning_rate": 1.806366236006041e-05, "loss": 0.5244, "step": 4514 }, { "epoch": 0.23, "learning_rate": 1.8062705809761633e-05, "loss": 0.4435, "step": 4515 }, { "epoch": 0.23, "learning_rate": 1.8061749048593346e-05, "loss": 0.3076, "step": 4516 }, { "epoch": 0.23, "learning_rate": 1.806079207658057e-05, "loss": 0.4983, "step": 4517 }, { "epoch": 0.23, "learning_rate": 1.8059834893748338e-05, "loss": 0.3402, "step": 4518 }, { "epoch": 0.23, "learning_rate": 1.8058877500121677e-05, "loss": 0.4856, "step": 4519 }, { "epoch": 0.23, "learning_rate": 1.8057919895725633e-05, "loss": 0.5664, "step": 4520 }, { "epoch": 0.23, "learning_rate": 1.8056962080585244e-05, "loss": 0.4433, "step": 4521 }, { "epoch": 0.23, "learning_rate": 1.805600405472557e-05, "loss": 0.3353, "step": 4522 }, { "epoch": 0.23, "learning_rate": 1.805504581817166e-05, "loss": 0.4499, "step": 4523 }, { "epoch": 0.23, "learning_rate": 1.8054087370948574e-05, "loss": 0.3272, "step": 4524 }, { "epoch": 0.23, "learning_rate": 1.8053128713081383e-05, "loss": 0.3407, "step": 4525 }, { "epoch": 0.23, "learning_rate": 1.805216984459516e-05, "loss": 0.4864, "step": 4526 }, { "epoch": 0.23, "learning_rate": 1.805121076551498e-05, "loss": 0.3904, "step": 4527 }, { "epoch": 0.23, "learning_rate": 1.8050251475865927e-05, "loss": 0.4529, "step": 4528 }, { "epoch": 0.23, "learning_rate": 1.804929197567309e-05, "loss": 0.2409, "step": 4529 }, { "epoch": 0.23, "learning_rate": 1.8048332264961564e-05, "loss": 0.2909, "step": 4530 }, { "epoch": 0.23, "learning_rate": 1.804737234375645e-05, "loss": 0.176, "step": 4531 }, { "epoch": 0.23, "learning_rate": 1.8046412212082854e-05, "loss": 0.4737, "step": 4532 }, { "epoch": 0.23, "learning_rate": 1.804545186996588e-05, "loss": 0.3607, "step": 4533 }, { "epoch": 0.23, "learning_rate": 1.804449131743066e-05, "loss": 0.4436, "step": 4534 }, { "epoch": 0.23, "learning_rate": 1.80435305545023e-05, "loss": 0.2598, "step": 4535 }, { "epoch": 0.23, "learning_rate": 1.8042569581205932e-05, "loss": 0.5919, "step": 4536 }, { "epoch": 0.23, "learning_rate": 1.8041608397566693e-05, "loss": 0.6284, "step": 4537 }, { "epoch": 0.23, "learning_rate": 1.8040647003609716e-05, "loss": 0.2018, "step": 4538 }, { "epoch": 0.23, "learning_rate": 1.803968539936015e-05, "loss": 0.4886, "step": 4539 }, { "epoch": 0.23, "learning_rate": 1.803872358484314e-05, "loss": 0.5841, "step": 4540 }, { "epoch": 0.23, "learning_rate": 1.803776156008385e-05, "loss": 0.6605, "step": 4541 }, { "epoch": 0.23, "learning_rate": 1.8036799325107427e-05, "loss": 0.5661, "step": 4542 }, { "epoch": 0.23, "learning_rate": 1.803583687993905e-05, "loss": 0.5871, "step": 4543 }, { "epoch": 0.23, "learning_rate": 1.803487422460388e-05, "loss": 0.3262, "step": 4544 }, { "epoch": 0.23, "learning_rate": 1.80339113591271e-05, "loss": 0.2638, "step": 4545 }, { "epoch": 0.23, "learning_rate": 1.8032948283533887e-05, "loss": 0.2589, "step": 4546 }, { "epoch": 0.23, "learning_rate": 1.803198499784944e-05, "loss": 0.2977, "step": 4547 }, { "epoch": 0.23, "learning_rate": 1.803102150209894e-05, "loss": 0.3715, "step": 4548 }, { "epoch": 0.23, "learning_rate": 1.8030057796307594e-05, "loss": 0.3659, "step": 4549 }, { "epoch": 0.23, "learning_rate": 1.8029093880500602e-05, "loss": 0.2014, "step": 4550 }, { "epoch": 0.23, "learning_rate": 1.802812975470318e-05, "loss": 0.3542, "step": 4551 }, { "epoch": 0.23, "learning_rate": 1.8027165418940534e-05, "loss": 0.271, "step": 4552 }, { "epoch": 0.23, "learning_rate": 1.8026200873237894e-05, "loss": 0.3789, "step": 4553 }, { "epoch": 0.23, "learning_rate": 1.802523611762048e-05, "loss": 0.4762, "step": 4554 }, { "epoch": 0.23, "learning_rate": 1.8024271152113528e-05, "loss": 0.3185, "step": 4555 }, { "epoch": 0.23, "learning_rate": 1.802330597674227e-05, "loss": 0.3011, "step": 4556 }, { "epoch": 0.23, "learning_rate": 1.802234059153196e-05, "loss": 0.6422, "step": 4557 }, { "epoch": 0.23, "learning_rate": 1.8021374996507834e-05, "loss": 0.5851, "step": 4558 }, { "epoch": 0.23, "learning_rate": 1.8020409191695153e-05, "loss": 0.4694, "step": 4559 }, { "epoch": 0.23, "learning_rate": 1.8019443177119173e-05, "loss": 0.2675, "step": 4560 }, { "epoch": 0.23, "learning_rate": 1.801847695280516e-05, "loss": 0.2598, "step": 4561 }, { "epoch": 0.23, "learning_rate": 1.801751051877839e-05, "loss": 0.4898, "step": 4562 }, { "epoch": 0.23, "learning_rate": 1.8016543875064124e-05, "loss": 0.412, "step": 4563 }, { "epoch": 0.23, "learning_rate": 1.801557702168766e-05, "loss": 0.3984, "step": 4564 }, { "epoch": 0.23, "learning_rate": 1.8014609958674275e-05, "loss": 0.4557, "step": 4565 }, { "epoch": 0.23, "learning_rate": 1.8013642686049265e-05, "loss": 0.3438, "step": 4566 }, { "epoch": 0.23, "learning_rate": 1.8012675203837926e-05, "loss": 0.3577, "step": 4567 }, { "epoch": 0.23, "learning_rate": 1.801170751206556e-05, "loss": 0.2784, "step": 4568 }, { "epoch": 0.23, "learning_rate": 1.801073961075748e-05, "loss": 0.5882, "step": 4569 }, { "epoch": 0.23, "learning_rate": 1.8009771499938997e-05, "loss": 0.4141, "step": 4570 }, { "epoch": 0.23, "learning_rate": 1.800880317963543e-05, "loss": 0.3804, "step": 4571 }, { "epoch": 0.23, "learning_rate": 1.8007834649872107e-05, "loss": 0.255, "step": 4572 }, { "epoch": 0.23, "learning_rate": 1.8006865910674357e-05, "loss": 0.1787, "step": 4573 }, { "epoch": 0.23, "learning_rate": 1.8005896962067515e-05, "loss": 0.4001, "step": 4574 }, { "epoch": 0.23, "learning_rate": 1.8004927804076927e-05, "loss": 0.242, "step": 4575 }, { "epoch": 0.23, "learning_rate": 1.8003958436727935e-05, "loss": 0.4632, "step": 4576 }, { "epoch": 0.23, "learning_rate": 1.800298886004589e-05, "loss": 0.3643, "step": 4577 }, { "epoch": 0.23, "learning_rate": 1.8002019074056155e-05, "loss": 0.3417, "step": 4578 }, { "epoch": 0.23, "learning_rate": 1.8001049078784096e-05, "loss": 0.5265, "step": 4579 }, { "epoch": 0.23, "learning_rate": 1.8000078874255072e-05, "loss": 0.3417, "step": 4580 }, { "epoch": 0.23, "learning_rate": 1.7999108460494466e-05, "loss": 0.4894, "step": 4581 }, { "epoch": 0.23, "learning_rate": 1.799813783752766e-05, "loss": 0.3973, "step": 4582 }, { "epoch": 0.23, "learning_rate": 1.799716700538003e-05, "loss": 0.5029, "step": 4583 }, { "epoch": 0.23, "learning_rate": 1.799619596407697e-05, "loss": 0.3279, "step": 4584 }, { "epoch": 0.23, "learning_rate": 1.7995224713643878e-05, "loss": 0.2992, "step": 4585 }, { "epoch": 0.23, "learning_rate": 1.7994253254106157e-05, "loss": 0.3076, "step": 4586 }, { "epoch": 0.23, "learning_rate": 1.799328158548921e-05, "loss": 0.3648, "step": 4587 }, { "epoch": 0.23, "learning_rate": 1.799230970781846e-05, "loss": 0.3567, "step": 4588 }, { "epoch": 0.23, "learning_rate": 1.799133762111931e-05, "loss": 0.3833, "step": 4589 }, { "epoch": 0.23, "learning_rate": 1.7990365325417194e-05, "loss": 0.2864, "step": 4590 }, { "epoch": 0.23, "learning_rate": 1.7989392820737537e-05, "loss": 0.2353, "step": 4591 }, { "epoch": 0.23, "learning_rate": 1.7988420107105775e-05, "loss": 0.5575, "step": 4592 }, { "epoch": 0.23, "learning_rate": 1.798744718454735e-05, "loss": 0.2434, "step": 4593 }, { "epoch": 0.23, "learning_rate": 1.7986474053087705e-05, "loss": 0.3336, "step": 4594 }, { "epoch": 0.23, "learning_rate": 1.798550071275229e-05, "loss": 0.4956, "step": 4595 }, { "epoch": 0.23, "learning_rate": 1.7984527163566566e-05, "loss": 0.5982, "step": 4596 }, { "epoch": 0.23, "learning_rate": 1.798355340555599e-05, "loss": 0.4555, "step": 4597 }, { "epoch": 0.23, "learning_rate": 1.798257943874603e-05, "loss": 0.2881, "step": 4598 }, { "epoch": 0.23, "learning_rate": 1.7981605263162164e-05, "loss": 0.2589, "step": 4599 }, { "epoch": 0.23, "learning_rate": 1.798063087882986e-05, "loss": 0.3477, "step": 4600 }, { "epoch": 0.23, "learning_rate": 1.7979656285774615e-05, "loss": 0.536, "step": 4601 }, { "epoch": 0.23, "learning_rate": 1.7978681484021906e-05, "loss": 0.2332, "step": 4602 }, { "epoch": 0.23, "learning_rate": 1.7977706473597238e-05, "loss": 0.2524, "step": 4603 }, { "epoch": 0.23, "learning_rate": 1.7976731254526104e-05, "loss": 0.1969, "step": 4604 }, { "epoch": 0.23, "learning_rate": 1.797575582683401e-05, "loss": 0.4083, "step": 4605 }, { "epoch": 0.23, "learning_rate": 1.7974780190546466e-05, "loss": 0.4852, "step": 4606 }, { "epoch": 0.23, "learning_rate": 1.7973804345688998e-05, "loss": 0.6056, "step": 4607 }, { "epoch": 0.23, "learning_rate": 1.7972828292287118e-05, "loss": 0.3071, "step": 4608 }, { "epoch": 0.23, "learning_rate": 1.7971852030366353e-05, "loss": 0.3919, "step": 4609 }, { "epoch": 0.23, "learning_rate": 1.7970875559952244e-05, "loss": 0.4923, "step": 4610 }, { "epoch": 0.23, "learning_rate": 1.7969898881070326e-05, "loss": 0.2725, "step": 4611 }, { "epoch": 0.23, "learning_rate": 1.796892199374614e-05, "loss": 0.4244, "step": 4612 }, { "epoch": 0.23, "learning_rate": 1.7967944898005233e-05, "loss": 0.1741, "step": 4613 }, { "epoch": 0.23, "learning_rate": 1.7966967593873165e-05, "loss": 0.5446, "step": 4614 }, { "epoch": 0.23, "learning_rate": 1.79659900813755e-05, "loss": 0.2661, "step": 4615 }, { "epoch": 0.23, "learning_rate": 1.796501236053779e-05, "loss": 0.2361, "step": 4616 }, { "epoch": 0.23, "learning_rate": 1.796403443138562e-05, "loss": 0.3776, "step": 4617 }, { "epoch": 0.23, "learning_rate": 1.7963056293944557e-05, "loss": 0.3874, "step": 4618 }, { "epoch": 0.23, "learning_rate": 1.7962077948240187e-05, "loss": 0.3321, "step": 4619 }, { "epoch": 0.23, "learning_rate": 1.7961099394298096e-05, "loss": 0.3158, "step": 4620 }, { "epoch": 0.23, "learning_rate": 1.796012063214388e-05, "loss": 0.4268, "step": 4621 }, { "epoch": 0.23, "learning_rate": 1.7959141661803133e-05, "loss": 0.3203, "step": 4622 }, { "epoch": 0.23, "learning_rate": 1.795816248330146e-05, "loss": 0.3515, "step": 4623 }, { "epoch": 0.23, "learning_rate": 1.7957183096664472e-05, "loss": 0.4003, "step": 4624 }, { "epoch": 0.23, "learning_rate": 1.7956203501917783e-05, "loss": 0.3393, "step": 4625 }, { "epoch": 0.23, "learning_rate": 1.7955223699087008e-05, "loss": 0.2304, "step": 4626 }, { "epoch": 0.23, "learning_rate": 1.7954243688197782e-05, "loss": 0.2647, "step": 4627 }, { "epoch": 0.23, "learning_rate": 1.7953263469275726e-05, "loss": 0.1403, "step": 4628 }, { "epoch": 0.23, "learning_rate": 1.795228304234648e-05, "loss": 0.2746, "step": 4629 }, { "epoch": 0.23, "learning_rate": 1.795130240743569e-05, "loss": 0.3558, "step": 4630 }, { "epoch": 0.23, "learning_rate": 1.7950321564569e-05, "loss": 0.342, "step": 4631 }, { "epoch": 0.23, "learning_rate": 1.7949340513772063e-05, "loss": 0.3465, "step": 4632 }, { "epoch": 0.23, "learning_rate": 1.794835925507053e-05, "loss": 0.4265, "step": 4633 }, { "epoch": 0.23, "learning_rate": 1.7947377788490083e-05, "loss": 0.4946, "step": 4634 }, { "epoch": 0.23, "learning_rate": 1.794639611405637e-05, "loss": 0.3942, "step": 4635 }, { "epoch": 0.23, "learning_rate": 1.7945414231795077e-05, "loss": 0.3314, "step": 4636 }, { "epoch": 0.23, "learning_rate": 1.7944432141731883e-05, "loss": 0.4063, "step": 4637 }, { "epoch": 0.23, "learning_rate": 1.794344984389247e-05, "loss": 0.4478, "step": 4638 }, { "epoch": 0.23, "learning_rate": 1.794246733830253e-05, "loss": 0.3478, "step": 4639 }, { "epoch": 0.23, "learning_rate": 1.794148462498776e-05, "loss": 0.4379, "step": 4640 }, { "epoch": 0.23, "learning_rate": 1.794050170397386e-05, "loss": 0.4855, "step": 4641 }, { "epoch": 0.23, "learning_rate": 1.7939518575286542e-05, "loss": 0.3494, "step": 4642 }, { "epoch": 0.23, "learning_rate": 1.793853523895151e-05, "loss": 0.3923, "step": 4643 }, { "epoch": 0.23, "learning_rate": 1.793755169499449e-05, "loss": 0.5834, "step": 4644 }, { "epoch": 0.23, "learning_rate": 1.7936567943441197e-05, "loss": 0.4225, "step": 4645 }, { "epoch": 0.23, "learning_rate": 1.793558398431737e-05, "loss": 0.3519, "step": 4646 }, { "epoch": 0.23, "learning_rate": 1.7934599817648734e-05, "loss": 0.5644, "step": 4647 }, { "epoch": 0.23, "learning_rate": 1.7933615443461035e-05, "loss": 0.2397, "step": 4648 }, { "epoch": 0.23, "learning_rate": 1.7932630861780015e-05, "loss": 0.2661, "step": 4649 }, { "epoch": 0.23, "learning_rate": 1.793164607263142e-05, "loss": 0.2462, "step": 4650 }, { "epoch": 0.23, "learning_rate": 1.7930661076041015e-05, "loss": 0.3575, "step": 4651 }, { "epoch": 0.23, "learning_rate": 1.7929675872034558e-05, "loss": 0.4454, "step": 4652 }, { "epoch": 0.23, "learning_rate": 1.792869046063781e-05, "loss": 0.3391, "step": 4653 }, { "epoch": 0.23, "learning_rate": 1.7927704841876548e-05, "loss": 0.4367, "step": 4654 }, { "epoch": 0.23, "learning_rate": 1.7926719015776552e-05, "loss": 0.5992, "step": 4655 }, { "epoch": 0.23, "learning_rate": 1.7925732982363606e-05, "loss": 0.3382, "step": 4656 }, { "epoch": 0.23, "learning_rate": 1.792474674166349e-05, "loss": 0.3301, "step": 4657 }, { "epoch": 0.23, "learning_rate": 1.7923760293702003e-05, "loss": 0.3466, "step": 4658 }, { "epoch": 0.23, "learning_rate": 1.7922773638504942e-05, "loss": 0.8226, "step": 4659 }, { "epoch": 0.23, "learning_rate": 1.7921786776098112e-05, "loss": 0.5023, "step": 4660 }, { "epoch": 0.23, "learning_rate": 1.792079970650733e-05, "loss": 0.297, "step": 4661 }, { "epoch": 0.23, "learning_rate": 1.7919812429758402e-05, "loss": 0.294, "step": 4662 }, { "epoch": 0.23, "learning_rate": 1.7918824945877156e-05, "loss": 0.3572, "step": 4663 }, { "epoch": 0.23, "learning_rate": 1.7917837254889416e-05, "loss": 0.3509, "step": 4664 }, { "epoch": 0.23, "learning_rate": 1.791684935682101e-05, "loss": 0.4364, "step": 4665 }, { "epoch": 0.23, "learning_rate": 1.791586125169778e-05, "loss": 0.257, "step": 4666 }, { "epoch": 0.23, "learning_rate": 1.7914872939545564e-05, "loss": 0.5246, "step": 4667 }, { "epoch": 0.23, "learning_rate": 1.7913884420390218e-05, "loss": 0.1443, "step": 4668 }, { "epoch": 0.23, "learning_rate": 1.7912895694257588e-05, "loss": 0.4727, "step": 4669 }, { "epoch": 0.23, "learning_rate": 1.791190676117353e-05, "loss": 0.509, "step": 4670 }, { "epoch": 0.23, "learning_rate": 1.7910917621163922e-05, "loss": 0.5719, "step": 4671 }, { "epoch": 0.23, "learning_rate": 1.790992827425462e-05, "loss": 0.3964, "step": 4672 }, { "epoch": 0.23, "learning_rate": 1.7908938720471505e-05, "loss": 0.2808, "step": 4673 }, { "epoch": 0.23, "learning_rate": 1.7907948959840456e-05, "loss": 0.3655, "step": 4674 }, { "epoch": 0.23, "learning_rate": 1.7906958992387364e-05, "loss": 0.411, "step": 4675 }, { "epoch": 0.23, "learning_rate": 1.7905968818138112e-05, "loss": 0.4801, "step": 4676 }, { "epoch": 0.23, "learning_rate": 1.7904978437118598e-05, "loss": 0.268, "step": 4677 }, { "epoch": 0.23, "learning_rate": 1.790398784935473e-05, "loss": 0.3312, "step": 4678 }, { "epoch": 0.23, "learning_rate": 1.790299705487241e-05, "loss": 0.4738, "step": 4679 }, { "epoch": 0.23, "learning_rate": 1.7902006053697558e-05, "loss": 0.4153, "step": 4680 }, { "epoch": 0.23, "learning_rate": 1.790101484585608e-05, "loss": 0.3366, "step": 4681 }, { "epoch": 0.23, "learning_rate": 1.7900023431373915e-05, "loss": 0.3064, "step": 4682 }, { "epoch": 0.23, "learning_rate": 1.789903181027698e-05, "loss": 0.3839, "step": 4683 }, { "epoch": 0.23, "learning_rate": 1.7898039982591213e-05, "loss": 0.1951, "step": 4684 }, { "epoch": 0.23, "learning_rate": 1.7897047948342557e-05, "loss": 0.6742, "step": 4685 }, { "epoch": 0.23, "learning_rate": 1.7896055707556955e-05, "loss": 0.5432, "step": 4686 }, { "epoch": 0.23, "learning_rate": 1.7895063260260356e-05, "loss": 0.2931, "step": 4687 }, { "epoch": 0.23, "learning_rate": 1.789407060647872e-05, "loss": 0.2835, "step": 4688 }, { "epoch": 0.23, "learning_rate": 1.7893077746238005e-05, "loss": 0.2648, "step": 4689 }, { "epoch": 0.23, "learning_rate": 1.789208467956418e-05, "loss": 0.5963, "step": 4690 }, { "epoch": 0.23, "learning_rate": 1.7891091406483218e-05, "loss": 0.2527, "step": 4691 }, { "epoch": 0.23, "learning_rate": 1.7890097927021096e-05, "loss": 0.5291, "step": 4692 }, { "epoch": 0.23, "learning_rate": 1.7889104241203798e-05, "loss": 0.4452, "step": 4693 }, { "epoch": 0.23, "learning_rate": 1.7888110349057307e-05, "loss": 0.4693, "step": 4694 }, { "epoch": 0.23, "learning_rate": 1.7887116250607627e-05, "loss": 0.5636, "step": 4695 }, { "epoch": 0.23, "learning_rate": 1.788612194588075e-05, "loss": 0.3438, "step": 4696 }, { "epoch": 0.23, "learning_rate": 1.788512743490268e-05, "loss": 0.3397, "step": 4697 }, { "epoch": 0.23, "learning_rate": 1.788413271769943e-05, "loss": 0.3548, "step": 4698 }, { "epoch": 0.23, "learning_rate": 1.7883137794297018e-05, "loss": 0.8039, "step": 4699 }, { "epoch": 0.23, "learning_rate": 1.788214266472146e-05, "loss": 0.3511, "step": 4700 }, { "epoch": 0.23, "learning_rate": 1.7881147328998783e-05, "loss": 0.344, "step": 4701 }, { "epoch": 0.23, "learning_rate": 1.788015178715502e-05, "loss": 0.3396, "step": 4702 }, { "epoch": 0.23, "learning_rate": 1.7879156039216212e-05, "loss": 0.3933, "step": 4703 }, { "epoch": 0.23, "learning_rate": 1.7878160085208396e-05, "loss": 0.5447, "step": 4704 }, { "epoch": 0.23, "learning_rate": 1.787716392515762e-05, "loss": 0.5977, "step": 4705 }, { "epoch": 0.23, "learning_rate": 1.7876167559089938e-05, "loss": 0.3803, "step": 4706 }, { "epoch": 0.24, "learning_rate": 1.7875170987031413e-05, "loss": 0.3534, "step": 4707 }, { "epoch": 0.24, "learning_rate": 1.7874174209008105e-05, "loss": 0.6025, "step": 4708 }, { "epoch": 0.24, "learning_rate": 1.787317722504608e-05, "loss": 0.2164, "step": 4709 }, { "epoch": 0.24, "learning_rate": 1.787218003517142e-05, "loss": 0.3691, "step": 4710 }, { "epoch": 0.24, "learning_rate": 1.78711826394102e-05, "loss": 0.2175, "step": 4711 }, { "epoch": 0.24, "learning_rate": 1.787018503778851e-05, "loss": 0.2513, "step": 4712 }, { "epoch": 0.24, "learning_rate": 1.7869187230332435e-05, "loss": 0.3315, "step": 4713 }, { "epoch": 0.24, "learning_rate": 1.7868189217068076e-05, "loss": 0.5337, "step": 4714 }, { "epoch": 0.24, "learning_rate": 1.7867190998021534e-05, "loss": 0.5148, "step": 4715 }, { "epoch": 0.24, "learning_rate": 1.786619257321892e-05, "loss": 0.7736, "step": 4716 }, { "epoch": 0.24, "learning_rate": 1.7865193942686335e-05, "loss": 0.2579, "step": 4717 }, { "epoch": 0.24, "learning_rate": 1.7864195106449907e-05, "loss": 0.2851, "step": 4718 }, { "epoch": 0.24, "learning_rate": 1.7863196064535756e-05, "loss": 0.4136, "step": 4719 }, { "epoch": 0.24, "learning_rate": 1.786219681697001e-05, "loss": 0.2365, "step": 4720 }, { "epoch": 0.24, "learning_rate": 1.7861197363778804e-05, "loss": 0.6418, "step": 4721 }, { "epoch": 0.24, "learning_rate": 1.786019770498828e-05, "loss": 0.2329, "step": 4722 }, { "epoch": 0.24, "learning_rate": 1.7859197840624576e-05, "loss": 0.3266, "step": 4723 }, { "epoch": 0.24, "learning_rate": 1.785819777071385e-05, "loss": 0.2734, "step": 4724 }, { "epoch": 0.24, "learning_rate": 1.7857197495282252e-05, "loss": 0.2639, "step": 4725 }, { "epoch": 0.24, "learning_rate": 1.7856197014355943e-05, "loss": 0.3279, "step": 4726 }, { "epoch": 0.24, "learning_rate": 1.785519632796109e-05, "loss": 0.4173, "step": 4727 }, { "epoch": 0.24, "learning_rate": 1.785419543612387e-05, "loss": 0.424, "step": 4728 }, { "epoch": 0.24, "learning_rate": 1.7853194338870456e-05, "loss": 0.4327, "step": 4729 }, { "epoch": 0.24, "learning_rate": 1.7852193036227026e-05, "loss": 0.5386, "step": 4730 }, { "epoch": 0.24, "learning_rate": 1.7851191528219773e-05, "loss": 0.5754, "step": 4731 }, { "epoch": 0.24, "learning_rate": 1.7850189814874888e-05, "loss": 0.4983, "step": 4732 }, { "epoch": 0.24, "learning_rate": 1.784918789621857e-05, "loss": 0.4592, "step": 4733 }, { "epoch": 0.24, "learning_rate": 1.7848185772277026e-05, "loss": 0.2868, "step": 4734 }, { "epoch": 0.24, "learning_rate": 1.784718344307646e-05, "loss": 0.5087, "step": 4735 }, { "epoch": 0.24, "learning_rate": 1.7846180908643092e-05, "loss": 0.4428, "step": 4736 }, { "epoch": 0.24, "learning_rate": 1.7845178169003134e-05, "loss": 0.6511, "step": 4737 }, { "epoch": 0.24, "learning_rate": 1.784417522418282e-05, "loss": 0.3174, "step": 4738 }, { "epoch": 0.24, "learning_rate": 1.7843172074208376e-05, "loss": 0.2635, "step": 4739 }, { "epoch": 0.24, "learning_rate": 1.784216871910604e-05, "loss": 0.4529, "step": 4740 }, { "epoch": 0.24, "learning_rate": 1.7841165158902048e-05, "loss": 0.4213, "step": 4741 }, { "epoch": 0.24, "learning_rate": 1.7840161393622657e-05, "loss": 0.4383, "step": 4742 }, { "epoch": 0.24, "learning_rate": 1.7839157423294113e-05, "loss": 0.8297, "step": 4743 }, { "epoch": 0.24, "learning_rate": 1.7838153247942674e-05, "loss": 0.4124, "step": 4744 }, { "epoch": 0.24, "learning_rate": 1.78371488675946e-05, "loss": 0.5362, "step": 4745 }, { "epoch": 0.24, "learning_rate": 1.7836144282276163e-05, "loss": 0.4888, "step": 4746 }, { "epoch": 0.24, "learning_rate": 1.7835139492013634e-05, "loss": 0.5025, "step": 4747 }, { "epoch": 0.24, "learning_rate": 1.78341344968333e-05, "loss": 0.4817, "step": 4748 }, { "epoch": 0.24, "learning_rate": 1.7833129296761435e-05, "loss": 0.3981, "step": 4749 }, { "epoch": 0.24, "learning_rate": 1.7832123891824332e-05, "loss": 0.4724, "step": 4750 }, { "epoch": 0.24, "learning_rate": 1.783111828204829e-05, "loss": 0.5403, "step": 4751 }, { "epoch": 0.24, "learning_rate": 1.7830112467459605e-05, "loss": 0.4228, "step": 4752 }, { "epoch": 0.24, "learning_rate": 1.782910644808458e-05, "loss": 0.4166, "step": 4753 }, { "epoch": 0.24, "learning_rate": 1.7828100223949538e-05, "loss": 0.483, "step": 4754 }, { "epoch": 0.24, "learning_rate": 1.782709379508078e-05, "loss": 0.5025, "step": 4755 }, { "epoch": 0.24, "learning_rate": 1.7826087161504636e-05, "loss": 0.3224, "step": 4756 }, { "epoch": 0.24, "learning_rate": 1.782508032324744e-05, "loss": 0.3632, "step": 4757 }, { "epoch": 0.24, "learning_rate": 1.782407328033551e-05, "loss": 0.395, "step": 4758 }, { "epoch": 0.24, "learning_rate": 1.7823066032795194e-05, "loss": 0.7165, "step": 4759 }, { "epoch": 0.24, "learning_rate": 1.7822058580652834e-05, "loss": 0.3759, "step": 4760 }, { "epoch": 0.24, "learning_rate": 1.782105092393477e-05, "loss": 0.2939, "step": 4761 }, { "epoch": 0.24, "learning_rate": 1.7820043062667366e-05, "loss": 0.3084, "step": 4762 }, { "epoch": 0.24, "learning_rate": 1.7819034996876982e-05, "loss": 0.3525, "step": 4763 }, { "epoch": 0.24, "learning_rate": 1.7818026726589977e-05, "loss": 0.5394, "step": 4764 }, { "epoch": 0.24, "learning_rate": 1.781701825183272e-05, "loss": 0.2141, "step": 4765 }, { "epoch": 0.24, "learning_rate": 1.781600957263159e-05, "loss": 0.5021, "step": 4766 }, { "epoch": 0.24, "learning_rate": 1.7815000689012972e-05, "loss": 0.5035, "step": 4767 }, { "epoch": 0.24, "learning_rate": 1.7813991601003244e-05, "loss": 0.3512, "step": 4768 }, { "epoch": 0.24, "learning_rate": 1.7812982308628795e-05, "loss": 0.3111, "step": 4769 }, { "epoch": 0.24, "learning_rate": 1.781197281191603e-05, "loss": 0.3564, "step": 4770 }, { "epoch": 0.24, "learning_rate": 1.781096311089135e-05, "loss": 0.4333, "step": 4771 }, { "epoch": 0.24, "learning_rate": 1.780995320558116e-05, "loss": 0.5136, "step": 4772 }, { "epoch": 0.24, "learning_rate": 1.780894309601187e-05, "loss": 0.2652, "step": 4773 }, { "epoch": 0.24, "learning_rate": 1.7807932782209906e-05, "loss": 0.1889, "step": 4774 }, { "epoch": 0.24, "learning_rate": 1.7806922264201683e-05, "loss": 0.4212, "step": 4775 }, { "epoch": 0.24, "learning_rate": 1.780591154201363e-05, "loss": 0.342, "step": 4776 }, { "epoch": 0.24, "learning_rate": 1.7804900615672192e-05, "loss": 0.3979, "step": 4777 }, { "epoch": 0.24, "learning_rate": 1.7803889485203797e-05, "loss": 0.1545, "step": 4778 }, { "epoch": 0.24, "learning_rate": 1.7802878150634894e-05, "loss": 0.3652, "step": 4779 }, { "epoch": 0.24, "learning_rate": 1.7801866611991933e-05, "loss": 0.5277, "step": 4780 }, { "epoch": 0.24, "learning_rate": 1.780085486930137e-05, "loss": 0.4023, "step": 4781 }, { "epoch": 0.24, "learning_rate": 1.779984292258967e-05, "loss": 0.4149, "step": 4782 }, { "epoch": 0.24, "learning_rate": 1.7798830771883285e-05, "loss": 0.2993, "step": 4783 }, { "epoch": 0.24, "learning_rate": 1.77978184172087e-05, "loss": 0.4967, "step": 4784 }, { "epoch": 0.24, "learning_rate": 1.779680585859239e-05, "loss": 0.251, "step": 4785 }, { "epoch": 0.24, "learning_rate": 1.779579309606083e-05, "loss": 0.4909, "step": 4786 }, { "epoch": 0.24, "learning_rate": 1.7794780129640513e-05, "loss": 0.2707, "step": 4787 }, { "epoch": 0.24, "learning_rate": 1.7793766959357935e-05, "loss": 0.22, "step": 4788 }, { "epoch": 0.24, "learning_rate": 1.7792753585239588e-05, "loss": 0.4747, "step": 4789 }, { "epoch": 0.24, "learning_rate": 1.779174000731198e-05, "loss": 0.4604, "step": 4790 }, { "epoch": 0.24, "learning_rate": 1.7790726225601615e-05, "loss": 0.3271, "step": 4791 }, { "epoch": 0.24, "learning_rate": 1.7789712240135006e-05, "loss": 0.4838, "step": 4792 }, { "epoch": 0.24, "learning_rate": 1.778869805093868e-05, "loss": 0.3885, "step": 4793 }, { "epoch": 0.24, "learning_rate": 1.7787683658039157e-05, "loss": 0.4475, "step": 4794 }, { "epoch": 0.24, "learning_rate": 1.7786669061462973e-05, "loss": 0.4416, "step": 4795 }, { "epoch": 0.24, "learning_rate": 1.7785654261236652e-05, "loss": 0.3101, "step": 4796 }, { "epoch": 0.24, "learning_rate": 1.7784639257386742e-05, "loss": 0.2571, "step": 4797 }, { "epoch": 0.24, "learning_rate": 1.7783624049939788e-05, "loss": 0.461, "step": 4798 }, { "epoch": 0.24, "learning_rate": 1.7782608638922343e-05, "loss": 0.4203, "step": 4799 }, { "epoch": 0.24, "learning_rate": 1.7781593024360962e-05, "loss": 0.3657, "step": 4800 }, { "epoch": 0.24, "learning_rate": 1.7780577206282207e-05, "loss": 0.5085, "step": 4801 }, { "epoch": 0.24, "learning_rate": 1.777956118471265e-05, "loss": 0.4556, "step": 4802 }, { "epoch": 0.24, "learning_rate": 1.7778544959678856e-05, "loss": 0.4527, "step": 4803 }, { "epoch": 0.24, "learning_rate": 1.7777528531207406e-05, "loss": 0.4295, "step": 4804 }, { "epoch": 0.24, "learning_rate": 1.7776511899324887e-05, "loss": 0.5301, "step": 4805 }, { "epoch": 0.24, "learning_rate": 1.7775495064057882e-05, "loss": 0.5424, "step": 4806 }, { "epoch": 0.24, "learning_rate": 1.7774478025432988e-05, "loss": 0.3517, "step": 4807 }, { "epoch": 0.24, "learning_rate": 1.7773460783476805e-05, "loss": 0.4488, "step": 4808 }, { "epoch": 0.24, "learning_rate": 1.7772443338215936e-05, "loss": 0.4444, "step": 4809 }, { "epoch": 0.24, "learning_rate": 1.777142568967699e-05, "loss": 0.6421, "step": 4810 }, { "epoch": 0.24, "learning_rate": 1.7770407837886585e-05, "loss": 0.3759, "step": 4811 }, { "epoch": 0.24, "learning_rate": 1.776938978287134e-05, "loss": 0.4909, "step": 4812 }, { "epoch": 0.24, "learning_rate": 1.7768371524657884e-05, "loss": 0.561, "step": 4813 }, { "epoch": 0.24, "learning_rate": 1.7767353063272846e-05, "loss": 0.419, "step": 4814 }, { "epoch": 0.24, "learning_rate": 1.776633439874286e-05, "loss": 0.6488, "step": 4815 }, { "epoch": 0.24, "learning_rate": 1.7765315531094566e-05, "loss": 0.6861, "step": 4816 }, { "epoch": 0.24, "learning_rate": 1.776429646035462e-05, "loss": 0.2169, "step": 4817 }, { "epoch": 0.24, "learning_rate": 1.776327718654967e-05, "loss": 0.205, "step": 4818 }, { "epoch": 0.24, "learning_rate": 1.776225770970637e-05, "loss": 0.4703, "step": 4819 }, { "epoch": 0.24, "learning_rate": 1.7761238029851385e-05, "loss": 0.4297, "step": 4820 }, { "epoch": 0.24, "learning_rate": 1.7760218147011386e-05, "loss": 0.1738, "step": 4821 }, { "epoch": 0.24, "learning_rate": 1.775919806121305e-05, "loss": 0.6337, "step": 4822 }, { "epoch": 0.24, "learning_rate": 1.7758177772483046e-05, "loss": 0.5159, "step": 4823 }, { "epoch": 0.24, "learning_rate": 1.7757157280848067e-05, "loss": 0.3627, "step": 4824 }, { "epoch": 0.24, "learning_rate": 1.7756136586334796e-05, "loss": 0.3731, "step": 4825 }, { "epoch": 0.24, "learning_rate": 1.7755115688969938e-05, "loss": 0.5895, "step": 4826 }, { "epoch": 0.24, "learning_rate": 1.775409458878018e-05, "loss": 0.5879, "step": 4827 }, { "epoch": 0.24, "learning_rate": 1.7753073285792237e-05, "loss": 0.3575, "step": 4828 }, { "epoch": 0.24, "learning_rate": 1.775205178003282e-05, "loss": 0.2792, "step": 4829 }, { "epoch": 0.24, "learning_rate": 1.7751030071528634e-05, "loss": 0.4138, "step": 4830 }, { "epoch": 0.24, "learning_rate": 1.7750008160306413e-05, "loss": 0.3769, "step": 4831 }, { "epoch": 0.24, "learning_rate": 1.774898604639288e-05, "loss": 0.3866, "step": 4832 }, { "epoch": 0.24, "learning_rate": 1.7747963729814764e-05, "loss": 0.4988, "step": 4833 }, { "epoch": 0.24, "learning_rate": 1.7746941210598808e-05, "loss": 0.448, "step": 4834 }, { "epoch": 0.24, "learning_rate": 1.7745918488771748e-05, "loss": 0.2295, "step": 4835 }, { "epoch": 0.24, "learning_rate": 1.7744895564360333e-05, "loss": 0.3403, "step": 4836 }, { "epoch": 0.24, "learning_rate": 1.7743872437391325e-05, "loss": 0.6292, "step": 4837 }, { "epoch": 0.24, "learning_rate": 1.7742849107891472e-05, "loss": 0.3826, "step": 4838 }, { "epoch": 0.24, "learning_rate": 1.7741825575887545e-05, "loss": 0.302, "step": 4839 }, { "epoch": 0.24, "learning_rate": 1.7740801841406305e-05, "loss": 0.3236, "step": 4840 }, { "epoch": 0.24, "learning_rate": 1.7739777904474537e-05, "loss": 0.4599, "step": 4841 }, { "epoch": 0.24, "learning_rate": 1.7738753765119012e-05, "loss": 0.3795, "step": 4842 }, { "epoch": 0.24, "learning_rate": 1.7737729423366516e-05, "loss": 0.5365, "step": 4843 }, { "epoch": 0.24, "learning_rate": 1.7736704879243845e-05, "loss": 0.4211, "step": 4844 }, { "epoch": 0.24, "learning_rate": 1.7735680132777794e-05, "loss": 0.2966, "step": 4845 }, { "epoch": 0.24, "learning_rate": 1.7734655183995155e-05, "loss": 0.451, "step": 4846 }, { "epoch": 0.24, "learning_rate": 1.7733630032922745e-05, "loss": 0.3151, "step": 4847 }, { "epoch": 0.24, "learning_rate": 1.7732604679587366e-05, "loss": 0.4132, "step": 4848 }, { "epoch": 0.24, "learning_rate": 1.773157912401584e-05, "loss": 0.3532, "step": 4849 }, { "epoch": 0.24, "learning_rate": 1.7730553366234994e-05, "loss": 0.2698, "step": 4850 }, { "epoch": 0.24, "learning_rate": 1.7729527406271646e-05, "loss": 0.3198, "step": 4851 }, { "epoch": 0.24, "learning_rate": 1.7728501244152633e-05, "loss": 0.26, "step": 4852 }, { "epoch": 0.24, "learning_rate": 1.772747487990479e-05, "loss": 0.3756, "step": 4853 }, { "epoch": 0.24, "learning_rate": 1.7726448313554966e-05, "loss": 0.4192, "step": 4854 }, { "epoch": 0.24, "learning_rate": 1.7725421545130007e-05, "loss": 0.3571, "step": 4855 }, { "epoch": 0.24, "learning_rate": 1.7724394574656763e-05, "loss": 0.4392, "step": 4856 }, { "epoch": 0.24, "learning_rate": 1.77233674021621e-05, "loss": 0.4459, "step": 4857 }, { "epoch": 0.24, "learning_rate": 1.7722340027672876e-05, "loss": 0.3532, "step": 4858 }, { "epoch": 0.24, "learning_rate": 1.772131245121596e-05, "loss": 0.4379, "step": 4859 }, { "epoch": 0.24, "learning_rate": 1.7720284672818233e-05, "loss": 0.2695, "step": 4860 }, { "epoch": 0.24, "learning_rate": 1.7719256692506573e-05, "loss": 0.2199, "step": 4861 }, { "epoch": 0.24, "learning_rate": 1.7718228510307863e-05, "loss": 0.4827, "step": 4862 }, { "epoch": 0.24, "learning_rate": 1.7717200126248997e-05, "loss": 0.2209, "step": 4863 }, { "epoch": 0.24, "learning_rate": 1.7716171540356868e-05, "loss": 0.4128, "step": 4864 }, { "epoch": 0.24, "learning_rate": 1.771514275265838e-05, "loss": 0.5707, "step": 4865 }, { "epoch": 0.24, "learning_rate": 1.7714113763180435e-05, "loss": 0.1768, "step": 4866 }, { "epoch": 0.24, "learning_rate": 1.7713084571949952e-05, "loss": 0.3428, "step": 4867 }, { "epoch": 0.24, "learning_rate": 1.771205517899384e-05, "loss": 0.2878, "step": 4868 }, { "epoch": 0.24, "learning_rate": 1.7711025584339033e-05, "loss": 0.543, "step": 4869 }, { "epoch": 0.24, "learning_rate": 1.7709995788012445e-05, "loss": 0.3101, "step": 4870 }, { "epoch": 0.24, "learning_rate": 1.7708965790041017e-05, "loss": 0.1733, "step": 4871 }, { "epoch": 0.24, "learning_rate": 1.7707935590451688e-05, "loss": 0.5882, "step": 4872 }, { "epoch": 0.24, "learning_rate": 1.7706905189271398e-05, "loss": 0.4204, "step": 4873 }, { "epoch": 0.24, "learning_rate": 1.7705874586527095e-05, "loss": 0.4904, "step": 4874 }, { "epoch": 0.24, "learning_rate": 1.7704843782245736e-05, "loss": 0.4257, "step": 4875 }, { "epoch": 0.24, "learning_rate": 1.770381277645428e-05, "loss": 0.5, "step": 4876 }, { "epoch": 0.24, "learning_rate": 1.7702781569179693e-05, "loss": 0.3723, "step": 4877 }, { "epoch": 0.24, "learning_rate": 1.7701750160448942e-05, "loss": 0.3562, "step": 4878 }, { "epoch": 0.24, "learning_rate": 1.7700718550289e-05, "loss": 0.2804, "step": 4879 }, { "epoch": 0.24, "learning_rate": 1.769968673872685e-05, "loss": 0.2601, "step": 4880 }, { "epoch": 0.24, "learning_rate": 1.7698654725789483e-05, "loss": 0.5026, "step": 4881 }, { "epoch": 0.24, "learning_rate": 1.7697622511503883e-05, "loss": 0.4907, "step": 4882 }, { "epoch": 0.24, "learning_rate": 1.7696590095897047e-05, "loss": 0.4432, "step": 4883 }, { "epoch": 0.24, "learning_rate": 1.7695557478995976e-05, "loss": 0.3196, "step": 4884 }, { "epoch": 0.24, "learning_rate": 1.7694524660827682e-05, "loss": 0.2829, "step": 4885 }, { "epoch": 0.24, "learning_rate": 1.7693491641419173e-05, "loss": 0.329, "step": 4886 }, { "epoch": 0.24, "learning_rate": 1.7692458420797466e-05, "loss": 0.262, "step": 4887 }, { "epoch": 0.24, "learning_rate": 1.7691424998989584e-05, "loss": 0.2212, "step": 4888 }, { "epoch": 0.24, "learning_rate": 1.7690391376022558e-05, "loss": 0.4056, "step": 4889 }, { "epoch": 0.24, "learning_rate": 1.7689357551923415e-05, "loss": 0.6488, "step": 4890 }, { "epoch": 0.24, "learning_rate": 1.76883235267192e-05, "loss": 0.4679, "step": 4891 }, { "epoch": 0.24, "learning_rate": 1.7687289300436948e-05, "loss": 0.2379, "step": 4892 }, { "epoch": 0.24, "learning_rate": 1.7686254873103715e-05, "loss": 0.3063, "step": 4893 }, { "epoch": 0.24, "learning_rate": 1.7685220244746554e-05, "loss": 0.3366, "step": 4894 }, { "epoch": 0.24, "learning_rate": 1.7684185415392525e-05, "loss": 0.2372, "step": 4895 }, { "epoch": 0.24, "learning_rate": 1.768315038506869e-05, "loss": 0.3958, "step": 4896 }, { "epoch": 0.24, "learning_rate": 1.7682115153802114e-05, "loss": 0.2753, "step": 4897 }, { "epoch": 0.24, "learning_rate": 1.7681079721619882e-05, "loss": 0.5888, "step": 4898 }, { "epoch": 0.24, "learning_rate": 1.7680044088549072e-05, "loss": 0.6188, "step": 4899 }, { "epoch": 0.24, "learning_rate": 1.7679008254616765e-05, "loss": 0.4364, "step": 4900 }, { "epoch": 0.24, "learning_rate": 1.7677972219850057e-05, "loss": 0.4127, "step": 4901 }, { "epoch": 0.24, "learning_rate": 1.767693598427604e-05, "loss": 0.3329, "step": 4902 }, { "epoch": 0.24, "learning_rate": 1.767589954792182e-05, "loss": 0.4833, "step": 4903 }, { "epoch": 0.24, "learning_rate": 1.76748629108145e-05, "loss": 0.2889, "step": 4904 }, { "epoch": 0.24, "learning_rate": 1.767382607298119e-05, "loss": 0.4288, "step": 4905 }, { "epoch": 0.24, "learning_rate": 1.7672789034449014e-05, "loss": 0.4973, "step": 4906 }, { "epoch": 0.25, "learning_rate": 1.767175179524509e-05, "loss": 0.4493, "step": 4907 }, { "epoch": 0.25, "learning_rate": 1.7670714355396544e-05, "loss": 0.3022, "step": 4908 }, { "epoch": 0.25, "learning_rate": 1.766967671493051e-05, "loss": 0.5031, "step": 4909 }, { "epoch": 0.25, "learning_rate": 1.766863887387413e-05, "loss": 0.4589, "step": 4910 }, { "epoch": 0.25, "learning_rate": 1.7667600832254543e-05, "loss": 0.3183, "step": 4911 }, { "epoch": 0.25, "learning_rate": 1.76665625900989e-05, "loss": 0.256, "step": 4912 }, { "epoch": 0.25, "learning_rate": 1.7665524147434353e-05, "loss": 0.3796, "step": 4913 }, { "epoch": 0.25, "learning_rate": 1.766448550428806e-05, "loss": 0.448, "step": 4914 }, { "epoch": 0.25, "learning_rate": 1.766344666068719e-05, "loss": 0.3216, "step": 4915 }, { "epoch": 0.25, "learning_rate": 1.7662407616658906e-05, "loss": 0.4137, "step": 4916 }, { "epoch": 0.25, "learning_rate": 1.7661368372230393e-05, "loss": 0.6522, "step": 4917 }, { "epoch": 0.25, "learning_rate": 1.766032892742882e-05, "loss": 0.4333, "step": 4918 }, { "epoch": 0.25, "learning_rate": 1.765928928228138e-05, "loss": 0.3408, "step": 4919 }, { "epoch": 0.25, "learning_rate": 1.7658249436815257e-05, "loss": 0.5625, "step": 4920 }, { "epoch": 0.25, "learning_rate": 1.765720939105765e-05, "loss": 0.6574, "step": 4921 }, { "epoch": 0.25, "learning_rate": 1.7656169145035765e-05, "loss": 0.4308, "step": 4922 }, { "epoch": 0.25, "learning_rate": 1.7655128698776804e-05, "loss": 0.3264, "step": 4923 }, { "epoch": 0.25, "learning_rate": 1.7654088052307976e-05, "loss": 0.4949, "step": 4924 }, { "epoch": 0.25, "learning_rate": 1.76530472056565e-05, "loss": 0.3619, "step": 4925 }, { "epoch": 0.25, "learning_rate": 1.76520061588496e-05, "loss": 0.317, "step": 4926 }, { "epoch": 0.25, "learning_rate": 1.7650964911914498e-05, "loss": 0.4379, "step": 4927 }, { "epoch": 0.25, "learning_rate": 1.7649923464878432e-05, "loss": 0.4296, "step": 4928 }, { "epoch": 0.25, "learning_rate": 1.7648881817768638e-05, "loss": 0.1983, "step": 4929 }, { "epoch": 0.25, "learning_rate": 1.7647839970612355e-05, "loss": 0.2926, "step": 4930 }, { "epoch": 0.25, "learning_rate": 1.7646797923436837e-05, "loss": 0.2787, "step": 4931 }, { "epoch": 0.25, "learning_rate": 1.7645755676269333e-05, "loss": 0.4654, "step": 4932 }, { "epoch": 0.25, "learning_rate": 1.7644713229137106e-05, "loss": 0.4379, "step": 4933 }, { "epoch": 0.25, "learning_rate": 1.7643670582067417e-05, "loss": 0.3768, "step": 4934 }, { "epoch": 0.25, "learning_rate": 1.7642627735087535e-05, "loss": 0.2144, "step": 4935 }, { "epoch": 0.25, "learning_rate": 1.7641584688224732e-05, "loss": 0.522, "step": 4936 }, { "epoch": 0.25, "learning_rate": 1.7640541441506295e-05, "loss": 0.3557, "step": 4937 }, { "epoch": 0.25, "learning_rate": 1.7639497994959497e-05, "loss": 0.3335, "step": 4938 }, { "epoch": 0.25, "learning_rate": 1.763845434861164e-05, "loss": 0.3506, "step": 4939 }, { "epoch": 0.25, "learning_rate": 1.7637410502490013e-05, "loss": 0.5294, "step": 4940 }, { "epoch": 0.25, "learning_rate": 1.763636645662192e-05, "loss": 0.5606, "step": 4941 }, { "epoch": 0.25, "learning_rate": 1.763532221103466e-05, "loss": 0.3815, "step": 4942 }, { "epoch": 0.25, "learning_rate": 1.763427776575555e-05, "loss": 0.271, "step": 4943 }, { "epoch": 0.25, "learning_rate": 1.7633233120811902e-05, "loss": 0.5313, "step": 4944 }, { "epoch": 0.25, "learning_rate": 1.763218827623104e-05, "loss": 0.5802, "step": 4945 }, { "epoch": 0.25, "learning_rate": 1.7631143232040294e-05, "loss": 0.4063, "step": 4946 }, { "epoch": 0.25, "learning_rate": 1.763009798826699e-05, "loss": 0.2832, "step": 4947 }, { "epoch": 0.25, "learning_rate": 1.7629052544938466e-05, "loss": 0.3244, "step": 4948 }, { "epoch": 0.25, "learning_rate": 1.7628006902082063e-05, "loss": 0.2816, "step": 4949 }, { "epoch": 0.25, "learning_rate": 1.7626961059725132e-05, "loss": 0.2174, "step": 4950 }, { "epoch": 0.25, "learning_rate": 1.7625915017895026e-05, "loss": 0.1821, "step": 4951 }, { "epoch": 0.25, "learning_rate": 1.76248687766191e-05, "loss": 0.2, "step": 4952 }, { "epoch": 0.25, "learning_rate": 1.7623822335924714e-05, "loss": 0.4112, "step": 4953 }, { "epoch": 0.25, "learning_rate": 1.762277569583924e-05, "loss": 0.6623, "step": 4954 }, { "epoch": 0.25, "learning_rate": 1.762172885639006e-05, "loss": 0.3115, "step": 4955 }, { "epoch": 0.25, "learning_rate": 1.762068181760454e-05, "loss": 0.455, "step": 4956 }, { "epoch": 0.25, "learning_rate": 1.761963457951007e-05, "loss": 0.427, "step": 4957 }, { "epoch": 0.25, "learning_rate": 1.761858714213403e-05, "loss": 0.4299, "step": 4958 }, { "epoch": 0.25, "learning_rate": 1.761753950550383e-05, "loss": 0.4053, "step": 4959 }, { "epoch": 0.25, "learning_rate": 1.7616491669646855e-05, "loss": 0.1998, "step": 4960 }, { "epoch": 0.25, "learning_rate": 1.761544363459052e-05, "loss": 0.4003, "step": 4961 }, { "epoch": 0.25, "learning_rate": 1.7614395400362233e-05, "loss": 0.2744, "step": 4962 }, { "epoch": 0.25, "learning_rate": 1.7613346966989406e-05, "loss": 0.4843, "step": 4963 }, { "epoch": 0.25, "learning_rate": 1.761229833449946e-05, "loss": 0.478, "step": 4964 }, { "epoch": 0.25, "learning_rate": 1.7611249502919818e-05, "loss": 0.4605, "step": 4965 }, { "epoch": 0.25, "learning_rate": 1.7610200472277917e-05, "loss": 0.3591, "step": 4966 }, { "epoch": 0.25, "learning_rate": 1.7609151242601193e-05, "loss": 0.662, "step": 4967 }, { "epoch": 0.25, "learning_rate": 1.760810181391708e-05, "loss": 0.5969, "step": 4968 }, { "epoch": 0.25, "learning_rate": 1.7607052186253036e-05, "loss": 0.2786, "step": 4969 }, { "epoch": 0.25, "learning_rate": 1.7606002359636497e-05, "loss": 0.4577, "step": 4970 }, { "epoch": 0.25, "learning_rate": 1.7604952334094935e-05, "loss": 0.297, "step": 4971 }, { "epoch": 0.25, "learning_rate": 1.76039021096558e-05, "loss": 0.2394, "step": 4972 }, { "epoch": 0.25, "learning_rate": 1.760285168634657e-05, "loss": 0.4347, "step": 4973 }, { "epoch": 0.25, "learning_rate": 1.7601801064194707e-05, "loss": 0.3053, "step": 4974 }, { "epoch": 0.25, "learning_rate": 1.76007502432277e-05, "loss": 0.4687, "step": 4975 }, { "epoch": 0.25, "learning_rate": 1.759969922347302e-05, "loss": 0.3434, "step": 4976 }, { "epoch": 0.25, "learning_rate": 1.7598648004958157e-05, "loss": 0.572, "step": 4977 }, { "epoch": 0.25, "learning_rate": 1.7597596587710614e-05, "loss": 0.1892, "step": 4978 }, { "epoch": 0.25, "learning_rate": 1.7596544971757885e-05, "loss": 0.3539, "step": 4979 }, { "epoch": 0.25, "learning_rate": 1.7595493157127467e-05, "loss": 0.5012, "step": 4980 }, { "epoch": 0.25, "learning_rate": 1.7594441143846877e-05, "loss": 0.5762, "step": 4981 }, { "epoch": 0.25, "learning_rate": 1.7593388931943623e-05, "loss": 0.1184, "step": 4982 }, { "epoch": 0.25, "learning_rate": 1.7592336521445232e-05, "loss": 0.4032, "step": 4983 }, { "epoch": 0.25, "learning_rate": 1.7591283912379217e-05, "loss": 0.2724, "step": 4984 }, { "epoch": 0.25, "learning_rate": 1.7590231104773118e-05, "loss": 0.3123, "step": 4985 }, { "epoch": 0.25, "learning_rate": 1.7589178098654467e-05, "loss": 0.4284, "step": 4986 }, { "epoch": 0.25, "learning_rate": 1.75881248940508e-05, "loss": 0.5373, "step": 4987 }, { "epoch": 0.25, "learning_rate": 1.7587071490989665e-05, "loss": 0.4267, "step": 4988 }, { "epoch": 0.25, "learning_rate": 1.7586017889498616e-05, "loss": 0.4765, "step": 4989 }, { "epoch": 0.25, "learning_rate": 1.7584964089605203e-05, "loss": 0.4952, "step": 4990 }, { "epoch": 0.25, "learning_rate": 1.7583910091336986e-05, "loss": 0.2873, "step": 4991 }, { "epoch": 0.25, "learning_rate": 1.7582855894721537e-05, "loss": 0.4239, "step": 4992 }, { "epoch": 0.25, "learning_rate": 1.7581801499786425e-05, "loss": 0.4028, "step": 4993 }, { "epoch": 0.25, "learning_rate": 1.758074690655922e-05, "loss": 0.2308, "step": 4994 }, { "epoch": 0.25, "learning_rate": 1.7579692115067513e-05, "loss": 0.4554, "step": 4995 }, { "epoch": 0.25, "learning_rate": 1.7578637125338883e-05, "loss": 0.3594, "step": 4996 }, { "epoch": 0.25, "learning_rate": 1.7577581937400928e-05, "loss": 0.5255, "step": 4997 }, { "epoch": 0.25, "learning_rate": 1.757652655128124e-05, "loss": 0.2487, "step": 4998 }, { "epoch": 0.25, "learning_rate": 1.757547096700743e-05, "loss": 0.3602, "step": 4999 }, { "epoch": 0.25, "learning_rate": 1.757441518460709e-05, "loss": 0.3849, "step": 5000 }, { "epoch": 0.25, "learning_rate": 1.7573359204107843e-05, "loss": 0.4218, "step": 5001 }, { "epoch": 0.25, "learning_rate": 1.757230302553731e-05, "loss": 0.2466, "step": 5002 }, { "epoch": 0.25, "learning_rate": 1.757124664892311e-05, "loss": 0.2993, "step": 5003 }, { "epoch": 0.25, "learning_rate": 1.757019007429287e-05, "loss": 0.3396, "step": 5004 }, { "epoch": 0.25, "learning_rate": 1.7569133301674218e-05, "loss": 0.2681, "step": 5005 }, { "epoch": 0.25, "learning_rate": 1.75680763310948e-05, "loss": 0.5618, "step": 5006 }, { "epoch": 0.25, "learning_rate": 1.756701916258226e-05, "loss": 0.4252, "step": 5007 }, { "epoch": 0.25, "learning_rate": 1.7565961796164246e-05, "loss": 0.3705, "step": 5008 }, { "epoch": 0.25, "learning_rate": 1.7564904231868408e-05, "loss": 0.2014, "step": 5009 }, { "epoch": 0.25, "learning_rate": 1.756384646972241e-05, "loss": 0.1004, "step": 5010 }, { "epoch": 0.25, "learning_rate": 1.7562788509753912e-05, "loss": 0.616, "step": 5011 }, { "epoch": 0.25, "learning_rate": 1.756173035199059e-05, "loss": 0.6117, "step": 5012 }, { "epoch": 0.25, "learning_rate": 1.7560671996460113e-05, "loss": 0.1821, "step": 5013 }, { "epoch": 0.25, "learning_rate": 1.755961344319016e-05, "loss": 0.3711, "step": 5014 }, { "epoch": 0.25, "learning_rate": 1.7558554692208422e-05, "loss": 0.5089, "step": 5015 }, { "epoch": 0.25, "learning_rate": 1.7557495743542586e-05, "loss": 0.3627, "step": 5016 }, { "epoch": 0.25, "learning_rate": 1.7556436597220344e-05, "loss": 0.2764, "step": 5017 }, { "epoch": 0.25, "learning_rate": 1.7555377253269402e-05, "loss": 0.4262, "step": 5018 }, { "epoch": 0.25, "learning_rate": 1.7554317711717464e-05, "loss": 0.4798, "step": 5019 }, { "epoch": 0.25, "learning_rate": 1.7553257972592242e-05, "loss": 0.2738, "step": 5020 }, { "epoch": 0.25, "learning_rate": 1.755219803592145e-05, "loss": 0.3412, "step": 5021 }, { "epoch": 0.25, "learning_rate": 1.755113790173281e-05, "loss": 0.3757, "step": 5022 }, { "epoch": 0.25, "learning_rate": 1.7550077570054045e-05, "loss": 0.4764, "step": 5023 }, { "epoch": 0.25, "learning_rate": 1.754901704091289e-05, "loss": 0.4874, "step": 5024 }, { "epoch": 0.25, "learning_rate": 1.7547956314337086e-05, "loss": 0.4863, "step": 5025 }, { "epoch": 0.25, "learning_rate": 1.7546895390354367e-05, "loss": 0.298, "step": 5026 }, { "epoch": 0.25, "learning_rate": 1.7545834268992487e-05, "loss": 0.4316, "step": 5027 }, { "epoch": 0.25, "learning_rate": 1.7544772950279198e-05, "loss": 0.399, "step": 5028 }, { "epoch": 0.25, "learning_rate": 1.7543711434242245e-05, "loss": 0.4599, "step": 5029 }, { "epoch": 0.25, "learning_rate": 1.7542649720909407e-05, "loss": 0.3049, "step": 5030 }, { "epoch": 0.25, "learning_rate": 1.7541587810308444e-05, "loss": 0.6286, "step": 5031 }, { "epoch": 0.25, "learning_rate": 1.7540525702467125e-05, "loss": 0.3837, "step": 5032 }, { "epoch": 0.25, "learning_rate": 1.7539463397413238e-05, "loss": 0.3696, "step": 5033 }, { "epoch": 0.25, "learning_rate": 1.7538400895174557e-05, "loss": 0.4513, "step": 5034 }, { "epoch": 0.25, "learning_rate": 1.7537338195778875e-05, "loss": 0.2678, "step": 5035 }, { "epoch": 0.25, "learning_rate": 1.7536275299253987e-05, "loss": 0.4488, "step": 5036 }, { "epoch": 0.25, "learning_rate": 1.7535212205627684e-05, "loss": 0.4915, "step": 5037 }, { "epoch": 0.25, "learning_rate": 1.753414891492778e-05, "loss": 0.436, "step": 5038 }, { "epoch": 0.25, "learning_rate": 1.753308542718208e-05, "loss": 0.3046, "step": 5039 }, { "epoch": 0.25, "learning_rate": 1.7532021742418393e-05, "loss": 0.2252, "step": 5040 }, { "epoch": 0.25, "learning_rate": 1.7530957860664544e-05, "loss": 0.4273, "step": 5041 }, { "epoch": 0.25, "learning_rate": 1.7529893781948356e-05, "loss": 0.3881, "step": 5042 }, { "epoch": 0.25, "learning_rate": 1.7528829506297657e-05, "loss": 0.465, "step": 5043 }, { "epoch": 0.25, "learning_rate": 1.7527765033740286e-05, "loss": 0.3571, "step": 5044 }, { "epoch": 0.25, "learning_rate": 1.752670036430408e-05, "loss": 0.3974, "step": 5045 }, { "epoch": 0.25, "learning_rate": 1.7525635498016883e-05, "loss": 0.2662, "step": 5046 }, { "epoch": 0.25, "learning_rate": 1.752457043490655e-05, "loss": 0.3667, "step": 5047 }, { "epoch": 0.25, "learning_rate": 1.7523505175000926e-05, "loss": 0.2556, "step": 5048 }, { "epoch": 0.25, "learning_rate": 1.752243971832788e-05, "loss": 0.7567, "step": 5049 }, { "epoch": 0.25, "learning_rate": 1.752137406491528e-05, "loss": 0.3626, "step": 5050 }, { "epoch": 0.25, "learning_rate": 1.7520308214790992e-05, "loss": 0.4315, "step": 5051 }, { "epoch": 0.25, "learning_rate": 1.7519242167982888e-05, "loss": 0.0817, "step": 5052 }, { "epoch": 0.25, "learning_rate": 1.7518175924518857e-05, "loss": 0.4626, "step": 5053 }, { "epoch": 0.25, "learning_rate": 1.751710948442678e-05, "loss": 0.3405, "step": 5054 }, { "epoch": 0.25, "learning_rate": 1.7516042847734553e-05, "loss": 0.2623, "step": 5055 }, { "epoch": 0.25, "learning_rate": 1.7514976014470066e-05, "loss": 0.5503, "step": 5056 }, { "epoch": 0.25, "learning_rate": 1.7513908984661222e-05, "loss": 0.3937, "step": 5057 }, { "epoch": 0.25, "learning_rate": 1.7512841758335938e-05, "loss": 0.5307, "step": 5058 }, { "epoch": 0.25, "learning_rate": 1.751177433552211e-05, "loss": 0.458, "step": 5059 }, { "epoch": 0.25, "learning_rate": 1.7510706716247667e-05, "loss": 0.2067, "step": 5060 }, { "epoch": 0.25, "learning_rate": 1.7509638900540527e-05, "loss": 0.17, "step": 5061 }, { "epoch": 0.25, "learning_rate": 1.7508570888428615e-05, "loss": 0.5349, "step": 5062 }, { "epoch": 0.25, "learning_rate": 1.750750267993987e-05, "loss": 0.2316, "step": 5063 }, { "epoch": 0.25, "learning_rate": 1.7506434275102222e-05, "loss": 0.3959, "step": 5064 }, { "epoch": 0.25, "learning_rate": 1.750536567394362e-05, "loss": 0.456, "step": 5065 }, { "epoch": 0.25, "learning_rate": 1.7504296876492008e-05, "loss": 0.164, "step": 5066 }, { "epoch": 0.25, "learning_rate": 1.7503227882775338e-05, "loss": 0.4498, "step": 5067 }, { "epoch": 0.25, "learning_rate": 1.750215869282157e-05, "loss": 0.2372, "step": 5068 }, { "epoch": 0.25, "learning_rate": 1.750108930665867e-05, "loss": 0.2719, "step": 5069 }, { "epoch": 0.25, "learning_rate": 1.7500019724314604e-05, "loss": 0.3783, "step": 5070 }, { "epoch": 0.25, "learning_rate": 1.749894994581734e-05, "loss": 0.281, "step": 5071 }, { "epoch": 0.25, "learning_rate": 1.7497879971194872e-05, "loss": 0.4105, "step": 5072 }, { "epoch": 0.25, "learning_rate": 1.7496809800475168e-05, "loss": 0.1941, "step": 5073 }, { "epoch": 0.25, "learning_rate": 1.749573943368622e-05, "loss": 0.4775, "step": 5074 }, { "epoch": 0.25, "learning_rate": 1.749466887085603e-05, "loss": 0.2967, "step": 5075 }, { "epoch": 0.25, "learning_rate": 1.7493598112012588e-05, "loss": 0.2079, "step": 5076 }, { "epoch": 0.25, "learning_rate": 1.7492527157183904e-05, "loss": 0.3643, "step": 5077 }, { "epoch": 0.25, "learning_rate": 1.7491456006397985e-05, "loss": 0.3654, "step": 5078 }, { "epoch": 0.25, "learning_rate": 1.7490384659682846e-05, "loss": 0.3046, "step": 5079 }, { "epoch": 0.25, "learning_rate": 1.748931311706651e-05, "loss": 0.491, "step": 5080 }, { "epoch": 0.25, "learning_rate": 1.7488241378576997e-05, "loss": 0.2442, "step": 5081 }, { "epoch": 0.25, "learning_rate": 1.7487169444242336e-05, "loss": 0.4984, "step": 5082 }, { "epoch": 0.25, "learning_rate": 1.7486097314090568e-05, "loss": 0.249, "step": 5083 }, { "epoch": 0.25, "learning_rate": 1.7485024988149726e-05, "loss": 0.6027, "step": 5084 }, { "epoch": 0.25, "learning_rate": 1.748395246644786e-05, "loss": 0.3935, "step": 5085 }, { "epoch": 0.25, "learning_rate": 1.7482879749013018e-05, "loss": 0.6433, "step": 5086 }, { "epoch": 0.25, "learning_rate": 1.7481806835873263e-05, "loss": 0.3246, "step": 5087 }, { "epoch": 0.25, "learning_rate": 1.7480733727056646e-05, "loss": 0.4586, "step": 5088 }, { "epoch": 0.25, "learning_rate": 1.7479660422591238e-05, "loss": 0.2648, "step": 5089 }, { "epoch": 0.25, "learning_rate": 1.7478586922505108e-05, "loss": 0.3926, "step": 5090 }, { "epoch": 0.25, "learning_rate": 1.747751322682633e-05, "loss": 0.3987, "step": 5091 }, { "epoch": 0.25, "learning_rate": 1.747643933558299e-05, "loss": 0.592, "step": 5092 }, { "epoch": 0.25, "learning_rate": 1.7475365248803168e-05, "loss": 0.42, "step": 5093 }, { "epoch": 0.25, "learning_rate": 1.7474290966514963e-05, "loss": 0.3303, "step": 5094 }, { "epoch": 0.25, "learning_rate": 1.7473216488746462e-05, "loss": 0.3169, "step": 5095 }, { "epoch": 0.25, "learning_rate": 1.7472141815525775e-05, "loss": 0.8461, "step": 5096 }, { "epoch": 0.25, "learning_rate": 1.7471066946881005e-05, "loss": 0.3589, "step": 5097 }, { "epoch": 0.25, "learning_rate": 1.7469991882840265e-05, "loss": 0.0701, "step": 5098 }, { "epoch": 0.25, "learning_rate": 1.7468916623431674e-05, "loss": 0.3563, "step": 5099 }, { "epoch": 0.25, "learning_rate": 1.7467841168683346e-05, "loss": 0.2123, "step": 5100 }, { "epoch": 0.25, "learning_rate": 1.7466765518623418e-05, "loss": 0.3607, "step": 5101 }, { "epoch": 0.25, "learning_rate": 1.7465689673280013e-05, "loss": 0.4248, "step": 5102 }, { "epoch": 0.25, "learning_rate": 1.7464613632681276e-05, "loss": 0.5665, "step": 5103 }, { "epoch": 0.25, "learning_rate": 1.7463537396855347e-05, "loss": 0.306, "step": 5104 }, { "epoch": 0.25, "learning_rate": 1.746246096583037e-05, "loss": 0.3237, "step": 5105 }, { "epoch": 0.25, "learning_rate": 1.74613843396345e-05, "loss": 0.6559, "step": 5106 }, { "epoch": 0.26, "learning_rate": 1.74603075182959e-05, "loss": 0.5551, "step": 5107 }, { "epoch": 0.26, "learning_rate": 1.7459230501842724e-05, "loss": 0.3405, "step": 5108 }, { "epoch": 0.26, "learning_rate": 1.7458153290303144e-05, "loss": 0.3483, "step": 5109 }, { "epoch": 0.26, "learning_rate": 1.7457075883705338e-05, "loss": 0.5024, "step": 5110 }, { "epoch": 0.26, "learning_rate": 1.7455998282077474e-05, "loss": 0.3346, "step": 5111 }, { "epoch": 0.26, "learning_rate": 1.745492048544774e-05, "loss": 0.57, "step": 5112 }, { "epoch": 0.26, "learning_rate": 1.745384249384433e-05, "loss": 0.3961, "step": 5113 }, { "epoch": 0.26, "learning_rate": 1.745276430729543e-05, "loss": 0.1424, "step": 5114 }, { "epoch": 0.26, "learning_rate": 1.7451685925829237e-05, "loss": 0.4379, "step": 5115 }, { "epoch": 0.26, "learning_rate": 1.7450607349473964e-05, "loss": 0.2522, "step": 5116 }, { "epoch": 0.26, "learning_rate": 1.7449528578257812e-05, "loss": 0.2343, "step": 5117 }, { "epoch": 0.26, "learning_rate": 1.7448449612208998e-05, "loss": 0.2873, "step": 5118 }, { "epoch": 0.26, "learning_rate": 1.744737045135574e-05, "loss": 0.3982, "step": 5119 }, { "epoch": 0.26, "learning_rate": 1.744629109572626e-05, "loss": 0.4104, "step": 5120 }, { "epoch": 0.26, "learning_rate": 1.7445211545348794e-05, "loss": 0.1353, "step": 5121 }, { "epoch": 0.26, "learning_rate": 1.7444131800251568e-05, "loss": 0.3366, "step": 5122 }, { "epoch": 0.26, "learning_rate": 1.7443051860462826e-05, "loss": 0.4802, "step": 5123 }, { "epoch": 0.26, "learning_rate": 1.744197172601081e-05, "loss": 0.3902, "step": 5124 }, { "epoch": 0.26, "learning_rate": 1.7440891396923772e-05, "loss": 0.1913, "step": 5125 }, { "epoch": 0.26, "learning_rate": 1.7439810873229965e-05, "loss": 0.257, "step": 5126 }, { "epoch": 0.26, "learning_rate": 1.7438730154957646e-05, "loss": 0.3446, "step": 5127 }, { "epoch": 0.26, "learning_rate": 1.7437649242135086e-05, "loss": 0.2957, "step": 5128 }, { "epoch": 0.26, "learning_rate": 1.743656813479055e-05, "loss": 0.4097, "step": 5129 }, { "epoch": 0.26, "learning_rate": 1.7435486832952316e-05, "loss": 0.4658, "step": 5130 }, { "epoch": 0.26, "learning_rate": 1.7434405336648665e-05, "loss": 0.2606, "step": 5131 }, { "epoch": 0.26, "learning_rate": 1.7433323645907875e-05, "loss": 0.4827, "step": 5132 }, { "epoch": 0.26, "learning_rate": 1.7432241760758245e-05, "loss": 0.2792, "step": 5133 }, { "epoch": 0.26, "learning_rate": 1.7431159681228062e-05, "loss": 0.257, "step": 5134 }, { "epoch": 0.26, "learning_rate": 1.7430077407345634e-05, "loss": 0.4501, "step": 5135 }, { "epoch": 0.26, "learning_rate": 1.7428994939139265e-05, "loss": 0.7121, "step": 5136 }, { "epoch": 0.26, "learning_rate": 1.7427912276637256e-05, "loss": 0.3937, "step": 5137 }, { "epoch": 0.26, "learning_rate": 1.7426829419867937e-05, "loss": 0.3196, "step": 5138 }, { "epoch": 0.26, "learning_rate": 1.7425746368859617e-05, "loss": 0.2513, "step": 5139 }, { "epoch": 0.26, "learning_rate": 1.7424663123640632e-05, "loss": 0.4087, "step": 5140 }, { "epoch": 0.26, "learning_rate": 1.7423579684239303e-05, "loss": 0.3999, "step": 5141 }, { "epoch": 0.26, "learning_rate": 1.7422496050683972e-05, "loss": 0.3946, "step": 5142 }, { "epoch": 0.26, "learning_rate": 1.742141222300298e-05, "loss": 0.4505, "step": 5143 }, { "epoch": 0.26, "learning_rate": 1.742032820122467e-05, "loss": 0.2605, "step": 5144 }, { "epoch": 0.26, "learning_rate": 1.7419243985377397e-05, "loss": 0.1621, "step": 5145 }, { "epoch": 0.26, "learning_rate": 1.7418159575489512e-05, "loss": 0.283, "step": 5146 }, { "epoch": 0.26, "learning_rate": 1.741707497158938e-05, "loss": 0.4719, "step": 5147 }, { "epoch": 0.26, "learning_rate": 1.741599017370537e-05, "loss": 0.2437, "step": 5148 }, { "epoch": 0.26, "learning_rate": 1.7414905181865846e-05, "loss": 0.2086, "step": 5149 }, { "epoch": 0.26, "learning_rate": 1.741381999609919e-05, "loss": 0.3643, "step": 5150 }, { "epoch": 0.26, "learning_rate": 1.7412734616433784e-05, "loss": 0.385, "step": 5151 }, { "epoch": 0.26, "learning_rate": 1.7411649042898013e-05, "loss": 0.1032, "step": 5152 }, { "epoch": 0.26, "learning_rate": 1.7410563275520266e-05, "loss": 0.4736, "step": 5153 }, { "epoch": 0.26, "learning_rate": 1.7409477314328943e-05, "loss": 0.3303, "step": 5154 }, { "epoch": 0.26, "learning_rate": 1.740839115935245e-05, "loss": 0.2022, "step": 5155 }, { "epoch": 0.26, "learning_rate": 1.7407304810619187e-05, "loss": 0.5733, "step": 5156 }, { "epoch": 0.26, "learning_rate": 1.7406218268157565e-05, "loss": 0.2814, "step": 5157 }, { "epoch": 0.26, "learning_rate": 1.7405131531996005e-05, "loss": 0.434, "step": 5158 }, { "epoch": 0.26, "learning_rate": 1.7404044602162934e-05, "loss": 0.5798, "step": 5159 }, { "epoch": 0.26, "learning_rate": 1.740295747868677e-05, "loss": 0.263, "step": 5160 }, { "epoch": 0.26, "learning_rate": 1.740187016159595e-05, "loss": 0.4647, "step": 5161 }, { "epoch": 0.26, "learning_rate": 1.7400782650918913e-05, "loss": 0.488, "step": 5162 }, { "epoch": 0.26, "learning_rate": 1.7399694946684096e-05, "loss": 0.3206, "step": 5163 }, { "epoch": 0.26, "learning_rate": 1.739860704891995e-05, "loss": 0.2628, "step": 5164 }, { "epoch": 0.26, "learning_rate": 1.7397518957654925e-05, "loss": 0.2882, "step": 5165 }, { "epoch": 0.26, "learning_rate": 1.7396430672917485e-05, "loss": 0.5616, "step": 5166 }, { "epoch": 0.26, "learning_rate": 1.7395342194736087e-05, "loss": 0.2496, "step": 5167 }, { "epoch": 0.26, "learning_rate": 1.7394253523139202e-05, "loss": 0.4468, "step": 5168 }, { "epoch": 0.26, "learning_rate": 1.7393164658155296e-05, "loss": 0.1763, "step": 5169 }, { "epoch": 0.26, "learning_rate": 1.7392075599812853e-05, "loss": 0.3533, "step": 5170 }, { "epoch": 0.26, "learning_rate": 1.739098634814036e-05, "loss": 0.2967, "step": 5171 }, { "epoch": 0.26, "learning_rate": 1.7389896903166295e-05, "loss": 0.3621, "step": 5172 }, { "epoch": 0.26, "learning_rate": 1.7388807264919154e-05, "loss": 0.3628, "step": 5173 }, { "epoch": 0.26, "learning_rate": 1.738771743342744e-05, "loss": 0.259, "step": 5174 }, { "epoch": 0.26, "learning_rate": 1.738662740871965e-05, "loss": 0.4224, "step": 5175 }, { "epoch": 0.26, "learning_rate": 1.7385537190824298e-05, "loss": 0.5377, "step": 5176 }, { "epoch": 0.26, "learning_rate": 1.738444677976989e-05, "loss": 0.3012, "step": 5177 }, { "epoch": 0.26, "learning_rate": 1.738335617558495e-05, "loss": 0.2837, "step": 5178 }, { "epoch": 0.26, "learning_rate": 1.7382265378298e-05, "loss": 0.5223, "step": 5179 }, { "epoch": 0.26, "learning_rate": 1.738117438793757e-05, "loss": 0.3774, "step": 5180 }, { "epoch": 0.26, "learning_rate": 1.738008320453219e-05, "loss": 0.265, "step": 5181 }, { "epoch": 0.26, "learning_rate": 1.73789918281104e-05, "loss": 0.3818, "step": 5182 }, { "epoch": 0.26, "learning_rate": 1.737790025870074e-05, "loss": 0.2372, "step": 5183 }, { "epoch": 0.26, "learning_rate": 1.7376808496331764e-05, "loss": 0.2931, "step": 5184 }, { "epoch": 0.26, "learning_rate": 1.7375716541032026e-05, "loss": 0.3459, "step": 5185 }, { "epoch": 0.26, "learning_rate": 1.7374624392830083e-05, "loss": 0.3175, "step": 5186 }, { "epoch": 0.26, "learning_rate": 1.7373532051754492e-05, "loss": 0.427, "step": 5187 }, { "epoch": 0.26, "learning_rate": 1.7372439517833836e-05, "loss": 0.752, "step": 5188 }, { "epoch": 0.26, "learning_rate": 1.7371346791096674e-05, "loss": 0.3776, "step": 5189 }, { "epoch": 0.26, "learning_rate": 1.7370253871571594e-05, "loss": 0.4794, "step": 5190 }, { "epoch": 0.26, "learning_rate": 1.7369160759287178e-05, "loss": 0.2935, "step": 5191 }, { "epoch": 0.26, "learning_rate": 1.7368067454272017e-05, "loss": 0.3819, "step": 5192 }, { "epoch": 0.26, "learning_rate": 1.7366973956554695e-05, "loss": 0.3613, "step": 5193 }, { "epoch": 0.26, "learning_rate": 1.7365880266163825e-05, "loss": 0.4067, "step": 5194 }, { "epoch": 0.26, "learning_rate": 1.7364786383128e-05, "loss": 0.1793, "step": 5195 }, { "epoch": 0.26, "learning_rate": 1.7363692307475832e-05, "loss": 0.6517, "step": 5196 }, { "epoch": 0.26, "learning_rate": 1.7362598039235938e-05, "loss": 0.5154, "step": 5197 }, { "epoch": 0.26, "learning_rate": 1.7361503578436937e-05, "loss": 0.3656, "step": 5198 }, { "epoch": 0.26, "learning_rate": 1.736040892510745e-05, "loss": 0.5963, "step": 5199 }, { "epoch": 0.26, "learning_rate": 1.735931407927611e-05, "loss": 0.2749, "step": 5200 }, { "epoch": 0.26, "learning_rate": 1.7358219040971547e-05, "loss": 0.3718, "step": 5201 }, { "epoch": 0.26, "learning_rate": 1.7357123810222406e-05, "loss": 0.5179, "step": 5202 }, { "epoch": 0.26, "learning_rate": 1.735602838705732e-05, "loss": 0.5935, "step": 5203 }, { "epoch": 0.26, "learning_rate": 1.7354932771504957e-05, "loss": 0.3364, "step": 5204 }, { "epoch": 0.26, "learning_rate": 1.7353836963593952e-05, "loss": 0.3246, "step": 5205 }, { "epoch": 0.26, "learning_rate": 1.7352740963352977e-05, "loss": 0.4236, "step": 5206 }, { "epoch": 0.26, "learning_rate": 1.735164477081069e-05, "loss": 0.365, "step": 5207 }, { "epoch": 0.26, "learning_rate": 1.7350548385995768e-05, "loss": 0.4906, "step": 5208 }, { "epoch": 0.26, "learning_rate": 1.7349451808936878e-05, "loss": 0.3814, "step": 5209 }, { "epoch": 0.26, "learning_rate": 1.7348355039662702e-05, "loss": 0.3548, "step": 5210 }, { "epoch": 0.26, "learning_rate": 1.734725807820192e-05, "loss": 0.2496, "step": 5211 }, { "epoch": 0.26, "learning_rate": 1.7346160924583232e-05, "loss": 0.3571, "step": 5212 }, { "epoch": 0.26, "learning_rate": 1.7345063578835323e-05, "loss": 0.4686, "step": 5213 }, { "epoch": 0.26, "learning_rate": 1.7343966040986897e-05, "loss": 0.304, "step": 5214 }, { "epoch": 0.26, "learning_rate": 1.7342868311066662e-05, "loss": 0.4588, "step": 5215 }, { "epoch": 0.26, "learning_rate": 1.734177038910332e-05, "loss": 0.5455, "step": 5216 }, { "epoch": 0.26, "learning_rate": 1.7340672275125593e-05, "loss": 0.4144, "step": 5217 }, { "epoch": 0.26, "learning_rate": 1.733957396916219e-05, "loss": 0.3254, "step": 5218 }, { "epoch": 0.26, "learning_rate": 1.733847547124185e-05, "loss": 0.2914, "step": 5219 }, { "epoch": 0.26, "learning_rate": 1.733737678139329e-05, "loss": 0.5122, "step": 5220 }, { "epoch": 0.26, "learning_rate": 1.7336277899645256e-05, "loss": 0.4026, "step": 5221 }, { "epoch": 0.26, "learning_rate": 1.733517882602648e-05, "loss": 0.4797, "step": 5222 }, { "epoch": 0.26, "learning_rate": 1.733407956056571e-05, "loss": 0.3132, "step": 5223 }, { "epoch": 0.26, "learning_rate": 1.7332980103291694e-05, "loss": 0.5331, "step": 5224 }, { "epoch": 0.26, "learning_rate": 1.7331880454233186e-05, "loss": 0.2343, "step": 5225 }, { "epoch": 0.26, "learning_rate": 1.7330780613418952e-05, "loss": 0.4162, "step": 5226 }, { "epoch": 0.26, "learning_rate": 1.732968058087775e-05, "loss": 0.6524, "step": 5227 }, { "epoch": 0.26, "learning_rate": 1.732858035663835e-05, "loss": 0.5479, "step": 5228 }, { "epoch": 0.26, "learning_rate": 1.7327479940729535e-05, "loss": 0.3728, "step": 5229 }, { "epoch": 0.26, "learning_rate": 1.7326379333180078e-05, "loss": 0.4298, "step": 5230 }, { "epoch": 0.26, "learning_rate": 1.732527853401876e-05, "loss": 0.4275, "step": 5231 }, { "epoch": 0.26, "learning_rate": 1.7324177543274382e-05, "loss": 0.3687, "step": 5232 }, { "epoch": 0.26, "learning_rate": 1.7323076360975736e-05, "loss": 0.4596, "step": 5233 }, { "epoch": 0.26, "learning_rate": 1.7321974987151614e-05, "loss": 0.2952, "step": 5234 }, { "epoch": 0.26, "learning_rate": 1.732087342183083e-05, "loss": 0.3367, "step": 5235 }, { "epoch": 0.26, "learning_rate": 1.7319771665042183e-05, "loss": 0.4251, "step": 5236 }, { "epoch": 0.26, "learning_rate": 1.7318669716814502e-05, "loss": 0.3298, "step": 5237 }, { "epoch": 0.26, "learning_rate": 1.7317567577176597e-05, "loss": 0.5332, "step": 5238 }, { "epoch": 0.26, "learning_rate": 1.7316465246157295e-05, "loss": 0.4313, "step": 5239 }, { "epoch": 0.26, "learning_rate": 1.731536272378543e-05, "loss": 0.4957, "step": 5240 }, { "epoch": 0.26, "learning_rate": 1.7314260010089835e-05, "loss": 0.4006, "step": 5241 }, { "epoch": 0.26, "learning_rate": 1.731315710509935e-05, "loss": 0.3207, "step": 5242 }, { "epoch": 0.26, "learning_rate": 1.731205400884282e-05, "loss": 0.3238, "step": 5243 }, { "epoch": 0.26, "learning_rate": 1.731095072134909e-05, "loss": 0.2976, "step": 5244 }, { "epoch": 0.26, "learning_rate": 1.7309847242647025e-05, "loss": 0.4494, "step": 5245 }, { "epoch": 0.26, "learning_rate": 1.7308743572765475e-05, "loss": 0.3906, "step": 5246 }, { "epoch": 0.26, "learning_rate": 1.7307639711733312e-05, "loss": 0.4156, "step": 5247 }, { "epoch": 0.26, "learning_rate": 1.7306535659579403e-05, "loss": 0.382, "step": 5248 }, { "epoch": 0.26, "learning_rate": 1.7305431416332625e-05, "loss": 0.3014, "step": 5249 }, { "epoch": 0.26, "learning_rate": 1.7304326982021855e-05, "loss": 0.3268, "step": 5250 }, { "epoch": 0.26, "learning_rate": 1.730322235667598e-05, "loss": 0.4553, "step": 5251 }, { "epoch": 0.26, "learning_rate": 1.7302117540323887e-05, "loss": 0.5277, "step": 5252 }, { "epoch": 0.26, "learning_rate": 1.7301012532994478e-05, "loss": 0.2527, "step": 5253 }, { "epoch": 0.26, "learning_rate": 1.729990733471665e-05, "loss": 0.3918, "step": 5254 }, { "epoch": 0.26, "learning_rate": 1.72988019455193e-05, "loss": 0.5003, "step": 5255 }, { "epoch": 0.26, "learning_rate": 1.729769636543135e-05, "loss": 0.4243, "step": 5256 }, { "epoch": 0.26, "learning_rate": 1.729659059448171e-05, "loss": 0.4395, "step": 5257 }, { "epoch": 0.26, "learning_rate": 1.72954846326993e-05, "loss": 0.6027, "step": 5258 }, { "epoch": 0.26, "learning_rate": 1.7294378480113045e-05, "loss": 0.6483, "step": 5259 }, { "epoch": 0.26, "learning_rate": 1.729327213675187e-05, "loss": 0.5441, "step": 5260 }, { "epoch": 0.26, "learning_rate": 1.729216560264472e-05, "loss": 0.4738, "step": 5261 }, { "epoch": 0.26, "learning_rate": 1.7291058877820524e-05, "loss": 0.3908, "step": 5262 }, { "epoch": 0.26, "learning_rate": 1.7289951962308237e-05, "loss": 0.4047, "step": 5263 }, { "epoch": 0.26, "learning_rate": 1.7288844856136803e-05, "loss": 0.2992, "step": 5264 }, { "epoch": 0.26, "learning_rate": 1.728773755933518e-05, "loss": 0.518, "step": 5265 }, { "epoch": 0.26, "learning_rate": 1.728663007193233e-05, "loss": 0.2124, "step": 5266 }, { "epoch": 0.26, "learning_rate": 1.728552239395721e-05, "loss": 0.4352, "step": 5267 }, { "epoch": 0.26, "learning_rate": 1.728441452543879e-05, "loss": 0.4223, "step": 5268 }, { "epoch": 0.26, "learning_rate": 1.7283306466406052e-05, "loss": 0.3893, "step": 5269 }, { "epoch": 0.26, "learning_rate": 1.7282198216887977e-05, "loss": 0.379, "step": 5270 }, { "epoch": 0.26, "learning_rate": 1.728108977691354e-05, "loss": 0.1255, "step": 5271 }, { "epoch": 0.26, "learning_rate": 1.7279981146511738e-05, "loss": 0.4916, "step": 5272 }, { "epoch": 0.26, "learning_rate": 1.7278872325711566e-05, "loss": 0.4708, "step": 5273 }, { "epoch": 0.26, "learning_rate": 1.7277763314542024e-05, "loss": 0.5494, "step": 5274 }, { "epoch": 0.26, "learning_rate": 1.7276654113032106e-05, "loss": 0.3979, "step": 5275 }, { "epoch": 0.26, "learning_rate": 1.7275544721210836e-05, "loss": 0.3985, "step": 5276 }, { "epoch": 0.26, "learning_rate": 1.7274435139107226e-05, "loss": 0.3895, "step": 5277 }, { "epoch": 0.26, "learning_rate": 1.7273325366750292e-05, "loss": 0.3111, "step": 5278 }, { "epoch": 0.26, "learning_rate": 1.7272215404169053e-05, "loss": 0.4237, "step": 5279 }, { "epoch": 0.26, "learning_rate": 1.727110525139255e-05, "loss": 0.5216, "step": 5280 }, { "epoch": 0.26, "learning_rate": 1.726999490844981e-05, "loss": 0.3001, "step": 5281 }, { "epoch": 0.26, "learning_rate": 1.726888437536988e-05, "loss": 0.3544, "step": 5282 }, { "epoch": 0.26, "learning_rate": 1.7267773652181797e-05, "loss": 0.2268, "step": 5283 }, { "epoch": 0.26, "learning_rate": 1.726666273891461e-05, "loss": 0.2713, "step": 5284 }, { "epoch": 0.26, "learning_rate": 1.726555163559738e-05, "loss": 0.3086, "step": 5285 }, { "epoch": 0.26, "learning_rate": 1.7264440342259163e-05, "loss": 0.3498, "step": 5286 }, { "epoch": 0.26, "learning_rate": 1.726332885892902e-05, "loss": 0.3691, "step": 5287 }, { "epoch": 0.26, "learning_rate": 1.726221718563603e-05, "loss": 0.3147, "step": 5288 }, { "epoch": 0.26, "learning_rate": 1.726110532240926e-05, "loss": 0.304, "step": 5289 }, { "epoch": 0.26, "learning_rate": 1.7259993269277786e-05, "loss": 0.3731, "step": 5290 }, { "epoch": 0.26, "learning_rate": 1.7258881026270703e-05, "loss": 0.5028, "step": 5291 }, { "epoch": 0.26, "learning_rate": 1.725776859341709e-05, "loss": 0.5759, "step": 5292 }, { "epoch": 0.26, "learning_rate": 1.7256655970746047e-05, "loss": 0.4769, "step": 5293 }, { "epoch": 0.26, "learning_rate": 1.725554315828667e-05, "loss": 0.331, "step": 5294 }, { "epoch": 0.26, "learning_rate": 1.7254430156068064e-05, "loss": 0.311, "step": 5295 }, { "epoch": 0.26, "learning_rate": 1.725331696411934e-05, "loss": 0.3586, "step": 5296 }, { "epoch": 0.26, "learning_rate": 1.7252203582469615e-05, "loss": 0.6016, "step": 5297 }, { "epoch": 0.26, "learning_rate": 1.7251090011147998e-05, "loss": 0.4911, "step": 5298 }, { "epoch": 0.26, "learning_rate": 1.7249976250183622e-05, "loss": 0.2031, "step": 5299 }, { "epoch": 0.26, "learning_rate": 1.7248862299605612e-05, "loss": 0.3083, "step": 5300 }, { "epoch": 0.26, "learning_rate": 1.7247748159443106e-05, "loss": 0.3113, "step": 5301 }, { "epoch": 0.26, "learning_rate": 1.7246633829725237e-05, "loss": 0.3355, "step": 5302 }, { "epoch": 0.26, "learning_rate": 1.724551931048115e-05, "loss": 0.6348, "step": 5303 }, { "epoch": 0.26, "learning_rate": 1.724440460174e-05, "loss": 0.3527, "step": 5304 }, { "epoch": 0.26, "learning_rate": 1.724328970353093e-05, "loss": 0.1918, "step": 5305 }, { "epoch": 0.26, "learning_rate": 1.724217461588311e-05, "loss": 0.3145, "step": 5306 }, { "epoch": 0.26, "learning_rate": 1.7241059338825696e-05, "loss": 0.2656, "step": 5307 }, { "epoch": 0.27, "learning_rate": 1.7239943872387857e-05, "loss": 0.5652, "step": 5308 }, { "epoch": 0.27, "learning_rate": 1.7238828216598774e-05, "loss": 0.6269, "step": 5309 }, { "epoch": 0.27, "learning_rate": 1.7237712371487616e-05, "loss": 0.2055, "step": 5310 }, { "epoch": 0.27, "learning_rate": 1.723659633708357e-05, "loss": 0.2883, "step": 5311 }, { "epoch": 0.27, "learning_rate": 1.723548011341583e-05, "loss": 0.4785, "step": 5312 }, { "epoch": 0.27, "learning_rate": 1.7234363700513582e-05, "loss": 0.3589, "step": 5313 }, { "epoch": 0.27, "learning_rate": 1.7233247098406023e-05, "loss": 0.5116, "step": 5314 }, { "epoch": 0.27, "learning_rate": 1.7232130307122362e-05, "loss": 0.4066, "step": 5315 }, { "epoch": 0.27, "learning_rate": 1.723101332669181e-05, "loss": 0.3876, "step": 5316 }, { "epoch": 0.27, "learning_rate": 1.722989615714357e-05, "loss": 0.3661, "step": 5317 }, { "epoch": 0.27, "learning_rate": 1.7228778798506867e-05, "loss": 0.3062, "step": 5318 }, { "epoch": 0.27, "learning_rate": 1.7227661250810922e-05, "loss": 0.5083, "step": 5319 }, { "epoch": 0.27, "learning_rate": 1.7226543514084966e-05, "loss": 0.5666, "step": 5320 }, { "epoch": 0.27, "learning_rate": 1.7225425588358228e-05, "loss": 0.3596, "step": 5321 }, { "epoch": 0.27, "learning_rate": 1.7224307473659946e-05, "loss": 0.4131, "step": 5322 }, { "epoch": 0.27, "learning_rate": 1.7223189170019365e-05, "loss": 0.6584, "step": 5323 }, { "epoch": 0.27, "learning_rate": 1.7222070677465734e-05, "loss": 0.377, "step": 5324 }, { "epoch": 0.27, "learning_rate": 1.72209519960283e-05, "loss": 0.4581, "step": 5325 }, { "epoch": 0.27, "learning_rate": 1.7219833125736326e-05, "loss": 0.4964, "step": 5326 }, { "epoch": 0.27, "learning_rate": 1.7218714066619073e-05, "loss": 0.3597, "step": 5327 }, { "epoch": 0.27, "learning_rate": 1.721759481870581e-05, "loss": 0.3832, "step": 5328 }, { "epoch": 0.27, "learning_rate": 1.7216475382025803e-05, "loss": 0.4495, "step": 5329 }, { "epoch": 0.27, "learning_rate": 1.721535575660834e-05, "loss": 0.557, "step": 5330 }, { "epoch": 0.27, "learning_rate": 1.7214235942482692e-05, "loss": 0.4867, "step": 5331 }, { "epoch": 0.27, "learning_rate": 1.7213115939678156e-05, "loss": 0.4561, "step": 5332 }, { "epoch": 0.27, "learning_rate": 1.721199574822402e-05, "loss": 0.4234, "step": 5333 }, { "epoch": 0.27, "learning_rate": 1.721087536814958e-05, "loss": 0.516, "step": 5334 }, { "epoch": 0.27, "learning_rate": 1.720975479948414e-05, "loss": 0.225, "step": 5335 }, { "epoch": 0.27, "learning_rate": 1.7208634042257006e-05, "loss": 0.2896, "step": 5336 }, { "epoch": 0.27, "learning_rate": 1.720751309649749e-05, "loss": 0.3788, "step": 5337 }, { "epoch": 0.27, "learning_rate": 1.7206391962234907e-05, "loss": 0.4975, "step": 5338 }, { "epoch": 0.27, "learning_rate": 1.7205270639498587e-05, "loss": 0.4957, "step": 5339 }, { "epoch": 0.27, "learning_rate": 1.7204149128317845e-05, "loss": 0.3696, "step": 5340 }, { "epoch": 0.27, "learning_rate": 1.7203027428722018e-05, "loss": 0.5495, "step": 5341 }, { "epoch": 0.27, "learning_rate": 1.7201905540740445e-05, "loss": 0.4394, "step": 5342 }, { "epoch": 0.27, "learning_rate": 1.7200783464402466e-05, "loss": 0.4821, "step": 5343 }, { "epoch": 0.27, "learning_rate": 1.7199661199737424e-05, "loss": 0.4997, "step": 5344 }, { "epoch": 0.27, "learning_rate": 1.7198538746774677e-05, "loss": 0.2327, "step": 5345 }, { "epoch": 0.27, "learning_rate": 1.7197416105543578e-05, "loss": 0.3849, "step": 5346 }, { "epoch": 0.27, "learning_rate": 1.7196293276073483e-05, "loss": 0.3231, "step": 5347 }, { "epoch": 0.27, "learning_rate": 1.7195170258393766e-05, "loss": 0.4581, "step": 5348 }, { "epoch": 0.27, "learning_rate": 1.7194047052533793e-05, "loss": 0.4396, "step": 5349 }, { "epoch": 0.27, "learning_rate": 1.719292365852294e-05, "loss": 0.4551, "step": 5350 }, { "epoch": 0.27, "learning_rate": 1.71918000763906e-05, "loss": 0.2357, "step": 5351 }, { "epoch": 0.27, "learning_rate": 1.7190676306166136e-05, "loss": 0.3744, "step": 5352 }, { "epoch": 0.27, "learning_rate": 1.718955234787896e-05, "loss": 0.491, "step": 5353 }, { "epoch": 0.27, "learning_rate": 1.7188428201558455e-05, "loss": 0.5664, "step": 5354 }, { "epoch": 0.27, "learning_rate": 1.7187303867234028e-05, "loss": 0.2006, "step": 5355 }, { "epoch": 0.27, "learning_rate": 1.7186179344935076e-05, "loss": 0.3707, "step": 5356 }, { "epoch": 0.27, "learning_rate": 1.718505463469102e-05, "loss": 0.4953, "step": 5357 }, { "epoch": 0.27, "learning_rate": 1.718392973653127e-05, "loss": 0.6105, "step": 5358 }, { "epoch": 0.27, "learning_rate": 1.718280465048525e-05, "loss": 0.3226, "step": 5359 }, { "epoch": 0.27, "learning_rate": 1.7181679376582376e-05, "loss": 0.3664, "step": 5360 }, { "epoch": 0.27, "learning_rate": 1.7180553914852088e-05, "loss": 0.351, "step": 5361 }, { "epoch": 0.27, "learning_rate": 1.7179428265323818e-05, "loss": 0.2162, "step": 5362 }, { "epoch": 0.27, "learning_rate": 1.7178302428027005e-05, "loss": 0.4027, "step": 5363 }, { "epoch": 0.27, "learning_rate": 1.7177176402991093e-05, "loss": 0.4197, "step": 5364 }, { "epoch": 0.27, "learning_rate": 1.717605019024553e-05, "loss": 0.314, "step": 5365 }, { "epoch": 0.27, "learning_rate": 1.7174923789819775e-05, "loss": 0.3123, "step": 5366 }, { "epoch": 0.27, "learning_rate": 1.717379720174329e-05, "loss": 0.2204, "step": 5367 }, { "epoch": 0.27, "learning_rate": 1.717267042604553e-05, "loss": 0.4767, "step": 5368 }, { "epoch": 0.27, "learning_rate": 1.717154346275597e-05, "loss": 0.4802, "step": 5369 }, { "epoch": 0.27, "learning_rate": 1.7170416311904085e-05, "loss": 0.149, "step": 5370 }, { "epoch": 0.27, "learning_rate": 1.7169288973519353e-05, "loss": 0.5789, "step": 5371 }, { "epoch": 0.27, "learning_rate": 1.7168161447631257e-05, "loss": 0.4072, "step": 5372 }, { "epoch": 0.27, "learning_rate": 1.7167033734269288e-05, "loss": 0.4462, "step": 5373 }, { "epoch": 0.27, "learning_rate": 1.716590583346294e-05, "loss": 0.4115, "step": 5374 }, { "epoch": 0.27, "learning_rate": 1.716477774524171e-05, "loss": 0.3693, "step": 5375 }, { "epoch": 0.27, "learning_rate": 1.71636494696351e-05, "loss": 0.3182, "step": 5376 }, { "epoch": 0.27, "learning_rate": 1.7162521006672622e-05, "loss": 0.4714, "step": 5377 }, { "epoch": 0.27, "learning_rate": 1.716139235638379e-05, "loss": 0.4452, "step": 5378 }, { "epoch": 0.27, "learning_rate": 1.7160263518798117e-05, "loss": 0.4715, "step": 5379 }, { "epoch": 0.27, "learning_rate": 1.7159134493945135e-05, "loss": 0.345, "step": 5380 }, { "epoch": 0.27, "learning_rate": 1.715800528185436e-05, "loss": 0.3156, "step": 5381 }, { "epoch": 0.27, "learning_rate": 1.7156875882555337e-05, "loss": 0.2964, "step": 5382 }, { "epoch": 0.27, "learning_rate": 1.7155746296077597e-05, "loss": 0.3364, "step": 5383 }, { "epoch": 0.27, "learning_rate": 1.7154616522450688e-05, "loss": 0.4714, "step": 5384 }, { "epoch": 0.27, "learning_rate": 1.7153486561704152e-05, "loss": 0.2132, "step": 5385 }, { "epoch": 0.27, "learning_rate": 1.7152356413867547e-05, "loss": 0.4152, "step": 5386 }, { "epoch": 0.27, "learning_rate": 1.7151226078970427e-05, "loss": 0.2927, "step": 5387 }, { "epoch": 0.27, "learning_rate": 1.7150095557042354e-05, "loss": 0.2746, "step": 5388 }, { "epoch": 0.27, "learning_rate": 1.7148964848112897e-05, "loss": 0.3883, "step": 5389 }, { "epoch": 0.27, "learning_rate": 1.714783395221163e-05, "loss": 0.3696, "step": 5390 }, { "epoch": 0.27, "learning_rate": 1.714670286936812e-05, "loss": 0.2528, "step": 5391 }, { "epoch": 0.27, "learning_rate": 1.7145571599611965e-05, "loss": 0.1504, "step": 5392 }, { "epoch": 0.27, "learning_rate": 1.7144440142972745e-05, "loss": 0.439, "step": 5393 }, { "epoch": 0.27, "learning_rate": 1.7143308499480046e-05, "loss": 0.4064, "step": 5394 }, { "epoch": 0.27, "learning_rate": 1.714217666916347e-05, "loss": 0.3703, "step": 5395 }, { "epoch": 0.27, "learning_rate": 1.7141044652052618e-05, "loss": 0.2009, "step": 5396 }, { "epoch": 0.27, "learning_rate": 1.7139912448177097e-05, "loss": 0.3814, "step": 5397 }, { "epoch": 0.27, "learning_rate": 1.713878005756652e-05, "loss": 0.7109, "step": 5398 }, { "epoch": 0.27, "learning_rate": 1.7137647480250495e-05, "loss": 0.4372, "step": 5399 }, { "epoch": 0.27, "learning_rate": 1.7136514716258652e-05, "loss": 0.2891, "step": 5400 }, { "epoch": 0.27, "learning_rate": 1.7135381765620613e-05, "loss": 0.3642, "step": 5401 }, { "epoch": 0.27, "learning_rate": 1.7134248628366012e-05, "loss": 0.4123, "step": 5402 }, { "epoch": 0.27, "learning_rate": 1.713311530452448e-05, "loss": 0.4886, "step": 5403 }, { "epoch": 0.27, "learning_rate": 1.713198179412566e-05, "loss": 0.3942, "step": 5404 }, { "epoch": 0.27, "learning_rate": 1.7130848097199196e-05, "loss": 0.6271, "step": 5405 }, { "epoch": 0.27, "learning_rate": 1.7129714213774746e-05, "loss": 0.3341, "step": 5406 }, { "epoch": 0.27, "learning_rate": 1.712858014388195e-05, "loss": 0.3557, "step": 5407 }, { "epoch": 0.27, "learning_rate": 1.712744588755048e-05, "loss": 0.5554, "step": 5408 }, { "epoch": 0.27, "learning_rate": 1.7126311444810003e-05, "loss": 0.2323, "step": 5409 }, { "epoch": 0.27, "learning_rate": 1.7125176815690178e-05, "loss": 0.4505, "step": 5410 }, { "epoch": 0.27, "learning_rate": 1.712404200022069e-05, "loss": 0.3816, "step": 5411 }, { "epoch": 0.27, "learning_rate": 1.712290699843121e-05, "loss": 0.4448, "step": 5412 }, { "epoch": 0.27, "learning_rate": 1.712177181035143e-05, "loss": 0.6929, "step": 5413 }, { "epoch": 0.27, "learning_rate": 1.7120636436011038e-05, "loss": 0.4801, "step": 5414 }, { "epoch": 0.27, "learning_rate": 1.7119500875439723e-05, "loss": 0.4141, "step": 5415 }, { "epoch": 0.27, "learning_rate": 1.711836512866719e-05, "loss": 0.3839, "step": 5416 }, { "epoch": 0.27, "learning_rate": 1.7117229195723136e-05, "loss": 0.4291, "step": 5417 }, { "epoch": 0.27, "learning_rate": 1.7116093076637276e-05, "loss": 0.3741, "step": 5418 }, { "epoch": 0.27, "learning_rate": 1.7114956771439327e-05, "loss": 0.2425, "step": 5419 }, { "epoch": 0.27, "learning_rate": 1.7113820280158997e-05, "loss": 0.4741, "step": 5420 }, { "epoch": 0.27, "learning_rate": 1.7112683602826016e-05, "loss": 0.5228, "step": 5421 }, { "epoch": 0.27, "learning_rate": 1.7111546739470113e-05, "loss": 0.4325, "step": 5422 }, { "epoch": 0.27, "learning_rate": 1.7110409690121018e-05, "loss": 0.1815, "step": 5423 }, { "epoch": 0.27, "learning_rate": 1.7109272454808472e-05, "loss": 0.228, "step": 5424 }, { "epoch": 0.27, "learning_rate": 1.7108135033562216e-05, "loss": 0.3444, "step": 5425 }, { "epoch": 0.27, "learning_rate": 1.7106997426411997e-05, "loss": 0.6674, "step": 5426 }, { "epoch": 0.27, "learning_rate": 1.7105859633387568e-05, "loss": 0.5859, "step": 5427 }, { "epoch": 0.27, "learning_rate": 1.710472165451869e-05, "loss": 0.2971, "step": 5428 }, { "epoch": 0.27, "learning_rate": 1.710358348983512e-05, "loss": 0.5096, "step": 5429 }, { "epoch": 0.27, "learning_rate": 1.710244513936663e-05, "loss": 0.2119, "step": 5430 }, { "epoch": 0.27, "learning_rate": 1.7101306603142988e-05, "loss": 0.4396, "step": 5431 }, { "epoch": 0.27, "learning_rate": 1.7100167881193975e-05, "loss": 0.3825, "step": 5432 }, { "epoch": 0.27, "learning_rate": 1.7099028973549366e-05, "loss": 0.2928, "step": 5433 }, { "epoch": 0.27, "learning_rate": 1.7097889880238957e-05, "loss": 0.3624, "step": 5434 }, { "epoch": 0.27, "learning_rate": 1.7096750601292537e-05, "loss": 0.4971, "step": 5435 }, { "epoch": 0.27, "learning_rate": 1.7095611136739893e-05, "loss": 0.3565, "step": 5436 }, { "epoch": 0.27, "learning_rate": 1.709447148661084e-05, "loss": 0.526, "step": 5437 }, { "epoch": 0.27, "learning_rate": 1.7093331650935173e-05, "loss": 0.2954, "step": 5438 }, { "epoch": 0.27, "learning_rate": 1.709219162974271e-05, "loss": 0.3913, "step": 5439 }, { "epoch": 0.27, "learning_rate": 1.709105142306326e-05, "loss": 0.6686, "step": 5440 }, { "epoch": 0.27, "learning_rate": 1.708991103092665e-05, "loss": 0.4264, "step": 5441 }, { "epoch": 0.27, "learning_rate": 1.7088770453362702e-05, "loss": 0.4607, "step": 5442 }, { "epoch": 0.27, "learning_rate": 1.708762969040125e-05, "loss": 0.5243, "step": 5443 }, { "epoch": 0.27, "learning_rate": 1.7086488742072128e-05, "loss": 0.3402, "step": 5444 }, { "epoch": 0.27, "learning_rate": 1.708534760840517e-05, "loss": 0.5927, "step": 5445 }, { "epoch": 0.27, "learning_rate": 1.7084206289430224e-05, "loss": 0.4568, "step": 5446 }, { "epoch": 0.27, "learning_rate": 1.7083064785177147e-05, "loss": 0.5431, "step": 5447 }, { "epoch": 0.27, "learning_rate": 1.7081923095675786e-05, "loss": 0.3247, "step": 5448 }, { "epoch": 0.27, "learning_rate": 1.7080781220956e-05, "loss": 0.6723, "step": 5449 }, { "epoch": 0.27, "learning_rate": 1.7079639161047657e-05, "loss": 0.75, "step": 5450 }, { "epoch": 0.27, "learning_rate": 1.7078496915980624e-05, "loss": 0.5482, "step": 5451 }, { "epoch": 0.27, "learning_rate": 1.7077354485784777e-05, "loss": 0.597, "step": 5452 }, { "epoch": 0.27, "learning_rate": 1.7076211870489988e-05, "loss": 0.3147, "step": 5453 }, { "epoch": 0.27, "learning_rate": 1.707506907012615e-05, "loss": 0.3456, "step": 5454 }, { "epoch": 0.27, "learning_rate": 1.7073926084723146e-05, "loss": 0.2902, "step": 5455 }, { "epoch": 0.27, "learning_rate": 1.707278291431087e-05, "loss": 0.5207, "step": 5456 }, { "epoch": 0.27, "learning_rate": 1.707163955891922e-05, "loss": 0.4344, "step": 5457 }, { "epoch": 0.27, "learning_rate": 1.7070496018578103e-05, "loss": 0.4044, "step": 5458 }, { "epoch": 0.27, "learning_rate": 1.7069352293317418e-05, "loss": 0.5392, "step": 5459 }, { "epoch": 0.27, "learning_rate": 1.706820838316709e-05, "loss": 0.2412, "step": 5460 }, { "epoch": 0.27, "learning_rate": 1.7067064288157023e-05, "loss": 0.4592, "step": 5461 }, { "epoch": 0.27, "learning_rate": 1.706592000831715e-05, "loss": 0.5424, "step": 5462 }, { "epoch": 0.27, "learning_rate": 1.7064775543677392e-05, "loss": 0.2995, "step": 5463 }, { "epoch": 0.27, "learning_rate": 1.706363089426768e-05, "loss": 0.381, "step": 5464 }, { "epoch": 0.27, "learning_rate": 1.706248606011796e-05, "loss": 0.4655, "step": 5465 }, { "epoch": 0.27, "learning_rate": 1.7061341041258163e-05, "loss": 0.6916, "step": 5466 }, { "epoch": 0.27, "learning_rate": 1.7060195837718237e-05, "loss": 0.4386, "step": 5467 }, { "epoch": 0.27, "learning_rate": 1.705905044952814e-05, "loss": 0.3935, "step": 5468 }, { "epoch": 0.27, "learning_rate": 1.7057904876717826e-05, "loss": 0.4321, "step": 5469 }, { "epoch": 0.27, "learning_rate": 1.7056759119317252e-05, "loss": 0.3532, "step": 5470 }, { "epoch": 0.27, "learning_rate": 1.705561317735639e-05, "loss": 0.4597, "step": 5471 }, { "epoch": 0.27, "learning_rate": 1.7054467050865196e-05, "loss": 0.4146, "step": 5472 }, { "epoch": 0.27, "learning_rate": 1.7053320739873663e-05, "loss": 0.4169, "step": 5473 }, { "epoch": 0.27, "learning_rate": 1.7052174244411765e-05, "loss": 0.4605, "step": 5474 }, { "epoch": 0.27, "learning_rate": 1.7051027564509482e-05, "loss": 0.1665, "step": 5475 }, { "epoch": 0.27, "learning_rate": 1.7049880700196814e-05, "loss": 0.377, "step": 5476 }, { "epoch": 0.27, "learning_rate": 1.7048733651503745e-05, "loss": 0.4475, "step": 5477 }, { "epoch": 0.27, "learning_rate": 1.7047586418460284e-05, "loss": 0.192, "step": 5478 }, { "epoch": 0.27, "learning_rate": 1.704643900109643e-05, "loss": 0.2968, "step": 5479 }, { "epoch": 0.27, "learning_rate": 1.704529139944219e-05, "loss": 0.4002, "step": 5480 }, { "epoch": 0.27, "learning_rate": 1.7044143613527583e-05, "loss": 0.3573, "step": 5481 }, { "epoch": 0.27, "learning_rate": 1.7042995643382627e-05, "loss": 0.4155, "step": 5482 }, { "epoch": 0.27, "learning_rate": 1.704184748903734e-05, "loss": 0.2506, "step": 5483 }, { "epoch": 0.27, "learning_rate": 1.704069915052176e-05, "loss": 0.5472, "step": 5484 }, { "epoch": 0.27, "learning_rate": 1.703955062786591e-05, "loss": 0.4873, "step": 5485 }, { "epoch": 0.27, "learning_rate": 1.703840192109984e-05, "loss": 0.4677, "step": 5486 }, { "epoch": 0.27, "learning_rate": 1.7037253030253586e-05, "loss": 0.5036, "step": 5487 }, { "epoch": 0.27, "learning_rate": 1.703610395535719e-05, "loss": 0.2541, "step": 5488 }, { "epoch": 0.27, "learning_rate": 1.7034954696440718e-05, "loss": 0.461, "step": 5489 }, { "epoch": 0.27, "learning_rate": 1.7033805253534217e-05, "loss": 0.5519, "step": 5490 }, { "epoch": 0.27, "learning_rate": 1.7032655626667752e-05, "loss": 0.2174, "step": 5491 }, { "epoch": 0.27, "learning_rate": 1.703150581587139e-05, "loss": 0.3157, "step": 5492 }, { "epoch": 0.27, "learning_rate": 1.7030355821175202e-05, "loss": 0.2699, "step": 5493 }, { "epoch": 0.27, "learning_rate": 1.7029205642609267e-05, "loss": 0.4694, "step": 5494 }, { "epoch": 0.27, "learning_rate": 1.7028055280203667e-05, "loss": 0.5, "step": 5495 }, { "epoch": 0.27, "learning_rate": 1.7026904733988485e-05, "loss": 0.3715, "step": 5496 }, { "epoch": 0.27, "learning_rate": 1.7025754003993807e-05, "loss": 0.4355, "step": 5497 }, { "epoch": 0.27, "learning_rate": 1.7024603090249742e-05, "loss": 0.236, "step": 5498 }, { "epoch": 0.27, "learning_rate": 1.7023451992786385e-05, "loss": 0.6318, "step": 5499 }, { "epoch": 0.27, "learning_rate": 1.7022300711633834e-05, "loss": 0.2286, "step": 5500 }, { "epoch": 0.27, "learning_rate": 1.7021149246822212e-05, "loss": 0.4082, "step": 5501 }, { "epoch": 0.27, "learning_rate": 1.7019997598381623e-05, "loss": 0.4773, "step": 5502 }, { "epoch": 0.27, "learning_rate": 1.701884576634219e-05, "loss": 0.3564, "step": 5503 }, { "epoch": 0.27, "learning_rate": 1.7017693750734045e-05, "loss": 0.3138, "step": 5504 }, { "epoch": 0.27, "learning_rate": 1.7016541551587304e-05, "loss": 0.2125, "step": 5505 }, { "epoch": 0.27, "learning_rate": 1.7015389168932115e-05, "loss": 0.4779, "step": 5506 }, { "epoch": 0.27, "learning_rate": 1.701423660279861e-05, "loss": 0.3328, "step": 5507 }, { "epoch": 0.28, "learning_rate": 1.701308385321693e-05, "loss": 0.3786, "step": 5508 }, { "epoch": 0.28, "learning_rate": 1.701193092021723e-05, "loss": 0.5191, "step": 5509 }, { "epoch": 0.28, "learning_rate": 1.7010777803829663e-05, "loss": 0.3735, "step": 5510 }, { "epoch": 0.28, "learning_rate": 1.700962450408438e-05, "loss": 0.5859, "step": 5511 }, { "epoch": 0.28, "learning_rate": 1.7008471021011552e-05, "loss": 0.2675, "step": 5512 }, { "epoch": 0.28, "learning_rate": 1.7007317354641344e-05, "loss": 0.2756, "step": 5513 }, { "epoch": 0.28, "learning_rate": 1.7006163505003925e-05, "loss": 0.2798, "step": 5514 }, { "epoch": 0.28, "learning_rate": 1.7005009472129482e-05, "loss": 0.3864, "step": 5515 }, { "epoch": 0.28, "learning_rate": 1.7003855256048185e-05, "loss": 0.2908, "step": 5516 }, { "epoch": 0.28, "learning_rate": 1.700270085679023e-05, "loss": 0.2531, "step": 5517 }, { "epoch": 0.28, "learning_rate": 1.700154627438581e-05, "loss": 0.3674, "step": 5518 }, { "epoch": 0.28, "learning_rate": 1.7000391508865112e-05, "loss": 0.1712, "step": 5519 }, { "epoch": 0.28, "learning_rate": 1.6999236560258342e-05, "loss": 0.3647, "step": 5520 }, { "epoch": 0.28, "learning_rate": 1.699808142859571e-05, "loss": 0.3499, "step": 5521 }, { "epoch": 0.28, "learning_rate": 1.6996926113907423e-05, "loss": 0.369, "step": 5522 }, { "epoch": 0.28, "learning_rate": 1.69957706162237e-05, "loss": 0.4723, "step": 5523 }, { "epoch": 0.28, "learning_rate": 1.6994614935574758e-05, "loss": 0.3863, "step": 5524 }, { "epoch": 0.28, "learning_rate": 1.6993459071990825e-05, "loss": 0.4517, "step": 5525 }, { "epoch": 0.28, "learning_rate": 1.6992303025502127e-05, "loss": 0.2591, "step": 5526 }, { "epoch": 0.28, "learning_rate": 1.6991146796138905e-05, "loss": 0.2573, "step": 5527 }, { "epoch": 0.28, "learning_rate": 1.6989990383931393e-05, "loss": 0.1942, "step": 5528 }, { "epoch": 0.28, "learning_rate": 1.6988833788909837e-05, "loss": 0.3989, "step": 5529 }, { "epoch": 0.28, "learning_rate": 1.6987677011104487e-05, "loss": 0.5476, "step": 5530 }, { "epoch": 0.28, "learning_rate": 1.69865200505456e-05, "loss": 0.2838, "step": 5531 }, { "epoch": 0.28, "learning_rate": 1.698536290726343e-05, "loss": 0.6143, "step": 5532 }, { "epoch": 0.28, "learning_rate": 1.698420558128824e-05, "loss": 0.2937, "step": 5533 }, { "epoch": 0.28, "learning_rate": 1.69830480726503e-05, "loss": 0.2018, "step": 5534 }, { "epoch": 0.28, "learning_rate": 1.698189038137989e-05, "loss": 0.4511, "step": 5535 }, { "epoch": 0.28, "learning_rate": 1.6980732507507277e-05, "loss": 0.31, "step": 5536 }, { "epoch": 0.28, "learning_rate": 1.6979574451062752e-05, "loss": 0.2323, "step": 5537 }, { "epoch": 0.28, "learning_rate": 1.6978416212076596e-05, "loss": 0.1565, "step": 5538 }, { "epoch": 0.28, "learning_rate": 1.6977257790579106e-05, "loss": 0.4166, "step": 5539 }, { "epoch": 0.28, "learning_rate": 1.6976099186600575e-05, "loss": 0.243, "step": 5540 }, { "epoch": 0.28, "learning_rate": 1.6974940400171306e-05, "loss": 0.3485, "step": 5541 }, { "epoch": 0.28, "learning_rate": 1.697378143132161e-05, "loss": 0.254, "step": 5542 }, { "epoch": 0.28, "learning_rate": 1.6972622280081792e-05, "loss": 0.458, "step": 5543 }, { "epoch": 0.28, "learning_rate": 1.6971462946482172e-05, "loss": 0.3841, "step": 5544 }, { "epoch": 0.28, "learning_rate": 1.697030343055307e-05, "loss": 0.4978, "step": 5545 }, { "epoch": 0.28, "learning_rate": 1.6969143732324805e-05, "loss": 0.1914, "step": 5546 }, { "epoch": 0.28, "learning_rate": 1.6967983851827724e-05, "loss": 0.3809, "step": 5547 }, { "epoch": 0.28, "learning_rate": 1.6966823789092144e-05, "loss": 0.2785, "step": 5548 }, { "epoch": 0.28, "learning_rate": 1.6965663544148414e-05, "loss": 0.4756, "step": 5549 }, { "epoch": 0.28, "learning_rate": 1.6964503117026876e-05, "loss": 0.57, "step": 5550 }, { "epoch": 0.28, "learning_rate": 1.6963342507757884e-05, "loss": 0.5556, "step": 5551 }, { "epoch": 0.28, "learning_rate": 1.6962181716371793e-05, "loss": 0.5458, "step": 5552 }, { "epoch": 0.28, "learning_rate": 1.696102074289895e-05, "loss": 0.3842, "step": 5553 }, { "epoch": 0.28, "learning_rate": 1.6959859587369733e-05, "loss": 0.325, "step": 5554 }, { "epoch": 0.28, "learning_rate": 1.69586982498145e-05, "loss": 0.3169, "step": 5555 }, { "epoch": 0.28, "learning_rate": 1.695753673026363e-05, "loss": 0.3294, "step": 5556 }, { "epoch": 0.28, "learning_rate": 1.69563750287475e-05, "loss": 0.4838, "step": 5557 }, { "epoch": 0.28, "learning_rate": 1.6955213145296493e-05, "loss": 0.4011, "step": 5558 }, { "epoch": 0.28, "learning_rate": 1.6954051079940998e-05, "loss": 0.2164, "step": 5559 }, { "epoch": 0.28, "learning_rate": 1.6952888832711398e-05, "loss": 0.509, "step": 5560 }, { "epoch": 0.28, "learning_rate": 1.69517264036381e-05, "loss": 0.2686, "step": 5561 }, { "epoch": 0.28, "learning_rate": 1.6950563792751507e-05, "loss": 0.3698, "step": 5562 }, { "epoch": 0.28, "learning_rate": 1.6949401000082017e-05, "loss": 0.2243, "step": 5563 }, { "epoch": 0.28, "learning_rate": 1.694823802566005e-05, "loss": 0.4502, "step": 5564 }, { "epoch": 0.28, "learning_rate": 1.6947074869516015e-05, "loss": 0.286, "step": 5565 }, { "epoch": 0.28, "learning_rate": 1.6945911531680335e-05, "loss": 0.3467, "step": 5566 }, { "epoch": 0.28, "learning_rate": 1.694474801218344e-05, "loss": 0.375, "step": 5567 }, { "epoch": 0.28, "learning_rate": 1.6943584311055758e-05, "loss": 0.4396, "step": 5568 }, { "epoch": 0.28, "learning_rate": 1.6942420428327716e-05, "loss": 0.4524, "step": 5569 }, { "epoch": 0.28, "learning_rate": 1.6941256364029763e-05, "loss": 0.2798, "step": 5570 }, { "epoch": 0.28, "learning_rate": 1.694009211819234e-05, "loss": 0.3555, "step": 5571 }, { "epoch": 0.28, "learning_rate": 1.6938927690845905e-05, "loss": 0.3689, "step": 5572 }, { "epoch": 0.28, "learning_rate": 1.69377630820209e-05, "loss": 0.4555, "step": 5573 }, { "epoch": 0.28, "learning_rate": 1.6936598291747784e-05, "loss": 0.2554, "step": 5574 }, { "epoch": 0.28, "learning_rate": 1.693543332005703e-05, "loss": 0.3042, "step": 5575 }, { "epoch": 0.28, "learning_rate": 1.69342681669791e-05, "loss": 0.2275, "step": 5576 }, { "epoch": 0.28, "learning_rate": 1.693310283254447e-05, "loss": 0.2847, "step": 5577 }, { "epoch": 0.28, "learning_rate": 1.6931937316783615e-05, "loss": 0.312, "step": 5578 }, { "epoch": 0.28, "learning_rate": 1.693077161972702e-05, "loss": 0.3336, "step": 5579 }, { "epoch": 0.28, "learning_rate": 1.692960574140517e-05, "loss": 0.2213, "step": 5580 }, { "epoch": 0.28, "learning_rate": 1.6928439681848555e-05, "loss": 0.5413, "step": 5581 }, { "epoch": 0.28, "learning_rate": 1.6927273441087677e-05, "loss": 0.2603, "step": 5582 }, { "epoch": 0.28, "learning_rate": 1.6926107019153034e-05, "loss": 0.3426, "step": 5583 }, { "epoch": 0.28, "learning_rate": 1.6924940416075137e-05, "loss": 0.7146, "step": 5584 }, { "epoch": 0.28, "learning_rate": 1.692377363188449e-05, "loss": 0.3869, "step": 5585 }, { "epoch": 0.28, "learning_rate": 1.6922606666611615e-05, "loss": 0.4758, "step": 5586 }, { "epoch": 0.28, "learning_rate": 1.6921439520287027e-05, "loss": 0.2777, "step": 5587 }, { "epoch": 0.28, "learning_rate": 1.6920272192941256e-05, "loss": 0.4897, "step": 5588 }, { "epoch": 0.28, "learning_rate": 1.6919104684604828e-05, "loss": 0.4017, "step": 5589 }, { "epoch": 0.28, "learning_rate": 1.6917936995308282e-05, "loss": 0.4843, "step": 5590 }, { "epoch": 0.28, "learning_rate": 1.691676912508215e-05, "loss": 0.3978, "step": 5591 }, { "epoch": 0.28, "learning_rate": 1.6915601073956985e-05, "loss": 0.3895, "step": 5592 }, { "epoch": 0.28, "learning_rate": 1.691443284196333e-05, "loss": 0.1736, "step": 5593 }, { "epoch": 0.28, "learning_rate": 1.691326442913174e-05, "loss": 0.5192, "step": 5594 }, { "epoch": 0.28, "learning_rate": 1.6912095835492774e-05, "loss": 0.4714, "step": 5595 }, { "epoch": 0.28, "learning_rate": 1.6910927061076996e-05, "loss": 0.3914, "step": 5596 }, { "epoch": 0.28, "learning_rate": 1.6909758105914972e-05, "loss": 0.2701, "step": 5597 }, { "epoch": 0.28, "learning_rate": 1.6908588970037273e-05, "loss": 0.305, "step": 5598 }, { "epoch": 0.28, "learning_rate": 1.690741965347448e-05, "loss": 0.1111, "step": 5599 }, { "epoch": 0.28, "learning_rate": 1.690625015625717e-05, "loss": 0.3735, "step": 5600 }, { "epoch": 0.28, "learning_rate": 1.690508047841594e-05, "loss": 0.3877, "step": 5601 }, { "epoch": 0.28, "learning_rate": 1.6903910619981368e-05, "loss": 0.4506, "step": 5602 }, { "epoch": 0.28, "learning_rate": 1.6902740580984055e-05, "loss": 0.1805, "step": 5603 }, { "epoch": 0.28, "learning_rate": 1.6901570361454606e-05, "loss": 0.3259, "step": 5604 }, { "epoch": 0.28, "learning_rate": 1.6900399961423624e-05, "loss": 0.3855, "step": 5605 }, { "epoch": 0.28, "learning_rate": 1.689922938092172e-05, "loss": 0.2961, "step": 5606 }, { "epoch": 0.28, "learning_rate": 1.6898058619979506e-05, "loss": 0.4421, "step": 5607 }, { "epoch": 0.28, "learning_rate": 1.68968876786276e-05, "loss": 0.5108, "step": 5608 }, { "epoch": 0.28, "learning_rate": 1.6895716556896636e-05, "loss": 0.4386, "step": 5609 }, { "epoch": 0.28, "learning_rate": 1.6894545254817233e-05, "loss": 0.5241, "step": 5610 }, { "epoch": 0.28, "learning_rate": 1.6893373772420034e-05, "loss": 0.3576, "step": 5611 }, { "epoch": 0.28, "learning_rate": 1.6892202109735666e-05, "loss": 0.2451, "step": 5612 }, { "epoch": 0.28, "learning_rate": 1.6891030266794785e-05, "loss": 0.5043, "step": 5613 }, { "epoch": 0.28, "learning_rate": 1.688985824362803e-05, "loss": 0.2919, "step": 5614 }, { "epoch": 0.28, "learning_rate": 1.6888686040266057e-05, "loss": 0.3591, "step": 5615 }, { "epoch": 0.28, "learning_rate": 1.6887513656739524e-05, "loss": 0.3705, "step": 5616 }, { "epoch": 0.28, "learning_rate": 1.6886341093079092e-05, "loss": 0.364, "step": 5617 }, { "epoch": 0.28, "learning_rate": 1.688516834931543e-05, "loss": 0.258, "step": 5618 }, { "epoch": 0.28, "learning_rate": 1.6883995425479204e-05, "loss": 0.3562, "step": 5619 }, { "epoch": 0.28, "learning_rate": 1.6882822321601094e-05, "loss": 0.3092, "step": 5620 }, { "epoch": 0.28, "learning_rate": 1.6881649037711786e-05, "loss": 0.3569, "step": 5621 }, { "epoch": 0.28, "learning_rate": 1.6880475573841956e-05, "loss": 0.4306, "step": 5622 }, { "epoch": 0.28, "learning_rate": 1.6879301930022304e-05, "loss": 0.2612, "step": 5623 }, { "epoch": 0.28, "learning_rate": 1.6878128106283512e-05, "loss": 0.5883, "step": 5624 }, { "epoch": 0.28, "learning_rate": 1.68769541026563e-05, "loss": 0.2809, "step": 5625 }, { "epoch": 0.28, "learning_rate": 1.687577991917135e-05, "loss": 0.309, "step": 5626 }, { "epoch": 0.28, "learning_rate": 1.6874605555859388e-05, "loss": 0.4879, "step": 5627 }, { "epoch": 0.28, "learning_rate": 1.6873431012751117e-05, "loss": 0.5485, "step": 5628 }, { "epoch": 0.28, "learning_rate": 1.6872256289877264e-05, "loss": 0.4885, "step": 5629 }, { "epoch": 0.28, "learning_rate": 1.687108138726855e-05, "loss": 0.5414, "step": 5630 }, { "epoch": 0.28, "learning_rate": 1.68699063049557e-05, "loss": 0.5204, "step": 5631 }, { "epoch": 0.28, "learning_rate": 1.6868731042969448e-05, "loss": 0.2885, "step": 5632 }, { "epoch": 0.28, "learning_rate": 1.6867555601340532e-05, "loss": 0.5166, "step": 5633 }, { "epoch": 0.28, "learning_rate": 1.6866379980099693e-05, "loss": 0.3205, "step": 5634 }, { "epoch": 0.28, "learning_rate": 1.6865204179277678e-05, "loss": 0.5682, "step": 5635 }, { "epoch": 0.28, "learning_rate": 1.6864028198905243e-05, "loss": 0.5116, "step": 5636 }, { "epoch": 0.28, "learning_rate": 1.686285203901314e-05, "loss": 0.3872, "step": 5637 }, { "epoch": 0.28, "learning_rate": 1.686167569963213e-05, "loss": 0.605, "step": 5638 }, { "epoch": 0.28, "learning_rate": 1.6860499180792982e-05, "loss": 0.4613, "step": 5639 }, { "epoch": 0.28, "learning_rate": 1.685932248252646e-05, "loss": 0.3349, "step": 5640 }, { "epoch": 0.28, "learning_rate": 1.6858145604863345e-05, "loss": 0.2684, "step": 5641 }, { "epoch": 0.28, "learning_rate": 1.685696854783441e-05, "loss": 0.213, "step": 5642 }, { "epoch": 0.28, "learning_rate": 1.6855791311470448e-05, "loss": 0.5463, "step": 5643 }, { "epoch": 0.28, "learning_rate": 1.685461389580224e-05, "loss": 0.1854, "step": 5644 }, { "epoch": 0.28, "learning_rate": 1.6853436300860586e-05, "loss": 0.4164, "step": 5645 }, { "epoch": 0.28, "learning_rate": 1.685225852667628e-05, "loss": 0.6555, "step": 5646 }, { "epoch": 0.28, "learning_rate": 1.6851080573280132e-05, "loss": 0.2636, "step": 5647 }, { "epoch": 0.28, "learning_rate": 1.684990244070294e-05, "loss": 0.1973, "step": 5648 }, { "epoch": 0.28, "learning_rate": 1.684872412897552e-05, "loss": 0.3612, "step": 5649 }, { "epoch": 0.28, "learning_rate": 1.6847545638128697e-05, "loss": 0.4534, "step": 5650 }, { "epoch": 0.28, "learning_rate": 1.6846366968193284e-05, "loss": 0.304, "step": 5651 }, { "epoch": 0.28, "learning_rate": 1.6845188119200107e-05, "loss": 0.4688, "step": 5652 }, { "epoch": 0.28, "learning_rate": 1.684400909118e-05, "loss": 0.3967, "step": 5653 }, { "epoch": 0.28, "learning_rate": 1.6842829884163805e-05, "loss": 0.3095, "step": 5654 }, { "epoch": 0.28, "learning_rate": 1.6841650498182355e-05, "loss": 0.3462, "step": 5655 }, { "epoch": 0.28, "learning_rate": 1.6840470933266495e-05, "loss": 0.336, "step": 5656 }, { "epoch": 0.28, "learning_rate": 1.683929118944708e-05, "loss": 0.2734, "step": 5657 }, { "epoch": 0.28, "learning_rate": 1.683811126675496e-05, "loss": 0.4694, "step": 5658 }, { "epoch": 0.28, "learning_rate": 1.6836931165220996e-05, "loss": 0.3071, "step": 5659 }, { "epoch": 0.28, "learning_rate": 1.6835750884876047e-05, "loss": 0.3831, "step": 5660 }, { "epoch": 0.28, "learning_rate": 1.6834570425750995e-05, "loss": 0.2873, "step": 5661 }, { "epoch": 0.28, "learning_rate": 1.6833389787876702e-05, "loss": 0.3196, "step": 5662 }, { "epoch": 0.28, "learning_rate": 1.683220897128405e-05, "loss": 0.571, "step": 5663 }, { "epoch": 0.28, "learning_rate": 1.683102797600392e-05, "loss": 0.4657, "step": 5664 }, { "epoch": 0.28, "learning_rate": 1.68298468020672e-05, "loss": 0.5369, "step": 5665 }, { "epoch": 0.28, "learning_rate": 1.6828665449504784e-05, "loss": 0.4232, "step": 5666 }, { "epoch": 0.28, "learning_rate": 1.6827483918347566e-05, "loss": 0.2983, "step": 5667 }, { "epoch": 0.28, "learning_rate": 1.682630220862645e-05, "loss": 0.3272, "step": 5668 }, { "epoch": 0.28, "learning_rate": 1.6825120320372337e-05, "loss": 0.5621, "step": 5669 }, { "epoch": 0.28, "learning_rate": 1.6823938253616142e-05, "loss": 0.2576, "step": 5670 }, { "epoch": 0.28, "learning_rate": 1.682275600838878e-05, "loss": 0.2464, "step": 5671 }, { "epoch": 0.28, "learning_rate": 1.6821573584721176e-05, "loss": 0.2738, "step": 5672 }, { "epoch": 0.28, "learning_rate": 1.6820390982644242e-05, "loss": 0.2201, "step": 5673 }, { "epoch": 0.28, "learning_rate": 1.681920820218892e-05, "loss": 0.2824, "step": 5674 }, { "epoch": 0.28, "learning_rate": 1.6818025243386138e-05, "loss": 0.5349, "step": 5675 }, { "epoch": 0.28, "learning_rate": 1.6816842106266835e-05, "loss": 0.4251, "step": 5676 }, { "epoch": 0.28, "learning_rate": 1.6815658790861958e-05, "loss": 0.4138, "step": 5677 }, { "epoch": 0.28, "learning_rate": 1.6814475297202446e-05, "loss": 0.4994, "step": 5678 }, { "epoch": 0.28, "learning_rate": 1.6813291625319264e-05, "loss": 0.3604, "step": 5679 }, { "epoch": 0.28, "learning_rate": 1.681210777524336e-05, "loss": 0.5623, "step": 5680 }, { "epoch": 0.28, "learning_rate": 1.6810923747005703e-05, "loss": 0.3802, "step": 5681 }, { "epoch": 0.28, "learning_rate": 1.6809739540637253e-05, "loss": 0.3797, "step": 5682 }, { "epoch": 0.28, "learning_rate": 1.6808555156168983e-05, "loss": 0.2071, "step": 5683 }, { "epoch": 0.28, "learning_rate": 1.6807370593631876e-05, "loss": 0.4583, "step": 5684 }, { "epoch": 0.28, "learning_rate": 1.68061858530569e-05, "loss": 0.3423, "step": 5685 }, { "epoch": 0.28, "learning_rate": 1.6805000934475054e-05, "loss": 0.5436, "step": 5686 }, { "epoch": 0.28, "learning_rate": 1.6803815837917315e-05, "loss": 0.5301, "step": 5687 }, { "epoch": 0.28, "learning_rate": 1.680263056341469e-05, "loss": 0.3359, "step": 5688 }, { "epoch": 0.28, "learning_rate": 1.680144511099817e-05, "loss": 0.5067, "step": 5689 }, { "epoch": 0.28, "learning_rate": 1.680025948069876e-05, "loss": 0.4662, "step": 5690 }, { "epoch": 0.28, "learning_rate": 1.6799073672547473e-05, "loss": 0.4526, "step": 5691 }, { "epoch": 0.28, "learning_rate": 1.6797887686575317e-05, "loss": 0.468, "step": 5692 }, { "epoch": 0.28, "learning_rate": 1.6796701522813313e-05, "loss": 0.3324, "step": 5693 }, { "epoch": 0.28, "learning_rate": 1.6795515181292484e-05, "loss": 0.3601, "step": 5694 }, { "epoch": 0.28, "learning_rate": 1.6794328662043855e-05, "loss": 0.3414, "step": 5695 }, { "epoch": 0.28, "learning_rate": 1.679314196509846e-05, "loss": 0.2799, "step": 5696 }, { "epoch": 0.28, "learning_rate": 1.6791955090487332e-05, "loss": 0.148, "step": 5697 }, { "epoch": 0.28, "learning_rate": 1.6790768038241516e-05, "loss": 0.5149, "step": 5698 }, { "epoch": 0.28, "learning_rate": 1.6789580808392058e-05, "loss": 0.3785, "step": 5699 }, { "epoch": 0.28, "learning_rate": 1.6788393400970006e-05, "loss": 0.3716, "step": 5700 }, { "epoch": 0.28, "learning_rate": 1.678720581600641e-05, "loss": 0.197, "step": 5701 }, { "epoch": 0.28, "learning_rate": 1.678601805353234e-05, "loss": 0.3304, "step": 5702 }, { "epoch": 0.28, "learning_rate": 1.6784830113578856e-05, "loss": 0.2174, "step": 5703 }, { "epoch": 0.28, "learning_rate": 1.6783641996177026e-05, "loss": 0.5025, "step": 5704 }, { "epoch": 0.28, "learning_rate": 1.678245370135793e-05, "loss": 0.5969, "step": 5705 }, { "epoch": 0.28, "learning_rate": 1.6781265229152633e-05, "loss": 0.4117, "step": 5706 }, { "epoch": 0.28, "learning_rate": 1.678007657959223e-05, "loss": 0.3496, "step": 5707 }, { "epoch": 0.29, "learning_rate": 1.6778887752707803e-05, "loss": 0.3257, "step": 5708 }, { "epoch": 0.29, "learning_rate": 1.6777698748530444e-05, "loss": 0.6106, "step": 5709 }, { "epoch": 0.29, "learning_rate": 1.6776509567091254e-05, "loss": 0.3412, "step": 5710 }, { "epoch": 0.29, "learning_rate": 1.6775320208421333e-05, "loss": 0.4016, "step": 5711 }, { "epoch": 0.29, "learning_rate": 1.6774130672551785e-05, "loss": 0.4974, "step": 5712 }, { "epoch": 0.29, "learning_rate": 1.6772940959513718e-05, "loss": 0.3334, "step": 5713 }, { "epoch": 0.29, "learning_rate": 1.6771751069338256e-05, "loss": 0.1952, "step": 5714 }, { "epoch": 0.29, "learning_rate": 1.6770561002056513e-05, "loss": 0.2788, "step": 5715 }, { "epoch": 0.29, "learning_rate": 1.676937075769961e-05, "loss": 0.2715, "step": 5716 }, { "epoch": 0.29, "learning_rate": 1.676818033629869e-05, "loss": 0.4213, "step": 5717 }, { "epoch": 0.29, "learning_rate": 1.676698973788487e-05, "loss": 0.362, "step": 5718 }, { "epoch": 0.29, "learning_rate": 1.67657989624893e-05, "loss": 0.2862, "step": 5719 }, { "epoch": 0.29, "learning_rate": 1.676460801014312e-05, "loss": 0.3844, "step": 5720 }, { "epoch": 0.29, "learning_rate": 1.6763416880877477e-05, "loss": 0.3096, "step": 5721 }, { "epoch": 0.29, "learning_rate": 1.6762225574723525e-05, "loss": 0.278, "step": 5722 }, { "epoch": 0.29, "learning_rate": 1.676103409171242e-05, "loss": 0.2925, "step": 5723 }, { "epoch": 0.29, "learning_rate": 1.6759842431875323e-05, "loss": 0.2978, "step": 5724 }, { "epoch": 0.29, "learning_rate": 1.67586505952434e-05, "loss": 0.6102, "step": 5725 }, { "epoch": 0.29, "learning_rate": 1.6757458581847825e-05, "loss": 0.5056, "step": 5726 }, { "epoch": 0.29, "learning_rate": 1.675626639171977e-05, "loss": 0.4241, "step": 5727 }, { "epoch": 0.29, "learning_rate": 1.6755074024890416e-05, "loss": 0.4763, "step": 5728 }, { "epoch": 0.29, "learning_rate": 1.675388148139095e-05, "loss": 0.4379, "step": 5729 }, { "epoch": 0.29, "learning_rate": 1.675268876125256e-05, "loss": 0.4714, "step": 5730 }, { "epoch": 0.29, "learning_rate": 1.6751495864506434e-05, "loss": 0.3408, "step": 5731 }, { "epoch": 0.29, "learning_rate": 1.6750302791183782e-05, "loss": 0.7276, "step": 5732 }, { "epoch": 0.29, "learning_rate": 1.6749109541315802e-05, "loss": 0.5023, "step": 5733 }, { "epoch": 0.29, "learning_rate": 1.67479161149337e-05, "loss": 0.3595, "step": 5734 }, { "epoch": 0.29, "learning_rate": 1.6746722512068693e-05, "loss": 0.2194, "step": 5735 }, { "epoch": 0.29, "learning_rate": 1.6745528732751995e-05, "loss": 0.5358, "step": 5736 }, { "epoch": 0.29, "learning_rate": 1.6744334777014827e-05, "loss": 0.5505, "step": 5737 }, { "epoch": 0.29, "learning_rate": 1.6743140644888414e-05, "loss": 0.3899, "step": 5738 }, { "epoch": 0.29, "learning_rate": 1.6741946336403992e-05, "loss": 0.1805, "step": 5739 }, { "epoch": 0.29, "learning_rate": 1.6740751851592793e-05, "loss": 0.3659, "step": 5740 }, { "epoch": 0.29, "learning_rate": 1.673955719048606e-05, "loss": 0.417, "step": 5741 }, { "epoch": 0.29, "learning_rate": 1.6738362353115034e-05, "loss": 0.5275, "step": 5742 }, { "epoch": 0.29, "learning_rate": 1.6737167339510967e-05, "loss": 0.3503, "step": 5743 }, { "epoch": 0.29, "learning_rate": 1.6735972149705112e-05, "loss": 0.2354, "step": 5744 }, { "epoch": 0.29, "learning_rate": 1.673477678372873e-05, "loss": 0.3273, "step": 5745 }, { "epoch": 0.29, "learning_rate": 1.673358124161308e-05, "loss": 0.34, "step": 5746 }, { "epoch": 0.29, "learning_rate": 1.673238552338943e-05, "loss": 0.2658, "step": 5747 }, { "epoch": 0.29, "learning_rate": 1.6731189629089056e-05, "loss": 0.607, "step": 5748 }, { "epoch": 0.29, "learning_rate": 1.6729993558743235e-05, "loss": 0.4635, "step": 5749 }, { "epoch": 0.29, "learning_rate": 1.6728797312383247e-05, "loss": 0.4527, "step": 5750 }, { "epoch": 0.29, "learning_rate": 1.6727600890040377e-05, "loss": 0.4976, "step": 5751 }, { "epoch": 0.29, "learning_rate": 1.672640429174592e-05, "loss": 0.2793, "step": 5752 }, { "epoch": 0.29, "learning_rate": 1.6725207517531164e-05, "loss": 0.3133, "step": 5753 }, { "epoch": 0.29, "learning_rate": 1.6724010567427417e-05, "loss": 0.2128, "step": 5754 }, { "epoch": 0.29, "learning_rate": 1.672281344146598e-05, "loss": 0.5985, "step": 5755 }, { "epoch": 0.29, "learning_rate": 1.6721616139678162e-05, "loss": 0.4767, "step": 5756 }, { "epoch": 0.29, "learning_rate": 1.672041866209528e-05, "loss": 0.5955, "step": 5757 }, { "epoch": 0.29, "learning_rate": 1.6719221008748646e-05, "loss": 0.4052, "step": 5758 }, { "epoch": 0.29, "learning_rate": 1.671802317966959e-05, "loss": 0.384, "step": 5759 }, { "epoch": 0.29, "learning_rate": 1.6716825174889438e-05, "loss": 0.3259, "step": 5760 }, { "epoch": 0.29, "learning_rate": 1.671562699443952e-05, "loss": 0.4481, "step": 5761 }, { "epoch": 0.29, "learning_rate": 1.6714428638351175e-05, "loss": 0.4531, "step": 5762 }, { "epoch": 0.29, "learning_rate": 1.671323010665574e-05, "loss": 0.3965, "step": 5763 }, { "epoch": 0.29, "learning_rate": 1.6712031399384566e-05, "loss": 0.3991, "step": 5764 }, { "epoch": 0.29, "learning_rate": 1.6710832516569003e-05, "loss": 0.2288, "step": 5765 }, { "epoch": 0.29, "learning_rate": 1.6709633458240403e-05, "loss": 0.4095, "step": 5766 }, { "epoch": 0.29, "learning_rate": 1.6708434224430128e-05, "loss": 0.427, "step": 5767 }, { "epoch": 0.29, "learning_rate": 1.6707234815169544e-05, "loss": 0.2246, "step": 5768 }, { "epoch": 0.29, "learning_rate": 1.670603523049002e-05, "loss": 0.4148, "step": 5769 }, { "epoch": 0.29, "learning_rate": 1.670483547042292e-05, "loss": 0.2978, "step": 5770 }, { "epoch": 0.29, "learning_rate": 1.670363553499964e-05, "loss": 0.6138, "step": 5771 }, { "epoch": 0.29, "learning_rate": 1.6702435424251546e-05, "loss": 0.415, "step": 5772 }, { "epoch": 0.29, "learning_rate": 1.6701235138210035e-05, "loss": 0.3506, "step": 5773 }, { "epoch": 0.29, "learning_rate": 1.6700034676906494e-05, "loss": 0.5248, "step": 5774 }, { "epoch": 0.29, "learning_rate": 1.6698834040372324e-05, "loss": 0.4589, "step": 5775 }, { "epoch": 0.29, "learning_rate": 1.6697633228638924e-05, "loss": 0.4624, "step": 5776 }, { "epoch": 0.29, "learning_rate": 1.6696432241737697e-05, "loss": 0.3815, "step": 5777 }, { "epoch": 0.29, "learning_rate": 1.6695231079700055e-05, "loss": 0.3204, "step": 5778 }, { "epoch": 0.29, "learning_rate": 1.6694029742557412e-05, "loss": 0.3028, "step": 5779 }, { "epoch": 0.29, "learning_rate": 1.6692828230341195e-05, "loss": 0.5616, "step": 5780 }, { "epoch": 0.29, "learning_rate": 1.6691626543082813e-05, "loss": 0.142, "step": 5781 }, { "epoch": 0.29, "learning_rate": 1.669042468081371e-05, "loss": 0.4055, "step": 5782 }, { "epoch": 0.29, "learning_rate": 1.6689222643565313e-05, "loss": 0.4326, "step": 5783 }, { "epoch": 0.29, "learning_rate": 1.6688020431369057e-05, "loss": 0.3547, "step": 5784 }, { "epoch": 0.29, "learning_rate": 1.6686818044256384e-05, "loss": 0.3184, "step": 5785 }, { "epoch": 0.29, "learning_rate": 1.6685615482258747e-05, "loss": 0.4638, "step": 5786 }, { "epoch": 0.29, "learning_rate": 1.6684412745407586e-05, "loss": 0.3026, "step": 5787 }, { "epoch": 0.29, "learning_rate": 1.6683209833734372e-05, "loss": 0.5122, "step": 5788 }, { "epoch": 0.29, "learning_rate": 1.668200674727056e-05, "loss": 0.4531, "step": 5789 }, { "epoch": 0.29, "learning_rate": 1.6680803486047606e-05, "loss": 0.3941, "step": 5790 }, { "epoch": 0.29, "learning_rate": 1.6679600050096993e-05, "loss": 0.1351, "step": 5791 }, { "epoch": 0.29, "learning_rate": 1.6678396439450185e-05, "loss": 0.6641, "step": 5792 }, { "epoch": 0.29, "learning_rate": 1.667719265413867e-05, "loss": 0.3779, "step": 5793 }, { "epoch": 0.29, "learning_rate": 1.6675988694193924e-05, "loss": 0.2233, "step": 5794 }, { "epoch": 0.29, "learning_rate": 1.6674784559647436e-05, "loss": 0.4361, "step": 5795 }, { "epoch": 0.29, "learning_rate": 1.6673580250530703e-05, "loss": 0.3733, "step": 5796 }, { "epoch": 0.29, "learning_rate": 1.667237576687522e-05, "loss": 0.316, "step": 5797 }, { "epoch": 0.29, "learning_rate": 1.6671171108712486e-05, "loss": 0.1669, "step": 5798 }, { "epoch": 0.29, "learning_rate": 1.666996627607401e-05, "loss": 0.4302, "step": 5799 }, { "epoch": 0.29, "learning_rate": 1.6668761268991303e-05, "loss": 0.3303, "step": 5800 }, { "epoch": 0.29, "learning_rate": 1.6667556087495877e-05, "loss": 0.3801, "step": 5801 }, { "epoch": 0.29, "learning_rate": 1.6666350731619257e-05, "loss": 0.2535, "step": 5802 }, { "epoch": 0.29, "learning_rate": 1.6665145201392963e-05, "loss": 0.4871, "step": 5803 }, { "epoch": 0.29, "learning_rate": 1.6663939496848525e-05, "loss": 0.4546, "step": 5804 }, { "epoch": 0.29, "learning_rate": 1.666273361801748e-05, "loss": 0.4098, "step": 5805 }, { "epoch": 0.29, "learning_rate": 1.6661527564931365e-05, "loss": 0.3532, "step": 5806 }, { "epoch": 0.29, "learning_rate": 1.666032133762172e-05, "loss": 0.3899, "step": 5807 }, { "epoch": 0.29, "learning_rate": 1.6659114936120094e-05, "loss": 0.5877, "step": 5808 }, { "epoch": 0.29, "learning_rate": 1.665790836045804e-05, "loss": 0.3434, "step": 5809 }, { "epoch": 0.29, "learning_rate": 1.665670161066711e-05, "loss": 0.4454, "step": 5810 }, { "epoch": 0.29, "learning_rate": 1.665549468677887e-05, "loss": 0.4687, "step": 5811 }, { "epoch": 0.29, "learning_rate": 1.6654287588824886e-05, "loss": 0.4121, "step": 5812 }, { "epoch": 0.29, "learning_rate": 1.665308031683672e-05, "loss": 0.4294, "step": 5813 }, { "epoch": 0.29, "learning_rate": 1.6651872870845956e-05, "loss": 0.5461, "step": 5814 }, { "epoch": 0.29, "learning_rate": 1.665066525088417e-05, "loss": 0.3441, "step": 5815 }, { "epoch": 0.29, "learning_rate": 1.664945745698294e-05, "loss": 0.346, "step": 5816 }, { "epoch": 0.29, "learning_rate": 1.6648249489173868e-05, "loss": 0.5468, "step": 5817 }, { "epoch": 0.29, "learning_rate": 1.664704134748853e-05, "loss": 0.3656, "step": 5818 }, { "epoch": 0.29, "learning_rate": 1.664583303195854e-05, "loss": 0.573, "step": 5819 }, { "epoch": 0.29, "learning_rate": 1.6644624542615485e-05, "loss": 0.415, "step": 5820 }, { "epoch": 0.29, "learning_rate": 1.6643415879490977e-05, "loss": 0.2646, "step": 5821 }, { "epoch": 0.29, "learning_rate": 1.6642207042616634e-05, "loss": 0.2703, "step": 5822 }, { "epoch": 0.29, "learning_rate": 1.6640998032024063e-05, "loss": 0.6512, "step": 5823 }, { "epoch": 0.29, "learning_rate": 1.6639788847744885e-05, "loss": 0.3056, "step": 5824 }, { "epoch": 0.29, "learning_rate": 1.6638579489810726e-05, "loss": 0.329, "step": 5825 }, { "epoch": 0.29, "learning_rate": 1.6637369958253218e-05, "loss": 0.3187, "step": 5826 }, { "epoch": 0.29, "learning_rate": 1.663616025310399e-05, "loss": 0.3643, "step": 5827 }, { "epoch": 0.29, "learning_rate": 1.6634950374394684e-05, "loss": 0.3565, "step": 5828 }, { "epoch": 0.29, "learning_rate": 1.663374032215694e-05, "loss": 0.3056, "step": 5829 }, { "epoch": 0.29, "learning_rate": 1.6632530096422405e-05, "loss": 0.5566, "step": 5830 }, { "epoch": 0.29, "learning_rate": 1.6631319697222736e-05, "loss": 0.3821, "step": 5831 }, { "epoch": 0.29, "learning_rate": 1.6630109124589585e-05, "loss": 0.4116, "step": 5832 }, { "epoch": 0.29, "learning_rate": 1.6628898378554613e-05, "loss": 0.3145, "step": 5833 }, { "epoch": 0.29, "learning_rate": 1.6627687459149484e-05, "loss": 0.4168, "step": 5834 }, { "epoch": 0.29, "learning_rate": 1.6626476366405872e-05, "loss": 0.4164, "step": 5835 }, { "epoch": 0.29, "learning_rate": 1.6625265100355453e-05, "loss": 0.4436, "step": 5836 }, { "epoch": 0.29, "learning_rate": 1.66240536610299e-05, "loss": 0.3663, "step": 5837 }, { "epoch": 0.29, "learning_rate": 1.6622842048460897e-05, "loss": 0.5173, "step": 5838 }, { "epoch": 0.29, "learning_rate": 1.662163026268014e-05, "loss": 0.316, "step": 5839 }, { "epoch": 0.29, "learning_rate": 1.662041830371931e-05, "loss": 0.5368, "step": 5840 }, { "epoch": 0.29, "learning_rate": 1.6619206171610113e-05, "loss": 0.4497, "step": 5841 }, { "epoch": 0.29, "learning_rate": 1.661799386638425e-05, "loss": 0.4172, "step": 5842 }, { "epoch": 0.29, "learning_rate": 1.6616781388073425e-05, "loss": 0.3111, "step": 5843 }, { "epoch": 0.29, "learning_rate": 1.661556873670935e-05, "loss": 0.1557, "step": 5844 }, { "epoch": 0.29, "learning_rate": 1.6614355912323732e-05, "loss": 0.5363, "step": 5845 }, { "epoch": 0.29, "learning_rate": 1.6613142914948307e-05, "loss": 0.0505, "step": 5846 }, { "epoch": 0.29, "learning_rate": 1.6611929744614787e-05, "loss": 0.4321, "step": 5847 }, { "epoch": 0.29, "learning_rate": 1.6610716401354907e-05, "loss": 0.6119, "step": 5848 }, { "epoch": 0.29, "learning_rate": 1.6609502885200397e-05, "loss": 0.3632, "step": 5849 }, { "epoch": 0.29, "learning_rate": 1.6608289196182993e-05, "loss": 0.4633, "step": 5850 }, { "epoch": 0.29, "learning_rate": 1.6607075334334445e-05, "loss": 0.3808, "step": 5851 }, { "epoch": 0.29, "learning_rate": 1.660586129968649e-05, "loss": 0.4437, "step": 5852 }, { "epoch": 0.29, "learning_rate": 1.6604647092270893e-05, "loss": 0.2225, "step": 5853 }, { "epoch": 0.29, "learning_rate": 1.6603432712119395e-05, "loss": 0.273, "step": 5854 }, { "epoch": 0.29, "learning_rate": 1.6602218159263764e-05, "loss": 0.3692, "step": 5855 }, { "epoch": 0.29, "learning_rate": 1.6601003433735768e-05, "loss": 0.4309, "step": 5856 }, { "epoch": 0.29, "learning_rate": 1.659978853556717e-05, "loss": 0.3998, "step": 5857 }, { "epoch": 0.29, "learning_rate": 1.6598573464789746e-05, "loss": 0.6189, "step": 5858 }, { "epoch": 0.29, "learning_rate": 1.6597358221435276e-05, "loss": 0.4242, "step": 5859 }, { "epoch": 0.29, "learning_rate": 1.6596142805535546e-05, "loss": 0.2904, "step": 5860 }, { "epoch": 0.29, "learning_rate": 1.6594927217122337e-05, "loss": 0.3117, "step": 5861 }, { "epoch": 0.29, "learning_rate": 1.659371145622745e-05, "loss": 0.3164, "step": 5862 }, { "epoch": 0.29, "learning_rate": 1.6592495522882668e-05, "loss": 0.2247, "step": 5863 }, { "epoch": 0.29, "learning_rate": 1.6591279417119807e-05, "loss": 0.646, "step": 5864 }, { "epoch": 0.29, "learning_rate": 1.6590063138970664e-05, "loss": 0.1961, "step": 5865 }, { "epoch": 0.29, "learning_rate": 1.658884668846705e-05, "loss": 0.2977, "step": 5866 }, { "epoch": 0.29, "learning_rate": 1.658763006564078e-05, "loss": 0.2703, "step": 5867 }, { "epoch": 0.29, "learning_rate": 1.658641327052367e-05, "loss": 0.557, "step": 5868 }, { "epoch": 0.29, "learning_rate": 1.6585196303147553e-05, "loss": 0.4791, "step": 5869 }, { "epoch": 0.29, "learning_rate": 1.6583979163544252e-05, "loss": 0.3076, "step": 5870 }, { "epoch": 0.29, "learning_rate": 1.6582761851745598e-05, "loss": 0.1236, "step": 5871 }, { "epoch": 0.29, "learning_rate": 1.658154436778343e-05, "loss": 0.4918, "step": 5872 }, { "epoch": 0.29, "learning_rate": 1.658032671168959e-05, "loss": 0.5124, "step": 5873 }, { "epoch": 0.29, "learning_rate": 1.6579108883495924e-05, "loss": 0.2627, "step": 5874 }, { "epoch": 0.29, "learning_rate": 1.657789088323428e-05, "loss": 0.4204, "step": 5875 }, { "epoch": 0.29, "learning_rate": 1.6576672710936518e-05, "loss": 0.5206, "step": 5876 }, { "epoch": 0.29, "learning_rate": 1.657545436663449e-05, "loss": 0.4567, "step": 5877 }, { "epoch": 0.29, "learning_rate": 1.657423585036007e-05, "loss": 0.3848, "step": 5878 }, { "epoch": 0.29, "learning_rate": 1.657301716214512e-05, "loss": 0.1711, "step": 5879 }, { "epoch": 0.29, "learning_rate": 1.6571798302021517e-05, "loss": 0.3183, "step": 5880 }, { "epoch": 0.29, "learning_rate": 1.6570579270021137e-05, "loss": 0.3454, "step": 5881 }, { "epoch": 0.29, "learning_rate": 1.656936006617586e-05, "loss": 0.4733, "step": 5882 }, { "epoch": 0.29, "learning_rate": 1.6568140690517577e-05, "loss": 0.4561, "step": 5883 }, { "epoch": 0.29, "learning_rate": 1.6566921143078176e-05, "loss": 0.2705, "step": 5884 }, { "epoch": 0.29, "learning_rate": 1.6565701423889553e-05, "loss": 0.4539, "step": 5885 }, { "epoch": 0.29, "learning_rate": 1.656448153298361e-05, "loss": 0.4506, "step": 5886 }, { "epoch": 0.29, "learning_rate": 1.656326147039225e-05, "loss": 0.3211, "step": 5887 }, { "epoch": 0.29, "learning_rate": 1.656204123614738e-05, "loss": 0.445, "step": 5888 }, { "epoch": 0.29, "learning_rate": 1.656082083028092e-05, "loss": 0.6052, "step": 5889 }, { "epoch": 0.29, "learning_rate": 1.6559600252824783e-05, "loss": 0.3144, "step": 5890 }, { "epoch": 0.29, "learning_rate": 1.655837950381089e-05, "loss": 0.3085, "step": 5891 }, { "epoch": 0.29, "learning_rate": 1.6557158583271177e-05, "loss": 0.4218, "step": 5892 }, { "epoch": 0.29, "learning_rate": 1.6555937491237567e-05, "loss": 0.5083, "step": 5893 }, { "epoch": 0.29, "learning_rate": 1.6554716227741996e-05, "loss": 0.2209, "step": 5894 }, { "epoch": 0.29, "learning_rate": 1.6553494792816408e-05, "loss": 0.448, "step": 5895 }, { "epoch": 0.29, "learning_rate": 1.6552273186492752e-05, "loss": 0.5494, "step": 5896 }, { "epoch": 0.29, "learning_rate": 1.6551051408802967e-05, "loss": 0.0982, "step": 5897 }, { "epoch": 0.29, "learning_rate": 1.6549829459779016e-05, "loss": 0.2953, "step": 5898 }, { "epoch": 0.29, "learning_rate": 1.6548607339452853e-05, "loss": 0.6134, "step": 5899 }, { "epoch": 0.29, "learning_rate": 1.6547385047856442e-05, "loss": 0.3125, "step": 5900 }, { "epoch": 0.29, "learning_rate": 1.6546162585021754e-05, "loss": 0.4433, "step": 5901 }, { "epoch": 0.29, "learning_rate": 1.654493995098075e-05, "loss": 0.3586, "step": 5902 }, { "epoch": 0.29, "learning_rate": 1.6543717145765417e-05, "loss": 0.2949, "step": 5903 }, { "epoch": 0.29, "learning_rate": 1.6542494169407738e-05, "loss": 0.3337, "step": 5904 }, { "epoch": 0.29, "learning_rate": 1.654127102193969e-05, "loss": 0.3447, "step": 5905 }, { "epoch": 0.29, "learning_rate": 1.6540047703393262e-05, "loss": 0.1343, "step": 5906 }, { "epoch": 0.29, "learning_rate": 1.6538824213800454e-05, "loss": 0.465, "step": 5907 }, { "epoch": 0.3, "learning_rate": 1.6537600553193266e-05, "loss": 0.2418, "step": 5908 }, { "epoch": 0.3, "learning_rate": 1.6536376721603696e-05, "loss": 0.3983, "step": 5909 }, { "epoch": 0.3, "learning_rate": 1.6535152719063758e-05, "loss": 0.4489, "step": 5910 }, { "epoch": 0.3, "learning_rate": 1.6533928545605452e-05, "loss": 0.4147, "step": 5911 }, { "epoch": 0.3, "learning_rate": 1.653270420126081e-05, "loss": 0.4787, "step": 5912 }, { "epoch": 0.3, "learning_rate": 1.6531479686061846e-05, "loss": 0.389, "step": 5913 }, { "epoch": 0.3, "learning_rate": 1.6530255000040583e-05, "loss": 0.2671, "step": 5914 }, { "epoch": 0.3, "learning_rate": 1.6529030143229052e-05, "loss": 0.3577, "step": 5915 }, { "epoch": 0.3, "learning_rate": 1.652780511565929e-05, "loss": 0.5772, "step": 5916 }, { "epoch": 0.3, "learning_rate": 1.6526579917363338e-05, "loss": 0.2698, "step": 5917 }, { "epoch": 0.3, "learning_rate": 1.6525354548373233e-05, "loss": 0.4061, "step": 5918 }, { "epoch": 0.3, "learning_rate": 1.6524129008721027e-05, "loss": 0.3207, "step": 5919 }, { "epoch": 0.3, "learning_rate": 1.6522903298438773e-05, "loss": 0.3907, "step": 5920 }, { "epoch": 0.3, "learning_rate": 1.6521677417558527e-05, "loss": 0.2361, "step": 5921 }, { "epoch": 0.3, "learning_rate": 1.652045136611235e-05, "loss": 0.4446, "step": 5922 }, { "epoch": 0.3, "learning_rate": 1.6519225144132306e-05, "loss": 0.3916, "step": 5923 }, { "epoch": 0.3, "learning_rate": 1.651799875165047e-05, "loss": 0.3642, "step": 5924 }, { "epoch": 0.3, "learning_rate": 1.651677218869891e-05, "loss": 0.4711, "step": 5925 }, { "epoch": 0.3, "learning_rate": 1.651554545530971e-05, "loss": 0.4071, "step": 5926 }, { "epoch": 0.3, "learning_rate": 1.651431855151495e-05, "loss": 0.2399, "step": 5927 }, { "epoch": 0.3, "learning_rate": 1.6513091477346727e-05, "loss": 0.448, "step": 5928 }, { "epoch": 0.3, "learning_rate": 1.651186423283712e-05, "loss": 0.3717, "step": 5929 }, { "epoch": 0.3, "learning_rate": 1.6510636818018243e-05, "loss": 0.2957, "step": 5930 }, { "epoch": 0.3, "learning_rate": 1.6509409232922178e-05, "loss": 0.3133, "step": 5931 }, { "epoch": 0.3, "learning_rate": 1.6508181477581045e-05, "loss": 0.165, "step": 5932 }, { "epoch": 0.3, "learning_rate": 1.6506953552026948e-05, "loss": 0.3901, "step": 5933 }, { "epoch": 0.3, "learning_rate": 1.6505725456292002e-05, "loss": 0.4468, "step": 5934 }, { "epoch": 0.3, "learning_rate": 1.6504497190408327e-05, "loss": 0.3335, "step": 5935 }, { "epoch": 0.3, "learning_rate": 1.6503268754408052e-05, "loss": 0.428, "step": 5936 }, { "epoch": 0.3, "learning_rate": 1.6502040148323297e-05, "loss": 0.3605, "step": 5937 }, { "epoch": 0.3, "learning_rate": 1.6500811372186202e-05, "loss": 0.2235, "step": 5938 }, { "epoch": 0.3, "learning_rate": 1.6499582426028896e-05, "loss": 0.448, "step": 5939 }, { "epoch": 0.3, "learning_rate": 1.6498353309883527e-05, "loss": 0.3293, "step": 5940 }, { "epoch": 0.3, "learning_rate": 1.6497124023782233e-05, "loss": 0.5295, "step": 5941 }, { "epoch": 0.3, "learning_rate": 1.6495894567757175e-05, "loss": 0.3689, "step": 5942 }, { "epoch": 0.3, "learning_rate": 1.6494664941840503e-05, "loss": 0.3548, "step": 5943 }, { "epoch": 0.3, "learning_rate": 1.6493435146064377e-05, "loss": 0.4376, "step": 5944 }, { "epoch": 0.3, "learning_rate": 1.6492205180460954e-05, "loss": 0.4017, "step": 5945 }, { "epoch": 0.3, "learning_rate": 1.649097504506241e-05, "loss": 0.2985, "step": 5946 }, { "epoch": 0.3, "learning_rate": 1.648974473990092e-05, "loss": 0.4052, "step": 5947 }, { "epoch": 0.3, "learning_rate": 1.6488514265008652e-05, "loss": 0.2792, "step": 5948 }, { "epoch": 0.3, "learning_rate": 1.648728362041779e-05, "loss": 0.4193, "step": 5949 }, { "epoch": 0.3, "learning_rate": 1.6486052806160528e-05, "loss": 0.3244, "step": 5950 }, { "epoch": 0.3, "learning_rate": 1.6484821822269044e-05, "loss": 0.2479, "step": 5951 }, { "epoch": 0.3, "learning_rate": 1.648359066877554e-05, "loss": 0.4075, "step": 5952 }, { "epoch": 0.3, "learning_rate": 1.6482359345712216e-05, "loss": 0.2004, "step": 5953 }, { "epoch": 0.3, "learning_rate": 1.6481127853111274e-05, "loss": 0.508, "step": 5954 }, { "epoch": 0.3, "learning_rate": 1.647989619100492e-05, "loss": 0.3199, "step": 5955 }, { "epoch": 0.3, "learning_rate": 1.647866435942537e-05, "loss": 0.5378, "step": 5956 }, { "epoch": 0.3, "learning_rate": 1.6477432358404838e-05, "loss": 0.2949, "step": 5957 }, { "epoch": 0.3, "learning_rate": 1.6476200187975547e-05, "loss": 0.134, "step": 5958 }, { "epoch": 0.3, "learning_rate": 1.647496784816972e-05, "loss": 0.5372, "step": 5959 }, { "epoch": 0.3, "learning_rate": 1.647373533901959e-05, "loss": 0.5327, "step": 5960 }, { "epoch": 0.3, "learning_rate": 1.6472502660557394e-05, "loss": 0.3752, "step": 5961 }, { "epoch": 0.3, "learning_rate": 1.647126981281537e-05, "loss": 0.4584, "step": 5962 }, { "epoch": 0.3, "learning_rate": 1.6470036795825755e-05, "loss": 0.2785, "step": 5963 }, { "epoch": 0.3, "learning_rate": 1.6468803609620805e-05, "loss": 0.2112, "step": 5964 }, { "epoch": 0.3, "learning_rate": 1.646757025423277e-05, "loss": 0.4383, "step": 5965 }, { "epoch": 0.3, "learning_rate": 1.64663367296939e-05, "loss": 0.4389, "step": 5966 }, { "epoch": 0.3, "learning_rate": 1.646510303603647e-05, "loss": 0.3533, "step": 5967 }, { "epoch": 0.3, "learning_rate": 1.646386917329273e-05, "loss": 0.3748, "step": 5968 }, { "epoch": 0.3, "learning_rate": 1.6462635141494966e-05, "loss": 0.4787, "step": 5969 }, { "epoch": 0.3, "learning_rate": 1.646140094067544e-05, "loss": 0.2453, "step": 5970 }, { "epoch": 0.3, "learning_rate": 1.6460166570866436e-05, "loss": 0.3006, "step": 5971 }, { "epoch": 0.3, "learning_rate": 1.6458932032100238e-05, "loss": 0.2023, "step": 5972 }, { "epoch": 0.3, "learning_rate": 1.645769732440913e-05, "loss": 0.4879, "step": 5973 }, { "epoch": 0.3, "learning_rate": 1.6456462447825408e-05, "loss": 0.5715, "step": 5974 }, { "epoch": 0.3, "learning_rate": 1.6455227402381364e-05, "loss": 0.5973, "step": 5975 }, { "epoch": 0.3, "learning_rate": 1.6453992188109307e-05, "loss": 0.314, "step": 5976 }, { "epoch": 0.3, "learning_rate": 1.6452756805041535e-05, "loss": 0.3735, "step": 5977 }, { "epoch": 0.3, "learning_rate": 1.645152125321036e-05, "loss": 0.6766, "step": 5978 }, { "epoch": 0.3, "learning_rate": 1.6450285532648097e-05, "loss": 0.5762, "step": 5979 }, { "epoch": 0.3, "learning_rate": 1.644904964338707e-05, "loss": 0.2495, "step": 5980 }, { "epoch": 0.3, "learning_rate": 1.6447813585459586e-05, "loss": 0.3687, "step": 5981 }, { "epoch": 0.3, "learning_rate": 1.6446577358897987e-05, "loss": 0.4752, "step": 5982 }, { "epoch": 0.3, "learning_rate": 1.64453409637346e-05, "loss": 0.3562, "step": 5983 }, { "epoch": 0.3, "learning_rate": 1.6444104400001767e-05, "loss": 0.4728, "step": 5984 }, { "epoch": 0.3, "learning_rate": 1.6442867667731818e-05, "loss": 0.671, "step": 5985 }, { "epoch": 0.3, "learning_rate": 1.6441630766957106e-05, "loss": 0.2803, "step": 5986 }, { "epoch": 0.3, "learning_rate": 1.6440393697709977e-05, "loss": 0.3008, "step": 5987 }, { "epoch": 0.3, "learning_rate": 1.643915646002279e-05, "loss": 0.3554, "step": 5988 }, { "epoch": 0.3, "learning_rate": 1.6437919053927897e-05, "loss": 0.2852, "step": 5989 }, { "epoch": 0.3, "learning_rate": 1.6436681479457663e-05, "loss": 0.0735, "step": 5990 }, { "epoch": 0.3, "learning_rate": 1.6435443736644455e-05, "loss": 0.4813, "step": 5991 }, { "epoch": 0.3, "learning_rate": 1.6434205825520644e-05, "loss": 0.5613, "step": 5992 }, { "epoch": 0.3, "learning_rate": 1.6432967746118607e-05, "loss": 0.1941, "step": 5993 }, { "epoch": 0.3, "learning_rate": 1.6431729498470727e-05, "loss": 0.4016, "step": 5994 }, { "epoch": 0.3, "learning_rate": 1.6430491082609385e-05, "loss": 0.4351, "step": 5995 }, { "epoch": 0.3, "learning_rate": 1.642925249856697e-05, "loss": 0.4572, "step": 5996 }, { "epoch": 0.3, "learning_rate": 1.642801374637588e-05, "loss": 0.577, "step": 5997 }, { "epoch": 0.3, "learning_rate": 1.6426774826068505e-05, "loss": 0.3134, "step": 5998 }, { "epoch": 0.3, "learning_rate": 1.6425535737677256e-05, "loss": 0.2059, "step": 5999 }, { "epoch": 0.3, "learning_rate": 1.6424296481234532e-05, "loss": 0.375, "step": 6000 }, { "epoch": 0.3, "learning_rate": 1.6423057056772754e-05, "loss": 0.4347, "step": 6001 }, { "epoch": 0.3, "learning_rate": 1.6421817464324325e-05, "loss": 0.3716, "step": 6002 }, { "epoch": 0.3, "learning_rate": 1.6420577703921673e-05, "loss": 0.6651, "step": 6003 }, { "epoch": 0.3, "learning_rate": 1.6419337775597223e-05, "loss": 0.4749, "step": 6004 }, { "epoch": 0.3, "learning_rate": 1.64180976793834e-05, "loss": 0.2387, "step": 6005 }, { "epoch": 0.3, "learning_rate": 1.6416857415312635e-05, "loss": 0.471, "step": 6006 }, { "epoch": 0.3, "learning_rate": 1.6415616983417374e-05, "loss": 0.375, "step": 6007 }, { "epoch": 0.3, "learning_rate": 1.641437638373005e-05, "loss": 0.7374, "step": 6008 }, { "epoch": 0.3, "learning_rate": 1.6413135616283113e-05, "loss": 0.3639, "step": 6009 }, { "epoch": 0.3, "learning_rate": 1.6411894681109017e-05, "loss": 0.3781, "step": 6010 }, { "epoch": 0.3, "learning_rate": 1.641065357824021e-05, "loss": 0.4422, "step": 6011 }, { "epoch": 0.3, "learning_rate": 1.640941230770916e-05, "loss": 0.4163, "step": 6012 }, { "epoch": 0.3, "learning_rate": 1.640817086954832e-05, "loss": 0.1762, "step": 6013 }, { "epoch": 0.3, "learning_rate": 1.640692926379017e-05, "loss": 0.3965, "step": 6014 }, { "epoch": 0.3, "learning_rate": 1.6405687490467173e-05, "loss": 0.3952, "step": 6015 }, { "epoch": 0.3, "learning_rate": 1.6404445549611807e-05, "loss": 0.3727, "step": 6016 }, { "epoch": 0.3, "learning_rate": 1.640320344125656e-05, "loss": 0.3102, "step": 6017 }, { "epoch": 0.3, "learning_rate": 1.6401961165433917e-05, "loss": 0.5019, "step": 6018 }, { "epoch": 0.3, "learning_rate": 1.640071872217636e-05, "loss": 0.3412, "step": 6019 }, { "epoch": 0.3, "learning_rate": 1.6399476111516386e-05, "loss": 0.483, "step": 6020 }, { "epoch": 0.3, "learning_rate": 1.63982333334865e-05, "loss": 0.9274, "step": 6021 }, { "epoch": 0.3, "learning_rate": 1.63969903881192e-05, "loss": 0.5033, "step": 6022 }, { "epoch": 0.3, "learning_rate": 1.6395747275446995e-05, "loss": 0.4035, "step": 6023 }, { "epoch": 0.3, "learning_rate": 1.63945039955024e-05, "loss": 0.247, "step": 6024 }, { "epoch": 0.3, "learning_rate": 1.6393260548317924e-05, "loss": 0.4099, "step": 6025 }, { "epoch": 0.3, "learning_rate": 1.6392016933926092e-05, "loss": 0.1083, "step": 6026 }, { "epoch": 0.3, "learning_rate": 1.6390773152359433e-05, "loss": 0.4508, "step": 6027 }, { "epoch": 0.3, "learning_rate": 1.6389529203650467e-05, "loss": 0.3129, "step": 6028 }, { "epoch": 0.3, "learning_rate": 1.638828508783174e-05, "loss": 0.3524, "step": 6029 }, { "epoch": 0.3, "learning_rate": 1.638704080493578e-05, "loss": 0.4155, "step": 6030 }, { "epoch": 0.3, "learning_rate": 1.638579635499513e-05, "loss": 0.1778, "step": 6031 }, { "epoch": 0.3, "learning_rate": 1.6384551738042346e-05, "loss": 0.2983, "step": 6032 }, { "epoch": 0.3, "learning_rate": 1.6383306954109973e-05, "loss": 0.2178, "step": 6033 }, { "epoch": 0.3, "learning_rate": 1.6382062003230563e-05, "loss": 0.1636, "step": 6034 }, { "epoch": 0.3, "learning_rate": 1.6380816885436687e-05, "loss": 0.2445, "step": 6035 }, { "epoch": 0.3, "learning_rate": 1.63795716007609e-05, "loss": 0.5397, "step": 6036 }, { "epoch": 0.3, "learning_rate": 1.6378326149235767e-05, "loss": 0.4864, "step": 6037 }, { "epoch": 0.3, "learning_rate": 1.6377080530893876e-05, "loss": 0.4651, "step": 6038 }, { "epoch": 0.3, "learning_rate": 1.637583474576779e-05, "loss": 0.3659, "step": 6039 }, { "epoch": 0.3, "learning_rate": 1.63745887938901e-05, "loss": 0.1018, "step": 6040 }, { "epoch": 0.3, "learning_rate": 1.637334267529339e-05, "loss": 0.423, "step": 6041 }, { "epoch": 0.3, "learning_rate": 1.637209639001025e-05, "loss": 0.3062, "step": 6042 }, { "epoch": 0.3, "learning_rate": 1.6370849938073275e-05, "loss": 0.3702, "step": 6043 }, { "epoch": 0.3, "learning_rate": 1.6369603319515063e-05, "loss": 0.324, "step": 6044 }, { "epoch": 0.3, "learning_rate": 1.636835653436822e-05, "loss": 0.351, "step": 6045 }, { "epoch": 0.3, "learning_rate": 1.636710958266535e-05, "loss": 0.2194, "step": 6046 }, { "epoch": 0.3, "learning_rate": 1.6365862464439073e-05, "loss": 0.5217, "step": 6047 }, { "epoch": 0.3, "learning_rate": 1.6364615179721997e-05, "loss": 0.2511, "step": 6048 }, { "epoch": 0.3, "learning_rate": 1.636336772854675e-05, "loss": 0.3997, "step": 6049 }, { "epoch": 0.3, "learning_rate": 1.6362120110945953e-05, "loss": 0.3104, "step": 6050 }, { "epoch": 0.3, "learning_rate": 1.636087232695224e-05, "loss": 0.4157, "step": 6051 }, { "epoch": 0.3, "learning_rate": 1.635962437659824e-05, "loss": 0.4683, "step": 6052 }, { "epoch": 0.3, "learning_rate": 1.635837625991659e-05, "loss": 0.363, "step": 6053 }, { "epoch": 0.3, "learning_rate": 1.6357127976939944e-05, "loss": 0.4671, "step": 6054 }, { "epoch": 0.3, "learning_rate": 1.6355879527700936e-05, "loss": 0.3919, "step": 6055 }, { "epoch": 0.3, "learning_rate": 1.635463091223223e-05, "loss": 0.338, "step": 6056 }, { "epoch": 0.3, "learning_rate": 1.6353382130566472e-05, "loss": 0.2949, "step": 6057 }, { "epoch": 0.3, "learning_rate": 1.6352133182736326e-05, "loss": 0.5063, "step": 6058 }, { "epoch": 0.3, "learning_rate": 1.6350884068774456e-05, "loss": 0.5479, "step": 6059 }, { "epoch": 0.3, "learning_rate": 1.6349634788713533e-05, "loss": 0.6918, "step": 6060 }, { "epoch": 0.3, "learning_rate": 1.6348385342586226e-05, "loss": 0.1961, "step": 6061 }, { "epoch": 0.3, "learning_rate": 1.6347135730425215e-05, "loss": 0.5558, "step": 6062 }, { "epoch": 0.3, "learning_rate": 1.634588595226319e-05, "loss": 0.3096, "step": 6063 }, { "epoch": 0.3, "learning_rate": 1.6344636008132822e-05, "loss": 0.5327, "step": 6064 }, { "epoch": 0.3, "learning_rate": 1.634338589806681e-05, "loss": 0.3897, "step": 6065 }, { "epoch": 0.3, "learning_rate": 1.634213562209785e-05, "loss": 0.3605, "step": 6066 }, { "epoch": 0.3, "learning_rate": 1.6340885180258637e-05, "loss": 0.3275, "step": 6067 }, { "epoch": 0.3, "learning_rate": 1.633963457258188e-05, "loss": 0.3813, "step": 6068 }, { "epoch": 0.3, "learning_rate": 1.633838379910028e-05, "loss": 0.3615, "step": 6069 }, { "epoch": 0.3, "learning_rate": 1.6337132859846557e-05, "loss": 0.3015, "step": 6070 }, { "epoch": 0.3, "learning_rate": 1.6335881754853423e-05, "loss": 0.333, "step": 6071 }, { "epoch": 0.3, "learning_rate": 1.63346304841536e-05, "loss": 0.4011, "step": 6072 }, { "epoch": 0.3, "learning_rate": 1.6333379047779815e-05, "loss": 0.2616, "step": 6073 }, { "epoch": 0.3, "learning_rate": 1.6332127445764796e-05, "loss": 0.3608, "step": 6074 }, { "epoch": 0.3, "learning_rate": 1.6330875678141277e-05, "loss": 0.1141, "step": 6075 }, { "epoch": 0.3, "learning_rate": 1.6329623744941997e-05, "loss": 0.398, "step": 6076 }, { "epoch": 0.3, "learning_rate": 1.6328371646199696e-05, "loss": 0.5433, "step": 6077 }, { "epoch": 0.3, "learning_rate": 1.6327119381947126e-05, "loss": 0.2891, "step": 6078 }, { "epoch": 0.3, "learning_rate": 1.6325866952217037e-05, "loss": 0.3341, "step": 6079 }, { "epoch": 0.3, "learning_rate": 1.632461435704218e-05, "loss": 0.3451, "step": 6080 }, { "epoch": 0.3, "learning_rate": 1.632336159645532e-05, "loss": 0.3414, "step": 6081 }, { "epoch": 0.3, "learning_rate": 1.6322108670489217e-05, "loss": 0.4393, "step": 6082 }, { "epoch": 0.3, "learning_rate": 1.6320855579176645e-05, "loss": 0.3471, "step": 6083 }, { "epoch": 0.3, "learning_rate": 1.6319602322550374e-05, "loss": 0.2891, "step": 6084 }, { "epoch": 0.3, "learning_rate": 1.631834890064318e-05, "loss": 0.5407, "step": 6085 }, { "epoch": 0.3, "learning_rate": 1.6317095313487846e-05, "loss": 0.3967, "step": 6086 }, { "epoch": 0.3, "learning_rate": 1.6315841561117158e-05, "loss": 0.3702, "step": 6087 }, { "epoch": 0.3, "learning_rate": 1.6314587643563906e-05, "loss": 0.385, "step": 6088 }, { "epoch": 0.3, "learning_rate": 1.6313333560860886e-05, "loss": 0.4197, "step": 6089 }, { "epoch": 0.3, "learning_rate": 1.6312079313040895e-05, "loss": 0.375, "step": 6090 }, { "epoch": 0.3, "learning_rate": 1.6310824900136738e-05, "loss": 0.4962, "step": 6091 }, { "epoch": 0.3, "learning_rate": 1.630957032218122e-05, "loss": 0.5062, "step": 6092 }, { "epoch": 0.3, "learning_rate": 1.6308315579207156e-05, "loss": 0.4075, "step": 6093 }, { "epoch": 0.3, "learning_rate": 1.6307060671247355e-05, "loss": 0.3085, "step": 6094 }, { "epoch": 0.3, "learning_rate": 1.630580559833465e-05, "loss": 0.2919, "step": 6095 }, { "epoch": 0.3, "learning_rate": 1.6304550360501857e-05, "loss": 0.2749, "step": 6096 }, { "epoch": 0.3, "learning_rate": 1.6303294957781804e-05, "loss": 0.6058, "step": 6097 }, { "epoch": 0.3, "learning_rate": 1.630203939020733e-05, "loss": 0.3546, "step": 6098 }, { "epoch": 0.3, "learning_rate": 1.6300783657811267e-05, "loss": 0.5803, "step": 6099 }, { "epoch": 0.3, "learning_rate": 1.6299527760626465e-05, "loss": 0.454, "step": 6100 }, { "epoch": 0.3, "learning_rate": 1.629827169868576e-05, "loss": 0.2506, "step": 6101 }, { "epoch": 0.3, "learning_rate": 1.629701547202201e-05, "loss": 0.5759, "step": 6102 }, { "epoch": 0.3, "learning_rate": 1.629575908066807e-05, "loss": 0.2781, "step": 6103 }, { "epoch": 0.3, "learning_rate": 1.6294502524656795e-05, "loss": 0.2774, "step": 6104 }, { "epoch": 0.3, "learning_rate": 1.629324580402105e-05, "loss": 0.3953, "step": 6105 }, { "epoch": 0.3, "learning_rate": 1.6291988918793704e-05, "loss": 0.2776, "step": 6106 }, { "epoch": 0.3, "learning_rate": 1.629073186900763e-05, "loss": 0.269, "step": 6107 }, { "epoch": 0.3, "learning_rate": 1.62894746546957e-05, "loss": 0.4488, "step": 6108 }, { "epoch": 0.31, "learning_rate": 1.6288217275890805e-05, "loss": 0.5204, "step": 6109 }, { "epoch": 0.31, "learning_rate": 1.628695973262582e-05, "loss": 0.225, "step": 6110 }, { "epoch": 0.31, "learning_rate": 1.6285702024933636e-05, "loss": 0.2234, "step": 6111 }, { "epoch": 0.31, "learning_rate": 1.6284444152847146e-05, "loss": 0.1876, "step": 6112 }, { "epoch": 0.31, "learning_rate": 1.6283186116399256e-05, "loss": 0.2524, "step": 6113 }, { "epoch": 0.31, "learning_rate": 1.6281927915622858e-05, "loss": 0.3271, "step": 6114 }, { "epoch": 0.31, "learning_rate": 1.6280669550550867e-05, "loss": 0.3109, "step": 6115 }, { "epoch": 0.31, "learning_rate": 1.627941102121619e-05, "loss": 0.3341, "step": 6116 }, { "epoch": 0.31, "learning_rate": 1.627815232765174e-05, "loss": 0.4945, "step": 6117 }, { "epoch": 0.31, "learning_rate": 1.627689346989044e-05, "loss": 0.4269, "step": 6118 }, { "epoch": 0.31, "learning_rate": 1.6275634447965212e-05, "loss": 0.5176, "step": 6119 }, { "epoch": 0.31, "learning_rate": 1.6274375261908987e-05, "loss": 0.3865, "step": 6120 }, { "epoch": 0.31, "learning_rate": 1.627311591175469e-05, "loss": 0.365, "step": 6121 }, { "epoch": 0.31, "learning_rate": 1.6271856397535265e-05, "loss": 0.0461, "step": 6122 }, { "epoch": 0.31, "learning_rate": 1.627059671928365e-05, "loss": 0.4007, "step": 6123 }, { "epoch": 0.31, "learning_rate": 1.626933687703279e-05, "loss": 0.3415, "step": 6124 }, { "epoch": 0.31, "learning_rate": 1.6268076870815642e-05, "loss": 0.4528, "step": 6125 }, { "epoch": 0.31, "learning_rate": 1.6266816700665144e-05, "loss": 0.3465, "step": 6126 }, { "epoch": 0.31, "learning_rate": 1.626555636661427e-05, "loss": 0.1609, "step": 6127 }, { "epoch": 0.31, "learning_rate": 1.6264295868695973e-05, "loss": 0.3037, "step": 6128 }, { "epoch": 0.31, "learning_rate": 1.626303520694322e-05, "loss": 0.2277, "step": 6129 }, { "epoch": 0.31, "learning_rate": 1.6261774381388988e-05, "loss": 0.5476, "step": 6130 }, { "epoch": 0.31, "learning_rate": 1.6260513392066245e-05, "loss": 0.4103, "step": 6131 }, { "epoch": 0.31, "learning_rate": 1.6259252239007982e-05, "loss": 0.3142, "step": 6132 }, { "epoch": 0.31, "learning_rate": 1.6257990922247168e-05, "loss": 0.354, "step": 6133 }, { "epoch": 0.31, "learning_rate": 1.62567294418168e-05, "loss": 0.4019, "step": 6134 }, { "epoch": 0.31, "learning_rate": 1.6255467797749863e-05, "loss": 0.401, "step": 6135 }, { "epoch": 0.31, "learning_rate": 1.6254205990079365e-05, "loss": 0.753, "step": 6136 }, { "epoch": 0.31, "learning_rate": 1.6252944018838296e-05, "loss": 0.406, "step": 6137 }, { "epoch": 0.31, "learning_rate": 1.6251681884059675e-05, "loss": 0.5871, "step": 6138 }, { "epoch": 0.31, "learning_rate": 1.6250419585776495e-05, "loss": 0.4747, "step": 6139 }, { "epoch": 0.31, "learning_rate": 1.6249157124021777e-05, "loss": 0.5239, "step": 6140 }, { "epoch": 0.31, "learning_rate": 1.6247894498828542e-05, "loss": 0.5028, "step": 6141 }, { "epoch": 0.31, "learning_rate": 1.6246631710229807e-05, "loss": 0.399, "step": 6142 }, { "epoch": 0.31, "learning_rate": 1.6245368758258603e-05, "loss": 0.5383, "step": 6143 }, { "epoch": 0.31, "learning_rate": 1.624410564294796e-05, "loss": 0.4542, "step": 6144 }, { "epoch": 0.31, "learning_rate": 1.6242842364330912e-05, "loss": 0.3369, "step": 6145 }, { "epoch": 0.31, "learning_rate": 1.6241578922440496e-05, "loss": 0.4633, "step": 6146 }, { "epoch": 0.31, "learning_rate": 1.6240315317309762e-05, "loss": 0.4523, "step": 6147 }, { "epoch": 0.31, "learning_rate": 1.6239051548971754e-05, "loss": 0.3103, "step": 6148 }, { "epoch": 0.31, "learning_rate": 1.6237787617459524e-05, "loss": 0.4695, "step": 6149 }, { "epoch": 0.31, "learning_rate": 1.6236523522806125e-05, "loss": 0.2414, "step": 6150 }, { "epoch": 0.31, "learning_rate": 1.6235259265044627e-05, "loss": 0.3644, "step": 6151 }, { "epoch": 0.31, "learning_rate": 1.6233994844208087e-05, "loss": 0.3236, "step": 6152 }, { "epoch": 0.31, "learning_rate": 1.6232730260329576e-05, "loss": 0.4357, "step": 6153 }, { "epoch": 0.31, "learning_rate": 1.623146551344217e-05, "loss": 0.3768, "step": 6154 }, { "epoch": 0.31, "learning_rate": 1.6230200603578947e-05, "loss": 0.4194, "step": 6155 }, { "epoch": 0.31, "learning_rate": 1.6228935530772985e-05, "loss": 0.3482, "step": 6156 }, { "epoch": 0.31, "learning_rate": 1.6227670295057373e-05, "loss": 0.4256, "step": 6157 }, { "epoch": 0.31, "learning_rate": 1.62264048964652e-05, "loss": 0.4163, "step": 6158 }, { "epoch": 0.31, "learning_rate": 1.6225139335029566e-05, "loss": 0.3746, "step": 6159 }, { "epoch": 0.31, "learning_rate": 1.622387361078356e-05, "loss": 0.5027, "step": 6160 }, { "epoch": 0.31, "learning_rate": 1.6222607723760294e-05, "loss": 0.3432, "step": 6161 }, { "epoch": 0.31, "learning_rate": 1.6221341673992873e-05, "loss": 0.4771, "step": 6162 }, { "epoch": 0.31, "learning_rate": 1.622007546151441e-05, "loss": 0.2667, "step": 6163 }, { "epoch": 0.31, "learning_rate": 1.6218809086358024e-05, "loss": 0.479, "step": 6164 }, { "epoch": 0.31, "learning_rate": 1.6217542548556826e-05, "loss": 0.3575, "step": 6165 }, { "epoch": 0.31, "learning_rate": 1.621627584814395e-05, "loss": 0.4777, "step": 6166 }, { "epoch": 0.31, "learning_rate": 1.6215008985152516e-05, "loss": 0.2395, "step": 6167 }, { "epoch": 0.31, "learning_rate": 1.6213741959615666e-05, "loss": 0.2957, "step": 6168 }, { "epoch": 0.31, "learning_rate": 1.6212474771566533e-05, "loss": 0.2627, "step": 6169 }, { "epoch": 0.31, "learning_rate": 1.621120742103826e-05, "loss": 0.2863, "step": 6170 }, { "epoch": 0.31, "learning_rate": 1.620993990806399e-05, "loss": 0.5581, "step": 6171 }, { "epoch": 0.31, "learning_rate": 1.6208672232676877e-05, "loss": 0.4587, "step": 6172 }, { "epoch": 0.31, "learning_rate": 1.620740439491007e-05, "loss": 0.3943, "step": 6173 }, { "epoch": 0.31, "learning_rate": 1.6206136394796733e-05, "loss": 0.427, "step": 6174 }, { "epoch": 0.31, "learning_rate": 1.620486823237003e-05, "loss": 0.4725, "step": 6175 }, { "epoch": 0.31, "learning_rate": 1.6203599907663124e-05, "loss": 0.5162, "step": 6176 }, { "epoch": 0.31, "learning_rate": 1.6202331420709187e-05, "loss": 0.2348, "step": 6177 }, { "epoch": 0.31, "learning_rate": 1.6201062771541396e-05, "loss": 0.44, "step": 6178 }, { "epoch": 0.31, "learning_rate": 1.6199793960192928e-05, "loss": 0.3742, "step": 6179 }, { "epoch": 0.31, "learning_rate": 1.6198524986696974e-05, "loss": 0.3418, "step": 6180 }, { "epoch": 0.31, "learning_rate": 1.6197255851086713e-05, "loss": 0.4326, "step": 6181 }, { "epoch": 0.31, "learning_rate": 1.6195986553395343e-05, "loss": 0.4241, "step": 6182 }, { "epoch": 0.31, "learning_rate": 1.6194717093656063e-05, "loss": 0.455, "step": 6183 }, { "epoch": 0.31, "learning_rate": 1.6193447471902067e-05, "loss": 0.6708, "step": 6184 }, { "epoch": 0.31, "learning_rate": 1.6192177688166566e-05, "loss": 0.343, "step": 6185 }, { "epoch": 0.31, "learning_rate": 1.6190907742482766e-05, "loss": 0.2891, "step": 6186 }, { "epoch": 0.31, "learning_rate": 1.6189637634883886e-05, "loss": 0.3927, "step": 6187 }, { "epoch": 0.31, "learning_rate": 1.618836736540314e-05, "loss": 0.6037, "step": 6188 }, { "epoch": 0.31, "learning_rate": 1.618709693407375e-05, "loss": 0.3116, "step": 6189 }, { "epoch": 0.31, "learning_rate": 1.6185826340928943e-05, "loss": 0.2351, "step": 6190 }, { "epoch": 0.31, "learning_rate": 1.618455558600195e-05, "loss": 0.4624, "step": 6191 }, { "epoch": 0.31, "learning_rate": 1.618328466932601e-05, "loss": 0.4516, "step": 6192 }, { "epoch": 0.31, "learning_rate": 1.6182013590934353e-05, "loss": 0.3581, "step": 6193 }, { "epoch": 0.31, "learning_rate": 1.6180742350860228e-05, "loss": 0.206, "step": 6194 }, { "epoch": 0.31, "learning_rate": 1.617947094913688e-05, "loss": 0.1963, "step": 6195 }, { "epoch": 0.31, "learning_rate": 1.617819938579757e-05, "loss": 0.2071, "step": 6196 }, { "epoch": 0.31, "learning_rate": 1.617692766087554e-05, "loss": 0.3832, "step": 6197 }, { "epoch": 0.31, "learning_rate": 1.617565577440406e-05, "loss": 0.3127, "step": 6198 }, { "epoch": 0.31, "learning_rate": 1.6174383726416393e-05, "loss": 0.4501, "step": 6199 }, { "epoch": 0.31, "learning_rate": 1.617311151694581e-05, "loss": 0.448, "step": 6200 }, { "epoch": 0.31, "learning_rate": 1.6171839146025573e-05, "loss": 0.4051, "step": 6201 }, { "epoch": 0.31, "learning_rate": 1.617056661368897e-05, "loss": 0.2618, "step": 6202 }, { "epoch": 0.31, "learning_rate": 1.616929391996928e-05, "loss": 0.0886, "step": 6203 }, { "epoch": 0.31, "learning_rate": 1.616802106489979e-05, "loss": 0.3021, "step": 6204 }, { "epoch": 0.31, "learning_rate": 1.6166748048513782e-05, "loss": 0.5752, "step": 6205 }, { "epoch": 0.31, "learning_rate": 1.6165474870844563e-05, "loss": 0.5625, "step": 6206 }, { "epoch": 0.31, "learning_rate": 1.6164201531925418e-05, "loss": 0.1779, "step": 6207 }, { "epoch": 0.31, "learning_rate": 1.616292803178966e-05, "loss": 0.3793, "step": 6208 }, { "epoch": 0.31, "learning_rate": 1.6161654370470587e-05, "loss": 0.361, "step": 6209 }, { "epoch": 0.31, "learning_rate": 1.6160380548001515e-05, "loss": 0.3192, "step": 6210 }, { "epoch": 0.31, "learning_rate": 1.6159106564415758e-05, "loss": 0.3088, "step": 6211 }, { "epoch": 0.31, "learning_rate": 1.615783241974664e-05, "loss": 0.5053, "step": 6212 }, { "epoch": 0.31, "learning_rate": 1.6156558114027476e-05, "loss": 0.4952, "step": 6213 }, { "epoch": 0.31, "learning_rate": 1.6155283647291602e-05, "loss": 0.3917, "step": 6214 }, { "epoch": 0.31, "learning_rate": 1.6154009019572342e-05, "loss": 0.142, "step": 6215 }, { "epoch": 0.31, "learning_rate": 1.615273423090304e-05, "loss": 0.3603, "step": 6216 }, { "epoch": 0.31, "learning_rate": 1.6151459281317034e-05, "loss": 0.3269, "step": 6217 }, { "epoch": 0.31, "learning_rate": 1.6150184170847663e-05, "loss": 0.4348, "step": 6218 }, { "epoch": 0.31, "learning_rate": 1.614890889952828e-05, "loss": 0.325, "step": 6219 }, { "epoch": 0.31, "learning_rate": 1.614763346739224e-05, "loss": 0.3311, "step": 6220 }, { "epoch": 0.31, "learning_rate": 1.61463578744729e-05, "loss": 0.361, "step": 6221 }, { "epoch": 0.31, "learning_rate": 1.614508212080362e-05, "loss": 0.5632, "step": 6222 }, { "epoch": 0.31, "learning_rate": 1.6143806206417764e-05, "loss": 0.4813, "step": 6223 }, { "epoch": 0.31, "learning_rate": 1.6142530131348703e-05, "loss": 0.344, "step": 6224 }, { "epoch": 0.31, "learning_rate": 1.6141253895629815e-05, "loss": 0.454, "step": 6225 }, { "epoch": 0.31, "learning_rate": 1.6139977499294472e-05, "loss": 0.364, "step": 6226 }, { "epoch": 0.31, "learning_rate": 1.6138700942376062e-05, "loss": 0.4628, "step": 6227 }, { "epoch": 0.31, "learning_rate": 1.6137424224907967e-05, "loss": 0.341, "step": 6228 }, { "epoch": 0.31, "learning_rate": 1.613614734692358e-05, "loss": 0.4016, "step": 6229 }, { "epoch": 0.31, "learning_rate": 1.61348703084563e-05, "loss": 0.435, "step": 6230 }, { "epoch": 0.31, "learning_rate": 1.6133593109539515e-05, "loss": 0.4946, "step": 6231 }, { "epoch": 0.31, "learning_rate": 1.613231575020664e-05, "loss": 0.2797, "step": 6232 }, { "epoch": 0.31, "learning_rate": 1.6131038230491078e-05, "loss": 0.284, "step": 6233 }, { "epoch": 0.31, "learning_rate": 1.6129760550426244e-05, "loss": 0.3934, "step": 6234 }, { "epoch": 0.31, "learning_rate": 1.612848271004555e-05, "loss": 0.3246, "step": 6235 }, { "epoch": 0.31, "learning_rate": 1.6127204709382414e-05, "loss": 0.7213, "step": 6236 }, { "epoch": 0.31, "learning_rate": 1.6125926548470268e-05, "loss": 0.5952, "step": 6237 }, { "epoch": 0.31, "learning_rate": 1.6124648227342534e-05, "loss": 0.52, "step": 6238 }, { "epoch": 0.31, "learning_rate": 1.6123369746032648e-05, "loss": 0.2607, "step": 6239 }, { "epoch": 0.31, "learning_rate": 1.6122091104574045e-05, "loss": 0.4487, "step": 6240 }, { "epoch": 0.31, "learning_rate": 1.612081230300017e-05, "loss": 0.3839, "step": 6241 }, { "epoch": 0.31, "learning_rate": 1.6119533341344466e-05, "loss": 0.3854, "step": 6242 }, { "epoch": 0.31, "learning_rate": 1.6118254219640385e-05, "loss": 0.1442, "step": 6243 }, { "epoch": 0.31, "learning_rate": 1.6116974937921375e-05, "loss": 0.3179, "step": 6244 }, { "epoch": 0.31, "learning_rate": 1.6115695496220898e-05, "loss": 0.4782, "step": 6245 }, { "epoch": 0.31, "learning_rate": 1.6114415894572414e-05, "loss": 0.454, "step": 6246 }, { "epoch": 0.31, "learning_rate": 1.6113136133009392e-05, "loss": 0.6094, "step": 6247 }, { "epoch": 0.31, "learning_rate": 1.6111856211565303e-05, "loss": 0.1349, "step": 6248 }, { "epoch": 0.31, "learning_rate": 1.6110576130273615e-05, "loss": 0.4174, "step": 6249 }, { "epoch": 0.31, "learning_rate": 1.6109295889167817e-05, "loss": 0.1639, "step": 6250 }, { "epoch": 0.31, "learning_rate": 1.6108015488281385e-05, "loss": 0.5998, "step": 6251 }, { "epoch": 0.31, "learning_rate": 1.610673492764781e-05, "loss": 0.4186, "step": 6252 }, { "epoch": 0.31, "learning_rate": 1.6105454207300578e-05, "loss": 0.3645, "step": 6253 }, { "epoch": 0.31, "learning_rate": 1.610417332727319e-05, "loss": 0.4888, "step": 6254 }, { "epoch": 0.31, "learning_rate": 1.610289228759914e-05, "loss": 0.6526, "step": 6255 }, { "epoch": 0.31, "learning_rate": 1.6101611088311938e-05, "loss": 0.2654, "step": 6256 }, { "epoch": 0.31, "learning_rate": 1.610032972944509e-05, "loss": 0.2396, "step": 6257 }, { "epoch": 0.31, "learning_rate": 1.6099048211032107e-05, "loss": 0.335, "step": 6258 }, { "epoch": 0.31, "learning_rate": 1.609776653310651e-05, "loss": 0.4055, "step": 6259 }, { "epoch": 0.31, "learning_rate": 1.6096484695701814e-05, "loss": 0.3517, "step": 6260 }, { "epoch": 0.31, "learning_rate": 1.6095202698851543e-05, "loss": 0.75, "step": 6261 }, { "epoch": 0.31, "learning_rate": 1.609392054258923e-05, "loss": 0.4614, "step": 6262 }, { "epoch": 0.31, "learning_rate": 1.6092638226948407e-05, "loss": 0.3221, "step": 6263 }, { "epoch": 0.31, "learning_rate": 1.6091355751962613e-05, "loss": 0.2568, "step": 6264 }, { "epoch": 0.31, "learning_rate": 1.6090073117665385e-05, "loss": 0.3999, "step": 6265 }, { "epoch": 0.31, "learning_rate": 1.6088790324090273e-05, "loss": 0.5755, "step": 6266 }, { "epoch": 0.31, "learning_rate": 1.608750737127082e-05, "loss": 0.4478, "step": 6267 }, { "epoch": 0.31, "learning_rate": 1.6086224259240594e-05, "loss": 0.4752, "step": 6268 }, { "epoch": 0.31, "learning_rate": 1.608494098803314e-05, "loss": 0.2136, "step": 6269 }, { "epoch": 0.31, "learning_rate": 1.608365755768202e-05, "loss": 0.2972, "step": 6270 }, { "epoch": 0.31, "learning_rate": 1.608237396822081e-05, "loss": 0.5158, "step": 6271 }, { "epoch": 0.31, "learning_rate": 1.6081090219683074e-05, "loss": 0.5157, "step": 6272 }, { "epoch": 0.31, "learning_rate": 1.607980631210239e-05, "loss": 0.2254, "step": 6273 }, { "epoch": 0.31, "learning_rate": 1.6078522245512326e-05, "loss": 0.423, "step": 6274 }, { "epoch": 0.31, "learning_rate": 1.6077238019946484e-05, "loss": 0.2498, "step": 6275 }, { "epoch": 0.31, "learning_rate": 1.607595363543844e-05, "loss": 0.3013, "step": 6276 }, { "epoch": 0.31, "learning_rate": 1.607466909202178e-05, "loss": 0.2866, "step": 6277 }, { "epoch": 0.31, "learning_rate": 1.6073384389730112e-05, "loss": 0.3155, "step": 6278 }, { "epoch": 0.31, "learning_rate": 1.607209952859703e-05, "loss": 0.3749, "step": 6279 }, { "epoch": 0.31, "learning_rate": 1.607081450865614e-05, "loss": 0.5394, "step": 6280 }, { "epoch": 0.31, "learning_rate": 1.606952932994104e-05, "loss": 0.2117, "step": 6281 }, { "epoch": 0.31, "learning_rate": 1.6068243992485358e-05, "loss": 0.5192, "step": 6282 }, { "epoch": 0.31, "learning_rate": 1.6066958496322698e-05, "loss": 0.4107, "step": 6283 }, { "epoch": 0.31, "learning_rate": 1.6065672841486685e-05, "loss": 0.6125, "step": 6284 }, { "epoch": 0.31, "learning_rate": 1.6064387028010945e-05, "loss": 0.4377, "step": 6285 }, { "epoch": 0.31, "learning_rate": 1.6063101055929106e-05, "loss": 0.4359, "step": 6286 }, { "epoch": 0.31, "learning_rate": 1.6061814925274798e-05, "loss": 0.4884, "step": 6287 }, { "epoch": 0.31, "learning_rate": 1.606052863608166e-05, "loss": 0.4165, "step": 6288 }, { "epoch": 0.31, "learning_rate": 1.605924218838333e-05, "loss": 0.3242, "step": 6289 }, { "epoch": 0.31, "learning_rate": 1.6057955582213464e-05, "loss": 0.5889, "step": 6290 }, { "epoch": 0.31, "learning_rate": 1.6056668817605702e-05, "loss": 0.1388, "step": 6291 }, { "epoch": 0.31, "learning_rate": 1.6055381894593702e-05, "loss": 0.2129, "step": 6292 }, { "epoch": 0.31, "learning_rate": 1.6054094813211114e-05, "loss": 0.3337, "step": 6293 }, { "epoch": 0.31, "learning_rate": 1.6052807573491608e-05, "loss": 0.4, "step": 6294 }, { "epoch": 0.31, "learning_rate": 1.605152017546885e-05, "loss": 0.161, "step": 6295 }, { "epoch": 0.31, "learning_rate": 1.6050232619176506e-05, "loss": 0.3393, "step": 6296 }, { "epoch": 0.31, "learning_rate": 1.6048944904648252e-05, "loss": 0.5749, "step": 6297 }, { "epoch": 0.31, "learning_rate": 1.6047657031917765e-05, "loss": 0.3476, "step": 6298 }, { "epoch": 0.31, "learning_rate": 1.604636900101873e-05, "loss": 0.3172, "step": 6299 }, { "epoch": 0.31, "learning_rate": 1.6045080811984837e-05, "loss": 0.1458, "step": 6300 }, { "epoch": 0.31, "learning_rate": 1.604379246484977e-05, "loss": 0.2627, "step": 6301 }, { "epoch": 0.31, "learning_rate": 1.6042503959647227e-05, "loss": 0.2872, "step": 6302 }, { "epoch": 0.31, "learning_rate": 1.604121529641091e-05, "loss": 0.5057, "step": 6303 }, { "epoch": 0.31, "learning_rate": 1.6039926475174516e-05, "loss": 0.1945, "step": 6304 }, { "epoch": 0.31, "learning_rate": 1.6038637495971757e-05, "loss": 0.3732, "step": 6305 }, { "epoch": 0.31, "learning_rate": 1.603734835883634e-05, "loss": 0.5426, "step": 6306 }, { "epoch": 0.31, "learning_rate": 1.603605906380199e-05, "loss": 0.3332, "step": 6307 }, { "epoch": 0.31, "learning_rate": 1.6034769610902418e-05, "loss": 0.3758, "step": 6308 }, { "epoch": 0.32, "learning_rate": 1.603348000017135e-05, "loss": 0.4195, "step": 6309 }, { "epoch": 0.32, "learning_rate": 1.6032190231642516e-05, "loss": 0.4867, "step": 6310 }, { "epoch": 0.32, "learning_rate": 1.6030900305349647e-05, "loss": 0.535, "step": 6311 }, { "epoch": 0.32, "learning_rate": 1.6029610221326484e-05, "loss": 0.6148, "step": 6312 }, { "epoch": 0.32, "learning_rate": 1.6028319979606755e-05, "loss": 0.5191, "step": 6313 }, { "epoch": 0.32, "learning_rate": 1.602702958022422e-05, "loss": 0.4723, "step": 6314 }, { "epoch": 0.32, "learning_rate": 1.6025739023212617e-05, "loss": 0.3423, "step": 6315 }, { "epoch": 0.32, "learning_rate": 1.6024448308605704e-05, "loss": 0.4384, "step": 6316 }, { "epoch": 0.32, "learning_rate": 1.6023157436437235e-05, "loss": 0.5134, "step": 6317 }, { "epoch": 0.32, "learning_rate": 1.6021866406740973e-05, "loss": 0.4732, "step": 6318 }, { "epoch": 0.32, "learning_rate": 1.602057521955068e-05, "loss": 0.6755, "step": 6319 }, { "epoch": 0.32, "learning_rate": 1.6019283874900126e-05, "loss": 0.4007, "step": 6320 }, { "epoch": 0.32, "learning_rate": 1.601799237282309e-05, "loss": 0.522, "step": 6321 }, { "epoch": 0.32, "learning_rate": 1.6016700713353346e-05, "loss": 0.4622, "step": 6322 }, { "epoch": 0.32, "learning_rate": 1.6015408896524673e-05, "loss": 0.5572, "step": 6323 }, { "epoch": 0.32, "learning_rate": 1.601411692237086e-05, "loss": 0.4236, "step": 6324 }, { "epoch": 0.32, "learning_rate": 1.6012824790925697e-05, "loss": 0.2918, "step": 6325 }, { "epoch": 0.32, "learning_rate": 1.6011532502222977e-05, "loss": 0.3577, "step": 6326 }, { "epoch": 0.32, "learning_rate": 1.6010240056296494e-05, "loss": 0.4544, "step": 6327 }, { "epoch": 0.32, "learning_rate": 1.600894745318006e-05, "loss": 0.3182, "step": 6328 }, { "epoch": 0.32, "learning_rate": 1.600765469290747e-05, "loss": 0.473, "step": 6329 }, { "epoch": 0.32, "learning_rate": 1.6006361775512544e-05, "loss": 0.2913, "step": 6330 }, { "epoch": 0.32, "learning_rate": 1.600506870102909e-05, "loss": 0.3119, "step": 6331 }, { "epoch": 0.32, "learning_rate": 1.600377546949093e-05, "loss": 0.3122, "step": 6332 }, { "epoch": 0.32, "learning_rate": 1.6002482080931893e-05, "loss": 0.2454, "step": 6333 }, { "epoch": 0.32, "learning_rate": 1.600118853538579e-05, "loss": 0.3277, "step": 6334 }, { "epoch": 0.32, "learning_rate": 1.5999894832886463e-05, "loss": 0.341, "step": 6335 }, { "epoch": 0.32, "learning_rate": 1.599860097346775e-05, "loss": 0.594, "step": 6336 }, { "epoch": 0.32, "learning_rate": 1.599730695716348e-05, "loss": 0.4129, "step": 6337 }, { "epoch": 0.32, "learning_rate": 1.5996012784007503e-05, "loss": 0.389, "step": 6338 }, { "epoch": 0.32, "learning_rate": 1.5994718454033665e-05, "loss": 0.2149, "step": 6339 }, { "epoch": 0.32, "learning_rate": 1.5993423967275818e-05, "loss": 0.4876, "step": 6340 }, { "epoch": 0.32, "learning_rate": 1.599212932376782e-05, "loss": 0.5197, "step": 6341 }, { "epoch": 0.32, "learning_rate": 1.5990834523543522e-05, "loss": 0.4178, "step": 6342 }, { "epoch": 0.32, "learning_rate": 1.5989539566636795e-05, "loss": 0.3902, "step": 6343 }, { "epoch": 0.32, "learning_rate": 1.598824445308151e-05, "loss": 0.4401, "step": 6344 }, { "epoch": 0.32, "learning_rate": 1.5986949182911532e-05, "loss": 0.4065, "step": 6345 }, { "epoch": 0.32, "learning_rate": 1.598565375616074e-05, "loss": 0.4374, "step": 6346 }, { "epoch": 0.32, "learning_rate": 1.598435817286301e-05, "loss": 0.3669, "step": 6347 }, { "epoch": 0.32, "learning_rate": 1.598306243305223e-05, "loss": 0.3112, "step": 6348 }, { "epoch": 0.32, "learning_rate": 1.5981766536762292e-05, "loss": 0.4706, "step": 6349 }, { "epoch": 0.32, "learning_rate": 1.5980470484027083e-05, "loss": 0.4532, "step": 6350 }, { "epoch": 0.32, "learning_rate": 1.5979174274880508e-05, "loss": 0.3294, "step": 6351 }, { "epoch": 0.32, "learning_rate": 1.5977877909356455e-05, "loss": 0.2276, "step": 6352 }, { "epoch": 0.32, "learning_rate": 1.5976581387488835e-05, "loss": 0.3879, "step": 6353 }, { "epoch": 0.32, "learning_rate": 1.5975284709311555e-05, "loss": 0.2942, "step": 6354 }, { "epoch": 0.32, "learning_rate": 1.5973987874858533e-05, "loss": 0.3322, "step": 6355 }, { "epoch": 0.32, "learning_rate": 1.5972690884163683e-05, "loss": 0.4755, "step": 6356 }, { "epoch": 0.32, "learning_rate": 1.597139373726092e-05, "loss": 0.2545, "step": 6357 }, { "epoch": 0.32, "learning_rate": 1.5970096434184187e-05, "loss": 0.1081, "step": 6358 }, { "epoch": 0.32, "learning_rate": 1.596879897496739e-05, "loss": 0.3397, "step": 6359 }, { "epoch": 0.32, "learning_rate": 1.596750135964448e-05, "loss": 0.4712, "step": 6360 }, { "epoch": 0.32, "learning_rate": 1.5966203588249384e-05, "loss": 0.2325, "step": 6361 }, { "epoch": 0.32, "learning_rate": 1.5964905660816048e-05, "loss": 0.3818, "step": 6362 }, { "epoch": 0.32, "learning_rate": 1.5963607577378416e-05, "loss": 0.4659, "step": 6363 }, { "epoch": 0.32, "learning_rate": 1.596230933797044e-05, "loss": 0.3922, "step": 6364 }, { "epoch": 0.32, "learning_rate": 1.5961010942626076e-05, "loss": 0.5098, "step": 6365 }, { "epoch": 0.32, "learning_rate": 1.5959712391379273e-05, "loss": 0.5681, "step": 6366 }, { "epoch": 0.32, "learning_rate": 1.5958413684264e-05, "loss": 0.2313, "step": 6367 }, { "epoch": 0.32, "learning_rate": 1.595711482131422e-05, "loss": 0.4139, "step": 6368 }, { "epoch": 0.32, "learning_rate": 1.595581580256391e-05, "loss": 0.2422, "step": 6369 }, { "epoch": 0.32, "learning_rate": 1.5954516628047034e-05, "loss": 0.401, "step": 6370 }, { "epoch": 0.32, "learning_rate": 1.5953217297797575e-05, "loss": 0.3732, "step": 6371 }, { "epoch": 0.32, "learning_rate": 1.5951917811849515e-05, "loss": 0.2579, "step": 6372 }, { "epoch": 0.32, "learning_rate": 1.595061817023684e-05, "loss": 0.3085, "step": 6373 }, { "epoch": 0.32, "learning_rate": 1.594931837299354e-05, "loss": 0.2825, "step": 6374 }, { "epoch": 0.32, "learning_rate": 1.594801842015361e-05, "loss": 0.2697, "step": 6375 }, { "epoch": 0.32, "learning_rate": 1.594671831175105e-05, "loss": 0.6699, "step": 6376 }, { "epoch": 0.32, "learning_rate": 1.594541804781986e-05, "loss": 0.3096, "step": 6377 }, { "epoch": 0.32, "learning_rate": 1.594411762839405e-05, "loss": 0.3229, "step": 6378 }, { "epoch": 0.32, "learning_rate": 1.594281705350763e-05, "loss": 0.4605, "step": 6379 }, { "epoch": 0.32, "learning_rate": 1.594151632319461e-05, "loss": 0.3091, "step": 6380 }, { "epoch": 0.32, "learning_rate": 1.5940215437489015e-05, "loss": 0.5655, "step": 6381 }, { "epoch": 0.32, "learning_rate": 1.5938914396424866e-05, "loss": 0.21, "step": 6382 }, { "epoch": 0.32, "learning_rate": 1.5937613200036185e-05, "loss": 0.243, "step": 6383 }, { "epoch": 0.32, "learning_rate": 1.5936311848357014e-05, "loss": 0.3994, "step": 6384 }, { "epoch": 0.32, "learning_rate": 1.5935010341421376e-05, "loss": 0.2533, "step": 6385 }, { "epoch": 0.32, "learning_rate": 1.593370867926332e-05, "loss": 0.3033, "step": 6386 }, { "epoch": 0.32, "learning_rate": 1.593240686191688e-05, "loss": 0.2753, "step": 6387 }, { "epoch": 0.32, "learning_rate": 1.5931104889416116e-05, "loss": 0.3187, "step": 6388 }, { "epoch": 0.32, "learning_rate": 1.5929802761795066e-05, "loss": 0.328, "step": 6389 }, { "epoch": 0.32, "learning_rate": 1.5928500479087796e-05, "loss": 0.2356, "step": 6390 }, { "epoch": 0.32, "learning_rate": 1.5927198041328357e-05, "loss": 0.4881, "step": 6391 }, { "epoch": 0.32, "learning_rate": 1.592589544855082e-05, "loss": 0.2075, "step": 6392 }, { "epoch": 0.32, "learning_rate": 1.5924592700789244e-05, "loss": 0.4448, "step": 6393 }, { "epoch": 0.32, "learning_rate": 1.5923289798077713e-05, "loss": 0.2627, "step": 6394 }, { "epoch": 0.32, "learning_rate": 1.592198674045029e-05, "loss": 0.4927, "step": 6395 }, { "epoch": 0.32, "learning_rate": 1.5920683527941062e-05, "loss": 0.4116, "step": 6396 }, { "epoch": 0.32, "learning_rate": 1.5919380160584112e-05, "loss": 0.5556, "step": 6397 }, { "epoch": 0.32, "learning_rate": 1.5918076638413523e-05, "loss": 0.1799, "step": 6398 }, { "epoch": 0.32, "learning_rate": 1.5916772961463395e-05, "loss": 0.264, "step": 6399 }, { "epoch": 0.32, "learning_rate": 1.5915469129767818e-05, "loss": 0.6381, "step": 6400 }, { "epoch": 0.32, "learning_rate": 1.5914165143360897e-05, "loss": 0.3476, "step": 6401 }, { "epoch": 0.32, "learning_rate": 1.5912861002276728e-05, "loss": 0.1658, "step": 6402 }, { "epoch": 0.32, "learning_rate": 1.5911556706549424e-05, "loss": 0.2311, "step": 6403 }, { "epoch": 0.32, "learning_rate": 1.59102522562131e-05, "loss": 0.2907, "step": 6404 }, { "epoch": 0.32, "learning_rate": 1.5908947651301866e-05, "loss": 0.4841, "step": 6405 }, { "epoch": 0.32, "learning_rate": 1.5907642891849848e-05, "loss": 0.4574, "step": 6406 }, { "epoch": 0.32, "learning_rate": 1.5906337977891167e-05, "loss": 0.485, "step": 6407 }, { "epoch": 0.32, "learning_rate": 1.5905032909459955e-05, "loss": 0.3999, "step": 6408 }, { "epoch": 0.32, "learning_rate": 1.5903727686590337e-05, "loss": 0.3849, "step": 6409 }, { "epoch": 0.32, "learning_rate": 1.5902422309316455e-05, "loss": 0.3455, "step": 6410 }, { "epoch": 0.32, "learning_rate": 1.5901116777672448e-05, "loss": 0.3249, "step": 6411 }, { "epoch": 0.32, "learning_rate": 1.589981109169246e-05, "loss": 0.3321, "step": 6412 }, { "epoch": 0.32, "learning_rate": 1.589850525141065e-05, "loss": 0.4411, "step": 6413 }, { "epoch": 0.32, "learning_rate": 1.589719925686115e-05, "loss": 0.3868, "step": 6414 }, { "epoch": 0.32, "learning_rate": 1.589589310807813e-05, "loss": 0.4258, "step": 6415 }, { "epoch": 0.32, "learning_rate": 1.5894586805095747e-05, "loss": 0.5304, "step": 6416 }, { "epoch": 0.32, "learning_rate": 1.589328034794817e-05, "loss": 0.2364, "step": 6417 }, { "epoch": 0.32, "learning_rate": 1.589197373666956e-05, "loss": 0.3326, "step": 6418 }, { "epoch": 0.32, "learning_rate": 1.5890666971294097e-05, "loss": 0.4049, "step": 6419 }, { "epoch": 0.32, "learning_rate": 1.588936005185595e-05, "loss": 0.291, "step": 6420 }, { "epoch": 0.32, "learning_rate": 1.5888052978389312e-05, "loss": 0.5405, "step": 6421 }, { "epoch": 0.32, "learning_rate": 1.588674575092836e-05, "loss": 0.2875, "step": 6422 }, { "epoch": 0.32, "learning_rate": 1.5885438369507277e-05, "loss": 0.3499, "step": 6423 }, { "epoch": 0.32, "learning_rate": 1.5884130834160264e-05, "loss": 0.3086, "step": 6424 }, { "epoch": 0.32, "learning_rate": 1.5882823144921516e-05, "loss": 0.4421, "step": 6425 }, { "epoch": 0.32, "learning_rate": 1.5881515301825237e-05, "loss": 0.3876, "step": 6426 }, { "epoch": 0.32, "learning_rate": 1.5880207304905623e-05, "loss": 0.2716, "step": 6427 }, { "epoch": 0.32, "learning_rate": 1.5878899154196895e-05, "loss": 0.391, "step": 6428 }, { "epoch": 0.32, "learning_rate": 1.5877590849733256e-05, "loss": 0.2443, "step": 6429 }, { "epoch": 0.32, "learning_rate": 1.5876282391548925e-05, "loss": 0.4793, "step": 6430 }, { "epoch": 0.32, "learning_rate": 1.5874973779678126e-05, "loss": 0.1646, "step": 6431 }, { "epoch": 0.32, "learning_rate": 1.5873665014155087e-05, "loss": 0.3114, "step": 6432 }, { "epoch": 0.32, "learning_rate": 1.5872356095014027e-05, "loss": 0.3994, "step": 6433 }, { "epoch": 0.32, "learning_rate": 1.5871047022289187e-05, "loss": 0.4628, "step": 6434 }, { "epoch": 0.32, "learning_rate": 1.58697377960148e-05, "loss": 0.367, "step": 6435 }, { "epoch": 0.32, "learning_rate": 1.586842841622511e-05, "loss": 0.3613, "step": 6436 }, { "epoch": 0.32, "learning_rate": 1.5867118882954363e-05, "loss": 0.2771, "step": 6437 }, { "epoch": 0.32, "learning_rate": 1.5865809196236806e-05, "loss": 0.387, "step": 6438 }, { "epoch": 0.32, "learning_rate": 1.586449935610669e-05, "loss": 0.3568, "step": 6439 }, { "epoch": 0.32, "learning_rate": 1.5863189362598275e-05, "loss": 0.3614, "step": 6440 }, { "epoch": 0.32, "learning_rate": 1.586187921574582e-05, "loss": 0.2904, "step": 6441 }, { "epoch": 0.32, "learning_rate": 1.5860568915583592e-05, "loss": 0.2046, "step": 6442 }, { "epoch": 0.32, "learning_rate": 1.5859258462145866e-05, "loss": 0.3055, "step": 6443 }, { "epoch": 0.32, "learning_rate": 1.5857947855466903e-05, "loss": 0.4936, "step": 6444 }, { "epoch": 0.32, "learning_rate": 1.5856637095580988e-05, "loss": 0.585, "step": 6445 }, { "epoch": 0.32, "learning_rate": 1.58553261825224e-05, "loss": 0.5388, "step": 6446 }, { "epoch": 0.32, "learning_rate": 1.585401511632543e-05, "loss": 0.4101, "step": 6447 }, { "epoch": 0.32, "learning_rate": 1.5852703897024354e-05, "loss": 0.3993, "step": 6448 }, { "epoch": 0.32, "learning_rate": 1.5851392524653478e-05, "loss": 0.4553, "step": 6449 }, { "epoch": 0.32, "learning_rate": 1.5850080999247093e-05, "loss": 0.4108, "step": 6450 }, { "epoch": 0.32, "learning_rate": 1.5848769320839504e-05, "loss": 0.3317, "step": 6451 }, { "epoch": 0.32, "learning_rate": 1.584745748946501e-05, "loss": 0.351, "step": 6452 }, { "epoch": 0.32, "learning_rate": 1.5846145505157928e-05, "loss": 0.3114, "step": 6453 }, { "epoch": 0.32, "learning_rate": 1.5844833367952564e-05, "loss": 0.7137, "step": 6454 }, { "epoch": 0.32, "learning_rate": 1.5843521077883242e-05, "loss": 0.3521, "step": 6455 }, { "epoch": 0.32, "learning_rate": 1.5842208634984277e-05, "loss": 0.2923, "step": 6456 }, { "epoch": 0.32, "learning_rate": 1.5840896039289998e-05, "loss": 0.5552, "step": 6457 }, { "epoch": 0.32, "learning_rate": 1.5839583290834736e-05, "loss": 0.1896, "step": 6458 }, { "epoch": 0.32, "learning_rate": 1.5838270389652813e-05, "loss": 0.3811, "step": 6459 }, { "epoch": 0.32, "learning_rate": 1.5836957335778582e-05, "loss": 0.3767, "step": 6460 }, { "epoch": 0.32, "learning_rate": 1.5835644129246377e-05, "loss": 0.2937, "step": 6461 }, { "epoch": 0.32, "learning_rate": 1.583433077009054e-05, "loss": 0.3843, "step": 6462 }, { "epoch": 0.32, "learning_rate": 1.5833017258345427e-05, "loss": 0.424, "step": 6463 }, { "epoch": 0.32, "learning_rate": 1.5831703594045383e-05, "loss": 0.2052, "step": 6464 }, { "epoch": 0.32, "learning_rate": 1.5830389777224774e-05, "loss": 0.4293, "step": 6465 }, { "epoch": 0.32, "learning_rate": 1.5829075807917954e-05, "loss": 0.516, "step": 6466 }, { "epoch": 0.32, "learning_rate": 1.5827761686159294e-05, "loss": 0.3096, "step": 6467 }, { "epoch": 0.32, "learning_rate": 1.5826447411983154e-05, "loss": 0.3187, "step": 6468 }, { "epoch": 0.32, "learning_rate": 1.582513298542392e-05, "loss": 0.3315, "step": 6469 }, { "epoch": 0.32, "learning_rate": 1.5823818406515962e-05, "loss": 0.4134, "step": 6470 }, { "epoch": 0.32, "learning_rate": 1.5822503675293654e-05, "loss": 0.4391, "step": 6471 }, { "epoch": 0.32, "learning_rate": 1.5821188791791394e-05, "loss": 0.5523, "step": 6472 }, { "epoch": 0.32, "learning_rate": 1.5819873756043567e-05, "loss": 0.3919, "step": 6473 }, { "epoch": 0.32, "learning_rate": 1.5818558568084564e-05, "loss": 0.3928, "step": 6474 }, { "epoch": 0.32, "learning_rate": 1.581724322794878e-05, "loss": 0.3349, "step": 6475 }, { "epoch": 0.32, "learning_rate": 1.5815927735670618e-05, "loss": 0.6186, "step": 6476 }, { "epoch": 0.32, "learning_rate": 1.5814612091284485e-05, "loss": 0.4018, "step": 6477 }, { "epoch": 0.32, "learning_rate": 1.581329629482479e-05, "loss": 0.3418, "step": 6478 }, { "epoch": 0.32, "learning_rate": 1.5811980346325943e-05, "loss": 0.4292, "step": 6479 }, { "epoch": 0.32, "learning_rate": 1.5810664245822364e-05, "loss": 0.3005, "step": 6480 }, { "epoch": 0.32, "learning_rate": 1.580934799334847e-05, "loss": 0.431, "step": 6481 }, { "epoch": 0.32, "learning_rate": 1.5808031588938688e-05, "loss": 0.4742, "step": 6482 }, { "epoch": 0.32, "learning_rate": 1.5806715032627448e-05, "loss": 0.2584, "step": 6483 }, { "epoch": 0.32, "learning_rate": 1.5805398324449178e-05, "loss": 0.482, "step": 6484 }, { "epoch": 0.32, "learning_rate": 1.5804081464438324e-05, "loss": 0.538, "step": 6485 }, { "epoch": 0.32, "learning_rate": 1.5802764452629323e-05, "loss": 0.3992, "step": 6486 }, { "epoch": 0.32, "learning_rate": 1.5801447289056613e-05, "loss": 0.4338, "step": 6487 }, { "epoch": 0.32, "learning_rate": 1.580012997375465e-05, "loss": 0.309, "step": 6488 }, { "epoch": 0.32, "learning_rate": 1.5798812506757882e-05, "loss": 0.2782, "step": 6489 }, { "epoch": 0.32, "learning_rate": 1.579749488810077e-05, "loss": 0.3611, "step": 6490 }, { "epoch": 0.32, "learning_rate": 1.5796177117817773e-05, "loss": 0.2206, "step": 6491 }, { "epoch": 0.32, "learning_rate": 1.5794859195943354e-05, "loss": 0.3297, "step": 6492 }, { "epoch": 0.32, "learning_rate": 1.5793541122511982e-05, "loss": 0.5357, "step": 6493 }, { "epoch": 0.32, "learning_rate": 1.5792222897558132e-05, "loss": 0.3045, "step": 6494 }, { "epoch": 0.32, "learning_rate": 1.5790904521116278e-05, "loss": 0.514, "step": 6495 }, { "epoch": 0.32, "learning_rate": 1.57895859932209e-05, "loss": 0.352, "step": 6496 }, { "epoch": 0.32, "learning_rate": 1.5788267313906482e-05, "loss": 0.4641, "step": 6497 }, { "epoch": 0.32, "learning_rate": 1.5786948483207517e-05, "loss": 0.3859, "step": 6498 }, { "epoch": 0.32, "learning_rate": 1.578562950115849e-05, "loss": 0.246, "step": 6499 }, { "epoch": 0.32, "learning_rate": 1.5784310367793904e-05, "loss": 0.373, "step": 6500 }, { "epoch": 0.32, "learning_rate": 1.5782991083148257e-05, "loss": 0.2231, "step": 6501 }, { "epoch": 0.32, "learning_rate": 1.5781671647256055e-05, "loss": 0.2937, "step": 6502 }, { "epoch": 0.32, "learning_rate": 1.5780352060151795e-05, "loss": 0.7356, "step": 6503 }, { "epoch": 0.32, "learning_rate": 1.5779032321870006e-05, "loss": 0.4047, "step": 6504 }, { "epoch": 0.32, "learning_rate": 1.5777712432445193e-05, "loss": 0.356, "step": 6505 }, { "epoch": 0.32, "learning_rate": 1.577639239191188e-05, "loss": 0.1726, "step": 6506 }, { "epoch": 0.32, "learning_rate": 1.5775072200304587e-05, "loss": 0.3449, "step": 6507 }, { "epoch": 0.32, "learning_rate": 1.5773751857657848e-05, "loss": 0.5959, "step": 6508 }, { "epoch": 0.33, "learning_rate": 1.5772431364006188e-05, "loss": 0.2288, "step": 6509 }, { "epoch": 0.33, "learning_rate": 1.5771110719384146e-05, "loss": 0.4968, "step": 6510 }, { "epoch": 0.33, "learning_rate": 1.5769789923826265e-05, "loss": 0.3389, "step": 6511 }, { "epoch": 0.33, "learning_rate": 1.576846897736708e-05, "loss": 0.2244, "step": 6512 }, { "epoch": 0.33, "learning_rate": 1.5767147880041152e-05, "loss": 0.3573, "step": 6513 }, { "epoch": 0.33, "learning_rate": 1.576582663188302e-05, "loss": 0.5291, "step": 6514 }, { "epoch": 0.33, "learning_rate": 1.5764505232927245e-05, "loss": 0.6726, "step": 6515 }, { "epoch": 0.33, "learning_rate": 1.5763183683208385e-05, "loss": 0.6933, "step": 6516 }, { "epoch": 0.33, "learning_rate": 1.5761861982761008e-05, "loss": 0.4614, "step": 6517 }, { "epoch": 0.33, "learning_rate": 1.5760540131619676e-05, "loss": 0.4189, "step": 6518 }, { "epoch": 0.33, "learning_rate": 1.575921812981896e-05, "loss": 0.5526, "step": 6519 }, { "epoch": 0.33, "learning_rate": 1.5757895977393435e-05, "loss": 0.3056, "step": 6520 }, { "epoch": 0.33, "learning_rate": 1.5756573674377685e-05, "loss": 0.5126, "step": 6521 }, { "epoch": 0.33, "learning_rate": 1.575525122080629e-05, "loss": 0.4241, "step": 6522 }, { "epoch": 0.33, "learning_rate": 1.575392861671384e-05, "loss": 0.2187, "step": 6523 }, { "epoch": 0.33, "learning_rate": 1.575260586213492e-05, "loss": 0.5001, "step": 6524 }, { "epoch": 0.33, "learning_rate": 1.5751282957104127e-05, "loss": 0.4191, "step": 6525 }, { "epoch": 0.33, "learning_rate": 1.5749959901656064e-05, "loss": 0.5683, "step": 6526 }, { "epoch": 0.33, "learning_rate": 1.5748636695825334e-05, "loss": 0.4011, "step": 6527 }, { "epoch": 0.33, "learning_rate": 1.5747313339646534e-05, "loss": 0.3857, "step": 6528 }, { "epoch": 0.33, "learning_rate": 1.5745989833154283e-05, "loss": 0.3309, "step": 6529 }, { "epoch": 0.33, "learning_rate": 1.5744666176383196e-05, "loss": 0.5122, "step": 6530 }, { "epoch": 0.33, "learning_rate": 1.574334236936789e-05, "loss": 0.1931, "step": 6531 }, { "epoch": 0.33, "learning_rate": 1.5742018412142985e-05, "loss": 0.4345, "step": 6532 }, { "epoch": 0.33, "learning_rate": 1.574069430474311e-05, "loss": 0.1554, "step": 6533 }, { "epoch": 0.33, "learning_rate": 1.573937004720289e-05, "loss": 0.3445, "step": 6534 }, { "epoch": 0.33, "learning_rate": 1.5738045639556967e-05, "loss": 0.2156, "step": 6535 }, { "epoch": 0.33, "learning_rate": 1.5736721081839976e-05, "loss": 0.3261, "step": 6536 }, { "epoch": 0.33, "learning_rate": 1.573539637408656e-05, "loss": 0.6386, "step": 6537 }, { "epoch": 0.33, "learning_rate": 1.573407151633136e-05, "loss": 0.4641, "step": 6538 }, { "epoch": 0.33, "learning_rate": 1.5732746508609033e-05, "loss": 0.5194, "step": 6539 }, { "epoch": 0.33, "learning_rate": 1.5731421350954228e-05, "loss": 0.33, "step": 6540 }, { "epoch": 0.33, "learning_rate": 1.5730096043401605e-05, "loss": 0.5346, "step": 6541 }, { "epoch": 0.33, "learning_rate": 1.5728770585985825e-05, "loss": 0.4375, "step": 6542 }, { "epoch": 0.33, "learning_rate": 1.5727444978741554e-05, "loss": 0.2717, "step": 6543 }, { "epoch": 0.33, "learning_rate": 1.572611922170346e-05, "loss": 0.5478, "step": 6544 }, { "epoch": 0.33, "learning_rate": 1.5724793314906223e-05, "loss": 0.4542, "step": 6545 }, { "epoch": 0.33, "learning_rate": 1.572346725838451e-05, "loss": 0.3887, "step": 6546 }, { "epoch": 0.33, "learning_rate": 1.5722141052173007e-05, "loss": 0.2501, "step": 6547 }, { "epoch": 0.33, "learning_rate": 1.57208146963064e-05, "loss": 0.4436, "step": 6548 }, { "epoch": 0.33, "learning_rate": 1.5719488190819377e-05, "loss": 0.5789, "step": 6549 }, { "epoch": 0.33, "learning_rate": 1.5718161535746633e-05, "loss": 0.2757, "step": 6550 }, { "epoch": 0.33, "learning_rate": 1.5716834731122866e-05, "loss": 0.2932, "step": 6551 }, { "epoch": 0.33, "learning_rate": 1.571550777698277e-05, "loss": 0.2446, "step": 6552 }, { "epoch": 0.33, "learning_rate": 1.5714180673361054e-05, "loss": 0.1477, "step": 6553 }, { "epoch": 0.33, "learning_rate": 1.571285342029243e-05, "loss": 0.3929, "step": 6554 }, { "epoch": 0.33, "learning_rate": 1.5711526017811603e-05, "loss": 0.1517, "step": 6555 }, { "epoch": 0.33, "learning_rate": 1.5710198465953297e-05, "loss": 0.306, "step": 6556 }, { "epoch": 0.33, "learning_rate": 1.5708870764752227e-05, "loss": 0.2661, "step": 6557 }, { "epoch": 0.33, "learning_rate": 1.5707542914243124e-05, "loss": 0.4181, "step": 6558 }, { "epoch": 0.33, "learning_rate": 1.5706214914460707e-05, "loss": 0.3142, "step": 6559 }, { "epoch": 0.33, "learning_rate": 1.5704886765439715e-05, "loss": 0.4774, "step": 6560 }, { "epoch": 0.33, "learning_rate": 1.570355846721488e-05, "loss": 0.3867, "step": 6561 }, { "epoch": 0.33, "learning_rate": 1.5702230019820946e-05, "loss": 0.3123, "step": 6562 }, { "epoch": 0.33, "learning_rate": 1.570090142329265e-05, "loss": 0.5192, "step": 6563 }, { "epoch": 0.33, "learning_rate": 1.569957267766475e-05, "loss": 0.1981, "step": 6564 }, { "epoch": 0.33, "learning_rate": 1.5698243782971987e-05, "loss": 0.335, "step": 6565 }, { "epoch": 0.33, "learning_rate": 1.5696914739249123e-05, "loss": 0.3016, "step": 6566 }, { "epoch": 0.33, "learning_rate": 1.5695585546530914e-05, "loss": 0.2287, "step": 6567 }, { "epoch": 0.33, "learning_rate": 1.569425620485213e-05, "loss": 0.4052, "step": 6568 }, { "epoch": 0.33, "learning_rate": 1.5692926714247527e-05, "loss": 0.3609, "step": 6569 }, { "epoch": 0.33, "learning_rate": 1.5691597074751883e-05, "loss": 0.2872, "step": 6570 }, { "epoch": 0.33, "learning_rate": 1.5690267286399974e-05, "loss": 0.425, "step": 6571 }, { "epoch": 0.33, "learning_rate": 1.5688937349226578e-05, "loss": 0.4565, "step": 6572 }, { "epoch": 0.33, "learning_rate": 1.5687607263266474e-05, "loss": 0.2779, "step": 6573 }, { "epoch": 0.33, "learning_rate": 1.5686277028554456e-05, "loss": 0.3785, "step": 6574 }, { "epoch": 0.33, "learning_rate": 1.5684946645125306e-05, "loss": 0.4046, "step": 6575 }, { "epoch": 0.33, "learning_rate": 1.5683616113013823e-05, "loss": 0.287, "step": 6576 }, { "epoch": 0.33, "learning_rate": 1.5682285432254807e-05, "loss": 0.3954, "step": 6577 }, { "epoch": 0.33, "learning_rate": 1.5680954602883054e-05, "loss": 0.3582, "step": 6578 }, { "epoch": 0.33, "learning_rate": 1.5679623624933378e-05, "loss": 0.435, "step": 6579 }, { "epoch": 0.33, "learning_rate": 1.567829249844058e-05, "loss": 0.3333, "step": 6580 }, { "epoch": 0.33, "learning_rate": 1.5676961223439488e-05, "loss": 0.0798, "step": 6581 }, { "epoch": 0.33, "learning_rate": 1.5675629799964904e-05, "loss": 0.4856, "step": 6582 }, { "epoch": 0.33, "learning_rate": 1.5674298228051655e-05, "loss": 0.2799, "step": 6583 }, { "epoch": 0.33, "learning_rate": 1.5672966507734572e-05, "loss": 0.337, "step": 6584 }, { "epoch": 0.33, "learning_rate": 1.567163463904848e-05, "loss": 0.3402, "step": 6585 }, { "epoch": 0.33, "learning_rate": 1.567030262202821e-05, "loss": 0.2601, "step": 6586 }, { "epoch": 0.33, "learning_rate": 1.56689704567086e-05, "loss": 0.4127, "step": 6587 }, { "epoch": 0.33, "learning_rate": 1.5667638143124495e-05, "loss": 0.3487, "step": 6588 }, { "epoch": 0.33, "learning_rate": 1.5666305681310737e-05, "loss": 0.3264, "step": 6589 }, { "epoch": 0.33, "learning_rate": 1.566497307130218e-05, "loss": 0.5078, "step": 6590 }, { "epoch": 0.33, "learning_rate": 1.5663640313133667e-05, "loss": 0.4464, "step": 6591 }, { "epoch": 0.33, "learning_rate": 1.5662307406840062e-05, "loss": 0.4784, "step": 6592 }, { "epoch": 0.33, "learning_rate": 1.566097435245622e-05, "loss": 0.3435, "step": 6593 }, { "epoch": 0.33, "learning_rate": 1.565964115001701e-05, "loss": 0.516, "step": 6594 }, { "epoch": 0.33, "learning_rate": 1.56583077995573e-05, "loss": 0.4624, "step": 6595 }, { "epoch": 0.33, "learning_rate": 1.565697430111196e-05, "loss": 0.4211, "step": 6596 }, { "epoch": 0.33, "learning_rate": 1.5655640654715867e-05, "loss": 0.3455, "step": 6597 }, { "epoch": 0.33, "learning_rate": 1.5654306860403898e-05, "loss": 0.2232, "step": 6598 }, { "epoch": 0.33, "learning_rate": 1.5652972918210943e-05, "loss": 0.4266, "step": 6599 }, { "epoch": 0.33, "learning_rate": 1.565163882817188e-05, "loss": 0.3007, "step": 6600 }, { "epoch": 0.33, "learning_rate": 1.565030459032161e-05, "loss": 0.4522, "step": 6601 }, { "epoch": 0.33, "learning_rate": 1.5648970204695022e-05, "loss": 0.3926, "step": 6602 }, { "epoch": 0.33, "learning_rate": 1.5647635671327016e-05, "loss": 0.3001, "step": 6603 }, { "epoch": 0.33, "learning_rate": 1.56463009902525e-05, "loss": 0.5073, "step": 6604 }, { "epoch": 0.33, "learning_rate": 1.5644966161506375e-05, "loss": 0.5745, "step": 6605 }, { "epoch": 0.33, "learning_rate": 1.5643631185123552e-05, "loss": 0.3404, "step": 6606 }, { "epoch": 0.33, "learning_rate": 1.564229606113895e-05, "loss": 0.5077, "step": 6607 }, { "epoch": 0.33, "learning_rate": 1.5640960789587483e-05, "loss": 0.3553, "step": 6608 }, { "epoch": 0.33, "learning_rate": 1.5639625370504073e-05, "loss": 0.2985, "step": 6609 }, { "epoch": 0.33, "learning_rate": 1.5638289803923648e-05, "loss": 0.5204, "step": 6610 }, { "epoch": 0.33, "learning_rate": 1.5636954089881143e-05, "loss": 0.3685, "step": 6611 }, { "epoch": 0.33, "learning_rate": 1.5635618228411483e-05, "loss": 0.5569, "step": 6612 }, { "epoch": 0.33, "learning_rate": 1.563428221954961e-05, "loss": 0.532, "step": 6613 }, { "epoch": 0.33, "learning_rate": 1.5632946063330465e-05, "loss": 0.1253, "step": 6614 }, { "epoch": 0.33, "learning_rate": 1.5631609759788993e-05, "loss": 0.709, "step": 6615 }, { "epoch": 0.33, "learning_rate": 1.5630273308960148e-05, "loss": 0.5101, "step": 6616 }, { "epoch": 0.33, "learning_rate": 1.5628936710878876e-05, "loss": 0.3811, "step": 6617 }, { "epoch": 0.33, "learning_rate": 1.5627599965580134e-05, "loss": 0.3823, "step": 6618 }, { "epoch": 0.33, "learning_rate": 1.5626263073098888e-05, "loss": 0.4313, "step": 6619 }, { "epoch": 0.33, "learning_rate": 1.56249260334701e-05, "loss": 0.3777, "step": 6620 }, { "epoch": 0.33, "learning_rate": 1.5623588846728738e-05, "loss": 0.2634, "step": 6621 }, { "epoch": 0.33, "learning_rate": 1.562225151290978e-05, "loss": 0.4519, "step": 6622 }, { "epoch": 0.33, "learning_rate": 1.5620914032048196e-05, "loss": 0.4017, "step": 6623 }, { "epoch": 0.33, "learning_rate": 1.5619576404178965e-05, "loss": 0.4534, "step": 6624 }, { "epoch": 0.33, "learning_rate": 1.5618238629337073e-05, "loss": 0.6275, "step": 6625 }, { "epoch": 0.33, "learning_rate": 1.5616900707557513e-05, "loss": 0.4536, "step": 6626 }, { "epoch": 0.33, "learning_rate": 1.5615562638875267e-05, "loss": 0.4048, "step": 6627 }, { "epoch": 0.33, "learning_rate": 1.5614224423325338e-05, "loss": 0.449, "step": 6628 }, { "epoch": 0.33, "learning_rate": 1.561288606094272e-05, "loss": 0.59, "step": 6629 }, { "epoch": 0.33, "learning_rate": 1.561154755176242e-05, "loss": 0.2498, "step": 6630 }, { "epoch": 0.33, "learning_rate": 1.5610208895819443e-05, "loss": 0.5311, "step": 6631 }, { "epoch": 0.33, "learning_rate": 1.5608870093148803e-05, "loss": 0.4122, "step": 6632 }, { "epoch": 0.33, "learning_rate": 1.5607531143785507e-05, "loss": 0.3886, "step": 6633 }, { "epoch": 0.33, "learning_rate": 1.5606192047764582e-05, "loss": 0.3611, "step": 6634 }, { "epoch": 0.33, "learning_rate": 1.5604852805121048e-05, "loss": 0.3943, "step": 6635 }, { "epoch": 0.33, "learning_rate": 1.5603513415889927e-05, "loss": 0.3659, "step": 6636 }, { "epoch": 0.33, "learning_rate": 1.5602173880106254e-05, "loss": 0.4416, "step": 6637 }, { "epoch": 0.33, "learning_rate": 1.5600834197805057e-05, "loss": 0.3715, "step": 6638 }, { "epoch": 0.33, "learning_rate": 1.559949436902138e-05, "loss": 0.2432, "step": 6639 }, { "epoch": 0.33, "learning_rate": 1.5598154393790262e-05, "loss": 0.2787, "step": 6640 }, { "epoch": 0.33, "learning_rate": 1.559681427214675e-05, "loss": 0.3258, "step": 6641 }, { "epoch": 0.33, "learning_rate": 1.5595474004125886e-05, "loss": 0.4131, "step": 6642 }, { "epoch": 0.33, "learning_rate": 1.5594133589762733e-05, "loss": 0.2844, "step": 6643 }, { "epoch": 0.33, "learning_rate": 1.559279302909234e-05, "loss": 0.275, "step": 6644 }, { "epoch": 0.33, "learning_rate": 1.5591452322149774e-05, "loss": 0.2066, "step": 6645 }, { "epoch": 0.33, "learning_rate": 1.5590111468970093e-05, "loss": 0.3452, "step": 6646 }, { "epoch": 0.33, "learning_rate": 1.5588770469588368e-05, "loss": 0.6433, "step": 6647 }, { "epoch": 0.33, "learning_rate": 1.5587429324039676e-05, "loss": 0.2759, "step": 6648 }, { "epoch": 0.33, "learning_rate": 1.558608803235908e-05, "loss": 0.2322, "step": 6649 }, { "epoch": 0.33, "learning_rate": 1.5584746594581678e-05, "loss": 0.4171, "step": 6650 }, { "epoch": 0.33, "learning_rate": 1.5583405010742537e-05, "loss": 0.2422, "step": 6651 }, { "epoch": 0.33, "learning_rate": 1.5582063280876752e-05, "loss": 0.4187, "step": 6652 }, { "epoch": 0.33, "learning_rate": 1.558072140501941e-05, "loss": 0.3681, "step": 6653 }, { "epoch": 0.33, "learning_rate": 1.5579379383205614e-05, "loss": 0.3298, "step": 6654 }, { "epoch": 0.33, "learning_rate": 1.5578037215470457e-05, "loss": 0.4241, "step": 6655 }, { "epoch": 0.33, "learning_rate": 1.557669490184904e-05, "loss": 0.3412, "step": 6656 }, { "epoch": 0.33, "learning_rate": 1.5575352442376474e-05, "loss": 0.4252, "step": 6657 }, { "epoch": 0.33, "learning_rate": 1.5574009837087867e-05, "loss": 0.4227, "step": 6658 }, { "epoch": 0.33, "learning_rate": 1.5572667086018338e-05, "loss": 0.5089, "step": 6659 }, { "epoch": 0.33, "learning_rate": 1.5571324189202993e-05, "loss": 0.3174, "step": 6660 }, { "epoch": 0.33, "learning_rate": 1.556998114667696e-05, "loss": 0.4654, "step": 6661 }, { "epoch": 0.33, "learning_rate": 1.556863795847537e-05, "loss": 0.3853, "step": 6662 }, { "epoch": 0.33, "learning_rate": 1.5567294624633347e-05, "loss": 0.4176, "step": 6663 }, { "epoch": 0.33, "learning_rate": 1.5565951145186027e-05, "loss": 0.2635, "step": 6664 }, { "epoch": 0.33, "learning_rate": 1.5564607520168543e-05, "loss": 0.359, "step": 6665 }, { "epoch": 0.33, "learning_rate": 1.5563263749616036e-05, "loss": 0.4366, "step": 6666 }, { "epoch": 0.33, "learning_rate": 1.5561919833563652e-05, "loss": 0.4352, "step": 6667 }, { "epoch": 0.33, "learning_rate": 1.5560575772046543e-05, "loss": 0.1482, "step": 6668 }, { "epoch": 0.33, "learning_rate": 1.5559231565099853e-05, "loss": 0.3472, "step": 6669 }, { "epoch": 0.33, "learning_rate": 1.555788721275875e-05, "loss": 0.5169, "step": 6670 }, { "epoch": 0.33, "learning_rate": 1.555654271505838e-05, "loss": 0.4966, "step": 6671 }, { "epoch": 0.33, "learning_rate": 1.5555198072033912e-05, "loss": 0.311, "step": 6672 }, { "epoch": 0.33, "learning_rate": 1.555385328372052e-05, "loss": 0.4008, "step": 6673 }, { "epoch": 0.33, "learning_rate": 1.5552508350153367e-05, "loss": 0.3609, "step": 6674 }, { "epoch": 0.33, "learning_rate": 1.555116327136763e-05, "loss": 0.1605, "step": 6675 }, { "epoch": 0.33, "learning_rate": 1.5549818047398487e-05, "loss": 0.5309, "step": 6676 }, { "epoch": 0.33, "learning_rate": 1.5548472678281126e-05, "loss": 0.1794, "step": 6677 }, { "epoch": 0.33, "learning_rate": 1.5547127164050725e-05, "loss": 0.425, "step": 6678 }, { "epoch": 0.33, "learning_rate": 1.5545781504742482e-05, "loss": 0.6444, "step": 6679 }, { "epoch": 0.33, "learning_rate": 1.5544435700391587e-05, "loss": 0.2743, "step": 6680 }, { "epoch": 0.33, "learning_rate": 1.554308975103324e-05, "loss": 0.4289, "step": 6681 }, { "epoch": 0.33, "learning_rate": 1.5541743656702636e-05, "loss": 0.3919, "step": 6682 }, { "epoch": 0.33, "learning_rate": 1.5540397417434983e-05, "loss": 0.2508, "step": 6683 }, { "epoch": 0.33, "learning_rate": 1.55390510332655e-05, "loss": 0.1284, "step": 6684 }, { "epoch": 0.33, "learning_rate": 1.5537704504229386e-05, "loss": 0.3743, "step": 6685 }, { "epoch": 0.33, "learning_rate": 1.5536357830361867e-05, "loss": 0.3252, "step": 6686 }, { "epoch": 0.33, "learning_rate": 1.553501101169816e-05, "loss": 0.387, "step": 6687 }, { "epoch": 0.33, "learning_rate": 1.5533664048273487e-05, "loss": 0.3377, "step": 6688 }, { "epoch": 0.33, "learning_rate": 1.5532316940123082e-05, "loss": 0.3244, "step": 6689 }, { "epoch": 0.33, "learning_rate": 1.553096968728217e-05, "loss": 0.5605, "step": 6690 }, { "epoch": 0.33, "learning_rate": 1.5529622289785992e-05, "loss": 0.2677, "step": 6691 }, { "epoch": 0.33, "learning_rate": 1.5528274747669786e-05, "loss": 0.4413, "step": 6692 }, { "epoch": 0.33, "learning_rate": 1.5526927060968794e-05, "loss": 0.4474, "step": 6693 }, { "epoch": 0.33, "learning_rate": 1.5525579229718264e-05, "loss": 0.2684, "step": 6694 }, { "epoch": 0.33, "learning_rate": 1.5524231253953445e-05, "loss": 0.6085, "step": 6695 }, { "epoch": 0.33, "learning_rate": 1.55228831337096e-05, "loss": 0.3373, "step": 6696 }, { "epoch": 0.33, "learning_rate": 1.552153486902197e-05, "loss": 0.4802, "step": 6697 }, { "epoch": 0.33, "learning_rate": 1.5520186459925835e-05, "loss": 0.4132, "step": 6698 }, { "epoch": 0.33, "learning_rate": 1.5518837906456448e-05, "loss": 0.1919, "step": 6699 }, { "epoch": 0.33, "learning_rate": 1.551748920864909e-05, "loss": 0.3892, "step": 6700 }, { "epoch": 0.33, "learning_rate": 1.5516140366539022e-05, "loss": 0.4545, "step": 6701 }, { "epoch": 0.33, "learning_rate": 1.551479138016153e-05, "loss": 0.3076, "step": 6702 }, { "epoch": 0.33, "learning_rate": 1.5513442249551893e-05, "loss": 0.4299, "step": 6703 }, { "epoch": 0.33, "learning_rate": 1.5512092974745395e-05, "loss": 0.2717, "step": 6704 }, { "epoch": 0.33, "learning_rate": 1.5510743555777326e-05, "loss": 0.3233, "step": 6705 }, { "epoch": 0.33, "learning_rate": 1.550939399268297e-05, "loss": 0.4967, "step": 6706 }, { "epoch": 0.33, "learning_rate": 1.550804428549764e-05, "loss": 0.4425, "step": 6707 }, { "epoch": 0.33, "learning_rate": 1.5506694434256623e-05, "loss": 0.2994, "step": 6708 }, { "epoch": 0.33, "learning_rate": 1.5505344438995224e-05, "loss": 0.452, "step": 6709 }, { "epoch": 0.34, "learning_rate": 1.550399429974875e-05, "loss": 0.3215, "step": 6710 }, { "epoch": 0.34, "learning_rate": 1.5502644016552518e-05, "loss": 0.3493, "step": 6711 }, { "epoch": 0.34, "learning_rate": 1.5501293589441837e-05, "loss": 0.2304, "step": 6712 }, { "epoch": 0.34, "learning_rate": 1.5499943018452028e-05, "loss": 0.3897, "step": 6713 }, { "epoch": 0.34, "learning_rate": 1.5498592303618414e-05, "loss": 0.3504, "step": 6714 }, { "epoch": 0.34, "learning_rate": 1.5497241444976318e-05, "loss": 0.4866, "step": 6715 }, { "epoch": 0.34, "learning_rate": 1.5495890442561075e-05, "loss": 0.469, "step": 6716 }, { "epoch": 0.34, "learning_rate": 1.549453929640801e-05, "loss": 0.2716, "step": 6717 }, { "epoch": 0.34, "learning_rate": 1.5493188006552474e-05, "loss": 0.3556, "step": 6718 }, { "epoch": 0.34, "learning_rate": 1.5491836573029794e-05, "loss": 0.3555, "step": 6719 }, { "epoch": 0.34, "learning_rate": 1.5490484995875325e-05, "loss": 0.4702, "step": 6720 }, { "epoch": 0.34, "learning_rate": 1.548913327512441e-05, "loss": 0.6291, "step": 6721 }, { "epoch": 0.34, "learning_rate": 1.5487781410812405e-05, "loss": 0.2324, "step": 6722 }, { "epoch": 0.34, "learning_rate": 1.548642940297467e-05, "loss": 0.2338, "step": 6723 }, { "epoch": 0.34, "learning_rate": 1.548507725164655e-05, "loss": 0.3905, "step": 6724 }, { "epoch": 0.34, "learning_rate": 1.5483724956863425e-05, "loss": 0.4508, "step": 6725 }, { "epoch": 0.34, "learning_rate": 1.548237251866065e-05, "loss": 0.347, "step": 6726 }, { "epoch": 0.34, "learning_rate": 1.5481019937073612e-05, "loss": 0.3844, "step": 6727 }, { "epoch": 0.34, "learning_rate": 1.5479667212137674e-05, "loss": 0.4374, "step": 6728 }, { "epoch": 0.34, "learning_rate": 1.547831434388821e-05, "loss": 0.148, "step": 6729 }, { "epoch": 0.34, "learning_rate": 1.5476961332360615e-05, "loss": 0.4926, "step": 6730 }, { "epoch": 0.34, "learning_rate": 1.547560817759027e-05, "loss": 0.5995, "step": 6731 }, { "epoch": 0.34, "learning_rate": 1.5474254879612565e-05, "loss": 0.242, "step": 6732 }, { "epoch": 0.34, "learning_rate": 1.547290143846289e-05, "loss": 0.2943, "step": 6733 }, { "epoch": 0.34, "learning_rate": 1.5471547854176648e-05, "loss": 0.2982, "step": 6734 }, { "epoch": 0.34, "learning_rate": 1.547019412678924e-05, "loss": 0.3272, "step": 6735 }, { "epoch": 0.34, "learning_rate": 1.546884025633607e-05, "loss": 0.7193, "step": 6736 }, { "epoch": 0.34, "learning_rate": 1.5467486242852544e-05, "loss": 0.2636, "step": 6737 }, { "epoch": 0.34, "learning_rate": 1.5466132086374076e-05, "loss": 0.2835, "step": 6738 }, { "epoch": 0.34, "learning_rate": 1.5464777786936083e-05, "loss": 0.4288, "step": 6739 }, { "epoch": 0.34, "learning_rate": 1.5463423344573983e-05, "loss": 0.1846, "step": 6740 }, { "epoch": 0.34, "learning_rate": 1.5462068759323202e-05, "loss": 0.3701, "step": 6741 }, { "epoch": 0.34, "learning_rate": 1.546071403121917e-05, "loss": 0.5548, "step": 6742 }, { "epoch": 0.34, "learning_rate": 1.545935916029731e-05, "loss": 0.2529, "step": 6743 }, { "epoch": 0.34, "learning_rate": 1.5458004146593066e-05, "loss": 0.3305, "step": 6744 }, { "epoch": 0.34, "learning_rate": 1.545664899014187e-05, "loss": 0.4728, "step": 6745 }, { "epoch": 0.34, "learning_rate": 1.5455293690979167e-05, "loss": 0.4446, "step": 6746 }, { "epoch": 0.34, "learning_rate": 1.54539382491404e-05, "loss": 0.4027, "step": 6747 }, { "epoch": 0.34, "learning_rate": 1.5452582664661027e-05, "loss": 0.4281, "step": 6748 }, { "epoch": 0.34, "learning_rate": 1.5451226937576493e-05, "loss": 0.3266, "step": 6749 }, { "epoch": 0.34, "learning_rate": 1.5449871067922257e-05, "loss": 0.3913, "step": 6750 }, { "epoch": 0.34, "learning_rate": 1.5448515055733783e-05, "loss": 0.4125, "step": 6751 }, { "epoch": 0.34, "learning_rate": 1.5447158901046533e-05, "loss": 0.3086, "step": 6752 }, { "epoch": 0.34, "learning_rate": 1.5445802603895978e-05, "loss": 0.5609, "step": 6753 }, { "epoch": 0.34, "learning_rate": 1.5444446164317585e-05, "loss": 0.326, "step": 6754 }, { "epoch": 0.34, "learning_rate": 1.5443089582346838e-05, "loss": 0.2716, "step": 6755 }, { "epoch": 0.34, "learning_rate": 1.544173285801921e-05, "loss": 0.3529, "step": 6756 }, { "epoch": 0.34, "learning_rate": 1.5440375991370188e-05, "loss": 0.3976, "step": 6757 }, { "epoch": 0.34, "learning_rate": 1.5439018982435255e-05, "loss": 0.1958, "step": 6758 }, { "epoch": 0.34, "learning_rate": 1.5437661831249907e-05, "loss": 0.5551, "step": 6759 }, { "epoch": 0.34, "learning_rate": 1.5436304537849632e-05, "loss": 0.2665, "step": 6760 }, { "epoch": 0.34, "learning_rate": 1.5434947102269937e-05, "loss": 0.5054, "step": 6761 }, { "epoch": 0.34, "learning_rate": 1.543358952454632e-05, "loss": 0.3, "step": 6762 }, { "epoch": 0.34, "learning_rate": 1.5432231804714276e-05, "loss": 0.3519, "step": 6763 }, { "epoch": 0.34, "learning_rate": 1.5430873942809332e-05, "loss": 0.3484, "step": 6764 }, { "epoch": 0.34, "learning_rate": 1.5429515938866992e-05, "loss": 0.5343, "step": 6765 }, { "epoch": 0.34, "learning_rate": 1.5428157792922773e-05, "loss": 0.4554, "step": 6766 }, { "epoch": 0.34, "learning_rate": 1.5426799505012197e-05, "loss": 0.6054, "step": 6767 }, { "epoch": 0.34, "learning_rate": 1.5425441075170788e-05, "loss": 0.4456, "step": 6768 }, { "epoch": 0.34, "learning_rate": 1.5424082503434074e-05, "loss": 0.2577, "step": 6769 }, { "epoch": 0.34, "learning_rate": 1.5422723789837586e-05, "loss": 0.654, "step": 6770 }, { "epoch": 0.34, "learning_rate": 1.542136493441686e-05, "loss": 0.4398, "step": 6771 }, { "epoch": 0.34, "learning_rate": 1.5420005937207436e-05, "loss": 0.5416, "step": 6772 }, { "epoch": 0.34, "learning_rate": 1.5418646798244853e-05, "loss": 0.5147, "step": 6773 }, { "epoch": 0.34, "learning_rate": 1.541728751756466e-05, "loss": 0.5488, "step": 6774 }, { "epoch": 0.34, "learning_rate": 1.541592809520241e-05, "loss": 0.4651, "step": 6775 }, { "epoch": 0.34, "learning_rate": 1.5414568531193656e-05, "loss": 0.2304, "step": 6776 }, { "epoch": 0.34, "learning_rate": 1.541320882557395e-05, "loss": 0.448, "step": 6777 }, { "epoch": 0.34, "learning_rate": 1.541184897837886e-05, "loss": 0.2345, "step": 6778 }, { "epoch": 0.34, "learning_rate": 1.5410488989643942e-05, "loss": 0.8309, "step": 6779 }, { "epoch": 0.34, "learning_rate": 1.5409128859404776e-05, "loss": 0.3968, "step": 6780 }, { "epoch": 0.34, "learning_rate": 1.5407768587696928e-05, "loss": 0.5056, "step": 6781 }, { "epoch": 0.34, "learning_rate": 1.5406408174555978e-05, "loss": 0.7936, "step": 6782 }, { "epoch": 0.34, "learning_rate": 1.54050476200175e-05, "loss": 0.6146, "step": 6783 }, { "epoch": 0.34, "learning_rate": 1.5403686924117084e-05, "loss": 0.4045, "step": 6784 }, { "epoch": 0.34, "learning_rate": 1.5402326086890315e-05, "loss": 0.4095, "step": 6785 }, { "epoch": 0.34, "learning_rate": 1.5400965108372775e-05, "loss": 0.481, "step": 6786 }, { "epoch": 0.34, "learning_rate": 1.5399603988600077e-05, "loss": 0.5105, "step": 6787 }, { "epoch": 0.34, "learning_rate": 1.5398242727607803e-05, "loss": 0.4948, "step": 6788 }, { "epoch": 0.34, "learning_rate": 1.5396881325431562e-05, "loss": 0.3107, "step": 6789 }, { "epoch": 0.34, "learning_rate": 1.5395519782106954e-05, "loss": 0.2969, "step": 6790 }, { "epoch": 0.34, "learning_rate": 1.5394158097669598e-05, "loss": 0.2785, "step": 6791 }, { "epoch": 0.34, "learning_rate": 1.5392796272155103e-05, "loss": 0.4751, "step": 6792 }, { "epoch": 0.34, "learning_rate": 1.539143430559908e-05, "loss": 0.2492, "step": 6793 }, { "epoch": 0.34, "learning_rate": 1.539007219803716e-05, "loss": 0.5364, "step": 6794 }, { "epoch": 0.34, "learning_rate": 1.538870994950496e-05, "loss": 0.1606, "step": 6795 }, { "epoch": 0.34, "learning_rate": 1.5387347560038106e-05, "loss": 0.3564, "step": 6796 }, { "epoch": 0.34, "learning_rate": 1.5385985029672235e-05, "loss": 0.521, "step": 6797 }, { "epoch": 0.34, "learning_rate": 1.538462235844298e-05, "loss": 0.5039, "step": 6798 }, { "epoch": 0.34, "learning_rate": 1.538325954638598e-05, "loss": 0.3951, "step": 6799 }, { "epoch": 0.34, "learning_rate": 1.5381896593536876e-05, "loss": 0.1252, "step": 6800 }, { "epoch": 0.34, "learning_rate": 1.5380533499931317e-05, "loss": 0.561, "step": 6801 }, { "epoch": 0.34, "learning_rate": 1.537917026560495e-05, "loss": 0.287, "step": 6802 }, { "epoch": 0.34, "learning_rate": 1.5377806890593433e-05, "loss": 0.5473, "step": 6803 }, { "epoch": 0.34, "learning_rate": 1.5376443374932414e-05, "loss": 0.4459, "step": 6804 }, { "epoch": 0.34, "learning_rate": 1.5375079718657566e-05, "loss": 0.3708, "step": 6805 }, { "epoch": 0.34, "learning_rate": 1.537371592180455e-05, "loss": 0.2589, "step": 6806 }, { "epoch": 0.34, "learning_rate": 1.5372351984409028e-05, "loss": 0.4286, "step": 6807 }, { "epoch": 0.34, "learning_rate": 1.537098790650668e-05, "loss": 0.4562, "step": 6808 }, { "epoch": 0.34, "learning_rate": 1.5369623688133174e-05, "loss": 0.4124, "step": 6809 }, { "epoch": 0.34, "learning_rate": 1.5368259329324195e-05, "loss": 0.574, "step": 6810 }, { "epoch": 0.34, "learning_rate": 1.5366894830115423e-05, "loss": 0.4794, "step": 6811 }, { "epoch": 0.34, "learning_rate": 1.5365530190542548e-05, "loss": 0.3037, "step": 6812 }, { "epoch": 0.34, "learning_rate": 1.5364165410641256e-05, "loss": 0.4301, "step": 6813 }, { "epoch": 0.34, "learning_rate": 1.536280049044725e-05, "loss": 0.3816, "step": 6814 }, { "epoch": 0.34, "learning_rate": 1.5361435429996215e-05, "loss": 0.5657, "step": 6815 }, { "epoch": 0.34, "learning_rate": 1.536007022932386e-05, "loss": 0.4366, "step": 6816 }, { "epoch": 0.34, "learning_rate": 1.5358704888465886e-05, "loss": 0.6126, "step": 6817 }, { "epoch": 0.34, "learning_rate": 1.535733940745801e-05, "loss": 0.5193, "step": 6818 }, { "epoch": 0.34, "learning_rate": 1.5355973786335935e-05, "loss": 0.4563, "step": 6819 }, { "epoch": 0.34, "learning_rate": 1.5354608025135383e-05, "loss": 0.4358, "step": 6820 }, { "epoch": 0.34, "learning_rate": 1.535324212389207e-05, "loss": 0.4032, "step": 6821 }, { "epoch": 0.34, "learning_rate": 1.535187608264172e-05, "loss": 0.362, "step": 6822 }, { "epoch": 0.34, "learning_rate": 1.535050990142006e-05, "loss": 0.2982, "step": 6823 }, { "epoch": 0.34, "learning_rate": 1.5349143580262824e-05, "loss": 0.4425, "step": 6824 }, { "epoch": 0.34, "learning_rate": 1.534777711920574e-05, "loss": 0.2929, "step": 6825 }, { "epoch": 0.34, "learning_rate": 1.5346410518284556e-05, "loss": 0.4267, "step": 6826 }, { "epoch": 0.34, "learning_rate": 1.5345043777535002e-05, "loss": 0.4554, "step": 6827 }, { "epoch": 0.34, "learning_rate": 1.5343676896992832e-05, "loss": 0.3024, "step": 6828 }, { "epoch": 0.34, "learning_rate": 1.534230987669379e-05, "loss": 0.2154, "step": 6829 }, { "epoch": 0.34, "learning_rate": 1.5340942716673627e-05, "loss": 0.4245, "step": 6830 }, { "epoch": 0.34, "learning_rate": 1.5339575416968106e-05, "loss": 0.5128, "step": 6831 }, { "epoch": 0.34, "learning_rate": 1.5338207977612987e-05, "loss": 0.4173, "step": 6832 }, { "epoch": 0.34, "learning_rate": 1.5336840398644027e-05, "loss": 0.1198, "step": 6833 }, { "epoch": 0.34, "learning_rate": 1.5335472680096993e-05, "loss": 0.3726, "step": 6834 }, { "epoch": 0.34, "learning_rate": 1.5334104822007664e-05, "loss": 0.3989, "step": 6835 }, { "epoch": 0.34, "learning_rate": 1.533273682441181e-05, "loss": 0.2962, "step": 6836 }, { "epoch": 0.34, "learning_rate": 1.5331368687345206e-05, "loss": 0.1775, "step": 6837 }, { "epoch": 0.34, "learning_rate": 1.5330000410843638e-05, "loss": 0.2087, "step": 6838 }, { "epoch": 0.34, "learning_rate": 1.532863199494289e-05, "loss": 0.3525, "step": 6839 }, { "epoch": 0.34, "learning_rate": 1.5327263439678754e-05, "loss": 0.3824, "step": 6840 }, { "epoch": 0.34, "learning_rate": 1.5325894745087015e-05, "loss": 0.3274, "step": 6841 }, { "epoch": 0.34, "learning_rate": 1.532452591120348e-05, "loss": 0.5309, "step": 6842 }, { "epoch": 0.34, "learning_rate": 1.532315693806394e-05, "loss": 0.6311, "step": 6843 }, { "epoch": 0.34, "learning_rate": 1.5321787825704208e-05, "loss": 0.4588, "step": 6844 }, { "epoch": 0.34, "learning_rate": 1.5320418574160076e-05, "loss": 0.3846, "step": 6845 }, { "epoch": 0.34, "learning_rate": 1.5319049183467375e-05, "loss": 0.3187, "step": 6846 }, { "epoch": 0.34, "learning_rate": 1.5317679653661902e-05, "loss": 0.3223, "step": 6847 }, { "epoch": 0.34, "learning_rate": 1.5316309984779484e-05, "loss": 0.3876, "step": 6848 }, { "epoch": 0.34, "learning_rate": 1.5314940176855943e-05, "loss": 0.3201, "step": 6849 }, { "epoch": 0.34, "learning_rate": 1.53135702299271e-05, "loss": 0.283, "step": 6850 }, { "epoch": 0.34, "learning_rate": 1.5312200144028793e-05, "loss": 0.4692, "step": 6851 }, { "epoch": 0.34, "learning_rate": 1.5310829919196843e-05, "loss": 0.3913, "step": 6852 }, { "epoch": 0.34, "learning_rate": 1.5309459555467094e-05, "loss": 0.4237, "step": 6853 }, { "epoch": 0.34, "learning_rate": 1.5308089052875388e-05, "loss": 0.4144, "step": 6854 }, { "epoch": 0.34, "learning_rate": 1.5306718411457563e-05, "loss": 0.6316, "step": 6855 }, { "epoch": 0.34, "learning_rate": 1.5305347631249467e-05, "loss": 0.4649, "step": 6856 }, { "epoch": 0.34, "learning_rate": 1.5303976712286955e-05, "loss": 0.4906, "step": 6857 }, { "epoch": 0.34, "learning_rate": 1.5302605654605878e-05, "loss": 0.2661, "step": 6858 }, { "epoch": 0.34, "learning_rate": 1.5301234458242095e-05, "loss": 0.325, "step": 6859 }, { "epoch": 0.34, "learning_rate": 1.5299863123231467e-05, "loss": 0.4065, "step": 6860 }, { "epoch": 0.34, "learning_rate": 1.529849164960986e-05, "loss": 0.6517, "step": 6861 }, { "epoch": 0.34, "learning_rate": 1.5297120037413145e-05, "loss": 0.3431, "step": 6862 }, { "epoch": 0.34, "learning_rate": 1.529574828667719e-05, "loss": 0.3809, "step": 6863 }, { "epoch": 0.34, "learning_rate": 1.5294376397437882e-05, "loss": 0.5464, "step": 6864 }, { "epoch": 0.34, "learning_rate": 1.529300436973109e-05, "loss": 0.1628, "step": 6865 }, { "epoch": 0.34, "learning_rate": 1.5291632203592702e-05, "loss": 0.3002, "step": 6866 }, { "epoch": 0.34, "learning_rate": 1.5290259899058605e-05, "loss": 0.4642, "step": 6867 }, { "epoch": 0.34, "learning_rate": 1.528888745616469e-05, "loss": 0.3276, "step": 6868 }, { "epoch": 0.34, "learning_rate": 1.528751487494685e-05, "loss": 0.2866, "step": 6869 }, { "epoch": 0.34, "learning_rate": 1.528614215544098e-05, "loss": 0.4285, "step": 6870 }, { "epoch": 0.34, "learning_rate": 1.528476929768299e-05, "loss": 0.406, "step": 6871 }, { "epoch": 0.34, "learning_rate": 1.5283396301708778e-05, "loss": 0.5897, "step": 6872 }, { "epoch": 0.34, "learning_rate": 1.5282023167554257e-05, "loss": 0.2308, "step": 6873 }, { "epoch": 0.34, "learning_rate": 1.5280649895255337e-05, "loss": 0.148, "step": 6874 }, { "epoch": 0.34, "learning_rate": 1.5279276484847936e-05, "loss": 0.3474, "step": 6875 }, { "epoch": 0.34, "learning_rate": 1.5277902936367978e-05, "loss": 0.2463, "step": 6876 }, { "epoch": 0.34, "learning_rate": 1.5276529249851373e-05, "loss": 0.4734, "step": 6877 }, { "epoch": 0.34, "learning_rate": 1.527515542533406e-05, "loss": 0.4797, "step": 6878 }, { "epoch": 0.34, "learning_rate": 1.527378146285197e-05, "loss": 0.3165, "step": 6879 }, { "epoch": 0.34, "learning_rate": 1.527240736244103e-05, "loss": 0.2945, "step": 6880 }, { "epoch": 0.34, "learning_rate": 1.527103312413718e-05, "loss": 0.2671, "step": 6881 }, { "epoch": 0.34, "learning_rate": 1.5269658747976363e-05, "loss": 0.5648, "step": 6882 }, { "epoch": 0.34, "learning_rate": 1.5268284233994526e-05, "loss": 0.3802, "step": 6883 }, { "epoch": 0.34, "learning_rate": 1.526690958222761e-05, "loss": 0.4606, "step": 6884 }, { "epoch": 0.34, "learning_rate": 1.5265534792711577e-05, "loss": 0.2568, "step": 6885 }, { "epoch": 0.34, "learning_rate": 1.5264159865482375e-05, "loss": 0.5796, "step": 6886 }, { "epoch": 0.34, "learning_rate": 1.526278480057597e-05, "loss": 0.4849, "step": 6887 }, { "epoch": 0.34, "learning_rate": 1.5261409598028318e-05, "loss": 0.3113, "step": 6888 }, { "epoch": 0.34, "learning_rate": 1.5260034257875392e-05, "loss": 0.1563, "step": 6889 }, { "epoch": 0.34, "learning_rate": 1.5258658780153157e-05, "loss": 0.3332, "step": 6890 }, { "epoch": 0.34, "learning_rate": 1.5257283164897591e-05, "loss": 0.1812, "step": 6891 }, { "epoch": 0.34, "learning_rate": 1.525590741214467e-05, "loss": 0.3393, "step": 6892 }, { "epoch": 0.34, "learning_rate": 1.5254531521930373e-05, "loss": 0.3559, "step": 6893 }, { "epoch": 0.34, "learning_rate": 1.5253155494290688e-05, "loss": 0.3062, "step": 6894 }, { "epoch": 0.34, "learning_rate": 1.5251779329261598e-05, "loss": 0.4728, "step": 6895 }, { "epoch": 0.34, "learning_rate": 1.5250403026879102e-05, "loss": 0.6543, "step": 6896 }, { "epoch": 0.34, "learning_rate": 1.5249026587179189e-05, "loss": 0.1731, "step": 6897 }, { "epoch": 0.34, "learning_rate": 1.5247650010197863e-05, "loss": 0.3914, "step": 6898 }, { "epoch": 0.34, "learning_rate": 1.5246273295971123e-05, "loss": 0.4513, "step": 6899 }, { "epoch": 0.34, "learning_rate": 1.5244896444534976e-05, "loss": 0.48, "step": 6900 }, { "epoch": 0.34, "learning_rate": 1.5243519455925432e-05, "loss": 0.2835, "step": 6901 }, { "epoch": 0.34, "learning_rate": 1.5242142330178504e-05, "loss": 0.4999, "step": 6902 }, { "epoch": 0.34, "learning_rate": 1.524076506733021e-05, "loss": 0.2638, "step": 6903 }, { "epoch": 0.34, "learning_rate": 1.5239387667416571e-05, "loss": 0.1347, "step": 6904 }, { "epoch": 0.34, "learning_rate": 1.5238010130473609e-05, "loss": 0.4327, "step": 6905 }, { "epoch": 0.34, "learning_rate": 1.5236632456537352e-05, "loss": 0.5705, "step": 6906 }, { "epoch": 0.34, "learning_rate": 1.5235254645643832e-05, "loss": 0.5496, "step": 6907 }, { "epoch": 0.34, "learning_rate": 1.5233876697829084e-05, "loss": 0.2819, "step": 6908 }, { "epoch": 0.34, "learning_rate": 1.5232498613129147e-05, "loss": 0.3604, "step": 6909 }, { "epoch": 0.35, "learning_rate": 1.523112039158006e-05, "loss": 0.3158, "step": 6910 }, { "epoch": 0.35, "learning_rate": 1.5229742033217872e-05, "loss": 0.5284, "step": 6911 }, { "epoch": 0.35, "learning_rate": 1.522836353807863e-05, "loss": 0.2503, "step": 6912 }, { "epoch": 0.35, "learning_rate": 1.522698490619839e-05, "loss": 0.2683, "step": 6913 }, { "epoch": 0.35, "learning_rate": 1.5225606137613205e-05, "loss": 0.4176, "step": 6914 }, { "epoch": 0.35, "learning_rate": 1.5224227232359135e-05, "loss": 0.151, "step": 6915 }, { "epoch": 0.35, "learning_rate": 1.5222848190472244e-05, "loss": 0.5448, "step": 6916 }, { "epoch": 0.35, "learning_rate": 1.52214690119886e-05, "loss": 0.1364, "step": 6917 }, { "epoch": 0.35, "learning_rate": 1.522008969694427e-05, "loss": 0.3812, "step": 6918 }, { "epoch": 0.35, "learning_rate": 1.5218710245375334e-05, "loss": 0.3796, "step": 6919 }, { "epoch": 0.35, "learning_rate": 1.5217330657317867e-05, "loss": 0.3452, "step": 6920 }, { "epoch": 0.35, "learning_rate": 1.521595093280795e-05, "loss": 0.2328, "step": 6921 }, { "epoch": 0.35, "learning_rate": 1.5214571071881668e-05, "loss": 0.4697, "step": 6922 }, { "epoch": 0.35, "learning_rate": 1.5213191074575107e-05, "loss": 0.4809, "step": 6923 }, { "epoch": 0.35, "learning_rate": 1.5211810940924365e-05, "loss": 0.266, "step": 6924 }, { "epoch": 0.35, "learning_rate": 1.5210430670965531e-05, "loss": 0.3757, "step": 6925 }, { "epoch": 0.35, "learning_rate": 1.5209050264734711e-05, "loss": 0.4684, "step": 6926 }, { "epoch": 0.35, "learning_rate": 1.5207669722268001e-05, "loss": 0.3427, "step": 6927 }, { "epoch": 0.35, "learning_rate": 1.5206289043601513e-05, "loss": 0.2538, "step": 6928 }, { "epoch": 0.35, "learning_rate": 1.5204908228771352e-05, "loss": 0.4375, "step": 6929 }, { "epoch": 0.35, "learning_rate": 1.5203527277813632e-05, "loss": 0.5299, "step": 6930 }, { "epoch": 0.35, "learning_rate": 1.5202146190764473e-05, "loss": 0.4509, "step": 6931 }, { "epoch": 0.35, "learning_rate": 1.5200764967659994e-05, "loss": 0.3738, "step": 6932 }, { "epoch": 0.35, "learning_rate": 1.5199383608536317e-05, "loss": 0.3372, "step": 6933 }, { "epoch": 0.35, "learning_rate": 1.5198002113429575e-05, "loss": 0.4284, "step": 6934 }, { "epoch": 0.35, "learning_rate": 1.5196620482375894e-05, "loss": 0.22, "step": 6935 }, { "epoch": 0.35, "learning_rate": 1.519523871541141e-05, "loss": 0.4982, "step": 6936 }, { "epoch": 0.35, "learning_rate": 1.5193856812572266e-05, "loss": 0.2791, "step": 6937 }, { "epoch": 0.35, "learning_rate": 1.5192474773894596e-05, "loss": 0.2007, "step": 6938 }, { "epoch": 0.35, "learning_rate": 1.519109259941455e-05, "loss": 0.1962, "step": 6939 }, { "epoch": 0.35, "learning_rate": 1.5189710289168275e-05, "loss": 0.3924, "step": 6940 }, { "epoch": 0.35, "learning_rate": 1.5188327843191928e-05, "loss": 0.4432, "step": 6941 }, { "epoch": 0.35, "learning_rate": 1.518694526152166e-05, "loss": 0.3218, "step": 6942 }, { "epoch": 0.35, "learning_rate": 1.518556254419363e-05, "loss": 0.3731, "step": 6943 }, { "epoch": 0.35, "learning_rate": 1.5184179691244007e-05, "loss": 0.2973, "step": 6944 }, { "epoch": 0.35, "learning_rate": 1.5182796702708955e-05, "loss": 0.1898, "step": 6945 }, { "epoch": 0.35, "learning_rate": 1.5181413578624638e-05, "loss": 0.2229, "step": 6946 }, { "epoch": 0.35, "learning_rate": 1.5180030319027239e-05, "loss": 0.3818, "step": 6947 }, { "epoch": 0.35, "learning_rate": 1.517864692395293e-05, "loss": 0.2809, "step": 6948 }, { "epoch": 0.35, "learning_rate": 1.5177263393437898e-05, "loss": 0.4219, "step": 6949 }, { "epoch": 0.35, "learning_rate": 1.5175879727518317e-05, "loss": 0.2256, "step": 6950 }, { "epoch": 0.35, "learning_rate": 1.5174495926230384e-05, "loss": 0.3161, "step": 6951 }, { "epoch": 0.35, "learning_rate": 1.517311198961029e-05, "loss": 0.4829, "step": 6952 }, { "epoch": 0.35, "learning_rate": 1.5171727917694223e-05, "loss": 0.2309, "step": 6953 }, { "epoch": 0.35, "learning_rate": 1.5170343710518389e-05, "loss": 0.3198, "step": 6954 }, { "epoch": 0.35, "learning_rate": 1.5168959368118984e-05, "loss": 0.2917, "step": 6955 }, { "epoch": 0.35, "learning_rate": 1.516757489053222e-05, "loss": 0.4768, "step": 6956 }, { "epoch": 0.35, "learning_rate": 1.51661902777943e-05, "loss": 0.3369, "step": 6957 }, { "epoch": 0.35, "learning_rate": 1.5164805529941442e-05, "loss": 0.4776, "step": 6958 }, { "epoch": 0.35, "learning_rate": 1.5163420647009858e-05, "loss": 0.3496, "step": 6959 }, { "epoch": 0.35, "learning_rate": 1.5162035629035771e-05, "loss": 0.3988, "step": 6960 }, { "epoch": 0.35, "learning_rate": 1.5160650476055405e-05, "loss": 0.2958, "step": 6961 }, { "epoch": 0.35, "learning_rate": 1.515926518810498e-05, "loss": 0.5148, "step": 6962 }, { "epoch": 0.35, "learning_rate": 1.5157879765220737e-05, "loss": 0.3605, "step": 6963 }, { "epoch": 0.35, "learning_rate": 1.5156494207438904e-05, "loss": 0.2057, "step": 6964 }, { "epoch": 0.35, "learning_rate": 1.5155108514795716e-05, "loss": 0.3379, "step": 6965 }, { "epoch": 0.35, "learning_rate": 1.515372268732742e-05, "loss": 0.4386, "step": 6966 }, { "epoch": 0.35, "learning_rate": 1.5152336725070256e-05, "loss": 0.4822, "step": 6967 }, { "epoch": 0.35, "learning_rate": 1.5150950628060471e-05, "loss": 0.3442, "step": 6968 }, { "epoch": 0.35, "learning_rate": 1.514956439633432e-05, "loss": 0.4841, "step": 6969 }, { "epoch": 0.35, "learning_rate": 1.5148178029928057e-05, "loss": 0.0917, "step": 6970 }, { "epoch": 0.35, "learning_rate": 1.514679152887794e-05, "loss": 0.3121, "step": 6971 }, { "epoch": 0.35, "learning_rate": 1.5145404893220233e-05, "loss": 0.4388, "step": 6972 }, { "epoch": 0.35, "learning_rate": 1.51440181229912e-05, "loss": 0.1347, "step": 6973 }, { "epoch": 0.35, "learning_rate": 1.5142631218227108e-05, "loss": 0.399, "step": 6974 }, { "epoch": 0.35, "learning_rate": 1.5141244178964237e-05, "loss": 0.4715, "step": 6975 }, { "epoch": 0.35, "learning_rate": 1.5139857005238854e-05, "loss": 0.1394, "step": 6976 }, { "epoch": 0.35, "learning_rate": 1.5138469697087246e-05, "loss": 0.6484, "step": 6977 }, { "epoch": 0.35, "learning_rate": 1.5137082254545691e-05, "loss": 0.3894, "step": 6978 }, { "epoch": 0.35, "learning_rate": 1.5135694677650481e-05, "loss": 0.2556, "step": 6979 }, { "epoch": 0.35, "learning_rate": 1.51343069664379e-05, "loss": 0.5312, "step": 6980 }, { "epoch": 0.35, "learning_rate": 1.5132919120944247e-05, "loss": 0.3521, "step": 6981 }, { "epoch": 0.35, "learning_rate": 1.5131531141205816e-05, "loss": 0.5037, "step": 6982 }, { "epoch": 0.35, "learning_rate": 1.5130143027258908e-05, "loss": 0.3455, "step": 6983 }, { "epoch": 0.35, "learning_rate": 1.5128754779139828e-05, "loss": 0.3785, "step": 6984 }, { "epoch": 0.35, "learning_rate": 1.5127366396884888e-05, "loss": 0.2745, "step": 6985 }, { "epoch": 0.35, "learning_rate": 1.5125977880530394e-05, "loss": 0.3176, "step": 6986 }, { "epoch": 0.35, "learning_rate": 1.5124589230112662e-05, "loss": 0.2458, "step": 6987 }, { "epoch": 0.35, "learning_rate": 1.512320044566801e-05, "loss": 0.0818, "step": 6988 }, { "epoch": 0.35, "learning_rate": 1.5121811527232764e-05, "loss": 0.3388, "step": 6989 }, { "epoch": 0.35, "learning_rate": 1.5120422474843243e-05, "loss": 0.3044, "step": 6990 }, { "epoch": 0.35, "learning_rate": 1.5119033288535775e-05, "loss": 0.4092, "step": 6991 }, { "epoch": 0.35, "learning_rate": 1.5117643968346705e-05, "loss": 0.4679, "step": 6992 }, { "epoch": 0.35, "learning_rate": 1.5116254514312351e-05, "loss": 0.1816, "step": 6993 }, { "epoch": 0.35, "learning_rate": 1.5114864926469066e-05, "loss": 0.2048, "step": 6994 }, { "epoch": 0.35, "learning_rate": 1.5113475204853187e-05, "loss": 0.3873, "step": 6995 }, { "epoch": 0.35, "learning_rate": 1.5112085349501062e-05, "loss": 0.1647, "step": 6996 }, { "epoch": 0.35, "learning_rate": 1.5110695360449042e-05, "loss": 0.3106, "step": 6997 }, { "epoch": 0.35, "learning_rate": 1.5109305237733476e-05, "loss": 0.3306, "step": 6998 }, { "epoch": 0.35, "learning_rate": 1.5107914981390723e-05, "loss": 0.3303, "step": 6999 }, { "epoch": 0.35, "learning_rate": 1.5106524591457146e-05, "loss": 0.3432, "step": 7000 }, { "epoch": 0.35, "learning_rate": 1.5105134067969105e-05, "loss": 0.5401, "step": 7001 }, { "epoch": 0.35, "learning_rate": 1.510374341096297e-05, "loss": 0.2956, "step": 7002 }, { "epoch": 0.35, "learning_rate": 1.5102352620475112e-05, "loss": 0.3643, "step": 7003 }, { "epoch": 0.35, "learning_rate": 1.5100961696541903e-05, "loss": 0.2111, "step": 7004 }, { "epoch": 0.35, "learning_rate": 1.5099570639199719e-05, "loss": 0.4313, "step": 7005 }, { "epoch": 0.35, "learning_rate": 1.5098179448484946e-05, "loss": 0.2646, "step": 7006 }, { "epoch": 0.35, "learning_rate": 1.5096788124433968e-05, "loss": 0.367, "step": 7007 }, { "epoch": 0.35, "learning_rate": 1.5095396667083173e-05, "loss": 0.5152, "step": 7008 }, { "epoch": 0.35, "learning_rate": 1.5094005076468947e-05, "loss": 0.3951, "step": 7009 }, { "epoch": 0.35, "learning_rate": 1.5092613352627696e-05, "loss": 0.4147, "step": 7010 }, { "epoch": 0.35, "learning_rate": 1.509122149559581e-05, "loss": 0.528, "step": 7011 }, { "epoch": 0.35, "learning_rate": 1.5089829505409696e-05, "loss": 0.4303, "step": 7012 }, { "epoch": 0.35, "learning_rate": 1.5088437382105757e-05, "loss": 0.3852, "step": 7013 }, { "epoch": 0.35, "learning_rate": 1.5087045125720404e-05, "loss": 0.3977, "step": 7014 }, { "epoch": 0.35, "learning_rate": 1.5085652736290046e-05, "loss": 0.417, "step": 7015 }, { "epoch": 0.35, "learning_rate": 1.5084260213851102e-05, "loss": 0.2099, "step": 7016 }, { "epoch": 0.35, "learning_rate": 1.5082867558439993e-05, "loss": 0.338, "step": 7017 }, { "epoch": 0.35, "learning_rate": 1.5081474770093142e-05, "loss": 0.3404, "step": 7018 }, { "epoch": 0.35, "learning_rate": 1.5080081848846968e-05, "loss": 0.3282, "step": 7019 }, { "epoch": 0.35, "learning_rate": 1.507868879473791e-05, "loss": 0.3725, "step": 7020 }, { "epoch": 0.35, "learning_rate": 1.5077295607802398e-05, "loss": 0.4716, "step": 7021 }, { "epoch": 0.35, "learning_rate": 1.507590228807687e-05, "loss": 0.4235, "step": 7022 }, { "epoch": 0.35, "learning_rate": 1.5074508835597765e-05, "loss": 0.4073, "step": 7023 }, { "epoch": 0.35, "learning_rate": 1.507311525040153e-05, "loss": 0.408, "step": 7024 }, { "epoch": 0.35, "learning_rate": 1.507172153252461e-05, "loss": 0.4452, "step": 7025 }, { "epoch": 0.35, "learning_rate": 1.5070327682003456e-05, "loss": 0.5618, "step": 7026 }, { "epoch": 0.35, "learning_rate": 1.5068933698874522e-05, "loss": 0.4622, "step": 7027 }, { "epoch": 0.35, "learning_rate": 1.5067539583174267e-05, "loss": 0.1419, "step": 7028 }, { "epoch": 0.35, "learning_rate": 1.5066145334939148e-05, "loss": 0.3844, "step": 7029 }, { "epoch": 0.35, "learning_rate": 1.5064750954205636e-05, "loss": 0.428, "step": 7030 }, { "epoch": 0.35, "learning_rate": 1.5063356441010198e-05, "loss": 0.1805, "step": 7031 }, { "epoch": 0.35, "learning_rate": 1.50619617953893e-05, "loss": 0.2627, "step": 7032 }, { "epoch": 0.35, "learning_rate": 1.5060567017379425e-05, "loss": 0.3922, "step": 7033 }, { "epoch": 0.35, "learning_rate": 1.5059172107017044e-05, "loss": 0.107, "step": 7034 }, { "epoch": 0.35, "learning_rate": 1.5057777064338646e-05, "loss": 0.3742, "step": 7035 }, { "epoch": 0.35, "learning_rate": 1.5056381889380714e-05, "loss": 0.3429, "step": 7036 }, { "epoch": 0.35, "learning_rate": 1.5054986582179732e-05, "loss": 0.3681, "step": 7037 }, { "epoch": 0.35, "learning_rate": 1.5053591142772199e-05, "loss": 0.691, "step": 7038 }, { "epoch": 0.35, "learning_rate": 1.505219557119461e-05, "loss": 0.2989, "step": 7039 }, { "epoch": 0.35, "learning_rate": 1.5050799867483464e-05, "loss": 0.3796, "step": 7040 }, { "epoch": 0.35, "learning_rate": 1.5049404031675259e-05, "loss": 0.3444, "step": 7041 }, { "epoch": 0.35, "learning_rate": 1.5048008063806507e-05, "loss": 0.3498, "step": 7042 }, { "epoch": 0.35, "learning_rate": 1.5046611963913717e-05, "loss": 0.1907, "step": 7043 }, { "epoch": 0.35, "learning_rate": 1.50452157320334e-05, "loss": 0.4373, "step": 7044 }, { "epoch": 0.35, "learning_rate": 1.5043819368202073e-05, "loss": 0.3184, "step": 7045 }, { "epoch": 0.35, "learning_rate": 1.5042422872456257e-05, "loss": 0.7157, "step": 7046 }, { "epoch": 0.35, "learning_rate": 1.5041026244832475e-05, "loss": 0.2945, "step": 7047 }, { "epoch": 0.35, "learning_rate": 1.5039629485367253e-05, "loss": 0.4701, "step": 7048 }, { "epoch": 0.35, "learning_rate": 1.5038232594097124e-05, "loss": 0.3735, "step": 7049 }, { "epoch": 0.35, "learning_rate": 1.503683557105862e-05, "loss": 0.3387, "step": 7050 }, { "epoch": 0.35, "learning_rate": 1.5035438416288281e-05, "loss": 0.4367, "step": 7051 }, { "epoch": 0.35, "learning_rate": 1.5034041129822643e-05, "loss": 0.3273, "step": 7052 }, { "epoch": 0.35, "learning_rate": 1.5032643711698253e-05, "loss": 0.3746, "step": 7053 }, { "epoch": 0.35, "learning_rate": 1.5031246161951658e-05, "loss": 0.3553, "step": 7054 }, { "epoch": 0.35, "learning_rate": 1.5029848480619409e-05, "loss": 0.2838, "step": 7055 }, { "epoch": 0.35, "learning_rate": 1.5028450667738063e-05, "loss": 0.5862, "step": 7056 }, { "epoch": 0.35, "learning_rate": 1.5027052723344173e-05, "loss": 0.2621, "step": 7057 }, { "epoch": 0.35, "learning_rate": 1.5025654647474306e-05, "loss": 0.3731, "step": 7058 }, { "epoch": 0.35, "learning_rate": 1.5024256440165024e-05, "loss": 0.4978, "step": 7059 }, { "epoch": 0.35, "learning_rate": 1.5022858101452893e-05, "loss": 0.2506, "step": 7060 }, { "epoch": 0.35, "learning_rate": 1.502145963137449e-05, "loss": 0.3613, "step": 7061 }, { "epoch": 0.35, "learning_rate": 1.5020061029966386e-05, "loss": 0.3763, "step": 7062 }, { "epoch": 0.35, "learning_rate": 1.5018662297265161e-05, "loss": 0.2809, "step": 7063 }, { "epoch": 0.35, "learning_rate": 1.5017263433307397e-05, "loss": 0.4258, "step": 7064 }, { "epoch": 0.35, "learning_rate": 1.5015864438129677e-05, "loss": 0.4618, "step": 7065 }, { "epoch": 0.35, "learning_rate": 1.5014465311768595e-05, "loss": 0.608, "step": 7066 }, { "epoch": 0.35, "learning_rate": 1.5013066054260738e-05, "loss": 0.475, "step": 7067 }, { "epoch": 0.35, "learning_rate": 1.5011666665642706e-05, "loss": 0.4092, "step": 7068 }, { "epoch": 0.35, "learning_rate": 1.5010267145951096e-05, "loss": 0.3005, "step": 7069 }, { "epoch": 0.35, "learning_rate": 1.5008867495222513e-05, "loss": 0.2163, "step": 7070 }, { "epoch": 0.35, "learning_rate": 1.5007467713493558e-05, "loss": 0.4847, "step": 7071 }, { "epoch": 0.35, "learning_rate": 1.5006067800800845e-05, "loss": 0.4404, "step": 7072 }, { "epoch": 0.35, "learning_rate": 1.5004667757180987e-05, "loss": 0.4912, "step": 7073 }, { "epoch": 0.35, "learning_rate": 1.5003267582670595e-05, "loss": 0.3608, "step": 7074 }, { "epoch": 0.35, "learning_rate": 1.5001867277306294e-05, "loss": 0.3811, "step": 7075 }, { "epoch": 0.35, "learning_rate": 1.500046684112471e-05, "loss": 0.3055, "step": 7076 }, { "epoch": 0.35, "learning_rate": 1.4999066274162461e-05, "loss": 0.5151, "step": 7077 }, { "epoch": 0.35, "learning_rate": 1.499766557645618e-05, "loss": 0.2222, "step": 7078 }, { "epoch": 0.35, "learning_rate": 1.4996264748042504e-05, "loss": 0.374, "step": 7079 }, { "epoch": 0.35, "learning_rate": 1.4994863788958068e-05, "loss": 0.1936, "step": 7080 }, { "epoch": 0.35, "learning_rate": 1.4993462699239509e-05, "loss": 0.2385, "step": 7081 }, { "epoch": 0.35, "learning_rate": 1.4992061478923477e-05, "loss": 0.4198, "step": 7082 }, { "epoch": 0.35, "learning_rate": 1.499066012804661e-05, "loss": 0.313, "step": 7083 }, { "epoch": 0.35, "learning_rate": 1.4989258646645568e-05, "loss": 0.4432, "step": 7084 }, { "epoch": 0.35, "learning_rate": 1.4987857034757001e-05, "loss": 0.3685, "step": 7085 }, { "epoch": 0.35, "learning_rate": 1.4986455292417566e-05, "loss": 0.2842, "step": 7086 }, { "epoch": 0.35, "learning_rate": 1.4985053419663923e-05, "loss": 0.4894, "step": 7087 }, { "epoch": 0.35, "learning_rate": 1.4983651416532739e-05, "loss": 0.4194, "step": 7088 }, { "epoch": 0.35, "learning_rate": 1.4982249283060677e-05, "loss": 0.5423, "step": 7089 }, { "epoch": 0.35, "learning_rate": 1.498084701928441e-05, "loss": 0.3366, "step": 7090 }, { "epoch": 0.35, "learning_rate": 1.4979444625240615e-05, "loss": 0.4065, "step": 7091 }, { "epoch": 0.35, "learning_rate": 1.4978042100965966e-05, "loss": 0.2769, "step": 7092 }, { "epoch": 0.35, "learning_rate": 1.4976639446497148e-05, "loss": 0.4781, "step": 7093 }, { "epoch": 0.35, "learning_rate": 1.4975236661870842e-05, "loss": 0.3591, "step": 7094 }, { "epoch": 0.35, "learning_rate": 1.4973833747123738e-05, "loss": 0.3639, "step": 7095 }, { "epoch": 0.35, "learning_rate": 1.4972430702292526e-05, "loss": 0.3416, "step": 7096 }, { "epoch": 0.35, "learning_rate": 1.4971027527413904e-05, "loss": 0.4809, "step": 7097 }, { "epoch": 0.35, "learning_rate": 1.4969624222524566e-05, "loss": 0.3141, "step": 7098 }, { "epoch": 0.35, "learning_rate": 1.4968220787661218e-05, "loss": 0.2982, "step": 7099 }, { "epoch": 0.35, "learning_rate": 1.496681722286056e-05, "loss": 0.4746, "step": 7100 }, { "epoch": 0.35, "learning_rate": 1.4965413528159303e-05, "loss": 0.4119, "step": 7101 }, { "epoch": 0.35, "learning_rate": 1.496400970359416e-05, "loss": 0.379, "step": 7102 }, { "epoch": 0.35, "learning_rate": 1.4962605749201845e-05, "loss": 0.2427, "step": 7103 }, { "epoch": 0.35, "learning_rate": 1.4961201665019073e-05, "loss": 0.2399, "step": 7104 }, { "epoch": 0.35, "learning_rate": 1.4959797451082572e-05, "loss": 0.3159, "step": 7105 }, { "epoch": 0.35, "learning_rate": 1.4958393107429065e-05, "loss": 0.4981, "step": 7106 }, { "epoch": 0.35, "learning_rate": 1.4956988634095281e-05, "loss": 0.6166, "step": 7107 }, { "epoch": 0.35, "learning_rate": 1.4955584031117952e-05, "loss": 0.5239, "step": 7108 }, { "epoch": 0.35, "learning_rate": 1.495417929853381e-05, "loss": 0.3315, "step": 7109 }, { "epoch": 0.36, "learning_rate": 1.49527744363796e-05, "loss": 0.3189, "step": 7110 }, { "epoch": 0.36, "learning_rate": 1.4951369444692061e-05, "loss": 0.4849, "step": 7111 }, { "epoch": 0.36, "learning_rate": 1.4949964323507941e-05, "loss": 0.3298, "step": 7112 }, { "epoch": 0.36, "learning_rate": 1.4948559072863986e-05, "loss": 0.2185, "step": 7113 }, { "epoch": 0.36, "learning_rate": 1.4947153692796951e-05, "loss": 0.3711, "step": 7114 }, { "epoch": 0.36, "learning_rate": 1.494574818334359e-05, "loss": 0.3609, "step": 7115 }, { "epoch": 0.36, "learning_rate": 1.4944342544540662e-05, "loss": 0.4098, "step": 7116 }, { "epoch": 0.36, "learning_rate": 1.494293677642493e-05, "loss": 0.4483, "step": 7117 }, { "epoch": 0.36, "learning_rate": 1.4941530879033164e-05, "loss": 0.363, "step": 7118 }, { "epoch": 0.36, "learning_rate": 1.4940124852402127e-05, "loss": 0.3094, "step": 7119 }, { "epoch": 0.36, "learning_rate": 1.4938718696568595e-05, "loss": 0.2767, "step": 7120 }, { "epoch": 0.36, "learning_rate": 1.4937312411569345e-05, "loss": 0.2495, "step": 7121 }, { "epoch": 0.36, "learning_rate": 1.4935905997441156e-05, "loss": 0.602, "step": 7122 }, { "epoch": 0.36, "learning_rate": 1.4934499454220813e-05, "loss": 0.3791, "step": 7123 }, { "epoch": 0.36, "learning_rate": 1.4933092781945095e-05, "loss": 0.577, "step": 7124 }, { "epoch": 0.36, "learning_rate": 1.4931685980650801e-05, "loss": 0.417, "step": 7125 }, { "epoch": 0.36, "learning_rate": 1.4930279050374714e-05, "loss": 0.1223, "step": 7126 }, { "epoch": 0.36, "learning_rate": 1.4928871991153639e-05, "loss": 0.4326, "step": 7127 }, { "epoch": 0.36, "learning_rate": 1.4927464803024374e-05, "loss": 0.4898, "step": 7128 }, { "epoch": 0.36, "learning_rate": 1.4926057486023719e-05, "loss": 0.4368, "step": 7129 }, { "epoch": 0.36, "learning_rate": 1.4924650040188482e-05, "loss": 0.2095, "step": 7130 }, { "epoch": 0.36, "learning_rate": 1.4923242465555474e-05, "loss": 0.352, "step": 7131 }, { "epoch": 0.36, "learning_rate": 1.492183476216151e-05, "loss": 0.435, "step": 7132 }, { "epoch": 0.36, "learning_rate": 1.49204269300434e-05, "loss": 0.4327, "step": 7133 }, { "epoch": 0.36, "learning_rate": 1.491901896923797e-05, "loss": 0.3188, "step": 7134 }, { "epoch": 0.36, "learning_rate": 1.4917610879782045e-05, "loss": 0.5074, "step": 7135 }, { "epoch": 0.36, "learning_rate": 1.4916202661712445e-05, "loss": 0.3105, "step": 7136 }, { "epoch": 0.36, "learning_rate": 1.4914794315066004e-05, "loss": 0.3453, "step": 7137 }, { "epoch": 0.36, "learning_rate": 1.4913385839879555e-05, "loss": 0.1787, "step": 7138 }, { "epoch": 0.36, "learning_rate": 1.4911977236189934e-05, "loss": 0.2615, "step": 7139 }, { "epoch": 0.36, "learning_rate": 1.4910568504033982e-05, "loss": 0.3222, "step": 7140 }, { "epoch": 0.36, "learning_rate": 1.4909159643448543e-05, "loss": 0.1611, "step": 7141 }, { "epoch": 0.36, "learning_rate": 1.4907750654470464e-05, "loss": 0.6309, "step": 7142 }, { "epoch": 0.36, "learning_rate": 1.4906341537136593e-05, "loss": 0.3615, "step": 7143 }, { "epoch": 0.36, "learning_rate": 1.4904932291483784e-05, "loss": 0.3896, "step": 7144 }, { "epoch": 0.36, "learning_rate": 1.4903522917548898e-05, "loss": 0.3159, "step": 7145 }, { "epoch": 0.36, "learning_rate": 1.4902113415368792e-05, "loss": 0.216, "step": 7146 }, { "epoch": 0.36, "learning_rate": 1.4900703784980331e-05, "loss": 0.3092, "step": 7147 }, { "epoch": 0.36, "learning_rate": 1.489929402642038e-05, "loss": 0.5791, "step": 7148 }, { "epoch": 0.36, "learning_rate": 1.4897884139725813e-05, "loss": 0.3975, "step": 7149 }, { "epoch": 0.36, "learning_rate": 1.48964741249335e-05, "loss": 0.2999, "step": 7150 }, { "epoch": 0.36, "learning_rate": 1.4895063982080317e-05, "loss": 0.5346, "step": 7151 }, { "epoch": 0.36, "learning_rate": 1.4893653711203151e-05, "loss": 0.4264, "step": 7152 }, { "epoch": 0.36, "learning_rate": 1.4892243312338877e-05, "loss": 0.4131, "step": 7153 }, { "epoch": 0.36, "learning_rate": 1.4890832785524392e-05, "loss": 0.3428, "step": 7154 }, { "epoch": 0.36, "learning_rate": 1.4889422130796578e-05, "loss": 0.3008, "step": 7155 }, { "epoch": 0.36, "learning_rate": 1.4888011348192333e-05, "loss": 0.4448, "step": 7156 }, { "epoch": 0.36, "learning_rate": 1.4886600437748552e-05, "loss": 0.1849, "step": 7157 }, { "epoch": 0.36, "learning_rate": 1.4885189399502138e-05, "loss": 0.5634, "step": 7158 }, { "epoch": 0.36, "learning_rate": 1.4883778233489997e-05, "loss": 0.358, "step": 7159 }, { "epoch": 0.36, "learning_rate": 1.4882366939749028e-05, "loss": 0.3047, "step": 7160 }, { "epoch": 0.36, "learning_rate": 1.488095551831615e-05, "loss": 0.3921, "step": 7161 }, { "epoch": 0.36, "learning_rate": 1.487954396922827e-05, "loss": 0.5099, "step": 7162 }, { "epoch": 0.36, "learning_rate": 1.4878132292522312e-05, "loss": 0.4113, "step": 7163 }, { "epoch": 0.36, "learning_rate": 1.487672048823519e-05, "loss": 0.4899, "step": 7164 }, { "epoch": 0.36, "learning_rate": 1.4875308556403832e-05, "loss": 0.5231, "step": 7165 }, { "epoch": 0.36, "learning_rate": 1.4873896497065163e-05, "loss": 0.2914, "step": 7166 }, { "epoch": 0.36, "learning_rate": 1.4872484310256117e-05, "loss": 0.3365, "step": 7167 }, { "epoch": 0.36, "learning_rate": 1.4871071996013625e-05, "loss": 0.2437, "step": 7168 }, { "epoch": 0.36, "learning_rate": 1.4869659554374623e-05, "loss": 0.5085, "step": 7169 }, { "epoch": 0.36, "learning_rate": 1.4868246985376055e-05, "loss": 0.2199, "step": 7170 }, { "epoch": 0.36, "learning_rate": 1.4866834289054865e-05, "loss": 0.6126, "step": 7171 }, { "epoch": 0.36, "learning_rate": 1.4865421465447999e-05, "loss": 0.3137, "step": 7172 }, { "epoch": 0.36, "learning_rate": 1.4864008514592404e-05, "loss": 0.1983, "step": 7173 }, { "epoch": 0.36, "learning_rate": 1.4862595436525038e-05, "loss": 0.1863, "step": 7174 }, { "epoch": 0.36, "learning_rate": 1.4861182231282858e-05, "loss": 0.317, "step": 7175 }, { "epoch": 0.36, "learning_rate": 1.4859768898902823e-05, "loss": 0.6332, "step": 7176 }, { "epoch": 0.36, "learning_rate": 1.48583554394219e-05, "loss": 0.463, "step": 7177 }, { "epoch": 0.36, "learning_rate": 1.485694185287705e-05, "loss": 0.5077, "step": 7178 }, { "epoch": 0.36, "learning_rate": 1.4855528139305246e-05, "loss": 0.6545, "step": 7179 }, { "epoch": 0.36, "learning_rate": 1.485411429874347e-05, "loss": 0.3781, "step": 7180 }, { "epoch": 0.36, "learning_rate": 1.4852700331228684e-05, "loss": 0.3153, "step": 7181 }, { "epoch": 0.36, "learning_rate": 1.4851286236797884e-05, "loss": 0.2606, "step": 7182 }, { "epoch": 0.36, "learning_rate": 1.4849872015488043e-05, "loss": 0.3818, "step": 7183 }, { "epoch": 0.36, "learning_rate": 1.4848457667336152e-05, "loss": 0.6144, "step": 7184 }, { "epoch": 0.36, "learning_rate": 1.48470431923792e-05, "loss": 0.3718, "step": 7185 }, { "epoch": 0.36, "learning_rate": 1.4845628590654184e-05, "loss": 0.4921, "step": 7186 }, { "epoch": 0.36, "learning_rate": 1.4844213862198098e-05, "loss": 0.3441, "step": 7187 }, { "epoch": 0.36, "learning_rate": 1.4842799007047943e-05, "loss": 0.477, "step": 7188 }, { "epoch": 0.36, "learning_rate": 1.4841384025240724e-05, "loss": 0.4875, "step": 7189 }, { "epoch": 0.36, "learning_rate": 1.4839968916813444e-05, "loss": 0.217, "step": 7190 }, { "epoch": 0.36, "learning_rate": 1.4838553681803118e-05, "loss": 0.3638, "step": 7191 }, { "epoch": 0.36, "learning_rate": 1.4837138320246758e-05, "loss": 0.3928, "step": 7192 }, { "epoch": 0.36, "learning_rate": 1.4835722832181381e-05, "loss": 0.491, "step": 7193 }, { "epoch": 0.36, "learning_rate": 1.4834307217644007e-05, "loss": 0.371, "step": 7194 }, { "epoch": 0.36, "learning_rate": 1.4832891476671659e-05, "loss": 0.4444, "step": 7195 }, { "epoch": 0.36, "learning_rate": 1.4831475609301365e-05, "loss": 0.3893, "step": 7196 }, { "epoch": 0.36, "learning_rate": 1.4830059615570154e-05, "loss": 0.2749, "step": 7197 }, { "epoch": 0.36, "learning_rate": 1.482864349551506e-05, "loss": 0.6611, "step": 7198 }, { "epoch": 0.36, "learning_rate": 1.482722724917312e-05, "loss": 0.5418, "step": 7199 }, { "epoch": 0.36, "learning_rate": 1.4825810876581375e-05, "loss": 0.3566, "step": 7200 }, { "epoch": 0.36, "learning_rate": 1.4824394377776862e-05, "loss": 0.206, "step": 7201 }, { "epoch": 0.36, "learning_rate": 1.482297775279664e-05, "loss": 0.4343, "step": 7202 }, { "epoch": 0.36, "learning_rate": 1.482156100167775e-05, "loss": 0.4403, "step": 7203 }, { "epoch": 0.36, "learning_rate": 1.4820144124457244e-05, "loss": 0.2202, "step": 7204 }, { "epoch": 0.36, "learning_rate": 1.4818727121172183e-05, "loss": 0.2807, "step": 7205 }, { "epoch": 0.36, "learning_rate": 1.4817309991859622e-05, "loss": 0.4892, "step": 7206 }, { "epoch": 0.36, "learning_rate": 1.4815892736556633e-05, "loss": 0.3598, "step": 7207 }, { "epoch": 0.36, "learning_rate": 1.4814475355300277e-05, "loss": 0.2374, "step": 7208 }, { "epoch": 0.36, "learning_rate": 1.4813057848127623e-05, "loss": 0.1279, "step": 7209 }, { "epoch": 0.36, "learning_rate": 1.4811640215075745e-05, "loss": 0.2807, "step": 7210 }, { "epoch": 0.36, "learning_rate": 1.481022245618172e-05, "loss": 0.3794, "step": 7211 }, { "epoch": 0.36, "learning_rate": 1.4808804571482625e-05, "loss": 0.4684, "step": 7212 }, { "epoch": 0.36, "learning_rate": 1.4807386561015544e-05, "loss": 0.2529, "step": 7213 }, { "epoch": 0.36, "learning_rate": 1.4805968424817568e-05, "loss": 0.5216, "step": 7214 }, { "epoch": 0.36, "learning_rate": 1.4804550162925776e-05, "loss": 0.2224, "step": 7215 }, { "epoch": 0.36, "learning_rate": 1.4803131775377276e-05, "loss": 0.1147, "step": 7216 }, { "epoch": 0.36, "learning_rate": 1.4801713262209147e-05, "loss": 0.3665, "step": 7217 }, { "epoch": 0.36, "learning_rate": 1.4800294623458502e-05, "loss": 0.2753, "step": 7218 }, { "epoch": 0.36, "learning_rate": 1.4798875859162439e-05, "loss": 0.5001, "step": 7219 }, { "epoch": 0.36, "learning_rate": 1.4797456969358059e-05, "loss": 0.2927, "step": 7220 }, { "epoch": 0.36, "learning_rate": 1.4796037954082479e-05, "loss": 0.2306, "step": 7221 }, { "epoch": 0.36, "learning_rate": 1.4794618813372808e-05, "loss": 0.5449, "step": 7222 }, { "epoch": 0.36, "learning_rate": 1.4793199547266161e-05, "loss": 0.2474, "step": 7223 }, { "epoch": 0.36, "learning_rate": 1.4791780155799656e-05, "loss": 0.3186, "step": 7224 }, { "epoch": 0.36, "learning_rate": 1.479036063901042e-05, "loss": 0.1183, "step": 7225 }, { "epoch": 0.36, "learning_rate": 1.4788940996935572e-05, "loss": 0.3978, "step": 7226 }, { "epoch": 0.36, "learning_rate": 1.4787521229612247e-05, "loss": 0.2683, "step": 7227 }, { "epoch": 0.36, "learning_rate": 1.4786101337077576e-05, "loss": 0.4416, "step": 7228 }, { "epoch": 0.36, "learning_rate": 1.478468131936869e-05, "loss": 0.5514, "step": 7229 }, { "epoch": 0.36, "learning_rate": 1.4783261176522735e-05, "loss": 0.5045, "step": 7230 }, { "epoch": 0.36, "learning_rate": 1.4781840908576844e-05, "loss": 0.3274, "step": 7231 }, { "epoch": 0.36, "learning_rate": 1.478042051556817e-05, "loss": 0.4158, "step": 7232 }, { "epoch": 0.36, "learning_rate": 1.477899999753386e-05, "loss": 0.2932, "step": 7233 }, { "epoch": 0.36, "learning_rate": 1.4777579354511064e-05, "loss": 0.262, "step": 7234 }, { "epoch": 0.36, "learning_rate": 1.4776158586536937e-05, "loss": 0.3866, "step": 7235 }, { "epoch": 0.36, "learning_rate": 1.4774737693648635e-05, "loss": 0.4254, "step": 7236 }, { "epoch": 0.36, "learning_rate": 1.4773316675883327e-05, "loss": 0.2725, "step": 7237 }, { "epoch": 0.36, "learning_rate": 1.4771895533278169e-05, "loss": 0.4189, "step": 7238 }, { "epoch": 0.36, "learning_rate": 1.4770474265870336e-05, "loss": 0.354, "step": 7239 }, { "epoch": 0.36, "learning_rate": 1.4769052873696995e-05, "loss": 0.1823, "step": 7240 }, { "epoch": 0.36, "learning_rate": 1.4767631356795323e-05, "loss": 0.2388, "step": 7241 }, { "epoch": 0.36, "learning_rate": 1.4766209715202497e-05, "loss": 0.2929, "step": 7242 }, { "epoch": 0.36, "learning_rate": 1.47647879489557e-05, "loss": 0.3126, "step": 7243 }, { "epoch": 0.36, "learning_rate": 1.4763366058092115e-05, "loss": 0.3764, "step": 7244 }, { "epoch": 0.36, "learning_rate": 1.4761944042648928e-05, "loss": 0.3852, "step": 7245 }, { "epoch": 0.36, "learning_rate": 1.4760521902663335e-05, "loss": 0.1574, "step": 7246 }, { "epoch": 0.36, "learning_rate": 1.475909963817252e-05, "loss": 0.2647, "step": 7247 }, { "epoch": 0.36, "learning_rate": 1.4757677249213693e-05, "loss": 0.3954, "step": 7248 }, { "epoch": 0.36, "learning_rate": 1.4756254735824047e-05, "loss": 0.3313, "step": 7249 }, { "epoch": 0.36, "learning_rate": 1.4754832098040788e-05, "loss": 0.3061, "step": 7250 }, { "epoch": 0.36, "learning_rate": 1.4753409335901125e-05, "loss": 0.3011, "step": 7251 }, { "epoch": 0.36, "learning_rate": 1.4751986449442264e-05, "loss": 0.4136, "step": 7252 }, { "epoch": 0.36, "learning_rate": 1.4750563438701424e-05, "loss": 0.27, "step": 7253 }, { "epoch": 0.36, "learning_rate": 1.4749140303715817e-05, "loss": 0.2408, "step": 7254 }, { "epoch": 0.36, "learning_rate": 1.4747717044522667e-05, "loss": 0.5012, "step": 7255 }, { "epoch": 0.36, "learning_rate": 1.4746293661159194e-05, "loss": 0.1707, "step": 7256 }, { "epoch": 0.36, "learning_rate": 1.4744870153662628e-05, "loss": 0.4496, "step": 7257 }, { "epoch": 0.36, "learning_rate": 1.47434465220702e-05, "loss": 0.2369, "step": 7258 }, { "epoch": 0.36, "learning_rate": 1.474202276641914e-05, "loss": 0.268, "step": 7259 }, { "epoch": 0.36, "learning_rate": 1.4740598886746687e-05, "loss": 0.4447, "step": 7260 }, { "epoch": 0.36, "learning_rate": 1.4739174883090074e-05, "loss": 0.4283, "step": 7261 }, { "epoch": 0.36, "learning_rate": 1.4737750755486553e-05, "loss": 0.4708, "step": 7262 }, { "epoch": 0.36, "learning_rate": 1.4736326503973364e-05, "loss": 0.4251, "step": 7263 }, { "epoch": 0.36, "learning_rate": 1.4734902128587762e-05, "loss": 0.5929, "step": 7264 }, { "epoch": 0.36, "learning_rate": 1.4733477629366992e-05, "loss": 0.3978, "step": 7265 }, { "epoch": 0.36, "learning_rate": 1.473205300634832e-05, "loss": 0.2037, "step": 7266 }, { "epoch": 0.36, "learning_rate": 1.4730628259568995e-05, "loss": 0.3558, "step": 7267 }, { "epoch": 0.36, "learning_rate": 1.4729203389066284e-05, "loss": 0.6048, "step": 7268 }, { "epoch": 0.36, "learning_rate": 1.4727778394877456e-05, "loss": 0.4751, "step": 7269 }, { "epoch": 0.36, "learning_rate": 1.4726353277039776e-05, "loss": 0.218, "step": 7270 }, { "epoch": 0.36, "learning_rate": 1.4724928035590514e-05, "loss": 0.3613, "step": 7271 }, { "epoch": 0.36, "learning_rate": 1.4723502670566947e-05, "loss": 0.6152, "step": 7272 }, { "epoch": 0.36, "learning_rate": 1.4722077182006358e-05, "loss": 0.2887, "step": 7273 }, { "epoch": 0.36, "learning_rate": 1.4720651569946021e-05, "loss": 0.4375, "step": 7274 }, { "epoch": 0.36, "learning_rate": 1.4719225834423227e-05, "loss": 0.5345, "step": 7275 }, { "epoch": 0.36, "learning_rate": 1.4717799975475264e-05, "loss": 0.5219, "step": 7276 }, { "epoch": 0.36, "learning_rate": 1.4716373993139418e-05, "loss": 0.2822, "step": 7277 }, { "epoch": 0.36, "learning_rate": 1.4714947887452992e-05, "loss": 0.524, "step": 7278 }, { "epoch": 0.36, "learning_rate": 1.4713521658453276e-05, "loss": 0.3991, "step": 7279 }, { "epoch": 0.36, "learning_rate": 1.4712095306177576e-05, "loss": 0.3755, "step": 7280 }, { "epoch": 0.36, "learning_rate": 1.4710668830663197e-05, "loss": 0.3884, "step": 7281 }, { "epoch": 0.36, "learning_rate": 1.4709242231947443e-05, "loss": 0.3734, "step": 7282 }, { "epoch": 0.36, "learning_rate": 1.4707815510067624e-05, "loss": 0.232, "step": 7283 }, { "epoch": 0.36, "learning_rate": 1.4706388665061057e-05, "loss": 0.3367, "step": 7284 }, { "epoch": 0.36, "learning_rate": 1.4704961696965063e-05, "loss": 0.3215, "step": 7285 }, { "epoch": 0.36, "learning_rate": 1.4703534605816953e-05, "loss": 0.4768, "step": 7286 }, { "epoch": 0.36, "learning_rate": 1.470210739165406e-05, "loss": 0.417, "step": 7287 }, { "epoch": 0.36, "learning_rate": 1.4700680054513706e-05, "loss": 0.3147, "step": 7288 }, { "epoch": 0.36, "learning_rate": 1.469925259443322e-05, "loss": 0.4393, "step": 7289 }, { "epoch": 0.36, "learning_rate": 1.4697825011449936e-05, "loss": 0.5701, "step": 7290 }, { "epoch": 0.36, "learning_rate": 1.4696397305601195e-05, "loss": 0.4436, "step": 7291 }, { "epoch": 0.36, "learning_rate": 1.4694969476924335e-05, "loss": 0.2159, "step": 7292 }, { "epoch": 0.36, "learning_rate": 1.4693541525456694e-05, "loss": 0.1706, "step": 7293 }, { "epoch": 0.36, "learning_rate": 1.4692113451235623e-05, "loss": 0.4813, "step": 7294 }, { "epoch": 0.36, "learning_rate": 1.4690685254298471e-05, "loss": 0.37, "step": 7295 }, { "epoch": 0.36, "learning_rate": 1.468925693468259e-05, "loss": 0.3716, "step": 7296 }, { "epoch": 0.36, "learning_rate": 1.4687828492425335e-05, "loss": 0.2096, "step": 7297 }, { "epoch": 0.36, "learning_rate": 1.4686399927564065e-05, "loss": 0.3024, "step": 7298 }, { "epoch": 0.36, "learning_rate": 1.4684971240136143e-05, "loss": 0.5483, "step": 7299 }, { "epoch": 0.36, "learning_rate": 1.4683542430178936e-05, "loss": 0.4353, "step": 7300 }, { "epoch": 0.36, "learning_rate": 1.4682113497729813e-05, "loss": 0.3071, "step": 7301 }, { "epoch": 0.36, "learning_rate": 1.468068444282614e-05, "loss": 0.5753, "step": 7302 }, { "epoch": 0.36, "learning_rate": 1.4679255265505297e-05, "loss": 0.5112, "step": 7303 }, { "epoch": 0.36, "learning_rate": 1.4677825965804661e-05, "loss": 0.4012, "step": 7304 }, { "epoch": 0.36, "learning_rate": 1.4676396543761614e-05, "loss": 0.4136, "step": 7305 }, { "epoch": 0.36, "learning_rate": 1.4674966999413542e-05, "loss": 0.4129, "step": 7306 }, { "epoch": 0.36, "learning_rate": 1.4673537332797831e-05, "loss": 0.3111, "step": 7307 }, { "epoch": 0.36, "learning_rate": 1.4672107543951872e-05, "loss": 0.2678, "step": 7308 }, { "epoch": 0.36, "learning_rate": 1.4670677632913058e-05, "loss": 0.1882, "step": 7309 }, { "epoch": 0.37, "learning_rate": 1.4669247599718788e-05, "loss": 0.5482, "step": 7310 }, { "epoch": 0.37, "learning_rate": 1.4667817444406463e-05, "loss": 0.3026, "step": 7311 }, { "epoch": 0.37, "learning_rate": 1.4666387167013488e-05, "loss": 0.4322, "step": 7312 }, { "epoch": 0.37, "learning_rate": 1.4664956767577265e-05, "loss": 0.514, "step": 7313 }, { "epoch": 0.37, "learning_rate": 1.4663526246135212e-05, "loss": 0.2046, "step": 7314 }, { "epoch": 0.37, "learning_rate": 1.4662095602724735e-05, "loss": 0.1999, "step": 7315 }, { "epoch": 0.37, "learning_rate": 1.4660664837383256e-05, "loss": 0.4013, "step": 7316 }, { "epoch": 0.37, "learning_rate": 1.465923395014819e-05, "loss": 0.269, "step": 7317 }, { "epoch": 0.37, "learning_rate": 1.4657802941056966e-05, "loss": 0.4469, "step": 7318 }, { "epoch": 0.37, "learning_rate": 1.4656371810147003e-05, "loss": 0.4652, "step": 7319 }, { "epoch": 0.37, "learning_rate": 1.4654940557455737e-05, "loss": 0.279, "step": 7320 }, { "epoch": 0.37, "learning_rate": 1.4653509183020599e-05, "loss": 0.3386, "step": 7321 }, { "epoch": 0.37, "learning_rate": 1.4652077686879019e-05, "loss": 0.3264, "step": 7322 }, { "epoch": 0.37, "learning_rate": 1.4650646069068441e-05, "loss": 0.3006, "step": 7323 }, { "epoch": 0.37, "learning_rate": 1.4649214329626309e-05, "loss": 0.3985, "step": 7324 }, { "epoch": 0.37, "learning_rate": 1.4647782468590061e-05, "loss": 0.4276, "step": 7325 }, { "epoch": 0.37, "learning_rate": 1.4646350485997153e-05, "loss": 0.3545, "step": 7326 }, { "epoch": 0.37, "learning_rate": 1.4644918381885033e-05, "loss": 0.4016, "step": 7327 }, { "epoch": 0.37, "learning_rate": 1.464348615629116e-05, "loss": 0.375, "step": 7328 }, { "epoch": 0.37, "learning_rate": 1.4642053809252984e-05, "loss": 0.3053, "step": 7329 }, { "epoch": 0.37, "learning_rate": 1.4640621340807973e-05, "loss": 0.4994, "step": 7330 }, { "epoch": 0.37, "learning_rate": 1.463918875099359e-05, "loss": 0.4035, "step": 7331 }, { "epoch": 0.37, "learning_rate": 1.46377560398473e-05, "loss": 0.4834, "step": 7332 }, { "epoch": 0.37, "learning_rate": 1.4636323207406573e-05, "loss": 0.2975, "step": 7333 }, { "epoch": 0.37, "learning_rate": 1.4634890253708887e-05, "loss": 0.2359, "step": 7334 }, { "epoch": 0.37, "learning_rate": 1.4633457178791717e-05, "loss": 0.1826, "step": 7335 }, { "epoch": 0.37, "learning_rate": 1.463202398269254e-05, "loss": 0.1638, "step": 7336 }, { "epoch": 0.37, "learning_rate": 1.4630590665448847e-05, "loss": 0.3098, "step": 7337 }, { "epoch": 0.37, "learning_rate": 1.4629157227098118e-05, "loss": 0.3572, "step": 7338 }, { "epoch": 0.37, "learning_rate": 1.4627723667677846e-05, "loss": 0.3259, "step": 7339 }, { "epoch": 0.37, "learning_rate": 1.4626289987225524e-05, "loss": 0.4301, "step": 7340 }, { "epoch": 0.37, "learning_rate": 1.4624856185778643e-05, "loss": 0.1708, "step": 7341 }, { "epoch": 0.37, "learning_rate": 1.4623422263374708e-05, "loss": 0.454, "step": 7342 }, { "epoch": 0.37, "learning_rate": 1.4621988220051222e-05, "loss": 0.4695, "step": 7343 }, { "epoch": 0.37, "learning_rate": 1.4620554055845685e-05, "loss": 0.5499, "step": 7344 }, { "epoch": 0.37, "learning_rate": 1.4619119770795607e-05, "loss": 0.4936, "step": 7345 }, { "epoch": 0.37, "learning_rate": 1.4617685364938504e-05, "loss": 0.5599, "step": 7346 }, { "epoch": 0.37, "learning_rate": 1.4616250838311885e-05, "loss": 0.4968, "step": 7347 }, { "epoch": 0.37, "learning_rate": 1.4614816190953276e-05, "loss": 0.2992, "step": 7348 }, { "epoch": 0.37, "learning_rate": 1.461338142290019e-05, "loss": 0.4887, "step": 7349 }, { "epoch": 0.37, "learning_rate": 1.4611946534190158e-05, "loss": 0.2871, "step": 7350 }, { "epoch": 0.37, "learning_rate": 1.4610511524860705e-05, "loss": 0.3146, "step": 7351 }, { "epoch": 0.37, "learning_rate": 1.4609076394949357e-05, "loss": 0.6349, "step": 7352 }, { "epoch": 0.37, "learning_rate": 1.4607641144493659e-05, "loss": 0.167, "step": 7353 }, { "epoch": 0.37, "learning_rate": 1.460620577353114e-05, "loss": 0.4969, "step": 7354 }, { "epoch": 0.37, "learning_rate": 1.4604770282099338e-05, "loss": 0.2105, "step": 7355 }, { "epoch": 0.37, "learning_rate": 1.4603334670235805e-05, "loss": 0.2327, "step": 7356 }, { "epoch": 0.37, "learning_rate": 1.4601898937978083e-05, "loss": 0.4764, "step": 7357 }, { "epoch": 0.37, "learning_rate": 1.460046308536372e-05, "loss": 0.4553, "step": 7358 }, { "epoch": 0.37, "learning_rate": 1.459902711243027e-05, "loss": 0.3382, "step": 7359 }, { "epoch": 0.37, "learning_rate": 1.459759101921529e-05, "loss": 0.5168, "step": 7360 }, { "epoch": 0.37, "learning_rate": 1.4596154805756339e-05, "loss": 0.2855, "step": 7361 }, { "epoch": 0.37, "learning_rate": 1.459471847209098e-05, "loss": 0.4766, "step": 7362 }, { "epoch": 0.37, "learning_rate": 1.4593282018256774e-05, "loss": 0.1492, "step": 7363 }, { "epoch": 0.37, "learning_rate": 1.4591845444291298e-05, "loss": 0.463, "step": 7364 }, { "epoch": 0.37, "learning_rate": 1.4590408750232115e-05, "loss": 0.3731, "step": 7365 }, { "epoch": 0.37, "learning_rate": 1.4588971936116803e-05, "loss": 0.5689, "step": 7366 }, { "epoch": 0.37, "learning_rate": 1.4587535001982943e-05, "loss": 0.4675, "step": 7367 }, { "epoch": 0.37, "learning_rate": 1.4586097947868112e-05, "loss": 0.2165, "step": 7368 }, { "epoch": 0.37, "learning_rate": 1.4584660773809898e-05, "loss": 0.3316, "step": 7369 }, { "epoch": 0.37, "learning_rate": 1.4583223479845883e-05, "loss": 0.447, "step": 7370 }, { "epoch": 0.37, "learning_rate": 1.4581786066013663e-05, "loss": 0.4591, "step": 7371 }, { "epoch": 0.37, "learning_rate": 1.458034853235083e-05, "loss": 0.3976, "step": 7372 }, { "epoch": 0.37, "learning_rate": 1.4578910878894979e-05, "loss": 0.4834, "step": 7373 }, { "epoch": 0.37, "learning_rate": 1.457747310568371e-05, "loss": 0.306, "step": 7374 }, { "epoch": 0.37, "learning_rate": 1.4576035212754629e-05, "loss": 0.1887, "step": 7375 }, { "epoch": 0.37, "learning_rate": 1.4574597200145342e-05, "loss": 0.3493, "step": 7376 }, { "epoch": 0.37, "learning_rate": 1.4573159067893454e-05, "loss": 0.1525, "step": 7377 }, { "epoch": 0.37, "learning_rate": 1.457172081603658e-05, "loss": 0.1556, "step": 7378 }, { "epoch": 0.37, "learning_rate": 1.457028244461234e-05, "loss": 0.386, "step": 7379 }, { "epoch": 0.37, "learning_rate": 1.4568843953658347e-05, "loss": 0.2526, "step": 7380 }, { "epoch": 0.37, "learning_rate": 1.4567405343212225e-05, "loss": 0.3946, "step": 7381 }, { "epoch": 0.37, "learning_rate": 1.4565966613311594e-05, "loss": 0.3231, "step": 7382 }, { "epoch": 0.37, "learning_rate": 1.4564527763994092e-05, "loss": 0.268, "step": 7383 }, { "epoch": 0.37, "learning_rate": 1.456308879529734e-05, "loss": 0.3208, "step": 7384 }, { "epoch": 0.37, "learning_rate": 1.4561649707258982e-05, "loss": 0.4518, "step": 7385 }, { "epoch": 0.37, "learning_rate": 1.4560210499916646e-05, "loss": 0.3966, "step": 7386 }, { "epoch": 0.37, "learning_rate": 1.455877117330798e-05, "loss": 0.5919, "step": 7387 }, { "epoch": 0.37, "learning_rate": 1.4557331727470623e-05, "loss": 0.2229, "step": 7388 }, { "epoch": 0.37, "learning_rate": 1.4555892162442227e-05, "loss": 0.4506, "step": 7389 }, { "epoch": 0.37, "learning_rate": 1.4554452478260436e-05, "loss": 0.4173, "step": 7390 }, { "epoch": 0.37, "learning_rate": 1.4553012674962906e-05, "loss": 0.3341, "step": 7391 }, { "epoch": 0.37, "learning_rate": 1.4551572752587293e-05, "loss": 0.2444, "step": 7392 }, { "epoch": 0.37, "learning_rate": 1.4550132711171259e-05, "loss": 0.6466, "step": 7393 }, { "epoch": 0.37, "learning_rate": 1.4548692550752459e-05, "loss": 0.4308, "step": 7394 }, { "epoch": 0.37, "learning_rate": 1.4547252271368564e-05, "loss": 0.38, "step": 7395 }, { "epoch": 0.37, "learning_rate": 1.4545811873057246e-05, "loss": 0.1065, "step": 7396 }, { "epoch": 0.37, "learning_rate": 1.4544371355856168e-05, "loss": 0.5284, "step": 7397 }, { "epoch": 0.37, "learning_rate": 1.4542930719803008e-05, "loss": 0.3288, "step": 7398 }, { "epoch": 0.37, "learning_rate": 1.4541489964935449e-05, "loss": 0.3506, "step": 7399 }, { "epoch": 0.37, "learning_rate": 1.4540049091291167e-05, "loss": 0.4622, "step": 7400 }, { "epoch": 0.37, "learning_rate": 1.4538608098907848e-05, "loss": 0.3315, "step": 7401 }, { "epoch": 0.37, "learning_rate": 1.4537166987823175e-05, "loss": 0.5312, "step": 7402 }, { "epoch": 0.37, "learning_rate": 1.4535725758074846e-05, "loss": 0.3361, "step": 7403 }, { "epoch": 0.37, "learning_rate": 1.453428440970055e-05, "loss": 0.4256, "step": 7404 }, { "epoch": 0.37, "learning_rate": 1.4532842942737984e-05, "loss": 0.3506, "step": 7405 }, { "epoch": 0.37, "learning_rate": 1.4531401357224848e-05, "loss": 0.2654, "step": 7406 }, { "epoch": 0.37, "learning_rate": 1.4529959653198842e-05, "loss": 0.3604, "step": 7407 }, { "epoch": 0.37, "learning_rate": 1.4528517830697678e-05, "loss": 0.6479, "step": 7408 }, { "epoch": 0.37, "learning_rate": 1.4527075889759056e-05, "loss": 0.2582, "step": 7409 }, { "epoch": 0.37, "learning_rate": 1.45256338304207e-05, "loss": 0.4598, "step": 7410 }, { "epoch": 0.37, "learning_rate": 1.4524191652720314e-05, "loss": 0.4938, "step": 7411 }, { "epoch": 0.37, "learning_rate": 1.4522749356695625e-05, "loss": 0.3295, "step": 7412 }, { "epoch": 0.37, "learning_rate": 1.4521306942384348e-05, "loss": 0.2638, "step": 7413 }, { "epoch": 0.37, "learning_rate": 1.451986440982421e-05, "loss": 0.1999, "step": 7414 }, { "epoch": 0.37, "learning_rate": 1.4518421759052938e-05, "loss": 0.5468, "step": 7415 }, { "epoch": 0.37, "learning_rate": 1.4516978990108264e-05, "loss": 0.1676, "step": 7416 }, { "epoch": 0.37, "learning_rate": 1.4515536103027921e-05, "loss": 0.1953, "step": 7417 }, { "epoch": 0.37, "learning_rate": 1.4514093097849644e-05, "loss": 0.4209, "step": 7418 }, { "epoch": 0.37, "learning_rate": 1.4512649974611174e-05, "loss": 0.2715, "step": 7419 }, { "epoch": 0.37, "learning_rate": 1.4511206733350256e-05, "loss": 0.3041, "step": 7420 }, { "epoch": 0.37, "learning_rate": 1.4509763374104632e-05, "loss": 0.2453, "step": 7421 }, { "epoch": 0.37, "learning_rate": 1.4508319896912058e-05, "loss": 0.3004, "step": 7422 }, { "epoch": 0.37, "learning_rate": 1.4506876301810278e-05, "loss": 0.3278, "step": 7423 }, { "epoch": 0.37, "learning_rate": 1.4505432588837053e-05, "loss": 0.4045, "step": 7424 }, { "epoch": 0.37, "learning_rate": 1.4503988758030138e-05, "loss": 0.5648, "step": 7425 }, { "epoch": 0.37, "learning_rate": 1.45025448094273e-05, "loss": 0.3852, "step": 7426 }, { "epoch": 0.37, "learning_rate": 1.4501100743066295e-05, "loss": 0.5083, "step": 7427 }, { "epoch": 0.37, "learning_rate": 1.4499656558984898e-05, "loss": 0.4667, "step": 7428 }, { "epoch": 0.37, "learning_rate": 1.4498212257220878e-05, "loss": 0.4752, "step": 7429 }, { "epoch": 0.37, "learning_rate": 1.4496767837812007e-05, "loss": 0.3499, "step": 7430 }, { "epoch": 0.37, "learning_rate": 1.4495323300796065e-05, "loss": 0.2688, "step": 7431 }, { "epoch": 0.37, "learning_rate": 1.4493878646210826e-05, "loss": 0.4389, "step": 7432 }, { "epoch": 0.37, "learning_rate": 1.4492433874094078e-05, "loss": 0.6972, "step": 7433 }, { "epoch": 0.37, "learning_rate": 1.4490988984483607e-05, "loss": 0.3291, "step": 7434 }, { "epoch": 0.37, "learning_rate": 1.4489543977417202e-05, "loss": 0.3134, "step": 7435 }, { "epoch": 0.37, "learning_rate": 1.4488098852932653e-05, "loss": 0.3384, "step": 7436 }, { "epoch": 0.37, "learning_rate": 1.4486653611067759e-05, "loss": 0.4533, "step": 7437 }, { "epoch": 0.37, "learning_rate": 1.4485208251860314e-05, "loss": 0.1342, "step": 7438 }, { "epoch": 0.37, "learning_rate": 1.448376277534812e-05, "loss": 0.3591, "step": 7439 }, { "epoch": 0.37, "learning_rate": 1.4482317181568988e-05, "loss": 0.2462, "step": 7440 }, { "epoch": 0.37, "learning_rate": 1.448087147056072e-05, "loss": 0.3256, "step": 7441 }, { "epoch": 0.37, "learning_rate": 1.4479425642361124e-05, "loss": 0.296, "step": 7442 }, { "epoch": 0.37, "learning_rate": 1.447797969700802e-05, "loss": 0.1687, "step": 7443 }, { "epoch": 0.37, "learning_rate": 1.4476533634539223e-05, "loss": 0.3295, "step": 7444 }, { "epoch": 0.37, "learning_rate": 1.447508745499255e-05, "loss": 0.3679, "step": 7445 }, { "epoch": 0.37, "learning_rate": 1.4473641158405827e-05, "loss": 0.3655, "step": 7446 }, { "epoch": 0.37, "learning_rate": 1.4472194744816879e-05, "loss": 0.3596, "step": 7447 }, { "epoch": 0.37, "learning_rate": 1.4470748214263533e-05, "loss": 0.3121, "step": 7448 }, { "epoch": 0.37, "learning_rate": 1.4469301566783625e-05, "loss": 0.5245, "step": 7449 }, { "epoch": 0.37, "learning_rate": 1.4467854802414987e-05, "loss": 0.3552, "step": 7450 }, { "epoch": 0.37, "learning_rate": 1.446640792119546e-05, "loss": 0.3266, "step": 7451 }, { "epoch": 0.37, "learning_rate": 1.4464960923162883e-05, "loss": 0.4304, "step": 7452 }, { "epoch": 0.37, "learning_rate": 1.4463513808355102e-05, "loss": 0.3836, "step": 7453 }, { "epoch": 0.37, "learning_rate": 1.4462066576809964e-05, "loss": 0.3093, "step": 7454 }, { "epoch": 0.37, "learning_rate": 1.4460619228565316e-05, "loss": 0.1667, "step": 7455 }, { "epoch": 0.37, "learning_rate": 1.4459171763659016e-05, "loss": 0.537, "step": 7456 }, { "epoch": 0.37, "learning_rate": 1.4457724182128919e-05, "loss": 0.2387, "step": 7457 }, { "epoch": 0.37, "learning_rate": 1.4456276484012883e-05, "loss": 0.3397, "step": 7458 }, { "epoch": 0.37, "learning_rate": 1.4454828669348775e-05, "loss": 0.308, "step": 7459 }, { "epoch": 0.37, "learning_rate": 1.4453380738174456e-05, "loss": 0.6708, "step": 7460 }, { "epoch": 0.37, "learning_rate": 1.4451932690527798e-05, "loss": 0.2356, "step": 7461 }, { "epoch": 0.37, "learning_rate": 1.445048452644667e-05, "loss": 0.5151, "step": 7462 }, { "epoch": 0.37, "learning_rate": 1.444903624596895e-05, "loss": 0.4512, "step": 7463 }, { "epoch": 0.37, "learning_rate": 1.4447587849132511e-05, "loss": 0.3598, "step": 7464 }, { "epoch": 0.37, "learning_rate": 1.4446139335975243e-05, "loss": 0.2919, "step": 7465 }, { "epoch": 0.37, "learning_rate": 1.444469070653502e-05, "loss": 0.0963, "step": 7466 }, { "epoch": 0.37, "learning_rate": 1.4443241960849735e-05, "loss": 0.4288, "step": 7467 }, { "epoch": 0.37, "learning_rate": 1.4441793098957272e-05, "loss": 0.3689, "step": 7468 }, { "epoch": 0.37, "learning_rate": 1.444034412089553e-05, "loss": 0.5775, "step": 7469 }, { "epoch": 0.37, "learning_rate": 1.4438895026702407e-05, "loss": 0.1492, "step": 7470 }, { "epoch": 0.37, "learning_rate": 1.4437445816415794e-05, "loss": 0.1913, "step": 7471 }, { "epoch": 0.37, "learning_rate": 1.44359964900736e-05, "loss": 0.4953, "step": 7472 }, { "epoch": 0.37, "learning_rate": 1.4434547047713726e-05, "loss": 0.4663, "step": 7473 }, { "epoch": 0.37, "learning_rate": 1.4433097489374084e-05, "loss": 0.1727, "step": 7474 }, { "epoch": 0.37, "learning_rate": 1.4431647815092583e-05, "loss": 0.2516, "step": 7475 }, { "epoch": 0.37, "learning_rate": 1.4430198024907138e-05, "loss": 0.2405, "step": 7476 }, { "epoch": 0.37, "learning_rate": 1.4428748118855666e-05, "loss": 0.3035, "step": 7477 }, { "epoch": 0.37, "learning_rate": 1.442729809697609e-05, "loss": 0.5186, "step": 7478 }, { "epoch": 0.37, "learning_rate": 1.4425847959306328e-05, "loss": 0.4125, "step": 7479 }, { "epoch": 0.37, "learning_rate": 1.4424397705884307e-05, "loss": 0.3742, "step": 7480 }, { "epoch": 0.37, "learning_rate": 1.4422947336747966e-05, "loss": 0.4314, "step": 7481 }, { "epoch": 0.37, "learning_rate": 1.4421496851935223e-05, "loss": 0.4519, "step": 7482 }, { "epoch": 0.37, "learning_rate": 1.4420046251484026e-05, "loss": 0.4788, "step": 7483 }, { "epoch": 0.37, "learning_rate": 1.4418595535432305e-05, "loss": 0.5445, "step": 7484 }, { "epoch": 0.37, "learning_rate": 1.4417144703818007e-05, "loss": 0.4284, "step": 7485 }, { "epoch": 0.37, "learning_rate": 1.4415693756679074e-05, "loss": 0.5994, "step": 7486 }, { "epoch": 0.37, "learning_rate": 1.4414242694053455e-05, "loss": 0.2994, "step": 7487 }, { "epoch": 0.37, "learning_rate": 1.44127915159791e-05, "loss": 0.4035, "step": 7488 }, { "epoch": 0.37, "learning_rate": 1.4411340222493964e-05, "loss": 0.1696, "step": 7489 }, { "epoch": 0.37, "learning_rate": 1.4409888813636001e-05, "loss": 0.3769, "step": 7490 }, { "epoch": 0.37, "learning_rate": 1.4408437289443171e-05, "loss": 0.1749, "step": 7491 }, { "epoch": 0.37, "learning_rate": 1.4406985649953439e-05, "loss": 0.4439, "step": 7492 }, { "epoch": 0.37, "learning_rate": 1.4405533895204768e-05, "loss": 0.1707, "step": 7493 }, { "epoch": 0.37, "learning_rate": 1.4404082025235128e-05, "loss": 0.4376, "step": 7494 }, { "epoch": 0.37, "learning_rate": 1.4402630040082496e-05, "loss": 0.4431, "step": 7495 }, { "epoch": 0.37, "learning_rate": 1.4401177939784834e-05, "loss": 0.2565, "step": 7496 }, { "epoch": 0.37, "learning_rate": 1.4399725724380134e-05, "loss": 0.3723, "step": 7497 }, { "epoch": 0.37, "learning_rate": 1.4398273393906365e-05, "loss": 0.5152, "step": 7498 }, { "epoch": 0.37, "learning_rate": 1.439682094840152e-05, "loss": 0.2731, "step": 7499 }, { "epoch": 0.37, "learning_rate": 1.4395368387903578e-05, "loss": 0.4692, "step": 7500 }, { "epoch": 0.37, "learning_rate": 1.4393915712450537e-05, "loss": 0.5314, "step": 7501 }, { "epoch": 0.37, "learning_rate": 1.4392462922080384e-05, "loss": 0.4088, "step": 7502 }, { "epoch": 0.37, "learning_rate": 1.4391010016831118e-05, "loss": 0.3694, "step": 7503 }, { "epoch": 0.37, "learning_rate": 1.4389556996740735e-05, "loss": 0.3121, "step": 7504 }, { "epoch": 0.37, "learning_rate": 1.4388103861847235e-05, "loss": 0.2301, "step": 7505 }, { "epoch": 0.37, "learning_rate": 1.4386650612188628e-05, "loss": 0.301, "step": 7506 }, { "epoch": 0.37, "learning_rate": 1.438519724780292e-05, "loss": 0.3742, "step": 7507 }, { "epoch": 0.37, "learning_rate": 1.4383743768728123e-05, "loss": 0.2213, "step": 7508 }, { "epoch": 0.37, "learning_rate": 1.4382290175002247e-05, "loss": 0.3365, "step": 7509 }, { "epoch": 0.37, "learning_rate": 1.438083646666331e-05, "loss": 0.158, "step": 7510 }, { "epoch": 0.38, "learning_rate": 1.4379382643749338e-05, "loss": 0.2058, "step": 7511 }, { "epoch": 0.38, "learning_rate": 1.4377928706298345e-05, "loss": 0.4182, "step": 7512 }, { "epoch": 0.38, "learning_rate": 1.4376474654348362e-05, "loss": 0.2353, "step": 7513 }, { "epoch": 0.38, "learning_rate": 1.4375020487937417e-05, "loss": 0.3062, "step": 7514 }, { "epoch": 0.38, "learning_rate": 1.4373566207103543e-05, "loss": 0.3589, "step": 7515 }, { "epoch": 0.38, "learning_rate": 1.4372111811884767e-05, "loss": 0.2319, "step": 7516 }, { "epoch": 0.38, "learning_rate": 1.4370657302319137e-05, "loss": 0.3463, "step": 7517 }, { "epoch": 0.38, "learning_rate": 1.4369202678444692e-05, "loss": 0.4276, "step": 7518 }, { "epoch": 0.38, "learning_rate": 1.436774794029947e-05, "loss": 0.1777, "step": 7519 }, { "epoch": 0.38, "learning_rate": 1.4366293087921523e-05, "loss": 0.5669, "step": 7520 }, { "epoch": 0.38, "learning_rate": 1.4364838121348899e-05, "loss": 0.4048, "step": 7521 }, { "epoch": 0.38, "learning_rate": 1.4363383040619652e-05, "loss": 0.3208, "step": 7522 }, { "epoch": 0.38, "learning_rate": 1.4361927845771833e-05, "loss": 0.4349, "step": 7523 }, { "epoch": 0.38, "learning_rate": 1.4360472536843508e-05, "loss": 0.277, "step": 7524 }, { "epoch": 0.38, "learning_rate": 1.4359017113872733e-05, "loss": 0.4112, "step": 7525 }, { "epoch": 0.38, "learning_rate": 1.4357561576897577e-05, "loss": 0.532, "step": 7526 }, { "epoch": 0.38, "learning_rate": 1.4356105925956103e-05, "loss": 0.2606, "step": 7527 }, { "epoch": 0.38, "learning_rate": 1.4354650161086383e-05, "loss": 0.5418, "step": 7528 }, { "epoch": 0.38, "learning_rate": 1.4353194282326495e-05, "loss": 0.2759, "step": 7529 }, { "epoch": 0.38, "learning_rate": 1.435173828971451e-05, "loss": 0.4123, "step": 7530 }, { "epoch": 0.38, "learning_rate": 1.435028218328851e-05, "loss": 0.2725, "step": 7531 }, { "epoch": 0.38, "learning_rate": 1.4348825963086577e-05, "loss": 0.312, "step": 7532 }, { "epoch": 0.38, "learning_rate": 1.4347369629146797e-05, "loss": 0.4947, "step": 7533 }, { "epoch": 0.38, "learning_rate": 1.4345913181507261e-05, "loss": 0.4255, "step": 7534 }, { "epoch": 0.38, "learning_rate": 1.4344456620206053e-05, "loss": 0.2951, "step": 7535 }, { "epoch": 0.38, "learning_rate": 1.4342999945281275e-05, "loss": 0.5285, "step": 7536 }, { "epoch": 0.38, "learning_rate": 1.434154315677102e-05, "loss": 0.2787, "step": 7537 }, { "epoch": 0.38, "learning_rate": 1.4340086254713392e-05, "loss": 0.2584, "step": 7538 }, { "epoch": 0.38, "learning_rate": 1.4338629239146493e-05, "loss": 0.3067, "step": 7539 }, { "epoch": 0.38, "learning_rate": 1.433717211010843e-05, "loss": 0.4036, "step": 7540 }, { "epoch": 0.38, "learning_rate": 1.4335714867637309e-05, "loss": 0.4301, "step": 7541 }, { "epoch": 0.38, "learning_rate": 1.4334257511771244e-05, "loss": 0.5519, "step": 7542 }, { "epoch": 0.38, "learning_rate": 1.4332800042548352e-05, "loss": 0.3061, "step": 7543 }, { "epoch": 0.38, "learning_rate": 1.4331342460006747e-05, "loss": 0.3481, "step": 7544 }, { "epoch": 0.38, "learning_rate": 1.4329884764184556e-05, "loss": 0.4407, "step": 7545 }, { "epoch": 0.38, "learning_rate": 1.4328426955119899e-05, "loss": 0.4201, "step": 7546 }, { "epoch": 0.38, "learning_rate": 1.4326969032850906e-05, "loss": 0.1434, "step": 7547 }, { "epoch": 0.38, "learning_rate": 1.4325510997415701e-05, "loss": 0.3068, "step": 7548 }, { "epoch": 0.38, "learning_rate": 1.4324052848852426e-05, "loss": 0.5283, "step": 7549 }, { "epoch": 0.38, "learning_rate": 1.432259458719921e-05, "loss": 0.4943, "step": 7550 }, { "epoch": 0.38, "learning_rate": 1.4321136212494196e-05, "loss": 0.4201, "step": 7551 }, { "epoch": 0.38, "learning_rate": 1.4319677724775522e-05, "loss": 0.3952, "step": 7552 }, { "epoch": 0.38, "learning_rate": 1.4318219124081335e-05, "loss": 0.3617, "step": 7553 }, { "epoch": 0.38, "learning_rate": 1.4316760410449783e-05, "loss": 0.3565, "step": 7554 }, { "epoch": 0.38, "learning_rate": 1.4315301583919014e-05, "loss": 0.1274, "step": 7555 }, { "epoch": 0.38, "learning_rate": 1.431384264452719e-05, "loss": 0.2553, "step": 7556 }, { "epoch": 0.38, "learning_rate": 1.4312383592312456e-05, "loss": 0.4792, "step": 7557 }, { "epoch": 0.38, "learning_rate": 1.431092442731298e-05, "loss": 0.3824, "step": 7558 }, { "epoch": 0.38, "learning_rate": 1.4309465149566921e-05, "loss": 0.5277, "step": 7559 }, { "epoch": 0.38, "learning_rate": 1.4308005759112445e-05, "loss": 0.3544, "step": 7560 }, { "epoch": 0.38, "learning_rate": 1.430654625598772e-05, "loss": 0.3907, "step": 7561 }, { "epoch": 0.38, "learning_rate": 1.4305086640230923e-05, "loss": 0.4417, "step": 7562 }, { "epoch": 0.38, "learning_rate": 1.4303626911880222e-05, "loss": 0.4274, "step": 7563 }, { "epoch": 0.38, "learning_rate": 1.4302167070973792e-05, "loss": 0.5877, "step": 7564 }, { "epoch": 0.38, "learning_rate": 1.4300707117549819e-05, "loss": 0.3542, "step": 7565 }, { "epoch": 0.38, "learning_rate": 1.4299247051646488e-05, "loss": 0.3022, "step": 7566 }, { "epoch": 0.38, "learning_rate": 1.4297786873301976e-05, "loss": 0.3926, "step": 7567 }, { "epoch": 0.38, "learning_rate": 1.4296326582554479e-05, "loss": 0.3083, "step": 7568 }, { "epoch": 0.38, "learning_rate": 1.4294866179442188e-05, "loss": 0.2878, "step": 7569 }, { "epoch": 0.38, "learning_rate": 1.4293405664003296e-05, "loss": 0.359, "step": 7570 }, { "epoch": 0.38, "learning_rate": 1.4291945036276003e-05, "loss": 0.4072, "step": 7571 }, { "epoch": 0.38, "learning_rate": 1.4290484296298511e-05, "loss": 0.1725, "step": 7572 }, { "epoch": 0.38, "learning_rate": 1.4289023444109018e-05, "loss": 0.5428, "step": 7573 }, { "epoch": 0.38, "learning_rate": 1.4287562479745737e-05, "loss": 0.2195, "step": 7574 }, { "epoch": 0.38, "learning_rate": 1.4286101403246877e-05, "loss": 0.4279, "step": 7575 }, { "epoch": 0.38, "learning_rate": 1.4284640214650647e-05, "loss": 0.2119, "step": 7576 }, { "epoch": 0.38, "learning_rate": 1.4283178913995265e-05, "loss": 0.3046, "step": 7577 }, { "epoch": 0.38, "learning_rate": 1.4281717501318944e-05, "loss": 0.1874, "step": 7578 }, { "epoch": 0.38, "learning_rate": 1.4280255976659913e-05, "loss": 0.3901, "step": 7579 }, { "epoch": 0.38, "learning_rate": 1.427879434005639e-05, "loss": 0.5403, "step": 7580 }, { "epoch": 0.38, "learning_rate": 1.4277332591546608e-05, "loss": 0.4681, "step": 7581 }, { "epoch": 0.38, "learning_rate": 1.4275870731168797e-05, "loss": 0.4221, "step": 7582 }, { "epoch": 0.38, "learning_rate": 1.4274408758961182e-05, "loss": 0.5118, "step": 7583 }, { "epoch": 0.38, "learning_rate": 1.427294667496201e-05, "loss": 0.2583, "step": 7584 }, { "epoch": 0.38, "learning_rate": 1.427148447920951e-05, "loss": 0.2931, "step": 7585 }, { "epoch": 0.38, "learning_rate": 1.4270022171741928e-05, "loss": 0.6938, "step": 7586 }, { "epoch": 0.38, "learning_rate": 1.426855975259751e-05, "loss": 0.3511, "step": 7587 }, { "epoch": 0.38, "learning_rate": 1.4267097221814506e-05, "loss": 0.2632, "step": 7588 }, { "epoch": 0.38, "learning_rate": 1.4265634579431157e-05, "loss": 0.4981, "step": 7589 }, { "epoch": 0.38, "learning_rate": 1.4264171825485727e-05, "loss": 0.3821, "step": 7590 }, { "epoch": 0.38, "learning_rate": 1.4262708960016466e-05, "loss": 0.2484, "step": 7591 }, { "epoch": 0.38, "learning_rate": 1.4261245983061636e-05, "loss": 0.4549, "step": 7592 }, { "epoch": 0.38, "learning_rate": 1.4259782894659497e-05, "loss": 0.1936, "step": 7593 }, { "epoch": 0.38, "learning_rate": 1.4258319694848317e-05, "loss": 0.2177, "step": 7594 }, { "epoch": 0.38, "learning_rate": 1.4256856383666362e-05, "loss": 0.5016, "step": 7595 }, { "epoch": 0.38, "learning_rate": 1.4255392961151904e-05, "loss": 0.3287, "step": 7596 }, { "epoch": 0.38, "learning_rate": 1.4253929427343218e-05, "loss": 0.3541, "step": 7597 }, { "epoch": 0.38, "learning_rate": 1.4252465782278577e-05, "loss": 0.5565, "step": 7598 }, { "epoch": 0.38, "learning_rate": 1.4251002025996268e-05, "loss": 0.377, "step": 7599 }, { "epoch": 0.38, "learning_rate": 1.4249538158534566e-05, "loss": 0.543, "step": 7600 }, { "epoch": 0.38, "learning_rate": 1.4248074179931757e-05, "loss": 0.3093, "step": 7601 }, { "epoch": 0.38, "learning_rate": 1.4246610090226133e-05, "loss": 0.4908, "step": 7602 }, { "epoch": 0.38, "learning_rate": 1.4245145889455984e-05, "loss": 0.2854, "step": 7603 }, { "epoch": 0.38, "learning_rate": 1.4243681577659608e-05, "loss": 0.538, "step": 7604 }, { "epoch": 0.38, "learning_rate": 1.4242217154875292e-05, "loss": 0.2922, "step": 7605 }, { "epoch": 0.38, "learning_rate": 1.4240752621141347e-05, "loss": 0.1932, "step": 7606 }, { "epoch": 0.38, "learning_rate": 1.4239287976496072e-05, "loss": 0.3807, "step": 7607 }, { "epoch": 0.38, "learning_rate": 1.4237823220977771e-05, "loss": 0.4349, "step": 7608 }, { "epoch": 0.38, "learning_rate": 1.4236358354624755e-05, "loss": 0.3236, "step": 7609 }, { "epoch": 0.38, "learning_rate": 1.4234893377475334e-05, "loss": 0.5478, "step": 7610 }, { "epoch": 0.38, "learning_rate": 1.4233428289567823e-05, "loss": 0.3313, "step": 7611 }, { "epoch": 0.38, "learning_rate": 1.4231963090940545e-05, "loss": 0.3702, "step": 7612 }, { "epoch": 0.38, "learning_rate": 1.423049778163181e-05, "loss": 0.4076, "step": 7613 }, { "epoch": 0.38, "learning_rate": 1.422903236167995e-05, "loss": 0.5625, "step": 7614 }, { "epoch": 0.38, "learning_rate": 1.4227566831123283e-05, "loss": 0.5257, "step": 7615 }, { "epoch": 0.38, "learning_rate": 1.4226101190000143e-05, "loss": 0.4265, "step": 7616 }, { "epoch": 0.38, "learning_rate": 1.4224635438348864e-05, "loss": 0.3163, "step": 7617 }, { "epoch": 0.38, "learning_rate": 1.4223169576207777e-05, "loss": 0.3388, "step": 7618 }, { "epoch": 0.38, "learning_rate": 1.422170360361522e-05, "loss": 0.6999, "step": 7619 }, { "epoch": 0.38, "learning_rate": 1.4220237520609539e-05, "loss": 0.4148, "step": 7620 }, { "epoch": 0.38, "learning_rate": 1.4218771327229069e-05, "loss": 0.4676, "step": 7621 }, { "epoch": 0.38, "learning_rate": 1.4217305023512163e-05, "loss": 0.3798, "step": 7622 }, { "epoch": 0.38, "learning_rate": 1.4215838609497168e-05, "loss": 0.2827, "step": 7623 }, { "epoch": 0.38, "learning_rate": 1.4214372085222433e-05, "loss": 0.2824, "step": 7624 }, { "epoch": 0.38, "learning_rate": 1.421290545072632e-05, "loss": 0.3724, "step": 7625 }, { "epoch": 0.38, "learning_rate": 1.4211438706047181e-05, "loss": 0.5282, "step": 7626 }, { "epoch": 0.38, "learning_rate": 1.4209971851223379e-05, "loss": 0.3975, "step": 7627 }, { "epoch": 0.38, "learning_rate": 1.4208504886293276e-05, "loss": 0.4188, "step": 7628 }, { "epoch": 0.38, "learning_rate": 1.4207037811295239e-05, "loss": 0.4652, "step": 7629 }, { "epoch": 0.38, "learning_rate": 1.4205570626267641e-05, "loss": 0.5508, "step": 7630 }, { "epoch": 0.38, "learning_rate": 1.420410333124885e-05, "loss": 0.5065, "step": 7631 }, { "epoch": 0.38, "learning_rate": 1.4202635926277241e-05, "loss": 0.2462, "step": 7632 }, { "epoch": 0.38, "learning_rate": 1.4201168411391197e-05, "loss": 0.3323, "step": 7633 }, { "epoch": 0.38, "learning_rate": 1.4199700786629095e-05, "loss": 0.2854, "step": 7634 }, { "epoch": 0.38, "learning_rate": 1.419823305202932e-05, "loss": 0.5331, "step": 7635 }, { "epoch": 0.38, "learning_rate": 1.419676520763026e-05, "loss": 0.5188, "step": 7636 }, { "epoch": 0.38, "learning_rate": 1.4195297253470298e-05, "loss": 0.3127, "step": 7637 }, { "epoch": 0.38, "learning_rate": 1.4193829189587833e-05, "loss": 0.5094, "step": 7638 }, { "epoch": 0.38, "learning_rate": 1.419236101602126e-05, "loss": 0.4711, "step": 7639 }, { "epoch": 0.38, "learning_rate": 1.4190892732808972e-05, "loss": 0.5318, "step": 7640 }, { "epoch": 0.38, "learning_rate": 1.4189424339989378e-05, "loss": 0.3375, "step": 7641 }, { "epoch": 0.38, "learning_rate": 1.4187955837600873e-05, "loss": 0.0802, "step": 7642 }, { "epoch": 0.38, "learning_rate": 1.4186487225681873e-05, "loss": 0.4241, "step": 7643 }, { "epoch": 0.38, "learning_rate": 1.418501850427078e-05, "loss": 0.2583, "step": 7644 }, { "epoch": 0.38, "learning_rate": 1.4183549673406009e-05, "loss": 0.3074, "step": 7645 }, { "epoch": 0.38, "learning_rate": 1.418208073312598e-05, "loss": 0.3842, "step": 7646 }, { "epoch": 0.38, "learning_rate": 1.4180611683469102e-05, "loss": 0.4173, "step": 7647 }, { "epoch": 0.38, "learning_rate": 1.4179142524473807e-05, "loss": 0.606, "step": 7648 }, { "epoch": 0.38, "learning_rate": 1.4177673256178506e-05, "loss": 0.1546, "step": 7649 }, { "epoch": 0.38, "learning_rate": 1.4176203878621636e-05, "loss": 0.2872, "step": 7650 }, { "epoch": 0.38, "learning_rate": 1.4174734391841622e-05, "loss": 0.4984, "step": 7651 }, { "epoch": 0.38, "learning_rate": 1.4173264795876902e-05, "loss": 0.2696, "step": 7652 }, { "epoch": 0.38, "learning_rate": 1.4171795090765903e-05, "loss": 0.5572, "step": 7653 }, { "epoch": 0.38, "learning_rate": 1.417032527654707e-05, "loss": 0.4095, "step": 7654 }, { "epoch": 0.38, "learning_rate": 1.4168855353258839e-05, "loss": 0.4121, "step": 7655 }, { "epoch": 0.38, "learning_rate": 1.4167385320939657e-05, "loss": 0.4266, "step": 7656 }, { "epoch": 0.38, "learning_rate": 1.4165915179627972e-05, "loss": 0.3682, "step": 7657 }, { "epoch": 0.38, "learning_rate": 1.416444492936223e-05, "loss": 0.3799, "step": 7658 }, { "epoch": 0.38, "learning_rate": 1.4162974570180888e-05, "loss": 0.2251, "step": 7659 }, { "epoch": 0.38, "learning_rate": 1.4161504102122396e-05, "loss": 0.5816, "step": 7660 }, { "epoch": 0.38, "learning_rate": 1.4160033525225217e-05, "loss": 0.3867, "step": 7661 }, { "epoch": 0.38, "learning_rate": 1.4158562839527808e-05, "loss": 0.326, "step": 7662 }, { "epoch": 0.38, "learning_rate": 1.4157092045068634e-05, "loss": 0.4995, "step": 7663 }, { "epoch": 0.38, "learning_rate": 1.4155621141886165e-05, "loss": 0.1764, "step": 7664 }, { "epoch": 0.38, "learning_rate": 1.4154150130018867e-05, "loss": 0.3416, "step": 7665 }, { "epoch": 0.38, "learning_rate": 1.4152679009505212e-05, "loss": 0.4269, "step": 7666 }, { "epoch": 0.38, "learning_rate": 1.4151207780383675e-05, "loss": 0.3473, "step": 7667 }, { "epoch": 0.38, "learning_rate": 1.4149736442692738e-05, "loss": 0.2051, "step": 7668 }, { "epoch": 0.38, "learning_rate": 1.414826499647088e-05, "loss": 0.1994, "step": 7669 }, { "epoch": 0.38, "learning_rate": 1.4146793441756583e-05, "loss": 0.5037, "step": 7670 }, { "epoch": 0.38, "learning_rate": 1.4145321778588337e-05, "loss": 0.4329, "step": 7671 }, { "epoch": 0.38, "learning_rate": 1.4143850007004629e-05, "loss": 0.4872, "step": 7672 }, { "epoch": 0.38, "learning_rate": 1.414237812704395e-05, "loss": 0.5262, "step": 7673 }, { "epoch": 0.38, "learning_rate": 1.4140906138744797e-05, "loss": 0.36, "step": 7674 }, { "epoch": 0.38, "learning_rate": 1.4139434042145668e-05, "loss": 0.2942, "step": 7675 }, { "epoch": 0.38, "learning_rate": 1.413796183728506e-05, "loss": 0.1427, "step": 7676 }, { "epoch": 0.38, "learning_rate": 1.4136489524201484e-05, "loss": 0.4943, "step": 7677 }, { "epoch": 0.38, "learning_rate": 1.4135017102933442e-05, "loss": 0.3608, "step": 7678 }, { "epoch": 0.38, "learning_rate": 1.4133544573519442e-05, "loss": 0.2294, "step": 7679 }, { "epoch": 0.38, "learning_rate": 1.4132071935997999e-05, "loss": 0.3909, "step": 7680 }, { "epoch": 0.38, "learning_rate": 1.4130599190407625e-05, "loss": 0.3393, "step": 7681 }, { "epoch": 0.38, "learning_rate": 1.412912633678684e-05, "loss": 0.1207, "step": 7682 }, { "epoch": 0.38, "learning_rate": 1.4127653375174164e-05, "loss": 0.1728, "step": 7683 }, { "epoch": 0.38, "learning_rate": 1.412618030560812e-05, "loss": 0.234, "step": 7684 }, { "epoch": 0.38, "learning_rate": 1.4124707128127235e-05, "loss": 0.3139, "step": 7685 }, { "epoch": 0.38, "learning_rate": 1.412323384277004e-05, "loss": 0.4475, "step": 7686 }, { "epoch": 0.38, "learning_rate": 1.4121760449575061e-05, "loss": 0.3555, "step": 7687 }, { "epoch": 0.38, "learning_rate": 1.4120286948580835e-05, "loss": 0.5396, "step": 7688 }, { "epoch": 0.38, "learning_rate": 1.4118813339825903e-05, "loss": 0.3272, "step": 7689 }, { "epoch": 0.38, "learning_rate": 1.41173396233488e-05, "loss": 0.5509, "step": 7690 }, { "epoch": 0.38, "learning_rate": 1.4115865799188073e-05, "loss": 0.2628, "step": 7691 }, { "epoch": 0.38, "learning_rate": 1.4114391867382267e-05, "loss": 0.4907, "step": 7692 }, { "epoch": 0.38, "learning_rate": 1.4112917827969932e-05, "loss": 0.3045, "step": 7693 }, { "epoch": 0.38, "learning_rate": 1.4111443680989617e-05, "loss": 0.3573, "step": 7694 }, { "epoch": 0.38, "learning_rate": 1.4109969426479877e-05, "loss": 0.3245, "step": 7695 }, { "epoch": 0.38, "learning_rate": 1.410849506447927e-05, "loss": 0.425, "step": 7696 }, { "epoch": 0.38, "learning_rate": 1.4107020595026358e-05, "loss": 0.3498, "step": 7697 }, { "epoch": 0.38, "learning_rate": 1.4105546018159698e-05, "loss": 0.4823, "step": 7698 }, { "epoch": 0.38, "learning_rate": 1.4104071333917862e-05, "loss": 0.328, "step": 7699 }, { "epoch": 0.38, "learning_rate": 1.4102596542339414e-05, "loss": 0.3756, "step": 7700 }, { "epoch": 0.38, "learning_rate": 1.4101121643462927e-05, "loss": 0.5174, "step": 7701 }, { "epoch": 0.38, "learning_rate": 1.4099646637326975e-05, "loss": 0.6143, "step": 7702 }, { "epoch": 0.38, "learning_rate": 1.4098171523970137e-05, "loss": 0.432, "step": 7703 }, { "epoch": 0.38, "learning_rate": 1.4096696303430987e-05, "loss": 0.3327, "step": 7704 }, { "epoch": 0.38, "learning_rate": 1.4095220975748114e-05, "loss": 0.2622, "step": 7705 }, { "epoch": 0.38, "learning_rate": 1.4093745540960097e-05, "loss": 0.5095, "step": 7706 }, { "epoch": 0.38, "learning_rate": 1.409226999910553e-05, "loss": 0.2414, "step": 7707 }, { "epoch": 0.38, "learning_rate": 1.4090794350223e-05, "loss": 0.2173, "step": 7708 }, { "epoch": 0.38, "learning_rate": 1.40893185943511e-05, "loss": 0.378, "step": 7709 }, { "epoch": 0.38, "learning_rate": 1.4087842731528432e-05, "loss": 0.3204, "step": 7710 }, { "epoch": 0.39, "learning_rate": 1.4086366761793589e-05, "loss": 0.5738, "step": 7711 }, { "epoch": 0.39, "learning_rate": 1.4084890685185177e-05, "loss": 0.6154, "step": 7712 }, { "epoch": 0.39, "learning_rate": 1.4083414501741797e-05, "loss": 0.2796, "step": 7713 }, { "epoch": 0.39, "learning_rate": 1.408193821150206e-05, "loss": 0.3074, "step": 7714 }, { "epoch": 0.39, "learning_rate": 1.4080461814504577e-05, "loss": 0.4544, "step": 7715 }, { "epoch": 0.39, "learning_rate": 1.4078985310787956e-05, "loss": 0.4075, "step": 7716 }, { "epoch": 0.39, "learning_rate": 1.4077508700390818e-05, "loss": 0.4061, "step": 7717 }, { "epoch": 0.39, "learning_rate": 1.4076031983351784e-05, "loss": 0.6384, "step": 7718 }, { "epoch": 0.39, "learning_rate": 1.4074555159709471e-05, "loss": 0.4372, "step": 7719 }, { "epoch": 0.39, "learning_rate": 1.4073078229502502e-05, "loss": 0.4271, "step": 7720 }, { "epoch": 0.39, "learning_rate": 1.4071601192769508e-05, "loss": 0.2481, "step": 7721 }, { "epoch": 0.39, "learning_rate": 1.4070124049549118e-05, "loss": 0.238, "step": 7722 }, { "epoch": 0.39, "learning_rate": 1.4068646799879963e-05, "loss": 0.3272, "step": 7723 }, { "epoch": 0.39, "learning_rate": 1.4067169443800682e-05, "loss": 0.3826, "step": 7724 }, { "epoch": 0.39, "learning_rate": 1.406569198134991e-05, "loss": 0.5407, "step": 7725 }, { "epoch": 0.39, "learning_rate": 1.4064214412566289e-05, "loss": 0.5022, "step": 7726 }, { "epoch": 0.39, "learning_rate": 1.4062736737488463e-05, "loss": 0.2552, "step": 7727 }, { "epoch": 0.39, "learning_rate": 1.406125895615508e-05, "loss": 0.3287, "step": 7728 }, { "epoch": 0.39, "learning_rate": 1.4059781068604788e-05, "loss": 0.4652, "step": 7729 }, { "epoch": 0.39, "learning_rate": 1.4058303074876239e-05, "loss": 0.2633, "step": 7730 }, { "epoch": 0.39, "learning_rate": 1.4056824975008087e-05, "loss": 0.4568, "step": 7731 }, { "epoch": 0.39, "learning_rate": 1.4055346769038994e-05, "loss": 0.4377, "step": 7732 }, { "epoch": 0.39, "learning_rate": 1.4053868457007618e-05, "loss": 0.3125, "step": 7733 }, { "epoch": 0.39, "learning_rate": 1.4052390038952622e-05, "loss": 0.326, "step": 7734 }, { "epoch": 0.39, "learning_rate": 1.405091151491267e-05, "loss": 0.3563, "step": 7735 }, { "epoch": 0.39, "learning_rate": 1.4049432884926435e-05, "loss": 0.4501, "step": 7736 }, { "epoch": 0.39, "learning_rate": 1.4047954149032585e-05, "loss": 0.4484, "step": 7737 }, { "epoch": 0.39, "learning_rate": 1.4046475307269796e-05, "loss": 0.4166, "step": 7738 }, { "epoch": 0.39, "learning_rate": 1.4044996359676748e-05, "loss": 0.1325, "step": 7739 }, { "epoch": 0.39, "learning_rate": 1.4043517306292116e-05, "loss": 0.1644, "step": 7740 }, { "epoch": 0.39, "learning_rate": 1.4042038147154586e-05, "loss": 0.4474, "step": 7741 }, { "epoch": 0.39, "learning_rate": 1.404055888230284e-05, "loss": 0.3593, "step": 7742 }, { "epoch": 0.39, "learning_rate": 1.403907951177557e-05, "loss": 0.275, "step": 7743 }, { "epoch": 0.39, "learning_rate": 1.4037600035611467e-05, "loss": 0.2513, "step": 7744 }, { "epoch": 0.39, "learning_rate": 1.4036120453849222e-05, "loss": 0.4136, "step": 7745 }, { "epoch": 0.39, "learning_rate": 1.4034640766527535e-05, "loss": 0.4283, "step": 7746 }, { "epoch": 0.39, "learning_rate": 1.40331609736851e-05, "loss": 0.3553, "step": 7747 }, { "epoch": 0.39, "learning_rate": 1.4031681075360626e-05, "loss": 0.27, "step": 7748 }, { "epoch": 0.39, "learning_rate": 1.4030201071592811e-05, "loss": 0.3493, "step": 7749 }, { "epoch": 0.39, "learning_rate": 1.4028720962420366e-05, "loss": 0.2577, "step": 7750 }, { "epoch": 0.39, "learning_rate": 1.4027240747882003e-05, "loss": 0.2673, "step": 7751 }, { "epoch": 0.39, "learning_rate": 1.4025760428016429e-05, "loss": 0.3338, "step": 7752 }, { "epoch": 0.39, "learning_rate": 1.4024280002862367e-05, "loss": 0.534, "step": 7753 }, { "epoch": 0.39, "learning_rate": 1.402279947245853e-05, "loss": 0.564, "step": 7754 }, { "epoch": 0.39, "learning_rate": 1.4021318836843642e-05, "loss": 0.429, "step": 7755 }, { "epoch": 0.39, "learning_rate": 1.4019838096056428e-05, "loss": 0.3111, "step": 7756 }, { "epoch": 0.39, "learning_rate": 1.4018357250135614e-05, "loss": 0.311, "step": 7757 }, { "epoch": 0.39, "learning_rate": 1.401687629911993e-05, "loss": 0.2241, "step": 7758 }, { "epoch": 0.39, "learning_rate": 1.4015395243048106e-05, "loss": 0.5297, "step": 7759 }, { "epoch": 0.39, "learning_rate": 1.4013914081958882e-05, "loss": 0.3046, "step": 7760 }, { "epoch": 0.39, "learning_rate": 1.4012432815890989e-05, "loss": 0.2889, "step": 7761 }, { "epoch": 0.39, "learning_rate": 1.4010951444883169e-05, "loss": 0.4366, "step": 7762 }, { "epoch": 0.39, "learning_rate": 1.4009469968974169e-05, "loss": 0.3031, "step": 7763 }, { "epoch": 0.39, "learning_rate": 1.4007988388202736e-05, "loss": 0.2712, "step": 7764 }, { "epoch": 0.39, "learning_rate": 1.4006506702607614e-05, "loss": 0.2908, "step": 7765 }, { "epoch": 0.39, "learning_rate": 1.4005024912227558e-05, "loss": 0.4119, "step": 7766 }, { "epoch": 0.39, "learning_rate": 1.4003543017101319e-05, "loss": 0.4221, "step": 7767 }, { "epoch": 0.39, "learning_rate": 1.4002061017267657e-05, "loss": 0.2741, "step": 7768 }, { "epoch": 0.39, "learning_rate": 1.4000578912765332e-05, "loss": 0.3555, "step": 7769 }, { "epoch": 0.39, "learning_rate": 1.3999096703633105e-05, "loss": 0.4225, "step": 7770 }, { "epoch": 0.39, "learning_rate": 1.3997614389909741e-05, "loss": 0.3039, "step": 7771 }, { "epoch": 0.39, "learning_rate": 1.3996131971634009e-05, "loss": 0.181, "step": 7772 }, { "epoch": 0.39, "learning_rate": 1.3994649448844679e-05, "loss": 0.5613, "step": 7773 }, { "epoch": 0.39, "learning_rate": 1.3993166821580524e-05, "loss": 0.7891, "step": 7774 }, { "epoch": 0.39, "learning_rate": 1.3991684089880324e-05, "loss": 0.4099, "step": 7775 }, { "epoch": 0.39, "learning_rate": 1.3990201253782851e-05, "loss": 0.4724, "step": 7776 }, { "epoch": 0.39, "learning_rate": 1.3988718313326892e-05, "loss": 0.3522, "step": 7777 }, { "epoch": 0.39, "learning_rate": 1.3987235268551231e-05, "loss": 0.6338, "step": 7778 }, { "epoch": 0.39, "learning_rate": 1.398575211949465e-05, "loss": 0.3843, "step": 7779 }, { "epoch": 0.39, "learning_rate": 1.3984268866195946e-05, "loss": 0.1212, "step": 7780 }, { "epoch": 0.39, "learning_rate": 1.3982785508693905e-05, "loss": 0.3707, "step": 7781 }, { "epoch": 0.39, "learning_rate": 1.398130204702733e-05, "loss": 0.4056, "step": 7782 }, { "epoch": 0.39, "learning_rate": 1.3979818481235015e-05, "loss": 0.4509, "step": 7783 }, { "epoch": 0.39, "learning_rate": 1.3978334811355758e-05, "loss": 0.2899, "step": 7784 }, { "epoch": 0.39, "learning_rate": 1.3976851037428366e-05, "loss": 0.2818, "step": 7785 }, { "epoch": 0.39, "learning_rate": 1.397536715949164e-05, "loss": 0.4139, "step": 7786 }, { "epoch": 0.39, "learning_rate": 1.3973883177584395e-05, "loss": 0.3926, "step": 7787 }, { "epoch": 0.39, "learning_rate": 1.3972399091745441e-05, "loss": 0.6678, "step": 7788 }, { "epoch": 0.39, "learning_rate": 1.397091490201359e-05, "loss": 0.4904, "step": 7789 }, { "epoch": 0.39, "learning_rate": 1.3969430608427662e-05, "loss": 0.453, "step": 7790 }, { "epoch": 0.39, "learning_rate": 1.3967946211026475e-05, "loss": 0.3484, "step": 7791 }, { "epoch": 0.39, "learning_rate": 1.3966461709848852e-05, "loss": 0.3396, "step": 7792 }, { "epoch": 0.39, "learning_rate": 1.396497710493362e-05, "loss": 0.3771, "step": 7793 }, { "epoch": 0.39, "learning_rate": 1.3963492396319602e-05, "loss": 0.4783, "step": 7794 }, { "epoch": 0.39, "learning_rate": 1.3962007584045633e-05, "loss": 0.3627, "step": 7795 }, { "epoch": 0.39, "learning_rate": 1.3960522668150547e-05, "loss": 0.3964, "step": 7796 }, { "epoch": 0.39, "learning_rate": 1.3959037648673175e-05, "loss": 0.5244, "step": 7797 }, { "epoch": 0.39, "learning_rate": 1.395755252565236e-05, "loss": 0.5134, "step": 7798 }, { "epoch": 0.39, "learning_rate": 1.395606729912694e-05, "loss": 0.4208, "step": 7799 }, { "epoch": 0.39, "learning_rate": 1.3954581969135764e-05, "loss": 0.4901, "step": 7800 }, { "epoch": 0.39, "learning_rate": 1.3953096535717676e-05, "loss": 0.2791, "step": 7801 }, { "epoch": 0.39, "learning_rate": 1.3951610998911524e-05, "loss": 0.3639, "step": 7802 }, { "epoch": 0.39, "learning_rate": 1.3950125358756164e-05, "loss": 0.3691, "step": 7803 }, { "epoch": 0.39, "learning_rate": 1.3948639615290448e-05, "loss": 0.2676, "step": 7804 }, { "epoch": 0.39, "learning_rate": 1.3947153768553235e-05, "loss": 0.4455, "step": 7805 }, { "epoch": 0.39, "learning_rate": 1.3945667818583383e-05, "loss": 0.4398, "step": 7806 }, { "epoch": 0.39, "learning_rate": 1.394418176541976e-05, "loss": 0.2895, "step": 7807 }, { "epoch": 0.39, "learning_rate": 1.3942695609101229e-05, "loss": 0.3737, "step": 7808 }, { "epoch": 0.39, "learning_rate": 1.3941209349666655e-05, "loss": 0.4566, "step": 7809 }, { "epoch": 0.39, "learning_rate": 1.3939722987154916e-05, "loss": 0.3463, "step": 7810 }, { "epoch": 0.39, "learning_rate": 1.393823652160488e-05, "loss": 0.3682, "step": 7811 }, { "epoch": 0.39, "learning_rate": 1.3936749953055427e-05, "loss": 0.3779, "step": 7812 }, { "epoch": 0.39, "learning_rate": 1.3935263281545433e-05, "loss": 0.1448, "step": 7813 }, { "epoch": 0.39, "learning_rate": 1.3933776507113785e-05, "loss": 0.1753, "step": 7814 }, { "epoch": 0.39, "learning_rate": 1.3932289629799363e-05, "loss": 0.4231, "step": 7815 }, { "epoch": 0.39, "learning_rate": 1.3930802649641058e-05, "loss": 0.2913, "step": 7816 }, { "epoch": 0.39, "learning_rate": 1.3929315566677757e-05, "loss": 0.5372, "step": 7817 }, { "epoch": 0.39, "learning_rate": 1.3927828380948354e-05, "loss": 0.487, "step": 7818 }, { "epoch": 0.39, "learning_rate": 1.3926341092491744e-05, "loss": 0.2637, "step": 7819 }, { "epoch": 0.39, "learning_rate": 1.3924853701346825e-05, "loss": 0.2127, "step": 7820 }, { "epoch": 0.39, "learning_rate": 1.39233662075525e-05, "loss": 0.5774, "step": 7821 }, { "epoch": 0.39, "learning_rate": 1.3921878611147668e-05, "loss": 0.1759, "step": 7822 }, { "epoch": 0.39, "learning_rate": 1.392039091217124e-05, "loss": 0.3599, "step": 7823 }, { "epoch": 0.39, "learning_rate": 1.391890311066212e-05, "loss": 0.4487, "step": 7824 }, { "epoch": 0.39, "learning_rate": 1.3917415206659223e-05, "loss": 0.1223, "step": 7825 }, { "epoch": 0.39, "learning_rate": 1.3915927200201464e-05, "loss": 0.4314, "step": 7826 }, { "epoch": 0.39, "learning_rate": 1.3914439091327757e-05, "loss": 0.2571, "step": 7827 }, { "epoch": 0.39, "learning_rate": 1.3912950880077021e-05, "loss": 0.4915, "step": 7828 }, { "epoch": 0.39, "learning_rate": 1.3911462566488178e-05, "loss": 0.32, "step": 7829 }, { "epoch": 0.39, "learning_rate": 1.3909974150600159e-05, "loss": 0.2556, "step": 7830 }, { "epoch": 0.39, "learning_rate": 1.3908485632451886e-05, "loss": 0.267, "step": 7831 }, { "epoch": 0.39, "learning_rate": 1.390699701208229e-05, "loss": 0.5065, "step": 7832 }, { "epoch": 0.39, "learning_rate": 1.3905508289530308e-05, "loss": 0.4146, "step": 7833 }, { "epoch": 0.39, "learning_rate": 1.3904019464834866e-05, "loss": 0.5662, "step": 7834 }, { "epoch": 0.39, "learning_rate": 1.3902530538034911e-05, "loss": 0.5283, "step": 7835 }, { "epoch": 0.39, "learning_rate": 1.3901041509169379e-05, "loss": 0.4423, "step": 7836 }, { "epoch": 0.39, "learning_rate": 1.3899552378277218e-05, "loss": 0.4262, "step": 7837 }, { "epoch": 0.39, "learning_rate": 1.3898063145397372e-05, "loss": 0.2715, "step": 7838 }, { "epoch": 0.39, "learning_rate": 1.389657381056879e-05, "loss": 0.4119, "step": 7839 }, { "epoch": 0.39, "learning_rate": 1.3895084373830425e-05, "loss": 0.5315, "step": 7840 }, { "epoch": 0.39, "learning_rate": 1.3893594835221227e-05, "loss": 0.5118, "step": 7841 }, { "epoch": 0.39, "learning_rate": 1.3892105194780157e-05, "loss": 0.4333, "step": 7842 }, { "epoch": 0.39, "learning_rate": 1.3890615452546175e-05, "loss": 0.4587, "step": 7843 }, { "epoch": 0.39, "learning_rate": 1.388912560855824e-05, "loss": 0.3979, "step": 7844 }, { "epoch": 0.39, "learning_rate": 1.3887635662855316e-05, "loss": 0.4561, "step": 7845 }, { "epoch": 0.39, "learning_rate": 1.3886145615476376e-05, "loss": 0.3603, "step": 7846 }, { "epoch": 0.39, "learning_rate": 1.3884655466460386e-05, "loss": 0.3958, "step": 7847 }, { "epoch": 0.39, "learning_rate": 1.3883165215846321e-05, "loss": 0.301, "step": 7848 }, { "epoch": 0.39, "learning_rate": 1.3881674863673159e-05, "loss": 0.5044, "step": 7849 }, { "epoch": 0.39, "learning_rate": 1.3880184409979868e-05, "loss": 0.2537, "step": 7850 }, { "epoch": 0.39, "learning_rate": 1.3878693854805443e-05, "loss": 0.4797, "step": 7851 }, { "epoch": 0.39, "learning_rate": 1.3877203198188855e-05, "loss": 0.4006, "step": 7852 }, { "epoch": 0.39, "learning_rate": 1.38757124401691e-05, "loss": 0.3718, "step": 7853 }, { "epoch": 0.39, "learning_rate": 1.3874221580785158e-05, "loss": 0.3581, "step": 7854 }, { "epoch": 0.39, "learning_rate": 1.387273062007603e-05, "loss": 0.3595, "step": 7855 }, { "epoch": 0.39, "learning_rate": 1.38712395580807e-05, "loss": 0.2636, "step": 7856 }, { "epoch": 0.39, "learning_rate": 1.3869748394838174e-05, "loss": 0.4331, "step": 7857 }, { "epoch": 0.39, "learning_rate": 1.3868257130387446e-05, "loss": 0.3653, "step": 7858 }, { "epoch": 0.39, "learning_rate": 1.3866765764767518e-05, "loss": 0.3997, "step": 7859 }, { "epoch": 0.39, "learning_rate": 1.3865274298017397e-05, "loss": 0.6369, "step": 7860 }, { "epoch": 0.39, "learning_rate": 1.3863782730176087e-05, "loss": 0.2175, "step": 7861 }, { "epoch": 0.39, "learning_rate": 1.3862291061282601e-05, "loss": 0.4166, "step": 7862 }, { "epoch": 0.39, "learning_rate": 1.3860799291375953e-05, "loss": 0.3659, "step": 7863 }, { "epoch": 0.39, "learning_rate": 1.3859307420495153e-05, "loss": 0.2636, "step": 7864 }, { "epoch": 0.39, "learning_rate": 1.3857815448679228e-05, "loss": 0.2406, "step": 7865 }, { "epoch": 0.39, "learning_rate": 1.3856323375967188e-05, "loss": 0.3785, "step": 7866 }, { "epoch": 0.39, "learning_rate": 1.3854831202398062e-05, "loss": 0.4635, "step": 7867 }, { "epoch": 0.39, "learning_rate": 1.3853338928010875e-05, "loss": 0.6373, "step": 7868 }, { "epoch": 0.39, "learning_rate": 1.3851846552844657e-05, "loss": 0.6034, "step": 7869 }, { "epoch": 0.39, "learning_rate": 1.3850354076938435e-05, "loss": 0.2862, "step": 7870 }, { "epoch": 0.39, "learning_rate": 1.3848861500331246e-05, "loss": 0.4764, "step": 7871 }, { "epoch": 0.39, "learning_rate": 1.3847368823062126e-05, "loss": 0.5399, "step": 7872 }, { "epoch": 0.39, "learning_rate": 1.3845876045170112e-05, "loss": 0.4426, "step": 7873 }, { "epoch": 0.39, "learning_rate": 1.3844383166694249e-05, "loss": 0.5512, "step": 7874 }, { "epoch": 0.39, "learning_rate": 1.3842890187673579e-05, "loss": 0.5002, "step": 7875 }, { "epoch": 0.39, "learning_rate": 1.384139710814715e-05, "loss": 0.2449, "step": 7876 }, { "epoch": 0.39, "learning_rate": 1.3839903928154008e-05, "loss": 0.2538, "step": 7877 }, { "epoch": 0.39, "learning_rate": 1.3838410647733212e-05, "loss": 0.4023, "step": 7878 }, { "epoch": 0.39, "learning_rate": 1.3836917266923813e-05, "loss": 0.3339, "step": 7879 }, { "epoch": 0.39, "learning_rate": 1.3835423785764865e-05, "loss": 0.3977, "step": 7880 }, { "epoch": 0.39, "learning_rate": 1.3833930204295435e-05, "loss": 0.2729, "step": 7881 }, { "epoch": 0.39, "learning_rate": 1.383243652255458e-05, "loss": 0.4459, "step": 7882 }, { "epoch": 0.39, "learning_rate": 1.3830942740581365e-05, "loss": 0.2813, "step": 7883 }, { "epoch": 0.39, "learning_rate": 1.3829448858414862e-05, "loss": 0.2441, "step": 7884 }, { "epoch": 0.39, "learning_rate": 1.3827954876094142e-05, "loss": 0.3399, "step": 7885 }, { "epoch": 0.39, "learning_rate": 1.3826460793658273e-05, "loss": 0.5264, "step": 7886 }, { "epoch": 0.39, "learning_rate": 1.3824966611146336e-05, "loss": 0.2793, "step": 7887 }, { "epoch": 0.39, "learning_rate": 1.3823472328597407e-05, "loss": 0.4752, "step": 7888 }, { "epoch": 0.39, "learning_rate": 1.3821977946050566e-05, "loss": 0.3019, "step": 7889 }, { "epoch": 0.39, "learning_rate": 1.3820483463544899e-05, "loss": 0.2379, "step": 7890 }, { "epoch": 0.39, "learning_rate": 1.3818988881119488e-05, "loss": 0.3182, "step": 7891 }, { "epoch": 0.39, "learning_rate": 1.3817494198813428e-05, "loss": 0.3394, "step": 7892 }, { "epoch": 0.39, "learning_rate": 1.3815999416665807e-05, "loss": 0.4537, "step": 7893 }, { "epoch": 0.39, "learning_rate": 1.3814504534715718e-05, "loss": 0.3047, "step": 7894 }, { "epoch": 0.39, "learning_rate": 1.3813009553002257e-05, "loss": 0.379, "step": 7895 }, { "epoch": 0.39, "learning_rate": 1.3811514471564529e-05, "loss": 0.3068, "step": 7896 }, { "epoch": 0.39, "learning_rate": 1.3810019290441631e-05, "loss": 0.2566, "step": 7897 }, { "epoch": 0.39, "learning_rate": 1.3808524009672668e-05, "loss": 0.2788, "step": 7898 }, { "epoch": 0.39, "learning_rate": 1.3807028629296748e-05, "loss": 0.2474, "step": 7899 }, { "epoch": 0.39, "learning_rate": 1.3805533149352982e-05, "loss": 0.3281, "step": 7900 }, { "epoch": 0.39, "learning_rate": 1.380403756988048e-05, "loss": 0.2589, "step": 7901 }, { "epoch": 0.39, "learning_rate": 1.3802541890918355e-05, "loss": 0.3355, "step": 7902 }, { "epoch": 0.39, "learning_rate": 1.3801046112505731e-05, "loss": 0.3735, "step": 7903 }, { "epoch": 0.39, "learning_rate": 1.3799550234681722e-05, "loss": 0.4085, "step": 7904 }, { "epoch": 0.39, "learning_rate": 1.3798054257485456e-05, "loss": 0.2692, "step": 7905 }, { "epoch": 0.39, "learning_rate": 1.3796558180956054e-05, "loss": 0.2225, "step": 7906 }, { "epoch": 0.39, "learning_rate": 1.3795062005132644e-05, "loss": 0.3249, "step": 7907 }, { "epoch": 0.39, "learning_rate": 1.3793565730054359e-05, "loss": 0.3094, "step": 7908 }, { "epoch": 0.39, "learning_rate": 1.3792069355760331e-05, "loss": 0.4305, "step": 7909 }, { "epoch": 0.39, "learning_rate": 1.3790572882289697e-05, "loss": 0.3707, "step": 7910 }, { "epoch": 0.4, "learning_rate": 1.378907630968159e-05, "loss": 0.7067, "step": 7911 }, { "epoch": 0.4, "learning_rate": 1.3787579637975159e-05, "loss": 0.3109, "step": 7912 }, { "epoch": 0.4, "learning_rate": 1.3786082867209542e-05, "loss": 0.4221, "step": 7913 }, { "epoch": 0.4, "learning_rate": 1.3784585997423885e-05, "loss": 0.3399, "step": 7914 }, { "epoch": 0.4, "learning_rate": 1.3783089028657339e-05, "loss": 0.1998, "step": 7915 }, { "epoch": 0.4, "learning_rate": 1.3781591960949057e-05, "loss": 0.3782, "step": 7916 }, { "epoch": 0.4, "learning_rate": 1.378009479433819e-05, "loss": 0.7025, "step": 7917 }, { "epoch": 0.4, "learning_rate": 1.3778597528863891e-05, "loss": 0.1678, "step": 7918 }, { "epoch": 0.4, "learning_rate": 1.3777100164565328e-05, "loss": 0.3146, "step": 7919 }, { "epoch": 0.4, "learning_rate": 1.377560270148165e-05, "loss": 0.2869, "step": 7920 }, { "epoch": 0.4, "learning_rate": 1.3774105139652034e-05, "loss": 0.3621, "step": 7921 }, { "epoch": 0.4, "learning_rate": 1.377260747911564e-05, "loss": 0.3036, "step": 7922 }, { "epoch": 0.4, "learning_rate": 1.3771109719911637e-05, "loss": 0.389, "step": 7923 }, { "epoch": 0.4, "learning_rate": 1.37696118620792e-05, "loss": 0.3641, "step": 7924 }, { "epoch": 0.4, "learning_rate": 1.37681139056575e-05, "loss": 0.4484, "step": 7925 }, { "epoch": 0.4, "learning_rate": 1.376661585068572e-05, "loss": 0.2216, "step": 7926 }, { "epoch": 0.4, "learning_rate": 1.3765117697203032e-05, "loss": 0.5404, "step": 7927 }, { "epoch": 0.4, "learning_rate": 1.3763619445248625e-05, "loss": 0.3819, "step": 7928 }, { "epoch": 0.4, "learning_rate": 1.376212109486168e-05, "loss": 0.4336, "step": 7929 }, { "epoch": 0.4, "learning_rate": 1.3760622646081387e-05, "loss": 0.4524, "step": 7930 }, { "epoch": 0.4, "learning_rate": 1.3759124098946932e-05, "loss": 0.3079, "step": 7931 }, { "epoch": 0.4, "learning_rate": 1.375762545349751e-05, "loss": 0.5645, "step": 7932 }, { "epoch": 0.4, "learning_rate": 1.3756126709772315e-05, "loss": 0.2466, "step": 7933 }, { "epoch": 0.4, "learning_rate": 1.3754627867810546e-05, "loss": 0.3218, "step": 7934 }, { "epoch": 0.4, "learning_rate": 1.3753128927651404e-05, "loss": 0.2263, "step": 7935 }, { "epoch": 0.4, "learning_rate": 1.3751629889334088e-05, "loss": 0.3002, "step": 7936 }, { "epoch": 0.4, "learning_rate": 1.375013075289781e-05, "loss": 0.5332, "step": 7937 }, { "epoch": 0.4, "learning_rate": 1.3748631518381772e-05, "loss": 0.4727, "step": 7938 }, { "epoch": 0.4, "learning_rate": 1.3747132185825185e-05, "loss": 0.2936, "step": 7939 }, { "epoch": 0.4, "learning_rate": 1.3745632755267266e-05, "loss": 0.3415, "step": 7940 }, { "epoch": 0.4, "learning_rate": 1.3744133226747227e-05, "loss": 0.3763, "step": 7941 }, { "epoch": 0.4, "learning_rate": 1.3742633600304288e-05, "loss": 0.3357, "step": 7942 }, { "epoch": 0.4, "learning_rate": 1.3741133875977667e-05, "loss": 0.2783, "step": 7943 }, { "epoch": 0.4, "learning_rate": 1.3739634053806592e-05, "loss": 0.4339, "step": 7944 }, { "epoch": 0.4, "learning_rate": 1.3738134133830285e-05, "loss": 0.2761, "step": 7945 }, { "epoch": 0.4, "learning_rate": 1.3736634116087977e-05, "loss": 0.288, "step": 7946 }, { "epoch": 0.4, "learning_rate": 1.3735134000618898e-05, "loss": 0.4677, "step": 7947 }, { "epoch": 0.4, "learning_rate": 1.3733633787462278e-05, "loss": 0.285, "step": 7948 }, { "epoch": 0.4, "learning_rate": 1.373213347665736e-05, "loss": 0.3535, "step": 7949 }, { "epoch": 0.4, "learning_rate": 1.3730633068243377e-05, "loss": 0.2561, "step": 7950 }, { "epoch": 0.4, "learning_rate": 1.3729132562259572e-05, "loss": 0.3605, "step": 7951 }, { "epoch": 0.4, "learning_rate": 1.3727631958745194e-05, "loss": 0.2829, "step": 7952 }, { "epoch": 0.4, "learning_rate": 1.3726131257739482e-05, "loss": 0.374, "step": 7953 }, { "epoch": 0.4, "learning_rate": 1.3724630459281685e-05, "loss": 0.405, "step": 7954 }, { "epoch": 0.4, "learning_rate": 1.3723129563411058e-05, "loss": 0.2486, "step": 7955 }, { "epoch": 0.4, "learning_rate": 1.3721628570166851e-05, "loss": 0.2338, "step": 7956 }, { "epoch": 0.4, "learning_rate": 1.3720127479588325e-05, "loss": 0.4993, "step": 7957 }, { "epoch": 0.4, "learning_rate": 1.371862629171474e-05, "loss": 0.4513, "step": 7958 }, { "epoch": 0.4, "learning_rate": 1.3717125006585352e-05, "loss": 0.3657, "step": 7959 }, { "epoch": 0.4, "learning_rate": 1.3715623624239429e-05, "loss": 0.4748, "step": 7960 }, { "epoch": 0.4, "learning_rate": 1.3714122144716238e-05, "loss": 0.396, "step": 7961 }, { "epoch": 0.4, "learning_rate": 1.3712620568055044e-05, "loss": 0.3809, "step": 7962 }, { "epoch": 0.4, "learning_rate": 1.3711118894295123e-05, "loss": 0.2962, "step": 7963 }, { "epoch": 0.4, "learning_rate": 1.3709617123475747e-05, "loss": 0.3254, "step": 7964 }, { "epoch": 0.4, "learning_rate": 1.3708115255636196e-05, "loss": 0.3633, "step": 7965 }, { "epoch": 0.4, "learning_rate": 1.370661329081575e-05, "loss": 0.3668, "step": 7966 }, { "epoch": 0.4, "learning_rate": 1.3705111229053683e-05, "loss": 0.3783, "step": 7967 }, { "epoch": 0.4, "learning_rate": 1.3703609070389285e-05, "loss": 0.504, "step": 7968 }, { "epoch": 0.4, "learning_rate": 1.3702106814861847e-05, "loss": 0.388, "step": 7969 }, { "epoch": 0.4, "learning_rate": 1.370060446251065e-05, "loss": 0.1329, "step": 7970 }, { "epoch": 0.4, "learning_rate": 1.369910201337499e-05, "loss": 0.3738, "step": 7971 }, { "epoch": 0.4, "learning_rate": 1.3697599467494163e-05, "loss": 0.4249, "step": 7972 }, { "epoch": 0.4, "learning_rate": 1.3696096824907464e-05, "loss": 0.3248, "step": 7973 }, { "epoch": 0.4, "learning_rate": 1.3694594085654194e-05, "loss": 0.401, "step": 7974 }, { "epoch": 0.4, "learning_rate": 1.3693091249773656e-05, "loss": 0.2667, "step": 7975 }, { "epoch": 0.4, "learning_rate": 1.3691588317305154e-05, "loss": 0.2679, "step": 7976 }, { "epoch": 0.4, "learning_rate": 1.3690085288287992e-05, "loss": 0.3076, "step": 7977 }, { "epoch": 0.4, "learning_rate": 1.3688582162761486e-05, "loss": 0.3206, "step": 7978 }, { "epoch": 0.4, "learning_rate": 1.3687078940764942e-05, "loss": 0.3595, "step": 7979 }, { "epoch": 0.4, "learning_rate": 1.3685575622337676e-05, "loss": 0.253, "step": 7980 }, { "epoch": 0.4, "learning_rate": 1.368407220751901e-05, "loss": 0.3884, "step": 7981 }, { "epoch": 0.4, "learning_rate": 1.3682568696348257e-05, "loss": 0.2142, "step": 7982 }, { "epoch": 0.4, "learning_rate": 1.3681065088864747e-05, "loss": 0.4535, "step": 7983 }, { "epoch": 0.4, "learning_rate": 1.3679561385107799e-05, "loss": 0.412, "step": 7984 }, { "epoch": 0.4, "learning_rate": 1.3678057585116742e-05, "loss": 0.5298, "step": 7985 }, { "epoch": 0.4, "learning_rate": 1.3676553688930909e-05, "loss": 0.2495, "step": 7986 }, { "epoch": 0.4, "learning_rate": 1.3675049696589626e-05, "loss": 0.3002, "step": 7987 }, { "epoch": 0.4, "learning_rate": 1.3673545608132234e-05, "loss": 0.485, "step": 7988 }, { "epoch": 0.4, "learning_rate": 1.3672041423598068e-05, "loss": 0.2255, "step": 7989 }, { "epoch": 0.4, "learning_rate": 1.3670537143026468e-05, "loss": 0.3155, "step": 7990 }, { "epoch": 0.4, "learning_rate": 1.3669032766456776e-05, "loss": 0.3116, "step": 7991 }, { "epoch": 0.4, "learning_rate": 1.3667528293928338e-05, "loss": 0.3433, "step": 7992 }, { "epoch": 0.4, "learning_rate": 1.3666023725480501e-05, "loss": 0.5288, "step": 7993 }, { "epoch": 0.4, "learning_rate": 1.3664519061152612e-05, "loss": 0.2958, "step": 7994 }, { "epoch": 0.4, "learning_rate": 1.3663014300984031e-05, "loss": 0.4634, "step": 7995 }, { "epoch": 0.4, "learning_rate": 1.3661509445014108e-05, "loss": 0.6244, "step": 7996 }, { "epoch": 0.4, "learning_rate": 1.3660004493282199e-05, "loss": 0.3179, "step": 7997 }, { "epoch": 0.4, "learning_rate": 1.3658499445827666e-05, "loss": 0.3032, "step": 7998 }, { "epoch": 0.4, "learning_rate": 1.3656994302689875e-05, "loss": 0.3778, "step": 7999 }, { "epoch": 0.4, "learning_rate": 1.3655489063908185e-05, "loss": 0.1383, "step": 8000 }, { "epoch": 0.4, "learning_rate": 1.3653983729521966e-05, "loss": 0.3263, "step": 8001 }, { "epoch": 0.4, "learning_rate": 1.3652478299570593e-05, "loss": 0.333, "step": 8002 }, { "epoch": 0.4, "learning_rate": 1.3650972774093431e-05, "loss": 0.4222, "step": 8003 }, { "epoch": 0.4, "learning_rate": 1.3649467153129859e-05, "loss": 0.2513, "step": 8004 }, { "epoch": 0.4, "learning_rate": 1.3647961436719251e-05, "loss": 0.3273, "step": 8005 }, { "epoch": 0.4, "learning_rate": 1.3646455624900989e-05, "loss": 0.2518, "step": 8006 }, { "epoch": 0.4, "learning_rate": 1.3644949717714458e-05, "loss": 0.6466, "step": 8007 }, { "epoch": 0.4, "learning_rate": 1.3643443715199043e-05, "loss": 0.2013, "step": 8008 }, { "epoch": 0.4, "learning_rate": 1.3641937617394128e-05, "loss": 0.4477, "step": 8009 }, { "epoch": 0.4, "learning_rate": 1.3640431424339102e-05, "loss": 0.6227, "step": 8010 }, { "epoch": 0.4, "learning_rate": 1.3638925136073364e-05, "loss": 0.5258, "step": 8011 }, { "epoch": 0.4, "learning_rate": 1.3637418752636302e-05, "loss": 0.4284, "step": 8012 }, { "epoch": 0.4, "learning_rate": 1.3635912274067323e-05, "loss": 0.467, "step": 8013 }, { "epoch": 0.4, "learning_rate": 1.3634405700405815e-05, "loss": 0.3836, "step": 8014 }, { "epoch": 0.4, "learning_rate": 1.3632899031691189e-05, "loss": 0.3896, "step": 8015 }, { "epoch": 0.4, "learning_rate": 1.3631392267962845e-05, "loss": 0.3516, "step": 8016 }, { "epoch": 0.4, "learning_rate": 1.3629885409260193e-05, "loss": 0.2974, "step": 8017 }, { "epoch": 0.4, "learning_rate": 1.3628378455622642e-05, "loss": 0.4423, "step": 8018 }, { "epoch": 0.4, "learning_rate": 1.3626871407089603e-05, "loss": 0.3296, "step": 8019 }, { "epoch": 0.4, "learning_rate": 1.3625364263700498e-05, "loss": 0.348, "step": 8020 }, { "epoch": 0.4, "learning_rate": 1.3623857025494734e-05, "loss": 0.6526, "step": 8021 }, { "epoch": 0.4, "learning_rate": 1.3622349692511738e-05, "loss": 0.4351, "step": 8022 }, { "epoch": 0.4, "learning_rate": 1.362084226479093e-05, "loss": 0.5902, "step": 8023 }, { "epoch": 0.4, "learning_rate": 1.3619334742371735e-05, "loss": 0.4681, "step": 8024 }, { "epoch": 0.4, "learning_rate": 1.3617827125293582e-05, "loss": 0.2455, "step": 8025 }, { "epoch": 0.4, "learning_rate": 1.3616319413595897e-05, "loss": 0.2495, "step": 8026 }, { "epoch": 0.4, "learning_rate": 1.3614811607318114e-05, "loss": 0.3916, "step": 8027 }, { "epoch": 0.4, "learning_rate": 1.361330370649967e-05, "loss": 0.4174, "step": 8028 }, { "epoch": 0.4, "learning_rate": 1.361179571118e-05, "loss": 0.298, "step": 8029 }, { "epoch": 0.4, "learning_rate": 1.361028762139854e-05, "loss": 0.3494, "step": 8030 }, { "epoch": 0.4, "learning_rate": 1.360877943719474e-05, "loss": 0.1757, "step": 8031 }, { "epoch": 0.4, "learning_rate": 1.3607271158608037e-05, "loss": 0.3921, "step": 8032 }, { "epoch": 0.4, "learning_rate": 1.3605762785677886e-05, "loss": 0.4335, "step": 8033 }, { "epoch": 0.4, "learning_rate": 1.3604254318443731e-05, "loss": 0.4698, "step": 8034 }, { "epoch": 0.4, "learning_rate": 1.360274575694502e-05, "loss": 0.5181, "step": 8035 }, { "epoch": 0.4, "learning_rate": 1.3601237101221215e-05, "loss": 0.2228, "step": 8036 }, { "epoch": 0.4, "learning_rate": 1.3599728351311771e-05, "loss": 0.3525, "step": 8037 }, { "epoch": 0.4, "learning_rate": 1.359821950725615e-05, "loss": 0.4498, "step": 8038 }, { "epoch": 0.4, "learning_rate": 1.3596710569093808e-05, "loss": 0.2685, "step": 8039 }, { "epoch": 0.4, "learning_rate": 1.3595201536864213e-05, "loss": 0.3298, "step": 8040 }, { "epoch": 0.4, "learning_rate": 1.3593692410606829e-05, "loss": 0.4521, "step": 8041 }, { "epoch": 0.4, "learning_rate": 1.3592183190361126e-05, "loss": 0.4882, "step": 8042 }, { "epoch": 0.4, "learning_rate": 1.359067387616658e-05, "loss": 0.538, "step": 8043 }, { "epoch": 0.4, "learning_rate": 1.3589164468062656e-05, "loss": 0.4709, "step": 8044 }, { "epoch": 0.4, "learning_rate": 1.3587654966088842e-05, "loss": 0.2856, "step": 8045 }, { "epoch": 0.4, "learning_rate": 1.3586145370284607e-05, "loss": 0.3965, "step": 8046 }, { "epoch": 0.4, "learning_rate": 1.358463568068944e-05, "loss": 0.4005, "step": 8047 }, { "epoch": 0.4, "learning_rate": 1.3583125897342819e-05, "loss": 0.2793, "step": 8048 }, { "epoch": 0.4, "learning_rate": 1.3581616020284236e-05, "loss": 0.4464, "step": 8049 }, { "epoch": 0.4, "learning_rate": 1.3580106049553176e-05, "loss": 0.2917, "step": 8050 }, { "epoch": 0.4, "learning_rate": 1.3578595985189134e-05, "loss": 0.1283, "step": 8051 }, { "epoch": 0.4, "learning_rate": 1.35770858272316e-05, "loss": 0.3831, "step": 8052 }, { "epoch": 0.4, "learning_rate": 1.3575575575720069e-05, "loss": 0.1743, "step": 8053 }, { "epoch": 0.4, "learning_rate": 1.3574065230694043e-05, "loss": 0.3308, "step": 8054 }, { "epoch": 0.4, "learning_rate": 1.357255479219302e-05, "loss": 0.5085, "step": 8055 }, { "epoch": 0.4, "learning_rate": 1.3571044260256509e-05, "loss": 0.4168, "step": 8056 }, { "epoch": 0.4, "learning_rate": 1.356953363492401e-05, "loss": 0.3471, "step": 8057 }, { "epoch": 0.4, "learning_rate": 1.3568022916235036e-05, "loss": 0.5061, "step": 8058 }, { "epoch": 0.4, "learning_rate": 1.3566512104229096e-05, "loss": 0.301, "step": 8059 }, { "epoch": 0.4, "learning_rate": 1.35650011989457e-05, "loss": 0.2125, "step": 8060 }, { "epoch": 0.4, "learning_rate": 1.356349020042437e-05, "loss": 0.2666, "step": 8061 }, { "epoch": 0.4, "learning_rate": 1.3561979108704621e-05, "loss": 0.3764, "step": 8062 }, { "epoch": 0.4, "learning_rate": 1.3560467923825972e-05, "loss": 0.4818, "step": 8063 }, { "epoch": 0.4, "learning_rate": 1.355895664582795e-05, "loss": 0.256, "step": 8064 }, { "epoch": 0.4, "learning_rate": 1.3557445274750078e-05, "loss": 0.4756, "step": 8065 }, { "epoch": 0.4, "learning_rate": 1.3555933810631885e-05, "loss": 0.3644, "step": 8066 }, { "epoch": 0.4, "learning_rate": 1.3554422253512897e-05, "loss": 0.363, "step": 8067 }, { "epoch": 0.4, "learning_rate": 1.3552910603432654e-05, "loss": 0.4713, "step": 8068 }, { "epoch": 0.4, "learning_rate": 1.3551398860430688e-05, "loss": 0.3931, "step": 8069 }, { "epoch": 0.4, "learning_rate": 1.3549887024546534e-05, "loss": 0.2547, "step": 8070 }, { "epoch": 0.4, "learning_rate": 1.3548375095819732e-05, "loss": 0.3561, "step": 8071 }, { "epoch": 0.4, "learning_rate": 1.354686307428983e-05, "loss": 0.4032, "step": 8072 }, { "epoch": 0.4, "learning_rate": 1.354535095999637e-05, "loss": 0.4184, "step": 8073 }, { "epoch": 0.4, "learning_rate": 1.35438387529789e-05, "loss": 0.3348, "step": 8074 }, { "epoch": 0.4, "learning_rate": 1.3542326453276968e-05, "loss": 0.4589, "step": 8075 }, { "epoch": 0.4, "learning_rate": 1.3540814060930128e-05, "loss": 0.5324, "step": 8076 }, { "epoch": 0.4, "learning_rate": 1.3539301575977934e-05, "loss": 0.3124, "step": 8077 }, { "epoch": 0.4, "learning_rate": 1.353778899845994e-05, "loss": 0.4631, "step": 8078 }, { "epoch": 0.4, "learning_rate": 1.3536276328415711e-05, "loss": 0.5279, "step": 8079 }, { "epoch": 0.4, "learning_rate": 1.3534763565884806e-05, "loss": 0.2526, "step": 8080 }, { "epoch": 0.4, "learning_rate": 1.3533250710906788e-05, "loss": 0.3875, "step": 8081 }, { "epoch": 0.4, "learning_rate": 1.3531737763521229e-05, "loss": 0.3566, "step": 8082 }, { "epoch": 0.4, "learning_rate": 1.353022472376769e-05, "loss": 0.1634, "step": 8083 }, { "epoch": 0.4, "learning_rate": 1.3528711591685751e-05, "loss": 0.3527, "step": 8084 }, { "epoch": 0.4, "learning_rate": 1.352719836731498e-05, "loss": 0.2207, "step": 8085 }, { "epoch": 0.4, "learning_rate": 1.3525685050694955e-05, "loss": 0.4759, "step": 8086 }, { "epoch": 0.4, "learning_rate": 1.3524171641865258e-05, "loss": 0.3851, "step": 8087 }, { "epoch": 0.4, "learning_rate": 1.3522658140865467e-05, "loss": 0.318, "step": 8088 }, { "epoch": 0.4, "learning_rate": 1.3521144547735164e-05, "loss": 0.4049, "step": 8089 }, { "epoch": 0.4, "learning_rate": 1.3519630862513937e-05, "loss": 0.3515, "step": 8090 }, { "epoch": 0.4, "learning_rate": 1.3518117085241378e-05, "loss": 0.4944, "step": 8091 }, { "epoch": 0.4, "learning_rate": 1.351660321595707e-05, "loss": 0.3537, "step": 8092 }, { "epoch": 0.4, "learning_rate": 1.3515089254700615e-05, "loss": 0.4194, "step": 8093 }, { "epoch": 0.4, "learning_rate": 1.3513575201511602e-05, "loss": 0.4316, "step": 8094 }, { "epoch": 0.4, "learning_rate": 1.3512061056429632e-05, "loss": 0.4376, "step": 8095 }, { "epoch": 0.4, "learning_rate": 1.3510546819494305e-05, "loss": 0.3782, "step": 8096 }, { "epoch": 0.4, "learning_rate": 1.3509032490745226e-05, "loss": 0.3119, "step": 8097 }, { "epoch": 0.4, "learning_rate": 1.3507518070221996e-05, "loss": 0.1524, "step": 8098 }, { "epoch": 0.4, "learning_rate": 1.3506003557964225e-05, "loss": 0.5348, "step": 8099 }, { "epoch": 0.4, "learning_rate": 1.3504488954011523e-05, "loss": 0.3847, "step": 8100 }, { "epoch": 0.4, "learning_rate": 1.3502974258403506e-05, "loss": 0.3214, "step": 8101 }, { "epoch": 0.4, "learning_rate": 1.350145947117978e-05, "loss": 0.5645, "step": 8102 }, { "epoch": 0.4, "learning_rate": 1.349994459237997e-05, "loss": 0.553, "step": 8103 }, { "epoch": 0.4, "learning_rate": 1.3498429622043694e-05, "loss": 0.504, "step": 8104 }, { "epoch": 0.4, "learning_rate": 1.349691456021057e-05, "loss": 0.2815, "step": 8105 }, { "epoch": 0.4, "learning_rate": 1.349539940692023e-05, "loss": 0.3981, "step": 8106 }, { "epoch": 0.4, "learning_rate": 1.3493884162212296e-05, "loss": 0.1878, "step": 8107 }, { "epoch": 0.4, "learning_rate": 1.3492368826126394e-05, "loss": 0.2463, "step": 8108 }, { "epoch": 0.4, "learning_rate": 1.3490853398702164e-05, "loss": 0.4728, "step": 8109 }, { "epoch": 0.4, "learning_rate": 1.3489337879979232e-05, "loss": 0.4347, "step": 8110 }, { "epoch": 0.41, "learning_rate": 1.3487822269997242e-05, "loss": 0.2213, "step": 8111 }, { "epoch": 0.41, "learning_rate": 1.3486306568795829e-05, "loss": 0.2061, "step": 8112 }, { "epoch": 0.41, "learning_rate": 1.3484790776414629e-05, "loss": 0.2345, "step": 8113 }, { "epoch": 0.41, "learning_rate": 1.3483274892893294e-05, "loss": 0.1861, "step": 8114 }, { "epoch": 0.41, "learning_rate": 1.3481758918271466e-05, "loss": 0.5839, "step": 8115 }, { "epoch": 0.41, "learning_rate": 1.3480242852588794e-05, "loss": 0.2395, "step": 8116 }, { "epoch": 0.41, "learning_rate": 1.3478726695884925e-05, "loss": 0.3475, "step": 8117 }, { "epoch": 0.41, "learning_rate": 1.347721044819952e-05, "loss": 0.3527, "step": 8118 }, { "epoch": 0.41, "learning_rate": 1.3475694109572227e-05, "loss": 0.4208, "step": 8119 }, { "epoch": 0.41, "learning_rate": 1.347417768004271e-05, "loss": 0.6638, "step": 8120 }, { "epoch": 0.41, "learning_rate": 1.3472661159650622e-05, "loss": 0.3339, "step": 8121 }, { "epoch": 0.41, "learning_rate": 1.3471144548435633e-05, "loss": 0.6452, "step": 8122 }, { "epoch": 0.41, "learning_rate": 1.3469627846437405e-05, "loss": 0.2386, "step": 8123 }, { "epoch": 0.41, "learning_rate": 1.3468111053695605e-05, "loss": 0.5423, "step": 8124 }, { "epoch": 0.41, "learning_rate": 1.3466594170249903e-05, "loss": 0.5675, "step": 8125 }, { "epoch": 0.41, "learning_rate": 1.3465077196139969e-05, "loss": 0.2463, "step": 8126 }, { "epoch": 0.41, "learning_rate": 1.346356013140548e-05, "loss": 0.2104, "step": 8127 }, { "epoch": 0.41, "learning_rate": 1.3462042976086112e-05, "loss": 0.4282, "step": 8128 }, { "epoch": 0.41, "learning_rate": 1.3460525730221545e-05, "loss": 0.5643, "step": 8129 }, { "epoch": 0.41, "learning_rate": 1.345900839385146e-05, "loss": 0.2835, "step": 8130 }, { "epoch": 0.41, "learning_rate": 1.3457490967015539e-05, "loss": 0.3851, "step": 8131 }, { "epoch": 0.41, "learning_rate": 1.3455973449753472e-05, "loss": 0.3003, "step": 8132 }, { "epoch": 0.41, "learning_rate": 1.3454455842104945e-05, "loss": 0.3829, "step": 8133 }, { "epoch": 0.41, "learning_rate": 1.3452938144109653e-05, "loss": 0.4988, "step": 8134 }, { "epoch": 0.41, "learning_rate": 1.3451420355807283e-05, "loss": 0.3314, "step": 8135 }, { "epoch": 0.41, "learning_rate": 1.3449902477237534e-05, "loss": 0.3553, "step": 8136 }, { "epoch": 0.41, "learning_rate": 1.3448384508440106e-05, "loss": 0.2098, "step": 8137 }, { "epoch": 0.41, "learning_rate": 1.3446866449454698e-05, "loss": 0.3825, "step": 8138 }, { "epoch": 0.41, "learning_rate": 1.3445348300321015e-05, "loss": 0.075, "step": 8139 }, { "epoch": 0.41, "learning_rate": 1.3443830061078755e-05, "loss": 0.4615, "step": 8140 }, { "epoch": 0.41, "learning_rate": 1.3442311731767632e-05, "loss": 0.2841, "step": 8141 }, { "epoch": 0.41, "learning_rate": 1.3440793312427353e-05, "loss": 0.3644, "step": 8142 }, { "epoch": 0.41, "learning_rate": 1.3439274803097635e-05, "loss": 0.131, "step": 8143 }, { "epoch": 0.41, "learning_rate": 1.3437756203818184e-05, "loss": 0.4207, "step": 8144 }, { "epoch": 0.41, "learning_rate": 1.3436237514628728e-05, "loss": 0.2067, "step": 8145 }, { "epoch": 0.41, "learning_rate": 1.3434718735568976e-05, "loss": 0.4289, "step": 8146 }, { "epoch": 0.41, "learning_rate": 1.3433199866678656e-05, "loss": 0.5081, "step": 8147 }, { "epoch": 0.41, "learning_rate": 1.343168090799749e-05, "loss": 0.4244, "step": 8148 }, { "epoch": 0.41, "learning_rate": 1.3430161859565205e-05, "loss": 0.388, "step": 8149 }, { "epoch": 0.41, "learning_rate": 1.342864272142153e-05, "loss": 0.4597, "step": 8150 }, { "epoch": 0.41, "learning_rate": 1.3427123493606194e-05, "loss": 0.5471, "step": 8151 }, { "epoch": 0.41, "learning_rate": 1.3425604176158933e-05, "loss": 0.3374, "step": 8152 }, { "epoch": 0.41, "learning_rate": 1.3424084769119479e-05, "loss": 0.4794, "step": 8153 }, { "epoch": 0.41, "learning_rate": 1.3422565272527574e-05, "loss": 0.5114, "step": 8154 }, { "epoch": 0.41, "learning_rate": 1.3421045686422958e-05, "loss": 0.3887, "step": 8155 }, { "epoch": 0.41, "learning_rate": 1.341952601084537e-05, "loss": 0.2826, "step": 8156 }, { "epoch": 0.41, "learning_rate": 1.3418006245834563e-05, "loss": 0.1125, "step": 8157 }, { "epoch": 0.41, "learning_rate": 1.3416486391430272e-05, "loss": 0.3513, "step": 8158 }, { "epoch": 0.41, "learning_rate": 1.3414966447672262e-05, "loss": 0.509, "step": 8159 }, { "epoch": 0.41, "learning_rate": 1.3413446414600273e-05, "loss": 0.333, "step": 8160 }, { "epoch": 0.41, "learning_rate": 1.3411926292254065e-05, "loss": 0.3309, "step": 8161 }, { "epoch": 0.41, "learning_rate": 1.3410406080673394e-05, "loss": 0.3668, "step": 8162 }, { "epoch": 0.41, "learning_rate": 1.3408885779898017e-05, "loss": 0.2481, "step": 8163 }, { "epoch": 0.41, "learning_rate": 1.3407365389967702e-05, "loss": 0.2454, "step": 8164 }, { "epoch": 0.41, "learning_rate": 1.3405844910922201e-05, "loss": 0.4568, "step": 8165 }, { "epoch": 0.41, "learning_rate": 1.3404324342801292e-05, "loss": 0.2557, "step": 8166 }, { "epoch": 0.41, "learning_rate": 1.3402803685644736e-05, "loss": 0.2608, "step": 8167 }, { "epoch": 0.41, "learning_rate": 1.3401282939492308e-05, "loss": 0.3628, "step": 8168 }, { "epoch": 0.41, "learning_rate": 1.339976210438378e-05, "loss": 0.5404, "step": 8169 }, { "epoch": 0.41, "learning_rate": 1.3398241180358926e-05, "loss": 0.5591, "step": 8170 }, { "epoch": 0.41, "learning_rate": 1.3396720167457526e-05, "loss": 0.402, "step": 8171 }, { "epoch": 0.41, "learning_rate": 1.3395199065719358e-05, "loss": 0.4964, "step": 8172 }, { "epoch": 0.41, "learning_rate": 1.3393677875184205e-05, "loss": 0.4285, "step": 8173 }, { "epoch": 0.41, "learning_rate": 1.3392156595891854e-05, "loss": 0.3005, "step": 8174 }, { "epoch": 0.41, "learning_rate": 1.339063522788209e-05, "loss": 0.4817, "step": 8175 }, { "epoch": 0.41, "learning_rate": 1.33891137711947e-05, "loss": 0.3817, "step": 8176 }, { "epoch": 0.41, "learning_rate": 1.338759222586948e-05, "loss": 0.4947, "step": 8177 }, { "epoch": 0.41, "learning_rate": 1.3386070591946222e-05, "loss": 0.2542, "step": 8178 }, { "epoch": 0.41, "learning_rate": 1.3384548869464723e-05, "loss": 0.4097, "step": 8179 }, { "epoch": 0.41, "learning_rate": 1.3383027058464782e-05, "loss": 0.3961, "step": 8180 }, { "epoch": 0.41, "learning_rate": 1.3381505158986196e-05, "loss": 0.3437, "step": 8181 }, { "epoch": 0.41, "learning_rate": 1.3379983171068776e-05, "loss": 0.4782, "step": 8182 }, { "epoch": 0.41, "learning_rate": 1.3378461094752318e-05, "loss": 0.2843, "step": 8183 }, { "epoch": 0.41, "learning_rate": 1.3376938930076638e-05, "loss": 0.4677, "step": 8184 }, { "epoch": 0.41, "learning_rate": 1.3375416677081546e-05, "loss": 0.2295, "step": 8185 }, { "epoch": 0.41, "learning_rate": 1.3373894335806847e-05, "loss": 0.4159, "step": 8186 }, { "epoch": 0.41, "learning_rate": 1.3372371906292365e-05, "loss": 0.295, "step": 8187 }, { "epoch": 0.41, "learning_rate": 1.337084938857791e-05, "loss": 0.3842, "step": 8188 }, { "epoch": 0.41, "learning_rate": 1.3369326782703304e-05, "loss": 0.3766, "step": 8189 }, { "epoch": 0.41, "learning_rate": 1.336780408870837e-05, "loss": 0.3139, "step": 8190 }, { "epoch": 0.41, "learning_rate": 1.336628130663293e-05, "loss": 0.2574, "step": 8191 }, { "epoch": 0.41, "learning_rate": 1.3364758436516811e-05, "loss": 0.2838, "step": 8192 }, { "epoch": 0.41, "learning_rate": 1.3363235478399843e-05, "loss": 0.1965, "step": 8193 }, { "epoch": 0.41, "learning_rate": 1.3361712432321856e-05, "loss": 0.3138, "step": 8194 }, { "epoch": 0.41, "learning_rate": 1.3360189298322686e-05, "loss": 0.2481, "step": 8195 }, { "epoch": 0.41, "learning_rate": 1.3358666076442164e-05, "loss": 0.3078, "step": 8196 }, { "epoch": 0.41, "learning_rate": 1.335714276672013e-05, "loss": 0.2448, "step": 8197 }, { "epoch": 0.41, "learning_rate": 1.3355619369196423e-05, "loss": 0.488, "step": 8198 }, { "epoch": 0.41, "learning_rate": 1.3354095883910888e-05, "loss": 0.3174, "step": 8199 }, { "epoch": 0.41, "learning_rate": 1.3352572310903367e-05, "loss": 0.3227, "step": 8200 }, { "epoch": 0.41, "learning_rate": 1.3351048650213708e-05, "loss": 0.4263, "step": 8201 }, { "epoch": 0.41, "learning_rate": 1.3349524901881763e-05, "loss": 0.3143, "step": 8202 }, { "epoch": 0.41, "learning_rate": 1.334800106594738e-05, "loss": 0.4373, "step": 8203 }, { "epoch": 0.41, "learning_rate": 1.3346477142450412e-05, "loss": 0.3291, "step": 8204 }, { "epoch": 0.41, "learning_rate": 1.3344953131430721e-05, "loss": 0.2673, "step": 8205 }, { "epoch": 0.41, "learning_rate": 1.3343429032928156e-05, "loss": 0.2934, "step": 8206 }, { "epoch": 0.41, "learning_rate": 1.3341904846982589e-05, "loss": 0.3247, "step": 8207 }, { "epoch": 0.41, "learning_rate": 1.3340380573633875e-05, "loss": 0.4479, "step": 8208 }, { "epoch": 0.41, "learning_rate": 1.3338856212921887e-05, "loss": 0.1604, "step": 8209 }, { "epoch": 0.41, "learning_rate": 1.3337331764886486e-05, "loss": 0.3521, "step": 8210 }, { "epoch": 0.41, "learning_rate": 1.3335807229567541e-05, "loss": 0.3544, "step": 8211 }, { "epoch": 0.41, "learning_rate": 1.3334282607004931e-05, "loss": 0.2842, "step": 8212 }, { "epoch": 0.41, "learning_rate": 1.3332757897238522e-05, "loss": 0.4117, "step": 8213 }, { "epoch": 0.41, "learning_rate": 1.3331233100308199e-05, "loss": 0.4466, "step": 8214 }, { "epoch": 0.41, "learning_rate": 1.3329708216253834e-05, "loss": 0.3997, "step": 8215 }, { "epoch": 0.41, "learning_rate": 1.3328183245115317e-05, "loss": 0.4878, "step": 8216 }, { "epoch": 0.41, "learning_rate": 1.332665818693252e-05, "loss": 0.177, "step": 8217 }, { "epoch": 0.41, "learning_rate": 1.332513304174534e-05, "loss": 0.5016, "step": 8218 }, { "epoch": 0.41, "learning_rate": 1.3323607809593657e-05, "loss": 0.3378, "step": 8219 }, { "epoch": 0.41, "learning_rate": 1.3322082490517367e-05, "loss": 0.212, "step": 8220 }, { "epoch": 0.41, "learning_rate": 1.332055708455636e-05, "loss": 0.2968, "step": 8221 }, { "epoch": 0.41, "learning_rate": 1.3319031591750533e-05, "loss": 0.6797, "step": 8222 }, { "epoch": 0.41, "learning_rate": 1.3317506012139782e-05, "loss": 0.3705, "step": 8223 }, { "epoch": 0.41, "learning_rate": 1.3315980345764004e-05, "loss": 0.3371, "step": 8224 }, { "epoch": 0.41, "learning_rate": 1.3314454592663104e-05, "loss": 0.472, "step": 8225 }, { "epoch": 0.41, "learning_rate": 1.3312928752876986e-05, "loss": 0.485, "step": 8226 }, { "epoch": 0.41, "learning_rate": 1.3311402826445554e-05, "loss": 0.3726, "step": 8227 }, { "epoch": 0.41, "learning_rate": 1.3309876813408721e-05, "loss": 0.2884, "step": 8228 }, { "epoch": 0.41, "learning_rate": 1.3308350713806392e-05, "loss": 0.3652, "step": 8229 }, { "epoch": 0.41, "learning_rate": 1.3306824527678485e-05, "loss": 0.1826, "step": 8230 }, { "epoch": 0.41, "learning_rate": 1.3305298255064913e-05, "loss": 0.4718, "step": 8231 }, { "epoch": 0.41, "learning_rate": 1.3303771896005596e-05, "loss": 0.3204, "step": 8232 }, { "epoch": 0.41, "learning_rate": 1.3302245450540453e-05, "loss": 0.4643, "step": 8233 }, { "epoch": 0.41, "learning_rate": 1.3300718918709403e-05, "loss": 0.5927, "step": 8234 }, { "epoch": 0.41, "learning_rate": 1.3299192300552375e-05, "loss": 0.4245, "step": 8235 }, { "epoch": 0.41, "learning_rate": 1.3297665596109288e-05, "loss": 0.5267, "step": 8236 }, { "epoch": 0.41, "learning_rate": 1.3296138805420082e-05, "loss": 0.4344, "step": 8237 }, { "epoch": 0.41, "learning_rate": 1.3294611928524679e-05, "loss": 0.5055, "step": 8238 }, { "epoch": 0.41, "learning_rate": 1.329308496546302e-05, "loss": 0.3882, "step": 8239 }, { "epoch": 0.41, "learning_rate": 1.3291557916275033e-05, "loss": 0.4092, "step": 8240 }, { "epoch": 0.41, "learning_rate": 1.3290030781000663e-05, "loss": 0.2279, "step": 8241 }, { "epoch": 0.41, "learning_rate": 1.3288503559679844e-05, "loss": 0.2301, "step": 8242 }, { "epoch": 0.41, "learning_rate": 1.3286976252352526e-05, "loss": 0.0974, "step": 8243 }, { "epoch": 0.41, "learning_rate": 1.3285448859058648e-05, "loss": 0.3283, "step": 8244 }, { "epoch": 0.41, "learning_rate": 1.3283921379838153e-05, "loss": 0.3884, "step": 8245 }, { "epoch": 0.41, "learning_rate": 1.3282393814731003e-05, "loss": 0.4387, "step": 8246 }, { "epoch": 0.41, "learning_rate": 1.3280866163777138e-05, "loss": 0.1957, "step": 8247 }, { "epoch": 0.41, "learning_rate": 1.3279338427016518e-05, "loss": 0.2865, "step": 8248 }, { "epoch": 0.41, "learning_rate": 1.3277810604489092e-05, "loss": 0.3798, "step": 8249 }, { "epoch": 0.41, "learning_rate": 1.3276282696234826e-05, "loss": 0.4684, "step": 8250 }, { "epoch": 0.41, "learning_rate": 1.3274754702293678e-05, "loss": 0.356, "step": 8251 }, { "epoch": 0.41, "learning_rate": 1.3273226622705609e-05, "loss": 0.4216, "step": 8252 }, { "epoch": 0.41, "learning_rate": 1.3271698457510584e-05, "loss": 0.4032, "step": 8253 }, { "epoch": 0.41, "learning_rate": 1.3270170206748572e-05, "loss": 0.3116, "step": 8254 }, { "epoch": 0.41, "learning_rate": 1.326864187045954e-05, "loss": 0.3109, "step": 8255 }, { "epoch": 0.41, "learning_rate": 1.326711344868346e-05, "loss": 0.4261, "step": 8256 }, { "epoch": 0.41, "learning_rate": 1.326558494146031e-05, "loss": 0.4036, "step": 8257 }, { "epoch": 0.41, "learning_rate": 1.3264056348830064e-05, "loss": 0.3077, "step": 8258 }, { "epoch": 0.41, "learning_rate": 1.3262527670832697e-05, "loss": 0.4003, "step": 8259 }, { "epoch": 0.41, "learning_rate": 1.3260998907508193e-05, "loss": 0.5373, "step": 8260 }, { "epoch": 0.41, "learning_rate": 1.325947005889653e-05, "loss": 0.4677, "step": 8261 }, { "epoch": 0.41, "learning_rate": 1.32579411250377e-05, "loss": 0.2276, "step": 8262 }, { "epoch": 0.41, "learning_rate": 1.3256412105971687e-05, "loss": 0.3147, "step": 8263 }, { "epoch": 0.41, "learning_rate": 1.3254883001738482e-05, "loss": 0.2691, "step": 8264 }, { "epoch": 0.41, "learning_rate": 1.3253353812378072e-05, "loss": 0.4414, "step": 8265 }, { "epoch": 0.41, "learning_rate": 1.3251824537930455e-05, "loss": 0.2792, "step": 8266 }, { "epoch": 0.41, "learning_rate": 1.3250295178435627e-05, "loss": 0.3207, "step": 8267 }, { "epoch": 0.41, "learning_rate": 1.3248765733933585e-05, "loss": 0.5206, "step": 8268 }, { "epoch": 0.41, "learning_rate": 1.324723620446433e-05, "loss": 0.4942, "step": 8269 }, { "epoch": 0.41, "learning_rate": 1.3245706590067868e-05, "loss": 0.2345, "step": 8270 }, { "epoch": 0.41, "learning_rate": 1.3244176890784202e-05, "loss": 0.2441, "step": 8271 }, { "epoch": 0.41, "learning_rate": 1.3242647106653333e-05, "loss": 0.2761, "step": 8272 }, { "epoch": 0.41, "learning_rate": 1.324111723771528e-05, "loss": 0.2085, "step": 8273 }, { "epoch": 0.41, "learning_rate": 1.323958728401005e-05, "loss": 0.388, "step": 8274 }, { "epoch": 0.41, "learning_rate": 1.3238057245577655e-05, "loss": 0.3763, "step": 8275 }, { "epoch": 0.41, "learning_rate": 1.3236527122458118e-05, "loss": 0.4209, "step": 8276 }, { "epoch": 0.41, "learning_rate": 1.323499691469145e-05, "loss": 0.4656, "step": 8277 }, { "epoch": 0.41, "learning_rate": 1.3233466622317676e-05, "loss": 0.4895, "step": 8278 }, { "epoch": 0.41, "learning_rate": 1.3231936245376818e-05, "loss": 0.426, "step": 8279 }, { "epoch": 0.41, "learning_rate": 1.3230405783908901e-05, "loss": 0.2118, "step": 8280 }, { "epoch": 0.41, "learning_rate": 1.3228875237953952e-05, "loss": 0.2876, "step": 8281 }, { "epoch": 0.41, "learning_rate": 1.3227344607552e-05, "loss": 0.5119, "step": 8282 }, { "epoch": 0.41, "learning_rate": 1.3225813892743075e-05, "loss": 0.296, "step": 8283 }, { "epoch": 0.41, "learning_rate": 1.3224283093567217e-05, "loss": 0.3768, "step": 8284 }, { "epoch": 0.41, "learning_rate": 1.3222752210064454e-05, "loss": 0.344, "step": 8285 }, { "epoch": 0.41, "learning_rate": 1.3221221242274827e-05, "loss": 0.3294, "step": 8286 }, { "epoch": 0.41, "learning_rate": 1.321969019023838e-05, "loss": 0.5471, "step": 8287 }, { "epoch": 0.41, "learning_rate": 1.321815905399515e-05, "loss": 0.1774, "step": 8288 }, { "epoch": 0.41, "learning_rate": 1.321662783358519e-05, "loss": 0.589, "step": 8289 }, { "epoch": 0.41, "learning_rate": 1.3215096529048536e-05, "loss": 0.2916, "step": 8290 }, { "epoch": 0.41, "learning_rate": 1.3213565140425248e-05, "loss": 0.4475, "step": 8291 }, { "epoch": 0.41, "learning_rate": 1.3212033667755371e-05, "loss": 0.5048, "step": 8292 }, { "epoch": 0.41, "learning_rate": 1.321050211107896e-05, "loss": 0.4764, "step": 8293 }, { "epoch": 0.41, "learning_rate": 1.3208970470436073e-05, "loss": 0.4152, "step": 8294 }, { "epoch": 0.41, "learning_rate": 1.3207438745866765e-05, "loss": 0.4391, "step": 8295 }, { "epoch": 0.41, "learning_rate": 1.32059069374111e-05, "loss": 0.3563, "step": 8296 }, { "epoch": 0.41, "learning_rate": 1.3204375045109133e-05, "loss": 0.7042, "step": 8297 }, { "epoch": 0.41, "learning_rate": 1.320284306900094e-05, "loss": 0.1849, "step": 8298 }, { "epoch": 0.41, "learning_rate": 1.3201311009126577e-05, "loss": 0.3805, "step": 8299 }, { "epoch": 0.41, "learning_rate": 1.3199778865526116e-05, "loss": 0.3287, "step": 8300 }, { "epoch": 0.41, "learning_rate": 1.3198246638239635e-05, "loss": 0.3279, "step": 8301 }, { "epoch": 0.41, "learning_rate": 1.3196714327307199e-05, "loss": 0.3048, "step": 8302 }, { "epoch": 0.41, "learning_rate": 1.3195181932768885e-05, "loss": 0.1423, "step": 8303 }, { "epoch": 0.41, "learning_rate": 1.3193649454664775e-05, "loss": 0.2944, "step": 8304 }, { "epoch": 0.41, "learning_rate": 1.3192116893034948e-05, "loss": 0.3545, "step": 8305 }, { "epoch": 0.41, "learning_rate": 1.3190584247919483e-05, "loss": 0.3441, "step": 8306 }, { "epoch": 0.41, "learning_rate": 1.3189051519358467e-05, "loss": 0.2475, "step": 8307 }, { "epoch": 0.41, "learning_rate": 1.3187518707391985e-05, "loss": 0.3581, "step": 8308 }, { "epoch": 0.41, "learning_rate": 1.3185985812060122e-05, "loss": 0.5468, "step": 8309 }, { "epoch": 0.41, "learning_rate": 1.3184452833402977e-05, "loss": 0.2947, "step": 8310 }, { "epoch": 0.41, "learning_rate": 1.3182919771460636e-05, "loss": 0.2003, "step": 8311 }, { "epoch": 0.42, "learning_rate": 1.31813866262732e-05, "loss": 0.3692, "step": 8312 }, { "epoch": 0.42, "learning_rate": 1.3179853397880763e-05, "loss": 0.3284, "step": 8313 }, { "epoch": 0.42, "learning_rate": 1.3178320086323427e-05, "loss": 0.1786, "step": 8314 }, { "epoch": 0.42, "learning_rate": 1.317678669164129e-05, "loss": 0.31, "step": 8315 }, { "epoch": 0.42, "learning_rate": 1.3175253213874458e-05, "loss": 0.4577, "step": 8316 }, { "epoch": 0.42, "learning_rate": 1.3173719653063037e-05, "loss": 0.5162, "step": 8317 }, { "epoch": 0.42, "learning_rate": 1.3172186009247134e-05, "loss": 0.5629, "step": 8318 }, { "epoch": 0.42, "learning_rate": 1.3170652282466863e-05, "loss": 0.3494, "step": 8319 }, { "epoch": 0.42, "learning_rate": 1.3169118472762334e-05, "loss": 0.1783, "step": 8320 }, { "epoch": 0.42, "learning_rate": 1.3167584580173662e-05, "loss": 0.4228, "step": 8321 }, { "epoch": 0.42, "learning_rate": 1.3166050604740962e-05, "loss": 0.4655, "step": 8322 }, { "epoch": 0.42, "learning_rate": 1.316451654650436e-05, "loss": 0.4296, "step": 8323 }, { "epoch": 0.42, "learning_rate": 1.316298240550397e-05, "loss": 0.4891, "step": 8324 }, { "epoch": 0.42, "learning_rate": 1.3161448181779917e-05, "loss": 0.4064, "step": 8325 }, { "epoch": 0.42, "learning_rate": 1.3159913875372328e-05, "loss": 0.3621, "step": 8326 }, { "epoch": 0.42, "learning_rate": 1.3158379486321329e-05, "loss": 0.3922, "step": 8327 }, { "epoch": 0.42, "learning_rate": 1.3156845014667052e-05, "loss": 0.4283, "step": 8328 }, { "epoch": 0.42, "learning_rate": 1.315531046044963e-05, "loss": 0.3017, "step": 8329 }, { "epoch": 0.42, "learning_rate": 1.3153775823709195e-05, "loss": 0.5155, "step": 8330 }, { "epoch": 0.42, "learning_rate": 1.3152241104485883e-05, "loss": 0.5152, "step": 8331 }, { "epoch": 0.42, "learning_rate": 1.3150706302819836e-05, "loss": 0.3696, "step": 8332 }, { "epoch": 0.42, "learning_rate": 1.3149171418751192e-05, "loss": 0.3965, "step": 8333 }, { "epoch": 0.42, "learning_rate": 1.3147636452320091e-05, "loss": 0.3558, "step": 8334 }, { "epoch": 0.42, "learning_rate": 1.3146101403566685e-05, "loss": 0.3512, "step": 8335 }, { "epoch": 0.42, "learning_rate": 1.3144566272531114e-05, "loss": 0.2384, "step": 8336 }, { "epoch": 0.42, "learning_rate": 1.3143031059253533e-05, "loss": 0.3333, "step": 8337 }, { "epoch": 0.42, "learning_rate": 1.314149576377409e-05, "loss": 0.3672, "step": 8338 }, { "epoch": 0.42, "learning_rate": 1.3139960386132939e-05, "loss": 0.3316, "step": 8339 }, { "epoch": 0.42, "learning_rate": 1.313842492637024e-05, "loss": 0.3921, "step": 8340 }, { "epoch": 0.42, "learning_rate": 1.3136889384526146e-05, "loss": 0.2434, "step": 8341 }, { "epoch": 0.42, "learning_rate": 1.3135353760640819e-05, "loss": 0.2919, "step": 8342 }, { "epoch": 0.42, "learning_rate": 1.3133818054754421e-05, "loss": 0.3969, "step": 8343 }, { "epoch": 0.42, "learning_rate": 1.3132282266907118e-05, "loss": 0.2817, "step": 8344 }, { "epoch": 0.42, "learning_rate": 1.3130746397139072e-05, "loss": 0.1639, "step": 8345 }, { "epoch": 0.42, "learning_rate": 1.3129210445490455e-05, "loss": 0.3229, "step": 8346 }, { "epoch": 0.42, "learning_rate": 1.3127674412001437e-05, "loss": 0.3132, "step": 8347 }, { "epoch": 0.42, "learning_rate": 1.3126138296712192e-05, "loss": 0.2596, "step": 8348 }, { "epoch": 0.42, "learning_rate": 1.3124602099662895e-05, "loss": 0.4706, "step": 8349 }, { "epoch": 0.42, "learning_rate": 1.312306582089372e-05, "loss": 0.4778, "step": 8350 }, { "epoch": 0.42, "learning_rate": 1.3121529460444851e-05, "loss": 0.3749, "step": 8351 }, { "epoch": 0.42, "learning_rate": 1.3119993018356465e-05, "loss": 0.5133, "step": 8352 }, { "epoch": 0.42, "learning_rate": 1.3118456494668752e-05, "loss": 0.1822, "step": 8353 }, { "epoch": 0.42, "learning_rate": 1.3116919889421891e-05, "loss": 0.1847, "step": 8354 }, { "epoch": 0.42, "learning_rate": 1.3115383202656075e-05, "loss": 0.2924, "step": 8355 }, { "epoch": 0.42, "learning_rate": 1.3113846434411493e-05, "loss": 0.3457, "step": 8356 }, { "epoch": 0.42, "learning_rate": 1.3112309584728334e-05, "loss": 0.4611, "step": 8357 }, { "epoch": 0.42, "learning_rate": 1.3110772653646792e-05, "loss": 0.375, "step": 8358 }, { "epoch": 0.42, "learning_rate": 1.3109235641207067e-05, "loss": 0.3087, "step": 8359 }, { "epoch": 0.42, "learning_rate": 1.3107698547449357e-05, "loss": 0.2213, "step": 8360 }, { "epoch": 0.42, "learning_rate": 1.310616137241386e-05, "loss": 0.4536, "step": 8361 }, { "epoch": 0.42, "learning_rate": 1.3104624116140782e-05, "loss": 0.5047, "step": 8362 }, { "epoch": 0.42, "learning_rate": 1.3103086778670326e-05, "loss": 0.3656, "step": 8363 }, { "epoch": 0.42, "learning_rate": 1.31015493600427e-05, "loss": 0.3176, "step": 8364 }, { "epoch": 0.42, "learning_rate": 1.3100011860298113e-05, "loss": 0.4452, "step": 8365 }, { "epoch": 0.42, "learning_rate": 1.3098474279476777e-05, "loss": 0.2762, "step": 8366 }, { "epoch": 0.42, "learning_rate": 1.3096936617618905e-05, "loss": 0.5057, "step": 8367 }, { "epoch": 0.42, "learning_rate": 1.309539887476471e-05, "loss": 0.32, "step": 8368 }, { "epoch": 0.42, "learning_rate": 1.3093861050954412e-05, "loss": 0.4659, "step": 8369 }, { "epoch": 0.42, "learning_rate": 1.3092323146228232e-05, "loss": 0.5217, "step": 8370 }, { "epoch": 0.42, "learning_rate": 1.3090785160626389e-05, "loss": 0.5072, "step": 8371 }, { "epoch": 0.42, "learning_rate": 1.3089247094189108e-05, "loss": 0.3832, "step": 8372 }, { "epoch": 0.42, "learning_rate": 1.3087708946956616e-05, "loss": 0.3789, "step": 8373 }, { "epoch": 0.42, "learning_rate": 1.308617071896914e-05, "loss": 0.445, "step": 8374 }, { "epoch": 0.42, "learning_rate": 1.308463241026691e-05, "loss": 0.5865, "step": 8375 }, { "epoch": 0.42, "learning_rate": 1.3083094020890164e-05, "loss": 0.4623, "step": 8376 }, { "epoch": 0.42, "learning_rate": 1.3081555550879126e-05, "loss": 0.3589, "step": 8377 }, { "epoch": 0.42, "learning_rate": 1.308001700027404e-05, "loss": 0.3317, "step": 8378 }, { "epoch": 0.42, "learning_rate": 1.3078478369115149e-05, "loss": 0.4358, "step": 8379 }, { "epoch": 0.42, "learning_rate": 1.3076939657442684e-05, "loss": 0.3111, "step": 8380 }, { "epoch": 0.42, "learning_rate": 1.3075400865296892e-05, "loss": 0.2376, "step": 8381 }, { "epoch": 0.42, "learning_rate": 1.307386199271802e-05, "loss": 0.5898, "step": 8382 }, { "epoch": 0.42, "learning_rate": 1.3072323039746313e-05, "loss": 0.2683, "step": 8383 }, { "epoch": 0.42, "learning_rate": 1.3070784006422019e-05, "loss": 0.4278, "step": 8384 }, { "epoch": 0.42, "learning_rate": 1.3069244892785395e-05, "loss": 0.3749, "step": 8385 }, { "epoch": 0.42, "learning_rate": 1.3067705698876687e-05, "loss": 0.2192, "step": 8386 }, { "epoch": 0.42, "learning_rate": 1.3066166424736158e-05, "loss": 0.2305, "step": 8387 }, { "epoch": 0.42, "learning_rate": 1.306462707040406e-05, "loss": 0.5061, "step": 8388 }, { "epoch": 0.42, "learning_rate": 1.3063087635920652e-05, "loss": 0.2654, "step": 8389 }, { "epoch": 0.42, "learning_rate": 1.3061548121326204e-05, "loss": 0.4918, "step": 8390 }, { "epoch": 0.42, "learning_rate": 1.3060008526660972e-05, "loss": 0.3164, "step": 8391 }, { "epoch": 0.42, "learning_rate": 1.3058468851965226e-05, "loss": 0.257, "step": 8392 }, { "epoch": 0.42, "learning_rate": 1.3056929097279233e-05, "loss": 0.4259, "step": 8393 }, { "epoch": 0.42, "learning_rate": 1.3055389262643265e-05, "loss": 0.3377, "step": 8394 }, { "epoch": 0.42, "learning_rate": 1.305384934809759e-05, "loss": 0.3409, "step": 8395 }, { "epoch": 0.42, "learning_rate": 1.3052309353682487e-05, "loss": 0.4675, "step": 8396 }, { "epoch": 0.42, "learning_rate": 1.3050769279438228e-05, "loss": 0.3485, "step": 8397 }, { "epoch": 0.42, "learning_rate": 1.3049229125405096e-05, "loss": 0.4164, "step": 8398 }, { "epoch": 0.42, "learning_rate": 1.3047688891623371e-05, "loss": 0.4989, "step": 8399 }, { "epoch": 0.42, "learning_rate": 1.3046148578133332e-05, "loss": 0.2892, "step": 8400 }, { "epoch": 0.42, "learning_rate": 1.3044608184975269e-05, "loss": 0.304, "step": 8401 }, { "epoch": 0.42, "learning_rate": 1.3043067712189466e-05, "loss": 0.2465, "step": 8402 }, { "epoch": 0.42, "learning_rate": 1.3041527159816212e-05, "loss": 0.5003, "step": 8403 }, { "epoch": 0.42, "learning_rate": 1.30399865278958e-05, "loss": 0.1977, "step": 8404 }, { "epoch": 0.42, "learning_rate": 1.3038445816468524e-05, "loss": 0.2485, "step": 8405 }, { "epoch": 0.42, "learning_rate": 1.3036905025574673e-05, "loss": 0.2976, "step": 8406 }, { "epoch": 0.42, "learning_rate": 1.3035364155254551e-05, "loss": 0.3819, "step": 8407 }, { "epoch": 0.42, "learning_rate": 1.3033823205548455e-05, "loss": 0.3992, "step": 8408 }, { "epoch": 0.42, "learning_rate": 1.3032282176496685e-05, "loss": 0.2693, "step": 8409 }, { "epoch": 0.42, "learning_rate": 1.3030741068139548e-05, "loss": 0.1812, "step": 8410 }, { "epoch": 0.42, "learning_rate": 1.3029199880517345e-05, "loss": 0.2949, "step": 8411 }, { "epoch": 0.42, "learning_rate": 1.302765861367039e-05, "loss": 0.2398, "step": 8412 }, { "epoch": 0.42, "learning_rate": 1.3026117267638991e-05, "loss": 0.4652, "step": 8413 }, { "epoch": 0.42, "learning_rate": 1.3024575842463454e-05, "loss": 0.4059, "step": 8414 }, { "epoch": 0.42, "learning_rate": 1.30230343381841e-05, "loss": 0.4959, "step": 8415 }, { "epoch": 0.42, "learning_rate": 1.3021492754841243e-05, "loss": 0.2289, "step": 8416 }, { "epoch": 0.42, "learning_rate": 1.3019951092475198e-05, "loss": 0.5961, "step": 8417 }, { "epoch": 0.42, "learning_rate": 1.3018409351126289e-05, "loss": 0.626, "step": 8418 }, { "epoch": 0.42, "learning_rate": 1.3016867530834839e-05, "loss": 0.3383, "step": 8419 }, { "epoch": 0.42, "learning_rate": 1.3015325631641167e-05, "loss": 0.1089, "step": 8420 }, { "epoch": 0.42, "learning_rate": 1.3013783653585602e-05, "loss": 0.4514, "step": 8421 }, { "epoch": 0.42, "learning_rate": 1.3012241596708473e-05, "loss": 0.2153, "step": 8422 }, { "epoch": 0.42, "learning_rate": 1.301069946105011e-05, "loss": 0.2491, "step": 8423 }, { "epoch": 0.42, "learning_rate": 1.3009157246650847e-05, "loss": 0.3875, "step": 8424 }, { "epoch": 0.42, "learning_rate": 1.3007614953551017e-05, "loss": 0.5717, "step": 8425 }, { "epoch": 0.42, "learning_rate": 1.300607258179096e-05, "loss": 0.2399, "step": 8426 }, { "epoch": 0.42, "learning_rate": 1.3004530131411008e-05, "loss": 0.3272, "step": 8427 }, { "epoch": 0.42, "learning_rate": 1.300298760245151e-05, "loss": 0.4031, "step": 8428 }, { "epoch": 0.42, "learning_rate": 1.30014449949528e-05, "loss": 0.2739, "step": 8429 }, { "epoch": 0.42, "learning_rate": 1.299990230895523e-05, "loss": 0.2488, "step": 8430 }, { "epoch": 0.42, "learning_rate": 1.2998359544499147e-05, "loss": 0.4346, "step": 8431 }, { "epoch": 0.42, "learning_rate": 1.2996816701624891e-05, "loss": 0.1827, "step": 8432 }, { "epoch": 0.42, "learning_rate": 1.2995273780372823e-05, "loss": 0.3488, "step": 8433 }, { "epoch": 0.42, "learning_rate": 1.299373078078329e-05, "loss": 0.3321, "step": 8434 }, { "epoch": 0.42, "learning_rate": 1.2992187702896653e-05, "loss": 0.4088, "step": 8435 }, { "epoch": 0.42, "learning_rate": 1.2990644546753263e-05, "loss": 0.3882, "step": 8436 }, { "epoch": 0.42, "learning_rate": 1.2989101312393481e-05, "loss": 0.2504, "step": 8437 }, { "epoch": 0.42, "learning_rate": 1.2987557999857672e-05, "loss": 0.4658, "step": 8438 }, { "epoch": 0.42, "learning_rate": 1.2986014609186195e-05, "loss": 0.3899, "step": 8439 }, { "epoch": 0.42, "learning_rate": 1.2984471140419418e-05, "loss": 0.2766, "step": 8440 }, { "epoch": 0.42, "learning_rate": 1.2982927593597706e-05, "loss": 0.3203, "step": 8441 }, { "epoch": 0.42, "learning_rate": 1.298138396876143e-05, "loss": 0.3242, "step": 8442 }, { "epoch": 0.42, "learning_rate": 1.2979840265950961e-05, "loss": 0.3244, "step": 8443 }, { "epoch": 0.42, "learning_rate": 1.2978296485206674e-05, "loss": 0.4131, "step": 8444 }, { "epoch": 0.42, "learning_rate": 1.2976752626568943e-05, "loss": 0.5345, "step": 8445 }, { "epoch": 0.42, "learning_rate": 1.2975208690078143e-05, "loss": 0.3816, "step": 8446 }, { "epoch": 0.42, "learning_rate": 1.297366467577466e-05, "loss": 0.2235, "step": 8447 }, { "epoch": 0.42, "learning_rate": 1.2972120583698871e-05, "loss": 0.1591, "step": 8448 }, { "epoch": 0.42, "learning_rate": 1.297057641389116e-05, "loss": 0.4458, "step": 8449 }, { "epoch": 0.42, "learning_rate": 1.2969032166391914e-05, "loss": 0.3026, "step": 8450 }, { "epoch": 0.42, "learning_rate": 1.2967487841241522e-05, "loss": 0.2376, "step": 8451 }, { "epoch": 0.42, "learning_rate": 1.2965943438480371e-05, "loss": 0.3527, "step": 8452 }, { "epoch": 0.42, "learning_rate": 1.2964398958148856e-05, "loss": 0.4728, "step": 8453 }, { "epoch": 0.42, "learning_rate": 1.2962854400287366e-05, "loss": 0.457, "step": 8454 }, { "epoch": 0.42, "learning_rate": 1.29613097649363e-05, "loss": 0.296, "step": 8455 }, { "epoch": 0.42, "learning_rate": 1.295976505213606e-05, "loss": 0.3495, "step": 8456 }, { "epoch": 0.42, "learning_rate": 1.2958220261927035e-05, "loss": 0.39, "step": 8457 }, { "epoch": 0.42, "learning_rate": 1.2956675394349639e-05, "loss": 0.3845, "step": 8458 }, { "epoch": 0.42, "learning_rate": 1.2955130449444267e-05, "loss": 0.3418, "step": 8459 }, { "epoch": 0.42, "learning_rate": 1.2953585427251331e-05, "loss": 0.2214, "step": 8460 }, { "epoch": 0.42, "learning_rate": 1.2952040327811237e-05, "loss": 0.36, "step": 8461 }, { "epoch": 0.42, "learning_rate": 1.295049515116439e-05, "loss": 0.4257, "step": 8462 }, { "epoch": 0.42, "learning_rate": 1.2948949897351214e-05, "loss": 0.2375, "step": 8463 }, { "epoch": 0.42, "learning_rate": 1.294740456641211e-05, "loss": 0.3012, "step": 8464 }, { "epoch": 0.42, "learning_rate": 1.2945859158387504e-05, "loss": 0.467, "step": 8465 }, { "epoch": 0.42, "learning_rate": 1.2944313673317807e-05, "loss": 0.4314, "step": 8466 }, { "epoch": 0.42, "learning_rate": 1.2942768111243444e-05, "loss": 0.5923, "step": 8467 }, { "epoch": 0.42, "learning_rate": 1.2941222472204835e-05, "loss": 0.2891, "step": 8468 }, { "epoch": 0.42, "learning_rate": 1.2939676756242405e-05, "loss": 0.3917, "step": 8469 }, { "epoch": 0.42, "learning_rate": 1.293813096339658e-05, "loss": 0.4137, "step": 8470 }, { "epoch": 0.42, "learning_rate": 1.2936585093707785e-05, "loss": 0.376, "step": 8471 }, { "epoch": 0.42, "learning_rate": 1.2935039147216456e-05, "loss": 0.449, "step": 8472 }, { "epoch": 0.42, "learning_rate": 1.2933493123963021e-05, "loss": 0.5293, "step": 8473 }, { "epoch": 0.42, "learning_rate": 1.2931947023987916e-05, "loss": 0.3024, "step": 8474 }, { "epoch": 0.42, "learning_rate": 1.2930400847331576e-05, "loss": 0.5994, "step": 8475 }, { "epoch": 0.42, "learning_rate": 1.2928854594034441e-05, "loss": 0.4302, "step": 8476 }, { "epoch": 0.42, "learning_rate": 1.2927308264136951e-05, "loss": 0.4203, "step": 8477 }, { "epoch": 0.42, "learning_rate": 1.2925761857679549e-05, "loss": 0.5195, "step": 8478 }, { "epoch": 0.42, "learning_rate": 1.2924215374702675e-05, "loss": 0.5074, "step": 8479 }, { "epoch": 0.42, "learning_rate": 1.2922668815246775e-05, "loss": 0.3931, "step": 8480 }, { "epoch": 0.42, "learning_rate": 1.2921122179352305e-05, "loss": 0.4528, "step": 8481 }, { "epoch": 0.42, "learning_rate": 1.2919575467059709e-05, "loss": 0.2366, "step": 8482 }, { "epoch": 0.42, "learning_rate": 1.2918028678409437e-05, "loss": 0.287, "step": 8483 }, { "epoch": 0.42, "learning_rate": 1.291648181344195e-05, "loss": 0.5367, "step": 8484 }, { "epoch": 0.42, "learning_rate": 1.2914934872197702e-05, "loss": 0.415, "step": 8485 }, { "epoch": 0.42, "learning_rate": 1.2913387854717149e-05, "loss": 0.4008, "step": 8486 }, { "epoch": 0.42, "learning_rate": 1.2911840761040753e-05, "loss": 0.1856, "step": 8487 }, { "epoch": 0.42, "learning_rate": 1.2910293591208975e-05, "loss": 0.3928, "step": 8488 }, { "epoch": 0.42, "learning_rate": 1.2908746345262283e-05, "loss": 0.3758, "step": 8489 }, { "epoch": 0.42, "learning_rate": 1.2907199023241135e-05, "loss": 0.266, "step": 8490 }, { "epoch": 0.42, "learning_rate": 1.290565162518601e-05, "loss": 0.3352, "step": 8491 }, { "epoch": 0.42, "learning_rate": 1.290410415113737e-05, "loss": 0.4225, "step": 8492 }, { "epoch": 0.42, "learning_rate": 1.290255660113569e-05, "loss": 0.4738, "step": 8493 }, { "epoch": 0.42, "learning_rate": 1.290100897522144e-05, "loss": 0.2727, "step": 8494 }, { "epoch": 0.42, "learning_rate": 1.2899461273435103e-05, "loss": 0.2149, "step": 8495 }, { "epoch": 0.42, "learning_rate": 1.2897913495817154e-05, "loss": 0.3299, "step": 8496 }, { "epoch": 0.42, "learning_rate": 1.2896365642408073e-05, "loss": 0.4049, "step": 8497 }, { "epoch": 0.42, "learning_rate": 1.289481771324834e-05, "loss": 0.3098, "step": 8498 }, { "epoch": 0.42, "learning_rate": 1.2893269708378444e-05, "loss": 0.4124, "step": 8499 }, { "epoch": 0.42, "learning_rate": 1.2891721627838867e-05, "loss": 0.2161, "step": 8500 }, { "epoch": 0.42, "learning_rate": 1.2890173471670097e-05, "loss": 0.6024, "step": 8501 }, { "epoch": 0.42, "learning_rate": 1.2888625239912628e-05, "loss": 0.5694, "step": 8502 }, { "epoch": 0.42, "learning_rate": 1.288707693260695e-05, "loss": 0.2086, "step": 8503 }, { "epoch": 0.42, "learning_rate": 1.2885528549793555e-05, "loss": 0.2944, "step": 8504 }, { "epoch": 0.42, "learning_rate": 1.2883980091512939e-05, "loss": 0.2943, "step": 8505 }, { "epoch": 0.42, "learning_rate": 1.2882431557805601e-05, "loss": 0.4354, "step": 8506 }, { "epoch": 0.42, "learning_rate": 1.288088294871204e-05, "loss": 0.5296, "step": 8507 }, { "epoch": 0.42, "learning_rate": 1.2879334264272762e-05, "loss": 0.4785, "step": 8508 }, { "epoch": 0.42, "learning_rate": 1.2877785504528268e-05, "loss": 0.4019, "step": 8509 }, { "epoch": 0.42, "learning_rate": 1.287623666951906e-05, "loss": 0.2269, "step": 8510 }, { "epoch": 0.42, "learning_rate": 1.2874687759285652e-05, "loss": 0.198, "step": 8511 }, { "epoch": 0.43, "learning_rate": 1.2873138773868545e-05, "loss": 0.2862, "step": 8512 }, { "epoch": 0.43, "learning_rate": 1.2871589713308263e-05, "loss": 0.447, "step": 8513 }, { "epoch": 0.43, "learning_rate": 1.2870040577645311e-05, "loss": 0.3243, "step": 8514 }, { "epoch": 0.43, "learning_rate": 1.2868491366920208e-05, "loss": 0.3276, "step": 8515 }, { "epoch": 0.43, "learning_rate": 1.2866942081173468e-05, "loss": 0.247, "step": 8516 }, { "epoch": 0.43, "learning_rate": 1.2865392720445615e-05, "loss": 0.4162, "step": 8517 }, { "epoch": 0.43, "learning_rate": 1.2863843284777167e-05, "loss": 0.4792, "step": 8518 }, { "epoch": 0.43, "learning_rate": 1.2862293774208649e-05, "loss": 0.3489, "step": 8519 }, { "epoch": 0.43, "learning_rate": 1.2860744188780587e-05, "loss": 0.2589, "step": 8520 }, { "epoch": 0.43, "learning_rate": 1.2859194528533505e-05, "loss": 0.2438, "step": 8521 }, { "epoch": 0.43, "learning_rate": 1.2857644793507937e-05, "loss": 0.1031, "step": 8522 }, { "epoch": 0.43, "learning_rate": 1.2856094983744411e-05, "loss": 0.476, "step": 8523 }, { "epoch": 0.43, "learning_rate": 1.2854545099283464e-05, "loss": 0.3639, "step": 8524 }, { "epoch": 0.43, "learning_rate": 1.285299514016563e-05, "loss": 0.5219, "step": 8525 }, { "epoch": 0.43, "learning_rate": 1.285144510643144e-05, "loss": 0.356, "step": 8526 }, { "epoch": 0.43, "learning_rate": 1.2849894998121442e-05, "loss": 0.3606, "step": 8527 }, { "epoch": 0.43, "learning_rate": 1.284834481527617e-05, "loss": 0.221, "step": 8528 }, { "epoch": 0.43, "learning_rate": 1.2846794557936173e-05, "loss": 0.3337, "step": 8529 }, { "epoch": 0.43, "learning_rate": 1.284524422614199e-05, "loss": 0.2741, "step": 8530 }, { "epoch": 0.43, "learning_rate": 1.2843693819934172e-05, "loss": 0.4266, "step": 8531 }, { "epoch": 0.43, "learning_rate": 1.2842143339353265e-05, "loss": 0.4263, "step": 8532 }, { "epoch": 0.43, "learning_rate": 1.2840592784439827e-05, "loss": 0.3036, "step": 8533 }, { "epoch": 0.43, "learning_rate": 1.2839042155234402e-05, "loss": 0.1816, "step": 8534 }, { "epoch": 0.43, "learning_rate": 1.2837491451777546e-05, "loss": 0.4315, "step": 8535 }, { "epoch": 0.43, "learning_rate": 1.2835940674109819e-05, "loss": 0.4464, "step": 8536 }, { "epoch": 0.43, "learning_rate": 1.2834389822271777e-05, "loss": 0.5816, "step": 8537 }, { "epoch": 0.43, "learning_rate": 1.2832838896303982e-05, "loss": 0.1012, "step": 8538 }, { "epoch": 0.43, "learning_rate": 1.2831287896247e-05, "loss": 0.1603, "step": 8539 }, { "epoch": 0.43, "learning_rate": 1.2829736822141387e-05, "loss": 0.4727, "step": 8540 }, { "epoch": 0.43, "learning_rate": 1.2828185674027717e-05, "loss": 0.4098, "step": 8541 }, { "epoch": 0.43, "learning_rate": 1.2826634451946552e-05, "loss": 0.2409, "step": 8542 }, { "epoch": 0.43, "learning_rate": 1.2825083155938466e-05, "loss": 0.4128, "step": 8543 }, { "epoch": 0.43, "learning_rate": 1.282353178604403e-05, "loss": 0.3282, "step": 8544 }, { "epoch": 0.43, "learning_rate": 1.2821980342303819e-05, "loss": 0.7772, "step": 8545 }, { "epoch": 0.43, "learning_rate": 1.2820428824758407e-05, "loss": 0.4379, "step": 8546 }, { "epoch": 0.43, "learning_rate": 1.2818877233448373e-05, "loss": 0.466, "step": 8547 }, { "epoch": 0.43, "learning_rate": 1.2817325568414299e-05, "loss": 0.2923, "step": 8548 }, { "epoch": 0.43, "learning_rate": 1.2815773829696763e-05, "loss": 0.3254, "step": 8549 }, { "epoch": 0.43, "learning_rate": 1.281422201733635e-05, "loss": 0.2849, "step": 8550 }, { "epoch": 0.43, "learning_rate": 1.281267013137365e-05, "loss": 0.4854, "step": 8551 }, { "epoch": 0.43, "learning_rate": 1.2811118171849244e-05, "loss": 0.4688, "step": 8552 }, { "epoch": 0.43, "learning_rate": 1.2809566138803723e-05, "loss": 0.4958, "step": 8553 }, { "epoch": 0.43, "learning_rate": 1.2808014032277683e-05, "loss": 0.3348, "step": 8554 }, { "epoch": 0.43, "learning_rate": 1.280646185231171e-05, "loss": 0.375, "step": 8555 }, { "epoch": 0.43, "learning_rate": 1.2804909598946404e-05, "loss": 0.3099, "step": 8556 }, { "epoch": 0.43, "learning_rate": 1.2803357272222363e-05, "loss": 0.5276, "step": 8557 }, { "epoch": 0.43, "learning_rate": 1.280180487218018e-05, "loss": 0.2195, "step": 8558 }, { "epoch": 0.43, "learning_rate": 1.2800252398860465e-05, "loss": 0.4729, "step": 8559 }, { "epoch": 0.43, "learning_rate": 1.2798699852303814e-05, "loss": 0.4097, "step": 8560 }, { "epoch": 0.43, "learning_rate": 1.2797147232550836e-05, "loss": 0.461, "step": 8561 }, { "epoch": 0.43, "learning_rate": 1.2795594539642131e-05, "loss": 0.5036, "step": 8562 }, { "epoch": 0.43, "learning_rate": 1.2794041773618317e-05, "loss": 0.2602, "step": 8563 }, { "epoch": 0.43, "learning_rate": 1.279248893452e-05, "loss": 0.2422, "step": 8564 }, { "epoch": 0.43, "learning_rate": 1.279093602238779e-05, "loss": 0.3903, "step": 8565 }, { "epoch": 0.43, "learning_rate": 1.2789383037262306e-05, "loss": 0.2852, "step": 8566 }, { "epoch": 0.43, "learning_rate": 1.278782997918416e-05, "loss": 0.422, "step": 8567 }, { "epoch": 0.43, "learning_rate": 1.2786276848193975e-05, "loss": 0.3195, "step": 8568 }, { "epoch": 0.43, "learning_rate": 1.2784723644332363e-05, "loss": 0.2273, "step": 8569 }, { "epoch": 0.43, "learning_rate": 1.2783170367639956e-05, "loss": 0.4919, "step": 8570 }, { "epoch": 0.43, "learning_rate": 1.278161701815737e-05, "loss": 0.4038, "step": 8571 }, { "epoch": 0.43, "learning_rate": 1.2780063595925238e-05, "loss": 0.2657, "step": 8572 }, { "epoch": 0.43, "learning_rate": 1.2778510100984185e-05, "loss": 0.6418, "step": 8573 }, { "epoch": 0.43, "learning_rate": 1.2776956533374834e-05, "loss": 0.2983, "step": 8574 }, { "epoch": 0.43, "learning_rate": 1.2775402893137827e-05, "loss": 0.4242, "step": 8575 }, { "epoch": 0.43, "learning_rate": 1.277384918031379e-05, "loss": 0.5186, "step": 8576 }, { "epoch": 0.43, "learning_rate": 1.2772295394943361e-05, "loss": 0.4153, "step": 8577 }, { "epoch": 0.43, "learning_rate": 1.2770741537067178e-05, "loss": 0.3029, "step": 8578 }, { "epoch": 0.43, "learning_rate": 1.2769187606725879e-05, "loss": 0.1502, "step": 8579 }, { "epoch": 0.43, "learning_rate": 1.2767633603960104e-05, "loss": 0.3446, "step": 8580 }, { "epoch": 0.43, "learning_rate": 1.2766079528810496e-05, "loss": 0.4079, "step": 8581 }, { "epoch": 0.43, "learning_rate": 1.2764525381317703e-05, "loss": 0.3267, "step": 8582 }, { "epoch": 0.43, "learning_rate": 1.2762971161522368e-05, "loss": 0.2659, "step": 8583 }, { "epoch": 0.43, "learning_rate": 1.2761416869465143e-05, "loss": 0.471, "step": 8584 }, { "epoch": 0.43, "learning_rate": 1.2759862505186673e-05, "loss": 0.2574, "step": 8585 }, { "epoch": 0.43, "learning_rate": 1.2758308068727618e-05, "loss": 0.3598, "step": 8586 }, { "epoch": 0.43, "learning_rate": 1.2756753560128627e-05, "loss": 0.2619, "step": 8587 }, { "epoch": 0.43, "learning_rate": 1.2755198979430357e-05, "loss": 0.4788, "step": 8588 }, { "epoch": 0.43, "learning_rate": 1.2753644326673467e-05, "loss": 0.3565, "step": 8589 }, { "epoch": 0.43, "learning_rate": 1.2752089601898617e-05, "loss": 0.9267, "step": 8590 }, { "epoch": 0.43, "learning_rate": 1.2750534805146466e-05, "loss": 0.3104, "step": 8591 }, { "epoch": 0.43, "learning_rate": 1.2748979936457677e-05, "loss": 0.5021, "step": 8592 }, { "epoch": 0.43, "learning_rate": 1.2747424995872923e-05, "loss": 0.3219, "step": 8593 }, { "epoch": 0.43, "learning_rate": 1.2745869983432866e-05, "loss": 0.5222, "step": 8594 }, { "epoch": 0.43, "learning_rate": 1.2744314899178173e-05, "loss": 0.3507, "step": 8595 }, { "epoch": 0.43, "learning_rate": 1.2742759743149522e-05, "loss": 0.4228, "step": 8596 }, { "epoch": 0.43, "learning_rate": 1.274120451538758e-05, "loss": 0.2998, "step": 8597 }, { "epoch": 0.43, "learning_rate": 1.2739649215933025e-05, "loss": 0.2468, "step": 8598 }, { "epoch": 0.43, "learning_rate": 1.2738093844826533e-05, "loss": 0.6032, "step": 8599 }, { "epoch": 0.43, "learning_rate": 1.2736538402108785e-05, "loss": 0.557, "step": 8600 }, { "epoch": 0.43, "learning_rate": 1.2734982887820457e-05, "loss": 0.5703, "step": 8601 }, { "epoch": 0.43, "learning_rate": 1.2733427302002235e-05, "loss": 0.3201, "step": 8602 }, { "epoch": 0.43, "learning_rate": 1.27318716446948e-05, "loss": 0.35, "step": 8603 }, { "epoch": 0.43, "learning_rate": 1.2730315915938842e-05, "loss": 0.4028, "step": 8604 }, { "epoch": 0.43, "learning_rate": 1.2728760115775045e-05, "loss": 0.4528, "step": 8605 }, { "epoch": 0.43, "learning_rate": 1.2727204244244102e-05, "loss": 0.3923, "step": 8606 }, { "epoch": 0.43, "learning_rate": 1.2725648301386706e-05, "loss": 0.3239, "step": 8607 }, { "epoch": 0.43, "learning_rate": 1.2724092287243545e-05, "loss": 0.5518, "step": 8608 }, { "epoch": 0.43, "learning_rate": 1.2722536201855323e-05, "loss": 0.5103, "step": 8609 }, { "epoch": 0.43, "learning_rate": 1.272098004526273e-05, "loss": 0.4061, "step": 8610 }, { "epoch": 0.43, "learning_rate": 1.2719423817506466e-05, "loss": 0.3509, "step": 8611 }, { "epoch": 0.43, "learning_rate": 1.2717867518627239e-05, "loss": 0.5593, "step": 8612 }, { "epoch": 0.43, "learning_rate": 1.2716311148665743e-05, "loss": 0.3299, "step": 8613 }, { "epoch": 0.43, "learning_rate": 1.2714754707662688e-05, "loss": 0.3013, "step": 8614 }, { "epoch": 0.43, "learning_rate": 1.2713198195658776e-05, "loss": 0.5549, "step": 8615 }, { "epoch": 0.43, "learning_rate": 1.2711641612694722e-05, "loss": 0.4991, "step": 8616 }, { "epoch": 0.43, "learning_rate": 1.271008495881123e-05, "loss": 0.2954, "step": 8617 }, { "epoch": 0.43, "learning_rate": 1.270852823404902e-05, "loss": 0.2583, "step": 8618 }, { "epoch": 0.43, "learning_rate": 1.2706971438448798e-05, "loss": 0.4482, "step": 8619 }, { "epoch": 0.43, "learning_rate": 1.2705414572051287e-05, "loss": 0.4767, "step": 8620 }, { "epoch": 0.43, "learning_rate": 1.2703857634897196e-05, "loss": 0.3446, "step": 8621 }, { "epoch": 0.43, "learning_rate": 1.2702300627027256e-05, "loss": 0.2978, "step": 8622 }, { "epoch": 0.43, "learning_rate": 1.2700743548482181e-05, "loss": 0.4415, "step": 8623 }, { "epoch": 0.43, "learning_rate": 1.2699186399302694e-05, "loss": 0.3712, "step": 8624 }, { "epoch": 0.43, "learning_rate": 1.2697629179529525e-05, "loss": 0.3602, "step": 8625 }, { "epoch": 0.43, "learning_rate": 1.2696071889203393e-05, "loss": 0.3284, "step": 8626 }, { "epoch": 0.43, "learning_rate": 1.2694514528365036e-05, "loss": 0.2794, "step": 8627 }, { "epoch": 0.43, "learning_rate": 1.2692957097055179e-05, "loss": 0.1797, "step": 8628 }, { "epoch": 0.43, "learning_rate": 1.2691399595314558e-05, "loss": 0.29, "step": 8629 }, { "epoch": 0.43, "learning_rate": 1.2689842023183905e-05, "loss": 0.2311, "step": 8630 }, { "epoch": 0.43, "learning_rate": 1.2688284380703955e-05, "loss": 0.3863, "step": 8631 }, { "epoch": 0.43, "learning_rate": 1.268672666791545e-05, "loss": 0.1628, "step": 8632 }, { "epoch": 0.43, "learning_rate": 1.2685168884859126e-05, "loss": 0.3019, "step": 8633 }, { "epoch": 0.43, "learning_rate": 1.2683611031575728e-05, "loss": 0.3209, "step": 8634 }, { "epoch": 0.43, "learning_rate": 1.2682053108105997e-05, "loss": 0.4645, "step": 8635 }, { "epoch": 0.43, "learning_rate": 1.2680495114490683e-05, "loss": 0.2527, "step": 8636 }, { "epoch": 0.43, "learning_rate": 1.2678937050770527e-05, "loss": 0.4185, "step": 8637 }, { "epoch": 0.43, "learning_rate": 1.2677378916986284e-05, "loss": 0.5223, "step": 8638 }, { "epoch": 0.43, "learning_rate": 1.2675820713178702e-05, "loss": 0.4419, "step": 8639 }, { "epoch": 0.43, "learning_rate": 1.2674262439388532e-05, "loss": 0.1752, "step": 8640 }, { "epoch": 0.43, "learning_rate": 1.2672704095656531e-05, "loss": 0.3594, "step": 8641 }, { "epoch": 0.43, "learning_rate": 1.2671145682023456e-05, "loss": 0.6862, "step": 8642 }, { "epoch": 0.43, "learning_rate": 1.2669587198530062e-05, "loss": 0.3413, "step": 8643 }, { "epoch": 0.43, "learning_rate": 1.2668028645217112e-05, "loss": 0.3934, "step": 8644 }, { "epoch": 0.43, "learning_rate": 1.2666470022125367e-05, "loss": 0.1796, "step": 8645 }, { "epoch": 0.43, "learning_rate": 1.2664911329295595e-05, "loss": 0.4055, "step": 8646 }, { "epoch": 0.43, "learning_rate": 1.2663352566768554e-05, "loss": 0.3178, "step": 8647 }, { "epoch": 0.43, "learning_rate": 1.2661793734585018e-05, "loss": 0.3297, "step": 8648 }, { "epoch": 0.43, "learning_rate": 1.2660234832785749e-05, "loss": 0.4773, "step": 8649 }, { "epoch": 0.43, "learning_rate": 1.2658675861411528e-05, "loss": 0.2436, "step": 8650 }, { "epoch": 0.43, "learning_rate": 1.2657116820503115e-05, "loss": 0.5477, "step": 8651 }, { "epoch": 0.43, "learning_rate": 1.2655557710101296e-05, "loss": 0.2678, "step": 8652 }, { "epoch": 0.43, "learning_rate": 1.265399853024684e-05, "loss": 0.4113, "step": 8653 }, { "epoch": 0.43, "learning_rate": 1.265243928098053e-05, "loss": 0.4184, "step": 8654 }, { "epoch": 0.43, "learning_rate": 1.2650879962343148e-05, "loss": 0.437, "step": 8655 }, { "epoch": 0.43, "learning_rate": 1.2649320574375468e-05, "loss": 0.5699, "step": 8656 }, { "epoch": 0.43, "learning_rate": 1.2647761117118283e-05, "loss": 0.3702, "step": 8657 }, { "epoch": 0.43, "learning_rate": 1.2646201590612368e-05, "loss": 0.4725, "step": 8658 }, { "epoch": 0.43, "learning_rate": 1.2644641994898521e-05, "loss": 0.1412, "step": 8659 }, { "epoch": 0.43, "learning_rate": 1.2643082330017526e-05, "loss": 0.56, "step": 8660 }, { "epoch": 0.43, "learning_rate": 1.2641522596010174e-05, "loss": 0.2413, "step": 8661 }, { "epoch": 0.43, "learning_rate": 1.263996279291726e-05, "loss": 0.2185, "step": 8662 }, { "epoch": 0.43, "learning_rate": 1.2638402920779574e-05, "loss": 0.4097, "step": 8663 }, { "epoch": 0.43, "learning_rate": 1.2636842979637916e-05, "loss": 0.3323, "step": 8664 }, { "epoch": 0.43, "learning_rate": 1.2635282969533083e-05, "loss": 0.3849, "step": 8665 }, { "epoch": 0.43, "learning_rate": 1.2633722890505876e-05, "loss": 0.326, "step": 8666 }, { "epoch": 0.43, "learning_rate": 1.2632162742597094e-05, "loss": 0.4914, "step": 8667 }, { "epoch": 0.43, "learning_rate": 1.2630602525847547e-05, "loss": 0.3884, "step": 8668 }, { "epoch": 0.43, "learning_rate": 1.2629042240298033e-05, "loss": 0.4242, "step": 8669 }, { "epoch": 0.43, "learning_rate": 1.2627481885989365e-05, "loss": 0.3465, "step": 8670 }, { "epoch": 0.43, "learning_rate": 1.2625921462962349e-05, "loss": 0.3937, "step": 8671 }, { "epoch": 0.43, "learning_rate": 1.2624360971257796e-05, "loss": 0.2656, "step": 8672 }, { "epoch": 0.43, "learning_rate": 1.262280041091652e-05, "loss": 0.293, "step": 8673 }, { "epoch": 0.43, "learning_rate": 1.2621239781979337e-05, "loss": 0.4077, "step": 8674 }, { "epoch": 0.43, "learning_rate": 1.2619679084487057e-05, "loss": 0.3168, "step": 8675 }, { "epoch": 0.43, "learning_rate": 1.2618118318480503e-05, "loss": 0.3663, "step": 8676 }, { "epoch": 0.43, "learning_rate": 1.2616557484000494e-05, "loss": 0.6097, "step": 8677 }, { "epoch": 0.43, "learning_rate": 1.261499658108785e-05, "loss": 0.3951, "step": 8678 }, { "epoch": 0.43, "learning_rate": 1.2613435609783395e-05, "loss": 0.4433, "step": 8679 }, { "epoch": 0.43, "learning_rate": 1.2611874570127956e-05, "loss": 0.3413, "step": 8680 }, { "epoch": 0.43, "learning_rate": 1.2610313462162356e-05, "loss": 0.2238, "step": 8681 }, { "epoch": 0.43, "learning_rate": 1.260875228592743e-05, "loss": 0.386, "step": 8682 }, { "epoch": 0.43, "learning_rate": 1.2607191041464e-05, "loss": 0.3849, "step": 8683 }, { "epoch": 0.43, "learning_rate": 1.2605629728812907e-05, "loss": 0.4371, "step": 8684 }, { "epoch": 0.43, "learning_rate": 1.2604068348014982e-05, "loss": 0.4165, "step": 8685 }, { "epoch": 0.43, "learning_rate": 1.2602506899111058e-05, "loss": 0.3924, "step": 8686 }, { "epoch": 0.43, "learning_rate": 1.2600945382141975e-05, "loss": 0.2524, "step": 8687 }, { "epoch": 0.43, "learning_rate": 1.2599383797148572e-05, "loss": 0.4394, "step": 8688 }, { "epoch": 0.43, "learning_rate": 1.2597822144171692e-05, "loss": 0.4426, "step": 8689 }, { "epoch": 0.43, "learning_rate": 1.2596260423252173e-05, "loss": 0.3986, "step": 8690 }, { "epoch": 0.43, "learning_rate": 1.2594698634430867e-05, "loss": 0.5312, "step": 8691 }, { "epoch": 0.43, "learning_rate": 1.2593136777748613e-05, "loss": 0.392, "step": 8692 }, { "epoch": 0.43, "learning_rate": 1.2591574853246265e-05, "loss": 0.2752, "step": 8693 }, { "epoch": 0.43, "learning_rate": 1.2590012860964672e-05, "loss": 0.3053, "step": 8694 }, { "epoch": 0.43, "learning_rate": 1.258845080094468e-05, "loss": 0.3617, "step": 8695 }, { "epoch": 0.43, "learning_rate": 1.2586888673227154e-05, "loss": 0.279, "step": 8696 }, { "epoch": 0.43, "learning_rate": 1.2585326477852942e-05, "loss": 0.3349, "step": 8697 }, { "epoch": 0.43, "learning_rate": 1.2583764214862898e-05, "loss": 0.4954, "step": 8698 }, { "epoch": 0.43, "learning_rate": 1.2582201884297888e-05, "loss": 0.317, "step": 8699 }, { "epoch": 0.43, "learning_rate": 1.258063948619877e-05, "loss": 0.2763, "step": 8700 }, { "epoch": 0.43, "learning_rate": 1.2579077020606402e-05, "loss": 0.2669, "step": 8701 }, { "epoch": 0.43, "learning_rate": 1.2577514487561659e-05, "loss": 0.5557, "step": 8702 }, { "epoch": 0.43, "learning_rate": 1.2575951887105397e-05, "loss": 0.3108, "step": 8703 }, { "epoch": 0.43, "learning_rate": 1.257438921927849e-05, "loss": 0.1677, "step": 8704 }, { "epoch": 0.43, "learning_rate": 1.25728264841218e-05, "loss": 0.5275, "step": 8705 }, { "epoch": 0.43, "learning_rate": 1.2571263681676205e-05, "loss": 0.2334, "step": 8706 }, { "epoch": 0.43, "learning_rate": 1.256970081198258e-05, "loss": 0.2393, "step": 8707 }, { "epoch": 0.43, "learning_rate": 1.256813787508179e-05, "loss": 0.3472, "step": 8708 }, { "epoch": 0.43, "learning_rate": 1.256657487101472e-05, "loss": 0.4855, "step": 8709 }, { "epoch": 0.43, "learning_rate": 1.2565011799822249e-05, "loss": 0.3958, "step": 8710 }, { "epoch": 0.43, "learning_rate": 1.256344866154525e-05, "loss": 0.5134, "step": 8711 }, { "epoch": 0.44, "learning_rate": 1.2561885456224612e-05, "loss": 0.1743, "step": 8712 }, { "epoch": 0.44, "learning_rate": 1.256032218390121e-05, "loss": 0.3498, "step": 8713 }, { "epoch": 0.44, "learning_rate": 1.255875884461594e-05, "loss": 0.4403, "step": 8714 }, { "epoch": 0.44, "learning_rate": 1.255719543840968e-05, "loss": 0.4936, "step": 8715 }, { "epoch": 0.44, "learning_rate": 1.2555631965323326e-05, "loss": 0.3657, "step": 8716 }, { "epoch": 0.44, "learning_rate": 1.2554068425397762e-05, "loss": 0.2773, "step": 8717 }, { "epoch": 0.44, "learning_rate": 1.2552504818673886e-05, "loss": 0.2432, "step": 8718 }, { "epoch": 0.44, "learning_rate": 1.255094114519259e-05, "loss": 0.4288, "step": 8719 }, { "epoch": 0.44, "learning_rate": 1.2549377404994768e-05, "loss": 0.5008, "step": 8720 }, { "epoch": 0.44, "learning_rate": 1.2547813598121319e-05, "loss": 0.1868, "step": 8721 }, { "epoch": 0.44, "learning_rate": 1.2546249724613145e-05, "loss": 0.4568, "step": 8722 }, { "epoch": 0.44, "learning_rate": 1.254468578451114e-05, "loss": 0.4511, "step": 8723 }, { "epoch": 0.44, "learning_rate": 1.2543121777856213e-05, "loss": 0.5655, "step": 8724 }, { "epoch": 0.44, "learning_rate": 1.2541557704689268e-05, "loss": 0.2385, "step": 8725 }, { "epoch": 0.44, "learning_rate": 1.253999356505121e-05, "loss": 0.2425, "step": 8726 }, { "epoch": 0.44, "learning_rate": 1.2538429358982947e-05, "loss": 0.3217, "step": 8727 }, { "epoch": 0.44, "learning_rate": 1.2536865086525389e-05, "loss": 0.2602, "step": 8728 }, { "epoch": 0.44, "learning_rate": 1.2535300747719444e-05, "loss": 0.3466, "step": 8729 }, { "epoch": 0.44, "learning_rate": 1.2533736342606033e-05, "loss": 0.446, "step": 8730 }, { "epoch": 0.44, "learning_rate": 1.2532171871226062e-05, "loss": 0.2128, "step": 8731 }, { "epoch": 0.44, "learning_rate": 1.2530607333620458e-05, "loss": 0.4499, "step": 8732 }, { "epoch": 0.44, "learning_rate": 1.252904272983013e-05, "loss": 0.344, "step": 8733 }, { "epoch": 0.44, "learning_rate": 1.2527478059896003e-05, "loss": 0.5266, "step": 8734 }, { "epoch": 0.44, "learning_rate": 1.2525913323858997e-05, "loss": 0.3332, "step": 8735 }, { "epoch": 0.44, "learning_rate": 1.2524348521760037e-05, "loss": 0.3514, "step": 8736 }, { "epoch": 0.44, "learning_rate": 1.252278365364005e-05, "loss": 0.3385, "step": 8737 }, { "epoch": 0.44, "learning_rate": 1.2521218719539956e-05, "loss": 0.6119, "step": 8738 }, { "epoch": 0.44, "learning_rate": 1.2519653719500691e-05, "loss": 0.2742, "step": 8739 }, { "epoch": 0.44, "learning_rate": 1.251808865356318e-05, "loss": 0.2013, "step": 8740 }, { "epoch": 0.44, "learning_rate": 1.2516523521768363e-05, "loss": 0.3993, "step": 8741 }, { "epoch": 0.44, "learning_rate": 1.2514958324157167e-05, "loss": 0.487, "step": 8742 }, { "epoch": 0.44, "learning_rate": 1.2513393060770532e-05, "loss": 0.5925, "step": 8743 }, { "epoch": 0.44, "learning_rate": 1.2511827731649392e-05, "loss": 0.51, "step": 8744 }, { "epoch": 0.44, "learning_rate": 1.2510262336834687e-05, "loss": 0.192, "step": 8745 }, { "epoch": 0.44, "learning_rate": 1.2508696876367358e-05, "loss": 0.4229, "step": 8746 }, { "epoch": 0.44, "learning_rate": 1.250713135028835e-05, "loss": 0.3202, "step": 8747 }, { "epoch": 0.44, "learning_rate": 1.2505565758638607e-05, "loss": 0.2683, "step": 8748 }, { "epoch": 0.44, "learning_rate": 1.2504000101459068e-05, "loss": 0.4999, "step": 8749 }, { "epoch": 0.44, "learning_rate": 1.2502434378790692e-05, "loss": 0.7432, "step": 8750 }, { "epoch": 0.44, "learning_rate": 1.2500868590674418e-05, "loss": 0.508, "step": 8751 }, { "epoch": 0.44, "learning_rate": 1.2499302737151201e-05, "loss": 0.3894, "step": 8752 }, { "epoch": 0.44, "learning_rate": 1.2497736818261998e-05, "loss": 0.3383, "step": 8753 }, { "epoch": 0.44, "learning_rate": 1.2496170834047756e-05, "loss": 0.5267, "step": 8754 }, { "epoch": 0.44, "learning_rate": 1.2494604784549437e-05, "loss": 0.3434, "step": 8755 }, { "epoch": 0.44, "learning_rate": 1.2493038669807997e-05, "loss": 0.2653, "step": 8756 }, { "epoch": 0.44, "learning_rate": 1.2491472489864397e-05, "loss": 0.3727, "step": 8757 }, { "epoch": 0.44, "learning_rate": 1.2489906244759598e-05, "loss": 0.2923, "step": 8758 }, { "epoch": 0.44, "learning_rate": 1.2488339934534561e-05, "loss": 0.2955, "step": 8759 }, { "epoch": 0.44, "learning_rate": 1.2486773559230254e-05, "loss": 0.4483, "step": 8760 }, { "epoch": 0.44, "learning_rate": 1.2485207118887638e-05, "loss": 0.3818, "step": 8761 }, { "epoch": 0.44, "learning_rate": 1.2483640613547686e-05, "loss": 0.3189, "step": 8762 }, { "epoch": 0.44, "learning_rate": 1.2482074043251368e-05, "loss": 0.3343, "step": 8763 }, { "epoch": 0.44, "learning_rate": 1.2480507408039655e-05, "loss": 0.4805, "step": 8764 }, { "epoch": 0.44, "learning_rate": 1.2478940707953515e-05, "loss": 0.6086, "step": 8765 }, { "epoch": 0.44, "learning_rate": 1.2477373943033933e-05, "loss": 0.5155, "step": 8766 }, { "epoch": 0.44, "learning_rate": 1.2475807113321878e-05, "loss": 0.2256, "step": 8767 }, { "epoch": 0.44, "learning_rate": 1.247424021885833e-05, "loss": 0.2434, "step": 8768 }, { "epoch": 0.44, "learning_rate": 1.2472673259684272e-05, "loss": 0.5, "step": 8769 }, { "epoch": 0.44, "learning_rate": 1.2471106235840683e-05, "loss": 0.3955, "step": 8770 }, { "epoch": 0.44, "learning_rate": 1.2469539147368547e-05, "loss": 0.3417, "step": 8771 }, { "epoch": 0.44, "learning_rate": 1.2467971994308847e-05, "loss": 0.2922, "step": 8772 }, { "epoch": 0.44, "learning_rate": 1.2466404776702574e-05, "loss": 0.31, "step": 8773 }, { "epoch": 0.44, "learning_rate": 1.2464837494590714e-05, "loss": 0.3376, "step": 8774 }, { "epoch": 0.44, "learning_rate": 1.2463270148014257e-05, "loss": 0.2901, "step": 8775 }, { "epoch": 0.44, "learning_rate": 1.2461702737014196e-05, "loss": 0.3866, "step": 8776 }, { "epoch": 0.44, "learning_rate": 1.2460135261631523e-05, "loss": 0.3715, "step": 8777 }, { "epoch": 0.44, "learning_rate": 1.2458567721907237e-05, "loss": 0.4263, "step": 8778 }, { "epoch": 0.44, "learning_rate": 1.245700011788233e-05, "loss": 0.4703, "step": 8779 }, { "epoch": 0.44, "learning_rate": 1.2455432449597804e-05, "loss": 0.2145, "step": 8780 }, { "epoch": 0.44, "learning_rate": 1.2453864717094657e-05, "loss": 0.5635, "step": 8781 }, { "epoch": 0.44, "learning_rate": 1.2452296920413893e-05, "loss": 0.1957, "step": 8782 }, { "epoch": 0.44, "learning_rate": 1.2450729059596518e-05, "loss": 0.4227, "step": 8783 }, { "epoch": 0.44, "learning_rate": 1.2449161134683532e-05, "loss": 0.3751, "step": 8784 }, { "epoch": 0.44, "learning_rate": 1.2447593145715945e-05, "loss": 0.3876, "step": 8785 }, { "epoch": 0.44, "learning_rate": 1.2446025092734764e-05, "loss": 0.4184, "step": 8786 }, { "epoch": 0.44, "learning_rate": 1.2444456975781e-05, "loss": 0.3968, "step": 8787 }, { "epoch": 0.44, "learning_rate": 1.2442888794895667e-05, "loss": 0.2469, "step": 8788 }, { "epoch": 0.44, "learning_rate": 1.2441320550119775e-05, "loss": 0.1842, "step": 8789 }, { "epoch": 0.44, "learning_rate": 1.2439752241494343e-05, "loss": 0.221, "step": 8790 }, { "epoch": 0.44, "learning_rate": 1.2438183869060387e-05, "loss": 0.4741, "step": 8791 }, { "epoch": 0.44, "learning_rate": 1.2436615432858925e-05, "loss": 0.212, "step": 8792 }, { "epoch": 0.44, "learning_rate": 1.2435046932930976e-05, "loss": 0.3657, "step": 8793 }, { "epoch": 0.44, "learning_rate": 1.2433478369317566e-05, "loss": 0.5958, "step": 8794 }, { "epoch": 0.44, "learning_rate": 1.2431909742059719e-05, "loss": 0.2463, "step": 8795 }, { "epoch": 0.44, "learning_rate": 1.2430341051198453e-05, "loss": 0.1612, "step": 8796 }, { "epoch": 0.44, "learning_rate": 1.2428772296774802e-05, "loss": 0.5534, "step": 8797 }, { "epoch": 0.44, "learning_rate": 1.2427203478829792e-05, "loss": 0.3947, "step": 8798 }, { "epoch": 0.44, "learning_rate": 1.2425634597404457e-05, "loss": 0.3121, "step": 8799 }, { "epoch": 0.44, "learning_rate": 1.2424065652539823e-05, "loss": 0.3998, "step": 8800 }, { "epoch": 0.44, "learning_rate": 1.2422496644276927e-05, "loss": 0.2989, "step": 8801 }, { "epoch": 0.44, "learning_rate": 1.2420927572656804e-05, "loss": 0.426, "step": 8802 }, { "epoch": 0.44, "learning_rate": 1.2419358437720493e-05, "loss": 0.1281, "step": 8803 }, { "epoch": 0.44, "learning_rate": 1.2417789239509028e-05, "loss": 0.3403, "step": 8804 }, { "epoch": 0.44, "learning_rate": 1.2416219978063456e-05, "loss": 0.4867, "step": 8805 }, { "epoch": 0.44, "learning_rate": 1.2414650653424814e-05, "loss": 0.5174, "step": 8806 }, { "epoch": 0.44, "learning_rate": 1.2413081265634146e-05, "loss": 0.858, "step": 8807 }, { "epoch": 0.44, "learning_rate": 1.2411511814732496e-05, "loss": 0.4695, "step": 8808 }, { "epoch": 0.44, "learning_rate": 1.2409942300760916e-05, "loss": 0.4194, "step": 8809 }, { "epoch": 0.44, "learning_rate": 1.240837272376045e-05, "loss": 0.2158, "step": 8810 }, { "epoch": 0.44, "learning_rate": 1.2406803083772149e-05, "loss": 0.4205, "step": 8811 }, { "epoch": 0.44, "learning_rate": 1.240523338083707e-05, "loss": 0.3793, "step": 8812 }, { "epoch": 0.44, "learning_rate": 1.2403663614996257e-05, "loss": 0.4641, "step": 8813 }, { "epoch": 0.44, "learning_rate": 1.2402093786290772e-05, "loss": 0.2356, "step": 8814 }, { "epoch": 0.44, "learning_rate": 1.2400523894761674e-05, "loss": 0.683, "step": 8815 }, { "epoch": 0.44, "learning_rate": 1.2398953940450011e-05, "loss": 0.3827, "step": 8816 }, { "epoch": 0.44, "learning_rate": 1.2397383923396855e-05, "loss": 0.3262, "step": 8817 }, { "epoch": 0.44, "learning_rate": 1.2395813843643258e-05, "loss": 0.3628, "step": 8818 }, { "epoch": 0.44, "learning_rate": 1.2394243701230293e-05, "loss": 0.1315, "step": 8819 }, { "epoch": 0.44, "learning_rate": 1.239267349619902e-05, "loss": 0.2714, "step": 8820 }, { "epoch": 0.44, "learning_rate": 1.2391103228590501e-05, "loss": 0.461, "step": 8821 }, { "epoch": 0.44, "learning_rate": 1.238953289844581e-05, "loss": 0.1175, "step": 8822 }, { "epoch": 0.44, "learning_rate": 1.2387962505806018e-05, "loss": 0.3939, "step": 8823 }, { "epoch": 0.44, "learning_rate": 1.2386392050712193e-05, "loss": 0.2626, "step": 8824 }, { "epoch": 0.44, "learning_rate": 1.2384821533205409e-05, "loss": 0.3616, "step": 8825 }, { "epoch": 0.44, "learning_rate": 1.2383250953326741e-05, "loss": 0.4673, "step": 8826 }, { "epoch": 0.44, "learning_rate": 1.2381680311117266e-05, "loss": 0.4871, "step": 8827 }, { "epoch": 0.44, "learning_rate": 1.2380109606618064e-05, "loss": 0.311, "step": 8828 }, { "epoch": 0.44, "learning_rate": 1.2378538839870209e-05, "loss": 0.2593, "step": 8829 }, { "epoch": 0.44, "learning_rate": 1.237696801091479e-05, "loss": 0.2705, "step": 8830 }, { "epoch": 0.44, "learning_rate": 1.2375397119792885e-05, "loss": 0.2832, "step": 8831 }, { "epoch": 0.44, "learning_rate": 1.237382616654558e-05, "loss": 0.4009, "step": 8832 }, { "epoch": 0.44, "learning_rate": 1.2372255151213959e-05, "loss": 0.2537, "step": 8833 }, { "epoch": 0.44, "learning_rate": 1.2370684073839109e-05, "loss": 0.5829, "step": 8834 }, { "epoch": 0.44, "learning_rate": 1.2369112934462129e-05, "loss": 0.3927, "step": 8835 }, { "epoch": 0.44, "learning_rate": 1.2367541733124097e-05, "loss": 0.4, "step": 8836 }, { "epoch": 0.44, "learning_rate": 1.2365970469866116e-05, "loss": 0.3577, "step": 8837 }, { "epoch": 0.44, "learning_rate": 1.236439914472927e-05, "loss": 0.1992, "step": 8838 }, { "epoch": 0.44, "learning_rate": 1.2362827757754667e-05, "loss": 0.6379, "step": 8839 }, { "epoch": 0.44, "learning_rate": 1.23612563089834e-05, "loss": 0.2445, "step": 8840 }, { "epoch": 0.44, "learning_rate": 1.2359684798456562e-05, "loss": 0.385, "step": 8841 }, { "epoch": 0.44, "learning_rate": 1.2358113226215263e-05, "loss": 0.3833, "step": 8842 }, { "epoch": 0.44, "learning_rate": 1.2356541592300599e-05, "loss": 0.2313, "step": 8843 }, { "epoch": 0.44, "learning_rate": 1.2354969896753674e-05, "loss": 0.5595, "step": 8844 }, { "epoch": 0.44, "learning_rate": 1.23533981396156e-05, "loss": 0.28, "step": 8845 }, { "epoch": 0.44, "learning_rate": 1.2351826320927477e-05, "loss": 0.585, "step": 8846 }, { "epoch": 0.44, "learning_rate": 1.2350254440730421e-05, "loss": 0.4067, "step": 8847 }, { "epoch": 0.44, "learning_rate": 1.2348682499065534e-05, "loss": 0.4034, "step": 8848 }, { "epoch": 0.44, "learning_rate": 1.2347110495973935e-05, "loss": 0.2769, "step": 8849 }, { "epoch": 0.44, "learning_rate": 1.2345538431496733e-05, "loss": 0.3652, "step": 8850 }, { "epoch": 0.44, "learning_rate": 1.2343966305675045e-05, "loss": 0.3302, "step": 8851 }, { "epoch": 0.44, "learning_rate": 1.234239411854999e-05, "loss": 0.2974, "step": 8852 }, { "epoch": 0.44, "learning_rate": 1.2340821870162684e-05, "loss": 0.3251, "step": 8853 }, { "epoch": 0.44, "learning_rate": 1.2339249560554249e-05, "loss": 0.5622, "step": 8854 }, { "epoch": 0.44, "learning_rate": 1.2337677189765807e-05, "loss": 0.325, "step": 8855 }, { "epoch": 0.44, "learning_rate": 1.2336104757838478e-05, "loss": 0.5451, "step": 8856 }, { "epoch": 0.44, "learning_rate": 1.233453226481339e-05, "loss": 0.2011, "step": 8857 }, { "epoch": 0.44, "learning_rate": 1.2332959710731666e-05, "loss": 0.1893, "step": 8858 }, { "epoch": 0.44, "learning_rate": 1.2331387095634438e-05, "loss": 0.4927, "step": 8859 }, { "epoch": 0.44, "learning_rate": 1.2329814419562834e-05, "loss": 0.217, "step": 8860 }, { "epoch": 0.44, "learning_rate": 1.2328241682557984e-05, "loss": 0.3063, "step": 8861 }, { "epoch": 0.44, "learning_rate": 1.2326668884661023e-05, "loss": 0.2665, "step": 8862 }, { "epoch": 0.44, "learning_rate": 1.2325096025913083e-05, "loss": 0.3694, "step": 8863 }, { "epoch": 0.44, "learning_rate": 1.2323523106355306e-05, "loss": 0.5317, "step": 8864 }, { "epoch": 0.44, "learning_rate": 1.2321950126028822e-05, "loss": 0.5733, "step": 8865 }, { "epoch": 0.44, "learning_rate": 1.2320377084974776e-05, "loss": 0.2993, "step": 8866 }, { "epoch": 0.44, "learning_rate": 1.2318803983234305e-05, "loss": 0.2936, "step": 8867 }, { "epoch": 0.44, "learning_rate": 1.2317230820848553e-05, "loss": 0.3265, "step": 8868 }, { "epoch": 0.44, "learning_rate": 1.2315657597858665e-05, "loss": 0.3555, "step": 8869 }, { "epoch": 0.44, "learning_rate": 1.2314084314305786e-05, "loss": 0.2361, "step": 8870 }, { "epoch": 0.44, "learning_rate": 1.231251097023106e-05, "loss": 0.5015, "step": 8871 }, { "epoch": 0.44, "learning_rate": 1.231093756567564e-05, "loss": 0.5082, "step": 8872 }, { "epoch": 0.44, "learning_rate": 1.2309364100680675e-05, "loss": 0.6769, "step": 8873 }, { "epoch": 0.44, "learning_rate": 1.2307790575287316e-05, "loss": 0.5679, "step": 8874 }, { "epoch": 0.44, "learning_rate": 1.2306216989536716e-05, "loss": 0.3407, "step": 8875 }, { "epoch": 0.44, "learning_rate": 1.2304643343470032e-05, "loss": 0.2973, "step": 8876 }, { "epoch": 0.44, "learning_rate": 1.230306963712842e-05, "loss": 0.2593, "step": 8877 }, { "epoch": 0.44, "learning_rate": 1.230149587055304e-05, "loss": 0.344, "step": 8878 }, { "epoch": 0.44, "learning_rate": 1.2299922043785048e-05, "loss": 0.3877, "step": 8879 }, { "epoch": 0.44, "learning_rate": 1.2298348156865608e-05, "loss": 0.4326, "step": 8880 }, { "epoch": 0.44, "learning_rate": 1.2296774209835881e-05, "loss": 0.3369, "step": 8881 }, { "epoch": 0.44, "learning_rate": 1.2295200202737037e-05, "loss": 0.4084, "step": 8882 }, { "epoch": 0.44, "learning_rate": 1.2293626135610236e-05, "loss": 0.3683, "step": 8883 }, { "epoch": 0.44, "learning_rate": 1.2292052008496643e-05, "loss": 0.4439, "step": 8884 }, { "epoch": 0.44, "learning_rate": 1.2290477821437437e-05, "loss": 0.4335, "step": 8885 }, { "epoch": 0.44, "learning_rate": 1.2288903574473778e-05, "loss": 0.4878, "step": 8886 }, { "epoch": 0.44, "learning_rate": 1.228732926764685e-05, "loss": 0.2893, "step": 8887 }, { "epoch": 0.44, "learning_rate": 1.2285754900997819e-05, "loss": 0.5031, "step": 8888 }, { "epoch": 0.44, "learning_rate": 1.228418047456786e-05, "loss": 0.3686, "step": 8889 }, { "epoch": 0.44, "learning_rate": 1.2282605988398156e-05, "loss": 0.2653, "step": 8890 }, { "epoch": 0.44, "learning_rate": 1.2281031442529878e-05, "loss": 0.2806, "step": 8891 }, { "epoch": 0.44, "learning_rate": 1.2279456837004214e-05, "loss": 0.4618, "step": 8892 }, { "epoch": 0.44, "learning_rate": 1.2277882171862338e-05, "loss": 0.3972, "step": 8893 }, { "epoch": 0.44, "learning_rate": 1.227630744714544e-05, "loss": 0.3885, "step": 8894 }, { "epoch": 0.44, "learning_rate": 1.2274732662894702e-05, "loss": 0.6523, "step": 8895 }, { "epoch": 0.44, "learning_rate": 1.2273157819151311e-05, "loss": 0.312, "step": 8896 }, { "epoch": 0.44, "learning_rate": 1.2271582915956456e-05, "loss": 0.295, "step": 8897 }, { "epoch": 0.44, "learning_rate": 1.2270007953351322e-05, "loss": 0.4008, "step": 8898 }, { "epoch": 0.44, "learning_rate": 1.2268432931377104e-05, "loss": 0.2917, "step": 8899 }, { "epoch": 0.44, "learning_rate": 1.2266857850074993e-05, "loss": 0.4867, "step": 8900 }, { "epoch": 0.44, "learning_rate": 1.2265282709486186e-05, "loss": 0.3184, "step": 8901 }, { "epoch": 0.44, "learning_rate": 1.2263707509651874e-05, "loss": 0.2757, "step": 8902 }, { "epoch": 0.44, "learning_rate": 1.226213225061326e-05, "loss": 0.4993, "step": 8903 }, { "epoch": 0.44, "learning_rate": 1.2260556932411538e-05, "loss": 0.4933, "step": 8904 }, { "epoch": 0.44, "learning_rate": 1.2258981555087913e-05, "loss": 0.2336, "step": 8905 }, { "epoch": 0.44, "learning_rate": 1.2257406118683582e-05, "loss": 0.3474, "step": 8906 }, { "epoch": 0.44, "learning_rate": 1.2255830623239749e-05, "loss": 0.4701, "step": 8907 }, { "epoch": 0.44, "learning_rate": 1.2254255068797623e-05, "loss": 0.4573, "step": 8908 }, { "epoch": 0.44, "learning_rate": 1.2252679455398405e-05, "loss": 0.6155, "step": 8909 }, { "epoch": 0.44, "learning_rate": 1.2251103783083312e-05, "loss": 0.279, "step": 8910 }, { "epoch": 0.44, "learning_rate": 1.2249528051893541e-05, "loss": 0.5183, "step": 8911 }, { "epoch": 0.44, "learning_rate": 1.2247952261870314e-05, "loss": 0.4714, "step": 8912 }, { "epoch": 0.45, "learning_rate": 1.2246376413054842e-05, "loss": 0.1706, "step": 8913 }, { "epoch": 0.45, "learning_rate": 1.2244800505488334e-05, "loss": 0.3797, "step": 8914 }, { "epoch": 0.45, "learning_rate": 1.2243224539212012e-05, "loss": 0.2994, "step": 8915 }, { "epoch": 0.45, "learning_rate": 1.2241648514267083e-05, "loss": 0.3341, "step": 8916 }, { "epoch": 0.45, "learning_rate": 1.224007243069478e-05, "loss": 0.2733, "step": 8917 }, { "epoch": 0.45, "learning_rate": 1.2238496288536317e-05, "loss": 0.485, "step": 8918 }, { "epoch": 0.45, "learning_rate": 1.2236920087832913e-05, "loss": 0.4855, "step": 8919 }, { "epoch": 0.45, "learning_rate": 1.2235343828625797e-05, "loss": 0.2541, "step": 8920 }, { "epoch": 0.45, "learning_rate": 1.2233767510956186e-05, "loss": 0.142, "step": 8921 }, { "epoch": 0.45, "learning_rate": 1.2232191134865314e-05, "loss": 0.2775, "step": 8922 }, { "epoch": 0.45, "learning_rate": 1.2230614700394403e-05, "loss": 0.1659, "step": 8923 }, { "epoch": 0.45, "learning_rate": 1.2229038207584692e-05, "loss": 0.5336, "step": 8924 }, { "epoch": 0.45, "learning_rate": 1.2227461656477401e-05, "loss": 0.4911, "step": 8925 }, { "epoch": 0.45, "learning_rate": 1.2225885047113773e-05, "loss": 0.3699, "step": 8926 }, { "epoch": 0.45, "learning_rate": 1.2224308379535031e-05, "loss": 0.3109, "step": 8927 }, { "epoch": 0.45, "learning_rate": 1.222273165378242e-05, "loss": 0.407, "step": 8928 }, { "epoch": 0.45, "learning_rate": 1.2221154869897175e-05, "loss": 0.4759, "step": 8929 }, { "epoch": 0.45, "learning_rate": 1.2219578027920534e-05, "loss": 0.3973, "step": 8930 }, { "epoch": 0.45, "learning_rate": 1.2218001127893735e-05, "loss": 0.404, "step": 8931 }, { "epoch": 0.45, "learning_rate": 1.2216424169858022e-05, "loss": 0.5292, "step": 8932 }, { "epoch": 0.45, "learning_rate": 1.2214847153854637e-05, "loss": 0.4003, "step": 8933 }, { "epoch": 0.45, "learning_rate": 1.2213270079924824e-05, "loss": 0.2622, "step": 8934 }, { "epoch": 0.45, "learning_rate": 1.2211692948109834e-05, "loss": 0.3265, "step": 8935 }, { "epoch": 0.45, "learning_rate": 1.221011575845091e-05, "loss": 0.2439, "step": 8936 }, { "epoch": 0.45, "learning_rate": 1.2208538510989306e-05, "loss": 0.4749, "step": 8937 }, { "epoch": 0.45, "learning_rate": 1.2206961205766268e-05, "loss": 0.3259, "step": 8938 }, { "epoch": 0.45, "learning_rate": 1.2205383842823046e-05, "loss": 0.1876, "step": 8939 }, { "epoch": 0.45, "learning_rate": 1.2203806422200906e-05, "loss": 0.5364, "step": 8940 }, { "epoch": 0.45, "learning_rate": 1.220222894394109e-05, "loss": 0.4891, "step": 8941 }, { "epoch": 0.45, "learning_rate": 1.2200651408084862e-05, "loss": 0.5763, "step": 8942 }, { "epoch": 0.45, "learning_rate": 1.219907381467348e-05, "loss": 0.3035, "step": 8943 }, { "epoch": 0.45, "learning_rate": 1.2197496163748201e-05, "loss": 0.397, "step": 8944 }, { "epoch": 0.45, "learning_rate": 1.2195918455350288e-05, "loss": 0.4419, "step": 8945 }, { "epoch": 0.45, "learning_rate": 1.2194340689521005e-05, "loss": 0.3416, "step": 8946 }, { "epoch": 0.45, "learning_rate": 1.2192762866301616e-05, "loss": 0.3567, "step": 8947 }, { "epoch": 0.45, "learning_rate": 1.2191184985733382e-05, "loss": 0.5282, "step": 8948 }, { "epoch": 0.45, "learning_rate": 1.2189607047857578e-05, "loss": 0.4339, "step": 8949 }, { "epoch": 0.45, "learning_rate": 1.2188029052715468e-05, "loss": 0.4288, "step": 8950 }, { "epoch": 0.45, "learning_rate": 1.2186451000348327e-05, "loss": 0.2462, "step": 8951 }, { "epoch": 0.45, "learning_rate": 1.2184872890797421e-05, "loss": 0.2499, "step": 8952 }, { "epoch": 0.45, "learning_rate": 1.2183294724104027e-05, "loss": 0.4951, "step": 8953 }, { "epoch": 0.45, "learning_rate": 1.218171650030942e-05, "loss": 0.3404, "step": 8954 }, { "epoch": 0.45, "learning_rate": 1.2180138219454876e-05, "loss": 0.3027, "step": 8955 }, { "epoch": 0.45, "learning_rate": 1.2178559881581671e-05, "loss": 0.2647, "step": 8956 }, { "epoch": 0.45, "learning_rate": 1.2176981486731083e-05, "loss": 0.3339, "step": 8957 }, { "epoch": 0.45, "learning_rate": 1.21754030349444e-05, "loss": 0.2837, "step": 8958 }, { "epoch": 0.45, "learning_rate": 1.2173824526262896e-05, "loss": 0.0882, "step": 8959 }, { "epoch": 0.45, "learning_rate": 1.2172245960727862e-05, "loss": 0.237, "step": 8960 }, { "epoch": 0.45, "learning_rate": 1.2170667338380579e-05, "loss": 0.4697, "step": 8961 }, { "epoch": 0.45, "learning_rate": 1.2169088659262333e-05, "loss": 0.3834, "step": 8962 }, { "epoch": 0.45, "learning_rate": 1.2167509923414417e-05, "loss": 0.2906, "step": 8963 }, { "epoch": 0.45, "learning_rate": 1.2165931130878116e-05, "loss": 0.4171, "step": 8964 }, { "epoch": 0.45, "learning_rate": 1.2164352281694725e-05, "loss": 0.382, "step": 8965 }, { "epoch": 0.45, "learning_rate": 1.2162773375905534e-05, "loss": 0.2887, "step": 8966 }, { "epoch": 0.45, "learning_rate": 1.2161194413551837e-05, "loss": 0.4085, "step": 8967 }, { "epoch": 0.45, "learning_rate": 1.2159615394674935e-05, "loss": 0.238, "step": 8968 }, { "epoch": 0.45, "learning_rate": 1.2158036319316115e-05, "loss": 0.5111, "step": 8969 }, { "epoch": 0.45, "learning_rate": 1.2156457187516686e-05, "loss": 0.6298, "step": 8970 }, { "epoch": 0.45, "learning_rate": 1.215487799931794e-05, "loss": 0.4081, "step": 8971 }, { "epoch": 0.45, "learning_rate": 1.2153298754761185e-05, "loss": 0.3889, "step": 8972 }, { "epoch": 0.45, "learning_rate": 1.2151719453887719e-05, "loss": 0.2673, "step": 8973 }, { "epoch": 0.45, "learning_rate": 1.2150140096738849e-05, "loss": 0.4442, "step": 8974 }, { "epoch": 0.45, "learning_rate": 1.2148560683355883e-05, "loss": 0.4069, "step": 8975 }, { "epoch": 0.45, "learning_rate": 1.2146981213780127e-05, "loss": 0.2423, "step": 8976 }, { "epoch": 0.45, "learning_rate": 1.214540168805289e-05, "loss": 0.4428, "step": 8977 }, { "epoch": 0.45, "learning_rate": 1.214382210621548e-05, "loss": 0.4751, "step": 8978 }, { "epoch": 0.45, "learning_rate": 1.2142242468309211e-05, "loss": 0.3673, "step": 8979 }, { "epoch": 0.45, "learning_rate": 1.2140662774375394e-05, "loss": 0.2848, "step": 8980 }, { "epoch": 0.45, "learning_rate": 1.2139083024455348e-05, "loss": 0.2998, "step": 8981 }, { "epoch": 0.45, "learning_rate": 1.2137503218590384e-05, "loss": 0.3113, "step": 8982 }, { "epoch": 0.45, "learning_rate": 1.2135923356821828e-05, "loss": 0.6275, "step": 8983 }, { "epoch": 0.45, "learning_rate": 1.213434343919099e-05, "loss": 0.2237, "step": 8984 }, { "epoch": 0.45, "learning_rate": 1.21327634657392e-05, "loss": 0.539, "step": 8985 }, { "epoch": 0.45, "learning_rate": 1.213118343650777e-05, "loss": 0.2543, "step": 8986 }, { "epoch": 0.45, "learning_rate": 1.212960335153803e-05, "loss": 0.2362, "step": 8987 }, { "epoch": 0.45, "learning_rate": 1.2128023210871305e-05, "loss": 0.2418, "step": 8988 }, { "epoch": 0.45, "learning_rate": 1.2126443014548916e-05, "loss": 0.6072, "step": 8989 }, { "epoch": 0.45, "learning_rate": 1.21248627626122e-05, "loss": 0.3124, "step": 8990 }, { "epoch": 0.45, "learning_rate": 1.212328245510248e-05, "loss": 0.391, "step": 8991 }, { "epoch": 0.45, "learning_rate": 1.2121702092061089e-05, "loss": 0.4885, "step": 8992 }, { "epoch": 0.45, "learning_rate": 1.212012167352936e-05, "loss": 0.3421, "step": 8993 }, { "epoch": 0.45, "learning_rate": 1.2118541199548621e-05, "loss": 0.3727, "step": 8994 }, { "epoch": 0.45, "learning_rate": 1.2116960670160215e-05, "loss": 0.3903, "step": 8995 }, { "epoch": 0.45, "learning_rate": 1.2115380085405473e-05, "loss": 0.2165, "step": 8996 }, { "epoch": 0.45, "learning_rate": 1.2113799445325739e-05, "loss": 0.244, "step": 8997 }, { "epoch": 0.45, "learning_rate": 1.2112218749962343e-05, "loss": 0.3673, "step": 8998 }, { "epoch": 0.45, "learning_rate": 1.2110637999356637e-05, "loss": 0.2153, "step": 8999 }, { "epoch": 0.45, "learning_rate": 1.2109057193549957e-05, "loss": 0.3984, "step": 9000 }, { "epoch": 0.45, "learning_rate": 1.210747633258365e-05, "loss": 0.2151, "step": 9001 }, { "epoch": 0.45, "learning_rate": 1.210589541649906e-05, "loss": 0.288, "step": 9002 }, { "epoch": 0.45, "learning_rate": 1.2104314445337533e-05, "loss": 0.4626, "step": 9003 }, { "epoch": 0.45, "learning_rate": 1.2102733419140417e-05, "loss": 0.348, "step": 9004 }, { "epoch": 0.45, "learning_rate": 1.2101152337949062e-05, "loss": 0.5001, "step": 9005 }, { "epoch": 0.45, "learning_rate": 1.2099571201804823e-05, "loss": 0.3444, "step": 9006 }, { "epoch": 0.45, "learning_rate": 1.2097990010749046e-05, "loss": 0.3919, "step": 9007 }, { "epoch": 0.45, "learning_rate": 1.2096408764823089e-05, "loss": 0.2545, "step": 9008 }, { "epoch": 0.45, "learning_rate": 1.2094827464068307e-05, "loss": 0.3156, "step": 9009 }, { "epoch": 0.45, "learning_rate": 1.2093246108526053e-05, "loss": 0.4597, "step": 9010 }, { "epoch": 0.45, "learning_rate": 1.2091664698237693e-05, "loss": 0.3242, "step": 9011 }, { "epoch": 0.45, "learning_rate": 1.2090083233244578e-05, "loss": 0.3876, "step": 9012 }, { "epoch": 0.45, "learning_rate": 1.2088501713588079e-05, "loss": 0.3491, "step": 9013 }, { "epoch": 0.45, "learning_rate": 1.208692013930955e-05, "loss": 0.6567, "step": 9014 }, { "epoch": 0.45, "learning_rate": 1.208533851045036e-05, "loss": 0.4868, "step": 9015 }, { "epoch": 0.45, "learning_rate": 1.2083756827051871e-05, "loss": 0.4653, "step": 9016 }, { "epoch": 0.45, "learning_rate": 1.2082175089155452e-05, "loss": 0.3821, "step": 9017 }, { "epoch": 0.45, "learning_rate": 1.2080593296802472e-05, "loss": 0.3603, "step": 9018 }, { "epoch": 0.45, "learning_rate": 1.2079011450034294e-05, "loss": 0.3843, "step": 9019 }, { "epoch": 0.45, "learning_rate": 1.2077429548892302e-05, "loss": 0.365, "step": 9020 }, { "epoch": 0.45, "learning_rate": 1.2075847593417855e-05, "loss": 0.2971, "step": 9021 }, { "epoch": 0.45, "learning_rate": 1.2074265583652336e-05, "loss": 0.286, "step": 9022 }, { "epoch": 0.45, "learning_rate": 1.2072683519637116e-05, "loss": 0.3872, "step": 9023 }, { "epoch": 0.45, "learning_rate": 1.2071101401413574e-05, "loss": 0.3307, "step": 9024 }, { "epoch": 0.45, "learning_rate": 1.206951922902309e-05, "loss": 0.3844, "step": 9025 }, { "epoch": 0.45, "learning_rate": 1.2067937002507038e-05, "loss": 0.4669, "step": 9026 }, { "epoch": 0.45, "learning_rate": 1.2066354721906804e-05, "loss": 0.2364, "step": 9027 }, { "epoch": 0.45, "learning_rate": 1.2064772387263769e-05, "loss": 0.5556, "step": 9028 }, { "epoch": 0.45, "learning_rate": 1.2063189998619314e-05, "loss": 0.2728, "step": 9029 }, { "epoch": 0.45, "learning_rate": 1.2061607556014828e-05, "loss": 0.1241, "step": 9030 }, { "epoch": 0.45, "learning_rate": 1.2060025059491696e-05, "loss": 0.3826, "step": 9031 }, { "epoch": 0.45, "learning_rate": 1.2058442509091305e-05, "loss": 0.5172, "step": 9032 }, { "epoch": 0.45, "learning_rate": 1.2056859904855051e-05, "loss": 0.3666, "step": 9033 }, { "epoch": 0.45, "learning_rate": 1.205527724682432e-05, "loss": 0.4086, "step": 9034 }, { "epoch": 0.45, "learning_rate": 1.20536945350405e-05, "loss": 0.3531, "step": 9035 }, { "epoch": 0.45, "learning_rate": 1.2052111769544992e-05, "loss": 0.1757, "step": 9036 }, { "epoch": 0.45, "learning_rate": 1.2050528950379187e-05, "loss": 0.398, "step": 9037 }, { "epoch": 0.45, "learning_rate": 1.2048946077584484e-05, "loss": 0.4735, "step": 9038 }, { "epoch": 0.45, "learning_rate": 1.2047363151202282e-05, "loss": 0.3805, "step": 9039 }, { "epoch": 0.45, "learning_rate": 1.2045780171273978e-05, "loss": 0.3694, "step": 9040 }, { "epoch": 0.45, "learning_rate": 1.2044197137840974e-05, "loss": 0.521, "step": 9041 }, { "epoch": 0.45, "learning_rate": 1.2042614050944665e-05, "loss": 0.4032, "step": 9042 }, { "epoch": 0.45, "learning_rate": 1.2041030910626467e-05, "loss": 0.5394, "step": 9043 }, { "epoch": 0.45, "learning_rate": 1.2039447716927774e-05, "loss": 0.2811, "step": 9044 }, { "epoch": 0.45, "learning_rate": 1.2037864469890001e-05, "loss": 0.5856, "step": 9045 }, { "epoch": 0.45, "learning_rate": 1.2036281169554549e-05, "loss": 0.2357, "step": 9046 }, { "epoch": 0.45, "learning_rate": 1.2034697815962835e-05, "loss": 0.3191, "step": 9047 }, { "epoch": 0.45, "learning_rate": 1.203311440915626e-05, "loss": 0.2824, "step": 9048 }, { "epoch": 0.45, "learning_rate": 1.203153094917624e-05, "loss": 0.2091, "step": 9049 }, { "epoch": 0.45, "learning_rate": 1.2029947436064193e-05, "loss": 0.4196, "step": 9050 }, { "epoch": 0.45, "learning_rate": 1.2028363869861527e-05, "loss": 0.2771, "step": 9051 }, { "epoch": 0.45, "learning_rate": 1.2026780250609661e-05, "loss": 0.247, "step": 9052 }, { "epoch": 0.45, "learning_rate": 1.2025196578350014e-05, "loss": 0.2294, "step": 9053 }, { "epoch": 0.45, "learning_rate": 1.2023612853124001e-05, "loss": 0.369, "step": 9054 }, { "epoch": 0.45, "learning_rate": 1.2022029074973043e-05, "loss": 0.6943, "step": 9055 }, { "epoch": 0.45, "learning_rate": 1.2020445243938563e-05, "loss": 0.2848, "step": 9056 }, { "epoch": 0.45, "learning_rate": 1.2018861360061989e-05, "loss": 0.479, "step": 9057 }, { "epoch": 0.45, "learning_rate": 1.2017277423384734e-05, "loss": 0.2552, "step": 9058 }, { "epoch": 0.45, "learning_rate": 1.2015693433948235e-05, "loss": 0.1968, "step": 9059 }, { "epoch": 0.45, "learning_rate": 1.201410939179391e-05, "loss": 0.1323, "step": 9060 }, { "epoch": 0.45, "learning_rate": 1.2012525296963193e-05, "loss": 0.2857, "step": 9061 }, { "epoch": 0.45, "learning_rate": 1.2010941149497513e-05, "loss": 0.3331, "step": 9062 }, { "epoch": 0.45, "learning_rate": 1.2009356949438302e-05, "loss": 0.5081, "step": 9063 }, { "epoch": 0.45, "learning_rate": 1.2007772696826994e-05, "loss": 0.4036, "step": 9064 }, { "epoch": 0.45, "learning_rate": 1.2006188391705019e-05, "loss": 0.396, "step": 9065 }, { "epoch": 0.45, "learning_rate": 1.2004604034113814e-05, "loss": 0.1362, "step": 9066 }, { "epoch": 0.45, "learning_rate": 1.2003019624094814e-05, "loss": 0.4347, "step": 9067 }, { "epoch": 0.45, "learning_rate": 1.2001435161689463e-05, "loss": 0.2497, "step": 9068 }, { "epoch": 0.45, "learning_rate": 1.1999850646939193e-05, "loss": 0.3097, "step": 9069 }, { "epoch": 0.45, "learning_rate": 1.1998266079885453e-05, "loss": 0.3355, "step": 9070 }, { "epoch": 0.45, "learning_rate": 1.1996681460569675e-05, "loss": 0.5029, "step": 9071 }, { "epoch": 0.45, "learning_rate": 1.1995096789033314e-05, "loss": 0.3734, "step": 9072 }, { "epoch": 0.45, "learning_rate": 1.1993512065317806e-05, "loss": 0.3244, "step": 9073 }, { "epoch": 0.45, "learning_rate": 1.1991927289464602e-05, "loss": 0.557, "step": 9074 }, { "epoch": 0.45, "learning_rate": 1.199034246151515e-05, "loss": 0.384, "step": 9075 }, { "epoch": 0.45, "learning_rate": 1.1988757581510897e-05, "loss": 0.321, "step": 9076 }, { "epoch": 0.45, "learning_rate": 1.1987172649493295e-05, "loss": 0.419, "step": 9077 }, { "epoch": 0.45, "learning_rate": 1.1985587665503795e-05, "loss": 0.4464, "step": 9078 }, { "epoch": 0.45, "learning_rate": 1.1984002629583848e-05, "loss": 0.5538, "step": 9079 }, { "epoch": 0.45, "learning_rate": 1.1982417541774911e-05, "loss": 0.5289, "step": 9080 }, { "epoch": 0.45, "learning_rate": 1.1980832402118442e-05, "loss": 0.4723, "step": 9081 }, { "epoch": 0.45, "learning_rate": 1.1979247210655897e-05, "loss": 0.5802, "step": 9082 }, { "epoch": 0.45, "learning_rate": 1.197766196742873e-05, "loss": 0.1503, "step": 9083 }, { "epoch": 0.45, "learning_rate": 1.1976076672478406e-05, "loss": 0.4815, "step": 9084 }, { "epoch": 0.45, "learning_rate": 1.1974491325846384e-05, "loss": 0.3433, "step": 9085 }, { "epoch": 0.45, "learning_rate": 1.197290592757413e-05, "loss": 0.1968, "step": 9086 }, { "epoch": 0.45, "learning_rate": 1.1971320477703104e-05, "loss": 0.3839, "step": 9087 }, { "epoch": 0.45, "learning_rate": 1.1969734976274774e-05, "loss": 0.3761, "step": 9088 }, { "epoch": 0.45, "learning_rate": 1.1968149423330606e-05, "loss": 0.3803, "step": 9089 }, { "epoch": 0.45, "learning_rate": 1.1966563818912065e-05, "loss": 0.3788, "step": 9090 }, { "epoch": 0.45, "learning_rate": 1.1964978163060623e-05, "loss": 0.5, "step": 9091 }, { "epoch": 0.45, "learning_rate": 1.1963392455817751e-05, "loss": 0.6514, "step": 9092 }, { "epoch": 0.45, "learning_rate": 1.1961806697224922e-05, "loss": 0.2813, "step": 9093 }, { "epoch": 0.45, "learning_rate": 1.1960220887323608e-05, "loss": 0.564, "step": 9094 }, { "epoch": 0.45, "learning_rate": 1.1958635026155285e-05, "loss": 0.279, "step": 9095 }, { "epoch": 0.45, "learning_rate": 1.1957049113761426e-05, "loss": 0.3513, "step": 9096 }, { "epoch": 0.45, "learning_rate": 1.1955463150183514e-05, "loss": 0.4032, "step": 9097 }, { "epoch": 0.45, "learning_rate": 1.1953877135463021e-05, "loss": 0.28, "step": 9098 }, { "epoch": 0.45, "learning_rate": 1.1952291069641432e-05, "loss": 0.3769, "step": 9099 }, { "epoch": 0.45, "learning_rate": 1.1950704952760228e-05, "loss": 0.3703, "step": 9100 }, { "epoch": 0.45, "learning_rate": 1.1949118784860891e-05, "loss": 0.1197, "step": 9101 }, { "epoch": 0.45, "learning_rate": 1.1947532565984904e-05, "loss": 0.3468, "step": 9102 }, { "epoch": 0.45, "learning_rate": 1.1945946296173753e-05, "loss": 0.4703, "step": 9103 }, { "epoch": 0.45, "learning_rate": 1.1944359975468929e-05, "loss": 0.4686, "step": 9104 }, { "epoch": 0.45, "learning_rate": 1.1942773603911914e-05, "loss": 0.4508, "step": 9105 }, { "epoch": 0.45, "learning_rate": 1.1941187181544198e-05, "loss": 0.4278, "step": 9106 }, { "epoch": 0.45, "learning_rate": 1.1939600708407278e-05, "loss": 0.3795, "step": 9107 }, { "epoch": 0.45, "learning_rate": 1.1938014184542636e-05, "loss": 0.251, "step": 9108 }, { "epoch": 0.45, "learning_rate": 1.1936427609991777e-05, "loss": 0.6745, "step": 9109 }, { "epoch": 0.45, "learning_rate": 1.1934840984796186e-05, "loss": 0.3755, "step": 9110 }, { "epoch": 0.45, "learning_rate": 1.1933254308997366e-05, "loss": 0.3042, "step": 9111 }, { "epoch": 0.45, "learning_rate": 1.1931667582636813e-05, "loss": 0.4483, "step": 9112 }, { "epoch": 0.46, "learning_rate": 1.1930080805756022e-05, "loss": 0.3632, "step": 9113 }, { "epoch": 0.46, "learning_rate": 1.1928493978396495e-05, "loss": 0.3586, "step": 9114 }, { "epoch": 0.46, "learning_rate": 1.1926907100599733e-05, "loss": 0.152, "step": 9115 }, { "epoch": 0.46, "learning_rate": 1.1925320172407245e-05, "loss": 0.4812, "step": 9116 }, { "epoch": 0.46, "learning_rate": 1.1923733193860523e-05, "loss": 0.3099, "step": 9117 }, { "epoch": 0.46, "learning_rate": 1.1922146165001083e-05, "loss": 0.2265, "step": 9118 }, { "epoch": 0.46, "learning_rate": 1.1920559085870425e-05, "loss": 0.5229, "step": 9119 }, { "epoch": 0.46, "learning_rate": 1.1918971956510061e-05, "loss": 0.4102, "step": 9120 }, { "epoch": 0.46, "learning_rate": 1.1917384776961498e-05, "loss": 0.2521, "step": 9121 }, { "epoch": 0.46, "learning_rate": 1.1915797547266248e-05, "loss": 0.323, "step": 9122 }, { "epoch": 0.46, "learning_rate": 1.1914210267465824e-05, "loss": 0.2695, "step": 9123 }, { "epoch": 0.46, "learning_rate": 1.1912622937601737e-05, "loss": 0.4344, "step": 9124 }, { "epoch": 0.46, "learning_rate": 1.1911035557715502e-05, "loss": 0.2899, "step": 9125 }, { "epoch": 0.46, "learning_rate": 1.1909448127848633e-05, "loss": 0.6419, "step": 9126 }, { "epoch": 0.46, "learning_rate": 1.190786064804265e-05, "loss": 0.1644, "step": 9127 }, { "epoch": 0.46, "learning_rate": 1.190627311833907e-05, "loss": 0.5334, "step": 9128 }, { "epoch": 0.46, "learning_rate": 1.1904685538779416e-05, "loss": 0.2678, "step": 9129 }, { "epoch": 0.46, "learning_rate": 1.1903097909405206e-05, "loss": 0.443, "step": 9130 }, { "epoch": 0.46, "learning_rate": 1.190151023025796e-05, "loss": 0.1733, "step": 9131 }, { "epoch": 0.46, "learning_rate": 1.1899922501379208e-05, "loss": 0.2795, "step": 9132 }, { "epoch": 0.46, "learning_rate": 1.1898334722810467e-05, "loss": 0.4229, "step": 9133 }, { "epoch": 0.46, "learning_rate": 1.1896746894593271e-05, "loss": 0.1315, "step": 9134 }, { "epoch": 0.46, "learning_rate": 1.1895159016769142e-05, "loss": 0.2763, "step": 9135 }, { "epoch": 0.46, "learning_rate": 1.1893571089379614e-05, "loss": 0.3872, "step": 9136 }, { "epoch": 0.46, "learning_rate": 1.1891983112466215e-05, "loss": 0.3384, "step": 9137 }, { "epoch": 0.46, "learning_rate": 1.1890395086070474e-05, "loss": 0.4012, "step": 9138 }, { "epoch": 0.46, "learning_rate": 1.1888807010233924e-05, "loss": 0.2507, "step": 9139 }, { "epoch": 0.46, "learning_rate": 1.1887218884998102e-05, "loss": 0.1487, "step": 9140 }, { "epoch": 0.46, "learning_rate": 1.1885630710404543e-05, "loss": 0.5108, "step": 9141 }, { "epoch": 0.46, "learning_rate": 1.188404248649478e-05, "loss": 0.405, "step": 9142 }, { "epoch": 0.46, "learning_rate": 1.1882454213310355e-05, "loss": 0.3292, "step": 9143 }, { "epoch": 0.46, "learning_rate": 1.1880865890892804e-05, "loss": 0.3877, "step": 9144 }, { "epoch": 0.46, "learning_rate": 1.1879277519283672e-05, "loss": 0.4335, "step": 9145 }, { "epoch": 0.46, "learning_rate": 1.1877689098524497e-05, "loss": 0.4186, "step": 9146 }, { "epoch": 0.46, "learning_rate": 1.187610062865682e-05, "loss": 0.3662, "step": 9147 }, { "epoch": 0.46, "learning_rate": 1.1874512109722192e-05, "loss": 0.3932, "step": 9148 }, { "epoch": 0.46, "learning_rate": 1.1872923541762153e-05, "loss": 0.4747, "step": 9149 }, { "epoch": 0.46, "learning_rate": 1.1871334924818254e-05, "loss": 0.47, "step": 9150 }, { "epoch": 0.46, "learning_rate": 1.1869746258932038e-05, "loss": 0.3563, "step": 9151 }, { "epoch": 0.46, "learning_rate": 1.1868157544145058e-05, "loss": 0.2394, "step": 9152 }, { "epoch": 0.46, "learning_rate": 1.1866568780498863e-05, "loss": 0.3076, "step": 9153 }, { "epoch": 0.46, "learning_rate": 1.186497996803501e-05, "loss": 0.4566, "step": 9154 }, { "epoch": 0.46, "learning_rate": 1.1863391106795048e-05, "loss": 0.3639, "step": 9155 }, { "epoch": 0.46, "learning_rate": 1.186180219682053e-05, "loss": 0.332, "step": 9156 }, { "epoch": 0.46, "learning_rate": 1.1860213238153016e-05, "loss": 0.4651, "step": 9157 }, { "epoch": 0.46, "learning_rate": 1.185862423083406e-05, "loss": 0.2798, "step": 9158 }, { "epoch": 0.46, "learning_rate": 1.1857035174905223e-05, "loss": 0.2329, "step": 9159 }, { "epoch": 0.46, "learning_rate": 1.1855446070408063e-05, "loss": 0.2786, "step": 9160 }, { "epoch": 0.46, "learning_rate": 1.1853856917384141e-05, "loss": 0.4811, "step": 9161 }, { "epoch": 0.46, "learning_rate": 1.1852267715875018e-05, "loss": 0.3895, "step": 9162 }, { "epoch": 0.46, "learning_rate": 1.1850678465922261e-05, "loss": 0.4125, "step": 9163 }, { "epoch": 0.46, "learning_rate": 1.1849089167567432e-05, "loss": 0.2529, "step": 9164 }, { "epoch": 0.46, "learning_rate": 1.1847499820852098e-05, "loss": 0.3689, "step": 9165 }, { "epoch": 0.46, "learning_rate": 1.1845910425817823e-05, "loss": 0.3757, "step": 9166 }, { "epoch": 0.46, "learning_rate": 1.1844320982506181e-05, "loss": 0.3645, "step": 9167 }, { "epoch": 0.46, "learning_rate": 1.184273149095874e-05, "loss": 0.361, "step": 9168 }, { "epoch": 0.46, "learning_rate": 1.1841141951217068e-05, "loss": 0.5119, "step": 9169 }, { "epoch": 0.46, "learning_rate": 1.1839552363322741e-05, "loss": 0.2191, "step": 9170 }, { "epoch": 0.46, "learning_rate": 1.1837962727317334e-05, "loss": 0.3791, "step": 9171 }, { "epoch": 0.46, "learning_rate": 1.1836373043242414e-05, "loss": 0.2506, "step": 9172 }, { "epoch": 0.46, "learning_rate": 1.1834783311139568e-05, "loss": 0.516, "step": 9173 }, { "epoch": 0.46, "learning_rate": 1.1833193531050364e-05, "loss": 0.283, "step": 9174 }, { "epoch": 0.46, "learning_rate": 1.1831603703016387e-05, "loss": 0.2185, "step": 9175 }, { "epoch": 0.46, "learning_rate": 1.183001382707921e-05, "loss": 0.3746, "step": 9176 }, { "epoch": 0.46, "learning_rate": 1.1828423903280422e-05, "loss": 0.3799, "step": 9177 }, { "epoch": 0.46, "learning_rate": 1.1826833931661602e-05, "loss": 0.4231, "step": 9178 }, { "epoch": 0.46, "learning_rate": 1.1825243912264328e-05, "loss": 0.5266, "step": 9179 }, { "epoch": 0.46, "learning_rate": 1.1823653845130194e-05, "loss": 0.3493, "step": 9180 }, { "epoch": 0.46, "learning_rate": 1.1822063730300779e-05, "loss": 0.5046, "step": 9181 }, { "epoch": 0.46, "learning_rate": 1.1820473567817678e-05, "loss": 0.365, "step": 9182 }, { "epoch": 0.46, "learning_rate": 1.1818883357722473e-05, "loss": 0.3868, "step": 9183 }, { "epoch": 0.46, "learning_rate": 1.1817293100056757e-05, "loss": 0.2079, "step": 9184 }, { "epoch": 0.46, "learning_rate": 1.1815702794862121e-05, "loss": 0.3027, "step": 9185 }, { "epoch": 0.46, "learning_rate": 1.1814112442180154e-05, "loss": 0.4755, "step": 9186 }, { "epoch": 0.46, "learning_rate": 1.1812522042052454e-05, "loss": 0.2228, "step": 9187 }, { "epoch": 0.46, "learning_rate": 1.1810931594520612e-05, "loss": 0.3035, "step": 9188 }, { "epoch": 0.46, "learning_rate": 1.1809341099626227e-05, "loss": 0.5315, "step": 9189 }, { "epoch": 0.46, "learning_rate": 1.1807750557410896e-05, "loss": 0.3967, "step": 9190 }, { "epoch": 0.46, "learning_rate": 1.1806159967916218e-05, "loss": 0.3254, "step": 9191 }, { "epoch": 0.46, "learning_rate": 1.1804569331183788e-05, "loss": 0.5075, "step": 9192 }, { "epoch": 0.46, "learning_rate": 1.1802978647255214e-05, "loss": 0.2427, "step": 9193 }, { "epoch": 0.46, "learning_rate": 1.1801387916172096e-05, "loss": 0.5351, "step": 9194 }, { "epoch": 0.46, "learning_rate": 1.1799797137976033e-05, "loss": 0.3343, "step": 9195 }, { "epoch": 0.46, "learning_rate": 1.1798206312708633e-05, "loss": 0.1977, "step": 9196 }, { "epoch": 0.46, "learning_rate": 1.1796615440411505e-05, "loss": 0.1336, "step": 9197 }, { "epoch": 0.46, "learning_rate": 1.179502452112625e-05, "loss": 0.3797, "step": 9198 }, { "epoch": 0.46, "learning_rate": 1.1793433554894484e-05, "loss": 0.4418, "step": 9199 }, { "epoch": 0.46, "learning_rate": 1.179184254175781e-05, "loss": 0.568, "step": 9200 }, { "epoch": 0.46, "learning_rate": 1.179025148175784e-05, "loss": 0.3703, "step": 9201 }, { "epoch": 0.46, "learning_rate": 1.178866037493619e-05, "loss": 0.3037, "step": 9202 }, { "epoch": 0.46, "learning_rate": 1.178706922133447e-05, "loss": 0.4346, "step": 9203 }, { "epoch": 0.46, "learning_rate": 1.1785478020994292e-05, "loss": 0.4787, "step": 9204 }, { "epoch": 0.46, "learning_rate": 1.1783886773957278e-05, "loss": 0.4073, "step": 9205 }, { "epoch": 0.46, "learning_rate": 1.178229548026504e-05, "loss": 0.4457, "step": 9206 }, { "epoch": 0.46, "learning_rate": 1.1780704139959201e-05, "loss": 0.2416, "step": 9207 }, { "epoch": 0.46, "learning_rate": 1.1779112753081374e-05, "loss": 0.3999, "step": 9208 }, { "epoch": 0.46, "learning_rate": 1.1777521319673186e-05, "loss": 0.4435, "step": 9209 }, { "epoch": 0.46, "learning_rate": 1.1775929839776255e-05, "loss": 0.6329, "step": 9210 }, { "epoch": 0.46, "learning_rate": 1.1774338313432205e-05, "loss": 0.2234, "step": 9211 }, { "epoch": 0.46, "learning_rate": 1.177274674068266e-05, "loss": 0.3041, "step": 9212 }, { "epoch": 0.46, "learning_rate": 1.1771155121569247e-05, "loss": 0.2587, "step": 9213 }, { "epoch": 0.46, "learning_rate": 1.1769563456133592e-05, "loss": 0.3354, "step": 9214 }, { "epoch": 0.46, "learning_rate": 1.1767971744417319e-05, "loss": 0.3556, "step": 9215 }, { "epoch": 0.46, "learning_rate": 1.1766379986462064e-05, "loss": 0.5829, "step": 9216 }, { "epoch": 0.46, "learning_rate": 1.1764788182309453e-05, "loss": 0.3103, "step": 9217 }, { "epoch": 0.46, "learning_rate": 1.1763196332001118e-05, "loss": 0.5504, "step": 9218 }, { "epoch": 0.46, "learning_rate": 1.1761604435578696e-05, "loss": 0.2756, "step": 9219 }, { "epoch": 0.46, "learning_rate": 1.176001249308381e-05, "loss": 0.4159, "step": 9220 }, { "epoch": 0.46, "learning_rate": 1.1758420504558108e-05, "loss": 0.2496, "step": 9221 }, { "epoch": 0.46, "learning_rate": 1.175682847004322e-05, "loss": 0.2846, "step": 9222 }, { "epoch": 0.46, "learning_rate": 1.1755236389580787e-05, "loss": 0.3812, "step": 9223 }, { "epoch": 0.46, "learning_rate": 1.1753644263212438e-05, "loss": 0.2558, "step": 9224 }, { "epoch": 0.46, "learning_rate": 1.1752052090979827e-05, "loss": 0.3791, "step": 9225 }, { "epoch": 0.46, "learning_rate": 1.1750459872924584e-05, "loss": 0.4153, "step": 9226 }, { "epoch": 0.46, "learning_rate": 1.1748867609088357e-05, "loss": 0.4972, "step": 9227 }, { "epoch": 0.46, "learning_rate": 1.1747275299512788e-05, "loss": 0.3736, "step": 9228 }, { "epoch": 0.46, "learning_rate": 1.1745682944239523e-05, "loss": 0.3967, "step": 9229 }, { "epoch": 0.46, "learning_rate": 1.1744090543310206e-05, "loss": 0.1496, "step": 9230 }, { "epoch": 0.46, "learning_rate": 1.1742498096766487e-05, "loss": 0.4919, "step": 9231 }, { "epoch": 0.46, "learning_rate": 1.174090560465001e-05, "loss": 0.3059, "step": 9232 }, { "epoch": 0.46, "learning_rate": 1.1739313067002429e-05, "loss": 0.3578, "step": 9233 }, { "epoch": 0.46, "learning_rate": 1.1737720483865392e-05, "loss": 0.2816, "step": 9234 }, { "epoch": 0.46, "learning_rate": 1.1736127855280553e-05, "loss": 0.3129, "step": 9235 }, { "epoch": 0.46, "learning_rate": 1.1734535181289564e-05, "loss": 0.3075, "step": 9236 }, { "epoch": 0.46, "learning_rate": 1.1732942461934079e-05, "loss": 0.3818, "step": 9237 }, { "epoch": 0.46, "learning_rate": 1.173134969725575e-05, "loss": 0.3115, "step": 9238 }, { "epoch": 0.46, "learning_rate": 1.172975688729624e-05, "loss": 0.2277, "step": 9239 }, { "epoch": 0.46, "learning_rate": 1.1728164032097204e-05, "loss": 0.5826, "step": 9240 }, { "epoch": 0.46, "learning_rate": 1.1726571131700302e-05, "loss": 0.3074, "step": 9241 }, { "epoch": 0.46, "learning_rate": 1.1724978186147193e-05, "loss": 0.3904, "step": 9242 }, { "epoch": 0.46, "learning_rate": 1.1723385195479537e-05, "loss": 0.3404, "step": 9243 }, { "epoch": 0.46, "learning_rate": 1.1721792159739001e-05, "loss": 0.45, "step": 9244 }, { "epoch": 0.46, "learning_rate": 1.1720199078967244e-05, "loss": 0.6159, "step": 9245 }, { "epoch": 0.46, "learning_rate": 1.1718605953205936e-05, "loss": 0.2024, "step": 9246 }, { "epoch": 0.46, "learning_rate": 1.1717012782496739e-05, "loss": 0.2976, "step": 9247 }, { "epoch": 0.46, "learning_rate": 1.1715419566881323e-05, "loss": 0.4167, "step": 9248 }, { "epoch": 0.46, "learning_rate": 1.1713826306401352e-05, "loss": 0.2826, "step": 9249 }, { "epoch": 0.46, "learning_rate": 1.1712233001098501e-05, "loss": 0.5843, "step": 9250 }, { "epoch": 0.46, "learning_rate": 1.1710639651014437e-05, "loss": 0.3738, "step": 9251 }, { "epoch": 0.46, "learning_rate": 1.1709046256190832e-05, "loss": 0.3442, "step": 9252 }, { "epoch": 0.46, "learning_rate": 1.1707452816669365e-05, "loss": 0.4609, "step": 9253 }, { "epoch": 0.46, "learning_rate": 1.1705859332491703e-05, "loss": 0.352, "step": 9254 }, { "epoch": 0.46, "learning_rate": 1.1704265803699524e-05, "loss": 0.4145, "step": 9255 }, { "epoch": 0.46, "learning_rate": 1.1702672230334504e-05, "loss": 0.2838, "step": 9256 }, { "epoch": 0.46, "learning_rate": 1.1701078612438326e-05, "loss": 0.392, "step": 9257 }, { "epoch": 0.46, "learning_rate": 1.1699484950052664e-05, "loss": 0.3402, "step": 9258 }, { "epoch": 0.46, "learning_rate": 1.1697891243219198e-05, "loss": 0.4004, "step": 9259 }, { "epoch": 0.46, "learning_rate": 1.1696297491979611e-05, "loss": 0.0988, "step": 9260 }, { "epoch": 0.46, "learning_rate": 1.1694703696375583e-05, "loss": 0.4276, "step": 9261 }, { "epoch": 0.46, "learning_rate": 1.16931098564488e-05, "loss": 0.3734, "step": 9262 }, { "epoch": 0.46, "learning_rate": 1.1691515972240948e-05, "loss": 0.3442, "step": 9263 }, { "epoch": 0.46, "learning_rate": 1.1689922043793709e-05, "loss": 0.5233, "step": 9264 }, { "epoch": 0.46, "learning_rate": 1.1688328071148773e-05, "loss": 0.2882, "step": 9265 }, { "epoch": 0.46, "learning_rate": 1.1686734054347828e-05, "loss": 0.3761, "step": 9266 }, { "epoch": 0.46, "learning_rate": 1.1685139993432564e-05, "loss": 0.2315, "step": 9267 }, { "epoch": 0.46, "learning_rate": 1.168354588844467e-05, "loss": 0.2482, "step": 9268 }, { "epoch": 0.46, "learning_rate": 1.1681951739425838e-05, "loss": 0.4103, "step": 9269 }, { "epoch": 0.46, "learning_rate": 1.1680357546417759e-05, "loss": 0.313, "step": 9270 }, { "epoch": 0.46, "learning_rate": 1.1678763309462133e-05, "loss": 0.4784, "step": 9271 }, { "epoch": 0.46, "learning_rate": 1.167716902860065e-05, "loss": 0.1409, "step": 9272 }, { "epoch": 0.46, "learning_rate": 1.167557470387501e-05, "loss": 0.2978, "step": 9273 }, { "epoch": 0.46, "learning_rate": 1.1673980335326904e-05, "loss": 0.3616, "step": 9274 }, { "epoch": 0.46, "learning_rate": 1.1672385922998038e-05, "loss": 0.2457, "step": 9275 }, { "epoch": 0.46, "learning_rate": 1.1670791466930108e-05, "loss": 0.3234, "step": 9276 }, { "epoch": 0.46, "learning_rate": 1.1669196967164814e-05, "loss": 0.1665, "step": 9277 }, { "epoch": 0.46, "learning_rate": 1.166760242374386e-05, "loss": 0.3516, "step": 9278 }, { "epoch": 0.46, "learning_rate": 1.1666007836708948e-05, "loss": 0.35, "step": 9279 }, { "epoch": 0.46, "learning_rate": 1.1664413206101785e-05, "loss": 0.3515, "step": 9280 }, { "epoch": 0.46, "learning_rate": 1.1662818531964073e-05, "loss": 0.409, "step": 9281 }, { "epoch": 0.46, "learning_rate": 1.1661223814337524e-05, "loss": 0.3127, "step": 9282 }, { "epoch": 0.46, "learning_rate": 1.165962905326384e-05, "loss": 0.2479, "step": 9283 }, { "epoch": 0.46, "learning_rate": 1.1658034248784731e-05, "loss": 0.4646, "step": 9284 }, { "epoch": 0.46, "learning_rate": 1.1656439400941908e-05, "loss": 0.3046, "step": 9285 }, { "epoch": 0.46, "learning_rate": 1.1654844509777078e-05, "loss": 0.2471, "step": 9286 }, { "epoch": 0.46, "learning_rate": 1.1653249575331964e-05, "loss": 0.2993, "step": 9287 }, { "epoch": 0.46, "learning_rate": 1.1651654597648268e-05, "loss": 0.343, "step": 9288 }, { "epoch": 0.46, "learning_rate": 1.1650059576767712e-05, "loss": 0.3195, "step": 9289 }, { "epoch": 0.46, "learning_rate": 1.1648464512732008e-05, "loss": 0.4136, "step": 9290 }, { "epoch": 0.46, "learning_rate": 1.1646869405582874e-05, "loss": 0.6847, "step": 9291 }, { "epoch": 0.46, "learning_rate": 1.1645274255362028e-05, "loss": 0.3459, "step": 9292 }, { "epoch": 0.46, "learning_rate": 1.1643679062111185e-05, "loss": 0.3234, "step": 9293 }, { "epoch": 0.46, "learning_rate": 1.1642083825872075e-05, "loss": 0.5714, "step": 9294 }, { "epoch": 0.46, "learning_rate": 1.164048854668641e-05, "loss": 0.2182, "step": 9295 }, { "epoch": 0.46, "learning_rate": 1.1638893224595918e-05, "loss": 0.2277, "step": 9296 }, { "epoch": 0.46, "learning_rate": 1.1637297859642318e-05, "loss": 0.3401, "step": 9297 }, { "epoch": 0.46, "learning_rate": 1.1635702451867339e-05, "loss": 0.225, "step": 9298 }, { "epoch": 0.46, "learning_rate": 1.1634107001312703e-05, "loss": 0.4174, "step": 9299 }, { "epoch": 0.46, "learning_rate": 1.1632511508020139e-05, "loss": 0.4598, "step": 9300 }, { "epoch": 0.46, "learning_rate": 1.1630915972031373e-05, "loss": 0.4965, "step": 9301 }, { "epoch": 0.46, "learning_rate": 1.1629320393388137e-05, "loss": 0.3106, "step": 9302 }, { "epoch": 0.46, "learning_rate": 1.1627724772132163e-05, "loss": 0.2097, "step": 9303 }, { "epoch": 0.46, "learning_rate": 1.1626129108305175e-05, "loss": 0.4628, "step": 9304 }, { "epoch": 0.46, "learning_rate": 1.1624533401948911e-05, "loss": 0.7539, "step": 9305 }, { "epoch": 0.46, "learning_rate": 1.1622937653105106e-05, "loss": 0.2976, "step": 9306 }, { "epoch": 0.46, "learning_rate": 1.1621341861815492e-05, "loss": 0.4356, "step": 9307 }, { "epoch": 0.46, "learning_rate": 1.1619746028121806e-05, "loss": 0.461, "step": 9308 }, { "epoch": 0.46, "learning_rate": 1.1618150152065782e-05, "loss": 0.261, "step": 9309 }, { "epoch": 0.46, "learning_rate": 1.161655423368916e-05, "loss": 0.3636, "step": 9310 }, { "epoch": 0.46, "learning_rate": 1.1614958273033681e-05, "loss": 0.3, "step": 9311 }, { "epoch": 0.46, "learning_rate": 1.1613362270141081e-05, "loss": 0.2016, "step": 9312 }, { "epoch": 0.47, "learning_rate": 1.1611766225053105e-05, "loss": 0.4642, "step": 9313 }, { "epoch": 0.47, "learning_rate": 1.1610170137811495e-05, "loss": 0.1936, "step": 9314 }, { "epoch": 0.47, "learning_rate": 1.1608574008457995e-05, "loss": 0.4815, "step": 9315 }, { "epoch": 0.47, "learning_rate": 1.1606977837034348e-05, "loss": 0.284, "step": 9316 }, { "epoch": 0.47, "learning_rate": 1.1605381623582299e-05, "loss": 0.7275, "step": 9317 }, { "epoch": 0.47, "learning_rate": 1.1603785368143597e-05, "loss": 0.3478, "step": 9318 }, { "epoch": 0.47, "learning_rate": 1.1602189070759991e-05, "loss": 0.2762, "step": 9319 }, { "epoch": 0.47, "learning_rate": 1.1600592731473227e-05, "loss": 0.6501, "step": 9320 }, { "epoch": 0.47, "learning_rate": 1.1598996350325057e-05, "loss": 0.5439, "step": 9321 }, { "epoch": 0.47, "learning_rate": 1.1597399927357231e-05, "loss": 0.1748, "step": 9322 }, { "epoch": 0.47, "learning_rate": 1.1595803462611501e-05, "loss": 0.3227, "step": 9323 }, { "epoch": 0.47, "learning_rate": 1.1594206956129625e-05, "loss": 0.4311, "step": 9324 }, { "epoch": 0.47, "learning_rate": 1.1592610407953352e-05, "loss": 0.3928, "step": 9325 }, { "epoch": 0.47, "learning_rate": 1.159101381812444e-05, "loss": 0.527, "step": 9326 }, { "epoch": 0.47, "learning_rate": 1.1589417186684645e-05, "loss": 0.3051, "step": 9327 }, { "epoch": 0.47, "learning_rate": 1.1587820513675725e-05, "loss": 0.466, "step": 9328 }, { "epoch": 0.47, "learning_rate": 1.1586223799139438e-05, "loss": 0.3129, "step": 9329 }, { "epoch": 0.47, "learning_rate": 1.1584627043117546e-05, "loss": 0.1947, "step": 9330 }, { "epoch": 0.47, "learning_rate": 1.1583030245651812e-05, "loss": 0.4597, "step": 9331 }, { "epoch": 0.47, "learning_rate": 1.1581433406783993e-05, "loss": 0.4263, "step": 9332 }, { "epoch": 0.47, "learning_rate": 1.1579836526555857e-05, "loss": 0.4089, "step": 9333 }, { "epoch": 0.47, "learning_rate": 1.157823960500916e-05, "loss": 0.3157, "step": 9334 }, { "epoch": 0.47, "learning_rate": 1.1576642642185678e-05, "loss": 0.5788, "step": 9335 }, { "epoch": 0.47, "learning_rate": 1.1575045638127172e-05, "loss": 0.626, "step": 9336 }, { "epoch": 0.47, "learning_rate": 1.1573448592875409e-05, "loss": 0.3713, "step": 9337 }, { "epoch": 0.47, "learning_rate": 1.1571851506472158e-05, "loss": 0.2693, "step": 9338 }, { "epoch": 0.47, "learning_rate": 1.1570254378959194e-05, "loss": 0.3614, "step": 9339 }, { "epoch": 0.47, "learning_rate": 1.156865721037828e-05, "loss": 0.3886, "step": 9340 }, { "epoch": 0.47, "learning_rate": 1.1567060000771192e-05, "loss": 0.4408, "step": 9341 }, { "epoch": 0.47, "learning_rate": 1.15654627501797e-05, "loss": 0.1726, "step": 9342 }, { "epoch": 0.47, "learning_rate": 1.1563865458645582e-05, "loss": 0.2881, "step": 9343 }, { "epoch": 0.47, "learning_rate": 1.1562268126210615e-05, "loss": 0.3678, "step": 9344 }, { "epoch": 0.47, "learning_rate": 1.1560670752916567e-05, "loss": 0.4497, "step": 9345 }, { "epoch": 0.47, "learning_rate": 1.155907333880522e-05, "loss": 0.3425, "step": 9346 }, { "epoch": 0.47, "learning_rate": 1.1557475883918355e-05, "loss": 0.2153, "step": 9347 }, { "epoch": 0.47, "learning_rate": 1.1555878388297744e-05, "loss": 0.0641, "step": 9348 }, { "epoch": 0.47, "learning_rate": 1.1554280851985174e-05, "loss": 0.356, "step": 9349 }, { "epoch": 0.47, "learning_rate": 1.155268327502242e-05, "loss": 0.2801, "step": 9350 }, { "epoch": 0.47, "learning_rate": 1.1551085657451272e-05, "loss": 0.4457, "step": 9351 }, { "epoch": 0.47, "learning_rate": 1.154948799931351e-05, "loss": 0.5628, "step": 9352 }, { "epoch": 0.47, "learning_rate": 1.1547890300650919e-05, "loss": 0.514, "step": 9353 }, { "epoch": 0.47, "learning_rate": 1.1546292561505284e-05, "loss": 0.3585, "step": 9354 }, { "epoch": 0.47, "learning_rate": 1.1544694781918392e-05, "loss": 0.4141, "step": 9355 }, { "epoch": 0.47, "learning_rate": 1.1543096961932032e-05, "loss": 0.3868, "step": 9356 }, { "epoch": 0.47, "learning_rate": 1.1541499101587994e-05, "loss": 0.2596, "step": 9357 }, { "epoch": 0.47, "learning_rate": 1.1539901200928066e-05, "loss": 0.4462, "step": 9358 }, { "epoch": 0.47, "learning_rate": 1.1538303259994035e-05, "loss": 0.3508, "step": 9359 }, { "epoch": 0.47, "learning_rate": 1.1536705278827698e-05, "loss": 0.4849, "step": 9360 }, { "epoch": 0.47, "learning_rate": 1.1535107257470848e-05, "loss": 0.4853, "step": 9361 }, { "epoch": 0.47, "learning_rate": 1.1533509195965277e-05, "loss": 0.8219, "step": 9362 }, { "epoch": 0.47, "learning_rate": 1.1531911094352786e-05, "loss": 0.3288, "step": 9363 }, { "epoch": 0.47, "learning_rate": 1.1530312952675162e-05, "loss": 0.5109, "step": 9364 }, { "epoch": 0.47, "learning_rate": 1.152871477097421e-05, "loss": 0.3444, "step": 9365 }, { "epoch": 0.47, "learning_rate": 1.152711654929172e-05, "loss": 0.2561, "step": 9366 }, { "epoch": 0.47, "learning_rate": 1.1525518287669503e-05, "loss": 0.3579, "step": 9367 }, { "epoch": 0.47, "learning_rate": 1.1523919986149352e-05, "loss": 0.4401, "step": 9368 }, { "epoch": 0.47, "learning_rate": 1.152232164477307e-05, "loss": 0.3354, "step": 9369 }, { "epoch": 0.47, "learning_rate": 1.1520723263582456e-05, "loss": 0.5109, "step": 9370 }, { "epoch": 0.47, "learning_rate": 1.151912484261932e-05, "loss": 0.6867, "step": 9371 }, { "epoch": 0.47, "learning_rate": 1.1517526381925464e-05, "loss": 0.4517, "step": 9372 }, { "epoch": 0.47, "learning_rate": 1.151592788154269e-05, "loss": 0.3361, "step": 9373 }, { "epoch": 0.47, "learning_rate": 1.151432934151281e-05, "loss": 0.1254, "step": 9374 }, { "epoch": 0.47, "learning_rate": 1.1512730761877628e-05, "loss": 0.5164, "step": 9375 }, { "epoch": 0.47, "learning_rate": 1.1511132142678956e-05, "loss": 0.7512, "step": 9376 }, { "epoch": 0.47, "learning_rate": 1.1509533483958599e-05, "loss": 0.5441, "step": 9377 }, { "epoch": 0.47, "learning_rate": 1.1507934785758374e-05, "loss": 0.47, "step": 9378 }, { "epoch": 0.47, "learning_rate": 1.150633604812009e-05, "loss": 0.4785, "step": 9379 }, { "epoch": 0.47, "learning_rate": 1.1504737271085556e-05, "loss": 0.3409, "step": 9380 }, { "epoch": 0.47, "learning_rate": 1.1503138454696594e-05, "loss": 0.3232, "step": 9381 }, { "epoch": 0.47, "learning_rate": 1.1501539598995014e-05, "loss": 0.326, "step": 9382 }, { "epoch": 0.47, "learning_rate": 1.149994070402263e-05, "loss": 0.3529, "step": 9383 }, { "epoch": 0.47, "learning_rate": 1.149834176982126e-05, "loss": 0.224, "step": 9384 }, { "epoch": 0.47, "learning_rate": 1.1496742796432726e-05, "loss": 0.3739, "step": 9385 }, { "epoch": 0.47, "learning_rate": 1.1495143783898844e-05, "loss": 0.3829, "step": 9386 }, { "epoch": 0.47, "learning_rate": 1.1493544732261434e-05, "loss": 0.408, "step": 9387 }, { "epoch": 0.47, "learning_rate": 1.1491945641562321e-05, "loss": 0.3217, "step": 9388 }, { "epoch": 0.47, "learning_rate": 1.149034651184332e-05, "loss": 0.3734, "step": 9389 }, { "epoch": 0.47, "learning_rate": 1.148874734314626e-05, "loss": 0.2164, "step": 9390 }, { "epoch": 0.47, "learning_rate": 1.1487148135512963e-05, "loss": 0.3702, "step": 9391 }, { "epoch": 0.47, "learning_rate": 1.1485548888985252e-05, "loss": 0.2746, "step": 9392 }, { "epoch": 0.47, "learning_rate": 1.1483949603604958e-05, "loss": 0.2983, "step": 9393 }, { "epoch": 0.47, "learning_rate": 1.1482350279413908e-05, "loss": 0.4634, "step": 9394 }, { "epoch": 0.47, "learning_rate": 1.1480750916453923e-05, "loss": 0.2253, "step": 9395 }, { "epoch": 0.47, "learning_rate": 1.147915151476684e-05, "loss": 0.1693, "step": 9396 }, { "epoch": 0.47, "learning_rate": 1.1477552074394487e-05, "loss": 0.4479, "step": 9397 }, { "epoch": 0.47, "learning_rate": 1.1475952595378691e-05, "loss": 0.4055, "step": 9398 }, { "epoch": 0.47, "learning_rate": 1.1474353077761292e-05, "loss": 0.4417, "step": 9399 }, { "epoch": 0.47, "learning_rate": 1.1472753521584116e-05, "loss": 0.6356, "step": 9400 }, { "epoch": 0.47, "learning_rate": 1.1471153926889002e-05, "loss": 0.4226, "step": 9401 }, { "epoch": 0.47, "learning_rate": 1.1469554293717784e-05, "loss": 0.3483, "step": 9402 }, { "epoch": 0.47, "learning_rate": 1.1467954622112301e-05, "loss": 0.4164, "step": 9403 }, { "epoch": 0.47, "learning_rate": 1.1466354912114385e-05, "loss": 0.5863, "step": 9404 }, { "epoch": 0.47, "learning_rate": 1.1464755163765878e-05, "loss": 0.3303, "step": 9405 }, { "epoch": 0.47, "learning_rate": 1.1463155377108617e-05, "loss": 0.5555, "step": 9406 }, { "epoch": 0.47, "learning_rate": 1.1461555552184442e-05, "loss": 0.497, "step": 9407 }, { "epoch": 0.47, "learning_rate": 1.1459955689035198e-05, "loss": 0.4358, "step": 9408 }, { "epoch": 0.47, "learning_rate": 1.1458355787702725e-05, "loss": 0.4219, "step": 9409 }, { "epoch": 0.47, "learning_rate": 1.1456755848228867e-05, "loss": 0.4143, "step": 9410 }, { "epoch": 0.47, "learning_rate": 1.1455155870655465e-05, "loss": 0.2351, "step": 9411 }, { "epoch": 0.47, "learning_rate": 1.145355585502437e-05, "loss": 0.1983, "step": 9412 }, { "epoch": 0.47, "learning_rate": 1.1451955801377426e-05, "loss": 0.4487, "step": 9413 }, { "epoch": 0.47, "learning_rate": 1.1450355709756476e-05, "loss": 0.3809, "step": 9414 }, { "epoch": 0.47, "learning_rate": 1.1448755580203374e-05, "loss": 0.4702, "step": 9415 }, { "epoch": 0.47, "learning_rate": 1.1447155412759966e-05, "loss": 0.4152, "step": 9416 }, { "epoch": 0.47, "learning_rate": 1.1445555207468106e-05, "loss": 0.2356, "step": 9417 }, { "epoch": 0.47, "learning_rate": 1.1443954964369643e-05, "loss": 0.2932, "step": 9418 }, { "epoch": 0.47, "learning_rate": 1.144235468350643e-05, "loss": 0.3739, "step": 9419 }, { "epoch": 0.47, "learning_rate": 1.1440754364920316e-05, "loss": 0.4051, "step": 9420 }, { "epoch": 0.47, "learning_rate": 1.143915400865316e-05, "loss": 0.179, "step": 9421 }, { "epoch": 0.47, "learning_rate": 1.1437553614746816e-05, "loss": 0.3915, "step": 9422 }, { "epoch": 0.47, "learning_rate": 1.143595318324314e-05, "loss": 0.5399, "step": 9423 }, { "epoch": 0.47, "learning_rate": 1.1434352714183989e-05, "loss": 0.3112, "step": 9424 }, { "epoch": 0.47, "learning_rate": 1.1432752207611222e-05, "loss": 0.2452, "step": 9425 }, { "epoch": 0.47, "learning_rate": 1.1431151663566698e-05, "loss": 0.2349, "step": 9426 }, { "epoch": 0.47, "learning_rate": 1.1429551082092274e-05, "loss": 0.4403, "step": 9427 }, { "epoch": 0.47, "learning_rate": 1.1427950463229818e-05, "loss": 0.4442, "step": 9428 }, { "epoch": 0.47, "learning_rate": 1.1426349807021186e-05, "loss": 0.2761, "step": 9429 }, { "epoch": 0.47, "learning_rate": 1.1424749113508244e-05, "loss": 0.1874, "step": 9430 }, { "epoch": 0.47, "learning_rate": 1.1423148382732854e-05, "loss": 0.4105, "step": 9431 }, { "epoch": 0.47, "learning_rate": 1.1421547614736881e-05, "loss": 0.458, "step": 9432 }, { "epoch": 0.47, "learning_rate": 1.1419946809562195e-05, "loss": 0.6561, "step": 9433 }, { "epoch": 0.47, "learning_rate": 1.1418345967250655e-05, "loss": 0.3674, "step": 9434 }, { "epoch": 0.47, "learning_rate": 1.1416745087844139e-05, "loss": 0.3761, "step": 9435 }, { "epoch": 0.47, "learning_rate": 1.1415144171384509e-05, "loss": 0.5086, "step": 9436 }, { "epoch": 0.47, "learning_rate": 1.1413543217913635e-05, "loss": 0.2838, "step": 9437 }, { "epoch": 0.47, "learning_rate": 1.1411942227473391e-05, "loss": 0.6753, "step": 9438 }, { "epoch": 0.47, "learning_rate": 1.1410341200105646e-05, "loss": 0.4802, "step": 9439 }, { "epoch": 0.47, "learning_rate": 1.1408740135852276e-05, "loss": 0.3402, "step": 9440 }, { "epoch": 0.47, "learning_rate": 1.1407139034755152e-05, "loss": 0.3155, "step": 9441 }, { "epoch": 0.47, "learning_rate": 1.1405537896856151e-05, "loss": 0.4333, "step": 9442 }, { "epoch": 0.47, "learning_rate": 1.1403936722197145e-05, "loss": 0.4051, "step": 9443 }, { "epoch": 0.47, "learning_rate": 1.1402335510820015e-05, "loss": 0.3928, "step": 9444 }, { "epoch": 0.47, "learning_rate": 1.1400734262766636e-05, "loss": 0.3907, "step": 9445 }, { "epoch": 0.47, "learning_rate": 1.1399132978078887e-05, "loss": 0.309, "step": 9446 }, { "epoch": 0.47, "learning_rate": 1.1397531656798646e-05, "loss": 0.1567, "step": 9447 }, { "epoch": 0.47, "learning_rate": 1.1395930298967794e-05, "loss": 0.3071, "step": 9448 }, { "epoch": 0.47, "learning_rate": 1.1394328904628217e-05, "loss": 0.2527, "step": 9449 }, { "epoch": 0.47, "learning_rate": 1.139272747382179e-05, "loss": 0.2681, "step": 9450 }, { "epoch": 0.47, "learning_rate": 1.1391126006590402e-05, "loss": 0.4565, "step": 9451 }, { "epoch": 0.47, "learning_rate": 1.1389524502975937e-05, "loss": 0.4213, "step": 9452 }, { "epoch": 0.47, "learning_rate": 1.1387922963020274e-05, "loss": 0.4769, "step": 9453 }, { "epoch": 0.47, "learning_rate": 1.1386321386765309e-05, "loss": 0.2869, "step": 9454 }, { "epoch": 0.47, "learning_rate": 1.1384719774252922e-05, "loss": 0.5085, "step": 9455 }, { "epoch": 0.47, "learning_rate": 1.1383118125525003e-05, "loss": 0.2868, "step": 9456 }, { "epoch": 0.47, "learning_rate": 1.1381516440623439e-05, "loss": 0.4434, "step": 9457 }, { "epoch": 0.47, "learning_rate": 1.1379914719590124e-05, "loss": 0.2273, "step": 9458 }, { "epoch": 0.47, "learning_rate": 1.1378312962466945e-05, "loss": 0.3198, "step": 9459 }, { "epoch": 0.47, "learning_rate": 1.1376711169295798e-05, "loss": 0.2571, "step": 9460 }, { "epoch": 0.47, "learning_rate": 1.1375109340118574e-05, "loss": 0.3058, "step": 9461 }, { "epoch": 0.47, "learning_rate": 1.1373507474977164e-05, "loss": 0.5665, "step": 9462 }, { "epoch": 0.47, "learning_rate": 1.1371905573913464e-05, "loss": 0.3966, "step": 9463 }, { "epoch": 0.47, "learning_rate": 1.1370303636969373e-05, "loss": 0.4508, "step": 9464 }, { "epoch": 0.47, "learning_rate": 1.1368701664186785e-05, "loss": 0.4148, "step": 9465 }, { "epoch": 0.47, "learning_rate": 1.1367099655607598e-05, "loss": 0.3778, "step": 9466 }, { "epoch": 0.47, "learning_rate": 1.136549761127371e-05, "loss": 0.2526, "step": 9467 }, { "epoch": 0.47, "learning_rate": 1.136389553122702e-05, "loss": 0.2911, "step": 9468 }, { "epoch": 0.47, "learning_rate": 1.1362293415509426e-05, "loss": 0.5166, "step": 9469 }, { "epoch": 0.47, "learning_rate": 1.1360691264162837e-05, "loss": 0.2496, "step": 9470 }, { "epoch": 0.47, "learning_rate": 1.1359089077229146e-05, "loss": 0.3614, "step": 9471 }, { "epoch": 0.47, "learning_rate": 1.1357486854750264e-05, "loss": 0.5792, "step": 9472 }, { "epoch": 0.47, "learning_rate": 1.1355884596768088e-05, "loss": 0.2102, "step": 9473 }, { "epoch": 0.47, "learning_rate": 1.1354282303324526e-05, "loss": 0.4529, "step": 9474 }, { "epoch": 0.47, "learning_rate": 1.1352679974461485e-05, "loss": 0.3046, "step": 9475 }, { "epoch": 0.47, "learning_rate": 1.1351077610220873e-05, "loss": 0.5229, "step": 9476 }, { "epoch": 0.47, "learning_rate": 1.1349475210644595e-05, "loss": 0.5207, "step": 9477 }, { "epoch": 0.47, "learning_rate": 1.134787277577456e-05, "loss": 0.264, "step": 9478 }, { "epoch": 0.47, "learning_rate": 1.1346270305652677e-05, "loss": 0.3729, "step": 9479 }, { "epoch": 0.47, "learning_rate": 1.1344667800320859e-05, "loss": 0.5282, "step": 9480 }, { "epoch": 0.47, "learning_rate": 1.1343065259821015e-05, "loss": 0.2941, "step": 9481 }, { "epoch": 0.47, "learning_rate": 1.1341462684195057e-05, "loss": 0.5569, "step": 9482 }, { "epoch": 0.47, "learning_rate": 1.1339860073484901e-05, "loss": 0.4075, "step": 9483 }, { "epoch": 0.47, "learning_rate": 1.133825742773246e-05, "loss": 0.5743, "step": 9484 }, { "epoch": 0.47, "learning_rate": 1.1336654746979647e-05, "loss": 0.3557, "step": 9485 }, { "epoch": 0.47, "learning_rate": 1.133505203126838e-05, "loss": 0.3665, "step": 9486 }, { "epoch": 0.47, "learning_rate": 1.1333449280640579e-05, "loss": 0.4108, "step": 9487 }, { "epoch": 0.47, "learning_rate": 1.1331846495138154e-05, "loss": 0.4941, "step": 9488 }, { "epoch": 0.47, "learning_rate": 1.1330243674803029e-05, "loss": 0.2715, "step": 9489 }, { "epoch": 0.47, "learning_rate": 1.1328640819677129e-05, "loss": 0.5452, "step": 9490 }, { "epoch": 0.47, "learning_rate": 1.1327037929802362e-05, "loss": 0.583, "step": 9491 }, { "epoch": 0.47, "learning_rate": 1.132543500522066e-05, "loss": 0.4545, "step": 9492 }, { "epoch": 0.47, "learning_rate": 1.1323832045973942e-05, "loss": 0.2408, "step": 9493 }, { "epoch": 0.47, "learning_rate": 1.1322229052104126e-05, "loss": 0.3527, "step": 9494 }, { "epoch": 0.47, "learning_rate": 1.1320626023653143e-05, "loss": 0.2313, "step": 9495 }, { "epoch": 0.47, "learning_rate": 1.1319022960662915e-05, "loss": 0.426, "step": 9496 }, { "epoch": 0.47, "learning_rate": 1.1317419863175372e-05, "loss": 0.6148, "step": 9497 }, { "epoch": 0.47, "learning_rate": 1.1315816731232433e-05, "loss": 0.4065, "step": 9498 }, { "epoch": 0.47, "learning_rate": 1.1314213564876038e-05, "loss": 0.2052, "step": 9499 }, { "epoch": 0.47, "learning_rate": 1.1312610364148105e-05, "loss": 0.6666, "step": 9500 }, { "epoch": 0.47, "learning_rate": 1.1311007129090565e-05, "loss": 0.4582, "step": 9501 }, { "epoch": 0.47, "learning_rate": 1.1309403859745354e-05, "loss": 0.7234, "step": 9502 }, { "epoch": 0.47, "learning_rate": 1.1307800556154398e-05, "loss": 0.2908, "step": 9503 }, { "epoch": 0.47, "learning_rate": 1.1306197218359634e-05, "loss": 0.3969, "step": 9504 }, { "epoch": 0.47, "learning_rate": 1.130459384640299e-05, "loss": 0.6263, "step": 9505 }, { "epoch": 0.47, "learning_rate": 1.1302990440326403e-05, "loss": 0.3017, "step": 9506 }, { "epoch": 0.47, "learning_rate": 1.1301387000171806e-05, "loss": 0.3828, "step": 9507 }, { "epoch": 0.47, "learning_rate": 1.1299783525981138e-05, "loss": 0.397, "step": 9508 }, { "epoch": 0.47, "learning_rate": 1.1298180017796337e-05, "loss": 0.4534, "step": 9509 }, { "epoch": 0.47, "learning_rate": 1.1296576475659333e-05, "loss": 0.1977, "step": 9510 }, { "epoch": 0.47, "learning_rate": 1.1294972899612073e-05, "loss": 0.412, "step": 9511 }, { "epoch": 0.47, "learning_rate": 1.1293369289696492e-05, "loss": 0.4179, "step": 9512 }, { "epoch": 0.48, "learning_rate": 1.129176564595453e-05, "loss": 0.5204, "step": 9513 }, { "epoch": 0.48, "learning_rate": 1.1290161968428133e-05, "loss": 0.304, "step": 9514 }, { "epoch": 0.48, "learning_rate": 1.1288558257159239e-05, "loss": 0.5754, "step": 9515 }, { "epoch": 0.48, "learning_rate": 1.1286954512189788e-05, "loss": 0.4236, "step": 9516 }, { "epoch": 0.48, "learning_rate": 1.1285350733561729e-05, "loss": 0.2941, "step": 9517 }, { "epoch": 0.48, "learning_rate": 1.1283746921317004e-05, "loss": 0.4854, "step": 9518 }, { "epoch": 0.48, "learning_rate": 1.1282143075497562e-05, "loss": 0.3426, "step": 9519 }, { "epoch": 0.48, "learning_rate": 1.1280539196145346e-05, "loss": 0.283, "step": 9520 }, { "epoch": 0.48, "learning_rate": 1.1278935283302304e-05, "loss": 0.542, "step": 9521 }, { "epoch": 0.48, "learning_rate": 1.1277331337010384e-05, "loss": 0.2653, "step": 9522 }, { "epoch": 0.48, "learning_rate": 1.1275727357311537e-05, "loss": 0.4187, "step": 9523 }, { "epoch": 0.48, "learning_rate": 1.1274123344247712e-05, "loss": 0.3164, "step": 9524 }, { "epoch": 0.48, "learning_rate": 1.1272519297860862e-05, "loss": 0.3545, "step": 9525 }, { "epoch": 0.48, "learning_rate": 1.1270915218192933e-05, "loss": 0.6129, "step": 9526 }, { "epoch": 0.48, "learning_rate": 1.1269311105285883e-05, "loss": 0.5795, "step": 9527 }, { "epoch": 0.48, "learning_rate": 1.1267706959181663e-05, "loss": 0.3536, "step": 9528 }, { "epoch": 0.48, "learning_rate": 1.1266102779922228e-05, "loss": 0.2654, "step": 9529 }, { "epoch": 0.48, "learning_rate": 1.126449856754953e-05, "loss": 0.4996, "step": 9530 }, { "epoch": 0.48, "learning_rate": 1.1262894322105532e-05, "loss": 0.3918, "step": 9531 }, { "epoch": 0.48, "learning_rate": 1.1261290043632188e-05, "loss": 0.477, "step": 9532 }, { "epoch": 0.48, "learning_rate": 1.1259685732171453e-05, "loss": 0.1975, "step": 9533 }, { "epoch": 0.48, "learning_rate": 1.1258081387765288e-05, "loss": 0.4238, "step": 9534 }, { "epoch": 0.48, "learning_rate": 1.125647701045565e-05, "loss": 0.5777, "step": 9535 }, { "epoch": 0.48, "learning_rate": 1.1254872600284502e-05, "loss": 0.5199, "step": 9536 }, { "epoch": 0.48, "learning_rate": 1.1253268157293807e-05, "loss": 0.2943, "step": 9537 }, { "epoch": 0.48, "learning_rate": 1.1251663681525521e-05, "loss": 0.3205, "step": 9538 }, { "epoch": 0.48, "learning_rate": 1.1250059173021616e-05, "loss": 0.2274, "step": 9539 }, { "epoch": 0.48, "learning_rate": 1.1248454631824049e-05, "loss": 0.1575, "step": 9540 }, { "epoch": 0.48, "learning_rate": 1.1246850057974788e-05, "loss": 0.3026, "step": 9541 }, { "epoch": 0.48, "learning_rate": 1.1245245451515792e-05, "loss": 0.2657, "step": 9542 }, { "epoch": 0.48, "learning_rate": 1.1243640812489034e-05, "loss": 0.2259, "step": 9543 }, { "epoch": 0.48, "learning_rate": 1.1242036140936479e-05, "loss": 0.2023, "step": 9544 }, { "epoch": 0.48, "learning_rate": 1.1240431436900097e-05, "loss": 0.5865, "step": 9545 }, { "epoch": 0.48, "learning_rate": 1.1238826700421853e-05, "loss": 0.4028, "step": 9546 }, { "epoch": 0.48, "learning_rate": 1.1237221931543722e-05, "loss": 0.1216, "step": 9547 }, { "epoch": 0.48, "learning_rate": 1.123561713030767e-05, "loss": 0.4301, "step": 9548 }, { "epoch": 0.48, "learning_rate": 1.1234012296755671e-05, "loss": 0.3564, "step": 9549 }, { "epoch": 0.48, "learning_rate": 1.1232407430929698e-05, "loss": 0.617, "step": 9550 }, { "epoch": 0.48, "learning_rate": 1.1230802532871724e-05, "loss": 0.4968, "step": 9551 }, { "epoch": 0.48, "learning_rate": 1.1229197602623716e-05, "loss": 0.343, "step": 9552 }, { "epoch": 0.48, "learning_rate": 1.1227592640227658e-05, "loss": 0.3545, "step": 9553 }, { "epoch": 0.48, "learning_rate": 1.1225987645725525e-05, "loss": 0.5067, "step": 9554 }, { "epoch": 0.48, "learning_rate": 1.1224382619159286e-05, "loss": 0.486, "step": 9555 }, { "epoch": 0.48, "learning_rate": 1.1222777560570925e-05, "loss": 0.4784, "step": 9556 }, { "epoch": 0.48, "learning_rate": 1.1221172470002421e-05, "loss": 0.3713, "step": 9557 }, { "epoch": 0.48, "learning_rate": 1.1219567347495748e-05, "loss": 0.3928, "step": 9558 }, { "epoch": 0.48, "learning_rate": 1.1217962193092888e-05, "loss": 0.4182, "step": 9559 }, { "epoch": 0.48, "learning_rate": 1.1216357006835822e-05, "loss": 0.1462, "step": 9560 }, { "epoch": 0.48, "learning_rate": 1.1214751788766534e-05, "loss": 0.5393, "step": 9561 }, { "epoch": 0.48, "learning_rate": 1.1213146538927e-05, "loss": 0.5067, "step": 9562 }, { "epoch": 0.48, "learning_rate": 1.121154125735921e-05, "loss": 0.4671, "step": 9563 }, { "epoch": 0.48, "learning_rate": 1.1209935944105147e-05, "loss": 0.3967, "step": 9564 }, { "epoch": 0.48, "learning_rate": 1.1208330599206793e-05, "loss": 0.5911, "step": 9565 }, { "epoch": 0.48, "learning_rate": 1.1206725222706137e-05, "loss": 0.1601, "step": 9566 }, { "epoch": 0.48, "learning_rate": 1.1205119814645158e-05, "loss": 0.3903, "step": 9567 }, { "epoch": 0.48, "learning_rate": 1.1203514375065852e-05, "loss": 0.3873, "step": 9568 }, { "epoch": 0.48, "learning_rate": 1.1201908904010201e-05, "loss": 0.1831, "step": 9569 }, { "epoch": 0.48, "learning_rate": 1.1200303401520199e-05, "loss": 0.38, "step": 9570 }, { "epoch": 0.48, "learning_rate": 1.119869786763783e-05, "loss": 0.2124, "step": 9571 }, { "epoch": 0.48, "learning_rate": 1.1197092302405095e-05, "loss": 0.2536, "step": 9572 }, { "epoch": 0.48, "learning_rate": 1.1195486705863976e-05, "loss": 0.5298, "step": 9573 }, { "epoch": 0.48, "learning_rate": 1.1193881078056466e-05, "loss": 0.4881, "step": 9574 }, { "epoch": 0.48, "learning_rate": 1.119227541902456e-05, "loss": 0.5116, "step": 9575 }, { "epoch": 0.48, "learning_rate": 1.1190669728810255e-05, "loss": 0.602, "step": 9576 }, { "epoch": 0.48, "learning_rate": 1.1189064007455544e-05, "loss": 0.4223, "step": 9577 }, { "epoch": 0.48, "learning_rate": 1.1187458255002414e-05, "loss": 0.3751, "step": 9578 }, { "epoch": 0.48, "learning_rate": 1.1185852471492875e-05, "loss": 0.2506, "step": 9579 }, { "epoch": 0.48, "learning_rate": 1.1184246656968916e-05, "loss": 0.3841, "step": 9580 }, { "epoch": 0.48, "learning_rate": 1.1182640811472535e-05, "loss": 0.3788, "step": 9581 }, { "epoch": 0.48, "learning_rate": 1.1181034935045735e-05, "loss": 0.4065, "step": 9582 }, { "epoch": 0.48, "learning_rate": 1.1179429027730513e-05, "loss": 0.1939, "step": 9583 }, { "epoch": 0.48, "learning_rate": 1.1177823089568868e-05, "loss": 0.3033, "step": 9584 }, { "epoch": 0.48, "learning_rate": 1.1176217120602804e-05, "loss": 0.5775, "step": 9585 }, { "epoch": 0.48, "learning_rate": 1.1174611120874322e-05, "loss": 0.3952, "step": 9586 }, { "epoch": 0.48, "learning_rate": 1.1173005090425424e-05, "loss": 0.4285, "step": 9587 }, { "epoch": 0.48, "learning_rate": 1.1171399029298114e-05, "loss": 0.2385, "step": 9588 }, { "epoch": 0.48, "learning_rate": 1.1169792937534397e-05, "loss": 0.4964, "step": 9589 }, { "epoch": 0.48, "learning_rate": 1.1168186815176278e-05, "loss": 0.6502, "step": 9590 }, { "epoch": 0.48, "learning_rate": 1.1166580662265765e-05, "loss": 0.3964, "step": 9591 }, { "epoch": 0.48, "learning_rate": 1.1164974478844858e-05, "loss": 0.3467, "step": 9592 }, { "epoch": 0.48, "learning_rate": 1.1163368264955572e-05, "loss": 0.3845, "step": 9593 }, { "epoch": 0.48, "learning_rate": 1.1161762020639913e-05, "loss": 0.5204, "step": 9594 }, { "epoch": 0.48, "learning_rate": 1.1160155745939889e-05, "loss": 0.28, "step": 9595 }, { "epoch": 0.48, "learning_rate": 1.115854944089751e-05, "loss": 0.3299, "step": 9596 }, { "epoch": 0.48, "learning_rate": 1.1156943105554789e-05, "loss": 0.3916, "step": 9597 }, { "epoch": 0.48, "learning_rate": 1.1155336739953739e-05, "loss": 0.4572, "step": 9598 }, { "epoch": 0.48, "learning_rate": 1.1153730344136366e-05, "loss": 0.4826, "step": 9599 }, { "epoch": 0.48, "learning_rate": 1.1152123918144688e-05, "loss": 0.2159, "step": 9600 }, { "epoch": 0.48, "learning_rate": 1.115051746202072e-05, "loss": 0.3761, "step": 9601 }, { "epoch": 0.48, "learning_rate": 1.1148910975806475e-05, "loss": 0.198, "step": 9602 }, { "epoch": 0.48, "learning_rate": 1.1147304459543965e-05, "loss": 0.1189, "step": 9603 }, { "epoch": 0.48, "learning_rate": 1.114569791327521e-05, "loss": 0.3509, "step": 9604 }, { "epoch": 0.48, "learning_rate": 1.1144091337042229e-05, "loss": 0.2563, "step": 9605 }, { "epoch": 0.48, "learning_rate": 1.1142484730887036e-05, "loss": 0.3625, "step": 9606 }, { "epoch": 0.48, "learning_rate": 1.114087809485165e-05, "loss": 0.5464, "step": 9607 }, { "epoch": 0.48, "learning_rate": 1.1139271428978094e-05, "loss": 0.2165, "step": 9608 }, { "epoch": 0.48, "learning_rate": 1.1137664733308384e-05, "loss": 0.2451, "step": 9609 }, { "epoch": 0.48, "learning_rate": 1.1136058007884542e-05, "loss": 0.2303, "step": 9610 }, { "epoch": 0.48, "learning_rate": 1.1134451252748595e-05, "loss": 0.6039, "step": 9611 }, { "epoch": 0.48, "learning_rate": 1.113284446794256e-05, "loss": 0.4484, "step": 9612 }, { "epoch": 0.48, "learning_rate": 1.1131237653508462e-05, "loss": 0.4158, "step": 9613 }, { "epoch": 0.48, "learning_rate": 1.1129630809488323e-05, "loss": 0.2818, "step": 9614 }, { "epoch": 0.48, "learning_rate": 1.112802393592417e-05, "loss": 0.291, "step": 9615 }, { "epoch": 0.48, "learning_rate": 1.112641703285803e-05, "loss": 0.4578, "step": 9616 }, { "epoch": 0.48, "learning_rate": 1.1124810100331925e-05, "loss": 0.4661, "step": 9617 }, { "epoch": 0.48, "learning_rate": 1.1123203138387888e-05, "loss": 0.3531, "step": 9618 }, { "epoch": 0.48, "learning_rate": 1.1121596147067941e-05, "loss": 0.2767, "step": 9619 }, { "epoch": 0.48, "learning_rate": 1.1119989126414116e-05, "loss": 0.3728, "step": 9620 }, { "epoch": 0.48, "learning_rate": 1.1118382076468446e-05, "loss": 0.2135, "step": 9621 }, { "epoch": 0.48, "learning_rate": 1.1116774997272954e-05, "loss": 0.3407, "step": 9622 }, { "epoch": 0.48, "learning_rate": 1.1115167888869678e-05, "loss": 0.1455, "step": 9623 }, { "epoch": 0.48, "learning_rate": 1.1113560751300642e-05, "loss": 0.459, "step": 9624 }, { "epoch": 0.48, "learning_rate": 1.1111953584607886e-05, "loss": 0.3164, "step": 9625 }, { "epoch": 0.48, "learning_rate": 1.1110346388833441e-05, "loss": 0.2152, "step": 9626 }, { "epoch": 0.48, "learning_rate": 1.1108739164019342e-05, "loss": 0.5321, "step": 9627 }, { "epoch": 0.48, "learning_rate": 1.1107131910207618e-05, "loss": 0.2933, "step": 9628 }, { "epoch": 0.48, "learning_rate": 1.1105524627440311e-05, "loss": 0.1894, "step": 9629 }, { "epoch": 0.48, "learning_rate": 1.1103917315759456e-05, "loss": 0.2073, "step": 9630 }, { "epoch": 0.48, "learning_rate": 1.110230997520709e-05, "loss": 0.2641, "step": 9631 }, { "epoch": 0.48, "learning_rate": 1.110070260582525e-05, "loss": 0.4458, "step": 9632 }, { "epoch": 0.48, "learning_rate": 1.109909520765597e-05, "loss": 0.4792, "step": 9633 }, { "epoch": 0.48, "learning_rate": 1.1097487780741299e-05, "loss": 0.376, "step": 9634 }, { "epoch": 0.48, "learning_rate": 1.1095880325123272e-05, "loss": 0.3577, "step": 9635 }, { "epoch": 0.48, "learning_rate": 1.109427284084393e-05, "loss": 0.351, "step": 9636 }, { "epoch": 0.48, "learning_rate": 1.1092665327945315e-05, "loss": 0.3982, "step": 9637 }, { "epoch": 0.48, "learning_rate": 1.1091057786469472e-05, "loss": 0.2733, "step": 9638 }, { "epoch": 0.48, "learning_rate": 1.1089450216458442e-05, "loss": 0.5874, "step": 9639 }, { "epoch": 0.48, "learning_rate": 1.1087842617954263e-05, "loss": 0.3677, "step": 9640 }, { "epoch": 0.48, "learning_rate": 1.1086234990998989e-05, "loss": 0.2595, "step": 9641 }, { "epoch": 0.48, "learning_rate": 1.1084627335634658e-05, "loss": 0.3593, "step": 9642 }, { "epoch": 0.48, "learning_rate": 1.1083019651903327e-05, "loss": 0.4753, "step": 9643 }, { "epoch": 0.48, "learning_rate": 1.1081411939847028e-05, "loss": 0.262, "step": 9644 }, { "epoch": 0.48, "learning_rate": 1.1079804199507819e-05, "loss": 0.4554, "step": 9645 }, { "epoch": 0.48, "learning_rate": 1.1078196430927748e-05, "loss": 0.2914, "step": 9646 }, { "epoch": 0.48, "learning_rate": 1.1076588634148857e-05, "loss": 0.4577, "step": 9647 }, { "epoch": 0.48, "learning_rate": 1.1074980809213202e-05, "loss": 0.3481, "step": 9648 }, { "epoch": 0.48, "learning_rate": 1.1073372956162834e-05, "loss": 0.2095, "step": 9649 }, { "epoch": 0.48, "learning_rate": 1.1071765075039801e-05, "loss": 0.341, "step": 9650 }, { "epoch": 0.48, "learning_rate": 1.1070157165886155e-05, "loss": 0.3144, "step": 9651 }, { "epoch": 0.48, "learning_rate": 1.1068549228743953e-05, "loss": 0.3612, "step": 9652 }, { "epoch": 0.48, "learning_rate": 1.106694126365524e-05, "loss": 0.4808, "step": 9653 }, { "epoch": 0.48, "learning_rate": 1.106533327066208e-05, "loss": 0.2491, "step": 9654 }, { "epoch": 0.48, "learning_rate": 1.1063725249806524e-05, "loss": 0.3703, "step": 9655 }, { "epoch": 0.48, "learning_rate": 1.1062117201130625e-05, "loss": 0.3024, "step": 9656 }, { "epoch": 0.48, "learning_rate": 1.1060509124676441e-05, "loss": 0.3369, "step": 9657 }, { "epoch": 0.48, "learning_rate": 1.105890102048603e-05, "loss": 0.5276, "step": 9658 }, { "epoch": 0.48, "learning_rate": 1.1057292888601451e-05, "loss": 0.2881, "step": 9659 }, { "epoch": 0.48, "learning_rate": 1.1055684729064762e-05, "loss": 0.3317, "step": 9660 }, { "epoch": 0.48, "learning_rate": 1.1054076541918021e-05, "loss": 0.1882, "step": 9661 }, { "epoch": 0.48, "learning_rate": 1.105246832720329e-05, "loss": 0.4199, "step": 9662 }, { "epoch": 0.48, "learning_rate": 1.105086008496263e-05, "loss": 0.4592, "step": 9663 }, { "epoch": 0.48, "learning_rate": 1.1049251815238097e-05, "loss": 0.2939, "step": 9664 }, { "epoch": 0.48, "learning_rate": 1.1047643518071756e-05, "loss": 0.5181, "step": 9665 }, { "epoch": 0.48, "learning_rate": 1.1046035193505676e-05, "loss": 0.305, "step": 9666 }, { "epoch": 0.48, "learning_rate": 1.104442684158191e-05, "loss": 0.4205, "step": 9667 }, { "epoch": 0.48, "learning_rate": 1.104281846234253e-05, "loss": 0.2596, "step": 9668 }, { "epoch": 0.48, "learning_rate": 1.10412100558296e-05, "loss": 0.3081, "step": 9669 }, { "epoch": 0.48, "learning_rate": 1.1039601622085184e-05, "loss": 0.3176, "step": 9670 }, { "epoch": 0.48, "learning_rate": 1.103799316115135e-05, "loss": 0.3568, "step": 9671 }, { "epoch": 0.48, "learning_rate": 1.1036384673070162e-05, "loss": 0.1863, "step": 9672 }, { "epoch": 0.48, "learning_rate": 1.1034776157883693e-05, "loss": 0.4573, "step": 9673 }, { "epoch": 0.48, "learning_rate": 1.1033167615634008e-05, "loss": 0.6388, "step": 9674 }, { "epoch": 0.48, "learning_rate": 1.1031559046363178e-05, "loss": 0.3315, "step": 9675 }, { "epoch": 0.48, "learning_rate": 1.1029950450113272e-05, "loss": 0.3912, "step": 9676 }, { "epoch": 0.48, "learning_rate": 1.1028341826926358e-05, "loss": 0.3172, "step": 9677 }, { "epoch": 0.48, "learning_rate": 1.1026733176844515e-05, "loss": 0.411, "step": 9678 }, { "epoch": 0.48, "learning_rate": 1.1025124499909807e-05, "loss": 0.3207, "step": 9679 }, { "epoch": 0.48, "learning_rate": 1.1023515796164312e-05, "loss": 0.3068, "step": 9680 }, { "epoch": 0.48, "learning_rate": 1.1021907065650101e-05, "loss": 0.4292, "step": 9681 }, { "epoch": 0.48, "learning_rate": 1.102029830840925e-05, "loss": 0.2029, "step": 9682 }, { "epoch": 0.48, "learning_rate": 1.1018689524483833e-05, "loss": 0.3604, "step": 9683 }, { "epoch": 0.48, "learning_rate": 1.1017080713915924e-05, "loss": 0.2328, "step": 9684 }, { "epoch": 0.48, "learning_rate": 1.1015471876747606e-05, "loss": 0.4086, "step": 9685 }, { "epoch": 0.48, "learning_rate": 1.1013863013020946e-05, "loss": 0.3268, "step": 9686 }, { "epoch": 0.48, "learning_rate": 1.101225412277803e-05, "loss": 0.3717, "step": 9687 }, { "epoch": 0.48, "learning_rate": 1.101064520606093e-05, "loss": 0.3606, "step": 9688 }, { "epoch": 0.48, "learning_rate": 1.1009036262911731e-05, "loss": 0.0964, "step": 9689 }, { "epoch": 0.48, "learning_rate": 1.1007427293372509e-05, "loss": 0.2454, "step": 9690 }, { "epoch": 0.48, "learning_rate": 1.1005818297485347e-05, "loss": 0.5358, "step": 9691 }, { "epoch": 0.48, "learning_rate": 1.1004209275292322e-05, "loss": 0.3763, "step": 9692 }, { "epoch": 0.48, "learning_rate": 1.100260022683552e-05, "loss": 0.3176, "step": 9693 }, { "epoch": 0.48, "learning_rate": 1.1000991152157023e-05, "loss": 0.4281, "step": 9694 }, { "epoch": 0.48, "learning_rate": 1.0999382051298913e-05, "loss": 0.2749, "step": 9695 }, { "epoch": 0.48, "learning_rate": 1.0997772924303275e-05, "loss": 0.2088, "step": 9696 }, { "epoch": 0.48, "learning_rate": 1.0996163771212191e-05, "loss": 0.6967, "step": 9697 }, { "epoch": 0.48, "learning_rate": 1.0994554592067752e-05, "loss": 0.3726, "step": 9698 }, { "epoch": 0.48, "learning_rate": 1.0992945386912037e-05, "loss": 0.5603, "step": 9699 }, { "epoch": 0.48, "learning_rate": 1.0991336155787138e-05, "loss": 0.4916, "step": 9700 }, { "epoch": 0.48, "learning_rate": 1.0989726898735137e-05, "loss": 0.4236, "step": 9701 }, { "epoch": 0.48, "learning_rate": 1.098811761579813e-05, "loss": 0.415, "step": 9702 }, { "epoch": 0.48, "learning_rate": 1.09865083070182e-05, "loss": 0.2762, "step": 9703 }, { "epoch": 0.48, "learning_rate": 1.0984898972437434e-05, "loss": 0.3991, "step": 9704 }, { "epoch": 0.48, "learning_rate": 1.0983289612097928e-05, "loss": 0.1669, "step": 9705 }, { "epoch": 0.48, "learning_rate": 1.098168022604177e-05, "loss": 0.3586, "step": 9706 }, { "epoch": 0.48, "learning_rate": 1.098007081431105e-05, "loss": 0.41, "step": 9707 }, { "epoch": 0.48, "learning_rate": 1.097846137694786e-05, "loss": 0.2409, "step": 9708 }, { "epoch": 0.48, "learning_rate": 1.0976851913994298e-05, "loss": 0.3932, "step": 9709 }, { "epoch": 0.48, "learning_rate": 1.0975242425492452e-05, "loss": 0.3231, "step": 9710 }, { "epoch": 0.48, "learning_rate": 1.097363291148442e-05, "loss": 0.2672, "step": 9711 }, { "epoch": 0.48, "learning_rate": 1.0972023372012292e-05, "loss": 0.2739, "step": 9712 }, { "epoch": 0.48, "learning_rate": 1.0970413807118166e-05, "loss": 0.5403, "step": 9713 }, { "epoch": 0.49, "learning_rate": 1.0968804216844138e-05, "loss": 0.3847, "step": 9714 }, { "epoch": 0.49, "learning_rate": 1.0967194601232303e-05, "loss": 0.196, "step": 9715 }, { "epoch": 0.49, "learning_rate": 1.0965584960324763e-05, "loss": 0.2619, "step": 9716 }, { "epoch": 0.49, "learning_rate": 1.096397529416361e-05, "loss": 0.1956, "step": 9717 }, { "epoch": 0.49, "learning_rate": 1.0962365602790947e-05, "loss": 0.2151, "step": 9718 }, { "epoch": 0.49, "learning_rate": 1.0960755886248876e-05, "loss": 0.4905, "step": 9719 }, { "epoch": 0.49, "learning_rate": 1.0959146144579487e-05, "loss": 0.4899, "step": 9720 }, { "epoch": 0.49, "learning_rate": 1.0957536377824889e-05, "loss": 0.4772, "step": 9721 }, { "epoch": 0.49, "learning_rate": 1.0955926586027184e-05, "loss": 0.2291, "step": 9722 }, { "epoch": 0.49, "learning_rate": 1.0954316769228468e-05, "loss": 0.405, "step": 9723 }, { "epoch": 0.49, "learning_rate": 1.0952706927470846e-05, "loss": 0.3516, "step": 9724 }, { "epoch": 0.49, "learning_rate": 1.0951097060796424e-05, "loss": 0.3655, "step": 9725 }, { "epoch": 0.49, "learning_rate": 1.0949487169247305e-05, "loss": 0.2463, "step": 9726 }, { "epoch": 0.49, "learning_rate": 1.0947877252865592e-05, "loss": 0.4042, "step": 9727 }, { "epoch": 0.49, "learning_rate": 1.0946267311693393e-05, "loss": 0.3878, "step": 9728 }, { "epoch": 0.49, "learning_rate": 1.0944657345772806e-05, "loss": 0.4183, "step": 9729 }, { "epoch": 0.49, "learning_rate": 1.0943047355145948e-05, "loss": 0.2677, "step": 9730 }, { "epoch": 0.49, "learning_rate": 1.0941437339854919e-05, "loss": 0.2558, "step": 9731 }, { "epoch": 0.49, "learning_rate": 1.0939827299941833e-05, "loss": 0.4042, "step": 9732 }, { "epoch": 0.49, "learning_rate": 1.0938217235448789e-05, "loss": 0.3802, "step": 9733 }, { "epoch": 0.49, "learning_rate": 1.0936607146417907e-05, "loss": 0.5053, "step": 9734 }, { "epoch": 0.49, "learning_rate": 1.0934997032891292e-05, "loss": 0.4287, "step": 9735 }, { "epoch": 0.49, "learning_rate": 1.0933386894911054e-05, "loss": 0.2561, "step": 9736 }, { "epoch": 0.49, "learning_rate": 1.0931776732519304e-05, "loss": 0.3245, "step": 9737 }, { "epoch": 0.49, "learning_rate": 1.0930166545758152e-05, "loss": 0.4248, "step": 9738 }, { "epoch": 0.49, "learning_rate": 1.0928556334669716e-05, "loss": 0.4985, "step": 9739 }, { "epoch": 0.49, "learning_rate": 1.0926946099296098e-05, "loss": 0.4117, "step": 9740 }, { "epoch": 0.49, "learning_rate": 1.0925335839679425e-05, "loss": 0.374, "step": 9741 }, { "epoch": 0.49, "learning_rate": 1.0923725555861806e-05, "loss": 0.2328, "step": 9742 }, { "epoch": 0.49, "learning_rate": 1.092211524788535e-05, "loss": 0.6326, "step": 9743 }, { "epoch": 0.49, "learning_rate": 1.092050491579218e-05, "loss": 0.5017, "step": 9744 }, { "epoch": 0.49, "learning_rate": 1.0918894559624409e-05, "loss": 0.2773, "step": 9745 }, { "epoch": 0.49, "learning_rate": 1.0917284179424158e-05, "loss": 0.2051, "step": 9746 }, { "epoch": 0.49, "learning_rate": 1.0915673775233538e-05, "loss": 0.2832, "step": 9747 }, { "epoch": 0.49, "learning_rate": 1.091406334709467e-05, "loss": 0.4946, "step": 9748 }, { "epoch": 0.49, "learning_rate": 1.091245289504967e-05, "loss": 0.409, "step": 9749 }, { "epoch": 0.49, "learning_rate": 1.0910842419140661e-05, "loss": 0.2589, "step": 9750 }, { "epoch": 0.49, "learning_rate": 1.0909231919409765e-05, "loss": 0.4261, "step": 9751 }, { "epoch": 0.49, "learning_rate": 1.0907621395899093e-05, "loss": 0.2821, "step": 9752 }, { "epoch": 0.49, "learning_rate": 1.0906010848650776e-05, "loss": 0.4003, "step": 9753 }, { "epoch": 0.49, "learning_rate": 1.090440027770693e-05, "loss": 0.4173, "step": 9754 }, { "epoch": 0.49, "learning_rate": 1.090278968310968e-05, "loss": 0.5794, "step": 9755 }, { "epoch": 0.49, "learning_rate": 1.090117906490115e-05, "loss": 0.4037, "step": 9756 }, { "epoch": 0.49, "learning_rate": 1.089956842312346e-05, "loss": 0.3612, "step": 9757 }, { "epoch": 0.49, "learning_rate": 1.0897957757818735e-05, "loss": 0.3495, "step": 9758 }, { "epoch": 0.49, "learning_rate": 1.0896347069029105e-05, "loss": 0.5248, "step": 9759 }, { "epoch": 0.49, "learning_rate": 1.0894736356796689e-05, "loss": 0.41, "step": 9760 }, { "epoch": 0.49, "learning_rate": 1.0893125621163613e-05, "loss": 0.4925, "step": 9761 }, { "epoch": 0.49, "learning_rate": 1.089151486217201e-05, "loss": 0.1849, "step": 9762 }, { "epoch": 0.49, "learning_rate": 1.0889904079864002e-05, "loss": 0.2423, "step": 9763 }, { "epoch": 0.49, "learning_rate": 1.0888293274281718e-05, "loss": 0.325, "step": 9764 }, { "epoch": 0.49, "learning_rate": 1.0886682445467288e-05, "loss": 0.3263, "step": 9765 }, { "epoch": 0.49, "learning_rate": 1.0885071593462839e-05, "loss": 0.5927, "step": 9766 }, { "epoch": 0.49, "learning_rate": 1.0883460718310503e-05, "loss": 0.2553, "step": 9767 }, { "epoch": 0.49, "learning_rate": 1.0881849820052406e-05, "loss": 0.2632, "step": 9768 }, { "epoch": 0.49, "learning_rate": 1.0880238898730684e-05, "loss": 0.2909, "step": 9769 }, { "epoch": 0.49, "learning_rate": 1.0878627954387467e-05, "loss": 0.3125, "step": 9770 }, { "epoch": 0.49, "learning_rate": 1.0877016987064886e-05, "loss": 0.2193, "step": 9771 }, { "epoch": 0.49, "learning_rate": 1.0875405996805075e-05, "loss": 0.3278, "step": 9772 }, { "epoch": 0.49, "learning_rate": 1.0873794983650167e-05, "loss": 0.4265, "step": 9773 }, { "epoch": 0.49, "learning_rate": 1.0872183947642297e-05, "loss": 0.3619, "step": 9774 }, { "epoch": 0.49, "learning_rate": 1.0870572888823593e-05, "loss": 0.4128, "step": 9775 }, { "epoch": 0.49, "learning_rate": 1.08689618072362e-05, "loss": 0.2207, "step": 9776 }, { "epoch": 0.49, "learning_rate": 1.0867350702922247e-05, "loss": 0.3572, "step": 9777 }, { "epoch": 0.49, "learning_rate": 1.0865739575923872e-05, "loss": 0.4252, "step": 9778 }, { "epoch": 0.49, "learning_rate": 1.0864128426283213e-05, "loss": 0.2392, "step": 9779 }, { "epoch": 0.49, "learning_rate": 1.0862517254042409e-05, "loss": 0.4645, "step": 9780 }, { "epoch": 0.49, "learning_rate": 1.0860906059243593e-05, "loss": 0.0627, "step": 9781 }, { "epoch": 0.49, "learning_rate": 1.0859294841928906e-05, "loss": 0.3311, "step": 9782 }, { "epoch": 0.49, "learning_rate": 1.0857683602140492e-05, "loss": 0.2664, "step": 9783 }, { "epoch": 0.49, "learning_rate": 1.0856072339920484e-05, "loss": 0.4524, "step": 9784 }, { "epoch": 0.49, "learning_rate": 1.0854461055311025e-05, "loss": 0.5509, "step": 9785 }, { "epoch": 0.49, "learning_rate": 1.0852849748354257e-05, "loss": 0.5822, "step": 9786 }, { "epoch": 0.49, "learning_rate": 1.085123841909232e-05, "loss": 0.2019, "step": 9787 }, { "epoch": 0.49, "learning_rate": 1.0849627067567357e-05, "loss": 0.3887, "step": 9788 }, { "epoch": 0.49, "learning_rate": 1.0848015693821512e-05, "loss": 0.2478, "step": 9789 }, { "epoch": 0.49, "learning_rate": 1.0846404297896925e-05, "loss": 0.3209, "step": 9790 }, { "epoch": 0.49, "learning_rate": 1.0844792879835745e-05, "loss": 0.2759, "step": 9791 }, { "epoch": 0.49, "learning_rate": 1.0843181439680113e-05, "loss": 0.2613, "step": 9792 }, { "epoch": 0.49, "learning_rate": 1.0841569977472172e-05, "loss": 0.3168, "step": 9793 }, { "epoch": 0.49, "learning_rate": 1.0839958493254075e-05, "loss": 0.1903, "step": 9794 }, { "epoch": 0.49, "learning_rate": 1.0838346987067963e-05, "loss": 0.4417, "step": 9795 }, { "epoch": 0.49, "learning_rate": 1.0836735458955983e-05, "loss": 0.4066, "step": 9796 }, { "epoch": 0.49, "learning_rate": 1.0835123908960282e-05, "loss": 0.2819, "step": 9797 }, { "epoch": 0.49, "learning_rate": 1.083351233712301e-05, "loss": 0.472, "step": 9798 }, { "epoch": 0.49, "learning_rate": 1.0831900743486312e-05, "loss": 0.4164, "step": 9799 }, { "epoch": 0.49, "learning_rate": 1.0830289128092343e-05, "loss": 0.3634, "step": 9800 }, { "epoch": 0.49, "learning_rate": 1.0828677490983246e-05, "loss": 0.3319, "step": 9801 }, { "epoch": 0.49, "learning_rate": 1.0827065832201176e-05, "loss": 0.5764, "step": 9802 }, { "epoch": 0.49, "learning_rate": 1.0825454151788281e-05, "loss": 0.307, "step": 9803 }, { "epoch": 0.49, "learning_rate": 1.0823842449786714e-05, "loss": 0.5083, "step": 9804 }, { "epoch": 0.49, "learning_rate": 1.0822230726238632e-05, "loss": 0.4042, "step": 9805 }, { "epoch": 0.49, "learning_rate": 1.0820618981186175e-05, "loss": 0.3853, "step": 9806 }, { "epoch": 0.49, "learning_rate": 1.0819007214671508e-05, "loss": 0.2229, "step": 9807 }, { "epoch": 0.49, "learning_rate": 1.0817395426736782e-05, "loss": 0.2557, "step": 9808 }, { "epoch": 0.49, "learning_rate": 1.0815783617424144e-05, "loss": 0.3129, "step": 9809 }, { "epoch": 0.49, "learning_rate": 1.0814171786775758e-05, "loss": 0.235, "step": 9810 }, { "epoch": 0.49, "learning_rate": 1.0812559934833774e-05, "loss": 0.4182, "step": 9811 }, { "epoch": 0.49, "learning_rate": 1.0810948061640347e-05, "loss": 0.1507, "step": 9812 }, { "epoch": 0.49, "learning_rate": 1.0809336167237636e-05, "loss": 0.1608, "step": 9813 }, { "epoch": 0.49, "learning_rate": 1.0807724251667799e-05, "loss": 0.3414, "step": 9814 }, { "epoch": 0.49, "learning_rate": 1.0806112314972994e-05, "loss": 0.3559, "step": 9815 }, { "epoch": 0.49, "learning_rate": 1.0804500357195374e-05, "loss": 0.2935, "step": 9816 }, { "epoch": 0.49, "learning_rate": 1.0802888378377104e-05, "loss": 0.2435, "step": 9817 }, { "epoch": 0.49, "learning_rate": 1.0801276378560334e-05, "loss": 0.2816, "step": 9818 }, { "epoch": 0.49, "learning_rate": 1.0799664357787236e-05, "loss": 0.3913, "step": 9819 }, { "epoch": 0.49, "learning_rate": 1.0798052316099963e-05, "loss": 0.5123, "step": 9820 }, { "epoch": 0.49, "learning_rate": 1.0796440253540676e-05, "loss": 0.6025, "step": 9821 }, { "epoch": 0.49, "learning_rate": 1.0794828170151537e-05, "loss": 0.3906, "step": 9822 }, { "epoch": 0.49, "learning_rate": 1.0793216065974707e-05, "loss": 0.2397, "step": 9823 }, { "epoch": 0.49, "learning_rate": 1.0791603941052353e-05, "loss": 0.378, "step": 9824 }, { "epoch": 0.49, "learning_rate": 1.078999179542663e-05, "loss": 0.4081, "step": 9825 }, { "epoch": 0.49, "learning_rate": 1.0788379629139711e-05, "loss": 0.6261, "step": 9826 }, { "epoch": 0.49, "learning_rate": 1.0786767442233752e-05, "loss": 0.3827, "step": 9827 }, { "epoch": 0.49, "learning_rate": 1.0785155234750924e-05, "loss": 0.2787, "step": 9828 }, { "epoch": 0.49, "learning_rate": 1.0783543006733385e-05, "loss": 0.4443, "step": 9829 }, { "epoch": 0.49, "learning_rate": 1.078193075822331e-05, "loss": 0.4644, "step": 9830 }, { "epoch": 0.49, "learning_rate": 1.0780318489262858e-05, "loss": 0.5302, "step": 9831 }, { "epoch": 0.49, "learning_rate": 1.07787061998942e-05, "loss": 0.3539, "step": 9832 }, { "epoch": 0.49, "learning_rate": 1.07770938901595e-05, "loss": 0.3673, "step": 9833 }, { "epoch": 0.49, "learning_rate": 1.0775481560100924e-05, "loss": 0.5162, "step": 9834 }, { "epoch": 0.49, "learning_rate": 1.0773869209760646e-05, "loss": 0.3807, "step": 9835 }, { "epoch": 0.49, "learning_rate": 1.077225683918083e-05, "loss": 0.3829, "step": 9836 }, { "epoch": 0.49, "learning_rate": 1.077064444840365e-05, "loss": 0.2279, "step": 9837 }, { "epoch": 0.49, "learning_rate": 1.0769032037471275e-05, "loss": 0.1828, "step": 9838 }, { "epoch": 0.49, "learning_rate": 1.0767419606425873e-05, "loss": 0.3498, "step": 9839 }, { "epoch": 0.49, "learning_rate": 1.0765807155309617e-05, "loss": 0.2645, "step": 9840 }, { "epoch": 0.49, "learning_rate": 1.0764194684164678e-05, "loss": 0.33, "step": 9841 }, { "epoch": 0.49, "learning_rate": 1.0762582193033227e-05, "loss": 0.2614, "step": 9842 }, { "epoch": 0.49, "learning_rate": 1.0760969681957436e-05, "loss": 0.3722, "step": 9843 }, { "epoch": 0.49, "learning_rate": 1.075935715097948e-05, "loss": 0.4523, "step": 9844 }, { "epoch": 0.49, "learning_rate": 1.0757744600141538e-05, "loss": 0.2722, "step": 9845 }, { "epoch": 0.49, "learning_rate": 1.0756132029485774e-05, "loss": 0.3201, "step": 9846 }, { "epoch": 0.49, "learning_rate": 1.0754519439054367e-05, "loss": 0.3293, "step": 9847 }, { "epoch": 0.49, "learning_rate": 1.0752906828889491e-05, "loss": 0.1456, "step": 9848 }, { "epoch": 0.49, "learning_rate": 1.0751294199033327e-05, "loss": 0.4542, "step": 9849 }, { "epoch": 0.49, "learning_rate": 1.0749681549528043e-05, "loss": 0.3642, "step": 9850 }, { "epoch": 0.49, "learning_rate": 1.0748068880415822e-05, "loss": 0.2445, "step": 9851 }, { "epoch": 0.49, "learning_rate": 1.074645619173884e-05, "loss": 0.4525, "step": 9852 }, { "epoch": 0.49, "learning_rate": 1.0744843483539271e-05, "loss": 0.2979, "step": 9853 }, { "epoch": 0.49, "learning_rate": 1.0743230755859298e-05, "loss": 0.3703, "step": 9854 }, { "epoch": 0.49, "learning_rate": 1.0741618008741099e-05, "loss": 0.5979, "step": 9855 }, { "epoch": 0.49, "learning_rate": 1.0740005242226851e-05, "loss": 0.3599, "step": 9856 }, { "epoch": 0.49, "learning_rate": 1.073839245635874e-05, "loss": 0.5924, "step": 9857 }, { "epoch": 0.49, "learning_rate": 1.0736779651178938e-05, "loss": 0.3799, "step": 9858 }, { "epoch": 0.49, "learning_rate": 1.0735166826729626e-05, "loss": 0.3043, "step": 9859 }, { "epoch": 0.49, "learning_rate": 1.0733553983052991e-05, "loss": 0.4568, "step": 9860 }, { "epoch": 0.49, "learning_rate": 1.0731941120191212e-05, "loss": 0.4054, "step": 9861 }, { "epoch": 0.49, "learning_rate": 1.0730328238186472e-05, "loss": 0.4572, "step": 9862 }, { "epoch": 0.49, "learning_rate": 1.0728715337080955e-05, "loss": 0.4575, "step": 9863 }, { "epoch": 0.49, "learning_rate": 1.072710241691684e-05, "loss": 0.3332, "step": 9864 }, { "epoch": 0.49, "learning_rate": 1.0725489477736315e-05, "loss": 0.0801, "step": 9865 }, { "epoch": 0.49, "learning_rate": 1.0723876519581562e-05, "loss": 0.5316, "step": 9866 }, { "epoch": 0.49, "learning_rate": 1.0722263542494768e-05, "loss": 0.3309, "step": 9867 }, { "epoch": 0.49, "learning_rate": 1.072065054651812e-05, "loss": 0.245, "step": 9868 }, { "epoch": 0.49, "learning_rate": 1.0719037531693797e-05, "loss": 0.287, "step": 9869 }, { "epoch": 0.49, "learning_rate": 1.0717424498063988e-05, "loss": 0.4292, "step": 9870 }, { "epoch": 0.49, "learning_rate": 1.0715811445670883e-05, "loss": 0.29, "step": 9871 }, { "epoch": 0.49, "learning_rate": 1.0714198374556668e-05, "loss": 0.331, "step": 9872 }, { "epoch": 0.49, "learning_rate": 1.0712585284763526e-05, "loss": 0.2434, "step": 9873 }, { "epoch": 0.49, "learning_rate": 1.0710972176333653e-05, "loss": 0.4223, "step": 9874 }, { "epoch": 0.49, "learning_rate": 1.0709359049309231e-05, "loss": 0.431, "step": 9875 }, { "epoch": 0.49, "learning_rate": 1.0707745903732456e-05, "loss": 0.3398, "step": 9876 }, { "epoch": 0.49, "learning_rate": 1.0706132739645509e-05, "loss": 0.4124, "step": 9877 }, { "epoch": 0.49, "learning_rate": 1.0704519557090589e-05, "loss": 0.3267, "step": 9878 }, { "epoch": 0.49, "learning_rate": 1.0702906356109883e-05, "loss": 0.2511, "step": 9879 }, { "epoch": 0.49, "learning_rate": 1.0701293136745578e-05, "loss": 0.4125, "step": 9880 }, { "epoch": 0.49, "learning_rate": 1.0699679899039873e-05, "loss": 0.2741, "step": 9881 }, { "epoch": 0.49, "learning_rate": 1.0698066643034957e-05, "loss": 0.519, "step": 9882 }, { "epoch": 0.49, "learning_rate": 1.0696453368773024e-05, "loss": 0.1905, "step": 9883 }, { "epoch": 0.49, "learning_rate": 1.069484007629626e-05, "loss": 0.2697, "step": 9884 }, { "epoch": 0.49, "learning_rate": 1.0693226765646866e-05, "loss": 0.4985, "step": 9885 }, { "epoch": 0.49, "learning_rate": 1.0691613436867033e-05, "loss": 0.3587, "step": 9886 }, { "epoch": 0.49, "learning_rate": 1.0690000089998958e-05, "loss": 0.3654, "step": 9887 }, { "epoch": 0.49, "learning_rate": 1.0688386725084835e-05, "loss": 0.2778, "step": 9888 }, { "epoch": 0.49, "learning_rate": 1.0686773342166857e-05, "loss": 0.4673, "step": 9889 }, { "epoch": 0.49, "learning_rate": 1.0685159941287221e-05, "loss": 0.6206, "step": 9890 }, { "epoch": 0.49, "learning_rate": 1.0683546522488126e-05, "loss": 0.3573, "step": 9891 }, { "epoch": 0.49, "learning_rate": 1.0681933085811767e-05, "loss": 0.473, "step": 9892 }, { "epoch": 0.49, "learning_rate": 1.0680319631300342e-05, "loss": 0.4821, "step": 9893 }, { "epoch": 0.49, "learning_rate": 1.0678706158996048e-05, "loss": 0.3059, "step": 9894 }, { "epoch": 0.49, "learning_rate": 1.0677092668941083e-05, "loss": 0.4657, "step": 9895 }, { "epoch": 0.49, "learning_rate": 1.0675479161177645e-05, "loss": 0.647, "step": 9896 }, { "epoch": 0.49, "learning_rate": 1.0673865635747935e-05, "loss": 0.2989, "step": 9897 }, { "epoch": 0.49, "learning_rate": 1.0672252092694151e-05, "loss": 0.4325, "step": 9898 }, { "epoch": 0.49, "learning_rate": 1.0670638532058496e-05, "loss": 0.5378, "step": 9899 }, { "epoch": 0.49, "learning_rate": 1.0669024953883165e-05, "loss": 0.3342, "step": 9900 }, { "epoch": 0.49, "learning_rate": 1.0667411358210364e-05, "loss": 0.5637, "step": 9901 }, { "epoch": 0.49, "learning_rate": 1.0665797745082294e-05, "loss": 0.4495, "step": 9902 }, { "epoch": 0.49, "learning_rate": 1.0664184114541157e-05, "loss": 0.3293, "step": 9903 }, { "epoch": 0.49, "learning_rate": 1.0662570466629151e-05, "loss": 0.3603, "step": 9904 }, { "epoch": 0.49, "learning_rate": 1.0660956801388486e-05, "loss": 0.4103, "step": 9905 }, { "epoch": 0.49, "learning_rate": 1.0659343118861361e-05, "loss": 0.3906, "step": 9906 }, { "epoch": 0.49, "learning_rate": 1.065772941908998e-05, "loss": 0.5969, "step": 9907 }, { "epoch": 0.49, "learning_rate": 1.0656115702116547e-05, "loss": 0.4807, "step": 9908 }, { "epoch": 0.49, "learning_rate": 1.0654501967983266e-05, "loss": 0.3543, "step": 9909 }, { "epoch": 0.49, "learning_rate": 1.0652888216732344e-05, "loss": 0.2571, "step": 9910 }, { "epoch": 0.49, "learning_rate": 1.0651274448405988e-05, "loss": 0.4155, "step": 9911 }, { "epoch": 0.49, "learning_rate": 1.06496606630464e-05, "loss": 0.3274, "step": 9912 }, { "epoch": 0.49, "learning_rate": 1.0648046860695786e-05, "loss": 0.367, "step": 9913 }, { "epoch": 0.5, "learning_rate": 1.0646433041396357e-05, "loss": 0.4381, "step": 9914 }, { "epoch": 0.5, "learning_rate": 1.064481920519032e-05, "loss": 0.4431, "step": 9915 }, { "epoch": 0.5, "learning_rate": 1.0643205352119879e-05, "loss": 0.24, "step": 9916 }, { "epoch": 0.5, "learning_rate": 1.0641591482227246e-05, "loss": 0.5791, "step": 9917 }, { "epoch": 0.5, "learning_rate": 1.0639977595554629e-05, "loss": 0.2024, "step": 9918 }, { "epoch": 0.5, "learning_rate": 1.0638363692144235e-05, "loss": 0.481, "step": 9919 }, { "epoch": 0.5, "learning_rate": 1.0636749772038275e-05, "loss": 0.2985, "step": 9920 }, { "epoch": 0.5, "learning_rate": 1.0635135835278957e-05, "loss": 0.3379, "step": 9921 }, { "epoch": 0.5, "learning_rate": 1.0633521881908497e-05, "loss": 0.4074, "step": 9922 }, { "epoch": 0.5, "learning_rate": 1.0631907911969097e-05, "loss": 0.4906, "step": 9923 }, { "epoch": 0.5, "learning_rate": 1.0630293925502977e-05, "loss": 0.4034, "step": 9924 }, { "epoch": 0.5, "learning_rate": 1.0628679922552343e-05, "loss": 0.3926, "step": 9925 }, { "epoch": 0.5, "learning_rate": 1.062706590315941e-05, "loss": 0.4587, "step": 9926 }, { "epoch": 0.5, "learning_rate": 1.0625451867366388e-05, "loss": 0.2612, "step": 9927 }, { "epoch": 0.5, "learning_rate": 1.0623837815215492e-05, "loss": 0.378, "step": 9928 }, { "epoch": 0.5, "learning_rate": 1.0622223746748936e-05, "loss": 0.4168, "step": 9929 }, { "epoch": 0.5, "learning_rate": 1.0620609662008936e-05, "loss": 0.3442, "step": 9930 }, { "epoch": 0.5, "learning_rate": 1.06189955610377e-05, "loss": 0.3607, "step": 9931 }, { "epoch": 0.5, "learning_rate": 1.0617381443877444e-05, "loss": 0.3007, "step": 9932 }, { "epoch": 0.5, "learning_rate": 1.0615767310570389e-05, "loss": 0.3218, "step": 9933 }, { "epoch": 0.5, "learning_rate": 1.0614153161158742e-05, "loss": 0.3471, "step": 9934 }, { "epoch": 0.5, "learning_rate": 1.0612538995684725e-05, "loss": 0.2983, "step": 9935 }, { "epoch": 0.5, "learning_rate": 1.0610924814190555e-05, "loss": 0.5395, "step": 9936 }, { "epoch": 0.5, "learning_rate": 1.0609310616718444e-05, "loss": 0.3484, "step": 9937 }, { "epoch": 0.5, "learning_rate": 1.0607696403310614e-05, "loss": 0.411, "step": 9938 }, { "epoch": 0.5, "learning_rate": 1.0606082174009277e-05, "loss": 0.3289, "step": 9939 }, { "epoch": 0.5, "learning_rate": 1.0604467928856658e-05, "loss": 0.1336, "step": 9940 }, { "epoch": 0.5, "learning_rate": 1.0602853667894972e-05, "loss": 0.2693, "step": 9941 }, { "epoch": 0.5, "learning_rate": 1.0601239391166437e-05, "loss": 0.2923, "step": 9942 }, { "epoch": 0.5, "learning_rate": 1.0599625098713273e-05, "loss": 0.3385, "step": 9943 }, { "epoch": 0.5, "learning_rate": 1.0598010790577701e-05, "loss": 0.7142, "step": 9944 }, { "epoch": 0.5, "learning_rate": 1.0596396466801939e-05, "loss": 0.2619, "step": 9945 }, { "epoch": 0.5, "learning_rate": 1.0594782127428209e-05, "loss": 0.4199, "step": 9946 }, { "epoch": 0.5, "learning_rate": 1.0593167772498732e-05, "loss": 0.2374, "step": 9947 }, { "epoch": 0.5, "learning_rate": 1.0591553402055725e-05, "loss": 0.2627, "step": 9948 }, { "epoch": 0.5, "learning_rate": 1.0589939016141417e-05, "loss": 0.466, "step": 9949 }, { "epoch": 0.5, "learning_rate": 1.0588324614798027e-05, "loss": 0.3513, "step": 9950 }, { "epoch": 0.5, "learning_rate": 1.0586710198067776e-05, "loss": 0.346, "step": 9951 }, { "epoch": 0.5, "learning_rate": 1.058509576599289e-05, "loss": 0.3698, "step": 9952 }, { "epoch": 0.5, "learning_rate": 1.0583481318615588e-05, "loss": 0.4709, "step": 9953 }, { "epoch": 0.5, "learning_rate": 1.0581866855978098e-05, "loss": 0.3357, "step": 9954 }, { "epoch": 0.5, "learning_rate": 1.0580252378122644e-05, "loss": 0.2653, "step": 9955 }, { "epoch": 0.5, "learning_rate": 1.0578637885091447e-05, "loss": 0.5332, "step": 9956 }, { "epoch": 0.5, "learning_rate": 1.0577023376926734e-05, "loss": 0.342, "step": 9957 }, { "epoch": 0.5, "learning_rate": 1.057540885367073e-05, "loss": 0.3133, "step": 9958 }, { "epoch": 0.5, "learning_rate": 1.0573794315365663e-05, "loss": 0.3762, "step": 9959 }, { "epoch": 0.5, "learning_rate": 1.0572179762053753e-05, "loss": 0.3984, "step": 9960 }, { "epoch": 0.5, "learning_rate": 1.0570565193777236e-05, "loss": 0.4439, "step": 9961 }, { "epoch": 0.5, "learning_rate": 1.056895061057833e-05, "loss": 0.2941, "step": 9962 }, { "epoch": 0.5, "learning_rate": 1.0567336012499267e-05, "loss": 0.3724, "step": 9963 }, { "epoch": 0.5, "learning_rate": 1.0565721399582274e-05, "loss": 0.3076, "step": 9964 }, { "epoch": 0.5, "learning_rate": 1.056410677186958e-05, "loss": 0.3848, "step": 9965 }, { "epoch": 0.5, "learning_rate": 1.056249212940341e-05, "loss": 0.3658, "step": 9966 }, { "epoch": 0.5, "learning_rate": 1.0560877472225997e-05, "loss": 0.3226, "step": 9967 }, { "epoch": 0.5, "learning_rate": 1.0559262800379569e-05, "loss": 0.5796, "step": 9968 }, { "epoch": 0.5, "learning_rate": 1.0557648113906351e-05, "loss": 0.5663, "step": 9969 }, { "epoch": 0.5, "learning_rate": 1.055603341284858e-05, "loss": 0.4417, "step": 9970 }, { "epoch": 0.5, "learning_rate": 1.055441869724848e-05, "loss": 0.4881, "step": 9971 }, { "epoch": 0.5, "learning_rate": 1.055280396714829e-05, "loss": 0.2802, "step": 9972 }, { "epoch": 0.5, "learning_rate": 1.0551189222590233e-05, "loss": 0.4515, "step": 9973 }, { "epoch": 0.5, "learning_rate": 1.0549574463616545e-05, "loss": 0.4034, "step": 9974 }, { "epoch": 0.5, "learning_rate": 1.0547959690269455e-05, "loss": 0.4866, "step": 9975 }, { "epoch": 0.5, "learning_rate": 1.0546344902591198e-05, "loss": 0.3595, "step": 9976 }, { "epoch": 0.5, "learning_rate": 1.054473010062401e-05, "loss": 0.4062, "step": 9977 }, { "epoch": 0.5, "learning_rate": 1.0543115284410113e-05, "loss": 0.5045, "step": 9978 }, { "epoch": 0.5, "learning_rate": 1.054150045399175e-05, "loss": 0.5124, "step": 9979 }, { "epoch": 0.5, "learning_rate": 1.0539885609411153e-05, "loss": 0.6157, "step": 9980 }, { "epoch": 0.5, "learning_rate": 1.0538270750710556e-05, "loss": 0.428, "step": 9981 }, { "epoch": 0.5, "learning_rate": 1.053665587793219e-05, "loss": 0.3267, "step": 9982 }, { "epoch": 0.5, "learning_rate": 1.0535040991118292e-05, "loss": 0.1926, "step": 9983 }, { "epoch": 0.5, "learning_rate": 1.05334260903111e-05, "loss": 0.4879, "step": 9984 }, { "epoch": 0.5, "learning_rate": 1.0531811175552843e-05, "loss": 0.3224, "step": 9985 }, { "epoch": 0.5, "learning_rate": 1.0530196246885764e-05, "loss": 0.4575, "step": 9986 }, { "epoch": 0.5, "learning_rate": 1.0528581304352097e-05, "loss": 0.2421, "step": 9987 }, { "epoch": 0.5, "learning_rate": 1.0526966347994079e-05, "loss": 0.3122, "step": 9988 }, { "epoch": 0.5, "learning_rate": 1.0525351377853942e-05, "loss": 0.2901, "step": 9989 }, { "epoch": 0.5, "learning_rate": 1.0523736393973934e-05, "loss": 0.4351, "step": 9990 }, { "epoch": 0.5, "learning_rate": 1.0522121396396282e-05, "loss": 0.2163, "step": 9991 }, { "epoch": 0.5, "learning_rate": 1.0520506385163232e-05, "loss": 0.242, "step": 9992 }, { "epoch": 0.5, "learning_rate": 1.0518891360317018e-05, "loss": 0.6433, "step": 9993 }, { "epoch": 0.5, "learning_rate": 1.0517276321899879e-05, "loss": 0.6342, "step": 9994 }, { "epoch": 0.5, "learning_rate": 1.0515661269954056e-05, "loss": 0.3082, "step": 9995 }, { "epoch": 0.5, "learning_rate": 1.051404620452179e-05, "loss": 0.2304, "step": 9996 }, { "epoch": 0.5, "learning_rate": 1.0512431125645314e-05, "loss": 0.5087, "step": 9997 }, { "epoch": 0.5, "learning_rate": 1.0510816033366876e-05, "loss": 0.4208, "step": 9998 }, { "epoch": 0.5, "learning_rate": 1.0509200927728713e-05, "loss": 0.3699, "step": 9999 }, { "epoch": 0.5, "learning_rate": 1.0507585808773069e-05, "loss": 0.4327, "step": 10000 }, { "epoch": 0.5, "learning_rate": 1.0505970676542178e-05, "loss": 0.2624, "step": 10001 }, { "epoch": 0.5, "learning_rate": 1.0504355531078292e-05, "loss": 0.3271, "step": 10002 }, { "epoch": 0.5, "learning_rate": 1.0502740372423646e-05, "loss": 0.3221, "step": 10003 }, { "epoch": 0.5, "learning_rate": 1.0501125200620484e-05, "loss": 0.3694, "step": 10004 }, { "epoch": 0.5, "learning_rate": 1.0499510015711047e-05, "loss": 0.4019, "step": 10005 }, { "epoch": 0.5, "learning_rate": 1.0497894817737581e-05, "loss": 0.5101, "step": 10006 }, { "epoch": 0.5, "learning_rate": 1.0496279606742328e-05, "loss": 0.4219, "step": 10007 }, { "epoch": 0.5, "learning_rate": 1.0494664382767532e-05, "loss": 0.2606, "step": 10008 }, { "epoch": 0.5, "learning_rate": 1.0493049145855438e-05, "loss": 0.4104, "step": 10009 }, { "epoch": 0.5, "learning_rate": 1.0491433896048287e-05, "loss": 0.5265, "step": 10010 }, { "epoch": 0.5, "learning_rate": 1.0489818633388329e-05, "loss": 0.4695, "step": 10011 }, { "epoch": 0.5, "learning_rate": 1.0488203357917803e-05, "loss": 0.399, "step": 10012 }, { "epoch": 0.5, "learning_rate": 1.048658806967896e-05, "loss": 0.31, "step": 10013 }, { "epoch": 0.5, "learning_rate": 1.0484972768714042e-05, "loss": 0.3127, "step": 10014 }, { "epoch": 0.5, "learning_rate": 1.0483357455065298e-05, "loss": 0.3454, "step": 10015 }, { "epoch": 0.5, "learning_rate": 1.0481742128774973e-05, "loss": 0.4331, "step": 10016 }, { "epoch": 0.5, "learning_rate": 1.0480126789885312e-05, "loss": 0.4217, "step": 10017 }, { "epoch": 0.5, "learning_rate": 1.0478511438438565e-05, "loss": 0.3991, "step": 10018 }, { "epoch": 0.5, "learning_rate": 1.0476896074476975e-05, "loss": 0.3342, "step": 10019 }, { "epoch": 0.5, "learning_rate": 1.0475280698042796e-05, "loss": 0.3127, "step": 10020 }, { "epoch": 0.5, "learning_rate": 1.0473665309178271e-05, "loss": 0.2628, "step": 10021 }, { "epoch": 0.5, "learning_rate": 1.0472049907925653e-05, "loss": 0.351, "step": 10022 }, { "epoch": 0.5, "learning_rate": 1.0470434494327185e-05, "loss": 0.2792, "step": 10023 }, { "epoch": 0.5, "learning_rate": 1.0468819068425118e-05, "loss": 0.4309, "step": 10024 }, { "epoch": 0.5, "learning_rate": 1.0467203630261707e-05, "loss": 0.3204, "step": 10025 }, { "epoch": 0.5, "learning_rate": 1.0465588179879192e-05, "loss": 0.4689, "step": 10026 }, { "epoch": 0.5, "learning_rate": 1.046397271731983e-05, "loss": 0.4425, "step": 10027 }, { "epoch": 0.5, "learning_rate": 1.0462357242625871e-05, "loss": 0.3601, "step": 10028 }, { "epoch": 0.5, "learning_rate": 1.0460741755839563e-05, "loss": 0.5779, "step": 10029 }, { "epoch": 0.5, "learning_rate": 1.0459126257003156e-05, "loss": 0.2071, "step": 10030 }, { "epoch": 0.5, "learning_rate": 1.0457510746158906e-05, "loss": 0.4515, "step": 10031 }, { "epoch": 0.5, "learning_rate": 1.045589522334906e-05, "loss": 0.3745, "step": 10032 }, { "epoch": 0.5, "learning_rate": 1.0454279688615872e-05, "loss": 0.3531, "step": 10033 }, { "epoch": 0.5, "learning_rate": 1.0452664142001591e-05, "loss": 0.293, "step": 10034 }, { "epoch": 0.5, "learning_rate": 1.0451048583548475e-05, "loss": 0.4818, "step": 10035 }, { "epoch": 0.5, "learning_rate": 1.0449433013298774e-05, "loss": 0.2653, "step": 10036 }, { "epoch": 0.5, "learning_rate": 1.044781743129474e-05, "loss": 0.3609, "step": 10037 }, { "epoch": 0.5, "learning_rate": 1.044620183757863e-05, "loss": 0.3094, "step": 10038 }, { "epoch": 0.5, "learning_rate": 1.0444586232192693e-05, "loss": 0.3902, "step": 10039 }, { "epoch": 0.5, "learning_rate": 1.0442970615179187e-05, "loss": 0.411, "step": 10040 }, { "epoch": 0.5, "learning_rate": 1.0441354986580365e-05, "loss": 0.1102, "step": 10041 }, { "epoch": 0.5, "learning_rate": 1.043973934643848e-05, "loss": 0.4199, "step": 10042 }, { "epoch": 0.5, "learning_rate": 1.0438123694795787e-05, "loss": 0.1551, "step": 10043 }, { "epoch": 0.5, "learning_rate": 1.0436508031694544e-05, "loss": 0.4784, "step": 10044 }, { "epoch": 0.5, "learning_rate": 1.0434892357177006e-05, "loss": 0.4076, "step": 10045 }, { "epoch": 0.5, "learning_rate": 1.0433276671285427e-05, "loss": 0.3246, "step": 10046 }, { "epoch": 0.5, "learning_rate": 1.0431660974062066e-05, "loss": 0.418, "step": 10047 }, { "epoch": 0.5, "learning_rate": 1.0430045265549174e-05, "loss": 0.2478, "step": 10048 }, { "epoch": 0.5, "learning_rate": 1.0428429545789014e-05, "loss": 0.3506, "step": 10049 }, { "epoch": 0.5, "learning_rate": 1.0426813814823842e-05, "loss": 0.2044, "step": 10050 }, { "epoch": 0.5, "learning_rate": 1.0425198072695907e-05, "loss": 0.3967, "step": 10051 }, { "epoch": 0.5, "learning_rate": 1.042358231944748e-05, "loss": 0.4306, "step": 10052 }, { "epoch": 0.5, "learning_rate": 1.0421966555120813e-05, "loss": 0.4056, "step": 10053 }, { "epoch": 0.5, "learning_rate": 1.042035077975816e-05, "loss": 0.4691, "step": 10054 }, { "epoch": 0.5, "learning_rate": 1.0418734993401783e-05, "loss": 0.2545, "step": 10055 }, { "epoch": 0.5, "learning_rate": 1.041711919609394e-05, "loss": 0.2802, "step": 10056 }, { "epoch": 0.5, "learning_rate": 1.0415503387876894e-05, "loss": 0.2904, "step": 10057 }, { "epoch": 0.5, "learning_rate": 1.0413887568792897e-05, "loss": 0.8372, "step": 10058 }, { "epoch": 0.5, "learning_rate": 1.0412271738884217e-05, "loss": 0.3434, "step": 10059 }, { "epoch": 0.5, "learning_rate": 1.0410655898193106e-05, "loss": 0.3727, "step": 10060 }, { "epoch": 0.5, "learning_rate": 1.0409040046761829e-05, "loss": 0.5687, "step": 10061 }, { "epoch": 0.5, "learning_rate": 1.0407424184632644e-05, "loss": 0.3485, "step": 10062 }, { "epoch": 0.5, "learning_rate": 1.0405808311847815e-05, "loss": 0.5685, "step": 10063 }, { "epoch": 0.5, "learning_rate": 1.0404192428449603e-05, "loss": 0.3968, "step": 10064 }, { "epoch": 0.5, "learning_rate": 1.0402576534480265e-05, "loss": 0.3112, "step": 10065 }, { "epoch": 0.5, "learning_rate": 1.0400960629982067e-05, "loss": 0.2865, "step": 10066 }, { "epoch": 0.5, "learning_rate": 1.0399344714997266e-05, "loss": 0.2289, "step": 10067 }, { "epoch": 0.5, "learning_rate": 1.0397728789568128e-05, "loss": 0.4246, "step": 10068 }, { "epoch": 0.5, "learning_rate": 1.0396112853736913e-05, "loss": 0.3925, "step": 10069 }, { "epoch": 0.5, "learning_rate": 1.0394496907545888e-05, "loss": 0.1329, "step": 10070 }, { "epoch": 0.5, "learning_rate": 1.039288095103731e-05, "loss": 0.5366, "step": 10071 }, { "epoch": 0.5, "learning_rate": 1.0391264984253448e-05, "loss": 0.3666, "step": 10072 }, { "epoch": 0.5, "learning_rate": 1.0389649007236562e-05, "loss": 0.3592, "step": 10073 }, { "epoch": 0.5, "learning_rate": 1.0388033020028915e-05, "loss": 0.161, "step": 10074 }, { "epoch": 0.5, "learning_rate": 1.0386417022672774e-05, "loss": 0.1973, "step": 10075 }, { "epoch": 0.5, "learning_rate": 1.0384801015210405e-05, "loss": 0.1679, "step": 10076 }, { "epoch": 0.5, "learning_rate": 1.0383184997684066e-05, "loss": 0.5558, "step": 10077 }, { "epoch": 0.5, "learning_rate": 1.0381568970136025e-05, "loss": 0.4647, "step": 10078 }, { "epoch": 0.5, "learning_rate": 1.0379952932608548e-05, "loss": 0.494, "step": 10079 }, { "epoch": 0.5, "learning_rate": 1.0378336885143897e-05, "loss": 0.2668, "step": 10080 }, { "epoch": 0.5, "learning_rate": 1.0376720827784345e-05, "loss": 0.4616, "step": 10081 }, { "epoch": 0.5, "learning_rate": 1.0375104760572152e-05, "loss": 0.3032, "step": 10082 }, { "epoch": 0.5, "learning_rate": 1.0373488683549583e-05, "loss": 0.4397, "step": 10083 }, { "epoch": 0.5, "learning_rate": 1.037187259675891e-05, "loss": 0.4756, "step": 10084 }, { "epoch": 0.5, "learning_rate": 1.0370256500242393e-05, "loss": 0.1201, "step": 10085 }, { "epoch": 0.5, "learning_rate": 1.0368640394042304e-05, "loss": 0.4101, "step": 10086 }, { "epoch": 0.5, "learning_rate": 1.0367024278200905e-05, "loss": 0.2978, "step": 10087 }, { "epoch": 0.5, "learning_rate": 1.0365408152760472e-05, "loss": 0.4269, "step": 10088 }, { "epoch": 0.5, "learning_rate": 1.0363792017763266e-05, "loss": 0.4224, "step": 10089 }, { "epoch": 0.5, "learning_rate": 1.0362175873251557e-05, "loss": 0.5225, "step": 10090 }, { "epoch": 0.5, "learning_rate": 1.036055971926761e-05, "loss": 0.4487, "step": 10091 }, { "epoch": 0.5, "learning_rate": 1.0358943555853695e-05, "loss": 0.2811, "step": 10092 }, { "epoch": 0.5, "learning_rate": 1.0357327383052086e-05, "loss": 0.3215, "step": 10093 }, { "epoch": 0.5, "learning_rate": 1.0355711200905041e-05, "loss": 0.4521, "step": 10094 }, { "epoch": 0.5, "learning_rate": 1.0354095009454839e-05, "loss": 0.3104, "step": 10095 }, { "epoch": 0.5, "learning_rate": 1.0352478808743744e-05, "loss": 0.2991, "step": 10096 }, { "epoch": 0.5, "learning_rate": 1.035086259881403e-05, "loss": 0.3593, "step": 10097 }, { "epoch": 0.5, "learning_rate": 1.0349246379707965e-05, "loss": 0.1502, "step": 10098 }, { "epoch": 0.5, "learning_rate": 1.0347630151467813e-05, "loss": 0.4135, "step": 10099 }, { "epoch": 0.5, "learning_rate": 1.0346013914135856e-05, "loss": 0.3541, "step": 10100 }, { "epoch": 0.5, "learning_rate": 1.0344397667754355e-05, "loss": 0.3259, "step": 10101 }, { "epoch": 0.5, "learning_rate": 1.0342781412365586e-05, "loss": 0.4883, "step": 10102 }, { "epoch": 0.5, "learning_rate": 1.0341165148011816e-05, "loss": 0.2372, "step": 10103 }, { "epoch": 0.5, "learning_rate": 1.0339548874735319e-05, "loss": 0.2272, "step": 10104 }, { "epoch": 0.5, "learning_rate": 1.0337932592578369e-05, "loss": 0.2895, "step": 10105 }, { "epoch": 0.5, "learning_rate": 1.033631630158323e-05, "loss": 0.4194, "step": 10106 }, { "epoch": 0.5, "learning_rate": 1.0334700001792179e-05, "loss": 0.4224, "step": 10107 }, { "epoch": 0.5, "learning_rate": 1.0333083693247488e-05, "loss": 0.4919, "step": 10108 }, { "epoch": 0.5, "learning_rate": 1.0331467375991431e-05, "loss": 0.3272, "step": 10109 }, { "epoch": 0.5, "learning_rate": 1.0329851050066278e-05, "loss": 0.3001, "step": 10110 }, { "epoch": 0.5, "learning_rate": 1.0328234715514303e-05, "loss": 0.2653, "step": 10111 }, { "epoch": 0.5, "learning_rate": 1.0326618372377779e-05, "loss": 0.4503, "step": 10112 }, { "epoch": 0.5, "learning_rate": 1.0325002020698982e-05, "loss": 0.3862, "step": 10113 }, { "epoch": 0.51, "learning_rate": 1.032338566052018e-05, "loss": 0.3245, "step": 10114 }, { "epoch": 0.51, "learning_rate": 1.0321769291883647e-05, "loss": 0.5169, "step": 10115 }, { "epoch": 0.51, "learning_rate": 1.0320152914831665e-05, "loss": 0.3966, "step": 10116 }, { "epoch": 0.51, "learning_rate": 1.0318536529406498e-05, "loss": 0.6949, "step": 10117 }, { "epoch": 0.51, "learning_rate": 1.0316920135650428e-05, "loss": 0.2703, "step": 10118 }, { "epoch": 0.51, "learning_rate": 1.0315303733605726e-05, "loss": 0.3357, "step": 10119 }, { "epoch": 0.51, "learning_rate": 1.031368732331467e-05, "loss": 0.4038, "step": 10120 }, { "epoch": 0.51, "learning_rate": 1.0312070904819531e-05, "loss": 0.3314, "step": 10121 }, { "epoch": 0.51, "learning_rate": 1.0310454478162584e-05, "loss": 0.1869, "step": 10122 }, { "epoch": 0.51, "learning_rate": 1.0308838043386111e-05, "loss": 0.2781, "step": 10123 }, { "epoch": 0.51, "learning_rate": 1.030722160053238e-05, "loss": 0.2584, "step": 10124 }, { "epoch": 0.51, "learning_rate": 1.0305605149643671e-05, "loss": 0.28, "step": 10125 }, { "epoch": 0.51, "learning_rate": 1.0303988690762264e-05, "loss": 0.1907, "step": 10126 }, { "epoch": 0.51, "learning_rate": 1.0302372223930429e-05, "loss": 0.0851, "step": 10127 }, { "epoch": 0.51, "learning_rate": 1.030075574919044e-05, "loss": 0.3605, "step": 10128 }, { "epoch": 0.51, "learning_rate": 1.0299139266584582e-05, "loss": 0.2975, "step": 10129 }, { "epoch": 0.51, "learning_rate": 1.0297522776155129e-05, "loss": 0.3964, "step": 10130 }, { "epoch": 0.51, "learning_rate": 1.0295906277944353e-05, "loss": 0.5119, "step": 10131 }, { "epoch": 0.51, "learning_rate": 1.0294289771994541e-05, "loss": 0.2476, "step": 10132 }, { "epoch": 0.51, "learning_rate": 1.0292673258347963e-05, "loss": 0.4491, "step": 10133 }, { "epoch": 0.51, "learning_rate": 1.02910567370469e-05, "loss": 0.3725, "step": 10134 }, { "epoch": 0.51, "learning_rate": 1.0289440208133628e-05, "loss": 0.4679, "step": 10135 }, { "epoch": 0.51, "learning_rate": 1.0287823671650428e-05, "loss": 0.3265, "step": 10136 }, { "epoch": 0.51, "learning_rate": 1.0286207127639575e-05, "loss": 0.3875, "step": 10137 }, { "epoch": 0.51, "learning_rate": 1.0284590576143355e-05, "loss": 0.24, "step": 10138 }, { "epoch": 0.51, "learning_rate": 1.0282974017204036e-05, "loss": 0.4866, "step": 10139 }, { "epoch": 0.51, "learning_rate": 1.0281357450863903e-05, "loss": 0.2744, "step": 10140 }, { "epoch": 0.51, "learning_rate": 1.0279740877165233e-05, "loss": 0.3167, "step": 10141 }, { "epoch": 0.51, "learning_rate": 1.0278124296150308e-05, "loss": 0.3201, "step": 10142 }, { "epoch": 0.51, "learning_rate": 1.0276507707861407e-05, "loss": 0.3193, "step": 10143 }, { "epoch": 0.51, "learning_rate": 1.0274891112340808e-05, "loss": 0.4252, "step": 10144 }, { "epoch": 0.51, "learning_rate": 1.0273274509630792e-05, "loss": 0.2569, "step": 10145 }, { "epoch": 0.51, "learning_rate": 1.0271657899773641e-05, "loss": 0.4007, "step": 10146 }, { "epoch": 0.51, "learning_rate": 1.0270041282811632e-05, "loss": 0.3723, "step": 10147 }, { "epoch": 0.51, "learning_rate": 1.0268424658787048e-05, "loss": 0.3468, "step": 10148 }, { "epoch": 0.51, "learning_rate": 1.0266808027742167e-05, "loss": 0.2694, "step": 10149 }, { "epoch": 0.51, "learning_rate": 1.0265191389719273e-05, "loss": 0.2367, "step": 10150 }, { "epoch": 0.51, "learning_rate": 1.0263574744760644e-05, "loss": 0.4921, "step": 10151 }, { "epoch": 0.51, "learning_rate": 1.0261958092908567e-05, "loss": 0.183, "step": 10152 }, { "epoch": 0.51, "learning_rate": 1.0260341434205314e-05, "loss": 0.4035, "step": 10153 }, { "epoch": 0.51, "learning_rate": 1.0258724768693173e-05, "loss": 0.3891, "step": 10154 }, { "epoch": 0.51, "learning_rate": 1.0257108096414425e-05, "loss": 0.3244, "step": 10155 }, { "epoch": 0.51, "learning_rate": 1.0255491417411352e-05, "loss": 0.3054, "step": 10156 }, { "epoch": 0.51, "learning_rate": 1.0253874731726234e-05, "loss": 0.5944, "step": 10157 }, { "epoch": 0.51, "learning_rate": 1.0252258039401356e-05, "loss": 0.3582, "step": 10158 }, { "epoch": 0.51, "learning_rate": 1.0250641340479e-05, "loss": 0.4801, "step": 10159 }, { "epoch": 0.51, "learning_rate": 1.0249024635001446e-05, "loss": 0.3912, "step": 10160 }, { "epoch": 0.51, "learning_rate": 1.0247407923010981e-05, "loss": 0.4422, "step": 10161 }, { "epoch": 0.51, "learning_rate": 1.0245791204549886e-05, "loss": 0.2772, "step": 10162 }, { "epoch": 0.51, "learning_rate": 1.0244174479660445e-05, "loss": 0.4023, "step": 10163 }, { "epoch": 0.51, "learning_rate": 1.0242557748384939e-05, "loss": 0.4967, "step": 10164 }, { "epoch": 0.51, "learning_rate": 1.024094101076565e-05, "loss": 0.2556, "step": 10165 }, { "epoch": 0.51, "learning_rate": 1.0239324266844868e-05, "loss": 0.3809, "step": 10166 }, { "epoch": 0.51, "learning_rate": 1.023770751666487e-05, "loss": 0.1998, "step": 10167 }, { "epoch": 0.51, "learning_rate": 1.0236090760267947e-05, "loss": 0.2854, "step": 10168 }, { "epoch": 0.51, "learning_rate": 1.0234473997696378e-05, "loss": 0.3285, "step": 10169 }, { "epoch": 0.51, "learning_rate": 1.0232857228992446e-05, "loss": 0.6043, "step": 10170 }, { "epoch": 0.51, "learning_rate": 1.023124045419844e-05, "loss": 0.5581, "step": 10171 }, { "epoch": 0.51, "learning_rate": 1.0229623673356643e-05, "loss": 0.2653, "step": 10172 }, { "epoch": 0.51, "learning_rate": 1.022800688650934e-05, "loss": 0.417, "step": 10173 }, { "epoch": 0.51, "learning_rate": 1.0226390093698817e-05, "loss": 0.6139, "step": 10174 }, { "epoch": 0.51, "learning_rate": 1.0224773294967355e-05, "loss": 0.4123, "step": 10175 }, { "epoch": 0.51, "learning_rate": 1.0223156490357241e-05, "loss": 0.1175, "step": 10176 }, { "epoch": 0.51, "learning_rate": 1.0221539679910765e-05, "loss": 0.2908, "step": 10177 }, { "epoch": 0.51, "learning_rate": 1.0219922863670208e-05, "loss": 0.3934, "step": 10178 }, { "epoch": 0.51, "learning_rate": 1.0218306041677854e-05, "loss": 0.3738, "step": 10179 }, { "epoch": 0.51, "learning_rate": 1.0216689213975994e-05, "loss": 0.3843, "step": 10180 }, { "epoch": 0.51, "learning_rate": 1.021507238060691e-05, "loss": 0.443, "step": 10181 }, { "epoch": 0.51, "learning_rate": 1.0213455541612889e-05, "loss": 0.5593, "step": 10182 }, { "epoch": 0.51, "learning_rate": 1.0211838697036218e-05, "loss": 0.2295, "step": 10183 }, { "epoch": 0.51, "learning_rate": 1.0210221846919186e-05, "loss": 0.2183, "step": 10184 }, { "epoch": 0.51, "learning_rate": 1.0208604991304075e-05, "loss": 0.464, "step": 10185 }, { "epoch": 0.51, "learning_rate": 1.0206988130233175e-05, "loss": 0.2669, "step": 10186 }, { "epoch": 0.51, "learning_rate": 1.0205371263748771e-05, "loss": 0.3552, "step": 10187 }, { "epoch": 0.51, "learning_rate": 1.0203754391893149e-05, "loss": 0.3863, "step": 10188 }, { "epoch": 0.51, "learning_rate": 1.02021375147086e-05, "loss": 0.2823, "step": 10189 }, { "epoch": 0.51, "learning_rate": 1.0200520632237406e-05, "loss": 0.4711, "step": 10190 }, { "epoch": 0.51, "learning_rate": 1.0198903744521857e-05, "loss": 0.2657, "step": 10191 }, { "epoch": 0.51, "learning_rate": 1.0197286851604242e-05, "loss": 0.2913, "step": 10192 }, { "epoch": 0.51, "learning_rate": 1.019566995352685e-05, "loss": 0.3158, "step": 10193 }, { "epoch": 0.51, "learning_rate": 1.0194053050331968e-05, "loss": 0.3309, "step": 10194 }, { "epoch": 0.51, "learning_rate": 1.0192436142061878e-05, "loss": 0.1872, "step": 10195 }, { "epoch": 0.51, "learning_rate": 1.0190819228758877e-05, "loss": 0.3714, "step": 10196 }, { "epoch": 0.51, "learning_rate": 1.0189202310465245e-05, "loss": 0.3206, "step": 10197 }, { "epoch": 0.51, "learning_rate": 1.0187585387223278e-05, "loss": 0.5368, "step": 10198 }, { "epoch": 0.51, "learning_rate": 1.018596845907526e-05, "loss": 0.4337, "step": 10199 }, { "epoch": 0.51, "learning_rate": 1.0184351526063481e-05, "loss": 0.1468, "step": 10200 }, { "epoch": 0.51, "learning_rate": 1.0182734588230226e-05, "loss": 0.2245, "step": 10201 }, { "epoch": 0.51, "learning_rate": 1.018111764561779e-05, "loss": 0.3599, "step": 10202 }, { "epoch": 0.51, "learning_rate": 1.0179500698268459e-05, "loss": 0.3304, "step": 10203 }, { "epoch": 0.51, "learning_rate": 1.017788374622452e-05, "loss": 0.3279, "step": 10204 }, { "epoch": 0.51, "learning_rate": 1.0176266789528268e-05, "loss": 0.4343, "step": 10205 }, { "epoch": 0.51, "learning_rate": 1.0174649828221985e-05, "loss": 0.3856, "step": 10206 }, { "epoch": 0.51, "learning_rate": 1.0173032862347969e-05, "loss": 0.2615, "step": 10207 }, { "epoch": 0.51, "learning_rate": 1.0171415891948502e-05, "loss": 0.4268, "step": 10208 }, { "epoch": 0.51, "learning_rate": 1.0169798917065879e-05, "loss": 0.2613, "step": 10209 }, { "epoch": 0.51, "learning_rate": 1.0168181937742387e-05, "loss": 0.3574, "step": 10210 }, { "epoch": 0.51, "learning_rate": 1.0166564954020317e-05, "loss": 0.42, "step": 10211 }, { "epoch": 0.51, "learning_rate": 1.0164947965941958e-05, "loss": 0.3172, "step": 10212 }, { "epoch": 0.51, "learning_rate": 1.0163330973549602e-05, "loss": 0.3383, "step": 10213 }, { "epoch": 0.51, "learning_rate": 1.0161713976885536e-05, "loss": 0.2157, "step": 10214 }, { "epoch": 0.51, "learning_rate": 1.0160096975992054e-05, "loss": 0.2253, "step": 10215 }, { "epoch": 0.51, "learning_rate": 1.0158479970911445e-05, "loss": 0.2298, "step": 10216 }, { "epoch": 0.51, "learning_rate": 1.0156862961686001e-05, "loss": 0.325, "step": 10217 }, { "epoch": 0.51, "learning_rate": 1.015524594835801e-05, "loss": 0.1868, "step": 10218 }, { "epoch": 0.51, "learning_rate": 1.0153628930969767e-05, "loss": 0.3756, "step": 10219 }, { "epoch": 0.51, "learning_rate": 1.0152011909563558e-05, "loss": 0.3776, "step": 10220 }, { "epoch": 0.51, "learning_rate": 1.0150394884181678e-05, "loss": 0.4473, "step": 10221 }, { "epoch": 0.51, "learning_rate": 1.0148777854866419e-05, "loss": 0.3466, "step": 10222 }, { "epoch": 0.51, "learning_rate": 1.0147160821660067e-05, "loss": 0.5545, "step": 10223 }, { "epoch": 0.51, "learning_rate": 1.0145543784604917e-05, "loss": 0.2399, "step": 10224 }, { "epoch": 0.51, "learning_rate": 1.014392674374326e-05, "loss": 0.2937, "step": 10225 }, { "epoch": 0.51, "learning_rate": 1.0142309699117387e-05, "loss": 0.4092, "step": 10226 }, { "epoch": 0.51, "learning_rate": 1.014069265076959e-05, "loss": 0.4356, "step": 10227 }, { "epoch": 0.51, "learning_rate": 1.0139075598742163e-05, "loss": 0.3512, "step": 10228 }, { "epoch": 0.51, "learning_rate": 1.0137458543077393e-05, "loss": 0.3128, "step": 10229 }, { "epoch": 0.51, "learning_rate": 1.0135841483817577e-05, "loss": 0.565, "step": 10230 }, { "epoch": 0.51, "learning_rate": 1.0134224421005003e-05, "loss": 0.1103, "step": 10231 }, { "epoch": 0.51, "learning_rate": 1.0132607354681968e-05, "loss": 0.2945, "step": 10232 }, { "epoch": 0.51, "learning_rate": 1.0130990284890758e-05, "loss": 0.3511, "step": 10233 }, { "epoch": 0.51, "learning_rate": 1.0129373211673672e-05, "loss": 0.1323, "step": 10234 }, { "epoch": 0.51, "learning_rate": 1.0127756135072996e-05, "loss": 0.3558, "step": 10235 }, { "epoch": 0.51, "learning_rate": 1.012613905513103e-05, "loss": 0.3515, "step": 10236 }, { "epoch": 0.51, "learning_rate": 1.012452197189006e-05, "loss": 0.4656, "step": 10237 }, { "epoch": 0.51, "learning_rate": 1.012290488539238e-05, "loss": 0.2517, "step": 10238 }, { "epoch": 0.51, "learning_rate": 1.0121287795680284e-05, "loss": 0.3215, "step": 10239 }, { "epoch": 0.51, "learning_rate": 1.0119670702796062e-05, "loss": 0.4317, "step": 10240 }, { "epoch": 0.51, "learning_rate": 1.0118053606782012e-05, "loss": 0.3917, "step": 10241 }, { "epoch": 0.51, "learning_rate": 1.0116436507680427e-05, "loss": 0.5679, "step": 10242 }, { "epoch": 0.51, "learning_rate": 1.0114819405533594e-05, "loss": 0.252, "step": 10243 }, { "epoch": 0.51, "learning_rate": 1.0113202300383811e-05, "loss": 0.1681, "step": 10244 }, { "epoch": 0.51, "learning_rate": 1.011158519227337e-05, "loss": 0.3747, "step": 10245 }, { "epoch": 0.51, "learning_rate": 1.0109968081244565e-05, "loss": 0.5378, "step": 10246 }, { "epoch": 0.51, "learning_rate": 1.010835096733969e-05, "loss": 0.2448, "step": 10247 }, { "epoch": 0.51, "learning_rate": 1.0106733850601036e-05, "loss": 0.2182, "step": 10248 }, { "epoch": 0.51, "learning_rate": 1.0105116731070896e-05, "loss": 0.3903, "step": 10249 }, { "epoch": 0.51, "learning_rate": 1.010349960879157e-05, "loss": 0.5696, "step": 10250 }, { "epoch": 0.51, "learning_rate": 1.0101882483805343e-05, "loss": 0.6131, "step": 10251 }, { "epoch": 0.51, "learning_rate": 1.0100265356154515e-05, "loss": 0.5222, "step": 10252 }, { "epoch": 0.51, "learning_rate": 1.0098648225881378e-05, "loss": 0.3489, "step": 10253 }, { "epoch": 0.51, "learning_rate": 1.0097031093028224e-05, "loss": 0.5603, "step": 10254 }, { "epoch": 0.51, "learning_rate": 1.009541395763735e-05, "loss": 0.2513, "step": 10255 }, { "epoch": 0.51, "learning_rate": 1.009379681975105e-05, "loss": 0.4287, "step": 10256 }, { "epoch": 0.51, "learning_rate": 1.0092179679411616e-05, "loss": 0.1898, "step": 10257 }, { "epoch": 0.51, "learning_rate": 1.0090562536661343e-05, "loss": 0.3215, "step": 10258 }, { "epoch": 0.51, "learning_rate": 1.0088945391542529e-05, "loss": 0.3554, "step": 10259 }, { "epoch": 0.51, "learning_rate": 1.0087328244097459e-05, "loss": 0.3558, "step": 10260 }, { "epoch": 0.51, "learning_rate": 1.0085711094368438e-05, "loss": 0.2687, "step": 10261 }, { "epoch": 0.51, "learning_rate": 1.0084093942397754e-05, "loss": 0.339, "step": 10262 }, { "epoch": 0.51, "learning_rate": 1.00824767882277e-05, "loss": 0.3957, "step": 10263 }, { "epoch": 0.51, "learning_rate": 1.0080859631900577e-05, "loss": 0.3305, "step": 10264 }, { "epoch": 0.51, "learning_rate": 1.0079242473458674e-05, "loss": 0.2025, "step": 10265 }, { "epoch": 0.51, "learning_rate": 1.007762531294429e-05, "loss": 0.4419, "step": 10266 }, { "epoch": 0.51, "learning_rate": 1.0076008150399719e-05, "loss": 0.4246, "step": 10267 }, { "epoch": 0.51, "learning_rate": 1.007439098586725e-05, "loss": 0.2297, "step": 10268 }, { "epoch": 0.51, "learning_rate": 1.0072773819389186e-05, "loss": 0.2605, "step": 10269 }, { "epoch": 0.51, "learning_rate": 1.0071156651007816e-05, "loss": 0.4024, "step": 10270 }, { "epoch": 0.51, "learning_rate": 1.0069539480765437e-05, "loss": 0.46, "step": 10271 }, { "epoch": 0.51, "learning_rate": 1.0067922308704345e-05, "loss": 0.2204, "step": 10272 }, { "epoch": 0.51, "learning_rate": 1.0066305134866836e-05, "loss": 0.52, "step": 10273 }, { "epoch": 0.51, "learning_rate": 1.00646879592952e-05, "loss": 0.4966, "step": 10274 }, { "epoch": 0.51, "learning_rate": 1.0063070782031737e-05, "loss": 0.4514, "step": 10275 }, { "epoch": 0.51, "learning_rate": 1.0061453603118738e-05, "loss": 0.1631, "step": 10276 }, { "epoch": 0.51, "learning_rate": 1.0059836422598502e-05, "loss": 0.3916, "step": 10277 }, { "epoch": 0.51, "learning_rate": 1.0058219240513323e-05, "loss": 0.3999, "step": 10278 }, { "epoch": 0.51, "learning_rate": 1.0056602056905493e-05, "loss": 0.4053, "step": 10279 }, { "epoch": 0.51, "learning_rate": 1.0054984871817313e-05, "loss": 0.3115, "step": 10280 }, { "epoch": 0.51, "learning_rate": 1.0053367685291073e-05, "loss": 0.5462, "step": 10281 }, { "epoch": 0.51, "learning_rate": 1.0051750497369075e-05, "loss": 0.4398, "step": 10282 }, { "epoch": 0.51, "learning_rate": 1.005013330809361e-05, "loss": 0.3675, "step": 10283 }, { "epoch": 0.51, "learning_rate": 1.0048516117506974e-05, "loss": 0.2, "step": 10284 }, { "epoch": 0.51, "learning_rate": 1.0046898925651462e-05, "loss": 0.5077, "step": 10285 }, { "epoch": 0.51, "learning_rate": 1.0045281732569367e-05, "loss": 0.3997, "step": 10286 }, { "epoch": 0.51, "learning_rate": 1.0043664538302992e-05, "loss": 0.3121, "step": 10287 }, { "epoch": 0.51, "learning_rate": 1.0042047342894626e-05, "loss": 0.2977, "step": 10288 }, { "epoch": 0.51, "learning_rate": 1.0040430146386567e-05, "loss": 0.3818, "step": 10289 }, { "epoch": 0.51, "learning_rate": 1.003881294882111e-05, "loss": 0.2705, "step": 10290 }, { "epoch": 0.51, "learning_rate": 1.0037195750240552e-05, "loss": 0.352, "step": 10291 }, { "epoch": 0.51, "learning_rate": 1.0035578550687189e-05, "loss": 0.4311, "step": 10292 }, { "epoch": 0.51, "learning_rate": 1.0033961350203311e-05, "loss": 0.2048, "step": 10293 }, { "epoch": 0.51, "learning_rate": 1.0032344148831223e-05, "loss": 0.2335, "step": 10294 }, { "epoch": 0.51, "learning_rate": 1.0030726946613217e-05, "loss": 0.2859, "step": 10295 }, { "epoch": 0.51, "learning_rate": 1.0029109743591583e-05, "loss": 0.3391, "step": 10296 }, { "epoch": 0.51, "learning_rate": 1.0027492539808625e-05, "loss": 0.268, "step": 10297 }, { "epoch": 0.51, "learning_rate": 1.0025875335306637e-05, "loss": 0.6828, "step": 10298 }, { "epoch": 0.51, "learning_rate": 1.0024258130127912e-05, "loss": 0.3606, "step": 10299 }, { "epoch": 0.51, "learning_rate": 1.0022640924314743e-05, "loss": 0.3064, "step": 10300 }, { "epoch": 0.51, "learning_rate": 1.0021023717909438e-05, "loss": 0.4271, "step": 10301 }, { "epoch": 0.51, "learning_rate": 1.001940651095428e-05, "loss": 0.4512, "step": 10302 }, { "epoch": 0.51, "learning_rate": 1.0017789303491572e-05, "loss": 0.507, "step": 10303 }, { "epoch": 0.51, "learning_rate": 1.0016172095563605e-05, "loss": 0.4544, "step": 10304 }, { "epoch": 0.51, "learning_rate": 1.0014554887212682e-05, "loss": 0.1495, "step": 10305 }, { "epoch": 0.51, "learning_rate": 1.0012937678481094e-05, "loss": 0.2041, "step": 10306 }, { "epoch": 0.51, "learning_rate": 1.0011320469411136e-05, "loss": 0.5599, "step": 10307 }, { "epoch": 0.51, "learning_rate": 1.000970326004511e-05, "loss": 0.2207, "step": 10308 }, { "epoch": 0.51, "learning_rate": 1.0008086050425305e-05, "loss": 0.3603, "step": 10309 }, { "epoch": 0.51, "learning_rate": 1.0006468840594022e-05, "loss": 0.3354, "step": 10310 }, { "epoch": 0.51, "learning_rate": 1.0004851630593551e-05, "loss": 0.2651, "step": 10311 }, { "epoch": 0.51, "learning_rate": 1.0003234420466195e-05, "loss": 0.2491, "step": 10312 }, { "epoch": 0.51, "learning_rate": 1.0001617210254246e-05, "loss": 0.4155, "step": 10313 }, { "epoch": 0.52, "learning_rate": 1e-05, "loss": 0.2035, "step": 10314 }, { "epoch": 0.52, "learning_rate": 9.998382789745757e-06, "loss": 0.5797, "step": 10315 }, { "epoch": 0.52, "learning_rate": 9.99676557953381e-06, "loss": 0.2642, "step": 10316 }, { "epoch": 0.52, "learning_rate": 9.995148369406449e-06, "loss": 0.3215, "step": 10317 }, { "epoch": 0.52, "learning_rate": 9.993531159405981e-06, "loss": 0.3683, "step": 10318 }, { "epoch": 0.52, "learning_rate": 9.991913949574697e-06, "loss": 0.4125, "step": 10319 }, { "epoch": 0.52, "learning_rate": 9.990296739954895e-06, "loss": 0.357, "step": 10320 }, { "epoch": 0.52, "learning_rate": 9.988679530588864e-06, "loss": 0.2532, "step": 10321 }, { "epoch": 0.52, "learning_rate": 9.987062321518908e-06, "loss": 0.2212, "step": 10322 }, { "epoch": 0.52, "learning_rate": 9.985445112787322e-06, "loss": 0.5094, "step": 10323 }, { "epoch": 0.52, "learning_rate": 9.983827904436397e-06, "loss": 0.3138, "step": 10324 }, { "epoch": 0.52, "learning_rate": 9.982210696508435e-06, "loss": 0.167, "step": 10325 }, { "epoch": 0.52, "learning_rate": 9.980593489045722e-06, "loss": 0.321, "step": 10326 }, { "epoch": 0.52, "learning_rate": 9.978976282090565e-06, "loss": 0.4529, "step": 10327 }, { "epoch": 0.52, "learning_rate": 9.977359075685258e-06, "loss": 0.4178, "step": 10328 }, { "epoch": 0.52, "learning_rate": 9.975741869872093e-06, "loss": 0.3128, "step": 10329 }, { "epoch": 0.52, "learning_rate": 9.974124664693366e-06, "loss": 0.4143, "step": 10330 }, { "epoch": 0.52, "learning_rate": 9.972507460191378e-06, "loss": 0.2939, "step": 10331 }, { "epoch": 0.52, "learning_rate": 9.97089025640842e-06, "loss": 0.2388, "step": 10332 }, { "epoch": 0.52, "learning_rate": 9.969273053386788e-06, "loss": 0.285, "step": 10333 }, { "epoch": 0.52, "learning_rate": 9.96765585116878e-06, "loss": 0.3594, "step": 10334 }, { "epoch": 0.52, "learning_rate": 9.966038649796689e-06, "loss": 0.3764, "step": 10335 }, { "epoch": 0.52, "learning_rate": 9.964421449312816e-06, "loss": 0.5797, "step": 10336 }, { "epoch": 0.52, "learning_rate": 9.962804249759452e-06, "loss": 0.3038, "step": 10337 }, { "epoch": 0.52, "learning_rate": 9.961187051178894e-06, "loss": 0.3358, "step": 10338 }, { "epoch": 0.52, "learning_rate": 9.959569853613438e-06, "loss": 0.4541, "step": 10339 }, { "epoch": 0.52, "learning_rate": 9.957952657105376e-06, "loss": 0.3757, "step": 10340 }, { "epoch": 0.52, "learning_rate": 9.956335461697011e-06, "loss": 0.3932, "step": 10341 }, { "epoch": 0.52, "learning_rate": 9.954718267430635e-06, "loss": 0.146, "step": 10342 }, { "epoch": 0.52, "learning_rate": 9.953101074348543e-06, "loss": 0.2123, "step": 10343 }, { "epoch": 0.52, "learning_rate": 9.95148388249303e-06, "loss": 0.4582, "step": 10344 }, { "epoch": 0.52, "learning_rate": 9.949866691906393e-06, "loss": 0.3065, "step": 10345 }, { "epoch": 0.52, "learning_rate": 9.948249502630928e-06, "loss": 0.3256, "step": 10346 }, { "epoch": 0.52, "learning_rate": 9.946632314708928e-06, "loss": 0.2225, "step": 10347 }, { "epoch": 0.52, "learning_rate": 9.945015128182692e-06, "loss": 0.3778, "step": 10348 }, { "epoch": 0.52, "learning_rate": 9.943397943094509e-06, "loss": 0.3902, "step": 10349 }, { "epoch": 0.52, "learning_rate": 9.941780759486682e-06, "loss": 0.4093, "step": 10350 }, { "epoch": 0.52, "learning_rate": 9.940163577401503e-06, "loss": 0.3552, "step": 10351 }, { "epoch": 0.52, "learning_rate": 9.938546396881267e-06, "loss": 0.2743, "step": 10352 }, { "epoch": 0.52, "learning_rate": 9.936929217968265e-06, "loss": 0.3274, "step": 10353 }, { "epoch": 0.52, "learning_rate": 9.935312040704803e-06, "loss": 0.4037, "step": 10354 }, { "epoch": 0.52, "learning_rate": 9.933694865133167e-06, "loss": 0.5383, "step": 10355 }, { "epoch": 0.52, "learning_rate": 9.932077691295657e-06, "loss": 0.3428, "step": 10356 }, { "epoch": 0.52, "learning_rate": 9.930460519234564e-06, "loss": 0.2599, "step": 10357 }, { "epoch": 0.52, "learning_rate": 9.928843348992185e-06, "loss": 0.24, "step": 10358 }, { "epoch": 0.52, "learning_rate": 9.927226180610816e-06, "loss": 0.4314, "step": 10359 }, { "epoch": 0.52, "learning_rate": 9.925609014132751e-06, "loss": 0.3969, "step": 10360 }, { "epoch": 0.52, "learning_rate": 9.923991849600286e-06, "loss": 0.2919, "step": 10361 }, { "epoch": 0.52, "learning_rate": 9.922374687055709e-06, "loss": 0.6486, "step": 10362 }, { "epoch": 0.52, "learning_rate": 9.920757526541325e-06, "loss": 0.3984, "step": 10363 }, { "epoch": 0.52, "learning_rate": 9.919140368099425e-06, "loss": 0.363, "step": 10364 }, { "epoch": 0.52, "learning_rate": 9.917523211772302e-06, "loss": 0.3398, "step": 10365 }, { "epoch": 0.52, "learning_rate": 9.915906057602251e-06, "loss": 0.3609, "step": 10366 }, { "epoch": 0.52, "learning_rate": 9.914288905631566e-06, "loss": 0.5215, "step": 10367 }, { "epoch": 0.52, "learning_rate": 9.912671755902543e-06, "loss": 0.3273, "step": 10368 }, { "epoch": 0.52, "learning_rate": 9.911054608457474e-06, "loss": 0.2159, "step": 10369 }, { "epoch": 0.52, "learning_rate": 9.909437463338659e-06, "loss": 0.323, "step": 10370 }, { "epoch": 0.52, "learning_rate": 9.907820320588387e-06, "loss": 0.4358, "step": 10371 }, { "epoch": 0.52, "learning_rate": 9.906203180248952e-06, "loss": 0.2841, "step": 10372 }, { "epoch": 0.52, "learning_rate": 9.904586042362652e-06, "loss": 0.1944, "step": 10373 }, { "epoch": 0.52, "learning_rate": 9.90296890697178e-06, "loss": 0.2552, "step": 10374 }, { "epoch": 0.52, "learning_rate": 9.901351774118627e-06, "loss": 0.32, "step": 10375 }, { "epoch": 0.52, "learning_rate": 9.899734643845487e-06, "loss": 0.4743, "step": 10376 }, { "epoch": 0.52, "learning_rate": 9.898117516194659e-06, "loss": 0.3645, "step": 10377 }, { "epoch": 0.52, "learning_rate": 9.896500391208434e-06, "loss": 0.3821, "step": 10378 }, { "epoch": 0.52, "learning_rate": 9.894883268929106e-06, "loss": 0.1999, "step": 10379 }, { "epoch": 0.52, "learning_rate": 9.89326614939897e-06, "loss": 0.2177, "step": 10380 }, { "epoch": 0.52, "learning_rate": 9.891649032660314e-06, "loss": 0.4811, "step": 10381 }, { "epoch": 0.52, "learning_rate": 9.890031918755438e-06, "loss": 0.3227, "step": 10382 }, { "epoch": 0.52, "learning_rate": 9.888414807726634e-06, "loss": 0.4553, "step": 10383 }, { "epoch": 0.52, "learning_rate": 9.886797699616194e-06, "loss": 0.3172, "step": 10384 }, { "epoch": 0.52, "learning_rate": 9.885180594466406e-06, "loss": 0.308, "step": 10385 }, { "epoch": 0.52, "learning_rate": 9.883563492319576e-06, "loss": 0.3883, "step": 10386 }, { "epoch": 0.52, "learning_rate": 9.88194639321799e-06, "loss": 0.3608, "step": 10387 }, { "epoch": 0.52, "learning_rate": 9.88032929720394e-06, "loss": 0.4658, "step": 10388 }, { "epoch": 0.52, "learning_rate": 9.87871220431972e-06, "loss": 0.259, "step": 10389 }, { "epoch": 0.52, "learning_rate": 9.877095114607622e-06, "loss": 0.4241, "step": 10390 }, { "epoch": 0.52, "learning_rate": 9.875478028109944e-06, "loss": 0.4508, "step": 10391 }, { "epoch": 0.52, "learning_rate": 9.873860944868974e-06, "loss": 0.3133, "step": 10392 }, { "epoch": 0.52, "learning_rate": 9.872243864927006e-06, "loss": 0.6442, "step": 10393 }, { "epoch": 0.52, "learning_rate": 9.87062678832633e-06, "loss": 0.4309, "step": 10394 }, { "epoch": 0.52, "learning_rate": 9.869009715109242e-06, "loss": 0.3892, "step": 10395 }, { "epoch": 0.52, "learning_rate": 9.867392645318036e-06, "loss": 0.5372, "step": 10396 }, { "epoch": 0.52, "learning_rate": 9.865775578995e-06, "loss": 0.526, "step": 10397 }, { "epoch": 0.52, "learning_rate": 9.864158516182428e-06, "loss": 0.3436, "step": 10398 }, { "epoch": 0.52, "learning_rate": 9.862541456922607e-06, "loss": 0.3166, "step": 10399 }, { "epoch": 0.52, "learning_rate": 9.86092440125784e-06, "loss": 0.4351, "step": 10400 }, { "epoch": 0.52, "learning_rate": 9.859307349230413e-06, "loss": 0.5552, "step": 10401 }, { "epoch": 0.52, "learning_rate": 9.857690300882618e-06, "loss": 0.5171, "step": 10402 }, { "epoch": 0.52, "learning_rate": 9.856073256256745e-06, "loss": 0.3283, "step": 10403 }, { "epoch": 0.52, "learning_rate": 9.854456215395088e-06, "loss": 0.2452, "step": 10404 }, { "epoch": 0.52, "learning_rate": 9.852839178339936e-06, "loss": 0.4082, "step": 10405 }, { "epoch": 0.52, "learning_rate": 9.851222145133585e-06, "loss": 0.4968, "step": 10406 }, { "epoch": 0.52, "learning_rate": 9.849605115818325e-06, "loss": 0.2464, "step": 10407 }, { "epoch": 0.52, "learning_rate": 9.847988090436444e-06, "loss": 0.2465, "step": 10408 }, { "epoch": 0.52, "learning_rate": 9.846371069030236e-06, "loss": 0.2847, "step": 10409 }, { "epoch": 0.52, "learning_rate": 9.844754051641993e-06, "loss": 0.1519, "step": 10410 }, { "epoch": 0.52, "learning_rate": 9.843137038314004e-06, "loss": 0.3346, "step": 10411 }, { "epoch": 0.52, "learning_rate": 9.84152002908856e-06, "loss": 0.2505, "step": 10412 }, { "epoch": 0.52, "learning_rate": 9.839903024007948e-06, "loss": 0.44, "step": 10413 }, { "epoch": 0.52, "learning_rate": 9.838286023114466e-06, "loss": 0.2653, "step": 10414 }, { "epoch": 0.52, "learning_rate": 9.836669026450403e-06, "loss": 0.4152, "step": 10415 }, { "epoch": 0.52, "learning_rate": 9.835052034058047e-06, "loss": 0.178, "step": 10416 }, { "epoch": 0.52, "learning_rate": 9.833435045979686e-06, "loss": 0.4826, "step": 10417 }, { "epoch": 0.52, "learning_rate": 9.831818062257617e-06, "loss": 0.1776, "step": 10418 }, { "epoch": 0.52, "learning_rate": 9.830201082934125e-06, "loss": 0.3026, "step": 10419 }, { "epoch": 0.52, "learning_rate": 9.828584108051503e-06, "loss": 0.3559, "step": 10420 }, { "epoch": 0.52, "learning_rate": 9.826967137652036e-06, "loss": 0.5536, "step": 10421 }, { "epoch": 0.52, "learning_rate": 9.825350171778015e-06, "loss": 0.2642, "step": 10422 }, { "epoch": 0.52, "learning_rate": 9.823733210471735e-06, "loss": 0.3998, "step": 10423 }, { "epoch": 0.52, "learning_rate": 9.822116253775481e-06, "loss": 0.4328, "step": 10424 }, { "epoch": 0.52, "learning_rate": 9.820499301731546e-06, "loss": 0.4104, "step": 10425 }, { "epoch": 0.52, "learning_rate": 9.818882354382211e-06, "loss": 0.3563, "step": 10426 }, { "epoch": 0.52, "learning_rate": 9.817265411769776e-06, "loss": 0.3954, "step": 10427 }, { "epoch": 0.52, "learning_rate": 9.815648473936522e-06, "loss": 0.4521, "step": 10428 }, { "epoch": 0.52, "learning_rate": 9.814031540924743e-06, "loss": 0.4128, "step": 10429 }, { "epoch": 0.52, "learning_rate": 9.812414612776725e-06, "loss": 0.3269, "step": 10430 }, { "epoch": 0.52, "learning_rate": 9.810797689534756e-06, "loss": 0.2621, "step": 10431 }, { "epoch": 0.52, "learning_rate": 9.809180771241126e-06, "loss": 0.4957, "step": 10432 }, { "epoch": 0.52, "learning_rate": 9.807563857938125e-06, "loss": 0.308, "step": 10433 }, { "epoch": 0.52, "learning_rate": 9.805946949668037e-06, "loss": 0.3091, "step": 10434 }, { "epoch": 0.52, "learning_rate": 9.804330046473155e-06, "loss": 0.5605, "step": 10435 }, { "epoch": 0.52, "learning_rate": 9.802713148395757e-06, "loss": 0.3685, "step": 10436 }, { "epoch": 0.52, "learning_rate": 9.801096255478145e-06, "loss": 0.2376, "step": 10437 }, { "epoch": 0.52, "learning_rate": 9.7994793677626e-06, "loss": 0.3943, "step": 10438 }, { "epoch": 0.52, "learning_rate": 9.797862485291407e-06, "loss": 0.6807, "step": 10439 }, { "epoch": 0.52, "learning_rate": 9.796245608106855e-06, "loss": 0.1497, "step": 10440 }, { "epoch": 0.52, "learning_rate": 9.794628736251232e-06, "loss": 0.3595, "step": 10441 }, { "epoch": 0.52, "learning_rate": 9.793011869766828e-06, "loss": 0.4107, "step": 10442 }, { "epoch": 0.52, "learning_rate": 9.791395008695927e-06, "loss": 0.4623, "step": 10443 }, { "epoch": 0.52, "learning_rate": 9.789778153080818e-06, "loss": 0.4671, "step": 10444 }, { "epoch": 0.52, "learning_rate": 9.788161302963782e-06, "loss": 0.4176, "step": 10445 }, { "epoch": 0.52, "learning_rate": 9.786544458387113e-06, "loss": 0.4424, "step": 10446 }, { "epoch": 0.52, "learning_rate": 9.784927619393094e-06, "loss": 0.5754, "step": 10447 }, { "epoch": 0.52, "learning_rate": 9.78331078602401e-06, "loss": 0.3936, "step": 10448 }, { "epoch": 0.52, "learning_rate": 9.781693958322145e-06, "loss": 0.4263, "step": 10449 }, { "epoch": 0.52, "learning_rate": 9.780077136329794e-06, "loss": 0.5849, "step": 10450 }, { "epoch": 0.52, "learning_rate": 9.778460320089236e-06, "loss": 0.3937, "step": 10451 }, { "epoch": 0.52, "learning_rate": 9.77684350964276e-06, "loss": 0.4254, "step": 10452 }, { "epoch": 0.52, "learning_rate": 9.77522670503265e-06, "loss": 0.3879, "step": 10453 }, { "epoch": 0.52, "learning_rate": 9.773609906301186e-06, "loss": 0.29, "step": 10454 }, { "epoch": 0.52, "learning_rate": 9.771993113490661e-06, "loss": 0.1652, "step": 10455 }, { "epoch": 0.52, "learning_rate": 9.77037632664336e-06, "loss": 0.4968, "step": 10456 }, { "epoch": 0.52, "learning_rate": 9.768759545801563e-06, "loss": 0.5978, "step": 10457 }, { "epoch": 0.52, "learning_rate": 9.767142771007554e-06, "loss": 0.4274, "step": 10458 }, { "epoch": 0.52, "learning_rate": 9.765526002303626e-06, "loss": 0.3497, "step": 10459 }, { "epoch": 0.52, "learning_rate": 9.763909239732057e-06, "loss": 0.5107, "step": 10460 }, { "epoch": 0.52, "learning_rate": 9.762292483335132e-06, "loss": 0.4257, "step": 10461 }, { "epoch": 0.52, "learning_rate": 9.760675733155137e-06, "loss": 0.4877, "step": 10462 }, { "epoch": 0.52, "learning_rate": 9.75905898923435e-06, "loss": 0.2891, "step": 10463 }, { "epoch": 0.52, "learning_rate": 9.757442251615065e-06, "loss": 0.3157, "step": 10464 }, { "epoch": 0.52, "learning_rate": 9.755825520339559e-06, "loss": 0.3434, "step": 10465 }, { "epoch": 0.52, "learning_rate": 9.754208795450117e-06, "loss": 0.3841, "step": 10466 }, { "epoch": 0.52, "learning_rate": 9.752592076989019e-06, "loss": 0.4312, "step": 10467 }, { "epoch": 0.52, "learning_rate": 9.750975364998556e-06, "loss": 0.4215, "step": 10468 }, { "epoch": 0.52, "learning_rate": 9.749358659521003e-06, "loss": 0.2495, "step": 10469 }, { "epoch": 0.52, "learning_rate": 9.747741960598646e-06, "loss": 0.2797, "step": 10470 }, { "epoch": 0.52, "learning_rate": 9.74612526827377e-06, "loss": 0.5238, "step": 10471 }, { "epoch": 0.52, "learning_rate": 9.74450858258865e-06, "loss": 0.1828, "step": 10472 }, { "epoch": 0.52, "learning_rate": 9.742891903585578e-06, "loss": 0.3261, "step": 10473 }, { "epoch": 0.52, "learning_rate": 9.74127523130683e-06, "loss": 0.459, "step": 10474 }, { "epoch": 0.52, "learning_rate": 9.739658565794691e-06, "loss": 0.3281, "step": 10475 }, { "epoch": 0.52, "learning_rate": 9.73804190709144e-06, "loss": 0.5726, "step": 10476 }, { "epoch": 0.52, "learning_rate": 9.736425255239356e-06, "loss": 0.2991, "step": 10477 }, { "epoch": 0.52, "learning_rate": 9.734808610280729e-06, "loss": 0.3032, "step": 10478 }, { "epoch": 0.52, "learning_rate": 9.733191972257835e-06, "loss": 0.338, "step": 10479 }, { "epoch": 0.52, "learning_rate": 9.731575341212955e-06, "loss": 0.6143, "step": 10480 }, { "epoch": 0.52, "learning_rate": 9.729958717188368e-06, "loss": 0.2176, "step": 10481 }, { "epoch": 0.52, "learning_rate": 9.72834210022636e-06, "loss": 0.3942, "step": 10482 }, { "epoch": 0.52, "learning_rate": 9.72672549036921e-06, "loss": 0.4456, "step": 10483 }, { "epoch": 0.52, "learning_rate": 9.725108887659195e-06, "loss": 0.2956, "step": 10484 }, { "epoch": 0.52, "learning_rate": 9.723492292138598e-06, "loss": 0.2611, "step": 10485 }, { "epoch": 0.52, "learning_rate": 9.721875703849692e-06, "loss": 0.3847, "step": 10486 }, { "epoch": 0.52, "learning_rate": 9.720259122834768e-06, "loss": 0.4054, "step": 10487 }, { "epoch": 0.52, "learning_rate": 9.718642549136102e-06, "loss": 0.5249, "step": 10488 }, { "epoch": 0.52, "learning_rate": 9.71702598279597e-06, "loss": 0.2857, "step": 10489 }, { "epoch": 0.52, "learning_rate": 9.71540942385665e-06, "loss": 0.2212, "step": 10490 }, { "epoch": 0.52, "learning_rate": 9.713792872360427e-06, "loss": 0.2473, "step": 10491 }, { "epoch": 0.52, "learning_rate": 9.712176328349575e-06, "loss": 0.3731, "step": 10492 }, { "epoch": 0.52, "learning_rate": 9.710559791866375e-06, "loss": 0.2175, "step": 10493 }, { "epoch": 0.52, "learning_rate": 9.708943262953105e-06, "loss": 0.5043, "step": 10494 }, { "epoch": 0.52, "learning_rate": 9.707326741652039e-06, "loss": 0.3693, "step": 10495 }, { "epoch": 0.52, "learning_rate": 9.705710228005462e-06, "loss": 0.2293, "step": 10496 }, { "epoch": 0.52, "learning_rate": 9.704093722055648e-06, "loss": 0.4165, "step": 10497 }, { "epoch": 0.52, "learning_rate": 9.702477223844876e-06, "loss": 0.4402, "step": 10498 }, { "epoch": 0.52, "learning_rate": 9.700860733415418e-06, "loss": 0.4294, "step": 10499 }, { "epoch": 0.52, "learning_rate": 9.699244250809562e-06, "loss": 0.4985, "step": 10500 }, { "epoch": 0.52, "learning_rate": 9.697627776069576e-06, "loss": 0.3063, "step": 10501 }, { "epoch": 0.52, "learning_rate": 9.696011309237741e-06, "loss": 0.5539, "step": 10502 }, { "epoch": 0.52, "learning_rate": 9.69439485035633e-06, "loss": 0.1443, "step": 10503 }, { "epoch": 0.52, "learning_rate": 9.69277839946762e-06, "loss": 0.1891, "step": 10504 }, { "epoch": 0.52, "learning_rate": 9.691161956613892e-06, "loss": 0.3914, "step": 10505 }, { "epoch": 0.52, "learning_rate": 9.689545521837418e-06, "loss": 0.5123, "step": 10506 }, { "epoch": 0.52, "learning_rate": 9.687929095180474e-06, "loss": 0.4542, "step": 10507 }, { "epoch": 0.52, "learning_rate": 9.686312676685337e-06, "loss": 0.536, "step": 10508 }, { "epoch": 0.52, "learning_rate": 9.684696266394276e-06, "loss": 0.4243, "step": 10509 }, { "epoch": 0.52, "learning_rate": 9.683079864349573e-06, "loss": 0.4048, "step": 10510 }, { "epoch": 0.52, "learning_rate": 9.681463470593505e-06, "loss": 0.1119, "step": 10511 }, { "epoch": 0.52, "learning_rate": 9.67984708516834e-06, "loss": 0.3324, "step": 10512 }, { "epoch": 0.52, "learning_rate": 9.678230708116353e-06, "loss": 0.2474, "step": 10513 }, { "epoch": 0.52, "learning_rate": 9.676614339479822e-06, "loss": 0.5131, "step": 10514 }, { "epoch": 0.53, "learning_rate": 9.674997979301021e-06, "loss": 0.306, "step": 10515 }, { "epoch": 0.53, "learning_rate": 9.673381627622223e-06, "loss": 0.0831, "step": 10516 }, { "epoch": 0.53, "learning_rate": 9.6717652844857e-06, "loss": 0.291, "step": 10517 }, { "epoch": 0.53, "learning_rate": 9.670148949933724e-06, "loss": 0.4614, "step": 10518 }, { "epoch": 0.53, "learning_rate": 9.668532624008572e-06, "loss": 0.308, "step": 10519 }, { "epoch": 0.53, "learning_rate": 9.666916306752515e-06, "loss": 0.6118, "step": 10520 }, { "epoch": 0.53, "learning_rate": 9.665299998207826e-06, "loss": 0.1573, "step": 10521 }, { "epoch": 0.53, "learning_rate": 9.663683698416772e-06, "loss": 0.5751, "step": 10522 }, { "epoch": 0.53, "learning_rate": 9.662067407421636e-06, "loss": 0.3555, "step": 10523 }, { "epoch": 0.53, "learning_rate": 9.660451125264683e-06, "loss": 0.3556, "step": 10524 }, { "epoch": 0.53, "learning_rate": 9.658834851988188e-06, "loss": 0.3407, "step": 10525 }, { "epoch": 0.53, "learning_rate": 9.657218587634419e-06, "loss": 0.3117, "step": 10526 }, { "epoch": 0.53, "learning_rate": 9.655602332245648e-06, "loss": 0.5678, "step": 10527 }, { "epoch": 0.53, "learning_rate": 9.653986085864148e-06, "loss": 0.318, "step": 10528 }, { "epoch": 0.53, "learning_rate": 9.652369848532188e-06, "loss": 0.5002, "step": 10529 }, { "epoch": 0.53, "learning_rate": 9.65075362029204e-06, "loss": 0.1111, "step": 10530 }, { "epoch": 0.53, "learning_rate": 9.649137401185971e-06, "loss": 0.4631, "step": 10531 }, { "epoch": 0.53, "learning_rate": 9.647521191256256e-06, "loss": 0.4228, "step": 10532 }, { "epoch": 0.53, "learning_rate": 9.645904990545163e-06, "loss": 0.5699, "step": 10533 }, { "epoch": 0.53, "learning_rate": 9.644288799094962e-06, "loss": 0.2806, "step": 10534 }, { "epoch": 0.53, "learning_rate": 9.642672616947921e-06, "loss": 0.2665, "step": 10535 }, { "epoch": 0.53, "learning_rate": 9.641056444146304e-06, "loss": 0.3882, "step": 10536 }, { "epoch": 0.53, "learning_rate": 9.639440280732393e-06, "loss": 0.3543, "step": 10537 }, { "epoch": 0.53, "learning_rate": 9.637824126748447e-06, "loss": 0.3813, "step": 10538 }, { "epoch": 0.53, "learning_rate": 9.636207982236738e-06, "loss": 0.2391, "step": 10539 }, { "epoch": 0.53, "learning_rate": 9.634591847239531e-06, "loss": 0.2551, "step": 10540 }, { "epoch": 0.53, "learning_rate": 9.632975721799095e-06, "loss": 0.2781, "step": 10541 }, { "epoch": 0.53, "learning_rate": 9.6313596059577e-06, "loss": 0.4232, "step": 10542 }, { "epoch": 0.53, "learning_rate": 9.62974349975761e-06, "loss": 0.2698, "step": 10543 }, { "epoch": 0.53, "learning_rate": 9.628127403241096e-06, "loss": 0.569, "step": 10544 }, { "epoch": 0.53, "learning_rate": 9.626511316450417e-06, "loss": 0.3525, "step": 10545 }, { "epoch": 0.53, "learning_rate": 9.62489523942785e-06, "loss": 0.3146, "step": 10546 }, { "epoch": 0.53, "learning_rate": 9.623279172215656e-06, "loss": 0.2909, "step": 10547 }, { "epoch": 0.53, "learning_rate": 9.621663114856104e-06, "loss": 0.3141, "step": 10548 }, { "epoch": 0.53, "learning_rate": 9.620047067391455e-06, "loss": 0.4637, "step": 10549 }, { "epoch": 0.53, "learning_rate": 9.618431029863975e-06, "loss": 0.3527, "step": 10550 }, { "epoch": 0.53, "learning_rate": 9.616815002315937e-06, "loss": 0.444, "step": 10551 }, { "epoch": 0.53, "learning_rate": 9.615198984789599e-06, "loss": 0.3736, "step": 10552 }, { "epoch": 0.53, "learning_rate": 9.613582977327228e-06, "loss": 0.3697, "step": 10553 }, { "epoch": 0.53, "learning_rate": 9.611966979971085e-06, "loss": 0.5535, "step": 10554 }, { "epoch": 0.53, "learning_rate": 9.61035099276344e-06, "loss": 0.2634, "step": 10555 }, { "epoch": 0.53, "learning_rate": 9.608735015746555e-06, "loss": 0.2402, "step": 10556 }, { "epoch": 0.53, "learning_rate": 9.607119048962692e-06, "loss": 0.3684, "step": 10557 }, { "epoch": 0.53, "learning_rate": 9.605503092454117e-06, "loss": 0.2052, "step": 10558 }, { "epoch": 0.53, "learning_rate": 9.603887146263087e-06, "loss": 0.4146, "step": 10559 }, { "epoch": 0.53, "learning_rate": 9.602271210431875e-06, "loss": 0.3378, "step": 10560 }, { "epoch": 0.53, "learning_rate": 9.600655285002737e-06, "loss": 0.4642, "step": 10561 }, { "epoch": 0.53, "learning_rate": 9.599039370017938e-06, "loss": 0.3804, "step": 10562 }, { "epoch": 0.53, "learning_rate": 9.597423465519737e-06, "loss": 0.2623, "step": 10563 }, { "epoch": 0.53, "learning_rate": 9.5958075715504e-06, "loss": 0.5233, "step": 10564 }, { "epoch": 0.53, "learning_rate": 9.594191688152186e-06, "loss": 0.2452, "step": 10565 }, { "epoch": 0.53, "learning_rate": 9.59257581536736e-06, "loss": 0.4629, "step": 10566 }, { "epoch": 0.53, "learning_rate": 9.590959953238176e-06, "loss": 0.273, "step": 10567 }, { "epoch": 0.53, "learning_rate": 9.589344101806896e-06, "loss": 0.347, "step": 10568 }, { "epoch": 0.53, "learning_rate": 9.587728261115788e-06, "loss": 0.3853, "step": 10569 }, { "epoch": 0.53, "learning_rate": 9.586112431207105e-06, "loss": 0.4858, "step": 10570 }, { "epoch": 0.53, "learning_rate": 9.584496612123111e-06, "loss": 0.4028, "step": 10571 }, { "epoch": 0.53, "learning_rate": 9.582880803906064e-06, "loss": 0.4338, "step": 10572 }, { "epoch": 0.53, "learning_rate": 9.581265006598217e-06, "loss": 0.2299, "step": 10573 }, { "epoch": 0.53, "learning_rate": 9.579649220241842e-06, "loss": 0.4429, "step": 10574 }, { "epoch": 0.53, "learning_rate": 9.57803344487919e-06, "loss": 0.3744, "step": 10575 }, { "epoch": 0.53, "learning_rate": 9.576417680552521e-06, "loss": 0.4278, "step": 10576 }, { "epoch": 0.53, "learning_rate": 9.574801927304091e-06, "loss": 0.3924, "step": 10577 }, { "epoch": 0.53, "learning_rate": 9.573186185176163e-06, "loss": 0.366, "step": 10578 }, { "epoch": 0.53, "learning_rate": 9.571570454210989e-06, "loss": 0.4746, "step": 10579 }, { "epoch": 0.53, "learning_rate": 9.56995473445083e-06, "loss": 0.3165, "step": 10580 }, { "epoch": 0.53, "learning_rate": 9.56833902593794e-06, "loss": 0.2784, "step": 10581 }, { "epoch": 0.53, "learning_rate": 9.566723328714575e-06, "loss": 0.3389, "step": 10582 }, { "epoch": 0.53, "learning_rate": 9.565107642822996e-06, "loss": 0.58, "step": 10583 }, { "epoch": 0.53, "learning_rate": 9.563491968305458e-06, "loss": 0.4232, "step": 10584 }, { "epoch": 0.53, "learning_rate": 9.561876305204216e-06, "loss": 0.4128, "step": 10585 }, { "epoch": 0.53, "learning_rate": 9.560260653561521e-06, "loss": 0.4501, "step": 10586 }, { "epoch": 0.53, "learning_rate": 9.558645013419638e-06, "loss": 0.2907, "step": 10587 }, { "epoch": 0.53, "learning_rate": 9.557029384820815e-06, "loss": 0.493, "step": 10588 }, { "epoch": 0.53, "learning_rate": 9.55541376780731e-06, "loss": 0.2502, "step": 10589 }, { "epoch": 0.53, "learning_rate": 9.553798162421374e-06, "loss": 0.4478, "step": 10590 }, { "epoch": 0.53, "learning_rate": 9.552182568705262e-06, "loss": 0.5321, "step": 10591 }, { "epoch": 0.53, "learning_rate": 9.550566986701228e-06, "loss": 0.35, "step": 10592 }, { "epoch": 0.53, "learning_rate": 9.548951416451528e-06, "loss": 0.4863, "step": 10593 }, { "epoch": 0.53, "learning_rate": 9.547335857998412e-06, "loss": 0.3188, "step": 10594 }, { "epoch": 0.53, "learning_rate": 9.54572031138413e-06, "loss": 0.6001, "step": 10595 }, { "epoch": 0.53, "learning_rate": 9.544104776650942e-06, "loss": 0.4154, "step": 10596 }, { "epoch": 0.53, "learning_rate": 9.542489253841097e-06, "loss": 0.2067, "step": 10597 }, { "epoch": 0.53, "learning_rate": 9.540873742996847e-06, "loss": 0.314, "step": 10598 }, { "epoch": 0.53, "learning_rate": 9.539258244160442e-06, "loss": 0.4821, "step": 10599 }, { "epoch": 0.53, "learning_rate": 9.537642757374132e-06, "loss": 0.362, "step": 10600 }, { "epoch": 0.53, "learning_rate": 9.536027282680172e-06, "loss": 0.3319, "step": 10601 }, { "epoch": 0.53, "learning_rate": 9.534411820120811e-06, "loss": 0.3245, "step": 10602 }, { "epoch": 0.53, "learning_rate": 9.532796369738298e-06, "loss": 0.5541, "step": 10603 }, { "epoch": 0.53, "learning_rate": 9.531180931574882e-06, "loss": 0.308, "step": 10604 }, { "epoch": 0.53, "learning_rate": 9.529565505672817e-06, "loss": 0.5345, "step": 10605 }, { "epoch": 0.53, "learning_rate": 9.527950092074352e-06, "loss": 0.4825, "step": 10606 }, { "epoch": 0.53, "learning_rate": 9.526334690821732e-06, "loss": 0.3412, "step": 10607 }, { "epoch": 0.53, "learning_rate": 9.524719301957209e-06, "loss": 0.3926, "step": 10608 }, { "epoch": 0.53, "learning_rate": 9.523103925523025e-06, "loss": 0.3937, "step": 10609 }, { "epoch": 0.53, "learning_rate": 9.521488561561439e-06, "loss": 0.3007, "step": 10610 }, { "epoch": 0.53, "learning_rate": 9.519873210114693e-06, "loss": 0.3009, "step": 10611 }, { "epoch": 0.53, "learning_rate": 9.518257871225032e-06, "loss": 0.3339, "step": 10612 }, { "epoch": 0.53, "learning_rate": 9.516642544934707e-06, "loss": 0.3958, "step": 10613 }, { "epoch": 0.53, "learning_rate": 9.51502723128596e-06, "loss": 0.3609, "step": 10614 }, { "epoch": 0.53, "learning_rate": 9.513411930321044e-06, "loss": 0.2736, "step": 10615 }, { "epoch": 0.53, "learning_rate": 9.5117966420822e-06, "loss": 0.198, "step": 10616 }, { "epoch": 0.53, "learning_rate": 9.510181366611678e-06, "loss": 0.4563, "step": 10617 }, { "epoch": 0.53, "learning_rate": 9.508566103951715e-06, "loss": 0.3574, "step": 10618 }, { "epoch": 0.53, "learning_rate": 9.506950854144565e-06, "loss": 0.3111, "step": 10619 }, { "epoch": 0.53, "learning_rate": 9.505335617232471e-06, "loss": 0.4256, "step": 10620 }, { "epoch": 0.53, "learning_rate": 9.503720393257677e-06, "loss": 0.4399, "step": 10621 }, { "epoch": 0.53, "learning_rate": 9.502105182262424e-06, "loss": 0.3996, "step": 10622 }, { "epoch": 0.53, "learning_rate": 9.500489984288954e-06, "loss": 0.3568, "step": 10623 }, { "epoch": 0.53, "learning_rate": 9.498874799379518e-06, "loss": 0.1853, "step": 10624 }, { "epoch": 0.53, "learning_rate": 9.497259627576358e-06, "loss": 0.2731, "step": 10625 }, { "epoch": 0.53, "learning_rate": 9.49564446892171e-06, "loss": 0.3965, "step": 10626 }, { "epoch": 0.53, "learning_rate": 9.494029323457822e-06, "loss": 0.3209, "step": 10627 }, { "epoch": 0.53, "learning_rate": 9.492414191226935e-06, "loss": 0.1978, "step": 10628 }, { "epoch": 0.53, "learning_rate": 9.49079907227129e-06, "loss": 0.2612, "step": 10629 }, { "epoch": 0.53, "learning_rate": 9.489183966633128e-06, "loss": 0.472, "step": 10630 }, { "epoch": 0.53, "learning_rate": 9.48756887435469e-06, "loss": 0.2337, "step": 10631 }, { "epoch": 0.53, "learning_rate": 9.485953795478212e-06, "loss": 0.5374, "step": 10632 }, { "epoch": 0.53, "learning_rate": 9.484338730045946e-06, "loss": 0.5872, "step": 10633 }, { "epoch": 0.53, "learning_rate": 9.482723678100125e-06, "loss": 0.2452, "step": 10634 }, { "epoch": 0.53, "learning_rate": 9.481108639682987e-06, "loss": 0.374, "step": 10635 }, { "epoch": 0.53, "learning_rate": 9.47949361483677e-06, "loss": 0.1996, "step": 10636 }, { "epoch": 0.53, "learning_rate": 9.47787860360372e-06, "loss": 0.3579, "step": 10637 }, { "epoch": 0.53, "learning_rate": 9.47626360602607e-06, "loss": 0.3354, "step": 10638 }, { "epoch": 0.53, "learning_rate": 9.47464862214606e-06, "loss": 0.3662, "step": 10639 }, { "epoch": 0.53, "learning_rate": 9.473033652005928e-06, "loss": 0.2661, "step": 10640 }, { "epoch": 0.53, "learning_rate": 9.471418695647903e-06, "loss": 0.2532, "step": 10641 }, { "epoch": 0.53, "learning_rate": 9.469803753114238e-06, "loss": 0.3434, "step": 10642 }, { "epoch": 0.53, "learning_rate": 9.46818882444716e-06, "loss": 0.3868, "step": 10643 }, { "epoch": 0.53, "learning_rate": 9.466573909688906e-06, "loss": 0.1156, "step": 10644 }, { "epoch": 0.53, "learning_rate": 9.464959008881713e-06, "loss": 0.2338, "step": 10645 }, { "epoch": 0.53, "learning_rate": 9.463344122067812e-06, "loss": 0.3744, "step": 10646 }, { "epoch": 0.53, "learning_rate": 9.461729249289447e-06, "loss": 0.3177, "step": 10647 }, { "epoch": 0.53, "learning_rate": 9.46011439058885e-06, "loss": 0.3027, "step": 10648 }, { "epoch": 0.53, "learning_rate": 9.458499546008253e-06, "loss": 0.3557, "step": 10649 }, { "epoch": 0.53, "learning_rate": 9.456884715589887e-06, "loss": 0.6746, "step": 10650 }, { "epoch": 0.53, "learning_rate": 9.455269899375995e-06, "loss": 0.2606, "step": 10651 }, { "epoch": 0.53, "learning_rate": 9.453655097408803e-06, "loss": 0.3832, "step": 10652 }, { "epoch": 0.53, "learning_rate": 9.452040309730546e-06, "loss": 0.3258, "step": 10653 }, { "epoch": 0.53, "learning_rate": 9.45042553638346e-06, "loss": 0.3183, "step": 10654 }, { "epoch": 0.53, "learning_rate": 9.448810777409769e-06, "loss": 0.5174, "step": 10655 }, { "epoch": 0.53, "learning_rate": 9.447196032851713e-06, "loss": 0.2396, "step": 10656 }, { "epoch": 0.53, "learning_rate": 9.445581302751521e-06, "loss": 0.2241, "step": 10657 }, { "epoch": 0.53, "learning_rate": 9.443966587151425e-06, "loss": 0.3246, "step": 10658 }, { "epoch": 0.53, "learning_rate": 9.442351886093649e-06, "loss": 0.6109, "step": 10659 }, { "epoch": 0.53, "learning_rate": 9.440737199620435e-06, "loss": 0.1898, "step": 10660 }, { "epoch": 0.53, "learning_rate": 9.439122527774004e-06, "loss": 0.2163, "step": 10661 }, { "epoch": 0.53, "learning_rate": 9.437507870596592e-06, "loss": 0.3318, "step": 10662 }, { "epoch": 0.53, "learning_rate": 9.435893228130425e-06, "loss": 0.2758, "step": 10663 }, { "epoch": 0.53, "learning_rate": 9.434278600417726e-06, "loss": 0.277, "step": 10664 }, { "epoch": 0.53, "learning_rate": 9.432663987500735e-06, "loss": 0.3867, "step": 10665 }, { "epoch": 0.53, "learning_rate": 9.431049389421672e-06, "loss": 0.3567, "step": 10666 }, { "epoch": 0.53, "learning_rate": 9.429434806222769e-06, "loss": 0.305, "step": 10667 }, { "epoch": 0.53, "learning_rate": 9.427820237946245e-06, "loss": 0.241, "step": 10668 }, { "epoch": 0.53, "learning_rate": 9.426205684634339e-06, "loss": 0.4609, "step": 10669 }, { "epoch": 0.53, "learning_rate": 9.424591146329272e-06, "loss": 0.2185, "step": 10670 }, { "epoch": 0.53, "learning_rate": 9.42297662307327e-06, "loss": 0.1131, "step": 10671 }, { "epoch": 0.53, "learning_rate": 9.421362114908558e-06, "loss": 0.4063, "step": 10672 }, { "epoch": 0.53, "learning_rate": 9.41974762187736e-06, "loss": 0.5921, "step": 10673 }, { "epoch": 0.53, "learning_rate": 9.418133144021906e-06, "loss": 0.5447, "step": 10674 }, { "epoch": 0.53, "learning_rate": 9.416518681384414e-06, "loss": 0.2454, "step": 10675 }, { "epoch": 0.53, "learning_rate": 9.414904234007115e-06, "loss": 0.3958, "step": 10676 }, { "epoch": 0.53, "learning_rate": 9.413289801932229e-06, "loss": 0.2587, "step": 10677 }, { "epoch": 0.53, "learning_rate": 9.411675385201975e-06, "loss": 0.4084, "step": 10678 }, { "epoch": 0.53, "learning_rate": 9.410060983858584e-06, "loss": 0.4598, "step": 10679 }, { "epoch": 0.53, "learning_rate": 9.408446597944278e-06, "loss": 0.3241, "step": 10680 }, { "epoch": 0.53, "learning_rate": 9.406832227501273e-06, "loss": 0.4903, "step": 10681 }, { "epoch": 0.53, "learning_rate": 9.405217872571793e-06, "loss": 0.2171, "step": 10682 }, { "epoch": 0.53, "learning_rate": 9.403603533198063e-06, "loss": 0.2783, "step": 10683 }, { "epoch": 0.53, "learning_rate": 9.401989209422302e-06, "loss": 0.5326, "step": 10684 }, { "epoch": 0.53, "learning_rate": 9.40037490128673e-06, "loss": 0.2613, "step": 10685 }, { "epoch": 0.53, "learning_rate": 9.398760608833567e-06, "loss": 0.2316, "step": 10686 }, { "epoch": 0.53, "learning_rate": 9.39714633210503e-06, "loss": 0.2867, "step": 10687 }, { "epoch": 0.53, "learning_rate": 9.395532071143343e-06, "loss": 0.297, "step": 10688 }, { "epoch": 0.53, "learning_rate": 9.393917825990724e-06, "loss": 0.2893, "step": 10689 }, { "epoch": 0.53, "learning_rate": 9.392303596689391e-06, "loss": 0.2325, "step": 10690 }, { "epoch": 0.53, "learning_rate": 9.390689383281558e-06, "loss": 0.2465, "step": 10691 }, { "epoch": 0.53, "learning_rate": 9.389075185809448e-06, "loss": 0.5634, "step": 10692 }, { "epoch": 0.53, "learning_rate": 9.387461004315277e-06, "loss": 0.23, "step": 10693 }, { "epoch": 0.53, "learning_rate": 9.385846838841261e-06, "loss": 0.4869, "step": 10694 }, { "epoch": 0.53, "learning_rate": 9.384232689429618e-06, "loss": 0.4233, "step": 10695 }, { "epoch": 0.53, "learning_rate": 9.382618556122556e-06, "loss": 0.3234, "step": 10696 }, { "epoch": 0.53, "learning_rate": 9.381004438962304e-06, "loss": 0.3121, "step": 10697 }, { "epoch": 0.53, "learning_rate": 9.379390337991067e-06, "loss": 0.2624, "step": 10698 }, { "epoch": 0.53, "learning_rate": 9.377776253251065e-06, "loss": 0.2099, "step": 10699 }, { "epoch": 0.53, "learning_rate": 9.376162184784508e-06, "loss": 0.3602, "step": 10700 }, { "epoch": 0.53, "learning_rate": 9.374548132633612e-06, "loss": 0.2521, "step": 10701 }, { "epoch": 0.53, "learning_rate": 9.372934096840593e-06, "loss": 0.4191, "step": 10702 }, { "epoch": 0.53, "learning_rate": 9.37132007744766e-06, "loss": 0.1988, "step": 10703 }, { "epoch": 0.53, "learning_rate": 9.369706074497028e-06, "loss": 0.2407, "step": 10704 }, { "epoch": 0.53, "learning_rate": 9.368092088030903e-06, "loss": 0.3126, "step": 10705 }, { "epoch": 0.53, "learning_rate": 9.366478118091506e-06, "loss": 0.1771, "step": 10706 }, { "epoch": 0.53, "learning_rate": 9.364864164721044e-06, "loss": 0.4315, "step": 10707 }, { "epoch": 0.53, "learning_rate": 9.363250227961729e-06, "loss": 0.4664, "step": 10708 }, { "epoch": 0.53, "learning_rate": 9.36163630785577e-06, "loss": 0.4199, "step": 10709 }, { "epoch": 0.53, "learning_rate": 9.360022404445373e-06, "loss": 0.1427, "step": 10710 }, { "epoch": 0.53, "learning_rate": 9.358408517772757e-06, "loss": 0.4036, "step": 10711 }, { "epoch": 0.53, "learning_rate": 9.356794647880125e-06, "loss": 0.2941, "step": 10712 }, { "epoch": 0.53, "learning_rate": 9.355180794809685e-06, "loss": 0.357, "step": 10713 }, { "epoch": 0.53, "learning_rate": 9.353566958603643e-06, "loss": 0.3791, "step": 10714 }, { "epoch": 0.54, "learning_rate": 9.351953139304215e-06, "loss": 0.2485, "step": 10715 }, { "epoch": 0.54, "learning_rate": 9.350339336953604e-06, "loss": 0.4704, "step": 10716 }, { "epoch": 0.54, "learning_rate": 9.348725551594017e-06, "loss": 0.4389, "step": 10717 }, { "epoch": 0.54, "learning_rate": 9.34711178326766e-06, "loss": 0.2868, "step": 10718 }, { "epoch": 0.54, "learning_rate": 9.345498032016736e-06, "loss": 0.5246, "step": 10719 }, { "epoch": 0.54, "learning_rate": 9.343884297883456e-06, "loss": 0.3253, "step": 10720 }, { "epoch": 0.54, "learning_rate": 9.342270580910023e-06, "loss": 0.5272, "step": 10721 }, { "epoch": 0.54, "learning_rate": 9.340656881138644e-06, "loss": 0.2714, "step": 10722 }, { "epoch": 0.54, "learning_rate": 9.339043198611516e-06, "loss": 0.3286, "step": 10723 }, { "epoch": 0.54, "learning_rate": 9.33742953337085e-06, "loss": 0.3896, "step": 10724 }, { "epoch": 0.54, "learning_rate": 9.335815885458847e-06, "loss": 0.3746, "step": 10725 }, { "epoch": 0.54, "learning_rate": 9.33420225491771e-06, "loss": 0.2561, "step": 10726 }, { "epoch": 0.54, "learning_rate": 9.33258864178964e-06, "loss": 0.2735, "step": 10727 }, { "epoch": 0.54, "learning_rate": 9.330975046116837e-06, "loss": 0.256, "step": 10728 }, { "epoch": 0.54, "learning_rate": 9.32936146794151e-06, "loss": 0.4344, "step": 10729 }, { "epoch": 0.54, "learning_rate": 9.327747907305852e-06, "loss": 0.5995, "step": 10730 }, { "epoch": 0.54, "learning_rate": 9.32613436425207e-06, "loss": 0.4072, "step": 10731 }, { "epoch": 0.54, "learning_rate": 9.324520838822357e-06, "loss": 0.4563, "step": 10732 }, { "epoch": 0.54, "learning_rate": 9.32290733105892e-06, "loss": 0.3463, "step": 10733 }, { "epoch": 0.54, "learning_rate": 9.321293841003956e-06, "loss": 0.4433, "step": 10734 }, { "epoch": 0.54, "learning_rate": 9.319680368699661e-06, "loss": 0.3763, "step": 10735 }, { "epoch": 0.54, "learning_rate": 9.318066914188235e-06, "loss": 0.29, "step": 10736 }, { "epoch": 0.54, "learning_rate": 9.316453477511874e-06, "loss": 0.234, "step": 10737 }, { "epoch": 0.54, "learning_rate": 9.31484005871278e-06, "loss": 0.392, "step": 10738 }, { "epoch": 0.54, "learning_rate": 9.313226657833146e-06, "loss": 0.406, "step": 10739 }, { "epoch": 0.54, "learning_rate": 9.31161327491517e-06, "loss": 0.2631, "step": 10740 }, { "epoch": 0.54, "learning_rate": 9.309999910001042e-06, "loss": 0.3462, "step": 10741 }, { "epoch": 0.54, "learning_rate": 9.308386563132967e-06, "loss": 0.4162, "step": 10742 }, { "epoch": 0.54, "learning_rate": 9.306773234353137e-06, "loss": 0.4117, "step": 10743 }, { "epoch": 0.54, "learning_rate": 9.305159923703744e-06, "loss": 0.2667, "step": 10744 }, { "epoch": 0.54, "learning_rate": 9.303546631226983e-06, "loss": 0.4085, "step": 10745 }, { "epoch": 0.54, "learning_rate": 9.301933356965044e-06, "loss": 0.3305, "step": 10746 }, { "epoch": 0.54, "learning_rate": 9.300320100960128e-06, "loss": 0.3691, "step": 10747 }, { "epoch": 0.54, "learning_rate": 9.298706863254424e-06, "loss": 0.2882, "step": 10748 }, { "epoch": 0.54, "learning_rate": 9.297093643890122e-06, "loss": 0.3124, "step": 10749 }, { "epoch": 0.54, "learning_rate": 9.295480442909416e-06, "loss": 0.2687, "step": 10750 }, { "epoch": 0.54, "learning_rate": 9.293867260354491e-06, "loss": 0.4083, "step": 10751 }, { "epoch": 0.54, "learning_rate": 9.292254096267548e-06, "loss": 0.2576, "step": 10752 }, { "epoch": 0.54, "learning_rate": 9.29064095069077e-06, "loss": 0.4644, "step": 10753 }, { "epoch": 0.54, "learning_rate": 9.289027823666352e-06, "loss": 0.331, "step": 10754 }, { "epoch": 0.54, "learning_rate": 9.287414715236474e-06, "loss": 0.3455, "step": 10755 }, { "epoch": 0.54, "learning_rate": 9.285801625443336e-06, "loss": 0.2364, "step": 10756 }, { "epoch": 0.54, "learning_rate": 9.28418855432912e-06, "loss": 0.3435, "step": 10757 }, { "epoch": 0.54, "learning_rate": 9.282575501936015e-06, "loss": 0.5333, "step": 10758 }, { "epoch": 0.54, "learning_rate": 9.28096246830621e-06, "loss": 0.6594, "step": 10759 }, { "epoch": 0.54, "learning_rate": 9.279349453481885e-06, "loss": 0.2465, "step": 10760 }, { "epoch": 0.54, "learning_rate": 9.277736457505234e-06, "loss": 0.2547, "step": 10761 }, { "epoch": 0.54, "learning_rate": 9.276123480418442e-06, "loss": 0.4788, "step": 10762 }, { "epoch": 0.54, "learning_rate": 9.27451052226369e-06, "loss": 0.3925, "step": 10763 }, { "epoch": 0.54, "learning_rate": 9.272897583083162e-06, "loss": 0.2206, "step": 10764 }, { "epoch": 0.54, "learning_rate": 9.271284662919049e-06, "loss": 0.3606, "step": 10765 }, { "epoch": 0.54, "learning_rate": 9.26967176181353e-06, "loss": 0.3141, "step": 10766 }, { "epoch": 0.54, "learning_rate": 9.268058879808791e-06, "loss": 0.2887, "step": 10767 }, { "epoch": 0.54, "learning_rate": 9.266446016947014e-06, "loss": 0.2181, "step": 10768 }, { "epoch": 0.54, "learning_rate": 9.264833173270376e-06, "loss": 0.3237, "step": 10769 }, { "epoch": 0.54, "learning_rate": 9.263220348821066e-06, "loss": 0.2552, "step": 10770 }, { "epoch": 0.54, "learning_rate": 9.261607543641265e-06, "loss": 0.2162, "step": 10771 }, { "epoch": 0.54, "learning_rate": 9.25999475777315e-06, "loss": 0.2478, "step": 10772 }, { "epoch": 0.54, "learning_rate": 9.258381991258901e-06, "loss": 0.3492, "step": 10773 }, { "epoch": 0.54, "learning_rate": 9.256769244140702e-06, "loss": 0.4146, "step": 10774 }, { "epoch": 0.54, "learning_rate": 9.25515651646073e-06, "loss": 0.3942, "step": 10775 }, { "epoch": 0.54, "learning_rate": 9.253543808261166e-06, "loss": 0.091, "step": 10776 }, { "epoch": 0.54, "learning_rate": 9.251931119584183e-06, "loss": 0.4862, "step": 10777 }, { "epoch": 0.54, "learning_rate": 9.250318450471958e-06, "loss": 0.2779, "step": 10778 }, { "epoch": 0.54, "learning_rate": 9.248705800966676e-06, "loss": 0.3992, "step": 10779 }, { "epoch": 0.54, "learning_rate": 9.24709317111051e-06, "loss": 0.5341, "step": 10780 }, { "epoch": 0.54, "learning_rate": 9.245480560945638e-06, "loss": 0.3424, "step": 10781 }, { "epoch": 0.54, "learning_rate": 9.243867970514231e-06, "loss": 0.4132, "step": 10782 }, { "epoch": 0.54, "learning_rate": 9.242255399858467e-06, "loss": 0.3552, "step": 10783 }, { "epoch": 0.54, "learning_rate": 9.240642849020521e-06, "loss": 0.303, "step": 10784 }, { "epoch": 0.54, "learning_rate": 9.239030318042567e-06, "loss": 0.3713, "step": 10785 }, { "epoch": 0.54, "learning_rate": 9.237417806966778e-06, "loss": 0.2877, "step": 10786 }, { "epoch": 0.54, "learning_rate": 9.235805315835324e-06, "loss": 0.1723, "step": 10787 }, { "epoch": 0.54, "learning_rate": 9.234192844690385e-06, "loss": 0.4084, "step": 10788 }, { "epoch": 0.54, "learning_rate": 9.23258039357413e-06, "loss": 0.2598, "step": 10789 }, { "epoch": 0.54, "learning_rate": 9.230967962528729e-06, "loss": 0.3114, "step": 10790 }, { "epoch": 0.54, "learning_rate": 9.229355551596353e-06, "loss": 0.4336, "step": 10791 }, { "epoch": 0.54, "learning_rate": 9.227743160819169e-06, "loss": 0.1989, "step": 10792 }, { "epoch": 0.54, "learning_rate": 9.226130790239356e-06, "loss": 0.1662, "step": 10793 }, { "epoch": 0.54, "learning_rate": 9.224518439899079e-06, "loss": 0.4259, "step": 10794 }, { "epoch": 0.54, "learning_rate": 9.222906109840506e-06, "loss": 0.2992, "step": 10795 }, { "epoch": 0.54, "learning_rate": 9.221293800105804e-06, "loss": 0.2462, "step": 10796 }, { "epoch": 0.54, "learning_rate": 9.219681510737144e-06, "loss": 0.2606, "step": 10797 }, { "epoch": 0.54, "learning_rate": 9.218069241776692e-06, "loss": 0.3569, "step": 10798 }, { "epoch": 0.54, "learning_rate": 9.216456993266616e-06, "loss": 0.4798, "step": 10799 }, { "epoch": 0.54, "learning_rate": 9.214844765249081e-06, "loss": 0.2379, "step": 10800 }, { "epoch": 0.54, "learning_rate": 9.213232557766248e-06, "loss": 0.2787, "step": 10801 }, { "epoch": 0.54, "learning_rate": 9.21162037086029e-06, "loss": 0.4219, "step": 10802 }, { "epoch": 0.54, "learning_rate": 9.210008204573371e-06, "loss": 0.145, "step": 10803 }, { "epoch": 0.54, "learning_rate": 9.208396058947653e-06, "loss": 0.3616, "step": 10804 }, { "epoch": 0.54, "learning_rate": 9.206783934025292e-06, "loss": 0.4134, "step": 10805 }, { "epoch": 0.54, "learning_rate": 9.205171829848465e-06, "loss": 0.4051, "step": 10806 }, { "epoch": 0.54, "learning_rate": 9.203559746459327e-06, "loss": 0.4593, "step": 10807 }, { "epoch": 0.54, "learning_rate": 9.20194768390004e-06, "loss": 0.3423, "step": 10808 }, { "epoch": 0.54, "learning_rate": 9.200335642212767e-06, "loss": 0.3984, "step": 10809 }, { "epoch": 0.54, "learning_rate": 9.198723621439666e-06, "loss": 0.2672, "step": 10810 }, { "epoch": 0.54, "learning_rate": 9.197111621622901e-06, "loss": 0.3858, "step": 10811 }, { "epoch": 0.54, "learning_rate": 9.19549964280463e-06, "loss": 0.3432, "step": 10812 }, { "epoch": 0.54, "learning_rate": 9.193887685027011e-06, "loss": 0.4852, "step": 10813 }, { "epoch": 0.54, "learning_rate": 9.192275748332204e-06, "loss": 0.3289, "step": 10814 }, { "epoch": 0.54, "learning_rate": 9.190663832762364e-06, "loss": 0.4104, "step": 10815 }, { "epoch": 0.54, "learning_rate": 9.189051938359656e-06, "loss": 0.44, "step": 10816 }, { "epoch": 0.54, "learning_rate": 9.187440065166231e-06, "loss": 0.4185, "step": 10817 }, { "epoch": 0.54, "learning_rate": 9.185828213224247e-06, "loss": 0.3192, "step": 10818 }, { "epoch": 0.54, "learning_rate": 9.184216382575857e-06, "loss": 0.3255, "step": 10819 }, { "epoch": 0.54, "learning_rate": 9.182604573263223e-06, "loss": 0.4436, "step": 10820 }, { "epoch": 0.54, "learning_rate": 9.180992785328495e-06, "loss": 0.3376, "step": 10821 }, { "epoch": 0.54, "learning_rate": 9.179381018813827e-06, "loss": 0.4016, "step": 10822 }, { "epoch": 0.54, "learning_rate": 9.177769273761374e-06, "loss": 0.4586, "step": 10823 }, { "epoch": 0.54, "learning_rate": 9.176157550213284e-06, "loss": 0.3602, "step": 10824 }, { "epoch": 0.54, "learning_rate": 9.17454584821172e-06, "loss": 0.2569, "step": 10825 }, { "epoch": 0.54, "learning_rate": 9.172934167798829e-06, "loss": 0.3948, "step": 10826 }, { "epoch": 0.54, "learning_rate": 9.171322509016759e-06, "loss": 0.3837, "step": 10827 }, { "epoch": 0.54, "learning_rate": 9.169710871907659e-06, "loss": 0.3487, "step": 10828 }, { "epoch": 0.54, "learning_rate": 9.168099256513691e-06, "loss": 0.4041, "step": 10829 }, { "epoch": 0.54, "learning_rate": 9.166487662876996e-06, "loss": 0.3315, "step": 10830 }, { "epoch": 0.54, "learning_rate": 9.164876091039723e-06, "loss": 0.3916, "step": 10831 }, { "epoch": 0.54, "learning_rate": 9.163264541044022e-06, "loss": 0.3335, "step": 10832 }, { "epoch": 0.54, "learning_rate": 9.16165301293204e-06, "loss": 0.2283, "step": 10833 }, { "epoch": 0.54, "learning_rate": 9.160041506745929e-06, "loss": 0.3519, "step": 10834 }, { "epoch": 0.54, "learning_rate": 9.15843002252783e-06, "loss": 0.2634, "step": 10835 }, { "epoch": 0.54, "learning_rate": 9.156818560319892e-06, "loss": 0.2004, "step": 10836 }, { "epoch": 0.54, "learning_rate": 9.155207120164255e-06, "loss": 0.4706, "step": 10837 }, { "epoch": 0.54, "learning_rate": 9.153595702103075e-06, "loss": 0.3463, "step": 10838 }, { "epoch": 0.54, "learning_rate": 9.151984306178491e-06, "loss": 0.2944, "step": 10839 }, { "epoch": 0.54, "learning_rate": 9.150372932432647e-06, "loss": 0.1222, "step": 10840 }, { "epoch": 0.54, "learning_rate": 9.148761580907685e-06, "loss": 0.4371, "step": 10841 }, { "epoch": 0.54, "learning_rate": 9.147150251645744e-06, "loss": 0.1519, "step": 10842 }, { "epoch": 0.54, "learning_rate": 9.145538944688976e-06, "loss": 0.6969, "step": 10843 }, { "epoch": 0.54, "learning_rate": 9.14392766007952e-06, "loss": 0.4111, "step": 10844 }, { "epoch": 0.54, "learning_rate": 9.142316397859511e-06, "loss": 0.523, "step": 10845 }, { "epoch": 0.54, "learning_rate": 9.140705158071096e-06, "loss": 0.3388, "step": 10846 }, { "epoch": 0.54, "learning_rate": 9.139093940756409e-06, "loss": 0.3376, "step": 10847 }, { "epoch": 0.54, "learning_rate": 9.137482745957596e-06, "loss": 0.3408, "step": 10848 }, { "epoch": 0.54, "learning_rate": 9.13587157371679e-06, "loss": 0.4314, "step": 10849 }, { "epoch": 0.54, "learning_rate": 9.134260424076133e-06, "loss": 0.3268, "step": 10850 }, { "epoch": 0.54, "learning_rate": 9.132649297077755e-06, "loss": 0.3152, "step": 10851 }, { "epoch": 0.54, "learning_rate": 9.131038192763803e-06, "loss": 0.2527, "step": 10852 }, { "epoch": 0.54, "learning_rate": 9.129427111176408e-06, "loss": 0.5059, "step": 10853 }, { "epoch": 0.54, "learning_rate": 9.127816052357708e-06, "loss": 0.439, "step": 10854 }, { "epoch": 0.54, "learning_rate": 9.126205016349838e-06, "loss": 0.3928, "step": 10855 }, { "epoch": 0.54, "learning_rate": 9.124594003194928e-06, "loss": 0.3796, "step": 10856 }, { "epoch": 0.54, "learning_rate": 9.122983012935117e-06, "loss": 0.4253, "step": 10857 }, { "epoch": 0.54, "learning_rate": 9.121372045612537e-06, "loss": 0.3818, "step": 10858 }, { "epoch": 0.54, "learning_rate": 9.11976110126932e-06, "loss": 0.3899, "step": 10859 }, { "epoch": 0.54, "learning_rate": 9.118150179947594e-06, "loss": 0.306, "step": 10860 }, { "epoch": 0.54, "learning_rate": 9.1165392816895e-06, "loss": 0.412, "step": 10861 }, { "epoch": 0.54, "learning_rate": 9.114928406537164e-06, "loss": 0.2678, "step": 10862 }, { "epoch": 0.54, "learning_rate": 9.113317554532717e-06, "loss": 0.3652, "step": 10863 }, { "epoch": 0.54, "learning_rate": 9.111706725718287e-06, "loss": 0.2506, "step": 10864 }, { "epoch": 0.54, "learning_rate": 9.110095920136e-06, "loss": 0.3588, "step": 10865 }, { "epoch": 0.54, "learning_rate": 9.108485137827992e-06, "loss": 0.618, "step": 10866 }, { "epoch": 0.54, "learning_rate": 9.106874378836388e-06, "loss": 0.3886, "step": 10867 }, { "epoch": 0.54, "learning_rate": 9.105263643203316e-06, "loss": 0.3672, "step": 10868 }, { "epoch": 0.54, "learning_rate": 9.1036529309709e-06, "loss": 0.3595, "step": 10869 }, { "epoch": 0.54, "learning_rate": 9.102042242181267e-06, "loss": 0.4092, "step": 10870 }, { "epoch": 0.54, "learning_rate": 9.100431576876544e-06, "loss": 0.4699, "step": 10871 }, { "epoch": 0.54, "learning_rate": 9.098820935098856e-06, "loss": 0.4027, "step": 10872 }, { "epoch": 0.54, "learning_rate": 9.097210316890325e-06, "loss": 0.1491, "step": 10873 }, { "epoch": 0.54, "learning_rate": 9.095599722293072e-06, "loss": 0.3157, "step": 10874 }, { "epoch": 0.54, "learning_rate": 9.093989151349227e-06, "loss": 0.5156, "step": 10875 }, { "epoch": 0.54, "learning_rate": 9.09237860410091e-06, "loss": 0.3251, "step": 10876 }, { "epoch": 0.54, "learning_rate": 9.090768080590242e-06, "loss": 0.5855, "step": 10877 }, { "epoch": 0.54, "learning_rate": 9.089157580859339e-06, "loss": 0.386, "step": 10878 }, { "epoch": 0.54, "learning_rate": 9.087547104950332e-06, "loss": 0.1972, "step": 10879 }, { "epoch": 0.54, "learning_rate": 9.085936652905334e-06, "loss": 0.3295, "step": 10880 }, { "epoch": 0.54, "learning_rate": 9.084326224766466e-06, "loss": 0.3294, "step": 10881 }, { "epoch": 0.54, "learning_rate": 9.082715820575846e-06, "loss": 0.318, "step": 10882 }, { "epoch": 0.54, "learning_rate": 9.081105440375591e-06, "loss": 0.3355, "step": 10883 }, { "epoch": 0.54, "learning_rate": 9.07949508420782e-06, "loss": 0.3676, "step": 10884 }, { "epoch": 0.54, "learning_rate": 9.077884752114652e-06, "loss": 0.3723, "step": 10885 }, { "epoch": 0.54, "learning_rate": 9.0762744441382e-06, "loss": 0.4489, "step": 10886 }, { "epoch": 0.54, "learning_rate": 9.074664160320578e-06, "loss": 0.3695, "step": 10887 }, { "epoch": 0.54, "learning_rate": 9.0730539007039e-06, "loss": 0.5185, "step": 10888 }, { "epoch": 0.54, "learning_rate": 9.071443665330289e-06, "loss": 0.3317, "step": 10889 }, { "epoch": 0.54, "learning_rate": 9.069833454241851e-06, "loss": 0.3401, "step": 10890 }, { "epoch": 0.54, "learning_rate": 9.068223267480701e-06, "loss": 0.3863, "step": 10891 }, { "epoch": 0.54, "learning_rate": 9.066613105088948e-06, "loss": 0.2062, "step": 10892 }, { "epoch": 0.54, "learning_rate": 9.06500296710871e-06, "loss": 0.3664, "step": 10893 }, { "epoch": 0.54, "learning_rate": 9.063392853582094e-06, "loss": 0.3895, "step": 10894 }, { "epoch": 0.54, "learning_rate": 9.061782764551213e-06, "loss": 0.2369, "step": 10895 }, { "epoch": 0.54, "learning_rate": 9.060172700058172e-06, "loss": 0.1797, "step": 10896 }, { "epoch": 0.54, "learning_rate": 9.058562660145081e-06, "loss": 0.2812, "step": 10897 }, { "epoch": 0.54, "learning_rate": 9.056952644854055e-06, "loss": 0.2193, "step": 10898 }, { "epoch": 0.54, "learning_rate": 9.055342654227196e-06, "loss": 0.3401, "step": 10899 }, { "epoch": 0.54, "learning_rate": 9.053732688306613e-06, "loss": 0.4027, "step": 10900 }, { "epoch": 0.54, "learning_rate": 9.05212274713441e-06, "loss": 0.4965, "step": 10901 }, { "epoch": 0.54, "learning_rate": 9.050512830752697e-06, "loss": 0.3196, "step": 10902 }, { "epoch": 0.54, "learning_rate": 9.048902939203577e-06, "loss": 0.3856, "step": 10903 }, { "epoch": 0.54, "learning_rate": 9.047293072529156e-06, "loss": 0.3794, "step": 10904 }, { "epoch": 0.54, "learning_rate": 9.045683230771535e-06, "loss": 0.5582, "step": 10905 }, { "epoch": 0.54, "learning_rate": 9.04407341397282e-06, "loss": 0.3616, "step": 10906 }, { "epoch": 0.54, "learning_rate": 9.042463622175113e-06, "loss": 0.4371, "step": 10907 }, { "epoch": 0.54, "learning_rate": 9.040853855420516e-06, "loss": 0.2972, "step": 10908 }, { "epoch": 0.54, "learning_rate": 9.039244113751131e-06, "loss": 0.3253, "step": 10909 }, { "epoch": 0.54, "learning_rate": 9.037634397209053e-06, "loss": 0.3779, "step": 10910 }, { "epoch": 0.54, "learning_rate": 9.036024705836392e-06, "loss": 0.2629, "step": 10911 }, { "epoch": 0.54, "learning_rate": 9.03441503967524e-06, "loss": 0.2074, "step": 10912 }, { "epoch": 0.54, "learning_rate": 9.0328053987677e-06, "loss": 0.1904, "step": 10913 }, { "epoch": 0.54, "learning_rate": 9.031195783155867e-06, "loss": 0.487, "step": 10914 }, { "epoch": 0.55, "learning_rate": 9.029586192881836e-06, "loss": 0.2961, "step": 10915 }, { "epoch": 0.55, "learning_rate": 9.027976627987711e-06, "loss": 0.5411, "step": 10916 }, { "epoch": 0.55, "learning_rate": 9.026367088515584e-06, "loss": 0.3969, "step": 10917 }, { "epoch": 0.55, "learning_rate": 9.02475757450755e-06, "loss": 0.3612, "step": 10918 }, { "epoch": 0.55, "learning_rate": 9.023148086005706e-06, "loss": 0.2958, "step": 10919 }, { "epoch": 0.55, "learning_rate": 9.02153862305214e-06, "loss": 0.2549, "step": 10920 }, { "epoch": 0.55, "learning_rate": 9.019929185688954e-06, "loss": 0.341, "step": 10921 }, { "epoch": 0.55, "learning_rate": 9.018319773958236e-06, "loss": 0.2176, "step": 10922 }, { "epoch": 0.55, "learning_rate": 9.016710387902077e-06, "loss": 0.1587, "step": 10923 }, { "epoch": 0.55, "learning_rate": 9.015101027562568e-06, "loss": 0.3085, "step": 10924 }, { "epoch": 0.55, "learning_rate": 9.013491692981804e-06, "loss": 0.2541, "step": 10925 }, { "epoch": 0.55, "learning_rate": 9.011882384201873e-06, "loss": 0.311, "step": 10926 }, { "epoch": 0.55, "learning_rate": 9.010273101264864e-06, "loss": 0.5905, "step": 10927 }, { "epoch": 0.55, "learning_rate": 9.008663844212866e-06, "loss": 0.2855, "step": 10928 }, { "epoch": 0.55, "learning_rate": 9.007054613087964e-06, "loss": 0.5371, "step": 10929 }, { "epoch": 0.55, "learning_rate": 9.005445407932252e-06, "loss": 0.167, "step": 10930 }, { "epoch": 0.55, "learning_rate": 9.00383622878781e-06, "loss": 0.3982, "step": 10931 }, { "epoch": 0.55, "learning_rate": 9.00222707569673e-06, "loss": 0.2972, "step": 10932 }, { "epoch": 0.55, "learning_rate": 9.000617948701088e-06, "loss": 0.237, "step": 10933 }, { "epoch": 0.55, "learning_rate": 8.999008847842978e-06, "loss": 0.4011, "step": 10934 }, { "epoch": 0.55, "learning_rate": 8.997399773164483e-06, "loss": 0.6933, "step": 10935 }, { "epoch": 0.55, "learning_rate": 8.995790724707682e-06, "loss": 0.373, "step": 10936 }, { "epoch": 0.55, "learning_rate": 8.994181702514658e-06, "loss": 0.4061, "step": 10937 }, { "epoch": 0.55, "learning_rate": 8.992572706627491e-06, "loss": 0.4185, "step": 10938 }, { "epoch": 0.55, "learning_rate": 8.99096373708827e-06, "loss": 0.4214, "step": 10939 }, { "epoch": 0.55, "learning_rate": 8.989354793939072e-06, "loss": 0.2792, "step": 10940 }, { "epoch": 0.55, "learning_rate": 8.987745877221976e-06, "loss": 0.4249, "step": 10941 }, { "epoch": 0.55, "learning_rate": 8.986136986979055e-06, "loss": 0.2832, "step": 10942 }, { "epoch": 0.55, "learning_rate": 8.984528123252398e-06, "loss": 0.3565, "step": 10943 }, { "epoch": 0.55, "learning_rate": 8.982919286084077e-06, "loss": 0.4268, "step": 10944 }, { "epoch": 0.55, "learning_rate": 8.98131047551617e-06, "loss": 0.1816, "step": 10945 }, { "epoch": 0.55, "learning_rate": 8.979701691590754e-06, "loss": 0.3796, "step": 10946 }, { "epoch": 0.55, "learning_rate": 8.9780929343499e-06, "loss": 0.2679, "step": 10947 }, { "epoch": 0.55, "learning_rate": 8.976484203835691e-06, "loss": 0.6122, "step": 10948 }, { "epoch": 0.55, "learning_rate": 8.974875500090196e-06, "loss": 0.3878, "step": 10949 }, { "epoch": 0.55, "learning_rate": 8.97326682315549e-06, "loss": 0.4498, "step": 10950 }, { "epoch": 0.55, "learning_rate": 8.971658173073645e-06, "loss": 0.2745, "step": 10951 }, { "epoch": 0.55, "learning_rate": 8.97004954988673e-06, "loss": 0.4537, "step": 10952 }, { "epoch": 0.55, "learning_rate": 8.968440953636824e-06, "loss": 0.1011, "step": 10953 }, { "epoch": 0.55, "learning_rate": 8.966832384365994e-06, "loss": 0.3962, "step": 10954 }, { "epoch": 0.55, "learning_rate": 8.96522384211631e-06, "loss": 0.3144, "step": 10955 }, { "epoch": 0.55, "learning_rate": 8.963615326929838e-06, "loss": 0.2633, "step": 10956 }, { "epoch": 0.55, "learning_rate": 8.962006838848654e-06, "loss": 0.5292, "step": 10957 }, { "epoch": 0.55, "learning_rate": 8.96039837791482e-06, "loss": 0.4136, "step": 10958 }, { "epoch": 0.55, "learning_rate": 8.958789944170404e-06, "loss": 0.3562, "step": 10959 }, { "epoch": 0.55, "learning_rate": 8.957181537657474e-06, "loss": 0.352, "step": 10960 }, { "epoch": 0.55, "learning_rate": 8.955573158418091e-06, "loss": 0.483, "step": 10961 }, { "epoch": 0.55, "learning_rate": 8.953964806494329e-06, "loss": 0.4169, "step": 10962 }, { "epoch": 0.55, "learning_rate": 8.952356481928246e-06, "loss": 0.3657, "step": 10963 }, { "epoch": 0.55, "learning_rate": 8.950748184761908e-06, "loss": 0.2774, "step": 10964 }, { "epoch": 0.55, "learning_rate": 8.949139915037374e-06, "loss": 0.3057, "step": 10965 }, { "epoch": 0.55, "learning_rate": 8.947531672796712e-06, "loss": 0.5649, "step": 10966 }, { "epoch": 0.55, "learning_rate": 8.94592345808198e-06, "loss": 0.5457, "step": 10967 }, { "epoch": 0.55, "learning_rate": 8.944315270935241e-06, "loss": 0.1951, "step": 10968 }, { "epoch": 0.55, "learning_rate": 8.94270711139855e-06, "loss": 0.3866, "step": 10969 }, { "epoch": 0.55, "learning_rate": 8.94109897951397e-06, "loss": 0.2396, "step": 10970 }, { "epoch": 0.55, "learning_rate": 8.93949087532356e-06, "loss": 0.2881, "step": 10971 }, { "epoch": 0.55, "learning_rate": 8.93788279886938e-06, "loss": 0.425, "step": 10972 }, { "epoch": 0.55, "learning_rate": 8.936274750193481e-06, "loss": 0.3664, "step": 10973 }, { "epoch": 0.55, "learning_rate": 8.93466672933792e-06, "loss": 0.3664, "step": 10974 }, { "epoch": 0.55, "learning_rate": 8.933058736344761e-06, "loss": 0.1755, "step": 10975 }, { "epoch": 0.55, "learning_rate": 8.931450771256052e-06, "loss": 0.5102, "step": 10976 }, { "epoch": 0.55, "learning_rate": 8.929842834113847e-06, "loss": 0.3255, "step": 10977 }, { "epoch": 0.55, "learning_rate": 8.928234924960204e-06, "loss": 0.4002, "step": 10978 }, { "epoch": 0.55, "learning_rate": 8.926627043837168e-06, "loss": 0.5039, "step": 10979 }, { "epoch": 0.55, "learning_rate": 8.9250191907868e-06, "loss": 0.4465, "step": 10980 }, { "epoch": 0.55, "learning_rate": 8.923411365851145e-06, "loss": 0.4189, "step": 10981 }, { "epoch": 0.55, "learning_rate": 8.921803569072258e-06, "loss": 0.2799, "step": 10982 }, { "epoch": 0.55, "learning_rate": 8.920195800492181e-06, "loss": 0.3192, "step": 10983 }, { "epoch": 0.55, "learning_rate": 8.918588060152973e-06, "loss": 0.2963, "step": 10984 }, { "epoch": 0.55, "learning_rate": 8.916980348096678e-06, "loss": 0.4015, "step": 10985 }, { "epoch": 0.55, "learning_rate": 8.915372664365343e-06, "loss": 0.2914, "step": 10986 }, { "epoch": 0.55, "learning_rate": 8.913765009001014e-06, "loss": 0.2953, "step": 10987 }, { "epoch": 0.55, "learning_rate": 8.912157382045737e-06, "loss": 0.531, "step": 10988 }, { "epoch": 0.55, "learning_rate": 8.910549783541563e-06, "loss": 0.2169, "step": 10989 }, { "epoch": 0.55, "learning_rate": 8.90894221353053e-06, "loss": 0.3289, "step": 10990 }, { "epoch": 0.55, "learning_rate": 8.907334672054687e-06, "loss": 0.2376, "step": 10991 }, { "epoch": 0.55, "learning_rate": 8.905727159156072e-06, "loss": 0.3603, "step": 10992 }, { "epoch": 0.55, "learning_rate": 8.90411967487673e-06, "loss": 0.271, "step": 10993 }, { "epoch": 0.55, "learning_rate": 8.902512219258703e-06, "loss": 0.3001, "step": 10994 }, { "epoch": 0.55, "learning_rate": 8.900904792344034e-06, "loss": 0.4158, "step": 10995 }, { "epoch": 0.55, "learning_rate": 8.899297394174757e-06, "loss": 0.3309, "step": 10996 }, { "epoch": 0.55, "learning_rate": 8.897690024792912e-06, "loss": 0.5528, "step": 10997 }, { "epoch": 0.55, "learning_rate": 8.896082684240547e-06, "loss": 0.2527, "step": 10998 }, { "epoch": 0.55, "learning_rate": 8.89447537255969e-06, "loss": 0.4739, "step": 10999 }, { "epoch": 0.55, "learning_rate": 8.892868089792384e-06, "loss": 0.2564, "step": 11000 }, { "epoch": 0.55, "learning_rate": 8.891260835980664e-06, "loss": 0.3671, "step": 11001 }, { "epoch": 0.55, "learning_rate": 8.88965361116656e-06, "loss": 0.4409, "step": 11002 }, { "epoch": 0.55, "learning_rate": 8.888046415392116e-06, "loss": 0.2617, "step": 11003 }, { "epoch": 0.55, "learning_rate": 8.88643924869936e-06, "loss": 0.2922, "step": 11004 }, { "epoch": 0.55, "learning_rate": 8.884832111130326e-06, "loss": 0.3065, "step": 11005 }, { "epoch": 0.55, "learning_rate": 8.883225002727046e-06, "loss": 0.4612, "step": 11006 }, { "epoch": 0.55, "learning_rate": 8.881617923531555e-06, "loss": 0.3507, "step": 11007 }, { "epoch": 0.55, "learning_rate": 8.880010873585886e-06, "loss": 0.5052, "step": 11008 }, { "epoch": 0.55, "learning_rate": 8.878403852932062e-06, "loss": 0.2954, "step": 11009 }, { "epoch": 0.55, "learning_rate": 8.876796861612117e-06, "loss": 0.2447, "step": 11010 }, { "epoch": 0.55, "learning_rate": 8.875189899668077e-06, "loss": 0.3859, "step": 11011 }, { "epoch": 0.55, "learning_rate": 8.873582967141974e-06, "loss": 0.2355, "step": 11012 }, { "epoch": 0.55, "learning_rate": 8.871976064075834e-06, "loss": 0.3975, "step": 11013 }, { "epoch": 0.55, "learning_rate": 8.87036919051168e-06, "loss": 0.2245, "step": 11014 }, { "epoch": 0.55, "learning_rate": 8.868762346491541e-06, "loss": 0.2687, "step": 11015 }, { "epoch": 0.55, "learning_rate": 8.867155532057443e-06, "loss": 0.3427, "step": 11016 }, { "epoch": 0.55, "learning_rate": 8.865548747251409e-06, "loss": 0.268, "step": 11017 }, { "epoch": 0.55, "learning_rate": 8.86394199211546e-06, "loss": 0.453, "step": 11018 }, { "epoch": 0.55, "learning_rate": 8.86233526669162e-06, "loss": 0.3188, "step": 11019 }, { "epoch": 0.55, "learning_rate": 8.860728571021908e-06, "loss": 0.3417, "step": 11020 }, { "epoch": 0.55, "learning_rate": 8.85912190514835e-06, "loss": 0.4313, "step": 11021 }, { "epoch": 0.55, "learning_rate": 8.857515269112968e-06, "loss": 0.4069, "step": 11022 }, { "epoch": 0.55, "learning_rate": 8.855908662957776e-06, "loss": 0.3399, "step": 11023 }, { "epoch": 0.55, "learning_rate": 8.854302086724794e-06, "loss": 0.3876, "step": 11024 }, { "epoch": 0.55, "learning_rate": 8.852695540456037e-06, "loss": 0.3416, "step": 11025 }, { "epoch": 0.55, "learning_rate": 8.851089024193528e-06, "loss": 0.4917, "step": 11026 }, { "epoch": 0.55, "learning_rate": 8.849482537979283e-06, "loss": 0.3927, "step": 11027 }, { "epoch": 0.55, "learning_rate": 8.847876081855314e-06, "loss": 0.5386, "step": 11028 }, { "epoch": 0.55, "learning_rate": 8.846269655863636e-06, "loss": 0.2488, "step": 11029 }, { "epoch": 0.55, "learning_rate": 8.844663260046264e-06, "loss": 0.2339, "step": 11030 }, { "epoch": 0.55, "learning_rate": 8.843056894445213e-06, "loss": 0.2895, "step": 11031 }, { "epoch": 0.55, "learning_rate": 8.841450559102493e-06, "loss": 0.3173, "step": 11032 }, { "epoch": 0.55, "learning_rate": 8.839844254060116e-06, "loss": 0.3567, "step": 11033 }, { "epoch": 0.55, "learning_rate": 8.838237979360089e-06, "loss": 0.5138, "step": 11034 }, { "epoch": 0.55, "learning_rate": 8.83663173504443e-06, "loss": 0.4292, "step": 11035 }, { "epoch": 0.55, "learning_rate": 8.835025521155145e-06, "loss": 0.2787, "step": 11036 }, { "epoch": 0.55, "learning_rate": 8.833419337734241e-06, "loss": 0.2961, "step": 11037 }, { "epoch": 0.55, "learning_rate": 8.831813184823724e-06, "loss": 0.326, "step": 11038 }, { "epoch": 0.55, "learning_rate": 8.830207062465605e-06, "loss": 0.3355, "step": 11039 }, { "epoch": 0.55, "learning_rate": 8.82860097070189e-06, "loss": 0.3508, "step": 11040 }, { "epoch": 0.55, "learning_rate": 8.82699490957458e-06, "loss": 0.3669, "step": 11041 }, { "epoch": 0.55, "learning_rate": 8.825388879125681e-06, "loss": 0.3669, "step": 11042 }, { "epoch": 0.55, "learning_rate": 8.823782879397196e-06, "loss": 0.5304, "step": 11043 }, { "epoch": 0.55, "learning_rate": 8.822176910431133e-06, "loss": 0.4655, "step": 11044 }, { "epoch": 0.55, "learning_rate": 8.82057097226949e-06, "loss": 0.2223, "step": 11045 }, { "epoch": 0.55, "learning_rate": 8.818965064954268e-06, "loss": 0.376, "step": 11046 }, { "epoch": 0.55, "learning_rate": 8.817359188527465e-06, "loss": 0.3589, "step": 11047 }, { "epoch": 0.55, "learning_rate": 8.815753343031085e-06, "loss": 0.255, "step": 11048 }, { "epoch": 0.55, "learning_rate": 8.814147528507128e-06, "loss": 0.4766, "step": 11049 }, { "epoch": 0.55, "learning_rate": 8.812541744997588e-06, "loss": 0.1935, "step": 11050 }, { "epoch": 0.55, "learning_rate": 8.810935992544463e-06, "loss": 0.2337, "step": 11051 }, { "epoch": 0.55, "learning_rate": 8.809330271189747e-06, "loss": 0.2179, "step": 11052 }, { "epoch": 0.55, "learning_rate": 8.807724580975442e-06, "loss": 0.4403, "step": 11053 }, { "epoch": 0.55, "learning_rate": 8.806118921943538e-06, "loss": 0.4032, "step": 11054 }, { "epoch": 0.55, "learning_rate": 8.80451329413603e-06, "loss": 0.5196, "step": 11055 }, { "epoch": 0.55, "learning_rate": 8.802907697594911e-06, "loss": 0.5192, "step": 11056 }, { "epoch": 0.55, "learning_rate": 8.80130213236217e-06, "loss": 0.2847, "step": 11057 }, { "epoch": 0.55, "learning_rate": 8.799696598479805e-06, "loss": 0.2945, "step": 11058 }, { "epoch": 0.55, "learning_rate": 8.798091095989802e-06, "loss": 0.375, "step": 11059 }, { "epoch": 0.55, "learning_rate": 8.796485624934154e-06, "loss": 0.328, "step": 11060 }, { "epoch": 0.55, "learning_rate": 8.794880185354844e-06, "loss": 0.4366, "step": 11061 }, { "epoch": 0.55, "learning_rate": 8.793274777293868e-06, "loss": 0.4512, "step": 11062 }, { "epoch": 0.55, "learning_rate": 8.791669400793208e-06, "loss": 0.1733, "step": 11063 }, { "epoch": 0.55, "learning_rate": 8.790064055894856e-06, "loss": 0.3734, "step": 11064 }, { "epoch": 0.55, "learning_rate": 8.788458742640791e-06, "loss": 0.2882, "step": 11065 }, { "epoch": 0.55, "learning_rate": 8.786853461073e-06, "loss": 0.3817, "step": 11066 }, { "epoch": 0.55, "learning_rate": 8.78524821123347e-06, "loss": 0.4364, "step": 11067 }, { "epoch": 0.55, "learning_rate": 8.78364299316418e-06, "loss": 0.1691, "step": 11068 }, { "epoch": 0.55, "learning_rate": 8.782037806907116e-06, "loss": 0.2809, "step": 11069 }, { "epoch": 0.55, "learning_rate": 8.780432652504254e-06, "loss": 0.3193, "step": 11070 }, { "epoch": 0.55, "learning_rate": 8.778827529997582e-06, "loss": 0.3835, "step": 11071 }, { "epoch": 0.55, "learning_rate": 8.777222439429077e-06, "loss": 0.1808, "step": 11072 }, { "epoch": 0.55, "learning_rate": 8.775617380840717e-06, "loss": 0.3095, "step": 11073 }, { "epoch": 0.55, "learning_rate": 8.77401235427448e-06, "loss": 0.312, "step": 11074 }, { "epoch": 0.55, "learning_rate": 8.772407359772343e-06, "loss": 0.3583, "step": 11075 }, { "epoch": 0.55, "learning_rate": 8.770802397376286e-06, "loss": 0.2318, "step": 11076 }, { "epoch": 0.55, "learning_rate": 8.769197467128282e-06, "loss": 0.1604, "step": 11077 }, { "epoch": 0.55, "learning_rate": 8.767592569070304e-06, "loss": 0.3308, "step": 11078 }, { "epoch": 0.55, "learning_rate": 8.76598770324433e-06, "loss": 0.4148, "step": 11079 }, { "epoch": 0.55, "learning_rate": 8.764382869692332e-06, "loss": 0.6008, "step": 11080 }, { "epoch": 0.55, "learning_rate": 8.762778068456281e-06, "loss": 0.3319, "step": 11081 }, { "epoch": 0.55, "learning_rate": 8.76117329957815e-06, "loss": 0.1638, "step": 11082 }, { "epoch": 0.55, "learning_rate": 8.759568563099908e-06, "loss": 0.2756, "step": 11083 }, { "epoch": 0.55, "learning_rate": 8.757963859063523e-06, "loss": 0.2257, "step": 11084 }, { "epoch": 0.55, "learning_rate": 8.75635918751097e-06, "loss": 0.4127, "step": 11085 }, { "epoch": 0.55, "learning_rate": 8.754754548484212e-06, "loss": 0.241, "step": 11086 }, { "epoch": 0.55, "learning_rate": 8.753149942025219e-06, "loss": 0.2934, "step": 11087 }, { "epoch": 0.55, "learning_rate": 8.751545368175956e-06, "loss": 0.2062, "step": 11088 }, { "epoch": 0.55, "learning_rate": 8.749940826978387e-06, "loss": 0.289, "step": 11089 }, { "epoch": 0.55, "learning_rate": 8.74833631847448e-06, "loss": 0.6398, "step": 11090 }, { "epoch": 0.55, "learning_rate": 8.746731842706198e-06, "loss": 0.375, "step": 11091 }, { "epoch": 0.55, "learning_rate": 8.745127399715501e-06, "loss": 0.4412, "step": 11092 }, { "epoch": 0.55, "learning_rate": 8.74352298954435e-06, "loss": 0.4223, "step": 11093 }, { "epoch": 0.55, "learning_rate": 8.741918612234716e-06, "loss": 0.198, "step": 11094 }, { "epoch": 0.55, "learning_rate": 8.740314267828552e-06, "loss": 0.4058, "step": 11095 }, { "epoch": 0.55, "learning_rate": 8.738709956367817e-06, "loss": 0.5505, "step": 11096 }, { "epoch": 0.55, "learning_rate": 8.737105677894473e-06, "loss": 0.2882, "step": 11097 }, { "epoch": 0.55, "learning_rate": 8.735501432450468e-06, "loss": 0.2771, "step": 11098 }, { "epoch": 0.55, "learning_rate": 8.733897220077774e-06, "loss": 0.4379, "step": 11099 }, { "epoch": 0.55, "learning_rate": 8.73229304081834e-06, "loss": 0.3113, "step": 11100 }, { "epoch": 0.55, "learning_rate": 8.73068889471412e-06, "loss": 0.3262, "step": 11101 }, { "epoch": 0.55, "learning_rate": 8.729084781807069e-06, "loss": 0.602, "step": 11102 }, { "epoch": 0.55, "learning_rate": 8.727480702139142e-06, "loss": 0.309, "step": 11103 }, { "epoch": 0.55, "learning_rate": 8.72587665575229e-06, "loss": 0.2993, "step": 11104 }, { "epoch": 0.55, "learning_rate": 8.724272642688467e-06, "loss": 0.2534, "step": 11105 }, { "epoch": 0.55, "learning_rate": 8.72266866298962e-06, "loss": 0.3024, "step": 11106 }, { "epoch": 0.55, "learning_rate": 8.721064716697697e-06, "loss": 0.4512, "step": 11107 }, { "epoch": 0.55, "learning_rate": 8.719460803854657e-06, "loss": 0.4072, "step": 11108 }, { "epoch": 0.55, "learning_rate": 8.717856924502441e-06, "loss": 0.23, "step": 11109 }, { "epoch": 0.55, "learning_rate": 8.716253078683e-06, "loss": 0.3342, "step": 11110 }, { "epoch": 0.55, "learning_rate": 8.714649266438274e-06, "loss": 0.4081, "step": 11111 }, { "epoch": 0.55, "learning_rate": 8.713045487810216e-06, "loss": 0.3969, "step": 11112 }, { "epoch": 0.55, "learning_rate": 8.711441742840764e-06, "loss": 0.3021, "step": 11113 }, { "epoch": 0.55, "learning_rate": 8.70983803157187e-06, "loss": 0.2348, "step": 11114 }, { "epoch": 0.56, "learning_rate": 8.708234354045471e-06, "loss": 0.3527, "step": 11115 }, { "epoch": 0.56, "learning_rate": 8.70663071030351e-06, "loss": 0.4603, "step": 11116 }, { "epoch": 0.56, "learning_rate": 8.70502710038793e-06, "loss": 0.6067, "step": 11117 }, { "epoch": 0.56, "learning_rate": 8.703423524340669e-06, "loss": 0.2777, "step": 11118 }, { "epoch": 0.56, "learning_rate": 8.701819982203668e-06, "loss": 0.4044, "step": 11119 }, { "epoch": 0.56, "learning_rate": 8.70021647401886e-06, "loss": 0.4287, "step": 11120 }, { "epoch": 0.56, "learning_rate": 8.698612999828194e-06, "loss": 0.3529, "step": 11121 }, { "epoch": 0.56, "learning_rate": 8.6970095596736e-06, "loss": 0.3394, "step": 11122 }, { "epoch": 0.56, "learning_rate": 8.695406153597015e-06, "loss": 0.3218, "step": 11123 }, { "epoch": 0.56, "learning_rate": 8.69380278164037e-06, "loss": 0.3681, "step": 11124 }, { "epoch": 0.56, "learning_rate": 8.692199443845603e-06, "loss": 0.2812, "step": 11125 }, { "epoch": 0.56, "learning_rate": 8.69059614025465e-06, "loss": 0.3835, "step": 11126 }, { "epoch": 0.56, "learning_rate": 8.688992870909438e-06, "loss": 0.4275, "step": 11127 }, { "epoch": 0.56, "learning_rate": 8.6873896358519e-06, "loss": 0.4947, "step": 11128 }, { "epoch": 0.56, "learning_rate": 8.685786435123968e-06, "loss": 0.4066, "step": 11129 }, { "epoch": 0.56, "learning_rate": 8.684183268767565e-06, "loss": 0.3051, "step": 11130 }, { "epoch": 0.56, "learning_rate": 8.682580136824631e-06, "loss": 0.3073, "step": 11131 }, { "epoch": 0.56, "learning_rate": 8.680977039337087e-06, "loss": 0.5835, "step": 11132 }, { "epoch": 0.56, "learning_rate": 8.679373976346862e-06, "loss": 0.3668, "step": 11133 }, { "epoch": 0.56, "learning_rate": 8.677770947895875e-06, "loss": 0.4161, "step": 11134 }, { "epoch": 0.56, "learning_rate": 8.676167954026063e-06, "loss": 0.2727, "step": 11135 }, { "epoch": 0.56, "learning_rate": 8.674564994779344e-06, "loss": 0.7502, "step": 11136 }, { "epoch": 0.56, "learning_rate": 8.67296207019764e-06, "loss": 0.2756, "step": 11137 }, { "epoch": 0.56, "learning_rate": 8.671359180322877e-06, "loss": 0.4742, "step": 11138 }, { "epoch": 0.56, "learning_rate": 8.66975632519697e-06, "loss": 0.5143, "step": 11139 }, { "epoch": 0.56, "learning_rate": 8.668153504861848e-06, "loss": 0.2728, "step": 11140 }, { "epoch": 0.56, "learning_rate": 8.666550719359426e-06, "loss": 0.5506, "step": 11141 }, { "epoch": 0.56, "learning_rate": 8.664947968731623e-06, "loss": 0.134, "step": 11142 }, { "epoch": 0.56, "learning_rate": 8.663345253020353e-06, "loss": 0.3022, "step": 11143 }, { "epoch": 0.56, "learning_rate": 8.661742572267542e-06, "loss": 0.2969, "step": 11144 }, { "epoch": 0.56, "learning_rate": 8.6601399265151e-06, "loss": 0.4455, "step": 11145 }, { "epoch": 0.56, "learning_rate": 8.658537315804947e-06, "loss": 0.3766, "step": 11146 }, { "epoch": 0.56, "learning_rate": 8.65693474017899e-06, "loss": 0.1627, "step": 11147 }, { "epoch": 0.56, "learning_rate": 8.655332199679145e-06, "loss": 0.3252, "step": 11148 }, { "epoch": 0.56, "learning_rate": 8.653729694347324e-06, "loss": 0.3971, "step": 11149 }, { "epoch": 0.56, "learning_rate": 8.652127224225442e-06, "loss": 0.5191, "step": 11150 }, { "epoch": 0.56, "learning_rate": 8.650524789355408e-06, "loss": 0.2449, "step": 11151 }, { "epoch": 0.56, "learning_rate": 8.648922389779129e-06, "loss": 0.3399, "step": 11152 }, { "epoch": 0.56, "learning_rate": 8.647320025538515e-06, "loss": 0.4066, "step": 11153 }, { "epoch": 0.56, "learning_rate": 8.645717696675476e-06, "loss": 0.25, "step": 11154 }, { "epoch": 0.56, "learning_rate": 8.644115403231917e-06, "loss": 0.4077, "step": 11155 }, { "epoch": 0.56, "learning_rate": 8.642513145249743e-06, "loss": 0.6358, "step": 11156 }, { "epoch": 0.56, "learning_rate": 8.640910922770854e-06, "loss": 0.4928, "step": 11157 }, { "epoch": 0.56, "learning_rate": 8.639308735837166e-06, "loss": 0.4059, "step": 11158 }, { "epoch": 0.56, "learning_rate": 8.637706584490576e-06, "loss": 0.4711, "step": 11159 }, { "epoch": 0.56, "learning_rate": 8.636104468772984e-06, "loss": 0.2836, "step": 11160 }, { "epoch": 0.56, "learning_rate": 8.634502388726295e-06, "loss": 0.5792, "step": 11161 }, { "epoch": 0.56, "learning_rate": 8.632900344392405e-06, "loss": 0.2867, "step": 11162 }, { "epoch": 0.56, "learning_rate": 8.631298335813218e-06, "loss": 0.4204, "step": 11163 }, { "epoch": 0.56, "learning_rate": 8.62969636303063e-06, "loss": 0.3167, "step": 11164 }, { "epoch": 0.56, "learning_rate": 8.62809442608654e-06, "loss": 0.4125, "step": 11165 }, { "epoch": 0.56, "learning_rate": 8.626492525022838e-06, "loss": 0.3409, "step": 11166 }, { "epoch": 0.56, "learning_rate": 8.62489065988143e-06, "loss": 0.3645, "step": 11167 }, { "epoch": 0.56, "learning_rate": 8.623288830704205e-06, "loss": 0.4024, "step": 11168 }, { "epoch": 0.56, "learning_rate": 8.621687037533057e-06, "loss": 0.3414, "step": 11169 }, { "epoch": 0.56, "learning_rate": 8.62008528040988e-06, "loss": 0.2707, "step": 11170 }, { "epoch": 0.56, "learning_rate": 8.618483559376561e-06, "loss": 0.5206, "step": 11171 }, { "epoch": 0.56, "learning_rate": 8.616881874475e-06, "loss": 0.3484, "step": 11172 }, { "epoch": 0.56, "learning_rate": 8.615280225747081e-06, "loss": 0.485, "step": 11173 }, { "epoch": 0.56, "learning_rate": 8.613678613234695e-06, "loss": 0.4492, "step": 11174 }, { "epoch": 0.56, "learning_rate": 8.612077036979726e-06, "loss": 0.3039, "step": 11175 }, { "epoch": 0.56, "learning_rate": 8.610475497024066e-06, "loss": 0.3632, "step": 11176 }, { "epoch": 0.56, "learning_rate": 8.608873993409601e-06, "loss": 0.3866, "step": 11177 }, { "epoch": 0.56, "learning_rate": 8.607272526178213e-06, "loss": 0.1315, "step": 11178 }, { "epoch": 0.56, "learning_rate": 8.605671095371788e-06, "loss": 0.392, "step": 11179 }, { "epoch": 0.56, "learning_rate": 8.604069701032207e-06, "loss": 0.2547, "step": 11180 }, { "epoch": 0.56, "learning_rate": 8.602468343201359e-06, "loss": 0.2826, "step": 11181 }, { "epoch": 0.56, "learning_rate": 8.600867021921118e-06, "loss": 0.2508, "step": 11182 }, { "epoch": 0.56, "learning_rate": 8.599265737233369e-06, "loss": 0.3682, "step": 11183 }, { "epoch": 0.56, "learning_rate": 8.597664489179988e-06, "loss": 0.4933, "step": 11184 }, { "epoch": 0.56, "learning_rate": 8.596063277802855e-06, "loss": 0.3679, "step": 11185 }, { "epoch": 0.56, "learning_rate": 8.59446210314385e-06, "loss": 0.2198, "step": 11186 }, { "epoch": 0.56, "learning_rate": 8.592860965244851e-06, "loss": 0.3179, "step": 11187 }, { "epoch": 0.56, "learning_rate": 8.591259864147727e-06, "loss": 0.2207, "step": 11188 }, { "epoch": 0.56, "learning_rate": 8.589658799894354e-06, "loss": 0.4749, "step": 11189 }, { "epoch": 0.56, "learning_rate": 8.588057772526612e-06, "loss": 0.5229, "step": 11190 }, { "epoch": 0.56, "learning_rate": 8.586456782086369e-06, "loss": 0.3072, "step": 11191 }, { "epoch": 0.56, "learning_rate": 8.584855828615497e-06, "loss": 0.3386, "step": 11192 }, { "epoch": 0.56, "learning_rate": 8.583254912155868e-06, "loss": 0.4043, "step": 11193 }, { "epoch": 0.56, "learning_rate": 8.581654032749345e-06, "loss": 0.43, "step": 11194 }, { "epoch": 0.56, "learning_rate": 8.580053190437809e-06, "loss": 0.3059, "step": 11195 }, { "epoch": 0.56, "learning_rate": 8.578452385263122e-06, "loss": 0.2297, "step": 11196 }, { "epoch": 0.56, "learning_rate": 8.576851617267151e-06, "loss": 0.4038, "step": 11197 }, { "epoch": 0.56, "learning_rate": 8.57525088649176e-06, "loss": 0.2, "step": 11198 }, { "epoch": 0.56, "learning_rate": 8.573650192978817e-06, "loss": 0.448, "step": 11199 }, { "epoch": 0.56, "learning_rate": 8.572049536770185e-06, "loss": 0.2717, "step": 11200 }, { "epoch": 0.56, "learning_rate": 8.570448917907728e-06, "loss": 0.5058, "step": 11201 }, { "epoch": 0.56, "learning_rate": 8.568848336433307e-06, "loss": 0.3145, "step": 11202 }, { "epoch": 0.56, "learning_rate": 8.567247792388778e-06, "loss": 0.3078, "step": 11203 }, { "epoch": 0.56, "learning_rate": 8.565647285816013e-06, "loss": 0.5187, "step": 11204 }, { "epoch": 0.56, "learning_rate": 8.564046816756863e-06, "loss": 0.2679, "step": 11205 }, { "epoch": 0.56, "learning_rate": 8.562446385253189e-06, "loss": 0.2682, "step": 11206 }, { "epoch": 0.56, "learning_rate": 8.56084599134684e-06, "loss": 0.3718, "step": 11207 }, { "epoch": 0.56, "learning_rate": 8.559245635079685e-06, "loss": 0.3636, "step": 11208 }, { "epoch": 0.56, "learning_rate": 8.557645316493573e-06, "loss": 0.4012, "step": 11209 }, { "epoch": 0.56, "learning_rate": 8.556045035630358e-06, "loss": 0.3884, "step": 11210 }, { "epoch": 0.56, "learning_rate": 8.554444792531896e-06, "loss": 0.0448, "step": 11211 }, { "epoch": 0.56, "learning_rate": 8.552844587240035e-06, "loss": 0.4347, "step": 11212 }, { "epoch": 0.56, "learning_rate": 8.551244419796628e-06, "loss": 0.2957, "step": 11213 }, { "epoch": 0.56, "learning_rate": 8.549644290243528e-06, "loss": 0.0277, "step": 11214 }, { "epoch": 0.56, "learning_rate": 8.54804419862258e-06, "loss": 0.4551, "step": 11215 }, { "epoch": 0.56, "learning_rate": 8.546444144975632e-06, "loss": 0.2692, "step": 11216 }, { "epoch": 0.56, "learning_rate": 8.544844129344537e-06, "loss": 0.3668, "step": 11217 }, { "epoch": 0.56, "learning_rate": 8.543244151771137e-06, "loss": 0.4664, "step": 11218 }, { "epoch": 0.56, "learning_rate": 8.541644212297278e-06, "loss": 0.2622, "step": 11219 }, { "epoch": 0.56, "learning_rate": 8.540044310964806e-06, "loss": 0.4694, "step": 11220 }, { "epoch": 0.56, "learning_rate": 8.538444447815558e-06, "loss": 0.5323, "step": 11221 }, { "epoch": 0.56, "learning_rate": 8.536844622891385e-06, "loss": 0.2085, "step": 11222 }, { "epoch": 0.56, "learning_rate": 8.535244836234125e-06, "loss": 0.1198, "step": 11223 }, { "epoch": 0.56, "learning_rate": 8.533645087885618e-06, "loss": 0.3755, "step": 11224 }, { "epoch": 0.56, "learning_rate": 8.532045377887704e-06, "loss": 0.3171, "step": 11225 }, { "epoch": 0.56, "learning_rate": 8.530445706282216e-06, "loss": 0.1634, "step": 11226 }, { "epoch": 0.56, "learning_rate": 8.528846073111e-06, "loss": 0.6116, "step": 11227 }, { "epoch": 0.56, "learning_rate": 8.527246478415887e-06, "loss": 0.2358, "step": 11228 }, { "epoch": 0.56, "learning_rate": 8.525646922238713e-06, "loss": 0.3029, "step": 11229 }, { "epoch": 0.56, "learning_rate": 8.52404740462131e-06, "loss": 0.3871, "step": 11230 }, { "epoch": 0.56, "learning_rate": 8.522447925605518e-06, "loss": 0.2628, "step": 11231 }, { "epoch": 0.56, "learning_rate": 8.520848485233162e-06, "loss": 0.1351, "step": 11232 }, { "epoch": 0.56, "learning_rate": 8.51924908354608e-06, "loss": 0.2569, "step": 11233 }, { "epoch": 0.56, "learning_rate": 8.517649720586099e-06, "loss": 0.262, "step": 11234 }, { "epoch": 0.56, "learning_rate": 8.516050396395044e-06, "loss": 0.511, "step": 11235 }, { "epoch": 0.56, "learning_rate": 8.514451111014749e-06, "loss": 0.3748, "step": 11236 }, { "epoch": 0.56, "learning_rate": 8.512851864487042e-06, "loss": 0.1887, "step": 11237 }, { "epoch": 0.56, "learning_rate": 8.511252656853745e-06, "loss": 0.5743, "step": 11238 }, { "epoch": 0.56, "learning_rate": 8.50965348815668e-06, "loss": 0.4148, "step": 11239 }, { "epoch": 0.56, "learning_rate": 8.508054358437682e-06, "loss": 0.3244, "step": 11240 }, { "epoch": 0.56, "learning_rate": 8.506455267738567e-06, "loss": 0.4225, "step": 11241 }, { "epoch": 0.56, "learning_rate": 8.50485621610116e-06, "loss": 0.3515, "step": 11242 }, { "epoch": 0.56, "learning_rate": 8.503257203567279e-06, "loss": 0.4827, "step": 11243 }, { "epoch": 0.56, "learning_rate": 8.50165823017874e-06, "loss": 0.282, "step": 11244 }, { "epoch": 0.56, "learning_rate": 8.500059295977374e-06, "loss": 0.3636, "step": 11245 }, { "epoch": 0.56, "learning_rate": 8.498460401004991e-06, "loss": 0.3522, "step": 11246 }, { "epoch": 0.56, "learning_rate": 8.49686154530341e-06, "loss": 0.238, "step": 11247 }, { "epoch": 0.56, "learning_rate": 8.495262728914444e-06, "loss": 0.2636, "step": 11248 }, { "epoch": 0.56, "learning_rate": 8.493663951879914e-06, "loss": 0.6261, "step": 11249 }, { "epoch": 0.56, "learning_rate": 8.49206521424163e-06, "loss": 0.5203, "step": 11250 }, { "epoch": 0.56, "learning_rate": 8.490466516041404e-06, "loss": 0.2982, "step": 11251 }, { "epoch": 0.56, "learning_rate": 8.488867857321049e-06, "loss": 0.3369, "step": 11252 }, { "epoch": 0.56, "learning_rate": 8.487269238122374e-06, "loss": 0.2337, "step": 11253 }, { "epoch": 0.56, "learning_rate": 8.485670658487193e-06, "loss": 0.3924, "step": 11254 }, { "epoch": 0.56, "learning_rate": 8.484072118457313e-06, "loss": 0.3932, "step": 11255 }, { "epoch": 0.56, "learning_rate": 8.482473618074541e-06, "loss": 0.5531, "step": 11256 }, { "epoch": 0.56, "learning_rate": 8.48087515738068e-06, "loss": 0.3896, "step": 11257 }, { "epoch": 0.56, "learning_rate": 8.479276736417544e-06, "loss": 0.2511, "step": 11258 }, { "epoch": 0.56, "learning_rate": 8.477678355226933e-06, "loss": 0.4388, "step": 11259 }, { "epoch": 0.56, "learning_rate": 8.476080013850651e-06, "loss": 0.2038, "step": 11260 }, { "epoch": 0.56, "learning_rate": 8.4744817123305e-06, "loss": 0.4957, "step": 11261 }, { "epoch": 0.56, "learning_rate": 8.47288345070828e-06, "loss": 0.2016, "step": 11262 }, { "epoch": 0.56, "learning_rate": 8.471285229025795e-06, "loss": 0.5428, "step": 11263 }, { "epoch": 0.56, "learning_rate": 8.469687047324841e-06, "loss": 0.385, "step": 11264 }, { "epoch": 0.56, "learning_rate": 8.46808890564722e-06, "loss": 0.3696, "step": 11265 }, { "epoch": 0.56, "learning_rate": 8.466490804034726e-06, "loss": 0.3302, "step": 11266 }, { "epoch": 0.56, "learning_rate": 8.464892742529154e-06, "loss": 0.5244, "step": 11267 }, { "epoch": 0.56, "learning_rate": 8.463294721172304e-06, "loss": 0.4275, "step": 11268 }, { "epoch": 0.56, "learning_rate": 8.461696740005968e-06, "loss": 0.2786, "step": 11269 }, { "epoch": 0.56, "learning_rate": 8.460098799071941e-06, "loss": 0.2743, "step": 11270 }, { "epoch": 0.56, "learning_rate": 8.458500898412008e-06, "loss": 0.1654, "step": 11271 }, { "epoch": 0.56, "learning_rate": 8.45690303806797e-06, "loss": 0.2423, "step": 11272 }, { "epoch": 0.56, "learning_rate": 8.45530521808161e-06, "loss": 0.3514, "step": 11273 }, { "epoch": 0.56, "learning_rate": 8.453707438494719e-06, "loss": 0.1902, "step": 11274 }, { "epoch": 0.56, "learning_rate": 8.452109699349086e-06, "loss": 0.2792, "step": 11275 }, { "epoch": 0.56, "learning_rate": 8.45051200068649e-06, "loss": 0.3601, "step": 11276 }, { "epoch": 0.56, "learning_rate": 8.44891434254873e-06, "loss": 0.3023, "step": 11277 }, { "epoch": 0.56, "learning_rate": 8.447316724977582e-06, "loss": 0.3614, "step": 11278 }, { "epoch": 0.56, "learning_rate": 8.445719148014831e-06, "loss": 0.4532, "step": 11279 }, { "epoch": 0.56, "learning_rate": 8.444121611702258e-06, "loss": 0.3694, "step": 11280 }, { "epoch": 0.56, "learning_rate": 8.442524116081649e-06, "loss": 0.6165, "step": 11281 }, { "epoch": 0.56, "learning_rate": 8.440926661194781e-06, "loss": 0.6837, "step": 11282 }, { "epoch": 0.56, "learning_rate": 8.439329247083436e-06, "loss": 0.9406, "step": 11283 }, { "epoch": 0.56, "learning_rate": 8.43773187378939e-06, "loss": 0.2973, "step": 11284 }, { "epoch": 0.56, "learning_rate": 8.436134541354417e-06, "loss": 0.3694, "step": 11285 }, { "epoch": 0.56, "learning_rate": 8.4345372498203e-06, "loss": 0.3812, "step": 11286 }, { "epoch": 0.56, "learning_rate": 8.432939999228812e-06, "loss": 0.324, "step": 11287 }, { "epoch": 0.56, "learning_rate": 8.431342789621723e-06, "loss": 0.4848, "step": 11288 }, { "epoch": 0.56, "learning_rate": 8.429745621040808e-06, "loss": 0.3133, "step": 11289 }, { "epoch": 0.56, "learning_rate": 8.42814849352784e-06, "loss": 0.402, "step": 11290 }, { "epoch": 0.56, "learning_rate": 8.426551407124593e-06, "loss": 0.2808, "step": 11291 }, { "epoch": 0.56, "learning_rate": 8.424954361872832e-06, "loss": 0.5128, "step": 11292 }, { "epoch": 0.56, "learning_rate": 8.423357357814326e-06, "loss": 0.4087, "step": 11293 }, { "epoch": 0.56, "learning_rate": 8.421760394990838e-06, "loss": 0.2677, "step": 11294 }, { "epoch": 0.56, "learning_rate": 8.420163473444148e-06, "loss": 0.2879, "step": 11295 }, { "epoch": 0.56, "learning_rate": 8.418566593216009e-06, "loss": 0.3648, "step": 11296 }, { "epoch": 0.56, "learning_rate": 8.416969754348191e-06, "loss": 0.3551, "step": 11297 }, { "epoch": 0.56, "learning_rate": 8.415372956882456e-06, "loss": 0.3374, "step": 11298 }, { "epoch": 0.56, "learning_rate": 8.413776200860563e-06, "loss": 0.3258, "step": 11299 }, { "epoch": 0.56, "learning_rate": 8.412179486324278e-06, "loss": 0.3728, "step": 11300 }, { "epoch": 0.56, "learning_rate": 8.41058281331536e-06, "loss": 0.3464, "step": 11301 }, { "epoch": 0.56, "learning_rate": 8.408986181875565e-06, "loss": 0.2427, "step": 11302 }, { "epoch": 0.56, "learning_rate": 8.40738959204665e-06, "loss": 0.4294, "step": 11303 }, { "epoch": 0.56, "learning_rate": 8.405793043870378e-06, "loss": 0.5415, "step": 11304 }, { "epoch": 0.56, "learning_rate": 8.4041965373885e-06, "loss": 0.6639, "step": 11305 }, { "epoch": 0.56, "learning_rate": 8.402600072642772e-06, "loss": 0.3484, "step": 11306 }, { "epoch": 0.56, "learning_rate": 8.401003649674948e-06, "loss": 0.206, "step": 11307 }, { "epoch": 0.56, "learning_rate": 8.399407268526775e-06, "loss": 0.4055, "step": 11308 }, { "epoch": 0.56, "learning_rate": 8.397810929240012e-06, "loss": 0.3559, "step": 11309 }, { "epoch": 0.56, "learning_rate": 8.396214631856407e-06, "loss": 0.2745, "step": 11310 }, { "epoch": 0.56, "learning_rate": 8.394618376417705e-06, "loss": 0.2311, "step": 11311 }, { "epoch": 0.56, "learning_rate": 8.393022162965654e-06, "loss": 0.5163, "step": 11312 }, { "epoch": 0.56, "learning_rate": 8.391425991542007e-06, "loss": 0.2945, "step": 11313 }, { "epoch": 0.56, "learning_rate": 8.389829862188506e-06, "loss": 0.3115, "step": 11314 }, { "epoch": 0.56, "learning_rate": 8.388233774946898e-06, "loss": 0.3277, "step": 11315 }, { "epoch": 0.57, "learning_rate": 8.386637729858922e-06, "loss": 0.3836, "step": 11316 }, { "epoch": 0.57, "learning_rate": 8.38504172696632e-06, "loss": 0.5098, "step": 11317 }, { "epoch": 0.57, "learning_rate": 8.383445766310841e-06, "loss": 0.3614, "step": 11318 }, { "epoch": 0.57, "learning_rate": 8.381849847934221e-06, "loss": 0.5216, "step": 11319 }, { "epoch": 0.57, "learning_rate": 8.380253971878199e-06, "loss": 0.3053, "step": 11320 }, { "epoch": 0.57, "learning_rate": 8.37865813818451e-06, "loss": 0.2319, "step": 11321 }, { "epoch": 0.57, "learning_rate": 8.377062346894896e-06, "loss": 0.3029, "step": 11322 }, { "epoch": 0.57, "learning_rate": 8.37546659805109e-06, "loss": 0.3368, "step": 11323 }, { "epoch": 0.57, "learning_rate": 8.373870891694829e-06, "loss": 0.2345, "step": 11324 }, { "epoch": 0.57, "learning_rate": 8.372275227867844e-06, "loss": 0.454, "step": 11325 }, { "epoch": 0.57, "learning_rate": 8.370679606611863e-06, "loss": 0.368, "step": 11326 }, { "epoch": 0.57, "learning_rate": 8.369084027968629e-06, "loss": 0.3863, "step": 11327 }, { "epoch": 0.57, "learning_rate": 8.367488491979864e-06, "loss": 0.3207, "step": 11328 }, { "epoch": 0.57, "learning_rate": 8.365892998687302e-06, "loss": 0.5057, "step": 11329 }, { "epoch": 0.57, "learning_rate": 8.364297548132668e-06, "loss": 0.4119, "step": 11330 }, { "epoch": 0.57, "learning_rate": 8.362702140357682e-06, "loss": 0.4231, "step": 11331 }, { "epoch": 0.57, "learning_rate": 8.361106775404084e-06, "loss": 0.313, "step": 11332 }, { "epoch": 0.57, "learning_rate": 8.359511453313592e-06, "loss": 0.5393, "step": 11333 }, { "epoch": 0.57, "learning_rate": 8.357916174127928e-06, "loss": 0.4329, "step": 11334 }, { "epoch": 0.57, "learning_rate": 8.356320937888813e-06, "loss": 0.5058, "step": 11335 }, { "epoch": 0.57, "learning_rate": 8.354725744637975e-06, "loss": 0.3047, "step": 11336 }, { "epoch": 0.57, "learning_rate": 8.35313059441713e-06, "loss": 0.4303, "step": 11337 }, { "epoch": 0.57, "learning_rate": 8.351535487267995e-06, "loss": 0.3897, "step": 11338 }, { "epoch": 0.57, "learning_rate": 8.349940423232293e-06, "loss": 0.359, "step": 11339 }, { "epoch": 0.57, "learning_rate": 8.348345402351732e-06, "loss": 0.385, "step": 11340 }, { "epoch": 0.57, "learning_rate": 8.346750424668038e-06, "loss": 0.3919, "step": 11341 }, { "epoch": 0.57, "learning_rate": 8.345155490222923e-06, "loss": 0.5338, "step": 11342 }, { "epoch": 0.57, "learning_rate": 8.343560599058099e-06, "loss": 0.3943, "step": 11343 }, { "epoch": 0.57, "learning_rate": 8.341965751215272e-06, "loss": 0.2925, "step": 11344 }, { "epoch": 0.57, "learning_rate": 8.340370946736164e-06, "loss": 0.348, "step": 11345 }, { "epoch": 0.57, "learning_rate": 8.33877618566248e-06, "loss": 0.3271, "step": 11346 }, { "epoch": 0.57, "learning_rate": 8.337181468035929e-06, "loss": 0.1907, "step": 11347 }, { "epoch": 0.57, "learning_rate": 8.33558679389822e-06, "loss": 0.2481, "step": 11348 }, { "epoch": 0.57, "learning_rate": 8.333992163291052e-06, "loss": 0.3793, "step": 11349 }, { "epoch": 0.57, "learning_rate": 8.332397576256142e-06, "loss": 0.3493, "step": 11350 }, { "epoch": 0.57, "learning_rate": 8.33080303283519e-06, "loss": 0.2999, "step": 11351 }, { "epoch": 0.57, "learning_rate": 8.329208533069897e-06, "loss": 0.2257, "step": 11352 }, { "epoch": 0.57, "learning_rate": 8.327614077001962e-06, "loss": 0.6804, "step": 11353 }, { "epoch": 0.57, "learning_rate": 8.326019664673098e-06, "loss": 0.3073, "step": 11354 }, { "epoch": 0.57, "learning_rate": 8.324425296124993e-06, "loss": 0.5032, "step": 11355 }, { "epoch": 0.57, "learning_rate": 8.322830971399352e-06, "loss": 0.6591, "step": 11356 }, { "epoch": 0.57, "learning_rate": 8.32123669053787e-06, "loss": 0.3696, "step": 11357 }, { "epoch": 0.57, "learning_rate": 8.31964245358224e-06, "loss": 0.359, "step": 11358 }, { "epoch": 0.57, "learning_rate": 8.318048260574165e-06, "loss": 0.5756, "step": 11359 }, { "epoch": 0.57, "learning_rate": 8.316454111555335e-06, "loss": 0.3584, "step": 11360 }, { "epoch": 0.57, "learning_rate": 8.31486000656744e-06, "loss": 0.5078, "step": 11361 }, { "epoch": 0.57, "learning_rate": 8.313265945652177e-06, "loss": 0.405, "step": 11362 }, { "epoch": 0.57, "learning_rate": 8.311671928851229e-06, "loss": 0.2524, "step": 11363 }, { "epoch": 0.57, "learning_rate": 8.310077956206293e-06, "loss": 0.2498, "step": 11364 }, { "epoch": 0.57, "learning_rate": 8.308484027759057e-06, "loss": 0.4434, "step": 11365 }, { "epoch": 0.57, "learning_rate": 8.306890143551205e-06, "loss": 0.4452, "step": 11366 }, { "epoch": 0.57, "learning_rate": 8.305296303624417e-06, "loss": 0.3082, "step": 11367 }, { "epoch": 0.57, "learning_rate": 8.30370250802039e-06, "loss": 0.2375, "step": 11368 }, { "epoch": 0.57, "learning_rate": 8.302108756780804e-06, "loss": 0.1529, "step": 11369 }, { "epoch": 0.57, "learning_rate": 8.30051504994734e-06, "loss": 0.3336, "step": 11370 }, { "epoch": 0.57, "learning_rate": 8.298921387561677e-06, "loss": 0.1305, "step": 11371 }, { "epoch": 0.57, "learning_rate": 8.297327769665495e-06, "loss": 0.2584, "step": 11372 }, { "epoch": 0.57, "learning_rate": 8.295734196300477e-06, "loss": 0.5068, "step": 11373 }, { "epoch": 0.57, "learning_rate": 8.294140667508302e-06, "loss": 0.2924, "step": 11374 }, { "epoch": 0.57, "learning_rate": 8.29254718333064e-06, "loss": 0.3439, "step": 11375 }, { "epoch": 0.57, "learning_rate": 8.290953743809168e-06, "loss": 0.32, "step": 11376 }, { "epoch": 0.57, "learning_rate": 8.289360348985566e-06, "loss": 0.1873, "step": 11377 }, { "epoch": 0.57, "learning_rate": 8.287766998901504e-06, "loss": 0.3131, "step": 11378 }, { "epoch": 0.57, "learning_rate": 8.286173693598651e-06, "loss": 0.3634, "step": 11379 }, { "epoch": 0.57, "learning_rate": 8.284580433118684e-06, "loss": 0.4502, "step": 11380 }, { "epoch": 0.57, "learning_rate": 8.282987217503263e-06, "loss": 0.3595, "step": 11381 }, { "epoch": 0.57, "learning_rate": 8.281394046794066e-06, "loss": 0.284, "step": 11382 }, { "epoch": 0.57, "learning_rate": 8.27980092103276e-06, "loss": 0.3951, "step": 11383 }, { "epoch": 0.57, "learning_rate": 8.278207840261004e-06, "loss": 0.3398, "step": 11384 }, { "epoch": 0.57, "learning_rate": 8.276614804520463e-06, "loss": 0.2332, "step": 11385 }, { "epoch": 0.57, "learning_rate": 8.275021813852809e-06, "loss": 0.332, "step": 11386 }, { "epoch": 0.57, "learning_rate": 8.273428868299701e-06, "loss": 0.3457, "step": 11387 }, { "epoch": 0.57, "learning_rate": 8.271835967902799e-06, "loss": 0.4733, "step": 11388 }, { "epoch": 0.57, "learning_rate": 8.270243112703765e-06, "loss": 0.3578, "step": 11389 }, { "epoch": 0.57, "learning_rate": 8.268650302744252e-06, "loss": 0.3006, "step": 11390 }, { "epoch": 0.57, "learning_rate": 8.267057538065926e-06, "loss": 0.2951, "step": 11391 }, { "epoch": 0.57, "learning_rate": 8.26546481871044e-06, "loss": 0.2848, "step": 11392 }, { "epoch": 0.57, "learning_rate": 8.26387214471945e-06, "loss": 0.5093, "step": 11393 }, { "epoch": 0.57, "learning_rate": 8.26227951613461e-06, "loss": 0.4813, "step": 11394 }, { "epoch": 0.57, "learning_rate": 8.260686932997574e-06, "loss": 0.3349, "step": 11395 }, { "epoch": 0.57, "learning_rate": 8.259094395349993e-06, "loss": 0.335, "step": 11396 }, { "epoch": 0.57, "learning_rate": 8.257501903233518e-06, "loss": 0.5397, "step": 11397 }, { "epoch": 0.57, "learning_rate": 8.255909456689799e-06, "loss": 0.4094, "step": 11398 }, { "epoch": 0.57, "learning_rate": 8.254317055760478e-06, "loss": 0.3877, "step": 11399 }, { "epoch": 0.57, "learning_rate": 8.252724700487213e-06, "loss": 0.3601, "step": 11400 }, { "epoch": 0.57, "learning_rate": 8.251132390911645e-06, "loss": 0.4286, "step": 11401 }, { "epoch": 0.57, "learning_rate": 8.249540127075419e-06, "loss": 0.2693, "step": 11402 }, { "epoch": 0.57, "learning_rate": 8.24794790902018e-06, "loss": 0.4701, "step": 11403 }, { "epoch": 0.57, "learning_rate": 8.246355736787562e-06, "loss": 0.4294, "step": 11404 }, { "epoch": 0.57, "learning_rate": 8.244763610419218e-06, "loss": 0.1832, "step": 11405 }, { "epoch": 0.57, "learning_rate": 8.243171529956782e-06, "loss": 0.4287, "step": 11406 }, { "epoch": 0.57, "learning_rate": 8.241579495441895e-06, "loss": 0.4748, "step": 11407 }, { "epoch": 0.57, "learning_rate": 8.23998750691619e-06, "loss": 0.417, "step": 11408 }, { "epoch": 0.57, "learning_rate": 8.23839556442131e-06, "loss": 0.222, "step": 11409 }, { "epoch": 0.57, "learning_rate": 8.236803667998885e-06, "loss": 0.3489, "step": 11410 }, { "epoch": 0.57, "learning_rate": 8.23521181769055e-06, "loss": 0.2719, "step": 11411 }, { "epoch": 0.57, "learning_rate": 8.233620013537941e-06, "loss": 0.2221, "step": 11412 }, { "epoch": 0.57, "learning_rate": 8.232028255582681e-06, "loss": 0.5268, "step": 11413 }, { "epoch": 0.57, "learning_rate": 8.230436543866411e-06, "loss": 0.3343, "step": 11414 }, { "epoch": 0.57, "learning_rate": 8.228844878430756e-06, "loss": 0.4515, "step": 11415 }, { "epoch": 0.57, "learning_rate": 8.227253259317343e-06, "loss": 0.361, "step": 11416 }, { "epoch": 0.57, "learning_rate": 8.225661686567797e-06, "loss": 0.2026, "step": 11417 }, { "epoch": 0.57, "learning_rate": 8.224070160223747e-06, "loss": 0.501, "step": 11418 }, { "epoch": 0.57, "learning_rate": 8.222478680326817e-06, "loss": 0.18, "step": 11419 }, { "epoch": 0.57, "learning_rate": 8.22088724691863e-06, "loss": 0.2587, "step": 11420 }, { "epoch": 0.57, "learning_rate": 8.219295860040804e-06, "loss": 0.2128, "step": 11421 }, { "epoch": 0.57, "learning_rate": 8.21770451973496e-06, "loss": 0.3814, "step": 11422 }, { "epoch": 0.57, "learning_rate": 8.216113226042724e-06, "loss": 0.3428, "step": 11423 }, { "epoch": 0.57, "learning_rate": 8.21452197900571e-06, "loss": 0.1657, "step": 11424 }, { "epoch": 0.57, "learning_rate": 8.212930778665536e-06, "loss": 0.4424, "step": 11425 }, { "epoch": 0.57, "learning_rate": 8.211339625063812e-06, "loss": 0.4058, "step": 11426 }, { "epoch": 0.57, "learning_rate": 8.20974851824216e-06, "loss": 0.4131, "step": 11427 }, { "epoch": 0.57, "learning_rate": 8.208157458242191e-06, "loss": 0.2844, "step": 11428 }, { "epoch": 0.57, "learning_rate": 8.20656644510552e-06, "loss": 0.5027, "step": 11429 }, { "epoch": 0.57, "learning_rate": 8.204975478873751e-06, "loss": 0.2597, "step": 11430 }, { "epoch": 0.57, "learning_rate": 8.203384559588498e-06, "loss": 0.4052, "step": 11431 }, { "epoch": 0.57, "learning_rate": 8.201793687291369e-06, "loss": 0.3571, "step": 11432 }, { "epoch": 0.57, "learning_rate": 8.200202862023972e-06, "loss": 0.4851, "step": 11433 }, { "epoch": 0.57, "learning_rate": 8.19861208382791e-06, "loss": 0.2076, "step": 11434 }, { "epoch": 0.57, "learning_rate": 8.197021352744791e-06, "loss": 0.5169, "step": 11435 }, { "epoch": 0.57, "learning_rate": 8.195430668816212e-06, "loss": 0.5615, "step": 11436 }, { "epoch": 0.57, "learning_rate": 8.193840032083786e-06, "loss": 0.5736, "step": 11437 }, { "epoch": 0.57, "learning_rate": 8.192249442589106e-06, "loss": 0.2534, "step": 11438 }, { "epoch": 0.57, "learning_rate": 8.190658900373776e-06, "loss": 0.4482, "step": 11439 }, { "epoch": 0.57, "learning_rate": 8.189068405479388e-06, "loss": 0.2689, "step": 11440 }, { "epoch": 0.57, "learning_rate": 8.187477957947549e-06, "loss": 0.4289, "step": 11441 }, { "epoch": 0.57, "learning_rate": 8.185887557819848e-06, "loss": 0.4675, "step": 11442 }, { "epoch": 0.57, "learning_rate": 8.184297205137884e-06, "loss": 0.2647, "step": 11443 }, { "epoch": 0.57, "learning_rate": 8.182706899943247e-06, "loss": 0.2775, "step": 11444 }, { "epoch": 0.57, "learning_rate": 8.181116642277529e-06, "loss": 0.3872, "step": 11445 }, { "epoch": 0.57, "learning_rate": 8.179526432182325e-06, "loss": 0.3503, "step": 11446 }, { "epoch": 0.57, "learning_rate": 8.177936269699223e-06, "loss": 0.143, "step": 11447 }, { "epoch": 0.57, "learning_rate": 8.17634615486981e-06, "loss": 0.4621, "step": 11448 }, { "epoch": 0.57, "learning_rate": 8.174756087735672e-06, "loss": 0.0603, "step": 11449 }, { "epoch": 0.57, "learning_rate": 8.173166068338403e-06, "loss": 0.5725, "step": 11450 }, { "epoch": 0.57, "learning_rate": 8.171576096719581e-06, "loss": 0.2938, "step": 11451 }, { "epoch": 0.57, "learning_rate": 8.169986172920793e-06, "loss": 0.3644, "step": 11452 }, { "epoch": 0.57, "learning_rate": 8.168396296983618e-06, "loss": 0.1707, "step": 11453 }, { "epoch": 0.57, "learning_rate": 8.166806468949638e-06, "loss": 0.4494, "step": 11454 }, { "epoch": 0.57, "learning_rate": 8.165216688860435e-06, "loss": 0.4484, "step": 11455 }, { "epoch": 0.57, "learning_rate": 8.163626956757587e-06, "loss": 0.2412, "step": 11456 }, { "epoch": 0.57, "learning_rate": 8.162037272682671e-06, "loss": 0.5849, "step": 11457 }, { "epoch": 0.57, "learning_rate": 8.160447636677257e-06, "loss": 0.2877, "step": 11458 }, { "epoch": 0.57, "learning_rate": 8.158858048782932e-06, "loss": 0.2698, "step": 11459 }, { "epoch": 0.57, "learning_rate": 8.157268509041263e-06, "loss": 0.5545, "step": 11460 }, { "epoch": 0.57, "learning_rate": 8.155679017493822e-06, "loss": 0.3759, "step": 11461 }, { "epoch": 0.57, "learning_rate": 8.15408957418218e-06, "loss": 0.2881, "step": 11462 }, { "epoch": 0.57, "learning_rate": 8.152500179147904e-06, "loss": 0.4274, "step": 11463 }, { "epoch": 0.57, "learning_rate": 8.15091083243257e-06, "loss": 0.3587, "step": 11464 }, { "epoch": 0.57, "learning_rate": 8.149321534077742e-06, "loss": 0.5112, "step": 11465 }, { "epoch": 0.57, "learning_rate": 8.147732284124985e-06, "loss": 0.3518, "step": 11466 }, { "epoch": 0.57, "learning_rate": 8.146143082615864e-06, "loss": 0.2248, "step": 11467 }, { "epoch": 0.57, "learning_rate": 8.144553929591939e-06, "loss": 0.2278, "step": 11468 }, { "epoch": 0.57, "learning_rate": 8.14296482509478e-06, "loss": 0.2254, "step": 11469 }, { "epoch": 0.57, "learning_rate": 8.141375769165942e-06, "loss": 0.2105, "step": 11470 }, { "epoch": 0.57, "learning_rate": 8.139786761846989e-06, "loss": 0.2243, "step": 11471 }, { "epoch": 0.57, "learning_rate": 8.138197803179471e-06, "loss": 0.2718, "step": 11472 }, { "epoch": 0.57, "learning_rate": 8.136608893204953e-06, "loss": 0.483, "step": 11473 }, { "epoch": 0.57, "learning_rate": 8.135020031964991e-06, "loss": 0.4741, "step": 11474 }, { "epoch": 0.57, "learning_rate": 8.133431219501138e-06, "loss": 0.6536, "step": 11475 }, { "epoch": 0.57, "learning_rate": 8.131842455854945e-06, "loss": 0.3537, "step": 11476 }, { "epoch": 0.57, "learning_rate": 8.130253741067964e-06, "loss": 0.5807, "step": 11477 }, { "epoch": 0.57, "learning_rate": 8.12866507518175e-06, "loss": 0.4124, "step": 11478 }, { "epoch": 0.57, "learning_rate": 8.12707645823785e-06, "loss": 0.4433, "step": 11479 }, { "epoch": 0.57, "learning_rate": 8.125487890277813e-06, "loss": 0.3537, "step": 11480 }, { "epoch": 0.57, "learning_rate": 8.123899371343182e-06, "loss": 0.6058, "step": 11481 }, { "epoch": 0.57, "learning_rate": 8.122310901475508e-06, "loss": 0.2804, "step": 11482 }, { "epoch": 0.57, "learning_rate": 8.120722480716333e-06, "loss": 0.2083, "step": 11483 }, { "epoch": 0.57, "learning_rate": 8.119134109107198e-06, "loss": 0.4965, "step": 11484 }, { "epoch": 0.57, "learning_rate": 8.11754578668965e-06, "loss": 0.3624, "step": 11485 }, { "epoch": 0.57, "learning_rate": 8.115957513505222e-06, "loss": 0.3746, "step": 11486 }, { "epoch": 0.57, "learning_rate": 8.11436928959546e-06, "loss": 0.1947, "step": 11487 }, { "epoch": 0.57, "learning_rate": 8.1127811150019e-06, "loss": 0.2661, "step": 11488 }, { "epoch": 0.57, "learning_rate": 8.111192989766079e-06, "loss": 0.2355, "step": 11489 }, { "epoch": 0.57, "learning_rate": 8.109604913929528e-06, "loss": 0.4818, "step": 11490 }, { "epoch": 0.57, "learning_rate": 8.108016887533788e-06, "loss": 0.3295, "step": 11491 }, { "epoch": 0.57, "learning_rate": 8.106428910620389e-06, "loss": 0.2475, "step": 11492 }, { "epoch": 0.57, "learning_rate": 8.10484098323086e-06, "loss": 0.4232, "step": 11493 }, { "epoch": 0.57, "learning_rate": 8.103253105406734e-06, "loss": 0.4263, "step": 11494 }, { "epoch": 0.57, "learning_rate": 8.101665277189533e-06, "loss": 0.3831, "step": 11495 }, { "epoch": 0.57, "learning_rate": 8.100077498620796e-06, "loss": 0.3171, "step": 11496 }, { "epoch": 0.57, "learning_rate": 8.098489769742042e-06, "loss": 0.2546, "step": 11497 }, { "epoch": 0.57, "learning_rate": 8.0969020905948e-06, "loss": 0.3692, "step": 11498 }, { "epoch": 0.57, "learning_rate": 8.095314461220589e-06, "loss": 0.3248, "step": 11499 }, { "epoch": 0.57, "learning_rate": 8.09372688166093e-06, "loss": 0.2688, "step": 11500 }, { "epoch": 0.57, "learning_rate": 8.092139351957351e-06, "loss": 0.4155, "step": 11501 }, { "epoch": 0.57, "learning_rate": 8.09055187215137e-06, "loss": 0.5539, "step": 11502 }, { "epoch": 0.57, "learning_rate": 8.088964442284504e-06, "loss": 0.3596, "step": 11503 }, { "epoch": 0.57, "learning_rate": 8.087377062398266e-06, "loss": 0.3935, "step": 11504 }, { "epoch": 0.57, "learning_rate": 8.085789732534179e-06, "loss": 0.6549, "step": 11505 }, { "epoch": 0.57, "learning_rate": 8.084202452733755e-06, "loss": 0.589, "step": 11506 }, { "epoch": 0.57, "learning_rate": 8.082615223038505e-06, "loss": 0.3611, "step": 11507 }, { "epoch": 0.57, "learning_rate": 8.081028043489945e-06, "loss": 0.4459, "step": 11508 }, { "epoch": 0.57, "learning_rate": 8.079440914129577e-06, "loss": 0.458, "step": 11509 }, { "epoch": 0.57, "learning_rate": 8.07785383499892e-06, "loss": 0.5516, "step": 11510 }, { "epoch": 0.57, "learning_rate": 8.07626680613948e-06, "loss": 0.1621, "step": 11511 }, { "epoch": 0.57, "learning_rate": 8.074679827592762e-06, "loss": 0.4297, "step": 11512 }, { "epoch": 0.57, "learning_rate": 8.073092899400267e-06, "loss": 0.2851, "step": 11513 }, { "epoch": 0.57, "learning_rate": 8.071506021603506e-06, "loss": 0.3276, "step": 11514 }, { "epoch": 0.57, "learning_rate": 8.069919194243981e-06, "loss": 0.3062, "step": 11515 }, { "epoch": 0.58, "learning_rate": 8.068332417363192e-06, "loss": 0.4134, "step": 11516 }, { "epoch": 0.58, "learning_rate": 8.066745691002636e-06, "loss": 0.3781, "step": 11517 }, { "epoch": 0.58, "learning_rate": 8.065159015203814e-06, "loss": 0.6735, "step": 11518 }, { "epoch": 0.58, "learning_rate": 8.063572390008227e-06, "loss": 0.1921, "step": 11519 }, { "epoch": 0.58, "learning_rate": 8.061985815457366e-06, "loss": 0.4337, "step": 11520 }, { "epoch": 0.58, "learning_rate": 8.060399291592729e-06, "loss": 0.2994, "step": 11521 }, { "epoch": 0.58, "learning_rate": 8.058812818455802e-06, "loss": 0.4766, "step": 11522 }, { "epoch": 0.58, "learning_rate": 8.057226396088089e-06, "loss": 0.384, "step": 11523 }, { "epoch": 0.58, "learning_rate": 8.055640024531074e-06, "loss": 0.2724, "step": 11524 }, { "epoch": 0.58, "learning_rate": 8.054053703826248e-06, "loss": 0.3171, "step": 11525 }, { "epoch": 0.58, "learning_rate": 8.052467434015099e-06, "loss": 0.3045, "step": 11526 }, { "epoch": 0.58, "learning_rate": 8.050881215139112e-06, "loss": 0.4127, "step": 11527 }, { "epoch": 0.58, "learning_rate": 8.049295047239775e-06, "loss": 0.2677, "step": 11528 }, { "epoch": 0.58, "learning_rate": 8.047708930358571e-06, "loss": 0.4906, "step": 11529 }, { "epoch": 0.58, "learning_rate": 8.046122864536982e-06, "loss": 0.3589, "step": 11530 }, { "epoch": 0.58, "learning_rate": 8.044536849816488e-06, "loss": 0.2005, "step": 11531 }, { "epoch": 0.58, "learning_rate": 8.042950886238576e-06, "loss": 0.5488, "step": 11532 }, { "epoch": 0.58, "learning_rate": 8.041364973844719e-06, "loss": 0.3378, "step": 11533 }, { "epoch": 0.58, "learning_rate": 8.039779112676397e-06, "loss": 0.4843, "step": 11534 }, { "epoch": 0.58, "learning_rate": 8.038193302775083e-06, "loss": 0.2779, "step": 11535 }, { "epoch": 0.58, "learning_rate": 8.036607544182249e-06, "loss": 0.3624, "step": 11536 }, { "epoch": 0.58, "learning_rate": 8.035021836939378e-06, "loss": 0.3029, "step": 11537 }, { "epoch": 0.58, "learning_rate": 8.03343618108794e-06, "loss": 0.4749, "step": 11538 }, { "epoch": 0.58, "learning_rate": 8.0318505766694e-06, "loss": 0.2872, "step": 11539 }, { "epoch": 0.58, "learning_rate": 8.030265023725231e-06, "loss": 0.3899, "step": 11540 }, { "epoch": 0.58, "learning_rate": 8.028679522296898e-06, "loss": 0.5577, "step": 11541 }, { "epoch": 0.58, "learning_rate": 8.027094072425872e-06, "loss": 0.5661, "step": 11542 }, { "epoch": 0.58, "learning_rate": 8.025508674153617e-06, "loss": 0.2445, "step": 11543 }, { "epoch": 0.58, "learning_rate": 8.023923327521597e-06, "loss": 0.445, "step": 11544 }, { "epoch": 0.58, "learning_rate": 8.022338032571271e-06, "loss": 0.6511, "step": 11545 }, { "epoch": 0.58, "learning_rate": 8.020752789344106e-06, "loss": 0.6627, "step": 11546 }, { "epoch": 0.58, "learning_rate": 8.01916759788156e-06, "loss": 0.2567, "step": 11547 }, { "epoch": 0.58, "learning_rate": 8.01758245822509e-06, "loss": 0.4807, "step": 11548 }, { "epoch": 0.58, "learning_rate": 8.015997370416155e-06, "loss": 0.3882, "step": 11549 }, { "epoch": 0.58, "learning_rate": 8.014412334496206e-06, "loss": 0.4665, "step": 11550 }, { "epoch": 0.58, "learning_rate": 8.012827350506707e-06, "loss": 0.4504, "step": 11551 }, { "epoch": 0.58, "learning_rate": 8.011242418489105e-06, "loss": 0.2121, "step": 11552 }, { "epoch": 0.58, "learning_rate": 8.009657538484853e-06, "loss": 0.3739, "step": 11553 }, { "epoch": 0.58, "learning_rate": 8.008072710535398e-06, "loss": 0.4071, "step": 11554 }, { "epoch": 0.58, "learning_rate": 8.006487934682195e-06, "loss": 0.3195, "step": 11555 }, { "epoch": 0.58, "learning_rate": 8.00490321096669e-06, "loss": 0.312, "step": 11556 }, { "epoch": 0.58, "learning_rate": 8.003318539430328e-06, "loss": 0.5511, "step": 11557 }, { "epoch": 0.58, "learning_rate": 8.001733920114554e-06, "loss": 0.2819, "step": 11558 }, { "epoch": 0.58, "learning_rate": 8.000149353060807e-06, "loss": 0.5589, "step": 11559 }, { "epoch": 0.58, "learning_rate": 7.99856483831054e-06, "loss": 0.3386, "step": 11560 }, { "epoch": 0.58, "learning_rate": 7.996980375905187e-06, "loss": 0.48, "step": 11561 }, { "epoch": 0.58, "learning_rate": 7.99539596588619e-06, "loss": 0.3946, "step": 11562 }, { "epoch": 0.58, "learning_rate": 7.993811608294983e-06, "loss": 0.2367, "step": 11563 }, { "epoch": 0.58, "learning_rate": 7.992227303173009e-06, "loss": 0.3563, "step": 11564 }, { "epoch": 0.58, "learning_rate": 7.9906430505617e-06, "loss": 0.2858, "step": 11565 }, { "epoch": 0.58, "learning_rate": 7.989058850502489e-06, "loss": 0.5508, "step": 11566 }, { "epoch": 0.58, "learning_rate": 7.98747470303681e-06, "loss": 0.3784, "step": 11567 }, { "epoch": 0.58, "learning_rate": 7.985890608206092e-06, "loss": 0.5067, "step": 11568 }, { "epoch": 0.58, "learning_rate": 7.98430656605177e-06, "loss": 0.4034, "step": 11569 }, { "epoch": 0.58, "learning_rate": 7.982722576615269e-06, "loss": 0.2123, "step": 11570 }, { "epoch": 0.58, "learning_rate": 7.981138639938018e-06, "loss": 0.1844, "step": 11571 }, { "epoch": 0.58, "learning_rate": 7.97955475606144e-06, "loss": 0.2418, "step": 11572 }, { "epoch": 0.58, "learning_rate": 7.977970925026959e-06, "loss": 0.461, "step": 11573 }, { "epoch": 0.58, "learning_rate": 7.976387146876002e-06, "loss": 0.43, "step": 11574 }, { "epoch": 0.58, "learning_rate": 7.974803421649991e-06, "loss": 0.278, "step": 11575 }, { "epoch": 0.58, "learning_rate": 7.973219749390342e-06, "loss": 0.427, "step": 11576 }, { "epoch": 0.58, "learning_rate": 7.971636130138474e-06, "loss": 0.406, "step": 11577 }, { "epoch": 0.58, "learning_rate": 7.97005256393581e-06, "loss": 0.3335, "step": 11578 }, { "epoch": 0.58, "learning_rate": 7.968469050823761e-06, "loss": 0.406, "step": 11579 }, { "epoch": 0.58, "learning_rate": 7.966885590843745e-06, "loss": 0.5576, "step": 11580 }, { "epoch": 0.58, "learning_rate": 7.965302184037171e-06, "loss": 0.4953, "step": 11581 }, { "epoch": 0.58, "learning_rate": 7.96371883044545e-06, "loss": 0.4336, "step": 11582 }, { "epoch": 0.58, "learning_rate": 7.962135530110002e-06, "loss": 0.5303, "step": 11583 }, { "epoch": 0.58, "learning_rate": 7.960552283072228e-06, "loss": 0.2933, "step": 11584 }, { "epoch": 0.58, "learning_rate": 7.958969089373538e-06, "loss": 0.3684, "step": 11585 }, { "epoch": 0.58, "learning_rate": 7.957385949055333e-06, "loss": 0.3128, "step": 11586 }, { "epoch": 0.58, "learning_rate": 7.95580286215903e-06, "loss": 0.2521, "step": 11587 }, { "epoch": 0.58, "learning_rate": 7.954219828726024e-06, "loss": 0.2743, "step": 11588 }, { "epoch": 0.58, "learning_rate": 7.952636848797721e-06, "loss": 0.2383, "step": 11589 }, { "epoch": 0.58, "learning_rate": 7.951053922415517e-06, "loss": 0.403, "step": 11590 }, { "epoch": 0.58, "learning_rate": 7.949471049620813e-06, "loss": 0.4122, "step": 11591 }, { "epoch": 0.58, "learning_rate": 7.94788823045501e-06, "loss": 0.5788, "step": 11592 }, { "epoch": 0.58, "learning_rate": 7.946305464959503e-06, "loss": 0.2906, "step": 11593 }, { "epoch": 0.58, "learning_rate": 7.944722753175687e-06, "loss": 0.4343, "step": 11594 }, { "epoch": 0.58, "learning_rate": 7.94314009514495e-06, "loss": 0.4535, "step": 11595 }, { "epoch": 0.58, "learning_rate": 7.941557490908693e-06, "loss": 0.3541, "step": 11596 }, { "epoch": 0.58, "learning_rate": 7.939974940508306e-06, "loss": 0.3554, "step": 11597 }, { "epoch": 0.58, "learning_rate": 7.938392443985175e-06, "loss": 0.3361, "step": 11598 }, { "epoch": 0.58, "learning_rate": 7.93681000138069e-06, "loss": 0.3408, "step": 11599 }, { "epoch": 0.58, "learning_rate": 7.935227612736236e-06, "loss": 0.5116, "step": 11600 }, { "epoch": 0.58, "learning_rate": 7.9336452780932e-06, "loss": 0.4186, "step": 11601 }, { "epoch": 0.58, "learning_rate": 7.932062997492966e-06, "loss": 0.5154, "step": 11602 }, { "epoch": 0.58, "learning_rate": 7.930480770976915e-06, "loss": 0.2079, "step": 11603 }, { "epoch": 0.58, "learning_rate": 7.928898598586431e-06, "loss": 0.3489, "step": 11604 }, { "epoch": 0.58, "learning_rate": 7.927316480362886e-06, "loss": 0.468, "step": 11605 }, { "epoch": 0.58, "learning_rate": 7.925734416347667e-06, "loss": 0.3182, "step": 11606 }, { "epoch": 0.58, "learning_rate": 7.924152406582149e-06, "loss": 0.1873, "step": 11607 }, { "epoch": 0.58, "learning_rate": 7.922570451107703e-06, "loss": 0.3118, "step": 11608 }, { "epoch": 0.58, "learning_rate": 7.920988549965704e-06, "loss": 0.422, "step": 11609 }, { "epoch": 0.58, "learning_rate": 7.919406703197533e-06, "loss": 0.4183, "step": 11610 }, { "epoch": 0.58, "learning_rate": 7.917824910844552e-06, "loss": 0.5234, "step": 11611 }, { "epoch": 0.58, "learning_rate": 7.916243172948132e-06, "loss": 0.4565, "step": 11612 }, { "epoch": 0.58, "learning_rate": 7.914661489549645e-06, "loss": 0.3602, "step": 11613 }, { "epoch": 0.58, "learning_rate": 7.913079860690452e-06, "loss": 0.1839, "step": 11614 }, { "epoch": 0.58, "learning_rate": 7.911498286411924e-06, "loss": 0.1909, "step": 11615 }, { "epoch": 0.58, "learning_rate": 7.909916766755424e-06, "loss": 0.4232, "step": 11616 }, { "epoch": 0.58, "learning_rate": 7.908335301762312e-06, "loss": 0.2901, "step": 11617 }, { "epoch": 0.58, "learning_rate": 7.906753891473947e-06, "loss": 0.3309, "step": 11618 }, { "epoch": 0.58, "learning_rate": 7.905172535931697e-06, "loss": 0.213, "step": 11619 }, { "epoch": 0.58, "learning_rate": 7.903591235176914e-06, "loss": 0.1918, "step": 11620 }, { "epoch": 0.58, "learning_rate": 7.902009989250958e-06, "loss": 0.3372, "step": 11621 }, { "epoch": 0.58, "learning_rate": 7.900428798195184e-06, "loss": 0.5074, "step": 11622 }, { "epoch": 0.58, "learning_rate": 7.898847662050938e-06, "loss": 0.2464, "step": 11623 }, { "epoch": 0.58, "learning_rate": 7.897266580859584e-06, "loss": 0.3265, "step": 11624 }, { "epoch": 0.58, "learning_rate": 7.89568555466247e-06, "loss": 0.2278, "step": 11625 }, { "epoch": 0.58, "learning_rate": 7.894104583500943e-06, "loss": 0.4415, "step": 11626 }, { "epoch": 0.58, "learning_rate": 7.89252366741635e-06, "loss": 0.3129, "step": 11627 }, { "epoch": 0.58, "learning_rate": 7.890942806450044e-06, "loss": 0.4685, "step": 11628 }, { "epoch": 0.58, "learning_rate": 7.889362000643364e-06, "loss": 0.392, "step": 11629 }, { "epoch": 0.58, "learning_rate": 7.887781250037659e-06, "loss": 0.3099, "step": 11630 }, { "epoch": 0.58, "learning_rate": 7.886200554674268e-06, "loss": 0.5054, "step": 11631 }, { "epoch": 0.58, "learning_rate": 7.884619914594527e-06, "loss": 0.602, "step": 11632 }, { "epoch": 0.58, "learning_rate": 7.883039329839787e-06, "loss": 0.4226, "step": 11633 }, { "epoch": 0.58, "learning_rate": 7.881458800451382e-06, "loss": 0.1463, "step": 11634 }, { "epoch": 0.58, "learning_rate": 7.879878326470645e-06, "loss": 0.2398, "step": 11635 }, { "epoch": 0.58, "learning_rate": 7.878297907938916e-06, "loss": 0.5802, "step": 11636 }, { "epoch": 0.58, "learning_rate": 7.876717544897522e-06, "loss": 0.409, "step": 11637 }, { "epoch": 0.58, "learning_rate": 7.875137237387802e-06, "loss": 0.3431, "step": 11638 }, { "epoch": 0.58, "learning_rate": 7.873556985451085e-06, "loss": 0.2059, "step": 11639 }, { "epoch": 0.58, "learning_rate": 7.8719767891287e-06, "loss": 0.4417, "step": 11640 }, { "epoch": 0.58, "learning_rate": 7.87039664846197e-06, "loss": 0.2871, "step": 11641 }, { "epoch": 0.58, "learning_rate": 7.868816563492231e-06, "loss": 0.5073, "step": 11642 }, { "epoch": 0.58, "learning_rate": 7.867236534260805e-06, "loss": 0.4179, "step": 11643 }, { "epoch": 0.58, "learning_rate": 7.865656560809011e-06, "loss": 0.7034, "step": 11644 }, { "epoch": 0.58, "learning_rate": 7.864076643178178e-06, "loss": 0.3779, "step": 11645 }, { "epoch": 0.58, "learning_rate": 7.862496781409616e-06, "loss": 0.3077, "step": 11646 }, { "epoch": 0.58, "learning_rate": 7.860916975544654e-06, "loss": 0.1208, "step": 11647 }, { "epoch": 0.58, "learning_rate": 7.859337225624609e-06, "loss": 0.393, "step": 11648 }, { "epoch": 0.58, "learning_rate": 7.857757531690796e-06, "loss": 0.399, "step": 11649 }, { "epoch": 0.58, "learning_rate": 7.856177893784524e-06, "loss": 0.404, "step": 11650 }, { "epoch": 0.58, "learning_rate": 7.854598311947115e-06, "loss": 0.4323, "step": 11651 }, { "epoch": 0.58, "learning_rate": 7.853018786219876e-06, "loss": 0.4941, "step": 11652 }, { "epoch": 0.58, "learning_rate": 7.85143931664412e-06, "loss": 0.2211, "step": 11653 }, { "epoch": 0.58, "learning_rate": 7.849859903261155e-06, "loss": 0.4875, "step": 11654 }, { "epoch": 0.58, "learning_rate": 7.848280546112283e-06, "loss": 0.5151, "step": 11655 }, { "epoch": 0.58, "learning_rate": 7.846701245238818e-06, "loss": 0.5026, "step": 11656 }, { "epoch": 0.58, "learning_rate": 7.845122000682064e-06, "loss": 0.4489, "step": 11657 }, { "epoch": 0.58, "learning_rate": 7.84354281248332e-06, "loss": 0.4584, "step": 11658 }, { "epoch": 0.58, "learning_rate": 7.841963680683887e-06, "loss": 0.2124, "step": 11659 }, { "epoch": 0.58, "learning_rate": 7.840384605325069e-06, "loss": 0.2517, "step": 11660 }, { "epoch": 0.58, "learning_rate": 7.838805586448164e-06, "loss": 0.4172, "step": 11661 }, { "epoch": 0.58, "learning_rate": 7.837226624094468e-06, "loss": 0.3374, "step": 11662 }, { "epoch": 0.58, "learning_rate": 7.835647718305278e-06, "loss": 0.4018, "step": 11663 }, { "epoch": 0.58, "learning_rate": 7.834068869121885e-06, "loss": 0.4894, "step": 11664 }, { "epoch": 0.58, "learning_rate": 7.832490076585584e-06, "loss": 0.4298, "step": 11665 }, { "epoch": 0.58, "learning_rate": 7.830911340737668e-06, "loss": 0.2887, "step": 11666 }, { "epoch": 0.58, "learning_rate": 7.829332661619425e-06, "loss": 0.3402, "step": 11667 }, { "epoch": 0.58, "learning_rate": 7.827754039272138e-06, "loss": 0.2463, "step": 11668 }, { "epoch": 0.58, "learning_rate": 7.826175473737102e-06, "loss": 0.307, "step": 11669 }, { "epoch": 0.58, "learning_rate": 7.824596965055603e-06, "loss": 0.3421, "step": 11670 }, { "epoch": 0.58, "learning_rate": 7.823018513268919e-06, "loss": 0.4207, "step": 11671 }, { "epoch": 0.58, "learning_rate": 7.821440118418334e-06, "loss": 0.4709, "step": 11672 }, { "epoch": 0.58, "learning_rate": 7.819861780545128e-06, "loss": 0.2467, "step": 11673 }, { "epoch": 0.58, "learning_rate": 7.818283499690583e-06, "loss": 0.4831, "step": 11674 }, { "epoch": 0.58, "learning_rate": 7.816705275895976e-06, "loss": 0.4815, "step": 11675 }, { "epoch": 0.58, "learning_rate": 7.815127109202582e-06, "loss": 0.3589, "step": 11676 }, { "epoch": 0.58, "learning_rate": 7.813548999651676e-06, "loss": 0.4324, "step": 11677 }, { "epoch": 0.58, "learning_rate": 7.811970947284533e-06, "loss": 0.3023, "step": 11678 }, { "epoch": 0.58, "learning_rate": 7.810392952142423e-06, "loss": 0.1583, "step": 11679 }, { "epoch": 0.58, "learning_rate": 7.808815014266622e-06, "loss": 0.5274, "step": 11680 }, { "epoch": 0.58, "learning_rate": 7.807237133698389e-06, "loss": 0.4115, "step": 11681 }, { "epoch": 0.58, "learning_rate": 7.805659310478997e-06, "loss": 0.2986, "step": 11682 }, { "epoch": 0.58, "learning_rate": 7.804081544649714e-06, "loss": 0.2487, "step": 11683 }, { "epoch": 0.58, "learning_rate": 7.802503836251802e-06, "loss": 0.2993, "step": 11684 }, { "epoch": 0.58, "learning_rate": 7.800926185326524e-06, "loss": 0.4852, "step": 11685 }, { "epoch": 0.58, "learning_rate": 7.799348591915141e-06, "loss": 0.4226, "step": 11686 }, { "epoch": 0.58, "learning_rate": 7.797771056058911e-06, "loss": 0.3472, "step": 11687 }, { "epoch": 0.58, "learning_rate": 7.796193577799098e-06, "loss": 0.5258, "step": 11688 }, { "epoch": 0.58, "learning_rate": 7.794616157176955e-06, "loss": 0.5032, "step": 11689 }, { "epoch": 0.58, "learning_rate": 7.793038794233738e-06, "loss": 0.3101, "step": 11690 }, { "epoch": 0.58, "learning_rate": 7.791461489010696e-06, "loss": 0.1857, "step": 11691 }, { "epoch": 0.58, "learning_rate": 7.789884241549091e-06, "loss": 0.1846, "step": 11692 }, { "epoch": 0.58, "learning_rate": 7.788307051890169e-06, "loss": 0.2943, "step": 11693 }, { "epoch": 0.58, "learning_rate": 7.786729920075177e-06, "loss": 0.3299, "step": 11694 }, { "epoch": 0.58, "learning_rate": 7.785152846145368e-06, "loss": 0.2385, "step": 11695 }, { "epoch": 0.58, "learning_rate": 7.78357583014198e-06, "loss": 0.2862, "step": 11696 }, { "epoch": 0.58, "learning_rate": 7.781998872106268e-06, "loss": 0.2852, "step": 11697 }, { "epoch": 0.58, "learning_rate": 7.78042197207947e-06, "loss": 0.473, "step": 11698 }, { "epoch": 0.58, "learning_rate": 7.778845130102828e-06, "loss": 0.5983, "step": 11699 }, { "epoch": 0.58, "learning_rate": 7.77726834621758e-06, "loss": 0.3327, "step": 11700 }, { "epoch": 0.58, "learning_rate": 7.775691620464969e-06, "loss": 0.2795, "step": 11701 }, { "epoch": 0.58, "learning_rate": 7.774114952886232e-06, "loss": 0.4056, "step": 11702 }, { "epoch": 0.58, "learning_rate": 7.772538343522602e-06, "loss": 0.4609, "step": 11703 }, { "epoch": 0.58, "learning_rate": 7.770961792415313e-06, "loss": 0.382, "step": 11704 }, { "epoch": 0.58, "learning_rate": 7.769385299605597e-06, "loss": 0.471, "step": 11705 }, { "epoch": 0.58, "learning_rate": 7.767808865134689e-06, "loss": 0.4518, "step": 11706 }, { "epoch": 0.58, "learning_rate": 7.766232489043818e-06, "loss": 0.2953, "step": 11707 }, { "epoch": 0.58, "learning_rate": 7.76465617137421e-06, "loss": 0.2584, "step": 11708 }, { "epoch": 0.58, "learning_rate": 7.763079912167092e-06, "loss": 0.3623, "step": 11709 }, { "epoch": 0.58, "learning_rate": 7.761503711463688e-06, "loss": 0.2891, "step": 11710 }, { "epoch": 0.58, "learning_rate": 7.759927569305222e-06, "loss": 0.1334, "step": 11711 }, { "epoch": 0.58, "learning_rate": 7.758351485732918e-06, "loss": 0.2522, "step": 11712 }, { "epoch": 0.58, "learning_rate": 7.756775460787993e-06, "loss": 0.4952, "step": 11713 }, { "epoch": 0.58, "learning_rate": 7.755199494511667e-06, "loss": 0.2483, "step": 11714 }, { "epoch": 0.58, "learning_rate": 7.75362358694516e-06, "loss": 0.2305, "step": 11715 }, { "epoch": 0.59, "learning_rate": 7.752047738129687e-06, "loss": 0.4028, "step": 11716 }, { "epoch": 0.59, "learning_rate": 7.750471948106462e-06, "loss": 0.3916, "step": 11717 }, { "epoch": 0.59, "learning_rate": 7.748896216916695e-06, "loss": 0.1622, "step": 11718 }, { "epoch": 0.59, "learning_rate": 7.747320544601596e-06, "loss": 0.4281, "step": 11719 }, { "epoch": 0.59, "learning_rate": 7.74574493120238e-06, "loss": 0.3704, "step": 11720 }, { "epoch": 0.59, "learning_rate": 7.744169376760255e-06, "loss": 0.1997, "step": 11721 }, { "epoch": 0.59, "learning_rate": 7.742593881316423e-06, "loss": 0.4237, "step": 11722 }, { "epoch": 0.59, "learning_rate": 7.74101844491209e-06, "loss": 0.3135, "step": 11723 }, { "epoch": 0.59, "learning_rate": 7.739443067588464e-06, "loss": 0.2913, "step": 11724 }, { "epoch": 0.59, "learning_rate": 7.737867749386744e-06, "loss": 0.3208, "step": 11725 }, { "epoch": 0.59, "learning_rate": 7.73629249034813e-06, "loss": 0.3031, "step": 11726 }, { "epoch": 0.59, "learning_rate": 7.734717290513821e-06, "loss": 0.5423, "step": 11727 }, { "epoch": 0.59, "learning_rate": 7.733142149925008e-06, "loss": 0.6328, "step": 11728 }, { "epoch": 0.59, "learning_rate": 7.731567068622898e-06, "loss": 0.484, "step": 11729 }, { "epoch": 0.59, "learning_rate": 7.729992046648681e-06, "loss": 0.2846, "step": 11730 }, { "epoch": 0.59, "learning_rate": 7.72841708404355e-06, "loss": 0.4841, "step": 11731 }, { "epoch": 0.59, "learning_rate": 7.72684218084869e-06, "loss": 0.3689, "step": 11732 }, { "epoch": 0.59, "learning_rate": 7.725267337105298e-06, "loss": 0.4495, "step": 11733 }, { "epoch": 0.59, "learning_rate": 7.723692552854561e-06, "loss": 0.2615, "step": 11734 }, { "epoch": 0.59, "learning_rate": 7.722117828137663e-06, "loss": 0.2497, "step": 11735 }, { "epoch": 0.59, "learning_rate": 7.720543162995791e-06, "loss": 0.4132, "step": 11736 }, { "epoch": 0.59, "learning_rate": 7.718968557470124e-06, "loss": 0.3284, "step": 11737 }, { "epoch": 0.59, "learning_rate": 7.717394011601848e-06, "loss": 0.4047, "step": 11738 }, { "epoch": 0.59, "learning_rate": 7.715819525432144e-06, "loss": 0.2693, "step": 11739 }, { "epoch": 0.59, "learning_rate": 7.714245099002186e-06, "loss": 0.2613, "step": 11740 }, { "epoch": 0.59, "learning_rate": 7.712670732353156e-06, "loss": 0.3797, "step": 11741 }, { "epoch": 0.59, "learning_rate": 7.711096425526222e-06, "loss": 0.264, "step": 11742 }, { "epoch": 0.59, "learning_rate": 7.709522178562568e-06, "loss": 0.1731, "step": 11743 }, { "epoch": 0.59, "learning_rate": 7.707947991503359e-06, "loss": 0.2571, "step": 11744 }, { "epoch": 0.59, "learning_rate": 7.706373864389771e-06, "loss": 0.3044, "step": 11745 }, { "epoch": 0.59, "learning_rate": 7.704799797262967e-06, "loss": 0.32, "step": 11746 }, { "epoch": 0.59, "learning_rate": 7.70322579016412e-06, "loss": 0.4421, "step": 11747 }, { "epoch": 0.59, "learning_rate": 7.701651843134396e-06, "loss": 0.3671, "step": 11748 }, { "epoch": 0.59, "learning_rate": 7.700077956214954e-06, "loss": 0.2082, "step": 11749 }, { "epoch": 0.59, "learning_rate": 7.698504129446963e-06, "loss": 0.4767, "step": 11750 }, { "epoch": 0.59, "learning_rate": 7.69693036287158e-06, "loss": 0.2365, "step": 11751 }, { "epoch": 0.59, "learning_rate": 7.69535665652997e-06, "loss": 0.2674, "step": 11752 }, { "epoch": 0.59, "learning_rate": 7.693783010463287e-06, "loss": 0.374, "step": 11753 }, { "epoch": 0.59, "learning_rate": 7.69220942471269e-06, "loss": 0.1771, "step": 11754 }, { "epoch": 0.59, "learning_rate": 7.690635899319327e-06, "loss": 0.1233, "step": 11755 }, { "epoch": 0.59, "learning_rate": 7.689062434324362e-06, "loss": 0.4496, "step": 11756 }, { "epoch": 0.59, "learning_rate": 7.687489029768941e-06, "loss": 0.5667, "step": 11757 }, { "epoch": 0.59, "learning_rate": 7.68591568569422e-06, "loss": 0.2504, "step": 11758 }, { "epoch": 0.59, "learning_rate": 7.68434240214134e-06, "loss": 0.401, "step": 11759 }, { "epoch": 0.59, "learning_rate": 7.682769179151448e-06, "loss": 0.2714, "step": 11760 }, { "epoch": 0.59, "learning_rate": 7.681196016765697e-06, "loss": 0.2468, "step": 11761 }, { "epoch": 0.59, "learning_rate": 7.679622915025228e-06, "loss": 0.2944, "step": 11762 }, { "epoch": 0.59, "learning_rate": 7.678049873971183e-06, "loss": 0.2276, "step": 11763 }, { "epoch": 0.59, "learning_rate": 7.676476893644695e-06, "loss": 0.3049, "step": 11764 }, { "epoch": 0.59, "learning_rate": 7.674903974086917e-06, "loss": 0.3687, "step": 11765 }, { "epoch": 0.59, "learning_rate": 7.67333111533898e-06, "loss": 0.545, "step": 11766 }, { "epoch": 0.59, "learning_rate": 7.67175831744202e-06, "loss": 0.3639, "step": 11767 }, { "epoch": 0.59, "learning_rate": 7.670185580437171e-06, "loss": 0.5911, "step": 11768 }, { "epoch": 0.59, "learning_rate": 7.668612904365562e-06, "loss": 0.4035, "step": 11769 }, { "epoch": 0.59, "learning_rate": 7.667040289268335e-06, "loss": 0.3023, "step": 11770 }, { "epoch": 0.59, "learning_rate": 7.665467735186613e-06, "loss": 0.3919, "step": 11771 }, { "epoch": 0.59, "learning_rate": 7.663895242161525e-06, "loss": 0.4505, "step": 11772 }, { "epoch": 0.59, "learning_rate": 7.662322810234197e-06, "loss": 0.395, "step": 11773 }, { "epoch": 0.59, "learning_rate": 7.660750439445753e-06, "loss": 0.2625, "step": 11774 }, { "epoch": 0.59, "learning_rate": 7.659178129837318e-06, "loss": 0.6034, "step": 11775 }, { "epoch": 0.59, "learning_rate": 7.657605881450011e-06, "loss": 0.2657, "step": 11776 }, { "epoch": 0.59, "learning_rate": 7.656033694324958e-06, "loss": 0.3313, "step": 11777 }, { "epoch": 0.59, "learning_rate": 7.654461568503269e-06, "loss": 0.5528, "step": 11778 }, { "epoch": 0.59, "learning_rate": 7.652889504026068e-06, "loss": 0.2365, "step": 11779 }, { "epoch": 0.59, "learning_rate": 7.651317500934469e-06, "loss": 0.2271, "step": 11780 }, { "epoch": 0.59, "learning_rate": 7.649745559269584e-06, "loss": 0.5145, "step": 11781 }, { "epoch": 0.59, "learning_rate": 7.648173679072526e-06, "loss": 0.2601, "step": 11782 }, { "epoch": 0.59, "learning_rate": 7.646601860384402e-06, "loss": 0.5301, "step": 11783 }, { "epoch": 0.59, "learning_rate": 7.645030103246327e-06, "loss": 0.4442, "step": 11784 }, { "epoch": 0.59, "learning_rate": 7.643458407699403e-06, "loss": 0.2612, "step": 11785 }, { "epoch": 0.59, "learning_rate": 7.641886773784742e-06, "loss": 0.3871, "step": 11786 }, { "epoch": 0.59, "learning_rate": 7.640315201543438e-06, "loss": 0.4864, "step": 11787 }, { "epoch": 0.59, "learning_rate": 7.638743691016603e-06, "loss": 0.406, "step": 11788 }, { "epoch": 0.59, "learning_rate": 7.637172242245335e-06, "loss": 0.1782, "step": 11789 }, { "epoch": 0.59, "learning_rate": 7.63560085527073e-06, "loss": 0.3372, "step": 11790 }, { "epoch": 0.59, "learning_rate": 7.63402953013389e-06, "loss": 0.1416, "step": 11791 }, { "epoch": 0.59, "learning_rate": 7.632458266875904e-06, "loss": 0.1999, "step": 11792 }, { "epoch": 0.59, "learning_rate": 7.630887065537874e-06, "loss": 0.2744, "step": 11793 }, { "epoch": 0.59, "learning_rate": 7.629315926160892e-06, "loss": 0.6253, "step": 11794 }, { "epoch": 0.59, "learning_rate": 7.627744848786045e-06, "loss": 0.4607, "step": 11795 }, { "epoch": 0.59, "learning_rate": 7.626173833454423e-06, "loss": 0.2771, "step": 11796 }, { "epoch": 0.59, "learning_rate": 7.624602880207118e-06, "loss": 0.3543, "step": 11797 }, { "epoch": 0.59, "learning_rate": 7.623031989085214e-06, "loss": 0.4453, "step": 11798 }, { "epoch": 0.59, "learning_rate": 7.621461160129793e-06, "loss": 0.2612, "step": 11799 }, { "epoch": 0.59, "learning_rate": 7.619890393381941e-06, "loss": 0.3051, "step": 11800 }, { "epoch": 0.59, "learning_rate": 7.6183196888827335e-06, "loss": 0.5059, "step": 11801 }, { "epoch": 0.59, "learning_rate": 7.6167490466732604e-06, "loss": 0.3148, "step": 11802 }, { "epoch": 0.59, "learning_rate": 7.615178466794594e-06, "loss": 0.6638, "step": 11803 }, { "epoch": 0.59, "learning_rate": 7.613607949287811e-06, "loss": 0.4176, "step": 11804 }, { "epoch": 0.59, "learning_rate": 7.612037494193982e-06, "loss": 0.2763, "step": 11805 }, { "epoch": 0.59, "learning_rate": 7.610467101554191e-06, "loss": 0.4351, "step": 11806 }, { "epoch": 0.59, "learning_rate": 7.608896771409501e-06, "loss": 0.2136, "step": 11807 }, { "epoch": 0.59, "learning_rate": 7.607326503800985e-06, "loss": 0.2518, "step": 11808 }, { "epoch": 0.59, "learning_rate": 7.60575629876971e-06, "loss": 0.4007, "step": 11809 }, { "epoch": 0.59, "learning_rate": 7.604186156356741e-06, "loss": 0.3166, "step": 11810 }, { "epoch": 0.59, "learning_rate": 7.602616076603148e-06, "loss": 0.4464, "step": 11811 }, { "epoch": 0.59, "learning_rate": 7.601046059549991e-06, "loss": 0.353, "step": 11812 }, { "epoch": 0.59, "learning_rate": 7.599476105238333e-06, "loss": 0.512, "step": 11813 }, { "epoch": 0.59, "learning_rate": 7.597906213709232e-06, "loss": 0.2633, "step": 11814 }, { "epoch": 0.59, "learning_rate": 7.5963363850037445e-06, "loss": 0.254, "step": 11815 }, { "epoch": 0.59, "learning_rate": 7.594766619162934e-06, "loss": 0.2868, "step": 11816 }, { "epoch": 0.59, "learning_rate": 7.5931969162278525e-06, "loss": 0.2167, "step": 11817 }, { "epoch": 0.59, "learning_rate": 7.591627276239553e-06, "loss": 0.4507, "step": 11818 }, { "epoch": 0.59, "learning_rate": 7.590057699239087e-06, "loss": 0.4321, "step": 11819 }, { "epoch": 0.59, "learning_rate": 7.5884881852675075e-06, "loss": 0.3466, "step": 11820 }, { "epoch": 0.59, "learning_rate": 7.586918734365858e-06, "loss": 0.2923, "step": 11821 }, { "epoch": 0.59, "learning_rate": 7.5853493465751905e-06, "loss": 0.3222, "step": 11822 }, { "epoch": 0.59, "learning_rate": 7.583780021936549e-06, "loss": 0.4308, "step": 11823 }, { "epoch": 0.59, "learning_rate": 7.582210760490973e-06, "loss": 0.4912, "step": 11824 }, { "epoch": 0.59, "learning_rate": 7.58064156227951e-06, "loss": 0.4481, "step": 11825 }, { "epoch": 0.59, "learning_rate": 7.579072427343199e-06, "loss": 0.2773, "step": 11826 }, { "epoch": 0.59, "learning_rate": 7.577503355723077e-06, "loss": 0.3158, "step": 11827 }, { "epoch": 0.59, "learning_rate": 7.5759343474601785e-06, "loss": 0.3262, "step": 11828 }, { "epoch": 0.59, "learning_rate": 7.574365402595546e-06, "loss": 0.2439, "step": 11829 }, { "epoch": 0.59, "learning_rate": 7.572796521170209e-06, "loss": 0.41, "step": 11830 }, { "epoch": 0.59, "learning_rate": 7.571227703225201e-06, "loss": 0.3702, "step": 11831 }, { "epoch": 0.59, "learning_rate": 7.56965894880155e-06, "loss": 0.4781, "step": 11832 }, { "epoch": 0.59, "learning_rate": 7.568090257940286e-06, "loss": 0.3663, "step": 11833 }, { "epoch": 0.59, "learning_rate": 7.566521630682436e-06, "loss": 0.3265, "step": 11834 }, { "epoch": 0.59, "learning_rate": 7.564953067069026e-06, "loss": 0.4358, "step": 11835 }, { "epoch": 0.59, "learning_rate": 7.56338456714108e-06, "loss": 0.4796, "step": 11836 }, { "epoch": 0.59, "learning_rate": 7.5618161309396145e-06, "loss": 0.3043, "step": 11837 }, { "epoch": 0.59, "learning_rate": 7.560247758505659e-06, "loss": 0.3315, "step": 11838 }, { "epoch": 0.59, "learning_rate": 7.558679449880228e-06, "loss": 0.389, "step": 11839 }, { "epoch": 0.59, "learning_rate": 7.5571112051043374e-06, "loss": 0.3068, "step": 11840 }, { "epoch": 0.59, "learning_rate": 7.555543024219004e-06, "loss": 0.176, "step": 11841 }, { "epoch": 0.59, "learning_rate": 7.553974907265238e-06, "loss": 0.3989, "step": 11842 }, { "epoch": 0.59, "learning_rate": 7.552406854284058e-06, "loss": 0.3906, "step": 11843 }, { "epoch": 0.59, "learning_rate": 7.55083886531647e-06, "loss": 0.5425, "step": 11844 }, { "epoch": 0.59, "learning_rate": 7.549270940403486e-06, "loss": 0.3799, "step": 11845 }, { "epoch": 0.59, "learning_rate": 7.547703079586108e-06, "loss": 0.4478, "step": 11846 }, { "epoch": 0.59, "learning_rate": 7.546135282905344e-06, "loss": 0.2407, "step": 11847 }, { "epoch": 0.59, "learning_rate": 7.544567550402198e-06, "loss": 0.3872, "step": 11848 }, { "epoch": 0.59, "learning_rate": 7.542999882117674e-06, "loss": 0.2982, "step": 11849 }, { "epoch": 0.59, "learning_rate": 7.541432278092769e-06, "loss": 0.4032, "step": 11850 }, { "epoch": 0.59, "learning_rate": 7.539864738368477e-06, "loss": 0.3002, "step": 11851 }, { "epoch": 0.59, "learning_rate": 7.5382972629858065e-06, "loss": 0.0782, "step": 11852 }, { "epoch": 0.59, "learning_rate": 7.536729851985747e-06, "loss": 0.2095, "step": 11853 }, { "epoch": 0.59, "learning_rate": 7.535162505409291e-06, "loss": 0.2801, "step": 11854 }, { "epoch": 0.59, "learning_rate": 7.533595223297431e-06, "loss": 0.401, "step": 11855 }, { "epoch": 0.59, "learning_rate": 7.532028005691155e-06, "loss": 0.413, "step": 11856 }, { "epoch": 0.59, "learning_rate": 7.530460852631456e-06, "loss": 0.269, "step": 11857 }, { "epoch": 0.59, "learning_rate": 7.52889376415932e-06, "loss": 0.3313, "step": 11858 }, { "epoch": 0.59, "learning_rate": 7.52732674031573e-06, "loss": 0.3353, "step": 11859 }, { "epoch": 0.59, "learning_rate": 7.52575978114167e-06, "loss": 0.404, "step": 11860 }, { "epoch": 0.59, "learning_rate": 7.524192886678123e-06, "loss": 0.4222, "step": 11861 }, { "epoch": 0.59, "learning_rate": 7.522626056966069e-06, "loss": 0.3061, "step": 11862 }, { "epoch": 0.59, "learning_rate": 7.521059292046486e-06, "loss": 0.3498, "step": 11863 }, { "epoch": 0.59, "learning_rate": 7.519492591960351e-06, "loss": 0.3943, "step": 11864 }, { "epoch": 0.59, "learning_rate": 7.517925956748633e-06, "loss": 0.0915, "step": 11865 }, { "epoch": 0.59, "learning_rate": 7.516359386452314e-06, "loss": 0.2383, "step": 11866 }, { "epoch": 0.59, "learning_rate": 7.514792881112365e-06, "loss": 0.5472, "step": 11867 }, { "epoch": 0.59, "learning_rate": 7.513226440769751e-06, "loss": 0.3888, "step": 11868 }, { "epoch": 0.59, "learning_rate": 7.51166006546544e-06, "loss": 0.2849, "step": 11869 }, { "epoch": 0.59, "learning_rate": 7.510093755240404e-06, "loss": 0.2548, "step": 11870 }, { "epoch": 0.59, "learning_rate": 7.508527510135605e-06, "loss": 0.425, "step": 11871 }, { "epoch": 0.59, "learning_rate": 7.506961330192005e-06, "loss": 0.5719, "step": 11872 }, { "epoch": 0.59, "learning_rate": 7.505395215450566e-06, "loss": 0.3453, "step": 11873 }, { "epoch": 0.59, "learning_rate": 7.5038291659522445e-06, "loss": 0.3903, "step": 11874 }, { "epoch": 0.59, "learning_rate": 7.502263181738006e-06, "loss": 0.4747, "step": 11875 }, { "epoch": 0.59, "learning_rate": 7.500697262848802e-06, "loss": 0.3152, "step": 11876 }, { "epoch": 0.59, "learning_rate": 7.4991314093255866e-06, "loss": 0.4794, "step": 11877 }, { "epoch": 0.59, "learning_rate": 7.497565621209315e-06, "loss": 0.1893, "step": 11878 }, { "epoch": 0.59, "learning_rate": 7.495999898540931e-06, "loss": 0.4046, "step": 11879 }, { "epoch": 0.59, "learning_rate": 7.4944342413613966e-06, "loss": 0.3891, "step": 11880 }, { "epoch": 0.59, "learning_rate": 7.492868649711651e-06, "loss": 0.2987, "step": 11881 }, { "epoch": 0.59, "learning_rate": 7.491303123632643e-06, "loss": 0.2263, "step": 11882 }, { "epoch": 0.59, "learning_rate": 7.489737663165314e-06, "loss": 0.3023, "step": 11883 }, { "epoch": 0.59, "learning_rate": 7.488172268350611e-06, "loss": 0.3381, "step": 11884 }, { "epoch": 0.59, "learning_rate": 7.486606939229472e-06, "loss": 0.3166, "step": 11885 }, { "epoch": 0.59, "learning_rate": 7.485041675842837e-06, "loss": 0.3386, "step": 11886 }, { "epoch": 0.59, "learning_rate": 7.4834764782316415e-06, "loss": 0.2797, "step": 11887 }, { "epoch": 0.59, "learning_rate": 7.481911346436819e-06, "loss": 0.3562, "step": 11888 }, { "epoch": 0.59, "learning_rate": 7.480346280499312e-06, "loss": 0.2996, "step": 11889 }, { "epoch": 0.59, "learning_rate": 7.478781280460048e-06, "loss": 0.3812, "step": 11890 }, { "epoch": 0.59, "learning_rate": 7.477216346359957e-06, "loss": 0.2867, "step": 11891 }, { "epoch": 0.59, "learning_rate": 7.475651478239966e-06, "loss": 0.426, "step": 11892 }, { "epoch": 0.59, "learning_rate": 7.4740866761410036e-06, "loss": 0.624, "step": 11893 }, { "epoch": 0.59, "learning_rate": 7.472521940103999e-06, "loss": 0.4435, "step": 11894 }, { "epoch": 0.59, "learning_rate": 7.470957270169872e-06, "loss": 0.5471, "step": 11895 }, { "epoch": 0.59, "learning_rate": 7.4693926663795464e-06, "loss": 0.1909, "step": 11896 }, { "epoch": 0.59, "learning_rate": 7.467828128773938e-06, "loss": 0.53, "step": 11897 }, { "epoch": 0.59, "learning_rate": 7.466263657393971e-06, "loss": 0.4563, "step": 11898 }, { "epoch": 0.59, "learning_rate": 7.464699252280559e-06, "loss": 0.3196, "step": 11899 }, { "epoch": 0.59, "learning_rate": 7.463134913474617e-06, "loss": 0.1676, "step": 11900 }, { "epoch": 0.59, "learning_rate": 7.461570641017055e-06, "loss": 0.4053, "step": 11901 }, { "epoch": 0.59, "learning_rate": 7.460006434948792e-06, "loss": 0.3859, "step": 11902 }, { "epoch": 0.59, "learning_rate": 7.4584422953107334e-06, "loss": 0.2799, "step": 11903 }, { "epoch": 0.59, "learning_rate": 7.456878222143789e-06, "loss": 0.3959, "step": 11904 }, { "epoch": 0.59, "learning_rate": 7.455314215488863e-06, "loss": 0.1876, "step": 11905 }, { "epoch": 0.59, "learning_rate": 7.453750275386859e-06, "loss": 0.3399, "step": 11906 }, { "epoch": 0.59, "learning_rate": 7.452186401878684e-06, "loss": 0.488, "step": 11907 }, { "epoch": 0.59, "learning_rate": 7.450622595005236e-06, "loss": 0.2819, "step": 11908 }, { "epoch": 0.59, "learning_rate": 7.449058854807415e-06, "loss": 0.2832, "step": 11909 }, { "epoch": 0.59, "learning_rate": 7.447495181326118e-06, "loss": 0.3646, "step": 11910 }, { "epoch": 0.59, "learning_rate": 7.4459315746022384e-06, "loss": 0.1432, "step": 11911 }, { "epoch": 0.59, "learning_rate": 7.444368034676677e-06, "loss": 0.1716, "step": 11912 }, { "epoch": 0.59, "learning_rate": 7.442804561590322e-06, "loss": 0.5288, "step": 11913 }, { "epoch": 0.59, "learning_rate": 7.4412411553840645e-06, "loss": 0.4549, "step": 11914 }, { "epoch": 0.59, "learning_rate": 7.4396778160987896e-06, "loss": 0.2226, "step": 11915 }, { "epoch": 0.59, "learning_rate": 7.4381145437753925e-06, "loss": 0.235, "step": 11916 }, { "epoch": 0.6, "learning_rate": 7.436551338454752e-06, "loss": 0.4203, "step": 11917 }, { "epoch": 0.6, "learning_rate": 7.434988200177755e-06, "loss": 0.2302, "step": 11918 }, { "epoch": 0.6, "learning_rate": 7.433425128985282e-06, "loss": 0.4801, "step": 11919 }, { "epoch": 0.6, "learning_rate": 7.431862124918211e-06, "loss": 0.3162, "step": 11920 }, { "epoch": 0.6, "learning_rate": 7.430299188017425e-06, "loss": 0.2953, "step": 11921 }, { "epoch": 0.6, "learning_rate": 7.428736318323797e-06, "loss": 0.4301, "step": 11922 }, { "epoch": 0.6, "learning_rate": 7.427173515878203e-06, "loss": 0.3502, "step": 11923 }, { "epoch": 0.6, "learning_rate": 7.425610780721513e-06, "loss": 0.279, "step": 11924 }, { "epoch": 0.6, "learning_rate": 7.424048112894604e-06, "loss": 0.373, "step": 11925 }, { "epoch": 0.6, "learning_rate": 7.422485512438344e-06, "loss": 0.3137, "step": 11926 }, { "epoch": 0.6, "learning_rate": 7.420922979393599e-06, "loss": 0.4127, "step": 11927 }, { "epoch": 0.6, "learning_rate": 7.419360513801235e-06, "loss": 0.3041, "step": 11928 }, { "epoch": 0.6, "learning_rate": 7.417798115702113e-06, "loss": 0.3357, "step": 11929 }, { "epoch": 0.6, "learning_rate": 7.416235785137102e-06, "loss": 0.3197, "step": 11930 }, { "epoch": 0.6, "learning_rate": 7.4146735221470625e-06, "loss": 0.3559, "step": 11931 }, { "epoch": 0.6, "learning_rate": 7.413111326772849e-06, "loss": 0.3651, "step": 11932 }, { "epoch": 0.6, "learning_rate": 7.411549199055319e-06, "loss": 0.4253, "step": 11933 }, { "epoch": 0.6, "learning_rate": 7.409987139035332e-06, "loss": 0.3568, "step": 11934 }, { "epoch": 0.6, "learning_rate": 7.408425146753739e-06, "loss": 0.3382, "step": 11935 }, { "epoch": 0.6, "learning_rate": 7.40686322225139e-06, "loss": 0.4312, "step": 11936 }, { "epoch": 0.6, "learning_rate": 7.40530136556914e-06, "loss": 0.4494, "step": 11937 }, { "epoch": 0.6, "learning_rate": 7.403739576747828e-06, "loss": 0.3663, "step": 11938 }, { "epoch": 0.6, "learning_rate": 7.402177855828312e-06, "loss": 0.3493, "step": 11939 }, { "epoch": 0.6, "learning_rate": 7.400616202851431e-06, "loss": 0.4151, "step": 11940 }, { "epoch": 0.6, "learning_rate": 7.399054617858028e-06, "loss": 0.4537, "step": 11941 }, { "epoch": 0.6, "learning_rate": 7.397493100888944e-06, "loss": 0.544, "step": 11942 }, { "epoch": 0.6, "learning_rate": 7.395931651985022e-06, "loss": 0.2647, "step": 11943 }, { "epoch": 0.6, "learning_rate": 7.394370271187095e-06, "loss": 0.2184, "step": 11944 }, { "epoch": 0.6, "learning_rate": 7.392808958536002e-06, "loss": 0.4091, "step": 11945 }, { "epoch": 0.6, "learning_rate": 7.391247714072576e-06, "loss": 0.34, "step": 11946 }, { "epoch": 0.6, "learning_rate": 7.389686537837645e-06, "loss": 0.2468, "step": 11947 }, { "epoch": 0.6, "learning_rate": 7.388125429872047e-06, "loss": 0.1782, "step": 11948 }, { "epoch": 0.6, "learning_rate": 7.386564390216609e-06, "loss": 0.3199, "step": 11949 }, { "epoch": 0.6, "learning_rate": 7.385003418912155e-06, "loss": 0.3616, "step": 11950 }, { "epoch": 0.6, "learning_rate": 7.383442515999511e-06, "loss": 0.4373, "step": 11951 }, { "epoch": 0.6, "learning_rate": 7.381881681519498e-06, "loss": 0.4518, "step": 11952 }, { "epoch": 0.6, "learning_rate": 7.380320915512944e-06, "loss": 0.3951, "step": 11953 }, { "epoch": 0.6, "learning_rate": 7.378760218020667e-06, "loss": 0.2695, "step": 11954 }, { "epoch": 0.6, "learning_rate": 7.377199589083481e-06, "loss": 0.3059, "step": 11955 }, { "epoch": 0.6, "learning_rate": 7.375639028742203e-06, "loss": 0.3827, "step": 11956 }, { "epoch": 0.6, "learning_rate": 7.374078537037652e-06, "loss": 0.3442, "step": 11957 }, { "epoch": 0.6, "learning_rate": 7.3725181140106375e-06, "loss": 0.3738, "step": 11958 }, { "epoch": 0.6, "learning_rate": 7.370957759701968e-06, "loss": 0.2922, "step": 11959 }, { "epoch": 0.6, "learning_rate": 7.3693974741524575e-06, "loss": 0.5582, "step": 11960 }, { "epoch": 0.6, "learning_rate": 7.3678372574029056e-06, "loss": 0.2937, "step": 11961 }, { "epoch": 0.6, "learning_rate": 7.366277109494127e-06, "loss": 0.3525, "step": 11962 }, { "epoch": 0.6, "learning_rate": 7.364717030466921e-06, "loss": 0.2172, "step": 11963 }, { "epoch": 0.6, "learning_rate": 7.363157020362089e-06, "loss": 0.2811, "step": 11964 }, { "epoch": 0.6, "learning_rate": 7.3615970792204285e-06, "loss": 0.4011, "step": 11965 }, { "epoch": 0.6, "learning_rate": 7.360037207082744e-06, "loss": 0.515, "step": 11966 }, { "epoch": 0.6, "learning_rate": 7.358477403989828e-06, "loss": 0.211, "step": 11967 }, { "epoch": 0.6, "learning_rate": 7.356917669982477e-06, "loss": 0.3845, "step": 11968 }, { "epoch": 0.6, "learning_rate": 7.355358005101481e-06, "loss": 0.4935, "step": 11969 }, { "epoch": 0.6, "learning_rate": 7.353798409387631e-06, "loss": 0.5332, "step": 11970 }, { "epoch": 0.6, "learning_rate": 7.352238882881721e-06, "loss": 0.3489, "step": 11971 }, { "epoch": 0.6, "learning_rate": 7.350679425624534e-06, "loss": 0.3696, "step": 11972 }, { "epoch": 0.6, "learning_rate": 7.349120037656857e-06, "loss": 0.7407, "step": 11973 }, { "epoch": 0.6, "learning_rate": 7.347560719019468e-06, "loss": 0.2813, "step": 11974 }, { "epoch": 0.6, "learning_rate": 7.346001469753161e-06, "loss": 0.2433, "step": 11975 }, { "epoch": 0.6, "learning_rate": 7.344442289898708e-06, "loss": 0.3217, "step": 11976 }, { "epoch": 0.6, "learning_rate": 7.342883179496889e-06, "loss": 0.3382, "step": 11977 }, { "epoch": 0.6, "learning_rate": 7.341324138588479e-06, "loss": 0.1607, "step": 11978 }, { "epoch": 0.6, "learning_rate": 7.3397651672142536e-06, "loss": 0.3049, "step": 11979 }, { "epoch": 0.6, "learning_rate": 7.3382062654149865e-06, "loss": 0.4126, "step": 11980 }, { "epoch": 0.6, "learning_rate": 7.33664743323145e-06, "loss": 0.1485, "step": 11981 }, { "epoch": 0.6, "learning_rate": 7.33508867070441e-06, "loss": 0.1785, "step": 11982 }, { "epoch": 0.6, "learning_rate": 7.333529977874636e-06, "loss": 0.3979, "step": 11983 }, { "epoch": 0.6, "learning_rate": 7.331971354782888e-06, "loss": 0.2087, "step": 11984 }, { "epoch": 0.6, "learning_rate": 7.330412801469939e-06, "loss": 0.305, "step": 11985 }, { "epoch": 0.6, "learning_rate": 7.328854317976548e-06, "loss": 0.4458, "step": 11986 }, { "epoch": 0.6, "learning_rate": 7.327295904343473e-06, "loss": 0.3095, "step": 11987 }, { "epoch": 0.6, "learning_rate": 7.325737560611469e-06, "loss": 0.3768, "step": 11988 }, { "epoch": 0.6, "learning_rate": 7.3241792868213e-06, "loss": 0.3143, "step": 11989 }, { "epoch": 0.6, "learning_rate": 7.322621083013718e-06, "loss": 0.455, "step": 11990 }, { "epoch": 0.6, "learning_rate": 7.321062949229474e-06, "loss": 0.5464, "step": 11991 }, { "epoch": 0.6, "learning_rate": 7.319504885509321e-06, "loss": 0.3585, "step": 11992 }, { "epoch": 0.6, "learning_rate": 7.317946891894003e-06, "loss": 0.3288, "step": 11993 }, { "epoch": 0.6, "learning_rate": 7.316388968424275e-06, "loss": 0.2775, "step": 11994 }, { "epoch": 0.6, "learning_rate": 7.314831115140877e-06, "loss": 0.1968, "step": 11995 }, { "epoch": 0.6, "learning_rate": 7.313273332084556e-06, "loss": 0.4834, "step": 11996 }, { "epoch": 0.6, "learning_rate": 7.311715619296047e-06, "loss": 0.302, "step": 11997 }, { "epoch": 0.6, "learning_rate": 7.310157976816099e-06, "loss": 0.4638, "step": 11998 }, { "epoch": 0.6, "learning_rate": 7.308600404685446e-06, "loss": 0.1981, "step": 11999 }, { "epoch": 0.6, "learning_rate": 7.307042902944824e-06, "loss": 0.2241, "step": 12000 }, { "epoch": 0.6, "learning_rate": 7.305485471634969e-06, "loss": 0.4321, "step": 12001 }, { "epoch": 0.6, "learning_rate": 7.3039281107966076e-06, "loss": 0.1717, "step": 12002 }, { "epoch": 0.6, "learning_rate": 7.302370820470479e-06, "loss": 0.2637, "step": 12003 }, { "epoch": 0.6, "learning_rate": 7.300813600697307e-06, "loss": 0.1887, "step": 12004 }, { "epoch": 0.6, "learning_rate": 7.299256451517822e-06, "loss": 0.4334, "step": 12005 }, { "epoch": 0.6, "learning_rate": 7.297699372972744e-06, "loss": 0.5358, "step": 12006 }, { "epoch": 0.6, "learning_rate": 7.296142365102803e-06, "loss": 0.4408, "step": 12007 }, { "epoch": 0.6, "learning_rate": 7.294585427948716e-06, "loss": 0.4474, "step": 12008 }, { "epoch": 0.6, "learning_rate": 7.293028561551204e-06, "loss": 0.3261, "step": 12009 }, { "epoch": 0.6, "learning_rate": 7.291471765950984e-06, "loss": 0.3054, "step": 12010 }, { "epoch": 0.6, "learning_rate": 7.289915041188769e-06, "loss": 0.3211, "step": 12011 }, { "epoch": 0.6, "learning_rate": 7.28835838730528e-06, "loss": 0.3479, "step": 12012 }, { "epoch": 0.6, "learning_rate": 7.286801804341226e-06, "loss": 0.3449, "step": 12013 }, { "epoch": 0.6, "learning_rate": 7.285245292337317e-06, "loss": 0.3036, "step": 12014 }, { "epoch": 0.6, "learning_rate": 7.283688851334263e-06, "loss": 0.3738, "step": 12015 }, { "epoch": 0.6, "learning_rate": 7.282132481372765e-06, "loss": 0.3208, "step": 12016 }, { "epoch": 0.6, "learning_rate": 7.280576182493535e-06, "loss": 0.1851, "step": 12017 }, { "epoch": 0.6, "learning_rate": 7.279019954737274e-06, "loss": 0.3297, "step": 12018 }, { "epoch": 0.6, "learning_rate": 7.277463798144681e-06, "loss": 0.1749, "step": 12019 }, { "epoch": 0.6, "learning_rate": 7.275907712756454e-06, "loss": 0.149, "step": 12020 }, { "epoch": 0.6, "learning_rate": 7.274351698613296e-06, "loss": 0.3078, "step": 12021 }, { "epoch": 0.6, "learning_rate": 7.272795755755899e-06, "loss": 0.2899, "step": 12022 }, { "epoch": 0.6, "learning_rate": 7.271239884224959e-06, "loss": 0.4033, "step": 12023 }, { "epoch": 0.6, "learning_rate": 7.269684084061164e-06, "loss": 0.3228, "step": 12024 }, { "epoch": 0.6, "learning_rate": 7.268128355305202e-06, "loss": 0.1538, "step": 12025 }, { "epoch": 0.6, "learning_rate": 7.266572697997768e-06, "loss": 0.6068, "step": 12026 }, { "epoch": 0.6, "learning_rate": 7.2650171121795466e-06, "loss": 0.4911, "step": 12027 }, { "epoch": 0.6, "learning_rate": 7.26346159789122e-06, "loss": 0.4012, "step": 12028 }, { "epoch": 0.6, "learning_rate": 7.261906155173468e-06, "loss": 0.2664, "step": 12029 }, { "epoch": 0.6, "learning_rate": 7.260350784066976e-06, "loss": 0.402, "step": 12030 }, { "epoch": 0.6, "learning_rate": 7.258795484612422e-06, "loss": 0.4951, "step": 12031 }, { "epoch": 0.6, "learning_rate": 7.257240256850482e-06, "loss": 0.1862, "step": 12032 }, { "epoch": 0.6, "learning_rate": 7.25568510082183e-06, "loss": 0.4788, "step": 12033 }, { "epoch": 0.6, "learning_rate": 7.254130016567136e-06, "loss": 0.3692, "step": 12034 }, { "epoch": 0.6, "learning_rate": 7.252575004127079e-06, "loss": 0.2438, "step": 12035 }, { "epoch": 0.6, "learning_rate": 7.251020063542325e-06, "loss": 0.4119, "step": 12036 }, { "epoch": 0.6, "learning_rate": 7.24946519485354e-06, "loss": 0.2499, "step": 12037 }, { "epoch": 0.6, "learning_rate": 7.247910398101385e-06, "loss": 0.376, "step": 12038 }, { "epoch": 0.6, "learning_rate": 7.246355673326535e-06, "loss": 0.4701, "step": 12039 }, { "epoch": 0.6, "learning_rate": 7.244801020569645e-06, "loss": 0.393, "step": 12040 }, { "epoch": 0.6, "learning_rate": 7.243246439871377e-06, "loss": 0.346, "step": 12041 }, { "epoch": 0.6, "learning_rate": 7.241691931272386e-06, "loss": 0.2727, "step": 12042 }, { "epoch": 0.6, "learning_rate": 7.240137494813327e-06, "loss": 0.4545, "step": 12043 }, { "epoch": 0.6, "learning_rate": 7.23858313053486e-06, "loss": 0.3074, "step": 12044 }, { "epoch": 0.6, "learning_rate": 7.237028838477635e-06, "loss": 0.4339, "step": 12045 }, { "epoch": 0.6, "learning_rate": 7.235474618682301e-06, "loss": 0.3007, "step": 12046 }, { "epoch": 0.6, "learning_rate": 7.233920471189504e-06, "loss": 0.2585, "step": 12047 }, { "epoch": 0.6, "learning_rate": 7.232366396039898e-06, "loss": 0.3361, "step": 12048 }, { "epoch": 0.6, "learning_rate": 7.230812393274125e-06, "loss": 0.2592, "step": 12049 }, { "epoch": 0.6, "learning_rate": 7.229258462932825e-06, "loss": 0.1816, "step": 12050 }, { "epoch": 0.6, "learning_rate": 7.227704605056643e-06, "loss": 0.4354, "step": 12051 }, { "epoch": 0.6, "learning_rate": 7.226150819686213e-06, "loss": 0.5056, "step": 12052 }, { "epoch": 0.6, "learning_rate": 7.224597106862177e-06, "loss": 0.504, "step": 12053 }, { "epoch": 0.6, "learning_rate": 7.223043466625168e-06, "loss": 0.5116, "step": 12054 }, { "epoch": 0.6, "learning_rate": 7.221489899015821e-06, "loss": 0.2638, "step": 12055 }, { "epoch": 0.6, "learning_rate": 7.2199364040747655e-06, "loss": 0.1933, "step": 12056 }, { "epoch": 0.6, "learning_rate": 7.218382981842629e-06, "loss": 0.4796, "step": 12057 }, { "epoch": 0.6, "learning_rate": 7.216829632360045e-06, "loss": 0.3028, "step": 12058 }, { "epoch": 0.6, "learning_rate": 7.215276355667638e-06, "loss": 0.2996, "step": 12059 }, { "epoch": 0.6, "learning_rate": 7.213723151806031e-06, "loss": 0.3093, "step": 12060 }, { "epoch": 0.6, "learning_rate": 7.21217002081584e-06, "loss": 0.233, "step": 12061 }, { "epoch": 0.6, "learning_rate": 7.210616962737697e-06, "loss": 0.3437, "step": 12062 }, { "epoch": 0.6, "learning_rate": 7.209063977612212e-06, "loss": 0.3382, "step": 12063 }, { "epoch": 0.6, "learning_rate": 7.207511065480004e-06, "loss": 0.4162, "step": 12064 }, { "epoch": 0.6, "learning_rate": 7.205958226381687e-06, "loss": 0.2515, "step": 12065 }, { "epoch": 0.6, "learning_rate": 7.204405460357869e-06, "loss": 0.29, "step": 12066 }, { "epoch": 0.6, "learning_rate": 7.202852767449168e-06, "loss": 0.2465, "step": 12067 }, { "epoch": 0.6, "learning_rate": 7.201300147696189e-06, "loss": 0.1897, "step": 12068 }, { "epoch": 0.6, "learning_rate": 7.199747601139539e-06, "loss": 0.5443, "step": 12069 }, { "epoch": 0.6, "learning_rate": 7.198195127819819e-06, "loss": 0.245, "step": 12070 }, { "epoch": 0.6, "learning_rate": 7.196642727777641e-06, "loss": 0.3629, "step": 12071 }, { "epoch": 0.6, "learning_rate": 7.1950904010535996e-06, "loss": 0.3067, "step": 12072 }, { "epoch": 0.6, "learning_rate": 7.193538147688295e-06, "loss": 0.4261, "step": 12073 }, { "epoch": 0.6, "learning_rate": 7.191985967722324e-06, "loss": 0.4272, "step": 12074 }, { "epoch": 0.6, "learning_rate": 7.1904338611962776e-06, "loss": 0.2728, "step": 12075 }, { "epoch": 0.6, "learning_rate": 7.188881828150758e-06, "loss": 0.1547, "step": 12076 }, { "epoch": 0.6, "learning_rate": 7.187329868626354e-06, "loss": 0.2391, "step": 12077 }, { "epoch": 0.6, "learning_rate": 7.185777982663651e-06, "loss": 0.2613, "step": 12078 }, { "epoch": 0.6, "learning_rate": 7.184226170303239e-06, "loss": 0.5043, "step": 12079 }, { "epoch": 0.6, "learning_rate": 7.182674431585703e-06, "loss": 0.6822, "step": 12080 }, { "epoch": 0.6, "learning_rate": 7.181122766551629e-06, "loss": 0.1447, "step": 12081 }, { "epoch": 0.6, "learning_rate": 7.1795711752415975e-06, "loss": 0.3406, "step": 12082 }, { "epoch": 0.6, "learning_rate": 7.178019657696187e-06, "loss": 0.4962, "step": 12083 }, { "epoch": 0.6, "learning_rate": 7.176468213955972e-06, "loss": 0.3753, "step": 12084 }, { "epoch": 0.6, "learning_rate": 7.174916844061537e-06, "loss": 0.4633, "step": 12085 }, { "epoch": 0.6, "learning_rate": 7.173365548053452e-06, "loss": 0.2541, "step": 12086 }, { "epoch": 0.6, "learning_rate": 7.171814325972289e-06, "loss": 0.2769, "step": 12087 }, { "epoch": 0.6, "learning_rate": 7.1702631778586165e-06, "loss": 0.4405, "step": 12088 }, { "epoch": 0.6, "learning_rate": 7.168712103753004e-06, "loss": 0.3909, "step": 12089 }, { "epoch": 0.6, "learning_rate": 7.167161103696019e-06, "loss": 0.2868, "step": 12090 }, { "epoch": 0.6, "learning_rate": 7.165610177728226e-06, "loss": 0.2926, "step": 12091 }, { "epoch": 0.6, "learning_rate": 7.1640593258901845e-06, "loss": 0.4896, "step": 12092 }, { "epoch": 0.6, "learning_rate": 7.162508548222455e-06, "loss": 0.1229, "step": 12093 }, { "epoch": 0.6, "learning_rate": 7.160957844765601e-06, "loss": 0.3515, "step": 12094 }, { "epoch": 0.6, "learning_rate": 7.159407215560177e-06, "loss": 0.2871, "step": 12095 }, { "epoch": 0.6, "learning_rate": 7.1578566606467355e-06, "loss": 0.5514, "step": 12096 }, { "epoch": 0.6, "learning_rate": 7.156306180065832e-06, "loss": 0.3206, "step": 12097 }, { "epoch": 0.6, "learning_rate": 7.1547557738580106e-06, "loss": 0.293, "step": 12098 }, { "epoch": 0.6, "learning_rate": 7.1532054420638295e-06, "loss": 0.3828, "step": 12099 }, { "epoch": 0.6, "learning_rate": 7.151655184723832e-06, "loss": 0.188, "step": 12100 }, { "epoch": 0.6, "learning_rate": 7.150105001878562e-06, "loss": 0.3137, "step": 12101 }, { "epoch": 0.6, "learning_rate": 7.14855489356856e-06, "loss": 0.527, "step": 12102 }, { "epoch": 0.6, "learning_rate": 7.147004859834374e-06, "loss": 0.4502, "step": 12103 }, { "epoch": 0.6, "learning_rate": 7.145454900716537e-06, "loss": 0.1166, "step": 12104 }, { "epoch": 0.6, "learning_rate": 7.14390501625559e-06, "loss": 0.3511, "step": 12105 }, { "epoch": 0.6, "learning_rate": 7.142355206492067e-06, "loss": 0.513, "step": 12106 }, { "epoch": 0.6, "learning_rate": 7.140805471466495e-06, "loss": 0.474, "step": 12107 }, { "epoch": 0.6, "learning_rate": 7.1392558112194166e-06, "loss": 0.3789, "step": 12108 }, { "epoch": 0.6, "learning_rate": 7.137706225791354e-06, "loss": 0.5847, "step": 12109 }, { "epoch": 0.6, "learning_rate": 7.136156715222838e-06, "loss": 0.3622, "step": 12110 }, { "epoch": 0.6, "learning_rate": 7.1346072795543865e-06, "loss": 0.3545, "step": 12111 }, { "epoch": 0.6, "learning_rate": 7.133057918826533e-06, "loss": 0.3301, "step": 12112 }, { "epoch": 0.6, "learning_rate": 7.1315086330797944e-06, "loss": 0.5154, "step": 12113 }, { "epoch": 0.6, "learning_rate": 7.129959422354691e-06, "loss": 0.5887, "step": 12114 }, { "epoch": 0.6, "learning_rate": 7.12841028669174e-06, "loss": 0.4035, "step": 12115 }, { "epoch": 0.6, "learning_rate": 7.126861226131454e-06, "loss": 0.2993, "step": 12116 }, { "epoch": 0.61, "learning_rate": 7.1253122407143526e-06, "loss": 0.5717, "step": 12117 }, { "epoch": 0.61, "learning_rate": 7.123763330480944e-06, "loss": 0.5574, "step": 12118 }, { "epoch": 0.61, "learning_rate": 7.122214495471738e-06, "loss": 0.3862, "step": 12119 }, { "epoch": 0.61, "learning_rate": 7.1206657357272415e-06, "loss": 0.4279, "step": 12120 }, { "epoch": 0.61, "learning_rate": 7.1191170512879585e-06, "loss": 0.5527, "step": 12121 }, { "epoch": 0.61, "learning_rate": 7.1175684421944e-06, "loss": 0.4496, "step": 12122 }, { "epoch": 0.61, "learning_rate": 7.116019908487064e-06, "loss": 0.174, "step": 12123 }, { "epoch": 0.61, "learning_rate": 7.11447145020645e-06, "loss": 0.2546, "step": 12124 }, { "epoch": 0.61, "learning_rate": 7.112923067393053e-06, "loss": 0.4833, "step": 12125 }, { "epoch": 0.61, "learning_rate": 7.1113747600873724e-06, "loss": 0.337, "step": 12126 }, { "epoch": 0.61, "learning_rate": 7.109826528329904e-06, "loss": 0.1275, "step": 12127 }, { "epoch": 0.61, "learning_rate": 7.108278372161137e-06, "loss": 0.2418, "step": 12128 }, { "epoch": 0.61, "learning_rate": 7.106730291621562e-06, "loss": 0.4569, "step": 12129 }, { "epoch": 0.61, "learning_rate": 7.105182286751661e-06, "loss": 0.1961, "step": 12130 }, { "epoch": 0.61, "learning_rate": 7.103634357591931e-06, "loss": 0.0986, "step": 12131 }, { "epoch": 0.61, "learning_rate": 7.10208650418285e-06, "loss": 0.3325, "step": 12132 }, { "epoch": 0.61, "learning_rate": 7.100538726564902e-06, "loss": 0.3863, "step": 12133 }, { "epoch": 0.61, "learning_rate": 7.098991024778561e-06, "loss": 0.3535, "step": 12134 }, { "epoch": 0.61, "learning_rate": 7.097443398864314e-06, "loss": 0.3536, "step": 12135 }, { "epoch": 0.61, "learning_rate": 7.095895848862634e-06, "loss": 0.2589, "step": 12136 }, { "epoch": 0.61, "learning_rate": 7.094348374813994e-06, "loss": 0.3968, "step": 12137 }, { "epoch": 0.61, "learning_rate": 7.092800976758867e-06, "loss": 0.385, "step": 12138 }, { "epoch": 0.61, "learning_rate": 7.09125365473772e-06, "loss": 0.462, "step": 12139 }, { "epoch": 0.61, "learning_rate": 7.089706408791026e-06, "loss": 0.1554, "step": 12140 }, { "epoch": 0.61, "learning_rate": 7.0881592389592495e-06, "loss": 0.4786, "step": 12141 }, { "epoch": 0.61, "learning_rate": 7.086612145282854e-06, "loss": 0.3845, "step": 12142 }, { "epoch": 0.61, "learning_rate": 7.085065127802297e-06, "loss": 0.2364, "step": 12143 }, { "epoch": 0.61, "learning_rate": 7.08351818655805e-06, "loss": 0.4179, "step": 12144 }, { "epoch": 0.61, "learning_rate": 7.081971321590564e-06, "loss": 0.2039, "step": 12145 }, { "epoch": 0.61, "learning_rate": 7.080424532940295e-06, "loss": 0.3781, "step": 12146 }, { "epoch": 0.61, "learning_rate": 7.0788778206477e-06, "loss": 0.4459, "step": 12147 }, { "epoch": 0.61, "learning_rate": 7.0773311847532245e-06, "loss": 0.3004, "step": 12148 }, { "epoch": 0.61, "learning_rate": 7.075784625297329e-06, "loss": 0.2971, "step": 12149 }, { "epoch": 0.61, "learning_rate": 7.074238142320455e-06, "loss": 0.6249, "step": 12150 }, { "epoch": 0.61, "learning_rate": 7.072691735863051e-06, "loss": 0.2178, "step": 12151 }, { "epoch": 0.61, "learning_rate": 7.07114540596556e-06, "loss": 0.3053, "step": 12152 }, { "epoch": 0.61, "learning_rate": 7.0695991526684245e-06, "loss": 0.5035, "step": 12153 }, { "epoch": 0.61, "learning_rate": 7.068052976012086e-06, "loss": 0.4955, "step": 12154 }, { "epoch": 0.61, "learning_rate": 7.066506876036982e-06, "loss": 0.2075, "step": 12155 }, { "epoch": 0.61, "learning_rate": 7.064960852783549e-06, "loss": 0.4688, "step": 12156 }, { "epoch": 0.61, "learning_rate": 7.063414906292215e-06, "loss": 0.4821, "step": 12157 }, { "epoch": 0.61, "learning_rate": 7.061869036603424e-06, "loss": 0.3559, "step": 12158 }, { "epoch": 0.61, "learning_rate": 7.060323243757599e-06, "loss": 0.3117, "step": 12159 }, { "epoch": 0.61, "learning_rate": 7.058777527795169e-06, "loss": 0.4795, "step": 12160 }, { "epoch": 0.61, "learning_rate": 7.057231888756559e-06, "loss": 0.6175, "step": 12161 }, { "epoch": 0.61, "learning_rate": 7.055686326682195e-06, "loss": 0.3506, "step": 12162 }, { "epoch": 0.61, "learning_rate": 7.0541408416125e-06, "loss": 0.3043, "step": 12163 }, { "epoch": 0.61, "learning_rate": 7.052595433587893e-06, "loss": 0.4801, "step": 12164 }, { "epoch": 0.61, "learning_rate": 7.051050102648792e-06, "loss": 0.3075, "step": 12165 }, { "epoch": 0.61, "learning_rate": 7.049504848835609e-06, "loss": 0.2963, "step": 12166 }, { "epoch": 0.61, "learning_rate": 7.047959672188767e-06, "loss": 0.3054, "step": 12167 }, { "epoch": 0.61, "learning_rate": 7.046414572748672e-06, "loss": 0.2069, "step": 12168 }, { "epoch": 0.61, "learning_rate": 7.044869550555735e-06, "loss": 0.2352, "step": 12169 }, { "epoch": 0.61, "learning_rate": 7.043324605650366e-06, "loss": 0.4584, "step": 12170 }, { "epoch": 0.61, "learning_rate": 7.041779738072964e-06, "loss": 0.2828, "step": 12171 }, { "epoch": 0.61, "learning_rate": 7.040234947863945e-06, "loss": 0.4816, "step": 12172 }, { "epoch": 0.61, "learning_rate": 7.0386902350637006e-06, "loss": 0.2418, "step": 12173 }, { "epoch": 0.61, "learning_rate": 7.037145599712637e-06, "loss": 0.3465, "step": 12174 }, { "epoch": 0.61, "learning_rate": 7.035601041851148e-06, "loss": 0.1383, "step": 12175 }, { "epoch": 0.61, "learning_rate": 7.034056561519631e-06, "loss": 0.3273, "step": 12176 }, { "epoch": 0.61, "learning_rate": 7.032512158758481e-06, "loss": 0.2842, "step": 12177 }, { "epoch": 0.61, "learning_rate": 7.030967833608089e-06, "loss": 0.2815, "step": 12178 }, { "epoch": 0.61, "learning_rate": 7.029423586108844e-06, "loss": 0.4452, "step": 12179 }, { "epoch": 0.61, "learning_rate": 7.027879416301131e-06, "loss": 0.3734, "step": 12180 }, { "epoch": 0.61, "learning_rate": 7.026335324225342e-06, "loss": 0.5629, "step": 12181 }, { "epoch": 0.61, "learning_rate": 7.024791309921859e-06, "loss": 0.3454, "step": 12182 }, { "epoch": 0.61, "learning_rate": 7.023247373431062e-06, "loss": 0.1447, "step": 12183 }, { "epoch": 0.61, "learning_rate": 7.0217035147933275e-06, "loss": 0.5161, "step": 12184 }, { "epoch": 0.61, "learning_rate": 7.02015973404904e-06, "loss": 0.4199, "step": 12185 }, { "epoch": 0.61, "learning_rate": 7.018616031238572e-06, "loss": 0.6216, "step": 12186 }, { "epoch": 0.61, "learning_rate": 7.017072406402297e-06, "loss": 0.334, "step": 12187 }, { "epoch": 0.61, "learning_rate": 7.015528859580586e-06, "loss": 0.3786, "step": 12188 }, { "epoch": 0.61, "learning_rate": 7.013985390813807e-06, "loss": 0.3003, "step": 12189 }, { "epoch": 0.61, "learning_rate": 7.012442000142331e-06, "loss": 0.6342, "step": 12190 }, { "epoch": 0.61, "learning_rate": 7.010898687606522e-06, "loss": 0.5846, "step": 12191 }, { "epoch": 0.61, "learning_rate": 7.009355453246743e-06, "loss": 0.4894, "step": 12192 }, { "epoch": 0.61, "learning_rate": 7.007812297103353e-06, "loss": 0.4066, "step": 12193 }, { "epoch": 0.61, "learning_rate": 7.00626921921671e-06, "loss": 0.2879, "step": 12194 }, { "epoch": 0.61, "learning_rate": 7.00472621962718e-06, "loss": 0.3086, "step": 12195 }, { "epoch": 0.61, "learning_rate": 7.003183298375111e-06, "loss": 0.2755, "step": 12196 }, { "epoch": 0.61, "learning_rate": 7.00164045550086e-06, "loss": 0.4458, "step": 12197 }, { "epoch": 0.61, "learning_rate": 7.000097691044772e-06, "loss": 0.3819, "step": 12198 }, { "epoch": 0.61, "learning_rate": 6.998555005047202e-06, "loss": 0.6291, "step": 12199 }, { "epoch": 0.61, "learning_rate": 6.9970123975484946e-06, "loss": 0.3736, "step": 12200 }, { "epoch": 0.61, "learning_rate": 6.995469868588994e-06, "loss": 0.2337, "step": 12201 }, { "epoch": 0.61, "learning_rate": 6.993927418209045e-06, "loss": 0.4401, "step": 12202 }, { "epoch": 0.61, "learning_rate": 6.992385046448982e-06, "loss": 0.5911, "step": 12203 }, { "epoch": 0.61, "learning_rate": 6.990842753349154e-06, "loss": 0.1889, "step": 12204 }, { "epoch": 0.61, "learning_rate": 6.989300538949893e-06, "loss": 0.4151, "step": 12205 }, { "epoch": 0.61, "learning_rate": 6.987758403291531e-06, "loss": 0.1322, "step": 12206 }, { "epoch": 0.61, "learning_rate": 6.9862163464144e-06, "loss": 0.5351, "step": 12207 }, { "epoch": 0.61, "learning_rate": 6.984674368358837e-06, "loss": 0.5339, "step": 12208 }, { "epoch": 0.61, "learning_rate": 6.9831324691651655e-06, "loss": 0.396, "step": 12209 }, { "epoch": 0.61, "learning_rate": 6.981590648873715e-06, "loss": 0.2862, "step": 12210 }, { "epoch": 0.61, "learning_rate": 6.980048907524806e-06, "loss": 0.3801, "step": 12211 }, { "epoch": 0.61, "learning_rate": 6.978507245158761e-06, "loss": 0.4436, "step": 12212 }, { "epoch": 0.61, "learning_rate": 6.976965661815903e-06, "loss": 0.2784, "step": 12213 }, { "epoch": 0.61, "learning_rate": 6.975424157536549e-06, "loss": 0.2971, "step": 12214 }, { "epoch": 0.61, "learning_rate": 6.973882732361015e-06, "loss": 0.3339, "step": 12215 }, { "epoch": 0.61, "learning_rate": 6.9723413863296095e-06, "loss": 0.2752, "step": 12216 }, { "epoch": 0.61, "learning_rate": 6.970800119482655e-06, "loss": 0.4024, "step": 12217 }, { "epoch": 0.61, "learning_rate": 6.969258931860454e-06, "loss": 0.2572, "step": 12218 }, { "epoch": 0.61, "learning_rate": 6.967717823503318e-06, "loss": 0.2817, "step": 12219 }, { "epoch": 0.61, "learning_rate": 6.9661767944515505e-06, "loss": 0.5279, "step": 12220 }, { "epoch": 0.61, "learning_rate": 6.964635844745451e-06, "loss": 0.3983, "step": 12221 }, { "epoch": 0.61, "learning_rate": 6.963094974425328e-06, "loss": 0.1319, "step": 12222 }, { "epoch": 0.61, "learning_rate": 6.96155418353148e-06, "loss": 0.1349, "step": 12223 }, { "epoch": 0.61, "learning_rate": 6.960013472104204e-06, "loss": 0.2945, "step": 12224 }, { "epoch": 0.61, "learning_rate": 6.958472840183791e-06, "loss": 0.4478, "step": 12225 }, { "epoch": 0.61, "learning_rate": 6.956932287810536e-06, "loss": 0.3582, "step": 12226 }, { "epoch": 0.61, "learning_rate": 6.955391815024734e-06, "loss": 0.3778, "step": 12227 }, { "epoch": 0.61, "learning_rate": 6.953851421866671e-06, "loss": 0.4508, "step": 12228 }, { "epoch": 0.61, "learning_rate": 6.9523111083766344e-06, "loss": 0.5783, "step": 12229 }, { "epoch": 0.61, "learning_rate": 6.9507708745949056e-06, "loss": 0.2282, "step": 12230 }, { "epoch": 0.61, "learning_rate": 6.949230720561773e-06, "loss": 0.5678, "step": 12231 }, { "epoch": 0.61, "learning_rate": 6.947690646317518e-06, "loss": 0.6002, "step": 12232 }, { "epoch": 0.61, "learning_rate": 6.946150651902413e-06, "loss": 0.2111, "step": 12233 }, { "epoch": 0.61, "learning_rate": 6.9446107373567405e-06, "loss": 0.5217, "step": 12234 }, { "epoch": 0.61, "learning_rate": 6.9430709027207685e-06, "loss": 0.35, "step": 12235 }, { "epoch": 0.61, "learning_rate": 6.941531148034777e-06, "loss": 0.233, "step": 12236 }, { "epoch": 0.61, "learning_rate": 6.939991473339032e-06, "loss": 0.2902, "step": 12237 }, { "epoch": 0.61, "learning_rate": 6.9384518786738e-06, "loss": 0.3062, "step": 12238 }, { "epoch": 0.61, "learning_rate": 6.936912364079347e-06, "loss": 0.5095, "step": 12239 }, { "epoch": 0.61, "learning_rate": 6.935372929595943e-06, "loss": 0.3802, "step": 12240 }, { "epoch": 0.61, "learning_rate": 6.9338335752638465e-06, "loss": 0.2246, "step": 12241 }, { "epoch": 0.61, "learning_rate": 6.932294301123317e-06, "loss": 0.4249, "step": 12242 }, { "epoch": 0.61, "learning_rate": 6.930755107214611e-06, "loss": 0.3285, "step": 12243 }, { "epoch": 0.61, "learning_rate": 6.929215993577981e-06, "loss": 0.3446, "step": 12244 }, { "epoch": 0.61, "learning_rate": 6.92767696025369e-06, "loss": 0.4445, "step": 12245 }, { "epoch": 0.61, "learning_rate": 6.926138007281984e-06, "loss": 0.3194, "step": 12246 }, { "epoch": 0.61, "learning_rate": 6.924599134703111e-06, "loss": 0.5569, "step": 12247 }, { "epoch": 0.61, "learning_rate": 6.923060342557318e-06, "loss": 0.1747, "step": 12248 }, { "epoch": 0.61, "learning_rate": 6.921521630884855e-06, "loss": 0.3789, "step": 12249 }, { "epoch": 0.61, "learning_rate": 6.91998299972596e-06, "loss": 0.4245, "step": 12250 }, { "epoch": 0.61, "learning_rate": 6.918444449120877e-06, "loss": 0.2382, "step": 12251 }, { "epoch": 0.61, "learning_rate": 6.916905979109843e-06, "loss": 0.2742, "step": 12252 }, { "epoch": 0.61, "learning_rate": 6.915367589733091e-06, "loss": 0.3362, "step": 12253 }, { "epoch": 0.61, "learning_rate": 6.913829281030861e-06, "loss": 0.327, "step": 12254 }, { "epoch": 0.61, "learning_rate": 6.912291053043387e-06, "loss": 0.499, "step": 12255 }, { "epoch": 0.61, "learning_rate": 6.910752905810896e-06, "loss": 0.2467, "step": 12256 }, { "epoch": 0.61, "learning_rate": 6.909214839373617e-06, "loss": 0.4681, "step": 12257 }, { "epoch": 0.61, "learning_rate": 6.9076768537717696e-06, "loss": 0.4726, "step": 12258 }, { "epoch": 0.61, "learning_rate": 6.9061389490455886e-06, "loss": 0.1711, "step": 12259 }, { "epoch": 0.61, "learning_rate": 6.904601125235293e-06, "loss": 0.2566, "step": 12260 }, { "epoch": 0.61, "learning_rate": 6.903063382381099e-06, "loss": 0.3719, "step": 12261 }, { "epoch": 0.61, "learning_rate": 6.901525720523223e-06, "loss": 0.3692, "step": 12262 }, { "epoch": 0.61, "learning_rate": 6.899988139701887e-06, "loss": 0.623, "step": 12263 }, { "epoch": 0.61, "learning_rate": 6.898450639957302e-06, "loss": 0.3013, "step": 12264 }, { "epoch": 0.61, "learning_rate": 6.896913221329677e-06, "loss": 0.349, "step": 12265 }, { "epoch": 0.61, "learning_rate": 6.895375883859221e-06, "loss": 0.3363, "step": 12266 }, { "epoch": 0.61, "learning_rate": 6.8938386275861405e-06, "loss": 0.4264, "step": 12267 }, { "epoch": 0.61, "learning_rate": 6.892301452550645e-06, "loss": 0.2728, "step": 12268 }, { "epoch": 0.61, "learning_rate": 6.890764358792936e-06, "loss": 0.2733, "step": 12269 }, { "epoch": 0.61, "learning_rate": 6.889227346353211e-06, "loss": 0.4593, "step": 12270 }, { "epoch": 0.61, "learning_rate": 6.88769041527167e-06, "loss": 0.5248, "step": 12271 }, { "epoch": 0.61, "learning_rate": 6.886153565588511e-06, "loss": 0.391, "step": 12272 }, { "epoch": 0.61, "learning_rate": 6.884616797343928e-06, "loss": 0.2797, "step": 12273 }, { "epoch": 0.61, "learning_rate": 6.883080110578111e-06, "loss": 0.6417, "step": 12274 }, { "epoch": 0.61, "learning_rate": 6.8815435053312516e-06, "loss": 0.3684, "step": 12275 }, { "epoch": 0.61, "learning_rate": 6.880006981643534e-06, "loss": 0.2653, "step": 12276 }, { "epoch": 0.61, "learning_rate": 6.878470539555151e-06, "loss": 0.3006, "step": 12277 }, { "epoch": 0.61, "learning_rate": 6.876934179106283e-06, "loss": 0.2497, "step": 12278 }, { "epoch": 0.61, "learning_rate": 6.8753979003371105e-06, "loss": 0.3474, "step": 12279 }, { "epoch": 0.61, "learning_rate": 6.87386170328781e-06, "loss": 0.171, "step": 12280 }, { "epoch": 0.61, "learning_rate": 6.8723255879985655e-06, "loss": 0.3868, "step": 12281 }, { "epoch": 0.61, "learning_rate": 6.870789554509549e-06, "loss": 0.3943, "step": 12282 }, { "epoch": 0.61, "learning_rate": 6.869253602860932e-06, "loss": 0.0774, "step": 12283 }, { "epoch": 0.61, "learning_rate": 6.867717733092889e-06, "loss": 0.2698, "step": 12284 }, { "epoch": 0.61, "learning_rate": 6.8661819452455824e-06, "loss": 0.3588, "step": 12285 }, { "epoch": 0.61, "learning_rate": 6.864646239359185e-06, "loss": 0.3434, "step": 12286 }, { "epoch": 0.61, "learning_rate": 6.863110615473858e-06, "loss": 0.3492, "step": 12287 }, { "epoch": 0.61, "learning_rate": 6.861575073629765e-06, "loss": 0.4104, "step": 12288 }, { "epoch": 0.61, "learning_rate": 6.860039613867064e-06, "loss": 0.3574, "step": 12289 }, { "epoch": 0.61, "learning_rate": 6.8585042362259115e-06, "loss": 0.547, "step": 12290 }, { "epoch": 0.61, "learning_rate": 6.85696894074647e-06, "loss": 0.4637, "step": 12291 }, { "epoch": 0.61, "learning_rate": 6.85543372746889e-06, "loss": 0.226, "step": 12292 }, { "epoch": 0.61, "learning_rate": 6.85389859643332e-06, "loss": 0.3672, "step": 12293 }, { "epoch": 0.61, "learning_rate": 6.85236354767991e-06, "loss": 0.4098, "step": 12294 }, { "epoch": 0.61, "learning_rate": 6.850828581248811e-06, "loss": 0.5696, "step": 12295 }, { "epoch": 0.61, "learning_rate": 6.849293697180166e-06, "loss": 0.3015, "step": 12296 }, { "epoch": 0.61, "learning_rate": 6.847758895514118e-06, "loss": 0.2889, "step": 12297 }, { "epoch": 0.61, "learning_rate": 6.846224176290808e-06, "loss": 0.3976, "step": 12298 }, { "epoch": 0.61, "learning_rate": 6.844689539550371e-06, "loss": 0.3186, "step": 12299 }, { "epoch": 0.61, "learning_rate": 6.84315498533295e-06, "loss": 0.333, "step": 12300 }, { "epoch": 0.61, "learning_rate": 6.8416205136786725e-06, "loss": 0.4673, "step": 12301 }, { "epoch": 0.61, "learning_rate": 6.840086124627676e-06, "loss": 0.4681, "step": 12302 }, { "epoch": 0.61, "learning_rate": 6.838551818220085e-06, "loss": 0.2714, "step": 12303 }, { "epoch": 0.61, "learning_rate": 6.837017594496033e-06, "loss": 0.3983, "step": 12304 }, { "epoch": 0.61, "learning_rate": 6.835483453495644e-06, "loss": 0.3738, "step": 12305 }, { "epoch": 0.61, "learning_rate": 6.833949395259038e-06, "loss": 0.4154, "step": 12306 }, { "epoch": 0.61, "learning_rate": 6.832415419826342e-06, "loss": 0.4996, "step": 12307 }, { "epoch": 0.61, "learning_rate": 6.830881527237669e-06, "loss": 0.3992, "step": 12308 }, { "epoch": 0.61, "learning_rate": 6.829347717533139e-06, "loss": 0.4851, "step": 12309 }, { "epoch": 0.61, "learning_rate": 6.8278139907528685e-06, "loss": 0.4062, "step": 12310 }, { "epoch": 0.61, "learning_rate": 6.826280346936969e-06, "loss": 0.6555, "step": 12311 }, { "epoch": 0.61, "learning_rate": 6.824746786125545e-06, "loss": 0.2982, "step": 12312 }, { "epoch": 0.61, "learning_rate": 6.823213308358712e-06, "loss": 0.3531, "step": 12313 }, { "epoch": 0.61, "learning_rate": 6.8216799136765774e-06, "loss": 0.3373, "step": 12314 }, { "epoch": 0.61, "learning_rate": 6.82014660211924e-06, "loss": 0.5972, "step": 12315 }, { "epoch": 0.61, "learning_rate": 6.818613373726804e-06, "loss": 0.2734, "step": 12316 }, { "epoch": 0.62, "learning_rate": 6.817080228539364e-06, "loss": 0.3596, "step": 12317 }, { "epoch": 0.62, "learning_rate": 6.815547166597025e-06, "loss": 0.55, "step": 12318 }, { "epoch": 0.62, "learning_rate": 6.8140141879398815e-06, "loss": 0.3443, "step": 12319 }, { "epoch": 0.62, "learning_rate": 6.812481292608022e-06, "loss": 0.5128, "step": 12320 }, { "epoch": 0.62, "learning_rate": 6.810948480641537e-06, "loss": 0.1609, "step": 12321 }, { "epoch": 0.62, "learning_rate": 6.809415752080521e-06, "loss": 0.3036, "step": 12322 }, { "epoch": 0.62, "learning_rate": 6.807883106965056e-06, "loss": 0.4675, "step": 12323 }, { "epoch": 0.62, "learning_rate": 6.806350545335227e-06, "loss": 0.391, "step": 12324 }, { "epoch": 0.62, "learning_rate": 6.804818067231118e-06, "loss": 0.6365, "step": 12325 }, { "epoch": 0.62, "learning_rate": 6.803285672692803e-06, "loss": 0.4977, "step": 12326 }, { "epoch": 0.62, "learning_rate": 6.801753361760368e-06, "loss": 0.4955, "step": 12327 }, { "epoch": 0.62, "learning_rate": 6.800221134473886e-06, "loss": 0.4626, "step": 12328 }, { "epoch": 0.62, "learning_rate": 6.798688990873428e-06, "loss": 0.2634, "step": 12329 }, { "epoch": 0.62, "learning_rate": 6.797156930999067e-06, "loss": 0.2946, "step": 12330 }, { "epoch": 0.62, "learning_rate": 6.795624954890867e-06, "loss": 0.5006, "step": 12331 }, { "epoch": 0.62, "learning_rate": 6.794093062588904e-06, "loss": 0.3642, "step": 12332 }, { "epoch": 0.62, "learning_rate": 6.792561254133238e-06, "loss": 0.3871, "step": 12333 }, { "epoch": 0.62, "learning_rate": 6.7910295295639305e-06, "loss": 0.2783, "step": 12334 }, { "epoch": 0.62, "learning_rate": 6.7894978889210415e-06, "loss": 0.4223, "step": 12335 }, { "epoch": 0.62, "learning_rate": 6.787966332244631e-06, "loss": 0.3388, "step": 12336 }, { "epoch": 0.62, "learning_rate": 6.786434859574755e-06, "loss": 0.2873, "step": 12337 }, { "epoch": 0.62, "learning_rate": 6.784903470951467e-06, "loss": 0.4273, "step": 12338 }, { "epoch": 0.62, "learning_rate": 6.783372166414817e-06, "loss": 0.396, "step": 12339 }, { "epoch": 0.62, "learning_rate": 6.781840946004851e-06, "loss": 0.3774, "step": 12340 }, { "epoch": 0.62, "learning_rate": 6.780309809761623e-06, "loss": 0.4376, "step": 12341 }, { "epoch": 0.62, "learning_rate": 6.778778757725176e-06, "loss": 0.3529, "step": 12342 }, { "epoch": 0.62, "learning_rate": 6.777247789935552e-06, "loss": 0.6555, "step": 12343 }, { "epoch": 0.62, "learning_rate": 6.7757169064327876e-06, "loss": 0.5551, "step": 12344 }, { "epoch": 0.62, "learning_rate": 6.774186107256928e-06, "loss": 0.4263, "step": 12345 }, { "epoch": 0.62, "learning_rate": 6.7726553924480046e-06, "loss": 0.2366, "step": 12346 }, { "epoch": 0.62, "learning_rate": 6.771124762046053e-06, "loss": 0.6779, "step": 12347 }, { "epoch": 0.62, "learning_rate": 6.769594216091104e-06, "loss": 0.4367, "step": 12348 }, { "epoch": 0.62, "learning_rate": 6.768063754623182e-06, "loss": 0.2192, "step": 12349 }, { "epoch": 0.62, "learning_rate": 6.766533377682325e-06, "loss": 0.279, "step": 12350 }, { "epoch": 0.62, "learning_rate": 6.765003085308553e-06, "loss": 0.4124, "step": 12351 }, { "epoch": 0.62, "learning_rate": 6.763472877541885e-06, "loss": 0.5502, "step": 12352 }, { "epoch": 0.62, "learning_rate": 6.761942754422344e-06, "loss": 0.445, "step": 12353 }, { "epoch": 0.62, "learning_rate": 6.760412715989952e-06, "loss": 0.282, "step": 12354 }, { "epoch": 0.62, "learning_rate": 6.758882762284722e-06, "loss": 0.2293, "step": 12355 }, { "epoch": 0.62, "learning_rate": 6.757352893346669e-06, "loss": 0.2185, "step": 12356 }, { "epoch": 0.62, "learning_rate": 6.755823109215804e-06, "loss": 0.5058, "step": 12357 }, { "epoch": 0.62, "learning_rate": 6.7542934099321335e-06, "loss": 0.1901, "step": 12358 }, { "epoch": 0.62, "learning_rate": 6.7527637955356704e-06, "loss": 0.5589, "step": 12359 }, { "epoch": 0.62, "learning_rate": 6.751234266066418e-06, "loss": 0.2959, "step": 12360 }, { "epoch": 0.62, "learning_rate": 6.749704821564377e-06, "loss": 0.3362, "step": 12361 }, { "epoch": 0.62, "learning_rate": 6.748175462069549e-06, "loss": 0.307, "step": 12362 }, { "epoch": 0.62, "learning_rate": 6.74664618762193e-06, "loss": 0.2886, "step": 12363 }, { "epoch": 0.62, "learning_rate": 6.7451169982615215e-06, "loss": 0.3015, "step": 12364 }, { "epoch": 0.62, "learning_rate": 6.743587894028316e-06, "loss": 0.294, "step": 12365 }, { "epoch": 0.62, "learning_rate": 6.742058874962303e-06, "loss": 0.3292, "step": 12366 }, { "epoch": 0.62, "learning_rate": 6.740529941103469e-06, "loss": 0.3373, "step": 12367 }, { "epoch": 0.62, "learning_rate": 6.7390010924918105e-06, "loss": 0.2382, "step": 12368 }, { "epoch": 0.62, "learning_rate": 6.737472329167305e-06, "loss": 0.3454, "step": 12369 }, { "epoch": 0.62, "learning_rate": 6.73594365116994e-06, "loss": 0.4282, "step": 12370 }, { "epoch": 0.62, "learning_rate": 6.734415058539692e-06, "loss": 0.4683, "step": 12371 }, { "epoch": 0.62, "learning_rate": 6.73288655131654e-06, "loss": 0.4687, "step": 12372 }, { "epoch": 0.62, "learning_rate": 6.731358129540463e-06, "loss": 0.6598, "step": 12373 }, { "epoch": 0.62, "learning_rate": 6.729829793251432e-06, "loss": 0.5141, "step": 12374 }, { "epoch": 0.62, "learning_rate": 6.728301542489422e-06, "loss": 0.3187, "step": 12375 }, { "epoch": 0.62, "learning_rate": 6.726773377294393e-06, "loss": 0.1929, "step": 12376 }, { "epoch": 0.62, "learning_rate": 6.7252452977063246e-06, "loss": 0.3279, "step": 12377 }, { "epoch": 0.62, "learning_rate": 6.723717303765176e-06, "loss": 0.3796, "step": 12378 }, { "epoch": 0.62, "learning_rate": 6.7221893955109095e-06, "loss": 0.4091, "step": 12379 }, { "epoch": 0.62, "learning_rate": 6.720661572983488e-06, "loss": 0.4215, "step": 12380 }, { "epoch": 0.62, "learning_rate": 6.719133836222864e-06, "loss": 0.5504, "step": 12381 }, { "epoch": 0.62, "learning_rate": 6.717606185269001e-06, "loss": 0.2876, "step": 12382 }, { "epoch": 0.62, "learning_rate": 6.7160786201618476e-06, "loss": 0.5762, "step": 12383 }, { "epoch": 0.62, "learning_rate": 6.714551140941359e-06, "loss": 0.2719, "step": 12384 }, { "epoch": 0.62, "learning_rate": 6.713023747647476e-06, "loss": 0.5487, "step": 12385 }, { "epoch": 0.62, "learning_rate": 6.711496440320155e-06, "loss": 0.1934, "step": 12386 }, { "epoch": 0.62, "learning_rate": 6.70996921899934e-06, "loss": 0.3308, "step": 12387 }, { "epoch": 0.62, "learning_rate": 6.70844208372497e-06, "loss": 0.2836, "step": 12388 }, { "epoch": 0.62, "learning_rate": 6.706915034536985e-06, "loss": 0.3672, "step": 12389 }, { "epoch": 0.62, "learning_rate": 6.705388071475321e-06, "loss": 0.2738, "step": 12390 }, { "epoch": 0.62, "learning_rate": 6.703861194579921e-06, "loss": 0.3227, "step": 12391 }, { "epoch": 0.62, "learning_rate": 6.702334403890713e-06, "loss": 0.3562, "step": 12392 }, { "epoch": 0.62, "learning_rate": 6.700807699447631e-06, "loss": 0.4219, "step": 12393 }, { "epoch": 0.62, "learning_rate": 6.6992810812906024e-06, "loss": 0.37, "step": 12394 }, { "epoch": 0.62, "learning_rate": 6.697754549459552e-06, "loss": 0.2218, "step": 12395 }, { "epoch": 0.62, "learning_rate": 6.696228103994407e-06, "loss": 0.1456, "step": 12396 }, { "epoch": 0.62, "learning_rate": 6.6947017449350905e-06, "loss": 0.4675, "step": 12397 }, { "epoch": 0.62, "learning_rate": 6.693175472321518e-06, "loss": 0.4593, "step": 12398 }, { "epoch": 0.62, "learning_rate": 6.691649286193609e-06, "loss": 0.376, "step": 12399 }, { "epoch": 0.62, "learning_rate": 6.690123186591282e-06, "loss": 0.3733, "step": 12400 }, { "epoch": 0.62, "learning_rate": 6.688597173554448e-06, "loss": 0.4954, "step": 12401 }, { "epoch": 0.62, "learning_rate": 6.687071247123018e-06, "loss": 0.4255, "step": 12402 }, { "epoch": 0.62, "learning_rate": 6.6855454073368996e-06, "loss": 0.5095, "step": 12403 }, { "epoch": 0.62, "learning_rate": 6.684019654235997e-06, "loss": 0.1918, "step": 12404 }, { "epoch": 0.62, "learning_rate": 6.6824939878602215e-06, "loss": 0.3965, "step": 12405 }, { "epoch": 0.62, "learning_rate": 6.68096840824947e-06, "loss": 0.376, "step": 12406 }, { "epoch": 0.62, "learning_rate": 6.679442915443641e-06, "loss": 0.5247, "step": 12407 }, { "epoch": 0.62, "learning_rate": 6.677917509482634e-06, "loss": 0.3288, "step": 12408 }, { "epoch": 0.62, "learning_rate": 6.6763921904063446e-06, "loss": 0.2745, "step": 12409 }, { "epoch": 0.62, "learning_rate": 6.674866958254665e-06, "loss": 0.3313, "step": 12410 }, { "epoch": 0.62, "learning_rate": 6.673341813067483e-06, "loss": 0.2721, "step": 12411 }, { "epoch": 0.62, "learning_rate": 6.67181675488469e-06, "loss": 0.4166, "step": 12412 }, { "epoch": 0.62, "learning_rate": 6.6702917837461656e-06, "loss": 0.4587, "step": 12413 }, { "epoch": 0.62, "learning_rate": 6.668766899691803e-06, "loss": 0.166, "step": 12414 }, { "epoch": 0.62, "learning_rate": 6.66724210276148e-06, "loss": 0.1485, "step": 12415 }, { "epoch": 0.62, "learning_rate": 6.665717392995075e-06, "loss": 0.2855, "step": 12416 }, { "epoch": 0.62, "learning_rate": 6.664192770432461e-06, "loss": 0.3723, "step": 12417 }, { "epoch": 0.62, "learning_rate": 6.662668235113518e-06, "loss": 0.3146, "step": 12418 }, { "epoch": 0.62, "learning_rate": 6.661143787078117e-06, "loss": 0.2666, "step": 12419 }, { "epoch": 0.62, "learning_rate": 6.659619426366126e-06, "loss": 0.4855, "step": 12420 }, { "epoch": 0.62, "learning_rate": 6.6580951530174145e-06, "loss": 0.563, "step": 12421 }, { "epoch": 0.62, "learning_rate": 6.656570967071843e-06, "loss": 0.2869, "step": 12422 }, { "epoch": 0.62, "learning_rate": 6.655046868569284e-06, "loss": 0.5446, "step": 12423 }, { "epoch": 0.62, "learning_rate": 6.6535228575495915e-06, "loss": 0.2967, "step": 12424 }, { "epoch": 0.62, "learning_rate": 6.651998934052626e-06, "loss": 0.5239, "step": 12425 }, { "epoch": 0.62, "learning_rate": 6.6504750981182435e-06, "loss": 0.4583, "step": 12426 }, { "epoch": 0.62, "learning_rate": 6.648951349786292e-06, "loss": 0.5046, "step": 12427 }, { "epoch": 0.62, "learning_rate": 6.647427689096635e-06, "loss": 0.3579, "step": 12428 }, { "epoch": 0.62, "learning_rate": 6.645904116089116e-06, "loss": 0.3704, "step": 12429 }, { "epoch": 0.62, "learning_rate": 6.644380630803581e-06, "loss": 0.389, "step": 12430 }, { "epoch": 0.62, "learning_rate": 6.642857233279874e-06, "loss": 0.3287, "step": 12431 }, { "epoch": 0.62, "learning_rate": 6.64133392355784e-06, "loss": 0.2584, "step": 12432 }, { "epoch": 0.62, "learning_rate": 6.639810701677318e-06, "loss": 0.6948, "step": 12433 }, { "epoch": 0.62, "learning_rate": 6.638287567678146e-06, "loss": 0.3107, "step": 12434 }, { "epoch": 0.62, "learning_rate": 6.636764521600161e-06, "loss": 0.1747, "step": 12435 }, { "epoch": 0.62, "learning_rate": 6.6352415634831905e-06, "loss": 0.2987, "step": 12436 }, { "epoch": 0.62, "learning_rate": 6.633718693367072e-06, "loss": 0.3751, "step": 12437 }, { "epoch": 0.62, "learning_rate": 6.632195911291634e-06, "loss": 0.4302, "step": 12438 }, { "epoch": 0.62, "learning_rate": 6.630673217296702e-06, "loss": 0.3748, "step": 12439 }, { "epoch": 0.62, "learning_rate": 6.629150611422093e-06, "loss": 0.1577, "step": 12440 }, { "epoch": 0.62, "learning_rate": 6.6276280937076385e-06, "loss": 0.3155, "step": 12441 }, { "epoch": 0.62, "learning_rate": 6.626105664193154e-06, "loss": 0.5451, "step": 12442 }, { "epoch": 0.62, "learning_rate": 6.6245833229184585e-06, "loss": 0.5092, "step": 12443 }, { "epoch": 0.62, "learning_rate": 6.6230610699233645e-06, "loss": 0.5311, "step": 12444 }, { "epoch": 0.62, "learning_rate": 6.621538905247683e-06, "loss": 0.0361, "step": 12445 }, { "epoch": 0.62, "learning_rate": 6.620016828931229e-06, "loss": 0.2, "step": 12446 }, { "epoch": 0.62, "learning_rate": 6.618494841013807e-06, "loss": 0.4172, "step": 12447 }, { "epoch": 0.62, "learning_rate": 6.616972941535224e-06, "loss": 0.6016, "step": 12448 }, { "epoch": 0.62, "learning_rate": 6.615451130535277e-06, "loss": 0.3758, "step": 12449 }, { "epoch": 0.62, "learning_rate": 6.61392940805378e-06, "loss": 0.2527, "step": 12450 }, { "epoch": 0.62, "learning_rate": 6.612407774130523e-06, "loss": 0.5415, "step": 12451 }, { "epoch": 0.62, "learning_rate": 6.610886228805302e-06, "loss": 0.3169, "step": 12452 }, { "epoch": 0.62, "learning_rate": 6.609364772117914e-06, "loss": 0.321, "step": 12453 }, { "epoch": 0.62, "learning_rate": 6.607843404108148e-06, "loss": 0.3142, "step": 12454 }, { "epoch": 0.62, "learning_rate": 6.606322124815797e-06, "loss": 0.2257, "step": 12455 }, { "epoch": 0.62, "learning_rate": 6.604800934280647e-06, "loss": 0.4104, "step": 12456 }, { "epoch": 0.62, "learning_rate": 6.6032798325424785e-06, "loss": 0.2027, "step": 12457 }, { "epoch": 0.62, "learning_rate": 6.601758819641074e-06, "loss": 0.5202, "step": 12458 }, { "epoch": 0.62, "learning_rate": 6.600237895616221e-06, "loss": 0.4305, "step": 12459 }, { "epoch": 0.62, "learning_rate": 6.598717060507693e-06, "loss": 0.1826, "step": 12460 }, { "epoch": 0.62, "learning_rate": 6.597196314355266e-06, "loss": 0.618, "step": 12461 }, { "epoch": 0.62, "learning_rate": 6.5956756571987126e-06, "loss": 0.2656, "step": 12462 }, { "epoch": 0.62, "learning_rate": 6.5941550890777995e-06, "loss": 0.3387, "step": 12463 }, { "epoch": 0.62, "learning_rate": 6.592634610032303e-06, "loss": 0.5137, "step": 12464 }, { "epoch": 0.62, "learning_rate": 6.591114220101984e-06, "loss": 0.2913, "step": 12465 }, { "epoch": 0.62, "learning_rate": 6.589593919326609e-06, "loss": 0.4066, "step": 12466 }, { "epoch": 0.62, "learning_rate": 6.588073707745939e-06, "loss": 0.4254, "step": 12467 }, { "epoch": 0.62, "learning_rate": 6.586553585399729e-06, "loss": 0.5149, "step": 12468 }, { "epoch": 0.62, "learning_rate": 6.585033552327742e-06, "loss": 0.3826, "step": 12469 }, { "epoch": 0.62, "learning_rate": 6.5835136085697296e-06, "loss": 0.4074, "step": 12470 }, { "epoch": 0.62, "learning_rate": 6.581993754165443e-06, "loss": 0.3984, "step": 12471 }, { "epoch": 0.62, "learning_rate": 6.58047398915463e-06, "loss": 0.4358, "step": 12472 }, { "epoch": 0.62, "learning_rate": 6.578954313577044e-06, "loss": 0.3828, "step": 12473 }, { "epoch": 0.62, "learning_rate": 6.577434727472429e-06, "loss": 0.4506, "step": 12474 }, { "epoch": 0.62, "learning_rate": 6.575915230880524e-06, "loss": 0.17, "step": 12475 }, { "epoch": 0.62, "learning_rate": 6.574395823841073e-06, "loss": 0.5785, "step": 12476 }, { "epoch": 0.62, "learning_rate": 6.572876506393807e-06, "loss": 0.4532, "step": 12477 }, { "epoch": 0.62, "learning_rate": 6.5713572785784715e-06, "loss": 0.3523, "step": 12478 }, { "epoch": 0.62, "learning_rate": 6.5698381404347965e-06, "loss": 0.3084, "step": 12479 }, { "epoch": 0.62, "learning_rate": 6.568319092002512e-06, "loss": 0.3953, "step": 12480 }, { "epoch": 0.62, "learning_rate": 6.566800133321345e-06, "loss": 0.3446, "step": 12481 }, { "epoch": 0.62, "learning_rate": 6.565281264431025e-06, "loss": 0.4275, "step": 12482 }, { "epoch": 0.62, "learning_rate": 6.563762485371276e-06, "loss": 0.2275, "step": 12483 }, { "epoch": 0.62, "learning_rate": 6.562243796181818e-06, "loss": 0.5254, "step": 12484 }, { "epoch": 0.62, "learning_rate": 6.560725196902371e-06, "loss": 0.1987, "step": 12485 }, { "epoch": 0.62, "learning_rate": 6.559206687572648e-06, "loss": 0.2187, "step": 12486 }, { "epoch": 0.62, "learning_rate": 6.557688268232369e-06, "loss": 0.1534, "step": 12487 }, { "epoch": 0.62, "learning_rate": 6.556169938921248e-06, "loss": 0.4143, "step": 12488 }, { "epoch": 0.62, "learning_rate": 6.554651699678991e-06, "loss": 0.3992, "step": 12489 }, { "epoch": 0.62, "learning_rate": 6.553133550545303e-06, "loss": 0.4454, "step": 12490 }, { "epoch": 0.62, "learning_rate": 6.551615491559895e-06, "loss": 0.3879, "step": 12491 }, { "epoch": 0.62, "learning_rate": 6.550097522762469e-06, "loss": 0.2063, "step": 12492 }, { "epoch": 0.62, "learning_rate": 6.548579644192721e-06, "loss": 0.3953, "step": 12493 }, { "epoch": 0.62, "learning_rate": 6.547061855890351e-06, "loss": 0.7388, "step": 12494 }, { "epoch": 0.62, "learning_rate": 6.545544157895055e-06, "loss": 0.5502, "step": 12495 }, { "epoch": 0.62, "learning_rate": 6.544026550246529e-06, "loss": 0.438, "step": 12496 }, { "epoch": 0.62, "learning_rate": 6.542509032984463e-06, "loss": 0.449, "step": 12497 }, { "epoch": 0.62, "learning_rate": 6.540991606148546e-06, "loss": 0.5527, "step": 12498 }, { "epoch": 0.62, "learning_rate": 6.539474269778461e-06, "loss": 0.2699, "step": 12499 }, { "epoch": 0.62, "learning_rate": 6.5379570239138905e-06, "loss": 0.471, "step": 12500 }, { "epoch": 0.62, "learning_rate": 6.536439868594523e-06, "loss": 0.3111, "step": 12501 }, { "epoch": 0.62, "learning_rate": 6.5349228038600355e-06, "loss": 0.3482, "step": 12502 }, { "epoch": 0.62, "learning_rate": 6.533405829750103e-06, "loss": 0.3545, "step": 12503 }, { "epoch": 0.62, "learning_rate": 6.531888946304398e-06, "loss": 0.4871, "step": 12504 }, { "epoch": 0.62, "learning_rate": 6.530372153562598e-06, "loss": 0.3489, "step": 12505 }, { "epoch": 0.62, "learning_rate": 6.5288554515643685e-06, "loss": 0.4117, "step": 12506 }, { "epoch": 0.62, "learning_rate": 6.5273388403493795e-06, "loss": 0.5443, "step": 12507 }, { "epoch": 0.62, "learning_rate": 6.525822319957296e-06, "loss": 0.4472, "step": 12508 }, { "epoch": 0.62, "learning_rate": 6.524305890427773e-06, "loss": 0.3822, "step": 12509 }, { "epoch": 0.62, "learning_rate": 6.522789551800482e-06, "loss": 0.3256, "step": 12510 }, { "epoch": 0.62, "learning_rate": 6.5212733041150765e-06, "loss": 0.0857, "step": 12511 }, { "epoch": 0.62, "learning_rate": 6.519757147411212e-06, "loss": 0.2436, "step": 12512 }, { "epoch": 0.62, "learning_rate": 6.5182410817285355e-06, "loss": 0.3593, "step": 12513 }, { "epoch": 0.62, "learning_rate": 6.516725107106707e-06, "loss": 0.2808, "step": 12514 }, { "epoch": 0.62, "learning_rate": 6.515209223585373e-06, "loss": 0.4327, "step": 12515 }, { "epoch": 0.62, "learning_rate": 6.513693431204175e-06, "loss": 0.3088, "step": 12516 }, { "epoch": 0.63, "learning_rate": 6.512177730002761e-06, "loss": 0.0943, "step": 12517 }, { "epoch": 0.63, "learning_rate": 6.5106621200207676e-06, "loss": 0.3244, "step": 12518 }, { "epoch": 0.63, "learning_rate": 6.509146601297838e-06, "loss": 0.5205, "step": 12519 }, { "epoch": 0.63, "learning_rate": 6.507631173873607e-06, "loss": 0.3461, "step": 12520 }, { "epoch": 0.63, "learning_rate": 6.50611583778771e-06, "loss": 0.2839, "step": 12521 }, { "epoch": 0.63, "learning_rate": 6.5046005930797705e-06, "loss": 0.3614, "step": 12522 }, { "epoch": 0.63, "learning_rate": 6.5030854397894295e-06, "loss": 0.2558, "step": 12523 }, { "epoch": 0.63, "learning_rate": 6.50157037795631e-06, "loss": 0.4011, "step": 12524 }, { "epoch": 0.63, "learning_rate": 6.500055407620033e-06, "loss": 0.4959, "step": 12525 }, { "epoch": 0.63, "learning_rate": 6.498540528820223e-06, "loss": 0.2862, "step": 12526 }, { "epoch": 0.63, "learning_rate": 6.497025741596499e-06, "loss": 0.518, "step": 12527 }, { "epoch": 0.63, "learning_rate": 6.495511045988478e-06, "loss": 0.5543, "step": 12528 }, { "epoch": 0.63, "learning_rate": 6.493996442035778e-06, "loss": 0.2498, "step": 12529 }, { "epoch": 0.63, "learning_rate": 6.4924819297780075e-06, "loss": 0.3741, "step": 12530 }, { "epoch": 0.63, "learning_rate": 6.490967509254777e-06, "loss": 0.5584, "step": 12531 }, { "epoch": 0.63, "learning_rate": 6.489453180505696e-06, "loss": 0.2469, "step": 12532 }, { "epoch": 0.63, "learning_rate": 6.48793894357037e-06, "loss": 0.2226, "step": 12533 }, { "epoch": 0.63, "learning_rate": 6.486424798488401e-06, "loss": 0.4411, "step": 12534 }, { "epoch": 0.63, "learning_rate": 6.48491074529939e-06, "loss": 0.3093, "step": 12535 }, { "epoch": 0.63, "learning_rate": 6.48339678404293e-06, "loss": 0.2704, "step": 12536 }, { "epoch": 0.63, "learning_rate": 6.481882914758625e-06, "loss": 0.2348, "step": 12537 }, { "epoch": 0.63, "learning_rate": 6.480369137486065e-06, "loss": 0.2657, "step": 12538 }, { "epoch": 0.63, "learning_rate": 6.47885545226484e-06, "loss": 0.4326, "step": 12539 }, { "epoch": 0.63, "learning_rate": 6.477341859134539e-06, "loss": 0.436, "step": 12540 }, { "epoch": 0.63, "learning_rate": 6.475828358134745e-06, "loss": 0.3489, "step": 12541 }, { "epoch": 0.63, "learning_rate": 6.474314949305047e-06, "loss": 0.0945, "step": 12542 }, { "epoch": 0.63, "learning_rate": 6.4728016326850244e-06, "loss": 0.5202, "step": 12543 }, { "epoch": 0.63, "learning_rate": 6.471288408314254e-06, "loss": 0.2124, "step": 12544 }, { "epoch": 0.63, "learning_rate": 6.469775276232311e-06, "loss": 0.262, "step": 12545 }, { "epoch": 0.63, "learning_rate": 6.468262236478774e-06, "loss": 0.2664, "step": 12546 }, { "epoch": 0.63, "learning_rate": 6.466749289093214e-06, "loss": 0.2686, "step": 12547 }, { "epoch": 0.63, "learning_rate": 6.465236434115198e-06, "loss": 0.218, "step": 12548 }, { "epoch": 0.63, "learning_rate": 6.463723671584293e-06, "loss": 0.3283, "step": 12549 }, { "epoch": 0.63, "learning_rate": 6.46221100154006e-06, "loss": 0.2264, "step": 12550 }, { "epoch": 0.63, "learning_rate": 6.4606984240220695e-06, "loss": 0.616, "step": 12551 }, { "epoch": 0.63, "learning_rate": 6.459185939069874e-06, "loss": 0.2295, "step": 12552 }, { "epoch": 0.63, "learning_rate": 6.457673546723035e-06, "loss": 0.4175, "step": 12553 }, { "epoch": 0.63, "learning_rate": 6.456161247021102e-06, "loss": 0.143, "step": 12554 }, { "epoch": 0.63, "learning_rate": 6.454649040003631e-06, "loss": 0.442, "step": 12555 }, { "epoch": 0.63, "learning_rate": 6.453136925710172e-06, "loss": 0.3465, "step": 12556 }, { "epoch": 0.63, "learning_rate": 6.4516249041802705e-06, "loss": 0.7173, "step": 12557 }, { "epoch": 0.63, "learning_rate": 6.450112975453472e-06, "loss": 0.5208, "step": 12558 }, { "epoch": 0.63, "learning_rate": 6.448601139569315e-06, "loss": 0.2993, "step": 12559 }, { "epoch": 0.63, "learning_rate": 6.447089396567348e-06, "loss": 0.5266, "step": 12560 }, { "epoch": 0.63, "learning_rate": 6.445577746487105e-06, "loss": 0.2779, "step": 12561 }, { "epoch": 0.63, "learning_rate": 6.44406618936812e-06, "loss": 0.4568, "step": 12562 }, { "epoch": 0.63, "learning_rate": 6.4425547252499255e-06, "loss": 0.4164, "step": 12563 }, { "epoch": 0.63, "learning_rate": 6.441043354172052e-06, "loss": 0.3221, "step": 12564 }, { "epoch": 0.63, "learning_rate": 6.4395320761740295e-06, "loss": 0.2721, "step": 12565 }, { "epoch": 0.63, "learning_rate": 6.43802089129538e-06, "loss": 0.3332, "step": 12566 }, { "epoch": 0.63, "learning_rate": 6.436509799575632e-06, "loss": 0.2325, "step": 12567 }, { "epoch": 0.63, "learning_rate": 6.4349988010542995e-06, "loss": 0.2886, "step": 12568 }, { "epoch": 0.63, "learning_rate": 6.433487895770907e-06, "loss": 0.3494, "step": 12569 }, { "epoch": 0.63, "learning_rate": 6.431977083764966e-06, "loss": 0.3797, "step": 12570 }, { "epoch": 0.63, "learning_rate": 6.430466365075994e-06, "loss": 0.2769, "step": 12571 }, { "epoch": 0.63, "learning_rate": 6.428955739743496e-06, "loss": 0.65, "step": 12572 }, { "epoch": 0.63, "learning_rate": 6.427445207806981e-06, "loss": 0.2443, "step": 12573 }, { "epoch": 0.63, "learning_rate": 6.425934769305959e-06, "loss": 0.4582, "step": 12574 }, { "epoch": 0.63, "learning_rate": 6.424424424279935e-06, "loss": 0.4663, "step": 12575 }, { "epoch": 0.63, "learning_rate": 6.4229141727684055e-06, "loss": 0.3314, "step": 12576 }, { "epoch": 0.63, "learning_rate": 6.4214040148108684e-06, "loss": 0.3812, "step": 12577 }, { "epoch": 0.63, "learning_rate": 6.419893950446825e-06, "loss": 0.2745, "step": 12578 }, { "epoch": 0.63, "learning_rate": 6.4183839797157655e-06, "loss": 0.2765, "step": 12579 }, { "epoch": 0.63, "learning_rate": 6.416874102657182e-06, "loss": 0.384, "step": 12580 }, { "epoch": 0.63, "learning_rate": 6.415364319310564e-06, "loss": 0.0923, "step": 12581 }, { "epoch": 0.63, "learning_rate": 6.413854629715392e-06, "loss": 0.199, "step": 12582 }, { "epoch": 0.63, "learning_rate": 6.412345033911161e-06, "loss": 0.4754, "step": 12583 }, { "epoch": 0.63, "learning_rate": 6.410835531937345e-06, "loss": 0.4849, "step": 12584 }, { "epoch": 0.63, "learning_rate": 6.409326123833426e-06, "loss": 0.4888, "step": 12585 }, { "epoch": 0.63, "learning_rate": 6.407816809638874e-06, "loss": 0.2653, "step": 12586 }, { "epoch": 0.63, "learning_rate": 6.4063075893931735e-06, "loss": 0.5515, "step": 12587 }, { "epoch": 0.63, "learning_rate": 6.4047984631357905e-06, "loss": 0.3303, "step": 12588 }, { "epoch": 0.63, "learning_rate": 6.403289430906194e-06, "loss": 0.3136, "step": 12589 }, { "epoch": 0.63, "learning_rate": 6.401780492743854e-06, "loss": 0.462, "step": 12590 }, { "epoch": 0.63, "learning_rate": 6.400271648688228e-06, "loss": 0.3913, "step": 12591 }, { "epoch": 0.63, "learning_rate": 6.398762898778785e-06, "loss": 0.3653, "step": 12592 }, { "epoch": 0.63, "learning_rate": 6.3972542430549825e-06, "loss": 0.2198, "step": 12593 }, { "epoch": 0.63, "learning_rate": 6.395745681556276e-06, "loss": 0.3288, "step": 12594 }, { "epoch": 0.63, "learning_rate": 6.394237214322117e-06, "loss": 0.8503, "step": 12595 }, { "epoch": 0.63, "learning_rate": 6.392728841391963e-06, "loss": 0.4295, "step": 12596 }, { "epoch": 0.63, "learning_rate": 6.391220562805263e-06, "loss": 0.2755, "step": 12597 }, { "epoch": 0.63, "learning_rate": 6.389712378601463e-06, "loss": 0.1865, "step": 12598 }, { "epoch": 0.63, "learning_rate": 6.388204288820005e-06, "loss": 0.2528, "step": 12599 }, { "epoch": 0.63, "learning_rate": 6.386696293500333e-06, "loss": 0.2775, "step": 12600 }, { "epoch": 0.63, "learning_rate": 6.385188392681889e-06, "loss": 0.3644, "step": 12601 }, { "epoch": 0.63, "learning_rate": 6.383680586404105e-06, "loss": 0.1928, "step": 12602 }, { "epoch": 0.63, "learning_rate": 6.382172874706421e-06, "loss": 0.2366, "step": 12603 }, { "epoch": 0.63, "learning_rate": 6.380665257628267e-06, "loss": 0.3319, "step": 12604 }, { "epoch": 0.63, "learning_rate": 6.379157735209072e-06, "loss": 0.5727, "step": 12605 }, { "epoch": 0.63, "learning_rate": 6.377650307488263e-06, "loss": 0.4878, "step": 12606 }, { "epoch": 0.63, "learning_rate": 6.376142974505268e-06, "loss": 0.4246, "step": 12607 }, { "epoch": 0.63, "learning_rate": 6.374635736299508e-06, "loss": 0.2412, "step": 12608 }, { "epoch": 0.63, "learning_rate": 6.373128592910396e-06, "loss": 0.3911, "step": 12609 }, { "epoch": 0.63, "learning_rate": 6.37162154437736e-06, "loss": 0.2403, "step": 12610 }, { "epoch": 0.63, "learning_rate": 6.370114590739811e-06, "loss": 0.484, "step": 12611 }, { "epoch": 0.63, "learning_rate": 6.36860773203716e-06, "loss": 0.219, "step": 12612 }, { "epoch": 0.63, "learning_rate": 6.367100968308817e-06, "loss": 0.1785, "step": 12613 }, { "epoch": 0.63, "learning_rate": 6.365594299594188e-06, "loss": 0.4425, "step": 12614 }, { "epoch": 0.63, "learning_rate": 6.3640877259326816e-06, "loss": 0.4114, "step": 12615 }, { "epoch": 0.63, "learning_rate": 6.362581247363699e-06, "loss": 0.4853, "step": 12616 }, { "epoch": 0.63, "learning_rate": 6.36107486392664e-06, "loss": 0.5148, "step": 12617 }, { "epoch": 0.63, "learning_rate": 6.3595685756608976e-06, "loss": 0.2647, "step": 12618 }, { "epoch": 0.63, "learning_rate": 6.358062382605874e-06, "loss": 0.4947, "step": 12619 }, { "epoch": 0.63, "learning_rate": 6.356556284800961e-06, "loss": 0.2608, "step": 12620 }, { "epoch": 0.63, "learning_rate": 6.355050282285544e-06, "loss": 0.4926, "step": 12621 }, { "epoch": 0.63, "learning_rate": 6.353544375099015e-06, "loss": 0.5007, "step": 12622 }, { "epoch": 0.63, "learning_rate": 6.352038563280751e-06, "loss": 0.3365, "step": 12623 }, { "epoch": 0.63, "learning_rate": 6.3505328468701455e-06, "loss": 0.2577, "step": 12624 }, { "epoch": 0.63, "learning_rate": 6.349027225906572e-06, "loss": 0.3453, "step": 12625 }, { "epoch": 0.63, "learning_rate": 6.347521700429411e-06, "loss": 0.2181, "step": 12626 }, { "epoch": 0.63, "learning_rate": 6.346016270478032e-06, "loss": 0.2218, "step": 12627 }, { "epoch": 0.63, "learning_rate": 6.3445109360918165e-06, "loss": 0.4415, "step": 12628 }, { "epoch": 0.63, "learning_rate": 6.343005697310128e-06, "loss": 0.1431, "step": 12629 }, { "epoch": 0.63, "learning_rate": 6.341500554172336e-06, "loss": 0.3042, "step": 12630 }, { "epoch": 0.63, "learning_rate": 6.339995506717805e-06, "loss": 0.4942, "step": 12631 }, { "epoch": 0.63, "learning_rate": 6.3384905549858945e-06, "loss": 0.4166, "step": 12632 }, { "epoch": 0.63, "learning_rate": 6.336985699015972e-06, "loss": 0.4769, "step": 12633 }, { "epoch": 0.63, "learning_rate": 6.335480938847389e-06, "loss": 0.5928, "step": 12634 }, { "epoch": 0.63, "learning_rate": 6.333976274519505e-06, "loss": 0.2642, "step": 12635 }, { "epoch": 0.63, "learning_rate": 6.332471706071668e-06, "loss": 0.3621, "step": 12636 }, { "epoch": 0.63, "learning_rate": 6.3309672335432286e-06, "loss": 0.3195, "step": 12637 }, { "epoch": 0.63, "learning_rate": 6.329462856973534e-06, "loss": 0.2026, "step": 12638 }, { "epoch": 0.63, "learning_rate": 6.327958576401935e-06, "loss": 0.2811, "step": 12639 }, { "epoch": 0.63, "learning_rate": 6.326454391867769e-06, "loss": 0.48, "step": 12640 }, { "epoch": 0.63, "learning_rate": 6.324950303410375e-06, "loss": 0.5259, "step": 12641 }, { "epoch": 0.63, "learning_rate": 6.3234463110690944e-06, "loss": 0.394, "step": 12642 }, { "epoch": 0.63, "learning_rate": 6.32194241488326e-06, "loss": 0.3631, "step": 12643 }, { "epoch": 0.63, "learning_rate": 6.320438614892204e-06, "loss": 0.3855, "step": 12644 }, { "epoch": 0.63, "learning_rate": 6.3189349111352575e-06, "loss": 0.3172, "step": 12645 }, { "epoch": 0.63, "learning_rate": 6.317431303651742e-06, "loss": 0.3064, "step": 12646 }, { "epoch": 0.63, "learning_rate": 6.315927792480993e-06, "loss": 0.2201, "step": 12647 }, { "epoch": 0.63, "learning_rate": 6.314424377662327e-06, "loss": 0.3948, "step": 12648 }, { "epoch": 0.63, "learning_rate": 6.312921059235063e-06, "loss": 0.3107, "step": 12649 }, { "epoch": 0.63, "learning_rate": 6.311417837238518e-06, "loss": 0.3375, "step": 12650 }, { "epoch": 0.63, "learning_rate": 6.309914711712011e-06, "loss": 0.5303, "step": 12651 }, { "epoch": 0.63, "learning_rate": 6.30841168269485e-06, "loss": 0.3394, "step": 12652 }, { "epoch": 0.63, "learning_rate": 6.306908750226348e-06, "loss": 0.2785, "step": 12653 }, { "epoch": 0.63, "learning_rate": 6.30540591434581e-06, "loss": 0.2622, "step": 12654 }, { "epoch": 0.63, "learning_rate": 6.303903175092536e-06, "loss": 0.2034, "step": 12655 }, { "epoch": 0.63, "learning_rate": 6.302400532505839e-06, "loss": 0.1731, "step": 12656 }, { "epoch": 0.63, "learning_rate": 6.300897986625013e-06, "loss": 0.3322, "step": 12657 }, { "epoch": 0.63, "learning_rate": 6.299395537489354e-06, "loss": 0.3916, "step": 12658 }, { "epoch": 0.63, "learning_rate": 6.297893185138155e-06, "loss": 0.2737, "step": 12659 }, { "epoch": 0.63, "learning_rate": 6.296390929610715e-06, "loss": 0.6084, "step": 12660 }, { "epoch": 0.63, "learning_rate": 6.294888770946319e-06, "loss": 0.2587, "step": 12661 }, { "epoch": 0.63, "learning_rate": 6.293386709184255e-06, "loss": 0.5086, "step": 12662 }, { "epoch": 0.63, "learning_rate": 6.291884744363805e-06, "loss": 0.5386, "step": 12663 }, { "epoch": 0.63, "learning_rate": 6.2903828765242525e-06, "loss": 0.2123, "step": 12664 }, { "epoch": 0.63, "learning_rate": 6.288881105704878e-06, "loss": 0.244, "step": 12665 }, { "epoch": 0.63, "learning_rate": 6.287379431944957e-06, "loss": 0.3257, "step": 12666 }, { "epoch": 0.63, "learning_rate": 6.285877855283766e-06, "loss": 0.377, "step": 12667 }, { "epoch": 0.63, "learning_rate": 6.2843763757605745e-06, "loss": 0.4024, "step": 12668 }, { "epoch": 0.63, "learning_rate": 6.2828749934146496e-06, "loss": 0.2803, "step": 12669 }, { "epoch": 0.63, "learning_rate": 6.281373708285263e-06, "loss": 0.2603, "step": 12670 }, { "epoch": 0.63, "learning_rate": 6.279872520411676e-06, "loss": 0.349, "step": 12671 }, { "epoch": 0.63, "learning_rate": 6.278371429833152e-06, "loss": 0.2904, "step": 12672 }, { "epoch": 0.63, "learning_rate": 6.276870436588947e-06, "loss": 0.2825, "step": 12673 }, { "epoch": 0.63, "learning_rate": 6.275369540718318e-06, "loss": 0.2482, "step": 12674 }, { "epoch": 0.63, "learning_rate": 6.2738687422605224e-06, "loss": 0.2864, "step": 12675 }, { "epoch": 0.63, "learning_rate": 6.27236804125481e-06, "loss": 0.3063, "step": 12676 }, { "epoch": 0.63, "learning_rate": 6.270867437740429e-06, "loss": 0.3033, "step": 12677 }, { "epoch": 0.63, "learning_rate": 6.269366931756624e-06, "loss": 0.3039, "step": 12678 }, { "epoch": 0.63, "learning_rate": 6.267866523342643e-06, "loss": 0.5118, "step": 12679 }, { "epoch": 0.63, "learning_rate": 6.266366212537724e-06, "loss": 0.3335, "step": 12680 }, { "epoch": 0.63, "learning_rate": 6.264865999381108e-06, "loss": 0.476, "step": 12681 }, { "epoch": 0.63, "learning_rate": 6.263365883912025e-06, "loss": 0.2893, "step": 12682 }, { "epoch": 0.63, "learning_rate": 6.261865866169716e-06, "loss": 0.4149, "step": 12683 }, { "epoch": 0.63, "learning_rate": 6.26036594619341e-06, "loss": 0.4192, "step": 12684 }, { "epoch": 0.63, "learning_rate": 6.258866124022336e-06, "loss": 0.2793, "step": 12685 }, { "epoch": 0.63, "learning_rate": 6.257366399695717e-06, "loss": 0.3664, "step": 12686 }, { "epoch": 0.63, "learning_rate": 6.255866773252776e-06, "loss": 0.2612, "step": 12687 }, { "epoch": 0.63, "learning_rate": 6.254367244732738e-06, "loss": 0.4284, "step": 12688 }, { "epoch": 0.63, "learning_rate": 6.252867814174818e-06, "loss": 0.4277, "step": 12689 }, { "epoch": 0.63, "learning_rate": 6.251368481618233e-06, "loss": 0.5034, "step": 12690 }, { "epoch": 0.63, "learning_rate": 6.249869247102191e-06, "loss": 0.4346, "step": 12691 }, { "epoch": 0.63, "learning_rate": 6.2483701106659114e-06, "loss": 0.4754, "step": 12692 }, { "epoch": 0.63, "learning_rate": 6.246871072348599e-06, "loss": 0.3269, "step": 12693 }, { "epoch": 0.63, "learning_rate": 6.245372132189456e-06, "loss": 0.6035, "step": 12694 }, { "epoch": 0.63, "learning_rate": 6.243873290227689e-06, "loss": 0.206, "step": 12695 }, { "epoch": 0.63, "learning_rate": 6.2423745465024925e-06, "loss": 0.5761, "step": 12696 }, { "epoch": 0.63, "learning_rate": 6.2408759010530695e-06, "loss": 0.4245, "step": 12697 }, { "epoch": 0.63, "learning_rate": 6.239377353918617e-06, "loss": 0.2392, "step": 12698 }, { "epoch": 0.63, "learning_rate": 6.237878905138321e-06, "loss": 0.4597, "step": 12699 }, { "epoch": 0.63, "learning_rate": 6.236380554751378e-06, "loss": 0.3966, "step": 12700 }, { "epoch": 0.63, "learning_rate": 6.234882302796967e-06, "loss": 0.4649, "step": 12701 }, { "epoch": 0.63, "learning_rate": 6.233384149314283e-06, "loss": 0.1839, "step": 12702 }, { "epoch": 0.63, "learning_rate": 6.2318860943425005e-06, "loss": 0.3078, "step": 12703 }, { "epoch": 0.63, "learning_rate": 6.230388137920805e-06, "loss": 0.4545, "step": 12704 }, { "epoch": 0.63, "learning_rate": 6.228890280088364e-06, "loss": 0.3753, "step": 12705 }, { "epoch": 0.63, "learning_rate": 6.227392520884363e-06, "loss": 0.306, "step": 12706 }, { "epoch": 0.63, "learning_rate": 6.22589486034797e-06, "loss": 0.4895, "step": 12707 }, { "epoch": 0.63, "learning_rate": 6.224397298518352e-06, "loss": 0.4068, "step": 12708 }, { "epoch": 0.63, "learning_rate": 6.222899835434679e-06, "loss": 0.45, "step": 12709 }, { "epoch": 0.63, "learning_rate": 6.22140247113611e-06, "loss": 0.3925, "step": 12710 }, { "epoch": 0.63, "learning_rate": 6.219905205661814e-06, "loss": 0.3084, "step": 12711 }, { "epoch": 0.63, "learning_rate": 6.218408039050945e-06, "loss": 0.2691, "step": 12712 }, { "epoch": 0.63, "learning_rate": 6.216910971342662e-06, "loss": 0.4249, "step": 12713 }, { "epoch": 0.63, "learning_rate": 6.215414002576116e-06, "loss": 0.3479, "step": 12714 }, { "epoch": 0.63, "learning_rate": 6.213917132790461e-06, "loss": 0.2852, "step": 12715 }, { "epoch": 0.63, "learning_rate": 6.212420362024844e-06, "loss": 0.2836, "step": 12716 }, { "epoch": 0.63, "learning_rate": 6.2109236903184135e-06, "loss": 0.615, "step": 12717 }, { "epoch": 0.64, "learning_rate": 6.209427117710309e-06, "loss": 0.3696, "step": 12718 }, { "epoch": 0.64, "learning_rate": 6.207930644239671e-06, "loss": 0.9274, "step": 12719 }, { "epoch": 0.64, "learning_rate": 6.206434269945643e-06, "loss": 0.4083, "step": 12720 }, { "epoch": 0.64, "learning_rate": 6.204937994867359e-06, "loss": 0.3091, "step": 12721 }, { "epoch": 0.64, "learning_rate": 6.203441819043951e-06, "loss": 0.2868, "step": 12722 }, { "epoch": 0.64, "learning_rate": 6.201945742514547e-06, "loss": 0.3604, "step": 12723 }, { "epoch": 0.64, "learning_rate": 6.200449765318279e-06, "loss": 0.3046, "step": 12724 }, { "epoch": 0.64, "learning_rate": 6.198953887494272e-06, "loss": 0.2021, "step": 12725 }, { "epoch": 0.64, "learning_rate": 6.197458109081648e-06, "loss": 0.3418, "step": 12726 }, { "epoch": 0.64, "learning_rate": 6.1959624301195264e-06, "loss": 0.3519, "step": 12727 }, { "epoch": 0.64, "learning_rate": 6.194466850647021e-06, "loss": 0.3843, "step": 12728 }, { "epoch": 0.64, "learning_rate": 6.192971370703253e-06, "loss": 0.5246, "step": 12729 }, { "epoch": 0.64, "learning_rate": 6.1914759903273345e-06, "loss": 0.5572, "step": 12730 }, { "epoch": 0.64, "learning_rate": 6.189980709558373e-06, "loss": 0.4689, "step": 12731 }, { "epoch": 0.64, "learning_rate": 6.188485528435471e-06, "loss": 0.3698, "step": 12732 }, { "epoch": 0.64, "learning_rate": 6.186990446997743e-06, "loss": 0.3097, "step": 12733 }, { "epoch": 0.64, "learning_rate": 6.185495465284285e-06, "loss": 0.4082, "step": 12734 }, { "epoch": 0.64, "learning_rate": 6.184000583334197e-06, "loss": 0.3747, "step": 12735 }, { "epoch": 0.64, "learning_rate": 6.182505801186576e-06, "loss": 0.403, "step": 12736 }, { "epoch": 0.64, "learning_rate": 6.181011118880513e-06, "loss": 0.6608, "step": 12737 }, { "epoch": 0.64, "learning_rate": 6.179516536455105e-06, "loss": 0.6467, "step": 12738 }, { "epoch": 0.64, "learning_rate": 6.178022053949437e-06, "loss": 0.4184, "step": 12739 }, { "epoch": 0.64, "learning_rate": 6.176527671402597e-06, "loss": 0.199, "step": 12740 }, { "epoch": 0.64, "learning_rate": 6.175033388853668e-06, "loss": 0.3858, "step": 12741 }, { "epoch": 0.64, "learning_rate": 6.173539206341727e-06, "loss": 0.3754, "step": 12742 }, { "epoch": 0.64, "learning_rate": 6.172045123905861e-06, "loss": 0.4288, "step": 12743 }, { "epoch": 0.64, "learning_rate": 6.170551141585139e-06, "loss": 0.2871, "step": 12744 }, { "epoch": 0.64, "learning_rate": 6.169057259418638e-06, "loss": 0.5416, "step": 12745 }, { "epoch": 0.64, "learning_rate": 6.167563477445422e-06, "loss": 0.4336, "step": 12746 }, { "epoch": 0.64, "learning_rate": 6.166069795704568e-06, "loss": 0.4888, "step": 12747 }, { "epoch": 0.64, "learning_rate": 6.164576214235136e-06, "loss": 0.3097, "step": 12748 }, { "epoch": 0.64, "learning_rate": 6.163082733076191e-06, "loss": 0.5978, "step": 12749 }, { "epoch": 0.64, "learning_rate": 6.161589352266792e-06, "loss": 0.3012, "step": 12750 }, { "epoch": 0.64, "learning_rate": 6.160096071845992e-06, "loss": 0.2637, "step": 12751 }, { "epoch": 0.64, "learning_rate": 6.158602891852854e-06, "loss": 0.4619, "step": 12752 }, { "epoch": 0.64, "learning_rate": 6.157109812326425e-06, "loss": 0.3647, "step": 12753 }, { "epoch": 0.64, "learning_rate": 6.1556168333057556e-06, "loss": 0.2697, "step": 12754 }, { "epoch": 0.64, "learning_rate": 6.1541239548298885e-06, "loss": 0.5209, "step": 12755 }, { "epoch": 0.64, "learning_rate": 6.152631176937875e-06, "loss": 0.7245, "step": 12756 }, { "epoch": 0.64, "learning_rate": 6.1511384996687565e-06, "loss": 0.1388, "step": 12757 }, { "epoch": 0.64, "learning_rate": 6.149645923061568e-06, "loss": 0.2593, "step": 12758 }, { "epoch": 0.64, "learning_rate": 6.148153447155348e-06, "loss": 0.2676, "step": 12759 }, { "epoch": 0.64, "learning_rate": 6.146661071989126e-06, "loss": 0.2393, "step": 12760 }, { "epoch": 0.64, "learning_rate": 6.14516879760194e-06, "loss": 0.2444, "step": 12761 }, { "epoch": 0.64, "learning_rate": 6.143676624032815e-06, "loss": 0.2911, "step": 12762 }, { "epoch": 0.64, "learning_rate": 6.142184551320778e-06, "loss": 0.2155, "step": 12763 }, { "epoch": 0.64, "learning_rate": 6.140692579504846e-06, "loss": 0.3195, "step": 12764 }, { "epoch": 0.64, "learning_rate": 6.139200708624049e-06, "loss": 0.1749, "step": 12765 }, { "epoch": 0.64, "learning_rate": 6.1377089387174e-06, "loss": 0.453, "step": 12766 }, { "epoch": 0.64, "learning_rate": 6.136217269823916e-06, "loss": 0.3374, "step": 12767 }, { "epoch": 0.64, "learning_rate": 6.134725701982609e-06, "loss": 0.2787, "step": 12768 }, { "epoch": 0.64, "learning_rate": 6.133234235232483e-06, "loss": 0.5224, "step": 12769 }, { "epoch": 0.64, "learning_rate": 6.131742869612557e-06, "loss": 0.3439, "step": 12770 }, { "epoch": 0.64, "learning_rate": 6.130251605161829e-06, "loss": 0.3052, "step": 12771 }, { "epoch": 0.64, "learning_rate": 6.1287604419193014e-06, "loss": 0.296, "step": 12772 }, { "epoch": 0.64, "learning_rate": 6.127269379923975e-06, "loss": 0.5701, "step": 12773 }, { "epoch": 0.64, "learning_rate": 6.125778419214843e-06, "loss": 0.2177, "step": 12774 }, { "epoch": 0.64, "learning_rate": 6.124287559830904e-06, "loss": 0.5745, "step": 12775 }, { "epoch": 0.64, "learning_rate": 6.1227968018111465e-06, "loss": 0.5531, "step": 12776 }, { "epoch": 0.64, "learning_rate": 6.1213061451945625e-06, "loss": 0.3485, "step": 12777 }, { "epoch": 0.64, "learning_rate": 6.119815590020131e-06, "loss": 0.3345, "step": 12778 }, { "epoch": 0.64, "learning_rate": 6.118325136326844e-06, "loss": 0.4419, "step": 12779 }, { "epoch": 0.64, "learning_rate": 6.11683478415368e-06, "loss": 0.2697, "step": 12780 }, { "epoch": 0.64, "learning_rate": 6.115344533539615e-06, "loss": 0.4672, "step": 12781 }, { "epoch": 0.64, "learning_rate": 6.113854384523628e-06, "loss": 0.59, "step": 12782 }, { "epoch": 0.64, "learning_rate": 6.112364337144684e-06, "loss": 0.4574, "step": 12783 }, { "epoch": 0.64, "learning_rate": 6.1108743914417634e-06, "loss": 0.3663, "step": 12784 }, { "epoch": 0.64, "learning_rate": 6.1093845474538295e-06, "loss": 0.2467, "step": 12785 }, { "epoch": 0.64, "learning_rate": 6.107894805219846e-06, "loss": 0.323, "step": 12786 }, { "epoch": 0.64, "learning_rate": 6.106405164778775e-06, "loss": 0.3461, "step": 12787 }, { "epoch": 0.64, "learning_rate": 6.104915626169579e-06, "loss": 0.2109, "step": 12788 }, { "epoch": 0.64, "learning_rate": 6.103426189431212e-06, "loss": 0.4512, "step": 12789 }, { "epoch": 0.64, "learning_rate": 6.10193685460263e-06, "loss": 0.5199, "step": 12790 }, { "epoch": 0.64, "learning_rate": 6.100447621722785e-06, "loss": 0.4959, "step": 12791 }, { "epoch": 0.64, "learning_rate": 6.098958490830621e-06, "loss": 0.5049, "step": 12792 }, { "epoch": 0.64, "learning_rate": 6.097469461965091e-06, "loss": 0.2444, "step": 12793 }, { "epoch": 0.64, "learning_rate": 6.095980535165137e-06, "loss": 0.4452, "step": 12794 }, { "epoch": 0.64, "learning_rate": 6.094491710469699e-06, "loss": 0.3518, "step": 12795 }, { "epoch": 0.64, "learning_rate": 6.093002987917711e-06, "loss": 0.2578, "step": 12796 }, { "epoch": 0.64, "learning_rate": 6.091514367548117e-06, "loss": 0.5189, "step": 12797 }, { "epoch": 0.64, "learning_rate": 6.090025849399843e-06, "loss": 0.3015, "step": 12798 }, { "epoch": 0.64, "learning_rate": 6.088537433511823e-06, "loss": 0.2402, "step": 12799 }, { "epoch": 0.64, "learning_rate": 6.087049119922984e-06, "loss": 0.2832, "step": 12800 }, { "epoch": 0.64, "learning_rate": 6.085560908672246e-06, "loss": 0.4995, "step": 12801 }, { "epoch": 0.64, "learning_rate": 6.0840727997985395e-06, "loss": 0.4239, "step": 12802 }, { "epoch": 0.64, "learning_rate": 6.0825847933407785e-06, "loss": 0.3719, "step": 12803 }, { "epoch": 0.64, "learning_rate": 6.081096889337883e-06, "loss": 0.1624, "step": 12804 }, { "epoch": 0.64, "learning_rate": 6.079609087828765e-06, "loss": 0.2092, "step": 12805 }, { "epoch": 0.64, "learning_rate": 6.078121388852332e-06, "loss": 0.5403, "step": 12806 }, { "epoch": 0.64, "learning_rate": 6.076633792447501e-06, "loss": 0.5523, "step": 12807 }, { "epoch": 0.64, "learning_rate": 6.0751462986531764e-06, "loss": 0.5642, "step": 12808 }, { "epoch": 0.64, "learning_rate": 6.073658907508259e-06, "loss": 0.3876, "step": 12809 }, { "epoch": 0.64, "learning_rate": 6.0721716190516475e-06, "loss": 0.4204, "step": 12810 }, { "epoch": 0.64, "learning_rate": 6.070684433322245e-06, "loss": 0.4406, "step": 12811 }, { "epoch": 0.64, "learning_rate": 6.069197350358945e-06, "loss": 0.3187, "step": 12812 }, { "epoch": 0.64, "learning_rate": 6.06771037020064e-06, "loss": 0.4942, "step": 12813 }, { "epoch": 0.64, "learning_rate": 6.06622349288622e-06, "loss": 0.6353, "step": 12814 }, { "epoch": 0.64, "learning_rate": 6.064736718454567e-06, "loss": 0.23, "step": 12815 }, { "epoch": 0.64, "learning_rate": 6.063250046944575e-06, "loss": 0.5163, "step": 12816 }, { "epoch": 0.64, "learning_rate": 6.061763478395124e-06, "loss": 0.2647, "step": 12817 }, { "epoch": 0.64, "learning_rate": 6.060277012845089e-06, "loss": 0.4566, "step": 12818 }, { "epoch": 0.64, "learning_rate": 6.058790650333346e-06, "loss": 0.2074, "step": 12819 }, { "epoch": 0.64, "learning_rate": 6.057304390898774e-06, "loss": 0.3388, "step": 12820 }, { "epoch": 0.64, "learning_rate": 6.055818234580242e-06, "loss": 0.24, "step": 12821 }, { "epoch": 0.64, "learning_rate": 6.054332181416619e-06, "loss": 0.2743, "step": 12822 }, { "epoch": 0.64, "learning_rate": 6.05284623144677e-06, "loss": 0.3309, "step": 12823 }, { "epoch": 0.64, "learning_rate": 6.051360384709555e-06, "loss": 0.4064, "step": 12824 }, { "epoch": 0.64, "learning_rate": 6.049874641243839e-06, "loss": 0.3819, "step": 12825 }, { "epoch": 0.64, "learning_rate": 6.048389001088479e-06, "loss": 0.2783, "step": 12826 }, { "epoch": 0.64, "learning_rate": 6.046903464282329e-06, "loss": 0.221, "step": 12827 }, { "epoch": 0.64, "learning_rate": 6.045418030864236e-06, "loss": 0.3646, "step": 12828 }, { "epoch": 0.64, "learning_rate": 6.043932700873061e-06, "loss": 0.5744, "step": 12829 }, { "epoch": 0.64, "learning_rate": 6.042447474347643e-06, "loss": 0.4764, "step": 12830 }, { "epoch": 0.64, "learning_rate": 6.040962351326828e-06, "loss": 0.5344, "step": 12831 }, { "epoch": 0.64, "learning_rate": 6.039477331849457e-06, "loss": 0.3043, "step": 12832 }, { "epoch": 0.64, "learning_rate": 6.037992415954368e-06, "loss": 0.375, "step": 12833 }, { "epoch": 0.64, "learning_rate": 6.036507603680399e-06, "loss": 0.3947, "step": 12834 }, { "epoch": 0.64, "learning_rate": 6.035022895066385e-06, "loss": 0.4695, "step": 12835 }, { "epoch": 0.64, "learning_rate": 6.033538290151151e-06, "loss": 0.3597, "step": 12836 }, { "epoch": 0.64, "learning_rate": 6.032053788973529e-06, "loss": 0.3434, "step": 12837 }, { "epoch": 0.64, "learning_rate": 6.030569391572338e-06, "loss": 0.2924, "step": 12838 }, { "epoch": 0.64, "learning_rate": 6.0290850979864114e-06, "loss": 0.3269, "step": 12839 }, { "epoch": 0.64, "learning_rate": 6.027600908254563e-06, "loss": 0.3394, "step": 12840 }, { "epoch": 0.64, "learning_rate": 6.026116822415609e-06, "loss": 0.3846, "step": 12841 }, { "epoch": 0.64, "learning_rate": 6.024632840508361e-06, "loss": 0.1958, "step": 12842 }, { "epoch": 0.64, "learning_rate": 6.0231489625716375e-06, "loss": 0.4184, "step": 12843 }, { "epoch": 0.64, "learning_rate": 6.0216651886442454e-06, "loss": 0.4585, "step": 12844 }, { "epoch": 0.64, "learning_rate": 6.020181518764989e-06, "loss": 0.5499, "step": 12845 }, { "epoch": 0.64, "learning_rate": 6.0186979529726715e-06, "loss": 0.3907, "step": 12846 }, { "epoch": 0.64, "learning_rate": 6.017214491306094e-06, "loss": 0.4559, "step": 12847 }, { "epoch": 0.64, "learning_rate": 6.015731133804057e-06, "loss": 0.4024, "step": 12848 }, { "epoch": 0.64, "learning_rate": 6.0142478805053526e-06, "loss": 0.4838, "step": 12849 }, { "epoch": 0.64, "learning_rate": 6.0127647314487745e-06, "loss": 0.368, "step": 12850 }, { "epoch": 0.64, "learning_rate": 6.011281686673109e-06, "loss": 0.3581, "step": 12851 }, { "epoch": 0.64, "learning_rate": 6.00979874621715e-06, "loss": 0.4733, "step": 12852 }, { "epoch": 0.64, "learning_rate": 6.00831591011968e-06, "loss": 0.4344, "step": 12853 }, { "epoch": 0.64, "learning_rate": 6.0068331784194775e-06, "loss": 0.2665, "step": 12854 }, { "epoch": 0.64, "learning_rate": 6.005350551155325e-06, "loss": 0.171, "step": 12855 }, { "epoch": 0.64, "learning_rate": 6.003868028365991e-06, "loss": 0.35, "step": 12856 }, { "epoch": 0.64, "learning_rate": 6.00238561009026e-06, "loss": 0.2274, "step": 12857 }, { "epoch": 0.64, "learning_rate": 6.000903296366898e-06, "loss": 0.2175, "step": 12858 }, { "epoch": 0.64, "learning_rate": 5.99942108723467e-06, "loss": 0.306, "step": 12859 }, { "epoch": 0.64, "learning_rate": 5.997938982732343e-06, "loss": 0.4251, "step": 12860 }, { "epoch": 0.64, "learning_rate": 5.996456982898682e-06, "loss": 0.2759, "step": 12861 }, { "epoch": 0.64, "learning_rate": 5.994975087772446e-06, "loss": 0.313, "step": 12862 }, { "epoch": 0.64, "learning_rate": 5.99349329739239e-06, "loss": 0.3011, "step": 12863 }, { "epoch": 0.64, "learning_rate": 5.992011611797268e-06, "loss": 0.2503, "step": 12864 }, { "epoch": 0.64, "learning_rate": 5.990530031025831e-06, "loss": 0.2808, "step": 12865 }, { "epoch": 0.64, "learning_rate": 5.989048555116832e-06, "loss": 0.368, "step": 12866 }, { "epoch": 0.64, "learning_rate": 5.987567184109015e-06, "loss": 0.2725, "step": 12867 }, { "epoch": 0.64, "learning_rate": 5.986085918041124e-06, "loss": 0.245, "step": 12868 }, { "epoch": 0.64, "learning_rate": 5.984604756951895e-06, "loss": 0.2904, "step": 12869 }, { "epoch": 0.64, "learning_rate": 5.9831237008800735e-06, "loss": 0.2178, "step": 12870 }, { "epoch": 0.64, "learning_rate": 5.981642749864388e-06, "loss": 0.3824, "step": 12871 }, { "epoch": 0.64, "learning_rate": 5.980161903943575e-06, "loss": 0.3513, "step": 12872 }, { "epoch": 0.64, "learning_rate": 5.978681163156361e-06, "loss": 0.4275, "step": 12873 }, { "epoch": 0.64, "learning_rate": 5.977200527541471e-06, "loss": 0.2942, "step": 12874 }, { "epoch": 0.64, "learning_rate": 5.975719997137637e-06, "loss": 0.4547, "step": 12875 }, { "epoch": 0.64, "learning_rate": 5.974239571983573e-06, "loss": 0.4424, "step": 12876 }, { "epoch": 0.64, "learning_rate": 5.972759252118003e-06, "loss": 0.3068, "step": 12877 }, { "epoch": 0.64, "learning_rate": 5.971279037579638e-06, "loss": 0.3895, "step": 12878 }, { "epoch": 0.64, "learning_rate": 5.969798928407191e-06, "loss": 0.3255, "step": 12879 }, { "epoch": 0.64, "learning_rate": 5.968318924639378e-06, "loss": 0.2099, "step": 12880 }, { "epoch": 0.64, "learning_rate": 5.9668390263149015e-06, "loss": 0.4021, "step": 12881 }, { "epoch": 0.64, "learning_rate": 5.965359233472469e-06, "loss": 0.3385, "step": 12882 }, { "epoch": 0.64, "learning_rate": 5.963879546150779e-06, "loss": 0.2927, "step": 12883 }, { "epoch": 0.64, "learning_rate": 5.962399964388534e-06, "loss": 0.5314, "step": 12884 }, { "epoch": 0.64, "learning_rate": 5.960920488224432e-06, "loss": 0.2723, "step": 12885 }, { "epoch": 0.64, "learning_rate": 5.959441117697162e-06, "loss": 0.4609, "step": 12886 }, { "epoch": 0.64, "learning_rate": 5.95796185284542e-06, "loss": 0.3294, "step": 12887 }, { "epoch": 0.64, "learning_rate": 5.956482693707886e-06, "loss": 0.3105, "step": 12888 }, { "epoch": 0.64, "learning_rate": 5.955003640323255e-06, "loss": 0.5197, "step": 12889 }, { "epoch": 0.64, "learning_rate": 5.953524692730206e-06, "loss": 0.3254, "step": 12890 }, { "epoch": 0.64, "learning_rate": 5.952045850967418e-06, "loss": 0.4481, "step": 12891 }, { "epoch": 0.64, "learning_rate": 5.950567115073567e-06, "loss": 0.3561, "step": 12892 }, { "epoch": 0.64, "learning_rate": 5.949088485087331e-06, "loss": 0.3675, "step": 12893 }, { "epoch": 0.64, "learning_rate": 5.947609961047382e-06, "loss": 0.3396, "step": 12894 }, { "epoch": 0.64, "learning_rate": 5.946131542992385e-06, "loss": 0.5031, "step": 12895 }, { "epoch": 0.64, "learning_rate": 5.944653230961008e-06, "loss": 0.3808, "step": 12896 }, { "epoch": 0.64, "learning_rate": 5.943175024991914e-06, "loss": 0.4768, "step": 12897 }, { "epoch": 0.64, "learning_rate": 5.941696925123763e-06, "loss": 0.2368, "step": 12898 }, { "epoch": 0.64, "learning_rate": 5.940218931395217e-06, "loss": 0.2622, "step": 12899 }, { "epoch": 0.64, "learning_rate": 5.938741043844924e-06, "loss": 0.3842, "step": 12900 }, { "epoch": 0.64, "learning_rate": 5.937263262511538e-06, "loss": 0.2847, "step": 12901 }, { "epoch": 0.64, "learning_rate": 5.935785587433713e-06, "loss": 0.1986, "step": 12902 }, { "epoch": 0.64, "learning_rate": 5.934308018650092e-06, "loss": 0.4, "step": 12903 }, { "epoch": 0.64, "learning_rate": 5.932830556199323e-06, "loss": 0.4483, "step": 12904 }, { "epoch": 0.64, "learning_rate": 5.93135320012004e-06, "loss": 0.3433, "step": 12905 }, { "epoch": 0.64, "learning_rate": 5.929875950450885e-06, "loss": 0.2141, "step": 12906 }, { "epoch": 0.64, "learning_rate": 5.928398807230496e-06, "loss": 0.5344, "step": 12907 }, { "epoch": 0.64, "learning_rate": 5.926921770497502e-06, "loss": 0.1641, "step": 12908 }, { "epoch": 0.64, "learning_rate": 5.925444840290535e-06, "loss": 0.448, "step": 12909 }, { "epoch": 0.64, "learning_rate": 5.923968016648222e-06, "loss": 0.1722, "step": 12910 }, { "epoch": 0.64, "learning_rate": 5.922491299609181e-06, "loss": 0.3826, "step": 12911 }, { "epoch": 0.64, "learning_rate": 5.921014689212045e-06, "loss": 0.2837, "step": 12912 }, { "epoch": 0.64, "learning_rate": 5.9195381854954285e-06, "loss": 0.4245, "step": 12913 }, { "epoch": 0.64, "learning_rate": 5.9180617884979444e-06, "loss": 0.2617, "step": 12914 }, { "epoch": 0.64, "learning_rate": 5.9165854982582035e-06, "loss": 0.2395, "step": 12915 }, { "epoch": 0.64, "learning_rate": 5.915109314814826e-06, "loss": 0.5699, "step": 12916 }, { "epoch": 0.64, "learning_rate": 5.9136332382064136e-06, "loss": 0.2648, "step": 12917 }, { "epoch": 0.65, "learning_rate": 5.912157268471572e-06, "loss": 0.2385, "step": 12918 }, { "epoch": 0.65, "learning_rate": 5.910681405648901e-06, "loss": 0.2727, "step": 12919 }, { "epoch": 0.65, "learning_rate": 5.9092056497770025e-06, "loss": 0.4026, "step": 12920 }, { "epoch": 0.65, "learning_rate": 5.907730000894474e-06, "loss": 0.1301, "step": 12921 }, { "epoch": 0.65, "learning_rate": 5.906254459039906e-06, "loss": 0.4828, "step": 12922 }, { "epoch": 0.65, "learning_rate": 5.904779024251891e-06, "loss": 0.4513, "step": 12923 }, { "epoch": 0.65, "learning_rate": 5.903303696569014e-06, "loss": 0.1894, "step": 12924 }, { "epoch": 0.65, "learning_rate": 5.901828476029865e-06, "loss": 0.4098, "step": 12925 }, { "epoch": 0.65, "learning_rate": 5.900353362673026e-06, "loss": 0.4121, "step": 12926 }, { "epoch": 0.65, "learning_rate": 5.898878356537076e-06, "loss": 0.4255, "step": 12927 }, { "epoch": 0.65, "learning_rate": 5.897403457660589e-06, "loss": 0.2802, "step": 12928 }, { "epoch": 0.65, "learning_rate": 5.895928666082138e-06, "loss": 0.0978, "step": 12929 }, { "epoch": 0.65, "learning_rate": 5.894453981840302e-06, "loss": 0.363, "step": 12930 }, { "epoch": 0.65, "learning_rate": 5.8929794049736445e-06, "loss": 0.1099, "step": 12931 }, { "epoch": 0.65, "learning_rate": 5.891504935520731e-06, "loss": 0.553, "step": 12932 }, { "epoch": 0.65, "learning_rate": 5.890030573520123e-06, "loss": 0.2753, "step": 12933 }, { "epoch": 0.65, "learning_rate": 5.888556319010386e-06, "loss": 0.4152, "step": 12934 }, { "epoch": 0.65, "learning_rate": 5.8870821720300706e-06, "loss": 0.3229, "step": 12935 }, { "epoch": 0.65, "learning_rate": 5.885608132617735e-06, "loss": 0.3335, "step": 12936 }, { "epoch": 0.65, "learning_rate": 5.884134200811931e-06, "loss": 0.3922, "step": 12937 }, { "epoch": 0.65, "learning_rate": 5.882660376651201e-06, "loss": 0.4559, "step": 12938 }, { "epoch": 0.65, "learning_rate": 5.881186660174101e-06, "loss": 0.361, "step": 12939 }, { "epoch": 0.65, "learning_rate": 5.879713051419168e-06, "loss": 0.2606, "step": 12940 }, { "epoch": 0.65, "learning_rate": 5.878239550424944e-06, "loss": 0.3715, "step": 12941 }, { "epoch": 0.65, "learning_rate": 5.876766157229966e-06, "loss": 0.3942, "step": 12942 }, { "epoch": 0.65, "learning_rate": 5.875292871872766e-06, "loss": 0.3214, "step": 12943 }, { "epoch": 0.65, "learning_rate": 5.873819694391882e-06, "loss": 0.2818, "step": 12944 }, { "epoch": 0.65, "learning_rate": 5.8723466248258396e-06, "loss": 0.2751, "step": 12945 }, { "epoch": 0.65, "learning_rate": 5.870873663213163e-06, "loss": 0.2804, "step": 12946 }, { "epoch": 0.65, "learning_rate": 5.869400809592376e-06, "loss": 0.2513, "step": 12947 }, { "epoch": 0.65, "learning_rate": 5.867928064002004e-06, "loss": 0.5588, "step": 12948 }, { "epoch": 0.65, "learning_rate": 5.86645542648056e-06, "loss": 0.5568, "step": 12949 }, { "epoch": 0.65, "learning_rate": 5.864982897066562e-06, "loss": 0.3287, "step": 12950 }, { "epoch": 0.65, "learning_rate": 5.86351047579852e-06, "loss": 0.2552, "step": 12951 }, { "epoch": 0.65, "learning_rate": 5.862038162714939e-06, "loss": 0.5082, "step": 12952 }, { "epoch": 0.65, "learning_rate": 5.860565957854334e-06, "loss": 0.3497, "step": 12953 }, { "epoch": 0.65, "learning_rate": 5.859093861255205e-06, "loss": 0.4522, "step": 12954 }, { "epoch": 0.65, "learning_rate": 5.857621872956054e-06, "loss": 0.3413, "step": 12955 }, { "epoch": 0.65, "learning_rate": 5.856149992995374e-06, "loss": 0.2033, "step": 12956 }, { "epoch": 0.65, "learning_rate": 5.8546782214116645e-06, "loss": 0.8633, "step": 12957 }, { "epoch": 0.65, "learning_rate": 5.853206558243418e-06, "loss": 0.2854, "step": 12958 }, { "epoch": 0.65, "learning_rate": 5.851735003529123e-06, "loss": 0.5054, "step": 12959 }, { "epoch": 0.65, "learning_rate": 5.850263557307265e-06, "loss": 0.4585, "step": 12960 }, { "epoch": 0.65, "learning_rate": 5.848792219616324e-06, "loss": 0.1297, "step": 12961 }, { "epoch": 0.65, "learning_rate": 5.847320990494792e-06, "loss": 0.3354, "step": 12962 }, { "epoch": 0.65, "learning_rate": 5.845849869981137e-06, "loss": 0.5494, "step": 12963 }, { "epoch": 0.65, "learning_rate": 5.844378858113839e-06, "loss": 0.552, "step": 12964 }, { "epoch": 0.65, "learning_rate": 5.842907954931365e-06, "loss": 0.3777, "step": 12965 }, { "epoch": 0.65, "learning_rate": 5.841437160472193e-06, "loss": 0.2862, "step": 12966 }, { "epoch": 0.65, "learning_rate": 5.839966474774785e-06, "loss": 0.4116, "step": 12967 }, { "epoch": 0.65, "learning_rate": 5.838495897877607e-06, "loss": 0.4051, "step": 12968 }, { "epoch": 0.65, "learning_rate": 5.837025429819117e-06, "loss": 0.2781, "step": 12969 }, { "epoch": 0.65, "learning_rate": 5.8355550706377704e-06, "loss": 0.2765, "step": 12970 }, { "epoch": 0.65, "learning_rate": 5.83408482037203e-06, "loss": 0.3483, "step": 12971 }, { "epoch": 0.65, "learning_rate": 5.832614679060345e-06, "loss": 0.4192, "step": 12972 }, { "epoch": 0.65, "learning_rate": 5.831144646741165e-06, "loss": 0.5099, "step": 12973 }, { "epoch": 0.65, "learning_rate": 5.829674723452936e-06, "loss": 0.4727, "step": 12974 }, { "epoch": 0.65, "learning_rate": 5.828204909234099e-06, "loss": 0.4333, "step": 12975 }, { "epoch": 0.65, "learning_rate": 5.826735204123102e-06, "loss": 0.3608, "step": 12976 }, { "epoch": 0.65, "learning_rate": 5.825265608158379e-06, "loss": 0.3396, "step": 12977 }, { "epoch": 0.65, "learning_rate": 5.823796121378367e-06, "loss": 0.1638, "step": 12978 }, { "epoch": 0.65, "learning_rate": 5.8223267438214935e-06, "loss": 0.4412, "step": 12979 }, { "epoch": 0.65, "learning_rate": 5.820857475526196e-06, "loss": 0.382, "step": 12980 }, { "epoch": 0.65, "learning_rate": 5.819388316530898e-06, "loss": 0.5155, "step": 12981 }, { "epoch": 0.65, "learning_rate": 5.817919266874024e-06, "loss": 0.5912, "step": 12982 }, { "epoch": 0.65, "learning_rate": 5.816450326593992e-06, "loss": 0.3868, "step": 12983 }, { "epoch": 0.65, "learning_rate": 5.81498149572922e-06, "loss": 0.481, "step": 12984 }, { "epoch": 0.65, "learning_rate": 5.813512774318127e-06, "loss": 0.4568, "step": 12985 }, { "epoch": 0.65, "learning_rate": 5.8120441623991265e-06, "loss": 0.8645, "step": 12986 }, { "epoch": 0.65, "learning_rate": 5.8105756600106245e-06, "loss": 0.1206, "step": 12987 }, { "epoch": 0.65, "learning_rate": 5.80910726719103e-06, "loss": 0.1596, "step": 12988 }, { "epoch": 0.65, "learning_rate": 5.8076389839787455e-06, "loss": 0.3807, "step": 12989 }, { "epoch": 0.65, "learning_rate": 5.8061708104121715e-06, "loss": 0.3695, "step": 12990 }, { "epoch": 0.65, "learning_rate": 5.804702746529708e-06, "loss": 0.311, "step": 12991 }, { "epoch": 0.65, "learning_rate": 5.803234792369748e-06, "loss": 0.3007, "step": 12992 }, { "epoch": 0.65, "learning_rate": 5.8017669479706825e-06, "loss": 0.2261, "step": 12993 }, { "epoch": 0.65, "learning_rate": 5.800299213370908e-06, "loss": 0.2762, "step": 12994 }, { "epoch": 0.65, "learning_rate": 5.798831588608807e-06, "loss": 0.2919, "step": 12995 }, { "epoch": 0.65, "learning_rate": 5.797364073722762e-06, "loss": 0.3497, "step": 12996 }, { "epoch": 0.65, "learning_rate": 5.7958966687511506e-06, "loss": 0.4932, "step": 12997 }, { "epoch": 0.65, "learning_rate": 5.794429373732362e-06, "loss": 0.236, "step": 12998 }, { "epoch": 0.65, "learning_rate": 5.792962188704763e-06, "loss": 0.4896, "step": 12999 }, { "epoch": 0.65, "learning_rate": 5.791495113706729e-06, "loss": 0.3118, "step": 13000 }, { "epoch": 0.65, "learning_rate": 5.790028148776627e-06, "loss": 0.2864, "step": 13001 }, { "epoch": 0.65, "learning_rate": 5.7885612939528215e-06, "loss": 0.344, "step": 13002 }, { "epoch": 0.65, "learning_rate": 5.787094549273682e-06, "loss": 0.2871, "step": 13003 }, { "epoch": 0.65, "learning_rate": 5.785627914777567e-06, "loss": 0.3597, "step": 13004 }, { "epoch": 0.65, "learning_rate": 5.784161390502836e-06, "loss": 0.4173, "step": 13005 }, { "epoch": 0.65, "learning_rate": 5.782694976487836e-06, "loss": 0.3654, "step": 13006 }, { "epoch": 0.65, "learning_rate": 5.78122867277093e-06, "loss": 0.2304, "step": 13007 }, { "epoch": 0.65, "learning_rate": 5.779762479390463e-06, "loss": 0.4608, "step": 13008 }, { "epoch": 0.65, "learning_rate": 5.778296396384779e-06, "loss": 0.2971, "step": 13009 }, { "epoch": 0.65, "learning_rate": 5.776830423792226e-06, "loss": 0.3479, "step": 13010 }, { "epoch": 0.65, "learning_rate": 5.77536456165114e-06, "loss": 0.2854, "step": 13011 }, { "epoch": 0.65, "learning_rate": 5.773898809999861e-06, "loss": 0.2752, "step": 13012 }, { "epoch": 0.65, "learning_rate": 5.772433168876723e-06, "loss": 0.587, "step": 13013 }, { "epoch": 0.65, "learning_rate": 5.77096763832006e-06, "loss": 0.3202, "step": 13014 }, { "epoch": 0.65, "learning_rate": 5.769502218368197e-06, "loss": 0.3889, "step": 13015 }, { "epoch": 0.65, "learning_rate": 5.7680369090594605e-06, "loss": 0.2541, "step": 13016 }, { "epoch": 0.65, "learning_rate": 5.766571710432178e-06, "loss": 0.3691, "step": 13017 }, { "epoch": 0.65, "learning_rate": 5.76510662252467e-06, "loss": 0.4868, "step": 13018 }, { "epoch": 0.65, "learning_rate": 5.763641645375249e-06, "loss": 0.2784, "step": 13019 }, { "epoch": 0.65, "learning_rate": 5.762176779022231e-06, "loss": 0.2991, "step": 13020 }, { "epoch": 0.65, "learning_rate": 5.760712023503929e-06, "loss": 0.1951, "step": 13021 }, { "epoch": 0.65, "learning_rate": 5.759247378858655e-06, "loss": 0.5359, "step": 13022 }, { "epoch": 0.65, "learning_rate": 5.75778284512471e-06, "loss": 0.4431, "step": 13023 }, { "epoch": 0.65, "learning_rate": 5.756318422340399e-06, "loss": 0.0871, "step": 13024 }, { "epoch": 0.65, "learning_rate": 5.754854110544016e-06, "loss": 0.2992, "step": 13025 }, { "epoch": 0.65, "learning_rate": 5.753389909773869e-06, "loss": 0.3847, "step": 13026 }, { "epoch": 0.65, "learning_rate": 5.751925820068246e-06, "loss": 0.131, "step": 13027 }, { "epoch": 0.65, "learning_rate": 5.75046184146544e-06, "loss": 0.401, "step": 13028 }, { "epoch": 0.65, "learning_rate": 5.748997974003734e-06, "loss": 0.2542, "step": 13029 }, { "epoch": 0.65, "learning_rate": 5.7475342177214226e-06, "loss": 0.4191, "step": 13030 }, { "epoch": 0.65, "learning_rate": 5.746070572656784e-06, "loss": 0.3982, "step": 13031 }, { "epoch": 0.65, "learning_rate": 5.744607038848098e-06, "loss": 0.4324, "step": 13032 }, { "epoch": 0.65, "learning_rate": 5.7431436163336394e-06, "loss": 0.1967, "step": 13033 }, { "epoch": 0.65, "learning_rate": 5.741680305151687e-06, "loss": 0.355, "step": 13034 }, { "epoch": 0.65, "learning_rate": 5.740217105340507e-06, "loss": 0.3469, "step": 13035 }, { "epoch": 0.65, "learning_rate": 5.73875401693837e-06, "loss": 0.2836, "step": 13036 }, { "epoch": 0.65, "learning_rate": 5.73729103998354e-06, "loss": 0.2333, "step": 13037 }, { "epoch": 0.65, "learning_rate": 5.735828174514275e-06, "loss": 0.4993, "step": 13038 }, { "epoch": 0.65, "learning_rate": 5.7343654205688445e-06, "loss": 0.4023, "step": 13039 }, { "epoch": 0.65, "learning_rate": 5.7329027781855e-06, "loss": 0.327, "step": 13040 }, { "epoch": 0.65, "learning_rate": 5.731440247402493e-06, "loss": 0.3433, "step": 13041 }, { "epoch": 0.65, "learning_rate": 5.7299778282580755e-06, "loss": 0.4225, "step": 13042 }, { "epoch": 0.65, "learning_rate": 5.728515520790492e-06, "loss": 0.3064, "step": 13043 }, { "epoch": 0.65, "learning_rate": 5.727053325037994e-06, "loss": 0.5515, "step": 13044 }, { "epoch": 0.65, "learning_rate": 5.725591241038819e-06, "loss": 0.327, "step": 13045 }, { "epoch": 0.65, "learning_rate": 5.724129268831208e-06, "loss": 0.5362, "step": 13046 }, { "epoch": 0.65, "learning_rate": 5.722667408453395e-06, "loss": 0.2873, "step": 13047 }, { "epoch": 0.65, "learning_rate": 5.72120565994361e-06, "loss": 0.2175, "step": 13048 }, { "epoch": 0.65, "learning_rate": 5.71974402334009e-06, "loss": 0.3649, "step": 13049 }, { "epoch": 0.65, "learning_rate": 5.718282498681059e-06, "loss": 0.3019, "step": 13050 }, { "epoch": 0.65, "learning_rate": 5.716821086004742e-06, "loss": 0.3398, "step": 13051 }, { "epoch": 0.65, "learning_rate": 5.715359785349355e-06, "loss": 0.245, "step": 13052 }, { "epoch": 0.65, "learning_rate": 5.713898596753125e-06, "loss": 0.3694, "step": 13053 }, { "epoch": 0.65, "learning_rate": 5.7124375202542635e-06, "loss": 0.1997, "step": 13054 }, { "epoch": 0.65, "learning_rate": 5.710976555890983e-06, "loss": 0.1411, "step": 13055 }, { "epoch": 0.65, "learning_rate": 5.7095157037014915e-06, "loss": 0.3093, "step": 13056 }, { "epoch": 0.65, "learning_rate": 5.7080549637239945e-06, "loss": 0.3448, "step": 13057 }, { "epoch": 0.65, "learning_rate": 5.706594335996704e-06, "loss": 0.3796, "step": 13058 }, { "epoch": 0.65, "learning_rate": 5.705133820557814e-06, "loss": 0.2922, "step": 13059 }, { "epoch": 0.65, "learning_rate": 5.703673417445523e-06, "loss": 0.4241, "step": 13060 }, { "epoch": 0.65, "learning_rate": 5.702213126698028e-06, "loss": 0.3171, "step": 13061 }, { "epoch": 0.65, "learning_rate": 5.700752948353518e-06, "loss": 0.2875, "step": 13062 }, { "epoch": 0.65, "learning_rate": 5.6992928824501845e-06, "loss": 0.2919, "step": 13063 }, { "epoch": 0.65, "learning_rate": 5.6978329290262125e-06, "loss": 0.4124, "step": 13064 }, { "epoch": 0.65, "learning_rate": 5.696373088119786e-06, "loss": 0.1562, "step": 13065 }, { "epoch": 0.65, "learning_rate": 5.69491335976908e-06, "loss": 0.5148, "step": 13066 }, { "epoch": 0.65, "learning_rate": 5.693453744012281e-06, "loss": 0.4006, "step": 13067 }, { "epoch": 0.65, "learning_rate": 5.691994240887557e-06, "loss": 0.4749, "step": 13068 }, { "epoch": 0.65, "learning_rate": 5.690534850433083e-06, "loss": 0.1149, "step": 13069 }, { "epoch": 0.65, "learning_rate": 5.689075572687022e-06, "loss": 0.4721, "step": 13070 }, { "epoch": 0.65, "learning_rate": 5.687616407687546e-06, "loss": 0.5936, "step": 13071 }, { "epoch": 0.65, "learning_rate": 5.686157355472814e-06, "loss": 0.2678, "step": 13072 }, { "epoch": 0.65, "learning_rate": 5.684698416080987e-06, "loss": 0.3371, "step": 13073 }, { "epoch": 0.65, "learning_rate": 5.683239589550222e-06, "loss": 0.2271, "step": 13074 }, { "epoch": 0.65, "learning_rate": 5.681780875918666e-06, "loss": 0.378, "step": 13075 }, { "epoch": 0.65, "learning_rate": 5.68032227522448e-06, "loss": 0.2554, "step": 13076 }, { "epoch": 0.65, "learning_rate": 5.678863787505807e-06, "loss": 0.3783, "step": 13077 }, { "epoch": 0.65, "learning_rate": 5.677405412800792e-06, "loss": 0.3905, "step": 13078 }, { "epoch": 0.65, "learning_rate": 5.6759471511475775e-06, "loss": 0.3872, "step": 13079 }, { "epoch": 0.65, "learning_rate": 5.674489002584297e-06, "loss": 0.4006, "step": 13080 }, { "epoch": 0.65, "learning_rate": 5.6730309671490955e-06, "loss": 0.2312, "step": 13081 }, { "epoch": 0.65, "learning_rate": 5.671573044880101e-06, "loss": 0.2048, "step": 13082 }, { "epoch": 0.65, "learning_rate": 5.670115235815446e-06, "loss": 0.4874, "step": 13083 }, { "epoch": 0.65, "learning_rate": 5.668657539993254e-06, "loss": 0.1122, "step": 13084 }, { "epoch": 0.65, "learning_rate": 5.667199957451653e-06, "loss": 0.3759, "step": 13085 }, { "epoch": 0.65, "learning_rate": 5.6657424882287605e-06, "loss": 0.4248, "step": 13086 }, { "epoch": 0.65, "learning_rate": 5.664285132362697e-06, "loss": 0.2533, "step": 13087 }, { "epoch": 0.65, "learning_rate": 5.662827889891577e-06, "loss": 0.2803, "step": 13088 }, { "epoch": 0.65, "learning_rate": 5.661370760853509e-06, "loss": 0.3654, "step": 13089 }, { "epoch": 0.65, "learning_rate": 5.65991374528661e-06, "loss": 0.3918, "step": 13090 }, { "epoch": 0.65, "learning_rate": 5.658456843228982e-06, "loss": 0.3855, "step": 13091 }, { "epoch": 0.65, "learning_rate": 5.657000054718728e-06, "loss": 0.3017, "step": 13092 }, { "epoch": 0.65, "learning_rate": 5.655543379793949e-06, "loss": 0.3804, "step": 13093 }, { "epoch": 0.65, "learning_rate": 5.654086818492744e-06, "loss": 0.3307, "step": 13094 }, { "epoch": 0.65, "learning_rate": 5.652630370853206e-06, "loss": 0.3658, "step": 13095 }, { "epoch": 0.65, "learning_rate": 5.651174036913427e-06, "loss": 0.3344, "step": 13096 }, { "epoch": 0.65, "learning_rate": 5.6497178167114955e-06, "loss": 0.4041, "step": 13097 }, { "epoch": 0.65, "learning_rate": 5.648261710285492e-06, "loss": 0.332, "step": 13098 }, { "epoch": 0.65, "learning_rate": 5.646805717673508e-06, "loss": 0.3865, "step": 13099 }, { "epoch": 0.65, "learning_rate": 5.645349838913619e-06, "loss": 0.3624, "step": 13100 }, { "epoch": 0.65, "learning_rate": 5.643894074043901e-06, "loss": 0.2312, "step": 13101 }, { "epoch": 0.65, "learning_rate": 5.642438423102425e-06, "loss": 0.2116, "step": 13102 }, { "epoch": 0.65, "learning_rate": 5.640982886127267e-06, "loss": 0.4286, "step": 13103 }, { "epoch": 0.65, "learning_rate": 5.639527463156494e-06, "loss": 0.343, "step": 13104 }, { "epoch": 0.65, "learning_rate": 5.638072154228168e-06, "loss": 0.512, "step": 13105 }, { "epoch": 0.65, "learning_rate": 5.636616959380352e-06, "loss": 0.4819, "step": 13106 }, { "epoch": 0.65, "learning_rate": 5.635161878651104e-06, "loss": 0.2691, "step": 13107 }, { "epoch": 0.65, "learning_rate": 5.633706912078481e-06, "loss": 0.1486, "step": 13108 }, { "epoch": 0.65, "learning_rate": 5.632252059700534e-06, "loss": 0.2165, "step": 13109 }, { "epoch": 0.65, "learning_rate": 5.630797321555315e-06, "loss": 0.4024, "step": 13110 }, { "epoch": 0.65, "learning_rate": 5.629342697680864e-06, "loss": 0.3541, "step": 13111 }, { "epoch": 0.65, "learning_rate": 5.6278881881152345e-06, "loss": 0.2709, "step": 13112 }, { "epoch": 0.65, "learning_rate": 5.626433792896464e-06, "loss": 0.2576, "step": 13113 }, { "epoch": 0.65, "learning_rate": 5.624979512062587e-06, "loss": 0.474, "step": 13114 }, { "epoch": 0.65, "learning_rate": 5.623525345651644e-06, "loss": 0.351, "step": 13115 }, { "epoch": 0.65, "learning_rate": 5.6220712937016566e-06, "loss": 0.3712, "step": 13116 }, { "epoch": 0.65, "learning_rate": 5.620617356250666e-06, "loss": 0.2703, "step": 13117 }, { "epoch": 0.66, "learning_rate": 5.6191635333366915e-06, "loss": 0.3382, "step": 13118 }, { "epoch": 0.66, "learning_rate": 5.617709824997757e-06, "loss": 0.4753, "step": 13119 }, { "epoch": 0.66, "learning_rate": 5.616256231271882e-06, "loss": 0.5093, "step": 13120 }, { "epoch": 0.66, "learning_rate": 5.6148027521970815e-06, "loss": 0.5455, "step": 13121 }, { "epoch": 0.66, "learning_rate": 5.6133493878113735e-06, "loss": 0.3561, "step": 13122 }, { "epoch": 0.66, "learning_rate": 5.611896138152768e-06, "loss": 0.4251, "step": 13123 }, { "epoch": 0.66, "learning_rate": 5.610443003259272e-06, "loss": 0.2163, "step": 13124 }, { "epoch": 0.66, "learning_rate": 5.608989983168885e-06, "loss": 0.6545, "step": 13125 }, { "epoch": 0.66, "learning_rate": 5.607537077919617e-06, "loss": 0.517, "step": 13126 }, { "epoch": 0.66, "learning_rate": 5.606084287549465e-06, "loss": 0.4263, "step": 13127 }, { "epoch": 0.66, "learning_rate": 5.604631612096422e-06, "loss": 0.222, "step": 13128 }, { "epoch": 0.66, "learning_rate": 5.603179051598484e-06, "loss": 0.2732, "step": 13129 }, { "epoch": 0.66, "learning_rate": 5.601726606093634e-06, "loss": 0.2756, "step": 13130 }, { "epoch": 0.66, "learning_rate": 5.600274275619868e-06, "loss": 0.3288, "step": 13131 }, { "epoch": 0.66, "learning_rate": 5.5988220602151655e-06, "loss": 0.3252, "step": 13132 }, { "epoch": 0.66, "learning_rate": 5.5973699599175094e-06, "loss": 0.5656, "step": 13133 }, { "epoch": 0.66, "learning_rate": 5.595917974764873e-06, "loss": 0.3384, "step": 13134 }, { "epoch": 0.66, "learning_rate": 5.5944661047952354e-06, "loss": 0.3949, "step": 13135 }, { "epoch": 0.66, "learning_rate": 5.593014350046566e-06, "loss": 0.5038, "step": 13136 }, { "epoch": 0.66, "learning_rate": 5.591562710556834e-06, "loss": 0.414, "step": 13137 }, { "epoch": 0.66, "learning_rate": 5.5901111863640046e-06, "loss": 0.2892, "step": 13138 }, { "epoch": 0.66, "learning_rate": 5.588659777506038e-06, "loss": 0.2198, "step": 13139 }, { "epoch": 0.66, "learning_rate": 5.587208484020902e-06, "loss": 0.5391, "step": 13140 }, { "epoch": 0.66, "learning_rate": 5.585757305946548e-06, "loss": 0.2633, "step": 13141 }, { "epoch": 0.66, "learning_rate": 5.58430624332093e-06, "loss": 0.5094, "step": 13142 }, { "epoch": 0.66, "learning_rate": 5.582855296181993e-06, "loss": 0.4075, "step": 13143 }, { "epoch": 0.66, "learning_rate": 5.581404464567696e-06, "loss": 0.455, "step": 13144 }, { "epoch": 0.66, "learning_rate": 5.579953748515977e-06, "loss": 0.3628, "step": 13145 }, { "epoch": 0.66, "learning_rate": 5.57850314806478e-06, "loss": 0.4836, "step": 13146 }, { "epoch": 0.66, "learning_rate": 5.577052663252041e-06, "loss": 0.5279, "step": 13147 }, { "epoch": 0.66, "learning_rate": 5.575602294115692e-06, "loss": 0.1549, "step": 13148 }, { "epoch": 0.66, "learning_rate": 5.574152040693675e-06, "loss": 0.4162, "step": 13149 }, { "epoch": 0.66, "learning_rate": 5.572701903023914e-06, "loss": 0.3771, "step": 13150 }, { "epoch": 0.66, "learning_rate": 5.571251881144336e-06, "loss": 0.1044, "step": 13151 }, { "epoch": 0.66, "learning_rate": 5.569801975092866e-06, "loss": 0.2929, "step": 13152 }, { "epoch": 0.66, "learning_rate": 5.568352184907416e-06, "loss": 0.5748, "step": 13153 }, { "epoch": 0.66, "learning_rate": 5.566902510625916e-06, "loss": 0.392, "step": 13154 }, { "epoch": 0.66, "learning_rate": 5.565452952286274e-06, "loss": 0.347, "step": 13155 }, { "epoch": 0.66, "learning_rate": 5.564003509926402e-06, "loss": 0.3274, "step": 13156 }, { "epoch": 0.66, "learning_rate": 5.562554183584208e-06, "loss": 0.3858, "step": 13157 }, { "epoch": 0.66, "learning_rate": 5.5611049732975975e-06, "loss": 0.3845, "step": 13158 }, { "epoch": 0.66, "learning_rate": 5.559655879104473e-06, "loss": 0.3694, "step": 13159 }, { "epoch": 0.66, "learning_rate": 5.558206901042733e-06, "loss": 0.3907, "step": 13160 }, { "epoch": 0.66, "learning_rate": 5.556758039150273e-06, "loss": 0.3273, "step": 13161 }, { "epoch": 0.66, "learning_rate": 5.5553092934649835e-06, "loss": 0.2421, "step": 13162 }, { "epoch": 0.66, "learning_rate": 5.553860664024761e-06, "loss": 0.3616, "step": 13163 }, { "epoch": 0.66, "learning_rate": 5.55241215086749e-06, "loss": 0.1164, "step": 13164 }, { "epoch": 0.66, "learning_rate": 5.550963754031054e-06, "loss": 0.2601, "step": 13165 }, { "epoch": 0.66, "learning_rate": 5.549515473553331e-06, "loss": 0.3206, "step": 13166 }, { "epoch": 0.66, "learning_rate": 5.548067309472205e-06, "loss": 0.2394, "step": 13167 }, { "epoch": 0.66, "learning_rate": 5.546619261825547e-06, "loss": 0.4975, "step": 13168 }, { "epoch": 0.66, "learning_rate": 5.545171330651229e-06, "loss": 0.4013, "step": 13169 }, { "epoch": 0.66, "learning_rate": 5.54372351598712e-06, "loss": 0.3082, "step": 13170 }, { "epoch": 0.66, "learning_rate": 5.542275817871084e-06, "loss": 0.3295, "step": 13171 }, { "epoch": 0.66, "learning_rate": 5.540828236340987e-06, "loss": 0.2577, "step": 13172 }, { "epoch": 0.66, "learning_rate": 5.5393807714346885e-06, "loss": 0.4019, "step": 13173 }, { "epoch": 0.66, "learning_rate": 5.537933423190042e-06, "loss": 0.5349, "step": 13174 }, { "epoch": 0.66, "learning_rate": 5.5364861916448985e-06, "loss": 0.3822, "step": 13175 }, { "epoch": 0.66, "learning_rate": 5.535039076837118e-06, "loss": 0.3043, "step": 13176 }, { "epoch": 0.66, "learning_rate": 5.533592078804542e-06, "loss": 0.4437, "step": 13177 }, { "epoch": 0.66, "learning_rate": 5.532145197585015e-06, "loss": 0.3372, "step": 13178 }, { "epoch": 0.66, "learning_rate": 5.530698433216378e-06, "loss": 0.3373, "step": 13179 }, { "epoch": 0.66, "learning_rate": 5.52925178573647e-06, "loss": 0.3321, "step": 13180 }, { "epoch": 0.66, "learning_rate": 5.527805255183126e-06, "loss": 0.2545, "step": 13181 }, { "epoch": 0.66, "learning_rate": 5.5263588415941795e-06, "loss": 0.1583, "step": 13182 }, { "epoch": 0.66, "learning_rate": 5.524912545007456e-06, "loss": 0.4098, "step": 13183 }, { "epoch": 0.66, "learning_rate": 5.523466365460781e-06, "loss": 0.4967, "step": 13184 }, { "epoch": 0.66, "learning_rate": 5.522020302991983e-06, "loss": 0.5382, "step": 13185 }, { "epoch": 0.66, "learning_rate": 5.520574357638879e-06, "loss": 0.3473, "step": 13186 }, { "epoch": 0.66, "learning_rate": 5.519128529439286e-06, "loss": 0.3544, "step": 13187 }, { "epoch": 0.66, "learning_rate": 5.517682818431017e-06, "loss": 0.3321, "step": 13188 }, { "epoch": 0.66, "learning_rate": 5.516237224651879e-06, "loss": 0.4762, "step": 13189 }, { "epoch": 0.66, "learning_rate": 5.5147917481396895e-06, "loss": 0.2126, "step": 13190 }, { "epoch": 0.66, "learning_rate": 5.513346388932246e-06, "loss": 0.1382, "step": 13191 }, { "epoch": 0.66, "learning_rate": 5.5119011470673515e-06, "loss": 0.42, "step": 13192 }, { "epoch": 0.66, "learning_rate": 5.510456022582803e-06, "loss": 0.343, "step": 13193 }, { "epoch": 0.66, "learning_rate": 5.5090110155163936e-06, "loss": 0.2518, "step": 13194 }, { "epoch": 0.66, "learning_rate": 5.507566125905923e-06, "loss": 0.4409, "step": 13195 }, { "epoch": 0.66, "learning_rate": 5.506121353789177e-06, "loss": 0.2084, "step": 13196 }, { "epoch": 0.66, "learning_rate": 5.50467669920394e-06, "loss": 0.2432, "step": 13197 }, { "epoch": 0.66, "learning_rate": 5.503232162187993e-06, "loss": 0.4364, "step": 13198 }, { "epoch": 0.66, "learning_rate": 5.501787742779123e-06, "loss": 0.3184, "step": 13199 }, { "epoch": 0.66, "learning_rate": 5.500343441015103e-06, "loss": 0.3109, "step": 13200 }, { "epoch": 0.66, "learning_rate": 5.498899256933705e-06, "loss": 0.3193, "step": 13201 }, { "epoch": 0.66, "learning_rate": 5.497455190572705e-06, "loss": 0.3618, "step": 13202 }, { "epoch": 0.66, "learning_rate": 5.496011241969861e-06, "loss": 0.4047, "step": 13203 }, { "epoch": 0.66, "learning_rate": 5.4945674111629475e-06, "loss": 0.2761, "step": 13204 }, { "epoch": 0.66, "learning_rate": 5.493123698189723e-06, "loss": 0.425, "step": 13205 }, { "epoch": 0.66, "learning_rate": 5.4916801030879455e-06, "loss": 0.6481, "step": 13206 }, { "epoch": 0.66, "learning_rate": 5.490236625895369e-06, "loss": 0.3527, "step": 13207 }, { "epoch": 0.66, "learning_rate": 5.488793266649748e-06, "loss": 0.2935, "step": 13208 }, { "epoch": 0.66, "learning_rate": 5.48735002538883e-06, "loss": 0.4812, "step": 13209 }, { "epoch": 0.66, "learning_rate": 5.485906902150362e-06, "loss": 0.3509, "step": 13210 }, { "epoch": 0.66, "learning_rate": 5.484463896972086e-06, "loss": 0.3727, "step": 13211 }, { "epoch": 0.66, "learning_rate": 5.483021009891738e-06, "loss": 0.3597, "step": 13212 }, { "epoch": 0.66, "learning_rate": 5.481578240947065e-06, "loss": 0.2758, "step": 13213 }, { "epoch": 0.66, "learning_rate": 5.480135590175793e-06, "loss": 0.26, "step": 13214 }, { "epoch": 0.66, "learning_rate": 5.478693057615656e-06, "loss": 0.2728, "step": 13215 }, { "epoch": 0.66, "learning_rate": 5.47725064330438e-06, "loss": 0.2247, "step": 13216 }, { "epoch": 0.66, "learning_rate": 5.475808347279685e-06, "loss": 0.2586, "step": 13217 }, { "epoch": 0.66, "learning_rate": 5.474366169579303e-06, "loss": 0.3479, "step": 13218 }, { "epoch": 0.66, "learning_rate": 5.472924110240945e-06, "loss": 0.2826, "step": 13219 }, { "epoch": 0.66, "learning_rate": 5.4714821693023264e-06, "loss": 0.3255, "step": 13220 }, { "epoch": 0.66, "learning_rate": 5.470040346801157e-06, "loss": 0.1607, "step": 13221 }, { "epoch": 0.66, "learning_rate": 5.468598642775155e-06, "loss": 0.3328, "step": 13222 }, { "epoch": 0.66, "learning_rate": 5.467157057262018e-06, "loss": 0.2128, "step": 13223 }, { "epoch": 0.66, "learning_rate": 5.465715590299453e-06, "loss": 0.3602, "step": 13224 }, { "epoch": 0.66, "learning_rate": 5.464274241925157e-06, "loss": 0.4891, "step": 13225 }, { "epoch": 0.66, "learning_rate": 5.4628330121768236e-06, "loss": 0.3157, "step": 13226 }, { "epoch": 0.66, "learning_rate": 5.461391901092155e-06, "loss": 0.4756, "step": 13227 }, { "epoch": 0.66, "learning_rate": 5.459950908708836e-06, "loss": 0.1243, "step": 13228 }, { "epoch": 0.66, "learning_rate": 5.458510035064554e-06, "loss": 0.3857, "step": 13229 }, { "epoch": 0.66, "learning_rate": 5.457069280196994e-06, "loss": 0.5309, "step": 13230 }, { "epoch": 0.66, "learning_rate": 5.455628644143836e-06, "loss": 0.5402, "step": 13231 }, { "epoch": 0.66, "learning_rate": 5.454188126942761e-06, "loss": 0.3998, "step": 13232 }, { "epoch": 0.66, "learning_rate": 5.45274772863144e-06, "loss": 0.2646, "step": 13233 }, { "epoch": 0.66, "learning_rate": 5.451307449247547e-06, "loss": 0.2433, "step": 13234 }, { "epoch": 0.66, "learning_rate": 5.4498672888287455e-06, "loss": 0.5624, "step": 13235 }, { "epoch": 0.66, "learning_rate": 5.448427247412709e-06, "loss": 0.4924, "step": 13236 }, { "epoch": 0.66, "learning_rate": 5.446987325037096e-06, "loss": 0.4296, "step": 13237 }, { "epoch": 0.66, "learning_rate": 5.445547521739568e-06, "loss": 0.304, "step": 13238 }, { "epoch": 0.66, "learning_rate": 5.444107837557775e-06, "loss": 0.3688, "step": 13239 }, { "epoch": 0.66, "learning_rate": 5.442668272529377e-06, "loss": 0.5164, "step": 13240 }, { "epoch": 0.66, "learning_rate": 5.441228826692022e-06, "loss": 0.4173, "step": 13241 }, { "epoch": 0.66, "learning_rate": 5.439789500083357e-06, "loss": 0.1738, "step": 13242 }, { "epoch": 0.66, "learning_rate": 5.438350292741023e-06, "loss": 0.2608, "step": 13243 }, { "epoch": 0.66, "learning_rate": 5.4369112047026595e-06, "loss": 0.4653, "step": 13244 }, { "epoch": 0.66, "learning_rate": 5.435472236005912e-06, "loss": 0.4173, "step": 13245 }, { "epoch": 0.66, "learning_rate": 5.434033386688408e-06, "loss": 0.2248, "step": 13246 }, { "epoch": 0.66, "learning_rate": 5.432594656787781e-06, "loss": 0.2974, "step": 13247 }, { "epoch": 0.66, "learning_rate": 5.431156046341655e-06, "loss": 0.3943, "step": 13248 }, { "epoch": 0.66, "learning_rate": 5.429717555387662e-06, "loss": 0.4218, "step": 13249 }, { "epoch": 0.66, "learning_rate": 5.428279183963418e-06, "loss": 0.5519, "step": 13250 }, { "epoch": 0.66, "learning_rate": 5.4268409321065475e-06, "loss": 0.3823, "step": 13251 }, { "epoch": 0.66, "learning_rate": 5.4254027998546624e-06, "loss": 0.2779, "step": 13252 }, { "epoch": 0.66, "learning_rate": 5.423964787245374e-06, "loss": 0.1975, "step": 13253 }, { "epoch": 0.66, "learning_rate": 5.422526894316294e-06, "loss": 0.3986, "step": 13254 }, { "epoch": 0.66, "learning_rate": 5.4210891211050265e-06, "loss": 0.2799, "step": 13255 }, { "epoch": 0.66, "learning_rate": 5.419651467649173e-06, "loss": 0.2438, "step": 13256 }, { "epoch": 0.66, "learning_rate": 5.418213933986339e-06, "loss": 0.2778, "step": 13257 }, { "epoch": 0.66, "learning_rate": 5.416776520154119e-06, "loss": 0.2912, "step": 13258 }, { "epoch": 0.66, "learning_rate": 5.415339226190106e-06, "loss": 0.1399, "step": 13259 }, { "epoch": 0.66, "learning_rate": 5.413902052131892e-06, "loss": 0.2735, "step": 13260 }, { "epoch": 0.66, "learning_rate": 5.4124649980170615e-06, "loss": 0.2364, "step": 13261 }, { "epoch": 0.66, "learning_rate": 5.411028063883198e-06, "loss": 0.3116, "step": 13262 }, { "epoch": 0.66, "learning_rate": 5.409591249767887e-06, "loss": 0.3815, "step": 13263 }, { "epoch": 0.66, "learning_rate": 5.408154555708707e-06, "loss": 0.4409, "step": 13264 }, { "epoch": 0.66, "learning_rate": 5.406717981743228e-06, "loss": 0.3399, "step": 13265 }, { "epoch": 0.66, "learning_rate": 5.405281527909026e-06, "loss": 0.3504, "step": 13266 }, { "epoch": 0.66, "learning_rate": 5.403845194243662e-06, "loss": 0.4861, "step": 13267 }, { "epoch": 0.66, "learning_rate": 5.402408980784713e-06, "loss": 0.3205, "step": 13268 }, { "epoch": 0.66, "learning_rate": 5.400972887569733e-06, "loss": 0.3943, "step": 13269 }, { "epoch": 0.66, "learning_rate": 5.399536914636284e-06, "loss": 0.2829, "step": 13270 }, { "epoch": 0.66, "learning_rate": 5.3981010620219185e-06, "loss": 0.1615, "step": 13271 }, { "epoch": 0.66, "learning_rate": 5.396665329764196e-06, "loss": 0.323, "step": 13272 }, { "epoch": 0.66, "learning_rate": 5.395229717900662e-06, "loss": 0.5202, "step": 13273 }, { "epoch": 0.66, "learning_rate": 5.393794226468865e-06, "loss": 0.2472, "step": 13274 }, { "epoch": 0.66, "learning_rate": 5.392358855506346e-06, "loss": 0.2313, "step": 13275 }, { "epoch": 0.66, "learning_rate": 5.390923605050641e-06, "loss": 0.4994, "step": 13276 }, { "epoch": 0.66, "learning_rate": 5.3894884751392976e-06, "loss": 0.4398, "step": 13277 }, { "epoch": 0.66, "learning_rate": 5.388053465809844e-06, "loss": 0.0946, "step": 13278 }, { "epoch": 0.66, "learning_rate": 5.386618577099811e-06, "loss": 0.41, "step": 13279 }, { "epoch": 0.66, "learning_rate": 5.385183809046728e-06, "loss": 0.2425, "step": 13280 }, { "epoch": 0.66, "learning_rate": 5.3837491616881174e-06, "loss": 0.4352, "step": 13281 }, { "epoch": 0.66, "learning_rate": 5.3823146350615005e-06, "loss": 0.2129, "step": 13282 }, { "epoch": 0.66, "learning_rate": 5.380880229204398e-06, "loss": 0.3778, "step": 13283 }, { "epoch": 0.66, "learning_rate": 5.379445944154322e-06, "loss": 0.5498, "step": 13284 }, { "epoch": 0.66, "learning_rate": 5.378011779948782e-06, "loss": 0.3902, "step": 13285 }, { "epoch": 0.66, "learning_rate": 5.376577736625295e-06, "loss": 0.1514, "step": 13286 }, { "epoch": 0.66, "learning_rate": 5.37514381422136e-06, "loss": 0.2652, "step": 13287 }, { "epoch": 0.66, "learning_rate": 5.373710012774481e-06, "loss": 0.177, "step": 13288 }, { "epoch": 0.66, "learning_rate": 5.372276332322158e-06, "loss": 0.4605, "step": 13289 }, { "epoch": 0.66, "learning_rate": 5.3708427729018835e-06, "loss": 0.3036, "step": 13290 }, { "epoch": 0.66, "learning_rate": 5.369409334551154e-06, "loss": 0.2341, "step": 13291 }, { "epoch": 0.66, "learning_rate": 5.367976017307461e-06, "loss": 0.3355, "step": 13292 }, { "epoch": 0.66, "learning_rate": 5.366542821208288e-06, "loss": 0.5603, "step": 13293 }, { "epoch": 0.66, "learning_rate": 5.365109746291115e-06, "loss": 0.2524, "step": 13294 }, { "epoch": 0.66, "learning_rate": 5.363676792593428e-06, "loss": 0.571, "step": 13295 }, { "epoch": 0.66, "learning_rate": 5.362243960152704e-06, "loss": 0.475, "step": 13296 }, { "epoch": 0.66, "learning_rate": 5.360811249006415e-06, "loss": 0.2565, "step": 13297 }, { "epoch": 0.66, "learning_rate": 5.3593786591920305e-06, "loss": 0.1762, "step": 13298 }, { "epoch": 0.66, "learning_rate": 5.357946190747015e-06, "loss": 0.3294, "step": 13299 }, { "epoch": 0.66, "learning_rate": 5.356513843708842e-06, "loss": 0.4887, "step": 13300 }, { "epoch": 0.66, "learning_rate": 5.355081618114967e-06, "loss": 0.5563, "step": 13301 }, { "epoch": 0.66, "learning_rate": 5.353649514002847e-06, "loss": 0.2219, "step": 13302 }, { "epoch": 0.66, "learning_rate": 5.35221753140994e-06, "loss": 0.4968, "step": 13303 }, { "epoch": 0.66, "learning_rate": 5.350785670373696e-06, "loss": 0.2539, "step": 13304 }, { "epoch": 0.66, "learning_rate": 5.349353930931564e-06, "loss": 0.2907, "step": 13305 }, { "epoch": 0.66, "learning_rate": 5.347922313120987e-06, "loss": 0.4402, "step": 13306 }, { "epoch": 0.66, "learning_rate": 5.346490816979409e-06, "loss": 0.5916, "step": 13307 }, { "epoch": 0.66, "learning_rate": 5.345059442544265e-06, "loss": 0.3021, "step": 13308 }, { "epoch": 0.66, "learning_rate": 5.343628189852998e-06, "loss": 0.1954, "step": 13309 }, { "epoch": 0.66, "learning_rate": 5.3421970589430395e-06, "loss": 0.2818, "step": 13310 }, { "epoch": 0.66, "learning_rate": 5.340766049851813e-06, "loss": 0.2501, "step": 13311 }, { "epoch": 0.66, "learning_rate": 5.339335162616745e-06, "loss": 0.3237, "step": 13312 }, { "epoch": 0.66, "learning_rate": 5.337904397275266e-06, "loss": 0.7396, "step": 13313 }, { "epoch": 0.66, "learning_rate": 5.3364737538647915e-06, "loss": 0.3133, "step": 13314 }, { "epoch": 0.66, "learning_rate": 5.335043232422737e-06, "loss": 0.3686, "step": 13315 }, { "epoch": 0.66, "learning_rate": 5.333612832986518e-06, "loss": 0.2663, "step": 13316 }, { "epoch": 0.66, "learning_rate": 5.332182555593538e-06, "loss": 0.3058, "step": 13317 }, { "epoch": 0.67, "learning_rate": 5.330752400281214e-06, "loss": 0.1795, "step": 13318 }, { "epoch": 0.67, "learning_rate": 5.3293223670869465e-06, "loss": 0.4013, "step": 13319 }, { "epoch": 0.67, "learning_rate": 5.3278924560481335e-06, "loss": 0.4488, "step": 13320 }, { "epoch": 0.67, "learning_rate": 5.326462667202175e-06, "loss": 0.386, "step": 13321 }, { "epoch": 0.67, "learning_rate": 5.3250330005864595e-06, "loss": 0.4443, "step": 13322 }, { "epoch": 0.67, "learning_rate": 5.323603456238386e-06, "loss": 0.3706, "step": 13323 }, { "epoch": 0.67, "learning_rate": 5.3221740341953394e-06, "loss": 0.301, "step": 13324 }, { "epoch": 0.67, "learning_rate": 5.320744734494705e-06, "loss": 0.3449, "step": 13325 }, { "epoch": 0.67, "learning_rate": 5.319315557173863e-06, "loss": 0.2875, "step": 13326 }, { "epoch": 0.67, "learning_rate": 5.317886502270193e-06, "loss": 0.4304, "step": 13327 }, { "epoch": 0.67, "learning_rate": 5.316457569821063e-06, "loss": 0.2986, "step": 13328 }, { "epoch": 0.67, "learning_rate": 5.3150287598638565e-06, "loss": 0.2962, "step": 13329 }, { "epoch": 0.67, "learning_rate": 5.313600072435936e-06, "loss": 0.5157, "step": 13330 }, { "epoch": 0.67, "learning_rate": 5.3121715075746685e-06, "loss": 0.1904, "step": 13331 }, { "epoch": 0.67, "learning_rate": 5.310743065317414e-06, "loss": 0.2926, "step": 13332 }, { "epoch": 0.67, "learning_rate": 5.309314745701533e-06, "loss": 0.4032, "step": 13333 }, { "epoch": 0.67, "learning_rate": 5.307886548764381e-06, "loss": 0.214, "step": 13334 }, { "epoch": 0.67, "learning_rate": 5.306458474543308e-06, "loss": 0.2953, "step": 13335 }, { "epoch": 0.67, "learning_rate": 5.305030523075668e-06, "loss": 0.4703, "step": 13336 }, { "epoch": 0.67, "learning_rate": 5.303602694398807e-06, "loss": 0.381, "step": 13337 }, { "epoch": 0.67, "learning_rate": 5.302174988550066e-06, "loss": 0.3459, "step": 13338 }, { "epoch": 0.67, "learning_rate": 5.300747405566785e-06, "loss": 0.355, "step": 13339 }, { "epoch": 0.67, "learning_rate": 5.299319945486296e-06, "loss": 0.4529, "step": 13340 }, { "epoch": 0.67, "learning_rate": 5.297892608345943e-06, "loss": 0.1675, "step": 13341 }, { "epoch": 0.67, "learning_rate": 5.296465394183048e-06, "loss": 0.264, "step": 13342 }, { "epoch": 0.67, "learning_rate": 5.295038303034942e-06, "loss": 0.499, "step": 13343 }, { "epoch": 0.67, "learning_rate": 5.293611334938941e-06, "loss": 0.3636, "step": 13344 }, { "epoch": 0.67, "learning_rate": 5.292184489932377e-06, "loss": 0.3611, "step": 13345 }, { "epoch": 0.67, "learning_rate": 5.290757768052561e-06, "loss": 0.3595, "step": 13346 }, { "epoch": 0.67, "learning_rate": 5.289331169336807e-06, "loss": 0.4163, "step": 13347 }, { "epoch": 0.67, "learning_rate": 5.287904693822426e-06, "loss": 0.3777, "step": 13348 }, { "epoch": 0.67, "learning_rate": 5.2864783415467236e-06, "loss": 0.193, "step": 13349 }, { "epoch": 0.67, "learning_rate": 5.285052112547009e-06, "loss": 0.2606, "step": 13350 }, { "epoch": 0.67, "learning_rate": 5.2836260068605815e-06, "loss": 0.3982, "step": 13351 }, { "epoch": 0.67, "learning_rate": 5.2822000245247396e-06, "loss": 0.3962, "step": 13352 }, { "epoch": 0.67, "learning_rate": 5.280774165576774e-06, "loss": 0.2994, "step": 13353 }, { "epoch": 0.67, "learning_rate": 5.279348430053981e-06, "loss": 0.2724, "step": 13354 }, { "epoch": 0.67, "learning_rate": 5.277922817993647e-06, "loss": 0.4452, "step": 13355 }, { "epoch": 0.67, "learning_rate": 5.276497329433057e-06, "loss": 0.3261, "step": 13356 }, { "epoch": 0.67, "learning_rate": 5.275071964409493e-06, "loss": 0.2383, "step": 13357 }, { "epoch": 0.67, "learning_rate": 5.2736467229602275e-06, "loss": 0.5069, "step": 13358 }, { "epoch": 0.67, "learning_rate": 5.2722216051225475e-06, "loss": 0.2922, "step": 13359 }, { "epoch": 0.67, "learning_rate": 5.270796610933717e-06, "loss": 0.2593, "step": 13360 }, { "epoch": 0.67, "learning_rate": 5.269371740431009e-06, "loss": 0.3573, "step": 13361 }, { "epoch": 0.67, "learning_rate": 5.267946993651685e-06, "loss": 0.3696, "step": 13362 }, { "epoch": 0.67, "learning_rate": 5.266522370633007e-06, "loss": 0.4533, "step": 13363 }, { "epoch": 0.67, "learning_rate": 5.265097871412241e-06, "loss": 0.3618, "step": 13364 }, { "epoch": 0.67, "learning_rate": 5.263673496026638e-06, "loss": 0.3793, "step": 13365 }, { "epoch": 0.67, "learning_rate": 5.262249244513451e-06, "loss": 0.3022, "step": 13366 }, { "epoch": 0.67, "learning_rate": 5.260825116909926e-06, "loss": 0.3161, "step": 13367 }, { "epoch": 0.67, "learning_rate": 5.259401113253316e-06, "loss": 0.4842, "step": 13368 }, { "epoch": 0.67, "learning_rate": 5.257977233580862e-06, "loss": 0.3072, "step": 13369 }, { "epoch": 0.67, "learning_rate": 5.256553477929802e-06, "loss": 0.4138, "step": 13370 }, { "epoch": 0.67, "learning_rate": 5.255129846337373e-06, "loss": 0.3563, "step": 13371 }, { "epoch": 0.67, "learning_rate": 5.253706338840805e-06, "loss": 0.4529, "step": 13372 }, { "epoch": 0.67, "learning_rate": 5.252282955477335e-06, "loss": 0.3853, "step": 13373 }, { "epoch": 0.67, "learning_rate": 5.250859696284185e-06, "loss": 0.5908, "step": 13374 }, { "epoch": 0.67, "learning_rate": 5.24943656129858e-06, "loss": 0.2409, "step": 13375 }, { "epoch": 0.67, "learning_rate": 5.248013550557738e-06, "loss": 0.6007, "step": 13376 }, { "epoch": 0.67, "learning_rate": 5.246590664098879e-06, "loss": 0.5085, "step": 13377 }, { "epoch": 0.67, "learning_rate": 5.245167901959216e-06, "loss": 0.4301, "step": 13378 }, { "epoch": 0.67, "learning_rate": 5.243745264175958e-06, "loss": 0.0635, "step": 13379 }, { "epoch": 0.67, "learning_rate": 5.242322750786313e-06, "loss": 0.2748, "step": 13380 }, { "epoch": 0.67, "learning_rate": 5.240900361827481e-06, "loss": 0.3019, "step": 13381 }, { "epoch": 0.67, "learning_rate": 5.239478097336671e-06, "loss": 0.5861, "step": 13382 }, { "epoch": 0.67, "learning_rate": 5.238055957351076e-06, "loss": 0.2855, "step": 13383 }, { "epoch": 0.67, "learning_rate": 5.236633941907891e-06, "loss": 0.386, "step": 13384 }, { "epoch": 0.67, "learning_rate": 5.235212051044302e-06, "loss": 0.3374, "step": 13385 }, { "epoch": 0.67, "learning_rate": 5.233790284797504e-06, "loss": 0.3759, "step": 13386 }, { "epoch": 0.67, "learning_rate": 5.2323686432046795e-06, "loss": 0.4378, "step": 13387 }, { "epoch": 0.67, "learning_rate": 5.2309471263030085e-06, "loss": 0.483, "step": 13388 }, { "epoch": 0.67, "learning_rate": 5.22952573412967e-06, "loss": 0.4637, "step": 13389 }, { "epoch": 0.67, "learning_rate": 5.228104466721833e-06, "loss": 0.2749, "step": 13390 }, { "epoch": 0.67, "learning_rate": 5.226683324116677e-06, "loss": 0.3015, "step": 13391 }, { "epoch": 0.67, "learning_rate": 5.225262306351367e-06, "loss": 0.2415, "step": 13392 }, { "epoch": 0.67, "learning_rate": 5.223841413463069e-06, "loss": 0.3289, "step": 13393 }, { "epoch": 0.67, "learning_rate": 5.222420645488942e-06, "loss": 0.5646, "step": 13394 }, { "epoch": 0.67, "learning_rate": 5.221000002466141e-06, "loss": 0.3781, "step": 13395 }, { "epoch": 0.67, "learning_rate": 5.219579484431829e-06, "loss": 0.2953, "step": 13396 }, { "epoch": 0.67, "learning_rate": 5.218159091423156e-06, "loss": 0.2665, "step": 13397 }, { "epoch": 0.67, "learning_rate": 5.216738823477269e-06, "loss": 0.3106, "step": 13398 }, { "epoch": 0.67, "learning_rate": 5.215318680631311e-06, "loss": 0.4172, "step": 13399 }, { "epoch": 0.67, "learning_rate": 5.213898662922425e-06, "loss": 0.4504, "step": 13400 }, { "epoch": 0.67, "learning_rate": 5.2124787703877526e-06, "loss": 0.5139, "step": 13401 }, { "epoch": 0.67, "learning_rate": 5.211059003064429e-06, "loss": 0.3557, "step": 13402 }, { "epoch": 0.67, "learning_rate": 5.209639360989583e-06, "loss": 0.3549, "step": 13403 }, { "epoch": 0.67, "learning_rate": 5.208219844200346e-06, "loss": 0.441, "step": 13404 }, { "epoch": 0.67, "learning_rate": 5.206800452733844e-06, "loss": 0.2016, "step": 13405 }, { "epoch": 0.67, "learning_rate": 5.205381186627196e-06, "loss": 0.2757, "step": 13406 }, { "epoch": 0.67, "learning_rate": 5.2039620459175255e-06, "loss": 0.4349, "step": 13407 }, { "epoch": 0.67, "learning_rate": 5.202543030641941e-06, "loss": 0.2818, "step": 13408 }, { "epoch": 0.67, "learning_rate": 5.201124140837565e-06, "loss": 0.3551, "step": 13409 }, { "epoch": 0.67, "learning_rate": 5.199705376541499e-06, "loss": 0.5098, "step": 13410 }, { "epoch": 0.67, "learning_rate": 5.198286737790854e-06, "loss": 0.3646, "step": 13411 }, { "epoch": 0.67, "learning_rate": 5.196868224622731e-06, "loss": 0.255, "step": 13412 }, { "epoch": 0.67, "learning_rate": 5.195449837074223e-06, "loss": 0.2932, "step": 13413 }, { "epoch": 0.67, "learning_rate": 5.194031575182435e-06, "loss": 0.3122, "step": 13414 }, { "epoch": 0.67, "learning_rate": 5.192613438984457e-06, "loss": 0.5743, "step": 13415 }, { "epoch": 0.67, "learning_rate": 5.191195428517379e-06, "loss": 0.2216, "step": 13416 }, { "epoch": 0.67, "learning_rate": 5.1897775438182815e-06, "loss": 0.27, "step": 13417 }, { "epoch": 0.67, "learning_rate": 5.188359784924256e-06, "loss": 0.471, "step": 13418 }, { "epoch": 0.67, "learning_rate": 5.186942151872378e-06, "loss": 0.0929, "step": 13419 }, { "epoch": 0.67, "learning_rate": 5.185524644699724e-06, "loss": 0.2319, "step": 13420 }, { "epoch": 0.67, "learning_rate": 5.184107263443369e-06, "loss": 0.3735, "step": 13421 }, { "epoch": 0.67, "learning_rate": 5.182690008140374e-06, "loss": 0.3466, "step": 13422 }, { "epoch": 0.67, "learning_rate": 5.181272878827819e-06, "loss": 0.5876, "step": 13423 }, { "epoch": 0.67, "learning_rate": 5.179855875542759e-06, "loss": 0.3597, "step": 13424 }, { "epoch": 0.67, "learning_rate": 5.178438998322255e-06, "loss": 0.5872, "step": 13425 }, { "epoch": 0.67, "learning_rate": 5.177022247203364e-06, "loss": 0.3872, "step": 13426 }, { "epoch": 0.67, "learning_rate": 5.1756056222231385e-06, "loss": 0.3754, "step": 13427 }, { "epoch": 0.67, "learning_rate": 5.174189123418632e-06, "loss": 0.3975, "step": 13428 }, { "epoch": 0.67, "learning_rate": 5.172772750826886e-06, "loss": 0.6008, "step": 13429 }, { "epoch": 0.67, "learning_rate": 5.171356504484946e-06, "loss": 0.4421, "step": 13430 }, { "epoch": 0.67, "learning_rate": 5.169940384429848e-06, "loss": 0.3006, "step": 13431 }, { "epoch": 0.67, "learning_rate": 5.168524390698638e-06, "loss": 0.5049, "step": 13432 }, { "epoch": 0.67, "learning_rate": 5.1671085233283444e-06, "loss": 0.5745, "step": 13433 }, { "epoch": 0.67, "learning_rate": 5.165692782355998e-06, "loss": 0.4802, "step": 13434 }, { "epoch": 0.67, "learning_rate": 5.164277167818623e-06, "loss": 0.3132, "step": 13435 }, { "epoch": 0.67, "learning_rate": 5.162861679753242e-06, "loss": 0.3866, "step": 13436 }, { "epoch": 0.67, "learning_rate": 5.161446318196883e-06, "loss": 0.3061, "step": 13437 }, { "epoch": 0.67, "learning_rate": 5.160031083186559e-06, "loss": 0.302, "step": 13438 }, { "epoch": 0.67, "learning_rate": 5.15861597475928e-06, "loss": 0.2636, "step": 13439 }, { "epoch": 0.67, "learning_rate": 5.157200992952058e-06, "loss": 0.5815, "step": 13440 }, { "epoch": 0.67, "learning_rate": 5.155786137801904e-06, "loss": 0.3442, "step": 13441 }, { "epoch": 0.67, "learning_rate": 5.154371409345817e-06, "loss": 0.3289, "step": 13442 }, { "epoch": 0.67, "learning_rate": 5.152956807620802e-06, "loss": 0.4578, "step": 13443 }, { "epoch": 0.67, "learning_rate": 5.1515423326638515e-06, "loss": 0.3008, "step": 13444 }, { "epoch": 0.67, "learning_rate": 5.150127984511957e-06, "loss": 0.4257, "step": 13445 }, { "epoch": 0.67, "learning_rate": 5.148713763202118e-06, "loss": 0.2667, "step": 13446 }, { "epoch": 0.67, "learning_rate": 5.147299668771315e-06, "loss": 0.3023, "step": 13447 }, { "epoch": 0.67, "learning_rate": 5.145885701256535e-06, "loss": 0.4483, "step": 13448 }, { "epoch": 0.67, "learning_rate": 5.1444718606947555e-06, "loss": 0.496, "step": 13449 }, { "epoch": 0.67, "learning_rate": 5.143058147122954e-06, "loss": 0.4053, "step": 13450 }, { "epoch": 0.67, "learning_rate": 5.1416445605781075e-06, "loss": 0.2363, "step": 13451 }, { "epoch": 0.67, "learning_rate": 5.140231101097181e-06, "loss": 0.3297, "step": 13452 }, { "epoch": 0.67, "learning_rate": 5.138817768717148e-06, "loss": 0.5088, "step": 13453 }, { "epoch": 0.67, "learning_rate": 5.137404563474965e-06, "loss": 0.4257, "step": 13454 }, { "epoch": 0.67, "learning_rate": 5.1359914854076e-06, "loss": 0.3237, "step": 13455 }, { "epoch": 0.67, "learning_rate": 5.134578534552006e-06, "loss": 0.3151, "step": 13456 }, { "epoch": 0.67, "learning_rate": 5.1331657109451394e-06, "loss": 0.2053, "step": 13457 }, { "epoch": 0.67, "learning_rate": 5.131753014623948e-06, "loss": 0.3887, "step": 13458 }, { "epoch": 0.67, "learning_rate": 5.130340445625377e-06, "loss": 0.2187, "step": 13459 }, { "epoch": 0.67, "learning_rate": 5.128928003986378e-06, "loss": 0.3884, "step": 13460 }, { "epoch": 0.67, "learning_rate": 5.127515689743887e-06, "loss": 0.2307, "step": 13461 }, { "epoch": 0.67, "learning_rate": 5.126103502934841e-06, "loss": 0.4515, "step": 13462 }, { "epoch": 0.67, "learning_rate": 5.12469144359617e-06, "loss": 0.1264, "step": 13463 }, { "epoch": 0.67, "learning_rate": 5.123279511764813e-06, "loss": 0.3391, "step": 13464 }, { "epoch": 0.67, "learning_rate": 5.121867707477693e-06, "loss": 0.4996, "step": 13465 }, { "epoch": 0.67, "learning_rate": 5.120456030771733e-06, "loss": 0.1614, "step": 13466 }, { "epoch": 0.67, "learning_rate": 5.119044481683855e-06, "loss": 0.3302, "step": 13467 }, { "epoch": 0.67, "learning_rate": 5.1176330602509725e-06, "loss": 0.3606, "step": 13468 }, { "epoch": 0.67, "learning_rate": 5.116221766510006e-06, "loss": 0.3951, "step": 13469 }, { "epoch": 0.67, "learning_rate": 5.114810600497862e-06, "loss": 0.3644, "step": 13470 }, { "epoch": 0.67, "learning_rate": 5.1133995622514484e-06, "loss": 0.3706, "step": 13471 }, { "epoch": 0.67, "learning_rate": 5.111988651807667e-06, "loss": 0.3198, "step": 13472 }, { "epoch": 0.67, "learning_rate": 5.1105778692034225e-06, "loss": 0.4412, "step": 13473 }, { "epoch": 0.67, "learning_rate": 5.10916721447561e-06, "loss": 0.1926, "step": 13474 }, { "epoch": 0.67, "learning_rate": 5.1077566876611225e-06, "loss": 0.1336, "step": 13475 }, { "epoch": 0.67, "learning_rate": 5.106346288796853e-06, "loss": 0.3097, "step": 13476 }, { "epoch": 0.67, "learning_rate": 5.104936017919685e-06, "loss": 0.463, "step": 13477 }, { "epoch": 0.67, "learning_rate": 5.103525875066504e-06, "loss": 0.3128, "step": 13478 }, { "epoch": 0.67, "learning_rate": 5.1021158602741915e-06, "loss": 0.4521, "step": 13479 }, { "epoch": 0.67, "learning_rate": 5.1007059735796235e-06, "loss": 0.4114, "step": 13480 }, { "epoch": 0.67, "learning_rate": 5.09929621501967e-06, "loss": 0.2732, "step": 13481 }, { "epoch": 0.67, "learning_rate": 5.0978865846312085e-06, "loss": 0.3164, "step": 13482 }, { "epoch": 0.67, "learning_rate": 5.096477082451103e-06, "loss": 0.1954, "step": 13483 }, { "epoch": 0.67, "learning_rate": 5.095067708516217e-06, "loss": 0.3079, "step": 13484 }, { "epoch": 0.67, "learning_rate": 5.0936584628634115e-06, "loss": 0.3073, "step": 13485 }, { "epoch": 0.67, "learning_rate": 5.092249345529539e-06, "loss": 0.3477, "step": 13486 }, { "epoch": 0.67, "learning_rate": 5.090840356551459e-06, "loss": 0.5487, "step": 13487 }, { "epoch": 0.67, "learning_rate": 5.089431495966021e-06, "loss": 0.5823, "step": 13488 }, { "epoch": 0.67, "learning_rate": 5.08802276381007e-06, "loss": 0.2551, "step": 13489 }, { "epoch": 0.67, "learning_rate": 5.08661416012045e-06, "loss": 0.2447, "step": 13490 }, { "epoch": 0.67, "learning_rate": 5.085205684933998e-06, "loss": 0.3257, "step": 13491 }, { "epoch": 0.67, "learning_rate": 5.083797338287557e-06, "loss": 0.2436, "step": 13492 }, { "epoch": 0.67, "learning_rate": 5.082389120217959e-06, "loss": 0.1121, "step": 13493 }, { "epoch": 0.67, "learning_rate": 5.080981030762031e-06, "loss": 0.554, "step": 13494 }, { "epoch": 0.67, "learning_rate": 5.0795730699565984e-06, "loss": 0.3515, "step": 13495 }, { "epoch": 0.67, "learning_rate": 5.0781652378384906e-06, "loss": 0.2715, "step": 13496 }, { "epoch": 0.67, "learning_rate": 5.076757534444525e-06, "loss": 0.2392, "step": 13497 }, { "epoch": 0.67, "learning_rate": 5.075349959811519e-06, "loss": 0.4395, "step": 13498 }, { "epoch": 0.67, "learning_rate": 5.0739425139762844e-06, "loss": 0.351, "step": 13499 }, { "epoch": 0.67, "learning_rate": 5.07253519697563e-06, "loss": 0.4471, "step": 13500 }, { "epoch": 0.67, "learning_rate": 5.071128008846364e-06, "loss": 0.4032, "step": 13501 }, { "epoch": 0.67, "learning_rate": 5.0697209496252895e-06, "loss": 0.6206, "step": 13502 }, { "epoch": 0.67, "learning_rate": 5.0683140193492075e-06, "loss": 0.2222, "step": 13503 }, { "epoch": 0.67, "learning_rate": 5.0669072180549075e-06, "loss": 0.2959, "step": 13504 }, { "epoch": 0.67, "learning_rate": 5.065500545779191e-06, "loss": 0.3241, "step": 13505 }, { "epoch": 0.67, "learning_rate": 5.064094002558847e-06, "loss": 0.1596, "step": 13506 }, { "epoch": 0.67, "learning_rate": 5.062687588430658e-06, "loss": 0.6015, "step": 13507 }, { "epoch": 0.67, "learning_rate": 5.061281303431409e-06, "loss": 0.3733, "step": 13508 }, { "epoch": 0.67, "learning_rate": 5.059875147597875e-06, "loss": 0.2248, "step": 13509 }, { "epoch": 0.67, "learning_rate": 5.05846912096684e-06, "loss": 0.3275, "step": 13510 }, { "epoch": 0.67, "learning_rate": 5.057063223575073e-06, "loss": 0.2013, "step": 13511 }, { "epoch": 0.67, "learning_rate": 5.055657455459342e-06, "loss": 0.4892, "step": 13512 }, { "epoch": 0.67, "learning_rate": 5.054251816656412e-06, "loss": 0.3261, "step": 13513 }, { "epoch": 0.67, "learning_rate": 5.052846307203052e-06, "loss": 0.3251, "step": 13514 }, { "epoch": 0.67, "learning_rate": 5.051440927136015e-06, "loss": 0.3858, "step": 13515 }, { "epoch": 0.67, "learning_rate": 5.050035676492062e-06, "loss": 0.5123, "step": 13516 }, { "epoch": 0.67, "learning_rate": 5.048630555307941e-06, "loss": 0.3781, "step": 13517 }, { "epoch": 0.67, "learning_rate": 5.047225563620399e-06, "loss": 0.2646, "step": 13518 }, { "epoch": 0.68, "learning_rate": 5.04582070146619e-06, "loss": 0.4471, "step": 13519 }, { "epoch": 0.68, "learning_rate": 5.044415968882051e-06, "loss": 0.2355, "step": 13520 }, { "epoch": 0.68, "learning_rate": 5.043011365904722e-06, "loss": 0.4457, "step": 13521 }, { "epoch": 0.68, "learning_rate": 5.0416068925709385e-06, "loss": 0.2534, "step": 13522 }, { "epoch": 0.68, "learning_rate": 5.0402025489174315e-06, "loss": 0.3455, "step": 13523 }, { "epoch": 0.68, "learning_rate": 5.03879833498093e-06, "loss": 0.4725, "step": 13524 }, { "epoch": 0.68, "learning_rate": 5.037394250798162e-06, "loss": 0.3255, "step": 13525 }, { "epoch": 0.68, "learning_rate": 5.035990296405846e-06, "loss": 0.3283, "step": 13526 }, { "epoch": 0.68, "learning_rate": 5.034586471840699e-06, "loss": 0.2563, "step": 13527 }, { "epoch": 0.68, "learning_rate": 5.033182777139444e-06, "loss": 0.2966, "step": 13528 }, { "epoch": 0.68, "learning_rate": 5.031779212338788e-06, "loss": 0.4437, "step": 13529 }, { "epoch": 0.68, "learning_rate": 5.030375777475438e-06, "loss": 0.0413, "step": 13530 }, { "epoch": 0.68, "learning_rate": 5.028972472586101e-06, "loss": 0.4237, "step": 13531 }, { "epoch": 0.68, "learning_rate": 5.027569297707474e-06, "loss": 0.4033, "step": 13532 }, { "epoch": 0.68, "learning_rate": 5.0261662528762634e-06, "loss": 0.2361, "step": 13533 }, { "epoch": 0.68, "learning_rate": 5.02476333812916e-06, "loss": 0.5401, "step": 13534 }, { "epoch": 0.68, "learning_rate": 5.023360553502856e-06, "loss": 0.2558, "step": 13535 }, { "epoch": 0.68, "learning_rate": 5.021957899034033e-06, "loss": 0.505, "step": 13536 }, { "epoch": 0.68, "learning_rate": 5.020555374759386e-06, "loss": 0.2289, "step": 13537 }, { "epoch": 0.68, "learning_rate": 5.0191529807155915e-06, "loss": 0.0834, "step": 13538 }, { "epoch": 0.68, "learning_rate": 5.0177507169393265e-06, "loss": 0.3168, "step": 13539 }, { "epoch": 0.68, "learning_rate": 5.016348583467267e-06, "loss": 0.4301, "step": 13540 }, { "epoch": 0.68, "learning_rate": 5.014946580336077e-06, "loss": 0.5121, "step": 13541 }, { "epoch": 0.68, "learning_rate": 5.013544707582436e-06, "loss": 0.4807, "step": 13542 }, { "epoch": 0.68, "learning_rate": 5.012142965243001e-06, "loss": 0.2929, "step": 13543 }, { "epoch": 0.68, "learning_rate": 5.010741353354434e-06, "loss": 0.4852, "step": 13544 }, { "epoch": 0.68, "learning_rate": 5.009339871953387e-06, "loss": 0.4878, "step": 13545 }, { "epoch": 0.68, "learning_rate": 5.0079385210765254e-06, "loss": 0.2634, "step": 13546 }, { "epoch": 0.68, "learning_rate": 5.006537300760491e-06, "loss": 0.2705, "step": 13547 }, { "epoch": 0.68, "learning_rate": 5.005136211041934e-06, "loss": 0.4954, "step": 13548 }, { "epoch": 0.68, "learning_rate": 5.003735251957498e-06, "loss": 0.2952, "step": 13549 }, { "epoch": 0.68, "learning_rate": 5.002334423543822e-06, "loss": 0.3729, "step": 13550 }, { "epoch": 0.68, "learning_rate": 5.000933725837544e-06, "loss": 0.4847, "step": 13551 }, { "epoch": 0.68, "learning_rate": 4.999533158875296e-06, "loss": 0.386, "step": 13552 }, { "epoch": 0.68, "learning_rate": 4.9981327226937085e-06, "loss": 0.3709, "step": 13553 }, { "epoch": 0.68, "learning_rate": 4.996732417329405e-06, "loss": 0.4359, "step": 13554 }, { "epoch": 0.68, "learning_rate": 4.995332242819017e-06, "loss": 0.286, "step": 13555 }, { "epoch": 0.68, "learning_rate": 4.993932199199157e-06, "loss": 0.3802, "step": 13556 }, { "epoch": 0.68, "learning_rate": 4.992532286506445e-06, "loss": 0.323, "step": 13557 }, { "epoch": 0.68, "learning_rate": 4.991132504777493e-06, "loss": 0.1991, "step": 13558 }, { "epoch": 0.68, "learning_rate": 4.989732854048904e-06, "loss": 0.5268, "step": 13559 }, { "epoch": 0.68, "learning_rate": 4.988333334357296e-06, "loss": 0.3665, "step": 13560 }, { "epoch": 0.68, "learning_rate": 4.986933945739264e-06, "loss": 0.3362, "step": 13561 }, { "epoch": 0.68, "learning_rate": 4.985534688231409e-06, "loss": 0.1819, "step": 13562 }, { "epoch": 0.68, "learning_rate": 4.984135561870327e-06, "loss": 0.4324, "step": 13563 }, { "epoch": 0.68, "learning_rate": 4.982736566692605e-06, "loss": 0.6445, "step": 13564 }, { "epoch": 0.68, "learning_rate": 4.9813377027348416e-06, "loss": 0.2783, "step": 13565 }, { "epoch": 0.68, "learning_rate": 4.979938970033618e-06, "loss": 0.3313, "step": 13566 }, { "epoch": 0.68, "learning_rate": 4.978540368625514e-06, "loss": 0.2104, "step": 13567 }, { "epoch": 0.68, "learning_rate": 4.9771418985471065e-06, "loss": 0.4545, "step": 13568 }, { "epoch": 0.68, "learning_rate": 4.975743559834977e-06, "loss": 0.4516, "step": 13569 }, { "epoch": 0.68, "learning_rate": 4.974345352525694e-06, "loss": 0.3543, "step": 13570 }, { "epoch": 0.68, "learning_rate": 4.972947276655827e-06, "loss": 0.6174, "step": 13571 }, { "epoch": 0.68, "learning_rate": 4.97154933226194e-06, "loss": 0.2758, "step": 13572 }, { "epoch": 0.68, "learning_rate": 4.9701515193805925e-06, "loss": 0.3452, "step": 13573 }, { "epoch": 0.68, "learning_rate": 4.9687538380483455e-06, "loss": 0.2789, "step": 13574 }, { "epoch": 0.68, "learning_rate": 4.967356288301751e-06, "loss": 0.4663, "step": 13575 }, { "epoch": 0.68, "learning_rate": 4.965958870177362e-06, "loss": 0.6601, "step": 13576 }, { "epoch": 0.68, "learning_rate": 4.964561583711721e-06, "loss": 0.5847, "step": 13577 }, { "epoch": 0.68, "learning_rate": 4.963164428941382e-06, "loss": 0.3586, "step": 13578 }, { "epoch": 0.68, "learning_rate": 4.9617674059028786e-06, "loss": 0.3529, "step": 13579 }, { "epoch": 0.68, "learning_rate": 4.960370514632749e-06, "loss": 0.2802, "step": 13580 }, { "epoch": 0.68, "learning_rate": 4.95897375516753e-06, "loss": 0.448, "step": 13581 }, { "epoch": 0.68, "learning_rate": 4.957577127543745e-06, "loss": 0.3909, "step": 13582 }, { "epoch": 0.68, "learning_rate": 4.95618063179793e-06, "loss": 0.2401, "step": 13583 }, { "epoch": 0.68, "learning_rate": 4.954784267966604e-06, "loss": 0.5119, "step": 13584 }, { "epoch": 0.68, "learning_rate": 4.953388036086287e-06, "loss": 0.3677, "step": 13585 }, { "epoch": 0.68, "learning_rate": 4.951991936193493e-06, "loss": 0.2311, "step": 13586 }, { "epoch": 0.68, "learning_rate": 4.950595968324742e-06, "loss": 0.2745, "step": 13587 }, { "epoch": 0.68, "learning_rate": 4.949200132516539e-06, "loss": 0.1627, "step": 13588 }, { "epoch": 0.68, "learning_rate": 4.947804428805393e-06, "loss": 0.2144, "step": 13589 }, { "epoch": 0.68, "learning_rate": 4.946408857227804e-06, "loss": 0.3143, "step": 13590 }, { "epoch": 0.68, "learning_rate": 4.945013417820268e-06, "loss": 0.5994, "step": 13591 }, { "epoch": 0.68, "learning_rate": 4.943618110619289e-06, "loss": 0.2161, "step": 13592 }, { "epoch": 0.68, "learning_rate": 4.942222935661356e-06, "loss": 0.4841, "step": 13593 }, { "epoch": 0.68, "learning_rate": 4.940827892982958e-06, "loss": 0.2101, "step": 13594 }, { "epoch": 0.68, "learning_rate": 4.939432982620579e-06, "loss": 0.4643, "step": 13595 }, { "epoch": 0.68, "learning_rate": 4.938038204610703e-06, "loss": 0.359, "step": 13596 }, { "epoch": 0.68, "learning_rate": 4.936643558989807e-06, "loss": 0.3072, "step": 13597 }, { "epoch": 0.68, "learning_rate": 4.935249045794369e-06, "loss": 0.4388, "step": 13598 }, { "epoch": 0.68, "learning_rate": 4.9338546650608564e-06, "loss": 0.4924, "step": 13599 }, { "epoch": 0.68, "learning_rate": 4.932460416825737e-06, "loss": 0.5436, "step": 13600 }, { "epoch": 0.68, "learning_rate": 4.9310663011254826e-06, "loss": 0.3194, "step": 13601 }, { "epoch": 0.68, "learning_rate": 4.929672317996548e-06, "loss": 0.4782, "step": 13602 }, { "epoch": 0.68, "learning_rate": 4.928278467475394e-06, "loss": 0.6623, "step": 13603 }, { "epoch": 0.68, "learning_rate": 4.926884749598474e-06, "loss": 0.3689, "step": 13604 }, { "epoch": 0.68, "learning_rate": 4.925491164402234e-06, "loss": 0.1967, "step": 13605 }, { "epoch": 0.68, "learning_rate": 4.924097711923131e-06, "loss": 0.4051, "step": 13606 }, { "epoch": 0.68, "learning_rate": 4.922704392197604e-06, "loss": 0.1121, "step": 13607 }, { "epoch": 0.68, "learning_rate": 4.921311205262094e-06, "loss": 0.3023, "step": 13608 }, { "epoch": 0.68, "learning_rate": 4.919918151153032e-06, "loss": 0.1273, "step": 13609 }, { "epoch": 0.68, "learning_rate": 4.9185252299068635e-06, "loss": 0.5122, "step": 13610 }, { "epoch": 0.68, "learning_rate": 4.91713244156001e-06, "loss": 0.3552, "step": 13611 }, { "epoch": 0.68, "learning_rate": 4.915739786148901e-06, "loss": 0.3613, "step": 13612 }, { "epoch": 0.68, "learning_rate": 4.914347263709959e-06, "loss": 0.2674, "step": 13613 }, { "epoch": 0.68, "learning_rate": 4.912954874279599e-06, "loss": 0.2997, "step": 13614 }, { "epoch": 0.68, "learning_rate": 4.911562617894244e-06, "loss": 0.5171, "step": 13615 }, { "epoch": 0.68, "learning_rate": 4.910170494590305e-06, "loss": 0.3491, "step": 13616 }, { "epoch": 0.68, "learning_rate": 4.908778504404192e-06, "loss": 0.2657, "step": 13617 }, { "epoch": 0.68, "learning_rate": 4.907386647372303e-06, "loss": 0.394, "step": 13618 }, { "epoch": 0.68, "learning_rate": 4.90599492353105e-06, "loss": 0.3842, "step": 13619 }, { "epoch": 0.68, "learning_rate": 4.904603332916829e-06, "loss": 0.4864, "step": 13620 }, { "epoch": 0.68, "learning_rate": 4.903211875566033e-06, "loss": 0.3896, "step": 13621 }, { "epoch": 0.68, "learning_rate": 4.901820551515055e-06, "loss": 0.3161, "step": 13622 }, { "epoch": 0.68, "learning_rate": 4.900429360800284e-06, "loss": 0.1688, "step": 13623 }, { "epoch": 0.68, "learning_rate": 4.899038303458102e-06, "loss": 0.4926, "step": 13624 }, { "epoch": 0.68, "learning_rate": 4.897647379524893e-06, "loss": 0.4917, "step": 13625 }, { "epoch": 0.68, "learning_rate": 4.896256589037035e-06, "loss": 0.4983, "step": 13626 }, { "epoch": 0.68, "learning_rate": 4.8948659320309e-06, "loss": 0.2962, "step": 13627 }, { "epoch": 0.68, "learning_rate": 4.893475408542857e-06, "loss": 0.5519, "step": 13628 }, { "epoch": 0.68, "learning_rate": 4.892085018609279e-06, "loss": 0.1783, "step": 13629 }, { "epoch": 0.68, "learning_rate": 4.890694762266528e-06, "loss": 0.4968, "step": 13630 }, { "epoch": 0.68, "learning_rate": 4.889304639550964e-06, "loss": 0.1791, "step": 13631 }, { "epoch": 0.68, "learning_rate": 4.887914650498939e-06, "loss": 0.2949, "step": 13632 }, { "epoch": 0.68, "learning_rate": 4.8865247951468145e-06, "loss": 0.3516, "step": 13633 }, { "epoch": 0.68, "learning_rate": 4.885135073530937e-06, "loss": 0.4913, "step": 13634 }, { "epoch": 0.68, "learning_rate": 4.883745485687651e-06, "loss": 0.1467, "step": 13635 }, { "epoch": 0.68, "learning_rate": 4.882356031653302e-06, "loss": 0.2444, "step": 13636 }, { "epoch": 0.68, "learning_rate": 4.880966711464224e-06, "loss": 0.3672, "step": 13637 }, { "epoch": 0.68, "learning_rate": 4.87957752515676e-06, "loss": 0.3055, "step": 13638 }, { "epoch": 0.68, "learning_rate": 4.878188472767241e-06, "loss": 0.3274, "step": 13639 }, { "epoch": 0.68, "learning_rate": 4.8767995543319915e-06, "loss": 0.2582, "step": 13640 }, { "epoch": 0.68, "learning_rate": 4.875410769887338e-06, "loss": 0.3885, "step": 13641 }, { "epoch": 0.68, "learning_rate": 4.874022119469606e-06, "loss": 0.2525, "step": 13642 }, { "epoch": 0.68, "learning_rate": 4.872633603115112e-06, "loss": 0.3209, "step": 13643 }, { "epoch": 0.68, "learning_rate": 4.8712452208601715e-06, "loss": 0.3748, "step": 13644 }, { "epoch": 0.68, "learning_rate": 4.869856972741094e-06, "loss": 0.3657, "step": 13645 }, { "epoch": 0.68, "learning_rate": 4.868468858794189e-06, "loss": 0.3476, "step": 13646 }, { "epoch": 0.68, "learning_rate": 4.867080879055759e-06, "loss": 0.4193, "step": 13647 }, { "epoch": 0.68, "learning_rate": 4.865693033562106e-06, "loss": 0.2797, "step": 13648 }, { "epoch": 0.68, "learning_rate": 4.864305322349526e-06, "loss": 0.1229, "step": 13649 }, { "epoch": 0.68, "learning_rate": 4.862917745454311e-06, "loss": 0.3553, "step": 13650 }, { "epoch": 0.68, "learning_rate": 4.8615303029127584e-06, "loss": 0.2739, "step": 13651 }, { "epoch": 0.68, "learning_rate": 4.860142994761149e-06, "loss": 0.3106, "step": 13652 }, { "epoch": 0.68, "learning_rate": 4.858755821035767e-06, "loss": 0.2891, "step": 13653 }, { "epoch": 0.68, "learning_rate": 4.857368781772895e-06, "loss": 0.3763, "step": 13654 }, { "epoch": 0.68, "learning_rate": 4.8559818770088015e-06, "loss": 0.194, "step": 13655 }, { "epoch": 0.68, "learning_rate": 4.854595106779769e-06, "loss": 0.3114, "step": 13656 }, { "epoch": 0.68, "learning_rate": 4.853208471122063e-06, "loss": 0.421, "step": 13657 }, { "epoch": 0.68, "learning_rate": 4.851821970071947e-06, "loss": 0.4623, "step": 13658 }, { "epoch": 0.68, "learning_rate": 4.8504356036656805e-06, "loss": 0.1824, "step": 13659 }, { "epoch": 0.68, "learning_rate": 4.849049371939532e-06, "loss": 0.7904, "step": 13660 }, { "epoch": 0.68, "learning_rate": 4.847663274929748e-06, "loss": 0.3326, "step": 13661 }, { "epoch": 0.68, "learning_rate": 4.8462773126725835e-06, "loss": 0.4156, "step": 13662 }, { "epoch": 0.68, "learning_rate": 4.844891485204287e-06, "loss": 0.2383, "step": 13663 }, { "epoch": 0.68, "learning_rate": 4.843505792561096e-06, "loss": 0.5582, "step": 13664 }, { "epoch": 0.68, "learning_rate": 4.842120234779263e-06, "loss": 0.3108, "step": 13665 }, { "epoch": 0.68, "learning_rate": 4.8407348118950185e-06, "loss": 0.2714, "step": 13666 }, { "epoch": 0.68, "learning_rate": 4.839349523944597e-06, "loss": 0.1665, "step": 13667 }, { "epoch": 0.68, "learning_rate": 4.837964370964231e-06, "loss": 0.1136, "step": 13668 }, { "epoch": 0.68, "learning_rate": 4.836579352990145e-06, "loss": 0.2121, "step": 13669 }, { "epoch": 0.68, "learning_rate": 4.8351944700585625e-06, "loss": 0.2851, "step": 13670 }, { "epoch": 0.68, "learning_rate": 4.833809722205704e-06, "loss": 0.1661, "step": 13671 }, { "epoch": 0.68, "learning_rate": 4.8324251094677875e-06, "loss": 0.3516, "step": 13672 }, { "epoch": 0.68, "learning_rate": 4.831040631881019e-06, "loss": 0.4123, "step": 13673 }, { "epoch": 0.68, "learning_rate": 4.829656289481616e-06, "loss": 0.3979, "step": 13674 }, { "epoch": 0.68, "learning_rate": 4.828272082305782e-06, "loss": 0.3427, "step": 13675 }, { "epoch": 0.68, "learning_rate": 4.826888010389716e-06, "loss": 0.3128, "step": 13676 }, { "epoch": 0.68, "learning_rate": 4.8255040737696206e-06, "loss": 0.517, "step": 13677 }, { "epoch": 0.68, "learning_rate": 4.824120272481684e-06, "loss": 0.3615, "step": 13678 }, { "epoch": 0.68, "learning_rate": 4.822736606562106e-06, "loss": 0.4071, "step": 13679 }, { "epoch": 0.68, "learning_rate": 4.821353076047072e-06, "loss": 0.2897, "step": 13680 }, { "epoch": 0.68, "learning_rate": 4.8199696809727645e-06, "loss": 0.3585, "step": 13681 }, { "epoch": 0.68, "learning_rate": 4.818586421375362e-06, "loss": 0.3509, "step": 13682 }, { "epoch": 0.68, "learning_rate": 4.817203297291049e-06, "loss": 0.5977, "step": 13683 }, { "epoch": 0.68, "learning_rate": 4.815820308755995e-06, "loss": 0.2589, "step": 13684 }, { "epoch": 0.68, "learning_rate": 4.814437455806372e-06, "loss": 0.3611, "step": 13685 }, { "epoch": 0.68, "learning_rate": 4.8130547384783445e-06, "loss": 0.2665, "step": 13686 }, { "epoch": 0.68, "learning_rate": 4.811672156808074e-06, "loss": 0.3957, "step": 13687 }, { "epoch": 0.68, "learning_rate": 4.810289710831725e-06, "loss": 0.3694, "step": 13688 }, { "epoch": 0.68, "learning_rate": 4.808907400585452e-06, "loss": 0.3783, "step": 13689 }, { "epoch": 0.68, "learning_rate": 4.807525226105407e-06, "loss": 0.2962, "step": 13690 }, { "epoch": 0.68, "learning_rate": 4.806143187427734e-06, "loss": 0.5183, "step": 13691 }, { "epoch": 0.68, "learning_rate": 4.804761284588589e-06, "loss": 0.4855, "step": 13692 }, { "epoch": 0.68, "learning_rate": 4.803379517624106e-06, "loss": 0.4011, "step": 13693 }, { "epoch": 0.68, "learning_rate": 4.801997886570427e-06, "loss": 0.294, "step": 13694 }, { "epoch": 0.68, "learning_rate": 4.800616391463684e-06, "loss": 0.3057, "step": 13695 }, { "epoch": 0.68, "learning_rate": 4.79923503234001e-06, "loss": 0.3556, "step": 13696 }, { "epoch": 0.68, "learning_rate": 4.797853809235531e-06, "loss": 0.4194, "step": 13697 }, { "epoch": 0.68, "learning_rate": 4.796472722186372e-06, "loss": 0.3173, "step": 13698 }, { "epoch": 0.68, "learning_rate": 4.795091771228655e-06, "loss": 0.4402, "step": 13699 }, { "epoch": 0.68, "learning_rate": 4.793710956398495e-06, "loss": 0.4981, "step": 13700 }, { "epoch": 0.68, "learning_rate": 4.792330277732001e-06, "loss": 0.5219, "step": 13701 }, { "epoch": 0.68, "learning_rate": 4.790949735265293e-06, "loss": 0.3161, "step": 13702 }, { "epoch": 0.68, "learning_rate": 4.7895693290344716e-06, "loss": 0.4158, "step": 13703 }, { "epoch": 0.68, "learning_rate": 4.78818905907564e-06, "loss": 0.3661, "step": 13704 }, { "epoch": 0.68, "learning_rate": 4.786808925424893e-06, "loss": 0.3546, "step": 13705 }, { "epoch": 0.68, "learning_rate": 4.785428928118334e-06, "loss": 0.3078, "step": 13706 }, { "epoch": 0.68, "learning_rate": 4.7840490671920525e-06, "loss": 0.0981, "step": 13707 }, { "epoch": 0.68, "learning_rate": 4.782669342682136e-06, "loss": 0.372, "step": 13708 }, { "epoch": 0.68, "learning_rate": 4.781289754624668e-06, "loss": 0.342, "step": 13709 }, { "epoch": 0.68, "learning_rate": 4.779910303055729e-06, "loss": 0.3389, "step": 13710 }, { "epoch": 0.68, "learning_rate": 4.778530988011403e-06, "loss": 0.3784, "step": 13711 }, { "epoch": 0.68, "learning_rate": 4.777151809527759e-06, "loss": 0.3534, "step": 13712 }, { "epoch": 0.68, "learning_rate": 4.775772767640869e-06, "loss": 0.3282, "step": 13713 }, { "epoch": 0.68, "learning_rate": 4.7743938623867954e-06, "loss": 0.1815, "step": 13714 }, { "epoch": 0.68, "learning_rate": 4.77301509380161e-06, "loss": 0.5932, "step": 13715 }, { "epoch": 0.68, "learning_rate": 4.77163646192137e-06, "loss": 0.4627, "step": 13716 }, { "epoch": 0.68, "learning_rate": 4.770257966782129e-06, "loss": 0.349, "step": 13717 }, { "epoch": 0.68, "learning_rate": 4.768879608419942e-06, "loss": 0.4484, "step": 13718 }, { "epoch": 0.69, "learning_rate": 4.767501386870856e-06, "loss": 0.4287, "step": 13719 }, { "epoch": 0.69, "learning_rate": 4.766123302170919e-06, "loss": 0.5535, "step": 13720 }, { "epoch": 0.69, "learning_rate": 4.764745354356172e-06, "loss": 0.3032, "step": 13721 }, { "epoch": 0.69, "learning_rate": 4.7633675434626525e-06, "loss": 0.2537, "step": 13722 }, { "epoch": 0.69, "learning_rate": 4.761989869526394e-06, "loss": 0.4507, "step": 13723 }, { "epoch": 0.69, "learning_rate": 4.760612332583432e-06, "loss": 0.362, "step": 13724 }, { "epoch": 0.69, "learning_rate": 4.7592349326697915e-06, "loss": 0.5183, "step": 13725 }, { "epoch": 0.69, "learning_rate": 4.757857669821499e-06, "loss": 0.5303, "step": 13726 }, { "epoch": 0.69, "learning_rate": 4.756480544074572e-06, "loss": 0.191, "step": 13727 }, { "epoch": 0.69, "learning_rate": 4.755103555465026e-06, "loss": 0.4059, "step": 13728 }, { "epoch": 0.69, "learning_rate": 4.753726704028879e-06, "loss": 0.4094, "step": 13729 }, { "epoch": 0.69, "learning_rate": 4.7523499898021395e-06, "loss": 0.2971, "step": 13730 }, { "epoch": 0.69, "learning_rate": 4.750973412820814e-06, "loss": 0.2701, "step": 13731 }, { "epoch": 0.69, "learning_rate": 4.749596973120902e-06, "loss": 0.5077, "step": 13732 }, { "epoch": 0.69, "learning_rate": 4.748220670738401e-06, "loss": 0.3945, "step": 13733 }, { "epoch": 0.69, "learning_rate": 4.746844505709315e-06, "loss": 0.3387, "step": 13734 }, { "epoch": 0.69, "learning_rate": 4.745468478069629e-06, "loss": 0.3217, "step": 13735 }, { "epoch": 0.69, "learning_rate": 4.744092587855333e-06, "loss": 0.3041, "step": 13736 }, { "epoch": 0.69, "learning_rate": 4.7427168351024085e-06, "loss": 0.3307, "step": 13737 }, { "epoch": 0.69, "learning_rate": 4.741341219846842e-06, "loss": 0.4088, "step": 13738 }, { "epoch": 0.69, "learning_rate": 4.7399657421246095e-06, "loss": 0.4712, "step": 13739 }, { "epoch": 0.69, "learning_rate": 4.738590401971683e-06, "loss": 0.2978, "step": 13740 }, { "epoch": 0.69, "learning_rate": 4.737215199424033e-06, "loss": 0.2057, "step": 13741 }, { "epoch": 0.69, "learning_rate": 4.735840134517627e-06, "loss": 0.5106, "step": 13742 }, { "epoch": 0.69, "learning_rate": 4.734465207288426e-06, "loss": 0.3152, "step": 13743 }, { "epoch": 0.69, "learning_rate": 4.733090417772393e-06, "loss": 0.3714, "step": 13744 }, { "epoch": 0.69, "learning_rate": 4.731715766005482e-06, "loss": 0.3372, "step": 13745 }, { "epoch": 0.69, "learning_rate": 4.730341252023639e-06, "loss": 0.2865, "step": 13746 }, { "epoch": 0.69, "learning_rate": 4.728966875862823e-06, "loss": 0.3122, "step": 13747 }, { "epoch": 0.69, "learning_rate": 4.727592637558975e-06, "loss": 0.2777, "step": 13748 }, { "epoch": 0.69, "learning_rate": 4.7262185371480355e-06, "loss": 0.3547, "step": 13749 }, { "epoch": 0.69, "learning_rate": 4.724844574665943e-06, "loss": 0.3049, "step": 13750 }, { "epoch": 0.69, "learning_rate": 4.723470750148628e-06, "loss": 0.2348, "step": 13751 }, { "epoch": 0.69, "learning_rate": 4.722097063632027e-06, "loss": 0.3848, "step": 13752 }, { "epoch": 0.69, "learning_rate": 4.7207235151520655e-06, "loss": 0.2297, "step": 13753 }, { "epoch": 0.69, "learning_rate": 4.719350104744667e-06, "loss": 0.2633, "step": 13754 }, { "epoch": 0.69, "learning_rate": 4.7179768324457445e-06, "loss": 0.2553, "step": 13755 }, { "epoch": 0.69, "learning_rate": 4.716603698291225e-06, "loss": 0.5406, "step": 13756 }, { "epoch": 0.69, "learning_rate": 4.715230702317014e-06, "loss": 0.7455, "step": 13757 }, { "epoch": 0.69, "learning_rate": 4.713857844559023e-06, "loss": 0.3456, "step": 13758 }, { "epoch": 0.69, "learning_rate": 4.712485125053157e-06, "loss": 0.1476, "step": 13759 }, { "epoch": 0.69, "learning_rate": 4.7111125438353125e-06, "loss": 0.3603, "step": 13760 }, { "epoch": 0.69, "learning_rate": 4.709740100941397e-06, "loss": 0.51, "step": 13761 }, { "epoch": 0.69, "learning_rate": 4.708367796407299e-06, "loss": 0.5579, "step": 13762 }, { "epoch": 0.69, "learning_rate": 4.706995630268911e-06, "loss": 0.2472, "step": 13763 }, { "epoch": 0.69, "learning_rate": 4.705623602562121e-06, "loss": 0.2699, "step": 13764 }, { "epoch": 0.69, "learning_rate": 4.704251713322805e-06, "loss": 0.2808, "step": 13765 }, { "epoch": 0.69, "learning_rate": 4.702879962586855e-06, "loss": 0.404, "step": 13766 }, { "epoch": 0.69, "learning_rate": 4.701508350390141e-06, "loss": 0.2539, "step": 13767 }, { "epoch": 0.69, "learning_rate": 4.700136876768536e-06, "loss": 0.5658, "step": 13768 }, { "epoch": 0.69, "learning_rate": 4.698765541757909e-06, "loss": 0.3442, "step": 13769 }, { "epoch": 0.69, "learning_rate": 4.697394345394127e-06, "loss": 0.7092, "step": 13770 }, { "epoch": 0.69, "learning_rate": 4.696023287713051e-06, "loss": 0.2778, "step": 13771 }, { "epoch": 0.69, "learning_rate": 4.694652368750537e-06, "loss": 0.3823, "step": 13772 }, { "epoch": 0.69, "learning_rate": 4.6932815885424434e-06, "loss": 0.5048, "step": 13773 }, { "epoch": 0.69, "learning_rate": 4.6919109471246146e-06, "loss": 0.3726, "step": 13774 }, { "epoch": 0.69, "learning_rate": 4.690540444532907e-06, "loss": 0.449, "step": 13775 }, { "epoch": 0.69, "learning_rate": 4.689170080803159e-06, "loss": 0.4422, "step": 13776 }, { "epoch": 0.69, "learning_rate": 4.6877998559712125e-06, "loss": 0.3135, "step": 13777 }, { "epoch": 0.69, "learning_rate": 4.686429770072899e-06, "loss": 0.4745, "step": 13778 }, { "epoch": 0.69, "learning_rate": 4.685059823144059e-06, "loss": 0.5202, "step": 13779 }, { "epoch": 0.69, "learning_rate": 4.683690015220519e-06, "loss": 0.2242, "step": 13780 }, { "epoch": 0.69, "learning_rate": 4.682320346338102e-06, "loss": 0.372, "step": 13781 }, { "epoch": 0.69, "learning_rate": 4.680950816532632e-06, "loss": 0.2437, "step": 13782 }, { "epoch": 0.69, "learning_rate": 4.679581425839923e-06, "loss": 0.4023, "step": 13783 }, { "epoch": 0.69, "learning_rate": 4.678212174295797e-06, "loss": 0.4323, "step": 13784 }, { "epoch": 0.69, "learning_rate": 4.676843061936061e-06, "loss": 0.3307, "step": 13785 }, { "epoch": 0.69, "learning_rate": 4.675474088796523e-06, "loss": 0.605, "step": 13786 }, { "epoch": 0.69, "learning_rate": 4.6741052549129826e-06, "loss": 0.3564, "step": 13787 }, { "epoch": 0.69, "learning_rate": 4.672736560321248e-06, "loss": 0.1209, "step": 13788 }, { "epoch": 0.69, "learning_rate": 4.671368005057111e-06, "loss": 0.3971, "step": 13789 }, { "epoch": 0.69, "learning_rate": 4.669999589156363e-06, "loss": 0.316, "step": 13790 }, { "epoch": 0.69, "learning_rate": 4.668631312654797e-06, "loss": 0.2565, "step": 13791 }, { "epoch": 0.69, "learning_rate": 4.667263175588193e-06, "loss": 0.4196, "step": 13792 }, { "epoch": 0.69, "learning_rate": 4.665895177992339e-06, "loss": 0.4795, "step": 13793 }, { "epoch": 0.69, "learning_rate": 4.66452731990301e-06, "loss": 0.3312, "step": 13794 }, { "epoch": 0.69, "learning_rate": 4.663159601355979e-06, "loss": 0.6362, "step": 13795 }, { "epoch": 0.69, "learning_rate": 4.661792022387016e-06, "loss": 0.1696, "step": 13796 }, { "epoch": 0.69, "learning_rate": 4.6604245830318936e-06, "loss": 0.4402, "step": 13797 }, { "epoch": 0.69, "learning_rate": 4.659057283326375e-06, "loss": 0.4145, "step": 13798 }, { "epoch": 0.69, "learning_rate": 4.657690123306215e-06, "loss": 0.1845, "step": 13799 }, { "epoch": 0.69, "learning_rate": 4.656323103007175e-06, "loss": 0.5018, "step": 13800 }, { "epoch": 0.69, "learning_rate": 4.654956222465e-06, "loss": 0.3926, "step": 13801 }, { "epoch": 0.69, "learning_rate": 4.653589481715448e-06, "loss": 0.3926, "step": 13802 }, { "epoch": 0.69, "learning_rate": 4.652222880794261e-06, "loss": 0.1328, "step": 13803 }, { "epoch": 0.69, "learning_rate": 4.650856419737181e-06, "loss": 0.3606, "step": 13804 }, { "epoch": 0.69, "learning_rate": 4.649490098579944e-06, "loss": 0.3273, "step": 13805 }, { "epoch": 0.69, "learning_rate": 4.648123917358281e-06, "loss": 0.2863, "step": 13806 }, { "epoch": 0.69, "learning_rate": 4.646757876107933e-06, "loss": 0.4061, "step": 13807 }, { "epoch": 0.69, "learning_rate": 4.64539197486462e-06, "loss": 0.2204, "step": 13808 }, { "epoch": 0.69, "learning_rate": 4.644026213664068e-06, "loss": 0.5172, "step": 13809 }, { "epoch": 0.69, "learning_rate": 4.64266059254199e-06, "loss": 0.648, "step": 13810 }, { "epoch": 0.69, "learning_rate": 4.641295111534112e-06, "loss": 0.568, "step": 13811 }, { "epoch": 0.69, "learning_rate": 4.639929770676141e-06, "loss": 0.297, "step": 13812 }, { "epoch": 0.69, "learning_rate": 4.638564570003787e-06, "loss": 0.2649, "step": 13813 }, { "epoch": 0.69, "learning_rate": 4.637199509552754e-06, "loss": 0.379, "step": 13814 }, { "epoch": 0.69, "learning_rate": 4.635834589358745e-06, "loss": 0.5612, "step": 13815 }, { "epoch": 0.69, "learning_rate": 4.634469809457456e-06, "loss": 0.4546, "step": 13816 }, { "epoch": 0.69, "learning_rate": 4.63310516988458e-06, "loss": 0.2356, "step": 13817 }, { "epoch": 0.69, "learning_rate": 4.631740670675812e-06, "loss": 0.4106, "step": 13818 }, { "epoch": 0.69, "learning_rate": 4.630376311866829e-06, "loss": 0.226, "step": 13819 }, { "epoch": 0.69, "learning_rate": 4.629012093493326e-06, "loss": 0.2593, "step": 13820 }, { "epoch": 0.69, "learning_rate": 4.627648015590976e-06, "loss": 0.4229, "step": 13821 }, { "epoch": 0.69, "learning_rate": 4.6262840781954555e-06, "loss": 0.5641, "step": 13822 }, { "epoch": 0.69, "learning_rate": 4.624920281342437e-06, "loss": 0.4229, "step": 13823 }, { "epoch": 0.69, "learning_rate": 4.623556625067585e-06, "loss": 0.2918, "step": 13824 }, { "epoch": 0.69, "learning_rate": 4.622193109406572e-06, "loss": 0.0518, "step": 13825 }, { "epoch": 0.69, "learning_rate": 4.620829734395054e-06, "loss": 0.3939, "step": 13826 }, { "epoch": 0.69, "learning_rate": 4.619466500068688e-06, "loss": 0.3651, "step": 13827 }, { "epoch": 0.69, "learning_rate": 4.618103406463124e-06, "loss": 0.302, "step": 13828 }, { "epoch": 0.69, "learning_rate": 4.616740453614023e-06, "loss": 0.3552, "step": 13829 }, { "epoch": 0.69, "learning_rate": 4.615377641557023e-06, "loss": 0.2077, "step": 13830 }, { "epoch": 0.69, "learning_rate": 4.614014970327768e-06, "loss": 0.3049, "step": 13831 }, { "epoch": 0.69, "learning_rate": 4.612652439961896e-06, "loss": 0.4618, "step": 13832 }, { "epoch": 0.69, "learning_rate": 4.611290050495042e-06, "loss": 0.2611, "step": 13833 }, { "epoch": 0.69, "learning_rate": 4.609927801962841e-06, "loss": 0.3752, "step": 13834 }, { "epoch": 0.69, "learning_rate": 4.608565694400919e-06, "loss": 0.435, "step": 13835 }, { "epoch": 0.69, "learning_rate": 4.6072037278449e-06, "loss": 0.2283, "step": 13836 }, { "epoch": 0.69, "learning_rate": 4.6058419023304044e-06, "loss": 0.2105, "step": 13837 }, { "epoch": 0.69, "learning_rate": 4.604480217893045e-06, "loss": 0.2912, "step": 13838 }, { "epoch": 0.69, "learning_rate": 4.603118674568441e-06, "loss": 0.3282, "step": 13839 }, { "epoch": 0.69, "learning_rate": 4.601757272392201e-06, "loss": 0.3455, "step": 13840 }, { "epoch": 0.69, "learning_rate": 4.600396011399928e-06, "loss": 0.1435, "step": 13841 }, { "epoch": 0.69, "learning_rate": 4.599034891627225e-06, "loss": 0.291, "step": 13842 }, { "epoch": 0.69, "learning_rate": 4.597673913109691e-06, "loss": 0.1873, "step": 13843 }, { "epoch": 0.69, "learning_rate": 4.596313075882921e-06, "loss": 0.2193, "step": 13844 }, { "epoch": 0.69, "learning_rate": 4.594952379982503e-06, "loss": 0.2104, "step": 13845 }, { "epoch": 0.69, "learning_rate": 4.593591825444028e-06, "loss": 0.2836, "step": 13846 }, { "epoch": 0.69, "learning_rate": 4.592231412303073e-06, "loss": 0.2082, "step": 13847 }, { "epoch": 0.69, "learning_rate": 4.5908711405952264e-06, "loss": 0.2644, "step": 13848 }, { "epoch": 0.69, "learning_rate": 4.58951101035606e-06, "loss": 0.2702, "step": 13849 }, { "epoch": 0.69, "learning_rate": 4.5881510216211475e-06, "loss": 0.3344, "step": 13850 }, { "epoch": 0.69, "learning_rate": 4.586791174426053e-06, "loss": 0.3992, "step": 13851 }, { "epoch": 0.69, "learning_rate": 4.585431468806349e-06, "loss": 0.3485, "step": 13852 }, { "epoch": 0.69, "learning_rate": 4.584071904797592e-06, "loss": 0.3787, "step": 13853 }, { "epoch": 0.69, "learning_rate": 4.5827124824353406e-06, "loss": 0.2173, "step": 13854 }, { "epoch": 0.69, "learning_rate": 4.581353201755151e-06, "loss": 0.4249, "step": 13855 }, { "epoch": 0.69, "learning_rate": 4.579994062792565e-06, "loss": 0.2232, "step": 13856 }, { "epoch": 0.69, "learning_rate": 4.578635065583142e-06, "loss": 0.4445, "step": 13857 }, { "epoch": 0.69, "learning_rate": 4.577276210162416e-06, "loss": 0.3667, "step": 13858 }, { "epoch": 0.69, "learning_rate": 4.575917496565928e-06, "loss": 0.3985, "step": 13859 }, { "epoch": 0.69, "learning_rate": 4.574558924829211e-06, "loss": 0.4643, "step": 13860 }, { "epoch": 0.69, "learning_rate": 4.573200494987803e-06, "loss": 0.4528, "step": 13861 }, { "epoch": 0.69, "learning_rate": 4.571842207077227e-06, "loss": 0.5033, "step": 13862 }, { "epoch": 0.69, "learning_rate": 4.570484061133009e-06, "loss": 0.3614, "step": 13863 }, { "epoch": 0.69, "learning_rate": 4.5691260571906705e-06, "loss": 0.3228, "step": 13864 }, { "epoch": 0.69, "learning_rate": 4.567768195285724e-06, "loss": 0.5078, "step": 13865 }, { "epoch": 0.69, "learning_rate": 4.566410475453688e-06, "loss": 0.3676, "step": 13866 }, { "epoch": 0.69, "learning_rate": 4.565052897730068e-06, "loss": 0.2693, "step": 13867 }, { "epoch": 0.69, "learning_rate": 4.563695462150371e-06, "loss": 0.1558, "step": 13868 }, { "epoch": 0.69, "learning_rate": 4.5623381687501e-06, "loss": 0.3401, "step": 13869 }, { "epoch": 0.69, "learning_rate": 4.5609810175647475e-06, "loss": 0.3033, "step": 13870 }, { "epoch": 0.69, "learning_rate": 4.559624008629816e-06, "loss": 0.1828, "step": 13871 }, { "epoch": 0.69, "learning_rate": 4.558267141980793e-06, "loss": 0.3811, "step": 13872 }, { "epoch": 0.69, "learning_rate": 4.556910417653166e-06, "loss": 0.3064, "step": 13873 }, { "epoch": 0.69, "learning_rate": 4.555553835682415e-06, "loss": 0.2057, "step": 13874 }, { "epoch": 0.69, "learning_rate": 4.554197396104024e-06, "loss": 0.2936, "step": 13875 }, { "epoch": 0.69, "learning_rate": 4.552841098953469e-06, "loss": 0.5332, "step": 13876 }, { "epoch": 0.69, "learning_rate": 4.55148494426622e-06, "loss": 0.1228, "step": 13877 }, { "epoch": 0.69, "learning_rate": 4.550128932077747e-06, "loss": 0.5328, "step": 13878 }, { "epoch": 0.69, "learning_rate": 4.548773062423509e-06, "loss": 0.3776, "step": 13879 }, { "epoch": 0.69, "learning_rate": 4.547417335338975e-06, "loss": 0.3506, "step": 13880 }, { "epoch": 0.69, "learning_rate": 4.5460617508596e-06, "loss": 0.3858, "step": 13881 }, { "epoch": 0.69, "learning_rate": 4.544706309020837e-06, "loss": 0.3691, "step": 13882 }, { "epoch": 0.69, "learning_rate": 4.54335100985813e-06, "loss": 0.4174, "step": 13883 }, { "epoch": 0.69, "learning_rate": 4.541995853406935e-06, "loss": 0.4603, "step": 13884 }, { "epoch": 0.69, "learning_rate": 4.54064083970269e-06, "loss": 0.2916, "step": 13885 }, { "epoch": 0.69, "learning_rate": 4.539285968780833e-06, "loss": 0.3235, "step": 13886 }, { "epoch": 0.69, "learning_rate": 4.5379312406768e-06, "loss": 0.2135, "step": 13887 }, { "epoch": 0.69, "learning_rate": 4.536576655426019e-06, "loss": 0.3944, "step": 13888 }, { "epoch": 0.69, "learning_rate": 4.535222213063922e-06, "loss": 0.5607, "step": 13889 }, { "epoch": 0.69, "learning_rate": 4.53386791362593e-06, "loss": 0.2421, "step": 13890 }, { "epoch": 0.69, "learning_rate": 4.532513757147463e-06, "loss": 0.3182, "step": 13891 }, { "epoch": 0.69, "learning_rate": 4.531159743663934e-06, "loss": 0.3842, "step": 13892 }, { "epoch": 0.69, "learning_rate": 4.529805873210763e-06, "loss": 0.2256, "step": 13893 }, { "epoch": 0.69, "learning_rate": 4.5284521458233544e-06, "loss": 0.2556, "step": 13894 }, { "epoch": 0.69, "learning_rate": 4.527098561537114e-06, "loss": 0.2919, "step": 13895 }, { "epoch": 0.69, "learning_rate": 4.525745120387442e-06, "loss": 0.2512, "step": 13896 }, { "epoch": 0.69, "learning_rate": 4.524391822409733e-06, "loss": 0.1932, "step": 13897 }, { "epoch": 0.69, "learning_rate": 4.523038667639388e-06, "loss": 0.4037, "step": 13898 }, { "epoch": 0.69, "learning_rate": 4.521685656111793e-06, "loss": 0.4103, "step": 13899 }, { "epoch": 0.69, "learning_rate": 4.520332787862334e-06, "loss": 0.4675, "step": 13900 }, { "epoch": 0.69, "learning_rate": 4.518980062926393e-06, "loss": 0.2478, "step": 13901 }, { "epoch": 0.69, "learning_rate": 4.5176274813393475e-06, "loss": 0.4158, "step": 13902 }, { "epoch": 0.69, "learning_rate": 4.516275043136578e-06, "loss": 0.3785, "step": 13903 }, { "epoch": 0.69, "learning_rate": 4.514922748353452e-06, "loss": 0.3155, "step": 13904 }, { "epoch": 0.69, "learning_rate": 4.5135705970253375e-06, "loss": 0.687, "step": 13905 }, { "epoch": 0.69, "learning_rate": 4.512218589187593e-06, "loss": 0.2469, "step": 13906 }, { "epoch": 0.69, "learning_rate": 4.51086672487559e-06, "loss": 0.5958, "step": 13907 }, { "epoch": 0.69, "learning_rate": 4.509515004124676e-06, "loss": 0.3595, "step": 13908 }, { "epoch": 0.69, "learning_rate": 4.508163426970207e-06, "loss": 0.3012, "step": 13909 }, { "epoch": 0.69, "learning_rate": 4.506811993447531e-06, "loss": 0.2895, "step": 13910 }, { "epoch": 0.69, "learning_rate": 4.505460703591988e-06, "loss": 0.5251, "step": 13911 }, { "epoch": 0.69, "learning_rate": 4.504109557438927e-06, "loss": 0.7143, "step": 13912 }, { "epoch": 0.69, "learning_rate": 4.502758555023683e-06, "loss": 0.2919, "step": 13913 }, { "epoch": 0.69, "learning_rate": 4.501407696381589e-06, "loss": 0.4519, "step": 13914 }, { "epoch": 0.69, "learning_rate": 4.5000569815479744e-06, "loss": 0.072, "step": 13915 }, { "epoch": 0.69, "learning_rate": 4.498706410558166e-06, "loss": 0.2526, "step": 13916 }, { "epoch": 0.69, "learning_rate": 4.497355983447485e-06, "loss": 0.3523, "step": 13917 }, { "epoch": 0.69, "learning_rate": 4.4960057002512534e-06, "loss": 0.2137, "step": 13918 }, { "epoch": 0.7, "learning_rate": 4.494655561004783e-06, "loss": 0.3247, "step": 13919 }, { "epoch": 0.7, "learning_rate": 4.493305565743381e-06, "loss": 0.1322, "step": 13920 }, { "epoch": 0.7, "learning_rate": 4.491955714502363e-06, "loss": 0.3702, "step": 13921 }, { "epoch": 0.7, "learning_rate": 4.49060600731703e-06, "loss": 0.2552, "step": 13922 }, { "epoch": 0.7, "learning_rate": 4.48925644422268e-06, "loss": 0.4681, "step": 13923 }, { "epoch": 0.7, "learning_rate": 4.487907025254607e-06, "loss": 0.4529, "step": 13924 }, { "epoch": 0.7, "learning_rate": 4.48655775044811e-06, "loss": 0.4407, "step": 13925 }, { "epoch": 0.7, "learning_rate": 4.4852086198384736e-06, "loss": 0.3919, "step": 13926 }, { "epoch": 0.7, "learning_rate": 4.4838596334609816e-06, "loss": 0.5186, "step": 13927 }, { "epoch": 0.7, "learning_rate": 4.482510791350917e-06, "loss": 0.1312, "step": 13928 }, { "epoch": 0.7, "learning_rate": 4.481162093543553e-06, "loss": 0.4747, "step": 13929 }, { "epoch": 0.7, "learning_rate": 4.479813540074168e-06, "loss": 0.5017, "step": 13930 }, { "epoch": 0.7, "learning_rate": 4.478465130978032e-06, "loss": 0.2799, "step": 13931 }, { "epoch": 0.7, "learning_rate": 4.477116866290406e-06, "loss": 0.5545, "step": 13932 }, { "epoch": 0.7, "learning_rate": 4.475768746046552e-06, "loss": 0.399, "step": 13933 }, { "epoch": 0.7, "learning_rate": 4.4744207702817355e-06, "loss": 0.2004, "step": 13934 }, { "epoch": 0.7, "learning_rate": 4.473072939031207e-06, "loss": 0.3792, "step": 13935 }, { "epoch": 0.7, "learning_rate": 4.471725252330215e-06, "loss": 0.5203, "step": 13936 }, { "epoch": 0.7, "learning_rate": 4.470377710214009e-06, "loss": 0.4702, "step": 13937 }, { "epoch": 0.7, "learning_rate": 4.469030312717832e-06, "loss": 0.3743, "step": 13938 }, { "epoch": 0.7, "learning_rate": 4.467683059876923e-06, "loss": 0.4655, "step": 13939 }, { "epoch": 0.7, "learning_rate": 4.466335951726517e-06, "loss": 0.4718, "step": 13940 }, { "epoch": 0.7, "learning_rate": 4.464988988301846e-06, "loss": 0.3446, "step": 13941 }, { "epoch": 0.7, "learning_rate": 4.463642169638139e-06, "loss": 0.268, "step": 13942 }, { "epoch": 0.7, "learning_rate": 4.462295495770615e-06, "loss": 0.2544, "step": 13943 }, { "epoch": 0.7, "learning_rate": 4.460948966734504e-06, "loss": 0.2443, "step": 13944 }, { "epoch": 0.7, "learning_rate": 4.459602582565018e-06, "loss": 0.3357, "step": 13945 }, { "epoch": 0.7, "learning_rate": 4.45825634329737e-06, "loss": 0.4544, "step": 13946 }, { "epoch": 0.7, "learning_rate": 4.456910248966765e-06, "loss": 0.4151, "step": 13947 }, { "epoch": 0.7, "learning_rate": 4.455564299608416e-06, "loss": 0.2435, "step": 13948 }, { "epoch": 0.7, "learning_rate": 4.45421849525752e-06, "loss": 0.3278, "step": 13949 }, { "epoch": 0.7, "learning_rate": 4.452872835949277e-06, "loss": 0.2914, "step": 13950 }, { "epoch": 0.7, "learning_rate": 4.4515273217188785e-06, "loss": 0.4092, "step": 13951 }, { "epoch": 0.7, "learning_rate": 4.450181952601512e-06, "loss": 0.2995, "step": 13952 }, { "epoch": 0.7, "learning_rate": 4.448836728632372e-06, "loss": 0.3091, "step": 13953 }, { "epoch": 0.7, "learning_rate": 4.4474916498466356e-06, "loss": 0.644, "step": 13954 }, { "epoch": 0.7, "learning_rate": 4.446146716279484e-06, "loss": 0.3166, "step": 13955 }, { "epoch": 0.7, "learning_rate": 4.444801927966086e-06, "loss": 0.2977, "step": 13956 }, { "epoch": 0.7, "learning_rate": 4.4434572849416215e-06, "loss": 0.5101, "step": 13957 }, { "epoch": 0.7, "learning_rate": 4.442112787241254e-06, "loss": 0.0847, "step": 13958 }, { "epoch": 0.7, "learning_rate": 4.440768434900148e-06, "loss": 0.4066, "step": 13959 }, { "epoch": 0.7, "learning_rate": 4.43942422795346e-06, "loss": 0.4029, "step": 13960 }, { "epoch": 0.7, "learning_rate": 4.43808016643635e-06, "loss": 0.3751, "step": 13961 }, { "epoch": 0.7, "learning_rate": 4.436736250383969e-06, "loss": 0.4095, "step": 13962 }, { "epoch": 0.7, "learning_rate": 4.435392479831463e-06, "loss": 0.3137, "step": 13963 }, { "epoch": 0.7, "learning_rate": 4.434048854813976e-06, "loss": 0.3191, "step": 13964 }, { "epoch": 0.7, "learning_rate": 4.432705375366655e-06, "loss": 0.1709, "step": 13965 }, { "epoch": 0.7, "learning_rate": 4.431362041524631e-06, "loss": 0.2429, "step": 13966 }, { "epoch": 0.7, "learning_rate": 4.430018853323042e-06, "loss": 0.4156, "step": 13967 }, { "epoch": 0.7, "learning_rate": 4.428675810797013e-06, "loss": 0.3067, "step": 13968 }, { "epoch": 0.7, "learning_rate": 4.427332913981672e-06, "loss": 0.2277, "step": 13969 }, { "epoch": 0.7, "learning_rate": 4.4259901629121335e-06, "loss": 0.2566, "step": 13970 }, { "epoch": 0.7, "learning_rate": 4.4246475576235285e-06, "loss": 0.1947, "step": 13971 }, { "epoch": 0.7, "learning_rate": 4.423305098150962e-06, "loss": 0.3326, "step": 13972 }, { "epoch": 0.7, "learning_rate": 4.421962784529548e-06, "loss": 0.3974, "step": 13973 }, { "epoch": 0.7, "learning_rate": 4.42062061679439e-06, "loss": 0.5076, "step": 13974 }, { "epoch": 0.7, "learning_rate": 4.419278594980589e-06, "loss": 0.2545, "step": 13975 }, { "epoch": 0.7, "learning_rate": 4.417936719123251e-06, "loss": 0.2839, "step": 13976 }, { "epoch": 0.7, "learning_rate": 4.416594989257468e-06, "loss": 0.2444, "step": 13977 }, { "epoch": 0.7, "learning_rate": 4.4152534054183286e-06, "loss": 0.2334, "step": 13978 }, { "epoch": 0.7, "learning_rate": 4.413911967640918e-06, "loss": 0.3373, "step": 13979 }, { "epoch": 0.7, "learning_rate": 4.412570675960327e-06, "loss": 0.357, "step": 13980 }, { "epoch": 0.7, "learning_rate": 4.411229530411633e-06, "loss": 0.3855, "step": 13981 }, { "epoch": 0.7, "learning_rate": 4.409888531029909e-06, "loss": 0.422, "step": 13982 }, { "epoch": 0.7, "learning_rate": 4.408547677850229e-06, "loss": 0.4292, "step": 13983 }, { "epoch": 0.7, "learning_rate": 4.407206970907658e-06, "loss": 0.5549, "step": 13984 }, { "epoch": 0.7, "learning_rate": 4.405866410237267e-06, "loss": 0.3675, "step": 13985 }, { "epoch": 0.7, "learning_rate": 4.404525995874114e-06, "loss": 0.3618, "step": 13986 }, { "epoch": 0.7, "learning_rate": 4.403185727853253e-06, "loss": 0.4389, "step": 13987 }, { "epoch": 0.7, "learning_rate": 4.401845606209739e-06, "loss": 0.3715, "step": 13988 }, { "epoch": 0.7, "learning_rate": 4.400505630978622e-06, "loss": 0.2581, "step": 13989 }, { "epoch": 0.7, "learning_rate": 4.399165802194946e-06, "loss": 0.362, "step": 13990 }, { "epoch": 0.7, "learning_rate": 4.397826119893752e-06, "loss": 0.2904, "step": 13991 }, { "epoch": 0.7, "learning_rate": 4.396486584110079e-06, "loss": 0.3673, "step": 13992 }, { "epoch": 0.7, "learning_rate": 4.395147194878956e-06, "loss": 0.1837, "step": 13993 }, { "epoch": 0.7, "learning_rate": 4.39380795223542e-06, "loss": 0.434, "step": 13994 }, { "epoch": 0.7, "learning_rate": 4.392468856214495e-06, "loss": 0.2494, "step": 13995 }, { "epoch": 0.7, "learning_rate": 4.391129906851204e-06, "loss": 0.2776, "step": 13996 }, { "epoch": 0.7, "learning_rate": 4.3897911041805584e-06, "loss": 0.3947, "step": 13997 }, { "epoch": 0.7, "learning_rate": 4.3884524482375824e-06, "loss": 0.1808, "step": 13998 }, { "epoch": 0.7, "learning_rate": 4.387113939057284e-06, "loss": 0.3717, "step": 13999 }, { "epoch": 0.7, "learning_rate": 4.385775576674667e-06, "loss": 0.1621, "step": 14000 }, { "epoch": 0.7, "learning_rate": 4.384437361124737e-06, "loss": 0.4443, "step": 14001 }, { "epoch": 0.7, "learning_rate": 4.38309929244249e-06, "loss": 0.232, "step": 14002 }, { "epoch": 0.7, "learning_rate": 4.381761370662928e-06, "loss": 0.4257, "step": 14003 }, { "epoch": 0.7, "learning_rate": 4.380423595821037e-06, "loss": 0.5191, "step": 14004 }, { "epoch": 0.7, "learning_rate": 4.379085967951808e-06, "loss": 0.5338, "step": 14005 }, { "epoch": 0.7, "learning_rate": 4.377748487090223e-06, "loss": 0.4139, "step": 14006 }, { "epoch": 0.7, "learning_rate": 4.376411153271259e-06, "loss": 0.22, "step": 14007 }, { "epoch": 0.7, "learning_rate": 4.3750739665299e-06, "loss": 0.393, "step": 14008 }, { "epoch": 0.7, "learning_rate": 4.373736926901113e-06, "loss": 0.4302, "step": 14009 }, { "epoch": 0.7, "learning_rate": 4.372400034419868e-06, "loss": 0.3988, "step": 14010 }, { "epoch": 0.7, "learning_rate": 4.371063289121129e-06, "loss": 0.3871, "step": 14011 }, { "epoch": 0.7, "learning_rate": 4.369726691039858e-06, "loss": 0.1304, "step": 14012 }, { "epoch": 0.7, "learning_rate": 4.36839024021101e-06, "loss": 0.1949, "step": 14013 }, { "epoch": 0.7, "learning_rate": 4.36705393666954e-06, "loss": 0.3263, "step": 14014 }, { "epoch": 0.7, "learning_rate": 4.3657177804503955e-06, "loss": 0.411, "step": 14015 }, { "epoch": 0.7, "learning_rate": 4.36438177158852e-06, "loss": 0.213, "step": 14016 }, { "epoch": 0.7, "learning_rate": 4.363045910118862e-06, "loss": 0.3749, "step": 14017 }, { "epoch": 0.7, "learning_rate": 4.361710196076354e-06, "loss": 0.3148, "step": 14018 }, { "epoch": 0.7, "learning_rate": 4.360374629495932e-06, "loss": 0.2792, "step": 14019 }, { "epoch": 0.7, "learning_rate": 4.35903921041252e-06, "loss": 0.3448, "step": 14020 }, { "epoch": 0.7, "learning_rate": 4.357703938861054e-06, "loss": 0.5851, "step": 14021 }, { "epoch": 0.7, "learning_rate": 4.3563688148764505e-06, "loss": 0.2981, "step": 14022 }, { "epoch": 0.7, "learning_rate": 4.3550338384936296e-06, "loss": 0.3194, "step": 14023 }, { "epoch": 0.7, "learning_rate": 4.353699009747505e-06, "loss": 0.337, "step": 14024 }, { "epoch": 0.7, "learning_rate": 4.352364328672983e-06, "loss": 0.5034, "step": 14025 }, { "epoch": 0.7, "learning_rate": 4.351029795304979e-06, "loss": 0.4066, "step": 14026 }, { "epoch": 0.7, "learning_rate": 4.349695409678393e-06, "loss": 0.5255, "step": 14027 }, { "epoch": 0.7, "learning_rate": 4.348361171828122e-06, "loss": 0.32, "step": 14028 }, { "epoch": 0.7, "learning_rate": 4.347027081789058e-06, "loss": 0.2188, "step": 14029 }, { "epoch": 0.7, "learning_rate": 4.345693139596101e-06, "loss": 0.4011, "step": 14030 }, { "epoch": 0.7, "learning_rate": 4.344359345284135e-06, "loss": 0.4001, "step": 14031 }, { "epoch": 0.7, "learning_rate": 4.343025698888042e-06, "loss": 0.4084, "step": 14032 }, { "epoch": 0.7, "learning_rate": 4.341692200442702e-06, "loss": 0.3722, "step": 14033 }, { "epoch": 0.7, "learning_rate": 4.34035884998299e-06, "loss": 0.5672, "step": 14034 }, { "epoch": 0.7, "learning_rate": 4.339025647543782e-06, "loss": 0.1605, "step": 14035 }, { "epoch": 0.7, "learning_rate": 4.33769259315994e-06, "loss": 0.1886, "step": 14036 }, { "epoch": 0.7, "learning_rate": 4.336359686866335e-06, "loss": 0.2837, "step": 14037 }, { "epoch": 0.7, "learning_rate": 4.335026928697824e-06, "loss": 0.4264, "step": 14038 }, { "epoch": 0.7, "learning_rate": 4.333694318689264e-06, "loss": 0.2975, "step": 14039 }, { "epoch": 0.7, "learning_rate": 4.332361856875508e-06, "loss": 0.4835, "step": 14040 }, { "epoch": 0.7, "learning_rate": 4.331029543291404e-06, "loss": 0.406, "step": 14041 }, { "epoch": 0.7, "learning_rate": 4.329697377971796e-06, "loss": 0.2236, "step": 14042 }, { "epoch": 0.7, "learning_rate": 4.328365360951525e-06, "loss": 0.2522, "step": 14043 }, { "epoch": 0.7, "learning_rate": 4.327033492265432e-06, "loss": 0.2508, "step": 14044 }, { "epoch": 0.7, "learning_rate": 4.325701771948347e-06, "loss": 0.4684, "step": 14045 }, { "epoch": 0.7, "learning_rate": 4.324370200035102e-06, "loss": 0.4457, "step": 14046 }, { "epoch": 0.7, "learning_rate": 4.32303877656052e-06, "loss": 0.4935, "step": 14047 }, { "epoch": 0.7, "learning_rate": 4.321707501559419e-06, "loss": 0.6176, "step": 14048 }, { "epoch": 0.7, "learning_rate": 4.320376375066625e-06, "loss": 0.3491, "step": 14049 }, { "epoch": 0.7, "learning_rate": 4.319045397116948e-06, "loss": 0.6225, "step": 14050 }, { "epoch": 0.7, "learning_rate": 4.317714567745198e-06, "loss": 0.4172, "step": 14051 }, { "epoch": 0.7, "learning_rate": 4.316383886986178e-06, "loss": 0.2383, "step": 14052 }, { "epoch": 0.7, "learning_rate": 4.315053354874696e-06, "loss": 0.292, "step": 14053 }, { "epoch": 0.7, "learning_rate": 4.313722971445546e-06, "loss": 0.4264, "step": 14054 }, { "epoch": 0.7, "learning_rate": 4.312392736733526e-06, "loss": 0.3839, "step": 14055 }, { "epoch": 0.7, "learning_rate": 4.311062650773426e-06, "loss": 0.4117, "step": 14056 }, { "epoch": 0.7, "learning_rate": 4.3097327136000246e-06, "loss": 0.5191, "step": 14057 }, { "epoch": 0.7, "learning_rate": 4.308402925248117e-06, "loss": 0.3742, "step": 14058 }, { "epoch": 0.7, "learning_rate": 4.307073285752475e-06, "loss": 0.407, "step": 14059 }, { "epoch": 0.7, "learning_rate": 4.305743795147874e-06, "loss": 0.2975, "step": 14060 }, { "epoch": 0.7, "learning_rate": 4.304414453469088e-06, "loss": 0.2926, "step": 14061 }, { "epoch": 0.7, "learning_rate": 4.303085260750881e-06, "loss": 0.3926, "step": 14062 }, { "epoch": 0.7, "learning_rate": 4.301756217028017e-06, "loss": 0.2648, "step": 14063 }, { "epoch": 0.7, "learning_rate": 4.300427322335257e-06, "loss": 0.4139, "step": 14064 }, { "epoch": 0.7, "learning_rate": 4.299098576707355e-06, "loss": 0.4323, "step": 14065 }, { "epoch": 0.7, "learning_rate": 4.297769980179058e-06, "loss": 0.3453, "step": 14066 }, { "epoch": 0.7, "learning_rate": 4.296441532785123e-06, "loss": 0.2901, "step": 14067 }, { "epoch": 0.7, "learning_rate": 4.2951132345602895e-06, "loss": 0.3964, "step": 14068 }, { "epoch": 0.7, "learning_rate": 4.293785085539297e-06, "loss": 0.3805, "step": 14069 }, { "epoch": 0.7, "learning_rate": 4.292457085756878e-06, "loss": 0.2106, "step": 14070 }, { "epoch": 0.7, "learning_rate": 4.291129235247773e-06, "loss": 0.2944, "step": 14071 }, { "epoch": 0.7, "learning_rate": 4.289801534046704e-06, "loss": 0.5184, "step": 14072 }, { "epoch": 0.7, "learning_rate": 4.288473982188399e-06, "loss": 0.4875, "step": 14073 }, { "epoch": 0.7, "learning_rate": 4.287146579707575e-06, "loss": 0.1959, "step": 14074 }, { "epoch": 0.7, "learning_rate": 4.2858193266389455e-06, "loss": 0.2997, "step": 14075 }, { "epoch": 0.7, "learning_rate": 4.284492223017232e-06, "loss": 0.2717, "step": 14076 }, { "epoch": 0.7, "learning_rate": 4.283165268877137e-06, "loss": 0.4071, "step": 14077 }, { "epoch": 0.7, "learning_rate": 4.281838464253369e-06, "loss": 0.4675, "step": 14078 }, { "epoch": 0.7, "learning_rate": 4.2805118091806254e-06, "loss": 0.2113, "step": 14079 }, { "epoch": 0.7, "learning_rate": 4.279185303693599e-06, "loss": 0.571, "step": 14080 }, { "epoch": 0.7, "learning_rate": 4.277858947826994e-06, "loss": 0.5085, "step": 14081 }, { "epoch": 0.7, "learning_rate": 4.276532741615492e-06, "loss": 0.3026, "step": 14082 }, { "epoch": 0.7, "learning_rate": 4.2752066850937815e-06, "loss": 0.1381, "step": 14083 }, { "epoch": 0.7, "learning_rate": 4.2738807782965405e-06, "loss": 0.399, "step": 14084 }, { "epoch": 0.7, "learning_rate": 4.272555021258449e-06, "loss": 0.3643, "step": 14085 }, { "epoch": 0.7, "learning_rate": 4.271229414014178e-06, "loss": 0.3748, "step": 14086 }, { "epoch": 0.7, "learning_rate": 4.269903956598399e-06, "loss": 0.484, "step": 14087 }, { "epoch": 0.7, "learning_rate": 4.268578649045777e-06, "loss": 0.3756, "step": 14088 }, { "epoch": 0.7, "learning_rate": 4.26725349139097e-06, "loss": 0.2871, "step": 14089 }, { "epoch": 0.7, "learning_rate": 4.265928483668642e-06, "loss": 0.2505, "step": 14090 }, { "epoch": 0.7, "learning_rate": 4.264603625913446e-06, "loss": 0.4048, "step": 14091 }, { "epoch": 0.7, "learning_rate": 4.2632789181600285e-06, "loss": 0.2386, "step": 14092 }, { "epoch": 0.7, "learning_rate": 4.261954360443034e-06, "loss": 0.2924, "step": 14093 }, { "epoch": 0.7, "learning_rate": 4.260629952797113e-06, "loss": 0.3046, "step": 14094 }, { "epoch": 0.7, "learning_rate": 4.259305695256897e-06, "loss": 0.3285, "step": 14095 }, { "epoch": 0.7, "learning_rate": 4.257981587857021e-06, "loss": 0.4353, "step": 14096 }, { "epoch": 0.7, "learning_rate": 4.256657630632116e-06, "loss": 0.3733, "step": 14097 }, { "epoch": 0.7, "learning_rate": 4.255333823616805e-06, "loss": 0.5323, "step": 14098 }, { "epoch": 0.7, "learning_rate": 4.254010166845718e-06, "loss": 0.4064, "step": 14099 }, { "epoch": 0.7, "learning_rate": 4.252686660353469e-06, "loss": 0.3246, "step": 14100 }, { "epoch": 0.7, "learning_rate": 4.251363304174671e-06, "loss": 0.3213, "step": 14101 }, { "epoch": 0.7, "learning_rate": 4.250040098343935e-06, "loss": 0.2027, "step": 14102 }, { "epoch": 0.7, "learning_rate": 4.248717042895872e-06, "loss": 0.366, "step": 14103 }, { "epoch": 0.7, "learning_rate": 4.247394137865082e-06, "loss": 0.3472, "step": 14104 }, { "epoch": 0.7, "learning_rate": 4.246071383286163e-06, "loss": 0.2881, "step": 14105 }, { "epoch": 0.7, "learning_rate": 4.244748779193712e-06, "loss": 0.3783, "step": 14106 }, { "epoch": 0.7, "learning_rate": 4.243426325622317e-06, "loss": 0.2591, "step": 14107 }, { "epoch": 0.7, "learning_rate": 4.2421040226065645e-06, "loss": 0.3576, "step": 14108 }, { "epoch": 0.7, "learning_rate": 4.240781870181043e-06, "loss": 0.5416, "step": 14109 }, { "epoch": 0.7, "learning_rate": 4.2394598683803275e-06, "loss": 0.2527, "step": 14110 }, { "epoch": 0.7, "learning_rate": 4.238138017238995e-06, "loss": 0.2208, "step": 14111 }, { "epoch": 0.7, "learning_rate": 4.236816316791617e-06, "loss": 0.4334, "step": 14112 }, { "epoch": 0.7, "learning_rate": 4.235494767072759e-06, "loss": 0.2779, "step": 14113 }, { "epoch": 0.7, "learning_rate": 4.2341733681169846e-06, "loss": 0.2337, "step": 14114 }, { "epoch": 0.7, "learning_rate": 4.232852119958854e-06, "loss": 0.4913, "step": 14115 }, { "epoch": 0.7, "learning_rate": 4.23153102263292e-06, "loss": 0.4795, "step": 14116 }, { "epoch": 0.7, "learning_rate": 4.2302100761737395e-06, "loss": 0.3666, "step": 14117 }, { "epoch": 0.7, "learning_rate": 4.228889280615858e-06, "loss": 0.3559, "step": 14118 }, { "epoch": 0.7, "learning_rate": 4.227568635993818e-06, "loss": 0.3448, "step": 14119 }, { "epoch": 0.71, "learning_rate": 4.22624814234216e-06, "loss": 0.3208, "step": 14120 }, { "epoch": 0.71, "learning_rate": 4.2249277996954155e-06, "loss": 0.2938, "step": 14121 }, { "epoch": 0.71, "learning_rate": 4.223607608088125e-06, "loss": 0.4367, "step": 14122 }, { "epoch": 0.71, "learning_rate": 4.222287567554811e-06, "loss": 0.459, "step": 14123 }, { "epoch": 0.71, "learning_rate": 4.220967678129998e-06, "loss": 0.3151, "step": 14124 }, { "epoch": 0.71, "learning_rate": 4.219647939848204e-06, "loss": 0.2954, "step": 14125 }, { "epoch": 0.71, "learning_rate": 4.218328352743949e-06, "loss": 0.458, "step": 14126 }, { "epoch": 0.71, "learning_rate": 4.217008916851744e-06, "loss": 0.3588, "step": 14127 }, { "epoch": 0.71, "learning_rate": 4.215689632206097e-06, "loss": 0.5038, "step": 14128 }, { "epoch": 0.71, "learning_rate": 4.214370498841511e-06, "loss": 0.3353, "step": 14129 }, { "epoch": 0.71, "learning_rate": 4.2130515167924834e-06, "loss": 0.3688, "step": 14130 }, { "epoch": 0.71, "learning_rate": 4.211732686093518e-06, "loss": 0.2984, "step": 14131 }, { "epoch": 0.71, "learning_rate": 4.210414006779102e-06, "loss": 0.5104, "step": 14132 }, { "epoch": 0.71, "learning_rate": 4.2090954788837255e-06, "loss": 0.5646, "step": 14133 }, { "epoch": 0.71, "learning_rate": 4.207777102441872e-06, "loss": 0.3362, "step": 14134 }, { "epoch": 0.71, "learning_rate": 4.206458877488021e-06, "loss": 0.4308, "step": 14135 }, { "epoch": 0.71, "learning_rate": 4.205140804056651e-06, "loss": 0.4222, "step": 14136 }, { "epoch": 0.71, "learning_rate": 4.2038228821822325e-06, "loss": 0.3468, "step": 14137 }, { "epoch": 0.71, "learning_rate": 4.202505111899235e-06, "loss": 0.5709, "step": 14138 }, { "epoch": 0.71, "learning_rate": 4.201187493242119e-06, "loss": 0.3379, "step": 14139 }, { "epoch": 0.71, "learning_rate": 4.199870026245354e-06, "loss": 0.2419, "step": 14140 }, { "epoch": 0.71, "learning_rate": 4.19855271094339e-06, "loss": 0.2735, "step": 14141 }, { "epoch": 0.71, "learning_rate": 4.1972355473706825e-06, "loss": 0.3932, "step": 14142 }, { "epoch": 0.71, "learning_rate": 4.1959185355616785e-06, "loss": 0.4055, "step": 14143 }, { "epoch": 0.71, "learning_rate": 4.19460167555082e-06, "loss": 0.1899, "step": 14144 }, { "epoch": 0.71, "learning_rate": 4.193284967372554e-06, "loss": 0.1963, "step": 14145 }, { "epoch": 0.71, "learning_rate": 4.191968411061315e-06, "loss": 0.2327, "step": 14146 }, { "epoch": 0.71, "learning_rate": 4.1906520066515345e-06, "loss": 0.4792, "step": 14147 }, { "epoch": 0.71, "learning_rate": 4.189335754177638e-06, "loss": 0.3354, "step": 14148 }, { "epoch": 0.71, "learning_rate": 4.188019653674058e-06, "loss": 0.2382, "step": 14149 }, { "epoch": 0.71, "learning_rate": 4.1867037051752124e-06, "loss": 0.2541, "step": 14150 }, { "epoch": 0.71, "learning_rate": 4.185387908715517e-06, "loss": 0.4938, "step": 14151 }, { "epoch": 0.71, "learning_rate": 4.184072264329385e-06, "loss": 0.3504, "step": 14152 }, { "epoch": 0.71, "learning_rate": 4.182756772051222e-06, "loss": 0.2614, "step": 14153 }, { "epoch": 0.71, "learning_rate": 4.18144143191544e-06, "loss": 0.6647, "step": 14154 }, { "epoch": 0.71, "learning_rate": 4.180126243956435e-06, "loss": 0.3924, "step": 14155 }, { "epoch": 0.71, "learning_rate": 4.178811208208606e-06, "loss": 0.2048, "step": 14156 }, { "epoch": 0.71, "learning_rate": 4.177496324706347e-06, "loss": 0.4404, "step": 14157 }, { "epoch": 0.71, "learning_rate": 4.176181593484044e-06, "loss": 0.1569, "step": 14158 }, { "epoch": 0.71, "learning_rate": 4.174867014576085e-06, "loss": 0.263, "step": 14159 }, { "epoch": 0.71, "learning_rate": 4.173552588016848e-06, "loss": 0.4456, "step": 14160 }, { "epoch": 0.71, "learning_rate": 4.172238313840713e-06, "loss": 0.313, "step": 14161 }, { "epoch": 0.71, "learning_rate": 4.170924192082048e-06, "loss": 0.337, "step": 14162 }, { "epoch": 0.71, "learning_rate": 4.169610222775229e-06, "loss": 0.4858, "step": 14163 }, { "epoch": 0.71, "learning_rate": 4.16829640595462e-06, "loss": 0.4579, "step": 14164 }, { "epoch": 0.71, "learning_rate": 4.166982741654579e-06, "loss": 0.4505, "step": 14165 }, { "epoch": 0.71, "learning_rate": 4.16566922990946e-06, "loss": 0.6273, "step": 14166 }, { "epoch": 0.71, "learning_rate": 4.164355870753626e-06, "loss": 0.3237, "step": 14167 }, { "epoch": 0.71, "learning_rate": 4.163042664221419e-06, "loss": 0.2219, "step": 14168 }, { "epoch": 0.71, "learning_rate": 4.161729610347189e-06, "loss": 0.4538, "step": 14169 }, { "epoch": 0.71, "learning_rate": 4.160416709165272e-06, "loss": 0.1451, "step": 14170 }, { "epoch": 0.71, "learning_rate": 4.159103960710004e-06, "loss": 0.3955, "step": 14171 }, { "epoch": 0.71, "learning_rate": 4.157791365015725e-06, "loss": 0.2119, "step": 14172 }, { "epoch": 0.71, "learning_rate": 4.156478922116762e-06, "loss": 0.429, "step": 14173 }, { "epoch": 0.71, "learning_rate": 4.1551666320474395e-06, "loss": 0.4004, "step": 14174 }, { "epoch": 0.71, "learning_rate": 4.153854494842073e-06, "loss": 0.3621, "step": 14175 }, { "epoch": 0.71, "learning_rate": 4.15254251053499e-06, "loss": 0.3002, "step": 14176 }, { "epoch": 0.71, "learning_rate": 4.1512306791604985e-06, "loss": 0.361, "step": 14177 }, { "epoch": 0.71, "learning_rate": 4.149919000752909e-06, "loss": 0.2669, "step": 14178 }, { "epoch": 0.71, "learning_rate": 4.148607475346524e-06, "loss": 0.3151, "step": 14179 }, { "epoch": 0.71, "learning_rate": 4.147296102975648e-06, "loss": 0.397, "step": 14180 }, { "epoch": 0.71, "learning_rate": 4.145984883674573e-06, "loss": 0.4372, "step": 14181 }, { "epoch": 0.71, "learning_rate": 4.144673817477599e-06, "loss": 0.5218, "step": 14182 }, { "epoch": 0.71, "learning_rate": 4.143362904419013e-06, "loss": 0.4155, "step": 14183 }, { "epoch": 0.71, "learning_rate": 4.1420521445330986e-06, "loss": 0.3387, "step": 14184 }, { "epoch": 0.71, "learning_rate": 4.140741537854138e-06, "loss": 0.2946, "step": 14185 }, { "epoch": 0.71, "learning_rate": 4.139431084416408e-06, "loss": 0.4545, "step": 14186 }, { "epoch": 0.71, "learning_rate": 4.138120784254183e-06, "loss": 0.3531, "step": 14187 }, { "epoch": 0.71, "learning_rate": 4.13681063740173e-06, "loss": 0.3292, "step": 14188 }, { "epoch": 0.71, "learning_rate": 4.1355006438933135e-06, "loss": 0.6129, "step": 14189 }, { "epoch": 0.71, "learning_rate": 4.134190803763199e-06, "loss": 0.5933, "step": 14190 }, { "epoch": 0.71, "learning_rate": 4.1328811170456405e-06, "loss": 0.5905, "step": 14191 }, { "epoch": 0.71, "learning_rate": 4.131571583774892e-06, "loss": 0.3532, "step": 14192 }, { "epoch": 0.71, "learning_rate": 4.130262203985204e-06, "loss": 0.2952, "step": 14193 }, { "epoch": 0.71, "learning_rate": 4.128952977710815e-06, "loss": 0.3568, "step": 14194 }, { "epoch": 0.71, "learning_rate": 4.127643904985975e-06, "loss": 0.3962, "step": 14195 }, { "epoch": 0.71, "learning_rate": 4.126334985844918e-06, "loss": 0.3646, "step": 14196 }, { "epoch": 0.71, "learning_rate": 4.125026220321876e-06, "loss": 0.32, "step": 14197 }, { "epoch": 0.71, "learning_rate": 4.123717608451074e-06, "loss": 0.5278, "step": 14198 }, { "epoch": 0.71, "learning_rate": 4.122409150266746e-06, "loss": 0.4032, "step": 14199 }, { "epoch": 0.71, "learning_rate": 4.121100845803107e-06, "loss": 0.2723, "step": 14200 }, { "epoch": 0.71, "learning_rate": 4.1197926950943775e-06, "loss": 0.4452, "step": 14201 }, { "epoch": 0.71, "learning_rate": 4.118484698174767e-06, "loss": 0.5822, "step": 14202 }, { "epoch": 0.71, "learning_rate": 4.117176855078483e-06, "loss": 0.4238, "step": 14203 }, { "epoch": 0.71, "learning_rate": 4.115869165839736e-06, "loss": 0.2105, "step": 14204 }, { "epoch": 0.71, "learning_rate": 4.114561630492725e-06, "loss": 0.3655, "step": 14205 }, { "epoch": 0.71, "learning_rate": 4.113254249071645e-06, "loss": 0.274, "step": 14206 }, { "epoch": 0.71, "learning_rate": 4.11194702161069e-06, "loss": 0.3019, "step": 14207 }, { "epoch": 0.71, "learning_rate": 4.1106399481440495e-06, "loss": 0.3703, "step": 14208 }, { "epoch": 0.71, "learning_rate": 4.109333028705908e-06, "loss": 0.3314, "step": 14209 }, { "epoch": 0.71, "learning_rate": 4.108026263330445e-06, "loss": 0.4895, "step": 14210 }, { "epoch": 0.71, "learning_rate": 4.106719652051837e-06, "loss": 0.3358, "step": 14211 }, { "epoch": 0.71, "learning_rate": 4.105413194904255e-06, "loss": 0.3416, "step": 14212 }, { "epoch": 0.71, "learning_rate": 4.104106891921874e-06, "loss": 0.3708, "step": 14213 }, { "epoch": 0.71, "learning_rate": 4.102800743138855e-06, "loss": 0.3947, "step": 14214 }, { "epoch": 0.71, "learning_rate": 4.101494748589358e-06, "loss": 0.3523, "step": 14215 }, { "epoch": 0.71, "learning_rate": 4.100188908307541e-06, "loss": 0.5911, "step": 14216 }, { "epoch": 0.71, "learning_rate": 4.098883222327552e-06, "loss": 0.3596, "step": 14217 }, { "epoch": 0.71, "learning_rate": 4.097577690683546e-06, "loss": 0.3844, "step": 14218 }, { "epoch": 0.71, "learning_rate": 4.096272313409666e-06, "loss": 0.3402, "step": 14219 }, { "epoch": 0.71, "learning_rate": 4.094967090540051e-06, "loss": 0.3406, "step": 14220 }, { "epoch": 0.71, "learning_rate": 4.093662022108833e-06, "loss": 0.3766, "step": 14221 }, { "epoch": 0.71, "learning_rate": 4.092357108150153e-06, "loss": 0.2847, "step": 14222 }, { "epoch": 0.71, "learning_rate": 4.091052348698136e-06, "loss": 0.3896, "step": 14223 }, { "epoch": 0.71, "learning_rate": 4.089747743786904e-06, "loss": 0.1934, "step": 14224 }, { "epoch": 0.71, "learning_rate": 4.088443293450579e-06, "loss": 0.2839, "step": 14225 }, { "epoch": 0.71, "learning_rate": 4.087138997723275e-06, "loss": 0.6544, "step": 14226 }, { "epoch": 0.71, "learning_rate": 4.0858348566391085e-06, "loss": 0.2437, "step": 14227 }, { "epoch": 0.71, "learning_rate": 4.0845308702321836e-06, "loss": 0.2732, "step": 14228 }, { "epoch": 0.71, "learning_rate": 4.083227038536608e-06, "loss": 0.3505, "step": 14229 }, { "epoch": 0.71, "learning_rate": 4.081923361586478e-06, "loss": 0.3573, "step": 14230 }, { "epoch": 0.71, "learning_rate": 4.080619839415893e-06, "loss": 0.2748, "step": 14231 }, { "epoch": 0.71, "learning_rate": 4.079316472058943e-06, "loss": 0.2506, "step": 14232 }, { "epoch": 0.71, "learning_rate": 4.078013259549715e-06, "loss": 0.3701, "step": 14233 }, { "epoch": 0.71, "learning_rate": 4.076710201922294e-06, "loss": 0.3549, "step": 14234 }, { "epoch": 0.71, "learning_rate": 4.075407299210755e-06, "loss": 0.6028, "step": 14235 }, { "epoch": 0.71, "learning_rate": 4.074104551449184e-06, "loss": 0.4088, "step": 14236 }, { "epoch": 0.71, "learning_rate": 4.072801958671645e-06, "loss": 0.3828, "step": 14237 }, { "epoch": 0.71, "learning_rate": 4.071499520912209e-06, "loss": 0.2137, "step": 14238 }, { "epoch": 0.71, "learning_rate": 4.070197238204934e-06, "loss": 0.2846, "step": 14239 }, { "epoch": 0.71, "learning_rate": 4.068895110583887e-06, "loss": 0.3481, "step": 14240 }, { "epoch": 0.71, "learning_rate": 4.0675931380831195e-06, "loss": 0.6611, "step": 14241 }, { "epoch": 0.71, "learning_rate": 4.066291320736684e-06, "loss": 0.4825, "step": 14242 }, { "epoch": 0.71, "learning_rate": 4.064989658578627e-06, "loss": 0.3929, "step": 14243 }, { "epoch": 0.71, "learning_rate": 4.063688151642989e-06, "loss": 0.231, "step": 14244 }, { "epoch": 0.71, "learning_rate": 4.062386799963815e-06, "loss": 0.646, "step": 14245 }, { "epoch": 0.71, "learning_rate": 4.0610856035751374e-06, "loss": 0.3591, "step": 14246 }, { "epoch": 0.71, "learning_rate": 4.059784562510989e-06, "loss": 0.3294, "step": 14247 }, { "epoch": 0.71, "learning_rate": 4.0584836768053935e-06, "loss": 0.2617, "step": 14248 }, { "epoch": 0.71, "learning_rate": 4.057182946492371e-06, "loss": 0.3948, "step": 14249 }, { "epoch": 0.71, "learning_rate": 4.0558823716059505e-06, "loss": 0.2863, "step": 14250 }, { "epoch": 0.71, "learning_rate": 4.05458195218014e-06, "loss": 0.614, "step": 14251 }, { "epoch": 0.71, "learning_rate": 4.053281688248951e-06, "loss": 0.4563, "step": 14252 }, { "epoch": 0.71, "learning_rate": 4.051981579846388e-06, "loss": 0.2842, "step": 14253 }, { "epoch": 0.71, "learning_rate": 4.05068162700646e-06, "loss": 0.5227, "step": 14254 }, { "epoch": 0.71, "learning_rate": 4.049381829763161e-06, "loss": 0.4851, "step": 14255 }, { "epoch": 0.71, "learning_rate": 4.048082188150487e-06, "loss": 0.2949, "step": 14256 }, { "epoch": 0.71, "learning_rate": 4.046782702202427e-06, "loss": 0.302, "step": 14257 }, { "epoch": 0.71, "learning_rate": 4.045483371952969e-06, "loss": 0.4315, "step": 14258 }, { "epoch": 0.71, "learning_rate": 4.044184197436094e-06, "loss": 0.443, "step": 14259 }, { "epoch": 0.71, "learning_rate": 4.04288517868578e-06, "loss": 0.4534, "step": 14260 }, { "epoch": 0.71, "learning_rate": 4.041586315736004e-06, "loss": 0.1695, "step": 14261 }, { "epoch": 0.71, "learning_rate": 4.040287608620728e-06, "loss": 0.3191, "step": 14262 }, { "epoch": 0.71, "learning_rate": 4.038989057373928e-06, "loss": 0.2072, "step": 14263 }, { "epoch": 0.71, "learning_rate": 4.037690662029561e-06, "loss": 0.4671, "step": 14264 }, { "epoch": 0.71, "learning_rate": 4.0363924226215865e-06, "loss": 0.3551, "step": 14265 }, { "epoch": 0.71, "learning_rate": 4.035094339183957e-06, "loss": 0.356, "step": 14266 }, { "epoch": 0.71, "learning_rate": 4.033796411750618e-06, "loss": 0.2629, "step": 14267 }, { "epoch": 0.71, "learning_rate": 4.032498640355524e-06, "loss": 0.2327, "step": 14268 }, { "epoch": 0.71, "learning_rate": 4.031201025032612e-06, "loss": 0.2233, "step": 14269 }, { "epoch": 0.71, "learning_rate": 4.029903565815819e-06, "loss": 0.4477, "step": 14270 }, { "epoch": 0.71, "learning_rate": 4.028606262739075e-06, "loss": 0.229, "step": 14271 }, { "epoch": 0.71, "learning_rate": 4.0273091158363185e-06, "loss": 0.5284, "step": 14272 }, { "epoch": 0.71, "learning_rate": 4.026012125141467e-06, "loss": 0.42, "step": 14273 }, { "epoch": 0.71, "learning_rate": 4.0247152906884455e-06, "loss": 0.523, "step": 14274 }, { "epoch": 0.71, "learning_rate": 4.023418612511169e-06, "loss": 0.3414, "step": 14275 }, { "epoch": 0.71, "learning_rate": 4.022122090643546e-06, "loss": 0.2074, "step": 14276 }, { "epoch": 0.71, "learning_rate": 4.020825725119495e-06, "loss": 0.3985, "step": 14277 }, { "epoch": 0.71, "learning_rate": 4.019529515972915e-06, "loss": 0.4048, "step": 14278 }, { "epoch": 0.71, "learning_rate": 4.018233463237709e-06, "loss": 0.2608, "step": 14279 }, { "epoch": 0.71, "learning_rate": 4.01693756694777e-06, "loss": 0.5354, "step": 14280 }, { "epoch": 0.71, "learning_rate": 4.0156418271369935e-06, "loss": 0.2581, "step": 14281 }, { "epoch": 0.71, "learning_rate": 4.014346243839267e-06, "loss": 0.3495, "step": 14282 }, { "epoch": 0.71, "learning_rate": 4.0130508170884755e-06, "loss": 0.3817, "step": 14283 }, { "epoch": 0.71, "learning_rate": 4.011755546918497e-06, "loss": 0.3403, "step": 14284 }, { "epoch": 0.71, "learning_rate": 4.010460433363206e-06, "loss": 0.368, "step": 14285 }, { "epoch": 0.71, "learning_rate": 4.0091654764564815e-06, "loss": 0.4572, "step": 14286 }, { "epoch": 0.71, "learning_rate": 4.007870676232186e-06, "loss": 0.3965, "step": 14287 }, { "epoch": 0.71, "learning_rate": 4.006576032724185e-06, "loss": 0.3944, "step": 14288 }, { "epoch": 0.71, "learning_rate": 4.005281545966338e-06, "loss": 0.2458, "step": 14289 }, { "epoch": 0.71, "learning_rate": 4.003987215992499e-06, "loss": 0.3328, "step": 14290 }, { "epoch": 0.71, "learning_rate": 4.002693042836522e-06, "loss": 0.5156, "step": 14291 }, { "epoch": 0.71, "learning_rate": 4.001399026532254e-06, "loss": 0.3921, "step": 14292 }, { "epoch": 0.71, "learning_rate": 4.000105167113539e-06, "loss": 0.1841, "step": 14293 }, { "epoch": 0.71, "learning_rate": 3.99881146461421e-06, "loss": 0.321, "step": 14294 }, { "epoch": 0.71, "learning_rate": 3.997517919068111e-06, "loss": 0.2372, "step": 14295 }, { "epoch": 0.71, "learning_rate": 3.996224530509068e-06, "loss": 0.4049, "step": 14296 }, { "epoch": 0.71, "learning_rate": 3.994931298970911e-06, "loss": 0.3759, "step": 14297 }, { "epoch": 0.71, "learning_rate": 3.99363822448746e-06, "loss": 0.2467, "step": 14298 }, { "epoch": 0.71, "learning_rate": 3.992345307092529e-06, "loss": 0.215, "step": 14299 }, { "epoch": 0.71, "learning_rate": 3.9910525468199425e-06, "loss": 0.2844, "step": 14300 }, { "epoch": 0.71, "learning_rate": 3.989759943703507e-06, "loss": 0.2133, "step": 14301 }, { "epoch": 0.71, "learning_rate": 3.9884674977770275e-06, "loss": 0.1144, "step": 14302 }, { "epoch": 0.71, "learning_rate": 3.987175209074307e-06, "loss": 0.5163, "step": 14303 }, { "epoch": 0.71, "learning_rate": 3.985883077629144e-06, "loss": 0.3316, "step": 14304 }, { "epoch": 0.71, "learning_rate": 3.9845911034753315e-06, "loss": 0.4196, "step": 14305 }, { "epoch": 0.71, "learning_rate": 3.9832992866466595e-06, "loss": 0.3721, "step": 14306 }, { "epoch": 0.71, "learning_rate": 3.982007627176915e-06, "loss": 0.4032, "step": 14307 }, { "epoch": 0.71, "learning_rate": 3.980716125099875e-06, "loss": 0.3178, "step": 14308 }, { "epoch": 0.71, "learning_rate": 3.979424780449325e-06, "loss": 0.3115, "step": 14309 }, { "epoch": 0.71, "learning_rate": 3.978133593259032e-06, "loss": 0.1549, "step": 14310 }, { "epoch": 0.71, "learning_rate": 3.9768425635627705e-06, "loss": 0.384, "step": 14311 }, { "epoch": 0.71, "learning_rate": 3.975551691394298e-06, "loss": 0.4412, "step": 14312 }, { "epoch": 0.71, "learning_rate": 3.974260976787384e-06, "loss": 0.5131, "step": 14313 }, { "epoch": 0.71, "learning_rate": 3.9729704197757825e-06, "loss": 0.3998, "step": 14314 }, { "epoch": 0.71, "learning_rate": 3.971680020393245e-06, "loss": 0.3328, "step": 14315 }, { "epoch": 0.71, "learning_rate": 3.970389778673521e-06, "loss": 0.4004, "step": 14316 }, { "epoch": 0.71, "learning_rate": 3.969099694650352e-06, "loss": 0.439, "step": 14317 }, { "epoch": 0.71, "learning_rate": 3.967809768357484e-06, "loss": 0.7074, "step": 14318 }, { "epoch": 0.71, "learning_rate": 3.966519999828651e-06, "loss": 0.4206, "step": 14319 }, { "epoch": 0.72, "learning_rate": 3.965230389097586e-06, "loss": 0.3617, "step": 14320 }, { "epoch": 0.72, "learning_rate": 3.963940936198014e-06, "loss": 0.3851, "step": 14321 }, { "epoch": 0.72, "learning_rate": 3.96265164116366e-06, "loss": 0.1929, "step": 14322 }, { "epoch": 0.72, "learning_rate": 3.961362504028246e-06, "loss": 0.4017, "step": 14323 }, { "epoch": 0.72, "learning_rate": 3.9600735248254875e-06, "loss": 0.3621, "step": 14324 }, { "epoch": 0.72, "learning_rate": 3.9587847035890955e-06, "loss": 0.3601, "step": 14325 }, { "epoch": 0.72, "learning_rate": 3.957496040352773e-06, "loss": 0.2838, "step": 14326 }, { "epoch": 0.72, "learning_rate": 3.956207535150232e-06, "loss": 0.2943, "step": 14327 }, { "epoch": 0.72, "learning_rate": 3.954919188015165e-06, "loss": 0.4682, "step": 14328 }, { "epoch": 0.72, "learning_rate": 3.9536309989812705e-06, "loss": 0.2149, "step": 14329 }, { "epoch": 0.72, "learning_rate": 3.952342968082237e-06, "loss": 0.2546, "step": 14330 }, { "epoch": 0.72, "learning_rate": 3.951055095351752e-06, "loss": 0.3142, "step": 14331 }, { "epoch": 0.72, "learning_rate": 3.9497673808234995e-06, "loss": 0.5232, "step": 14332 }, { "epoch": 0.72, "learning_rate": 3.948479824531155e-06, "loss": 0.5021, "step": 14333 }, { "epoch": 0.72, "learning_rate": 3.947192426508396e-06, "loss": 0.5287, "step": 14334 }, { "epoch": 0.72, "learning_rate": 3.945905186788887e-06, "loss": 0.5479, "step": 14335 }, { "epoch": 0.72, "learning_rate": 3.9446181054063025e-06, "loss": 0.2598, "step": 14336 }, { "epoch": 0.72, "learning_rate": 3.943331182394301e-06, "loss": 0.4636, "step": 14337 }, { "epoch": 0.72, "learning_rate": 3.942044417786539e-06, "loss": 0.3632, "step": 14338 }, { "epoch": 0.72, "learning_rate": 3.94075781161667e-06, "loss": 0.242, "step": 14339 }, { "epoch": 0.72, "learning_rate": 3.939471363918341e-06, "loss": 0.2874, "step": 14340 }, { "epoch": 0.72, "learning_rate": 3.938185074725205e-06, "loss": 0.112, "step": 14341 }, { "epoch": 0.72, "learning_rate": 3.936898944070898e-06, "loss": 0.2714, "step": 14342 }, { "epoch": 0.72, "learning_rate": 3.935612971989059e-06, "loss": 0.4082, "step": 14343 }, { "epoch": 0.72, "learning_rate": 3.934327158513315e-06, "loss": 0.2946, "step": 14344 }, { "epoch": 0.72, "learning_rate": 3.933041503677304e-06, "loss": 0.3749, "step": 14345 }, { "epoch": 0.72, "learning_rate": 3.931756007514646e-06, "loss": 0.556, "step": 14346 }, { "epoch": 0.72, "learning_rate": 3.93047067005896e-06, "loss": 0.2676, "step": 14347 }, { "epoch": 0.72, "learning_rate": 3.929185491343866e-06, "loss": 0.374, "step": 14348 }, { "epoch": 0.72, "learning_rate": 3.92790047140297e-06, "loss": 0.3082, "step": 14349 }, { "epoch": 0.72, "learning_rate": 3.926615610269887e-06, "loss": 0.1246, "step": 14350 }, { "epoch": 0.72, "learning_rate": 3.925330907978219e-06, "loss": 0.5968, "step": 14351 }, { "epoch": 0.72, "learning_rate": 3.924046364561564e-06, "loss": 0.4297, "step": 14352 }, { "epoch": 0.72, "learning_rate": 3.922761980053519e-06, "loss": 0.5683, "step": 14353 }, { "epoch": 0.72, "learning_rate": 3.921477754487675e-06, "loss": 0.487, "step": 14354 }, { "epoch": 0.72, "learning_rate": 3.920193687897618e-06, "loss": 0.4229, "step": 14355 }, { "epoch": 0.72, "learning_rate": 3.918909780316932e-06, "loss": 0.245, "step": 14356 }, { "epoch": 0.72, "learning_rate": 3.917626031779196e-06, "loss": 0.5828, "step": 14357 }, { "epoch": 0.72, "learning_rate": 3.916342442317981e-06, "loss": 0.3431, "step": 14358 }, { "epoch": 0.72, "learning_rate": 3.915059011966865e-06, "loss": 0.3423, "step": 14359 }, { "epoch": 0.72, "learning_rate": 3.91377574075941e-06, "loss": 0.3041, "step": 14360 }, { "epoch": 0.72, "learning_rate": 3.9124926287291795e-06, "loss": 0.3094, "step": 14361 }, { "epoch": 0.72, "learning_rate": 3.9112096759097316e-06, "loss": 0.2347, "step": 14362 }, { "epoch": 0.72, "learning_rate": 3.909926882334616e-06, "loss": 0.2501, "step": 14363 }, { "epoch": 0.72, "learning_rate": 3.9086442480373895e-06, "loss": 0.4439, "step": 14364 }, { "epoch": 0.72, "learning_rate": 3.907361773051595e-06, "loss": 0.2925, "step": 14365 }, { "epoch": 0.72, "learning_rate": 3.906079457410773e-06, "loss": 0.4813, "step": 14366 }, { "epoch": 0.72, "learning_rate": 3.9047973011484575e-06, "loss": 0.4346, "step": 14367 }, { "epoch": 0.72, "learning_rate": 3.903515304298189e-06, "loss": 0.4116, "step": 14368 }, { "epoch": 0.72, "learning_rate": 3.9022334668934924e-06, "loss": 0.2027, "step": 14369 }, { "epoch": 0.72, "learning_rate": 3.900951788967894e-06, "loss": 0.4583, "step": 14370 }, { "epoch": 0.72, "learning_rate": 3.899670270554912e-06, "loss": 0.4313, "step": 14371 }, { "epoch": 0.72, "learning_rate": 3.898388911688062e-06, "loss": 0.5589, "step": 14372 }, { "epoch": 0.72, "learning_rate": 3.8971077124008596e-06, "loss": 0.4642, "step": 14373 }, { "epoch": 0.72, "learning_rate": 3.895826672726814e-06, "loss": 0.3483, "step": 14374 }, { "epoch": 0.72, "learning_rate": 3.8945457926994246e-06, "loss": 0.2648, "step": 14375 }, { "epoch": 0.72, "learning_rate": 3.893265072352195e-06, "loss": 0.7598, "step": 14376 }, { "epoch": 0.72, "learning_rate": 3.891984511718617e-06, "loss": 0.3524, "step": 14377 }, { "epoch": 0.72, "learning_rate": 3.890704110832187e-06, "loss": 0.2773, "step": 14378 }, { "epoch": 0.72, "learning_rate": 3.889423869726386e-06, "loss": 0.4053, "step": 14379 }, { "epoch": 0.72, "learning_rate": 3.8881437884347025e-06, "loss": 0.3788, "step": 14380 }, { "epoch": 0.72, "learning_rate": 3.886863866990609e-06, "loss": 0.3964, "step": 14381 }, { "epoch": 0.72, "learning_rate": 3.885584105427588e-06, "loss": 0.4698, "step": 14382 }, { "epoch": 0.72, "learning_rate": 3.884304503779106e-06, "loss": 0.3448, "step": 14383 }, { "epoch": 0.72, "learning_rate": 3.883025062078629e-06, "loss": 0.2713, "step": 14384 }, { "epoch": 0.72, "learning_rate": 3.881745780359621e-06, "loss": 0.3726, "step": 14385 }, { "epoch": 0.72, "learning_rate": 3.880466658655534e-06, "loss": 0.4208, "step": 14386 }, { "epoch": 0.72, "learning_rate": 3.879187696999831e-06, "loss": 0.3261, "step": 14387 }, { "epoch": 0.72, "learning_rate": 3.8779088954259554e-06, "loss": 0.3817, "step": 14388 }, { "epoch": 0.72, "learning_rate": 3.876630253967355e-06, "loss": 0.3286, "step": 14389 }, { "epoch": 0.72, "learning_rate": 3.875351772657467e-06, "loss": 0.4292, "step": 14390 }, { "epoch": 0.72, "learning_rate": 3.8740734515297355e-06, "loss": 0.3388, "step": 14391 }, { "epoch": 0.72, "learning_rate": 3.872795290617588e-06, "loss": 0.3374, "step": 14392 }, { "epoch": 0.72, "learning_rate": 3.871517289954455e-06, "loss": 0.2752, "step": 14393 }, { "epoch": 0.72, "learning_rate": 3.870239449573761e-06, "loss": 0.2494, "step": 14394 }, { "epoch": 0.72, "learning_rate": 3.868961769508921e-06, "loss": 0.3503, "step": 14395 }, { "epoch": 0.72, "learning_rate": 3.86768424979336e-06, "loss": 0.3757, "step": 14396 }, { "epoch": 0.72, "learning_rate": 3.866406890460485e-06, "loss": 0.254, "step": 14397 }, { "epoch": 0.72, "learning_rate": 3.865129691543705e-06, "loss": 0.2873, "step": 14398 }, { "epoch": 0.72, "learning_rate": 3.863852653076419e-06, "loss": 0.3148, "step": 14399 }, { "epoch": 0.72, "learning_rate": 3.8625757750920335e-06, "loss": 0.3925, "step": 14400 }, { "epoch": 0.72, "learning_rate": 3.86129905762394e-06, "loss": 0.1959, "step": 14401 }, { "epoch": 0.72, "learning_rate": 3.8600225007055295e-06, "loss": 0.2672, "step": 14402 }, { "epoch": 0.72, "learning_rate": 3.858746104370188e-06, "loss": 0.2747, "step": 14403 }, { "epoch": 0.72, "learning_rate": 3.857469868651298e-06, "loss": 0.2479, "step": 14404 }, { "epoch": 0.72, "learning_rate": 3.856193793582239e-06, "loss": 0.2489, "step": 14405 }, { "epoch": 0.72, "learning_rate": 3.854917879196385e-06, "loss": 0.3855, "step": 14406 }, { "epoch": 0.72, "learning_rate": 3.853642125527104e-06, "loss": 0.4213, "step": 14407 }, { "epoch": 0.72, "learning_rate": 3.8523665326077604e-06, "loss": 0.4405, "step": 14408 }, { "epoch": 0.72, "learning_rate": 3.8510911004717226e-06, "loss": 0.2851, "step": 14409 }, { "epoch": 0.72, "learning_rate": 3.849815829152341e-06, "loss": 0.2827, "step": 14410 }, { "epoch": 0.72, "learning_rate": 3.848540718682973e-06, "loss": 0.4358, "step": 14411 }, { "epoch": 0.72, "learning_rate": 3.847265769096965e-06, "loss": 0.2524, "step": 14412 }, { "epoch": 0.72, "learning_rate": 3.8459909804276585e-06, "loss": 0.2621, "step": 14413 }, { "epoch": 0.72, "learning_rate": 3.844716352708401e-06, "loss": 0.2853, "step": 14414 }, { "epoch": 0.72, "learning_rate": 3.843441885972526e-06, "loss": 0.3389, "step": 14415 }, { "epoch": 0.72, "learning_rate": 3.8421675802533645e-06, "loss": 0.4569, "step": 14416 }, { "epoch": 0.72, "learning_rate": 3.840893435584244e-06, "loss": 0.3727, "step": 14417 }, { "epoch": 0.72, "learning_rate": 3.839619451998485e-06, "loss": 0.3503, "step": 14418 }, { "epoch": 0.72, "learning_rate": 3.8383456295294155e-06, "loss": 0.2063, "step": 14419 }, { "epoch": 0.72, "learning_rate": 3.837071968210345e-06, "loss": 0.2646, "step": 14420 }, { "epoch": 0.72, "learning_rate": 3.835798468074585e-06, "loss": 0.539, "step": 14421 }, { "epoch": 0.72, "learning_rate": 3.834525129155439e-06, "loss": 0.4412, "step": 14422 }, { "epoch": 0.72, "learning_rate": 3.8332519514862165e-06, "loss": 0.2167, "step": 14423 }, { "epoch": 0.72, "learning_rate": 3.831978935100213e-06, "loss": 0.2939, "step": 14424 }, { "epoch": 0.72, "learning_rate": 3.830706080030721e-06, "loss": 0.4014, "step": 14425 }, { "epoch": 0.72, "learning_rate": 3.829433386311031e-06, "loss": 0.4801, "step": 14426 }, { "epoch": 0.72, "learning_rate": 3.828160853974429e-06, "loss": 0.6491, "step": 14427 }, { "epoch": 0.72, "learning_rate": 3.826888483054196e-06, "loss": 0.1566, "step": 14428 }, { "epoch": 0.72, "learning_rate": 3.82561627358361e-06, "loss": 0.3949, "step": 14429 }, { "epoch": 0.72, "learning_rate": 3.8243442255959425e-06, "loss": 0.4106, "step": 14430 }, { "epoch": 0.72, "learning_rate": 3.823072339124461e-06, "loss": 0.4896, "step": 14431 }, { "epoch": 0.72, "learning_rate": 3.821800614202434e-06, "loss": 0.2209, "step": 14432 }, { "epoch": 0.72, "learning_rate": 3.820529050863121e-06, "loss": 0.3232, "step": 14433 }, { "epoch": 0.72, "learning_rate": 3.819257649139777e-06, "loss": 0.3862, "step": 14434 }, { "epoch": 0.72, "learning_rate": 3.8179864090656525e-06, "loss": 0.4537, "step": 14435 }, { "epoch": 0.72, "learning_rate": 3.816715330673995e-06, "loss": 0.242, "step": 14436 }, { "epoch": 0.72, "learning_rate": 3.815444413998052e-06, "loss": 0.4475, "step": 14437 }, { "epoch": 0.72, "learning_rate": 3.81417365907106e-06, "loss": 0.5619, "step": 14438 }, { "epoch": 0.72, "learning_rate": 3.8129030659262533e-06, "loss": 0.2259, "step": 14439 }, { "epoch": 0.72, "learning_rate": 3.8116326345968614e-06, "loss": 0.3032, "step": 14440 }, { "epoch": 0.72, "learning_rate": 3.810362365116115e-06, "loss": 0.5209, "step": 14441 }, { "epoch": 0.72, "learning_rate": 3.809092257517234e-06, "loss": 0.271, "step": 14442 }, { "epoch": 0.72, "learning_rate": 3.8078223118334367e-06, "loss": 0.3881, "step": 14443 }, { "epoch": 0.72, "learning_rate": 3.8065525280979363e-06, "loss": 0.3797, "step": 14444 }, { "epoch": 0.72, "learning_rate": 3.805282906343939e-06, "loss": 0.1566, "step": 14445 }, { "epoch": 0.72, "learning_rate": 3.8040134466046576e-06, "loss": 0.5551, "step": 14446 }, { "epoch": 0.72, "learning_rate": 3.802744148913289e-06, "loss": 0.3892, "step": 14447 }, { "epoch": 0.72, "learning_rate": 3.8014750133030297e-06, "loss": 0.4719, "step": 14448 }, { "epoch": 0.72, "learning_rate": 3.800206039807073e-06, "loss": 0.416, "step": 14449 }, { "epoch": 0.72, "learning_rate": 3.798937228458608e-06, "loss": 0.2313, "step": 14450 }, { "epoch": 0.72, "learning_rate": 3.797668579290816e-06, "loss": 0.1814, "step": 14451 }, { "epoch": 0.72, "learning_rate": 3.7964000923368804e-06, "loss": 0.4836, "step": 14452 }, { "epoch": 0.72, "learning_rate": 3.795131767629975e-06, "loss": 0.3429, "step": 14453 }, { "epoch": 0.72, "learning_rate": 3.7938636052032674e-06, "loss": 0.5238, "step": 14454 }, { "epoch": 0.72, "learning_rate": 3.792595605089933e-06, "loss": 0.4669, "step": 14455 }, { "epoch": 0.72, "learning_rate": 3.7913277673231287e-06, "loss": 0.4304, "step": 14456 }, { "epoch": 0.72, "learning_rate": 3.790060091936015e-06, "loss": 0.4973, "step": 14457 }, { "epoch": 0.72, "learning_rate": 3.7887925789617475e-06, "loss": 0.2345, "step": 14458 }, { "epoch": 0.72, "learning_rate": 3.7875252284334697e-06, "loss": 0.5359, "step": 14459 }, { "epoch": 0.72, "learning_rate": 3.7862580403843373e-06, "loss": 0.3149, "step": 14460 }, { "epoch": 0.72, "learning_rate": 3.7849910148474865e-06, "loss": 0.3879, "step": 14461 }, { "epoch": 0.72, "learning_rate": 3.7837241518560563e-06, "loss": 0.4942, "step": 14462 }, { "epoch": 0.72, "learning_rate": 3.7824574514431754e-06, "loss": 0.3052, "step": 14463 }, { "epoch": 0.72, "learning_rate": 3.781190913641979e-06, "loss": 0.5207, "step": 14464 }, { "epoch": 0.72, "learning_rate": 3.77992453848559e-06, "loss": 0.2709, "step": 14465 }, { "epoch": 0.72, "learning_rate": 3.7786583260071275e-06, "loss": 0.4162, "step": 14466 }, { "epoch": 0.72, "learning_rate": 3.7773922762397086e-06, "loss": 0.2586, "step": 14467 }, { "epoch": 0.72, "learning_rate": 3.7761263892164403e-06, "loss": 0.3828, "step": 14468 }, { "epoch": 0.72, "learning_rate": 3.7748606649704377e-06, "loss": 0.3252, "step": 14469 }, { "epoch": 0.72, "learning_rate": 3.7735951035348017e-06, "loss": 0.394, "step": 14470 }, { "epoch": 0.72, "learning_rate": 3.77232970494263e-06, "loss": 0.552, "step": 14471 }, { "epoch": 0.72, "learning_rate": 3.771064469227015e-06, "loss": 0.2588, "step": 14472 }, { "epoch": 0.72, "learning_rate": 3.7697993964210544e-06, "loss": 0.3019, "step": 14473 }, { "epoch": 0.72, "learning_rate": 3.76853448655783e-06, "loss": 0.246, "step": 14474 }, { "epoch": 0.72, "learning_rate": 3.767269739670424e-06, "loss": 0.2913, "step": 14475 }, { "epoch": 0.72, "learning_rate": 3.7660051557919154e-06, "loss": 0.4057, "step": 14476 }, { "epoch": 0.72, "learning_rate": 3.7647407349553755e-06, "loss": 0.3172, "step": 14477 }, { "epoch": 0.72, "learning_rate": 3.763476477193876e-06, "loss": 0.345, "step": 14478 }, { "epoch": 0.72, "learning_rate": 3.762212382540481e-06, "loss": 0.4155, "step": 14479 }, { "epoch": 0.72, "learning_rate": 3.76094845102825e-06, "loss": 0.595, "step": 14480 }, { "epoch": 0.72, "learning_rate": 3.7596846826902377e-06, "loss": 0.4421, "step": 14481 }, { "epoch": 0.72, "learning_rate": 3.758421077559503e-06, "loss": 0.3044, "step": 14482 }, { "epoch": 0.72, "learning_rate": 3.75715763566909e-06, "loss": 0.266, "step": 14483 }, { "epoch": 0.72, "learning_rate": 3.755894357052042e-06, "loss": 0.3878, "step": 14484 }, { "epoch": 0.72, "learning_rate": 3.7546312417413988e-06, "loss": 0.5407, "step": 14485 }, { "epoch": 0.72, "learning_rate": 3.7533682897701926e-06, "loss": 0.5169, "step": 14486 }, { "epoch": 0.72, "learning_rate": 3.75210550117146e-06, "loss": 0.5191, "step": 14487 }, { "epoch": 0.72, "learning_rate": 3.7508428759782256e-06, "loss": 0.6481, "step": 14488 }, { "epoch": 0.72, "learning_rate": 3.749580414223509e-06, "loss": 0.3547, "step": 14489 }, { "epoch": 0.72, "learning_rate": 3.7483181159403314e-06, "loss": 0.2527, "step": 14490 }, { "epoch": 0.72, "learning_rate": 3.7470559811617024e-06, "loss": 0.5395, "step": 14491 }, { "epoch": 0.72, "learning_rate": 3.745794009920637e-06, "loss": 0.2298, "step": 14492 }, { "epoch": 0.72, "learning_rate": 3.7445322022501373e-06, "loss": 0.3234, "step": 14493 }, { "epoch": 0.72, "learning_rate": 3.743270558183205e-06, "loss": 0.257, "step": 14494 }, { "epoch": 0.72, "learning_rate": 3.742009077752834e-06, "loss": 0.3228, "step": 14495 }, { "epoch": 0.72, "learning_rate": 3.7407477609920218e-06, "loss": 0.527, "step": 14496 }, { "epoch": 0.72, "learning_rate": 3.739486607933753e-06, "loss": 0.5706, "step": 14497 }, { "epoch": 0.72, "learning_rate": 3.738225618611013e-06, "loss": 0.3777, "step": 14498 }, { "epoch": 0.72, "learning_rate": 3.7369647930567796e-06, "loss": 0.492, "step": 14499 }, { "epoch": 0.72, "learning_rate": 3.7357041313040297e-06, "loss": 0.3419, "step": 14500 }, { "epoch": 0.72, "learning_rate": 3.7344436333857337e-06, "loss": 0.2697, "step": 14501 }, { "epoch": 0.72, "learning_rate": 3.733183299334858e-06, "loss": 0.2141, "step": 14502 }, { "epoch": 0.72, "learning_rate": 3.731923129184365e-06, "loss": 0.3207, "step": 14503 }, { "epoch": 0.72, "learning_rate": 3.7306631229672095e-06, "loss": 0.5955, "step": 14504 }, { "epoch": 0.72, "learning_rate": 3.7294032807163517e-06, "loss": 0.5027, "step": 14505 }, { "epoch": 0.72, "learning_rate": 3.7281436024647377e-06, "loss": 0.4676, "step": 14506 }, { "epoch": 0.72, "learning_rate": 3.726884088245314e-06, "loss": 0.441, "step": 14507 }, { "epoch": 0.72, "learning_rate": 3.7256247380910206e-06, "loss": 0.3813, "step": 14508 }, { "epoch": 0.72, "learning_rate": 3.72436555203479e-06, "loss": 0.2956, "step": 14509 }, { "epoch": 0.72, "learning_rate": 3.7231065301095626e-06, "loss": 0.3089, "step": 14510 }, { "epoch": 0.72, "learning_rate": 3.7218476723482633e-06, "loss": 0.428, "step": 14511 }, { "epoch": 0.72, "learning_rate": 3.7205889787838147e-06, "loss": 0.2377, "step": 14512 }, { "epoch": 0.72, "learning_rate": 3.7193304494491332e-06, "loss": 0.2383, "step": 14513 }, { "epoch": 0.72, "learning_rate": 3.718072084377141e-06, "loss": 0.4691, "step": 14514 }, { "epoch": 0.72, "learning_rate": 3.716813883600746e-06, "loss": 0.4431, "step": 14515 }, { "epoch": 0.72, "learning_rate": 3.715555847152854e-06, "loss": 0.3687, "step": 14516 }, { "epoch": 0.72, "learning_rate": 3.7142979750663677e-06, "loss": 0.3774, "step": 14517 }, { "epoch": 0.72, "learning_rate": 3.713040267374182e-06, "loss": 0.4183, "step": 14518 }, { "epoch": 0.72, "learning_rate": 3.7117827241091976e-06, "loss": 0.317, "step": 14519 }, { "epoch": 0.73, "learning_rate": 3.710525345304299e-06, "loss": 0.2222, "step": 14520 }, { "epoch": 0.73, "learning_rate": 3.7092681309923715e-06, "loss": 0.4257, "step": 14521 }, { "epoch": 0.73, "learning_rate": 3.7080110812062984e-06, "loss": 0.391, "step": 14522 }, { "epoch": 0.73, "learning_rate": 3.7067541959789535e-06, "loss": 0.441, "step": 14523 }, { "epoch": 0.73, "learning_rate": 3.7054974753432103e-06, "loss": 0.2048, "step": 14524 }, { "epoch": 0.73, "learning_rate": 3.704240919331936e-06, "loss": 0.4094, "step": 14525 }, { "epoch": 0.73, "learning_rate": 3.702984527977995e-06, "loss": 0.2493, "step": 14526 }, { "epoch": 0.73, "learning_rate": 3.7017283013142426e-06, "loss": 0.534, "step": 14527 }, { "epoch": 0.73, "learning_rate": 3.700472239373539e-06, "loss": 0.4696, "step": 14528 }, { "epoch": 0.73, "learning_rate": 3.699216342188735e-06, "loss": 0.4133, "step": 14529 }, { "epoch": 0.73, "learning_rate": 3.6979606097926748e-06, "loss": 0.5202, "step": 14530 }, { "epoch": 0.73, "learning_rate": 3.6967050422182003e-06, "loss": 0.3183, "step": 14531 }, { "epoch": 0.73, "learning_rate": 3.6954496394981464e-06, "loss": 0.1924, "step": 14532 }, { "epoch": 0.73, "learning_rate": 3.694194401665353e-06, "loss": 0.2338, "step": 14533 }, { "epoch": 0.73, "learning_rate": 3.692939328752645e-06, "loss": 0.3085, "step": 14534 }, { "epoch": 0.73, "learning_rate": 3.691684420792849e-06, "loss": 0.4643, "step": 14535 }, { "epoch": 0.73, "learning_rate": 3.690429677818781e-06, "loss": 0.5042, "step": 14536 }, { "epoch": 0.73, "learning_rate": 3.689175099863265e-06, "loss": 0.363, "step": 14537 }, { "epoch": 0.73, "learning_rate": 3.6879206869591066e-06, "loss": 0.6566, "step": 14538 }, { "epoch": 0.73, "learning_rate": 3.6866664391391172e-06, "loss": 0.1223, "step": 14539 }, { "epoch": 0.73, "learning_rate": 3.6854123564360977e-06, "loss": 0.4954, "step": 14540 }, { "epoch": 0.73, "learning_rate": 3.6841584388828434e-06, "loss": 0.196, "step": 14541 }, { "epoch": 0.73, "learning_rate": 3.6829046865121565e-06, "loss": 0.5702, "step": 14542 }, { "epoch": 0.73, "learning_rate": 3.6816510993568232e-06, "loss": 0.4083, "step": 14543 }, { "epoch": 0.73, "learning_rate": 3.680397677449631e-06, "loss": 0.4021, "step": 14544 }, { "epoch": 0.73, "learning_rate": 3.6791444208233552e-06, "loss": 0.2949, "step": 14545 }, { "epoch": 0.73, "learning_rate": 3.677891329510783e-06, "loss": 0.6332, "step": 14546 }, { "epoch": 0.73, "learning_rate": 3.676638403544682e-06, "loss": 0.4318, "step": 14547 }, { "epoch": 0.73, "learning_rate": 3.6753856429578217e-06, "loss": 0.1888, "step": 14548 }, { "epoch": 0.73, "learning_rate": 3.674133047782966e-06, "loss": 0.4854, "step": 14549 }, { "epoch": 0.73, "learning_rate": 3.6728806180528754e-06, "loss": 0.3671, "step": 14550 }, { "epoch": 0.73, "learning_rate": 3.671628353800305e-06, "loss": 0.3596, "step": 14551 }, { "epoch": 0.73, "learning_rate": 3.670376255058007e-06, "loss": 0.3323, "step": 14552 }, { "epoch": 0.73, "learning_rate": 3.669124321858727e-06, "loss": 0.0813, "step": 14553 }, { "epoch": 0.73, "learning_rate": 3.6678725542352077e-06, "loss": 0.3888, "step": 14554 }, { "epoch": 0.73, "learning_rate": 3.6666209522201857e-06, "loss": 0.3533, "step": 14555 }, { "epoch": 0.73, "learning_rate": 3.665369515846401e-06, "loss": 0.2164, "step": 14556 }, { "epoch": 0.73, "learning_rate": 3.6641182451465797e-06, "loss": 0.4692, "step": 14557 }, { "epoch": 0.73, "learning_rate": 3.6628671401534467e-06, "loss": 0.2486, "step": 14558 }, { "epoch": 0.73, "learning_rate": 3.6616162008997202e-06, "loss": 0.3694, "step": 14559 }, { "epoch": 0.73, "learning_rate": 3.660365427418123e-06, "loss": 0.2744, "step": 14560 }, { "epoch": 0.73, "learning_rate": 3.6591148197413663e-06, "loss": 0.2176, "step": 14561 }, { "epoch": 0.73, "learning_rate": 3.657864377902155e-06, "loss": 0.543, "step": 14562 }, { "epoch": 0.73, "learning_rate": 3.6566141019331945e-06, "loss": 0.554, "step": 14563 }, { "epoch": 0.73, "learning_rate": 3.65536399186718e-06, "loss": 0.3866, "step": 14564 }, { "epoch": 0.73, "learning_rate": 3.654114047736814e-06, "loss": 0.4666, "step": 14565 }, { "epoch": 0.73, "learning_rate": 3.652864269574784e-06, "loss": 0.2158, "step": 14566 }, { "epoch": 0.73, "learning_rate": 3.6516146574137756e-06, "loss": 0.4258, "step": 14567 }, { "epoch": 0.73, "learning_rate": 3.6503652112864673e-06, "loss": 0.6851, "step": 14568 }, { "epoch": 0.73, "learning_rate": 3.6491159312255443e-06, "loss": 0.5407, "step": 14569 }, { "epoch": 0.73, "learning_rate": 3.6478668172636755e-06, "loss": 0.3098, "step": 14570 }, { "epoch": 0.73, "learning_rate": 3.6466178694335295e-06, "loss": 0.2961, "step": 14571 }, { "epoch": 0.73, "learning_rate": 3.6453690877677727e-06, "loss": 0.3521, "step": 14572 }, { "epoch": 0.73, "learning_rate": 3.644120472299064e-06, "loss": 0.2795, "step": 14573 }, { "epoch": 0.73, "learning_rate": 3.64287202306006e-06, "loss": 0.2922, "step": 14574 }, { "epoch": 0.73, "learning_rate": 3.641623740083412e-06, "loss": 0.4153, "step": 14575 }, { "epoch": 0.73, "learning_rate": 3.6403756234017673e-06, "loss": 0.3555, "step": 14576 }, { "epoch": 0.73, "learning_rate": 3.639127673047764e-06, "loss": 0.2021, "step": 14577 }, { "epoch": 0.73, "learning_rate": 3.637879889054049e-06, "loss": 0.298, "step": 14578 }, { "epoch": 0.73, "learning_rate": 3.636632271453253e-06, "loss": 0.3942, "step": 14579 }, { "epoch": 0.73, "learning_rate": 3.635384820278005e-06, "loss": 0.2757, "step": 14580 }, { "epoch": 0.73, "learning_rate": 3.6341375355609323e-06, "loss": 0.4974, "step": 14581 }, { "epoch": 0.73, "learning_rate": 3.632890417334649e-06, "loss": 0.4099, "step": 14582 }, { "epoch": 0.73, "learning_rate": 3.631643465631782e-06, "loss": 0.5592, "step": 14583 }, { "epoch": 0.73, "learning_rate": 3.630396680484939e-06, "loss": 0.2866, "step": 14584 }, { "epoch": 0.73, "learning_rate": 3.6291500619267285e-06, "loss": 0.4119, "step": 14585 }, { "epoch": 0.73, "learning_rate": 3.6279036099897503e-06, "loss": 0.3609, "step": 14586 }, { "epoch": 0.73, "learning_rate": 3.6266573247066094e-06, "loss": 0.497, "step": 14587 }, { "epoch": 0.73, "learning_rate": 3.6254112061099002e-06, "loss": 0.6154, "step": 14588 }, { "epoch": 0.73, "learning_rate": 3.62416525423221e-06, "loss": 0.4337, "step": 14589 }, { "epoch": 0.73, "learning_rate": 3.622919469106129e-06, "loss": 0.3885, "step": 14590 }, { "epoch": 0.73, "learning_rate": 3.6216738507642314e-06, "loss": 0.2684, "step": 14591 }, { "epoch": 0.73, "learning_rate": 3.620428399239104e-06, "loss": 0.3213, "step": 14592 }, { "epoch": 0.73, "learning_rate": 3.619183114563317e-06, "loss": 0.1918, "step": 14593 }, { "epoch": 0.73, "learning_rate": 3.6179379967694374e-06, "loss": 0.3314, "step": 14594 }, { "epoch": 0.73, "learning_rate": 3.6166930458900306e-06, "loss": 0.5697, "step": 14595 }, { "epoch": 0.73, "learning_rate": 3.615448261957657e-06, "loss": 0.3512, "step": 14596 }, { "epoch": 0.73, "learning_rate": 3.6142036450048713e-06, "loss": 0.4593, "step": 14597 }, { "epoch": 0.73, "learning_rate": 3.612959195064226e-06, "loss": 0.1759, "step": 14598 }, { "epoch": 0.73, "learning_rate": 3.6117149121682673e-06, "loss": 0.312, "step": 14599 }, { "epoch": 0.73, "learning_rate": 3.6104707963495344e-06, "loss": 0.4186, "step": 14600 }, { "epoch": 0.73, "learning_rate": 3.6092268476405722e-06, "loss": 0.3258, "step": 14601 }, { "epoch": 0.73, "learning_rate": 3.6079830660739115e-06, "loss": 0.4591, "step": 14602 }, { "epoch": 0.73, "learning_rate": 3.6067394516820807e-06, "loss": 0.1967, "step": 14603 }, { "epoch": 0.73, "learning_rate": 3.6054960044976072e-06, "loss": 0.4255, "step": 14604 }, { "epoch": 0.73, "learning_rate": 3.6042527245530066e-06, "loss": 0.3861, "step": 14605 }, { "epoch": 0.73, "learning_rate": 3.603009611880803e-06, "loss": 0.4074, "step": 14606 }, { "epoch": 0.73, "learning_rate": 3.6017666665135044e-06, "loss": 0.5445, "step": 14607 }, { "epoch": 0.73, "learning_rate": 3.6005238884836168e-06, "loss": 0.2386, "step": 14608 }, { "epoch": 0.73, "learning_rate": 3.599281277823643e-06, "loss": 0.5024, "step": 14609 }, { "epoch": 0.73, "learning_rate": 3.5980388345660875e-06, "loss": 0.2875, "step": 14610 }, { "epoch": 0.73, "learning_rate": 3.5967965587434417e-06, "loss": 0.4551, "step": 14611 }, { "epoch": 0.73, "learning_rate": 3.5955544503881944e-06, "loss": 0.3872, "step": 14612 }, { "epoch": 0.73, "learning_rate": 3.5943125095328324e-06, "loss": 0.3032, "step": 14613 }, { "epoch": 0.73, "learning_rate": 3.5930707362098326e-06, "loss": 0.2246, "step": 14614 }, { "epoch": 0.73, "learning_rate": 3.5918291304516804e-06, "loss": 0.2329, "step": 14615 }, { "epoch": 0.73, "learning_rate": 3.5905876922908434e-06, "loss": 0.2528, "step": 14616 }, { "epoch": 0.73, "learning_rate": 3.5893464217597905e-06, "loss": 0.2331, "step": 14617 }, { "epoch": 0.73, "learning_rate": 3.588105318890983e-06, "loss": 0.4148, "step": 14618 }, { "epoch": 0.73, "learning_rate": 3.586864383716886e-06, "loss": 0.4431, "step": 14619 }, { "epoch": 0.73, "learning_rate": 3.585623616269951e-06, "loss": 0.3472, "step": 14620 }, { "epoch": 0.73, "learning_rate": 3.584383016582629e-06, "loss": 0.5576, "step": 14621 }, { "epoch": 0.73, "learning_rate": 3.583142584687366e-06, "loss": 0.4086, "step": 14622 }, { "epoch": 0.73, "learning_rate": 3.5819023206166047e-06, "loss": 0.278, "step": 14623 }, { "epoch": 0.73, "learning_rate": 3.580662224402782e-06, "loss": 0.1974, "step": 14624 }, { "epoch": 0.73, "learning_rate": 3.579422296078331e-06, "loss": 0.3798, "step": 14625 }, { "epoch": 0.73, "learning_rate": 3.5781825356756796e-06, "loss": 0.4559, "step": 14626 }, { "epoch": 0.73, "learning_rate": 3.576942943227253e-06, "loss": 0.2001, "step": 14627 }, { "epoch": 0.73, "learning_rate": 3.5757035187654685e-06, "loss": 0.5134, "step": 14628 }, { "epoch": 0.73, "learning_rate": 3.5744642623227467e-06, "loss": 0.2161, "step": 14629 }, { "epoch": 0.73, "learning_rate": 3.573225173931497e-06, "loss": 0.368, "step": 14630 }, { "epoch": 0.73, "learning_rate": 3.5719862536241246e-06, "loss": 0.3361, "step": 14631 }, { "epoch": 0.73, "learning_rate": 3.5707475014330293e-06, "loss": 0.2121, "step": 14632 }, { "epoch": 0.73, "learning_rate": 3.569508917390616e-06, "loss": 0.3114, "step": 14633 }, { "epoch": 0.73, "learning_rate": 3.568270501529275e-06, "loss": 0.2972, "step": 14634 }, { "epoch": 0.73, "learning_rate": 3.567032253881394e-06, "loss": 0.3613, "step": 14635 }, { "epoch": 0.73, "learning_rate": 3.565794174479359e-06, "loss": 0.443, "step": 14636 }, { "epoch": 0.73, "learning_rate": 3.5645562633555476e-06, "loss": 0.2495, "step": 14637 }, { "epoch": 0.73, "learning_rate": 3.5633185205423405e-06, "loss": 0.2572, "step": 14638 }, { "epoch": 0.73, "learning_rate": 3.5620809460721074e-06, "loss": 0.1986, "step": 14639 }, { "epoch": 0.73, "learning_rate": 3.5608435399772146e-06, "loss": 0.3584, "step": 14640 }, { "epoch": 0.73, "learning_rate": 3.559606302290022e-06, "loss": 0.4806, "step": 14641 }, { "epoch": 0.73, "learning_rate": 3.5583692330428944e-06, "loss": 0.5129, "step": 14642 }, { "epoch": 0.73, "learning_rate": 3.5571323322681827e-06, "loss": 0.3948, "step": 14643 }, { "epoch": 0.73, "learning_rate": 3.555895599998236e-06, "loss": 0.2653, "step": 14644 }, { "epoch": 0.73, "learning_rate": 3.5546590362654e-06, "loss": 0.6204, "step": 14645 }, { "epoch": 0.73, "learning_rate": 3.5534226411020145e-06, "loss": 0.3098, "step": 14646 }, { "epoch": 0.73, "learning_rate": 3.5521864145404173e-06, "loss": 0.2283, "step": 14647 }, { "epoch": 0.73, "learning_rate": 3.5509503566129388e-06, "loss": 0.4043, "step": 14648 }, { "epoch": 0.73, "learning_rate": 3.5497144673519068e-06, "loss": 0.7299, "step": 14649 }, { "epoch": 0.73, "learning_rate": 3.5484787467896407e-06, "loss": 0.3501, "step": 14650 }, { "epoch": 0.73, "learning_rate": 3.547243194958467e-06, "loss": 0.2732, "step": 14651 }, { "epoch": 0.73, "learning_rate": 3.546007811890696e-06, "loss": 0.3673, "step": 14652 }, { "epoch": 0.73, "learning_rate": 3.5447725976186377e-06, "loss": 0.2499, "step": 14653 }, { "epoch": 0.73, "learning_rate": 3.5435375521745975e-06, "loss": 0.2077, "step": 14654 }, { "epoch": 0.73, "learning_rate": 3.542302675590872e-06, "loss": 0.2014, "step": 14655 }, { "epoch": 0.73, "learning_rate": 3.5410679678997652e-06, "loss": 0.2496, "step": 14656 }, { "epoch": 0.73, "learning_rate": 3.539833429133567e-06, "loss": 0.3758, "step": 14657 }, { "epoch": 0.73, "learning_rate": 3.5385990593245635e-06, "loss": 0.3078, "step": 14658 }, { "epoch": 0.73, "learning_rate": 3.537364858505038e-06, "loss": 0.4244, "step": 14659 }, { "epoch": 0.73, "learning_rate": 3.536130826707268e-06, "loss": 0.3193, "step": 14660 }, { "epoch": 0.73, "learning_rate": 3.534896963963532e-06, "loss": 0.3251, "step": 14661 }, { "epoch": 0.73, "learning_rate": 3.533663270306099e-06, "loss": 0.4675, "step": 14662 }, { "epoch": 0.73, "learning_rate": 3.532429745767234e-06, "loss": 0.4143, "step": 14663 }, { "epoch": 0.73, "learning_rate": 3.5311963903791946e-06, "loss": 0.3734, "step": 14664 }, { "epoch": 0.73, "learning_rate": 3.529963204174245e-06, "loss": 0.3079, "step": 14665 }, { "epoch": 0.73, "learning_rate": 3.5287301871846325e-06, "loss": 0.1181, "step": 14666 }, { "epoch": 0.73, "learning_rate": 3.527497339442606e-06, "loss": 0.219, "step": 14667 }, { "epoch": 0.73, "learning_rate": 3.5262646609804094e-06, "loss": 0.3386, "step": 14668 }, { "epoch": 0.73, "learning_rate": 3.525032151830282e-06, "loss": 0.5799, "step": 14669 }, { "epoch": 0.73, "learning_rate": 3.5237998120244574e-06, "loss": 0.4381, "step": 14670 }, { "epoch": 0.73, "learning_rate": 3.522567641595167e-06, "loss": 0.3328, "step": 14671 }, { "epoch": 0.73, "learning_rate": 3.521335640574636e-06, "loss": 0.2912, "step": 14672 }, { "epoch": 0.73, "learning_rate": 3.5201038089950823e-06, "loss": 0.5036, "step": 14673 }, { "epoch": 0.73, "learning_rate": 3.518872146888729e-06, "loss": 0.2039, "step": 14674 }, { "epoch": 0.73, "learning_rate": 3.5176406542877874e-06, "loss": 0.4707, "step": 14675 }, { "epoch": 0.73, "learning_rate": 3.516409331224463e-06, "loss": 0.3984, "step": 14676 }, { "epoch": 0.73, "learning_rate": 3.51517817773096e-06, "loss": 0.4605, "step": 14677 }, { "epoch": 0.73, "learning_rate": 3.5139471938394765e-06, "loss": 0.4328, "step": 14678 }, { "epoch": 0.73, "learning_rate": 3.5127163795822107e-06, "loss": 0.3717, "step": 14679 }, { "epoch": 0.73, "learning_rate": 3.5114857349913524e-06, "loss": 0.3897, "step": 14680 }, { "epoch": 0.73, "learning_rate": 3.510255260099086e-06, "loss": 0.3854, "step": 14681 }, { "epoch": 0.73, "learning_rate": 3.5090249549375887e-06, "loss": 0.34, "step": 14682 }, { "epoch": 0.73, "learning_rate": 3.5077948195390467e-06, "loss": 0.3951, "step": 14683 }, { "epoch": 0.73, "learning_rate": 3.506564853935628e-06, "loss": 0.3469, "step": 14684 }, { "epoch": 0.73, "learning_rate": 3.5053350581595e-06, "loss": 0.1966, "step": 14685 }, { "epoch": 0.73, "learning_rate": 3.504105432242827e-06, "loss": 0.5696, "step": 14686 }, { "epoch": 0.73, "learning_rate": 3.5028759762177654e-06, "loss": 0.4326, "step": 14687 }, { "epoch": 0.73, "learning_rate": 3.5016466901164757e-06, "loss": 0.2613, "step": 14688 }, { "epoch": 0.73, "learning_rate": 3.5004175739711064e-06, "loss": 0.3199, "step": 14689 }, { "epoch": 0.73, "learning_rate": 3.499188627813802e-06, "loss": 0.2455, "step": 14690 }, { "epoch": 0.73, "learning_rate": 3.4979598516767043e-06, "loss": 0.4741, "step": 14691 }, { "epoch": 0.73, "learning_rate": 3.4967312455919478e-06, "loss": 0.3384, "step": 14692 }, { "epoch": 0.73, "learning_rate": 3.495502809591671e-06, "loss": 0.383, "step": 14693 }, { "epoch": 0.73, "learning_rate": 3.4942745437079982e-06, "loss": 0.453, "step": 14694 }, { "epoch": 0.73, "learning_rate": 3.4930464479730552e-06, "loss": 0.4694, "step": 14695 }, { "epoch": 0.73, "learning_rate": 3.4918185224189584e-06, "loss": 0.3098, "step": 14696 }, { "epoch": 0.73, "learning_rate": 3.490590767077824e-06, "loss": 0.3326, "step": 14697 }, { "epoch": 0.73, "learning_rate": 3.489363181981763e-06, "loss": 0.4156, "step": 14698 }, { "epoch": 0.73, "learning_rate": 3.48813576716288e-06, "loss": 0.3788, "step": 14699 }, { "epoch": 0.73, "learning_rate": 3.4869085226532774e-06, "loss": 0.4738, "step": 14700 }, { "epoch": 0.73, "learning_rate": 3.4856814484850487e-06, "loss": 0.3315, "step": 14701 }, { "epoch": 0.73, "learning_rate": 3.484454544690292e-06, "loss": 0.4488, "step": 14702 }, { "epoch": 0.73, "learning_rate": 3.4832278113010933e-06, "loss": 0.1694, "step": 14703 }, { "epoch": 0.73, "learning_rate": 3.482001248349536e-06, "loss": 0.4609, "step": 14704 }, { "epoch": 0.73, "learning_rate": 3.4807748558676956e-06, "loss": 0.4186, "step": 14705 }, { "epoch": 0.73, "learning_rate": 3.479548633887653e-06, "loss": 0.4302, "step": 14706 }, { "epoch": 0.73, "learning_rate": 3.478322582441476e-06, "loss": 0.3294, "step": 14707 }, { "epoch": 0.73, "learning_rate": 3.47709670156123e-06, "loss": 0.2398, "step": 14708 }, { "epoch": 0.73, "learning_rate": 3.4758709912789766e-06, "loss": 0.3194, "step": 14709 }, { "epoch": 0.73, "learning_rate": 3.474645451626768e-06, "loss": 0.2882, "step": 14710 }, { "epoch": 0.73, "learning_rate": 3.473420082636665e-06, "loss": 0.4926, "step": 14711 }, { "epoch": 0.73, "learning_rate": 3.4721948843407114e-06, "loss": 0.2544, "step": 14712 }, { "epoch": 0.73, "learning_rate": 3.4709698567709506e-06, "loss": 0.2868, "step": 14713 }, { "epoch": 0.73, "learning_rate": 3.469744999959419e-06, "loss": 0.4033, "step": 14714 }, { "epoch": 0.73, "learning_rate": 3.468520313938156e-06, "loss": 0.4703, "step": 14715 }, { "epoch": 0.73, "learning_rate": 3.467295798739191e-06, "loss": 0.4367, "step": 14716 }, { "epoch": 0.73, "learning_rate": 3.4660714543945473e-06, "loss": 0.5272, "step": 14717 }, { "epoch": 0.73, "learning_rate": 3.4648472809362465e-06, "loss": 0.4421, "step": 14718 }, { "epoch": 0.73, "learning_rate": 3.4636232783963052e-06, "loss": 0.3011, "step": 14719 }, { "epoch": 0.74, "learning_rate": 3.4623994468067366e-06, "loss": 0.4456, "step": 14720 }, { "epoch": 0.74, "learning_rate": 3.4611757861995475e-06, "loss": 0.3042, "step": 14721 }, { "epoch": 0.74, "learning_rate": 3.459952296606742e-06, "loss": 0.3657, "step": 14722 }, { "epoch": 0.74, "learning_rate": 3.458728978060314e-06, "loss": 0.3269, "step": 14723 }, { "epoch": 0.74, "learning_rate": 3.457505830592266e-06, "loss": 0.4209, "step": 14724 }, { "epoch": 0.74, "learning_rate": 3.4562828542345838e-06, "loss": 0.2081, "step": 14725 }, { "epoch": 0.74, "learning_rate": 3.455060049019253e-06, "loss": 0.3624, "step": 14726 }, { "epoch": 0.74, "learning_rate": 3.453837414978254e-06, "loss": 0.2437, "step": 14727 }, { "epoch": 0.74, "learning_rate": 3.4526149521435605e-06, "loss": 0.3525, "step": 14728 }, { "epoch": 0.74, "learning_rate": 3.4513926605471504e-06, "loss": 0.4175, "step": 14729 }, { "epoch": 0.74, "learning_rate": 3.4501705402209874e-06, "loss": 0.3042, "step": 14730 }, { "epoch": 0.74, "learning_rate": 3.4489485911970367e-06, "loss": 0.2723, "step": 14731 }, { "epoch": 0.74, "learning_rate": 3.447726813507254e-06, "loss": 0.3151, "step": 14732 }, { "epoch": 0.74, "learning_rate": 3.4465052071835913e-06, "loss": 0.4635, "step": 14733 }, { "epoch": 0.74, "learning_rate": 3.4452837722580057e-06, "loss": 0.5533, "step": 14734 }, { "epoch": 0.74, "learning_rate": 3.444062508762437e-06, "loss": 0.3916, "step": 14735 }, { "epoch": 0.74, "learning_rate": 3.442841416728827e-06, "loss": 0.369, "step": 14736 }, { "epoch": 0.74, "learning_rate": 3.4416204961891086e-06, "loss": 0.3681, "step": 14737 }, { "epoch": 0.74, "learning_rate": 3.4403997471752182e-06, "loss": 0.1702, "step": 14738 }, { "epoch": 0.74, "learning_rate": 3.4391791697190824e-06, "loss": 0.3972, "step": 14739 }, { "epoch": 0.74, "learning_rate": 3.4379587638526212e-06, "loss": 0.4972, "step": 14740 }, { "epoch": 0.74, "learning_rate": 3.4367385296077537e-06, "loss": 0.3075, "step": 14741 }, { "epoch": 0.74, "learning_rate": 3.4355184670163945e-06, "loss": 0.4018, "step": 14742 }, { "epoch": 0.74, "learning_rate": 3.4342985761104508e-06, "loss": 0.1289, "step": 14743 }, { "epoch": 0.74, "learning_rate": 3.43307885692183e-06, "loss": 0.3957, "step": 14744 }, { "epoch": 0.74, "learning_rate": 3.431859309482426e-06, "loss": 0.2295, "step": 14745 }, { "epoch": 0.74, "learning_rate": 3.4306399338241425e-06, "loss": 0.3466, "step": 14746 }, { "epoch": 0.74, "learning_rate": 3.429420729978867e-06, "loss": 0.1788, "step": 14747 }, { "epoch": 0.74, "learning_rate": 3.428201697978487e-06, "loss": 0.2028, "step": 14748 }, { "epoch": 0.74, "learning_rate": 3.426982837854883e-06, "loss": 0.4664, "step": 14749 }, { "epoch": 0.74, "learning_rate": 3.4257641496399353e-06, "loss": 0.2676, "step": 14750 }, { "epoch": 0.74, "learning_rate": 3.4245456333655115e-06, "loss": 0.177, "step": 14751 }, { "epoch": 0.74, "learning_rate": 3.4233272890634873e-06, "loss": 0.5459, "step": 14752 }, { "epoch": 0.74, "learning_rate": 3.422109116765724e-06, "loss": 0.2301, "step": 14753 }, { "epoch": 0.74, "learning_rate": 3.4208911165040813e-06, "loss": 0.2619, "step": 14754 }, { "epoch": 0.74, "learning_rate": 3.4196732883104123e-06, "loss": 0.2574, "step": 14755 }, { "epoch": 0.74, "learning_rate": 3.418455632216572e-06, "loss": 0.5267, "step": 14756 }, { "epoch": 0.74, "learning_rate": 3.4172381482544038e-06, "loss": 0.4577, "step": 14757 }, { "epoch": 0.74, "learning_rate": 3.4160208364557513e-06, "loss": 0.5589, "step": 14758 }, { "epoch": 0.74, "learning_rate": 3.414803696852449e-06, "loss": 0.3232, "step": 14759 }, { "epoch": 0.74, "learning_rate": 3.413586729476328e-06, "loss": 0.2912, "step": 14760 }, { "epoch": 0.74, "learning_rate": 3.412369934359223e-06, "loss": 0.3881, "step": 14761 }, { "epoch": 0.74, "learning_rate": 3.411153311532954e-06, "loss": 0.421, "step": 14762 }, { "epoch": 0.74, "learning_rate": 3.4099368610293416e-06, "loss": 0.3933, "step": 14763 }, { "epoch": 0.74, "learning_rate": 3.4087205828801972e-06, "loss": 0.391, "step": 14764 }, { "epoch": 0.74, "learning_rate": 3.4075044771173315e-06, "loss": 0.3581, "step": 14765 }, { "epoch": 0.74, "learning_rate": 3.4062885437725533e-06, "loss": 0.3332, "step": 14766 }, { "epoch": 0.74, "learning_rate": 3.405072782877663e-06, "loss": 0.2405, "step": 14767 }, { "epoch": 0.74, "learning_rate": 3.403857194464456e-06, "loss": 0.3333, "step": 14768 }, { "epoch": 0.74, "learning_rate": 3.4026417785647236e-06, "loss": 0.586, "step": 14769 }, { "epoch": 0.74, "learning_rate": 3.401426535210256e-06, "loss": 0.4277, "step": 14770 }, { "epoch": 0.74, "learning_rate": 3.4002114644328344e-06, "loss": 0.5323, "step": 14771 }, { "epoch": 0.74, "learning_rate": 3.398996566264238e-06, "loss": 0.3166, "step": 14772 }, { "epoch": 0.74, "learning_rate": 3.3977818407362394e-06, "loss": 0.3782, "step": 14773 }, { "epoch": 0.74, "learning_rate": 3.396567287880608e-06, "loss": 0.3422, "step": 14774 }, { "epoch": 0.74, "learning_rate": 3.3953529077291112e-06, "loss": 0.3961, "step": 14775 }, { "epoch": 0.74, "learning_rate": 3.3941387003135106e-06, "loss": 0.1829, "step": 14776 }, { "epoch": 0.74, "learning_rate": 3.392924665665559e-06, "loss": 0.3363, "step": 14777 }, { "epoch": 0.74, "learning_rate": 3.3917108038170065e-06, "loss": 0.4563, "step": 14778 }, { "epoch": 0.74, "learning_rate": 3.3904971147996058e-06, "loss": 0.4682, "step": 14779 }, { "epoch": 0.74, "learning_rate": 3.3892835986450955e-06, "loss": 0.3356, "step": 14780 }, { "epoch": 0.74, "learning_rate": 3.3880702553852153e-06, "loss": 0.328, "step": 14781 }, { "epoch": 0.74, "learning_rate": 3.3868570850516967e-06, "loss": 0.5447, "step": 14782 }, { "epoch": 0.74, "learning_rate": 3.385644087676266e-06, "loss": 0.3063, "step": 14783 }, { "epoch": 0.74, "learning_rate": 3.3844312632906553e-06, "loss": 0.441, "step": 14784 }, { "epoch": 0.74, "learning_rate": 3.3832186119265786e-06, "loss": 0.3294, "step": 14785 }, { "epoch": 0.74, "learning_rate": 3.382006133615753e-06, "loss": 0.385, "step": 14786 }, { "epoch": 0.74, "learning_rate": 3.3807938283898867e-06, "loss": 0.5552, "step": 14787 }, { "epoch": 0.74, "learning_rate": 3.3795816962806905e-06, "loss": 0.4157, "step": 14788 }, { "epoch": 0.74, "learning_rate": 3.378369737319864e-06, "loss": 0.3539, "step": 14789 }, { "epoch": 0.74, "learning_rate": 3.377157951539104e-06, "loss": 0.3188, "step": 14790 }, { "epoch": 0.74, "learning_rate": 3.375946338970104e-06, "loss": 0.4429, "step": 14791 }, { "epoch": 0.74, "learning_rate": 3.3747348996445515e-06, "loss": 0.1589, "step": 14792 }, { "epoch": 0.74, "learning_rate": 3.37352363359413e-06, "loss": 0.4391, "step": 14793 }, { "epoch": 0.74, "learning_rate": 3.3723125408505185e-06, "loss": 0.4544, "step": 14794 }, { "epoch": 0.74, "learning_rate": 3.371101621445393e-06, "loss": 0.2637, "step": 14795 }, { "epoch": 0.74, "learning_rate": 3.369890875410421e-06, "loss": 0.3132, "step": 14796 }, { "epoch": 0.74, "learning_rate": 3.368680302777266e-06, "loss": 0.2894, "step": 14797 }, { "epoch": 0.74, "learning_rate": 3.367469903577596e-06, "loss": 0.4957, "step": 14798 }, { "epoch": 0.74, "learning_rate": 3.3662596778430633e-06, "loss": 0.3798, "step": 14799 }, { "epoch": 0.74, "learning_rate": 3.36504962560532e-06, "loss": 0.5009, "step": 14800 }, { "epoch": 0.74, "learning_rate": 3.3638397468960104e-06, "loss": 0.359, "step": 14801 }, { "epoch": 0.74, "learning_rate": 3.362630041746784e-06, "loss": 0.4587, "step": 14802 }, { "epoch": 0.74, "learning_rate": 3.361420510189276e-06, "loss": 0.1492, "step": 14803 }, { "epoch": 0.74, "learning_rate": 3.3602111522551185e-06, "loss": 0.4463, "step": 14804 }, { "epoch": 0.74, "learning_rate": 3.3590019679759425e-06, "loss": 0.3266, "step": 14805 }, { "epoch": 0.74, "learning_rate": 3.357792957383368e-06, "loss": 0.2746, "step": 14806 }, { "epoch": 0.74, "learning_rate": 3.356584120509022e-06, "loss": 0.2886, "step": 14807 }, { "epoch": 0.74, "learning_rate": 3.3553754573845177e-06, "loss": 0.4447, "step": 14808 }, { "epoch": 0.74, "learning_rate": 3.3541669680414644e-06, "loss": 0.3912, "step": 14809 }, { "epoch": 0.74, "learning_rate": 3.3529586525114676e-06, "loss": 0.4479, "step": 14810 }, { "epoch": 0.74, "learning_rate": 3.3517505108261338e-06, "loss": 0.2936, "step": 14811 }, { "epoch": 0.74, "learning_rate": 3.3505425430170578e-06, "loss": 0.5771, "step": 14812 }, { "epoch": 0.74, "learning_rate": 3.349334749115832e-06, "loss": 0.3098, "step": 14813 }, { "epoch": 0.74, "learning_rate": 3.3481271291540463e-06, "loss": 0.3896, "step": 14814 }, { "epoch": 0.74, "learning_rate": 3.346919683163282e-06, "loss": 0.4775, "step": 14815 }, { "epoch": 0.74, "learning_rate": 3.3457124111751192e-06, "loss": 0.2805, "step": 14816 }, { "epoch": 0.74, "learning_rate": 3.3445053132211303e-06, "loss": 0.4041, "step": 14817 }, { "epoch": 0.74, "learning_rate": 3.343298389332891e-06, "loss": 0.382, "step": 14818 }, { "epoch": 0.74, "learning_rate": 3.3420916395419634e-06, "loss": 0.0857, "step": 14819 }, { "epoch": 0.74, "learning_rate": 3.3408850638799086e-06, "loss": 0.4358, "step": 14820 }, { "epoch": 0.74, "learning_rate": 3.3396786623782827e-06, "loss": 0.105, "step": 14821 }, { "epoch": 0.74, "learning_rate": 3.3384724350686393e-06, "loss": 0.3258, "step": 14822 }, { "epoch": 0.74, "learning_rate": 3.3372663819825236e-06, "loss": 0.3084, "step": 14823 }, { "epoch": 0.74, "learning_rate": 3.3360605031514746e-06, "loss": 0.3747, "step": 14824 }, { "epoch": 0.74, "learning_rate": 3.334854798607039e-06, "loss": 0.1477, "step": 14825 }, { "epoch": 0.74, "learning_rate": 3.3336492683807474e-06, "loss": 0.44, "step": 14826 }, { "epoch": 0.74, "learning_rate": 3.332443912504126e-06, "loss": 0.2269, "step": 14827 }, { "epoch": 0.74, "learning_rate": 3.3312387310087024e-06, "loss": 0.4363, "step": 14828 }, { "epoch": 0.74, "learning_rate": 3.3300337239259916e-06, "loss": 0.3455, "step": 14829 }, { "epoch": 0.74, "learning_rate": 3.3288288912875165e-06, "loss": 0.3189, "step": 14830 }, { "epoch": 0.74, "learning_rate": 3.3276242331247843e-06, "loss": 0.4281, "step": 14831 }, { "epoch": 0.74, "learning_rate": 3.3264197494692994e-06, "loss": 0.5088, "step": 14832 }, { "epoch": 0.74, "learning_rate": 3.325215440352564e-06, "loss": 0.2063, "step": 14833 }, { "epoch": 0.74, "learning_rate": 3.3240113058060773e-06, "loss": 0.1175, "step": 14834 }, { "epoch": 0.74, "learning_rate": 3.322807345861333e-06, "loss": 0.2306, "step": 14835 }, { "epoch": 0.74, "learning_rate": 3.3216035605498166e-06, "loss": 0.3118, "step": 14836 }, { "epoch": 0.74, "learning_rate": 3.3203999499030105e-06, "loss": 0.3528, "step": 14837 }, { "epoch": 0.74, "learning_rate": 3.319196513952393e-06, "loss": 0.5375, "step": 14838 }, { "epoch": 0.74, "learning_rate": 3.3179932527294435e-06, "loss": 0.366, "step": 14839 }, { "epoch": 0.74, "learning_rate": 3.316790166265629e-06, "loss": 0.3947, "step": 14840 }, { "epoch": 0.74, "learning_rate": 3.315587254592413e-06, "loss": 0.2645, "step": 14841 }, { "epoch": 0.74, "learning_rate": 3.314384517741258e-06, "loss": 0.2695, "step": 14842 }, { "epoch": 0.74, "learning_rate": 3.3131819557436197e-06, "loss": 0.1234, "step": 14843 }, { "epoch": 0.74, "learning_rate": 3.3119795686309496e-06, "loss": 0.2861, "step": 14844 }, { "epoch": 0.74, "learning_rate": 3.3107773564346937e-06, "loss": 0.2808, "step": 14845 }, { "epoch": 0.74, "learning_rate": 3.3095753191862943e-06, "loss": 0.4979, "step": 14846 }, { "epoch": 0.74, "learning_rate": 3.3083734569171866e-06, "loss": 0.2707, "step": 14847 }, { "epoch": 0.74, "learning_rate": 3.307171769658809e-06, "loss": 0.3713, "step": 14848 }, { "epoch": 0.74, "learning_rate": 3.305970257442589e-06, "loss": 0.2812, "step": 14849 }, { "epoch": 0.74, "learning_rate": 3.3047689202999488e-06, "loss": 0.1971, "step": 14850 }, { "epoch": 0.74, "learning_rate": 3.303567758262306e-06, "loss": 0.1919, "step": 14851 }, { "epoch": 0.74, "learning_rate": 3.3023667713610798e-06, "loss": 0.3701, "step": 14852 }, { "epoch": 0.74, "learning_rate": 3.301165959627679e-06, "loss": 0.1898, "step": 14853 }, { "epoch": 0.74, "learning_rate": 3.299965323093508e-06, "loss": 0.5343, "step": 14854 }, { "epoch": 0.74, "learning_rate": 3.298764861789969e-06, "loss": 0.3618, "step": 14855 }, { "epoch": 0.74, "learning_rate": 3.2975645757484543e-06, "loss": 0.3992, "step": 14856 }, { "epoch": 0.74, "learning_rate": 3.2963644650003634e-06, "loss": 0.2965, "step": 14857 }, { "epoch": 0.74, "learning_rate": 3.2951645295770795e-06, "loss": 0.3264, "step": 14858 }, { "epoch": 0.74, "learning_rate": 3.293964769509985e-06, "loss": 0.257, "step": 14859 }, { "epoch": 0.74, "learning_rate": 3.2927651848304565e-06, "loss": 0.3034, "step": 14860 }, { "epoch": 0.74, "learning_rate": 3.2915657755698715e-06, "loss": 0.3679, "step": 14861 }, { "epoch": 0.74, "learning_rate": 3.2903665417595988e-06, "loss": 0.5284, "step": 14862 }, { "epoch": 0.74, "learning_rate": 3.2891674834310005e-06, "loss": 0.3336, "step": 14863 }, { "epoch": 0.74, "learning_rate": 3.2879686006154365e-06, "loss": 0.2892, "step": 14864 }, { "epoch": 0.74, "learning_rate": 3.2867698933442625e-06, "loss": 0.356, "step": 14865 }, { "epoch": 0.74, "learning_rate": 3.2855713616488303e-06, "loss": 0.4261, "step": 14866 }, { "epoch": 0.74, "learning_rate": 3.284373005560484e-06, "loss": 0.3736, "step": 14867 }, { "epoch": 0.74, "learning_rate": 3.283174825110567e-06, "loss": 0.5639, "step": 14868 }, { "epoch": 0.74, "learning_rate": 3.281976820330414e-06, "loss": 0.379, "step": 14869 }, { "epoch": 0.74, "learning_rate": 3.280778991251354e-06, "loss": 0.4013, "step": 14870 }, { "epoch": 0.74, "learning_rate": 3.279581337904724e-06, "loss": 0.508, "step": 14871 }, { "epoch": 0.74, "learning_rate": 3.27838386032184e-06, "loss": 0.3822, "step": 14872 }, { "epoch": 0.74, "learning_rate": 3.2771865585340237e-06, "loss": 0.182, "step": 14873 }, { "epoch": 0.74, "learning_rate": 3.275989432572584e-06, "loss": 0.2289, "step": 14874 }, { "epoch": 0.74, "learning_rate": 3.274792482468837e-06, "loss": 0.5009, "step": 14875 }, { "epoch": 0.74, "learning_rate": 3.273595708254085e-06, "loss": 0.3386, "step": 14876 }, { "epoch": 0.74, "learning_rate": 3.2723991099596263e-06, "loss": 0.4395, "step": 14877 }, { "epoch": 0.74, "learning_rate": 3.271202687616758e-06, "loss": 0.4842, "step": 14878 }, { "epoch": 0.74, "learning_rate": 3.2700064412567666e-06, "loss": 0.2664, "step": 14879 }, { "epoch": 0.74, "learning_rate": 3.2688103709109455e-06, "loss": 0.3732, "step": 14880 }, { "epoch": 0.74, "learning_rate": 3.2676144766105734e-06, "loss": 0.1637, "step": 14881 }, { "epoch": 0.74, "learning_rate": 3.2664187583869258e-06, "loss": 0.278, "step": 14882 }, { "epoch": 0.74, "learning_rate": 3.2652232162712725e-06, "loss": 0.2736, "step": 14883 }, { "epoch": 0.74, "learning_rate": 3.2640278502948895e-06, "loss": 0.1813, "step": 14884 }, { "epoch": 0.74, "learning_rate": 3.2628326604890347e-06, "loss": 0.2809, "step": 14885 }, { "epoch": 0.74, "learning_rate": 3.261637646884969e-06, "loss": 0.1696, "step": 14886 }, { "epoch": 0.74, "learning_rate": 3.260442809513943e-06, "loss": 0.241, "step": 14887 }, { "epoch": 0.74, "learning_rate": 3.259248148407209e-06, "loss": 0.4054, "step": 14888 }, { "epoch": 0.74, "learning_rate": 3.258053663596008e-06, "loss": 0.2385, "step": 14889 }, { "epoch": 0.74, "learning_rate": 3.2568593551115857e-06, "loss": 0.269, "step": 14890 }, { "epoch": 0.74, "learning_rate": 3.255665222985176e-06, "loss": 0.3179, "step": 14891 }, { "epoch": 0.74, "learning_rate": 3.2544712672480074e-06, "loss": 0.0857, "step": 14892 }, { "epoch": 0.74, "learning_rate": 3.253277487931309e-06, "loss": 0.4376, "step": 14893 }, { "epoch": 0.74, "learning_rate": 3.2520838850663007e-06, "loss": 0.5747, "step": 14894 }, { "epoch": 0.74, "learning_rate": 3.2508904586842006e-06, "loss": 0.4216, "step": 14895 }, { "epoch": 0.74, "learning_rate": 3.2496972088162205e-06, "loss": 0.4744, "step": 14896 }, { "epoch": 0.74, "learning_rate": 3.2485041354935654e-06, "loss": 0.4706, "step": 14897 }, { "epoch": 0.74, "learning_rate": 3.247311238747445e-06, "loss": 0.6005, "step": 14898 }, { "epoch": 0.74, "learning_rate": 3.246118518609055e-06, "loss": 0.3212, "step": 14899 }, { "epoch": 0.74, "learning_rate": 3.2449259751095886e-06, "loss": 0.3562, "step": 14900 }, { "epoch": 0.74, "learning_rate": 3.243733608280236e-06, "loss": 0.3839, "step": 14901 }, { "epoch": 0.74, "learning_rate": 3.2425414181521788e-06, "loss": 0.4099, "step": 14902 }, { "epoch": 0.74, "learning_rate": 3.241349404756603e-06, "loss": 0.1903, "step": 14903 }, { "epoch": 0.74, "learning_rate": 3.2401575681246813e-06, "loss": 0.3512, "step": 14904 }, { "epoch": 0.74, "learning_rate": 3.2389659082875844e-06, "loss": 0.4377, "step": 14905 }, { "epoch": 0.74, "learning_rate": 3.2377744252764754e-06, "loss": 0.416, "step": 14906 }, { "epoch": 0.74, "learning_rate": 3.236583119122524e-06, "loss": 0.4053, "step": 14907 }, { "epoch": 0.74, "learning_rate": 3.235391989856881e-06, "loss": 0.6508, "step": 14908 }, { "epoch": 0.74, "learning_rate": 3.2342010375107013e-06, "loss": 0.4759, "step": 14909 }, { "epoch": 0.74, "learning_rate": 3.2330102621151316e-06, "loss": 0.3035, "step": 14910 }, { "epoch": 0.74, "learning_rate": 3.2318196637013123e-06, "loss": 0.5178, "step": 14911 }, { "epoch": 0.74, "learning_rate": 3.2306292423003884e-06, "loss": 0.3058, "step": 14912 }, { "epoch": 0.74, "learning_rate": 3.2294389979434904e-06, "loss": 0.1202, "step": 14913 }, { "epoch": 0.74, "learning_rate": 3.228248930661747e-06, "loss": 0.4805, "step": 14914 }, { "epoch": 0.74, "learning_rate": 3.2270590404862834e-06, "loss": 0.2451, "step": 14915 }, { "epoch": 0.74, "learning_rate": 3.22586932744822e-06, "loss": 0.3757, "step": 14916 }, { "epoch": 0.74, "learning_rate": 3.224679791578672e-06, "loss": 0.4588, "step": 14917 }, { "epoch": 0.74, "learning_rate": 3.223490432908749e-06, "loss": 0.5558, "step": 14918 }, { "epoch": 0.74, "learning_rate": 3.2223012514695595e-06, "loss": 0.4344, "step": 14919 }, { "epoch": 0.74, "learning_rate": 3.2211122472921997e-06, "loss": 0.5257, "step": 14920 }, { "epoch": 0.75, "learning_rate": 3.2199234204077732e-06, "loss": 0.3125, "step": 14921 }, { "epoch": 0.75, "learning_rate": 3.2187347708473704e-06, "loss": 0.2863, "step": 14922 }, { "epoch": 0.75, "learning_rate": 3.217546298642077e-06, "loss": 0.2203, "step": 14923 }, { "epoch": 0.75, "learning_rate": 3.2163580038229734e-06, "loss": 0.5493, "step": 14924 }, { "epoch": 0.75, "learning_rate": 3.2151698864211443e-06, "loss": 0.3956, "step": 14925 }, { "epoch": 0.75, "learning_rate": 3.2139819464676604e-06, "loss": 0.3334, "step": 14926 }, { "epoch": 0.75, "learning_rate": 3.212794183993592e-06, "loss": 0.3715, "step": 14927 }, { "epoch": 0.75, "learning_rate": 3.2116065990300004e-06, "loss": 0.2485, "step": 14928 }, { "epoch": 0.75, "learning_rate": 3.210419191607944e-06, "loss": 0.2518, "step": 14929 }, { "epoch": 0.75, "learning_rate": 3.2092319617584845e-06, "loss": 0.1536, "step": 14930 }, { "epoch": 0.75, "learning_rate": 3.2080449095126686e-06, "loss": 0.3463, "step": 14931 }, { "epoch": 0.75, "learning_rate": 3.2068580349015422e-06, "loss": 0.392, "step": 14932 }, { "epoch": 0.75, "learning_rate": 3.2056713379561466e-06, "loss": 0.4368, "step": 14933 }, { "epoch": 0.75, "learning_rate": 3.204484818707515e-06, "loss": 0.4587, "step": 14934 }, { "epoch": 0.75, "learning_rate": 3.203298477186686e-06, "loss": 0.3483, "step": 14935 }, { "epoch": 0.75, "learning_rate": 3.2021123134246823e-06, "loss": 0.3272, "step": 14936 }, { "epoch": 0.75, "learning_rate": 3.2009263274525282e-06, "loss": 0.663, "step": 14937 }, { "epoch": 0.75, "learning_rate": 3.199740519301241e-06, "loss": 0.3406, "step": 14938 }, { "epoch": 0.75, "learning_rate": 3.1985548890018325e-06, "loss": 0.4375, "step": 14939 }, { "epoch": 0.75, "learning_rate": 3.1973694365853135e-06, "loss": 0.1951, "step": 14940 }, { "epoch": 0.75, "learning_rate": 3.1961841620826873e-06, "loss": 0.5509, "step": 14941 }, { "epoch": 0.75, "learning_rate": 3.194999065524953e-06, "loss": 0.3506, "step": 14942 }, { "epoch": 0.75, "learning_rate": 3.1938141469431006e-06, "loss": 0.367, "step": 14943 }, { "epoch": 0.75, "learning_rate": 3.1926294063681297e-06, "loss": 0.348, "step": 14944 }, { "epoch": 0.75, "learning_rate": 3.1914448438310196e-06, "loss": 0.4035, "step": 14945 }, { "epoch": 0.75, "learning_rate": 3.1902604593627516e-06, "loss": 0.3231, "step": 14946 }, { "epoch": 0.75, "learning_rate": 3.1890762529942997e-06, "loss": 0.472, "step": 14947 }, { "epoch": 0.75, "learning_rate": 3.187892224756641e-06, "loss": 0.3985, "step": 14948 }, { "epoch": 0.75, "learning_rate": 3.1867083746807392e-06, "loss": 0.3568, "step": 14949 }, { "epoch": 0.75, "learning_rate": 3.1855247027975555e-06, "loss": 0.1932, "step": 14950 }, { "epoch": 0.75, "learning_rate": 3.1843412091380478e-06, "loss": 0.4087, "step": 14951 }, { "epoch": 0.75, "learning_rate": 3.1831578937331654e-06, "loss": 0.3345, "step": 14952 }, { "epoch": 0.75, "learning_rate": 3.1819747566138635e-06, "loss": 0.6246, "step": 14953 }, { "epoch": 0.75, "learning_rate": 3.1807917978110814e-06, "loss": 0.2245, "step": 14954 }, { "epoch": 0.75, "learning_rate": 3.179609017355758e-06, "loss": 0.5268, "step": 14955 }, { "epoch": 0.75, "learning_rate": 3.1784264152788256e-06, "loss": 0.4126, "step": 14956 }, { "epoch": 0.75, "learning_rate": 3.1772439916112184e-06, "loss": 0.4049, "step": 14957 }, { "epoch": 0.75, "learning_rate": 3.1760617463838583e-06, "loss": 0.2961, "step": 14958 }, { "epoch": 0.75, "learning_rate": 3.174879679627665e-06, "loss": 0.3866, "step": 14959 }, { "epoch": 0.75, "learning_rate": 3.1736977913735546e-06, "loss": 0.3281, "step": 14960 }, { "epoch": 0.75, "learning_rate": 3.1725160816524347e-06, "loss": 0.4982, "step": 14961 }, { "epoch": 0.75, "learning_rate": 3.1713345504952176e-06, "loss": 0.1612, "step": 14962 }, { "epoch": 0.75, "learning_rate": 3.1701531979328015e-06, "loss": 0.1845, "step": 14963 }, { "epoch": 0.75, "learning_rate": 3.168972023996082e-06, "loss": 0.3151, "step": 14964 }, { "epoch": 0.75, "learning_rate": 3.167791028715953e-06, "loss": 0.37, "step": 14965 }, { "epoch": 0.75, "learning_rate": 3.166610212123301e-06, "loss": 0.1428, "step": 14966 }, { "epoch": 0.75, "learning_rate": 3.1654295742490083e-06, "loss": 0.2206, "step": 14967 }, { "epoch": 0.75, "learning_rate": 3.164249115123954e-06, "loss": 0.2862, "step": 14968 }, { "epoch": 0.75, "learning_rate": 3.1630688347790108e-06, "loss": 0.4832, "step": 14969 }, { "epoch": 0.75, "learning_rate": 3.1618887332450444e-06, "loss": 0.2613, "step": 14970 }, { "epoch": 0.75, "learning_rate": 3.160708810552925e-06, "loss": 0.4222, "step": 14971 }, { "epoch": 0.75, "learning_rate": 3.159529066733508e-06, "loss": 0.4125, "step": 14972 }, { "epoch": 0.75, "learning_rate": 3.158349501817649e-06, "loss": 0.3148, "step": 14973 }, { "epoch": 0.75, "learning_rate": 3.157170115836199e-06, "loss": 0.4981, "step": 14974 }, { "epoch": 0.75, "learning_rate": 3.155990908819998e-06, "loss": 0.437, "step": 14975 }, { "epoch": 0.75, "learning_rate": 3.154811880799894e-06, "loss": 0.4844, "step": 14976 }, { "epoch": 0.75, "learning_rate": 3.15363303180672e-06, "loss": 0.3267, "step": 14977 }, { "epoch": 0.75, "learning_rate": 3.1524543618713066e-06, "loss": 0.5339, "step": 14978 }, { "epoch": 0.75, "learning_rate": 3.151275871024477e-06, "loss": 0.3344, "step": 14979 }, { "epoch": 0.75, "learning_rate": 3.1500975592970607e-06, "loss": 0.4489, "step": 14980 }, { "epoch": 0.75, "learning_rate": 3.148919426719871e-06, "loss": 0.2975, "step": 14981 }, { "epoch": 0.75, "learning_rate": 3.147741473323721e-06, "loss": 0.458, "step": 14982 }, { "epoch": 0.75, "learning_rate": 3.1465636991394168e-06, "loss": 0.3958, "step": 14983 }, { "epoch": 0.75, "learning_rate": 3.1453861041977597e-06, "loss": 0.4745, "step": 14984 }, { "epoch": 0.75, "learning_rate": 3.144208688529554e-06, "loss": 0.1598, "step": 14985 }, { "epoch": 0.75, "learning_rate": 3.1430314521655915e-06, "loss": 0.3061, "step": 14986 }, { "epoch": 0.75, "learning_rate": 3.1418543951366597e-06, "loss": 0.3875, "step": 14987 }, { "epoch": 0.75, "learning_rate": 3.140677517473544e-06, "loss": 0.238, "step": 14988 }, { "epoch": 0.75, "learning_rate": 3.139500819207023e-06, "loss": 0.302, "step": 14989 }, { "epoch": 0.75, "learning_rate": 3.1383243003678733e-06, "loss": 0.6901, "step": 14990 }, { "epoch": 0.75, "learning_rate": 3.1371479609868636e-06, "loss": 0.3168, "step": 14991 }, { "epoch": 0.75, "learning_rate": 3.1359718010947616e-06, "loss": 0.4316, "step": 14992 }, { "epoch": 0.75, "learning_rate": 3.1347958207223227e-06, "loss": 0.5909, "step": 14993 }, { "epoch": 0.75, "learning_rate": 3.1336200199003107e-06, "loss": 0.3664, "step": 14994 }, { "epoch": 0.75, "learning_rate": 3.1324443986594734e-06, "loss": 0.4746, "step": 14995 }, { "epoch": 0.75, "learning_rate": 3.1312689570305575e-06, "loss": 0.0598, "step": 14996 }, { "epoch": 0.75, "learning_rate": 3.130093695044304e-06, "loss": 0.3694, "step": 14997 }, { "epoch": 0.75, "learning_rate": 3.128918612731454e-06, "loss": 0.3181, "step": 14998 }, { "epoch": 0.75, "learning_rate": 3.127743710122738e-06, "loss": 0.0898, "step": 14999 }, { "epoch": 0.75, "learning_rate": 3.1265689872488846e-06, "loss": 0.4744, "step": 15000 }, { "epoch": 0.75, "learning_rate": 3.125394444140617e-06, "loss": 0.4846, "step": 15001 }, { "epoch": 0.75, "learning_rate": 3.1242200808286503e-06, "loss": 0.327, "step": 15002 }, { "epoch": 0.75, "learning_rate": 3.123045897343705e-06, "loss": 0.4148, "step": 15003 }, { "epoch": 0.75, "learning_rate": 3.121871893716487e-06, "loss": 0.4718, "step": 15004 }, { "epoch": 0.75, "learning_rate": 3.1206980699777e-06, "loss": 0.2319, "step": 15005 }, { "epoch": 0.75, "learning_rate": 3.119524426158046e-06, "loss": 0.4523, "step": 15006 }, { "epoch": 0.75, "learning_rate": 3.1183509622882147e-06, "loss": 0.2851, "step": 15007 }, { "epoch": 0.75, "learning_rate": 3.1171776783989048e-06, "loss": 0.2988, "step": 15008 }, { "epoch": 0.75, "learning_rate": 3.1160045745207967e-06, "loss": 0.3784, "step": 15009 }, { "epoch": 0.75, "learning_rate": 3.1148316506845734e-06, "loss": 0.6538, "step": 15010 }, { "epoch": 0.75, "learning_rate": 3.11365890692091e-06, "loss": 0.3226, "step": 15011 }, { "epoch": 0.75, "learning_rate": 3.112486343260479e-06, "loss": 0.5104, "step": 15012 }, { "epoch": 0.75, "learning_rate": 3.111313959733946e-06, "loss": 0.3928, "step": 15013 }, { "epoch": 0.75, "learning_rate": 3.1101417563719736e-06, "loss": 0.2595, "step": 15014 }, { "epoch": 0.75, "learning_rate": 3.1089697332052204e-06, "loss": 0.1818, "step": 15015 }, { "epoch": 0.75, "learning_rate": 3.1077978902643347e-06, "loss": 0.3069, "step": 15016 }, { "epoch": 0.75, "learning_rate": 3.1066262275799706e-06, "loss": 0.2174, "step": 15017 }, { "epoch": 0.75, "learning_rate": 3.1054547451827688e-06, "loss": 0.2584, "step": 15018 }, { "epoch": 0.75, "learning_rate": 3.1042834431033688e-06, "loss": 0.3063, "step": 15019 }, { "epoch": 0.75, "learning_rate": 3.1031123213724003e-06, "loss": 0.4304, "step": 15020 }, { "epoch": 0.75, "learning_rate": 3.101941380020498e-06, "loss": 0.47, "step": 15021 }, { "epoch": 0.75, "learning_rate": 3.1007706190782848e-06, "loss": 0.4021, "step": 15022 }, { "epoch": 0.75, "learning_rate": 3.099600038576379e-06, "loss": 0.1888, "step": 15023 }, { "epoch": 0.75, "learning_rate": 3.0984296385453973e-06, "loss": 0.4277, "step": 15024 }, { "epoch": 0.75, "learning_rate": 3.0972594190159444e-06, "loss": 0.5127, "step": 15025 }, { "epoch": 0.75, "learning_rate": 3.0960893800186355e-06, "loss": 0.3969, "step": 15026 }, { "epoch": 0.75, "learning_rate": 3.0949195215840646e-06, "loss": 0.3127, "step": 15027 }, { "epoch": 0.75, "learning_rate": 3.0937498437428303e-06, "loss": 0.3577, "step": 15028 }, { "epoch": 0.75, "learning_rate": 3.09258034652552e-06, "loss": 0.4186, "step": 15029 }, { "epoch": 0.75, "learning_rate": 3.0914110299627277e-06, "loss": 0.4079, "step": 15030 }, { "epoch": 0.75, "learning_rate": 3.09024189408503e-06, "loss": 0.5044, "step": 15031 }, { "epoch": 0.75, "learning_rate": 3.0890729389230055e-06, "loss": 0.3323, "step": 15032 }, { "epoch": 0.75, "learning_rate": 3.087904164507227e-06, "loss": 0.3707, "step": 15033 }, { "epoch": 0.75, "learning_rate": 3.086735570868259e-06, "loss": 0.396, "step": 15034 }, { "epoch": 0.75, "learning_rate": 3.085567158036671e-06, "loss": 0.4374, "step": 15035 }, { "epoch": 0.75, "learning_rate": 3.0843989260430163e-06, "loss": 0.323, "step": 15036 }, { "epoch": 0.75, "learning_rate": 3.0832308749178508e-06, "loss": 0.3011, "step": 15037 }, { "epoch": 0.75, "learning_rate": 3.082063004691722e-06, "loss": 0.2811, "step": 15038 }, { "epoch": 0.75, "learning_rate": 3.080895315395175e-06, "loss": 0.5266, "step": 15039 }, { "epoch": 0.75, "learning_rate": 3.0797278070587478e-06, "loss": 0.4428, "step": 15040 }, { "epoch": 0.75, "learning_rate": 3.0785604797129764e-06, "loss": 0.5857, "step": 15041 }, { "epoch": 0.75, "learning_rate": 3.07739333338839e-06, "loss": 0.3067, "step": 15042 }, { "epoch": 0.75, "learning_rate": 3.0762263681155114e-06, "loss": 0.3278, "step": 15043 }, { "epoch": 0.75, "learning_rate": 3.0750595839248664e-06, "loss": 0.2937, "step": 15044 }, { "epoch": 0.75, "learning_rate": 3.0738929808469676e-06, "loss": 0.2043, "step": 15045 }, { "epoch": 0.75, "learning_rate": 3.0727265589123267e-06, "loss": 0.4314, "step": 15046 }, { "epoch": 0.75, "learning_rate": 3.0715603181514497e-06, "loss": 0.3046, "step": 15047 }, { "epoch": 0.75, "learning_rate": 3.0703942585948344e-06, "loss": 0.2148, "step": 15048 }, { "epoch": 0.75, "learning_rate": 3.0692283802729837e-06, "loss": 0.4168, "step": 15049 }, { "epoch": 0.75, "learning_rate": 3.0680626832163883e-06, "loss": 0.3352, "step": 15050 }, { "epoch": 0.75, "learning_rate": 3.0668971674555337e-06, "loss": 0.3448, "step": 15051 }, { "epoch": 0.75, "learning_rate": 3.0657318330208996e-06, "loss": 0.3078, "step": 15052 }, { "epoch": 0.75, "learning_rate": 3.0645666799429706e-06, "loss": 0.2349, "step": 15053 }, { "epoch": 0.75, "learning_rate": 3.063401708252216e-06, "loss": 0.2117, "step": 15054 }, { "epoch": 0.75, "learning_rate": 3.0622369179791045e-06, "loss": 0.2205, "step": 15055 }, { "epoch": 0.75, "learning_rate": 3.0610723091540995e-06, "loss": 0.222, "step": 15056 }, { "epoch": 0.75, "learning_rate": 3.0599078818076567e-06, "loss": 0.3242, "step": 15057 }, { "epoch": 0.75, "learning_rate": 3.0587436359702373e-06, "loss": 0.3924, "step": 15058 }, { "epoch": 0.75, "learning_rate": 3.0575795716722857e-06, "loss": 0.4121, "step": 15059 }, { "epoch": 0.75, "learning_rate": 3.0564156889442476e-06, "loss": 0.4682, "step": 15060 }, { "epoch": 0.75, "learning_rate": 3.0552519878165633e-06, "loss": 0.2845, "step": 15061 }, { "epoch": 0.75, "learning_rate": 3.054088468319667e-06, "loss": 0.4286, "step": 15062 }, { "epoch": 0.75, "learning_rate": 3.052925130483989e-06, "loss": 0.3047, "step": 15063 }, { "epoch": 0.75, "learning_rate": 3.051761974339955e-06, "loss": 0.2229, "step": 15064 }, { "epoch": 0.75, "learning_rate": 3.0505989999179863e-06, "loss": 0.3692, "step": 15065 }, { "epoch": 0.75, "learning_rate": 3.0494362072484963e-06, "loss": 0.3548, "step": 15066 }, { "epoch": 0.75, "learning_rate": 3.0482735963619016e-06, "loss": 0.1938, "step": 15067 }, { "epoch": 0.75, "learning_rate": 3.0471111672886044e-06, "loss": 0.1901, "step": 15068 }, { "epoch": 0.75, "learning_rate": 3.0459489200590097e-06, "loss": 0.6047, "step": 15069 }, { "epoch": 0.75, "learning_rate": 3.044786854703512e-06, "loss": 0.3469, "step": 15070 }, { "epoch": 0.75, "learning_rate": 3.0436249712525014e-06, "loss": 0.3124, "step": 15071 }, { "epoch": 0.75, "learning_rate": 3.0424632697363708e-06, "loss": 0.4987, "step": 15072 }, { "epoch": 0.75, "learning_rate": 3.0413017501855025e-06, "loss": 0.4863, "step": 15073 }, { "epoch": 0.75, "learning_rate": 3.0401404126302713e-06, "loss": 0.4135, "step": 15074 }, { "epoch": 0.75, "learning_rate": 3.038979257101049e-06, "loss": 0.3738, "step": 15075 }, { "epoch": 0.75, "learning_rate": 3.03781828362821e-06, "loss": 0.4559, "step": 15076 }, { "epoch": 0.75, "learning_rate": 3.0366574922421155e-06, "loss": 0.6302, "step": 15077 }, { "epoch": 0.75, "learning_rate": 3.0354968829731234e-06, "loss": 0.2332, "step": 15078 }, { "epoch": 0.75, "learning_rate": 3.034336455851589e-06, "loss": 0.2661, "step": 15079 }, { "epoch": 0.75, "learning_rate": 3.0331762109078566e-06, "loss": 0.3655, "step": 15080 }, { "epoch": 0.75, "learning_rate": 3.03201614817228e-06, "loss": 0.3795, "step": 15081 }, { "epoch": 0.75, "learning_rate": 3.030856267675194e-06, "loss": 0.3685, "step": 15082 }, { "epoch": 0.75, "learning_rate": 3.029696569446935e-06, "loss": 0.3907, "step": 15083 }, { "epoch": 0.75, "learning_rate": 3.0285370535178327e-06, "loss": 0.3978, "step": 15084 }, { "epoch": 0.75, "learning_rate": 3.027377719918212e-06, "loss": 0.365, "step": 15085 }, { "epoch": 0.75, "learning_rate": 3.026218568678395e-06, "loss": 0.3739, "step": 15086 }, { "epoch": 0.75, "learning_rate": 3.0250595998286982e-06, "loss": 0.1947, "step": 15087 }, { "epoch": 0.75, "learning_rate": 3.023900813399431e-06, "loss": 0.3353, "step": 15088 }, { "epoch": 0.75, "learning_rate": 3.022742209420898e-06, "loss": 0.1402, "step": 15089 }, { "epoch": 0.75, "learning_rate": 3.0215837879234067e-06, "loss": 0.4558, "step": 15090 }, { "epoch": 0.75, "learning_rate": 3.0204255489372524e-06, "loss": 0.441, "step": 15091 }, { "epoch": 0.75, "learning_rate": 3.0192674924927257e-06, "loss": 0.2725, "step": 15092 }, { "epoch": 0.75, "learning_rate": 3.018109618620111e-06, "loss": 0.2579, "step": 15093 }, { "epoch": 0.75, "learning_rate": 3.0169519273496983e-06, "loss": 0.2403, "step": 15094 }, { "epoch": 0.75, "learning_rate": 3.0157944187117616e-06, "loss": 0.4415, "step": 15095 }, { "epoch": 0.75, "learning_rate": 3.014637092736574e-06, "loss": 0.2839, "step": 15096 }, { "epoch": 0.75, "learning_rate": 3.0134799494544053e-06, "loss": 0.416, "step": 15097 }, { "epoch": 0.75, "learning_rate": 3.012322988895513e-06, "loss": 0.2149, "step": 15098 }, { "epoch": 0.75, "learning_rate": 3.011166211090165e-06, "loss": 0.7054, "step": 15099 }, { "epoch": 0.75, "learning_rate": 3.010009616068611e-06, "loss": 0.1286, "step": 15100 }, { "epoch": 0.75, "learning_rate": 3.0088532038610995e-06, "loss": 0.3068, "step": 15101 }, { "epoch": 0.75, "learning_rate": 3.0076969744978756e-06, "loss": 0.3004, "step": 15102 }, { "epoch": 0.75, "learning_rate": 3.0065409280091773e-06, "loss": 0.5146, "step": 15103 }, { "epoch": 0.75, "learning_rate": 3.0053850644252437e-06, "loss": 0.0878, "step": 15104 }, { "epoch": 0.75, "learning_rate": 3.0042293837763014e-06, "loss": 0.5124, "step": 15105 }, { "epoch": 0.75, "learning_rate": 3.0030738860925778e-06, "loss": 0.2461, "step": 15106 }, { "epoch": 0.75, "learning_rate": 3.0019185714042885e-06, "loss": 0.0972, "step": 15107 }, { "epoch": 0.75, "learning_rate": 3.0007634397416576e-06, "loss": 0.4848, "step": 15108 }, { "epoch": 0.75, "learning_rate": 2.99960849113489e-06, "loss": 0.3634, "step": 15109 }, { "epoch": 0.75, "learning_rate": 2.9984537256141945e-06, "loss": 0.4716, "step": 15110 }, { "epoch": 0.75, "learning_rate": 2.9972991432097707e-06, "loss": 0.3597, "step": 15111 }, { "epoch": 0.75, "learning_rate": 2.9961447439518166e-06, "loss": 0.2758, "step": 15112 }, { "epoch": 0.75, "learning_rate": 2.994990527870524e-06, "loss": 0.4428, "step": 15113 }, { "epoch": 0.75, "learning_rate": 2.9938364949960774e-06, "loss": 0.4728, "step": 15114 }, { "epoch": 0.75, "learning_rate": 2.9926826453586623e-06, "loss": 0.2322, "step": 15115 }, { "epoch": 0.75, "learning_rate": 2.9915289789884505e-06, "loss": 0.3016, "step": 15116 }, { "epoch": 0.75, "learning_rate": 2.9903754959156226e-06, "loss": 0.2688, "step": 15117 }, { "epoch": 0.75, "learning_rate": 2.9892221961703427e-06, "loss": 0.3337, "step": 15118 }, { "epoch": 0.75, "learning_rate": 2.9880690797827736e-06, "loss": 0.2555, "step": 15119 }, { "epoch": 0.75, "learning_rate": 2.986916146783073e-06, "loss": 0.5493, "step": 15120 }, { "epoch": 0.76, "learning_rate": 2.9857633972013935e-06, "loss": 0.2875, "step": 15121 }, { "epoch": 0.76, "learning_rate": 2.9846108310678878e-06, "loss": 0.355, "step": 15122 }, { "epoch": 0.76, "learning_rate": 2.9834584484126973e-06, "loss": 0.4923, "step": 15123 }, { "epoch": 0.76, "learning_rate": 2.9823062492659606e-06, "loss": 0.4532, "step": 15124 }, { "epoch": 0.76, "learning_rate": 2.981154233657809e-06, "loss": 0.3377, "step": 15125 }, { "epoch": 0.76, "learning_rate": 2.9800024016183793e-06, "loss": 0.1956, "step": 15126 }, { "epoch": 0.76, "learning_rate": 2.978850753177792e-06, "loss": 0.8095, "step": 15127 }, { "epoch": 0.76, "learning_rate": 2.977699288366167e-06, "loss": 0.3767, "step": 15128 }, { "epoch": 0.76, "learning_rate": 2.97654800721362e-06, "loss": 0.429, "step": 15129 }, { "epoch": 0.76, "learning_rate": 2.9753969097502578e-06, "loss": 0.5052, "step": 15130 }, { "epoch": 0.76, "learning_rate": 2.974245996006192e-06, "loss": 0.4871, "step": 15131 }, { "epoch": 0.76, "learning_rate": 2.9730952660115196e-06, "loss": 0.585, "step": 15132 }, { "epoch": 0.76, "learning_rate": 2.9719447197963367e-06, "loss": 0.299, "step": 15133 }, { "epoch": 0.76, "learning_rate": 2.9707943573907348e-06, "loss": 0.4089, "step": 15134 }, { "epoch": 0.76, "learning_rate": 2.9696441788248e-06, "loss": 0.4922, "step": 15135 }, { "epoch": 0.76, "learning_rate": 2.968494184128614e-06, "loss": 0.2198, "step": 15136 }, { "epoch": 0.76, "learning_rate": 2.9673443733322528e-06, "loss": 0.3974, "step": 15137 }, { "epoch": 0.76, "learning_rate": 2.966194746465788e-06, "loss": 0.493, "step": 15138 }, { "epoch": 0.76, "learning_rate": 2.965045303559284e-06, "loss": 0.6321, "step": 15139 }, { "epoch": 0.76, "learning_rate": 2.963896044642809e-06, "loss": 0.454, "step": 15140 }, { "epoch": 0.76, "learning_rate": 2.9627469697464173e-06, "loss": 0.3875, "step": 15141 }, { "epoch": 0.76, "learning_rate": 2.961598078900162e-06, "loss": 0.3931, "step": 15142 }, { "epoch": 0.76, "learning_rate": 2.96044937213409e-06, "loss": 0.5161, "step": 15143 }, { "epoch": 0.76, "learning_rate": 2.9593008494782405e-06, "loss": 0.5756, "step": 15144 }, { "epoch": 0.76, "learning_rate": 2.9581525109626595e-06, "loss": 0.2751, "step": 15145 }, { "epoch": 0.76, "learning_rate": 2.957004356617377e-06, "loss": 0.1637, "step": 15146 }, { "epoch": 0.76, "learning_rate": 2.95585638647242e-06, "loss": 0.2301, "step": 15147 }, { "epoch": 0.76, "learning_rate": 2.9547086005578107e-06, "loss": 0.2096, "step": 15148 }, { "epoch": 0.76, "learning_rate": 2.953560998903573e-06, "loss": 0.4604, "step": 15149 }, { "epoch": 0.76, "learning_rate": 2.9524135815397183e-06, "loss": 0.2237, "step": 15150 }, { "epoch": 0.76, "learning_rate": 2.9512663484962558e-06, "loss": 0.4769, "step": 15151 }, { "epoch": 0.76, "learning_rate": 2.950119299803189e-06, "loss": 0.284, "step": 15152 }, { "epoch": 0.76, "learning_rate": 2.9489724354905156e-06, "loss": 0.3635, "step": 15153 }, { "epoch": 0.76, "learning_rate": 2.9478257555882363e-06, "loss": 0.5872, "step": 15154 }, { "epoch": 0.76, "learning_rate": 2.9466792601263373e-06, "loss": 0.3394, "step": 15155 }, { "epoch": 0.76, "learning_rate": 2.945532949134804e-06, "loss": 0.4289, "step": 15156 }, { "epoch": 0.76, "learning_rate": 2.9443868226436167e-06, "loss": 0.3865, "step": 15157 }, { "epoch": 0.76, "learning_rate": 2.943240880682752e-06, "loss": 0.4159, "step": 15158 }, { "epoch": 0.76, "learning_rate": 2.942095123282178e-06, "loss": 0.3603, "step": 15159 }, { "epoch": 0.76, "learning_rate": 2.940949550471862e-06, "loss": 0.2984, "step": 15160 }, { "epoch": 0.76, "learning_rate": 2.9398041622817654e-06, "loss": 0.4236, "step": 15161 }, { "epoch": 0.76, "learning_rate": 2.9386589587418403e-06, "loss": 0.3932, "step": 15162 }, { "epoch": 0.76, "learning_rate": 2.937513939882045e-06, "loss": 0.4066, "step": 15163 }, { "epoch": 0.76, "learning_rate": 2.936369105732322e-06, "loss": 0.1534, "step": 15164 }, { "epoch": 0.76, "learning_rate": 2.935224456322613e-06, "loss": 0.3253, "step": 15165 }, { "epoch": 0.76, "learning_rate": 2.934079991682852e-06, "loss": 0.4064, "step": 15166 }, { "epoch": 0.76, "learning_rate": 2.932935711842978e-06, "loss": 0.375, "step": 15167 }, { "epoch": 0.76, "learning_rate": 2.9317916168329143e-06, "loss": 0.3293, "step": 15168 }, { "epoch": 0.76, "learning_rate": 2.9306477066825823e-06, "loss": 0.248, "step": 15169 }, { "epoch": 0.76, "learning_rate": 2.9295039814219016e-06, "loss": 0.4263, "step": 15170 }, { "epoch": 0.76, "learning_rate": 2.9283604410807796e-06, "loss": 0.2431, "step": 15171 }, { "epoch": 0.76, "learning_rate": 2.9272170856891313e-06, "loss": 0.4771, "step": 15172 }, { "epoch": 0.76, "learning_rate": 2.926073915276856e-06, "loss": 0.4327, "step": 15173 }, { "epoch": 0.76, "learning_rate": 2.9249309298738526e-06, "loss": 0.38, "step": 15174 }, { "epoch": 0.76, "learning_rate": 2.9237881295100145e-06, "loss": 0.4966, "step": 15175 }, { "epoch": 0.76, "learning_rate": 2.9226455142152267e-06, "loss": 0.4304, "step": 15176 }, { "epoch": 0.76, "learning_rate": 2.921503084019378e-06, "loss": 0.4635, "step": 15177 }, { "epoch": 0.76, "learning_rate": 2.920360838952345e-06, "loss": 0.4204, "step": 15178 }, { "epoch": 0.76, "learning_rate": 2.919218779044003e-06, "loss": 0.5666, "step": 15179 }, { "epoch": 0.76, "learning_rate": 2.918076904324215e-06, "loss": 0.3062, "step": 15180 }, { "epoch": 0.76, "learning_rate": 2.9169352148228536e-06, "loss": 0.2246, "step": 15181 }, { "epoch": 0.76, "learning_rate": 2.9157937105697753e-06, "loss": 0.2553, "step": 15182 }, { "epoch": 0.76, "learning_rate": 2.9146523915948323e-06, "loss": 0.3511, "step": 15183 }, { "epoch": 0.76, "learning_rate": 2.913511257927877e-06, "loss": 0.1936, "step": 15184 }, { "epoch": 0.76, "learning_rate": 2.9123703095987532e-06, "loss": 0.1567, "step": 15185 }, { "epoch": 0.76, "learning_rate": 2.9112295466372995e-06, "loss": 0.1798, "step": 15186 }, { "epoch": 0.76, "learning_rate": 2.9100889690733535e-06, "loss": 0.3065, "step": 15187 }, { "epoch": 0.76, "learning_rate": 2.908948576936744e-06, "loss": 0.2173, "step": 15188 }, { "epoch": 0.76, "learning_rate": 2.9078083702572947e-06, "loss": 0.4317, "step": 15189 }, { "epoch": 0.76, "learning_rate": 2.9066683490648305e-06, "loss": 0.5366, "step": 15190 }, { "epoch": 0.76, "learning_rate": 2.9055285133891654e-06, "loss": 0.4656, "step": 15191 }, { "epoch": 0.76, "learning_rate": 2.904388863260109e-06, "loss": 0.3057, "step": 15192 }, { "epoch": 0.76, "learning_rate": 2.9032493987074693e-06, "loss": 0.3746, "step": 15193 }, { "epoch": 0.76, "learning_rate": 2.9021101197610435e-06, "loss": 0.2431, "step": 15194 }, { "epoch": 0.76, "learning_rate": 2.9009710264506342e-06, "loss": 0.3696, "step": 15195 }, { "epoch": 0.76, "learning_rate": 2.8998321188060285e-06, "loss": 0.3109, "step": 15196 }, { "epoch": 0.76, "learning_rate": 2.8986933968570153e-06, "loss": 0.3643, "step": 15197 }, { "epoch": 0.76, "learning_rate": 2.8975548606333716e-06, "loss": 0.2784, "step": 15198 }, { "epoch": 0.76, "learning_rate": 2.896416510164881e-06, "loss": 0.0805, "step": 15199 }, { "epoch": 0.76, "learning_rate": 2.8952783454813127e-06, "loss": 0.3867, "step": 15200 }, { "epoch": 0.76, "learning_rate": 2.8941403666124345e-06, "loss": 0.3731, "step": 15201 }, { "epoch": 0.76, "learning_rate": 2.8930025735880073e-06, "loss": 0.264, "step": 15202 }, { "epoch": 0.76, "learning_rate": 2.8918649664377864e-06, "loss": 0.4817, "step": 15203 }, { "epoch": 0.76, "learning_rate": 2.8907275451915307e-06, "loss": 0.2763, "step": 15204 }, { "epoch": 0.76, "learning_rate": 2.8895903098789836e-06, "loss": 0.4186, "step": 15205 }, { "epoch": 0.76, "learning_rate": 2.8884532605298897e-06, "loss": 0.1881, "step": 15206 }, { "epoch": 0.76, "learning_rate": 2.887316397173986e-06, "loss": 0.2232, "step": 15207 }, { "epoch": 0.76, "learning_rate": 2.8861797198410057e-06, "loss": 0.4073, "step": 15208 }, { "epoch": 0.76, "learning_rate": 2.885043228560679e-06, "loss": 0.3912, "step": 15209 }, { "epoch": 0.76, "learning_rate": 2.883906923362726e-06, "loss": 0.6077, "step": 15210 }, { "epoch": 0.76, "learning_rate": 2.8827708042768677e-06, "loss": 0.3637, "step": 15211 }, { "epoch": 0.76, "learning_rate": 2.8816348713328137e-06, "loss": 0.3908, "step": 15212 }, { "epoch": 0.76, "learning_rate": 2.8804991245602797e-06, "loss": 0.4756, "step": 15213 }, { "epoch": 0.76, "learning_rate": 2.879363563988966e-06, "loss": 0.4063, "step": 15214 }, { "epoch": 0.76, "learning_rate": 2.878228189648572e-06, "loss": 0.3801, "step": 15215 }, { "epoch": 0.76, "learning_rate": 2.8770930015687914e-06, "loss": 0.3441, "step": 15216 }, { "epoch": 0.76, "learning_rate": 2.8759579997793117e-06, "loss": 0.2659, "step": 15217 }, { "epoch": 0.76, "learning_rate": 2.874823184309823e-06, "loss": 0.2721, "step": 15218 }, { "epoch": 0.76, "learning_rate": 2.8736885551900007e-06, "loss": 0.6588, "step": 15219 }, { "epoch": 0.76, "learning_rate": 2.8725541124495215e-06, "loss": 0.5308, "step": 15220 }, { "epoch": 0.76, "learning_rate": 2.8714198561180495e-06, "loss": 0.426, "step": 15221 }, { "epoch": 0.76, "learning_rate": 2.870285786225259e-06, "loss": 0.4262, "step": 15222 }, { "epoch": 0.76, "learning_rate": 2.869151902800805e-06, "loss": 0.3474, "step": 15223 }, { "epoch": 0.76, "learning_rate": 2.868018205874343e-06, "loss": 0.4306, "step": 15224 }, { "epoch": 0.76, "learning_rate": 2.8668846954755235e-06, "loss": 0.6132, "step": 15225 }, { "epoch": 0.76, "learning_rate": 2.8657513716339902e-06, "loss": 0.2835, "step": 15226 }, { "epoch": 0.76, "learning_rate": 2.8646182343793873e-06, "loss": 0.294, "step": 15227 }, { "epoch": 0.76, "learning_rate": 2.8634852837413498e-06, "loss": 0.4396, "step": 15228 }, { "epoch": 0.76, "learning_rate": 2.8623525197495074e-06, "loss": 0.5817, "step": 15229 }, { "epoch": 0.76, "learning_rate": 2.861219942433485e-06, "loss": 0.2689, "step": 15230 }, { "epoch": 0.76, "learning_rate": 2.8600875518229054e-06, "loss": 0.4313, "step": 15231 }, { "epoch": 0.76, "learning_rate": 2.858955347947385e-06, "loss": 0.6609, "step": 15232 }, { "epoch": 0.76, "learning_rate": 2.8578233308365345e-06, "loss": 0.4884, "step": 15233 }, { "epoch": 0.76, "learning_rate": 2.856691500519959e-06, "loss": 0.3206, "step": 15234 }, { "epoch": 0.76, "learning_rate": 2.855559857027259e-06, "loss": 0.3147, "step": 15235 }, { "epoch": 0.76, "learning_rate": 2.8544284003880362e-06, "loss": 0.2863, "step": 15236 }, { "epoch": 0.76, "learning_rate": 2.85329713063188e-06, "loss": 0.2129, "step": 15237 }, { "epoch": 0.76, "learning_rate": 2.8521660477883774e-06, "loss": 0.3064, "step": 15238 }, { "epoch": 0.76, "learning_rate": 2.851035151887106e-06, "loss": 0.4483, "step": 15239 }, { "epoch": 0.76, "learning_rate": 2.8499044429576494e-06, "loss": 0.1471, "step": 15240 }, { "epoch": 0.76, "learning_rate": 2.8487739210295774e-06, "loss": 0.1762, "step": 15241 }, { "epoch": 0.76, "learning_rate": 2.847643586132457e-06, "loss": 0.2041, "step": 15242 }, { "epoch": 0.76, "learning_rate": 2.8465134382958513e-06, "loss": 0.3672, "step": 15243 }, { "epoch": 0.76, "learning_rate": 2.845383477549314e-06, "loss": 0.5324, "step": 15244 }, { "epoch": 0.76, "learning_rate": 2.844253703922404e-06, "loss": 0.2049, "step": 15245 }, { "epoch": 0.76, "learning_rate": 2.8431241174446657e-06, "loss": 0.3514, "step": 15246 }, { "epoch": 0.76, "learning_rate": 2.841994718145642e-06, "loss": 0.3076, "step": 15247 }, { "epoch": 0.76, "learning_rate": 2.840865506054872e-06, "loss": 0.4116, "step": 15248 }, { "epoch": 0.76, "learning_rate": 2.8397364812018846e-06, "loss": 0.329, "step": 15249 }, { "epoch": 0.76, "learning_rate": 2.8386076436162136e-06, "loss": 0.5528, "step": 15250 }, { "epoch": 0.76, "learning_rate": 2.8374789933273806e-06, "loss": 0.4366, "step": 15251 }, { "epoch": 0.76, "learning_rate": 2.836350530364903e-06, "loss": 0.2296, "step": 15252 }, { "epoch": 0.76, "learning_rate": 2.835222254758292e-06, "loss": 0.6401, "step": 15253 }, { "epoch": 0.76, "learning_rate": 2.8340941665370616e-06, "loss": 0.3509, "step": 15254 }, { "epoch": 0.76, "learning_rate": 2.832966265730712e-06, "loss": 0.2888, "step": 15255 }, { "epoch": 0.76, "learning_rate": 2.831838552368744e-06, "loss": 0.4783, "step": 15256 }, { "epoch": 0.76, "learning_rate": 2.830711026480649e-06, "loss": 0.3211, "step": 15257 }, { "epoch": 0.76, "learning_rate": 2.8295836880959173e-06, "loss": 0.4451, "step": 15258 }, { "epoch": 0.76, "learning_rate": 2.828456537244033e-06, "loss": 0.3976, "step": 15259 }, { "epoch": 0.76, "learning_rate": 2.8273295739544736e-06, "loss": 0.2311, "step": 15260 }, { "epoch": 0.76, "learning_rate": 2.8262027982567165e-06, "loss": 0.3284, "step": 15261 }, { "epoch": 0.76, "learning_rate": 2.8250762101802255e-06, "loss": 0.4525, "step": 15262 }, { "epoch": 0.76, "learning_rate": 2.8239498097544716e-06, "loss": 0.377, "step": 15263 }, { "epoch": 0.76, "learning_rate": 2.8228235970089115e-06, "loss": 0.6374, "step": 15264 }, { "epoch": 0.76, "learning_rate": 2.821697571973e-06, "loss": 0.5594, "step": 15265 }, { "epoch": 0.76, "learning_rate": 2.8205717346761863e-06, "loss": 0.1918, "step": 15266 }, { "epoch": 0.76, "learning_rate": 2.819446085147912e-06, "loss": 0.4429, "step": 15267 }, { "epoch": 0.76, "learning_rate": 2.8183206234176242e-06, "loss": 0.1355, "step": 15268 }, { "epoch": 0.76, "learning_rate": 2.817195349514754e-06, "loss": 0.6822, "step": 15269 }, { "epoch": 0.76, "learning_rate": 2.816070263468732e-06, "loss": 0.287, "step": 15270 }, { "epoch": 0.76, "learning_rate": 2.81494536530898e-06, "loss": 0.4137, "step": 15271 }, { "epoch": 0.76, "learning_rate": 2.813820655064924e-06, "loss": 0.4559, "step": 15272 }, { "epoch": 0.76, "learning_rate": 2.8126961327659763e-06, "loss": 0.3434, "step": 15273 }, { "epoch": 0.76, "learning_rate": 2.811571798441548e-06, "loss": 0.2754, "step": 15274 }, { "epoch": 0.76, "learning_rate": 2.8104476521210443e-06, "loss": 0.3807, "step": 15275 }, { "epoch": 0.76, "learning_rate": 2.809323693833862e-06, "loss": 0.3523, "step": 15276 }, { "epoch": 0.76, "learning_rate": 2.8081999236094037e-06, "loss": 0.2784, "step": 15277 }, { "epoch": 0.76, "learning_rate": 2.8070763414770572e-06, "loss": 0.4639, "step": 15278 }, { "epoch": 0.76, "learning_rate": 2.805952947466208e-06, "loss": 0.317, "step": 15279 }, { "epoch": 0.76, "learning_rate": 2.8048297416062363e-06, "loss": 0.3541, "step": 15280 }, { "epoch": 0.76, "learning_rate": 2.8037067239265192e-06, "loss": 0.2024, "step": 15281 }, { "epoch": 0.76, "learning_rate": 2.802583894456428e-06, "loss": 0.4436, "step": 15282 }, { "epoch": 0.76, "learning_rate": 2.801461253225326e-06, "loss": 0.4631, "step": 15283 }, { "epoch": 0.76, "learning_rate": 2.8003388002625788e-06, "loss": 0.3554, "step": 15284 }, { "epoch": 0.76, "learning_rate": 2.7992165355975353e-06, "loss": 0.1969, "step": 15285 }, { "epoch": 0.76, "learning_rate": 2.798094459259556e-06, "loss": 0.4146, "step": 15286 }, { "epoch": 0.76, "learning_rate": 2.7969725712779837e-06, "loss": 0.5215, "step": 15287 }, { "epoch": 0.76, "learning_rate": 2.7958508716821596e-06, "loss": 0.4327, "step": 15288 }, { "epoch": 0.76, "learning_rate": 2.7947293605014193e-06, "loss": 0.3614, "step": 15289 }, { "epoch": 0.76, "learning_rate": 2.793608037765093e-06, "loss": 0.4349, "step": 15290 }, { "epoch": 0.76, "learning_rate": 2.7924869035025124e-06, "loss": 0.397, "step": 15291 }, { "epoch": 0.76, "learning_rate": 2.7913659577429975e-06, "loss": 0.2822, "step": 15292 }, { "epoch": 0.76, "learning_rate": 2.7902452005158643e-06, "loss": 0.3443, "step": 15293 }, { "epoch": 0.76, "learning_rate": 2.7891246318504204e-06, "loss": 0.3829, "step": 15294 }, { "epoch": 0.76, "learning_rate": 2.788004251775982e-06, "loss": 0.3118, "step": 15295 }, { "epoch": 0.76, "learning_rate": 2.786884060321845e-06, "loss": 0.5398, "step": 15296 }, { "epoch": 0.76, "learning_rate": 2.7857640575173085e-06, "loss": 0.1592, "step": 15297 }, { "epoch": 0.76, "learning_rate": 2.7846442433916643e-06, "loss": 0.2094, "step": 15298 }, { "epoch": 0.76, "learning_rate": 2.783524617974196e-06, "loss": 0.4171, "step": 15299 }, { "epoch": 0.76, "learning_rate": 2.782405181294193e-06, "loss": 0.4288, "step": 15300 }, { "epoch": 0.76, "learning_rate": 2.781285933380928e-06, "loss": 0.3727, "step": 15301 }, { "epoch": 0.76, "learning_rate": 2.7801668742636767e-06, "loss": 0.2397, "step": 15302 }, { "epoch": 0.76, "learning_rate": 2.779048003971703e-06, "loss": 0.3293, "step": 15303 }, { "epoch": 0.76, "learning_rate": 2.7779293225342707e-06, "loss": 0.4775, "step": 15304 }, { "epoch": 0.76, "learning_rate": 2.7768108299806385e-06, "loss": 0.2475, "step": 15305 }, { "epoch": 0.76, "learning_rate": 2.775692526340058e-06, "loss": 0.4022, "step": 15306 }, { "epoch": 0.76, "learning_rate": 2.7745744116417783e-06, "loss": 0.4086, "step": 15307 }, { "epoch": 0.76, "learning_rate": 2.7734564859150372e-06, "loss": 0.4616, "step": 15308 }, { "epoch": 0.76, "learning_rate": 2.7723387491890797e-06, "loss": 0.3039, "step": 15309 }, { "epoch": 0.76, "learning_rate": 2.771221201493136e-06, "loss": 0.2989, "step": 15310 }, { "epoch": 0.76, "learning_rate": 2.770103842856433e-06, "loss": 0.4538, "step": 15311 }, { "epoch": 0.76, "learning_rate": 2.7689866733081962e-06, "loss": 0.4845, "step": 15312 }, { "epoch": 0.76, "learning_rate": 2.767869692877637e-06, "loss": 0.271, "step": 15313 }, { "epoch": 0.76, "learning_rate": 2.7667529015939777e-06, "loss": 0.3227, "step": 15314 }, { "epoch": 0.76, "learning_rate": 2.765636299486423e-06, "loss": 0.2276, "step": 15315 }, { "epoch": 0.76, "learning_rate": 2.764519886584175e-06, "loss": 0.3967, "step": 15316 }, { "epoch": 0.76, "learning_rate": 2.7634036629164295e-06, "loss": 0.2037, "step": 15317 }, { "epoch": 0.76, "learning_rate": 2.7622876285123857e-06, "loss": 0.3926, "step": 15318 }, { "epoch": 0.76, "learning_rate": 2.7611717834012295e-06, "loss": 0.2682, "step": 15319 }, { "epoch": 0.76, "learning_rate": 2.760056127612144e-06, "loss": 0.4297, "step": 15320 }, { "epoch": 0.77, "learning_rate": 2.7589406611743088e-06, "loss": 0.2432, "step": 15321 }, { "epoch": 0.77, "learning_rate": 2.757825384116892e-06, "loss": 0.3674, "step": 15322 }, { "epoch": 0.77, "learning_rate": 2.7567102964690706e-06, "loss": 0.3077, "step": 15323 }, { "epoch": 0.77, "learning_rate": 2.7555953982600035e-06, "loss": 0.1668, "step": 15324 }, { "epoch": 0.77, "learning_rate": 2.7544806895188515e-06, "loss": 0.2963, "step": 15325 }, { "epoch": 0.77, "learning_rate": 2.753366170274764e-06, "loss": 0.3873, "step": 15326 }, { "epoch": 0.77, "learning_rate": 2.7522518405568955e-06, "loss": 0.4239, "step": 15327 }, { "epoch": 0.77, "learning_rate": 2.751137700394387e-06, "loss": 0.5117, "step": 15328 }, { "epoch": 0.77, "learning_rate": 2.7500237498163784e-06, "loss": 0.4779, "step": 15329 }, { "epoch": 0.77, "learning_rate": 2.748909988852002e-06, "loss": 0.3484, "step": 15330 }, { "epoch": 0.77, "learning_rate": 2.7477964175303885e-06, "loss": 0.214, "step": 15331 }, { "epoch": 0.77, "learning_rate": 2.7466830358806604e-06, "loss": 0.3867, "step": 15332 }, { "epoch": 0.77, "learning_rate": 2.745569843931938e-06, "loss": 0.4522, "step": 15333 }, { "epoch": 0.77, "learning_rate": 2.7444568417133345e-06, "loss": 0.3194, "step": 15334 }, { "epoch": 0.77, "learning_rate": 2.7433440292539557e-06, "loss": 0.2453, "step": 15335 }, { "epoch": 0.77, "learning_rate": 2.7422314065829124e-06, "loss": 0.346, "step": 15336 }, { "epoch": 0.77, "learning_rate": 2.741118973729302e-06, "loss": 0.4496, "step": 15337 }, { "epoch": 0.77, "learning_rate": 2.7400067307222155e-06, "loss": 0.2614, "step": 15338 }, { "epoch": 0.77, "learning_rate": 2.7388946775907456e-06, "loss": 0.4244, "step": 15339 }, { "epoch": 0.77, "learning_rate": 2.7377828143639717e-06, "loss": 0.6261, "step": 15340 }, { "epoch": 0.77, "learning_rate": 2.736671141070979e-06, "loss": 0.3685, "step": 15341 }, { "epoch": 0.77, "learning_rate": 2.7355596577408394e-06, "loss": 0.5151, "step": 15342 }, { "epoch": 0.77, "learning_rate": 2.734448364402622e-06, "loss": 0.2428, "step": 15343 }, { "epoch": 0.77, "learning_rate": 2.7333372610853925e-06, "loss": 0.3309, "step": 15344 }, { "epoch": 0.77, "learning_rate": 2.7322263478182053e-06, "loss": 0.389, "step": 15345 }, { "epoch": 0.77, "learning_rate": 2.731115624630123e-06, "loss": 0.5277, "step": 15346 }, { "epoch": 0.77, "learning_rate": 2.7300050915501895e-06, "loss": 0.5017, "step": 15347 }, { "epoch": 0.77, "learning_rate": 2.7288947486074525e-06, "loss": 0.3939, "step": 15348 }, { "epoch": 0.77, "learning_rate": 2.7277845958309458e-06, "loss": 0.2527, "step": 15349 }, { "epoch": 0.77, "learning_rate": 2.726674633249712e-06, "loss": 0.4214, "step": 15350 }, { "epoch": 0.77, "learning_rate": 2.7255648608927755e-06, "loss": 0.4898, "step": 15351 }, { "epoch": 0.77, "learning_rate": 2.7244552787891633e-06, "loss": 0.4058, "step": 15352 }, { "epoch": 0.77, "learning_rate": 2.723345886967893e-06, "loss": 0.1998, "step": 15353 }, { "epoch": 0.77, "learning_rate": 2.7222366854579818e-06, "loss": 0.3061, "step": 15354 }, { "epoch": 0.77, "learning_rate": 2.7211276742884375e-06, "loss": 0.3082, "step": 15355 }, { "epoch": 0.77, "learning_rate": 2.7200188534882644e-06, "loss": 0.308, "step": 15356 }, { "epoch": 0.77, "learning_rate": 2.718910223086464e-06, "loss": 0.3557, "step": 15357 }, { "epoch": 0.77, "learning_rate": 2.7178017831120263e-06, "loss": 0.3183, "step": 15358 }, { "epoch": 0.77, "learning_rate": 2.716693533593949e-06, "loss": 0.3956, "step": 15359 }, { "epoch": 0.77, "learning_rate": 2.715585474561212e-06, "loss": 0.3838, "step": 15360 }, { "epoch": 0.77, "learning_rate": 2.714477606042797e-06, "loss": 0.3787, "step": 15361 }, { "epoch": 0.77, "learning_rate": 2.7133699280676774e-06, "loss": 0.5181, "step": 15362 }, { "epoch": 0.77, "learning_rate": 2.71226244066482e-06, "loss": 0.377, "step": 15363 }, { "epoch": 0.77, "learning_rate": 2.7111551438631977e-06, "loss": 0.38, "step": 15364 }, { "epoch": 0.77, "learning_rate": 2.710048037691765e-06, "loss": 0.4148, "step": 15365 }, { "epoch": 0.77, "learning_rate": 2.708941122179477e-06, "loss": 0.2308, "step": 15366 }, { "epoch": 0.77, "learning_rate": 2.7078343973552823e-06, "loss": 0.2817, "step": 15367 }, { "epoch": 0.77, "learning_rate": 2.706727863248131e-06, "loss": 0.3562, "step": 15368 }, { "epoch": 0.77, "learning_rate": 2.7056215198869593e-06, "loss": 0.4488, "step": 15369 }, { "epoch": 0.77, "learning_rate": 2.704515367300704e-06, "loss": 0.2947, "step": 15370 }, { "epoch": 0.77, "learning_rate": 2.7034094055182935e-06, "loss": 0.4672, "step": 15371 }, { "epoch": 0.77, "learning_rate": 2.7023036345686494e-06, "loss": 0.4384, "step": 15372 }, { "epoch": 0.77, "learning_rate": 2.7011980544806992e-06, "loss": 0.4481, "step": 15373 }, { "epoch": 0.77, "learning_rate": 2.700092665283354e-06, "loss": 0.5324, "step": 15374 }, { "epoch": 0.77, "learning_rate": 2.698987467005524e-06, "loss": 0.399, "step": 15375 }, { "epoch": 0.77, "learning_rate": 2.6978824596761144e-06, "loss": 0.4907, "step": 15376 }, { "epoch": 0.77, "learning_rate": 2.6967776433240245e-06, "loss": 0.3368, "step": 15377 }, { "epoch": 0.77, "learning_rate": 2.69567301797815e-06, "loss": 0.3113, "step": 15378 }, { "epoch": 0.77, "learning_rate": 2.6945685836673807e-06, "loss": 0.2109, "step": 15379 }, { "epoch": 0.77, "learning_rate": 2.6934643404206018e-06, "loss": 0.5398, "step": 15380 }, { "epoch": 0.77, "learning_rate": 2.69236028826669e-06, "loss": 0.4121, "step": 15381 }, { "epoch": 0.77, "learning_rate": 2.6912564272345265e-06, "loss": 0.1818, "step": 15382 }, { "epoch": 0.77, "learning_rate": 2.6901527573529795e-06, "loss": 0.4022, "step": 15383 }, { "epoch": 0.77, "learning_rate": 2.6890492786509116e-06, "loss": 0.2765, "step": 15384 }, { "epoch": 0.77, "learning_rate": 2.687945991157185e-06, "loss": 0.2206, "step": 15385 }, { "epoch": 0.77, "learning_rate": 2.6868428949006508e-06, "loss": 0.4657, "step": 15386 }, { "epoch": 0.77, "learning_rate": 2.6857399899101654e-06, "loss": 0.4863, "step": 15387 }, { "epoch": 0.77, "learning_rate": 2.68463727621457e-06, "loss": 0.6112, "step": 15388 }, { "epoch": 0.77, "learning_rate": 2.683534753842706e-06, "loss": 0.456, "step": 15389 }, { "epoch": 0.77, "learning_rate": 2.6824324228234033e-06, "loss": 0.2624, "step": 15390 }, { "epoch": 0.77, "learning_rate": 2.6813302831855005e-06, "loss": 0.4257, "step": 15391 }, { "epoch": 0.77, "learning_rate": 2.6802283349578183e-06, "loss": 0.3144, "step": 15392 }, { "epoch": 0.77, "learning_rate": 2.6791265781691766e-06, "loss": 0.4479, "step": 15393 }, { "epoch": 0.77, "learning_rate": 2.678025012848391e-06, "loss": 0.4449, "step": 15394 }, { "epoch": 0.77, "learning_rate": 2.676923639024267e-06, "loss": 0.3187, "step": 15395 }, { "epoch": 0.77, "learning_rate": 2.6758224567256176e-06, "loss": 0.2999, "step": 15396 }, { "epoch": 0.77, "learning_rate": 2.674721465981239e-06, "loss": 0.2433, "step": 15397 }, { "epoch": 0.77, "learning_rate": 2.673620666819926e-06, "loss": 0.4602, "step": 15398 }, { "epoch": 0.77, "learning_rate": 2.6725200592704647e-06, "loss": 0.4663, "step": 15399 }, { "epoch": 0.77, "learning_rate": 2.671419643361648e-06, "loss": 0.3586, "step": 15400 }, { "epoch": 0.77, "learning_rate": 2.6703194191222513e-06, "loss": 0.2306, "step": 15401 }, { "epoch": 0.77, "learning_rate": 2.669219386581051e-06, "loss": 0.3164, "step": 15402 }, { "epoch": 0.77, "learning_rate": 2.668119545766815e-06, "loss": 0.2858, "step": 15403 }, { "epoch": 0.77, "learning_rate": 2.6670198967083095e-06, "loss": 0.1342, "step": 15404 }, { "epoch": 0.77, "learning_rate": 2.665920439434294e-06, "loss": 0.2383, "step": 15405 }, { "epoch": 0.77, "learning_rate": 2.664821173973524e-06, "loss": 0.3593, "step": 15406 }, { "epoch": 0.77, "learning_rate": 2.6637221003547485e-06, "loss": 0.3937, "step": 15407 }, { "epoch": 0.77, "learning_rate": 2.66262321860671e-06, "loss": 0.3264, "step": 15408 }, { "epoch": 0.77, "learning_rate": 2.6615245287581537e-06, "loss": 0.3744, "step": 15409 }, { "epoch": 0.77, "learning_rate": 2.6604260308378115e-06, "loss": 0.2592, "step": 15410 }, { "epoch": 0.77, "learning_rate": 2.659327724874413e-06, "loss": 0.3859, "step": 15411 }, { "epoch": 0.77, "learning_rate": 2.658229610896684e-06, "loss": 0.2341, "step": 15412 }, { "epoch": 0.77, "learning_rate": 2.657131688933341e-06, "loss": 0.3477, "step": 15413 }, { "epoch": 0.77, "learning_rate": 2.6560339590131033e-06, "loss": 0.2831, "step": 15414 }, { "epoch": 0.77, "learning_rate": 2.654936421164679e-06, "loss": 0.3454, "step": 15415 }, { "epoch": 0.77, "learning_rate": 2.653839075416772e-06, "loss": 0.3952, "step": 15416 }, { "epoch": 0.77, "learning_rate": 2.652741921798082e-06, "loss": 0.1484, "step": 15417 }, { "epoch": 0.77, "learning_rate": 2.651644960337302e-06, "loss": 0.2932, "step": 15418 }, { "epoch": 0.77, "learning_rate": 2.6505481910631246e-06, "loss": 0.2109, "step": 15419 }, { "epoch": 0.77, "learning_rate": 2.649451614004235e-06, "loss": 0.4737, "step": 15420 }, { "epoch": 0.77, "learning_rate": 2.6483552291893098e-06, "loss": 0.3526, "step": 15421 }, { "epoch": 0.77, "learning_rate": 2.647259036647022e-06, "loss": 0.4191, "step": 15422 }, { "epoch": 0.77, "learning_rate": 2.6461630364060476e-06, "loss": 0.3536, "step": 15423 }, { "epoch": 0.77, "learning_rate": 2.6450672284950463e-06, "loss": 0.4065, "step": 15424 }, { "epoch": 0.77, "learning_rate": 2.6439716129426786e-06, "loss": 0.4854, "step": 15425 }, { "epoch": 0.77, "learning_rate": 2.642876189777598e-06, "loss": 0.3824, "step": 15426 }, { "epoch": 0.77, "learning_rate": 2.6417809590284547e-06, "loss": 0.3584, "step": 15427 }, { "epoch": 0.77, "learning_rate": 2.6406859207238944e-06, "loss": 0.6529, "step": 15428 }, { "epoch": 0.77, "learning_rate": 2.6395910748925535e-06, "loss": 0.344, "step": 15429 }, { "epoch": 0.77, "learning_rate": 2.638496421563068e-06, "loss": 0.3494, "step": 15430 }, { "epoch": 0.77, "learning_rate": 2.6374019607640634e-06, "loss": 0.3754, "step": 15431 }, { "epoch": 0.77, "learning_rate": 2.6363076925241703e-06, "loss": 0.5138, "step": 15432 }, { "epoch": 0.77, "learning_rate": 2.6352136168720056e-06, "loss": 0.2001, "step": 15433 }, { "epoch": 0.77, "learning_rate": 2.6341197338361823e-06, "loss": 0.1644, "step": 15434 }, { "epoch": 0.77, "learning_rate": 2.633026043445309e-06, "loss": 0.4124, "step": 15435 }, { "epoch": 0.77, "learning_rate": 2.6319325457279888e-06, "loss": 0.333, "step": 15436 }, { "epoch": 0.77, "learning_rate": 2.6308392407128234e-06, "loss": 0.2887, "step": 15437 }, { "epoch": 0.77, "learning_rate": 2.6297461284284077e-06, "loss": 0.3986, "step": 15438 }, { "epoch": 0.77, "learning_rate": 2.628653208903328e-06, "loss": 0.4292, "step": 15439 }, { "epoch": 0.77, "learning_rate": 2.627560482166166e-06, "loss": 0.276, "step": 15440 }, { "epoch": 0.77, "learning_rate": 2.626467948245507e-06, "loss": 0.3049, "step": 15441 }, { "epoch": 0.77, "learning_rate": 2.62537560716992e-06, "loss": 0.3816, "step": 15442 }, { "epoch": 0.77, "learning_rate": 2.6242834589679756e-06, "loss": 0.38, "step": 15443 }, { "epoch": 0.77, "learning_rate": 2.6231915036682378e-06, "loss": 0.3937, "step": 15444 }, { "epoch": 0.77, "learning_rate": 2.6220997412992598e-06, "loss": 0.3889, "step": 15445 }, { "epoch": 0.77, "learning_rate": 2.621008171889604e-06, "loss": 0.2956, "step": 15446 }, { "epoch": 0.77, "learning_rate": 2.619916795467814e-06, "loss": 0.5079, "step": 15447 }, { "epoch": 0.77, "learning_rate": 2.6188256120624335e-06, "loss": 0.4841, "step": 15448 }, { "epoch": 0.77, "learning_rate": 2.6177346217020028e-06, "loss": 0.4373, "step": 15449 }, { "epoch": 0.77, "learning_rate": 2.6166438244150527e-06, "loss": 0.1918, "step": 15450 }, { "epoch": 0.77, "learning_rate": 2.6155532202301136e-06, "loss": 0.3273, "step": 15451 }, { "epoch": 0.77, "learning_rate": 2.6144628091757083e-06, "loss": 0.339, "step": 15452 }, { "epoch": 0.77, "learning_rate": 2.613372591280352e-06, "loss": 0.3498, "step": 15453 }, { "epoch": 0.77, "learning_rate": 2.612282566572564e-06, "loss": 0.3002, "step": 15454 }, { "epoch": 0.77, "learning_rate": 2.6111927350808485e-06, "loss": 0.2365, "step": 15455 }, { "epoch": 0.77, "learning_rate": 2.6101030968337104e-06, "loss": 0.4446, "step": 15456 }, { "epoch": 0.77, "learning_rate": 2.6090136518596453e-06, "loss": 0.4186, "step": 15457 }, { "epoch": 0.77, "learning_rate": 2.6079244001871494e-06, "loss": 0.1532, "step": 15458 }, { "epoch": 0.77, "learning_rate": 2.6068353418447044e-06, "loss": 0.2175, "step": 15459 }, { "epoch": 0.77, "learning_rate": 2.6057464768608022e-06, "loss": 0.262, "step": 15460 }, { "epoch": 0.77, "learning_rate": 2.6046578052639147e-06, "loss": 0.475, "step": 15461 }, { "epoch": 0.77, "learning_rate": 2.603569327082517e-06, "loss": 0.2394, "step": 15462 }, { "epoch": 0.77, "learning_rate": 2.602481042345073e-06, "loss": 0.3637, "step": 15463 }, { "epoch": 0.77, "learning_rate": 2.6013929510800515e-06, "loss": 0.3784, "step": 15464 }, { "epoch": 0.77, "learning_rate": 2.6003050533159068e-06, "loss": 0.5555, "step": 15465 }, { "epoch": 0.77, "learning_rate": 2.5992173490810913e-06, "loss": 0.3241, "step": 15466 }, { "epoch": 0.77, "learning_rate": 2.5981298384040523e-06, "loss": 0.1983, "step": 15467 }, { "epoch": 0.77, "learning_rate": 2.59704252131323e-06, "loss": 0.2589, "step": 15468 }, { "epoch": 0.77, "learning_rate": 2.5959553978370677e-06, "loss": 0.3075, "step": 15469 }, { "epoch": 0.77, "learning_rate": 2.594868468003995e-06, "loss": 0.4277, "step": 15470 }, { "epoch": 0.77, "learning_rate": 2.5937817318424373e-06, "loss": 0.2779, "step": 15471 }, { "epoch": 0.77, "learning_rate": 2.5926951893808163e-06, "loss": 0.576, "step": 15472 }, { "epoch": 0.77, "learning_rate": 2.5916088406475526e-06, "loss": 0.3, "step": 15473 }, { "epoch": 0.77, "learning_rate": 2.5905226856710564e-06, "loss": 0.2855, "step": 15474 }, { "epoch": 0.77, "learning_rate": 2.5894367244797357e-06, "loss": 0.3182, "step": 15475 }, { "epoch": 0.77, "learning_rate": 2.5883509571019903e-06, "loss": 0.4552, "step": 15476 }, { "epoch": 0.77, "learning_rate": 2.5872653835662187e-06, "loss": 0.3339, "step": 15477 }, { "epoch": 0.77, "learning_rate": 2.5861800039008122e-06, "loss": 0.41, "step": 15478 }, { "epoch": 0.77, "learning_rate": 2.5850948181341574e-06, "loss": 0.4474, "step": 15479 }, { "epoch": 0.77, "learning_rate": 2.5840098262946358e-06, "loss": 0.6382, "step": 15480 }, { "epoch": 0.77, "learning_rate": 2.5829250284106234e-06, "loss": 0.2941, "step": 15481 }, { "epoch": 0.77, "learning_rate": 2.58184042451049e-06, "loss": 0.5838, "step": 15482 }, { "epoch": 0.77, "learning_rate": 2.5807560146226063e-06, "loss": 0.3764, "step": 15483 }, { "epoch": 0.77, "learning_rate": 2.579671798775333e-06, "loss": 0.3202, "step": 15484 }, { "epoch": 0.77, "learning_rate": 2.5785877769970237e-06, "loss": 0.3771, "step": 15485 }, { "epoch": 0.77, "learning_rate": 2.577503949316028e-06, "loss": 0.3592, "step": 15486 }, { "epoch": 0.77, "learning_rate": 2.576420315760698e-06, "loss": 0.1442, "step": 15487 }, { "epoch": 0.77, "learning_rate": 2.575336876359371e-06, "loss": 0.3146, "step": 15488 }, { "epoch": 0.77, "learning_rate": 2.5742536311403844e-06, "loss": 0.3445, "step": 15489 }, { "epoch": 0.77, "learning_rate": 2.573170580132067e-06, "loss": 0.2417, "step": 15490 }, { "epoch": 0.77, "learning_rate": 2.572087723362744e-06, "loss": 0.3391, "step": 15491 }, { "epoch": 0.77, "learning_rate": 2.5710050608607394e-06, "loss": 0.2861, "step": 15492 }, { "epoch": 0.77, "learning_rate": 2.569922592654368e-06, "loss": 0.4541, "step": 15493 }, { "epoch": 0.77, "learning_rate": 2.5688403187719403e-06, "loss": 0.3752, "step": 15494 }, { "epoch": 0.77, "learning_rate": 2.5677582392417568e-06, "loss": 0.3499, "step": 15495 }, { "epoch": 0.77, "learning_rate": 2.5666763540921245e-06, "loss": 0.2542, "step": 15496 }, { "epoch": 0.77, "learning_rate": 2.5655946633513373e-06, "loss": 0.2866, "step": 15497 }, { "epoch": 0.77, "learning_rate": 2.5645131670476843e-06, "loss": 0.2139, "step": 15498 }, { "epoch": 0.77, "learning_rate": 2.5634318652094502e-06, "loss": 0.2817, "step": 15499 }, { "epoch": 0.77, "learning_rate": 2.5623507578649166e-06, "loss": 0.515, "step": 15500 }, { "epoch": 0.77, "learning_rate": 2.561269845042357e-06, "loss": 0.3334, "step": 15501 }, { "epoch": 0.77, "learning_rate": 2.560189126770041e-06, "loss": 0.3156, "step": 15502 }, { "epoch": 0.77, "learning_rate": 2.5591086030762336e-06, "loss": 0.3201, "step": 15503 }, { "epoch": 0.77, "learning_rate": 2.5580282739891925e-06, "loss": 0.4396, "step": 15504 }, { "epoch": 0.77, "learning_rate": 2.5569481395371776e-06, "loss": 0.3035, "step": 15505 }, { "epoch": 0.77, "learning_rate": 2.5558681997484357e-06, "loss": 0.2467, "step": 15506 }, { "epoch": 0.77, "learning_rate": 2.5547884546512103e-06, "loss": 0.4435, "step": 15507 }, { "epoch": 0.77, "learning_rate": 2.553708904273742e-06, "loss": 0.5241, "step": 15508 }, { "epoch": 0.77, "learning_rate": 2.552629548644262e-06, "loss": 0.3786, "step": 15509 }, { "epoch": 0.77, "learning_rate": 2.5515503877910042e-06, "loss": 0.3856, "step": 15510 }, { "epoch": 0.77, "learning_rate": 2.55047142174219e-06, "loss": 0.3538, "step": 15511 }, { "epoch": 0.77, "learning_rate": 2.549392650526039e-06, "loss": 0.318, "step": 15512 }, { "epoch": 0.77, "learning_rate": 2.5483140741707612e-06, "loss": 0.3261, "step": 15513 }, { "epoch": 0.77, "learning_rate": 2.547235692704573e-06, "loss": 0.5271, "step": 15514 }, { "epoch": 0.77, "learning_rate": 2.546157506155672e-06, "loss": 0.3844, "step": 15515 }, { "epoch": 0.77, "learning_rate": 2.54507951455226e-06, "loss": 0.2688, "step": 15516 }, { "epoch": 0.77, "learning_rate": 2.5440017179225285e-06, "loss": 0.5537, "step": 15517 }, { "epoch": 0.77, "learning_rate": 2.5429241162946638e-06, "loss": 0.3696, "step": 15518 }, { "epoch": 0.77, "learning_rate": 2.5418467096968547e-06, "loss": 0.2945, "step": 15519 }, { "epoch": 0.77, "learning_rate": 2.540769498157277e-06, "loss": 0.3159, "step": 15520 }, { "epoch": 0.78, "learning_rate": 2.5396924817041022e-06, "loss": 0.5416, "step": 15521 }, { "epoch": 0.78, "learning_rate": 2.538615660365501e-06, "loss": 0.2829, "step": 15522 }, { "epoch": 0.78, "learning_rate": 2.5375390341696325e-06, "loss": 0.4805, "step": 15523 }, { "epoch": 0.78, "learning_rate": 2.536462603144658e-06, "loss": 0.3272, "step": 15524 }, { "epoch": 0.78, "learning_rate": 2.535386367318725e-06, "loss": 0.1968, "step": 15525 }, { "epoch": 0.78, "learning_rate": 2.5343103267199878e-06, "loss": 0.3303, "step": 15526 }, { "epoch": 0.78, "learning_rate": 2.5332344813765864e-06, "loss": 0.256, "step": 15527 }, { "epoch": 0.78, "learning_rate": 2.532158831316657e-06, "loss": 0.3769, "step": 15528 }, { "epoch": 0.78, "learning_rate": 2.5310833765683308e-06, "loss": 0.3667, "step": 15529 }, { "epoch": 0.78, "learning_rate": 2.5300081171597377e-06, "loss": 0.269, "step": 15530 }, { "epoch": 0.78, "learning_rate": 2.5289330531189984e-06, "loss": 0.5308, "step": 15531 }, { "epoch": 0.78, "learning_rate": 2.5278581844742265e-06, "loss": 0.3465, "step": 15532 }, { "epoch": 0.78, "learning_rate": 2.52678351125354e-06, "loss": 0.4137, "step": 15533 }, { "epoch": 0.78, "learning_rate": 2.525709033485043e-06, "loss": 0.3923, "step": 15534 }, { "epoch": 0.78, "learning_rate": 2.524634751196836e-06, "loss": 0.2811, "step": 15535 }, { "epoch": 0.78, "learning_rate": 2.523560664417014e-06, "loss": 0.2866, "step": 15536 }, { "epoch": 0.78, "learning_rate": 2.5224867731736724e-06, "loss": 0.4272, "step": 15537 }, { "epoch": 0.78, "learning_rate": 2.5214130774948964e-06, "loss": 0.2488, "step": 15538 }, { "epoch": 0.78, "learning_rate": 2.520339577408766e-06, "loss": 0.5318, "step": 15539 }, { "epoch": 0.78, "learning_rate": 2.519266272943357e-06, "loss": 0.4653, "step": 15540 }, { "epoch": 0.78, "learning_rate": 2.518193164126738e-06, "loss": 0.3603, "step": 15541 }, { "epoch": 0.78, "learning_rate": 2.5171202509869797e-06, "loss": 0.3862, "step": 15542 }, { "epoch": 0.78, "learning_rate": 2.5160475335521406e-06, "loss": 0.4073, "step": 15543 }, { "epoch": 0.78, "learning_rate": 2.514975011850277e-06, "loss": 0.2984, "step": 15544 }, { "epoch": 0.78, "learning_rate": 2.513902685909434e-06, "loss": 0.3201, "step": 15545 }, { "epoch": 0.78, "learning_rate": 2.512830555757665e-06, "loss": 0.2403, "step": 15546 }, { "epoch": 0.78, "learning_rate": 2.511758621423006e-06, "loss": 0.6945, "step": 15547 }, { "epoch": 0.78, "learning_rate": 2.510686882933493e-06, "loss": 0.5513, "step": 15548 }, { "epoch": 0.78, "learning_rate": 2.5096153403171552e-06, "loss": 0.3464, "step": 15549 }, { "epoch": 0.78, "learning_rate": 2.5085439936020173e-06, "loss": 0.3824, "step": 15550 }, { "epoch": 0.78, "learning_rate": 2.507472842816099e-06, "loss": 0.4712, "step": 15551 }, { "epoch": 0.78, "learning_rate": 2.506401887987415e-06, "loss": 0.3017, "step": 15552 }, { "epoch": 0.78, "learning_rate": 2.5053311291439754e-06, "loss": 0.3635, "step": 15553 }, { "epoch": 0.78, "learning_rate": 2.5042605663137844e-06, "loss": 0.182, "step": 15554 }, { "epoch": 0.78, "learning_rate": 2.5031901995248366e-06, "loss": 0.5562, "step": 15555 }, { "epoch": 0.78, "learning_rate": 2.5021200288051327e-06, "loss": 0.3408, "step": 15556 }, { "epoch": 0.78, "learning_rate": 2.50105005418266e-06, "loss": 0.3493, "step": 15557 }, { "epoch": 0.78, "learning_rate": 2.4999802756854008e-06, "loss": 0.2285, "step": 15558 }, { "epoch": 0.78, "learning_rate": 2.498910693341331e-06, "loss": 0.2517, "step": 15559 }, { "epoch": 0.78, "learning_rate": 2.49784130717843e-06, "loss": 0.1784, "step": 15560 }, { "epoch": 0.78, "learning_rate": 2.4967721172246653e-06, "loss": 0.4109, "step": 15561 }, { "epoch": 0.78, "learning_rate": 2.4957031235079967e-06, "loss": 0.5043, "step": 15562 }, { "epoch": 0.78, "learning_rate": 2.4946343260563845e-06, "loss": 0.3355, "step": 15563 }, { "epoch": 0.78, "learning_rate": 2.4935657248977786e-06, "loss": 0.1538, "step": 15564 }, { "epoch": 0.78, "learning_rate": 2.492497320060132e-06, "loss": 0.143, "step": 15565 }, { "epoch": 0.78, "learning_rate": 2.491429111571385e-06, "loss": 0.3074, "step": 15566 }, { "epoch": 0.78, "learning_rate": 2.490361099459475e-06, "loss": 0.1941, "step": 15567 }, { "epoch": 0.78, "learning_rate": 2.489293283752332e-06, "loss": 0.2116, "step": 15568 }, { "epoch": 0.78, "learning_rate": 2.488225664477889e-06, "loss": 0.2939, "step": 15569 }, { "epoch": 0.78, "learning_rate": 2.487158241664065e-06, "loss": 0.4057, "step": 15570 }, { "epoch": 0.78, "learning_rate": 2.4860910153387775e-06, "loss": 0.4915, "step": 15571 }, { "epoch": 0.78, "learning_rate": 2.4850239855299373e-06, "loss": 0.4156, "step": 15572 }, { "epoch": 0.78, "learning_rate": 2.4839571522654528e-06, "loss": 0.7119, "step": 15573 }, { "epoch": 0.78, "learning_rate": 2.482890515573223e-06, "loss": 0.4377, "step": 15574 }, { "epoch": 0.78, "learning_rate": 2.481824075481147e-06, "loss": 0.2672, "step": 15575 }, { "epoch": 0.78, "learning_rate": 2.480757832017117e-06, "loss": 0.3183, "step": 15576 }, { "epoch": 0.78, "learning_rate": 2.479691785209012e-06, "loss": 0.4124, "step": 15577 }, { "epoch": 0.78, "learning_rate": 2.4786259350847232e-06, "loss": 0.3464, "step": 15578 }, { "epoch": 0.78, "learning_rate": 2.4775602816721202e-06, "loss": 0.3515, "step": 15579 }, { "epoch": 0.78, "learning_rate": 2.476494824999076e-06, "loss": 0.2594, "step": 15580 }, { "epoch": 0.78, "learning_rate": 2.4754295650934566e-06, "loss": 0.2053, "step": 15581 }, { "epoch": 0.78, "learning_rate": 2.4743645019831175e-06, "loss": 0.5018, "step": 15582 }, { "epoch": 0.78, "learning_rate": 2.4732996356959216e-06, "loss": 0.2224, "step": 15583 }, { "epoch": 0.78, "learning_rate": 2.4722349662597158e-06, "loss": 0.367, "step": 15584 }, { "epoch": 0.78, "learning_rate": 2.471170493702344e-06, "loss": 0.3433, "step": 15585 }, { "epoch": 0.78, "learning_rate": 2.4701062180516476e-06, "loss": 0.3705, "step": 15586 }, { "epoch": 0.78, "learning_rate": 2.469042139335457e-06, "loss": 0.3154, "step": 15587 }, { "epoch": 0.78, "learning_rate": 2.467978257581609e-06, "loss": 0.3648, "step": 15588 }, { "epoch": 0.78, "learning_rate": 2.4669145728179246e-06, "loss": 0.4492, "step": 15589 }, { "epoch": 0.78, "learning_rate": 2.465851085072223e-06, "loss": 0.2797, "step": 15590 }, { "epoch": 0.78, "learning_rate": 2.4647877943723143e-06, "loss": 0.3215, "step": 15591 }, { "epoch": 0.78, "learning_rate": 2.4637247007460154e-06, "loss": 0.3643, "step": 15592 }, { "epoch": 0.78, "learning_rate": 2.462661804221126e-06, "loss": 0.5052, "step": 15593 }, { "epoch": 0.78, "learning_rate": 2.461599104825445e-06, "loss": 0.3521, "step": 15594 }, { "epoch": 0.78, "learning_rate": 2.460536602586766e-06, "loss": 0.2695, "step": 15595 }, { "epoch": 0.78, "learning_rate": 2.4594742975328766e-06, "loss": 0.2941, "step": 15596 }, { "epoch": 0.78, "learning_rate": 2.4584121896915593e-06, "loss": 0.4516, "step": 15597 }, { "epoch": 0.78, "learning_rate": 2.4573502790905945e-06, "loss": 0.3699, "step": 15598 }, { "epoch": 0.78, "learning_rate": 2.4562885657577553e-06, "loss": 0.3472, "step": 15599 }, { "epoch": 0.78, "learning_rate": 2.4552270497208077e-06, "loss": 0.3536, "step": 15600 }, { "epoch": 0.78, "learning_rate": 2.454165731007515e-06, "loss": 0.4242, "step": 15601 }, { "epoch": 0.78, "learning_rate": 2.4531046096456335e-06, "loss": 0.2701, "step": 15602 }, { "epoch": 0.78, "learning_rate": 2.452043685662917e-06, "loss": 0.3861, "step": 15603 }, { "epoch": 0.78, "learning_rate": 2.4509829590871127e-06, "loss": 0.2798, "step": 15604 }, { "epoch": 0.78, "learning_rate": 2.4499224299459566e-06, "loss": 0.3977, "step": 15605 }, { "epoch": 0.78, "learning_rate": 2.448862098267195e-06, "loss": 0.3417, "step": 15606 }, { "epoch": 0.78, "learning_rate": 2.4478019640785543e-06, "loss": 0.4973, "step": 15607 }, { "epoch": 0.78, "learning_rate": 2.4467420274077624e-06, "loss": 0.3514, "step": 15608 }, { "epoch": 0.78, "learning_rate": 2.445682288282537e-06, "loss": 0.4381, "step": 15609 }, { "epoch": 0.78, "learning_rate": 2.4446227467305994e-06, "loss": 0.1311, "step": 15610 }, { "epoch": 0.78, "learning_rate": 2.443563402779657e-06, "loss": 0.3984, "step": 15611 }, { "epoch": 0.78, "learning_rate": 2.4425042564574186e-06, "loss": 0.3214, "step": 15612 }, { "epoch": 0.78, "learning_rate": 2.4414453077915813e-06, "loss": 0.3406, "step": 15613 }, { "epoch": 0.78, "learning_rate": 2.4403865568098396e-06, "loss": 0.4068, "step": 15614 }, { "epoch": 0.78, "learning_rate": 2.439328003539889e-06, "loss": 0.5388, "step": 15615 }, { "epoch": 0.78, "learning_rate": 2.4382696480094127e-06, "loss": 0.4448, "step": 15616 }, { "epoch": 0.78, "learning_rate": 2.4372114902460886e-06, "loss": 0.4263, "step": 15617 }, { "epoch": 0.78, "learning_rate": 2.436153530277593e-06, "loss": 0.4845, "step": 15618 }, { "epoch": 0.78, "learning_rate": 2.435095768131592e-06, "loss": 0.1947, "step": 15619 }, { "epoch": 0.78, "learning_rate": 2.4340382038357557e-06, "loss": 0.3576, "step": 15620 }, { "epoch": 0.78, "learning_rate": 2.432980837417741e-06, "loss": 0.3219, "step": 15621 }, { "epoch": 0.78, "learning_rate": 2.4319236689052007e-06, "loss": 0.3578, "step": 15622 }, { "epoch": 0.78, "learning_rate": 2.4308666983257854e-06, "loss": 0.3808, "step": 15623 }, { "epoch": 0.78, "learning_rate": 2.4298099257071373e-06, "loss": 0.4133, "step": 15624 }, { "epoch": 0.78, "learning_rate": 2.4287533510768947e-06, "loss": 0.5755, "step": 15625 }, { "epoch": 0.78, "learning_rate": 2.4276969744626933e-06, "loss": 0.2816, "step": 15626 }, { "epoch": 0.78, "learning_rate": 2.426640795892158e-06, "loss": 0.3949, "step": 15627 }, { "epoch": 0.78, "learning_rate": 2.425584815392911e-06, "loss": 0.5034, "step": 15628 }, { "epoch": 0.78, "learning_rate": 2.4245290329925753e-06, "loss": 0.3709, "step": 15629 }, { "epoch": 0.78, "learning_rate": 2.4234734487187617e-06, "loss": 0.4187, "step": 15630 }, { "epoch": 0.78, "learning_rate": 2.4224180625990755e-06, "loss": 0.6031, "step": 15631 }, { "epoch": 0.78, "learning_rate": 2.4213628746611173e-06, "loss": 0.3412, "step": 15632 }, { "epoch": 0.78, "learning_rate": 2.420307884932489e-06, "loss": 0.4772, "step": 15633 }, { "epoch": 0.78, "learning_rate": 2.4192530934407822e-06, "loss": 0.3621, "step": 15634 }, { "epoch": 0.78, "learning_rate": 2.41819850021358e-06, "loss": 0.3073, "step": 15635 }, { "epoch": 0.78, "learning_rate": 2.4171441052784662e-06, "loss": 0.5197, "step": 15636 }, { "epoch": 0.78, "learning_rate": 2.416089908663014e-06, "loss": 0.2182, "step": 15637 }, { "epoch": 0.78, "learning_rate": 2.4150359103947997e-06, "loss": 0.6747, "step": 15638 }, { "epoch": 0.78, "learning_rate": 2.4139821105013874e-06, "loss": 0.3353, "step": 15639 }, { "epoch": 0.78, "learning_rate": 2.412928509010336e-06, "loss": 0.3468, "step": 15640 }, { "epoch": 0.78, "learning_rate": 2.4118751059491995e-06, "loss": 0.2064, "step": 15641 }, { "epoch": 0.78, "learning_rate": 2.410821901345535e-06, "loss": 0.2679, "step": 15642 }, { "epoch": 0.78, "learning_rate": 2.4097688952268826e-06, "loss": 0.4297, "step": 15643 }, { "epoch": 0.78, "learning_rate": 2.4087160876207827e-06, "loss": 0.3399, "step": 15644 }, { "epoch": 0.78, "learning_rate": 2.4076634785547716e-06, "loss": 0.5755, "step": 15645 }, { "epoch": 0.78, "learning_rate": 2.406611068056378e-06, "loss": 0.1265, "step": 15646 }, { "epoch": 0.78, "learning_rate": 2.4055588561531263e-06, "loss": 0.4073, "step": 15647 }, { "epoch": 0.78, "learning_rate": 2.404506842872536e-06, "loss": 0.5842, "step": 15648 }, { "epoch": 0.78, "learning_rate": 2.403455028242121e-06, "loss": 0.2004, "step": 15649 }, { "epoch": 0.78, "learning_rate": 2.402403412289387e-06, "loss": 0.2518, "step": 15650 }, { "epoch": 0.78, "learning_rate": 2.401351995041843e-06, "loss": 0.4623, "step": 15651 }, { "epoch": 0.78, "learning_rate": 2.4003007765269858e-06, "loss": 0.3766, "step": 15652 }, { "epoch": 0.78, "learning_rate": 2.399249756772307e-06, "loss": 0.335, "step": 15653 }, { "epoch": 0.78, "learning_rate": 2.398198935805297e-06, "loss": 0.2249, "step": 15654 }, { "epoch": 0.78, "learning_rate": 2.3971483136534335e-06, "loss": 0.3324, "step": 15655 }, { "epoch": 0.78, "learning_rate": 2.396097890344201e-06, "loss": 0.3724, "step": 15656 }, { "epoch": 0.78, "learning_rate": 2.395047665905069e-06, "loss": 0.2129, "step": 15657 }, { "epoch": 0.78, "learning_rate": 2.3939976403635044e-06, "loss": 0.2111, "step": 15658 }, { "epoch": 0.78, "learning_rate": 2.3929478137469697e-06, "loss": 0.4274, "step": 15659 }, { "epoch": 0.78, "learning_rate": 2.391898186082918e-06, "loss": 0.2236, "step": 15660 }, { "epoch": 0.78, "learning_rate": 2.3908487573988083e-06, "loss": 0.348, "step": 15661 }, { "epoch": 0.78, "learning_rate": 2.389799527722082e-06, "loss": 0.3408, "step": 15662 }, { "epoch": 0.78, "learning_rate": 2.388750497080182e-06, "loss": 0.2902, "step": 15663 }, { "epoch": 0.78, "learning_rate": 2.3877016655005414e-06, "loss": 0.5249, "step": 15664 }, { "epoch": 0.78, "learning_rate": 2.3866530330105965e-06, "loss": 0.3113, "step": 15665 }, { "epoch": 0.78, "learning_rate": 2.3856045996377687e-06, "loss": 0.5092, "step": 15666 }, { "epoch": 0.78, "learning_rate": 2.3845563654094807e-06, "loss": 0.3539, "step": 15667 }, { "epoch": 0.78, "learning_rate": 2.383508330353146e-06, "loss": 0.1914, "step": 15668 }, { "epoch": 0.78, "learning_rate": 2.3824604944961717e-06, "loss": 0.1583, "step": 15669 }, { "epoch": 0.78, "learning_rate": 2.381412857865969e-06, "loss": 0.3901, "step": 15670 }, { "epoch": 0.78, "learning_rate": 2.3803654204899353e-06, "loss": 0.5031, "step": 15671 }, { "epoch": 0.78, "learning_rate": 2.3793181823954627e-06, "loss": 0.2429, "step": 15672 }, { "epoch": 0.78, "learning_rate": 2.3782711436099427e-06, "loss": 0.5118, "step": 15673 }, { "epoch": 0.78, "learning_rate": 2.377224304160758e-06, "loss": 0.3004, "step": 15674 }, { "epoch": 0.78, "learning_rate": 2.3761776640752878e-06, "loss": 0.6387, "step": 15675 }, { "epoch": 0.78, "learning_rate": 2.3751312233809054e-06, "loss": 0.4095, "step": 15676 }, { "epoch": 0.78, "learning_rate": 2.3740849821049794e-06, "loss": 0.2413, "step": 15677 }, { "epoch": 0.78, "learning_rate": 2.3730389402748687e-06, "loss": 0.4748, "step": 15678 }, { "epoch": 0.78, "learning_rate": 2.3719930979179382e-06, "loss": 0.3716, "step": 15679 }, { "epoch": 0.78, "learning_rate": 2.3709474550615373e-06, "loss": 0.3426, "step": 15680 }, { "epoch": 0.78, "learning_rate": 2.369902011733014e-06, "loss": 0.2561, "step": 15681 }, { "epoch": 0.78, "learning_rate": 2.3688567679597073e-06, "loss": 0.2369, "step": 15682 }, { "epoch": 0.78, "learning_rate": 2.3678117237689592e-06, "loss": 0.1937, "step": 15683 }, { "epoch": 0.78, "learning_rate": 2.3667668791880982e-06, "loss": 0.4094, "step": 15684 }, { "epoch": 0.78, "learning_rate": 2.365722234244453e-06, "loss": 0.4298, "step": 15685 }, { "epoch": 0.78, "learning_rate": 2.3646777889653426e-06, "loss": 0.3984, "step": 15686 }, { "epoch": 0.78, "learning_rate": 2.363633543378082e-06, "loss": 0.4145, "step": 15687 }, { "epoch": 0.78, "learning_rate": 2.362589497509987e-06, "loss": 0.6006, "step": 15688 }, { "epoch": 0.78, "learning_rate": 2.3615456513883607e-06, "loss": 0.3532, "step": 15689 }, { "epoch": 0.78, "learning_rate": 2.3605020050405035e-06, "loss": 0.2639, "step": 15690 }, { "epoch": 0.78, "learning_rate": 2.3594585584937103e-06, "loss": 0.3147, "step": 15691 }, { "epoch": 0.78, "learning_rate": 2.358415311775267e-06, "loss": 0.2974, "step": 15692 }, { "epoch": 0.78, "learning_rate": 2.3573722649124674e-06, "loss": 0.3581, "step": 15693 }, { "epoch": 0.78, "learning_rate": 2.3563294179325846e-06, "loss": 0.3824, "step": 15694 }, { "epoch": 0.78, "learning_rate": 2.3552867708628955e-06, "loss": 0.3895, "step": 15695 }, { "epoch": 0.78, "learning_rate": 2.3542443237306678e-06, "loss": 0.2853, "step": 15696 }, { "epoch": 0.78, "learning_rate": 2.3532020765631656e-06, "loss": 0.3625, "step": 15697 }, { "epoch": 0.78, "learning_rate": 2.352160029387648e-06, "loss": 0.1671, "step": 15698 }, { "epoch": 0.78, "learning_rate": 2.3511181822313665e-06, "loss": 0.258, "step": 15699 }, { "epoch": 0.78, "learning_rate": 2.350076535121573e-06, "loss": 0.4933, "step": 15700 }, { "epoch": 0.78, "learning_rate": 2.349035088085503e-06, "loss": 0.3336, "step": 15701 }, { "epoch": 0.78, "learning_rate": 2.347993841150403e-06, "loss": 0.2917, "step": 15702 }, { "epoch": 0.78, "learning_rate": 2.346952794343502e-06, "loss": 0.3804, "step": 15703 }, { "epoch": 0.78, "learning_rate": 2.3459119476920277e-06, "loss": 0.462, "step": 15704 }, { "epoch": 0.78, "learning_rate": 2.344871301223197e-06, "loss": 0.3897, "step": 15705 }, { "epoch": 0.78, "learning_rate": 2.3438308549642354e-06, "loss": 0.2109, "step": 15706 }, { "epoch": 0.78, "learning_rate": 2.342790608942349e-06, "loss": 0.4685, "step": 15707 }, { "epoch": 0.78, "learning_rate": 2.3417505631847448e-06, "loss": 0.3825, "step": 15708 }, { "epoch": 0.78, "learning_rate": 2.340710717718625e-06, "loss": 0.0893, "step": 15709 }, { "epoch": 0.78, "learning_rate": 2.3396710725711813e-06, "loss": 0.4681, "step": 15710 }, { "epoch": 0.78, "learning_rate": 2.33863162776961e-06, "loss": 0.4987, "step": 15711 }, { "epoch": 0.78, "learning_rate": 2.337592383341094e-06, "loss": 0.7006, "step": 15712 }, { "epoch": 0.78, "learning_rate": 2.3365533393128135e-06, "loss": 0.4952, "step": 15713 }, { "epoch": 0.78, "learning_rate": 2.33551449571194e-06, "loss": 0.2045, "step": 15714 }, { "epoch": 0.78, "learning_rate": 2.3344758525656498e-06, "loss": 0.2443, "step": 15715 }, { "epoch": 0.78, "learning_rate": 2.3334374099011025e-06, "loss": 0.3116, "step": 15716 }, { "epoch": 0.78, "learning_rate": 2.3323991677454584e-06, "loss": 0.3855, "step": 15717 }, { "epoch": 0.78, "learning_rate": 2.3313611261258717e-06, "loss": 0.4525, "step": 15718 }, { "epoch": 0.78, "learning_rate": 2.330323285069491e-06, "loss": 0.3993, "step": 15719 }, { "epoch": 0.78, "learning_rate": 2.3292856446034596e-06, "loss": 0.5208, "step": 15720 }, { "epoch": 0.78, "learning_rate": 2.3282482047549147e-06, "loss": 0.1472, "step": 15721 }, { "epoch": 0.79, "learning_rate": 2.32721096555099e-06, "loss": 0.4881, "step": 15722 }, { "epoch": 0.79, "learning_rate": 2.326173927018812e-06, "loss": 0.1837, "step": 15723 }, { "epoch": 0.79, "learning_rate": 2.325137089185503e-06, "loss": 0.237, "step": 15724 }, { "epoch": 0.79, "learning_rate": 2.324100452078183e-06, "loss": 0.302, "step": 15725 }, { "epoch": 0.79, "learning_rate": 2.3230640157239614e-06, "loss": 0.2577, "step": 15726 }, { "epoch": 0.79, "learning_rate": 2.322027780149947e-06, "loss": 0.4843, "step": 15727 }, { "epoch": 0.79, "learning_rate": 2.3209917453832353e-06, "loss": 0.5544, "step": 15728 }, { "epoch": 0.79, "learning_rate": 2.3199559114509308e-06, "loss": 0.4674, "step": 15729 }, { "epoch": 0.79, "learning_rate": 2.31892027838012e-06, "loss": 0.4053, "step": 15730 }, { "epoch": 0.79, "learning_rate": 2.3178848461978887e-06, "loss": 0.5676, "step": 15731 }, { "epoch": 0.79, "learning_rate": 2.3168496149313167e-06, "loss": 0.1569, "step": 15732 }, { "epoch": 0.79, "learning_rate": 2.3158145846074788e-06, "loss": 0.3006, "step": 15733 }, { "epoch": 0.79, "learning_rate": 2.3147797552534477e-06, "loss": 0.3216, "step": 15734 }, { "epoch": 0.79, "learning_rate": 2.313745126896286e-06, "loss": 0.3295, "step": 15735 }, { "epoch": 0.79, "learning_rate": 2.3127106995630544e-06, "loss": 0.3514, "step": 15736 }, { "epoch": 0.79, "learning_rate": 2.3116764732808028e-06, "loss": 0.3526, "step": 15737 }, { "epoch": 0.79, "learning_rate": 2.3106424480765855e-06, "loss": 0.3649, "step": 15738 }, { "epoch": 0.79, "learning_rate": 2.309608623977444e-06, "loss": 0.2215, "step": 15739 }, { "epoch": 0.79, "learning_rate": 2.308575001010417e-06, "loss": 0.1407, "step": 15740 }, { "epoch": 0.79, "learning_rate": 2.307541579202536e-06, "loss": 0.2468, "step": 15741 }, { "epoch": 0.79, "learning_rate": 2.306508358580827e-06, "loss": 0.4705, "step": 15742 }, { "epoch": 0.79, "learning_rate": 2.3054753391723185e-06, "loss": 0.2961, "step": 15743 }, { "epoch": 0.79, "learning_rate": 2.3044425210040233e-06, "loss": 0.3173, "step": 15744 }, { "epoch": 0.79, "learning_rate": 2.3034099041029558e-06, "loss": 0.1932, "step": 15745 }, { "epoch": 0.79, "learning_rate": 2.3023774884961214e-06, "loss": 0.3963, "step": 15746 }, { "epoch": 0.79, "learning_rate": 2.301345274210521e-06, "loss": 0.6649, "step": 15747 }, { "epoch": 0.79, "learning_rate": 2.3003132612731516e-06, "loss": 0.2194, "step": 15748 }, { "epoch": 0.79, "learning_rate": 2.299281449711005e-06, "loss": 0.3687, "step": 15749 }, { "epoch": 0.79, "learning_rate": 2.2982498395510656e-06, "loss": 0.2932, "step": 15750 }, { "epoch": 0.79, "learning_rate": 2.2972184308203107e-06, "loss": 0.3847, "step": 15751 }, { "epoch": 0.79, "learning_rate": 2.2961872235457226e-06, "loss": 0.2955, "step": 15752 }, { "epoch": 0.79, "learning_rate": 2.2951562177542663e-06, "loss": 0.2392, "step": 15753 }, { "epoch": 0.79, "learning_rate": 2.2941254134729086e-06, "loss": 0.3231, "step": 15754 }, { "epoch": 0.79, "learning_rate": 2.2930948107286077e-06, "loss": 0.6267, "step": 15755 }, { "epoch": 0.79, "learning_rate": 2.2920644095483145e-06, "loss": 0.417, "step": 15756 }, { "epoch": 0.79, "learning_rate": 2.2910342099589845e-06, "loss": 0.3508, "step": 15757 }, { "epoch": 0.79, "learning_rate": 2.2900042119875576e-06, "loss": 0.3947, "step": 15758 }, { "epoch": 0.79, "learning_rate": 2.288974415660973e-06, "loss": 0.2417, "step": 15759 }, { "epoch": 0.79, "learning_rate": 2.2879448210061584e-06, "loss": 0.2054, "step": 15760 }, { "epoch": 0.79, "learning_rate": 2.2869154280500505e-06, "loss": 0.3316, "step": 15761 }, { "epoch": 0.79, "learning_rate": 2.2858862368195667e-06, "loss": 0.4395, "step": 15762 }, { "epoch": 0.79, "learning_rate": 2.284857247341624e-06, "loss": 0.4454, "step": 15763 }, { "epoch": 0.79, "learning_rate": 2.283828459643136e-06, "loss": 0.2158, "step": 15764 }, { "epoch": 0.79, "learning_rate": 2.2827998737510047e-06, "loss": 0.3273, "step": 15765 }, { "epoch": 0.79, "learning_rate": 2.2817714896921383e-06, "loss": 0.3638, "step": 15766 }, { "epoch": 0.79, "learning_rate": 2.2807433074934294e-06, "loss": 0.4379, "step": 15767 }, { "epoch": 0.79, "learning_rate": 2.2797153271817684e-06, "loss": 0.2388, "step": 15768 }, { "epoch": 0.79, "learning_rate": 2.278687548784042e-06, "loss": 0.2905, "step": 15769 }, { "epoch": 0.79, "learning_rate": 2.2776599723271287e-06, "loss": 0.3066, "step": 15770 }, { "epoch": 0.79, "learning_rate": 2.2766325978379057e-06, "loss": 0.4235, "step": 15771 }, { "epoch": 0.79, "learning_rate": 2.27560542534324e-06, "loss": 0.3468, "step": 15772 }, { "epoch": 0.79, "learning_rate": 2.274578454869998e-06, "loss": 0.4149, "step": 15773 }, { "epoch": 0.79, "learning_rate": 2.2735516864450345e-06, "loss": 0.2805, "step": 15774 }, { "epoch": 0.79, "learning_rate": 2.2725251200952102e-06, "loss": 0.4354, "step": 15775 }, { "epoch": 0.79, "learning_rate": 2.271498755847371e-06, "loss": 0.4372, "step": 15776 }, { "epoch": 0.79, "learning_rate": 2.270472593728358e-06, "loss": 0.3664, "step": 15777 }, { "epoch": 0.79, "learning_rate": 2.269446633765008e-06, "loss": 0.3082, "step": 15778 }, { "epoch": 0.79, "learning_rate": 2.2684208759841597e-06, "loss": 0.2614, "step": 15779 }, { "epoch": 0.79, "learning_rate": 2.267395320412635e-06, "loss": 0.5232, "step": 15780 }, { "epoch": 0.79, "learning_rate": 2.26636996707726e-06, "loss": 0.3709, "step": 15781 }, { "epoch": 0.79, "learning_rate": 2.2653448160048485e-06, "loss": 0.2606, "step": 15782 }, { "epoch": 0.79, "learning_rate": 2.2643198672222088e-06, "loss": 0.226, "step": 15783 }, { "epoch": 0.79, "learning_rate": 2.263295120756155e-06, "loss": 0.2178, "step": 15784 }, { "epoch": 0.79, "learning_rate": 2.262270576633484e-06, "loss": 0.3033, "step": 15785 }, { "epoch": 0.79, "learning_rate": 2.2612462348809916e-06, "loss": 0.3359, "step": 15786 }, { "epoch": 0.79, "learning_rate": 2.260222095525465e-06, "loss": 0.3706, "step": 15787 }, { "epoch": 0.79, "learning_rate": 2.259198158593695e-06, "loss": 0.2778, "step": 15788 }, { "epoch": 0.79, "learning_rate": 2.2581744241124582e-06, "loss": 0.1779, "step": 15789 }, { "epoch": 0.79, "learning_rate": 2.2571508921085294e-06, "loss": 0.461, "step": 15790 }, { "epoch": 0.79, "learning_rate": 2.2561275626086776e-06, "loss": 0.3735, "step": 15791 }, { "epoch": 0.79, "learning_rate": 2.2551044356396678e-06, "loss": 0.1875, "step": 15792 }, { "epoch": 0.79, "learning_rate": 2.2540815112282564e-06, "loss": 0.2889, "step": 15793 }, { "epoch": 0.79, "learning_rate": 2.253058789401198e-06, "loss": 0.2273, "step": 15794 }, { "epoch": 0.79, "learning_rate": 2.2520362701852405e-06, "loss": 0.2503, "step": 15795 }, { "epoch": 0.79, "learning_rate": 2.2510139536071264e-06, "loss": 0.2562, "step": 15796 }, { "epoch": 0.79, "learning_rate": 2.2499918396935893e-06, "loss": 0.4554, "step": 15797 }, { "epoch": 0.79, "learning_rate": 2.2489699284713685e-06, "loss": 0.3842, "step": 15798 }, { "epoch": 0.79, "learning_rate": 2.2479482199671878e-06, "loss": 0.3202, "step": 15799 }, { "epoch": 0.79, "learning_rate": 2.2469267142077668e-06, "loss": 0.403, "step": 15800 }, { "epoch": 0.79, "learning_rate": 2.245905411219821e-06, "loss": 0.2505, "step": 15801 }, { "epoch": 0.79, "learning_rate": 2.2448843110300657e-06, "loss": 0.4844, "step": 15802 }, { "epoch": 0.79, "learning_rate": 2.2438634136652037e-06, "loss": 0.4443, "step": 15803 }, { "epoch": 0.79, "learning_rate": 2.2428427191519353e-06, "loss": 0.5303, "step": 15804 }, { "epoch": 0.79, "learning_rate": 2.2418222275169553e-06, "loss": 0.5098, "step": 15805 }, { "epoch": 0.79, "learning_rate": 2.240801938786952e-06, "loss": 0.0825, "step": 15806 }, { "epoch": 0.79, "learning_rate": 2.2397818529886128e-06, "loss": 0.3895, "step": 15807 }, { "epoch": 0.79, "learning_rate": 2.238761970148616e-06, "loss": 0.6422, "step": 15808 }, { "epoch": 0.79, "learning_rate": 2.237742290293634e-06, "loss": 0.429, "step": 15809 }, { "epoch": 0.79, "learning_rate": 2.236722813450333e-06, "loss": 0.159, "step": 15810 }, { "epoch": 0.79, "learning_rate": 2.2357035396453818e-06, "loss": 0.4734, "step": 15811 }, { "epoch": 0.79, "learning_rate": 2.234684468905435e-06, "loss": 0.4182, "step": 15812 }, { "epoch": 0.79, "learning_rate": 2.2336656012571443e-06, "loss": 0.2575, "step": 15813 }, { "epoch": 0.79, "learning_rate": 2.2326469367271588e-06, "loss": 0.3542, "step": 15814 }, { "epoch": 0.79, "learning_rate": 2.2316284753421158e-06, "loss": 0.2928, "step": 15815 }, { "epoch": 0.79, "learning_rate": 2.230610217128658e-06, "loss": 0.3131, "step": 15816 }, { "epoch": 0.79, "learning_rate": 2.2295921621134155e-06, "loss": 0.3634, "step": 15817 }, { "epoch": 0.79, "learning_rate": 2.228574310323011e-06, "loss": 0.4168, "step": 15818 }, { "epoch": 0.79, "learning_rate": 2.2275566617840674e-06, "loss": 0.3407, "step": 15819 }, { "epoch": 0.79, "learning_rate": 2.226539216523198e-06, "loss": 0.2956, "step": 15820 }, { "epoch": 0.79, "learning_rate": 2.2255219745670153e-06, "loss": 0.1965, "step": 15821 }, { "epoch": 0.79, "learning_rate": 2.2245049359421224e-06, "loss": 0.2731, "step": 15822 }, { "epoch": 0.79, "learning_rate": 2.2234881006751187e-06, "loss": 0.3574, "step": 15823 }, { "epoch": 0.79, "learning_rate": 2.222471468792595e-06, "loss": 0.2826, "step": 15824 }, { "epoch": 0.79, "learning_rate": 2.2214550403211464e-06, "loss": 0.2445, "step": 15825 }, { "epoch": 0.79, "learning_rate": 2.2204388152873536e-06, "loss": 0.4687, "step": 15826 }, { "epoch": 0.79, "learning_rate": 2.219422793717794e-06, "loss": 0.1469, "step": 15827 }, { "epoch": 0.79, "learning_rate": 2.2184069756390404e-06, "loss": 0.5774, "step": 15828 }, { "epoch": 0.79, "learning_rate": 2.2173913610776577e-06, "loss": 0.4633, "step": 15829 }, { "epoch": 0.79, "learning_rate": 2.2163759500602133e-06, "loss": 0.5701, "step": 15830 }, { "epoch": 0.79, "learning_rate": 2.2153607426132604e-06, "loss": 0.344, "step": 15831 }, { "epoch": 0.79, "learning_rate": 2.2143457387633527e-06, "loss": 0.3565, "step": 15832 }, { "epoch": 0.79, "learning_rate": 2.2133309385370306e-06, "loss": 0.2123, "step": 15833 }, { "epoch": 0.79, "learning_rate": 2.2123163419608427e-06, "loss": 0.375, "step": 15834 }, { "epoch": 0.79, "learning_rate": 2.2113019490613206e-06, "loss": 0.4354, "step": 15835 }, { "epoch": 0.79, "learning_rate": 2.210287759864995e-06, "loss": 0.3952, "step": 15836 }, { "epoch": 0.79, "learning_rate": 2.2092737743983894e-06, "loss": 0.2806, "step": 15837 }, { "epoch": 0.79, "learning_rate": 2.2082599926880222e-06, "loss": 0.3443, "step": 15838 }, { "epoch": 0.79, "learning_rate": 2.2072464147604122e-06, "loss": 0.3531, "step": 15839 }, { "epoch": 0.79, "learning_rate": 2.2062330406420664e-06, "loss": 0.2212, "step": 15840 }, { "epoch": 0.79, "learning_rate": 2.2052198703594862e-06, "loss": 0.248, "step": 15841 }, { "epoch": 0.79, "learning_rate": 2.2042069039391712e-06, "loss": 0.2462, "step": 15842 }, { "epoch": 0.79, "learning_rate": 2.203194141407614e-06, "loss": 0.2999, "step": 15843 }, { "epoch": 0.79, "learning_rate": 2.202181582791303e-06, "loss": 0.4221, "step": 15844 }, { "epoch": 0.79, "learning_rate": 2.2011692281167186e-06, "loss": 0.6181, "step": 15845 }, { "epoch": 0.79, "learning_rate": 2.2001570774103385e-06, "loss": 0.35, "step": 15846 }, { "epoch": 0.79, "learning_rate": 2.199145130698631e-06, "loss": 0.2797, "step": 15847 }, { "epoch": 0.79, "learning_rate": 2.198133388008068e-06, "loss": 0.2415, "step": 15848 }, { "epoch": 0.79, "learning_rate": 2.197121849365108e-06, "loss": 0.212, "step": 15849 }, { "epoch": 0.79, "learning_rate": 2.1961105147962057e-06, "loss": 0.4224, "step": 15850 }, { "epoch": 0.79, "learning_rate": 2.19509938432781e-06, "loss": 0.4101, "step": 15851 }, { "epoch": 0.79, "learning_rate": 2.1940884579863684e-06, "loss": 0.6134, "step": 15852 }, { "epoch": 0.79, "learning_rate": 2.1930777357983214e-06, "loss": 0.4574, "step": 15853 }, { "epoch": 0.79, "learning_rate": 2.1920672177900993e-06, "loss": 0.4369, "step": 15854 }, { "epoch": 0.79, "learning_rate": 2.1910569039881322e-06, "loss": 0.4157, "step": 15855 }, { "epoch": 0.79, "learning_rate": 2.190046794418843e-06, "loss": 0.272, "step": 15856 }, { "epoch": 0.79, "learning_rate": 2.1890368891086523e-06, "loss": 0.3361, "step": 15857 }, { "epoch": 0.79, "learning_rate": 2.188027188083971e-06, "loss": 0.386, "step": 15858 }, { "epoch": 0.79, "learning_rate": 2.187017691371207e-06, "loss": 0.4016, "step": 15859 }, { "epoch": 0.79, "learning_rate": 2.186008398996763e-06, "loss": 0.2329, "step": 15860 }, { "epoch": 0.79, "learning_rate": 2.1849993109870306e-06, "loss": 0.4158, "step": 15861 }, { "epoch": 0.79, "learning_rate": 2.183990427368409e-06, "loss": 0.3866, "step": 15862 }, { "epoch": 0.79, "learning_rate": 2.18298174816728e-06, "loss": 0.351, "step": 15863 }, { "epoch": 0.79, "learning_rate": 2.181973273410026e-06, "loss": 0.5803, "step": 15864 }, { "epoch": 0.79, "learning_rate": 2.1809650031230203e-06, "loss": 0.347, "step": 15865 }, { "epoch": 0.79, "learning_rate": 2.1799569373326348e-06, "loss": 0.4231, "step": 15866 }, { "epoch": 0.79, "learning_rate": 2.1789490760652323e-06, "loss": 0.3542, "step": 15867 }, { "epoch": 0.79, "learning_rate": 2.1779414193471738e-06, "loss": 0.3577, "step": 15868 }, { "epoch": 0.79, "learning_rate": 2.1769339672048117e-06, "loss": 0.311, "step": 15869 }, { "epoch": 0.79, "learning_rate": 2.175926719664493e-06, "loss": 0.4307, "step": 15870 }, { "epoch": 0.79, "learning_rate": 2.174919676752565e-06, "loss": 0.4822, "step": 15871 }, { "epoch": 0.79, "learning_rate": 2.1739128384953646e-06, "loss": 0.3229, "step": 15872 }, { "epoch": 0.79, "learning_rate": 2.172906204919224e-06, "loss": 0.4791, "step": 15873 }, { "epoch": 0.79, "learning_rate": 2.1718997760504666e-06, "loss": 0.3464, "step": 15874 }, { "epoch": 0.79, "learning_rate": 2.1708935519154205e-06, "loss": 0.3077, "step": 15875 }, { "epoch": 0.79, "learning_rate": 2.169887532540399e-06, "loss": 0.3322, "step": 15876 }, { "epoch": 0.79, "learning_rate": 2.168881717951714e-06, "loss": 0.4893, "step": 15877 }, { "epoch": 0.79, "learning_rate": 2.1678761081756706e-06, "loss": 0.1689, "step": 15878 }, { "epoch": 0.79, "learning_rate": 2.166870703238566e-06, "loss": 0.4512, "step": 15879 }, { "epoch": 0.79, "learning_rate": 2.165865503166702e-06, "loss": 0.6058, "step": 15880 }, { "epoch": 0.79, "learning_rate": 2.164860507986365e-06, "loss": 0.1789, "step": 15881 }, { "epoch": 0.79, "learning_rate": 2.1638557177238396e-06, "loss": 0.5785, "step": 15882 }, { "epoch": 0.79, "learning_rate": 2.1628511324054004e-06, "loss": 0.5265, "step": 15883 }, { "epoch": 0.79, "learning_rate": 2.1618467520573284e-06, "loss": 0.472, "step": 15884 }, { "epoch": 0.79, "learning_rate": 2.1608425767058885e-06, "loss": 0.3366, "step": 15885 }, { "epoch": 0.79, "learning_rate": 2.1598386063773436e-06, "loss": 0.38, "step": 15886 }, { "epoch": 0.79, "learning_rate": 2.158834841097951e-06, "loss": 0.3269, "step": 15887 }, { "epoch": 0.79, "learning_rate": 2.1578312808939606e-06, "loss": 0.2521, "step": 15888 }, { "epoch": 0.79, "learning_rate": 2.1568279257916245e-06, "loss": 0.3566, "step": 15889 }, { "epoch": 0.79, "learning_rate": 2.1558247758171813e-06, "loss": 0.3661, "step": 15890 }, { "epoch": 0.79, "learning_rate": 2.154821830996866e-06, "loss": 0.2756, "step": 15891 }, { "epoch": 0.79, "learning_rate": 2.1538190913569112e-06, "loss": 0.4321, "step": 15892 }, { "epoch": 0.79, "learning_rate": 2.1528165569235415e-06, "loss": 0.3652, "step": 15893 }, { "epoch": 0.79, "learning_rate": 2.151814227722977e-06, "loss": 0.4355, "step": 15894 }, { "epoch": 0.79, "learning_rate": 2.1508121037814313e-06, "loss": 0.3795, "step": 15895 }, { "epoch": 0.79, "learning_rate": 2.1498101851251154e-06, "loss": 0.2126, "step": 15896 }, { "epoch": 0.79, "learning_rate": 2.1488084717802283e-06, "loss": 0.181, "step": 15897 }, { "epoch": 0.79, "learning_rate": 2.147806963772976e-06, "loss": 0.4578, "step": 15898 }, { "epoch": 0.79, "learning_rate": 2.1468056611295486e-06, "loss": 0.6473, "step": 15899 }, { "epoch": 0.79, "learning_rate": 2.1458045638761315e-06, "loss": 0.3108, "step": 15900 }, { "epoch": 0.79, "learning_rate": 2.1448036720389097e-06, "loss": 0.4169, "step": 15901 }, { "epoch": 0.79, "learning_rate": 2.143802985644058e-06, "loss": 0.2733, "step": 15902 }, { "epoch": 0.79, "learning_rate": 2.14280250471775e-06, "loss": 0.314, "step": 15903 }, { "epoch": 0.79, "learning_rate": 2.141802229286153e-06, "loss": 0.3457, "step": 15904 }, { "epoch": 0.79, "learning_rate": 2.140802159375426e-06, "loss": 0.2931, "step": 15905 }, { "epoch": 0.79, "learning_rate": 2.1398022950117215e-06, "loss": 0.223, "step": 15906 }, { "epoch": 0.79, "learning_rate": 2.1388026362211967e-06, "loss": 0.2636, "step": 15907 }, { "epoch": 0.79, "learning_rate": 2.1378031830299916e-06, "loss": 0.3646, "step": 15908 }, { "epoch": 0.79, "learning_rate": 2.136803935464247e-06, "loss": 0.4718, "step": 15909 }, { "epoch": 0.79, "learning_rate": 2.1358048935500965e-06, "loss": 0.1692, "step": 15910 }, { "epoch": 0.79, "learning_rate": 2.1348060573136666e-06, "loss": 0.3347, "step": 15911 }, { "epoch": 0.79, "learning_rate": 2.133807426781085e-06, "loss": 0.4158, "step": 15912 }, { "epoch": 0.79, "learning_rate": 2.1328090019784663e-06, "loss": 0.5242, "step": 15913 }, { "epoch": 0.79, "learning_rate": 2.1318107829319256e-06, "loss": 0.2949, "step": 15914 }, { "epoch": 0.79, "learning_rate": 2.1308127696675675e-06, "loss": 0.34, "step": 15915 }, { "epoch": 0.79, "learning_rate": 2.129814962211495e-06, "loss": 0.3382, "step": 15916 }, { "epoch": 0.79, "learning_rate": 2.128817360589803e-06, "loss": 0.6617, "step": 15917 }, { "epoch": 0.79, "learning_rate": 2.1278199648285848e-06, "loss": 0.3232, "step": 15918 }, { "epoch": 0.79, "learning_rate": 2.126822774953925e-06, "loss": 0.37, "step": 15919 }, { "epoch": 0.79, "learning_rate": 2.1258257909919e-06, "loss": 0.5037, "step": 15920 }, { "epoch": 0.79, "learning_rate": 2.1248290129685912e-06, "loss": 0.3317, "step": 15921 }, { "epoch": 0.8, "learning_rate": 2.1238324409100643e-06, "loss": 0.6684, "step": 15922 }, { "epoch": 0.8, "learning_rate": 2.1228360748423836e-06, "loss": 0.3396, "step": 15923 }, { "epoch": 0.8, "learning_rate": 2.121839914791607e-06, "loss": 0.5428, "step": 15924 }, { "epoch": 0.8, "learning_rate": 2.1208439607837904e-06, "loss": 0.4972, "step": 15925 }, { "epoch": 0.8, "learning_rate": 2.1198482128449806e-06, "loss": 0.338, "step": 15926 }, { "epoch": 0.8, "learning_rate": 2.1188526710012203e-06, "loss": 0.4752, "step": 15927 }, { "epoch": 0.8, "learning_rate": 2.1178573352785447e-06, "loss": 0.471, "step": 15928 }, { "epoch": 0.8, "learning_rate": 2.1168622057029853e-06, "loss": 0.4075, "step": 15929 }, { "epoch": 0.8, "learning_rate": 2.115867282300571e-06, "loss": 0.5694, "step": 15930 }, { "epoch": 0.8, "learning_rate": 2.114872565097322e-06, "loss": 0.4898, "step": 15931 }, { "epoch": 0.8, "learning_rate": 2.113878054119255e-06, "loss": 0.1567, "step": 15932 }, { "epoch": 0.8, "learning_rate": 2.112883749392377e-06, "loss": 0.2283, "step": 15933 }, { "epoch": 0.8, "learning_rate": 2.111889650942692e-06, "loss": 0.3953, "step": 15934 }, { "epoch": 0.8, "learning_rate": 2.1108957587962042e-06, "loss": 0.3095, "step": 15935 }, { "epoch": 0.8, "learning_rate": 2.1099020729789056e-06, "loss": 0.102, "step": 15936 }, { "epoch": 0.8, "learning_rate": 2.108908593516783e-06, "loss": 0.276, "step": 15937 }, { "epoch": 0.8, "learning_rate": 2.1079153204358216e-06, "loss": 0.4069, "step": 15938 }, { "epoch": 0.8, "learning_rate": 2.106922253761997e-06, "loss": 0.3262, "step": 15939 }, { "epoch": 0.8, "learning_rate": 2.1059293935212842e-06, "loss": 0.372, "step": 15940 }, { "epoch": 0.8, "learning_rate": 2.104936739739648e-06, "loss": 0.4067, "step": 15941 }, { "epoch": 0.8, "learning_rate": 2.1039442924430508e-06, "loss": 0.3365, "step": 15942 }, { "epoch": 0.8, "learning_rate": 2.1029520516574465e-06, "loss": 0.3686, "step": 15943 }, { "epoch": 0.8, "learning_rate": 2.1019600174087896e-06, "loss": 0.3611, "step": 15944 }, { "epoch": 0.8, "learning_rate": 2.1009681897230248e-06, "loss": 0.4159, "step": 15945 }, { "epoch": 0.8, "learning_rate": 2.099976568626091e-06, "loss": 0.4268, "step": 15946 }, { "epoch": 0.8, "learning_rate": 2.09898515414392e-06, "loss": 0.5027, "step": 15947 }, { "epoch": 0.8, "learning_rate": 2.0979939463024457e-06, "loss": 0.4051, "step": 15948 }, { "epoch": 0.8, "learning_rate": 2.0970029451275907e-06, "loss": 0.3751, "step": 15949 }, { "epoch": 0.8, "learning_rate": 2.0960121506452737e-06, "loss": 0.2923, "step": 15950 }, { "epoch": 0.8, "learning_rate": 2.095021562881405e-06, "loss": 0.2912, "step": 15951 }, { "epoch": 0.8, "learning_rate": 2.094031181861891e-06, "loss": 0.2885, "step": 15952 }, { "epoch": 0.8, "learning_rate": 2.0930410076126393e-06, "loss": 0.3974, "step": 15953 }, { "epoch": 0.8, "learning_rate": 2.092051040159545e-06, "loss": 0.4453, "step": 15954 }, { "epoch": 0.8, "learning_rate": 2.091061279528497e-06, "loss": 0.2254, "step": 15955 }, { "epoch": 0.8, "learning_rate": 2.0900717257453805e-06, "loss": 0.402, "step": 15956 }, { "epoch": 0.8, "learning_rate": 2.0890823788360794e-06, "loss": 0.3209, "step": 15957 }, { "epoch": 0.8, "learning_rate": 2.0880932388264685e-06, "loss": 0.254, "step": 15958 }, { "epoch": 0.8, "learning_rate": 2.0871043057424156e-06, "loss": 0.3662, "step": 15959 }, { "epoch": 0.8, "learning_rate": 2.0861155796097855e-06, "loss": 0.4777, "step": 15960 }, { "epoch": 0.8, "learning_rate": 2.0851270604544337e-06, "loss": 0.4316, "step": 15961 }, { "epoch": 0.8, "learning_rate": 2.0841387483022214e-06, "loss": 0.3332, "step": 15962 }, { "epoch": 0.8, "learning_rate": 2.0831506431789904e-06, "loss": 0.3221, "step": 15963 }, { "epoch": 0.8, "learning_rate": 2.0821627451105865e-06, "loss": 0.4592, "step": 15964 }, { "epoch": 0.8, "learning_rate": 2.0811750541228448e-06, "loss": 0.3737, "step": 15965 }, { "epoch": 0.8, "learning_rate": 2.0801875702415984e-06, "loss": 0.3185, "step": 15966 }, { "epoch": 0.8, "learning_rate": 2.0792002934926725e-06, "loss": 0.3044, "step": 15967 }, { "epoch": 0.8, "learning_rate": 2.078213223901888e-06, "loss": 0.2381, "step": 15968 }, { "epoch": 0.8, "learning_rate": 2.077226361495063e-06, "loss": 0.3367, "step": 15969 }, { "epoch": 0.8, "learning_rate": 2.076239706298001e-06, "loss": 0.4726, "step": 15970 }, { "epoch": 0.8, "learning_rate": 2.075253258336514e-06, "loss": 0.3069, "step": 15971 }, { "epoch": 0.8, "learning_rate": 2.074267017636399e-06, "loss": 0.4328, "step": 15972 }, { "epoch": 0.8, "learning_rate": 2.073280984223449e-06, "loss": 0.1486, "step": 15973 }, { "epoch": 0.8, "learning_rate": 2.072295158123453e-06, "loss": 0.6163, "step": 15974 }, { "epoch": 0.8, "learning_rate": 2.071309539362191e-06, "loss": 0.6198, "step": 15975 }, { "epoch": 0.8, "learning_rate": 2.070324127965445e-06, "loss": 0.6192, "step": 15976 }, { "epoch": 0.8, "learning_rate": 2.0693389239589867e-06, "loss": 0.3085, "step": 15977 }, { "epoch": 0.8, "learning_rate": 2.0683539273685815e-06, "loss": 0.356, "step": 15978 }, { "epoch": 0.8, "learning_rate": 2.067369138219988e-06, "loss": 0.3408, "step": 15979 }, { "epoch": 0.8, "learning_rate": 2.0663845565389663e-06, "loss": 0.491, "step": 15980 }, { "epoch": 0.8, "learning_rate": 2.065400182351267e-06, "loss": 0.494, "step": 15981 }, { "epoch": 0.8, "learning_rate": 2.064416015682632e-06, "loss": 0.4331, "step": 15982 }, { "epoch": 0.8, "learning_rate": 2.0634320565588027e-06, "loss": 0.5054, "step": 15983 }, { "epoch": 0.8, "learning_rate": 2.062448305005511e-06, "loss": 0.2307, "step": 15984 }, { "epoch": 0.8, "learning_rate": 2.0614647610484893e-06, "loss": 0.2187, "step": 15985 }, { "epoch": 0.8, "learning_rate": 2.060481424713461e-06, "loss": 0.2825, "step": 15986 }, { "epoch": 0.8, "learning_rate": 2.05949829602614e-06, "loss": 0.4338, "step": 15987 }, { "epoch": 0.8, "learning_rate": 2.058515375012242e-06, "loss": 0.2752, "step": 15988 }, { "epoch": 0.8, "learning_rate": 2.0575326616974723e-06, "loss": 0.4219, "step": 15989 }, { "epoch": 0.8, "learning_rate": 2.056550156107534e-06, "loss": 0.1967, "step": 15990 }, { "epoch": 0.8, "learning_rate": 2.0555678582681214e-06, "loss": 0.45, "step": 15991 }, { "epoch": 0.8, "learning_rate": 2.0545857682049263e-06, "loss": 0.4823, "step": 15992 }, { "epoch": 0.8, "learning_rate": 2.0536038859436313e-06, "loss": 0.1873, "step": 15993 }, { "epoch": 0.8, "learning_rate": 2.0526222115099213e-06, "loss": 0.3875, "step": 15994 }, { "epoch": 0.8, "learning_rate": 2.0516407449294683e-06, "loss": 0.3032, "step": 15995 }, { "epoch": 0.8, "learning_rate": 2.0506594862279407e-06, "loss": 0.3742, "step": 15996 }, { "epoch": 0.8, "learning_rate": 2.049678435431003e-06, "loss": 0.4951, "step": 15997 }, { "epoch": 0.8, "learning_rate": 2.0486975925643104e-06, "loss": 0.2765, "step": 15998 }, { "epoch": 0.8, "learning_rate": 2.0477169576535193e-06, "loss": 0.4518, "step": 15999 }, { "epoch": 0.8, "learning_rate": 2.046736530724277e-06, "loss": 0.1646, "step": 16000 }, { "epoch": 0.8, "learning_rate": 2.0457563118022228e-06, "loss": 0.311, "step": 16001 }, { "epoch": 0.8, "learning_rate": 2.044776300912993e-06, "loss": 0.598, "step": 16002 }, { "epoch": 0.8, "learning_rate": 2.04379649808222e-06, "loss": 0.515, "step": 16003 }, { "epoch": 0.8, "learning_rate": 2.0428169033355307e-06, "loss": 0.2555, "step": 16004 }, { "epoch": 0.8, "learning_rate": 2.041837516698543e-06, "loss": 0.3303, "step": 16005 }, { "epoch": 0.8, "learning_rate": 2.0408583381968704e-06, "loss": 0.2694, "step": 16006 }, { "epoch": 0.8, "learning_rate": 2.039879367856121e-06, "loss": 0.1566, "step": 16007 }, { "epoch": 0.8, "learning_rate": 2.0389006057019045e-06, "loss": 0.3079, "step": 16008 }, { "epoch": 0.8, "learning_rate": 2.0379220517598153e-06, "loss": 0.394, "step": 16009 }, { "epoch": 0.8, "learning_rate": 2.0369437060554466e-06, "loss": 0.5321, "step": 16010 }, { "epoch": 0.8, "learning_rate": 2.035965568614384e-06, "loss": 0.2621, "step": 16011 }, { "epoch": 0.8, "learning_rate": 2.0349876394622125e-06, "loss": 0.2982, "step": 16012 }, { "epoch": 0.8, "learning_rate": 2.0340099186245065e-06, "loss": 0.4425, "step": 16013 }, { "epoch": 0.8, "learning_rate": 2.033032406126837e-06, "loss": 0.4748, "step": 16014 }, { "epoch": 0.8, "learning_rate": 2.032055101994771e-06, "loss": 0.3001, "step": 16015 }, { "epoch": 0.8, "learning_rate": 2.031078006253865e-06, "loss": 0.4888, "step": 16016 }, { "epoch": 0.8, "learning_rate": 2.0301011189296782e-06, "loss": 0.5327, "step": 16017 }, { "epoch": 0.8, "learning_rate": 2.029124440047758e-06, "loss": 0.5901, "step": 16018 }, { "epoch": 0.8, "learning_rate": 2.028147969633648e-06, "loss": 0.3005, "step": 16019 }, { "epoch": 0.8, "learning_rate": 2.0271717077128848e-06, "loss": 0.2701, "step": 16020 }, { "epoch": 0.8, "learning_rate": 2.026195654311005e-06, "loss": 0.2726, "step": 16021 }, { "epoch": 0.8, "learning_rate": 2.0252198094535348e-06, "loss": 0.0984, "step": 16022 }, { "epoch": 0.8, "learning_rate": 2.0242441731659944e-06, "loss": 0.2688, "step": 16023 }, { "epoch": 0.8, "learning_rate": 2.023268745473902e-06, "loss": 0.1673, "step": 16024 }, { "epoch": 0.8, "learning_rate": 2.0222935264027656e-06, "loss": 0.2843, "step": 16025 }, { "epoch": 0.8, "learning_rate": 2.0213185159780947e-06, "loss": 0.4017, "step": 16026 }, { "epoch": 0.8, "learning_rate": 2.020343714225389e-06, "loss": 0.2087, "step": 16027 }, { "epoch": 0.8, "learning_rate": 2.0193691211701416e-06, "loss": 0.3348, "step": 16028 }, { "epoch": 0.8, "learning_rate": 2.018394736837841e-06, "loss": 0.1477, "step": 16029 }, { "epoch": 0.8, "learning_rate": 2.017420561253971e-06, "loss": 0.3578, "step": 16030 }, { "epoch": 0.8, "learning_rate": 2.0164465944440127e-06, "loss": 0.3654, "step": 16031 }, { "epoch": 0.8, "learning_rate": 2.0154728364334362e-06, "loss": 0.2537, "step": 16032 }, { "epoch": 0.8, "learning_rate": 2.014499287247711e-06, "loss": 0.2512, "step": 16033 }, { "epoch": 0.8, "learning_rate": 2.0135259469122957e-06, "loss": 0.3719, "step": 16034 }, { "epoch": 0.8, "learning_rate": 2.0125528154526506e-06, "loss": 0.2904, "step": 16035 }, { "epoch": 0.8, "learning_rate": 2.0115798928942252e-06, "loss": 0.3633, "step": 16036 }, { "epoch": 0.8, "learning_rate": 2.0106071792624648e-06, "loss": 0.4446, "step": 16037 }, { "epoch": 0.8, "learning_rate": 2.0096346745828086e-06, "loss": 0.362, "step": 16038 }, { "epoch": 0.8, "learning_rate": 2.0086623788806937e-06, "loss": 0.1405, "step": 16039 }, { "epoch": 0.8, "learning_rate": 2.0076902921815456e-06, "loss": 0.4175, "step": 16040 }, { "epoch": 0.8, "learning_rate": 2.006718414510791e-06, "loss": 0.3672, "step": 16041 }, { "epoch": 0.8, "learning_rate": 2.0057467458938474e-06, "loss": 0.4446, "step": 16042 }, { "epoch": 0.8, "learning_rate": 2.0047752863561233e-06, "loss": 0.3963, "step": 16043 }, { "epoch": 0.8, "learning_rate": 2.0038040359230327e-06, "loss": 0.1257, "step": 16044 }, { "epoch": 0.8, "learning_rate": 2.002832994619974e-06, "loss": 0.2509, "step": 16045 }, { "epoch": 0.8, "learning_rate": 2.0018621624723445e-06, "loss": 0.4312, "step": 16046 }, { "epoch": 0.8, "learning_rate": 2.000891539505535e-06, "loss": 0.3446, "step": 16047 }, { "epoch": 0.8, "learning_rate": 1.999921125744927e-06, "loss": 0.3474, "step": 16048 }, { "epoch": 0.8, "learning_rate": 1.9989509212159066e-06, "loss": 0.1965, "step": 16049 }, { "epoch": 0.8, "learning_rate": 1.9979809259438453e-06, "loss": 0.293, "step": 16050 }, { "epoch": 0.8, "learning_rate": 1.9970111399541114e-06, "loss": 0.3617, "step": 16051 }, { "epoch": 0.8, "learning_rate": 1.9960415632720675e-06, "loss": 0.2955, "step": 16052 }, { "epoch": 0.8, "learning_rate": 1.9950721959230755e-06, "loss": 0.6474, "step": 16053 }, { "epoch": 0.8, "learning_rate": 1.9941030379324854e-06, "loss": 0.3721, "step": 16054 }, { "epoch": 0.8, "learning_rate": 1.9931340893256447e-06, "loss": 0.5233, "step": 16055 }, { "epoch": 0.8, "learning_rate": 1.992165350127895e-06, "loss": 0.5716, "step": 16056 }, { "epoch": 0.8, "learning_rate": 1.9911968203645693e-06, "loss": 0.3419, "step": 16057 }, { "epoch": 0.8, "learning_rate": 1.9902285000610034e-06, "loss": 0.2077, "step": 16058 }, { "epoch": 0.8, "learning_rate": 1.989260389242521e-06, "loss": 0.2824, "step": 16059 }, { "epoch": 0.8, "learning_rate": 1.98829248793444e-06, "loss": 0.1403, "step": 16060 }, { "epoch": 0.8, "learning_rate": 1.9873247961620765e-06, "loss": 0.1825, "step": 16061 }, { "epoch": 0.8, "learning_rate": 1.9863573139507375e-06, "loss": 0.4472, "step": 16062 }, { "epoch": 0.8, "learning_rate": 1.9853900413257276e-06, "loss": 0.3195, "step": 16063 }, { "epoch": 0.8, "learning_rate": 1.9844229783123435e-06, "loss": 0.3049, "step": 16064 }, { "epoch": 0.8, "learning_rate": 1.9834561249358776e-06, "loss": 0.5135, "step": 16065 }, { "epoch": 0.8, "learning_rate": 1.9824894812216142e-06, "loss": 0.2061, "step": 16066 }, { "epoch": 0.8, "learning_rate": 1.981523047194841e-06, "loss": 0.2389, "step": 16067 }, { "epoch": 0.8, "learning_rate": 1.980556822880829e-06, "loss": 0.2715, "step": 16068 }, { "epoch": 0.8, "learning_rate": 1.979590808304851e-06, "loss": 0.2682, "step": 16069 }, { "epoch": 0.8, "learning_rate": 1.9786250034921695e-06, "loss": 0.2534, "step": 16070 }, { "epoch": 0.8, "learning_rate": 1.9776594084680436e-06, "loss": 0.4433, "step": 16071 }, { "epoch": 0.8, "learning_rate": 1.97669402325773e-06, "loss": 0.5385, "step": 16072 }, { "epoch": 0.8, "learning_rate": 1.9757288478864756e-06, "loss": 0.5105, "step": 16073 }, { "epoch": 0.8, "learning_rate": 1.9747638823795235e-06, "loss": 0.3288, "step": 16074 }, { "epoch": 0.8, "learning_rate": 1.9737991267621082e-06, "loss": 0.2532, "step": 16075 }, { "epoch": 0.8, "learning_rate": 1.972834581059467e-06, "loss": 0.5634, "step": 16076 }, { "epoch": 0.8, "learning_rate": 1.971870245296824e-06, "loss": 0.4599, "step": 16077 }, { "epoch": 0.8, "learning_rate": 1.9709061194993995e-06, "loss": 0.3089, "step": 16078 }, { "epoch": 0.8, "learning_rate": 1.9699422036924086e-06, "loss": 0.2775, "step": 16079 }, { "epoch": 0.8, "learning_rate": 1.9689784979010596e-06, "loss": 0.2624, "step": 16080 }, { "epoch": 0.8, "learning_rate": 1.968015002150562e-06, "loss": 0.3231, "step": 16081 }, { "epoch": 0.8, "learning_rate": 1.967051716466112e-06, "loss": 0.3554, "step": 16082 }, { "epoch": 0.8, "learning_rate": 1.966088640872903e-06, "loss": 0.3691, "step": 16083 }, { "epoch": 0.8, "learning_rate": 1.965125775396124e-06, "loss": 0.2884, "step": 16084 }, { "epoch": 0.8, "learning_rate": 1.9641631200609557e-06, "loss": 0.5295, "step": 16085 }, { "epoch": 0.8, "learning_rate": 1.963200674892576e-06, "loss": 0.5939, "step": 16086 }, { "epoch": 0.8, "learning_rate": 1.9622384399161553e-06, "loss": 0.3909, "step": 16087 }, { "epoch": 0.8, "learning_rate": 1.961276415156862e-06, "loss": 0.3042, "step": 16088 }, { "epoch": 0.8, "learning_rate": 1.960314600639852e-06, "loss": 0.282, "step": 16089 }, { "epoch": 0.8, "learning_rate": 1.959352996390287e-06, "loss": 0.2707, "step": 16090 }, { "epoch": 0.8, "learning_rate": 1.9583916024333116e-06, "loss": 0.3515, "step": 16091 }, { "epoch": 0.8, "learning_rate": 1.9574304187940727e-06, "loss": 0.4355, "step": 16092 }, { "epoch": 0.8, "learning_rate": 1.956469445497704e-06, "loss": 0.4526, "step": 16093 }, { "epoch": 0.8, "learning_rate": 1.9555086825693447e-06, "loss": 0.3585, "step": 16094 }, { "epoch": 0.8, "learning_rate": 1.9545481300341185e-06, "loss": 0.5213, "step": 16095 }, { "epoch": 0.8, "learning_rate": 1.9535877879171496e-06, "loss": 0.2359, "step": 16096 }, { "epoch": 0.8, "learning_rate": 1.952627656243552e-06, "loss": 0.403, "step": 16097 }, { "epoch": 0.8, "learning_rate": 1.951667735038436e-06, "loss": 0.423, "step": 16098 }, { "epoch": 0.8, "learning_rate": 1.9507080243269116e-06, "loss": 0.3233, "step": 16099 }, { "epoch": 0.8, "learning_rate": 1.9497485241340764e-06, "loss": 0.5242, "step": 16100 }, { "epoch": 0.8, "learning_rate": 1.9487892344850233e-06, "loss": 0.6711, "step": 16101 }, { "epoch": 0.8, "learning_rate": 1.947830155404844e-06, "loss": 0.2363, "step": 16102 }, { "epoch": 0.8, "learning_rate": 1.946871286918618e-06, "loss": 0.5251, "step": 16103 }, { "epoch": 0.8, "learning_rate": 1.9459126290514274e-06, "loss": 0.4052, "step": 16104 }, { "epoch": 0.8, "learning_rate": 1.9449541818283433e-06, "loss": 0.4873, "step": 16105 }, { "epoch": 0.8, "learning_rate": 1.9439959452744327e-06, "loss": 0.4699, "step": 16106 }, { "epoch": 0.8, "learning_rate": 1.9430379194147542e-06, "loss": 0.5247, "step": 16107 }, { "epoch": 0.8, "learning_rate": 1.9420801042743674e-06, "loss": 0.4233, "step": 16108 }, { "epoch": 0.8, "learning_rate": 1.941122499878323e-06, "loss": 0.5039, "step": 16109 }, { "epoch": 0.8, "learning_rate": 1.940165106251665e-06, "loss": 0.2871, "step": 16110 }, { "epoch": 0.8, "learning_rate": 1.9392079234194304e-06, "loss": 0.4849, "step": 16111 }, { "epoch": 0.8, "learning_rate": 1.938250951406655e-06, "loss": 0.1698, "step": 16112 }, { "epoch": 0.8, "learning_rate": 1.937294190238368e-06, "loss": 0.2643, "step": 16113 }, { "epoch": 0.8, "learning_rate": 1.9363376399395906e-06, "loss": 0.4113, "step": 16114 }, { "epoch": 0.8, "learning_rate": 1.9353813005353406e-06, "loss": 0.323, "step": 16115 }, { "epoch": 0.8, "learning_rate": 1.934425172050628e-06, "loss": 0.2393, "step": 16116 }, { "epoch": 0.8, "learning_rate": 1.9334692545104626e-06, "loss": 0.2478, "step": 16117 }, { "epoch": 0.8, "learning_rate": 1.9325135479398447e-06, "loss": 0.3394, "step": 16118 }, { "epoch": 0.8, "learning_rate": 1.9315580523637678e-06, "loss": 0.2645, "step": 16119 }, { "epoch": 0.8, "learning_rate": 1.9306027678072226e-06, "loss": 0.3758, "step": 16120 }, { "epoch": 0.8, "learning_rate": 1.9296476942951903e-06, "loss": 0.1681, "step": 16121 }, { "epoch": 0.81, "learning_rate": 1.9286928318526545e-06, "loss": 0.5096, "step": 16122 }, { "epoch": 0.81, "learning_rate": 1.927738180504587e-06, "loss": 0.2777, "step": 16123 }, { "epoch": 0.81, "learning_rate": 1.926783740275954e-06, "loss": 0.432, "step": 16124 }, { "epoch": 0.81, "learning_rate": 1.9258295111917156e-06, "loss": 0.6058, "step": 16125 }, { "epoch": 0.81, "learning_rate": 1.924875493276833e-06, "loss": 0.3492, "step": 16126 }, { "epoch": 0.81, "learning_rate": 1.923921686556257e-06, "loss": 0.4578, "step": 16127 }, { "epoch": 0.81, "learning_rate": 1.922968091054931e-06, "loss": 0.3188, "step": 16128 }, { "epoch": 0.81, "learning_rate": 1.9220147067977947e-06, "loss": 0.3589, "step": 16129 }, { "epoch": 0.81, "learning_rate": 1.9210615338097815e-06, "loss": 0.4423, "step": 16130 }, { "epoch": 0.81, "learning_rate": 1.920108572115825e-06, "loss": 0.2675, "step": 16131 }, { "epoch": 0.81, "learning_rate": 1.919155821740846e-06, "loss": 0.3504, "step": 16132 }, { "epoch": 0.81, "learning_rate": 1.918203282709763e-06, "loss": 0.3821, "step": 16133 }, { "epoch": 0.81, "learning_rate": 1.917250955047487e-06, "loss": 0.6351, "step": 16134 }, { "epoch": 0.81, "learning_rate": 1.916298838778926e-06, "loss": 0.4942, "step": 16135 }, { "epoch": 0.81, "learning_rate": 1.915346933928981e-06, "loss": 0.5265, "step": 16136 }, { "epoch": 0.81, "learning_rate": 1.9143952405225485e-06, "loss": 0.309, "step": 16137 }, { "epoch": 0.81, "learning_rate": 1.913443758584518e-06, "loss": 0.3814, "step": 16138 }, { "epoch": 0.81, "learning_rate": 1.912492488139772e-06, "loss": 0.4224, "step": 16139 }, { "epoch": 0.81, "learning_rate": 1.911541429213194e-06, "loss": 0.5276, "step": 16140 }, { "epoch": 0.81, "learning_rate": 1.9105905818296565e-06, "loss": 0.2899, "step": 16141 }, { "epoch": 0.81, "learning_rate": 1.909639946014028e-06, "loss": 0.5438, "step": 16142 }, { "epoch": 0.81, "learning_rate": 1.9086895217911695e-06, "loss": 0.5416, "step": 16143 }, { "epoch": 0.81, "learning_rate": 1.9077393091859366e-06, "loss": 0.2189, "step": 16144 }, { "epoch": 0.81, "learning_rate": 1.9067893082231848e-06, "loss": 0.5953, "step": 16145 }, { "epoch": 0.81, "learning_rate": 1.905839518927759e-06, "loss": 0.4294, "step": 16146 }, { "epoch": 0.81, "learning_rate": 1.9048899413244992e-06, "loss": 0.4372, "step": 16147 }, { "epoch": 0.81, "learning_rate": 1.903940575438238e-06, "loss": 0.3935, "step": 16148 }, { "epoch": 0.81, "learning_rate": 1.9029914212938083e-06, "loss": 0.4174, "step": 16149 }, { "epoch": 0.81, "learning_rate": 1.902042478916034e-06, "loss": 0.3982, "step": 16150 }, { "epoch": 0.81, "learning_rate": 1.901093748329732e-06, "loss": 0.3501, "step": 16151 }, { "epoch": 0.81, "learning_rate": 1.9001452295597156e-06, "loss": 0.4893, "step": 16152 }, { "epoch": 0.81, "learning_rate": 1.8991969226307893e-06, "loss": 0.3586, "step": 16153 }, { "epoch": 0.81, "learning_rate": 1.8982488275677601e-06, "loss": 0.2919, "step": 16154 }, { "epoch": 0.81, "learning_rate": 1.8973009443954205e-06, "loss": 0.5065, "step": 16155 }, { "epoch": 0.81, "learning_rate": 1.8963532731385625e-06, "loss": 0.3356, "step": 16156 }, { "epoch": 0.81, "learning_rate": 1.895405813821971e-06, "loss": 0.3292, "step": 16157 }, { "epoch": 0.81, "learning_rate": 1.8944585664704252e-06, "loss": 0.4181, "step": 16158 }, { "epoch": 0.81, "learning_rate": 1.8935115311086993e-06, "loss": 0.2985, "step": 16159 }, { "epoch": 0.81, "learning_rate": 1.8925647077615617e-06, "loss": 0.3603, "step": 16160 }, { "epoch": 0.81, "learning_rate": 1.8916180964537723e-06, "loss": 0.2903, "step": 16161 }, { "epoch": 0.81, "learning_rate": 1.8906716972100948e-06, "loss": 0.4979, "step": 16162 }, { "epoch": 0.81, "learning_rate": 1.889725510055278e-06, "loss": 0.1701, "step": 16163 }, { "epoch": 0.81, "learning_rate": 1.888779535014067e-06, "loss": 0.2676, "step": 16164 }, { "epoch": 0.81, "learning_rate": 1.8878337721112038e-06, "loss": 0.1608, "step": 16165 }, { "epoch": 0.81, "learning_rate": 1.8868882213714236e-06, "loss": 0.2401, "step": 16166 }, { "epoch": 0.81, "learning_rate": 1.8859428828194526e-06, "loss": 0.3372, "step": 16167 }, { "epoch": 0.81, "learning_rate": 1.8849977564800204e-06, "loss": 0.3109, "step": 16168 }, { "epoch": 0.81, "learning_rate": 1.8840528423778437e-06, "loss": 0.4405, "step": 16169 }, { "epoch": 0.81, "learning_rate": 1.883108140537635e-06, "loss": 0.3587, "step": 16170 }, { "epoch": 0.81, "learning_rate": 1.882163650984099e-06, "loss": 0.3117, "step": 16171 }, { "epoch": 0.81, "learning_rate": 1.8812193737419426e-06, "loss": 0.3003, "step": 16172 }, { "epoch": 0.81, "learning_rate": 1.8802753088358593e-06, "loss": 0.2932, "step": 16173 }, { "epoch": 0.81, "learning_rate": 1.8793314562905417e-06, "loss": 0.3783, "step": 16174 }, { "epoch": 0.81, "learning_rate": 1.8783878161306723e-06, "loss": 0.2292, "step": 16175 }, { "epoch": 0.81, "learning_rate": 1.87744438838093e-06, "loss": 0.4451, "step": 16176 }, { "epoch": 0.81, "learning_rate": 1.8765011730659943e-06, "loss": 0.1201, "step": 16177 }, { "epoch": 0.81, "learning_rate": 1.8755581702105309e-06, "loss": 0.2753, "step": 16178 }, { "epoch": 0.81, "learning_rate": 1.8746153798392009e-06, "loss": 0.3161, "step": 16179 }, { "epoch": 0.81, "learning_rate": 1.8736728019766615e-06, "loss": 0.3487, "step": 16180 }, { "epoch": 0.81, "learning_rate": 1.8727304366475696e-06, "loss": 0.2962, "step": 16181 }, { "epoch": 0.81, "learning_rate": 1.8717882838765677e-06, "loss": 0.3049, "step": 16182 }, { "epoch": 0.81, "learning_rate": 1.8708463436882973e-06, "loss": 0.4022, "step": 16183 }, { "epoch": 0.81, "learning_rate": 1.869904616107393e-06, "loss": 0.2962, "step": 16184 }, { "epoch": 0.81, "learning_rate": 1.8689631011584853e-06, "loss": 0.3365, "step": 16185 }, { "epoch": 0.81, "learning_rate": 1.8680217988661975e-06, "loss": 0.2279, "step": 16186 }, { "epoch": 0.81, "learning_rate": 1.867080709255149e-06, "loss": 0.3515, "step": 16187 }, { "epoch": 0.81, "learning_rate": 1.8661398323499537e-06, "loss": 0.3959, "step": 16188 }, { "epoch": 0.81, "learning_rate": 1.8651991681752135e-06, "loss": 0.4836, "step": 16189 }, { "epoch": 0.81, "learning_rate": 1.8642587167555371e-06, "loss": 0.2823, "step": 16190 }, { "epoch": 0.81, "learning_rate": 1.8633184781155178e-06, "loss": 0.4022, "step": 16191 }, { "epoch": 0.81, "learning_rate": 1.8623784522797472e-06, "loss": 0.4281, "step": 16192 }, { "epoch": 0.81, "learning_rate": 1.861438639272809e-06, "loss": 0.3571, "step": 16193 }, { "epoch": 0.81, "learning_rate": 1.8604990391192824e-06, "loss": 0.2593, "step": 16194 }, { "epoch": 0.81, "learning_rate": 1.8595596518437441e-06, "loss": 0.3466, "step": 16195 }, { "epoch": 0.81, "learning_rate": 1.8586204774707618e-06, "loss": 0.466, "step": 16196 }, { "epoch": 0.81, "learning_rate": 1.8576815160248963e-06, "loss": 0.4132, "step": 16197 }, { "epoch": 0.81, "learning_rate": 1.8567427675307049e-06, "loss": 0.3023, "step": 16198 }, { "epoch": 0.81, "learning_rate": 1.8558042320127423e-06, "loss": 0.4649, "step": 16199 }, { "epoch": 0.81, "learning_rate": 1.8548659094955534e-06, "loss": 0.494, "step": 16200 }, { "epoch": 0.81, "learning_rate": 1.8539278000036787e-06, "loss": 0.4068, "step": 16201 }, { "epoch": 0.81, "learning_rate": 1.8529899035616527e-06, "loss": 0.3108, "step": 16202 }, { "epoch": 0.81, "learning_rate": 1.8520522201940029e-06, "loss": 0.7383, "step": 16203 }, { "epoch": 0.81, "learning_rate": 1.851114749925258e-06, "loss": 0.305, "step": 16204 }, { "epoch": 0.81, "learning_rate": 1.8501774927799333e-06, "loss": 0.581, "step": 16205 }, { "epoch": 0.81, "learning_rate": 1.8492404487825422e-06, "loss": 0.3918, "step": 16206 }, { "epoch": 0.81, "learning_rate": 1.848303617957591e-06, "loss": 0.1707, "step": 16207 }, { "epoch": 0.81, "learning_rate": 1.8473670003295829e-06, "loss": 0.663, "step": 16208 }, { "epoch": 0.81, "learning_rate": 1.8464305959230122e-06, "loss": 0.2734, "step": 16209 }, { "epoch": 0.81, "learning_rate": 1.8454944047623703e-06, "loss": 0.3713, "step": 16210 }, { "epoch": 0.81, "learning_rate": 1.8445584268721405e-06, "loss": 0.1965, "step": 16211 }, { "epoch": 0.81, "learning_rate": 1.8436226622768028e-06, "loss": 0.228, "step": 16212 }, { "epoch": 0.81, "learning_rate": 1.8426871110008327e-06, "loss": 0.3839, "step": 16213 }, { "epoch": 0.81, "learning_rate": 1.8417517730686973e-06, "loss": 0.4051, "step": 16214 }, { "epoch": 0.81, "learning_rate": 1.8408166485048584e-06, "loss": 0.4, "step": 16215 }, { "epoch": 0.81, "learning_rate": 1.839881737333774e-06, "loss": 0.093, "step": 16216 }, { "epoch": 0.81, "learning_rate": 1.8389470395798925e-06, "loss": 0.3454, "step": 16217 }, { "epoch": 0.81, "learning_rate": 1.8380125552676653e-06, "loss": 0.5375, "step": 16218 }, { "epoch": 0.81, "learning_rate": 1.8370782844215284e-06, "loss": 0.4391, "step": 16219 }, { "epoch": 0.81, "learning_rate": 1.836144227065918e-06, "loss": 0.6318, "step": 16220 }, { "epoch": 0.81, "learning_rate": 1.8352103832252598e-06, "loss": 0.4801, "step": 16221 }, { "epoch": 0.81, "learning_rate": 1.8342767529239824e-06, "loss": 0.4474, "step": 16222 }, { "epoch": 0.81, "learning_rate": 1.8333433361865016e-06, "loss": 0.424, "step": 16223 }, { "epoch": 0.81, "learning_rate": 1.832410133037229e-06, "loss": 0.4252, "step": 16224 }, { "epoch": 0.81, "learning_rate": 1.831477143500573e-06, "loss": 0.3359, "step": 16225 }, { "epoch": 0.81, "learning_rate": 1.8305443676009294e-06, "loss": 0.2766, "step": 16226 }, { "epoch": 0.81, "learning_rate": 1.829611805362701e-06, "loss": 0.2764, "step": 16227 }, { "epoch": 0.81, "learning_rate": 1.828679456810275e-06, "loss": 0.4478, "step": 16228 }, { "epoch": 0.81, "learning_rate": 1.8277473219680343e-06, "loss": 0.35, "step": 16229 }, { "epoch": 0.81, "learning_rate": 1.82681540086036e-06, "loss": 0.3559, "step": 16230 }, { "epoch": 0.81, "learning_rate": 1.8258836935116231e-06, "loss": 0.4227, "step": 16231 }, { "epoch": 0.81, "learning_rate": 1.8249521999461917e-06, "loss": 0.1981, "step": 16232 }, { "epoch": 0.81, "learning_rate": 1.8240209201884285e-06, "loss": 0.1871, "step": 16233 }, { "epoch": 0.81, "learning_rate": 1.8230898542626874e-06, "loss": 0.3571, "step": 16234 }, { "epoch": 0.81, "learning_rate": 1.8221590021933232e-06, "loss": 0.4985, "step": 16235 }, { "epoch": 0.81, "learning_rate": 1.82122836400468e-06, "loss": 0.3367, "step": 16236 }, { "epoch": 0.81, "learning_rate": 1.8202979397210963e-06, "loss": 0.4198, "step": 16237 }, { "epoch": 0.81, "learning_rate": 1.8193677293669065e-06, "loss": 0.2535, "step": 16238 }, { "epoch": 0.81, "learning_rate": 1.8184377329664382e-06, "loss": 0.6249, "step": 16239 }, { "epoch": 0.81, "learning_rate": 1.8175079505440141e-06, "loss": 0.3241, "step": 16240 }, { "epoch": 0.81, "learning_rate": 1.8165783821239536e-06, "loss": 0.1962, "step": 16241 }, { "epoch": 0.81, "learning_rate": 1.815649027730567e-06, "loss": 0.4535, "step": 16242 }, { "epoch": 0.81, "learning_rate": 1.8147198873881622e-06, "loss": 0.2534, "step": 16243 }, { "epoch": 0.81, "learning_rate": 1.8137909611210336e-06, "loss": 0.3191, "step": 16244 }, { "epoch": 0.81, "learning_rate": 1.8128622489534841e-06, "loss": 0.4755, "step": 16245 }, { "epoch": 0.81, "learning_rate": 1.811933750909799e-06, "loss": 0.317, "step": 16246 }, { "epoch": 0.81, "learning_rate": 1.8110054670142618e-06, "loss": 0.4565, "step": 16247 }, { "epoch": 0.81, "learning_rate": 1.8100773972911522e-06, "loss": 0.387, "step": 16248 }, { "epoch": 0.81, "learning_rate": 1.8091495417647376e-06, "loss": 0.1438, "step": 16249 }, { "epoch": 0.81, "learning_rate": 1.8082219004592915e-06, "loss": 0.5928, "step": 16250 }, { "epoch": 0.81, "learning_rate": 1.8072944733990728e-06, "loss": 0.1702, "step": 16251 }, { "epoch": 0.81, "learning_rate": 1.8063672606083361e-06, "loss": 0.6351, "step": 16252 }, { "epoch": 0.81, "learning_rate": 1.8054402621113288e-06, "loss": 0.3737, "step": 16253 }, { "epoch": 0.81, "learning_rate": 1.8045134779323015e-06, "loss": 0.3915, "step": 16254 }, { "epoch": 0.81, "learning_rate": 1.8035869080954904e-06, "loss": 0.3751, "step": 16255 }, { "epoch": 0.81, "learning_rate": 1.8026605526251285e-06, "loss": 0.3599, "step": 16256 }, { "epoch": 0.81, "learning_rate": 1.8017344115454427e-06, "loss": 0.1115, "step": 16257 }, { "epoch": 0.81, "learning_rate": 1.8008084848806562e-06, "loss": 0.2686, "step": 16258 }, { "epoch": 0.81, "learning_rate": 1.799882772654984e-06, "loss": 0.5175, "step": 16259 }, { "epoch": 0.81, "learning_rate": 1.7989572748926388e-06, "loss": 0.4631, "step": 16260 }, { "epoch": 0.81, "learning_rate": 1.7980319916178236e-06, "loss": 0.267, "step": 16261 }, { "epoch": 0.81, "learning_rate": 1.7971069228547367e-06, "loss": 0.5261, "step": 16262 }, { "epoch": 0.81, "learning_rate": 1.7961820686275767e-06, "loss": 0.414, "step": 16263 }, { "epoch": 0.81, "learning_rate": 1.7952574289605306e-06, "loss": 0.2462, "step": 16264 }, { "epoch": 0.81, "learning_rate": 1.7943330038777795e-06, "loss": 0.204, "step": 16265 }, { "epoch": 0.81, "learning_rate": 1.7934087934035005e-06, "loss": 0.313, "step": 16266 }, { "epoch": 0.81, "learning_rate": 1.7924847975618643e-06, "loss": 0.4851, "step": 16267 }, { "epoch": 0.81, "learning_rate": 1.7915610163770404e-06, "loss": 0.4239, "step": 16268 }, { "epoch": 0.81, "learning_rate": 1.790637449873187e-06, "loss": 0.322, "step": 16269 }, { "epoch": 0.81, "learning_rate": 1.7897140980744587e-06, "loss": 0.4469, "step": 16270 }, { "epoch": 0.81, "learning_rate": 1.7887909610050048e-06, "loss": 0.3138, "step": 16271 }, { "epoch": 0.81, "learning_rate": 1.7878680386889658e-06, "loss": 0.3876, "step": 16272 }, { "epoch": 0.81, "learning_rate": 1.7869453311504847e-06, "loss": 0.3946, "step": 16273 }, { "epoch": 0.81, "learning_rate": 1.7860228384136923e-06, "loss": 0.4914, "step": 16274 }, { "epoch": 0.81, "learning_rate": 1.7851005605027138e-06, "loss": 0.4168, "step": 16275 }, { "epoch": 0.81, "learning_rate": 1.7841784974416676e-06, "loss": 0.2465, "step": 16276 }, { "epoch": 0.81, "learning_rate": 1.7832566492546743e-06, "loss": 0.3756, "step": 16277 }, { "epoch": 0.81, "learning_rate": 1.7823350159658415e-06, "loss": 0.2285, "step": 16278 }, { "epoch": 0.81, "learning_rate": 1.781413597599273e-06, "loss": 0.557, "step": 16279 }, { "epoch": 0.81, "learning_rate": 1.7804923941790686e-06, "loss": 0.3611, "step": 16280 }, { "epoch": 0.81, "learning_rate": 1.7795714057293189e-06, "loss": 0.4122, "step": 16281 }, { "epoch": 0.81, "learning_rate": 1.7786506322741126e-06, "loss": 0.251, "step": 16282 }, { "epoch": 0.81, "learning_rate": 1.7777300738375314e-06, "loss": 0.262, "step": 16283 }, { "epoch": 0.81, "learning_rate": 1.7768097304436494e-06, "loss": 0.4062, "step": 16284 }, { "epoch": 0.81, "learning_rate": 1.7758896021165373e-06, "loss": 0.2866, "step": 16285 }, { "epoch": 0.81, "learning_rate": 1.7749696888802636e-06, "loss": 0.5082, "step": 16286 }, { "epoch": 0.81, "learning_rate": 1.7740499907588837e-06, "loss": 0.6293, "step": 16287 }, { "epoch": 0.81, "learning_rate": 1.7731305077764538e-06, "loss": 0.332, "step": 16288 }, { "epoch": 0.81, "learning_rate": 1.772211239957019e-06, "loss": 0.2966, "step": 16289 }, { "epoch": 0.81, "learning_rate": 1.7712921873246203e-06, "loss": 0.4634, "step": 16290 }, { "epoch": 0.81, "learning_rate": 1.7703733499032992e-06, "loss": 0.3762, "step": 16291 }, { "epoch": 0.81, "learning_rate": 1.769454727717085e-06, "loss": 0.2763, "step": 16292 }, { "epoch": 0.81, "learning_rate": 1.768536320790002e-06, "loss": 0.6531, "step": 16293 }, { "epoch": 0.81, "learning_rate": 1.7676181291460681e-06, "loss": 0.4804, "step": 16294 }, { "epoch": 0.81, "learning_rate": 1.7667001528093009e-06, "loss": 0.3599, "step": 16295 }, { "epoch": 0.81, "learning_rate": 1.7657823918037088e-06, "loss": 0.483, "step": 16296 }, { "epoch": 0.81, "learning_rate": 1.7648648461532924e-06, "loss": 0.1684, "step": 16297 }, { "epoch": 0.81, "learning_rate": 1.7639475158820506e-06, "loss": 0.2514, "step": 16298 }, { "epoch": 0.81, "learning_rate": 1.763030401013971e-06, "loss": 0.3952, "step": 16299 }, { "epoch": 0.81, "learning_rate": 1.7621135015730462e-06, "loss": 0.4528, "step": 16300 }, { "epoch": 0.81, "learning_rate": 1.7611968175832516e-06, "loss": 0.3758, "step": 16301 }, { "epoch": 0.81, "learning_rate": 1.7602803490685649e-06, "loss": 0.3916, "step": 16302 }, { "epoch": 0.81, "learning_rate": 1.7593640960529534e-06, "loss": 0.209, "step": 16303 }, { "epoch": 0.81, "learning_rate": 1.7584480585603803e-06, "loss": 0.3077, "step": 16304 }, { "epoch": 0.81, "learning_rate": 1.757532236614803e-06, "loss": 0.1575, "step": 16305 }, { "epoch": 0.81, "learning_rate": 1.7566166302401722e-06, "loss": 0.2522, "step": 16306 }, { "epoch": 0.81, "learning_rate": 1.75570123946044e-06, "loss": 0.3158, "step": 16307 }, { "epoch": 0.81, "learning_rate": 1.754786064299543e-06, "loss": 0.2002, "step": 16308 }, { "epoch": 0.81, "learning_rate": 1.7538711047814172e-06, "loss": 0.2453, "step": 16309 }, { "epoch": 0.81, "learning_rate": 1.7529563609299927e-06, "loss": 0.3949, "step": 16310 }, { "epoch": 0.81, "learning_rate": 1.7520418327691924e-06, "loss": 0.3186, "step": 16311 }, { "epoch": 0.81, "learning_rate": 1.751127520322935e-06, "loss": 0.3963, "step": 16312 }, { "epoch": 0.81, "learning_rate": 1.7502134236151313e-06, "loss": 0.355, "step": 16313 }, { "epoch": 0.81, "learning_rate": 1.7492995426696924e-06, "loss": 0.3407, "step": 16314 }, { "epoch": 0.81, "learning_rate": 1.7483858775105167e-06, "loss": 0.3394, "step": 16315 }, { "epoch": 0.81, "learning_rate": 1.7474724281615018e-06, "loss": 0.4888, "step": 16316 }, { "epoch": 0.81, "learning_rate": 1.7465591946465342e-06, "loss": 0.36, "step": 16317 }, { "epoch": 0.81, "learning_rate": 1.7456461769895027e-06, "loss": 0.2689, "step": 16318 }, { "epoch": 0.81, "learning_rate": 1.7447333752142848e-06, "loss": 0.4961, "step": 16319 }, { "epoch": 0.81, "learning_rate": 1.7438207893447522e-06, "loss": 0.3814, "step": 16320 }, { "epoch": 0.81, "learning_rate": 1.742908419404774e-06, "loss": 0.2315, "step": 16321 }, { "epoch": 0.81, "learning_rate": 1.7419962654182076e-06, "loss": 0.2235, "step": 16322 }, { "epoch": 0.82, "learning_rate": 1.7410843274089161e-06, "loss": 0.3398, "step": 16323 }, { "epoch": 0.82, "learning_rate": 1.7401726054007473e-06, "loss": 0.2797, "step": 16324 }, { "epoch": 0.82, "learning_rate": 1.7392610994175451e-06, "loss": 0.6737, "step": 16325 }, { "epoch": 0.82, "learning_rate": 1.7383498094831464e-06, "loss": 0.1879, "step": 16326 }, { "epoch": 0.82, "learning_rate": 1.7374387356213907e-06, "loss": 0.3781, "step": 16327 }, { "epoch": 0.82, "learning_rate": 1.7365278778561024e-06, "loss": 0.3295, "step": 16328 }, { "epoch": 0.82, "learning_rate": 1.7356172362111045e-06, "loss": 0.3994, "step": 16329 }, { "epoch": 0.82, "learning_rate": 1.7347068107102128e-06, "loss": 0.4444, "step": 16330 }, { "epoch": 0.82, "learning_rate": 1.73379660137724e-06, "loss": 0.4218, "step": 16331 }, { "epoch": 0.82, "learning_rate": 1.7328866082359896e-06, "loss": 0.324, "step": 16332 }, { "epoch": 0.82, "learning_rate": 1.7319768313102614e-06, "loss": 0.4476, "step": 16333 }, { "epoch": 0.82, "learning_rate": 1.7310672706238507e-06, "loss": 0.318, "step": 16334 }, { "epoch": 0.82, "learning_rate": 1.7301579262005431e-06, "loss": 0.3582, "step": 16335 }, { "epoch": 0.82, "learning_rate": 1.7292487980641247e-06, "loss": 0.616, "step": 16336 }, { "epoch": 0.82, "learning_rate": 1.7283398862383727e-06, "loss": 0.2601, "step": 16337 }, { "epoch": 0.82, "learning_rate": 1.7274311907470564e-06, "loss": 0.2432, "step": 16338 }, { "epoch": 0.82, "learning_rate": 1.7265227116139416e-06, "loss": 0.2952, "step": 16339 }, { "epoch": 0.82, "learning_rate": 1.7256144488627869e-06, "loss": 0.3336, "step": 16340 }, { "epoch": 0.82, "learning_rate": 1.7247064025173522e-06, "loss": 0.2694, "step": 16341 }, { "epoch": 0.82, "learning_rate": 1.7237985726013818e-06, "loss": 0.492, "step": 16342 }, { "epoch": 0.82, "learning_rate": 1.72289095913862e-06, "loss": 0.3056, "step": 16343 }, { "epoch": 0.82, "learning_rate": 1.7219835621528035e-06, "loss": 0.447, "step": 16344 }, { "epoch": 0.82, "learning_rate": 1.7210763816676635e-06, "loss": 0.271, "step": 16345 }, { "epoch": 0.82, "learning_rate": 1.7201694177069294e-06, "loss": 0.4278, "step": 16346 }, { "epoch": 0.82, "learning_rate": 1.7192626702943193e-06, "loss": 0.5538, "step": 16347 }, { "epoch": 0.82, "learning_rate": 1.7183561394535476e-06, "loss": 0.2695, "step": 16348 }, { "epoch": 0.82, "learning_rate": 1.7174498252083227e-06, "loss": 0.3731, "step": 16349 }, { "epoch": 0.82, "learning_rate": 1.7165437275823516e-06, "loss": 0.3467, "step": 16350 }, { "epoch": 0.82, "learning_rate": 1.7156378465993296e-06, "loss": 0.453, "step": 16351 }, { "epoch": 0.82, "learning_rate": 1.71473218228295e-06, "loss": 0.2681, "step": 16352 }, { "epoch": 0.82, "learning_rate": 1.7138267346568971e-06, "loss": 0.4445, "step": 16353 }, { "epoch": 0.82, "learning_rate": 1.7129215037448532e-06, "loss": 0.3149, "step": 16354 }, { "epoch": 0.82, "learning_rate": 1.712016489570494e-06, "loss": 0.256, "step": 16355 }, { "epoch": 0.82, "learning_rate": 1.7111116921574876e-06, "loss": 0.1536, "step": 16356 }, { "epoch": 0.82, "learning_rate": 1.7102071115294994e-06, "loss": 0.4836, "step": 16357 }, { "epoch": 0.82, "learning_rate": 1.7093027477101831e-06, "loss": 0.4158, "step": 16358 }, { "epoch": 0.82, "learning_rate": 1.7083986007231978e-06, "loss": 0.5131, "step": 16359 }, { "epoch": 0.82, "learning_rate": 1.7074946705921869e-06, "loss": 0.4, "step": 16360 }, { "epoch": 0.82, "learning_rate": 1.7065909573407912e-06, "loss": 0.6043, "step": 16361 }, { "epoch": 0.82, "learning_rate": 1.7056874609926467e-06, "loss": 0.3564, "step": 16362 }, { "epoch": 0.82, "learning_rate": 1.7047841815713805e-06, "loss": 0.37, "step": 16363 }, { "epoch": 0.82, "learning_rate": 1.7038811191006221e-06, "loss": 0.2032, "step": 16364 }, { "epoch": 0.82, "learning_rate": 1.702978273603987e-06, "loss": 0.2734, "step": 16365 }, { "epoch": 0.82, "learning_rate": 1.7020756451050868e-06, "loss": 0.4584, "step": 16366 }, { "epoch": 0.82, "learning_rate": 1.701173233627529e-06, "loss": 0.3816, "step": 16367 }, { "epoch": 0.82, "learning_rate": 1.7002710391949172e-06, "loss": 0.1736, "step": 16368 }, { "epoch": 0.82, "learning_rate": 1.6993690618308456e-06, "loss": 0.4579, "step": 16369 }, { "epoch": 0.82, "learning_rate": 1.6984673015589047e-06, "loss": 0.2199, "step": 16370 }, { "epoch": 0.82, "learning_rate": 1.6975657584026794e-06, "loss": 0.3673, "step": 16371 }, { "epoch": 0.82, "learning_rate": 1.696664432385744e-06, "loss": 0.2641, "step": 16372 }, { "epoch": 0.82, "learning_rate": 1.6957633235316773e-06, "loss": 0.3349, "step": 16373 }, { "epoch": 0.82, "learning_rate": 1.694862431864045e-06, "loss": 0.38, "step": 16374 }, { "epoch": 0.82, "learning_rate": 1.6939617574064071e-06, "loss": 0.6809, "step": 16375 }, { "epoch": 0.82, "learning_rate": 1.6930613001823204e-06, "loss": 0.5145, "step": 16376 }, { "epoch": 0.82, "learning_rate": 1.692161060215336e-06, "loss": 0.306, "step": 16377 }, { "epoch": 0.82, "learning_rate": 1.691261037528995e-06, "loss": 0.4487, "step": 16378 }, { "epoch": 0.82, "learning_rate": 1.6903612321468422e-06, "loss": 0.2557, "step": 16379 }, { "epoch": 0.82, "learning_rate": 1.6894616440924071e-06, "loss": 0.3688, "step": 16380 }, { "epoch": 0.82, "learning_rate": 1.688562273389217e-06, "loss": 0.3973, "step": 16381 }, { "epoch": 0.82, "learning_rate": 1.6876631200607962e-06, "loss": 0.166, "step": 16382 }, { "epoch": 0.82, "learning_rate": 1.6867641841306582e-06, "loss": 0.3174, "step": 16383 }, { "epoch": 0.82, "learning_rate": 1.6858654656223151e-06, "loss": 0.1548, "step": 16384 }, { "epoch": 0.82, "learning_rate": 1.6849669645592715e-06, "loss": 0.2475, "step": 16385 }, { "epoch": 0.82, "learning_rate": 1.6840686809650231e-06, "loss": 0.5191, "step": 16386 }, { "epoch": 0.82, "learning_rate": 1.683170614863069e-06, "loss": 0.4536, "step": 16387 }, { "epoch": 0.82, "learning_rate": 1.6822727662768945e-06, "loss": 0.6333, "step": 16388 }, { "epoch": 0.82, "learning_rate": 1.6813751352299811e-06, "loss": 0.2419, "step": 16389 }, { "epoch": 0.82, "learning_rate": 1.6804777217458034e-06, "loss": 0.2045, "step": 16390 }, { "epoch": 0.82, "learning_rate": 1.679580525847837e-06, "loss": 0.1903, "step": 16391 }, { "epoch": 0.82, "learning_rate": 1.6786835475595443e-06, "loss": 0.355, "step": 16392 }, { "epoch": 0.82, "learning_rate": 1.677786786904384e-06, "loss": 0.331, "step": 16393 }, { "epoch": 0.82, "learning_rate": 1.67689024390581e-06, "loss": 0.3318, "step": 16394 }, { "epoch": 0.82, "learning_rate": 1.6759939185872686e-06, "loss": 0.5609, "step": 16395 }, { "epoch": 0.82, "learning_rate": 1.675097810972206e-06, "loss": 0.3365, "step": 16396 }, { "epoch": 0.82, "learning_rate": 1.6742019210840554e-06, "loss": 0.4548, "step": 16397 }, { "epoch": 0.82, "learning_rate": 1.6733062489462503e-06, "loss": 0.3577, "step": 16398 }, { "epoch": 0.82, "learning_rate": 1.672410794582211e-06, "loss": 0.2803, "step": 16399 }, { "epoch": 0.82, "learning_rate": 1.671515558015363e-06, "loss": 0.537, "step": 16400 }, { "epoch": 0.82, "learning_rate": 1.6706205392691166e-06, "loss": 0.567, "step": 16401 }, { "epoch": 0.82, "learning_rate": 1.6697257383668807e-06, "loss": 0.2714, "step": 16402 }, { "epoch": 0.82, "learning_rate": 1.668831155332058e-06, "loss": 0.3794, "step": 16403 }, { "epoch": 0.82, "learning_rate": 1.6679367901880438e-06, "loss": 0.043, "step": 16404 }, { "epoch": 0.82, "learning_rate": 1.667042642958231e-06, "loss": 0.3344, "step": 16405 }, { "epoch": 0.82, "learning_rate": 1.6661487136660025e-06, "loss": 0.5626, "step": 16406 }, { "epoch": 0.82, "learning_rate": 1.6652550023347392e-06, "loss": 0.3716, "step": 16407 }, { "epoch": 0.82, "learning_rate": 1.6643615089878151e-06, "loss": 0.5113, "step": 16408 }, { "epoch": 0.82, "learning_rate": 1.6634682336485953e-06, "loss": 0.3891, "step": 16409 }, { "epoch": 0.82, "learning_rate": 1.6625751763404474e-06, "loss": 0.4222, "step": 16410 }, { "epoch": 0.82, "learning_rate": 1.6616823370867253e-06, "loss": 0.1461, "step": 16411 }, { "epoch": 0.82, "learning_rate": 1.6607897159107812e-06, "loss": 0.4631, "step": 16412 }, { "epoch": 0.82, "learning_rate": 1.6598973128359563e-06, "loss": 0.3852, "step": 16413 }, { "epoch": 0.82, "learning_rate": 1.6590051278855969e-06, "loss": 0.5637, "step": 16414 }, { "epoch": 0.82, "learning_rate": 1.6581131610830336e-06, "loss": 0.326, "step": 16415 }, { "epoch": 0.82, "learning_rate": 1.6572214124515939e-06, "loss": 0.2856, "step": 16416 }, { "epoch": 0.82, "learning_rate": 1.6563298820146023e-06, "loss": 0.5549, "step": 16417 }, { "epoch": 0.82, "learning_rate": 1.6554385697953712e-06, "loss": 0.325, "step": 16418 }, { "epoch": 0.82, "learning_rate": 1.6545474758172175e-06, "loss": 0.3536, "step": 16419 }, { "epoch": 0.82, "learning_rate": 1.6536566001034437e-06, "loss": 0.2214, "step": 16420 }, { "epoch": 0.82, "learning_rate": 1.6527659426773502e-06, "loss": 0.1585, "step": 16421 }, { "epoch": 0.82, "learning_rate": 1.651875503562228e-06, "loss": 0.4964, "step": 16422 }, { "epoch": 0.82, "learning_rate": 1.65098528278137e-06, "loss": 0.3692, "step": 16423 }, { "epoch": 0.82, "learning_rate": 1.6500952803580573e-06, "loss": 0.3575, "step": 16424 }, { "epoch": 0.82, "learning_rate": 1.6492054963155658e-06, "loss": 0.3024, "step": 16425 }, { "epoch": 0.82, "learning_rate": 1.6483159306771668e-06, "loss": 0.5004, "step": 16426 }, { "epoch": 0.82, "learning_rate": 1.6474265834661263e-06, "loss": 0.3449, "step": 16427 }, { "epoch": 0.82, "learning_rate": 1.6465374547057033e-06, "loss": 0.2775, "step": 16428 }, { "epoch": 0.82, "learning_rate": 1.6456485444191516e-06, "loss": 0.4952, "step": 16429 }, { "epoch": 0.82, "learning_rate": 1.6447598526297205e-06, "loss": 0.39, "step": 16430 }, { "epoch": 0.82, "learning_rate": 1.6438713793606498e-06, "loss": 0.36, "step": 16431 }, { "epoch": 0.82, "learning_rate": 1.6429831246351813e-06, "loss": 0.3756, "step": 16432 }, { "epoch": 0.82, "learning_rate": 1.642095088476543e-06, "loss": 0.6487, "step": 16433 }, { "epoch": 0.82, "learning_rate": 1.6412072709079606e-06, "loss": 0.4519, "step": 16434 }, { "epoch": 0.82, "learning_rate": 1.6403196719526549e-06, "loss": 0.4883, "step": 16435 }, { "epoch": 0.82, "learning_rate": 1.6394322916338357e-06, "loss": 0.2845, "step": 16436 }, { "epoch": 0.82, "learning_rate": 1.6385451299747168e-06, "loss": 0.104, "step": 16437 }, { "epoch": 0.82, "learning_rate": 1.6376581869984987e-06, "loss": 0.3184, "step": 16438 }, { "epoch": 0.82, "learning_rate": 1.6367714627283782e-06, "loss": 0.4071, "step": 16439 }, { "epoch": 0.82, "learning_rate": 1.6358849571875435e-06, "loss": 0.3181, "step": 16440 }, { "epoch": 0.82, "learning_rate": 1.634998670399186e-06, "loss": 0.29, "step": 16441 }, { "epoch": 0.82, "learning_rate": 1.6341126023864818e-06, "loss": 0.2548, "step": 16442 }, { "epoch": 0.82, "learning_rate": 1.6332267531726042e-06, "loss": 0.2794, "step": 16443 }, { "epoch": 0.82, "learning_rate": 1.6323411227807229e-06, "loss": 0.4471, "step": 16444 }, { "epoch": 0.82, "learning_rate": 1.6314557112339979e-06, "loss": 0.1895, "step": 16445 }, { "epoch": 0.82, "learning_rate": 1.6305705185555898e-06, "loss": 0.2853, "step": 16446 }, { "epoch": 0.82, "learning_rate": 1.6296855447686478e-06, "loss": 0.3431, "step": 16447 }, { "epoch": 0.82, "learning_rate": 1.6288007898963175e-06, "loss": 0.2568, "step": 16448 }, { "epoch": 0.82, "learning_rate": 1.6279162539617389e-06, "loss": 0.3079, "step": 16449 }, { "epoch": 0.82, "learning_rate": 1.6270319369880416e-06, "loss": 0.3568, "step": 16450 }, { "epoch": 0.82, "learning_rate": 1.62614783899836e-06, "loss": 0.274, "step": 16451 }, { "epoch": 0.82, "learning_rate": 1.6252639600158137e-06, "loss": 0.4782, "step": 16452 }, { "epoch": 0.82, "learning_rate": 1.6243803000635205e-06, "loss": 0.2748, "step": 16453 }, { "epoch": 0.82, "learning_rate": 1.6234968591645905e-06, "loss": 0.2767, "step": 16454 }, { "epoch": 0.82, "learning_rate": 1.6226136373421286e-06, "loss": 0.3822, "step": 16455 }, { "epoch": 0.82, "learning_rate": 1.621730634619234e-06, "loss": 0.4528, "step": 16456 }, { "epoch": 0.82, "learning_rate": 1.6208478510190018e-06, "loss": 0.3749, "step": 16457 }, { "epoch": 0.82, "learning_rate": 1.6199652865645189e-06, "loss": 0.267, "step": 16458 }, { "epoch": 0.82, "learning_rate": 1.6190829412788656e-06, "loss": 0.4814, "step": 16459 }, { "epoch": 0.82, "learning_rate": 1.6182008151851236e-06, "loss": 0.4596, "step": 16460 }, { "epoch": 0.82, "learning_rate": 1.6173189083063612e-06, "loss": 0.1722, "step": 16461 }, { "epoch": 0.82, "learning_rate": 1.6164372206656432e-06, "loss": 0.5279, "step": 16462 }, { "epoch": 0.82, "learning_rate": 1.6155557522860266e-06, "loss": 0.2846, "step": 16463 }, { "epoch": 0.82, "learning_rate": 1.6146745031905709e-06, "loss": 0.5499, "step": 16464 }, { "epoch": 0.82, "learning_rate": 1.6137934734023197e-06, "loss": 0.444, "step": 16465 }, { "epoch": 0.82, "learning_rate": 1.612912662944317e-06, "loss": 0.4314, "step": 16466 }, { "epoch": 0.82, "learning_rate": 1.6120320718395977e-06, "loss": 0.4787, "step": 16467 }, { "epoch": 0.82, "learning_rate": 1.611151700111192e-06, "loss": 0.3624, "step": 16468 }, { "epoch": 0.82, "learning_rate": 1.6102715477821274e-06, "loss": 0.3329, "step": 16469 }, { "epoch": 0.82, "learning_rate": 1.609391614875423e-06, "loss": 0.2688, "step": 16470 }, { "epoch": 0.82, "learning_rate": 1.6085119014140905e-06, "loss": 0.3915, "step": 16471 }, { "epoch": 0.82, "learning_rate": 1.6076324074211359e-06, "loss": 0.4395, "step": 16472 }, { "epoch": 0.82, "learning_rate": 1.6067531329195663e-06, "loss": 0.325, "step": 16473 }, { "epoch": 0.82, "learning_rate": 1.6058740779323756e-06, "loss": 0.3536, "step": 16474 }, { "epoch": 0.82, "learning_rate": 1.6049952424825532e-06, "loss": 0.2458, "step": 16475 }, { "epoch": 0.82, "learning_rate": 1.6041166265930853e-06, "loss": 0.5378, "step": 16476 }, { "epoch": 0.82, "learning_rate": 1.603238230286951e-06, "loss": 0.4334, "step": 16477 }, { "epoch": 0.82, "learning_rate": 1.6023600535871219e-06, "loss": 0.4824, "step": 16478 }, { "epoch": 0.82, "learning_rate": 1.6014820965165678e-06, "loss": 0.565, "step": 16479 }, { "epoch": 0.82, "learning_rate": 1.600604359098249e-06, "loss": 0.4866, "step": 16480 }, { "epoch": 0.82, "learning_rate": 1.5997268413551215e-06, "loss": 0.4259, "step": 16481 }, { "epoch": 0.82, "learning_rate": 1.5988495433101336e-06, "loss": 0.2951, "step": 16482 }, { "epoch": 0.82, "learning_rate": 1.597972464986235e-06, "loss": 0.2579, "step": 16483 }, { "epoch": 0.82, "learning_rate": 1.5970956064063624e-06, "loss": 0.3341, "step": 16484 }, { "epoch": 0.82, "learning_rate": 1.5962189675934482e-06, "loss": 0.3101, "step": 16485 }, { "epoch": 0.82, "learning_rate": 1.5953425485704165e-06, "loss": 0.4484, "step": 16486 }, { "epoch": 0.82, "learning_rate": 1.5944663493601952e-06, "loss": 0.6192, "step": 16487 }, { "epoch": 0.82, "learning_rate": 1.5935903699856981e-06, "loss": 0.5155, "step": 16488 }, { "epoch": 0.82, "learning_rate": 1.5927146104698344e-06, "loss": 0.4808, "step": 16489 }, { "epoch": 0.82, "learning_rate": 1.591839070835508e-06, "loss": 0.3417, "step": 16490 }, { "epoch": 0.82, "learning_rate": 1.5909637511056153e-06, "loss": 0.4343, "step": 16491 }, { "epoch": 0.82, "learning_rate": 1.5900886513030545e-06, "loss": 0.5183, "step": 16492 }, { "epoch": 0.82, "learning_rate": 1.5892137714507095e-06, "loss": 0.3342, "step": 16493 }, { "epoch": 0.82, "learning_rate": 1.588339111571463e-06, "loss": 0.3706, "step": 16494 }, { "epoch": 0.82, "learning_rate": 1.587464671688187e-06, "loss": 0.3651, "step": 16495 }, { "epoch": 0.82, "learning_rate": 1.5865904518237562e-06, "loss": 0.4946, "step": 16496 }, { "epoch": 0.82, "learning_rate": 1.5857164520010337e-06, "loss": 0.5154, "step": 16497 }, { "epoch": 0.82, "learning_rate": 1.5848426722428755e-06, "loss": 0.365, "step": 16498 }, { "epoch": 0.82, "learning_rate": 1.5839691125721358e-06, "loss": 0.2352, "step": 16499 }, { "epoch": 0.82, "learning_rate": 1.5830957730116602e-06, "loss": 0.3143, "step": 16500 }, { "epoch": 0.82, "learning_rate": 1.5822226535842922e-06, "loss": 0.312, "step": 16501 }, { "epoch": 0.82, "learning_rate": 1.5813497543128642e-06, "loss": 0.4221, "step": 16502 }, { "epoch": 0.82, "learning_rate": 1.580477075220207e-06, "loss": 0.449, "step": 16503 }, { "epoch": 0.82, "learning_rate": 1.5796046163291412e-06, "loss": 0.2658, "step": 16504 }, { "epoch": 0.82, "learning_rate": 1.5787323776624918e-06, "loss": 0.1582, "step": 16505 }, { "epoch": 0.82, "learning_rate": 1.5778603592430663e-06, "loss": 0.4919, "step": 16506 }, { "epoch": 0.82, "learning_rate": 1.576988561093672e-06, "loss": 0.3604, "step": 16507 }, { "epoch": 0.82, "learning_rate": 1.5761169832371103e-06, "loss": 0.2092, "step": 16508 }, { "epoch": 0.82, "learning_rate": 1.5752456256961723e-06, "loss": 0.3197, "step": 16509 }, { "epoch": 0.82, "learning_rate": 1.574374488493654e-06, "loss": 0.0533, "step": 16510 }, { "epoch": 0.82, "learning_rate": 1.573503571652334e-06, "loss": 0.1386, "step": 16511 }, { "epoch": 0.82, "learning_rate": 1.5726328751949926e-06, "loss": 0.2961, "step": 16512 }, { "epoch": 0.82, "learning_rate": 1.5717623991443999e-06, "loss": 0.2317, "step": 16513 }, { "epoch": 0.82, "learning_rate": 1.5708921435233205e-06, "loss": 0.2461, "step": 16514 }, { "epoch": 0.82, "learning_rate": 1.5700221083545197e-06, "loss": 0.3652, "step": 16515 }, { "epoch": 0.82, "learning_rate": 1.5691522936607484e-06, "loss": 0.4237, "step": 16516 }, { "epoch": 0.82, "learning_rate": 1.5682826994647571e-06, "loss": 0.3707, "step": 16517 }, { "epoch": 0.82, "learning_rate": 1.5674133257892854e-06, "loss": 0.3383, "step": 16518 }, { "epoch": 0.82, "learning_rate": 1.5665441726570762e-06, "loss": 0.2825, "step": 16519 }, { "epoch": 0.82, "learning_rate": 1.5656752400908581e-06, "loss": 0.3657, "step": 16520 }, { "epoch": 0.82, "learning_rate": 1.5648065281133574e-06, "loss": 0.3991, "step": 16521 }, { "epoch": 0.82, "learning_rate": 1.5639380367472934e-06, "loss": 0.1549, "step": 16522 }, { "epoch": 0.83, "learning_rate": 1.5630697660153793e-06, "loss": 0.2693, "step": 16523 }, { "epoch": 0.83, "learning_rate": 1.5622017159403269e-06, "loss": 0.3066, "step": 16524 }, { "epoch": 0.83, "learning_rate": 1.5613338865448369e-06, "loss": 0.314, "step": 16525 }, { "epoch": 0.83, "learning_rate": 1.5604662778516066e-06, "loss": 0.2952, "step": 16526 }, { "epoch": 0.83, "learning_rate": 1.559598889883327e-06, "loss": 0.4475, "step": 16527 }, { "epoch": 0.83, "learning_rate": 1.558731722662683e-06, "loss": 0.2863, "step": 16528 }, { "epoch": 0.83, "learning_rate": 1.5578647762123555e-06, "loss": 0.3708, "step": 16529 }, { "epoch": 0.83, "learning_rate": 1.5569980505550175e-06, "loss": 0.4293, "step": 16530 }, { "epoch": 0.83, "learning_rate": 1.5561315457133375e-06, "loss": 0.3441, "step": 16531 }, { "epoch": 0.83, "learning_rate": 1.555265261709974e-06, "loss": 0.529, "step": 16532 }, { "epoch": 0.83, "learning_rate": 1.55439919856759e-06, "loss": 0.4324, "step": 16533 }, { "epoch": 0.83, "learning_rate": 1.5535333563088328e-06, "loss": 0.2099, "step": 16534 }, { "epoch": 0.83, "learning_rate": 1.5526677349563479e-06, "loss": 0.2414, "step": 16535 }, { "epoch": 0.83, "learning_rate": 1.5518023345327726e-06, "loss": 0.4655, "step": 16536 }, { "epoch": 0.83, "learning_rate": 1.5509371550607433e-06, "loss": 0.2104, "step": 16537 }, { "epoch": 0.83, "learning_rate": 1.5500721965628874e-06, "loss": 0.2924, "step": 16538 }, { "epoch": 0.83, "learning_rate": 1.5492074590618256e-06, "loss": 0.4589, "step": 16539 }, { "epoch": 0.83, "learning_rate": 1.5483429425801744e-06, "loss": 0.4086, "step": 16540 }, { "epoch": 0.83, "learning_rate": 1.5474786471405411e-06, "loss": 0.1705, "step": 16541 }, { "epoch": 0.83, "learning_rate": 1.5466145727655347e-06, "loss": 0.1463, "step": 16542 }, { "epoch": 0.83, "learning_rate": 1.5457507194777532e-06, "loss": 0.5091, "step": 16543 }, { "epoch": 0.83, "learning_rate": 1.544887087299788e-06, "loss": 0.3783, "step": 16544 }, { "epoch": 0.83, "learning_rate": 1.544023676254227e-06, "loss": 0.1595, "step": 16545 }, { "epoch": 0.83, "learning_rate": 1.5431604863636496e-06, "loss": 0.2214, "step": 16546 }, { "epoch": 0.83, "learning_rate": 1.542297517650635e-06, "loss": 0.309, "step": 16547 }, { "epoch": 0.83, "learning_rate": 1.5414347701377507e-06, "loss": 0.4936, "step": 16548 }, { "epoch": 0.83, "learning_rate": 1.5405722438475623e-06, "loss": 0.3164, "step": 16549 }, { "epoch": 0.83, "learning_rate": 1.5397099388026271e-06, "loss": 0.5003, "step": 16550 }, { "epoch": 0.83, "learning_rate": 1.538847855025496e-06, "loss": 0.6394, "step": 16551 }, { "epoch": 0.83, "learning_rate": 1.5379859925387186e-06, "loss": 0.3722, "step": 16552 }, { "epoch": 0.83, "learning_rate": 1.5371243513648338e-06, "loss": 0.3596, "step": 16553 }, { "epoch": 0.83, "learning_rate": 1.5362629315263767e-06, "loss": 0.4177, "step": 16554 }, { "epoch": 0.83, "learning_rate": 1.535401733045876e-06, "loss": 0.4256, "step": 16555 }, { "epoch": 0.83, "learning_rate": 1.534540755945858e-06, "loss": 0.4229, "step": 16556 }, { "epoch": 0.83, "learning_rate": 1.5336800002488383e-06, "loss": 0.2108, "step": 16557 }, { "epoch": 0.83, "learning_rate": 1.5328194659773299e-06, "loss": 0.2293, "step": 16558 }, { "epoch": 0.83, "learning_rate": 1.5319591531538359e-06, "loss": 0.4003, "step": 16559 }, { "epoch": 0.83, "learning_rate": 1.5310990618008615e-06, "loss": 0.4054, "step": 16560 }, { "epoch": 0.83, "learning_rate": 1.5302391919408988e-06, "loss": 0.3654, "step": 16561 }, { "epoch": 0.83, "learning_rate": 1.5293795435964365e-06, "loss": 0.3586, "step": 16562 }, { "epoch": 0.83, "learning_rate": 1.5285201167899577e-06, "loss": 0.3501, "step": 16563 }, { "epoch": 0.83, "learning_rate": 1.5276609115439368e-06, "loss": 0.2584, "step": 16564 }, { "epoch": 0.83, "learning_rate": 1.526801927880851e-06, "loss": 0.3123, "step": 16565 }, { "epoch": 0.83, "learning_rate": 1.5259431658231628e-06, "loss": 0.3805, "step": 16566 }, { "epoch": 0.83, "learning_rate": 1.5250846253933327e-06, "loss": 0.3856, "step": 16567 }, { "epoch": 0.83, "learning_rate": 1.5242263066138108e-06, "loss": 0.2619, "step": 16568 }, { "epoch": 0.83, "learning_rate": 1.5233682095070512e-06, "loss": 0.4347, "step": 16569 }, { "epoch": 0.83, "learning_rate": 1.522510334095494e-06, "loss": 0.3845, "step": 16570 }, { "epoch": 0.83, "learning_rate": 1.5216526804015752e-06, "loss": 0.324, "step": 16571 }, { "epoch": 0.83, "learning_rate": 1.5207952484477251e-06, "loss": 0.4125, "step": 16572 }, { "epoch": 0.83, "learning_rate": 1.519938038256371e-06, "loss": 0.4996, "step": 16573 }, { "epoch": 0.83, "learning_rate": 1.5190810498499298e-06, "loss": 0.2669, "step": 16574 }, { "epoch": 0.83, "learning_rate": 1.5182242832508154e-06, "loss": 0.4902, "step": 16575 }, { "epoch": 0.83, "learning_rate": 1.5173677384814367e-06, "loss": 0.1566, "step": 16576 }, { "epoch": 0.83, "learning_rate": 1.5165114155641913e-06, "loss": 0.432, "step": 16577 }, { "epoch": 0.83, "learning_rate": 1.5156553145214814e-06, "loss": 0.5661, "step": 16578 }, { "epoch": 0.83, "learning_rate": 1.5147994353756945e-06, "loss": 0.3341, "step": 16579 }, { "epoch": 0.83, "learning_rate": 1.513943778149214e-06, "loss": 0.3635, "step": 16580 }, { "epoch": 0.83, "learning_rate": 1.5130883428644194e-06, "loss": 0.4233, "step": 16581 }, { "epoch": 0.83, "learning_rate": 1.512233129543682e-06, "loss": 0.2597, "step": 16582 }, { "epoch": 0.83, "learning_rate": 1.5113781382093706e-06, "loss": 0.5333, "step": 16583 }, { "epoch": 0.83, "learning_rate": 1.5105233688838472e-06, "loss": 0.1523, "step": 16584 }, { "epoch": 0.83, "learning_rate": 1.5096688215894661e-06, "loss": 0.1831, "step": 16585 }, { "epoch": 0.83, "learning_rate": 1.5088144963485762e-06, "loss": 0.303, "step": 16586 }, { "epoch": 0.83, "learning_rate": 1.5079603931835186e-06, "loss": 0.4841, "step": 16587 }, { "epoch": 0.83, "learning_rate": 1.5071065121166373e-06, "loss": 0.286, "step": 16588 }, { "epoch": 0.83, "learning_rate": 1.5062528531702626e-06, "loss": 0.3647, "step": 16589 }, { "epoch": 0.83, "learning_rate": 1.5053994163667185e-06, "loss": 0.4259, "step": 16590 }, { "epoch": 0.83, "learning_rate": 1.5045462017283242e-06, "loss": 0.2866, "step": 16591 }, { "epoch": 0.83, "learning_rate": 1.5036932092773992e-06, "loss": 0.4758, "step": 16592 }, { "epoch": 0.83, "learning_rate": 1.5028404390362506e-06, "loss": 0.5171, "step": 16593 }, { "epoch": 0.83, "learning_rate": 1.5019878910271813e-06, "loss": 0.427, "step": 16594 }, { "epoch": 0.83, "learning_rate": 1.501135565272488e-06, "loss": 0.2223, "step": 16595 }, { "epoch": 0.83, "learning_rate": 1.5002834617944607e-06, "loss": 0.3108, "step": 16596 }, { "epoch": 0.83, "learning_rate": 1.499431580615388e-06, "loss": 0.3122, "step": 16597 }, { "epoch": 0.83, "learning_rate": 1.4985799217575493e-06, "loss": 0.3283, "step": 16598 }, { "epoch": 0.83, "learning_rate": 1.4977284852432183e-06, "loss": 0.4051, "step": 16599 }, { "epoch": 0.83, "learning_rate": 1.4968772710946621e-06, "loss": 0.3787, "step": 16600 }, { "epoch": 0.83, "learning_rate": 1.496026279334144e-06, "loss": 0.2033, "step": 16601 }, { "epoch": 0.83, "learning_rate": 1.4951755099839206e-06, "loss": 0.4447, "step": 16602 }, { "epoch": 0.83, "learning_rate": 1.4943249630662414e-06, "loss": 0.4509, "step": 16603 }, { "epoch": 0.83, "learning_rate": 1.493474638603354e-06, "loss": 0.4364, "step": 16604 }, { "epoch": 0.83, "learning_rate": 1.492624536617493e-06, "loss": 0.537, "step": 16605 }, { "epoch": 0.83, "learning_rate": 1.4917746571308955e-06, "loss": 0.4129, "step": 16606 }, { "epoch": 0.83, "learning_rate": 1.4909250001657894e-06, "loss": 0.4441, "step": 16607 }, { "epoch": 0.83, "learning_rate": 1.4900755657443955e-06, "loss": 0.4873, "step": 16608 }, { "epoch": 0.83, "learning_rate": 1.4892263538889262e-06, "loss": 0.2274, "step": 16609 }, { "epoch": 0.83, "learning_rate": 1.488377364621596e-06, "loss": 0.3454, "step": 16610 }, { "epoch": 0.83, "learning_rate": 1.4875285979646092e-06, "loss": 0.4015, "step": 16611 }, { "epoch": 0.83, "learning_rate": 1.4866800539401615e-06, "loss": 0.3793, "step": 16612 }, { "epoch": 0.83, "learning_rate": 1.4858317325704463e-06, "loss": 0.4587, "step": 16613 }, { "epoch": 0.83, "learning_rate": 1.4849836338776492e-06, "loss": 0.1868, "step": 16614 }, { "epoch": 0.83, "learning_rate": 1.4841357578839544e-06, "loss": 0.4489, "step": 16615 }, { "epoch": 0.83, "learning_rate": 1.4832881046115354e-06, "loss": 0.2792, "step": 16616 }, { "epoch": 0.83, "learning_rate": 1.4824406740825602e-06, "loss": 0.4265, "step": 16617 }, { "epoch": 0.83, "learning_rate": 1.4815934663191944e-06, "loss": 0.5659, "step": 16618 }, { "epoch": 0.83, "learning_rate": 1.4807464813435913e-06, "loss": 0.3943, "step": 16619 }, { "epoch": 0.83, "learning_rate": 1.4798997191779074e-06, "loss": 0.3308, "step": 16620 }, { "epoch": 0.83, "learning_rate": 1.4790531798442886e-06, "loss": 0.2314, "step": 16621 }, { "epoch": 0.83, "learning_rate": 1.4782068633648727e-06, "loss": 0.1822, "step": 16622 }, { "epoch": 0.83, "learning_rate": 1.4773607697617942e-06, "loss": 0.2578, "step": 16623 }, { "epoch": 0.83, "learning_rate": 1.476514899057183e-06, "loss": 0.3597, "step": 16624 }, { "epoch": 0.83, "learning_rate": 1.4756692512731608e-06, "loss": 0.2499, "step": 16625 }, { "epoch": 0.83, "learning_rate": 1.4748238264318437e-06, "loss": 0.395, "step": 16626 }, { "epoch": 0.83, "learning_rate": 1.4739786245553444e-06, "loss": 0.3431, "step": 16627 }, { "epoch": 0.83, "learning_rate": 1.4731336456657642e-06, "loss": 0.3925, "step": 16628 }, { "epoch": 0.83, "learning_rate": 1.4722888897852073e-06, "loss": 0.4254, "step": 16629 }, { "epoch": 0.83, "learning_rate": 1.4714443569357662e-06, "loss": 0.5405, "step": 16630 }, { "epoch": 0.83, "learning_rate": 1.4706000471395265e-06, "loss": 0.5578, "step": 16631 }, { "epoch": 0.83, "learning_rate": 1.4697559604185696e-06, "loss": 0.4999, "step": 16632 }, { "epoch": 0.83, "learning_rate": 1.4689120967949744e-06, "loss": 0.248, "step": 16633 }, { "epoch": 0.83, "learning_rate": 1.4680684562908098e-06, "loss": 0.369, "step": 16634 }, { "epoch": 0.83, "learning_rate": 1.4672250389281407e-06, "loss": 0.4868, "step": 16635 }, { "epoch": 0.83, "learning_rate": 1.4663818447290235e-06, "loss": 0.3327, "step": 16636 }, { "epoch": 0.83, "learning_rate": 1.4655388737155108e-06, "loss": 0.3771, "step": 16637 }, { "epoch": 0.83, "learning_rate": 1.4646961259096526e-06, "loss": 0.1795, "step": 16638 }, { "epoch": 0.83, "learning_rate": 1.463853601333488e-06, "loss": 0.4291, "step": 16639 }, { "epoch": 0.83, "learning_rate": 1.4630113000090528e-06, "loss": 0.1046, "step": 16640 }, { "epoch": 0.83, "learning_rate": 1.4621692219583727e-06, "loss": 0.2514, "step": 16641 }, { "epoch": 0.83, "learning_rate": 1.4613273672034767e-06, "loss": 0.3978, "step": 16642 }, { "epoch": 0.83, "learning_rate": 1.4604857357663794e-06, "loss": 0.3358, "step": 16643 }, { "epoch": 0.83, "learning_rate": 1.459644327669093e-06, "loss": 0.3727, "step": 16644 }, { "epoch": 0.83, "learning_rate": 1.4588031429336235e-06, "loss": 0.353, "step": 16645 }, { "epoch": 0.83, "learning_rate": 1.45796218158197e-06, "loss": 0.2413, "step": 16646 }, { "epoch": 0.83, "learning_rate": 1.457121443636128e-06, "loss": 0.369, "step": 16647 }, { "epoch": 0.83, "learning_rate": 1.4562809291180869e-06, "loss": 0.2874, "step": 16648 }, { "epoch": 0.83, "learning_rate": 1.4554406380498264e-06, "loss": 0.5527, "step": 16649 }, { "epoch": 0.83, "learning_rate": 1.454600570453325e-06, "loss": 0.2155, "step": 16650 }, { "epoch": 0.83, "learning_rate": 1.4537607263505505e-06, "loss": 0.2189, "step": 16651 }, { "epoch": 0.83, "learning_rate": 1.4529211057634728e-06, "loss": 0.494, "step": 16652 }, { "epoch": 0.83, "learning_rate": 1.4520817087140504e-06, "loss": 0.3528, "step": 16653 }, { "epoch": 0.83, "learning_rate": 1.4512425352242331e-06, "loss": 0.2926, "step": 16654 }, { "epoch": 0.83, "learning_rate": 1.4504035853159692e-06, "loss": 0.4156, "step": 16655 }, { "epoch": 0.83, "learning_rate": 1.4495648590112043e-06, "loss": 0.2827, "step": 16656 }, { "epoch": 0.83, "learning_rate": 1.448726356331871e-06, "loss": 0.3191, "step": 16657 }, { "epoch": 0.83, "learning_rate": 1.4478880772998994e-06, "loss": 0.2821, "step": 16658 }, { "epoch": 0.83, "learning_rate": 1.4470500219372142e-06, "loss": 0.3056, "step": 16659 }, { "epoch": 0.83, "learning_rate": 1.4462121902657312e-06, "loss": 0.1608, "step": 16660 }, { "epoch": 0.83, "learning_rate": 1.4453745823073672e-06, "loss": 0.4564, "step": 16661 }, { "epoch": 0.83, "learning_rate": 1.4445371980840262e-06, "loss": 0.2662, "step": 16662 }, { "epoch": 0.83, "learning_rate": 1.443700037617609e-06, "loss": 0.4027, "step": 16663 }, { "epoch": 0.83, "learning_rate": 1.4428631009300087e-06, "loss": 0.3, "step": 16664 }, { "epoch": 0.83, "learning_rate": 1.4420263880431183e-06, "loss": 0.3094, "step": 16665 }, { "epoch": 0.83, "learning_rate": 1.4411898989788186e-06, "loss": 0.3104, "step": 16666 }, { "epoch": 0.83, "learning_rate": 1.4403536337589873e-06, "loss": 0.389, "step": 16667 }, { "epoch": 0.83, "learning_rate": 1.439517592405496e-06, "loss": 0.3526, "step": 16668 }, { "epoch": 0.83, "learning_rate": 1.438681774940207e-06, "loss": 0.3977, "step": 16669 }, { "epoch": 0.83, "learning_rate": 1.4378461813849853e-06, "loss": 0.4587, "step": 16670 }, { "epoch": 0.83, "learning_rate": 1.4370108117616822e-06, "loss": 0.2732, "step": 16671 }, { "epoch": 0.83, "learning_rate": 1.4361756660921455e-06, "loss": 0.335, "step": 16672 }, { "epoch": 0.83, "learning_rate": 1.4353407443982182e-06, "loss": 0.5058, "step": 16673 }, { "epoch": 0.83, "learning_rate": 1.4345060467017358e-06, "loss": 0.1585, "step": 16674 }, { "epoch": 0.83, "learning_rate": 1.4336715730245288e-06, "loss": 0.3869, "step": 16675 }, { "epoch": 0.83, "learning_rate": 1.4328373233884229e-06, "loss": 0.2611, "step": 16676 }, { "epoch": 0.83, "learning_rate": 1.4320032978152343e-06, "loss": 0.324, "step": 16677 }, { "epoch": 0.83, "learning_rate": 1.4311694963267765e-06, "loss": 0.3917, "step": 16678 }, { "epoch": 0.83, "learning_rate": 1.4303359189448584e-06, "loss": 0.3059, "step": 16679 }, { "epoch": 0.83, "learning_rate": 1.4295025656912809e-06, "loss": 0.3332, "step": 16680 }, { "epoch": 0.83, "learning_rate": 1.4286694365878384e-06, "loss": 0.3557, "step": 16681 }, { "epoch": 0.83, "learning_rate": 1.4278365316563203e-06, "loss": 0.6071, "step": 16682 }, { "epoch": 0.83, "learning_rate": 1.427003850918508e-06, "loss": 0.3999, "step": 16683 }, { "epoch": 0.83, "learning_rate": 1.4261713943961842e-06, "loss": 0.405, "step": 16684 }, { "epoch": 0.83, "learning_rate": 1.4253391621111169e-06, "loss": 0.3224, "step": 16685 }, { "epoch": 0.83, "learning_rate": 1.424507154085073e-06, "loss": 0.3576, "step": 16686 }, { "epoch": 0.83, "learning_rate": 1.423675370339811e-06, "loss": 0.335, "step": 16687 }, { "epoch": 0.83, "learning_rate": 1.4228438108970888e-06, "loss": 0.2437, "step": 16688 }, { "epoch": 0.83, "learning_rate": 1.4220124757786523e-06, "loss": 0.2364, "step": 16689 }, { "epoch": 0.83, "learning_rate": 1.4211813650062444e-06, "loss": 0.3144, "step": 16690 }, { "epoch": 0.83, "learning_rate": 1.420350478601602e-06, "loss": 0.265, "step": 16691 }, { "epoch": 0.83, "learning_rate": 1.4195198165864532e-06, "loss": 0.3563, "step": 16692 }, { "epoch": 0.83, "learning_rate": 1.4186893789825274e-06, "loss": 0.3672, "step": 16693 }, { "epoch": 0.83, "learning_rate": 1.4178591658115414e-06, "loss": 0.4892, "step": 16694 }, { "epoch": 0.83, "learning_rate": 1.4170291770952082e-06, "loss": 0.4945, "step": 16695 }, { "epoch": 0.83, "learning_rate": 1.416199412855236e-06, "loss": 0.3192, "step": 16696 }, { "epoch": 0.83, "learning_rate": 1.4153698731133236e-06, "loss": 0.4311, "step": 16697 }, { "epoch": 0.83, "learning_rate": 1.4145405578911697e-06, "loss": 0.2944, "step": 16698 }, { "epoch": 0.83, "learning_rate": 1.4137114672104624e-06, "loss": 0.4534, "step": 16699 }, { "epoch": 0.83, "learning_rate": 1.4128826010928854e-06, "loss": 0.0611, "step": 16700 }, { "epoch": 0.83, "learning_rate": 1.4120539595601147e-06, "loss": 0.268, "step": 16701 }, { "epoch": 0.83, "learning_rate": 1.4112255426338272e-06, "loss": 0.4656, "step": 16702 }, { "epoch": 0.83, "learning_rate": 1.410397350335686e-06, "loss": 0.3882, "step": 16703 }, { "epoch": 0.83, "learning_rate": 1.4095693826873513e-06, "loss": 0.2491, "step": 16704 }, { "epoch": 0.83, "learning_rate": 1.408741639710477e-06, "loss": 0.5152, "step": 16705 }, { "epoch": 0.83, "learning_rate": 1.4079141214267134e-06, "loss": 0.2946, "step": 16706 }, { "epoch": 0.83, "learning_rate": 1.4070868278577032e-06, "loss": 0.4631, "step": 16707 }, { "epoch": 0.83, "learning_rate": 1.4062597590250815e-06, "loss": 0.4945, "step": 16708 }, { "epoch": 0.83, "learning_rate": 1.4054329149504808e-06, "loss": 0.2901, "step": 16709 }, { "epoch": 0.83, "learning_rate": 1.404606295655523e-06, "loss": 0.4504, "step": 16710 }, { "epoch": 0.83, "learning_rate": 1.4037799011618314e-06, "loss": 0.0837, "step": 16711 }, { "epoch": 0.83, "learning_rate": 1.402953731491018e-06, "loss": 0.4129, "step": 16712 }, { "epoch": 0.83, "learning_rate": 1.4021277866646887e-06, "loss": 0.1907, "step": 16713 }, { "epoch": 0.83, "learning_rate": 1.4013020667044452e-06, "loss": 0.3402, "step": 16714 }, { "epoch": 0.83, "learning_rate": 1.4004765716318847e-06, "loss": 0.2752, "step": 16715 }, { "epoch": 0.83, "learning_rate": 1.3996513014685974e-06, "loss": 0.2833, "step": 16716 }, { "epoch": 0.83, "learning_rate": 1.3988262562361655e-06, "loss": 0.4182, "step": 16717 }, { "epoch": 0.83, "learning_rate": 1.3980014359561666e-06, "loss": 0.34, "step": 16718 }, { "epoch": 0.83, "learning_rate": 1.397176840650174e-06, "loss": 0.4543, "step": 16719 }, { "epoch": 0.83, "learning_rate": 1.3963524703397534e-06, "loss": 0.1092, "step": 16720 }, { "epoch": 0.83, "learning_rate": 1.3955283250464656e-06, "loss": 0.4247, "step": 16721 }, { "epoch": 0.83, "learning_rate": 1.3947044047918645e-06, "loss": 0.0915, "step": 16722 }, { "epoch": 0.84, "learning_rate": 1.3938807095974982e-06, "loss": 0.2933, "step": 16723 }, { "epoch": 0.84, "learning_rate": 1.3930572394849084e-06, "loss": 0.5199, "step": 16724 }, { "epoch": 0.84, "learning_rate": 1.3922339944756347e-06, "loss": 0.4127, "step": 16725 }, { "epoch": 0.84, "learning_rate": 1.3914109745912075e-06, "loss": 0.64, "step": 16726 }, { "epoch": 0.84, "learning_rate": 1.3905881798531507e-06, "loss": 0.2583, "step": 16727 }, { "epoch": 0.84, "learning_rate": 1.3897656102829814e-06, "loss": 0.5867, "step": 16728 }, { "epoch": 0.84, "learning_rate": 1.3889432659022185e-06, "loss": 0.4499, "step": 16729 }, { "epoch": 0.84, "learning_rate": 1.3881211467323652e-06, "loss": 0.3809, "step": 16730 }, { "epoch": 0.84, "learning_rate": 1.3872992527949236e-06, "loss": 0.4701, "step": 16731 }, { "epoch": 0.84, "learning_rate": 1.3864775841113897e-06, "loss": 0.3222, "step": 16732 }, { "epoch": 0.84, "learning_rate": 1.3856561407032509e-06, "loss": 0.3769, "step": 16733 }, { "epoch": 0.84, "learning_rate": 1.3848349225919944e-06, "loss": 0.2304, "step": 16734 }, { "epoch": 0.84, "learning_rate": 1.3840139297990973e-06, "loss": 0.1676, "step": 16735 }, { "epoch": 0.84, "learning_rate": 1.3831931623460315e-06, "loss": 0.304, "step": 16736 }, { "epoch": 0.84, "learning_rate": 1.3823726202542597e-06, "loss": 0.3319, "step": 16737 }, { "epoch": 0.84, "learning_rate": 1.381552303545247e-06, "loss": 0.4092, "step": 16738 }, { "epoch": 0.84, "learning_rate": 1.3807322122404455e-06, "loss": 0.3134, "step": 16739 }, { "epoch": 0.84, "learning_rate": 1.379912346361304e-06, "loss": 0.3802, "step": 16740 }, { "epoch": 0.84, "learning_rate": 1.3790927059292658e-06, "loss": 0.3524, "step": 16741 }, { "epoch": 0.84, "learning_rate": 1.3782732909657625e-06, "loss": 0.3085, "step": 16742 }, { "epoch": 0.84, "learning_rate": 1.3774541014922326e-06, "loss": 0.5606, "step": 16743 }, { "epoch": 0.84, "learning_rate": 1.376635137530097e-06, "loss": 0.3419, "step": 16744 }, { "epoch": 0.84, "learning_rate": 1.3758163991007733e-06, "loss": 0.3423, "step": 16745 }, { "epoch": 0.84, "learning_rate": 1.3749978862256774e-06, "loss": 0.3934, "step": 16746 }, { "epoch": 0.84, "learning_rate": 1.3741795989262153e-06, "loss": 0.35, "step": 16747 }, { "epoch": 0.84, "learning_rate": 1.3733615372237873e-06, "loss": 0.3272, "step": 16748 }, { "epoch": 0.84, "learning_rate": 1.372543701139789e-06, "loss": 0.2631, "step": 16749 }, { "epoch": 0.84, "learning_rate": 1.371726090695611e-06, "loss": 0.2755, "step": 16750 }, { "epoch": 0.84, "learning_rate": 1.3709087059126335e-06, "loss": 0.4045, "step": 16751 }, { "epoch": 0.84, "learning_rate": 1.3700915468122388e-06, "loss": 0.1748, "step": 16752 }, { "epoch": 0.84, "learning_rate": 1.3692746134157976e-06, "loss": 0.2092, "step": 16753 }, { "epoch": 0.84, "learning_rate": 1.3684579057446734e-06, "loss": 0.5658, "step": 16754 }, { "epoch": 0.84, "learning_rate": 1.3676414238202284e-06, "loss": 0.2805, "step": 16755 }, { "epoch": 0.84, "learning_rate": 1.3668251676638122e-06, "loss": 0.4811, "step": 16756 }, { "epoch": 0.84, "learning_rate": 1.36600913729678e-06, "loss": 0.2324, "step": 16757 }, { "epoch": 0.84, "learning_rate": 1.3651933327404688e-06, "loss": 0.3279, "step": 16758 }, { "epoch": 0.84, "learning_rate": 1.3643777540162173e-06, "loss": 0.5108, "step": 16759 }, { "epoch": 0.84, "learning_rate": 1.363562401145353e-06, "loss": 0.2713, "step": 16760 }, { "epoch": 0.84, "learning_rate": 1.3627472741492043e-06, "loss": 0.3096, "step": 16761 }, { "epoch": 0.84, "learning_rate": 1.3619323730490875e-06, "loss": 0.1698, "step": 16762 }, { "epoch": 0.84, "learning_rate": 1.3611176978663166e-06, "loss": 0.3432, "step": 16763 }, { "epoch": 0.84, "learning_rate": 1.3603032486221968e-06, "loss": 0.2942, "step": 16764 }, { "epoch": 0.84, "learning_rate": 1.3594890253380267e-06, "loss": 0.471, "step": 16765 }, { "epoch": 0.84, "learning_rate": 1.3586750280351068e-06, "loss": 0.335, "step": 16766 }, { "epoch": 0.84, "learning_rate": 1.3578612567347238e-06, "loss": 0.2244, "step": 16767 }, { "epoch": 0.84, "learning_rate": 1.3570477114581604e-06, "loss": 0.3251, "step": 16768 }, { "epoch": 0.84, "learning_rate": 1.356234392226694e-06, "loss": 0.249, "step": 16769 }, { "epoch": 0.84, "learning_rate": 1.3554212990615957e-06, "loss": 0.2926, "step": 16770 }, { "epoch": 0.84, "learning_rate": 1.3546084319841302e-06, "loss": 0.305, "step": 16771 }, { "epoch": 0.84, "learning_rate": 1.3537957910155587e-06, "loss": 0.1641, "step": 16772 }, { "epoch": 0.84, "learning_rate": 1.352983376177134e-06, "loss": 0.4322, "step": 16773 }, { "epoch": 0.84, "learning_rate": 1.3521711874901e-06, "loss": 0.2536, "step": 16774 }, { "epoch": 0.84, "learning_rate": 1.3513592249757047e-06, "loss": 0.2334, "step": 16775 }, { "epoch": 0.84, "learning_rate": 1.3505474886551816e-06, "loss": 0.5614, "step": 16776 }, { "epoch": 0.84, "learning_rate": 1.34973597854976e-06, "loss": 0.4406, "step": 16777 }, { "epoch": 0.84, "learning_rate": 1.3489246946806622e-06, "loss": 0.5802, "step": 16778 }, { "epoch": 0.84, "learning_rate": 1.3481136370691095e-06, "loss": 0.2717, "step": 16779 }, { "epoch": 0.84, "learning_rate": 1.3473028057363136e-06, "loss": 0.4341, "step": 16780 }, { "epoch": 0.84, "learning_rate": 1.3464922007034787e-06, "loss": 0.3086, "step": 16781 }, { "epoch": 0.84, "learning_rate": 1.3456818219918077e-06, "loss": 0.2275, "step": 16782 }, { "epoch": 0.84, "learning_rate": 1.3448716696224905e-06, "loss": 0.1242, "step": 16783 }, { "epoch": 0.84, "learning_rate": 1.3440617436167202e-06, "loss": 0.4462, "step": 16784 }, { "epoch": 0.84, "learning_rate": 1.3432520439956787e-06, "loss": 0.5248, "step": 16785 }, { "epoch": 0.84, "learning_rate": 1.3424425707805412e-06, "loss": 0.4056, "step": 16786 }, { "epoch": 0.84, "learning_rate": 1.3416333239924794e-06, "loss": 0.4075, "step": 16787 }, { "epoch": 0.84, "learning_rate": 1.3408243036526547e-06, "loss": 0.5307, "step": 16788 }, { "epoch": 0.84, "learning_rate": 1.340015509782232e-06, "loss": 0.335, "step": 16789 }, { "epoch": 0.84, "learning_rate": 1.3392069424023601e-06, "loss": 0.6, "step": 16790 }, { "epoch": 0.84, "learning_rate": 1.3383986015341878e-06, "loss": 0.4007, "step": 16791 }, { "epoch": 0.84, "learning_rate": 1.3375904871988554e-06, "loss": 0.552, "step": 16792 }, { "epoch": 0.84, "learning_rate": 1.3367825994174976e-06, "loss": 0.3555, "step": 16793 }, { "epoch": 0.84, "learning_rate": 1.3359749382112442e-06, "loss": 0.3349, "step": 16794 }, { "epoch": 0.84, "learning_rate": 1.335167503601219e-06, "loss": 0.592, "step": 16795 }, { "epoch": 0.84, "learning_rate": 1.3343602956085376e-06, "loss": 0.3063, "step": 16796 }, { "epoch": 0.84, "learning_rate": 1.3335533142543111e-06, "loss": 0.1451, "step": 16797 }, { "epoch": 0.84, "learning_rate": 1.33274655955965e-06, "loss": 0.2778, "step": 16798 }, { "epoch": 0.84, "learning_rate": 1.3319400315456488e-06, "loss": 0.3568, "step": 16799 }, { "epoch": 0.84, "learning_rate": 1.3311337302334048e-06, "loss": 0.6156, "step": 16800 }, { "epoch": 0.84, "learning_rate": 1.3303276556440003e-06, "loss": 0.3327, "step": 16801 }, { "epoch": 0.84, "learning_rate": 1.3295218077985229e-06, "loss": 0.5556, "step": 16802 }, { "epoch": 0.84, "learning_rate": 1.3287161867180475e-06, "loss": 0.5242, "step": 16803 }, { "epoch": 0.84, "learning_rate": 1.3279107924236423e-06, "loss": 0.2861, "step": 16804 }, { "epoch": 0.84, "learning_rate": 1.3271056249363712e-06, "loss": 0.2974, "step": 16805 }, { "epoch": 0.84, "learning_rate": 1.326300684277292e-06, "loss": 0.2532, "step": 16806 }, { "epoch": 0.84, "learning_rate": 1.3254959704674597e-06, "loss": 0.4737, "step": 16807 }, { "epoch": 0.84, "learning_rate": 1.3246914835279179e-06, "loss": 0.1781, "step": 16808 }, { "epoch": 0.84, "learning_rate": 1.3238872234797085e-06, "loss": 0.3498, "step": 16809 }, { "epoch": 0.84, "learning_rate": 1.3230831903438635e-06, "loss": 0.5568, "step": 16810 }, { "epoch": 0.84, "learning_rate": 1.3222793841414138e-06, "loss": 0.1342, "step": 16811 }, { "epoch": 0.84, "learning_rate": 1.3214758048933818e-06, "loss": 0.2229, "step": 16812 }, { "epoch": 0.84, "learning_rate": 1.3206724526207836e-06, "loss": 0.3783, "step": 16813 }, { "epoch": 0.84, "learning_rate": 1.3198693273446283e-06, "loss": 0.3789, "step": 16814 }, { "epoch": 0.84, "learning_rate": 1.3190664290859212e-06, "loss": 0.2425, "step": 16815 }, { "epoch": 0.84, "learning_rate": 1.318263757865663e-06, "loss": 0.2506, "step": 16816 }, { "epoch": 0.84, "learning_rate": 1.3174613137048454e-06, "loss": 0.2333, "step": 16817 }, { "epoch": 0.84, "learning_rate": 1.3166590966244552e-06, "loss": 0.4442, "step": 16818 }, { "epoch": 0.84, "learning_rate": 1.3158571066454728e-06, "loss": 0.472, "step": 16819 }, { "epoch": 0.84, "learning_rate": 1.3150553437888736e-06, "loss": 0.3584, "step": 16820 }, { "epoch": 0.84, "learning_rate": 1.3142538080756263e-06, "loss": 0.5429, "step": 16821 }, { "epoch": 0.84, "learning_rate": 1.313452499526695e-06, "loss": 0.3936, "step": 16822 }, { "epoch": 0.84, "learning_rate": 1.3126514181630368e-06, "loss": 0.3175, "step": 16823 }, { "epoch": 0.84, "learning_rate": 1.311850564005599e-06, "loss": 0.4507, "step": 16824 }, { "epoch": 0.84, "learning_rate": 1.3110499370753326e-06, "loss": 0.4392, "step": 16825 }, { "epoch": 0.84, "learning_rate": 1.3102495373931755e-06, "loss": 0.457, "step": 16826 }, { "epoch": 0.84, "learning_rate": 1.3094493649800589e-06, "loss": 0.3635, "step": 16827 }, { "epoch": 0.84, "learning_rate": 1.3086494198569132e-06, "loss": 0.4225, "step": 16828 }, { "epoch": 0.84, "learning_rate": 1.3078497020446546e-06, "loss": 0.2765, "step": 16829 }, { "epoch": 0.84, "learning_rate": 1.307050211564206e-06, "loss": 0.3282, "step": 16830 }, { "epoch": 0.84, "learning_rate": 1.3062509484364717e-06, "loss": 0.3309, "step": 16831 }, { "epoch": 0.84, "learning_rate": 1.3054519126823584e-06, "loss": 0.365, "step": 16832 }, { "epoch": 0.84, "learning_rate": 1.3046531043227595e-06, "loss": 0.3428, "step": 16833 }, { "epoch": 0.84, "learning_rate": 1.3038545233785727e-06, "loss": 0.3941, "step": 16834 }, { "epoch": 0.84, "learning_rate": 1.3030561698706813e-06, "loss": 0.251, "step": 16835 }, { "epoch": 0.84, "learning_rate": 1.3022580438199638e-06, "loss": 0.3049, "step": 16836 }, { "epoch": 0.84, "learning_rate": 1.3014601452472952e-06, "loss": 0.3408, "step": 16837 }, { "epoch": 0.84, "learning_rate": 1.300662474173542e-06, "loss": 0.3778, "step": 16838 }, { "epoch": 0.84, "learning_rate": 1.2998650306195693e-06, "loss": 0.2672, "step": 16839 }, { "epoch": 0.84, "learning_rate": 1.299067814606232e-06, "loss": 0.5756, "step": 16840 }, { "epoch": 0.84, "learning_rate": 1.2982708261543798e-06, "loss": 0.5441, "step": 16841 }, { "epoch": 0.84, "learning_rate": 1.297474065284856e-06, "loss": 0.2179, "step": 16842 }, { "epoch": 0.84, "learning_rate": 1.296677532018501e-06, "loss": 0.4975, "step": 16843 }, { "epoch": 0.84, "learning_rate": 1.2958812263761456e-06, "loss": 0.2039, "step": 16844 }, { "epoch": 0.84, "learning_rate": 1.2950851483786163e-06, "loss": 0.2575, "step": 16845 }, { "epoch": 0.84, "learning_rate": 1.2942892980467325e-06, "loss": 0.2256, "step": 16846 }, { "epoch": 0.84, "learning_rate": 1.2934936754013094e-06, "loss": 0.5125, "step": 16847 }, { "epoch": 0.84, "learning_rate": 1.292698280463156e-06, "loss": 0.6011, "step": 16848 }, { "epoch": 0.84, "learning_rate": 1.2919031132530757e-06, "loss": 0.1681, "step": 16849 }, { "epoch": 0.84, "learning_rate": 1.2911081737918641e-06, "loss": 0.2264, "step": 16850 }, { "epoch": 0.84, "learning_rate": 1.290313462100309e-06, "loss": 0.5227, "step": 16851 }, { "epoch": 0.84, "learning_rate": 1.2895189781992e-06, "loss": 0.1752, "step": 16852 }, { "epoch": 0.84, "learning_rate": 1.2887247221093136e-06, "loss": 0.2538, "step": 16853 }, { "epoch": 0.84, "learning_rate": 1.2879306938514213e-06, "loss": 0.1814, "step": 16854 }, { "epoch": 0.84, "learning_rate": 1.2871368934462924e-06, "loss": 0.4414, "step": 16855 }, { "epoch": 0.84, "learning_rate": 1.2863433209146826e-06, "loss": 0.4791, "step": 16856 }, { "epoch": 0.84, "learning_rate": 1.2855499762773538e-06, "loss": 0.5168, "step": 16857 }, { "epoch": 0.84, "learning_rate": 1.2847568595550508e-06, "loss": 0.3637, "step": 16858 }, { "epoch": 0.84, "learning_rate": 1.2839639707685181e-06, "loss": 0.4593, "step": 16859 }, { "epoch": 0.84, "learning_rate": 1.2831713099384912e-06, "loss": 0.3127, "step": 16860 }, { "epoch": 0.84, "learning_rate": 1.2823788770856993e-06, "loss": 0.2043, "step": 16861 }, { "epoch": 0.84, "learning_rate": 1.2815866722308723e-06, "loss": 0.1453, "step": 16862 }, { "epoch": 0.84, "learning_rate": 1.280794695394727e-06, "loss": 0.1909, "step": 16863 }, { "epoch": 0.84, "learning_rate": 1.2800029465979757e-06, "loss": 0.4149, "step": 16864 }, { "epoch": 0.84, "learning_rate": 1.2792114258613264e-06, "loss": 0.2923, "step": 16865 }, { "epoch": 0.84, "learning_rate": 1.2784201332054802e-06, "loss": 0.5668, "step": 16866 }, { "epoch": 0.84, "learning_rate": 1.2776290686511327e-06, "loss": 0.2627, "step": 16867 }, { "epoch": 0.84, "learning_rate": 1.276838232218972e-06, "loss": 0.5099, "step": 16868 }, { "epoch": 0.84, "learning_rate": 1.2760476239296826e-06, "loss": 0.6301, "step": 16869 }, { "epoch": 0.84, "learning_rate": 1.275257243803938e-06, "loss": 0.1046, "step": 16870 }, { "epoch": 0.84, "learning_rate": 1.274467091862416e-06, "loss": 0.3173, "step": 16871 }, { "epoch": 0.84, "learning_rate": 1.2736771681257786e-06, "loss": 0.182, "step": 16872 }, { "epoch": 0.84, "learning_rate": 1.2728874726146857e-06, "loss": 0.3951, "step": 16873 }, { "epoch": 0.84, "learning_rate": 1.2720980053497877e-06, "loss": 0.3591, "step": 16874 }, { "epoch": 0.84, "learning_rate": 1.2713087663517377e-06, "loss": 0.2756, "step": 16875 }, { "epoch": 0.84, "learning_rate": 1.2705197556411742e-06, "loss": 0.3157, "step": 16876 }, { "epoch": 0.84, "learning_rate": 1.269730973238732e-06, "loss": 0.2922, "step": 16877 }, { "epoch": 0.84, "learning_rate": 1.2689424191650423e-06, "loss": 0.3854, "step": 16878 }, { "epoch": 0.84, "learning_rate": 1.2681540934407265e-06, "loss": 0.3344, "step": 16879 }, { "epoch": 0.84, "learning_rate": 1.2673659960864048e-06, "loss": 0.2909, "step": 16880 }, { "epoch": 0.84, "learning_rate": 1.2665781271226885e-06, "loss": 0.3201, "step": 16881 }, { "epoch": 0.84, "learning_rate": 1.265790486570182e-06, "loss": 0.4489, "step": 16882 }, { "epoch": 0.84, "learning_rate": 1.2650030744494845e-06, "loss": 0.2393, "step": 16883 }, { "epoch": 0.84, "learning_rate": 1.264215890781192e-06, "loss": 0.5705, "step": 16884 }, { "epoch": 0.84, "learning_rate": 1.2634289355858909e-06, "loss": 0.6087, "step": 16885 }, { "epoch": 0.84, "learning_rate": 1.262642208884164e-06, "loss": 0.4123, "step": 16886 }, { "epoch": 0.84, "learning_rate": 1.2618557106965857e-06, "loss": 0.56, "step": 16887 }, { "epoch": 0.84, "learning_rate": 1.2610694410437241e-06, "loss": 0.6077, "step": 16888 }, { "epoch": 0.84, "learning_rate": 1.2602833999461473e-06, "loss": 0.3997, "step": 16889 }, { "epoch": 0.84, "learning_rate": 1.259497587424412e-06, "loss": 0.4497, "step": 16890 }, { "epoch": 0.84, "learning_rate": 1.2587120034990686e-06, "loss": 0.4243, "step": 16891 }, { "epoch": 0.84, "learning_rate": 1.2579266481906639e-06, "loss": 0.4734, "step": 16892 }, { "epoch": 0.84, "learning_rate": 1.257141521519738e-06, "loss": 0.4662, "step": 16893 }, { "epoch": 0.84, "learning_rate": 1.2563566235068237e-06, "loss": 0.2121, "step": 16894 }, { "epoch": 0.84, "learning_rate": 1.2555719541724498e-06, "loss": 0.2065, "step": 16895 }, { "epoch": 0.84, "learning_rate": 1.254787513537139e-06, "loss": 0.2846, "step": 16896 }, { "epoch": 0.84, "learning_rate": 1.2540033016214047e-06, "loss": 0.3941, "step": 16897 }, { "epoch": 0.84, "learning_rate": 1.2532193184457607e-06, "loss": 0.1922, "step": 16898 }, { "epoch": 0.84, "learning_rate": 1.2524355640307085e-06, "loss": 0.3323, "step": 16899 }, { "epoch": 0.84, "learning_rate": 1.2516520383967478e-06, "loss": 0.4805, "step": 16900 }, { "epoch": 0.84, "learning_rate": 1.2508687415643693e-06, "loss": 0.34, "step": 16901 }, { "epoch": 0.84, "learning_rate": 1.250085673554058e-06, "loss": 0.5434, "step": 16902 }, { "epoch": 0.84, "learning_rate": 1.2493028343862967e-06, "loss": 0.3386, "step": 16903 }, { "epoch": 0.84, "learning_rate": 1.2485202240815586e-06, "loss": 0.4852, "step": 16904 }, { "epoch": 0.84, "learning_rate": 1.2477378426603126e-06, "loss": 0.1806, "step": 16905 }, { "epoch": 0.84, "learning_rate": 1.246955690143017e-06, "loss": 0.2126, "step": 16906 }, { "epoch": 0.84, "learning_rate": 1.2461737665501339e-06, "loss": 0.3665, "step": 16907 }, { "epoch": 0.84, "learning_rate": 1.2453920719021095e-06, "loss": 0.2149, "step": 16908 }, { "epoch": 0.84, "learning_rate": 1.2446106062193886e-06, "loss": 0.4227, "step": 16909 }, { "epoch": 0.84, "learning_rate": 1.2438293695224114e-06, "loss": 0.5446, "step": 16910 }, { "epoch": 0.84, "learning_rate": 1.2430483618316047e-06, "loss": 0.1607, "step": 16911 }, { "epoch": 0.84, "learning_rate": 1.2422675831674014e-06, "loss": 0.3559, "step": 16912 }, { "epoch": 0.84, "learning_rate": 1.2414870335502194e-06, "loss": 0.2096, "step": 16913 }, { "epoch": 0.84, "learning_rate": 1.2407067130004724e-06, "loss": 0.5334, "step": 16914 }, { "epoch": 0.84, "learning_rate": 1.2399266215385676e-06, "loss": 0.2284, "step": 16915 }, { "epoch": 0.84, "learning_rate": 1.2391467591849094e-06, "loss": 0.3774, "step": 16916 }, { "epoch": 0.84, "learning_rate": 1.2383671259598928e-06, "loss": 0.4579, "step": 16917 }, { "epoch": 0.84, "learning_rate": 1.2375877218839084e-06, "loss": 0.4248, "step": 16918 }, { "epoch": 0.84, "learning_rate": 1.2368085469773405e-06, "loss": 0.3783, "step": 16919 }, { "epoch": 0.84, "learning_rate": 1.2360296012605643e-06, "loss": 0.5011, "step": 16920 }, { "epoch": 0.84, "learning_rate": 1.235250884753958e-06, "loss": 0.313, "step": 16921 }, { "epoch": 0.84, "learning_rate": 1.234472397477885e-06, "loss": 0.3724, "step": 16922 }, { "epoch": 0.85, "learning_rate": 1.2336941394527058e-06, "loss": 0.2346, "step": 16923 }, { "epoch": 0.85, "learning_rate": 1.2329161106987752e-06, "loss": 0.3001, "step": 16924 }, { "epoch": 0.85, "learning_rate": 1.2321383112364382e-06, "loss": 0.2918, "step": 16925 }, { "epoch": 0.85, "learning_rate": 1.2313607410860418e-06, "loss": 0.5063, "step": 16926 }, { "epoch": 0.85, "learning_rate": 1.2305834002679195e-06, "loss": 0.467, "step": 16927 }, { "epoch": 0.85, "learning_rate": 1.2298062888024043e-06, "loss": 0.3047, "step": 16928 }, { "epoch": 0.85, "learning_rate": 1.2290294067098153e-06, "loss": 0.2888, "step": 16929 }, { "epoch": 0.85, "learning_rate": 1.2282527540104773e-06, "loss": 0.1789, "step": 16930 }, { "epoch": 0.85, "learning_rate": 1.2274763307246983e-06, "loss": 0.2444, "step": 16931 }, { "epoch": 0.85, "learning_rate": 1.226700136872787e-06, "loss": 0.3287, "step": 16932 }, { "epoch": 0.85, "learning_rate": 1.2259241724750425e-06, "loss": 0.3789, "step": 16933 }, { "epoch": 0.85, "learning_rate": 1.2251484375517564e-06, "loss": 0.1902, "step": 16934 }, { "epoch": 0.85, "learning_rate": 1.2243729321232233e-06, "loss": 0.2675, "step": 16935 }, { "epoch": 0.85, "learning_rate": 1.2235976562097218e-06, "loss": 0.3334, "step": 16936 }, { "epoch": 0.85, "learning_rate": 1.2228226098315277e-06, "loss": 0.2071, "step": 16937 }, { "epoch": 0.85, "learning_rate": 1.2220477930089137e-06, "loss": 0.4006, "step": 16938 }, { "epoch": 0.85, "learning_rate": 1.2212732057621414e-06, "loss": 0.3279, "step": 16939 }, { "epoch": 0.85, "learning_rate": 1.220498848111471e-06, "loss": 0.5242, "step": 16940 }, { "epoch": 0.85, "learning_rate": 1.2197247200771533e-06, "loss": 0.3759, "step": 16941 }, { "epoch": 0.85, "learning_rate": 1.2189508216794343e-06, "loss": 0.2506, "step": 16942 }, { "epoch": 0.85, "learning_rate": 1.2181771529385567e-06, "loss": 0.4662, "step": 16943 }, { "epoch": 0.85, "learning_rate": 1.217403713874754e-06, "loss": 0.2722, "step": 16944 }, { "epoch": 0.85, "learning_rate": 1.216630504508254e-06, "loss": 0.2159, "step": 16945 }, { "epoch": 0.85, "learning_rate": 1.2158575248592785e-06, "loss": 0.3663, "step": 16946 }, { "epoch": 0.85, "learning_rate": 1.2150847749480432e-06, "loss": 0.3208, "step": 16947 }, { "epoch": 0.85, "learning_rate": 1.2143122547947606e-06, "loss": 0.5286, "step": 16948 }, { "epoch": 0.85, "learning_rate": 1.2135399644196343e-06, "loss": 0.2951, "step": 16949 }, { "epoch": 0.85, "learning_rate": 1.212767903842862e-06, "loss": 0.4456, "step": 16950 }, { "epoch": 0.85, "learning_rate": 1.2119960730846358e-06, "loss": 0.295, "step": 16951 }, { "epoch": 0.85, "learning_rate": 1.2112244721651411e-06, "loss": 0.4589, "step": 16952 }, { "epoch": 0.85, "learning_rate": 1.2104531011045596e-06, "loss": 0.4717, "step": 16953 }, { "epoch": 0.85, "learning_rate": 1.2096819599230657e-06, "loss": 0.3875, "step": 16954 }, { "epoch": 0.85, "learning_rate": 1.2089110486408273e-06, "loss": 0.4525, "step": 16955 }, { "epoch": 0.85, "learning_rate": 1.2081403672780057e-06, "loss": 0.3126, "step": 16956 }, { "epoch": 0.85, "learning_rate": 1.2073699158547548e-06, "loss": 0.4125, "step": 16957 }, { "epoch": 0.85, "learning_rate": 1.2065996943912305e-06, "loss": 0.2164, "step": 16958 }, { "epoch": 0.85, "learning_rate": 1.205829702907574e-06, "loss": 0.4801, "step": 16959 }, { "epoch": 0.85, "learning_rate": 1.2050599414239228e-06, "loss": 0.3042, "step": 16960 }, { "epoch": 0.85, "learning_rate": 1.204290409960407e-06, "loss": 0.4447, "step": 16961 }, { "epoch": 0.85, "learning_rate": 1.203521108537158e-06, "loss": 0.246, "step": 16962 }, { "epoch": 0.85, "learning_rate": 1.2027520371742928e-06, "loss": 0.3394, "step": 16963 }, { "epoch": 0.85, "learning_rate": 1.201983195891926e-06, "loss": 0.4361, "step": 16964 }, { "epoch": 0.85, "learning_rate": 1.2012145847101653e-06, "loss": 0.2272, "step": 16965 }, { "epoch": 0.85, "learning_rate": 1.2004462036491127e-06, "loss": 0.3286, "step": 16966 }, { "epoch": 0.85, "learning_rate": 1.1996780527288644e-06, "loss": 0.4364, "step": 16967 }, { "epoch": 0.85, "learning_rate": 1.1989101319695106e-06, "loss": 0.4907, "step": 16968 }, { "epoch": 0.85, "learning_rate": 1.1981424413911347e-06, "loss": 0.4061, "step": 16969 }, { "epoch": 0.85, "learning_rate": 1.1973749810138125e-06, "loss": 0.1418, "step": 16970 }, { "epoch": 0.85, "learning_rate": 1.19660775085762e-06, "loss": 0.3436, "step": 16971 }, { "epoch": 0.85, "learning_rate": 1.1958407509426228e-06, "loss": 0.4411, "step": 16972 }, { "epoch": 0.85, "learning_rate": 1.1950739812888778e-06, "loss": 0.4048, "step": 16973 }, { "epoch": 0.85, "learning_rate": 1.1943074419164413e-06, "loss": 0.2962, "step": 16974 }, { "epoch": 0.85, "learning_rate": 1.1935411328453584e-06, "loss": 0.342, "step": 16975 }, { "epoch": 0.85, "learning_rate": 1.192775054095675e-06, "loss": 0.2932, "step": 16976 }, { "epoch": 0.85, "learning_rate": 1.1920092056874244e-06, "loss": 0.2594, "step": 16977 }, { "epoch": 0.85, "learning_rate": 1.191243587640636e-06, "loss": 0.3703, "step": 16978 }, { "epoch": 0.85, "learning_rate": 1.1904781999753323e-06, "loss": 0.4306, "step": 16979 }, { "epoch": 0.85, "learning_rate": 1.1897130427115355e-06, "loss": 0.2692, "step": 16980 }, { "epoch": 0.85, "learning_rate": 1.188948115869254e-06, "loss": 0.2789, "step": 16981 }, { "epoch": 0.85, "learning_rate": 1.188183419468495e-06, "loss": 0.2554, "step": 16982 }, { "epoch": 0.85, "learning_rate": 1.1874189535292568e-06, "loss": 0.5756, "step": 16983 }, { "epoch": 0.85, "learning_rate": 1.1866547180715316e-06, "loss": 0.4346, "step": 16984 }, { "epoch": 0.85, "learning_rate": 1.1858907131153096e-06, "loss": 0.3993, "step": 16985 }, { "epoch": 0.85, "learning_rate": 1.185126938680573e-06, "loss": 0.3754, "step": 16986 }, { "epoch": 0.85, "learning_rate": 1.1843633947872958e-06, "loss": 0.2436, "step": 16987 }, { "epoch": 0.85, "learning_rate": 1.183600081455447e-06, "loss": 0.4057, "step": 16988 }, { "epoch": 0.85, "learning_rate": 1.1828369987049915e-06, "loss": 0.2762, "step": 16989 }, { "epoch": 0.85, "learning_rate": 1.182074146555885e-06, "loss": 0.431, "step": 16990 }, { "epoch": 0.85, "learning_rate": 1.1813115250280804e-06, "loss": 0.4818, "step": 16991 }, { "epoch": 0.85, "learning_rate": 1.1805491341415222e-06, "loss": 0.3072, "step": 16992 }, { "epoch": 0.85, "learning_rate": 1.1797869739161483e-06, "loss": 0.4712, "step": 16993 }, { "epoch": 0.85, "learning_rate": 1.1790250443718954e-06, "loss": 0.4266, "step": 16994 }, { "epoch": 0.85, "learning_rate": 1.1782633455286895e-06, "loss": 0.3504, "step": 16995 }, { "epoch": 0.85, "learning_rate": 1.1775018774064505e-06, "loss": 0.3168, "step": 16996 }, { "epoch": 0.85, "learning_rate": 1.1767406400250948e-06, "loss": 0.457, "step": 16997 }, { "epoch": 0.85, "learning_rate": 1.1759796334045293e-06, "loss": 0.3295, "step": 16998 }, { "epoch": 0.85, "learning_rate": 1.1752188575646617e-06, "loss": 0.588, "step": 16999 }, { "epoch": 0.85, "learning_rate": 1.174458312525386e-06, "loss": 0.2514, "step": 17000 }, { "epoch": 0.85, "learning_rate": 1.1736979983065944e-06, "loss": 0.303, "step": 17001 }, { "epoch": 0.85, "learning_rate": 1.1729379149281682e-06, "loss": 0.1961, "step": 17002 }, { "epoch": 0.85, "learning_rate": 1.1721780624099921e-06, "loss": 0.4267, "step": 17003 }, { "epoch": 0.85, "learning_rate": 1.1714184407719364e-06, "loss": 0.2357, "step": 17004 }, { "epoch": 0.85, "learning_rate": 1.1706590500338677e-06, "loss": 0.49, "step": 17005 }, { "epoch": 0.85, "learning_rate": 1.1698998902156466e-06, "loss": 0.2288, "step": 17006 }, { "epoch": 0.85, "learning_rate": 1.1691409613371274e-06, "loss": 0.622, "step": 17007 }, { "epoch": 0.85, "learning_rate": 1.1683822634181618e-06, "loss": 0.4433, "step": 17008 }, { "epoch": 0.85, "learning_rate": 1.1676237964785907e-06, "loss": 0.274, "step": 17009 }, { "epoch": 0.85, "learning_rate": 1.1668655605382518e-06, "loss": 0.5203, "step": 17010 }, { "epoch": 0.85, "learning_rate": 1.1661075556169743e-06, "loss": 0.2742, "step": 17011 }, { "epoch": 0.85, "learning_rate": 1.1653497817345827e-06, "loss": 0.2029, "step": 17012 }, { "epoch": 0.85, "learning_rate": 1.1645922389108965e-06, "loss": 0.4327, "step": 17013 }, { "epoch": 0.85, "learning_rate": 1.1638349271657268e-06, "loss": 0.2045, "step": 17014 }, { "epoch": 0.85, "learning_rate": 1.1630778465188818e-06, "loss": 0.6032, "step": 17015 }, { "epoch": 0.85, "learning_rate": 1.1623209969901627e-06, "loss": 0.3918, "step": 17016 }, { "epoch": 0.85, "learning_rate": 1.1615643785993624e-06, "loss": 0.1965, "step": 17017 }, { "epoch": 0.85, "learning_rate": 1.1608079913662696e-06, "loss": 0.2889, "step": 17018 }, { "epoch": 0.85, "learning_rate": 1.160051835310666e-06, "loss": 0.1914, "step": 17019 }, { "epoch": 0.85, "learning_rate": 1.159295910452326e-06, "loss": 0.3363, "step": 17020 }, { "epoch": 0.85, "learning_rate": 1.158540216811025e-06, "loss": 0.3399, "step": 17021 }, { "epoch": 0.85, "learning_rate": 1.1577847544065247e-06, "loss": 0.2023, "step": 17022 }, { "epoch": 0.85, "learning_rate": 1.1570295232585826e-06, "loss": 0.6331, "step": 17023 }, { "epoch": 0.85, "learning_rate": 1.156274523386951e-06, "loss": 0.2347, "step": 17024 }, { "epoch": 0.85, "learning_rate": 1.155519754811374e-06, "loss": 0.3388, "step": 17025 }, { "epoch": 0.85, "learning_rate": 1.1547652175515956e-06, "loss": 0.3432, "step": 17026 }, { "epoch": 0.85, "learning_rate": 1.154010911627348e-06, "loss": 0.4293, "step": 17027 }, { "epoch": 0.85, "learning_rate": 1.1532568370583585e-06, "loss": 0.2846, "step": 17028 }, { "epoch": 0.85, "learning_rate": 1.1525029938643494e-06, "loss": 0.3561, "step": 17029 }, { "epoch": 0.85, "learning_rate": 1.1517493820650338e-06, "loss": 0.1456, "step": 17030 }, { "epoch": 0.85, "learning_rate": 1.1509960016801269e-06, "loss": 0.3482, "step": 17031 }, { "epoch": 0.85, "learning_rate": 1.1502428527293297e-06, "loss": 0.3523, "step": 17032 }, { "epoch": 0.85, "learning_rate": 1.1494899352323386e-06, "loss": 0.3021, "step": 17033 }, { "epoch": 0.85, "learning_rate": 1.1487372492088445e-06, "loss": 0.5112, "step": 17034 }, { "epoch": 0.85, "learning_rate": 1.147984794678536e-06, "loss": 0.4218, "step": 17035 }, { "epoch": 0.85, "learning_rate": 1.1472325716610922e-06, "loss": 0.255, "step": 17036 }, { "epoch": 0.85, "learning_rate": 1.1464805801761858e-06, "loss": 0.2563, "step": 17037 }, { "epoch": 0.85, "learning_rate": 1.1457288202434824e-06, "loss": 0.2471, "step": 17038 }, { "epoch": 0.85, "learning_rate": 1.1449772918826451e-06, "loss": 0.6494, "step": 17039 }, { "epoch": 0.85, "learning_rate": 1.1442259951133294e-06, "loss": 0.2656, "step": 17040 }, { "epoch": 0.85, "learning_rate": 1.143474929955184e-06, "loss": 0.3739, "step": 17041 }, { "epoch": 0.85, "learning_rate": 1.1427240964278508e-06, "loss": 0.2426, "step": 17042 }, { "epoch": 0.85, "learning_rate": 1.1419734945509664e-06, "loss": 0.378, "step": 17043 }, { "epoch": 0.85, "learning_rate": 1.1412231243441652e-06, "loss": 0.2366, "step": 17044 }, { "epoch": 0.85, "learning_rate": 1.1404729858270702e-06, "loss": 0.0967, "step": 17045 }, { "epoch": 0.85, "learning_rate": 1.1397230790193003e-06, "loss": 0.2561, "step": 17046 }, { "epoch": 0.85, "learning_rate": 1.1389734039404687e-06, "loss": 0.5105, "step": 17047 }, { "epoch": 0.85, "learning_rate": 1.1382239606101797e-06, "loss": 0.3329, "step": 17048 }, { "epoch": 0.85, "learning_rate": 1.1374747490480387e-06, "loss": 0.2405, "step": 17049 }, { "epoch": 0.85, "learning_rate": 1.1367257692736366e-06, "loss": 0.2288, "step": 17050 }, { "epoch": 0.85, "learning_rate": 1.1359770213065647e-06, "loss": 0.2192, "step": 17051 }, { "epoch": 0.85, "learning_rate": 1.1352285051664014e-06, "loss": 0.2586, "step": 17052 }, { "epoch": 0.85, "learning_rate": 1.1344802208727268e-06, "loss": 0.2277, "step": 17053 }, { "epoch": 0.85, "learning_rate": 1.1337321684451118e-06, "loss": 0.3192, "step": 17054 }, { "epoch": 0.85, "learning_rate": 1.1329843479031178e-06, "loss": 0.2593, "step": 17055 }, { "epoch": 0.85, "learning_rate": 1.1322367592663052e-06, "loss": 0.4059, "step": 17056 }, { "epoch": 0.85, "learning_rate": 1.1314894025542234e-06, "loss": 0.2732, "step": 17057 }, { "epoch": 0.85, "learning_rate": 1.130742277786423e-06, "loss": 0.4721, "step": 17058 }, { "epoch": 0.85, "learning_rate": 1.129995384982442e-06, "loss": 0.3439, "step": 17059 }, { "epoch": 0.85, "learning_rate": 1.1292487241618132e-06, "loss": 0.4902, "step": 17060 }, { "epoch": 0.85, "learning_rate": 1.128502295344066e-06, "loss": 0.2841, "step": 17061 }, { "epoch": 0.85, "learning_rate": 1.1277560985487213e-06, "loss": 0.4827, "step": 17062 }, { "epoch": 0.85, "learning_rate": 1.1270101337952955e-06, "loss": 0.2766, "step": 17063 }, { "epoch": 0.85, "learning_rate": 1.1262644011032974e-06, "loss": 0.3615, "step": 17064 }, { "epoch": 0.85, "learning_rate": 1.1255189004922318e-06, "loss": 0.2351, "step": 17065 }, { "epoch": 0.85, "learning_rate": 1.124773631981595e-06, "loss": 0.4327, "step": 17066 }, { "epoch": 0.85, "learning_rate": 1.1240285955908803e-06, "loss": 0.4567, "step": 17067 }, { "epoch": 0.85, "learning_rate": 1.1232837913395722e-06, "loss": 0.4427, "step": 17068 }, { "epoch": 0.85, "learning_rate": 1.1225392192471508e-06, "loss": 0.4716, "step": 17069 }, { "epoch": 0.85, "learning_rate": 1.121794879333088e-06, "loss": 0.2585, "step": 17070 }, { "epoch": 0.85, "learning_rate": 1.1210507716168494e-06, "loss": 0.2304, "step": 17071 }, { "epoch": 0.85, "learning_rate": 1.120306896117901e-06, "loss": 0.2439, "step": 17072 }, { "epoch": 0.85, "learning_rate": 1.1195632528556966e-06, "loss": 0.5, "step": 17073 }, { "epoch": 0.85, "learning_rate": 1.1188198418496821e-06, "loss": 0.2233, "step": 17074 }, { "epoch": 0.85, "learning_rate": 1.118076663119302e-06, "loss": 0.3727, "step": 17075 }, { "epoch": 0.85, "learning_rate": 1.1173337166839948e-06, "loss": 0.4331, "step": 17076 }, { "epoch": 0.85, "learning_rate": 1.1165910025631899e-06, "loss": 0.3784, "step": 17077 }, { "epoch": 0.85, "learning_rate": 1.1158485207763125e-06, "loss": 0.3484, "step": 17078 }, { "epoch": 0.85, "learning_rate": 1.1151062713427808e-06, "loss": 0.3554, "step": 17079 }, { "epoch": 0.85, "learning_rate": 1.1143642542820054e-06, "loss": 0.2775, "step": 17080 }, { "epoch": 0.85, "learning_rate": 1.1136224696133968e-06, "loss": 0.3211, "step": 17081 }, { "epoch": 0.85, "learning_rate": 1.1128809173563536e-06, "loss": 0.4593, "step": 17082 }, { "epoch": 0.85, "learning_rate": 1.1121395975302696e-06, "loss": 0.5047, "step": 17083 }, { "epoch": 0.85, "learning_rate": 1.1113985101545333e-06, "loss": 0.3599, "step": 17084 }, { "epoch": 0.85, "learning_rate": 1.1106576552485282e-06, "loss": 0.3271, "step": 17085 }, { "epoch": 0.85, "learning_rate": 1.1099170328316255e-06, "loss": 0.2398, "step": 17086 }, { "epoch": 0.85, "learning_rate": 1.109176642923202e-06, "loss": 0.3934, "step": 17087 }, { "epoch": 0.85, "learning_rate": 1.1084364855426188e-06, "loss": 0.4605, "step": 17088 }, { "epoch": 0.85, "learning_rate": 1.1076965607092328e-06, "loss": 0.4026, "step": 17089 }, { "epoch": 0.85, "learning_rate": 1.106956868442396e-06, "loss": 0.3294, "step": 17090 }, { "epoch": 0.85, "learning_rate": 1.1062174087614552e-06, "loss": 0.4648, "step": 17091 }, { "epoch": 0.85, "learning_rate": 1.1054781816857496e-06, "loss": 0.3455, "step": 17092 }, { "epoch": 0.85, "learning_rate": 1.1047391872346113e-06, "loss": 0.3855, "step": 17093 }, { "epoch": 0.85, "learning_rate": 1.1040004254273685e-06, "loss": 0.1513, "step": 17094 }, { "epoch": 0.85, "learning_rate": 1.103261896283343e-06, "loss": 0.3092, "step": 17095 }, { "epoch": 0.85, "learning_rate": 1.102523599821852e-06, "loss": 0.2965, "step": 17096 }, { "epoch": 0.85, "learning_rate": 1.101785536062201e-06, "loss": 0.5609, "step": 17097 }, { "epoch": 0.85, "learning_rate": 1.1010477050236935e-06, "loss": 0.3968, "step": 17098 }, { "epoch": 0.85, "learning_rate": 1.1003101067256295e-06, "loss": 0.22, "step": 17099 }, { "epoch": 0.85, "learning_rate": 1.099572741187298e-06, "loss": 0.2482, "step": 17100 }, { "epoch": 0.85, "learning_rate": 1.098835608427984e-06, "loss": 0.3966, "step": 17101 }, { "epoch": 0.85, "learning_rate": 1.098098708466967e-06, "loss": 0.2933, "step": 17102 }, { "epoch": 0.85, "learning_rate": 1.0973620413235165e-06, "loss": 0.4706, "step": 17103 }, { "epoch": 0.85, "learning_rate": 1.0966256070169034e-06, "loss": 0.2857, "step": 17104 }, { "epoch": 0.85, "learning_rate": 1.0958894055663861e-06, "loss": 0.3557, "step": 17105 }, { "epoch": 0.85, "learning_rate": 1.0951534369912187e-06, "loss": 0.3103, "step": 17106 }, { "epoch": 0.85, "learning_rate": 1.0944177013106483e-06, "loss": 0.2587, "step": 17107 }, { "epoch": 0.85, "learning_rate": 1.0936821985439205e-06, "loss": 0.2323, "step": 17108 }, { "epoch": 0.85, "learning_rate": 1.09294692871027e-06, "loss": 0.3177, "step": 17109 }, { "epoch": 0.85, "learning_rate": 1.0922118918289259e-06, "loss": 0.2295, "step": 17110 }, { "epoch": 0.85, "learning_rate": 1.0914770879191138e-06, "loss": 0.2128, "step": 17111 }, { "epoch": 0.85, "learning_rate": 1.0907425170000497e-06, "loss": 0.2631, "step": 17112 }, { "epoch": 0.85, "learning_rate": 1.0900081790909456e-06, "loss": 0.459, "step": 17113 }, { "epoch": 0.85, "learning_rate": 1.0892740742110074e-06, "loss": 0.5894, "step": 17114 }, { "epoch": 0.85, "learning_rate": 1.0885402023794356e-06, "loss": 0.3431, "step": 17115 }, { "epoch": 0.85, "learning_rate": 1.087806563615421e-06, "loss": 0.4512, "step": 17116 }, { "epoch": 0.85, "learning_rate": 1.0870731579381543e-06, "loss": 0.3804, "step": 17117 }, { "epoch": 0.85, "learning_rate": 1.0863399853668154e-06, "loss": 0.4706, "step": 17118 }, { "epoch": 0.85, "learning_rate": 1.0856070459205803e-06, "loss": 0.3118, "step": 17119 }, { "epoch": 0.85, "learning_rate": 1.0848743396186157e-06, "loss": 0.1329, "step": 17120 }, { "epoch": 0.85, "learning_rate": 1.084141866480085e-06, "loss": 0.2235, "step": 17121 }, { "epoch": 0.85, "learning_rate": 1.0834096265241479e-06, "loss": 0.2554, "step": 17122 }, { "epoch": 0.85, "learning_rate": 1.0826776197699528e-06, "loss": 0.4327, "step": 17123 }, { "epoch": 0.86, "learning_rate": 1.081945846236646e-06, "loss": 0.2528, "step": 17124 }, { "epoch": 0.86, "learning_rate": 1.0812143059433622e-06, "loss": 0.4807, "step": 17125 }, { "epoch": 0.86, "learning_rate": 1.080482998909239e-06, "loss": 0.3908, "step": 17126 }, { "epoch": 0.86, "learning_rate": 1.0797519251534006e-06, "loss": 0.2116, "step": 17127 }, { "epoch": 0.86, "learning_rate": 1.0790210846949679e-06, "loss": 0.4499, "step": 17128 }, { "epoch": 0.86, "learning_rate": 1.0782904775530534e-06, "loss": 0.4347, "step": 17129 }, { "epoch": 0.86, "learning_rate": 1.0775601037467642e-06, "loss": 0.1568, "step": 17130 }, { "epoch": 0.86, "learning_rate": 1.0768299632952062e-06, "loss": 0.3917, "step": 17131 }, { "epoch": 0.86, "learning_rate": 1.076100056217473e-06, "loss": 0.3097, "step": 17132 }, { "epoch": 0.86, "learning_rate": 1.0753703825326556e-06, "loss": 0.4176, "step": 17133 }, { "epoch": 0.86, "learning_rate": 1.0746409422598357e-06, "loss": 0.2033, "step": 17134 }, { "epoch": 0.86, "learning_rate": 1.0739117354180918e-06, "loss": 0.4376, "step": 17135 }, { "epoch": 0.86, "learning_rate": 1.0731827620264957e-06, "loss": 0.2992, "step": 17136 }, { "epoch": 0.86, "learning_rate": 1.0724540221041124e-06, "loss": 0.2658, "step": 17137 }, { "epoch": 0.86, "learning_rate": 1.0717255156700013e-06, "loss": 0.3602, "step": 17138 }, { "epoch": 0.86, "learning_rate": 1.0709972427432135e-06, "loss": 0.4247, "step": 17139 }, { "epoch": 0.86, "learning_rate": 1.0702692033427997e-06, "loss": 0.2752, "step": 17140 }, { "epoch": 0.86, "learning_rate": 1.0695413974877989e-06, "loss": 0.1489, "step": 17141 }, { "epoch": 0.86, "learning_rate": 1.0688138251972468e-06, "loss": 0.5404, "step": 17142 }, { "epoch": 0.86, "learning_rate": 1.0680864864901709e-06, "loss": 0.5143, "step": 17143 }, { "epoch": 0.86, "learning_rate": 1.0673593813855921e-06, "loss": 0.4142, "step": 17144 }, { "epoch": 0.86, "learning_rate": 1.066632509902531e-06, "loss": 0.2722, "step": 17145 }, { "epoch": 0.86, "learning_rate": 1.0659058720599957e-06, "loss": 0.4604, "step": 17146 }, { "epoch": 0.86, "learning_rate": 1.0651794678769912e-06, "loss": 0.2557, "step": 17147 }, { "epoch": 0.86, "learning_rate": 1.064453297372513e-06, "loss": 0.2961, "step": 17148 }, { "epoch": 0.86, "learning_rate": 1.0637273605655573e-06, "loss": 0.554, "step": 17149 }, { "epoch": 0.86, "learning_rate": 1.063001657475109e-06, "loss": 0.4938, "step": 17150 }, { "epoch": 0.86, "learning_rate": 1.062276188120146e-06, "loss": 0.3519, "step": 17151 }, { "epoch": 0.86, "learning_rate": 1.0615509525196432e-06, "loss": 0.3563, "step": 17152 }, { "epoch": 0.86, "learning_rate": 1.0608259506925655e-06, "loss": 0.3375, "step": 17153 }, { "epoch": 0.86, "learning_rate": 1.06010118265788e-06, "loss": 0.5136, "step": 17154 }, { "epoch": 0.86, "learning_rate": 1.059376648434538e-06, "loss": 0.2954, "step": 17155 }, { "epoch": 0.86, "learning_rate": 1.05865234804149e-06, "loss": 0.349, "step": 17156 }, { "epoch": 0.86, "learning_rate": 1.0579282814976787e-06, "loss": 0.3131, "step": 17157 }, { "epoch": 0.86, "learning_rate": 1.057204448822039e-06, "loss": 0.2654, "step": 17158 }, { "epoch": 0.86, "learning_rate": 1.0564808500335056e-06, "loss": 0.4234, "step": 17159 }, { "epoch": 0.86, "learning_rate": 1.0557574851510022e-06, "loss": 0.4086, "step": 17160 }, { "epoch": 0.86, "learning_rate": 1.055034354193447e-06, "loss": 0.2026, "step": 17161 }, { "epoch": 0.86, "learning_rate": 1.054311457179753e-06, "loss": 0.3594, "step": 17162 }, { "epoch": 0.86, "learning_rate": 1.0535887941288248e-06, "loss": 0.4242, "step": 17163 }, { "epoch": 0.86, "learning_rate": 1.052866365059565e-06, "loss": 0.4134, "step": 17164 }, { "epoch": 0.86, "learning_rate": 1.0521441699908664e-06, "loss": 0.2678, "step": 17165 }, { "epoch": 0.86, "learning_rate": 1.051422208941617e-06, "loss": 0.3981, "step": 17166 }, { "epoch": 0.86, "learning_rate": 1.0507004819306977e-06, "loss": 0.4339, "step": 17167 }, { "epoch": 0.86, "learning_rate": 1.0499789889769885e-06, "loss": 0.3482, "step": 17168 }, { "epoch": 0.86, "learning_rate": 1.0492577300993555e-06, "loss": 0.3446, "step": 17169 }, { "epoch": 0.86, "learning_rate": 1.048536705316663e-06, "loss": 0.4599, "step": 17170 }, { "epoch": 0.86, "learning_rate": 1.0478159146477673e-06, "loss": 0.5713, "step": 17171 }, { "epoch": 0.86, "learning_rate": 1.047095358111523e-06, "loss": 0.6059, "step": 17172 }, { "epoch": 0.86, "learning_rate": 1.0463750357267731e-06, "loss": 0.2852, "step": 17173 }, { "epoch": 0.86, "learning_rate": 1.0456549475123579e-06, "loss": 0.3614, "step": 17174 }, { "epoch": 0.86, "learning_rate": 1.0449350934871084e-06, "loss": 0.1842, "step": 17175 }, { "epoch": 0.86, "learning_rate": 1.0442154736698506e-06, "loss": 0.2603, "step": 17176 }, { "epoch": 0.86, "learning_rate": 1.0434960880794098e-06, "loss": 0.3157, "step": 17177 }, { "epoch": 0.86, "learning_rate": 1.0427769367345976e-06, "loss": 0.4249, "step": 17178 }, { "epoch": 0.86, "learning_rate": 1.0420580196542218e-06, "loss": 0.2592, "step": 17179 }, { "epoch": 0.86, "learning_rate": 1.0413393368570845e-06, "loss": 0.3806, "step": 17180 }, { "epoch": 0.86, "learning_rate": 1.040620888361986e-06, "loss": 0.6422, "step": 17181 }, { "epoch": 0.86, "learning_rate": 1.0399026741877128e-06, "loss": 0.4716, "step": 17182 }, { "epoch": 0.86, "learning_rate": 1.0391846943530492e-06, "loss": 0.1773, "step": 17183 }, { "epoch": 0.86, "learning_rate": 1.038466948876773e-06, "loss": 0.4953, "step": 17184 }, { "epoch": 0.86, "learning_rate": 1.0377494377776577e-06, "loss": 0.2766, "step": 17185 }, { "epoch": 0.86, "learning_rate": 1.0370321610744661e-06, "loss": 0.2483, "step": 17186 }, { "epoch": 0.86, "learning_rate": 1.0363151187859588e-06, "loss": 0.3079, "step": 17187 }, { "epoch": 0.86, "learning_rate": 1.0355983109308897e-06, "loss": 0.6612, "step": 17188 }, { "epoch": 0.86, "learning_rate": 1.034881737528004e-06, "loss": 0.5065, "step": 17189 }, { "epoch": 0.86, "learning_rate": 1.0341653985960465e-06, "loss": 0.2596, "step": 17190 }, { "epoch": 0.86, "learning_rate": 1.0334492941537488e-06, "loss": 0.4535, "step": 17191 }, { "epoch": 0.86, "learning_rate": 1.0327334242198406e-06, "loss": 0.3276, "step": 17192 }, { "epoch": 0.86, "learning_rate": 1.0320177888130455e-06, "loss": 0.4455, "step": 17193 }, { "epoch": 0.86, "learning_rate": 1.0313023879520778e-06, "loss": 0.4283, "step": 17194 }, { "epoch": 0.86, "learning_rate": 1.0305872216556501e-06, "loss": 0.1915, "step": 17195 }, { "epoch": 0.86, "learning_rate": 1.0298722899424663e-06, "loss": 0.4466, "step": 17196 }, { "epoch": 0.86, "learning_rate": 1.0291575928312236e-06, "loss": 0.3999, "step": 17197 }, { "epoch": 0.86, "learning_rate": 1.028443130340614e-06, "loss": 0.336, "step": 17198 }, { "epoch": 0.86, "learning_rate": 1.0277289024893222e-06, "loss": 0.2314, "step": 17199 }, { "epoch": 0.86, "learning_rate": 1.0270149092960324e-06, "loss": 0.279, "step": 17200 }, { "epoch": 0.86, "learning_rate": 1.026301150779414e-06, "loss": 0.5307, "step": 17201 }, { "epoch": 0.86, "learning_rate": 1.0255876269581355e-06, "loss": 0.5151, "step": 17202 }, { "epoch": 0.86, "learning_rate": 1.0248743378508563e-06, "loss": 0.4649, "step": 17203 }, { "epoch": 0.86, "learning_rate": 1.024161283476236e-06, "loss": 0.2885, "step": 17204 }, { "epoch": 0.86, "learning_rate": 1.0234484638529208e-06, "loss": 0.4963, "step": 17205 }, { "epoch": 0.86, "learning_rate": 1.0227358789995545e-06, "loss": 0.2255, "step": 17206 }, { "epoch": 0.86, "learning_rate": 1.0220235289347724e-06, "loss": 0.4748, "step": 17207 }, { "epoch": 0.86, "learning_rate": 1.0213114136772063e-06, "loss": 0.3131, "step": 17208 }, { "epoch": 0.86, "learning_rate": 1.0205995332454798e-06, "loss": 0.3755, "step": 17209 }, { "epoch": 0.86, "learning_rate": 1.0198878876582108e-06, "loss": 0.1265, "step": 17210 }, { "epoch": 0.86, "learning_rate": 1.019176476934014e-06, "loss": 0.3344, "step": 17211 }, { "epoch": 0.86, "learning_rate": 1.0184653010914902e-06, "loss": 0.2532, "step": 17212 }, { "epoch": 0.86, "learning_rate": 1.0177543601492456e-06, "loss": 0.581, "step": 17213 }, { "epoch": 0.86, "learning_rate": 1.017043654125871e-06, "loss": 0.248, "step": 17214 }, { "epoch": 0.86, "learning_rate": 1.0163331830399549e-06, "loss": 0.1197, "step": 17215 }, { "epoch": 0.86, "learning_rate": 1.0156229469100764e-06, "loss": 0.3208, "step": 17216 }, { "epoch": 0.86, "learning_rate": 1.0149129457548112e-06, "loss": 0.5076, "step": 17217 }, { "epoch": 0.86, "learning_rate": 1.014203179592731e-06, "loss": 0.4186, "step": 17218 }, { "epoch": 0.86, "learning_rate": 1.0134936484423974e-06, "loss": 0.2548, "step": 17219 }, { "epoch": 0.86, "learning_rate": 1.0127843523223679e-06, "loss": 0.411, "step": 17220 }, { "epoch": 0.86, "learning_rate": 1.01207529125119e-06, "loss": 0.4289, "step": 17221 }, { "epoch": 0.86, "learning_rate": 1.0113664652474132e-06, "loss": 0.1618, "step": 17222 }, { "epoch": 0.86, "learning_rate": 1.0106578743295725e-06, "loss": 0.1711, "step": 17223 }, { "epoch": 0.86, "learning_rate": 1.0099495185162023e-06, "loss": 0.3475, "step": 17224 }, { "epoch": 0.86, "learning_rate": 1.0092413978258264e-06, "loss": 0.3816, "step": 17225 }, { "epoch": 0.86, "learning_rate": 1.0085335122769647e-06, "loss": 0.4471, "step": 17226 }, { "epoch": 0.86, "learning_rate": 1.0078258618881342e-06, "loss": 0.2558, "step": 17227 }, { "epoch": 0.86, "learning_rate": 1.0071184466778416e-06, "loss": 0.105, "step": 17228 }, { "epoch": 0.86, "learning_rate": 1.0064112666645865e-06, "loss": 0.3235, "step": 17229 }, { "epoch": 0.86, "learning_rate": 1.005704321866865e-06, "loss": 0.3338, "step": 17230 }, { "epoch": 0.86, "learning_rate": 1.0049976123031647e-06, "loss": 0.4816, "step": 17231 }, { "epoch": 0.86, "learning_rate": 1.0042911379919728e-06, "loss": 0.215, "step": 17232 }, { "epoch": 0.86, "learning_rate": 1.0035848989517638e-06, "loss": 0.3396, "step": 17233 }, { "epoch": 0.86, "learning_rate": 1.0028788952010083e-06, "loss": 0.5421, "step": 17234 }, { "epoch": 0.86, "learning_rate": 1.0021731267581714e-06, "loss": 0.338, "step": 17235 }, { "epoch": 0.86, "learning_rate": 1.0014675936417117e-06, "loss": 0.45, "step": 17236 }, { "epoch": 0.86, "learning_rate": 1.0007622958700813e-06, "loss": 0.4029, "step": 17237 }, { "epoch": 0.86, "learning_rate": 1.0000572334617254e-06, "loss": 0.6207, "step": 17238 }, { "epoch": 0.86, "learning_rate": 9.993524064350847e-07, "loss": 0.2601, "step": 17239 }, { "epoch": 0.86, "learning_rate": 9.986478148085921e-07, "loss": 0.2904, "step": 17240 }, { "epoch": 0.86, "learning_rate": 9.979434586006775e-07, "loss": 0.1973, "step": 17241 }, { "epoch": 0.86, "learning_rate": 9.972393378297607e-07, "loss": 0.3779, "step": 17242 }, { "epoch": 0.86, "learning_rate": 9.96535452514259e-07, "loss": 0.5013, "step": 17243 }, { "epoch": 0.86, "learning_rate": 9.958318026725766e-07, "loss": 0.2891, "step": 17244 }, { "epoch": 0.86, "learning_rate": 9.951283883231232e-07, "loss": 0.2161, "step": 17245 }, { "epoch": 0.86, "learning_rate": 9.944252094842922e-07, "loss": 0.4085, "step": 17246 }, { "epoch": 0.86, "learning_rate": 9.93722266174474e-07, "loss": 0.4148, "step": 17247 }, { "epoch": 0.86, "learning_rate": 9.930195584120539e-07, "loss": 0.4229, "step": 17248 }, { "epoch": 0.86, "learning_rate": 9.923170862154085e-07, "loss": 0.4385, "step": 17249 }, { "epoch": 0.86, "learning_rate": 9.916148496029144e-07, "loss": 0.5518, "step": 17250 }, { "epoch": 0.86, "learning_rate": 9.909128485929342e-07, "loss": 0.3551, "step": 17251 }, { "epoch": 0.86, "learning_rate": 9.90211083203829e-07, "loss": 0.4309, "step": 17252 }, { "epoch": 0.86, "learning_rate": 9.8950955345395e-07, "loss": 0.667, "step": 17253 }, { "epoch": 0.86, "learning_rate": 9.8880825936165e-07, "loss": 0.4638, "step": 17254 }, { "epoch": 0.86, "learning_rate": 9.881072009452663e-07, "loss": 0.4817, "step": 17255 }, { "epoch": 0.86, "learning_rate": 9.874063782231357e-07, "loss": 0.2627, "step": 17256 }, { "epoch": 0.86, "learning_rate": 9.86705791213587e-07, "loss": 0.4763, "step": 17257 }, { "epoch": 0.86, "learning_rate": 9.860054399349428e-07, "loss": 0.3253, "step": 17258 }, { "epoch": 0.86, "learning_rate": 9.8530532440552e-07, "loss": 0.4012, "step": 17259 }, { "epoch": 0.86, "learning_rate": 9.846054446436293e-07, "loss": 0.437, "step": 17260 }, { "epoch": 0.86, "learning_rate": 9.83905800667575e-07, "loss": 0.2507, "step": 17261 }, { "epoch": 0.86, "learning_rate": 9.832063924956547e-07, "loss": 0.2377, "step": 17262 }, { "epoch": 0.86, "learning_rate": 9.825072201461628e-07, "loss": 0.4838, "step": 17263 }, { "epoch": 0.86, "learning_rate": 9.818082836373831e-07, "loss": 0.4291, "step": 17264 }, { "epoch": 0.86, "learning_rate": 9.81109582987596e-07, "loss": 0.4765, "step": 17265 }, { "epoch": 0.86, "learning_rate": 9.804111182150755e-07, "loss": 0.3125, "step": 17266 }, { "epoch": 0.86, "learning_rate": 9.79712889338087e-07, "loss": 0.2634, "step": 17267 }, { "epoch": 0.86, "learning_rate": 9.790148963748946e-07, "loss": 0.4095, "step": 17268 }, { "epoch": 0.86, "learning_rate": 9.78317139343753e-07, "loss": 0.4281, "step": 17269 }, { "epoch": 0.86, "learning_rate": 9.776196182629105e-07, "loss": 0.5371, "step": 17270 }, { "epoch": 0.86, "learning_rate": 9.769223331506094e-07, "loss": 0.5821, "step": 17271 }, { "epoch": 0.86, "learning_rate": 9.762252840250841e-07, "loss": 0.4769, "step": 17272 }, { "epoch": 0.86, "learning_rate": 9.755284709045697e-07, "loss": 0.2863, "step": 17273 }, { "epoch": 0.86, "learning_rate": 9.748318938072886e-07, "loss": 0.2854, "step": 17274 }, { "epoch": 0.86, "learning_rate": 9.741355527514585e-07, "loss": 0.405, "step": 17275 }, { "epoch": 0.86, "learning_rate": 9.734394477552888e-07, "loss": 0.3301, "step": 17276 }, { "epoch": 0.86, "learning_rate": 9.727435788369898e-07, "loss": 0.2325, "step": 17277 }, { "epoch": 0.86, "learning_rate": 9.720479460147592e-07, "loss": 0.2056, "step": 17278 }, { "epoch": 0.86, "learning_rate": 9.713525493067899e-07, "loss": 0.2693, "step": 17279 }, { "epoch": 0.86, "learning_rate": 9.706573887312686e-07, "loss": 0.3294, "step": 17280 }, { "epoch": 0.86, "learning_rate": 9.699624643063776e-07, "loss": 0.4416, "step": 17281 }, { "epoch": 0.86, "learning_rate": 9.69267776050291e-07, "loss": 0.4389, "step": 17282 }, { "epoch": 0.86, "learning_rate": 9.685733239811778e-07, "loss": 0.3587, "step": 17283 }, { "epoch": 0.86, "learning_rate": 9.678791081172e-07, "loss": 0.4765, "step": 17284 }, { "epoch": 0.86, "learning_rate": 9.671851284765122e-07, "loss": 0.4165, "step": 17285 }, { "epoch": 0.86, "learning_rate": 9.664913850772684e-07, "loss": 0.451, "step": 17286 }, { "epoch": 0.86, "learning_rate": 9.65797877937611e-07, "loss": 0.4446, "step": 17287 }, { "epoch": 0.86, "learning_rate": 9.651046070756775e-07, "loss": 0.571, "step": 17288 }, { "epoch": 0.86, "learning_rate": 9.644115725095992e-07, "loss": 0.2886, "step": 17289 }, { "epoch": 0.86, "learning_rate": 9.637187742575005e-07, "loss": 0.4637, "step": 17290 }, { "epoch": 0.86, "learning_rate": 9.630262123375023e-07, "loss": 0.2594, "step": 17291 }, { "epoch": 0.86, "learning_rate": 9.623338867677178e-07, "loss": 0.4625, "step": 17292 }, { "epoch": 0.86, "learning_rate": 9.61641797566254e-07, "loss": 0.4789, "step": 17293 }, { "epoch": 0.86, "learning_rate": 9.609499447512082e-07, "loss": 0.2421, "step": 17294 }, { "epoch": 0.86, "learning_rate": 9.602583283406796e-07, "loss": 0.2927, "step": 17295 }, { "epoch": 0.86, "learning_rate": 9.595669483527548e-07, "loss": 0.3955, "step": 17296 }, { "epoch": 0.86, "learning_rate": 9.588758048055158e-07, "loss": 0.2925, "step": 17297 }, { "epoch": 0.86, "learning_rate": 9.58184897717037e-07, "loss": 0.3317, "step": 17298 }, { "epoch": 0.86, "learning_rate": 9.57494227105389e-07, "loss": 0.3845, "step": 17299 }, { "epoch": 0.86, "learning_rate": 9.568037929886377e-07, "loss": 0.2291, "step": 17300 }, { "epoch": 0.86, "learning_rate": 9.561135953848387e-07, "loss": 0.2178, "step": 17301 }, { "epoch": 0.86, "learning_rate": 9.554236343120427e-07, "loss": 0.2194, "step": 17302 }, { "epoch": 0.86, "learning_rate": 9.547339097882957e-07, "loss": 0.5062, "step": 17303 }, { "epoch": 0.86, "learning_rate": 9.540444218316347e-07, "loss": 0.2992, "step": 17304 }, { "epoch": 0.86, "learning_rate": 9.533551704600952e-07, "loss": 0.3905, "step": 17305 }, { "epoch": 0.86, "learning_rate": 9.526661556917017e-07, "loss": 0.3416, "step": 17306 }, { "epoch": 0.86, "learning_rate": 9.519773775444752e-07, "loss": 0.2917, "step": 17307 }, { "epoch": 0.86, "learning_rate": 9.512888360364303e-07, "loss": 0.6498, "step": 17308 }, { "epoch": 0.86, "learning_rate": 9.506005311855737e-07, "loss": 0.3445, "step": 17309 }, { "epoch": 0.86, "learning_rate": 9.499124630099076e-07, "loss": 0.3563, "step": 17310 }, { "epoch": 0.86, "learning_rate": 9.492246315274279e-07, "loss": 0.3102, "step": 17311 }, { "epoch": 0.86, "learning_rate": 9.485370367561231e-07, "loss": 0.3734, "step": 17312 }, { "epoch": 0.86, "learning_rate": 9.478496787139757e-07, "loss": 0.1843, "step": 17313 }, { "epoch": 0.86, "learning_rate": 9.471625574189647e-07, "loss": 0.1989, "step": 17314 }, { "epoch": 0.86, "learning_rate": 9.464756728890601e-07, "loss": 0.2622, "step": 17315 }, { "epoch": 0.86, "learning_rate": 9.457890251422274e-07, "loss": 0.4947, "step": 17316 }, { "epoch": 0.86, "learning_rate": 9.451026141964204e-07, "loss": 0.3613, "step": 17317 }, { "epoch": 0.86, "learning_rate": 9.444164400695977e-07, "loss": 0.4502, "step": 17318 }, { "epoch": 0.86, "learning_rate": 9.437305027797017e-07, "loss": 0.2562, "step": 17319 }, { "epoch": 0.86, "learning_rate": 9.430448023446737e-07, "loss": 0.3259, "step": 17320 }, { "epoch": 0.86, "learning_rate": 9.423593387824459e-07, "loss": 0.5154, "step": 17321 }, { "epoch": 0.86, "learning_rate": 9.416741121109441e-07, "loss": 0.4555, "step": 17322 }, { "epoch": 0.86, "learning_rate": 9.409891223480938e-07, "loss": 0.2536, "step": 17323 }, { "epoch": 0.87, "learning_rate": 9.403043695118075e-07, "loss": 0.2311, "step": 17324 }, { "epoch": 0.87, "learning_rate": 9.396198536199941e-07, "loss": 0.3632, "step": 17325 }, { "epoch": 0.87, "learning_rate": 9.389355746905548e-07, "loss": 0.3133, "step": 17326 }, { "epoch": 0.87, "learning_rate": 9.3825153274139e-07, "loss": 0.4529, "step": 17327 }, { "epoch": 0.87, "learning_rate": 9.375677277903872e-07, "loss": 0.2736, "step": 17328 }, { "epoch": 0.87, "learning_rate": 9.368841598554313e-07, "loss": 0.5281, "step": 17329 }, { "epoch": 0.87, "learning_rate": 9.362008289544001e-07, "loss": 0.5095, "step": 17330 }, { "epoch": 0.87, "learning_rate": 9.355177351051636e-07, "loss": 0.4053, "step": 17331 }, { "epoch": 0.87, "learning_rate": 9.348348783255889e-07, "loss": 0.1342, "step": 17332 }, { "epoch": 0.87, "learning_rate": 9.34152258633535e-07, "loss": 0.6125, "step": 17333 }, { "epoch": 0.87, "learning_rate": 9.334698760468541e-07, "loss": 0.3473, "step": 17334 }, { "epoch": 0.87, "learning_rate": 9.327877305833933e-07, "loss": 0.2902, "step": 17335 }, { "epoch": 0.87, "learning_rate": 9.321058222609924e-07, "loss": 0.1499, "step": 17336 }, { "epoch": 0.87, "learning_rate": 9.314241510974887e-07, "loss": 0.2842, "step": 17337 }, { "epoch": 0.87, "learning_rate": 9.307427171107087e-07, "loss": 0.5956, "step": 17338 }, { "epoch": 0.87, "learning_rate": 9.300615203184738e-07, "loss": 0.2283, "step": 17339 }, { "epoch": 0.87, "learning_rate": 9.293805607385986e-07, "loss": 0.5037, "step": 17340 }, { "epoch": 0.87, "learning_rate": 9.286998383888968e-07, "loss": 0.361, "step": 17341 }, { "epoch": 0.87, "learning_rate": 9.280193532871684e-07, "loss": 0.4854, "step": 17342 }, { "epoch": 0.87, "learning_rate": 9.273391054512126e-07, "loss": 0.1886, "step": 17343 }, { "epoch": 0.87, "learning_rate": 9.266590948988197e-07, "loss": 0.3867, "step": 17344 }, { "epoch": 0.87, "learning_rate": 9.259793216477708e-07, "loss": 0.343, "step": 17345 }, { "epoch": 0.87, "learning_rate": 9.252997857158519e-07, "loss": 0.4112, "step": 17346 }, { "epoch": 0.87, "learning_rate": 9.246204871208298e-07, "loss": 0.1573, "step": 17347 }, { "epoch": 0.87, "learning_rate": 9.239414258804724e-07, "loss": 0.6697, "step": 17348 }, { "epoch": 0.87, "learning_rate": 9.23262602012539e-07, "loss": 0.289, "step": 17349 }, { "epoch": 0.87, "learning_rate": 9.225840155347843e-07, "loss": 0.4933, "step": 17350 }, { "epoch": 0.87, "learning_rate": 9.219056664649562e-07, "loss": 0.5022, "step": 17351 }, { "epoch": 0.87, "learning_rate": 9.21227554820795e-07, "loss": 0.4833, "step": 17352 }, { "epoch": 0.87, "learning_rate": 9.205496806200353e-07, "loss": 0.3282, "step": 17353 }, { "epoch": 0.87, "learning_rate": 9.198720438804076e-07, "loss": 0.3916, "step": 17354 }, { "epoch": 0.87, "learning_rate": 9.191946446196332e-07, "loss": 0.5594, "step": 17355 }, { "epoch": 0.87, "learning_rate": 9.1851748285543e-07, "loss": 0.327, "step": 17356 }, { "epoch": 0.87, "learning_rate": 9.178405586055061e-07, "loss": 0.4434, "step": 17357 }, { "epoch": 0.87, "learning_rate": 9.171638718875664e-07, "loss": 0.3666, "step": 17358 }, { "epoch": 0.87, "learning_rate": 9.16487422719311e-07, "loss": 0.4835, "step": 17359 }, { "epoch": 0.87, "learning_rate": 9.158112111184292e-07, "loss": 0.3626, "step": 17360 }, { "epoch": 0.87, "learning_rate": 9.151352371026068e-07, "loss": 0.2706, "step": 17361 }, { "epoch": 0.87, "learning_rate": 9.14459500689524e-07, "loss": 0.3252, "step": 17362 }, { "epoch": 0.87, "learning_rate": 9.137840018968503e-07, "loss": 0.4399, "step": 17363 }, { "epoch": 0.87, "learning_rate": 9.131087407422579e-07, "loss": 0.1766, "step": 17364 }, { "epoch": 0.87, "learning_rate": 9.124337172434039e-07, "loss": 0.1792, "step": 17365 }, { "epoch": 0.87, "learning_rate": 9.117589314179443e-07, "loss": 0.4212, "step": 17366 }, { "epoch": 0.87, "learning_rate": 9.110843832835247e-07, "loss": 0.317, "step": 17367 }, { "epoch": 0.87, "learning_rate": 9.104100728577903e-07, "loss": 0.4088, "step": 17368 }, { "epoch": 0.87, "learning_rate": 9.097360001583744e-07, "loss": 0.3105, "step": 17369 }, { "epoch": 0.87, "learning_rate": 9.090621652029075e-07, "loss": 0.4196, "step": 17370 }, { "epoch": 0.87, "learning_rate": 9.083885680090132e-07, "loss": 0.3961, "step": 17371 }, { "epoch": 0.87, "learning_rate": 9.077152085943064e-07, "loss": 0.2506, "step": 17372 }, { "epoch": 0.87, "learning_rate": 9.070420869764008e-07, "loss": 0.2947, "step": 17373 }, { "epoch": 0.87, "learning_rate": 9.063692031728999e-07, "loss": 0.4486, "step": 17374 }, { "epoch": 0.87, "learning_rate": 9.056965572014021e-07, "loss": 0.4097, "step": 17375 }, { "epoch": 0.87, "learning_rate": 9.050241490794997e-07, "loss": 0.1662, "step": 17376 }, { "epoch": 0.87, "learning_rate": 9.043519788247768e-07, "loss": 0.4391, "step": 17377 }, { "epoch": 0.87, "learning_rate": 9.036800464548157e-07, "loss": 0.5761, "step": 17378 }, { "epoch": 0.87, "learning_rate": 9.03008351987189e-07, "loss": 0.1696, "step": 17379 }, { "epoch": 0.87, "learning_rate": 9.023368954394651e-07, "loss": 0.2443, "step": 17380 }, { "epoch": 0.87, "learning_rate": 9.016656768292031e-07, "loss": 0.3363, "step": 17381 }, { "epoch": 0.87, "learning_rate": 9.00994696173958e-07, "loss": 0.3198, "step": 17382 }, { "epoch": 0.87, "learning_rate": 9.0032395349128e-07, "loss": 0.3755, "step": 17383 }, { "epoch": 0.87, "learning_rate": 8.996534487987107e-07, "loss": 0.2625, "step": 17384 }, { "epoch": 0.87, "learning_rate": 8.989831821137851e-07, "loss": 0.442, "step": 17385 }, { "epoch": 0.87, "learning_rate": 8.983131534540335e-07, "loss": 0.2145, "step": 17386 }, { "epoch": 0.87, "learning_rate": 8.976433628369807e-07, "loss": 0.1934, "step": 17387 }, { "epoch": 0.87, "learning_rate": 8.96973810280145e-07, "loss": 0.5573, "step": 17388 }, { "epoch": 0.87, "learning_rate": 8.963044958010359e-07, "loss": 0.2111, "step": 17389 }, { "epoch": 0.87, "learning_rate": 8.956354194171579e-07, "loss": 0.3852, "step": 17390 }, { "epoch": 0.87, "learning_rate": 8.949665811460118e-07, "loss": 0.4592, "step": 17391 }, { "epoch": 0.87, "learning_rate": 8.9429798100509e-07, "loss": 0.2026, "step": 17392 }, { "epoch": 0.87, "learning_rate": 8.936296190118788e-07, "loss": 0.3863, "step": 17393 }, { "epoch": 0.87, "learning_rate": 8.929614951838561e-07, "loss": 0.1213, "step": 17394 }, { "epoch": 0.87, "learning_rate": 8.922936095384971e-07, "loss": 0.3162, "step": 17395 }, { "epoch": 0.87, "learning_rate": 8.91625962093271e-07, "loss": 0.2413, "step": 17396 }, { "epoch": 0.87, "learning_rate": 8.909585528656373e-07, "loss": 0.4265, "step": 17397 }, { "epoch": 0.87, "learning_rate": 8.90291381873053e-07, "loss": 0.45, "step": 17398 }, { "epoch": 0.87, "learning_rate": 8.896244491329631e-07, "loss": 0.2778, "step": 17399 }, { "epoch": 0.87, "learning_rate": 8.889577546628158e-07, "loss": 0.3976, "step": 17400 }, { "epoch": 0.87, "learning_rate": 8.882912984800451e-07, "loss": 0.494, "step": 17401 }, { "epoch": 0.87, "learning_rate": 8.876250806020814e-07, "loss": 0.4316, "step": 17402 }, { "epoch": 0.87, "learning_rate": 8.869591010463485e-07, "loss": 0.3964, "step": 17403 }, { "epoch": 0.87, "learning_rate": 8.862933598302648e-07, "loss": 0.2996, "step": 17404 }, { "epoch": 0.87, "learning_rate": 8.856278569712417e-07, "loss": 0.2712, "step": 17405 }, { "epoch": 0.87, "learning_rate": 8.849625924866834e-07, "loss": 0.704, "step": 17406 }, { "epoch": 0.87, "learning_rate": 8.842975663939901e-07, "loss": 0.2405, "step": 17407 }, { "epoch": 0.87, "learning_rate": 8.836327787105548e-07, "loss": 0.3387, "step": 17408 }, { "epoch": 0.87, "learning_rate": 8.829682294537622e-07, "loss": 0.4563, "step": 17409 }, { "epoch": 0.87, "learning_rate": 8.823039186409954e-07, "loss": 0.4794, "step": 17410 }, { "epoch": 0.87, "learning_rate": 8.816398462896281e-07, "loss": 0.2945, "step": 17411 }, { "epoch": 0.87, "learning_rate": 8.809760124170275e-07, "loss": 0.3418, "step": 17412 }, { "epoch": 0.87, "learning_rate": 8.803124170405541e-07, "loss": 0.4927, "step": 17413 }, { "epoch": 0.87, "learning_rate": 8.796490601775665e-07, "loss": 0.3208, "step": 17414 }, { "epoch": 0.87, "learning_rate": 8.789859418454128e-07, "loss": 0.2523, "step": 17415 }, { "epoch": 0.87, "learning_rate": 8.783230620614347e-07, "loss": 0.5575, "step": 17416 }, { "epoch": 0.87, "learning_rate": 8.776604208429695e-07, "loss": 0.5473, "step": 17417 }, { "epoch": 0.87, "learning_rate": 8.769980182073467e-07, "loss": 0.3149, "step": 17418 }, { "epoch": 0.87, "learning_rate": 8.763358541718924e-07, "loss": 0.4034, "step": 17419 }, { "epoch": 0.87, "learning_rate": 8.756739287539251e-07, "loss": 0.3301, "step": 17420 }, { "epoch": 0.87, "learning_rate": 8.750122419707552e-07, "loss": 0.3145, "step": 17421 }, { "epoch": 0.87, "learning_rate": 8.743507938396867e-07, "loss": 0.4445, "step": 17422 }, { "epoch": 0.87, "learning_rate": 8.736895843780224e-07, "loss": 0.4728, "step": 17423 }, { "epoch": 0.87, "learning_rate": 8.730286136030541e-07, "loss": 0.2991, "step": 17424 }, { "epoch": 0.87, "learning_rate": 8.723678815320679e-07, "loss": 0.1787, "step": 17425 }, { "epoch": 0.87, "learning_rate": 8.717073881823446e-07, "loss": 0.1932, "step": 17426 }, { "epoch": 0.87, "learning_rate": 8.71047133571159e-07, "loss": 0.2049, "step": 17427 }, { "epoch": 0.87, "learning_rate": 8.703871177157785e-07, "loss": 0.2468, "step": 17428 }, { "epoch": 0.87, "learning_rate": 8.69727340633465e-07, "loss": 0.3448, "step": 17429 }, { "epoch": 0.87, "learning_rate": 8.690678023414745e-07, "loss": 0.2441, "step": 17430 }, { "epoch": 0.87, "learning_rate": 8.684085028570555e-07, "loss": 0.3254, "step": 17431 }, { "epoch": 0.87, "learning_rate": 8.677494421974519e-07, "loss": 0.3248, "step": 17432 }, { "epoch": 0.87, "learning_rate": 8.670906203799023e-07, "loss": 0.3711, "step": 17433 }, { "epoch": 0.87, "learning_rate": 8.66432037421635e-07, "loss": 0.3666, "step": 17434 }, { "epoch": 0.87, "learning_rate": 8.657736933398742e-07, "loss": 0.3471, "step": 17435 }, { "epoch": 0.87, "learning_rate": 8.651155881518381e-07, "loss": 0.112, "step": 17436 }, { "epoch": 0.87, "learning_rate": 8.644577218747396e-07, "loss": 0.4495, "step": 17437 }, { "epoch": 0.87, "learning_rate": 8.638000945257851e-07, "loss": 0.2743, "step": 17438 }, { "epoch": 0.87, "learning_rate": 8.631427061221731e-07, "loss": 0.192, "step": 17439 }, { "epoch": 0.87, "learning_rate": 8.624855566810952e-07, "loss": 0.291, "step": 17440 }, { "epoch": 0.87, "learning_rate": 8.61828646219739e-07, "loss": 0.1273, "step": 17441 }, { "epoch": 0.87, "learning_rate": 8.611719747552871e-07, "loss": 0.3479, "step": 17442 }, { "epoch": 0.87, "learning_rate": 8.605155423049127e-07, "loss": 0.5803, "step": 17443 }, { "epoch": 0.87, "learning_rate": 8.598593488857832e-07, "loss": 0.2628, "step": 17444 }, { "epoch": 0.87, "learning_rate": 8.59203394515059e-07, "loss": 0.3736, "step": 17445 }, { "epoch": 0.87, "learning_rate": 8.585476792099001e-07, "loss": 0.1826, "step": 17446 }, { "epoch": 0.87, "learning_rate": 8.578922029874537e-07, "loss": 0.3136, "step": 17447 }, { "epoch": 0.87, "learning_rate": 8.572369658648627e-07, "loss": 0.4216, "step": 17448 }, { "epoch": 0.87, "learning_rate": 8.565819678592635e-07, "loss": 0.1709, "step": 17449 }, { "epoch": 0.87, "learning_rate": 8.559272089877857e-07, "loss": 0.3448, "step": 17450 }, { "epoch": 0.87, "learning_rate": 8.552726892675567e-07, "loss": 0.3618, "step": 17451 }, { "epoch": 0.87, "learning_rate": 8.546184087156928e-07, "loss": 0.3294, "step": 17452 }, { "epoch": 0.87, "learning_rate": 8.53964367349307e-07, "loss": 0.2669, "step": 17453 }, { "epoch": 0.87, "learning_rate": 8.533105651855033e-07, "loss": 0.4305, "step": 17454 }, { "epoch": 0.87, "learning_rate": 8.526570022413827e-07, "loss": 0.3264, "step": 17455 }, { "epoch": 0.87, "learning_rate": 8.520036785340358e-07, "loss": 0.2999, "step": 17456 }, { "epoch": 0.87, "learning_rate": 8.513505940805522e-07, "loss": 0.2187, "step": 17457 }, { "epoch": 0.87, "learning_rate": 8.506977488980106e-07, "loss": 0.3362, "step": 17458 }, { "epoch": 0.87, "learning_rate": 8.500451430034851e-07, "loss": 0.3921, "step": 17459 }, { "epoch": 0.87, "learning_rate": 8.493927764140453e-07, "loss": 0.4442, "step": 17460 }, { "epoch": 0.87, "learning_rate": 8.487406491467531e-07, "loss": 0.4637, "step": 17461 }, { "epoch": 0.87, "learning_rate": 8.480887612186628e-07, "loss": 0.3067, "step": 17462 }, { "epoch": 0.87, "learning_rate": 8.474371126468218e-07, "loss": 0.4817, "step": 17463 }, { "epoch": 0.87, "learning_rate": 8.467857034482785e-07, "loss": 0.2767, "step": 17464 }, { "epoch": 0.87, "learning_rate": 8.46134533640065e-07, "loss": 0.1821, "step": 17465 }, { "epoch": 0.87, "learning_rate": 8.454836032392144e-07, "loss": 0.3914, "step": 17466 }, { "epoch": 0.87, "learning_rate": 8.448329122627496e-07, "loss": 0.3296, "step": 17467 }, { "epoch": 0.87, "learning_rate": 8.441824607276872e-07, "loss": 0.4301, "step": 17468 }, { "epoch": 0.87, "learning_rate": 8.435322486510423e-07, "loss": 0.2719, "step": 17469 }, { "epoch": 0.87, "learning_rate": 8.42882276049819e-07, "loss": 0.3589, "step": 17470 }, { "epoch": 0.87, "learning_rate": 8.422325429410161e-07, "loss": 0.2533, "step": 17471 }, { "epoch": 0.87, "learning_rate": 8.415830493416255e-07, "loss": 0.2929, "step": 17472 }, { "epoch": 0.87, "learning_rate": 8.409337952686336e-07, "loss": 0.1464, "step": 17473 }, { "epoch": 0.87, "learning_rate": 8.402847807390247e-07, "loss": 0.4069, "step": 17474 }, { "epoch": 0.87, "learning_rate": 8.396360057697695e-07, "loss": 0.4632, "step": 17475 }, { "epoch": 0.87, "learning_rate": 8.389874703778366e-07, "loss": 0.3345, "step": 17476 }, { "epoch": 0.87, "learning_rate": 8.383391745801872e-07, "loss": 0.4694, "step": 17477 }, { "epoch": 0.87, "learning_rate": 8.376911183937775e-07, "loss": 0.5178, "step": 17478 }, { "epoch": 0.87, "learning_rate": 8.370433018355573e-07, "loss": 0.2923, "step": 17479 }, { "epoch": 0.87, "learning_rate": 8.363957249224663e-07, "loss": 0.4233, "step": 17480 }, { "epoch": 0.87, "learning_rate": 8.357483876714445e-07, "loss": 0.4865, "step": 17481 }, { "epoch": 0.87, "learning_rate": 8.351012900994182e-07, "loss": 0.4833, "step": 17482 }, { "epoch": 0.87, "learning_rate": 8.34454432223315e-07, "loss": 0.3239, "step": 17483 }, { "epoch": 0.87, "learning_rate": 8.338078140600526e-07, "loss": 0.2862, "step": 17484 }, { "epoch": 0.87, "learning_rate": 8.331614356265405e-07, "loss": 0.2722, "step": 17485 }, { "epoch": 0.87, "learning_rate": 8.325152969396821e-07, "loss": 0.3006, "step": 17486 }, { "epoch": 0.87, "learning_rate": 8.318693980163817e-07, "loss": 0.309, "step": 17487 }, { "epoch": 0.87, "learning_rate": 8.312237388735289e-07, "loss": 0.4301, "step": 17488 }, { "epoch": 0.87, "learning_rate": 8.305783195280103e-07, "loss": 0.4211, "step": 17489 }, { "epoch": 0.87, "learning_rate": 8.29933139996707e-07, "loss": 0.2413, "step": 17490 }, { "epoch": 0.87, "learning_rate": 8.292882002964885e-07, "loss": 0.4706, "step": 17491 }, { "epoch": 0.87, "learning_rate": 8.286435004442284e-07, "loss": 0.3537, "step": 17492 }, { "epoch": 0.87, "learning_rate": 8.27999040456784e-07, "loss": 0.247, "step": 17493 }, { "epoch": 0.87, "learning_rate": 8.273548203510129e-07, "loss": 0.3185, "step": 17494 }, { "epoch": 0.87, "learning_rate": 8.267108401437596e-07, "loss": 0.2543, "step": 17495 }, { "epoch": 0.87, "learning_rate": 8.260670998518716e-07, "loss": 0.3655, "step": 17496 }, { "epoch": 0.87, "learning_rate": 8.254235994921833e-07, "loss": 0.2326, "step": 17497 }, { "epoch": 0.87, "learning_rate": 8.247803390815234e-07, "loss": 0.1691, "step": 17498 }, { "epoch": 0.87, "learning_rate": 8.241373186367175e-07, "loss": 0.2596, "step": 17499 }, { "epoch": 0.87, "learning_rate": 8.234945381745807e-07, "loss": 0.1715, "step": 17500 }, { "epoch": 0.87, "learning_rate": 8.228519977119254e-07, "loss": 0.432, "step": 17501 }, { "epoch": 0.87, "learning_rate": 8.222096972655569e-07, "loss": 0.4311, "step": 17502 }, { "epoch": 0.87, "learning_rate": 8.215676368522729e-07, "loss": 0.6258, "step": 17503 }, { "epoch": 0.87, "learning_rate": 8.209258164888634e-07, "loss": 0.3154, "step": 17504 }, { "epoch": 0.87, "learning_rate": 8.202842361921192e-07, "loss": 0.3782, "step": 17505 }, { "epoch": 0.87, "learning_rate": 8.196428959788172e-07, "loss": 0.3164, "step": 17506 }, { "epoch": 0.87, "learning_rate": 8.190017958657303e-07, "loss": 0.4159, "step": 17507 }, { "epoch": 0.87, "learning_rate": 8.183609358696276e-07, "loss": 0.404, "step": 17508 }, { "epoch": 0.87, "learning_rate": 8.177203160072666e-07, "loss": 0.4086, "step": 17509 }, { "epoch": 0.87, "learning_rate": 8.170799362954063e-07, "loss": 0.5458, "step": 17510 }, { "epoch": 0.87, "learning_rate": 8.16439796750792e-07, "loss": 0.2743, "step": 17511 }, { "epoch": 0.87, "learning_rate": 8.15799897390167e-07, "loss": 0.2214, "step": 17512 }, { "epoch": 0.87, "learning_rate": 8.15160238230267e-07, "loss": 0.3752, "step": 17513 }, { "epoch": 0.87, "learning_rate": 8.145208192878184e-07, "loss": 0.2268, "step": 17514 }, { "epoch": 0.87, "learning_rate": 8.138816405795491e-07, "loss": 0.5395, "step": 17515 }, { "epoch": 0.87, "learning_rate": 8.132427021221744e-07, "loss": 0.5401, "step": 17516 }, { "epoch": 0.87, "learning_rate": 8.126040039324035e-07, "loss": 0.3069, "step": 17517 }, { "epoch": 0.87, "learning_rate": 8.119655460269405e-07, "loss": 0.4732, "step": 17518 }, { "epoch": 0.87, "learning_rate": 8.113273284224854e-07, "loss": 0.2732, "step": 17519 }, { "epoch": 0.87, "learning_rate": 8.106893511357294e-07, "loss": 0.2344, "step": 17520 }, { "epoch": 0.87, "learning_rate": 8.100516141833581e-07, "loss": 0.3829, "step": 17521 }, { "epoch": 0.87, "learning_rate": 8.094141175820492e-07, "loss": 0.403, "step": 17522 }, { "epoch": 0.87, "learning_rate": 8.087768613484759e-07, "loss": 0.373, "step": 17523 }, { "epoch": 0.88, "learning_rate": 8.081398454993062e-07, "loss": 0.3811, "step": 17524 }, { "epoch": 0.88, "learning_rate": 8.075030700511999e-07, "loss": 0.3044, "step": 17525 }, { "epoch": 0.88, "learning_rate": 8.068665350208104e-07, "loss": 0.394, "step": 17526 }, { "epoch": 0.88, "learning_rate": 8.062302404247868e-07, "loss": 0.2091, "step": 17527 }, { "epoch": 0.88, "learning_rate": 8.055941862797689e-07, "loss": 0.2965, "step": 17528 }, { "epoch": 0.88, "learning_rate": 8.049583726023935e-07, "loss": 0.3027, "step": 17529 }, { "epoch": 0.88, "learning_rate": 8.043227994092873e-07, "loss": 0.3775, "step": 17530 }, { "epoch": 0.88, "learning_rate": 8.036874667170747e-07, "loss": 0.3193, "step": 17531 }, { "epoch": 0.88, "learning_rate": 8.030523745423701e-07, "loss": 0.5456, "step": 17532 }, { "epoch": 0.88, "learning_rate": 8.024175229017861e-07, "loss": 0.1346, "step": 17533 }, { "epoch": 0.88, "learning_rate": 8.017829118119246e-07, "loss": 0.4189, "step": 17534 }, { "epoch": 0.88, "learning_rate": 8.011485412893838e-07, "loss": 0.151, "step": 17535 }, { "epoch": 0.88, "learning_rate": 8.005144113507535e-07, "loss": 0.5581, "step": 17536 }, { "epoch": 0.88, "learning_rate": 7.998805220126205e-07, "loss": 0.2039, "step": 17537 }, { "epoch": 0.88, "learning_rate": 7.992468732915637e-07, "loss": 0.1441, "step": 17538 }, { "epoch": 0.88, "learning_rate": 7.986134652041533e-07, "loss": 0.5834, "step": 17539 }, { "epoch": 0.88, "learning_rate": 7.97980297766956e-07, "loss": 0.5903, "step": 17540 }, { "epoch": 0.88, "learning_rate": 7.973473709965296e-07, "loss": 0.2557, "step": 17541 }, { "epoch": 0.88, "learning_rate": 7.967146849094309e-07, "loss": 0.363, "step": 17542 }, { "epoch": 0.88, "learning_rate": 7.960822395222068e-07, "loss": 0.2926, "step": 17543 }, { "epoch": 0.88, "learning_rate": 7.954500348513949e-07, "loss": 0.4178, "step": 17544 }, { "epoch": 0.88, "learning_rate": 7.948180709135334e-07, "loss": 0.3084, "step": 17545 }, { "epoch": 0.88, "learning_rate": 7.941863477251455e-07, "loss": 0.3575, "step": 17546 }, { "epoch": 0.88, "learning_rate": 7.935548653027592e-07, "loss": 0.4757, "step": 17547 }, { "epoch": 0.88, "learning_rate": 7.929236236628857e-07, "loss": 0.4108, "step": 17548 }, { "epoch": 0.88, "learning_rate": 7.922926228220362e-07, "loss": 0.3189, "step": 17549 }, { "epoch": 0.88, "learning_rate": 7.916618627967143e-07, "loss": 0.2224, "step": 17550 }, { "epoch": 0.88, "learning_rate": 7.910313436034145e-07, "loss": 0.3035, "step": 17551 }, { "epoch": 0.88, "learning_rate": 7.90401065258628e-07, "loss": 0.6525, "step": 17552 }, { "epoch": 0.88, "learning_rate": 7.897710277788406e-07, "loss": 0.5438, "step": 17553 }, { "epoch": 0.88, "learning_rate": 7.89141231180528e-07, "loss": 0.4202, "step": 17554 }, { "epoch": 0.88, "learning_rate": 7.885116754801603e-07, "loss": 0.4808, "step": 17555 }, { "epoch": 0.88, "learning_rate": 7.878823606942076e-07, "loss": 0.3674, "step": 17556 }, { "epoch": 0.88, "learning_rate": 7.872532868391248e-07, "loss": 0.6477, "step": 17557 }, { "epoch": 0.88, "learning_rate": 7.866244539313661e-07, "loss": 0.4859, "step": 17558 }, { "epoch": 0.88, "learning_rate": 7.859958619873765e-07, "loss": 0.2976, "step": 17559 }, { "epoch": 0.88, "learning_rate": 7.853675110235981e-07, "loss": 0.324, "step": 17560 }, { "epoch": 0.88, "learning_rate": 7.847394010564635e-07, "loss": 0.2771, "step": 17561 }, { "epoch": 0.88, "learning_rate": 7.841115321024006e-07, "loss": 0.3367, "step": 17562 }, { "epoch": 0.88, "learning_rate": 7.834839041778286e-07, "loss": 0.2854, "step": 17563 }, { "epoch": 0.88, "learning_rate": 7.828565172991631e-07, "loss": 0.3774, "step": 17564 }, { "epoch": 0.88, "learning_rate": 7.822293714828144e-07, "loss": 0.4458, "step": 17565 }, { "epoch": 0.88, "learning_rate": 7.816024667451838e-07, "loss": 0.3456, "step": 17566 }, { "epoch": 0.88, "learning_rate": 7.809758031026659e-07, "loss": 0.2566, "step": 17567 }, { "epoch": 0.88, "learning_rate": 7.8034938057165e-07, "loss": 0.3839, "step": 17568 }, { "epoch": 0.88, "learning_rate": 7.797231991685217e-07, "loss": 0.2514, "step": 17569 }, { "epoch": 0.88, "learning_rate": 7.790972589096579e-07, "loss": 0.2533, "step": 17570 }, { "epoch": 0.88, "learning_rate": 7.784715598114268e-07, "loss": 0.4335, "step": 17571 }, { "epoch": 0.88, "learning_rate": 7.77846101890194e-07, "loss": 0.3272, "step": 17572 }, { "epoch": 0.88, "learning_rate": 7.772208851623186e-07, "loss": 0.3347, "step": 17573 }, { "epoch": 0.88, "learning_rate": 7.765959096441511e-07, "loss": 0.199, "step": 17574 }, { "epoch": 0.88, "learning_rate": 7.75971175352036e-07, "loss": 0.2725, "step": 17575 }, { "epoch": 0.88, "learning_rate": 7.753466823023148e-07, "loss": 0.4658, "step": 17576 }, { "epoch": 0.88, "learning_rate": 7.747224305113177e-07, "loss": 0.3829, "step": 17577 }, { "epoch": 0.88, "learning_rate": 7.740984199953716e-07, "loss": 0.5233, "step": 17578 }, { "epoch": 0.88, "learning_rate": 7.734746507707991e-07, "loss": 0.4913, "step": 17579 }, { "epoch": 0.88, "learning_rate": 7.728511228539126e-07, "loss": 0.3022, "step": 17580 }, { "epoch": 0.88, "learning_rate": 7.722278362610203e-07, "loss": 0.2683, "step": 17581 }, { "epoch": 0.88, "learning_rate": 7.716047910084201e-07, "loss": 0.344, "step": 17582 }, { "epoch": 0.88, "learning_rate": 7.709819871124113e-07, "loss": 0.3244, "step": 17583 }, { "epoch": 0.88, "learning_rate": 7.703594245892809e-07, "loss": 0.192, "step": 17584 }, { "epoch": 0.88, "learning_rate": 7.697371034553114e-07, "loss": 0.3235, "step": 17585 }, { "epoch": 0.88, "learning_rate": 7.691150237267786e-07, "loss": 0.5452, "step": 17586 }, { "epoch": 0.88, "learning_rate": 7.684931854199496e-07, "loss": 0.3946, "step": 17587 }, { "epoch": 0.88, "learning_rate": 7.678715885510924e-07, "loss": 0.31, "step": 17588 }, { "epoch": 0.88, "learning_rate": 7.672502331364617e-07, "loss": 0.3426, "step": 17589 }, { "epoch": 0.88, "learning_rate": 7.666291191923081e-07, "loss": 0.3541, "step": 17590 }, { "epoch": 0.88, "learning_rate": 7.660082467348751e-07, "loss": 0.4578, "step": 17591 }, { "epoch": 0.88, "learning_rate": 7.653876157804041e-07, "loss": 0.5708, "step": 17592 }, { "epoch": 0.88, "learning_rate": 7.647672263451245e-07, "loss": 0.4557, "step": 17593 }, { "epoch": 0.88, "learning_rate": 7.64147078445262e-07, "loss": 0.2737, "step": 17594 }, { "epoch": 0.88, "learning_rate": 7.635271720970361e-07, "loss": 0.4813, "step": 17595 }, { "epoch": 0.88, "learning_rate": 7.62907507316658e-07, "loss": 0.3754, "step": 17596 }, { "epoch": 0.88, "learning_rate": 7.622880841203361e-07, "loss": 0.3405, "step": 17597 }, { "epoch": 0.88, "learning_rate": 7.616689025242718e-07, "loss": 0.3233, "step": 17598 }, { "epoch": 0.88, "learning_rate": 7.610499625446566e-07, "loss": 0.5453, "step": 17599 }, { "epoch": 0.88, "learning_rate": 7.604312641976785e-07, "loss": 0.381, "step": 17600 }, { "epoch": 0.88, "learning_rate": 7.598128074995192e-07, "loss": 0.1257, "step": 17601 }, { "epoch": 0.88, "learning_rate": 7.591945924663535e-07, "loss": 0.6106, "step": 17602 }, { "epoch": 0.88, "learning_rate": 7.585766191143496e-07, "loss": 0.4149, "step": 17603 }, { "epoch": 0.88, "learning_rate": 7.5795888745967e-07, "loss": 0.4862, "step": 17604 }, { "epoch": 0.88, "learning_rate": 7.573413975184696e-07, "loss": 0.3241, "step": 17605 }, { "epoch": 0.88, "learning_rate": 7.56724149306901e-07, "loss": 0.3691, "step": 17606 }, { "epoch": 0.88, "learning_rate": 7.56107142841106e-07, "loss": 0.443, "step": 17607 }, { "epoch": 0.88, "learning_rate": 7.554903781372202e-07, "loss": 0.2857, "step": 17608 }, { "epoch": 0.88, "learning_rate": 7.548738552113765e-07, "loss": 0.2449, "step": 17609 }, { "epoch": 0.88, "learning_rate": 7.542575740796964e-07, "loss": 0.1991, "step": 17610 }, { "epoch": 0.88, "learning_rate": 7.536415347583015e-07, "loss": 0.4523, "step": 17611 }, { "epoch": 0.88, "learning_rate": 7.530257372633009e-07, "loss": 0.3366, "step": 17612 }, { "epoch": 0.88, "learning_rate": 7.524101816108009e-07, "loss": 0.4517, "step": 17613 }, { "epoch": 0.88, "learning_rate": 7.517948678168996e-07, "loss": 0.2759, "step": 17614 }, { "epoch": 0.88, "learning_rate": 7.511797958976918e-07, "loss": 0.1816, "step": 17615 }, { "epoch": 0.88, "learning_rate": 7.505649658692626e-07, "loss": 0.3063, "step": 17616 }, { "epoch": 0.88, "learning_rate": 7.499503777476924e-07, "loss": 0.2509, "step": 17617 }, { "epoch": 0.88, "learning_rate": 7.493360315490539e-07, "loss": 0.3084, "step": 17618 }, { "epoch": 0.88, "learning_rate": 7.487219272894142e-07, "loss": 0.2255, "step": 17619 }, { "epoch": 0.88, "learning_rate": 7.481080649848372e-07, "loss": 0.2554, "step": 17620 }, { "epoch": 0.88, "learning_rate": 7.474944446513754e-07, "loss": 0.3769, "step": 17621 }, { "epoch": 0.88, "learning_rate": 7.468810663050785e-07, "loss": 0.3258, "step": 17622 }, { "epoch": 0.88, "learning_rate": 7.462679299619879e-07, "loss": 0.3492, "step": 17623 }, { "epoch": 0.88, "learning_rate": 7.456550356381387e-07, "loss": 0.3294, "step": 17624 }, { "epoch": 0.88, "learning_rate": 7.450423833495623e-07, "loss": 0.2614, "step": 17625 }, { "epoch": 0.88, "learning_rate": 7.444299731122795e-07, "loss": 0.482, "step": 17626 }, { "epoch": 0.88, "learning_rate": 7.438178049423084e-07, "loss": 0.4857, "step": 17627 }, { "epoch": 0.88, "learning_rate": 7.432058788556573e-07, "loss": 0.35, "step": 17628 }, { "epoch": 0.88, "learning_rate": 7.425941948683347e-07, "loss": 0.331, "step": 17629 }, { "epoch": 0.88, "learning_rate": 7.419827529963353e-07, "loss": 0.2858, "step": 17630 }, { "epoch": 0.88, "learning_rate": 7.413715532556509e-07, "loss": 0.2931, "step": 17631 }, { "epoch": 0.88, "learning_rate": 7.407605956622655e-07, "loss": 0.408, "step": 17632 }, { "epoch": 0.88, "learning_rate": 7.401498802321605e-07, "loss": 0.3567, "step": 17633 }, { "epoch": 0.88, "learning_rate": 7.395394069813078e-07, "loss": 0.4333, "step": 17634 }, { "epoch": 0.88, "learning_rate": 7.389291759256722e-07, "loss": 0.4139, "step": 17635 }, { "epoch": 0.88, "learning_rate": 7.383191870812157e-07, "loss": 0.3744, "step": 17636 }, { "epoch": 0.88, "learning_rate": 7.377094404638873e-07, "loss": 0.3765, "step": 17637 }, { "epoch": 0.88, "learning_rate": 7.370999360896391e-07, "loss": 0.2885, "step": 17638 }, { "epoch": 0.88, "learning_rate": 7.364906739744093e-07, "loss": 0.2784, "step": 17639 }, { "epoch": 0.88, "learning_rate": 7.358816541341329e-07, "loss": 0.5095, "step": 17640 }, { "epoch": 0.88, "learning_rate": 7.352728765847373e-07, "loss": 0.3376, "step": 17641 }, { "epoch": 0.88, "learning_rate": 7.346643413421461e-07, "loss": 0.1678, "step": 17642 }, { "epoch": 0.88, "learning_rate": 7.340560484222737e-07, "loss": 0.2864, "step": 17643 }, { "epoch": 0.88, "learning_rate": 7.334479978410292e-07, "loss": 0.2854, "step": 17644 }, { "epoch": 0.88, "learning_rate": 7.328401896143156e-07, "loss": 0.2738, "step": 17645 }, { "epoch": 0.88, "learning_rate": 7.322326237580279e-07, "loss": 0.5541, "step": 17646 }, { "epoch": 0.88, "learning_rate": 7.316253002880591e-07, "loss": 0.2064, "step": 17647 }, { "epoch": 0.88, "learning_rate": 7.310182192202898e-07, "loss": 0.418, "step": 17648 }, { "epoch": 0.88, "learning_rate": 7.304113805705992e-07, "loss": 0.4267, "step": 17649 }, { "epoch": 0.88, "learning_rate": 7.298047843548562e-07, "loss": 0.1755, "step": 17650 }, { "epoch": 0.88, "learning_rate": 7.291984305889288e-07, "loss": 0.2166, "step": 17651 }, { "epoch": 0.88, "learning_rate": 7.285923192886745e-07, "loss": 0.4816, "step": 17652 }, { "epoch": 0.88, "learning_rate": 7.279864504699441e-07, "loss": 0.1624, "step": 17653 }, { "epoch": 0.88, "learning_rate": 7.273808241485847e-07, "loss": 0.3262, "step": 17654 }, { "epoch": 0.88, "learning_rate": 7.267754403404325e-07, "loss": 0.2307, "step": 17655 }, { "epoch": 0.88, "learning_rate": 7.26170299061325e-07, "loss": 0.4403, "step": 17656 }, { "epoch": 0.88, "learning_rate": 7.255654003270863e-07, "loss": 0.4939, "step": 17657 }, { "epoch": 0.88, "learning_rate": 7.249607441535378e-07, "loss": 0.3877, "step": 17658 }, { "epoch": 0.88, "learning_rate": 7.243563305564927e-07, "loss": 0.325, "step": 17659 }, { "epoch": 0.88, "learning_rate": 7.237521595517571e-07, "loss": 0.4703, "step": 17660 }, { "epoch": 0.88, "learning_rate": 7.231482311551352e-07, "loss": 0.3323, "step": 17661 }, { "epoch": 0.88, "learning_rate": 7.22544545382422e-07, "loss": 0.6138, "step": 17662 }, { "epoch": 0.88, "learning_rate": 7.219411022494038e-07, "loss": 0.3677, "step": 17663 }, { "epoch": 0.88, "learning_rate": 7.213379017718636e-07, "loss": 0.3486, "step": 17664 }, { "epoch": 0.88, "learning_rate": 7.207349439655787e-07, "loss": 0.3527, "step": 17665 }, { "epoch": 0.88, "learning_rate": 7.201322288463175e-07, "loss": 0.4395, "step": 17666 }, { "epoch": 0.88, "learning_rate": 7.195297564298432e-07, "loss": 0.5518, "step": 17667 }, { "epoch": 0.88, "learning_rate": 7.189275267319129e-07, "loss": 0.4757, "step": 17668 }, { "epoch": 0.88, "learning_rate": 7.183255397682764e-07, "loss": 0.3506, "step": 17669 }, { "epoch": 0.88, "learning_rate": 7.177237955546801e-07, "loss": 0.2732, "step": 17670 }, { "epoch": 0.88, "learning_rate": 7.17122294106859e-07, "loss": 0.2809, "step": 17671 }, { "epoch": 0.88, "learning_rate": 7.165210354405471e-07, "loss": 0.1863, "step": 17672 }, { "epoch": 0.88, "learning_rate": 7.159200195714677e-07, "loss": 0.1765, "step": 17673 }, { "epoch": 0.88, "learning_rate": 7.153192465153413e-07, "loss": 0.4481, "step": 17674 }, { "epoch": 0.88, "learning_rate": 7.147187162878777e-07, "loss": 0.4096, "step": 17675 }, { "epoch": 0.88, "learning_rate": 7.141184289047864e-07, "loss": 0.3872, "step": 17676 }, { "epoch": 0.88, "learning_rate": 7.135183843817639e-07, "loss": 0.4711, "step": 17677 }, { "epoch": 0.88, "learning_rate": 7.129185827345031e-07, "loss": 0.3973, "step": 17678 }, { "epoch": 0.88, "learning_rate": 7.123190239786959e-07, "loss": 0.4083, "step": 17679 }, { "epoch": 0.88, "learning_rate": 7.117197081300186e-07, "loss": 0.2478, "step": 17680 }, { "epoch": 0.88, "learning_rate": 7.111206352041478e-07, "loss": 0.3413, "step": 17681 }, { "epoch": 0.88, "learning_rate": 7.105218052167495e-07, "loss": 0.4408, "step": 17682 }, { "epoch": 0.88, "learning_rate": 7.099232181834859e-07, "loss": 0.2216, "step": 17683 }, { "epoch": 0.88, "learning_rate": 7.093248741200132e-07, "loss": 0.4993, "step": 17684 }, { "epoch": 0.88, "learning_rate": 7.087267730419811e-07, "loss": 0.2984, "step": 17685 }, { "epoch": 0.88, "learning_rate": 7.081289149650294e-07, "loss": 0.5494, "step": 17686 }, { "epoch": 0.88, "learning_rate": 7.075312999047957e-07, "loss": 0.346, "step": 17687 }, { "epoch": 0.88, "learning_rate": 7.069339278769105e-07, "loss": 0.3596, "step": 17688 }, { "epoch": 0.88, "learning_rate": 7.063367988969971e-07, "loss": 0.2484, "step": 17689 }, { "epoch": 0.88, "learning_rate": 7.057399129806719e-07, "loss": 0.4671, "step": 17690 }, { "epoch": 0.88, "learning_rate": 7.051432701435457e-07, "loss": 0.4172, "step": 17691 }, { "epoch": 0.88, "learning_rate": 7.045468704012226e-07, "loss": 0.4456, "step": 17692 }, { "epoch": 0.88, "learning_rate": 7.039507137693024e-07, "loss": 0.1985, "step": 17693 }, { "epoch": 0.88, "learning_rate": 7.03354800263375e-07, "loss": 0.3961, "step": 17694 }, { "epoch": 0.88, "learning_rate": 7.027591298990277e-07, "loss": 0.4713, "step": 17695 }, { "epoch": 0.88, "learning_rate": 7.021637026918371e-07, "loss": 0.3321, "step": 17696 }, { "epoch": 0.88, "learning_rate": 7.015685186573773e-07, "loss": 0.4782, "step": 17697 }, { "epoch": 0.88, "learning_rate": 7.009735778112136e-07, "loss": 0.2952, "step": 17698 }, { "epoch": 0.88, "learning_rate": 7.003788801689071e-07, "loss": 0.565, "step": 17699 }, { "epoch": 0.88, "learning_rate": 6.997844257460107e-07, "loss": 0.1594, "step": 17700 }, { "epoch": 0.88, "learning_rate": 6.991902145580687e-07, "loss": 0.2943, "step": 17701 }, { "epoch": 0.88, "learning_rate": 6.985962466206275e-07, "loss": 0.388, "step": 17702 }, { "epoch": 0.88, "learning_rate": 6.980025219492193e-07, "loss": 0.3141, "step": 17703 }, { "epoch": 0.88, "learning_rate": 6.974090405593703e-07, "loss": 0.2728, "step": 17704 }, { "epoch": 0.88, "learning_rate": 6.968158024666027e-07, "loss": 0.1884, "step": 17705 }, { "epoch": 0.88, "learning_rate": 6.962228076864352e-07, "loss": 0.4378, "step": 17706 }, { "epoch": 0.88, "learning_rate": 6.956300562343732e-07, "loss": 0.5954, "step": 17707 }, { "epoch": 0.88, "learning_rate": 6.950375481259208e-07, "loss": 0.3256, "step": 17708 }, { "epoch": 0.88, "learning_rate": 6.944452833765747e-07, "loss": 0.2765, "step": 17709 }, { "epoch": 0.88, "learning_rate": 6.938532620018223e-07, "loss": 0.4037, "step": 17710 }, { "epoch": 0.88, "learning_rate": 6.932614840171503e-07, "loss": 0.6688, "step": 17711 }, { "epoch": 0.88, "learning_rate": 6.92669949438034e-07, "loss": 0.3094, "step": 17712 }, { "epoch": 0.88, "learning_rate": 6.920786582799454e-07, "loss": 0.2501, "step": 17713 }, { "epoch": 0.88, "learning_rate": 6.91487610558349e-07, "loss": 0.3204, "step": 17714 }, { "epoch": 0.88, "learning_rate": 6.908968062887e-07, "loss": 0.3236, "step": 17715 }, { "epoch": 0.88, "learning_rate": 6.903062454864529e-07, "loss": 0.4064, "step": 17716 }, { "epoch": 0.88, "learning_rate": 6.89715928167054e-07, "loss": 0.2926, "step": 17717 }, { "epoch": 0.88, "learning_rate": 6.891258543459401e-07, "loss": 0.4521, "step": 17718 }, { "epoch": 0.88, "learning_rate": 6.885360240385441e-07, "loss": 0.5536, "step": 17719 }, { "epoch": 0.88, "learning_rate": 6.879464372602929e-07, "loss": 0.382, "step": 17720 }, { "epoch": 0.88, "learning_rate": 6.873570940266061e-07, "loss": 0.4787, "step": 17721 }, { "epoch": 0.88, "learning_rate": 6.867679943528949e-07, "loss": 0.353, "step": 17722 }, { "epoch": 0.88, "learning_rate": 6.861791382545713e-07, "loss": 0.2554, "step": 17723 }, { "epoch": 0.89, "learning_rate": 6.855905257470319e-07, "loss": 0.3252, "step": 17724 }, { "epoch": 0.89, "learning_rate": 6.850021568456744e-07, "loss": 0.3103, "step": 17725 }, { "epoch": 0.89, "learning_rate": 6.844140315658842e-07, "loss": 0.3862, "step": 17726 }, { "epoch": 0.89, "learning_rate": 6.838261499230437e-07, "loss": 0.2144, "step": 17727 }, { "epoch": 0.89, "learning_rate": 6.832385119325269e-07, "loss": 0.3864, "step": 17728 }, { "epoch": 0.89, "learning_rate": 6.826511176097051e-07, "loss": 0.4608, "step": 17729 }, { "epoch": 0.89, "learning_rate": 6.820639669699402e-07, "loss": 0.364, "step": 17730 }, { "epoch": 0.89, "learning_rate": 6.814770600285881e-07, "loss": 0.2115, "step": 17731 }, { "epoch": 0.89, "learning_rate": 6.808903968009983e-07, "loss": 0.3414, "step": 17732 }, { "epoch": 0.89, "learning_rate": 6.803039773025122e-07, "loss": 0.4808, "step": 17733 }, { "epoch": 0.89, "learning_rate": 6.797178015484717e-07, "loss": 0.2444, "step": 17734 }, { "epoch": 0.89, "learning_rate": 6.791318695542038e-07, "loss": 0.4169, "step": 17735 }, { "epoch": 0.89, "learning_rate": 6.785461813350336e-07, "loss": 0.3061, "step": 17736 }, { "epoch": 0.89, "learning_rate": 6.779607369062779e-07, "loss": 0.1902, "step": 17737 }, { "epoch": 0.89, "learning_rate": 6.773755362832513e-07, "loss": 0.3446, "step": 17738 }, { "epoch": 0.89, "learning_rate": 6.767905794812569e-07, "loss": 0.4531, "step": 17739 }, { "epoch": 0.89, "learning_rate": 6.762058665155935e-07, "loss": 0.3628, "step": 17740 }, { "epoch": 0.89, "learning_rate": 6.756213974015535e-07, "loss": 0.2816, "step": 17741 }, { "epoch": 0.89, "learning_rate": 6.750371721544224e-07, "loss": 0.3899, "step": 17742 }, { "epoch": 0.89, "learning_rate": 6.744531907894814e-07, "loss": 0.4999, "step": 17743 }, { "epoch": 0.89, "learning_rate": 6.738694533220025e-07, "loss": 0.1183, "step": 17744 }, { "epoch": 0.89, "learning_rate": 6.732859597672536e-07, "loss": 0.5037, "step": 17745 }, { "epoch": 0.89, "learning_rate": 6.727027101404948e-07, "loss": 0.3837, "step": 17746 }, { "epoch": 0.89, "learning_rate": 6.721197044569805e-07, "loss": 0.192, "step": 17747 }, { "epoch": 0.89, "learning_rate": 6.715369427319574e-07, "loss": 0.3838, "step": 17748 }, { "epoch": 0.89, "learning_rate": 6.709544249806676e-07, "loss": 0.4628, "step": 17749 }, { "epoch": 0.89, "learning_rate": 6.703721512183459e-07, "loss": 0.212, "step": 17750 }, { "epoch": 0.89, "learning_rate": 6.697901214602187e-07, "loss": 0.6383, "step": 17751 }, { "epoch": 0.89, "learning_rate": 6.692083357215129e-07, "loss": 0.2893, "step": 17752 }, { "epoch": 0.89, "learning_rate": 6.686267940174418e-07, "loss": 0.491, "step": 17753 }, { "epoch": 0.89, "learning_rate": 6.680454963632144e-07, "loss": 0.2764, "step": 17754 }, { "epoch": 0.89, "learning_rate": 6.674644427740351e-07, "loss": 0.4481, "step": 17755 }, { "epoch": 0.89, "learning_rate": 6.668836332650985e-07, "loss": 0.361, "step": 17756 }, { "epoch": 0.89, "learning_rate": 6.66303067851597e-07, "loss": 0.3947, "step": 17757 }, { "epoch": 0.89, "learning_rate": 6.657227465487148e-07, "loss": 0.293, "step": 17758 }, { "epoch": 0.89, "learning_rate": 6.651426693716279e-07, "loss": 0.4651, "step": 17759 }, { "epoch": 0.89, "learning_rate": 6.645628363355061e-07, "loss": 0.528, "step": 17760 }, { "epoch": 0.89, "learning_rate": 6.639832474555186e-07, "loss": 0.3004, "step": 17761 }, { "epoch": 0.89, "learning_rate": 6.634039027468198e-07, "loss": 0.2961, "step": 17762 }, { "epoch": 0.89, "learning_rate": 6.628248022245642e-07, "loss": 0.2908, "step": 17763 }, { "epoch": 0.89, "learning_rate": 6.622459459038966e-07, "loss": 0.4751, "step": 17764 }, { "epoch": 0.89, "learning_rate": 6.616673337999546e-07, "loss": 0.3529, "step": 17765 }, { "epoch": 0.89, "learning_rate": 6.610889659278729e-07, "loss": 0.3619, "step": 17766 }, { "epoch": 0.89, "learning_rate": 6.605108423027784e-07, "loss": 0.3729, "step": 17767 }, { "epoch": 0.89, "learning_rate": 6.599329629397899e-07, "loss": 0.4005, "step": 17768 }, { "epoch": 0.89, "learning_rate": 6.593553278540222e-07, "loss": 0.4711, "step": 17769 }, { "epoch": 0.89, "learning_rate": 6.587779370605807e-07, "loss": 0.2685, "step": 17770 }, { "epoch": 0.89, "learning_rate": 6.58200790574568e-07, "loss": 0.4153, "step": 17771 }, { "epoch": 0.89, "learning_rate": 6.576238884110786e-07, "loss": 0.1788, "step": 17772 }, { "epoch": 0.89, "learning_rate": 6.570472305851993e-07, "loss": 0.3724, "step": 17773 }, { "epoch": 0.89, "learning_rate": 6.564708171120115e-07, "loss": 0.4762, "step": 17774 }, { "epoch": 0.89, "learning_rate": 6.558946480065931e-07, "loss": 0.1763, "step": 17775 }, { "epoch": 0.89, "learning_rate": 6.553187232840119e-07, "loss": 0.4472, "step": 17776 }, { "epoch": 0.89, "learning_rate": 6.547430429593293e-07, "loss": 0.2373, "step": 17777 }, { "epoch": 0.89, "learning_rate": 6.541676070476011e-07, "loss": 0.5898, "step": 17778 }, { "epoch": 0.89, "learning_rate": 6.535924155638806e-07, "loss": 0.3059, "step": 17779 }, { "epoch": 0.89, "learning_rate": 6.530174685232082e-07, "loss": 0.5751, "step": 17780 }, { "epoch": 0.89, "learning_rate": 6.524427659406218e-07, "loss": 0.145, "step": 17781 }, { "epoch": 0.89, "learning_rate": 6.518683078311516e-07, "loss": 0.6483, "step": 17782 }, { "epoch": 0.89, "learning_rate": 6.5129409420982e-07, "loss": 0.2155, "step": 17783 }, { "epoch": 0.89, "learning_rate": 6.507201250916495e-07, "loss": 0.6671, "step": 17784 }, { "epoch": 0.89, "learning_rate": 6.50146400491648e-07, "loss": 0.277, "step": 17785 }, { "epoch": 0.89, "learning_rate": 6.495729204248213e-07, "loss": 0.4137, "step": 17786 }, { "epoch": 0.89, "learning_rate": 6.489996849061675e-07, "loss": 0.2685, "step": 17787 }, { "epoch": 0.89, "learning_rate": 6.484266939506789e-07, "loss": 0.3977, "step": 17788 }, { "epoch": 0.89, "learning_rate": 6.478539475733425e-07, "loss": 0.2882, "step": 17789 }, { "epoch": 0.89, "learning_rate": 6.472814457891375e-07, "loss": 0.4119, "step": 17790 }, { "epoch": 0.89, "learning_rate": 6.467091886130361e-07, "loss": 0.3657, "step": 17791 }, { "epoch": 0.89, "learning_rate": 6.461371760600055e-07, "loss": 0.386, "step": 17792 }, { "epoch": 0.89, "learning_rate": 6.455654081450047e-07, "loss": 0.4473, "step": 17793 }, { "epoch": 0.89, "learning_rate": 6.449938848829885e-07, "loss": 0.3606, "step": 17794 }, { "epoch": 0.89, "learning_rate": 6.444226062889047e-07, "loss": 0.2978, "step": 17795 }, { "epoch": 0.89, "learning_rate": 6.43851572377695e-07, "loss": 0.4016, "step": 17796 }, { "epoch": 0.89, "learning_rate": 6.432807831642928e-07, "loss": 0.2303, "step": 17797 }, { "epoch": 0.89, "learning_rate": 6.427102386636264e-07, "loss": 0.3925, "step": 17798 }, { "epoch": 0.89, "learning_rate": 6.42139938890618e-07, "loss": 0.2669, "step": 17799 }, { "epoch": 0.89, "learning_rate": 6.415698838601824e-07, "loss": 0.3297, "step": 17800 }, { "epoch": 0.89, "learning_rate": 6.41000073587228e-07, "loss": 0.5446, "step": 17801 }, { "epoch": 0.89, "learning_rate": 6.404305080866602e-07, "loss": 0.1768, "step": 17802 }, { "epoch": 0.89, "learning_rate": 6.39861187373374e-07, "loss": 0.5677, "step": 17803 }, { "epoch": 0.89, "learning_rate": 6.392921114622575e-07, "loss": 0.4046, "step": 17804 }, { "epoch": 0.89, "learning_rate": 6.387232803681964e-07, "loss": 0.3062, "step": 17805 }, { "epoch": 0.89, "learning_rate": 6.381546941060656e-07, "loss": 0.4156, "step": 17806 }, { "epoch": 0.89, "learning_rate": 6.375863526907388e-07, "loss": 0.264, "step": 17807 }, { "epoch": 0.89, "learning_rate": 6.370182561370775e-07, "loss": 0.2833, "step": 17808 }, { "epoch": 0.89, "learning_rate": 6.364504044599406e-07, "loss": 0.323, "step": 17809 }, { "epoch": 0.89, "learning_rate": 6.358827976741778e-07, "loss": 0.4544, "step": 17810 }, { "epoch": 0.89, "learning_rate": 6.353154357946379e-07, "loss": 0.5867, "step": 17811 }, { "epoch": 0.89, "learning_rate": 6.347483188361559e-07, "loss": 0.4991, "step": 17812 }, { "epoch": 0.89, "learning_rate": 6.341814468135665e-07, "loss": 0.3587, "step": 17813 }, { "epoch": 0.89, "learning_rate": 6.336148197416936e-07, "loss": 0.5291, "step": 17814 }, { "epoch": 0.89, "learning_rate": 6.330484376353563e-07, "loss": 0.3834, "step": 17815 }, { "epoch": 0.89, "learning_rate": 6.324823005093705e-07, "loss": 0.3191, "step": 17816 }, { "epoch": 0.89, "learning_rate": 6.3191640837854e-07, "loss": 0.608, "step": 17817 }, { "epoch": 0.89, "learning_rate": 6.313507612576664e-07, "loss": 0.2644, "step": 17818 }, { "epoch": 0.89, "learning_rate": 6.30785359161542e-07, "loss": 0.26, "step": 17819 }, { "epoch": 0.89, "learning_rate": 6.302202021049564e-07, "loss": 0.2257, "step": 17820 }, { "epoch": 0.89, "learning_rate": 6.296552901026887e-07, "loss": 0.1285, "step": 17821 }, { "epoch": 0.89, "learning_rate": 6.290906231695138e-07, "loss": 0.2664, "step": 17822 }, { "epoch": 0.89, "learning_rate": 6.285262013201998e-07, "loss": 0.3685, "step": 17823 }, { "epoch": 0.89, "learning_rate": 6.279620245695073e-07, "loss": 0.2409, "step": 17824 }, { "epoch": 0.89, "learning_rate": 6.273980929321944e-07, "loss": 0.2612, "step": 17825 }, { "epoch": 0.89, "learning_rate": 6.268344064230081e-07, "loss": 0.4707, "step": 17826 }, { "epoch": 0.89, "learning_rate": 6.262709650566923e-07, "loss": 0.2646, "step": 17827 }, { "epoch": 0.89, "learning_rate": 6.257077688479807e-07, "loss": 0.4071, "step": 17828 }, { "epoch": 0.89, "learning_rate": 6.251448178116038e-07, "loss": 0.3608, "step": 17829 }, { "epoch": 0.89, "learning_rate": 6.245821119622863e-07, "loss": 0.2134, "step": 17830 }, { "epoch": 0.89, "learning_rate": 6.240196513147445e-07, "loss": 0.5025, "step": 17831 }, { "epoch": 0.89, "learning_rate": 6.234574358836876e-07, "loss": 0.6335, "step": 17832 }, { "epoch": 0.89, "learning_rate": 6.228954656838193e-07, "loss": 0.3737, "step": 17833 }, { "epoch": 0.89, "learning_rate": 6.223337407298402e-07, "loss": 0.1696, "step": 17834 }, { "epoch": 0.89, "learning_rate": 6.217722610364397e-07, "loss": 0.5407, "step": 17835 }, { "epoch": 0.89, "learning_rate": 6.212110266183014e-07, "loss": 0.2706, "step": 17836 }, { "epoch": 0.89, "learning_rate": 6.206500374901048e-07, "loss": 0.4961, "step": 17837 }, { "epoch": 0.89, "learning_rate": 6.200892936665215e-07, "loss": 0.2951, "step": 17838 }, { "epoch": 0.89, "learning_rate": 6.195287951622175e-07, "loss": 0.2989, "step": 17839 }, { "epoch": 0.89, "learning_rate": 6.189685419918523e-07, "loss": 0.242, "step": 17840 }, { "epoch": 0.89, "learning_rate": 6.184085341700774e-07, "loss": 0.5139, "step": 17841 }, { "epoch": 0.89, "learning_rate": 6.1784877171154e-07, "loss": 0.4437, "step": 17842 }, { "epoch": 0.89, "learning_rate": 6.172892546308795e-07, "loss": 0.3321, "step": 17843 }, { "epoch": 0.89, "learning_rate": 6.167299829427298e-07, "loss": 0.4049, "step": 17844 }, { "epoch": 0.89, "learning_rate": 6.161709566617169e-07, "loss": 0.3429, "step": 17845 }, { "epoch": 0.89, "learning_rate": 6.156121758024625e-07, "loss": 0.5746, "step": 17846 }, { "epoch": 0.89, "learning_rate": 6.150536403795793e-07, "loss": 0.3038, "step": 17847 }, { "epoch": 0.89, "learning_rate": 6.144953504076767e-07, "loss": 0.5439, "step": 17848 }, { "epoch": 0.89, "learning_rate": 6.139373059013554e-07, "loss": 0.353, "step": 17849 }, { "epoch": 0.89, "learning_rate": 6.133795068752101e-07, "loss": 0.2814, "step": 17850 }, { "epoch": 0.89, "learning_rate": 6.128219533438307e-07, "loss": 0.2849, "step": 17851 }, { "epoch": 0.89, "learning_rate": 6.12264645321795e-07, "loss": 0.2071, "step": 17852 }, { "epoch": 0.89, "learning_rate": 6.11707582823684e-07, "loss": 0.3401, "step": 17853 }, { "epoch": 0.89, "learning_rate": 6.111507658640658e-07, "loss": 0.2347, "step": 17854 }, { "epoch": 0.89, "learning_rate": 6.105941944575012e-07, "loss": 0.4973, "step": 17855 }, { "epoch": 0.89, "learning_rate": 6.100378686185459e-07, "loss": 0.1897, "step": 17856 }, { "epoch": 0.89, "learning_rate": 6.094817883617532e-07, "loss": 0.5499, "step": 17857 }, { "epoch": 0.89, "learning_rate": 6.089259537016645e-07, "loss": 0.2384, "step": 17858 }, { "epoch": 0.89, "learning_rate": 6.083703646528183e-07, "loss": 0.3403, "step": 17859 }, { "epoch": 0.89, "learning_rate": 6.078150212297429e-07, "loss": 0.4355, "step": 17860 }, { "epoch": 0.89, "learning_rate": 6.072599234469644e-07, "loss": 0.2659, "step": 17861 }, { "epoch": 0.89, "learning_rate": 6.067050713190004e-07, "loss": 0.3273, "step": 17862 }, { "epoch": 0.89, "learning_rate": 6.061504648603633e-07, "loss": 0.5738, "step": 17863 }, { "epoch": 0.89, "learning_rate": 6.055961040855563e-07, "loss": 0.2352, "step": 17864 }, { "epoch": 0.89, "learning_rate": 6.050419890090786e-07, "loss": 0.276, "step": 17865 }, { "epoch": 0.89, "learning_rate": 6.044881196454222e-07, "loss": 0.43, "step": 17866 }, { "epoch": 0.89, "learning_rate": 6.039344960090731e-07, "loss": 0.3762, "step": 17867 }, { "epoch": 0.89, "learning_rate": 6.03381118114511e-07, "loss": 0.2152, "step": 17868 }, { "epoch": 0.89, "learning_rate": 6.028279859762099e-07, "loss": 0.3299, "step": 17869 }, { "epoch": 0.89, "learning_rate": 6.022750996086335e-07, "loss": 0.3786, "step": 17870 }, { "epoch": 0.89, "learning_rate": 6.017224590262427e-07, "loss": 0.2763, "step": 17871 }, { "epoch": 0.89, "learning_rate": 6.011700642434915e-07, "loss": 0.3682, "step": 17872 }, { "epoch": 0.89, "learning_rate": 6.006179152748282e-07, "loss": 0.2089, "step": 17873 }, { "epoch": 0.89, "learning_rate": 6.000660121346891e-07, "loss": 0.3742, "step": 17874 }, { "epoch": 0.89, "learning_rate": 5.995143548375149e-07, "loss": 0.392, "step": 17875 }, { "epoch": 0.89, "learning_rate": 5.989629433977296e-07, "loss": 0.2855, "step": 17876 }, { "epoch": 0.89, "learning_rate": 5.984117778297549e-07, "loss": 0.1793, "step": 17877 }, { "epoch": 0.89, "learning_rate": 5.978608581480072e-07, "loss": 0.4967, "step": 17878 }, { "epoch": 0.89, "learning_rate": 5.973101843668916e-07, "loss": 0.3942, "step": 17879 }, { "epoch": 0.89, "learning_rate": 5.967597565008154e-07, "loss": 0.4312, "step": 17880 }, { "epoch": 0.89, "learning_rate": 5.962095745641716e-07, "loss": 0.3837, "step": 17881 }, { "epoch": 0.89, "learning_rate": 5.956596385713487e-07, "loss": 0.1498, "step": 17882 }, { "epoch": 0.89, "learning_rate": 5.951099485367318e-07, "loss": 0.5365, "step": 17883 }, { "epoch": 0.89, "learning_rate": 5.945605044746939e-07, "loss": 0.3886, "step": 17884 }, { "epoch": 0.89, "learning_rate": 5.940113063996078e-07, "loss": 0.208, "step": 17885 }, { "epoch": 0.89, "learning_rate": 5.934623543258377e-07, "loss": 0.1473, "step": 17886 }, { "epoch": 0.89, "learning_rate": 5.929136482677389e-07, "loss": 0.2975, "step": 17887 }, { "epoch": 0.89, "learning_rate": 5.923651882396608e-07, "loss": 0.3375, "step": 17888 }, { "epoch": 0.89, "learning_rate": 5.91816974255951e-07, "loss": 0.3413, "step": 17889 }, { "epoch": 0.89, "learning_rate": 5.912690063309456e-07, "loss": 0.4803, "step": 17890 }, { "epoch": 0.89, "learning_rate": 5.907212844789756e-07, "loss": 0.452, "step": 17891 }, { "epoch": 0.89, "learning_rate": 5.901738087143671e-07, "loss": 0.2453, "step": 17892 }, { "epoch": 0.89, "learning_rate": 5.896265790514388e-07, "loss": 0.4876, "step": 17893 }, { "epoch": 0.89, "learning_rate": 5.890795955045003e-07, "loss": 0.2736, "step": 17894 }, { "epoch": 0.89, "learning_rate": 5.885328580878602e-07, "loss": 0.259, "step": 17895 }, { "epoch": 0.89, "learning_rate": 5.879863668158159e-07, "loss": 0.3505, "step": 17896 }, { "epoch": 0.89, "learning_rate": 5.874401217026593e-07, "loss": 0.3129, "step": 17897 }, { "epoch": 0.89, "learning_rate": 5.86894122762679e-07, "loss": 0.4275, "step": 17898 }, { "epoch": 0.89, "learning_rate": 5.863483700101547e-07, "loss": 0.252, "step": 17899 }, { "epoch": 0.89, "learning_rate": 5.858028634593593e-07, "loss": 0.1275, "step": 17900 }, { "epoch": 0.89, "learning_rate": 5.852576031245583e-07, "loss": 0.287, "step": 17901 }, { "epoch": 0.89, "learning_rate": 5.847125890200134e-07, "loss": 0.3124, "step": 17902 }, { "epoch": 0.89, "learning_rate": 5.8416782115998e-07, "loss": 0.406, "step": 17903 }, { "epoch": 0.89, "learning_rate": 5.836232995587043e-07, "loss": 0.4593, "step": 17904 }, { "epoch": 0.89, "learning_rate": 5.830790242304285e-07, "loss": 0.5083, "step": 17905 }, { "epoch": 0.89, "learning_rate": 5.825349951893855e-07, "loss": 0.4132, "step": 17906 }, { "epoch": 0.89, "learning_rate": 5.819912124498061e-07, "loss": 0.2136, "step": 17907 }, { "epoch": 0.89, "learning_rate": 5.814476760259114e-07, "loss": 0.2059, "step": 17908 }, { "epoch": 0.89, "learning_rate": 5.809043859319164e-07, "loss": 0.4226, "step": 17909 }, { "epoch": 0.89, "learning_rate": 5.803613421820309e-07, "loss": 0.2297, "step": 17910 }, { "epoch": 0.89, "learning_rate": 5.798185447904547e-07, "loss": 0.502, "step": 17911 }, { "epoch": 0.89, "learning_rate": 5.792759937713887e-07, "loss": 0.3887, "step": 17912 }, { "epoch": 0.89, "learning_rate": 5.787336891390194e-07, "loss": 0.3394, "step": 17913 }, { "epoch": 0.89, "learning_rate": 5.781916309075308e-07, "loss": 0.3856, "step": 17914 }, { "epoch": 0.89, "learning_rate": 5.776498190911006e-07, "loss": 0.3509, "step": 17915 }, { "epoch": 0.89, "learning_rate": 5.771082537038974e-07, "loss": 0.3123, "step": 17916 }, { "epoch": 0.89, "learning_rate": 5.765669347600866e-07, "loss": 0.2596, "step": 17917 }, { "epoch": 0.89, "learning_rate": 5.760258622738247e-07, "loss": 0.4218, "step": 17918 }, { "epoch": 0.89, "learning_rate": 5.754850362592634e-07, "loss": 0.3649, "step": 17919 }, { "epoch": 0.89, "learning_rate": 5.749444567305451e-07, "loss": 0.5151, "step": 17920 }, { "epoch": 0.89, "learning_rate": 5.744041237018127e-07, "loss": 0.4129, "step": 17921 }, { "epoch": 0.89, "learning_rate": 5.73864037187194e-07, "loss": 0.4404, "step": 17922 }, { "epoch": 0.89, "learning_rate": 5.733241972008152e-07, "loss": 0.4454, "step": 17923 }, { "epoch": 0.89, "learning_rate": 5.727846037567963e-07, "loss": 0.3836, "step": 17924 }, { "epoch": 0.9, "learning_rate": 5.72245256869246e-07, "loss": 0.4883, "step": 17925 }, { "epoch": 0.9, "learning_rate": 5.717061565522753e-07, "loss": 0.4336, "step": 17926 }, { "epoch": 0.9, "learning_rate": 5.711673028199816e-07, "loss": 0.3803, "step": 17927 }, { "epoch": 0.9, "learning_rate": 5.706286956864571e-07, "loss": 0.4115, "step": 17928 }, { "epoch": 0.9, "learning_rate": 5.700903351657872e-07, "loss": 0.4017, "step": 17929 }, { "epoch": 0.9, "learning_rate": 5.695522212720539e-07, "loss": 0.5035, "step": 17930 }, { "epoch": 0.9, "learning_rate": 5.690143540193327e-07, "loss": 0.3102, "step": 17931 }, { "epoch": 0.9, "learning_rate": 5.684767334216867e-07, "loss": 0.4462, "step": 17932 }, { "epoch": 0.9, "learning_rate": 5.679393594931803e-07, "loss": 0.5695, "step": 17933 }, { "epoch": 0.9, "learning_rate": 5.674022322478645e-07, "loss": 0.4916, "step": 17934 }, { "epoch": 0.9, "learning_rate": 5.668653516997891e-07, "loss": 0.1678, "step": 17935 }, { "epoch": 0.9, "learning_rate": 5.663287178629961e-07, "loss": 0.2895, "step": 17936 }, { "epoch": 0.9, "learning_rate": 5.657923307515201e-07, "loss": 0.5541, "step": 17937 }, { "epoch": 0.9, "learning_rate": 5.652561903793863e-07, "loss": 0.268, "step": 17938 }, { "epoch": 0.9, "learning_rate": 5.647202967606225e-07, "loss": 0.3532, "step": 17939 }, { "epoch": 0.9, "learning_rate": 5.641846499092407e-07, "loss": 0.3103, "step": 17940 }, { "epoch": 0.9, "learning_rate": 5.636492498392498e-07, "loss": 0.32, "step": 17941 }, { "epoch": 0.9, "learning_rate": 5.631140965646542e-07, "loss": 0.6425, "step": 17942 }, { "epoch": 0.9, "learning_rate": 5.625791900994493e-07, "loss": 0.3241, "step": 17943 }, { "epoch": 0.9, "learning_rate": 5.620445304576249e-07, "loss": 0.2664, "step": 17944 }, { "epoch": 0.9, "learning_rate": 5.615101176531646e-07, "loss": 0.3365, "step": 17945 }, { "epoch": 0.9, "learning_rate": 5.609759517000446e-07, "loss": 0.4826, "step": 17946 }, { "epoch": 0.9, "learning_rate": 5.60442032612234e-07, "loss": 0.23, "step": 17947 }, { "epoch": 0.9, "learning_rate": 5.599083604037003e-07, "loss": 0.4112, "step": 17948 }, { "epoch": 0.9, "learning_rate": 5.593749350883992e-07, "loss": 0.5148, "step": 17949 }, { "epoch": 0.9, "learning_rate": 5.588417566802817e-07, "loss": 0.2652, "step": 17950 }, { "epoch": 0.9, "learning_rate": 5.583088251932922e-07, "loss": 0.2849, "step": 17951 }, { "epoch": 0.9, "learning_rate": 5.577761406413673e-07, "loss": 0.2672, "step": 17952 }, { "epoch": 0.9, "learning_rate": 5.572437030384414e-07, "loss": 0.4658, "step": 17953 }, { "epoch": 0.9, "learning_rate": 5.56711512398439e-07, "loss": 0.3738, "step": 17954 }, { "epoch": 0.9, "learning_rate": 5.561795687352789e-07, "loss": 0.3721, "step": 17955 }, { "epoch": 0.9, "learning_rate": 5.556478720628733e-07, "loss": 0.6205, "step": 17956 }, { "epoch": 0.9, "learning_rate": 5.551164223951256e-07, "loss": 0.6232, "step": 17957 }, { "epoch": 0.9, "learning_rate": 5.545852197459389e-07, "loss": 0.3622, "step": 17958 }, { "epoch": 0.9, "learning_rate": 5.540542641292046e-07, "loss": 0.1991, "step": 17959 }, { "epoch": 0.9, "learning_rate": 5.535235555588092e-07, "loss": 0.3473, "step": 17960 }, { "epoch": 0.9, "learning_rate": 5.529930940486306e-07, "loss": 0.2348, "step": 17961 }, { "epoch": 0.9, "learning_rate": 5.524628796125464e-07, "loss": 0.3399, "step": 17962 }, { "epoch": 0.9, "learning_rate": 5.519329122644213e-07, "loss": 0.2786, "step": 17963 }, { "epoch": 0.9, "learning_rate": 5.514031920181151e-07, "loss": 0.6233, "step": 17964 }, { "epoch": 0.9, "learning_rate": 5.508737188874846e-07, "loss": 0.245, "step": 17965 }, { "epoch": 0.9, "learning_rate": 5.503444928863743e-07, "loss": 0.3013, "step": 17966 }, { "epoch": 0.9, "learning_rate": 5.498155140286277e-07, "loss": 0.1311, "step": 17967 }, { "epoch": 0.9, "learning_rate": 5.492867823280789e-07, "loss": 0.2869, "step": 17968 }, { "epoch": 0.9, "learning_rate": 5.487582977985562e-07, "loss": 0.1932, "step": 17969 }, { "epoch": 0.9, "learning_rate": 5.482300604538804e-07, "loss": 0.4116, "step": 17970 }, { "epoch": 0.9, "learning_rate": 5.477020703078684e-07, "loss": 0.2619, "step": 17971 }, { "epoch": 0.9, "learning_rate": 5.47174327374328e-07, "loss": 0.4359, "step": 17972 }, { "epoch": 0.9, "learning_rate": 5.466468316670637e-07, "loss": 0.3795, "step": 17973 }, { "epoch": 0.9, "learning_rate": 5.461195831998689e-07, "loss": 0.2949, "step": 17974 }, { "epoch": 0.9, "learning_rate": 5.455925819865326e-07, "loss": 0.4092, "step": 17975 }, { "epoch": 0.9, "learning_rate": 5.450658280408405e-07, "loss": 0.3893, "step": 17976 }, { "epoch": 0.9, "learning_rate": 5.445393213765681e-07, "loss": 0.4256, "step": 17977 }, { "epoch": 0.9, "learning_rate": 5.440130620074847e-07, "loss": 0.2927, "step": 17978 }, { "epoch": 0.9, "learning_rate": 5.434870499473544e-07, "loss": 0.3121, "step": 17979 }, { "epoch": 0.9, "learning_rate": 5.429612852099353e-07, "loss": 0.3205, "step": 17980 }, { "epoch": 0.9, "learning_rate": 5.424357678089764e-07, "loss": 0.4442, "step": 17981 }, { "epoch": 0.9, "learning_rate": 5.419104977582245e-07, "loss": 0.4816, "step": 17982 }, { "epoch": 0.9, "learning_rate": 5.413854750714143e-07, "loss": 0.3525, "step": 17983 }, { "epoch": 0.9, "learning_rate": 5.408606997622778e-07, "loss": 0.2012, "step": 17984 }, { "epoch": 0.9, "learning_rate": 5.403361718445421e-07, "loss": 0.5687, "step": 17985 }, { "epoch": 0.9, "learning_rate": 5.398118913319228e-07, "loss": 0.3391, "step": 17986 }, { "epoch": 0.9, "learning_rate": 5.392878582381333e-07, "loss": 0.3312, "step": 17987 }, { "epoch": 0.9, "learning_rate": 5.387640725768795e-07, "loss": 0.2454, "step": 17988 }, { "epoch": 0.9, "learning_rate": 5.382405343618579e-07, "loss": 0.4332, "step": 17989 }, { "epoch": 0.9, "learning_rate": 5.377172436067635e-07, "loss": 0.2849, "step": 17990 }, { "epoch": 0.9, "learning_rate": 5.371942003252806e-07, "loss": 0.3376, "step": 17991 }, { "epoch": 0.9, "learning_rate": 5.366714045310905e-07, "loss": 0.2658, "step": 17992 }, { "epoch": 0.9, "learning_rate": 5.361488562378625e-07, "loss": 0.5327, "step": 17993 }, { "epoch": 0.9, "learning_rate": 5.356265554592677e-07, "loss": 0.5022, "step": 17994 }, { "epoch": 0.9, "learning_rate": 5.351045022089641e-07, "loss": 0.4765, "step": 17995 }, { "epoch": 0.9, "learning_rate": 5.345826965006051e-07, "loss": 0.3668, "step": 17996 }, { "epoch": 0.9, "learning_rate": 5.340611383478389e-07, "loss": 0.2727, "step": 17997 }, { "epoch": 0.9, "learning_rate": 5.335398277643034e-07, "loss": 0.4077, "step": 17998 }, { "epoch": 0.9, "learning_rate": 5.330187647636364e-07, "loss": 0.287, "step": 17999 }, { "epoch": 0.9, "learning_rate": 5.32497949359464e-07, "loss": 0.4025, "step": 18000 }, { "epoch": 0.9, "learning_rate": 5.319773815654083e-07, "loss": 0.4636, "step": 18001 }, { "epoch": 0.9, "learning_rate": 5.314570613950809e-07, "loss": 0.3023, "step": 18002 }, { "epoch": 0.9, "learning_rate": 5.309369888620941e-07, "loss": 0.3329, "step": 18003 }, { "epoch": 0.9, "learning_rate": 5.304171639800482e-07, "loss": 0.3157, "step": 18004 }, { "epoch": 0.9, "learning_rate": 5.298975867625389e-07, "loss": 0.3989, "step": 18005 }, { "epoch": 0.9, "learning_rate": 5.293782572231543e-07, "loss": 0.4534, "step": 18006 }, { "epoch": 0.9, "learning_rate": 5.288591753754757e-07, "loss": 0.3838, "step": 18007 }, { "epoch": 0.9, "learning_rate": 5.283403412330823e-07, "loss": 0.341, "step": 18008 }, { "epoch": 0.9, "learning_rate": 5.27821754809541e-07, "loss": 0.3799, "step": 18009 }, { "epoch": 0.9, "learning_rate": 5.273034161184154e-07, "loss": 0.1026, "step": 18010 }, { "epoch": 0.9, "learning_rate": 5.267853251732613e-07, "loss": 0.2511, "step": 18011 }, { "epoch": 0.9, "learning_rate": 5.262674819876301e-07, "loss": 0.2714, "step": 18012 }, { "epoch": 0.9, "learning_rate": 5.257498865750643e-07, "loss": 0.263, "step": 18013 }, { "epoch": 0.9, "learning_rate": 5.252325389491009e-07, "loss": 0.5136, "step": 18014 }, { "epoch": 0.9, "learning_rate": 5.247154391232712e-07, "loss": 0.2829, "step": 18015 }, { "epoch": 0.9, "learning_rate": 5.241985871110988e-07, "loss": 0.1868, "step": 18016 }, { "epoch": 0.9, "learning_rate": 5.236819829261019e-07, "loss": 0.5427, "step": 18017 }, { "epoch": 0.9, "learning_rate": 5.231656265817908e-07, "loss": 0.2296, "step": 18018 }, { "epoch": 0.9, "learning_rate": 5.226495180916702e-07, "loss": 0.3195, "step": 18019 }, { "epoch": 0.9, "learning_rate": 5.22133657469237e-07, "loss": 0.3851, "step": 18020 }, { "epoch": 0.9, "learning_rate": 5.21618044727985e-07, "loss": 0.4994, "step": 18021 }, { "epoch": 0.9, "learning_rate": 5.211026798813989e-07, "loss": 0.218, "step": 18022 }, { "epoch": 0.9, "learning_rate": 5.205875629429569e-07, "loss": 0.288, "step": 18023 }, { "epoch": 0.9, "learning_rate": 5.200726939261314e-07, "loss": 0.3212, "step": 18024 }, { "epoch": 0.9, "learning_rate": 5.195580728443872e-07, "loss": 0.5086, "step": 18025 }, { "epoch": 0.9, "learning_rate": 5.190436997111858e-07, "loss": 0.3958, "step": 18026 }, { "epoch": 0.9, "learning_rate": 5.185295745399777e-07, "loss": 0.387, "step": 18027 }, { "epoch": 0.9, "learning_rate": 5.180156973442107e-07, "loss": 0.0625, "step": 18028 }, { "epoch": 0.9, "learning_rate": 5.175020681373244e-07, "loss": 0.3272, "step": 18029 }, { "epoch": 0.9, "learning_rate": 5.169886869327501e-07, "loss": 0.3434, "step": 18030 }, { "epoch": 0.9, "learning_rate": 5.164755537439171e-07, "loss": 0.2431, "step": 18031 }, { "epoch": 0.9, "learning_rate": 5.159626685842456e-07, "loss": 0.5709, "step": 18032 }, { "epoch": 0.9, "learning_rate": 5.154500314671473e-07, "loss": 0.314, "step": 18033 }, { "epoch": 0.9, "learning_rate": 5.149376424060304e-07, "loss": 0.2892, "step": 18034 }, { "epoch": 0.9, "learning_rate": 5.144255014142973e-07, "loss": 0.4226, "step": 18035 }, { "epoch": 0.9, "learning_rate": 5.13913608505342e-07, "loss": 0.386, "step": 18036 }, { "epoch": 0.9, "learning_rate": 5.134019636925513e-07, "loss": 0.3118, "step": 18037 }, { "epoch": 0.9, "learning_rate": 5.128905669893069e-07, "loss": 0.2239, "step": 18038 }, { "epoch": 0.9, "learning_rate": 5.123794184089847e-07, "loss": 0.45, "step": 18039 }, { "epoch": 0.9, "learning_rate": 5.118685179649519e-07, "loss": 0.2515, "step": 18040 }, { "epoch": 0.9, "learning_rate": 5.113578656705698e-07, "loss": 0.2086, "step": 18041 }, { "epoch": 0.9, "learning_rate": 5.108474615391956e-07, "loss": 0.3139, "step": 18042 }, { "epoch": 0.9, "learning_rate": 5.103373055841765e-07, "loss": 0.4497, "step": 18043 }, { "epoch": 0.9, "learning_rate": 5.098273978188561e-07, "loss": 0.3147, "step": 18044 }, { "epoch": 0.9, "learning_rate": 5.093177382565717e-07, "loss": 0.2975, "step": 18045 }, { "epoch": 0.9, "learning_rate": 5.088083269106503e-07, "loss": 0.3412, "step": 18046 }, { "epoch": 0.9, "learning_rate": 5.082991637944167e-07, "loss": 0.582, "step": 18047 }, { "epoch": 0.9, "learning_rate": 5.077902489211839e-07, "loss": 0.59, "step": 18048 }, { "epoch": 0.9, "learning_rate": 5.072815823042665e-07, "loss": 0.3024, "step": 18049 }, { "epoch": 0.9, "learning_rate": 5.067731639569673e-07, "loss": 0.4544, "step": 18050 }, { "epoch": 0.9, "learning_rate": 5.062649938925801e-07, "loss": 0.4538, "step": 18051 }, { "epoch": 0.9, "learning_rate": 5.057570721243976e-07, "loss": 0.2822, "step": 18052 }, { "epoch": 0.9, "learning_rate": 5.052493986657048e-07, "loss": 0.1431, "step": 18053 }, { "epoch": 0.9, "learning_rate": 5.047419735297776e-07, "loss": 0.3849, "step": 18054 }, { "epoch": 0.9, "learning_rate": 5.042347967298878e-07, "loss": 0.4122, "step": 18055 }, { "epoch": 0.9, "learning_rate": 5.03727868279299e-07, "loss": 0.2392, "step": 18056 }, { "epoch": 0.9, "learning_rate": 5.032211881912697e-07, "loss": 0.5474, "step": 18057 }, { "epoch": 0.9, "learning_rate": 5.027147564790524e-07, "loss": 0.3502, "step": 18058 }, { "epoch": 0.9, "learning_rate": 5.022085731558912e-07, "loss": 0.5738, "step": 18059 }, { "epoch": 0.9, "learning_rate": 5.017026382350254e-07, "loss": 0.3514, "step": 18060 }, { "epoch": 0.9, "learning_rate": 5.011969517296866e-07, "loss": 0.2654, "step": 18061 }, { "epoch": 0.9, "learning_rate": 5.006915136530998e-07, "loss": 0.3593, "step": 18062 }, { "epoch": 0.9, "learning_rate": 5.001863240184845e-07, "loss": 0.3452, "step": 18063 }, { "epoch": 0.9, "learning_rate": 4.996813828390545e-07, "loss": 0.4315, "step": 18064 }, { "epoch": 0.9, "learning_rate": 4.991766901280137e-07, "loss": 0.3137, "step": 18065 }, { "epoch": 0.9, "learning_rate": 4.986722458985616e-07, "loss": 0.3432, "step": 18066 }, { "epoch": 0.9, "learning_rate": 4.981680501638941e-07, "loss": 0.3733, "step": 18067 }, { "epoch": 0.9, "learning_rate": 4.976641029371954e-07, "loss": 0.3078, "step": 18068 }, { "epoch": 0.9, "learning_rate": 4.971604042316469e-07, "loss": 0.2379, "step": 18069 }, { "epoch": 0.9, "learning_rate": 4.966569540604216e-07, "loss": 0.3192, "step": 18070 }, { "epoch": 0.9, "learning_rate": 4.961537524366844e-07, "loss": 0.3418, "step": 18071 }, { "epoch": 0.9, "learning_rate": 4.956507993736004e-07, "loss": 0.4186, "step": 18072 }, { "epoch": 0.9, "learning_rate": 4.951480948843212e-07, "loss": 0.485, "step": 18073 }, { "epoch": 0.9, "learning_rate": 4.946456389819931e-07, "loss": 0.4392, "step": 18074 }, { "epoch": 0.9, "learning_rate": 4.941434316797588e-07, "loss": 0.395, "step": 18075 }, { "epoch": 0.9, "learning_rate": 4.936414729907535e-07, "loss": 0.3769, "step": 18076 }, { "epoch": 0.9, "learning_rate": 4.931397629281031e-07, "loss": 0.3151, "step": 18077 }, { "epoch": 0.9, "learning_rate": 4.926383015049319e-07, "loss": 0.4807, "step": 18078 }, { "epoch": 0.9, "learning_rate": 4.921370887343535e-07, "loss": 0.4033, "step": 18079 }, { "epoch": 0.9, "learning_rate": 4.916361246294743e-07, "loss": 0.255, "step": 18080 }, { "epoch": 0.9, "learning_rate": 4.911354092034004e-07, "loss": 0.3695, "step": 18081 }, { "epoch": 0.9, "learning_rate": 4.906349424692258e-07, "loss": 0.2783, "step": 18082 }, { "epoch": 0.9, "learning_rate": 4.90134724440039e-07, "loss": 0.3002, "step": 18083 }, { "epoch": 0.9, "learning_rate": 4.896347551289216e-07, "loss": 0.3177, "step": 18084 }, { "epoch": 0.9, "learning_rate": 4.89135034548951e-07, "loss": 0.4353, "step": 18085 }, { "epoch": 0.9, "learning_rate": 4.886355627131978e-07, "loss": 0.1246, "step": 18086 }, { "epoch": 0.9, "learning_rate": 4.881363396347239e-07, "loss": 0.2764, "step": 18087 }, { "epoch": 0.9, "learning_rate": 4.876373653265853e-07, "loss": 0.3327, "step": 18088 }, { "epoch": 0.9, "learning_rate": 4.871386398018318e-07, "loss": 0.3779, "step": 18089 }, { "epoch": 0.9, "learning_rate": 4.866401630735085e-07, "loss": 0.2545, "step": 18090 }, { "epoch": 0.9, "learning_rate": 4.861419351546514e-07, "loss": 0.3796, "step": 18091 }, { "epoch": 0.9, "learning_rate": 4.856439560582915e-07, "loss": 0.5732, "step": 18092 }, { "epoch": 0.9, "learning_rate": 4.851462257974515e-07, "loss": 0.3118, "step": 18093 }, { "epoch": 0.9, "learning_rate": 4.846487443851488e-07, "loss": 0.2768, "step": 18094 }, { "epoch": 0.9, "learning_rate": 4.841515118343964e-07, "loss": 0.5024, "step": 18095 }, { "epoch": 0.9, "learning_rate": 4.836545281581972e-07, "loss": 0.4581, "step": 18096 }, { "epoch": 0.9, "learning_rate": 4.831577933695508e-07, "loss": 0.3451, "step": 18097 }, { "epoch": 0.9, "learning_rate": 4.826613074814457e-07, "loss": 0.4913, "step": 18098 }, { "epoch": 0.9, "learning_rate": 4.821650705068692e-07, "loss": 0.5582, "step": 18099 }, { "epoch": 0.9, "learning_rate": 4.816690824588e-07, "loss": 0.4622, "step": 18100 }, { "epoch": 0.9, "learning_rate": 4.811733433502097e-07, "loss": 0.469, "step": 18101 }, { "epoch": 0.9, "learning_rate": 4.806778531940626e-07, "loss": 0.2733, "step": 18102 }, { "epoch": 0.9, "learning_rate": 4.801826120033159e-07, "loss": 0.2087, "step": 18103 }, { "epoch": 0.9, "learning_rate": 4.796876197909273e-07, "loss": 0.2798, "step": 18104 }, { "epoch": 0.9, "learning_rate": 4.791928765698383e-07, "loss": 0.2366, "step": 18105 }, { "epoch": 0.9, "learning_rate": 4.7869838235299e-07, "loss": 0.3881, "step": 18106 }, { "epoch": 0.9, "learning_rate": 4.782041371533131e-07, "loss": 0.4268, "step": 18107 }, { "epoch": 0.9, "learning_rate": 4.777101409837382e-07, "loss": 0.3926, "step": 18108 }, { "epoch": 0.9, "learning_rate": 4.772163938571817e-07, "loss": 0.3114, "step": 18109 }, { "epoch": 0.9, "learning_rate": 4.767228957865577e-07, "loss": 0.4766, "step": 18110 }, { "epoch": 0.9, "learning_rate": 4.762296467847738e-07, "loss": 0.5029, "step": 18111 }, { "epoch": 0.9, "learning_rate": 4.7573664686472954e-07, "loss": 0.4216, "step": 18112 }, { "epoch": 0.9, "learning_rate": 4.7524389603931796e-07, "loss": 0.3417, "step": 18113 }, { "epoch": 0.9, "learning_rate": 4.747513943214277e-07, "loss": 0.4219, "step": 18114 }, { "epoch": 0.9, "learning_rate": 4.742591417239384e-07, "loss": 0.2056, "step": 18115 }, { "epoch": 0.9, "learning_rate": 4.737671382597242e-07, "loss": 0.5166, "step": 18116 }, { "epoch": 0.9, "learning_rate": 4.732753839416537e-07, "loss": 0.297, "step": 18117 }, { "epoch": 0.9, "learning_rate": 4.727838787825889e-07, "loss": 0.2315, "step": 18118 }, { "epoch": 0.9, "learning_rate": 4.7229262279538277e-07, "loss": 0.2693, "step": 18119 }, { "epoch": 0.9, "learning_rate": 4.71801615992884e-07, "loss": 0.1429, "step": 18120 }, { "epoch": 0.9, "learning_rate": 4.7131085838793224e-07, "loss": 0.3705, "step": 18121 }, { "epoch": 0.9, "learning_rate": 4.7082034999336723e-07, "loss": 0.3907, "step": 18122 }, { "epoch": 0.9, "learning_rate": 4.703300908220143e-07, "loss": 0.3378, "step": 18123 }, { "epoch": 0.9, "learning_rate": 4.6984008088669654e-07, "loss": 0.1722, "step": 18124 }, { "epoch": 0.91, "learning_rate": 4.693503202002292e-07, "loss": 0.19, "step": 18125 }, { "epoch": 0.91, "learning_rate": 4.6886080877541986e-07, "loss": 0.1849, "step": 18126 }, { "epoch": 0.91, "learning_rate": 4.683715466250738e-07, "loss": 0.4945, "step": 18127 }, { "epoch": 0.91, "learning_rate": 4.6788253376198524e-07, "loss": 0.2803, "step": 18128 }, { "epoch": 0.91, "learning_rate": 4.6739377019894503e-07, "loss": 0.1999, "step": 18129 }, { "epoch": 0.91, "learning_rate": 4.6690525594873416e-07, "loss": 0.3393, "step": 18130 }, { "epoch": 0.91, "learning_rate": 4.6641699102413007e-07, "loss": 0.4354, "step": 18131 }, { "epoch": 0.91, "learning_rate": 4.6592897543790373e-07, "loss": 0.4406, "step": 18132 }, { "epoch": 0.91, "learning_rate": 4.6544120920281824e-07, "loss": 0.3616, "step": 18133 }, { "epoch": 0.91, "learning_rate": 4.649536923316289e-07, "loss": 0.5201, "step": 18134 }, { "epoch": 0.91, "learning_rate": 4.644664248370878e-07, "loss": 0.3634, "step": 18135 }, { "epoch": 0.91, "learning_rate": 4.639794067319381e-07, "loss": 0.2373, "step": 18136 }, { "epoch": 0.91, "learning_rate": 4.634926380289162e-07, "loss": 0.325, "step": 18137 }, { "epoch": 0.91, "learning_rate": 4.630061187407542e-07, "loss": 0.3112, "step": 18138 }, { "epoch": 0.91, "learning_rate": 4.625198488801752e-07, "loss": 0.2696, "step": 18139 }, { "epoch": 0.91, "learning_rate": 4.6203382845989906e-07, "loss": 0.4713, "step": 18140 }, { "epoch": 0.91, "learning_rate": 4.615480574926345e-07, "loss": 0.3785, "step": 18141 }, { "epoch": 0.91, "learning_rate": 4.6106253599108806e-07, "loss": 0.3791, "step": 18142 }, { "epoch": 0.91, "learning_rate": 4.605772639679573e-07, "loss": 0.5101, "step": 18143 }, { "epoch": 0.91, "learning_rate": 4.6009224143593214e-07, "loss": 0.3064, "step": 18144 }, { "epoch": 0.91, "learning_rate": 4.596074684077012e-07, "loss": 0.5612, "step": 18145 }, { "epoch": 0.91, "learning_rate": 4.591229448959411e-07, "loss": 0.3432, "step": 18146 }, { "epoch": 0.91, "learning_rate": 4.5863867091332394e-07, "loss": 0.6021, "step": 18147 }, { "epoch": 0.91, "learning_rate": 4.5815464647251397e-07, "loss": 0.4635, "step": 18148 }, { "epoch": 0.91, "learning_rate": 4.576708715861733e-07, "loss": 0.5117, "step": 18149 }, { "epoch": 0.91, "learning_rate": 4.5718734626695296e-07, "loss": 0.2728, "step": 18150 }, { "epoch": 0.91, "learning_rate": 4.5670407052749834e-07, "loss": 0.4618, "step": 18151 }, { "epoch": 0.91, "learning_rate": 4.5622104438044826e-07, "loss": 0.1728, "step": 18152 }, { "epoch": 0.91, "learning_rate": 4.55738267838437e-07, "loss": 0.3957, "step": 18153 }, { "epoch": 0.91, "learning_rate": 4.552557409140912e-07, "loss": 0.2664, "step": 18154 }, { "epoch": 0.91, "learning_rate": 4.547734636200307e-07, "loss": 0.4048, "step": 18155 }, { "epoch": 0.91, "learning_rate": 4.542914359688677e-07, "loss": 0.2053, "step": 18156 }, { "epoch": 0.91, "learning_rate": 4.5380965797320876e-07, "loss": 0.5913, "step": 18157 }, { "epoch": 0.91, "learning_rate": 4.533281296456549e-07, "loss": 0.3272, "step": 18158 }, { "epoch": 0.91, "learning_rate": 4.5284685099880044e-07, "loss": 0.3122, "step": 18159 }, { "epoch": 0.91, "learning_rate": 4.5236582204523315e-07, "loss": 0.4208, "step": 18160 }, { "epoch": 0.91, "learning_rate": 4.518850427975319e-07, "loss": 0.4305, "step": 18161 }, { "epoch": 0.91, "learning_rate": 4.5140451326827205e-07, "loss": 0.3939, "step": 18162 }, { "epoch": 0.91, "learning_rate": 4.5092423347002035e-07, "loss": 0.4902, "step": 18163 }, { "epoch": 0.91, "learning_rate": 4.504442034153378e-07, "loss": 0.2221, "step": 18164 }, { "epoch": 0.91, "learning_rate": 4.499644231167799e-07, "loss": 0.6849, "step": 18165 }, { "epoch": 0.91, "learning_rate": 4.4948489258689445e-07, "loss": 0.5101, "step": 18166 }, { "epoch": 0.91, "learning_rate": 4.4900561183822135e-07, "loss": 0.3736, "step": 18167 }, { "epoch": 0.91, "learning_rate": 4.485265808832973e-07, "loss": 0.1953, "step": 18168 }, { "epoch": 0.91, "learning_rate": 4.4804779973465105e-07, "loss": 0.2399, "step": 18169 }, { "epoch": 0.91, "learning_rate": 4.4756926840480384e-07, "loss": 0.4786, "step": 18170 }, { "epoch": 0.91, "learning_rate": 4.470909869062701e-07, "loss": 0.33, "step": 18171 }, { "epoch": 0.91, "learning_rate": 4.4661295525156076e-07, "loss": 0.5177, "step": 18172 }, { "epoch": 0.91, "learning_rate": 4.4613517345317603e-07, "loss": 0.4062, "step": 18173 }, { "epoch": 0.91, "learning_rate": 4.456576415236125e-07, "loss": 0.2552, "step": 18174 }, { "epoch": 0.91, "learning_rate": 4.451803594753601e-07, "loss": 0.5777, "step": 18175 }, { "epoch": 0.91, "learning_rate": 4.447033273209e-07, "loss": 0.4987, "step": 18176 }, { "epoch": 0.91, "learning_rate": 4.44226545072709e-07, "loss": 0.3156, "step": 18177 }, { "epoch": 0.91, "learning_rate": 4.4375001274325815e-07, "loss": 0.2987, "step": 18178 }, { "epoch": 0.91, "learning_rate": 4.4327373034500854e-07, "loss": 0.3526, "step": 18179 }, { "epoch": 0.91, "learning_rate": 4.4279769789041695e-07, "loss": 0.4893, "step": 18180 }, { "epoch": 0.91, "learning_rate": 4.423219153919345e-07, "loss": 0.2696, "step": 18181 }, { "epoch": 0.91, "learning_rate": 4.4184638286200453e-07, "loss": 0.276, "step": 18182 }, { "epoch": 0.91, "learning_rate": 4.4137110031306275e-07, "loss": 0.353, "step": 18183 }, { "epoch": 0.91, "learning_rate": 4.4089606775754133e-07, "loss": 0.4129, "step": 18184 }, { "epoch": 0.91, "learning_rate": 4.4042128520786265e-07, "loss": 0.3917, "step": 18185 }, { "epoch": 0.91, "learning_rate": 4.3994675267644447e-07, "loss": 0.4383, "step": 18186 }, { "epoch": 0.91, "learning_rate": 4.3947247017569694e-07, "loss": 0.3571, "step": 18187 }, { "epoch": 0.91, "learning_rate": 4.3899843771802564e-07, "loss": 0.4438, "step": 18188 }, { "epoch": 0.91, "learning_rate": 4.385246553158262e-07, "loss": 0.1113, "step": 18189 }, { "epoch": 0.91, "learning_rate": 4.380511229814921e-07, "loss": 0.3, "step": 18190 }, { "epoch": 0.91, "learning_rate": 4.3757784072740673e-07, "loss": 0.4759, "step": 18191 }, { "epoch": 0.91, "learning_rate": 4.371048085659491e-07, "loss": 0.6175, "step": 18192 }, { "epoch": 0.91, "learning_rate": 4.366320265094892e-07, "loss": 0.4204, "step": 18193 }, { "epoch": 0.91, "learning_rate": 4.3615949457039175e-07, "loss": 0.4468, "step": 18194 }, { "epoch": 0.91, "learning_rate": 4.35687212761019e-07, "loss": 0.2516, "step": 18195 }, { "epoch": 0.91, "learning_rate": 4.3521518109371884e-07, "loss": 0.4865, "step": 18196 }, { "epoch": 0.91, "learning_rate": 4.347433995808381e-07, "loss": 0.4053, "step": 18197 }, { "epoch": 0.91, "learning_rate": 4.342718682347158e-07, "loss": 0.2964, "step": 18198 }, { "epoch": 0.91, "learning_rate": 4.338005870676831e-07, "loss": 0.2208, "step": 18199 }, { "epoch": 0.91, "learning_rate": 4.3332955609206693e-07, "loss": 0.3205, "step": 18200 }, { "epoch": 0.91, "learning_rate": 4.3285877532018627e-07, "loss": 0.2129, "step": 18201 }, { "epoch": 0.91, "learning_rate": 4.323882447643546e-07, "loss": 0.5208, "step": 18202 }, { "epoch": 0.91, "learning_rate": 4.319179644368743e-07, "loss": 0.1797, "step": 18203 }, { "epoch": 0.91, "learning_rate": 4.3144793435004996e-07, "loss": 0.2757, "step": 18204 }, { "epoch": 0.91, "learning_rate": 4.3097815451617286e-07, "loss": 0.2661, "step": 18205 }, { "epoch": 0.91, "learning_rate": 4.3050862494752763e-07, "loss": 0.3443, "step": 18206 }, { "epoch": 0.91, "learning_rate": 4.300393456563967e-07, "loss": 0.3469, "step": 18207 }, { "epoch": 0.91, "learning_rate": 4.295703166550513e-07, "loss": 0.2047, "step": 18208 }, { "epoch": 0.91, "learning_rate": 4.291015379557595e-07, "loss": 0.4714, "step": 18209 }, { "epoch": 0.91, "learning_rate": 4.2863300957078247e-07, "loss": 0.4954, "step": 18210 }, { "epoch": 0.91, "learning_rate": 4.281647315123716e-07, "loss": 0.263, "step": 18211 }, { "epoch": 0.91, "learning_rate": 4.276967037927748e-07, "loss": 0.4044, "step": 18212 }, { "epoch": 0.91, "learning_rate": 4.272289264242335e-07, "loss": 0.1787, "step": 18213 }, { "epoch": 0.91, "learning_rate": 4.267613994189823e-07, "loss": 0.4639, "step": 18214 }, { "epoch": 0.91, "learning_rate": 4.262941227892481e-07, "loss": 0.3306, "step": 18215 }, { "epoch": 0.91, "learning_rate": 4.258270965472511e-07, "loss": 0.4628, "step": 18216 }, { "epoch": 0.91, "learning_rate": 4.2536032070520595e-07, "loss": 0.4175, "step": 18217 }, { "epoch": 0.91, "learning_rate": 4.248937952753218e-07, "loss": 0.339, "step": 18218 }, { "epoch": 0.91, "learning_rate": 4.2442752026980006e-07, "loss": 0.4475, "step": 18219 }, { "epoch": 0.91, "learning_rate": 4.2396149570083423e-07, "loss": 0.3635, "step": 18220 }, { "epoch": 0.91, "learning_rate": 4.234957215806124e-07, "loss": 0.3092, "step": 18221 }, { "epoch": 0.91, "learning_rate": 4.230301979213181e-07, "loss": 0.2972, "step": 18222 }, { "epoch": 0.91, "learning_rate": 4.225649247351249e-07, "loss": 0.1489, "step": 18223 }, { "epoch": 0.91, "learning_rate": 4.2209990203420204e-07, "loss": 0.4293, "step": 18224 }, { "epoch": 0.91, "learning_rate": 4.2163512983071086e-07, "loss": 0.5886, "step": 18225 }, { "epoch": 0.91, "learning_rate": 4.211706081368072e-07, "loss": 0.2576, "step": 18226 }, { "epoch": 0.91, "learning_rate": 4.2070633696464136e-07, "loss": 0.2993, "step": 18227 }, { "epoch": 0.91, "learning_rate": 4.202423163263547e-07, "loss": 0.4047, "step": 18228 }, { "epoch": 0.91, "learning_rate": 4.197785462340831e-07, "loss": 0.2676, "step": 18229 }, { "epoch": 0.91, "learning_rate": 4.193150266999546e-07, "loss": 0.2921, "step": 18230 }, { "epoch": 0.91, "learning_rate": 4.188517577360929e-07, "loss": 0.2301, "step": 18231 }, { "epoch": 0.91, "learning_rate": 4.18388739354616e-07, "loss": 0.4025, "step": 18232 }, { "epoch": 0.91, "learning_rate": 4.17925971567632e-07, "loss": 0.4124, "step": 18233 }, { "epoch": 0.91, "learning_rate": 4.174634543872436e-07, "loss": 0.183, "step": 18234 }, { "epoch": 0.91, "learning_rate": 4.1700118782554754e-07, "loss": 0.2556, "step": 18235 }, { "epoch": 0.91, "learning_rate": 4.1653917189463324e-07, "loss": 0.316, "step": 18236 }, { "epoch": 0.91, "learning_rate": 4.1607740660658535e-07, "loss": 0.1673, "step": 18237 }, { "epoch": 0.91, "learning_rate": 4.1561589197348097e-07, "loss": 0.4188, "step": 18238 }, { "epoch": 0.91, "learning_rate": 4.1515462800738927e-07, "loss": 0.302, "step": 18239 }, { "epoch": 0.91, "learning_rate": 4.146936147203728e-07, "loss": 0.1624, "step": 18240 }, { "epoch": 0.91, "learning_rate": 4.14232852124492e-07, "loss": 0.2086, "step": 18241 }, { "epoch": 0.91, "learning_rate": 4.13772340231795e-07, "loss": 0.3917, "step": 18242 }, { "epoch": 0.91, "learning_rate": 4.133120790543277e-07, "loss": 0.3569, "step": 18243 }, { "epoch": 0.91, "learning_rate": 4.1285206860412484e-07, "loss": 0.4937, "step": 18244 }, { "epoch": 0.91, "learning_rate": 4.1239230889322135e-07, "loss": 0.4001, "step": 18245 }, { "epoch": 0.91, "learning_rate": 4.119327999336387e-07, "loss": 0.3099, "step": 18246 }, { "epoch": 0.91, "learning_rate": 4.1147354173739496e-07, "loss": 0.2794, "step": 18247 }, { "epoch": 0.91, "learning_rate": 4.1101453431650287e-07, "loss": 0.1719, "step": 18248 }, { "epoch": 0.91, "learning_rate": 4.1055577768296497e-07, "loss": 0.2373, "step": 18249 }, { "epoch": 0.91, "learning_rate": 4.100972718487828e-07, "loss": 0.6586, "step": 18250 }, { "epoch": 0.91, "learning_rate": 4.096390168259445e-07, "loss": 0.1624, "step": 18251 }, { "epoch": 0.91, "learning_rate": 4.091810126264373e-07, "loss": 0.3831, "step": 18252 }, { "epoch": 0.91, "learning_rate": 4.087232592622381e-07, "loss": 0.4751, "step": 18253 }, { "epoch": 0.91, "learning_rate": 4.0826575674532075e-07, "loss": 0.4132, "step": 18254 }, { "epoch": 0.91, "learning_rate": 4.0780850508764904e-07, "loss": 0.3971, "step": 18255 }, { "epoch": 0.91, "learning_rate": 4.073515043011833e-07, "loss": 0.6941, "step": 18256 }, { "epoch": 0.91, "learning_rate": 4.068947543978741e-07, "loss": 0.3464, "step": 18257 }, { "epoch": 0.91, "learning_rate": 4.064382553896684e-07, "loss": 0.3669, "step": 18258 }, { "epoch": 0.91, "learning_rate": 4.0598200728850457e-07, "loss": 0.4506, "step": 18259 }, { "epoch": 0.91, "learning_rate": 4.055260101063152e-07, "loss": 0.1776, "step": 18260 }, { "epoch": 0.91, "learning_rate": 4.0507026385502747e-07, "loss": 0.1631, "step": 18261 }, { "epoch": 0.91, "learning_rate": 4.0461476854655844e-07, "loss": 0.3131, "step": 18262 }, { "epoch": 0.91, "learning_rate": 4.0415952419282203e-07, "loss": 0.2978, "step": 18263 }, { "epoch": 0.91, "learning_rate": 4.0370453080572634e-07, "loss": 0.1828, "step": 18264 }, { "epoch": 0.91, "learning_rate": 4.0324978839716976e-07, "loss": 0.3003, "step": 18265 }, { "epoch": 0.91, "learning_rate": 4.0279529697904495e-07, "loss": 0.4004, "step": 18266 }, { "epoch": 0.91, "learning_rate": 4.0234105656323907e-07, "loss": 0.4028, "step": 18267 }, { "epoch": 0.91, "learning_rate": 4.0188706716163264e-07, "loss": 0.3401, "step": 18268 }, { "epoch": 0.91, "learning_rate": 4.014333287860983e-07, "loss": 0.4835, "step": 18269 }, { "epoch": 0.91, "learning_rate": 4.0097984144850335e-07, "loss": 0.3849, "step": 18270 }, { "epoch": 0.91, "learning_rate": 4.0052660516070816e-07, "loss": 0.5165, "step": 18271 }, { "epoch": 0.91, "learning_rate": 4.000736199345667e-07, "loss": 0.1574, "step": 18272 }, { "epoch": 0.91, "learning_rate": 3.9962088578192614e-07, "loss": 0.4078, "step": 18273 }, { "epoch": 0.91, "learning_rate": 3.99168402714627e-07, "loss": 0.6085, "step": 18274 }, { "epoch": 0.91, "learning_rate": 3.987161707445031e-07, "loss": 0.15, "step": 18275 }, { "epoch": 0.91, "learning_rate": 3.982641898833817e-07, "loss": 0.2596, "step": 18276 }, { "epoch": 0.91, "learning_rate": 3.978124601430866e-07, "loss": 0.3105, "step": 18277 }, { "epoch": 0.91, "learning_rate": 3.9736098153542844e-07, "loss": 0.389, "step": 18278 }, { "epoch": 0.91, "learning_rate": 3.9690975407221663e-07, "loss": 0.3469, "step": 18279 }, { "epoch": 0.91, "learning_rate": 3.96458777765254e-07, "loss": 0.4243, "step": 18280 }, { "epoch": 0.91, "learning_rate": 3.9600805262633215e-07, "loss": 0.4167, "step": 18281 }, { "epoch": 0.91, "learning_rate": 3.9555757866724166e-07, "loss": 0.4279, "step": 18282 }, { "epoch": 0.91, "learning_rate": 3.95107355899762e-07, "loss": 0.4193, "step": 18283 }, { "epoch": 0.91, "learning_rate": 3.946573843356705e-07, "loss": 0.3116, "step": 18284 }, { "epoch": 0.91, "learning_rate": 3.9420766398673205e-07, "loss": 0.2044, "step": 18285 }, { "epoch": 0.91, "learning_rate": 3.937581948647129e-07, "loss": 0.255, "step": 18286 }, { "epoch": 0.91, "learning_rate": 3.933089769813658e-07, "loss": 0.2567, "step": 18287 }, { "epoch": 0.91, "learning_rate": 3.9286001034844037e-07, "loss": 0.4399, "step": 18288 }, { "epoch": 0.91, "learning_rate": 3.924112949776782e-07, "loss": 0.5569, "step": 18289 }, { "epoch": 0.91, "learning_rate": 3.919628308808132e-07, "loss": 0.3432, "step": 18290 }, { "epoch": 0.91, "learning_rate": 3.9151461806957726e-07, "loss": 0.5134, "step": 18291 }, { "epoch": 0.91, "learning_rate": 3.91066656555692e-07, "loss": 0.236, "step": 18292 }, { "epoch": 0.91, "learning_rate": 3.906189463508725e-07, "loss": 0.4345, "step": 18293 }, { "epoch": 0.91, "learning_rate": 3.901714874668294e-07, "loss": 0.3499, "step": 18294 }, { "epoch": 0.91, "learning_rate": 3.897242799152623e-07, "loss": 0.2866, "step": 18295 }, { "epoch": 0.91, "learning_rate": 3.8927732370787065e-07, "loss": 0.3447, "step": 18296 }, { "epoch": 0.91, "learning_rate": 3.88830618856344e-07, "loss": 0.3802, "step": 18297 }, { "epoch": 0.91, "learning_rate": 3.88384165372363e-07, "loss": 0.3513, "step": 18298 }, { "epoch": 0.91, "learning_rate": 3.879379632676039e-07, "loss": 0.4185, "step": 18299 }, { "epoch": 0.91, "learning_rate": 3.8749201255373956e-07, "loss": 0.1654, "step": 18300 }, { "epoch": 0.91, "learning_rate": 3.870463132424318e-07, "loss": 0.7191, "step": 18301 }, { "epoch": 0.91, "learning_rate": 3.866008653453357e-07, "loss": 0.2845, "step": 18302 }, { "epoch": 0.91, "learning_rate": 3.8615566887410414e-07, "loss": 0.4392, "step": 18303 }, { "epoch": 0.91, "learning_rate": 3.8571072384037676e-07, "loss": 0.3852, "step": 18304 }, { "epoch": 0.91, "learning_rate": 3.8526603025579533e-07, "loss": 0.1041, "step": 18305 }, { "epoch": 0.91, "learning_rate": 3.8482158813198723e-07, "loss": 0.3333, "step": 18306 }, { "epoch": 0.91, "learning_rate": 3.8437739748057646e-07, "loss": 0.259, "step": 18307 }, { "epoch": 0.91, "learning_rate": 3.839334583131815e-07, "loss": 0.2632, "step": 18308 }, { "epoch": 0.91, "learning_rate": 3.83489770641412e-07, "loss": 0.1945, "step": 18309 }, { "epoch": 0.91, "learning_rate": 3.83046334476872e-07, "loss": 0.154, "step": 18310 }, { "epoch": 0.91, "learning_rate": 3.8260314983115885e-07, "loss": 0.2499, "step": 18311 }, { "epoch": 0.91, "learning_rate": 3.8216021671586446e-07, "loss": 0.2098, "step": 18312 }, { "epoch": 0.91, "learning_rate": 3.817175351425706e-07, "loss": 0.3603, "step": 18313 }, { "epoch": 0.91, "learning_rate": 3.81275105122858e-07, "loss": 0.3275, "step": 18314 }, { "epoch": 0.91, "learning_rate": 3.8083292666829754e-07, "loss": 0.4141, "step": 18315 }, { "epoch": 0.91, "learning_rate": 3.803909997904531e-07, "loss": 0.3838, "step": 18316 }, { "epoch": 0.91, "learning_rate": 3.799493245008812e-07, "loss": 0.1948, "step": 18317 }, { "epoch": 0.91, "learning_rate": 3.7950790081113574e-07, "loss": 0.304, "step": 18318 }, { "epoch": 0.91, "learning_rate": 3.7906672873275985e-07, "loss": 0.268, "step": 18319 }, { "epoch": 0.91, "learning_rate": 3.786258082772931e-07, "loss": 0.3883, "step": 18320 }, { "epoch": 0.91, "learning_rate": 3.7818513945626745e-07, "loss": 0.2759, "step": 18321 }, { "epoch": 0.91, "learning_rate": 3.7774472228120474e-07, "loss": 0.4746, "step": 18322 }, { "epoch": 0.91, "learning_rate": 3.77304556763628e-07, "loss": 0.3232, "step": 18323 }, { "epoch": 0.91, "learning_rate": 3.768646429150469e-07, "loss": 0.4435, "step": 18324 }, { "epoch": 0.92, "learning_rate": 3.7642498074696664e-07, "loss": 0.3681, "step": 18325 }, { "epoch": 0.92, "learning_rate": 3.759855702708848e-07, "loss": 0.4208, "step": 18326 }, { "epoch": 0.92, "learning_rate": 3.755464114982976e-07, "loss": 0.2032, "step": 18327 }, { "epoch": 0.92, "learning_rate": 3.7510750444068713e-07, "loss": 0.2927, "step": 18328 }, { "epoch": 0.92, "learning_rate": 3.74668849109534e-07, "loss": 0.2658, "step": 18329 }, { "epoch": 0.92, "learning_rate": 3.7423044551631037e-07, "loss": 0.2902, "step": 18330 }, { "epoch": 0.92, "learning_rate": 3.737922936724825e-07, "loss": 0.3546, "step": 18331 }, { "epoch": 0.92, "learning_rate": 3.7335439358950787e-07, "loss": 0.4554, "step": 18332 }, { "epoch": 0.92, "learning_rate": 3.729167452788407e-07, "loss": 0.5563, "step": 18333 }, { "epoch": 0.92, "learning_rate": 3.724793487519274e-07, "loss": 0.4176, "step": 18334 }, { "epoch": 0.92, "learning_rate": 3.7204220402020654e-07, "loss": 0.2827, "step": 18335 }, { "epoch": 0.92, "learning_rate": 3.716053110951101e-07, "loss": 0.4351, "step": 18336 }, { "epoch": 0.92, "learning_rate": 3.711686699880679e-07, "loss": 0.4717, "step": 18337 }, { "epoch": 0.92, "learning_rate": 3.707322807104974e-07, "loss": 0.43, "step": 18338 }, { "epoch": 0.92, "learning_rate": 3.7029614327381167e-07, "loss": 0.2968, "step": 18339 }, { "epoch": 0.92, "learning_rate": 3.6986025768941614e-07, "loss": 0.3011, "step": 18340 }, { "epoch": 0.92, "learning_rate": 3.694246239687138e-07, "loss": 0.4914, "step": 18341 }, { "epoch": 0.92, "learning_rate": 3.6898924212309673e-07, "loss": 0.3631, "step": 18342 }, { "epoch": 0.92, "learning_rate": 3.685541121639513e-07, "loss": 0.6264, "step": 18343 }, { "epoch": 0.92, "learning_rate": 3.681192341026585e-07, "loss": 0.2594, "step": 18344 }, { "epoch": 0.92, "learning_rate": 3.6768460795059136e-07, "loss": 0.3176, "step": 18345 }, { "epoch": 0.92, "learning_rate": 3.672502337191175e-07, "loss": 0.2128, "step": 18346 }, { "epoch": 0.92, "learning_rate": 3.668161114195967e-07, "loss": 0.2665, "step": 18347 }, { "epoch": 0.92, "learning_rate": 3.663822410633844e-07, "loss": 0.4572, "step": 18348 }, { "epoch": 0.92, "learning_rate": 3.6594862266182474e-07, "loss": 0.2797, "step": 18349 }, { "epoch": 0.92, "learning_rate": 3.6551525622626204e-07, "loss": 0.4103, "step": 18350 }, { "epoch": 0.92, "learning_rate": 3.6508214176802835e-07, "loss": 0.3521, "step": 18351 }, { "epoch": 0.92, "learning_rate": 3.646492792984524e-07, "loss": 0.2201, "step": 18352 }, { "epoch": 0.92, "learning_rate": 3.64216668828854e-07, "loss": 0.2488, "step": 18353 }, { "epoch": 0.92, "learning_rate": 3.637843103705485e-07, "loss": 0.2748, "step": 18354 }, { "epoch": 0.92, "learning_rate": 3.633522039348436e-07, "loss": 0.5176, "step": 18355 }, { "epoch": 0.92, "learning_rate": 3.629203495330391e-07, "loss": 0.4524, "step": 18356 }, { "epoch": 0.92, "learning_rate": 3.6248874717643047e-07, "loss": 0.2518, "step": 18357 }, { "epoch": 0.92, "learning_rate": 3.6205739687630527e-07, "loss": 0.5213, "step": 18358 }, { "epoch": 0.92, "learning_rate": 3.616262986439467e-07, "loss": 0.4071, "step": 18359 }, { "epoch": 0.92, "learning_rate": 3.6119545249062806e-07, "loss": 0.1997, "step": 18360 }, { "epoch": 0.92, "learning_rate": 3.6076485842761797e-07, "loss": 0.4112, "step": 18361 }, { "epoch": 0.92, "learning_rate": 3.603345164661776e-07, "loss": 0.3214, "step": 18362 }, { "epoch": 0.92, "learning_rate": 3.599044266175611e-07, "loss": 0.2948, "step": 18363 }, { "epoch": 0.92, "learning_rate": 3.594745888930196e-07, "loss": 0.3867, "step": 18364 }, { "epoch": 0.92, "learning_rate": 3.590450033037929e-07, "loss": 0.4741, "step": 18365 }, { "epoch": 0.92, "learning_rate": 3.586156698611165e-07, "loss": 0.4139, "step": 18366 }, { "epoch": 0.92, "learning_rate": 3.5818658857622033e-07, "loss": 0.2746, "step": 18367 }, { "epoch": 0.92, "learning_rate": 3.577577594603232e-07, "loss": 0.2449, "step": 18368 }, { "epoch": 0.92, "learning_rate": 3.57329182524645e-07, "loss": 0.3687, "step": 18369 }, { "epoch": 0.92, "learning_rate": 3.5690085778039116e-07, "loss": 0.5205, "step": 18370 }, { "epoch": 0.92, "learning_rate": 3.5647278523876615e-07, "loss": 0.3987, "step": 18371 }, { "epoch": 0.92, "learning_rate": 3.560449649109632e-07, "loss": 0.3574, "step": 18372 }, { "epoch": 0.92, "learning_rate": 3.5561739680817445e-07, "loss": 0.4031, "step": 18373 }, { "epoch": 0.92, "learning_rate": 3.551900809415798e-07, "loss": 0.4824, "step": 18374 }, { "epoch": 0.92, "learning_rate": 3.547630173223571e-07, "loss": 0.5454, "step": 18375 }, { "epoch": 0.92, "learning_rate": 3.543362059616751e-07, "loss": 0.4748, "step": 18376 }, { "epoch": 0.92, "learning_rate": 3.539096468706937e-07, "loss": 0.3456, "step": 18377 }, { "epoch": 0.92, "learning_rate": 3.534833400605731e-07, "loss": 0.3733, "step": 18378 }, { "epoch": 0.92, "learning_rate": 3.5305728554246077e-07, "loss": 0.1638, "step": 18379 }, { "epoch": 0.92, "learning_rate": 3.526314833275002e-07, "loss": 0.2703, "step": 18380 }, { "epoch": 0.92, "learning_rate": 3.5220593342682797e-07, "loss": 0.3476, "step": 18381 }, { "epoch": 0.92, "learning_rate": 3.5178063585157296e-07, "loss": 0.2197, "step": 18382 }, { "epoch": 0.92, "learning_rate": 3.513555906128574e-07, "loss": 0.2215, "step": 18383 }, { "epoch": 0.92, "learning_rate": 3.5093079772180014e-07, "loss": 0.5819, "step": 18384 }, { "epoch": 0.92, "learning_rate": 3.50506257189509e-07, "loss": 0.3456, "step": 18385 }, { "epoch": 0.92, "learning_rate": 3.5008196902708734e-07, "loss": 0.1906, "step": 18386 }, { "epoch": 0.92, "learning_rate": 3.4965793324563404e-07, "loss": 0.163, "step": 18387 }, { "epoch": 0.92, "learning_rate": 3.4923414985623795e-07, "loss": 0.4563, "step": 18388 }, { "epoch": 0.92, "learning_rate": 3.488106188699825e-07, "loss": 0.1507, "step": 18389 }, { "epoch": 0.92, "learning_rate": 3.483873402979432e-07, "loss": 0.2653, "step": 18390 }, { "epoch": 0.92, "learning_rate": 3.4796431415119234e-07, "loss": 0.3513, "step": 18391 }, { "epoch": 0.92, "learning_rate": 3.475415404407934e-07, "loss": 0.3624, "step": 18392 }, { "epoch": 0.92, "learning_rate": 3.471190191778029e-07, "loss": 0.4468, "step": 18393 }, { "epoch": 0.92, "learning_rate": 3.4669675037327097e-07, "loss": 0.3926, "step": 18394 }, { "epoch": 0.92, "learning_rate": 3.4627473403824106e-07, "loss": 0.3638, "step": 18395 }, { "epoch": 0.92, "learning_rate": 3.458529701837532e-07, "loss": 0.421, "step": 18396 }, { "epoch": 0.92, "learning_rate": 3.4543145882083516e-07, "loss": 0.3469, "step": 18397 }, { "epoch": 0.92, "learning_rate": 3.450101999605138e-07, "loss": 0.3287, "step": 18398 }, { "epoch": 0.92, "learning_rate": 3.445891936138035e-07, "loss": 0.3413, "step": 18399 }, { "epoch": 0.92, "learning_rate": 3.4416843979171555e-07, "loss": 0.394, "step": 18400 }, { "epoch": 0.92, "learning_rate": 3.4374793850525666e-07, "loss": 0.3898, "step": 18401 }, { "epoch": 0.92, "learning_rate": 3.433276897654225e-07, "loss": 0.3532, "step": 18402 }, { "epoch": 0.92, "learning_rate": 3.429076935832054e-07, "loss": 0.2447, "step": 18403 }, { "epoch": 0.92, "learning_rate": 3.424879499695888e-07, "loss": 0.1844, "step": 18404 }, { "epoch": 0.92, "learning_rate": 3.420684589355505e-07, "loss": 0.2566, "step": 18405 }, { "epoch": 0.92, "learning_rate": 3.4164922049206184e-07, "loss": 0.3692, "step": 18406 }, { "epoch": 0.92, "learning_rate": 3.412302346500884e-07, "loss": 0.3127, "step": 18407 }, { "epoch": 0.92, "learning_rate": 3.4081150142058705e-07, "loss": 0.4669, "step": 18408 }, { "epoch": 0.92, "learning_rate": 3.403930208145079e-07, "loss": 0.4029, "step": 18409 }, { "epoch": 0.92, "learning_rate": 3.3997479284279986e-07, "loss": 0.721, "step": 18410 }, { "epoch": 0.92, "learning_rate": 3.395568175163977e-07, "loss": 0.226, "step": 18411 }, { "epoch": 0.92, "learning_rate": 3.391390948462347e-07, "loss": 0.2665, "step": 18412 }, { "epoch": 0.92, "learning_rate": 3.3872162484323344e-07, "loss": 0.6069, "step": 18413 }, { "epoch": 0.92, "learning_rate": 3.383044075183162e-07, "loss": 0.3059, "step": 18414 }, { "epoch": 0.92, "learning_rate": 3.3788744288239197e-07, "loss": 0.2892, "step": 18415 }, { "epoch": 0.92, "learning_rate": 3.3747073094636653e-07, "loss": 0.41, "step": 18416 }, { "epoch": 0.92, "learning_rate": 3.370542717211389e-07, "loss": 0.3259, "step": 18417 }, { "epoch": 0.92, "learning_rate": 3.366380652175993e-07, "loss": 0.3025, "step": 18418 }, { "epoch": 0.92, "learning_rate": 3.3622211144663577e-07, "loss": 0.3327, "step": 18419 }, { "epoch": 0.92, "learning_rate": 3.3580641041912497e-07, "loss": 0.2378, "step": 18420 }, { "epoch": 0.92, "learning_rate": 3.3539096214593947e-07, "loss": 0.3207, "step": 18421 }, { "epoch": 0.92, "learning_rate": 3.34975766637945e-07, "loss": 0.4036, "step": 18422 }, { "epoch": 0.92, "learning_rate": 3.3456082390600054e-07, "loss": 0.4217, "step": 18423 }, { "epoch": 0.92, "learning_rate": 3.341461339609586e-07, "loss": 0.4551, "step": 18424 }, { "epoch": 0.92, "learning_rate": 3.33731696813665e-07, "loss": 0.2639, "step": 18425 }, { "epoch": 0.92, "learning_rate": 3.333175124749577e-07, "loss": 0.4419, "step": 18426 }, { "epoch": 0.92, "learning_rate": 3.329035809556702e-07, "loss": 0.2979, "step": 18427 }, { "epoch": 0.92, "learning_rate": 3.3248990226662834e-07, "loss": 0.4562, "step": 18428 }, { "epoch": 0.92, "learning_rate": 3.3207647641865013e-07, "loss": 0.406, "step": 18429 }, { "epoch": 0.92, "learning_rate": 3.3166330342254917e-07, "loss": 0.5029, "step": 18430 }, { "epoch": 0.92, "learning_rate": 3.3125038328913006e-07, "loss": 0.1676, "step": 18431 }, { "epoch": 0.92, "learning_rate": 3.3083771602919425e-07, "loss": 0.4308, "step": 18432 }, { "epoch": 0.92, "learning_rate": 3.304253016535341e-07, "loss": 0.3446, "step": 18433 }, { "epoch": 0.92, "learning_rate": 3.3001314017293563e-07, "loss": 0.2973, "step": 18434 }, { "epoch": 0.92, "learning_rate": 3.296012315981778e-07, "loss": 0.4885, "step": 18435 }, { "epoch": 0.92, "learning_rate": 3.291895759400321e-07, "loss": 0.29, "step": 18436 }, { "epoch": 0.92, "learning_rate": 3.2877817320926875e-07, "loss": 0.4026, "step": 18437 }, { "epoch": 0.92, "learning_rate": 3.2836702341664473e-07, "loss": 0.4803, "step": 18438 }, { "epoch": 0.92, "learning_rate": 3.2795612657291367e-07, "loss": 0.3479, "step": 18439 }, { "epoch": 0.92, "learning_rate": 3.2754548268882134e-07, "loss": 0.3071, "step": 18440 }, { "epoch": 0.92, "learning_rate": 3.2713509177510813e-07, "loss": 0.5597, "step": 18441 }, { "epoch": 0.92, "learning_rate": 3.267249538425077e-07, "loss": 0.3416, "step": 18442 }, { "epoch": 0.92, "learning_rate": 3.263150689017469e-07, "loss": 0.3649, "step": 18443 }, { "epoch": 0.92, "learning_rate": 3.2590543696354503e-07, "loss": 0.2846, "step": 18444 }, { "epoch": 0.92, "learning_rate": 3.2549605803861463e-07, "loss": 0.5095, "step": 18445 }, { "epoch": 0.92, "learning_rate": 3.250869321376648e-07, "loss": 0.4154, "step": 18446 }, { "epoch": 0.92, "learning_rate": 3.2467805927139493e-07, "loss": 0.2809, "step": 18447 }, { "epoch": 0.92, "learning_rate": 3.242694394504975e-07, "loss": 0.3453, "step": 18448 }, { "epoch": 0.92, "learning_rate": 3.2386107268565947e-07, "loss": 0.4685, "step": 18449 }, { "epoch": 0.92, "learning_rate": 3.2345295898756126e-07, "loss": 0.581, "step": 18450 }, { "epoch": 0.92, "learning_rate": 3.230450983668776e-07, "loss": 0.1229, "step": 18451 }, { "epoch": 0.92, "learning_rate": 3.2263749083427444e-07, "loss": 0.4695, "step": 18452 }, { "epoch": 0.92, "learning_rate": 3.2223013640041324e-07, "loss": 0.4618, "step": 18453 }, { "epoch": 0.92, "learning_rate": 3.2182303507594657e-07, "loss": 0.5257, "step": 18454 }, { "epoch": 0.92, "learning_rate": 3.2141618687152264e-07, "loss": 0.1475, "step": 18455 }, { "epoch": 0.92, "learning_rate": 3.210095917977829e-07, "loss": 0.3892, "step": 18456 }, { "epoch": 0.92, "learning_rate": 3.2060324986535885e-07, "loss": 0.305, "step": 18457 }, { "epoch": 0.92, "learning_rate": 3.201971610848786e-07, "loss": 0.4804, "step": 18458 }, { "epoch": 0.92, "learning_rate": 3.1979132546696267e-07, "loss": 0.3205, "step": 18459 }, { "epoch": 0.92, "learning_rate": 3.19385743022228e-07, "loss": 0.2737, "step": 18460 }, { "epoch": 0.92, "learning_rate": 3.189804137612784e-07, "loss": 0.3749, "step": 18461 }, { "epoch": 0.92, "learning_rate": 3.185753376947165e-07, "loss": 0.4542, "step": 18462 }, { "epoch": 0.92, "learning_rate": 3.18170514833136e-07, "loss": 0.5507, "step": 18463 }, { "epoch": 0.92, "learning_rate": 3.177659451871251e-07, "loss": 0.3091, "step": 18464 }, { "epoch": 0.92, "learning_rate": 3.1736162876726537e-07, "loss": 0.3429, "step": 18465 }, { "epoch": 0.92, "learning_rate": 3.169575655841295e-07, "loss": 0.524, "step": 18466 }, { "epoch": 0.92, "learning_rate": 3.165537556482856e-07, "loss": 0.3954, "step": 18467 }, { "epoch": 0.92, "learning_rate": 3.161501989702942e-07, "loss": 0.4319, "step": 18468 }, { "epoch": 0.92, "learning_rate": 3.157468955607124e-07, "loss": 0.251, "step": 18469 }, { "epoch": 0.92, "learning_rate": 3.153438454300872e-07, "loss": 0.3542, "step": 18470 }, { "epoch": 0.92, "learning_rate": 3.1494104858895813e-07, "loss": 0.2702, "step": 18471 }, { "epoch": 0.92, "learning_rate": 3.1453850504786e-07, "loss": 0.5809, "step": 18472 }, { "epoch": 0.92, "learning_rate": 3.1413621481732216e-07, "loss": 0.2715, "step": 18473 }, { "epoch": 0.92, "learning_rate": 3.137341779078651e-07, "loss": 0.3928, "step": 18474 }, { "epoch": 0.92, "learning_rate": 3.1333239433000485e-07, "loss": 0.5313, "step": 18475 }, { "epoch": 0.92, "learning_rate": 3.1293086409424856e-07, "loss": 0.1519, "step": 18476 }, { "epoch": 0.92, "learning_rate": 3.125295872110967e-07, "loss": 0.4375, "step": 18477 }, { "epoch": 0.92, "learning_rate": 3.121285636910465e-07, "loss": 0.6303, "step": 18478 }, { "epoch": 0.92, "learning_rate": 3.1172779354458394e-07, "loss": 0.3772, "step": 18479 }, { "epoch": 0.92, "learning_rate": 3.1132727678219177e-07, "loss": 0.3742, "step": 18480 }, { "epoch": 0.92, "learning_rate": 3.1092701341434497e-07, "loss": 0.2445, "step": 18481 }, { "epoch": 0.92, "learning_rate": 3.1052700345151064e-07, "loss": 0.2496, "step": 18482 }, { "epoch": 0.92, "learning_rate": 3.101272469041527e-07, "loss": 0.2621, "step": 18483 }, { "epoch": 0.92, "learning_rate": 3.0972774378272507e-07, "loss": 0.3026, "step": 18484 }, { "epoch": 0.92, "learning_rate": 3.0932849409767705e-07, "loss": 0.2983, "step": 18485 }, { "epoch": 0.92, "learning_rate": 3.0892949785944815e-07, "loss": 0.2008, "step": 18486 }, { "epoch": 0.92, "learning_rate": 3.085307550784755e-07, "loss": 0.3359, "step": 18487 }, { "epoch": 0.92, "learning_rate": 3.081322657651886e-07, "loss": 0.427, "step": 18488 }, { "epoch": 0.92, "learning_rate": 3.077340299300069e-07, "loss": 0.2456, "step": 18489 }, { "epoch": 0.92, "learning_rate": 3.073360475833487e-07, "loss": 0.229, "step": 18490 }, { "epoch": 0.92, "learning_rate": 3.0693831873561895e-07, "loss": 0.531, "step": 18491 }, { "epoch": 0.92, "learning_rate": 3.065408433972228e-07, "loss": 0.3076, "step": 18492 }, { "epoch": 0.92, "learning_rate": 3.0614362157855513e-07, "loss": 0.1185, "step": 18493 }, { "epoch": 0.92, "learning_rate": 3.057466532900044e-07, "loss": 0.2999, "step": 18494 }, { "epoch": 0.92, "learning_rate": 3.0534993854195117e-07, "loss": 0.3873, "step": 18495 }, { "epoch": 0.92, "learning_rate": 3.0495347734477486e-07, "loss": 0.2362, "step": 18496 }, { "epoch": 0.92, "learning_rate": 3.045572697088406e-07, "loss": 0.4964, "step": 18497 }, { "epoch": 0.92, "learning_rate": 3.0416131564451335e-07, "loss": 0.3761, "step": 18498 }, { "epoch": 0.92, "learning_rate": 3.037656151621471e-07, "loss": 0.3765, "step": 18499 }, { "epoch": 0.92, "learning_rate": 3.033701682720913e-07, "loss": 0.3578, "step": 18500 }, { "epoch": 0.92, "learning_rate": 3.029749749846889e-07, "loss": 0.3258, "step": 18501 }, { "epoch": 0.92, "learning_rate": 3.025800353102748e-07, "loss": 0.2416, "step": 18502 }, { "epoch": 0.92, "learning_rate": 3.021853492591775e-07, "loss": 0.2521, "step": 18503 }, { "epoch": 0.92, "learning_rate": 3.0179091684172215e-07, "loss": 0.3505, "step": 18504 }, { "epoch": 0.92, "learning_rate": 3.013967380682226e-07, "loss": 0.2777, "step": 18505 }, { "epoch": 0.92, "learning_rate": 3.010028129489884e-07, "loss": 0.4134, "step": 18506 }, { "epoch": 0.92, "learning_rate": 3.006091414943224e-07, "loss": 0.3828, "step": 18507 }, { "epoch": 0.92, "learning_rate": 3.002157237145198e-07, "loss": 0.1746, "step": 18508 }, { "epoch": 0.92, "learning_rate": 2.9982255961987005e-07, "loss": 0.1799, "step": 18509 }, { "epoch": 0.92, "learning_rate": 2.994296492206572e-07, "loss": 0.4567, "step": 18510 }, { "epoch": 0.92, "learning_rate": 2.9903699252715636e-07, "loss": 0.2923, "step": 18511 }, { "epoch": 0.92, "learning_rate": 2.9864458954963704e-07, "loss": 0.3728, "step": 18512 }, { "epoch": 0.92, "learning_rate": 2.982524402983622e-07, "loss": 0.2141, "step": 18513 }, { "epoch": 0.92, "learning_rate": 2.9786054478358696e-07, "loss": 0.3958, "step": 18514 }, { "epoch": 0.92, "learning_rate": 2.97468903015562e-07, "loss": 0.3892, "step": 18515 }, { "epoch": 0.92, "learning_rate": 2.970775150045302e-07, "loss": 0.3807, "step": 18516 }, { "epoch": 0.92, "learning_rate": 2.9668638076072785e-07, "loss": 0.3461, "step": 18517 }, { "epoch": 0.92, "learning_rate": 2.962955002943824e-07, "loss": 0.4063, "step": 18518 }, { "epoch": 0.92, "learning_rate": 2.9590487361571994e-07, "loss": 0.403, "step": 18519 }, { "epoch": 0.92, "learning_rate": 2.955145007349558e-07, "loss": 0.1272, "step": 18520 }, { "epoch": 0.92, "learning_rate": 2.951243816622984e-07, "loss": 0.2074, "step": 18521 }, { "epoch": 0.92, "learning_rate": 2.9473451640795184e-07, "loss": 0.1707, "step": 18522 }, { "epoch": 0.92, "learning_rate": 2.9434490498211234e-07, "loss": 0.2966, "step": 18523 }, { "epoch": 0.92, "learning_rate": 2.9395554739496957e-07, "loss": 0.397, "step": 18524 }, { "epoch": 0.93, "learning_rate": 2.935664436567076e-07, "loss": 0.3985, "step": 18525 }, { "epoch": 0.93, "learning_rate": 2.9317759377750164e-07, "loss": 0.4312, "step": 18526 }, { "epoch": 0.93, "learning_rate": 2.9278899776752247e-07, "loss": 0.4207, "step": 18527 }, { "epoch": 0.93, "learning_rate": 2.92400655636933e-07, "loss": 0.4381, "step": 18528 }, { "epoch": 0.93, "learning_rate": 2.9201256739588957e-07, "loss": 0.414, "step": 18529 }, { "epoch": 0.93, "learning_rate": 2.9162473305454186e-07, "loss": 0.2817, "step": 18530 }, { "epoch": 0.93, "learning_rate": 2.9123715262303397e-07, "loss": 0.5025, "step": 18531 }, { "epoch": 0.93, "learning_rate": 2.9084982611150113e-07, "loss": 0.3643, "step": 18532 }, { "epoch": 0.93, "learning_rate": 2.9046275353007523e-07, "loss": 0.3634, "step": 18533 }, { "epoch": 0.93, "learning_rate": 2.900759348888793e-07, "loss": 0.4723, "step": 18534 }, { "epoch": 0.93, "learning_rate": 2.8968937019802856e-07, "loss": 0.3751, "step": 18535 }, { "epoch": 0.93, "learning_rate": 2.8930305946763493e-07, "loss": 0.3198, "step": 18536 }, { "epoch": 0.93, "learning_rate": 2.889170027078003e-07, "loss": 0.4818, "step": 18537 }, { "epoch": 0.93, "learning_rate": 2.8853119992862224e-07, "loss": 0.3791, "step": 18538 }, { "epoch": 0.93, "learning_rate": 2.8814565114019147e-07, "loss": 0.3553, "step": 18539 }, { "epoch": 0.93, "learning_rate": 2.87760356352591e-07, "loss": 0.4554, "step": 18540 }, { "epoch": 0.93, "learning_rate": 2.8737531557589624e-07, "loss": 0.4897, "step": 18541 }, { "epoch": 0.93, "learning_rate": 2.869905288201802e-07, "loss": 0.3484, "step": 18542 }, { "epoch": 0.93, "learning_rate": 2.8660599609550585e-07, "loss": 0.5307, "step": 18543 }, { "epoch": 0.93, "learning_rate": 2.862217174119286e-07, "loss": 0.3775, "step": 18544 }, { "epoch": 0.93, "learning_rate": 2.8583769277949924e-07, "loss": 0.1404, "step": 18545 }, { "epoch": 0.93, "learning_rate": 2.8545392220826083e-07, "loss": 0.161, "step": 18546 }, { "epoch": 0.93, "learning_rate": 2.8507040570825315e-07, "loss": 0.4929, "step": 18547 }, { "epoch": 0.93, "learning_rate": 2.8468714328950374e-07, "loss": 0.3749, "step": 18548 }, { "epoch": 0.93, "learning_rate": 2.843041349620379e-07, "loss": 0.5156, "step": 18549 }, { "epoch": 0.93, "learning_rate": 2.8392138073587204e-07, "loss": 0.2857, "step": 18550 }, { "epoch": 0.93, "learning_rate": 2.835388806210171e-07, "loss": 0.2792, "step": 18551 }, { "epoch": 0.93, "learning_rate": 2.831566346274761e-07, "loss": 0.3453, "step": 18552 }, { "epoch": 0.93, "learning_rate": 2.8277464276524666e-07, "loss": 0.4386, "step": 18553 }, { "epoch": 0.93, "learning_rate": 2.823929050443186e-07, "loss": 0.3371, "step": 18554 }, { "epoch": 0.93, "learning_rate": 2.820114214746761e-07, "loss": 0.2599, "step": 18555 }, { "epoch": 0.93, "learning_rate": 2.816301920662967e-07, "loss": 0.5718, "step": 18556 }, { "epoch": 0.93, "learning_rate": 2.812492168291514e-07, "loss": 0.2418, "step": 18557 }, { "epoch": 0.93, "learning_rate": 2.808684957732044e-07, "loss": 0.2789, "step": 18558 }, { "epoch": 0.93, "learning_rate": 2.8048802890841e-07, "loss": 0.2361, "step": 18559 }, { "epoch": 0.93, "learning_rate": 2.801078162447224e-07, "loss": 0.4678, "step": 18560 }, { "epoch": 0.93, "learning_rate": 2.797278577920837e-07, "loss": 0.392, "step": 18561 }, { "epoch": 0.93, "learning_rate": 2.7934815356043146e-07, "loss": 0.2566, "step": 18562 }, { "epoch": 0.93, "learning_rate": 2.7896870355969663e-07, "loss": 0.2808, "step": 18563 }, { "epoch": 0.93, "learning_rate": 2.785895077998024e-07, "loss": 0.4228, "step": 18564 }, { "epoch": 0.93, "learning_rate": 2.7821056629066755e-07, "loss": 0.2235, "step": 18565 }, { "epoch": 0.93, "learning_rate": 2.77831879042203e-07, "loss": 0.4057, "step": 18566 }, { "epoch": 0.93, "learning_rate": 2.774534460643108e-07, "loss": 0.4771, "step": 18567 }, { "epoch": 0.93, "learning_rate": 2.770752673668886e-07, "loss": 0.3358, "step": 18568 }, { "epoch": 0.93, "learning_rate": 2.7669734295982965e-07, "loss": 0.382, "step": 18569 }, { "epoch": 0.93, "learning_rate": 2.7631967285301597e-07, "loss": 0.2631, "step": 18570 }, { "epoch": 0.93, "learning_rate": 2.7594225705632526e-07, "loss": 0.5528, "step": 18571 }, { "epoch": 0.93, "learning_rate": 2.7556509557962964e-07, "loss": 0.4999, "step": 18572 }, { "epoch": 0.93, "learning_rate": 2.751881884327912e-07, "loss": 0.2899, "step": 18573 }, { "epoch": 0.93, "learning_rate": 2.7481153562566864e-07, "loss": 0.4389, "step": 18574 }, { "epoch": 0.93, "learning_rate": 2.74435137168112e-07, "loss": 0.4908, "step": 18575 }, { "epoch": 0.93, "learning_rate": 2.740589930699666e-07, "loss": 0.2616, "step": 18576 }, { "epoch": 0.93, "learning_rate": 2.736831033410703e-07, "loss": 0.3852, "step": 18577 }, { "epoch": 0.93, "learning_rate": 2.7330746799125283e-07, "loss": 0.246, "step": 18578 }, { "epoch": 0.93, "learning_rate": 2.729320870303387e-07, "loss": 0.6562, "step": 18579 }, { "epoch": 0.93, "learning_rate": 2.725569604681466e-07, "loss": 0.4144, "step": 18580 }, { "epoch": 0.93, "learning_rate": 2.7218208831448544e-07, "loss": 0.2794, "step": 18581 }, { "epoch": 0.93, "learning_rate": 2.7180747057915955e-07, "loss": 0.3278, "step": 18582 }, { "epoch": 0.93, "learning_rate": 2.714331072719689e-07, "loss": 0.5377, "step": 18583 }, { "epoch": 0.93, "learning_rate": 2.7105899840270346e-07, "loss": 0.5418, "step": 18584 }, { "epoch": 0.93, "learning_rate": 2.7068514398114755e-07, "loss": 0.4335, "step": 18585 }, { "epoch": 0.93, "learning_rate": 2.703115440170778e-07, "loss": 0.259, "step": 18586 }, { "epoch": 0.93, "learning_rate": 2.699381985202654e-07, "loss": 0.4248, "step": 18587 }, { "epoch": 0.93, "learning_rate": 2.695651075004768e-07, "loss": 0.4111, "step": 18588 }, { "epoch": 0.93, "learning_rate": 2.691922709674677e-07, "loss": 0.3142, "step": 18589 }, { "epoch": 0.93, "learning_rate": 2.688196889309902e-07, "loss": 0.4705, "step": 18590 }, { "epoch": 0.93, "learning_rate": 2.684473614007865e-07, "loss": 0.5887, "step": 18591 }, { "epoch": 0.93, "learning_rate": 2.680752883865967e-07, "loss": 0.3657, "step": 18592 }, { "epoch": 0.93, "learning_rate": 2.677034698981518e-07, "loss": 0.2955, "step": 18593 }, { "epoch": 0.93, "learning_rate": 2.673319059451762e-07, "loss": 0.2665, "step": 18594 }, { "epoch": 0.93, "learning_rate": 2.669605965373867e-07, "loss": 0.3402, "step": 18595 }, { "epoch": 0.93, "learning_rate": 2.6658954168449324e-07, "loss": 0.3332, "step": 18596 }, { "epoch": 0.93, "learning_rate": 2.6621874139620363e-07, "loss": 0.4089, "step": 18597 }, { "epoch": 0.93, "learning_rate": 2.6584819568221344e-07, "loss": 0.3383, "step": 18598 }, { "epoch": 0.93, "learning_rate": 2.654779045522138e-07, "loss": 0.512, "step": 18599 }, { "epoch": 0.93, "learning_rate": 2.651078680158903e-07, "loss": 0.3902, "step": 18600 }, { "epoch": 0.93, "learning_rate": 2.647380860829196e-07, "loss": 0.4707, "step": 18601 }, { "epoch": 0.93, "learning_rate": 2.643685587629741e-07, "loss": 0.4212, "step": 18602 }, { "epoch": 0.93, "learning_rate": 2.63999286065717e-07, "loss": 0.6347, "step": 18603 }, { "epoch": 0.93, "learning_rate": 2.636302680008063e-07, "loss": 0.2284, "step": 18604 }, { "epoch": 0.93, "learning_rate": 2.6326150457789304e-07, "loss": 0.3492, "step": 18605 }, { "epoch": 0.93, "learning_rate": 2.62892995806624e-07, "loss": 0.3388, "step": 18606 }, { "epoch": 0.93, "learning_rate": 2.6252474169663366e-07, "loss": 0.2849, "step": 18607 }, { "epoch": 0.93, "learning_rate": 2.621567422575566e-07, "loss": 0.6015, "step": 18608 }, { "epoch": 0.93, "learning_rate": 2.617889974990151e-07, "loss": 0.4424, "step": 18609 }, { "epoch": 0.93, "learning_rate": 2.6142150743062587e-07, "loss": 0.3483, "step": 18610 }, { "epoch": 0.93, "learning_rate": 2.610542720620035e-07, "loss": 0.2259, "step": 18611 }, { "epoch": 0.93, "learning_rate": 2.606872914027514e-07, "loss": 0.4089, "step": 18612 }, { "epoch": 0.93, "learning_rate": 2.603205654624663e-07, "loss": 0.1526, "step": 18613 }, { "epoch": 0.93, "learning_rate": 2.5995409425073946e-07, "loss": 0.1774, "step": 18614 }, { "epoch": 0.93, "learning_rate": 2.595878777771565e-07, "loss": 0.257, "step": 18615 }, { "epoch": 0.93, "learning_rate": 2.5922191605129544e-07, "loss": 0.1972, "step": 18616 }, { "epoch": 0.93, "learning_rate": 2.5885620908272737e-07, "loss": 0.3017, "step": 18617 }, { "epoch": 0.93, "learning_rate": 2.5849075688101577e-07, "loss": 0.3902, "step": 18618 }, { "epoch": 0.93, "learning_rate": 2.581255594557186e-07, "loss": 0.2999, "step": 18619 }, { "epoch": 0.93, "learning_rate": 2.577606168163893e-07, "loss": 0.4211, "step": 18620 }, { "epoch": 0.93, "learning_rate": 2.573959289725714e-07, "loss": 0.1491, "step": 18621 }, { "epoch": 0.93, "learning_rate": 2.570314959338016e-07, "loss": 0.323, "step": 18622 }, { "epoch": 0.93, "learning_rate": 2.566673177096124e-07, "loss": 0.2583, "step": 18623 }, { "epoch": 0.93, "learning_rate": 2.5630339430952724e-07, "loss": 0.4211, "step": 18624 }, { "epoch": 0.93, "learning_rate": 2.559397257430651e-07, "loss": 0.4055, "step": 18625 }, { "epoch": 0.93, "learning_rate": 2.5557631201973743e-07, "loss": 0.5078, "step": 18626 }, { "epoch": 0.93, "learning_rate": 2.552131531490487e-07, "loss": 0.4204, "step": 18627 }, { "epoch": 0.93, "learning_rate": 2.548502491404958e-07, "loss": 0.331, "step": 18628 }, { "epoch": 0.93, "learning_rate": 2.544876000035712e-07, "loss": 0.5384, "step": 18629 }, { "epoch": 0.93, "learning_rate": 2.5412520574775834e-07, "loss": 0.1702, "step": 18630 }, { "epoch": 0.93, "learning_rate": 2.5376306638253746e-07, "loss": 0.4014, "step": 18631 }, { "epoch": 0.93, "learning_rate": 2.534011819173765e-07, "loss": 0.1938, "step": 18632 }, { "epoch": 0.93, "learning_rate": 2.5303955236174237e-07, "loss": 0.3737, "step": 18633 }, { "epoch": 0.93, "learning_rate": 2.5267817772509304e-07, "loss": 0.3873, "step": 18634 }, { "epoch": 0.93, "learning_rate": 2.523170580168799e-07, "loss": 0.4337, "step": 18635 }, { "epoch": 0.93, "learning_rate": 2.519561932465464e-07, "loss": 0.4264, "step": 18636 }, { "epoch": 0.93, "learning_rate": 2.515955834235295e-07, "loss": 0.3583, "step": 18637 }, { "epoch": 0.93, "learning_rate": 2.512352285572639e-07, "loss": 0.3828, "step": 18638 }, { "epoch": 0.93, "learning_rate": 2.5087512865717087e-07, "loss": 0.4467, "step": 18639 }, { "epoch": 0.93, "learning_rate": 2.505152837326708e-07, "loss": 0.3062, "step": 18640 }, { "epoch": 0.93, "learning_rate": 2.5015569379317484e-07, "loss": 0.3769, "step": 18641 }, { "epoch": 0.93, "learning_rate": 2.497963588480845e-07, "loss": 0.2396, "step": 18642 }, { "epoch": 0.93, "learning_rate": 2.4943727890680114e-07, "loss": 0.1878, "step": 18643 }, { "epoch": 0.93, "learning_rate": 2.4907845397871497e-07, "loss": 0.1932, "step": 18644 }, { "epoch": 0.93, "learning_rate": 2.487198840732108e-07, "loss": 0.3175, "step": 18645 }, { "epoch": 0.93, "learning_rate": 2.483615691996655e-07, "loss": 0.5031, "step": 18646 }, { "epoch": 0.93, "learning_rate": 2.4800350936745043e-07, "loss": 0.3452, "step": 18647 }, { "epoch": 0.93, "learning_rate": 2.476457045859315e-07, "loss": 0.34, "step": 18648 }, { "epoch": 0.93, "learning_rate": 2.472881548644657e-07, "loss": 0.1982, "step": 18649 }, { "epoch": 0.93, "learning_rate": 2.469308602124043e-07, "loss": 0.4967, "step": 18650 }, { "epoch": 0.93, "learning_rate": 2.465738206390922e-07, "loss": 0.3877, "step": 18651 }, { "epoch": 0.93, "learning_rate": 2.462170361538674e-07, "loss": 0.1697, "step": 18652 }, { "epoch": 0.93, "learning_rate": 2.4586050676606135e-07, "loss": 0.3424, "step": 18653 }, { "epoch": 0.93, "learning_rate": 2.4550423248499764e-07, "loss": 0.3375, "step": 18654 }, { "epoch": 0.93, "learning_rate": 2.451482133199934e-07, "loss": 0.4464, "step": 18655 }, { "epoch": 0.93, "learning_rate": 2.447924492803622e-07, "loss": 0.4618, "step": 18656 }, { "epoch": 0.93, "learning_rate": 2.4443694037540765e-07, "loss": 0.3779, "step": 18657 }, { "epoch": 0.93, "learning_rate": 2.440816866144269e-07, "loss": 0.3642, "step": 18658 }, { "epoch": 0.93, "learning_rate": 2.437266880067124e-07, "loss": 0.2616, "step": 18659 }, { "epoch": 0.93, "learning_rate": 2.433719445615468e-07, "loss": 0.325, "step": 18660 }, { "epoch": 0.93, "learning_rate": 2.4301745628820926e-07, "loss": 0.3527, "step": 18661 }, { "epoch": 0.93, "learning_rate": 2.4266322319597136e-07, "loss": 0.336, "step": 18662 }, { "epoch": 0.93, "learning_rate": 2.4230924529409785e-07, "loss": 0.2824, "step": 18663 }, { "epoch": 0.93, "learning_rate": 2.419555225918435e-07, "loss": 0.4937, "step": 18664 }, { "epoch": 0.93, "learning_rate": 2.416020550984632e-07, "loss": 0.3925, "step": 18665 }, { "epoch": 0.93, "learning_rate": 2.412488428232007e-07, "loss": 0.3993, "step": 18666 }, { "epoch": 0.93, "learning_rate": 2.408958857752919e-07, "loss": 0.4623, "step": 18667 }, { "epoch": 0.93, "learning_rate": 2.4054318396396935e-07, "loss": 0.271, "step": 18668 }, { "epoch": 0.93, "learning_rate": 2.401907373984569e-07, "loss": 0.4342, "step": 18669 }, { "epoch": 0.93, "learning_rate": 2.3983854608797266e-07, "loss": 0.4172, "step": 18670 }, { "epoch": 0.93, "learning_rate": 2.3948661004172926e-07, "loss": 0.3797, "step": 18671 }, { "epoch": 0.93, "learning_rate": 2.3913492926892823e-07, "loss": 0.4786, "step": 18672 }, { "epoch": 0.93, "learning_rate": 2.387835037787689e-07, "loss": 0.1812, "step": 18673 }, { "epoch": 0.93, "learning_rate": 2.384323335804428e-07, "loss": 0.3645, "step": 18674 }, { "epoch": 0.93, "learning_rate": 2.380814186831326e-07, "loss": 0.5229, "step": 18675 }, { "epoch": 0.93, "learning_rate": 2.3773075909601762e-07, "loss": 0.3386, "step": 18676 }, { "epoch": 0.93, "learning_rate": 2.3738035482826827e-07, "loss": 0.2225, "step": 18677 }, { "epoch": 0.93, "learning_rate": 2.3703020588904835e-07, "loss": 0.3176, "step": 18678 }, { "epoch": 0.93, "learning_rate": 2.366803122875172e-07, "loss": 0.2935, "step": 18679 }, { "epoch": 0.93, "learning_rate": 2.3633067403282418e-07, "loss": 0.5365, "step": 18680 }, { "epoch": 0.93, "learning_rate": 2.3598129113411417e-07, "loss": 0.2257, "step": 18681 }, { "epoch": 0.93, "learning_rate": 2.3563216360052543e-07, "loss": 0.4374, "step": 18682 }, { "epoch": 0.93, "learning_rate": 2.3528329144118844e-07, "loss": 0.6051, "step": 18683 }, { "epoch": 0.93, "learning_rate": 2.34934674665227e-07, "loss": 0.3326, "step": 18684 }, { "epoch": 0.93, "learning_rate": 2.345863132817594e-07, "loss": 0.3893, "step": 18685 }, { "epoch": 0.93, "learning_rate": 2.3423820729989722e-07, "loss": 0.3483, "step": 18686 }, { "epoch": 0.93, "learning_rate": 2.3389035672874205e-07, "loss": 0.3162, "step": 18687 }, { "epoch": 0.93, "learning_rate": 2.335427615773944e-07, "loss": 0.2077, "step": 18688 }, { "epoch": 0.93, "learning_rate": 2.3319542185494483e-07, "loss": 0.2772, "step": 18689 }, { "epoch": 0.93, "learning_rate": 2.3284833757047598e-07, "loss": 0.2139, "step": 18690 }, { "epoch": 0.93, "learning_rate": 2.3250150873306621e-07, "loss": 0.1571, "step": 18691 }, { "epoch": 0.93, "learning_rate": 2.321549353517849e-07, "loss": 0.2466, "step": 18692 }, { "epoch": 0.93, "learning_rate": 2.3180861743570038e-07, "loss": 0.312, "step": 18693 }, { "epoch": 0.93, "learning_rate": 2.3146255499386538e-07, "loss": 0.3413, "step": 18694 }, { "epoch": 0.93, "learning_rate": 2.311167480353338e-07, "loss": 0.432, "step": 18695 }, { "epoch": 0.93, "learning_rate": 2.307711965691495e-07, "loss": 0.452, "step": 18696 }, { "epoch": 0.93, "learning_rate": 2.3042590060434743e-07, "loss": 0.2243, "step": 18697 }, { "epoch": 0.93, "learning_rate": 2.3008086014996155e-07, "loss": 0.3163, "step": 18698 }, { "epoch": 0.93, "learning_rate": 2.297360752150135e-07, "loss": 0.3587, "step": 18699 }, { "epoch": 0.93, "learning_rate": 2.2939154580852274e-07, "loss": 0.2389, "step": 18700 }, { "epoch": 0.93, "learning_rate": 2.290472719394965e-07, "loss": 0.2651, "step": 18701 }, { "epoch": 0.93, "learning_rate": 2.2870325361694312e-07, "loss": 0.67, "step": 18702 }, { "epoch": 0.93, "learning_rate": 2.2835949084985875e-07, "loss": 0.2627, "step": 18703 }, { "epoch": 0.93, "learning_rate": 2.2801598364723176e-07, "loss": 0.3042, "step": 18704 }, { "epoch": 0.93, "learning_rate": 2.276727320180483e-07, "loss": 0.526, "step": 18705 }, { "epoch": 0.93, "learning_rate": 2.273297359712845e-07, "loss": 0.2463, "step": 18706 }, { "epoch": 0.93, "learning_rate": 2.2698699551591207e-07, "loss": 0.3353, "step": 18707 }, { "epoch": 0.93, "learning_rate": 2.26644510660895e-07, "loss": 0.2417, "step": 18708 }, { "epoch": 0.93, "learning_rate": 2.263022814151894e-07, "loss": 0.3988, "step": 18709 }, { "epoch": 0.93, "learning_rate": 2.259603077877459e-07, "loss": 0.2529, "step": 18710 }, { "epoch": 0.93, "learning_rate": 2.256185897875096e-07, "loss": 0.4008, "step": 18711 }, { "epoch": 0.93, "learning_rate": 2.2527712742341668e-07, "loss": 0.2469, "step": 18712 }, { "epoch": 0.93, "learning_rate": 2.2493592070439773e-07, "loss": 0.7226, "step": 18713 }, { "epoch": 0.93, "learning_rate": 2.2459496963937787e-07, "loss": 0.2249, "step": 18714 }, { "epoch": 0.93, "learning_rate": 2.2425427423727109e-07, "loss": 0.3716, "step": 18715 }, { "epoch": 0.93, "learning_rate": 2.2391383450699134e-07, "loss": 0.3906, "step": 18716 }, { "epoch": 0.93, "learning_rate": 2.235736504574404e-07, "loss": 0.2634, "step": 18717 }, { "epoch": 0.93, "learning_rate": 2.232337220975156e-07, "loss": 0.4378, "step": 18718 }, { "epoch": 0.93, "learning_rate": 2.228940494361076e-07, "loss": 0.3462, "step": 18719 }, { "epoch": 0.93, "learning_rate": 2.225546324821004e-07, "loss": 0.2384, "step": 18720 }, { "epoch": 0.93, "learning_rate": 2.2221547124437027e-07, "loss": 0.5246, "step": 18721 }, { "epoch": 0.93, "learning_rate": 2.2187656573178896e-07, "loss": 0.607, "step": 18722 }, { "epoch": 0.93, "learning_rate": 2.2153791595321829e-07, "loss": 0.4067, "step": 18723 }, { "epoch": 0.93, "learning_rate": 2.211995219175156e-07, "loss": 0.3036, "step": 18724 }, { "epoch": 0.93, "learning_rate": 2.208613836335316e-07, "loss": 0.2647, "step": 18725 }, { "epoch": 0.94, "learning_rate": 2.2052350111011033e-07, "loss": 0.3549, "step": 18726 }, { "epoch": 0.94, "learning_rate": 2.2018587435608808e-07, "loss": 0.4013, "step": 18727 }, { "epoch": 0.94, "learning_rate": 2.1984850338029328e-07, "loss": 0.5023, "step": 18728 }, { "epoch": 0.94, "learning_rate": 2.1951138819155337e-07, "loss": 0.391, "step": 18729 }, { "epoch": 0.94, "learning_rate": 2.1917452879868238e-07, "loss": 0.3178, "step": 18730 }, { "epoch": 0.94, "learning_rate": 2.1883792521048997e-07, "loss": 0.3399, "step": 18731 }, { "epoch": 0.94, "learning_rate": 2.1850157743578237e-07, "loss": 0.317, "step": 18732 }, { "epoch": 0.94, "learning_rate": 2.1816548548335258e-07, "loss": 0.4719, "step": 18733 }, { "epoch": 0.94, "learning_rate": 2.1782964936199357e-07, "loss": 0.4009, "step": 18734 }, { "epoch": 0.94, "learning_rate": 2.174940690804872e-07, "loss": 0.3906, "step": 18735 }, { "epoch": 0.94, "learning_rate": 2.17158744647612e-07, "loss": 0.4414, "step": 18736 }, { "epoch": 0.94, "learning_rate": 2.1682367607213428e-07, "loss": 0.4229, "step": 18737 }, { "epoch": 0.94, "learning_rate": 2.164888633628215e-07, "loss": 0.3243, "step": 18738 }, { "epoch": 0.94, "learning_rate": 2.1615430652842883e-07, "loss": 0.3415, "step": 18739 }, { "epoch": 0.94, "learning_rate": 2.1582000557770487e-07, "loss": 0.597, "step": 18740 }, { "epoch": 0.94, "learning_rate": 2.1548596051939374e-07, "loss": 0.2628, "step": 18741 }, { "epoch": 0.94, "learning_rate": 2.1515217136223177e-07, "loss": 0.3108, "step": 18742 }, { "epoch": 0.94, "learning_rate": 2.1481863811494864e-07, "loss": 0.2588, "step": 18743 }, { "epoch": 0.94, "learning_rate": 2.1448536078626846e-07, "loss": 0.2287, "step": 18744 }, { "epoch": 0.94, "learning_rate": 2.1415233938490655e-07, "loss": 0.4357, "step": 18745 }, { "epoch": 0.94, "learning_rate": 2.1381957391957365e-07, "loss": 0.4945, "step": 18746 }, { "epoch": 0.94, "learning_rate": 2.1348706439897171e-07, "loss": 0.3259, "step": 18747 }, { "epoch": 0.94, "learning_rate": 2.1315481083179824e-07, "loss": 0.2083, "step": 18748 }, { "epoch": 0.94, "learning_rate": 2.128228132267418e-07, "loss": 0.146, "step": 18749 }, { "epoch": 0.94, "learning_rate": 2.124910715924855e-07, "loss": 0.2696, "step": 18750 }, { "epoch": 0.94, "learning_rate": 2.121595859377057e-07, "loss": 0.2973, "step": 18751 }, { "epoch": 0.94, "learning_rate": 2.1182835627107212e-07, "loss": 0.333, "step": 18752 }, { "epoch": 0.94, "learning_rate": 2.1149738260124785e-07, "loss": 0.3506, "step": 18753 }, { "epoch": 0.94, "learning_rate": 2.111666649368893e-07, "loss": 0.4499, "step": 18754 }, { "epoch": 0.94, "learning_rate": 2.108362032866462e-07, "loss": 0.3611, "step": 18755 }, { "epoch": 0.94, "learning_rate": 2.1050599765915835e-07, "loss": 0.3162, "step": 18756 }, { "epoch": 0.94, "learning_rate": 2.1017604806306658e-07, "loss": 0.1998, "step": 18757 }, { "epoch": 0.94, "learning_rate": 2.098463545069962e-07, "loss": 0.3547, "step": 18758 }, { "epoch": 0.94, "learning_rate": 2.0951691699957254e-07, "loss": 0.1583, "step": 18759 }, { "epoch": 0.94, "learning_rate": 2.0918773554940984e-07, "loss": 0.3191, "step": 18760 }, { "epoch": 0.94, "learning_rate": 2.08858810165119e-07, "loss": 0.2563, "step": 18761 }, { "epoch": 0.94, "learning_rate": 2.0853014085530198e-07, "loss": 0.3197, "step": 18762 }, { "epoch": 0.94, "learning_rate": 2.082017276285553e-07, "loss": 0.4275, "step": 18763 }, { "epoch": 0.94, "learning_rate": 2.0787357049346647e-07, "loss": 0.565, "step": 18764 }, { "epoch": 0.94, "learning_rate": 2.0754566945861865e-07, "loss": 0.293, "step": 18765 }, { "epoch": 0.94, "learning_rate": 2.0721802453258833e-07, "loss": 0.382, "step": 18766 }, { "epoch": 0.94, "learning_rate": 2.068906357239453e-07, "loss": 0.4102, "step": 18767 }, { "epoch": 0.94, "learning_rate": 2.065635030412505e-07, "loss": 0.1545, "step": 18768 }, { "epoch": 0.94, "learning_rate": 2.062366264930604e-07, "loss": 0.2405, "step": 18769 }, { "epoch": 0.94, "learning_rate": 2.0591000608792378e-07, "loss": 0.303, "step": 18770 }, { "epoch": 0.94, "learning_rate": 2.055836418343826e-07, "loss": 0.4069, "step": 18771 }, { "epoch": 0.94, "learning_rate": 2.0525753374097235e-07, "loss": 0.3903, "step": 18772 }, { "epoch": 0.94, "learning_rate": 2.049316818162239e-07, "loss": 0.4782, "step": 18773 }, { "epoch": 0.94, "learning_rate": 2.0460608606865606e-07, "loss": 0.3462, "step": 18774 }, { "epoch": 0.94, "learning_rate": 2.0428074650678752e-07, "loss": 0.3385, "step": 18775 }, { "epoch": 0.94, "learning_rate": 2.0395566313912707e-07, "loss": 0.2835, "step": 18776 }, { "epoch": 0.94, "learning_rate": 2.0363083597417454e-07, "loss": 0.4428, "step": 18777 }, { "epoch": 0.94, "learning_rate": 2.033062650204265e-07, "loss": 0.648, "step": 18778 }, { "epoch": 0.94, "learning_rate": 2.029819502863717e-07, "loss": 0.4752, "step": 18779 }, { "epoch": 0.94, "learning_rate": 2.0265789178049222e-07, "loss": 0.4048, "step": 18780 }, { "epoch": 0.94, "learning_rate": 2.0233408951126355e-07, "loss": 0.3721, "step": 18781 }, { "epoch": 0.94, "learning_rate": 2.0201054348715443e-07, "loss": 0.37, "step": 18782 }, { "epoch": 0.94, "learning_rate": 2.016872537166259e-07, "loss": 0.2336, "step": 18783 }, { "epoch": 0.94, "learning_rate": 2.013642202081345e-07, "loss": 0.3178, "step": 18784 }, { "epoch": 0.94, "learning_rate": 2.0104144297012684e-07, "loss": 0.2985, "step": 18785 }, { "epoch": 0.94, "learning_rate": 2.0071892201104726e-07, "loss": 0.4206, "step": 18786 }, { "epoch": 0.94, "learning_rate": 2.00396657339329e-07, "loss": 0.4037, "step": 18787 }, { "epoch": 0.94, "learning_rate": 2.0007464896339978e-07, "loss": 0.2912, "step": 18788 }, { "epoch": 0.94, "learning_rate": 1.9975289689168397e-07, "loss": 0.211, "step": 18789 }, { "epoch": 0.94, "learning_rate": 1.9943140113259486e-07, "loss": 0.5564, "step": 18790 }, { "epoch": 0.94, "learning_rate": 1.9911016169454122e-07, "loss": 0.3725, "step": 18791 }, { "epoch": 0.94, "learning_rate": 1.9878917858592417e-07, "loss": 0.6866, "step": 18792 }, { "epoch": 0.94, "learning_rate": 1.984684518151392e-07, "loss": 0.4307, "step": 18793 }, { "epoch": 0.94, "learning_rate": 1.98147981390574e-07, "loss": 0.4031, "step": 18794 }, { "epoch": 0.94, "learning_rate": 1.9782776732060972e-07, "loss": 0.3281, "step": 18795 }, { "epoch": 0.94, "learning_rate": 1.9750780961362181e-07, "loss": 0.4536, "step": 18796 }, { "epoch": 0.94, "learning_rate": 1.971881082779792e-07, "loss": 0.4031, "step": 18797 }, { "epoch": 0.94, "learning_rate": 1.968686633220418e-07, "loss": 0.2479, "step": 18798 }, { "epoch": 0.94, "learning_rate": 1.965494747541641e-07, "loss": 0.2412, "step": 18799 }, { "epoch": 0.94, "learning_rate": 1.9623054258269491e-07, "loss": 0.6597, "step": 18800 }, { "epoch": 0.94, "learning_rate": 1.959118668159743e-07, "loss": 0.2131, "step": 18801 }, { "epoch": 0.94, "learning_rate": 1.9559344746233888e-07, "loss": 0.3734, "step": 18802 }, { "epoch": 0.94, "learning_rate": 1.9527528453011536e-07, "loss": 0.2994, "step": 18803 }, { "epoch": 0.94, "learning_rate": 1.9495737802762594e-07, "loss": 0.4023, "step": 18804 }, { "epoch": 0.94, "learning_rate": 1.9463972796318286e-07, "loss": 0.385, "step": 18805 }, { "epoch": 0.94, "learning_rate": 1.9432233434509396e-07, "loss": 0.4514, "step": 18806 }, { "epoch": 0.94, "learning_rate": 1.9400519718166254e-07, "loss": 0.2479, "step": 18807 }, { "epoch": 0.94, "learning_rate": 1.936883164811809e-07, "loss": 0.3112, "step": 18808 }, { "epoch": 0.94, "learning_rate": 1.9337169225193797e-07, "loss": 0.3602, "step": 18809 }, { "epoch": 0.94, "learning_rate": 1.9305532450221488e-07, "loss": 0.4201, "step": 18810 }, { "epoch": 0.94, "learning_rate": 1.9273921324028278e-07, "loss": 0.0851, "step": 18811 }, { "epoch": 0.94, "learning_rate": 1.9242335847441283e-07, "loss": 0.3926, "step": 18812 }, { "epoch": 0.94, "learning_rate": 1.9210776021286403e-07, "loss": 0.1647, "step": 18813 }, { "epoch": 0.94, "learning_rate": 1.9179241846389086e-07, "loss": 0.2144, "step": 18814 }, { "epoch": 0.94, "learning_rate": 1.9147733323574002e-07, "loss": 0.3077, "step": 18815 }, { "epoch": 0.94, "learning_rate": 1.9116250453665385e-07, "loss": 0.1959, "step": 18816 }, { "epoch": 0.94, "learning_rate": 1.9084793237486466e-07, "loss": 0.3885, "step": 18817 }, { "epoch": 0.94, "learning_rate": 1.9053361675859915e-07, "loss": 0.2971, "step": 18818 }, { "epoch": 0.94, "learning_rate": 1.9021955769607968e-07, "loss": 0.1401, "step": 18819 }, { "epoch": 0.94, "learning_rate": 1.8990575519551968e-07, "loss": 0.2261, "step": 18820 }, { "epoch": 0.94, "learning_rate": 1.895922092651248e-07, "loss": 0.3414, "step": 18821 }, { "epoch": 0.94, "learning_rate": 1.8927891991309732e-07, "loss": 0.2866, "step": 18822 }, { "epoch": 0.94, "learning_rate": 1.8896588714762854e-07, "loss": 0.3995, "step": 18823 }, { "epoch": 0.94, "learning_rate": 1.8865311097690741e-07, "loss": 0.3245, "step": 18824 }, { "epoch": 0.94, "learning_rate": 1.8834059140911409e-07, "loss": 0.3293, "step": 18825 }, { "epoch": 0.94, "learning_rate": 1.880283284524209e-07, "loss": 0.3685, "step": 18826 }, { "epoch": 0.94, "learning_rate": 1.877163221149958e-07, "loss": 0.5896, "step": 18827 }, { "epoch": 0.94, "learning_rate": 1.8740457240499886e-07, "loss": 0.3911, "step": 18828 }, { "epoch": 0.94, "learning_rate": 1.8709307933058252e-07, "loss": 0.415, "step": 18829 }, { "epoch": 0.94, "learning_rate": 1.8678184289989465e-07, "loss": 0.1952, "step": 18830 }, { "epoch": 0.94, "learning_rate": 1.8647086312107432e-07, "loss": 0.3078, "step": 18831 }, { "epoch": 0.94, "learning_rate": 1.86160140002255e-07, "loss": 0.4368, "step": 18832 }, { "epoch": 0.94, "learning_rate": 1.8584967355156357e-07, "loss": 0.2646, "step": 18833 }, { "epoch": 0.94, "learning_rate": 1.8553946377711908e-07, "loss": 0.3827, "step": 18834 }, { "epoch": 0.94, "learning_rate": 1.852295106870361e-07, "loss": 0.3657, "step": 18835 }, { "epoch": 0.94, "learning_rate": 1.8491981428942042e-07, "loss": 0.1504, "step": 18836 }, { "epoch": 0.94, "learning_rate": 1.8461037459237108e-07, "loss": 0.2944, "step": 18837 }, { "epoch": 0.94, "learning_rate": 1.8430119160398164e-07, "loss": 0.4773, "step": 18838 }, { "epoch": 0.94, "learning_rate": 1.8399226533233783e-07, "loss": 0.2772, "step": 18839 }, { "epoch": 0.94, "learning_rate": 1.8368359578552097e-07, "loss": 0.3485, "step": 18840 }, { "epoch": 0.94, "learning_rate": 1.8337518297160128e-07, "loss": 0.3752, "step": 18841 }, { "epoch": 0.94, "learning_rate": 1.8306702689864674e-07, "loss": 0.2491, "step": 18842 }, { "epoch": 0.94, "learning_rate": 1.8275912757471647e-07, "loss": 0.2086, "step": 18843 }, { "epoch": 0.94, "learning_rate": 1.824514850078629e-07, "loss": 0.1396, "step": 18844 }, { "epoch": 0.94, "learning_rate": 1.8214409920613185e-07, "loss": 0.3334, "step": 18845 }, { "epoch": 0.94, "learning_rate": 1.8183697017756353e-07, "loss": 0.4346, "step": 18846 }, { "epoch": 0.94, "learning_rate": 1.8153009793018818e-07, "loss": 0.4734, "step": 18847 }, { "epoch": 0.94, "learning_rate": 1.8122348247203382e-07, "loss": 0.44, "step": 18848 }, { "epoch": 0.94, "learning_rate": 1.809171238111196e-07, "loss": 0.2843, "step": 18849 }, { "epoch": 0.94, "learning_rate": 1.8061102195545687e-07, "loss": 0.2433, "step": 18850 }, { "epoch": 0.94, "learning_rate": 1.803051769130526e-07, "loss": 0.3495, "step": 18851 }, { "epoch": 0.94, "learning_rate": 1.799995886919037e-07, "loss": 0.2924, "step": 18852 }, { "epoch": 0.94, "learning_rate": 1.7969425730000377e-07, "loss": 0.4928, "step": 18853 }, { "epoch": 0.94, "learning_rate": 1.7938918274533867e-07, "loss": 0.2499, "step": 18854 }, { "epoch": 0.94, "learning_rate": 1.7908436503588756e-07, "loss": 0.1692, "step": 18855 }, { "epoch": 0.94, "learning_rate": 1.787798041796196e-07, "loss": 0.5702, "step": 18856 }, { "epoch": 0.94, "learning_rate": 1.78475500184504e-07, "loss": 0.219, "step": 18857 }, { "epoch": 0.94, "learning_rate": 1.781714530584977e-07, "loss": 0.3923, "step": 18858 }, { "epoch": 0.94, "learning_rate": 1.7786766280955325e-07, "loss": 0.4887, "step": 18859 }, { "epoch": 0.94, "learning_rate": 1.775641294456154e-07, "loss": 0.2912, "step": 18860 }, { "epoch": 0.94, "learning_rate": 1.7726085297462116e-07, "loss": 0.2413, "step": 18861 }, { "epoch": 0.94, "learning_rate": 1.7695783340450523e-07, "loss": 0.2752, "step": 18862 }, { "epoch": 0.94, "learning_rate": 1.766550707431902e-07, "loss": 0.3059, "step": 18863 }, { "epoch": 0.94, "learning_rate": 1.7635256499859644e-07, "loss": 0.235, "step": 18864 }, { "epoch": 0.94, "learning_rate": 1.7605031617863422e-07, "loss": 0.5005, "step": 18865 }, { "epoch": 0.94, "learning_rate": 1.7574832429120837e-07, "loss": 0.3182, "step": 18866 }, { "epoch": 0.94, "learning_rate": 1.7544658934421922e-07, "loss": 0.3966, "step": 18867 }, { "epoch": 0.94, "learning_rate": 1.7514511134555601e-07, "loss": 0.4103, "step": 18868 }, { "epoch": 0.94, "learning_rate": 1.7484389030310357e-07, "loss": 0.2631, "step": 18869 }, { "epoch": 0.94, "learning_rate": 1.745429262247411e-07, "loss": 0.4069, "step": 18870 }, { "epoch": 0.94, "learning_rate": 1.7424221911833906e-07, "loss": 0.3525, "step": 18871 }, { "epoch": 0.94, "learning_rate": 1.7394176899176329e-07, "loss": 0.2277, "step": 18872 }, { "epoch": 0.94, "learning_rate": 1.7364157585286978e-07, "loss": 0.3387, "step": 18873 }, { "epoch": 0.94, "learning_rate": 1.7334163970951e-07, "loss": 0.3503, "step": 18874 }, { "epoch": 0.94, "learning_rate": 1.7304196056952993e-07, "loss": 0.3184, "step": 18875 }, { "epoch": 0.94, "learning_rate": 1.727425384407666e-07, "loss": 0.227, "step": 18876 }, { "epoch": 0.94, "learning_rate": 1.7244337333105045e-07, "loss": 0.5882, "step": 18877 }, { "epoch": 0.94, "learning_rate": 1.7214446524820628e-07, "loss": 0.2345, "step": 18878 }, { "epoch": 0.94, "learning_rate": 1.718458142000501e-07, "loss": 0.278, "step": 18879 }, { "epoch": 0.94, "learning_rate": 1.7154742019439563e-07, "loss": 0.2937, "step": 18880 }, { "epoch": 0.94, "learning_rate": 1.7124928323904556e-07, "loss": 0.4327, "step": 18881 }, { "epoch": 0.94, "learning_rate": 1.7095140334179694e-07, "loss": 0.1651, "step": 18882 }, { "epoch": 0.94, "learning_rate": 1.7065378051044025e-07, "loss": 0.5607, "step": 18883 }, { "epoch": 0.94, "learning_rate": 1.7035641475275923e-07, "loss": 0.2974, "step": 18884 }, { "epoch": 0.94, "learning_rate": 1.700593060765321e-07, "loss": 0.3428, "step": 18885 }, { "epoch": 0.94, "learning_rate": 1.6976245448952934e-07, "loss": 0.2961, "step": 18886 }, { "epoch": 0.94, "learning_rate": 1.694658599995136e-07, "loss": 0.3629, "step": 18887 }, { "epoch": 0.94, "learning_rate": 1.6916952261424313e-07, "loss": 0.295, "step": 18888 }, { "epoch": 0.94, "learning_rate": 1.688734423414673e-07, "loss": 0.3034, "step": 18889 }, { "epoch": 0.94, "learning_rate": 1.68577619188931e-07, "loss": 0.5401, "step": 18890 }, { "epoch": 0.94, "learning_rate": 1.6828205316436918e-07, "loss": 0.2927, "step": 18891 }, { "epoch": 0.94, "learning_rate": 1.6798674427551343e-07, "loss": 0.3734, "step": 18892 }, { "epoch": 0.94, "learning_rate": 1.6769169253008755e-07, "loss": 0.3581, "step": 18893 }, { "epoch": 0.94, "learning_rate": 1.673968979358065e-07, "loss": 0.4164, "step": 18894 }, { "epoch": 0.94, "learning_rate": 1.6710236050038187e-07, "loss": 0.3744, "step": 18895 }, { "epoch": 0.94, "learning_rate": 1.668080802315153e-07, "loss": 0.4084, "step": 18896 }, { "epoch": 0.94, "learning_rate": 1.6651405713690395e-07, "loss": 0.502, "step": 18897 }, { "epoch": 0.94, "learning_rate": 1.6622029122423944e-07, "loss": 0.431, "step": 18898 }, { "epoch": 0.94, "learning_rate": 1.6592678250120232e-07, "loss": 0.209, "step": 18899 }, { "epoch": 0.94, "learning_rate": 1.6563353097546975e-07, "loss": 0.2593, "step": 18900 }, { "epoch": 0.94, "learning_rate": 1.6534053665471227e-07, "loss": 0.2577, "step": 18901 }, { "epoch": 0.94, "learning_rate": 1.6504779954659044e-07, "loss": 0.3302, "step": 18902 }, { "epoch": 0.94, "learning_rate": 1.647553196587637e-07, "loss": 0.1964, "step": 18903 }, { "epoch": 0.94, "learning_rate": 1.6446309699887918e-07, "loss": 0.2876, "step": 18904 }, { "epoch": 0.94, "learning_rate": 1.6417113157457976e-07, "loss": 0.402, "step": 18905 }, { "epoch": 0.94, "learning_rate": 1.638794233935015e-07, "loss": 0.334, "step": 18906 }, { "epoch": 0.94, "learning_rate": 1.6358797246327384e-07, "loss": 0.4093, "step": 18907 }, { "epoch": 0.94, "learning_rate": 1.6329677879151962e-07, "loss": 0.346, "step": 18908 }, { "epoch": 0.94, "learning_rate": 1.6300584238585493e-07, "loss": 0.3407, "step": 18909 }, { "epoch": 0.94, "learning_rate": 1.6271516325388814e-07, "loss": 0.457, "step": 18910 }, { "epoch": 0.94, "learning_rate": 1.6242474140322094e-07, "loss": 0.4945, "step": 18911 }, { "epoch": 0.94, "learning_rate": 1.6213457684145062e-07, "loss": 0.6342, "step": 18912 }, { "epoch": 0.94, "learning_rate": 1.6184466957616552e-07, "loss": 0.2032, "step": 18913 }, { "epoch": 0.94, "learning_rate": 1.6155501961494625e-07, "loss": 0.5322, "step": 18914 }, { "epoch": 0.94, "learning_rate": 1.6126562696537007e-07, "loss": 0.3059, "step": 18915 }, { "epoch": 0.94, "learning_rate": 1.6097649163500539e-07, "loss": 0.3772, "step": 18916 }, { "epoch": 0.94, "learning_rate": 1.6068761363141284e-07, "loss": 0.333, "step": 18917 }, { "epoch": 0.94, "learning_rate": 1.6039899296214968e-07, "loss": 0.1845, "step": 18918 }, { "epoch": 0.94, "learning_rate": 1.601106296347621e-07, "loss": 0.1429, "step": 18919 }, { "epoch": 0.94, "learning_rate": 1.5982252365679295e-07, "loss": 0.2692, "step": 18920 }, { "epoch": 0.94, "learning_rate": 1.5953467503577845e-07, "loss": 0.4459, "step": 18921 }, { "epoch": 0.94, "learning_rate": 1.592470837792448e-07, "loss": 0.3313, "step": 18922 }, { "epoch": 0.94, "learning_rate": 1.5895974989471596e-07, "loss": 0.2694, "step": 18923 }, { "epoch": 0.94, "learning_rate": 1.586726733897048e-07, "loss": 0.1535, "step": 18924 }, { "epoch": 0.94, "learning_rate": 1.5838585427171872e-07, "loss": 0.4643, "step": 18925 }, { "epoch": 0.95, "learning_rate": 1.5809929254826163e-07, "loss": 0.3748, "step": 18926 }, { "epoch": 0.95, "learning_rate": 1.5781298822682756e-07, "loss": 0.53, "step": 18927 }, { "epoch": 0.95, "learning_rate": 1.5752694131490276e-07, "loss": 0.2581, "step": 18928 }, { "epoch": 0.95, "learning_rate": 1.5724115181996902e-07, "loss": 0.2821, "step": 18929 }, { "epoch": 0.95, "learning_rate": 1.5695561974950257e-07, "loss": 0.2781, "step": 18930 }, { "epoch": 0.95, "learning_rate": 1.5667034511096967e-07, "loss": 0.2383, "step": 18931 }, { "epoch": 0.95, "learning_rate": 1.5638532791183102e-07, "loss": 0.2516, "step": 18932 }, { "epoch": 0.95, "learning_rate": 1.5610056815954178e-07, "loss": 0.2015, "step": 18933 }, { "epoch": 0.95, "learning_rate": 1.5581606586154817e-07, "loss": 0.214, "step": 18934 }, { "epoch": 0.95, "learning_rate": 1.5553182102529206e-07, "loss": 0.3802, "step": 18935 }, { "epoch": 0.95, "learning_rate": 1.552478336582075e-07, "loss": 0.4526, "step": 18936 }, { "epoch": 0.95, "learning_rate": 1.5496410376772185e-07, "loss": 0.3297, "step": 18937 }, { "epoch": 0.95, "learning_rate": 1.5468063136125367e-07, "loss": 0.3646, "step": 18938 }, { "epoch": 0.95, "learning_rate": 1.543974164462203e-07, "loss": 0.5673, "step": 18939 }, { "epoch": 0.95, "learning_rate": 1.5411445903002697e-07, "loss": 0.2205, "step": 18940 }, { "epoch": 0.95, "learning_rate": 1.5383175912007332e-07, "loss": 0.3643, "step": 18941 }, { "epoch": 0.95, "learning_rate": 1.535493167237545e-07, "loss": 0.3175, "step": 18942 }, { "epoch": 0.95, "learning_rate": 1.5326713184845688e-07, "loss": 0.3214, "step": 18943 }, { "epoch": 0.95, "learning_rate": 1.5298520450156118e-07, "loss": 0.3259, "step": 18944 }, { "epoch": 0.95, "learning_rate": 1.527035346904393e-07, "loss": 0.34, "step": 18945 }, { "epoch": 0.95, "learning_rate": 1.524221224224587e-07, "loss": 0.4321, "step": 18946 }, { "epoch": 0.95, "learning_rate": 1.521409677049801e-07, "loss": 0.2437, "step": 18947 }, { "epoch": 0.95, "learning_rate": 1.5186007054535435e-07, "loss": 0.4045, "step": 18948 }, { "epoch": 0.95, "learning_rate": 1.5157943095093108e-07, "loss": 0.5371, "step": 18949 }, { "epoch": 0.95, "learning_rate": 1.512990489290489e-07, "loss": 0.2199, "step": 18950 }, { "epoch": 0.95, "learning_rate": 1.5101892448704082e-07, "loss": 0.1869, "step": 18951 }, { "epoch": 0.95, "learning_rate": 1.5073905763223096e-07, "loss": 0.3562, "step": 18952 }, { "epoch": 0.95, "learning_rate": 1.5045944837194238e-07, "loss": 0.3439, "step": 18953 }, { "epoch": 0.95, "learning_rate": 1.50180096713487e-07, "loss": 0.3819, "step": 18954 }, { "epoch": 0.95, "learning_rate": 1.4990100266416896e-07, "loss": 0.3895, "step": 18955 }, { "epoch": 0.95, "learning_rate": 1.4962216623129022e-07, "loss": 0.432, "step": 18956 }, { "epoch": 0.95, "learning_rate": 1.493435874221405e-07, "loss": 0.2725, "step": 18957 }, { "epoch": 0.95, "learning_rate": 1.490652662440073e-07, "loss": 0.3454, "step": 18958 }, { "epoch": 0.95, "learning_rate": 1.4878720270417034e-07, "loss": 0.2985, "step": 18959 }, { "epoch": 0.95, "learning_rate": 1.4850939680990162e-07, "loss": 0.3401, "step": 18960 }, { "epoch": 0.95, "learning_rate": 1.4823184856846528e-07, "loss": 0.1294, "step": 18961 }, { "epoch": 0.95, "learning_rate": 1.4795455798712223e-07, "loss": 0.4744, "step": 18962 }, { "epoch": 0.95, "learning_rate": 1.4767752507312439e-07, "loss": 0.42, "step": 18963 }, { "epoch": 0.95, "learning_rate": 1.47400749833716e-07, "loss": 0.4104, "step": 18964 }, { "epoch": 0.95, "learning_rate": 1.4712423227613683e-07, "loss": 0.3511, "step": 18965 }, { "epoch": 0.95, "learning_rate": 1.4684797240761883e-07, "loss": 0.3077, "step": 18966 }, { "epoch": 0.95, "learning_rate": 1.4657197023538627e-07, "loss": 0.2529, "step": 18967 }, { "epoch": 0.95, "learning_rate": 1.462962257666578e-07, "loss": 0.3404, "step": 18968 }, { "epoch": 0.95, "learning_rate": 1.4602073900864655e-07, "loss": 0.2612, "step": 18969 }, { "epoch": 0.95, "learning_rate": 1.4574550996855452e-07, "loss": 0.4997, "step": 18970 }, { "epoch": 0.95, "learning_rate": 1.4547053865358374e-07, "loss": 0.2716, "step": 18971 }, { "epoch": 0.95, "learning_rate": 1.451958250709229e-07, "loss": 0.3284, "step": 18972 }, { "epoch": 0.95, "learning_rate": 1.4492136922775844e-07, "loss": 0.3064, "step": 18973 }, { "epoch": 0.95, "learning_rate": 1.4464717113126692e-07, "loss": 0.333, "step": 18974 }, { "epoch": 0.95, "learning_rate": 1.4437323078862032e-07, "loss": 0.4806, "step": 18975 }, { "epoch": 0.95, "learning_rate": 1.4409954820698513e-07, "loss": 0.2914, "step": 18976 }, { "epoch": 0.95, "learning_rate": 1.4382612339351565e-07, "loss": 0.1508, "step": 18977 }, { "epoch": 0.95, "learning_rate": 1.4355295635536504e-07, "loss": 0.2868, "step": 18978 }, { "epoch": 0.95, "learning_rate": 1.4328004709967757e-07, "loss": 0.4879, "step": 18979 }, { "epoch": 0.95, "learning_rate": 1.4300739563358978e-07, "loss": 0.4637, "step": 18980 }, { "epoch": 0.95, "learning_rate": 1.4273500196423374e-07, "loss": 0.1796, "step": 18981 }, { "epoch": 0.95, "learning_rate": 1.4246286609873372e-07, "loss": 0.2922, "step": 18982 }, { "epoch": 0.95, "learning_rate": 1.4219098804420518e-07, "loss": 0.3792, "step": 18983 }, { "epoch": 0.95, "learning_rate": 1.4191936780776016e-07, "loss": 0.195, "step": 18984 }, { "epoch": 0.95, "learning_rate": 1.4164800539650302e-07, "loss": 0.3146, "step": 18985 }, { "epoch": 0.95, "learning_rate": 1.413769008175292e-07, "loss": 0.2314, "step": 18986 }, { "epoch": 0.95, "learning_rate": 1.4110605407793077e-07, "loss": 0.4839, "step": 18987 }, { "epoch": 0.95, "learning_rate": 1.4083546518479095e-07, "loss": 0.5653, "step": 18988 }, { "epoch": 0.95, "learning_rate": 1.4056513414518526e-07, "loss": 0.3412, "step": 18989 }, { "epoch": 0.95, "learning_rate": 1.402950609661857e-07, "loss": 0.3814, "step": 18990 }, { "epoch": 0.95, "learning_rate": 1.4002524565485565e-07, "loss": 0.3694, "step": 18991 }, { "epoch": 0.95, "learning_rate": 1.3975568821825048e-07, "loss": 0.2868, "step": 18992 }, { "epoch": 0.95, "learning_rate": 1.39486388663419e-07, "loss": 0.2912, "step": 18993 }, { "epoch": 0.95, "learning_rate": 1.3921734699740785e-07, "loss": 0.4663, "step": 18994 }, { "epoch": 0.95, "learning_rate": 1.3894856322725135e-07, "loss": 0.2877, "step": 18995 }, { "epoch": 0.95, "learning_rate": 1.3868003735997947e-07, "loss": 0.1636, "step": 18996 }, { "epoch": 0.95, "learning_rate": 1.3841176940261548e-07, "loss": 0.2956, "step": 18997 }, { "epoch": 0.95, "learning_rate": 1.3814375936217485e-07, "loss": 0.4311, "step": 18998 }, { "epoch": 0.95, "learning_rate": 1.3787600724566752e-07, "loss": 0.4398, "step": 18999 }, { "epoch": 0.95, "learning_rate": 1.376085130600968e-07, "loss": 0.2785, "step": 19000 }, { "epoch": 0.95, "learning_rate": 1.3734127681245712e-07, "loss": 0.4581, "step": 19001 }, { "epoch": 0.95, "learning_rate": 1.3707429850973842e-07, "loss": 0.2261, "step": 19002 }, { "epoch": 0.95, "learning_rate": 1.3680757815892397e-07, "loss": 0.3454, "step": 19003 }, { "epoch": 0.95, "learning_rate": 1.3654111576698936e-07, "loss": 0.4079, "step": 19004 }, { "epoch": 0.95, "learning_rate": 1.3627491134090232e-07, "loss": 0.3937, "step": 19005 }, { "epoch": 0.95, "learning_rate": 1.3600896488762617e-07, "loss": 0.5206, "step": 19006 }, { "epoch": 0.95, "learning_rate": 1.3574327641411533e-07, "loss": 0.3509, "step": 19007 }, { "epoch": 0.95, "learning_rate": 1.3547784592731982e-07, "loss": 0.2937, "step": 19008 }, { "epoch": 0.95, "learning_rate": 1.3521267343418076e-07, "loss": 0.3786, "step": 19009 }, { "epoch": 0.95, "learning_rate": 1.349477589416337e-07, "loss": 0.31, "step": 19010 }, { "epoch": 0.95, "learning_rate": 1.3468310245660755e-07, "loss": 0.191, "step": 19011 }, { "epoch": 0.95, "learning_rate": 1.3441870398602342e-07, "loss": 0.3078, "step": 19012 }, { "epoch": 0.95, "learning_rate": 1.3415456353679578e-07, "loss": 0.2683, "step": 19013 }, { "epoch": 0.95, "learning_rate": 1.338906811158347e-07, "loss": 0.4479, "step": 19014 }, { "epoch": 0.95, "learning_rate": 1.3362705673004017e-07, "loss": 0.3267, "step": 19015 }, { "epoch": 0.95, "learning_rate": 1.3336369038630781e-07, "loss": 0.3843, "step": 19016 }, { "epoch": 0.95, "learning_rate": 1.3310058209152433e-07, "loss": 0.3468, "step": 19017 }, { "epoch": 0.95, "learning_rate": 1.3283773185257198e-07, "loss": 0.0793, "step": 19018 }, { "epoch": 0.95, "learning_rate": 1.325751396763264e-07, "loss": 0.2065, "step": 19019 }, { "epoch": 0.95, "learning_rate": 1.3231280556965209e-07, "loss": 0.33, "step": 19020 }, { "epoch": 0.95, "learning_rate": 1.3205072953941244e-07, "loss": 0.2413, "step": 19021 }, { "epoch": 0.95, "learning_rate": 1.31788911592462e-07, "loss": 0.2067, "step": 19022 }, { "epoch": 0.95, "learning_rate": 1.3152735173564747e-07, "loss": 0.3136, "step": 19023 }, { "epoch": 0.95, "learning_rate": 1.3126604997580895e-07, "loss": 0.3207, "step": 19024 }, { "epoch": 0.95, "learning_rate": 1.3100500631978098e-07, "loss": 0.3955, "step": 19025 }, { "epoch": 0.95, "learning_rate": 1.3074422077439143e-07, "loss": 0.3586, "step": 19026 }, { "epoch": 0.95, "learning_rate": 1.3048369334646148e-07, "loss": 0.2266, "step": 19027 }, { "epoch": 0.95, "learning_rate": 1.3022342404280242e-07, "loss": 0.4084, "step": 19028 }, { "epoch": 0.95, "learning_rate": 1.2996341287022317e-07, "loss": 0.3798, "step": 19029 }, { "epoch": 0.95, "learning_rate": 1.297036598355228e-07, "loss": 0.3954, "step": 19030 }, { "epoch": 0.95, "learning_rate": 1.2944416494549694e-07, "loss": 0.3124, "step": 19031 }, { "epoch": 0.95, "learning_rate": 1.2918492820692907e-07, "loss": 0.4267, "step": 19032 }, { "epoch": 0.95, "learning_rate": 1.2892594962660265e-07, "loss": 0.0886, "step": 19033 }, { "epoch": 0.95, "learning_rate": 1.2866722921128782e-07, "loss": 0.0547, "step": 19034 }, { "epoch": 0.95, "learning_rate": 1.284087669677525e-07, "loss": 0.2482, "step": 19035 }, { "epoch": 0.95, "learning_rate": 1.2815056290275686e-07, "loss": 0.4551, "step": 19036 }, { "epoch": 0.95, "learning_rate": 1.2789261702305434e-07, "loss": 0.3785, "step": 19037 }, { "epoch": 0.95, "learning_rate": 1.2763492933538958e-07, "loss": 0.1843, "step": 19038 }, { "epoch": 0.95, "learning_rate": 1.273774998465027e-07, "loss": 0.3011, "step": 19039 }, { "epoch": 0.95, "learning_rate": 1.2712032856312618e-07, "loss": 0.3054, "step": 19040 }, { "epoch": 0.95, "learning_rate": 1.2686341549198676e-07, "loss": 0.4363, "step": 19041 }, { "epoch": 0.95, "learning_rate": 1.266067606398036e-07, "loss": 0.1395, "step": 19042 }, { "epoch": 0.95, "learning_rate": 1.2635036401328792e-07, "loss": 0.3528, "step": 19043 }, { "epoch": 0.95, "learning_rate": 1.2609422561914663e-07, "loss": 0.3367, "step": 19044 }, { "epoch": 0.95, "learning_rate": 1.258383454640777e-07, "loss": 0.3459, "step": 19045 }, { "epoch": 0.95, "learning_rate": 1.2558272355477574e-07, "loss": 0.3313, "step": 19046 }, { "epoch": 0.95, "learning_rate": 1.2532735989792322e-07, "loss": 0.4474, "step": 19047 }, { "epoch": 0.95, "learning_rate": 1.2507225450019923e-07, "loss": 0.2808, "step": 19048 }, { "epoch": 0.95, "learning_rate": 1.248174073682784e-07, "loss": 0.3718, "step": 19049 }, { "epoch": 0.95, "learning_rate": 1.2456281850882323e-07, "loss": 0.5555, "step": 19050 }, { "epoch": 0.95, "learning_rate": 1.2430848792849393e-07, "loss": 0.2659, "step": 19051 }, { "epoch": 0.95, "learning_rate": 1.2405441563394072e-07, "loss": 0.1902, "step": 19052 }, { "epoch": 0.95, "learning_rate": 1.238006016318083e-07, "loss": 0.2072, "step": 19053 }, { "epoch": 0.95, "learning_rate": 1.2354704592873578e-07, "loss": 0.3745, "step": 19054 }, { "epoch": 0.95, "learning_rate": 1.2329374853135568e-07, "loss": 0.4565, "step": 19055 }, { "epoch": 0.95, "learning_rate": 1.2304070944629044e-07, "loss": 0.296, "step": 19056 }, { "epoch": 0.95, "learning_rate": 1.227879286801592e-07, "loss": 0.3143, "step": 19057 }, { "epoch": 0.95, "learning_rate": 1.2253540623957228e-07, "loss": 0.1081, "step": 19058 }, { "epoch": 0.95, "learning_rate": 1.2228314213113547e-07, "loss": 0.3241, "step": 19059 }, { "epoch": 0.95, "learning_rate": 1.220311363614457e-07, "loss": 0.3684, "step": 19060 }, { "epoch": 0.95, "learning_rate": 1.217793889370933e-07, "loss": 0.3306, "step": 19061 }, { "epoch": 0.95, "learning_rate": 1.2152789986466408e-07, "loss": 0.2051, "step": 19062 }, { "epoch": 0.95, "learning_rate": 1.2127666915073277e-07, "loss": 0.4686, "step": 19063 }, { "epoch": 0.95, "learning_rate": 1.2102569680187193e-07, "loss": 0.3594, "step": 19064 }, { "epoch": 0.95, "learning_rate": 1.2077498282464405e-07, "loss": 0.4273, "step": 19065 }, { "epoch": 0.95, "learning_rate": 1.2052452722560726e-07, "loss": 0.2651, "step": 19066 }, { "epoch": 0.95, "learning_rate": 1.2027433001131182e-07, "loss": 0.2541, "step": 19067 }, { "epoch": 0.95, "learning_rate": 1.2002439118830144e-07, "loss": 0.2447, "step": 19068 }, { "epoch": 0.95, "learning_rate": 1.1977471076311308e-07, "loss": 0.4716, "step": 19069 }, { "epoch": 0.95, "learning_rate": 1.1952528874227598e-07, "loss": 0.3192, "step": 19070 }, { "epoch": 0.95, "learning_rate": 1.192761251323138e-07, "loss": 0.313, "step": 19071 }, { "epoch": 0.95, "learning_rate": 1.1902721993974242e-07, "loss": 0.3587, "step": 19072 }, { "epoch": 0.95, "learning_rate": 1.1877857317107333e-07, "loss": 0.3654, "step": 19073 }, { "epoch": 0.95, "learning_rate": 1.1853018483280909e-07, "loss": 0.2152, "step": 19074 }, { "epoch": 0.95, "learning_rate": 1.182820549314434e-07, "loss": 0.1046, "step": 19075 }, { "epoch": 0.95, "learning_rate": 1.1803418347346995e-07, "loss": 0.3507, "step": 19076 }, { "epoch": 0.95, "learning_rate": 1.1778657046536801e-07, "loss": 0.2873, "step": 19077 }, { "epoch": 0.95, "learning_rate": 1.1753921591361572e-07, "loss": 0.4093, "step": 19078 }, { "epoch": 0.95, "learning_rate": 1.1729211982468125e-07, "loss": 0.284, "step": 19079 }, { "epoch": 0.95, "learning_rate": 1.170452822050272e-07, "loss": 0.5789, "step": 19080 }, { "epoch": 0.95, "learning_rate": 1.1679870306110952e-07, "loss": 0.4673, "step": 19081 }, { "epoch": 0.95, "learning_rate": 1.165523823993775e-07, "loss": 0.3323, "step": 19082 }, { "epoch": 0.95, "learning_rate": 1.1630632022627264e-07, "loss": 0.2449, "step": 19083 }, { "epoch": 0.95, "learning_rate": 1.1606051654823092e-07, "loss": 0.2763, "step": 19084 }, { "epoch": 0.95, "learning_rate": 1.158149713716794e-07, "loss": 0.2887, "step": 19085 }, { "epoch": 0.95, "learning_rate": 1.1556968470304297e-07, "loss": 0.4617, "step": 19086 }, { "epoch": 0.95, "learning_rate": 1.1532465654873537e-07, "loss": 0.2519, "step": 19087 }, { "epoch": 0.95, "learning_rate": 1.1507988691516369e-07, "loss": 0.3671, "step": 19088 }, { "epoch": 0.95, "learning_rate": 1.1483537580873172e-07, "loss": 0.3208, "step": 19089 }, { "epoch": 0.95, "learning_rate": 1.145911232358321e-07, "loss": 0.1571, "step": 19090 }, { "epoch": 0.95, "learning_rate": 1.1434712920285418e-07, "loss": 0.3239, "step": 19091 }, { "epoch": 0.95, "learning_rate": 1.1410339371618062e-07, "loss": 0.4848, "step": 19092 }, { "epoch": 0.95, "learning_rate": 1.13859916782183e-07, "loss": 0.3506, "step": 19093 }, { "epoch": 0.95, "learning_rate": 1.1361669840723066e-07, "loss": 0.4365, "step": 19094 }, { "epoch": 0.95, "learning_rate": 1.133737385976863e-07, "loss": 0.2374, "step": 19095 }, { "epoch": 0.95, "learning_rate": 1.1313103735990149e-07, "loss": 0.234, "step": 19096 }, { "epoch": 0.95, "learning_rate": 1.1288859470022562e-07, "loss": 0.434, "step": 19097 }, { "epoch": 0.95, "learning_rate": 1.1264641062499693e-07, "loss": 0.3816, "step": 19098 }, { "epoch": 0.95, "learning_rate": 1.1240448514055369e-07, "loss": 0.295, "step": 19099 }, { "epoch": 0.95, "learning_rate": 1.1216281825321973e-07, "loss": 0.2839, "step": 19100 }, { "epoch": 0.95, "learning_rate": 1.1192140996931666e-07, "loss": 0.2446, "step": 19101 }, { "epoch": 0.95, "learning_rate": 1.1168026029515721e-07, "loss": 0.4399, "step": 19102 }, { "epoch": 0.95, "learning_rate": 1.1143936923704967e-07, "loss": 0.2483, "step": 19103 }, { "epoch": 0.95, "learning_rate": 1.1119873680129344e-07, "loss": 0.0606, "step": 19104 }, { "epoch": 0.95, "learning_rate": 1.109583629941835e-07, "loss": 0.2547, "step": 19105 }, { "epoch": 0.95, "learning_rate": 1.107182478220048e-07, "loss": 0.2825, "step": 19106 }, { "epoch": 0.95, "learning_rate": 1.1047839129103677e-07, "loss": 0.3457, "step": 19107 }, { "epoch": 0.95, "learning_rate": 1.1023879340755439e-07, "loss": 0.5433, "step": 19108 }, { "epoch": 0.95, "learning_rate": 1.0999945417782265e-07, "loss": 0.4313, "step": 19109 }, { "epoch": 0.95, "learning_rate": 1.0976037360810209e-07, "loss": 0.444, "step": 19110 }, { "epoch": 0.95, "learning_rate": 1.095215517046444e-07, "loss": 0.2548, "step": 19111 }, { "epoch": 0.95, "learning_rate": 1.0928298847369678e-07, "loss": 0.4683, "step": 19112 }, { "epoch": 0.95, "learning_rate": 1.0904468392149758e-07, "loss": 0.2435, "step": 19113 }, { "epoch": 0.95, "learning_rate": 1.0880663805428072e-07, "loss": 0.354, "step": 19114 }, { "epoch": 0.95, "learning_rate": 1.0856885087827119e-07, "loss": 0.2887, "step": 19115 }, { "epoch": 0.95, "learning_rate": 1.0833132239968624e-07, "loss": 0.2225, "step": 19116 }, { "epoch": 0.95, "learning_rate": 1.0809405262474093e-07, "loss": 0.4045, "step": 19117 }, { "epoch": 0.95, "learning_rate": 1.0785704155964028e-07, "loss": 0.3612, "step": 19118 }, { "epoch": 0.95, "learning_rate": 1.0762028921058155e-07, "loss": 0.4678, "step": 19119 }, { "epoch": 0.95, "learning_rate": 1.0738379558375756e-07, "loss": 0.2552, "step": 19120 }, { "epoch": 0.95, "learning_rate": 1.0714756068535337e-07, "loss": 0.4101, "step": 19121 }, { "epoch": 0.95, "learning_rate": 1.0691158452154738e-07, "loss": 0.4168, "step": 19122 }, { "epoch": 0.95, "learning_rate": 1.066758670985113e-07, "loss": 0.4441, "step": 19123 }, { "epoch": 0.95, "learning_rate": 1.0644040842241021e-07, "loss": 0.2788, "step": 19124 }, { "epoch": 0.95, "learning_rate": 1.0620520849940142e-07, "loss": 0.4128, "step": 19125 }, { "epoch": 0.96, "learning_rate": 1.0597026733563665e-07, "loss": 0.4461, "step": 19126 }, { "epoch": 0.96, "learning_rate": 1.057355849372621e-07, "loss": 0.3983, "step": 19127 }, { "epoch": 0.96, "learning_rate": 1.0550116131041288e-07, "loss": 0.6518, "step": 19128 }, { "epoch": 0.96, "learning_rate": 1.0526699646122185e-07, "loss": 0.2111, "step": 19129 }, { "epoch": 0.96, "learning_rate": 1.0503309039581189e-07, "loss": 0.293, "step": 19130 }, { "epoch": 0.96, "learning_rate": 1.0479944312030254e-07, "loss": 0.4315, "step": 19131 }, { "epoch": 0.96, "learning_rate": 1.0456605464080338e-07, "loss": 0.2147, "step": 19132 }, { "epoch": 0.96, "learning_rate": 1.043329249634184e-07, "loss": 0.4394, "step": 19133 }, { "epoch": 0.96, "learning_rate": 1.0410005409424385e-07, "loss": 0.3194, "step": 19134 }, { "epoch": 0.96, "learning_rate": 1.0386744203937261e-07, "loss": 0.4136, "step": 19135 }, { "epoch": 0.96, "learning_rate": 1.0363508880488537e-07, "loss": 0.0697, "step": 19136 }, { "epoch": 0.96, "learning_rate": 1.0340299439686063e-07, "loss": 0.2506, "step": 19137 }, { "epoch": 0.96, "learning_rate": 1.0317115882136908e-07, "loss": 0.4332, "step": 19138 }, { "epoch": 0.96, "learning_rate": 1.0293958208447252e-07, "loss": 0.3642, "step": 19139 }, { "epoch": 0.96, "learning_rate": 1.0270826419222946e-07, "loss": 0.3405, "step": 19140 }, { "epoch": 0.96, "learning_rate": 1.0247720515068727e-07, "loss": 0.3385, "step": 19141 }, { "epoch": 0.96, "learning_rate": 1.0224640496589111e-07, "loss": 0.1463, "step": 19142 }, { "epoch": 0.96, "learning_rate": 1.0201586364387728e-07, "loss": 0.4332, "step": 19143 }, { "epoch": 0.96, "learning_rate": 1.0178558119067316e-07, "loss": 0.1582, "step": 19144 }, { "epoch": 0.96, "learning_rate": 1.0155555761230284e-07, "loss": 0.2663, "step": 19145 }, { "epoch": 0.96, "learning_rate": 1.0132579291478262e-07, "loss": 0.5376, "step": 19146 }, { "epoch": 0.96, "learning_rate": 1.0109628710412211e-07, "loss": 0.4139, "step": 19147 }, { "epoch": 0.96, "learning_rate": 1.008670401863221e-07, "loss": 0.1832, "step": 19148 }, { "epoch": 0.96, "learning_rate": 1.0063805216737888e-07, "loss": 0.4858, "step": 19149 }, { "epoch": 0.96, "learning_rate": 1.0040932305328211e-07, "loss": 0.4662, "step": 19150 }, { "epoch": 0.96, "learning_rate": 1.0018085285001366e-07, "loss": 0.3681, "step": 19151 }, { "epoch": 0.96, "learning_rate": 9.995264156354878e-08, "loss": 0.5612, "step": 19152 }, { "epoch": 0.96, "learning_rate": 9.97246891998549e-08, "loss": 0.3569, "step": 19153 }, { "epoch": 0.96, "learning_rate": 9.949699576489503e-08, "loss": 0.3523, "step": 19154 }, { "epoch": 0.96, "learning_rate": 9.926956126462439e-08, "loss": 0.3649, "step": 19155 }, { "epoch": 0.96, "learning_rate": 9.904238570499047e-08, "loss": 0.3254, "step": 19156 }, { "epoch": 0.96, "learning_rate": 9.881546909193407e-08, "loss": 0.2844, "step": 19157 }, { "epoch": 0.96, "learning_rate": 9.858881143139154e-08, "loss": 0.2913, "step": 19158 }, { "epoch": 0.96, "learning_rate": 9.836241272929037e-08, "loss": 0.4175, "step": 19159 }, { "epoch": 0.96, "learning_rate": 9.813627299155137e-08, "loss": 0.4623, "step": 19160 }, { "epoch": 0.96, "learning_rate": 9.791039222408982e-08, "loss": 0.2428, "step": 19161 }, { "epoch": 0.96, "learning_rate": 9.7684770432811e-08, "loss": 0.3119, "step": 19162 }, { "epoch": 0.96, "learning_rate": 9.745940762361905e-08, "loss": 0.2681, "step": 19163 }, { "epoch": 0.96, "learning_rate": 9.723430380240484e-08, "loss": 0.311, "step": 19164 }, { "epoch": 0.96, "learning_rate": 9.700945897505809e-08, "loss": 0.3323, "step": 19165 }, { "epoch": 0.96, "learning_rate": 9.678487314745854e-08, "loss": 0.3275, "step": 19166 }, { "epoch": 0.96, "learning_rate": 9.656054632547817e-08, "loss": 0.4273, "step": 19167 }, { "epoch": 0.96, "learning_rate": 9.633647851498672e-08, "loss": 0.2356, "step": 19168 }, { "epoch": 0.96, "learning_rate": 9.611266972184285e-08, "loss": 0.3423, "step": 19169 }, { "epoch": 0.96, "learning_rate": 9.588911995190075e-08, "loss": 0.6357, "step": 19170 }, { "epoch": 0.96, "learning_rate": 9.566582921100576e-08, "loss": 0.1744, "step": 19171 }, { "epoch": 0.96, "learning_rate": 9.544279750499985e-08, "loss": 0.4995, "step": 19172 }, { "epoch": 0.96, "learning_rate": 9.522002483971393e-08, "loss": 0.469, "step": 19173 }, { "epoch": 0.96, "learning_rate": 9.499751122097556e-08, "loss": 0.3964, "step": 19174 }, { "epoch": 0.96, "learning_rate": 9.477525665460452e-08, "loss": 0.2466, "step": 19175 }, { "epoch": 0.96, "learning_rate": 9.455326114641283e-08, "loss": 0.3005, "step": 19176 }, { "epoch": 0.96, "learning_rate": 9.433152470220697e-08, "loss": 0.4288, "step": 19177 }, { "epoch": 0.96, "learning_rate": 9.411004732778562e-08, "loss": 0.3393, "step": 19178 }, { "epoch": 0.96, "learning_rate": 9.388882902894191e-08, "loss": 0.5485, "step": 19179 }, { "epoch": 0.96, "learning_rate": 9.366786981146015e-08, "loss": 0.3344, "step": 19180 }, { "epoch": 0.96, "learning_rate": 9.344716968112233e-08, "loss": 0.5684, "step": 19181 }, { "epoch": 0.96, "learning_rate": 9.32267286436972e-08, "loss": 0.148, "step": 19182 }, { "epoch": 0.96, "learning_rate": 9.300654670495124e-08, "loss": 0.2479, "step": 19183 }, { "epoch": 0.96, "learning_rate": 9.27866238706443e-08, "loss": 0.2968, "step": 19184 }, { "epoch": 0.96, "learning_rate": 9.256696014652622e-08, "loss": 0.3765, "step": 19185 }, { "epoch": 0.96, "learning_rate": 9.234755553834352e-08, "loss": 0.4676, "step": 19186 }, { "epoch": 0.96, "learning_rate": 9.212841005183382e-08, "loss": 0.6293, "step": 19187 }, { "epoch": 0.96, "learning_rate": 9.190952369272809e-08, "loss": 0.3683, "step": 19188 }, { "epoch": 0.96, "learning_rate": 9.169089646675178e-08, "loss": 0.3828, "step": 19189 }, { "epoch": 0.96, "learning_rate": 9.14725283796214e-08, "loss": 0.5937, "step": 19190 }, { "epoch": 0.96, "learning_rate": 9.125441943705016e-08, "loss": 0.1812, "step": 19191 }, { "epoch": 0.96, "learning_rate": 9.103656964474239e-08, "loss": 0.2917, "step": 19192 }, { "epoch": 0.96, "learning_rate": 9.081897900839353e-08, "loss": 0.297, "step": 19193 }, { "epoch": 0.96, "learning_rate": 9.060164753369571e-08, "loss": 0.498, "step": 19194 }, { "epoch": 0.96, "learning_rate": 9.038457522633215e-08, "loss": 0.4583, "step": 19195 }, { "epoch": 0.96, "learning_rate": 9.016776209198164e-08, "loss": 0.3609, "step": 19196 }, { "epoch": 0.96, "learning_rate": 8.995120813631408e-08, "loss": 0.3628, "step": 19197 }, { "epoch": 0.96, "learning_rate": 8.973491336499274e-08, "loss": 0.3431, "step": 19198 }, { "epoch": 0.96, "learning_rate": 8.951887778367418e-08, "loss": 0.3216, "step": 19199 }, { "epoch": 0.96, "learning_rate": 8.930310139800835e-08, "loss": 0.4332, "step": 19200 }, { "epoch": 0.96, "learning_rate": 8.908758421364072e-08, "loss": 0.3901, "step": 19201 }, { "epoch": 0.96, "learning_rate": 8.887232623620568e-08, "loss": 0.1958, "step": 19202 }, { "epoch": 0.96, "learning_rate": 8.865732747133316e-08, "loss": 0.3808, "step": 19203 }, { "epoch": 0.96, "learning_rate": 8.844258792464755e-08, "loss": 0.5432, "step": 19204 }, { "epoch": 0.96, "learning_rate": 8.822810760176326e-08, "loss": 0.3341, "step": 19205 }, { "epoch": 0.96, "learning_rate": 8.801388650829135e-08, "loss": 0.1927, "step": 19206 }, { "epoch": 0.96, "learning_rate": 8.7799924649834e-08, "loss": 0.2876, "step": 19207 }, { "epoch": 0.96, "learning_rate": 8.758622203198785e-08, "loss": 0.4223, "step": 19208 }, { "epoch": 0.96, "learning_rate": 8.737277866033955e-08, "loss": 0.2661, "step": 19209 }, { "epoch": 0.96, "learning_rate": 8.715959454047352e-08, "loss": 0.1386, "step": 19210 }, { "epoch": 0.96, "learning_rate": 8.694666967796528e-08, "loss": 0.342, "step": 19211 }, { "epoch": 0.96, "learning_rate": 8.673400407838262e-08, "loss": 0.1862, "step": 19212 }, { "epoch": 0.96, "learning_rate": 8.652159774728774e-08, "loss": 0.5041, "step": 19213 }, { "epoch": 0.96, "learning_rate": 8.630945069023733e-08, "loss": 0.3289, "step": 19214 }, { "epoch": 0.96, "learning_rate": 8.609756291277804e-08, "loss": 0.3606, "step": 19215 }, { "epoch": 0.96, "learning_rate": 8.588593442045212e-08, "loss": 0.3464, "step": 19216 }, { "epoch": 0.96, "learning_rate": 8.567456521879291e-08, "loss": 0.2944, "step": 19217 }, { "epoch": 0.96, "learning_rate": 8.546345531333155e-08, "loss": 0.3226, "step": 19218 }, { "epoch": 0.96, "learning_rate": 8.525260470958807e-08, "loss": 0.3616, "step": 19219 }, { "epoch": 0.96, "learning_rate": 8.504201341307584e-08, "loss": 0.5493, "step": 19220 }, { "epoch": 0.96, "learning_rate": 8.483168142930487e-08, "loss": 0.3321, "step": 19221 }, { "epoch": 0.96, "learning_rate": 8.4621608763773e-08, "loss": 0.4951, "step": 19222 }, { "epoch": 0.96, "learning_rate": 8.441179542197698e-08, "loss": 0.4374, "step": 19223 }, { "epoch": 0.96, "learning_rate": 8.420224140940348e-08, "loss": 0.3166, "step": 19224 }, { "epoch": 0.96, "learning_rate": 8.399294673153257e-08, "loss": 0.1896, "step": 19225 }, { "epoch": 0.96, "learning_rate": 8.37839113938388e-08, "loss": 0.2265, "step": 19226 }, { "epoch": 0.96, "learning_rate": 8.357513540178885e-08, "loss": 0.3365, "step": 19227 }, { "epoch": 0.96, "learning_rate": 8.336661876084284e-08, "loss": 0.2386, "step": 19228 }, { "epoch": 0.96, "learning_rate": 8.315836147645529e-08, "loss": 0.2626, "step": 19229 }, { "epoch": 0.96, "learning_rate": 8.295036355407182e-08, "loss": 0.1762, "step": 19230 }, { "epoch": 0.96, "learning_rate": 8.274262499913144e-08, "loss": 0.1657, "step": 19231 }, { "epoch": 0.96, "learning_rate": 8.253514581706867e-08, "loss": 0.1611, "step": 19232 }, { "epoch": 0.96, "learning_rate": 8.23279260133103e-08, "loss": 0.344, "step": 19233 }, { "epoch": 0.96, "learning_rate": 8.212096559327531e-08, "loss": 0.3615, "step": 19234 }, { "epoch": 0.96, "learning_rate": 8.191426456237717e-08, "loss": 0.5584, "step": 19235 }, { "epoch": 0.96, "learning_rate": 8.170782292601931e-08, "loss": 0.1993, "step": 19236 }, { "epoch": 0.96, "learning_rate": 8.15016406896041e-08, "loss": 0.4464, "step": 19237 }, { "epoch": 0.96, "learning_rate": 8.129571785852163e-08, "loss": 0.1647, "step": 19238 }, { "epoch": 0.96, "learning_rate": 8.109005443815877e-08, "loss": 0.6145, "step": 19239 }, { "epoch": 0.96, "learning_rate": 8.088465043389448e-08, "loss": 0.4147, "step": 19240 }, { "epoch": 0.96, "learning_rate": 8.067950585110006e-08, "loss": 0.4087, "step": 19241 }, { "epoch": 0.96, "learning_rate": 8.047462069514122e-08, "loss": 0.4061, "step": 19242 }, { "epoch": 0.96, "learning_rate": 8.026999497137588e-08, "loss": 0.3548, "step": 19243 }, { "epoch": 0.96, "learning_rate": 8.006562868515644e-08, "loss": 0.3349, "step": 19244 }, { "epoch": 0.96, "learning_rate": 7.98615218418286e-08, "loss": 0.4116, "step": 19245 }, { "epoch": 0.96, "learning_rate": 7.965767444672923e-08, "loss": 0.4091, "step": 19246 }, { "epoch": 0.96, "learning_rate": 7.945408650519071e-08, "loss": 0.1806, "step": 19247 }, { "epoch": 0.96, "learning_rate": 7.925075802253545e-08, "loss": 0.2152, "step": 19248 }, { "epoch": 0.96, "learning_rate": 7.904768900408366e-08, "loss": 0.5112, "step": 19249 }, { "epoch": 0.96, "learning_rate": 7.884487945514552e-08, "loss": 0.3875, "step": 19250 }, { "epoch": 0.96, "learning_rate": 7.864232938102456e-08, "loss": 0.2916, "step": 19251 }, { "epoch": 0.96, "learning_rate": 7.84400387870199e-08, "loss": 0.4187, "step": 19252 }, { "epoch": 0.96, "learning_rate": 7.823800767841949e-08, "loss": 0.1671, "step": 19253 }, { "epoch": 0.96, "learning_rate": 7.803623606051025e-08, "loss": 0.1452, "step": 19254 }, { "epoch": 0.96, "learning_rate": 7.783472393856794e-08, "loss": 0.2512, "step": 19255 }, { "epoch": 0.96, "learning_rate": 7.763347131786169e-08, "loss": 0.3144, "step": 19256 }, { "epoch": 0.96, "learning_rate": 7.74324782036573e-08, "loss": 0.2214, "step": 19257 }, { "epoch": 0.96, "learning_rate": 7.723174460120941e-08, "loss": 0.4792, "step": 19258 }, { "epoch": 0.96, "learning_rate": 7.703127051576831e-08, "loss": 0.3437, "step": 19259 }, { "epoch": 0.96, "learning_rate": 7.683105595257867e-08, "loss": 0.1475, "step": 19260 }, { "epoch": 0.96, "learning_rate": 7.66311009168752e-08, "loss": 0.4227, "step": 19261 }, { "epoch": 0.96, "learning_rate": 7.643140541388816e-08, "loss": 0.2294, "step": 19262 }, { "epoch": 0.96, "learning_rate": 7.623196944883893e-08, "loss": 0.5404, "step": 19263 }, { "epoch": 0.96, "learning_rate": 7.603279302694667e-08, "loss": 0.2036, "step": 19264 }, { "epoch": 0.96, "learning_rate": 7.583387615341719e-08, "loss": 0.2759, "step": 19265 }, { "epoch": 0.96, "learning_rate": 7.563521883345526e-08, "loss": 0.4622, "step": 19266 }, { "epoch": 0.96, "learning_rate": 7.543682107225448e-08, "loss": 0.2881, "step": 19267 }, { "epoch": 0.96, "learning_rate": 7.523868287500624e-08, "loss": 0.1892, "step": 19268 }, { "epoch": 0.96, "learning_rate": 7.504080424688976e-08, "loss": 0.3158, "step": 19269 }, { "epoch": 0.96, "learning_rate": 7.484318519308309e-08, "loss": 0.2311, "step": 19270 }, { "epoch": 0.96, "learning_rate": 7.464582571875212e-08, "loss": 0.3138, "step": 19271 }, { "epoch": 0.96, "learning_rate": 7.444872582906049e-08, "loss": 0.4858, "step": 19272 }, { "epoch": 0.96, "learning_rate": 7.425188552916185e-08, "loss": 0.2629, "step": 19273 }, { "epoch": 0.96, "learning_rate": 7.405530482420542e-08, "loss": 0.3829, "step": 19274 }, { "epoch": 0.96, "learning_rate": 7.385898371933043e-08, "loss": 0.222, "step": 19275 }, { "epoch": 0.96, "learning_rate": 7.366292221967386e-08, "loss": 0.3683, "step": 19276 }, { "epoch": 0.96, "learning_rate": 7.346712033036275e-08, "loss": 0.249, "step": 19277 }, { "epoch": 0.96, "learning_rate": 7.32715780565163e-08, "loss": 0.0663, "step": 19278 }, { "epoch": 0.96, "learning_rate": 7.307629540325156e-08, "loss": 0.3775, "step": 19279 }, { "epoch": 0.96, "learning_rate": 7.288127237567332e-08, "loss": 0.2979, "step": 19280 }, { "epoch": 0.96, "learning_rate": 7.268650897888418e-08, "loss": 0.2834, "step": 19281 }, { "epoch": 0.96, "learning_rate": 7.249200521797672e-08, "loss": 0.4839, "step": 19282 }, { "epoch": 0.96, "learning_rate": 7.229776109803687e-08, "loss": 0.4692, "step": 19283 }, { "epoch": 0.96, "learning_rate": 7.210377662414724e-08, "loss": 0.6654, "step": 19284 }, { "epoch": 0.96, "learning_rate": 7.191005180138044e-08, "loss": 0.198, "step": 19285 }, { "epoch": 0.96, "learning_rate": 7.171658663480241e-08, "loss": 0.3852, "step": 19286 }, { "epoch": 0.96, "learning_rate": 7.152338112947354e-08, "loss": 0.5247, "step": 19287 }, { "epoch": 0.96, "learning_rate": 7.133043529044648e-08, "loss": 0.4061, "step": 19288 }, { "epoch": 0.96, "learning_rate": 7.113774912276717e-08, "loss": 0.542, "step": 19289 }, { "epoch": 0.96, "learning_rate": 7.094532263147603e-08, "loss": 0.1521, "step": 19290 }, { "epoch": 0.96, "learning_rate": 7.075315582160459e-08, "loss": 0.4211, "step": 19291 }, { "epoch": 0.96, "learning_rate": 7.056124869817994e-08, "loss": 0.3575, "step": 19292 }, { "epoch": 0.96, "learning_rate": 7.036960126622139e-08, "loss": 0.3447, "step": 19293 }, { "epoch": 0.96, "learning_rate": 7.01782135307394e-08, "loss": 0.3085, "step": 19294 }, { "epoch": 0.96, "learning_rate": 6.998708549673994e-08, "loss": 0.2705, "step": 19295 }, { "epoch": 0.96, "learning_rate": 6.979621716922346e-08, "loss": 0.2674, "step": 19296 }, { "epoch": 0.96, "learning_rate": 6.960560855318044e-08, "loss": 0.2445, "step": 19297 }, { "epoch": 0.96, "learning_rate": 6.941525965359686e-08, "loss": 0.3825, "step": 19298 }, { "epoch": 0.96, "learning_rate": 6.922517047544875e-08, "loss": 0.3707, "step": 19299 }, { "epoch": 0.96, "learning_rate": 6.903534102371101e-08, "loss": 0.2945, "step": 19300 }, { "epoch": 0.96, "learning_rate": 6.884577130334525e-08, "loss": 0.329, "step": 19301 }, { "epoch": 0.96, "learning_rate": 6.865646131931192e-08, "loss": 0.2679, "step": 19302 }, { "epoch": 0.96, "learning_rate": 6.846741107656152e-08, "loss": 0.4395, "step": 19303 }, { "epoch": 0.96, "learning_rate": 6.827862058003676e-08, "loss": 0.3125, "step": 19304 }, { "epoch": 0.96, "learning_rate": 6.809008983467701e-08, "loss": 0.2896, "step": 19305 }, { "epoch": 0.96, "learning_rate": 6.790181884541281e-08, "loss": 0.4004, "step": 19306 }, { "epoch": 0.96, "learning_rate": 6.771380761716683e-08, "loss": 0.5036, "step": 19307 }, { "epoch": 0.96, "learning_rate": 6.752605615485852e-08, "loss": 0.1912, "step": 19308 }, { "epoch": 0.96, "learning_rate": 6.733856446339615e-08, "loss": 0.1049, "step": 19309 }, { "epoch": 0.96, "learning_rate": 6.715133254768469e-08, "loss": 0.4726, "step": 19310 }, { "epoch": 0.96, "learning_rate": 6.696436041262023e-08, "loss": 0.3798, "step": 19311 }, { "epoch": 0.96, "learning_rate": 6.677764806309329e-08, "loss": 0.3266, "step": 19312 }, { "epoch": 0.96, "learning_rate": 6.659119550398551e-08, "loss": 0.3872, "step": 19313 }, { "epoch": 0.96, "learning_rate": 6.640500274017636e-08, "loss": 0.3086, "step": 19314 }, { "epoch": 0.96, "learning_rate": 6.621906977653302e-08, "loss": 0.4436, "step": 19315 }, { "epoch": 0.96, "learning_rate": 6.60333966179183e-08, "loss": 0.4019, "step": 19316 }, { "epoch": 0.96, "learning_rate": 6.58479832691894e-08, "loss": 0.2331, "step": 19317 }, { "epoch": 0.96, "learning_rate": 6.566282973519467e-08, "loss": 0.3498, "step": 19318 }, { "epoch": 0.96, "learning_rate": 6.547793602077801e-08, "loss": 0.4473, "step": 19319 }, { "epoch": 0.96, "learning_rate": 6.529330213077334e-08, "loss": 0.3773, "step": 19320 }, { "epoch": 0.96, "learning_rate": 6.510892807001124e-08, "loss": 0.2717, "step": 19321 }, { "epoch": 0.96, "learning_rate": 6.492481384331117e-08, "loss": 0.3628, "step": 19322 }, { "epoch": 0.96, "learning_rate": 6.474095945549041e-08, "loss": 0.2868, "step": 19323 }, { "epoch": 0.96, "learning_rate": 6.455736491135733e-08, "loss": 0.492, "step": 19324 }, { "epoch": 0.96, "learning_rate": 6.437403021571365e-08, "loss": 0.4888, "step": 19325 }, { "epoch": 0.96, "learning_rate": 6.41909553733533e-08, "loss": 0.4074, "step": 19326 }, { "epoch": 0.97, "learning_rate": 6.400814038906467e-08, "loss": 0.147, "step": 19327 }, { "epoch": 0.97, "learning_rate": 6.38255852676306e-08, "loss": 0.1823, "step": 19328 }, { "epoch": 0.97, "learning_rate": 6.364329001382285e-08, "loss": 0.1959, "step": 19329 }, { "epoch": 0.97, "learning_rate": 6.346125463241204e-08, "loss": 0.5147, "step": 19330 }, { "epoch": 0.97, "learning_rate": 6.327947912815657e-08, "loss": 0.4164, "step": 19331 }, { "epoch": 0.97, "learning_rate": 6.309796350581155e-08, "loss": 0.3517, "step": 19332 }, { "epoch": 0.97, "learning_rate": 6.291670777012426e-08, "loss": 0.2358, "step": 19333 }, { "epoch": 0.97, "learning_rate": 6.27357119258365e-08, "loss": 0.3067, "step": 19334 }, { "epoch": 0.97, "learning_rate": 6.25549759776789e-08, "loss": 0.2696, "step": 19335 }, { "epoch": 0.97, "learning_rate": 6.237449993038103e-08, "loss": 0.4739, "step": 19336 }, { "epoch": 0.97, "learning_rate": 6.219428378866244e-08, "loss": 0.3884, "step": 19337 }, { "epoch": 0.97, "learning_rate": 6.201432755723603e-08, "loss": 0.1683, "step": 19338 }, { "epoch": 0.97, "learning_rate": 6.183463124080802e-08, "loss": 0.1509, "step": 19339 }, { "epoch": 0.97, "learning_rate": 6.1655194844078e-08, "loss": 0.2993, "step": 19340 }, { "epoch": 0.97, "learning_rate": 6.147601837174e-08, "loss": 0.1688, "step": 19341 }, { "epoch": 0.97, "learning_rate": 6.129710182848026e-08, "loss": 0.4499, "step": 19342 }, { "epoch": 0.97, "learning_rate": 6.111844521897725e-08, "loss": 0.4085, "step": 19343 }, { "epoch": 0.97, "learning_rate": 6.094004854790391e-08, "loss": 0.4365, "step": 19344 }, { "epoch": 0.97, "learning_rate": 6.076191181992541e-08, "loss": 0.2731, "step": 19345 }, { "epoch": 0.97, "learning_rate": 6.058403503970023e-08, "loss": 0.5844, "step": 19346 }, { "epoch": 0.97, "learning_rate": 6.040641821188242e-08, "loss": 0.2772, "step": 19347 }, { "epoch": 0.97, "learning_rate": 6.022906134111606e-08, "loss": 0.3851, "step": 19348 }, { "epoch": 0.97, "learning_rate": 6.005196443203964e-08, "loss": 0.2281, "step": 19349 }, { "epoch": 0.97, "learning_rate": 5.987512748928503e-08, "loss": 0.1785, "step": 19350 }, { "epoch": 0.97, "learning_rate": 5.969855051747742e-08, "loss": 0.4526, "step": 19351 }, { "epoch": 0.97, "learning_rate": 5.9522233521234205e-08, "loss": 0.448, "step": 19352 }, { "epoch": 0.97, "learning_rate": 5.934617650516838e-08, "loss": 0.3826, "step": 19353 }, { "epoch": 0.97, "learning_rate": 5.917037947388293e-08, "loss": 0.2192, "step": 19354 }, { "epoch": 0.97, "learning_rate": 5.8994842431975285e-08, "loss": 0.3743, "step": 19355 }, { "epoch": 0.97, "learning_rate": 5.8819565384038435e-08, "loss": 0.4468, "step": 19356 }, { "epoch": 0.97, "learning_rate": 5.8644548334654275e-08, "loss": 0.5681, "step": 19357 }, { "epoch": 0.97, "learning_rate": 5.8469791288401354e-08, "loss": 0.3964, "step": 19358 }, { "epoch": 0.97, "learning_rate": 5.8295294249850474e-08, "loss": 0.4181, "step": 19359 }, { "epoch": 0.97, "learning_rate": 5.812105722356465e-08, "loss": 0.2295, "step": 19360 }, { "epoch": 0.97, "learning_rate": 5.794708021410134e-08, "loss": 0.4734, "step": 19361 }, { "epoch": 0.97, "learning_rate": 5.7773363226009127e-08, "loss": 0.2905, "step": 19362 }, { "epoch": 0.97, "learning_rate": 5.759990626383327e-08, "loss": 0.2048, "step": 19363 }, { "epoch": 0.97, "learning_rate": 5.742670933211014e-08, "loss": 0.253, "step": 19364 }, { "epoch": 0.97, "learning_rate": 5.7253772435368335e-08, "loss": 0.2939, "step": 19365 }, { "epoch": 0.97, "learning_rate": 5.7081095578132016e-08, "loss": 0.3753, "step": 19366 }, { "epoch": 0.97, "learning_rate": 5.690867876491646e-08, "loss": 0.2525, "step": 19367 }, { "epoch": 0.97, "learning_rate": 5.6736522000231386e-08, "loss": 0.2564, "step": 19368 }, { "epoch": 0.97, "learning_rate": 5.6564625288579864e-08, "loss": 0.5291, "step": 19369 }, { "epoch": 0.97, "learning_rate": 5.639298863445608e-08, "loss": 0.5459, "step": 19370 }, { "epoch": 0.97, "learning_rate": 5.622161204234977e-08, "loss": 0.4062, "step": 19371 }, { "epoch": 0.97, "learning_rate": 5.605049551674402e-08, "loss": 0.3728, "step": 19372 }, { "epoch": 0.97, "learning_rate": 5.587963906211302e-08, "loss": 0.4361, "step": 19373 }, { "epoch": 0.97, "learning_rate": 5.570904268292543e-08, "loss": 0.4021, "step": 19374 }, { "epoch": 0.97, "learning_rate": 5.553870638364323e-08, "loss": 0.3917, "step": 19375 }, { "epoch": 0.97, "learning_rate": 5.536863016872063e-08, "loss": 0.3028, "step": 19376 }, { "epoch": 0.97, "learning_rate": 5.519881404260741e-08, "loss": 0.2354, "step": 19377 }, { "epoch": 0.97, "learning_rate": 5.5029258009743346e-08, "loss": 0.1901, "step": 19378 }, { "epoch": 0.97, "learning_rate": 5.485996207456379e-08, "loss": 0.2351, "step": 19379 }, { "epoch": 0.97, "learning_rate": 5.46909262414963e-08, "loss": 0.2797, "step": 19380 }, { "epoch": 0.97, "learning_rate": 5.452215051496068e-08, "loss": 0.3866, "step": 19381 }, { "epoch": 0.97, "learning_rate": 5.4353634899372285e-08, "loss": 0.2382, "step": 19382 }, { "epoch": 0.97, "learning_rate": 5.4185379399138703e-08, "loss": 0.5312, "step": 19383 }, { "epoch": 0.97, "learning_rate": 5.401738401865975e-08, "loss": 0.2314, "step": 19384 }, { "epoch": 0.97, "learning_rate": 5.3849648762328563e-08, "loss": 0.2972, "step": 19385 }, { "epoch": 0.97, "learning_rate": 5.3682173634532765e-08, "loss": 0.175, "step": 19386 }, { "epoch": 0.97, "learning_rate": 5.351495863965328e-08, "loss": 0.3613, "step": 19387 }, { "epoch": 0.97, "learning_rate": 5.334800378206217e-08, "loss": 0.2251, "step": 19388 }, { "epoch": 0.97, "learning_rate": 5.318130906612595e-08, "loss": 0.4181, "step": 19389 }, { "epoch": 0.97, "learning_rate": 5.301487449620446e-08, "loss": 0.2704, "step": 19390 }, { "epoch": 0.97, "learning_rate": 5.2848700076650886e-08, "loss": 0.3101, "step": 19391 }, { "epoch": 0.97, "learning_rate": 5.2682785811811746e-08, "loss": 0.5462, "step": 19392 }, { "epoch": 0.97, "learning_rate": 5.2517131706024684e-08, "loss": 0.3894, "step": 19393 }, { "epoch": 0.97, "learning_rate": 5.235173776362401e-08, "loss": 0.5403, "step": 19394 }, { "epoch": 0.97, "learning_rate": 5.2186603988934046e-08, "loss": 0.3305, "step": 19395 }, { "epoch": 0.97, "learning_rate": 5.202173038627356e-08, "loss": 0.3568, "step": 19396 }, { "epoch": 0.97, "learning_rate": 5.185711695995577e-08, "loss": 0.3432, "step": 19397 }, { "epoch": 0.97, "learning_rate": 5.169276371428611e-08, "loss": 0.313, "step": 19398 }, { "epoch": 0.97, "learning_rate": 5.152867065356226e-08, "loss": 0.3157, "step": 19399 }, { "epoch": 0.97, "learning_rate": 5.136483778207413e-08, "loss": 0.3527, "step": 19400 }, { "epoch": 0.97, "learning_rate": 5.120126510410939e-08, "loss": 0.4098, "step": 19401 }, { "epoch": 0.97, "learning_rate": 5.103795262394573e-08, "loss": 0.4342, "step": 19402 }, { "epoch": 0.97, "learning_rate": 5.0874900345853073e-08, "loss": 0.263, "step": 19403 }, { "epoch": 0.97, "learning_rate": 5.0712108274095784e-08, "loss": 0.4689, "step": 19404 }, { "epoch": 0.97, "learning_rate": 5.0549576412932674e-08, "loss": 0.2993, "step": 19405 }, { "epoch": 0.97, "learning_rate": 5.0387304766614796e-08, "loss": 0.2994, "step": 19406 }, { "epoch": 0.97, "learning_rate": 5.022529333938431e-08, "loss": 0.3511, "step": 19407 }, { "epoch": 0.97, "learning_rate": 5.006354213547893e-08, "loss": 0.6438, "step": 19408 }, { "epoch": 0.97, "learning_rate": 4.9902051159129716e-08, "loss": 0.2232, "step": 19409 }, { "epoch": 0.97, "learning_rate": 4.9740820414561076e-08, "loss": 0.482, "step": 19410 }, { "epoch": 0.97, "learning_rate": 4.957984990598741e-08, "loss": 0.4227, "step": 19411 }, { "epoch": 0.97, "learning_rate": 4.941913963762091e-08, "loss": 0.3901, "step": 19412 }, { "epoch": 0.97, "learning_rate": 4.925868961366376e-08, "loss": 0.403, "step": 19413 }, { "epoch": 0.97, "learning_rate": 4.909849983831261e-08, "loss": 0.4545, "step": 19414 }, { "epoch": 0.97, "learning_rate": 4.893857031575633e-08, "loss": 0.3569, "step": 19415 }, { "epoch": 0.97, "learning_rate": 4.877890105017935e-08, "loss": 0.5179, "step": 19416 }, { "epoch": 0.97, "learning_rate": 4.8619492045755004e-08, "loss": 0.367, "step": 19417 }, { "epoch": 0.97, "learning_rate": 4.846034330665439e-08, "loss": 0.4323, "step": 19418 }, { "epoch": 0.97, "learning_rate": 4.830145483703974e-08, "loss": 0.5764, "step": 19419 }, { "epoch": 0.97, "learning_rate": 4.814282664106551e-08, "loss": 0.4264, "step": 19420 }, { "epoch": 0.97, "learning_rate": 4.798445872288171e-08, "loss": 0.1002, "step": 19421 }, { "epoch": 0.97, "learning_rate": 4.782635108662948e-08, "loss": 0.3425, "step": 19422 }, { "epoch": 0.97, "learning_rate": 4.766850373644327e-08, "loss": 0.3443, "step": 19423 }, { "epoch": 0.97, "learning_rate": 4.751091667645313e-08, "loss": 0.2679, "step": 19424 }, { "epoch": 0.97, "learning_rate": 4.73535899107791e-08, "loss": 0.5439, "step": 19425 }, { "epoch": 0.97, "learning_rate": 4.719652344353676e-08, "loss": 0.569, "step": 19426 }, { "epoch": 0.97, "learning_rate": 4.703971727883283e-08, "loss": 0.4909, "step": 19427 }, { "epoch": 0.97, "learning_rate": 4.688317142076848e-08, "loss": 0.3171, "step": 19428 }, { "epoch": 0.97, "learning_rate": 4.6726885873439324e-08, "loss": 0.2381, "step": 19429 }, { "epoch": 0.97, "learning_rate": 4.657086064093319e-08, "loss": 0.2873, "step": 19430 }, { "epoch": 0.97, "learning_rate": 4.6415095727327945e-08, "loss": 0.3746, "step": 19431 }, { "epoch": 0.97, "learning_rate": 4.62595911366992e-08, "loss": 0.3509, "step": 19432 }, { "epoch": 0.97, "learning_rate": 4.6104346873114824e-08, "loss": 0.5542, "step": 19433 }, { "epoch": 0.97, "learning_rate": 4.594936294063268e-08, "loss": 0.4259, "step": 19434 }, { "epoch": 0.97, "learning_rate": 4.579463934330841e-08, "loss": 0.4052, "step": 19435 }, { "epoch": 0.97, "learning_rate": 4.564017608518767e-08, "loss": 0.3538, "step": 19436 }, { "epoch": 0.97, "learning_rate": 4.548597317030945e-08, "loss": 0.3594, "step": 19437 }, { "epoch": 0.97, "learning_rate": 4.53320306027083e-08, "loss": 0.2269, "step": 19438 }, { "epoch": 0.97, "learning_rate": 4.517834838640989e-08, "loss": 0.4316, "step": 19439 }, { "epoch": 0.97, "learning_rate": 4.502492652543211e-08, "loss": 0.2893, "step": 19440 }, { "epoch": 0.97, "learning_rate": 4.4871765023789537e-08, "loss": 0.2133, "step": 19441 }, { "epoch": 0.97, "learning_rate": 4.471886388548674e-08, "loss": 0.3968, "step": 19442 }, { "epoch": 0.97, "learning_rate": 4.4566223114522744e-08, "loss": 0.4756, "step": 19443 }, { "epoch": 0.97, "learning_rate": 4.441384271489102e-08, "loss": 0.2531, "step": 19444 }, { "epoch": 0.97, "learning_rate": 4.426172269057394e-08, "loss": 0.2627, "step": 19445 }, { "epoch": 0.97, "learning_rate": 4.410986304555387e-08, "loss": 0.2944, "step": 19446 }, { "epoch": 0.97, "learning_rate": 4.395826378379875e-08, "loss": 0.2553, "step": 19447 }, { "epoch": 0.97, "learning_rate": 4.3806924909276514e-08, "loss": 0.1638, "step": 19448 }, { "epoch": 0.97, "learning_rate": 4.365584642594289e-08, "loss": 0.2779, "step": 19449 }, { "epoch": 0.97, "learning_rate": 4.350502833775028e-08, "loss": 0.1457, "step": 19450 }, { "epoch": 0.97, "learning_rate": 4.335447064864329e-08, "loss": 0.5179, "step": 19451 }, { "epoch": 0.97, "learning_rate": 4.3204173362559886e-08, "loss": 0.3698, "step": 19452 }, { "epoch": 0.97, "learning_rate": 4.305413648342915e-08, "loss": 0.608, "step": 19453 }, { "epoch": 0.97, "learning_rate": 4.290436001517684e-08, "loss": 0.2994, "step": 19454 }, { "epoch": 0.97, "learning_rate": 4.2754843961719806e-08, "loss": 0.3655, "step": 19455 }, { "epoch": 0.97, "learning_rate": 4.260558832696715e-08, "loss": 0.3578, "step": 19456 }, { "epoch": 0.97, "learning_rate": 4.245659311482464e-08, "loss": 0.3347, "step": 19457 }, { "epoch": 0.97, "learning_rate": 4.230785832918693e-08, "loss": 0.3479, "step": 19458 }, { "epoch": 0.97, "learning_rate": 4.215938397394537e-08, "loss": 0.2789, "step": 19459 }, { "epoch": 0.97, "learning_rate": 4.2011170052982387e-08, "loss": 0.2885, "step": 19460 }, { "epoch": 0.97, "learning_rate": 4.1863216570176005e-08, "loss": 0.1878, "step": 19461 }, { "epoch": 0.97, "learning_rate": 4.171552352939312e-08, "loss": 0.4493, "step": 19462 }, { "epoch": 0.97, "learning_rate": 4.156809093449732e-08, "loss": 0.2966, "step": 19463 }, { "epoch": 0.97, "learning_rate": 4.142091878934551e-08, "loss": 0.4413, "step": 19464 }, { "epoch": 0.97, "learning_rate": 4.127400709778684e-08, "loss": 0.3223, "step": 19465 }, { "epoch": 0.97, "learning_rate": 4.1127355863661566e-08, "loss": 0.3366, "step": 19466 }, { "epoch": 0.97, "learning_rate": 4.0980965090807736e-08, "loss": 0.315, "step": 19467 }, { "epoch": 0.97, "learning_rate": 4.083483478305117e-08, "loss": 0.2078, "step": 19468 }, { "epoch": 0.97, "learning_rate": 4.0688964944216594e-08, "loss": 0.2774, "step": 19469 }, { "epoch": 0.97, "learning_rate": 4.0543355578117614e-08, "loss": 0.474, "step": 19470 }, { "epoch": 0.97, "learning_rate": 4.039800668856342e-08, "loss": 0.4285, "step": 19471 }, { "epoch": 0.97, "learning_rate": 4.0252918279353183e-08, "loss": 0.3924, "step": 19472 }, { "epoch": 0.97, "learning_rate": 4.010809035428387e-08, "loss": 0.2316, "step": 19473 }, { "epoch": 0.97, "learning_rate": 3.996352291714134e-08, "loss": 0.1959, "step": 19474 }, { "epoch": 0.97, "learning_rate": 3.9819215971709235e-08, "loss": 0.4597, "step": 19475 }, { "epoch": 0.97, "learning_rate": 3.967516952175898e-08, "loss": 0.3189, "step": 19476 }, { "epoch": 0.97, "learning_rate": 3.953138357105868e-08, "loss": 0.2985, "step": 19477 }, { "epoch": 0.97, "learning_rate": 3.938785812336976e-08, "loss": 0.4657, "step": 19478 }, { "epoch": 0.97, "learning_rate": 3.924459318244589e-08, "loss": 0.2357, "step": 19479 }, { "epoch": 0.97, "learning_rate": 3.910158875203185e-08, "loss": 0.3584, "step": 19480 }, { "epoch": 0.97, "learning_rate": 3.89588448358702e-08, "loss": 0.547, "step": 19481 }, { "epoch": 0.97, "learning_rate": 3.8816361437693516e-08, "loss": 0.4834, "step": 19482 }, { "epoch": 0.97, "learning_rate": 3.86741385612277e-08, "loss": 0.3281, "step": 19483 }, { "epoch": 0.97, "learning_rate": 3.853217621019312e-08, "loss": 0.4568, "step": 19484 }, { "epoch": 0.97, "learning_rate": 3.8390474388302343e-08, "loss": 0.3728, "step": 19485 }, { "epoch": 0.97, "learning_rate": 3.824903309926131e-08, "loss": 0.2085, "step": 19486 }, { "epoch": 0.97, "learning_rate": 3.810785234676928e-08, "loss": 0.2414, "step": 19487 }, { "epoch": 0.97, "learning_rate": 3.796693213451774e-08, "loss": 0.2583, "step": 19488 }, { "epoch": 0.97, "learning_rate": 3.7826272466194855e-08, "loss": 0.4451, "step": 19489 }, { "epoch": 0.97, "learning_rate": 3.7685873345476575e-08, "loss": 0.4163, "step": 19490 }, { "epoch": 0.97, "learning_rate": 3.754573477603551e-08, "loss": 0.371, "step": 19491 }, { "epoch": 0.97, "learning_rate": 3.740585676153874e-08, "loss": 0.4033, "step": 19492 }, { "epoch": 0.97, "learning_rate": 3.7266239305642215e-08, "loss": 0.4589, "step": 19493 }, { "epoch": 0.97, "learning_rate": 3.712688241199858e-08, "loss": 0.2766, "step": 19494 }, { "epoch": 0.97, "learning_rate": 3.698778608425157e-08, "loss": 0.2357, "step": 19495 }, { "epoch": 0.97, "learning_rate": 3.684895032604052e-08, "loss": 0.4578, "step": 19496 }, { "epoch": 0.97, "learning_rate": 3.6710375140995847e-08, "loss": 0.5526, "step": 19497 }, { "epoch": 0.97, "learning_rate": 3.657206053274243e-08, "loss": 0.3447, "step": 19498 }, { "epoch": 0.97, "learning_rate": 3.643400650489626e-08, "loss": 0.3923, "step": 19499 }, { "epoch": 0.97, "learning_rate": 3.629621306106779e-08, "loss": 0.2215, "step": 19500 }, { "epoch": 0.97, "learning_rate": 3.6158680204863015e-08, "loss": 0.3413, "step": 19501 }, { "epoch": 0.97, "learning_rate": 3.6021407939876853e-08, "loss": 0.5107, "step": 19502 }, { "epoch": 0.97, "learning_rate": 3.588439626970086e-08, "loss": 0.2747, "step": 19503 }, { "epoch": 0.97, "learning_rate": 3.574764519791884e-08, "loss": 0.3382, "step": 19504 }, { "epoch": 0.97, "learning_rate": 3.5611154728104614e-08, "loss": 0.4535, "step": 19505 }, { "epoch": 0.97, "learning_rate": 3.547492486383086e-08, "loss": 0.2294, "step": 19506 }, { "epoch": 0.97, "learning_rate": 3.5338955608658075e-08, "loss": 0.3311, "step": 19507 }, { "epoch": 0.97, "learning_rate": 3.520324696614452e-08, "loss": 0.3401, "step": 19508 }, { "epoch": 0.97, "learning_rate": 3.506779893983847e-08, "loss": 0.2622, "step": 19509 }, { "epoch": 0.97, "learning_rate": 3.493261153328153e-08, "loss": 0.3902, "step": 19510 }, { "epoch": 0.97, "learning_rate": 3.4797684750010885e-08, "loss": 0.4596, "step": 19511 }, { "epoch": 0.97, "learning_rate": 3.466301859355481e-08, "loss": 0.4335, "step": 19512 }, { "epoch": 0.97, "learning_rate": 3.4528613067434937e-08, "loss": 0.2214, "step": 19513 }, { "epoch": 0.97, "learning_rate": 3.439446817516734e-08, "loss": 0.4757, "step": 19514 }, { "epoch": 0.97, "learning_rate": 3.426058392025922e-08, "loss": 0.3567, "step": 19515 }, { "epoch": 0.97, "learning_rate": 3.412696030621221e-08, "loss": 0.3774, "step": 19516 }, { "epoch": 0.97, "learning_rate": 3.399359733652241e-08, "loss": 0.4751, "step": 19517 }, { "epoch": 0.97, "learning_rate": 3.386049501467703e-08, "loss": 0.2337, "step": 19518 }, { "epoch": 0.97, "learning_rate": 3.372765334415662e-08, "loss": 0.417, "step": 19519 }, { "epoch": 0.97, "learning_rate": 3.3595072328437284e-08, "loss": 0.2395, "step": 19520 }, { "epoch": 0.97, "learning_rate": 3.346275197098403e-08, "loss": 0.5393, "step": 19521 }, { "epoch": 0.97, "learning_rate": 3.3330692275258534e-08, "loss": 0.3111, "step": 19522 }, { "epoch": 0.97, "learning_rate": 3.31988932447147e-08, "loss": 0.2471, "step": 19523 }, { "epoch": 0.97, "learning_rate": 3.3067354882799775e-08, "loss": 0.3863, "step": 19524 }, { "epoch": 0.97, "learning_rate": 3.293607719295433e-08, "loss": 0.3628, "step": 19525 }, { "epoch": 0.97, "learning_rate": 3.280506017861118e-08, "loss": 0.2278, "step": 19526 }, { "epoch": 0.98, "learning_rate": 3.267430384319759e-08, "loss": 0.4621, "step": 19527 }, { "epoch": 0.98, "learning_rate": 3.254380819013192e-08, "loss": 0.2501, "step": 19528 }, { "epoch": 0.98, "learning_rate": 3.241357322282812e-08, "loss": 0.2924, "step": 19529 }, { "epoch": 0.98, "learning_rate": 3.228359894469235e-08, "loss": 0.4496, "step": 19530 }, { "epoch": 0.98, "learning_rate": 3.2153885359123e-08, "loss": 0.2458, "step": 19531 }, { "epoch": 0.98, "learning_rate": 3.2024432469512925e-08, "loss": 0.2426, "step": 19532 }, { "epoch": 0.98, "learning_rate": 3.18952402792494e-08, "loss": 0.2875, "step": 19533 }, { "epoch": 0.98, "learning_rate": 3.176630879170972e-08, "loss": 0.3778, "step": 19534 }, { "epoch": 0.98, "learning_rate": 3.163763801026565e-08, "loss": 0.3249, "step": 19535 }, { "epoch": 0.98, "learning_rate": 3.1509227938283374e-08, "loss": 0.4025, "step": 19536 }, { "epoch": 0.98, "learning_rate": 3.1381078579120205e-08, "loss": 0.4435, "step": 19537 }, { "epoch": 0.98, "learning_rate": 3.125318993612903e-08, "loss": 0.5329, "step": 19538 }, { "epoch": 0.98, "learning_rate": 3.112556201265382e-08, "loss": 0.3011, "step": 19539 }, { "epoch": 0.98, "learning_rate": 3.0998194812031924e-08, "loss": 0.4011, "step": 19540 }, { "epoch": 0.98, "learning_rate": 3.0871088337596225e-08, "loss": 0.2095, "step": 19541 }, { "epoch": 0.98, "learning_rate": 3.074424259266961e-08, "loss": 0.6699, "step": 19542 }, { "epoch": 0.98, "learning_rate": 3.061765758057056e-08, "loss": 0.1915, "step": 19543 }, { "epoch": 0.98, "learning_rate": 3.049133330460863e-08, "loss": 0.229, "step": 19544 }, { "epoch": 0.98, "learning_rate": 3.0365269768087846e-08, "loss": 0.5338, "step": 19545 }, { "epoch": 0.98, "learning_rate": 3.023946697430558e-08, "loss": 0.3251, "step": 19546 }, { "epoch": 0.98, "learning_rate": 3.0113924926552516e-08, "loss": 0.3683, "step": 19547 }, { "epoch": 0.98, "learning_rate": 2.998864362811049e-08, "loss": 0.2258, "step": 19548 }, { "epoch": 0.98, "learning_rate": 2.986362308225799e-08, "loss": 0.8301, "step": 19549 }, { "epoch": 0.98, "learning_rate": 2.9738863292264608e-08, "loss": 0.3676, "step": 19550 }, { "epoch": 0.98, "learning_rate": 2.9614364261391082e-08, "loss": 0.3896, "step": 19551 }, { "epoch": 0.98, "learning_rate": 2.9490125992895913e-08, "loss": 0.2231, "step": 19552 }, { "epoch": 0.98, "learning_rate": 2.9366148490027612e-08, "loss": 0.3797, "step": 19553 }, { "epoch": 0.98, "learning_rate": 2.9242431756028035e-08, "loss": 0.3555, "step": 19554 }, { "epoch": 0.98, "learning_rate": 2.9118975794133476e-08, "loss": 0.1943, "step": 19555 }, { "epoch": 0.98, "learning_rate": 2.8995780607572465e-08, "loss": 0.3128, "step": 19556 }, { "epoch": 0.98, "learning_rate": 2.8872846199566872e-08, "loss": 0.6774, "step": 19557 }, { "epoch": 0.98, "learning_rate": 2.8750172573333012e-08, "loss": 0.266, "step": 19558 }, { "epoch": 0.98, "learning_rate": 2.8627759732077208e-08, "loss": 0.1595, "step": 19559 }, { "epoch": 0.98, "learning_rate": 2.8505607679003565e-08, "loss": 0.3734, "step": 19560 }, { "epoch": 0.98, "learning_rate": 2.8383716417303973e-08, "loss": 0.3014, "step": 19561 }, { "epoch": 0.98, "learning_rate": 2.8262085950169217e-08, "loss": 0.2287, "step": 19562 }, { "epoch": 0.98, "learning_rate": 2.814071628077786e-08, "loss": 0.4636, "step": 19563 }, { "epoch": 0.98, "learning_rate": 2.801960741230625e-08, "loss": 0.4565, "step": 19564 }, { "epoch": 0.98, "learning_rate": 2.7898759347919637e-08, "loss": 0.6001, "step": 19565 }, { "epoch": 0.98, "learning_rate": 2.777817209078104e-08, "loss": 0.2083, "step": 19566 }, { "epoch": 0.98, "learning_rate": 2.765784564404239e-08, "loss": 0.4275, "step": 19567 }, { "epoch": 0.98, "learning_rate": 2.7537780010852276e-08, "loss": 0.287, "step": 19568 }, { "epoch": 0.98, "learning_rate": 2.74179751943493e-08, "loss": 0.4554, "step": 19569 }, { "epoch": 0.98, "learning_rate": 2.7298431197667618e-08, "loss": 0.2568, "step": 19570 }, { "epoch": 0.98, "learning_rate": 2.717914802393251e-08, "loss": 0.3335, "step": 19571 }, { "epoch": 0.98, "learning_rate": 2.7060125676265926e-08, "loss": 0.4845, "step": 19572 }, { "epoch": 0.98, "learning_rate": 2.6941364157778703e-08, "loss": 0.2433, "step": 19573 }, { "epoch": 0.98, "learning_rate": 2.6822863471578365e-08, "loss": 0.4019, "step": 19574 }, { "epoch": 0.98, "learning_rate": 2.6704623620763536e-08, "loss": 0.1449, "step": 19575 }, { "epoch": 0.98, "learning_rate": 2.6586644608427304e-08, "loss": 0.2928, "step": 19576 }, { "epoch": 0.98, "learning_rate": 2.6468926437653864e-08, "loss": 0.345, "step": 19577 }, { "epoch": 0.98, "learning_rate": 2.6351469111521866e-08, "loss": 0.4425, "step": 19578 }, { "epoch": 0.98, "learning_rate": 2.6234272633105518e-08, "loss": 0.1963, "step": 19579 }, { "epoch": 0.98, "learning_rate": 2.6117337005467925e-08, "loss": 0.2293, "step": 19580 }, { "epoch": 0.98, "learning_rate": 2.600066223166775e-08, "loss": 0.4795, "step": 19581 }, { "epoch": 0.98, "learning_rate": 2.588424831475589e-08, "loss": 0.5023, "step": 19582 }, { "epoch": 0.98, "learning_rate": 2.576809525777879e-08, "loss": 0.409, "step": 19583 }, { "epoch": 0.98, "learning_rate": 2.5652203063772918e-08, "loss": 0.3731, "step": 19584 }, { "epoch": 0.98, "learning_rate": 2.5536571735770287e-08, "loss": 0.2084, "step": 19585 }, { "epoch": 0.98, "learning_rate": 2.542120127679404e-08, "loss": 0.1851, "step": 19586 }, { "epoch": 0.98, "learning_rate": 2.5306091689860645e-08, "loss": 0.365, "step": 19587 }, { "epoch": 0.98, "learning_rate": 2.5191242977983253e-08, "loss": 0.3982, "step": 19588 }, { "epoch": 0.98, "learning_rate": 2.507665514416391e-08, "loss": 0.2958, "step": 19589 }, { "epoch": 0.98, "learning_rate": 2.4962328191400208e-08, "loss": 0.4907, "step": 19590 }, { "epoch": 0.98, "learning_rate": 2.4848262122681988e-08, "loss": 0.2306, "step": 19591 }, { "epoch": 0.98, "learning_rate": 2.4734456940991303e-08, "loss": 0.2123, "step": 19592 }, { "epoch": 0.98, "learning_rate": 2.4620912649306884e-08, "loss": 0.1012, "step": 19593 }, { "epoch": 0.98, "learning_rate": 2.4507629250596353e-08, "loss": 0.3095, "step": 19594 }, { "epoch": 0.98, "learning_rate": 2.4394606747824013e-08, "loss": 0.2433, "step": 19595 }, { "epoch": 0.98, "learning_rate": 2.428184514394416e-08, "loss": 0.3179, "step": 19596 }, { "epoch": 0.98, "learning_rate": 2.416934444190666e-08, "loss": 0.2592, "step": 19597 }, { "epoch": 0.98, "learning_rate": 2.4057104644654715e-08, "loss": 0.4841, "step": 19598 }, { "epoch": 0.98, "learning_rate": 2.3945125755122643e-08, "loss": 0.4128, "step": 19599 }, { "epoch": 0.98, "learning_rate": 2.3833407776240326e-08, "loss": 0.5573, "step": 19600 }, { "epoch": 0.98, "learning_rate": 2.3721950710927643e-08, "loss": 0.441, "step": 19601 }, { "epoch": 0.98, "learning_rate": 2.361075456210116e-08, "loss": 0.178, "step": 19602 }, { "epoch": 0.98, "learning_rate": 2.3499819332668538e-08, "loss": 0.3818, "step": 19603 }, { "epoch": 0.98, "learning_rate": 2.338914502553191e-08, "loss": 0.4496, "step": 19604 }, { "epoch": 0.98, "learning_rate": 2.3278731643584517e-08, "loss": 0.265, "step": 19605 }, { "epoch": 0.98, "learning_rate": 2.3168579189715156e-08, "loss": 0.2684, "step": 19606 }, { "epoch": 0.98, "learning_rate": 2.3058687666803746e-08, "loss": 0.3061, "step": 19607 }, { "epoch": 0.98, "learning_rate": 2.2949057077725766e-08, "loss": 0.3789, "step": 19608 }, { "epoch": 0.98, "learning_rate": 2.28396874253467e-08, "loss": 0.515, "step": 19609 }, { "epoch": 0.98, "learning_rate": 2.2730578712528706e-08, "loss": 0.3427, "step": 19610 }, { "epoch": 0.98, "learning_rate": 2.2621730942123942e-08, "loss": 0.1925, "step": 19611 }, { "epoch": 0.98, "learning_rate": 2.2513144116980134e-08, "loss": 0.3237, "step": 19612 }, { "epoch": 0.98, "learning_rate": 2.2404818239937232e-08, "loss": 0.2578, "step": 19613 }, { "epoch": 0.98, "learning_rate": 2.2296753313827413e-08, "loss": 0.3725, "step": 19614 }, { "epoch": 0.98, "learning_rate": 2.2188949341477307e-08, "loss": 0.3823, "step": 19615 }, { "epoch": 0.98, "learning_rate": 2.2081406325706877e-08, "loss": 0.0792, "step": 19616 }, { "epoch": 0.98, "learning_rate": 2.1974124269328322e-08, "loss": 0.4525, "step": 19617 }, { "epoch": 0.98, "learning_rate": 2.1867103175148285e-08, "loss": 0.4196, "step": 19618 }, { "epoch": 0.98, "learning_rate": 2.176034304596453e-08, "loss": 0.4571, "step": 19619 }, { "epoch": 0.98, "learning_rate": 2.165384388457037e-08, "loss": 0.4455, "step": 19620 }, { "epoch": 0.98, "learning_rate": 2.1547605693751362e-08, "loss": 0.4669, "step": 19621 }, { "epoch": 0.98, "learning_rate": 2.144162847628417e-08, "loss": 0.5703, "step": 19622 }, { "epoch": 0.98, "learning_rate": 2.1335912234942136e-08, "loss": 0.2583, "step": 19623 }, { "epoch": 0.98, "learning_rate": 2.123045697248971e-08, "loss": 0.3695, "step": 19624 }, { "epoch": 0.98, "learning_rate": 2.112526269168358e-08, "loss": 0.536, "step": 19625 }, { "epoch": 0.98, "learning_rate": 2.1020329395278206e-08, "loss": 0.5307, "step": 19626 }, { "epoch": 0.98, "learning_rate": 2.091565708601473e-08, "loss": 0.3919, "step": 19627 }, { "epoch": 0.98, "learning_rate": 2.0811245766632073e-08, "loss": 0.4091, "step": 19628 }, { "epoch": 0.98, "learning_rate": 2.070709543986027e-08, "loss": 0.2911, "step": 19629 }, { "epoch": 0.98, "learning_rate": 2.060320610842492e-08, "loss": 0.5196, "step": 19630 }, { "epoch": 0.98, "learning_rate": 2.0499577775041634e-08, "loss": 0.4406, "step": 19631 }, { "epoch": 0.98, "learning_rate": 2.039621044242157e-08, "loss": 0.4838, "step": 19632 }, { "epoch": 0.98, "learning_rate": 2.0293104113267014e-08, "loss": 0.3287, "step": 19633 }, { "epoch": 0.98, "learning_rate": 2.0190258790275808e-08, "loss": 0.0834, "step": 19634 }, { "epoch": 0.98, "learning_rate": 2.008767447613691e-08, "loss": 0.4669, "step": 19635 }, { "epoch": 0.98, "learning_rate": 1.998535117353373e-08, "loss": 0.1689, "step": 19636 }, { "epoch": 0.98, "learning_rate": 1.9883288885141905e-08, "loss": 0.2583, "step": 19637 }, { "epoch": 0.98, "learning_rate": 1.9781487613631523e-08, "loss": 0.3961, "step": 19638 }, { "epoch": 0.98, "learning_rate": 1.9679947361663786e-08, "loss": 0.3719, "step": 19639 }, { "epoch": 0.98, "learning_rate": 1.9578668131896572e-08, "loss": 0.3666, "step": 19640 }, { "epoch": 0.98, "learning_rate": 1.9477649926975538e-08, "loss": 0.4159, "step": 19641 }, { "epoch": 0.98, "learning_rate": 1.937689274954524e-08, "loss": 0.3287, "step": 19642 }, { "epoch": 0.98, "learning_rate": 1.9276396602240234e-08, "loss": 0.3231, "step": 19643 }, { "epoch": 0.98, "learning_rate": 1.9176161487688415e-08, "loss": 0.1318, "step": 19644 }, { "epoch": 0.98, "learning_rate": 1.9076187408512137e-08, "loss": 0.2938, "step": 19645 }, { "epoch": 0.98, "learning_rate": 1.897647436732486e-08, "loss": 0.415, "step": 19646 }, { "epoch": 0.98, "learning_rate": 1.8877022366735608e-08, "loss": 0.3944, "step": 19647 }, { "epoch": 0.98, "learning_rate": 1.8777831409344527e-08, "loss": 0.4527, "step": 19648 }, { "epoch": 0.98, "learning_rate": 1.8678901497746206e-08, "loss": 0.3567, "step": 19649 }, { "epoch": 0.98, "learning_rate": 1.8580232634528573e-08, "loss": 0.3743, "step": 19650 }, { "epoch": 0.98, "learning_rate": 1.8481824822270677e-08, "loss": 0.2912, "step": 19651 }, { "epoch": 0.98, "learning_rate": 1.8383678063548238e-08, "loss": 0.3918, "step": 19652 }, { "epoch": 0.98, "learning_rate": 1.828579236092587e-08, "loss": 0.3383, "step": 19653 }, { "epoch": 0.98, "learning_rate": 1.8188167716965967e-08, "loss": 0.2136, "step": 19654 }, { "epoch": 0.98, "learning_rate": 1.809080413422093e-08, "loss": 0.3918, "step": 19655 }, { "epoch": 0.98, "learning_rate": 1.7993701615236504e-08, "loss": 0.2109, "step": 19656 }, { "epoch": 0.98, "learning_rate": 1.7896860162551767e-08, "loss": 0.57, "step": 19657 }, { "epoch": 0.98, "learning_rate": 1.7800279778701357e-08, "loss": 0.1833, "step": 19658 }, { "epoch": 0.98, "learning_rate": 1.7703960466211035e-08, "loss": 0.3452, "step": 19659 }, { "epoch": 0.98, "learning_rate": 1.7607902227597673e-08, "loss": 0.2689, "step": 19660 }, { "epoch": 0.98, "learning_rate": 1.7512105065375927e-08, "loss": 0.2026, "step": 19661 }, { "epoch": 0.98, "learning_rate": 1.7416568982050463e-08, "loss": 0.4733, "step": 19662 }, { "epoch": 0.98, "learning_rate": 1.732129398011928e-08, "loss": 0.4947, "step": 19663 }, { "epoch": 0.98, "learning_rate": 1.722628006207483e-08, "loss": 0.5114, "step": 19664 }, { "epoch": 0.98, "learning_rate": 1.71315272304029e-08, "loss": 0.376, "step": 19665 }, { "epoch": 0.98, "learning_rate": 1.703703548757929e-08, "loss": 0.319, "step": 19666 }, { "epoch": 0.98, "learning_rate": 1.6942804836077574e-08, "loss": 0.2185, "step": 19667 }, { "epoch": 0.98, "learning_rate": 1.6848835278361342e-08, "loss": 0.5258, "step": 19668 }, { "epoch": 0.98, "learning_rate": 1.6755126816887512e-08, "loss": 0.183, "step": 19669 }, { "epoch": 0.98, "learning_rate": 1.666167945410746e-08, "loss": 0.5526, "step": 19670 }, { "epoch": 0.98, "learning_rate": 1.6568493192467007e-08, "loss": 0.392, "step": 19671 }, { "epoch": 0.98, "learning_rate": 1.6475568034399758e-08, "loss": 0.2282, "step": 19672 }, { "epoch": 0.98, "learning_rate": 1.6382903982339328e-08, "loss": 0.4444, "step": 19673 }, { "epoch": 0.98, "learning_rate": 1.6290501038707106e-08, "loss": 0.404, "step": 19674 }, { "epoch": 0.98, "learning_rate": 1.6198359205921166e-08, "loss": 0.4174, "step": 19675 }, { "epoch": 0.98, "learning_rate": 1.6106478486390686e-08, "loss": 0.3662, "step": 19676 }, { "epoch": 0.98, "learning_rate": 1.601485888251819e-08, "loss": 0.4847, "step": 19677 }, { "epoch": 0.98, "learning_rate": 1.5923500396699543e-08, "loss": 0.3596, "step": 19678 }, { "epoch": 0.98, "learning_rate": 1.583240303132616e-08, "loss": 0.2332, "step": 19679 }, { "epoch": 0.98, "learning_rate": 1.574156678877947e-08, "loss": 0.4169, "step": 19680 }, { "epoch": 0.98, "learning_rate": 1.5650991671435357e-08, "loss": 0.3451, "step": 19681 }, { "epoch": 0.98, "learning_rate": 1.5560677681660808e-08, "loss": 0.3638, "step": 19682 }, { "epoch": 0.98, "learning_rate": 1.54706248218206e-08, "loss": 0.245, "step": 19683 }, { "epoch": 0.98, "learning_rate": 1.538083309426841e-08, "loss": 0.29, "step": 19684 }, { "epoch": 0.98, "learning_rate": 1.5291302501353466e-08, "loss": 0.2441, "step": 19685 }, { "epoch": 0.98, "learning_rate": 1.5202033045417232e-08, "loss": 0.2843, "step": 19686 }, { "epoch": 0.98, "learning_rate": 1.511302472879339e-08, "loss": 0.3544, "step": 19687 }, { "epoch": 0.98, "learning_rate": 1.5024277553810084e-08, "loss": 0.7413, "step": 19688 }, { "epoch": 0.98, "learning_rate": 1.4935791522789898e-08, "loss": 0.3646, "step": 19689 }, { "epoch": 0.98, "learning_rate": 1.484756663804543e-08, "loss": 0.3297, "step": 19690 }, { "epoch": 0.98, "learning_rate": 1.4759602901884829e-08, "loss": 0.4399, "step": 19691 }, { "epoch": 0.98, "learning_rate": 1.4671900316608478e-08, "loss": 0.4324, "step": 19692 }, { "epoch": 0.98, "learning_rate": 1.45844588845101e-08, "loss": 0.3272, "step": 19693 }, { "epoch": 0.98, "learning_rate": 1.4497278607876753e-08, "loss": 0.0741, "step": 19694 }, { "epoch": 0.98, "learning_rate": 1.4410359488988835e-08, "loss": 0.3717, "step": 19695 }, { "epoch": 0.98, "learning_rate": 1.4323701530117862e-08, "loss": 0.3564, "step": 19696 }, { "epoch": 0.98, "learning_rate": 1.4237304733533131e-08, "loss": 0.4785, "step": 19697 }, { "epoch": 0.98, "learning_rate": 1.4151169101491724e-08, "loss": 0.2582, "step": 19698 }, { "epoch": 0.98, "learning_rate": 1.4065294636247395e-08, "loss": 0.5004, "step": 19699 }, { "epoch": 0.98, "learning_rate": 1.3979681340047236e-08, "loss": 0.2403, "step": 19700 }, { "epoch": 0.98, "learning_rate": 1.3894329215127233e-08, "loss": 0.188, "step": 19701 }, { "epoch": 0.98, "learning_rate": 1.3809238263723378e-08, "loss": 0.1619, "step": 19702 }, { "epoch": 0.98, "learning_rate": 1.3724408488058338e-08, "loss": 0.3518, "step": 19703 }, { "epoch": 0.98, "learning_rate": 1.3639839890352558e-08, "loss": 0.3721, "step": 19704 }, { "epoch": 0.98, "learning_rate": 1.355553247281649e-08, "loss": 0.5166, "step": 19705 }, { "epoch": 0.98, "learning_rate": 1.347148623765504e-08, "loss": 0.328, "step": 19706 }, { "epoch": 0.98, "learning_rate": 1.3387701187067558e-08, "loss": 0.5421, "step": 19707 }, { "epoch": 0.98, "learning_rate": 1.3304177323243406e-08, "loss": 0.527, "step": 19708 }, { "epoch": 0.98, "learning_rate": 1.3220914648369721e-08, "loss": 0.1557, "step": 19709 }, { "epoch": 0.98, "learning_rate": 1.313791316462254e-08, "loss": 0.335, "step": 19710 }, { "epoch": 0.98, "learning_rate": 1.305517287417235e-08, "loss": 0.4912, "step": 19711 }, { "epoch": 0.98, "learning_rate": 1.2972693779182977e-08, "loss": 0.2543, "step": 19712 }, { "epoch": 0.98, "learning_rate": 1.2890475881813802e-08, "loss": 0.4461, "step": 19713 }, { "epoch": 0.98, "learning_rate": 1.2808519184211999e-08, "loss": 0.3857, "step": 19714 }, { "epoch": 0.98, "learning_rate": 1.2726823688523626e-08, "loss": 0.2426, "step": 19715 }, { "epoch": 0.98, "learning_rate": 1.2645389396883645e-08, "loss": 0.4034, "step": 19716 }, { "epoch": 0.98, "learning_rate": 1.2564216311421462e-08, "loss": 0.3381, "step": 19717 }, { "epoch": 0.98, "learning_rate": 1.2483304434262045e-08, "loss": 0.3243, "step": 19718 }, { "epoch": 0.98, "learning_rate": 1.2402653767520367e-08, "loss": 0.2024, "step": 19719 }, { "epoch": 0.98, "learning_rate": 1.2322264313305855e-08, "loss": 0.3485, "step": 19720 }, { "epoch": 0.98, "learning_rate": 1.2242136073721266e-08, "loss": 0.4056, "step": 19721 }, { "epoch": 0.98, "learning_rate": 1.2162269050861597e-08, "loss": 0.4584, "step": 19722 }, { "epoch": 0.98, "learning_rate": 1.2082663246816283e-08, "loss": 0.4135, "step": 19723 }, { "epoch": 0.98, "learning_rate": 1.2003318663666996e-08, "loss": 0.4498, "step": 19724 }, { "epoch": 0.98, "learning_rate": 1.192423530348985e-08, "loss": 0.3961, "step": 19725 }, { "epoch": 0.98, "learning_rate": 1.1845413168350972e-08, "loss": 0.2559, "step": 19726 }, { "epoch": 0.99, "learning_rate": 1.1766852260314266e-08, "loss": 0.4997, "step": 19727 }, { "epoch": 0.99, "learning_rate": 1.1688552581432533e-08, "loss": 0.4033, "step": 19728 }, { "epoch": 0.99, "learning_rate": 1.1610514133754137e-08, "loss": 0.3382, "step": 19729 }, { "epoch": 0.99, "learning_rate": 1.1532736919320775e-08, "loss": 0.1661, "step": 19730 }, { "epoch": 0.99, "learning_rate": 1.1455220940166378e-08, "loss": 0.3406, "step": 19731 }, { "epoch": 0.99, "learning_rate": 1.1377966198318213e-08, "loss": 0.3124, "step": 19732 }, { "epoch": 0.99, "learning_rate": 1.1300972695795776e-08, "loss": 0.2754, "step": 19733 }, { "epoch": 0.99, "learning_rate": 1.122424043461412e-08, "loss": 0.2639, "step": 19734 }, { "epoch": 0.99, "learning_rate": 1.114776941677942e-08, "loss": 0.1881, "step": 19735 }, { "epoch": 0.99, "learning_rate": 1.1071559644291186e-08, "loss": 0.3741, "step": 19736 }, { "epoch": 0.99, "learning_rate": 1.099561111914449e-08, "loss": 0.3743, "step": 19737 }, { "epoch": 0.99, "learning_rate": 1.0919923843323298e-08, "loss": 0.5426, "step": 19738 }, { "epoch": 0.99, "learning_rate": 1.0844497818808253e-08, "loss": 0.5116, "step": 19739 }, { "epoch": 0.99, "learning_rate": 1.0769333047572216e-08, "loss": 0.182, "step": 19740 }, { "epoch": 0.99, "learning_rate": 1.0694429531580286e-08, "loss": 0.3047, "step": 19741 }, { "epoch": 0.99, "learning_rate": 1.0619787272792003e-08, "loss": 0.25, "step": 19742 }, { "epoch": 0.99, "learning_rate": 1.0545406273159142e-08, "loss": 0.3106, "step": 19743 }, { "epoch": 0.99, "learning_rate": 1.0471286534627923e-08, "loss": 0.465, "step": 19744 }, { "epoch": 0.99, "learning_rate": 1.0397428059136794e-08, "loss": 0.3066, "step": 19745 }, { "epoch": 0.99, "learning_rate": 1.0323830848616435e-08, "loss": 0.1865, "step": 19746 }, { "epoch": 0.99, "learning_rate": 1.025049490499197e-08, "loss": 0.3885, "step": 19747 }, { "epoch": 0.99, "learning_rate": 1.0177420230181867e-08, "loss": 0.2631, "step": 19748 }, { "epoch": 0.99, "learning_rate": 1.010460682609682e-08, "loss": 0.2074, "step": 19749 }, { "epoch": 0.99, "learning_rate": 1.0032054694640858e-08, "loss": 0.2879, "step": 19750 }, { "epoch": 0.99, "learning_rate": 9.959763837712466e-09, "loss": 0.3135, "step": 19751 }, { "epoch": 0.99, "learning_rate": 9.887734257202352e-09, "loss": 0.4212, "step": 19752 }, { "epoch": 0.99, "learning_rate": 9.815965954992346e-09, "loss": 0.5279, "step": 19753 }, { "epoch": 0.99, "learning_rate": 9.744458932962052e-09, "loss": 0.5012, "step": 19754 }, { "epoch": 0.99, "learning_rate": 9.67321319298109e-09, "loss": 0.2414, "step": 19755 }, { "epoch": 0.99, "learning_rate": 9.602228736911301e-09, "loss": 0.4357, "step": 19756 }, { "epoch": 0.99, "learning_rate": 9.531505566610089e-09, "loss": 0.0996, "step": 19757 }, { "epoch": 0.99, "learning_rate": 9.461043683928195e-09, "loss": 0.6647, "step": 19758 }, { "epoch": 0.99, "learning_rate": 9.39084309070637e-09, "loss": 0.2871, "step": 19759 }, { "epoch": 0.99, "learning_rate": 9.320903788782032e-09, "loss": 0.2752, "step": 19760 }, { "epoch": 0.99, "learning_rate": 9.251225779984829e-09, "loss": 0.2765, "step": 19761 }, { "epoch": 0.99, "learning_rate": 9.181809066135528e-09, "loss": 0.2991, "step": 19762 }, { "epoch": 0.99, "learning_rate": 9.112653649050451e-09, "loss": 0.3581, "step": 19763 }, { "epoch": 0.99, "learning_rate": 9.043759530538154e-09, "loss": 0.5552, "step": 19764 }, { "epoch": 0.99, "learning_rate": 8.975126712400528e-09, "loss": 0.4657, "step": 19765 }, { "epoch": 0.99, "learning_rate": 8.906755196432803e-09, "loss": 0.5206, "step": 19766 }, { "epoch": 0.99, "learning_rate": 8.83864498442244e-09, "loss": 0.31, "step": 19767 }, { "epoch": 0.99, "learning_rate": 8.770796078151345e-09, "loss": 0.2577, "step": 19768 }, { "epoch": 0.99, "learning_rate": 8.703208479394764e-09, "loss": 0.5007, "step": 19769 }, { "epoch": 0.99, "learning_rate": 8.635882189919065e-09, "loss": 0.4854, "step": 19770 }, { "epoch": 0.99, "learning_rate": 8.568817211485059e-09, "loss": 0.2119, "step": 19771 }, { "epoch": 0.99, "learning_rate": 8.502013545848009e-09, "loss": 0.3266, "step": 19772 }, { "epoch": 0.99, "learning_rate": 8.435471194753187e-09, "loss": 0.3189, "step": 19773 }, { "epoch": 0.99, "learning_rate": 8.369190159942531e-09, "loss": 0.5605, "step": 19774 }, { "epoch": 0.99, "learning_rate": 8.3031704431491e-09, "loss": 0.2604, "step": 19775 }, { "epoch": 0.99, "learning_rate": 8.237412046099292e-09, "loss": 0.3669, "step": 19776 }, { "epoch": 0.99, "learning_rate": 8.17191497051395e-09, "loss": 0.3398, "step": 19777 }, { "epoch": 0.99, "learning_rate": 8.10667921810393e-09, "loss": 0.2999, "step": 19778 }, { "epoch": 0.99, "learning_rate": 8.041704790577864e-09, "loss": 0.2577, "step": 19779 }, { "epoch": 0.99, "learning_rate": 7.976991689633284e-09, "loss": 0.5636, "step": 19780 }, { "epoch": 0.99, "learning_rate": 7.91253991696328e-09, "loss": 0.3802, "step": 19781 }, { "epoch": 0.99, "learning_rate": 7.84834947425317e-09, "loss": 0.1574, "step": 19782 }, { "epoch": 0.99, "learning_rate": 7.78442036318272e-09, "loss": 0.1448, "step": 19783 }, { "epoch": 0.99, "learning_rate": 7.72075258542282e-09, "loss": 0.3797, "step": 19784 }, { "epoch": 0.99, "learning_rate": 7.65734614263991e-09, "loss": 0.603, "step": 19785 }, { "epoch": 0.99, "learning_rate": 7.594201036491555e-09, "loss": 0.2334, "step": 19786 }, { "epoch": 0.99, "learning_rate": 7.531317268628657e-09, "loss": 0.2059, "step": 19787 }, { "epoch": 0.99, "learning_rate": 7.468694840695456e-09, "loss": 0.6443, "step": 19788 }, { "epoch": 0.99, "learning_rate": 7.406333754331751e-09, "loss": 0.2521, "step": 19789 }, { "epoch": 0.99, "learning_rate": 7.344234011167351e-09, "loss": 0.3382, "step": 19790 }, { "epoch": 0.99, "learning_rate": 7.28239561282651e-09, "loss": 0.3426, "step": 19791 }, { "epoch": 0.99, "learning_rate": 7.220818560925713e-09, "loss": 0.2251, "step": 19792 }, { "epoch": 0.99, "learning_rate": 7.159502857077005e-09, "loss": 0.2092, "step": 19793 }, { "epoch": 0.99, "learning_rate": 7.098448502883548e-09, "loss": 0.2619, "step": 19794 }, { "epoch": 0.99, "learning_rate": 7.037655499941843e-09, "loss": 0.3032, "step": 19795 }, { "epoch": 0.99, "learning_rate": 6.977123849840617e-09, "loss": 0.3576, "step": 19796 }, { "epoch": 0.99, "learning_rate": 6.91685355416527e-09, "loss": 0.3789, "step": 19797 }, { "epoch": 0.99, "learning_rate": 6.856844614491209e-09, "loss": 0.5212, "step": 19798 }, { "epoch": 0.99, "learning_rate": 6.797097032387179e-09, "loss": 0.3771, "step": 19799 }, { "epoch": 0.99, "learning_rate": 6.737610809416373e-09, "loss": 0.6184, "step": 19800 }, { "epoch": 0.99, "learning_rate": 6.678385947134214e-09, "loss": 0.4896, "step": 19801 }, { "epoch": 0.99, "learning_rate": 6.619422447090574e-09, "loss": 0.4761, "step": 19802 }, { "epoch": 0.99, "learning_rate": 6.56072031082644e-09, "loss": 0.5252, "step": 19803 }, { "epoch": 0.99, "learning_rate": 6.502279539877254e-09, "loss": 0.2292, "step": 19804 }, { "epoch": 0.99, "learning_rate": 6.4441001357729014e-09, "loss": 0.4293, "step": 19805 }, { "epoch": 0.99, "learning_rate": 6.386182100033278e-09, "loss": 0.4717, "step": 19806 }, { "epoch": 0.99, "learning_rate": 6.328525434173838e-09, "loss": 0.3384, "step": 19807 }, { "epoch": 0.99, "learning_rate": 6.271130139702264e-09, "loss": 0.4978, "step": 19808 }, { "epoch": 0.99, "learning_rate": 6.213996218120688e-09, "loss": 0.4786, "step": 19809 }, { "epoch": 0.99, "learning_rate": 6.1571236709212504e-09, "loss": 0.2536, "step": 19810 }, { "epoch": 0.99, "learning_rate": 6.100512499593869e-09, "loss": 0.1436, "step": 19811 }, { "epoch": 0.99, "learning_rate": 6.044162705617362e-09, "loss": 0.5221, "step": 19812 }, { "epoch": 0.99, "learning_rate": 5.988074290466106e-09, "loss": 0.3976, "step": 19813 }, { "epoch": 0.99, "learning_rate": 5.932247255606705e-09, "loss": 0.3037, "step": 19814 }, { "epoch": 0.99, "learning_rate": 5.876681602500212e-09, "loss": 0.1836, "step": 19815 }, { "epoch": 0.99, "learning_rate": 5.821377332598799e-09, "loss": 0.452, "step": 19816 }, { "epoch": 0.99, "learning_rate": 5.766334447349086e-09, "loss": 0.4322, "step": 19817 }, { "epoch": 0.99, "learning_rate": 5.7115529481910345e-09, "loss": 0.4239, "step": 19818 }, { "epoch": 0.99, "learning_rate": 5.657032836556831e-09, "loss": 0.1878, "step": 19819 }, { "epoch": 0.99, "learning_rate": 5.602774113873111e-09, "loss": 0.185, "step": 19820 }, { "epoch": 0.99, "learning_rate": 5.54877678155763e-09, "loss": 0.3033, "step": 19821 }, { "epoch": 0.99, "learning_rate": 5.495040841024813e-09, "loss": 0.2271, "step": 19822 }, { "epoch": 0.99, "learning_rate": 5.44156629367798e-09, "loss": 0.2139, "step": 19823 }, { "epoch": 0.99, "learning_rate": 5.388353140916014e-09, "loss": 0.2355, "step": 19824 }, { "epoch": 0.99, "learning_rate": 5.335401384132244e-09, "loss": 0.3058, "step": 19825 }, { "epoch": 0.99, "learning_rate": 5.282711024710008e-09, "loss": 0.4744, "step": 19826 }, { "epoch": 0.99, "learning_rate": 5.230282064027092e-09, "loss": 0.3094, "step": 19827 }, { "epoch": 0.99, "learning_rate": 5.178114503455733e-09, "loss": 0.2702, "step": 19828 }, { "epoch": 0.99, "learning_rate": 5.126208344360395e-09, "loss": 0.3593, "step": 19829 }, { "epoch": 0.99, "learning_rate": 5.074563588097769e-09, "loss": 0.1496, "step": 19830 }, { "epoch": 0.99, "learning_rate": 5.023180236018998e-09, "loss": 0.479, "step": 19831 }, { "epoch": 0.99, "learning_rate": 4.9720582894674515e-09, "loss": 0.3194, "step": 19832 }, { "epoch": 0.99, "learning_rate": 4.9211977497809485e-09, "loss": 0.6724, "step": 19833 }, { "epoch": 0.99, "learning_rate": 4.8705986182884245e-09, "loss": 0.4432, "step": 19834 }, { "epoch": 0.99, "learning_rate": 4.820260896315488e-09, "loss": 0.2966, "step": 19835 }, { "epoch": 0.99, "learning_rate": 4.7701845851766405e-09, "loss": 0.3765, "step": 19836 }, { "epoch": 0.99, "learning_rate": 4.720369686183057e-09, "loss": 0.3766, "step": 19837 }, { "epoch": 0.99, "learning_rate": 4.670816200635919e-09, "loss": 0.5453, "step": 19838 }, { "epoch": 0.99, "learning_rate": 4.6215241298319665e-09, "loss": 0.3711, "step": 19839 }, { "epoch": 0.99, "learning_rate": 4.572493475060169e-09, "loss": 0.5216, "step": 19840 }, { "epoch": 0.99, "learning_rate": 4.523724237603944e-09, "loss": 0.5128, "step": 19841 }, { "epoch": 0.99, "learning_rate": 4.475216418737827e-09, "loss": 0.3774, "step": 19842 }, { "epoch": 0.99, "learning_rate": 4.426970019730803e-09, "loss": 0.2888, "step": 19843 }, { "epoch": 0.99, "learning_rate": 4.378985041844086e-09, "loss": 0.2371, "step": 19844 }, { "epoch": 0.99, "learning_rate": 4.331261486333338e-09, "loss": 0.1044, "step": 19845 }, { "epoch": 0.99, "learning_rate": 4.283799354446449e-09, "loss": 0.5945, "step": 19846 }, { "epoch": 0.99, "learning_rate": 4.23659864742465e-09, "loss": 0.4638, "step": 19847 }, { "epoch": 0.99, "learning_rate": 4.1896593665025076e-09, "loss": 0.5117, "step": 19848 }, { "epoch": 0.99, "learning_rate": 4.142981512907929e-09, "loss": 0.4404, "step": 19849 }, { "epoch": 0.99, "learning_rate": 4.096565087861049e-09, "loss": 0.4917, "step": 19850 }, { "epoch": 0.99, "learning_rate": 4.050410092576451e-09, "loss": 0.1456, "step": 19851 }, { "epoch": 0.99, "learning_rate": 4.004516528260949e-09, "loss": 0.3191, "step": 19852 }, { "epoch": 0.99, "learning_rate": 3.9588843961146926e-09, "loss": 0.386, "step": 19853 }, { "epoch": 0.99, "learning_rate": 3.913513697331173e-09, "loss": 0.3288, "step": 19854 }, { "epoch": 0.99, "learning_rate": 3.868404433098327e-09, "loss": 0.2932, "step": 19855 }, { "epoch": 0.99, "learning_rate": 3.823556604594103e-09, "loss": 0.4786, "step": 19856 }, { "epoch": 0.99, "learning_rate": 3.7789702129920055e-09, "loss": 0.4867, "step": 19857 }, { "epoch": 0.99, "learning_rate": 3.73464525945777e-09, "loss": 0.5565, "step": 19858 }, { "epoch": 0.99, "learning_rate": 3.690581745152688e-09, "loss": 0.4311, "step": 19859 }, { "epoch": 0.99, "learning_rate": 3.646779671225842e-09, "loss": 0.4737, "step": 19860 }, { "epoch": 0.99, "learning_rate": 3.6032390388252016e-09, "loss": 0.1799, "step": 19861 }, { "epoch": 0.99, "learning_rate": 3.5599598490898558e-09, "loss": 0.2532, "step": 19862 }, { "epoch": 0.99, "learning_rate": 3.5169421031500116e-09, "loss": 0.3466, "step": 19863 }, { "epoch": 0.99, "learning_rate": 3.4741858021314355e-09, "loss": 0.3508, "step": 19864 }, { "epoch": 0.99, "learning_rate": 3.431690947152122e-09, "loss": 0.2287, "step": 19865 }, { "epoch": 0.99, "learning_rate": 3.389457539324514e-09, "loss": 0.4267, "step": 19866 }, { "epoch": 0.99, "learning_rate": 3.3474855797521744e-09, "loss": 0.4403, "step": 19867 }, { "epoch": 0.99, "learning_rate": 3.3057750695331127e-09, "loss": 0.4791, "step": 19868 }, { "epoch": 0.99, "learning_rate": 3.264326009757568e-09, "loss": 0.2864, "step": 19869 }, { "epoch": 0.99, "learning_rate": 3.223138401511339e-09, "loss": 0.243, "step": 19870 }, { "epoch": 0.99, "learning_rate": 3.182212245870231e-09, "loss": 0.339, "step": 19871 }, { "epoch": 0.99, "learning_rate": 3.1415475439044997e-09, "loss": 0.3195, "step": 19872 }, { "epoch": 0.99, "learning_rate": 3.1011442966777383e-09, "loss": 0.3744, "step": 19873 }, { "epoch": 0.99, "learning_rate": 3.0610025052468795e-09, "loss": 0.3254, "step": 19874 }, { "epoch": 0.99, "learning_rate": 3.021122170662194e-09, "loss": 0.3186, "step": 19875 }, { "epoch": 0.99, "learning_rate": 2.9815032939672915e-09, "loss": 0.5595, "step": 19876 }, { "epoch": 0.99, "learning_rate": 2.9421458761969003e-09, "loss": 0.274, "step": 19877 }, { "epoch": 0.99, "learning_rate": 2.903049918380196e-09, "loss": 0.5268, "step": 19878 }, { "epoch": 0.99, "learning_rate": 2.8642154215408056e-09, "loss": 0.3124, "step": 19879 }, { "epoch": 0.99, "learning_rate": 2.8256423866934723e-09, "loss": 0.3201, "step": 19880 }, { "epoch": 0.99, "learning_rate": 2.787330814848499e-09, "loss": 0.0777, "step": 19881 }, { "epoch": 0.99, "learning_rate": 2.7492807070061967e-09, "loss": 0.3163, "step": 19882 }, { "epoch": 0.99, "learning_rate": 2.7114920641624355e-09, "loss": 0.3631, "step": 19883 }, { "epoch": 0.99, "learning_rate": 2.673964887306424e-09, "loss": 0.4191, "step": 19884 }, { "epoch": 0.99, "learning_rate": 2.636699177417379e-09, "loss": 0.3529, "step": 19885 }, { "epoch": 0.99, "learning_rate": 2.599694935472297e-09, "loss": 0.3276, "step": 19886 }, { "epoch": 0.99, "learning_rate": 2.562952162437071e-09, "loss": 0.2982, "step": 19887 }, { "epoch": 0.99, "learning_rate": 2.526470859274266e-09, "loss": 0.42, "step": 19888 }, { "epoch": 0.99, "learning_rate": 2.4902510269364522e-09, "loss": 0.2169, "step": 19889 }, { "epoch": 0.99, "learning_rate": 2.4542926663728704e-09, "loss": 0.3324, "step": 19890 }, { "epoch": 0.99, "learning_rate": 2.4185957785216596e-09, "loss": 0.6271, "step": 19891 }, { "epoch": 0.99, "learning_rate": 2.3831603643187375e-09, "loss": 0.4052, "step": 19892 }, { "epoch": 0.99, "learning_rate": 2.3479864246889193e-09, "loss": 0.1875, "step": 19893 }, { "epoch": 0.99, "learning_rate": 2.313073960553691e-09, "loss": 0.4844, "step": 19894 }, { "epoch": 0.99, "learning_rate": 2.2784229728245454e-09, "loss": 0.2376, "step": 19895 }, { "epoch": 0.99, "learning_rate": 2.244033462409645e-09, "loss": 0.4934, "step": 19896 }, { "epoch": 0.99, "learning_rate": 2.20990543020605e-09, "loss": 0.5505, "step": 19897 }, { "epoch": 0.99, "learning_rate": 2.1760388771085995e-09, "loss": 0.1625, "step": 19898 }, { "epoch": 0.99, "learning_rate": 2.1424338040010317e-09, "loss": 0.3402, "step": 19899 }, { "epoch": 0.99, "learning_rate": 2.1090902117648636e-09, "loss": 0.181, "step": 19900 }, { "epoch": 0.99, "learning_rate": 2.0760081012694e-09, "loss": 0.6981, "step": 19901 }, { "epoch": 0.99, "learning_rate": 2.0431874733806143e-09, "loss": 0.2037, "step": 19902 }, { "epoch": 0.99, "learning_rate": 2.0106283289578198e-09, "loss": 0.3762, "step": 19903 }, { "epoch": 0.99, "learning_rate": 1.9783306688525574e-09, "loss": 0.4573, "step": 19904 }, { "epoch": 0.99, "learning_rate": 1.9462944939085957e-09, "loss": 0.5816, "step": 19905 }, { "epoch": 0.99, "learning_rate": 1.9145198049641545e-09, "loss": 0.3871, "step": 19906 }, { "epoch": 0.99, "learning_rate": 1.8830066028507897e-09, "loss": 0.4419, "step": 19907 }, { "epoch": 0.99, "learning_rate": 1.8517548883911774e-09, "loss": 0.2569, "step": 19908 }, { "epoch": 0.99, "learning_rate": 1.8207646624046615e-09, "loss": 0.4659, "step": 19909 }, { "epoch": 0.99, "learning_rate": 1.790035925700595e-09, "loss": 0.2599, "step": 19910 }, { "epoch": 0.99, "learning_rate": 1.7595686790816691e-09, "loss": 0.5048, "step": 19911 }, { "epoch": 0.99, "learning_rate": 1.7293629233472441e-09, "loss": 0.3768, "step": 19912 }, { "epoch": 0.99, "learning_rate": 1.6994186592855787e-09, "loss": 0.1637, "step": 19913 }, { "epoch": 0.99, "learning_rate": 1.6697358876804904e-09, "loss": 0.449, "step": 19914 }, { "epoch": 0.99, "learning_rate": 1.6403146093069144e-09, "loss": 0.4051, "step": 19915 }, { "epoch": 0.99, "learning_rate": 1.611154824936456e-09, "loss": 0.2796, "step": 19916 }, { "epoch": 0.99, "learning_rate": 1.5822565353307284e-09, "loss": 0.3288, "step": 19917 }, { "epoch": 0.99, "learning_rate": 1.553619741244683e-09, "loss": 0.7536, "step": 19918 }, { "epoch": 0.99, "learning_rate": 1.5252444434288304e-09, "loss": 0.2371, "step": 19919 }, { "epoch": 0.99, "learning_rate": 1.4971306426248e-09, "loss": 0.3935, "step": 19920 }, { "epoch": 0.99, "learning_rate": 1.469278339566449e-09, "loss": 0.2361, "step": 19921 }, { "epoch": 0.99, "learning_rate": 1.441687534984304e-09, "loss": 0.3642, "step": 19922 }, { "epoch": 0.99, "learning_rate": 1.4143582295989e-09, "loss": 0.3304, "step": 19923 }, { "epoch": 0.99, "learning_rate": 1.3872904241252205e-09, "loss": 0.4138, "step": 19924 }, { "epoch": 0.99, "learning_rate": 1.3604841192715878e-09, "loss": 0.2323, "step": 19925 }, { "epoch": 0.99, "learning_rate": 1.3339393157374425e-09, "loss": 0.4931, "step": 19926 }, { "epoch": 1.0, "learning_rate": 1.3076560142188943e-09, "loss": 0.3064, "step": 19927 }, { "epoch": 1.0, "learning_rate": 1.2816342154031713e-09, "loss": 0.3524, "step": 19928 }, { "epoch": 1.0, "learning_rate": 1.2558739199697302e-09, "loss": 0.2773, "step": 19929 }, { "epoch": 1.0, "learning_rate": 1.230375128592476e-09, "loss": 0.1674, "step": 19930 }, { "epoch": 1.0, "learning_rate": 1.205137841939763e-09, "loss": 0.4823, "step": 19931 }, { "epoch": 1.0, "learning_rate": 1.1801620606699537e-09, "loss": 0.3939, "step": 19932 }, { "epoch": 1.0, "learning_rate": 1.1554477854369695e-09, "loss": 0.4218, "step": 19933 }, { "epoch": 1.0, "learning_rate": 1.1309950168869598e-09, "loss": 0.351, "step": 19934 }, { "epoch": 1.0, "learning_rate": 1.1068037556605238e-09, "loss": 0.298, "step": 19935 }, { "epoch": 1.0, "learning_rate": 1.0828740023882677e-09, "loss": 0.3341, "step": 19936 }, { "epoch": 1.0, "learning_rate": 1.059205757697468e-09, "loss": 0.2974, "step": 19937 }, { "epoch": 1.0, "learning_rate": 1.0357990222076287e-09, "loss": 0.2967, "step": 19938 }, { "epoch": 1.0, "learning_rate": 1.0126537965293726e-09, "loss": 0.4562, "step": 19939 }, { "epoch": 1.0, "learning_rate": 9.897700812688815e-10, "loss": 0.4178, "step": 19940 }, { "epoch": 1.0, "learning_rate": 9.671478770256758e-10, "loss": 0.5147, "step": 19941 }, { "epoch": 1.0, "learning_rate": 9.447871843892843e-10, "loss": 0.4705, "step": 19942 }, { "epoch": 1.0, "learning_rate": 9.22688003945904e-10, "loss": 0.2307, "step": 19943 }, { "epoch": 1.0, "learning_rate": 9.008503362728516e-10, "loss": 0.3384, "step": 19944 }, { "epoch": 1.0, "learning_rate": 8.792741819418915e-10, "loss": 0.3876, "step": 19945 }, { "epoch": 1.0, "learning_rate": 8.579595415170172e-10, "loss": 0.4209, "step": 19946 }, { "epoch": 1.0, "learning_rate": 8.369064155555607e-10, "loss": 0.4278, "step": 19947 }, { "epoch": 1.0, "learning_rate": 8.161148046081923e-10, "loss": 0.386, "step": 19948 }, { "epoch": 1.0, "learning_rate": 7.955847092189217e-10, "loss": 0.4253, "step": 19949 }, { "epoch": 1.0, "learning_rate": 7.753161299239864e-10, "loss": 0.398, "step": 19950 }, { "epoch": 1.0, "learning_rate": 7.553090672551833e-10, "loss": 0.4073, "step": 19951 }, { "epoch": 1.0, "learning_rate": 7.355635217343171e-10, "loss": 0.3662, "step": 19952 }, { "epoch": 1.0, "learning_rate": 7.160794938776416e-10, "loss": 0.2463, "step": 19953 }, { "epoch": 1.0, "learning_rate": 6.968569841947492e-10, "loss": 0.1756, "step": 19954 }, { "epoch": 1.0, "learning_rate": 6.778959931896812e-10, "loss": 0.2995, "step": 19955 }, { "epoch": 1.0, "learning_rate": 6.591965213575968e-10, "loss": 0.3941, "step": 19956 }, { "epoch": 1.0, "learning_rate": 6.407585691869944e-10, "loss": 0.4507, "step": 19957 }, { "epoch": 1.0, "learning_rate": 6.225821371597107e-10, "loss": 0.1382, "step": 19958 }, { "epoch": 1.0, "learning_rate": 6.046672257531416e-10, "loss": 0.2422, "step": 19959 }, { "epoch": 1.0, "learning_rate": 5.87013835434691e-10, "loss": 0.3258, "step": 19960 }, { "epoch": 1.0, "learning_rate": 5.696219666651015e-10, "loss": 0.3627, "step": 19961 }, { "epoch": 1.0, "learning_rate": 5.524916199006746e-10, "loss": 0.3519, "step": 19962 }, { "epoch": 1.0, "learning_rate": 5.356227955888305e-10, "loss": 0.4591, "step": 19963 }, { "epoch": 1.0, "learning_rate": 5.190154941714376e-10, "loss": 0.3476, "step": 19964 }, { "epoch": 1.0, "learning_rate": 5.026697160814831e-10, "loss": 0.4341, "step": 19965 }, { "epoch": 1.0, "learning_rate": 4.865854617475129e-10, "loss": 0.3684, "step": 19966 }, { "epoch": 1.0, "learning_rate": 4.707627315903019e-10, "loss": 0.4239, "step": 19967 }, { "epoch": 1.0, "learning_rate": 4.5520152602285263e-10, "loss": 0.3072, "step": 19968 }, { "epoch": 1.0, "learning_rate": 4.3990184545261715e-10, "loss": 0.3448, "step": 19969 }, { "epoch": 1.0, "learning_rate": 4.248636902792758e-10, "loss": 0.4346, "step": 19970 }, { "epoch": 1.0, "learning_rate": 4.100870608969576e-10, "loss": 0.3461, "step": 19971 }, { "epoch": 1.0, "learning_rate": 3.9557195769091007e-10, "loss": 0.3277, "step": 19972 }, { "epoch": 1.0, "learning_rate": 3.8131838104304984e-10, "loss": 0.2416, "step": 19973 }, { "epoch": 1.0, "learning_rate": 3.673263313230813e-10, "loss": 0.2419, "step": 19974 }, { "epoch": 1.0, "learning_rate": 3.535958088995983e-10, "loss": 0.4528, "step": 19975 }, { "epoch": 1.0, "learning_rate": 3.4012681413009284e-10, "loss": 0.4159, "step": 19976 }, { "epoch": 1.0, "learning_rate": 3.2691934736761577e-10, "loss": 0.2704, "step": 19977 }, { "epoch": 1.0, "learning_rate": 3.139734089574464e-10, "loss": 0.4188, "step": 19978 }, { "epoch": 1.0, "learning_rate": 3.012889992382029e-10, "loss": 0.4005, "step": 19979 }, { "epoch": 1.0, "learning_rate": 2.8886611854073154e-10, "loss": 0.3669, "step": 19980 }, { "epoch": 1.0, "learning_rate": 2.7670476719143803e-10, "loss": 0.4301, "step": 19981 }, { "epoch": 1.0, "learning_rate": 2.648049455078461e-10, "loss": 0.5729, "step": 19982 }, { "epoch": 1.0, "learning_rate": 2.5316665379970794e-10, "loss": 0.4794, "step": 19983 }, { "epoch": 1.0, "learning_rate": 2.417898923734452e-10, "loss": 0.2533, "step": 19984 }, { "epoch": 1.0, "learning_rate": 2.3067466152659756e-10, "loss": 0.4217, "step": 19985 }, { "epoch": 1.0, "learning_rate": 2.1982096154782306e-10, "loss": 0.4128, "step": 19986 }, { "epoch": 1.0, "learning_rate": 2.0922879272355922e-10, "loss": 0.2792, "step": 19987 }, { "epoch": 1.0, "learning_rate": 1.9889815532803115e-10, "loss": 0.3348, "step": 19988 }, { "epoch": 1.0, "learning_rate": 1.8882904963435367e-10, "loss": 0.4219, "step": 19989 }, { "epoch": 1.0, "learning_rate": 1.7902147590342921e-10, "loss": 0.3497, "step": 19990 }, { "epoch": 1.0, "learning_rate": 1.6947543439393976e-10, "loss": 0.3748, "step": 19991 }, { "epoch": 1.0, "learning_rate": 1.6019092535346503e-10, "loss": 0.2691, "step": 19992 }, { "epoch": 1.0, "learning_rate": 1.511679490262541e-10, "loss": 0.3825, "step": 19993 }, { "epoch": 1.0, "learning_rate": 1.42406505646564e-10, "loss": 0.218, "step": 19994 }, { "epoch": 1.0, "learning_rate": 1.3390659544643135e-10, "loss": 0.3536, "step": 19995 }, { "epoch": 1.0, "learning_rate": 1.2566821864568034e-10, "loss": 0.3515, "step": 19996 }, { "epoch": 1.0, "learning_rate": 1.1769137546080444e-10, "loss": 0.3821, "step": 19997 }, { "epoch": 1.0, "learning_rate": 1.0997606609941536e-10, "loss": 0.3211, "step": 19998 }, { "epoch": 1.0, "learning_rate": 1.025222907646839e-10, "loss": 0.2866, "step": 19999 }, { "epoch": 1.0, "learning_rate": 9.533004965200932e-11, "loss": 0.3342, "step": 20000 }, { "epoch": 1.0, "learning_rate": 8.839934294679886e-11, "loss": 0.4268, "step": 20001 }, { "epoch": 1.0, "learning_rate": 8.173017083334955e-11, "loss": 0.354, "step": 20002 }, { "epoch": 1.0, "learning_rate": 7.532253348374596e-11, "loss": 0.4704, "step": 20003 }, { "epoch": 1.0, "learning_rate": 6.917643106674199e-11, "loss": 0.3524, "step": 20004 }, { "epoch": 1.0, "learning_rate": 6.329186374331997e-11, "loss": 0.1508, "step": 20005 }, { "epoch": 1.0, "learning_rate": 5.7668831666690684e-11, "loss": 0.2556, "step": 20006 }, { "epoch": 1.0, "learning_rate": 5.230733498340357e-11, "loss": 0.3908, "step": 20007 }, { "epoch": 1.0, "learning_rate": 4.720737383556717e-11, "loss": 0.271, "step": 20008 }, { "epoch": 1.0, "learning_rate": 4.2368948355298034e-11, "loss": 0.2612, "step": 20009 }, { "epoch": 1.0, "learning_rate": 3.779205866805136e-11, "loss": 0.2866, "step": 20010 }, { "epoch": 1.0, "learning_rate": 3.347670489595167e-11, "loss": 0.3542, "step": 20011 }, { "epoch": 1.0, "learning_rate": 2.942288715002128e-11, "loss": 0.3166, "step": 20012 }, { "epoch": 1.0, "learning_rate": 2.5630605537951825e-11, "loss": 0.2158, "step": 20013 }, { "epoch": 1.0, "learning_rate": 2.209986015744292e-11, "loss": 0.3376, "step": 20014 }, { "epoch": 1.0, "learning_rate": 1.8830651100643083e-11, "loss": 0.3274, "step": 20015 }, { "epoch": 1.0, "learning_rate": 1.5822978454149705e-11, "loss": 0.2354, "step": 20016 }, { "epoch": 1.0, "learning_rate": 1.3076842295678404e-11, "loss": 0.7431, "step": 20017 }, { "epoch": 1.0, "learning_rate": 1.0592242698503896e-11, "loss": 0.3189, "step": 20018 }, { "epoch": 1.0, "learning_rate": 8.369179725908894e-12, "loss": 0.5439, "step": 20019 }, { "epoch": 1.0, "learning_rate": 6.40765343784544e-12, "loss": 0.3935, "step": 20020 }, { "epoch": 1.0, "learning_rate": 4.707663883163349e-12, "loss": 0.2578, "step": 20021 }, { "epoch": 1.0, "learning_rate": 3.2692111084919874e-12, "loss": 0.3454, "step": 20022 }, { "epoch": 1.0, "learning_rate": 2.092295150468715e-12, "loss": 0.3697, "step": 20023 }, { "epoch": 1.0, "learning_rate": 1.1769160401797763e-12, "loss": 0.1388, "step": 20024 }, { "epoch": 1.0, "learning_rate": 5.230738009398551e-13, "loss": 0.1661, "step": 20025 }, { "epoch": 1.0, "learning_rate": 1.3076845162274254e-13, "loss": 0.4048, "step": 20026 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.4797, "step": 20027 }, { "epoch": 1.0, "step": 20027, "total_flos": 4.346936596497181e+20, "train_loss": 0.3907263813688313, "train_runtime": 63489.0413, "train_samples_per_second": 5.047, "train_steps_per_second": 0.315 } ], "logging_steps": 1.0, "max_steps": 20027, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 200, "total_flos": 4.346936596497181e+20, "train_batch_size": 1, "trial_name": null, "trial_params": null }