{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.9995180722891566, "global_step": 4668, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.4184397163120568e-07, "loss": 0.0229, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.8368794326241136e-07, "loss": 0.0184, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.2553191489361704e-07, "loss": 0.0183, "step": 3 }, { "epoch": 0.0, "learning_rate": 5.673758865248227e-07, "loss": 0.0198, "step": 4 }, { "epoch": 0.0, "learning_rate": 7.092198581560285e-07, "loss": 0.0175, "step": 5 }, { "epoch": 0.0, "learning_rate": 8.510638297872341e-07, "loss": 0.0159, "step": 6 }, { "epoch": 0.0, "learning_rate": 9.929078014184399e-07, "loss": 0.0179, "step": 7 }, { "epoch": 0.01, "learning_rate": 1.1347517730496454e-06, "loss": 0.0176, "step": 8 }, { "epoch": 0.01, "learning_rate": 1.276595744680851e-06, "loss": 0.0159, "step": 9 }, { "epoch": 0.01, "learning_rate": 1.418439716312057e-06, "loss": 0.0164, "step": 10 }, { "epoch": 0.01, "learning_rate": 1.5602836879432626e-06, "loss": 0.0168, "step": 11 }, { "epoch": 0.01, "learning_rate": 1.7021276595744682e-06, "loss": 0.0465, "step": 12 }, { "epoch": 0.01, "learning_rate": 1.8439716312056737e-06, "loss": 0.0146, "step": 13 }, { "epoch": 0.01, "learning_rate": 1.9858156028368797e-06, "loss": 0.0141, "step": 14 }, { "epoch": 0.01, "learning_rate": 2.1276595744680853e-06, "loss": 0.0136, "step": 15 }, { "epoch": 0.01, "learning_rate": 2.269503546099291e-06, "loss": 0.0141, "step": 16 }, { "epoch": 0.01, "learning_rate": 2.4113475177304965e-06, "loss": 0.0127, "step": 17 }, { "epoch": 0.01, "learning_rate": 2.553191489361702e-06, "loss": 0.0124, "step": 18 }, { "epoch": 0.01, "learning_rate": 2.695035460992908e-06, "loss": 0.0151, "step": 19 }, { "epoch": 0.01, "learning_rate": 2.836879432624114e-06, "loss": 0.0146, "step": 20 }, { "epoch": 0.01, "learning_rate": 2.978723404255319e-06, "loss": 0.0118, "step": 21 }, { "epoch": 0.01, "learning_rate": 3.120567375886525e-06, "loss": 0.0135, "step": 22 }, { "epoch": 0.01, "learning_rate": 3.262411347517731e-06, "loss": 0.0133, "step": 23 }, { "epoch": 0.02, "learning_rate": 3.4042553191489363e-06, "loss": 0.0138, "step": 24 }, { "epoch": 0.02, "learning_rate": 3.5460992907801423e-06, "loss": 0.0138, "step": 25 }, { "epoch": 0.02, "learning_rate": 3.6879432624113475e-06, "loss": 0.012, "step": 26 }, { "epoch": 0.02, "learning_rate": 3.8297872340425535e-06, "loss": 0.0117, "step": 27 }, { "epoch": 0.02, "learning_rate": 3.9716312056737595e-06, "loss": 0.0147, "step": 28 }, { "epoch": 0.02, "learning_rate": 4.113475177304965e-06, "loss": 0.0142, "step": 29 }, { "epoch": 0.02, "learning_rate": 4.255319148936171e-06, "loss": 0.0127, "step": 30 }, { "epoch": 0.02, "learning_rate": 4.397163120567377e-06, "loss": 0.0148, "step": 31 }, { "epoch": 0.02, "learning_rate": 4.539007092198582e-06, "loss": 0.0107, "step": 32 }, { "epoch": 0.02, "learning_rate": 4.680851063829788e-06, "loss": 0.0112, "step": 33 }, { "epoch": 0.02, "learning_rate": 4.822695035460993e-06, "loss": 0.0124, "step": 34 }, { "epoch": 0.02, "learning_rate": 4.964539007092199e-06, "loss": 0.0144, "step": 35 }, { "epoch": 0.02, "learning_rate": 5.106382978723404e-06, "loss": 0.0112, "step": 36 }, { "epoch": 0.02, "learning_rate": 5.24822695035461e-06, "loss": 0.0133, "step": 37 }, { "epoch": 0.02, "learning_rate": 5.390070921985816e-06, "loss": 0.0122, "step": 38 }, { "epoch": 0.03, "learning_rate": 5.531914893617022e-06, "loss": 0.0113, "step": 39 }, { "epoch": 0.03, "learning_rate": 5.673758865248228e-06, "loss": 0.0124, "step": 40 }, { "epoch": 0.03, "learning_rate": 5.815602836879432e-06, "loss": 0.0116, "step": 41 }, { "epoch": 0.03, "learning_rate": 5.957446808510638e-06, "loss": 0.0114, "step": 42 }, { "epoch": 0.03, "learning_rate": 6.099290780141844e-06, "loss": 0.0123, "step": 43 }, { "epoch": 0.03, "learning_rate": 6.24113475177305e-06, "loss": 0.0107, "step": 44 }, { "epoch": 0.03, "learning_rate": 6.382978723404256e-06, "loss": 0.0106, "step": 45 }, { "epoch": 0.03, "learning_rate": 6.524822695035462e-06, "loss": 0.0121, "step": 46 }, { "epoch": 0.03, "learning_rate": 6.666666666666667e-06, "loss": 0.0096, "step": 47 }, { "epoch": 0.03, "learning_rate": 6.808510638297873e-06, "loss": 0.0103, "step": 48 }, { "epoch": 0.03, "learning_rate": 6.950354609929079e-06, "loss": 0.0116, "step": 49 }, { "epoch": 0.03, "learning_rate": 7.092198581560285e-06, "loss": 0.0114, "step": 50 }, { "epoch": 0.03, "learning_rate": 7.234042553191491e-06, "loss": 0.0112, "step": 51 }, { "epoch": 0.03, "learning_rate": 7.375886524822695e-06, "loss": 0.0097, "step": 52 }, { "epoch": 0.03, "learning_rate": 7.517730496453901e-06, "loss": 0.0101, "step": 53 }, { "epoch": 0.03, "learning_rate": 7.659574468085107e-06, "loss": 0.012, "step": 54 }, { "epoch": 0.04, "learning_rate": 7.801418439716313e-06, "loss": 0.0086, "step": 55 }, { "epoch": 0.04, "learning_rate": 7.943262411347519e-06, "loss": 0.0091, "step": 56 }, { "epoch": 0.04, "learning_rate": 8.085106382978723e-06, "loss": 0.0097, "step": 57 }, { "epoch": 0.04, "learning_rate": 8.22695035460993e-06, "loss": 0.0095, "step": 58 }, { "epoch": 0.04, "learning_rate": 8.368794326241135e-06, "loss": 0.0071, "step": 59 }, { "epoch": 0.04, "learning_rate": 8.510638297872341e-06, "loss": 0.0084, "step": 60 }, { "epoch": 0.04, "learning_rate": 8.652482269503547e-06, "loss": 0.007, "step": 61 }, { "epoch": 0.04, "learning_rate": 8.794326241134753e-06, "loss": 0.0069, "step": 62 }, { "epoch": 0.04, "learning_rate": 8.936170212765958e-06, "loss": 0.0071, "step": 63 }, { "epoch": 0.04, "learning_rate": 9.078014184397164e-06, "loss": 0.0051, "step": 64 }, { "epoch": 0.04, "learning_rate": 9.21985815602837e-06, "loss": 0.0059, "step": 65 }, { "epoch": 0.04, "learning_rate": 9.361702127659576e-06, "loss": 0.0049, "step": 66 }, { "epoch": 0.04, "learning_rate": 9.503546099290782e-06, "loss": 0.0042, "step": 67 }, { "epoch": 0.04, "learning_rate": 9.645390070921986e-06, "loss": 0.0037, "step": 68 }, { "epoch": 0.04, "learning_rate": 9.787234042553192e-06, "loss": 0.0061, "step": 69 }, { "epoch": 0.04, "learning_rate": 9.929078014184398e-06, "loss": 0.0026, "step": 70 }, { "epoch": 0.05, "learning_rate": 1.0070921985815602e-05, "loss": 0.0017, "step": 71 }, { "epoch": 0.05, "learning_rate": 1.0212765957446808e-05, "loss": 0.0017, "step": 72 }, { "epoch": 0.05, "learning_rate": 1.0354609929078014e-05, "loss": 0.0023, "step": 73 }, { "epoch": 0.05, "learning_rate": 1.049645390070922e-05, "loss": 0.0101, "step": 74 }, { "epoch": 0.05, "learning_rate": 1.0638297872340426e-05, "loss": 0.0023, "step": 75 }, { "epoch": 0.05, "learning_rate": 1.0780141843971632e-05, "loss": 0.0038, "step": 76 }, { "epoch": 0.05, "learning_rate": 1.0921985815602838e-05, "loss": 0.0016, "step": 77 }, { "epoch": 0.05, "learning_rate": 1.1063829787234044e-05, "loss": 0.0033, "step": 78 }, { "epoch": 0.05, "learning_rate": 1.120567375886525e-05, "loss": 0.002, "step": 79 }, { "epoch": 0.05, "learning_rate": 1.1347517730496456e-05, "loss": 0.0016, "step": 80 }, { "epoch": 0.05, "learning_rate": 1.1489361702127662e-05, "loss": 0.0024, "step": 81 }, { "epoch": 0.05, "learning_rate": 1.1631205673758865e-05, "loss": 0.0028, "step": 82 }, { "epoch": 0.05, "learning_rate": 1.177304964539007e-05, "loss": 0.0015, "step": 83 }, { "epoch": 0.05, "learning_rate": 1.1914893617021277e-05, "loss": 0.0037, "step": 84 }, { "epoch": 0.05, "learning_rate": 1.2056737588652483e-05, "loss": 0.0032, "step": 85 }, { "epoch": 0.06, "learning_rate": 1.2198581560283689e-05, "loss": 0.0015, "step": 86 }, { "epoch": 0.06, "learning_rate": 1.2340425531914895e-05, "loss": 0.0027, "step": 87 }, { "epoch": 0.06, "learning_rate": 1.24822695035461e-05, "loss": 0.0023, "step": 88 }, { "epoch": 0.06, "learning_rate": 1.2624113475177307e-05, "loss": 0.0032, "step": 89 }, { "epoch": 0.06, "learning_rate": 1.2765957446808513e-05, "loss": 0.0044, "step": 90 }, { "epoch": 0.06, "learning_rate": 1.2907801418439719e-05, "loss": 0.0021, "step": 91 }, { "epoch": 0.06, "learning_rate": 1.3049645390070925e-05, "loss": 0.0024, "step": 92 }, { "epoch": 0.06, "learning_rate": 1.3191489361702127e-05, "loss": 0.0029, "step": 93 }, { "epoch": 0.06, "learning_rate": 1.3333333333333333e-05, "loss": 0.0036, "step": 94 }, { "epoch": 0.06, "learning_rate": 1.347517730496454e-05, "loss": 0.003, "step": 95 }, { "epoch": 0.06, "learning_rate": 1.3617021276595745e-05, "loss": 0.0018, "step": 96 }, { "epoch": 0.06, "learning_rate": 1.3758865248226951e-05, "loss": 0.0018, "step": 97 }, { "epoch": 0.06, "learning_rate": 1.3900709219858157e-05, "loss": 0.0028, "step": 98 }, { "epoch": 0.06, "learning_rate": 1.4042553191489363e-05, "loss": 0.0023, "step": 99 }, { "epoch": 0.06, "learning_rate": 1.418439716312057e-05, "loss": 0.0022, "step": 100 }, { "epoch": 0.06, "learning_rate": 1.4326241134751775e-05, "loss": 0.0022, "step": 101 }, { "epoch": 0.07, "learning_rate": 1.4468085106382981e-05, "loss": 0.0025, "step": 102 }, { "epoch": 0.07, "learning_rate": 1.4609929078014187e-05, "loss": 0.0018, "step": 103 }, { "epoch": 0.07, "learning_rate": 1.475177304964539e-05, "loss": 0.0053, "step": 104 }, { "epoch": 0.07, "learning_rate": 1.4893617021276596e-05, "loss": 0.0017, "step": 105 }, { "epoch": 0.07, "learning_rate": 1.5035460992907802e-05, "loss": 0.0021, "step": 106 }, { "epoch": 0.07, "learning_rate": 1.5177304964539008e-05, "loss": 0.002, "step": 107 }, { "epoch": 0.07, "learning_rate": 1.5319148936170214e-05, "loss": 0.0017, "step": 108 }, { "epoch": 0.07, "learning_rate": 1.546099290780142e-05, "loss": 0.0028, "step": 109 }, { "epoch": 0.07, "learning_rate": 1.5602836879432626e-05, "loss": 0.0018, "step": 110 }, { "epoch": 0.07, "learning_rate": 1.5744680851063832e-05, "loss": 0.0038, "step": 111 }, { "epoch": 0.07, "learning_rate": 1.5886524822695038e-05, "loss": 0.0024, "step": 112 }, { "epoch": 0.07, "learning_rate": 1.6028368794326244e-05, "loss": 0.0026, "step": 113 }, { "epoch": 0.07, "learning_rate": 1.6170212765957446e-05, "loss": 0.0018, "step": 114 }, { "epoch": 0.07, "learning_rate": 1.6312056737588652e-05, "loss": 0.0012, "step": 115 }, { "epoch": 0.07, "learning_rate": 1.645390070921986e-05, "loss": 0.0076, "step": 116 }, { "epoch": 0.08, "learning_rate": 1.6595744680851064e-05, "loss": 0.0023, "step": 117 }, { "epoch": 0.08, "learning_rate": 1.673758865248227e-05, "loss": 0.0029, "step": 118 }, { "epoch": 0.08, "learning_rate": 1.6879432624113476e-05, "loss": 0.0036, "step": 119 }, { "epoch": 0.08, "learning_rate": 1.7021276595744682e-05, "loss": 0.0022, "step": 120 }, { "epoch": 0.08, "learning_rate": 1.716312056737589e-05, "loss": 0.0029, "step": 121 }, { "epoch": 0.08, "learning_rate": 1.7304964539007094e-05, "loss": 0.0021, "step": 122 }, { "epoch": 0.08, "learning_rate": 1.74468085106383e-05, "loss": 0.0021, "step": 123 }, { "epoch": 0.08, "learning_rate": 1.7588652482269506e-05, "loss": 0.0023, "step": 124 }, { "epoch": 0.08, "learning_rate": 1.773049645390071e-05, "loss": 0.0017, "step": 125 }, { "epoch": 0.08, "learning_rate": 1.7872340425531915e-05, "loss": 0.0038, "step": 126 }, { "epoch": 0.08, "learning_rate": 1.801418439716312e-05, "loss": 0.0018, "step": 127 }, { "epoch": 0.08, "learning_rate": 1.8156028368794327e-05, "loss": 0.0035, "step": 128 }, { "epoch": 0.08, "learning_rate": 1.8297872340425533e-05, "loss": 0.0017, "step": 129 }, { "epoch": 0.08, "learning_rate": 1.843971631205674e-05, "loss": 0.0017, "step": 130 }, { "epoch": 0.08, "learning_rate": 1.8581560283687945e-05, "loss": 0.0023, "step": 131 }, { "epoch": 0.08, "learning_rate": 1.872340425531915e-05, "loss": 0.0024, "step": 132 }, { "epoch": 0.09, "learning_rate": 1.8865248226950357e-05, "loss": 0.0059, "step": 133 }, { "epoch": 0.09, "learning_rate": 1.9007092198581563e-05, "loss": 0.0047, "step": 134 }, { "epoch": 0.09, "learning_rate": 1.914893617021277e-05, "loss": 0.0054, "step": 135 }, { "epoch": 0.09, "learning_rate": 1.929078014184397e-05, "loss": 0.0032, "step": 136 }, { "epoch": 0.09, "learning_rate": 1.9432624113475178e-05, "loss": 0.0014, "step": 137 }, { "epoch": 0.09, "learning_rate": 1.9574468085106384e-05, "loss": 0.0019, "step": 138 }, { "epoch": 0.09, "learning_rate": 1.971631205673759e-05, "loss": 0.0023, "step": 139 }, { "epoch": 0.09, "learning_rate": 1.9858156028368796e-05, "loss": 0.0018, "step": 140 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 0.003, "step": 141 }, { "epoch": 0.09, "learning_rate": 1.999999759204291e-05, "loss": 0.0018, "step": 142 }, { "epoch": 0.09, "learning_rate": 1.9999990368172805e-05, "loss": 0.0028, "step": 143 }, { "epoch": 0.09, "learning_rate": 1.999997832839316e-05, "loss": 0.0028, "step": 144 }, { "epoch": 0.09, "learning_rate": 1.9999961472709774e-05, "loss": 0.002, "step": 145 }, { "epoch": 0.09, "learning_rate": 1.999993980113076e-05, "loss": 0.0016, "step": 146 }, { "epoch": 0.09, "learning_rate": 1.999991331366656e-05, "loss": 0.0014, "step": 147 }, { "epoch": 0.1, "learning_rate": 1.9999882010329935e-05, "loss": 0.003, "step": 148 }, { "epoch": 0.1, "learning_rate": 1.9999845891135948e-05, "loss": 0.0022, "step": 149 }, { "epoch": 0.1, "learning_rate": 1.9999804956102004e-05, "loss": 0.0017, "step": 150 }, { "epoch": 0.1, "learning_rate": 1.9999759205247815e-05, "loss": 0.0018, "step": 151 }, { "epoch": 0.1, "learning_rate": 1.999970863859541e-05, "loss": 0.002, "step": 152 }, { "epoch": 0.1, "learning_rate": 1.9999653256169146e-05, "loss": 0.0016, "step": 153 }, { "epoch": 0.1, "learning_rate": 1.9999593057995694e-05, "loss": 0.0024, "step": 154 }, { "epoch": 0.1, "learning_rate": 1.999952804410404e-05, "loss": 0.0018, "step": 155 }, { "epoch": 0.1, "learning_rate": 1.9999458214525503e-05, "loss": 0.0017, "step": 156 }, { "epoch": 0.1, "learning_rate": 1.9999383569293704e-05, "loss": 0.0026, "step": 157 }, { "epoch": 0.1, "learning_rate": 1.9999304108444596e-05, "loss": 0.0019, "step": 158 }, { "epoch": 0.1, "learning_rate": 1.9999219832016444e-05, "loss": 0.0028, "step": 159 }, { "epoch": 0.1, "learning_rate": 1.999913074004984e-05, "loss": 0.0019, "step": 160 }, { "epoch": 0.1, "learning_rate": 1.9999036832587683e-05, "loss": 0.0016, "step": 161 }, { "epoch": 0.1, "learning_rate": 1.9998938109675198e-05, "loss": 0.0022, "step": 162 }, { "epoch": 0.1, "learning_rate": 1.9998834571359937e-05, "loss": 0.0022, "step": 163 }, { "epoch": 0.11, "learning_rate": 1.9998726217691757e-05, "loss": 0.0017, "step": 164 }, { "epoch": 0.11, "learning_rate": 1.999861304872284e-05, "loss": 0.0025, "step": 165 }, { "epoch": 0.11, "learning_rate": 1.9998495064507693e-05, "loss": 0.0014, "step": 166 }, { "epoch": 0.11, "learning_rate": 1.999837226510313e-05, "loss": 0.002, "step": 167 }, { "epoch": 0.11, "learning_rate": 1.999824465056829e-05, "loss": 0.0031, "step": 168 }, { "epoch": 0.11, "learning_rate": 1.9998112220964638e-05, "loss": 0.003, "step": 169 }, { "epoch": 0.11, "learning_rate": 1.999797497635594e-05, "loss": 0.0016, "step": 170 }, { "epoch": 0.11, "learning_rate": 1.9997832916808304e-05, "loss": 0.0028, "step": 171 }, { "epoch": 0.11, "learning_rate": 1.9997686042390135e-05, "loss": 0.0021, "step": 172 }, { "epoch": 0.11, "learning_rate": 1.9997534353172172e-05, "loss": 0.0019, "step": 173 }, { "epoch": 0.11, "learning_rate": 1.9997377849227467e-05, "loss": 0.0027, "step": 174 }, { "epoch": 0.11, "learning_rate": 1.9997216530631385e-05, "loss": 0.0029, "step": 175 }, { "epoch": 0.11, "learning_rate": 1.9997050397461624e-05, "loss": 0.0038, "step": 176 }, { "epoch": 0.11, "learning_rate": 1.999687944979819e-05, "loss": 0.0017, "step": 177 }, { "epoch": 0.11, "learning_rate": 1.9996703687723403e-05, "loss": 0.002, "step": 178 }, { "epoch": 0.12, "learning_rate": 1.9996523111321915e-05, "loss": 0.0019, "step": 179 }, { "epoch": 0.12, "learning_rate": 1.9996337720680688e-05, "loss": 0.0013, "step": 180 }, { "epoch": 0.12, "learning_rate": 1.999614751588901e-05, "loss": 0.0028, "step": 181 }, { "epoch": 0.12, "learning_rate": 1.9995952497038474e-05, "loss": 0.004, "step": 182 }, { "epoch": 0.12, "learning_rate": 1.9995752664223e-05, "loss": 0.0022, "step": 183 }, { "epoch": 0.12, "learning_rate": 1.9995548017538833e-05, "loss": 0.0013, "step": 184 }, { "epoch": 0.12, "learning_rate": 1.9995338557084522e-05, "loss": 0.0016, "step": 185 }, { "epoch": 0.12, "learning_rate": 1.9995124282960944e-05, "loss": 0.0025, "step": 186 }, { "epoch": 0.12, "learning_rate": 1.9994905195271294e-05, "loss": 0.0019, "step": 187 }, { "epoch": 0.12, "learning_rate": 1.999468129412108e-05, "loss": 0.0026, "step": 188 }, { "epoch": 0.12, "learning_rate": 1.9994452579618132e-05, "loss": 0.0011, "step": 189 }, { "epoch": 0.12, "learning_rate": 1.9994219051872594e-05, "loss": 0.0019, "step": 190 }, { "epoch": 0.12, "learning_rate": 1.9993980710996932e-05, "loss": 0.0031, "step": 191 }, { "epoch": 0.12, "learning_rate": 1.9993737557105932e-05, "loss": 0.002, "step": 192 }, { "epoch": 0.12, "learning_rate": 1.9993489590316693e-05, "loss": 0.002, "step": 193 }, { "epoch": 0.12, "learning_rate": 1.999323681074863e-05, "loss": 0.0036, "step": 194 }, { "epoch": 0.13, "learning_rate": 1.9992979218523488e-05, "loss": 0.0018, "step": 195 }, { "epoch": 0.13, "learning_rate": 1.999271681376531e-05, "loss": 0.0027, "step": 196 }, { "epoch": 0.13, "learning_rate": 1.999244959660048e-05, "loss": 0.0018, "step": 197 }, { "epoch": 0.13, "learning_rate": 1.9992177567157678e-05, "loss": 0.0036, "step": 198 }, { "epoch": 0.13, "learning_rate": 1.9991900725567915e-05, "loss": 0.0026, "step": 199 }, { "epoch": 0.13, "learning_rate": 1.9991619071964518e-05, "loss": 0.0009, "step": 200 }, { "epoch": 0.13, "learning_rate": 1.9991332606483126e-05, "loss": 0.0022, "step": 201 }, { "epoch": 0.13, "learning_rate": 1.9991041329261694e-05, "loss": 0.0032, "step": 202 }, { "epoch": 0.13, "learning_rate": 1.9990745240440505e-05, "loss": 0.0023, "step": 203 }, { "epoch": 0.13, "learning_rate": 1.999044434016215e-05, "loss": 0.0018, "step": 204 }, { "epoch": 0.13, "learning_rate": 1.9990138628571543e-05, "loss": 0.0033, "step": 205 }, { "epoch": 0.13, "learning_rate": 1.9989828105815912e-05, "loss": 0.0012, "step": 206 }, { "epoch": 0.13, "learning_rate": 1.9989512772044794e-05, "loss": 0.0029, "step": 207 }, { "epoch": 0.13, "learning_rate": 1.998919262741006e-05, "loss": 0.0017, "step": 208 }, { "epoch": 0.13, "learning_rate": 1.998886767206589e-05, "loss": 0.0019, "step": 209 }, { "epoch": 0.13, "learning_rate": 1.998853790616877e-05, "loss": 0.0022, "step": 210 }, { "epoch": 0.14, "learning_rate": 1.998820332987752e-05, "loss": 0.0012, "step": 211 }, { "epoch": 0.14, "learning_rate": 1.9987863943353266e-05, "loss": 0.0012, "step": 212 }, { "epoch": 0.14, "learning_rate": 1.9987519746759456e-05, "loss": 0.0019, "step": 213 }, { "epoch": 0.14, "learning_rate": 1.998717074026185e-05, "loss": 0.0022, "step": 214 }, { "epoch": 0.14, "learning_rate": 1.998681692402853e-05, "loss": 0.0021, "step": 215 }, { "epoch": 0.14, "learning_rate": 1.9986458298229887e-05, "loss": 0.002, "step": 216 }, { "epoch": 0.14, "learning_rate": 1.998609486303863e-05, "loss": 0.0019, "step": 217 }, { "epoch": 0.14, "learning_rate": 1.9985726618629793e-05, "loss": 0.0012, "step": 218 }, { "epoch": 0.14, "learning_rate": 1.9985353565180716e-05, "loss": 0.0024, "step": 219 }, { "epoch": 0.14, "learning_rate": 1.9984975702871058e-05, "loss": 0.003, "step": 220 }, { "epoch": 0.14, "learning_rate": 1.9984593031882794e-05, "loss": 0.0027, "step": 221 }, { "epoch": 0.14, "learning_rate": 1.9984205552400215e-05, "loss": 0.001, "step": 222 }, { "epoch": 0.14, "learning_rate": 1.998381326460993e-05, "loss": 0.0014, "step": 223 }, { "epoch": 0.14, "learning_rate": 1.998341616870086e-05, "loss": 0.0051, "step": 224 }, { "epoch": 0.14, "learning_rate": 1.998301426486424e-05, "loss": 0.002, "step": 225 }, { "epoch": 0.15, "learning_rate": 1.998260755329363e-05, "loss": 0.0016, "step": 226 }, { "epoch": 0.15, "learning_rate": 1.9982196034184893e-05, "loss": 0.0013, "step": 227 }, { "epoch": 0.15, "learning_rate": 1.998177970773622e-05, "loss": 0.0053, "step": 228 }, { "epoch": 0.15, "learning_rate": 1.99813585741481e-05, "loss": 0.0011, "step": 229 }, { "epoch": 0.15, "learning_rate": 1.9980932633623353e-05, "loss": 0.004, "step": 230 }, { "epoch": 0.15, "learning_rate": 1.998050188636711e-05, "loss": 0.003, "step": 231 }, { "epoch": 0.15, "learning_rate": 1.9980066332586815e-05, "loss": 0.0018, "step": 232 }, { "epoch": 0.15, "learning_rate": 1.997962597249222e-05, "loss": 0.0022, "step": 233 }, { "epoch": 0.15, "learning_rate": 1.9979180806295408e-05, "loss": 0.0018, "step": 234 }, { "epoch": 0.15, "learning_rate": 1.9978730834210763e-05, "loss": 0.0022, "step": 235 }, { "epoch": 0.15, "learning_rate": 1.9978276056454982e-05, "loss": 0.0018, "step": 236 }, { "epoch": 0.15, "learning_rate": 1.9977816473247093e-05, "loss": 0.0019, "step": 237 }, { "epoch": 0.15, "learning_rate": 1.9977352084808423e-05, "loss": 0.0023, "step": 238 }, { "epoch": 0.15, "learning_rate": 1.9976882891362616e-05, "loss": 0.0038, "step": 239 }, { "epoch": 0.15, "learning_rate": 1.9976408893135632e-05, "loss": 0.0016, "step": 240 }, { "epoch": 0.15, "learning_rate": 1.9975930090355744e-05, "loss": 0.0027, "step": 241 }, { "epoch": 0.16, "learning_rate": 1.997544648325354e-05, "loss": 0.0018, "step": 242 }, { "epoch": 0.16, "learning_rate": 1.997495807206192e-05, "loss": 0.0016, "step": 243 }, { "epoch": 0.16, "learning_rate": 1.99744648570161e-05, "loss": 0.0014, "step": 244 }, { "epoch": 0.16, "learning_rate": 1.9973966838353606e-05, "loss": 0.0018, "step": 245 }, { "epoch": 0.16, "learning_rate": 1.9973464016314284e-05, "loss": 0.0026, "step": 246 }, { "epoch": 0.16, "learning_rate": 1.997295639114029e-05, "loss": 0.0067, "step": 247 }, { "epoch": 0.16, "learning_rate": 1.997244396307608e-05, "loss": 0.0023, "step": 248 }, { "epoch": 0.16, "learning_rate": 1.9971926732368446e-05, "loss": 0.0016, "step": 249 }, { "epoch": 0.16, "learning_rate": 1.9971404699266482e-05, "loss": 0.0012, "step": 250 }, { "epoch": 0.16, "learning_rate": 1.9970877864021588e-05, "loss": 0.0012, "step": 251 }, { "epoch": 0.16, "learning_rate": 1.997034622688749e-05, "loss": 0.0018, "step": 252 }, { "epoch": 0.16, "learning_rate": 1.996980978812021e-05, "loss": 0.0021, "step": 253 }, { "epoch": 0.16, "learning_rate": 1.9969268547978107e-05, "loss": 0.0027, "step": 254 }, { "epoch": 0.16, "learning_rate": 1.9968722506721825e-05, "loss": 0.0018, "step": 255 }, { "epoch": 0.16, "learning_rate": 1.9968171664614336e-05, "loss": 0.0018, "step": 256 }, { "epoch": 0.17, "learning_rate": 1.9967616021920927e-05, "loss": 0.0022, "step": 257 }, { "epoch": 0.17, "learning_rate": 1.9967055578909185e-05, "loss": 0.0014, "step": 258 }, { "epoch": 0.17, "learning_rate": 1.9966490335849013e-05, "loss": 0.0031, "step": 259 }, { "epoch": 0.17, "learning_rate": 1.996592029301263e-05, "loss": 0.0021, "step": 260 }, { "epoch": 0.17, "learning_rate": 1.9965345450674566e-05, "loss": 0.0023, "step": 261 }, { "epoch": 0.17, "learning_rate": 1.9964765809111655e-05, "loss": 0.0018, "step": 262 }, { "epoch": 0.17, "learning_rate": 1.9964181368603052e-05, "loss": 0.0015, "step": 263 }, { "epoch": 0.17, "learning_rate": 1.9963592129430217e-05, "loss": 0.0021, "step": 264 }, { "epoch": 0.17, "learning_rate": 1.9962998091876918e-05, "loss": 0.0027, "step": 265 }, { "epoch": 0.17, "learning_rate": 1.9962399256229246e-05, "loss": 0.0016, "step": 266 }, { "epoch": 0.17, "learning_rate": 1.996179562277559e-05, "loss": 0.0022, "step": 267 }, { "epoch": 0.17, "learning_rate": 1.996118719180666e-05, "loss": 0.0031, "step": 268 }, { "epoch": 0.17, "learning_rate": 1.996057396361546e-05, "loss": 0.0019, "step": 269 }, { "epoch": 0.17, "learning_rate": 1.995995593849733e-05, "loss": 0.0009, "step": 270 }, { "epoch": 0.17, "learning_rate": 1.9959333116749896e-05, "loss": 0.0017, "step": 271 }, { "epoch": 0.17, "learning_rate": 1.9958705498673107e-05, "loss": 0.0015, "step": 272 }, { "epoch": 0.18, "learning_rate": 1.995807308456921e-05, "loss": 0.0011, "step": 273 }, { "epoch": 0.18, "learning_rate": 1.995743587474279e-05, "loss": 0.002, "step": 274 }, { "epoch": 0.18, "learning_rate": 1.99567938695007e-05, "loss": 0.0014, "step": 275 }, { "epoch": 0.18, "learning_rate": 1.9956147069152138e-05, "loss": 0.0015, "step": 276 }, { "epoch": 0.18, "learning_rate": 1.995549547400859e-05, "loss": 0.0089, "step": 277 }, { "epoch": 0.18, "learning_rate": 1.9954839084383866e-05, "loss": 0.0023, "step": 278 }, { "epoch": 0.18, "learning_rate": 1.9954177900594073e-05, "loss": 0.0014, "step": 279 }, { "epoch": 0.18, "learning_rate": 1.9953511922957628e-05, "loss": 0.0012, "step": 280 }, { "epoch": 0.18, "learning_rate": 1.9952841151795265e-05, "loss": 0.0019, "step": 281 }, { "epoch": 0.18, "learning_rate": 1.9952165587430023e-05, "loss": 0.0047, "step": 282 }, { "epoch": 0.18, "learning_rate": 1.9951485230187244e-05, "loss": 0.0028, "step": 283 }, { "epoch": 0.18, "learning_rate": 1.9950800080394583e-05, "loss": 0.0021, "step": 284 }, { "epoch": 0.18, "learning_rate": 1.9950110138382007e-05, "loss": 0.0023, "step": 285 }, { "epoch": 0.18, "learning_rate": 1.9949415404481777e-05, "loss": 0.0017, "step": 286 }, { "epoch": 0.18, "learning_rate": 1.994871587902848e-05, "loss": 0.0023, "step": 287 }, { "epoch": 0.19, "learning_rate": 1.9948011562358997e-05, "loss": 0.0013, "step": 288 }, { "epoch": 0.19, "learning_rate": 1.994730245481252e-05, "loss": 0.0009, "step": 289 }, { "epoch": 0.19, "learning_rate": 1.994658855673055e-05, "loss": 0.0026, "step": 290 }, { "epoch": 0.19, "learning_rate": 1.9945869868456897e-05, "loss": 0.0012, "step": 291 }, { "epoch": 0.19, "learning_rate": 1.9945146390337674e-05, "loss": 0.0023, "step": 292 }, { "epoch": 0.19, "learning_rate": 1.99444181227213e-05, "loss": 0.0012, "step": 293 }, { "epoch": 0.19, "learning_rate": 1.99436850659585e-05, "loss": 0.0021, "step": 294 }, { "epoch": 0.19, "learning_rate": 1.9942947220402313e-05, "loss": 0.0352, "step": 295 }, { "epoch": 0.19, "learning_rate": 1.9942204586408076e-05, "loss": 0.0037, "step": 296 }, { "epoch": 0.19, "learning_rate": 1.994145716433344e-05, "loss": 0.002, "step": 297 }, { "epoch": 0.19, "learning_rate": 1.9940704954538347e-05, "loss": 0.0013, "step": 298 }, { "epoch": 0.19, "learning_rate": 1.9939947957385068e-05, "loss": 0.0021, "step": 299 }, { "epoch": 0.19, "learning_rate": 1.9939186173238155e-05, "loss": 0.0019, "step": 300 }, { "epoch": 0.19, "learning_rate": 1.993841960246448e-05, "loss": 0.0028, "step": 301 }, { "epoch": 0.19, "learning_rate": 1.9937648245433223e-05, "loss": 0.0017, "step": 302 }, { "epoch": 0.19, "learning_rate": 1.9936872102515854e-05, "loss": 0.0017, "step": 303 }, { "epoch": 0.2, "learning_rate": 1.993609117408616e-05, "loss": 0.0018, "step": 304 }, { "epoch": 0.2, "learning_rate": 1.9935305460520232e-05, "loss": 0.0018, "step": 305 }, { "epoch": 0.2, "learning_rate": 1.9934514962196464e-05, "loss": 0.0023, "step": 306 }, { "epoch": 0.2, "learning_rate": 1.9933719679495547e-05, "loss": 0.0015, "step": 307 }, { "epoch": 0.2, "learning_rate": 1.9932919612800482e-05, "loss": 0.0016, "step": 308 }, { "epoch": 0.2, "learning_rate": 1.9932114762496582e-05, "loss": 0.002, "step": 309 }, { "epoch": 0.2, "learning_rate": 1.9931305128971453e-05, "loss": 0.0015, "step": 310 }, { "epoch": 0.2, "learning_rate": 1.9930490712615006e-05, "loss": 0.0011, "step": 311 }, { "epoch": 0.2, "learning_rate": 1.9929671513819456e-05, "loss": 0.0014, "step": 312 }, { "epoch": 0.2, "learning_rate": 1.9928847532979324e-05, "loss": 0.0033, "step": 313 }, { "epoch": 0.2, "learning_rate": 1.9928018770491427e-05, "loss": 0.0016, "step": 314 }, { "epoch": 0.2, "learning_rate": 1.99271852267549e-05, "loss": 0.0014, "step": 315 }, { "epoch": 0.2, "learning_rate": 1.9926346902171165e-05, "loss": 0.001, "step": 316 }, { "epoch": 0.2, "learning_rate": 1.992550379714395e-05, "loss": 0.0011, "step": 317 }, { "epoch": 0.2, "learning_rate": 1.992465591207929e-05, "loss": 0.0027, "step": 318 }, { "epoch": 0.2, "learning_rate": 1.992380324738552e-05, "loss": 0.0028, "step": 319 }, { "epoch": 0.21, "learning_rate": 1.992294580347327e-05, "loss": 0.0019, "step": 320 }, { "epoch": 0.21, "learning_rate": 1.9922083580755483e-05, "loss": 0.0009, "step": 321 }, { "epoch": 0.21, "learning_rate": 1.99212165796474e-05, "loss": 0.0011, "step": 322 }, { "epoch": 0.21, "learning_rate": 1.9920344800566553e-05, "loss": 0.0007, "step": 323 }, { "epoch": 0.21, "learning_rate": 1.991946824393279e-05, "loss": 0.0018, "step": 324 }, { "epoch": 0.21, "learning_rate": 1.9918586910168253e-05, "loss": 0.001, "step": 325 }, { "epoch": 0.21, "learning_rate": 1.9917700799697382e-05, "loss": 0.0009, "step": 326 }, { "epoch": 0.21, "learning_rate": 1.9916809912946924e-05, "loss": 0.0023, "step": 327 }, { "epoch": 0.21, "learning_rate": 1.991591425034592e-05, "loss": 0.001, "step": 328 }, { "epoch": 0.21, "learning_rate": 1.9915013812325712e-05, "loss": 0.0015, "step": 329 }, { "epoch": 0.21, "learning_rate": 1.9914108599319942e-05, "loss": 0.0024, "step": 330 }, { "epoch": 0.21, "learning_rate": 1.991319861176456e-05, "loss": 0.0034, "step": 331 }, { "epoch": 0.21, "learning_rate": 1.9912283850097803e-05, "loss": 0.0008, "step": 332 }, { "epoch": 0.21, "learning_rate": 1.9911364314760213e-05, "loss": 0.0022, "step": 333 }, { "epoch": 0.21, "learning_rate": 1.9910440006194627e-05, "loss": 0.0014, "step": 334 }, { "epoch": 0.22, "learning_rate": 1.9909510924846192e-05, "loss": 0.001, "step": 335 }, { "epoch": 0.22, "learning_rate": 1.990857707116234e-05, "loss": 0.0016, "step": 336 }, { "epoch": 0.22, "learning_rate": 1.9907638445592808e-05, "loss": 0.0019, "step": 337 }, { "epoch": 0.22, "learning_rate": 1.9906695048589628e-05, "loss": 0.0016, "step": 338 }, { "epoch": 0.22, "learning_rate": 1.9905746880607135e-05, "loss": 0.0018, "step": 339 }, { "epoch": 0.22, "learning_rate": 1.9904793942101957e-05, "loss": 0.0012, "step": 340 }, { "epoch": 0.22, "learning_rate": 1.990383623353302e-05, "loss": 0.0014, "step": 341 }, { "epoch": 0.22, "learning_rate": 1.990287375536155e-05, "loss": 0.0014, "step": 342 }, { "epoch": 0.22, "learning_rate": 1.9901906508051073e-05, "loss": 0.0016, "step": 343 }, { "epoch": 0.22, "learning_rate": 1.99009344920674e-05, "loss": 0.0021, "step": 344 }, { "epoch": 0.22, "learning_rate": 1.9899957707878643e-05, "loss": 0.0014, "step": 345 }, { "epoch": 0.22, "learning_rate": 1.989897615595522e-05, "loss": 0.0024, "step": 346 }, { "epoch": 0.22, "learning_rate": 1.9897989836769837e-05, "loss": 0.0011, "step": 347 }, { "epoch": 0.22, "learning_rate": 1.9896998750797493e-05, "loss": 0.0029, "step": 348 }, { "epoch": 0.22, "learning_rate": 1.989600289851549e-05, "loss": 0.0024, "step": 349 }, { "epoch": 0.22, "learning_rate": 1.989500228040342e-05, "loss": 0.0014, "step": 350 }, { "epoch": 0.23, "learning_rate": 1.9893996896943173e-05, "loss": 0.002, "step": 351 }, { "epoch": 0.23, "learning_rate": 1.9892986748618933e-05, "loss": 0.0041, "step": 352 }, { "epoch": 0.23, "learning_rate": 1.9891971835917177e-05, "loss": 0.0013, "step": 353 }, { "epoch": 0.23, "learning_rate": 1.9890952159326682e-05, "loss": 0.0011, "step": 354 }, { "epoch": 0.23, "learning_rate": 1.988992771933851e-05, "loss": 0.001, "step": 355 }, { "epoch": 0.23, "learning_rate": 1.988889851644603e-05, "loss": 0.0016, "step": 356 }, { "epoch": 0.23, "learning_rate": 1.9887864551144892e-05, "loss": 0.0018, "step": 357 }, { "epoch": 0.23, "learning_rate": 1.9886825823933043e-05, "loss": 0.0033, "step": 358 }, { "epoch": 0.23, "learning_rate": 1.9885782335310727e-05, "loss": 0.0008, "step": 359 }, { "epoch": 0.23, "learning_rate": 1.9884734085780483e-05, "loss": 0.0008, "step": 360 }, { "epoch": 0.23, "learning_rate": 1.988368107584714e-05, "loss": 0.0017, "step": 361 }, { "epoch": 0.23, "learning_rate": 1.988262330601781e-05, "loss": 0.0027, "step": 362 }, { "epoch": 0.23, "learning_rate": 1.988156077680191e-05, "loss": 0.0014, "step": 363 }, { "epoch": 0.23, "learning_rate": 1.9880493488711144e-05, "loss": 0.002, "step": 364 }, { "epoch": 0.23, "learning_rate": 1.9879421442259516e-05, "loss": 0.0032, "step": 365 }, { "epoch": 0.24, "learning_rate": 1.9878344637963304e-05, "loss": 0.0011, "step": 366 }, { "epoch": 0.24, "learning_rate": 1.9877263076341093e-05, "loss": 0.0018, "step": 367 }, { "epoch": 0.24, "learning_rate": 1.9876176757913757e-05, "loss": 0.0011, "step": 368 }, { "epoch": 0.24, "learning_rate": 1.987508568320445e-05, "loss": 0.0022, "step": 369 }, { "epoch": 0.24, "learning_rate": 1.987398985273863e-05, "loss": 0.0017, "step": 370 }, { "epoch": 0.24, "learning_rate": 1.9872889267044034e-05, "loss": 0.0015, "step": 371 }, { "epoch": 0.24, "learning_rate": 1.9871783926650702e-05, "loss": 0.002, "step": 372 }, { "epoch": 0.24, "learning_rate": 1.987067383209095e-05, "loss": 0.0011, "step": 373 }, { "epoch": 0.24, "learning_rate": 1.9869558983899396e-05, "loss": 0.0015, "step": 374 }, { "epoch": 0.24, "learning_rate": 1.9868439382612935e-05, "loss": 0.0007, "step": 375 }, { "epoch": 0.24, "learning_rate": 1.9867315028770758e-05, "loss": 0.0012, "step": 376 }, { "epoch": 0.24, "learning_rate": 1.9866185922914352e-05, "loss": 0.0014, "step": 377 }, { "epoch": 0.24, "learning_rate": 1.9865052065587475e-05, "loss": 0.0009, "step": 378 }, { "epoch": 0.24, "learning_rate": 1.986391345733619e-05, "loss": 0.0017, "step": 379 }, { "epoch": 0.24, "learning_rate": 1.9862770098708836e-05, "loss": 0.0012, "step": 380 }, { "epoch": 0.24, "learning_rate": 1.9861621990256048e-05, "loss": 0.0005, "step": 381 }, { "epoch": 0.25, "learning_rate": 1.9860469132530745e-05, "loss": 0.0007, "step": 382 }, { "epoch": 0.25, "learning_rate": 1.985931152608813e-05, "loss": 0.0015, "step": 383 }, { "epoch": 0.25, "learning_rate": 1.98581491714857e-05, "loss": 0.0019, "step": 384 }, { "epoch": 0.25, "learning_rate": 1.9856982069283233e-05, "loss": 0.0005, "step": 385 }, { "epoch": 0.25, "learning_rate": 1.9855810220042796e-05, "loss": 0.0015, "step": 386 }, { "epoch": 0.25, "learning_rate": 1.9854633624328742e-05, "loss": 0.0011, "step": 387 }, { "epoch": 0.25, "learning_rate": 1.985345228270771e-05, "loss": 0.0022, "step": 388 }, { "epoch": 0.25, "learning_rate": 1.9852266195748623e-05, "loss": 0.0022, "step": 389 }, { "epoch": 0.25, "learning_rate": 1.9851075364022687e-05, "loss": 0.0018, "step": 390 }, { "epoch": 0.25, "learning_rate": 1.9849879788103402e-05, "loss": 0.0022, "step": 391 }, { "epoch": 0.25, "learning_rate": 1.9848679468566545e-05, "loss": 0.0008, "step": 392 }, { "epoch": 0.25, "learning_rate": 1.9847474405990178e-05, "loss": 0.0011, "step": 393 }, { "epoch": 0.25, "learning_rate": 1.984626460095465e-05, "loss": 0.0018, "step": 394 }, { "epoch": 0.25, "learning_rate": 1.9845050054042597e-05, "loss": 0.0012, "step": 395 }, { "epoch": 0.25, "learning_rate": 1.9843830765838924e-05, "loss": 0.0027, "step": 396 }, { "epoch": 0.26, "learning_rate": 1.984260673693084e-05, "loss": 0.0073, "step": 397 }, { "epoch": 0.26, "learning_rate": 1.9841377967907825e-05, "loss": 0.0025, "step": 398 }, { "epoch": 0.26, "learning_rate": 1.9840144459361638e-05, "loss": 0.0016, "step": 399 }, { "epoch": 0.26, "learning_rate": 1.983890621188633e-05, "loss": 0.0016, "step": 400 }, { "epoch": 0.26, "learning_rate": 1.983766322607823e-05, "loss": 0.0014, "step": 401 }, { "epoch": 0.26, "learning_rate": 1.983641550253595e-05, "loss": 0.0019, "step": 402 }, { "epoch": 0.26, "learning_rate": 1.9835163041860377e-05, "loss": 0.0014, "step": 403 }, { "epoch": 0.26, "learning_rate": 1.9833905844654696e-05, "loss": 0.0024, "step": 404 }, { "epoch": 0.26, "learning_rate": 1.9832643911524357e-05, "loss": 0.0016, "step": 405 }, { "epoch": 0.26, "learning_rate": 1.9831377243077092e-05, "loss": 0.0007, "step": 406 }, { "epoch": 0.26, "learning_rate": 1.9830105839922924e-05, "loss": 0.0015, "step": 407 }, { "epoch": 0.26, "learning_rate": 1.9828829702674147e-05, "loss": 0.0009, "step": 408 }, { "epoch": 0.26, "learning_rate": 1.982754883194534e-05, "loss": 0.0008, "step": 409 }, { "epoch": 0.26, "learning_rate": 1.9826263228353356e-05, "loss": 0.002, "step": 410 }, { "epoch": 0.26, "learning_rate": 1.9824972892517333e-05, "loss": 0.0013, "step": 411 }, { "epoch": 0.26, "learning_rate": 1.982367782505868e-05, "loss": 0.001, "step": 412 }, { "epoch": 0.27, "learning_rate": 1.9822378026601102e-05, "loss": 0.0006, "step": 413 }, { "epoch": 0.27, "learning_rate": 1.9821073497770565e-05, "loss": 0.0012, "step": 414 }, { "epoch": 0.27, "learning_rate": 1.9819764239195312e-05, "loss": 0.0017, "step": 415 }, { "epoch": 0.27, "learning_rate": 1.9818450251505878e-05, "loss": 0.0022, "step": 416 }, { "epoch": 0.27, "learning_rate": 1.981713153533507e-05, "loss": 0.0018, "step": 417 }, { "epoch": 0.27, "learning_rate": 1.9815808091317964e-05, "loss": 0.0016, "step": 418 }, { "epoch": 0.27, "learning_rate": 1.9814479920091925e-05, "loss": 0.0011, "step": 419 }, { "epoch": 0.27, "learning_rate": 1.9813147022296584e-05, "loss": 0.0011, "step": 420 }, { "epoch": 0.27, "learning_rate": 1.9811809398573857e-05, "loss": 0.0026, "step": 421 }, { "epoch": 0.27, "learning_rate": 1.981046704956793e-05, "loss": 0.0011, "step": 422 }, { "epoch": 0.27, "learning_rate": 1.9809119975925268e-05, "loss": 0.0014, "step": 423 }, { "epoch": 0.27, "learning_rate": 1.980776817829461e-05, "loss": 0.0013, "step": 424 }, { "epoch": 0.27, "learning_rate": 1.980641165732697e-05, "loss": 0.0013, "step": 425 }, { "epoch": 0.27, "learning_rate": 1.9805050413675636e-05, "loss": 0.002, "step": 426 }, { "epoch": 0.27, "learning_rate": 1.980368444799617e-05, "loss": 0.001, "step": 427 }, { "epoch": 0.28, "learning_rate": 1.980231376094641e-05, "loss": 0.0024, "step": 428 }, { "epoch": 0.28, "learning_rate": 1.980093835318647e-05, "loss": 0.0015, "step": 429 }, { "epoch": 0.28, "learning_rate": 1.9799558225378734e-05, "loss": 0.0022, "step": 430 }, { "epoch": 0.28, "learning_rate": 1.979817337818786e-05, "loss": 0.001, "step": 431 }, { "epoch": 0.28, "learning_rate": 1.9796783812280775e-05, "loss": 0.0015, "step": 432 }, { "epoch": 0.28, "learning_rate": 1.9795389528326684e-05, "loss": 0.0011, "step": 433 }, { "epoch": 0.28, "learning_rate": 1.979399052699706e-05, "loss": 0.0015, "step": 434 }, { "epoch": 0.28, "learning_rate": 1.9792586808965653e-05, "loss": 0.0015, "step": 435 }, { "epoch": 0.28, "learning_rate": 1.9791178374908483e-05, "loss": 0.0016, "step": 436 }, { "epoch": 0.28, "learning_rate": 1.9789765225503836e-05, "loss": 0.0009, "step": 437 }, { "epoch": 0.28, "learning_rate": 1.9788347361432276e-05, "loss": 0.0011, "step": 438 }, { "epoch": 0.28, "learning_rate": 1.978692478337663e-05, "loss": 0.0007, "step": 439 }, { "epoch": 0.28, "learning_rate": 1.9785497492022002e-05, "loss": 0.001, "step": 440 }, { "epoch": 0.28, "learning_rate": 1.9784065488055763e-05, "loss": 0.0009, "step": 441 }, { "epoch": 0.28, "learning_rate": 1.9782628772167557e-05, "loss": 0.0014, "step": 442 }, { "epoch": 0.28, "learning_rate": 1.978118734504929e-05, "loss": 0.001, "step": 443 }, { "epoch": 0.29, "learning_rate": 1.9779741207395138e-05, "loss": 0.002, "step": 444 }, { "epoch": 0.29, "learning_rate": 1.9778290359901554e-05, "loss": 0.0014, "step": 445 }, { "epoch": 0.29, "learning_rate": 1.977683480326725e-05, "loss": 0.0014, "step": 446 }, { "epoch": 0.29, "learning_rate": 1.9775374538193218e-05, "loss": 0.0018, "step": 447 }, { "epoch": 0.29, "learning_rate": 1.97739095653827e-05, "loss": 0.0019, "step": 448 }, { "epoch": 0.29, "learning_rate": 1.9772439885541214e-05, "loss": 0.0017, "step": 449 }, { "epoch": 0.29, "learning_rate": 1.977096549937655e-05, "loss": 0.002, "step": 450 }, { "epoch": 0.29, "learning_rate": 1.9769486407598755e-05, "loss": 0.001, "step": 451 }, { "epoch": 0.29, "learning_rate": 1.9768002610920155e-05, "loss": 0.0005, "step": 452 }, { "epoch": 0.29, "learning_rate": 1.9766514110055322e-05, "loss": 0.0012, "step": 453 }, { "epoch": 0.29, "learning_rate": 1.976502090572112e-05, "loss": 0.0013, "step": 454 }, { "epoch": 0.29, "learning_rate": 1.976352299863665e-05, "loss": 0.001, "step": 455 }, { "epoch": 0.29, "learning_rate": 1.9762020389523296e-05, "loss": 0.0011, "step": 456 }, { "epoch": 0.29, "learning_rate": 1.9760513079104706e-05, "loss": 0.0011, "step": 457 }, { "epoch": 0.29, "learning_rate": 1.9759001068106783e-05, "loss": 0.0018, "step": 458 }, { "epoch": 0.29, "learning_rate": 1.9757484357257695e-05, "loss": 0.0011, "step": 459 }, { "epoch": 0.3, "learning_rate": 1.9755962947287887e-05, "loss": 0.0009, "step": 460 }, { "epoch": 0.3, "learning_rate": 1.9754436838930047e-05, "loss": 0.0028, "step": 461 }, { "epoch": 0.3, "learning_rate": 1.975290603291914e-05, "loss": 0.0015, "step": 462 }, { "epoch": 0.3, "learning_rate": 1.9751370529992394e-05, "loss": 0.0008, "step": 463 }, { "epoch": 0.3, "learning_rate": 1.9749830330889285e-05, "loss": 0.0011, "step": 464 }, { "epoch": 0.3, "learning_rate": 1.9748285436351565e-05, "loss": 0.0014, "step": 465 }, { "epoch": 0.3, "learning_rate": 1.974673584712324e-05, "loss": 0.0012, "step": 466 }, { "epoch": 0.3, "learning_rate": 1.974518156395058e-05, "loss": 0.0009, "step": 467 }, { "epoch": 0.3, "learning_rate": 1.9743622587582112e-05, "loss": 0.0013, "step": 468 }, { "epoch": 0.3, "learning_rate": 1.974205891876863e-05, "loss": 0.0006, "step": 469 }, { "epoch": 0.3, "learning_rate": 1.9740490558263177e-05, "loss": 0.0008, "step": 470 }, { "epoch": 0.3, "learning_rate": 1.9738917506821068e-05, "loss": 0.0023, "step": 471 }, { "epoch": 0.3, "learning_rate": 1.9737339765199873e-05, "loss": 0.0012, "step": 472 }, { "epoch": 0.3, "learning_rate": 1.9735757334159415e-05, "loss": 0.0017, "step": 473 }, { "epoch": 0.3, "learning_rate": 1.9734170214461776e-05, "loss": 0.0042, "step": 474 }, { "epoch": 0.31, "learning_rate": 1.9732578406871302e-05, "loss": 0.0021, "step": 475 }, { "epoch": 0.31, "learning_rate": 1.973098191215459e-05, "loss": 0.0021, "step": 476 }, { "epoch": 0.31, "learning_rate": 1.972938073108051e-05, "loss": 0.0013, "step": 477 }, { "epoch": 0.31, "learning_rate": 1.9727774864420165e-05, "loss": 0.0008, "step": 478 }, { "epoch": 0.31, "learning_rate": 1.9726164312946933e-05, "loss": 0.0013, "step": 479 }, { "epoch": 0.31, "learning_rate": 1.972454907743644e-05, "loss": 0.0029, "step": 480 }, { "epoch": 0.31, "learning_rate": 1.972292915866657e-05, "loss": 0.0012, "step": 481 }, { "epoch": 0.31, "learning_rate": 1.972130455741746e-05, "loss": 0.0022, "step": 482 }, { "epoch": 0.31, "learning_rate": 1.9719675274471504e-05, "loss": 0.0015, "step": 483 }, { "epoch": 0.31, "learning_rate": 1.9718041310613357e-05, "loss": 0.0013, "step": 484 }, { "epoch": 0.31, "learning_rate": 1.9716402666629914e-05, "loss": 0.0012, "step": 485 }, { "epoch": 0.31, "learning_rate": 1.9714759343310335e-05, "loss": 0.0021, "step": 486 }, { "epoch": 0.31, "learning_rate": 1.9713111341446028e-05, "loss": 0.0016, "step": 487 }, { "epoch": 0.31, "learning_rate": 1.971145866183066e-05, "loss": 0.0013, "step": 488 }, { "epoch": 0.31, "learning_rate": 1.970980130526015e-05, "loss": 0.0012, "step": 489 }, { "epoch": 0.31, "learning_rate": 1.970813927253266e-05, "loss": 0.0012, "step": 490 }, { "epoch": 0.32, "learning_rate": 1.9706472564448612e-05, "loss": 0.0006, "step": 491 }, { "epoch": 0.32, "learning_rate": 1.9704801181810683e-05, "loss": 0.0006, "step": 492 }, { "epoch": 0.32, "learning_rate": 1.9703125125423793e-05, "loss": 0.0019, "step": 493 }, { "epoch": 0.32, "learning_rate": 1.9701444396095115e-05, "loss": 0.0009, "step": 494 }, { "epoch": 0.32, "learning_rate": 1.969975899463408e-05, "loss": 0.0018, "step": 495 }, { "epoch": 0.32, "learning_rate": 1.9698068921852353e-05, "loss": 0.0024, "step": 496 }, { "epoch": 0.32, "learning_rate": 1.9696374178563865e-05, "loss": 0.0018, "step": 497 }, { "epoch": 0.32, "learning_rate": 1.969467476558479e-05, "loss": 0.001, "step": 498 }, { "epoch": 0.32, "learning_rate": 1.9692970683733547e-05, "loss": 0.0008, "step": 499 }, { "epoch": 0.32, "learning_rate": 1.9691261933830815e-05, "loss": 0.0005, "step": 500 }, { "epoch": 0.32, "learning_rate": 1.9689548516699503e-05, "loss": 0.0004, "step": 501 }, { "epoch": 0.32, "learning_rate": 1.968783043316479e-05, "loss": 0.0007, "step": 502 }, { "epoch": 0.32, "learning_rate": 1.9686107684054074e-05, "loss": 0.0013, "step": 503 }, { "epoch": 0.32, "learning_rate": 1.968438027019703e-05, "loss": 0.0008, "step": 504 }, { "epoch": 0.32, "learning_rate": 1.9682648192425565e-05, "loss": 0.002, "step": 505 }, { "epoch": 0.33, "learning_rate": 1.9680911451573827e-05, "loss": 0.0028, "step": 506 }, { "epoch": 0.33, "learning_rate": 1.9679170048478214e-05, "loss": 0.0011, "step": 507 }, { "epoch": 0.33, "learning_rate": 1.9677423983977374e-05, "loss": 0.0023, "step": 508 }, { "epoch": 0.33, "learning_rate": 1.9675673258912203e-05, "loss": 0.0004, "step": 509 }, { "epoch": 0.33, "learning_rate": 1.9673917874125823e-05, "loss": 0.0022, "step": 510 }, { "epoch": 0.33, "learning_rate": 1.967215783046362e-05, "loss": 0.0012, "step": 511 }, { "epoch": 0.33, "learning_rate": 1.9670393128773218e-05, "loss": 0.0014, "step": 512 }, { "epoch": 0.33, "learning_rate": 1.9668623769904475e-05, "loss": 0.001, "step": 513 }, { "epoch": 0.33, "learning_rate": 1.9666849754709504e-05, "loss": 0.0027, "step": 514 }, { "epoch": 0.33, "learning_rate": 1.9665071084042653e-05, "loss": 0.0014, "step": 515 }, { "epoch": 0.33, "learning_rate": 1.9663287758760513e-05, "loss": 0.0009, "step": 516 }, { "epoch": 0.33, "learning_rate": 1.9661499779721924e-05, "loss": 0.0014, "step": 517 }, { "epoch": 0.33, "learning_rate": 1.9659707147787957e-05, "loss": 0.0004, "step": 518 }, { "epoch": 0.33, "learning_rate": 1.965790986382193e-05, "loss": 0.0014, "step": 519 }, { "epoch": 0.33, "learning_rate": 1.96561079286894e-05, "loss": 0.0004, "step": 520 }, { "epoch": 0.33, "learning_rate": 1.965430134325816e-05, "loss": 0.0012, "step": 521 }, { "epoch": 0.34, "learning_rate": 1.9652490108398247e-05, "loss": 0.0016, "step": 522 }, { "epoch": 0.34, "learning_rate": 1.9650674224981943e-05, "loss": 0.0009, "step": 523 }, { "epoch": 0.34, "learning_rate": 1.964885369388375e-05, "loss": 0.0012, "step": 524 }, { "epoch": 0.34, "learning_rate": 1.964702851598043e-05, "loss": 0.001, "step": 525 }, { "epoch": 0.34, "learning_rate": 1.964519869215097e-05, "loss": 0.001, "step": 526 }, { "epoch": 0.34, "learning_rate": 1.9643364223276595e-05, "loss": 0.0013, "step": 527 }, { "epoch": 0.34, "learning_rate": 1.9641525110240772e-05, "loss": 0.0012, "step": 528 }, { "epoch": 0.34, "learning_rate": 1.96396813539292e-05, "loss": 0.001, "step": 529 }, { "epoch": 0.34, "learning_rate": 1.963783295522982e-05, "loss": 0.0008, "step": 530 }, { "epoch": 0.34, "learning_rate": 1.9635979915032802e-05, "loss": 0.0007, "step": 531 }, { "epoch": 0.34, "learning_rate": 1.963412223423056e-05, "loss": 0.0016, "step": 532 }, { "epoch": 0.34, "learning_rate": 1.9632259913717723e-05, "loss": 0.0009, "step": 533 }, { "epoch": 0.34, "learning_rate": 1.963039295439118e-05, "loss": 0.001, "step": 534 }, { "epoch": 0.34, "learning_rate": 1.9628521357150044e-05, "loss": 0.0009, "step": 535 }, { "epoch": 0.34, "learning_rate": 1.9626645122895654e-05, "loss": 0.0005, "step": 536 }, { "epoch": 0.35, "learning_rate": 1.962476425253159e-05, "loss": 0.0009, "step": 537 }, { "epoch": 0.35, "learning_rate": 1.9622878746963666e-05, "loss": 0.0007, "step": 538 }, { "epoch": 0.35, "learning_rate": 1.962098860709992e-05, "loss": 0.0009, "step": 539 }, { "epoch": 0.35, "learning_rate": 1.961909383385063e-05, "loss": 0.0007, "step": 540 }, { "epoch": 0.35, "learning_rate": 1.9617194428128307e-05, "loss": 0.0009, "step": 541 }, { "epoch": 0.35, "learning_rate": 1.9615290390847678e-05, "loss": 0.0008, "step": 542 }, { "epoch": 0.35, "learning_rate": 1.961338172292572e-05, "loss": 0.0006, "step": 543 }, { "epoch": 0.35, "learning_rate": 1.961146842528163e-05, "loss": 0.0013, "step": 544 }, { "epoch": 0.35, "learning_rate": 1.9609550498836835e-05, "loss": 0.0012, "step": 545 }, { "epoch": 0.35, "learning_rate": 1.9607627944514983e-05, "loss": 0.0015, "step": 546 }, { "epoch": 0.35, "learning_rate": 1.9605700763241976e-05, "loss": 0.001, "step": 547 }, { "epoch": 0.35, "learning_rate": 1.9603768955945917e-05, "loss": 0.0008, "step": 548 }, { "epoch": 0.35, "learning_rate": 1.960183252355715e-05, "loss": 0.0014, "step": 549 }, { "epoch": 0.35, "learning_rate": 1.959989146700824e-05, "loss": 0.0007, "step": 550 }, { "epoch": 0.35, "learning_rate": 1.9597945787233997e-05, "loss": 0.0013, "step": 551 }, { "epoch": 0.35, "learning_rate": 1.959599548517143e-05, "loss": 0.001, "step": 552 }, { "epoch": 0.36, "learning_rate": 1.959404056175979e-05, "loss": 0.0014, "step": 553 }, { "epoch": 0.36, "learning_rate": 1.9592081017940557e-05, "loss": 0.0008, "step": 554 }, { "epoch": 0.36, "learning_rate": 1.9590116854657424e-05, "loss": 0.0007, "step": 555 }, { "epoch": 0.36, "learning_rate": 1.958814807285632e-05, "loss": 0.0005, "step": 556 }, { "epoch": 0.36, "learning_rate": 1.958617467348539e-05, "loss": 0.0006, "step": 557 }, { "epoch": 0.36, "learning_rate": 1.958419665749501e-05, "loss": 0.0011, "step": 558 }, { "epoch": 0.36, "learning_rate": 1.958221402583777e-05, "loss": 0.0009, "step": 559 }, { "epoch": 0.36, "learning_rate": 1.9580226779468497e-05, "loss": 0.0012, "step": 560 }, { "epoch": 0.36, "learning_rate": 1.9578234919344222e-05, "loss": 0.0011, "step": 561 }, { "epoch": 0.36, "learning_rate": 1.9576238446424213e-05, "loss": 0.001, "step": 562 }, { "epoch": 0.36, "learning_rate": 1.9574237361669954e-05, "loss": 0.0006, "step": 563 }, { "epoch": 0.36, "learning_rate": 1.957223166604515e-05, "loss": 0.0009, "step": 564 }, { "epoch": 0.36, "learning_rate": 1.9570221360515727e-05, "loss": 0.0014, "step": 565 }, { "epoch": 0.36, "learning_rate": 1.956820644604983e-05, "loss": 0.0008, "step": 566 }, { "epoch": 0.36, "learning_rate": 1.956618692361782e-05, "loss": 0.0016, "step": 567 }, { "epoch": 0.36, "learning_rate": 1.9564162794192295e-05, "loss": 0.001, "step": 568 }, { "epoch": 0.37, "learning_rate": 1.9562134058748047e-05, "loss": 0.0014, "step": 569 }, { "epoch": 0.37, "learning_rate": 1.9560100718262095e-05, "loss": 0.0011, "step": 570 }, { "epoch": 0.37, "learning_rate": 1.955806277371369e-05, "loss": 0.0011, "step": 571 }, { "epoch": 0.37, "learning_rate": 1.955602022608428e-05, "loss": 0.0007, "step": 572 }, { "epoch": 0.37, "learning_rate": 1.9553973076357546e-05, "loss": 0.0007, "step": 573 }, { "epoch": 0.37, "learning_rate": 1.9551921325519365e-05, "loss": 0.001, "step": 574 }, { "epoch": 0.37, "learning_rate": 1.9549864974557853e-05, "loss": 0.0007, "step": 575 }, { "epoch": 0.37, "learning_rate": 1.954780402446333e-05, "loss": 0.0015, "step": 576 }, { "epoch": 0.37, "learning_rate": 1.954573847622833e-05, "loss": 0.0006, "step": 577 }, { "epoch": 0.37, "learning_rate": 1.95436683308476e-05, "loss": 0.0005, "step": 578 }, { "epoch": 0.37, "learning_rate": 1.9541593589318108e-05, "loss": 0.0016, "step": 579 }, { "epoch": 0.37, "learning_rate": 1.9539514252639035e-05, "loss": 0.0015, "step": 580 }, { "epoch": 0.37, "learning_rate": 1.9537430321811763e-05, "loss": 0.0007, "step": 581 }, { "epoch": 0.37, "learning_rate": 1.9535341797839904e-05, "loss": 0.0012, "step": 582 }, { "epoch": 0.37, "learning_rate": 1.953324868172927e-05, "loss": 0.0022, "step": 583 }, { "epoch": 0.38, "learning_rate": 1.9531150974487886e-05, "loss": 0.0008, "step": 584 }, { "epoch": 0.38, "learning_rate": 1.952904867712599e-05, "loss": 0.0008, "step": 585 }, { "epoch": 0.38, "learning_rate": 1.952694179065603e-05, "loss": 0.0005, "step": 586 }, { "epoch": 0.38, "learning_rate": 1.9524830316092667e-05, "loss": 0.0007, "step": 587 }, { "epoch": 0.38, "learning_rate": 1.952271425445277e-05, "loss": 0.001, "step": 588 }, { "epoch": 0.38, "learning_rate": 1.9520593606755414e-05, "loss": 0.0007, "step": 589 }, { "epoch": 0.38, "learning_rate": 1.951846837402188e-05, "loss": 0.0008, "step": 590 }, { "epoch": 0.38, "learning_rate": 1.951633855727567e-05, "loss": 0.001, "step": 591 }, { "epoch": 0.38, "learning_rate": 1.951420415754248e-05, "loss": 0.0013, "step": 592 }, { "epoch": 0.38, "learning_rate": 1.9512065175850227e-05, "loss": 0.0007, "step": 593 }, { "epoch": 0.38, "learning_rate": 1.9509921613229013e-05, "loss": 0.0006, "step": 594 }, { "epoch": 0.38, "learning_rate": 1.950777347071117e-05, "loss": 0.0011, "step": 595 }, { "epoch": 0.38, "learning_rate": 1.9505620749331217e-05, "loss": 0.0012, "step": 596 }, { "epoch": 0.38, "learning_rate": 1.9503463450125894e-05, "loss": 0.0007, "step": 597 }, { "epoch": 0.38, "learning_rate": 1.9501301574134133e-05, "loss": 0.0006, "step": 598 }, { "epoch": 0.38, "learning_rate": 1.9499135122397078e-05, "loss": 0.0008, "step": 599 }, { "epoch": 0.39, "learning_rate": 1.9496964095958067e-05, "loss": 0.001, "step": 600 }, { "epoch": 0.39, "learning_rate": 1.9494788495862654e-05, "loss": 0.0011, "step": 601 }, { "epoch": 0.39, "learning_rate": 1.949260832315859e-05, "loss": 0.0006, "step": 602 }, { "epoch": 0.39, "learning_rate": 1.9490423578895827e-05, "loss": 0.0019, "step": 603 }, { "epoch": 0.39, "learning_rate": 1.948823426412651e-05, "loss": 0.0004, "step": 604 }, { "epoch": 0.39, "learning_rate": 1.9486040379905006e-05, "loss": 0.0014, "step": 605 }, { "epoch": 0.39, "learning_rate": 1.9483841927287862e-05, "loss": 0.001, "step": 606 }, { "epoch": 0.39, "learning_rate": 1.9481638907333843e-05, "loss": 0.0004, "step": 607 }, { "epoch": 0.39, "learning_rate": 1.94794313211039e-05, "loss": 0.0009, "step": 608 }, { "epoch": 0.39, "learning_rate": 1.9477219169661183e-05, "loss": 0.001, "step": 609 }, { "epoch": 0.39, "learning_rate": 1.9475002454071054e-05, "loss": 0.0009, "step": 610 }, { "epoch": 0.39, "learning_rate": 1.9472781175401058e-05, "loss": 0.0015, "step": 611 }, { "epoch": 0.39, "learning_rate": 1.9470555334720942e-05, "loss": 0.0011, "step": 612 }, { "epoch": 0.39, "learning_rate": 1.946832493310266e-05, "loss": 0.0007, "step": 613 }, { "epoch": 0.39, "learning_rate": 1.9466089971620343e-05, "loss": 0.0008, "step": 614 }, { "epoch": 0.4, "learning_rate": 1.946385045135034e-05, "loss": 0.001, "step": 615 }, { "epoch": 0.4, "learning_rate": 1.946160637337118e-05, "loss": 0.0008, "step": 616 }, { "epoch": 0.4, "learning_rate": 1.945935773876359e-05, "loss": 0.0005, "step": 617 }, { "epoch": 0.4, "learning_rate": 1.94571045486105e-05, "loss": 0.0008, "step": 618 }, { "epoch": 0.4, "learning_rate": 1.9454846803997017e-05, "loss": 0.001, "step": 619 }, { "epoch": 0.4, "learning_rate": 1.945258450601046e-05, "loss": 0.001, "step": 620 }, { "epoch": 0.4, "learning_rate": 1.9450317655740326e-05, "loss": 0.001, "step": 621 }, { "epoch": 0.4, "learning_rate": 1.9448046254278315e-05, "loss": 0.0011, "step": 622 }, { "epoch": 0.4, "learning_rate": 1.9445770302718317e-05, "loss": 0.0005, "step": 623 }, { "epoch": 0.4, "learning_rate": 1.94434898021564e-05, "loss": 0.0018, "step": 624 }, { "epoch": 0.4, "learning_rate": 1.9441204753690845e-05, "loss": 0.0006, "step": 625 }, { "epoch": 0.4, "learning_rate": 1.9438915158422106e-05, "loss": 0.0007, "step": 626 }, { "epoch": 0.4, "learning_rate": 1.9436621017452833e-05, "loss": 0.0006, "step": 627 }, { "epoch": 0.4, "learning_rate": 1.9434322331887863e-05, "loss": 0.001, "step": 628 }, { "epoch": 0.4, "learning_rate": 1.9432019102834226e-05, "loss": 0.0007, "step": 629 }, { "epoch": 0.4, "learning_rate": 1.9429711331401137e-05, "loss": 0.0008, "step": 630 }, { "epoch": 0.41, "learning_rate": 1.94273990187e-05, "loss": 0.0063, "step": 631 }, { "epoch": 0.41, "learning_rate": 1.94250821658444e-05, "loss": 0.0005, "step": 632 }, { "epoch": 0.41, "learning_rate": 1.942276077395012e-05, "loss": 0.0011, "step": 633 }, { "epoch": 0.41, "learning_rate": 1.942043484413512e-05, "loss": 0.001, "step": 634 }, { "epoch": 0.41, "learning_rate": 1.9418104377519546e-05, "loss": 0.0003, "step": 635 }, { "epoch": 0.41, "learning_rate": 1.9415769375225733e-05, "loss": 0.0011, "step": 636 }, { "epoch": 0.41, "learning_rate": 1.9413429838378194e-05, "loss": 0.0006, "step": 637 }, { "epoch": 0.41, "learning_rate": 1.9411085768103636e-05, "loss": 0.0005, "step": 638 }, { "epoch": 0.41, "learning_rate": 1.9408737165530937e-05, "loss": 0.0014, "step": 639 }, { "epoch": 0.41, "learning_rate": 1.940638403179117e-05, "loss": 0.0007, "step": 640 }, { "epoch": 0.41, "learning_rate": 1.9404026368017578e-05, "loss": 0.0008, "step": 641 }, { "epoch": 0.41, "learning_rate": 1.9401664175345597e-05, "loss": 0.0008, "step": 642 }, { "epoch": 0.41, "learning_rate": 1.9399297454912835e-05, "loss": 0.001, "step": 643 }, { "epoch": 0.41, "learning_rate": 1.9396926207859085e-05, "loss": 0.0009, "step": 644 }, { "epoch": 0.41, "learning_rate": 1.939455043532632e-05, "loss": 0.0007, "step": 645 }, { "epoch": 0.42, "learning_rate": 1.9392170138458694e-05, "loss": 0.0007, "step": 646 }, { "epoch": 0.42, "learning_rate": 1.938978531840253e-05, "loss": 0.0006, "step": 647 }, { "epoch": 0.42, "learning_rate": 1.9387395976306346e-05, "loss": 0.0011, "step": 648 }, { "epoch": 0.42, "learning_rate": 1.938500211332082e-05, "loss": 0.0011, "step": 649 }, { "epoch": 0.42, "learning_rate": 1.9382603730598824e-05, "loss": 0.0003, "step": 650 }, { "epoch": 0.42, "learning_rate": 1.9380200829295394e-05, "loss": 0.0009, "step": 651 }, { "epoch": 0.42, "learning_rate": 1.9377793410567746e-05, "loss": 0.0006, "step": 652 }, { "epoch": 0.42, "learning_rate": 1.9375381475575273e-05, "loss": 0.001, "step": 653 }, { "epoch": 0.42, "learning_rate": 1.9372965025479544e-05, "loss": 0.0007, "step": 654 }, { "epoch": 0.42, "learning_rate": 1.93705440614443e-05, "loss": 0.0012, "step": 655 }, { "epoch": 0.42, "learning_rate": 1.9368118584635453e-05, "loss": 0.0079, "step": 656 }, { "epoch": 0.42, "learning_rate": 1.9365688596221098e-05, "loss": 0.0006, "step": 657 }, { "epoch": 0.42, "learning_rate": 1.9363254097371495e-05, "loss": 0.0009, "step": 658 }, { "epoch": 0.42, "learning_rate": 1.9360815089259072e-05, "loss": 0.0005, "step": 659 }, { "epoch": 0.42, "learning_rate": 1.935837157305844e-05, "loss": 0.0006, "step": 660 }, { "epoch": 0.42, "learning_rate": 1.9355923549946375e-05, "loss": 0.0004, "step": 661 }, { "epoch": 0.43, "learning_rate": 1.935347102110182e-05, "loss": 0.0011, "step": 662 }, { "epoch": 0.43, "learning_rate": 1.93510139877059e-05, "loss": 0.0006, "step": 663 }, { "epoch": 0.43, "learning_rate": 1.9348552450941888e-05, "loss": 0.0003, "step": 664 }, { "epoch": 0.43, "learning_rate": 1.9346086411995253e-05, "loss": 0.001, "step": 665 }, { "epoch": 0.43, "learning_rate": 1.934361587205361e-05, "loss": 0.0014, "step": 666 }, { "epoch": 0.43, "learning_rate": 1.9341140832306754e-05, "loss": 0.0006, "step": 667 }, { "epoch": 0.43, "learning_rate": 1.9338661293946638e-05, "loss": 0.0005, "step": 668 }, { "epoch": 0.43, "learning_rate": 1.933617725816739e-05, "loss": 0.0018, "step": 669 }, { "epoch": 0.43, "learning_rate": 1.9333688726165298e-05, "loss": 0.0012, "step": 670 }, { "epoch": 0.43, "learning_rate": 1.933119569913882e-05, "loss": 0.0007, "step": 671 }, { "epoch": 0.43, "learning_rate": 1.9328698178288578e-05, "loss": 0.0015, "step": 672 }, { "epoch": 0.43, "learning_rate": 1.9326196164817352e-05, "loss": 0.0007, "step": 673 }, { "epoch": 0.43, "learning_rate": 1.932368965993009e-05, "loss": 0.001, "step": 674 }, { "epoch": 0.43, "learning_rate": 1.932117866483391e-05, "loss": 0.0019, "step": 675 }, { "epoch": 0.43, "learning_rate": 1.931866318073808e-05, "loss": 0.0005, "step": 676 }, { "epoch": 0.44, "learning_rate": 1.9316143208854036e-05, "loss": 0.0004, "step": 677 }, { "epoch": 0.44, "learning_rate": 1.931361875039538e-05, "loss": 0.0005, "step": 678 }, { "epoch": 0.44, "learning_rate": 1.931108980657786e-05, "loss": 0.0006, "step": 679 }, { "epoch": 0.44, "learning_rate": 1.9308556378619397e-05, "loss": 0.0008, "step": 680 }, { "epoch": 0.44, "learning_rate": 1.9306018467740074e-05, "loss": 0.0006, "step": 681 }, { "epoch": 0.44, "learning_rate": 1.9303476075162123e-05, "loss": 0.0008, "step": 682 }, { "epoch": 0.44, "learning_rate": 1.9300929202109935e-05, "loss": 0.0006, "step": 683 }, { "epoch": 0.44, "learning_rate": 1.9298377849810072e-05, "loss": 0.0004, "step": 684 }, { "epoch": 0.44, "learning_rate": 1.9295822019491234e-05, "loss": 0.0011, "step": 685 }, { "epoch": 0.44, "learning_rate": 1.9293261712384287e-05, "loss": 0.0009, "step": 686 }, { "epoch": 0.44, "learning_rate": 1.9290696929722256e-05, "loss": 0.0007, "step": 687 }, { "epoch": 0.44, "learning_rate": 1.9288127672740323e-05, "loss": 0.0011, "step": 688 }, { "epoch": 0.44, "learning_rate": 1.9285553942675813e-05, "loss": 0.0011, "step": 689 }, { "epoch": 0.44, "learning_rate": 1.9282975740768215e-05, "loss": 0.0009, "step": 690 }, { "epoch": 0.44, "learning_rate": 1.928039306825917e-05, "loss": 0.0014, "step": 691 }, { "epoch": 0.44, "learning_rate": 1.9277805926392468e-05, "loss": 0.0014, "step": 692 }, { "epoch": 0.45, "learning_rate": 1.9275214316414056e-05, "loss": 0.0011, "step": 693 }, { "epoch": 0.45, "learning_rate": 1.9272618239572032e-05, "loss": 0.0048, "step": 694 }, { "epoch": 0.45, "learning_rate": 1.927001769711664e-05, "loss": 0.0009, "step": 695 }, { "epoch": 0.45, "learning_rate": 1.926741269030029e-05, "loss": 0.0016, "step": 696 }, { "epoch": 0.45, "learning_rate": 1.9264803220377516e-05, "loss": 0.001, "step": 697 }, { "epoch": 0.45, "learning_rate": 1.9262189288605023e-05, "loss": 0.001, "step": 698 }, { "epoch": 0.45, "learning_rate": 1.9259570896241664e-05, "loss": 0.001, "step": 699 }, { "epoch": 0.45, "learning_rate": 1.9256948044548424e-05, "loss": 0.0008, "step": 700 }, { "epoch": 0.45, "learning_rate": 1.9254320734788455e-05, "loss": 0.0006, "step": 701 }, { "epoch": 0.45, "learning_rate": 1.925168896822704e-05, "loss": 0.0006, "step": 702 }, { "epoch": 0.45, "learning_rate": 1.9249052746131618e-05, "loss": 0.0007, "step": 703 }, { "epoch": 0.45, "learning_rate": 1.9246412069771773e-05, "loss": 0.0007, "step": 704 }, { "epoch": 0.45, "learning_rate": 1.9243766940419225e-05, "loss": 0.0012, "step": 705 }, { "epoch": 0.45, "learning_rate": 1.9241117359347854e-05, "loss": 0.0005, "step": 706 }, { "epoch": 0.45, "learning_rate": 1.923846332783367e-05, "loss": 0.0004, "step": 707 }, { "epoch": 0.45, "learning_rate": 1.9235804847154836e-05, "loss": 0.0007, "step": 708 }, { "epoch": 0.46, "learning_rate": 1.9233141918591652e-05, "loss": 0.0023, "step": 709 }, { "epoch": 0.46, "learning_rate": 1.9230474543426554e-05, "loss": 0.0009, "step": 710 }, { "epoch": 0.46, "learning_rate": 1.9227802722944143e-05, "loss": 0.0008, "step": 711 }, { "epoch": 0.46, "learning_rate": 1.9225126458431132e-05, "loss": 0.0007, "step": 712 }, { "epoch": 0.46, "learning_rate": 1.9222445751176387e-05, "loss": 0.0015, "step": 713 }, { "epoch": 0.46, "learning_rate": 1.921976060247092e-05, "loss": 0.0006, "step": 714 }, { "epoch": 0.46, "learning_rate": 1.9217071013607872e-05, "loss": 0.0014, "step": 715 }, { "epoch": 0.46, "learning_rate": 1.921437698588253e-05, "loss": 0.0019, "step": 716 }, { "epoch": 0.46, "learning_rate": 1.921167852059231e-05, "loss": 0.0009, "step": 717 }, { "epoch": 0.46, "learning_rate": 1.9208975619036767e-05, "loss": 0.0007, "step": 718 }, { "epoch": 0.46, "learning_rate": 1.9206268282517603e-05, "loss": 0.0009, "step": 719 }, { "epoch": 0.46, "learning_rate": 1.9203556512338642e-05, "loss": 0.0008, "step": 720 }, { "epoch": 0.46, "learning_rate": 1.920084030980585e-05, "loss": 0.0006, "step": 721 }, { "epoch": 0.46, "learning_rate": 1.919811967622733e-05, "loss": 0.0015, "step": 722 }, { "epoch": 0.46, "learning_rate": 1.9195394612913315e-05, "loss": 0.0008, "step": 723 }, { "epoch": 0.47, "learning_rate": 1.9192665121176172e-05, "loss": 0.0008, "step": 724 }, { "epoch": 0.47, "learning_rate": 1.91899312023304e-05, "loss": 0.0009, "step": 725 }, { "epoch": 0.47, "learning_rate": 1.918719285769263e-05, "loss": 0.0008, "step": 726 }, { "epoch": 0.47, "learning_rate": 1.9184450088581624e-05, "loss": 0.0006, "step": 727 }, { "epoch": 0.47, "learning_rate": 1.9181702896318283e-05, "loss": 0.0006, "step": 728 }, { "epoch": 0.47, "learning_rate": 1.917895128222562e-05, "loss": 0.0004, "step": 729 }, { "epoch": 0.47, "learning_rate": 1.9176195247628802e-05, "loss": 0.0007, "step": 730 }, { "epoch": 0.47, "learning_rate": 1.91734347938551e-05, "loss": 0.0008, "step": 731 }, { "epoch": 0.47, "learning_rate": 1.917066992223393e-05, "loss": 0.001, "step": 732 }, { "epoch": 0.47, "learning_rate": 1.9167900634096833e-05, "loss": 0.0009, "step": 733 }, { "epoch": 0.47, "learning_rate": 1.916512693077747e-05, "loss": 0.0011, "step": 734 }, { "epoch": 0.47, "learning_rate": 1.9162348813611632e-05, "loss": 0.0007, "step": 735 }, { "epoch": 0.47, "learning_rate": 1.915956628393724e-05, "loss": 0.0008, "step": 736 }, { "epoch": 0.47, "learning_rate": 1.9156779343094336e-05, "loss": 0.001, "step": 737 }, { "epoch": 0.47, "learning_rate": 1.9153987992425086e-05, "loss": 0.0012, "step": 738 }, { "epoch": 0.47, "learning_rate": 1.9151192233273777e-05, "loss": 0.0005, "step": 739 }, { "epoch": 0.48, "learning_rate": 1.9148392066986827e-05, "loss": 0.0007, "step": 740 }, { "epoch": 0.48, "learning_rate": 1.9145587494912772e-05, "loss": 0.0018, "step": 741 }, { "epoch": 0.48, "learning_rate": 1.914277851840227e-05, "loss": 0.001, "step": 742 }, { "epoch": 0.48, "learning_rate": 1.9139965138808097e-05, "loss": 0.0011, "step": 743 }, { "epoch": 0.48, "learning_rate": 1.913714735748515e-05, "loss": 0.0006, "step": 744 }, { "epoch": 0.48, "learning_rate": 1.913432517579046e-05, "loss": 0.0008, "step": 745 }, { "epoch": 0.48, "learning_rate": 1.9131498595083156e-05, "loss": 0.0023, "step": 746 }, { "epoch": 0.48, "learning_rate": 1.9128667616724494e-05, "loss": 0.0006, "step": 747 }, { "epoch": 0.48, "learning_rate": 1.9125832242077856e-05, "loss": 0.0009, "step": 748 }, { "epoch": 0.48, "learning_rate": 1.912299247250873e-05, "loss": 0.0006, "step": 749 }, { "epoch": 0.48, "learning_rate": 1.912014830938472e-05, "loss": 0.0004, "step": 750 }, { "epoch": 0.48, "learning_rate": 1.911729975407556e-05, "loss": 0.0005, "step": 751 }, { "epoch": 0.48, "learning_rate": 1.9114446807953084e-05, "loss": 0.0005, "step": 752 }, { "epoch": 0.48, "learning_rate": 1.9111589472391242e-05, "loss": 0.0005, "step": 753 }, { "epoch": 0.48, "learning_rate": 1.9108727748766113e-05, "loss": 0.0004, "step": 754 }, { "epoch": 0.49, "learning_rate": 1.910586163845587e-05, "loss": 0.0004, "step": 755 }, { "epoch": 0.49, "learning_rate": 1.9102991142840812e-05, "loss": 0.001, "step": 756 }, { "epoch": 0.49, "learning_rate": 1.9100116263303347e-05, "loss": 0.0009, "step": 757 }, { "epoch": 0.49, "learning_rate": 1.9097237001227983e-05, "loss": 0.0004, "step": 758 }, { "epoch": 0.49, "learning_rate": 1.9094353358001358e-05, "loss": 0.0004, "step": 759 }, { "epoch": 0.49, "learning_rate": 1.90914653350122e-05, "loss": 0.0013, "step": 760 }, { "epoch": 0.49, "learning_rate": 1.9088572933651365e-05, "loss": 0.0011, "step": 761 }, { "epoch": 0.49, "learning_rate": 1.9085676155311806e-05, "loss": 0.0003, "step": 762 }, { "epoch": 0.49, "learning_rate": 1.9082775001388586e-05, "loss": 0.0006, "step": 763 }, { "epoch": 0.49, "learning_rate": 1.9079869473278872e-05, "loss": 0.0005, "step": 764 }, { "epoch": 0.49, "learning_rate": 1.9076959572381947e-05, "loss": 0.0007, "step": 765 }, { "epoch": 0.49, "learning_rate": 1.9074045300099193e-05, "loss": 0.0003, "step": 766 }, { "epoch": 0.49, "learning_rate": 1.9071126657834097e-05, "loss": 0.001, "step": 767 }, { "epoch": 0.49, "learning_rate": 1.9068203646992252e-05, "loss": 0.0008, "step": 768 }, { "epoch": 0.49, "learning_rate": 1.906527626898136e-05, "loss": 0.0006, "step": 769 }, { "epoch": 0.49, "learning_rate": 1.906234452521121e-05, "loss": 0.0005, "step": 770 }, { "epoch": 0.5, "learning_rate": 1.905940841709372e-05, "loss": 0.0005, "step": 771 }, { "epoch": 0.5, "learning_rate": 1.905646794604288e-05, "loss": 0.0005, "step": 772 }, { "epoch": 0.5, "learning_rate": 1.90535231134748e-05, "loss": 0.0019, "step": 773 }, { "epoch": 0.5, "learning_rate": 1.9050573920807694e-05, "loss": 0.0005, "step": 774 }, { "epoch": 0.5, "learning_rate": 1.904762036946186e-05, "loss": 0.0009, "step": 775 }, { "epoch": 0.5, "learning_rate": 1.9044662460859703e-05, "loss": 0.0008, "step": 776 }, { "epoch": 0.5, "learning_rate": 1.904170019642573e-05, "loss": 0.0006, "step": 777 }, { "epoch": 0.5, "learning_rate": 1.9038733577586534e-05, "loss": 0.0005, "step": 778 }, { "epoch": 0.5, "learning_rate": 1.9035762605770826e-05, "loss": 0.0011, "step": 779 }, { "epoch": 0.5, "learning_rate": 1.903278728240939e-05, "loss": 0.0007, "step": 780 }, { "epoch": 0.5, "learning_rate": 1.902980760893512e-05, "loss": 0.0005, "step": 781 }, { "epoch": 0.5, "learning_rate": 1.9026823586783004e-05, "loss": 0.0007, "step": 782 }, { "epoch": 0.5, "learning_rate": 1.9023835217390116e-05, "loss": 0.0011, "step": 783 }, { "epoch": 0.5, "learning_rate": 1.9020842502195634e-05, "loss": 0.0004, "step": 784 }, { "epoch": 0.5, "learning_rate": 1.901784544264082e-05, "loss": 0.0007, "step": 785 }, { "epoch": 0.51, "learning_rate": 1.9014844040169033e-05, "loss": 0.0006, "step": 786 }, { "epoch": 0.51, "learning_rate": 1.9011838296225725e-05, "loss": 0.001, "step": 787 }, { "epoch": 0.51, "learning_rate": 1.9008828212258437e-05, "loss": 0.0004, "step": 788 }, { "epoch": 0.51, "learning_rate": 1.9005813789716795e-05, "loss": 0.0005, "step": 789 }, { "epoch": 0.51, "learning_rate": 1.9002795030052522e-05, "loss": 0.0008, "step": 790 }, { "epoch": 0.51, "learning_rate": 1.8999771934719426e-05, "loss": 0.0006, "step": 791 }, { "epoch": 0.51, "learning_rate": 1.8996744505173405e-05, "loss": 0.0006, "step": 792 }, { "epoch": 0.51, "learning_rate": 1.8993712742872442e-05, "loss": 0.0007, "step": 793 }, { "epoch": 0.51, "learning_rate": 1.8990676649276608e-05, "loss": 0.0004, "step": 794 }, { "epoch": 0.51, "learning_rate": 1.898763622584806e-05, "loss": 0.0009, "step": 795 }, { "epoch": 0.51, "learning_rate": 1.898459147405104e-05, "loss": 0.0005, "step": 796 }, { "epoch": 0.51, "learning_rate": 1.898154239535187e-05, "loss": 0.0006, "step": 797 }, { "epoch": 0.51, "learning_rate": 1.8978488991218967e-05, "loss": 0.0006, "step": 798 }, { "epoch": 0.51, "learning_rate": 1.897543126312282e-05, "loss": 0.0006, "step": 799 }, { "epoch": 0.51, "learning_rate": 1.8972369212536006e-05, "loss": 0.0009, "step": 800 }, { "epoch": 0.51, "learning_rate": 1.8969302840933182e-05, "loss": 0.0013, "step": 801 }, { "epoch": 0.52, "learning_rate": 1.8966232149791088e-05, "loss": 0.0007, "step": 802 }, { "epoch": 0.52, "learning_rate": 1.8963157140588537e-05, "loss": 0.0009, "step": 803 }, { "epoch": 0.52, "learning_rate": 1.8960077814806432e-05, "loss": 0.0007, "step": 804 }, { "epoch": 0.52, "learning_rate": 1.895699417392775e-05, "loss": 0.0008, "step": 805 }, { "epoch": 0.52, "learning_rate": 1.8953906219437544e-05, "loss": 0.0004, "step": 806 }, { "epoch": 0.52, "learning_rate": 1.8950813952822948e-05, "loss": 0.001, "step": 807 }, { "epoch": 0.52, "learning_rate": 1.8947717375573167e-05, "loss": 0.0005, "step": 808 }, { "epoch": 0.52, "learning_rate": 1.8944616489179493e-05, "loss": 0.001, "step": 809 }, { "epoch": 0.52, "learning_rate": 1.894151129513528e-05, "loss": 0.0016, "step": 810 }, { "epoch": 0.52, "learning_rate": 1.8938401794935966e-05, "loss": 0.0005, "step": 811 }, { "epoch": 0.52, "learning_rate": 1.8935287990079057e-05, "loss": 0.0003, "step": 812 }, { "epoch": 0.52, "learning_rate": 1.893216988206414e-05, "loss": 0.0004, "step": 813 }, { "epoch": 0.52, "learning_rate": 1.8929047472392863e-05, "loss": 0.0005, "step": 814 }, { "epoch": 0.52, "learning_rate": 1.8925920762568952e-05, "loss": 0.0005, "step": 815 }, { "epoch": 0.52, "learning_rate": 1.892278975409821e-05, "loss": 0.0004, "step": 816 }, { "epoch": 0.52, "learning_rate": 1.8919654448488498e-05, "loss": 0.0007, "step": 817 }, { "epoch": 0.53, "learning_rate": 1.8916514847249753e-05, "loss": 0.0004, "step": 818 }, { "epoch": 0.53, "learning_rate": 1.8913370951893985e-05, "loss": 0.0003, "step": 819 }, { "epoch": 0.53, "learning_rate": 1.891022276393526e-05, "loss": 0.0005, "step": 820 }, { "epoch": 0.53, "learning_rate": 1.890707028488972e-05, "loss": 0.0007, "step": 821 }, { "epoch": 0.53, "learning_rate": 1.8903913516275573e-05, "loss": 0.0005, "step": 822 }, { "epoch": 0.53, "learning_rate": 1.8900752459613093e-05, "loss": 0.0005, "step": 823 }, { "epoch": 0.53, "learning_rate": 1.8897587116424615e-05, "loss": 0.0007, "step": 824 }, { "epoch": 0.53, "learning_rate": 1.8894417488234547e-05, "loss": 0.0008, "step": 825 }, { "epoch": 0.53, "learning_rate": 1.8891243576569348e-05, "loss": 0.0003, "step": 826 }, { "epoch": 0.53, "learning_rate": 1.8888065382957545e-05, "loss": 0.0008, "step": 827 }, { "epoch": 0.53, "learning_rate": 1.8884882908929734e-05, "loss": 0.0008, "step": 828 }, { "epoch": 0.53, "learning_rate": 1.888169615601857e-05, "loss": 0.0004, "step": 829 }, { "epoch": 0.53, "learning_rate": 1.887850512575876e-05, "loss": 0.0004, "step": 830 }, { "epoch": 0.53, "learning_rate": 1.887530981968708e-05, "loss": 0.0007, "step": 831 }, { "epoch": 0.53, "learning_rate": 1.887211023934236e-05, "loss": 0.0006, "step": 832 }, { "epoch": 0.54, "learning_rate": 1.8868906386265487e-05, "loss": 0.001, "step": 833 }, { "epoch": 0.54, "learning_rate": 1.886569826199942e-05, "loss": 0.0003, "step": 834 }, { "epoch": 0.54, "learning_rate": 1.8862485868089153e-05, "loss": 0.0006, "step": 835 }, { "epoch": 0.54, "learning_rate": 1.8859269206081753e-05, "loss": 0.0017, "step": 836 }, { "epoch": 0.54, "learning_rate": 1.8856048277526335e-05, "loss": 0.0006, "step": 837 }, { "epoch": 0.54, "learning_rate": 1.885282308397407e-05, "loss": 0.0003, "step": 838 }, { "epoch": 0.54, "learning_rate": 1.8849593626978183e-05, "loss": 0.0007, "step": 839 }, { "epoch": 0.54, "learning_rate": 1.8846359908093957e-05, "loss": 0.0006, "step": 840 }, { "epoch": 0.54, "learning_rate": 1.884312192887872e-05, "loss": 0.0011, "step": 841 }, { "epoch": 0.54, "learning_rate": 1.8839879690891858e-05, "loss": 0.0007, "step": 842 }, { "epoch": 0.54, "learning_rate": 1.8836633195694798e-05, "loss": 0.0008, "step": 843 }, { "epoch": 0.54, "learning_rate": 1.883338244485103e-05, "loss": 0.0011, "step": 844 }, { "epoch": 0.54, "learning_rate": 1.883012743992609e-05, "loss": 0.0006, "step": 845 }, { "epoch": 0.54, "learning_rate": 1.8826868182487553e-05, "loss": 0.0006, "step": 846 }, { "epoch": 0.54, "learning_rate": 1.8823604674105054e-05, "loss": 0.0006, "step": 847 }, { "epoch": 0.54, "learning_rate": 1.882033691635027e-05, "loss": 0.0004, "step": 848 }, { "epoch": 0.55, "learning_rate": 1.8817064910796928e-05, "loss": 0.0018, "step": 849 }, { "epoch": 0.55, "learning_rate": 1.8813788659020794e-05, "loss": 0.0005, "step": 850 }, { "epoch": 0.55, "learning_rate": 1.8810508162599678e-05, "loss": 0.0003, "step": 851 }, { "epoch": 0.55, "learning_rate": 1.880722342311345e-05, "loss": 0.0004, "step": 852 }, { "epoch": 0.55, "learning_rate": 1.8803934442144006e-05, "loss": 0.0009, "step": 853 }, { "epoch": 0.55, "learning_rate": 1.8800641221275292e-05, "loss": 0.0005, "step": 854 }, { "epoch": 0.55, "learning_rate": 1.879734376209329e-05, "loss": 0.0009, "step": 855 }, { "epoch": 0.55, "learning_rate": 1.8794042066186038e-05, "loss": 0.0004, "step": 856 }, { "epoch": 0.55, "learning_rate": 1.87907361351436e-05, "loss": 0.0003, "step": 857 }, { "epoch": 0.55, "learning_rate": 1.878742597055808e-05, "loss": 0.0008, "step": 858 }, { "epoch": 0.55, "learning_rate": 1.878411157402363e-05, "loss": 0.0005, "step": 859 }, { "epoch": 0.55, "learning_rate": 1.878079294713643e-05, "loss": 0.0005, "step": 860 }, { "epoch": 0.55, "learning_rate": 1.877747009149471e-05, "loss": 0.0006, "step": 861 }, { "epoch": 0.55, "learning_rate": 1.877414300869872e-05, "loss": 0.0011, "step": 862 }, { "epoch": 0.55, "learning_rate": 1.877081170035076e-05, "loss": 0.0013, "step": 863 }, { "epoch": 0.56, "learning_rate": 1.8767476168055156e-05, "loss": 0.0005, "step": 864 }, { "epoch": 0.56, "learning_rate": 1.8764136413418276e-05, "loss": 0.0005, "step": 865 }, { "epoch": 0.56, "learning_rate": 1.8760792438048516e-05, "loss": 0.0008, "step": 866 }, { "epoch": 0.56, "learning_rate": 1.8757444243556306e-05, "loss": 0.0005, "step": 867 }, { "epoch": 0.56, "learning_rate": 1.87540918315541e-05, "loss": 0.0004, "step": 868 }, { "epoch": 0.56, "learning_rate": 1.8750735203656403e-05, "loss": 0.0007, "step": 869 }, { "epoch": 0.56, "learning_rate": 1.874737436147973e-05, "loss": 0.0004, "step": 870 }, { "epoch": 0.56, "learning_rate": 1.8744009306642637e-05, "loss": 0.0006, "step": 871 }, { "epoch": 0.56, "learning_rate": 1.8740640040765704e-05, "loss": 0.0005, "step": 872 }, { "epoch": 0.56, "learning_rate": 1.873726656547154e-05, "loss": 0.0012, "step": 873 }, { "epoch": 0.56, "learning_rate": 1.8733888882384786e-05, "loss": 0.0005, "step": 874 }, { "epoch": 0.56, "learning_rate": 1.8730506993132103e-05, "loss": 0.0004, "step": 875 }, { "epoch": 0.56, "learning_rate": 1.872712089934218e-05, "loss": 0.0004, "step": 876 }, { "epoch": 0.56, "learning_rate": 1.8723730602645728e-05, "loss": 0.001, "step": 877 }, { "epoch": 0.56, "learning_rate": 1.8720336104675484e-05, "loss": 0.0008, "step": 878 }, { "epoch": 0.56, "learning_rate": 1.8716937407066214e-05, "loss": 0.0003, "step": 879 }, { "epoch": 0.57, "learning_rate": 1.8713534511454698e-05, "loss": 0.0005, "step": 880 }, { "epoch": 0.57, "learning_rate": 1.8710127419479744e-05, "loss": 0.0007, "step": 881 }, { "epoch": 0.57, "learning_rate": 1.8706716132782178e-05, "loss": 0.001, "step": 882 }, { "epoch": 0.57, "learning_rate": 1.8703300653004842e-05, "loss": 0.0004, "step": 883 }, { "epoch": 0.57, "learning_rate": 1.8699880981792605e-05, "loss": 0.0003, "step": 884 }, { "epoch": 0.57, "learning_rate": 1.8696457120792355e-05, "loss": 0.0003, "step": 885 }, { "epoch": 0.57, "learning_rate": 1.869302907165299e-05, "loss": 0.0004, "step": 886 }, { "epoch": 0.57, "learning_rate": 1.8689596836025426e-05, "loss": 0.0004, "step": 887 }, { "epoch": 0.57, "learning_rate": 1.8686160415562605e-05, "loss": 0.0004, "step": 888 }, { "epoch": 0.57, "learning_rate": 1.8682719811919473e-05, "loss": 0.0004, "step": 889 }, { "epoch": 0.57, "learning_rate": 1.8679275026752998e-05, "loss": 0.0009, "step": 890 }, { "epoch": 0.57, "learning_rate": 1.8675826061722155e-05, "loss": 0.0003, "step": 891 }, { "epoch": 0.57, "learning_rate": 1.867237291848794e-05, "loss": 0.0005, "step": 892 }, { "epoch": 0.57, "learning_rate": 1.8668915598713354e-05, "loss": 0.0004, "step": 893 }, { "epoch": 0.57, "learning_rate": 1.8665454104063415e-05, "loss": 0.0005, "step": 894 }, { "epoch": 0.58, "learning_rate": 1.8661988436205148e-05, "loss": 0.0005, "step": 895 }, { "epoch": 0.58, "learning_rate": 1.865851859680759e-05, "loss": 0.0008, "step": 896 }, { "epoch": 0.58, "learning_rate": 1.8655044587541785e-05, "loss": 0.0004, "step": 897 }, { "epoch": 0.58, "learning_rate": 1.8651566410080786e-05, "loss": 0.0005, "step": 898 }, { "epoch": 0.58, "learning_rate": 1.864808406609965e-05, "loss": 0.0005, "step": 899 }, { "epoch": 0.58, "learning_rate": 1.8644597557275454e-05, "loss": 0.0002, "step": 900 }, { "epoch": 0.58, "learning_rate": 1.864110688528726e-05, "loss": 0.0004, "step": 901 }, { "epoch": 0.58, "learning_rate": 1.8637612051816148e-05, "loss": 0.0003, "step": 902 }, { "epoch": 0.58, "learning_rate": 1.8634113058545204e-05, "loss": 0.0006, "step": 903 }, { "epoch": 0.58, "learning_rate": 1.8630609907159512e-05, "loss": 0.0004, "step": 904 }, { "epoch": 0.58, "learning_rate": 1.8627102599346157e-05, "loss": 0.0006, "step": 905 }, { "epoch": 0.58, "learning_rate": 1.8623591136794226e-05, "loss": 0.0004, "step": 906 }, { "epoch": 0.58, "learning_rate": 1.862007552119482e-05, "loss": 0.0003, "step": 907 }, { "epoch": 0.58, "learning_rate": 1.8616555754241017e-05, "loss": 0.0005, "step": 908 }, { "epoch": 0.58, "learning_rate": 1.8613031837627915e-05, "loss": 0.0003, "step": 909 }, { "epoch": 0.58, "learning_rate": 1.8609503773052595e-05, "loss": 0.0005, "step": 910 }, { "epoch": 0.59, "learning_rate": 1.860597156221415e-05, "loss": 0.0009, "step": 911 }, { "epoch": 0.59, "learning_rate": 1.8602435206813657e-05, "loss": 0.0008, "step": 912 }, { "epoch": 0.59, "learning_rate": 1.8598894708554197e-05, "loss": 0.0004, "step": 913 }, { "epoch": 0.59, "learning_rate": 1.859535006914084e-05, "loss": 0.0007, "step": 914 }, { "epoch": 0.59, "learning_rate": 1.8591801290280664e-05, "loss": 0.0005, "step": 915 }, { "epoch": 0.59, "learning_rate": 1.8588248373682717e-05, "loss": 0.0005, "step": 916 }, { "epoch": 0.59, "learning_rate": 1.8584691321058064e-05, "loss": 0.0005, "step": 917 }, { "epoch": 0.59, "learning_rate": 1.8581130134119742e-05, "loss": 0.0009, "step": 918 }, { "epoch": 0.59, "learning_rate": 1.8577564814582795e-05, "loss": 0.0006, "step": 919 }, { "epoch": 0.59, "learning_rate": 1.8573995364164247e-05, "loss": 0.0003, "step": 920 }, { "epoch": 0.59, "learning_rate": 1.8570421784583114e-05, "loss": 0.0002, "step": 921 }, { "epoch": 0.59, "learning_rate": 1.85668440775604e-05, "loss": 0.0004, "step": 922 }, { "epoch": 0.59, "learning_rate": 1.8563262244819107e-05, "loss": 0.0004, "step": 923 }, { "epoch": 0.59, "learning_rate": 1.855967628808421e-05, "loss": 0.0005, "step": 924 }, { "epoch": 0.59, "learning_rate": 1.8556086209082665e-05, "loss": 0.0007, "step": 925 }, { "epoch": 0.6, "learning_rate": 1.8552492009543436e-05, "loss": 0.0008, "step": 926 }, { "epoch": 0.6, "learning_rate": 1.8548893691197457e-05, "loss": 0.0002, "step": 927 }, { "epoch": 0.6, "learning_rate": 1.8545291255777643e-05, "loss": 0.0007, "step": 928 }, { "epoch": 0.6, "learning_rate": 1.8541684705018897e-05, "loss": 0.0007, "step": 929 }, { "epoch": 0.6, "learning_rate": 1.85380740406581e-05, "loss": 0.0004, "step": 930 }, { "epoch": 0.6, "learning_rate": 1.8534459264434126e-05, "loss": 0.0005, "step": 931 }, { "epoch": 0.6, "learning_rate": 1.8530840378087815e-05, "loss": 0.0006, "step": 932 }, { "epoch": 0.6, "learning_rate": 1.852721738336199e-05, "loss": 0.0008, "step": 933 }, { "epoch": 0.6, "learning_rate": 1.8523590282001452e-05, "loss": 0.0005, "step": 934 }, { "epoch": 0.6, "learning_rate": 1.851995907575299e-05, "loss": 0.0002, "step": 935 }, { "epoch": 0.6, "learning_rate": 1.8516323766365355e-05, "loss": 0.0007, "step": 936 }, { "epoch": 0.6, "learning_rate": 1.8512684355589282e-05, "loss": 0.0011, "step": 937 }, { "epoch": 0.6, "learning_rate": 1.8509040845177486e-05, "loss": 0.0002, "step": 938 }, { "epoch": 0.6, "learning_rate": 1.8505393236884637e-05, "loss": 0.0002, "step": 939 }, { "epoch": 0.6, "learning_rate": 1.8501741532467406e-05, "loss": 0.0005, "step": 940 }, { "epoch": 0.6, "learning_rate": 1.8498085733684415e-05, "loss": 0.0006, "step": 941 }, { "epoch": 0.61, "learning_rate": 1.849442584229627e-05, "loss": 0.0007, "step": 942 }, { "epoch": 0.61, "learning_rate": 1.8490761860065534e-05, "loss": 0.0008, "step": 943 }, { "epoch": 0.61, "learning_rate": 1.848709378875676e-05, "loss": 0.0003, "step": 944 }, { "epoch": 0.61, "learning_rate": 1.8483421630136457e-05, "loss": 0.0003, "step": 945 }, { "epoch": 0.61, "learning_rate": 1.84797453859731e-05, "loss": 0.0006, "step": 946 }, { "epoch": 0.61, "learning_rate": 1.847606505803714e-05, "loss": 0.0005, "step": 947 }, { "epoch": 0.61, "learning_rate": 1.8472380648100993e-05, "loss": 0.0003, "step": 948 }, { "epoch": 0.61, "learning_rate": 1.8468692157939037e-05, "loss": 0.0003, "step": 949 }, { "epoch": 0.61, "learning_rate": 1.846499958932762e-05, "loss": 0.0008, "step": 950 }, { "epoch": 0.61, "learning_rate": 1.8461302944045045e-05, "loss": 0.0006, "step": 951 }, { "epoch": 0.61, "learning_rate": 1.845760222387159e-05, "loss": 0.001, "step": 952 }, { "epoch": 0.61, "learning_rate": 1.845389743058949e-05, "loss": 0.0004, "step": 953 }, { "epoch": 0.61, "learning_rate": 1.8450188565982942e-05, "loss": 0.0005, "step": 954 }, { "epoch": 0.61, "learning_rate": 1.84464756318381e-05, "loss": 0.0003, "step": 955 }, { "epoch": 0.61, "learning_rate": 1.844275862994308e-05, "loss": 0.0006, "step": 956 }, { "epoch": 0.61, "learning_rate": 1.8439037562087966e-05, "loss": 0.0008, "step": 957 }, { "epoch": 0.62, "learning_rate": 1.8435312430064786e-05, "loss": 0.0005, "step": 958 }, { "epoch": 0.62, "learning_rate": 1.8431583235667532e-05, "loss": 0.0006, "step": 959 }, { "epoch": 0.62, "learning_rate": 1.8427849980692155e-05, "loss": 0.0006, "step": 960 }, { "epoch": 0.62, "learning_rate": 1.8424112666936556e-05, "loss": 0.0002, "step": 961 }, { "epoch": 0.62, "learning_rate": 1.8420371296200596e-05, "loss": 0.0005, "step": 962 }, { "epoch": 0.62, "learning_rate": 1.8416625870286083e-05, "loss": 0.0033, "step": 963 }, { "epoch": 0.62, "learning_rate": 1.8412876390996787e-05, "loss": 0.0003, "step": 964 }, { "epoch": 0.62, "learning_rate": 1.840912286013842e-05, "loss": 0.0006, "step": 965 }, { "epoch": 0.62, "learning_rate": 1.8405365279518653e-05, "loss": 0.0003, "step": 966 }, { "epoch": 0.62, "learning_rate": 1.8401603650947102e-05, "loss": 0.0011, "step": 967 }, { "epoch": 0.62, "learning_rate": 1.8397837976235338e-05, "loss": 0.001, "step": 968 }, { "epoch": 0.62, "learning_rate": 1.839406825719688e-05, "loss": 0.0009, "step": 969 }, { "epoch": 0.62, "learning_rate": 1.8390294495647183e-05, "loss": 0.0008, "step": 970 }, { "epoch": 0.62, "learning_rate": 1.8386516693403666e-05, "loss": 0.0004, "step": 971 }, { "epoch": 0.62, "learning_rate": 1.8382734852285686e-05, "loss": 0.0004, "step": 972 }, { "epoch": 0.63, "learning_rate": 1.837894897411454e-05, "loss": 0.001, "step": 973 }, { "epoch": 0.63, "learning_rate": 1.837515906071348e-05, "loss": 0.0008, "step": 974 }, { "epoch": 0.63, "learning_rate": 1.83713651139077e-05, "loss": 0.0005, "step": 975 }, { "epoch": 0.63, "learning_rate": 1.836756713552432e-05, "loss": 0.0005, "step": 976 }, { "epoch": 0.63, "learning_rate": 1.836376512739242e-05, "loss": 0.0004, "step": 977 }, { "epoch": 0.63, "learning_rate": 1.835995909134301e-05, "loss": 0.0004, "step": 978 }, { "epoch": 0.63, "learning_rate": 1.8356149029209053e-05, "loss": 0.0005, "step": 979 }, { "epoch": 0.63, "learning_rate": 1.835233494282544e-05, "loss": 0.0005, "step": 980 }, { "epoch": 0.63, "learning_rate": 1.8348516834029e-05, "loss": 0.0003, "step": 981 }, { "epoch": 0.63, "learning_rate": 1.8344694704658494e-05, "loss": 0.0005, "step": 982 }, { "epoch": 0.63, "learning_rate": 1.834086855655464e-05, "loss": 0.0009, "step": 983 }, { "epoch": 0.63, "learning_rate": 1.833703839156007e-05, "loss": 0.0005, "step": 984 }, { "epoch": 0.63, "learning_rate": 1.8333204211519364e-05, "loss": 0.0007, "step": 985 }, { "epoch": 0.63, "learning_rate": 1.8329366018279023e-05, "loss": 0.0007, "step": 986 }, { "epoch": 0.63, "learning_rate": 1.8325523813687495e-05, "loss": 0.0009, "step": 987 }, { "epoch": 0.63, "learning_rate": 1.832167759959515e-05, "loss": 0.0003, "step": 988 }, { "epoch": 0.64, "learning_rate": 1.831782737785429e-05, "loss": 0.0002, "step": 989 }, { "epoch": 0.64, "learning_rate": 1.831397315031915e-05, "loss": 0.0004, "step": 990 }, { "epoch": 0.64, "learning_rate": 1.8310114918845892e-05, "loss": 0.002, "step": 991 }, { "epoch": 0.64, "learning_rate": 1.830625268529261e-05, "loss": 0.0006, "step": 992 }, { "epoch": 0.64, "learning_rate": 1.830238645151932e-05, "loss": 0.0004, "step": 993 }, { "epoch": 0.64, "learning_rate": 1.829851621938797e-05, "loss": 0.0005, "step": 994 }, { "epoch": 0.64, "learning_rate": 1.8294641990762426e-05, "loss": 0.0004, "step": 995 }, { "epoch": 0.64, "learning_rate": 1.8290763767508483e-05, "loss": 0.0007, "step": 996 }, { "epoch": 0.64, "learning_rate": 1.8286881551493866e-05, "loss": 0.0003, "step": 997 }, { "epoch": 0.64, "learning_rate": 1.8282995344588213e-05, "loss": 0.0004, "step": 998 }, { "epoch": 0.64, "learning_rate": 1.827910514866309e-05, "loss": 0.0009, "step": 999 }, { "epoch": 0.64, "learning_rate": 1.8275210965591978e-05, "loss": 0.0001, "step": 1000 }, { "epoch": 0.64, "learning_rate": 1.8271312797250285e-05, "loss": 0.0118, "step": 1001 }, { "epoch": 0.64, "learning_rate": 1.8267410645515333e-05, "loss": 0.0005, "step": 1002 }, { "epoch": 0.64, "learning_rate": 1.826350451226637e-05, "loss": 0.0006, "step": 1003 }, { "epoch": 0.65, "learning_rate": 1.8259594399384553e-05, "loss": 0.0005, "step": 1004 }, { "epoch": 0.65, "learning_rate": 1.825568030875296e-05, "loss": 0.0008, "step": 1005 }, { "epoch": 0.65, "learning_rate": 1.8251762242256576e-05, "loss": 0.0003, "step": 1006 }, { "epoch": 0.65, "learning_rate": 1.8247840201782318e-05, "loss": 0.0002, "step": 1007 }, { "epoch": 0.65, "learning_rate": 1.8243914189219007e-05, "loss": 0.0006, "step": 1008 }, { "epoch": 0.65, "learning_rate": 1.8239984206457367e-05, "loss": 0.0007, "step": 1009 }, { "epoch": 0.65, "learning_rate": 1.8236050255390056e-05, "loss": 0.0006, "step": 1010 }, { "epoch": 0.65, "learning_rate": 1.823211233791162e-05, "loss": 0.0008, "step": 1011 }, { "epoch": 0.65, "learning_rate": 1.8228170455918537e-05, "loss": 0.0004, "step": 1012 }, { "epoch": 0.65, "learning_rate": 1.8224224611309177e-05, "loss": 0.0005, "step": 1013 }, { "epoch": 0.65, "learning_rate": 1.8220274805983828e-05, "loss": 0.0006, "step": 1014 }, { "epoch": 0.65, "learning_rate": 1.8216321041844676e-05, "loss": 0.0008, "step": 1015 }, { "epoch": 0.65, "learning_rate": 1.8212363320795827e-05, "loss": 0.0003, "step": 1016 }, { "epoch": 0.65, "learning_rate": 1.820840164474328e-05, "loss": 0.0005, "step": 1017 }, { "epoch": 0.65, "learning_rate": 1.820443601559495e-05, "loss": 0.0004, "step": 1018 }, { "epoch": 0.65, "learning_rate": 1.8200466435260646e-05, "loss": 0.0004, "step": 1019 }, { "epoch": 0.66, "learning_rate": 1.8196492905652085e-05, "loss": 0.0004, "step": 1020 }, { "epoch": 0.66, "learning_rate": 1.8192515428682887e-05, "loss": 0.0005, "step": 1021 }, { "epoch": 0.66, "learning_rate": 1.8188534006268566e-05, "loss": 0.0006, "step": 1022 }, { "epoch": 0.66, "learning_rate": 1.8184548640326543e-05, "loss": 0.0003, "step": 1023 }, { "epoch": 0.66, "learning_rate": 1.818055933277614e-05, "loss": 0.0002, "step": 1024 }, { "epoch": 0.66, "learning_rate": 1.8176566085538562e-05, "loss": 0.0004, "step": 1025 }, { "epoch": 0.66, "learning_rate": 1.8172568900536934e-05, "loss": 0.0005, "step": 1026 }, { "epoch": 0.66, "learning_rate": 1.8168567779696265e-05, "loss": 0.0005, "step": 1027 }, { "epoch": 0.66, "learning_rate": 1.8164562724943456e-05, "loss": 0.0005, "step": 1028 }, { "epoch": 0.66, "learning_rate": 1.8160553738207308e-05, "loss": 0.0005, "step": 1029 }, { "epoch": 0.66, "learning_rate": 1.8156540821418516e-05, "loss": 0.0004, "step": 1030 }, { "epoch": 0.66, "learning_rate": 1.8152523976509664e-05, "loss": 0.0004, "step": 1031 }, { "epoch": 0.66, "learning_rate": 1.814850320541523e-05, "loss": 0.0012, "step": 1032 }, { "epoch": 0.66, "learning_rate": 1.814447851007159e-05, "loss": 0.0003, "step": 1033 }, { "epoch": 0.66, "learning_rate": 1.8140449892416994e-05, "loss": 0.0005, "step": 1034 }, { "epoch": 0.67, "learning_rate": 1.813641735439159e-05, "loss": 0.0005, "step": 1035 }, { "epoch": 0.67, "learning_rate": 1.813238089793742e-05, "loss": 0.0005, "step": 1036 }, { "epoch": 0.67, "learning_rate": 1.8128340524998403e-05, "loss": 0.0012, "step": 1037 }, { "epoch": 0.67, "learning_rate": 1.812429623752035e-05, "loss": 0.0004, "step": 1038 }, { "epoch": 0.67, "learning_rate": 1.8120248037450953e-05, "loss": 0.0004, "step": 1039 }, { "epoch": 0.67, "learning_rate": 1.8116195926739785e-05, "loss": 0.0004, "step": 1040 }, { "epoch": 0.67, "learning_rate": 1.811213990733832e-05, "loss": 0.0007, "step": 1041 }, { "epoch": 0.67, "learning_rate": 1.810807998119989e-05, "loss": 0.0004, "step": 1042 }, { "epoch": 0.67, "learning_rate": 1.8104016150279734e-05, "loss": 0.0004, "step": 1043 }, { "epoch": 0.67, "learning_rate": 1.8099948416534944e-05, "loss": 0.0005, "step": 1044 }, { "epoch": 0.67, "learning_rate": 1.809587678192451e-05, "loss": 0.0005, "step": 1045 }, { "epoch": 0.67, "learning_rate": 1.8091801248409306e-05, "loss": 0.0004, "step": 1046 }, { "epoch": 0.67, "learning_rate": 1.808772181795206e-05, "loss": 0.0007, "step": 1047 }, { "epoch": 0.67, "learning_rate": 1.8083638492517402e-05, "loss": 0.0011, "step": 1048 }, { "epoch": 0.67, "learning_rate": 1.807955127407182e-05, "loss": 0.0004, "step": 1049 }, { "epoch": 0.67, "learning_rate": 1.807546016458368e-05, "loss": 0.0001, "step": 1050 }, { "epoch": 0.68, "learning_rate": 1.8071365166023237e-05, "loss": 0.0015, "step": 1051 }, { "epoch": 0.68, "learning_rate": 1.8067266280362596e-05, "loss": 0.0002, "step": 1052 }, { "epoch": 0.68, "learning_rate": 1.806316350957575e-05, "loss": 0.0003, "step": 1053 }, { "epoch": 0.68, "learning_rate": 1.8059056855638556e-05, "loss": 0.0003, "step": 1054 }, { "epoch": 0.68, "learning_rate": 1.805494632052875e-05, "loss": 0.0004, "step": 1055 }, { "epoch": 0.68, "learning_rate": 1.805083190622592e-05, "loss": 0.0003, "step": 1056 }, { "epoch": 0.68, "learning_rate": 1.804671361471154e-05, "loss": 0.0003, "step": 1057 }, { "epoch": 0.68, "learning_rate": 1.8042591447968944e-05, "loss": 0.0008, "step": 1058 }, { "epoch": 0.68, "learning_rate": 1.8038465407983327e-05, "loss": 0.0003, "step": 1059 }, { "epoch": 0.68, "learning_rate": 1.8034335496741754e-05, "loss": 0.0007, "step": 1060 }, { "epoch": 0.68, "learning_rate": 1.8030201716233162e-05, "loss": 0.0006, "step": 1061 }, { "epoch": 0.68, "learning_rate": 1.8026064068448338e-05, "loss": 0.0002, "step": 1062 }, { "epoch": 0.68, "learning_rate": 1.802192255537994e-05, "loss": 0.0003, "step": 1063 }, { "epoch": 0.68, "learning_rate": 1.801777717902249e-05, "loss": 0.0003, "step": 1064 }, { "epoch": 0.68, "learning_rate": 1.8013627941372357e-05, "loss": 0.0003, "step": 1065 }, { "epoch": 0.68, "learning_rate": 1.8009474844427782e-05, "loss": 0.0008, "step": 1066 }, { "epoch": 0.69, "learning_rate": 1.8005317890188864e-05, "loss": 0.0003, "step": 1067 }, { "epoch": 0.69, "learning_rate": 1.800115708065755e-05, "loss": 0.0006, "step": 1068 }, { "epoch": 0.69, "learning_rate": 1.7996992417837655e-05, "loss": 0.0007, "step": 1069 }, { "epoch": 0.69, "learning_rate": 1.7992823903734845e-05, "loss": 0.0003, "step": 1070 }, { "epoch": 0.69, "learning_rate": 1.7988651540356638e-05, "loss": 0.0004, "step": 1071 }, { "epoch": 0.69, "learning_rate": 1.7984475329712405e-05, "loss": 0.0002, "step": 1072 }, { "epoch": 0.69, "learning_rate": 1.7980295273813383e-05, "loss": 0.0003, "step": 1073 }, { "epoch": 0.69, "learning_rate": 1.7976111374672643e-05, "loss": 0.0003, "step": 1074 }, { "epoch": 0.69, "learning_rate": 1.7971923634305118e-05, "loss": 0.0007, "step": 1075 }, { "epoch": 0.69, "learning_rate": 1.796773205472759e-05, "loss": 0.0003, "step": 1076 }, { "epoch": 0.69, "learning_rate": 1.796353663795868e-05, "loss": 0.0003, "step": 1077 }, { "epoch": 0.69, "learning_rate": 1.7959337386018874e-05, "loss": 0.0004, "step": 1078 }, { "epoch": 0.69, "learning_rate": 1.795513430093049e-05, "loss": 0.0007, "step": 1079 }, { "epoch": 0.69, "learning_rate": 1.79509273847177e-05, "loss": 0.0006, "step": 1080 }, { "epoch": 0.69, "learning_rate": 1.7946716639406516e-05, "loss": 0.0005, "step": 1081 }, { "epoch": 0.7, "learning_rate": 1.79425020670248e-05, "loss": 0.0003, "step": 1082 }, { "epoch": 0.7, "learning_rate": 1.7938283669602253e-05, "loss": 0.0005, "step": 1083 }, { "epoch": 0.7, "learning_rate": 1.793406144917042e-05, "loss": 0.0008, "step": 1084 }, { "epoch": 0.7, "learning_rate": 1.7929835407762682e-05, "loss": 0.0004, "step": 1085 }, { "epoch": 0.7, "learning_rate": 1.792560554741427e-05, "loss": 0.0005, "step": 1086 }, { "epoch": 0.7, "learning_rate": 1.7921371870162244e-05, "loss": 0.001, "step": 1087 }, { "epoch": 0.7, "learning_rate": 1.791713437804551e-05, "loss": 0.0004, "step": 1088 }, { "epoch": 0.7, "learning_rate": 1.7912893073104805e-05, "loss": 0.0006, "step": 1089 }, { "epoch": 0.7, "learning_rate": 1.7908647957382704e-05, "loss": 0.0003, "step": 1090 }, { "epoch": 0.7, "learning_rate": 1.7904399032923623e-05, "loss": 0.0002, "step": 1091 }, { "epoch": 0.7, "learning_rate": 1.7900146301773804e-05, "loss": 0.0003, "step": 1092 }, { "epoch": 0.7, "learning_rate": 1.7895889765981326e-05, "loss": 0.001, "step": 1093 }, { "epoch": 0.7, "learning_rate": 1.7891629427596102e-05, "loss": 0.0006, "step": 1094 }, { "epoch": 0.7, "learning_rate": 1.7887365288669874e-05, "loss": 0.0002, "step": 1095 }, { "epoch": 0.7, "learning_rate": 1.7883097351256215e-05, "loss": 0.0008, "step": 1096 }, { "epoch": 0.7, "learning_rate": 1.787882561741052e-05, "loss": 0.0007, "step": 1097 }, { "epoch": 0.71, "learning_rate": 1.787455008919003e-05, "loss": 0.0004, "step": 1098 }, { "epoch": 0.71, "learning_rate": 1.7870270768653796e-05, "loss": 0.0005, "step": 1099 }, { "epoch": 0.71, "learning_rate": 1.7865987657862707e-05, "loss": 0.0002, "step": 1100 }, { "epoch": 0.71, "learning_rate": 1.7861700758879466e-05, "loss": 0.0005, "step": 1101 }, { "epoch": 0.71, "learning_rate": 1.7857410073768615e-05, "loss": 0.0005, "step": 1102 }, { "epoch": 0.71, "learning_rate": 1.78531156045965e-05, "loss": 0.0004, "step": 1103 }, { "epoch": 0.71, "learning_rate": 1.784881735343131e-05, "loss": 0.0002, "step": 1104 }, { "epoch": 0.71, "learning_rate": 1.7844515322343046e-05, "loss": 0.0015, "step": 1105 }, { "epoch": 0.71, "learning_rate": 1.7840209513403522e-05, "loss": 0.0006, "step": 1106 }, { "epoch": 0.71, "learning_rate": 1.783589992868638e-05, "loss": 0.0004, "step": 1107 }, { "epoch": 0.71, "learning_rate": 1.783158657026709e-05, "loss": 0.0003, "step": 1108 }, { "epoch": 0.71, "learning_rate": 1.7827269440222914e-05, "loss": 0.0005, "step": 1109 }, { "epoch": 0.71, "learning_rate": 1.782294854063295e-05, "loss": 0.0004, "step": 1110 }, { "epoch": 0.71, "learning_rate": 1.781862387357811e-05, "loss": 0.0004, "step": 1111 }, { "epoch": 0.71, "learning_rate": 1.7814295441141115e-05, "loss": 0.0003, "step": 1112 }, { "epoch": 0.72, "learning_rate": 1.7809963245406498e-05, "loss": 0.0007, "step": 1113 }, { "epoch": 0.72, "learning_rate": 1.780562728846061e-05, "loss": 0.0005, "step": 1114 }, { "epoch": 0.72, "learning_rate": 1.7801287572391604e-05, "loss": 0.0004, "step": 1115 }, { "epoch": 0.72, "learning_rate": 1.7796944099289456e-05, "loss": 0.0008, "step": 1116 }, { "epoch": 0.72, "learning_rate": 1.779259687124595e-05, "loss": 0.0002, "step": 1117 }, { "epoch": 0.72, "learning_rate": 1.7788245890354664e-05, "loss": 0.0008, "step": 1118 }, { "epoch": 0.72, "learning_rate": 1.7783891158711e-05, "loss": 0.0002, "step": 1119 }, { "epoch": 0.72, "learning_rate": 1.7779532678412156e-05, "loss": 0.0012, "step": 1120 }, { "epoch": 0.72, "learning_rate": 1.777517045155714e-05, "loss": 0.0008, "step": 1121 }, { "epoch": 0.72, "learning_rate": 1.7770804480246763e-05, "loss": 0.0002, "step": 1122 }, { "epoch": 0.72, "learning_rate": 1.776643476658364e-05, "loss": 0.0002, "step": 1123 }, { "epoch": 0.72, "learning_rate": 1.7762061312672184e-05, "loss": 0.0002, "step": 1124 }, { "epoch": 0.72, "learning_rate": 1.7757684120618617e-05, "loss": 0.0004, "step": 1125 }, { "epoch": 0.72, "learning_rate": 1.7753303192530955e-05, "loss": 0.0001, "step": 1126 }, { "epoch": 0.72, "learning_rate": 1.7748918530519013e-05, "loss": 0.0005, "step": 1127 }, { "epoch": 0.72, "learning_rate": 1.7744530136694416e-05, "loss": 0.0005, "step": 1128 }, { "epoch": 0.73, "learning_rate": 1.7740138013170563e-05, "loss": 0.0007, "step": 1129 }, { "epoch": 0.73, "learning_rate": 1.7735742162062677e-05, "loss": 0.0004, "step": 1130 }, { "epoch": 0.73, "learning_rate": 1.7731342585487753e-05, "loss": 0.0004, "step": 1131 }, { "epoch": 0.73, "learning_rate": 1.772693928556459e-05, "loss": 0.0002, "step": 1132 }, { "epoch": 0.73, "learning_rate": 1.7722532264413782e-05, "loss": 0.0003, "step": 1133 }, { "epoch": 0.73, "learning_rate": 1.7718121524157713e-05, "loss": 0.0005, "step": 1134 }, { "epoch": 0.73, "learning_rate": 1.7713707066920558e-05, "loss": 0.0002, "step": 1135 }, { "epoch": 0.73, "learning_rate": 1.770928889482828e-05, "loss": 0.002, "step": 1136 }, { "epoch": 0.73, "learning_rate": 1.770486701000863e-05, "loss": 0.0008, "step": 1137 }, { "epoch": 0.73, "learning_rate": 1.770044141459115e-05, "loss": 0.0074, "step": 1138 }, { "epoch": 0.73, "learning_rate": 1.7696012110707177e-05, "loss": 0.0004, "step": 1139 }, { "epoch": 0.73, "learning_rate": 1.7691579100489818e-05, "loss": 0.0003, "step": 1140 }, { "epoch": 0.73, "learning_rate": 1.7687142386073975e-05, "loss": 0.0004, "step": 1141 }, { "epoch": 0.73, "learning_rate": 1.7682701969596333e-05, "loss": 0.0006, "step": 1142 }, { "epoch": 0.73, "learning_rate": 1.7678257853195358e-05, "loss": 0.0005, "step": 1143 }, { "epoch": 0.74, "learning_rate": 1.7673810039011296e-05, "loss": 0.0011, "step": 1144 }, { "epoch": 0.74, "learning_rate": 1.7669358529186177e-05, "loss": 0.0006, "step": 1145 }, { "epoch": 0.74, "learning_rate": 1.766490332586381e-05, "loss": 0.0005, "step": 1146 }, { "epoch": 0.74, "learning_rate": 1.766044443118978e-05, "loss": 0.0005, "step": 1147 }, { "epoch": 0.74, "learning_rate": 1.765598184731146e-05, "loss": 0.0006, "step": 1148 }, { "epoch": 0.74, "learning_rate": 1.7651515576377985e-05, "loss": 0.0003, "step": 1149 }, { "epoch": 0.74, "learning_rate": 1.764704562054028e-05, "loss": 0.0005, "step": 1150 }, { "epoch": 0.74, "learning_rate": 1.764257198195103e-05, "loss": 0.0005, "step": 1151 }, { "epoch": 0.74, "learning_rate": 1.76380946627647e-05, "loss": 0.0003, "step": 1152 }, { "epoch": 0.74, "learning_rate": 1.7633613665137534e-05, "loss": 0.0006, "step": 1153 }, { "epoch": 0.74, "learning_rate": 1.7629128991227544e-05, "loss": 0.0006, "step": 1154 }, { "epoch": 0.74, "learning_rate": 1.7624640643194502e-05, "loss": 0.0002, "step": 1155 }, { "epoch": 0.74, "learning_rate": 1.7620148623199963e-05, "loss": 0.0004, "step": 1156 }, { "epoch": 0.74, "learning_rate": 1.761565293340725e-05, "loss": 0.0004, "step": 1157 }, { "epoch": 0.74, "learning_rate": 1.7611153575981435e-05, "loss": 0.0006, "step": 1158 }, { "epoch": 0.74, "learning_rate": 1.7606650553089382e-05, "loss": 0.001, "step": 1159 }, { "epoch": 0.75, "learning_rate": 1.7602143866899705e-05, "loss": 0.0002, "step": 1160 }, { "epoch": 0.75, "learning_rate": 1.7597633519582786e-05, "loss": 0.0005, "step": 1161 }, { "epoch": 0.75, "learning_rate": 1.7593119513310766e-05, "loss": 0.0085, "step": 1162 }, { "epoch": 0.75, "learning_rate": 1.7588601850257557e-05, "loss": 0.0008, "step": 1163 }, { "epoch": 0.75, "learning_rate": 1.7584080532598822e-05, "loss": 0.0005, "step": 1164 }, { "epoch": 0.75, "learning_rate": 1.757955556251199e-05, "loss": 0.0002, "step": 1165 }, { "epoch": 0.75, "learning_rate": 1.757502694217625e-05, "loss": 0.0004, "step": 1166 }, { "epoch": 0.75, "learning_rate": 1.7570494673772545e-05, "loss": 0.0006, "step": 1167 }, { "epoch": 0.75, "learning_rate": 1.7565958759483574e-05, "loss": 0.0003, "step": 1168 }, { "epoch": 0.75, "learning_rate": 1.75614192014938e-05, "loss": 0.0006, "step": 1169 }, { "epoch": 0.75, "learning_rate": 1.7556876001989432e-05, "loss": 0.0006, "step": 1170 }, { "epoch": 0.75, "learning_rate": 1.7552329163158435e-05, "loss": 0.0011, "step": 1171 }, { "epoch": 0.75, "learning_rate": 1.754777868719053e-05, "loss": 0.0003, "step": 1172 }, { "epoch": 0.75, "learning_rate": 1.7543224576277182e-05, "loss": 0.0004, "step": 1173 }, { "epoch": 0.75, "learning_rate": 1.753866683261162e-05, "loss": 0.0006, "step": 1174 }, { "epoch": 0.76, "learning_rate": 1.753410545838881e-05, "loss": 0.0003, "step": 1175 }, { "epoch": 0.76, "learning_rate": 1.7529540455805466e-05, "loss": 0.0006, "step": 1176 }, { "epoch": 0.76, "learning_rate": 1.7524971827060063e-05, "loss": 0.0006, "step": 1177 }, { "epoch": 0.76, "learning_rate": 1.7520399574352807e-05, "loss": 0.0002, "step": 1178 }, { "epoch": 0.76, "learning_rate": 1.7515823699885657e-05, "loss": 0.0004, "step": 1179 }, { "epoch": 0.76, "learning_rate": 1.7511244205862318e-05, "loss": 0.0002, "step": 1180 }, { "epoch": 0.76, "learning_rate": 1.750666109448823e-05, "loss": 0.0003, "step": 1181 }, { "epoch": 0.76, "learning_rate": 1.7502074367970582e-05, "loss": 0.0004, "step": 1182 }, { "epoch": 0.76, "learning_rate": 1.7497484028518304e-05, "loss": 0.0003, "step": 1183 }, { "epoch": 0.76, "learning_rate": 1.7492890078342063e-05, "loss": 0.0003, "step": 1184 }, { "epoch": 0.76, "learning_rate": 1.7488292519654267e-05, "loss": 0.0005, "step": 1185 }, { "epoch": 0.76, "learning_rate": 1.748369135466906e-05, "loss": 0.0002, "step": 1186 }, { "epoch": 0.76, "learning_rate": 1.7479086585602322e-05, "loss": 0.0003, "step": 1187 }, { "epoch": 0.76, "learning_rate": 1.7474478214671668e-05, "loss": 0.0004, "step": 1188 }, { "epoch": 0.76, "learning_rate": 1.7469866244096458e-05, "loss": 0.0002, "step": 1189 }, { "epoch": 0.76, "learning_rate": 1.7465250676097772e-05, "loss": 0.0002, "step": 1190 }, { "epoch": 0.77, "learning_rate": 1.7460631512898424e-05, "loss": 0.0008, "step": 1191 }, { "epoch": 0.77, "learning_rate": 1.7456008756722974e-05, "loss": 0.0003, "step": 1192 }, { "epoch": 0.77, "learning_rate": 1.7451382409797695e-05, "loss": 0.0003, "step": 1193 }, { "epoch": 0.77, "learning_rate": 1.7446752474350593e-05, "loss": 0.0003, "step": 1194 }, { "epoch": 0.77, "learning_rate": 1.7442118952611412e-05, "loss": 0.0005, "step": 1195 }, { "epoch": 0.77, "learning_rate": 1.7437481846811612e-05, "loss": 0.0003, "step": 1196 }, { "epoch": 0.77, "learning_rate": 1.7432841159184385e-05, "loss": 0.0003, "step": 1197 }, { "epoch": 0.77, "learning_rate": 1.742819689196465e-05, "loss": 0.0002, "step": 1198 }, { "epoch": 0.77, "learning_rate": 1.7423549047389034e-05, "loss": 0.001, "step": 1199 }, { "epoch": 0.77, "learning_rate": 1.7418897627695912e-05, "loss": 0.0227, "step": 1200 }, { "epoch": 0.77, "learning_rate": 1.7414242635125363e-05, "loss": 0.0005, "step": 1201 }, { "epoch": 0.77, "learning_rate": 1.740958407191919e-05, "loss": 0.0009, "step": 1202 }, { "epoch": 0.77, "learning_rate": 1.7404921940320922e-05, "loss": 0.0003, "step": 1203 }, { "epoch": 0.77, "learning_rate": 1.7400256242575792e-05, "loss": 0.0003, "step": 1204 }, { "epoch": 0.77, "learning_rate": 1.739558698093077e-05, "loss": 0.0005, "step": 1205 }, { "epoch": 0.77, "learning_rate": 1.7390914157634527e-05, "loss": 0.0002, "step": 1206 }, { "epoch": 0.78, "learning_rate": 1.738623777493746e-05, "loss": 0.0001, "step": 1207 }, { "epoch": 0.78, "learning_rate": 1.7381557835091665e-05, "loss": 0.0008, "step": 1208 }, { "epoch": 0.78, "learning_rate": 1.7376874340350972e-05, "loss": 0.0003, "step": 1209 }, { "epoch": 0.78, "learning_rate": 1.7372187292970903e-05, "loss": 0.0008, "step": 1210 }, { "epoch": 0.78, "learning_rate": 1.7367496695208707e-05, "loss": 0.0003, "step": 1211 }, { "epoch": 0.78, "learning_rate": 1.7362802549323324e-05, "loss": 0.0002, "step": 1212 }, { "epoch": 0.78, "learning_rate": 1.735810485757543e-05, "loss": 0.0003, "step": 1213 }, { "epoch": 0.78, "learning_rate": 1.7353403622227385e-05, "loss": 0.0009, "step": 1214 }, { "epoch": 0.78, "learning_rate": 1.734869884554326e-05, "loss": 0.0006, "step": 1215 }, { "epoch": 0.78, "learning_rate": 1.7343990529788844e-05, "loss": 0.0005, "step": 1216 }, { "epoch": 0.78, "learning_rate": 1.733927867723161e-05, "loss": 0.0009, "step": 1217 }, { "epoch": 0.78, "learning_rate": 1.7334563290140755e-05, "loss": 0.0006, "step": 1218 }, { "epoch": 0.78, "learning_rate": 1.732984437078717e-05, "loss": 0.0002, "step": 1219 }, { "epoch": 0.78, "learning_rate": 1.732512192144344e-05, "loss": 0.0006, "step": 1220 }, { "epoch": 0.78, "learning_rate": 1.7320395944383857e-05, "loss": 0.0004, "step": 1221 }, { "epoch": 0.79, "learning_rate": 1.7315666441884415e-05, "loss": 0.0004, "step": 1222 }, { "epoch": 0.79, "learning_rate": 1.7310933416222796e-05, "loss": 0.0004, "step": 1223 }, { "epoch": 0.79, "learning_rate": 1.7306196869678386e-05, "loss": 0.0003, "step": 1224 }, { "epoch": 0.79, "learning_rate": 1.730145680453227e-05, "loss": 0.0003, "step": 1225 }, { "epoch": 0.79, "learning_rate": 1.729671322306722e-05, "loss": 0.001, "step": 1226 }, { "epoch": 0.79, "learning_rate": 1.7291966127567702e-05, "loss": 0.0007, "step": 1227 }, { "epoch": 0.79, "learning_rate": 1.728721552031988e-05, "loss": 0.0004, "step": 1228 }, { "epoch": 0.79, "learning_rate": 1.7282461403611602e-05, "loss": 0.0006, "step": 1229 }, { "epoch": 0.79, "learning_rate": 1.7277703779732412e-05, "loss": 0.0004, "step": 1230 }, { "epoch": 0.79, "learning_rate": 1.727294265097354e-05, "loss": 0.0003, "step": 1231 }, { "epoch": 0.79, "learning_rate": 1.7268178019627906e-05, "loss": 0.0005, "step": 1232 }, { "epoch": 0.79, "learning_rate": 1.7263409887990117e-05, "loss": 0.001, "step": 1233 }, { "epoch": 0.79, "learning_rate": 1.725863825835646e-05, "loss": 0.0002, "step": 1234 }, { "epoch": 0.79, "learning_rate": 1.725386313302491e-05, "loss": 0.0005, "step": 1235 }, { "epoch": 0.79, "learning_rate": 1.7249084514295133e-05, "loss": 0.0005, "step": 1236 }, { "epoch": 0.79, "learning_rate": 1.7244302404468463e-05, "loss": 0.0007, "step": 1237 }, { "epoch": 0.8, "learning_rate": 1.7239516805847932e-05, "loss": 0.0003, "step": 1238 }, { "epoch": 0.8, "learning_rate": 1.7234727720738238e-05, "loss": 0.0003, "step": 1239 }, { "epoch": 0.8, "learning_rate": 1.7229935151445762e-05, "loss": 0.0004, "step": 1240 }, { "epoch": 0.8, "learning_rate": 1.7225139100278566e-05, "loss": 0.0002, "step": 1241 }, { "epoch": 0.8, "learning_rate": 1.7220339569546385e-05, "loss": 0.0004, "step": 1242 }, { "epoch": 0.8, "learning_rate": 1.7215536561560635e-05, "loss": 0.0006, "step": 1243 }, { "epoch": 0.8, "learning_rate": 1.7210730078634402e-05, "loss": 0.0002, "step": 1244 }, { "epoch": 0.8, "learning_rate": 1.720592012308245e-05, "loss": 0.0004, "step": 1245 }, { "epoch": 0.8, "learning_rate": 1.7201106697221208e-05, "loss": 0.0006, "step": 1246 }, { "epoch": 0.8, "learning_rate": 1.719628980336878e-05, "loss": 0.0087, "step": 1247 }, { "epoch": 0.8, "learning_rate": 1.7191469443844944e-05, "loss": 0.0013, "step": 1248 }, { "epoch": 0.8, "learning_rate": 1.7186645620971142e-05, "loss": 0.0003, "step": 1249 }, { "epoch": 0.8, "learning_rate": 1.7181818337070488e-05, "loss": 0.0002, "step": 1250 }, { "epoch": 0.8, "learning_rate": 1.717698759446776e-05, "loss": 0.0003, "step": 1251 }, { "epoch": 0.8, "learning_rate": 1.7172153395489396e-05, "loss": 0.0003, "step": 1252 }, { "epoch": 0.81, "learning_rate": 1.7167315742463513e-05, "loss": 0.0005, "step": 1253 }, { "epoch": 0.81, "learning_rate": 1.7162474637719877e-05, "loss": 0.0006, "step": 1254 }, { "epoch": 0.81, "learning_rate": 1.7157630083589926e-05, "loss": 0.0006, "step": 1255 }, { "epoch": 0.81, "learning_rate": 1.7152782082406755e-05, "loss": 0.0002, "step": 1256 }, { "epoch": 0.81, "learning_rate": 1.7147930636505125e-05, "loss": 0.0004, "step": 1257 }, { "epoch": 0.81, "learning_rate": 1.714307574822144e-05, "loss": 0.0005, "step": 1258 }, { "epoch": 0.81, "learning_rate": 1.713821741989378e-05, "loss": 0.0003, "step": 1259 }, { "epoch": 0.81, "learning_rate": 1.713335565386187e-05, "loss": 0.0008, "step": 1260 }, { "epoch": 0.81, "learning_rate": 1.71284904524671e-05, "loss": 0.0003, "step": 1261 }, { "epoch": 0.81, "learning_rate": 1.71236218180525e-05, "loss": 0.0012, "step": 1262 }, { "epoch": 0.81, "learning_rate": 1.711874975296278e-05, "loss": 0.0008, "step": 1263 }, { "epoch": 0.81, "learning_rate": 1.7113874259544262e-05, "loss": 0.0007, "step": 1264 }, { "epoch": 0.81, "learning_rate": 1.710899534014496e-05, "loss": 0.0005, "step": 1265 }, { "epoch": 0.81, "learning_rate": 1.710411299711451e-05, "loss": 0.0003, "step": 1266 }, { "epoch": 0.81, "learning_rate": 1.709922723280421e-05, "loss": 0.0011, "step": 1267 }, { "epoch": 0.81, "learning_rate": 1.7094338049567007e-05, "loss": 0.0003, "step": 1268 }, { "epoch": 0.82, "learning_rate": 1.708944544975748e-05, "loss": 0.0006, "step": 1269 }, { "epoch": 0.82, "learning_rate": 1.7084549435731864e-05, "loss": 0.001, "step": 1270 }, { "epoch": 0.82, "learning_rate": 1.7079650009848047e-05, "loss": 0.0004, "step": 1271 }, { "epoch": 0.82, "learning_rate": 1.707474717446554e-05, "loss": 0.0003, "step": 1272 }, { "epoch": 0.82, "learning_rate": 1.7069840931945517e-05, "loss": 0.0006, "step": 1273 }, { "epoch": 0.82, "learning_rate": 1.706493128465077e-05, "loss": 0.0004, "step": 1274 }, { "epoch": 0.82, "learning_rate": 1.706001823494575e-05, "loss": 0.0003, "step": 1275 }, { "epoch": 0.82, "learning_rate": 1.705510178519654e-05, "loss": 0.0002, "step": 1276 }, { "epoch": 0.82, "learning_rate": 1.7050181937770857e-05, "loss": 0.0002, "step": 1277 }, { "epoch": 0.82, "learning_rate": 1.704525869503806e-05, "loss": 0.0004, "step": 1278 }, { "epoch": 0.82, "learning_rate": 1.7040332059369137e-05, "loss": 0.0002, "step": 1279 }, { "epoch": 0.82, "learning_rate": 1.703540203313672e-05, "loss": 0.0002, "step": 1280 }, { "epoch": 0.82, "learning_rate": 1.703046861871506e-05, "loss": 0.0004, "step": 1281 }, { "epoch": 0.82, "learning_rate": 1.7025531818480052e-05, "loss": 0.0004, "step": 1282 }, { "epoch": 0.82, "learning_rate": 1.702059163480921e-05, "loss": 0.0001, "step": 1283 }, { "epoch": 0.83, "learning_rate": 1.7015648070081695e-05, "loss": 0.0005, "step": 1284 }, { "epoch": 0.83, "learning_rate": 1.7010701126678276e-05, "loss": 0.0005, "step": 1285 }, { "epoch": 0.83, "learning_rate": 1.7005750806981366e-05, "loss": 0.0031, "step": 1286 }, { "epoch": 0.83, "learning_rate": 1.7000797113374984e-05, "loss": 0.0003, "step": 1287 }, { "epoch": 0.83, "learning_rate": 1.69958400482448e-05, "loss": 0.0004, "step": 1288 }, { "epoch": 0.83, "learning_rate": 1.699087961397809e-05, "loss": 0.0002, "step": 1289 }, { "epoch": 0.83, "learning_rate": 1.698591581296375e-05, "loss": 0.0004, "step": 1290 }, { "epoch": 0.83, "learning_rate": 1.698094864759231e-05, "loss": 0.001, "step": 1291 }, { "epoch": 0.83, "learning_rate": 1.697597812025592e-05, "loss": 0.0007, "step": 1292 }, { "epoch": 0.83, "learning_rate": 1.697100423334833e-05, "loss": 0.0003, "step": 1293 }, { "epoch": 0.83, "learning_rate": 1.696602698926493e-05, "loss": 0.0002, "step": 1294 }, { "epoch": 0.83, "learning_rate": 1.6961046390402715e-05, "loss": 0.0002, "step": 1295 }, { "epoch": 0.83, "learning_rate": 1.69560624391603e-05, "loss": 0.0003, "step": 1296 }, { "epoch": 0.83, "learning_rate": 1.695107513793791e-05, "loss": 0.001, "step": 1297 }, { "epoch": 0.83, "learning_rate": 1.6946084489137393e-05, "loss": 0.0003, "step": 1298 }, { "epoch": 0.83, "learning_rate": 1.69410904951622e-05, "loss": 0.0002, "step": 1299 }, { "epoch": 0.84, "learning_rate": 1.693609315841739e-05, "loss": 0.0003, "step": 1300 }, { "epoch": 0.84, "learning_rate": 1.6931092481309643e-05, "loss": 0.0004, "step": 1301 }, { "epoch": 0.84, "learning_rate": 1.692608846624724e-05, "loss": 0.0005, "step": 1302 }, { "epoch": 0.84, "learning_rate": 1.6921081115640076e-05, "loss": 0.0002, "step": 1303 }, { "epoch": 0.84, "learning_rate": 1.691607043189964e-05, "loss": 0.0003, "step": 1304 }, { "epoch": 0.84, "learning_rate": 1.6911056417439044e-05, "loss": 0.0002, "step": 1305 }, { "epoch": 0.84, "learning_rate": 1.690603907467299e-05, "loss": 0.0004, "step": 1306 }, { "epoch": 0.84, "learning_rate": 1.6901018406017785e-05, "loss": 0.0002, "step": 1307 }, { "epoch": 0.84, "learning_rate": 1.689599441389134e-05, "loss": 0.0003, "step": 1308 }, { "epoch": 0.84, "learning_rate": 1.6890967100713167e-05, "loss": 0.0003, "step": 1309 }, { "epoch": 0.84, "learning_rate": 1.6885936468904384e-05, "loss": 0.0003, "step": 1310 }, { "epoch": 0.84, "learning_rate": 1.6880902520887687e-05, "loss": 0.0007, "step": 1311 }, { "epoch": 0.84, "learning_rate": 1.6875865259087395e-05, "loss": 0.0001, "step": 1312 }, { "epoch": 0.84, "learning_rate": 1.6870824685929402e-05, "loss": 0.0004, "step": 1313 }, { "epoch": 0.84, "learning_rate": 1.6865780803841208e-05, "loss": 0.0004, "step": 1314 }, { "epoch": 0.84, "learning_rate": 1.6860733615251898e-05, "loss": 0.0004, "step": 1315 }, { "epoch": 0.85, "learning_rate": 1.6855683122592165e-05, "loss": 0.0002, "step": 1316 }, { "epoch": 0.85, "learning_rate": 1.6850629328294273e-05, "loss": 0.001, "step": 1317 }, { "epoch": 0.85, "learning_rate": 1.6845572234792092e-05, "loss": 0.0007, "step": 1318 }, { "epoch": 0.85, "learning_rate": 1.6840511844521074e-05, "loss": 0.0007, "step": 1319 }, { "epoch": 0.85, "learning_rate": 1.6835448159918255e-05, "loss": 0.0003, "step": 1320 }, { "epoch": 0.85, "learning_rate": 1.6830381183422266e-05, "loss": 0.0004, "step": 1321 }, { "epoch": 0.85, "learning_rate": 1.6825310917473322e-05, "loss": 0.0004, "step": 1322 }, { "epoch": 0.85, "learning_rate": 1.6820237364513215e-05, "loss": 0.0001, "step": 1323 }, { "epoch": 0.85, "learning_rate": 1.6815160526985324e-05, "loss": 0.0001, "step": 1324 }, { "epoch": 0.85, "learning_rate": 1.6810080407334616e-05, "loss": 0.0003, "step": 1325 }, { "epoch": 0.85, "learning_rate": 1.6804997008007623e-05, "loss": 0.0003, "step": 1326 }, { "epoch": 0.85, "learning_rate": 1.679991033145248e-05, "loss": 0.0003, "step": 1327 }, { "epoch": 0.85, "learning_rate": 1.6794820380118878e-05, "loss": 0.0004, "step": 1328 }, { "epoch": 0.85, "learning_rate": 1.678972715645809e-05, "loss": 0.0003, "step": 1329 }, { "epoch": 0.85, "learning_rate": 1.6784630662922978e-05, "loss": 0.0004, "step": 1330 }, { "epoch": 0.86, "learning_rate": 1.677953090196797e-05, "loss": 0.0002, "step": 1331 }, { "epoch": 0.86, "learning_rate": 1.6774427876049054e-05, "loss": 0.0006, "step": 1332 }, { "epoch": 0.86, "learning_rate": 1.6769321587623818e-05, "loss": 0.0015, "step": 1333 }, { "epoch": 0.86, "learning_rate": 1.6764212039151402e-05, "loss": 0.0002, "step": 1334 }, { "epoch": 0.86, "learning_rate": 1.6759099233092517e-05, "loss": 0.0005, "step": 1335 }, { "epoch": 0.86, "learning_rate": 1.6753983171909453e-05, "loss": 0.0002, "step": 1336 }, { "epoch": 0.86, "learning_rate": 1.6748863858066053e-05, "loss": 0.0004, "step": 1337 }, { "epoch": 0.86, "learning_rate": 1.674374129402774e-05, "loss": 0.0006, "step": 1338 }, { "epoch": 0.86, "learning_rate": 1.67386154822615e-05, "loss": 0.0004, "step": 1339 }, { "epoch": 0.86, "learning_rate": 1.6733486425235873e-05, "loss": 0.0005, "step": 1340 }, { "epoch": 0.86, "learning_rate": 1.6728354125420973e-05, "loss": 0.0004, "step": 1341 }, { "epoch": 0.86, "learning_rate": 1.672321858528847e-05, "loss": 0.0002, "step": 1342 }, { "epoch": 0.86, "learning_rate": 1.6718079807311592e-05, "loss": 0.0002, "step": 1343 }, { "epoch": 0.86, "learning_rate": 1.671293779396514e-05, "loss": 0.0005, "step": 1344 }, { "epoch": 0.86, "learning_rate": 1.6707792547725457e-05, "loss": 0.0004, "step": 1345 }, { "epoch": 0.86, "learning_rate": 1.670264407107045e-05, "loss": 0.0002, "step": 1346 }, { "epoch": 0.87, "learning_rate": 1.669749236647958e-05, "loss": 0.0003, "step": 1347 }, { "epoch": 0.87, "learning_rate": 1.669233743643387e-05, "loss": 0.0002, "step": 1348 }, { "epoch": 0.87, "learning_rate": 1.6687179283415883e-05, "loss": 0.0003, "step": 1349 }, { "epoch": 0.87, "learning_rate": 1.668201790990974e-05, "loss": 0.0004, "step": 1350 }, { "epoch": 0.87, "learning_rate": 1.6676853318401126e-05, "loss": 0.0003, "step": 1351 }, { "epoch": 0.87, "learning_rate": 1.667168551137725e-05, "loss": 0.0004, "step": 1352 }, { "epoch": 0.87, "learning_rate": 1.6666514491326894e-05, "loss": 0.0004, "step": 1353 }, { "epoch": 0.87, "learning_rate": 1.666134026074037e-05, "loss": 0.0002, "step": 1354 }, { "epoch": 0.87, "learning_rate": 1.6656162822109542e-05, "loss": 0.0002, "step": 1355 }, { "epoch": 0.87, "learning_rate": 1.665098217792783e-05, "loss": 0.0002, "step": 1356 }, { "epoch": 0.87, "learning_rate": 1.6645798330690177e-05, "loss": 0.0002, "step": 1357 }, { "epoch": 0.87, "learning_rate": 1.6640611282893085e-05, "loss": 0.0003, "step": 1358 }, { "epoch": 0.87, "learning_rate": 1.663542103703459e-05, "loss": 0.0002, "step": 1359 }, { "epoch": 0.87, "learning_rate": 1.663022759561427e-05, "loss": 0.0003, "step": 1360 }, { "epoch": 0.87, "learning_rate": 1.6625030961133242e-05, "loss": 0.0006, "step": 1361 }, { "epoch": 0.88, "learning_rate": 1.661983113609416e-05, "loss": 0.0004, "step": 1362 }, { "epoch": 0.88, "learning_rate": 1.6614628123001217e-05, "loss": 0.0002, "step": 1363 }, { "epoch": 0.88, "learning_rate": 1.6609421924360135e-05, "loss": 0.0001, "step": 1364 }, { "epoch": 0.88, "learning_rate": 1.660421254267818e-05, "loss": 0.0003, "step": 1365 }, { "epoch": 0.88, "learning_rate": 1.6598999980464142e-05, "loss": 0.0001, "step": 1366 }, { "epoch": 0.88, "learning_rate": 1.659378424022835e-05, "loss": 0.0003, "step": 1367 }, { "epoch": 0.88, "learning_rate": 1.6588565324482655e-05, "loss": 0.0003, "step": 1368 }, { "epoch": 0.88, "learning_rate": 1.6583343235740445e-05, "loss": 0.0002, "step": 1369 }, { "epoch": 0.88, "learning_rate": 1.6578117976516634e-05, "loss": 0.0002, "step": 1370 }, { "epoch": 0.88, "learning_rate": 1.6572889549327655e-05, "loss": 0.0006, "step": 1371 }, { "epoch": 0.88, "learning_rate": 1.6567657956691486e-05, "loss": 0.0001, "step": 1372 }, { "epoch": 0.88, "learning_rate": 1.6562423201127603e-05, "loss": 0.0002, "step": 1373 }, { "epoch": 0.88, "learning_rate": 1.655718528515703e-05, "loss": 0.0003, "step": 1374 }, { "epoch": 0.88, "learning_rate": 1.6551944211302296e-05, "loss": 0.0007, "step": 1375 }, { "epoch": 0.88, "learning_rate": 1.6546699982087464e-05, "loss": 0.0003, "step": 1376 }, { "epoch": 0.88, "learning_rate": 1.6541452600038105e-05, "loss": 0.0003, "step": 1377 }, { "epoch": 0.89, "learning_rate": 1.6536202067681312e-05, "loss": 0.0002, "step": 1378 }, { "epoch": 0.89, "learning_rate": 1.65309483875457e-05, "loss": 0.0003, "step": 1379 }, { "epoch": 0.89, "learning_rate": 1.6525691562161393e-05, "loss": 0.0003, "step": 1380 }, { "epoch": 0.89, "learning_rate": 1.6520431594060035e-05, "loss": 0.0003, "step": 1381 }, { "epoch": 0.89, "learning_rate": 1.651516848577478e-05, "loss": 0.0002, "step": 1382 }, { "epoch": 0.89, "learning_rate": 1.6509902239840298e-05, "loss": 0.0003, "step": 1383 }, { "epoch": 0.89, "learning_rate": 1.6504632858792772e-05, "loss": 0.0002, "step": 1384 }, { "epoch": 0.89, "learning_rate": 1.6499360345169877e-05, "loss": 0.0008, "step": 1385 }, { "epoch": 0.89, "learning_rate": 1.6494084701510818e-05, "loss": 0.0002, "step": 1386 }, { "epoch": 0.89, "learning_rate": 1.6488805930356307e-05, "loss": 0.0002, "step": 1387 }, { "epoch": 0.89, "learning_rate": 1.6483524034248545e-05, "loss": 0.0003, "step": 1388 }, { "epoch": 0.89, "learning_rate": 1.647823901573125e-05, "loss": 0.0003, "step": 1389 }, { "epoch": 0.89, "learning_rate": 1.6472950877349645e-05, "loss": 0.0009, "step": 1390 }, { "epoch": 0.89, "learning_rate": 1.646765962165045e-05, "loss": 0.0004, "step": 1391 }, { "epoch": 0.89, "learning_rate": 1.6462365251181885e-05, "loss": 0.0001, "step": 1392 }, { "epoch": 0.9, "learning_rate": 1.645706776849368e-05, "loss": 0.0002, "step": 1393 }, { "epoch": 0.9, "learning_rate": 1.645176717613705e-05, "loss": 0.0002, "step": 1394 }, { "epoch": 0.9, "learning_rate": 1.644646347666472e-05, "loss": 0.0002, "step": 1395 }, { "epoch": 0.9, "learning_rate": 1.64411566726309e-05, "loss": 0.0002, "step": 1396 }, { "epoch": 0.9, "learning_rate": 1.643584676659131e-05, "loss": 0.0001, "step": 1397 }, { "epoch": 0.9, "learning_rate": 1.6430533761103154e-05, "loss": 0.0003, "step": 1398 }, { "epoch": 0.9, "learning_rate": 1.6425217658725122e-05, "loss": 0.0002, "step": 1399 }, { "epoch": 0.9, "learning_rate": 1.641989846201741e-05, "loss": 0.0001, "step": 1400 }, { "epoch": 0.9, "learning_rate": 1.6414576173541697e-05, "loss": 0.0003, "step": 1401 }, { "epoch": 0.9, "learning_rate": 1.6409250795861148e-05, "loss": 0.0002, "step": 1402 }, { "epoch": 0.9, "learning_rate": 1.640392233154042e-05, "loss": 0.0001, "step": 1403 }, { "epoch": 0.9, "learning_rate": 1.6398590783145662e-05, "loss": 0.0003, "step": 1404 }, { "epoch": 0.9, "learning_rate": 1.6393256153244493e-05, "loss": 0.0001, "step": 1405 }, { "epoch": 0.9, "learning_rate": 1.638791844440603e-05, "loss": 0.0002, "step": 1406 }, { "epoch": 0.9, "learning_rate": 1.6382577659200865e-05, "loss": 0.0002, "step": 1407 }, { "epoch": 0.9, "learning_rate": 1.6377233800201075e-05, "loss": 0.0003, "step": 1408 }, { "epoch": 0.91, "learning_rate": 1.6371886869980222e-05, "loss": 0.0003, "step": 1409 }, { "epoch": 0.91, "learning_rate": 1.6366536871113333e-05, "loss": 0.0003, "step": 1410 }, { "epoch": 0.91, "learning_rate": 1.6361183806176925e-05, "loss": 0.0002, "step": 1411 }, { "epoch": 0.91, "learning_rate": 1.6355827677748992e-05, "loss": 0.0002, "step": 1412 }, { "epoch": 0.91, "learning_rate": 1.6350468488409e-05, "loss": 0.0006, "step": 1413 }, { "epoch": 0.91, "learning_rate": 1.6345106240737877e-05, "loss": 0.0002, "step": 1414 }, { "epoch": 0.91, "learning_rate": 1.633974093731805e-05, "loss": 0.0002, "step": 1415 }, { "epoch": 0.91, "learning_rate": 1.6334372580733392e-05, "loss": 0.0012, "step": 1416 }, { "epoch": 0.91, "learning_rate": 1.6329001173569264e-05, "loss": 0.0005, "step": 1417 }, { "epoch": 0.91, "learning_rate": 1.6323626718412487e-05, "loss": 0.0002, "step": 1418 }, { "epoch": 0.91, "learning_rate": 1.6318249217851355e-05, "loss": 0.0001, "step": 1419 }, { "epoch": 0.91, "learning_rate": 1.6312868674475622e-05, "loss": 0.0003, "step": 1420 }, { "epoch": 0.91, "learning_rate": 1.6307485090876513e-05, "loss": 0.0002, "step": 1421 }, { "epoch": 0.91, "learning_rate": 1.630209846964672e-05, "loss": 0.0003, "step": 1422 }, { "epoch": 0.91, "learning_rate": 1.6296708813380387e-05, "loss": 0.0002, "step": 1423 }, { "epoch": 0.92, "learning_rate": 1.629131612467313e-05, "loss": 0.0002, "step": 1424 }, { "epoch": 0.92, "learning_rate": 1.628592040612202e-05, "loss": 0.0005, "step": 1425 }, { "epoch": 0.92, "learning_rate": 1.628052166032559e-05, "loss": 0.0009, "step": 1426 }, { "epoch": 0.92, "learning_rate": 1.627511988988383e-05, "loss": 0.0001, "step": 1427 }, { "epoch": 0.92, "learning_rate": 1.6269715097398183e-05, "loss": 0.0003, "step": 1428 }, { "epoch": 0.92, "learning_rate": 1.6264307285471555e-05, "loss": 0.0003, "step": 1429 }, { "epoch": 0.92, "learning_rate": 1.6258896456708304e-05, "loss": 0.0002, "step": 1430 }, { "epoch": 0.92, "learning_rate": 1.6253482613714226e-05, "loss": 0.0002, "step": 1431 }, { "epoch": 0.92, "learning_rate": 1.6248065759096594e-05, "loss": 0.0001, "step": 1432 }, { "epoch": 0.92, "learning_rate": 1.6242645895464115e-05, "loss": 0.0003, "step": 1433 }, { "epoch": 0.92, "learning_rate": 1.623722302542695e-05, "loss": 0.0001, "step": 1434 }, { "epoch": 0.92, "learning_rate": 1.62317971515967e-05, "loss": 0.0002, "step": 1435 }, { "epoch": 0.92, "learning_rate": 1.622636827658643e-05, "loss": 0.0007, "step": 1436 }, { "epoch": 0.92, "learning_rate": 1.622093640301063e-05, "loss": 0.0002, "step": 1437 }, { "epoch": 0.92, "learning_rate": 1.621550153348525e-05, "loss": 0.0004, "step": 1438 }, { "epoch": 0.92, "learning_rate": 1.6210063670627675e-05, "loss": 0.0005, "step": 1439 }, { "epoch": 0.93, "learning_rate": 1.620462281705673e-05, "loss": 0.0003, "step": 1440 }, { "epoch": 0.93, "learning_rate": 1.6199178975392685e-05, "loss": 0.0004, "step": 1441 }, { "epoch": 0.93, "learning_rate": 1.619373214825725e-05, "loss": 0.0002, "step": 1442 }, { "epoch": 0.93, "learning_rate": 1.6188282338273568e-05, "loss": 0.0003, "step": 1443 }, { "epoch": 0.93, "learning_rate": 1.618282954806622e-05, "loss": 0.0003, "step": 1444 }, { "epoch": 0.93, "learning_rate": 1.6177373780261226e-05, "loss": 0.0001, "step": 1445 }, { "epoch": 0.93, "learning_rate": 1.617191503748603e-05, "loss": 0.0003, "step": 1446 }, { "epoch": 0.93, "learning_rate": 1.6166453322369525e-05, "loss": 0.0004, "step": 1447 }, { "epoch": 0.93, "learning_rate": 1.616098863754202e-05, "loss": 0.0002, "step": 1448 }, { "epoch": 0.93, "learning_rate": 1.6155520985635265e-05, "loss": 0.0002, "step": 1449 }, { "epoch": 0.93, "learning_rate": 1.6150050369282427e-05, "loss": 0.0002, "step": 1450 }, { "epoch": 0.93, "learning_rate": 1.6144576791118116e-05, "loss": 0.0002, "step": 1451 }, { "epoch": 0.93, "learning_rate": 1.6139100253778355e-05, "loss": 0.0001, "step": 1452 }, { "epoch": 0.93, "learning_rate": 1.6133620759900598e-05, "loss": 0.0003, "step": 1453 }, { "epoch": 0.93, "learning_rate": 1.6128138312123726e-05, "loss": 0.0001, "step": 1454 }, { "epoch": 0.93, "learning_rate": 1.6122652913088032e-05, "loss": 0.0002, "step": 1455 }, { "epoch": 0.94, "learning_rate": 1.6117164565435238e-05, "loss": 0.0001, "step": 1456 }, { "epoch": 0.94, "learning_rate": 1.6111673271808495e-05, "loss": 0.0005, "step": 1457 }, { "epoch": 0.94, "learning_rate": 1.6106179034852354e-05, "loss": 0.0001, "step": 1458 }, { "epoch": 0.94, "learning_rate": 1.6100681857212794e-05, "loss": 0.0004, "step": 1459 }, { "epoch": 0.94, "learning_rate": 1.609518174153721e-05, "loss": 0.0002, "step": 1460 }, { "epoch": 0.94, "learning_rate": 1.6089678690474407e-05, "loss": 0.0001, "step": 1461 }, { "epoch": 0.94, "learning_rate": 1.6084172706674612e-05, "loss": 0.0003, "step": 1462 }, { "epoch": 0.94, "learning_rate": 1.6078663792789453e-05, "loss": 0.0002, "step": 1463 }, { "epoch": 0.94, "learning_rate": 1.607315195147198e-05, "loss": 0.0002, "step": 1464 }, { "epoch": 0.94, "learning_rate": 1.6067637185376653e-05, "loss": 0.0005, "step": 1465 }, { "epoch": 0.94, "learning_rate": 1.606211949715933e-05, "loss": 0.0002, "step": 1466 }, { "epoch": 0.94, "learning_rate": 1.605659888947728e-05, "loss": 0.0002, "step": 1467 }, { "epoch": 0.94, "learning_rate": 1.6051075364989185e-05, "loss": 0.0002, "step": 1468 }, { "epoch": 0.94, "learning_rate": 1.604554892635513e-05, "loss": 0.0006, "step": 1469 }, { "epoch": 0.94, "learning_rate": 1.6040019576236593e-05, "loss": 0.0001, "step": 1470 }, { "epoch": 0.95, "learning_rate": 1.6034487317296463e-05, "loss": 0.0004, "step": 1471 }, { "epoch": 0.95, "learning_rate": 1.602895215219903e-05, "loss": 0.0002, "step": 1472 }, { "epoch": 0.95, "learning_rate": 1.6023414083609987e-05, "loss": 0.0001, "step": 1473 }, { "epoch": 0.95, "learning_rate": 1.601787311419641e-05, "loss": 0.0001, "step": 1474 }, { "epoch": 0.95, "learning_rate": 1.6012329246626794e-05, "loss": 0.0002, "step": 1475 }, { "epoch": 0.95, "learning_rate": 1.6006782483571006e-05, "loss": 0.0012, "step": 1476 }, { "epoch": 0.95, "learning_rate": 1.600123282770033e-05, "loss": 0.0001, "step": 1477 }, { "epoch": 0.95, "learning_rate": 1.5995680281687426e-05, "loss": 0.0003, "step": 1478 }, { "epoch": 0.95, "learning_rate": 1.5990124848206356e-05, "loss": 0.0001, "step": 1479 }, { "epoch": 0.95, "learning_rate": 1.5984566529932563e-05, "loss": 0.0002, "step": 1480 }, { "epoch": 0.95, "learning_rate": 1.5979005329542892e-05, "loss": 0.0004, "step": 1481 }, { "epoch": 0.95, "learning_rate": 1.597344124971557e-05, "loss": 0.0001, "step": 1482 }, { "epoch": 0.95, "learning_rate": 1.5967874293130204e-05, "loss": 0.0003, "step": 1483 }, { "epoch": 0.95, "learning_rate": 1.5962304462467798e-05, "loss": 0.0002, "step": 1484 }, { "epoch": 0.95, "learning_rate": 1.595673176041073e-05, "loss": 0.0001, "step": 1485 }, { "epoch": 0.95, "learning_rate": 1.5951156189642773e-05, "loss": 0.0005, "step": 1486 }, { "epoch": 0.96, "learning_rate": 1.5945577752849066e-05, "loss": 0.0002, "step": 1487 }, { "epoch": 0.96, "learning_rate": 1.5939996452716138e-05, "loss": 0.0002, "step": 1488 }, { "epoch": 0.96, "learning_rate": 1.59344122919319e-05, "loss": 0.0003, "step": 1489 }, { "epoch": 0.96, "learning_rate": 1.592882527318563e-05, "loss": 0.0001, "step": 1490 }, { "epoch": 0.96, "learning_rate": 1.5923235399167993e-05, "loss": 0.0002, "step": 1491 }, { "epoch": 0.96, "learning_rate": 1.591764267257102e-05, "loss": 0.0003, "step": 1492 }, { "epoch": 0.96, "learning_rate": 1.5912047096088124e-05, "loss": 0.0006, "step": 1493 }, { "epoch": 0.96, "learning_rate": 1.5906448672414085e-05, "loss": 0.0004, "step": 1494 }, { "epoch": 0.96, "learning_rate": 1.5900847404245055e-05, "loss": 0.0004, "step": 1495 }, { "epoch": 0.96, "learning_rate": 1.5895243294278556e-05, "loss": 0.0009, "step": 1496 }, { "epoch": 0.96, "learning_rate": 1.588963634521348e-05, "loss": 0.0003, "step": 1497 }, { "epoch": 0.96, "learning_rate": 1.5884026559750092e-05, "loss": 0.0002, "step": 1498 }, { "epoch": 0.96, "learning_rate": 1.587841394059001e-05, "loss": 0.0001, "step": 1499 }, { "epoch": 0.96, "learning_rate": 1.587279849043622e-05, "loss": 0.0005, "step": 1500 }, { "epoch": 0.96, "learning_rate": 1.5867180211993076e-05, "loss": 0.0005, "step": 1501 }, { "epoch": 0.97, "learning_rate": 1.58615591079663e-05, "loss": 0.0001, "step": 1502 }, { "epoch": 0.97, "learning_rate": 1.585593518106296e-05, "loss": 0.0004, "step": 1503 }, { "epoch": 0.97, "learning_rate": 1.5850308433991498e-05, "loss": 0.0009, "step": 1504 }, { "epoch": 0.97, "learning_rate": 1.58446788694617e-05, "loss": 0.0003, "step": 1505 }, { "epoch": 0.97, "learning_rate": 1.5839046490184714e-05, "loss": 0.0003, "step": 1506 }, { "epoch": 0.97, "learning_rate": 1.5833411298873058e-05, "loss": 0.0001, "step": 1507 }, { "epoch": 0.97, "learning_rate": 1.582777329824058e-05, "loss": 0.0002, "step": 1508 }, { "epoch": 0.97, "learning_rate": 1.58221324910025e-05, "loss": 0.0002, "step": 1509 }, { "epoch": 0.97, "learning_rate": 1.5816488879875373e-05, "loss": 0.0001, "step": 1510 }, { "epoch": 0.97, "learning_rate": 1.5810842467577125e-05, "loss": 0.0003, "step": 1511 }, { "epoch": 0.97, "learning_rate": 1.5805193256827013e-05, "loss": 0.0006, "step": 1512 }, { "epoch": 0.97, "learning_rate": 1.579954125034565e-05, "loss": 0.0002, "step": 1513 }, { "epoch": 0.97, "learning_rate": 1.579388645085499e-05, "loss": 0.0003, "step": 1514 }, { "epoch": 0.97, "learning_rate": 1.5788228861078345e-05, "loss": 0.0002, "step": 1515 }, { "epoch": 0.97, "learning_rate": 1.5782568483740356e-05, "loss": 0.0002, "step": 1516 }, { "epoch": 0.97, "learning_rate": 1.577690532156701e-05, "loss": 0.0002, "step": 1517 }, { "epoch": 0.98, "learning_rate": 1.577123937728564e-05, "loss": 0.0003, "step": 1518 }, { "epoch": 0.98, "learning_rate": 1.576557065362492e-05, "loss": 0.0001, "step": 1519 }, { "epoch": 0.98, "learning_rate": 1.575989915331485e-05, "loss": 0.0002, "step": 1520 }, { "epoch": 0.98, "learning_rate": 1.575422487908678e-05, "loss": 0.0003, "step": 1521 }, { "epoch": 0.98, "learning_rate": 1.5748547833673393e-05, "loss": 0.0004, "step": 1522 }, { "epoch": 0.98, "learning_rate": 1.5742868019808706e-05, "loss": 0.0003, "step": 1523 }, { "epoch": 0.98, "learning_rate": 1.5737185440228066e-05, "loss": 0.0001, "step": 1524 }, { "epoch": 0.98, "learning_rate": 1.5731500097668154e-05, "loss": 0.0003, "step": 1525 }, { "epoch": 0.98, "learning_rate": 1.5725811994866988e-05, "loss": 0.0006, "step": 1526 }, { "epoch": 0.98, "learning_rate": 1.5720121134563902e-05, "loss": 0.0002, "step": 1527 }, { "epoch": 0.98, "learning_rate": 1.5714427519499568e-05, "loss": 0.0004, "step": 1528 }, { "epoch": 0.98, "learning_rate": 1.5708731152415985e-05, "loss": 0.0002, "step": 1529 }, { "epoch": 0.98, "learning_rate": 1.5703032036056474e-05, "loss": 0.0004, "step": 1530 }, { "epoch": 0.98, "learning_rate": 1.5697330173165676e-05, "loss": 0.0002, "step": 1531 }, { "epoch": 0.98, "learning_rate": 1.5691625566489563e-05, "loss": 0.0005, "step": 1532 }, { "epoch": 0.99, "learning_rate": 1.5685918218775428e-05, "loss": 0.0003, "step": 1533 }, { "epoch": 0.99, "learning_rate": 1.568020813277187e-05, "loss": 0.0001, "step": 1534 }, { "epoch": 0.99, "learning_rate": 1.567449531122883e-05, "loss": 0.0002, "step": 1535 }, { "epoch": 0.99, "learning_rate": 1.5668779756897543e-05, "loss": 0.0008, "step": 1536 }, { "epoch": 0.99, "learning_rate": 1.566306147253058e-05, "loss": 0.0002, "step": 1537 }, { "epoch": 0.99, "learning_rate": 1.565734046088181e-05, "loss": 0.0002, "step": 1538 }, { "epoch": 0.99, "learning_rate": 1.5651616724706426e-05, "loss": 0.0002, "step": 1539 }, { "epoch": 0.99, "learning_rate": 1.5645890266760933e-05, "loss": 0.0002, "step": 1540 }, { "epoch": 0.99, "learning_rate": 1.5640161089803138e-05, "loss": 0.0002, "step": 1541 }, { "epoch": 0.99, "learning_rate": 1.5634429196592173e-05, "loss": 0.0003, "step": 1542 }, { "epoch": 0.99, "learning_rate": 1.5628694589888458e-05, "loss": 0.0003, "step": 1543 }, { "epoch": 0.99, "learning_rate": 1.562295727245374e-05, "loss": 0.0001, "step": 1544 }, { "epoch": 0.99, "learning_rate": 1.5617217247051054e-05, "loss": 0.0004, "step": 1545 }, { "epoch": 0.99, "learning_rate": 1.5611474516444746e-05, "loss": 0.0001, "step": 1546 }, { "epoch": 0.99, "learning_rate": 1.5605729083400475e-05, "loss": 0.0001, "step": 1547 }, { "epoch": 0.99, "learning_rate": 1.5599980950685184e-05, "loss": 0.0001, "step": 1548 }, { "epoch": 1.0, "learning_rate": 1.559423012106713e-05, "loss": 0.0003, "step": 1549 }, { "epoch": 1.0, "learning_rate": 1.558847659731586e-05, "loss": 0.0002, "step": 1550 }, { "epoch": 1.0, "learning_rate": 1.5582720382202222e-05, "loss": 0.0002, "step": 1551 }, { "epoch": 1.0, "learning_rate": 1.557696147849836e-05, "loss": 0.0002, "step": 1552 }, { "epoch": 1.0, "learning_rate": 1.5571199888977712e-05, "loss": 0.0002, "step": 1553 }, { "epoch": 1.0, "learning_rate": 1.556543561641501e-05, "loss": 0.0003, "step": 1554 }, { "epoch": 1.0, "learning_rate": 1.555966866358628e-05, "loss": 0.0003, "step": 1555 }, { "epoch": 1.0, "learning_rate": 1.5553899033268836e-05, "loss": 0.0006, "step": 1556 }, { "epoch": 1.0, "learning_rate": 1.554812672824128e-05, "loss": 0.0001, "step": 1557 }, { "epoch": 1.0, "learning_rate": 1.5542351751283507e-05, "loss": 0.0001, "step": 1558 }, { "epoch": 1.0, "learning_rate": 1.55365741051767e-05, "loss": 0.0001, "step": 1559 }, { "epoch": 1.0, "learning_rate": 1.553079379270332e-05, "loss": 0.0001, "step": 1560 }, { "epoch": 1.0, "learning_rate": 1.5525010816647113e-05, "loss": 0.0002, "step": 1561 }, { "epoch": 1.0, "learning_rate": 1.5519225179793115e-05, "loss": 0.0001, "step": 1562 }, { "epoch": 1.0, "learning_rate": 1.5513436884927634e-05, "loss": 0.0002, "step": 1563 }, { "epoch": 1.0, "learning_rate": 1.550764593483827e-05, "loss": 0.0001, "step": 1564 }, { "epoch": 1.01, "learning_rate": 1.5501852332313895e-05, "loss": 0.0002, "step": 1565 }, { "epoch": 1.01, "learning_rate": 1.5496056080144647e-05, "loss": 0.001, "step": 1566 }, { "epoch": 1.01, "learning_rate": 1.5490257181121965e-05, "loss": 0.0002, "step": 1567 }, { "epoch": 1.01, "learning_rate": 1.548445563803854e-05, "loss": 0.0001, "step": 1568 }, { "epoch": 1.01, "learning_rate": 1.5478651453688347e-05, "loss": 0.0001, "step": 1569 }, { "epoch": 1.01, "learning_rate": 1.5472844630866638e-05, "loss": 0.0002, "step": 1570 }, { "epoch": 1.01, "learning_rate": 1.546703517236992e-05, "loss": 0.0001, "step": 1571 }, { "epoch": 1.01, "learning_rate": 1.546122308099598e-05, "loss": 0.0003, "step": 1572 }, { "epoch": 1.01, "learning_rate": 1.5455408359543876e-05, "loss": 0.0002, "step": 1573 }, { "epoch": 1.01, "learning_rate": 1.5449591010813928e-05, "loss": 0.0001, "step": 1574 }, { "epoch": 1.01, "learning_rate": 1.5443771037607717e-05, "loss": 0.0002, "step": 1575 }, { "epoch": 1.01, "learning_rate": 1.5437948442728097e-05, "loss": 0.0001, "step": 1576 }, { "epoch": 1.01, "learning_rate": 1.543212322897917e-05, "loss": 0.0001, "step": 1577 }, { "epoch": 1.01, "learning_rate": 1.542629539916632e-05, "loss": 0.0002, "step": 1578 }, { "epoch": 1.01, "learning_rate": 1.5420464956096175e-05, "loss": 0.0002, "step": 1579 }, { "epoch": 1.02, "learning_rate": 1.5414631902576626e-05, "loss": 0.0001, "step": 1580 }, { "epoch": 1.02, "learning_rate": 1.540879624141682e-05, "loss": 0.0002, "step": 1581 }, { "epoch": 1.02, "learning_rate": 1.5402957975427167e-05, "loss": 0.0002, "step": 1582 }, { "epoch": 1.02, "learning_rate": 1.539711710741932e-05, "loss": 0.0001, "step": 1583 }, { "epoch": 1.02, "learning_rate": 1.5391273640206193e-05, "loss": 0.0002, "step": 1584 }, { "epoch": 1.02, "learning_rate": 1.5385427576601952e-05, "loss": 0.0001, "step": 1585 }, { "epoch": 1.02, "learning_rate": 1.537957891942201e-05, "loss": 0.0, "step": 1586 }, { "epoch": 1.02, "learning_rate": 1.5373727671483026e-05, "loss": 0.0001, "step": 1587 }, { "epoch": 1.02, "learning_rate": 1.5367873835602917e-05, "loss": 0.0001, "step": 1588 }, { "epoch": 1.02, "learning_rate": 1.5362017414600835e-05, "loss": 0.0001, "step": 1589 }, { "epoch": 1.02, "learning_rate": 1.5356158411297185e-05, "loss": 0.0001, "step": 1590 }, { "epoch": 1.02, "learning_rate": 1.5350296828513605e-05, "loss": 0.0001, "step": 1591 }, { "epoch": 1.02, "learning_rate": 1.5344432669072998e-05, "loss": 0.0001, "step": 1592 }, { "epoch": 1.02, "learning_rate": 1.5338565935799483e-05, "loss": 0.0, "step": 1593 }, { "epoch": 1.02, "learning_rate": 1.533269663151843e-05, "loss": 0.0001, "step": 1594 }, { "epoch": 1.02, "learning_rate": 1.5326824759056442e-05, "loss": 0.0001, "step": 1595 }, { "epoch": 1.03, "learning_rate": 1.5320950321241366e-05, "loss": 0.0003, "step": 1596 }, { "epoch": 1.03, "learning_rate": 1.5315073320902286e-05, "loss": 0.0001, "step": 1597 }, { "epoch": 1.03, "learning_rate": 1.5309193760869507e-05, "loss": 0.0001, "step": 1598 }, { "epoch": 1.03, "learning_rate": 1.530331164397458e-05, "loss": 0.0001, "step": 1599 }, { "epoch": 1.03, "learning_rate": 1.5297426973050272e-05, "loss": 0.0002, "step": 1600 }, { "epoch": 1.03, "learning_rate": 1.5291539750930606e-05, "loss": 0.0001, "step": 1601 }, { "epoch": 1.03, "learning_rate": 1.5285649980450803e-05, "loss": 0.0001, "step": 1602 }, { "epoch": 1.03, "learning_rate": 1.5279757664447336e-05, "loss": 0.0001, "step": 1603 }, { "epoch": 1.03, "learning_rate": 1.5273862805757888e-05, "loss": 0.0001, "step": 1604 }, { "epoch": 1.03, "learning_rate": 1.5267965407221373e-05, "loss": 0.0002, "step": 1605 }, { "epoch": 1.03, "learning_rate": 1.5262065471677932e-05, "loss": 0.0001, "step": 1606 }, { "epoch": 1.03, "learning_rate": 1.5256163001968918e-05, "loss": 0.0001, "step": 1607 }, { "epoch": 1.03, "learning_rate": 1.525025800093691e-05, "loss": 0.0001, "step": 1608 }, { "epoch": 1.03, "learning_rate": 1.5244350471425708e-05, "loss": 0.0002, "step": 1609 }, { "epoch": 1.03, "learning_rate": 1.5238440416280331e-05, "loss": 0.0001, "step": 1610 }, { "epoch": 1.04, "learning_rate": 1.5232527838347001e-05, "loss": 0.0001, "step": 1611 }, { "epoch": 1.04, "learning_rate": 1.5226612740473177e-05, "loss": 0.0002, "step": 1612 }, { "epoch": 1.04, "learning_rate": 1.5220695125507506e-05, "loss": 0.0001, "step": 1613 }, { "epoch": 1.04, "learning_rate": 1.5214774996299872e-05, "loss": 0.0001, "step": 1614 }, { "epoch": 1.04, "learning_rate": 1.5208852355701355e-05, "loss": 0.0002, "step": 1615 }, { "epoch": 1.04, "learning_rate": 1.5202927206564245e-05, "loss": 0.0001, "step": 1616 }, { "epoch": 1.04, "learning_rate": 1.5196999551742045e-05, "loss": 0.0001, "step": 1617 }, { "epoch": 1.04, "learning_rate": 1.5191069394089461e-05, "loss": 0.0001, "step": 1618 }, { "epoch": 1.04, "learning_rate": 1.5185136736462411e-05, "loss": 0.0001, "step": 1619 }, { "epoch": 1.04, "learning_rate": 1.5179201581718008e-05, "loss": 0.0002, "step": 1620 }, { "epoch": 1.04, "learning_rate": 1.517326393271457e-05, "loss": 0.0, "step": 1621 }, { "epoch": 1.04, "learning_rate": 1.5167323792311623e-05, "loss": 0.0002, "step": 1622 }, { "epoch": 1.04, "learning_rate": 1.5161381163369882e-05, "loss": 0.0002, "step": 1623 }, { "epoch": 1.04, "learning_rate": 1.5155436048751268e-05, "loss": 0.0004, "step": 1624 }, { "epoch": 1.04, "learning_rate": 1.51494884513189e-05, "loss": 0.0003, "step": 1625 }, { "epoch": 1.04, "learning_rate": 1.5143538373937088e-05, "loss": 0.0, "step": 1626 }, { "epoch": 1.05, "learning_rate": 1.5137585819471338e-05, "loss": 0.0001, "step": 1627 }, { "epoch": 1.05, "learning_rate": 1.513163079078835e-05, "loss": 0.0001, "step": 1628 }, { "epoch": 1.05, "learning_rate": 1.5125673290756012e-05, "loss": 0.0001, "step": 1629 }, { "epoch": 1.05, "learning_rate": 1.5119713322243406e-05, "loss": 0.0001, "step": 1630 }, { "epoch": 1.05, "learning_rate": 1.5113750888120804e-05, "loss": 0.0003, "step": 1631 }, { "epoch": 1.05, "learning_rate": 1.5107785991259663e-05, "loss": 0.0001, "step": 1632 }, { "epoch": 1.05, "learning_rate": 1.510181863453262e-05, "loss": 0.0, "step": 1633 }, { "epoch": 1.05, "learning_rate": 1.5095848820813512e-05, "loss": 0.0005, "step": 1634 }, { "epoch": 1.05, "learning_rate": 1.5089876552977345e-05, "loss": 0.0001, "step": 1635 }, { "epoch": 1.05, "learning_rate": 1.5083901833900312e-05, "loss": 0.0, "step": 1636 }, { "epoch": 1.05, "learning_rate": 1.5077924666459783e-05, "loss": 0.0024, "step": 1637 }, { "epoch": 1.05, "learning_rate": 1.5071945053534319e-05, "loss": 0.0003, "step": 1638 }, { "epoch": 1.05, "learning_rate": 1.5065962998003647e-05, "loss": 0.0001, "step": 1639 }, { "epoch": 1.05, "learning_rate": 1.5059978502748669e-05, "loss": 0.0001, "step": 1640 }, { "epoch": 1.05, "learning_rate": 1.5053991570651469e-05, "loss": 0.0001, "step": 1641 }, { "epoch": 1.06, "learning_rate": 1.5048002204595304e-05, "loss": 0.0001, "step": 1642 }, { "epoch": 1.06, "learning_rate": 1.5042010407464603e-05, "loss": 0.0001, "step": 1643 }, { "epoch": 1.06, "learning_rate": 1.5036016182144957e-05, "loss": 0.0, "step": 1644 }, { "epoch": 1.06, "learning_rate": 1.5030019531523141e-05, "loss": 0.0002, "step": 1645 }, { "epoch": 1.06, "learning_rate": 1.5024020458487088e-05, "loss": 0.0001, "step": 1646 }, { "epoch": 1.06, "learning_rate": 1.5018018965925895e-05, "loss": 0.0001, "step": 1647 }, { "epoch": 1.06, "learning_rate": 1.5012015056729836e-05, "loss": 0.0, "step": 1648 }, { "epoch": 1.06, "learning_rate": 1.500600873379034e-05, "loss": 0.0001, "step": 1649 }, { "epoch": 1.06, "learning_rate": 1.5000000000000002e-05, "loss": 0.0001, "step": 1650 }, { "epoch": 1.06, "learning_rate": 1.4993988858252572e-05, "loss": 0.0001, "step": 1651 }, { "epoch": 1.06, "learning_rate": 1.498797531144297e-05, "loss": 0.0001, "step": 1652 }, { "epoch": 1.06, "learning_rate": 1.4981959362467265e-05, "loss": 0.0004, "step": 1653 }, { "epoch": 1.06, "learning_rate": 1.4975941014222688e-05, "loss": 0.0001, "step": 1654 }, { "epoch": 1.06, "learning_rate": 1.4969920269607621e-05, "loss": 0.0002, "step": 1655 }, { "epoch": 1.06, "learning_rate": 1.4963897131521605e-05, "loss": 0.0001, "step": 1656 }, { "epoch": 1.06, "learning_rate": 1.4957871602865334e-05, "loss": 0.0002, "step": 1657 }, { "epoch": 1.07, "learning_rate": 1.4951843686540648e-05, "loss": 0.0002, "step": 1658 }, { "epoch": 1.07, "learning_rate": 1.4945813385450536e-05, "loss": 0.0001, "step": 1659 }, { "epoch": 1.07, "learning_rate": 1.4939780702499149e-05, "loss": 0.0001, "step": 1660 }, { "epoch": 1.07, "learning_rate": 1.493374564059177e-05, "loss": 0.0, "step": 1661 }, { "epoch": 1.07, "learning_rate": 1.4927708202634828e-05, "loss": 0.0001, "step": 1662 }, { "epoch": 1.07, "learning_rate": 1.492166839153591e-05, "loss": 0.0003, "step": 1663 }, { "epoch": 1.07, "learning_rate": 1.4915626210203736e-05, "loss": 0.0001, "step": 1664 }, { "epoch": 1.07, "learning_rate": 1.490958166154816e-05, "loss": 0.0002, "step": 1665 }, { "epoch": 1.07, "learning_rate": 1.4903534748480193e-05, "loss": 0.0001, "step": 1666 }, { "epoch": 1.07, "learning_rate": 1.4897485473911974e-05, "loss": 0.0001, "step": 1667 }, { "epoch": 1.07, "learning_rate": 1.4891433840756783e-05, "loss": 0.0002, "step": 1668 }, { "epoch": 1.07, "learning_rate": 1.4885379851929034e-05, "loss": 0.0002, "step": 1669 }, { "epoch": 1.07, "learning_rate": 1.4879323510344277e-05, "loss": 0.0, "step": 1670 }, { "epoch": 1.07, "learning_rate": 1.4873264818919189e-05, "loss": 0.0002, "step": 1671 }, { "epoch": 1.07, "learning_rate": 1.4867203780571587e-05, "loss": 0.0001, "step": 1672 }, { "epoch": 1.08, "learning_rate": 1.486114039822042e-05, "loss": 0.0002, "step": 1673 }, { "epoch": 1.08, "learning_rate": 1.4855074674785755e-05, "loss": 0.0, "step": 1674 }, { "epoch": 1.08, "learning_rate": 1.4849006613188795e-05, "loss": 0.0001, "step": 1675 }, { "epoch": 1.08, "learning_rate": 1.4842936216351862e-05, "loss": 0.0001, "step": 1676 }, { "epoch": 1.08, "learning_rate": 1.483686348719841e-05, "loss": 0.0001, "step": 1677 }, { "epoch": 1.08, "learning_rate": 1.4830788428653016e-05, "loss": 0.0001, "step": 1678 }, { "epoch": 1.08, "learning_rate": 1.4824711043641373e-05, "loss": 0.0002, "step": 1679 }, { "epoch": 1.08, "learning_rate": 1.4818631335090299e-05, "loss": 0.0001, "step": 1680 }, { "epoch": 1.08, "learning_rate": 1.4812549305927728e-05, "loss": 0.0002, "step": 1681 }, { "epoch": 1.08, "learning_rate": 1.4806464959082712e-05, "loss": 0.0, "step": 1682 }, { "epoch": 1.08, "learning_rate": 1.4800378297485423e-05, "loss": 0.0001, "step": 1683 }, { "epoch": 1.08, "learning_rate": 1.4794289324067143e-05, "loss": 0.0002, "step": 1684 }, { "epoch": 1.08, "learning_rate": 1.478819804176027e-05, "loss": 0.0001, "step": 1685 }, { "epoch": 1.08, "learning_rate": 1.4782104453498316e-05, "loss": 0.0002, "step": 1686 }, { "epoch": 1.08, "learning_rate": 1.4776008562215895e-05, "loss": 0.0002, "step": 1687 }, { "epoch": 1.08, "learning_rate": 1.4769910370848744e-05, "loss": 0.0001, "step": 1688 }, { "epoch": 1.09, "learning_rate": 1.4763809882333688e-05, "loss": 0.0001, "step": 1689 }, { "epoch": 1.09, "learning_rate": 1.4757707099608681e-05, "loss": 0.0, "step": 1690 }, { "epoch": 1.09, "learning_rate": 1.4751602025612765e-05, "loss": 0.0001, "step": 1691 }, { "epoch": 1.09, "learning_rate": 1.4745494663286093e-05, "loss": 0.0, "step": 1692 }, { "epoch": 1.09, "learning_rate": 1.473938501556992e-05, "loss": 0.0001, "step": 1693 }, { "epoch": 1.09, "learning_rate": 1.4733273085406594e-05, "loss": 0.0002, "step": 1694 }, { "epoch": 1.09, "learning_rate": 1.4727158875739573e-05, "loss": 0.0001, "step": 1695 }, { "epoch": 1.09, "learning_rate": 1.4721042389513408e-05, "loss": 0.0, "step": 1696 }, { "epoch": 1.09, "learning_rate": 1.4714923629673746e-05, "loss": 0.0001, "step": 1697 }, { "epoch": 1.09, "learning_rate": 1.4708802599167327e-05, "loss": 0.0001, "step": 1698 }, { "epoch": 1.09, "learning_rate": 1.4702679300941988e-05, "loss": 0.0001, "step": 1699 }, { "epoch": 1.09, "learning_rate": 1.4696553737946658e-05, "loss": 0.0001, "step": 1700 }, { "epoch": 1.09, "learning_rate": 1.4690425913131352e-05, "loss": 0.0001, "step": 1701 }, { "epoch": 1.09, "learning_rate": 1.4684295829447184e-05, "loss": 0.0001, "step": 1702 }, { "epoch": 1.09, "learning_rate": 1.4678163489846346e-05, "loss": 0.0002, "step": 1703 }, { "epoch": 1.09, "learning_rate": 1.4672028897282117e-05, "loss": 0.0001, "step": 1704 }, { "epoch": 1.1, "learning_rate": 1.466589205470887e-05, "loss": 0.0001, "step": 1705 }, { "epoch": 1.1, "learning_rate": 1.4659752965082052e-05, "loss": 0.0001, "step": 1706 }, { "epoch": 1.1, "learning_rate": 1.4653611631358198e-05, "loss": 0.0002, "step": 1707 }, { "epoch": 1.1, "learning_rate": 1.464746805649492e-05, "loss": 0.0001, "step": 1708 }, { "epoch": 1.1, "learning_rate": 1.4641322243450913e-05, "loss": 0.0003, "step": 1709 }, { "epoch": 1.1, "learning_rate": 1.4635174195185945e-05, "loss": 0.0001, "step": 1710 }, { "epoch": 1.1, "learning_rate": 1.4629023914660862e-05, "loss": 0.0001, "step": 1711 }, { "epoch": 1.1, "learning_rate": 1.462287140483759e-05, "loss": 0.0005, "step": 1712 }, { "epoch": 1.1, "learning_rate": 1.4616716668679127e-05, "loss": 0.0002, "step": 1713 }, { "epoch": 1.1, "learning_rate": 1.4610559709149536e-05, "loss": 0.0001, "step": 1714 }, { "epoch": 1.1, "learning_rate": 1.4604400529213958e-05, "loss": 0.0, "step": 1715 }, { "epoch": 1.1, "learning_rate": 1.4598239131838603e-05, "loss": 0.0001, "step": 1716 }, { "epoch": 1.1, "learning_rate": 1.4592075519990743e-05, "loss": 0.0001, "step": 1717 }, { "epoch": 1.1, "learning_rate": 1.4585909696638722e-05, "loss": 0.0001, "step": 1718 }, { "epoch": 1.1, "learning_rate": 1.4579741664751947e-05, "loss": 0.0001, "step": 1719 }, { "epoch": 1.11, "learning_rate": 1.4573571427300896e-05, "loss": 0.0001, "step": 1720 }, { "epoch": 1.11, "learning_rate": 1.4567398987257093e-05, "loss": 0.0001, "step": 1721 }, { "epoch": 1.11, "learning_rate": 1.4561224347593137e-05, "loss": 0.0001, "step": 1722 }, { "epoch": 1.11, "learning_rate": 1.455504751128268e-05, "loss": 0.0001, "step": 1723 }, { "epoch": 1.11, "learning_rate": 1.4548868481300432e-05, "loss": 0.0002, "step": 1724 }, { "epoch": 1.11, "learning_rate": 1.4542687260622167e-05, "loss": 0.0001, "step": 1725 }, { "epoch": 1.11, "learning_rate": 1.4536503852224702e-05, "loss": 0.0001, "step": 1726 }, { "epoch": 1.11, "learning_rate": 1.4530318259085916e-05, "loss": 0.0002, "step": 1727 }, { "epoch": 1.11, "learning_rate": 1.4524130484184735e-05, "loss": 0.0001, "step": 1728 }, { "epoch": 1.11, "learning_rate": 1.451794053050114e-05, "loss": 0.0003, "step": 1729 }, { "epoch": 1.11, "learning_rate": 1.451174840101616e-05, "loss": 0.0001, "step": 1730 }, { "epoch": 1.11, "learning_rate": 1.450555409871187e-05, "loss": 0.0001, "step": 1731 }, { "epoch": 1.11, "learning_rate": 1.4499357626571394e-05, "loss": 0.0, "step": 1732 }, { "epoch": 1.11, "learning_rate": 1.4493158987578899e-05, "loss": 0.0001, "step": 1733 }, { "epoch": 1.11, "learning_rate": 1.4486958184719597e-05, "loss": 0.0002, "step": 1734 }, { "epoch": 1.11, "learning_rate": 1.4480755220979741e-05, "loss": 0.0001, "step": 1735 }, { "epoch": 1.12, "learning_rate": 1.4474550099346624e-05, "loss": 0.0002, "step": 1736 }, { "epoch": 1.12, "learning_rate": 1.4468342822808582e-05, "loss": 0.0003, "step": 1737 }, { "epoch": 1.12, "learning_rate": 1.4462133394354986e-05, "loss": 0.0002, "step": 1738 }, { "epoch": 1.12, "learning_rate": 1.4455921816976243e-05, "loss": 0.0001, "step": 1739 }, { "epoch": 1.12, "learning_rate": 1.4449708093663792e-05, "loss": 0.0001, "step": 1740 }, { "epoch": 1.12, "learning_rate": 1.4443492227410113e-05, "loss": 0.0001, "step": 1741 }, { "epoch": 1.12, "learning_rate": 1.4437274221208713e-05, "loss": 0.0, "step": 1742 }, { "epoch": 1.12, "learning_rate": 1.4431054078054126e-05, "loss": 0.0001, "step": 1743 }, { "epoch": 1.12, "learning_rate": 1.4424831800941928e-05, "loss": 0.0001, "step": 1744 }, { "epoch": 1.12, "learning_rate": 1.4418607392868706e-05, "loss": 0.0, "step": 1745 }, { "epoch": 1.12, "learning_rate": 1.4412380856832086e-05, "loss": 0.0001, "step": 1746 }, { "epoch": 1.12, "learning_rate": 1.4406152195830714e-05, "loss": 0.0001, "step": 1747 }, { "epoch": 1.12, "learning_rate": 1.4399921412864257e-05, "loss": 0.0002, "step": 1748 }, { "epoch": 1.12, "learning_rate": 1.439368851093341e-05, "loss": 0.0, "step": 1749 }, { "epoch": 1.12, "learning_rate": 1.4387453493039883e-05, "loss": 0.0002, "step": 1750 }, { "epoch": 1.13, "learning_rate": 1.438121636218641e-05, "loss": 0.0001, "step": 1751 }, { "epoch": 1.13, "learning_rate": 1.4374977121376736e-05, "loss": 0.0001, "step": 1752 }, { "epoch": 1.13, "learning_rate": 1.4368735773615625e-05, "loss": 0.0001, "step": 1753 }, { "epoch": 1.13, "learning_rate": 1.4362492321908861e-05, "loss": 0.0001, "step": 1754 }, { "epoch": 1.13, "learning_rate": 1.4356246769263236e-05, "loss": 0.0001, "step": 1755 }, { "epoch": 1.13, "learning_rate": 1.4349999118686553e-05, "loss": 0.0002, "step": 1756 }, { "epoch": 1.13, "learning_rate": 1.4343749373187623e-05, "loss": 0.0001, "step": 1757 }, { "epoch": 1.13, "learning_rate": 1.4337497535776277e-05, "loss": 0.0001, "step": 1758 }, { "epoch": 1.13, "learning_rate": 1.4331243609463343e-05, "loss": 0.0, "step": 1759 }, { "epoch": 1.13, "learning_rate": 1.4324987597260658e-05, "loss": 0.0, "step": 1760 }, { "epoch": 1.13, "learning_rate": 1.4318729502181064e-05, "loss": 0.0001, "step": 1761 }, { "epoch": 1.13, "learning_rate": 1.4312469327238407e-05, "loss": 0.0001, "step": 1762 }, { "epoch": 1.13, "learning_rate": 1.4306207075447531e-05, "loss": 0.0002, "step": 1763 }, { "epoch": 1.13, "learning_rate": 1.4299942749824285e-05, "loss": 0.0001, "step": 1764 }, { "epoch": 1.13, "learning_rate": 1.4293676353385513e-05, "loss": 0.0002, "step": 1765 }, { "epoch": 1.13, "learning_rate": 1.4287407889149059e-05, "loss": 0.0001, "step": 1766 }, { "epoch": 1.14, "learning_rate": 1.4281137360133761e-05, "loss": 0.0, "step": 1767 }, { "epoch": 1.14, "learning_rate": 1.4274864769359451e-05, "loss": 0.0002, "step": 1768 }, { "epoch": 1.14, "learning_rate": 1.4268590119846956e-05, "loss": 0.0002, "step": 1769 }, { "epoch": 1.14, "learning_rate": 1.4262313414618096e-05, "loss": 0.0001, "step": 1770 }, { "epoch": 1.14, "learning_rate": 1.425603465669567e-05, "loss": 0.0001, "step": 1771 }, { "epoch": 1.14, "learning_rate": 1.4249753849103486e-05, "loss": 0.0, "step": 1772 }, { "epoch": 1.14, "learning_rate": 1.4243470994866315e-05, "loss": 0.0002, "step": 1773 }, { "epoch": 1.14, "learning_rate": 1.4237186097009936e-05, "loss": 0.0001, "step": 1774 }, { "epoch": 1.14, "learning_rate": 1.4230899158561093e-05, "loss": 0.0002, "step": 1775 }, { "epoch": 1.14, "learning_rate": 1.422461018254753e-05, "loss": 0.0001, "step": 1776 }, { "epoch": 1.14, "learning_rate": 1.4218319171997954e-05, "loss": 0.0001, "step": 1777 }, { "epoch": 1.14, "learning_rate": 1.421202612994207e-05, "loss": 0.0001, "step": 1778 }, { "epoch": 1.14, "learning_rate": 1.420573105941055e-05, "loss": 0.0001, "step": 1779 }, { "epoch": 1.14, "learning_rate": 1.4199433963435048e-05, "loss": 0.0001, "step": 1780 }, { "epoch": 1.14, "learning_rate": 1.419313484504819e-05, "loss": 0.0002, "step": 1781 }, { "epoch": 1.15, "learning_rate": 1.4186833707283575e-05, "loss": 0.0001, "step": 1782 }, { "epoch": 1.15, "learning_rate": 1.4180530553175777e-05, "loss": 0.0002, "step": 1783 }, { "epoch": 1.15, "learning_rate": 1.4174225385760347e-05, "loss": 0.0001, "step": 1784 }, { "epoch": 1.15, "learning_rate": 1.4167918208073794e-05, "loss": 0.0001, "step": 1785 }, { "epoch": 1.15, "learning_rate": 1.41616090231536e-05, "loss": 0.0001, "step": 1786 }, { "epoch": 1.15, "learning_rate": 1.4155297834038217e-05, "loss": 0.0001, "step": 1787 }, { "epoch": 1.15, "learning_rate": 1.4148984643767058e-05, "loss": 0.0001, "step": 1788 }, { "epoch": 1.15, "learning_rate": 1.4142669455380502e-05, "loss": 0.0001, "step": 1789 }, { "epoch": 1.15, "learning_rate": 1.4136352271919891e-05, "loss": 0.0001, "step": 1790 }, { "epoch": 1.15, "learning_rate": 1.4130033096427527e-05, "loss": 0.0001, "step": 1791 }, { "epoch": 1.15, "learning_rate": 1.4123711931946667e-05, "loss": 0.0002, "step": 1792 }, { "epoch": 1.15, "learning_rate": 1.4117388781521529e-05, "loss": 0.0001, "step": 1793 }, { "epoch": 1.15, "learning_rate": 1.4111063648197289e-05, "loss": 0.0002, "step": 1794 }, { "epoch": 1.15, "learning_rate": 1.4104736535020083e-05, "loss": 0.0001, "step": 1795 }, { "epoch": 1.15, "learning_rate": 1.4098407445036983e-05, "loss": 0.0001, "step": 1796 }, { "epoch": 1.15, "learning_rate": 1.4092076381296038e-05, "loss": 0.0001, "step": 1797 }, { "epoch": 1.16, "learning_rate": 1.4085743346846223e-05, "loss": 0.0001, "step": 1798 }, { "epoch": 1.16, "learning_rate": 1.407940834473748e-05, "loss": 0.0001, "step": 1799 }, { "epoch": 1.16, "learning_rate": 1.4073071378020684e-05, "loss": 0.0002, "step": 1800 }, { "epoch": 1.16, "learning_rate": 1.4066732449747669e-05, "loss": 0.0001, "step": 1801 }, { "epoch": 1.16, "learning_rate": 1.406039156297121e-05, "loss": 0.0001, "step": 1802 }, { "epoch": 1.16, "learning_rate": 1.405404872074502e-05, "loss": 0.0001, "step": 1803 }, { "epoch": 1.16, "learning_rate": 1.4047703926123757e-05, "loss": 0.0001, "step": 1804 }, { "epoch": 1.16, "learning_rate": 1.4041357182163024e-05, "loss": 0.0002, "step": 1805 }, { "epoch": 1.16, "learning_rate": 1.4035008491919355e-05, "loss": 0.0001, "step": 1806 }, { "epoch": 1.16, "learning_rate": 1.4028657858450225e-05, "loss": 0.0, "step": 1807 }, { "epoch": 1.16, "learning_rate": 1.4022305284814046e-05, "loss": 0.0001, "step": 1808 }, { "epoch": 1.16, "learning_rate": 1.4015950774070164e-05, "loss": 0.0001, "step": 1809 }, { "epoch": 1.16, "learning_rate": 1.400959432927885e-05, "loss": 0.0001, "step": 1810 }, { "epoch": 1.16, "learning_rate": 1.4003235953501324e-05, "loss": 0.0003, "step": 1811 }, { "epoch": 1.16, "learning_rate": 1.3996875649799714e-05, "loss": 0.0001, "step": 1812 }, { "epoch": 1.16, "learning_rate": 1.3990513421237098e-05, "loss": 0.0001, "step": 1813 }, { "epoch": 1.17, "learning_rate": 1.398414927087746e-05, "loss": 0.0001, "step": 1814 }, { "epoch": 1.17, "learning_rate": 1.3977783201785732e-05, "loss": 0.0001, "step": 1815 }, { "epoch": 1.17, "learning_rate": 1.397141521702775e-05, "loss": 0.0001, "step": 1816 }, { "epoch": 1.17, "learning_rate": 1.396504531967028e-05, "loss": 0.0, "step": 1817 }, { "epoch": 1.17, "learning_rate": 1.3958673512781016e-05, "loss": 0.0001, "step": 1818 }, { "epoch": 1.17, "learning_rate": 1.3952299799428561e-05, "loss": 0.0001, "step": 1819 }, { "epoch": 1.17, "learning_rate": 1.3945924182682444e-05, "loss": 0.0001, "step": 1820 }, { "epoch": 1.17, "learning_rate": 1.3939546665613105e-05, "loss": 0.0001, "step": 1821 }, { "epoch": 1.17, "learning_rate": 1.3933167251291901e-05, "loss": 0.0001, "step": 1822 }, { "epoch": 1.17, "learning_rate": 1.3926785942791102e-05, "loss": 0.0001, "step": 1823 }, { "epoch": 1.17, "learning_rate": 1.3920402743183896e-05, "loss": 0.0002, "step": 1824 }, { "epoch": 1.17, "learning_rate": 1.3914017655544374e-05, "loss": 0.0001, "step": 1825 }, { "epoch": 1.17, "learning_rate": 1.390763068294754e-05, "loss": 0.0003, "step": 1826 }, { "epoch": 1.17, "learning_rate": 1.3901241828469308e-05, "loss": 0.0001, "step": 1827 }, { "epoch": 1.17, "learning_rate": 1.3894851095186485e-05, "loss": 0.0001, "step": 1828 }, { "epoch": 1.18, "learning_rate": 1.3888458486176808e-05, "loss": 0.0002, "step": 1829 }, { "epoch": 1.18, "learning_rate": 1.3882064004518894e-05, "loss": 0.0, "step": 1830 }, { "epoch": 1.18, "learning_rate": 1.3875667653292272e-05, "loss": 0.0002, "step": 1831 }, { "epoch": 1.18, "learning_rate": 1.3869269435577369e-05, "loss": 0.0, "step": 1832 }, { "epoch": 1.18, "learning_rate": 1.3862869354455513e-05, "loss": 0.0, "step": 1833 }, { "epoch": 1.18, "learning_rate": 1.3856467413008927e-05, "loss": 0.0002, "step": 1834 }, { "epoch": 1.18, "learning_rate": 1.3850063614320732e-05, "loss": 0.0001, "step": 1835 }, { "epoch": 1.18, "learning_rate": 1.3843657961474944e-05, "loss": 0.0001, "step": 1836 }, { "epoch": 1.18, "learning_rate": 1.3837250457556468e-05, "loss": 0.0001, "step": 1837 }, { "epoch": 1.18, "learning_rate": 1.3830841105651104e-05, "loss": 0.0001, "step": 1838 }, { "epoch": 1.18, "learning_rate": 1.382442990884554e-05, "loss": 0.0001, "step": 1839 }, { "epoch": 1.18, "learning_rate": 1.3818016870227355e-05, "loss": 0.0002, "step": 1840 }, { "epoch": 1.18, "learning_rate": 1.381160199288501e-05, "loss": 0.0, "step": 1841 }, { "epoch": 1.18, "learning_rate": 1.3805185279907858e-05, "loss": 0.0, "step": 1842 }, { "epoch": 1.18, "learning_rate": 1.3798766734386133e-05, "loss": 0.0005, "step": 1843 }, { "epoch": 1.18, "learning_rate": 1.3792346359410953e-05, "loss": 0.0002, "step": 1844 }, { "epoch": 1.19, "learning_rate": 1.378592415807431e-05, "loss": 0.0002, "step": 1845 }, { "epoch": 1.19, "learning_rate": 1.3779500133469083e-05, "loss": 0.0001, "step": 1846 }, { "epoch": 1.19, "learning_rate": 1.377307428868903e-05, "loss": 0.0001, "step": 1847 }, { "epoch": 1.19, "learning_rate": 1.3766646626828781e-05, "loss": 0.0001, "step": 1848 }, { "epoch": 1.19, "learning_rate": 1.376021715098384e-05, "loss": 0.0002, "step": 1849 }, { "epoch": 1.19, "learning_rate": 1.3753785864250592e-05, "loss": 0.0002, "step": 1850 }, { "epoch": 1.19, "learning_rate": 1.3747352769726292e-05, "loss": 0.0, "step": 1851 }, { "epoch": 1.19, "learning_rate": 1.3740917870509054e-05, "loss": 0.0001, "step": 1852 }, { "epoch": 1.19, "learning_rate": 1.373448116969787e-05, "loss": 0.0, "step": 1853 }, { "epoch": 1.19, "learning_rate": 1.3728042670392615e-05, "loss": 0.0, "step": 1854 }, { "epoch": 1.19, "learning_rate": 1.3721602375694e-05, "loss": 0.0, "step": 1855 }, { "epoch": 1.19, "learning_rate": 1.3715160288703617e-05, "loss": 0.0001, "step": 1856 }, { "epoch": 1.19, "learning_rate": 1.3708716412523928e-05, "loss": 0.0001, "step": 1857 }, { "epoch": 1.19, "learning_rate": 1.3702270750258243e-05, "loss": 0.0002, "step": 1858 }, { "epoch": 1.19, "learning_rate": 1.3695823305010737e-05, "loss": 0.0001, "step": 1859 }, { "epoch": 1.2, "learning_rate": 1.3689374079886444e-05, "loss": 0.0001, "step": 1860 }, { "epoch": 1.2, "learning_rate": 1.3682923077991262e-05, "loss": 0.0001, "step": 1861 }, { "epoch": 1.2, "learning_rate": 1.367647030243193e-05, "loss": 0.0002, "step": 1862 }, { "epoch": 1.2, "learning_rate": 1.3670015756316053e-05, "loss": 0.0001, "step": 1863 }, { "epoch": 1.2, "learning_rate": 1.3663559442752084e-05, "loss": 0.0002, "step": 1864 }, { "epoch": 1.2, "learning_rate": 1.3657101364849328e-05, "loss": 0.0001, "step": 1865 }, { "epoch": 1.2, "learning_rate": 1.3650641525717943e-05, "loss": 0.0002, "step": 1866 }, { "epoch": 1.2, "learning_rate": 1.3644179928468927e-05, "loss": 0.0, "step": 1867 }, { "epoch": 1.2, "learning_rate": 1.3637716576214137e-05, "loss": 0.0001, "step": 1868 }, { "epoch": 1.2, "learning_rate": 1.3631251472066263e-05, "loss": 0.0002, "step": 1869 }, { "epoch": 1.2, "learning_rate": 1.362478461913884e-05, "loss": 0.0001, "step": 1870 }, { "epoch": 1.2, "learning_rate": 1.3618316020546255e-05, "loss": 0.0, "step": 1871 }, { "epoch": 1.2, "learning_rate": 1.3611845679403728e-05, "loss": 0.0001, "step": 1872 }, { "epoch": 1.2, "learning_rate": 1.3605373598827321e-05, "loss": 0.0, "step": 1873 }, { "epoch": 1.2, "learning_rate": 1.3598899781933925e-05, "loss": 0.0, "step": 1874 }, { "epoch": 1.2, "learning_rate": 1.3592424231841288e-05, "loss": 0.0, "step": 1875 }, { "epoch": 1.21, "learning_rate": 1.3585946951667968e-05, "loss": 0.0002, "step": 1876 }, { "epoch": 1.21, "learning_rate": 1.357946794453337e-05, "loss": 0.0002, "step": 1877 }, { "epoch": 1.21, "learning_rate": 1.3572987213557731e-05, "loss": 0.0001, "step": 1878 }, { "epoch": 1.21, "learning_rate": 1.3566504761862116e-05, "loss": 0.0001, "step": 1879 }, { "epoch": 1.21, "learning_rate": 1.3560020592568414e-05, "loss": 0.0001, "step": 1880 }, { "epoch": 1.21, "learning_rate": 1.355353470879935e-05, "loss": 0.0, "step": 1881 }, { "epoch": 1.21, "learning_rate": 1.3547047113678465e-05, "loss": 0.0001, "step": 1882 }, { "epoch": 1.21, "learning_rate": 1.3540557810330133e-05, "loss": 0.0002, "step": 1883 }, { "epoch": 1.21, "learning_rate": 1.3534066801879544e-05, "loss": 0.0001, "step": 1884 }, { "epoch": 1.21, "learning_rate": 1.3527574091452714e-05, "loss": 0.0001, "step": 1885 }, { "epoch": 1.21, "learning_rate": 1.3521079682176477e-05, "loss": 0.0001, "step": 1886 }, { "epoch": 1.21, "learning_rate": 1.3514583577178482e-05, "loss": 0.0002, "step": 1887 }, { "epoch": 1.21, "learning_rate": 1.35080857795872e-05, "loss": 0.0001, "step": 1888 }, { "epoch": 1.21, "learning_rate": 1.350158629253191e-05, "loss": 0.0, "step": 1889 }, { "epoch": 1.21, "learning_rate": 1.3495085119142717e-05, "loss": 0.0001, "step": 1890 }, { "epoch": 1.22, "learning_rate": 1.3488582262550528e-05, "loss": 0.0001, "step": 1891 }, { "epoch": 1.22, "learning_rate": 1.3482077725887056e-05, "loss": 0.0001, "step": 1892 }, { "epoch": 1.22, "learning_rate": 1.3475571512284838e-05, "loss": 0.0001, "step": 1893 }, { "epoch": 1.22, "learning_rate": 1.3469063624877205e-05, "loss": 0.0002, "step": 1894 }, { "epoch": 1.22, "learning_rate": 1.3462554066798305e-05, "loss": 0.0001, "step": 1895 }, { "epoch": 1.22, "learning_rate": 1.3456042841183083e-05, "loss": 0.0002, "step": 1896 }, { "epoch": 1.22, "learning_rate": 1.3449529951167288e-05, "loss": 0.0002, "step": 1897 }, { "epoch": 1.22, "learning_rate": 1.3443015399887476e-05, "loss": 0.0002, "step": 1898 }, { "epoch": 1.22, "learning_rate": 1.3436499190480991e-05, "loss": 0.0003, "step": 1899 }, { "epoch": 1.22, "learning_rate": 1.3429981326085992e-05, "loss": 0.0001, "step": 1900 }, { "epoch": 1.22, "learning_rate": 1.3423461809841424e-05, "loss": 0.0001, "step": 1901 }, { "epoch": 1.22, "learning_rate": 1.3416940644887026e-05, "loss": 0.0, "step": 1902 }, { "epoch": 1.22, "learning_rate": 1.3410417834363341e-05, "loss": 0.0001, "step": 1903 }, { "epoch": 1.22, "learning_rate": 1.3403893381411697e-05, "loss": 0.0001, "step": 1904 }, { "epoch": 1.22, "learning_rate": 1.339736728917421e-05, "loss": 0.0001, "step": 1905 }, { "epoch": 1.22, "learning_rate": 1.3390839560793793e-05, "loss": 0.0001, "step": 1906 }, { "epoch": 1.23, "learning_rate": 1.3384310199414148e-05, "loss": 0.0001, "step": 1907 }, { "epoch": 1.23, "learning_rate": 1.3377779208179753e-05, "loss": 0.0001, "step": 1908 }, { "epoch": 1.23, "learning_rate": 1.337124659023588e-05, "loss": 0.0001, "step": 1909 }, { "epoch": 1.23, "learning_rate": 1.3364712348728586e-05, "loss": 0.0, "step": 1910 }, { "epoch": 1.23, "learning_rate": 1.33581764868047e-05, "loss": 0.0001, "step": 1911 }, { "epoch": 1.23, "learning_rate": 1.3351639007611836e-05, "loss": 0.0001, "step": 1912 }, { "epoch": 1.23, "learning_rate": 1.334509991429839e-05, "loss": 0.0001, "step": 1913 }, { "epoch": 1.23, "learning_rate": 1.3338559210013534e-05, "loss": 0.0001, "step": 1914 }, { "epoch": 1.23, "learning_rate": 1.3332016897907219e-05, "loss": 0.0001, "step": 1915 }, { "epoch": 1.23, "learning_rate": 1.3325472981130158e-05, "loss": 0.0, "step": 1916 }, { "epoch": 1.23, "learning_rate": 1.3318927462833848e-05, "loss": 0.0002, "step": 1917 }, { "epoch": 1.23, "learning_rate": 1.331238034617056e-05, "loss": 0.0001, "step": 1918 }, { "epoch": 1.23, "learning_rate": 1.3305831634293323e-05, "loss": 0.0001, "step": 1919 }, { "epoch": 1.23, "learning_rate": 1.3299281330355939e-05, "loss": 0.0001, "step": 1920 }, { "epoch": 1.23, "learning_rate": 1.3292729437512985e-05, "loss": 0.0001, "step": 1921 }, { "epoch": 1.24, "learning_rate": 1.3286175958919794e-05, "loss": 0.0, "step": 1922 }, { "epoch": 1.24, "learning_rate": 1.3279620897732459e-05, "loss": 0.0001, "step": 1923 }, { "epoch": 1.24, "learning_rate": 1.3273064257107849e-05, "loss": 0.0001, "step": 1924 }, { "epoch": 1.24, "learning_rate": 1.3266506040203583e-05, "loss": 0.0001, "step": 1925 }, { "epoch": 1.24, "learning_rate": 1.3259946250178043e-05, "loss": 0.0001, "step": 1926 }, { "epoch": 1.24, "learning_rate": 1.3253384890190362e-05, "loss": 0.0001, "step": 1927 }, { "epoch": 1.24, "learning_rate": 1.3246821963400442e-05, "loss": 0.0001, "step": 1928 }, { "epoch": 1.24, "learning_rate": 1.3240257472968929e-05, "loss": 0.0001, "step": 1929 }, { "epoch": 1.24, "learning_rate": 1.3233691422057226e-05, "loss": 0.0003, "step": 1930 }, { "epoch": 1.24, "learning_rate": 1.3227123813827487e-05, "loss": 0.0002, "step": 1931 }, { "epoch": 1.24, "learning_rate": 1.3220554651442614e-05, "loss": 0.0001, "step": 1932 }, { "epoch": 1.24, "learning_rate": 1.321398393806626e-05, "loss": 0.0001, "step": 1933 }, { "epoch": 1.24, "learning_rate": 1.3207411676862827e-05, "loss": 0.0001, "step": 1934 }, { "epoch": 1.24, "learning_rate": 1.3200837870997454e-05, "loss": 0.0001, "step": 1935 }, { "epoch": 1.24, "learning_rate": 1.3194262523636035e-05, "loss": 0.0, "step": 1936 }, { "epoch": 1.24, "learning_rate": 1.3187685637945198e-05, "loss": 0.0001, "step": 1937 }, { "epoch": 1.25, "learning_rate": 1.3181107217092312e-05, "loss": 0.0001, "step": 1938 }, { "epoch": 1.25, "learning_rate": 1.3174527264245495e-05, "loss": 0.0001, "step": 1939 }, { "epoch": 1.25, "learning_rate": 1.3167945782573592e-05, "loss": 0.0001, "step": 1940 }, { "epoch": 1.25, "learning_rate": 1.3161362775246185e-05, "loss": 0.0, "step": 1941 }, { "epoch": 1.25, "learning_rate": 1.3154778245433597e-05, "loss": 0.0001, "step": 1942 }, { "epoch": 1.25, "learning_rate": 1.3148192196306884e-05, "loss": 0.0002, "step": 1943 }, { "epoch": 1.25, "learning_rate": 1.3141604631037827e-05, "loss": 0.0001, "step": 1944 }, { "epoch": 1.25, "learning_rate": 1.313501555279894e-05, "loss": 0.0002, "step": 1945 }, { "epoch": 1.25, "learning_rate": 1.3128424964763471e-05, "loss": 0.0001, "step": 1946 }, { "epoch": 1.25, "learning_rate": 1.3121832870105382e-05, "loss": 0.0002, "step": 1947 }, { "epoch": 1.25, "learning_rate": 1.3115239271999379e-05, "loss": 0.0, "step": 1948 }, { "epoch": 1.25, "learning_rate": 1.3108644173620879e-05, "loss": 0.0004, "step": 1949 }, { "epoch": 1.25, "learning_rate": 1.3102047578146024e-05, "loss": 0.0, "step": 1950 }, { "epoch": 1.25, "learning_rate": 1.3095449488751677e-05, "loss": 0.0, "step": 1951 }, { "epoch": 1.25, "learning_rate": 1.308884990861542e-05, "loss": 0.0003, "step": 1952 }, { "epoch": 1.25, "learning_rate": 1.3082248840915556e-05, "loss": 0.0002, "step": 1953 }, { "epoch": 1.26, "learning_rate": 1.3075646288831108e-05, "loss": 0.0002, "step": 1954 }, { "epoch": 1.26, "learning_rate": 1.3069042255541798e-05, "loss": 0.0001, "step": 1955 }, { "epoch": 1.26, "learning_rate": 1.3062436744228078e-05, "loss": 0.0001, "step": 1956 }, { "epoch": 1.26, "learning_rate": 1.3055829758071105e-05, "loss": 0.0001, "step": 1957 }, { "epoch": 1.26, "learning_rate": 1.3049221300252745e-05, "loss": 0.0001, "step": 1958 }, { "epoch": 1.26, "learning_rate": 1.3042611373955578e-05, "loss": 0.0001, "step": 1959 }, { "epoch": 1.26, "learning_rate": 1.3035999982362885e-05, "loss": 0.0, "step": 1960 }, { "epoch": 1.26, "learning_rate": 1.3029387128658657e-05, "loss": 0.0, "step": 1961 }, { "epoch": 1.26, "learning_rate": 1.3022772816027588e-05, "loss": 0.0001, "step": 1962 }, { "epoch": 1.26, "learning_rate": 1.3016157047655066e-05, "loss": 0.0001, "step": 1963 }, { "epoch": 1.26, "learning_rate": 1.3009539826727202e-05, "loss": 0.0003, "step": 1964 }, { "epoch": 1.26, "learning_rate": 1.3002921156430781e-05, "loss": 0.0002, "step": 1965 }, { "epoch": 1.26, "learning_rate": 1.2996301039953305e-05, "loss": 0.0, "step": 1966 }, { "epoch": 1.26, "learning_rate": 1.2989679480482964e-05, "loss": 0.0, "step": 1967 }, { "epoch": 1.26, "learning_rate": 1.2983056481208642e-05, "loss": 0.0001, "step": 1968 }, { "epoch": 1.27, "learning_rate": 1.2976432045319923e-05, "loss": 0.0002, "step": 1969 }, { "epoch": 1.27, "learning_rate": 1.2969806176007072e-05, "loss": 0.0003, "step": 1970 }, { "epoch": 1.27, "learning_rate": 1.2963178876461053e-05, "loss": 0.0001, "step": 1971 }, { "epoch": 1.27, "learning_rate": 1.2956550149873519e-05, "loss": 0.0002, "step": 1972 }, { "epoch": 1.27, "learning_rate": 1.2949919999436807e-05, "loss": 0.0001, "step": 1973 }, { "epoch": 1.27, "learning_rate": 1.294328842834394e-05, "loss": 0.0001, "step": 1974 }, { "epoch": 1.27, "learning_rate": 1.2936655439788628e-05, "loss": 0.0, "step": 1975 }, { "epoch": 1.27, "learning_rate": 1.2930021036965257e-05, "loss": 0.0, "step": 1976 }, { "epoch": 1.27, "learning_rate": 1.2923385223068898e-05, "loss": 0.0001, "step": 1977 }, { "epoch": 1.27, "learning_rate": 1.2916748001295307e-05, "loss": 0.0, "step": 1978 }, { "epoch": 1.27, "learning_rate": 1.2910109374840914e-05, "loss": 0.0001, "step": 1979 }, { "epoch": 1.27, "learning_rate": 1.2903469346902817e-05, "loss": 0.0001, "step": 1980 }, { "epoch": 1.27, "learning_rate": 1.2896827920678803e-05, "loss": 0.0, "step": 1981 }, { "epoch": 1.27, "learning_rate": 1.289018509936732e-05, "loss": 0.0001, "step": 1982 }, { "epoch": 1.27, "learning_rate": 1.28835408861675e-05, "loss": 0.0001, "step": 1983 }, { "epoch": 1.27, "learning_rate": 1.2876895284279136e-05, "loss": 0.0001, "step": 1984 }, { "epoch": 1.28, "learning_rate": 1.2870248296902697e-05, "loss": 0.0001, "step": 1985 }, { "epoch": 1.28, "learning_rate": 1.2863599927239307e-05, "loss": 0.0001, "step": 1986 }, { "epoch": 1.28, "learning_rate": 1.2856950178490768e-05, "loss": 0.0002, "step": 1987 }, { "epoch": 1.28, "learning_rate": 1.285029905385954e-05, "loss": 0.0001, "step": 1988 }, { "epoch": 1.28, "learning_rate": 1.2843646556548754e-05, "loss": 0.0005, "step": 1989 }, { "epoch": 1.28, "learning_rate": 1.2836992689762191e-05, "loss": 0.0001, "step": 1990 }, { "epoch": 1.28, "learning_rate": 1.2830337456704294e-05, "loss": 0.0, "step": 1991 }, { "epoch": 1.28, "learning_rate": 1.2823680860580168e-05, "loss": 0.0001, "step": 1992 }, { "epoch": 1.28, "learning_rate": 1.2817022904595575e-05, "loss": 0.0002, "step": 1993 }, { "epoch": 1.28, "learning_rate": 1.2810363591956926e-05, "loss": 0.0, "step": 1994 }, { "epoch": 1.28, "learning_rate": 1.2803702925871289e-05, "loss": 0.0, "step": 1995 }, { "epoch": 1.28, "learning_rate": 1.2797040909546388e-05, "loss": 0.0, "step": 1996 }, { "epoch": 1.28, "learning_rate": 1.279037754619059e-05, "loss": 0.0001, "step": 1997 }, { "epoch": 1.28, "learning_rate": 1.278371283901291e-05, "loss": 0.0, "step": 1998 }, { "epoch": 1.28, "learning_rate": 1.277704679122302e-05, "loss": 0.0001, "step": 1999 }, { "epoch": 1.29, "learning_rate": 1.2770379406031227e-05, "loss": 0.0001, "step": 2000 }, { "epoch": 1.29, "learning_rate": 1.2763710686648485e-05, "loss": 0.0, "step": 2001 }, { "epoch": 1.29, "learning_rate": 1.27570406362864e-05, "loss": 0.0001, "step": 2002 }, { "epoch": 1.29, "learning_rate": 1.2750369258157205e-05, "loss": 0.0001, "step": 2003 }, { "epoch": 1.29, "learning_rate": 1.2743696555473779e-05, "loss": 0.0002, "step": 2004 }, { "epoch": 1.29, "learning_rate": 1.2737022531449639e-05, "loss": 0.0001, "step": 2005 }, { "epoch": 1.29, "learning_rate": 1.2730347189298936e-05, "loss": 0.0, "step": 2006 }, { "epoch": 1.29, "learning_rate": 1.272367053223646e-05, "loss": 0.0001, "step": 2007 }, { "epoch": 1.29, "learning_rate": 1.2716992563477633e-05, "loss": 0.0001, "step": 2008 }, { "epoch": 1.29, "learning_rate": 1.2710313286238503e-05, "loss": 0.0001, "step": 2009 }, { "epoch": 1.29, "learning_rate": 1.2703632703735756e-05, "loss": 0.0003, "step": 2010 }, { "epoch": 1.29, "learning_rate": 1.2696950819186699e-05, "loss": 0.0001, "step": 2011 }, { "epoch": 1.29, "learning_rate": 1.2690267635809275e-05, "loss": 0.0, "step": 2012 }, { "epoch": 1.29, "learning_rate": 1.2683583156822045e-05, "loss": 0.0001, "step": 2013 }, { "epoch": 1.29, "learning_rate": 1.2676897385444202e-05, "loss": 0.0001, "step": 2014 }, { "epoch": 1.29, "learning_rate": 1.2670210324895548e-05, "loss": 0.0002, "step": 2015 }, { "epoch": 1.3, "learning_rate": 1.266352197839652e-05, "loss": 0.0002, "step": 2016 }, { "epoch": 1.3, "learning_rate": 1.2656832349168165e-05, "loss": 0.0, "step": 2017 }, { "epoch": 1.3, "learning_rate": 1.2650141440432152e-05, "loss": 0.0001, "step": 2018 }, { "epoch": 1.3, "learning_rate": 1.2643449255410765e-05, "loss": 0.0001, "step": 2019 }, { "epoch": 1.3, "learning_rate": 1.2636755797326903e-05, "loss": 0.0001, "step": 2020 }, { "epoch": 1.3, "learning_rate": 1.263006106940408e-05, "loss": 0.0, "step": 2021 }, { "epoch": 1.3, "learning_rate": 1.2623365074866418e-05, "loss": 0.0001, "step": 2022 }, { "epoch": 1.3, "learning_rate": 1.2616667816938648e-05, "loss": 0.0001, "step": 2023 }, { "epoch": 1.3, "learning_rate": 1.2609969298846112e-05, "loss": 0.0001, "step": 2024 }, { "epoch": 1.3, "learning_rate": 1.2603269523814765e-05, "loss": 0.0001, "step": 2025 }, { "epoch": 1.3, "learning_rate": 1.2596568495071155e-05, "loss": 0.0001, "step": 2026 }, { "epoch": 1.3, "learning_rate": 1.2589866215842443e-05, "loss": 0.0001, "step": 2027 }, { "epoch": 1.3, "learning_rate": 1.258316268935639e-05, "loss": 0.0001, "step": 2028 }, { "epoch": 1.3, "learning_rate": 1.2576457918841354e-05, "loss": 0.0001, "step": 2029 }, { "epoch": 1.3, "learning_rate": 1.2569751907526293e-05, "loss": 0.0001, "step": 2030 }, { "epoch": 1.31, "learning_rate": 1.256304465864077e-05, "loss": 0.0001, "step": 2031 }, { "epoch": 1.31, "learning_rate": 1.2556336175414939e-05, "loss": 0.0001, "step": 2032 }, { "epoch": 1.31, "learning_rate": 1.2549626461079537e-05, "loss": 0.0001, "step": 2033 }, { "epoch": 1.31, "learning_rate": 1.254291551886592e-05, "loss": 0.0001, "step": 2034 }, { "epoch": 1.31, "learning_rate": 1.2536203352006009e-05, "loss": 0.0001, "step": 2035 }, { "epoch": 1.31, "learning_rate": 1.2529489963732333e-05, "loss": 0.0001, "step": 2036 }, { "epoch": 1.31, "learning_rate": 1.2522775357277997e-05, "loss": 0.0002, "step": 2037 }, { "epoch": 1.31, "learning_rate": 1.2516059535876698e-05, "loss": 0.0002, "step": 2038 }, { "epoch": 1.31, "learning_rate": 1.2509342502762725e-05, "loss": 0.0001, "step": 2039 }, { "epoch": 1.31, "learning_rate": 1.2502624261170934e-05, "loss": 0.0, "step": 2040 }, { "epoch": 1.31, "learning_rate": 1.2495904814336777e-05, "loss": 0.0002, "step": 2041 }, { "epoch": 1.31, "learning_rate": 1.2489184165496284e-05, "loss": 0.0002, "step": 2042 }, { "epoch": 1.31, "learning_rate": 1.248246231788606e-05, "loss": 0.0001, "step": 2043 }, { "epoch": 1.31, "learning_rate": 1.2475739274743286e-05, "loss": 0.0001, "step": 2044 }, { "epoch": 1.31, "learning_rate": 1.2469015039305727e-05, "loss": 0.0, "step": 2045 }, { "epoch": 1.31, "learning_rate": 1.2462289614811715e-05, "loss": 0.0001, "step": 2046 }, { "epoch": 1.32, "learning_rate": 1.2455563004500152e-05, "loss": 0.0002, "step": 2047 }, { "epoch": 1.32, "learning_rate": 1.2448835211610524e-05, "loss": 0.0002, "step": 2048 }, { "epoch": 1.32, "learning_rate": 1.2442106239382873e-05, "loss": 0.0, "step": 2049 }, { "epoch": 1.32, "learning_rate": 1.2435376091057818e-05, "loss": 0.0001, "step": 2050 }, { "epoch": 1.32, "learning_rate": 1.2428644769876537e-05, "loss": 0.0002, "step": 2051 }, { "epoch": 1.32, "learning_rate": 1.242191227908078e-05, "loss": 0.0001, "step": 2052 }, { "epoch": 1.32, "learning_rate": 1.2415178621912851e-05, "loss": 0.0, "step": 2053 }, { "epoch": 1.32, "learning_rate": 1.2408443801615626e-05, "loss": 0.0002, "step": 2054 }, { "epoch": 1.32, "learning_rate": 1.2401707821432536e-05, "loss": 0.0, "step": 2055 }, { "epoch": 1.32, "learning_rate": 1.2394970684607572e-05, "loss": 0.0002, "step": 2056 }, { "epoch": 1.32, "learning_rate": 1.2388232394385283e-05, "loss": 0.0001, "step": 2057 }, { "epoch": 1.32, "learning_rate": 1.2381492954010765e-05, "loss": 0.0001, "step": 2058 }, { "epoch": 1.32, "learning_rate": 1.2374752366729678e-05, "loss": 0.0002, "step": 2059 }, { "epoch": 1.32, "learning_rate": 1.2368010635788238e-05, "loss": 0.0002, "step": 2060 }, { "epoch": 1.32, "learning_rate": 1.2361267764433194e-05, "loss": 0.0, "step": 2061 }, { "epoch": 1.32, "learning_rate": 1.2354523755911858e-05, "loss": 0.0, "step": 2062 }, { "epoch": 1.33, "learning_rate": 1.2347778613472093e-05, "loss": 0.0001, "step": 2063 }, { "epoch": 1.33, "learning_rate": 1.2341032340362294e-05, "loss": 0.0001, "step": 2064 }, { "epoch": 1.33, "learning_rate": 1.233428493983141e-05, "loss": 0.0001, "step": 2065 }, { "epoch": 1.33, "learning_rate": 1.2327536415128933e-05, "loss": 0.0001, "step": 2066 }, { "epoch": 1.33, "learning_rate": 1.2320786769504895e-05, "loss": 0.0002, "step": 2067 }, { "epoch": 1.33, "learning_rate": 1.2314036006209866e-05, "loss": 0.0, "step": 2068 }, { "epoch": 1.33, "learning_rate": 1.2307284128494957e-05, "loss": 0.0001, "step": 2069 }, { "epoch": 1.33, "learning_rate": 1.2300531139611809e-05, "loss": 0.0, "step": 2070 }, { "epoch": 1.33, "learning_rate": 1.2293777042812614e-05, "loss": 0.0002, "step": 2071 }, { "epoch": 1.33, "learning_rate": 1.2287021841350075e-05, "loss": 0.0001, "step": 2072 }, { "epoch": 1.33, "learning_rate": 1.228026553847745e-05, "loss": 0.0001, "step": 2073 }, { "epoch": 1.33, "learning_rate": 1.227350813744851e-05, "loss": 0.0002, "step": 2074 }, { "epoch": 1.33, "learning_rate": 1.2266749641517561e-05, "loss": 0.0001, "step": 2075 }, { "epoch": 1.33, "learning_rate": 1.225999005393944e-05, "loss": 0.0, "step": 2076 }, { "epoch": 1.33, "learning_rate": 1.22532293779695e-05, "loss": 0.0001, "step": 2077 }, { "epoch": 1.34, "learning_rate": 1.2246467616863635e-05, "loss": 0.0001, "step": 2078 }, { "epoch": 1.34, "learning_rate": 1.2239704773878242e-05, "loss": 0.0001, "step": 2079 }, { "epoch": 1.34, "learning_rate": 1.2232940852270254e-05, "loss": 0.0001, "step": 2080 }, { "epoch": 1.34, "learning_rate": 1.2226175855297116e-05, "loss": 0.0002, "step": 2081 }, { "epoch": 1.34, "learning_rate": 1.2219409786216788e-05, "loss": 0.0002, "step": 2082 }, { "epoch": 1.34, "learning_rate": 1.2212642648287754e-05, "loss": 0.0, "step": 2083 }, { "epoch": 1.34, "learning_rate": 1.2205874444769013e-05, "loss": 0.0001, "step": 2084 }, { "epoch": 1.34, "learning_rate": 1.219910517892007e-05, "loss": 0.0002, "step": 2085 }, { "epoch": 1.34, "learning_rate": 1.2192334854000945e-05, "loss": 0.0001, "step": 2086 }, { "epoch": 1.34, "learning_rate": 1.2185563473272171e-05, "loss": 0.0, "step": 2087 }, { "epoch": 1.34, "learning_rate": 1.2178791039994781e-05, "loss": 0.0, "step": 2088 }, { "epoch": 1.34, "learning_rate": 1.217201755743033e-05, "loss": 0.0001, "step": 2089 }, { "epoch": 1.34, "learning_rate": 1.2165243028840861e-05, "loss": 0.0002, "step": 2090 }, { "epoch": 1.34, "learning_rate": 1.2158467457488934e-05, "loss": 0.0, "step": 2091 }, { "epoch": 1.34, "learning_rate": 1.2151690846637605e-05, "loss": 0.0001, "step": 2092 }, { "epoch": 1.34, "learning_rate": 1.2144913199550427e-05, "loss": 0.0002, "step": 2093 }, { "epoch": 1.35, "learning_rate": 1.213813451949146e-05, "loss": 0.0001, "step": 2094 }, { "epoch": 1.35, "learning_rate": 1.2131354809725262e-05, "loss": 0.0001, "step": 2095 }, { "epoch": 1.35, "learning_rate": 1.2124574073516878e-05, "loss": 0.0, "step": 2096 }, { "epoch": 1.35, "learning_rate": 1.2117792314131851e-05, "loss": 0.0, "step": 2097 }, { "epoch": 1.35, "learning_rate": 1.2111009534836225e-05, "loss": 0.0, "step": 2098 }, { "epoch": 1.35, "learning_rate": 1.210422573889652e-05, "loss": 0.0, "step": 2099 }, { "epoch": 1.35, "learning_rate": 1.2097440929579762e-05, "loss": 0.0, "step": 2100 }, { "epoch": 1.35, "learning_rate": 1.2090655110153448e-05, "loss": 0.0002, "step": 2101 }, { "epoch": 1.35, "learning_rate": 1.208386828388558e-05, "loss": 0.0001, "step": 2102 }, { "epoch": 1.35, "learning_rate": 1.2077080454044632e-05, "loss": 0.0001, "step": 2103 }, { "epoch": 1.35, "learning_rate": 1.2070291623899557e-05, "loss": 0.0, "step": 2104 }, { "epoch": 1.35, "learning_rate": 1.206350179671981e-05, "loss": 0.0001, "step": 2105 }, { "epoch": 1.35, "learning_rate": 1.2056710975775303e-05, "loss": 0.0001, "step": 2106 }, { "epoch": 1.35, "learning_rate": 1.2049919164336443e-05, "loss": 0.0, "step": 2107 }, { "epoch": 1.35, "learning_rate": 1.2043126365674102e-05, "loss": 0.0003, "step": 2108 }, { "epoch": 1.36, "learning_rate": 1.2036332583059645e-05, "loss": 0.0002, "step": 2109 }, { "epoch": 1.36, "learning_rate": 1.202953781976489e-05, "loss": 0.0, "step": 2110 }, { "epoch": 1.36, "learning_rate": 1.2022742079062136e-05, "loss": 0.0002, "step": 2111 }, { "epoch": 1.36, "learning_rate": 1.2015945364224155e-05, "loss": 0.0001, "step": 2112 }, { "epoch": 1.36, "learning_rate": 1.2009147678524191e-05, "loss": 0.0001, "step": 2113 }, { "epoch": 1.36, "learning_rate": 1.2002349025235952e-05, "loss": 0.0, "step": 2114 }, { "epoch": 1.36, "learning_rate": 1.19955494076336e-05, "loss": 0.0001, "step": 2115 }, { "epoch": 1.36, "learning_rate": 1.1988748828991783e-05, "loss": 0.0, "step": 2116 }, { "epoch": 1.36, "learning_rate": 1.1981947292585597e-05, "loss": 0.0003, "step": 2117 }, { "epoch": 1.36, "learning_rate": 1.1975144801690601e-05, "loss": 0.0, "step": 2118 }, { "epoch": 1.36, "learning_rate": 1.1968341359582824e-05, "loss": 0.0, "step": 2119 }, { "epoch": 1.36, "learning_rate": 1.196153696953874e-05, "loss": 0.0001, "step": 2120 }, { "epoch": 1.36, "learning_rate": 1.1954731634835287e-05, "loss": 0.0001, "step": 2121 }, { "epoch": 1.36, "learning_rate": 1.1947925358749852e-05, "loss": 0.0001, "step": 2122 }, { "epoch": 1.36, "learning_rate": 1.1941118144560281e-05, "loss": 0.0001, "step": 2123 }, { "epoch": 1.36, "learning_rate": 1.1934309995544875e-05, "loss": 0.0002, "step": 2124 }, { "epoch": 1.37, "learning_rate": 1.1927500914982373e-05, "loss": 0.0001, "step": 2125 }, { "epoch": 1.37, "learning_rate": 1.1920690906151973e-05, "loss": 0.0002, "step": 2126 }, { "epoch": 1.37, "learning_rate": 1.1913879972333317e-05, "loss": 0.0001, "step": 2127 }, { "epoch": 1.37, "learning_rate": 1.1907068116806493e-05, "loss": 0.0, "step": 2128 }, { "epoch": 1.37, "learning_rate": 1.1900255342852026e-05, "loss": 0.0001, "step": 2129 }, { "epoch": 1.37, "learning_rate": 1.1893441653750897e-05, "loss": 0.0002, "step": 2130 }, { "epoch": 1.37, "learning_rate": 1.1886627052784518e-05, "loss": 0.0, "step": 2131 }, { "epoch": 1.37, "learning_rate": 1.1879811543234741e-05, "loss": 0.0001, "step": 2132 }, { "epoch": 1.37, "learning_rate": 1.187299512838386e-05, "loss": 0.0001, "step": 2133 }, { "epoch": 1.37, "learning_rate": 1.1866177811514596e-05, "loss": 0.0001, "step": 2134 }, { "epoch": 1.37, "learning_rate": 1.1859359595910116e-05, "loss": 0.0, "step": 2135 }, { "epoch": 1.37, "learning_rate": 1.1852540484854009e-05, "loss": 0.0001, "step": 2136 }, { "epoch": 1.37, "learning_rate": 1.1845720481630307e-05, "loss": 0.0001, "step": 2137 }, { "epoch": 1.37, "learning_rate": 1.1838899589523462e-05, "loss": 0.0, "step": 2138 }, { "epoch": 1.37, "learning_rate": 1.1832077811818357e-05, "loss": 0.0002, "step": 2139 }, { "epoch": 1.38, "learning_rate": 1.1825255151800297e-05, "loss": 0.0001, "step": 2140 }, { "epoch": 1.38, "learning_rate": 1.1818431612755022e-05, "loss": 0.0001, "step": 2141 }, { "epoch": 1.38, "learning_rate": 1.181160719796869e-05, "loss": 0.0002, "step": 2142 }, { "epoch": 1.38, "learning_rate": 1.1804781910727874e-05, "loss": 0.0001, "step": 2143 }, { "epoch": 1.38, "learning_rate": 1.1797955754319585e-05, "loss": 0.0001, "step": 2144 }, { "epoch": 1.38, "learning_rate": 1.1791128732031234e-05, "loss": 0.0, "step": 2145 }, { "epoch": 1.38, "learning_rate": 1.1784300847150656e-05, "loss": 0.0001, "step": 2146 }, { "epoch": 1.38, "learning_rate": 1.1777472102966101e-05, "loss": 0.0001, "step": 2147 }, { "epoch": 1.38, "learning_rate": 1.177064250276624e-05, "loss": 0.0, "step": 2148 }, { "epoch": 1.38, "learning_rate": 1.1763812049840144e-05, "loss": 0.0001, "step": 2149 }, { "epoch": 1.38, "learning_rate": 1.1756980747477302e-05, "loss": 0.0001, "step": 2150 }, { "epoch": 1.38, "learning_rate": 1.1750148598967613e-05, "loss": 0.0001, "step": 2151 }, { "epoch": 1.38, "learning_rate": 1.1743315607601376e-05, "loss": 0.0001, "step": 2152 }, { "epoch": 1.38, "learning_rate": 1.1736481776669307e-05, "loss": 0.0001, "step": 2153 }, { "epoch": 1.38, "learning_rate": 1.1729647109462512e-05, "loss": 0.0001, "step": 2154 }, { "epoch": 1.38, "learning_rate": 1.1722811609272517e-05, "loss": 0.0002, "step": 2155 }, { "epoch": 1.39, "learning_rate": 1.1715975279391239e-05, "loss": 0.0, "step": 2156 }, { "epoch": 1.39, "learning_rate": 1.1709138123110987e-05, "loss": 0.0001, "step": 2157 }, { "epoch": 1.39, "learning_rate": 1.1702300143724488e-05, "loss": 0.0001, "step": 2158 }, { "epoch": 1.39, "learning_rate": 1.1695461344524846e-05, "loss": 0.0, "step": 2159 }, { "epoch": 1.39, "learning_rate": 1.1688621728805573e-05, "loss": 0.0005, "step": 2160 }, { "epoch": 1.39, "learning_rate": 1.1681781299860567e-05, "loss": 0.0001, "step": 2161 }, { "epoch": 1.39, "learning_rate": 1.1674940060984121e-05, "loss": 0.0001, "step": 2162 }, { "epoch": 1.39, "learning_rate": 1.1668098015470917e-05, "loss": 0.0002, "step": 2163 }, { "epoch": 1.39, "learning_rate": 1.1661255166616024e-05, "loss": 0.0, "step": 2164 }, { "epoch": 1.39, "learning_rate": 1.1654411517714899e-05, "loss": 0.0001, "step": 2165 }, { "epoch": 1.39, "learning_rate": 1.1647567072063386e-05, "loss": 0.0, "step": 2166 }, { "epoch": 1.39, "learning_rate": 1.164072183295771e-05, "loss": 0.0001, "step": 2167 }, { "epoch": 1.39, "learning_rate": 1.1633875803694483e-05, "loss": 0.0, "step": 2168 }, { "epoch": 1.39, "learning_rate": 1.162702898757069e-05, "loss": 0.0003, "step": 2169 }, { "epoch": 1.39, "learning_rate": 1.1620181387883697e-05, "loss": 0.0, "step": 2170 }, { "epoch": 1.4, "learning_rate": 1.1613333007931255e-05, "loss": 0.0003, "step": 2171 }, { "epoch": 1.4, "learning_rate": 1.1606483851011483e-05, "loss": 0.0001, "step": 2172 }, { "epoch": 1.4, "learning_rate": 1.1599633920422877e-05, "loss": 0.0002, "step": 2173 }, { "epoch": 1.4, "learning_rate": 1.15927832194643e-05, "loss": 0.0001, "step": 2174 }, { "epoch": 1.4, "learning_rate": 1.1585931751434996e-05, "loss": 0.0001, "step": 2175 }, { "epoch": 1.4, "learning_rate": 1.1579079519634569e-05, "loss": 0.0001, "step": 2176 }, { "epoch": 1.4, "learning_rate": 1.1572226527363e-05, "loss": 0.0001, "step": 2177 }, { "epoch": 1.4, "learning_rate": 1.1565372777920627e-05, "loss": 0.0, "step": 2178 }, { "epoch": 1.4, "learning_rate": 1.155851827460816e-05, "loss": 0.0001, "step": 2179 }, { "epoch": 1.4, "learning_rate": 1.1551663020726666e-05, "loss": 0.0, "step": 2180 }, { "epoch": 1.4, "learning_rate": 1.1544807019577577e-05, "loss": 0.0001, "step": 2181 }, { "epoch": 1.4, "learning_rate": 1.1537950274462685e-05, "loss": 0.0001, "step": 2182 }, { "epoch": 1.4, "learning_rate": 1.1531092788684144e-05, "loss": 0.0008, "step": 2183 }, { "epoch": 1.4, "learning_rate": 1.1524234565544455e-05, "loss": 0.0003, "step": 2184 }, { "epoch": 1.4, "learning_rate": 1.151737560834648e-05, "loss": 0.0001, "step": 2185 }, { "epoch": 1.4, "learning_rate": 1.1510515920393433e-05, "loss": 0.0, "step": 2186 }, { "epoch": 1.41, "learning_rate": 1.1503655504988881e-05, "loss": 0.0001, "step": 2187 }, { "epoch": 1.41, "learning_rate": 1.1496794365436746e-05, "loss": 0.0001, "step": 2188 }, { "epoch": 1.41, "learning_rate": 1.1489932505041286e-05, "loss": 0.0001, "step": 2189 }, { "epoch": 1.41, "learning_rate": 1.1483069927107123e-05, "loss": 0.0001, "step": 2190 }, { "epoch": 1.41, "learning_rate": 1.1476206634939211e-05, "loss": 0.0002, "step": 2191 }, { "epoch": 1.41, "learning_rate": 1.146934263184285e-05, "loss": 0.0001, "step": 2192 }, { "epoch": 1.41, "learning_rate": 1.146247792112369e-05, "loss": 0.0001, "step": 2193 }, { "epoch": 1.41, "learning_rate": 1.145561250608771e-05, "loss": 0.0001, "step": 2194 }, { "epoch": 1.41, "learning_rate": 1.1448746390041246e-05, "loss": 0.0, "step": 2195 }, { "epoch": 1.41, "learning_rate": 1.1441879576290948e-05, "loss": 0.0002, "step": 2196 }, { "epoch": 1.41, "learning_rate": 1.1435012068143825e-05, "loss": 0.0002, "step": 2197 }, { "epoch": 1.41, "learning_rate": 1.1428143868907205e-05, "loss": 0.0001, "step": 2198 }, { "epoch": 1.41, "learning_rate": 1.142127498188875e-05, "loss": 0.0001, "step": 2199 }, { "epoch": 1.41, "learning_rate": 1.1414405410396463e-05, "loss": 0.0002, "step": 2200 }, { "epoch": 1.41, "learning_rate": 1.1407535157738671e-05, "loss": 0.0002, "step": 2201 }, { "epoch": 1.41, "learning_rate": 1.1400664227224029e-05, "loss": 0.0001, "step": 2202 }, { "epoch": 1.42, "learning_rate": 1.1393792622161513e-05, "loss": 0.0001, "step": 2203 }, { "epoch": 1.42, "learning_rate": 1.1386920345860433e-05, "loss": 0.0, "step": 2204 }, { "epoch": 1.42, "learning_rate": 1.1380047401630414e-05, "loss": 0.0001, "step": 2205 }, { "epoch": 1.42, "learning_rate": 1.1373173792781416e-05, "loss": 0.0001, "step": 2206 }, { "epoch": 1.42, "learning_rate": 1.1366299522623698e-05, "loss": 0.0001, "step": 2207 }, { "epoch": 1.42, "learning_rate": 1.135942459446786e-05, "loss": 0.0001, "step": 2208 }, { "epoch": 1.42, "learning_rate": 1.1352549011624805e-05, "loss": 0.0001, "step": 2209 }, { "epoch": 1.42, "learning_rate": 1.134567277740575e-05, "loss": 0.0001, "step": 2210 }, { "epoch": 1.42, "learning_rate": 1.1338795895122235e-05, "loss": 0.0001, "step": 2211 }, { "epoch": 1.42, "learning_rate": 1.133191836808611e-05, "loss": 0.0001, "step": 2212 }, { "epoch": 1.42, "learning_rate": 1.1325040199609526e-05, "loss": 0.0001, "step": 2213 }, { "epoch": 1.42, "learning_rate": 1.1318161393004956e-05, "loss": 0.0001, "step": 2214 }, { "epoch": 1.42, "learning_rate": 1.1311281951585171e-05, "loss": 0.0001, "step": 2215 }, { "epoch": 1.42, "learning_rate": 1.1304401878663254e-05, "loss": 0.0007, "step": 2216 }, { "epoch": 1.42, "learning_rate": 1.1297521177552584e-05, "loss": 0.0, "step": 2217 }, { "epoch": 1.43, "learning_rate": 1.1290639851566849e-05, "loss": 0.0001, "step": 2218 }, { "epoch": 1.43, "learning_rate": 1.128375790402004e-05, "loss": 0.0001, "step": 2219 }, { "epoch": 1.43, "learning_rate": 1.1276875338226445e-05, "loss": 0.0001, "step": 2220 }, { "epoch": 1.43, "learning_rate": 1.1269992157500642e-05, "loss": 0.0001, "step": 2221 }, { "epoch": 1.43, "learning_rate": 1.1263108365157515e-05, "loss": 0.0, "step": 2222 }, { "epoch": 1.43, "learning_rate": 1.125622396451224e-05, "loss": 0.0001, "step": 2223 }, { "epoch": 1.43, "learning_rate": 1.1249338958880285e-05, "loss": 0.0002, "step": 2224 }, { "epoch": 1.43, "learning_rate": 1.124245335157741e-05, "loss": 0.0001, "step": 2225 }, { "epoch": 1.43, "learning_rate": 1.1235567145919666e-05, "loss": 0.0001, "step": 2226 }, { "epoch": 1.43, "learning_rate": 1.1228680345223385e-05, "loss": 0.0002, "step": 2227 }, { "epoch": 1.43, "learning_rate": 1.1221792952805195e-05, "loss": 0.0001, "step": 2228 }, { "epoch": 1.43, "learning_rate": 1.1214904971982005e-05, "loss": 0.0001, "step": 2229 }, { "epoch": 1.43, "learning_rate": 1.1208016406071009e-05, "loss": 0.0, "step": 2230 }, { "epoch": 1.43, "learning_rate": 1.120112725838968e-05, "loss": 0.0002, "step": 2231 }, { "epoch": 1.43, "learning_rate": 1.1194237532255768e-05, "loss": 0.0001, "step": 2232 }, { "epoch": 1.43, "learning_rate": 1.1187347230987312e-05, "loss": 0.0001, "step": 2233 }, { "epoch": 1.44, "learning_rate": 1.1180456357902615e-05, "loss": 0.0, "step": 2234 }, { "epoch": 1.44, "learning_rate": 1.117356491632027e-05, "loss": 0.0, "step": 2235 }, { "epoch": 1.44, "learning_rate": 1.1166672909559135e-05, "loss": 0.0, "step": 2236 }, { "epoch": 1.44, "learning_rate": 1.115978034093834e-05, "loss": 0.0001, "step": 2237 }, { "epoch": 1.44, "learning_rate": 1.1152887213777285e-05, "loss": 0.0001, "step": 2238 }, { "epoch": 1.44, "learning_rate": 1.114599353139564e-05, "loss": 0.0001, "step": 2239 }, { "epoch": 1.44, "learning_rate": 1.1139099297113348e-05, "loss": 0.0, "step": 2240 }, { "epoch": 1.44, "learning_rate": 1.1132204514250608e-05, "loss": 0.0003, "step": 2241 }, { "epoch": 1.44, "learning_rate": 1.112530918612789e-05, "loss": 0.0003, "step": 2242 }, { "epoch": 1.44, "learning_rate": 1.111841331606593e-05, "loss": 0.0001, "step": 2243 }, { "epoch": 1.44, "learning_rate": 1.1111516907385712e-05, "loss": 0.0, "step": 2244 }, { "epoch": 1.44, "learning_rate": 1.1104619963408494e-05, "loss": 0.0001, "step": 2245 }, { "epoch": 1.44, "learning_rate": 1.109772248745578e-05, "loss": 0.0001, "step": 2246 }, { "epoch": 1.44, "learning_rate": 1.1090824482849334e-05, "loss": 0.0001, "step": 2247 }, { "epoch": 1.44, "learning_rate": 1.1083925952911184e-05, "loss": 0.0001, "step": 2248 }, { "epoch": 1.45, "learning_rate": 1.1077026900963593e-05, "loss": 0.0, "step": 2249 }, { "epoch": 1.45, "learning_rate": 1.1070127330329092e-05, "loss": 0.0001, "step": 2250 }, { "epoch": 1.45, "learning_rate": 1.1063227244330452e-05, "loss": 0.0001, "step": 2251 }, { "epoch": 1.45, "learning_rate": 1.1056326646290696e-05, "loss": 0.0, "step": 2252 }, { "epoch": 1.45, "learning_rate": 1.1049425539533092e-05, "loss": 0.0, "step": 2253 }, { "epoch": 1.45, "learning_rate": 1.1042523927381157e-05, "loss": 0.0001, "step": 2254 }, { "epoch": 1.45, "learning_rate": 1.1035621813158643e-05, "loss": 0.0001, "step": 2255 }, { "epoch": 1.45, "learning_rate": 1.1028719200189553e-05, "loss": 0.0001, "step": 2256 }, { "epoch": 1.45, "learning_rate": 1.1021816091798126e-05, "loss": 0.0, "step": 2257 }, { "epoch": 1.45, "learning_rate": 1.1014912491308835e-05, "loss": 0.0002, "step": 2258 }, { "epoch": 1.45, "learning_rate": 1.1008008402046405e-05, "loss": 0.0, "step": 2259 }, { "epoch": 1.45, "learning_rate": 1.1001103827335772e-05, "loss": 0.0001, "step": 2260 }, { "epoch": 1.45, "learning_rate": 1.0994198770502131e-05, "loss": 0.0002, "step": 2261 }, { "epoch": 1.45, "learning_rate": 1.0987293234870898e-05, "loss": 0.0001, "step": 2262 }, { "epoch": 1.45, "learning_rate": 1.098038722376771e-05, "loss": 0.0, "step": 2263 }, { "epoch": 1.45, "learning_rate": 1.097348074051845e-05, "loss": 0.0, "step": 2264 }, { "epoch": 1.46, "learning_rate": 1.0966573788449223e-05, "loss": 0.0, "step": 2265 }, { "epoch": 1.46, "learning_rate": 1.0959666370886356e-05, "loss": 0.0002, "step": 2266 }, { "epoch": 1.46, "learning_rate": 1.0952758491156394e-05, "loss": 0.0001, "step": 2267 }, { "epoch": 1.46, "learning_rate": 1.0945850152586126e-05, "loss": 0.0, "step": 2268 }, { "epoch": 1.46, "learning_rate": 1.093894135850254e-05, "loss": 0.0, "step": 2269 }, { "epoch": 1.46, "learning_rate": 1.0932032112232854e-05, "loss": 0.0001, "step": 2270 }, { "epoch": 1.46, "learning_rate": 1.0925122417104499e-05, "loss": 0.0001, "step": 2271 }, { "epoch": 1.46, "learning_rate": 1.0918212276445131e-05, "loss": 0.0003, "step": 2272 }, { "epoch": 1.46, "learning_rate": 1.091130169358261e-05, "loss": 0.0001, "step": 2273 }, { "epoch": 1.46, "learning_rate": 1.0904390671845013e-05, "loss": 0.0001, "step": 2274 }, { "epoch": 1.46, "learning_rate": 1.0897479214560631e-05, "loss": 0.0, "step": 2275 }, { "epoch": 1.46, "learning_rate": 1.0890567325057962e-05, "loss": 0.0, "step": 2276 }, { "epoch": 1.46, "learning_rate": 1.0883655006665711e-05, "loss": 0.0001, "step": 2277 }, { "epoch": 1.46, "learning_rate": 1.0876742262712793e-05, "loss": 0.0001, "step": 2278 }, { "epoch": 1.46, "learning_rate": 1.0869829096528329e-05, "loss": 0.0001, "step": 2279 }, { "epoch": 1.47, "learning_rate": 1.0862915511441634e-05, "loss": 0.0, "step": 2280 }, { "epoch": 1.47, "learning_rate": 1.0856001510782235e-05, "loss": 0.0001, "step": 2281 }, { "epoch": 1.47, "learning_rate": 1.0849087097879851e-05, "loss": 0.0001, "step": 2282 }, { "epoch": 1.47, "learning_rate": 1.0842172276064413e-05, "loss": 0.0, "step": 2283 }, { "epoch": 1.47, "learning_rate": 1.0835257048666034e-05, "loss": 0.0, "step": 2284 }, { "epoch": 1.47, "learning_rate": 1.0828341419015027e-05, "loss": 0.0003, "step": 2285 }, { "epoch": 1.47, "learning_rate": 1.0821425390441902e-05, "loss": 0.0001, "step": 2286 }, { "epoch": 1.47, "learning_rate": 1.0814508966277358e-05, "loss": 0.0001, "step": 2287 }, { "epoch": 1.47, "learning_rate": 1.0807592149852286e-05, "loss": 0.0001, "step": 2288 }, { "epoch": 1.47, "learning_rate": 1.0800674944497766e-05, "loss": 0.0001, "step": 2289 }, { "epoch": 1.47, "learning_rate": 1.0793757353545066e-05, "loss": 0.0001, "step": 2290 }, { "epoch": 1.47, "learning_rate": 1.0786839380325638e-05, "loss": 0.0, "step": 2291 }, { "epoch": 1.47, "learning_rate": 1.0779921028171112e-05, "loss": 0.0001, "step": 2292 }, { "epoch": 1.47, "learning_rate": 1.0773002300413313e-05, "loss": 0.0001, "step": 2293 }, { "epoch": 1.47, "learning_rate": 1.0766083200384243e-05, "loss": 0.0, "step": 2294 }, { "epoch": 1.47, "learning_rate": 1.0759163731416078e-05, "loss": 0.0, "step": 2295 }, { "epoch": 1.48, "learning_rate": 1.0752243896841174e-05, "loss": 0.0001, "step": 2296 }, { "epoch": 1.48, "learning_rate": 1.0745323699992068e-05, "loss": 0.0002, "step": 2297 }, { "epoch": 1.48, "learning_rate": 1.0738403144201464e-05, "loss": 0.0001, "step": 2298 }, { "epoch": 1.48, "learning_rate": 1.073148223280224e-05, "loss": 0.0, "step": 2299 }, { "epoch": 1.48, "learning_rate": 1.0724560969127449e-05, "loss": 0.0, "step": 2300 }, { "epoch": 1.48, "learning_rate": 1.0717639356510318e-05, "loss": 0.0001, "step": 2301 }, { "epoch": 1.48, "learning_rate": 1.071071739828423e-05, "loss": 0.0011, "step": 2302 }, { "epoch": 1.48, "learning_rate": 1.0703795097782743e-05, "loss": 0.0, "step": 2303 }, { "epoch": 1.48, "learning_rate": 1.0696872458339578e-05, "loss": 0.0001, "step": 2304 }, { "epoch": 1.48, "learning_rate": 1.0689949483288615e-05, "loss": 0.0001, "step": 2305 }, { "epoch": 1.48, "learning_rate": 1.0683026175963902e-05, "loss": 0.0, "step": 2306 }, { "epoch": 1.48, "learning_rate": 1.067610253969965e-05, "loss": 0.0001, "step": 2307 }, { "epoch": 1.48, "learning_rate": 1.0669178577830213e-05, "loss": 0.0001, "step": 2308 }, { "epoch": 1.48, "learning_rate": 1.0662254293690115e-05, "loss": 0.0001, "step": 2309 }, { "epoch": 1.48, "learning_rate": 1.0655329690614033e-05, "loss": 0.0001, "step": 2310 }, { "epoch": 1.48, "learning_rate": 1.0648404771936797e-05, "loss": 0.0002, "step": 2311 }, { "epoch": 1.49, "learning_rate": 1.064147954099339e-05, "loss": 0.0001, "step": 2312 }, { "epoch": 1.49, "learning_rate": 1.0634554001118938e-05, "loss": 0.0, "step": 2313 }, { "epoch": 1.49, "learning_rate": 1.0627628155648728e-05, "loss": 0.0, "step": 2314 }, { "epoch": 1.49, "learning_rate": 1.0620702007918183e-05, "loss": 0.0002, "step": 2315 }, { "epoch": 1.49, "learning_rate": 1.0613775561262877e-05, "loss": 0.0001, "step": 2316 }, { "epoch": 1.49, "learning_rate": 1.0606848819018526e-05, "loss": 0.0001, "step": 2317 }, { "epoch": 1.49, "learning_rate": 1.0599921784520998e-05, "loss": 0.0002, "step": 2318 }, { "epoch": 1.49, "learning_rate": 1.0592994461106285e-05, "loss": 0.0001, "step": 2319 }, { "epoch": 1.49, "learning_rate": 1.0586066852110527e-05, "loss": 0.0002, "step": 2320 }, { "epoch": 1.49, "learning_rate": 1.0579138960870005e-05, "loss": 0.0001, "step": 2321 }, { "epoch": 1.49, "learning_rate": 1.0572210790721129e-05, "loss": 0.0001, "step": 2322 }, { "epoch": 1.49, "learning_rate": 1.0565282345000446e-05, "loss": 0.0, "step": 2323 }, { "epoch": 1.49, "learning_rate": 1.0558353627044635e-05, "loss": 0.0, "step": 2324 }, { "epoch": 1.49, "learning_rate": 1.0551424640190513e-05, "loss": 0.0, "step": 2325 }, { "epoch": 1.49, "learning_rate": 1.0544495387775017e-05, "loss": 0.0, "step": 2326 }, { "epoch": 1.5, "learning_rate": 1.0537565873135211e-05, "loss": 0.0, "step": 2327 }, { "epoch": 1.5, "learning_rate": 1.0530636099608299e-05, "loss": 0.0001, "step": 2328 }, { "epoch": 1.5, "learning_rate": 1.0523706070531592e-05, "loss": 0.0001, "step": 2329 }, { "epoch": 1.5, "learning_rate": 1.0516775789242538e-05, "loss": 0.0, "step": 2330 }, { "epoch": 1.5, "learning_rate": 1.0509845259078697e-05, "loss": 0.0001, "step": 2331 }, { "epoch": 1.5, "learning_rate": 1.0502914483377758e-05, "loss": 0.0, "step": 2332 }, { "epoch": 1.5, "learning_rate": 1.0495983465477518e-05, "loss": 0.0, "step": 2333 }, { "epoch": 1.5, "learning_rate": 1.0489052208715896e-05, "loss": 0.0, "step": 2334 }, { "epoch": 1.5, "learning_rate": 1.0482120716430928e-05, "loss": 0.0001, "step": 2335 }, { "epoch": 1.5, "learning_rate": 1.047518899196076e-05, "loss": 0.0, "step": 2336 }, { "epoch": 1.5, "learning_rate": 1.0468257038643657e-05, "loss": 0.0, "step": 2337 }, { "epoch": 1.5, "learning_rate": 1.0461324859817976e-05, "loss": 0.0, "step": 2338 }, { "epoch": 1.5, "learning_rate": 1.0454392458822207e-05, "loss": 0.0001, "step": 2339 }, { "epoch": 1.5, "learning_rate": 1.0447459838994924e-05, "loss": 0.0, "step": 2340 }, { "epoch": 1.5, "learning_rate": 1.0440527003674825e-05, "loss": 0.0002, "step": 2341 }, { "epoch": 1.5, "learning_rate": 1.0433593956200705e-05, "loss": 0.0001, "step": 2342 }, { "epoch": 1.51, "learning_rate": 1.0426660699911454e-05, "loss": 0.0, "step": 2343 }, { "epoch": 1.51, "learning_rate": 1.0419727238146073e-05, "loss": 0.0, "step": 2344 }, { "epoch": 1.51, "learning_rate": 1.0412793574243656e-05, "loss": 0.0001, "step": 2345 }, { "epoch": 1.51, "learning_rate": 1.0405859711543393e-05, "loss": 0.0003, "step": 2346 }, { "epoch": 1.51, "learning_rate": 1.0398925653384581e-05, "loss": 0.0, "step": 2347 }, { "epoch": 1.51, "learning_rate": 1.0391991403106593e-05, "loss": 0.0001, "step": 2348 }, { "epoch": 1.51, "learning_rate": 1.0385056964048914e-05, "loss": 0.0001, "step": 2349 }, { "epoch": 1.51, "learning_rate": 1.0378122339551107e-05, "loss": 0.0001, "step": 2350 }, { "epoch": 1.51, "learning_rate": 1.0371187532952825e-05, "loss": 0.0, "step": 2351 }, { "epoch": 1.51, "learning_rate": 1.0364252547593809e-05, "loss": 0.0001, "step": 2352 }, { "epoch": 1.51, "learning_rate": 1.0357317386813892e-05, "loss": 0.0, "step": 2353 }, { "epoch": 1.51, "learning_rate": 1.0350382053952993e-05, "loss": 0.0, "step": 2354 }, { "epoch": 1.51, "learning_rate": 1.0343446552351099e-05, "loss": 0.0, "step": 2355 }, { "epoch": 1.51, "learning_rate": 1.0336510885348297e-05, "loss": 0.0, "step": 2356 }, { "epoch": 1.51, "learning_rate": 1.032957505628474e-05, "loss": 0.0001, "step": 2357 }, { "epoch": 1.52, "learning_rate": 1.0322639068500662e-05, "loss": 0.0, "step": 2358 }, { "epoch": 1.52, "learning_rate": 1.0315702925336377e-05, "loss": 0.0001, "step": 2359 }, { "epoch": 1.52, "learning_rate": 1.0308766630132274e-05, "loss": 0.0001, "step": 2360 }, { "epoch": 1.52, "learning_rate": 1.0301830186228813e-05, "loss": 0.0, "step": 2361 }, { "epoch": 1.52, "learning_rate": 1.0294893596966527e-05, "loss": 0.0, "step": 2362 }, { "epoch": 1.52, "learning_rate": 1.028795686568601e-05, "loss": 0.0001, "step": 2363 }, { "epoch": 1.52, "learning_rate": 1.0281019995727939e-05, "loss": 0.0, "step": 2364 }, { "epoch": 1.52, "learning_rate": 1.0274082990433051e-05, "loss": 0.0001, "step": 2365 }, { "epoch": 1.52, "learning_rate": 1.0267145853142147e-05, "loss": 0.0, "step": 2366 }, { "epoch": 1.52, "learning_rate": 1.0260208587196094e-05, "loss": 0.0, "step": 2367 }, { "epoch": 1.52, "learning_rate": 1.0253271195935818e-05, "loss": 0.0, "step": 2368 }, { "epoch": 1.52, "learning_rate": 1.0246333682702305e-05, "loss": 0.0001, "step": 2369 }, { "epoch": 1.52, "learning_rate": 1.0239396050836608e-05, "loss": 0.0001, "step": 2370 }, { "epoch": 1.52, "learning_rate": 1.0232458303679827e-05, "loss": 0.0001, "step": 2371 }, { "epoch": 1.52, "learning_rate": 1.022552044457312e-05, "loss": 0.0001, "step": 2372 }, { "epoch": 1.52, "learning_rate": 1.0218582476857704e-05, "loss": 0.0, "step": 2373 }, { "epoch": 1.53, "learning_rate": 1.0211644403874843e-05, "loss": 0.0, "step": 2374 }, { "epoch": 1.53, "learning_rate": 1.020470622896585e-05, "loss": 0.0, "step": 2375 }, { "epoch": 1.53, "learning_rate": 1.01977679554721e-05, "loss": 0.0, "step": 2376 }, { "epoch": 1.53, "learning_rate": 1.0190829586734994e-05, "loss": 0.0001, "step": 2377 }, { "epoch": 1.53, "learning_rate": 1.0183891126096002e-05, "loss": 0.0002, "step": 2378 }, { "epoch": 1.53, "learning_rate": 1.017695257689662e-05, "loss": 0.0, "step": 2379 }, { "epoch": 1.53, "learning_rate": 1.0170013942478393e-05, "loss": 0.0006, "step": 2380 }, { "epoch": 1.53, "learning_rate": 1.0163075226182914e-05, "loss": 0.0001, "step": 2381 }, { "epoch": 1.53, "learning_rate": 1.0156136431351802e-05, "loss": 0.0, "step": 2382 }, { "epoch": 1.53, "learning_rate": 1.014919756132673e-05, "loss": 0.0, "step": 2383 }, { "epoch": 1.53, "learning_rate": 1.0142258619449388e-05, "loss": 0.0, "step": 2384 }, { "epoch": 1.53, "learning_rate": 1.0135319609061519e-05, "loss": 0.0, "step": 2385 }, { "epoch": 1.53, "learning_rate": 1.0128380533504884e-05, "loss": 0.0001, "step": 2386 }, { "epoch": 1.53, "learning_rate": 1.0121441396121288e-05, "loss": 0.0001, "step": 2387 }, { "epoch": 1.53, "learning_rate": 1.0114502200252556e-05, "loss": 0.0, "step": 2388 }, { "epoch": 1.54, "learning_rate": 1.010756294924055e-05, "loss": 0.0001, "step": 2389 }, { "epoch": 1.54, "learning_rate": 1.010062364642715e-05, "loss": 0.0003, "step": 2390 }, { "epoch": 1.54, "learning_rate": 1.0093684295154264e-05, "loss": 0.0, "step": 2391 }, { "epoch": 1.54, "learning_rate": 1.0086744898763828e-05, "loss": 0.0001, "step": 2392 }, { "epoch": 1.54, "learning_rate": 1.007980546059779e-05, "loss": 0.0001, "step": 2393 }, { "epoch": 1.54, "learning_rate": 1.0072865983998128e-05, "loss": 0.0001, "step": 2394 }, { "epoch": 1.54, "learning_rate": 1.0065926472306834e-05, "loss": 0.0001, "step": 2395 }, { "epoch": 1.54, "learning_rate": 1.005898692886592e-05, "loss": 0.0, "step": 2396 }, { "epoch": 1.54, "learning_rate": 1.0052047357017404e-05, "loss": 0.0, "step": 2397 }, { "epoch": 1.54, "learning_rate": 1.0045107760103327e-05, "loss": 0.0002, "step": 2398 }, { "epoch": 1.54, "learning_rate": 1.0038168141465741e-05, "loss": 0.0, "step": 2399 }, { "epoch": 1.54, "learning_rate": 1.0031228504446704e-05, "loss": 0.0002, "step": 2400 }, { "epoch": 1.54, "learning_rate": 1.0024288852388288e-05, "loss": 0.0, "step": 2401 }, { "epoch": 1.54, "learning_rate": 1.0017349188632571e-05, "loss": 0.0001, "step": 2402 }, { "epoch": 1.54, "learning_rate": 1.0010409516521633e-05, "loss": 0.0001, "step": 2403 }, { "epoch": 1.54, "learning_rate": 1.0003469839397562e-05, "loss": 0.0001, "step": 2404 }, { "epoch": 1.55, "learning_rate": 9.996530160602441e-06, "loss": 0.0001, "step": 2405 }, { "epoch": 1.55, "learning_rate": 9.989590483478372e-06, "loss": 0.0, "step": 2406 }, { "epoch": 1.55, "learning_rate": 9.98265081136743e-06, "loss": 0.0011, "step": 2407 }, { "epoch": 1.55, "learning_rate": 9.975711147611712e-06, "loss": 0.0, "step": 2408 }, { "epoch": 1.55, "learning_rate": 9.9687714955533e-06, "loss": 0.0001, "step": 2409 }, { "epoch": 1.55, "learning_rate": 9.961831858534264e-06, "loss": 0.0001, "step": 2410 }, { "epoch": 1.55, "learning_rate": 9.954892239896675e-06, "loss": 0.0001, "step": 2411 }, { "epoch": 1.55, "learning_rate": 9.9479526429826e-06, "loss": 0.0, "step": 2412 }, { "epoch": 1.55, "learning_rate": 9.941013071134083e-06, "loss": 0.0001, "step": 2413 }, { "epoch": 1.55, "learning_rate": 9.934073527693166e-06, "loss": 0.0, "step": 2414 }, { "epoch": 1.55, "learning_rate": 9.927134016001875e-06, "loss": 0.0, "step": 2415 }, { "epoch": 1.55, "learning_rate": 9.920194539402213e-06, "loss": 0.0001, "step": 2416 }, { "epoch": 1.55, "learning_rate": 9.913255101236174e-06, "loss": 0.0, "step": 2417 }, { "epoch": 1.55, "learning_rate": 9.906315704845738e-06, "loss": 0.0, "step": 2418 }, { "epoch": 1.55, "learning_rate": 9.899376353572851e-06, "loss": 0.0, "step": 2419 }, { "epoch": 1.56, "learning_rate": 9.892437050759456e-06, "loss": 0.0001, "step": 2420 }, { "epoch": 1.56, "learning_rate": 9.885497799747447e-06, "loss": 0.0001, "step": 2421 }, { "epoch": 1.56, "learning_rate": 9.878558603878715e-06, "loss": 0.0001, "step": 2422 }, { "epoch": 1.56, "learning_rate": 9.87161946649512e-06, "loss": 0.0, "step": 2423 }, { "epoch": 1.56, "learning_rate": 9.864680390938486e-06, "loss": 0.0, "step": 2424 }, { "epoch": 1.56, "learning_rate": 9.857741380550613e-06, "loss": 0.0002, "step": 2425 }, { "epoch": 1.56, "learning_rate": 9.850802438673277e-06, "loss": 0.0, "step": 2426 }, { "epoch": 1.56, "learning_rate": 9.8438635686482e-06, "loss": 0.0001, "step": 2427 }, { "epoch": 1.56, "learning_rate": 9.83692477381709e-06, "loss": 0.0001, "step": 2428 }, { "epoch": 1.56, "learning_rate": 9.82998605752161e-06, "loss": 0.0, "step": 2429 }, { "epoch": 1.56, "learning_rate": 9.823047423103383e-06, "loss": 0.0, "step": 2430 }, { "epoch": 1.56, "learning_rate": 9.816108873904005e-06, "loss": 0.0001, "step": 2431 }, { "epoch": 1.56, "learning_rate": 9.809170413265007e-06, "loss": 0.0003, "step": 2432 }, { "epoch": 1.56, "learning_rate": 9.802232044527904e-06, "loss": 0.0001, "step": 2433 }, { "epoch": 1.56, "learning_rate": 9.795293771034153e-06, "loss": 0.0001, "step": 2434 }, { "epoch": 1.56, "learning_rate": 9.78835559612516e-06, "loss": 0.0, "step": 2435 }, { "epoch": 1.57, "learning_rate": 9.781417523142298e-06, "loss": 0.0002, "step": 2436 }, { "epoch": 1.57, "learning_rate": 9.774479555426885e-06, "loss": 0.0001, "step": 2437 }, { "epoch": 1.57, "learning_rate": 9.767541696320178e-06, "loss": 0.0002, "step": 2438 }, { "epoch": 1.57, "learning_rate": 9.760603949163394e-06, "loss": 0.0, "step": 2439 }, { "epoch": 1.57, "learning_rate": 9.753666317297697e-06, "loss": 0.0, "step": 2440 }, { "epoch": 1.57, "learning_rate": 9.746728804064185e-06, "loss": 0.0001, "step": 2441 }, { "epoch": 1.57, "learning_rate": 9.739791412803907e-06, "loss": 0.0001, "step": 2442 }, { "epoch": 1.57, "learning_rate": 9.732854146857856e-06, "loss": 0.0, "step": 2443 }, { "epoch": 1.57, "learning_rate": 9.72591700956695e-06, "loss": 0.0001, "step": 2444 }, { "epoch": 1.57, "learning_rate": 9.718980004272063e-06, "loss": 0.0, "step": 2445 }, { "epoch": 1.57, "learning_rate": 9.712043134313993e-06, "loss": 0.0001, "step": 2446 }, { "epoch": 1.57, "learning_rate": 9.705106403033478e-06, "loss": 0.0001, "step": 2447 }, { "epoch": 1.57, "learning_rate": 9.698169813771192e-06, "loss": 0.0, "step": 2448 }, { "epoch": 1.57, "learning_rate": 9.691233369867728e-06, "loss": 0.0, "step": 2449 }, { "epoch": 1.57, "learning_rate": 9.684297074663625e-06, "loss": 0.0001, "step": 2450 }, { "epoch": 1.57, "learning_rate": 9.677360931499343e-06, "loss": 0.0001, "step": 2451 }, { "epoch": 1.58, "learning_rate": 9.670424943715264e-06, "loss": 0.0, "step": 2452 }, { "epoch": 1.58, "learning_rate": 9.663489114651705e-06, "loss": 0.0, "step": 2453 }, { "epoch": 1.58, "learning_rate": 9.656553447648905e-06, "loss": 0.0001, "step": 2454 }, { "epoch": 1.58, "learning_rate": 9.64961794604701e-06, "loss": 0.0001, "step": 2455 }, { "epoch": 1.58, "learning_rate": 9.642682613186108e-06, "loss": 0.0, "step": 2456 }, { "epoch": 1.58, "learning_rate": 9.635747452406194e-06, "loss": 0.0, "step": 2457 }, { "epoch": 1.58, "learning_rate": 9.62881246704718e-06, "loss": 0.0, "step": 2458 }, { "epoch": 1.58, "learning_rate": 9.6218776604489e-06, "loss": 0.0001, "step": 2459 }, { "epoch": 1.58, "learning_rate": 9.614943035951088e-06, "loss": 0.0, "step": 2460 }, { "epoch": 1.58, "learning_rate": 9.608008596893405e-06, "loss": 0.0, "step": 2461 }, { "epoch": 1.58, "learning_rate": 9.601074346615424e-06, "loss": 0.0, "step": 2462 }, { "epoch": 1.58, "learning_rate": 9.594140288456609e-06, "loss": 0.0, "step": 2463 }, { "epoch": 1.58, "learning_rate": 9.587206425756347e-06, "loss": 0.0, "step": 2464 }, { "epoch": 1.58, "learning_rate": 9.58027276185393e-06, "loss": 0.0001, "step": 2465 }, { "epoch": 1.58, "learning_rate": 9.573339300088547e-06, "loss": 0.0, "step": 2466 }, { "epoch": 1.59, "learning_rate": 9.566406043799296e-06, "loss": 0.0, "step": 2467 }, { "epoch": 1.59, "learning_rate": 9.559472996325176e-06, "loss": 0.0, "step": 2468 }, { "epoch": 1.59, "learning_rate": 9.552540161005078e-06, "loss": 0.0, "step": 2469 }, { "epoch": 1.59, "learning_rate": 9.545607541177796e-06, "loss": 0.0, "step": 2470 }, { "epoch": 1.59, "learning_rate": 9.538675140182027e-06, "loss": 0.0001, "step": 2471 }, { "epoch": 1.59, "learning_rate": 9.531742961356348e-06, "loss": 0.0, "step": 2472 }, { "epoch": 1.59, "learning_rate": 9.524811008039243e-06, "loss": 0.0, "step": 2473 }, { "epoch": 1.59, "learning_rate": 9.517879283569075e-06, "loss": 0.0, "step": 2474 }, { "epoch": 1.59, "learning_rate": 9.510947791284108e-06, "loss": 0.0001, "step": 2475 }, { "epoch": 1.59, "learning_rate": 9.504016534522487e-06, "loss": 0.0, "step": 2476 }, { "epoch": 1.59, "learning_rate": 9.497085516622247e-06, "loss": 0.0001, "step": 2477 }, { "epoch": 1.59, "learning_rate": 9.490154740921303e-06, "loss": 0.0, "step": 2478 }, { "epoch": 1.59, "learning_rate": 9.483224210757467e-06, "loss": 0.0, "step": 2479 }, { "epoch": 1.59, "learning_rate": 9.476293929468411e-06, "loss": 0.0003, "step": 2480 }, { "epoch": 1.59, "learning_rate": 9.469363900391703e-06, "loss": 0.0001, "step": 2481 }, { "epoch": 1.59, "learning_rate": 9.46243412686479e-06, "loss": 0.0001, "step": 2482 }, { "epoch": 1.6, "learning_rate": 9.455504612224986e-06, "loss": 0.0001, "step": 2483 }, { "epoch": 1.6, "learning_rate": 9.44857535980949e-06, "loss": 0.0, "step": 2484 }, { "epoch": 1.6, "learning_rate": 9.441646372955366e-06, "loss": 0.0001, "step": 2485 }, { "epoch": 1.6, "learning_rate": 9.434717654999556e-06, "loss": 0.0, "step": 2486 }, { "epoch": 1.6, "learning_rate": 9.427789209278876e-06, "loss": 0.0, "step": 2487 }, { "epoch": 1.6, "learning_rate": 9.420861039129998e-06, "loss": 0.0, "step": 2488 }, { "epoch": 1.6, "learning_rate": 9.413933147889474e-06, "loss": 0.0, "step": 2489 }, { "epoch": 1.6, "learning_rate": 9.40700553889372e-06, "loss": 0.0, "step": 2490 }, { "epoch": 1.6, "learning_rate": 9.400078215479005e-06, "loss": 0.0002, "step": 2491 }, { "epoch": 1.6, "learning_rate": 9.393151180981472e-06, "loss": 0.0, "step": 2492 }, { "epoch": 1.6, "learning_rate": 9.386224438737126e-06, "loss": 0.0, "step": 2493 }, { "epoch": 1.6, "learning_rate": 9.379297992081819e-06, "loss": 0.0, "step": 2494 }, { "epoch": 1.6, "learning_rate": 9.372371844351274e-06, "loss": 0.0, "step": 2495 }, { "epoch": 1.6, "learning_rate": 9.365445998881065e-06, "loss": 0.0001, "step": 2496 }, { "epoch": 1.6, "learning_rate": 9.358520459006612e-06, "loss": 0.0002, "step": 2497 }, { "epoch": 1.61, "learning_rate": 9.351595228063204e-06, "loss": 0.0001, "step": 2498 }, { "epoch": 1.61, "learning_rate": 9.344670309385969e-06, "loss": 0.0, "step": 2499 }, { "epoch": 1.61, "learning_rate": 9.337745706309887e-06, "loss": 0.0001, "step": 2500 }, { "epoch": 1.61, "learning_rate": 9.330821422169792e-06, "loss": 0.0002, "step": 2501 }, { "epoch": 1.61, "learning_rate": 9.323897460300356e-06, "loss": 0.0001, "step": 2502 }, { "epoch": 1.61, "learning_rate": 9.316973824036098e-06, "loss": 0.0, "step": 2503 }, { "epoch": 1.61, "learning_rate": 9.310050516711388e-06, "loss": 0.0001, "step": 2504 }, { "epoch": 1.61, "learning_rate": 9.303127541660425e-06, "loss": 0.0001, "step": 2505 }, { "epoch": 1.61, "learning_rate": 9.296204902217257e-06, "loss": 0.0, "step": 2506 }, { "epoch": 1.61, "learning_rate": 9.289282601715773e-06, "loss": 0.0, "step": 2507 }, { "epoch": 1.61, "learning_rate": 9.282360643489685e-06, "loss": 0.0, "step": 2508 }, { "epoch": 1.61, "learning_rate": 9.275439030872551e-06, "loss": 0.0, "step": 2509 }, { "epoch": 1.61, "learning_rate": 9.268517767197764e-06, "loss": 0.0001, "step": 2510 }, { "epoch": 1.61, "learning_rate": 9.26159685579854e-06, "loss": 0.0, "step": 2511 }, { "epoch": 1.61, "learning_rate": 9.254676300007937e-06, "loss": 0.0, "step": 2512 }, { "epoch": 1.61, "learning_rate": 9.247756103158827e-06, "loss": 0.0001, "step": 2513 }, { "epoch": 1.62, "learning_rate": 9.240836268583924e-06, "loss": 0.0001, "step": 2514 }, { "epoch": 1.62, "learning_rate": 9.23391679961576e-06, "loss": 0.0001, "step": 2515 }, { "epoch": 1.62, "learning_rate": 9.22699769958669e-06, "loss": 0.0001, "step": 2516 }, { "epoch": 1.62, "learning_rate": 9.220078971828888e-06, "loss": 0.0, "step": 2517 }, { "epoch": 1.62, "learning_rate": 9.213160619674368e-06, "loss": 0.0, "step": 2518 }, { "epoch": 1.62, "learning_rate": 9.206242646454937e-06, "loss": 0.0001, "step": 2519 }, { "epoch": 1.62, "learning_rate": 9.199325055502234e-06, "loss": 0.0002, "step": 2520 }, { "epoch": 1.62, "learning_rate": 9.192407850147717e-06, "loss": 0.0, "step": 2521 }, { "epoch": 1.62, "learning_rate": 9.185491033722646e-06, "loss": 0.0001, "step": 2522 }, { "epoch": 1.62, "learning_rate": 9.178574609558103e-06, "loss": 0.0001, "step": 2523 }, { "epoch": 1.62, "learning_rate": 9.171658580984976e-06, "loss": 0.0001, "step": 2524 }, { "epoch": 1.62, "learning_rate": 9.16474295133397e-06, "loss": 0.0002, "step": 2525 }, { "epoch": 1.62, "learning_rate": 9.157827723935592e-06, "loss": 0.0, "step": 2526 }, { "epoch": 1.62, "learning_rate": 9.15091290212015e-06, "loss": 0.0, "step": 2527 }, { "epoch": 1.62, "learning_rate": 9.143998489217769e-06, "loss": 0.0, "step": 2528 }, { "epoch": 1.63, "learning_rate": 9.13708448855837e-06, "loss": 0.0, "step": 2529 }, { "epoch": 1.63, "learning_rate": 9.130170903471675e-06, "loss": 0.0, "step": 2530 }, { "epoch": 1.63, "learning_rate": 9.123257737287207e-06, "loss": 0.0001, "step": 2531 }, { "epoch": 1.63, "learning_rate": 9.116344993334294e-06, "loss": 0.0, "step": 2532 }, { "epoch": 1.63, "learning_rate": 9.109432674942041e-06, "loss": 0.0003, "step": 2533 }, { "epoch": 1.63, "learning_rate": 9.10252078543937e-06, "loss": 0.0002, "step": 2534 }, { "epoch": 1.63, "learning_rate": 9.09560932815499e-06, "loss": 0.0, "step": 2535 }, { "epoch": 1.63, "learning_rate": 9.088698306417394e-06, "loss": 0.0001, "step": 2536 }, { "epoch": 1.63, "learning_rate": 9.081787723554874e-06, "loss": 0.0003, "step": 2537 }, { "epoch": 1.63, "learning_rate": 9.074877582895503e-06, "loss": 0.0, "step": 2538 }, { "epoch": 1.63, "learning_rate": 9.06796788776715e-06, "loss": 0.0, "step": 2539 }, { "epoch": 1.63, "learning_rate": 9.061058641497464e-06, "loss": 0.0001, "step": 2540 }, { "epoch": 1.63, "learning_rate": 9.054149847413876e-06, "loss": 0.0, "step": 2541 }, { "epoch": 1.63, "learning_rate": 9.047241508843606e-06, "loss": 0.0, "step": 2542 }, { "epoch": 1.63, "learning_rate": 9.040333629113651e-06, "loss": 0.0002, "step": 2543 }, { "epoch": 1.63, "learning_rate": 9.03342621155078e-06, "loss": 0.0, "step": 2544 }, { "epoch": 1.64, "learning_rate": 9.02651925948155e-06, "loss": 0.0, "step": 2545 }, { "epoch": 1.64, "learning_rate": 9.019612776232293e-06, "loss": 0.0001, "step": 2546 }, { "epoch": 1.64, "learning_rate": 9.012706765129106e-06, "loss": 0.0, "step": 2547 }, { "epoch": 1.64, "learning_rate": 9.005801229497869e-06, "loss": 0.0001, "step": 2548 }, { "epoch": 1.64, "learning_rate": 8.99889617266423e-06, "loss": 0.0001, "step": 2549 }, { "epoch": 1.64, "learning_rate": 8.9919915979536e-06, "loss": 0.0001, "step": 2550 }, { "epoch": 1.64, "learning_rate": 8.985087508691167e-06, "loss": 0.0, "step": 2551 }, { "epoch": 1.64, "learning_rate": 8.978183908201875e-06, "loss": 0.0, "step": 2552 }, { "epoch": 1.64, "learning_rate": 8.971280799810447e-06, "loss": 0.0, "step": 2553 }, { "epoch": 1.64, "learning_rate": 8.96437818684136e-06, "loss": 0.0001, "step": 2554 }, { "epoch": 1.64, "learning_rate": 8.957476072618846e-06, "loss": 0.0001, "step": 2555 }, { "epoch": 1.64, "learning_rate": 8.950574460466908e-06, "loss": 0.0002, "step": 2556 }, { "epoch": 1.64, "learning_rate": 8.943673353709307e-06, "loss": 0.0, "step": 2557 }, { "epoch": 1.64, "learning_rate": 8.93677275566955e-06, "loss": 0.0, "step": 2558 }, { "epoch": 1.64, "learning_rate": 8.92987266967091e-06, "loss": 0.0, "step": 2559 }, { "epoch": 1.64, "learning_rate": 8.92297309903641e-06, "loss": 0.0, "step": 2560 }, { "epoch": 1.65, "learning_rate": 8.916074047088821e-06, "loss": 0.0, "step": 2561 }, { "epoch": 1.65, "learning_rate": 8.90917551715067e-06, "loss": 0.0, "step": 2562 }, { "epoch": 1.65, "learning_rate": 8.902277512544224e-06, "loss": 0.0001, "step": 2563 }, { "epoch": 1.65, "learning_rate": 8.895380036591508e-06, "loss": 0.0, "step": 2564 }, { "epoch": 1.65, "learning_rate": 8.888483092614291e-06, "loss": 0.0, "step": 2565 }, { "epoch": 1.65, "learning_rate": 8.881586683934073e-06, "loss": 0.0, "step": 2566 }, { "epoch": 1.65, "learning_rate": 8.87469081387211e-06, "loss": 0.0001, "step": 2567 }, { "epoch": 1.65, "learning_rate": 8.867795485749397e-06, "loss": 0.0001, "step": 2568 }, { "epoch": 1.65, "learning_rate": 8.860900702886656e-06, "loss": 0.0001, "step": 2569 }, { "epoch": 1.65, "learning_rate": 8.854006468604361e-06, "loss": 0.0001, "step": 2570 }, { "epoch": 1.65, "learning_rate": 8.84711278622272e-06, "loss": 0.0, "step": 2571 }, { "epoch": 1.65, "learning_rate": 8.840219659061665e-06, "loss": 0.0, "step": 2572 }, { "epoch": 1.65, "learning_rate": 8.833327090440867e-06, "loss": 0.0001, "step": 2573 }, { "epoch": 1.65, "learning_rate": 8.826435083679733e-06, "loss": 0.0, "step": 2574 }, { "epoch": 1.65, "learning_rate": 8.819543642097386e-06, "loss": 0.0, "step": 2575 }, { "epoch": 1.66, "learning_rate": 8.812652769012695e-06, "loss": 0.0, "step": 2576 }, { "epoch": 1.66, "learning_rate": 8.805762467744237e-06, "loss": 0.0001, "step": 2577 }, { "epoch": 1.66, "learning_rate": 8.798872741610325e-06, "loss": 0.0001, "step": 2578 }, { "epoch": 1.66, "learning_rate": 8.791983593928996e-06, "loss": 0.0001, "step": 2579 }, { "epoch": 1.66, "learning_rate": 8.785095028017998e-06, "loss": 0.0001, "step": 2580 }, { "epoch": 1.66, "learning_rate": 8.778207047194806e-06, "loss": 0.0001, "step": 2581 }, { "epoch": 1.66, "learning_rate": 8.77131965477662e-06, "loss": 0.0001, "step": 2582 }, { "epoch": 1.66, "learning_rate": 8.764432854080339e-06, "loss": 0.0, "step": 2583 }, { "epoch": 1.66, "learning_rate": 8.757546648422591e-06, "loss": 0.0, "step": 2584 }, { "epoch": 1.66, "learning_rate": 8.750661041119718e-06, "loss": 0.0, "step": 2585 }, { "epoch": 1.66, "learning_rate": 8.743776035487762e-06, "loss": 0.0, "step": 2586 }, { "epoch": 1.66, "learning_rate": 8.736891634842488e-06, "loss": 0.0, "step": 2587 }, { "epoch": 1.66, "learning_rate": 8.730007842499361e-06, "loss": 0.0, "step": 2588 }, { "epoch": 1.66, "learning_rate": 8.723124661773559e-06, "loss": 0.0001, "step": 2589 }, { "epoch": 1.66, "learning_rate": 8.716242095979962e-06, "loss": 0.0001, "step": 2590 }, { "epoch": 1.66, "learning_rate": 8.709360148433153e-06, "loss": 0.0, "step": 2591 }, { "epoch": 1.67, "learning_rate": 8.70247882244742e-06, "loss": 0.0, "step": 2592 }, { "epoch": 1.67, "learning_rate": 8.695598121336751e-06, "loss": 0.0001, "step": 2593 }, { "epoch": 1.67, "learning_rate": 8.68871804841483e-06, "loss": 0.0001, "step": 2594 }, { "epoch": 1.67, "learning_rate": 8.681838606995046e-06, "loss": 0.0, "step": 2595 }, { "epoch": 1.67, "learning_rate": 8.674959800390479e-06, "loss": 0.0001, "step": 2596 }, { "epoch": 1.67, "learning_rate": 8.668081631913894e-06, "loss": 0.0001, "step": 2597 }, { "epoch": 1.67, "learning_rate": 8.661204104877765e-06, "loss": 0.0, "step": 2598 }, { "epoch": 1.67, "learning_rate": 8.654327222594251e-06, "loss": 0.0, "step": 2599 }, { "epoch": 1.67, "learning_rate": 8.647450988375199e-06, "loss": 0.0, "step": 2600 }, { "epoch": 1.67, "learning_rate": 8.64057540553214e-06, "loss": 0.0, "step": 2601 }, { "epoch": 1.67, "learning_rate": 8.633700477376304e-06, "loss": 0.0001, "step": 2602 }, { "epoch": 1.67, "learning_rate": 8.626826207218589e-06, "loss": 0.0, "step": 2603 }, { "epoch": 1.67, "learning_rate": 8.619952598369587e-06, "loss": 0.0001, "step": 2604 }, { "epoch": 1.67, "learning_rate": 8.61307965413957e-06, "loss": 0.0, "step": 2605 }, { "epoch": 1.67, "learning_rate": 8.606207377838489e-06, "loss": 0.0001, "step": 2606 }, { "epoch": 1.68, "learning_rate": 8.599335772775976e-06, "loss": 0.0, "step": 2607 }, { "epoch": 1.68, "learning_rate": 8.59246484226133e-06, "loss": 0.0, "step": 2608 }, { "epoch": 1.68, "learning_rate": 8.585594589603536e-06, "loss": 0.0001, "step": 2609 }, { "epoch": 1.68, "learning_rate": 8.578725018111253e-06, "loss": 0.0001, "step": 2610 }, { "epoch": 1.68, "learning_rate": 8.571856131092799e-06, "loss": 0.0, "step": 2611 }, { "epoch": 1.68, "learning_rate": 8.564987931856176e-06, "loss": 0.0001, "step": 2612 }, { "epoch": 1.68, "learning_rate": 8.558120423709055e-06, "loss": 0.0002, "step": 2613 }, { "epoch": 1.68, "learning_rate": 8.551253609958758e-06, "loss": 0.0, "step": 2614 }, { "epoch": 1.68, "learning_rate": 8.544387493912291e-06, "loss": 0.0, "step": 2615 }, { "epoch": 1.68, "learning_rate": 8.537522078876314e-06, "loss": 0.0001, "step": 2616 }, { "epoch": 1.68, "learning_rate": 8.530657368157152e-06, "loss": 0.0, "step": 2617 }, { "epoch": 1.68, "learning_rate": 8.523793365060796e-06, "loss": 0.0, "step": 2618 }, { "epoch": 1.68, "learning_rate": 8.51693007289288e-06, "loss": 0.0001, "step": 2619 }, { "epoch": 1.68, "learning_rate": 8.510067494958713e-06, "loss": 0.0, "step": 2620 }, { "epoch": 1.68, "learning_rate": 8.50320563456326e-06, "loss": 0.0, "step": 2621 }, { "epoch": 1.68, "learning_rate": 8.49634449501112e-06, "loss": 0.0, "step": 2622 }, { "epoch": 1.69, "learning_rate": 8.489484079606569e-06, "loss": 0.0001, "step": 2623 }, { "epoch": 1.69, "learning_rate": 8.482624391653525e-06, "loss": 0.0, "step": 2624 }, { "epoch": 1.69, "learning_rate": 8.475765434455549e-06, "loss": 0.0, "step": 2625 }, { "epoch": 1.69, "learning_rate": 8.468907211315857e-06, "loss": 0.0, "step": 2626 }, { "epoch": 1.69, "learning_rate": 8.462049725537317e-06, "loss": 0.0002, "step": 2627 }, { "epoch": 1.69, "learning_rate": 8.455192980422425e-06, "loss": 0.0, "step": 2628 }, { "epoch": 1.69, "learning_rate": 8.448336979273339e-06, "loss": 0.0, "step": 2629 }, { "epoch": 1.69, "learning_rate": 8.441481725391843e-06, "loss": 0.0, "step": 2630 }, { "epoch": 1.69, "learning_rate": 8.434627222079375e-06, "loss": 0.0, "step": 2631 }, { "epoch": 1.69, "learning_rate": 8.427773472637006e-06, "loss": 0.0001, "step": 2632 }, { "epoch": 1.69, "learning_rate": 8.420920480365433e-06, "loss": 0.0, "step": 2633 }, { "epoch": 1.69, "learning_rate": 8.414068248565007e-06, "loss": 0.0001, "step": 2634 }, { "epoch": 1.69, "learning_rate": 8.407216780535703e-06, "loss": 0.0001, "step": 2635 }, { "epoch": 1.69, "learning_rate": 8.400366079577126e-06, "loss": 0.0, "step": 2636 }, { "epoch": 1.69, "learning_rate": 8.393516148988517e-06, "loss": 0.0, "step": 2637 }, { "epoch": 1.7, "learning_rate": 8.386666992068748e-06, "loss": 0.0, "step": 2638 }, { "epoch": 1.7, "learning_rate": 8.379818612116304e-06, "loss": 0.0001, "step": 2639 }, { "epoch": 1.7, "learning_rate": 8.372971012429312e-06, "loss": 0.0, "step": 2640 }, { "epoch": 1.7, "learning_rate": 8.36612419630552e-06, "loss": 0.0, "step": 2641 }, { "epoch": 1.7, "learning_rate": 8.359278167042291e-06, "loss": 0.0, "step": 2642 }, { "epoch": 1.7, "learning_rate": 8.352432927936619e-06, "loss": 0.0001, "step": 2643 }, { "epoch": 1.7, "learning_rate": 8.345588482285106e-06, "loss": 0.0, "step": 2644 }, { "epoch": 1.7, "learning_rate": 8.33874483338398e-06, "loss": 0.0, "step": 2645 }, { "epoch": 1.7, "learning_rate": 8.331901984529087e-06, "loss": 0.0, "step": 2646 }, { "epoch": 1.7, "learning_rate": 8.32505993901588e-06, "loss": 0.0, "step": 2647 }, { "epoch": 1.7, "learning_rate": 8.318218700139433e-06, "loss": 0.0, "step": 2648 }, { "epoch": 1.7, "learning_rate": 8.311378271194432e-06, "loss": 0.0001, "step": 2649 }, { "epoch": 1.7, "learning_rate": 8.304538655475156e-06, "loss": 0.0, "step": 2650 }, { "epoch": 1.7, "learning_rate": 8.297699856275516e-06, "loss": 0.0, "step": 2651 }, { "epoch": 1.7, "learning_rate": 8.290861876889016e-06, "loss": 0.0, "step": 2652 }, { "epoch": 1.7, "learning_rate": 8.284024720608765e-06, "loss": 0.0, "step": 2653 }, { "epoch": 1.71, "learning_rate": 8.277188390727488e-06, "loss": 0.0001, "step": 2654 }, { "epoch": 1.71, "learning_rate": 8.270352890537491e-06, "loss": 0.0001, "step": 2655 }, { "epoch": 1.71, "learning_rate": 8.263518223330698e-06, "loss": 0.0, "step": 2656 }, { "epoch": 1.71, "learning_rate": 8.256684392398627e-06, "loss": 0.0, "step": 2657 }, { "epoch": 1.71, "learning_rate": 8.24985140103239e-06, "loss": 0.0, "step": 2658 }, { "epoch": 1.71, "learning_rate": 8.243019252522698e-06, "loss": 0.0001, "step": 2659 }, { "epoch": 1.71, "learning_rate": 8.23618795015986e-06, "loss": 0.0001, "step": 2660 }, { "epoch": 1.71, "learning_rate": 8.229357497233763e-06, "loss": 0.0, "step": 2661 }, { "epoch": 1.71, "learning_rate": 8.2225278970339e-06, "loss": 0.0, "step": 2662 }, { "epoch": 1.71, "learning_rate": 8.215699152849347e-06, "loss": 0.0, "step": 2663 }, { "epoch": 1.71, "learning_rate": 8.208871267968769e-06, "loss": 0.0, "step": 2664 }, { "epoch": 1.71, "learning_rate": 8.202044245680416e-06, "loss": 0.0001, "step": 2665 }, { "epoch": 1.71, "learning_rate": 8.195218089272128e-06, "loss": 0.0001, "step": 2666 }, { "epoch": 1.71, "learning_rate": 8.188392802031314e-06, "loss": 0.0, "step": 2667 }, { "epoch": 1.71, "learning_rate": 8.181568387244982e-06, "loss": 0.0001, "step": 2668 }, { "epoch": 1.72, "learning_rate": 8.174744848199707e-06, "loss": 0.0001, "step": 2669 }, { "epoch": 1.72, "learning_rate": 8.167922188181648e-06, "loss": 0.0, "step": 2670 }, { "epoch": 1.72, "learning_rate": 8.161100410476544e-06, "loss": 0.0, "step": 2671 }, { "epoch": 1.72, "learning_rate": 8.154279518369695e-06, "loss": 0.0, "step": 2672 }, { "epoch": 1.72, "learning_rate": 8.147459515145991e-06, "loss": 0.0, "step": 2673 }, { "epoch": 1.72, "learning_rate": 8.140640404089887e-06, "loss": 0.0, "step": 2674 }, { "epoch": 1.72, "learning_rate": 8.133822188485407e-06, "loss": 0.0, "step": 2675 }, { "epoch": 1.72, "learning_rate": 8.127004871616142e-06, "loss": 0.0, "step": 2676 }, { "epoch": 1.72, "learning_rate": 8.120188456765262e-06, "loss": 0.0, "step": 2677 }, { "epoch": 1.72, "learning_rate": 8.113372947215484e-06, "loss": 0.0001, "step": 2678 }, { "epoch": 1.72, "learning_rate": 8.106558346249103e-06, "loss": 0.0001, "step": 2679 }, { "epoch": 1.72, "learning_rate": 8.099744657147978e-06, "loss": 0.0001, "step": 2680 }, { "epoch": 1.72, "learning_rate": 8.09293188319351e-06, "loss": 0.0, "step": 2681 }, { "epoch": 1.72, "learning_rate": 8.086120027666688e-06, "loss": 0.0, "step": 2682 }, { "epoch": 1.72, "learning_rate": 8.079309093848029e-06, "loss": 0.0, "step": 2683 }, { "epoch": 1.72, "learning_rate": 8.072499085017629e-06, "loss": 0.0002, "step": 2684 }, { "epoch": 1.73, "learning_rate": 8.06569000445513e-06, "loss": 0.0, "step": 2685 }, { "epoch": 1.73, "learning_rate": 8.05888185543972e-06, "loss": 0.0001, "step": 2686 }, { "epoch": 1.73, "learning_rate": 8.05207464125015e-06, "loss": 0.0002, "step": 2687 }, { "epoch": 1.73, "learning_rate": 8.045268365164718e-06, "loss": 0.0, "step": 2688 }, { "epoch": 1.73, "learning_rate": 8.038463030461261e-06, "loss": 0.0001, "step": 2689 }, { "epoch": 1.73, "learning_rate": 8.031658640417178e-06, "loss": 0.0, "step": 2690 }, { "epoch": 1.73, "learning_rate": 8.024855198309402e-06, "loss": 0.0001, "step": 2691 }, { "epoch": 1.73, "learning_rate": 8.018052707414408e-06, "loss": 0.0, "step": 2692 }, { "epoch": 1.73, "learning_rate": 8.01125117100822e-06, "loss": 0.0001, "step": 2693 }, { "epoch": 1.73, "learning_rate": 8.004450592366404e-06, "loss": 0.0001, "step": 2694 }, { "epoch": 1.73, "learning_rate": 7.997650974764053e-06, "loss": 0.0, "step": 2695 }, { "epoch": 1.73, "learning_rate": 7.990852321475812e-06, "loss": 0.0001, "step": 2696 }, { "epoch": 1.73, "learning_rate": 7.984054635775847e-06, "loss": 0.0001, "step": 2697 }, { "epoch": 1.73, "learning_rate": 7.977257920937868e-06, "loss": 0.0001, "step": 2698 }, { "epoch": 1.73, "learning_rate": 7.970462180235117e-06, "loss": 0.0, "step": 2699 }, { "epoch": 1.73, "learning_rate": 7.96366741694036e-06, "loss": 0.0001, "step": 2700 }, { "epoch": 1.74, "learning_rate": 7.956873634325897e-06, "loss": 0.0, "step": 2701 }, { "epoch": 1.74, "learning_rate": 7.950080835663563e-06, "loss": 0.0001, "step": 2702 }, { "epoch": 1.74, "learning_rate": 7.9432890242247e-06, "loss": 0.0001, "step": 2703 }, { "epoch": 1.74, "learning_rate": 7.936498203280195e-06, "loss": 0.0001, "step": 2704 }, { "epoch": 1.74, "learning_rate": 7.929708376100445e-06, "loss": 0.0001, "step": 2705 }, { "epoch": 1.74, "learning_rate": 7.922919545955373e-06, "loss": 0.0, "step": 2706 }, { "epoch": 1.74, "learning_rate": 7.916131716114424e-06, "loss": 0.0, "step": 2707 }, { "epoch": 1.74, "learning_rate": 7.909344889846554e-06, "loss": 0.0, "step": 2708 }, { "epoch": 1.74, "learning_rate": 7.902559070420241e-06, "loss": 0.0001, "step": 2709 }, { "epoch": 1.74, "learning_rate": 7.895774261103483e-06, "loss": 0.0, "step": 2710 }, { "epoch": 1.74, "learning_rate": 7.888990465163779e-06, "loss": 0.0, "step": 2711 }, { "epoch": 1.74, "learning_rate": 7.882207685868149e-06, "loss": 0.0, "step": 2712 }, { "epoch": 1.74, "learning_rate": 7.875425926483128e-06, "loss": 0.0, "step": 2713 }, { "epoch": 1.74, "learning_rate": 7.868645190274741e-06, "loss": 0.0001, "step": 2714 }, { "epoch": 1.74, "learning_rate": 7.861865480508541e-06, "loss": 0.0, "step": 2715 }, { "epoch": 1.75, "learning_rate": 7.855086800449576e-06, "loss": 0.0, "step": 2716 }, { "epoch": 1.75, "learning_rate": 7.848309153362398e-06, "loss": 0.0, "step": 2717 }, { "epoch": 1.75, "learning_rate": 7.841532542511066e-06, "loss": 0.0001, "step": 2718 }, { "epoch": 1.75, "learning_rate": 7.83475697115914e-06, "loss": 0.0, "step": 2719 }, { "epoch": 1.75, "learning_rate": 7.827982442569673e-06, "loss": 0.0, "step": 2720 }, { "epoch": 1.75, "learning_rate": 7.82120896000522e-06, "loss": 0.0002, "step": 2721 }, { "epoch": 1.75, "learning_rate": 7.814436526727832e-06, "loss": 0.0001, "step": 2722 }, { "epoch": 1.75, "learning_rate": 7.807665145999057e-06, "loss": 0.0, "step": 2723 }, { "epoch": 1.75, "learning_rate": 7.800894821079934e-06, "loss": 0.0, "step": 2724 }, { "epoch": 1.75, "learning_rate": 7.79412555523099e-06, "loss": 0.0, "step": 2725 }, { "epoch": 1.75, "learning_rate": 7.787357351712247e-06, "loss": 0.0, "step": 2726 }, { "epoch": 1.75, "learning_rate": 7.780590213783215e-06, "loss": 0.0001, "step": 2727 }, { "epoch": 1.75, "learning_rate": 7.773824144702888e-06, "loss": 0.0, "step": 2728 }, { "epoch": 1.75, "learning_rate": 7.767059147729748e-06, "loss": 0.0, "step": 2729 }, { "epoch": 1.75, "learning_rate": 7.760295226121761e-06, "loss": 0.0, "step": 2730 }, { "epoch": 1.75, "learning_rate": 7.753532383136369e-06, "loss": 0.0001, "step": 2731 }, { "epoch": 1.76, "learning_rate": 7.7467706220305e-06, "loss": 0.0, "step": 2732 }, { "epoch": 1.76, "learning_rate": 7.740009946060565e-06, "loss": 0.0, "step": 2733 }, { "epoch": 1.76, "learning_rate": 7.733250358482442e-06, "loss": 0.0001, "step": 2734 }, { "epoch": 1.76, "learning_rate": 7.726491862551495e-06, "loss": 0.0001, "step": 2735 }, { "epoch": 1.76, "learning_rate": 7.719734461522554e-06, "loss": 0.0, "step": 2736 }, { "epoch": 1.76, "learning_rate": 7.712978158649925e-06, "loss": 0.0, "step": 2737 }, { "epoch": 1.76, "learning_rate": 7.706222957187393e-06, "loss": 0.0, "step": 2738 }, { "epoch": 1.76, "learning_rate": 7.699468860388192e-06, "loss": 0.0001, "step": 2739 }, { "epoch": 1.76, "learning_rate": 7.692715871505048e-06, "loss": 0.0001, "step": 2740 }, { "epoch": 1.76, "learning_rate": 7.685963993790138e-06, "loss": 0.0001, "step": 2741 }, { "epoch": 1.76, "learning_rate": 7.679213230495107e-06, "loss": 0.0, "step": 2742 }, { "epoch": 1.76, "learning_rate": 7.672463584871068e-06, "loss": 0.0, "step": 2743 }, { "epoch": 1.76, "learning_rate": 7.665715060168595e-06, "loss": 0.0, "step": 2744 }, { "epoch": 1.76, "learning_rate": 7.658967659637711e-06, "loss": 0.0, "step": 2745 }, { "epoch": 1.76, "learning_rate": 7.652221386527912e-06, "loss": 0.0, "step": 2746 }, { "epoch": 1.77, "learning_rate": 7.645476244088144e-06, "loss": 0.0001, "step": 2747 }, { "epoch": 1.77, "learning_rate": 7.63873223556681e-06, "loss": 0.0, "step": 2748 }, { "epoch": 1.77, "learning_rate": 7.631989364211769e-06, "loss": 0.0, "step": 2749 }, { "epoch": 1.77, "learning_rate": 7.625247633270324e-06, "loss": 0.0002, "step": 2750 }, { "epoch": 1.77, "learning_rate": 7.618507045989239e-06, "loss": 0.0, "step": 2751 }, { "epoch": 1.77, "learning_rate": 7.611767605614723e-06, "loss": 0.0001, "step": 2752 }, { "epoch": 1.77, "learning_rate": 7.605029315392429e-06, "loss": 0.0001, "step": 2753 }, { "epoch": 1.77, "learning_rate": 7.598292178567465e-06, "loss": 0.0002, "step": 2754 }, { "epoch": 1.77, "learning_rate": 7.591556198384378e-06, "loss": 0.0001, "step": 2755 }, { "epoch": 1.77, "learning_rate": 7.584821378087153e-06, "loss": 0.0, "step": 2756 }, { "epoch": 1.77, "learning_rate": 7.5780877209192246e-06, "loss": 0.0001, "step": 2757 }, { "epoch": 1.77, "learning_rate": 7.571355230123466e-06, "loss": 0.0, "step": 2758 }, { "epoch": 1.77, "learning_rate": 7.564623908942184e-06, "loss": 0.0, "step": 2759 }, { "epoch": 1.77, "learning_rate": 7.55789376061713e-06, "loss": 0.0001, "step": 2760 }, { "epoch": 1.77, "learning_rate": 7.551164788389479e-06, "loss": 0.0, "step": 2761 }, { "epoch": 1.77, "learning_rate": 7.544436995499848e-06, "loss": 0.0001, "step": 2762 }, { "epoch": 1.78, "learning_rate": 7.53771038518829e-06, "loss": 0.0001, "step": 2763 }, { "epoch": 1.78, "learning_rate": 7.530984960694276e-06, "loss": 0.0, "step": 2764 }, { "epoch": 1.78, "learning_rate": 7.524260725256714e-06, "loss": 0.0, "step": 2765 }, { "epoch": 1.78, "learning_rate": 7.517537682113944e-06, "loss": 0.0001, "step": 2766 }, { "epoch": 1.78, "learning_rate": 7.510815834503718e-06, "loss": 0.0, "step": 2767 }, { "epoch": 1.78, "learning_rate": 7.504095185663223e-06, "loss": 0.0003, "step": 2768 }, { "epoch": 1.78, "learning_rate": 7.497375738829069e-06, "loss": 0.0, "step": 2769 }, { "epoch": 1.78, "learning_rate": 7.490657497237277e-06, "loss": 0.0001, "step": 2770 }, { "epoch": 1.78, "learning_rate": 7.483940464123301e-06, "loss": 0.0, "step": 2771 }, { "epoch": 1.78, "learning_rate": 7.477224642722007e-06, "loss": 0.0, "step": 2772 }, { "epoch": 1.78, "learning_rate": 7.470510036267671e-06, "loss": 0.0001, "step": 2773 }, { "epoch": 1.78, "learning_rate": 7.463796647993993e-06, "loss": 0.0, "step": 2774 }, { "epoch": 1.78, "learning_rate": 7.457084481134083e-06, "loss": 0.0, "step": 2775 }, { "epoch": 1.78, "learning_rate": 7.450373538920462e-06, "loss": 0.0001, "step": 2776 }, { "epoch": 1.78, "learning_rate": 7.443663824585068e-06, "loss": 0.0001, "step": 2777 }, { "epoch": 1.79, "learning_rate": 7.436955341359232e-06, "loss": 0.0001, "step": 2778 }, { "epoch": 1.79, "learning_rate": 7.430248092473707e-06, "loss": 0.0001, "step": 2779 }, { "epoch": 1.79, "learning_rate": 7.42354208115865e-06, "loss": 0.0, "step": 2780 }, { "epoch": 1.79, "learning_rate": 7.416837310643612e-06, "loss": 0.0001, "step": 2781 }, { "epoch": 1.79, "learning_rate": 7.410133784157556e-06, "loss": 0.0001, "step": 2782 }, { "epoch": 1.79, "learning_rate": 7.403431504928847e-06, "loss": 0.0001, "step": 2783 }, { "epoch": 1.79, "learning_rate": 7.396730476185237e-06, "loss": 0.0002, "step": 2784 }, { "epoch": 1.79, "learning_rate": 7.390030701153887e-06, "loss": 0.0, "step": 2785 }, { "epoch": 1.79, "learning_rate": 7.383332183061356e-06, "loss": 0.0, "step": 2786 }, { "epoch": 1.79, "learning_rate": 7.376634925133586e-06, "loss": 0.0, "step": 2787 }, { "epoch": 1.79, "learning_rate": 7.369938930595925e-06, "loss": 0.0, "step": 2788 }, { "epoch": 1.79, "learning_rate": 7.363244202673099e-06, "loss": 0.0001, "step": 2789 }, { "epoch": 1.79, "learning_rate": 7.356550744589237e-06, "loss": 0.0, "step": 2790 }, { "epoch": 1.79, "learning_rate": 7.349858559567853e-06, "loss": 0.0001, "step": 2791 }, { "epoch": 1.79, "learning_rate": 7.343167650831839e-06, "loss": 0.0001, "step": 2792 }, { "epoch": 1.79, "learning_rate": 7.336478021603481e-06, "loss": 0.0, "step": 2793 }, { "epoch": 1.8, "learning_rate": 7.329789675104455e-06, "loss": 0.0, "step": 2794 }, { "epoch": 1.8, "learning_rate": 7.323102614555801e-06, "loss": 0.0, "step": 2795 }, { "epoch": 1.8, "learning_rate": 7.316416843177953e-06, "loss": 0.0001, "step": 2796 }, { "epoch": 1.8, "learning_rate": 7.309732364190728e-06, "loss": 0.0001, "step": 2797 }, { "epoch": 1.8, "learning_rate": 7.303049180813304e-06, "loss": 0.0, "step": 2798 }, { "epoch": 1.8, "learning_rate": 7.29636729626425e-06, "loss": 0.0001, "step": 2799 }, { "epoch": 1.8, "learning_rate": 7.2896867137615015e-06, "loss": 0.0, "step": 2800 }, { "epoch": 1.8, "learning_rate": 7.283007436522369e-06, "loss": 0.0001, "step": 2801 }, { "epoch": 1.8, "learning_rate": 7.276329467763543e-06, "loss": 0.0, "step": 2802 }, { "epoch": 1.8, "learning_rate": 7.269652810701067e-06, "loss": 0.0, "step": 2803 }, { "epoch": 1.8, "learning_rate": 7.262977468550364e-06, "loss": 0.0, "step": 2804 }, { "epoch": 1.8, "learning_rate": 7.256303444526225e-06, "loss": 0.0001, "step": 2805 }, { "epoch": 1.8, "learning_rate": 7.2496307418427965e-06, "loss": 0.0001, "step": 2806 }, { "epoch": 1.8, "learning_rate": 7.242959363713601e-06, "loss": 0.0001, "step": 2807 }, { "epoch": 1.8, "learning_rate": 7.236289313351516e-06, "loss": 0.0001, "step": 2808 }, { "epoch": 1.8, "learning_rate": 7.229620593968778e-06, "loss": 0.0, "step": 2809 }, { "epoch": 1.81, "learning_rate": 7.222953208776984e-06, "loss": 0.0, "step": 2810 }, { "epoch": 1.81, "learning_rate": 7.216287160987093e-06, "loss": 0.0001, "step": 2811 }, { "epoch": 1.81, "learning_rate": 7.209622453809413e-06, "loss": 0.0, "step": 2812 }, { "epoch": 1.81, "learning_rate": 7.202959090453617e-06, "loss": 0.0001, "step": 2813 }, { "epoch": 1.81, "learning_rate": 7.196297074128713e-06, "loss": 0.0, "step": 2814 }, { "epoch": 1.81, "learning_rate": 7.189636408043077e-06, "loss": 0.0, "step": 2815 }, { "epoch": 1.81, "learning_rate": 7.182977095404429e-06, "loss": 0.0, "step": 2816 }, { "epoch": 1.81, "learning_rate": 7.176319139419834e-06, "loss": 0.0, "step": 2817 }, { "epoch": 1.81, "learning_rate": 7.169662543295708e-06, "loss": 0.0, "step": 2818 }, { "epoch": 1.81, "learning_rate": 7.163007310237815e-06, "loss": 0.0001, "step": 2819 }, { "epoch": 1.81, "learning_rate": 7.15635344345125e-06, "loss": 0.0, "step": 2820 }, { "epoch": 1.81, "learning_rate": 7.14970094614046e-06, "loss": 0.0001, "step": 2821 }, { "epoch": 1.81, "learning_rate": 7.143049821509236e-06, "loss": 0.0001, "step": 2822 }, { "epoch": 1.81, "learning_rate": 7.136400072760696e-06, "loss": 0.0, "step": 2823 }, { "epoch": 1.81, "learning_rate": 7.129751703097305e-06, "loss": 0.0001, "step": 2824 }, { "epoch": 1.82, "learning_rate": 7.123104715720866e-06, "loss": 0.0001, "step": 2825 }, { "epoch": 1.82, "learning_rate": 7.116459113832501e-06, "loss": 0.0, "step": 2826 }, { "epoch": 1.82, "learning_rate": 7.109814900632682e-06, "loss": 0.0, "step": 2827 }, { "epoch": 1.82, "learning_rate": 7.103172079321201e-06, "loss": 0.0, "step": 2828 }, { "epoch": 1.82, "learning_rate": 7.096530653097184e-06, "loss": 0.0001, "step": 2829 }, { "epoch": 1.82, "learning_rate": 7.089890625159091e-06, "loss": 0.0, "step": 2830 }, { "epoch": 1.82, "learning_rate": 7.083251998704694e-06, "loss": 0.0001, "step": 2831 }, { "epoch": 1.82, "learning_rate": 7.076614776931103e-06, "loss": 0.0001, "step": 2832 }, { "epoch": 1.82, "learning_rate": 7.069978963034747e-06, "loss": 0.0, "step": 2833 }, { "epoch": 1.82, "learning_rate": 7.063344560211376e-06, "loss": 0.0, "step": 2834 }, { "epoch": 1.82, "learning_rate": 7.0567115716560595e-06, "loss": 0.0002, "step": 2835 }, { "epoch": 1.82, "learning_rate": 7.0500800005631955e-06, "loss": 0.0001, "step": 2836 }, { "epoch": 1.82, "learning_rate": 7.043449850126483e-06, "loss": 0.0, "step": 2837 }, { "epoch": 1.82, "learning_rate": 7.036821123538951e-06, "loss": 0.0001, "step": 2838 }, { "epoch": 1.82, "learning_rate": 7.030193823992932e-06, "loss": 0.0001, "step": 2839 }, { "epoch": 1.82, "learning_rate": 7.02356795468008e-06, "loss": 0.0001, "step": 2840 }, { "epoch": 1.83, "learning_rate": 7.016943518791361e-06, "loss": 0.0, "step": 2841 }, { "epoch": 1.83, "learning_rate": 7.010320519517039e-06, "loss": 0.0, "step": 2842 }, { "epoch": 1.83, "learning_rate": 7.003698960046696e-06, "loss": 0.0001, "step": 2843 }, { "epoch": 1.83, "learning_rate": 6.997078843569223e-06, "loss": 0.0, "step": 2844 }, { "epoch": 1.83, "learning_rate": 6.990460173272802e-06, "loss": 0.0001, "step": 2845 }, { "epoch": 1.83, "learning_rate": 6.983842952344933e-06, "loss": 0.0, "step": 2846 }, { "epoch": 1.83, "learning_rate": 6.977227183972419e-06, "loss": 0.0, "step": 2847 }, { "epoch": 1.83, "learning_rate": 6.970612871341347e-06, "loss": 0.0, "step": 2848 }, { "epoch": 1.83, "learning_rate": 6.964000017637116e-06, "loss": 0.0001, "step": 2849 }, { "epoch": 1.83, "learning_rate": 6.957388626044426e-06, "loss": 0.0001, "step": 2850 }, { "epoch": 1.83, "learning_rate": 6.950778699747256e-06, "loss": 0.0, "step": 2851 }, { "epoch": 1.83, "learning_rate": 6.944170241928898e-06, "loss": 0.0002, "step": 2852 }, { "epoch": 1.83, "learning_rate": 6.9375632557719245e-06, "loss": 0.0001, "step": 2853 }, { "epoch": 1.83, "learning_rate": 6.930957744458204e-06, "loss": 0.0001, "step": 2854 }, { "epoch": 1.83, "learning_rate": 6.924353711168899e-06, "loss": 0.0, "step": 2855 }, { "epoch": 1.84, "learning_rate": 6.917751159084446e-06, "loss": 0.0, "step": 2856 }, { "epoch": 1.84, "learning_rate": 6.911150091384582e-06, "loss": 0.0, "step": 2857 }, { "epoch": 1.84, "learning_rate": 6.904550511248328e-06, "loss": 0.0, "step": 2858 }, { "epoch": 1.84, "learning_rate": 6.897952421853979e-06, "loss": 0.0, "step": 2859 }, { "epoch": 1.84, "learning_rate": 6.891355826379123e-06, "loss": 0.0, "step": 2860 }, { "epoch": 1.84, "learning_rate": 6.884760728000624e-06, "loss": 0.0, "step": 2861 }, { "epoch": 1.84, "learning_rate": 6.878167129894619e-06, "loss": 0.0, "step": 2862 }, { "epoch": 1.84, "learning_rate": 6.871575035236533e-06, "loss": 0.0001, "step": 2863 }, { "epoch": 1.84, "learning_rate": 6.864984447201063e-06, "loss": 0.0001, "step": 2864 }, { "epoch": 1.84, "learning_rate": 6.8583953689621765e-06, "loss": 0.0001, "step": 2865 }, { "epoch": 1.84, "learning_rate": 6.85180780369312e-06, "loss": 0.0, "step": 2866 }, { "epoch": 1.84, "learning_rate": 6.845221754566404e-06, "loss": 0.0, "step": 2867 }, { "epoch": 1.84, "learning_rate": 6.8386372247538165e-06, "loss": 0.0, "step": 2868 }, { "epoch": 1.84, "learning_rate": 6.832054217426414e-06, "loss": 0.0, "step": 2869 }, { "epoch": 1.84, "learning_rate": 6.825472735754508e-06, "loss": 0.0, "step": 2870 }, { "epoch": 1.84, "learning_rate": 6.818892782907688e-06, "loss": 0.0001, "step": 2871 }, { "epoch": 1.85, "learning_rate": 6.812314362054808e-06, "loss": 0.0, "step": 2872 }, { "epoch": 1.85, "learning_rate": 6.805737476363969e-06, "loss": 0.0001, "step": 2873 }, { "epoch": 1.85, "learning_rate": 6.799162129002548e-06, "loss": 0.0, "step": 2874 }, { "epoch": 1.85, "learning_rate": 6.792588323137177e-06, "loss": 0.0, "step": 2875 }, { "epoch": 1.85, "learning_rate": 6.78601606193374e-06, "loss": 0.0, "step": 2876 }, { "epoch": 1.85, "learning_rate": 6.77944534855739e-06, "loss": 0.0001, "step": 2877 }, { "epoch": 1.85, "learning_rate": 6.772876186172517e-06, "loss": 0.0, "step": 2878 }, { "epoch": 1.85, "learning_rate": 6.766308577942775e-06, "loss": 0.0, "step": 2879 }, { "epoch": 1.85, "learning_rate": 6.759742527031074e-06, "loss": 0.0001, "step": 2880 }, { "epoch": 1.85, "learning_rate": 6.753178036599559e-06, "loss": 0.0001, "step": 2881 }, { "epoch": 1.85, "learning_rate": 6.746615109809638e-06, "loss": 0.0, "step": 2882 }, { "epoch": 1.85, "learning_rate": 6.740053749821963e-06, "loss": 0.0, "step": 2883 }, { "epoch": 1.85, "learning_rate": 6.73349395979642e-06, "loss": 0.0, "step": 2884 }, { "epoch": 1.85, "learning_rate": 6.726935742892154e-06, "loss": 0.0001, "step": 2885 }, { "epoch": 1.85, "learning_rate": 6.720379102267544e-06, "loss": 0.0001, "step": 2886 }, { "epoch": 1.86, "learning_rate": 6.71382404108021e-06, "loss": 0.0, "step": 2887 }, { "epoch": 1.86, "learning_rate": 6.707270562487016e-06, "loss": 0.0, "step": 2888 }, { "epoch": 1.86, "learning_rate": 6.700718669644064e-06, "loss": 0.0001, "step": 2889 }, { "epoch": 1.86, "learning_rate": 6.694168365706681e-06, "loss": 0.0002, "step": 2890 }, { "epoch": 1.86, "learning_rate": 6.6876196538294445e-06, "loss": 0.0, "step": 2891 }, { "epoch": 1.86, "learning_rate": 6.681072537166153e-06, "loss": 0.0001, "step": 2892 }, { "epoch": 1.86, "learning_rate": 6.674527018869844e-06, "loss": 0.0, "step": 2893 }, { "epoch": 1.86, "learning_rate": 6.667983102092788e-06, "loss": 0.0001, "step": 2894 }, { "epoch": 1.86, "learning_rate": 6.661440789986467e-06, "loss": 0.0001, "step": 2895 }, { "epoch": 1.86, "learning_rate": 6.6549000857016124e-06, "loss": 0.0001, "step": 2896 }, { "epoch": 1.86, "learning_rate": 6.648360992388168e-06, "loss": 0.0, "step": 2897 }, { "epoch": 1.86, "learning_rate": 6.641823513195304e-06, "loss": 0.0, "step": 2898 }, { "epoch": 1.86, "learning_rate": 6.635287651271416e-06, "loss": 0.0, "step": 2899 }, { "epoch": 1.86, "learning_rate": 6.628753409764122e-06, "loss": 0.0, "step": 2900 }, { "epoch": 1.86, "learning_rate": 6.6222207918202485e-06, "loss": 0.0001, "step": 2901 }, { "epoch": 1.86, "learning_rate": 6.615689800585854e-06, "loss": 0.0001, "step": 2902 }, { "epoch": 1.87, "learning_rate": 6.6091604392062105e-06, "loss": 0.0, "step": 2903 }, { "epoch": 1.87, "learning_rate": 6.602632710825794e-06, "loss": 0.0, "step": 2904 }, { "epoch": 1.87, "learning_rate": 6.59610661858831e-06, "loss": 0.0, "step": 2905 }, { "epoch": 1.87, "learning_rate": 6.589582165636661e-06, "loss": 0.0001, "step": 2906 }, { "epoch": 1.87, "learning_rate": 6.583059355112975e-06, "loss": 0.0, "step": 2907 }, { "epoch": 1.87, "learning_rate": 6.576538190158581e-06, "loss": 0.0, "step": 2908 }, { "epoch": 1.87, "learning_rate": 6.570018673914011e-06, "loss": 0.0, "step": 2909 }, { "epoch": 1.87, "learning_rate": 6.56350080951901e-06, "loss": 0.0, "step": 2910 }, { "epoch": 1.87, "learning_rate": 6.5569846001125284e-06, "loss": 0.0, "step": 2911 }, { "epoch": 1.87, "learning_rate": 6.550470048832713e-06, "loss": 0.0, "step": 2912 }, { "epoch": 1.87, "learning_rate": 6.5439571588169186e-06, "loss": 0.0, "step": 2913 }, { "epoch": 1.87, "learning_rate": 6.537445933201698e-06, "loss": 0.0001, "step": 2914 }, { "epoch": 1.87, "learning_rate": 6.530936375122796e-06, "loss": 0.0, "step": 2915 }, { "epoch": 1.87, "learning_rate": 6.524428487715165e-06, "loss": 0.0, "step": 2916 }, { "epoch": 1.87, "learning_rate": 6.517922274112947e-06, "loss": 0.0, "step": 2917 }, { "epoch": 1.88, "learning_rate": 6.511417737449477e-06, "loss": 0.0, "step": 2918 }, { "epoch": 1.88, "learning_rate": 6.504914880857287e-06, "loss": 0.0001, "step": 2919 }, { "epoch": 1.88, "learning_rate": 6.498413707468092e-06, "loss": 0.0, "step": 2920 }, { "epoch": 1.88, "learning_rate": 6.491914220412805e-06, "loss": 0.0, "step": 2921 }, { "epoch": 1.88, "learning_rate": 6.485416422821522e-06, "loss": 0.0001, "step": 2922 }, { "epoch": 1.88, "learning_rate": 6.478920317823527e-06, "loss": 0.0002, "step": 2923 }, { "epoch": 1.88, "learning_rate": 6.472425908547289e-06, "loss": 0.0, "step": 2924 }, { "epoch": 1.88, "learning_rate": 6.465933198120461e-06, "loss": 0.0, "step": 2925 }, { "epoch": 1.88, "learning_rate": 6.459442189669871e-06, "loss": 0.0, "step": 2926 }, { "epoch": 1.88, "learning_rate": 6.452952886321537e-06, "loss": 0.0001, "step": 2927 }, { "epoch": 1.88, "learning_rate": 6.446465291200654e-06, "loss": 0.0, "step": 2928 }, { "epoch": 1.88, "learning_rate": 6.439979407431587e-06, "loss": 0.0002, "step": 2929 }, { "epoch": 1.88, "learning_rate": 6.433495238137888e-06, "loss": 0.0001, "step": 2930 }, { "epoch": 1.88, "learning_rate": 6.427012786442271e-06, "loss": 0.0, "step": 2931 }, { "epoch": 1.88, "learning_rate": 6.420532055466632e-06, "loss": 0.0, "step": 2932 }, { "epoch": 1.88, "learning_rate": 6.414053048332036e-06, "loss": 0.0, "step": 2933 }, { "epoch": 1.89, "learning_rate": 6.407575768158716e-06, "loss": 0.0, "step": 2934 }, { "epoch": 1.89, "learning_rate": 6.401100218066074e-06, "loss": 0.0002, "step": 2935 }, { "epoch": 1.89, "learning_rate": 6.3946264011726856e-06, "loss": 0.0, "step": 2936 }, { "epoch": 1.89, "learning_rate": 6.388154320596276e-06, "loss": 0.0001, "step": 2937 }, { "epoch": 1.89, "learning_rate": 6.381683979453748e-06, "loss": 0.0, "step": 2938 }, { "epoch": 1.89, "learning_rate": 6.375215380861164e-06, "loss": 0.0, "step": 2939 }, { "epoch": 1.89, "learning_rate": 6.368748527933741e-06, "loss": 0.0, "step": 2940 }, { "epoch": 1.89, "learning_rate": 6.362283423785864e-06, "loss": 0.0, "step": 2941 }, { "epoch": 1.89, "learning_rate": 6.355820071531073e-06, "loss": 0.0001, "step": 2942 }, { "epoch": 1.89, "learning_rate": 6.349358474282059e-06, "loss": 0.0002, "step": 2943 }, { "epoch": 1.89, "learning_rate": 6.342898635150674e-06, "loss": 0.0, "step": 2944 }, { "epoch": 1.89, "learning_rate": 6.3364405572479186e-06, "loss": 0.0, "step": 2945 }, { "epoch": 1.89, "learning_rate": 6.329984243683949e-06, "loss": 0.0, "step": 2946 }, { "epoch": 1.89, "learning_rate": 6.323529697568075e-06, "loss": 0.0, "step": 2947 }, { "epoch": 1.89, "learning_rate": 6.317076922008742e-06, "loss": 0.0, "step": 2948 }, { "epoch": 1.89, "learning_rate": 6.310625920113556e-06, "loss": 0.0001, "step": 2949 }, { "epoch": 1.9, "learning_rate": 6.304176694989266e-06, "loss": 0.0, "step": 2950 }, { "epoch": 1.9, "learning_rate": 6.29772924974176e-06, "loss": 0.0, "step": 2951 }, { "epoch": 1.9, "learning_rate": 6.291283587476072e-06, "loss": 0.0, "step": 2952 }, { "epoch": 1.9, "learning_rate": 6.2848397112963846e-06, "loss": 0.0002, "step": 2953 }, { "epoch": 1.9, "learning_rate": 6.278397624306004e-06, "loss": 0.0002, "step": 2954 }, { "epoch": 1.9, "learning_rate": 6.271957329607388e-06, "loss": 0.0, "step": 2955 }, { "epoch": 1.9, "learning_rate": 6.265518830302131e-06, "loss": 0.0001, "step": 2956 }, { "epoch": 1.9, "learning_rate": 6.2590821294909486e-06, "loss": 0.0, "step": 2957 }, { "epoch": 1.9, "learning_rate": 6.252647230273714e-06, "loss": 0.0001, "step": 2958 }, { "epoch": 1.9, "learning_rate": 6.246214135749409e-06, "loss": 0.0, "step": 2959 }, { "epoch": 1.9, "learning_rate": 6.2397828490161604e-06, "loss": 0.0001, "step": 2960 }, { "epoch": 1.9, "learning_rate": 6.233353373171224e-06, "loss": 0.0001, "step": 2961 }, { "epoch": 1.9, "learning_rate": 6.226925711310974e-06, "loss": 0.0, "step": 2962 }, { "epoch": 1.9, "learning_rate": 6.220499866530918e-06, "loss": 0.0, "step": 2963 }, { "epoch": 1.9, "learning_rate": 6.214075841925694e-06, "loss": 0.0, "step": 2964 }, { "epoch": 1.91, "learning_rate": 6.207653640589051e-06, "loss": 0.0, "step": 2965 }, { "epoch": 1.91, "learning_rate": 6.201233265613867e-06, "loss": 0.0, "step": 2966 }, { "epoch": 1.91, "learning_rate": 6.194814720092144e-06, "loss": 0.0001, "step": 2967 }, { "epoch": 1.91, "learning_rate": 6.1883980071149915e-06, "loss": 0.0001, "step": 2968 }, { "epoch": 1.91, "learning_rate": 6.1819831297726494e-06, "loss": 0.0, "step": 2969 }, { "epoch": 1.91, "learning_rate": 6.175570091154462e-06, "loss": 0.0001, "step": 2970 }, { "epoch": 1.91, "learning_rate": 6.1691588943488974e-06, "loss": 0.0001, "step": 2971 }, { "epoch": 1.91, "learning_rate": 6.162749542443537e-06, "loss": 0.0, "step": 2972 }, { "epoch": 1.91, "learning_rate": 6.15634203852506e-06, "loss": 0.0, "step": 2973 }, { "epoch": 1.91, "learning_rate": 6.149936385679269e-06, "loss": 0.0, "step": 2974 }, { "epoch": 1.91, "learning_rate": 6.143532586991077e-06, "loss": 0.0001, "step": 2975 }, { "epoch": 1.91, "learning_rate": 6.13713064554449e-06, "loss": 0.0, "step": 2976 }, { "epoch": 1.91, "learning_rate": 6.130730564422633e-06, "loss": 0.0001, "step": 2977 }, { "epoch": 1.91, "learning_rate": 6.1243323467077335e-06, "loss": 0.0, "step": 2978 }, { "epoch": 1.91, "learning_rate": 6.11793599548111e-06, "loss": 0.0, "step": 2979 }, { "epoch": 1.91, "learning_rate": 6.111541513823194e-06, "loss": 0.0, "step": 2980 }, { "epoch": 1.92, "learning_rate": 6.105148904813516e-06, "loss": 0.0, "step": 2981 }, { "epoch": 1.92, "learning_rate": 6.0987581715306966e-06, "loss": 0.0, "step": 2982 }, { "epoch": 1.92, "learning_rate": 6.092369317052464e-06, "loss": 0.0, "step": 2983 }, { "epoch": 1.92, "learning_rate": 6.08598234445563e-06, "loss": 0.0, "step": 2984 }, { "epoch": 1.92, "learning_rate": 6.0795972568161065e-06, "loss": 0.0, "step": 2985 }, { "epoch": 1.92, "learning_rate": 6.0732140572089006e-06, "loss": 0.0002, "step": 2986 }, { "epoch": 1.92, "learning_rate": 6.0668327487081025e-06, "loss": 0.0, "step": 2987 }, { "epoch": 1.92, "learning_rate": 6.060453334386898e-06, "loss": 0.0, "step": 2988 }, { "epoch": 1.92, "learning_rate": 6.0540758173175595e-06, "loss": 0.0, "step": 2989 }, { "epoch": 1.92, "learning_rate": 6.04770020057144e-06, "loss": 0.0001, "step": 2990 }, { "epoch": 1.92, "learning_rate": 6.041326487218986e-06, "loss": 0.0001, "step": 2991 }, { "epoch": 1.92, "learning_rate": 6.034954680329721e-06, "loss": 0.0, "step": 2992 }, { "epoch": 1.92, "learning_rate": 6.0285847829722535e-06, "loss": 0.0, "step": 2993 }, { "epoch": 1.92, "learning_rate": 6.022216798214269e-06, "loss": 0.0001, "step": 2994 }, { "epoch": 1.92, "learning_rate": 6.015850729122541e-06, "loss": 0.0, "step": 2995 }, { "epoch": 1.93, "learning_rate": 6.009486578762906e-06, "loss": 0.0001, "step": 2996 }, { "epoch": 1.93, "learning_rate": 6.0031243502002875e-06, "loss": 0.0, "step": 2997 }, { "epoch": 1.93, "learning_rate": 5.99676404649868e-06, "loss": 0.0, "step": 2998 }, { "epoch": 1.93, "learning_rate": 5.9904056707211485e-06, "loss": 0.0, "step": 2999 }, { "epoch": 1.93, "learning_rate": 5.984049225929841e-06, "loss": 0.0001, "step": 3000 }, { "epoch": 1.93, "learning_rate": 5.977694715185956e-06, "loss": 0.0001, "step": 3001 }, { "epoch": 1.93, "learning_rate": 5.971342141549776e-06, "loss": 0.0, "step": 3002 }, { "epoch": 1.93, "learning_rate": 5.9649915080806475e-06, "loss": 0.0, "step": 3003 }, { "epoch": 1.93, "learning_rate": 5.958642817836977e-06, "loss": 0.0001, "step": 3004 }, { "epoch": 1.93, "learning_rate": 5.952296073876242e-06, "loss": 0.0001, "step": 3005 }, { "epoch": 1.93, "learning_rate": 5.945951279254984e-06, "loss": 0.0001, "step": 3006 }, { "epoch": 1.93, "learning_rate": 5.939608437028794e-06, "loss": 0.0, "step": 3007 }, { "epoch": 1.93, "learning_rate": 5.933267550252333e-06, "loss": 0.0, "step": 3008 }, { "epoch": 1.93, "learning_rate": 5.9269286219793195e-06, "loss": 0.0, "step": 3009 }, { "epoch": 1.93, "learning_rate": 5.920591655262524e-06, "loss": 0.0, "step": 3010 }, { "epoch": 1.93, "learning_rate": 5.91425665315378e-06, "loss": 0.0, "step": 3011 }, { "epoch": 1.94, "learning_rate": 5.907923618703965e-06, "loss": 0.0, "step": 3012 }, { "epoch": 1.94, "learning_rate": 5.901592554963016e-06, "loss": 0.0, "step": 3013 }, { "epoch": 1.94, "learning_rate": 5.895263464979924e-06, "loss": 0.0001, "step": 3014 }, { "epoch": 1.94, "learning_rate": 5.888936351802712e-06, "loss": 0.0001, "step": 3015 }, { "epoch": 1.94, "learning_rate": 5.882611218478474e-06, "loss": 0.0, "step": 3016 }, { "epoch": 1.94, "learning_rate": 5.876288068053337e-06, "loss": 0.0001, "step": 3017 }, { "epoch": 1.94, "learning_rate": 5.869966903572476e-06, "loss": 0.0, "step": 3018 }, { "epoch": 1.94, "learning_rate": 5.8636477280801085e-06, "loss": 0.0001, "step": 3019 }, { "epoch": 1.94, "learning_rate": 5.857330544619499e-06, "loss": 0.0, "step": 3020 }, { "epoch": 1.94, "learning_rate": 5.851015356232944e-06, "loss": 0.0002, "step": 3021 }, { "epoch": 1.94, "learning_rate": 5.844702165961787e-06, "loss": 0.0, "step": 3022 }, { "epoch": 1.94, "learning_rate": 5.8383909768464055e-06, "loss": 0.0, "step": 3023 }, { "epoch": 1.94, "learning_rate": 5.8320817919262095e-06, "loss": 0.0, "step": 3024 }, { "epoch": 1.94, "learning_rate": 5.825774614239657e-06, "loss": 0.0, "step": 3025 }, { "epoch": 1.94, "learning_rate": 5.819469446824226e-06, "loss": 0.0, "step": 3026 }, { "epoch": 1.95, "learning_rate": 5.813166292716427e-06, "loss": 0.0, "step": 3027 }, { "epoch": 1.95, "learning_rate": 5.8068651549518175e-06, "loss": 0.0, "step": 3028 }, { "epoch": 1.95, "learning_rate": 5.800566036564954e-06, "loss": 0.0, "step": 3029 }, { "epoch": 1.95, "learning_rate": 5.794268940589451e-06, "loss": 0.0, "step": 3030 }, { "epoch": 1.95, "learning_rate": 5.787973870057933e-06, "loss": 0.0, "step": 3031 }, { "epoch": 1.95, "learning_rate": 5.781680828002047e-06, "loss": 0.0, "step": 3032 }, { "epoch": 1.95, "learning_rate": 5.7753898174524744e-06, "loss": 0.0, "step": 3033 }, { "epoch": 1.95, "learning_rate": 5.769100841438909e-06, "loss": 0.0, "step": 3034 }, { "epoch": 1.95, "learning_rate": 5.76281390299007e-06, "loss": 0.0, "step": 3035 }, { "epoch": 1.95, "learning_rate": 5.756529005133689e-06, "loss": 0.0001, "step": 3036 }, { "epoch": 1.95, "learning_rate": 5.750246150896517e-06, "loss": 0.0, "step": 3037 }, { "epoch": 1.95, "learning_rate": 5.743965343304331e-06, "loss": 0.0, "step": 3038 }, { "epoch": 1.95, "learning_rate": 5.7376865853819094e-06, "loss": 0.0, "step": 3039 }, { "epoch": 1.95, "learning_rate": 5.731409880153047e-06, "loss": 0.0, "step": 3040 }, { "epoch": 1.95, "learning_rate": 5.725135230640551e-06, "loss": 0.0, "step": 3041 }, { "epoch": 1.95, "learning_rate": 5.718862639866242e-06, "loss": 0.0, "step": 3042 }, { "epoch": 1.96, "learning_rate": 5.712592110850945e-06, "loss": 0.0, "step": 3043 }, { "epoch": 1.96, "learning_rate": 5.706323646614486e-06, "loss": 0.0, "step": 3044 }, { "epoch": 1.96, "learning_rate": 5.700057250175721e-06, "loss": 0.0001, "step": 3045 }, { "epoch": 1.96, "learning_rate": 5.69379292455247e-06, "loss": 0.0001, "step": 3046 }, { "epoch": 1.96, "learning_rate": 5.687530672761597e-06, "loss": 0.0, "step": 3047 }, { "epoch": 1.96, "learning_rate": 5.6812704978189396e-06, "loss": 0.0, "step": 3048 }, { "epoch": 1.96, "learning_rate": 5.675012402739342e-06, "loss": 0.0, "step": 3049 }, { "epoch": 1.96, "learning_rate": 5.6687563905366625e-06, "loss": 0.0001, "step": 3050 }, { "epoch": 1.96, "learning_rate": 5.662502464223725e-06, "loss": 0.0, "step": 3051 }, { "epoch": 1.96, "learning_rate": 5.656250626812378e-06, "loss": 0.0002, "step": 3052 }, { "epoch": 1.96, "learning_rate": 5.650000881313453e-06, "loss": 0.0001, "step": 3053 }, { "epoch": 1.96, "learning_rate": 5.643753230736766e-06, "loss": 0.0002, "step": 3054 }, { "epoch": 1.96, "learning_rate": 5.63750767809114e-06, "loss": 0.0, "step": 3055 }, { "epoch": 1.96, "learning_rate": 5.631264226384376e-06, "loss": 0.0, "step": 3056 }, { "epoch": 1.96, "learning_rate": 5.6250228786232686e-06, "loss": 0.0, "step": 3057 }, { "epoch": 1.96, "learning_rate": 5.618783637813592e-06, "loss": 0.0, "step": 3058 }, { "epoch": 1.97, "learning_rate": 5.612546506960117e-06, "loss": 0.0, "step": 3059 }, { "epoch": 1.97, "learning_rate": 5.606311489066592e-06, "loss": 0.0, "step": 3060 }, { "epoch": 1.97, "learning_rate": 5.600078587135747e-06, "loss": 0.0001, "step": 3061 }, { "epoch": 1.97, "learning_rate": 5.593847804169291e-06, "loss": 0.0, "step": 3062 }, { "epoch": 1.97, "learning_rate": 5.587619143167915e-06, "loss": 0.0001, "step": 3063 }, { "epoch": 1.97, "learning_rate": 5.5813926071313e-06, "loss": 0.0, "step": 3064 }, { "epoch": 1.97, "learning_rate": 5.575168199058077e-06, "loss": 0.0, "step": 3065 }, { "epoch": 1.97, "learning_rate": 5.568945921945872e-06, "loss": 0.0001, "step": 3066 }, { "epoch": 1.97, "learning_rate": 5.562725778791293e-06, "loss": 0.0, "step": 3067 }, { "epoch": 1.97, "learning_rate": 5.556507772589889e-06, "loss": 0.0, "step": 3068 }, { "epoch": 1.97, "learning_rate": 5.5502919063362095e-06, "loss": 0.0001, "step": 3069 }, { "epoch": 1.97, "learning_rate": 5.54407818302376e-06, "loss": 0.0, "step": 3070 }, { "epoch": 1.97, "learning_rate": 5.537866605645017e-06, "loss": 0.0001, "step": 3071 }, { "epoch": 1.97, "learning_rate": 5.5316571771914175e-06, "loss": 0.0, "step": 3072 }, { "epoch": 1.97, "learning_rate": 5.5254499006533765e-06, "loss": 0.0, "step": 3073 }, { "epoch": 1.98, "learning_rate": 5.519244779020262e-06, "loss": 0.0001, "step": 3074 }, { "epoch": 1.98, "learning_rate": 5.513041815280407e-06, "loss": 0.0001, "step": 3075 }, { "epoch": 1.98, "learning_rate": 5.506841012421107e-06, "loss": 0.0001, "step": 3076 }, { "epoch": 1.98, "learning_rate": 5.500642373428609e-06, "loss": 0.0, "step": 3077 }, { "epoch": 1.98, "learning_rate": 5.494445901288133e-06, "loss": 0.0, "step": 3078 }, { "epoch": 1.98, "learning_rate": 5.4882515989838444e-06, "loss": 0.0, "step": 3079 }, { "epoch": 1.98, "learning_rate": 5.4820594694988616e-06, "loss": 0.0001, "step": 3080 }, { "epoch": 1.98, "learning_rate": 5.475869515815271e-06, "loss": 0.0001, "step": 3081 }, { "epoch": 1.98, "learning_rate": 5.469681740914088e-06, "loss": 0.0001, "step": 3082 }, { "epoch": 1.98, "learning_rate": 5.4634961477753e-06, "loss": 0.0, "step": 3083 }, { "epoch": 1.98, "learning_rate": 5.457312739377836e-06, "loss": 0.0001, "step": 3084 }, { "epoch": 1.98, "learning_rate": 5.451131518699566e-06, "loss": 0.0001, "step": 3085 }, { "epoch": 1.98, "learning_rate": 5.444952488717321e-06, "loss": 0.0, "step": 3086 }, { "epoch": 1.98, "learning_rate": 5.438775652406866e-06, "loss": 0.0, "step": 3087 }, { "epoch": 1.98, "learning_rate": 5.432601012742911e-06, "loss": 0.0, "step": 3088 }, { "epoch": 1.98, "learning_rate": 5.42642857269911e-06, "loss": 0.0, "step": 3089 }, { "epoch": 1.99, "learning_rate": 5.420258335248052e-06, "loss": 0.0, "step": 3090 }, { "epoch": 1.99, "learning_rate": 5.41409030336128e-06, "loss": 0.0, "step": 3091 }, { "epoch": 1.99, "learning_rate": 5.407924480009261e-06, "loss": 0.0, "step": 3092 }, { "epoch": 1.99, "learning_rate": 5.4017608681614034e-06, "loss": 0.0001, "step": 3093 }, { "epoch": 1.99, "learning_rate": 5.395599470786043e-06, "loss": 0.0, "step": 3094 }, { "epoch": 1.99, "learning_rate": 5.389440290850465e-06, "loss": 0.0, "step": 3095 }, { "epoch": 1.99, "learning_rate": 5.383283331320875e-06, "loss": 0.0, "step": 3096 }, { "epoch": 1.99, "learning_rate": 5.377128595162408e-06, "loss": 0.0001, "step": 3097 }, { "epoch": 1.99, "learning_rate": 5.370976085339141e-06, "loss": 0.0001, "step": 3098 }, { "epoch": 1.99, "learning_rate": 5.364825804814058e-06, "loss": 0.0, "step": 3099 }, { "epoch": 1.99, "learning_rate": 5.358677756549091e-06, "loss": 0.0, "step": 3100 }, { "epoch": 1.99, "learning_rate": 5.3525319435050836e-06, "loss": 0.0004, "step": 3101 }, { "epoch": 1.99, "learning_rate": 5.3463883686418014e-06, "loss": 0.0001, "step": 3102 }, { "epoch": 1.99, "learning_rate": 5.340247034917953e-06, "loss": 0.0, "step": 3103 }, { "epoch": 1.99, "learning_rate": 5.334107945291132e-06, "loss": 0.0, "step": 3104 }, { "epoch": 2.0, "learning_rate": 5.327971102717886e-06, "loss": 0.0, "step": 3105 }, { "epoch": 2.0, "learning_rate": 5.3218365101536595e-06, "loss": 0.0, "step": 3106 }, { "epoch": 2.0, "learning_rate": 5.315704170552816e-06, "loss": 0.0, "step": 3107 }, { "epoch": 2.0, "learning_rate": 5.309574086868648e-06, "loss": 0.0001, "step": 3108 }, { "epoch": 2.0, "learning_rate": 5.303446262053345e-06, "loss": 0.0, "step": 3109 }, { "epoch": 2.0, "learning_rate": 5.297320699058015e-06, "loss": 0.0001, "step": 3110 }, { "epoch": 2.0, "learning_rate": 5.291197400832674e-06, "loss": 0.0001, "step": 3111 }, { "epoch": 2.0, "learning_rate": 5.285076370326257e-06, "loss": 0.0, "step": 3112 }, { "epoch": 2.0, "learning_rate": 5.278957610486595e-06, "loss": 0.0, "step": 3113 }, { "epoch": 2.0, "learning_rate": 5.272841124260431e-06, "loss": 0.0, "step": 3114 }, { "epoch": 2.0, "learning_rate": 5.26672691459341e-06, "loss": 0.0, "step": 3115 }, { "epoch": 2.0, "learning_rate": 5.260614984430084e-06, "loss": 0.0, "step": 3116 }, { "epoch": 2.0, "learning_rate": 5.254505336713908e-06, "loss": 0.0, "step": 3117 }, { "epoch": 2.0, "learning_rate": 5.248397974387238e-06, "loss": 0.0, "step": 3118 }, { "epoch": 2.0, "learning_rate": 5.242292900391319e-06, "loss": 0.0, "step": 3119 }, { "epoch": 2.0, "learning_rate": 5.2361901176663155e-06, "loss": 0.0, "step": 3120 }, { "epoch": 2.01, "learning_rate": 5.23008962915126e-06, "loss": 0.0001, "step": 3121 }, { "epoch": 2.01, "learning_rate": 5.2239914377841065e-06, "loss": 0.0, "step": 3122 }, { "epoch": 2.01, "learning_rate": 5.217895546501688e-06, "loss": 0.0001, "step": 3123 }, { "epoch": 2.01, "learning_rate": 5.211801958239733e-06, "loss": 0.0, "step": 3124 }, { "epoch": 2.01, "learning_rate": 5.2057106759328576e-06, "loss": 0.0, "step": 3125 }, { "epoch": 2.01, "learning_rate": 5.199621702514579e-06, "loss": 0.0, "step": 3126 }, { "epoch": 2.01, "learning_rate": 5.1935350409172905e-06, "loss": 0.0, "step": 3127 }, { "epoch": 2.01, "learning_rate": 5.1874506940722764e-06, "loss": 0.0, "step": 3128 }, { "epoch": 2.01, "learning_rate": 5.1813686649097065e-06, "loss": 0.0, "step": 3129 }, { "epoch": 2.01, "learning_rate": 5.175288956358628e-06, "loss": 0.0, "step": 3130 }, { "epoch": 2.01, "learning_rate": 5.169211571346986e-06, "loss": 0.0, "step": 3131 }, { "epoch": 2.01, "learning_rate": 5.163136512801592e-06, "loss": 0.0001, "step": 3132 }, { "epoch": 2.01, "learning_rate": 5.15706378364814e-06, "loss": 0.0, "step": 3133 }, { "epoch": 2.01, "learning_rate": 5.150993386811213e-06, "loss": 0.0, "step": 3134 }, { "epoch": 2.01, "learning_rate": 5.144925325214248e-06, "loss": 0.0, "step": 3135 }, { "epoch": 2.02, "learning_rate": 5.138859601779583e-06, "loss": 0.0, "step": 3136 }, { "epoch": 2.02, "learning_rate": 5.132796219428415e-06, "loss": 0.0, "step": 3137 }, { "epoch": 2.02, "learning_rate": 5.126735181080813e-06, "loss": 0.0, "step": 3138 }, { "epoch": 2.02, "learning_rate": 5.120676489655727e-06, "loss": 0.0, "step": 3139 }, { "epoch": 2.02, "learning_rate": 5.114620148070969e-06, "loss": 0.0, "step": 3140 }, { "epoch": 2.02, "learning_rate": 5.10856615924322e-06, "loss": 0.0, "step": 3141 }, { "epoch": 2.02, "learning_rate": 5.102514526088029e-06, "loss": 0.0, "step": 3142 }, { "epoch": 2.02, "learning_rate": 5.0964652515198085e-06, "loss": 0.0, "step": 3143 }, { "epoch": 2.02, "learning_rate": 5.090418338451844e-06, "loss": 0.0, "step": 3144 }, { "epoch": 2.02, "learning_rate": 5.084373789796271e-06, "loss": 0.0, "step": 3145 }, { "epoch": 2.02, "learning_rate": 5.078331608464093e-06, "loss": 0.0, "step": 3146 }, { "epoch": 2.02, "learning_rate": 5.072291797365172e-06, "loss": 0.0, "step": 3147 }, { "epoch": 2.02, "learning_rate": 5.066254359408234e-06, "loss": 0.0, "step": 3148 }, { "epoch": 2.02, "learning_rate": 5.060219297500853e-06, "loss": 0.0, "step": 3149 }, { "epoch": 2.02, "learning_rate": 5.054186614549461e-06, "loss": 0.0, "step": 3150 }, { "epoch": 2.02, "learning_rate": 5.048156313459358e-06, "loss": 0.0, "step": 3151 }, { "epoch": 2.03, "learning_rate": 5.042128397134669e-06, "loss": 0.0, "step": 3152 }, { "epoch": 2.03, "learning_rate": 5.036102868478396e-06, "loss": 0.0, "step": 3153 }, { "epoch": 2.03, "learning_rate": 5.030079730392383e-06, "loss": 0.0, "step": 3154 }, { "epoch": 2.03, "learning_rate": 5.024058985777313e-06, "loss": 0.0, "step": 3155 }, { "epoch": 2.03, "learning_rate": 5.0180406375327405e-06, "loss": 0.0, "step": 3156 }, { "epoch": 2.03, "learning_rate": 5.012024688557032e-06, "loss": 0.0, "step": 3157 }, { "epoch": 2.03, "learning_rate": 5.00601114174743e-06, "loss": 0.0, "step": 3158 }, { "epoch": 2.03, "learning_rate": 5.000000000000003e-06, "loss": 0.0, "step": 3159 }, { "epoch": 2.03, "learning_rate": 4.99399126620966e-06, "loss": 0.0, "step": 3160 }, { "epoch": 2.03, "learning_rate": 4.987984943270166e-06, "loss": 0.0002, "step": 3161 }, { "epoch": 2.03, "learning_rate": 4.981981034074107e-06, "loss": 0.0, "step": 3162 }, { "epoch": 2.03, "learning_rate": 4.975979541512918e-06, "loss": 0.0, "step": 3163 }, { "epoch": 2.03, "learning_rate": 4.969980468476859e-06, "loss": 0.0, "step": 3164 }, { "epoch": 2.03, "learning_rate": 4.963983817855044e-06, "loss": 0.0, "step": 3165 }, { "epoch": 2.03, "learning_rate": 4.9579895925354e-06, "loss": 0.0, "step": 3166 }, { "epoch": 2.04, "learning_rate": 4.951997795404698e-06, "loss": 0.0, "step": 3167 }, { "epoch": 2.04, "learning_rate": 4.946008429348534e-06, "loss": 0.0, "step": 3168 }, { "epoch": 2.04, "learning_rate": 4.940021497251334e-06, "loss": 0.0, "step": 3169 }, { "epoch": 2.04, "learning_rate": 4.934037001996357e-06, "loss": 0.0, "step": 3170 }, { "epoch": 2.04, "learning_rate": 4.928054946465684e-06, "loss": 0.0, "step": 3171 }, { "epoch": 2.04, "learning_rate": 4.922075333540215e-06, "loss": 0.0, "step": 3172 }, { "epoch": 2.04, "learning_rate": 4.916098166099695e-06, "loss": 0.0, "step": 3173 }, { "epoch": 2.04, "learning_rate": 4.910123447022658e-06, "loss": 0.0, "step": 3174 }, { "epoch": 2.04, "learning_rate": 4.90415117918649e-06, "loss": 0.0, "step": 3175 }, { "epoch": 2.04, "learning_rate": 4.898181365467382e-06, "loss": 0.0, "step": 3176 }, { "epoch": 2.04, "learning_rate": 4.892214008740339e-06, "loss": 0.0, "step": 3177 }, { "epoch": 2.04, "learning_rate": 4.886249111879196e-06, "loss": 0.0, "step": 3178 }, { "epoch": 2.04, "learning_rate": 4.880286677756595e-06, "loss": 0.0, "step": 3179 }, { "epoch": 2.04, "learning_rate": 4.8743267092439915e-06, "loss": 0.0, "step": 3180 }, { "epoch": 2.04, "learning_rate": 4.868369209211655e-06, "loss": 0.0, "step": 3181 }, { "epoch": 2.04, "learning_rate": 4.862414180528666e-06, "loss": 0.0, "step": 3182 }, { "epoch": 2.05, "learning_rate": 4.856461626062913e-06, "loss": 0.0, "step": 3183 }, { "epoch": 2.05, "learning_rate": 4.850511548681101e-06, "loss": 0.0, "step": 3184 }, { "epoch": 2.05, "learning_rate": 4.844563951248734e-06, "loss": 0.0, "step": 3185 }, { "epoch": 2.05, "learning_rate": 4.838618836630119e-06, "loss": 0.0, "step": 3186 }, { "epoch": 2.05, "learning_rate": 4.832676207688384e-06, "loss": 0.0, "step": 3187 }, { "epoch": 2.05, "learning_rate": 4.826736067285431e-06, "loss": 0.0, "step": 3188 }, { "epoch": 2.05, "learning_rate": 4.8207984182819955e-06, "loss": 0.0, "step": 3189 }, { "epoch": 2.05, "learning_rate": 4.814863263537593e-06, "loss": 0.0, "step": 3190 }, { "epoch": 2.05, "learning_rate": 4.808930605910539e-06, "loss": 0.0, "step": 3191 }, { "epoch": 2.05, "learning_rate": 4.803000448257961e-06, "loss": 0.0, "step": 3192 }, { "epoch": 2.05, "learning_rate": 4.797072793435759e-06, "loss": 0.0, "step": 3193 }, { "epoch": 2.05, "learning_rate": 4.79114764429865e-06, "loss": 0.0, "step": 3194 }, { "epoch": 2.05, "learning_rate": 4.785225003700132e-06, "loss": 0.0, "step": 3195 }, { "epoch": 2.05, "learning_rate": 4.779304874492494e-06, "loss": 0.0, "step": 3196 }, { "epoch": 2.05, "learning_rate": 4.773387259526827e-06, "loss": 0.0, "step": 3197 }, { "epoch": 2.05, "learning_rate": 4.767472161653e-06, "loss": 0.0, "step": 3198 }, { "epoch": 2.06, "learning_rate": 4.761559583719675e-06, "loss": 0.0, "step": 3199 }, { "epoch": 2.06, "learning_rate": 4.755649528574292e-06, "loss": 0.0, "step": 3200 }, { "epoch": 2.06, "learning_rate": 4.749741999063092e-06, "loss": 0.0, "step": 3201 }, { "epoch": 2.06, "learning_rate": 4.743836998031086e-06, "loss": 0.0, "step": 3202 }, { "epoch": 2.06, "learning_rate": 4.737934528322069e-06, "loss": 0.0, "step": 3203 }, { "epoch": 2.06, "learning_rate": 4.732034592778631e-06, "loss": 0.0, "step": 3204 }, { "epoch": 2.06, "learning_rate": 4.726137194242113e-06, "loss": 0.0, "step": 3205 }, { "epoch": 2.06, "learning_rate": 4.720242335552666e-06, "loss": 0.0, "step": 3206 }, { "epoch": 2.06, "learning_rate": 4.7143500195491986e-06, "loss": 0.0001, "step": 3207 }, { "epoch": 2.06, "learning_rate": 4.708460249069394e-06, "loss": 0.0, "step": 3208 }, { "epoch": 2.06, "learning_rate": 4.70257302694973e-06, "loss": 0.0001, "step": 3209 }, { "epoch": 2.06, "learning_rate": 4.696688356025424e-06, "loss": 0.0, "step": 3210 }, { "epoch": 2.06, "learning_rate": 4.690806239130494e-06, "loss": 0.0, "step": 3211 }, { "epoch": 2.06, "learning_rate": 4.684926679097717e-06, "loss": 0.0, "step": 3212 }, { "epoch": 2.06, "learning_rate": 4.679049678758631e-06, "loss": 0.0, "step": 3213 }, { "epoch": 2.07, "learning_rate": 4.673175240943558e-06, "loss": 0.0, "step": 3214 }, { "epoch": 2.07, "learning_rate": 4.6673033684815735e-06, "loss": 0.0, "step": 3215 }, { "epoch": 2.07, "learning_rate": 4.6614340642005205e-06, "loss": 0.0, "step": 3216 }, { "epoch": 2.07, "learning_rate": 4.655567330927002e-06, "loss": 0.0, "step": 3217 }, { "epoch": 2.07, "learning_rate": 4.649703171486394e-06, "loss": 0.0, "step": 3218 }, { "epoch": 2.07, "learning_rate": 4.64384158870282e-06, "loss": 0.0, "step": 3219 }, { "epoch": 2.07, "learning_rate": 4.63798258539917e-06, "loss": 0.0, "step": 3220 }, { "epoch": 2.07, "learning_rate": 4.63212616439709e-06, "loss": 0.0, "step": 3221 }, { "epoch": 2.07, "learning_rate": 4.626272328516975e-06, "loss": 0.0, "step": 3222 }, { "epoch": 2.07, "learning_rate": 4.6204210805779925e-06, "loss": 0.0, "step": 3223 }, { "epoch": 2.07, "learning_rate": 4.614572423398051e-06, "loss": 0.0, "step": 3224 }, { "epoch": 2.07, "learning_rate": 4.608726359793805e-06, "loss": 0.0, "step": 3225 }, { "epoch": 2.07, "learning_rate": 4.602882892580683e-06, "loss": 0.0, "step": 3226 }, { "epoch": 2.07, "learning_rate": 4.597042024572835e-06, "loss": 0.0, "step": 3227 }, { "epoch": 2.07, "learning_rate": 4.591203758583182e-06, "loss": 0.0, "step": 3228 }, { "epoch": 2.07, "learning_rate": 4.585368097423378e-06, "loss": 0.0, "step": 3229 }, { "epoch": 2.08, "learning_rate": 4.579535043903827e-06, "loss": 0.0, "step": 3230 }, { "epoch": 2.08, "learning_rate": 4.573704600833682e-06, "loss": 0.0001, "step": 3231 }, { "epoch": 2.08, "learning_rate": 4.5678767710208315e-06, "loss": 0.0, "step": 3232 }, { "epoch": 2.08, "learning_rate": 4.562051557271908e-06, "loss": 0.0, "step": 3233 }, { "epoch": 2.08, "learning_rate": 4.556228962392286e-06, "loss": 0.0, "step": 3234 }, { "epoch": 2.08, "learning_rate": 4.550408989186076e-06, "loss": 0.0001, "step": 3235 }, { "epoch": 2.08, "learning_rate": 4.544591640456123e-06, "loss": 0.0, "step": 3236 }, { "epoch": 2.08, "learning_rate": 4.538776919004021e-06, "loss": 0.0, "step": 3237 }, { "epoch": 2.08, "learning_rate": 4.532964827630084e-06, "loss": 0.0, "step": 3238 }, { "epoch": 2.08, "learning_rate": 4.527155369133364e-06, "loss": 0.0, "step": 3239 }, { "epoch": 2.08, "learning_rate": 4.521348546311657e-06, "loss": 0.0, "step": 3240 }, { "epoch": 2.08, "learning_rate": 4.515544361961463e-06, "loss": 0.0, "step": 3241 }, { "epoch": 2.08, "learning_rate": 4.50974281887804e-06, "loss": 0.0, "step": 3242 }, { "epoch": 2.08, "learning_rate": 4.503943919855357e-06, "loss": 0.0, "step": 3243 }, { "epoch": 2.08, "learning_rate": 4.49814766768611e-06, "loss": 0.0, "step": 3244 }, { "epoch": 2.09, "learning_rate": 4.492354065161735e-06, "loss": 0.0, "step": 3245 }, { "epoch": 2.09, "learning_rate": 4.486563115072368e-06, "loss": 0.0, "step": 3246 }, { "epoch": 2.09, "learning_rate": 4.48077482020689e-06, "loss": 0.0, "step": 3247 }, { "epoch": 2.09, "learning_rate": 4.474989183352892e-06, "loss": 0.0, "step": 3248 }, { "epoch": 2.09, "learning_rate": 4.469206207296682e-06, "loss": 0.0, "step": 3249 }, { "epoch": 2.09, "learning_rate": 4.463425894823301e-06, "loss": 0.0, "step": 3250 }, { "epoch": 2.09, "learning_rate": 4.457648248716493e-06, "loss": 0.0, "step": 3251 }, { "epoch": 2.09, "learning_rate": 4.451873271758723e-06, "loss": 0.0, "step": 3252 }, { "epoch": 2.09, "learning_rate": 4.4461009667311665e-06, "loss": 0.0, "step": 3253 }, { "epoch": 2.09, "learning_rate": 4.440331336413722e-06, "loss": 0.0, "step": 3254 }, { "epoch": 2.09, "learning_rate": 4.434564383584993e-06, "loss": 0.0, "step": 3255 }, { "epoch": 2.09, "learning_rate": 4.428800111022289e-06, "loss": 0.0, "step": 3256 }, { "epoch": 2.09, "learning_rate": 4.423038521501646e-06, "loss": 0.0, "step": 3257 }, { "epoch": 2.09, "learning_rate": 4.41727961779778e-06, "loss": 0.0, "step": 3258 }, { "epoch": 2.09, "learning_rate": 4.411523402684143e-06, "loss": 0.0, "step": 3259 }, { "epoch": 2.09, "learning_rate": 4.4057698789328726e-06, "loss": 0.0, "step": 3260 }, { "epoch": 2.1, "learning_rate": 4.400019049314814e-06, "loss": 0.0, "step": 3261 }, { "epoch": 2.1, "learning_rate": 4.39427091659953e-06, "loss": 0.0, "step": 3262 }, { "epoch": 2.1, "learning_rate": 4.3885254835552545e-06, "loss": 0.0, "step": 3263 }, { "epoch": 2.1, "learning_rate": 4.382782752948951e-06, "loss": 0.0, "step": 3264 }, { "epoch": 2.1, "learning_rate": 4.377042727546265e-06, "loss": 0.0, "step": 3265 }, { "epoch": 2.1, "learning_rate": 4.371305410111542e-06, "loss": 0.0, "step": 3266 }, { "epoch": 2.1, "learning_rate": 4.3655708034078305e-06, "loss": 0.0, "step": 3267 }, { "epoch": 2.1, "learning_rate": 4.359838910196863e-06, "loss": 0.0, "step": 3268 }, { "epoch": 2.1, "learning_rate": 4.354109733239072e-06, "loss": 0.0, "step": 3269 }, { "epoch": 2.1, "learning_rate": 4.348383275293575e-06, "loss": 0.0, "step": 3270 }, { "epoch": 2.1, "learning_rate": 4.342659539118194e-06, "loss": 0.0, "step": 3271 }, { "epoch": 2.1, "learning_rate": 4.336938527469425e-06, "loss": 0.0, "step": 3272 }, { "epoch": 2.1, "learning_rate": 4.3312202431024605e-06, "loss": 0.0, "step": 3273 }, { "epoch": 2.1, "learning_rate": 4.325504688771176e-06, "loss": 0.0, "step": 3274 }, { "epoch": 2.1, "learning_rate": 4.3197918672281305e-06, "loss": 0.0, "step": 3275 }, { "epoch": 2.11, "learning_rate": 4.314081781224576e-06, "loss": 0.0, "step": 3276 }, { "epoch": 2.11, "learning_rate": 4.308374433510438e-06, "loss": 0.0, "step": 3277 }, { "epoch": 2.11, "learning_rate": 4.302669826834324e-06, "loss": 0.0, "step": 3278 }, { "epoch": 2.11, "learning_rate": 4.296967963943532e-06, "loss": 0.0, "step": 3279 }, { "epoch": 2.11, "learning_rate": 4.291268847584016e-06, "loss": 0.0, "step": 3280 }, { "epoch": 2.11, "learning_rate": 4.285572480500434e-06, "loss": 0.0, "step": 3281 }, { "epoch": 2.11, "learning_rate": 4.279878865436103e-06, "loss": 0.0, "step": 3282 }, { "epoch": 2.11, "learning_rate": 4.274188005133014e-06, "loss": 0.0, "step": 3283 }, { "epoch": 2.11, "learning_rate": 4.268499902331851e-06, "loss": 0.0, "step": 3284 }, { "epoch": 2.11, "learning_rate": 4.262814559771938e-06, "loss": 0.0, "step": 3285 }, { "epoch": 2.11, "learning_rate": 4.257131980191299e-06, "loss": 0.0, "step": 3286 }, { "epoch": 2.11, "learning_rate": 4.251452166326612e-06, "loss": 0.0, "step": 3287 }, { "epoch": 2.11, "learning_rate": 4.245775120913226e-06, "loss": 0.0, "step": 3288 }, { "epoch": 2.11, "learning_rate": 4.240100846685154e-06, "loss": 0.0, "step": 3289 }, { "epoch": 2.11, "learning_rate": 4.234429346375085e-06, "loss": 0.0, "step": 3290 }, { "epoch": 2.11, "learning_rate": 4.228760622714363e-06, "loss": 0.0, "step": 3291 }, { "epoch": 2.12, "learning_rate": 4.22309467843299e-06, "loss": 0.0, "step": 3292 }, { "epoch": 2.12, "learning_rate": 4.21743151625965e-06, "loss": 0.0, "step": 3293 }, { "epoch": 2.12, "learning_rate": 4.211771138921657e-06, "loss": 0.0, "step": 3294 }, { "epoch": 2.12, "learning_rate": 4.20611354914501e-06, "loss": 0.0, "step": 3295 }, { "epoch": 2.12, "learning_rate": 4.200458749654355e-06, "loss": 0.0, "step": 3296 }, { "epoch": 2.12, "learning_rate": 4.1948067431729875e-06, "loss": 0.0, "step": 3297 }, { "epoch": 2.12, "learning_rate": 4.1891575324228805e-06, "loss": 0.0, "step": 3298 }, { "epoch": 2.12, "learning_rate": 4.183511120124628e-06, "loss": 0.0, "step": 3299 }, { "epoch": 2.12, "learning_rate": 4.1778675089975055e-06, "loss": 0.0, "step": 3300 }, { "epoch": 2.12, "learning_rate": 4.172226701759424e-06, "loss": 0.0, "step": 3301 }, { "epoch": 2.12, "learning_rate": 4.166588701126944e-06, "loss": 0.0, "step": 3302 }, { "epoch": 2.12, "learning_rate": 4.160953509815286e-06, "loss": 0.0, "step": 3303 }, { "epoch": 2.12, "learning_rate": 4.155321130538305e-06, "loss": 0.0, "step": 3304 }, { "epoch": 2.12, "learning_rate": 4.149691566008508e-06, "loss": 0.0001, "step": 3305 }, { "epoch": 2.12, "learning_rate": 4.14406481893704e-06, "loss": 0.0, "step": 3306 }, { "epoch": 2.12, "learning_rate": 4.138440892033702e-06, "loss": 0.0, "step": 3307 }, { "epoch": 2.13, "learning_rate": 4.132819788006927e-06, "loss": 0.0, "step": 3308 }, { "epoch": 2.13, "learning_rate": 4.127201509563783e-06, "loss": 0.0, "step": 3309 }, { "epoch": 2.13, "learning_rate": 4.121586059409999e-06, "loss": 0.0, "step": 3310 }, { "epoch": 2.13, "learning_rate": 4.1159734402499106e-06, "loss": 0.0, "step": 3311 }, { "epoch": 2.13, "learning_rate": 4.11036365478652e-06, "loss": 0.0001, "step": 3312 }, { "epoch": 2.13, "learning_rate": 4.104756705721447e-06, "loss": 0.0, "step": 3313 }, { "epoch": 2.13, "learning_rate": 4.099152595754946e-06, "loss": 0.0, "step": 3314 }, { "epoch": 2.13, "learning_rate": 4.093551327585921e-06, "loss": 0.0002, "step": 3315 }, { "epoch": 2.13, "learning_rate": 4.087952903911878e-06, "loss": 0.0, "step": 3316 }, { "epoch": 2.13, "learning_rate": 4.0823573274289815e-06, "loss": 0.0, "step": 3317 }, { "epoch": 2.13, "learning_rate": 4.076764600832011e-06, "loss": 0.0, "step": 3318 }, { "epoch": 2.13, "learning_rate": 4.07117472681437e-06, "loss": 0.0, "step": 3319 }, { "epoch": 2.13, "learning_rate": 4.065587708068102e-06, "loss": 0.0, "step": 3320 }, { "epoch": 2.13, "learning_rate": 4.060003547283863e-06, "loss": 0.0, "step": 3321 }, { "epoch": 2.13, "learning_rate": 4.0544222471509375e-06, "loss": 0.0, "step": 3322 }, { "epoch": 2.14, "learning_rate": 4.048843810357228e-06, "loss": 0.0, "step": 3323 }, { "epoch": 2.14, "learning_rate": 4.043268239589269e-06, "loss": 0.0, "step": 3324 }, { "epoch": 2.14, "learning_rate": 4.037695537532205e-06, "loss": 0.0001, "step": 3325 }, { "epoch": 2.14, "learning_rate": 4.032125706869798e-06, "loss": 0.0, "step": 3326 }, { "epoch": 2.14, "learning_rate": 4.026558750284435e-06, "loss": 0.0, "step": 3327 }, { "epoch": 2.14, "learning_rate": 4.020994670457109e-06, "loss": 0.0, "step": 3328 }, { "epoch": 2.14, "learning_rate": 4.015433470067439e-06, "loss": 0.0, "step": 3329 }, { "epoch": 2.14, "learning_rate": 4.009875151793649e-06, "loss": 0.0, "step": 3330 }, { "epoch": 2.14, "learning_rate": 4.004319718312575e-06, "loss": 0.0, "step": 3331 }, { "epoch": 2.14, "learning_rate": 3.998767172299676e-06, "loss": 0.0, "step": 3332 }, { "epoch": 2.14, "learning_rate": 3.993217516428994e-06, "loss": 0.0, "step": 3333 }, { "epoch": 2.14, "learning_rate": 3.9876707533732095e-06, "loss": 0.0001, "step": 3334 }, { "epoch": 2.14, "learning_rate": 3.982126885803591e-06, "loss": 0.0, "step": 3335 }, { "epoch": 2.14, "learning_rate": 3.976585916390014e-06, "loss": 0.0, "step": 3336 }, { "epoch": 2.14, "learning_rate": 3.971047847800973e-06, "loss": 0.0, "step": 3337 }, { "epoch": 2.14, "learning_rate": 3.965512682703538e-06, "loss": 0.0, "step": 3338 }, { "epoch": 2.15, "learning_rate": 3.959980423763411e-06, "loss": 0.0, "step": 3339 }, { "epoch": 2.15, "learning_rate": 3.954451073644875e-06, "loss": 0.0, "step": 3340 }, { "epoch": 2.15, "learning_rate": 3.948924635010814e-06, "loss": 0.0, "step": 3341 }, { "epoch": 2.15, "learning_rate": 3.943401110522721e-06, "loss": 0.0, "step": 3342 }, { "epoch": 2.15, "learning_rate": 3.937880502840674e-06, "loss": 0.0, "step": 3343 }, { "epoch": 2.15, "learning_rate": 3.932362814623351e-06, "loss": 0.0, "step": 3344 }, { "epoch": 2.15, "learning_rate": 3.926848048528018e-06, "loss": 0.0001, "step": 3345 }, { "epoch": 2.15, "learning_rate": 3.921336207210552e-06, "loss": 0.0, "step": 3346 }, { "epoch": 2.15, "learning_rate": 3.9158272933253924e-06, "loss": 0.0, "step": 3347 }, { "epoch": 2.15, "learning_rate": 3.910321309525596e-06, "loss": 0.0, "step": 3348 }, { "epoch": 2.15, "learning_rate": 3.904818258462796e-06, "loss": 0.0, "step": 3349 }, { "epoch": 2.15, "learning_rate": 3.899318142787207e-06, "loss": 0.0, "step": 3350 }, { "epoch": 2.15, "learning_rate": 3.893820965147651e-06, "loss": 0.0, "step": 3351 }, { "epoch": 2.15, "learning_rate": 3.8883267281915075e-06, "loss": 0.0, "step": 3352 }, { "epoch": 2.15, "learning_rate": 3.8828354345647625e-06, "loss": 0.0, "step": 3353 }, { "epoch": 2.16, "learning_rate": 3.877347086911973e-06, "loss": 0.0, "step": 3354 }, { "epoch": 2.16, "learning_rate": 3.871861687876278e-06, "loss": 0.0, "step": 3355 }, { "epoch": 2.16, "learning_rate": 3.866379240099404e-06, "loss": 0.0002, "step": 3356 }, { "epoch": 2.16, "learning_rate": 3.860899746221648e-06, "loss": 0.0, "step": 3357 }, { "epoch": 2.16, "learning_rate": 3.855423208881887e-06, "loss": 0.0, "step": 3358 }, { "epoch": 2.16, "learning_rate": 3.8499496307175724e-06, "loss": 0.0, "step": 3359 }, { "epoch": 2.16, "learning_rate": 3.844479014364737e-06, "loss": 0.0, "step": 3360 }, { "epoch": 2.16, "learning_rate": 3.839011362457982e-06, "loss": 0.0, "step": 3361 }, { "epoch": 2.16, "learning_rate": 3.833546677630474e-06, "loss": 0.0, "step": 3362 }, { "epoch": 2.16, "learning_rate": 3.828084962513973e-06, "loss": 0.0, "step": 3363 }, { "epoch": 2.16, "learning_rate": 3.822626219738778e-06, "loss": 0.0, "step": 3364 }, { "epoch": 2.16, "learning_rate": 3.817170451933783e-06, "loss": 0.0, "step": 3365 }, { "epoch": 2.16, "learning_rate": 3.811717661726436e-06, "loss": 0.0, "step": 3366 }, { "epoch": 2.16, "learning_rate": 3.8062678517427507e-06, "loss": 0.0001, "step": 3367 }, { "epoch": 2.16, "learning_rate": 3.80082102460732e-06, "loss": 0.0, "step": 3368 }, { "epoch": 2.16, "learning_rate": 3.795377182943274e-06, "loss": 0.0, "step": 3369 }, { "epoch": 2.17, "learning_rate": 3.78993632937233e-06, "loss": 0.0, "step": 3370 }, { "epoch": 2.17, "learning_rate": 3.7844984665147543e-06, "loss": 0.0, "step": 3371 }, { "epoch": 2.17, "learning_rate": 3.779063596989371e-06, "loss": 0.0, "step": 3372 }, { "epoch": 2.17, "learning_rate": 3.7736317234135733e-06, "loss": 0.0, "step": 3373 }, { "epoch": 2.17, "learning_rate": 3.7682028484033015e-06, "loss": 0.0, "step": 3374 }, { "epoch": 2.17, "learning_rate": 3.762776974573056e-06, "loss": 0.0, "step": 3375 }, { "epoch": 2.17, "learning_rate": 3.7573541045358896e-06, "loss": 0.0, "step": 3376 }, { "epoch": 2.17, "learning_rate": 3.751934240903408e-06, "loss": 0.0, "step": 3377 }, { "epoch": 2.17, "learning_rate": 3.746517386285776e-06, "loss": 0.0, "step": 3378 }, { "epoch": 2.17, "learning_rate": 3.7411035432917032e-06, "loss": 0.0, "step": 3379 }, { "epoch": 2.17, "learning_rate": 3.7356927145284483e-06, "loss": 0.0001, "step": 3380 }, { "epoch": 2.17, "learning_rate": 3.7302849026018175e-06, "loss": 0.0, "step": 3381 }, { "epoch": 2.17, "learning_rate": 3.7248801101161724e-06, "loss": 0.0, "step": 3382 }, { "epoch": 2.17, "learning_rate": 3.7194783396744127e-06, "loss": 0.0, "step": 3383 }, { "epoch": 2.17, "learning_rate": 3.7140795938779807e-06, "loss": 0.0, "step": 3384 }, { "epoch": 2.18, "learning_rate": 3.708683875326875e-06, "loss": 0.0, "step": 3385 }, { "epoch": 2.18, "learning_rate": 3.7032911866196154e-06, "loss": 0.0, "step": 3386 }, { "epoch": 2.18, "learning_rate": 3.697901530353284e-06, "loss": 0.0, "step": 3387 }, { "epoch": 2.18, "learning_rate": 3.692514909123489e-06, "loss": 0.0, "step": 3388 }, { "epoch": 2.18, "learning_rate": 3.687131325524379e-06, "loss": 0.0, "step": 3389 }, { "epoch": 2.18, "learning_rate": 3.681750782148651e-06, "loss": 0.0, "step": 3390 }, { "epoch": 2.18, "learning_rate": 3.676373281587515e-06, "loss": 0.0, "step": 3391 }, { "epoch": 2.18, "learning_rate": 3.6709988264307394e-06, "loss": 0.0, "step": 3392 }, { "epoch": 2.18, "learning_rate": 3.665627419266612e-06, "loss": 0.0, "step": 3393 }, { "epoch": 2.18, "learning_rate": 3.6602590626819524e-06, "loss": 0.0, "step": 3394 }, { "epoch": 2.18, "learning_rate": 3.6548937592621235e-06, "loss": 0.0, "step": 3395 }, { "epoch": 2.18, "learning_rate": 3.649531511591006e-06, "loss": 0.0, "step": 3396 }, { "epoch": 2.18, "learning_rate": 3.6441723222510105e-06, "loss": 0.0, "step": 3397 }, { "epoch": 2.18, "learning_rate": 3.6388161938230737e-06, "loss": 0.0, "step": 3398 }, { "epoch": 2.18, "learning_rate": 3.633463128886673e-06, "loss": 0.0, "step": 3399 }, { "epoch": 2.18, "learning_rate": 3.6281131300197825e-06, "loss": 0.0, "step": 3400 }, { "epoch": 2.19, "learning_rate": 3.6227661997989238e-06, "loss": 0.0, "step": 3401 }, { "epoch": 2.19, "learning_rate": 3.6174223407991405e-06, "loss": 0.0, "step": 3402 }, { "epoch": 2.19, "learning_rate": 3.6120815555939736e-06, "loss": 0.0, "step": 3403 }, { "epoch": 2.19, "learning_rate": 3.6067438467555137e-06, "loss": 0.0, "step": 3404 }, { "epoch": 2.19, "learning_rate": 3.6014092168543423e-06, "loss": 0.0, "step": 3405 }, { "epoch": 2.19, "learning_rate": 3.5960776684595824e-06, "loss": 0.0, "step": 3406 }, { "epoch": 2.19, "learning_rate": 3.590749204138857e-06, "loss": 0.0, "step": 3407 }, { "epoch": 2.19, "learning_rate": 3.5854238264583052e-06, "loss": 0.0, "step": 3408 }, { "epoch": 2.19, "learning_rate": 3.580101537982592e-06, "loss": 0.0, "step": 3409 }, { "epoch": 2.19, "learning_rate": 3.5747823412748807e-06, "loss": 0.0, "step": 3410 }, { "epoch": 2.19, "learning_rate": 3.5694662388968504e-06, "loss": 0.0, "step": 3411 }, { "epoch": 2.19, "learning_rate": 3.5641532334086894e-06, "loss": 0.0, "step": 3412 }, { "epoch": 2.19, "learning_rate": 3.5588433273691004e-06, "loss": 0.0, "step": 3413 }, { "epoch": 2.19, "learning_rate": 3.553536523335285e-06, "loss": 0.0, "step": 3414 }, { "epoch": 2.19, "learning_rate": 3.548232823862955e-06, "loss": 0.0, "step": 3415 }, { "epoch": 2.2, "learning_rate": 3.5429322315063276e-06, "loss": 0.0, "step": 3416 }, { "epoch": 2.2, "learning_rate": 3.5376347488181174e-06, "loss": 0.0, "step": 3417 }, { "epoch": 2.2, "learning_rate": 3.5323403783495546e-06, "loss": 0.0, "step": 3418 }, { "epoch": 2.2, "learning_rate": 3.527049122650359e-06, "loss": 0.0, "step": 3419 }, { "epoch": 2.2, "learning_rate": 3.5217609842687494e-06, "loss": 0.0, "step": 3420 }, { "epoch": 2.2, "learning_rate": 3.51647596575146e-06, "loss": 0.0001, "step": 3421 }, { "epoch": 2.2, "learning_rate": 3.5111940696436954e-06, "loss": 0.0, "step": 3422 }, { "epoch": 2.2, "learning_rate": 3.505915298489182e-06, "loss": 0.0, "step": 3423 }, { "epoch": 2.2, "learning_rate": 3.500639654830127e-06, "loss": 0.0001, "step": 3424 }, { "epoch": 2.2, "learning_rate": 3.4953671412072323e-06, "loss": 0.0, "step": 3425 }, { "epoch": 2.2, "learning_rate": 3.490097760159702e-06, "loss": 0.0, "step": 3426 }, { "epoch": 2.2, "learning_rate": 3.48483151422522e-06, "loss": 0.0, "step": 3427 }, { "epoch": 2.2, "learning_rate": 3.479568405939967e-06, "loss": 0.0001, "step": 3428 }, { "epoch": 2.2, "learning_rate": 3.4743084378386104e-06, "loss": 0.0, "step": 3429 }, { "epoch": 2.2, "learning_rate": 3.4690516124543017e-06, "loss": 0.0, "step": 3430 }, { "epoch": 2.2, "learning_rate": 3.463797932318689e-06, "loss": 0.0, "step": 3431 }, { "epoch": 2.21, "learning_rate": 3.458547399961898e-06, "loss": 0.0, "step": 3432 }, { "epoch": 2.21, "learning_rate": 3.4533000179125387e-06, "loss": 0.0, "step": 3433 }, { "epoch": 2.21, "learning_rate": 3.448055788697703e-06, "loss": 0.0, "step": 3434 }, { "epoch": 2.21, "learning_rate": 3.4428147148429715e-06, "loss": 0.0, "step": 3435 }, { "epoch": 2.21, "learning_rate": 3.4375767988723984e-06, "loss": 0.0, "step": 3436 }, { "epoch": 2.21, "learning_rate": 3.4323420433085164e-06, "loss": 0.0, "step": 3437 }, { "epoch": 2.21, "learning_rate": 3.4271104506723474e-06, "loss": 0.0, "step": 3438 }, { "epoch": 2.21, "learning_rate": 3.4218820234833694e-06, "loss": 0.0, "step": 3439 }, { "epoch": 2.21, "learning_rate": 3.4166567642595573e-06, "loss": 0.0, "step": 3440 }, { "epoch": 2.21, "learning_rate": 3.4114346755173467e-06, "loss": 0.0, "step": 3441 }, { "epoch": 2.21, "learning_rate": 3.4062157597716493e-06, "loss": 0.0001, "step": 3442 }, { "epoch": 2.21, "learning_rate": 3.4010000195358605e-06, "loss": 0.0, "step": 3443 }, { "epoch": 2.21, "learning_rate": 3.3957874573218207e-06, "loss": 0.0, "step": 3444 }, { "epoch": 2.21, "learning_rate": 3.3905780756398664e-06, "loss": 0.0, "step": 3445 }, { "epoch": 2.21, "learning_rate": 3.3853718769987876e-06, "loss": 0.0, "step": 3446 }, { "epoch": 2.21, "learning_rate": 3.3801688639058404e-06, "loss": 0.0, "step": 3447 }, { "epoch": 2.22, "learning_rate": 3.37496903886676e-06, "loss": 0.0, "step": 3448 }, { "epoch": 2.22, "learning_rate": 3.3697724043857326e-06, "loss": 0.0, "step": 3449 }, { "epoch": 2.22, "learning_rate": 3.3645789629654135e-06, "loss": 0.0, "step": 3450 }, { "epoch": 2.22, "learning_rate": 3.359388717106915e-06, "loss": 0.0, "step": 3451 }, { "epoch": 2.22, "learning_rate": 3.354201669309827e-06, "loss": 0.0, "step": 3452 }, { "epoch": 2.22, "learning_rate": 3.3490178220721724e-06, "loss": 0.0, "step": 3453 }, { "epoch": 2.22, "learning_rate": 3.3438371778904554e-06, "loss": 0.0, "step": 3454 }, { "epoch": 2.22, "learning_rate": 3.3386597392596343e-06, "loss": 0.0, "step": 3455 }, { "epoch": 2.22, "learning_rate": 3.3334855086731077e-06, "loss": 0.0, "step": 3456 }, { "epoch": 2.22, "learning_rate": 3.3283144886227535e-06, "loss": 0.0, "step": 3457 }, { "epoch": 2.22, "learning_rate": 3.323146681598878e-06, "loss": 0.0, "step": 3458 }, { "epoch": 2.22, "learning_rate": 3.3179820900902605e-06, "loss": 0.0, "step": 3459 }, { "epoch": 2.22, "learning_rate": 3.312820716584122e-06, "loss": 0.0, "step": 3460 }, { "epoch": 2.22, "learning_rate": 3.307662563566133e-06, "loss": 0.0, "step": 3461 }, { "epoch": 2.22, "learning_rate": 3.302507633520421e-06, "loss": 0.0, "step": 3462 }, { "epoch": 2.23, "learning_rate": 3.297355928929554e-06, "loss": 0.0, "step": 3463 }, { "epoch": 2.23, "learning_rate": 3.2922074522745483e-06, "loss": 0.0001, "step": 3464 }, { "epoch": 2.23, "learning_rate": 3.2870622060348624e-06, "loss": 0.0, "step": 3465 }, { "epoch": 2.23, "learning_rate": 3.2819201926884094e-06, "loss": 0.0, "step": 3466 }, { "epoch": 2.23, "learning_rate": 3.276781414711535e-06, "loss": 0.0, "step": 3467 }, { "epoch": 2.23, "learning_rate": 3.2716458745790325e-06, "loss": 0.0, "step": 3468 }, { "epoch": 2.23, "learning_rate": 3.266513574764132e-06, "loss": 0.0, "step": 3469 }, { "epoch": 2.23, "learning_rate": 3.261384517738503e-06, "loss": 0.0, "step": 3470 }, { "epoch": 2.23, "learning_rate": 3.2562587059722605e-06, "loss": 0.0, "step": 3471 }, { "epoch": 2.23, "learning_rate": 3.2511361419339493e-06, "loss": 0.0001, "step": 3472 }, { "epoch": 2.23, "learning_rate": 3.246016828090549e-06, "loss": 0.0, "step": 3473 }, { "epoch": 2.23, "learning_rate": 3.2409007669074867e-06, "loss": 0.0, "step": 3474 }, { "epoch": 2.23, "learning_rate": 3.2357879608486008e-06, "loss": 0.0, "step": 3475 }, { "epoch": 2.23, "learning_rate": 3.2306784123761837e-06, "loss": 0.0, "step": 3476 }, { "epoch": 2.23, "learning_rate": 3.2255721239509472e-06, "loss": 0.0, "step": 3477 }, { "epoch": 2.23, "learning_rate": 3.220469098032034e-06, "loss": 0.0, "step": 3478 }, { "epoch": 2.24, "learning_rate": 3.2153693370770235e-06, "loss": 0.0, "step": 3479 }, { "epoch": 2.24, "learning_rate": 3.2102728435419116e-06, "loss": 0.0, "step": 3480 }, { "epoch": 2.24, "learning_rate": 3.205179619881128e-06, "loss": 0.0, "step": 3481 }, { "epoch": 2.24, "learning_rate": 3.2000896685475258e-06, "loss": 0.0, "step": 3482 }, { "epoch": 2.24, "learning_rate": 3.1950029919923777e-06, "loss": 0.0, "step": 3483 }, { "epoch": 2.24, "learning_rate": 3.1899195926653893e-06, "loss": 0.0, "step": 3484 }, { "epoch": 2.24, "learning_rate": 3.1848394730146794e-06, "loss": 0.0, "step": 3485 }, { "epoch": 2.24, "learning_rate": 3.1797626354867904e-06, "loss": 0.0, "step": 3486 }, { "epoch": 2.24, "learning_rate": 3.1746890825266794e-06, "loss": 0.0, "step": 3487 }, { "epoch": 2.24, "learning_rate": 3.169618816577735e-06, "loss": 0.0, "step": 3488 }, { "epoch": 2.24, "learning_rate": 3.1645518400817475e-06, "loss": 0.0, "step": 3489 }, { "epoch": 2.24, "learning_rate": 3.159488155478928e-06, "loss": 0.0, "step": 3490 }, { "epoch": 2.24, "learning_rate": 3.1544277652079123e-06, "loss": 0.0, "step": 3491 }, { "epoch": 2.24, "learning_rate": 3.1493706717057283e-06, "loss": 0.0, "step": 3492 }, { "epoch": 2.24, "learning_rate": 3.144316877407839e-06, "loss": 0.0, "step": 3493 }, { "epoch": 2.25, "learning_rate": 3.1392663847481044e-06, "loss": 0.0, "step": 3494 }, { "epoch": 2.25, "learning_rate": 3.134219196158795e-06, "loss": 0.0, "step": 3495 }, { "epoch": 2.25, "learning_rate": 3.1291753140706037e-06, "loss": 0.0001, "step": 3496 }, { "epoch": 2.25, "learning_rate": 3.1241347409126086e-06, "loss": 0.0, "step": 3497 }, { "epoch": 2.25, "learning_rate": 3.119097479112315e-06, "loss": 0.0, "step": 3498 }, { "epoch": 2.25, "learning_rate": 3.1140635310956223e-06, "loss": 0.0, "step": 3499 }, { "epoch": 2.25, "learning_rate": 3.1090328992868323e-06, "loss": 0.0, "step": 3500 }, { "epoch": 2.25, "learning_rate": 3.1040055861086625e-06, "loss": 0.0, "step": 3501 }, { "epoch": 2.25, "learning_rate": 3.0989815939822198e-06, "loss": 0.0, "step": 3502 }, { "epoch": 2.25, "learning_rate": 3.0939609253270152e-06, "loss": 0.0, "step": 3503 }, { "epoch": 2.25, "learning_rate": 3.088943582560956e-06, "loss": 0.0, "step": 3504 }, { "epoch": 2.25, "learning_rate": 3.083929568100359e-06, "loss": 0.0, "step": 3505 }, { "epoch": 2.25, "learning_rate": 3.0789188843599273e-06, "loss": 0.0, "step": 3506 }, { "epoch": 2.25, "learning_rate": 3.0739115337527626e-06, "loss": 0.0, "step": 3507 }, { "epoch": 2.25, "learning_rate": 3.068907518690362e-06, "loss": 0.0, "step": 3508 }, { "epoch": 2.25, "learning_rate": 3.0639068415826123e-06, "loss": 0.0, "step": 3509 }, { "epoch": 2.26, "learning_rate": 3.058909504837808e-06, "loss": 0.0, "step": 3510 }, { "epoch": 2.26, "learning_rate": 3.0539155108626106e-06, "loss": 0.0, "step": 3511 }, { "epoch": 2.26, "learning_rate": 3.0489248620620925e-06, "loss": 0.0, "step": 3512 }, { "epoch": 2.26, "learning_rate": 3.0439375608397058e-06, "loss": 0.0, "step": 3513 }, { "epoch": 2.26, "learning_rate": 3.038953609597287e-06, "loss": 0.0, "step": 3514 }, { "epoch": 2.26, "learning_rate": 3.033973010735074e-06, "loss": 0.0, "step": 3515 }, { "epoch": 2.26, "learning_rate": 3.0289957666516735e-06, "loss": 0.0, "step": 3516 }, { "epoch": 2.26, "learning_rate": 3.024021879744086e-06, "loss": 0.0001, "step": 3517 }, { "epoch": 2.26, "learning_rate": 3.019051352407688e-06, "loss": 0.0, "step": 3518 }, { "epoch": 2.26, "learning_rate": 3.0140841870362505e-06, "loss": 0.0, "step": 3519 }, { "epoch": 2.26, "learning_rate": 3.009120386021914e-06, "loss": 0.0, "step": 3520 }, { "epoch": 2.26, "learning_rate": 3.0041599517552024e-06, "loss": 0.0, "step": 3521 }, { "epoch": 2.26, "learning_rate": 2.9992028866250177e-06, "loss": 0.0, "step": 3522 }, { "epoch": 2.26, "learning_rate": 2.994249193018638e-06, "loss": 0.0, "step": 3523 }, { "epoch": 2.26, "learning_rate": 2.9892988733217255e-06, "loss": 0.0, "step": 3524 }, { "epoch": 2.27, "learning_rate": 2.9843519299183076e-06, "loss": 0.0, "step": 3525 }, { "epoch": 2.27, "learning_rate": 2.9794083651907878e-06, "loss": 0.0, "step": 3526 }, { "epoch": 2.27, "learning_rate": 2.974468181519954e-06, "loss": 0.0, "step": 3527 }, { "epoch": 2.27, "learning_rate": 2.9695313812849426e-06, "loss": 0.0, "step": 3528 }, { "epoch": 2.27, "learning_rate": 2.964597966863284e-06, "loss": 0.0, "step": 3529 }, { "epoch": 2.27, "learning_rate": 2.959667940630865e-06, "loss": 0.0, "step": 3530 }, { "epoch": 2.27, "learning_rate": 2.954741304961941e-06, "loss": 0.0, "step": 3531 }, { "epoch": 2.27, "learning_rate": 2.949818062229145e-06, "loss": 0.0, "step": 3532 }, { "epoch": 2.27, "learning_rate": 2.9448982148034633e-06, "loss": 0.0, "step": 3533 }, { "epoch": 2.27, "learning_rate": 2.939981765054253e-06, "loss": 0.0, "step": 3534 }, { "epoch": 2.27, "learning_rate": 2.935068715349234e-06, "loss": 0.0, "step": 3535 }, { "epoch": 2.27, "learning_rate": 2.9301590680544868e-06, "loss": 0.0, "step": 3536 }, { "epoch": 2.27, "learning_rate": 2.925252825534459e-06, "loss": 0.0, "step": 3537 }, { "epoch": 2.27, "learning_rate": 2.920349990151955e-06, "loss": 0.0, "step": 3538 }, { "epoch": 2.27, "learning_rate": 2.9154505642681364e-06, "loss": 0.0, "step": 3539 }, { "epoch": 2.27, "learning_rate": 2.9105545502425214e-06, "loss": 0.0, "step": 3540 }, { "epoch": 2.28, "learning_rate": 2.9056619504329954e-06, "loss": 0.0, "step": 3541 }, { "epoch": 2.28, "learning_rate": 2.9007727671957885e-06, "loss": 0.0, "step": 3542 }, { "epoch": 2.28, "learning_rate": 2.895887002885488e-06, "loss": 0.0001, "step": 3543 }, { "epoch": 2.28, "learning_rate": 2.8910046598550433e-06, "loss": 0.0, "step": 3544 }, { "epoch": 2.28, "learning_rate": 2.8861257404557376e-06, "loss": 0.0, "step": 3545 }, { "epoch": 2.28, "learning_rate": 2.881250247037225e-06, "loss": 0.0, "step": 3546 }, { "epoch": 2.28, "learning_rate": 2.8763781819475e-06, "loss": 0.0, "step": 3547 }, { "epoch": 2.28, "learning_rate": 2.8715095475329023e-06, "loss": 0.0, "step": 3548 }, { "epoch": 2.28, "learning_rate": 2.8666443461381345e-06, "loss": 0.0, "step": 3549 }, { "epoch": 2.28, "learning_rate": 2.8617825801062236e-06, "loss": 0.0, "step": 3550 }, { "epoch": 2.28, "learning_rate": 2.8569242517785636e-06, "loss": 0.0, "step": 3551 }, { "epoch": 2.28, "learning_rate": 2.8520693634948804e-06, "loss": 0.0, "step": 3552 }, { "epoch": 2.28, "learning_rate": 2.8472179175932434e-06, "loss": 0.0, "step": 3553 }, { "epoch": 2.28, "learning_rate": 2.8423699164100747e-06, "loss": 0.0, "step": 3554 }, { "epoch": 2.28, "learning_rate": 2.837525362280126e-06, "loss": 0.0, "step": 3555 }, { "epoch": 2.28, "learning_rate": 2.832684257536492e-06, "loss": 0.0, "step": 3556 }, { "epoch": 2.29, "learning_rate": 2.8278466045106058e-06, "loss": 0.0, "step": 3557 }, { "epoch": 2.29, "learning_rate": 2.823012405532244e-06, "loss": 0.0, "step": 3558 }, { "epoch": 2.29, "learning_rate": 2.8181816629295146e-06, "loss": 0.0, "step": 3559 }, { "epoch": 2.29, "learning_rate": 2.81335437902886e-06, "loss": 0.0, "step": 3560 }, { "epoch": 2.29, "learning_rate": 2.8085305561550593e-06, "loss": 0.0, "step": 3561 }, { "epoch": 2.29, "learning_rate": 2.803710196631221e-06, "loss": 0.0, "step": 3562 }, { "epoch": 2.29, "learning_rate": 2.7988933027787977e-06, "loss": 0.0, "step": 3563 }, { "epoch": 2.29, "learning_rate": 2.7940798769175526e-06, "loss": 0.0, "step": 3564 }, { "epoch": 2.29, "learning_rate": 2.789269921365596e-06, "loss": 0.0, "step": 3565 }, { "epoch": 2.29, "learning_rate": 2.7844634384393678e-06, "loss": 0.0, "step": 3566 }, { "epoch": 2.29, "learning_rate": 2.779660430453616e-06, "loss": 0.0, "step": 3567 }, { "epoch": 2.29, "learning_rate": 2.7748608997214366e-06, "loss": 0.0, "step": 3568 }, { "epoch": 2.29, "learning_rate": 2.7700648485542415e-06, "loss": 0.0, "step": 3569 }, { "epoch": 2.29, "learning_rate": 2.7652722792617657e-06, "loss": 0.0, "step": 3570 }, { "epoch": 2.29, "learning_rate": 2.760483194152068e-06, "loss": 0.0, "step": 3571 }, { "epoch": 2.3, "learning_rate": 2.7556975955315356e-06, "loss": 0.0, "step": 3572 }, { "epoch": 2.3, "learning_rate": 2.7509154857048703e-06, "loss": 0.0, "step": 3573 }, { "epoch": 2.3, "learning_rate": 2.746136866975092e-06, "loss": 0.0, "step": 3574 }, { "epoch": 2.3, "learning_rate": 2.7413617416435457e-06, "loss": 0.0, "step": 3575 }, { "epoch": 2.3, "learning_rate": 2.736590112009886e-06, "loss": 0.0, "step": 3576 }, { "epoch": 2.3, "learning_rate": 2.7318219803720956e-06, "loss": 0.0, "step": 3577 }, { "epoch": 2.3, "learning_rate": 2.7270573490264616e-06, "loss": 0.0, "step": 3578 }, { "epoch": 2.3, "learning_rate": 2.7222962202675883e-06, "loss": 0.0, "step": 3579 }, { "epoch": 2.3, "learning_rate": 2.7175385963884028e-06, "loss": 0.0, "step": 3580 }, { "epoch": 2.3, "learning_rate": 2.712784479680123e-06, "loss": 0.0, "step": 3581 }, { "epoch": 2.3, "learning_rate": 2.708033872432301e-06, "loss": 0.0, "step": 3582 }, { "epoch": 2.3, "learning_rate": 2.7032867769327844e-06, "loss": 0.0, "step": 3583 }, { "epoch": 2.3, "learning_rate": 2.698543195467731e-06, "loss": 0.0001, "step": 3584 }, { "epoch": 2.3, "learning_rate": 2.693803130321615e-06, "loss": 0.0, "step": 3585 }, { "epoch": 2.3, "learning_rate": 2.689066583777209e-06, "loss": 0.0, "step": 3586 }, { "epoch": 2.3, "learning_rate": 2.6843335581155915e-06, "loss": 0.0, "step": 3587 }, { "epoch": 2.31, "learning_rate": 2.6796040556161484e-06, "loss": 0.0, "step": 3588 }, { "epoch": 2.31, "learning_rate": 2.6748780785565632e-06, "loss": 0.0, "step": 3589 }, { "epoch": 2.31, "learning_rate": 2.670155629212833e-06, "loss": 0.0, "step": 3590 }, { "epoch": 2.31, "learning_rate": 2.665436709859246e-06, "loss": 0.0, "step": 3591 }, { "epoch": 2.31, "learning_rate": 2.6607213227683926e-06, "loss": 0.0, "step": 3592 }, { "epoch": 2.31, "learning_rate": 2.6560094702111593e-06, "loss": 0.0, "step": 3593 }, { "epoch": 2.31, "learning_rate": 2.6513011544567414e-06, "loss": 0.0, "step": 3594 }, { "epoch": 2.31, "learning_rate": 2.646596377772619e-06, "loss": 0.0, "step": 3595 }, { "epoch": 2.31, "learning_rate": 2.6418951424245698e-06, "loss": 0.0, "step": 3596 }, { "epoch": 2.31, "learning_rate": 2.6371974506766775e-06, "loss": 0.0, "step": 3597 }, { "epoch": 2.31, "learning_rate": 2.6325033047912972e-06, "loss": 0.0, "step": 3598 }, { "epoch": 2.31, "learning_rate": 2.6278127070290995e-06, "loss": 0.0, "step": 3599 }, { "epoch": 2.31, "learning_rate": 2.6231256596490317e-06, "loss": 0.0, "step": 3600 }, { "epoch": 2.31, "learning_rate": 2.6184421649083334e-06, "loss": 0.0, "step": 3601 }, { "epoch": 2.31, "learning_rate": 2.613762225062546e-06, "loss": 0.0, "step": 3602 }, { "epoch": 2.32, "learning_rate": 2.609085842365473e-06, "loss": 0.0, "step": 3603 }, { "epoch": 2.32, "learning_rate": 2.6044130190692317e-06, "loss": 0.0, "step": 3604 }, { "epoch": 2.32, "learning_rate": 2.5997437574242103e-06, "loss": 0.0, "step": 3605 }, { "epoch": 2.32, "learning_rate": 2.5950780596790816e-06, "loss": 0.0, "step": 3606 }, { "epoch": 2.32, "learning_rate": 2.5904159280808116e-06, "loss": 0.0, "step": 3607 }, { "epoch": 2.32, "learning_rate": 2.5857573648746405e-06, "loss": 0.0, "step": 3608 }, { "epoch": 2.32, "learning_rate": 2.5811023723040906e-06, "loss": 0.0, "step": 3609 }, { "epoch": 2.32, "learning_rate": 2.576450952610966e-06, "loss": 0.0, "step": 3610 }, { "epoch": 2.32, "learning_rate": 2.5718031080353547e-06, "loss": 0.0, "step": 3611 }, { "epoch": 2.32, "learning_rate": 2.567158840815617e-06, "loss": 0.0, "step": 3612 }, { "epoch": 2.32, "learning_rate": 2.5625181531883904e-06, "loss": 0.0, "step": 3613 }, { "epoch": 2.32, "learning_rate": 2.5578810473885917e-06, "loss": 0.0, "step": 3614 }, { "epoch": 2.32, "learning_rate": 2.5532475256494073e-06, "loss": 0.0, "step": 3615 }, { "epoch": 2.32, "learning_rate": 2.548617590202308e-06, "loss": 0.0, "step": 3616 }, { "epoch": 2.32, "learning_rate": 2.5439912432770287e-06, "loss": 0.0, "step": 3617 }, { "epoch": 2.32, "learning_rate": 2.5393684871015744e-06, "loss": 0.0, "step": 3618 }, { "epoch": 2.33, "learning_rate": 2.5347493239022336e-06, "loss": 0.0, "step": 3619 }, { "epoch": 2.33, "learning_rate": 2.5301337559035443e-06, "loss": 0.0, "step": 3620 }, { "epoch": 2.33, "learning_rate": 2.525521785328333e-06, "loss": 0.0, "step": 3621 }, { "epoch": 2.33, "learning_rate": 2.5209134143976833e-06, "loss": 0.0, "step": 3622 }, { "epoch": 2.33, "learning_rate": 2.516308645330945e-06, "loss": 0.0, "step": 3623 }, { "epoch": 2.33, "learning_rate": 2.5117074803457343e-06, "loss": 0.0, "step": 3624 }, { "epoch": 2.33, "learning_rate": 2.5071099216579375e-06, "loss": 0.0, "step": 3625 }, { "epoch": 2.33, "learning_rate": 2.5025159714816972e-06, "loss": 0.0, "step": 3626 }, { "epoch": 2.33, "learning_rate": 2.497925632029421e-06, "loss": 0.0, "step": 3627 }, { "epoch": 2.33, "learning_rate": 2.4933389055117742e-06, "loss": 0.0, "step": 3628 }, { "epoch": 2.33, "learning_rate": 2.488755794137685e-06, "loss": 0.0, "step": 3629 }, { "epoch": 2.33, "learning_rate": 2.4841763001143447e-06, "loss": 0.0001, "step": 3630 }, { "epoch": 2.33, "learning_rate": 2.479600425647196e-06, "loss": 0.0, "step": 3631 }, { "epoch": 2.33, "learning_rate": 2.475028172939938e-06, "loss": 0.0, "step": 3632 }, { "epoch": 2.33, "learning_rate": 2.4704595441945377e-06, "loss": 0.0, "step": 3633 }, { "epoch": 2.34, "learning_rate": 2.4658945416111935e-06, "loss": 0.0, "step": 3634 }, { "epoch": 2.34, "learning_rate": 2.461333167388382e-06, "loss": 0.0, "step": 3635 }, { "epoch": 2.34, "learning_rate": 2.4567754237228205e-06, "loss": 0.0, "step": 3636 }, { "epoch": 2.34, "learning_rate": 2.4522213128094728e-06, "loss": 0.0, "step": 3637 }, { "epoch": 2.34, "learning_rate": 2.4476708368415704e-06, "loss": 0.0, "step": 3638 }, { "epoch": 2.34, "learning_rate": 2.4431239980105714e-06, "loss": 0.0, "step": 3639 }, { "epoch": 2.34, "learning_rate": 2.438580798506203e-06, "loss": 0.0, "step": 3640 }, { "epoch": 2.34, "learning_rate": 2.4340412405164292e-06, "loss": 0.0, "step": 3641 }, { "epoch": 2.34, "learning_rate": 2.429505326227457e-06, "loss": 0.0, "step": 3642 }, { "epoch": 2.34, "learning_rate": 2.4249730578237527e-06, "loss": 0.0, "step": 3643 }, { "epoch": 2.34, "learning_rate": 2.420444437488012e-06, "loss": 0.0, "step": 3644 }, { "epoch": 2.34, "learning_rate": 2.415919467401182e-06, "loss": 0.0, "step": 3645 }, { "epoch": 2.34, "learning_rate": 2.411398149742444e-06, "loss": 0.0, "step": 3646 }, { "epoch": 2.34, "learning_rate": 2.406880486689235e-06, "loss": 0.0, "step": 3647 }, { "epoch": 2.34, "learning_rate": 2.4023664804172163e-06, "loss": 0.0001, "step": 3648 }, { "epoch": 2.34, "learning_rate": 2.3978561331002946e-06, "loss": 0.0, "step": 3649 }, { "epoch": 2.35, "learning_rate": 2.393349446910621e-06, "loss": 0.0, "step": 3650 }, { "epoch": 2.35, "learning_rate": 2.388846424018566e-06, "loss": 0.0, "step": 3651 }, { "epoch": 2.35, "learning_rate": 2.384347066592756e-06, "loss": 0.0, "step": 3652 }, { "epoch": 2.35, "learning_rate": 2.3798513768000396e-06, "loss": 0.0, "step": 3653 }, { "epoch": 2.35, "learning_rate": 2.375359356805499e-06, "loss": 0.0, "step": 3654 }, { "epoch": 2.35, "learning_rate": 2.3708710087724618e-06, "loss": 0.0, "step": 3655 }, { "epoch": 2.35, "learning_rate": 2.3663863348624673e-06, "loss": 0.0, "step": 3656 }, { "epoch": 2.35, "learning_rate": 2.3619053372353028e-06, "loss": 0.0, "step": 3657 }, { "epoch": 2.35, "learning_rate": 2.357428018048976e-06, "loss": 0.0, "step": 3658 }, { "epoch": 2.35, "learning_rate": 2.3529543794597233e-06, "loss": 0.0, "step": 3659 }, { "epoch": 2.35, "learning_rate": 2.348484423622015e-06, "loss": 0.0, "step": 3660 }, { "epoch": 2.35, "learning_rate": 2.3440181526885418e-06, "loss": 0.0, "step": 3661 }, { "epoch": 2.35, "learning_rate": 2.339555568810221e-06, "loss": 0.0, "step": 3662 }, { "epoch": 2.35, "learning_rate": 2.3350966741361926e-06, "loss": 0.0, "step": 3663 }, { "epoch": 2.35, "learning_rate": 2.330641470813826e-06, "loss": 0.0, "step": 3664 }, { "epoch": 2.36, "learning_rate": 2.326189960988707e-06, "loss": 0.0, "step": 3665 }, { "epoch": 2.36, "learning_rate": 2.3217421468046452e-06, "loss": 0.0, "step": 3666 }, { "epoch": 2.36, "learning_rate": 2.3172980304036685e-06, "loss": 0.0, "step": 3667 }, { "epoch": 2.36, "learning_rate": 2.312857613926024e-06, "loss": 0.0001, "step": 3668 }, { "epoch": 2.36, "learning_rate": 2.308420899510182e-06, "loss": 0.0, "step": 3669 }, { "epoch": 2.36, "learning_rate": 2.3039878892928247e-06, "loss": 0.0, "step": 3670 }, { "epoch": 2.36, "learning_rate": 2.2995585854088475e-06, "loss": 0.0, "step": 3671 }, { "epoch": 2.36, "learning_rate": 2.2951329899913754e-06, "loss": 0.0, "step": 3672 }, { "epoch": 2.36, "learning_rate": 2.2907111051717235e-06, "loss": 0.0, "step": 3673 }, { "epoch": 2.36, "learning_rate": 2.2862929330794446e-06, "loss": 0.0, "step": 3674 }, { "epoch": 2.36, "learning_rate": 2.2818784758422884e-06, "loss": 0.0, "step": 3675 }, { "epoch": 2.36, "learning_rate": 2.2774677355862194e-06, "loss": 0.0, "step": 3676 }, { "epoch": 2.36, "learning_rate": 2.27306071443541e-06, "loss": 0.0, "step": 3677 }, { "epoch": 2.36, "learning_rate": 2.26865741451225e-06, "loss": 0.0, "step": 3678 }, { "epoch": 2.36, "learning_rate": 2.2642578379373257e-06, "loss": 0.0, "step": 3679 }, { "epoch": 2.36, "learning_rate": 2.259861986829437e-06, "loss": 0.0, "step": 3680 }, { "epoch": 2.37, "learning_rate": 2.255469863305588e-06, "loss": 0.0, "step": 3681 }, { "epoch": 2.37, "learning_rate": 2.251081469480986e-06, "loss": 0.0, "step": 3682 }, { "epoch": 2.37, "learning_rate": 2.2466968074690474e-06, "loss": 0.0, "step": 3683 }, { "epoch": 2.37, "learning_rate": 2.242315879381387e-06, "loss": 0.0002, "step": 3684 }, { "epoch": 2.37, "learning_rate": 2.2379386873278175e-06, "loss": 0.0, "step": 3685 }, { "epoch": 2.37, "learning_rate": 2.233565233416366e-06, "loss": 0.0, "step": 3686 }, { "epoch": 2.37, "learning_rate": 2.229195519753239e-06, "loss": 0.0, "step": 3687 }, { "epoch": 2.37, "learning_rate": 2.2248295484428618e-06, "loss": 0.0, "step": 3688 }, { "epoch": 2.37, "learning_rate": 2.220467321587847e-06, "loss": 0.0, "step": 3689 }, { "epoch": 2.37, "learning_rate": 2.216108841289001e-06, "loss": 0.0, "step": 3690 }, { "epoch": 2.37, "learning_rate": 2.21175410964534e-06, "loss": 0.0, "step": 3691 }, { "epoch": 2.37, "learning_rate": 2.207403128754053e-06, "loss": 0.0, "step": 3692 }, { "epoch": 2.37, "learning_rate": 2.2030559007105457e-06, "loss": 0.0, "step": 3693 }, { "epoch": 2.37, "learning_rate": 2.1987124276084006e-06, "loss": 0.0, "step": 3694 }, { "epoch": 2.37, "learning_rate": 2.194372711539395e-06, "loss": 0.0, "step": 3695 }, { "epoch": 2.37, "learning_rate": 2.1900367545935066e-06, "loss": 0.0, "step": 3696 }, { "epoch": 2.38, "learning_rate": 2.1857045588588888e-06, "loss": 0.0, "step": 3697 }, { "epoch": 2.38, "learning_rate": 2.1813761264218923e-06, "loss": 0.0, "step": 3698 }, { "epoch": 2.38, "learning_rate": 2.17705145936705e-06, "loss": 0.0, "step": 3699 }, { "epoch": 2.38, "learning_rate": 2.1727305597770887e-06, "loss": 0.0, "step": 3700 }, { "epoch": 2.38, "learning_rate": 2.168413429732914e-06, "loss": 0.0, "step": 3701 }, { "epoch": 2.38, "learning_rate": 2.1641000713136174e-06, "loss": 0.0, "step": 3702 }, { "epoch": 2.38, "learning_rate": 2.1597904865964824e-06, "loss": 0.0, "step": 3703 }, { "epoch": 2.38, "learning_rate": 2.1554846776569572e-06, "loss": 0.0, "step": 3704 }, { "epoch": 2.38, "learning_rate": 2.1511826465686913e-06, "loss": 0.0, "step": 3705 }, { "epoch": 2.38, "learning_rate": 2.1468843954035023e-06, "loss": 0.0, "step": 3706 }, { "epoch": 2.38, "learning_rate": 2.1425899262313875e-06, "loss": 0.0, "step": 3707 }, { "epoch": 2.38, "learning_rate": 2.138299241120537e-06, "loss": 0.0, "step": 3708 }, { "epoch": 2.38, "learning_rate": 2.134012342137296e-06, "loss": 0.0002, "step": 3709 }, { "epoch": 2.38, "learning_rate": 2.129729231346206e-06, "loss": 0.0, "step": 3710 }, { "epoch": 2.38, "learning_rate": 2.1254499108099736e-06, "loss": 0.0, "step": 3711 }, { "epoch": 2.39, "learning_rate": 2.1211743825894805e-06, "loss": 0.0, "step": 3712 }, { "epoch": 2.39, "learning_rate": 2.1169026487437905e-06, "loss": 0.0, "step": 3713 }, { "epoch": 2.39, "learning_rate": 2.112634711330129e-06, "loss": 0.0, "step": 3714 }, { "epoch": 2.39, "learning_rate": 2.1083705724039004e-06, "loss": 0.0001, "step": 3715 }, { "epoch": 2.39, "learning_rate": 2.1041102340186736e-06, "loss": 0.0, "step": 3716 }, { "epoch": 2.39, "learning_rate": 2.0998536982261976e-06, "loss": 0.0, "step": 3717 }, { "epoch": 2.39, "learning_rate": 2.0956009670763788e-06, "loss": 0.0, "step": 3718 }, { "epoch": 2.39, "learning_rate": 2.091352042617297e-06, "loss": 0.0, "step": 3719 }, { "epoch": 2.39, "learning_rate": 2.087106926895198e-06, "loss": 0.0, "step": 3720 }, { "epoch": 2.39, "learning_rate": 2.0828656219544906e-06, "loss": 0.0, "step": 3721 }, { "epoch": 2.39, "learning_rate": 2.0786281298377564e-06, "loss": 0.0, "step": 3722 }, { "epoch": 2.39, "learning_rate": 2.0743944525857307e-06, "loss": 0.0, "step": 3723 }, { "epoch": 2.39, "learning_rate": 2.0701645922373157e-06, "loss": 0.0, "step": 3724 }, { "epoch": 2.39, "learning_rate": 2.0659385508295827e-06, "loss": 0.0, "step": 3725 }, { "epoch": 2.39, "learning_rate": 2.061716330397746e-06, "loss": 0.0, "step": 3726 }, { "epoch": 2.39, "learning_rate": 2.0574979329752e-06, "loss": 0.0, "step": 3727 }, { "epoch": 2.4, "learning_rate": 2.0532833605934856e-06, "loss": 0.0, "step": 3728 }, { "epoch": 2.4, "learning_rate": 2.0490726152823003e-06, "loss": 0.0, "step": 3729 }, { "epoch": 2.4, "learning_rate": 2.0448656990695138e-06, "loss": 0.0, "step": 3730 }, { "epoch": 2.4, "learning_rate": 2.040662613981128e-06, "loss": 0.0, "step": 3731 }, { "epoch": 2.4, "learning_rate": 2.036463362041321e-06, "loss": 0.0, "step": 3732 }, { "epoch": 2.4, "learning_rate": 2.0322679452724148e-06, "loss": 0.0, "step": 3733 }, { "epoch": 2.4, "learning_rate": 2.0280763656948853e-06, "loss": 0.0, "step": 3734 }, { "epoch": 2.4, "learning_rate": 2.0238886253273583e-06, "loss": 0.0, "step": 3735 }, { "epoch": 2.4, "learning_rate": 2.01970472618662e-06, "loss": 0.0, "step": 3736 }, { "epoch": 2.4, "learning_rate": 2.0155246702875964e-06, "loss": 0.0, "step": 3737 }, { "epoch": 2.4, "learning_rate": 2.0113484596433654e-06, "loss": 0.0, "step": 3738 }, { "epoch": 2.4, "learning_rate": 2.00717609626516e-06, "loss": 0.0, "step": 3739 }, { "epoch": 2.4, "learning_rate": 2.003007582162346e-06, "loss": 0.0, "step": 3740 }, { "epoch": 2.4, "learning_rate": 1.998842919342451e-06, "loss": 0.0, "step": 3741 }, { "epoch": 2.4, "learning_rate": 1.994682109811139e-06, "loss": 0.0, "step": 3742 }, { "epoch": 2.41, "learning_rate": 1.990525155572216e-06, "loss": 0.0, "step": 3743 }, { "epoch": 2.41, "learning_rate": 1.9863720586276457e-06, "loss": 0.0, "step": 3744 }, { "epoch": 2.41, "learning_rate": 1.9822228209775108e-06, "loss": 0.0, "step": 3745 }, { "epoch": 2.41, "learning_rate": 1.978077444620059e-06, "loss": 0.0, "step": 3746 }, { "epoch": 2.41, "learning_rate": 1.973935931551664e-06, "loss": 0.0, "step": 3747 }, { "epoch": 2.41, "learning_rate": 1.9697982837668397e-06, "loss": 0.0, "step": 3748 }, { "epoch": 2.41, "learning_rate": 1.965664503258248e-06, "loss": 0.0, "step": 3749 }, { "epoch": 2.41, "learning_rate": 1.9615345920166784e-06, "loss": 0.0, "step": 3750 }, { "epoch": 2.41, "learning_rate": 1.9574085520310626e-06, "loss": 0.0, "step": 3751 }, { "epoch": 2.41, "learning_rate": 1.953286385288461e-06, "loss": 0.0, "step": 3752 }, { "epoch": 2.41, "learning_rate": 1.949168093774082e-06, "loss": 0.0, "step": 3753 }, { "epoch": 2.41, "learning_rate": 1.9450536794712536e-06, "loss": 0.0, "step": 3754 }, { "epoch": 2.41, "learning_rate": 1.9409431443614422e-06, "loss": 0.0, "step": 3755 }, { "epoch": 2.41, "learning_rate": 1.9368364904242532e-06, "loss": 0.0, "step": 3756 }, { "epoch": 2.41, "learning_rate": 1.932733719637405e-06, "loss": 0.0, "step": 3757 }, { "epoch": 2.41, "learning_rate": 1.928634833976766e-06, "loss": 0.0, "step": 3758 }, { "epoch": 2.42, "learning_rate": 1.9245398354163202e-06, "loss": 0.0, "step": 3759 }, { "epoch": 2.42, "learning_rate": 1.9204487259281824e-06, "loss": 0.0, "step": 3760 }, { "epoch": 2.42, "learning_rate": 1.9163615074826023e-06, "loss": 0.0, "step": 3761 }, { "epoch": 2.42, "learning_rate": 1.9122781820479397e-06, "loss": 0.0, "step": 3762 }, { "epoch": 2.42, "learning_rate": 1.9081987515906975e-06, "loss": 0.0, "step": 3763 }, { "epoch": 2.42, "learning_rate": 1.9041232180754898e-06, "loss": 0.0, "step": 3764 }, { "epoch": 2.42, "learning_rate": 1.900051583465058e-06, "loss": 0.0, "step": 3765 }, { "epoch": 2.42, "learning_rate": 1.89598384972027e-06, "loss": 0.0, "step": 3766 }, { "epoch": 2.42, "learning_rate": 1.8919200188001107e-06, "loss": 0.0001, "step": 3767 }, { "epoch": 2.42, "learning_rate": 1.8878600926616842e-06, "loss": 0.0, "step": 3768 }, { "epoch": 2.42, "learning_rate": 1.8838040732602148e-06, "loss": 0.0, "step": 3769 }, { "epoch": 2.42, "learning_rate": 1.8797519625490512e-06, "loss": 0.0, "step": 3770 }, { "epoch": 2.42, "learning_rate": 1.8757037624796538e-06, "loss": 0.0, "step": 3771 }, { "epoch": 2.42, "learning_rate": 1.8716594750015993e-06, "loss": 0.0, "step": 3772 }, { "epoch": 2.42, "learning_rate": 1.8676191020625821e-06, "loss": 0.0001, "step": 3773 }, { "epoch": 2.43, "learning_rate": 1.8635826456084094e-06, "loss": 0.0, "step": 3774 }, { "epoch": 2.43, "learning_rate": 1.859550107583009e-06, "loss": 0.0, "step": 3775 }, { "epoch": 2.43, "learning_rate": 1.8555214899284135e-06, "loss": 0.0, "step": 3776 }, { "epoch": 2.43, "learning_rate": 1.8514967945847685e-06, "loss": 0.0, "step": 3777 }, { "epoch": 2.43, "learning_rate": 1.84747602349034e-06, "loss": 0.0, "step": 3778 }, { "epoch": 2.43, "learning_rate": 1.843459178581487e-06, "loss": 0.0, "step": 3779 }, { "epoch": 2.43, "learning_rate": 1.8394462617926944e-06, "loss": 0.0, "step": 3780 }, { "epoch": 2.43, "learning_rate": 1.8354372750565475e-06, "loss": 0.0, "step": 3781 }, { "epoch": 2.43, "learning_rate": 1.831432220303736e-06, "loss": 0.0, "step": 3782 }, { "epoch": 2.43, "learning_rate": 1.8274310994630673e-06, "loss": 0.0, "step": 3783 }, { "epoch": 2.43, "learning_rate": 1.8234339144614387e-06, "loss": 0.0, "step": 3784 }, { "epoch": 2.43, "learning_rate": 1.8194406672238652e-06, "loss": 0.0, "step": 3785 }, { "epoch": 2.43, "learning_rate": 1.8154513596734603e-06, "loss": 0.0, "step": 3786 }, { "epoch": 2.43, "learning_rate": 1.8114659937314383e-06, "loss": 0.0, "step": 3787 }, { "epoch": 2.43, "learning_rate": 1.8074845713171153e-06, "loss": 0.0, "step": 3788 }, { "epoch": 2.43, "learning_rate": 1.8035070943479161e-06, "loss": 0.0, "step": 3789 }, { "epoch": 2.44, "learning_rate": 1.7995335647393552e-06, "loss": 0.0, "step": 3790 }, { "epoch": 2.44, "learning_rate": 1.79556398440505e-06, "loss": 0.0, "step": 3791 }, { "epoch": 2.44, "learning_rate": 1.791598355256723e-06, "loss": 0.0, "step": 3792 }, { "epoch": 2.44, "learning_rate": 1.7876366792041767e-06, "loss": 0.0, "step": 3793 }, { "epoch": 2.44, "learning_rate": 1.783678958155327e-06, "loss": 0.0, "step": 3794 }, { "epoch": 2.44, "learning_rate": 1.7797251940161775e-06, "loss": 0.0, "step": 3795 }, { "epoch": 2.44, "learning_rate": 1.7757753886908236e-06, "loss": 0.0, "step": 3796 }, { "epoch": 2.44, "learning_rate": 1.771829544081467e-06, "loss": 0.0, "step": 3797 }, { "epoch": 2.44, "learning_rate": 1.76788766208838e-06, "loss": 0.0, "step": 3798 }, { "epoch": 2.44, "learning_rate": 1.7639497446099474e-06, "loss": 0.0, "step": 3799 }, { "epoch": 2.44, "learning_rate": 1.7600157935426344e-06, "loss": 0.0, "step": 3800 }, { "epoch": 2.44, "learning_rate": 1.7560858107809963e-06, "loss": 0.0, "step": 3801 }, { "epoch": 2.44, "learning_rate": 1.752159798217683e-06, "loss": 0.0, "step": 3802 }, { "epoch": 2.44, "learning_rate": 1.748237757743425e-06, "loss": 0.0, "step": 3803 }, { "epoch": 2.44, "learning_rate": 1.7443196912470462e-06, "loss": 0.0, "step": 3804 }, { "epoch": 2.44, "learning_rate": 1.7404056006154491e-06, "loss": 0.0, "step": 3805 }, { "epoch": 2.45, "learning_rate": 1.7364954877336316e-06, "loss": 0.0, "step": 3806 }, { "epoch": 2.45, "learning_rate": 1.7325893544846683e-06, "loss": 0.0, "step": 3807 }, { "epoch": 2.45, "learning_rate": 1.7286872027497159e-06, "loss": 0.0, "step": 3808 }, { "epoch": 2.45, "learning_rate": 1.7247890344080264e-06, "loss": 0.0, "step": 3809 }, { "epoch": 2.45, "learning_rate": 1.7208948513369128e-06, "loss": 0.0, "step": 3810 }, { "epoch": 2.45, "learning_rate": 1.7170046554117892e-06, "loss": 0.0, "step": 3811 }, { "epoch": 2.45, "learning_rate": 1.7131184485061369e-06, "loss": 0.0, "step": 3812 }, { "epoch": 2.45, "learning_rate": 1.709236232491517e-06, "loss": 0.0, "step": 3813 }, { "epoch": 2.45, "learning_rate": 1.7053580092375799e-06, "loss": 0.0, "step": 3814 }, { "epoch": 2.45, "learning_rate": 1.7014837806120344e-06, "loss": 0.0, "step": 3815 }, { "epoch": 2.45, "learning_rate": 1.6976135484806822e-06, "loss": 0.0, "step": 3816 }, { "epoch": 2.45, "learning_rate": 1.6937473147073935e-06, "loss": 0.0, "step": 3817 }, { "epoch": 2.45, "learning_rate": 1.689885081154109e-06, "loss": 0.0, "step": 3818 }, { "epoch": 2.45, "learning_rate": 1.686026849680853e-06, "loss": 0.0, "step": 3819 }, { "epoch": 2.45, "learning_rate": 1.6821726221457136e-06, "loss": 0.0, "step": 3820 }, { "epoch": 2.46, "learning_rate": 1.6783224004048538e-06, "loss": 0.0, "step": 3821 }, { "epoch": 2.46, "learning_rate": 1.6744761863125082e-06, "loss": 0.0, "step": 3822 }, { "epoch": 2.46, "learning_rate": 1.6706339817209771e-06, "loss": 0.0, "step": 3823 }, { "epoch": 2.46, "learning_rate": 1.6667957884806385e-06, "loss": 0.0, "step": 3824 }, { "epoch": 2.46, "learning_rate": 1.6629616084399314e-06, "loss": 0.0, "step": 3825 }, { "epoch": 2.46, "learning_rate": 1.659131443445363e-06, "loss": 0.0, "step": 3826 }, { "epoch": 2.46, "learning_rate": 1.655305295341506e-06, "loss": 0.0, "step": 3827 }, { "epoch": 2.46, "learning_rate": 1.6514831659710051e-06, "loss": 0.0, "step": 3828 }, { "epoch": 2.46, "learning_rate": 1.6476650571745634e-06, "loss": 0.0, "step": 3829 }, { "epoch": 2.46, "learning_rate": 1.6438509707909457e-06, "loss": 0.0, "step": 3830 }, { "epoch": 2.46, "learning_rate": 1.6400409086569913e-06, "loss": 0.0, "step": 3831 }, { "epoch": 2.46, "learning_rate": 1.636234872607584e-06, "loss": 0.0, "step": 3832 }, { "epoch": 2.46, "learning_rate": 1.6324328644756848e-06, "loss": 0.0, "step": 3833 }, { "epoch": 2.46, "learning_rate": 1.628634886092305e-06, "loss": 0.0, "step": 3834 }, { "epoch": 2.46, "learning_rate": 1.6248409392865183e-06, "loss": 0.0, "step": 3835 }, { "epoch": 2.46, "learning_rate": 1.6210510258854617e-06, "loss": 0.0, "step": 3836 }, { "epoch": 2.47, "learning_rate": 1.6172651477143165e-06, "loss": 0.0, "step": 3837 }, { "epoch": 2.47, "learning_rate": 1.6134833065963363e-06, "loss": 0.0, "step": 3838 }, { "epoch": 2.47, "learning_rate": 1.6097055043528209e-06, "loss": 0.0, "step": 3839 }, { "epoch": 2.47, "learning_rate": 1.6059317428031273e-06, "loss": 0.0, "step": 3840 }, { "epoch": 2.47, "learning_rate": 1.602162023764664e-06, "loss": 0.0, "step": 3841 }, { "epoch": 2.47, "learning_rate": 1.5983963490529008e-06, "loss": 0.0, "step": 3842 }, { "epoch": 2.47, "learning_rate": 1.5946347204813518e-06, "loss": 0.0, "step": 3843 }, { "epoch": 2.47, "learning_rate": 1.5908771398615819e-06, "loss": 0.0, "step": 3844 }, { "epoch": 2.47, "learning_rate": 1.5871236090032183e-06, "loss": 0.0, "step": 3845 }, { "epoch": 2.47, "learning_rate": 1.5833741297139181e-06, "loss": 0.0, "step": 3846 }, { "epoch": 2.47, "learning_rate": 1.5796287037994063e-06, "loss": 0.0, "step": 3847 }, { "epoch": 2.47, "learning_rate": 1.575887333063445e-06, "loss": 0.0, "step": 3848 }, { "epoch": 2.47, "learning_rate": 1.5721500193078455e-06, "loss": 0.0, "step": 3849 }, { "epoch": 2.47, "learning_rate": 1.568416764332471e-06, "loss": 0.0, "step": 3850 }, { "epoch": 2.47, "learning_rate": 1.564687569935216e-06, "loss": 0.0, "step": 3851 }, { "epoch": 2.48, "learning_rate": 1.560962437912037e-06, "loss": 0.0, "step": 3852 }, { "epoch": 2.48, "learning_rate": 1.5572413700569221e-06, "loss": 0.0, "step": 3853 }, { "epoch": 2.48, "learning_rate": 1.5535243681619029e-06, "loss": 0.0, "step": 3854 }, { "epoch": 2.48, "learning_rate": 1.5498114340170621e-06, "loss": 0.0, "step": 3855 }, { "epoch": 2.48, "learning_rate": 1.5461025694105124e-06, "loss": 0.0, "step": 3856 }, { "epoch": 2.48, "learning_rate": 1.5423977761284126e-06, "loss": 0.0, "step": 3857 }, { "epoch": 2.48, "learning_rate": 1.5386970559549564e-06, "loss": 0.0, "step": 3858 }, { "epoch": 2.48, "learning_rate": 1.5350004106723838e-06, "loss": 0.0, "step": 3859 }, { "epoch": 2.48, "learning_rate": 1.5313078420609651e-06, "loss": 0.0, "step": 3860 }, { "epoch": 2.48, "learning_rate": 1.527619351899008e-06, "loss": 0.0, "step": 3861 }, { "epoch": 2.48, "learning_rate": 1.5239349419628636e-06, "loss": 0.0, "step": 3862 }, { "epoch": 2.48, "learning_rate": 1.5202546140269026e-06, "loss": 0.0, "step": 3863 }, { "epoch": 2.48, "learning_rate": 1.5165783698635471e-06, "loss": 0.0, "step": 3864 }, { "epoch": 2.48, "learning_rate": 1.5129062112432424e-06, "loss": 0.0001, "step": 3865 }, { "epoch": 2.48, "learning_rate": 1.5092381399344659e-06, "loss": 0.0, "step": 3866 }, { "epoch": 2.48, "learning_rate": 1.5055741577037364e-06, "loss": 0.0, "step": 3867 }, { "epoch": 2.49, "learning_rate": 1.5019142663155872e-06, "loss": 0.0, "step": 3868 }, { "epoch": 2.49, "learning_rate": 1.4982584675325962e-06, "loss": 0.0, "step": 3869 }, { "epoch": 2.49, "learning_rate": 1.494606763115365e-06, "loss": 0.0, "step": 3870 }, { "epoch": 2.49, "learning_rate": 1.4909591548225188e-06, "loss": 0.0, "step": 3871 }, { "epoch": 2.49, "learning_rate": 1.4873156444107194e-06, "loss": 0.0, "step": 3872 }, { "epoch": 2.49, "learning_rate": 1.483676233634649e-06, "loss": 0.0, "step": 3873 }, { "epoch": 2.49, "learning_rate": 1.4800409242470147e-06, "loss": 0.0, "step": 3874 }, { "epoch": 2.49, "learning_rate": 1.4764097179985514e-06, "loss": 0.0, "step": 3875 }, { "epoch": 2.49, "learning_rate": 1.4727826166380144e-06, "loss": 0.0, "step": 3876 }, { "epoch": 2.49, "learning_rate": 1.4691596219121884e-06, "loss": 0.0, "step": 3877 }, { "epoch": 2.49, "learning_rate": 1.465540735565877e-06, "loss": 0.0, "step": 3878 }, { "epoch": 2.49, "learning_rate": 1.461925959341901e-06, "loss": 0.0, "step": 3879 }, { "epoch": 2.49, "learning_rate": 1.4583152949811063e-06, "loss": 0.0, "step": 3880 }, { "epoch": 2.49, "learning_rate": 1.4547087442223607e-06, "loss": 0.0, "step": 3881 }, { "epoch": 2.49, "learning_rate": 1.4511063088025468e-06, "loss": 0.0, "step": 3882 }, { "epoch": 2.5, "learning_rate": 1.4475079904565636e-06, "loss": 0.0, "step": 3883 }, { "epoch": 2.5, "learning_rate": 1.4439137909173385e-06, "loss": 0.0, "step": 3884 }, { "epoch": 2.5, "learning_rate": 1.4403237119157954e-06, "loss": 0.0, "step": 3885 }, { "epoch": 2.5, "learning_rate": 1.4367377551808947e-06, "loss": 0.0, "step": 3886 }, { "epoch": 2.5, "learning_rate": 1.4331559224395996e-06, "loss": 0.0, "step": 3887 }, { "epoch": 2.5, "learning_rate": 1.429578215416888e-06, "loss": 0.0, "step": 3888 }, { "epoch": 2.5, "learning_rate": 1.4260046358357593e-06, "loss": 0.0, "step": 3889 }, { "epoch": 2.5, "learning_rate": 1.4224351854172092e-06, "loss": 0.0, "step": 3890 }, { "epoch": 2.5, "learning_rate": 1.4188698658802613e-06, "loss": 0.0, "step": 3891 }, { "epoch": 2.5, "learning_rate": 1.4153086789419412e-06, "loss": 0.0, "step": 3892 }, { "epoch": 2.5, "learning_rate": 1.4117516263172838e-06, "loss": 0.0, "step": 3893 }, { "epoch": 2.5, "learning_rate": 1.4081987097193383e-06, "loss": 0.0, "step": 3894 }, { "epoch": 2.5, "learning_rate": 1.4046499308591588e-06, "loss": 0.0, "step": 3895 }, { "epoch": 2.5, "learning_rate": 1.4011052914458056e-06, "loss": 0.0, "step": 3896 }, { "epoch": 2.5, "learning_rate": 1.3975647931863435e-06, "loss": 0.0, "step": 3897 }, { "epoch": 2.5, "learning_rate": 1.3940284377858537e-06, "loss": 0.0, "step": 3898 }, { "epoch": 2.51, "learning_rate": 1.3904962269474054e-06, "loss": 0.0, "step": 3899 }, { "epoch": 2.51, "learning_rate": 1.3869681623720888e-06, "loss": 0.0, "step": 3900 }, { "epoch": 2.51, "learning_rate": 1.3834442457589858e-06, "loss": 0.0, "step": 3901 }, { "epoch": 2.51, "learning_rate": 1.3799244788051825e-06, "loss": 0.0, "step": 3902 }, { "epoch": 2.51, "learning_rate": 1.376408863205776e-06, "loss": 0.0, "step": 3903 }, { "epoch": 2.51, "learning_rate": 1.3728974006538464e-06, "loss": 0.0, "step": 3904 }, { "epoch": 2.51, "learning_rate": 1.3693900928404924e-06, "loss": 0.0, "step": 3905 }, { "epoch": 2.51, "learning_rate": 1.3658869414547992e-06, "loss": 0.0, "step": 3906 }, { "epoch": 2.51, "learning_rate": 1.3623879481838531e-06, "loss": 0.0, "step": 3907 }, { "epoch": 2.51, "learning_rate": 1.3588931147127437e-06, "loss": 0.0, "step": 3908 }, { "epoch": 2.51, "learning_rate": 1.3554024427245504e-06, "loss": 0.0, "step": 3909 }, { "epoch": 2.51, "learning_rate": 1.3519159339003507e-06, "loss": 0.0, "step": 3910 }, { "epoch": 2.51, "learning_rate": 1.3484335899192159e-06, "loss": 0.0, "step": 3911 }, { "epoch": 2.51, "learning_rate": 1.3449554124582176e-06, "loss": 0.0, "step": 3912 }, { "epoch": 2.51, "learning_rate": 1.3414814031924118e-06, "loss": 0.0, "step": 3913 }, { "epoch": 2.52, "learning_rate": 1.3380115637948532e-06, "loss": 0.0, "step": 3914 }, { "epoch": 2.52, "learning_rate": 1.3345458959365877e-06, "loss": 0.0, "step": 3915 }, { "epoch": 2.52, "learning_rate": 1.3310844012866464e-06, "loss": 0.0, "step": 3916 }, { "epoch": 2.52, "learning_rate": 1.3276270815120618e-06, "loss": 0.0, "step": 3917 }, { "epoch": 2.52, "learning_rate": 1.3241739382778474e-06, "loss": 0.0, "step": 3918 }, { "epoch": 2.52, "learning_rate": 1.3207249732470028e-06, "loss": 0.0, "step": 3919 }, { "epoch": 2.52, "learning_rate": 1.3172801880805297e-06, "loss": 0.0, "step": 3920 }, { "epoch": 2.52, "learning_rate": 1.3138395844373964e-06, "loss": 0.0, "step": 3921 }, { "epoch": 2.52, "learning_rate": 1.310403163974574e-06, "loss": 0.0, "step": 3922 }, { "epoch": 2.52, "learning_rate": 1.3069709283470132e-06, "loss": 0.0, "step": 3923 }, { "epoch": 2.52, "learning_rate": 1.3035428792076454e-06, "loss": 0.0, "step": 3924 }, { "epoch": 2.52, "learning_rate": 1.3001190182073941e-06, "loss": 0.0, "step": 3925 }, { "epoch": 2.52, "learning_rate": 1.2966993469951595e-06, "loss": 0.0, "step": 3926 }, { "epoch": 2.52, "learning_rate": 1.2932838672178262e-06, "loss": 0.0, "step": 3927 }, { "epoch": 2.52, "learning_rate": 1.2898725805202584e-06, "loss": 0.0, "step": 3928 }, { "epoch": 2.52, "learning_rate": 1.2864654885453031e-06, "loss": 0.0, "step": 3929 }, { "epoch": 2.53, "learning_rate": 1.283062592933788e-06, "loss": 0.0, "step": 3930 }, { "epoch": 2.53, "learning_rate": 1.2796638953245189e-06, "loss": 0.0, "step": 3931 }, { "epoch": 2.53, "learning_rate": 1.2762693973542773e-06, "loss": 0.0, "step": 3932 }, { "epoch": 2.53, "learning_rate": 1.2728791006578235e-06, "loss": 0.0, "step": 3933 }, { "epoch": 2.53, "learning_rate": 1.2694930068678978e-06, "loss": 0.0, "step": 3934 }, { "epoch": 2.53, "learning_rate": 1.266111117615214e-06, "loss": 0.0, "step": 3935 }, { "epoch": 2.53, "learning_rate": 1.262733434528458e-06, "loss": 0.0001, "step": 3936 }, { "epoch": 2.53, "learning_rate": 1.2593599592342987e-06, "loss": 0.0, "step": 3937 }, { "epoch": 2.53, "learning_rate": 1.255990693357365e-06, "loss": 0.0, "step": 3938 }, { "epoch": 2.53, "learning_rate": 1.2526256385202717e-06, "loss": 0.0, "step": 3939 }, { "epoch": 2.53, "learning_rate": 1.2492647963436e-06, "loss": 0.0, "step": 3940 }, { "epoch": 2.53, "learning_rate": 1.2459081684458995e-06, "loss": 0.0, "step": 3941 }, { "epoch": 2.53, "learning_rate": 1.2425557564437008e-06, "loss": 0.0, "step": 3942 }, { "epoch": 2.53, "learning_rate": 1.2392075619514864e-06, "loss": 0.0, "step": 3943 }, { "epoch": 2.53, "learning_rate": 1.2358635865817248e-06, "loss": 0.0, "step": 3944 }, { "epoch": 2.53, "learning_rate": 1.2325238319448451e-06, "loss": 0.0, "step": 3945 }, { "epoch": 2.54, "learning_rate": 1.229188299649241e-06, "loss": 0.0, "step": 3946 }, { "epoch": 2.54, "learning_rate": 1.225856991301282e-06, "loss": 0.0, "step": 3947 }, { "epoch": 2.54, "learning_rate": 1.2225299085052933e-06, "loss": 0.0, "step": 3948 }, { "epoch": 2.54, "learning_rate": 1.2192070528635713e-06, "loss": 0.0, "step": 3949 }, { "epoch": 2.54, "learning_rate": 1.2158884259763727e-06, "loss": 0.0, "step": 3950 }, { "epoch": 2.54, "learning_rate": 1.2125740294419242e-06, "loss": 0.0, "step": 3951 }, { "epoch": 2.54, "learning_rate": 1.2092638648564037e-06, "loss": 0.0, "step": 3952 }, { "epoch": 2.54, "learning_rate": 1.2059579338139637e-06, "loss": 0.0, "step": 3953 }, { "epoch": 2.54, "learning_rate": 1.2026562379067108e-06, "loss": 0.0, "step": 3954 }, { "epoch": 2.54, "learning_rate": 1.1993587787247107e-06, "loss": 0.0, "step": 3955 }, { "epoch": 2.54, "learning_rate": 1.1960655578559988e-06, "loss": 0.0, "step": 3956 }, { "epoch": 2.54, "learning_rate": 1.1927765768865519e-06, "loss": 0.0, "step": 3957 }, { "epoch": 2.54, "learning_rate": 1.1894918374003227e-06, "loss": 0.0, "step": 3958 }, { "epoch": 2.54, "learning_rate": 1.186211340979212e-06, "loss": 0.0, "step": 3959 }, { "epoch": 2.54, "learning_rate": 1.1829350892030745e-06, "loss": 0.0, "step": 3960 }, { "epoch": 2.55, "learning_rate": 1.1796630836497303e-06, "loss": 0.0, "step": 3961 }, { "epoch": 2.55, "learning_rate": 1.1763953258949468e-06, "loss": 0.0, "step": 3962 }, { "epoch": 2.55, "learning_rate": 1.1731318175124496e-06, "loss": 0.0, "step": 3963 }, { "epoch": 2.55, "learning_rate": 1.1698725600739125e-06, "loss": 0.0, "step": 3964 }, { "epoch": 2.55, "learning_rate": 1.16661755514897e-06, "loss": 0.0, "step": 3965 }, { "epoch": 2.55, "learning_rate": 1.1633668043052027e-06, "loss": 0.0, "step": 3966 }, { "epoch": 2.55, "learning_rate": 1.1601203091081458e-06, "loss": 0.0, "step": 3967 }, { "epoch": 2.55, "learning_rate": 1.1568780711212812e-06, "loss": 0.0, "step": 3968 }, { "epoch": 2.55, "learning_rate": 1.153640091906043e-06, "loss": 0.0, "step": 3969 }, { "epoch": 2.55, "learning_rate": 1.1504063730218174e-06, "loss": 0.0, "step": 3970 }, { "epoch": 2.55, "learning_rate": 1.1471769160259327e-06, "loss": 0.0, "step": 3971 }, { "epoch": 2.55, "learning_rate": 1.1439517224736674e-06, "loss": 0.0, "step": 3972 }, { "epoch": 2.55, "learning_rate": 1.140730793918251e-06, "loss": 0.0, "step": 3973 }, { "epoch": 2.55, "learning_rate": 1.1375141319108484e-06, "loss": 0.0, "step": 3974 }, { "epoch": 2.55, "learning_rate": 1.1343017380005828e-06, "loss": 0.0, "step": 3975 }, { "epoch": 2.55, "learning_rate": 1.131093613734513e-06, "loss": 0.0, "step": 3976 }, { "epoch": 2.56, "learning_rate": 1.1278897606576412e-06, "loss": 0.0, "step": 3977 }, { "epoch": 2.56, "learning_rate": 1.1246901803129217e-06, "loss": 0.0, "step": 3978 }, { "epoch": 2.56, "learning_rate": 1.12149487424124e-06, "loss": 0.0, "step": 3979 }, { "epoch": 2.56, "learning_rate": 1.1183038439814308e-06, "loss": 0.0, "step": 3980 }, { "epoch": 2.56, "learning_rate": 1.1151170910702658e-06, "loss": 0.0, "step": 3981 }, { "epoch": 2.56, "learning_rate": 1.111934617042455e-06, "loss": 0.0, "step": 3982 }, { "epoch": 2.56, "learning_rate": 1.1087564234306568e-06, "loss": 0.0, "step": 3983 }, { "epoch": 2.56, "learning_rate": 1.1055825117654574e-06, "loss": 0.0, "step": 3984 }, { "epoch": 2.56, "learning_rate": 1.1024128835753866e-06, "loss": 0.0, "step": 3985 }, { "epoch": 2.56, "learning_rate": 1.0992475403869084e-06, "loss": 0.0, "step": 3986 }, { "epoch": 2.56, "learning_rate": 1.096086483724429e-06, "loss": 0.0, "step": 3987 }, { "epoch": 2.56, "learning_rate": 1.0929297151102835e-06, "loss": 0.0, "step": 3988 }, { "epoch": 2.56, "learning_rate": 1.089777236064744e-06, "loss": 0.0, "step": 3989 }, { "epoch": 2.56, "learning_rate": 1.0866290481060204e-06, "loss": 0.0, "step": 3990 }, { "epoch": 2.56, "learning_rate": 1.0834851527502476e-06, "loss": 0.0, "step": 3991 }, { "epoch": 2.57, "learning_rate": 1.0803455515115046e-06, "loss": 0.0, "step": 3992 }, { "epoch": 2.57, "learning_rate": 1.077210245901793e-06, "loss": 0.0, "step": 3993 }, { "epoch": 2.57, "learning_rate": 1.0740792374310483e-06, "loss": 0.0, "step": 3994 }, { "epoch": 2.57, "learning_rate": 1.070952527607142e-06, "loss": 0.0, "step": 3995 }, { "epoch": 2.57, "learning_rate": 1.067830117935864e-06, "loss": 0.0, "step": 3996 }, { "epoch": 2.57, "learning_rate": 1.0647120099209452e-06, "loss": 0.0, "step": 3997 }, { "epoch": 2.57, "learning_rate": 1.0615982050640373e-06, "loss": 0.0, "step": 3998 }, { "epoch": 2.57, "learning_rate": 1.058488704864722e-06, "loss": 0.0, "step": 3999 }, { "epoch": 2.57, "learning_rate": 1.055383510820509e-06, "loss": 0.0, "step": 4000 }, { "epoch": 2.57, "learning_rate": 1.0522826244268336e-06, "loss": 0.0, "step": 4001 }, { "epoch": 2.57, "learning_rate": 1.0491860471770554e-06, "loss": 0.0, "step": 4002 }, { "epoch": 2.57, "learning_rate": 1.0460937805624571e-06, "loss": 0.0001, "step": 4003 }, { "epoch": 2.57, "learning_rate": 1.0430058260722531e-06, "loss": 0.0, "step": 4004 }, { "epoch": 2.57, "learning_rate": 1.0399221851935692e-06, "loss": 0.0, "step": 4005 }, { "epoch": 2.57, "learning_rate": 1.0368428594114654e-06, "loss": 0.0001, "step": 4006 }, { "epoch": 2.57, "learning_rate": 1.0337678502089167e-06, "loss": 0.0, "step": 4007 }, { "epoch": 2.58, "learning_rate": 1.0306971590668191e-06, "loss": 0.0, "step": 4008 }, { "epoch": 2.58, "learning_rate": 1.027630787463998e-06, "loss": 0.0, "step": 4009 }, { "epoch": 2.58, "learning_rate": 1.0245687368771818e-06, "loss": 0.0, "step": 4010 }, { "epoch": 2.58, "learning_rate": 1.0215110087810366e-06, "loss": 0.0, "step": 4011 }, { "epoch": 2.58, "learning_rate": 1.0184576046481331e-06, "loss": 0.0, "step": 4012 }, { "epoch": 2.58, "learning_rate": 1.0154085259489632e-06, "loss": 0.0, "step": 4013 }, { "epoch": 2.58, "learning_rate": 1.0123637741519432e-06, "loss": 0.0, "step": 4014 }, { "epoch": 2.58, "learning_rate": 1.0093233507233945e-06, "loss": 0.0, "step": 4015 }, { "epoch": 2.58, "learning_rate": 1.0062872571275607e-06, "loss": 0.0, "step": 4016 }, { "epoch": 2.58, "learning_rate": 1.003255494826596e-06, "loss": 0.0, "step": 4017 }, { "epoch": 2.58, "learning_rate": 1.0002280652805752e-06, "loss": 0.0, "step": 4018 }, { "epoch": 2.58, "learning_rate": 9.972049699474806e-07, "loss": 0.0, "step": 4019 }, { "epoch": 2.58, "learning_rate": 9.94186210283208e-07, "loss": 0.0, "step": 4020 }, { "epoch": 2.58, "learning_rate": 9.911717877415672e-07, "loss": 0.0, "step": 4021 }, { "epoch": 2.58, "learning_rate": 9.881617037742752e-07, "loss": 0.0, "step": 4022 }, { "epoch": 2.59, "learning_rate": 9.851559598309678e-07, "loss": 0.0, "step": 4023 }, { "epoch": 2.59, "learning_rate": 9.82154557359183e-07, "loss": 0.0, "step": 4024 }, { "epoch": 2.59, "learning_rate": 9.791574978043683e-07, "loss": 0.0, "step": 4025 }, { "epoch": 2.59, "learning_rate": 9.761647826098874e-07, "loss": 0.0, "step": 4026 }, { "epoch": 2.59, "learning_rate": 9.731764132169996e-07, "loss": 0.0, "step": 4027 }, { "epoch": 2.59, "learning_rate": 9.701923910648826e-07, "loss": 0.0, "step": 4028 }, { "epoch": 2.59, "learning_rate": 9.672127175906144e-07, "loss": 0.0, "step": 4029 }, { "epoch": 2.59, "learning_rate": 9.642373942291782e-07, "loss": 0.0, "step": 4030 }, { "epoch": 2.59, "learning_rate": 9.612664224134683e-07, "loss": 0.0, "step": 4031 }, { "epoch": 2.59, "learning_rate": 9.582998035742763e-07, "loss": 0.0, "step": 4032 }, { "epoch": 2.59, "learning_rate": 9.55337539140302e-07, "loss": 0.0, "step": 4033 }, { "epoch": 2.59, "learning_rate": 9.52379630538145e-07, "loss": 0.0, "step": 4034 }, { "epoch": 2.59, "learning_rate": 9.494260791923083e-07, "loss": 0.0, "step": 4035 }, { "epoch": 2.59, "learning_rate": 9.464768865251995e-07, "loss": 0.0, "step": 4036 }, { "epoch": 2.59, "learning_rate": 9.435320539571236e-07, "loss": 0.0, "step": 4037 }, { "epoch": 2.59, "learning_rate": 9.40591582906285e-07, "loss": 0.0, "step": 4038 }, { "epoch": 2.6, "learning_rate": 9.376554747887889e-07, "loss": 0.0, "step": 4039 }, { "epoch": 2.6, "learning_rate": 9.34723731018643e-07, "loss": 0.0, "step": 4040 }, { "epoch": 2.6, "learning_rate": 9.31796353007749e-07, "loss": 0.0, "step": 4041 }, { "epoch": 2.6, "learning_rate": 9.288733421659036e-07, "loss": 0.0, "step": 4042 }, { "epoch": 2.6, "learning_rate": 9.259546999008096e-07, "loss": 0.0, "step": 4043 }, { "epoch": 2.6, "learning_rate": 9.230404276180538e-07, "loss": 0.0, "step": 4044 }, { "epoch": 2.6, "learning_rate": 9.201305267211291e-07, "loss": 0.0, "step": 4045 }, { "epoch": 2.6, "learning_rate": 9.172249986114178e-07, "loss": 0.0, "step": 4046 }, { "epoch": 2.6, "learning_rate": 9.143238446881953e-07, "loss": 0.0, "step": 4047 }, { "epoch": 2.6, "learning_rate": 9.114270663486369e-07, "loss": 0.0, "step": 4048 }, { "epoch": 2.6, "learning_rate": 9.085346649878002e-07, "loss": 0.0, "step": 4049 }, { "epoch": 2.6, "learning_rate": 9.056466419986453e-07, "loss": 0.0, "step": 4050 }, { "epoch": 2.6, "learning_rate": 9.027629987720187e-07, "loss": 0.0, "step": 4051 }, { "epoch": 2.6, "learning_rate": 8.998837366966562e-07, "loss": 0.0, "step": 4052 }, { "epoch": 2.6, "learning_rate": 8.970088571591884e-07, "loss": 0.0, "step": 4053 }, { "epoch": 2.6, "learning_rate": 8.941383615441301e-07, "loss": 0.0, "step": 4054 }, { "epoch": 2.61, "learning_rate": 8.9127225123389e-07, "loss": 0.0, "step": 4055 }, { "epoch": 2.61, "learning_rate": 8.884105276087585e-07, "loss": 0.0, "step": 4056 }, { "epoch": 2.61, "learning_rate": 8.855531920469207e-07, "loss": 0.0, "step": 4057 }, { "epoch": 2.61, "learning_rate": 8.827002459244438e-07, "loss": 0.0, "step": 4058 }, { "epoch": 2.61, "learning_rate": 8.798516906152821e-07, "loss": 0.0, "step": 4059 }, { "epoch": 2.61, "learning_rate": 8.770075274912748e-07, "loss": 0.0, "step": 4060 }, { "epoch": 2.61, "learning_rate": 8.741677579221452e-07, "loss": 0.0, "step": 4061 }, { "epoch": 2.61, "learning_rate": 8.713323832755072e-07, "loss": 0.0, "step": 4062 }, { "epoch": 2.61, "learning_rate": 8.685014049168461e-07, "loss": 0.0, "step": 4063 }, { "epoch": 2.61, "learning_rate": 8.656748242095414e-07, "loss": 0.0, "step": 4064 }, { "epoch": 2.61, "learning_rate": 8.628526425148498e-07, "loss": 0.0, "step": 4065 }, { "epoch": 2.61, "learning_rate": 8.600348611919051e-07, "loss": 0.0, "step": 4066 }, { "epoch": 2.61, "learning_rate": 8.572214815977331e-07, "loss": 0.0, "step": 4067 }, { "epoch": 2.61, "learning_rate": 8.544125050872299e-07, "loss": 0.0001, "step": 4068 }, { "epoch": 2.61, "learning_rate": 8.516079330131755e-07, "loss": 0.0, "step": 4069 }, { "epoch": 2.62, "learning_rate": 8.488077667262251e-07, "loss": 0.0, "step": 4070 }, { "epoch": 2.62, "learning_rate": 8.460120075749179e-07, "loss": 0.0, "step": 4071 }, { "epoch": 2.62, "learning_rate": 8.432206569056677e-07, "loss": 0.0, "step": 4072 }, { "epoch": 2.62, "learning_rate": 8.40433716062763e-07, "loss": 0.0, "step": 4073 }, { "epoch": 2.62, "learning_rate": 8.376511863883718e-07, "loss": 0.0, "step": 4074 }, { "epoch": 2.62, "learning_rate": 8.348730692225337e-07, "loss": 0.0, "step": 4075 }, { "epoch": 2.62, "learning_rate": 8.320993659031707e-07, "loss": 0.0, "step": 4076 }, { "epoch": 2.62, "learning_rate": 8.293300777660717e-07, "loss": 0.0, "step": 4077 }, { "epoch": 2.62, "learning_rate": 8.265652061449014e-07, "loss": 0.0, "step": 4078 }, { "epoch": 2.62, "learning_rate": 8.238047523712023e-07, "loss": 0.0, "step": 4079 }, { "epoch": 2.62, "learning_rate": 8.210487177743798e-07, "loss": 0.0, "step": 4080 }, { "epoch": 2.62, "learning_rate": 8.182971036817211e-07, "loss": 0.0, "step": 4081 }, { "epoch": 2.62, "learning_rate": 8.155499114183774e-07, "loss": 0.0, "step": 4082 }, { "epoch": 2.62, "learning_rate": 8.128071423073724e-07, "loss": 0.0, "step": 4083 }, { "epoch": 2.62, "learning_rate": 8.10068797669602e-07, "loss": 0.0, "step": 4084 }, { "epoch": 2.62, "learning_rate": 8.0733487882383e-07, "loss": 0.0, "step": 4085 }, { "epoch": 2.63, "learning_rate": 8.046053870866854e-07, "loss": 0.0, "step": 4086 }, { "epoch": 2.63, "learning_rate": 8.018803237726714e-07, "loss": 0.0, "step": 4087 }, { "epoch": 2.63, "learning_rate": 7.991596901941501e-07, "loss": 0.0, "step": 4088 }, { "epoch": 2.63, "learning_rate": 7.964434876613614e-07, "loss": 0.0, "step": 4089 }, { "epoch": 2.63, "learning_rate": 7.937317174824011e-07, "loss": 0.0, "step": 4090 }, { "epoch": 2.63, "learning_rate": 7.910243809632367e-07, "loss": 0.0, "step": 4091 }, { "epoch": 2.63, "learning_rate": 7.883214794076943e-07, "loss": 0.0, "step": 4092 }, { "epoch": 2.63, "learning_rate": 7.856230141174736e-07, "loss": 0.0, "step": 4093 }, { "epoch": 2.63, "learning_rate": 7.829289863921286e-07, "loss": 0.0, "step": 4094 }, { "epoch": 2.63, "learning_rate": 7.802393975290801e-07, "loss": 0.0, "step": 4095 }, { "epoch": 2.63, "learning_rate": 7.775542488236154e-07, "loss": 0.0, "step": 4096 }, { "epoch": 2.63, "learning_rate": 7.748735415688713e-07, "loss": 0.0, "step": 4097 }, { "epoch": 2.63, "learning_rate": 7.721972770558605e-07, "loss": 0.0, "step": 4098 }, { "epoch": 2.63, "learning_rate": 7.695254565734445e-07, "loss": 0.0, "step": 4099 }, { "epoch": 2.63, "learning_rate": 7.668580814083503e-07, "loss": 0.0, "step": 4100 }, { "epoch": 2.64, "learning_rate": 7.641951528451663e-07, "loss": 0.0, "step": 4101 }, { "epoch": 2.64, "learning_rate": 7.615366721663297e-07, "loss": 0.0, "step": 4102 }, { "epoch": 2.64, "learning_rate": 7.588826406521477e-07, "loss": 0.0, "step": 4103 }, { "epoch": 2.64, "learning_rate": 7.562330595807765e-07, "loss": 0.0, "step": 4104 }, { "epoch": 2.64, "learning_rate": 7.535879302282301e-07, "loss": 0.0, "step": 4105 }, { "epoch": 2.64, "learning_rate": 7.509472538683838e-07, "loss": 0.0003, "step": 4106 }, { "epoch": 2.64, "learning_rate": 7.483110317729624e-07, "loss": 0.0, "step": 4107 }, { "epoch": 2.64, "learning_rate": 7.456792652115486e-07, "loss": 0.0, "step": 4108 }, { "epoch": 2.64, "learning_rate": 7.430519554515758e-07, "loss": 0.0, "step": 4109 }, { "epoch": 2.64, "learning_rate": 7.404291037583389e-07, "loss": 0.0001, "step": 4110 }, { "epoch": 2.64, "learning_rate": 7.378107113949773e-07, "loss": 0.0, "step": 4111 }, { "epoch": 2.64, "learning_rate": 7.351967796224868e-07, "loss": 0.0, "step": 4112 }, { "epoch": 2.64, "learning_rate": 7.32587309699715e-07, "loss": 0.0, "step": 4113 }, { "epoch": 2.64, "learning_rate": 7.299823028833586e-07, "loss": 0.0, "step": 4114 }, { "epoch": 2.64, "learning_rate": 7.27381760427972e-07, "loss": 0.0, "step": 4115 }, { "epoch": 2.64, "learning_rate": 7.247856835859457e-07, "loss": 0.0, "step": 4116 }, { "epoch": 2.65, "learning_rate": 7.221940736075328e-07, "loss": 0.0, "step": 4117 }, { "epoch": 2.65, "learning_rate": 7.196069317408327e-07, "loss": 0.0, "step": 4118 }, { "epoch": 2.65, "learning_rate": 7.170242592317856e-07, "loss": 0.0, "step": 4119 }, { "epoch": 2.65, "learning_rate": 7.144460573241885e-07, "loss": 0.0, "step": 4120 }, { "epoch": 2.65, "learning_rate": 7.118723272596784e-07, "loss": 0.0, "step": 4121 }, { "epoch": 2.65, "learning_rate": 7.093030702777437e-07, "loss": 0.0, "step": 4122 }, { "epoch": 2.65, "learning_rate": 7.067382876157147e-07, "loss": 0.0, "step": 4123 }, { "epoch": 2.65, "learning_rate": 7.0417798050877e-07, "loss": 0.0, "step": 4124 }, { "epoch": 2.65, "learning_rate": 7.016221501899323e-07, "loss": 0.0, "step": 4125 }, { "epoch": 2.65, "learning_rate": 6.990707978900668e-07, "loss": 0.0, "step": 4126 }, { "epoch": 2.65, "learning_rate": 6.965239248378819e-07, "loss": 0.0, "step": 4127 }, { "epoch": 2.65, "learning_rate": 6.939815322599285e-07, "loss": 0.0, "step": 4128 }, { "epoch": 2.65, "learning_rate": 6.914436213806053e-07, "loss": 0.0, "step": 4129 }, { "epoch": 2.65, "learning_rate": 6.889101934221453e-07, "loss": 0.0, "step": 4130 }, { "epoch": 2.65, "learning_rate": 6.863812496046252e-07, "loss": 0.0, "step": 4131 }, { "epoch": 2.66, "learning_rate": 6.838567911459671e-07, "loss": 0.0, "step": 4132 }, { "epoch": 2.66, "learning_rate": 6.813368192619218e-07, "loss": 0.0, "step": 4133 }, { "epoch": 2.66, "learning_rate": 6.788213351660922e-07, "loss": 0.0, "step": 4134 }, { "epoch": 2.66, "learning_rate": 6.7631034006991e-07, "loss": 0.0, "step": 4135 }, { "epoch": 2.66, "learning_rate": 6.738038351826503e-07, "loss": 0.0, "step": 4136 }, { "epoch": 2.66, "learning_rate": 6.713018217114264e-07, "loss": 0.0001, "step": 4137 }, { "epoch": 2.66, "learning_rate": 6.688043008611811e-07, "loss": 0.0, "step": 4138 }, { "epoch": 2.66, "learning_rate": 6.663112738347033e-07, "loss": 0.0, "step": 4139 }, { "epoch": 2.66, "learning_rate": 6.638227418326137e-07, "loss": 0.0, "step": 4140 }, { "epoch": 2.66, "learning_rate": 6.613387060533639e-07, "loss": 0.0, "step": 4141 }, { "epoch": 2.66, "learning_rate": 6.588591676932498e-07, "loss": 0.0, "step": 4142 }, { "epoch": 2.66, "learning_rate": 6.563841279463922e-07, "loss": 0.0, "step": 4143 }, { "epoch": 2.66, "learning_rate": 6.539135880047498e-07, "loss": 0.0, "step": 4144 }, { "epoch": 2.66, "learning_rate": 6.514475490581118e-07, "loss": 0.0, "step": 4145 }, { "epoch": 2.66, "learning_rate": 6.489860122941038e-07, "loss": 0.0, "step": 4146 }, { "epoch": 2.66, "learning_rate": 6.46528978898181e-07, "loss": 0.0, "step": 4147 }, { "epoch": 2.67, "learning_rate": 6.440764500536267e-07, "loss": 0.0, "step": 4148 }, { "epoch": 2.67, "learning_rate": 6.416284269415629e-07, "loss": 0.0, "step": 4149 }, { "epoch": 2.67, "learning_rate": 6.391849107409298e-07, "loss": 0.0, "step": 4150 }, { "epoch": 2.67, "learning_rate": 6.367459026285083e-07, "loss": 0.0, "step": 4151 }, { "epoch": 2.67, "learning_rate": 6.343114037789033e-07, "loss": 0.0, "step": 4152 }, { "epoch": 2.67, "learning_rate": 6.318814153645458e-07, "loss": 0.0, "step": 4153 }, { "epoch": 2.67, "learning_rate": 6.294559385557042e-07, "loss": 0.0, "step": 4154 }, { "epoch": 2.67, "learning_rate": 6.270349745204585e-07, "loss": 0.0, "step": 4155 }, { "epoch": 2.67, "learning_rate": 6.246185244247294e-07, "loss": 0.0, "step": 4156 }, { "epoch": 2.67, "learning_rate": 6.222065894322582e-07, "loss": 0.0, "step": 4157 }, { "epoch": 2.67, "learning_rate": 6.197991707046102e-07, "loss": 0.0, "step": 4158 }, { "epoch": 2.67, "learning_rate": 6.173962694011793e-07, "loss": 0.0, "step": 4159 }, { "epoch": 2.67, "learning_rate": 6.149978866791817e-07, "loss": 0.0, "step": 4160 }, { "epoch": 2.67, "learning_rate": 6.126040236936593e-07, "loss": 0.0, "step": 4161 }, { "epoch": 2.67, "learning_rate": 6.102146815974718e-07, "loss": 0.0, "step": 4162 }, { "epoch": 2.68, "learning_rate": 6.078298615413103e-07, "loss": 0.0, "step": 4163 }, { "epoch": 2.68, "learning_rate": 6.054495646736814e-07, "loss": 0.0, "step": 4164 }, { "epoch": 2.68, "learning_rate": 6.030737921409169e-07, "loss": 0.0, "step": 4165 }, { "epoch": 2.68, "learning_rate": 6.007025450871673e-07, "loss": 0.0, "step": 4166 }, { "epoch": 2.68, "learning_rate": 5.983358246544036e-07, "loss": 0.0, "step": 4167 }, { "epoch": 2.68, "learning_rate": 5.959736319824217e-07, "loss": 0.0, "step": 4168 }, { "epoch": 2.68, "learning_rate": 5.936159682088316e-07, "loss": 0.0, "step": 4169 }, { "epoch": 2.68, "learning_rate": 5.912628344690619e-07, "loss": 0.0, "step": 4170 }, { "epoch": 2.68, "learning_rate": 5.889142318963669e-07, "loss": 0.0, "step": 4171 }, { "epoch": 2.68, "learning_rate": 5.865701616218056e-07, "loss": 0.0, "step": 4172 }, { "epoch": 2.68, "learning_rate": 5.842306247742691e-07, "loss": 0.0, "step": 4173 }, { "epoch": 2.68, "learning_rate": 5.818956224804551e-07, "loss": 0.0, "step": 4174 }, { "epoch": 2.68, "learning_rate": 5.795651558648818e-07, "loss": 0.0, "step": 4175 }, { "epoch": 2.68, "learning_rate": 5.772392260498804e-07, "loss": 0.0, "step": 4176 }, { "epoch": 2.68, "learning_rate": 5.749178341555994e-07, "loss": 0.0, "step": 4177 }, { "epoch": 2.68, "learning_rate": 5.726009813000022e-07, "loss": 0.0002, "step": 4178 }, { "epoch": 2.69, "learning_rate": 5.70288668598864e-07, "loss": 0.0, "step": 4179 }, { "epoch": 2.69, "learning_rate": 5.679808971657763e-07, "loss": 0.0, "step": 4180 }, { "epoch": 2.69, "learning_rate": 5.656776681121389e-07, "loss": 0.0, "step": 4181 }, { "epoch": 2.69, "learning_rate": 5.633789825471714e-07, "loss": 0.0, "step": 4182 }, { "epoch": 2.69, "learning_rate": 5.610848415778979e-07, "loss": 0.0, "step": 4183 }, { "epoch": 2.69, "learning_rate": 5.587952463091572e-07, "loss": 0.0, "step": 4184 }, { "epoch": 2.69, "learning_rate": 5.565101978436027e-07, "loss": 0.0, "step": 4185 }, { "epoch": 2.69, "learning_rate": 5.542296972816874e-07, "loss": 0.0, "step": 4186 }, { "epoch": 2.69, "learning_rate": 5.519537457216861e-07, "loss": 0.0, "step": 4187 }, { "epoch": 2.69, "learning_rate": 5.496823442596754e-07, "loss": 0.0, "step": 4188 }, { "epoch": 2.69, "learning_rate": 5.474154939895415e-07, "loss": 0.0, "step": 4189 }, { "epoch": 2.69, "learning_rate": 5.451531960029854e-07, "loss": 0.0, "step": 4190 }, { "epoch": 2.69, "learning_rate": 5.428954513895035e-07, "loss": 0.0, "step": 4191 }, { "epoch": 2.69, "learning_rate": 5.406422612364104e-07, "loss": 0.0, "step": 4192 }, { "epoch": 2.69, "learning_rate": 5.383936266288225e-07, "loss": 0.0, "step": 4193 }, { "epoch": 2.69, "learning_rate": 5.361495486496604e-07, "loss": 0.0, "step": 4194 }, { "epoch": 2.7, "learning_rate": 5.339100283796573e-07, "loss": 0.0, "step": 4195 }, { "epoch": 2.7, "learning_rate": 5.316750668973447e-07, "loss": 0.0, "step": 4196 }, { "epoch": 2.7, "learning_rate": 5.294446652790608e-07, "loss": 0.0001, "step": 4197 }, { "epoch": 2.7, "learning_rate": 5.272188245989452e-07, "loss": 0.0, "step": 4198 }, { "epoch": 2.7, "learning_rate": 5.249975459289491e-07, "loss": 0.0, "step": 4199 }, { "epoch": 2.7, "learning_rate": 5.227808303388182e-07, "loss": 0.0, "step": 4200 }, { "epoch": 2.7, "learning_rate": 5.205686788961017e-07, "loss": 0.0001, "step": 4201 }, { "epoch": 2.7, "learning_rate": 5.183610926661587e-07, "loss": 0.0, "step": 4202 }, { "epoch": 2.7, "learning_rate": 5.161580727121363e-07, "loss": 0.0, "step": 4203 }, { "epoch": 2.7, "learning_rate": 5.139596200949959e-07, "loss": 0.0, "step": 4204 }, { "epoch": 2.7, "learning_rate": 5.117657358734917e-07, "loss": 0.0, "step": 4205 }, { "epoch": 2.7, "learning_rate": 5.095764211041776e-07, "loss": 0.0, "step": 4206 }, { "epoch": 2.7, "learning_rate": 5.073916768414122e-07, "loss": 0.0, "step": 4207 }, { "epoch": 2.7, "learning_rate": 5.052115041373462e-07, "loss": 0.0, "step": 4208 }, { "epoch": 2.7, "learning_rate": 5.030359040419342e-07, "loss": 0.0, "step": 4209 }, { "epoch": 2.71, "learning_rate": 5.008648776029257e-07, "loss": 0.0, "step": 4210 }, { "epoch": 2.71, "learning_rate": 4.986984258658678e-07, "loss": 0.0, "step": 4211 }, { "epoch": 2.71, "learning_rate": 4.96536549874107e-07, "loss": 0.0, "step": 4212 }, { "epoch": 2.71, "learning_rate": 4.943792506687828e-07, "loss": 0.0, "step": 4213 }, { "epoch": 2.71, "learning_rate": 4.922265292888329e-07, "loss": 0.0, "step": 4214 }, { "epoch": 2.71, "learning_rate": 4.900783867709869e-07, "loss": 0.0, "step": 4215 }, { "epoch": 2.71, "learning_rate": 4.879348241497761e-07, "loss": 0.0, "step": 4216 }, { "epoch": 2.71, "learning_rate": 4.857958424575182e-07, "loss": 0.0, "step": 4217 }, { "epoch": 2.71, "learning_rate": 4.836614427243303e-07, "loss": 0.0, "step": 4218 }, { "epoch": 2.71, "learning_rate": 4.8153162597812e-07, "loss": 0.0, "step": 4219 }, { "epoch": 2.71, "learning_rate": 4.794063932445881e-07, "loss": 0.0, "step": 4220 }, { "epoch": 2.71, "learning_rate": 4.772857455472313e-07, "loss": 0.0, "step": 4221 }, { "epoch": 2.71, "learning_rate": 4.7516968390733275e-07, "loss": 0.0, "step": 4222 }, { "epoch": 2.71, "learning_rate": 4.7305820934396954e-07, "loss": 0.0, "step": 4223 }, { "epoch": 2.71, "learning_rate": 4.709513228740137e-07, "loss": 0.0001, "step": 4224 }, { "epoch": 2.71, "learning_rate": 4.68849025512117e-07, "loss": 0.0, "step": 4225 }, { "epoch": 2.72, "learning_rate": 4.6675131827073284e-07, "loss": 0.0001, "step": 4226 }, { "epoch": 2.72, "learning_rate": 4.646582021600976e-07, "loss": 0.0, "step": 4227 }, { "epoch": 2.72, "learning_rate": 4.625696781882383e-07, "loss": 0.0, "step": 4228 }, { "epoch": 2.72, "learning_rate": 4.6048574736096917e-07, "loss": 0.0, "step": 4229 }, { "epoch": 2.72, "learning_rate": 4.5840641068189306e-07, "loss": 0.0, "step": 4230 }, { "epoch": 2.72, "learning_rate": 4.5633166915240333e-07, "loss": 0.0002, "step": 4231 }, { "epoch": 2.72, "learning_rate": 4.542615237716752e-07, "loss": 0.0, "step": 4232 }, { "epoch": 2.72, "learning_rate": 4.521959755366734e-07, "loss": 0.0, "step": 4233 }, { "epoch": 2.72, "learning_rate": 4.501350254421488e-07, "loss": 0.0, "step": 4234 }, { "epoch": 2.72, "learning_rate": 4.4807867448063755e-07, "loss": 0.0, "step": 4235 }, { "epoch": 2.72, "learning_rate": 4.460269236424597e-07, "loss": 0.0, "step": 4236 }, { "epoch": 2.72, "learning_rate": 4.4397977391572034e-07, "loss": 0.0, "step": 4237 }, { "epoch": 2.72, "learning_rate": 4.4193722628631195e-07, "loss": 0.0, "step": 4238 }, { "epoch": 2.72, "learning_rate": 4.398992817379044e-07, "loss": 0.0, "step": 4239 }, { "epoch": 2.72, "learning_rate": 4.3786594125195703e-07, "loss": 0.0, "step": 4240 }, { "epoch": 2.73, "learning_rate": 4.3583720580770874e-07, "loss": 0.0, "step": 4241 }, { "epoch": 2.73, "learning_rate": 4.338130763821791e-07, "loss": 0.0, "step": 4242 }, { "epoch": 2.73, "learning_rate": 4.31793553950175e-07, "loss": 0.0, "step": 4243 }, { "epoch": 2.73, "learning_rate": 4.297786394842762e-07, "loss": 0.0, "step": 4244 }, { "epoch": 2.73, "learning_rate": 4.2776833395485215e-07, "loss": 0.0, "step": 4245 }, { "epoch": 2.73, "learning_rate": 4.257626383300484e-07, "loss": 0.0, "step": 4246 }, { "epoch": 2.73, "learning_rate": 4.2376155357578886e-07, "loss": 0.0001, "step": 4247 }, { "epoch": 2.73, "learning_rate": 4.2176508065578047e-07, "loss": 0.0, "step": 4248 }, { "epoch": 2.73, "learning_rate": 4.197732205315064e-07, "loss": 0.0, "step": 4249 }, { "epoch": 2.73, "learning_rate": 4.177859741622314e-07, "loss": 0.0, "step": 4250 }, { "epoch": 2.73, "learning_rate": 4.158033425049923e-07, "loss": 0.0, "step": 4251 }, { "epoch": 2.73, "learning_rate": 4.138253265146108e-07, "loss": 0.0, "step": 4252 }, { "epoch": 2.73, "learning_rate": 4.118519271436827e-07, "loss": 0.0, "step": 4253 }, { "epoch": 2.73, "learning_rate": 4.098831453425767e-07, "loss": 0.0, "step": 4254 }, { "epoch": 2.73, "learning_rate": 4.0791898205944644e-07, "loss": 0.0, "step": 4255 }, { "epoch": 2.73, "learning_rate": 4.05959438240211e-07, "loss": 0.0, "step": 4256 }, { "epoch": 2.74, "learning_rate": 4.0400451482857317e-07, "loss": 0.0, "step": 4257 }, { "epoch": 2.74, "learning_rate": 4.020542127660065e-07, "loss": 0.0, "step": 4258 }, { "epoch": 2.74, "learning_rate": 4.0010853299175755e-07, "loss": 0.0, "step": 4259 }, { "epoch": 2.74, "learning_rate": 3.9816747644285336e-07, "loss": 0.0, "step": 4260 }, { "epoch": 2.74, "learning_rate": 3.962310440540851e-07, "loss": 0.0, "step": 4261 }, { "epoch": 2.74, "learning_rate": 3.9429923675802674e-07, "loss": 0.0, "step": 4262 }, { "epoch": 2.74, "learning_rate": 3.923720554850163e-07, "loss": 0.0, "step": 4263 }, { "epoch": 2.74, "learning_rate": 3.9044950116316925e-07, "loss": 0.0, "step": 4264 }, { "epoch": 2.74, "learning_rate": 3.8853157471837266e-07, "loss": 0.0, "step": 4265 }, { "epoch": 2.74, "learning_rate": 3.8661827707428103e-07, "loss": 0.0, "step": 4266 }, { "epoch": 2.74, "learning_rate": 3.847096091523239e-07, "loss": 0.0, "step": 4267 }, { "epoch": 2.74, "learning_rate": 3.8280557187169833e-07, "loss": 0.0, "step": 4268 }, { "epoch": 2.74, "learning_rate": 3.8090616614937183e-07, "loss": 0.0, "step": 4269 }, { "epoch": 2.74, "learning_rate": 3.790113929000827e-07, "loss": 0.0, "step": 4270 }, { "epoch": 2.74, "learning_rate": 3.7712125303633885e-07, "loss": 0.0, "step": 4271 }, { "epoch": 2.75, "learning_rate": 3.752357474684132e-07, "loss": 0.0, "step": 4272 }, { "epoch": 2.75, "learning_rate": 3.733548771043494e-07, "loss": 0.0, "step": 4273 }, { "epoch": 2.75, "learning_rate": 3.714786428499595e-07, "loss": 0.0, "step": 4274 }, { "epoch": 2.75, "learning_rate": 3.6960704560882077e-07, "loss": 0.0, "step": 4275 }, { "epoch": 2.75, "learning_rate": 3.6774008628227884e-07, "loss": 0.0, "step": 4276 }, { "epoch": 2.75, "learning_rate": 3.658777657694468e-07, "loss": 0.0, "step": 4277 }, { "epoch": 2.75, "learning_rate": 3.640200849671982e-07, "loss": 0.0, "step": 4278 }, { "epoch": 2.75, "learning_rate": 3.6216704477018084e-07, "loss": 0.0, "step": 4279 }, { "epoch": 2.75, "learning_rate": 3.6031864607079967e-07, "loss": 0.0, "step": 4280 }, { "epoch": 2.75, "learning_rate": 3.584748897592294e-07, "loss": 0.0, "step": 4281 }, { "epoch": 2.75, "learning_rate": 3.5663577672340746e-07, "loss": 0.0, "step": 4282 }, { "epoch": 2.75, "learning_rate": 3.54801307849032e-07, "loss": 0.0, "step": 4283 }, { "epoch": 2.75, "learning_rate": 3.529714840195708e-07, "loss": 0.0, "step": 4284 }, { "epoch": 2.75, "learning_rate": 3.511463061162512e-07, "loss": 0.0, "step": 4285 }, { "epoch": 2.75, "learning_rate": 3.493257750180612e-07, "loss": 0.0, "step": 4286 }, { "epoch": 2.75, "learning_rate": 3.4750989160175274e-07, "loss": 0.0, "step": 4287 }, { "epoch": 2.76, "learning_rate": 3.456986567418419e-07, "loss": 0.0, "step": 4288 }, { "epoch": 2.76, "learning_rate": 3.438920713106031e-07, "loss": 0.0, "step": 4289 }, { "epoch": 2.76, "learning_rate": 3.420901361780704e-07, "loss": 0.0, "step": 4290 }, { "epoch": 2.76, "learning_rate": 3.402928522120441e-07, "loss": 0.0, "step": 4291 }, { "epoch": 2.76, "learning_rate": 3.385002202780774e-07, "loss": 0.0, "step": 4292 }, { "epoch": 2.76, "learning_rate": 3.3671224123948743e-07, "loss": 0.0, "step": 4293 }, { "epoch": 2.76, "learning_rate": 3.3492891595735097e-07, "loss": 0.0, "step": 4294 }, { "epoch": 2.76, "learning_rate": 3.331502452904989e-07, "loss": 0.0, "step": 4295 }, { "epoch": 2.76, "learning_rate": 3.313762300955292e-07, "loss": 0.0, "step": 4296 }, { "epoch": 2.76, "learning_rate": 3.2960687122678527e-07, "loss": 0.0, "step": 4297 }, { "epoch": 2.76, "learning_rate": 3.2784216953638095e-07, "loss": 0.0001, "step": 4298 }, { "epoch": 2.76, "learning_rate": 3.2608212587417866e-07, "loss": 0.0, "step": 4299 }, { "epoch": 2.76, "learning_rate": 3.243267410878004e-07, "loss": 0.0, "step": 4300 }, { "epoch": 2.76, "learning_rate": 3.2257601602262655e-07, "loss": 0.0, "step": 4301 }, { "epoch": 2.76, "learning_rate": 3.208299515217883e-07, "loss": 0.0, "step": 4302 }, { "epoch": 2.76, "learning_rate": 3.1908854842617855e-07, "loss": 0.0, "step": 4303 }, { "epoch": 2.77, "learning_rate": 3.1735180757443753e-07, "loss": 0.0, "step": 4304 }, { "epoch": 2.77, "learning_rate": 3.156197298029695e-07, "loss": 0.0, "step": 4305 }, { "epoch": 2.77, "learning_rate": 3.1389231594592616e-07, "loss": 0.0, "step": 4306 }, { "epoch": 2.77, "learning_rate": 3.121695668352143e-07, "loss": 0.0, "step": 4307 }, { "epoch": 2.77, "learning_rate": 3.104514833004979e-07, "loss": 0.0, "step": 4308 }, { "epoch": 2.77, "learning_rate": 3.0873806616918744e-07, "loss": 0.0, "step": 4309 }, { "epoch": 2.77, "learning_rate": 3.070293162664528e-07, "loss": 0.0, "step": 4310 }, { "epoch": 2.77, "learning_rate": 3.0532523441521356e-07, "loss": 0.0001, "step": 4311 }, { "epoch": 2.77, "learning_rate": 3.036258214361376e-07, "loss": 0.0, "step": 4312 }, { "epoch": 2.77, "learning_rate": 3.0193107814765145e-07, "loss": 0.0, "step": 4313 }, { "epoch": 2.77, "learning_rate": 3.0024100536592547e-07, "loss": 0.0, "step": 4314 }, { "epoch": 2.77, "learning_rate": 2.985556039048876e-07, "loss": 0.0, "step": 4315 }, { "epoch": 2.77, "learning_rate": 2.9687487457621066e-07, "loss": 0.0, "step": 4316 }, { "epoch": 2.77, "learning_rate": 2.9519881818931837e-07, "loss": 0.0, "step": 4317 }, { "epoch": 2.77, "learning_rate": 2.935274355513884e-07, "loss": 0.0, "step": 4318 }, { "epoch": 2.78, "learning_rate": 2.918607274673424e-07, "loss": 0.0, "step": 4319 }, { "epoch": 2.78, "learning_rate": 2.9019869473985275e-07, "loss": 0.0, "step": 4320 }, { "epoch": 2.78, "learning_rate": 2.8854133816934025e-07, "loss": 0.0, "step": 4321 }, { "epoch": 2.78, "learning_rate": 2.868886585539732e-07, "loss": 0.0, "step": 4322 }, { "epoch": 2.78, "learning_rate": 2.852406566896682e-07, "loss": 0.0, "step": 4323 }, { "epoch": 2.78, "learning_rate": 2.835973333700892e-07, "loss": 0.0, "step": 4324 }, { "epoch": 2.78, "learning_rate": 2.819586893866466e-07, "loss": 0.0, "step": 4325 }, { "epoch": 2.78, "learning_rate": 2.803247255284958e-07, "loss": 0.0, "step": 4326 }, { "epoch": 2.78, "learning_rate": 2.786954425825417e-07, "loss": 0.0, "step": 4327 }, { "epoch": 2.78, "learning_rate": 2.770708413334322e-07, "loss": 0.0, "step": 4328 }, { "epoch": 2.78, "learning_rate": 2.754509225635604e-07, "loss": 0.0, "step": 4329 }, { "epoch": 2.78, "learning_rate": 2.738356870530678e-07, "loss": 0.0, "step": 4330 }, { "epoch": 2.78, "learning_rate": 2.7222513557983553e-07, "loss": 0.0, "step": 4331 }, { "epoch": 2.78, "learning_rate": 2.7061926891949197e-07, "loss": 0.0, "step": 4332 }, { "epoch": 2.78, "learning_rate": 2.690180878454096e-07, "loss": 0.0, "step": 4333 }, { "epoch": 2.78, "learning_rate": 2.6742159312870163e-07, "loss": 0.0, "step": 4334 }, { "epoch": 2.79, "learning_rate": 2.658297855382297e-07, "loss": 0.0, "step": 4335 }, { "epoch": 2.79, "learning_rate": 2.6424266584059054e-07, "loss": 0.0, "step": 4336 }, { "epoch": 2.79, "learning_rate": 2.626602348001295e-07, "loss": 0.0, "step": 4337 }, { "epoch": 2.79, "learning_rate": 2.610824931789313e-07, "loss": 0.0, "step": 4338 }, { "epoch": 2.79, "learning_rate": 2.595094417368238e-07, "loss": 0.0, "step": 4339 }, { "epoch": 2.79, "learning_rate": 2.579410812313732e-07, "loss": 0.0, "step": 4340 }, { "epoch": 2.79, "learning_rate": 2.563774124178908e-07, "loss": 0.0, "step": 4341 }, { "epoch": 2.79, "learning_rate": 2.5481843604942437e-07, "loss": 0.0, "step": 4342 }, { "epoch": 2.79, "learning_rate": 2.5326415287676323e-07, "loss": 0.0, "step": 4343 }, { "epoch": 2.79, "learning_rate": 2.517145636484386e-07, "loss": 0.0, "step": 4344 }, { "epoch": 2.79, "learning_rate": 2.5016966911071693e-07, "loss": 0.0, "step": 4345 }, { "epoch": 2.79, "learning_rate": 2.4862947000760864e-07, "loss": 0.0, "step": 4346 }, { "epoch": 2.79, "learning_rate": 2.470939670808603e-07, "loss": 0.0, "step": 4347 }, { "epoch": 2.79, "learning_rate": 2.455631610699538e-07, "loss": 0.0, "step": 4348 }, { "epoch": 2.79, "learning_rate": 2.4403705271211606e-07, "loss": 0.0, "step": 4349 }, { "epoch": 2.8, "learning_rate": 2.425156427423048e-07, "loss": 0.0, "step": 4350 }, { "epoch": 2.8, "learning_rate": 2.409989318932204e-07, "loss": 0.0, "step": 4351 }, { "epoch": 2.8, "learning_rate": 2.3948692089529544e-07, "loss": 0.0, "step": 4352 }, { "epoch": 2.8, "learning_rate": 2.3797961047670292e-07, "loss": 0.0, "step": 4353 }, { "epoch": 2.8, "learning_rate": 2.3647700136335105e-07, "loss": 0.0, "step": 4354 }, { "epoch": 2.8, "learning_rate": 2.349790942788821e-07, "loss": 0.0, "step": 4355 }, { "epoch": 2.8, "learning_rate": 2.334858899446768e-07, "loss": 0.0, "step": 4356 }, { "epoch": 2.8, "learning_rate": 2.319973890798477e-07, "loss": 0.0, "step": 4357 }, { "epoch": 2.8, "learning_rate": 2.3051359240124583e-07, "loss": 0.0, "step": 4358 }, { "epoch": 2.8, "learning_rate": 2.2903450062345288e-07, "loss": 0.0, "step": 4359 }, { "epoch": 2.8, "learning_rate": 2.2756011445878912e-07, "loss": 0.0, "step": 4360 }, { "epoch": 2.8, "learning_rate": 2.260904346173054e-07, "loss": 0.0, "step": 4361 }, { "epoch": 2.8, "learning_rate": 2.2462546180678556e-07, "loss": 0.0, "step": 4362 }, { "epoch": 2.8, "learning_rate": 2.2316519673274972e-07, "loss": 0.0, "step": 4363 }, { "epoch": 2.8, "learning_rate": 2.2170964009844753e-07, "loss": 0.0, "step": 4364 }, { "epoch": 2.8, "learning_rate": 2.2025879260486382e-07, "loss": 0.0, "step": 4365 }, { "epoch": 2.81, "learning_rate": 2.1881265495071524e-07, "loss": 0.0, "step": 4366 }, { "epoch": 2.81, "learning_rate": 2.173712278324458e-07, "loss": 0.0, "step": 4367 }, { "epoch": 2.81, "learning_rate": 2.1593451194423798e-07, "loss": 0.0, "step": 4368 }, { "epoch": 2.81, "learning_rate": 2.1450250797799942e-07, "loss": 0.0, "step": 4369 }, { "epoch": 2.81, "learning_rate": 2.1307521662337182e-07, "loss": 0.0, "step": 4370 }, { "epoch": 2.81, "learning_rate": 2.1165263856772643e-07, "loss": 0.0, "step": 4371 }, { "epoch": 2.81, "learning_rate": 2.1023477449616525e-07, "loss": 0.0, "step": 4372 }, { "epoch": 2.81, "learning_rate": 2.088216250915187e-07, "loss": 0.0, "step": 4373 }, { "epoch": 2.81, "learning_rate": 2.07413191034348e-07, "loss": 0.0, "step": 4374 }, { "epoch": 2.81, "learning_rate": 2.060094730029416e-07, "loss": 0.0, "step": 4375 }, { "epoch": 2.81, "learning_rate": 2.0461047167331993e-07, "loss": 0.0, "step": 4376 }, { "epoch": 2.81, "learning_rate": 2.0321618771922846e-07, "loss": 0.0, "step": 4377 }, { "epoch": 2.81, "learning_rate": 2.0182662181214342e-07, "loss": 0.0, "step": 4378 }, { "epoch": 2.81, "learning_rate": 2.0044177462126613e-07, "loss": 0.0, "step": 4379 }, { "epoch": 2.81, "learning_rate": 1.990616468135298e-07, "loss": 0.0, "step": 4380 }, { "epoch": 2.82, "learning_rate": 1.976862390535905e-07, "loss": 0.0, "step": 4381 }, { "epoch": 2.82, "learning_rate": 1.9631555200383178e-07, "loss": 0.0, "step": 4382 }, { "epoch": 2.82, "learning_rate": 1.949495863243678e-07, "loss": 0.0, "step": 4383 }, { "epoch": 2.82, "learning_rate": 1.9358834267303229e-07, "loss": 0.0001, "step": 4384 }, { "epoch": 2.82, "learning_rate": 1.922318217053909e-07, "loss": 0.0, "step": 4385 }, { "epoch": 2.82, "learning_rate": 1.9088002407473327e-07, "loss": 0.0, "step": 4386 }, { "epoch": 2.82, "learning_rate": 1.8953295043206977e-07, "loss": 0.0, "step": 4387 }, { "epoch": 2.82, "learning_rate": 1.8819060142614477e-07, "loss": 0.0, "step": 4388 }, { "epoch": 2.82, "learning_rate": 1.868529777034167e-07, "loss": 0.0, "step": 4389 }, { "epoch": 2.82, "learning_rate": 1.8552007990807698e-07, "loss": 0.0, "step": 4390 }, { "epoch": 2.82, "learning_rate": 1.841919086820365e-07, "loss": 0.0, "step": 4391 }, { "epoch": 2.82, "learning_rate": 1.8286846466493257e-07, "loss": 0.0, "step": 4392 }, { "epoch": 2.82, "learning_rate": 1.8154974849412198e-07, "loss": 0.0, "step": 4393 }, { "epoch": 2.82, "learning_rate": 1.8023576080468895e-07, "loss": 0.0, "step": 4394 }, { "epoch": 2.82, "learning_rate": 1.789265022294395e-07, "loss": 0.0, "step": 4395 }, { "epoch": 2.82, "learning_rate": 1.7762197339889818e-07, "loss": 0.0, "step": 4396 }, { "epoch": 2.83, "learning_rate": 1.763221749413191e-07, "loss": 0.0, "step": 4397 }, { "epoch": 2.83, "learning_rate": 1.7502710748266927e-07, "loss": 0.0, "step": 4398 }, { "epoch": 2.83, "learning_rate": 1.7373677164664648e-07, "loss": 0.0, "step": 4399 }, { "epoch": 2.83, "learning_rate": 1.7245116805466254e-07, "loss": 0.0, "step": 4400 }, { "epoch": 2.83, "learning_rate": 1.7117029732585332e-07, "loss": 0.0, "step": 4401 }, { "epoch": 2.83, "learning_rate": 1.698941600770787e-07, "loss": 0.0, "step": 4402 }, { "epoch": 2.83, "learning_rate": 1.6862275692290931e-07, "loss": 0.0, "step": 4403 }, { "epoch": 2.83, "learning_rate": 1.6735608847564644e-07, "loss": 0.0, "step": 4404 }, { "epoch": 2.83, "learning_rate": 1.660941553453066e-07, "loss": 0.0, "step": 4405 }, { "epoch": 2.83, "learning_rate": 1.6483695813962363e-07, "loss": 0.0, "step": 4406 }, { "epoch": 2.83, "learning_rate": 1.6358449746405435e-07, "loss": 0.0, "step": 4407 }, { "epoch": 2.83, "learning_rate": 1.6233677392177405e-07, "loss": 0.0, "step": 4408 }, { "epoch": 2.83, "learning_rate": 1.6109378811367314e-07, "loss": 0.0, "step": 4409 }, { "epoch": 2.83, "learning_rate": 1.5985554063836506e-07, "loss": 0.0, "step": 4410 }, { "epoch": 2.83, "learning_rate": 1.586220320921783e-07, "loss": 0.0, "step": 4411 }, { "epoch": 2.84, "learning_rate": 1.573932630691599e-07, "loss": 0.0, "step": 4412 }, { "epoch": 2.84, "learning_rate": 1.5616923416107544e-07, "loss": 0.0, "step": 4413 }, { "epoch": 2.84, "learning_rate": 1.5494994595740664e-07, "loss": 0.0, "step": 4414 }, { "epoch": 2.84, "learning_rate": 1.537353990453494e-07, "loss": 0.0, "step": 4415 }, { "epoch": 2.84, "learning_rate": 1.5252559400982247e-07, "loss": 0.0, "step": 4416 }, { "epoch": 2.84, "learning_rate": 1.513205314334576e-07, "loss": 0.0, "step": 4417 }, { "epoch": 2.84, "learning_rate": 1.5012021189659943e-07, "loss": 0.0, "step": 4418 }, { "epoch": 2.84, "learning_rate": 1.4892463597731555e-07, "loss": 0.0, "step": 4419 }, { "epoch": 2.84, "learning_rate": 1.4773380425138095e-07, "loss": 0.0, "step": 4420 }, { "epoch": 2.84, "learning_rate": 1.4654771729229355e-07, "loss": 0.0, "step": 4421 }, { "epoch": 2.84, "learning_rate": 1.4536637567125978e-07, "loss": 0.0, "step": 4422 }, { "epoch": 2.84, "learning_rate": 1.441897799572056e-07, "loss": 0.0, "step": 4423 }, { "epoch": 2.84, "learning_rate": 1.4301793071676894e-07, "loss": 0.0, "step": 4424 }, { "epoch": 2.84, "learning_rate": 1.4185082851430276e-07, "loss": 0.0, "step": 4425 }, { "epoch": 2.84, "learning_rate": 1.406884739118719e-07, "loss": 0.0, "step": 4426 }, { "epoch": 2.84, "learning_rate": 1.3953086746925858e-07, "loss": 0.0, "step": 4427 }, { "epoch": 2.85, "learning_rate": 1.3837800974395354e-07, "loss": 0.0, "step": 4428 }, { "epoch": 2.85, "learning_rate": 1.3722990129116488e-07, "loss": 0.0, "step": 4429 }, { "epoch": 2.85, "learning_rate": 1.360865426638125e-07, "loss": 0.0, "step": 4430 }, { "epoch": 2.85, "learning_rate": 1.3494793441252595e-07, "loss": 0.0, "step": 4431 }, { "epoch": 2.85, "learning_rate": 1.3381407708564998e-07, "loss": 0.0, "step": 4432 }, { "epoch": 2.85, "learning_rate": 1.326849712292422e-07, "loss": 0.0, "step": 4433 }, { "epoch": 2.85, "learning_rate": 1.315606173870676e-07, "loss": 0.0, "step": 4434 }, { "epoch": 2.85, "learning_rate": 1.3044101610060645e-07, "loss": 0.0, "step": 4435 }, { "epoch": 2.85, "learning_rate": 1.2932616790905183e-07, "loss": 0.0, "step": 4436 }, { "epoch": 2.85, "learning_rate": 1.2821607334929986e-07, "loss": 0.0, "step": 4437 }, { "epoch": 2.85, "learning_rate": 1.2711073295596621e-07, "loss": 0.0, "step": 4438 }, { "epoch": 2.85, "learning_rate": 1.2601014726137283e-07, "loss": 0.0, "step": 4439 }, { "epoch": 2.85, "learning_rate": 1.2491431679555133e-07, "loss": 0.0001, "step": 4440 }, { "epoch": 2.85, "learning_rate": 1.2382324208624618e-07, "loss": 0.0, "step": 4441 }, { "epoch": 2.85, "learning_rate": 1.227369236589071e-07, "loss": 0.0, "step": 4442 }, { "epoch": 2.85, "learning_rate": 1.216553620366967e-07, "loss": 0.0, "step": 4443 }, { "epoch": 2.86, "learning_rate": 1.205785577404861e-07, "loss": 0.0, "step": 4444 }, { "epoch": 2.86, "learning_rate": 1.1950651128885492e-07, "loss": 0.0, "step": 4445 }, { "epoch": 2.86, "learning_rate": 1.1843922319809131e-07, "loss": 0.0, "step": 4446 }, { "epoch": 2.86, "learning_rate": 1.1737669398219298e-07, "loss": 0.0, "step": 4447 }, { "epoch": 2.86, "learning_rate": 1.1631892415286394e-07, "loss": 0.0, "step": 4448 }, { "epoch": 2.86, "learning_rate": 1.152659142195156e-07, "loss": 0.0, "step": 4449 }, { "epoch": 2.86, "learning_rate": 1.1421766468927232e-07, "loss": 0.0, "step": 4450 }, { "epoch": 2.86, "learning_rate": 1.1317417606695913e-07, "loss": 0.0001, "step": 4451 }, { "epoch": 2.86, "learning_rate": 1.1213544885511186e-07, "loss": 0.0, "step": 4452 }, { "epoch": 2.86, "learning_rate": 1.1110148355397255e-07, "loss": 0.0, "step": 4453 }, { "epoch": 2.86, "learning_rate": 1.1007228066149068e-07, "loss": 0.0, "step": 4454 }, { "epoch": 2.86, "learning_rate": 1.0904784067332086e-07, "loss": 0.0, "step": 4455 }, { "epoch": 2.86, "learning_rate": 1.0802816408282401e-07, "loss": 0.0, "step": 4456 }, { "epoch": 2.86, "learning_rate": 1.0701325138106955e-07, "loss": 0.0, "step": 4457 }, { "epoch": 2.86, "learning_rate": 1.0600310305682981e-07, "loss": 0.0, "step": 4458 }, { "epoch": 2.87, "learning_rate": 1.0499771959658122e-07, "loss": 0.0, "step": 4459 }, { "epoch": 2.87, "learning_rate": 1.0399710148451203e-07, "loss": 0.0, "step": 4460 }, { "epoch": 2.87, "learning_rate": 1.0300124920250898e-07, "loss": 0.0, "step": 4461 }, { "epoch": 2.87, "learning_rate": 1.020101632301651e-07, "loss": 0.0, "step": 4462 }, { "epoch": 2.87, "learning_rate": 1.0102384404478083e-07, "loss": 0.0, "step": 4463 }, { "epoch": 2.87, "learning_rate": 1.0004229212135841e-07, "loss": 0.0, "step": 4464 }, { "epoch": 2.87, "learning_rate": 9.906550793260417e-08, "loss": 0.0, "step": 4465 }, { "epoch": 2.87, "learning_rate": 9.809349194892959e-08, "loss": 0.0, "step": 4466 }, { "epoch": 2.87, "learning_rate": 9.71262446384491e-08, "loss": 0.0, "step": 4467 }, { "epoch": 2.87, "learning_rate": 9.616376646698011e-08, "loss": 0.0, "step": 4468 }, { "epoch": 2.87, "learning_rate": 9.520605789804516e-08, "loss": 0.0, "step": 4469 }, { "epoch": 2.87, "learning_rate": 9.425311939286752e-08, "loss": 0.0, "step": 4470 }, { "epoch": 2.87, "learning_rate": 9.330495141037454e-08, "loss": 0.0, "step": 4471 }, { "epoch": 2.87, "learning_rate": 9.23615544071954e-08, "loss": 0.0, "step": 4472 }, { "epoch": 2.87, "learning_rate": 9.142292883766224e-08, "loss": 0.0, "step": 4473 }, { "epoch": 2.87, "learning_rate": 9.048907515381011e-08, "loss": 0.0, "step": 4474 }, { "epoch": 2.88, "learning_rate": 8.955999380537261e-08, "loss": 0.0, "step": 4475 }, { "epoch": 2.88, "learning_rate": 8.86356852397885e-08, "loss": 0.0, "step": 4476 }, { "epoch": 2.88, "learning_rate": 8.771614990219835e-08, "loss": 0.0002, "step": 4477 }, { "epoch": 2.88, "learning_rate": 8.680138823544016e-08, "loss": 0.0, "step": 4478 }, { "epoch": 2.88, "learning_rate": 8.589140068005708e-08, "loss": 0.0, "step": 4479 }, { "epoch": 2.88, "learning_rate": 8.498618767428968e-08, "loss": 0.0, "step": 4480 }, { "epoch": 2.88, "learning_rate": 8.408574965408145e-08, "loss": 0.0, "step": 4481 }, { "epoch": 2.88, "learning_rate": 8.319008705307662e-08, "loss": 0.0, "step": 4482 }, { "epoch": 2.88, "learning_rate": 8.229920030261795e-08, "loss": 0.0, "step": 4483 }, { "epoch": 2.88, "learning_rate": 8.141308983174778e-08, "loss": 0.0, "step": 4484 }, { "epoch": 2.88, "learning_rate": 8.053175606721032e-08, "loss": 0.0, "step": 4485 }, { "epoch": 2.88, "learning_rate": 7.965519943344824e-08, "loss": 0.0, "step": 4486 }, { "epoch": 2.88, "learning_rate": 7.87834203526039e-08, "loss": 0.0, "step": 4487 }, { "epoch": 2.88, "learning_rate": 7.791641924451809e-08, "loss": 0.0, "step": 4488 }, { "epoch": 2.88, "learning_rate": 7.705419652673241e-08, "loss": 0.0, "step": 4489 }, { "epoch": 2.89, "learning_rate": 7.619675261448356e-08, "loss": 0.0, "step": 4490 }, { "epoch": 2.89, "learning_rate": 7.534408792071124e-08, "loss": 0.0, "step": 4491 }, { "epoch": 2.89, "learning_rate": 7.449620285605141e-08, "loss": 0.0, "step": 4492 }, { "epoch": 2.89, "learning_rate": 7.365309782883634e-08, "loss": 0.0, "step": 4493 }, { "epoch": 2.89, "learning_rate": 7.281477324510012e-08, "loss": 0.0, "step": 4494 }, { "epoch": 2.89, "learning_rate": 7.1981229508572e-08, "loss": 0.0, "step": 4495 }, { "epoch": 2.89, "learning_rate": 7.115246702067868e-08, "loss": 0.0, "step": 4496 }, { "epoch": 2.89, "learning_rate": 7.032848618054644e-08, "loss": 0.0, "step": 4497 }, { "epoch": 2.89, "learning_rate": 6.950928738499674e-08, "loss": 0.0, "step": 4498 }, { "epoch": 2.89, "learning_rate": 6.869487102854844e-08, "loss": 0.0, "step": 4499 }, { "epoch": 2.89, "learning_rate": 6.78852375034178e-08, "loss": 0.0, "step": 4500 }, { "epoch": 2.89, "learning_rate": 6.708038719951737e-08, "loss": 0.0, "step": 4501 }, { "epoch": 2.89, "learning_rate": 6.628032050445598e-08, "loss": 0.0, "step": 4502 }, { "epoch": 2.89, "learning_rate": 6.548503780353988e-08, "loss": 0.0001, "step": 4503 }, { "epoch": 2.89, "learning_rate": 6.469453947976823e-08, "loss": 0.0, "step": 4504 }, { "epoch": 2.89, "learning_rate": 6.390882591383985e-08, "loss": 0.0, "step": 4505 }, { "epoch": 2.9, "learning_rate": 6.312789748414871e-08, "loss": 0.0, "step": 4506 }, { "epoch": 2.9, "learning_rate": 6.235175456677956e-08, "loss": 0.0, "step": 4507 }, { "epoch": 2.9, "learning_rate": 6.158039753552115e-08, "loss": 0.0, "step": 4508 }, { "epoch": 2.9, "learning_rate": 6.081382676184743e-08, "loss": 0.0, "step": 4509 }, { "epoch": 2.9, "learning_rate": 6.005204261493535e-08, "loss": 0.0, "step": 4510 }, { "epoch": 2.9, "learning_rate": 5.9295045461652544e-08, "loss": 0.0, "step": 4511 }, { "epoch": 2.9, "learning_rate": 5.854283566656294e-08, "loss": 0.0, "step": 4512 }, { "epoch": 2.9, "learning_rate": 5.779541359192453e-08, "loss": 0.0, "step": 4513 }, { "epoch": 2.9, "learning_rate": 5.705277959768829e-08, "loss": 0.0, "step": 4514 }, { "epoch": 2.9, "learning_rate": 5.6314934041501455e-08, "loss": 0.0, "step": 4515 }, { "epoch": 2.9, "learning_rate": 5.558187727870312e-08, "loss": 0.0, "step": 4516 }, { "epoch": 2.9, "learning_rate": 5.4853609662327554e-08, "loss": 0.0, "step": 4517 }, { "epoch": 2.9, "learning_rate": 5.4130131543103094e-08, "loss": 0.0, "step": 4518 }, { "epoch": 2.9, "learning_rate": 5.3411443269449915e-08, "loss": 0.0, "step": 4519 }, { "epoch": 2.9, "learning_rate": 5.269754518748116e-08, "loss": 0.0, "step": 4520 }, { "epoch": 2.91, "learning_rate": 5.198843764100403e-08, "loss": 0.0, "step": 4521 }, { "epoch": 2.91, "learning_rate": 5.128412097152091e-08, "loss": 0.0, "step": 4522 }, { "epoch": 2.91, "learning_rate": 5.0584595518222704e-08, "loss": 0.0, "step": 4523 }, { "epoch": 2.91, "learning_rate": 4.9889861617995474e-08, "loss": 0.0, "step": 4524 }, { "epoch": 2.91, "learning_rate": 4.9199919605417144e-08, "loss": 0.0, "step": 4525 }, { "epoch": 2.91, "learning_rate": 4.851476981275749e-08, "loss": 0.0, "step": 4526 }, { "epoch": 2.91, "learning_rate": 4.7834412569979226e-08, "loss": 0.0001, "step": 4527 }, { "epoch": 2.91, "learning_rate": 4.715884820473582e-08, "loss": 0.0, "step": 4528 }, { "epoch": 2.91, "learning_rate": 4.648807704237368e-08, "loss": 0.0, "step": 4529 }, { "epoch": 2.91, "learning_rate": 4.582209940593108e-08, "loss": 0.0, "step": 4530 }, { "epoch": 2.91, "learning_rate": 4.5160915616137e-08, "loss": 0.0, "step": 4531 }, { "epoch": 2.91, "learning_rate": 4.4504525991411193e-08, "loss": 0.0, "step": 4532 }, { "epoch": 2.91, "learning_rate": 4.3852930847865236e-08, "loss": 0.0, "step": 4533 }, { "epoch": 2.91, "learning_rate": 4.320613049930145e-08, "loss": 0.0, "step": 4534 }, { "epoch": 2.91, "learning_rate": 4.2564125257215094e-08, "loss": 0.0, "step": 4535 }, { "epoch": 2.91, "learning_rate": 4.192691543078886e-08, "loss": 0.0, "step": 4536 }, { "epoch": 2.92, "learning_rate": 4.1294501326898387e-08, "loss": 0.0, "step": 4537 }, { "epoch": 2.92, "learning_rate": 4.06668832501067e-08, "loss": 0.0, "step": 4538 }, { "epoch": 2.92, "learning_rate": 4.0044061502673146e-08, "loss": 0.0, "step": 4539 }, { "epoch": 2.92, "learning_rate": 3.942603638454001e-08, "loss": 0.0, "step": 4540 }, { "epoch": 2.92, "learning_rate": 3.881280819334365e-08, "loss": 0.0, "step": 4541 }, { "epoch": 2.92, "learning_rate": 3.820437722441117e-08, "loss": 0.0, "step": 4542 }, { "epoch": 2.92, "learning_rate": 3.760074377075596e-08, "loss": 0.0, "step": 4543 }, { "epoch": 2.92, "learning_rate": 3.7001908123083244e-08, "loss": 0.0, "step": 4544 }, { "epoch": 2.92, "learning_rate": 3.6407870569786784e-08, "loss": 0.0, "step": 4545 }, { "epoch": 2.92, "learning_rate": 3.581863139694996e-08, "loss": 0.0, "step": 4546 }, { "epoch": 2.92, "learning_rate": 3.5234190888345786e-08, "loss": 0.0, "step": 4547 }, { "epoch": 2.92, "learning_rate": 3.465454932543577e-08, "loss": 0.0, "step": 4548 }, { "epoch": 2.92, "learning_rate": 3.407970698736995e-08, "loss": 0.0, "step": 4549 }, { "epoch": 2.92, "learning_rate": 3.3509664150987996e-08, "loss": 0.0, "step": 4550 }, { "epoch": 2.92, "learning_rate": 3.294442109081697e-08, "loss": 0.0, "step": 4551 }, { "epoch": 2.92, "learning_rate": 3.238397807907356e-08, "loss": 0.0, "step": 4552 }, { "epoch": 2.93, "learning_rate": 3.182833538566299e-08, "loss": 0.0, "step": 4553 }, { "epoch": 2.93, "learning_rate": 3.1277493278176755e-08, "loss": 0.0, "step": 4554 }, { "epoch": 2.93, "learning_rate": 3.073145202189487e-08, "loss": 0.0, "step": 4555 }, { "epoch": 2.93, "learning_rate": 3.019021187978921e-08, "loss": 0.0001, "step": 4556 }, { "epoch": 2.93, "learning_rate": 2.9653773112513497e-08, "loss": 0.0, "step": 4557 }, { "epoch": 2.93, "learning_rate": 2.912213597841329e-08, "loss": 0.0, "step": 4558 }, { "epoch": 2.93, "learning_rate": 2.859530073352046e-08, "loss": 0.0, "step": 4559 }, { "epoch": 2.93, "learning_rate": 2.8073267631553157e-08, "loss": 0.0, "step": 4560 }, { "epoch": 2.93, "learning_rate": 2.7556036923920282e-08, "loss": 0.0, "step": 4561 }, { "epoch": 2.93, "learning_rate": 2.7043608859713687e-08, "loss": 0.0, "step": 4562 }, { "epoch": 2.93, "learning_rate": 2.653598368571597e-08, "loss": 0.0001, "step": 4563 }, { "epoch": 2.93, "learning_rate": 2.6033161646393802e-08, "loss": 0.0, "step": 4564 }, { "epoch": 2.93, "learning_rate": 2.553514298390125e-08, "loss": 0.0, "step": 4565 }, { "epoch": 2.93, "learning_rate": 2.504192793808202e-08, "loss": 0.0, "step": 4566 }, { "epoch": 2.93, "learning_rate": 2.455351674646389e-08, "loss": 0.0, "step": 4567 }, { "epoch": 2.94, "learning_rate": 2.4069909644259815e-08, "loss": 0.0, "step": 4568 }, { "epoch": 2.94, "learning_rate": 2.359110686437127e-08, "loss": 0.0, "step": 4569 }, { "epoch": 2.94, "learning_rate": 2.3117108637386032e-08, "loss": 0.0, "step": 4570 }, { "epoch": 2.94, "learning_rate": 2.2647915191579273e-08, "loss": 0.0, "step": 4571 }, { "epoch": 2.94, "learning_rate": 2.2183526752906914e-08, "loss": 0.0, "step": 4572 }, { "epoch": 2.94, "learning_rate": 2.1723943545017832e-08, "loss": 0.0, "step": 4573 }, { "epoch": 2.94, "learning_rate": 2.126916578924054e-08, "loss": 0.0, "step": 4574 }, { "epoch": 2.94, "learning_rate": 2.0819193704594287e-08, "loss": 0.0, "step": 4575 }, { "epoch": 2.94, "learning_rate": 2.037402750778128e-08, "loss": 0.0, "step": 4576 }, { "epoch": 2.94, "learning_rate": 1.9933667413188917e-08, "loss": 0.0, "step": 4577 }, { "epoch": 2.94, "learning_rate": 1.9498113632892003e-08, "loss": 0.0, "step": 4578 }, { "epoch": 2.94, "learning_rate": 1.90673663766483e-08, "loss": 0.0, "step": 4579 }, { "epoch": 2.94, "learning_rate": 1.8641425851902984e-08, "loss": 0.0, "step": 4580 }, { "epoch": 2.94, "learning_rate": 1.822029226378419e-08, "loss": 0.0, "step": 4581 }, { "epoch": 2.94, "learning_rate": 1.780396581510635e-08, "loss": 0.0, "step": 4582 }, { "epoch": 2.94, "learning_rate": 1.7392446706371302e-08, "loss": 0.0, "step": 4583 }, { "epoch": 2.95, "learning_rate": 1.6985735135759406e-08, "loss": 0.0, "step": 4584 }, { "epoch": 2.95, "learning_rate": 1.6583831299141763e-08, "loss": 0.0, "step": 4585 }, { "epoch": 2.95, "learning_rate": 1.6186735390071318e-08, "loss": 0.0, "step": 4586 }, { "epoch": 2.95, "learning_rate": 1.579444759978621e-08, "loss": 0.0, "step": 4587 }, { "epoch": 2.95, "learning_rate": 1.5406968117208654e-08, "loss": 0.0, "step": 4588 }, { "epoch": 2.95, "learning_rate": 1.5024297128944932e-08, "loss": 0.0, "step": 4589 }, { "epoch": 2.95, "learning_rate": 1.4646434819286514e-08, "loss": 0.0, "step": 4590 }, { "epoch": 2.95, "learning_rate": 1.427338137020895e-08, "loss": 0.0, "step": 4591 }, { "epoch": 2.95, "learning_rate": 1.3905136961371857e-08, "loss": 0.0, "step": 4592 }, { "epoch": 2.95, "learning_rate": 1.3541701770117821e-08, "loss": 0.0, "step": 4593 }, { "epoch": 2.95, "learning_rate": 1.3183075971473503e-08, "loss": 0.0, "step": 4594 }, { "epoch": 2.95, "learning_rate": 1.2829259738151857e-08, "loss": 0.0, "step": 4595 }, { "epoch": 2.95, "learning_rate": 1.248025324054658e-08, "loss": 0.0, "step": 4596 }, { "epoch": 2.95, "learning_rate": 1.213605664673656e-08, "loss": 0.0, "step": 4597 }, { "epoch": 2.95, "learning_rate": 1.1796670122482535e-08, "loss": 0.0, "step": 4598 }, { "epoch": 2.96, "learning_rate": 1.1462093831232646e-08, "loss": 0.0, "step": 4599 }, { "epoch": 2.96, "learning_rate": 1.1132327934114672e-08, "loss": 0.0, "step": 4600 }, { "epoch": 2.96, "learning_rate": 1.0807372589940468e-08, "loss": 0.0, "step": 4601 }, { "epoch": 2.96, "learning_rate": 1.048722795520596e-08, "loss": 0.0, "step": 4602 }, { "epoch": 2.96, "learning_rate": 1.0171894184091147e-08, "loss": 0.0, "step": 4603 }, { "epoch": 2.96, "learning_rate": 9.861371428457889e-09, "loss": 0.0, "step": 4604 }, { "epoch": 2.96, "learning_rate": 9.555659837849896e-09, "loss": 0.0, "step": 4605 }, { "epoch": 2.96, "learning_rate": 9.254759559496063e-09, "loss": 0.0, "step": 4606 }, { "epoch": 2.96, "learning_rate": 8.95867073830714e-09, "loss": 0.0, "step": 4607 }, { "epoch": 2.96, "learning_rate": 8.667393516877953e-09, "loss": 0.0, "step": 4608 }, { "epoch": 2.96, "learning_rate": 8.38092803548407e-09, "loss": 0.0, "step": 4609 }, { "epoch": 2.96, "learning_rate": 8.099274432085135e-09, "loss": 0.0, "step": 4610 }, { "epoch": 2.96, "learning_rate": 7.822432842322648e-09, "loss": 0.0, "step": 4611 }, { "epoch": 2.96, "learning_rate": 7.550403399522177e-09, "loss": 0.0001, "step": 4612 }, { "epoch": 2.96, "learning_rate": 7.283186234690043e-09, "loss": 0.0, "step": 4613 }, { "epoch": 2.96, "learning_rate": 7.0207814765155255e-09, "loss": 0.0, "step": 4614 }, { "epoch": 2.97, "learning_rate": 6.763189251370872e-09, "loss": 0.0001, "step": 4615 }, { "epoch": 2.97, "learning_rate": 6.510409683310182e-09, "loss": 0.0, "step": 4616 }, { "epoch": 2.97, "learning_rate": 6.262442894070519e-09, "loss": 0.0, "step": 4617 }, { "epoch": 2.97, "learning_rate": 6.019289003069695e-09, "loss": 0.0, "step": 4618 }, { "epoch": 2.97, "learning_rate": 5.780948127409591e-09, "loss": 0.0, "step": 4619 }, { "epoch": 2.97, "learning_rate": 5.547420381871726e-09, "loss": 0.0, "step": 4620 }, { "epoch": 2.97, "learning_rate": 5.318705878921693e-09, "loss": 0.0, "step": 4621 }, { "epoch": 2.97, "learning_rate": 5.094804728706937e-09, "loss": 0.0, "step": 4622 }, { "epoch": 2.97, "learning_rate": 4.8757170390556495e-09, "loss": 0.0, "step": 4623 }, { "epoch": 2.97, "learning_rate": 4.661442915477876e-09, "loss": 0.0, "step": 4624 }, { "epoch": 2.97, "learning_rate": 4.451982461168847e-09, "loss": 0.0, "step": 4625 }, { "epoch": 2.97, "learning_rate": 4.2473357770000944e-09, "loss": 0.0, "step": 4626 }, { "epoch": 2.97, "learning_rate": 4.047502961529448e-09, "loss": 0.0, "step": 4627 }, { "epoch": 2.97, "learning_rate": 3.85248411099326e-09, "loss": 0.0, "step": 4628 }, { "epoch": 2.97, "learning_rate": 3.6622793193119567e-09, "loss": 0.0, "step": 4629 }, { "epoch": 2.98, "learning_rate": 3.4768886780867096e-09, "loss": 0.0, "step": 4630 }, { "epoch": 2.98, "learning_rate": 3.2963122765994336e-09, "loss": 0.0, "step": 4631 }, { "epoch": 2.98, "learning_rate": 3.1205502018150092e-09, "loss": 0.0, "step": 4632 }, { "epoch": 2.98, "learning_rate": 2.9496025383779492e-09, "loss": 0.0, "step": 4633 }, { "epoch": 2.98, "learning_rate": 2.7834693686146218e-09, "loss": 0.0, "step": 4634 }, { "epoch": 2.98, "learning_rate": 2.62215077253547e-09, "loss": 0.0, "step": 4635 }, { "epoch": 2.98, "learning_rate": 2.4656468278283496e-09, "loss": 0.0, "step": 4636 }, { "epoch": 2.98, "learning_rate": 2.3139576098651915e-09, "loss": 0.0, "step": 4637 }, { "epoch": 2.98, "learning_rate": 2.167083191697561e-09, "loss": 0.0, "step": 4638 }, { "epoch": 2.98, "learning_rate": 2.0250236440588766e-09, "loss": 0.0, "step": 4639 }, { "epoch": 2.98, "learning_rate": 1.887779035364412e-09, "loss": 0.0, "step": 4640 }, { "epoch": 2.98, "learning_rate": 1.7553494317101848e-09, "loss": 0.0, "step": 4641 }, { "epoch": 2.98, "learning_rate": 1.6277348968718466e-09, "loss": 0.0, "step": 4642 }, { "epoch": 2.98, "learning_rate": 1.504935492309123e-09, "loss": 0.0, "step": 4643 }, { "epoch": 2.98, "learning_rate": 1.3869512771602645e-09, "loss": 0.0, "step": 4644 }, { "epoch": 2.98, "learning_rate": 1.2737823082453748e-09, "loss": 0.0, "step": 4645 }, { "epoch": 2.99, "learning_rate": 1.1654286400653025e-09, "loss": 0.0, "step": 4646 }, { "epoch": 2.99, "learning_rate": 1.0618903248027501e-09, "loss": 0.0, "step": 4647 }, { "epoch": 2.99, "learning_rate": 9.631674123211643e-10, "loss": 0.0, "step": 4648 }, { "epoch": 2.99, "learning_rate": 8.692599501647359e-10, "loss": 0.0, "step": 4649 }, { "epoch": 2.99, "learning_rate": 7.801679835572895e-10, "loss": 0.0, "step": 4650 }, { "epoch": 2.99, "learning_rate": 6.958915554067247e-10, "loss": 0.0, "step": 4651 }, { "epoch": 2.99, "learning_rate": 6.164307062983543e-10, "loss": 0.0, "step": 4652 }, { "epoch": 2.99, "learning_rate": 5.417854744993456e-10, "loss": 0.0, "step": 4653 }, { "epoch": 2.99, "learning_rate": 4.719558959598303e-10, "loss": 0.0, "step": 4654 }, { "epoch": 2.99, "learning_rate": 4.069420043073535e-10, "loss": 0.0, "step": 4655 }, { "epoch": 2.99, "learning_rate": 3.467438308546456e-10, "loss": 0.0, "step": 4656 }, { "epoch": 2.99, "learning_rate": 2.9136140458962956e-10, "loss": 0.0, "step": 4657 }, { "epoch": 2.99, "learning_rate": 2.407947521865239e-10, "loss": 0.0, "step": 4658 }, { "epoch": 2.99, "learning_rate": 1.950438979958502e-10, "loss": 0.0, "step": 4659 }, { "epoch": 2.99, "learning_rate": 1.541088640510946e-10, "loss": 0.0, "step": 4660 }, { "epoch": 3.0, "learning_rate": 1.1798967006759754e-10, "loss": 0.0, "step": 4661 }, { "epoch": 3.0, "learning_rate": 8.668633343922317e-11, "loss": 0.0, "step": 4662 }, { "epoch": 3.0, "learning_rate": 6.019886924168994e-11, "loss": 0.0, "step": 4663 }, { "epoch": 3.0, "learning_rate": 3.8527290230350176e-11, "loss": 0.0, "step": 4664 }, { "epoch": 3.0, "learning_rate": 2.1671606842410543e-11, "loss": 0.0, "step": 4665 }, { "epoch": 3.0, "learning_rate": 9.631827195821786e-12, "loss": 0.0, "step": 4666 }, { "epoch": 3.0, "learning_rate": 2.407957089278767e-12, "loss": 0.0, "step": 4667 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.0, "step": 4668 }, { "epoch": 3.0, "step": 4668, "total_flos": 8.749105718541418e+18, "train_loss": 0.0005586616116672645, "train_runtime": 138680.0685, "train_samples_per_second": 2.154, "train_steps_per_second": 0.034 } ], "max_steps": 4668, "num_train_epochs": 3, "total_flos": 8.749105718541418e+18, "trial_name": null, "trial_params": null }