| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9945300845350572, | |
| "eval_steps": 500, | |
| "global_step": 2000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 1.7647, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 1.6789, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.2e-05, | |
| "loss": 1.7601, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.6000000000000003e-05, | |
| "loss": 1.7238, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2e-05, | |
| "loss": 1.7794, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.4e-05, | |
| "loss": 1.7026, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.8000000000000003e-05, | |
| "loss": 1.4233, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.2000000000000005e-05, | |
| "loss": 1.6603, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.6e-05, | |
| "loss": 1.2334, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4e-05, | |
| "loss": 1.368, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.4000000000000006e-05, | |
| "loss": 1.3571, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.8e-05, | |
| "loss": 1.31, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.2000000000000004e-05, | |
| "loss": 1.2953, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.6000000000000006e-05, | |
| "loss": 1.0765, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6e-05, | |
| "loss": 0.9163, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.400000000000001e-05, | |
| "loss": 1.0587, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.800000000000001e-05, | |
| "loss": 1.1125, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.2e-05, | |
| "loss": 1.1693, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.6e-05, | |
| "loss": 0.763, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8e-05, | |
| "loss": 1.081, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.4e-05, | |
| "loss": 0.7426, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.800000000000001e-05, | |
| "loss": 0.7081, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.200000000000001e-05, | |
| "loss": 0.8163, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.6e-05, | |
| "loss": 0.7577, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001, | |
| "loss": 0.8087, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00010400000000000001, | |
| "loss": 0.983, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00010800000000000001, | |
| "loss": 0.7019, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00011200000000000001, | |
| "loss": 0.9995, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000116, | |
| "loss": 0.7865, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00012, | |
| "loss": 0.6138, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000124, | |
| "loss": 0.7568, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00012800000000000002, | |
| "loss": 0.8195, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000132, | |
| "loss": 0.7583, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00013600000000000003, | |
| "loss": 0.7332, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00014, | |
| "loss": 0.7058, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000144, | |
| "loss": 0.7036, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000148, | |
| "loss": 0.6442, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000152, | |
| "loss": 0.9267, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015600000000000002, | |
| "loss": 0.7025, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016, | |
| "loss": 0.6675, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000164, | |
| "loss": 0.6326, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000168, | |
| "loss": 0.8249, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000172, | |
| "loss": 0.6033, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00017600000000000002, | |
| "loss": 0.7918, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00018, | |
| "loss": 0.6245, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00018400000000000003, | |
| "loss": 0.8179, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000188, | |
| "loss": 0.6853, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000192, | |
| "loss": 0.6894, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000196, | |
| "loss": 0.7317, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6921, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00020400000000000003, | |
| "loss": 0.7591, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00020800000000000001, | |
| "loss": 0.626, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00021200000000000003, | |
| "loss": 0.6851, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00021600000000000002, | |
| "loss": 0.761, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022000000000000003, | |
| "loss": 0.6841, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022400000000000002, | |
| "loss": 0.7289, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022799999999999999, | |
| "loss": 0.545, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000232, | |
| "loss": 0.761, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000236, | |
| "loss": 0.6678, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00024, | |
| "loss": 0.7483, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000244, | |
| "loss": 0.7225, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000248, | |
| "loss": 0.7489, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000252, | |
| "loss": 0.8577, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00025600000000000004, | |
| "loss": 0.795, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00026000000000000003, | |
| "loss": 0.6954, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000264, | |
| "loss": 0.8429, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000268, | |
| "loss": 0.5476, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00027200000000000005, | |
| "loss": 0.6681, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000276, | |
| "loss": 0.7104, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00028, | |
| "loss": 0.6646, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000284, | |
| "loss": 0.8192, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000288, | |
| "loss": 0.6286, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000292, | |
| "loss": 0.6914, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000296, | |
| "loss": 0.6344, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00030000000000000003, | |
| "loss": 0.6412, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000304, | |
| "loss": 0.8665, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000308, | |
| "loss": 0.6683, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00031200000000000005, | |
| "loss": 0.6472, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00031600000000000004, | |
| "loss": 0.6689, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00032, | |
| "loss": 0.6126, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000324, | |
| "loss": 0.8707, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000328, | |
| "loss": 0.7206, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000332, | |
| "loss": 0.7253, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000336, | |
| "loss": 0.5838, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00034, | |
| "loss": 0.7274, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000344, | |
| "loss": 0.6212, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000348, | |
| "loss": 0.8331, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00035200000000000005, | |
| "loss": 0.9408, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00035600000000000003, | |
| "loss": 0.713, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00036, | |
| "loss": 0.6373, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000364, | |
| "loss": 0.6545, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00036800000000000005, | |
| "loss": 0.6066, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00037200000000000004, | |
| "loss": 0.798, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000376, | |
| "loss": 0.5667, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00038, | |
| "loss": 0.7304, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000384, | |
| "loss": 0.7651, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000388, | |
| "loss": 0.7812, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000392, | |
| "loss": 0.7129, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039600000000000003, | |
| "loss": 0.7782, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004, | |
| "loss": 0.9215, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999997297421774, | |
| "loss": 0.6187, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999989189694399, | |
| "loss": 0.6336, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999975676839788, | |
| "loss": 0.7769, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039999567588944597, | |
| "loss": 0.6579, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999932435909542, | |
| "loss": 0.5559, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999902707950768, | |
| "loss": 0.7106, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999867575098482, | |
| "loss": 0.6364, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039998270374476315, | |
| "loss": 0.5283, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039997810951077745, | |
| "loss": 0.7988, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999729748203072, | |
| "loss": 0.5788, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999672996872294, | |
| "loss": 0.577, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999610841268815, | |
| "loss": 0.7407, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999543281560616, | |
| "loss": 0.8146, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039994703179302817, | |
| "loss": 0.6977, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999391950575002, | |
| "loss": 0.6826, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999308179706572, | |
| "loss": 0.6363, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039992190055513873, | |
| "loss": 0.8227, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999124428350449, | |
| "loss": 0.7266, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039990244483593593, | |
| "loss": 0.8752, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039989190658483215, | |
| "loss": 0.5351, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998808281102141, | |
| "loss": 0.6784, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998692094420221, | |
| "loss": 0.56, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998570506116567, | |
| "loss": 0.56, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998443516519779, | |
| "loss": 0.7041, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998311125973057, | |
| "loss": 0.4968, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039981733348341966, | |
| "loss": 0.6442, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039980301434755896, | |
| "loss": 0.755, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039978815522842217, | |
| "loss": 0.8611, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003997727561661672, | |
| "loss": 0.7475, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039975681720241124, | |
| "loss": 0.6555, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039974033838023057, | |
| "loss": 0.7254, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003997233197441605, | |
| "loss": 0.5879, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039970576134019524, | |
| "loss": 0.6784, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003996876632157877, | |
| "loss": 0.5117, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039966902541984956, | |
| "loss": 0.7618, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003996498480027509, | |
| "loss": 0.6151, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039963013101632013, | |
| "loss": 0.7617, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000399609874513844, | |
| "loss": 0.74, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003995890785500673, | |
| "loss": 0.7561, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003995677431811927, | |
| "loss": 0.7816, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039954586846488075, | |
| "loss": 0.7096, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003995234544602496, | |
| "loss": 0.5191, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003995005012278747, | |
| "loss": 0.6101, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039947700882978914, | |
| "loss": 0.5408, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003994529773294829, | |
| "loss": 0.7153, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039942840679190296, | |
| "loss": 0.7468, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003994032972834532, | |
| "loss": 0.6848, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003993776488719939, | |
| "loss": 0.6419, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039935146162684205, | |
| "loss": 0.6881, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039932473561877054, | |
| "loss": 0.7313, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003992974709200087, | |
| "loss": 0.5955, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039926966760424137, | |
| "loss": 0.6141, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003992413257466093, | |
| "loss": 0.5981, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003992124454237085, | |
| "loss": 0.6241, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003991830267135903, | |
| "loss": 0.5943, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039915306969576106, | |
| "loss": 0.6664, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039912257445118195, | |
| "loss": 0.682, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039909154106226885, | |
| "loss": 0.7231, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003990599696128919, | |
| "loss": 0.5938, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039902786018837527, | |
| "loss": 0.8645, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039899521287549736, | |
| "loss": 0.7242, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039896202776248995, | |
| "loss": 0.7245, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003989283049390386, | |
| "loss": 0.7995, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039889404449628167, | |
| "loss": 0.6952, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039885924652681074, | |
| "loss": 0.5638, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003988239111246701, | |
| "loss": 0.8103, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039878803838535643, | |
| "loss": 0.6228, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039875162840581857, | |
| "loss": 0.9623, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003987146812844574, | |
| "loss": 0.7025, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003986771971211253, | |
| "loss": 0.6623, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039863917601712633, | |
| "loss": 0.7049, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003986006180752153, | |
| "loss": 0.6288, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003985615233995982, | |
| "loss": 0.6368, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003985218920959314, | |
| "loss": 0.5796, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039848172427132153, | |
| "loss": 0.6072, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039844102003432547, | |
| "loss": 0.7546, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039839977949494946, | |
| "loss": 0.7092, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003983580027646492, | |
| "loss": 0.6492, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003983156899563298, | |
| "loss": 0.3913, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003982728411843448, | |
| "loss": 0.5999, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039822945656449633, | |
| "loss": 0.6012, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003981855362140348, | |
| "loss": 0.5822, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003981410802516583, | |
| "loss": 0.5619, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039809608879751266, | |
| "loss": 0.6232, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039805056197319075, | |
| "loss": 0.6429, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003980044999017323, | |
| "loss": 0.6316, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039795790270762376, | |
| "loss": 0.7693, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003979107705167977, | |
| "loss": 0.7608, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003978631034566325, | |
| "loss": 0.8001, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039781490165595213, | |
| "loss": 0.7098, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003977661652450257, | |
| "loss": 0.7594, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039771689435556725, | |
| "loss": 0.6571, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039766708912073514, | |
| "loss": 0.658, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039761674967513197, | |
| "loss": 0.6272, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000397565876154804, | |
| "loss": 0.7159, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039751446869724087, | |
| "loss": 0.5958, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003974625274413753, | |
| "loss": 0.6282, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039741005252758255, | |
| "loss": 0.5312, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003973570440976803, | |
| "loss": 0.7235, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003973035022949278, | |
| "loss": 0.7214, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003972494272640261, | |
| "loss": 0.5673, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039719481915111713, | |
| "loss": 0.6782, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003971396781037836, | |
| "loss": 0.8209, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003970840042710485, | |
| "loss": 0.583, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003970277978033748, | |
| "loss": 0.7465, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039697105885266474, | |
| "loss": 0.5445, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039691378757225984, | |
| "loss": 0.638, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003968559841169403, | |
| "loss": 0.6875, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039679764864292426, | |
| "loss": 0.5552, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003967387813078681, | |
| "loss": 0.6606, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039667938227086534, | |
| "loss": 0.6827, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003966194516924465, | |
| "loss": 0.6292, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039655898973457867, | |
| "loss": 0.6688, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039649799656066504, | |
| "loss": 0.6292, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039643647233554444, | |
| "loss": 0.7615, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003963744172254908, | |
| "loss": 0.6115, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039631183139821307, | |
| "loss": 0.6144, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039624871502285427, | |
| "loss": 0.61, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003961850682699913, | |
| "loss": 0.7783, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039612089131163455, | |
| "loss": 0.7403, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003960561843212272, | |
| "loss": 0.6587, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000395990947473645, | |
| "loss": 0.5576, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039592518094519566, | |
| "loss": 0.6095, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003958588849136183, | |
| "loss": 0.65, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003957920595580832, | |
| "loss": 0.5773, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039572470505919107, | |
| "loss": 0.5764, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039565682159897276, | |
| "loss": 0.5655, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039558840936088855, | |
| "loss": 0.7379, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003955194685298279, | |
| "loss": 0.6677, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003954499992921088, | |
| "loss": 0.5464, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003953800018354774, | |
| "loss": 0.5902, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003953094763491071, | |
| "loss": 0.6901, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039523842302359875, | |
| "loss": 0.5261, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003951668420509793, | |
| "loss": 0.6494, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003950947336247022, | |
| "loss": 0.6419, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039502209793964584, | |
| "loss": 0.5659, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000394948935192114, | |
| "loss": 0.7916, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003948752455798348, | |
| "loss": 0.7088, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003948010293019599, | |
| "loss": 0.7528, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039472628655906486, | |
| "loss": 0.847, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003946510175531477, | |
| "loss": 0.7424, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039457522248762883, | |
| "loss": 0.724, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039449890156735025, | |
| "loss": 0.5163, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039442205499857526, | |
| "loss": 0.5648, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003943446829889879, | |
| "loss": 0.6173, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039426678574769173, | |
| "loss": 0.585, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003941883634852104, | |
| "loss": 0.7929, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003941094164134862, | |
| "loss": 0.705, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003940299447458796, | |
| "loss": 0.5889, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003939499486971692, | |
| "loss": 0.7188, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003938694284835505, | |
| "loss": 0.6511, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003937883843226355, | |
| "loss": 0.6225, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039370681643345265, | |
| "loss": 0.6195, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039362472503644546, | |
| "loss": 0.597, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003935421103534723, | |
| "loss": 0.6274, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003934589726078059, | |
| "loss": 0.7368, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039337531202413245, | |
| "loss": 0.7585, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003932911288285512, | |
| "loss": 0.5838, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039320642324857394, | |
| "loss": 0.5379, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039312119551312405, | |
| "loss": 0.7431, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039303544585253606, | |
| "loss": 0.4641, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039294917449855533, | |
| "loss": 0.701, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039286238168433676, | |
| "loss": 0.7539, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003927750676444449, | |
| "loss": 0.5141, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003926872326148526, | |
| "loss": 0.6841, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039259887683294097, | |
| "loss": 0.5684, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003925100005374984, | |
| "loss": 0.7382, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039242060396872013, | |
| "loss": 0.7834, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003923306873682073, | |
| "loss": 0.5423, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003922402509789665, | |
| "loss": 0.5786, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003921492950454093, | |
| "loss": 0.6305, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039205781981335107, | |
| "loss": 0.6927, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039196582553001087, | |
| "loss": 0.7202, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039187331244401044, | |
| "loss": 0.6246, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003917802808053735, | |
| "loss": 0.4526, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039168673086552563, | |
| "loss": 0.7272, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003915926628772925, | |
| "loss": 0.7071, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003914980770949005, | |
| "loss": 0.6837, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000391402973773975, | |
| "loss": 0.6223, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039130735317154005, | |
| "loss": 0.6601, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039121121554601797, | |
| "loss": 0.7047, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003911145611572282, | |
| "loss": 0.672, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039101739026638666, | |
| "loss": 0.8582, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003909197031361054, | |
| "loss": 0.6171, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003908215000303915, | |
| "loss": 0.4937, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039072278121464655, | |
| "loss": 0.5383, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003906235469556658, | |
| "loss": 0.5295, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003905237975216377, | |
| "loss": 0.6685, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039042353318214284, | |
| "loss": 0.5457, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039032275420815347, | |
| "loss": 0.6101, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003902214608720326, | |
| "loss": 0.6333, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003901196534475334, | |
| "loss": 0.6954, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00039001733220979845, | |
| "loss": 0.7586, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003899144974353589, | |
| "loss": 0.4571, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038981114940213373, | |
| "loss": 0.651, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003897072883894291, | |
| "loss": 0.7007, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038960291467793757, | |
| "loss": 0.6685, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003894980285497371, | |
| "loss": 0.6625, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038939263028829087, | |
| "loss": 0.5978, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003892867201784458, | |
| "loss": 0.5614, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003891802985064323, | |
| "loss": 0.495, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003890733655598632, | |
| "loss": 0.6905, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038896592162773313, | |
| "loss": 0.564, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038885796700041783, | |
| "loss": 0.6111, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003887495019696731, | |
| "loss": 0.6665, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003886405268286341, | |
| "loss": 0.6585, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003885310418718148, | |
| "loss": 0.3871, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003884210473951067, | |
| "loss": 0.6848, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038831054369577856, | |
| "loss": 0.772, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003881995310724753, | |
| "loss": 0.5701, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038808800982521717, | |
| "loss": 0.5739, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003879759802553991, | |
| "loss": 0.5448, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003878634426657898, | |
| "loss": 0.6247, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003877503973605308, | |
| "loss": 0.8148, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003876368446451359, | |
| "loss": 0.498, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003875227848264902, | |
| "loss": 0.5991, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038740821821284943, | |
| "loss": 0.5773, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003872931451138387, | |
| "loss": 0.6099, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000387177565840452, | |
| "loss": 0.6782, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038706148070505146, | |
| "loss": 0.6256, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038694489002136625, | |
| "loss": 0.6732, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003868277941044917, | |
| "loss": 0.8662, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003867101932708888, | |
| "loss": 0.6355, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000386592087838383, | |
| "loss": 0.5827, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038647347812616336, | |
| "loss": 0.5385, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038635436445478205, | |
| "loss": 0.7274, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000386234747146153, | |
| "loss": 0.7166, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003861146265235513, | |
| "loss": 0.5965, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038599400291161243, | |
| "loss": 0.6544, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038587287663633107, | |
| "loss": 0.5942, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038575124802506044, | |
| "loss": 0.686, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003856291174065115, | |
| "loss": 0.6384, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003855064851107517, | |
| "loss": 0.5287, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003853833514692044, | |
| "loss": 0.5819, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003852597168146479, | |
| "loss": 0.5371, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038513558148121457, | |
| "loss": 0.6539, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038501094580438986, | |
| "loss": 0.7534, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038488581012101136, | |
| "loss": 0.69, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003847601747692681, | |
| "loss": 0.5101, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003846340400886995, | |
| "loss": 0.5204, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038450740642019434, | |
| "loss": 0.7896, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038438027410599, | |
| "loss": 0.5985, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038425264348967156, | |
| "loss": 0.6492, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038412451491617064, | |
| "loss": 0.6247, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038399588873176487, | |
| "loss": 0.5555, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038386676528407647, | |
| "loss": 0.7643, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003837371449220717, | |
| "loss": 0.6722, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038360702799605974, | |
| "loss": 0.6424, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038347641485769176, | |
| "loss": 0.6874, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003833453058599599, | |
| "loss": 0.5627, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038321370135719664, | |
| "loss": 0.5754, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003830816017050733, | |
| "loss": 0.6212, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003829490072605996, | |
| "loss": 0.604, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038281591838212236, | |
| "loss": 0.6161, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003826823354293247, | |
| "loss": 0.6257, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038254825876322496, | |
| "loss": 0.6736, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038241368874617594, | |
| "loss": 0.512, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038227862574186355, | |
| "loss": 0.6189, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038214307011530613, | |
| "loss": 0.6199, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003820070222328533, | |
| "loss": 0.7591, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038187048246218525, | |
| "loss": 0.6766, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038173345117231125, | |
| "loss": 0.6218, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038159592873356925, | |
| "loss": 0.6951, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003814579155176242, | |
| "loss": 0.5625, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038131941189746777, | |
| "loss": 0.6293, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003811804182474167, | |
| "loss": 0.6459, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003810409349431123, | |
| "loss": 0.7419, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000380900962361519, | |
| "loss": 0.8274, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003807605008809239, | |
| "loss": 0.5752, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038061955088093485, | |
| "loss": 0.6657, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038047811274248035, | |
| "loss": 0.751, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038033618684780806, | |
| "loss": 0.5926, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003801937735804838, | |
| "loss": 0.6257, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003800508733253906, | |
| "loss": 0.574, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037990748646872757, | |
| "loss": 0.523, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037976361339800885, | |
| "loss": 0.5428, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037961925450206275, | |
| "loss": 0.5206, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037947441017103037, | |
| "loss": 0.6535, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003793290807963649, | |
| "loss": 0.6768, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037918326677083037, | |
| "loss": 0.6109, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037903696848850057, | |
| "loss": 0.5639, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003788901863447581, | |
| "loss": 0.6201, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003787429207362931, | |
| "loss": 0.7355, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003785951720611024, | |
| "loss": 0.7389, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003784469407184884, | |
| "loss": 0.5913, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003782982271090579, | |
| "loss": 0.7367, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003781490316347211, | |
| "loss": 0.5414, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003779993546986903, | |
| "loss": 0.5817, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037784919670547925, | |
| "loss": 0.5148, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003776985580609017, | |
| "loss": 0.6823, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003775474391720703, | |
| "loss": 0.7489, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037739584044739564, | |
| "loss": 0.5502, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037724376229658516, | |
| "loss": 0.7147, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.000377091205130642, | |
| "loss": 0.6231, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003769381693618638, | |
| "loss": 0.6253, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037678465540384173, | |
| "loss": 0.7029, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003766306636714592, | |
| "loss": 0.7591, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037647619458089094, | |
| "loss": 0.5443, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003763212485496018, | |
| "loss": 0.6599, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037616582599634557, | |
| "loss": 0.8092, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003760099273411637, | |
| "loss": 0.6074, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003758535530053847, | |
| "loss": 0.5847, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003756967034116223, | |
| "loss": 0.6226, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003755393789837749, | |
| "loss": 0.8133, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.000375381580147024, | |
| "loss": 0.7402, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037522330732783326, | |
| "loss": 0.5815, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037506456095394745, | |
| "loss": 0.5198, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003749053414543911, | |
| "loss": 0.4467, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037474564925946727, | |
| "loss": 0.6378, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003745854848007566, | |
| "loss": 0.5427, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003744248485111161, | |
| "loss": 0.5876, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003742637408246779, | |
| "loss": 0.6719, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037410216217684814, | |
| "loss": 0.5377, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037394011300430574, | |
| "loss": 0.6441, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037377759374500135, | |
| "loss": 0.6824, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037361460483815583, | |
| "loss": 0.6374, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037345114672425955, | |
| "loss": 0.5061, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037328721984507084, | |
| "loss": 0.6868, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003731228246436149, | |
| "loss": 1.1785, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037295796156418265, | |
| "loss": 0.5278, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037279263105232936, | |
| "loss": 0.6359, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003726268335548738, | |
| "loss": 0.6748, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037246056951989665, | |
| "loss": 0.7672, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037229383939673934, | |
| "loss": 0.6781, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037229383939673934, | |
| "loss": 1.0859, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003721266436360033, | |
| "loss": 1.4568, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037195898268954793, | |
| "loss": 0.6202, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003717908570104902, | |
| "loss": 0.5115, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003716222670532028, | |
| "loss": 0.6034, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003714532132733134, | |
| "loss": 0.6322, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037128369612770293, | |
| "loss": 0.5529, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003711137160745049, | |
| "loss": 0.5303, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003709432735731035, | |
| "loss": 0.4894, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003707723690841331, | |
| "loss": 0.8528, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003706010030694764, | |
| "loss": 0.6607, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003704291759922634, | |
| "loss": 0.6919, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00037025688831687027, | |
| "loss": 0.5845, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003700841405089179, | |
| "loss": 0.4882, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003699109330352708, | |
| "loss": 0.6078, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003697372663640357, | |
| "loss": 0.5976, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003695631409645603, | |
| "loss": 0.7385, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003693885573074322, | |
| "loss": 0.5573, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036921351586447735, | |
| "loss": 0.8106, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000369038017108759, | |
| "loss": 0.5469, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003688620615145761, | |
| "loss": 0.5426, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036868564955746267, | |
| "loss": 0.5748, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003685087817141856, | |
| "loss": 0.7693, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003683314584627441, | |
| "loss": 0.6121, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003681536802823683, | |
| "loss": 0.6382, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036797544765351743, | |
| "loss": 0.621, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003677967610578792, | |
| "loss": 0.6809, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003676176209783681, | |
| "loss": 0.6904, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003674380278991243, | |
| "loss": 0.7222, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000367257982305512, | |
| "loss": 0.6531, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036707748468411863, | |
| "loss": 0.6264, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000366896535522753, | |
| "loss": 0.6397, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003667151353104445, | |
| "loss": 0.7211, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003665332845374413, | |
| "loss": 0.612, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036635098369520937, | |
| "loss": 0.7429, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036616823327643104, | |
| "loss": 0.6543, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003659850337750035, | |
| "loss": 0.682, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036580138568603787, | |
| "loss": 0.573, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036561728950585743, | |
| "loss": 0.6211, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036543274573199645, | |
| "loss": 0.714, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000365247754863199, | |
| "loss": 0.5405, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003650623173994173, | |
| "loss": 0.6542, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036487643384181066, | |
| "loss": 0.5903, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003646901046927439, | |
| "loss": 0.6483, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003645033304557861, | |
| "loss": 0.6035, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003643161116357093, | |
| "loss": 0.7037, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036412844873848704, | |
| "loss": 0.4548, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003639403422712928, | |
| "loss": 0.6608, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036375179274249924, | |
| "loss": 0.7457, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003635628006616761, | |
| "loss": 0.7015, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036337336653958927, | |
| "loss": 0.6911, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003631834908881993, | |
| "loss": 0.7825, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036299317422065997, | |
| "loss": 0.8098, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.000362802417051317, | |
| "loss": 0.5504, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003626112198957065, | |
| "loss": 0.593, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036241958327055376, | |
| "loss": 0.7801, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036222750769377186, | |
| "loss": 0.7113, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003620349936844599, | |
| "loss": 0.6126, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036184204176290224, | |
| "loss": 0.6875, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036164865245056634, | |
| "loss": 0.5044, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.000361454826270102, | |
| "loss": 0.5396, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003612605637453397, | |
| "loss": 0.7856, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003610658654012892, | |
| "loss": 0.5603, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036087073176413773, | |
| "loss": 0.6771, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036067516336124947, | |
| "loss": 0.7139, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003604791607211632, | |
| "loss": 0.5865, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036028272437359135, | |
| "loss": 0.5851, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003600858548494186, | |
| "loss": 0.7633, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003598885526807003, | |
| "loss": 0.7342, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003596908184006608, | |
| "loss": 0.6061, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035949265254369263, | |
| "loss": 0.6854, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035929405564535445, | |
| "loss": 0.6038, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003590950282423699, | |
| "loss": 0.6915, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003588955708726261, | |
| "loss": 0.6072, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003586956840751723, | |
| "loss": 0.6412, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003584953683902181, | |
| "loss": 0.7201, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035829462435913237, | |
| "loss": 0.6801, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003580934525244415, | |
| "loss": 0.7302, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003578918534298281, | |
| "loss": 0.7046, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035768982762012956, | |
| "loss": 0.746, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003574873756413364, | |
| "loss": 0.6652, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035728449804059095, | |
| "loss": 0.4887, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003570811953661858, | |
| "loss": 0.71, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003568774681675623, | |
| "loss": 0.7497, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035667331699530913, | |
| "loss": 0.529, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003564687424011608, | |
| "loss": 0.658, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003562637449379962, | |
| "loss": 0.6412, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035605832515983707, | |
| "loss": 0.774, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035585248362184627, | |
| "loss": 0.5033, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035564622088032677, | |
| "loss": 0.8501, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003554395374927197, | |
| "loss": 0.606, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035523243401760314, | |
| "loss": 0.5026, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003550249110146904, | |
| "loss": 0.6111, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003548169690448286, | |
| "loss": 0.5991, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035460860866999723, | |
| "loss": 0.7183, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003543998304533066, | |
| "loss": 0.5178, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003541906349589959, | |
| "loss": 0.6164, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003539810227524326, | |
| "loss": 0.522, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035377099440010994, | |
| "loss": 0.6188, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035356055046964594, | |
| "loss": 0.7164, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003533496915297819, | |
| "loss": 0.5172, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003531384181503805, | |
| "loss": 0.6411, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003529267309024246, | |
| "loss": 0.5529, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035271463035801563, | |
| "loss": 0.5118, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035250211709037183, | |
| "loss": 0.7478, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003522891916738269, | |
| "loss": 0.6035, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035207585468382846, | |
| "loss": 0.6822, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003518621066969365, | |
| "loss": 0.5875, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00035164794829082153, | |
| "loss": 0.7102, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003514333800442635, | |
| "loss": 0.4673, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003512184025371498, | |
| "loss": 0.6669, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00035100301635047405, | |
| "loss": 0.5954, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003507872220663342, | |
| "loss": 0.7105, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003505710202679312, | |
| "loss": 0.5827, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00035035441153956733, | |
| "loss": 0.4864, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003501373964666447, | |
| "loss": 0.5347, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003499199756356633, | |
| "loss": 0.5432, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034970214963422024, | |
| "loss": 0.6944, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003494839190510071, | |
| "loss": 0.7033, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003492652844758092, | |
| "loss": 0.635, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034904624649950347, | |
| "loss": 0.6312, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003488268057140573, | |
| "loss": 0.6411, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003486069627125265, | |
| "loss": 0.5184, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003483867180890541, | |
| "loss": 0.7392, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034816607243886834, | |
| "loss": 0.5188, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034794502635828123, | |
| "loss": 0.5766, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003477235804446873, | |
| "loss": 0.8635, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003475017352965614, | |
| "loss": 0.7471, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003472794915134573, | |
| "loss": 0.8053, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034705684969600626, | |
| "loss": 0.687, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003468338104459153, | |
| "loss": 0.7805, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003466103743659653, | |
| "loss": 0.6063, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034638654206000985, | |
| "loss": 0.5344, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034616231413297317, | |
| "loss": 0.5276, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003459376911908488, | |
| "loss": 0.7777, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034571267384069795, | |
| "loss": 0.5223, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034548726269064745, | |
| "loss": 0.6222, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003452614583498886, | |
| "loss": 0.7613, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003450352614286753, | |
| "loss": 0.5035, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003448086725383225, | |
| "loss": 0.6905, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003445816922912043, | |
| "loss": 0.6635, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034435432130075255, | |
| "loss": 0.6324, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034412656018145525, | |
| "loss": 0.6451, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003438984095488546, | |
| "loss": 0.622, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034366987001954556, | |
| "loss": 0.5876, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.000343440942211174, | |
| "loss": 0.7518, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003432116267424353, | |
| "loss": 0.5941, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034298192423307236, | |
| "loss": 0.6457, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003427518353038743, | |
| "loss": 0.6172, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00034252136057667446, | |
| "loss": 0.5411, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003422905006743487, | |
| "loss": 0.4871, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00034205925622081396, | |
| "loss": 0.5874, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003418276278410266, | |
| "loss": 0.8631, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00034159561616098046, | |
| "loss": 0.6571, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00034136322180770503, | |
| "loss": 0.5722, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003411304454092644, | |
| "loss": 0.5915, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.000340897287594755, | |
| "loss": 0.8315, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003406637489943039, | |
| "loss": 0.7307, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003404298302390677, | |
| "loss": 0.9096, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003401955319612299, | |
| "loss": 0.6186, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033996085479400006, | |
| "loss": 0.5719, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003397257993716115, | |
| "loss": 0.7004, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003394903663293199, | |
| "loss": 0.5368, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033925455630340155, | |
| "loss": 0.5887, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003390183699311515, | |
| "loss": 0.5134, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003387818078508817, | |
| "loss": 0.5697, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033854487070191994, | |
| "loss": 0.4265, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033830755912460717, | |
| "loss": 0.6071, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003380698737602966, | |
| "loss": 0.5267, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003378318152513516, | |
| "loss": 0.6264, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033759338424114375, | |
| "loss": 0.5747, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003373545813740516, | |
| "loss": 0.5607, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033711540729545856, | |
| "loss": 0.5068, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033687586265175135, | |
| "loss": 0.6233, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033663594809031796, | |
| "loss": 0.4596, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003363956642595464, | |
| "loss": 0.7301, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003361550118088225, | |
| "loss": 0.5716, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003359139913885282, | |
| "loss": 0.4822, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003356726036500402, | |
| "loss": 0.5038, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033543084924572773, | |
| "loss": 0.7338, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003351887288289509, | |
| "loss": 0.718, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033494624305405906, | |
| "loss": 0.7541, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033470339257638915, | |
| "loss": 0.6289, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003344601780522634, | |
| "loss": 0.7546, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003342166001389881, | |
| "loss": 0.5442, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003339726594948518, | |
| "loss": 0.5831, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033372835677912294, | |
| "loss": 0.6044, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003334836926520489, | |
| "loss": 0.5945, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003332386677748535, | |
| "loss": 0.6161, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003329932828097357, | |
| "loss": 0.7493, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003327475384198676, | |
| "loss": 0.6404, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033250143526939254, | |
| "loss": 0.542, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003322549740234236, | |
| "loss": 0.6989, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033200815534804157, | |
| "loss": 0.6645, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033176097991029316, | |
| "loss": 0.5754, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003315134483781894, | |
| "loss": 0.5178, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003312655614207036, | |
| "loss": 0.5631, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033101731970776977, | |
| "loss": 0.5439, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003307687239102803, | |
| "loss": 0.5434, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033051977470008505, | |
| "loss": 0.5761, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033027047274998854, | |
| "loss": 0.3953, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033002081873374886, | |
| "loss": 0.4845, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032977081332607555, | |
| "loss": 0.6734, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032952045720262773, | |
| "loss": 0.7418, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032926975104001244, | |
| "loss": 0.751, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003290186955157827, | |
| "loss": 0.5736, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032876729130843564, | |
| "loss": 0.6946, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003285155390974109, | |
| "loss": 0.6579, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003282634395630884, | |
| "loss": 0.463, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000328010993386787, | |
| "loss": 0.5171, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032775820125076214, | |
| "loss": 0.696, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003275050638382044, | |
| "loss": 0.6108, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032725158183323727, | |
| "loss": 0.6624, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000326997755920916, | |
| "loss": 0.5566, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003267435867872247, | |
| "loss": 0.5459, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032648907511907547, | |
| "loss": 0.618, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032623422160430593, | |
| "loss": 0.6795, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032597902693167776, | |
| "loss": 0.7755, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032572349179087445, | |
| "loss": 0.6377, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032546761687249966, | |
| "loss": 0.5181, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032521140286807543, | |
| "loss": 0.5466, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032495485047004017, | |
| "loss": 0.6179, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032469796037174674, | |
| "loss": 0.5162, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003244407332674608, | |
| "loss": 0.6455, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003241831698523586, | |
| "loss": 0.5336, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003239252708225256, | |
| "loss": 0.6631, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003236670368749539, | |
| "loss": 0.5471, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003234084687075411, | |
| "loss": 0.726, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003231495670190879, | |
| "loss": 0.6812, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032289033250929624, | |
| "loss": 0.4997, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003226307658787677, | |
| "loss": 0.742, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003223708678290015, | |
| "loss": 0.6021, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032211063906239243, | |
| "loss": 0.6174, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032185008028222893, | |
| "loss": 0.4748, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003215891921926917, | |
| "loss": 0.5445, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032132797549885107, | |
| "loss": 0.6354, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032106643090666566, | |
| "loss": 0.5978, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003208045591229802, | |
| "loss": 0.6035, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003205423608555235, | |
| "loss": 0.6836, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003202798368129071, | |
| "loss": 0.5718, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003200169877046227, | |
| "loss": 0.6273, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031975381424104046, | |
| "loss": 0.6191, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031949031713340744, | |
| "loss": 0.7143, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031922649709384507, | |
| "loss": 0.5616, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003189623548353477, | |
| "loss": 0.7009, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031869789107178044, | |
| "loss": 0.6305, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003184331065178772, | |
| "loss": 0.7242, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031816800188923906, | |
| "loss": 0.5754, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.000317902577902332, | |
| "loss": 0.6522, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031763683527448504, | |
| "loss": 0.6287, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031737077472388854, | |
| "loss": 0.6262, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003171043969695918, | |
| "loss": 0.5627, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003168377027315017, | |
| "loss": 0.6852, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003165706927303802, | |
| "loss": 0.4525, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003163033676878426, | |
| "loss": 0.5041, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000316035728326356, | |
| "loss": 0.7958, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031576777536923654, | |
| "loss": 0.7252, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000315499509540648, | |
| "loss": 0.4905, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003152309315655999, | |
| "loss": 0.5408, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031496204216994517, | |
| "loss": 0.5442, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003146928420803784, | |
| "loss": 0.7182, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031442333202443397, | |
| "loss": 0.6432, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003141535127304838, | |
| "loss": 0.638, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003138833849277357, | |
| "loss": 0.6878, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031361294934623114, | |
| "loss": 0.6994, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031334220671684355, | |
| "loss": 0.5861, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003130711577712759, | |
| "loss": 0.5591, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003127998032420593, | |
| "loss": 0.704, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003125281438625505, | |
| "loss": 0.7457, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003122561803669303, | |
| "loss": 0.6541, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003119839134902013, | |
| "loss": 0.6377, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031171134396818596, | |
| "loss": 0.6298, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003114384725375248, | |
| "loss": 0.568, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031116529993567424, | |
| "loss": 0.6788, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003108918269009046, | |
| "loss": 0.5118, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003106180541722981, | |
| "loss": 0.5467, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031034398248974693, | |
| "loss": 0.6975, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031006961259395126, | |
| "loss": 0.7416, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030979494522641733, | |
| "loss": 0.639, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030951998112945504, | |
| "loss": 0.6382, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003092447210461764, | |
| "loss": 0.6499, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030896916572049335, | |
| "loss": 0.6502, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003086933158971157, | |
| "loss": 0.5365, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003084171723215492, | |
| "loss": 0.727, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030814073574009346, | |
| "loss": 0.6185, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003078640068998399, | |
| "loss": 0.6577, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003075869865486699, | |
| "loss": 0.654, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003073096754352526, | |
| "loss": 0.5244, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003070320743090431, | |
| "loss": 0.7939, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030675418392028, | |
| "loss": 0.5754, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003064760050199839, | |
| "loss": 0.7041, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.000306197538359955, | |
| "loss": 0.6668, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003059187846927713, | |
| "loss": 0.5769, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030563974477178636, | |
| "loss": 0.6367, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030536041935112735, | |
| "loss": 0.5151, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003050808091856931, | |
| "loss": 0.7165, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030480091503115196, | |
| "loss": 0.58, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030452073764393976, | |
| "loss": 0.6932, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003042402777812579, | |
| "loss": 0.5797, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030395953620107087, | |
| "loss": 0.6333, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003036785136621051, | |
| "loss": 0.4602, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003033972109238456, | |
| "loss": 0.6508, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030311562874653526, | |
| "loss": 0.5522, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030283376789117195, | |
| "loss": 0.6039, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003025516291195065, | |
| "loss": 0.5461, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003022692131940413, | |
| "loss": 0.5432, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003019865208780272, | |
| "loss": 0.6246, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003017035529354624, | |
| "loss": 0.6223, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030142031013109, | |
| "loss": 0.5961, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030113679323039567, | |
| "loss": 0.6369, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0003008530029996061, | |
| "loss": 0.6262, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00030056894020568666, | |
| "loss": 0.6475, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0003002846056163391, | |
| "loss": 0.5914, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00030000000000000003, | |
| "loss": 0.6132, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029971512412583834, | |
| "loss": 0.5984, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029942997876375335, | |
| "loss": 0.5443, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029914456468437266, | |
| "loss": 0.7574, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002988588826590503, | |
| "loss": 0.695, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029857293345986424, | |
| "loss": 0.4555, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029828671785961445, | |
| "loss": 0.4875, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.000298000236631821, | |
| "loss": 0.5362, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029771349055072196, | |
| "loss": 0.8078, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002974264803912709, | |
| "loss": 0.5168, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002971392069291353, | |
| "loss": 0.553, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029685167094069416, | |
| "loss": 0.6403, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029656387320303593, | |
| "loss": 0.6318, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029627581449395663, | |
| "loss": 0.6543, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029598749559195733, | |
| "loss": 0.6347, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029569891727624243, | |
| "loss": 0.6335, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002954100803267174, | |
| "loss": 0.5285, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002951209855239868, | |
| "loss": 0.4819, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029483163364935186, | |
| "loss": 0.6122, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029454202548480856, | |
| "loss": 0.624, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002942521618130459, | |
| "loss": 0.6652, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002939620434174428, | |
| "loss": 0.4588, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029367167108206716, | |
| "loss": 0.5946, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002933810455916728, | |
| "loss": 0.5821, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002930901677316979, | |
| "loss": 0.6065, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029279903828826264, | |
| "loss": 0.7857, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029250765804816713, | |
| "loss": 0.4992, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002922160277988891, | |
| "loss": 0.6143, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002919241483285824, | |
| "loss": 0.5624, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002916320204260738, | |
| "loss": 0.5454, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002913396448808621, | |
| "loss": 0.7938, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002910470224831149, | |
| "loss": 0.6054, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029075415402366714, | |
| "loss": 0.6346, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002904610402940189, | |
| "loss": 0.6516, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029016768208633276, | |
| "loss": 0.6173, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00028987408019343227, | |
| "loss": 0.6376, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002895802354087996, | |
| "loss": 0.6175, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028928614852657324, | |
| "loss": 0.7132, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.000288991820341546, | |
| "loss": 0.6448, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002886972516491627, | |
| "loss": 0.482, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002884024432455184, | |
| "loss": 0.5998, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002881073959273559, | |
| "loss": 0.4999, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002878121104920636, | |
| "loss": 0.6943, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028751658773767345, | |
| "loss": 0.6817, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028722082846285884, | |
| "loss": 0.7338, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002869248334669324, | |
| "loss": 0.6351, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002866286035498437, | |
| "loss": 0.6638, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002863321395121773, | |
| "loss": 0.681, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002860354421551504, | |
| "loss": 0.5534, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028573851228061084, | |
| "loss": 0.6712, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028544135069103485, | |
| "loss": 0.6651, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002851439581895249, | |
| "loss": 0.6093, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028484633557980744, | |
| "loss": 0.5459, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002845484836662309, | |
| "loss": 0.5464, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028425040325376333, | |
| "loss": 0.6008, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002839520951479904, | |
| "loss": 0.6463, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.000283653560155113, | |
| "loss": 0.7204, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002833547990819455, | |
| "loss": 0.6987, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002830558127359129, | |
| "loss": 0.7399, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028275660192504925, | |
| "loss": 0.6109, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028245716745799507, | |
| "loss": 0.6211, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002821575101439957, | |
| "loss": 0.6248, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028185763079289816, | |
| "loss": 0.549, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028155753021515004, | |
| "loss": 0.6633, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002812572092217966, | |
| "loss": 0.7503, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028095666862447877, | |
| "loss": 0.7072, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028065590923543113, | |
| "loss": 0.4936, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028035493186747925, | |
| "loss": 0.4967, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002800537373340382, | |
| "loss": 0.6648, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002797523264491097, | |
| "loss": 0.7149, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027945070002728027, | |
| "loss": 0.7318, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002791488588837189, | |
| "loss": 0.4862, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002788468038341748, | |
| "loss": 0.656, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002785445356949756, | |
| "loss": 0.573, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002782420552830243, | |
| "loss": 0.6089, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002779393634157981, | |
| "loss": 0.5132, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002776364609113454, | |
| "loss": 0.612, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002773333485882838, | |
| "loss": 0.5797, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002770300272657981, | |
| "loss": 0.5597, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002767264977636381, | |
| "loss": 0.7143, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002764227609021158, | |
| "loss": 0.4842, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027611881750210394, | |
| "loss": 0.7061, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002758146683850333, | |
| "loss": 0.5978, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002755103143728906, | |
| "loss": 0.623, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027520575628821645, | |
| "loss": 0.6494, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027490099495410284, | |
| "loss": 0.8597, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002745960311941912, | |
| "loss": 0.5277, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002742908658326699, | |
| "loss": 0.6807, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027398549969427215, | |
| "loss": 0.5962, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002736799336042739, | |
| "loss": 0.7133, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002733741683884914, | |
| "loss": 0.6638, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027306820487327905, | |
| "loss": 0.6981, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002727620438855271, | |
| "loss": 0.551, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027245568625265974, | |
| "loss": 0.4927, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027214913280263235, | |
| "loss": 0.6758, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002718423843639296, | |
| "loss": 0.634, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027153544176556314, | |
| "loss": 0.5044, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00027122830583706936, | |
| "loss": 0.6553, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002709209774085071, | |
| "loss": 0.592, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002706134573104556, | |
| "loss": 0.5193, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002703057463740119, | |
| "loss": 0.511, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002699978454307888, | |
| "loss": 0.5747, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002696897553129128, | |
| "loss": 0.5736, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002693814768530215, | |
| "loss": 0.569, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002690730108842615, | |
| "loss": 0.575, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026876435824028634, | |
| "loss": 0.6588, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002684555197552539, | |
| "loss": 0.4873, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026814649626382423, | |
| "loss": 0.7184, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026783728860115756, | |
| "loss": 0.6923, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026752789760291185, | |
| "loss": 0.5479, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026721832410524037, | |
| "loss": 0.6508, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002669085689447898, | |
| "loss": 0.5326, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026659863295869766, | |
| "loss": 0.5271, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026628851698459017, | |
| "loss": 0.6588, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026597822186058003, | |
| "loss": 0.5319, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002656677484252641, | |
| "loss": 0.668, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026535709751772114, | |
| "loss": 0.5484, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026504626997750943, | |
| "loss": 0.6673, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026473526664466484, | |
| "loss": 0.7688, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026442408835969817, | |
| "loss": 0.5884, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000264112735963593, | |
| "loss": 0.5721, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026380121029780356, | |
| "loss": 0.6351, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002634895122042525, | |
| "loss": 0.5884, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026317764252532805, | |
| "loss": 0.6249, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026286560210388263, | |
| "loss": 0.5571, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026255339178322967, | |
| "loss": 0.496, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002622410124071422, | |
| "loss": 0.6426, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002619284648198498, | |
| "loss": 0.5665, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026161574986603676, | |
| "loss": 0.5385, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026130286839083975, | |
| "loss": 0.6971, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002609898212398455, | |
| "loss": 0.6886, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026067660925908827, | |
| "loss": 0.4864, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026036323329504805, | |
| "loss": 0.5795, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002600496941946478, | |
| "loss": 0.674, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002597359928052516, | |
| "loss": 0.5623, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025942212997466193, | |
| "loss": 0.7926, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002591081065511176, | |
| "loss": 0.6203, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025879392338329154, | |
| "loss": 0.5183, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025847958132028834, | |
| "loss": 0.5908, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025816508121164194, | |
| "loss": 0.5671, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025785042390731355, | |
| "loss": 0.5284, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025753561025768914, | |
| "loss": 0.4666, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002572206411135772, | |
| "loss": 0.6031, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002569055173262065, | |
| "loss": 0.4675, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025659023974722376, | |
| "loss": 0.6052, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002562748092286912, | |
| "loss": 0.5744, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002559592266230846, | |
| "loss": 0.631, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002556434927832906, | |
| "loss": 0.5999, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002553276085626045, | |
| "loss": 0.5661, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002550115748147282, | |
| "loss": 0.5969, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002546953923937676, | |
| "loss": 0.5902, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002543790621542305, | |
| "loss": 0.4696, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002540625849510241, | |
| "loss": 0.6571, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002537459616394527, | |
| "loss": 0.6716, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002534291930752156, | |
| "loss": 0.6623, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025311228011440465, | |
| "loss": 0.6643, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025279522361350195, | |
| "loss": 0.661, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002524780244293775, | |
| "loss": 0.6256, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00025216068341928684, | |
| "loss": 0.5152, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00025184320144086894, | |
| "loss": 0.6576, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002515255793521436, | |
| "loss": 0.5751, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002512078180115094, | |
| "loss": 0.7199, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00025088991827774135, | |
| "loss": 0.6782, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002505718810099881, | |
| "loss": 0.5683, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00025025370706777047, | |
| "loss": 0.5592, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002499353973109783, | |
| "loss": 0.6514, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002496169525998686, | |
| "loss": 0.513, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002492983737950632, | |
| "loss": 0.8558, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024897966175754614, | |
| "loss": 0.606, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002486608173486618, | |
| "loss": 0.5086, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002483418414301119, | |
| "loss": 0.5997, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024802273486395395, | |
| "loss": 0.6479, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002477034985125985, | |
| "loss": 0.5839, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002473841332388066, | |
| "loss": 0.5507, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002470646399056879, | |
| "loss": 0.5209, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024674501937669825, | |
| "loss": 0.5869, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000246425272515637, | |
| "loss": 0.5121, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024610540018664515, | |
| "loss": 0.6656, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002457854032542026, | |
| "loss": 0.5784, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002454652825831261, | |
| "loss": 0.5954, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002451450390385669, | |
| "loss": 0.641, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002448246734860082, | |
| "loss": 0.6801, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002445041867912629, | |
| "loss": 0.4268, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002441835798204714, | |
| "loss": 0.6886, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024386285344009915, | |
| "loss": 0.6441, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024354200851693425, | |
| "loss": 0.6536, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024322104591808518, | |
| "loss": 0.6856, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024289996651097852, | |
| "loss": 0.6311, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024257877116335647, | |
| "loss": 0.5318, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024225746074327457, | |
| "loss": 0.5349, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024193603611909938, | |
| "loss": 0.4851, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024161449815950604, | |
| "loss": 0.7092, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024129284773347616, | |
| "loss": 0.5582, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024097108571029504, | |
| "loss": 0.751, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002406492129595498, | |
| "loss": 0.4319, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002403272303511267, | |
| "loss": 0.6059, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002400051387552089, | |
| "loss": 0.5332, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023968293904227416, | |
| "loss": 0.6963, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023936063208309248, | |
| "loss": 0.6334, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002390382187487235, | |
| "loss": 0.535, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023871569991051463, | |
| "loss": 0.5836, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023839307644009809, | |
| "loss": 0.5793, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023807034920938916, | |
| "loss": 0.6611, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023774751909058336, | |
| "loss": 0.6133, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023742458695615438, | |
| "loss": 0.6956, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023710155367885163, | |
| "loss": 0.5664, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023677842013169768, | |
| "loss": 0.6474, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023645518718798627, | |
| "loss": 0.6103, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002361318557212798, | |
| "loss": 0.5126, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023580842660540676, | |
| "loss": 0.5481, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023548490071445967, | |
| "loss": 0.6181, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023516127892279256, | |
| "loss": 0.5051, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002348375621050186, | |
| "loss": 0.6425, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002345137511360079, | |
| "loss": 0.6452, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023418984689088483, | |
| "loss": 0.648, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023386585024502598, | |
| "loss": 0.464, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023354176207405765, | |
| "loss": 0.7175, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002332175832538535, | |
| "loss": 0.6119, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023289331466053214, | |
| "loss": 0.4535, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002325689571704547, | |
| "loss": 0.5197, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002322445116602229, | |
| "loss": 0.5404, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023191997900667586, | |
| "loss": 0.6062, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023159536008688863, | |
| "loss": 0.7049, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023127065577816915, | |
| "loss": 0.5183, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023094586695805624, | |
| "loss": 0.557, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023062099450431718, | |
| "loss": 0.7013, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023029603929494504, | |
| "loss": 0.4784, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002299710022081567, | |
| "loss": 0.6412, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002296458841223905, | |
| "loss": 0.7029, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022932068591630334, | |
| "loss": 0.5793, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022899540846876883, | |
| "loss": 0.5633, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002286700526588747, | |
| "loss": 0.5717, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022834461936592054, | |
| "loss": 0.5916, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002280191094694153, | |
| "loss": 0.5211, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002276935238490748, | |
| "loss": 0.5772, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002273678633848198, | |
| "loss": 0.538, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002270421289567731, | |
| "loss": 0.7536, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002267163214452575, | |
| "loss": 0.5432, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022639044173079333, | |
| "loss": 0.7121, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022606449069409587, | |
| "loss": 0.4753, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022573846921607337, | |
| "loss": 0.5248, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022541237817782448, | |
| "loss": 0.5956, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022508621846063561, | |
| "loss": 0.5833, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022475999094597894, | |
| "loss": 0.6158, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022443369651550988, | |
| "loss": 0.3833, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002241073360510646, | |
| "loss": 0.5666, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002237809104346579, | |
| "loss": 0.5257, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002234544205484804, | |
| "loss": 0.4698, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002231278672748966, | |
| "loss": 0.5947, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022280125149644228, | |
| "loss": 0.6016, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002224745740958222, | |
| "loss": 0.5976, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022214783595590746, | |
| "loss": 0.6418, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022182103795973356, | |
| "loss": 0.516, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022149418099049758, | |
| "loss": 0.6745, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022116726593155605, | |
| "loss": 0.5929, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022084029366642248, | |
| "loss": 0.5449, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002205132650787651, | |
| "loss": 0.6999, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002201861810524041, | |
| "loss": 0.6399, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002198590424713098, | |
| "loss": 0.636, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021953185021959964, | |
| "loss": 0.5715, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021920460518153636, | |
| "loss": 0.6085, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021887730824152532, | |
| "loss": 0.6042, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021854996028411203, | |
| "loss": 0.5206, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021822256219398005, | |
| "loss": 0.4998, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021789511485594814, | |
| "loss": 0.5396, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021756761915496853, | |
| "loss": 0.7216, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002172400759761239, | |
| "loss": 0.555, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021691248620462532, | |
| "loss": 0.7088, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021658485072580978, | |
| "loss": 0.649, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021625717042513776, | |
| "loss": 0.5802, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021592944618819093, | |
| "loss": 0.5908, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021560167890066976, | |
| "loss": 0.6346, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021527386944839076, | |
| "loss": 0.6521, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021494601871728487, | |
| "loss": 0.5974, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021461812759339418, | |
| "loss": 0.577, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021429019696287015, | |
| "loss": 0.617, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021396222771197102, | |
| "loss": 0.496, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002136342207270592, | |
| "loss": 0.6628, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021330617689459933, | |
| "loss": 0.622, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002129780971011556, | |
| "loss": 0.5631, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021264998223338908, | |
| "loss": 0.5748, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021232183317805613, | |
| "loss": 0.6545, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002119936508220051, | |
| "loss": 0.5119, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021166543605217447, | |
| "loss": 0.5616, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002113371897555904, | |
| "loss": 0.63, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002110089128193641, | |
| "loss": 0.5225, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002106806061306898, | |
| "loss": 0.675, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021035227057684193, | |
| "loss": 0.5959, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021002390704517293, | |
| "loss": 0.5498, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020969551642311104, | |
| "loss": 0.6352, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002093670995981576, | |
| "loss": 0.6417, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020903865745788472, | |
| "loss": 0.5401, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020871019088993298, | |
| "loss": 0.5388, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020838170078200902, | |
| "loss": 0.5584, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020805318802188304, | |
| "loss": 0.5231, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020772465349738645, | |
| "loss": 0.6236, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020739609809640949, | |
| "loss": 0.6752, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002070675227068989, | |
| "loss": 0.5411, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002067389282168553, | |
| "loss": 0.6577, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002064103155143311, | |
| "loss": 0.5116, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020608168548742767, | |
| "loss": 0.5038, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002057530390242935, | |
| "loss": 0.5372, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020542437701312133, | |
| "loss": 0.5425, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020509570034214596, | |
| "loss": 0.6684, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002047670098996418, | |
| "loss": 0.5784, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020443830657392042, | |
| "loss": 0.5851, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020410959125332845, | |
| "loss": 0.566, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020378086482624456, | |
| "loss": 0.6186, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020345212818107772, | |
| "loss": 0.5748, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002031233822062644, | |
| "loss": 0.5842, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020279462779026642, | |
| "loss": 0.6554, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002024658658215682, | |
| "loss": 0.5159, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002021370971886747, | |
| "loss": 0.6017, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020180832278010878, | |
| "loss": 0.5976, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020147954348440918, | |
| "loss": 0.5319, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020115076019012755, | |
| "loss": 0.54, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020082197378582653, | |
| "loss": 0.5083, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020049318516007696, | |
| "loss": 0.52, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00020016439520145594, | |
| "loss": 0.7114, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001998356047985441, | |
| "loss": 0.5675, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001995068148399231, | |
| "loss": 0.6798, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019917802621417351, | |
| "loss": 0.5685, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001988492398098725, | |
| "loss": 0.6611, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019852045651559092, | |
| "loss": 0.6799, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019819167721989124, | |
| "loss": 0.5562, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019786290281132538, | |
| "loss": 0.5373, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001975341341784319, | |
| "loss": 0.6438, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001972053722097336, | |
| "loss": 0.6821, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019687661779373562, | |
| "loss": 0.5554, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019654787181892233, | |
| "loss": 0.5566, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019621913517375552, | |
| "loss": 0.6173, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001958904087466716, | |
| "loss": 0.4909, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001955616934260796, | |
| "loss": 0.6469, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019523299010035827, | |
| "loss": 0.6703, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019490429965785406, | |
| "loss": 0.7082, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019457562298687872, | |
| "loss": 0.4897, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019424696097570659, | |
| "loss": 0.6879, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019391831451257235, | |
| "loss": 0.5814, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019358968448566897, | |
| "loss": 0.5914, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019326107178314474, | |
| "loss": 0.5879, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001929324772931011, | |
| "loss": 0.5253, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019260390190359053, | |
| "loss": 0.4781, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019227534650261357, | |
| "loss": 0.5881, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019194681197811704, | |
| "loss": 0.494, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.000191618299217991, | |
| "loss": 0.6606, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019128980911006706, | |
| "loss": 0.6415, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019096134254211535, | |
| "loss": 0.6102, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001906329004018424, | |
| "loss": 0.4848, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019030448357688898, | |
| "loss": 0.686, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001899760929548271, | |
| "loss": 0.5898, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001896477294231581, | |
| "loss": 0.5556, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001893193938693102, | |
| "loss": 0.6058, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018899108718063593, | |
| "loss": 0.4957, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018866281024440966, | |
| "loss": 0.5546, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018833456394782552, | |
| "loss": 0.4813, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018800634917799495, | |
| "loss": 0.6084, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018767816682194394, | |
| "loss": 0.5452, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001873500177666109, | |
| "loss": 0.5506, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018702190289884447, | |
| "loss": 0.6016, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018669382310540072, | |
| "loss": 0.6112, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001863657792729408, | |
| "loss": 0.6409, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018603777228802905, | |
| "loss": 0.4469, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018570980303712987, | |
| "loss": 0.7357, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001853818724066059, | |
| "loss": 0.5013, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018505398128271515, | |
| "loss": 0.8275, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018472613055160928, | |
| "loss": 0.6174, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018439832109933032, | |
| "loss": 0.5529, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018407055381180904, | |
| "loss": 0.4092, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018374282957486226, | |
| "loss": 0.4563, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001834151492741903, | |
| "loss": 0.5041, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001830875137953747, | |
| "loss": 0.5483, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001827599240238761, | |
| "loss": 0.648, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018243238084503154, | |
| "loss": 0.6578, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001821048851440519, | |
| "loss": 0.6169, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018177743780602003, | |
| "loss": 0.5239, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018145003971588801, | |
| "loss": 0.5951, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018112269175847475, | |
| "loss": 0.656, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018079539481846365, | |
| "loss": 0.5225, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018046814978040038, | |
| "loss": 0.4178, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001801409575286903, | |
| "loss": 0.7248, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001798138189475959, | |
| "loss": 0.618, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017948673492123495, | |
| "loss": 0.4661, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017915970633357753, | |
| "loss": 0.6421, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017883273406844394, | |
| "loss": 0.4908, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017850581900950244, | |
| "loss": 0.5955, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017817896204026652, | |
| "loss": 0.6161, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017785216404409256, | |
| "loss": 0.7025, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017752542590417783, | |
| "loss": 0.5538, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017719874850355774, | |
| "loss": 0.598, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001768721327251035, | |
| "loss": 0.7377, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017654557945151966, | |
| "loss": 0.6071, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017621908956534215, | |
| "loss": 0.6327, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001758926639489354, | |
| "loss": 0.5189, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017556630348449014, | |
| "loss": 0.5354, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001752400090540211, | |
| "loss": 0.5147, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017491378153936443, | |
| "loss": 0.6035, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017458762182217562, | |
| "loss": 0.5917, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017426153078392662, | |
| "loss": 0.6357, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001739355093059042, | |
| "loss": 0.7021, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017360955826920677, | |
| "loss": 0.6531, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001732836785547425, | |
| "loss": 0.6321, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017295787104322694, | |
| "loss": 0.5125, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017263213661518022, | |
| "loss": 0.526, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017230647615092517, | |
| "loss": 0.5105, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017198089053058476, | |
| "loss": 0.6109, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001716553806340795, | |
| "loss": 0.6327, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017132994734112534, | |
| "loss": 0.7332, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001710045915312312, | |
| "loss": 0.5146, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017067931408369674, | |
| "loss": 0.5959, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001703541158776096, | |
| "loss": 0.6762, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017002899779184328, | |
| "loss": 0.5781, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.000169703960705055, | |
| "loss": 0.5869, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001693790054956829, | |
| "loss": 0.66, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016905413304194372, | |
| "loss": 0.478, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001687293442218309, | |
| "loss": 0.6168, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016840463991311141, | |
| "loss": 0.6429, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016808002099332422, | |
| "loss": 0.532, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016775548833977715, | |
| "loss": 0.6642, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001674310428295453, | |
| "loss": 0.6117, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016710668533946794, | |
| "loss": 0.6529, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001667824167461465, | |
| "loss": 0.5162, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016645823792594237, | |
| "loss": 0.682, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001661341497549741, | |
| "loss": 0.6162, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016581015310911522, | |
| "loss": 0.6077, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016548624886399217, | |
| "loss": 0.6012, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016516243789498145, | |
| "loss": 0.4475, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016483872107720746, | |
| "loss": 0.7236, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016451509928554035, | |
| "loss": 0.7156, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016419157339459325, | |
| "loss": 0.5773, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016386814427872025, | |
| "loss": 0.4481, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016354481281201372, | |
| "loss": 0.6861, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016322157986830234, | |
| "loss": 0.5268, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016289844632114847, | |
| "loss": 0.5019, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016257541304384563, | |
| "loss": 0.5413, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001622524809094167, | |
| "loss": 0.7214, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001619296507906109, | |
| "loss": 0.6356, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001616069235599019, | |
| "loss": 0.6104, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016128430008948544, | |
| "loss": 0.6021, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016096178125127655, | |
| "loss": 0.6504, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001606393679169076, | |
| "loss": 0.551, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001603170609577258, | |
| "loss": 0.5265, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015999486124479114, | |
| "loss": 0.6799, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001596727696488734, | |
| "loss": 0.6665, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015935078704045025, | |
| "loss": 0.57, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015902891428970498, | |
| "loss": 0.6211, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015870715226652392, | |
| "loss": 0.5729, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015838550184049395, | |
| "loss": 0.6553, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015806396388090067, | |
| "loss": 0.5665, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015774253925672545, | |
| "loss": 0.5231, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001574212288366436, | |
| "loss": 0.4747, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001571000334890215, | |
| "loss": 0.4959, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015677895408191487, | |
| "loss": 0.6132, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015645799148306582, | |
| "loss": 0.6648, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015613714655990087, | |
| "loss": 0.5834, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015581642017952865, | |
| "loss": 0.6344, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015549581320873715, | |
| "loss": 0.6147, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015517532651399186, | |
| "loss": 0.6592, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001548549609614331, | |
| "loss": 0.589, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015453471741687394, | |
| "loss": 0.6146, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015421459674579748, | |
| "loss": 0.5862, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015389459981335487, | |
| "loss": 0.5358, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015357472748436302, | |
| "loss": 0.6281, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015325498062330185, | |
| "loss": 0.6898, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015293536009431207, | |
| "loss": 0.4841, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015261586676119343, | |
| "loss": 0.7174, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015229650148740155, | |
| "loss": 0.6463, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.000151977265136046, | |
| "loss": 0.5454, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015165815856988814, | |
| "loss": 0.5697, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001513391826513383, | |
| "loss": 0.8046, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015102033824245382, | |
| "loss": 0.6944, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015070162620493683, | |
| "loss": 0.5571, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015038304740013148, | |
| "loss": 0.6686, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015006460268902178, | |
| "loss": 0.5048, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014974629293222958, | |
| "loss": 0.4968, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014942811899001191, | |
| "loss": 0.5873, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014911008172225878, | |
| "loss": 0.6541, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001487921819884906, | |
| "loss": 0.5694, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014847442064785642, | |
| "loss": 0.8085, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014815679855913113, | |
| "loss": 0.5661, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014783931658071317, | |
| "loss": 0.563, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014752197557062256, | |
| "loss": 0.5192, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014720477638649807, | |
| "loss": 0.6208, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001468877198855954, | |
| "loss": 0.61, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014657080692478445, | |
| "loss": 0.6043, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001462540383605474, | |
| "loss": 0.7895, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014593741504897602, | |
| "loss": 0.5809, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001456209378457695, | |
| "loss": 0.4988, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001453046076062324, | |
| "loss": 0.517, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014498842518527184, | |
| "loss": 0.4019, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014467239143739553, | |
| "loss": 0.621, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014435650721670947, | |
| "loss": 0.418, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014404077337691547, | |
| "loss": 0.6342, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014372519077130884, | |
| "loss": 0.6217, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014340976025277628, | |
| "loss": 0.5439, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014309448267379353, | |
| "loss": 0.6359, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014277935888642288, | |
| "loss": 0.6488, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014246438974231088, | |
| "loss": 0.5722, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001421495760926865, | |
| "loss": 0.6628, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014183491878835814, | |
| "loss": 0.4864, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00014152041867971168, | |
| "loss": 0.5799, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00014120607661670848, | |
| "loss": 0.4748, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001408918934488824, | |
| "loss": 0.5015, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00014057787002533806, | |
| "loss": 0.6417, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001402640071947484, | |
| "loss": 0.6394, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013995030580535222, | |
| "loss": 0.5146, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013963676670495202, | |
| "loss": 0.6583, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013932339074091177, | |
| "loss": 0.602, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013901017876015459, | |
| "loss": 0.4739, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013869713160916027, | |
| "loss": 0.7806, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013838425013396326, | |
| "loss": 0.5962, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013807153518015025, | |
| "loss": 0.6247, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013775898759285787, | |
| "loss": 0.5588, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013744660821677032, | |
| "loss": 0.5648, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013713439789611739, | |
| "loss": 0.5577, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013682235747467195, | |
| "loss": 0.6871, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001365104877957476, | |
| "loss": 0.7031, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013619878970219643, | |
| "loss": 0.4076, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.000135887264036407, | |
| "loss": 0.6394, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013557591164030193, | |
| "loss": 0.5669, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013526473335533518, | |
| "loss": 0.6243, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001349537300224906, | |
| "loss": 0.4863, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013464290248227893, | |
| "loss": 0.5245, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013433225157473587, | |
| "loss": 0.544, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013402177813942, | |
| "loss": 0.5648, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001337114830154099, | |
| "loss": 0.62, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001334013670413024, | |
| "loss": 0.5993, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001330914310552102, | |
| "loss": 0.6768, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013278167589475965, | |
| "loss": 0.6884, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013247210239708822, | |
| "loss": 0.696, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013216271139884244, | |
| "loss": 0.5382, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001318535037361758, | |
| "loss": 0.6814, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001315444802447462, | |
| "loss": 0.5858, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013123564175971362, | |
| "loss": 0.6106, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013092698911573852, | |
| "loss": 0.5073, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013061852314697855, | |
| "loss": 0.6604, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001303102446870873, | |
| "loss": 0.5642, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013000215456921124, | |
| "loss": 0.6113, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00012969425362598815, | |
| "loss": 0.477, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00012938654268954448, | |
| "loss": 0.684, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012907902259149287, | |
| "loss": 0.6332, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012877169416293069, | |
| "loss": 0.4346, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001284645582344369, | |
| "loss": 0.6421, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012815761563607042, | |
| "loss": 0.4018, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012785086719736767, | |
| "loss": 0.5223, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001275443137473403, | |
| "loss": 0.5652, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001272379561144729, | |
| "loss": 0.5142, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000126931795126721, | |
| "loss": 0.5714, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012662583161150866, | |
| "loss": 0.8653, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012632006639572618, | |
| "loss": 0.6586, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012601450030572787, | |
| "loss": 0.5177, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012570913416733016, | |
| "loss": 0.6673, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012540396880580886, | |
| "loss": 0.5853, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012509900504589715, | |
| "loss": 0.5235, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012479424371178362, | |
| "loss": 0.4654, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012448968562710946, | |
| "loss": 0.6356, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012418533161496673, | |
| "loss": 0.682, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012388118249789608, | |
| "loss": 0.5594, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012357723909788425, | |
| "loss": 0.6996, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012327350223636193, | |
| "loss": 0.6306, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012296997273420184, | |
| "loss": 0.5851, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012266665141171624, | |
| "loss": 0.5441, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0001223635390886547, | |
| "loss": 0.6795, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012206063658420194, | |
| "loss": 0.61, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0001217579447169757, | |
| "loss": 0.5037, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012145546430502453, | |
| "loss": 0.5703, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012115319616582518, | |
| "loss": 0.6322, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012085114111628115, | |
| "loss": 0.7315, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0001205492999727198, | |
| "loss": 0.5698, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012024767355089037, | |
| "loss": 0.5622, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011994626266596185, | |
| "loss": 0.6408, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011964506813252078, | |
| "loss": 0.5888, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011934409076456901, | |
| "loss": 0.558, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011904333137552124, | |
| "loss": 0.633, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011874279077820345, | |
| "loss": 0.5286, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011844246978484998, | |
| "loss": 0.6405, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0001181423692071018, | |
| "loss": 0.7323, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011784248985600434, | |
| "loss": 0.5947, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011754283254200493, | |
| "loss": 0.6061, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011724339807495081, | |
| "loss": 0.479, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011694418726408711, | |
| "loss": 0.5226, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011664520091805455, | |
| "loss": 0.5983, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011634643984488702, | |
| "loss": 0.5292, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011604790485200965, | |
| "loss": 0.5651, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011574959674623671, | |
| "loss": 0.5867, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011545151633376909, | |
| "loss": 0.4943, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011515366442019249, | |
| "loss": 0.436, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011485604181047507, | |
| "loss": 0.5399, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011455864930896517, | |
| "loss": 0.5985, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011426148771938914, | |
| "loss": 0.5151, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011396455784484965, | |
| "loss": 0.6318, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011366786048782278, | |
| "loss": 0.4892, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011337139645015642, | |
| "loss": 0.5246, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011307516653306766, | |
| "loss": 0.4829, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011277917153714117, | |
| "loss": 0.5657, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011248341226232661, | |
| "loss": 0.5778, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011218788950793641, | |
| "loss": 0.6199, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011189260407264412, | |
| "loss": 0.5737, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011159755675448162, | |
| "loss": 0.6709, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011130274835083731, | |
| "loss": 0.7244, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011100817965845412, | |
| "loss": 0.6206, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00011071385147342685, | |
| "loss": 0.4948, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00011041976459120043, | |
| "loss": 0.5182, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001101259198065677, | |
| "loss": 0.5773, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010983231791366733, | |
| "loss": 0.4056, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010953895970598122, | |
| "loss": 0.5953, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010924584597633285, | |
| "loss": 0.6621, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001089529775168852, | |
| "loss": 0.4612, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010866035511913804, | |
| "loss": 0.5649, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010836797957392623, | |
| "loss": 0.6143, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010807585167141767, | |
| "loss": 0.6086, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010778397220111091, | |
| "loss": 0.5377, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.000107492341951833, | |
| "loss": 0.6693, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010720096171173736, | |
| "loss": 0.501, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010690983226830213, | |
| "loss": 0.5379, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010661895440832727, | |
| "loss": 0.5287, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010632832891793289, | |
| "loss": 0.543, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010603795658255727, | |
| "loss": 0.5684, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001057478381869542, | |
| "loss": 0.5377, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010545797451519139, | |
| "loss": 0.6862, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010516836635064819, | |
| "loss": 0.6075, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010487901447601327, | |
| "loss": 0.533, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010458991967328266, | |
| "loss": 0.5118, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010430108272375764, | |
| "loss": 0.5511, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010401250440804278, | |
| "loss": 0.5975, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010372418550604351, | |
| "loss": 0.5916, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010343612679696409, | |
| "loss": 0.4844, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010314832905930585, | |
| "loss": 0.6386, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010286079307086473, | |
| "loss": 0.5655, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0001025735196087291, | |
| "loss": 0.4526, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0001022865094492781, | |
| "loss": 0.6151, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010199976336817905, | |
| "loss": 0.723, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010171328214038562, | |
| "loss": 0.567, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010142706654013587, | |
| "loss": 0.4526, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010114111734094969, | |
| "loss": 0.7902, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010085543531562733, | |
| "loss": 0.4727, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010057002123624668, | |
| "loss": 0.6169, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010028487587416174, | |
| "loss": 0.6267, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010000000000000005, | |
| "loss": 0.6472, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.97153943836609e-05, | |
| "loss": 0.6187, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.943105979431343e-05, | |
| "loss": 0.5263, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.914699700039393e-05, | |
| "loss": 0.4554, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.886320676960435e-05, | |
| "loss": 0.5333, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.857968986891002e-05, | |
| "loss": 0.6132, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.829644706453761e-05, | |
| "loss": 0.4738, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.801347912197288e-05, | |
| "loss": 0.5376, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.773078680595875e-05, | |
| "loss": 0.6361, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.74483708804935e-05, | |
| "loss": 0.5269, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.716623210882815e-05, | |
| "loss": 0.5961, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.688437125346477e-05, | |
| "loss": 0.5514, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.660278907615448e-05, | |
| "loss": 0.6386, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.632148633789499e-05, | |
| "loss": 0.6139, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.604046379892908e-05, | |
| "loss": 0.6664, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.575972221874217e-05, | |
| "loss": 0.62, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.547926235606026e-05, | |
| "loss": 0.6849, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.51990849688481e-05, | |
| "loss": 0.554, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.491919081430693e-05, | |
| "loss": 0.5524, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.463958064887273e-05, | |
| "loss": 0.5242, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.436025522821368e-05, | |
| "loss": 0.5359, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.408121530722867e-05, | |
| "loss": 0.7573, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.380246164004498e-05, | |
| "loss": 0.6348, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.352399498001614e-05, | |
| "loss": 0.5284, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.324581607972e-05, | |
| "loss": 0.4824, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.296792569095696e-05, | |
| "loss": 0.5186, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.269032456474744e-05, | |
| "loss": 0.7306, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.241301345133022e-05, | |
| "loss": 0.6135, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.213599310016018e-05, | |
| "loss": 0.7198, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.18592642599066e-05, | |
| "loss": 0.6388, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.158282767845086e-05, | |
| "loss": 0.7146, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.130668410288428e-05, | |
| "loss": 0.5117, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.103083427950667e-05, | |
| "loss": 0.5372, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.075527895382366e-05, | |
| "loss": 0.5141, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.048001887054498e-05, | |
| "loss": 0.5709, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.020505477358272e-05, | |
| "loss": 0.5992, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.993038740604878e-05, | |
| "loss": 0.5346, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.965601751025312e-05, | |
| "loss": 0.5806, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.938194582770192e-05, | |
| "loss": 0.4386, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.910817309909543e-05, | |
| "loss": 0.6664, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.883470006432579e-05, | |
| "loss": 0.5542, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.856152746247517e-05, | |
| "loss": 0.6192, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.828865603181407e-05, | |
| "loss": 0.5777, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.80160865097988e-05, | |
| "loss": 0.5084, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.774381963306972e-05, | |
| "loss": 0.5847, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.747185613744955e-05, | |
| "loss": 0.5339, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.720019675794073e-05, | |
| "loss": 0.5967, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.692884222872406e-05, | |
| "loss": 0.6195, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.665779328315648e-05, | |
| "loss": 0.5682, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.638705065376887e-05, | |
| "loss": 0.5585, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.611661507226439e-05, | |
| "loss": 0.525, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.584648726951623e-05, | |
| "loss": 0.6005, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.557666797556612e-05, | |
| "loss": 0.5778, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.530715791962162e-05, | |
| "loss": 0.8494, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.503795783005483e-05, | |
| "loss": 0.5017, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.476906843440012e-05, | |
| "loss": 0.5674, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.450049045935204e-05, | |
| "loss": 0.5665, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.42322246307635e-05, | |
| "loss": 0.4734, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.396427167364404e-05, | |
| "loss": 0.665, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.36966323121574e-05, | |
| "loss": 0.6272, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.342930726961992e-05, | |
| "loss": 0.6055, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.316229726849835e-05, | |
| "loss": 0.6692, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.28956030304082e-05, | |
| "loss": 0.5904, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.262922527611152e-05, | |
| "loss": 0.4628, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.236316472551495e-05, | |
| "loss": 0.6658, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.209742209766807e-05, | |
| "loss": 0.553, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.183199811076102e-05, | |
| "loss": 0.611, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.156689348212283e-05, | |
| "loss": 0.5117, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.130210892821965e-05, | |
| "loss": 0.5457, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.103764516465238e-05, | |
| "loss": 0.5706, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.077350290615495e-05, | |
| "loss": 0.6841, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.050968286659255e-05, | |
| "loss": 0.5295, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.024618575895954e-05, | |
| "loss": 0.6741, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.998301229537739e-05, | |
| "loss": 0.4948, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.972016318709292e-05, | |
| "loss": 0.7006, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.945763914447654e-05, | |
| "loss": 0.5699, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.919544087701995e-05, | |
| "loss": 0.523, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.893356909333439e-05, | |
| "loss": 0.5678, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.867202450114892e-05, | |
| "loss": 0.563, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.841080780730831e-05, | |
| "loss": 0.5472, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.814991971777102e-05, | |
| "loss": 0.5659, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.788936093760761e-05, | |
| "loss": 0.6069, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.762913217099852e-05, | |
| "loss": 0.5612, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.736923412123234e-05, | |
| "loss": 0.4909, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.71096674907038e-05, | |
| "loss": 0.6087, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.685043298091219e-05, | |
| "loss": 0.6558, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.65915312924589e-05, | |
| "loss": 0.5878, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.633296312504608e-05, | |
| "loss": 0.578, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.607472917747444e-05, | |
| "loss": 0.6769, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.581683014764143e-05, | |
| "loss": 0.6759, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.555926673253925e-05, | |
| "loss": 0.5574, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.530203962825331e-05, | |
| "loss": 0.4509, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.504514952995992e-05, | |
| "loss": 0.6259, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.478859713192468e-05, | |
| "loss": 0.561, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.45323831275004e-05, | |
| "loss": 0.5794, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.42765082091256e-05, | |
| "loss": 0.5938, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.402097306832227e-05, | |
| "loss": 0.6182, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.376577839569403e-05, | |
| "loss": 0.5978, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.351092488092457e-05, | |
| "loss": 0.6754, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.325641321277536e-05, | |
| "loss": 0.5047, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.300224407908404e-05, | |
| "loss": 0.3988, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.274841816676275e-05, | |
| "loss": 0.5145, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.249493616179565e-05, | |
| "loss": 0.5139, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.224179874923791e-05, | |
| "loss": 0.5856, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.198900661321299e-05, | |
| "loss": 0.4739, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.173656043691159e-05, | |
| "loss": 0.6046, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.148446090258917e-05, | |
| "loss": 0.5799, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.123270869156437e-05, | |
| "loss": 0.4513, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.098130448421738e-05, | |
| "loss": 0.5846, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.073024895998766e-05, | |
| "loss": 0.4879, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.047954279737232e-05, | |
| "loss": 0.488, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.022918667392447e-05, | |
| "loss": 0.5757, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.997918126625116e-05, | |
| "loss": 0.5933, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.972952725001146e-05, | |
| "loss": 0.5535, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.948022529991497e-05, | |
| "loss": 0.5882, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.92312760897197e-05, | |
| "loss": 0.6387, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.898268029223034e-05, | |
| "loss": 0.6817, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.873443857929638e-05, | |
| "loss": 0.5143, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.848655162181063e-05, | |
| "loss": 0.5678, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.823902008970686e-05, | |
| "loss": 0.6752, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.799184465195845e-05, | |
| "loss": 0.3535, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.774502597657644e-05, | |
| "loss": 0.5071, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.749856473060752e-05, | |
| "loss": 0.5279, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.725246158013243e-05, | |
| "loss": 0.5461, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.700671719026432e-05, | |
| "loss": 0.5052, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.676133222514656e-05, | |
| "loss": 0.6445, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.651630734795114e-05, | |
| "loss": 0.6836, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.627164322087709e-05, | |
| "loss": 0.5863, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.602734050514823e-05, | |
| "loss": 0.5827, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.57833998610119e-05, | |
| "loss": 0.6316, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.553982194773662e-05, | |
| "loss": 0.6316, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.529660742361093e-05, | |
| "loss": 0.6263, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.505375694594096e-05, | |
| "loss": 0.6869, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.481127117104913e-05, | |
| "loss": 0.6391, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.456915075427232e-05, | |
| "loss": 0.4967, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.432739634995977e-05, | |
| "loss": 0.7022, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.40860086114718e-05, | |
| "loss": 0.5281, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.384498819117752e-05, | |
| "loss": 0.5553, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.360433574045361e-05, | |
| "loss": 0.5955, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.336405190968206e-05, | |
| "loss": 0.6587, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.312413734824869e-05, | |
| "loss": 0.6174, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.288459270454146e-05, | |
| "loss": 0.6349, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.264541862594846e-05, | |
| "loss": 0.5279, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.24066157588563e-05, | |
| "loss": 0.5298, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.216818474864843e-05, | |
| "loss": 0.5281, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.19301262397034e-05, | |
| "loss": 0.6515, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.169244087539289e-05, | |
| "loss": 0.4388, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.145512929808012e-05, | |
| "loss": 0.5272, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.121819214911832e-05, | |
| "loss": 0.5215, | |
| "step": 1523 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.098163006884863e-05, | |
| "loss": 0.4931, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0745443696598467e-05, | |
| "loss": 0.6555, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.050963367068012e-05, | |
| "loss": 0.4934, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0274200628388536e-05, | |
| "loss": 0.58, | |
| "step": 1527 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.003914520599996e-05, | |
| "loss": 0.5399, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.980446803877011e-05, | |
| "loss": 0.7155, | |
| "step": 1529 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.957016976093237e-05, | |
| "loss": 0.6412, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.933625100569606e-05, | |
| "loss": 0.6489, | |
| "step": 1531 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.910271240524508e-05, | |
| "loss": 0.6347, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.886955459073566e-05, | |
| "loss": 0.5478, | |
| "step": 1533 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.8636778192295006e-05, | |
| "loss": 0.6453, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.840438383901958e-05, | |
| "loss": 0.5159, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.8172372158973375e-05, | |
| "loss": 0.5361, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.794074377918604e-05, | |
| "loss": 0.52, | |
| "step": 1537 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.770949932565133e-05, | |
| "loss": 0.5644, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.747863942332561e-05, | |
| "loss": 0.6857, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.724816469612575e-05, | |
| "loss": 0.5358, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.701807576692766e-05, | |
| "loss": 0.4923, | |
| "step": 1541 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.678837325756479e-05, | |
| "loss": 0.5579, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.6559057788826044e-05, | |
| "loss": 0.5854, | |
| "step": 1543 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.6330129980454505e-05, | |
| "loss": 0.6019, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.610159045114538e-05, | |
| "loss": 0.6511, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.587343981854476e-05, | |
| "loss": 0.5693, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.5645678699247485e-05, | |
| "loss": 0.5621, | |
| "step": 1547 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.541830770879577e-05, | |
| "loss": 0.5177, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.5191327461677587e-05, | |
| "loss": 0.5729, | |
| "step": 1549 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.4964738571324756e-05, | |
| "loss": 0.6283, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.473854165011145e-05, | |
| "loss": 0.5288, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.451273730935258e-05, | |
| "loss": 0.5495, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.42873261593021e-05, | |
| "loss": 0.3946, | |
| "step": 1553 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.4062308809151216e-05, | |
| "loss": 0.4393, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.383768586702687e-05, | |
| "loss": 0.56, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.361345793999024e-05, | |
| "loss": 0.5791, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.338962563403478e-05, | |
| "loss": 0.6294, | |
| "step": 1557 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.316618955408477e-05, | |
| "loss": 0.4667, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.294315030399373e-05, | |
| "loss": 0.5795, | |
| "step": 1559 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.272050848654275e-05, | |
| "loss": 0.6031, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.2498264703438637e-05, | |
| "loss": 0.6122, | |
| "step": 1561 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.2276419555312705e-05, | |
| "loss": 0.6102, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.205497364171878e-05, | |
| "loss": 0.5454, | |
| "step": 1563 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.183392756113179e-05, | |
| "loss": 0.5792, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.1613281910945944e-05, | |
| "loss": 0.6078, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.139303728747351e-05, | |
| "loss": 0.4588, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.1173194285942714e-05, | |
| "loss": 0.5037, | |
| "step": 1567 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.095375350049651e-05, | |
| "loss": 0.4136, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.073471552419087e-05, | |
| "loss": 0.6526, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.051608094899296e-05, | |
| "loss": 0.5529, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.029785036577976e-05, | |
| "loss": 0.462, | |
| "step": 1571 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.0080024364336674e-05, | |
| "loss": 0.5289, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.986260353335537e-05, | |
| "loss": 0.6423, | |
| "step": 1573 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.9645588460432655e-05, | |
| "loss": 0.5589, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.942897973206884e-05, | |
| "loss": 0.5389, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.921277793366581e-05, | |
| "loss": 0.6005, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.8996983649526005e-05, | |
| "loss": 0.5417, | |
| "step": 1577 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.87815974628502e-05, | |
| "loss": 0.5732, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.856661995573655e-05, | |
| "loss": 0.4226, | |
| "step": 1579 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.835205170917854e-05, | |
| "loss": 0.4314, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.813789330306355e-05, | |
| "loss": 0.6286, | |
| "step": 1581 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.7924145316171555e-05, | |
| "loss": 0.6571, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.7710808326173115e-05, | |
| "loss": 0.5803, | |
| "step": 1583 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.7497882909628223e-05, | |
| "loss": 0.402, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.728536964198436e-05, | |
| "loss": 0.63, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.707326909757539e-05, | |
| "loss": 0.6197, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.686158184961955e-05, | |
| "loss": 0.4823, | |
| "step": 1587 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.665030847021814e-05, | |
| "loss": 0.7108, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.64394495303541e-05, | |
| "loss": 0.5695, | |
| "step": 1589 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.622900559989016e-05, | |
| "loss": 0.6441, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.6018977247567454e-05, | |
| "loss": 0.479, | |
| "step": 1591 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.580936504100408e-05, | |
| "loss": 0.5757, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.560016954669346e-05, | |
| "loss": 0.4321, | |
| "step": 1593 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.539139133000272e-05, | |
| "loss": 0.5224, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.51830309551714e-05, | |
| "loss": 0.6232, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.497508898530966e-05, | |
| "loss": 0.4942, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.476756598239695e-05, | |
| "loss": 0.6305, | |
| "step": 1597 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.4560462507280345e-05, | |
| "loss": 0.6654, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.435377911967331e-05, | |
| "loss": 0.443, | |
| "step": 1599 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.4147516378153774e-05, | |
| "loss": 0.4838, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.394167484016296e-05, | |
| "loss": 0.5394, | |
| "step": 1601 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.373625506200378e-05, | |
| "loss": 0.6359, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.3531257598839226e-05, | |
| "loss": 0.4968, | |
| "step": 1603 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.3326683004690895e-05, | |
| "loss": 0.6698, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.3122531832437754e-05, | |
| "loss": 0.5289, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.291880463381426e-05, | |
| "loss": 0.7246, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.2715501959409056e-05, | |
| "loss": 0.6669, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.2512624358663636e-05, | |
| "loss": 0.4569, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.231017237987045e-05, | |
| "loss": 0.5208, | |
| "step": 1609 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.2108146570171924e-05, | |
| "loss": 0.5082, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.1906547475558535e-05, | |
| "loss": 0.5933, | |
| "step": 1611 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.1705375640867697e-05, | |
| "loss": 0.3717, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.150463160978195e-05, | |
| "loss": 0.6177, | |
| "step": 1613 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.130431592482773e-05, | |
| "loss": 0.5659, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.110442912737391e-05, | |
| "loss": 0.5284, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.090497175763013e-05, | |
| "loss": 0.6168, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.0705944354645566e-05, | |
| "loss": 0.4887, | |
| "step": 1617 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.0507347456307395e-05, | |
| "loss": 0.5434, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.0309181599339255e-05, | |
| "loss": 0.4851, | |
| "step": 1619 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.011144731929981e-05, | |
| "loss": 0.6417, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.9914145150581405e-05, | |
| "loss": 0.6536, | |
| "step": 1621 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.97172756264087e-05, | |
| "loss": 0.5332, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.952083927883692e-05, | |
| "loss": 0.699, | |
| "step": 1623 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.932483663875059e-05, | |
| "loss": 0.5081, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.912926823586225e-05, | |
| "loss": 0.5237, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.893413459871087e-05, | |
| "loss": 0.5732, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.8739436254660234e-05, | |
| "loss": 0.5799, | |
| "step": 1627 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.8545173729897985e-05, | |
| "loss": 0.5414, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.8351347549433724e-05, | |
| "loss": 0.5114, | |
| "step": 1629 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.8157958237097866e-05, | |
| "loss": 0.6165, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.79650063155401e-05, | |
| "loss": 0.549, | |
| "step": 1631 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.777249230622819e-05, | |
| "loss": 0.5154, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.758041672944621e-05, | |
| "loss": 0.5012, | |
| "step": 1633 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.7388780104293475e-05, | |
| "loss": 0.5567, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.719758294868303e-05, | |
| "loss": 0.6292, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.700682577934007e-05, | |
| "loss": 0.6107, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.6816509111800726e-05, | |
| "loss": 0.5705, | |
| "step": 1637 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.662663346041077e-05, | |
| "loss": 0.7666, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.643719933832397e-05, | |
| "loss": 0.5697, | |
| "step": 1639 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.6248207257500776e-05, | |
| "loss": 0.765, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.6059657728707166e-05, | |
| "loss": 0.5403, | |
| "step": 1641 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.587155126151302e-05, | |
| "loss": 0.6573, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.568388836429073e-05, | |
| "loss": 0.521, | |
| "step": 1643 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.549666954421391e-05, | |
| "loss": 0.5346, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.5309895307256145e-05, | |
| "loss": 0.4645, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.512356615818941e-05, | |
| "loss": 0.5445, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.493768260058272e-05, | |
| "loss": 0.4568, | |
| "step": 1647 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4752245136801065e-05, | |
| "loss": 0.6272, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.456725426800356e-05, | |
| "loss": 0.5433, | |
| "step": 1649 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.438271049414257e-05, | |
| "loss": 0.5988, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.419861431396212e-05, | |
| "loss": 0.6232, | |
| "step": 1651 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4014966224996493e-05, | |
| "loss": 0.5488, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.383176672356902e-05, | |
| "loss": 0.6025, | |
| "step": 1653 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.364901630479065e-05, | |
| "loss": 0.6566, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.3466715462558753e-05, | |
| "loss": 0.498, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.328486468955558e-05, | |
| "loss": 0.6569, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.310346447724701e-05, | |
| "loss": 0.3958, | |
| "step": 1657 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.29225153158814e-05, | |
| "loss": 0.6333, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.274201769448801e-05, | |
| "loss": 0.5077, | |
| "step": 1659 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.25619721008757e-05, | |
| "loss": 0.6296, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.238237902163188e-05, | |
| "loss": 0.5996, | |
| "step": 1661 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.2203238942120826e-05, | |
| "loss": 0.577, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.202455234648265e-05, | |
| "loss": 0.6299, | |
| "step": 1663 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.184631971763177e-05, | |
| "loss": 0.6003, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.166854153725587e-05, | |
| "loss": 0.5744, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.149121828581445e-05, | |
| "loss": 0.6802, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.131435044253737e-05, | |
| "loss": 0.5213, | |
| "step": 1667 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.113793848542388e-05, | |
| "loss": 0.6491, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0961982891241084e-05, | |
| "loss": 0.4335, | |
| "step": 1669 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.078648413552265e-05, | |
| "loss": 0.601, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.061144269256784e-05, | |
| "loss": 0.4452, | |
| "step": 1671 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0436859035439758e-05, | |
| "loss": 0.6428, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.026273363596437e-05, | |
| "loss": 0.6778, | |
| "step": 1673 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.008906696472922e-05, | |
| "loss": 0.5271, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.991585949108211e-05, | |
| "loss": 0.5103, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9743111683129775e-05, | |
| "loss": 0.5444, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.957082400773663e-05, | |
| "loss": 0.5378, | |
| "step": 1677 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.939899693052366e-05, | |
| "loss": 0.4916, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9227630915866934e-05, | |
| "loss": 0.7527, | |
| "step": 1679 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.9056726426896498e-05, | |
| "loss": 0.6649, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8886283925495173e-05, | |
| "loss": 0.5591, | |
| "step": 1681 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8716303872297067e-05, | |
| "loss": 0.6211, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8546786726686646e-05, | |
| "loss": 0.4996, | |
| "step": 1683 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8377732946797198e-05, | |
| "loss": 0.5575, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.820914298950985e-05, | |
| "loss": 0.5656, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.804101731045212e-05, | |
| "loss": 0.6139, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.787335636399675e-05, | |
| "loss": 0.598, | |
| "step": 1687 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7706160603260676e-05, | |
| "loss": 0.5393, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.753943048010339e-05, | |
| "loss": 0.5141, | |
| "step": 1689 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.737316644512622e-05, | |
| "loss": 0.5761, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.720736894767064e-05, | |
| "loss": 0.5729, | |
| "step": 1691 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7042038435817408e-05, | |
| "loss": 0.501, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6877175356385097e-05, | |
| "loss": 0.6074, | |
| "step": 1693 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6712780154929175e-05, | |
| "loss": 0.5049, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.654885327574048e-05, | |
| "loss": 0.4351, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6385395161844218e-05, | |
| "loss": 0.602, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.62224062549987e-05, | |
| "loss": 0.5546, | |
| "step": 1697 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.605988699569424e-05, | |
| "loss": 0.63, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.5897837823151893e-05, | |
| "loss": 0.562, | |
| "step": 1699 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.573625917532212e-05, | |
| "loss": 0.56, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5575151488883942e-05, | |
| "loss": 0.5293, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.541451519924347e-05, | |
| "loss": 0.5314, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5254350740532794e-05, | |
| "loss": 0.5842, | |
| "step": 1703 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5094658545608952e-05, | |
| "loss": 0.563, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4935439046052577e-05, | |
| "loss": 0.5209, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4776692672166757e-05, | |
| "loss": 0.623, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.461841985297604e-05, | |
| "loss": 0.5986, | |
| "step": 1707 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4460621016225148e-05, | |
| "loss": 0.6124, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4303296588377732e-05, | |
| "loss": 0.5768, | |
| "step": 1709 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4146446994615325e-05, | |
| "loss": 0.6432, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3990072658836303e-05, | |
| "loss": 0.6841, | |
| "step": 1711 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.383417400365451e-05, | |
| "loss": 0.5113, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3678751450398196e-05, | |
| "loss": 0.5389, | |
| "step": 1713 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.352380541910908e-05, | |
| "loss": 0.7349, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3369336328540836e-05, | |
| "loss": 0.5495, | |
| "step": 1715 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3215344596158327e-05, | |
| "loss": 0.604, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3061830638136207e-05, | |
| "loss": 0.5903, | |
| "step": 1717 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.2908794869358042e-05, | |
| "loss": 0.52, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.2756237703414884e-05, | |
| "loss": 0.6684, | |
| "step": 1719 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2604159552604397e-05, | |
| "loss": 0.4203, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.245256082792977e-05, | |
| "loss": 0.4637, | |
| "step": 1721 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2301441939098312e-05, | |
| "loss": 0.6435, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2150803294520704e-05, | |
| "loss": 0.6166, | |
| "step": 1723 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2000645301309674e-05, | |
| "loss": 0.6948, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1850968365278934e-05, | |
| "loss": 0.5734, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1701772890942108e-05, | |
| "loss": 0.4455, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.155305928151159e-05, | |
| "loss": 0.6263, | |
| "step": 1727 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1404827938897644e-05, | |
| "loss": 0.6431, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.125707926370699e-05, | |
| "loss": 0.503, | |
| "step": 1729 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1109813655241962e-05, | |
| "loss": 0.6417, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.096303151149941e-05, | |
| "loss": 0.5697, | |
| "step": 1731 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.081673322916964e-05, | |
| "loss": 0.4416, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0670919203635087e-05, | |
| "loss": 0.5068, | |
| "step": 1733 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.052558982896966e-05, | |
| "loss": 0.4305, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0380745497937316e-05, | |
| "loss": 0.6147, | |
| "step": 1735 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0236386601991164e-05, | |
| "loss": 0.5966, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.009251353127246e-05, | |
| "loss": 0.596, | |
| "step": 1737 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.994912667460942e-05, | |
| "loss": 0.4954, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.9806226419516192e-05, | |
| "loss": 0.5827, | |
| "step": 1739 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9663813152191924e-05, | |
| "loss": 0.5452, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9521887257519667e-05, | |
| "loss": 0.5277, | |
| "step": 1741 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9380449119065225e-05, | |
| "loss": 0.578, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9239499119076164e-05, | |
| "loss": 0.5792, | |
| "step": 1743 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9099037638480977e-05, | |
| "loss": 0.5469, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.895906505688776e-05, | |
| "loss": 0.5786, | |
| "step": 1745 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8819581752583315e-05, | |
| "loss": 0.5904, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.868058810253226e-05, | |
| "loss": 0.7136, | |
| "step": 1747 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.85420844823758e-05, | |
| "loss": 0.5348, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.84040712664308e-05, | |
| "loss": 0.5283, | |
| "step": 1749 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.82665488276887e-05, | |
| "loss": 0.4777, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8129517537814754e-05, | |
| "loss": 0.4548, | |
| "step": 1751 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.79929777671467e-05, | |
| "loss": 0.6258, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7856929884693897e-05, | |
| "loss": 0.5721, | |
| "step": 1753 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.772137425813649e-05, | |
| "loss": 0.5414, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7586311253824063e-05, | |
| "loss": 0.511, | |
| "step": 1755 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7451741236775e-05, | |
| "loss": 0.631, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.731766457067532e-05, | |
| "loss": 0.5598, | |
| "step": 1757 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.718408161787768e-05, | |
| "loss": 0.5395, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7050992739400472e-05, | |
| "loss": 0.6625, | |
| "step": 1759 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.691839829492674e-05, | |
| "loss": 0.5838, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6786298642803434e-05, | |
| "loss": 0.5057, | |
| "step": 1761 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.665469414004013e-05, | |
| "loss": 0.4622, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6523585142308296e-05, | |
| "loss": 0.5737, | |
| "step": 1763 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6392972003940276e-05, | |
| "loss": 0.3944, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.626285507792833e-05, | |
| "loss": 0.5092, | |
| "step": 1765 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.613323471592354e-05, | |
| "loss": 0.5509, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6004111268235156e-05, | |
| "loss": 0.5805, | |
| "step": 1767 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.587548508382939e-05, | |
| "loss": 0.5997, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.574735651032848e-05, | |
| "loss": 0.6007, | |
| "step": 1769 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.561972589401002e-05, | |
| "loss": 0.5367, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.549259357980568e-05, | |
| "loss": 0.5417, | |
| "step": 1771 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5365959911300498e-05, | |
| "loss": 0.578, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5239825230731864e-05, | |
| "loss": 0.5198, | |
| "step": 1773 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.511418987898865e-05, | |
| "loss": 0.5761, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4989054195610209e-05, | |
| "loss": 0.4781, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4864418518785461e-05, | |
| "loss": 0.4777, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4740283185352144e-05, | |
| "loss": 0.4594, | |
| "step": 1777 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4616648530795673e-05, | |
| "loss": 0.5537, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.449351488924835e-05, | |
| "loss": 0.5342, | |
| "step": 1779 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.437088259348851e-05, | |
| "loss": 0.5502, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4248751974939534e-05, | |
| "loss": 0.5159, | |
| "step": 1781 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.412712336366897e-05, | |
| "loss": 0.46, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.400599708838759e-05, | |
| "loss": 0.5663, | |
| "step": 1783 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3885373476448737e-05, | |
| "loss": 0.4346, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3765252853847065e-05, | |
| "loss": 0.5582, | |
| "step": 1785 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3645635545217982e-05, | |
| "loss": 0.6823, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.352652187383665e-05, | |
| "loss": 0.5231, | |
| "step": 1787 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.340791216161703e-05, | |
| "loss": 0.4572, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3289806729111176e-05, | |
| "loss": 0.4247, | |
| "step": 1789 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3172205895508293e-05, | |
| "loss": 0.5043, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3055109978633794e-05, | |
| "loss": 0.5619, | |
| "step": 1791 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2938519294948582e-05, | |
| "loss": 0.5167, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2822434159548025e-05, | |
| "loss": 0.515, | |
| "step": 1793 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.270685488616139e-05, | |
| "loss": 0.7149, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2591781787150591e-05, | |
| "loss": 0.6149, | |
| "step": 1795 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2477215173509792e-05, | |
| "loss": 0.4558, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2363155354864142e-05, | |
| "loss": 0.6418, | |
| "step": 1797 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2249602639469281e-05, | |
| "loss": 0.5086, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2136557334210264e-05, | |
| "loss": 0.7389, | |
| "step": 1799 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.2024019744600923e-05, | |
| "loss": 0.5536, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1911990174782861e-05, | |
| "loss": 0.6159, | |
| "step": 1801 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1800468927524732e-05, | |
| "loss": 0.6106, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1689456304221446e-05, | |
| "loss": 0.579, | |
| "step": 1803 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1578952604893323e-05, | |
| "loss": 0.6558, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1468958128185248e-05, | |
| "loss": 0.6442, | |
| "step": 1805 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1359473171365875e-05, | |
| "loss": 0.6614, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1250498030326917e-05, | |
| "loss": 0.692, | |
| "step": 1807 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1142032999582185e-05, | |
| "loss": 0.487, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1034078372266865e-05, | |
| "loss": 0.7443, | |
| "step": 1809 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0926634440136884e-05, | |
| "loss": 0.7088, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0819701493567791e-05, | |
| "loss": 0.6358, | |
| "step": 1811 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0713279821554234e-05, | |
| "loss": 0.5139, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0607369711709125e-05, | |
| "loss": 0.5366, | |
| "step": 1813 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0501971450262882e-05, | |
| "loss": 0.4908, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0397085322062494e-05, | |
| "loss": 0.5129, | |
| "step": 1815 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0292711610570904e-05, | |
| "loss": 0.5163, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0188850597866273e-05, | |
| "loss": 0.5462, | |
| "step": 1817 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0085502564641113e-05, | |
| "loss": 0.5973, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.98266779020154e-06, | |
| "loss": 0.5908, | |
| "step": 1819 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.880346552466613e-06, | |
| "loss": 0.4985, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.778539127967423e-06, | |
| "loss": 0.6789, | |
| "step": 1821 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.677245791846546e-06, | |
| "loss": 0.5844, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.576466817857178e-06, | |
| "loss": 0.6839, | |
| "step": 1823 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.476202478362317e-06, | |
| "loss": 0.6839, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.37645304433421e-06, | |
| "loss": 0.6216, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.277218785353481e-06, | |
| "loss": 0.6172, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.178499969608534e-06, | |
| "loss": 0.6143, | |
| "step": 1827 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.080296863894621e-06, | |
| "loss": 0.6968, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.982609733613334e-06, | |
| "loss": 0.5413, | |
| "step": 1829 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.885438842771843e-06, | |
| "loss": 0.522, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.788784453982057e-06, | |
| "loss": 0.6453, | |
| "step": 1831 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.69264682845996e-06, | |
| "loss": 0.558, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.597026226025074e-06, | |
| "loss": 0.4673, | |
| "step": 1833 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.501922905099547e-06, | |
| "loss": 0.5758, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.407337122707514e-06, | |
| "loss": 0.6695, | |
| "step": 1835 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.313269134474455e-06, | |
| "loss": 0.6309, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.219719194626474e-06, | |
| "loss": 0.5425, | |
| "step": 1837 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.126687555989642e-06, | |
| "loss": 0.6955, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.034174469989176e-06, | |
| "loss": 0.5157, | |
| "step": 1839 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.942180186648962e-06, | |
| "loss": 0.5601, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.85070495459077e-06, | |
| "loss": 0.5988, | |
| "step": 1841 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.759749021033514e-06, | |
| "loss": 0.6228, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.669312631792757e-06, | |
| "loss": 0.5371, | |
| "step": 1843 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.579396031279906e-06, | |
| "loss": 0.5842, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.489999462501595e-06, | |
| "loss": 0.4797, | |
| "step": 1845 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.401123167059054e-06, | |
| "loss": 0.5301, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.312767385147457e-06, | |
| "loss": 0.6387, | |
| "step": 1847 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.224932355555192e-06, | |
| "loss": 0.6006, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.137618315663242e-06, | |
| "loss": 0.5388, | |
| "step": 1849 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.050825501444736e-06, | |
| "loss": 0.5434, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.964554147463975e-06, | |
| "loss": 0.6283, | |
| "step": 1851 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.878804486876011e-06, | |
| "loss": 0.5419, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.7935767514260894e-06, | |
| "loss": 0.5315, | |
| "step": 1853 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.708871171448783e-06, | |
| "loss": 0.6914, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.624687975867571e-06, | |
| "loss": 0.4933, | |
| "step": 1855 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.54102739219411e-06, | |
| "loss": 0.532, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.457889646527693e-06, | |
| "loss": 0.6382, | |
| "step": 1857 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.375274963554567e-06, | |
| "loss": 0.4447, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.293183566547334e-06, | |
| "loss": 0.7039, | |
| "step": 1859 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.211615677364502e-06, | |
| "loss": 0.5065, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.1305715164495795e-06, | |
| "loss": 0.4635, | |
| "step": 1861 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.050051302830828e-06, | |
| "loss": 0.639, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.970055254120399e-06, | |
| "loss": 0.5991, | |
| "step": 1863 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.890583586513843e-06, | |
| "loss": 0.5698, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.811636514789598e-06, | |
| "loss": 0.6724, | |
| "step": 1865 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.733214252308284e-06, | |
| "loss": 0.4552, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.6553170110122064e-06, | |
| "loss": 0.5713, | |
| "step": 1867 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.577945001424744e-06, | |
| "loss": 0.5379, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.501098432649787e-06, | |
| "loss": 0.7229, | |
| "step": 1869 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.4247775123712265e-06, | |
| "loss": 0.5831, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.348982446852335e-06, | |
| "loss": 0.5144, | |
| "step": 1871 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.273713440935146e-06, | |
| "loss": 0.5958, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.198970698040118e-06, | |
| "loss": 0.6183, | |
| "step": 1873 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.124754420165312e-06, | |
| "loss": 0.3951, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.051064807885997e-06, | |
| "loss": 0.5846, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.9779020603542e-06, | |
| "loss": 0.5642, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.9052663752978676e-06, | |
| "loss": 0.5108, | |
| "step": 1877 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.833157949020706e-06, | |
| "loss": 0.6426, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.761576976401316e-06, | |
| "loss": 0.4332, | |
| "step": 1879 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.6905236508929085e-06, | |
| "loss": 0.5255, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.6199981645226764e-06, | |
| "loss": 0.583, | |
| "step": 1881 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.5500007078911996e-06, | |
| "loss": 0.6851, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.480531470172134e-06, | |
| "loss": 0.5351, | |
| "step": 1883 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.411590639111518e-06, | |
| "loss": 0.6308, | |
| "step": 1884 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.343178401027292e-06, | |
| "loss": 0.623, | |
| "step": 1885 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.275294940808938e-06, | |
| "loss": 0.5406, | |
| "step": 1886 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.207940441916836e-06, | |
| "loss": 0.5566, | |
| "step": 1887 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.1411150863817306e-06, | |
| "loss": 0.6149, | |
| "step": 1888 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.07481905480438e-06, | |
| "loss": 0.6237, | |
| "step": 1889 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.009052526355039e-06, | |
| "loss": 0.4733, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.9438156787728445e-06, | |
| "loss": 0.5993, | |
| "step": 1891 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.8791086883655185e-06, | |
| "loss": 0.6099, | |
| "step": 1892 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.814931730008753e-06, | |
| "loss": 0.4444, | |
| "step": 1893 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.751284977145764e-06, | |
| "loss": 0.6946, | |
| "step": 1894 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.6881686017869123e-06, | |
| "loss": 0.5838, | |
| "step": 1895 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.625582774509173e-06, | |
| "loss": 0.679, | |
| "step": 1896 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.5635276644556015e-06, | |
| "loss": 0.5991, | |
| "step": 1897 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.5020034393349553e-06, | |
| "loss": 0.4727, | |
| "step": 1898 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.4410102654213405e-06, | |
| "loss": 0.5221, | |
| "step": 1899 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.3805483075535217e-06, | |
| "loss": 0.5863, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.3206177291347005e-06, | |
| "loss": 0.5829, | |
| "step": 1901 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.2612186921319176e-06, | |
| "loss": 0.5908, | |
| "step": 1902 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.2023513570757613e-06, | |
| "loss": 0.6148, | |
| "step": 1903 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.144015883059792e-06, | |
| "loss": 0.6208, | |
| "step": 1904 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.086212427740165e-06, | |
| "loss": 0.485, | |
| "step": 1905 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.0289411473352736e-06, | |
| "loss": 0.4281, | |
| "step": 1906 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.97220219662524e-06, | |
| "loss": 0.5932, | |
| "step": 1907 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.9159957289514927e-06, | |
| "loss": 0.5337, | |
| "step": 1908 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.860321896216411e-06, | |
| "loss": 0.4847, | |
| "step": 1909 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.805180848882882e-06, | |
| "loss": 0.6465, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.750572735973922e-06, | |
| "loss": 0.5542, | |
| "step": 1911 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.6964977050722095e-06, | |
| "loss": 0.5421, | |
| "step": 1912 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.6429559023197328e-06, | |
| "loss": 0.4713, | |
| "step": 1913 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.589947472417431e-06, | |
| "loss": 0.5003, | |
| "step": 1914 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.5374725586247093e-06, | |
| "loss": 0.472, | |
| "step": 1915 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.4855313027591478e-06, | |
| "loss": 0.5863, | |
| "step": 1916 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.4341238451960345e-06, | |
| "loss": 0.5076, | |
| "step": 1917 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.383250324868036e-06, | |
| "loss": 0.4888, | |
| "step": 1918 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.332910879264838e-06, | |
| "loss": 0.6435, | |
| "step": 1919 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.2831056444327482e-06, | |
| "loss": 0.7292, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.2338347549742956e-06, | |
| "loss": 0.5122, | |
| "step": 1921 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.185098344047898e-06, | |
| "loss": 0.6734, | |
| "step": 1922 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.136896543367528e-06, | |
| "loss": 0.6434, | |
| "step": 1923 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.0892294832023153e-06, | |
| "loss": 0.6012, | |
| "step": 1924 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.042097292376233e-06, | |
| "loss": 0.5877, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.995500098267722e-06, | |
| "loss": 0.6125, | |
| "step": 1926 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.9494380268093137e-06, | |
| "loss": 0.6093, | |
| "step": 1927 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.903911202487363e-06, | |
| "loss": 0.5182, | |
| "step": 1928 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.8589197483417142e-06, | |
| "loss": 0.5058, | |
| "step": 1929 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.814463785965237e-06, | |
| "loss": 0.568, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.770543435503691e-06, | |
| "loss": 0.4434, | |
| "step": 1931 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7271588156552388e-06, | |
| "loss": 0.5758, | |
| "step": 1932 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.684310043670223e-06, | |
| "loss": 0.6148, | |
| "step": 1933 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.6419972353507895e-06, | |
| "loss": 0.6292, | |
| "step": 1934 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.600220505050598e-06, | |
| "loss": 0.5371, | |
| "step": 1935 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.5589799656745564e-06, | |
| "loss": 0.5317, | |
| "step": 1936 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.5182757286784422e-06, | |
| "loss": 0.5533, | |
| "step": 1937 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.4781079040686374e-06, | |
| "loss": 0.4745, | |
| "step": 1938 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.438476600401839e-06, | |
| "loss": 0.5246, | |
| "step": 1939 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.3993819247847262e-06, | |
| "loss": 0.6321, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.3608239828737157e-06, | |
| "loss": 0.5588, | |
| "step": 1941 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.322802878874696e-06, | |
| "loss": 0.4707, | |
| "step": 1942 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2853187155426271e-06, | |
| "loss": 0.596, | |
| "step": 1943 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2483715941814523e-06, | |
| "loss": 0.5407, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2119616146435863e-06, | |
| "loss": 0.5982, | |
| "step": 1945 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.176088875329917e-06, | |
| "loss": 0.4775, | |
| "step": 1946 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.1407534731892933e-06, | |
| "loss": 0.5262, | |
| "step": 1947 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.1059555037183923e-06, | |
| "loss": 0.5687, | |
| "step": 1948 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.071695060961475e-06, | |
| "loss": 0.4545, | |
| "step": 1949 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.037972237510032e-06, | |
| "loss": 0.5397, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.0047871245026708e-06, | |
| "loss": 0.5624, | |
| "step": 1951 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.721398116247394e-07, | |
| "loss": 0.5166, | |
| "step": 1952 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.400303871081484e-07, | |
| "loss": 0.5667, | |
| "step": 1953 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.084589377311492e-07, | |
| "loss": 0.5282, | |
| "step": 1954 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.774255488180449e-07, | |
| "loss": 0.5333, | |
| "step": 1955 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.469303042389687e-07, | |
| "loss": 0.6061, | |
| "step": 1956 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.169732864097501e-07, | |
| "loss": 0.5567, | |
| "step": 1957 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.875545762915382e-07, | |
| "loss": 0.5602, | |
| "step": 1958 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.586742533907121e-07, | |
| "loss": 0.5493, | |
| "step": 1959 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.30332395758615e-07, | |
| "loss": 0.5494, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 7.025290799913098e-07, | |
| "loss": 0.661, | |
| "step": 1961 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 6.752643812294456e-07, | |
| "loss": 0.5367, | |
| "step": 1962 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 6.485383731580141e-07, | |
| "loss": 0.4913, | |
| "step": 1963 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 6.223511280061267e-07, | |
| "loss": 0.5806, | |
| "step": 1964 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.967027165468375e-07, | |
| "loss": 0.5477, | |
| "step": 1965 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.715932080970543e-07, | |
| "loss": 0.5165, | |
| "step": 1966 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.470226705171167e-07, | |
| "loss": 0.4641, | |
| "step": 1967 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.229911702108626e-07, | |
| "loss": 0.5667, | |
| "step": 1968 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.994987721253175e-07, | |
| "loss": 0.5601, | |
| "step": 1969 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.765455397504948e-07, | |
| "loss": 0.7066, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.541315351192843e-07, | |
| "loss": 0.5087, | |
| "step": 1971 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.322568188073195e-07, | |
| "loss": 0.5452, | |
| "step": 1972 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.10921449932733e-07, | |
| "loss": 0.4866, | |
| "step": 1973 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.9012548615600107e-07, | |
| "loss": 0.5847, | |
| "step": 1974 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.6986898367987743e-07, | |
| "loss": 0.65, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.501519972491263e-07, | |
| "loss": 0.4749, | |
| "step": 1976 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.309745801504338e-07, | |
| "loss": 0.7296, | |
| "step": 1977 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.1233678421229704e-07, | |
| "loss": 0.6065, | |
| "step": 1978 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.9423865980477974e-07, | |
| "loss": 0.595, | |
| "step": 1979 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.7668025583953426e-07, | |
| "loss": 0.6535, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.59661619769469e-07, | |
| "loss": 0.5453, | |
| "step": 1981 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.4318279758879237e-07, | |
| "loss": 0.4655, | |
| "step": 1982 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.2724383383283532e-07, | |
| "loss": 0.5071, | |
| "step": 1983 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.1184477157785153e-07, | |
| "loss": 0.5046, | |
| "step": 1984 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.9698565244106183e-07, | |
| "loss": 0.5824, | |
| "step": 1985 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.8266651658034316e-07, | |
| "loss": 0.595, | |
| "step": 1986 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.688874026943399e-07, | |
| "loss": 0.518, | |
| "step": 1987 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.5564834802213047e-07, | |
| "loss": 0.599, | |
| "step": 1988 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.429493883433386e-07, | |
| "loss": 0.5967, | |
| "step": 1989 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.3079055797786677e-07, | |
| "loss": 0.4572, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.1917188978591842e-07, | |
| "loss": 0.6873, | |
| "step": 1991 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.0809341516784255e-07, | |
| "loss": 0.5536, | |
| "step": 1992 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.755516406408927e-08, | |
| "loss": 0.5285, | |
| "step": 1993 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.7557164955121e-08, | |
| "loss": 0.5842, | |
| "step": 1994 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.80994448613015e-08, | |
| "loss": 0.5293, | |
| "step": 1995 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.918202934285135e-08, | |
| "loss": 0.3683, | |
| "step": 1996 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.080494249978141e-08, | |
| "loss": 0.4227, | |
| "step": 1997 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.2968206971848454e-08, | |
| "loss": 0.5427, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.567184393842183e-08, | |
| "loss": 0.6272, | |
| "step": 1999 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.891587311848355e-08, | |
| "loss": 0.5241, | |
| "step": 2000 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 2011, | |
| "num_train_epochs": 1, | |
| "save_steps": 1000, | |
| "total_flos": 1.390090385400791e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |