| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 32.52244048393391, |
| "global_step": 1000000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.02, |
| "learning_rate": 5.000000000000001e-07, |
| "loss": 10.4451, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 1.0000000000000002e-06, |
| "loss": 10.1476, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1.5e-06, |
| "loss": 9.8078, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 2.0000000000000003e-06, |
| "loss": 9.4816, |
| "step": 2000 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 2.5e-06, |
| "loss": 9.1839, |
| "step": 2500 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 3e-06, |
| "loss": 8.9761, |
| "step": 3000 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 3.5e-06, |
| "loss": 8.863, |
| "step": 3500 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 4.000000000000001e-06, |
| "loss": 8.8057, |
| "step": 4000 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 4.5e-06, |
| "loss": 8.7534, |
| "step": 4500 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 5e-06, |
| "loss": 8.6688, |
| "step": 5000 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 5.500000000000001e-06, |
| "loss": 8.5886, |
| "step": 5500 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 6e-06, |
| "loss": 8.5122, |
| "step": 6000 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 6.5000000000000004e-06, |
| "loss": 8.4138, |
| "step": 6500 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 7e-06, |
| "loss": 8.2859, |
| "step": 7000 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 7.500000000000001e-06, |
| "loss": 8.1421, |
| "step": 7500 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 8.000000000000001e-06, |
| "loss": 7.9556, |
| "step": 8000 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 8.5e-06, |
| "loss": 7.7768, |
| "step": 8500 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 9e-06, |
| "loss": 7.6378, |
| "step": 9000 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 9.5e-06, |
| "loss": 7.5256, |
| "step": 9500 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 1e-05, |
| "loss": 7.4265, |
| "step": 10000 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 9.994949494949497e-06, |
| "loss": 7.3388, |
| "step": 10500 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 9.989898989898991e-06, |
| "loss": 7.2598, |
| "step": 11000 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 9.984848484848485e-06, |
| "loss": 7.1871, |
| "step": 11500 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 9.97979797979798e-06, |
| "loss": 7.118, |
| "step": 12000 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 9.974747474747476e-06, |
| "loss": 7.053, |
| "step": 12500 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 9.96969696969697e-06, |
| "loss": 6.9952, |
| "step": 13000 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 9.964646464646466e-06, |
| "loss": 6.937, |
| "step": 13500 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 9.95959595959596e-06, |
| "loss": 6.8813, |
| "step": 14000 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 9.954545454545456e-06, |
| "loss": 6.8295, |
| "step": 14500 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 9.94949494949495e-06, |
| "loss": 6.7739, |
| "step": 15000 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 9.944444444444445e-06, |
| "loss": 6.7107, |
| "step": 15500 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 9.939393939393939e-06, |
| "loss": 6.6159, |
| "step": 16000 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 9.934343434343435e-06, |
| "loss": 6.4972, |
| "step": 16500 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 9.92929292929293e-06, |
| "loss": 6.3819, |
| "step": 17000 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 9.924242424242425e-06, |
| "loss": 6.2707, |
| "step": 17500 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 9.91919191919192e-06, |
| "loss": 6.1748, |
| "step": 18000 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 9.914141414141416e-06, |
| "loss": 6.0899, |
| "step": 18500 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 9.90909090909091e-06, |
| "loss": 6.0077, |
| "step": 19000 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 9.904040404040404e-06, |
| "loss": 5.93, |
| "step": 19500 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 9.8989898989899e-06, |
| "loss": 5.8585, |
| "step": 20000 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 9.893939393939395e-06, |
| "loss": 5.788, |
| "step": 20500 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 9.88888888888889e-06, |
| "loss": 5.7231, |
| "step": 21000 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 9.883838383838385e-06, |
| "loss": 5.6584, |
| "step": 21500 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 9.87878787878788e-06, |
| "loss": 5.5957, |
| "step": 22000 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 9.873737373737373e-06, |
| "loss": 5.5366, |
| "step": 22500 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 9.86868686868687e-06, |
| "loss": 5.4811, |
| "step": 23000 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 9.863636363636364e-06, |
| "loss": 5.4286, |
| "step": 23500 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 9.85858585858586e-06, |
| "loss": 5.3739, |
| "step": 24000 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 9.853535353535354e-06, |
| "loss": 5.3187, |
| "step": 24500 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 9.84848484848485e-06, |
| "loss": 5.2669, |
| "step": 25000 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 9.843434343434344e-06, |
| "loss": 5.2168, |
| "step": 25500 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 9.838383838383839e-06, |
| "loss": 5.1705, |
| "step": 26000 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 9.833333333333333e-06, |
| "loss": 5.1234, |
| "step": 26500 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 9.828282828282829e-06, |
| "loss": 5.0796, |
| "step": 27000 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 9.823232323232325e-06, |
| "loss": 5.0307, |
| "step": 27500 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 9.81818181818182e-06, |
| "loss": 4.9889, |
| "step": 28000 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 9.813131313131315e-06, |
| "loss": 4.9411, |
| "step": 28500 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 9.80808080808081e-06, |
| "loss": 4.9048, |
| "step": 29000 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 9.803030303030304e-06, |
| "loss": 4.865, |
| "step": 29500 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 9.797979797979798e-06, |
| "loss": 4.8235, |
| "step": 30000 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 9.792929292929294e-06, |
| "loss": 4.7869, |
| "step": 30500 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 9.787878787878788e-06, |
| "loss": 4.7453, |
| "step": 31000 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 9.782828282828284e-06, |
| "loss": 4.7062, |
| "step": 31500 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 9.777777777777779e-06, |
| "loss": 4.6705, |
| "step": 32000 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 9.772727272727273e-06, |
| "loss": 4.636, |
| "step": 32500 |
| }, |
| { |
| "epoch": 1.07, |
| "learning_rate": 9.767676767676767e-06, |
| "loss": 4.5991, |
| "step": 33000 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 9.762626262626263e-06, |
| "loss": 4.5671, |
| "step": 33500 |
| }, |
| { |
| "epoch": 1.11, |
| "learning_rate": 9.757575757575758e-06, |
| "loss": 4.5312, |
| "step": 34000 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 9.752525252525254e-06, |
| "loss": 4.4905, |
| "step": 34500 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 9.747474747474748e-06, |
| "loss": 4.4666, |
| "step": 35000 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 9.742424242424244e-06, |
| "loss": 4.4342, |
| "step": 35500 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 9.737373737373738e-06, |
| "loss": 4.4024, |
| "step": 36000 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 9.732323232323232e-06, |
| "loss": 4.3676, |
| "step": 36500 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 9.727272727272728e-06, |
| "loss": 4.3405, |
| "step": 37000 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 9.722222222222223e-06, |
| "loss": 4.3077, |
| "step": 37500 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 9.717171717171719e-06, |
| "loss": 4.2787, |
| "step": 38000 |
| }, |
| { |
| "epoch": 1.25, |
| "learning_rate": 9.712121212121213e-06, |
| "loss": 4.2466, |
| "step": 38500 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 9.707070707070709e-06, |
| "loss": 4.2114, |
| "step": 39000 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 9.702020202020203e-06, |
| "loss": 4.1992, |
| "step": 39500 |
| }, |
| { |
| "epoch": 1.3, |
| "learning_rate": 9.696969696969698e-06, |
| "loss": 4.1587, |
| "step": 40000 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 9.691919191919192e-06, |
| "loss": 4.1242, |
| "step": 40500 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 9.686868686868688e-06, |
| "loss": 4.0931, |
| "step": 41000 |
| }, |
| { |
| "epoch": 1.35, |
| "learning_rate": 9.681818181818182e-06, |
| "loss": 4.0625, |
| "step": 41500 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 9.676767676767678e-06, |
| "loss": 4.0334, |
| "step": 42000 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 9.671717171717172e-06, |
| "loss": 3.9965, |
| "step": 42500 |
| }, |
| { |
| "epoch": 1.4, |
| "learning_rate": 9.666666666666667e-06, |
| "loss": 3.9718, |
| "step": 43000 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 9.661616161616163e-06, |
| "loss": 3.9466, |
| "step": 43500 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 9.656565656565657e-06, |
| "loss": 3.9127, |
| "step": 44000 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 9.651515151515153e-06, |
| "loss": 3.8866, |
| "step": 44500 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 9.646464646464647e-06, |
| "loss": 3.856, |
| "step": 45000 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 9.641414141414143e-06, |
| "loss": 3.8336, |
| "step": 45500 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 9.636363636363638e-06, |
| "loss": 3.8032, |
| "step": 46000 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 9.631313131313132e-06, |
| "loss": 3.7802, |
| "step": 46500 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 9.626262626262626e-06, |
| "loss": 3.7482, |
| "step": 47000 |
| }, |
| { |
| "epoch": 1.54, |
| "learning_rate": 9.621212121212122e-06, |
| "loss": 3.7237, |
| "step": 47500 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 9.616161616161616e-06, |
| "loss": 3.701, |
| "step": 48000 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 9.611111111111112e-06, |
| "loss": 3.6887, |
| "step": 48500 |
| }, |
| { |
| "epoch": 1.59, |
| "learning_rate": 9.606060606060607e-06, |
| "loss": 3.6558, |
| "step": 49000 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 9.601010101010103e-06, |
| "loss": 3.625, |
| "step": 49500 |
| }, |
| { |
| "epoch": 1.63, |
| "learning_rate": 9.595959595959597e-06, |
| "loss": 3.6123, |
| "step": 50000 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 9.590909090909091e-06, |
| "loss": 3.5798, |
| "step": 50500 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 9.585858585858586e-06, |
| "loss": 3.5625, |
| "step": 51000 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 9.580808080808082e-06, |
| "loss": 3.5352, |
| "step": 51500 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 9.575757575757576e-06, |
| "loss": 3.5153, |
| "step": 52000 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 9.570707070707072e-06, |
| "loss": 3.492, |
| "step": 52500 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 9.565656565656566e-06, |
| "loss": 3.4653, |
| "step": 53000 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 9.56060606060606e-06, |
| "loss": 3.4461, |
| "step": 53500 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 9.555555555555556e-06, |
| "loss": 3.4219, |
| "step": 54000 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 9.55050505050505e-06, |
| "loss": 3.4078, |
| "step": 54500 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 9.545454545454547e-06, |
| "loss": 3.3775, |
| "step": 55000 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 9.540404040404041e-06, |
| "loss": 3.3512, |
| "step": 55500 |
| }, |
| { |
| "epoch": 1.82, |
| "learning_rate": 9.535353535353537e-06, |
| "loss": 3.3344, |
| "step": 56000 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 9.530303030303031e-06, |
| "loss": 3.3091, |
| "step": 56500 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 9.525252525252526e-06, |
| "loss": 3.2932, |
| "step": 57000 |
| }, |
| { |
| "epoch": 1.87, |
| "learning_rate": 9.52020202020202e-06, |
| "loss": 3.2704, |
| "step": 57500 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 9.515151515151516e-06, |
| "loss": 3.249, |
| "step": 58000 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 9.51010101010101e-06, |
| "loss": 3.2267, |
| "step": 58500 |
| }, |
| { |
| "epoch": 1.92, |
| "learning_rate": 9.505050505050506e-06, |
| "loss": 3.2072, |
| "step": 59000 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 9.5e-06, |
| "loss": 3.1906, |
| "step": 59500 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 9.494949494949497e-06, |
| "loss": 3.1637, |
| "step": 60000 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 9.48989898989899e-06, |
| "loss": 3.1467, |
| "step": 60500 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 9.484848484848485e-06, |
| "loss": 3.1268, |
| "step": 61000 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 9.479797979797981e-06, |
| "loss": 3.1122, |
| "step": 61500 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 9.474747474747475e-06, |
| "loss": 3.0908, |
| "step": 62000 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 9.469696969696971e-06, |
| "loss": 3.0709, |
| "step": 62500 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 9.464646464646466e-06, |
| "loss": 3.0544, |
| "step": 63000 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 9.45959595959596e-06, |
| "loss": 3.0413, |
| "step": 63500 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 9.454545454545456e-06, |
| "loss": 3.0243, |
| "step": 64000 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 9.44949494949495e-06, |
| "loss": 2.9992, |
| "step": 64500 |
| }, |
| { |
| "epoch": 2.11, |
| "learning_rate": 9.444444444444445e-06, |
| "loss": 2.9798, |
| "step": 65000 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 9.43939393939394e-06, |
| "loss": 2.9626, |
| "step": 65500 |
| }, |
| { |
| "epoch": 2.15, |
| "learning_rate": 9.434343434343435e-06, |
| "loss": 2.9465, |
| "step": 66000 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 9.42929292929293e-06, |
| "loss": 2.927, |
| "step": 66500 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 9.424242424242425e-06, |
| "loss": 2.9166, |
| "step": 67000 |
| }, |
| { |
| "epoch": 2.2, |
| "learning_rate": 9.41919191919192e-06, |
| "loss": 2.9035, |
| "step": 67500 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 9.414141414141414e-06, |
| "loss": 2.8826, |
| "step": 68000 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 9.40909090909091e-06, |
| "loss": 2.8696, |
| "step": 68500 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 9.404040404040404e-06, |
| "loss": 2.8516, |
| "step": 69000 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 9.3989898989899e-06, |
| "loss": 2.8337, |
| "step": 69500 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 9.393939393939396e-06, |
| "loss": 2.82, |
| "step": 70000 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 9.38888888888889e-06, |
| "loss": 2.8056, |
| "step": 70500 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 9.383838383838385e-06, |
| "loss": 2.7897, |
| "step": 71000 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 9.378787878787879e-06, |
| "loss": 2.7751, |
| "step": 71500 |
| }, |
| { |
| "epoch": 2.34, |
| "learning_rate": 9.373737373737375e-06, |
| "loss": 2.7571, |
| "step": 72000 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 9.36868686868687e-06, |
| "loss": 2.7505, |
| "step": 72500 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 9.363636363636365e-06, |
| "loss": 2.736, |
| "step": 73000 |
| }, |
| { |
| "epoch": 2.39, |
| "learning_rate": 9.35858585858586e-06, |
| "loss": 2.7176, |
| "step": 73500 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 9.353535353535354e-06, |
| "loss": 2.7094, |
| "step": 74000 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 9.34848484848485e-06, |
| "loss": 2.6899, |
| "step": 74500 |
| }, |
| { |
| "epoch": 2.44, |
| "learning_rate": 9.343434343434344e-06, |
| "loss": 2.6751, |
| "step": 75000 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 9.338383838383838e-06, |
| "loss": 2.6588, |
| "step": 75500 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 9.333333333333334e-06, |
| "loss": 2.6447, |
| "step": 76000 |
| }, |
| { |
| "epoch": 2.49, |
| "learning_rate": 9.328282828282829e-06, |
| "loss": 2.6373, |
| "step": 76500 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 9.323232323232325e-06, |
| "loss": 2.6218, |
| "step": 77000 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 9.318181818181819e-06, |
| "loss": 2.6093, |
| "step": 77500 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 9.313131313131313e-06, |
| "loss": 2.5918, |
| "step": 78000 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 9.30808080808081e-06, |
| "loss": 2.5866, |
| "step": 78500 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 9.303030303030303e-06, |
| "loss": 2.5693, |
| "step": 79000 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 9.2979797979798e-06, |
| "loss": 2.5579, |
| "step": 79500 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 9.292929292929294e-06, |
| "loss": 2.5494, |
| "step": 80000 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 9.28787878787879e-06, |
| "loss": 2.5327, |
| "step": 80500 |
| }, |
| { |
| "epoch": 2.63, |
| "learning_rate": 9.282828282828284e-06, |
| "loss": 2.5212, |
| "step": 81000 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 9.277777777777778e-06, |
| "loss": 2.5063, |
| "step": 81500 |
| }, |
| { |
| "epoch": 2.67, |
| "learning_rate": 9.272727272727273e-06, |
| "loss": 2.4969, |
| "step": 82000 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 9.267676767676769e-06, |
| "loss": 2.4869, |
| "step": 82500 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 9.262626262626263e-06, |
| "loss": 2.4717, |
| "step": 83000 |
| }, |
| { |
| "epoch": 2.72, |
| "learning_rate": 9.257575757575759e-06, |
| "loss": 2.4601, |
| "step": 83500 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 9.252525252525253e-06, |
| "loss": 2.4501, |
| "step": 84000 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 9.24747474747475e-06, |
| "loss": 2.4375, |
| "step": 84500 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 9.242424242424244e-06, |
| "loss": 2.4225, |
| "step": 85000 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 9.237373737373738e-06, |
| "loss": 2.4121, |
| "step": 85500 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 9.232323232323232e-06, |
| "loss": 2.4009, |
| "step": 86000 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 9.227272727272728e-06, |
| "loss": 2.3894, |
| "step": 86500 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 9.222222222222224e-06, |
| "loss": 2.3715, |
| "step": 87000 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 9.217171717171718e-06, |
| "loss": 2.3563, |
| "step": 87500 |
| }, |
| { |
| "epoch": 2.86, |
| "learning_rate": 9.212121212121213e-06, |
| "loss": 2.3484, |
| "step": 88000 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 9.207070707070707e-06, |
| "loss": 2.3344, |
| "step": 88500 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 9.202020202020203e-06, |
| "loss": 2.3213, |
| "step": 89000 |
| }, |
| { |
| "epoch": 2.91, |
| "learning_rate": 9.196969696969697e-06, |
| "loss": 2.3136, |
| "step": 89500 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 9.191919191919193e-06, |
| "loss": 2.2937, |
| "step": 90000 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 9.186868686868688e-06, |
| "loss": 2.2817, |
| "step": 90500 |
| }, |
| { |
| "epoch": 2.96, |
| "learning_rate": 9.181818181818184e-06, |
| "loss": 2.2733, |
| "step": 91000 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 9.176767676767678e-06, |
| "loss": 2.2603, |
| "step": 91500 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 9.171717171717172e-06, |
| "loss": 2.2498, |
| "step": 92000 |
| }, |
| { |
| "epoch": 3.01, |
| "learning_rate": 9.166666666666666e-06, |
| "loss": 2.2374, |
| "step": 92500 |
| }, |
| { |
| "epoch": 3.02, |
| "learning_rate": 9.161616161616162e-06, |
| "loss": 2.2276, |
| "step": 93000 |
| }, |
| { |
| "epoch": 3.04, |
| "learning_rate": 9.156565656565657e-06, |
| "loss": 2.2111, |
| "step": 93500 |
| }, |
| { |
| "epoch": 3.06, |
| "learning_rate": 9.151515151515153e-06, |
| "loss": 2.2021, |
| "step": 94000 |
| }, |
| { |
| "epoch": 3.07, |
| "learning_rate": 9.146464646464647e-06, |
| "loss": 2.1931, |
| "step": 94500 |
| }, |
| { |
| "epoch": 3.09, |
| "learning_rate": 9.141414141414143e-06, |
| "loss": 2.1768, |
| "step": 95000 |
| }, |
| { |
| "epoch": 3.11, |
| "learning_rate": 9.136363636363637e-06, |
| "loss": 2.1694, |
| "step": 95500 |
| }, |
| { |
| "epoch": 3.12, |
| "learning_rate": 9.131313131313132e-06, |
| "loss": 2.158, |
| "step": 96000 |
| }, |
| { |
| "epoch": 3.14, |
| "learning_rate": 9.126262626262628e-06, |
| "loss": 2.1448, |
| "step": 96500 |
| }, |
| { |
| "epoch": 3.15, |
| "learning_rate": 9.121212121212122e-06, |
| "loss": 2.1369, |
| "step": 97000 |
| }, |
| { |
| "epoch": 3.17, |
| "learning_rate": 9.116161616161618e-06, |
| "loss": 2.1274, |
| "step": 97500 |
| }, |
| { |
| "epoch": 3.19, |
| "learning_rate": 9.111111111111112e-06, |
| "loss": 2.1111, |
| "step": 98000 |
| }, |
| { |
| "epoch": 3.2, |
| "learning_rate": 9.106060606060606e-06, |
| "loss": 2.1003, |
| "step": 98500 |
| }, |
| { |
| "epoch": 3.22, |
| "learning_rate": 9.1010101010101e-06, |
| "loss": 2.0863, |
| "step": 99000 |
| }, |
| { |
| "epoch": 3.24, |
| "learning_rate": 9.095959595959597e-06, |
| "loss": 2.079, |
| "step": 99500 |
| }, |
| { |
| "epoch": 3.25, |
| "learning_rate": 9.090909090909091e-06, |
| "loss": 2.0716, |
| "step": 100000 |
| }, |
| { |
| "epoch": 3.27, |
| "learning_rate": 9.085858585858587e-06, |
| "loss": 2.0595, |
| "step": 100500 |
| }, |
| { |
| "epoch": 3.28, |
| "learning_rate": 9.080808080808081e-06, |
| "loss": 2.0456, |
| "step": 101000 |
| }, |
| { |
| "epoch": 3.3, |
| "learning_rate": 9.075757575757577e-06, |
| "loss": 2.0385, |
| "step": 101500 |
| }, |
| { |
| "epoch": 3.32, |
| "learning_rate": 9.070707070707072e-06, |
| "loss": 2.0254, |
| "step": 102000 |
| }, |
| { |
| "epoch": 3.33, |
| "learning_rate": 9.065656565656566e-06, |
| "loss": 2.0141, |
| "step": 102500 |
| }, |
| { |
| "epoch": 3.35, |
| "learning_rate": 9.06060606060606e-06, |
| "loss": 2.0025, |
| "step": 103000 |
| }, |
| { |
| "epoch": 3.37, |
| "learning_rate": 9.055555555555556e-06, |
| "loss": 1.993, |
| "step": 103500 |
| }, |
| { |
| "epoch": 3.38, |
| "learning_rate": 9.050505050505052e-06, |
| "loss": 1.9813, |
| "step": 104000 |
| }, |
| { |
| "epoch": 3.4, |
| "learning_rate": 9.045454545454546e-06, |
| "loss": 1.9709, |
| "step": 104500 |
| }, |
| { |
| "epoch": 3.41, |
| "learning_rate": 9.040404040404042e-06, |
| "loss": 1.9632, |
| "step": 105000 |
| }, |
| { |
| "epoch": 3.43, |
| "learning_rate": 9.035353535353537e-06, |
| "loss": 1.9532, |
| "step": 105500 |
| }, |
| { |
| "epoch": 3.45, |
| "learning_rate": 9.030303030303031e-06, |
| "loss": 1.9382, |
| "step": 106000 |
| }, |
| { |
| "epoch": 3.46, |
| "learning_rate": 9.025252525252525e-06, |
| "loss": 1.9306, |
| "step": 106500 |
| }, |
| { |
| "epoch": 3.48, |
| "learning_rate": 9.020202020202021e-06, |
| "loss": 1.919, |
| "step": 107000 |
| }, |
| { |
| "epoch": 3.5, |
| "learning_rate": 9.015151515151516e-06, |
| "loss": 1.9093, |
| "step": 107500 |
| }, |
| { |
| "epoch": 3.51, |
| "learning_rate": 9.010101010101012e-06, |
| "loss": 1.9006, |
| "step": 108000 |
| }, |
| { |
| "epoch": 3.53, |
| "learning_rate": 9.005050505050506e-06, |
| "loss": 1.8927, |
| "step": 108500 |
| }, |
| { |
| "epoch": 3.54, |
| "learning_rate": 9e-06, |
| "loss": 1.88, |
| "step": 109000 |
| }, |
| { |
| "epoch": 3.56, |
| "learning_rate": 8.994949494949495e-06, |
| "loss": 1.8681, |
| "step": 109500 |
| }, |
| { |
| "epoch": 3.58, |
| "learning_rate": 8.98989898989899e-06, |
| "loss": 1.8623, |
| "step": 110000 |
| }, |
| { |
| "epoch": 3.59, |
| "learning_rate": 8.984848484848485e-06, |
| "loss": 1.8515, |
| "step": 110500 |
| }, |
| { |
| "epoch": 3.61, |
| "learning_rate": 8.97979797979798e-06, |
| "loss": 1.8424, |
| "step": 111000 |
| }, |
| { |
| "epoch": 3.63, |
| "learning_rate": 8.974747474747475e-06, |
| "loss": 1.8322, |
| "step": 111500 |
| }, |
| { |
| "epoch": 3.64, |
| "learning_rate": 8.969696969696971e-06, |
| "loss": 1.8241, |
| "step": 112000 |
| }, |
| { |
| "epoch": 3.66, |
| "learning_rate": 8.964646464646465e-06, |
| "loss": 1.8149, |
| "step": 112500 |
| }, |
| { |
| "epoch": 3.68, |
| "learning_rate": 8.95959595959596e-06, |
| "loss": 1.8053, |
| "step": 113000 |
| }, |
| { |
| "epoch": 3.69, |
| "learning_rate": 8.954545454545456e-06, |
| "loss": 1.7971, |
| "step": 113500 |
| }, |
| { |
| "epoch": 3.71, |
| "learning_rate": 8.94949494949495e-06, |
| "loss": 1.789, |
| "step": 114000 |
| }, |
| { |
| "epoch": 3.72, |
| "learning_rate": 8.944444444444446e-06, |
| "loss": 1.7804, |
| "step": 114500 |
| }, |
| { |
| "epoch": 3.74, |
| "learning_rate": 8.93939393939394e-06, |
| "loss": 1.7719, |
| "step": 115000 |
| }, |
| { |
| "epoch": 3.76, |
| "learning_rate": 8.934343434343436e-06, |
| "loss": 1.7608, |
| "step": 115500 |
| }, |
| { |
| "epoch": 3.77, |
| "learning_rate": 8.92929292929293e-06, |
| "loss": 1.7512, |
| "step": 116000 |
| }, |
| { |
| "epoch": 3.79, |
| "learning_rate": 8.924242424242425e-06, |
| "loss": 1.7448, |
| "step": 116500 |
| }, |
| { |
| "epoch": 3.81, |
| "learning_rate": 8.919191919191919e-06, |
| "loss": 1.7366, |
| "step": 117000 |
| }, |
| { |
| "epoch": 3.82, |
| "learning_rate": 8.914141414141415e-06, |
| "loss": 1.7252, |
| "step": 117500 |
| }, |
| { |
| "epoch": 3.84, |
| "learning_rate": 8.90909090909091e-06, |
| "loss": 1.7187, |
| "step": 118000 |
| }, |
| { |
| "epoch": 3.85, |
| "learning_rate": 8.904040404040405e-06, |
| "loss": 1.7151, |
| "step": 118500 |
| }, |
| { |
| "epoch": 3.87, |
| "learning_rate": 8.8989898989899e-06, |
| "loss": 1.7092, |
| "step": 119000 |
| }, |
| { |
| "epoch": 3.89, |
| "learning_rate": 8.893939393939394e-06, |
| "loss": 1.6961, |
| "step": 119500 |
| }, |
| { |
| "epoch": 3.9, |
| "learning_rate": 8.888888888888888e-06, |
| "loss": 1.6896, |
| "step": 120000 |
| }, |
| { |
| "epoch": 3.92, |
| "learning_rate": 8.883838383838384e-06, |
| "loss": 1.6813, |
| "step": 120500 |
| }, |
| { |
| "epoch": 3.94, |
| "learning_rate": 8.87878787878788e-06, |
| "loss": 1.6744, |
| "step": 121000 |
| }, |
| { |
| "epoch": 3.95, |
| "learning_rate": 8.873737373737375e-06, |
| "loss": 1.6657, |
| "step": 121500 |
| }, |
| { |
| "epoch": 3.97, |
| "learning_rate": 8.86868686868687e-06, |
| "loss": 1.659, |
| "step": 122000 |
| }, |
| { |
| "epoch": 3.98, |
| "learning_rate": 8.863636363636365e-06, |
| "loss": 1.6521, |
| "step": 122500 |
| }, |
| { |
| "epoch": 4.0, |
| "learning_rate": 8.85858585858586e-06, |
| "loss": 1.6431, |
| "step": 123000 |
| }, |
| { |
| "epoch": 4.02, |
| "learning_rate": 8.853535353535353e-06, |
| "loss": 1.6365, |
| "step": 123500 |
| }, |
| { |
| "epoch": 4.03, |
| "learning_rate": 8.84848484848485e-06, |
| "loss": 1.6282, |
| "step": 124000 |
| }, |
| { |
| "epoch": 4.05, |
| "learning_rate": 8.843434343434344e-06, |
| "loss": 1.622, |
| "step": 124500 |
| }, |
| { |
| "epoch": 4.07, |
| "learning_rate": 8.83838383838384e-06, |
| "loss": 1.6115, |
| "step": 125000 |
| }, |
| { |
| "epoch": 4.08, |
| "learning_rate": 8.833333333333334e-06, |
| "loss": 1.6057, |
| "step": 125500 |
| }, |
| { |
| "epoch": 4.1, |
| "learning_rate": 8.82828282828283e-06, |
| "loss": 1.5967, |
| "step": 126000 |
| }, |
| { |
| "epoch": 4.11, |
| "learning_rate": 8.823232323232324e-06, |
| "loss": 1.5929, |
| "step": 126500 |
| }, |
| { |
| "epoch": 4.13, |
| "learning_rate": 8.818181818181819e-06, |
| "loss": 1.585, |
| "step": 127000 |
| }, |
| { |
| "epoch": 4.15, |
| "learning_rate": 8.813131313131313e-06, |
| "loss": 1.5804, |
| "step": 127500 |
| }, |
| { |
| "epoch": 4.16, |
| "learning_rate": 8.808080808080809e-06, |
| "loss": 1.572, |
| "step": 128000 |
| }, |
| { |
| "epoch": 4.18, |
| "learning_rate": 8.803030303030303e-06, |
| "loss": 1.569, |
| "step": 128500 |
| }, |
| { |
| "epoch": 4.2, |
| "learning_rate": 8.7979797979798e-06, |
| "loss": 1.5566, |
| "step": 129000 |
| }, |
| { |
| "epoch": 4.21, |
| "learning_rate": 8.792929292929293e-06, |
| "loss": 1.5525, |
| "step": 129500 |
| }, |
| { |
| "epoch": 4.23, |
| "learning_rate": 8.787878787878788e-06, |
| "loss": 1.544, |
| "step": 130000 |
| }, |
| { |
| "epoch": 4.24, |
| "learning_rate": 8.782828282828284e-06, |
| "loss": 1.5399, |
| "step": 130500 |
| }, |
| { |
| "epoch": 4.26, |
| "learning_rate": 8.777777777777778e-06, |
| "loss": 1.534, |
| "step": 131000 |
| }, |
| { |
| "epoch": 4.28, |
| "learning_rate": 8.772727272727274e-06, |
| "loss": 1.5271, |
| "step": 131500 |
| }, |
| { |
| "epoch": 4.29, |
| "learning_rate": 8.767676767676768e-06, |
| "loss": 1.5183, |
| "step": 132000 |
| }, |
| { |
| "epoch": 4.31, |
| "learning_rate": 8.762626262626264e-06, |
| "loss": 1.513, |
| "step": 132500 |
| }, |
| { |
| "epoch": 4.33, |
| "learning_rate": 8.757575757575759e-06, |
| "loss": 1.5109, |
| "step": 133000 |
| }, |
| { |
| "epoch": 4.34, |
| "learning_rate": 8.752525252525253e-06, |
| "loss": 1.5012, |
| "step": 133500 |
| }, |
| { |
| "epoch": 4.36, |
| "learning_rate": 8.747474747474747e-06, |
| "loss": 1.4959, |
| "step": 134000 |
| }, |
| { |
| "epoch": 4.37, |
| "learning_rate": 8.742424242424243e-06, |
| "loss": 1.4882, |
| "step": 134500 |
| }, |
| { |
| "epoch": 4.39, |
| "learning_rate": 8.737373737373738e-06, |
| "loss": 1.4823, |
| "step": 135000 |
| }, |
| { |
| "epoch": 4.41, |
| "learning_rate": 8.732323232323234e-06, |
| "loss": 1.4805, |
| "step": 135500 |
| }, |
| { |
| "epoch": 4.42, |
| "learning_rate": 8.727272727272728e-06, |
| "loss": 1.4717, |
| "step": 136000 |
| }, |
| { |
| "epoch": 4.44, |
| "learning_rate": 8.722222222222224e-06, |
| "loss": 1.4684, |
| "step": 136500 |
| }, |
| { |
| "epoch": 4.46, |
| "learning_rate": 8.717171717171718e-06, |
| "loss": 1.4615, |
| "step": 137000 |
| }, |
| { |
| "epoch": 4.47, |
| "learning_rate": 8.712121212121212e-06, |
| "loss": 1.4562, |
| "step": 137500 |
| }, |
| { |
| "epoch": 4.49, |
| "learning_rate": 8.707070707070707e-06, |
| "loss": 1.4514, |
| "step": 138000 |
| }, |
| { |
| "epoch": 4.5, |
| "learning_rate": 8.702020202020203e-06, |
| "loss": 1.4456, |
| "step": 138500 |
| }, |
| { |
| "epoch": 4.52, |
| "learning_rate": 8.696969696969699e-06, |
| "loss": 1.4383, |
| "step": 139000 |
| }, |
| { |
| "epoch": 4.54, |
| "learning_rate": 8.691919191919193e-06, |
| "loss": 1.4344, |
| "step": 139500 |
| }, |
| { |
| "epoch": 4.55, |
| "learning_rate": 8.686868686868687e-06, |
| "loss": 1.431, |
| "step": 140000 |
| }, |
| { |
| "epoch": 4.57, |
| "learning_rate": 8.681818181818182e-06, |
| "loss": 1.4245, |
| "step": 140500 |
| }, |
| { |
| "epoch": 4.59, |
| "learning_rate": 8.676767676767678e-06, |
| "loss": 1.4198, |
| "step": 141000 |
| }, |
| { |
| "epoch": 4.6, |
| "learning_rate": 8.671717171717172e-06, |
| "loss": 1.4127, |
| "step": 141500 |
| }, |
| { |
| "epoch": 4.62, |
| "learning_rate": 8.666666666666668e-06, |
| "loss": 1.4075, |
| "step": 142000 |
| }, |
| { |
| "epoch": 4.63, |
| "learning_rate": 8.661616161616162e-06, |
| "loss": 1.405, |
| "step": 142500 |
| }, |
| { |
| "epoch": 4.65, |
| "learning_rate": 8.656565656565658e-06, |
| "loss": 1.4003, |
| "step": 143000 |
| }, |
| { |
| "epoch": 4.67, |
| "learning_rate": 8.651515151515152e-06, |
| "loss": 1.394, |
| "step": 143500 |
| }, |
| { |
| "epoch": 4.68, |
| "learning_rate": 8.646464646464647e-06, |
| "loss": 1.39, |
| "step": 144000 |
| }, |
| { |
| "epoch": 4.7, |
| "learning_rate": 8.641414141414141e-06, |
| "loss": 1.3857, |
| "step": 144500 |
| }, |
| { |
| "epoch": 4.72, |
| "learning_rate": 8.636363636363637e-06, |
| "loss": 1.3816, |
| "step": 145000 |
| }, |
| { |
| "epoch": 4.73, |
| "learning_rate": 8.631313131313131e-06, |
| "loss": 1.3764, |
| "step": 145500 |
| }, |
| { |
| "epoch": 4.75, |
| "learning_rate": 8.626262626262627e-06, |
| "loss": 1.3719, |
| "step": 146000 |
| }, |
| { |
| "epoch": 4.76, |
| "learning_rate": 8.621212121212122e-06, |
| "loss": 1.3675, |
| "step": 146500 |
| }, |
| { |
| "epoch": 4.78, |
| "learning_rate": 8.616161616161618e-06, |
| "loss": 1.3663, |
| "step": 147000 |
| }, |
| { |
| "epoch": 4.8, |
| "learning_rate": 8.611111111111112e-06, |
| "loss": 1.3599, |
| "step": 147500 |
| }, |
| { |
| "epoch": 4.81, |
| "learning_rate": 8.606060606060606e-06, |
| "loss": 1.3566, |
| "step": 148000 |
| }, |
| { |
| "epoch": 4.83, |
| "learning_rate": 8.601010101010102e-06, |
| "loss": 1.3528, |
| "step": 148500 |
| }, |
| { |
| "epoch": 4.85, |
| "learning_rate": 8.595959595959596e-06, |
| "loss": 1.3466, |
| "step": 149000 |
| }, |
| { |
| "epoch": 4.86, |
| "learning_rate": 8.590909090909092e-06, |
| "loss": 1.344, |
| "step": 149500 |
| }, |
| { |
| "epoch": 4.88, |
| "learning_rate": 8.585858585858587e-06, |
| "loss": 1.3414, |
| "step": 150000 |
| }, |
| { |
| "epoch": 4.89, |
| "learning_rate": 8.580808080808081e-06, |
| "loss": 1.3367, |
| "step": 150500 |
| }, |
| { |
| "epoch": 4.91, |
| "learning_rate": 8.575757575757575e-06, |
| "loss": 1.3321, |
| "step": 151000 |
| }, |
| { |
| "epoch": 4.93, |
| "learning_rate": 8.570707070707071e-06, |
| "loss": 1.3288, |
| "step": 151500 |
| }, |
| { |
| "epoch": 4.94, |
| "learning_rate": 8.565656565656566e-06, |
| "loss": 1.3233, |
| "step": 152000 |
| }, |
| { |
| "epoch": 4.96, |
| "learning_rate": 8.560606060606062e-06, |
| "loss": 1.3211, |
| "step": 152500 |
| }, |
| { |
| "epoch": 4.98, |
| "learning_rate": 8.555555555555556e-06, |
| "loss": 1.3145, |
| "step": 153000 |
| }, |
| { |
| "epoch": 4.99, |
| "learning_rate": 8.550505050505052e-06, |
| "loss": 1.3113, |
| "step": 153500 |
| }, |
| { |
| "epoch": 5.01, |
| "learning_rate": 8.545454545454546e-06, |
| "loss": 1.3084, |
| "step": 154000 |
| }, |
| { |
| "epoch": 5.02, |
| "learning_rate": 8.54040404040404e-06, |
| "loss": 1.3065, |
| "step": 154500 |
| }, |
| { |
| "epoch": 5.04, |
| "learning_rate": 8.535353535353535e-06, |
| "loss": 1.3017, |
| "step": 155000 |
| }, |
| { |
| "epoch": 5.06, |
| "learning_rate": 8.53030303030303e-06, |
| "loss": 1.2981, |
| "step": 155500 |
| }, |
| { |
| "epoch": 5.07, |
| "learning_rate": 8.525252525252527e-06, |
| "loss": 1.2935, |
| "step": 156000 |
| }, |
| { |
| "epoch": 5.09, |
| "learning_rate": 8.520202020202021e-06, |
| "loss": 1.2921, |
| "step": 156500 |
| }, |
| { |
| "epoch": 5.11, |
| "learning_rate": 8.515151515151517e-06, |
| "loss": 1.2873, |
| "step": 157000 |
| }, |
| { |
| "epoch": 5.12, |
| "learning_rate": 8.510101010101011e-06, |
| "loss": 1.2847, |
| "step": 157500 |
| }, |
| { |
| "epoch": 5.14, |
| "learning_rate": 8.505050505050506e-06, |
| "loss": 1.2801, |
| "step": 158000 |
| }, |
| { |
| "epoch": 5.15, |
| "learning_rate": 8.5e-06, |
| "loss": 1.2767, |
| "step": 158500 |
| }, |
| { |
| "epoch": 5.17, |
| "learning_rate": 8.494949494949496e-06, |
| "loss": 1.2743, |
| "step": 159000 |
| }, |
| { |
| "epoch": 5.19, |
| "learning_rate": 8.48989898989899e-06, |
| "loss": 1.2707, |
| "step": 159500 |
| }, |
| { |
| "epoch": 5.2, |
| "learning_rate": 8.484848484848486e-06, |
| "loss": 1.2664, |
| "step": 160000 |
| }, |
| { |
| "epoch": 5.22, |
| "learning_rate": 8.47979797979798e-06, |
| "loss": 1.2654, |
| "step": 160500 |
| }, |
| { |
| "epoch": 5.24, |
| "learning_rate": 8.474747474747475e-06, |
| "loss": 1.2622, |
| "step": 161000 |
| }, |
| { |
| "epoch": 5.25, |
| "learning_rate": 8.46969696969697e-06, |
| "loss": 1.258, |
| "step": 161500 |
| }, |
| { |
| "epoch": 5.27, |
| "learning_rate": 8.464646464646465e-06, |
| "loss": 1.254, |
| "step": 162000 |
| }, |
| { |
| "epoch": 5.28, |
| "learning_rate": 8.45959595959596e-06, |
| "loss": 1.2513, |
| "step": 162500 |
| }, |
| { |
| "epoch": 5.3, |
| "learning_rate": 8.454545454545455e-06, |
| "loss": 1.2492, |
| "step": 163000 |
| }, |
| { |
| "epoch": 5.32, |
| "learning_rate": 8.44949494949495e-06, |
| "loss": 1.2449, |
| "step": 163500 |
| }, |
| { |
| "epoch": 5.33, |
| "learning_rate": 8.444444444444446e-06, |
| "loss": 1.2428, |
| "step": 164000 |
| }, |
| { |
| "epoch": 5.35, |
| "learning_rate": 8.43939393939394e-06, |
| "loss": 1.2373, |
| "step": 164500 |
| }, |
| { |
| "epoch": 5.37, |
| "learning_rate": 8.434343434343434e-06, |
| "loss": 1.2362, |
| "step": 165000 |
| }, |
| { |
| "epoch": 5.38, |
| "learning_rate": 8.42929292929293e-06, |
| "loss": 1.2341, |
| "step": 165500 |
| }, |
| { |
| "epoch": 5.4, |
| "learning_rate": 8.424242424242425e-06, |
| "loss": 1.2298, |
| "step": 166000 |
| }, |
| { |
| "epoch": 5.41, |
| "learning_rate": 8.41919191919192e-06, |
| "loss": 1.2279, |
| "step": 166500 |
| }, |
| { |
| "epoch": 5.43, |
| "learning_rate": 8.414141414141415e-06, |
| "loss": 1.2244, |
| "step": 167000 |
| }, |
| { |
| "epoch": 5.45, |
| "learning_rate": 8.40909090909091e-06, |
| "loss": 1.221, |
| "step": 167500 |
| }, |
| { |
| "epoch": 5.46, |
| "learning_rate": 8.404040404040405e-06, |
| "loss": 1.2183, |
| "step": 168000 |
| }, |
| { |
| "epoch": 5.48, |
| "learning_rate": 8.3989898989899e-06, |
| "loss": 1.2159, |
| "step": 168500 |
| }, |
| { |
| "epoch": 5.5, |
| "learning_rate": 8.393939393939394e-06, |
| "loss": 1.2118, |
| "step": 169000 |
| }, |
| { |
| "epoch": 5.51, |
| "learning_rate": 8.38888888888889e-06, |
| "loss": 1.2101, |
| "step": 169500 |
| }, |
| { |
| "epoch": 5.53, |
| "learning_rate": 8.383838383838384e-06, |
| "loss": 1.2067, |
| "step": 170000 |
| }, |
| { |
| "epoch": 5.55, |
| "learning_rate": 8.37878787878788e-06, |
| "loss": 1.204, |
| "step": 170500 |
| }, |
| { |
| "epoch": 5.56, |
| "learning_rate": 8.373737373737374e-06, |
| "loss": 1.2018, |
| "step": 171000 |
| }, |
| { |
| "epoch": 5.58, |
| "learning_rate": 8.368686868686869e-06, |
| "loss": 1.1997, |
| "step": 171500 |
| }, |
| { |
| "epoch": 5.59, |
| "learning_rate": 8.363636363636365e-06, |
| "loss": 1.1964, |
| "step": 172000 |
| }, |
| { |
| "epoch": 5.61, |
| "learning_rate": 8.358585858585859e-06, |
| "loss": 1.1936, |
| "step": 172500 |
| }, |
| { |
| "epoch": 5.63, |
| "learning_rate": 8.353535353535355e-06, |
| "loss": 1.1923, |
| "step": 173000 |
| }, |
| { |
| "epoch": 5.64, |
| "learning_rate": 8.348484848484849e-06, |
| "loss": 1.189, |
| "step": 173500 |
| }, |
| { |
| "epoch": 5.66, |
| "learning_rate": 8.343434343434345e-06, |
| "loss": 1.1857, |
| "step": 174000 |
| }, |
| { |
| "epoch": 5.68, |
| "learning_rate": 8.33838383838384e-06, |
| "loss": 1.1831, |
| "step": 174500 |
| }, |
| { |
| "epoch": 5.69, |
| "learning_rate": 8.333333333333334e-06, |
| "loss": 1.1813, |
| "step": 175000 |
| }, |
| { |
| "epoch": 5.71, |
| "learning_rate": 8.328282828282828e-06, |
| "loss": 1.1794, |
| "step": 175500 |
| }, |
| { |
| "epoch": 5.72, |
| "learning_rate": 8.323232323232324e-06, |
| "loss": 1.1751, |
| "step": 176000 |
| }, |
| { |
| "epoch": 5.74, |
| "learning_rate": 8.318181818181818e-06, |
| "loss": 1.1736, |
| "step": 176500 |
| }, |
| { |
| "epoch": 5.76, |
| "learning_rate": 8.313131313131314e-06, |
| "loss": 1.1716, |
| "step": 177000 |
| }, |
| { |
| "epoch": 5.77, |
| "learning_rate": 8.308080808080809e-06, |
| "loss": 1.1689, |
| "step": 177500 |
| }, |
| { |
| "epoch": 5.79, |
| "learning_rate": 8.303030303030305e-06, |
| "loss": 1.1675, |
| "step": 178000 |
| }, |
| { |
| "epoch": 5.81, |
| "learning_rate": 8.297979797979799e-06, |
| "loss": 1.1652, |
| "step": 178500 |
| }, |
| { |
| "epoch": 5.82, |
| "learning_rate": 8.292929292929293e-06, |
| "loss": 1.163, |
| "step": 179000 |
| }, |
| { |
| "epoch": 5.84, |
| "learning_rate": 8.287878787878787e-06, |
| "loss": 1.1602, |
| "step": 179500 |
| }, |
| { |
| "epoch": 5.85, |
| "learning_rate": 8.282828282828283e-06, |
| "loss": 1.1577, |
| "step": 180000 |
| }, |
| { |
| "epoch": 5.87, |
| "learning_rate": 8.277777777777778e-06, |
| "loss": 1.1565, |
| "step": 180500 |
| }, |
| { |
| "epoch": 5.89, |
| "learning_rate": 8.272727272727274e-06, |
| "loss": 1.1525, |
| "step": 181000 |
| }, |
| { |
| "epoch": 5.9, |
| "learning_rate": 8.267676767676768e-06, |
| "loss": 1.152, |
| "step": 181500 |
| }, |
| { |
| "epoch": 5.92, |
| "learning_rate": 8.262626262626264e-06, |
| "loss": 1.1506, |
| "step": 182000 |
| }, |
| { |
| "epoch": 5.94, |
| "learning_rate": 8.257575757575758e-06, |
| "loss": 1.1469, |
| "step": 182500 |
| }, |
| { |
| "epoch": 5.95, |
| "learning_rate": 8.252525252525253e-06, |
| "loss": 1.1449, |
| "step": 183000 |
| }, |
| { |
| "epoch": 5.97, |
| "learning_rate": 8.247474747474749e-06, |
| "loss": 1.1422, |
| "step": 183500 |
| }, |
| { |
| "epoch": 5.98, |
| "learning_rate": 8.242424242424243e-06, |
| "loss": 1.1404, |
| "step": 184000 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 8.237373737373739e-06, |
| "loss": 1.1403, |
| "step": 184500 |
| }, |
| { |
| "epoch": 6.02, |
| "learning_rate": 8.232323232323233e-06, |
| "loss": 1.1371, |
| "step": 185000 |
| }, |
| { |
| "epoch": 6.03, |
| "learning_rate": 8.227272727272728e-06, |
| "loss": 1.1372, |
| "step": 185500 |
| }, |
| { |
| "epoch": 6.05, |
| "learning_rate": 8.222222222222222e-06, |
| "loss": 1.1341, |
| "step": 186000 |
| }, |
| { |
| "epoch": 6.07, |
| "learning_rate": 8.217171717171718e-06, |
| "loss": 1.1302, |
| "step": 186500 |
| }, |
| { |
| "epoch": 6.08, |
| "learning_rate": 8.212121212121212e-06, |
| "loss": 1.1303, |
| "step": 187000 |
| }, |
| { |
| "epoch": 6.1, |
| "learning_rate": 8.207070707070708e-06, |
| "loss": 1.1276, |
| "step": 187500 |
| }, |
| { |
| "epoch": 6.11, |
| "learning_rate": 8.202020202020202e-06, |
| "loss": 1.1267, |
| "step": 188000 |
| }, |
| { |
| "epoch": 6.13, |
| "learning_rate": 8.196969696969698e-06, |
| "loss": 1.123, |
| "step": 188500 |
| }, |
| { |
| "epoch": 6.15, |
| "learning_rate": 8.191919191919193e-06, |
| "loss": 1.1219, |
| "step": 189000 |
| }, |
| { |
| "epoch": 6.16, |
| "learning_rate": 8.186868686868687e-06, |
| "loss": 1.119, |
| "step": 189500 |
| }, |
| { |
| "epoch": 6.18, |
| "learning_rate": 8.181818181818183e-06, |
| "loss": 1.1191, |
| "step": 190000 |
| }, |
| { |
| "epoch": 6.2, |
| "learning_rate": 8.176767676767677e-06, |
| "loss": 1.1166, |
| "step": 190500 |
| }, |
| { |
| "epoch": 6.21, |
| "learning_rate": 8.171717171717173e-06, |
| "loss": 1.114, |
| "step": 191000 |
| }, |
| { |
| "epoch": 6.23, |
| "learning_rate": 8.166666666666668e-06, |
| "loss": 1.113, |
| "step": 191500 |
| }, |
| { |
| "epoch": 6.24, |
| "learning_rate": 8.161616161616162e-06, |
| "loss": 1.1112, |
| "step": 192000 |
| }, |
| { |
| "epoch": 6.26, |
| "learning_rate": 8.156565656565658e-06, |
| "loss": 1.1104, |
| "step": 192500 |
| }, |
| { |
| "epoch": 6.28, |
| "learning_rate": 8.151515151515152e-06, |
| "loss": 1.1083, |
| "step": 193000 |
| }, |
| { |
| "epoch": 6.29, |
| "learning_rate": 8.146464646464646e-06, |
| "loss": 1.1073, |
| "step": 193500 |
| }, |
| { |
| "epoch": 6.31, |
| "learning_rate": 8.141414141414142e-06, |
| "loss": 1.1051, |
| "step": 194000 |
| }, |
| { |
| "epoch": 6.33, |
| "learning_rate": 8.136363636363637e-06, |
| "loss": 1.1036, |
| "step": 194500 |
| }, |
| { |
| "epoch": 6.34, |
| "learning_rate": 8.131313131313133e-06, |
| "loss": 1.102, |
| "step": 195000 |
| }, |
| { |
| "epoch": 6.36, |
| "learning_rate": 8.126262626262627e-06, |
| "loss": 1.0998, |
| "step": 195500 |
| }, |
| { |
| "epoch": 6.37, |
| "learning_rate": 8.121212121212121e-06, |
| "loss": 1.0982, |
| "step": 196000 |
| }, |
| { |
| "epoch": 6.39, |
| "learning_rate": 8.116161616161616e-06, |
| "loss": 1.0972, |
| "step": 196500 |
| }, |
| { |
| "epoch": 6.41, |
| "learning_rate": 8.111111111111112e-06, |
| "loss": 1.0951, |
| "step": 197000 |
| }, |
| { |
| "epoch": 6.42, |
| "learning_rate": 8.106060606060606e-06, |
| "loss": 1.0942, |
| "step": 197500 |
| }, |
| { |
| "epoch": 6.44, |
| "learning_rate": 8.101010101010102e-06, |
| "loss": 1.0922, |
| "step": 198000 |
| }, |
| { |
| "epoch": 6.46, |
| "learning_rate": 8.095959595959598e-06, |
| "loss": 1.0919, |
| "step": 198500 |
| }, |
| { |
| "epoch": 6.47, |
| "learning_rate": 8.090909090909092e-06, |
| "loss": 1.0898, |
| "step": 199000 |
| }, |
| { |
| "epoch": 6.49, |
| "learning_rate": 8.085858585858586e-06, |
| "loss": 1.0885, |
| "step": 199500 |
| }, |
| { |
| "epoch": 6.5, |
| "learning_rate": 8.08080808080808e-06, |
| "loss": 1.0876, |
| "step": 200000 |
| }, |
| { |
| "epoch": 6.52, |
| "learning_rate": 8.075757575757577e-06, |
| "loss": 1.0854, |
| "step": 200500 |
| }, |
| { |
| "epoch": 6.54, |
| "learning_rate": 8.070707070707071e-06, |
| "loss": 1.0846, |
| "step": 201000 |
| }, |
| { |
| "epoch": 6.55, |
| "learning_rate": 8.065656565656567e-06, |
| "loss": 1.0831, |
| "step": 201500 |
| }, |
| { |
| "epoch": 6.57, |
| "learning_rate": 8.060606060606061e-06, |
| "loss": 1.0814, |
| "step": 202000 |
| }, |
| { |
| "epoch": 6.59, |
| "learning_rate": 8.055555555555557e-06, |
| "loss": 1.0799, |
| "step": 202500 |
| }, |
| { |
| "epoch": 6.6, |
| "learning_rate": 8.050505050505052e-06, |
| "loss": 1.0787, |
| "step": 203000 |
| }, |
| { |
| "epoch": 6.62, |
| "learning_rate": 8.045454545454546e-06, |
| "loss": 1.0768, |
| "step": 203500 |
| }, |
| { |
| "epoch": 6.63, |
| "learning_rate": 8.04040404040404e-06, |
| "loss": 1.0763, |
| "step": 204000 |
| }, |
| { |
| "epoch": 6.65, |
| "learning_rate": 8.035353535353536e-06, |
| "loss": 1.0751, |
| "step": 204500 |
| }, |
| { |
| "epoch": 6.67, |
| "learning_rate": 8.03030303030303e-06, |
| "loss": 1.0738, |
| "step": 205000 |
| }, |
| { |
| "epoch": 6.68, |
| "learning_rate": 8.025252525252526e-06, |
| "loss": 1.0708, |
| "step": 205500 |
| }, |
| { |
| "epoch": 6.7, |
| "learning_rate": 8.02020202020202e-06, |
| "loss": 1.0701, |
| "step": 206000 |
| }, |
| { |
| "epoch": 6.72, |
| "learning_rate": 8.015151515151515e-06, |
| "loss": 1.0708, |
| "step": 206500 |
| }, |
| { |
| "epoch": 6.73, |
| "learning_rate": 8.010101010101011e-06, |
| "loss": 1.0684, |
| "step": 207000 |
| }, |
| { |
| "epoch": 6.75, |
| "learning_rate": 8.005050505050505e-06, |
| "loss": 1.0668, |
| "step": 207500 |
| }, |
| { |
| "epoch": 6.76, |
| "learning_rate": 8.000000000000001e-06, |
| "loss": 1.0668, |
| "step": 208000 |
| }, |
| { |
| "epoch": 6.78, |
| "learning_rate": 7.994949494949496e-06, |
| "loss": 1.0646, |
| "step": 208500 |
| }, |
| { |
| "epoch": 6.8, |
| "learning_rate": 7.989898989898992e-06, |
| "loss": 1.0641, |
| "step": 209000 |
| }, |
| { |
| "epoch": 6.81, |
| "learning_rate": 7.984848484848486e-06, |
| "loss": 1.0613, |
| "step": 209500 |
| }, |
| { |
| "epoch": 6.83, |
| "learning_rate": 7.97979797979798e-06, |
| "loss": 1.0613, |
| "step": 210000 |
| }, |
| { |
| "epoch": 6.85, |
| "learning_rate": 7.974747474747475e-06, |
| "loss": 1.0604, |
| "step": 210500 |
| }, |
| { |
| "epoch": 6.86, |
| "learning_rate": 7.96969696969697e-06, |
| "loss": 1.0593, |
| "step": 211000 |
| }, |
| { |
| "epoch": 6.88, |
| "learning_rate": 7.964646464646465e-06, |
| "loss": 1.0575, |
| "step": 211500 |
| }, |
| { |
| "epoch": 6.89, |
| "learning_rate": 7.95959595959596e-06, |
| "loss": 1.0564, |
| "step": 212000 |
| }, |
| { |
| "epoch": 6.91, |
| "learning_rate": 7.954545454545455e-06, |
| "loss": 1.0554, |
| "step": 212500 |
| }, |
| { |
| "epoch": 6.93, |
| "learning_rate": 7.949494949494951e-06, |
| "loss": 1.0544, |
| "step": 213000 |
| }, |
| { |
| "epoch": 6.94, |
| "learning_rate": 7.944444444444445e-06, |
| "loss": 1.053, |
| "step": 213500 |
| }, |
| { |
| "epoch": 6.96, |
| "learning_rate": 7.93939393939394e-06, |
| "loss": 1.0526, |
| "step": 214000 |
| }, |
| { |
| "epoch": 6.98, |
| "learning_rate": 7.934343434343434e-06, |
| "loss": 1.0503, |
| "step": 214500 |
| }, |
| { |
| "epoch": 6.99, |
| "learning_rate": 7.92929292929293e-06, |
| "loss": 1.0505, |
| "step": 215000 |
| }, |
| { |
| "epoch": 7.01, |
| "learning_rate": 7.924242424242426e-06, |
| "loss": 1.0482, |
| "step": 215500 |
| }, |
| { |
| "epoch": 7.02, |
| "learning_rate": 7.91919191919192e-06, |
| "loss": 1.0478, |
| "step": 216000 |
| }, |
| { |
| "epoch": 7.04, |
| "learning_rate": 7.914141414141415e-06, |
| "loss": 1.0466, |
| "step": 216500 |
| }, |
| { |
| "epoch": 7.06, |
| "learning_rate": 7.909090909090909e-06, |
| "loss": 1.0461, |
| "step": 217000 |
| }, |
| { |
| "epoch": 7.07, |
| "learning_rate": 7.904040404040405e-06, |
| "loss": 1.0442, |
| "step": 217500 |
| }, |
| { |
| "epoch": 7.09, |
| "learning_rate": 7.898989898989899e-06, |
| "loss": 1.0436, |
| "step": 218000 |
| }, |
| { |
| "epoch": 7.11, |
| "learning_rate": 7.893939393939395e-06, |
| "loss": 1.0426, |
| "step": 218500 |
| }, |
| { |
| "epoch": 7.12, |
| "learning_rate": 7.88888888888889e-06, |
| "loss": 1.043, |
| "step": 219000 |
| }, |
| { |
| "epoch": 7.14, |
| "learning_rate": 7.883838383838385e-06, |
| "loss": 1.0406, |
| "step": 219500 |
| }, |
| { |
| "epoch": 7.15, |
| "learning_rate": 7.87878787878788e-06, |
| "loss": 1.0398, |
| "step": 220000 |
| }, |
| { |
| "epoch": 7.17, |
| "learning_rate": 7.873737373737374e-06, |
| "loss": 1.0386, |
| "step": 220500 |
| }, |
| { |
| "epoch": 7.19, |
| "learning_rate": 7.868686868686868e-06, |
| "loss": 1.0379, |
| "step": 221000 |
| }, |
| { |
| "epoch": 7.2, |
| "learning_rate": 7.863636363636364e-06, |
| "loss": 1.0377, |
| "step": 221500 |
| }, |
| { |
| "epoch": 7.22, |
| "learning_rate": 7.858585858585859e-06, |
| "loss": 1.0362, |
| "step": 222000 |
| }, |
| { |
| "epoch": 7.24, |
| "learning_rate": 7.853535353535355e-06, |
| "loss": 1.0357, |
| "step": 222500 |
| }, |
| { |
| "epoch": 7.25, |
| "learning_rate": 7.848484848484849e-06, |
| "loss": 1.0342, |
| "step": 223000 |
| }, |
| { |
| "epoch": 7.27, |
| "learning_rate": 7.843434343434345e-06, |
| "loss": 1.0333, |
| "step": 223500 |
| }, |
| { |
| "epoch": 7.29, |
| "learning_rate": 7.838383838383839e-06, |
| "loss": 1.0331, |
| "step": 224000 |
| }, |
| { |
| "epoch": 7.3, |
| "learning_rate": 7.833333333333333e-06, |
| "loss": 1.0314, |
| "step": 224500 |
| }, |
| { |
| "epoch": 7.32, |
| "learning_rate": 7.82828282828283e-06, |
| "loss": 1.0313, |
| "step": 225000 |
| }, |
| { |
| "epoch": 7.33, |
| "learning_rate": 7.823232323232324e-06, |
| "loss": 1.0294, |
| "step": 225500 |
| }, |
| { |
| "epoch": 7.35, |
| "learning_rate": 7.81818181818182e-06, |
| "loss": 1.0285, |
| "step": 226000 |
| }, |
| { |
| "epoch": 7.37, |
| "learning_rate": 7.813131313131314e-06, |
| "loss": 1.0283, |
| "step": 226500 |
| }, |
| { |
| "epoch": 7.38, |
| "learning_rate": 7.808080808080808e-06, |
| "loss": 1.0278, |
| "step": 227000 |
| }, |
| { |
| "epoch": 7.4, |
| "learning_rate": 7.803030303030303e-06, |
| "loss": 1.0265, |
| "step": 227500 |
| }, |
| { |
| "epoch": 7.42, |
| "learning_rate": 7.797979797979799e-06, |
| "loss": 1.0255, |
| "step": 228000 |
| }, |
| { |
| "epoch": 7.43, |
| "learning_rate": 7.792929292929293e-06, |
| "loss": 1.0243, |
| "step": 228500 |
| }, |
| { |
| "epoch": 7.45, |
| "learning_rate": 7.787878787878789e-06, |
| "loss": 1.0229, |
| "step": 229000 |
| }, |
| { |
| "epoch": 7.46, |
| "learning_rate": 7.782828282828283e-06, |
| "loss": 1.0222, |
| "step": 229500 |
| }, |
| { |
| "epoch": 7.48, |
| "learning_rate": 7.77777777777778e-06, |
| "loss": 1.0219, |
| "step": 230000 |
| }, |
| { |
| "epoch": 7.5, |
| "learning_rate": 7.772727272727273e-06, |
| "loss": 1.0224, |
| "step": 230500 |
| }, |
| { |
| "epoch": 7.51, |
| "learning_rate": 7.767676767676768e-06, |
| "loss": 1.0212, |
| "step": 231000 |
| }, |
| { |
| "epoch": 7.53, |
| "learning_rate": 7.762626262626262e-06, |
| "loss": 1.0203, |
| "step": 231500 |
| }, |
| { |
| "epoch": 7.55, |
| "learning_rate": 7.757575757575758e-06, |
| "loss": 1.0185, |
| "step": 232000 |
| }, |
| { |
| "epoch": 7.56, |
| "learning_rate": 7.752525252525254e-06, |
| "loss": 1.0183, |
| "step": 232500 |
| }, |
| { |
| "epoch": 7.58, |
| "learning_rate": 7.747474747474748e-06, |
| "loss": 1.0165, |
| "step": 233000 |
| }, |
| { |
| "epoch": 7.59, |
| "learning_rate": 7.742424242424244e-06, |
| "loss": 1.0161, |
| "step": 233500 |
| }, |
| { |
| "epoch": 7.61, |
| "learning_rate": 7.737373737373739e-06, |
| "loss": 1.0157, |
| "step": 234000 |
| }, |
| { |
| "epoch": 7.63, |
| "learning_rate": 7.732323232323233e-06, |
| "loss": 1.0154, |
| "step": 234500 |
| }, |
| { |
| "epoch": 7.64, |
| "learning_rate": 7.727272727272727e-06, |
| "loss": 1.0134, |
| "step": 235000 |
| }, |
| { |
| "epoch": 7.66, |
| "learning_rate": 7.722222222222223e-06, |
| "loss": 1.0125, |
| "step": 235500 |
| }, |
| { |
| "epoch": 7.68, |
| "learning_rate": 7.717171717171717e-06, |
| "loss": 1.0127, |
| "step": 236000 |
| }, |
| { |
| "epoch": 7.69, |
| "learning_rate": 7.712121212121213e-06, |
| "loss": 1.0119, |
| "step": 236500 |
| }, |
| { |
| "epoch": 7.71, |
| "learning_rate": 7.707070707070708e-06, |
| "loss": 1.0101, |
| "step": 237000 |
| }, |
| { |
| "epoch": 7.72, |
| "learning_rate": 7.702020202020202e-06, |
| "loss": 1.0104, |
| "step": 237500 |
| }, |
| { |
| "epoch": 7.74, |
| "learning_rate": 7.696969696969696e-06, |
| "loss": 1.0098, |
| "step": 238000 |
| }, |
| { |
| "epoch": 7.76, |
| "learning_rate": 7.691919191919192e-06, |
| "loss": 1.0094, |
| "step": 238500 |
| }, |
| { |
| "epoch": 7.77, |
| "learning_rate": 7.686868686868687e-06, |
| "loss": 1.0075, |
| "step": 239000 |
| }, |
| { |
| "epoch": 7.79, |
| "learning_rate": 7.681818181818183e-06, |
| "loss": 1.0078, |
| "step": 239500 |
| }, |
| { |
| "epoch": 7.81, |
| "learning_rate": 7.676767676767677e-06, |
| "loss": 1.0068, |
| "step": 240000 |
| }, |
| { |
| "epoch": 7.82, |
| "learning_rate": 7.671717171717173e-06, |
| "loss": 1.0074, |
| "step": 240500 |
| }, |
| { |
| "epoch": 7.84, |
| "learning_rate": 7.666666666666667e-06, |
| "loss": 1.0056, |
| "step": 241000 |
| }, |
| { |
| "epoch": 7.85, |
| "learning_rate": 7.661616161616162e-06, |
| "loss": 1.0044, |
| "step": 241500 |
| }, |
| { |
| "epoch": 7.87, |
| "learning_rate": 7.656565656565658e-06, |
| "loss": 1.0039, |
| "step": 242000 |
| }, |
| { |
| "epoch": 7.89, |
| "learning_rate": 7.651515151515152e-06, |
| "loss": 1.0028, |
| "step": 242500 |
| }, |
| { |
| "epoch": 7.9, |
| "learning_rate": 7.646464646464648e-06, |
| "loss": 1.0032, |
| "step": 243000 |
| }, |
| { |
| "epoch": 7.92, |
| "learning_rate": 7.641414141414142e-06, |
| "loss": 1.0012, |
| "step": 243500 |
| }, |
| { |
| "epoch": 7.94, |
| "learning_rate": 7.636363636363638e-06, |
| "loss": 1.0012, |
| "step": 244000 |
| }, |
| { |
| "epoch": 7.95, |
| "learning_rate": 7.631313131313132e-06, |
| "loss": 1.0008, |
| "step": 244500 |
| }, |
| { |
| "epoch": 7.97, |
| "learning_rate": 7.6262626262626275e-06, |
| "loss": 0.9994, |
| "step": 245000 |
| }, |
| { |
| "epoch": 7.98, |
| "learning_rate": 7.621212121212122e-06, |
| "loss": 0.9993, |
| "step": 245500 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 7.616161616161617e-06, |
| "loss": 0.9978, |
| "step": 246000 |
| }, |
| { |
| "epoch": 8.02, |
| "learning_rate": 7.611111111111111e-06, |
| "loss": 0.9977, |
| "step": 246500 |
| }, |
| { |
| "epoch": 8.03, |
| "learning_rate": 7.606060606060606e-06, |
| "loss": 0.9968, |
| "step": 247000 |
| }, |
| { |
| "epoch": 8.05, |
| "learning_rate": 7.6010101010101016e-06, |
| "loss": 0.9956, |
| "step": 247500 |
| }, |
| { |
| "epoch": 8.07, |
| "learning_rate": 7.595959595959597e-06, |
| "loss": 0.9955, |
| "step": 248000 |
| }, |
| { |
| "epoch": 8.08, |
| "learning_rate": 7.590909090909091e-06, |
| "loss": 0.9951, |
| "step": 248500 |
| }, |
| { |
| "epoch": 8.1, |
| "learning_rate": 7.585858585858586e-06, |
| "loss": 0.9949, |
| "step": 249000 |
| }, |
| { |
| "epoch": 8.11, |
| "learning_rate": 7.580808080808082e-06, |
| "loss": 0.9926, |
| "step": 249500 |
| }, |
| { |
| "epoch": 8.13, |
| "learning_rate": 7.5757575757575764e-06, |
| "loss": 0.9937, |
| "step": 250000 |
| }, |
| { |
| "epoch": 8.15, |
| "learning_rate": 7.5707070707070716e-06, |
| "loss": 0.9924, |
| "step": 250500 |
| }, |
| { |
| "epoch": 8.16, |
| "learning_rate": 7.565656565656566e-06, |
| "loss": 0.9919, |
| "step": 251000 |
| }, |
| { |
| "epoch": 8.18, |
| "learning_rate": 7.560606060606062e-06, |
| "loss": 0.9913, |
| "step": 251500 |
| }, |
| { |
| "epoch": 8.2, |
| "learning_rate": 7.555555555555556e-06, |
| "loss": 0.9907, |
| "step": 252000 |
| }, |
| { |
| "epoch": 8.21, |
| "learning_rate": 7.550505050505051e-06, |
| "loss": 0.9898, |
| "step": 252500 |
| }, |
| { |
| "epoch": 8.23, |
| "learning_rate": 7.545454545454546e-06, |
| "loss": 0.9891, |
| "step": 253000 |
| }, |
| { |
| "epoch": 8.24, |
| "learning_rate": 7.540404040404042e-06, |
| "loss": 0.9891, |
| "step": 253500 |
| }, |
| { |
| "epoch": 8.26, |
| "learning_rate": 7.535353535353536e-06, |
| "loss": 0.9882, |
| "step": 254000 |
| }, |
| { |
| "epoch": 8.28, |
| "learning_rate": 7.530303030303031e-06, |
| "loss": 0.987, |
| "step": 254500 |
| }, |
| { |
| "epoch": 8.29, |
| "learning_rate": 7.525252525252525e-06, |
| "loss": 0.987, |
| "step": 255000 |
| }, |
| { |
| "epoch": 8.31, |
| "learning_rate": 7.520202020202021e-06, |
| "loss": 0.9869, |
| "step": 255500 |
| }, |
| { |
| "epoch": 8.33, |
| "learning_rate": 7.515151515151516e-06, |
| "loss": 0.9846, |
| "step": 256000 |
| }, |
| { |
| "epoch": 8.34, |
| "learning_rate": 7.510101010101011e-06, |
| "loss": 0.9855, |
| "step": 256500 |
| }, |
| { |
| "epoch": 8.36, |
| "learning_rate": 7.505050505050505e-06, |
| "loss": 0.9838, |
| "step": 257000 |
| }, |
| { |
| "epoch": 8.37, |
| "learning_rate": 7.500000000000001e-06, |
| "loss": 0.9844, |
| "step": 257500 |
| }, |
| { |
| "epoch": 8.39, |
| "learning_rate": 7.494949494949496e-06, |
| "loss": 0.9836, |
| "step": 258000 |
| }, |
| { |
| "epoch": 8.41, |
| "learning_rate": 7.4898989898989905e-06, |
| "loss": 0.9824, |
| "step": 258500 |
| }, |
| { |
| "epoch": 8.42, |
| "learning_rate": 7.484848484848486e-06, |
| "loss": 0.9816, |
| "step": 259000 |
| }, |
| { |
| "epoch": 8.44, |
| "learning_rate": 7.47979797979798e-06, |
| "loss": 0.9821, |
| "step": 259500 |
| }, |
| { |
| "epoch": 8.46, |
| "learning_rate": 7.474747474747476e-06, |
| "loss": 0.981, |
| "step": 260000 |
| }, |
| { |
| "epoch": 8.47, |
| "learning_rate": 7.46969696969697e-06, |
| "loss": 0.9804, |
| "step": 260500 |
| }, |
| { |
| "epoch": 8.49, |
| "learning_rate": 7.464646464646465e-06, |
| "loss": 0.9799, |
| "step": 261000 |
| }, |
| { |
| "epoch": 8.5, |
| "learning_rate": 7.45959595959596e-06, |
| "loss": 0.9785, |
| "step": 261500 |
| }, |
| { |
| "epoch": 8.52, |
| "learning_rate": 7.454545454545456e-06, |
| "loss": 0.9784, |
| "step": 262000 |
| }, |
| { |
| "epoch": 8.54, |
| "learning_rate": 7.44949494949495e-06, |
| "loss": 0.9782, |
| "step": 262500 |
| }, |
| { |
| "epoch": 8.55, |
| "learning_rate": 7.444444444444445e-06, |
| "loss": 0.977, |
| "step": 263000 |
| }, |
| { |
| "epoch": 8.57, |
| "learning_rate": 7.439393939393939e-06, |
| "loss": 0.9768, |
| "step": 263500 |
| }, |
| { |
| "epoch": 8.59, |
| "learning_rate": 7.434343434343435e-06, |
| "loss": 0.9761, |
| "step": 264000 |
| }, |
| { |
| "epoch": 8.6, |
| "learning_rate": 7.42929292929293e-06, |
| "loss": 0.9753, |
| "step": 264500 |
| }, |
| { |
| "epoch": 8.62, |
| "learning_rate": 7.424242424242425e-06, |
| "loss": 0.9747, |
| "step": 265000 |
| }, |
| { |
| "epoch": 8.63, |
| "learning_rate": 7.419191919191919e-06, |
| "loss": 0.9745, |
| "step": 265500 |
| }, |
| { |
| "epoch": 8.65, |
| "learning_rate": 7.414141414141415e-06, |
| "loss": 0.9743, |
| "step": 266000 |
| }, |
| { |
| "epoch": 8.67, |
| "learning_rate": 7.40909090909091e-06, |
| "loss": 0.9732, |
| "step": 266500 |
| }, |
| { |
| "epoch": 8.68, |
| "learning_rate": 7.4040404040404045e-06, |
| "loss": 0.9734, |
| "step": 267000 |
| }, |
| { |
| "epoch": 8.7, |
| "learning_rate": 7.3989898989899e-06, |
| "loss": 0.9727, |
| "step": 267500 |
| }, |
| { |
| "epoch": 8.72, |
| "learning_rate": 7.393939393939395e-06, |
| "loss": 0.9712, |
| "step": 268000 |
| }, |
| { |
| "epoch": 8.73, |
| "learning_rate": 7.38888888888889e-06, |
| "loss": 0.9707, |
| "step": 268500 |
| }, |
| { |
| "epoch": 8.75, |
| "learning_rate": 7.383838383838384e-06, |
| "loss": 0.9708, |
| "step": 269000 |
| }, |
| { |
| "epoch": 8.76, |
| "learning_rate": 7.378787878787879e-06, |
| "loss": 0.9704, |
| "step": 269500 |
| }, |
| { |
| "epoch": 8.78, |
| "learning_rate": 7.373737373737374e-06, |
| "loss": 0.9695, |
| "step": 270000 |
| }, |
| { |
| "epoch": 8.8, |
| "learning_rate": 7.36868686868687e-06, |
| "loss": 0.9694, |
| "step": 270500 |
| }, |
| { |
| "epoch": 8.81, |
| "learning_rate": 7.363636363636364e-06, |
| "loss": 0.9689, |
| "step": 271000 |
| }, |
| { |
| "epoch": 8.83, |
| "learning_rate": 7.358585858585859e-06, |
| "loss": 0.9683, |
| "step": 271500 |
| }, |
| { |
| "epoch": 8.85, |
| "learning_rate": 7.353535353535353e-06, |
| "loss": 0.9667, |
| "step": 272000 |
| }, |
| { |
| "epoch": 8.86, |
| "learning_rate": 7.348484848484849e-06, |
| "loss": 0.9673, |
| "step": 272500 |
| }, |
| { |
| "epoch": 8.88, |
| "learning_rate": 7.343434343434344e-06, |
| "loss": 0.9663, |
| "step": 273000 |
| }, |
| { |
| "epoch": 8.89, |
| "learning_rate": 7.338383838383839e-06, |
| "loss": 0.9662, |
| "step": 273500 |
| }, |
| { |
| "epoch": 8.91, |
| "learning_rate": 7.333333333333333e-06, |
| "loss": 0.9654, |
| "step": 274000 |
| }, |
| { |
| "epoch": 8.93, |
| "learning_rate": 7.328282828282829e-06, |
| "loss": 0.9646, |
| "step": 274500 |
| }, |
| { |
| "epoch": 8.94, |
| "learning_rate": 7.323232323232324e-06, |
| "loss": 0.9645, |
| "step": 275000 |
| }, |
| { |
| "epoch": 8.96, |
| "learning_rate": 7.3181818181818186e-06, |
| "loss": 0.9641, |
| "step": 275500 |
| }, |
| { |
| "epoch": 8.98, |
| "learning_rate": 7.3131313131313146e-06, |
| "loss": 0.9641, |
| "step": 276000 |
| }, |
| { |
| "epoch": 8.99, |
| "learning_rate": 7.308080808080809e-06, |
| "loss": 0.9626, |
| "step": 276500 |
| }, |
| { |
| "epoch": 9.01, |
| "learning_rate": 7.303030303030304e-06, |
| "loss": 0.9622, |
| "step": 277000 |
| }, |
| { |
| "epoch": 9.02, |
| "learning_rate": 7.297979797979798e-06, |
| "loss": 0.9629, |
| "step": 277500 |
| }, |
| { |
| "epoch": 9.04, |
| "learning_rate": 7.2929292929292934e-06, |
| "loss": 0.9619, |
| "step": 278000 |
| }, |
| { |
| "epoch": 9.06, |
| "learning_rate": 7.287878787878789e-06, |
| "loss": 0.9613, |
| "step": 278500 |
| }, |
| { |
| "epoch": 9.07, |
| "learning_rate": 7.282828282828284e-06, |
| "loss": 0.9609, |
| "step": 279000 |
| }, |
| { |
| "epoch": 9.09, |
| "learning_rate": 7.277777777777778e-06, |
| "loss": 0.9594, |
| "step": 279500 |
| }, |
| { |
| "epoch": 9.11, |
| "learning_rate": 7.272727272727273e-06, |
| "loss": 0.9596, |
| "step": 280000 |
| }, |
| { |
| "epoch": 9.12, |
| "learning_rate": 7.2676767676767675e-06, |
| "loss": 0.9599, |
| "step": 280500 |
| }, |
| { |
| "epoch": 9.14, |
| "learning_rate": 7.2626262626262635e-06, |
| "loss": 0.9588, |
| "step": 281000 |
| }, |
| { |
| "epoch": 9.16, |
| "learning_rate": 7.257575757575758e-06, |
| "loss": 0.9583, |
| "step": 281500 |
| }, |
| { |
| "epoch": 9.17, |
| "learning_rate": 7.252525252525253e-06, |
| "loss": 0.9579, |
| "step": 282000 |
| }, |
| { |
| "epoch": 9.19, |
| "learning_rate": 7.247474747474747e-06, |
| "loss": 0.957, |
| "step": 282500 |
| }, |
| { |
| "epoch": 9.2, |
| "learning_rate": 7.242424242424243e-06, |
| "loss": 0.958, |
| "step": 283000 |
| }, |
| { |
| "epoch": 9.22, |
| "learning_rate": 7.237373737373738e-06, |
| "loss": 0.9561, |
| "step": 283500 |
| }, |
| { |
| "epoch": 9.24, |
| "learning_rate": 7.232323232323233e-06, |
| "loss": 0.9555, |
| "step": 284000 |
| }, |
| { |
| "epoch": 9.25, |
| "learning_rate": 7.227272727272729e-06, |
| "loss": 0.9558, |
| "step": 284500 |
| }, |
| { |
| "epoch": 9.27, |
| "learning_rate": 7.222222222222223e-06, |
| "loss": 0.9542, |
| "step": 285000 |
| }, |
| { |
| "epoch": 9.29, |
| "learning_rate": 7.217171717171718e-06, |
| "loss": 0.9543, |
| "step": 285500 |
| }, |
| { |
| "epoch": 9.3, |
| "learning_rate": 7.212121212121212e-06, |
| "loss": 0.9542, |
| "step": 286000 |
| }, |
| { |
| "epoch": 9.32, |
| "learning_rate": 7.207070707070708e-06, |
| "loss": 0.9538, |
| "step": 286500 |
| }, |
| { |
| "epoch": 9.33, |
| "learning_rate": 7.202020202020203e-06, |
| "loss": 0.9521, |
| "step": 287000 |
| }, |
| { |
| "epoch": 9.35, |
| "learning_rate": 7.196969696969698e-06, |
| "loss": 0.9522, |
| "step": 287500 |
| }, |
| { |
| "epoch": 9.37, |
| "learning_rate": 7.191919191919192e-06, |
| "loss": 0.9519, |
| "step": 288000 |
| }, |
| { |
| "epoch": 9.38, |
| "learning_rate": 7.186868686868688e-06, |
| "loss": 0.953, |
| "step": 288500 |
| }, |
| { |
| "epoch": 9.4, |
| "learning_rate": 7.181818181818182e-06, |
| "loss": 0.9514, |
| "step": 289000 |
| }, |
| { |
| "epoch": 9.42, |
| "learning_rate": 7.1767676767676775e-06, |
| "loss": 0.9514, |
| "step": 289500 |
| }, |
| { |
| "epoch": 9.43, |
| "learning_rate": 7.171717171717172e-06, |
| "loss": 0.9502, |
| "step": 290000 |
| }, |
| { |
| "epoch": 9.45, |
| "learning_rate": 7.166666666666667e-06, |
| "loss": 0.9501, |
| "step": 290500 |
| }, |
| { |
| "epoch": 9.46, |
| "learning_rate": 7.161616161616162e-06, |
| "loss": 0.9494, |
| "step": 291000 |
| }, |
| { |
| "epoch": 9.48, |
| "learning_rate": 7.156565656565657e-06, |
| "loss": 0.9492, |
| "step": 291500 |
| }, |
| { |
| "epoch": 9.5, |
| "learning_rate": 7.151515151515152e-06, |
| "loss": 0.9492, |
| "step": 292000 |
| }, |
| { |
| "epoch": 9.51, |
| "learning_rate": 7.146464646464647e-06, |
| "loss": 0.948, |
| "step": 292500 |
| }, |
| { |
| "epoch": 9.53, |
| "learning_rate": 7.141414141414143e-06, |
| "loss": 0.9473, |
| "step": 293000 |
| }, |
| { |
| "epoch": 9.55, |
| "learning_rate": 7.136363636363637e-06, |
| "loss": 0.9472, |
| "step": 293500 |
| }, |
| { |
| "epoch": 9.56, |
| "learning_rate": 7.131313131313132e-06, |
| "loss": 0.9468, |
| "step": 294000 |
| }, |
| { |
| "epoch": 9.58, |
| "learning_rate": 7.126262626262626e-06, |
| "loss": 0.9464, |
| "step": 294500 |
| }, |
| { |
| "epoch": 9.59, |
| "learning_rate": 7.121212121212122e-06, |
| "loss": 0.9457, |
| "step": 295000 |
| }, |
| { |
| "epoch": 9.61, |
| "learning_rate": 7.116161616161617e-06, |
| "loss": 0.9456, |
| "step": 295500 |
| }, |
| { |
| "epoch": 9.63, |
| "learning_rate": 7.111111111111112e-06, |
| "loss": 0.9454, |
| "step": 296000 |
| }, |
| { |
| "epoch": 9.64, |
| "learning_rate": 7.106060606060606e-06, |
| "loss": 0.9451, |
| "step": 296500 |
| }, |
| { |
| "epoch": 9.66, |
| "learning_rate": 7.101010101010102e-06, |
| "loss": 0.9441, |
| "step": 297000 |
| }, |
| { |
| "epoch": 9.68, |
| "learning_rate": 7.095959595959596e-06, |
| "loss": 0.9443, |
| "step": 297500 |
| }, |
| { |
| "epoch": 9.69, |
| "learning_rate": 7.0909090909090916e-06, |
| "loss": 0.9437, |
| "step": 298000 |
| }, |
| { |
| "epoch": 9.71, |
| "learning_rate": 7.085858585858586e-06, |
| "loss": 0.9422, |
| "step": 298500 |
| }, |
| { |
| "epoch": 9.72, |
| "learning_rate": 7.080808080808082e-06, |
| "loss": 0.9429, |
| "step": 299000 |
| }, |
| { |
| "epoch": 9.74, |
| "learning_rate": 7.075757575757576e-06, |
| "loss": 0.9428, |
| "step": 299500 |
| }, |
| { |
| "epoch": 9.76, |
| "learning_rate": 7.070707070707071e-06, |
| "loss": 0.9421, |
| "step": 300000 |
| }, |
| { |
| "epoch": 9.77, |
| "learning_rate": 7.065656565656566e-06, |
| "loss": 0.9414, |
| "step": 300500 |
| }, |
| { |
| "epoch": 9.79, |
| "learning_rate": 7.060606060606061e-06, |
| "loss": 0.9416, |
| "step": 301000 |
| }, |
| { |
| "epoch": 9.81, |
| "learning_rate": 7.055555555555557e-06, |
| "loss": 0.9408, |
| "step": 301500 |
| }, |
| { |
| "epoch": 9.82, |
| "learning_rate": 7.050505050505051e-06, |
| "loss": 0.9401, |
| "step": 302000 |
| }, |
| { |
| "epoch": 9.84, |
| "learning_rate": 7.045454545454546e-06, |
| "loss": 0.9405, |
| "step": 302500 |
| }, |
| { |
| "epoch": 9.85, |
| "learning_rate": 7.0404040404040404e-06, |
| "loss": 0.9399, |
| "step": 303000 |
| }, |
| { |
| "epoch": 9.87, |
| "learning_rate": 7.0353535353535364e-06, |
| "loss": 0.9386, |
| "step": 303500 |
| }, |
| { |
| "epoch": 9.89, |
| "learning_rate": 7.030303030303031e-06, |
| "loss": 0.939, |
| "step": 304000 |
| }, |
| { |
| "epoch": 9.9, |
| "learning_rate": 7.025252525252526e-06, |
| "loss": 0.9384, |
| "step": 304500 |
| }, |
| { |
| "epoch": 9.92, |
| "learning_rate": 7.02020202020202e-06, |
| "loss": 0.937, |
| "step": 305000 |
| }, |
| { |
| "epoch": 9.94, |
| "learning_rate": 7.015151515151516e-06, |
| "loss": 0.9371, |
| "step": 305500 |
| }, |
| { |
| "epoch": 9.95, |
| "learning_rate": 7.0101010101010105e-06, |
| "loss": 0.9369, |
| "step": 306000 |
| }, |
| { |
| "epoch": 9.97, |
| "learning_rate": 7.005050505050506e-06, |
| "loss": 0.9363, |
| "step": 306500 |
| }, |
| { |
| "epoch": 9.98, |
| "learning_rate": 7e-06, |
| "loss": 0.9357, |
| "step": 307000 |
| }, |
| { |
| "epoch": 10.0, |
| "learning_rate": 6.994949494949496e-06, |
| "loss": 0.9361, |
| "step": 307500 |
| }, |
| { |
| "epoch": 10.02, |
| "learning_rate": 6.98989898989899e-06, |
| "loss": 0.9351, |
| "step": 308000 |
| }, |
| { |
| "epoch": 10.03, |
| "learning_rate": 6.984848484848485e-06, |
| "loss": 0.9361, |
| "step": 308500 |
| }, |
| { |
| "epoch": 10.05, |
| "learning_rate": 6.979797979797981e-06, |
| "loss": 0.9351, |
| "step": 309000 |
| }, |
| { |
| "epoch": 10.07, |
| "learning_rate": 6.974747474747476e-06, |
| "loss": 0.9343, |
| "step": 309500 |
| }, |
| { |
| "epoch": 10.08, |
| "learning_rate": 6.969696969696971e-06, |
| "loss": 0.9339, |
| "step": 310000 |
| }, |
| { |
| "epoch": 10.1, |
| "learning_rate": 6.964646464646465e-06, |
| "loss": 0.9335, |
| "step": 310500 |
| }, |
| { |
| "epoch": 10.11, |
| "learning_rate": 6.95959595959596e-06, |
| "loss": 0.9338, |
| "step": 311000 |
| }, |
| { |
| "epoch": 10.13, |
| "learning_rate": 6.954545454545455e-06, |
| "loss": 0.9328, |
| "step": 311500 |
| }, |
| { |
| "epoch": 10.15, |
| "learning_rate": 6.9494949494949505e-06, |
| "loss": 0.9331, |
| "step": 312000 |
| }, |
| { |
| "epoch": 10.16, |
| "learning_rate": 6.944444444444445e-06, |
| "loss": 0.9319, |
| "step": 312500 |
| }, |
| { |
| "epoch": 10.18, |
| "learning_rate": 6.93939393939394e-06, |
| "loss": 0.9313, |
| "step": 313000 |
| }, |
| { |
| "epoch": 10.2, |
| "learning_rate": 6.934343434343434e-06, |
| "loss": 0.9313, |
| "step": 313500 |
| }, |
| { |
| "epoch": 10.21, |
| "learning_rate": 6.92929292929293e-06, |
| "loss": 0.931, |
| "step": 314000 |
| }, |
| { |
| "epoch": 10.23, |
| "learning_rate": 6.9242424242424245e-06, |
| "loss": 0.9298, |
| "step": 314500 |
| }, |
| { |
| "epoch": 10.24, |
| "learning_rate": 6.91919191919192e-06, |
| "loss": 0.931, |
| "step": 315000 |
| }, |
| { |
| "epoch": 10.26, |
| "learning_rate": 6.914141414141414e-06, |
| "loss": 0.9302, |
| "step": 315500 |
| }, |
| { |
| "epoch": 10.28, |
| "learning_rate": 6.90909090909091e-06, |
| "loss": 0.929, |
| "step": 316000 |
| }, |
| { |
| "epoch": 10.29, |
| "learning_rate": 6.904040404040404e-06, |
| "loss": 0.9292, |
| "step": 316500 |
| }, |
| { |
| "epoch": 10.31, |
| "learning_rate": 6.898989898989899e-06, |
| "loss": 0.9284, |
| "step": 317000 |
| }, |
| { |
| "epoch": 10.33, |
| "learning_rate": 6.893939393939395e-06, |
| "loss": 0.9292, |
| "step": 317500 |
| }, |
| { |
| "epoch": 10.34, |
| "learning_rate": 6.88888888888889e-06, |
| "loss": 0.9282, |
| "step": 318000 |
| }, |
| { |
| "epoch": 10.36, |
| "learning_rate": 6.883838383838385e-06, |
| "loss": 0.9279, |
| "step": 318500 |
| }, |
| { |
| "epoch": 10.37, |
| "learning_rate": 6.878787878787879e-06, |
| "loss": 0.9272, |
| "step": 319000 |
| }, |
| { |
| "epoch": 10.39, |
| "learning_rate": 6.873737373737375e-06, |
| "loss": 0.927, |
| "step": 319500 |
| }, |
| { |
| "epoch": 10.41, |
| "learning_rate": 6.868686868686869e-06, |
| "loss": 0.9271, |
| "step": 320000 |
| }, |
| { |
| "epoch": 10.42, |
| "learning_rate": 6.8636363636363645e-06, |
| "loss": 0.9266, |
| "step": 320500 |
| }, |
| { |
| "epoch": 10.44, |
| "learning_rate": 6.858585858585859e-06, |
| "loss": 0.9256, |
| "step": 321000 |
| }, |
| { |
| "epoch": 10.46, |
| "learning_rate": 6.853535353535354e-06, |
| "loss": 0.9257, |
| "step": 321500 |
| }, |
| { |
| "epoch": 10.47, |
| "learning_rate": 6.848484848484849e-06, |
| "loss": 0.9259, |
| "step": 322000 |
| }, |
| { |
| "epoch": 10.49, |
| "learning_rate": 6.843434343434344e-06, |
| "loss": 0.9252, |
| "step": 322500 |
| }, |
| { |
| "epoch": 10.5, |
| "learning_rate": 6.8383838383838386e-06, |
| "loss": 0.9246, |
| "step": 323000 |
| }, |
| { |
| "epoch": 10.52, |
| "learning_rate": 6.833333333333334e-06, |
| "loss": 0.9241, |
| "step": 323500 |
| }, |
| { |
| "epoch": 10.54, |
| "learning_rate": 6.828282828282828e-06, |
| "loss": 0.9239, |
| "step": 324000 |
| }, |
| { |
| "epoch": 10.55, |
| "learning_rate": 6.823232323232324e-06, |
| "loss": 0.9237, |
| "step": 324500 |
| }, |
| { |
| "epoch": 10.57, |
| "learning_rate": 6.818181818181818e-06, |
| "loss": 0.9231, |
| "step": 325000 |
| }, |
| { |
| "epoch": 10.59, |
| "learning_rate": 6.813131313131313e-06, |
| "loss": 0.9231, |
| "step": 325500 |
| }, |
| { |
| "epoch": 10.6, |
| "learning_rate": 6.808080808080809e-06, |
| "loss": 0.9227, |
| "step": 326000 |
| }, |
| { |
| "epoch": 10.62, |
| "learning_rate": 6.803030303030304e-06, |
| "loss": 0.9222, |
| "step": 326500 |
| }, |
| { |
| "epoch": 10.63, |
| "learning_rate": 6.797979797979799e-06, |
| "loss": 0.9215, |
| "step": 327000 |
| }, |
| { |
| "epoch": 10.65, |
| "learning_rate": 6.792929292929293e-06, |
| "loss": 0.9217, |
| "step": 327500 |
| }, |
| { |
| "epoch": 10.67, |
| "learning_rate": 6.787878787878789e-06, |
| "loss": 0.9211, |
| "step": 328000 |
| }, |
| { |
| "epoch": 10.68, |
| "learning_rate": 6.7828282828282834e-06, |
| "loss": 0.9204, |
| "step": 328500 |
| }, |
| { |
| "epoch": 10.7, |
| "learning_rate": 6.777777777777779e-06, |
| "loss": 0.9211, |
| "step": 329000 |
| }, |
| { |
| "epoch": 10.72, |
| "learning_rate": 6.772727272727273e-06, |
| "loss": 0.9195, |
| "step": 329500 |
| }, |
| { |
| "epoch": 10.73, |
| "learning_rate": 6.767676767676769e-06, |
| "loss": 0.9195, |
| "step": 330000 |
| }, |
| { |
| "epoch": 10.75, |
| "learning_rate": 6.762626262626263e-06, |
| "loss": 0.9195, |
| "step": 330500 |
| }, |
| { |
| "epoch": 10.76, |
| "learning_rate": 6.757575757575758e-06, |
| "loss": 0.9188, |
| "step": 331000 |
| }, |
| { |
| "epoch": 10.78, |
| "learning_rate": 6.752525252525253e-06, |
| "loss": 0.9187, |
| "step": 331500 |
| }, |
| { |
| "epoch": 10.8, |
| "learning_rate": 6.747474747474749e-06, |
| "loss": 0.9193, |
| "step": 332000 |
| }, |
| { |
| "epoch": 10.81, |
| "learning_rate": 6.742424242424243e-06, |
| "loss": 0.9183, |
| "step": 332500 |
| }, |
| { |
| "epoch": 10.83, |
| "learning_rate": 6.737373737373738e-06, |
| "loss": 0.9176, |
| "step": 333000 |
| }, |
| { |
| "epoch": 10.85, |
| "learning_rate": 6.732323232323232e-06, |
| "loss": 0.9173, |
| "step": 333500 |
| }, |
| { |
| "epoch": 10.86, |
| "learning_rate": 6.7272727272727275e-06, |
| "loss": 0.9164, |
| "step": 334000 |
| }, |
| { |
| "epoch": 10.88, |
| "learning_rate": 6.7222222222222235e-06, |
| "loss": 0.9165, |
| "step": 334500 |
| }, |
| { |
| "epoch": 10.9, |
| "learning_rate": 6.717171717171718e-06, |
| "loss": 0.9169, |
| "step": 335000 |
| }, |
| { |
| "epoch": 10.91, |
| "learning_rate": 6.712121212121213e-06, |
| "loss": 0.9156, |
| "step": 335500 |
| }, |
| { |
| "epoch": 10.93, |
| "learning_rate": 6.707070707070707e-06, |
| "loss": 0.916, |
| "step": 336000 |
| }, |
| { |
| "epoch": 10.94, |
| "learning_rate": 6.702020202020203e-06, |
| "loss": 0.9148, |
| "step": 336500 |
| }, |
| { |
| "epoch": 10.96, |
| "learning_rate": 6.6969696969696975e-06, |
| "loss": 0.9154, |
| "step": 337000 |
| }, |
| { |
| "epoch": 10.98, |
| "learning_rate": 6.691919191919193e-06, |
| "loss": 0.9147, |
| "step": 337500 |
| }, |
| { |
| "epoch": 10.99, |
| "learning_rate": 6.686868686868687e-06, |
| "loss": 0.9145, |
| "step": 338000 |
| }, |
| { |
| "epoch": 11.01, |
| "learning_rate": 6.681818181818183e-06, |
| "loss": 0.9144, |
| "step": 338500 |
| }, |
| { |
| "epoch": 11.03, |
| "learning_rate": 6.676767676767677e-06, |
| "loss": 0.9144, |
| "step": 339000 |
| }, |
| { |
| "epoch": 11.04, |
| "learning_rate": 6.671717171717172e-06, |
| "loss": 0.9127, |
| "step": 339500 |
| }, |
| { |
| "epoch": 11.06, |
| "learning_rate": 6.666666666666667e-06, |
| "loss": 0.9133, |
| "step": 340000 |
| }, |
| { |
| "epoch": 11.07, |
| "learning_rate": 6.661616161616163e-06, |
| "loss": 0.9129, |
| "step": 340500 |
| }, |
| { |
| "epoch": 11.09, |
| "learning_rate": 6.656565656565657e-06, |
| "loss": 0.9133, |
| "step": 341000 |
| }, |
| { |
| "epoch": 11.11, |
| "learning_rate": 6.651515151515152e-06, |
| "loss": 0.9128, |
| "step": 341500 |
| }, |
| { |
| "epoch": 11.12, |
| "learning_rate": 6.646464646464646e-06, |
| "loss": 0.9123, |
| "step": 342000 |
| }, |
| { |
| "epoch": 11.14, |
| "learning_rate": 6.641414141414142e-06, |
| "loss": 0.9114, |
| "step": 342500 |
| }, |
| { |
| "epoch": 11.16, |
| "learning_rate": 6.6363636363636375e-06, |
| "loss": 0.9113, |
| "step": 343000 |
| }, |
| { |
| "epoch": 11.17, |
| "learning_rate": 6.631313131313132e-06, |
| "loss": 0.9117, |
| "step": 343500 |
| }, |
| { |
| "epoch": 11.19, |
| "learning_rate": 6.626262626262627e-06, |
| "loss": 0.9111, |
| "step": 344000 |
| }, |
| { |
| "epoch": 11.2, |
| "learning_rate": 6.621212121212121e-06, |
| "loss": 0.91, |
| "step": 344500 |
| }, |
| { |
| "epoch": 11.22, |
| "learning_rate": 6.616161616161617e-06, |
| "loss": 0.9098, |
| "step": 345000 |
| }, |
| { |
| "epoch": 11.24, |
| "learning_rate": 6.6111111111111115e-06, |
| "loss": 0.9109, |
| "step": 345500 |
| }, |
| { |
| "epoch": 11.25, |
| "learning_rate": 6.606060606060607e-06, |
| "loss": 0.9105, |
| "step": 346000 |
| }, |
| { |
| "epoch": 11.27, |
| "learning_rate": 6.601010101010101e-06, |
| "loss": 0.9095, |
| "step": 346500 |
| }, |
| { |
| "epoch": 11.29, |
| "learning_rate": 6.595959595959597e-06, |
| "loss": 0.9081, |
| "step": 347000 |
| }, |
| { |
| "epoch": 11.3, |
| "learning_rate": 6.590909090909091e-06, |
| "loss": 0.9081, |
| "step": 347500 |
| }, |
| { |
| "epoch": 11.32, |
| "learning_rate": 6.585858585858586e-06, |
| "loss": 0.9084, |
| "step": 348000 |
| }, |
| { |
| "epoch": 11.33, |
| "learning_rate": 6.580808080808081e-06, |
| "loss": 0.9084, |
| "step": 348500 |
| }, |
| { |
| "epoch": 11.35, |
| "learning_rate": 6.575757575757577e-06, |
| "loss": 0.9075, |
| "step": 349000 |
| }, |
| { |
| "epoch": 11.37, |
| "learning_rate": 6.570707070707071e-06, |
| "loss": 0.9076, |
| "step": 349500 |
| }, |
| { |
| "epoch": 11.38, |
| "learning_rate": 6.565656565656566e-06, |
| "loss": 0.9079, |
| "step": 350000 |
| }, |
| { |
| "epoch": 11.4, |
| "learning_rate": 6.56060606060606e-06, |
| "loss": 0.9069, |
| "step": 350500 |
| }, |
| { |
| "epoch": 11.42, |
| "learning_rate": 6.555555555555556e-06, |
| "loss": 0.9071, |
| "step": 351000 |
| }, |
| { |
| "epoch": 11.43, |
| "learning_rate": 6.5505050505050516e-06, |
| "loss": 0.9054, |
| "step": 351500 |
| }, |
| { |
| "epoch": 11.45, |
| "learning_rate": 6.545454545454546e-06, |
| "loss": 0.9064, |
| "step": 352000 |
| }, |
| { |
| "epoch": 11.46, |
| "learning_rate": 6.540404040404042e-06, |
| "loss": 0.9053, |
| "step": 352500 |
| }, |
| { |
| "epoch": 11.48, |
| "learning_rate": 6.535353535353536e-06, |
| "loss": 0.9053, |
| "step": 353000 |
| }, |
| { |
| "epoch": 11.5, |
| "learning_rate": 6.530303030303031e-06, |
| "loss": 0.9046, |
| "step": 353500 |
| }, |
| { |
| "epoch": 11.51, |
| "learning_rate": 6.525252525252526e-06, |
| "loss": 0.9054, |
| "step": 354000 |
| }, |
| { |
| "epoch": 11.53, |
| "learning_rate": 6.520202020202021e-06, |
| "loss": 0.9042, |
| "step": 354500 |
| }, |
| { |
| "epoch": 11.55, |
| "learning_rate": 6.515151515151516e-06, |
| "loss": 0.9043, |
| "step": 355000 |
| }, |
| { |
| "epoch": 11.56, |
| "learning_rate": 6.510101010101011e-06, |
| "loss": 0.904, |
| "step": 355500 |
| }, |
| { |
| "epoch": 11.58, |
| "learning_rate": 6.505050505050505e-06, |
| "loss": 0.9029, |
| "step": 356000 |
| }, |
| { |
| "epoch": 11.59, |
| "learning_rate": 6.5000000000000004e-06, |
| "loss": 0.9035, |
| "step": 356500 |
| }, |
| { |
| "epoch": 11.61, |
| "learning_rate": 6.494949494949495e-06, |
| "loss": 0.9036, |
| "step": 357000 |
| }, |
| { |
| "epoch": 11.63, |
| "learning_rate": 6.489898989898991e-06, |
| "loss": 0.9025, |
| "step": 357500 |
| }, |
| { |
| "epoch": 11.64, |
| "learning_rate": 6.484848484848485e-06, |
| "loss": 0.903, |
| "step": 358000 |
| }, |
| { |
| "epoch": 11.66, |
| "learning_rate": 6.47979797979798e-06, |
| "loss": 0.9022, |
| "step": 358500 |
| }, |
| { |
| "epoch": 11.68, |
| "learning_rate": 6.4747474747474745e-06, |
| "loss": 0.9018, |
| "step": 359000 |
| }, |
| { |
| "epoch": 11.69, |
| "learning_rate": 6.4696969696969705e-06, |
| "loss": 0.9017, |
| "step": 359500 |
| }, |
| { |
| "epoch": 11.71, |
| "learning_rate": 6.464646464646466e-06, |
| "loss": 0.9003, |
| "step": 360000 |
| }, |
| { |
| "epoch": 11.72, |
| "learning_rate": 6.45959595959596e-06, |
| "loss": 0.9013, |
| "step": 360500 |
| }, |
| { |
| "epoch": 11.74, |
| "learning_rate": 6.454545454545456e-06, |
| "loss": 0.9001, |
| "step": 361000 |
| }, |
| { |
| "epoch": 11.76, |
| "learning_rate": 6.44949494949495e-06, |
| "loss": 0.8999, |
| "step": 361500 |
| }, |
| { |
| "epoch": 11.77, |
| "learning_rate": 6.444444444444445e-06, |
| "loss": 0.9004, |
| "step": 362000 |
| }, |
| { |
| "epoch": 11.79, |
| "learning_rate": 6.43939393939394e-06, |
| "loss": 0.8995, |
| "step": 362500 |
| }, |
| { |
| "epoch": 11.81, |
| "learning_rate": 6.434343434343436e-06, |
| "loss": 0.8995, |
| "step": 363000 |
| }, |
| { |
| "epoch": 11.82, |
| "learning_rate": 6.42929292929293e-06, |
| "loss": 0.8992, |
| "step": 363500 |
| }, |
| { |
| "epoch": 11.84, |
| "learning_rate": 6.424242424242425e-06, |
| "loss": 0.8985, |
| "step": 364000 |
| }, |
| { |
| "epoch": 11.85, |
| "learning_rate": 6.419191919191919e-06, |
| "loss": 0.8987, |
| "step": 364500 |
| }, |
| { |
| "epoch": 11.87, |
| "learning_rate": 6.4141414141414145e-06, |
| "loss": 0.898, |
| "step": 365000 |
| }, |
| { |
| "epoch": 11.89, |
| "learning_rate": 6.40909090909091e-06, |
| "loss": 0.8981, |
| "step": 365500 |
| }, |
| { |
| "epoch": 11.9, |
| "learning_rate": 6.404040404040405e-06, |
| "loss": 0.8977, |
| "step": 366000 |
| }, |
| { |
| "epoch": 11.92, |
| "learning_rate": 6.398989898989899e-06, |
| "loss": 0.8975, |
| "step": 366500 |
| }, |
| { |
| "epoch": 11.94, |
| "learning_rate": 6.393939393939394e-06, |
| "loss": 0.8977, |
| "step": 367000 |
| }, |
| { |
| "epoch": 11.95, |
| "learning_rate": 6.3888888888888885e-06, |
| "loss": 0.8966, |
| "step": 367500 |
| }, |
| { |
| "epoch": 11.97, |
| "learning_rate": 6.3838383838383845e-06, |
| "loss": 0.8967, |
| "step": 368000 |
| }, |
| { |
| "epoch": 11.98, |
| "learning_rate": 6.37878787878788e-06, |
| "loss": 0.8961, |
| "step": 368500 |
| }, |
| { |
| "epoch": 12.0, |
| "learning_rate": 6.373737373737374e-06, |
| "loss": 0.8963, |
| "step": 369000 |
| }, |
| { |
| "epoch": 12.02, |
| "learning_rate": 6.36868686868687e-06, |
| "loss": 0.8957, |
| "step": 369500 |
| }, |
| { |
| "epoch": 12.03, |
| "learning_rate": 6.363636363636364e-06, |
| "loss": 0.8962, |
| "step": 370000 |
| }, |
| { |
| "epoch": 12.05, |
| "learning_rate": 6.358585858585859e-06, |
| "loss": 0.8947, |
| "step": 370500 |
| }, |
| { |
| "epoch": 12.07, |
| "learning_rate": 6.353535353535354e-06, |
| "loss": 0.8947, |
| "step": 371000 |
| }, |
| { |
| "epoch": 12.08, |
| "learning_rate": 6.34848484848485e-06, |
| "loss": 0.8938, |
| "step": 371500 |
| }, |
| { |
| "epoch": 12.1, |
| "learning_rate": 6.343434343434344e-06, |
| "loss": 0.8951, |
| "step": 372000 |
| }, |
| { |
| "epoch": 12.11, |
| "learning_rate": 6.338383838383839e-06, |
| "loss": 0.8942, |
| "step": 372500 |
| }, |
| { |
| "epoch": 12.13, |
| "learning_rate": 6.333333333333333e-06, |
| "loss": 0.8938, |
| "step": 373000 |
| }, |
| { |
| "epoch": 12.15, |
| "learning_rate": 6.328282828282829e-06, |
| "loss": 0.8934, |
| "step": 373500 |
| }, |
| { |
| "epoch": 12.16, |
| "learning_rate": 6.323232323232324e-06, |
| "loss": 0.8931, |
| "step": 374000 |
| }, |
| { |
| "epoch": 12.18, |
| "learning_rate": 6.318181818181819e-06, |
| "loss": 0.893, |
| "step": 374500 |
| }, |
| { |
| "epoch": 12.2, |
| "learning_rate": 6.313131313131313e-06, |
| "loss": 0.8923, |
| "step": 375000 |
| }, |
| { |
| "epoch": 12.21, |
| "learning_rate": 6.308080808080809e-06, |
| "loss": 0.8934, |
| "step": 375500 |
| }, |
| { |
| "epoch": 12.23, |
| "learning_rate": 6.303030303030303e-06, |
| "loss": 0.8923, |
| "step": 376000 |
| }, |
| { |
| "epoch": 12.24, |
| "learning_rate": 6.2979797979797986e-06, |
| "loss": 0.8921, |
| "step": 376500 |
| }, |
| { |
| "epoch": 12.26, |
| "learning_rate": 6.292929292929294e-06, |
| "loss": 0.8915, |
| "step": 377000 |
| }, |
| { |
| "epoch": 12.28, |
| "learning_rate": 6.287878787878788e-06, |
| "loss": 0.8915, |
| "step": 377500 |
| }, |
| { |
| "epoch": 12.29, |
| "learning_rate": 6.282828282828284e-06, |
| "loss": 0.8909, |
| "step": 378000 |
| }, |
| { |
| "epoch": 12.31, |
| "learning_rate": 6.277777777777778e-06, |
| "loss": 0.8911, |
| "step": 378500 |
| }, |
| { |
| "epoch": 12.33, |
| "learning_rate": 6.2727272727272734e-06, |
| "loss": 0.8907, |
| "step": 379000 |
| }, |
| { |
| "epoch": 12.34, |
| "learning_rate": 6.267676767676768e-06, |
| "loss": 0.8904, |
| "step": 379500 |
| }, |
| { |
| "epoch": 12.36, |
| "learning_rate": 6.262626262626264e-06, |
| "loss": 0.8905, |
| "step": 380000 |
| }, |
| { |
| "epoch": 12.37, |
| "learning_rate": 6.257575757575758e-06, |
| "loss": 0.89, |
| "step": 380500 |
| }, |
| { |
| "epoch": 12.39, |
| "learning_rate": 6.252525252525253e-06, |
| "loss": 0.8896, |
| "step": 381000 |
| }, |
| { |
| "epoch": 12.41, |
| "learning_rate": 6.2474747474747474e-06, |
| "loss": 0.8889, |
| "step": 381500 |
| }, |
| { |
| "epoch": 12.42, |
| "learning_rate": 6.2424242424242434e-06, |
| "loss": 0.8885, |
| "step": 382000 |
| }, |
| { |
| "epoch": 12.44, |
| "learning_rate": 6.237373737373738e-06, |
| "loss": 0.8887, |
| "step": 382500 |
| }, |
| { |
| "epoch": 12.46, |
| "learning_rate": 6.232323232323233e-06, |
| "loss": 0.8885, |
| "step": 383000 |
| }, |
| { |
| "epoch": 12.47, |
| "learning_rate": 6.227272727272727e-06, |
| "loss": 0.8887, |
| "step": 383500 |
| }, |
| { |
| "epoch": 12.49, |
| "learning_rate": 6.222222222222223e-06, |
| "loss": 0.888, |
| "step": 384000 |
| }, |
| { |
| "epoch": 12.5, |
| "learning_rate": 6.2171717171717175e-06, |
| "loss": 0.8884, |
| "step": 384500 |
| }, |
| { |
| "epoch": 12.52, |
| "learning_rate": 6.212121212121213e-06, |
| "loss": 0.888, |
| "step": 385000 |
| }, |
| { |
| "epoch": 12.54, |
| "learning_rate": 6.207070707070707e-06, |
| "loss": 0.8869, |
| "step": 385500 |
| }, |
| { |
| "epoch": 12.55, |
| "learning_rate": 6.202020202020203e-06, |
| "loss": 0.8869, |
| "step": 386000 |
| }, |
| { |
| "epoch": 12.57, |
| "learning_rate": 6.196969696969698e-06, |
| "loss": 0.8865, |
| "step": 386500 |
| }, |
| { |
| "epoch": 12.59, |
| "learning_rate": 6.191919191919192e-06, |
| "loss": 0.8863, |
| "step": 387000 |
| }, |
| { |
| "epoch": 12.6, |
| "learning_rate": 6.1868686868686875e-06, |
| "loss": 0.8865, |
| "step": 387500 |
| }, |
| { |
| "epoch": 12.62, |
| "learning_rate": 6.181818181818182e-06, |
| "loss": 0.8856, |
| "step": 388000 |
| }, |
| { |
| "epoch": 12.63, |
| "learning_rate": 6.176767676767678e-06, |
| "loss": 0.885, |
| "step": 388500 |
| }, |
| { |
| "epoch": 12.65, |
| "learning_rate": 6.171717171717172e-06, |
| "loss": 0.8857, |
| "step": 389000 |
| }, |
| { |
| "epoch": 12.67, |
| "learning_rate": 6.166666666666667e-06, |
| "loss": 0.8857, |
| "step": 389500 |
| }, |
| { |
| "epoch": 12.68, |
| "learning_rate": 6.1616161616161615e-06, |
| "loss": 0.8857, |
| "step": 390000 |
| }, |
| { |
| "epoch": 12.7, |
| "learning_rate": 6.1565656565656575e-06, |
| "loss": 0.885, |
| "step": 390500 |
| }, |
| { |
| "epoch": 12.72, |
| "learning_rate": 6.151515151515152e-06, |
| "loss": 0.8848, |
| "step": 391000 |
| }, |
| { |
| "epoch": 12.73, |
| "learning_rate": 6.146464646464647e-06, |
| "loss": 0.8839, |
| "step": 391500 |
| }, |
| { |
| "epoch": 12.75, |
| "learning_rate": 6.141414141414141e-06, |
| "loss": 0.8841, |
| "step": 392000 |
| }, |
| { |
| "epoch": 12.77, |
| "learning_rate": 6.136363636363637e-06, |
| "loss": 0.8837, |
| "step": 392500 |
| }, |
| { |
| "epoch": 12.78, |
| "learning_rate": 6.1313131313131315e-06, |
| "loss": 0.8839, |
| "step": 393000 |
| }, |
| { |
| "epoch": 12.8, |
| "learning_rate": 6.126262626262627e-06, |
| "loss": 0.8832, |
| "step": 393500 |
| }, |
| { |
| "epoch": 12.81, |
| "learning_rate": 6.121212121212121e-06, |
| "loss": 0.8831, |
| "step": 394000 |
| }, |
| { |
| "epoch": 12.83, |
| "learning_rate": 6.116161616161617e-06, |
| "loss": 0.8832, |
| "step": 394500 |
| }, |
| { |
| "epoch": 12.85, |
| "learning_rate": 6.111111111111112e-06, |
| "loss": 0.8829, |
| "step": 395000 |
| }, |
| { |
| "epoch": 12.86, |
| "learning_rate": 6.106060606060606e-06, |
| "loss": 0.8823, |
| "step": 395500 |
| }, |
| { |
| "epoch": 12.88, |
| "learning_rate": 6.1010101010101015e-06, |
| "loss": 0.8816, |
| "step": 396000 |
| }, |
| { |
| "epoch": 12.9, |
| "learning_rate": 6.095959595959597e-06, |
| "loss": 0.8818, |
| "step": 396500 |
| }, |
| { |
| "epoch": 12.91, |
| "learning_rate": 6.090909090909092e-06, |
| "loss": 0.8808, |
| "step": 397000 |
| }, |
| { |
| "epoch": 12.93, |
| "learning_rate": 6.085858585858586e-06, |
| "loss": 0.8813, |
| "step": 397500 |
| }, |
| { |
| "epoch": 12.94, |
| "learning_rate": 6.080808080808081e-06, |
| "loss": 0.8815, |
| "step": 398000 |
| }, |
| { |
| "epoch": 12.96, |
| "learning_rate": 6.0757575757575755e-06, |
| "loss": 0.8807, |
| "step": 398500 |
| }, |
| { |
| "epoch": 12.98, |
| "learning_rate": 6.0707070707070715e-06, |
| "loss": 0.8805, |
| "step": 399000 |
| }, |
| { |
| "epoch": 12.99, |
| "learning_rate": 6.065656565656566e-06, |
| "loss": 0.88, |
| "step": 399500 |
| }, |
| { |
| "epoch": 13.01, |
| "learning_rate": 6.060606060606061e-06, |
| "loss": 0.8797, |
| "step": 400000 |
| }, |
| { |
| "epoch": 13.03, |
| "learning_rate": 6.055555555555555e-06, |
| "loss": 0.8797, |
| "step": 400500 |
| }, |
| { |
| "epoch": 13.04, |
| "learning_rate": 6.050505050505051e-06, |
| "loss": 0.8797, |
| "step": 401000 |
| }, |
| { |
| "epoch": 13.06, |
| "learning_rate": 6.0454545454545456e-06, |
| "loss": 0.8792, |
| "step": 401500 |
| }, |
| { |
| "epoch": 13.07, |
| "learning_rate": 6.040404040404041e-06, |
| "loss": 0.8785, |
| "step": 402000 |
| }, |
| { |
| "epoch": 13.09, |
| "learning_rate": 6.035353535353535e-06, |
| "loss": 0.8788, |
| "step": 402500 |
| }, |
| { |
| "epoch": 13.11, |
| "learning_rate": 6.030303030303031e-06, |
| "loss": 0.8789, |
| "step": 403000 |
| }, |
| { |
| "epoch": 13.12, |
| "learning_rate": 6.025252525252526e-06, |
| "loss": 0.8788, |
| "step": 403500 |
| }, |
| { |
| "epoch": 13.14, |
| "learning_rate": 6.0202020202020204e-06, |
| "loss": 0.8779, |
| "step": 404000 |
| }, |
| { |
| "epoch": 13.16, |
| "learning_rate": 6.015151515151516e-06, |
| "loss": 0.8779, |
| "step": 404500 |
| }, |
| { |
| "epoch": 13.17, |
| "learning_rate": 6.010101010101011e-06, |
| "loss": 0.8768, |
| "step": 405000 |
| }, |
| { |
| "epoch": 13.19, |
| "learning_rate": 6.005050505050506e-06, |
| "loss": 0.8775, |
| "step": 405500 |
| }, |
| { |
| "epoch": 13.2, |
| "learning_rate": 6e-06, |
| "loss": 0.8773, |
| "step": 406000 |
| }, |
| { |
| "epoch": 13.22, |
| "learning_rate": 5.994949494949496e-06, |
| "loss": 0.8766, |
| "step": 406500 |
| }, |
| { |
| "epoch": 13.24, |
| "learning_rate": 5.9898989898989904e-06, |
| "loss": 0.8764, |
| "step": 407000 |
| }, |
| { |
| "epoch": 13.25, |
| "learning_rate": 5.984848484848486e-06, |
| "loss": 0.8772, |
| "step": 407500 |
| }, |
| { |
| "epoch": 13.27, |
| "learning_rate": 5.97979797979798e-06, |
| "loss": 0.876, |
| "step": 408000 |
| }, |
| { |
| "epoch": 13.29, |
| "learning_rate": 5.974747474747475e-06, |
| "loss": 0.8756, |
| "step": 408500 |
| }, |
| { |
| "epoch": 13.3, |
| "learning_rate": 5.96969696969697e-06, |
| "loss": 0.8755, |
| "step": 409000 |
| }, |
| { |
| "epoch": 13.32, |
| "learning_rate": 5.964646464646465e-06, |
| "loss": 0.8751, |
| "step": 409500 |
| }, |
| { |
| "epoch": 13.33, |
| "learning_rate": 5.95959595959596e-06, |
| "loss": 0.8749, |
| "step": 410000 |
| }, |
| { |
| "epoch": 13.35, |
| "learning_rate": 5.954545454545455e-06, |
| "loss": 0.8748, |
| "step": 410500 |
| }, |
| { |
| "epoch": 13.37, |
| "learning_rate": 5.949494949494949e-06, |
| "loss": 0.875, |
| "step": 411000 |
| }, |
| { |
| "epoch": 13.38, |
| "learning_rate": 5.944444444444445e-06, |
| "loss": 0.8751, |
| "step": 411500 |
| }, |
| { |
| "epoch": 13.4, |
| "learning_rate": 5.93939393939394e-06, |
| "loss": 0.8737, |
| "step": 412000 |
| }, |
| { |
| "epoch": 13.42, |
| "learning_rate": 5.9343434343434345e-06, |
| "loss": 0.8743, |
| "step": 412500 |
| }, |
| { |
| "epoch": 13.43, |
| "learning_rate": 5.9292929292929305e-06, |
| "loss": 0.8733, |
| "step": 413000 |
| }, |
| { |
| "epoch": 13.45, |
| "learning_rate": 5.924242424242425e-06, |
| "loss": 0.8735, |
| "step": 413500 |
| }, |
| { |
| "epoch": 13.46, |
| "learning_rate": 5.91919191919192e-06, |
| "loss": 0.8732, |
| "step": 414000 |
| }, |
| { |
| "epoch": 13.48, |
| "learning_rate": 5.914141414141414e-06, |
| "loss": 0.8734, |
| "step": 414500 |
| }, |
| { |
| "epoch": 13.5, |
| "learning_rate": 5.90909090909091e-06, |
| "loss": 0.8724, |
| "step": 415000 |
| }, |
| { |
| "epoch": 13.51, |
| "learning_rate": 5.9040404040404045e-06, |
| "loss": 0.8724, |
| "step": 415500 |
| }, |
| { |
| "epoch": 13.53, |
| "learning_rate": 5.8989898989899e-06, |
| "loss": 0.8719, |
| "step": 416000 |
| }, |
| { |
| "epoch": 13.55, |
| "learning_rate": 5.893939393939394e-06, |
| "loss": 0.8714, |
| "step": 416500 |
| }, |
| { |
| "epoch": 13.56, |
| "learning_rate": 5.88888888888889e-06, |
| "loss": 0.8723, |
| "step": 417000 |
| }, |
| { |
| "epoch": 13.58, |
| "learning_rate": 5.883838383838384e-06, |
| "loss": 0.8719, |
| "step": 417500 |
| }, |
| { |
| "epoch": 13.59, |
| "learning_rate": 5.878787878787879e-06, |
| "loss": 0.8709, |
| "step": 418000 |
| }, |
| { |
| "epoch": 13.61, |
| "learning_rate": 5.873737373737374e-06, |
| "loss": 0.8709, |
| "step": 418500 |
| }, |
| { |
| "epoch": 13.63, |
| "learning_rate": 5.868686868686869e-06, |
| "loss": 0.8707, |
| "step": 419000 |
| }, |
| { |
| "epoch": 13.64, |
| "learning_rate": 5.863636363636364e-06, |
| "loss": 0.871, |
| "step": 419500 |
| }, |
| { |
| "epoch": 13.66, |
| "learning_rate": 5.858585858585859e-06, |
| "loss": 0.8703, |
| "step": 420000 |
| }, |
| { |
| "epoch": 13.68, |
| "learning_rate": 5.853535353535354e-06, |
| "loss": 0.8698, |
| "step": 420500 |
| }, |
| { |
| "epoch": 13.69, |
| "learning_rate": 5.8484848484848485e-06, |
| "loss": 0.8698, |
| "step": 421000 |
| }, |
| { |
| "epoch": 13.71, |
| "learning_rate": 5.8434343434343445e-06, |
| "loss": 0.8696, |
| "step": 421500 |
| }, |
| { |
| "epoch": 13.72, |
| "learning_rate": 5.838383838383839e-06, |
| "loss": 0.869, |
| "step": 422000 |
| }, |
| { |
| "epoch": 13.74, |
| "learning_rate": 5.833333333333334e-06, |
| "loss": 0.8695, |
| "step": 422500 |
| }, |
| { |
| "epoch": 13.76, |
| "learning_rate": 5.828282828282828e-06, |
| "loss": 0.8691, |
| "step": 423000 |
| }, |
| { |
| "epoch": 13.77, |
| "learning_rate": 5.823232323232324e-06, |
| "loss": 0.8693, |
| "step": 423500 |
| }, |
| { |
| "epoch": 13.79, |
| "learning_rate": 5.8181818181818185e-06, |
| "loss": 0.8686, |
| "step": 424000 |
| }, |
| { |
| "epoch": 13.81, |
| "learning_rate": 5.813131313131314e-06, |
| "loss": 0.8684, |
| "step": 424500 |
| }, |
| { |
| "epoch": 13.82, |
| "learning_rate": 5.808080808080808e-06, |
| "loss": 0.8679, |
| "step": 425000 |
| }, |
| { |
| "epoch": 13.84, |
| "learning_rate": 5.803030303030304e-06, |
| "loss": 0.8679, |
| "step": 425500 |
| }, |
| { |
| "epoch": 13.85, |
| "learning_rate": 5.797979797979798e-06, |
| "loss": 0.8673, |
| "step": 426000 |
| }, |
| { |
| "epoch": 13.87, |
| "learning_rate": 5.792929292929293e-06, |
| "loss": 0.8678, |
| "step": 426500 |
| }, |
| { |
| "epoch": 13.89, |
| "learning_rate": 5.787878787878788e-06, |
| "loss": 0.8677, |
| "step": 427000 |
| }, |
| { |
| "epoch": 13.9, |
| "learning_rate": 5.782828282828284e-06, |
| "loss": 0.867, |
| "step": 427500 |
| }, |
| { |
| "epoch": 13.92, |
| "learning_rate": 5.777777777777778e-06, |
| "loss": 0.8667, |
| "step": 428000 |
| }, |
| { |
| "epoch": 13.94, |
| "learning_rate": 5.772727272727273e-06, |
| "loss": 0.8662, |
| "step": 428500 |
| }, |
| { |
| "epoch": 13.95, |
| "learning_rate": 5.767676767676768e-06, |
| "loss": 0.8665, |
| "step": 429000 |
| }, |
| { |
| "epoch": 13.97, |
| "learning_rate": 5.762626262626263e-06, |
| "loss": 0.8656, |
| "step": 429500 |
| }, |
| { |
| "epoch": 13.98, |
| "learning_rate": 5.7575757575757586e-06, |
| "loss": 0.8659, |
| "step": 430000 |
| }, |
| { |
| "epoch": 14.0, |
| "learning_rate": 5.752525252525253e-06, |
| "loss": 0.8656, |
| "step": 430500 |
| }, |
| { |
| "epoch": 14.02, |
| "learning_rate": 5.747474747474748e-06, |
| "loss": 0.8647, |
| "step": 431000 |
| }, |
| { |
| "epoch": 14.03, |
| "learning_rate": 5.742424242424242e-06, |
| "loss": 0.865, |
| "step": 431500 |
| }, |
| { |
| "epoch": 14.05, |
| "learning_rate": 5.737373737373738e-06, |
| "loss": 0.8646, |
| "step": 432000 |
| }, |
| { |
| "epoch": 14.07, |
| "learning_rate": 5.732323232323233e-06, |
| "loss": 0.8644, |
| "step": 432500 |
| }, |
| { |
| "epoch": 14.08, |
| "learning_rate": 5.727272727272728e-06, |
| "loss": 0.8645, |
| "step": 433000 |
| }, |
| { |
| "epoch": 14.1, |
| "learning_rate": 5.722222222222222e-06, |
| "loss": 0.8641, |
| "step": 433500 |
| }, |
| { |
| "epoch": 14.11, |
| "learning_rate": 5.717171717171718e-06, |
| "loss": 0.8643, |
| "step": 434000 |
| }, |
| { |
| "epoch": 14.13, |
| "learning_rate": 5.712121212121212e-06, |
| "loss": 0.8635, |
| "step": 434500 |
| }, |
| { |
| "epoch": 14.15, |
| "learning_rate": 5.7070707070707075e-06, |
| "loss": 0.8631, |
| "step": 435000 |
| }, |
| { |
| "epoch": 14.16, |
| "learning_rate": 5.702020202020202e-06, |
| "loss": 0.8635, |
| "step": 435500 |
| }, |
| { |
| "epoch": 14.18, |
| "learning_rate": 5.696969696969698e-06, |
| "loss": 0.8627, |
| "step": 436000 |
| }, |
| { |
| "epoch": 14.2, |
| "learning_rate": 5.691919191919192e-06, |
| "loss": 0.8623, |
| "step": 436500 |
| }, |
| { |
| "epoch": 14.21, |
| "learning_rate": 5.686868686868687e-06, |
| "loss": 0.8622, |
| "step": 437000 |
| }, |
| { |
| "epoch": 14.23, |
| "learning_rate": 5.681818181818183e-06, |
| "loss": 0.8619, |
| "step": 437500 |
| }, |
| { |
| "epoch": 14.24, |
| "learning_rate": 5.6767676767676775e-06, |
| "loss": 0.8616, |
| "step": 438000 |
| }, |
| { |
| "epoch": 14.26, |
| "learning_rate": 5.671717171717173e-06, |
| "loss": 0.8621, |
| "step": 438500 |
| }, |
| { |
| "epoch": 14.28, |
| "learning_rate": 5.666666666666667e-06, |
| "loss": 0.8622, |
| "step": 439000 |
| }, |
| { |
| "epoch": 14.29, |
| "learning_rate": 5.661616161616162e-06, |
| "loss": 0.8616, |
| "step": 439500 |
| }, |
| { |
| "epoch": 14.31, |
| "learning_rate": 5.656565656565657e-06, |
| "loss": 0.8612, |
| "step": 440000 |
| }, |
| { |
| "epoch": 14.33, |
| "learning_rate": 5.651515151515152e-06, |
| "loss": 0.8612, |
| "step": 440500 |
| }, |
| { |
| "epoch": 14.34, |
| "learning_rate": 5.646464646464647e-06, |
| "loss": 0.8607, |
| "step": 441000 |
| }, |
| { |
| "epoch": 14.36, |
| "learning_rate": 5.641414141414142e-06, |
| "loss": 0.8607, |
| "step": 441500 |
| }, |
| { |
| "epoch": 14.37, |
| "learning_rate": 5.636363636363636e-06, |
| "loss": 0.8594, |
| "step": 442000 |
| }, |
| { |
| "epoch": 14.39, |
| "learning_rate": 5.631313131313132e-06, |
| "loss": 0.8602, |
| "step": 442500 |
| }, |
| { |
| "epoch": 14.41, |
| "learning_rate": 5.626262626262626e-06, |
| "loss": 0.8597, |
| "step": 443000 |
| }, |
| { |
| "epoch": 14.42, |
| "learning_rate": 5.6212121212121215e-06, |
| "loss": 0.8594, |
| "step": 443500 |
| }, |
| { |
| "epoch": 14.44, |
| "learning_rate": 5.616161616161616e-06, |
| "loss": 0.8598, |
| "step": 444000 |
| }, |
| { |
| "epoch": 14.46, |
| "learning_rate": 5.611111111111112e-06, |
| "loss": 0.8588, |
| "step": 444500 |
| }, |
| { |
| "epoch": 14.47, |
| "learning_rate": 5.606060606060606e-06, |
| "loss": 0.8587, |
| "step": 445000 |
| }, |
| { |
| "epoch": 14.49, |
| "learning_rate": 5.601010101010101e-06, |
| "loss": 0.8583, |
| "step": 445500 |
| }, |
| { |
| "epoch": 14.51, |
| "learning_rate": 5.595959595959597e-06, |
| "loss": 0.8589, |
| "step": 446000 |
| }, |
| { |
| "epoch": 14.52, |
| "learning_rate": 5.5909090909090915e-06, |
| "loss": 0.8582, |
| "step": 446500 |
| }, |
| { |
| "epoch": 14.54, |
| "learning_rate": 5.585858585858587e-06, |
| "loss": 0.8573, |
| "step": 447000 |
| }, |
| { |
| "epoch": 14.55, |
| "learning_rate": 5.580808080808081e-06, |
| "loss": 0.8581, |
| "step": 447500 |
| }, |
| { |
| "epoch": 14.57, |
| "learning_rate": 5.575757575757577e-06, |
| "loss": 0.8578, |
| "step": 448000 |
| }, |
| { |
| "epoch": 14.59, |
| "learning_rate": 5.570707070707071e-06, |
| "loss": 0.858, |
| "step": 448500 |
| }, |
| { |
| "epoch": 14.6, |
| "learning_rate": 5.565656565656566e-06, |
| "loss": 0.8566, |
| "step": 449000 |
| }, |
| { |
| "epoch": 14.62, |
| "learning_rate": 5.560606060606061e-06, |
| "loss": 0.8567, |
| "step": 449500 |
| }, |
| { |
| "epoch": 14.64, |
| "learning_rate": 5.555555555555557e-06, |
| "loss": 0.8566, |
| "step": 450000 |
| }, |
| { |
| "epoch": 14.65, |
| "learning_rate": 5.550505050505051e-06, |
| "loss": 0.8566, |
| "step": 450500 |
| }, |
| { |
| "epoch": 14.67, |
| "learning_rate": 5.545454545454546e-06, |
| "loss": 0.8566, |
| "step": 451000 |
| }, |
| { |
| "epoch": 14.68, |
| "learning_rate": 5.54040404040404e-06, |
| "loss": 0.8563, |
| "step": 451500 |
| }, |
| { |
| "epoch": 14.7, |
| "learning_rate": 5.5353535353535355e-06, |
| "loss": 0.8565, |
| "step": 452000 |
| }, |
| { |
| "epoch": 14.72, |
| "learning_rate": 5.530303030303031e-06, |
| "loss": 0.8566, |
| "step": 452500 |
| }, |
| { |
| "epoch": 14.73, |
| "learning_rate": 5.525252525252526e-06, |
| "loss": 0.8555, |
| "step": 453000 |
| }, |
| { |
| "epoch": 14.75, |
| "learning_rate": 5.52020202020202e-06, |
| "loss": 0.8557, |
| "step": 453500 |
| }, |
| { |
| "epoch": 14.77, |
| "learning_rate": 5.515151515151515e-06, |
| "loss": 0.8556, |
| "step": 454000 |
| }, |
| { |
| "epoch": 14.78, |
| "learning_rate": 5.510101010101011e-06, |
| "loss": 0.8553, |
| "step": 454500 |
| }, |
| { |
| "epoch": 14.8, |
| "learning_rate": 5.5050505050505056e-06, |
| "loss": 0.8544, |
| "step": 455000 |
| }, |
| { |
| "epoch": 14.81, |
| "learning_rate": 5.500000000000001e-06, |
| "loss": 0.8542, |
| "step": 455500 |
| }, |
| { |
| "epoch": 14.83, |
| "learning_rate": 5.494949494949495e-06, |
| "loss": 0.8538, |
| "step": 456000 |
| }, |
| { |
| "epoch": 14.85, |
| "learning_rate": 5.489898989898991e-06, |
| "loss": 0.8538, |
| "step": 456500 |
| }, |
| { |
| "epoch": 14.86, |
| "learning_rate": 5.484848484848485e-06, |
| "loss": 0.8543, |
| "step": 457000 |
| }, |
| { |
| "epoch": 14.88, |
| "learning_rate": 5.4797979797979804e-06, |
| "loss": 0.8538, |
| "step": 457500 |
| }, |
| { |
| "epoch": 14.9, |
| "learning_rate": 5.474747474747475e-06, |
| "loss": 0.8535, |
| "step": 458000 |
| }, |
| { |
| "epoch": 14.91, |
| "learning_rate": 5.469696969696971e-06, |
| "loss": 0.8531, |
| "step": 458500 |
| }, |
| { |
| "epoch": 14.93, |
| "learning_rate": 5.464646464646465e-06, |
| "loss": 0.8533, |
| "step": 459000 |
| }, |
| { |
| "epoch": 14.94, |
| "learning_rate": 5.45959595959596e-06, |
| "loss": 0.8527, |
| "step": 459500 |
| }, |
| { |
| "epoch": 14.96, |
| "learning_rate": 5.4545454545454545e-06, |
| "loss": 0.8525, |
| "step": 460000 |
| }, |
| { |
| "epoch": 14.98, |
| "learning_rate": 5.4494949494949504e-06, |
| "loss": 0.8522, |
| "step": 460500 |
| }, |
| { |
| "epoch": 14.99, |
| "learning_rate": 5.444444444444445e-06, |
| "loss": 0.8525, |
| "step": 461000 |
| }, |
| { |
| "epoch": 15.01, |
| "learning_rate": 5.43939393939394e-06, |
| "loss": 0.8517, |
| "step": 461500 |
| }, |
| { |
| "epoch": 15.03, |
| "learning_rate": 5.434343434343434e-06, |
| "loss": 0.852, |
| "step": 462000 |
| }, |
| { |
| "epoch": 15.04, |
| "learning_rate": 5.429292929292929e-06, |
| "loss": 0.852, |
| "step": 462500 |
| }, |
| { |
| "epoch": 15.06, |
| "learning_rate": 5.424242424242425e-06, |
| "loss": 0.8509, |
| "step": 463000 |
| }, |
| { |
| "epoch": 15.07, |
| "learning_rate": 5.41919191919192e-06, |
| "loss": 0.8512, |
| "step": 463500 |
| }, |
| { |
| "epoch": 15.09, |
| "learning_rate": 5.414141414141415e-06, |
| "loss": 0.8511, |
| "step": 464000 |
| }, |
| { |
| "epoch": 15.11, |
| "learning_rate": 5.409090909090909e-06, |
| "loss": 0.8509, |
| "step": 464500 |
| }, |
| { |
| "epoch": 15.12, |
| "learning_rate": 5.404040404040405e-06, |
| "loss": 0.8508, |
| "step": 465000 |
| }, |
| { |
| "epoch": 15.14, |
| "learning_rate": 5.398989898989899e-06, |
| "loss": 0.8512, |
| "step": 465500 |
| }, |
| { |
| "epoch": 15.16, |
| "learning_rate": 5.3939393939393945e-06, |
| "loss": 0.8497, |
| "step": 466000 |
| }, |
| { |
| "epoch": 15.17, |
| "learning_rate": 5.388888888888889e-06, |
| "loss": 0.8502, |
| "step": 466500 |
| }, |
| { |
| "epoch": 15.19, |
| "learning_rate": 5.383838383838385e-06, |
| "loss": 0.8493, |
| "step": 467000 |
| }, |
| { |
| "epoch": 15.2, |
| "learning_rate": 5.378787878787879e-06, |
| "loss": 0.8501, |
| "step": 467500 |
| }, |
| { |
| "epoch": 15.22, |
| "learning_rate": 5.373737373737374e-06, |
| "loss": 0.8488, |
| "step": 468000 |
| }, |
| { |
| "epoch": 15.24, |
| "learning_rate": 5.3686868686868685e-06, |
| "loss": 0.8491, |
| "step": 468500 |
| }, |
| { |
| "epoch": 15.25, |
| "learning_rate": 5.3636363636363645e-06, |
| "loss": 0.8496, |
| "step": 469000 |
| }, |
| { |
| "epoch": 15.27, |
| "learning_rate": 5.358585858585859e-06, |
| "loss": 0.8484, |
| "step": 469500 |
| }, |
| { |
| "epoch": 15.29, |
| "learning_rate": 5.353535353535354e-06, |
| "loss": 0.8489, |
| "step": 470000 |
| }, |
| { |
| "epoch": 15.3, |
| "learning_rate": 5.348484848484848e-06, |
| "loss": 0.849, |
| "step": 470500 |
| }, |
| { |
| "epoch": 15.32, |
| "learning_rate": 5.343434343434344e-06, |
| "loss": 0.8483, |
| "step": 471000 |
| }, |
| { |
| "epoch": 15.33, |
| "learning_rate": 5.338383838383839e-06, |
| "loss": 0.8481, |
| "step": 471500 |
| }, |
| { |
| "epoch": 15.35, |
| "learning_rate": 5.333333333333334e-06, |
| "loss": 0.8478, |
| "step": 472000 |
| }, |
| { |
| "epoch": 15.37, |
| "learning_rate": 5.328282828282829e-06, |
| "loss": 0.8474, |
| "step": 472500 |
| }, |
| { |
| "epoch": 15.38, |
| "learning_rate": 5.323232323232324e-06, |
| "loss": 0.8474, |
| "step": 473000 |
| }, |
| { |
| "epoch": 15.4, |
| "learning_rate": 5.318181818181819e-06, |
| "loss": 0.8468, |
| "step": 473500 |
| }, |
| { |
| "epoch": 15.42, |
| "learning_rate": 5.313131313131313e-06, |
| "loss": 0.8468, |
| "step": 474000 |
| }, |
| { |
| "epoch": 15.43, |
| "learning_rate": 5.3080808080808085e-06, |
| "loss": 0.8466, |
| "step": 474500 |
| }, |
| { |
| "epoch": 15.45, |
| "learning_rate": 5.303030303030303e-06, |
| "loss": 0.8468, |
| "step": 475000 |
| }, |
| { |
| "epoch": 15.46, |
| "learning_rate": 5.297979797979799e-06, |
| "loss": 0.8463, |
| "step": 475500 |
| }, |
| { |
| "epoch": 15.48, |
| "learning_rate": 5.292929292929293e-06, |
| "loss": 0.8461, |
| "step": 476000 |
| }, |
| { |
| "epoch": 15.5, |
| "learning_rate": 5.287878787878788e-06, |
| "loss": 0.8461, |
| "step": 476500 |
| }, |
| { |
| "epoch": 15.51, |
| "learning_rate": 5.2828282828282825e-06, |
| "loss": 0.8461, |
| "step": 477000 |
| }, |
| { |
| "epoch": 15.53, |
| "learning_rate": 5.2777777777777785e-06, |
| "loss": 0.8453, |
| "step": 477500 |
| }, |
| { |
| "epoch": 15.55, |
| "learning_rate": 5.272727272727273e-06, |
| "loss": 0.8454, |
| "step": 478000 |
| }, |
| { |
| "epoch": 15.56, |
| "learning_rate": 5.267676767676768e-06, |
| "loss": 0.8452, |
| "step": 478500 |
| }, |
| { |
| "epoch": 15.58, |
| "learning_rate": 5.262626262626262e-06, |
| "loss": 0.8456, |
| "step": 479000 |
| }, |
| { |
| "epoch": 15.59, |
| "learning_rate": 5.257575757575758e-06, |
| "loss": 0.8445, |
| "step": 479500 |
| }, |
| { |
| "epoch": 15.61, |
| "learning_rate": 5.252525252525253e-06, |
| "loss": 0.8446, |
| "step": 480000 |
| }, |
| { |
| "epoch": 15.63, |
| "learning_rate": 5.247474747474748e-06, |
| "loss": 0.8444, |
| "step": 480500 |
| }, |
| { |
| "epoch": 15.64, |
| "learning_rate": 5.242424242424244e-06, |
| "loss": 0.8442, |
| "step": 481000 |
| }, |
| { |
| "epoch": 15.66, |
| "learning_rate": 5.237373737373738e-06, |
| "loss": 0.8442, |
| "step": 481500 |
| }, |
| { |
| "epoch": 15.68, |
| "learning_rate": 5.232323232323233e-06, |
| "loss": 0.8438, |
| "step": 482000 |
| }, |
| { |
| "epoch": 15.69, |
| "learning_rate": 5.2272727272727274e-06, |
| "loss": 0.8439, |
| "step": 482500 |
| }, |
| { |
| "epoch": 15.71, |
| "learning_rate": 5.2222222222222226e-06, |
| "loss": 0.8443, |
| "step": 483000 |
| }, |
| { |
| "epoch": 15.72, |
| "learning_rate": 5.217171717171718e-06, |
| "loss": 0.8432, |
| "step": 483500 |
| }, |
| { |
| "epoch": 15.74, |
| "learning_rate": 5.212121212121213e-06, |
| "loss": 0.8439, |
| "step": 484000 |
| }, |
| { |
| "epoch": 15.76, |
| "learning_rate": 5.207070707070707e-06, |
| "loss": 0.8428, |
| "step": 484500 |
| }, |
| { |
| "epoch": 15.77, |
| "learning_rate": 5.202020202020202e-06, |
| "loss": 0.8434, |
| "step": 485000 |
| }, |
| { |
| "epoch": 15.79, |
| "learning_rate": 5.196969696969697e-06, |
| "loss": 0.8428, |
| "step": 485500 |
| }, |
| { |
| "epoch": 15.81, |
| "learning_rate": 5.191919191919193e-06, |
| "loss": 0.8425, |
| "step": 486000 |
| }, |
| { |
| "epoch": 15.82, |
| "learning_rate": 5.186868686868687e-06, |
| "loss": 0.8431, |
| "step": 486500 |
| }, |
| { |
| "epoch": 15.84, |
| "learning_rate": 5.181818181818182e-06, |
| "loss": 0.8422, |
| "step": 487000 |
| }, |
| { |
| "epoch": 15.85, |
| "learning_rate": 5.176767676767676e-06, |
| "loss": 0.8421, |
| "step": 487500 |
| }, |
| { |
| "epoch": 15.87, |
| "learning_rate": 5.171717171717172e-06, |
| "loss": 0.842, |
| "step": 488000 |
| }, |
| { |
| "epoch": 15.89, |
| "learning_rate": 5.1666666666666675e-06, |
| "loss": 0.8418, |
| "step": 488500 |
| }, |
| { |
| "epoch": 15.9, |
| "learning_rate": 5.161616161616162e-06, |
| "loss": 0.8412, |
| "step": 489000 |
| }, |
| { |
| "epoch": 15.92, |
| "learning_rate": 5.156565656565658e-06, |
| "loss": 0.8414, |
| "step": 489500 |
| }, |
| { |
| "epoch": 15.94, |
| "learning_rate": 5.151515151515152e-06, |
| "loss": 0.8409, |
| "step": 490000 |
| }, |
| { |
| "epoch": 15.95, |
| "learning_rate": 5.146464646464647e-06, |
| "loss": 0.8407, |
| "step": 490500 |
| }, |
| { |
| "epoch": 15.97, |
| "learning_rate": 5.1414141414141415e-06, |
| "loss": 0.8408, |
| "step": 491000 |
| }, |
| { |
| "epoch": 15.98, |
| "learning_rate": 5.1363636363636375e-06, |
| "loss": 0.8407, |
| "step": 491500 |
| }, |
| { |
| "epoch": 16.0, |
| "learning_rate": 5.131313131313132e-06, |
| "loss": 0.8407, |
| "step": 492000 |
| }, |
| { |
| "epoch": 16.02, |
| "learning_rate": 5.126262626262627e-06, |
| "loss": 0.8407, |
| "step": 492500 |
| }, |
| { |
| "epoch": 16.03, |
| "learning_rate": 5.121212121212121e-06, |
| "loss": 0.8395, |
| "step": 493000 |
| }, |
| { |
| "epoch": 16.05, |
| "learning_rate": 5.116161616161617e-06, |
| "loss": 0.84, |
| "step": 493500 |
| }, |
| { |
| "epoch": 16.07, |
| "learning_rate": 5.1111111111111115e-06, |
| "loss": 0.8394, |
| "step": 494000 |
| }, |
| { |
| "epoch": 16.08, |
| "learning_rate": 5.106060606060607e-06, |
| "loss": 0.8395, |
| "step": 494500 |
| }, |
| { |
| "epoch": 16.1, |
| "learning_rate": 5.101010101010101e-06, |
| "loss": 0.8391, |
| "step": 495000 |
| }, |
| { |
| "epoch": 16.11, |
| "learning_rate": 5.095959595959596e-06, |
| "loss": 0.8396, |
| "step": 495500 |
| }, |
| { |
| "epoch": 16.13, |
| "learning_rate": 5.090909090909091e-06, |
| "loss": 0.839, |
| "step": 496000 |
| }, |
| { |
| "epoch": 16.15, |
| "learning_rate": 5.085858585858586e-06, |
| "loss": 0.8383, |
| "step": 496500 |
| }, |
| { |
| "epoch": 16.16, |
| "learning_rate": 5.0808080808080815e-06, |
| "loss": 0.8389, |
| "step": 497000 |
| }, |
| { |
| "epoch": 16.18, |
| "learning_rate": 5.075757575757576e-06, |
| "loss": 0.8385, |
| "step": 497500 |
| }, |
| { |
| "epoch": 16.2, |
| "learning_rate": 5.070707070707072e-06, |
| "loss": 0.838, |
| "step": 498000 |
| }, |
| { |
| "epoch": 16.21, |
| "learning_rate": 5.065656565656566e-06, |
| "loss": 0.8377, |
| "step": 498500 |
| }, |
| { |
| "epoch": 16.23, |
| "learning_rate": 5.060606060606061e-06, |
| "loss": 0.8379, |
| "step": 499000 |
| }, |
| { |
| "epoch": 16.24, |
| "learning_rate": 5.0555555555555555e-06, |
| "loss": 0.8376, |
| "step": 499500 |
| }, |
| { |
| "epoch": 16.26, |
| "learning_rate": 5.0505050505050515e-06, |
| "loss": 0.8377, |
| "step": 500000 |
| }, |
| { |
| "epoch": 16.28, |
| "learning_rate": 5.045454545454546e-06, |
| "loss": 0.8374, |
| "step": 500500 |
| }, |
| { |
| "epoch": 16.29, |
| "learning_rate": 5.040404040404041e-06, |
| "loss": 0.8372, |
| "step": 501000 |
| }, |
| { |
| "epoch": 16.31, |
| "learning_rate": 5.035353535353535e-06, |
| "loss": 0.8362, |
| "step": 501500 |
| }, |
| { |
| "epoch": 16.33, |
| "learning_rate": 5.030303030303031e-06, |
| "loss": 0.8366, |
| "step": 502000 |
| }, |
| { |
| "epoch": 16.34, |
| "learning_rate": 5.0252525252525255e-06, |
| "loss": 0.837, |
| "step": 502500 |
| }, |
| { |
| "epoch": 16.36, |
| "learning_rate": 5.020202020202021e-06, |
| "loss": 0.8366, |
| "step": 503000 |
| }, |
| { |
| "epoch": 16.38, |
| "learning_rate": 5.015151515151515e-06, |
| "loss": 0.8363, |
| "step": 503500 |
| }, |
| { |
| "epoch": 16.39, |
| "learning_rate": 5.010101010101011e-06, |
| "loss": 0.8367, |
| "step": 504000 |
| }, |
| { |
| "epoch": 16.41, |
| "learning_rate": 5.005050505050505e-06, |
| "loss": 0.8362, |
| "step": 504500 |
| }, |
| { |
| "epoch": 16.42, |
| "learning_rate": 5e-06, |
| "loss": 0.8354, |
| "step": 505000 |
| }, |
| { |
| "epoch": 16.44, |
| "learning_rate": 4.9949494949494956e-06, |
| "loss": 0.8354, |
| "step": 505500 |
| }, |
| { |
| "epoch": 16.46, |
| "learning_rate": 4.98989898989899e-06, |
| "loss": 0.8359, |
| "step": 506000 |
| }, |
| { |
| "epoch": 16.47, |
| "learning_rate": 4.984848484848485e-06, |
| "loss": 0.8354, |
| "step": 506500 |
| }, |
| { |
| "epoch": 16.49, |
| "learning_rate": 4.97979797979798e-06, |
| "loss": 0.8358, |
| "step": 507000 |
| }, |
| { |
| "epoch": 16.51, |
| "learning_rate": 4.974747474747475e-06, |
| "loss": 0.8346, |
| "step": 507500 |
| }, |
| { |
| "epoch": 16.52, |
| "learning_rate": 4.9696969696969696e-06, |
| "loss": 0.8345, |
| "step": 508000 |
| }, |
| { |
| "epoch": 16.54, |
| "learning_rate": 4.964646464646465e-06, |
| "loss": 0.8346, |
| "step": 508500 |
| }, |
| { |
| "epoch": 16.55, |
| "learning_rate": 4.95959595959596e-06, |
| "loss": 0.834, |
| "step": 509000 |
| }, |
| { |
| "epoch": 16.57, |
| "learning_rate": 4.954545454545455e-06, |
| "loss": 0.8336, |
| "step": 509500 |
| }, |
| { |
| "epoch": 16.59, |
| "learning_rate": 4.94949494949495e-06, |
| "loss": 0.8342, |
| "step": 510000 |
| }, |
| { |
| "epoch": 16.6, |
| "learning_rate": 4.944444444444445e-06, |
| "loss": 0.8342, |
| "step": 510500 |
| }, |
| { |
| "epoch": 16.62, |
| "learning_rate": 4.93939393939394e-06, |
| "loss": 0.8339, |
| "step": 511000 |
| }, |
| { |
| "epoch": 16.64, |
| "learning_rate": 4.934343434343435e-06, |
| "loss": 0.8332, |
| "step": 511500 |
| }, |
| { |
| "epoch": 16.65, |
| "learning_rate": 4.92929292929293e-06, |
| "loss": 0.8334, |
| "step": 512000 |
| }, |
| { |
| "epoch": 16.67, |
| "learning_rate": 4.924242424242425e-06, |
| "loss": 0.834, |
| "step": 512500 |
| }, |
| { |
| "epoch": 16.68, |
| "learning_rate": 4.919191919191919e-06, |
| "loss": 0.8329, |
| "step": 513000 |
| }, |
| { |
| "epoch": 16.7, |
| "learning_rate": 4.9141414141414145e-06, |
| "loss": 0.833, |
| "step": 513500 |
| }, |
| { |
| "epoch": 16.72, |
| "learning_rate": 4.90909090909091e-06, |
| "loss": 0.8321, |
| "step": 514000 |
| }, |
| { |
| "epoch": 16.73, |
| "learning_rate": 4.904040404040405e-06, |
| "loss": 0.8329, |
| "step": 514500 |
| }, |
| { |
| "epoch": 16.75, |
| "learning_rate": 4.898989898989899e-06, |
| "loss": 0.8331, |
| "step": 515000 |
| }, |
| { |
| "epoch": 16.77, |
| "learning_rate": 4.893939393939394e-06, |
| "loss": 0.8321, |
| "step": 515500 |
| }, |
| { |
| "epoch": 16.78, |
| "learning_rate": 4.888888888888889e-06, |
| "loss": 0.8318, |
| "step": 516000 |
| }, |
| { |
| "epoch": 16.8, |
| "learning_rate": 4.883838383838384e-06, |
| "loss": 0.8313, |
| "step": 516500 |
| }, |
| { |
| "epoch": 16.81, |
| "learning_rate": 4.878787878787879e-06, |
| "loss": 0.8314, |
| "step": 517000 |
| }, |
| { |
| "epoch": 16.83, |
| "learning_rate": 4.873737373737374e-06, |
| "loss": 0.8314, |
| "step": 517500 |
| }, |
| { |
| "epoch": 16.85, |
| "learning_rate": 4.868686868686869e-06, |
| "loss": 0.831, |
| "step": 518000 |
| }, |
| { |
| "epoch": 16.86, |
| "learning_rate": 4.863636363636364e-06, |
| "loss": 0.8312, |
| "step": 518500 |
| }, |
| { |
| "epoch": 16.88, |
| "learning_rate": 4.858585858585859e-06, |
| "loss": 0.8313, |
| "step": 519000 |
| }, |
| { |
| "epoch": 16.9, |
| "learning_rate": 4.8535353535353545e-06, |
| "loss": 0.8309, |
| "step": 519500 |
| }, |
| { |
| "epoch": 16.91, |
| "learning_rate": 4.848484848484849e-06, |
| "loss": 0.8306, |
| "step": 520000 |
| }, |
| { |
| "epoch": 16.93, |
| "learning_rate": 4.843434343434344e-06, |
| "loss": 0.8304, |
| "step": 520500 |
| }, |
| { |
| "epoch": 16.94, |
| "learning_rate": 4.838383838383839e-06, |
| "loss": 0.83, |
| "step": 521000 |
| }, |
| { |
| "epoch": 16.96, |
| "learning_rate": 4.833333333333333e-06, |
| "loss": 0.8299, |
| "step": 521500 |
| }, |
| { |
| "epoch": 16.98, |
| "learning_rate": 4.8282828282828285e-06, |
| "loss": 0.8296, |
| "step": 522000 |
| }, |
| { |
| "epoch": 16.99, |
| "learning_rate": 4.823232323232324e-06, |
| "loss": 0.8294, |
| "step": 522500 |
| }, |
| { |
| "epoch": 17.01, |
| "learning_rate": 4.818181818181819e-06, |
| "loss": 0.8299, |
| "step": 523000 |
| }, |
| { |
| "epoch": 17.03, |
| "learning_rate": 4.813131313131313e-06, |
| "loss": 0.8294, |
| "step": 523500 |
| }, |
| { |
| "epoch": 17.04, |
| "learning_rate": 4.808080808080808e-06, |
| "loss": 0.8291, |
| "step": 524000 |
| }, |
| { |
| "epoch": 17.06, |
| "learning_rate": 4.803030303030303e-06, |
| "loss": 0.8286, |
| "step": 524500 |
| }, |
| { |
| "epoch": 17.07, |
| "learning_rate": 4.7979797979797985e-06, |
| "loss": 0.8286, |
| "step": 525000 |
| }, |
| { |
| "epoch": 17.09, |
| "learning_rate": 4.792929292929293e-06, |
| "loss": 0.829, |
| "step": 525500 |
| }, |
| { |
| "epoch": 17.11, |
| "learning_rate": 4.787878787878788e-06, |
| "loss": 0.8281, |
| "step": 526000 |
| }, |
| { |
| "epoch": 17.12, |
| "learning_rate": 4.782828282828283e-06, |
| "loss": 0.8277, |
| "step": 526500 |
| }, |
| { |
| "epoch": 17.14, |
| "learning_rate": 4.777777777777778e-06, |
| "loss": 0.8277, |
| "step": 527000 |
| }, |
| { |
| "epoch": 17.16, |
| "learning_rate": 4.772727272727273e-06, |
| "loss": 0.8269, |
| "step": 527500 |
| }, |
| { |
| "epoch": 17.17, |
| "learning_rate": 4.7676767676767685e-06, |
| "loss": 0.828, |
| "step": 528000 |
| }, |
| { |
| "epoch": 17.19, |
| "learning_rate": 4.762626262626263e-06, |
| "loss": 0.8279, |
| "step": 528500 |
| }, |
| { |
| "epoch": 17.2, |
| "learning_rate": 4.757575757575758e-06, |
| "loss": 0.8272, |
| "step": 529000 |
| }, |
| { |
| "epoch": 17.22, |
| "learning_rate": 4.752525252525253e-06, |
| "loss": 0.8272, |
| "step": 529500 |
| }, |
| { |
| "epoch": 17.24, |
| "learning_rate": 4.747474747474748e-06, |
| "loss": 0.8269, |
| "step": 530000 |
| }, |
| { |
| "epoch": 17.25, |
| "learning_rate": 4.7424242424242426e-06, |
| "loss": 0.8267, |
| "step": 530500 |
| }, |
| { |
| "epoch": 17.27, |
| "learning_rate": 4.737373737373738e-06, |
| "loss": 0.8268, |
| "step": 531000 |
| }, |
| { |
| "epoch": 17.29, |
| "learning_rate": 4.732323232323233e-06, |
| "loss": 0.8265, |
| "step": 531500 |
| }, |
| { |
| "epoch": 17.3, |
| "learning_rate": 4.727272727272728e-06, |
| "loss": 0.8263, |
| "step": 532000 |
| }, |
| { |
| "epoch": 17.32, |
| "learning_rate": 4.722222222222222e-06, |
| "loss": 0.8261, |
| "step": 532500 |
| }, |
| { |
| "epoch": 17.33, |
| "learning_rate": 4.717171717171717e-06, |
| "loss": 0.827, |
| "step": 533000 |
| }, |
| { |
| "epoch": 17.35, |
| "learning_rate": 4.7121212121212126e-06, |
| "loss": 0.8258, |
| "step": 533500 |
| }, |
| { |
| "epoch": 17.37, |
| "learning_rate": 4.707070707070707e-06, |
| "loss": 0.8255, |
| "step": 534000 |
| }, |
| { |
| "epoch": 17.38, |
| "learning_rate": 4.702020202020202e-06, |
| "loss": 0.8257, |
| "step": 534500 |
| }, |
| { |
| "epoch": 17.4, |
| "learning_rate": 4.696969696969698e-06, |
| "loss": 0.8253, |
| "step": 535000 |
| }, |
| { |
| "epoch": 17.42, |
| "learning_rate": 4.691919191919192e-06, |
| "loss": 0.8253, |
| "step": 535500 |
| }, |
| { |
| "epoch": 17.43, |
| "learning_rate": 4.6868686868686874e-06, |
| "loss": 0.8251, |
| "step": 536000 |
| }, |
| { |
| "epoch": 17.45, |
| "learning_rate": 4.681818181818183e-06, |
| "loss": 0.8246, |
| "step": 536500 |
| }, |
| { |
| "epoch": 17.46, |
| "learning_rate": 4.676767676767677e-06, |
| "loss": 0.8249, |
| "step": 537000 |
| }, |
| { |
| "epoch": 17.48, |
| "learning_rate": 4.671717171717172e-06, |
| "loss": 0.8245, |
| "step": 537500 |
| }, |
| { |
| "epoch": 17.5, |
| "learning_rate": 4.666666666666667e-06, |
| "loss": 0.825, |
| "step": 538000 |
| }, |
| { |
| "epoch": 17.51, |
| "learning_rate": 4.661616161616162e-06, |
| "loss": 0.8242, |
| "step": 538500 |
| }, |
| { |
| "epoch": 17.53, |
| "learning_rate": 4.656565656565657e-06, |
| "loss": 0.8234, |
| "step": 539000 |
| }, |
| { |
| "epoch": 17.55, |
| "learning_rate": 4.651515151515152e-06, |
| "loss": 0.8242, |
| "step": 539500 |
| }, |
| { |
| "epoch": 17.56, |
| "learning_rate": 4.646464646464647e-06, |
| "loss": 0.8236, |
| "step": 540000 |
| }, |
| { |
| "epoch": 17.58, |
| "learning_rate": 4.641414141414142e-06, |
| "loss": 0.8235, |
| "step": 540500 |
| }, |
| { |
| "epoch": 17.59, |
| "learning_rate": 4.636363636363636e-06, |
| "loss": 0.8235, |
| "step": 541000 |
| }, |
| { |
| "epoch": 17.61, |
| "learning_rate": 4.6313131313131315e-06, |
| "loss": 0.8239, |
| "step": 541500 |
| }, |
| { |
| "epoch": 17.63, |
| "learning_rate": 4.626262626262627e-06, |
| "loss": 0.8229, |
| "step": 542000 |
| }, |
| { |
| "epoch": 17.64, |
| "learning_rate": 4.621212121212122e-06, |
| "loss": 0.823, |
| "step": 542500 |
| }, |
| { |
| "epoch": 17.66, |
| "learning_rate": 4.616161616161616e-06, |
| "loss": 0.8229, |
| "step": 543000 |
| }, |
| { |
| "epoch": 17.68, |
| "learning_rate": 4.611111111111112e-06, |
| "loss": 0.8225, |
| "step": 543500 |
| }, |
| { |
| "epoch": 17.69, |
| "learning_rate": 4.606060606060606e-06, |
| "loss": 0.8231, |
| "step": 544000 |
| }, |
| { |
| "epoch": 17.71, |
| "learning_rate": 4.6010101010101015e-06, |
| "loss": 0.8214, |
| "step": 544500 |
| }, |
| { |
| "epoch": 17.72, |
| "learning_rate": 4.595959595959597e-06, |
| "loss": 0.8226, |
| "step": 545000 |
| }, |
| { |
| "epoch": 17.74, |
| "learning_rate": 4.590909090909092e-06, |
| "loss": 0.8219, |
| "step": 545500 |
| }, |
| { |
| "epoch": 17.76, |
| "learning_rate": 4.585858585858586e-06, |
| "loss": 0.8215, |
| "step": 546000 |
| }, |
| { |
| "epoch": 17.77, |
| "learning_rate": 4.580808080808081e-06, |
| "loss": 0.8221, |
| "step": 546500 |
| }, |
| { |
| "epoch": 17.79, |
| "learning_rate": 4.575757575757576e-06, |
| "loss": 0.8219, |
| "step": 547000 |
| }, |
| { |
| "epoch": 17.81, |
| "learning_rate": 4.5707070707070715e-06, |
| "loss": 0.8214, |
| "step": 547500 |
| }, |
| { |
| "epoch": 17.82, |
| "learning_rate": 4.565656565656566e-06, |
| "loss": 0.8215, |
| "step": 548000 |
| }, |
| { |
| "epoch": 17.84, |
| "learning_rate": 4.560606060606061e-06, |
| "loss": 0.8215, |
| "step": 548500 |
| }, |
| { |
| "epoch": 17.85, |
| "learning_rate": 4.555555555555556e-06, |
| "loss": 0.8204, |
| "step": 549000 |
| }, |
| { |
| "epoch": 17.87, |
| "learning_rate": 4.55050505050505e-06, |
| "loss": 0.8205, |
| "step": 549500 |
| }, |
| { |
| "epoch": 17.89, |
| "learning_rate": 4.5454545454545455e-06, |
| "loss": 0.8211, |
| "step": 550000 |
| }, |
| { |
| "epoch": 17.9, |
| "learning_rate": 4.540404040404041e-06, |
| "loss": 0.8205, |
| "step": 550500 |
| }, |
| { |
| "epoch": 17.92, |
| "learning_rate": 4.535353535353536e-06, |
| "loss": 0.8207, |
| "step": 551000 |
| }, |
| { |
| "epoch": 17.94, |
| "learning_rate": 4.53030303030303e-06, |
| "loss": 0.8197, |
| "step": 551500 |
| }, |
| { |
| "epoch": 17.95, |
| "learning_rate": 4.525252525252526e-06, |
| "loss": 0.8198, |
| "step": 552000 |
| }, |
| { |
| "epoch": 17.97, |
| "learning_rate": 4.520202020202021e-06, |
| "loss": 0.82, |
| "step": 552500 |
| }, |
| { |
| "epoch": 17.98, |
| "learning_rate": 4.5151515151515155e-06, |
| "loss": 0.8198, |
| "step": 553000 |
| }, |
| { |
| "epoch": 18.0, |
| "learning_rate": 4.510101010101011e-06, |
| "loss": 0.8198, |
| "step": 553500 |
| }, |
| { |
| "epoch": 18.02, |
| "learning_rate": 4.505050505050506e-06, |
| "loss": 0.8194, |
| "step": 554000 |
| }, |
| { |
| "epoch": 18.03, |
| "learning_rate": 4.5e-06, |
| "loss": 0.8195, |
| "step": 554500 |
| }, |
| { |
| "epoch": 18.05, |
| "learning_rate": 4.494949494949495e-06, |
| "loss": 0.8192, |
| "step": 555000 |
| }, |
| { |
| "epoch": 18.07, |
| "learning_rate": 4.48989898989899e-06, |
| "loss": 0.8188, |
| "step": 555500 |
| }, |
| { |
| "epoch": 18.08, |
| "learning_rate": 4.4848484848484855e-06, |
| "loss": 0.8188, |
| "step": 556000 |
| }, |
| { |
| "epoch": 18.1, |
| "learning_rate": 4.47979797979798e-06, |
| "loss": 0.8183, |
| "step": 556500 |
| }, |
| { |
| "epoch": 18.11, |
| "learning_rate": 4.474747474747475e-06, |
| "loss": 0.8184, |
| "step": 557000 |
| }, |
| { |
| "epoch": 18.13, |
| "learning_rate": 4.46969696969697e-06, |
| "loss": 0.8182, |
| "step": 557500 |
| }, |
| { |
| "epoch": 18.15, |
| "learning_rate": 4.464646464646465e-06, |
| "loss": 0.8185, |
| "step": 558000 |
| }, |
| { |
| "epoch": 18.16, |
| "learning_rate": 4.4595959595959596e-06, |
| "loss": 0.818, |
| "step": 558500 |
| }, |
| { |
| "epoch": 18.18, |
| "learning_rate": 4.454545454545455e-06, |
| "loss": 0.8176, |
| "step": 559000 |
| }, |
| { |
| "epoch": 18.2, |
| "learning_rate": 4.44949494949495e-06, |
| "loss": 0.8177, |
| "step": 559500 |
| }, |
| { |
| "epoch": 18.21, |
| "learning_rate": 4.444444444444444e-06, |
| "loss": 0.8175, |
| "step": 560000 |
| }, |
| { |
| "epoch": 18.23, |
| "learning_rate": 4.43939393939394e-06, |
| "loss": 0.8176, |
| "step": 560500 |
| }, |
| { |
| "epoch": 18.25, |
| "learning_rate": 4.434343434343435e-06, |
| "loss": 0.817, |
| "step": 561000 |
| }, |
| { |
| "epoch": 18.26, |
| "learning_rate": 4.42929292929293e-06, |
| "loss": 0.8165, |
| "step": 561500 |
| }, |
| { |
| "epoch": 18.28, |
| "learning_rate": 4.424242424242425e-06, |
| "loss": 0.8167, |
| "step": 562000 |
| }, |
| { |
| "epoch": 18.29, |
| "learning_rate": 4.41919191919192e-06, |
| "loss": 0.8175, |
| "step": 562500 |
| }, |
| { |
| "epoch": 18.31, |
| "learning_rate": 4.414141414141415e-06, |
| "loss": 0.8166, |
| "step": 563000 |
| }, |
| { |
| "epoch": 18.33, |
| "learning_rate": 4.409090909090909e-06, |
| "loss": 0.8167, |
| "step": 563500 |
| }, |
| { |
| "epoch": 18.34, |
| "learning_rate": 4.4040404040404044e-06, |
| "loss": 0.8167, |
| "step": 564000 |
| }, |
| { |
| "epoch": 18.36, |
| "learning_rate": 4.3989898989899e-06, |
| "loss": 0.8166, |
| "step": 564500 |
| }, |
| { |
| "epoch": 18.38, |
| "learning_rate": 4.393939393939394e-06, |
| "loss": 0.8163, |
| "step": 565000 |
| }, |
| { |
| "epoch": 18.39, |
| "learning_rate": 4.388888888888889e-06, |
| "loss": 0.8156, |
| "step": 565500 |
| }, |
| { |
| "epoch": 18.41, |
| "learning_rate": 4.383838383838384e-06, |
| "loss": 0.8155, |
| "step": 566000 |
| }, |
| { |
| "epoch": 18.42, |
| "learning_rate": 4.378787878787879e-06, |
| "loss": 0.8156, |
| "step": 566500 |
| }, |
| { |
| "epoch": 18.44, |
| "learning_rate": 4.373737373737374e-06, |
| "loss": 0.8152, |
| "step": 567000 |
| }, |
| { |
| "epoch": 18.46, |
| "learning_rate": 4.368686868686869e-06, |
| "loss": 0.8155, |
| "step": 567500 |
| }, |
| { |
| "epoch": 18.47, |
| "learning_rate": 4.363636363636364e-06, |
| "loss": 0.8149, |
| "step": 568000 |
| }, |
| { |
| "epoch": 18.49, |
| "learning_rate": 4.358585858585859e-06, |
| "loss": 0.8152, |
| "step": 568500 |
| }, |
| { |
| "epoch": 18.51, |
| "learning_rate": 4.353535353535353e-06, |
| "loss": 0.8145, |
| "step": 569000 |
| }, |
| { |
| "epoch": 18.52, |
| "learning_rate": 4.348484848484849e-06, |
| "loss": 0.8148, |
| "step": 569500 |
| }, |
| { |
| "epoch": 18.54, |
| "learning_rate": 4.343434343434344e-06, |
| "loss": 0.8147, |
| "step": 570000 |
| }, |
| { |
| "epoch": 18.55, |
| "learning_rate": 4.338383838383839e-06, |
| "loss": 0.8143, |
| "step": 570500 |
| }, |
| { |
| "epoch": 18.57, |
| "learning_rate": 4.333333333333334e-06, |
| "loss": 0.8143, |
| "step": 571000 |
| }, |
| { |
| "epoch": 18.59, |
| "learning_rate": 4.328282828282829e-06, |
| "loss": 0.8144, |
| "step": 571500 |
| }, |
| { |
| "epoch": 18.6, |
| "learning_rate": 4.323232323232323e-06, |
| "loss": 0.8138, |
| "step": 572000 |
| }, |
| { |
| "epoch": 18.62, |
| "learning_rate": 4.3181818181818185e-06, |
| "loss": 0.8138, |
| "step": 572500 |
| }, |
| { |
| "epoch": 18.64, |
| "learning_rate": 4.313131313131314e-06, |
| "loss": 0.8141, |
| "step": 573000 |
| }, |
| { |
| "epoch": 18.65, |
| "learning_rate": 4.308080808080809e-06, |
| "loss": 0.8134, |
| "step": 573500 |
| }, |
| { |
| "epoch": 18.67, |
| "learning_rate": 4.303030303030303e-06, |
| "loss": 0.8137, |
| "step": 574000 |
| }, |
| { |
| "epoch": 18.68, |
| "learning_rate": 4.297979797979798e-06, |
| "loss": 0.8135, |
| "step": 574500 |
| }, |
| { |
| "epoch": 18.7, |
| "learning_rate": 4.292929292929293e-06, |
| "loss": 0.8127, |
| "step": 575000 |
| }, |
| { |
| "epoch": 18.72, |
| "learning_rate": 4.287878787878788e-06, |
| "loss": 0.8135, |
| "step": 575500 |
| }, |
| { |
| "epoch": 18.73, |
| "learning_rate": 4.282828282828283e-06, |
| "loss": 0.8134, |
| "step": 576000 |
| }, |
| { |
| "epoch": 18.75, |
| "learning_rate": 4.277777777777778e-06, |
| "loss": 0.8127, |
| "step": 576500 |
| }, |
| { |
| "epoch": 18.77, |
| "learning_rate": 4.272727272727273e-06, |
| "loss": 0.812, |
| "step": 577000 |
| }, |
| { |
| "epoch": 18.78, |
| "learning_rate": 4.267676767676767e-06, |
| "loss": 0.8122, |
| "step": 577500 |
| }, |
| { |
| "epoch": 18.8, |
| "learning_rate": 4.262626262626263e-06, |
| "loss": 0.8124, |
| "step": 578000 |
| }, |
| { |
| "epoch": 18.81, |
| "learning_rate": 4.2575757575757585e-06, |
| "loss": 0.8125, |
| "step": 578500 |
| }, |
| { |
| "epoch": 18.83, |
| "learning_rate": 4.252525252525253e-06, |
| "loss": 0.8117, |
| "step": 579000 |
| }, |
| { |
| "epoch": 18.85, |
| "learning_rate": 4.247474747474748e-06, |
| "loss": 0.8114, |
| "step": 579500 |
| }, |
| { |
| "epoch": 18.86, |
| "learning_rate": 4.242424242424243e-06, |
| "loss": 0.8118, |
| "step": 580000 |
| }, |
| { |
| "epoch": 18.88, |
| "learning_rate": 4.237373737373737e-06, |
| "loss": 0.8115, |
| "step": 580500 |
| }, |
| { |
| "epoch": 18.9, |
| "learning_rate": 4.2323232323232325e-06, |
| "loss": 0.8118, |
| "step": 581000 |
| }, |
| { |
| "epoch": 18.91, |
| "learning_rate": 4.227272727272728e-06, |
| "loss": 0.8113, |
| "step": 581500 |
| }, |
| { |
| "epoch": 18.93, |
| "learning_rate": 4.222222222222223e-06, |
| "loss": 0.8115, |
| "step": 582000 |
| }, |
| { |
| "epoch": 18.94, |
| "learning_rate": 4.217171717171717e-06, |
| "loss": 0.8107, |
| "step": 582500 |
| }, |
| { |
| "epoch": 18.96, |
| "learning_rate": 4.212121212121212e-06, |
| "loss": 0.811, |
| "step": 583000 |
| }, |
| { |
| "epoch": 18.98, |
| "learning_rate": 4.207070707070707e-06, |
| "loss": 0.811, |
| "step": 583500 |
| }, |
| { |
| "epoch": 18.99, |
| "learning_rate": 4.2020202020202026e-06, |
| "loss": 0.8107, |
| "step": 584000 |
| }, |
| { |
| "epoch": 19.01, |
| "learning_rate": 4.196969696969697e-06, |
| "loss": 0.8102, |
| "step": 584500 |
| }, |
| { |
| "epoch": 19.03, |
| "learning_rate": 4.191919191919192e-06, |
| "loss": 0.8105, |
| "step": 585000 |
| }, |
| { |
| "epoch": 19.04, |
| "learning_rate": 4.186868686868687e-06, |
| "loss": 0.8094, |
| "step": 585500 |
| }, |
| { |
| "epoch": 19.06, |
| "learning_rate": 4.181818181818182e-06, |
| "loss": 0.8096, |
| "step": 586000 |
| }, |
| { |
| "epoch": 19.07, |
| "learning_rate": 4.1767676767676774e-06, |
| "loss": 0.8095, |
| "step": 586500 |
| }, |
| { |
| "epoch": 19.09, |
| "learning_rate": 4.1717171717171726e-06, |
| "loss": 0.8097, |
| "step": 587000 |
| }, |
| { |
| "epoch": 19.11, |
| "learning_rate": 4.166666666666667e-06, |
| "loss": 0.8101, |
| "step": 587500 |
| }, |
| { |
| "epoch": 19.12, |
| "learning_rate": 4.161616161616162e-06, |
| "loss": 0.8092, |
| "step": 588000 |
| }, |
| { |
| "epoch": 19.14, |
| "learning_rate": 4.156565656565657e-06, |
| "loss": 0.8089, |
| "step": 588500 |
| }, |
| { |
| "epoch": 19.16, |
| "learning_rate": 4.151515151515152e-06, |
| "loss": 0.8092, |
| "step": 589000 |
| }, |
| { |
| "epoch": 19.17, |
| "learning_rate": 4.146464646464647e-06, |
| "loss": 0.809, |
| "step": 589500 |
| }, |
| { |
| "epoch": 19.19, |
| "learning_rate": 4.141414141414142e-06, |
| "loss": 0.8097, |
| "step": 590000 |
| }, |
| { |
| "epoch": 19.2, |
| "learning_rate": 4.136363636363637e-06, |
| "loss": 0.8085, |
| "step": 590500 |
| }, |
| { |
| "epoch": 19.22, |
| "learning_rate": 4.131313131313132e-06, |
| "loss": 0.8091, |
| "step": 591000 |
| }, |
| { |
| "epoch": 19.24, |
| "learning_rate": 4.126262626262626e-06, |
| "loss": 0.8085, |
| "step": 591500 |
| }, |
| { |
| "epoch": 19.25, |
| "learning_rate": 4.1212121212121215e-06, |
| "loss": 0.8084, |
| "step": 592000 |
| }, |
| { |
| "epoch": 19.27, |
| "learning_rate": 4.116161616161617e-06, |
| "loss": 0.8078, |
| "step": 592500 |
| }, |
| { |
| "epoch": 19.29, |
| "learning_rate": 4.111111111111111e-06, |
| "loss": 0.8082, |
| "step": 593000 |
| }, |
| { |
| "epoch": 19.3, |
| "learning_rate": 4.106060606060606e-06, |
| "loss": 0.8083, |
| "step": 593500 |
| }, |
| { |
| "epoch": 19.32, |
| "learning_rate": 4.101010101010101e-06, |
| "loss": 0.8081, |
| "step": 594000 |
| }, |
| { |
| "epoch": 19.33, |
| "learning_rate": 4.095959595959596e-06, |
| "loss": 0.8076, |
| "step": 594500 |
| }, |
| { |
| "epoch": 19.35, |
| "learning_rate": 4.0909090909090915e-06, |
| "loss": 0.8077, |
| "step": 595000 |
| }, |
| { |
| "epoch": 19.37, |
| "learning_rate": 4.085858585858587e-06, |
| "loss": 0.8072, |
| "step": 595500 |
| }, |
| { |
| "epoch": 19.38, |
| "learning_rate": 4.080808080808081e-06, |
| "loss": 0.8069, |
| "step": 596000 |
| }, |
| { |
| "epoch": 19.4, |
| "learning_rate": 4.075757575757576e-06, |
| "loss": 0.807, |
| "step": 596500 |
| }, |
| { |
| "epoch": 19.42, |
| "learning_rate": 4.070707070707071e-06, |
| "loss": 0.8069, |
| "step": 597000 |
| }, |
| { |
| "epoch": 19.43, |
| "learning_rate": 4.065656565656566e-06, |
| "loss": 0.8068, |
| "step": 597500 |
| }, |
| { |
| "epoch": 19.45, |
| "learning_rate": 4.060606060606061e-06, |
| "loss": 0.8061, |
| "step": 598000 |
| }, |
| { |
| "epoch": 19.46, |
| "learning_rate": 4.055555555555556e-06, |
| "loss": 0.8072, |
| "step": 598500 |
| }, |
| { |
| "epoch": 19.48, |
| "learning_rate": 4.050505050505051e-06, |
| "loss": 0.8064, |
| "step": 599000 |
| }, |
| { |
| "epoch": 19.5, |
| "learning_rate": 4.045454545454546e-06, |
| "loss": 0.8063, |
| "step": 599500 |
| }, |
| { |
| "epoch": 19.51, |
| "learning_rate": 4.04040404040404e-06, |
| "loss": 0.8064, |
| "step": 600000 |
| }, |
| { |
| "epoch": 19.53, |
| "learning_rate": 4.0353535353535355e-06, |
| "loss": 0.806, |
| "step": 600500 |
| }, |
| { |
| "epoch": 19.55, |
| "learning_rate": 4.030303030303031e-06, |
| "loss": 0.805, |
| "step": 601000 |
| }, |
| { |
| "epoch": 19.56, |
| "learning_rate": 4.025252525252526e-06, |
| "loss": 0.806, |
| "step": 601500 |
| }, |
| { |
| "epoch": 19.58, |
| "learning_rate": 4.02020202020202e-06, |
| "loss": 0.806, |
| "step": 602000 |
| }, |
| { |
| "epoch": 19.59, |
| "learning_rate": 4.015151515151515e-06, |
| "loss": 0.8049, |
| "step": 602500 |
| }, |
| { |
| "epoch": 19.61, |
| "learning_rate": 4.01010101010101e-06, |
| "loss": 0.8054, |
| "step": 603000 |
| }, |
| { |
| "epoch": 19.63, |
| "learning_rate": 4.0050505050505055e-06, |
| "loss": 0.8054, |
| "step": 603500 |
| }, |
| { |
| "epoch": 19.64, |
| "learning_rate": 4.000000000000001e-06, |
| "loss": 0.8051, |
| "step": 604000 |
| }, |
| { |
| "epoch": 19.66, |
| "learning_rate": 3.994949494949496e-06, |
| "loss": 0.8047, |
| "step": 604500 |
| }, |
| { |
| "epoch": 19.68, |
| "learning_rate": 3.98989898989899e-06, |
| "loss": 0.8048, |
| "step": 605000 |
| }, |
| { |
| "epoch": 19.69, |
| "learning_rate": 3.984848484848485e-06, |
| "loss": 0.8046, |
| "step": 605500 |
| }, |
| { |
| "epoch": 19.71, |
| "learning_rate": 3.97979797979798e-06, |
| "loss": 0.8048, |
| "step": 606000 |
| }, |
| { |
| "epoch": 19.72, |
| "learning_rate": 3.9747474747474755e-06, |
| "loss": 0.8044, |
| "step": 606500 |
| }, |
| { |
| "epoch": 19.74, |
| "learning_rate": 3.96969696969697e-06, |
| "loss": 0.8045, |
| "step": 607000 |
| }, |
| { |
| "epoch": 19.76, |
| "learning_rate": 3.964646464646465e-06, |
| "loss": 0.8049, |
| "step": 607500 |
| }, |
| { |
| "epoch": 19.77, |
| "learning_rate": 3.95959595959596e-06, |
| "loss": 0.8036, |
| "step": 608000 |
| }, |
| { |
| "epoch": 19.79, |
| "learning_rate": 3.954545454545454e-06, |
| "loss": 0.8035, |
| "step": 608500 |
| }, |
| { |
| "epoch": 19.81, |
| "learning_rate": 3.9494949494949496e-06, |
| "loss": 0.8034, |
| "step": 609000 |
| }, |
| { |
| "epoch": 19.82, |
| "learning_rate": 3.944444444444445e-06, |
| "loss": 0.8033, |
| "step": 609500 |
| }, |
| { |
| "epoch": 19.84, |
| "learning_rate": 3.93939393939394e-06, |
| "loss": 0.8026, |
| "step": 610000 |
| }, |
| { |
| "epoch": 19.85, |
| "learning_rate": 3.934343434343434e-06, |
| "loss": 0.8033, |
| "step": 610500 |
| }, |
| { |
| "epoch": 19.87, |
| "learning_rate": 3.929292929292929e-06, |
| "loss": 0.8036, |
| "step": 611000 |
| }, |
| { |
| "epoch": 19.89, |
| "learning_rate": 3.9242424242424244e-06, |
| "loss": 0.8033, |
| "step": 611500 |
| }, |
| { |
| "epoch": 19.9, |
| "learning_rate": 3.9191919191919196e-06, |
| "loss": 0.803, |
| "step": 612000 |
| }, |
| { |
| "epoch": 19.92, |
| "learning_rate": 3.914141414141415e-06, |
| "loss": 0.8024, |
| "step": 612500 |
| }, |
| { |
| "epoch": 19.94, |
| "learning_rate": 3.90909090909091e-06, |
| "loss": 0.8028, |
| "step": 613000 |
| }, |
| { |
| "epoch": 19.95, |
| "learning_rate": 3.904040404040404e-06, |
| "loss": 0.8027, |
| "step": 613500 |
| }, |
| { |
| "epoch": 19.97, |
| "learning_rate": 3.898989898989899e-06, |
| "loss": 0.8025, |
| "step": 614000 |
| }, |
| { |
| "epoch": 19.99, |
| "learning_rate": 3.8939393939393944e-06, |
| "loss": 0.802, |
| "step": 614500 |
| }, |
| { |
| "epoch": 20.0, |
| "learning_rate": 3.88888888888889e-06, |
| "loss": 0.8017, |
| "step": 615000 |
| }, |
| { |
| "epoch": 20.02, |
| "learning_rate": 3.883838383838384e-06, |
| "loss": 0.802, |
| "step": 615500 |
| }, |
| { |
| "epoch": 20.03, |
| "learning_rate": 3.878787878787879e-06, |
| "loss": 0.8021, |
| "step": 616000 |
| }, |
| { |
| "epoch": 20.05, |
| "learning_rate": 3.873737373737374e-06, |
| "loss": 0.8017, |
| "step": 616500 |
| }, |
| { |
| "epoch": 20.07, |
| "learning_rate": 3.868686868686869e-06, |
| "loss": 0.8017, |
| "step": 617000 |
| }, |
| { |
| "epoch": 20.08, |
| "learning_rate": 3.863636363636364e-06, |
| "loss": 0.8013, |
| "step": 617500 |
| }, |
| { |
| "epoch": 20.1, |
| "learning_rate": 3.858585858585859e-06, |
| "loss": 0.8011, |
| "step": 618000 |
| }, |
| { |
| "epoch": 20.12, |
| "learning_rate": 3.853535353535354e-06, |
| "loss": 0.801, |
| "step": 618500 |
| }, |
| { |
| "epoch": 20.13, |
| "learning_rate": 3.848484848484848e-06, |
| "loss": 0.8009, |
| "step": 619000 |
| }, |
| { |
| "epoch": 20.15, |
| "learning_rate": 3.843434343434343e-06, |
| "loss": 0.8008, |
| "step": 619500 |
| }, |
| { |
| "epoch": 20.16, |
| "learning_rate": 3.8383838383838385e-06, |
| "loss": 0.8013, |
| "step": 620000 |
| }, |
| { |
| "epoch": 20.18, |
| "learning_rate": 3.833333333333334e-06, |
| "loss": 0.8004, |
| "step": 620500 |
| }, |
| { |
| "epoch": 20.2, |
| "learning_rate": 3.828282828282829e-06, |
| "loss": 0.8003, |
| "step": 621000 |
| }, |
| { |
| "epoch": 20.21, |
| "learning_rate": 3.823232323232324e-06, |
| "loss": 0.8002, |
| "step": 621500 |
| }, |
| { |
| "epoch": 20.23, |
| "learning_rate": 3.818181818181819e-06, |
| "loss": 0.8004, |
| "step": 622000 |
| }, |
| { |
| "epoch": 20.25, |
| "learning_rate": 3.8131313131313138e-06, |
| "loss": 0.8002, |
| "step": 622500 |
| }, |
| { |
| "epoch": 20.26, |
| "learning_rate": 3.8080808080808085e-06, |
| "loss": 0.8, |
| "step": 623000 |
| }, |
| { |
| "epoch": 20.28, |
| "learning_rate": 3.803030303030303e-06, |
| "loss": 0.7997, |
| "step": 623500 |
| }, |
| { |
| "epoch": 20.29, |
| "learning_rate": 3.7979797979797984e-06, |
| "loss": 0.7992, |
| "step": 624000 |
| }, |
| { |
| "epoch": 20.31, |
| "learning_rate": 3.792929292929293e-06, |
| "loss": 0.7993, |
| "step": 624500 |
| }, |
| { |
| "epoch": 20.33, |
| "learning_rate": 3.7878787878787882e-06, |
| "loss": 0.7995, |
| "step": 625000 |
| }, |
| { |
| "epoch": 20.34, |
| "learning_rate": 3.782828282828283e-06, |
| "loss": 0.7992, |
| "step": 625500 |
| }, |
| { |
| "epoch": 20.36, |
| "learning_rate": 3.777777777777778e-06, |
| "loss": 0.7988, |
| "step": 626000 |
| }, |
| { |
| "epoch": 20.38, |
| "learning_rate": 3.772727272727273e-06, |
| "loss": 0.7995, |
| "step": 626500 |
| }, |
| { |
| "epoch": 20.39, |
| "learning_rate": 3.767676767676768e-06, |
| "loss": 0.7983, |
| "step": 627000 |
| }, |
| { |
| "epoch": 20.41, |
| "learning_rate": 3.7626262626262627e-06, |
| "loss": 0.7985, |
| "step": 627500 |
| }, |
| { |
| "epoch": 20.42, |
| "learning_rate": 3.757575757575758e-06, |
| "loss": 0.7979, |
| "step": 628000 |
| }, |
| { |
| "epoch": 20.44, |
| "learning_rate": 3.7525252525252525e-06, |
| "loss": 0.7981, |
| "step": 628500 |
| }, |
| { |
| "epoch": 20.46, |
| "learning_rate": 3.747474747474748e-06, |
| "loss": 0.7981, |
| "step": 629000 |
| }, |
| { |
| "epoch": 20.47, |
| "learning_rate": 3.742424242424243e-06, |
| "loss": 0.7981, |
| "step": 629500 |
| }, |
| { |
| "epoch": 20.49, |
| "learning_rate": 3.737373737373738e-06, |
| "loss": 0.7975, |
| "step": 630000 |
| }, |
| { |
| "epoch": 20.51, |
| "learning_rate": 3.7323232323232327e-06, |
| "loss": 0.7974, |
| "step": 630500 |
| }, |
| { |
| "epoch": 20.52, |
| "learning_rate": 3.727272727272728e-06, |
| "loss": 0.7975, |
| "step": 631000 |
| }, |
| { |
| "epoch": 20.54, |
| "learning_rate": 3.7222222222222225e-06, |
| "loss": 0.7972, |
| "step": 631500 |
| }, |
| { |
| "epoch": 20.55, |
| "learning_rate": 3.7171717171717177e-06, |
| "loss": 0.7977, |
| "step": 632000 |
| }, |
| { |
| "epoch": 20.57, |
| "learning_rate": 3.7121212121212124e-06, |
| "loss": 0.7972, |
| "step": 632500 |
| }, |
| { |
| "epoch": 20.59, |
| "learning_rate": 3.7070707070707075e-06, |
| "loss": 0.7966, |
| "step": 633000 |
| }, |
| { |
| "epoch": 20.6, |
| "learning_rate": 3.7020202020202023e-06, |
| "loss": 0.7968, |
| "step": 633500 |
| }, |
| { |
| "epoch": 20.62, |
| "learning_rate": 3.6969696969696974e-06, |
| "loss": 0.7971, |
| "step": 634000 |
| }, |
| { |
| "epoch": 20.64, |
| "learning_rate": 3.691919191919192e-06, |
| "loss": 0.7968, |
| "step": 634500 |
| }, |
| { |
| "epoch": 20.65, |
| "learning_rate": 3.686868686868687e-06, |
| "loss": 0.7963, |
| "step": 635000 |
| }, |
| { |
| "epoch": 20.67, |
| "learning_rate": 3.681818181818182e-06, |
| "loss": 0.7959, |
| "step": 635500 |
| }, |
| { |
| "epoch": 20.68, |
| "learning_rate": 3.6767676767676767e-06, |
| "loss": 0.796, |
| "step": 636000 |
| }, |
| { |
| "epoch": 20.7, |
| "learning_rate": 3.671717171717172e-06, |
| "loss": 0.7956, |
| "step": 636500 |
| }, |
| { |
| "epoch": 20.72, |
| "learning_rate": 3.6666666666666666e-06, |
| "loss": 0.7961, |
| "step": 637000 |
| }, |
| { |
| "epoch": 20.73, |
| "learning_rate": 3.661616161616162e-06, |
| "loss": 0.7957, |
| "step": 637500 |
| }, |
| { |
| "epoch": 20.75, |
| "learning_rate": 3.6565656565656573e-06, |
| "loss": 0.7957, |
| "step": 638000 |
| }, |
| { |
| "epoch": 20.77, |
| "learning_rate": 3.651515151515152e-06, |
| "loss": 0.7955, |
| "step": 638500 |
| }, |
| { |
| "epoch": 20.78, |
| "learning_rate": 3.6464646464646467e-06, |
| "loss": 0.7953, |
| "step": 639000 |
| }, |
| { |
| "epoch": 20.8, |
| "learning_rate": 3.641414141414142e-06, |
| "loss": 0.7947, |
| "step": 639500 |
| }, |
| { |
| "epoch": 20.81, |
| "learning_rate": 3.6363636363636366e-06, |
| "loss": 0.7946, |
| "step": 640000 |
| }, |
| { |
| "epoch": 20.83, |
| "learning_rate": 3.6313131313131317e-06, |
| "loss": 0.795, |
| "step": 640500 |
| }, |
| { |
| "epoch": 20.85, |
| "learning_rate": 3.6262626262626264e-06, |
| "loss": 0.795, |
| "step": 641000 |
| }, |
| { |
| "epoch": 20.86, |
| "learning_rate": 3.6212121212121216e-06, |
| "loss": 0.7944, |
| "step": 641500 |
| }, |
| { |
| "epoch": 20.88, |
| "learning_rate": 3.6161616161616163e-06, |
| "loss": 0.7939, |
| "step": 642000 |
| }, |
| { |
| "epoch": 20.9, |
| "learning_rate": 3.6111111111111115e-06, |
| "loss": 0.7944, |
| "step": 642500 |
| }, |
| { |
| "epoch": 20.91, |
| "learning_rate": 3.606060606060606e-06, |
| "loss": 0.7939, |
| "step": 643000 |
| }, |
| { |
| "epoch": 20.93, |
| "learning_rate": 3.6010101010101013e-06, |
| "loss": 0.794, |
| "step": 643500 |
| }, |
| { |
| "epoch": 20.94, |
| "learning_rate": 3.595959595959596e-06, |
| "loss": 0.7935, |
| "step": 644000 |
| }, |
| { |
| "epoch": 20.96, |
| "learning_rate": 3.590909090909091e-06, |
| "loss": 0.7937, |
| "step": 644500 |
| }, |
| { |
| "epoch": 20.98, |
| "learning_rate": 3.585858585858586e-06, |
| "loss": 0.7936, |
| "step": 645000 |
| }, |
| { |
| "epoch": 20.99, |
| "learning_rate": 3.580808080808081e-06, |
| "loss": 0.793, |
| "step": 645500 |
| }, |
| { |
| "epoch": 21.01, |
| "learning_rate": 3.575757575757576e-06, |
| "loss": 0.7927, |
| "step": 646000 |
| }, |
| { |
| "epoch": 21.03, |
| "learning_rate": 3.5707070707070713e-06, |
| "loss": 0.7927, |
| "step": 646500 |
| }, |
| { |
| "epoch": 21.04, |
| "learning_rate": 3.565656565656566e-06, |
| "loss": 0.7928, |
| "step": 647000 |
| }, |
| { |
| "epoch": 21.06, |
| "learning_rate": 3.560606060606061e-06, |
| "loss": 0.7925, |
| "step": 647500 |
| }, |
| { |
| "epoch": 21.07, |
| "learning_rate": 3.555555555555556e-06, |
| "loss": 0.7924, |
| "step": 648000 |
| }, |
| { |
| "epoch": 21.09, |
| "learning_rate": 3.550505050505051e-06, |
| "loss": 0.7921, |
| "step": 648500 |
| }, |
| { |
| "epoch": 21.11, |
| "learning_rate": 3.5454545454545458e-06, |
| "loss": 0.7915, |
| "step": 649000 |
| }, |
| { |
| "epoch": 21.12, |
| "learning_rate": 3.540404040404041e-06, |
| "loss": 0.7922, |
| "step": 649500 |
| }, |
| { |
| "epoch": 21.14, |
| "learning_rate": 3.5353535353535356e-06, |
| "loss": 0.7916, |
| "step": 650000 |
| }, |
| { |
| "epoch": 21.16, |
| "learning_rate": 3.5303030303030304e-06, |
| "loss": 0.7914, |
| "step": 650500 |
| }, |
| { |
| "epoch": 21.17, |
| "learning_rate": 3.5252525252525255e-06, |
| "loss": 0.7914, |
| "step": 651000 |
| }, |
| { |
| "epoch": 21.19, |
| "learning_rate": 3.5202020202020202e-06, |
| "loss": 0.7912, |
| "step": 651500 |
| }, |
| { |
| "epoch": 21.2, |
| "learning_rate": 3.5151515151515154e-06, |
| "loss": 0.7904, |
| "step": 652000 |
| }, |
| { |
| "epoch": 21.22, |
| "learning_rate": 3.51010101010101e-06, |
| "loss": 0.7909, |
| "step": 652500 |
| }, |
| { |
| "epoch": 21.24, |
| "learning_rate": 3.5050505050505052e-06, |
| "loss": 0.7908, |
| "step": 653000 |
| }, |
| { |
| "epoch": 21.25, |
| "learning_rate": 3.5e-06, |
| "loss": 0.7907, |
| "step": 653500 |
| }, |
| { |
| "epoch": 21.27, |
| "learning_rate": 3.494949494949495e-06, |
| "loss": 0.7903, |
| "step": 654000 |
| }, |
| { |
| "epoch": 21.29, |
| "learning_rate": 3.4898989898989907e-06, |
| "loss": 0.7902, |
| "step": 654500 |
| }, |
| { |
| "epoch": 21.3, |
| "learning_rate": 3.4848484848484854e-06, |
| "loss": 0.7908, |
| "step": 655000 |
| }, |
| { |
| "epoch": 21.32, |
| "learning_rate": 3.47979797979798e-06, |
| "loss": 0.7897, |
| "step": 655500 |
| }, |
| { |
| "epoch": 21.33, |
| "learning_rate": 3.4747474747474752e-06, |
| "loss": 0.7897, |
| "step": 656000 |
| }, |
| { |
| "epoch": 21.35, |
| "learning_rate": 3.46969696969697e-06, |
| "loss": 0.7897, |
| "step": 656500 |
| }, |
| { |
| "epoch": 21.37, |
| "learning_rate": 3.464646464646465e-06, |
| "loss": 0.7891, |
| "step": 657000 |
| }, |
| { |
| "epoch": 21.38, |
| "learning_rate": 3.45959595959596e-06, |
| "loss": 0.7894, |
| "step": 657500 |
| }, |
| { |
| "epoch": 21.4, |
| "learning_rate": 3.454545454545455e-06, |
| "loss": 0.7893, |
| "step": 658000 |
| }, |
| { |
| "epoch": 21.42, |
| "learning_rate": 3.4494949494949497e-06, |
| "loss": 0.7893, |
| "step": 658500 |
| }, |
| { |
| "epoch": 21.43, |
| "learning_rate": 3.444444444444445e-06, |
| "loss": 0.789, |
| "step": 659000 |
| }, |
| { |
| "epoch": 21.45, |
| "learning_rate": 3.4393939393939395e-06, |
| "loss": 0.7888, |
| "step": 659500 |
| }, |
| { |
| "epoch": 21.46, |
| "learning_rate": 3.4343434343434347e-06, |
| "loss": 0.7885, |
| "step": 660000 |
| }, |
| { |
| "epoch": 21.48, |
| "learning_rate": 3.4292929292929294e-06, |
| "loss": 0.788, |
| "step": 660500 |
| }, |
| { |
| "epoch": 21.5, |
| "learning_rate": 3.4242424242424246e-06, |
| "loss": 0.7885, |
| "step": 661000 |
| }, |
| { |
| "epoch": 21.51, |
| "learning_rate": 3.4191919191919193e-06, |
| "loss": 0.7874, |
| "step": 661500 |
| }, |
| { |
| "epoch": 21.53, |
| "learning_rate": 3.414141414141414e-06, |
| "loss": 0.7878, |
| "step": 662000 |
| }, |
| { |
| "epoch": 21.55, |
| "learning_rate": 3.409090909090909e-06, |
| "loss": 0.7879, |
| "step": 662500 |
| }, |
| { |
| "epoch": 21.56, |
| "learning_rate": 3.4040404040404047e-06, |
| "loss": 0.7874, |
| "step": 663000 |
| }, |
| { |
| "epoch": 21.58, |
| "learning_rate": 3.3989898989898994e-06, |
| "loss": 0.7872, |
| "step": 663500 |
| }, |
| { |
| "epoch": 21.59, |
| "learning_rate": 3.3939393939393946e-06, |
| "loss": 0.7869, |
| "step": 664000 |
| }, |
| { |
| "epoch": 21.61, |
| "learning_rate": 3.3888888888888893e-06, |
| "loss": 0.787, |
| "step": 664500 |
| }, |
| { |
| "epoch": 21.63, |
| "learning_rate": 3.3838383838383844e-06, |
| "loss": 0.787, |
| "step": 665000 |
| }, |
| { |
| "epoch": 21.64, |
| "learning_rate": 3.378787878787879e-06, |
| "loss": 0.787, |
| "step": 665500 |
| }, |
| { |
| "epoch": 21.66, |
| "learning_rate": 3.3737373737373743e-06, |
| "loss": 0.7864, |
| "step": 666000 |
| }, |
| { |
| "epoch": 21.68, |
| "learning_rate": 3.368686868686869e-06, |
| "loss": 0.7861, |
| "step": 666500 |
| }, |
| { |
| "epoch": 21.69, |
| "learning_rate": 3.3636363636363637e-06, |
| "loss": 0.7861, |
| "step": 667000 |
| }, |
| { |
| "epoch": 21.71, |
| "learning_rate": 3.358585858585859e-06, |
| "loss": 0.7856, |
| "step": 667500 |
| }, |
| { |
| "epoch": 21.72, |
| "learning_rate": 3.3535353535353536e-06, |
| "loss": 0.7854, |
| "step": 668000 |
| }, |
| { |
| "epoch": 21.74, |
| "learning_rate": 3.3484848484848487e-06, |
| "loss": 0.786, |
| "step": 668500 |
| }, |
| { |
| "epoch": 21.76, |
| "learning_rate": 3.3434343434343435e-06, |
| "loss": 0.7851, |
| "step": 669000 |
| }, |
| { |
| "epoch": 21.77, |
| "learning_rate": 3.3383838383838386e-06, |
| "loss": 0.7858, |
| "step": 669500 |
| }, |
| { |
| "epoch": 21.79, |
| "learning_rate": 3.3333333333333333e-06, |
| "loss": 0.7851, |
| "step": 670000 |
| }, |
| { |
| "epoch": 21.81, |
| "learning_rate": 3.3282828282828285e-06, |
| "loss": 0.7853, |
| "step": 670500 |
| }, |
| { |
| "epoch": 21.82, |
| "learning_rate": 3.323232323232323e-06, |
| "loss": 0.7847, |
| "step": 671000 |
| }, |
| { |
| "epoch": 21.84, |
| "learning_rate": 3.3181818181818188e-06, |
| "loss": 0.7843, |
| "step": 671500 |
| }, |
| { |
| "epoch": 21.86, |
| "learning_rate": 3.3131313131313135e-06, |
| "loss": 0.7842, |
| "step": 672000 |
| }, |
| { |
| "epoch": 21.87, |
| "learning_rate": 3.3080808080808086e-06, |
| "loss": 0.7843, |
| "step": 672500 |
| }, |
| { |
| "epoch": 21.89, |
| "learning_rate": 3.3030303030303033e-06, |
| "loss": 0.7841, |
| "step": 673000 |
| }, |
| { |
| "epoch": 21.9, |
| "learning_rate": 3.2979797979797985e-06, |
| "loss": 0.7837, |
| "step": 673500 |
| }, |
| { |
| "epoch": 21.92, |
| "learning_rate": 3.292929292929293e-06, |
| "loss": 0.7839, |
| "step": 674000 |
| }, |
| { |
| "epoch": 21.94, |
| "learning_rate": 3.2878787878787883e-06, |
| "loss": 0.7832, |
| "step": 674500 |
| }, |
| { |
| "epoch": 21.95, |
| "learning_rate": 3.282828282828283e-06, |
| "loss": 0.7832, |
| "step": 675000 |
| }, |
| { |
| "epoch": 21.97, |
| "learning_rate": 3.277777777777778e-06, |
| "loss": 0.7829, |
| "step": 675500 |
| }, |
| { |
| "epoch": 21.99, |
| "learning_rate": 3.272727272727273e-06, |
| "loss": 0.7829, |
| "step": 676000 |
| }, |
| { |
| "epoch": 22.0, |
| "learning_rate": 3.267676767676768e-06, |
| "loss": 0.7821, |
| "step": 676500 |
| }, |
| { |
| "epoch": 22.02, |
| "learning_rate": 3.262626262626263e-06, |
| "loss": 0.7827, |
| "step": 677000 |
| }, |
| { |
| "epoch": 22.03, |
| "learning_rate": 3.257575757575758e-06, |
| "loss": 0.7831, |
| "step": 677500 |
| }, |
| { |
| "epoch": 22.05, |
| "learning_rate": 3.2525252525252527e-06, |
| "loss": 0.7816, |
| "step": 678000 |
| }, |
| { |
| "epoch": 22.07, |
| "learning_rate": 3.2474747474747474e-06, |
| "loss": 0.7816, |
| "step": 678500 |
| }, |
| { |
| "epoch": 22.08, |
| "learning_rate": 3.2424242424242425e-06, |
| "loss": 0.7818, |
| "step": 679000 |
| }, |
| { |
| "epoch": 22.1, |
| "learning_rate": 3.2373737373737372e-06, |
| "loss": 0.7822, |
| "step": 679500 |
| }, |
| { |
| "epoch": 22.12, |
| "learning_rate": 3.232323232323233e-06, |
| "loss": 0.7815, |
| "step": 680000 |
| }, |
| { |
| "epoch": 22.13, |
| "learning_rate": 3.227272727272728e-06, |
| "loss": 0.7815, |
| "step": 680500 |
| }, |
| { |
| "epoch": 22.15, |
| "learning_rate": 3.2222222222222227e-06, |
| "loss": 0.7808, |
| "step": 681000 |
| }, |
| { |
| "epoch": 22.16, |
| "learning_rate": 3.217171717171718e-06, |
| "loss": 0.781, |
| "step": 681500 |
| }, |
| { |
| "epoch": 22.18, |
| "learning_rate": 3.2121212121212125e-06, |
| "loss": 0.781, |
| "step": 682000 |
| }, |
| { |
| "epoch": 22.2, |
| "learning_rate": 3.2070707070707072e-06, |
| "loss": 0.7803, |
| "step": 682500 |
| }, |
| { |
| "epoch": 22.21, |
| "learning_rate": 3.2020202020202024e-06, |
| "loss": 0.7797, |
| "step": 683000 |
| }, |
| { |
| "epoch": 22.23, |
| "learning_rate": 3.196969696969697e-06, |
| "loss": 0.7802, |
| "step": 683500 |
| }, |
| { |
| "epoch": 22.25, |
| "learning_rate": 3.1919191919191923e-06, |
| "loss": 0.7798, |
| "step": 684000 |
| }, |
| { |
| "epoch": 22.26, |
| "learning_rate": 3.186868686868687e-06, |
| "loss": 0.7799, |
| "step": 684500 |
| }, |
| { |
| "epoch": 22.28, |
| "learning_rate": 3.181818181818182e-06, |
| "loss": 0.7797, |
| "step": 685000 |
| }, |
| { |
| "epoch": 22.29, |
| "learning_rate": 3.176767676767677e-06, |
| "loss": 0.7794, |
| "step": 685500 |
| }, |
| { |
| "epoch": 22.31, |
| "learning_rate": 3.171717171717172e-06, |
| "loss": 0.779, |
| "step": 686000 |
| }, |
| { |
| "epoch": 22.33, |
| "learning_rate": 3.1666666666666667e-06, |
| "loss": 0.779, |
| "step": 686500 |
| }, |
| { |
| "epoch": 22.34, |
| "learning_rate": 3.161616161616162e-06, |
| "loss": 0.7788, |
| "step": 687000 |
| }, |
| { |
| "epoch": 22.36, |
| "learning_rate": 3.1565656565656566e-06, |
| "loss": 0.7787, |
| "step": 687500 |
| }, |
| { |
| "epoch": 22.38, |
| "learning_rate": 3.1515151515151517e-06, |
| "loss": 0.7785, |
| "step": 688000 |
| }, |
| { |
| "epoch": 22.39, |
| "learning_rate": 3.146464646464647e-06, |
| "loss": 0.7787, |
| "step": 688500 |
| }, |
| { |
| "epoch": 22.41, |
| "learning_rate": 3.141414141414142e-06, |
| "loss": 0.7783, |
| "step": 689000 |
| }, |
| { |
| "epoch": 22.42, |
| "learning_rate": 3.1363636363636367e-06, |
| "loss": 0.7781, |
| "step": 689500 |
| }, |
| { |
| "epoch": 22.44, |
| "learning_rate": 3.131313131313132e-06, |
| "loss": 0.7777, |
| "step": 690000 |
| }, |
| { |
| "epoch": 22.46, |
| "learning_rate": 3.1262626262626266e-06, |
| "loss": 0.7778, |
| "step": 690500 |
| }, |
| { |
| "epoch": 22.47, |
| "learning_rate": 3.1212121212121217e-06, |
| "loss": 0.7771, |
| "step": 691000 |
| }, |
| { |
| "epoch": 22.49, |
| "learning_rate": 3.1161616161616164e-06, |
| "loss": 0.7774, |
| "step": 691500 |
| }, |
| { |
| "epoch": 22.51, |
| "learning_rate": 3.1111111111111116e-06, |
| "loss": 0.7769, |
| "step": 692000 |
| }, |
| { |
| "epoch": 22.52, |
| "learning_rate": 3.1060606060606063e-06, |
| "loss": 0.7769, |
| "step": 692500 |
| }, |
| { |
| "epoch": 22.54, |
| "learning_rate": 3.1010101010101014e-06, |
| "loss": 0.7762, |
| "step": 693000 |
| }, |
| { |
| "epoch": 22.55, |
| "learning_rate": 3.095959595959596e-06, |
| "loss": 0.7763, |
| "step": 693500 |
| }, |
| { |
| "epoch": 22.57, |
| "learning_rate": 3.090909090909091e-06, |
| "loss": 0.7759, |
| "step": 694000 |
| }, |
| { |
| "epoch": 22.59, |
| "learning_rate": 3.085858585858586e-06, |
| "loss": 0.7763, |
| "step": 694500 |
| }, |
| { |
| "epoch": 22.6, |
| "learning_rate": 3.0808080808080807e-06, |
| "loss": 0.7759, |
| "step": 695000 |
| }, |
| { |
| "epoch": 22.62, |
| "learning_rate": 3.075757575757576e-06, |
| "loss": 0.7756, |
| "step": 695500 |
| }, |
| { |
| "epoch": 22.64, |
| "learning_rate": 3.0707070707070706e-06, |
| "loss": 0.7751, |
| "step": 696000 |
| }, |
| { |
| "epoch": 22.65, |
| "learning_rate": 3.0656565656565658e-06, |
| "loss": 0.7751, |
| "step": 696500 |
| }, |
| { |
| "epoch": 22.67, |
| "learning_rate": 3.0606060606060605e-06, |
| "loss": 0.7749, |
| "step": 697000 |
| }, |
| { |
| "epoch": 22.68, |
| "learning_rate": 3.055555555555556e-06, |
| "loss": 0.7747, |
| "step": 697500 |
| }, |
| { |
| "epoch": 22.7, |
| "learning_rate": 3.0505050505050508e-06, |
| "loss": 0.7752, |
| "step": 698000 |
| }, |
| { |
| "epoch": 22.72, |
| "learning_rate": 3.045454545454546e-06, |
| "loss": 0.7747, |
| "step": 698500 |
| }, |
| { |
| "epoch": 22.73, |
| "learning_rate": 3.0404040404040406e-06, |
| "loss": 0.7744, |
| "step": 699000 |
| }, |
| { |
| "epoch": 22.75, |
| "learning_rate": 3.0353535353535358e-06, |
| "loss": 0.7746, |
| "step": 699500 |
| }, |
| { |
| "epoch": 22.77, |
| "learning_rate": 3.0303030303030305e-06, |
| "loss": 0.7746, |
| "step": 700000 |
| }, |
| { |
| "epoch": 22.78, |
| "learning_rate": 3.0252525252525256e-06, |
| "loss": 0.7737, |
| "step": 700500 |
| }, |
| { |
| "epoch": 22.8, |
| "learning_rate": 3.0202020202020203e-06, |
| "loss": 0.7739, |
| "step": 701000 |
| }, |
| { |
| "epoch": 22.81, |
| "learning_rate": 3.0151515151515155e-06, |
| "loss": 0.7731, |
| "step": 701500 |
| }, |
| { |
| "epoch": 22.83, |
| "learning_rate": 3.0101010101010102e-06, |
| "loss": 0.7734, |
| "step": 702000 |
| }, |
| { |
| "epoch": 22.85, |
| "learning_rate": 3.0050505050505054e-06, |
| "loss": 0.7733, |
| "step": 702500 |
| }, |
| { |
| "epoch": 22.86, |
| "learning_rate": 3e-06, |
| "loss": 0.7729, |
| "step": 703000 |
| }, |
| { |
| "epoch": 22.88, |
| "learning_rate": 2.9949494949494952e-06, |
| "loss": 0.7729, |
| "step": 703500 |
| }, |
| { |
| "epoch": 22.9, |
| "learning_rate": 2.98989898989899e-06, |
| "loss": 0.7729, |
| "step": 704000 |
| }, |
| { |
| "epoch": 22.91, |
| "learning_rate": 2.984848484848485e-06, |
| "loss": 0.7724, |
| "step": 704500 |
| }, |
| { |
| "epoch": 22.93, |
| "learning_rate": 2.97979797979798e-06, |
| "loss": 0.7724, |
| "step": 705000 |
| }, |
| { |
| "epoch": 22.94, |
| "learning_rate": 2.9747474747474745e-06, |
| "loss": 0.7727, |
| "step": 705500 |
| }, |
| { |
| "epoch": 22.96, |
| "learning_rate": 2.96969696969697e-06, |
| "loss": 0.7721, |
| "step": 706000 |
| }, |
| { |
| "epoch": 22.98, |
| "learning_rate": 2.9646464646464652e-06, |
| "loss": 0.7719, |
| "step": 706500 |
| }, |
| { |
| "epoch": 22.99, |
| "learning_rate": 2.95959595959596e-06, |
| "loss": 0.7724, |
| "step": 707000 |
| }, |
| { |
| "epoch": 23.01, |
| "learning_rate": 2.954545454545455e-06, |
| "loss": 0.7717, |
| "step": 707500 |
| }, |
| { |
| "epoch": 23.03, |
| "learning_rate": 2.94949494949495e-06, |
| "loss": 0.7714, |
| "step": 708000 |
| }, |
| { |
| "epoch": 23.04, |
| "learning_rate": 2.944444444444445e-06, |
| "loss": 0.7715, |
| "step": 708500 |
| }, |
| { |
| "epoch": 23.06, |
| "learning_rate": 2.9393939393939397e-06, |
| "loss": 0.7705, |
| "step": 709000 |
| }, |
| { |
| "epoch": 23.07, |
| "learning_rate": 2.9343434343434344e-06, |
| "loss": 0.7708, |
| "step": 709500 |
| }, |
| { |
| "epoch": 23.09, |
| "learning_rate": 2.9292929292929295e-06, |
| "loss": 0.7707, |
| "step": 710000 |
| }, |
| { |
| "epoch": 23.11, |
| "learning_rate": 2.9242424242424243e-06, |
| "loss": 0.7705, |
| "step": 710500 |
| }, |
| { |
| "epoch": 23.12, |
| "learning_rate": 2.9191919191919194e-06, |
| "loss": 0.77, |
| "step": 711000 |
| }, |
| { |
| "epoch": 23.14, |
| "learning_rate": 2.914141414141414e-06, |
| "loss": 0.7698, |
| "step": 711500 |
| }, |
| { |
| "epoch": 23.16, |
| "learning_rate": 2.9090909090909093e-06, |
| "loss": 0.7696, |
| "step": 712000 |
| }, |
| { |
| "epoch": 23.17, |
| "learning_rate": 2.904040404040404e-06, |
| "loss": 0.7698, |
| "step": 712500 |
| }, |
| { |
| "epoch": 23.19, |
| "learning_rate": 2.898989898989899e-06, |
| "loss": 0.7695, |
| "step": 713000 |
| }, |
| { |
| "epoch": 23.2, |
| "learning_rate": 2.893939393939394e-06, |
| "loss": 0.7695, |
| "step": 713500 |
| }, |
| { |
| "epoch": 23.22, |
| "learning_rate": 2.888888888888889e-06, |
| "loss": 0.7692, |
| "step": 714000 |
| }, |
| { |
| "epoch": 23.24, |
| "learning_rate": 2.883838383838384e-06, |
| "loss": 0.7686, |
| "step": 714500 |
| }, |
| { |
| "epoch": 23.25, |
| "learning_rate": 2.8787878787878793e-06, |
| "loss": 0.7681, |
| "step": 715000 |
| }, |
| { |
| "epoch": 23.27, |
| "learning_rate": 2.873737373737374e-06, |
| "loss": 0.7681, |
| "step": 715500 |
| }, |
| { |
| "epoch": 23.29, |
| "learning_rate": 2.868686868686869e-06, |
| "loss": 0.7677, |
| "step": 716000 |
| }, |
| { |
| "epoch": 23.3, |
| "learning_rate": 2.863636363636364e-06, |
| "loss": 0.768, |
| "step": 716500 |
| }, |
| { |
| "epoch": 23.32, |
| "learning_rate": 2.858585858585859e-06, |
| "loss": 0.7676, |
| "step": 717000 |
| }, |
| { |
| "epoch": 23.33, |
| "learning_rate": 2.8535353535353537e-06, |
| "loss": 0.7683, |
| "step": 717500 |
| }, |
| { |
| "epoch": 23.35, |
| "learning_rate": 2.848484848484849e-06, |
| "loss": 0.7679, |
| "step": 718000 |
| }, |
| { |
| "epoch": 23.37, |
| "learning_rate": 2.8434343434343436e-06, |
| "loss": 0.7675, |
| "step": 718500 |
| }, |
| { |
| "epoch": 23.38, |
| "learning_rate": 2.8383838383838387e-06, |
| "loss": 0.767, |
| "step": 719000 |
| }, |
| { |
| "epoch": 23.4, |
| "learning_rate": 2.8333333333333335e-06, |
| "loss": 0.7678, |
| "step": 719500 |
| }, |
| { |
| "epoch": 23.42, |
| "learning_rate": 2.8282828282828286e-06, |
| "loss": 0.7666, |
| "step": 720000 |
| }, |
| { |
| "epoch": 23.43, |
| "learning_rate": 2.8232323232323233e-06, |
| "loss": 0.7663, |
| "step": 720500 |
| }, |
| { |
| "epoch": 23.45, |
| "learning_rate": 2.818181818181818e-06, |
| "loss": 0.767, |
| "step": 721000 |
| }, |
| { |
| "epoch": 23.46, |
| "learning_rate": 2.813131313131313e-06, |
| "loss": 0.7664, |
| "step": 721500 |
| }, |
| { |
| "epoch": 23.48, |
| "learning_rate": 2.808080808080808e-06, |
| "loss": 0.7665, |
| "step": 722000 |
| }, |
| { |
| "epoch": 23.5, |
| "learning_rate": 2.803030303030303e-06, |
| "loss": 0.7664, |
| "step": 722500 |
| }, |
| { |
| "epoch": 23.51, |
| "learning_rate": 2.7979797979797986e-06, |
| "loss": 0.7659, |
| "step": 723000 |
| }, |
| { |
| "epoch": 23.53, |
| "learning_rate": 2.7929292929292933e-06, |
| "loss": 0.7659, |
| "step": 723500 |
| }, |
| { |
| "epoch": 23.55, |
| "learning_rate": 2.7878787878787885e-06, |
| "loss": 0.7656, |
| "step": 724000 |
| }, |
| { |
| "epoch": 23.56, |
| "learning_rate": 2.782828282828283e-06, |
| "loss": 0.7651, |
| "step": 724500 |
| }, |
| { |
| "epoch": 23.58, |
| "learning_rate": 2.7777777777777783e-06, |
| "loss": 0.7648, |
| "step": 725000 |
| }, |
| { |
| "epoch": 23.6, |
| "learning_rate": 2.772727272727273e-06, |
| "loss": 0.7655, |
| "step": 725500 |
| }, |
| { |
| "epoch": 23.61, |
| "learning_rate": 2.7676767676767678e-06, |
| "loss": 0.7651, |
| "step": 726000 |
| }, |
| { |
| "epoch": 23.63, |
| "learning_rate": 2.762626262626263e-06, |
| "loss": 0.7643, |
| "step": 726500 |
| }, |
| { |
| "epoch": 23.64, |
| "learning_rate": 2.7575757575757576e-06, |
| "loss": 0.7637, |
| "step": 727000 |
| }, |
| { |
| "epoch": 23.66, |
| "learning_rate": 2.7525252525252528e-06, |
| "loss": 0.7646, |
| "step": 727500 |
| }, |
| { |
| "epoch": 23.68, |
| "learning_rate": 2.7474747474747475e-06, |
| "loss": 0.7648, |
| "step": 728000 |
| }, |
| { |
| "epoch": 23.69, |
| "learning_rate": 2.7424242424242426e-06, |
| "loss": 0.7643, |
| "step": 728500 |
| }, |
| { |
| "epoch": 23.71, |
| "learning_rate": 2.7373737373737374e-06, |
| "loss": 0.7643, |
| "step": 729000 |
| }, |
| { |
| "epoch": 23.73, |
| "learning_rate": 2.7323232323232325e-06, |
| "loss": 0.7641, |
| "step": 729500 |
| }, |
| { |
| "epoch": 23.74, |
| "learning_rate": 2.7272727272727272e-06, |
| "loss": 0.7633, |
| "step": 730000 |
| }, |
| { |
| "epoch": 23.76, |
| "learning_rate": 2.7222222222222224e-06, |
| "loss": 0.764, |
| "step": 730500 |
| }, |
| { |
| "epoch": 23.77, |
| "learning_rate": 2.717171717171717e-06, |
| "loss": 0.7634, |
| "step": 731000 |
| }, |
| { |
| "epoch": 23.79, |
| "learning_rate": 2.7121212121212127e-06, |
| "loss": 0.7629, |
| "step": 731500 |
| }, |
| { |
| "epoch": 23.81, |
| "learning_rate": 2.7070707070707074e-06, |
| "loss": 0.7633, |
| "step": 732000 |
| }, |
| { |
| "epoch": 23.82, |
| "learning_rate": 2.7020202020202025e-06, |
| "loss": 0.7622, |
| "step": 732500 |
| }, |
| { |
| "epoch": 23.84, |
| "learning_rate": 2.6969696969696972e-06, |
| "loss": 0.7628, |
| "step": 733000 |
| }, |
| { |
| "epoch": 23.86, |
| "learning_rate": 2.6919191919191924e-06, |
| "loss": 0.762, |
| "step": 733500 |
| }, |
| { |
| "epoch": 23.87, |
| "learning_rate": 2.686868686868687e-06, |
| "loss": 0.7631, |
| "step": 734000 |
| }, |
| { |
| "epoch": 23.89, |
| "learning_rate": 2.6818181818181822e-06, |
| "loss": 0.7623, |
| "step": 734500 |
| }, |
| { |
| "epoch": 23.9, |
| "learning_rate": 2.676767676767677e-06, |
| "loss": 0.7622, |
| "step": 735000 |
| }, |
| { |
| "epoch": 23.92, |
| "learning_rate": 2.671717171717172e-06, |
| "loss": 0.7621, |
| "step": 735500 |
| }, |
| { |
| "epoch": 23.94, |
| "learning_rate": 2.666666666666667e-06, |
| "loss": 0.7625, |
| "step": 736000 |
| }, |
| { |
| "epoch": 23.95, |
| "learning_rate": 2.661616161616162e-06, |
| "loss": 0.7619, |
| "step": 736500 |
| }, |
| { |
| "epoch": 23.97, |
| "learning_rate": 2.6565656565656567e-06, |
| "loss": 0.761, |
| "step": 737000 |
| }, |
| { |
| "epoch": 23.99, |
| "learning_rate": 2.6515151515151514e-06, |
| "loss": 0.761, |
| "step": 737500 |
| }, |
| { |
| "epoch": 24.0, |
| "learning_rate": 2.6464646464646466e-06, |
| "loss": 0.7614, |
| "step": 738000 |
| }, |
| { |
| "epoch": 24.02, |
| "learning_rate": 2.6414141414141413e-06, |
| "loss": 0.7612, |
| "step": 738500 |
| }, |
| { |
| "epoch": 24.03, |
| "learning_rate": 2.6363636363636364e-06, |
| "loss": 0.7612, |
| "step": 739000 |
| }, |
| { |
| "epoch": 24.05, |
| "learning_rate": 2.631313131313131e-06, |
| "loss": 0.7606, |
| "step": 739500 |
| }, |
| { |
| "epoch": 24.07, |
| "learning_rate": 2.6262626262626267e-06, |
| "loss": 0.7606, |
| "step": 740000 |
| }, |
| { |
| "epoch": 24.08, |
| "learning_rate": 2.621212121212122e-06, |
| "loss": 0.7604, |
| "step": 740500 |
| }, |
| { |
| "epoch": 24.1, |
| "learning_rate": 2.6161616161616166e-06, |
| "loss": 0.7607, |
| "step": 741000 |
| }, |
| { |
| "epoch": 24.12, |
| "learning_rate": 2.6111111111111113e-06, |
| "loss": 0.76, |
| "step": 741500 |
| }, |
| { |
| "epoch": 24.13, |
| "learning_rate": 2.6060606060606064e-06, |
| "loss": 0.7601, |
| "step": 742000 |
| }, |
| { |
| "epoch": 24.15, |
| "learning_rate": 2.601010101010101e-06, |
| "loss": 0.7592, |
| "step": 742500 |
| }, |
| { |
| "epoch": 24.16, |
| "learning_rate": 2.5959595959595963e-06, |
| "loss": 0.7591, |
| "step": 743000 |
| }, |
| { |
| "epoch": 24.18, |
| "learning_rate": 2.590909090909091e-06, |
| "loss": 0.7597, |
| "step": 743500 |
| }, |
| { |
| "epoch": 24.2, |
| "learning_rate": 2.585858585858586e-06, |
| "loss": 0.7592, |
| "step": 744000 |
| }, |
| { |
| "epoch": 24.21, |
| "learning_rate": 2.580808080808081e-06, |
| "loss": 0.7587, |
| "step": 744500 |
| }, |
| { |
| "epoch": 24.23, |
| "learning_rate": 2.575757575757576e-06, |
| "loss": 0.7591, |
| "step": 745000 |
| }, |
| { |
| "epoch": 24.25, |
| "learning_rate": 2.5707070707070707e-06, |
| "loss": 0.7591, |
| "step": 745500 |
| }, |
| { |
| "epoch": 24.26, |
| "learning_rate": 2.565656565656566e-06, |
| "loss": 0.7586, |
| "step": 746000 |
| }, |
| { |
| "epoch": 24.28, |
| "learning_rate": 2.5606060606060606e-06, |
| "loss": 0.7583, |
| "step": 746500 |
| }, |
| { |
| "epoch": 24.29, |
| "learning_rate": 2.5555555555555557e-06, |
| "loss": 0.7586, |
| "step": 747000 |
| }, |
| { |
| "epoch": 24.31, |
| "learning_rate": 2.5505050505050505e-06, |
| "loss": 0.7583, |
| "step": 747500 |
| }, |
| { |
| "epoch": 24.33, |
| "learning_rate": 2.5454545454545456e-06, |
| "loss": 0.7584, |
| "step": 748000 |
| }, |
| { |
| "epoch": 24.34, |
| "learning_rate": 2.5404040404040408e-06, |
| "loss": 0.7578, |
| "step": 748500 |
| }, |
| { |
| "epoch": 24.36, |
| "learning_rate": 2.535353535353536e-06, |
| "loss": 0.7582, |
| "step": 749000 |
| }, |
| { |
| "epoch": 24.38, |
| "learning_rate": 2.5303030303030306e-06, |
| "loss": 0.7573, |
| "step": 749500 |
| }, |
| { |
| "epoch": 24.39, |
| "learning_rate": 2.5252525252525258e-06, |
| "loss": 0.7573, |
| "step": 750000 |
| }, |
| { |
| "epoch": 24.41, |
| "learning_rate": 2.5202020202020205e-06, |
| "loss": 0.7573, |
| "step": 750500 |
| }, |
| { |
| "epoch": 24.42, |
| "learning_rate": 2.5151515151515156e-06, |
| "loss": 0.7574, |
| "step": 751000 |
| }, |
| { |
| "epoch": 24.44, |
| "learning_rate": 2.5101010101010103e-06, |
| "loss": 0.7567, |
| "step": 751500 |
| }, |
| { |
| "epoch": 24.46, |
| "learning_rate": 2.5050505050505055e-06, |
| "loss": 0.7566, |
| "step": 752000 |
| }, |
| { |
| "epoch": 24.47, |
| "learning_rate": 2.5e-06, |
| "loss": 0.7567, |
| "step": 752500 |
| }, |
| { |
| "epoch": 24.49, |
| "learning_rate": 2.494949494949495e-06, |
| "loss": 0.7561, |
| "step": 753000 |
| }, |
| { |
| "epoch": 24.51, |
| "learning_rate": 2.48989898989899e-06, |
| "loss": 0.7563, |
| "step": 753500 |
| }, |
| { |
| "epoch": 24.52, |
| "learning_rate": 2.4848484848484848e-06, |
| "loss": 0.7557, |
| "step": 754000 |
| }, |
| { |
| "epoch": 24.54, |
| "learning_rate": 2.47979797979798e-06, |
| "loss": 0.7566, |
| "step": 754500 |
| }, |
| { |
| "epoch": 24.55, |
| "learning_rate": 2.474747474747475e-06, |
| "loss": 0.7557, |
| "step": 755000 |
| }, |
| { |
| "epoch": 24.57, |
| "learning_rate": 2.46969696969697e-06, |
| "loss": 0.7557, |
| "step": 755500 |
| }, |
| { |
| "epoch": 24.59, |
| "learning_rate": 2.464646464646465e-06, |
| "loss": 0.7554, |
| "step": 756000 |
| }, |
| { |
| "epoch": 24.6, |
| "learning_rate": 2.4595959595959597e-06, |
| "loss": 0.7552, |
| "step": 756500 |
| }, |
| { |
| "epoch": 24.62, |
| "learning_rate": 2.454545454545455e-06, |
| "loss": 0.7556, |
| "step": 757000 |
| }, |
| { |
| "epoch": 24.64, |
| "learning_rate": 2.4494949494949495e-06, |
| "loss": 0.7551, |
| "step": 757500 |
| }, |
| { |
| "epoch": 24.65, |
| "learning_rate": 2.4444444444444447e-06, |
| "loss": 0.7551, |
| "step": 758000 |
| }, |
| { |
| "epoch": 24.67, |
| "learning_rate": 2.4393939393939394e-06, |
| "loss": 0.755, |
| "step": 758500 |
| }, |
| { |
| "epoch": 24.68, |
| "learning_rate": 2.4343434343434345e-06, |
| "loss": 0.7551, |
| "step": 759000 |
| }, |
| { |
| "epoch": 24.7, |
| "learning_rate": 2.4292929292929297e-06, |
| "loss": 0.7543, |
| "step": 759500 |
| }, |
| { |
| "epoch": 24.72, |
| "learning_rate": 2.4242424242424244e-06, |
| "loss": 0.7547, |
| "step": 760000 |
| }, |
| { |
| "epoch": 24.73, |
| "learning_rate": 2.4191919191919195e-06, |
| "loss": 0.7545, |
| "step": 760500 |
| }, |
| { |
| "epoch": 24.75, |
| "learning_rate": 2.4141414141414143e-06, |
| "loss": 0.7544, |
| "step": 761000 |
| }, |
| { |
| "epoch": 24.77, |
| "learning_rate": 2.4090909090909094e-06, |
| "loss": 0.7542, |
| "step": 761500 |
| }, |
| { |
| "epoch": 24.78, |
| "learning_rate": 2.404040404040404e-06, |
| "loss": 0.7543, |
| "step": 762000 |
| }, |
| { |
| "epoch": 24.8, |
| "learning_rate": 2.3989898989898993e-06, |
| "loss": 0.754, |
| "step": 762500 |
| }, |
| { |
| "epoch": 24.81, |
| "learning_rate": 2.393939393939394e-06, |
| "loss": 0.754, |
| "step": 763000 |
| }, |
| { |
| "epoch": 24.83, |
| "learning_rate": 2.388888888888889e-06, |
| "loss": 0.7535, |
| "step": 763500 |
| }, |
| { |
| "epoch": 24.85, |
| "learning_rate": 2.3838383838383843e-06, |
| "loss": 0.7537, |
| "step": 764000 |
| }, |
| { |
| "epoch": 24.86, |
| "learning_rate": 2.378787878787879e-06, |
| "loss": 0.7539, |
| "step": 764500 |
| }, |
| { |
| "epoch": 24.88, |
| "learning_rate": 2.373737373737374e-06, |
| "loss": 0.7533, |
| "step": 765000 |
| }, |
| { |
| "epoch": 24.9, |
| "learning_rate": 2.368686868686869e-06, |
| "loss": 0.7532, |
| "step": 765500 |
| }, |
| { |
| "epoch": 24.91, |
| "learning_rate": 2.363636363636364e-06, |
| "loss": 0.7531, |
| "step": 766000 |
| }, |
| { |
| "epoch": 24.93, |
| "learning_rate": 2.3585858585858587e-06, |
| "loss": 0.7529, |
| "step": 766500 |
| }, |
| { |
| "epoch": 24.94, |
| "learning_rate": 2.3535353535353534e-06, |
| "loss": 0.7525, |
| "step": 767000 |
| }, |
| { |
| "epoch": 24.96, |
| "learning_rate": 2.348484848484849e-06, |
| "loss": 0.7524, |
| "step": 767500 |
| }, |
| { |
| "epoch": 24.98, |
| "learning_rate": 2.3434343434343437e-06, |
| "loss": 0.7523, |
| "step": 768000 |
| }, |
| { |
| "epoch": 24.99, |
| "learning_rate": 2.3383838383838384e-06, |
| "loss": 0.7522, |
| "step": 768500 |
| }, |
| { |
| "epoch": 25.01, |
| "learning_rate": 2.3333333333333336e-06, |
| "loss": 0.7521, |
| "step": 769000 |
| }, |
| { |
| "epoch": 25.03, |
| "learning_rate": 2.3282828282828283e-06, |
| "loss": 0.7519, |
| "step": 769500 |
| }, |
| { |
| "epoch": 25.04, |
| "learning_rate": 2.3232323232323234e-06, |
| "loss": 0.7519, |
| "step": 770000 |
| }, |
| { |
| "epoch": 25.06, |
| "learning_rate": 2.318181818181818e-06, |
| "loss": 0.7517, |
| "step": 770500 |
| }, |
| { |
| "epoch": 25.07, |
| "learning_rate": 2.3131313131313133e-06, |
| "loss": 0.7516, |
| "step": 771000 |
| }, |
| { |
| "epoch": 25.09, |
| "learning_rate": 2.308080808080808e-06, |
| "loss": 0.7515, |
| "step": 771500 |
| }, |
| { |
| "epoch": 25.11, |
| "learning_rate": 2.303030303030303e-06, |
| "loss": 0.7508, |
| "step": 772000 |
| }, |
| { |
| "epoch": 25.12, |
| "learning_rate": 2.2979797979797983e-06, |
| "loss": 0.7518, |
| "step": 772500 |
| }, |
| { |
| "epoch": 25.14, |
| "learning_rate": 2.292929292929293e-06, |
| "loss": 0.7511, |
| "step": 773000 |
| }, |
| { |
| "epoch": 25.16, |
| "learning_rate": 2.287878787878788e-06, |
| "loss": 0.751, |
| "step": 773500 |
| }, |
| { |
| "epoch": 25.17, |
| "learning_rate": 2.282828282828283e-06, |
| "loss": 0.7509, |
| "step": 774000 |
| }, |
| { |
| "epoch": 25.19, |
| "learning_rate": 2.277777777777778e-06, |
| "loss": 0.7506, |
| "step": 774500 |
| }, |
| { |
| "epoch": 25.2, |
| "learning_rate": 2.2727272727272728e-06, |
| "loss": 0.7505, |
| "step": 775000 |
| }, |
| { |
| "epoch": 25.22, |
| "learning_rate": 2.267676767676768e-06, |
| "loss": 0.7505, |
| "step": 775500 |
| }, |
| { |
| "epoch": 25.24, |
| "learning_rate": 2.262626262626263e-06, |
| "loss": 0.7498, |
| "step": 776000 |
| }, |
| { |
| "epoch": 25.25, |
| "learning_rate": 2.2575757575757578e-06, |
| "loss": 0.7503, |
| "step": 776500 |
| }, |
| { |
| "epoch": 25.27, |
| "learning_rate": 2.252525252525253e-06, |
| "loss": 0.7497, |
| "step": 777000 |
| }, |
| { |
| "epoch": 25.29, |
| "learning_rate": 2.2474747474747476e-06, |
| "loss": 0.7504, |
| "step": 777500 |
| }, |
| { |
| "epoch": 25.3, |
| "learning_rate": 2.2424242424242428e-06, |
| "loss": 0.7497, |
| "step": 778000 |
| }, |
| { |
| "epoch": 25.32, |
| "learning_rate": 2.2373737373737375e-06, |
| "loss": 0.7499, |
| "step": 778500 |
| }, |
| { |
| "epoch": 25.33, |
| "learning_rate": 2.2323232323232326e-06, |
| "loss": 0.749, |
| "step": 779000 |
| }, |
| { |
| "epoch": 25.35, |
| "learning_rate": 2.2272727272727274e-06, |
| "loss": 0.7496, |
| "step": 779500 |
| }, |
| { |
| "epoch": 25.37, |
| "learning_rate": 2.222222222222222e-06, |
| "loss": 0.7496, |
| "step": 780000 |
| }, |
| { |
| "epoch": 25.38, |
| "learning_rate": 2.2171717171717176e-06, |
| "loss": 0.7491, |
| "step": 780500 |
| }, |
| { |
| "epoch": 25.4, |
| "learning_rate": 2.2121212121212124e-06, |
| "loss": 0.7491, |
| "step": 781000 |
| }, |
| { |
| "epoch": 25.42, |
| "learning_rate": 2.2070707070707075e-06, |
| "loss": 0.7494, |
| "step": 781500 |
| }, |
| { |
| "epoch": 25.43, |
| "learning_rate": 2.2020202020202022e-06, |
| "loss": 0.7489, |
| "step": 782000 |
| }, |
| { |
| "epoch": 25.45, |
| "learning_rate": 2.196969696969697e-06, |
| "loss": 0.7487, |
| "step": 782500 |
| }, |
| { |
| "epoch": 25.47, |
| "learning_rate": 2.191919191919192e-06, |
| "loss": 0.7484, |
| "step": 783000 |
| }, |
| { |
| "epoch": 25.48, |
| "learning_rate": 2.186868686868687e-06, |
| "loss": 0.7488, |
| "step": 783500 |
| }, |
| { |
| "epoch": 25.5, |
| "learning_rate": 2.181818181818182e-06, |
| "loss": 0.7488, |
| "step": 784000 |
| }, |
| { |
| "epoch": 25.51, |
| "learning_rate": 2.1767676767676767e-06, |
| "loss": 0.7486, |
| "step": 784500 |
| }, |
| { |
| "epoch": 25.53, |
| "learning_rate": 2.171717171717172e-06, |
| "loss": 0.748, |
| "step": 785000 |
| }, |
| { |
| "epoch": 25.55, |
| "learning_rate": 2.166666666666667e-06, |
| "loss": 0.7481, |
| "step": 785500 |
| }, |
| { |
| "epoch": 25.56, |
| "learning_rate": 2.1616161616161617e-06, |
| "loss": 0.7478, |
| "step": 786000 |
| }, |
| { |
| "epoch": 25.58, |
| "learning_rate": 2.156565656565657e-06, |
| "loss": 0.7484, |
| "step": 786500 |
| }, |
| { |
| "epoch": 25.6, |
| "learning_rate": 2.1515151515151515e-06, |
| "loss": 0.7477, |
| "step": 787000 |
| }, |
| { |
| "epoch": 25.61, |
| "learning_rate": 2.1464646464646467e-06, |
| "loss": 0.7474, |
| "step": 787500 |
| }, |
| { |
| "epoch": 25.63, |
| "learning_rate": 2.1414141414141414e-06, |
| "loss": 0.7472, |
| "step": 788000 |
| }, |
| { |
| "epoch": 25.64, |
| "learning_rate": 2.1363636363636365e-06, |
| "loss": 0.7473, |
| "step": 788500 |
| }, |
| { |
| "epoch": 25.66, |
| "learning_rate": 2.1313131313131317e-06, |
| "loss": 0.7471, |
| "step": 789000 |
| }, |
| { |
| "epoch": 25.68, |
| "learning_rate": 2.1262626262626264e-06, |
| "loss": 0.7471, |
| "step": 789500 |
| }, |
| { |
| "epoch": 25.69, |
| "learning_rate": 2.1212121212121216e-06, |
| "loss": 0.7471, |
| "step": 790000 |
| }, |
| { |
| "epoch": 25.71, |
| "learning_rate": 2.1161616161616163e-06, |
| "loss": 0.7471, |
| "step": 790500 |
| }, |
| { |
| "epoch": 25.73, |
| "learning_rate": 2.1111111111111114e-06, |
| "loss": 0.7467, |
| "step": 791000 |
| }, |
| { |
| "epoch": 25.74, |
| "learning_rate": 2.106060606060606e-06, |
| "loss": 0.7469, |
| "step": 791500 |
| }, |
| { |
| "epoch": 25.76, |
| "learning_rate": 2.1010101010101013e-06, |
| "loss": 0.746, |
| "step": 792000 |
| }, |
| { |
| "epoch": 25.77, |
| "learning_rate": 2.095959595959596e-06, |
| "loss": 0.746, |
| "step": 792500 |
| }, |
| { |
| "epoch": 25.79, |
| "learning_rate": 2.090909090909091e-06, |
| "loss": 0.7464, |
| "step": 793000 |
| }, |
| { |
| "epoch": 25.81, |
| "learning_rate": 2.0858585858585863e-06, |
| "loss": 0.7459, |
| "step": 793500 |
| }, |
| { |
| "epoch": 25.82, |
| "learning_rate": 2.080808080808081e-06, |
| "loss": 0.746, |
| "step": 794000 |
| }, |
| { |
| "epoch": 25.84, |
| "learning_rate": 2.075757575757576e-06, |
| "loss": 0.7459, |
| "step": 794500 |
| }, |
| { |
| "epoch": 25.86, |
| "learning_rate": 2.070707070707071e-06, |
| "loss": 0.7458, |
| "step": 795000 |
| }, |
| { |
| "epoch": 25.87, |
| "learning_rate": 2.065656565656566e-06, |
| "loss": 0.745, |
| "step": 795500 |
| }, |
| { |
| "epoch": 25.89, |
| "learning_rate": 2.0606060606060607e-06, |
| "loss": 0.7456, |
| "step": 796000 |
| }, |
| { |
| "epoch": 25.9, |
| "learning_rate": 2.0555555555555555e-06, |
| "loss": 0.7457, |
| "step": 796500 |
| }, |
| { |
| "epoch": 25.92, |
| "learning_rate": 2.0505050505050506e-06, |
| "loss": 0.7448, |
| "step": 797000 |
| }, |
| { |
| "epoch": 25.94, |
| "learning_rate": 2.0454545454545457e-06, |
| "loss": 0.7452, |
| "step": 797500 |
| }, |
| { |
| "epoch": 25.95, |
| "learning_rate": 2.0404040404040405e-06, |
| "loss": 0.7459, |
| "step": 798000 |
| }, |
| { |
| "epoch": 25.97, |
| "learning_rate": 2.0353535353535356e-06, |
| "loss": 0.7447, |
| "step": 798500 |
| }, |
| { |
| "epoch": 25.99, |
| "learning_rate": 2.0303030303030303e-06, |
| "loss": 0.7451, |
| "step": 799000 |
| }, |
| { |
| "epoch": 26.0, |
| "learning_rate": 2.0252525252525255e-06, |
| "loss": 0.7446, |
| "step": 799500 |
| }, |
| { |
| "epoch": 26.02, |
| "learning_rate": 2.02020202020202e-06, |
| "loss": 0.7443, |
| "step": 800000 |
| }, |
| { |
| "epoch": 26.03, |
| "learning_rate": 2.0151515151515153e-06, |
| "loss": 0.7446, |
| "step": 800500 |
| }, |
| { |
| "epoch": 26.05, |
| "learning_rate": 2.01010101010101e-06, |
| "loss": 0.7445, |
| "step": 801000 |
| }, |
| { |
| "epoch": 26.07, |
| "learning_rate": 2.005050505050505e-06, |
| "loss": 0.7444, |
| "step": 801500 |
| }, |
| { |
| "epoch": 26.08, |
| "learning_rate": 2.0000000000000003e-06, |
| "loss": 0.7443, |
| "step": 802000 |
| }, |
| { |
| "epoch": 26.1, |
| "learning_rate": 1.994949494949495e-06, |
| "loss": 0.744, |
| "step": 802500 |
| }, |
| { |
| "epoch": 26.12, |
| "learning_rate": 1.98989898989899e-06, |
| "loss": 0.7439, |
| "step": 803000 |
| }, |
| { |
| "epoch": 26.13, |
| "learning_rate": 1.984848484848485e-06, |
| "loss": 0.7432, |
| "step": 803500 |
| }, |
| { |
| "epoch": 26.15, |
| "learning_rate": 1.97979797979798e-06, |
| "loss": 0.7442, |
| "step": 804000 |
| }, |
| { |
| "epoch": 26.16, |
| "learning_rate": 1.9747474747474748e-06, |
| "loss": 0.7434, |
| "step": 804500 |
| }, |
| { |
| "epoch": 26.18, |
| "learning_rate": 1.96969696969697e-06, |
| "loss": 0.7433, |
| "step": 805000 |
| }, |
| { |
| "epoch": 26.2, |
| "learning_rate": 1.9646464646464646e-06, |
| "loss": 0.7433, |
| "step": 805500 |
| }, |
| { |
| "epoch": 26.21, |
| "learning_rate": 1.9595959595959598e-06, |
| "loss": 0.7432, |
| "step": 806000 |
| }, |
| { |
| "epoch": 26.23, |
| "learning_rate": 1.954545454545455e-06, |
| "loss": 0.7428, |
| "step": 806500 |
| }, |
| { |
| "epoch": 26.25, |
| "learning_rate": 1.9494949494949496e-06, |
| "loss": 0.7429, |
| "step": 807000 |
| }, |
| { |
| "epoch": 26.26, |
| "learning_rate": 1.944444444444445e-06, |
| "loss": 0.7432, |
| "step": 807500 |
| }, |
| { |
| "epoch": 26.28, |
| "learning_rate": 1.9393939393939395e-06, |
| "loss": 0.7434, |
| "step": 808000 |
| }, |
| { |
| "epoch": 26.29, |
| "learning_rate": 1.9343434343434347e-06, |
| "loss": 0.7423, |
| "step": 808500 |
| }, |
| { |
| "epoch": 26.31, |
| "learning_rate": 1.9292929292929294e-06, |
| "loss": 0.743, |
| "step": 809000 |
| }, |
| { |
| "epoch": 26.33, |
| "learning_rate": 1.924242424242424e-06, |
| "loss": 0.7426, |
| "step": 809500 |
| }, |
| { |
| "epoch": 26.34, |
| "learning_rate": 1.9191919191919192e-06, |
| "loss": 0.7424, |
| "step": 810000 |
| }, |
| { |
| "epoch": 26.36, |
| "learning_rate": 1.9141414141414144e-06, |
| "loss": 0.7424, |
| "step": 810500 |
| }, |
| { |
| "epoch": 26.38, |
| "learning_rate": 1.9090909090909095e-06, |
| "loss": 0.7422, |
| "step": 811000 |
| }, |
| { |
| "epoch": 26.39, |
| "learning_rate": 1.9040404040404042e-06, |
| "loss": 0.7417, |
| "step": 811500 |
| }, |
| { |
| "epoch": 26.41, |
| "learning_rate": 1.8989898989898992e-06, |
| "loss": 0.7419, |
| "step": 812000 |
| }, |
| { |
| "epoch": 26.42, |
| "learning_rate": 1.8939393939393941e-06, |
| "loss": 0.7422, |
| "step": 812500 |
| }, |
| { |
| "epoch": 26.44, |
| "learning_rate": 1.888888888888889e-06, |
| "loss": 0.742, |
| "step": 813000 |
| }, |
| { |
| "epoch": 26.46, |
| "learning_rate": 1.883838383838384e-06, |
| "loss": 0.7416, |
| "step": 813500 |
| }, |
| { |
| "epoch": 26.47, |
| "learning_rate": 1.878787878787879e-06, |
| "loss": 0.7416, |
| "step": 814000 |
| }, |
| { |
| "epoch": 26.49, |
| "learning_rate": 1.873737373737374e-06, |
| "loss": 0.7417, |
| "step": 814500 |
| }, |
| { |
| "epoch": 26.51, |
| "learning_rate": 1.868686868686869e-06, |
| "loss": 0.7412, |
| "step": 815000 |
| }, |
| { |
| "epoch": 26.52, |
| "learning_rate": 1.863636363636364e-06, |
| "loss": 0.7417, |
| "step": 815500 |
| }, |
| { |
| "epoch": 26.54, |
| "learning_rate": 1.8585858585858588e-06, |
| "loss": 0.7409, |
| "step": 816000 |
| }, |
| { |
| "epoch": 26.55, |
| "learning_rate": 1.8535353535353538e-06, |
| "loss": 0.7412, |
| "step": 816500 |
| }, |
| { |
| "epoch": 26.57, |
| "learning_rate": 1.8484848484848487e-06, |
| "loss": 0.7414, |
| "step": 817000 |
| }, |
| { |
| "epoch": 26.59, |
| "learning_rate": 1.8434343434343434e-06, |
| "loss": 0.7409, |
| "step": 817500 |
| }, |
| { |
| "epoch": 26.6, |
| "learning_rate": 1.8383838383838384e-06, |
| "loss": 0.7404, |
| "step": 818000 |
| }, |
| { |
| "epoch": 26.62, |
| "learning_rate": 1.8333333333333333e-06, |
| "loss": 0.7405, |
| "step": 818500 |
| }, |
| { |
| "epoch": 26.64, |
| "learning_rate": 1.8282828282828286e-06, |
| "loss": 0.7403, |
| "step": 819000 |
| }, |
| { |
| "epoch": 26.65, |
| "learning_rate": 1.8232323232323234e-06, |
| "loss": 0.7404, |
| "step": 819500 |
| }, |
| { |
| "epoch": 26.67, |
| "learning_rate": 1.8181818181818183e-06, |
| "loss": 0.74, |
| "step": 820000 |
| }, |
| { |
| "epoch": 26.68, |
| "learning_rate": 1.8131313131313132e-06, |
| "loss": 0.74, |
| "step": 820500 |
| }, |
| { |
| "epoch": 26.7, |
| "learning_rate": 1.8080808080808082e-06, |
| "loss": 0.7399, |
| "step": 821000 |
| }, |
| { |
| "epoch": 26.72, |
| "learning_rate": 1.803030303030303e-06, |
| "loss": 0.74, |
| "step": 821500 |
| }, |
| { |
| "epoch": 26.73, |
| "learning_rate": 1.797979797979798e-06, |
| "loss": 0.7403, |
| "step": 822000 |
| }, |
| { |
| "epoch": 26.75, |
| "learning_rate": 1.792929292929293e-06, |
| "loss": 0.7403, |
| "step": 822500 |
| }, |
| { |
| "epoch": 26.77, |
| "learning_rate": 1.787878787878788e-06, |
| "loss": 0.7401, |
| "step": 823000 |
| }, |
| { |
| "epoch": 26.78, |
| "learning_rate": 1.782828282828283e-06, |
| "loss": 0.7394, |
| "step": 823500 |
| }, |
| { |
| "epoch": 26.8, |
| "learning_rate": 1.777777777777778e-06, |
| "loss": 0.7397, |
| "step": 824000 |
| }, |
| { |
| "epoch": 26.81, |
| "learning_rate": 1.7727272727272729e-06, |
| "loss": 0.7399, |
| "step": 824500 |
| }, |
| { |
| "epoch": 26.83, |
| "learning_rate": 1.7676767676767678e-06, |
| "loss": 0.7395, |
| "step": 825000 |
| }, |
| { |
| "epoch": 26.85, |
| "learning_rate": 1.7626262626262628e-06, |
| "loss": 0.7392, |
| "step": 825500 |
| }, |
| { |
| "epoch": 26.86, |
| "learning_rate": 1.7575757575757577e-06, |
| "loss": 0.739, |
| "step": 826000 |
| }, |
| { |
| "epoch": 26.88, |
| "learning_rate": 1.7525252525252526e-06, |
| "loss": 0.7388, |
| "step": 826500 |
| }, |
| { |
| "epoch": 26.9, |
| "learning_rate": 1.7474747474747475e-06, |
| "loss": 0.739, |
| "step": 827000 |
| }, |
| { |
| "epoch": 26.91, |
| "learning_rate": 1.7424242424242427e-06, |
| "loss": 0.739, |
| "step": 827500 |
| }, |
| { |
| "epoch": 26.93, |
| "learning_rate": 1.7373737373737376e-06, |
| "loss": 0.7385, |
| "step": 828000 |
| }, |
| { |
| "epoch": 26.94, |
| "learning_rate": 1.7323232323232326e-06, |
| "loss": 0.7385, |
| "step": 828500 |
| }, |
| { |
| "epoch": 26.96, |
| "learning_rate": 1.7272727272727275e-06, |
| "loss": 0.7386, |
| "step": 829000 |
| }, |
| { |
| "epoch": 26.98, |
| "learning_rate": 1.7222222222222224e-06, |
| "loss": 0.7383, |
| "step": 829500 |
| }, |
| { |
| "epoch": 26.99, |
| "learning_rate": 1.7171717171717173e-06, |
| "loss": 0.7378, |
| "step": 830000 |
| }, |
| { |
| "epoch": 27.01, |
| "learning_rate": 1.7121212121212123e-06, |
| "loss": 0.7385, |
| "step": 830500 |
| }, |
| { |
| "epoch": 27.03, |
| "learning_rate": 1.707070707070707e-06, |
| "loss": 0.738, |
| "step": 831000 |
| }, |
| { |
| "epoch": 27.04, |
| "learning_rate": 1.7020202020202024e-06, |
| "loss": 0.7381, |
| "step": 831500 |
| }, |
| { |
| "epoch": 27.06, |
| "learning_rate": 1.6969696969696973e-06, |
| "loss": 0.7384, |
| "step": 832000 |
| }, |
| { |
| "epoch": 27.07, |
| "learning_rate": 1.6919191919191922e-06, |
| "loss": 0.7379, |
| "step": 832500 |
| }, |
| { |
| "epoch": 27.09, |
| "learning_rate": 1.6868686868686871e-06, |
| "loss": 0.7377, |
| "step": 833000 |
| }, |
| { |
| "epoch": 27.11, |
| "learning_rate": 1.6818181818181819e-06, |
| "loss": 0.7379, |
| "step": 833500 |
| }, |
| { |
| "epoch": 27.12, |
| "learning_rate": 1.6767676767676768e-06, |
| "loss": 0.7377, |
| "step": 834000 |
| }, |
| { |
| "epoch": 27.14, |
| "learning_rate": 1.6717171717171717e-06, |
| "loss": 0.7375, |
| "step": 834500 |
| }, |
| { |
| "epoch": 27.16, |
| "learning_rate": 1.6666666666666667e-06, |
| "loss": 0.7375, |
| "step": 835000 |
| }, |
| { |
| "epoch": 27.17, |
| "learning_rate": 1.6616161616161616e-06, |
| "loss": 0.7369, |
| "step": 835500 |
| }, |
| { |
| "epoch": 27.19, |
| "learning_rate": 1.6565656565656567e-06, |
| "loss": 0.738, |
| "step": 836000 |
| }, |
| { |
| "epoch": 27.21, |
| "learning_rate": 1.6515151515151517e-06, |
| "loss": 0.7377, |
| "step": 836500 |
| }, |
| { |
| "epoch": 27.22, |
| "learning_rate": 1.6464646464646466e-06, |
| "loss": 0.7368, |
| "step": 837000 |
| }, |
| { |
| "epoch": 27.24, |
| "learning_rate": 1.6414141414141415e-06, |
| "loss": 0.7369, |
| "step": 837500 |
| }, |
| { |
| "epoch": 27.25, |
| "learning_rate": 1.6363636363636365e-06, |
| "loss": 0.7372, |
| "step": 838000 |
| }, |
| { |
| "epoch": 27.27, |
| "learning_rate": 1.6313131313131314e-06, |
| "loss": 0.7366, |
| "step": 838500 |
| }, |
| { |
| "epoch": 27.29, |
| "learning_rate": 1.6262626262626263e-06, |
| "loss": 0.7366, |
| "step": 839000 |
| }, |
| { |
| "epoch": 27.3, |
| "learning_rate": 1.6212121212121213e-06, |
| "loss": 0.7363, |
| "step": 839500 |
| }, |
| { |
| "epoch": 27.32, |
| "learning_rate": 1.6161616161616164e-06, |
| "loss": 0.7364, |
| "step": 840000 |
| }, |
| { |
| "epoch": 27.34, |
| "learning_rate": 1.6111111111111113e-06, |
| "loss": 0.7364, |
| "step": 840500 |
| }, |
| { |
| "epoch": 27.35, |
| "learning_rate": 1.6060606060606063e-06, |
| "loss": 0.7365, |
| "step": 841000 |
| }, |
| { |
| "epoch": 27.37, |
| "learning_rate": 1.6010101010101012e-06, |
| "loss": 0.7367, |
| "step": 841500 |
| }, |
| { |
| "epoch": 27.38, |
| "learning_rate": 1.5959595959595961e-06, |
| "loss": 0.7358, |
| "step": 842000 |
| }, |
| { |
| "epoch": 27.4, |
| "learning_rate": 1.590909090909091e-06, |
| "loss": 0.736, |
| "step": 842500 |
| }, |
| { |
| "epoch": 27.42, |
| "learning_rate": 1.585858585858586e-06, |
| "loss": 0.7366, |
| "step": 843000 |
| }, |
| { |
| "epoch": 27.43, |
| "learning_rate": 1.580808080808081e-06, |
| "loss": 0.7358, |
| "step": 843500 |
| }, |
| { |
| "epoch": 27.45, |
| "learning_rate": 1.5757575757575759e-06, |
| "loss": 0.7359, |
| "step": 844000 |
| }, |
| { |
| "epoch": 27.47, |
| "learning_rate": 1.570707070707071e-06, |
| "loss": 0.7362, |
| "step": 844500 |
| }, |
| { |
| "epoch": 27.48, |
| "learning_rate": 1.565656565656566e-06, |
| "loss": 0.7359, |
| "step": 845000 |
| }, |
| { |
| "epoch": 27.5, |
| "learning_rate": 1.5606060606060609e-06, |
| "loss": 0.7352, |
| "step": 845500 |
| }, |
| { |
| "epoch": 27.51, |
| "learning_rate": 1.5555555555555558e-06, |
| "loss": 0.7354, |
| "step": 846000 |
| }, |
| { |
| "epoch": 27.53, |
| "learning_rate": 1.5505050505050507e-06, |
| "loss": 0.7358, |
| "step": 846500 |
| }, |
| { |
| "epoch": 27.55, |
| "learning_rate": 1.5454545454545454e-06, |
| "loss": 0.7356, |
| "step": 847000 |
| }, |
| { |
| "epoch": 27.56, |
| "learning_rate": 1.5404040404040404e-06, |
| "loss": 0.7353, |
| "step": 847500 |
| }, |
| { |
| "epoch": 27.58, |
| "learning_rate": 1.5353535353535353e-06, |
| "loss": 0.7354, |
| "step": 848000 |
| }, |
| { |
| "epoch": 27.6, |
| "learning_rate": 1.5303030303030302e-06, |
| "loss": 0.7354, |
| "step": 848500 |
| }, |
| { |
| "epoch": 27.61, |
| "learning_rate": 1.5252525252525254e-06, |
| "loss": 0.7348, |
| "step": 849000 |
| }, |
| { |
| "epoch": 27.63, |
| "learning_rate": 1.5202020202020203e-06, |
| "loss": 0.7354, |
| "step": 849500 |
| }, |
| { |
| "epoch": 27.64, |
| "learning_rate": 1.5151515151515152e-06, |
| "loss": 0.7349, |
| "step": 850000 |
| }, |
| { |
| "epoch": 27.66, |
| "learning_rate": 1.5101010101010102e-06, |
| "loss": 0.7347, |
| "step": 850500 |
| }, |
| { |
| "epoch": 27.68, |
| "learning_rate": 1.5050505050505051e-06, |
| "loss": 0.7348, |
| "step": 851000 |
| }, |
| { |
| "epoch": 27.69, |
| "learning_rate": 1.5e-06, |
| "loss": 0.7346, |
| "step": 851500 |
| }, |
| { |
| "epoch": 27.71, |
| "learning_rate": 1.494949494949495e-06, |
| "loss": 0.7346, |
| "step": 852000 |
| }, |
| { |
| "epoch": 27.73, |
| "learning_rate": 1.48989898989899e-06, |
| "loss": 0.7343, |
| "step": 852500 |
| }, |
| { |
| "epoch": 27.74, |
| "learning_rate": 1.484848484848485e-06, |
| "loss": 0.7347, |
| "step": 853000 |
| }, |
| { |
| "epoch": 27.76, |
| "learning_rate": 1.47979797979798e-06, |
| "loss": 0.7347, |
| "step": 853500 |
| }, |
| { |
| "epoch": 27.77, |
| "learning_rate": 1.474747474747475e-06, |
| "loss": 0.7347, |
| "step": 854000 |
| }, |
| { |
| "epoch": 27.79, |
| "learning_rate": 1.4696969696969698e-06, |
| "loss": 0.7343, |
| "step": 854500 |
| }, |
| { |
| "epoch": 27.81, |
| "learning_rate": 1.4646464646464648e-06, |
| "loss": 0.7344, |
| "step": 855000 |
| }, |
| { |
| "epoch": 27.82, |
| "learning_rate": 1.4595959595959597e-06, |
| "loss": 0.7338, |
| "step": 855500 |
| }, |
| { |
| "epoch": 27.84, |
| "learning_rate": 1.4545454545454546e-06, |
| "loss": 0.7339, |
| "step": 856000 |
| }, |
| { |
| "epoch": 27.86, |
| "learning_rate": 1.4494949494949496e-06, |
| "loss": 0.7336, |
| "step": 856500 |
| }, |
| { |
| "epoch": 27.87, |
| "learning_rate": 1.4444444444444445e-06, |
| "loss": 0.734, |
| "step": 857000 |
| }, |
| { |
| "epoch": 27.89, |
| "learning_rate": 1.4393939393939396e-06, |
| "loss": 0.7339, |
| "step": 857500 |
| }, |
| { |
| "epoch": 27.9, |
| "learning_rate": 1.4343434343434346e-06, |
| "loss": 0.7335, |
| "step": 858000 |
| }, |
| { |
| "epoch": 27.92, |
| "learning_rate": 1.4292929292929295e-06, |
| "loss": 0.7337, |
| "step": 858500 |
| }, |
| { |
| "epoch": 27.94, |
| "learning_rate": 1.4242424242424244e-06, |
| "loss": 0.7334, |
| "step": 859000 |
| }, |
| { |
| "epoch": 27.95, |
| "learning_rate": 1.4191919191919194e-06, |
| "loss": 0.7335, |
| "step": 859500 |
| }, |
| { |
| "epoch": 27.97, |
| "learning_rate": 1.4141414141414143e-06, |
| "loss": 0.7335, |
| "step": 860000 |
| }, |
| { |
| "epoch": 27.99, |
| "learning_rate": 1.409090909090909e-06, |
| "loss": 0.7331, |
| "step": 860500 |
| }, |
| { |
| "epoch": 28.0, |
| "learning_rate": 1.404040404040404e-06, |
| "loss": 0.7332, |
| "step": 861000 |
| }, |
| { |
| "epoch": 28.02, |
| "learning_rate": 1.3989898989898993e-06, |
| "loss": 0.7329, |
| "step": 861500 |
| }, |
| { |
| "epoch": 28.03, |
| "learning_rate": 1.3939393939393942e-06, |
| "loss": 0.733, |
| "step": 862000 |
| }, |
| { |
| "epoch": 28.05, |
| "learning_rate": 1.3888888888888892e-06, |
| "loss": 0.7332, |
| "step": 862500 |
| }, |
| { |
| "epoch": 28.07, |
| "learning_rate": 1.3838383838383839e-06, |
| "loss": 0.7336, |
| "step": 863000 |
| }, |
| { |
| "epoch": 28.08, |
| "learning_rate": 1.3787878787878788e-06, |
| "loss": 0.7325, |
| "step": 863500 |
| }, |
| { |
| "epoch": 28.1, |
| "learning_rate": 1.3737373737373738e-06, |
| "loss": 0.7331, |
| "step": 864000 |
| }, |
| { |
| "epoch": 28.12, |
| "learning_rate": 1.3686868686868687e-06, |
| "loss": 0.7329, |
| "step": 864500 |
| }, |
| { |
| "epoch": 28.13, |
| "learning_rate": 1.3636363636363636e-06, |
| "loss": 0.733, |
| "step": 865000 |
| }, |
| { |
| "epoch": 28.15, |
| "learning_rate": 1.3585858585858585e-06, |
| "loss": 0.7318, |
| "step": 865500 |
| }, |
| { |
| "epoch": 28.16, |
| "learning_rate": 1.3535353535353537e-06, |
| "loss": 0.7324, |
| "step": 866000 |
| }, |
| { |
| "epoch": 28.18, |
| "learning_rate": 1.3484848484848486e-06, |
| "loss": 0.7327, |
| "step": 866500 |
| }, |
| { |
| "epoch": 28.2, |
| "learning_rate": 1.3434343434343436e-06, |
| "loss": 0.7319, |
| "step": 867000 |
| }, |
| { |
| "epoch": 28.21, |
| "learning_rate": 1.3383838383838385e-06, |
| "loss": 0.7326, |
| "step": 867500 |
| }, |
| { |
| "epoch": 28.23, |
| "learning_rate": 1.3333333333333334e-06, |
| "loss": 0.732, |
| "step": 868000 |
| }, |
| { |
| "epoch": 28.25, |
| "learning_rate": 1.3282828282828283e-06, |
| "loss": 0.7321, |
| "step": 868500 |
| }, |
| { |
| "epoch": 28.26, |
| "learning_rate": 1.3232323232323233e-06, |
| "loss": 0.7319, |
| "step": 869000 |
| }, |
| { |
| "epoch": 28.28, |
| "learning_rate": 1.3181818181818182e-06, |
| "loss": 0.7328, |
| "step": 869500 |
| }, |
| { |
| "epoch": 28.29, |
| "learning_rate": 1.3131313131313134e-06, |
| "loss": 0.7317, |
| "step": 870000 |
| }, |
| { |
| "epoch": 28.31, |
| "learning_rate": 1.3080808080808083e-06, |
| "loss": 0.7326, |
| "step": 870500 |
| }, |
| { |
| "epoch": 28.33, |
| "learning_rate": 1.3030303030303032e-06, |
| "loss": 0.7322, |
| "step": 871000 |
| }, |
| { |
| "epoch": 28.34, |
| "learning_rate": 1.2979797979797981e-06, |
| "loss": 0.7318, |
| "step": 871500 |
| }, |
| { |
| "epoch": 28.36, |
| "learning_rate": 1.292929292929293e-06, |
| "loss": 0.7319, |
| "step": 872000 |
| }, |
| { |
| "epoch": 28.38, |
| "learning_rate": 1.287878787878788e-06, |
| "loss": 0.7317, |
| "step": 872500 |
| }, |
| { |
| "epoch": 28.39, |
| "learning_rate": 1.282828282828283e-06, |
| "loss": 0.7317, |
| "step": 873000 |
| }, |
| { |
| "epoch": 28.41, |
| "learning_rate": 1.2777777777777779e-06, |
| "loss": 0.732, |
| "step": 873500 |
| }, |
| { |
| "epoch": 28.42, |
| "learning_rate": 1.2727272727272728e-06, |
| "loss": 0.7315, |
| "step": 874000 |
| }, |
| { |
| "epoch": 28.44, |
| "learning_rate": 1.267676767676768e-06, |
| "loss": 0.7317, |
| "step": 874500 |
| }, |
| { |
| "epoch": 28.46, |
| "learning_rate": 1.2626262626262629e-06, |
| "loss": 0.7317, |
| "step": 875000 |
| }, |
| { |
| "epoch": 28.47, |
| "learning_rate": 1.2575757575757578e-06, |
| "loss": 0.7311, |
| "step": 875500 |
| }, |
| { |
| "epoch": 28.49, |
| "learning_rate": 1.2525252525252527e-06, |
| "loss": 0.7312, |
| "step": 876000 |
| }, |
| { |
| "epoch": 28.51, |
| "learning_rate": 1.2474747474747475e-06, |
| "loss": 0.731, |
| "step": 876500 |
| }, |
| { |
| "epoch": 28.52, |
| "learning_rate": 1.2424242424242424e-06, |
| "loss": 0.7316, |
| "step": 877000 |
| }, |
| { |
| "epoch": 28.54, |
| "learning_rate": 1.2373737373737375e-06, |
| "loss": 0.7311, |
| "step": 877500 |
| }, |
| { |
| "epoch": 28.55, |
| "learning_rate": 1.2323232323232325e-06, |
| "loss": 0.7313, |
| "step": 878000 |
| }, |
| { |
| "epoch": 28.57, |
| "learning_rate": 1.2272727272727274e-06, |
| "loss": 0.7317, |
| "step": 878500 |
| }, |
| { |
| "epoch": 28.59, |
| "learning_rate": 1.2222222222222223e-06, |
| "loss": 0.7309, |
| "step": 879000 |
| }, |
| { |
| "epoch": 28.6, |
| "learning_rate": 1.2171717171717173e-06, |
| "loss": 0.7308, |
| "step": 879500 |
| }, |
| { |
| "epoch": 28.62, |
| "learning_rate": 1.2121212121212122e-06, |
| "loss": 0.7302, |
| "step": 880000 |
| }, |
| { |
| "epoch": 28.64, |
| "learning_rate": 1.2070707070707071e-06, |
| "loss": 0.731, |
| "step": 880500 |
| }, |
| { |
| "epoch": 28.65, |
| "learning_rate": 1.202020202020202e-06, |
| "loss": 0.7309, |
| "step": 881000 |
| }, |
| { |
| "epoch": 28.67, |
| "learning_rate": 1.196969696969697e-06, |
| "loss": 0.7312, |
| "step": 881500 |
| }, |
| { |
| "epoch": 28.68, |
| "learning_rate": 1.1919191919191921e-06, |
| "loss": 0.7308, |
| "step": 882000 |
| }, |
| { |
| "epoch": 28.7, |
| "learning_rate": 1.186868686868687e-06, |
| "loss": 0.7306, |
| "step": 882500 |
| }, |
| { |
| "epoch": 28.72, |
| "learning_rate": 1.181818181818182e-06, |
| "loss": 0.7308, |
| "step": 883000 |
| }, |
| { |
| "epoch": 28.73, |
| "learning_rate": 1.1767676767676767e-06, |
| "loss": 0.7303, |
| "step": 883500 |
| }, |
| { |
| "epoch": 28.75, |
| "learning_rate": 1.1717171717171719e-06, |
| "loss": 0.7307, |
| "step": 884000 |
| }, |
| { |
| "epoch": 28.77, |
| "learning_rate": 1.1666666666666668e-06, |
| "loss": 0.7306, |
| "step": 884500 |
| }, |
| { |
| "epoch": 28.78, |
| "learning_rate": 1.1616161616161617e-06, |
| "loss": 0.7307, |
| "step": 885000 |
| }, |
| { |
| "epoch": 28.8, |
| "learning_rate": 1.1565656565656567e-06, |
| "loss": 0.7302, |
| "step": 885500 |
| }, |
| { |
| "epoch": 28.81, |
| "learning_rate": 1.1515151515151516e-06, |
| "loss": 0.7302, |
| "step": 886000 |
| }, |
| { |
| "epoch": 28.83, |
| "learning_rate": 1.1464646464646465e-06, |
| "loss": 0.7304, |
| "step": 886500 |
| }, |
| { |
| "epoch": 28.85, |
| "learning_rate": 1.1414141414141414e-06, |
| "loss": 0.73, |
| "step": 887000 |
| }, |
| { |
| "epoch": 28.86, |
| "learning_rate": 1.1363636363636364e-06, |
| "loss": 0.7298, |
| "step": 887500 |
| }, |
| { |
| "epoch": 28.88, |
| "learning_rate": 1.1313131313131315e-06, |
| "loss": 0.7299, |
| "step": 888000 |
| }, |
| { |
| "epoch": 28.9, |
| "learning_rate": 1.1262626262626265e-06, |
| "loss": 0.7302, |
| "step": 888500 |
| }, |
| { |
| "epoch": 28.91, |
| "learning_rate": 1.1212121212121214e-06, |
| "loss": 0.7303, |
| "step": 889000 |
| }, |
| { |
| "epoch": 28.93, |
| "learning_rate": 1.1161616161616163e-06, |
| "loss": 0.7299, |
| "step": 889500 |
| }, |
| { |
| "epoch": 28.94, |
| "learning_rate": 1.111111111111111e-06, |
| "loss": 0.7296, |
| "step": 890000 |
| }, |
| { |
| "epoch": 28.96, |
| "learning_rate": 1.1060606060606062e-06, |
| "loss": 0.7298, |
| "step": 890500 |
| }, |
| { |
| "epoch": 28.98, |
| "learning_rate": 1.1010101010101011e-06, |
| "loss": 0.7298, |
| "step": 891000 |
| }, |
| { |
| "epoch": 28.99, |
| "learning_rate": 1.095959595959596e-06, |
| "loss": 0.73, |
| "step": 891500 |
| }, |
| { |
| "epoch": 29.01, |
| "learning_rate": 1.090909090909091e-06, |
| "loss": 0.7298, |
| "step": 892000 |
| }, |
| { |
| "epoch": 29.03, |
| "learning_rate": 1.085858585858586e-06, |
| "loss": 0.7295, |
| "step": 892500 |
| }, |
| { |
| "epoch": 29.04, |
| "learning_rate": 1.0808080808080808e-06, |
| "loss": 0.7292, |
| "step": 893000 |
| }, |
| { |
| "epoch": 29.06, |
| "learning_rate": 1.0757575757575758e-06, |
| "loss": 0.7294, |
| "step": 893500 |
| }, |
| { |
| "epoch": 29.08, |
| "learning_rate": 1.0707070707070707e-06, |
| "loss": 0.7294, |
| "step": 894000 |
| }, |
| { |
| "epoch": 29.09, |
| "learning_rate": 1.0656565656565658e-06, |
| "loss": 0.7298, |
| "step": 894500 |
| }, |
| { |
| "epoch": 29.11, |
| "learning_rate": 1.0606060606060608e-06, |
| "loss": 0.7296, |
| "step": 895000 |
| }, |
| { |
| "epoch": 29.12, |
| "learning_rate": 1.0555555555555557e-06, |
| "loss": 0.7298, |
| "step": 895500 |
| }, |
| { |
| "epoch": 29.14, |
| "learning_rate": 1.0505050505050506e-06, |
| "loss": 0.7298, |
| "step": 896000 |
| }, |
| { |
| "epoch": 29.16, |
| "learning_rate": 1.0454545454545456e-06, |
| "loss": 0.7293, |
| "step": 896500 |
| }, |
| { |
| "epoch": 29.17, |
| "learning_rate": 1.0404040404040405e-06, |
| "loss": 0.7288, |
| "step": 897000 |
| }, |
| { |
| "epoch": 29.19, |
| "learning_rate": 1.0353535353535354e-06, |
| "loss": 0.7295, |
| "step": 897500 |
| }, |
| { |
| "epoch": 29.21, |
| "learning_rate": 1.0303030303030304e-06, |
| "loss": 0.7294, |
| "step": 898000 |
| }, |
| { |
| "epoch": 29.22, |
| "learning_rate": 1.0252525252525253e-06, |
| "loss": 0.7293, |
| "step": 898500 |
| }, |
| { |
| "epoch": 29.24, |
| "learning_rate": 1.0202020202020202e-06, |
| "loss": 0.7295, |
| "step": 899000 |
| }, |
| { |
| "epoch": 29.25, |
| "learning_rate": 1.0151515151515152e-06, |
| "loss": 0.7291, |
| "step": 899500 |
| }, |
| { |
| "epoch": 29.27, |
| "learning_rate": 1.01010101010101e-06, |
| "loss": 0.7292, |
| "step": 900000 |
| }, |
| { |
| "epoch": 29.29, |
| "learning_rate": 1.005050505050505e-06, |
| "loss": 0.7289, |
| "step": 900500 |
| }, |
| { |
| "epoch": 29.3, |
| "learning_rate": 1.0000000000000002e-06, |
| "loss": 0.7289, |
| "step": 901000 |
| }, |
| { |
| "epoch": 29.32, |
| "learning_rate": 9.94949494949495e-07, |
| "loss": 0.7292, |
| "step": 901500 |
| }, |
| { |
| "epoch": 29.34, |
| "learning_rate": 9.8989898989899e-07, |
| "loss": 0.7291, |
| "step": 902000 |
| }, |
| { |
| "epoch": 29.35, |
| "learning_rate": 9.84848484848485e-07, |
| "loss": 0.7288, |
| "step": 902500 |
| }, |
| { |
| "epoch": 29.37, |
| "learning_rate": 9.797979797979799e-07, |
| "loss": 0.7287, |
| "step": 903000 |
| }, |
| { |
| "epoch": 29.38, |
| "learning_rate": 9.747474747474748e-07, |
| "loss": 0.7287, |
| "step": 903500 |
| }, |
| { |
| "epoch": 29.4, |
| "learning_rate": 9.696969696969698e-07, |
| "loss": 0.7286, |
| "step": 904000 |
| }, |
| { |
| "epoch": 29.42, |
| "learning_rate": 9.646464646464647e-07, |
| "loss": 0.729, |
| "step": 904500 |
| }, |
| { |
| "epoch": 29.43, |
| "learning_rate": 9.595959595959596e-07, |
| "loss": 0.7284, |
| "step": 905000 |
| }, |
| { |
| "epoch": 29.45, |
| "learning_rate": 9.545454545454548e-07, |
| "loss": 0.7285, |
| "step": 905500 |
| }, |
| { |
| "epoch": 29.47, |
| "learning_rate": 9.494949494949496e-07, |
| "loss": 0.7286, |
| "step": 906000 |
| }, |
| { |
| "epoch": 29.48, |
| "learning_rate": 9.444444444444445e-07, |
| "loss": 0.7281, |
| "step": 906500 |
| }, |
| { |
| "epoch": 29.5, |
| "learning_rate": 9.393939393939395e-07, |
| "loss": 0.7286, |
| "step": 907000 |
| }, |
| { |
| "epoch": 29.51, |
| "learning_rate": 9.343434343434345e-07, |
| "loss": 0.7283, |
| "step": 907500 |
| }, |
| { |
| "epoch": 29.53, |
| "learning_rate": 9.292929292929294e-07, |
| "loss": 0.7284, |
| "step": 908000 |
| }, |
| { |
| "epoch": 29.55, |
| "learning_rate": 9.242424242424244e-07, |
| "loss": 0.7282, |
| "step": 908500 |
| }, |
| { |
| "epoch": 29.56, |
| "learning_rate": 9.191919191919192e-07, |
| "loss": 0.7281, |
| "step": 909000 |
| }, |
| { |
| "epoch": 29.58, |
| "learning_rate": 9.141414141414143e-07, |
| "loss": 0.7287, |
| "step": 909500 |
| }, |
| { |
| "epoch": 29.6, |
| "learning_rate": 9.090909090909091e-07, |
| "loss": 0.7282, |
| "step": 910000 |
| }, |
| { |
| "epoch": 29.61, |
| "learning_rate": 9.040404040404041e-07, |
| "loss": 0.728, |
| "step": 910500 |
| }, |
| { |
| "epoch": 29.63, |
| "learning_rate": 8.98989898989899e-07, |
| "loss": 0.7278, |
| "step": 911000 |
| }, |
| { |
| "epoch": 29.64, |
| "learning_rate": 8.93939393939394e-07, |
| "loss": 0.7287, |
| "step": 911500 |
| }, |
| { |
| "epoch": 29.66, |
| "learning_rate": 8.88888888888889e-07, |
| "loss": 0.7282, |
| "step": 912000 |
| }, |
| { |
| "epoch": 29.68, |
| "learning_rate": 8.838383838383839e-07, |
| "loss": 0.7282, |
| "step": 912500 |
| }, |
| { |
| "epoch": 29.69, |
| "learning_rate": 8.787878787878788e-07, |
| "loss": 0.7281, |
| "step": 913000 |
| }, |
| { |
| "epoch": 29.71, |
| "learning_rate": 8.737373737373738e-07, |
| "loss": 0.7279, |
| "step": 913500 |
| }, |
| { |
| "epoch": 29.73, |
| "learning_rate": 8.686868686868688e-07, |
| "loss": 0.7285, |
| "step": 914000 |
| }, |
| { |
| "epoch": 29.74, |
| "learning_rate": 8.636363636363637e-07, |
| "loss": 0.7277, |
| "step": 914500 |
| }, |
| { |
| "epoch": 29.76, |
| "learning_rate": 8.585858585858587e-07, |
| "loss": 0.7281, |
| "step": 915000 |
| }, |
| { |
| "epoch": 29.77, |
| "learning_rate": 8.535353535353535e-07, |
| "loss": 0.7276, |
| "step": 915500 |
| }, |
| { |
| "epoch": 29.79, |
| "learning_rate": 8.484848484848486e-07, |
| "loss": 0.7278, |
| "step": 916000 |
| }, |
| { |
| "epoch": 29.81, |
| "learning_rate": 8.434343434343436e-07, |
| "loss": 0.7277, |
| "step": 916500 |
| }, |
| { |
| "epoch": 29.82, |
| "learning_rate": 8.383838383838384e-07, |
| "loss": 0.7279, |
| "step": 917000 |
| }, |
| { |
| "epoch": 29.84, |
| "learning_rate": 8.333333333333333e-07, |
| "loss": 0.7276, |
| "step": 917500 |
| }, |
| { |
| "epoch": 29.86, |
| "learning_rate": 8.282828282828284e-07, |
| "loss": 0.7276, |
| "step": 918000 |
| }, |
| { |
| "epoch": 29.87, |
| "learning_rate": 8.232323232323233e-07, |
| "loss": 0.7277, |
| "step": 918500 |
| }, |
| { |
| "epoch": 29.89, |
| "learning_rate": 8.181818181818182e-07, |
| "loss": 0.7276, |
| "step": 919000 |
| }, |
| { |
| "epoch": 29.9, |
| "learning_rate": 8.131313131313132e-07, |
| "loss": 0.7273, |
| "step": 919500 |
| }, |
| { |
| "epoch": 29.92, |
| "learning_rate": 8.080808080808082e-07, |
| "loss": 0.7274, |
| "step": 920000 |
| }, |
| { |
| "epoch": 29.94, |
| "learning_rate": 8.030303030303031e-07, |
| "loss": 0.7277, |
| "step": 920500 |
| }, |
| { |
| "epoch": 29.95, |
| "learning_rate": 7.979797979797981e-07, |
| "loss": 0.7275, |
| "step": 921000 |
| }, |
| { |
| "epoch": 29.97, |
| "learning_rate": 7.92929292929293e-07, |
| "loss": 0.7274, |
| "step": 921500 |
| }, |
| { |
| "epoch": 29.99, |
| "learning_rate": 7.878787878787879e-07, |
| "loss": 0.7269, |
| "step": 922000 |
| }, |
| { |
| "epoch": 30.0, |
| "learning_rate": 7.82828282828283e-07, |
| "loss": 0.7276, |
| "step": 922500 |
| }, |
| { |
| "epoch": 30.02, |
| "learning_rate": 7.777777777777779e-07, |
| "loss": 0.7273, |
| "step": 923000 |
| }, |
| { |
| "epoch": 30.03, |
| "learning_rate": 7.727272727272727e-07, |
| "loss": 0.7275, |
| "step": 923500 |
| }, |
| { |
| "epoch": 30.05, |
| "learning_rate": 7.676767676767677e-07, |
| "loss": 0.7276, |
| "step": 924000 |
| }, |
| { |
| "epoch": 30.07, |
| "learning_rate": 7.626262626262627e-07, |
| "loss": 0.7272, |
| "step": 924500 |
| }, |
| { |
| "epoch": 30.08, |
| "learning_rate": 7.575757575757576e-07, |
| "loss": 0.7269, |
| "step": 925000 |
| }, |
| { |
| "epoch": 30.1, |
| "learning_rate": 7.525252525252526e-07, |
| "loss": 0.7269, |
| "step": 925500 |
| }, |
| { |
| "epoch": 30.12, |
| "learning_rate": 7.474747474747475e-07, |
| "loss": 0.7274, |
| "step": 926000 |
| }, |
| { |
| "epoch": 30.13, |
| "learning_rate": 7.424242424242425e-07, |
| "loss": 0.7272, |
| "step": 926500 |
| }, |
| { |
| "epoch": 30.15, |
| "learning_rate": 7.373737373737375e-07, |
| "loss": 0.7269, |
| "step": 927000 |
| }, |
| { |
| "epoch": 30.16, |
| "learning_rate": 7.323232323232324e-07, |
| "loss": 0.727, |
| "step": 927500 |
| }, |
| { |
| "epoch": 30.18, |
| "learning_rate": 7.272727272727273e-07, |
| "loss": 0.7265, |
| "step": 928000 |
| }, |
| { |
| "epoch": 30.2, |
| "learning_rate": 7.222222222222222e-07, |
| "loss": 0.727, |
| "step": 928500 |
| }, |
| { |
| "epoch": 30.21, |
| "learning_rate": 7.171717171717173e-07, |
| "loss": 0.7272, |
| "step": 929000 |
| }, |
| { |
| "epoch": 30.23, |
| "learning_rate": 7.121212121212122e-07, |
| "loss": 0.7275, |
| "step": 929500 |
| }, |
| { |
| "epoch": 30.25, |
| "learning_rate": 7.070707070707071e-07, |
| "loss": 0.7272, |
| "step": 930000 |
| }, |
| { |
| "epoch": 30.26, |
| "learning_rate": 7.02020202020202e-07, |
| "loss": 0.7266, |
| "step": 930500 |
| }, |
| { |
| "epoch": 30.28, |
| "learning_rate": 6.969696969696971e-07, |
| "loss": 0.7271, |
| "step": 931000 |
| }, |
| { |
| "epoch": 30.29, |
| "learning_rate": 6.919191919191919e-07, |
| "loss": 0.7265, |
| "step": 931500 |
| }, |
| { |
| "epoch": 30.31, |
| "learning_rate": 6.868686868686869e-07, |
| "loss": 0.7268, |
| "step": 932000 |
| }, |
| { |
| "epoch": 30.33, |
| "learning_rate": 6.818181818181818e-07, |
| "loss": 0.7268, |
| "step": 932500 |
| }, |
| { |
| "epoch": 30.34, |
| "learning_rate": 6.767676767676768e-07, |
| "loss": 0.7266, |
| "step": 933000 |
| }, |
| { |
| "epoch": 30.36, |
| "learning_rate": 6.717171717171718e-07, |
| "loss": 0.7267, |
| "step": 933500 |
| }, |
| { |
| "epoch": 30.38, |
| "learning_rate": 6.666666666666667e-07, |
| "loss": 0.7264, |
| "step": 934000 |
| }, |
| { |
| "epoch": 30.39, |
| "learning_rate": 6.616161616161616e-07, |
| "loss": 0.7264, |
| "step": 934500 |
| }, |
| { |
| "epoch": 30.41, |
| "learning_rate": 6.565656565656567e-07, |
| "loss": 0.726, |
| "step": 935000 |
| }, |
| { |
| "epoch": 30.42, |
| "learning_rate": 6.515151515151516e-07, |
| "loss": 0.7263, |
| "step": 935500 |
| }, |
| { |
| "epoch": 30.44, |
| "learning_rate": 6.464646464646465e-07, |
| "loss": 0.7265, |
| "step": 936000 |
| }, |
| { |
| "epoch": 30.46, |
| "learning_rate": 6.414141414141415e-07, |
| "loss": 0.7267, |
| "step": 936500 |
| }, |
| { |
| "epoch": 30.47, |
| "learning_rate": 6.363636363636364e-07, |
| "loss": 0.7266, |
| "step": 937000 |
| }, |
| { |
| "epoch": 30.49, |
| "learning_rate": 6.313131313131314e-07, |
| "loss": 0.7266, |
| "step": 937500 |
| }, |
| { |
| "epoch": 30.51, |
| "learning_rate": 6.262626262626264e-07, |
| "loss": 0.7265, |
| "step": 938000 |
| }, |
| { |
| "epoch": 30.52, |
| "learning_rate": 6.212121212121212e-07, |
| "loss": 0.7266, |
| "step": 938500 |
| }, |
| { |
| "epoch": 30.54, |
| "learning_rate": 6.161616161616162e-07, |
| "loss": 0.7263, |
| "step": 939000 |
| }, |
| { |
| "epoch": 30.55, |
| "learning_rate": 6.111111111111112e-07, |
| "loss": 0.7269, |
| "step": 939500 |
| }, |
| { |
| "epoch": 30.57, |
| "learning_rate": 6.060606060606061e-07, |
| "loss": 0.7264, |
| "step": 940000 |
| }, |
| { |
| "epoch": 30.59, |
| "learning_rate": 6.01010101010101e-07, |
| "loss": 0.7259, |
| "step": 940500 |
| }, |
| { |
| "epoch": 30.6, |
| "learning_rate": 5.959595959595961e-07, |
| "loss": 0.727, |
| "step": 941000 |
| }, |
| { |
| "epoch": 30.62, |
| "learning_rate": 5.90909090909091e-07, |
| "loss": 0.7261, |
| "step": 941500 |
| }, |
| { |
| "epoch": 30.64, |
| "learning_rate": 5.858585858585859e-07, |
| "loss": 0.7261, |
| "step": 942000 |
| }, |
| { |
| "epoch": 30.65, |
| "learning_rate": 5.808080808080809e-07, |
| "loss": 0.7262, |
| "step": 942500 |
| }, |
| { |
| "epoch": 30.67, |
| "learning_rate": 5.757575757575758e-07, |
| "loss": 0.7271, |
| "step": 943000 |
| }, |
| { |
| "epoch": 30.68, |
| "learning_rate": 5.707070707070707e-07, |
| "loss": 0.726, |
| "step": 943500 |
| }, |
| { |
| "epoch": 30.7, |
| "learning_rate": 5.656565656565658e-07, |
| "loss": 0.7262, |
| "step": 944000 |
| }, |
| { |
| "epoch": 30.72, |
| "learning_rate": 5.606060606060607e-07, |
| "loss": 0.7261, |
| "step": 944500 |
| }, |
| { |
| "epoch": 30.73, |
| "learning_rate": 5.555555555555555e-07, |
| "loss": 0.7267, |
| "step": 945000 |
| }, |
| { |
| "epoch": 30.75, |
| "learning_rate": 5.505050505050506e-07, |
| "loss": 0.7259, |
| "step": 945500 |
| }, |
| { |
| "epoch": 30.77, |
| "learning_rate": 5.454545454545455e-07, |
| "loss": 0.7263, |
| "step": 946000 |
| }, |
| { |
| "epoch": 30.78, |
| "learning_rate": 5.404040404040404e-07, |
| "loss": 0.726, |
| "step": 946500 |
| }, |
| { |
| "epoch": 30.8, |
| "learning_rate": 5.353535353535354e-07, |
| "loss": 0.7261, |
| "step": 947000 |
| }, |
| { |
| "epoch": 30.82, |
| "learning_rate": 5.303030303030304e-07, |
| "loss": 0.7256, |
| "step": 947500 |
| }, |
| { |
| "epoch": 30.83, |
| "learning_rate": 5.252525252525253e-07, |
| "loss": 0.7256, |
| "step": 948000 |
| }, |
| { |
| "epoch": 30.85, |
| "learning_rate": 5.202020202020203e-07, |
| "loss": 0.7262, |
| "step": 948500 |
| }, |
| { |
| "epoch": 30.86, |
| "learning_rate": 5.151515151515152e-07, |
| "loss": 0.7256, |
| "step": 949000 |
| }, |
| { |
| "epoch": 30.88, |
| "learning_rate": 5.101010101010101e-07, |
| "loss": 0.726, |
| "step": 949500 |
| }, |
| { |
| "epoch": 30.9, |
| "learning_rate": 5.05050505050505e-07, |
| "loss": 0.7259, |
| "step": 950000 |
| }, |
| { |
| "epoch": 30.91, |
| "learning_rate": 5.000000000000001e-07, |
| "loss": 0.7258, |
| "step": 950500 |
| }, |
| { |
| "epoch": 30.93, |
| "learning_rate": 4.94949494949495e-07, |
| "loss": 0.726, |
| "step": 951000 |
| }, |
| { |
| "epoch": 30.95, |
| "learning_rate": 4.898989898989899e-07, |
| "loss": 0.7255, |
| "step": 951500 |
| }, |
| { |
| "epoch": 30.96, |
| "learning_rate": 4.848484848484849e-07, |
| "loss": 0.7261, |
| "step": 952000 |
| }, |
| { |
| "epoch": 30.98, |
| "learning_rate": 4.797979797979798e-07, |
| "loss": 0.7259, |
| "step": 952500 |
| }, |
| { |
| "epoch": 30.99, |
| "learning_rate": 4.747474747474748e-07, |
| "loss": 0.7255, |
| "step": 953000 |
| }, |
| { |
| "epoch": 31.01, |
| "learning_rate": 4.696969696969697e-07, |
| "loss": 0.7256, |
| "step": 953500 |
| }, |
| { |
| "epoch": 31.03, |
| "learning_rate": 4.646464646464647e-07, |
| "loss": 0.7252, |
| "step": 954000 |
| }, |
| { |
| "epoch": 31.04, |
| "learning_rate": 4.595959595959596e-07, |
| "loss": 0.7259, |
| "step": 954500 |
| }, |
| { |
| "epoch": 31.06, |
| "learning_rate": 4.5454545454545457e-07, |
| "loss": 0.7257, |
| "step": 955000 |
| }, |
| { |
| "epoch": 31.08, |
| "learning_rate": 4.494949494949495e-07, |
| "loss": 0.7259, |
| "step": 955500 |
| }, |
| { |
| "epoch": 31.09, |
| "learning_rate": 4.444444444444445e-07, |
| "loss": 0.7258, |
| "step": 956000 |
| }, |
| { |
| "epoch": 31.11, |
| "learning_rate": 4.393939393939394e-07, |
| "loss": 0.7261, |
| "step": 956500 |
| }, |
| { |
| "epoch": 31.12, |
| "learning_rate": 4.343434343434344e-07, |
| "loss": 0.7258, |
| "step": 957000 |
| }, |
| { |
| "epoch": 31.14, |
| "learning_rate": 4.2929292929292934e-07, |
| "loss": 0.7256, |
| "step": 957500 |
| }, |
| { |
| "epoch": 31.16, |
| "learning_rate": 4.242424242424243e-07, |
| "loss": 0.7256, |
| "step": 958000 |
| }, |
| { |
| "epoch": 31.17, |
| "learning_rate": 4.191919191919192e-07, |
| "loss": 0.7262, |
| "step": 958500 |
| }, |
| { |
| "epoch": 31.19, |
| "learning_rate": 4.141414141414142e-07, |
| "loss": 0.7258, |
| "step": 959000 |
| }, |
| { |
| "epoch": 31.21, |
| "learning_rate": 4.090909090909091e-07, |
| "loss": 0.7248, |
| "step": 959500 |
| }, |
| { |
| "epoch": 31.22, |
| "learning_rate": 4.040404040404041e-07, |
| "loss": 0.7253, |
| "step": 960000 |
| }, |
| { |
| "epoch": 31.24, |
| "learning_rate": 3.9898989898989903e-07, |
| "loss": 0.7253, |
| "step": 960500 |
| }, |
| { |
| "epoch": 31.25, |
| "learning_rate": 3.9393939393939396e-07, |
| "loss": 0.7258, |
| "step": 961000 |
| }, |
| { |
| "epoch": 31.27, |
| "learning_rate": 3.8888888888888895e-07, |
| "loss": 0.7255, |
| "step": 961500 |
| }, |
| { |
| "epoch": 31.29, |
| "learning_rate": 3.838383838383838e-07, |
| "loss": 0.7255, |
| "step": 962000 |
| }, |
| { |
| "epoch": 31.3, |
| "learning_rate": 3.787878787878788e-07, |
| "loss": 0.7256, |
| "step": 962500 |
| }, |
| { |
| "epoch": 31.32, |
| "learning_rate": 3.7373737373737374e-07, |
| "loss": 0.7252, |
| "step": 963000 |
| }, |
| { |
| "epoch": 31.34, |
| "learning_rate": 3.6868686868686873e-07, |
| "loss": 0.7257, |
| "step": 963500 |
| }, |
| { |
| "epoch": 31.35, |
| "learning_rate": 3.6363636363636366e-07, |
| "loss": 0.7254, |
| "step": 964000 |
| }, |
| { |
| "epoch": 31.37, |
| "learning_rate": 3.5858585858585864e-07, |
| "loss": 0.7256, |
| "step": 964500 |
| }, |
| { |
| "epoch": 31.38, |
| "learning_rate": 3.535353535353536e-07, |
| "loss": 0.7257, |
| "step": 965000 |
| }, |
| { |
| "epoch": 31.4, |
| "learning_rate": 3.4848484848484856e-07, |
| "loss": 0.7256, |
| "step": 965500 |
| }, |
| { |
| "epoch": 31.42, |
| "learning_rate": 3.4343434343434344e-07, |
| "loss": 0.7255, |
| "step": 966000 |
| }, |
| { |
| "epoch": 31.43, |
| "learning_rate": 3.383838383838384e-07, |
| "loss": 0.7255, |
| "step": 966500 |
| }, |
| { |
| "epoch": 31.45, |
| "learning_rate": 3.3333333333333335e-07, |
| "loss": 0.7252, |
| "step": 967000 |
| }, |
| { |
| "epoch": 31.47, |
| "learning_rate": 3.2828282828282834e-07, |
| "loss": 0.7256, |
| "step": 967500 |
| }, |
| { |
| "epoch": 31.48, |
| "learning_rate": 3.2323232323232327e-07, |
| "loss": 0.7257, |
| "step": 968000 |
| }, |
| { |
| "epoch": 31.5, |
| "learning_rate": 3.181818181818182e-07, |
| "loss": 0.7259, |
| "step": 968500 |
| }, |
| { |
| "epoch": 31.51, |
| "learning_rate": 3.131313131313132e-07, |
| "loss": 0.7252, |
| "step": 969000 |
| }, |
| { |
| "epoch": 31.53, |
| "learning_rate": 3.080808080808081e-07, |
| "loss": 0.7257, |
| "step": 969500 |
| }, |
| { |
| "epoch": 31.55, |
| "learning_rate": 3.0303030303030305e-07, |
| "loss": 0.725, |
| "step": 970000 |
| }, |
| { |
| "epoch": 31.56, |
| "learning_rate": 2.9797979797979803e-07, |
| "loss": 0.7252, |
| "step": 970500 |
| }, |
| { |
| "epoch": 31.58, |
| "learning_rate": 2.9292929292929296e-07, |
| "loss": 0.7253, |
| "step": 971000 |
| }, |
| { |
| "epoch": 31.6, |
| "learning_rate": 2.878787878787879e-07, |
| "loss": 0.7251, |
| "step": 971500 |
| }, |
| { |
| "epoch": 31.61, |
| "learning_rate": 2.828282828282829e-07, |
| "loss": 0.725, |
| "step": 972000 |
| }, |
| { |
| "epoch": 31.63, |
| "learning_rate": 2.7777777777777776e-07, |
| "loss": 0.7256, |
| "step": 972500 |
| }, |
| { |
| "epoch": 31.64, |
| "learning_rate": 2.7272727272727274e-07, |
| "loss": 0.7257, |
| "step": 973000 |
| }, |
| { |
| "epoch": 31.66, |
| "learning_rate": 2.676767676767677e-07, |
| "loss": 0.7248, |
| "step": 973500 |
| }, |
| { |
| "epoch": 31.68, |
| "learning_rate": 2.6262626262626266e-07, |
| "loss": 0.7248, |
| "step": 974000 |
| }, |
| { |
| "epoch": 31.69, |
| "learning_rate": 2.575757575757576e-07, |
| "loss": 0.7246, |
| "step": 974500 |
| }, |
| { |
| "epoch": 31.71, |
| "learning_rate": 2.525252525252525e-07, |
| "loss": 0.7255, |
| "step": 975000 |
| }, |
| { |
| "epoch": 31.73, |
| "learning_rate": 2.474747474747475e-07, |
| "loss": 0.7248, |
| "step": 975500 |
| }, |
| { |
| "epoch": 31.74, |
| "learning_rate": 2.4242424242424244e-07, |
| "loss": 0.7249, |
| "step": 976000 |
| }, |
| { |
| "epoch": 31.76, |
| "learning_rate": 2.373737373737374e-07, |
| "loss": 0.7253, |
| "step": 976500 |
| }, |
| { |
| "epoch": 31.77, |
| "learning_rate": 2.3232323232323235e-07, |
| "loss": 0.7248, |
| "step": 977000 |
| }, |
| { |
| "epoch": 31.79, |
| "learning_rate": 2.2727272727272729e-07, |
| "loss": 0.725, |
| "step": 977500 |
| }, |
| { |
| "epoch": 31.81, |
| "learning_rate": 2.2222222222222224e-07, |
| "loss": 0.7255, |
| "step": 978000 |
| }, |
| { |
| "epoch": 31.82, |
| "learning_rate": 2.171717171717172e-07, |
| "loss": 0.7251, |
| "step": 978500 |
| }, |
| { |
| "epoch": 31.84, |
| "learning_rate": 2.1212121212121216e-07, |
| "loss": 0.7252, |
| "step": 979000 |
| }, |
| { |
| "epoch": 31.86, |
| "learning_rate": 2.070707070707071e-07, |
| "loss": 0.7252, |
| "step": 979500 |
| }, |
| { |
| "epoch": 31.87, |
| "learning_rate": 2.0202020202020205e-07, |
| "loss": 0.7246, |
| "step": 980000 |
| }, |
| { |
| "epoch": 31.89, |
| "learning_rate": 1.9696969696969698e-07, |
| "loss": 0.725, |
| "step": 980500 |
| }, |
| { |
| "epoch": 31.9, |
| "learning_rate": 1.919191919191919e-07, |
| "loss": 0.7248, |
| "step": 981000 |
| }, |
| { |
| "epoch": 31.92, |
| "learning_rate": 1.8686868686868687e-07, |
| "loss": 0.7255, |
| "step": 981500 |
| }, |
| { |
| "epoch": 31.94, |
| "learning_rate": 1.8181818181818183e-07, |
| "loss": 0.7248, |
| "step": 982000 |
| }, |
| { |
| "epoch": 31.95, |
| "learning_rate": 1.767676767676768e-07, |
| "loss": 0.7252, |
| "step": 982500 |
| }, |
| { |
| "epoch": 31.97, |
| "learning_rate": 1.7171717171717172e-07, |
| "loss": 0.7248, |
| "step": 983000 |
| }, |
| { |
| "epoch": 31.99, |
| "learning_rate": 1.6666666666666668e-07, |
| "loss": 0.7254, |
| "step": 983500 |
| }, |
| { |
| "epoch": 32.0, |
| "learning_rate": 1.6161616161616163e-07, |
| "loss": 0.725, |
| "step": 984000 |
| }, |
| { |
| "epoch": 32.02, |
| "learning_rate": 1.565656565656566e-07, |
| "loss": 0.7259, |
| "step": 984500 |
| }, |
| { |
| "epoch": 32.03, |
| "learning_rate": 1.5151515151515152e-07, |
| "loss": 0.7246, |
| "step": 985000 |
| }, |
| { |
| "epoch": 32.05, |
| "learning_rate": 1.4646464646464648e-07, |
| "loss": 0.7247, |
| "step": 985500 |
| }, |
| { |
| "epoch": 32.07, |
| "learning_rate": 1.4141414141414144e-07, |
| "loss": 0.7248, |
| "step": 986000 |
| }, |
| { |
| "epoch": 32.08, |
| "learning_rate": 1.3636363636363637e-07, |
| "loss": 0.7252, |
| "step": 986500 |
| }, |
| { |
| "epoch": 32.1, |
| "learning_rate": 1.3131313131313133e-07, |
| "loss": 0.7255, |
| "step": 987000 |
| }, |
| { |
| "epoch": 32.12, |
| "learning_rate": 1.2626262626262626e-07, |
| "loss": 0.7247, |
| "step": 987500 |
| }, |
| { |
| "epoch": 32.13, |
| "learning_rate": 1.2121212121212122e-07, |
| "loss": 0.7244, |
| "step": 988000 |
| }, |
| { |
| "epoch": 32.15, |
| "learning_rate": 1.1616161616161618e-07, |
| "loss": 0.7251, |
| "step": 988500 |
| }, |
| { |
| "epoch": 32.16, |
| "learning_rate": 1.1111111111111112e-07, |
| "loss": 0.7248, |
| "step": 989000 |
| }, |
| { |
| "epoch": 32.18, |
| "learning_rate": 1.0606060606060608e-07, |
| "loss": 0.7248, |
| "step": 989500 |
| }, |
| { |
| "epoch": 32.2, |
| "learning_rate": 1.0101010101010103e-07, |
| "loss": 0.7251, |
| "step": 990000 |
| }, |
| { |
| "epoch": 32.21, |
| "learning_rate": 9.595959595959596e-08, |
| "loss": 0.7248, |
| "step": 990500 |
| }, |
| { |
| "epoch": 32.23, |
| "learning_rate": 9.090909090909091e-08, |
| "loss": 0.7247, |
| "step": 991000 |
| }, |
| { |
| "epoch": 32.25, |
| "learning_rate": 8.585858585858586e-08, |
| "loss": 0.7249, |
| "step": 991500 |
| }, |
| { |
| "epoch": 32.26, |
| "learning_rate": 8.080808080808082e-08, |
| "loss": 0.7252, |
| "step": 992000 |
| }, |
| { |
| "epoch": 32.28, |
| "learning_rate": 7.575757575757576e-08, |
| "loss": 0.725, |
| "step": 992500 |
| }, |
| { |
| "epoch": 32.29, |
| "learning_rate": 7.070707070707072e-08, |
| "loss": 0.7253, |
| "step": 993000 |
| }, |
| { |
| "epoch": 32.31, |
| "learning_rate": 6.565656565656566e-08, |
| "loss": 0.7255, |
| "step": 993500 |
| }, |
| { |
| "epoch": 32.33, |
| "learning_rate": 6.060606060606061e-08, |
| "loss": 0.7248, |
| "step": 994000 |
| }, |
| { |
| "epoch": 32.34, |
| "learning_rate": 5.555555555555556e-08, |
| "loss": 0.7249, |
| "step": 994500 |
| }, |
| { |
| "epoch": 32.36, |
| "learning_rate": 5.050505050505051e-08, |
| "loss": 0.724, |
| "step": 995000 |
| }, |
| { |
| "epoch": 32.38, |
| "learning_rate": 4.545454545454546e-08, |
| "loss": 0.725, |
| "step": 995500 |
| }, |
| { |
| "epoch": 32.39, |
| "learning_rate": 4.040404040404041e-08, |
| "loss": 0.7244, |
| "step": 996000 |
| }, |
| { |
| "epoch": 32.41, |
| "learning_rate": 3.535353535353536e-08, |
| "loss": 0.725, |
| "step": 996500 |
| }, |
| { |
| "epoch": 32.42, |
| "learning_rate": 3.0303030303030305e-08, |
| "loss": 0.7254, |
| "step": 997000 |
| }, |
| { |
| "epoch": 32.44, |
| "learning_rate": 2.5252525252525256e-08, |
| "loss": 0.7251, |
| "step": 997500 |
| }, |
| { |
| "epoch": 32.46, |
| "learning_rate": 2.0202020202020204e-08, |
| "loss": 0.7244, |
| "step": 998000 |
| }, |
| { |
| "epoch": 32.47, |
| "learning_rate": 1.5151515151515152e-08, |
| "loss": 0.7251, |
| "step": 998500 |
| }, |
| { |
| "epoch": 32.49, |
| "learning_rate": 1.0101010101010102e-08, |
| "loss": 0.7251, |
| "step": 999000 |
| }, |
| { |
| "epoch": 32.51, |
| "learning_rate": 5.050505050505051e-09, |
| "loss": 0.725, |
| "step": 999500 |
| }, |
| { |
| "epoch": 32.52, |
| "learning_rate": 0.0, |
| "loss": 0.7246, |
| "step": 1000000 |
| }, |
| { |
| "epoch": 32.52, |
| "step": 1000000, |
| "total_flos": 1.1571044548608e+19, |
| "train_loss": 1.2262278026428222, |
| "train_runtime": 587774.1428, |
| "train_samples_per_second": 435.541, |
| "train_steps_per_second": 1.701 |
| } |
| ], |
| "max_steps": 1000000, |
| "num_train_epochs": 33, |
| "total_flos": 1.1571044548608e+19, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|