| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 10.0, | |
| "eval_steps": 500, | |
| "global_step": 27650, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.8083182640144665e-09, | |
| "loss": 13.6388, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.616636528028933e-09, | |
| "loss": 13.7324, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.4249547920433996e-09, | |
| "loss": 13.5866, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 7.233273056057866e-09, | |
| "loss": 13.6427, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.041591320072333e-09, | |
| "loss": 13.6357, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.0849909584086799e-08, | |
| "loss": 13.6714, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.2658227848101265e-08, | |
| "loss": 13.6575, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.4466546112115732e-08, | |
| "loss": 13.6404, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.62748643761302e-08, | |
| "loss": 13.7001, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.8083182640144666e-08, | |
| "loss": 13.6391, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9891500904159132e-08, | |
| "loss": 13.7209, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.1699819168173598e-08, | |
| "loss": 13.7199, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.3508137432188065e-08, | |
| "loss": 13.6763, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.531645569620253e-08, | |
| "loss": 13.6408, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.7124773960216997e-08, | |
| "loss": 13.6908, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.8933092224231463e-08, | |
| "loss": 13.7436, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.0741410488245926e-08, | |
| "loss": 13.6329, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 3.25497287522604e-08, | |
| "loss": 13.678, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 3.4358047016274865e-08, | |
| "loss": 13.7068, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 3.616636528028933e-08, | |
| "loss": 13.7395, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 3.79746835443038e-08, | |
| "loss": 13.5979, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 3.9783001808318264e-08, | |
| "loss": 13.6808, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.159132007233273e-08, | |
| "loss": 13.6201, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.3399638336347197e-08, | |
| "loss": 13.6224, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.520795660036166e-08, | |
| "loss": 13.7064, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.701627486437613e-08, | |
| "loss": 13.7375, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.8824593128390595e-08, | |
| "loss": 13.7072, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 5.063291139240506e-08, | |
| "loss": 13.719, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 5.244122965641953e-08, | |
| "loss": 13.6121, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 5.4249547920433994e-08, | |
| "loss": 13.6514, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 5.605786618444846e-08, | |
| "loss": 13.7277, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 5.7866184448462927e-08, | |
| "loss": 13.6584, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 5.967450271247739e-08, | |
| "loss": 13.6968, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 6.148282097649185e-08, | |
| "loss": 13.6416, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 6.329113924050633e-08, | |
| "loss": 13.773, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 6.50994575045208e-08, | |
| "loss": 13.6784, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 6.690777576853526e-08, | |
| "loss": 13.6289, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 6.871609403254973e-08, | |
| "loss": 13.5546, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 7.05244122965642e-08, | |
| "loss": 13.5832, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.233273056057866e-08, | |
| "loss": 13.5718, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.414104882459313e-08, | |
| "loss": 13.6693, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 7.59493670886076e-08, | |
| "loss": 13.7198, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 7.775768535262206e-08, | |
| "loss": 13.6503, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 7.956600361663653e-08, | |
| "loss": 13.6857, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 8.1374321880651e-08, | |
| "loss": 13.6539, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 8.318264014466546e-08, | |
| "loss": 13.6969, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 8.499095840867993e-08, | |
| "loss": 13.7703, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 8.679927667269439e-08, | |
| "loss": 13.6366, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 8.860759493670886e-08, | |
| "loss": 13.7623, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 9.041591320072333e-08, | |
| "loss": 13.668, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 9.222423146473779e-08, | |
| "loss": 13.6859, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 9.403254972875226e-08, | |
| "loss": 13.6488, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 9.584086799276672e-08, | |
| "loss": 13.604, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 9.764918625678119e-08, | |
| "loss": 13.63, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 9.945750452079566e-08, | |
| "loss": 13.6844, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.0126582278481012e-07, | |
| "loss": 13.6574, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.0307414104882459e-07, | |
| "loss": 13.6688, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.0488245931283906e-07, | |
| "loss": 13.6358, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.0669077757685352e-07, | |
| "loss": 13.6021, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.0849909584086799e-07, | |
| "loss": 13.6597, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.1030741410488245e-07, | |
| "loss": 13.5985, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1211573236889692e-07, | |
| "loss": 13.6186, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1392405063291139e-07, | |
| "loss": 13.6391, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1573236889692585e-07, | |
| "loss": 13.599, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1754068716094032e-07, | |
| "loss": 13.7044, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1934900542495477e-07, | |
| "loss": 13.7581, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.2115732368896925e-07, | |
| "loss": 13.648, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.229656419529837e-07, | |
| "loss": 13.6057, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.2477396021699818e-07, | |
| "loss": 13.627, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2658227848101266e-07, | |
| "loss": 13.6813, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2839059674502712e-07, | |
| "loss": 13.5936, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.301989150090416e-07, | |
| "loss": 13.6134, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.3200723327305605e-07, | |
| "loss": 13.6773, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.3381555153707053e-07, | |
| "loss": 13.6927, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.3562386980108498e-07, | |
| "loss": 13.5613, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.3743218806509946e-07, | |
| "loss": 13.6399, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.3924050632911391e-07, | |
| "loss": 13.6222, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.410488245931284e-07, | |
| "loss": 13.6306, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.4285714285714285e-07, | |
| "loss": 13.6322, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.4466546112115733e-07, | |
| "loss": 13.5773, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.4647377938517178e-07, | |
| "loss": 13.6057, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.4828209764918626e-07, | |
| "loss": 13.5897, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.500904159132007e-07, | |
| "loss": 13.508, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.518987341772152e-07, | |
| "loss": 13.6156, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.5370705244122964e-07, | |
| "loss": 13.6333, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.5551537070524412e-07, | |
| "loss": 13.6283, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5732368896925858e-07, | |
| "loss": 13.454, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.5913200723327306e-07, | |
| "loss": 13.6243, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.609403254972875e-07, | |
| "loss": 13.5886, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.62748643761302e-07, | |
| "loss": 13.546, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.6455696202531644e-07, | |
| "loss": 13.5578, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6636528028933092e-07, | |
| "loss": 13.5799, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.6817359855334537e-07, | |
| "loss": 13.6162, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.6998191681735985e-07, | |
| "loss": 13.5781, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.717902350813743e-07, | |
| "loss": 13.5051, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.7359855334538879e-07, | |
| "loss": 13.5705, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.7540687160940324e-07, | |
| "loss": 13.5822, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7721518987341772e-07, | |
| "loss": 13.4456, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.7902350813743217e-07, | |
| "loss": 13.5349, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8083182640144665e-07, | |
| "loss": 13.5085, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8264014466546113e-07, | |
| "loss": 13.6023, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8444846292947558e-07, | |
| "loss": 13.5162, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8625678119349006e-07, | |
| "loss": 13.4671, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8806509945750452e-07, | |
| "loss": 13.5219, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.89873417721519e-07, | |
| "loss": 13.5798, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.9168173598553345e-07, | |
| "loss": 13.4227, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.9349005424954793e-07, | |
| "loss": 13.603, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.9529837251356238e-07, | |
| "loss": 13.4905, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.9710669077757686e-07, | |
| "loss": 13.6152, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9891500904159131e-07, | |
| "loss": 13.5049, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.007233273056058e-07, | |
| "loss": 13.574, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.0253164556962025e-07, | |
| "loss": 13.5087, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.0433996383363473e-07, | |
| "loss": 13.5461, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.0614828209764918e-07, | |
| "loss": 13.5421, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.0795660036166366e-07, | |
| "loss": 13.5746, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.097649186256781e-07, | |
| "loss": 13.5038, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.115732368896926e-07, | |
| "loss": 13.534, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.1338155515370704e-07, | |
| "loss": 13.4248, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1518987341772152e-07, | |
| "loss": 13.588, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.1699819168173598e-07, | |
| "loss": 13.5216, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 2.1880650994575046e-07, | |
| "loss": 13.539, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 2.206148282097649e-07, | |
| "loss": 13.5719, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 2.224231464737794e-07, | |
| "loss": 13.4977, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.2423146473779384e-07, | |
| "loss": 13.4069, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.2603978300180832e-07, | |
| "loss": 13.533, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.2784810126582277e-07, | |
| "loss": 13.5443, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.2965641952983725e-07, | |
| "loss": 13.5105, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.314647377938517e-07, | |
| "loss": 13.516, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.3327305605786619e-07, | |
| "loss": 13.5084, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.3508137432188064e-07, | |
| "loss": 13.4837, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.3688969258589512e-07, | |
| "loss": 13.5951, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.3869801084990954e-07, | |
| "loss": 13.4639, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.40506329113924e-07, | |
| "loss": 13.4462, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.423146473779385e-07, | |
| "loss": 13.4621, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.44122965641953e-07, | |
| "loss": 13.5816, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.459312839059674e-07, | |
| "loss": 13.5002, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.477396021699819e-07, | |
| "loss": 13.5648, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.4954792043399637e-07, | |
| "loss": 13.513, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.513562386980108e-07, | |
| "loss": 13.5086, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.5316455696202533e-07, | |
| "loss": 13.5994, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.5497287522603975e-07, | |
| "loss": 13.4693, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.5678119349005423e-07, | |
| "loss": 13.5481, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.5858951175406866e-07, | |
| "loss": 13.4964, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 2.603978300180832e-07, | |
| "loss": 13.4156, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 2.622061482820976e-07, | |
| "loss": 13.5032, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.640144665461121e-07, | |
| "loss": 13.5745, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.658227848101265e-07, | |
| "loss": 13.5202, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 2.6763110307414106e-07, | |
| "loss": 13.4655, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.694394213381555e-07, | |
| "loss": 13.5322, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.7124773960216996e-07, | |
| "loss": 13.5317, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 2.7305605786618444e-07, | |
| "loss": 13.4784, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.748643761301989e-07, | |
| "loss": 13.4989, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.7667269439421335e-07, | |
| "loss": 13.421, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.7848101265822783e-07, | |
| "loss": 13.4547, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 2.802893309222423e-07, | |
| "loss": 13.5269, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.820976491862568e-07, | |
| "loss": 13.4043, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.839059674502712e-07, | |
| "loss": 13.4519, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.857142857142857e-07, | |
| "loss": 13.454, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 2.8752260397830017e-07, | |
| "loss": 13.4185, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.8933092224231465e-07, | |
| "loss": 13.4596, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.911392405063291e-07, | |
| "loss": 13.4783, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.9294755877034356e-07, | |
| "loss": 13.406, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.9475587703435804e-07, | |
| "loss": 13.4561, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.965641952983725e-07, | |
| "loss": 13.4658, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.9837251356238694e-07, | |
| "loss": 13.3853, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 3.001808318264014e-07, | |
| "loss": 13.4351, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.019891500904159e-07, | |
| "loss": 13.4493, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.037974683544304e-07, | |
| "loss": 13.493, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.056057866184448e-07, | |
| "loss": 13.4442, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 3.074141048824593e-07, | |
| "loss": 13.424, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.0922242314647377e-07, | |
| "loss": 13.3822, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.1103074141048825e-07, | |
| "loss": 13.4068, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 3.128390596745027e-07, | |
| "loss": 13.3222, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 3.1464737793851715e-07, | |
| "loss": 13.5169, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.1645569620253163e-07, | |
| "loss": 13.395, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.182640144665461e-07, | |
| "loss": 13.3751, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.2007233273056054e-07, | |
| "loss": 13.4999, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.21880650994575e-07, | |
| "loss": 13.3497, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.236889692585895e-07, | |
| "loss": 13.3344, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 3.25497287522604e-07, | |
| "loss": 13.396, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 3.273056057866184e-07, | |
| "loss": 13.4008, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 3.291139240506329e-07, | |
| "loss": 13.3652, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 3.3092224231464736e-07, | |
| "loss": 13.3697, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 3.3273056057866184e-07, | |
| "loss": 13.2918, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 3.3453887884267627e-07, | |
| "loss": 13.3176, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 3.3634719710669075e-07, | |
| "loss": 13.3925, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 3.3815551537070523e-07, | |
| "loss": 13.374, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 3.399638336347197e-07, | |
| "loss": 13.2872, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.4177215189873413e-07, | |
| "loss": 13.3694, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.435804701627486e-07, | |
| "loss": 13.4143, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 3.453887884267631e-07, | |
| "loss": 13.2812, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 3.4719710669077757e-07, | |
| "loss": 13.2149, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 3.49005424954792e-07, | |
| "loss": 13.3274, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 3.508137432188065e-07, | |
| "loss": 13.2607, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 3.5262206148282096e-07, | |
| "loss": 13.2853, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 3.5443037974683544e-07, | |
| "loss": 13.3039, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 3.5623869801084986e-07, | |
| "loss": 13.2795, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 3.5804701627486434e-07, | |
| "loss": 13.3223, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 3.598553345388788e-07, | |
| "loss": 13.3381, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 3.616636528028933e-07, | |
| "loss": 13.2402, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 3.6347197106690773e-07, | |
| "loss": 13.276, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 3.6528028933092226e-07, | |
| "loss": 13.236, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 3.670886075949367e-07, | |
| "loss": 13.3119, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 3.6889692585895117e-07, | |
| "loss": 13.2363, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 3.707052441229656e-07, | |
| "loss": 13.1797, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 3.7251356238698013e-07, | |
| "loss": 13.2306, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 3.7432188065099455e-07, | |
| "loss": 13.1583, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 3.7613019891500903e-07, | |
| "loss": 13.2315, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 3.7793851717902346e-07, | |
| "loss": 13.1212, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 3.79746835443038e-07, | |
| "loss": 13.2316, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 3.815551537070524e-07, | |
| "loss": 13.2637, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 3.833634719710669e-07, | |
| "loss": 13.2633, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 3.851717902350813e-07, | |
| "loss": 13.2313, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 3.8698010849909586e-07, | |
| "loss": 13.2121, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 3.887884267631103e-07, | |
| "loss": 13.2068, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 3.9059674502712476e-07, | |
| "loss": 13.1658, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 3.924050632911392e-07, | |
| "loss": 13.2193, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 3.942133815551537e-07, | |
| "loss": 13.1056, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 3.9602169981916815e-07, | |
| "loss": 13.1592, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 3.9783001808318263e-07, | |
| "loss": 13.2159, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 3.9963833634719705e-07, | |
| "loss": 13.2012, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 4.014466546112116e-07, | |
| "loss": 13.1279, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 4.03254972875226e-07, | |
| "loss": 13.3183, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 4.050632911392405e-07, | |
| "loss": 13.2495, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 4.068716094032549e-07, | |
| "loss": 13.0892, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 4.0867992766726945e-07, | |
| "loss": 13.1179, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 4.104882459312839e-07, | |
| "loss": 13.0991, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 4.1229656419529836e-07, | |
| "loss": 13.1761, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 4.141048824593128e-07, | |
| "loss": 13.1728, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 4.159132007233273e-07, | |
| "loss": 13.1226, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 4.1772151898734174e-07, | |
| "loss": 13.193, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 4.195298372513562e-07, | |
| "loss": 13.0945, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 4.2133815551537065e-07, | |
| "loss": 13.2452, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 4.231464737793852e-07, | |
| "loss": 13.1626, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 4.249547920433996e-07, | |
| "loss": 12.9736, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 4.267631103074141e-07, | |
| "loss": 13.1536, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 4.285714285714285e-07, | |
| "loss": 13.0694, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 4.3037974683544305e-07, | |
| "loss": 13.1939, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 4.3218806509945747e-07, | |
| "loss": 13.089, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 4.3399638336347195e-07, | |
| "loss": 13.1015, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 4.358047016274864e-07, | |
| "loss": 13.1876, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 4.376130198915009e-07, | |
| "loss": 13.1382, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 4.3942133815551534e-07, | |
| "loss": 13.1007, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 4.412296564195298e-07, | |
| "loss": 13.0365, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 4.4303797468354424e-07, | |
| "loss": 13.0515, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 4.448462929475588e-07, | |
| "loss": 13.0274, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 4.466546112115732e-07, | |
| "loss": 13.0163, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 4.484629294755877e-07, | |
| "loss": 13.1132, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 4.502712477396021e-07, | |
| "loss": 13.0237, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 4.5207956600361664e-07, | |
| "loss": 13.0806, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 4.5388788426763107e-07, | |
| "loss": 13.1263, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 4.5569620253164555e-07, | |
| "loss": 12.9992, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 4.5750452079566003e-07, | |
| "loss": 12.97, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 4.593128390596745e-07, | |
| "loss": 12.9025, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 4.6112115732368893e-07, | |
| "loss": 12.9131, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.629294755877034e-07, | |
| "loss": 12.8724, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 4.647377938517179e-07, | |
| "loss": 12.7369, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 4.6654611211573237e-07, | |
| "loss": 12.6747, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 4.683544303797468e-07, | |
| "loss": 12.7058, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 4.701627486437613e-07, | |
| "loss": 12.5785, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 4.7197106690777576e-07, | |
| "loss": 12.6113, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 4.7377938517179024e-07, | |
| "loss": 12.5997, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 4.7558770343580466e-07, | |
| "loss": 12.4582, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 4.773960216998191e-07, | |
| "loss": 12.4912, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 4.792043399638337e-07, | |
| "loss": 12.4618, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 4.81012658227848e-07, | |
| "loss": 12.4053, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 4.828209764918625e-07, | |
| "loss": 12.1477, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 4.84629294755877e-07, | |
| "loss": 12.2996, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 4.864376130198915e-07, | |
| "loss": 12.3589, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 4.88245931283906e-07, | |
| "loss": 12.2919, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 4.900542495479204e-07, | |
| "loss": 12.2481, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 4.918625678119348e-07, | |
| "loss": 12.2361, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 4.936708860759494e-07, | |
| "loss": 12.1804, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 4.954792043399638e-07, | |
| "loss": 12.0909, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 4.972875226039783e-07, | |
| "loss": 12.2437, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.990958408679927e-07, | |
| "loss": 12.2058, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 4.999997478613401e-07, | |
| "loss": 12.0753, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 4.99997730755113e-07, | |
| "loss": 11.8849, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 4.999936965589338e-07, | |
| "loss": 11.9003, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.99987645305352e-07, | |
| "loss": 12.1372, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 4.999795770431919e-07, | |
| "loss": 11.8718, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 4.999694918375516e-07, | |
| "loss": 11.945, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 4.99957389769803e-07, | |
| "loss": 11.8009, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 4.999432709375907e-07, | |
| "loss": 11.8197, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 4.999271354548316e-07, | |
| "loss": 11.8286, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 4.999089834517138e-07, | |
| "loss": 11.8144, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 4.998888150746957e-07, | |
| "loss": 11.8192, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 4.99866630486504e-07, | |
| "loss": 11.7335, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 4.99842429866134e-07, | |
| "loss": 11.7849, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 4.998162134088466e-07, | |
| "loss": 11.6888, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 4.997879813261676e-07, | |
| "loss": 11.7469, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 4.997577338458857e-07, | |
| "loss": 11.7298, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 4.997254712120507e-07, | |
| "loss": 11.633, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 4.996911936849713e-07, | |
| "loss": 11.7687, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 4.996549015412135e-07, | |
| "loss": 11.628, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 4.996165950735983e-07, | |
| "loss": 11.6068, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 4.995762745911985e-07, | |
| "loss": 11.6469, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 4.995339404193373e-07, | |
| "loss": 11.6589, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 4.994895928995854e-07, | |
| "loss": 11.5968, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 4.994432323897575e-07, | |
| "loss": 11.6637, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 4.993948592639104e-07, | |
| "loss": 11.6312, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 4.993444739123394e-07, | |
| "loss": 11.6247, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 4.992920767415752e-07, | |
| "loss": 11.6806, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 4.992376681743811e-07, | |
| "loss": 11.5952, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 4.991812486497489e-07, | |
| "loss": 11.5932, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 4.991228186228956e-07, | |
| "loss": 11.526, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 4.990623785652603e-07, | |
| "loss": 11.5652, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 4.989999289644991e-07, | |
| "loss": 11.5451, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 4.989354703244829e-07, | |
| "loss": 11.5461, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 4.988690031652916e-07, | |
| "loss": 11.5855, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 4.98800528023211e-07, | |
| "loss": 11.5853, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 4.987300454507285e-07, | |
| "loss": 11.4898, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 4.986575560165277e-07, | |
| "loss": 11.5472, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 4.985830603054849e-07, | |
| "loss": 11.5545, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 4.985065589186638e-07, | |
| "loss": 11.5627, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 4.984280524733107e-07, | |
| "loss": 11.6308, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 4.983475416028494e-07, | |
| "loss": 11.5904, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 4.982650269568766e-07, | |
| "loss": 11.6035, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.981805092011564e-07, | |
| "loss": 11.6106, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 4.980939890176143e-07, | |
| "loss": 11.5128, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 4.980054671043329e-07, | |
| "loss": 11.5708, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 4.979149441755452e-07, | |
| "loss": 11.4375, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.978224209616292e-07, | |
| "loss": 11.6326, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 4.977278982091027e-07, | |
| "loss": 11.6381, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 4.976313766806159e-07, | |
| "loss": 11.5119, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 4.975328571549462e-07, | |
| "loss": 11.594, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.974323404269921e-07, | |
| "loss": 11.5035, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 4.973298273077657e-07, | |
| "loss": 11.5414, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 4.972253186243876e-07, | |
| "loss": 11.5682, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.971188152200791e-07, | |
| "loss": 11.4559, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 4.970103179541556e-07, | |
| "loss": 11.6374, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 4.9689982770202e-07, | |
| "loss": 11.6862, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 4.967873453551557e-07, | |
| "loss": 11.6331, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.966728718211188e-07, | |
| "loss": 11.5836, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 4.965564080235315e-07, | |
| "loss": 11.5568, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.964379549020741e-07, | |
| "loss": 11.4982, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.963175134124775e-07, | |
| "loss": 11.5727, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 4.961950845265162e-07, | |
| "loss": 11.4852, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 4.960706692319991e-07, | |
| "loss": 11.5736, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 4.959442685327627e-07, | |
| "loss": 11.5491, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 4.958158834486628e-07, | |
| "loss": 11.5888, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 4.956855150155657e-07, | |
| "loss": 11.5538, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 4.955531642853403e-07, | |
| "loss": 11.5338, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 4.954188323258498e-07, | |
| "loss": 11.6751, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 4.952825202209426e-07, | |
| "loss": 11.5477, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 4.951442290704437e-07, | |
| "loss": 11.613, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 4.950039599901459e-07, | |
| "loss": 11.5106, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 4.94861714111801e-07, | |
| "loss": 11.5909, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 4.947174925831103e-07, | |
| "loss": 11.6413, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 4.945712965677158e-07, | |
| "loss": 11.6069, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 4.944231272451899e-07, | |
| "loss": 11.5424, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 4.94272985811027e-07, | |
| "loss": 11.6403, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 4.941208734766332e-07, | |
| "loss": 11.4936, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 4.939667914693168e-07, | |
| "loss": 11.5997, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 4.93810741032278e-07, | |
| "loss": 11.5182, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 4.936527234245994e-07, | |
| "loss": 11.6751, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 4.934927399212354e-07, | |
| "loss": 11.5386, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 4.933307918130022e-07, | |
| "loss": 11.5771, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 4.931668804065674e-07, | |
| "loss": 11.5864, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 4.93001007024439e-07, | |
| "loss": 11.4768, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 4.928331730049555e-07, | |
| "loss": 11.5559, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 4.926633797022744e-07, | |
| "loss": 11.5335, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 4.924916284863614e-07, | |
| "loss": 11.6058, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 4.923179207429798e-07, | |
| "loss": 11.5176, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 4.92142257873679e-07, | |
| "loss": 11.5609, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 4.919646412957829e-07, | |
| "loss": 11.7039, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 4.917850724423792e-07, | |
| "loss": 11.5618, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 4.91603552762307e-07, | |
| "loss": 11.5667, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 4.914200837201458e-07, | |
| "loss": 11.5145, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 4.912346667962032e-07, | |
| "loss": 11.4585, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 4.910473034865032e-07, | |
| "loss": 11.5594, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 4.908579953027743e-07, | |
| "loss": 11.6038, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 4.906667437724366e-07, | |
| "loss": 11.6523, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.904735504385906e-07, | |
| "loss": 11.6437, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.902784168600036e-07, | |
| "loss": 11.7565, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.900813446110978e-07, | |
| "loss": 11.6141, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.898823352819375e-07, | |
| "loss": 11.6731, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.896813904782162e-07, | |
| "loss": 11.647, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.894785118212435e-07, | |
| "loss": 11.5461, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.892737009479322e-07, | |
| "loss": 11.6451, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.890669595107853e-07, | |
| "loss": 11.6812, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.888582891778821e-07, | |
| "loss": 11.5704, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.886476916328654e-07, | |
| "loss": 11.5842, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.88435168574927e-07, | |
| "loss": 11.4895, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.882207217187954e-07, | |
| "loss": 11.5317, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.880043527947205e-07, | |
| "loss": 11.5456, | |
| "step": 7720 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.877860635484606e-07, | |
| "loss": 11.6659, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.875658557412676e-07, | |
| "loss": 11.4757, | |
| "step": 7760 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.873437311498736e-07, | |
| "loss": 11.58, | |
| "step": 7780 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.871196915664761e-07, | |
| "loss": 11.7097, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.868937387987233e-07, | |
| "loss": 11.6675, | |
| "step": 7820 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.866658746697001e-07, | |
| "loss": 11.5867, | |
| "step": 7840 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.864361010179128e-07, | |
| "loss": 11.6793, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.862044196972751e-07, | |
| "loss": 11.4478, | |
| "step": 7880 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.859708325770919e-07, | |
| "loss": 11.6853, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.857353415420452e-07, | |
| "loss": 11.4939, | |
| "step": 7920 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 4.854979484921789e-07, | |
| "loss": 11.6505, | |
| "step": 7940 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 4.852586553428828e-07, | |
| "loss": 11.7574, | |
| "step": 7960 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 4.850174640248775e-07, | |
| "loss": 11.5477, | |
| "step": 7980 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 4.847743764841993e-07, | |
| "loss": 11.5399, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 4.845293946821836e-07, | |
| "loss": 11.6456, | |
| "step": 8020 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 4.842825205954495e-07, | |
| "loss": 11.5402, | |
| "step": 8040 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 4.840337562158843e-07, | |
| "loss": 11.6021, | |
| "step": 8060 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 4.837831035506267e-07, | |
| "loss": 11.5701, | |
| "step": 8080 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 4.835305646220509e-07, | |
| "loss": 11.5934, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 4.832761414677502e-07, | |
| "loss": 11.6171, | |
| "step": 8120 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 4.83019836140521e-07, | |
| "loss": 11.5625, | |
| "step": 8140 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 4.827616507083456e-07, | |
| "loss": 11.5389, | |
| "step": 8160 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 4.825015872543758e-07, | |
| "loss": 11.5511, | |
| "step": 8180 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 4.822396478769162e-07, | |
| "loss": 11.597, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 4.819758346894072e-07, | |
| "loss": 11.6375, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 4.817101498204078e-07, | |
| "loss": 11.5426, | |
| "step": 8240 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 4.814425954135785e-07, | |
| "loss": 11.5441, | |
| "step": 8260 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 4.811731736276643e-07, | |
| "loss": 11.5702, | |
| "step": 8280 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 4.809018866364766e-07, | |
| "loss": 11.4933, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.806287366288766e-07, | |
| "loss": 11.559, | |
| "step": 8320 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.803537258087566e-07, | |
| "loss": 11.63, | |
| "step": 8340 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.800768563950231e-07, | |
| "loss": 11.613, | |
| "step": 8360 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 4.797981306215784e-07, | |
| "loss": 11.578, | |
| "step": 8380 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 4.795175507373028e-07, | |
| "loss": 11.6106, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.792351190060363e-07, | |
| "loss": 11.5344, | |
| "step": 8420 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.789508377065603e-07, | |
| "loss": 11.6361, | |
| "step": 8440 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 4.786647091325796e-07, | |
| "loss": 11.5861, | |
| "step": 8460 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 4.783767355927033e-07, | |
| "loss": 11.5699, | |
| "step": 8480 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 4.780869194104268e-07, | |
| "loss": 11.5232, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 4.777952629241122e-07, | |
| "loss": 11.5817, | |
| "step": 8520 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 4.775017684869707e-07, | |
| "loss": 11.4567, | |
| "step": 8540 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 4.772064384670424e-07, | |
| "loss": 11.5982, | |
| "step": 8560 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 4.769092752471778e-07, | |
| "loss": 11.4741, | |
| "step": 8580 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 4.766102812250183e-07, | |
| "loss": 11.4757, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 4.7630945881297746e-07, | |
| "loss": 11.6331, | |
| "step": 8620 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 4.7600681043822044e-07, | |
| "loss": 11.4811, | |
| "step": 8640 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 4.7570233854264564e-07, | |
| "loss": 11.5495, | |
| "step": 8660 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 4.7539604558286395e-07, | |
| "loss": 11.6934, | |
| "step": 8680 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 4.7508793403017976e-07, | |
| "loss": 11.516, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 4.747780063705705e-07, | |
| "loss": 11.6319, | |
| "step": 8720 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 4.744662651046666e-07, | |
| "loss": 11.6796, | |
| "step": 8740 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 4.741527127477317e-07, | |
| "loss": 11.5041, | |
| "step": 8760 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.738373518296421e-07, | |
| "loss": 11.5453, | |
| "step": 8780 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.7352018489486606e-07, | |
| "loss": 11.4937, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 4.732012145024439e-07, | |
| "loss": 11.5725, | |
| "step": 8820 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 4.7288044322596663e-07, | |
| "loss": 11.5179, | |
| "step": 8840 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 4.725578736535562e-07, | |
| "loss": 11.4851, | |
| "step": 8860 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 4.722335083878433e-07, | |
| "loss": 11.4871, | |
| "step": 8880 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 4.7190735004594753e-07, | |
| "loss": 11.5277, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 4.715794012594555e-07, | |
| "loss": 11.5012, | |
| "step": 8920 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 4.712496646744002e-07, | |
| "loss": 11.6283, | |
| "step": 8940 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 4.70918142951239e-07, | |
| "loss": 11.6502, | |
| "step": 8960 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 4.705848387648329e-07, | |
| "loss": 11.5598, | |
| "step": 8980 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 4.702497548044243e-07, | |
| "loss": 11.5682, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 4.699128937736157e-07, | |
| "loss": 11.5341, | |
| "step": 9020 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 4.695742583903478e-07, | |
| "loss": 11.4734, | |
| "step": 9040 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 4.692338513868776e-07, | |
| "loss": 11.502, | |
| "step": 9060 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 4.6889167550975613e-07, | |
| "loss": 11.5905, | |
| "step": 9080 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 4.6854773351980647e-07, | |
| "loss": 11.5555, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 4.682020281921017e-07, | |
| "loss": 11.5055, | |
| "step": 9120 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 4.67854562315942e-07, | |
| "loss": 11.6565, | |
| "step": 9140 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 4.6750533869483257e-07, | |
| "loss": 11.5252, | |
| "step": 9160 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 4.6715436014646077e-07, | |
| "loss": 11.5533, | |
| "step": 9180 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 4.6680162950267356e-07, | |
| "loss": 11.6982, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 4.6644714960945453e-07, | |
| "loss": 11.5075, | |
| "step": 9220 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 4.660909233269009e-07, | |
| "loss": 11.6, | |
| "step": 9240 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 4.657329535292007e-07, | |
| "loss": 11.5212, | |
| "step": 9260 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 4.653732431046092e-07, | |
| "loss": 11.5537, | |
| "step": 9280 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 4.6501179495542585e-07, | |
| "loss": 11.5932, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 4.646486119979709e-07, | |
| "loss": 11.6961, | |
| "step": 9320 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 4.642836971625616e-07, | |
| "loss": 11.5336, | |
| "step": 9340 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 4.639170533934891e-07, | |
| "loss": 11.5924, | |
| "step": 9360 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 4.635486836489938e-07, | |
| "loss": 11.5276, | |
| "step": 9380 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 4.631785909012426e-07, | |
| "loss": 11.5576, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 4.6280677813630397e-07, | |
| "loss": 11.6092, | |
| "step": 9420 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 4.624332483541242e-07, | |
| "loss": 11.6501, | |
| "step": 9440 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 4.6205800456850343e-07, | |
| "loss": 11.5759, | |
| "step": 9460 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 4.6168104980707103e-07, | |
| "loss": 11.5808, | |
| "step": 9480 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 4.6130238711126123e-07, | |
| "loss": 11.5683, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 4.609220195362886e-07, | |
| "loss": 11.626, | |
| "step": 9520 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 4.6053995015112343e-07, | |
| "loss": 11.6191, | |
| "step": 9540 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 4.601561820384671e-07, | |
| "loss": 11.5689, | |
| "step": 9560 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 4.597707182947268e-07, | |
| "loss": 11.524, | |
| "step": 9580 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 4.593835620299911e-07, | |
| "loss": 11.6196, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 4.589947163680041e-07, | |
| "loss": 11.5477, | |
| "step": 9620 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 4.5860418444614133e-07, | |
| "loss": 11.5832, | |
| "step": 9640 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 4.5821196941538334e-07, | |
| "loss": 11.5997, | |
| "step": 9660 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 4.5781807444029075e-07, | |
| "loss": 11.4988, | |
| "step": 9680 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 4.5742250269897884e-07, | |
| "loss": 11.6131, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 4.570252573830918e-07, | |
| "loss": 11.5342, | |
| "step": 9720 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 4.5662634169777674e-07, | |
| "loss": 11.6138, | |
| "step": 9740 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 4.5622575886165826e-07, | |
| "loss": 11.5204, | |
| "step": 9760 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 4.55823512106812e-07, | |
| "loss": 11.475, | |
| "step": 9780 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 4.554196046787392e-07, | |
| "loss": 11.5602, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 4.550140398363398e-07, | |
| "loss": 11.5868, | |
| "step": 9820 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 4.546068208518865e-07, | |
| "loss": 11.5379, | |
| "step": 9840 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 4.5419795101099847e-07, | |
| "loss": 11.6708, | |
| "step": 9860 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 4.537874336126146e-07, | |
| "loss": 11.5728, | |
| "step": 9880 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 4.5337527196896715e-07, | |
| "loss": 11.543, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 4.529614694055546e-07, | |
| "loss": 11.5632, | |
| "step": 9920 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 4.5254602926111533e-07, | |
| "loss": 11.483, | |
| "step": 9940 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 4.521289548876003e-07, | |
| "loss": 11.5291, | |
| "step": 9960 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.517102496501462e-07, | |
| "loss": 11.58, | |
| "step": 9980 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.512899169270481e-07, | |
| "loss": 11.5208, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.508679601097326e-07, | |
| "loss": 11.5706, | |
| "step": 10020 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.504443826027298e-07, | |
| "loss": 11.5883, | |
| "step": 10040 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.5001918782364665e-07, | |
| "loss": 11.5798, | |
| "step": 10060 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.4959237920313877e-07, | |
| "loss": 11.51, | |
| "step": 10080 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.491639601848828e-07, | |
| "loss": 11.6322, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 4.4873393422554894e-07, | |
| "loss": 11.5523, | |
| "step": 10120 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 4.483023047947729e-07, | |
| "loss": 11.6259, | |
| "step": 10140 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 4.478690753751278e-07, | |
| "loss": 11.5645, | |
| "step": 10160 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 4.4743424946209627e-07, | |
| "loss": 11.5416, | |
| "step": 10180 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 4.46997830564042e-07, | |
| "loss": 11.4724, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 4.4655982220218167e-07, | |
| "loss": 11.498, | |
| "step": 10220 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 4.461202279105565e-07, | |
| "loss": 11.5235, | |
| "step": 10240 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 4.4567905123600345e-07, | |
| "loss": 11.5966, | |
| "step": 10260 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 4.4523629573812705e-07, | |
| "loss": 11.613, | |
| "step": 10280 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 4.447919649892704e-07, | |
| "loss": 11.6004, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 4.443460625744865e-07, | |
| "loss": 11.7358, | |
| "step": 10320 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 4.438985920915089e-07, | |
| "loss": 11.5892, | |
| "step": 10340 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 4.434495571507234e-07, | |
| "loss": 11.4706, | |
| "step": 10360 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 4.4299896137513837e-07, | |
| "loss": 11.6245, | |
| "step": 10380 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 4.4254680840035554e-07, | |
| "loss": 11.4879, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 4.42093101874541e-07, | |
| "loss": 11.6351, | |
| "step": 10420 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 4.4163784545839543e-07, | |
| "loss": 11.5064, | |
| "step": 10440 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 4.411810428251248e-07, | |
| "loss": 11.6152, | |
| "step": 10460 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 4.407226976604105e-07, | |
| "loss": 11.7646, | |
| "step": 10480 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 4.402628136623798e-07, | |
| "loss": 11.5066, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 4.3980139454157607e-07, | |
| "loss": 11.5206, | |
| "step": 10520 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 4.393384440209284e-07, | |
| "loss": 11.5648, | |
| "step": 10540 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 4.3887396583572225e-07, | |
| "loss": 11.6255, | |
| "step": 10560 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 4.3840796373356864e-07, | |
| "loss": 11.5703, | |
| "step": 10580 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 4.3794044147437437e-07, | |
| "loss": 11.5165, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 4.3747140283031153e-07, | |
| "loss": 11.4863, | |
| "step": 10620 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 4.3700085158578694e-07, | |
| "loss": 11.4701, | |
| "step": 10640 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 4.365287915374118e-07, | |
| "loss": 11.4714, | |
| "step": 10660 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 4.360552264939712e-07, | |
| "loss": 11.5868, | |
| "step": 10680 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 4.355801602763927e-07, | |
| "loss": 11.5528, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.3510359671771647e-07, | |
| "loss": 11.691, | |
| "step": 10720 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.3462553966306357e-07, | |
| "loss": 11.5223, | |
| "step": 10740 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.341459929696054e-07, | |
| "loss": 11.66, | |
| "step": 10760 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 4.3366496050653235e-07, | |
| "loss": 11.6234, | |
| "step": 10780 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 4.3318244615502254e-07, | |
| "loss": 11.4636, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 4.326984538082108e-07, | |
| "loss": 11.6062, | |
| "step": 10820 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 4.32212987371157e-07, | |
| "loss": 11.5589, | |
| "step": 10840 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 4.3172605076081456e-07, | |
| "loss": 11.5099, | |
| "step": 10860 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 4.312376479059988e-07, | |
| "loss": 11.5904, | |
| "step": 10880 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 4.307477827473556e-07, | |
| "loss": 11.5902, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 4.302564592373292e-07, | |
| "loss": 11.5599, | |
| "step": 10920 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 4.2976368134013033e-07, | |
| "loss": 11.5479, | |
| "step": 10940 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 4.292694530317046e-07, | |
| "loss": 11.4897, | |
| "step": 10960 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 4.2877377829969983e-07, | |
| "loss": 11.5071, | |
| "step": 10980 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 4.2827666114343463e-07, | |
| "loss": 11.5219, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 4.2777810557386534e-07, | |
| "loss": 11.5207, | |
| "step": 11020 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 4.2727811561355423e-07, | |
| "loss": 11.5808, | |
| "step": 11040 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.2677669529663686e-07, | |
| "loss": 11.5662, | |
| "step": 11060 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 4.262738486687895e-07, | |
| "loss": 11.6212, | |
| "step": 11080 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 4.2576957978719636e-07, | |
| "loss": 11.7402, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 4.252638927205172e-07, | |
| "loss": 11.5713, | |
| "step": 11120 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 4.2475679154885443e-07, | |
| "loss": 11.5501, | |
| "step": 11140 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 4.242482803637197e-07, | |
| "loss": 11.6619, | |
| "step": 11160 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 4.237383632680015e-07, | |
| "loss": 11.4901, | |
| "step": 11180 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 4.232270443759319e-07, | |
| "loss": 11.5512, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.2271432781305293e-07, | |
| "loss": 11.712, | |
| "step": 11220 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.222002177161841e-07, | |
| "loss": 11.5573, | |
| "step": 11240 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.216847182333881e-07, | |
| "loss": 11.685, | |
| "step": 11260 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.2116783352393803e-07, | |
| "loss": 11.4884, | |
| "step": 11280 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.2064956775828366e-07, | |
| "loss": 11.5112, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.201299251180176e-07, | |
| "loss": 11.513, | |
| "step": 11320 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.1960890979584155e-07, | |
| "loss": 11.5298, | |
| "step": 11340 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.1908652599553293e-07, | |
| "loss": 11.5555, | |
| "step": 11360 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.1856277793191044e-07, | |
| "loss": 11.5323, | |
| "step": 11380 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.1803766983080006e-07, | |
| "loss": 11.5696, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 4.1751120592900156e-07, | |
| "loss": 11.6017, | |
| "step": 11420 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.169833904742537e-07, | |
| "loss": 11.5497, | |
| "step": 11440 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.164542277252e-07, | |
| "loss": 11.6159, | |
| "step": 11460 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 4.159237219513547e-07, | |
| "loss": 11.5327, | |
| "step": 11480 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 4.153918774330682e-07, | |
| "loss": 11.6313, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 4.1485869846149233e-07, | |
| "loss": 11.5099, | |
| "step": 11520 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 4.1432418933854586e-07, | |
| "loss": 11.5367, | |
| "step": 11540 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 4.1378835437687996e-07, | |
| "loss": 11.5341, | |
| "step": 11560 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 4.132511978998432e-07, | |
| "loss": 11.5623, | |
| "step": 11580 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 4.1271272424144645e-07, | |
| "loss": 11.5277, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 4.1217293774632844e-07, | |
| "loss": 11.5259, | |
| "step": 11620 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 4.116318427697205e-07, | |
| "loss": 11.5667, | |
| "step": 11640 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 4.1108944367741105e-07, | |
| "loss": 11.6065, | |
| "step": 11660 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 4.1054574484571105e-07, | |
| "loss": 11.579, | |
| "step": 11680 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 4.100007506614178e-07, | |
| "loss": 11.7267, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 4.094544655217807e-07, | |
| "loss": 11.5985, | |
| "step": 11720 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 4.0890689383446476e-07, | |
| "loss": 11.5627, | |
| "step": 11740 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 4.083580400175153e-07, | |
| "loss": 11.4979, | |
| "step": 11760 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 4.078079084993227e-07, | |
| "loss": 11.6263, | |
| "step": 11780 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.0725650371858646e-07, | |
| "loss": 11.6158, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.0670383012427877e-07, | |
| "loss": 11.5151, | |
| "step": 11820 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 4.0614989217560983e-07, | |
| "loss": 11.6298, | |
| "step": 11840 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 4.0559469434199077e-07, | |
| "loss": 11.5004, | |
| "step": 11860 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 4.050382411029981e-07, | |
| "loss": 11.5026, | |
| "step": 11880 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 4.044805369483377e-07, | |
| "loss": 11.5553, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 4.0392158637780794e-07, | |
| "loss": 11.5134, | |
| "step": 11920 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 4.0336139390126424e-07, | |
| "loss": 11.5529, | |
| "step": 11940 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 4.027999640385821e-07, | |
| "loss": 11.6092, | |
| "step": 11960 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 4.022373013196206e-07, | |
| "loss": 11.4795, | |
| "step": 11980 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 4.0167341028418655e-07, | |
| "loss": 11.5562, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.0110829548199667e-07, | |
| "loss": 11.5685, | |
| "step": 12020 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.005419614726421e-07, | |
| "loss": 11.5652, | |
| "step": 12040 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 3.999744128255508e-07, | |
| "loss": 11.5517, | |
| "step": 12060 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 3.994056541199511e-07, | |
| "loss": 11.6305, | |
| "step": 12080 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 3.988356899448344e-07, | |
| "loss": 11.6278, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 3.982645248989186e-07, | |
| "loss": 11.6063, | |
| "step": 12120 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 3.9769216359061063e-07, | |
| "loss": 11.6833, | |
| "step": 12140 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 3.971186106379693e-07, | |
| "loss": 11.5823, | |
| "step": 12160 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 3.9654387066866833e-07, | |
| "loss": 11.512, | |
| "step": 12180 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 3.9596794831995863e-07, | |
| "loss": 11.5795, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 3.953908482386311e-07, | |
| "loss": 11.7476, | |
| "step": 12220 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 3.94812575080979e-07, | |
| "loss": 11.5925, | |
| "step": 12240 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 3.9423313351276075e-07, | |
| "loss": 11.6184, | |
| "step": 12260 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 3.9365252820916186e-07, | |
| "loss": 11.5744, | |
| "step": 12280 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 3.930707638547571e-07, | |
| "loss": 11.5405, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 3.924878451434735e-07, | |
| "loss": 11.5852, | |
| "step": 12320 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 3.9190377677855155e-07, | |
| "loss": 11.5914, | |
| "step": 12340 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 3.913185634725077e-07, | |
| "loss": 11.5288, | |
| "step": 12360 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 3.907322099470963e-07, | |
| "loss": 11.6778, | |
| "step": 12380 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 3.9014472093327164e-07, | |
| "loss": 11.5544, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 3.8955610117114946e-07, | |
| "loss": 11.4673, | |
| "step": 12420 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 3.889663554099688e-07, | |
| "loss": 11.4688, | |
| "step": 12440 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 3.883754884080539e-07, | |
| "loss": 11.5434, | |
| "step": 12460 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 3.8778350493277566e-07, | |
| "loss": 11.4583, | |
| "step": 12480 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 3.871904097605131e-07, | |
| "loss": 11.6132, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 3.8659620767661483e-07, | |
| "loss": 11.6718, | |
| "step": 12520 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 3.8600090347536064e-07, | |
| "loss": 11.5192, | |
| "step": 12540 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 3.8540450195992255e-07, | |
| "loss": 11.552, | |
| "step": 12560 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 3.8480700794232634e-07, | |
| "loss": 11.4474, | |
| "step": 12580 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 3.842084262434125e-07, | |
| "loss": 11.5033, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 3.8360876169279734e-07, | |
| "loss": 11.5274, | |
| "step": 12620 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 3.8300801912883414e-07, | |
| "loss": 11.5262, | |
| "step": 12640 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 3.82406203398574e-07, | |
| "loss": 11.4807, | |
| "step": 12660 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 3.81803319357727e-07, | |
| "loss": 11.6133, | |
| "step": 12680 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 3.8119937187062254e-07, | |
| "loss": 11.5197, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 3.8059436581017044e-07, | |
| "loss": 11.5477, | |
| "step": 12720 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 3.7998830605782175e-07, | |
| "loss": 11.5387, | |
| "step": 12740 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 3.7938119750352885e-07, | |
| "loss": 11.5181, | |
| "step": 12760 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 3.787730450457065e-07, | |
| "loss": 11.5785, | |
| "step": 12780 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 3.781638535911922e-07, | |
| "loss": 11.5045, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 3.775536280552063e-07, | |
| "loss": 11.5656, | |
| "step": 12820 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 3.769423733613126e-07, | |
| "loss": 11.5141, | |
| "step": 12840 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 3.76330094441379e-07, | |
| "loss": 11.5804, | |
| "step": 12860 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 3.757167962355365e-07, | |
| "loss": 11.5743, | |
| "step": 12880 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 3.7510248369214093e-07, | |
| "loss": 11.5083, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 3.744871617677319e-07, | |
| "loss": 11.638, | |
| "step": 12920 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 3.73870835426993e-07, | |
| "loss": 11.5635, | |
| "step": 12940 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 3.732535096427123e-07, | |
| "loss": 11.662, | |
| "step": 12960 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 3.7263518939574136e-07, | |
| "loss": 11.5764, | |
| "step": 12980 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 3.720158796749556e-07, | |
| "loss": 11.5007, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 3.713955854772143e-07, | |
| "loss": 11.5644, | |
| "step": 13020 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 3.707743118073195e-07, | |
| "loss": 11.5123, | |
| "step": 13040 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 3.7015206367797627e-07, | |
| "loss": 11.5952, | |
| "step": 13060 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 3.695288461097519e-07, | |
| "loss": 11.5475, | |
| "step": 13080 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 3.6890466413103574e-07, | |
| "loss": 11.5173, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 3.682795227779981e-07, | |
| "loss": 11.5033, | |
| "step": 13120 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 3.6765342709455035e-07, | |
| "loss": 11.5651, | |
| "step": 13140 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 3.670263821323034e-07, | |
| "loss": 11.4879, | |
| "step": 13160 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 3.6639839295052776e-07, | |
| "loss": 11.6856, | |
| "step": 13180 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 3.657694646161119e-07, | |
| "loss": 11.5647, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 3.6513960220352204e-07, | |
| "loss": 11.5979, | |
| "step": 13220 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 3.645088107947609e-07, | |
| "loss": 11.6209, | |
| "step": 13240 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 3.638770954793268e-07, | |
| "loss": 11.5795, | |
| "step": 13260 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 3.632444613541723e-07, | |
| "loss": 11.5411, | |
| "step": 13280 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 3.6261091352366363e-07, | |
| "loss": 11.4905, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 3.6197645709953895e-07, | |
| "loss": 11.4929, | |
| "step": 13320 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 3.613410972008674e-07, | |
| "loss": 11.6179, | |
| "step": 13340 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 3.60704838954008e-07, | |
| "loss": 11.5728, | |
| "step": 13360 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 3.6006768749256755e-07, | |
| "loss": 11.5603, | |
| "step": 13380 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 3.594296479573602e-07, | |
| "loss": 11.5548, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 3.5879072549636494e-07, | |
| "loss": 11.5005, | |
| "step": 13420 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 3.581509252646851e-07, | |
| "loss": 11.5154, | |
| "step": 13440 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 3.5751025242450596e-07, | |
| "loss": 11.6091, | |
| "step": 13460 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 3.568687121450533e-07, | |
| "loss": 11.565, | |
| "step": 13480 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 3.5622630960255215e-07, | |
| "loss": 11.5348, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 3.5558304998018426e-07, | |
| "loss": 11.5913, | |
| "step": 13520 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 3.5493893846804673e-07, | |
| "loss": 11.4063, | |
| "step": 13540 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 3.5429398026311037e-07, | |
| "loss": 11.5915, | |
| "step": 13560 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 3.5364818056917724e-07, | |
| "loss": 11.576, | |
| "step": 13580 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 3.530015445968388e-07, | |
| "loss": 11.6471, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 3.5235407756343416e-07, | |
| "loss": 11.5045, | |
| "step": 13620 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 3.517057846930078e-07, | |
| "loss": 11.5106, | |
| "step": 13640 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 3.510566712162673e-07, | |
| "loss": 11.5982, | |
| "step": 13660 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 3.5040674237054125e-07, | |
| "loss": 11.5696, | |
| "step": 13680 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 3.49756003399737e-07, | |
| "loss": 11.6319, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 3.491044595542985e-07, | |
| "loss": 11.5474, | |
| "step": 13720 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 3.4845211609116354e-07, | |
| "loss": 11.6031, | |
| "step": 13740 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 3.4779897827372164e-07, | |
| "loss": 11.4964, | |
| "step": 13760 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 3.4714505137177163e-07, | |
| "loss": 11.4955, | |
| "step": 13780 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 3.4649034066147894e-07, | |
| "loss": 11.527, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 3.4583485142533303e-07, | |
| "loss": 11.6848, | |
| "step": 13820 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 3.4517858895210493e-07, | |
| "loss": 11.5795, | |
| "step": 13840 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 3.4452155853680454e-07, | |
| "loss": 11.4507, | |
| "step": 13860 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 3.438637654806378e-07, | |
| "loss": 11.4203, | |
| "step": 13880 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 3.432052150909637e-07, | |
| "loss": 11.4583, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 3.4254591268125214e-07, | |
| "loss": 11.4966, | |
| "step": 13920 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 3.418858635710406e-07, | |
| "loss": 11.4683, | |
| "step": 13940 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 3.412250730858909e-07, | |
| "loss": 11.5058, | |
| "step": 13960 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 3.4056354655734686e-07, | |
| "loss": 11.6324, | |
| "step": 13980 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 3.399012893228912e-07, | |
| "loss": 11.5019, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 3.392383067259018e-07, | |
| "loss": 11.6269, | |
| "step": 14020 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 3.3857460411560943e-07, | |
| "loss": 11.4724, | |
| "step": 14040 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 3.379101868470543e-07, | |
| "loss": 11.5622, | |
| "step": 14060 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 3.372450602810426e-07, | |
| "loss": 11.6062, | |
| "step": 14080 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 3.3657922978410335e-07, | |
| "loss": 11.6198, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.3591270072844547e-07, | |
| "loss": 11.5821, | |
| "step": 14120 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 3.3524547849191396e-07, | |
| "loss": 11.662, | |
| "step": 14140 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 3.3457756845794687e-07, | |
| "loss": 11.6006, | |
| "step": 14160 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 3.3390897601553146e-07, | |
| "loss": 11.5646, | |
| "step": 14180 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 3.3323970655916115e-07, | |
| "loss": 11.4826, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 3.325697654887918e-07, | |
| "loss": 11.4648, | |
| "step": 14220 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 3.3189915820979785e-07, | |
| "loss": 11.5364, | |
| "step": 14240 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 3.312278901329295e-07, | |
| "loss": 11.5953, | |
| "step": 14260 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 3.305559666742682e-07, | |
| "loss": 11.6724, | |
| "step": 14280 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 3.298833932551832e-07, | |
| "loss": 11.493, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 3.2921017530228845e-07, | |
| "loss": 11.5222, | |
| "step": 14320 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 3.2853631824739756e-07, | |
| "loss": 11.5317, | |
| "step": 14340 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 3.278618275274814e-07, | |
| "loss": 11.5487, | |
| "step": 14360 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 3.2718670858462296e-07, | |
| "loss": 11.5281, | |
| "step": 14380 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 3.2651096686597423e-07, | |
| "loss": 11.5604, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 3.2583460782371215e-07, | |
| "loss": 11.5536, | |
| "step": 14420 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 3.2515763691499425e-07, | |
| "loss": 11.4892, | |
| "step": 14440 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 3.2448005960191507e-07, | |
| "loss": 11.5586, | |
| "step": 14460 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 3.2380188135146173e-07, | |
| "loss": 11.5617, | |
| "step": 14480 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 3.2312310763547005e-07, | |
| "loss": 11.5167, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 3.224437439305803e-07, | |
| "loss": 11.5548, | |
| "step": 14520 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 3.2176379571819314e-07, | |
| "loss": 11.6007, | |
| "step": 14540 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 3.2108326848442503e-07, | |
| "loss": 11.5821, | |
| "step": 14560 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 3.2040216772006457e-07, | |
| "loss": 11.6232, | |
| "step": 14580 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 3.197204989205276e-07, | |
| "loss": 11.5639, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 3.190382675858131e-07, | |
| "loss": 11.5587, | |
| "step": 14620 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 3.18355479220459e-07, | |
| "loss": 11.6086, | |
| "step": 14640 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 3.1767213933349756e-07, | |
| "loss": 11.5826, | |
| "step": 14660 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 3.1698825343841086e-07, | |
| "loss": 11.517, | |
| "step": 14680 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 3.1630382705308637e-07, | |
| "loss": 11.7103, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 3.156188656997727e-07, | |
| "loss": 11.5983, | |
| "step": 14720 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 3.1493337490503457e-07, | |
| "loss": 11.5032, | |
| "step": 14740 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 3.142473601997086e-07, | |
| "loss": 11.6589, | |
| "step": 14760 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 3.1356082711885846e-07, | |
| "loss": 11.6167, | |
| "step": 14780 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 3.1287378120173045e-07, | |
| "loss": 11.4824, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 3.121862279917084e-07, | |
| "loss": 11.5116, | |
| "step": 14820 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 3.1149817303626947e-07, | |
| "loss": 11.5834, | |
| "step": 14840 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 3.1080962188693907e-07, | |
| "loss": 11.6011, | |
| "step": 14860 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 3.101205800992458e-07, | |
| "loss": 11.6644, | |
| "step": 14880 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 3.0943105323267746e-07, | |
| "loss": 11.6467, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 3.0874104685063515e-07, | |
| "loss": 11.5182, | |
| "step": 14920 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 3.080505665203893e-07, | |
| "loss": 11.604, | |
| "step": 14940 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 3.073596178130342e-07, | |
| "loss": 11.6015, | |
| "step": 14960 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 3.066682063034433e-07, | |
| "loss": 11.476, | |
| "step": 14980 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 3.059763375702241e-07, | |
| "loss": 11.4862, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 3.05284017195673e-07, | |
| "loss": 11.5193, | |
| "step": 15020 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 3.0459125076573063e-07, | |
| "loss": 11.4783, | |
| "step": 15040 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 3.038980438699366e-07, | |
| "loss": 11.5304, | |
| "step": 15060 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 3.0320440210138433e-07, | |
| "loss": 11.6162, | |
| "step": 15080 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 3.0251033105667594e-07, | |
| "loss": 11.5835, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 3.018158363358773e-07, | |
| "loss": 11.4436, | |
| "step": 15120 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 3.0112092354247235e-07, | |
| "loss": 11.5809, | |
| "step": 15140 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 3.004255982833186e-07, | |
| "loss": 11.5485, | |
| "step": 15160 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 2.997298661686014e-07, | |
| "loss": 11.5227, | |
| "step": 15180 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 2.990337328117886e-07, | |
| "loss": 11.474, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 2.983372038295855e-07, | |
| "loss": 11.6214, | |
| "step": 15220 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 2.9764028484188985e-07, | |
| "loss": 11.5108, | |
| "step": 15240 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 2.969429814717456e-07, | |
| "loss": 11.607, | |
| "step": 15260 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 2.9624529934529845e-07, | |
| "loss": 11.4842, | |
| "step": 15280 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 2.955472440917498e-07, | |
| "loss": 11.5665, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 2.948488213433118e-07, | |
| "loss": 11.5481, | |
| "step": 15320 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 2.9415003673516165e-07, | |
| "loss": 11.5603, | |
| "step": 15340 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 2.9345089590539605e-07, | |
| "loss": 11.5924, | |
| "step": 15360 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 2.927514044949861e-07, | |
| "loss": 11.6252, | |
| "step": 15380 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 2.9205156814773143e-07, | |
| "loss": 11.5732, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 2.913513925102146e-07, | |
| "loss": 11.5231, | |
| "step": 15420 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 2.9065088323175594e-07, | |
| "loss": 11.5965, | |
| "step": 15440 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 2.8995004596436774e-07, | |
| "loss": 11.5837, | |
| "step": 15460 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 2.892488863627085e-07, | |
| "loss": 11.5682, | |
| "step": 15480 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 2.8854741008403753e-07, | |
| "loss": 11.5165, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 2.878456227881692e-07, | |
| "loss": 11.5324, | |
| "step": 15520 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 2.871435301374273e-07, | |
| "loss": 11.5491, | |
| "step": 15540 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 2.864411377965995e-07, | |
| "loss": 11.592, | |
| "step": 15560 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 2.8573845143289123e-07, | |
| "loss": 11.5268, | |
| "step": 15580 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 2.850354767158804e-07, | |
| "loss": 11.6144, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 2.843322193174715e-07, | |
| "loss": 11.5777, | |
| "step": 15620 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 2.8362868491184965e-07, | |
| "loss": 11.5478, | |
| "step": 15640 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 2.829248791754353e-07, | |
| "loss": 11.5472, | |
| "step": 15660 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 2.8222080778683766e-07, | |
| "loss": 11.4977, | |
| "step": 15680 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 2.8151647642680976e-07, | |
| "loss": 11.5799, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 2.8081189077820206e-07, | |
| "loss": 11.6459, | |
| "step": 15720 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 2.801070565259165e-07, | |
| "loss": 11.412, | |
| "step": 15740 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 2.7940197935686123e-07, | |
| "loss": 11.6231, | |
| "step": 15760 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 2.78696664959904e-07, | |
| "loss": 11.6703, | |
| "step": 15780 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 2.7799111902582693e-07, | |
| "loss": 11.5974, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 2.7728534724728023e-07, | |
| "loss": 11.5836, | |
| "step": 15820 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 2.7657935531873606e-07, | |
| "loss": 11.5461, | |
| "step": 15840 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 2.758731489364431e-07, | |
| "loss": 11.5596, | |
| "step": 15860 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 2.751667337983803e-07, | |
| "loss": 11.5528, | |
| "step": 15880 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 2.7446011560421087e-07, | |
| "loss": 11.5686, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 2.737533000552363e-07, | |
| "loss": 11.5344, | |
| "step": 15920 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 2.730462928543507e-07, | |
| "loss": 11.5748, | |
| "step": 15940 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 2.7233909970599426e-07, | |
| "loss": 11.5089, | |
| "step": 15960 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 2.716317263161076e-07, | |
| "loss": 11.5718, | |
| "step": 15980 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 2.7092417839208537e-07, | |
| "loss": 11.4656, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 2.7021646164273084e-07, | |
| "loss": 11.5286, | |
| "step": 16020 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 2.695085817782091e-07, | |
| "loss": 11.5443, | |
| "step": 16040 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 2.6880054451000144e-07, | |
| "loss": 11.627, | |
| "step": 16060 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 2.6809235555085923e-07, | |
| "loss": 11.6072, | |
| "step": 16080 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 2.673840206147576e-07, | |
| "loss": 11.605, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 2.666755454168495e-07, | |
| "loss": 11.7521, | |
| "step": 16120 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 2.659669356734198e-07, | |
| "loss": 11.6228, | |
| "step": 16140 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 2.6525819710183867e-07, | |
| "loss": 11.5816, | |
| "step": 16160 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 2.645493354205158e-07, | |
| "loss": 11.6014, | |
| "step": 16180 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 2.638403563488542e-07, | |
| "loss": 11.5542, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 2.6313126560720413e-07, | |
| "loss": 11.547, | |
| "step": 16220 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 2.6242206891681663e-07, | |
| "loss": 11.4594, | |
| "step": 16240 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 2.6171277199979785e-07, | |
| "loss": 11.6306, | |
| "step": 16260 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 2.6100338057906243e-07, | |
| "loss": 11.5441, | |
| "step": 16280 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 2.602939003782875e-07, | |
| "loss": 11.5436, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 2.5958433712186656e-07, | |
| "loss": 11.5212, | |
| "step": 16320 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 2.5887469653486327e-07, | |
| "loss": 11.4528, | |
| "step": 16340 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 2.5816498434296513e-07, | |
| "loss": 11.6732, | |
| "step": 16360 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 2.5745520627243756e-07, | |
| "loss": 11.6237, | |
| "step": 16380 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 2.567453680500774e-07, | |
| "loss": 11.5789, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 2.560354754031667e-07, | |
| "loss": 11.6132, | |
| "step": 16420 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 2.553255340594268e-07, | |
| "loss": 11.547, | |
| "step": 16440 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 2.54615549746972e-07, | |
| "loss": 11.5666, | |
| "step": 16460 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 2.53905528194263e-07, | |
| "loss": 11.6524, | |
| "step": 16480 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 2.5319547513006124e-07, | |
| "loss": 11.6698, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 2.524853962833824e-07, | |
| "loss": 11.5607, | |
| "step": 16520 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 2.5177529738345005e-07, | |
| "loss": 11.5959, | |
| "step": 16540 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 2.510651841596496e-07, | |
| "loss": 11.615, | |
| "step": 16560 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.5035506234148213e-07, | |
| "loss": 11.5368, | |
| "step": 16580 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 2.4964493765851795e-07, | |
| "loss": 11.6266, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 2.4893481584035043e-07, | |
| "loss": 11.5118, | |
| "step": 16620 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 2.4822470261655e-07, | |
| "loss": 11.4948, | |
| "step": 16640 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.475146037166176e-07, | |
| "loss": 11.621, | |
| "step": 16660 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 2.4680452486993874e-07, | |
| "loss": 11.5366, | |
| "step": 16680 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 2.460944718057371e-07, | |
| "loss": 11.7582, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.453844502530281e-07, | |
| "loss": 11.5283, | |
| "step": 16720 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 2.446744659405732e-07, | |
| "loss": 11.572, | |
| "step": 16740 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 2.439645245968333e-07, | |
| "loss": 11.551, | |
| "step": 16760 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 2.432546319499226e-07, | |
| "loss": 11.6219, | |
| "step": 16780 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 2.4254479372756236e-07, | |
| "loss": 11.5473, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 2.4183501565703485e-07, | |
| "loss": 11.6384, | |
| "step": 16820 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 2.4112530346513676e-07, | |
| "loss": 11.639, | |
| "step": 16840 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 2.404156628781335e-07, | |
| "loss": 11.5192, | |
| "step": 16860 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 2.3970609962171255e-07, | |
| "loss": 11.5126, | |
| "step": 16880 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 2.3899661942093755e-07, | |
| "loss": 11.5228, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 2.382872280002022e-07, | |
| "loss": 11.5192, | |
| "step": 16920 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 2.3757793108318337e-07, | |
| "loss": 11.4967, | |
| "step": 16940 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 2.368687343927959e-07, | |
| "loss": 11.5445, | |
| "step": 16960 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 2.361596436511458e-07, | |
| "loss": 11.4839, | |
| "step": 16980 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 2.354506645794842e-07, | |
| "loss": 11.4847, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 2.347418028981614e-07, | |
| "loss": 11.532, | |
| "step": 17020 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 2.3403306432658023e-07, | |
| "loss": 11.5848, | |
| "step": 17040 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 2.3332445458315048e-07, | |
| "loss": 11.6084, | |
| "step": 17060 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 2.3261597938524244e-07, | |
| "loss": 11.5482, | |
| "step": 17080 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 2.3190764444914078e-07, | |
| "loss": 11.5448, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 2.311994554899985e-07, | |
| "loss": 11.4889, | |
| "step": 17120 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 2.3049141822179097e-07, | |
| "loss": 11.5677, | |
| "step": 17140 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 2.2978353835726919e-07, | |
| "loss": 11.5334, | |
| "step": 17160 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 2.290758216079146e-07, | |
| "loss": 11.659, | |
| "step": 17180 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 2.2836827368389245e-07, | |
| "loss": 11.5961, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 2.276609002940057e-07, | |
| "loss": 11.5773, | |
| "step": 17220 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 2.2695370714564925e-07, | |
| "loss": 11.5673, | |
| "step": 17240 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 2.2624669994476368e-07, | |
| "loss": 11.4363, | |
| "step": 17260 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 2.2553988439578914e-07, | |
| "loss": 11.6368, | |
| "step": 17280 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 2.2483326620161975e-07, | |
| "loss": 11.4992, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 2.2412685106355693e-07, | |
| "loss": 11.6233, | |
| "step": 17320 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 2.2342064468126395e-07, | |
| "loss": 11.5791, | |
| "step": 17340 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 2.2271465275271983e-07, | |
| "loss": 11.65, | |
| "step": 17360 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 2.2200888097417302e-07, | |
| "loss": 11.5825, | |
| "step": 17380 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 2.21303335040096e-07, | |
| "loss": 11.5799, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 2.2059802064313882e-07, | |
| "loss": 11.512, | |
| "step": 17420 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 2.1989294347408347e-07, | |
| "loss": 11.4995, | |
| "step": 17440 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 2.1918810922179803e-07, | |
| "loss": 11.5878, | |
| "step": 17460 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 2.1848352357319022e-07, | |
| "loss": 11.55, | |
| "step": 17480 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 2.1777919221316232e-07, | |
| "loss": 11.5182, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 2.1707512082456473e-07, | |
| "loss": 11.5603, | |
| "step": 17520 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 2.1637131508815027e-07, | |
| "loss": 11.482, | |
| "step": 17540 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 2.1566778068252858e-07, | |
| "loss": 11.5986, | |
| "step": 17560 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 2.1496452328411964e-07, | |
| "loss": 11.552, | |
| "step": 17580 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 2.142615485671088e-07, | |
| "loss": 11.6, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 2.135588622034005e-07, | |
| "loss": 11.5519, | |
| "step": 17620 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 2.128564698625726e-07, | |
| "loss": 11.5955, | |
| "step": 17640 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 2.1215437721183074e-07, | |
| "loss": 11.6683, | |
| "step": 17660 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 2.1145258991596245e-07, | |
| "loss": 11.5436, | |
| "step": 17680 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 2.1075111363729154e-07, | |
| "loss": 11.5688, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 2.1004995403563224e-07, | |
| "loss": 11.5736, | |
| "step": 17720 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 2.0934911676824403e-07, | |
| "loss": 11.5119, | |
| "step": 17740 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 2.086486074897854e-07, | |
| "loss": 11.5214, | |
| "step": 17760 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 2.0794843185226865e-07, | |
| "loss": 11.5909, | |
| "step": 17780 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 2.0724859550501393e-07, | |
| "loss": 11.5332, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 2.0654910409460396e-07, | |
| "loss": 11.4678, | |
| "step": 17820 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 2.0584996326483838e-07, | |
| "loss": 11.6677, | |
| "step": 17840 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 2.0515117865668815e-07, | |
| "loss": 11.5631, | |
| "step": 17860 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 2.0445275590825024e-07, | |
| "loss": 11.5571, | |
| "step": 17880 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 2.0375470065470158e-07, | |
| "loss": 11.5439, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 2.0305701852825438e-07, | |
| "loss": 11.5207, | |
| "step": 17920 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 2.0235971515811013e-07, | |
| "loss": 11.5164, | |
| "step": 17940 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 2.016627961704144e-07, | |
| "loss": 11.5409, | |
| "step": 17960 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 2.0096626718821143e-07, | |
| "loss": 11.5803, | |
| "step": 17980 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 2.002701338313987e-07, | |
| "loss": 11.6046, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 1.995744017166814e-07, | |
| "loss": 11.6438, | |
| "step": 18020 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 1.9887907645752765e-07, | |
| "loss": 11.5637, | |
| "step": 18040 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 1.9818416366412275e-07, | |
| "loss": 11.8021, | |
| "step": 18060 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 1.9748966894332404e-07, | |
| "loss": 11.6382, | |
| "step": 18080 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 1.9679559789861575e-07, | |
| "loss": 11.6082, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 1.9610195613006343e-07, | |
| "loss": 11.7247, | |
| "step": 18120 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 1.954087492342694e-07, | |
| "loss": 11.4888, | |
| "step": 18140 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 1.9471598280432705e-07, | |
| "loss": 11.5055, | |
| "step": 18160 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 1.9402366242977592e-07, | |
| "loss": 11.5259, | |
| "step": 18180 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 1.933317936965566e-07, | |
| "loss": 11.5196, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 1.9264038218696576e-07, | |
| "loss": 11.5081, | |
| "step": 18220 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 1.919494334796107e-07, | |
| "loss": 11.6072, | |
| "step": 18240 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 1.9125895314936488e-07, | |
| "loss": 11.5958, | |
| "step": 18260 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 1.905689467673226e-07, | |
| "loss": 11.5092, | |
| "step": 18280 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 1.8987941990075415e-07, | |
| "loss": 11.6284, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 1.8919037811306104e-07, | |
| "loss": 11.5005, | |
| "step": 18320 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 1.885018269637305e-07, | |
| "loss": 11.5771, | |
| "step": 18340 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 1.8781377200829156e-07, | |
| "loss": 11.5173, | |
| "step": 18360 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 1.8712621879826955e-07, | |
| "loss": 11.5931, | |
| "step": 18380 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 1.8643917288114146e-07, | |
| "loss": 11.5598, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 1.8575263980029147e-07, | |
| "loss": 11.5043, | |
| "step": 18420 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 1.8506662509496546e-07, | |
| "loss": 11.6294, | |
| "step": 18440 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 1.8438113430022733e-07, | |
| "loss": 11.5551, | |
| "step": 18460 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 1.8369617294691358e-07, | |
| "loss": 11.4976, | |
| "step": 18480 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 1.8301174656158912e-07, | |
| "loss": 11.5757, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 1.823278606665024e-07, | |
| "loss": 11.5967, | |
| "step": 18520 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 1.81644520779541e-07, | |
| "loss": 11.5046, | |
| "step": 18540 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 1.8096173241418695e-07, | |
| "loss": 11.5908, | |
| "step": 18560 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 1.8027950107947246e-07, | |
| "loss": 11.548, | |
| "step": 18580 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 1.795978322799354e-07, | |
| "loss": 11.4977, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 1.7891673151557492e-07, | |
| "loss": 11.6205, | |
| "step": 18620 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 1.7823620428180692e-07, | |
| "loss": 11.63, | |
| "step": 18640 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 1.775562560694197e-07, | |
| "loss": 11.5467, | |
| "step": 18660 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 1.7687689236452995e-07, | |
| "loss": 11.4762, | |
| "step": 18680 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 1.7619811864853827e-07, | |
| "loss": 11.6024, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 1.75519940398085e-07, | |
| "loss": 11.4667, | |
| "step": 18720 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 1.748423630850058e-07, | |
| "loss": 11.6476, | |
| "step": 18740 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 1.741653921762879e-07, | |
| "loss": 11.6168, | |
| "step": 18760 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 1.734890331340258e-07, | |
| "loss": 11.5395, | |
| "step": 18780 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 1.728132914153771e-07, | |
| "loss": 11.4921, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 1.7213817247251862e-07, | |
| "loss": 11.6036, | |
| "step": 18820 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 1.7146368175260234e-07, | |
| "loss": 11.6459, | |
| "step": 18840 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 1.7078982469771163e-07, | |
| "loss": 11.5501, | |
| "step": 18860 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 1.7011660674481676e-07, | |
| "loss": 11.456, | |
| "step": 18880 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 1.6944403332573185e-07, | |
| "loss": 11.6167, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 1.6877210986707046e-07, | |
| "loss": 11.5092, | |
| "step": 18920 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 1.6810084179020208e-07, | |
| "loss": 11.5915, | |
| "step": 18940 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 1.674302345112083e-07, | |
| "loss": 11.6085, | |
| "step": 18960 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 1.6676029344083885e-07, | |
| "loss": 11.5425, | |
| "step": 18980 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 1.6609102398446852e-07, | |
| "loss": 11.5581, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 1.654224315420531e-07, | |
| "loss": 11.5621, | |
| "step": 19020 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 1.6475452150808597e-07, | |
| "loss": 11.5826, | |
| "step": 19040 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 1.6408729927155453e-07, | |
| "loss": 11.5437, | |
| "step": 19060 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 1.6342077021589669e-07, | |
| "loss": 11.5366, | |
| "step": 19080 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 1.6275493971895743e-07, | |
| "loss": 11.5912, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 1.620898131529457e-07, | |
| "loss": 11.4859, | |
| "step": 19120 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 1.6142539588439052e-07, | |
| "loss": 11.5597, | |
| "step": 19140 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 1.607616932740982e-07, | |
| "loss": 11.4989, | |
| "step": 19160 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 1.600987106771089e-07, | |
| "loss": 11.5415, | |
| "step": 19180 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 1.5943645344265312e-07, | |
| "loss": 11.5759, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 1.5877492691410913e-07, | |
| "loss": 11.6497, | |
| "step": 19220 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 1.5811413642895943e-07, | |
| "loss": 11.5942, | |
| "step": 19240 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 1.5745408731874776e-07, | |
| "loss": 11.6553, | |
| "step": 19260 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 1.5679478490903635e-07, | |
| "loss": 11.5545, | |
| "step": 19280 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 1.5613623451936232e-07, | |
| "loss": 11.5778, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 1.5547844146319544e-07, | |
| "loss": 11.6732, | |
| "step": 19320 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 1.5482141104789504e-07, | |
| "loss": 11.5742, | |
| "step": 19340 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 1.5416514857466695e-07, | |
| "loss": 11.5561, | |
| "step": 19360 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 1.5350965933852104e-07, | |
| "loss": 11.5486, | |
| "step": 19380 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 1.528549486282284e-07, | |
| "loss": 11.5493, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 1.5220102172627837e-07, | |
| "loss": 11.5734, | |
| "step": 19420 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 1.515478839088365e-07, | |
| "loss": 11.5701, | |
| "step": 19440 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 1.5089554044570149e-07, | |
| "loss": 11.5146, | |
| "step": 19460 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.5024399660026294e-07, | |
| "loss": 11.4944, | |
| "step": 19480 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 1.495932576294588e-07, | |
| "loss": 11.5861, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 1.4894332878373276e-07, | |
| "loss": 11.4813, | |
| "step": 19520 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 1.4829421530699222e-07, | |
| "loss": 11.507, | |
| "step": 19540 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 1.4764592243656582e-07, | |
| "loss": 11.6154, | |
| "step": 19560 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "learning_rate": 1.4699845540316123e-07, | |
| "loss": 11.5995, | |
| "step": 19580 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 1.4635181943082284e-07, | |
| "loss": 11.6338, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 1.4570601973688966e-07, | |
| "loss": 11.5402, | |
| "step": 19620 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 1.450610615319533e-07, | |
| "loss": 11.5688, | |
| "step": 19640 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 1.4441695001981585e-07, | |
| "loss": 11.6236, | |
| "step": 19660 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 1.437736903974479e-07, | |
| "loss": 11.6191, | |
| "step": 19680 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 1.4313128785494667e-07, | |
| "loss": 11.6409, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 1.4248974757549415e-07, | |
| "loss": 11.6191, | |
| "step": 19720 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 1.4184907473531496e-07, | |
| "loss": 11.6084, | |
| "step": 19740 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 1.412092745036351e-07, | |
| "loss": 11.5322, | |
| "step": 19760 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 1.405703520426399e-07, | |
| "loss": 11.4722, | |
| "step": 19780 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 1.3993231250743243e-07, | |
| "loss": 11.5774, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 1.3929516104599202e-07, | |
| "loss": 11.4938, | |
| "step": 19820 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 1.386589027991325e-07, | |
| "loss": 11.5817, | |
| "step": 19840 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 1.3802354290046103e-07, | |
| "loss": 11.5777, | |
| "step": 19860 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 1.3738908647633634e-07, | |
| "loss": 11.5889, | |
| "step": 19880 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 1.367555386458276e-07, | |
| "loss": 11.5573, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 1.3612290452067322e-07, | |
| "loss": 11.5237, | |
| "step": 19920 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 1.3549118920523905e-07, | |
| "loss": 11.5466, | |
| "step": 19940 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 1.3486039779647793e-07, | |
| "loss": 11.6006, | |
| "step": 19960 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 1.3423053538388808e-07, | |
| "loss": 11.5048, | |
| "step": 19980 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 1.3360160704947221e-07, | |
| "loss": 11.5967, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 1.329736178676965e-07, | |
| "loss": 11.5465, | |
| "step": 20020 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.323465729054497e-07, | |
| "loss": 11.5456, | |
| "step": 20040 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 1.317204772220019e-07, | |
| "loss": 11.6243, | |
| "step": 20060 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "learning_rate": 1.3109533586896432e-07, | |
| "loss": 11.5195, | |
| "step": 20080 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 1.304711538902481e-07, | |
| "loss": 11.5885, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 1.2984793632202373e-07, | |
| "loss": 11.5375, | |
| "step": 20120 | |
| }, | |
| { | |
| "epoch": 7.28, | |
| "learning_rate": 1.2922568819268054e-07, | |
| "loss": 11.5209, | |
| "step": 20140 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 1.2860441452278574e-07, | |
| "loss": 11.5969, | |
| "step": 20160 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 1.2798412032504437e-07, | |
| "loss": 11.5644, | |
| "step": 20180 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 1.273648106042587e-07, | |
| "loss": 11.4928, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 1.2674649035728768e-07, | |
| "loss": 11.6097, | |
| "step": 20220 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 1.2612916457300687e-07, | |
| "loss": 11.5373, | |
| "step": 20240 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 1.2551283823226812e-07, | |
| "loss": 11.4943, | |
| "step": 20260 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 1.2489751630785905e-07, | |
| "loss": 11.5864, | |
| "step": 20280 | |
| }, | |
| { | |
| "epoch": 7.34, | |
| "learning_rate": 1.2428320376446348e-07, | |
| "loss": 11.4818, | |
| "step": 20300 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 1.2366990555862106e-07, | |
| "loss": 11.5295, | |
| "step": 20320 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 1.2305762663868728e-07, | |
| "loss": 11.6868, | |
| "step": 20340 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 1.2244637194479376e-07, | |
| "loss": 11.5521, | |
| "step": 20360 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 1.2183614640880783e-07, | |
| "loss": 11.5967, | |
| "step": 20380 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 1.2122695495429347e-07, | |
| "loss": 11.7017, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 1.2061880249647113e-07, | |
| "loss": 11.526, | |
| "step": 20420 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 1.2001169394217825e-07, | |
| "loss": 11.5515, | |
| "step": 20440 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 1.1940563418982959e-07, | |
| "loss": 11.6034, | |
| "step": 20460 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 1.1880062812937753e-07, | |
| "loss": 11.6085, | |
| "step": 20480 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 1.1819668064227303e-07, | |
| "loss": 11.5318, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 1.1759379660142597e-07, | |
| "loss": 11.581, | |
| "step": 20520 | |
| }, | |
| { | |
| "epoch": 7.43, | |
| "learning_rate": 1.1699198087116588e-07, | |
| "loss": 11.4248, | |
| "step": 20540 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 1.1639123830720265e-07, | |
| "loss": 11.6121, | |
| "step": 20560 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 1.1579157375658755e-07, | |
| "loss": 11.5246, | |
| "step": 20580 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 1.151929920576737e-07, | |
| "loss": 11.6043, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 1.1459549804007748e-07, | |
| "loss": 11.5643, | |
| "step": 20620 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 1.1399909652463943e-07, | |
| "loss": 11.4834, | |
| "step": 20640 | |
| }, | |
| { | |
| "epoch": 7.47, | |
| "learning_rate": 1.134037923233852e-07, | |
| "loss": 11.5314, | |
| "step": 20660 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 1.1280959023948692e-07, | |
| "loss": 11.5545, | |
| "step": 20680 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 1.122164950672243e-07, | |
| "loss": 11.5737, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 1.1162451159194614e-07, | |
| "loss": 11.5333, | |
| "step": 20720 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "learning_rate": 1.1103364459003126e-07, | |
| "loss": 11.543, | |
| "step": 20740 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 1.1044389882885058e-07, | |
| "loss": 11.657, | |
| "step": 20760 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 1.0985527906672834e-07, | |
| "loss": 11.5753, | |
| "step": 20780 | |
| }, | |
| { | |
| "epoch": 7.52, | |
| "learning_rate": 1.0926779005290365e-07, | |
| "loss": 11.6339, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 1.0868143652749228e-07, | |
| "loss": 11.5556, | |
| "step": 20820 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 1.0809622322144843e-07, | |
| "loss": 11.6215, | |
| "step": 20840 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 1.0751215485652643e-07, | |
| "loss": 11.7268, | |
| "step": 20860 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 1.0692923614524279e-07, | |
| "loss": 11.4722, | |
| "step": 20880 | |
| }, | |
| { | |
| "epoch": 7.56, | |
| "learning_rate": 1.063474717908382e-07, | |
| "loss": 11.5266, | |
| "step": 20900 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 1.0576686648723923e-07, | |
| "loss": 11.5344, | |
| "step": 20920 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 1.0518742491902097e-07, | |
| "loss": 11.6412, | |
| "step": 20940 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 1.0460915176136892e-07, | |
| "loss": 11.5482, | |
| "step": 20960 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 1.0403205168004132e-07, | |
| "loss": 11.5908, | |
| "step": 20980 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 1.0345612933133166e-07, | |
| "loss": 11.5425, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 1.0288138936203067e-07, | |
| "loss": 11.6107, | |
| "step": 21020 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 1.0230783640938936e-07, | |
| "loss": 11.5566, | |
| "step": 21040 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 1.0173547510108136e-07, | |
| "loss": 11.6423, | |
| "step": 21060 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 1.0116431005516557e-07, | |
| "loss": 11.642, | |
| "step": 21080 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 1.0059434588004897e-07, | |
| "loss": 11.6133, | |
| "step": 21100 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 1.0002558717444922e-07, | |
| "loss": 11.5334, | |
| "step": 21120 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 9.945803852735793e-08, | |
| "loss": 11.5411, | |
| "step": 21140 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 9.889170451800332e-08, | |
| "loss": 11.6101, | |
| "step": 21160 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 9.832658971581346e-08, | |
| "loss": 11.4722, | |
| "step": 21180 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 9.776269868037928e-08, | |
| "loss": 11.5552, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 7.67, | |
| "learning_rate": 9.720003596141796e-08, | |
| "loss": 11.6173, | |
| "step": 21220 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 9.663860609873575e-08, | |
| "loss": 11.5692, | |
| "step": 21240 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 9.607841362219207e-08, | |
| "loss": 11.5422, | |
| "step": 21260 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 9.551946305166233e-08, | |
| "loss": 11.6599, | |
| "step": 21280 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 9.496175889700184e-08, | |
| "loss": 11.4644, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 7.71, | |
| "learning_rate": 9.440530565800927e-08, | |
| "loss": 11.5522, | |
| "step": 21320 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 9.385010782439018e-08, | |
| "loss": 11.509, | |
| "step": 21340 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 9.329616987572122e-08, | |
| "loss": 11.5604, | |
| "step": 21360 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 9.274349628141359e-08, | |
| "loss": 11.6646, | |
| "step": 21380 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 9.219209150067725e-08, | |
| "loss": 11.5983, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 9.164195998248469e-08, | |
| "loss": 11.5241, | |
| "step": 21420 | |
| }, | |
| { | |
| "epoch": 7.75, | |
| "learning_rate": 9.109310616553534e-08, | |
| "loss": 11.5351, | |
| "step": 21440 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 9.05455344782193e-08, | |
| "loss": 11.5137, | |
| "step": 21460 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "learning_rate": 8.999924933858219e-08, | |
| "loss": 11.6381, | |
| "step": 21480 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 8.945425515428904e-08, | |
| "loss": 11.5419, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 7.78, | |
| "learning_rate": 8.891055632258892e-08, | |
| "loss": 11.5143, | |
| "step": 21520 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 8.836815723027957e-08, | |
| "loss": 11.5709, | |
| "step": 21540 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 8.78270622536716e-08, | |
| "loss": 11.6181, | |
| "step": 21560 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 8.728727575855363e-08, | |
| "loss": 11.5301, | |
| "step": 21580 | |
| }, | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 8.67488021001569e-08, | |
| "loss": 11.6804, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 8.621164562312003e-08, | |
| "loss": 11.4996, | |
| "step": 21620 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 8.567581066145413e-08, | |
| "loss": 11.546, | |
| "step": 21640 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 8.514130153850768e-08, | |
| "loss": 11.6006, | |
| "step": 21660 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 8.460812256693178e-08, | |
| "loss": 11.627, | |
| "step": 21680 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 8.407627804864523e-08, | |
| "loss": 11.5404, | |
| "step": 21700 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 8.354577227479995e-08, | |
| "loss": 11.6296, | |
| "step": 21720 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "learning_rate": 8.301660952574633e-08, | |
| "loss": 11.5375, | |
| "step": 21740 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 8.24887940709984e-08, | |
| "loss": 11.5262, | |
| "step": 21760 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 8.19623301691999e-08, | |
| "loss": 11.5798, | |
| "step": 21780 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 8.143722206808959e-08, | |
| "loss": 11.5828, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 8.091347400446702e-08, | |
| "loss": 11.5238, | |
| "step": 21820 | |
| }, | |
| { | |
| "epoch": 7.9, | |
| "learning_rate": 8.039109020415838e-08, | |
| "loss": 11.5729, | |
| "step": 21840 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 7.987007488198244e-08, | |
| "loss": 11.5649, | |
| "step": 21860 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 7.935043224171631e-08, | |
| "loss": 11.5577, | |
| "step": 21880 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 7.883216647606192e-08, | |
| "loss": 11.5743, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 7.831528176661189e-08, | |
| "loss": 11.5481, | |
| "step": 21920 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 7.77997822838159e-08, | |
| "loss": 11.5684, | |
| "step": 21940 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 7.728567218694706e-08, | |
| "loss": 11.5378, | |
| "step": 21960 | |
| }, | |
| { | |
| "epoch": 7.95, | |
| "learning_rate": 7.677295562406812e-08, | |
| "loss": 11.5941, | |
| "step": 21980 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 7.626163673199848e-08, | |
| "loss": 11.5268, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 7.57517196362803e-08, | |
| "loss": 11.5047, | |
| "step": 22020 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 7.524320845114557e-08, | |
| "loss": 11.5994, | |
| "step": 22040 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 7.473610727948271e-08, | |
| "loss": 11.5535, | |
| "step": 22060 | |
| }, | |
| { | |
| "epoch": 7.99, | |
| "learning_rate": 7.423042021280369e-08, | |
| "loss": 11.6226, | |
| "step": 22080 | |
| }, | |
| { | |
| "epoch": 7.99, | |
| "learning_rate": 7.372615133121057e-08, | |
| "loss": 11.6075, | |
| "step": 22100 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 7.322330470336313e-08, | |
| "loss": 11.4413, | |
| "step": 22120 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 7.272188438644574e-08, | |
| "loss": 11.5181, | |
| "step": 22140 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 7.222189442613464e-08, | |
| "loss": 11.5709, | |
| "step": 22160 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 7.172333885656542e-08, | |
| "loss": 11.53, | |
| "step": 22180 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 7.122622170030016e-08, | |
| "loss": 11.5248, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 7.073054696829545e-08, | |
| "loss": 11.573, | |
| "step": 22220 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 7.023631865986965e-08, | |
| "loss": 11.6162, | |
| "step": 22240 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 6.97435407626708e-08, | |
| "loss": 11.5448, | |
| "step": 22260 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 6.925221725264436e-08, | |
| "loss": 11.5652, | |
| "step": 22280 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 6.876235209400123e-08, | |
| "loss": 11.6287, | |
| "step": 22300 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 6.827394923918553e-08, | |
| "loss": 11.6471, | |
| "step": 22320 | |
| }, | |
| { | |
| "epoch": 8.08, | |
| "learning_rate": 6.778701262884304e-08, | |
| "loss": 11.5994, | |
| "step": 22340 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 6.730154619178918e-08, | |
| "loss": 11.5979, | |
| "step": 22360 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 6.681755384497748e-08, | |
| "loss": 11.5242, | |
| "step": 22380 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "learning_rate": 6.633503949346775e-08, | |
| "loss": 11.6002, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 6.585400703039465e-08, | |
| "loss": 11.4928, | |
| "step": 22420 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 6.537446033693645e-08, | |
| "loss": 11.5, | |
| "step": 22440 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 6.489640328228354e-08, | |
| "loss": 11.453, | |
| "step": 22460 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "learning_rate": 6.441983972360729e-08, | |
| "loss": 11.5869, | |
| "step": 22480 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 6.39447735060288e-08, | |
| "loss": 11.5977, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 8.14, | |
| "learning_rate": 6.347120846258818e-08, | |
| "loss": 11.6181, | |
| "step": 22520 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 6.299914841421309e-08, | |
| "loss": 11.585, | |
| "step": 22540 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "learning_rate": 6.25285971696885e-08, | |
| "loss": 11.6489, | |
| "step": 22560 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 6.20595585256256e-08, | |
| "loss": 11.5211, | |
| "step": 22580 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 6.159203626643137e-08, | |
| "loss": 11.6789, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 6.112603416427776e-08, | |
| "loss": 11.5727, | |
| "step": 22620 | |
| }, | |
| { | |
| "epoch": 8.19, | |
| "learning_rate": 6.066155597907157e-08, | |
| "loss": 11.5466, | |
| "step": 22640 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 6.019860545842392e-08, | |
| "loss": 11.6394, | |
| "step": 22660 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 5.973718633762015e-08, | |
| "loss": 11.4351, | |
| "step": 22680 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 5.927730233958947e-08, | |
| "loss": 11.5098, | |
| "step": 22700 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 5.881895717487523e-08, | |
| "loss": 11.5179, | |
| "step": 22720 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 5.836215454160453e-08, | |
| "loss": 11.5002, | |
| "step": 22740 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 5.7906898125458984e-08, | |
| "loss": 11.5277, | |
| "step": 22760 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 5.7453191599644405e-08, | |
| "loss": 11.567, | |
| "step": 22780 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 5.700103862486158e-08, | |
| "loss": 11.674, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 5.655044284927657e-08, | |
| "loss": 11.5012, | |
| "step": 22820 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 5.610140790849108e-08, | |
| "loss": 11.4852, | |
| "step": 22840 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 5.5653937425513526e-08, | |
| "loss": 11.5356, | |
| "step": 22860 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 5.520803501072954e-08, | |
| "loss": 11.5343, | |
| "step": 22880 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 5.4763704261872906e-08, | |
| "loss": 11.6046, | |
| "step": 22900 | |
| }, | |
| { | |
| "epoch": 8.29, | |
| "learning_rate": 5.432094876399654e-08, | |
| "loss": 11.5436, | |
| "step": 22920 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 5.387977208944355e-08, | |
| "loss": 11.5449, | |
| "step": 22940 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 5.344017779781834e-08, | |
| "loss": 11.5867, | |
| "step": 22960 | |
| }, | |
| { | |
| "epoch": 8.31, | |
| "learning_rate": 5.3002169435958e-08, | |
| "loss": 11.714, | |
| "step": 22980 | |
| }, | |
| { | |
| "epoch": 8.32, | |
| "learning_rate": 5.2565750537903716e-08, | |
| "loss": 11.5761, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 5.213092462487215e-08, | |
| "loss": 11.5768, | |
| "step": 23020 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 5.1697695205227126e-08, | |
| "loss": 11.724, | |
| "step": 23040 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "learning_rate": 5.1266065774451086e-08, | |
| "loss": 11.5739, | |
| "step": 23060 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 5.0836039815117224e-08, | |
| "loss": 11.4955, | |
| "step": 23080 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 5.040762079686123e-08, | |
| "loss": 11.5552, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 4.9980812176353274e-08, | |
| "loss": 11.5925, | |
| "step": 23120 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 4.955561739727013e-08, | |
| "loss": 11.5741, | |
| "step": 23140 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 4.9132039890267456e-08, | |
| "loss": 11.5445, | |
| "step": 23160 | |
| }, | |
| { | |
| "epoch": 8.38, | |
| "learning_rate": 4.8710083072951904e-08, | |
| "loss": 11.5156, | |
| "step": 23180 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 4.8289750349853834e-08, | |
| "loss": 11.5152, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "learning_rate": 4.7871045112399674e-08, | |
| "loss": 11.4828, | |
| "step": 23220 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 4.745397073888463e-08, | |
| "loss": 11.4675, | |
| "step": 23240 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 4.703853059444543e-08, | |
| "loss": 11.4411, | |
| "step": 23260 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 4.66247280310329e-08, | |
| "loss": 11.5138, | |
| "step": 23280 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 4.621256638738541e-08, | |
| "loss": 11.5501, | |
| "step": 23300 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 4.5802048989001556e-08, | |
| "loss": 11.5232, | |
| "step": 23320 | |
| }, | |
| { | |
| "epoch": 8.44, | |
| "learning_rate": 4.539317914811353e-08, | |
| "loss": 11.5105, | |
| "step": 23340 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 4.498596016366027e-08, | |
| "loss": 11.5695, | |
| "step": 23360 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 4.458039532126082e-08, | |
| "loss": 11.6165, | |
| "step": 23380 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "learning_rate": 4.4176487893187956e-08, | |
| "loss": 11.4926, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 8.47, | |
| "learning_rate": 4.3774241138341775e-08, | |
| "loss": 11.6075, | |
| "step": 23420 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 4.3373658302223253e-08, | |
| "loss": 11.6337, | |
| "step": 23440 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 4.29747426169082e-08, | |
| "loss": 11.565, | |
| "step": 23460 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 4.257749730102112e-08, | |
| "loss": 11.6841, | |
| "step": 23480 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 4.218192555970923e-08, | |
| "loss": 11.5039, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 4.1788030584616634e-08, | |
| "loss": 11.539, | |
| "step": 23520 | |
| }, | |
| { | |
| "epoch": 8.51, | |
| "learning_rate": 4.1395815553858624e-08, | |
| "loss": 11.5355, | |
| "step": 23540 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 4.1005283631995816e-08, | |
| "loss": 11.6099, | |
| "step": 23560 | |
| }, | |
| { | |
| "epoch": 8.53, | |
| "learning_rate": 4.061643797000894e-08, | |
| "loss": 11.6437, | |
| "step": 23580 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 4.022928170527315e-08, | |
| "loss": 11.5887, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 3.984381796153288e-08, | |
| "loss": 11.4992, | |
| "step": 23620 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 3.946004984887652e-08, | |
| "loss": 11.5405, | |
| "step": 23640 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 3.9077980463711384e-08, | |
| "loss": 11.6498, | |
| "step": 23660 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 3.869761288873882e-08, | |
| "loss": 11.6689, | |
| "step": 23680 | |
| }, | |
| { | |
| "epoch": 8.57, | |
| "learning_rate": 3.831895019292897e-08, | |
| "loss": 11.5665, | |
| "step": 23700 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 3.7941995431496535e-08, | |
| "loss": 11.5487, | |
| "step": 23720 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 3.7566751645875776e-08, | |
| "loss": 11.5458, | |
| "step": 23740 | |
| }, | |
| { | |
| "epoch": 8.59, | |
| "learning_rate": 3.7193221863696026e-08, | |
| "loss": 11.5934, | |
| "step": 23760 | |
| }, | |
| { | |
| "epoch": 8.6, | |
| "learning_rate": 3.6821409098757336e-08, | |
| "loss": 11.5329, | |
| "step": 23780 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 3.6451316351006153e-08, | |
| "loss": 11.4771, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 3.6082946606510956e-08, | |
| "loss": 11.5533, | |
| "step": 23820 | |
| }, | |
| { | |
| "epoch": 8.62, | |
| "learning_rate": 3.571630283743837e-08, | |
| "loss": 11.4931, | |
| "step": 23840 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 3.535138800202911e-08, | |
| "loss": 11.5671, | |
| "step": 23860 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 3.498820504457414e-08, | |
| "loss": 11.5643, | |
| "step": 23880 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 3.4626756895390824e-08, | |
| "loss": 11.507, | |
| "step": 23900 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 3.426704647079928e-08, | |
| "loss": 11.5686, | |
| "step": 23920 | |
| }, | |
| { | |
| "epoch": 8.66, | |
| "learning_rate": 3.390907667309906e-08, | |
| "loss": 11.6062, | |
| "step": 23940 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 3.355285039054545e-08, | |
| "loss": 11.5904, | |
| "step": 23960 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 3.3198370497326405e-08, | |
| "loss": 11.4588, | |
| "step": 23980 | |
| }, | |
| { | |
| "epoch": 8.68, | |
| "learning_rate": 3.284563985353925e-08, | |
| "loss": 11.4522, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 3.249466130516745e-08, | |
| "loss": 11.5236, | |
| "step": 24020 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 3.2145437684058e-08, | |
| "loss": 11.5657, | |
| "step": 24040 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 3.179797180789831e-08, | |
| "loss": 11.6823, | |
| "step": 24060 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 3.14522664801935e-08, | |
| "loss": 11.5531, | |
| "step": 24080 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 3.1108324490243864e-08, | |
| "loss": 11.5316, | |
| "step": 24100 | |
| }, | |
| { | |
| "epoch": 8.72, | |
| "learning_rate": 3.07661486131224e-08, | |
| "loss": 11.6228, | |
| "step": 24120 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 3.0425741609652166e-08, | |
| "loss": 11.5026, | |
| "step": 24140 | |
| }, | |
| { | |
| "epoch": 8.74, | |
| "learning_rate": 3.00871062263843e-08, | |
| "loss": 11.5837, | |
| "step": 24160 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 2.9750245195575703e-08, | |
| "loss": 11.4998, | |
| "step": 24180 | |
| }, | |
| { | |
| "epoch": 8.75, | |
| "learning_rate": 2.94151612351671e-08, | |
| "loss": 11.5767, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 2.908185704876101e-08, | |
| "loss": 11.5251, | |
| "step": 24220 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 2.875033532559984e-08, | |
| "loss": 11.5134, | |
| "step": 24240 | |
| }, | |
| { | |
| "epoch": 8.77, | |
| "learning_rate": 2.8420598740544476e-08, | |
| "loss": 11.5617, | |
| "step": 24260 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 2.8092649954052473e-08, | |
| "loss": 11.6258, | |
| "step": 24280 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 2.7766491612156663e-08, | |
| "loss": 11.563, | |
| "step": 24300 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 2.74421263464438e-08, | |
| "loss": 11.5837, | |
| "step": 24320 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 2.7119556774033327e-08, | |
| "loss": 11.554, | |
| "step": 24340 | |
| }, | |
| { | |
| "epoch": 8.81, | |
| "learning_rate": 2.679878549755618e-08, | |
| "loss": 11.5175, | |
| "step": 24360 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 2.6479815105133974e-08, | |
| "loss": 11.5168, | |
| "step": 24380 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 2.616264817035793e-08, | |
| "loss": 11.5535, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 8.83, | |
| "learning_rate": 2.5847287252268228e-08, | |
| "loss": 11.5774, | |
| "step": 24420 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 2.5533734895333363e-08, | |
| "loss": 11.6685, | |
| "step": 24440 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 2.5221993629429505e-08, | |
| "loss": 11.6426, | |
| "step": 24460 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 2.4912065969820206e-08, | |
| "loss": 11.6459, | |
| "step": 24480 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 2.4603954417136e-08, | |
| "loss": 11.5111, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 8.87, | |
| "learning_rate": 2.4297661457354346e-08, | |
| "loss": 11.5756, | |
| "step": 24520 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 2.3993189561779537e-08, | |
| "loss": 11.508, | |
| "step": 24540 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 2.3690541187022545e-08, | |
| "loss": 11.5202, | |
| "step": 24560 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 2.338971877498161e-08, | |
| "loss": 11.6627, | |
| "step": 24580 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 2.30907247528222e-08, | |
| "loss": 11.6358, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 8.9, | |
| "learning_rate": 2.2793561532957555e-08, | |
| "loss": 11.5407, | |
| "step": 24620 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 2.2498231513029236e-08, | |
| "loss": 11.6465, | |
| "step": 24640 | |
| }, | |
| { | |
| "epoch": 8.92, | |
| "learning_rate": 2.220473707588777e-08, | |
| "loss": 11.5609, | |
| "step": 24660 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 2.191308058957328e-08, | |
| "loss": 11.47, | |
| "step": 24680 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 2.1623264407296642e-08, | |
| "loss": 11.5498, | |
| "step": 24700 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 2.1335290867420337e-08, | |
| "loss": 11.5768, | |
| "step": 24720 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 2.1049162293439587e-08, | |
| "loss": 11.5281, | |
| "step": 24740 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 2.0764880993963675e-08, | |
| "loss": 11.5445, | |
| "step": 24760 | |
| }, | |
| { | |
| "epoch": 8.96, | |
| "learning_rate": 2.0482449262697126e-08, | |
| "loss": 11.5208, | |
| "step": 24780 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 2.0201869378421497e-08, | |
| "loss": 11.6154, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 1.9923143604976823e-08, | |
| "loss": 11.6076, | |
| "step": 24820 | |
| }, | |
| { | |
| "epoch": 8.98, | |
| "learning_rate": 1.9646274191243318e-08, | |
| "loss": 11.6224, | |
| "step": 24840 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 1.937126337112338e-08, | |
| "loss": 11.4927, | |
| "step": 24860 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 1.909811336352332e-08, | |
| "loss": 11.5288, | |
| "step": 24880 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 1.882682637233568e-08, | |
| "loss": 11.5851, | |
| "step": 24900 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 1.855740458642141e-08, | |
| "loss": 11.5949, | |
| "step": 24920 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.828985017959217e-08, | |
| "loss": 11.7067, | |
| "step": 24940 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 1.8024165310592754e-08, | |
| "loss": 11.5169, | |
| "step": 24960 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 1.7760352123083798e-08, | |
| "loss": 11.6754, | |
| "step": 24980 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 1.749841274562422e-08, | |
| "loss": 11.5465, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 9.05, | |
| "learning_rate": 1.7238349291654435e-08, | |
| "loss": 11.563, | |
| "step": 25020 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.6980163859479007e-08, | |
| "loss": 11.4924, | |
| "step": 25040 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.6723858532249778e-08, | |
| "loss": 11.4883, | |
| "step": 25060 | |
| }, | |
| { | |
| "epoch": 9.07, | |
| "learning_rate": 1.6469435377949175e-08, | |
| "loss": 11.5703, | |
| "step": 25080 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.6216896449373295e-08, | |
| "loss": 11.5845, | |
| "step": 25100 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.596624378411565e-08, | |
| "loss": 11.5394, | |
| "step": 25120 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 1.5717479404550455e-08, | |
| "loss": 11.4934, | |
| "step": 25140 | |
| }, | |
| { | |
| "epoch": 9.1, | |
| "learning_rate": 1.5470605317816436e-08, | |
| "loss": 11.4668, | |
| "step": 25160 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 1.5225623515800673e-08, | |
| "loss": 11.724, | |
| "step": 25180 | |
| }, | |
| { | |
| "epoch": 9.11, | |
| "learning_rate": 1.4982535975122474e-08, | |
| "loss": 11.5385, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "learning_rate": 1.4741344657117238e-08, | |
| "loss": 11.615, | |
| "step": 25220 | |
| }, | |
| { | |
| "epoch": 9.13, | |
| "learning_rate": 1.4502051507821106e-08, | |
| "loss": 11.5736, | |
| "step": 25240 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 1.4264658457954743e-08, | |
| "loss": 11.5491, | |
| "step": 25260 | |
| }, | |
| { | |
| "epoch": 9.14, | |
| "learning_rate": 1.4029167422908105e-08, | |
| "loss": 11.5916, | |
| "step": 25280 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 1.3795580302724874e-08, | |
| "loss": 11.6231, | |
| "step": 25300 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 1.3563898982087069e-08, | |
| "loss": 11.5331, | |
| "step": 25320 | |
| }, | |
| { | |
| "epoch": 9.16, | |
| "learning_rate": 1.3334125330299928e-08, | |
| "loss": 11.514, | |
| "step": 25340 | |
| }, | |
| { | |
| "epoch": 9.17, | |
| "learning_rate": 1.3106261201276724e-08, | |
| "loss": 11.6518, | |
| "step": 25360 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "learning_rate": 1.2880308433523945e-08, | |
| "loss": 11.6731, | |
| "step": 25380 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 1.2656268850126411e-08, | |
| "loss": 11.49, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 1.243414425873246e-08, | |
| "loss": 11.5648, | |
| "step": 25420 | |
| }, | |
| { | |
| "epoch": 9.2, | |
| "learning_rate": 1.221393645153948e-08, | |
| "loss": 11.6055, | |
| "step": 25440 | |
| }, | |
| { | |
| "epoch": 9.21, | |
| "learning_rate": 1.1995647205279457e-08, | |
| "loss": 11.5351, | |
| "step": 25460 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 1.1779278281204536e-08, | |
| "loss": 11.5114, | |
| "step": 25480 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 1.156483142507289e-08, | |
| "loss": 11.5505, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 9.23, | |
| "learning_rate": 1.135230836713466e-08, | |
| "loss": 11.596, | |
| "step": 25520 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 1.1141710822117872e-08, | |
| "loss": 11.8107, | |
| "step": 25540 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "learning_rate": 1.0933040489214674e-08, | |
| "loss": 11.6673, | |
| "step": 25560 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 1.0726299052067761e-08, | |
| "loss": 11.5999, | |
| "step": 25580 | |
| }, | |
| { | |
| "epoch": 9.26, | |
| "learning_rate": 1.0521488178756532e-08, | |
| "loss": 11.5096, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 1.0318609521783817e-08, | |
| "loss": 11.5442, | |
| "step": 25620 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 1.0117664718062469e-08, | |
| "loss": 11.472, | |
| "step": 25640 | |
| }, | |
| { | |
| "epoch": 9.28, | |
| "learning_rate": 9.918655388902158e-09, | |
| "loss": 11.501, | |
| "step": 25660 | |
| }, | |
| { | |
| "epoch": 9.29, | |
| "learning_rate": 9.721583139996382e-09, | |
| "loss": 11.6153, | |
| "step": 25680 | |
| }, | |
| { | |
| "epoch": 9.29, | |
| "learning_rate": 9.526449561409356e-09, | |
| "loss": 11.5813, | |
| "step": 25700 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 9.333256227563341e-09, | |
| "loss": 11.5631, | |
| "step": 25720 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 9.14200469722573e-09, | |
| "loss": 11.6075, | |
| "step": 25740 | |
| }, | |
| { | |
| "epoch": 9.32, | |
| "learning_rate": 8.952696513496755e-09, | |
| "loss": 11.5646, | |
| "step": 25760 | |
| }, | |
| { | |
| "epoch": 9.32, | |
| "learning_rate": 8.76533320379677e-09, | |
| "loss": 11.6741, | |
| "step": 25780 | |
| }, | |
| { | |
| "epoch": 9.33, | |
| "learning_rate": 8.57991627985416e-09, | |
| "loss": 11.5443, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 8.396447237692921e-09, | |
| "loss": 11.505, | |
| "step": 25820 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 8.214927557620766e-09, | |
| "loss": 11.5103, | |
| "step": 25840 | |
| }, | |
| { | |
| "epoch": 9.35, | |
| "learning_rate": 8.035358704217039e-09, | |
| "loss": 11.5984, | |
| "step": 25860 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "learning_rate": 7.85774212632101e-09, | |
| "loss": 11.5731, | |
| "step": 25880 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 7.682079257020163e-09, | |
| "loss": 11.5109, | |
| "step": 25900 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 7.508371513638618e-09, | |
| "loss": 11.5923, | |
| "step": 25920 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 7.336620297725666e-09, | |
| "loss": 11.5269, | |
| "step": 25940 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "learning_rate": 7.1668269950444784e-09, | |
| "loss": 11.5616, | |
| "step": 25960 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 6.998992975560919e-09, | |
| "loss": 11.5697, | |
| "step": 25980 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 6.833119593432607e-09, | |
| "loss": 11.5153, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 9.41, | |
| "learning_rate": 6.66920818699776e-09, | |
| "loss": 11.5417, | |
| "step": 26020 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 6.50726007876462e-09, | |
| "loss": 11.5555, | |
| "step": 26040 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "learning_rate": 6.347276575400628e-09, | |
| "loss": 11.6729, | |
| "step": 26060 | |
| }, | |
| { | |
| "epoch": 9.43, | |
| "learning_rate": 6.189258967721989e-09, | |
| "loss": 11.5967, | |
| "step": 26080 | |
| }, | |
| { | |
| "epoch": 9.44, | |
| "learning_rate": 6.033208530683204e-09, | |
| "loss": 11.5563, | |
| "step": 26100 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 5.879126523366751e-09, | |
| "loss": 11.6089, | |
| "step": 26120 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 5.727014188972979e-09, | |
| "loss": 11.5541, | |
| "step": 26140 | |
| }, | |
| { | |
| "epoch": 9.46, | |
| "learning_rate": 5.576872754810113e-09, | |
| "loss": 11.5303, | |
| "step": 26160 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 5.428703432284243e-09, | |
| "loss": 11.5068, | |
| "step": 26180 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 5.282507416889625e-09, | |
| "loss": 11.4943, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "learning_rate": 5.138285888199007e-09, | |
| "loss": 11.5372, | |
| "step": 26220 | |
| }, | |
| { | |
| "epoch": 9.49, | |
| "learning_rate": 4.996040009854152e-09, | |
| "loss": 11.5843, | |
| "step": 26240 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 4.855770929556385e-09, | |
| "loss": 11.4974, | |
| "step": 26260 | |
| }, | |
| { | |
| "epoch": 9.5, | |
| "learning_rate": 4.7174797790574264e-09, | |
| "loss": 11.5472, | |
| "step": 26280 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 4.5811676741501496e-09, | |
| "loss": 11.6288, | |
| "step": 26300 | |
| }, | |
| { | |
| "epoch": 9.52, | |
| "learning_rate": 4.446835714659647e-09, | |
| "loss": 11.5523, | |
| "step": 26320 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 4.314484984434319e-09, | |
| "loss": 11.5398, | |
| "step": 26340 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 4.184116551337241e-09, | |
| "loss": 11.5505, | |
| "step": 26360 | |
| }, | |
| { | |
| "epoch": 9.54, | |
| "learning_rate": 4.055731467237283e-09, | |
| "loss": 11.5857, | |
| "step": 26380 | |
| }, | |
| { | |
| "epoch": 9.55, | |
| "learning_rate": 3.929330768000949e-09, | |
| "loss": 11.5889, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 9.56, | |
| "learning_rate": 3.8049154734838275e-09, | |
| "loss": 11.563, | |
| "step": 26420 | |
| }, | |
| { | |
| "epoch": 9.56, | |
| "learning_rate": 3.6824865875224043e-09, | |
| "loss": 11.5881, | |
| "step": 26440 | |
| }, | |
| { | |
| "epoch": 9.57, | |
| "learning_rate": 3.562045097925903e-09, | |
| "loss": 11.614, | |
| "step": 26460 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 3.4435919764684572e-09, | |
| "loss": 11.6156, | |
| "step": 26480 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 3.3271281788811444e-09, | |
| "loss": 11.4977, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 9.59, | |
| "learning_rate": 3.2126546448442704e-09, | |
| "loss": 11.5597, | |
| "step": 26520 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 3.1001722979799306e-09, | |
| "loss": 11.5895, | |
| "step": 26540 | |
| }, | |
| { | |
| "epoch": 9.61, | |
| "learning_rate": 2.989682045844405e-09, | |
| "loss": 11.5509, | |
| "step": 26560 | |
| }, | |
| { | |
| "epoch": 9.61, | |
| "learning_rate": 2.8811847799208868e-09, | |
| "loss": 11.4969, | |
| "step": 26580 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 2.774681375612292e-09, | |
| "loss": 11.6286, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 2.6701726922342126e-09, | |
| "loss": 11.5598, | |
| "step": 26620 | |
| }, | |
| { | |
| "epoch": 9.63, | |
| "learning_rate": 2.5676595730079742e-09, | |
| "loss": 11.5438, | |
| "step": 26640 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 2.4671428450537824e-09, | |
| "loss": 11.6316, | |
| "step": 26660 | |
| }, | |
| { | |
| "epoch": 9.65, | |
| "learning_rate": 2.3686233193841722e-09, | |
| "loss": 11.5677, | |
| "step": 26680 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 2.272101790897346e-09, | |
| "loss": 11.5419, | |
| "step": 26700 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 2.177579038370736e-09, | |
| "loss": 11.6155, | |
| "step": 26720 | |
| }, | |
| { | |
| "epoch": 9.67, | |
| "learning_rate": 2.085055824454868e-09, | |
| "loss": 11.5047, | |
| "step": 26740 | |
| }, | |
| { | |
| "epoch": 9.68, | |
| "learning_rate": 1.994532895667117e-09, | |
| "loss": 11.6007, | |
| "step": 26760 | |
| }, | |
| { | |
| "epoch": 9.69, | |
| "learning_rate": 1.9060109823856583e-09, | |
| "loss": 11.4755, | |
| "step": 26780 | |
| }, | |
| { | |
| "epoch": 9.69, | |
| "learning_rate": 1.8194907988436093e-09, | |
| "loss": 11.5557, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 9.7, | |
| "learning_rate": 1.7349730431233111e-09, | |
| "loss": 11.5062, | |
| "step": 26820 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 1.6524583971505857e-09, | |
| "loss": 11.6292, | |
| "step": 26840 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 1.5719475266893489e-09, | |
| "loss": 11.4641, | |
| "step": 26860 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "learning_rate": 1.493441081336172e-09, | |
| "loss": 11.5438, | |
| "step": 26880 | |
| }, | |
| { | |
| "epoch": 9.73, | |
| "learning_rate": 1.4169396945150346e-09, | |
| "loss": 11.5255, | |
| "step": 26900 | |
| }, | |
| { | |
| "epoch": 9.74, | |
| "learning_rate": 1.3424439834722746e-09, | |
| "loss": 11.5248, | |
| "step": 26920 | |
| }, | |
| { | |
| "epoch": 9.74, | |
| "learning_rate": 1.269954549271507e-09, | |
| "loss": 11.5856, | |
| "step": 26940 | |
| }, | |
| { | |
| "epoch": 9.75, | |
| "learning_rate": 1.199471976788935e-09, | |
| "loss": 11.5466, | |
| "step": 26960 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 1.1309968347084364e-09, | |
| "loss": 11.534, | |
| "step": 26980 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 1.0645296755171229e-09, | |
| "loss": 11.5456, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 1.0000710355008157e-09, | |
| "loss": 11.5889, | |
| "step": 27020 | |
| }, | |
| { | |
| "epoch": 9.78, | |
| "learning_rate": 9.376214347397437e-10, | |
| "loss": 11.5358, | |
| "step": 27040 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 8.771813771042968e-10, | |
| "loss": 11.5742, | |
| "step": 27060 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 8.187513502510846e-10, | |
| "loss": 11.5168, | |
| "step": 27080 | |
| }, | |
| { | |
| "epoch": 9.8, | |
| "learning_rate": 7.62331825618856e-10, | |
| "loss": 11.6472, | |
| "step": 27100 | |
| }, | |
| { | |
| "epoch": 9.81, | |
| "learning_rate": 7.079232584247252e-10, | |
| "loss": 11.6027, | |
| "step": 27120 | |
| }, | |
| { | |
| "epoch": 9.82, | |
| "learning_rate": 6.555260876606183e-10, | |
| "loss": 11.5031, | |
| "step": 27140 | |
| }, | |
| { | |
| "epoch": 9.82, | |
| "learning_rate": 6.051407360895822e-10, | |
| "loss": 11.6224, | |
| "step": 27160 | |
| }, | |
| { | |
| "epoch": 9.83, | |
| "learning_rate": 5.567676102424534e-10, | |
| "loss": 11.4651, | |
| "step": 27180 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "learning_rate": 5.10407100414556e-10, | |
| "loss": 11.5794, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "learning_rate": 4.660595806625645e-10, | |
| "loss": 11.6067, | |
| "step": 27220 | |
| }, | |
| { | |
| "epoch": 9.85, | |
| "learning_rate": 4.2372540880147854e-10, | |
| "loss": 11.5273, | |
| "step": 27240 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 3.834049264017092e-10, | |
| "loss": 11.5379, | |
| "step": 27260 | |
| }, | |
| { | |
| "epoch": 9.87, | |
| "learning_rate": 3.450984587863859e-10, | |
| "loss": 11.5575, | |
| "step": 27280 | |
| }, | |
| { | |
| "epoch": 9.87, | |
| "learning_rate": 3.088063150286924e-10, | |
| "loss": 11.4664, | |
| "step": 27300 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 2.74528787949313e-10, | |
| "loss": 11.6222, | |
| "step": 27320 | |
| }, | |
| { | |
| "epoch": 9.89, | |
| "learning_rate": 2.4226615411424013e-10, | |
| "loss": 11.5381, | |
| "step": 27340 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "learning_rate": 2.1201867383233153e-10, | |
| "loss": 11.6545, | |
| "step": 27360 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "learning_rate": 1.8378659115333984e-10, | |
| "loss": 11.6301, | |
| "step": 27380 | |
| }, | |
| { | |
| "epoch": 9.91, | |
| "learning_rate": 1.5757013386599738e-10, | |
| "loss": 11.4795, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 1.3336951349599e-10, | |
| "loss": 11.6273, | |
| "step": 27420 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 1.1118492530443058e-10, | |
| "loss": 11.4725, | |
| "step": 27440 | |
| }, | |
| { | |
| "epoch": 9.93, | |
| "learning_rate": 9.101654828613803e-11, | |
| "loss": 11.556, | |
| "step": 27460 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 7.286454516833296e-11, | |
| "loss": 11.5383, | |
| "step": 27480 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 5.672906240927755e-11, | |
| "loss": 11.6122, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 9.95, | |
| "learning_rate": 4.261023019697108e-11, | |
| "loss": 11.5628, | |
| "step": 27520 | |
| }, | |
| { | |
| "epoch": 9.96, | |
| "learning_rate": 3.050816244831722e-11, | |
| "loss": 11.6629, | |
| "step": 27540 | |
| }, | |
| { | |
| "epoch": 9.97, | |
| "learning_rate": 2.0422956808013868e-11, | |
| "loss": 11.4987, | |
| "step": 27560 | |
| }, | |
| { | |
| "epoch": 9.97, | |
| "learning_rate": 1.235469464785921e-11, | |
| "loss": 11.5243, | |
| "step": 27580 | |
| }, | |
| { | |
| "epoch": 9.98, | |
| "learning_rate": 6.3034410661133574e-12, | |
| "loss": 11.5907, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 9.99, | |
| "learning_rate": 2.2692448868877334e-12, | |
| "loss": 11.56, | |
| "step": 27620 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 2.521386598119957e-13, | |
| "loss": 11.5704, | |
| "step": 27640 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 27650, | |
| "total_flos": 8.165662585242993e+17, | |
| "train_loss": 11.932003958565849, | |
| "train_runtime": 24504.6855, | |
| "train_samples_per_second": 2.257, | |
| "train_steps_per_second": 1.128 | |
| } | |
| ], | |
| "logging_steps": 20, | |
| "max_steps": 27650, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 10, | |
| "save_steps": 100, | |
| "total_flos": 8.165662585242993e+17, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |