{ "best_metric": null, "best_model_checkpoint": null, "epoch": 81.91420564776891, "global_step": 380000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.01, "learning_rate": 3.233455486096142e-08, "loss": 0.0716, "step": 50 }, { "epoch": 0.02, "learning_rate": 6.466910972192284e-08, "loss": 0.0704, "step": 100 }, { "epoch": 0.03, "learning_rate": 9.700366458288424e-08, "loss": 0.0544, "step": 150 }, { "epoch": 0.04, "learning_rate": 1.2933821944384568e-07, "loss": 0.0418, "step": 200 }, { "epoch": 0.05, "learning_rate": 1.6167277430480708e-07, "loss": 0.0306, "step": 250 }, { "epoch": 0.06, "learning_rate": 1.9400732916576848e-07, "loss": 0.0263, "step": 300 }, { "epoch": 0.08, "learning_rate": 2.263418840267299e-07, "loss": 0.0205, "step": 350 }, { "epoch": 0.09, "learning_rate": 2.5867643888769136e-07, "loss": 0.0191, "step": 400 }, { "epoch": 0.1, "learning_rate": 2.9101099374865273e-07, "loss": 0.0167, "step": 450 }, { "epoch": 0.11, "learning_rate": 3.2334554860961416e-07, "loss": 0.0148, "step": 500 }, { "epoch": 0.12, "learning_rate": 3.556801034705756e-07, "loss": 0.0145, "step": 550 }, { "epoch": 0.13, "learning_rate": 3.8801465833153695e-07, "loss": 0.014, "step": 600 }, { "epoch": 0.14, "learning_rate": 4.203492131924984e-07, "loss": 0.0127, "step": 650 }, { "epoch": 0.15, "learning_rate": 4.526837680534598e-07, "loss": 0.0131, "step": 700 }, { "epoch": 0.16, "learning_rate": 4.850183229144212e-07, "loss": 0.0126, "step": 750 }, { "epoch": 0.17, "learning_rate": 5.173528777753827e-07, "loss": 0.0116, "step": 800 }, { "epoch": 0.18, "learning_rate": 5.49687432636344e-07, "loss": 0.012, "step": 850 }, { "epoch": 0.19, "learning_rate": 5.820219874973055e-07, "loss": 0.0106, "step": 900 }, { "epoch": 0.2, "learning_rate": 6.143565423582669e-07, "loss": 0.0105, "step": 950 }, { "epoch": 0.22, "learning_rate": 6.466910972192283e-07, "loss": 0.0095, "step": 1000 }, { "epoch": 0.23, "learning_rate": 6.790256520801896e-07, "loss": 0.0094, "step": 1050 }, { "epoch": 0.24, "learning_rate": 7.113602069411512e-07, "loss": 0.0094, "step": 1100 }, { "epoch": 0.25, "learning_rate": 7.436947618021126e-07, "loss": 0.0087, "step": 1150 }, { "epoch": 0.26, "learning_rate": 7.760293166630739e-07, "loss": 0.0083, "step": 1200 }, { "epoch": 0.27, "learning_rate": 8.083638715240353e-07, "loss": 0.0085, "step": 1250 }, { "epoch": 0.28, "learning_rate": 8.406984263849968e-07, "loss": 0.0081, "step": 1300 }, { "epoch": 0.29, "learning_rate": 8.730329812459582e-07, "loss": 0.008, "step": 1350 }, { "epoch": 0.3, "learning_rate": 9.053675361069196e-07, "loss": 0.0075, "step": 1400 }, { "epoch": 0.31, "learning_rate": 9.37702090967881e-07, "loss": 0.0072, "step": 1450 }, { "epoch": 0.32, "learning_rate": 9.700366458288424e-07, "loss": 0.0076, "step": 1500 }, { "epoch": 0.33, "learning_rate": 1.002371200689804e-06, "loss": 0.0072, "step": 1550 }, { "epoch": 0.34, "learning_rate": 1.0347057555507654e-06, "loss": 0.007, "step": 1600 }, { "epoch": 0.36, "learning_rate": 1.0670403104117266e-06, "loss": 0.0068, "step": 1650 }, { "epoch": 0.37, "learning_rate": 1.099374865272688e-06, "loss": 0.0066, "step": 1700 }, { "epoch": 0.38, "learning_rate": 1.1317094201336495e-06, "loss": 0.0067, "step": 1750 }, { "epoch": 0.39, "learning_rate": 1.164043974994611e-06, "loss": 0.0065, "step": 1800 }, { "epoch": 0.4, "learning_rate": 1.1963785298555723e-06, "loss": 0.0064, "step": 1850 }, { "epoch": 0.41, "learning_rate": 1.2287130847165338e-06, "loss": 0.0064, "step": 1900 }, { "epoch": 0.42, "learning_rate": 1.2610476395774952e-06, "loss": 0.0065, "step": 1950 }, { "epoch": 0.43, "learning_rate": 1.2933821944384566e-06, "loss": 0.0061, "step": 2000 }, { "epoch": 0.44, "learning_rate": 1.325716749299418e-06, "loss": 0.0059, "step": 2050 }, { "epoch": 0.45, "learning_rate": 1.3580513041603793e-06, "loss": 0.0062, "step": 2100 }, { "epoch": 0.46, "learning_rate": 1.3903858590213407e-06, "loss": 0.006, "step": 2150 }, { "epoch": 0.47, "learning_rate": 1.4227204138823023e-06, "loss": 0.006, "step": 2200 }, { "epoch": 0.49, "learning_rate": 1.4550549687432638e-06, "loss": 0.0062, "step": 2250 }, { "epoch": 0.5, "learning_rate": 1.4873895236042252e-06, "loss": 0.0056, "step": 2300 }, { "epoch": 0.51, "learning_rate": 1.5197240784651864e-06, "loss": 0.0058, "step": 2350 }, { "epoch": 0.52, "learning_rate": 1.5520586333261478e-06, "loss": 0.0056, "step": 2400 }, { "epoch": 0.53, "learning_rate": 1.5843931881871092e-06, "loss": 0.0058, "step": 2450 }, { "epoch": 0.54, "learning_rate": 1.6167277430480707e-06, "loss": 0.0056, "step": 2500 }, { "epoch": 0.55, "learning_rate": 1.6490622979090323e-06, "loss": 0.0055, "step": 2550 }, { "epoch": 0.56, "learning_rate": 1.6813968527699935e-06, "loss": 0.0052, "step": 2600 }, { "epoch": 0.57, "learning_rate": 1.713731407630955e-06, "loss": 0.0052, "step": 2650 }, { "epoch": 0.58, "learning_rate": 1.7460659624919164e-06, "loss": 0.0053, "step": 2700 }, { "epoch": 0.59, "learning_rate": 1.7784005173528778e-06, "loss": 0.0052, "step": 2750 }, { "epoch": 0.6, "learning_rate": 1.8107350722138392e-06, "loss": 0.0051, "step": 2800 }, { "epoch": 0.61, "learning_rate": 1.8430696270748007e-06, "loss": 0.0051, "step": 2850 }, { "epoch": 0.63, "learning_rate": 1.875404181935762e-06, "loss": 0.005, "step": 2900 }, { "epoch": 0.64, "learning_rate": 1.9077387367967233e-06, "loss": 0.005, "step": 2950 }, { "epoch": 0.65, "learning_rate": 1.9400732916576847e-06, "loss": 0.0049, "step": 3000 }, { "epoch": 0.66, "learning_rate": 1.9724078465186466e-06, "loss": 0.0049, "step": 3050 }, { "epoch": 0.67, "learning_rate": 2.004742401379608e-06, "loss": 0.0049, "step": 3100 }, { "epoch": 0.68, "learning_rate": 2.0370769562405694e-06, "loss": 0.0048, "step": 3150 }, { "epoch": 0.69, "learning_rate": 2.069411511101531e-06, "loss": 0.0049, "step": 3200 }, { "epoch": 0.7, "learning_rate": 2.101746065962492e-06, "loss": 0.0049, "step": 3250 }, { "epoch": 0.71, "learning_rate": 2.1340806208234533e-06, "loss": 0.0048, "step": 3300 }, { "epoch": 0.72, "learning_rate": 2.1664151756844147e-06, "loss": 0.0046, "step": 3350 }, { "epoch": 0.73, "learning_rate": 2.198749730545376e-06, "loss": 0.0048, "step": 3400 }, { "epoch": 0.74, "learning_rate": 2.2310842854063376e-06, "loss": 0.0047, "step": 3450 }, { "epoch": 0.75, "learning_rate": 2.263418840267299e-06, "loss": 0.0047, "step": 3500 }, { "epoch": 0.77, "learning_rate": 2.2957533951282604e-06, "loss": 0.0046, "step": 3550 }, { "epoch": 0.78, "learning_rate": 2.328087949989222e-06, "loss": 0.0044, "step": 3600 }, { "epoch": 0.79, "learning_rate": 2.3604225048501833e-06, "loss": 0.0044, "step": 3650 }, { "epoch": 0.8, "learning_rate": 2.3927570597111447e-06, "loss": 0.0045, "step": 3700 }, { "epoch": 0.81, "learning_rate": 2.425091614572106e-06, "loss": 0.0044, "step": 3750 }, { "epoch": 0.82, "learning_rate": 2.4574261694330675e-06, "loss": 0.0045, "step": 3800 }, { "epoch": 0.83, "learning_rate": 2.489760724294029e-06, "loss": 0.0045, "step": 3850 }, { "epoch": 0.84, "learning_rate": 2.5220952791549904e-06, "loss": 0.0044, "step": 3900 }, { "epoch": 0.85, "learning_rate": 2.554429834015952e-06, "loss": 0.0044, "step": 3950 }, { "epoch": 0.86, "learning_rate": 2.5867643888769132e-06, "loss": 0.0043, "step": 4000 }, { "epoch": 0.87, "learning_rate": 2.6190989437378747e-06, "loss": 0.0041, "step": 4050 }, { "epoch": 0.88, "learning_rate": 2.651433498598836e-06, "loss": 0.0044, "step": 4100 }, { "epoch": 0.89, "learning_rate": 2.6837680534597975e-06, "loss": 0.0039, "step": 4150 }, { "epoch": 0.91, "learning_rate": 2.7161026083207585e-06, "loss": 0.0041, "step": 4200 }, { "epoch": 0.92, "learning_rate": 2.74843716318172e-06, "loss": 0.004, "step": 4250 }, { "epoch": 0.93, "learning_rate": 2.7807717180426814e-06, "loss": 0.0039, "step": 4300 }, { "epoch": 0.94, "learning_rate": 2.8131062729036432e-06, "loss": 0.0039, "step": 4350 }, { "epoch": 0.95, "learning_rate": 2.8454408277646047e-06, "loss": 0.0041, "step": 4400 }, { "epoch": 0.96, "learning_rate": 2.877775382625566e-06, "loss": 0.0038, "step": 4450 }, { "epoch": 0.97, "learning_rate": 2.9101099374865275e-06, "loss": 0.0039, "step": 4500 }, { "epoch": 0.98, "learning_rate": 2.942444492347489e-06, "loss": 0.0039, "step": 4550 }, { "epoch": 0.99, "learning_rate": 2.9747790472084504e-06, "loss": 0.0038, "step": 4600 }, { "epoch": 1.0, "learning_rate": 3.0071136020694118e-06, "loss": 0.0039, "step": 4650 }, { "epoch": 1.01, "learning_rate": 3.0394481569303728e-06, "loss": 0.0037, "step": 4700 }, { "epoch": 1.02, "learning_rate": 3.0717827117913342e-06, "loss": 0.0039, "step": 4750 }, { "epoch": 1.03, "learning_rate": 3.1041172666522956e-06, "loss": 0.0038, "step": 4800 }, { "epoch": 1.05, "learning_rate": 3.136451821513257e-06, "loss": 0.0036, "step": 4850 }, { "epoch": 1.06, "learning_rate": 3.1687863763742185e-06, "loss": 0.0036, "step": 4900 }, { "epoch": 1.07, "learning_rate": 3.20112093123518e-06, "loss": 0.0037, "step": 4950 }, { "epoch": 1.08, "learning_rate": 3.2334554860961413e-06, "loss": 0.0035, "step": 5000 }, { "epoch": 1.09, "learning_rate": 3.265790040957103e-06, "loss": 0.0035, "step": 5050 }, { "epoch": 1.1, "learning_rate": 3.2981245958180646e-06, "loss": 0.0035, "step": 5100 }, { "epoch": 1.11, "learning_rate": 3.330459150679026e-06, "loss": 0.0033, "step": 5150 }, { "epoch": 1.12, "learning_rate": 3.362793705539987e-06, "loss": 0.0035, "step": 5200 }, { "epoch": 1.13, "learning_rate": 3.3951282604009485e-06, "loss": 0.0033, "step": 5250 }, { "epoch": 1.14, "learning_rate": 3.42746281526191e-06, "loss": 0.0034, "step": 5300 }, { "epoch": 1.15, "learning_rate": 3.4597973701228713e-06, "loss": 0.0034, "step": 5350 }, { "epoch": 1.16, "learning_rate": 3.4921319249838328e-06, "loss": 0.0033, "step": 5400 }, { "epoch": 1.17, "learning_rate": 3.524466479844794e-06, "loss": 0.0033, "step": 5450 }, { "epoch": 1.19, "learning_rate": 3.5568010347057556e-06, "loss": 0.0034, "step": 5500 }, { "epoch": 1.2, "learning_rate": 3.589135589566717e-06, "loss": 0.0034, "step": 5550 }, { "epoch": 1.21, "learning_rate": 3.6214701444276785e-06, "loss": 0.0032, "step": 5600 }, { "epoch": 1.22, "learning_rate": 3.65380469928864e-06, "loss": 0.0033, "step": 5650 }, { "epoch": 1.23, "learning_rate": 3.6861392541496013e-06, "loss": 0.0032, "step": 5700 }, { "epoch": 1.24, "learning_rate": 3.7184738090105627e-06, "loss": 0.0032, "step": 5750 }, { "epoch": 1.25, "learning_rate": 3.750808363871524e-06, "loss": 0.003, "step": 5800 }, { "epoch": 1.26, "learning_rate": 3.783142918732486e-06, "loss": 0.0031, "step": 5850 }, { "epoch": 1.27, "learning_rate": 3.815477473593447e-06, "loss": 0.003, "step": 5900 }, { "epoch": 1.28, "learning_rate": 3.847812028454408e-06, "loss": 0.003, "step": 5950 }, { "epoch": 1.29, "learning_rate": 3.8801465833153694e-06, "loss": 0.003, "step": 6000 }, { "epoch": 1.3, "learning_rate": 3.912481138176331e-06, "loss": 0.003, "step": 6050 }, { "epoch": 1.31, "learning_rate": 3.944815693037293e-06, "loss": 0.0028, "step": 6100 }, { "epoch": 1.33, "learning_rate": 3.977150247898254e-06, "loss": 0.0029, "step": 6150 }, { "epoch": 1.34, "learning_rate": 4.009484802759216e-06, "loss": 0.003, "step": 6200 }, { "epoch": 1.35, "learning_rate": 4.041819357620177e-06, "loss": 0.0028, "step": 6250 }, { "epoch": 1.36, "learning_rate": 4.074153912481139e-06, "loss": 0.0029, "step": 6300 }, { "epoch": 1.37, "learning_rate": 4.1064884673421e-06, "loss": 0.0028, "step": 6350 }, { "epoch": 1.38, "learning_rate": 4.138823022203062e-06, "loss": 0.0027, "step": 6400 }, { "epoch": 1.39, "learning_rate": 4.171157577064023e-06, "loss": 0.0026, "step": 6450 }, { "epoch": 1.4, "learning_rate": 4.203492131924984e-06, "loss": 0.0027, "step": 6500 }, { "epoch": 1.41, "learning_rate": 4.2358266867859456e-06, "loss": 0.0026, "step": 6550 }, { "epoch": 1.42, "learning_rate": 4.2681612416469066e-06, "loss": 0.0026, "step": 6600 }, { "epoch": 1.43, "learning_rate": 4.300495796507868e-06, "loss": 0.0026, "step": 6650 }, { "epoch": 1.44, "learning_rate": 4.332830351368829e-06, "loss": 0.0025, "step": 6700 }, { "epoch": 1.46, "learning_rate": 4.365164906229791e-06, "loss": 0.0024, "step": 6750 }, { "epoch": 1.47, "learning_rate": 4.397499461090752e-06, "loss": 0.0023, "step": 6800 }, { "epoch": 1.48, "learning_rate": 4.429834015951714e-06, "loss": 0.0024, "step": 6850 }, { "epoch": 1.49, "learning_rate": 4.462168570812675e-06, "loss": 0.0023, "step": 6900 }, { "epoch": 1.5, "learning_rate": 4.494503125673636e-06, "loss": 0.0024, "step": 6950 }, { "epoch": 1.51, "learning_rate": 4.526837680534598e-06, "loss": 0.0023, "step": 7000 }, { "epoch": 1.52, "learning_rate": 4.559172235395559e-06, "loss": 0.0023, "step": 7050 }, { "epoch": 1.53, "learning_rate": 4.591506790256521e-06, "loss": 0.0022, "step": 7100 }, { "epoch": 1.54, "learning_rate": 4.623841345117482e-06, "loss": 0.0022, "step": 7150 }, { "epoch": 1.55, "learning_rate": 4.656175899978444e-06, "loss": 0.0024, "step": 7200 }, { "epoch": 1.56, "learning_rate": 4.688510454839405e-06, "loss": 0.0022, "step": 7250 }, { "epoch": 1.57, "learning_rate": 4.7208450097003665e-06, "loss": 0.0022, "step": 7300 }, { "epoch": 1.58, "learning_rate": 4.753179564561328e-06, "loss": 0.0023, "step": 7350 }, { "epoch": 1.6, "learning_rate": 4.785514119422289e-06, "loss": 0.0022, "step": 7400 }, { "epoch": 1.61, "learning_rate": 4.817848674283251e-06, "loss": 0.0022, "step": 7450 }, { "epoch": 1.62, "learning_rate": 4.850183229144212e-06, "loss": 0.002, "step": 7500 }, { "epoch": 1.63, "learning_rate": 4.882517784005174e-06, "loss": 0.0022, "step": 7550 }, { "epoch": 1.64, "learning_rate": 4.914852338866135e-06, "loss": 0.002, "step": 7600 }, { "epoch": 1.65, "learning_rate": 4.947186893727097e-06, "loss": 0.002, "step": 7650 }, { "epoch": 1.66, "learning_rate": 4.979521448588058e-06, "loss": 0.0021, "step": 7700 }, { "epoch": 1.67, "learning_rate": 5.01185600344902e-06, "loss": 0.002, "step": 7750 }, { "epoch": 1.68, "learning_rate": 5.044190558309981e-06, "loss": 0.002, "step": 7800 }, { "epoch": 1.69, "learning_rate": 5.076525113170943e-06, "loss": 0.0019, "step": 7850 }, { "epoch": 1.7, "learning_rate": 5.108859668031904e-06, "loss": 0.0022, "step": 7900 }, { "epoch": 1.71, "learning_rate": 5.141194222892865e-06, "loss": 0.002, "step": 7950 }, { "epoch": 1.72, "learning_rate": 5.1735287777538265e-06, "loss": 0.0021, "step": 8000 }, { "epoch": 1.74, "learning_rate": 5.2058633326147875e-06, "loss": 0.002, "step": 8050 }, { "epoch": 1.75, "learning_rate": 5.238197887475749e-06, "loss": 0.0019, "step": 8100 }, { "epoch": 1.76, "learning_rate": 5.27053244233671e-06, "loss": 0.0019, "step": 8150 }, { "epoch": 1.77, "learning_rate": 5.302866997197672e-06, "loss": 0.0019, "step": 8200 }, { "epoch": 1.78, "learning_rate": 5.335201552058633e-06, "loss": 0.002, "step": 8250 }, { "epoch": 1.79, "learning_rate": 5.367536106919595e-06, "loss": 0.0017, "step": 8300 }, { "epoch": 1.8, "learning_rate": 5.399870661780556e-06, "loss": 0.0018, "step": 8350 }, { "epoch": 1.81, "learning_rate": 5.432205216641517e-06, "loss": 0.0018, "step": 8400 }, { "epoch": 1.82, "learning_rate": 5.464539771502479e-06, "loss": 0.0019, "step": 8450 }, { "epoch": 1.83, "learning_rate": 5.49687432636344e-06, "loss": 0.0018, "step": 8500 }, { "epoch": 1.84, "learning_rate": 5.529208881224402e-06, "loss": 0.0019, "step": 8550 }, { "epoch": 1.85, "learning_rate": 5.561543436085363e-06, "loss": 0.0017, "step": 8600 }, { "epoch": 1.86, "learning_rate": 5.593877990946325e-06, "loss": 0.0018, "step": 8650 }, { "epoch": 1.88, "learning_rate": 5.6262125458072865e-06, "loss": 0.0019, "step": 8700 }, { "epoch": 1.89, "learning_rate": 5.658547100668248e-06, "loss": 0.0017, "step": 8750 }, { "epoch": 1.9, "learning_rate": 5.690881655529209e-06, "loss": 0.0018, "step": 8800 }, { "epoch": 1.91, "learning_rate": 5.723216210390171e-06, "loss": 0.0016, "step": 8850 }, { "epoch": 1.92, "learning_rate": 5.755550765251132e-06, "loss": 0.0017, "step": 8900 }, { "epoch": 1.93, "learning_rate": 5.787885320112093e-06, "loss": 0.0018, "step": 8950 }, { "epoch": 1.94, "learning_rate": 5.820219874973055e-06, "loss": 0.0018, "step": 9000 }, { "epoch": 1.95, "learning_rate": 5.852554429834016e-06, "loss": 0.0017, "step": 9050 }, { "epoch": 1.96, "learning_rate": 5.884888984694978e-06, "loss": 0.0017, "step": 9100 }, { "epoch": 1.97, "learning_rate": 5.917223539555939e-06, "loss": 0.0017, "step": 9150 }, { "epoch": 1.98, "learning_rate": 5.949558094416901e-06, "loss": 0.0017, "step": 9200 }, { "epoch": 1.99, "learning_rate": 5.981892649277862e-06, "loss": 0.0017, "step": 9250 }, { "epoch": 2.0, "learning_rate": 6.0142272041388236e-06, "loss": 0.0016, "step": 9300 }, { "epoch": 2.02, "learning_rate": 6.0465617589997846e-06, "loss": 0.0016, "step": 9350 }, { "epoch": 2.03, "learning_rate": 6.0788963138607456e-06, "loss": 0.0016, "step": 9400 }, { "epoch": 2.04, "learning_rate": 6.1112308687217074e-06, "loss": 0.0016, "step": 9450 }, { "epoch": 2.05, "learning_rate": 6.1435654235826684e-06, "loss": 0.0016, "step": 9500 }, { "epoch": 2.06, "learning_rate": 6.17589997844363e-06, "loss": 0.0015, "step": 9550 }, { "epoch": 2.07, "learning_rate": 6.208234533304591e-06, "loss": 0.0015, "step": 9600 }, { "epoch": 2.08, "learning_rate": 6.240569088165553e-06, "loss": 0.0017, "step": 9650 }, { "epoch": 2.09, "learning_rate": 6.272903643026514e-06, "loss": 0.0015, "step": 9700 }, { "epoch": 2.1, "learning_rate": 6.305238197887476e-06, "loss": 0.0014, "step": 9750 }, { "epoch": 2.11, "learning_rate": 6.337572752748437e-06, "loss": 0.0016, "step": 9800 }, { "epoch": 2.12, "learning_rate": 6.369907307609398e-06, "loss": 0.0015, "step": 9850 }, { "epoch": 2.13, "learning_rate": 6.40224186247036e-06, "loss": 0.0015, "step": 9900 }, { "epoch": 2.14, "learning_rate": 6.434576417331321e-06, "loss": 0.0015, "step": 9950 }, { "epoch": 2.16, "learning_rate": 6.466910972192283e-06, "loss": 0.0014, "step": 10000 }, { "epoch": 2.17, "learning_rate": 6.4992455270532445e-06, "loss": 0.0015, "step": 10050 }, { "epoch": 2.18, "learning_rate": 6.531580081914206e-06, "loss": 0.0016, "step": 10100 }, { "epoch": 2.19, "learning_rate": 6.563914636775167e-06, "loss": 0.0016, "step": 10150 }, { "epoch": 2.2, "learning_rate": 6.596249191636129e-06, "loss": 0.0014, "step": 10200 }, { "epoch": 2.21, "learning_rate": 6.62858374649709e-06, "loss": 0.0013, "step": 10250 }, { "epoch": 2.22, "learning_rate": 6.660918301358052e-06, "loss": 0.0014, "step": 10300 }, { "epoch": 2.23, "learning_rate": 6.693252856219013e-06, "loss": 0.0014, "step": 10350 }, { "epoch": 2.24, "learning_rate": 6.725587411079974e-06, "loss": 0.0015, "step": 10400 }, { "epoch": 2.25, "learning_rate": 6.757921965940936e-06, "loss": 0.0016, "step": 10450 }, { "epoch": 2.26, "learning_rate": 6.790256520801897e-06, "loss": 0.0013, "step": 10500 }, { "epoch": 2.27, "learning_rate": 6.822591075662859e-06, "loss": 0.0013, "step": 10550 }, { "epoch": 2.28, "learning_rate": 6.85492563052382e-06, "loss": 0.0014, "step": 10600 }, { "epoch": 2.3, "learning_rate": 6.887260185384782e-06, "loss": 0.0014, "step": 10650 }, { "epoch": 2.31, "learning_rate": 6.919594740245743e-06, "loss": 0.0014, "step": 10700 }, { "epoch": 2.32, "learning_rate": 6.9519292951067045e-06, "loss": 0.0014, "step": 10750 }, { "epoch": 2.33, "learning_rate": 6.9842638499676655e-06, "loss": 0.0014, "step": 10800 }, { "epoch": 2.34, "learning_rate": 7.0165984048286265e-06, "loss": 0.0014, "step": 10850 }, { "epoch": 2.35, "learning_rate": 7.048932959689588e-06, "loss": 0.0014, "step": 10900 }, { "epoch": 2.36, "learning_rate": 7.081267514550549e-06, "loss": 0.0013, "step": 10950 }, { "epoch": 2.37, "learning_rate": 7.113602069411511e-06, "loss": 0.0014, "step": 11000 }, { "epoch": 2.38, "learning_rate": 7.145936624272472e-06, "loss": 0.0014, "step": 11050 }, { "epoch": 2.39, "learning_rate": 7.178271179133434e-06, "loss": 0.0014, "step": 11100 }, { "epoch": 2.4, "learning_rate": 7.210605733994395e-06, "loss": 0.0013, "step": 11150 }, { "epoch": 2.41, "learning_rate": 7.242940288855357e-06, "loss": 0.0013, "step": 11200 }, { "epoch": 2.43, "learning_rate": 7.275274843716318e-06, "loss": 0.0015, "step": 11250 }, { "epoch": 2.44, "learning_rate": 7.30760939857728e-06, "loss": 0.0015, "step": 11300 }, { "epoch": 2.45, "learning_rate": 7.339943953438242e-06, "loss": 0.0013, "step": 11350 }, { "epoch": 2.46, "learning_rate": 7.372278508299203e-06, "loss": 0.0012, "step": 11400 }, { "epoch": 2.47, "learning_rate": 7.4046130631601645e-06, "loss": 0.0013, "step": 11450 }, { "epoch": 2.48, "learning_rate": 7.4369476180211255e-06, "loss": 0.0013, "step": 11500 }, { "epoch": 2.49, "learning_rate": 7.469282172882087e-06, "loss": 0.0013, "step": 11550 }, { "epoch": 2.5, "learning_rate": 7.501616727743048e-06, "loss": 0.0013, "step": 11600 }, { "epoch": 2.51, "learning_rate": 7.533951282604009e-06, "loss": 0.0015, "step": 11650 }, { "epoch": 2.52, "learning_rate": 7.566285837464972e-06, "loss": 0.0012, "step": 11700 }, { "epoch": 2.53, "learning_rate": 7.598620392325933e-06, "loss": 0.0013, "step": 11750 }, { "epoch": 2.54, "learning_rate": 7.630954947186893e-06, "loss": 0.0014, "step": 11800 }, { "epoch": 2.55, "learning_rate": 7.663289502047854e-06, "loss": 0.0011, "step": 11850 }, { "epoch": 2.57, "learning_rate": 7.695624056908815e-06, "loss": 0.0012, "step": 11900 }, { "epoch": 2.58, "learning_rate": 7.727958611769778e-06, "loss": 0.0014, "step": 11950 }, { "epoch": 2.59, "learning_rate": 7.760293166630739e-06, "loss": 0.0012, "step": 12000 }, { "epoch": 2.6, "learning_rate": 7.792627721491702e-06, "loss": 0.0013, "step": 12050 }, { "epoch": 2.61, "learning_rate": 7.824962276352663e-06, "loss": 0.0012, "step": 12100 }, { "epoch": 2.62, "learning_rate": 7.857296831213625e-06, "loss": 0.0012, "step": 12150 }, { "epoch": 2.63, "learning_rate": 7.889631386074586e-06, "loss": 0.0013, "step": 12200 }, { "epoch": 2.64, "learning_rate": 7.921965940935547e-06, "loss": 0.0012, "step": 12250 }, { "epoch": 2.65, "learning_rate": 7.954300495796508e-06, "loss": 0.0013, "step": 12300 }, { "epoch": 2.66, "learning_rate": 7.98663505065747e-06, "loss": 0.0011, "step": 12350 }, { "epoch": 2.67, "learning_rate": 8.018969605518432e-06, "loss": 0.0012, "step": 12400 }, { "epoch": 2.68, "learning_rate": 8.051304160379393e-06, "loss": 0.0012, "step": 12450 }, { "epoch": 2.69, "learning_rate": 8.083638715240354e-06, "loss": 0.0013, "step": 12500 }, { "epoch": 2.71, "learning_rate": 8.115973270101315e-06, "loss": 0.0013, "step": 12550 }, { "epoch": 2.72, "learning_rate": 8.148307824962278e-06, "loss": 0.0012, "step": 12600 }, { "epoch": 2.73, "learning_rate": 8.180642379823239e-06, "loss": 0.001, "step": 12650 }, { "epoch": 2.74, "learning_rate": 8.2129769346842e-06, "loss": 0.0013, "step": 12700 }, { "epoch": 2.75, "learning_rate": 8.24531148954516e-06, "loss": 0.0012, "step": 12750 }, { "epoch": 2.76, "learning_rate": 8.277646044406123e-06, "loss": 0.0012, "step": 12800 }, { "epoch": 2.77, "learning_rate": 8.309980599267084e-06, "loss": 0.0011, "step": 12850 }, { "epoch": 2.78, "learning_rate": 8.342315154128045e-06, "loss": 0.0013, "step": 12900 }, { "epoch": 2.79, "learning_rate": 8.374649708989006e-06, "loss": 0.0012, "step": 12950 }, { "epoch": 2.8, "learning_rate": 8.406984263849967e-06, "loss": 0.0012, "step": 13000 }, { "epoch": 2.81, "learning_rate": 8.43931881871093e-06, "loss": 0.0012, "step": 13050 }, { "epoch": 2.82, "learning_rate": 8.471653373571891e-06, "loss": 0.0012, "step": 13100 }, { "epoch": 2.83, "learning_rate": 8.503987928432852e-06, "loss": 0.0011, "step": 13150 }, { "epoch": 2.85, "learning_rate": 8.536322483293813e-06, "loss": 0.0011, "step": 13200 }, { "epoch": 2.86, "learning_rate": 8.568657038154776e-06, "loss": 0.0012, "step": 13250 }, { "epoch": 2.87, "learning_rate": 8.600991593015737e-06, "loss": 0.0012, "step": 13300 }, { "epoch": 2.88, "learning_rate": 8.633326147876698e-06, "loss": 0.0011, "step": 13350 }, { "epoch": 2.89, "learning_rate": 8.665660702737659e-06, "loss": 0.0011, "step": 13400 }, { "epoch": 2.9, "learning_rate": 8.69799525759862e-06, "loss": 0.0011, "step": 13450 }, { "epoch": 2.91, "learning_rate": 8.730329812459583e-06, "loss": 0.0012, "step": 13500 }, { "epoch": 2.92, "learning_rate": 8.762664367320544e-06, "loss": 0.0011, "step": 13550 }, { "epoch": 2.93, "learning_rate": 8.794998922181505e-06, "loss": 0.0012, "step": 13600 }, { "epoch": 2.94, "learning_rate": 8.827333477042466e-06, "loss": 0.0013, "step": 13650 }, { "epoch": 2.95, "learning_rate": 8.859668031903428e-06, "loss": 0.001, "step": 13700 }, { "epoch": 2.96, "learning_rate": 8.89200258676439e-06, "loss": 0.0012, "step": 13750 }, { "epoch": 2.97, "learning_rate": 8.92433714162535e-06, "loss": 0.0011, "step": 13800 }, { "epoch": 2.99, "learning_rate": 8.956671696486311e-06, "loss": 0.0012, "step": 13850 }, { "epoch": 3.0, "learning_rate": 8.989006251347272e-06, "loss": 0.0012, "step": 13900 }, { "epoch": 3.01, "learning_rate": 9.021340806208235e-06, "loss": 0.001, "step": 13950 }, { "epoch": 3.02, "learning_rate": 9.053675361069196e-06, "loss": 0.0009, "step": 14000 }, { "epoch": 3.03, "learning_rate": 9.086009915930157e-06, "loss": 0.0008, "step": 14050 }, { "epoch": 3.04, "learning_rate": 9.118344470791118e-06, "loss": 0.0009, "step": 14100 }, { "epoch": 3.05, "learning_rate": 9.15067902565208e-06, "loss": 0.0009, "step": 14150 }, { "epoch": 3.06, "learning_rate": 9.183013580513042e-06, "loss": 0.001, "step": 14200 }, { "epoch": 3.07, "learning_rate": 9.215348135374003e-06, "loss": 0.001, "step": 14250 }, { "epoch": 3.08, "learning_rate": 9.247682690234964e-06, "loss": 0.0009, "step": 14300 }, { "epoch": 3.09, "learning_rate": 9.280017245095925e-06, "loss": 0.0009, "step": 14350 }, { "epoch": 3.1, "learning_rate": 9.312351799956887e-06, "loss": 0.001, "step": 14400 }, { "epoch": 3.11, "learning_rate": 9.344686354817848e-06, "loss": 0.001, "step": 14450 }, { "epoch": 3.13, "learning_rate": 9.37702090967881e-06, "loss": 0.001, "step": 14500 }, { "epoch": 3.14, "learning_rate": 9.40935546453977e-06, "loss": 0.0009, "step": 14550 }, { "epoch": 3.15, "learning_rate": 9.441690019400733e-06, "loss": 0.0009, "step": 14600 }, { "epoch": 3.16, "learning_rate": 9.474024574261694e-06, "loss": 0.0009, "step": 14650 }, { "epoch": 3.17, "learning_rate": 9.506359129122657e-06, "loss": 0.001, "step": 14700 }, { "epoch": 3.18, "learning_rate": 9.538693683983618e-06, "loss": 0.001, "step": 14750 }, { "epoch": 3.19, "learning_rate": 9.571028238844579e-06, "loss": 0.001, "step": 14800 }, { "epoch": 3.2, "learning_rate": 9.603362793705541e-06, "loss": 0.0011, "step": 14850 }, { "epoch": 3.21, "learning_rate": 9.635697348566502e-06, "loss": 0.0009, "step": 14900 }, { "epoch": 3.22, "learning_rate": 9.668031903427463e-06, "loss": 0.0009, "step": 14950 }, { "epoch": 3.23, "learning_rate": 9.700366458288424e-06, "loss": 0.001, "step": 15000 }, { "epoch": 3.24, "learning_rate": 9.732701013149387e-06, "loss": 0.001, "step": 15050 }, { "epoch": 3.26, "learning_rate": 9.765035568010348e-06, "loss": 0.0011, "step": 15100 }, { "epoch": 3.27, "learning_rate": 9.79737012287131e-06, "loss": 0.001, "step": 15150 }, { "epoch": 3.28, "learning_rate": 9.82970467773227e-06, "loss": 0.0009, "step": 15200 }, { "epoch": 3.29, "learning_rate": 9.862039232593233e-06, "loss": 0.0009, "step": 15250 }, { "epoch": 3.3, "learning_rate": 9.894373787454194e-06, "loss": 0.0009, "step": 15300 }, { "epoch": 3.31, "learning_rate": 9.926708342315155e-06, "loss": 0.001, "step": 15350 }, { "epoch": 3.32, "learning_rate": 9.959042897176116e-06, "loss": 0.0009, "step": 15400 }, { "epoch": 3.33, "learning_rate": 9.991377452037077e-06, "loss": 0.0009, "step": 15450 }, { "epoch": 3.34, "learning_rate": 1.002371200689804e-05, "loss": 0.001, "step": 15500 }, { "epoch": 3.35, "learning_rate": 1.0056046561759e-05, "loss": 0.0008, "step": 15550 }, { "epoch": 3.36, "learning_rate": 1.0088381116619962e-05, "loss": 0.0009, "step": 15600 }, { "epoch": 3.37, "learning_rate": 1.0120715671480923e-05, "loss": 0.001, "step": 15650 }, { "epoch": 3.38, "learning_rate": 1.0153050226341885e-05, "loss": 0.001, "step": 15700 }, { "epoch": 3.4, "learning_rate": 1.0185384781202846e-05, "loss": 0.0009, "step": 15750 }, { "epoch": 3.41, "learning_rate": 1.0217719336063807e-05, "loss": 0.0009, "step": 15800 }, { "epoch": 3.42, "learning_rate": 1.0250053890924768e-05, "loss": 0.0008, "step": 15850 }, { "epoch": 3.43, "learning_rate": 1.028238844578573e-05, "loss": 0.0009, "step": 15900 }, { "epoch": 3.44, "learning_rate": 1.0314723000646692e-05, "loss": 0.0009, "step": 15950 }, { "epoch": 3.45, "learning_rate": 1.0347057555507653e-05, "loss": 0.0008, "step": 16000 }, { "epoch": 3.46, "learning_rate": 1.0379392110368614e-05, "loss": 0.0009, "step": 16050 }, { "epoch": 3.47, "learning_rate": 1.0411726665229575e-05, "loss": 0.0009, "step": 16100 }, { "epoch": 3.48, "learning_rate": 1.0444061220090538e-05, "loss": 0.0016, "step": 16150 }, { "epoch": 3.49, "learning_rate": 1.0476395774951499e-05, "loss": 0.0009, "step": 16200 }, { "epoch": 3.5, "learning_rate": 1.050873032981246e-05, "loss": 0.0008, "step": 16250 }, { "epoch": 3.51, "learning_rate": 1.054106488467342e-05, "loss": 0.001, "step": 16300 }, { "epoch": 3.52, "learning_rate": 1.0573399439534382e-05, "loss": 0.0009, "step": 16350 }, { "epoch": 3.54, "learning_rate": 1.0605733994395344e-05, "loss": 0.0009, "step": 16400 }, { "epoch": 3.55, "learning_rate": 1.0638068549256305e-05, "loss": 0.001, "step": 16450 }, { "epoch": 3.56, "learning_rate": 1.0670403104117266e-05, "loss": 0.0009, "step": 16500 }, { "epoch": 3.57, "learning_rate": 1.0702737658978227e-05, "loss": 0.0009, "step": 16550 }, { "epoch": 3.58, "learning_rate": 1.073507221383919e-05, "loss": 0.0009, "step": 16600 }, { "epoch": 3.59, "learning_rate": 1.0767406768700151e-05, "loss": 0.0009, "step": 16650 }, { "epoch": 3.6, "learning_rate": 1.0799741323561112e-05, "loss": 0.0009, "step": 16700 }, { "epoch": 3.61, "learning_rate": 1.0832075878422073e-05, "loss": 0.0008, "step": 16750 }, { "epoch": 3.62, "learning_rate": 1.0864410433283034e-05, "loss": 0.0009, "step": 16800 }, { "epoch": 3.63, "learning_rate": 1.0896744988143997e-05, "loss": 0.0008, "step": 16850 }, { "epoch": 3.64, "learning_rate": 1.0929079543004958e-05, "loss": 0.0008, "step": 16900 }, { "epoch": 3.65, "learning_rate": 1.0961414097865919e-05, "loss": 0.0009, "step": 16950 }, { "epoch": 3.66, "learning_rate": 1.099374865272688e-05, "loss": 0.0009, "step": 17000 }, { "epoch": 3.68, "learning_rate": 1.1026083207587843e-05, "loss": 0.0008, "step": 17050 }, { "epoch": 3.69, "learning_rate": 1.1058417762448804e-05, "loss": 0.0008, "step": 17100 }, { "epoch": 3.7, "learning_rate": 1.1090752317309765e-05, "loss": 0.001, "step": 17150 }, { "epoch": 3.71, "learning_rate": 1.1123086872170726e-05, "loss": 0.0009, "step": 17200 }, { "epoch": 3.72, "learning_rate": 1.1155421427031687e-05, "loss": 0.0009, "step": 17250 }, { "epoch": 3.73, "learning_rate": 1.118775598189265e-05, "loss": 0.0009, "step": 17300 }, { "epoch": 3.74, "learning_rate": 1.1220090536753612e-05, "loss": 0.0008, "step": 17350 }, { "epoch": 3.75, "learning_rate": 1.1252425091614573e-05, "loss": 0.0009, "step": 17400 }, { "epoch": 3.76, "learning_rate": 1.1284759646475534e-05, "loss": 0.0008, "step": 17450 }, { "epoch": 3.77, "learning_rate": 1.1317094201336497e-05, "loss": 0.0008, "step": 17500 }, { "epoch": 3.78, "learning_rate": 1.1349428756197458e-05, "loss": 0.0009, "step": 17550 }, { "epoch": 3.79, "learning_rate": 1.1381763311058419e-05, "loss": 0.0009, "step": 17600 }, { "epoch": 3.8, "learning_rate": 1.141409786591938e-05, "loss": 0.0009, "step": 17650 }, { "epoch": 3.82, "learning_rate": 1.1446432420780342e-05, "loss": 0.0009, "step": 17700 }, { "epoch": 3.83, "learning_rate": 1.1478766975641303e-05, "loss": 0.0009, "step": 17750 }, { "epoch": 3.84, "learning_rate": 1.1511101530502264e-05, "loss": 0.0008, "step": 17800 }, { "epoch": 3.85, "learning_rate": 1.1543436085363225e-05, "loss": 0.0009, "step": 17850 }, { "epoch": 3.86, "learning_rate": 1.1575770640224186e-05, "loss": 0.0008, "step": 17900 }, { "epoch": 3.87, "learning_rate": 1.1608105195085149e-05, "loss": 0.0008, "step": 17950 }, { "epoch": 3.88, "learning_rate": 1.164043974994611e-05, "loss": 0.0008, "step": 18000 }, { "epoch": 3.89, "learning_rate": 1.1672774304807071e-05, "loss": 0.0008, "step": 18050 }, { "epoch": 3.9, "learning_rate": 1.1705108859668032e-05, "loss": 0.0009, "step": 18100 }, { "epoch": 3.91, "learning_rate": 1.1737443414528995e-05, "loss": 0.001, "step": 18150 }, { "epoch": 3.92, "learning_rate": 1.1769777969389956e-05, "loss": 0.001, "step": 18200 }, { "epoch": 3.93, "learning_rate": 1.1802112524250917e-05, "loss": 0.0009, "step": 18250 }, { "epoch": 3.94, "learning_rate": 1.1834447079111878e-05, "loss": 0.0008, "step": 18300 }, { "epoch": 3.96, "learning_rate": 1.1866781633972839e-05, "loss": 0.0009, "step": 18350 }, { "epoch": 3.97, "learning_rate": 1.1899116188833801e-05, "loss": 0.0007, "step": 18400 }, { "epoch": 3.98, "learning_rate": 1.1931450743694762e-05, "loss": 0.001, "step": 18450 }, { "epoch": 3.99, "learning_rate": 1.1963785298555723e-05, "loss": 0.0008, "step": 18500 }, { "epoch": 4.0, "learning_rate": 1.1996119853416684e-05, "loss": 0.0007, "step": 18550 }, { "epoch": 4.01, "learning_rate": 1.2028454408277647e-05, "loss": 0.0007, "step": 18600 }, { "epoch": 4.02, "learning_rate": 1.2060788963138608e-05, "loss": 0.0007, "step": 18650 }, { "epoch": 4.03, "learning_rate": 1.2093123517999569e-05, "loss": 0.0008, "step": 18700 }, { "epoch": 4.04, "learning_rate": 1.212545807286053e-05, "loss": 0.0006, "step": 18750 }, { "epoch": 4.05, "learning_rate": 1.2157792627721491e-05, "loss": 0.0007, "step": 18800 }, { "epoch": 4.06, "learning_rate": 1.2190127182582454e-05, "loss": 0.0007, "step": 18850 }, { "epoch": 4.07, "learning_rate": 1.2222461737443415e-05, "loss": 0.0007, "step": 18900 }, { "epoch": 4.08, "learning_rate": 1.2254796292304376e-05, "loss": 0.0007, "step": 18950 }, { "epoch": 4.1, "learning_rate": 1.2287130847165337e-05, "loss": 0.0007, "step": 19000 }, { "epoch": 4.11, "learning_rate": 1.23194654020263e-05, "loss": 0.0006, "step": 19050 }, { "epoch": 4.12, "learning_rate": 1.235179995688726e-05, "loss": 0.0006, "step": 19100 }, { "epoch": 4.13, "learning_rate": 1.2384134511748222e-05, "loss": 0.0006, "step": 19150 }, { "epoch": 4.14, "learning_rate": 1.2416469066609183e-05, "loss": 0.0006, "step": 19200 }, { "epoch": 4.15, "learning_rate": 1.2448803621470144e-05, "loss": 0.0007, "step": 19250 }, { "epoch": 4.16, "learning_rate": 1.2481138176331106e-05, "loss": 0.0007, "step": 19300 }, { "epoch": 4.17, "learning_rate": 1.2513472731192067e-05, "loss": 0.0006, "step": 19350 }, { "epoch": 4.18, "learning_rate": 1.2545807286053028e-05, "loss": 0.0007, "step": 19400 }, { "epoch": 4.19, "learning_rate": 1.257814184091399e-05, "loss": 0.0007, "step": 19450 }, { "epoch": 4.2, "learning_rate": 1.2610476395774952e-05, "loss": 0.0006, "step": 19500 }, { "epoch": 4.21, "learning_rate": 1.2642810950635913e-05, "loss": 0.0007, "step": 19550 }, { "epoch": 4.23, "learning_rate": 1.2675145505496874e-05, "loss": 0.0007, "step": 19600 }, { "epoch": 4.24, "learning_rate": 1.2707480060357835e-05, "loss": 0.0008, "step": 19650 }, { "epoch": 4.25, "learning_rate": 1.2739814615218796e-05, "loss": 0.0008, "step": 19700 }, { "epoch": 4.26, "learning_rate": 1.2772149170079759e-05, "loss": 0.0008, "step": 19750 }, { "epoch": 4.27, "learning_rate": 1.280448372494072e-05, "loss": 0.0007, "step": 19800 }, { "epoch": 4.28, "learning_rate": 1.283681827980168e-05, "loss": 0.0008, "step": 19850 }, { "epoch": 4.29, "learning_rate": 1.2869152834662642e-05, "loss": 0.0007, "step": 19900 }, { "epoch": 4.3, "learning_rate": 1.2901487389523604e-05, "loss": 0.0007, "step": 19950 }, { "epoch": 4.31, "learning_rate": 1.2933821944384565e-05, "loss": 0.0007, "step": 20000 }, { "epoch": 4.32, "learning_rate": 1.2966156499245528e-05, "loss": 0.0006, "step": 20050 }, { "epoch": 4.33, "learning_rate": 1.2998491054106489e-05, "loss": 0.0007, "step": 20100 }, { "epoch": 4.34, "learning_rate": 1.3030825608967452e-05, "loss": 0.0007, "step": 20150 }, { "epoch": 4.35, "learning_rate": 1.3063160163828413e-05, "loss": 0.0006, "step": 20200 }, { "epoch": 4.37, "learning_rate": 1.3095494718689374e-05, "loss": 0.0006, "step": 20250 }, { "epoch": 4.38, "learning_rate": 1.3127829273550335e-05, "loss": 0.0007, "step": 20300 }, { "epoch": 4.39, "learning_rate": 1.3160163828411296e-05, "loss": 0.0006, "step": 20350 }, { "epoch": 4.4, "learning_rate": 1.3192498383272258e-05, "loss": 0.0007, "step": 20400 }, { "epoch": 4.41, "learning_rate": 1.322483293813322e-05, "loss": 0.0007, "step": 20450 }, { "epoch": 4.42, "learning_rate": 1.325716749299418e-05, "loss": 0.0008, "step": 20500 }, { "epoch": 4.43, "learning_rate": 1.3289502047855141e-05, "loss": 0.0007, "step": 20550 }, { "epoch": 4.44, "learning_rate": 1.3321836602716104e-05, "loss": 0.0006, "step": 20600 }, { "epoch": 4.45, "learning_rate": 1.3354171157577065e-05, "loss": 0.0006, "step": 20650 }, { "epoch": 4.46, "learning_rate": 1.3386505712438026e-05, "loss": 0.0008, "step": 20700 }, { "epoch": 4.47, "learning_rate": 1.3418840267298987e-05, "loss": 0.0006, "step": 20750 }, { "epoch": 4.48, "learning_rate": 1.3451174822159948e-05, "loss": 0.0007, "step": 20800 }, { "epoch": 4.49, "learning_rate": 1.3483509377020911e-05, "loss": 0.0006, "step": 20850 }, { "epoch": 4.51, "learning_rate": 1.3515843931881872e-05, "loss": 0.0008, "step": 20900 }, { "epoch": 4.52, "learning_rate": 1.3548178486742833e-05, "loss": 0.0007, "step": 20950 }, { "epoch": 4.53, "learning_rate": 1.3580513041603794e-05, "loss": 0.0006, "step": 21000 }, { "epoch": 4.54, "learning_rate": 1.3612847596464757e-05, "loss": 0.0006, "step": 21050 }, { "epoch": 4.55, "learning_rate": 1.3645182151325718e-05, "loss": 0.0006, "step": 21100 }, { "epoch": 4.56, "learning_rate": 1.3677516706186679e-05, "loss": 0.0006, "step": 21150 }, { "epoch": 4.57, "learning_rate": 1.370985126104764e-05, "loss": 0.0007, "step": 21200 }, { "epoch": 4.58, "learning_rate": 1.37421858159086e-05, "loss": 0.0009, "step": 21250 }, { "epoch": 4.59, "learning_rate": 1.3774520370769563e-05, "loss": 0.0006, "step": 21300 }, { "epoch": 4.6, "learning_rate": 1.3806854925630524e-05, "loss": 0.0007, "step": 21350 }, { "epoch": 4.61, "learning_rate": 1.3839189480491485e-05, "loss": 0.0007, "step": 21400 }, { "epoch": 4.62, "learning_rate": 1.3871524035352446e-05, "loss": 0.0006, "step": 21450 }, { "epoch": 4.63, "learning_rate": 1.3903858590213409e-05, "loss": 0.0006, "step": 21500 }, { "epoch": 4.65, "learning_rate": 1.393619314507437e-05, "loss": 0.0007, "step": 21550 }, { "epoch": 4.66, "learning_rate": 1.3968527699935331e-05, "loss": 0.0006, "step": 21600 }, { "epoch": 4.67, "learning_rate": 1.4000862254796292e-05, "loss": 0.0007, "step": 21650 }, { "epoch": 4.68, "learning_rate": 1.4033196809657253e-05, "loss": 0.0006, "step": 21700 }, { "epoch": 4.69, "learning_rate": 1.4065531364518216e-05, "loss": 0.0006, "step": 21750 }, { "epoch": 4.7, "learning_rate": 1.4097865919379177e-05, "loss": 0.0006, "step": 21800 }, { "epoch": 4.71, "learning_rate": 1.4130200474240138e-05, "loss": 0.0006, "step": 21850 }, { "epoch": 4.72, "learning_rate": 1.4162535029101099e-05, "loss": 0.0006, "step": 21900 }, { "epoch": 4.73, "learning_rate": 1.4194869583962061e-05, "loss": 0.0007, "step": 21950 }, { "epoch": 4.74, "learning_rate": 1.4227204138823022e-05, "loss": 0.0008, "step": 22000 }, { "epoch": 4.75, "learning_rate": 1.4259538693683983e-05, "loss": 0.0008, "step": 22050 }, { "epoch": 4.76, "learning_rate": 1.4291873248544944e-05, "loss": 0.0008, "step": 22100 }, { "epoch": 4.77, "learning_rate": 1.4324207803405907e-05, "loss": 0.0007, "step": 22150 }, { "epoch": 4.79, "learning_rate": 1.4356542358266868e-05, "loss": 0.0007, "step": 22200 }, { "epoch": 4.8, "learning_rate": 1.4388876913127829e-05, "loss": 0.0006, "step": 22250 }, { "epoch": 4.81, "learning_rate": 1.442121146798879e-05, "loss": 0.0006, "step": 22300 }, { "epoch": 4.82, "learning_rate": 1.4453546022849751e-05, "loss": 0.0007, "step": 22350 }, { "epoch": 4.83, "learning_rate": 1.4485880577710714e-05, "loss": 0.0006, "step": 22400 }, { "epoch": 4.84, "learning_rate": 1.4518215132571675e-05, "loss": 0.0007, "step": 22450 }, { "epoch": 4.85, "learning_rate": 1.4550549687432636e-05, "loss": 0.0007, "step": 22500 }, { "epoch": 4.86, "learning_rate": 1.4582884242293597e-05, "loss": 0.0007, "step": 22550 }, { "epoch": 4.87, "learning_rate": 1.461521879715456e-05, "loss": 0.0006, "step": 22600 }, { "epoch": 4.88, "learning_rate": 1.464755335201552e-05, "loss": 0.0006, "step": 22650 }, { "epoch": 4.89, "learning_rate": 1.4679887906876483e-05, "loss": 0.0006, "step": 22700 }, { "epoch": 4.9, "learning_rate": 1.4712222461737444e-05, "loss": 0.0008, "step": 22750 }, { "epoch": 4.91, "learning_rate": 1.4744557016598405e-05, "loss": 0.0006, "step": 22800 }, { "epoch": 4.93, "learning_rate": 1.4776891571459368e-05, "loss": 0.0006, "step": 22850 }, { "epoch": 4.94, "learning_rate": 1.4809226126320329e-05, "loss": 0.0007, "step": 22900 }, { "epoch": 4.95, "learning_rate": 1.484156068118129e-05, "loss": 0.0007, "step": 22950 }, { "epoch": 4.96, "learning_rate": 1.4873895236042251e-05, "loss": 0.0006, "step": 23000 }, { "epoch": 4.97, "learning_rate": 1.4906229790903214e-05, "loss": 0.0006, "step": 23050 }, { "epoch": 4.98, "learning_rate": 1.4938564345764175e-05, "loss": 0.0006, "step": 23100 }, { "epoch": 4.99, "learning_rate": 1.4970898900625136e-05, "loss": 0.0007, "step": 23150 }, { "epoch": 5.0, "learning_rate": 1.5003233455486097e-05, "loss": 0.0006, "step": 23200 }, { "epoch": 5.01, "learning_rate": 1.5035568010347058e-05, "loss": 0.0006, "step": 23250 }, { "epoch": 5.02, "learning_rate": 1.5067902565208019e-05, "loss": 0.0006, "step": 23300 }, { "epoch": 5.03, "learning_rate": 1.510023712006898e-05, "loss": 0.0004, "step": 23350 }, { "epoch": 5.04, "learning_rate": 1.5132571674929944e-05, "loss": 0.0005, "step": 23400 }, { "epoch": 5.05, "learning_rate": 1.5164906229790905e-05, "loss": 0.0005, "step": 23450 }, { "epoch": 5.07, "learning_rate": 1.5197240784651866e-05, "loss": 0.0004, "step": 23500 }, { "epoch": 5.08, "learning_rate": 1.5229575339512827e-05, "loss": 0.0005, "step": 23550 }, { "epoch": 5.09, "learning_rate": 1.5261909894373786e-05, "loss": 0.0005, "step": 23600 }, { "epoch": 5.1, "learning_rate": 1.529424444923475e-05, "loss": 0.0005, "step": 23650 }, { "epoch": 5.11, "learning_rate": 1.532657900409571e-05, "loss": 0.0005, "step": 23700 }, { "epoch": 5.12, "learning_rate": 1.535891355895667e-05, "loss": 0.0006, "step": 23750 }, { "epoch": 5.13, "learning_rate": 1.539124811381763e-05, "loss": 0.0005, "step": 23800 }, { "epoch": 5.14, "learning_rate": 1.5423582668678596e-05, "loss": 0.0005, "step": 23850 }, { "epoch": 5.15, "learning_rate": 1.5455917223539556e-05, "loss": 0.0005, "step": 23900 }, { "epoch": 5.16, "learning_rate": 1.548825177840052e-05, "loss": 0.0005, "step": 23950 }, { "epoch": 5.17, "learning_rate": 1.5520586333261478e-05, "loss": 0.0006, "step": 24000 }, { "epoch": 5.18, "learning_rate": 1.555292088812244e-05, "loss": 0.0006, "step": 24050 }, { "epoch": 5.2, "learning_rate": 1.5585255442983403e-05, "loss": 0.0004, "step": 24100 }, { "epoch": 5.21, "learning_rate": 1.5617589997844362e-05, "loss": 0.0005, "step": 24150 }, { "epoch": 5.22, "learning_rate": 1.5649924552705325e-05, "loss": 0.0005, "step": 24200 }, { "epoch": 5.23, "learning_rate": 1.5682259107566284e-05, "loss": 0.0005, "step": 24250 }, { "epoch": 5.24, "learning_rate": 1.571459366242725e-05, "loss": 0.0005, "step": 24300 }, { "epoch": 5.25, "learning_rate": 1.574692821728821e-05, "loss": 0.0005, "step": 24350 }, { "epoch": 5.26, "learning_rate": 1.5779262772149173e-05, "loss": 0.0004, "step": 24400 }, { "epoch": 5.27, "learning_rate": 1.5811597327010132e-05, "loss": 0.0005, "step": 24450 }, { "epoch": 5.28, "learning_rate": 1.5843931881871095e-05, "loss": 0.0005, "step": 24500 }, { "epoch": 5.29, "learning_rate": 1.5876266436732054e-05, "loss": 0.0005, "step": 24550 }, { "epoch": 5.3, "learning_rate": 1.5908600991593017e-05, "loss": 0.0005, "step": 24600 }, { "epoch": 5.31, "learning_rate": 1.5940935546453976e-05, "loss": 0.0005, "step": 24650 }, { "epoch": 5.32, "learning_rate": 1.597327010131494e-05, "loss": 0.0005, "step": 24700 }, { "epoch": 5.34, "learning_rate": 1.60056046561759e-05, "loss": 0.0005, "step": 24750 }, { "epoch": 5.35, "learning_rate": 1.6037939211036864e-05, "loss": 0.0006, "step": 24800 }, { "epoch": 5.36, "learning_rate": 1.6070273765897823e-05, "loss": 0.0005, "step": 24850 }, { "epoch": 5.37, "learning_rate": 1.6102608320758786e-05, "loss": 0.0005, "step": 24900 }, { "epoch": 5.38, "learning_rate": 1.6134942875619745e-05, "loss": 0.0007, "step": 24950 }, { "epoch": 5.39, "learning_rate": 1.6167277430480708e-05, "loss": 0.0005, "step": 25000 }, { "epoch": 5.4, "learning_rate": 1.6199611985341667e-05, "loss": 0.0005, "step": 25050 }, { "epoch": 5.41, "learning_rate": 1.623194654020263e-05, "loss": 0.0005, "step": 25100 }, { "epoch": 5.42, "learning_rate": 1.626428109506359e-05, "loss": 0.0007, "step": 25150 }, { "epoch": 5.43, "learning_rate": 1.6296615649924555e-05, "loss": 0.0005, "step": 25200 }, { "epoch": 5.44, "learning_rate": 1.6328950204785515e-05, "loss": 0.0006, "step": 25250 }, { "epoch": 5.45, "learning_rate": 1.6361284759646477e-05, "loss": 0.0005, "step": 25300 }, { "epoch": 5.46, "learning_rate": 1.6393619314507437e-05, "loss": 0.0005, "step": 25350 }, { "epoch": 5.48, "learning_rate": 1.64259538693684e-05, "loss": 0.0004, "step": 25400 }, { "epoch": 5.49, "learning_rate": 1.645828842422936e-05, "loss": 0.0005, "step": 25450 }, { "epoch": 5.5, "learning_rate": 1.649062297909032e-05, "loss": 0.0005, "step": 25500 }, { "epoch": 5.51, "learning_rate": 1.652295753395128e-05, "loss": 0.0005, "step": 25550 }, { "epoch": 5.52, "learning_rate": 1.6555292088812247e-05, "loss": 0.0005, "step": 25600 }, { "epoch": 5.53, "learning_rate": 1.6587626643673206e-05, "loss": 0.0005, "step": 25650 }, { "epoch": 5.54, "learning_rate": 1.661996119853417e-05, "loss": 0.0005, "step": 25700 }, { "epoch": 5.55, "learning_rate": 1.6652295753395128e-05, "loss": 0.0006, "step": 25750 }, { "epoch": 5.56, "learning_rate": 1.668463030825609e-05, "loss": 0.0005, "step": 25800 }, { "epoch": 5.57, "learning_rate": 1.671696486311705e-05, "loss": 0.0004, "step": 25850 }, { "epoch": 5.58, "learning_rate": 1.6749299417978013e-05, "loss": 0.0005, "step": 25900 }, { "epoch": 5.59, "learning_rate": 1.6781633972838972e-05, "loss": 0.0005, "step": 25950 }, { "epoch": 5.6, "learning_rate": 1.6813968527699935e-05, "loss": 0.0005, "step": 26000 }, { "epoch": 5.62, "learning_rate": 1.6846303082560898e-05, "loss": 0.0005, "step": 26050 }, { "epoch": 5.63, "learning_rate": 1.687863763742186e-05, "loss": 0.0005, "step": 26100 }, { "epoch": 5.64, "learning_rate": 1.691097219228282e-05, "loss": 0.0005, "step": 26150 }, { "epoch": 5.65, "learning_rate": 1.6943306747143782e-05, "loss": 0.0005, "step": 26200 }, { "epoch": 5.66, "learning_rate": 1.697564130200474e-05, "loss": 0.0005, "step": 26250 }, { "epoch": 5.67, "learning_rate": 1.7007975856865704e-05, "loss": 0.0005, "step": 26300 }, { "epoch": 5.68, "learning_rate": 1.7040310411726664e-05, "loss": 0.0006, "step": 26350 }, { "epoch": 5.69, "learning_rate": 1.7072644966587626e-05, "loss": 0.0005, "step": 26400 }, { "epoch": 5.7, "learning_rate": 1.7104979521448586e-05, "loss": 0.0005, "step": 26450 }, { "epoch": 5.71, "learning_rate": 1.713731407630955e-05, "loss": 0.0006, "step": 26500 }, { "epoch": 5.72, "learning_rate": 1.716964863117051e-05, "loss": 0.0004, "step": 26550 }, { "epoch": 5.73, "learning_rate": 1.7201983186031474e-05, "loss": 0.0008, "step": 26600 }, { "epoch": 5.74, "learning_rate": 1.7234317740892433e-05, "loss": 0.0006, "step": 26650 }, { "epoch": 5.76, "learning_rate": 1.7266652295753396e-05, "loss": 0.0005, "step": 26700 }, { "epoch": 5.77, "learning_rate": 1.729898685061436e-05, "loss": 0.0005, "step": 26750 }, { "epoch": 5.78, "learning_rate": 1.7331321405475318e-05, "loss": 0.0006, "step": 26800 }, { "epoch": 5.79, "learning_rate": 1.736365596033628e-05, "loss": 0.0006, "step": 26850 }, { "epoch": 5.8, "learning_rate": 1.739599051519724e-05, "loss": 0.0005, "step": 26900 }, { "epoch": 5.81, "learning_rate": 1.7428325070058206e-05, "loss": 0.0005, "step": 26950 }, { "epoch": 5.82, "learning_rate": 1.7460659624919165e-05, "loss": 0.0005, "step": 27000 }, { "epoch": 5.83, "learning_rate": 1.7492994179780128e-05, "loss": 0.0006, "step": 27050 }, { "epoch": 5.84, "learning_rate": 1.7525328734641087e-05, "loss": 0.0006, "step": 27100 }, { "epoch": 5.85, "learning_rate": 1.755766328950205e-05, "loss": 0.0005, "step": 27150 }, { "epoch": 5.86, "learning_rate": 1.758999784436301e-05, "loss": 0.0005, "step": 27200 }, { "epoch": 5.87, "learning_rate": 1.7622332399223972e-05, "loss": 0.0005, "step": 27250 }, { "epoch": 5.88, "learning_rate": 1.765466695408493e-05, "loss": 0.0005, "step": 27300 }, { "epoch": 5.9, "learning_rate": 1.7687001508945894e-05, "loss": 0.0005, "step": 27350 }, { "epoch": 5.91, "learning_rate": 1.7719336063806856e-05, "loss": 0.0005, "step": 27400 }, { "epoch": 5.92, "learning_rate": 1.775167061866782e-05, "loss": 0.0006, "step": 27450 }, { "epoch": 5.93, "learning_rate": 1.778400517352878e-05, "loss": 0.0005, "step": 27500 }, { "epoch": 5.94, "learning_rate": 1.781633972838974e-05, "loss": 0.0005, "step": 27550 }, { "epoch": 5.95, "learning_rate": 1.78486742832507e-05, "loss": 0.0005, "step": 27600 }, { "epoch": 5.96, "learning_rate": 1.7881008838111663e-05, "loss": 0.0005, "step": 27650 }, { "epoch": 5.97, "learning_rate": 1.7913343392972622e-05, "loss": 0.0006, "step": 27700 }, { "epoch": 5.98, "learning_rate": 1.7945677947833585e-05, "loss": 0.0005, "step": 27750 }, { "epoch": 5.99, "learning_rate": 1.7978012502694544e-05, "loss": 0.0005, "step": 27800 }, { "epoch": 6.0, "learning_rate": 1.801034705755551e-05, "loss": 0.0005, "step": 27850 }, { "epoch": 6.01, "learning_rate": 1.804268161241647e-05, "loss": 0.0005, "step": 27900 }, { "epoch": 6.03, "learning_rate": 1.8075016167277433e-05, "loss": 0.0004, "step": 27950 }, { "epoch": 6.04, "learning_rate": 1.8107350722138392e-05, "loss": 0.0004, "step": 28000 }, { "epoch": 6.05, "learning_rate": 1.8139685276999355e-05, "loss": 0.0004, "step": 28050 }, { "epoch": 6.06, "learning_rate": 1.8172019831860314e-05, "loss": 0.0005, "step": 28100 }, { "epoch": 6.07, "learning_rate": 1.8204354386721277e-05, "loss": 0.0004, "step": 28150 }, { "epoch": 6.08, "learning_rate": 1.8236688941582236e-05, "loss": 0.0004, "step": 28200 }, { "epoch": 6.09, "learning_rate": 1.82690234964432e-05, "loss": 0.0004, "step": 28250 }, { "epoch": 6.1, "learning_rate": 1.830135805130416e-05, "loss": 0.0004, "step": 28300 }, { "epoch": 6.11, "learning_rate": 1.8333692606165124e-05, "loss": 0.0003, "step": 28350 }, { "epoch": 6.12, "learning_rate": 1.8366027161026083e-05, "loss": 0.0003, "step": 28400 }, { "epoch": 6.13, "learning_rate": 1.8398361715887046e-05, "loss": 0.0004, "step": 28450 }, { "epoch": 6.14, "learning_rate": 1.8430696270748005e-05, "loss": 0.0003, "step": 28500 }, { "epoch": 6.15, "learning_rate": 1.8463030825608968e-05, "loss": 0.0004, "step": 28550 }, { "epoch": 6.17, "learning_rate": 1.8495365380469927e-05, "loss": 0.0004, "step": 28600 }, { "epoch": 6.18, "learning_rate": 1.852769993533089e-05, "loss": 0.0004, "step": 28650 }, { "epoch": 6.19, "learning_rate": 1.856003449019185e-05, "loss": 0.0004, "step": 28700 }, { "epoch": 6.2, "learning_rate": 1.8592369045052815e-05, "loss": 0.0004, "step": 28750 }, { "epoch": 6.21, "learning_rate": 1.8624703599913775e-05, "loss": 0.0004, "step": 28800 }, { "epoch": 6.22, "learning_rate": 1.8657038154774737e-05, "loss": 0.0003, "step": 28850 }, { "epoch": 6.23, "learning_rate": 1.8689372709635697e-05, "loss": 0.0003, "step": 28900 }, { "epoch": 6.24, "learning_rate": 1.872170726449666e-05, "loss": 0.0005, "step": 28950 }, { "epoch": 6.25, "learning_rate": 1.875404181935762e-05, "loss": 0.0004, "step": 29000 }, { "epoch": 6.26, "learning_rate": 1.878637637421858e-05, "loss": 0.0004, "step": 29050 }, { "epoch": 6.27, "learning_rate": 1.881871092907954e-05, "loss": 0.0003, "step": 29100 }, { "epoch": 6.28, "learning_rate": 1.8851045483940503e-05, "loss": 0.0004, "step": 29150 }, { "epoch": 6.29, "learning_rate": 1.8883380038801466e-05, "loss": 0.0004, "step": 29200 }, { "epoch": 6.31, "learning_rate": 1.891571459366243e-05, "loss": 0.0004, "step": 29250 }, { "epoch": 6.32, "learning_rate": 1.8948049148523388e-05, "loss": 0.0004, "step": 29300 }, { "epoch": 6.33, "learning_rate": 1.898038370338435e-05, "loss": 0.0004, "step": 29350 }, { "epoch": 6.34, "learning_rate": 1.9012718258245313e-05, "loss": 0.0004, "step": 29400 }, { "epoch": 6.35, "learning_rate": 1.9045052813106273e-05, "loss": 0.0004, "step": 29450 }, { "epoch": 6.36, "learning_rate": 1.9077387367967235e-05, "loss": 0.0003, "step": 29500 }, { "epoch": 6.37, "learning_rate": 1.9109721922828195e-05, "loss": 0.0004, "step": 29550 }, { "epoch": 6.38, "learning_rate": 1.9142056477689158e-05, "loss": 0.0004, "step": 29600 }, { "epoch": 6.39, "learning_rate": 1.917439103255012e-05, "loss": 0.0004, "step": 29650 }, { "epoch": 6.4, "learning_rate": 1.9206725587411083e-05, "loss": 0.0003, "step": 29700 }, { "epoch": 6.41, "learning_rate": 1.9239060142272042e-05, "loss": 0.0004, "step": 29750 }, { "epoch": 6.42, "learning_rate": 1.9271394697133005e-05, "loss": 0.0005, "step": 29800 }, { "epoch": 6.43, "learning_rate": 1.9303729251993964e-05, "loss": 0.0004, "step": 29850 }, { "epoch": 6.45, "learning_rate": 1.9336063806854927e-05, "loss": 0.0004, "step": 29900 }, { "epoch": 6.46, "learning_rate": 1.9368398361715886e-05, "loss": 0.0004, "step": 29950 }, { "epoch": 6.47, "learning_rate": 1.940073291657685e-05, "loss": 0.0004, "step": 30000 }, { "epoch": 6.48, "learning_rate": 1.9433067471437808e-05, "loss": 0.0004, "step": 30050 }, { "epoch": 6.49, "learning_rate": 1.9465402026298774e-05, "loss": 0.0004, "step": 30100 }, { "epoch": 6.5, "learning_rate": 1.9497736581159734e-05, "loss": 0.0003, "step": 30150 }, { "epoch": 6.51, "learning_rate": 1.9530071136020696e-05, "loss": 0.0004, "step": 30200 }, { "epoch": 6.52, "learning_rate": 1.9562405690881656e-05, "loss": 0.0004, "step": 30250 }, { "epoch": 6.53, "learning_rate": 1.959474024574262e-05, "loss": 0.0004, "step": 30300 }, { "epoch": 6.54, "learning_rate": 1.9627074800603578e-05, "loss": 0.0004, "step": 30350 }, { "epoch": 6.55, "learning_rate": 1.965940935546454e-05, "loss": 0.0004, "step": 30400 }, { "epoch": 6.56, "learning_rate": 1.96917439103255e-05, "loss": 0.0003, "step": 30450 }, { "epoch": 6.57, "learning_rate": 1.9724078465186466e-05, "loss": 0.0004, "step": 30500 }, { "epoch": 6.59, "learning_rate": 1.9756413020047425e-05, "loss": 0.0005, "step": 30550 }, { "epoch": 6.6, "learning_rate": 1.9788747574908388e-05, "loss": 0.0004, "step": 30600 }, { "epoch": 6.61, "learning_rate": 1.9821082129769347e-05, "loss": 0.0003, "step": 30650 }, { "epoch": 6.62, "learning_rate": 1.985341668463031e-05, "loss": 0.0004, "step": 30700 }, { "epoch": 6.63, "learning_rate": 1.988575123949127e-05, "loss": 0.0004, "step": 30750 }, { "epoch": 6.64, "learning_rate": 1.9918085794352232e-05, "loss": 0.0005, "step": 30800 }, { "epoch": 6.65, "learning_rate": 1.995042034921319e-05, "loss": 0.0004, "step": 30850 }, { "epoch": 6.66, "learning_rate": 1.9982754904074154e-05, "loss": 0.0004, "step": 30900 }, { "epoch": 6.67, "learning_rate": 2.0015089458935116e-05, "loss": 0.0003, "step": 30950 }, { "epoch": 6.68, "learning_rate": 2.004742401379608e-05, "loss": 0.0004, "step": 31000 }, { "epoch": 6.69, "learning_rate": 2.007975856865704e-05, "loss": 0.0004, "step": 31050 }, { "epoch": 6.7, "learning_rate": 2.0112093123518e-05, "loss": 0.0004, "step": 31100 }, { "epoch": 6.71, "learning_rate": 2.014442767837896e-05, "loss": 0.0004, "step": 31150 }, { "epoch": 6.73, "learning_rate": 2.0176762233239923e-05, "loss": 0.0005, "step": 31200 }, { "epoch": 6.74, "learning_rate": 2.0209096788100882e-05, "loss": 0.0005, "step": 31250 }, { "epoch": 6.75, "learning_rate": 2.0241431342961845e-05, "loss": 0.0004, "step": 31300 }, { "epoch": 6.76, "learning_rate": 2.0273765897822804e-05, "loss": 0.0004, "step": 31350 }, { "epoch": 6.77, "learning_rate": 2.030610045268377e-05, "loss": 0.0004, "step": 31400 }, { "epoch": 6.78, "learning_rate": 2.033843500754473e-05, "loss": 0.0004, "step": 31450 }, { "epoch": 6.79, "learning_rate": 2.0370769562405693e-05, "loss": 0.0004, "step": 31500 }, { "epoch": 6.8, "learning_rate": 2.0403104117266652e-05, "loss": 0.0004, "step": 31550 }, { "epoch": 6.81, "learning_rate": 2.0435438672127615e-05, "loss": 0.0003, "step": 31600 }, { "epoch": 6.82, "learning_rate": 2.0467773226988574e-05, "loss": 0.0004, "step": 31650 }, { "epoch": 6.83, "learning_rate": 2.0500107781849537e-05, "loss": 0.0004, "step": 31700 }, { "epoch": 6.84, "learning_rate": 2.0532442336710496e-05, "loss": 0.0004, "step": 31750 }, { "epoch": 6.85, "learning_rate": 2.056477689157146e-05, "loss": 0.0004, "step": 31800 }, { "epoch": 6.87, "learning_rate": 2.059711144643242e-05, "loss": 0.0004, "step": 31850 }, { "epoch": 6.88, "learning_rate": 2.0629446001293384e-05, "loss": 0.0004, "step": 31900 }, { "epoch": 6.89, "learning_rate": 2.0661780556154343e-05, "loss": 0.0004, "step": 31950 }, { "epoch": 6.9, "learning_rate": 2.0694115111015306e-05, "loss": 0.0004, "step": 32000 }, { "epoch": 6.91, "learning_rate": 2.072644966587627e-05, "loss": 0.0004, "step": 32050 }, { "epoch": 6.92, "learning_rate": 2.0758784220737228e-05, "loss": 0.0004, "step": 32100 }, { "epoch": 6.93, "learning_rate": 2.079111877559819e-05, "loss": 0.0005, "step": 32150 }, { "epoch": 6.94, "learning_rate": 2.082345333045915e-05, "loss": 0.0005, "step": 32200 }, { "epoch": 6.95, "learning_rate": 2.0855787885320113e-05, "loss": 0.0005, "step": 32250 }, { "epoch": 6.96, "learning_rate": 2.0888122440181075e-05, "loss": 0.0003, "step": 32300 }, { "epoch": 6.97, "learning_rate": 2.0920456995042038e-05, "loss": 0.0004, "step": 32350 }, { "epoch": 6.98, "learning_rate": 2.0952791549902997e-05, "loss": 0.0004, "step": 32400 }, { "epoch": 7.0, "learning_rate": 2.098512610476396e-05, "loss": 0.0005, "step": 32450 }, { "epoch": 7.01, "learning_rate": 2.101746065962492e-05, "loss": 0.0004, "step": 32500 }, { "epoch": 7.02, "learning_rate": 2.1049795214485882e-05, "loss": 0.0003, "step": 32550 }, { "epoch": 7.03, "learning_rate": 2.108212976934684e-05, "loss": 0.0003, "step": 32600 }, { "epoch": 7.04, "learning_rate": 2.1114464324207804e-05, "loss": 0.0003, "step": 32650 }, { "epoch": 7.05, "learning_rate": 2.1146798879068763e-05, "loss": 0.0003, "step": 32700 }, { "epoch": 7.06, "learning_rate": 2.117913343392973e-05, "loss": 0.0004, "step": 32750 }, { "epoch": 7.07, "learning_rate": 2.121146798879069e-05, "loss": 0.0003, "step": 32800 }, { "epoch": 7.08, "learning_rate": 2.124380254365165e-05, "loss": 0.0004, "step": 32850 }, { "epoch": 7.09, "learning_rate": 2.127613709851261e-05, "loss": 0.0003, "step": 32900 }, { "epoch": 7.1, "learning_rate": 2.1308471653373573e-05, "loss": 0.0003, "step": 32950 }, { "epoch": 7.11, "learning_rate": 2.1340806208234533e-05, "loss": 0.0003, "step": 33000 }, { "epoch": 7.12, "learning_rate": 2.1373140763095495e-05, "loss": 0.0003, "step": 33050 }, { "epoch": 7.14, "learning_rate": 2.1405475317956455e-05, "loss": 0.0003, "step": 33100 }, { "epoch": 7.15, "learning_rate": 2.1437809872817417e-05, "loss": 0.0003, "step": 33150 }, { "epoch": 7.16, "learning_rate": 2.147014442767838e-05, "loss": 0.0003, "step": 33200 }, { "epoch": 7.17, "learning_rate": 2.1502478982539343e-05, "loss": 0.0003, "step": 33250 }, { "epoch": 7.18, "learning_rate": 2.1534813537400302e-05, "loss": 0.0003, "step": 33300 }, { "epoch": 7.19, "learning_rate": 2.1567148092261265e-05, "loss": 0.0003, "step": 33350 }, { "epoch": 7.2, "learning_rate": 2.1599482647122224e-05, "loss": 0.0003, "step": 33400 }, { "epoch": 7.21, "learning_rate": 2.1631817201983187e-05, "loss": 0.0003, "step": 33450 }, { "epoch": 7.22, "learning_rate": 2.1664151756844146e-05, "loss": 0.0003, "step": 33500 }, { "epoch": 7.23, "learning_rate": 2.169648631170511e-05, "loss": 0.0003, "step": 33550 }, { "epoch": 7.24, "learning_rate": 2.1728820866566068e-05, "loss": 0.0004, "step": 33600 }, { "epoch": 7.25, "learning_rate": 2.1761155421427034e-05, "loss": 0.0003, "step": 33650 }, { "epoch": 7.26, "learning_rate": 2.1793489976287994e-05, "loss": 0.0003, "step": 33700 }, { "epoch": 7.28, "learning_rate": 2.1825824531148956e-05, "loss": 0.0004, "step": 33750 }, { "epoch": 7.29, "learning_rate": 2.1858159086009916e-05, "loss": 0.0003, "step": 33800 }, { "epoch": 7.3, "learning_rate": 2.1890493640870878e-05, "loss": 0.0003, "step": 33850 }, { "epoch": 7.31, "learning_rate": 2.1922828195731838e-05, "loss": 0.0004, "step": 33900 }, { "epoch": 7.32, "learning_rate": 2.19551627505928e-05, "loss": 0.0004, "step": 33950 }, { "epoch": 7.33, "learning_rate": 2.198749730545376e-05, "loss": 0.0004, "step": 34000 }, { "epoch": 7.34, "learning_rate": 2.2019831860314722e-05, "loss": 0.0003, "step": 34050 }, { "epoch": 7.35, "learning_rate": 2.2052166415175685e-05, "loss": 0.0002, "step": 34100 }, { "epoch": 7.36, "learning_rate": 2.2084500970036648e-05, "loss": 0.0004, "step": 34150 }, { "epoch": 7.37, "learning_rate": 2.2116835524897607e-05, "loss": 0.0003, "step": 34200 }, { "epoch": 7.38, "learning_rate": 2.214917007975857e-05, "loss": 0.0005, "step": 34250 }, { "epoch": 7.39, "learning_rate": 2.218150463461953e-05, "loss": 0.0003, "step": 34300 }, { "epoch": 7.4, "learning_rate": 2.2213839189480492e-05, "loss": 0.0004, "step": 34350 }, { "epoch": 7.42, "learning_rate": 2.224617374434145e-05, "loss": 0.0003, "step": 34400 }, { "epoch": 7.43, "learning_rate": 2.2278508299202414e-05, "loss": 0.0003, "step": 34450 }, { "epoch": 7.44, "learning_rate": 2.2310842854063373e-05, "loss": 0.0004, "step": 34500 }, { "epoch": 7.45, "learning_rate": 2.234317740892434e-05, "loss": 0.0003, "step": 34550 }, { "epoch": 7.46, "learning_rate": 2.23755119637853e-05, "loss": 0.0003, "step": 34600 }, { "epoch": 7.47, "learning_rate": 2.240784651864626e-05, "loss": 0.0003, "step": 34650 }, { "epoch": 7.48, "learning_rate": 2.2440181073507224e-05, "loss": 0.0003, "step": 34700 }, { "epoch": 7.49, "learning_rate": 2.2472515628368183e-05, "loss": 0.0003, "step": 34750 }, { "epoch": 7.5, "learning_rate": 2.2504850183229146e-05, "loss": 0.0003, "step": 34800 }, { "epoch": 7.51, "learning_rate": 2.2537184738090105e-05, "loss": 0.0003, "step": 34850 }, { "epoch": 7.52, "learning_rate": 2.2569519292951068e-05, "loss": 0.0004, "step": 34900 }, { "epoch": 7.53, "learning_rate": 2.260185384781203e-05, "loss": 0.0003, "step": 34950 }, { "epoch": 7.54, "learning_rate": 2.2634188402672993e-05, "loss": 0.0004, "step": 35000 }, { "epoch": 7.56, "learning_rate": 2.2666522957533953e-05, "loss": 0.0003, "step": 35050 }, { "epoch": 7.57, "learning_rate": 2.2698857512394915e-05, "loss": 0.0002, "step": 35100 }, { "epoch": 7.58, "learning_rate": 2.2731192067255875e-05, "loss": 0.0003, "step": 35150 }, { "epoch": 7.59, "learning_rate": 2.2763526622116837e-05, "loss": 0.0004, "step": 35200 }, { "epoch": 7.6, "learning_rate": 2.2795861176977797e-05, "loss": 0.0003, "step": 35250 }, { "epoch": 7.61, "learning_rate": 2.282819573183876e-05, "loss": 0.0003, "step": 35300 }, { "epoch": 7.62, "learning_rate": 2.286053028669972e-05, "loss": 0.0003, "step": 35350 }, { "epoch": 7.63, "learning_rate": 2.2892864841560685e-05, "loss": 0.0003, "step": 35400 }, { "epoch": 7.64, "learning_rate": 2.2925199396421644e-05, "loss": 0.0003, "step": 35450 }, { "epoch": 7.65, "learning_rate": 2.2957533951282607e-05, "loss": 0.0004, "step": 35500 }, { "epoch": 7.66, "learning_rate": 2.2989868506143566e-05, "loss": 0.0004, "step": 35550 }, { "epoch": 7.67, "learning_rate": 2.302220306100453e-05, "loss": 0.0003, "step": 35600 }, { "epoch": 7.68, "learning_rate": 2.3054537615865488e-05, "loss": 0.0003, "step": 35650 }, { "epoch": 7.7, "learning_rate": 2.308687217072645e-05, "loss": 0.0003, "step": 35700 }, { "epoch": 7.71, "learning_rate": 2.311920672558741e-05, "loss": 0.0003, "step": 35750 }, { "epoch": 7.72, "learning_rate": 2.3151541280448373e-05, "loss": 0.0003, "step": 35800 }, { "epoch": 7.73, "learning_rate": 2.3183875835309335e-05, "loss": 0.0003, "step": 35850 }, { "epoch": 7.74, "learning_rate": 2.3216210390170298e-05, "loss": 0.0004, "step": 35900 }, { "epoch": 7.75, "learning_rate": 2.3248544945031257e-05, "loss": 0.0004, "step": 35950 }, { "epoch": 7.76, "learning_rate": 2.328087949989222e-05, "loss": 0.0003, "step": 36000 }, { "epoch": 7.77, "learning_rate": 2.331321405475318e-05, "loss": 0.0004, "step": 36050 }, { "epoch": 7.78, "learning_rate": 2.3345548609614142e-05, "loss": 0.0004, "step": 36100 }, { "epoch": 7.79, "learning_rate": 2.33778831644751e-05, "loss": 0.0003, "step": 36150 }, { "epoch": 7.8, "learning_rate": 2.3410217719336064e-05, "loss": 0.0003, "step": 36200 }, { "epoch": 7.81, "learning_rate": 2.3442552274197023e-05, "loss": 0.0004, "step": 36250 }, { "epoch": 7.82, "learning_rate": 2.347488682905799e-05, "loss": 0.0003, "step": 36300 }, { "epoch": 7.84, "learning_rate": 2.350722138391895e-05, "loss": 0.0003, "step": 36350 }, { "epoch": 7.85, "learning_rate": 2.353955593877991e-05, "loss": 0.0003, "step": 36400 }, { "epoch": 7.86, "learning_rate": 2.357189049364087e-05, "loss": 0.0004, "step": 36450 }, { "epoch": 7.87, "learning_rate": 2.3604225048501833e-05, "loss": 0.0004, "step": 36500 }, { "epoch": 7.88, "learning_rate": 2.3636559603362793e-05, "loss": 0.0004, "step": 36550 }, { "epoch": 7.89, "learning_rate": 2.3668894158223755e-05, "loss": 0.0004, "step": 36600 }, { "epoch": 7.9, "learning_rate": 2.3701228713084715e-05, "loss": 0.0004, "step": 36650 }, { "epoch": 7.91, "learning_rate": 2.3733563267945677e-05, "loss": 0.0003, "step": 36700 }, { "epoch": 7.92, "learning_rate": 2.376589782280664e-05, "loss": 0.0003, "step": 36750 }, { "epoch": 7.93, "learning_rate": 2.3798232377667603e-05, "loss": 0.0004, "step": 36800 }, { "epoch": 7.94, "learning_rate": 2.3830566932528562e-05, "loss": 0.0003, "step": 36850 }, { "epoch": 7.95, "learning_rate": 2.3862901487389525e-05, "loss": 0.0003, "step": 36900 }, { "epoch": 7.97, "learning_rate": 2.3895236042250484e-05, "loss": 0.0003, "step": 36950 }, { "epoch": 7.98, "learning_rate": 2.3927570597111447e-05, "loss": 0.0003, "step": 37000 }, { "epoch": 7.99, "learning_rate": 2.3959905151972406e-05, "loss": 0.0003, "step": 37050 }, { "epoch": 8.0, "learning_rate": 2.399223970683337e-05, "loss": 0.0004, "step": 37100 }, { "epoch": 8.01, "learning_rate": 2.4024574261694328e-05, "loss": 0.0003, "step": 37150 }, { "epoch": 8.02, "learning_rate": 2.4056908816555294e-05, "loss": 0.0003, "step": 37200 }, { "epoch": 8.03, "learning_rate": 2.4089243371416254e-05, "loss": 0.0002, "step": 37250 }, { "epoch": 8.04, "learning_rate": 2.4121577926277216e-05, "loss": 0.0003, "step": 37300 }, { "epoch": 8.05, "learning_rate": 2.415391248113818e-05, "loss": 0.0002, "step": 37350 }, { "epoch": 8.06, "learning_rate": 2.4186247035999138e-05, "loss": 0.0003, "step": 37400 }, { "epoch": 8.07, "learning_rate": 2.42185815908601e-05, "loss": 0.0003, "step": 37450 }, { "epoch": 8.08, "learning_rate": 2.425091614572106e-05, "loss": 0.0003, "step": 37500 }, { "epoch": 8.09, "learning_rate": 2.4283250700582023e-05, "loss": 0.0002, "step": 37550 }, { "epoch": 8.11, "learning_rate": 2.4315585255442982e-05, "loss": 0.0003, "step": 37600 }, { "epoch": 8.12, "learning_rate": 2.434791981030395e-05, "loss": 0.0003, "step": 37650 }, { "epoch": 8.13, "learning_rate": 2.4380254365164908e-05, "loss": 0.0003, "step": 37700 }, { "epoch": 8.14, "learning_rate": 2.441258892002587e-05, "loss": 0.0003, "step": 37750 }, { "epoch": 8.15, "learning_rate": 2.444492347488683e-05, "loss": 0.0002, "step": 37800 }, { "epoch": 8.16, "learning_rate": 2.4477258029747792e-05, "loss": 0.0003, "step": 37850 }, { "epoch": 8.17, "learning_rate": 2.4509592584608752e-05, "loss": 0.0002, "step": 37900 }, { "epoch": 8.18, "learning_rate": 2.4541927139469714e-05, "loss": 0.0003, "step": 37950 }, { "epoch": 8.19, "learning_rate": 2.4574261694330674e-05, "loss": 0.0003, "step": 38000 }, { "epoch": 8.2, "learning_rate": 2.4606596249191636e-05, "loss": 0.0003, "step": 38050 }, { "epoch": 8.21, "learning_rate": 2.46389308040526e-05, "loss": 0.0003, "step": 38100 }, { "epoch": 8.22, "learning_rate": 2.4671265358913562e-05, "loss": 0.0003, "step": 38150 }, { "epoch": 8.23, "learning_rate": 2.470359991377452e-05, "loss": 0.0003, "step": 38200 }, { "epoch": 8.25, "learning_rate": 2.4735934468635484e-05, "loss": 0.0003, "step": 38250 }, { "epoch": 8.26, "learning_rate": 2.4768269023496443e-05, "loss": 0.0003, "step": 38300 }, { "epoch": 8.27, "learning_rate": 2.4800603578357406e-05, "loss": 0.0002, "step": 38350 }, { "epoch": 8.28, "learning_rate": 2.4832938133218365e-05, "loss": 0.0003, "step": 38400 }, { "epoch": 8.29, "learning_rate": 2.4865272688079328e-05, "loss": 0.0003, "step": 38450 }, { "epoch": 8.3, "learning_rate": 2.4897607242940287e-05, "loss": 0.0002, "step": 38500 }, { "epoch": 8.31, "learning_rate": 2.4929941797801253e-05, "loss": 0.0003, "step": 38550 }, { "epoch": 8.32, "learning_rate": 2.4962276352662213e-05, "loss": 0.0003, "step": 38600 }, { "epoch": 8.33, "learning_rate": 2.4994610907523175e-05, "loss": 0.0003, "step": 38650 }, { "epoch": 8.34, "learning_rate": 2.5026945462384135e-05, "loss": 0.0003, "step": 38700 }, { "epoch": 8.35, "learning_rate": 2.5059280017245097e-05, "loss": 0.0002, "step": 38750 }, { "epoch": 8.36, "learning_rate": 2.5091614572106057e-05, "loss": 0.0003, "step": 38800 }, { "epoch": 8.37, "learning_rate": 2.512394912696702e-05, "loss": 0.0003, "step": 38850 }, { "epoch": 8.39, "learning_rate": 2.515628368182798e-05, "loss": 0.0003, "step": 38900 }, { "epoch": 8.4, "learning_rate": 2.518861823668894e-05, "loss": 0.0003, "step": 38950 }, { "epoch": 8.41, "learning_rate": 2.5220952791549904e-05, "loss": 0.0002, "step": 39000 }, { "epoch": 8.42, "learning_rate": 2.5253287346410867e-05, "loss": 0.0003, "step": 39050 }, { "epoch": 8.43, "learning_rate": 2.5285621901271826e-05, "loss": 0.0003, "step": 39100 }, { "epoch": 8.44, "learning_rate": 2.531795645613279e-05, "loss": 0.0003, "step": 39150 }, { "epoch": 8.45, "learning_rate": 2.5350291010993748e-05, "loss": 0.0003, "step": 39200 }, { "epoch": 8.46, "learning_rate": 2.538262556585471e-05, "loss": 0.0002, "step": 39250 }, { "epoch": 8.47, "learning_rate": 2.541496012071567e-05, "loss": 0.0003, "step": 39300 }, { "epoch": 8.48, "learning_rate": 2.5447294675576633e-05, "loss": 0.0003, "step": 39350 }, { "epoch": 8.49, "learning_rate": 2.5479629230437592e-05, "loss": 0.0003, "step": 39400 }, { "epoch": 8.5, "learning_rate": 2.5511963785298558e-05, "loss": 0.0003, "step": 39450 }, { "epoch": 8.51, "learning_rate": 2.5544298340159517e-05, "loss": 0.0003, "step": 39500 }, { "epoch": 8.53, "learning_rate": 2.557663289502048e-05, "loss": 0.0002, "step": 39550 }, { "epoch": 8.54, "learning_rate": 2.560896744988144e-05, "loss": 0.0003, "step": 39600 }, { "epoch": 8.55, "learning_rate": 2.5641302004742402e-05, "loss": 0.0003, "step": 39650 }, { "epoch": 8.56, "learning_rate": 2.567363655960336e-05, "loss": 0.0003, "step": 39700 }, { "epoch": 8.57, "learning_rate": 2.5705971114464324e-05, "loss": 0.0003, "step": 39750 }, { "epoch": 8.58, "learning_rate": 2.5738305669325283e-05, "loss": 0.0003, "step": 39800 }, { "epoch": 8.59, "learning_rate": 2.577064022418625e-05, "loss": 0.0003, "step": 39850 }, { "epoch": 8.6, "learning_rate": 2.580297477904721e-05, "loss": 0.0002, "step": 39900 }, { "epoch": 8.61, "learning_rate": 2.583530933390817e-05, "loss": 0.0003, "step": 39950 }, { "epoch": 8.62, "learning_rate": 2.586764388876913e-05, "loss": 0.0002, "step": 40000 }, { "epoch": 8.63, "learning_rate": 2.5899978443630093e-05, "loss": 0.0003, "step": 40050 }, { "epoch": 8.64, "learning_rate": 2.5932312998491056e-05, "loss": 0.0003, "step": 40100 }, { "epoch": 8.65, "learning_rate": 2.5964647553352015e-05, "loss": 0.0003, "step": 40150 }, { "epoch": 8.67, "learning_rate": 2.5996982108212978e-05, "loss": 0.0003, "step": 40200 }, { "epoch": 8.68, "learning_rate": 2.6029316663073937e-05, "loss": 0.0003, "step": 40250 }, { "epoch": 8.69, "learning_rate": 2.6061651217934904e-05, "loss": 0.0003, "step": 40300 }, { "epoch": 8.7, "learning_rate": 2.6093985772795863e-05, "loss": 0.0003, "step": 40350 }, { "epoch": 8.71, "learning_rate": 2.6126320327656826e-05, "loss": 0.0003, "step": 40400 }, { "epoch": 8.72, "learning_rate": 2.6158654882517785e-05, "loss": 0.0002, "step": 40450 }, { "epoch": 8.73, "learning_rate": 2.6190989437378748e-05, "loss": 0.0002, "step": 40500 }, { "epoch": 8.74, "learning_rate": 2.6223323992239707e-05, "loss": 0.0002, "step": 40550 }, { "epoch": 8.75, "learning_rate": 2.625565854710067e-05, "loss": 0.0003, "step": 40600 }, { "epoch": 8.76, "learning_rate": 2.628799310196163e-05, "loss": 0.0003, "step": 40650 }, { "epoch": 8.77, "learning_rate": 2.632032765682259e-05, "loss": 0.0004, "step": 40700 }, { "epoch": 8.78, "learning_rate": 2.6352662211683554e-05, "loss": 0.0003, "step": 40750 }, { "epoch": 8.79, "learning_rate": 2.6384996766544517e-05, "loss": 0.0003, "step": 40800 }, { "epoch": 8.81, "learning_rate": 2.6417331321405476e-05, "loss": 0.0003, "step": 40850 }, { "epoch": 8.82, "learning_rate": 2.644966587626644e-05, "loss": 0.0003, "step": 40900 }, { "epoch": 8.83, "learning_rate": 2.6482000431127398e-05, "loss": 0.0003, "step": 40950 }, { "epoch": 8.84, "learning_rate": 2.651433498598836e-05, "loss": 0.0002, "step": 41000 }, { "epoch": 8.85, "learning_rate": 2.654666954084932e-05, "loss": 0.0004, "step": 41050 }, { "epoch": 8.86, "learning_rate": 2.6579004095710283e-05, "loss": 0.0003, "step": 41100 }, { "epoch": 8.87, "learning_rate": 2.6611338650571242e-05, "loss": 0.0002, "step": 41150 }, { "epoch": 8.88, "learning_rate": 2.664367320543221e-05, "loss": 0.0002, "step": 41200 }, { "epoch": 8.89, "learning_rate": 2.6676007760293168e-05, "loss": 0.0003, "step": 41250 }, { "epoch": 8.9, "learning_rate": 2.670834231515413e-05, "loss": 0.0002, "step": 41300 }, { "epoch": 8.91, "learning_rate": 2.674067687001509e-05, "loss": 0.0003, "step": 41350 }, { "epoch": 8.92, "learning_rate": 2.6773011424876052e-05, "loss": 0.0003, "step": 41400 }, { "epoch": 8.94, "learning_rate": 2.680534597973701e-05, "loss": 0.0002, "step": 41450 }, { "epoch": 8.95, "learning_rate": 2.6837680534597974e-05, "loss": 0.0003, "step": 41500 }, { "epoch": 8.96, "learning_rate": 2.6870015089458934e-05, "loss": 0.0003, "step": 41550 }, { "epoch": 8.97, "learning_rate": 2.6902349644319896e-05, "loss": 0.0003, "step": 41600 }, { "epoch": 8.98, "learning_rate": 2.693468419918086e-05, "loss": 0.0004, "step": 41650 }, { "epoch": 8.99, "learning_rate": 2.6967018754041822e-05, "loss": 0.0002, "step": 41700 }, { "epoch": 9.0, "learning_rate": 2.699935330890278e-05, "loss": 0.0002, "step": 41750 }, { "epoch": 9.01, "learning_rate": 2.7031687863763744e-05, "loss": 0.0002, "step": 41800 }, { "epoch": 9.02, "learning_rate": 2.7064022418624703e-05, "loss": 0.0002, "step": 41850 }, { "epoch": 9.03, "learning_rate": 2.7096356973485666e-05, "loss": 0.0003, "step": 41900 }, { "epoch": 9.04, "learning_rate": 2.7128691528346625e-05, "loss": 0.0002, "step": 41950 }, { "epoch": 9.05, "learning_rate": 2.7161026083207588e-05, "loss": 0.0002, "step": 42000 }, { "epoch": 9.06, "learning_rate": 2.7193360638068547e-05, "loss": 0.0002, "step": 42050 }, { "epoch": 9.08, "learning_rate": 2.7225695192929513e-05, "loss": 0.0002, "step": 42100 }, { "epoch": 9.09, "learning_rate": 2.7258029747790473e-05, "loss": 0.0002, "step": 42150 }, { "epoch": 9.1, "learning_rate": 2.7290364302651435e-05, "loss": 0.0002, "step": 42200 }, { "epoch": 9.11, "learning_rate": 2.7322698857512395e-05, "loss": 0.0002, "step": 42250 }, { "epoch": 9.12, "learning_rate": 2.7355033412373357e-05, "loss": 0.0003, "step": 42300 }, { "epoch": 9.13, "learning_rate": 2.7387367967234317e-05, "loss": 0.0002, "step": 42350 }, { "epoch": 9.14, "learning_rate": 2.741970252209528e-05, "loss": 0.0002, "step": 42400 }, { "epoch": 9.15, "learning_rate": 2.745203707695624e-05, "loss": 0.0003, "step": 42450 }, { "epoch": 9.16, "learning_rate": 2.74843716318172e-05, "loss": 0.0002, "step": 42500 }, { "epoch": 9.17, "learning_rate": 2.7516706186678164e-05, "loss": 0.0002, "step": 42550 }, { "epoch": 9.18, "learning_rate": 2.7549040741539127e-05, "loss": 0.0002, "step": 42600 }, { "epoch": 9.19, "learning_rate": 2.7581375296400086e-05, "loss": 0.0002, "step": 42650 }, { "epoch": 9.2, "learning_rate": 2.761370985126105e-05, "loss": 0.0003, "step": 42700 }, { "epoch": 9.22, "learning_rate": 2.764604440612201e-05, "loss": 0.0002, "step": 42750 }, { "epoch": 9.23, "learning_rate": 2.767837896098297e-05, "loss": 0.0002, "step": 42800 }, { "epoch": 9.24, "learning_rate": 2.7710713515843933e-05, "loss": 0.0002, "step": 42850 }, { "epoch": 9.25, "learning_rate": 2.7743048070704893e-05, "loss": 0.0002, "step": 42900 }, { "epoch": 9.26, "learning_rate": 2.7775382625565855e-05, "loss": 0.0002, "step": 42950 }, { "epoch": 9.27, "learning_rate": 2.7807717180426818e-05, "loss": 0.0002, "step": 43000 }, { "epoch": 9.28, "learning_rate": 2.784005173528778e-05, "loss": 0.0003, "step": 43050 }, { "epoch": 9.29, "learning_rate": 2.787238629014874e-05, "loss": 0.0002, "step": 43100 }, { "epoch": 9.3, "learning_rate": 2.7904720845009703e-05, "loss": 0.0002, "step": 43150 }, { "epoch": 9.31, "learning_rate": 2.7937055399870662e-05, "loss": 0.0002, "step": 43200 }, { "epoch": 9.32, "learning_rate": 2.7969389954731625e-05, "loss": 0.0003, "step": 43250 }, { "epoch": 9.33, "learning_rate": 2.8001724509592584e-05, "loss": 0.0002, "step": 43300 }, { "epoch": 9.34, "learning_rate": 2.8034059064453547e-05, "loss": 0.0003, "step": 43350 }, { "epoch": 9.36, "learning_rate": 2.8066393619314506e-05, "loss": 0.0003, "step": 43400 }, { "epoch": 9.37, "learning_rate": 2.8098728174175472e-05, "loss": 0.0002, "step": 43450 }, { "epoch": 9.38, "learning_rate": 2.813106272903643e-05, "loss": 0.0002, "step": 43500 }, { "epoch": 9.39, "learning_rate": 2.8163397283897394e-05, "loss": 0.0003, "step": 43550 }, { "epoch": 9.4, "learning_rate": 2.8195731838758353e-05, "loss": 0.0003, "step": 43600 }, { "epoch": 9.41, "learning_rate": 2.8228066393619316e-05, "loss": 0.0003, "step": 43650 }, { "epoch": 9.42, "learning_rate": 2.8260400948480275e-05, "loss": 0.0003, "step": 43700 }, { "epoch": 9.43, "learning_rate": 2.8292735503341238e-05, "loss": 0.0003, "step": 43750 }, { "epoch": 9.44, "learning_rate": 2.8325070058202197e-05, "loss": 0.0002, "step": 43800 }, { "epoch": 9.45, "learning_rate": 2.835740461306316e-05, "loss": 0.0002, "step": 43850 }, { "epoch": 9.46, "learning_rate": 2.8389739167924123e-05, "loss": 0.0002, "step": 43900 }, { "epoch": 9.47, "learning_rate": 2.8422073722785086e-05, "loss": 0.0002, "step": 43950 }, { "epoch": 9.48, "learning_rate": 2.8454408277646045e-05, "loss": 0.0002, "step": 44000 }, { "epoch": 9.5, "learning_rate": 2.8486742832507008e-05, "loss": 0.0003, "step": 44050 }, { "epoch": 9.51, "learning_rate": 2.8519077387367967e-05, "loss": 0.0003, "step": 44100 }, { "epoch": 9.52, "learning_rate": 2.855141194222893e-05, "loss": 0.0002, "step": 44150 }, { "epoch": 9.53, "learning_rate": 2.858374649708989e-05, "loss": 0.0003, "step": 44200 }, { "epoch": 9.54, "learning_rate": 2.861608105195085e-05, "loss": 0.0002, "step": 44250 }, { "epoch": 9.55, "learning_rate": 2.8648415606811814e-05, "loss": 0.0002, "step": 44300 }, { "epoch": 9.56, "learning_rate": 2.8680750161672777e-05, "loss": 0.0002, "step": 44350 }, { "epoch": 9.57, "learning_rate": 2.8713084716533736e-05, "loss": 0.0003, "step": 44400 }, { "epoch": 9.58, "learning_rate": 2.87454192713947e-05, "loss": 0.0002, "step": 44450 }, { "epoch": 9.59, "learning_rate": 2.8777753826255658e-05, "loss": 0.0002, "step": 44500 }, { "epoch": 9.6, "learning_rate": 2.881008838111662e-05, "loss": 0.0002, "step": 44550 }, { "epoch": 9.61, "learning_rate": 2.884242293597758e-05, "loss": 0.0002, "step": 44600 }, { "epoch": 9.62, "learning_rate": 2.8874757490838543e-05, "loss": 0.0003, "step": 44650 }, { "epoch": 9.64, "learning_rate": 2.8907092045699502e-05, "loss": 0.0003, "step": 44700 }, { "epoch": 9.65, "learning_rate": 2.893942660056047e-05, "loss": 0.0003, "step": 44750 }, { "epoch": 9.66, "learning_rate": 2.8971761155421428e-05, "loss": 0.0002, "step": 44800 }, { "epoch": 9.67, "learning_rate": 2.900409571028239e-05, "loss": 0.0002, "step": 44850 }, { "epoch": 9.68, "learning_rate": 2.903643026514335e-05, "loss": 0.0002, "step": 44900 }, { "epoch": 9.69, "learning_rate": 2.9068764820004312e-05, "loss": 0.0002, "step": 44950 }, { "epoch": 9.7, "learning_rate": 2.910109937486527e-05, "loss": 0.0002, "step": 45000 }, { "epoch": 9.71, "learning_rate": 2.9133433929726234e-05, "loss": 0.0003, "step": 45050 }, { "epoch": 9.72, "learning_rate": 2.9165768484587194e-05, "loss": 0.0002, "step": 45100 }, { "epoch": 9.73, "learning_rate": 2.9198103039448156e-05, "loss": 0.0003, "step": 45150 }, { "epoch": 9.74, "learning_rate": 2.923043759430912e-05, "loss": 0.0003, "step": 45200 }, { "epoch": 9.75, "learning_rate": 2.9262772149170082e-05, "loss": 0.0003, "step": 45250 }, { "epoch": 9.77, "learning_rate": 2.929510670403104e-05, "loss": 0.0003, "step": 45300 }, { "epoch": 9.78, "learning_rate": 2.9327441258892004e-05, "loss": 0.0002, "step": 45350 }, { "epoch": 9.79, "learning_rate": 2.9359775813752966e-05, "loss": 0.0002, "step": 45400 }, { "epoch": 9.8, "learning_rate": 2.9392110368613926e-05, "loss": 0.0002, "step": 45450 }, { "epoch": 9.81, "learning_rate": 2.942444492347489e-05, "loss": 0.0002, "step": 45500 }, { "epoch": 9.82, "learning_rate": 2.9456779478335848e-05, "loss": 0.0003, "step": 45550 }, { "epoch": 9.83, "learning_rate": 2.948911403319681e-05, "loss": 0.0002, "step": 45600 }, { "epoch": 9.84, "learning_rate": 2.9521448588057773e-05, "loss": 0.0003, "step": 45650 }, { "epoch": 9.85, "learning_rate": 2.9553783142918736e-05, "loss": 0.0004, "step": 45700 }, { "epoch": 9.86, "learning_rate": 2.9586117697779695e-05, "loss": 0.0003, "step": 45750 }, { "epoch": 9.87, "learning_rate": 2.9618452252640658e-05, "loss": 0.0002, "step": 45800 }, { "epoch": 9.88, "learning_rate": 2.9650786807501617e-05, "loss": 0.0002, "step": 45850 }, { "epoch": 9.89, "learning_rate": 2.968312136236258e-05, "loss": 0.0003, "step": 45900 }, { "epoch": 9.91, "learning_rate": 2.971545591722354e-05, "loss": 0.0002, "step": 45950 }, { "epoch": 9.92, "learning_rate": 2.9747790472084502e-05, "loss": 0.0002, "step": 46000 }, { "epoch": 9.93, "learning_rate": 2.978012502694546e-05, "loss": 0.0002, "step": 46050 }, { "epoch": 9.94, "learning_rate": 2.9812459581806427e-05, "loss": 0.0003, "step": 46100 }, { "epoch": 9.95, "learning_rate": 2.9844794136667387e-05, "loss": 0.0003, "step": 46150 }, { "epoch": 9.96, "learning_rate": 2.987712869152835e-05, "loss": 0.0003, "step": 46200 }, { "epoch": 9.97, "learning_rate": 2.990946324638931e-05, "loss": 0.0003, "step": 46250 }, { "epoch": 9.98, "learning_rate": 2.994179780125027e-05, "loss": 0.0003, "step": 46300 }, { "epoch": 9.99, "learning_rate": 2.997413235611123e-05, "loss": 0.0003, "step": 46350 }, { "epoch": 10.0, "learning_rate": 2.9999999957535383e-05, "loss": 0.0003, "step": 46400 }, { "epoch": 10.01, "learning_rate": 2.9999998471273694e-05, "loss": 0.0002, "step": 46450 }, { "epoch": 10.02, "learning_rate": 2.9999994861781242e-05, "loss": 0.0002, "step": 46500 }, { "epoch": 10.03, "learning_rate": 2.9999989129058528e-05, "loss": 0.0002, "step": 46550 }, { "epoch": 10.05, "learning_rate": 2.999998127310636e-05, "loss": 0.0002, "step": 46600 }, { "epoch": 10.06, "learning_rate": 2.9999971293925855e-05, "loss": 0.0002, "step": 46650 }, { "epoch": 10.07, "learning_rate": 2.9999959191518427e-05, "loss": 0.0002, "step": 46700 }, { "epoch": 10.08, "learning_rate": 2.9999944965885782e-05, "loss": 0.0002, "step": 46750 }, { "epoch": 10.09, "learning_rate": 2.9999928617029946e-05, "loss": 0.0002, "step": 46800 }, { "epoch": 10.1, "learning_rate": 2.9999910144953217e-05, "loss": 0.0002, "step": 46850 }, { "epoch": 10.11, "learning_rate": 2.9999889549658223e-05, "loss": 0.0002, "step": 46900 }, { "epoch": 10.12, "learning_rate": 2.9999866831147875e-05, "loss": 0.0002, "step": 46950 }, { "epoch": 10.13, "learning_rate": 2.999984198942539e-05, "loss": 0.0002, "step": 47000 }, { "epoch": 10.14, "learning_rate": 2.999981502449428e-05, "loss": 0.0003, "step": 47050 }, { "epoch": 10.15, "learning_rate": 2.9999785936358368e-05, "loss": 0.0002, "step": 47100 }, { "epoch": 10.16, "learning_rate": 2.9999754725021766e-05, "loss": 0.0002, "step": 47150 }, { "epoch": 10.17, "learning_rate": 2.999972139048889e-05, "loss": 0.0002, "step": 47200 }, { "epoch": 10.19, "learning_rate": 2.9999685932764468e-05, "loss": 0.0003, "step": 47250 }, { "epoch": 10.2, "learning_rate": 2.9999648351853513e-05, "loss": 0.0002, "step": 47300 }, { "epoch": 10.21, "learning_rate": 2.9999608647761343e-05, "loss": 0.0003, "step": 47350 }, { "epoch": 10.22, "learning_rate": 2.9999566820493576e-05, "loss": 0.0002, "step": 47400 }, { "epoch": 10.23, "learning_rate": 2.9999522870056143e-05, "loss": 0.0003, "step": 47450 }, { "epoch": 10.24, "learning_rate": 2.9999476796455255e-05, "loss": 0.0002, "step": 47500 }, { "epoch": 10.25, "learning_rate": 2.999942859969744e-05, "loss": 0.0002, "step": 47550 }, { "epoch": 10.26, "learning_rate": 2.9999378279789513e-05, "loss": 0.0002, "step": 47600 }, { "epoch": 10.27, "learning_rate": 2.9999325836738604e-05, "loss": 0.0002, "step": 47650 }, { "epoch": 10.28, "learning_rate": 2.9999271270552135e-05, "loss": 0.0002, "step": 47700 }, { "epoch": 10.29, "learning_rate": 2.9999214581237827e-05, "loss": 0.0001, "step": 47750 }, { "epoch": 10.3, "learning_rate": 2.9999155768803708e-05, "loss": 0.0002, "step": 47800 }, { "epoch": 10.31, "learning_rate": 2.9999094833258097e-05, "loss": 0.0003, "step": 47850 }, { "epoch": 10.33, "learning_rate": 2.9999031774609625e-05, "loss": 0.0002, "step": 47900 }, { "epoch": 10.34, "learning_rate": 2.9998966592867215e-05, "loss": 0.0002, "step": 47950 }, { "epoch": 10.35, "learning_rate": 2.999889928804009e-05, "loss": 0.0002, "step": 48000 }, { "epoch": 10.36, "learning_rate": 2.9998829860137784e-05, "loss": 0.0002, "step": 48050 }, { "epoch": 10.37, "learning_rate": 2.9998758309170126e-05, "loss": 0.0002, "step": 48100 }, { "epoch": 10.38, "learning_rate": 2.9998684635147233e-05, "loss": 0.0002, "step": 48150 }, { "epoch": 10.39, "learning_rate": 2.999860883807954e-05, "loss": 0.0002, "step": 48200 }, { "epoch": 10.4, "learning_rate": 2.9998530917977777e-05, "loss": 0.0002, "step": 48250 }, { "epoch": 10.41, "learning_rate": 2.9998450874852973e-05, "loss": 0.0002, "step": 48300 }, { "epoch": 10.42, "learning_rate": 2.9998368708716458e-05, "loss": 0.0003, "step": 48350 }, { "epoch": 10.43, "learning_rate": 2.9998284419579857e-05, "loss": 0.0002, "step": 48400 }, { "epoch": 10.44, "learning_rate": 2.9998198007455114e-05, "loss": 0.0002, "step": 48450 }, { "epoch": 10.45, "learning_rate": 2.9998109472354446e-05, "loss": 0.0002, "step": 48500 }, { "epoch": 10.47, "learning_rate": 2.9998018814290393e-05, "loss": 0.0002, "step": 48550 }, { "epoch": 10.48, "learning_rate": 2.9997926033275785e-05, "loss": 0.0002, "step": 48600 }, { "epoch": 10.49, "learning_rate": 2.9997831129323758e-05, "loss": 0.0002, "step": 48650 }, { "epoch": 10.5, "learning_rate": 2.9997734102447743e-05, "loss": 0.0003, "step": 48700 }, { "epoch": 10.51, "learning_rate": 2.999763495266147e-05, "loss": 0.0003, "step": 48750 }, { "epoch": 10.52, "learning_rate": 2.9997533679978984e-05, "loss": 0.0002, "step": 48800 }, { "epoch": 10.53, "learning_rate": 2.999743028441461e-05, "loss": 0.0002, "step": 48850 }, { "epoch": 10.54, "learning_rate": 2.9997324765982988e-05, "loss": 0.0003, "step": 48900 }, { "epoch": 10.55, "learning_rate": 2.9997217124699053e-05, "loss": 0.0002, "step": 48950 }, { "epoch": 10.56, "learning_rate": 2.999710736057805e-05, "loss": 0.0002, "step": 49000 }, { "epoch": 10.57, "learning_rate": 2.99969954736355e-05, "loss": 0.0002, "step": 49050 }, { "epoch": 10.58, "learning_rate": 2.999688146388725e-05, "loss": 0.0002, "step": 49100 }, { "epoch": 10.59, "learning_rate": 2.9996765331349438e-05, "loss": 0.0002, "step": 49150 }, { "epoch": 10.61, "learning_rate": 2.9996647076038495e-05, "loss": 0.0002, "step": 49200 }, { "epoch": 10.62, "learning_rate": 2.9996526697971172e-05, "loss": 0.0003, "step": 49250 }, { "epoch": 10.63, "learning_rate": 2.99964041971645e-05, "loss": 0.0002, "step": 49300 }, { "epoch": 10.64, "learning_rate": 2.9996279573635817e-05, "loss": 0.0003, "step": 49350 }, { "epoch": 10.65, "learning_rate": 2.9996152827402774e-05, "loss": 0.0003, "step": 49400 }, { "epoch": 10.66, "learning_rate": 2.9996023958483298e-05, "loss": 0.0002, "step": 49450 }, { "epoch": 10.67, "learning_rate": 2.999589296689564e-05, "loss": 0.0002, "step": 49500 }, { "epoch": 10.68, "learning_rate": 2.9995759852658343e-05, "loss": 0.0002, "step": 49550 }, { "epoch": 10.69, "learning_rate": 2.9995624615790242e-05, "loss": 0.0002, "step": 49600 }, { "epoch": 10.7, "learning_rate": 2.999548725631048e-05, "loss": 0.0002, "step": 49650 }, { "epoch": 10.71, "learning_rate": 2.9995347774238508e-05, "loss": 0.0002, "step": 49700 }, { "epoch": 10.72, "learning_rate": 2.999520616959406e-05, "loss": 0.0002, "step": 49750 }, { "epoch": 10.74, "learning_rate": 2.9995062442397187e-05, "loss": 0.0002, "step": 49800 }, { "epoch": 10.75, "learning_rate": 2.999491659266823e-05, "loss": 0.0002, "step": 49850 }, { "epoch": 10.76, "learning_rate": 2.9994768620427834e-05, "loss": 0.0002, "step": 49900 }, { "epoch": 10.77, "learning_rate": 2.9994618525696943e-05, "loss": 0.0002, "step": 49950 }, { "epoch": 10.78, "learning_rate": 2.9994466308496805e-05, "loss": 0.0002, "step": 50000 }, { "epoch": 10.79, "learning_rate": 2.9994311968848965e-05, "loss": 0.0002, "step": 50050 }, { "epoch": 10.8, "learning_rate": 2.999415550677527e-05, "loss": 0.0003, "step": 50100 }, { "epoch": 10.81, "learning_rate": 2.999399692229787e-05, "loss": 0.0002, "step": 50150 }, { "epoch": 10.82, "learning_rate": 2.9993836215439207e-05, "loss": 0.0003, "step": 50200 }, { "epoch": 10.83, "learning_rate": 2.9993673386222033e-05, "loss": 0.0002, "step": 50250 }, { "epoch": 10.84, "learning_rate": 2.9993508434669396e-05, "loss": 0.0003, "step": 50300 }, { "epoch": 10.85, "learning_rate": 2.999334136080464e-05, "loss": 0.0002, "step": 50350 }, { "epoch": 10.86, "learning_rate": 2.9993172164651417e-05, "loss": 0.0002, "step": 50400 }, { "epoch": 10.88, "learning_rate": 2.999300084623368e-05, "loss": 0.0003, "step": 50450 }, { "epoch": 10.89, "learning_rate": 2.9992827405575675e-05, "loss": 0.0002, "step": 50500 }, { "epoch": 10.9, "learning_rate": 2.999265184270195e-05, "loss": 0.0002, "step": 50550 }, { "epoch": 10.91, "learning_rate": 2.999247415763736e-05, "loss": 0.0002, "step": 50600 }, { "epoch": 10.92, "learning_rate": 2.999229435040706e-05, "loss": 0.0002, "step": 50650 }, { "epoch": 10.93, "learning_rate": 2.9992112421036497e-05, "loss": 0.0003, "step": 50700 }, { "epoch": 10.94, "learning_rate": 2.9991928369551415e-05, "loss": 0.0003, "step": 50750 }, { "epoch": 10.95, "learning_rate": 2.999174219597788e-05, "loss": 0.0002, "step": 50800 }, { "epoch": 10.96, "learning_rate": 2.999155390034223e-05, "loss": 0.0002, "step": 50850 }, { "epoch": 10.97, "learning_rate": 2.9991363482671135e-05, "loss": 0.0002, "step": 50900 }, { "epoch": 10.98, "learning_rate": 2.9991170942991534e-05, "loss": 0.0002, "step": 50950 }, { "epoch": 10.99, "learning_rate": 2.999097628133069e-05, "loss": 0.0002, "step": 51000 }, { "epoch": 11.0, "learning_rate": 2.999077949771615e-05, "loss": 0.0002, "step": 51050 }, { "epoch": 11.02, "learning_rate": 2.9990580592175776e-05, "loss": 0.0001, "step": 51100 }, { "epoch": 11.03, "learning_rate": 2.9990379564737716e-05, "loss": 0.0002, "step": 51150 }, { "epoch": 11.04, "learning_rate": 2.9990176415430427e-05, "loss": 0.0001, "step": 51200 }, { "epoch": 11.05, "learning_rate": 2.9989971144282667e-05, "loss": 0.0002, "step": 51250 }, { "epoch": 11.06, "learning_rate": 2.9989763751323494e-05, "loss": 0.0002, "step": 51300 }, { "epoch": 11.07, "learning_rate": 2.9989554236582254e-05, "loss": 0.0002, "step": 51350 }, { "epoch": 11.08, "learning_rate": 2.9989342600088615e-05, "loss": 0.0003, "step": 51400 }, { "epoch": 11.09, "learning_rate": 2.9989128841872525e-05, "loss": 0.0002, "step": 51450 }, { "epoch": 11.1, "learning_rate": 2.998891296196425e-05, "loss": 0.0002, "step": 51500 }, { "epoch": 11.11, "learning_rate": 2.9988694960394338e-05, "loss": 0.0002, "step": 51550 }, { "epoch": 11.12, "learning_rate": 2.9988474837193655e-05, "loss": 0.0002, "step": 51600 }, { "epoch": 11.13, "learning_rate": 2.998825259239335e-05, "loss": 0.0001, "step": 51650 }, { "epoch": 11.14, "learning_rate": 2.9988028226024892e-05, "loss": 0.0002, "step": 51700 }, { "epoch": 11.16, "learning_rate": 2.9987801738120035e-05, "loss": 0.0002, "step": 51750 }, { "epoch": 11.17, "learning_rate": 2.9987573128710837e-05, "loss": 0.0001, "step": 51800 }, { "epoch": 11.18, "learning_rate": 2.9987342397829662e-05, "loss": 0.0002, "step": 51850 }, { "epoch": 11.19, "learning_rate": 2.998710954550916e-05, "loss": 0.0002, "step": 51900 }, { "epoch": 11.2, "learning_rate": 2.9986874571782305e-05, "loss": 0.0002, "step": 51950 }, { "epoch": 11.21, "learning_rate": 2.998663747668234e-05, "loss": 0.0002, "step": 52000 }, { "epoch": 11.22, "learning_rate": 2.998639826024284e-05, "loss": 0.0002, "step": 52050 }, { "epoch": 11.23, "learning_rate": 2.9986156922497658e-05, "loss": 0.0002, "step": 52100 }, { "epoch": 11.24, "learning_rate": 2.9985913463480958e-05, "loss": 0.0001, "step": 52150 }, { "epoch": 11.25, "learning_rate": 2.9985667883227207e-05, "loss": 0.0002, "step": 52200 }, { "epoch": 11.26, "learning_rate": 2.9985420181771154e-05, "loss": 0.0002, "step": 52250 }, { "epoch": 11.27, "learning_rate": 2.998517035914787e-05, "loss": 0.0002, "step": 52300 }, { "epoch": 11.28, "learning_rate": 2.9984918415392714e-05, "loss": 0.0002, "step": 52350 }, { "epoch": 11.3, "learning_rate": 2.9984664350541345e-05, "loss": 0.0001, "step": 52400 }, { "epoch": 11.31, "learning_rate": 2.998440816462973e-05, "loss": 0.0001, "step": 52450 }, { "epoch": 11.32, "learning_rate": 2.9984149857694138e-05, "loss": 0.0002, "step": 52500 }, { "epoch": 11.33, "learning_rate": 2.9983889429771117e-05, "loss": 0.0002, "step": 52550 }, { "epoch": 11.34, "learning_rate": 2.998362688089754e-05, "loss": 0.0001, "step": 52600 }, { "epoch": 11.35, "learning_rate": 2.9983362211110573e-05, "loss": 0.0001, "step": 52650 }, { "epoch": 11.36, "learning_rate": 2.9983095420447674e-05, "loss": 0.0002, "step": 52700 }, { "epoch": 11.37, "learning_rate": 2.9982826508946605e-05, "loss": 0.0002, "step": 52750 }, { "epoch": 11.38, "learning_rate": 2.9982555476645436e-05, "loss": 0.0002, "step": 52800 }, { "epoch": 11.39, "learning_rate": 2.9982282323582527e-05, "loss": 0.0002, "step": 52850 }, { "epoch": 11.4, "learning_rate": 2.998200704979655e-05, "loss": 0.0001, "step": 52900 }, { "epoch": 11.41, "learning_rate": 2.9981729655326458e-05, "loss": 0.0002, "step": 52950 }, { "epoch": 11.42, "learning_rate": 2.9981450140211523e-05, "loss": 0.0002, "step": 53000 }, { "epoch": 11.44, "learning_rate": 2.9981168504491308e-05, "loss": 0.0002, "step": 53050 }, { "epoch": 11.45, "learning_rate": 2.9980884748205682e-05, "loss": 0.0001, "step": 53100 }, { "epoch": 11.46, "learning_rate": 2.9980598871394804e-05, "loss": 0.0001, "step": 53150 }, { "epoch": 11.47, "learning_rate": 2.9980310874099142e-05, "loss": 0.0002, "step": 53200 }, { "epoch": 11.48, "learning_rate": 2.9980020756359465e-05, "loss": 0.0002, "step": 53250 }, { "epoch": 11.49, "learning_rate": 2.9979728518216834e-05, "loss": 0.0002, "step": 53300 }, { "epoch": 11.5, "learning_rate": 2.997943415971262e-05, "loss": 0.0002, "step": 53350 }, { "epoch": 11.51, "learning_rate": 2.9979137680888482e-05, "loss": 0.0002, "step": 53400 }, { "epoch": 11.52, "learning_rate": 2.9978839081786396e-05, "loss": 0.0002, "step": 53450 }, { "epoch": 11.53, "learning_rate": 2.997853836244862e-05, "loss": 0.0001, "step": 53500 }, { "epoch": 11.54, "learning_rate": 2.9978235522917717e-05, "loss": 0.0002, "step": 53550 }, { "epoch": 11.55, "learning_rate": 2.9977930563236566e-05, "loss": 0.0001, "step": 53600 }, { "epoch": 11.56, "learning_rate": 2.9977623483448326e-05, "loss": 0.0002, "step": 53650 }, { "epoch": 11.58, "learning_rate": 2.9977314283596467e-05, "loss": 0.0002, "step": 53700 }, { "epoch": 11.59, "learning_rate": 2.9977002963724753e-05, "loss": 0.0002, "step": 53750 }, { "epoch": 11.6, "learning_rate": 2.997668952387725e-05, "loss": 0.0002, "step": 53800 }, { "epoch": 11.61, "learning_rate": 2.997637396409833e-05, "loss": 0.0002, "step": 53850 }, { "epoch": 11.62, "learning_rate": 2.9976056284432656e-05, "loss": 0.0002, "step": 53900 }, { "epoch": 11.63, "learning_rate": 2.9975736484925197e-05, "loss": 0.0002, "step": 53950 }, { "epoch": 11.64, "learning_rate": 2.9975414565621215e-05, "loss": 0.0002, "step": 54000 }, { "epoch": 11.65, "learning_rate": 2.9975090526566285e-05, "loss": 0.0002, "step": 54050 }, { "epoch": 11.66, "learning_rate": 2.9974764367806268e-05, "loss": 0.0002, "step": 54100 }, { "epoch": 11.67, "learning_rate": 2.9974436089387337e-05, "loss": 0.0002, "step": 54150 }, { "epoch": 11.68, "learning_rate": 2.9974105691355957e-05, "loss": 0.0002, "step": 54200 }, { "epoch": 11.69, "learning_rate": 2.9973773173758892e-05, "loss": 0.0001, "step": 54250 }, { "epoch": 11.71, "learning_rate": 2.9973438536643217e-05, "loss": 0.0002, "step": 54300 }, { "epoch": 11.72, "learning_rate": 2.9973101780056292e-05, "loss": 0.0002, "step": 54350 }, { "epoch": 11.73, "learning_rate": 2.9972762904045788e-05, "loss": 0.0002, "step": 54400 }, { "epoch": 11.74, "learning_rate": 2.997242190865967e-05, "loss": 0.0003, "step": 54450 }, { "epoch": 11.75, "learning_rate": 2.997207879394621e-05, "loss": 0.0002, "step": 54500 }, { "epoch": 11.76, "learning_rate": 2.9971733559953976e-05, "loss": 0.0002, "step": 54550 }, { "epoch": 11.77, "learning_rate": 2.9971386206731825e-05, "loss": 0.0002, "step": 54600 }, { "epoch": 11.78, "learning_rate": 2.9971036734328935e-05, "loss": 0.0002, "step": 54650 }, { "epoch": 11.79, "learning_rate": 2.9970685142794772e-05, "loss": 0.0002, "step": 54700 }, { "epoch": 11.8, "learning_rate": 2.9970331432179098e-05, "loss": 0.0002, "step": 54750 }, { "epoch": 11.81, "learning_rate": 2.996997560253199e-05, "loss": 0.0002, "step": 54800 }, { "epoch": 11.82, "learning_rate": 2.99696176539038e-05, "loss": 0.0001, "step": 54850 }, { "epoch": 11.83, "learning_rate": 2.9969257586345212e-05, "loss": 0.0001, "step": 54900 }, { "epoch": 11.85, "learning_rate": 2.9968895399907182e-05, "loss": 0.0002, "step": 54950 }, { "epoch": 11.86, "learning_rate": 2.996853109464098e-05, "loss": 0.0002, "step": 55000 }, { "epoch": 11.87, "learning_rate": 2.996816467059817e-05, "loss": 0.0001, "step": 55050 }, { "epoch": 11.88, "learning_rate": 2.9967796127830628e-05, "loss": 0.0002, "step": 55100 }, { "epoch": 11.89, "learning_rate": 2.996742546639051e-05, "loss": 0.0002, "step": 55150 }, { "epoch": 11.9, "learning_rate": 2.996705268633029e-05, "loss": 0.0002, "step": 55200 }, { "epoch": 11.91, "learning_rate": 2.9966677787702734e-05, "loss": 0.0002, "step": 55250 }, { "epoch": 11.92, "learning_rate": 2.99663007705609e-05, "loss": 0.0002, "step": 55300 }, { "epoch": 11.93, "learning_rate": 2.996592163495817e-05, "loss": 0.0002, "step": 55350 }, { "epoch": 11.94, "learning_rate": 2.996554038094819e-05, "loss": 0.0002, "step": 55400 }, { "epoch": 11.95, "learning_rate": 2.9965157008584945e-05, "loss": 0.0001, "step": 55450 }, { "epoch": 11.96, "learning_rate": 2.9964771517922692e-05, "loss": 0.0002, "step": 55500 }, { "epoch": 11.97, "learning_rate": 2.9964383909015997e-05, "loss": 0.0003, "step": 55550 }, { "epoch": 11.99, "learning_rate": 2.9963994181919722e-05, "loss": 0.0002, "step": 55600 }, { "epoch": 12.0, "learning_rate": 2.996360233668904e-05, "loss": 0.0002, "step": 55650 }, { "epoch": 12.01, "learning_rate": 2.9963208373379414e-05, "loss": 0.0002, "step": 55700 }, { "epoch": 12.02, "learning_rate": 2.9962812292046603e-05, "loss": 0.0001, "step": 55750 }, { "epoch": 12.03, "learning_rate": 2.9962414092746684e-05, "loss": 0.0001, "step": 55800 }, { "epoch": 12.04, "learning_rate": 2.996201377553601e-05, "loss": 0.0002, "step": 55850 }, { "epoch": 12.05, "learning_rate": 2.9961611340471244e-05, "loss": 0.0001, "step": 55900 }, { "epoch": 12.06, "learning_rate": 2.996120678760936e-05, "loss": 0.0002, "step": 55950 }, { "epoch": 12.07, "learning_rate": 2.996080011700762e-05, "loss": 0.0001, "step": 56000 }, { "epoch": 12.08, "learning_rate": 2.9960391328723584e-05, "loss": 0.0002, "step": 56050 }, { "epoch": 12.09, "learning_rate": 2.9959980422815117e-05, "loss": 0.0002, "step": 56100 }, { "epoch": 12.1, "learning_rate": 2.995956739934038e-05, "loss": 0.0001, "step": 56150 }, { "epoch": 12.11, "learning_rate": 2.995915225835784e-05, "loss": 0.0001, "step": 56200 }, { "epoch": 12.13, "learning_rate": 2.9958734999926255e-05, "loss": 0.0001, "step": 56250 }, { "epoch": 12.14, "learning_rate": 2.9958315624104692e-05, "loss": 0.0001, "step": 56300 }, { "epoch": 12.15, "learning_rate": 2.995789413095251e-05, "loss": 0.0002, "step": 56350 }, { "epoch": 12.16, "learning_rate": 2.995747052052937e-05, "loss": 0.0001, "step": 56400 }, { "epoch": 12.17, "learning_rate": 2.995704479289524e-05, "loss": 0.0001, "step": 56450 }, { "epoch": 12.18, "learning_rate": 2.995661694811038e-05, "loss": 0.0001, "step": 56500 }, { "epoch": 12.19, "learning_rate": 2.995618698623534e-05, "loss": 0.0001, "step": 56550 }, { "epoch": 12.2, "learning_rate": 2.995575490733099e-05, "loss": 0.0001, "step": 56600 }, { "epoch": 12.21, "learning_rate": 2.9955320711458486e-05, "loss": 0.0001, "step": 56650 }, { "epoch": 12.22, "learning_rate": 2.9954884398679292e-05, "loss": 0.0001, "step": 56700 }, { "epoch": 12.23, "learning_rate": 2.9954445969055167e-05, "loss": 0.0001, "step": 56750 }, { "epoch": 12.24, "learning_rate": 2.995400542264817e-05, "loss": 0.0002, "step": 56800 }, { "epoch": 12.25, "learning_rate": 2.9953562759520658e-05, "loss": 0.0001, "step": 56850 }, { "epoch": 12.27, "learning_rate": 2.9953117979735288e-05, "loss": 0.0002, "step": 56900 }, { "epoch": 12.28, "learning_rate": 2.9952671083355018e-05, "loss": 0.0002, "step": 56950 }, { "epoch": 12.29, "learning_rate": 2.9952222070443113e-05, "loss": 0.0002, "step": 57000 }, { "epoch": 12.3, "learning_rate": 2.995177094106312e-05, "loss": 0.0002, "step": 57050 }, { "epoch": 12.31, "learning_rate": 2.9951317695278906e-05, "loss": 0.0001, "step": 57100 }, { "epoch": 12.32, "learning_rate": 2.995086233315462e-05, "loss": 0.0002, "step": 57150 }, { "epoch": 12.33, "learning_rate": 2.995040485475472e-05, "loss": 0.0001, "step": 57200 }, { "epoch": 12.34, "learning_rate": 2.994994526014396e-05, "loss": 0.0002, "step": 57250 }, { "epoch": 12.35, "learning_rate": 2.9949483549387398e-05, "loss": 0.0002, "step": 57300 }, { "epoch": 12.36, "learning_rate": 2.9949019722550386e-05, "loss": 0.0001, "step": 57350 }, { "epoch": 12.37, "learning_rate": 2.994855377969858e-05, "loss": 0.0001, "step": 57400 }, { "epoch": 12.38, "learning_rate": 2.994808572089793e-05, "loss": 0.0002, "step": 57450 }, { "epoch": 12.39, "learning_rate": 2.99476155462147e-05, "loss": 0.0001, "step": 57500 }, { "epoch": 12.41, "learning_rate": 2.9947143255715426e-05, "loss": 0.0002, "step": 57550 }, { "epoch": 12.42, "learning_rate": 2.9946668849466973e-05, "loss": 0.0002, "step": 57600 }, { "epoch": 12.43, "learning_rate": 2.9946192327536486e-05, "loss": 0.0002, "step": 57650 }, { "epoch": 12.44, "learning_rate": 2.994571368999142e-05, "loss": 0.0002, "step": 57700 }, { "epoch": 12.45, "learning_rate": 2.994523293689952e-05, "loss": 0.0002, "step": 57750 }, { "epoch": 12.46, "learning_rate": 2.994475006832885e-05, "loss": 0.0002, "step": 57800 }, { "epoch": 12.47, "learning_rate": 2.994426508434774e-05, "loss": 0.0001, "step": 57850 }, { "epoch": 12.48, "learning_rate": 2.994377798502485e-05, "loss": 0.0002, "step": 57900 }, { "epoch": 12.49, "learning_rate": 2.994328877042913e-05, "loss": 0.0001, "step": 57950 }, { "epoch": 12.5, "learning_rate": 2.9942797440629818e-05, "loss": 0.0001, "step": 58000 }, { "epoch": 12.51, "learning_rate": 2.994230399569647e-05, "loss": 0.0002, "step": 58050 }, { "epoch": 12.52, "learning_rate": 2.9941808435698932e-05, "loss": 0.0001, "step": 58100 }, { "epoch": 12.54, "learning_rate": 2.994131076070735e-05, "loss": 0.0002, "step": 58150 }, { "epoch": 12.55, "learning_rate": 2.994081097079216e-05, "loss": 0.0002, "step": 58200 }, { "epoch": 12.56, "learning_rate": 2.9940309066024114e-05, "loss": 0.0001, "step": 58250 }, { "epoch": 12.57, "learning_rate": 2.993980504647426e-05, "loss": 0.0002, "step": 58300 }, { "epoch": 12.58, "learning_rate": 2.9939298912213936e-05, "loss": 0.0001, "step": 58350 }, { "epoch": 12.59, "learning_rate": 2.9938790663314785e-05, "loss": 0.0001, "step": 58400 }, { "epoch": 12.6, "learning_rate": 2.9938280299848748e-05, "loss": 0.0002, "step": 58450 }, { "epoch": 12.61, "learning_rate": 2.9937767821888072e-05, "loss": 0.0002, "step": 58500 }, { "epoch": 12.62, "learning_rate": 2.993725322950529e-05, "loss": 0.0001, "step": 58550 }, { "epoch": 12.63, "learning_rate": 2.9936736522773244e-05, "loss": 0.0002, "step": 58600 }, { "epoch": 12.64, "learning_rate": 2.9936217701765075e-05, "loss": 0.0001, "step": 58650 }, { "epoch": 12.65, "learning_rate": 2.9935696766554227e-05, "loss": 0.0002, "step": 58700 }, { "epoch": 12.66, "learning_rate": 2.993517371721443e-05, "loss": 0.0002, "step": 58750 }, { "epoch": 12.68, "learning_rate": 2.9934648553819717e-05, "loss": 0.0001, "step": 58800 }, { "epoch": 12.69, "learning_rate": 2.9934121276444435e-05, "loss": 0.0002, "step": 58850 }, { "epoch": 12.7, "learning_rate": 2.993359188516321e-05, "loss": 0.0002, "step": 58900 }, { "epoch": 12.71, "learning_rate": 2.9933060380050985e-05, "loss": 0.0002, "step": 58950 }, { "epoch": 12.72, "learning_rate": 2.9932526761182986e-05, "loss": 0.0002, "step": 59000 }, { "epoch": 12.73, "learning_rate": 2.9931991028634755e-05, "loss": 0.0002, "step": 59050 }, { "epoch": 12.74, "learning_rate": 2.9931453182482116e-05, "loss": 0.0002, "step": 59100 }, { "epoch": 12.75, "learning_rate": 2.9930913222801202e-05, "loss": 0.0001, "step": 59150 }, { "epoch": 12.76, "learning_rate": 2.9930371149668446e-05, "loss": 0.0001, "step": 59200 }, { "epoch": 12.77, "learning_rate": 2.9929826963160582e-05, "loss": 0.0002, "step": 59250 }, { "epoch": 12.78, "learning_rate": 2.992928066335463e-05, "loss": 0.0001, "step": 59300 }, { "epoch": 12.79, "learning_rate": 2.992873225032792e-05, "loss": 0.0002, "step": 59350 }, { "epoch": 12.8, "learning_rate": 2.9928181724158083e-05, "loss": 0.0002, "step": 59400 }, { "epoch": 12.82, "learning_rate": 2.9927629084923046e-05, "loss": 0.0002, "step": 59450 }, { "epoch": 12.83, "learning_rate": 2.992707433270103e-05, "loss": 0.0001, "step": 59500 }, { "epoch": 12.84, "learning_rate": 2.992651746757056e-05, "loss": 0.0001, "step": 59550 }, { "epoch": 12.85, "learning_rate": 2.9925958489610464e-05, "loss": 0.0001, "step": 59600 }, { "epoch": 12.86, "learning_rate": 2.9925397398899858e-05, "loss": 0.0001, "step": 59650 }, { "epoch": 12.87, "learning_rate": 2.992483419551817e-05, "loss": 0.0002, "step": 59700 }, { "epoch": 12.88, "learning_rate": 2.9924268879545118e-05, "loss": 0.0001, "step": 59750 }, { "epoch": 12.89, "learning_rate": 2.992370145106072e-05, "loss": 0.0002, "step": 59800 }, { "epoch": 12.9, "learning_rate": 2.9923131910145297e-05, "loss": 0.0002, "step": 59850 }, { "epoch": 12.91, "learning_rate": 2.9922560256879466e-05, "loss": 0.0001, "step": 59900 }, { "epoch": 12.92, "learning_rate": 2.9921986491344142e-05, "loss": 0.0002, "step": 59950 }, { "epoch": 12.93, "learning_rate": 2.9921410613620545e-05, "loss": 0.0002, "step": 60000 }, { "epoch": 12.94, "learning_rate": 2.9920832623790188e-05, "loss": 0.0002, "step": 60050 }, { "epoch": 12.96, "learning_rate": 2.9920252521934886e-05, "loss": 0.0001, "step": 60100 }, { "epoch": 12.97, "learning_rate": 2.9919670308136748e-05, "loss": 0.0001, "step": 60150 }, { "epoch": 12.98, "learning_rate": 2.9919085982478183e-05, "loss": 0.0001, "step": 60200 }, { "epoch": 12.99, "learning_rate": 2.9918499545041912e-05, "loss": 0.0002, "step": 60250 }, { "epoch": 13.0, "learning_rate": 2.991791099591094e-05, "loss": 0.0002, "step": 60300 }, { "epoch": 13.01, "learning_rate": 2.9917320335168566e-05, "loss": 0.0002, "step": 60350 }, { "epoch": 13.02, "learning_rate": 2.9916727562898413e-05, "loss": 0.0001, "step": 60400 }, { "epoch": 13.03, "learning_rate": 2.9916132679184377e-05, "loss": 0.0001, "step": 60450 }, { "epoch": 13.04, "learning_rate": 2.9915535684110664e-05, "loss": 0.0001, "step": 60500 }, { "epoch": 13.05, "learning_rate": 2.991493657776178e-05, "loss": 0.0001, "step": 60550 }, { "epoch": 13.06, "learning_rate": 2.9914335360222528e-05, "loss": 0.0001, "step": 60600 }, { "epoch": 13.07, "learning_rate": 2.9913732031578007e-05, "loss": 0.0001, "step": 60650 }, { "epoch": 13.08, "learning_rate": 2.991312659191362e-05, "loss": 0.0001, "step": 60700 }, { "epoch": 13.1, "learning_rate": 2.9912519041315066e-05, "loss": 0.0001, "step": 60750 }, { "epoch": 13.11, "learning_rate": 2.9911909379868346e-05, "loss": 0.0001, "step": 60800 }, { "epoch": 13.12, "learning_rate": 2.9911297607659747e-05, "loss": 0.0001, "step": 60850 }, { "epoch": 13.13, "learning_rate": 2.9910683724775872e-05, "loss": 0.0002, "step": 60900 }, { "epoch": 13.14, "learning_rate": 2.9910067731303615e-05, "loss": 0.0001, "step": 60950 }, { "epoch": 13.15, "learning_rate": 2.990944962733017e-05, "loss": 0.0001, "step": 61000 }, { "epoch": 13.16, "learning_rate": 2.9908829412943022e-05, "loss": 0.0001, "step": 61050 }, { "epoch": 13.17, "learning_rate": 2.9908207088229976e-05, "loss": 0.0001, "step": 61100 }, { "epoch": 13.18, "learning_rate": 2.9907582653279103e-05, "loss": 0.0001, "step": 61150 }, { "epoch": 13.19, "learning_rate": 2.9906956108178802e-05, "loss": 0.0001, "step": 61200 }, { "epoch": 13.2, "learning_rate": 2.990632745301776e-05, "loss": 0.0002, "step": 61250 }, { "epoch": 13.21, "learning_rate": 2.9905696687884956e-05, "loss": 0.0001, "step": 61300 }, { "epoch": 13.22, "learning_rate": 2.990506381286968e-05, "loss": 0.0001, "step": 61350 }, { "epoch": 13.24, "learning_rate": 2.990442882806151e-05, "loss": 0.0001, "step": 61400 }, { "epoch": 13.25, "learning_rate": 2.9903791733550335e-05, "loss": 0.0001, "step": 61450 }, { "epoch": 13.26, "learning_rate": 2.9903152529426326e-05, "loss": 0.0001, "step": 61500 }, { "epoch": 13.27, "learning_rate": 2.9902511215779967e-05, "loss": 0.0002, "step": 61550 }, { "epoch": 13.28, "learning_rate": 2.9901867792702034e-05, "loss": 0.0001, "step": 61600 }, { "epoch": 13.29, "learning_rate": 2.99012222602836e-05, "loss": 0.0001, "step": 61650 }, { "epoch": 13.3, "learning_rate": 2.990057461861604e-05, "loss": 0.0002, "step": 61700 }, { "epoch": 13.31, "learning_rate": 2.989992486779103e-05, "loss": 0.0001, "step": 61750 }, { "epoch": 13.32, "learning_rate": 2.989927300790054e-05, "loss": 0.0001, "step": 61800 }, { "epoch": 13.33, "learning_rate": 2.9898619039036842e-05, "loss": 0.0002, "step": 61850 }, { "epoch": 13.34, "learning_rate": 2.9897962961292498e-05, "loss": 0.0002, "step": 61900 }, { "epoch": 13.35, "learning_rate": 2.989730477476038e-05, "loss": 0.0002, "step": 61950 }, { "epoch": 13.36, "learning_rate": 2.9896644479533654e-05, "loss": 0.0002, "step": 62000 }, { "epoch": 13.38, "learning_rate": 2.9895982075705783e-05, "loss": 0.0002, "step": 62050 }, { "epoch": 13.39, "learning_rate": 2.9895317563370525e-05, "loss": 0.0002, "step": 62100 }, { "epoch": 13.4, "learning_rate": 2.9894650942621947e-05, "loss": 0.0001, "step": 62150 }, { "epoch": 13.41, "learning_rate": 2.989398221355441e-05, "loss": 0.0002, "step": 62200 }, { "epoch": 13.42, "learning_rate": 2.989331137626256e-05, "loss": 0.0001, "step": 62250 }, { "epoch": 13.43, "learning_rate": 2.9892638430841367e-05, "loss": 0.0002, "step": 62300 }, { "epoch": 13.44, "learning_rate": 2.9891963377386077e-05, "loss": 0.0001, "step": 62350 }, { "epoch": 13.45, "learning_rate": 2.9891286215992248e-05, "loss": 0.0001, "step": 62400 }, { "epoch": 13.46, "learning_rate": 2.9890606946755723e-05, "loss": 0.0001, "step": 62450 }, { "epoch": 13.47, "learning_rate": 2.9889925569772665e-05, "loss": 0.0001, "step": 62500 }, { "epoch": 13.48, "learning_rate": 2.9889242085139506e-05, "loss": 0.0002, "step": 62550 }, { "epoch": 13.49, "learning_rate": 2.9888556492953013e-05, "loss": 0.0002, "step": 62600 }, { "epoch": 13.51, "learning_rate": 2.9887868793310206e-05, "loss": 0.0001, "step": 62650 }, { "epoch": 13.52, "learning_rate": 2.9887178986308448e-05, "loss": 0.0001, "step": 62700 }, { "epoch": 13.53, "learning_rate": 2.9886487072045374e-05, "loss": 0.0001, "step": 62750 }, { "epoch": 13.54, "learning_rate": 2.988579305061892e-05, "loss": 0.0001, "step": 62800 }, { "epoch": 13.55, "learning_rate": 2.9885096922127323e-05, "loss": 0.0001, "step": 62850 }, { "epoch": 13.56, "learning_rate": 2.9884398686669128e-05, "loss": 0.0001, "step": 62900 }, { "epoch": 13.57, "learning_rate": 2.9883698344343163e-05, "loss": 0.0001, "step": 62950 }, { "epoch": 13.58, "learning_rate": 2.988299589524856e-05, "loss": 0.0001, "step": 63000 }, { "epoch": 13.59, "learning_rate": 2.988229133948475e-05, "loss": 0.0002, "step": 63050 }, { "epoch": 13.6, "learning_rate": 2.988158467715147e-05, "loss": 0.0002, "step": 63100 }, { "epoch": 13.61, "learning_rate": 2.9880875908348733e-05, "loss": 0.0001, "step": 63150 }, { "epoch": 13.62, "learning_rate": 2.9880165033176874e-05, "loss": 0.0001, "step": 63200 }, { "epoch": 13.63, "learning_rate": 2.9879452051736518e-05, "loss": 0.0001, "step": 63250 }, { "epoch": 13.65, "learning_rate": 2.987873696412858e-05, "loss": 0.0001, "step": 63300 }, { "epoch": 13.66, "learning_rate": 2.987801977045428e-05, "loss": 0.0002, "step": 63350 }, { "epoch": 13.67, "learning_rate": 2.987730047081514e-05, "loss": 0.0002, "step": 63400 }, { "epoch": 13.68, "learning_rate": 2.9876579065312973e-05, "loss": 0.0001, "step": 63450 }, { "epoch": 13.69, "learning_rate": 2.9875855554049895e-05, "loss": 0.0001, "step": 63500 }, { "epoch": 13.7, "learning_rate": 2.987512993712832e-05, "loss": 0.0001, "step": 63550 }, { "epoch": 13.71, "learning_rate": 2.9874402214650955e-05, "loss": 0.0002, "step": 63600 }, { "epoch": 13.72, "learning_rate": 2.987367238672081e-05, "loss": 0.0001, "step": 63650 }, { "epoch": 13.73, "learning_rate": 2.9872940453441185e-05, "loss": 0.0001, "step": 63700 }, { "epoch": 13.74, "learning_rate": 2.9872206414915696e-05, "loss": 0.0001, "step": 63750 }, { "epoch": 13.75, "learning_rate": 2.9871470271248232e-05, "loss": 0.0002, "step": 63800 }, { "epoch": 13.76, "learning_rate": 2.9870732022543002e-05, "loss": 0.0001, "step": 63850 }, { "epoch": 13.77, "learning_rate": 2.9869991668904503e-05, "loss": 0.0001, "step": 63900 }, { "epoch": 13.79, "learning_rate": 2.986924921043753e-05, "loss": 0.0001, "step": 63950 }, { "epoch": 13.8, "learning_rate": 2.986850464724718e-05, "loss": 0.0001, "step": 64000 }, { "epoch": 13.81, "learning_rate": 2.986775797943884e-05, "loss": 0.0002, "step": 64050 }, { "epoch": 13.82, "learning_rate": 2.9867009207118196e-05, "loss": 0.0001, "step": 64100 }, { "epoch": 13.83, "learning_rate": 2.9866258330391246e-05, "loss": 0.0002, "step": 64150 }, { "epoch": 13.84, "learning_rate": 2.9865505349364273e-05, "loss": 0.0002, "step": 64200 }, { "epoch": 13.85, "learning_rate": 2.9864750264143856e-05, "loss": 0.0002, "step": 64250 }, { "epoch": 13.86, "learning_rate": 2.9863993074836884e-05, "loss": 0.0002, "step": 64300 }, { "epoch": 13.87, "learning_rate": 2.9863233781550527e-05, "loss": 0.0001, "step": 64350 }, { "epoch": 13.88, "learning_rate": 2.9862472384392267e-05, "loss": 0.0001, "step": 64400 }, { "epoch": 13.89, "learning_rate": 2.986170888346988e-05, "loss": 0.0002, "step": 64450 }, { "epoch": 13.9, "learning_rate": 2.9860943278891434e-05, "loss": 0.0001, "step": 64500 }, { "epoch": 13.91, "learning_rate": 2.9860175570765305e-05, "loss": 0.0001, "step": 64550 }, { "epoch": 13.93, "learning_rate": 2.9859405759200155e-05, "loss": 0.0002, "step": 64600 }, { "epoch": 13.94, "learning_rate": 2.985863384430496e-05, "loss": 0.0002, "step": 64650 }, { "epoch": 13.95, "learning_rate": 2.9857859826188973e-05, "loss": 0.0002, "step": 64700 }, { "epoch": 13.96, "learning_rate": 2.9857083704961754e-05, "loss": 0.0002, "step": 64750 }, { "epoch": 13.97, "learning_rate": 2.9856305480733174e-05, "loss": 0.0002, "step": 64800 }, { "epoch": 13.98, "learning_rate": 2.985552515361338e-05, "loss": 0.0002, "step": 64850 }, { "epoch": 13.99, "learning_rate": 2.9854742723712826e-05, "loss": 0.0001, "step": 64900 }, { "epoch": 14.0, "learning_rate": 2.985395819114227e-05, "loss": 0.0002, "step": 64950 }, { "epoch": 14.01, "learning_rate": 2.9853171556012757e-05, "loss": 0.0001, "step": 65000 }, { "epoch": 14.02, "learning_rate": 2.9852382818435634e-05, "loss": 0.0001, "step": 65050 }, { "epoch": 14.03, "learning_rate": 2.9851591978522553e-05, "loss": 0.0001, "step": 65100 }, { "epoch": 14.04, "learning_rate": 2.9850799036385448e-05, "loss": 0.0001, "step": 65150 }, { "epoch": 14.05, "learning_rate": 2.9850003992136558e-05, "loss": 0.0002, "step": 65200 }, { "epoch": 14.07, "learning_rate": 2.9849206845888428e-05, "loss": 0.0001, "step": 65250 }, { "epoch": 14.08, "learning_rate": 2.984840759775389e-05, "loss": 0.0001, "step": 65300 }, { "epoch": 14.09, "learning_rate": 2.9847606247846074e-05, "loss": 0.0001, "step": 65350 }, { "epoch": 14.1, "learning_rate": 2.9846802796278414e-05, "loss": 0.0001, "step": 65400 }, { "epoch": 14.11, "learning_rate": 2.9845997243164637e-05, "loss": 0.0002, "step": 65450 }, { "epoch": 14.12, "learning_rate": 2.9845189588618764e-05, "loss": 0.0001, "step": 65500 }, { "epoch": 14.13, "learning_rate": 2.984437983275512e-05, "loss": 0.0002, "step": 65550 }, { "epoch": 14.14, "learning_rate": 2.9843567975688327e-05, "loss": 0.0001, "step": 65600 }, { "epoch": 14.15, "learning_rate": 2.9842754017533297e-05, "loss": 0.0002, "step": 65650 }, { "epoch": 14.16, "learning_rate": 2.984193795840525e-05, "loss": 0.0001, "step": 65700 }, { "epoch": 14.17, "learning_rate": 2.9841119798419697e-05, "loss": 0.0001, "step": 65750 }, { "epoch": 14.18, "learning_rate": 2.9840299537692446e-05, "loss": 0.0001, "step": 65800 }, { "epoch": 14.19, "learning_rate": 2.9839477176339603e-05, "loss": 0.0001, "step": 65850 }, { "epoch": 14.21, "learning_rate": 2.9838652714477575e-05, "loss": 0.0001, "step": 65900 }, { "epoch": 14.22, "learning_rate": 2.9837826152223064e-05, "loss": 0.0001, "step": 65950 }, { "epoch": 14.23, "learning_rate": 2.983699748969306e-05, "loss": 0.0001, "step": 66000 }, { "epoch": 14.24, "learning_rate": 2.9836166727004873e-05, "loss": 0.0001, "step": 66050 }, { "epoch": 14.25, "learning_rate": 2.983533386427609e-05, "loss": 0.0001, "step": 66100 }, { "epoch": 14.26, "learning_rate": 2.98344989016246e-05, "loss": 0.0001, "step": 66150 }, { "epoch": 14.27, "learning_rate": 2.983366183916859e-05, "loss": 0.0001, "step": 66200 }, { "epoch": 14.28, "learning_rate": 2.983282267702655e-05, "loss": 0.0001, "step": 66250 }, { "epoch": 14.29, "learning_rate": 2.983198141531726e-05, "loss": 0.0001, "step": 66300 }, { "epoch": 14.3, "learning_rate": 2.9831138054159795e-05, "loss": 0.0001, "step": 66350 }, { "epoch": 14.31, "learning_rate": 2.9830292593673544e-05, "loss": 0.0001, "step": 66400 }, { "epoch": 14.32, "learning_rate": 2.982944503397817e-05, "loss": 0.0001, "step": 66450 }, { "epoch": 14.33, "learning_rate": 2.982859537519364e-05, "loss": 0.0001, "step": 66500 }, { "epoch": 14.35, "learning_rate": 2.982774361744024e-05, "loss": 0.0001, "step": 66550 }, { "epoch": 14.36, "learning_rate": 2.9826889760838513e-05, "loss": 0.0001, "step": 66600 }, { "epoch": 14.37, "learning_rate": 2.9826033805509338e-05, "loss": 0.0001, "step": 66650 }, { "epoch": 14.38, "learning_rate": 2.9825175751573868e-05, "loss": 0.0001, "step": 66700 }, { "epoch": 14.39, "learning_rate": 2.9824315599153562e-05, "loss": 0.0001, "step": 66750 }, { "epoch": 14.4, "learning_rate": 2.9823453348370172e-05, "loss": 0.0001, "step": 66800 }, { "epoch": 14.41, "learning_rate": 2.9822588999345743e-05, "loss": 0.0001, "step": 66850 }, { "epoch": 14.42, "learning_rate": 2.982172255220263e-05, "loss": 0.0001, "step": 66900 }, { "epoch": 14.43, "learning_rate": 2.982085400706348e-05, "loss": 0.0002, "step": 66950 }, { "epoch": 14.44, "learning_rate": 2.9819983364051226e-05, "loss": 0.0001, "step": 67000 }, { "epoch": 14.45, "learning_rate": 2.981911062328911e-05, "loss": 0.0001, "step": 67050 }, { "epoch": 14.46, "learning_rate": 2.981823578490067e-05, "loss": 0.0002, "step": 67100 }, { "epoch": 14.48, "learning_rate": 2.981735884900973e-05, "loss": 0.0001, "step": 67150 }, { "epoch": 14.49, "learning_rate": 2.981647981574043e-05, "loss": 0.0002, "step": 67200 }, { "epoch": 14.5, "learning_rate": 2.981559868521719e-05, "loss": 0.0001, "step": 67250 }, { "epoch": 14.51, "learning_rate": 2.9814715457564734e-05, "loss": 0.0001, "step": 67300 }, { "epoch": 14.52, "learning_rate": 2.9813830132908083e-05, "loss": 0.0001, "step": 67350 }, { "epoch": 14.53, "learning_rate": 2.9812942711372552e-05, "loss": 0.0001, "step": 67400 }, { "epoch": 14.54, "learning_rate": 2.9812053193083752e-05, "loss": 0.0002, "step": 67450 }, { "epoch": 14.55, "learning_rate": 2.9811161578167597e-05, "loss": 0.0002, "step": 67500 }, { "epoch": 14.56, "learning_rate": 2.9810267866750297e-05, "loss": 0.0001, "step": 67550 }, { "epoch": 14.57, "learning_rate": 2.9809372058958345e-05, "loss": 0.0001, "step": 67600 }, { "epoch": 14.58, "learning_rate": 2.980847415491855e-05, "loss": 0.0001, "step": 67650 }, { "epoch": 14.59, "learning_rate": 2.9807574154758006e-05, "loss": 0.0001, "step": 67700 }, { "epoch": 14.6, "learning_rate": 2.980667205860411e-05, "loss": 0.0001, "step": 67750 }, { "epoch": 14.62, "learning_rate": 2.9805767866584553e-05, "loss": 0.0001, "step": 67800 }, { "epoch": 14.63, "learning_rate": 2.9804861578827317e-05, "loss": 0.0002, "step": 67850 }, { "epoch": 14.64, "learning_rate": 2.980395319546069e-05, "loss": 0.0001, "step": 67900 }, { "epoch": 14.65, "learning_rate": 2.980304271661325e-05, "loss": 0.0001, "step": 67950 }, { "epoch": 14.66, "learning_rate": 2.9802130142413875e-05, "loss": 0.0001, "step": 68000 }, { "epoch": 14.67, "learning_rate": 2.980121547299174e-05, "loss": 0.0001, "step": 68050 }, { "epoch": 14.68, "learning_rate": 2.9800298708476312e-05, "loss": 0.0001, "step": 68100 }, { "epoch": 14.69, "learning_rate": 2.9799379848997366e-05, "loss": 0.0001, "step": 68150 }, { "epoch": 14.7, "learning_rate": 2.9798458894684955e-05, "loss": 0.0001, "step": 68200 }, { "epoch": 14.71, "learning_rate": 2.9797535845669444e-05, "loss": 0.0002, "step": 68250 }, { "epoch": 14.72, "learning_rate": 2.979661070208149e-05, "loss": 0.0001, "step": 68300 }, { "epoch": 14.73, "learning_rate": 2.9795683464052045e-05, "loss": 0.0001, "step": 68350 }, { "epoch": 14.74, "learning_rate": 2.9794754131712355e-05, "loss": 0.0001, "step": 68400 }, { "epoch": 14.76, "learning_rate": 2.9793822705193973e-05, "loss": 0.0001, "step": 68450 }, { "epoch": 14.77, "learning_rate": 2.9792889184628734e-05, "loss": 0.0001, "step": 68500 }, { "epoch": 14.78, "learning_rate": 2.9791953570148783e-05, "loss": 0.0001, "step": 68550 }, { "epoch": 14.79, "learning_rate": 2.9791015861886552e-05, "loss": 0.0001, "step": 68600 }, { "epoch": 14.8, "learning_rate": 2.979007605997477e-05, "loss": 0.0002, "step": 68650 }, { "epoch": 14.81, "learning_rate": 2.978913416454647e-05, "loss": 0.0002, "step": 68700 }, { "epoch": 14.82, "learning_rate": 2.978819017573497e-05, "loss": 0.0001, "step": 68750 }, { "epoch": 14.83, "learning_rate": 2.9787244093673897e-05, "loss": 0.0001, "step": 68800 }, { "epoch": 14.84, "learning_rate": 2.9786295918497164e-05, "loss": 0.0002, "step": 68850 }, { "epoch": 14.85, "learning_rate": 2.9785345650338982e-05, "loss": 0.0001, "step": 68900 }, { "epoch": 14.86, "learning_rate": 2.9784393289333864e-05, "loss": 0.0001, "step": 68950 }, { "epoch": 14.87, "learning_rate": 2.9783438835616614e-05, "loss": 0.0001, "step": 69000 }, { "epoch": 14.88, "learning_rate": 2.978248228932233e-05, "loss": 0.0001, "step": 69050 }, { "epoch": 14.9, "learning_rate": 2.9781523650586422e-05, "loss": 0.0001, "step": 69100 }, { "epoch": 14.91, "learning_rate": 2.9780562919544568e-05, "loss": 0.0001, "step": 69150 }, { "epoch": 14.92, "learning_rate": 2.977960009633277e-05, "loss": 0.0002, "step": 69200 }, { "epoch": 14.93, "learning_rate": 2.977863518108731e-05, "loss": 0.0001, "step": 69250 }, { "epoch": 14.94, "learning_rate": 2.9777668173944772e-05, "loss": 0.0001, "step": 69300 }, { "epoch": 14.95, "learning_rate": 2.9776699075042033e-05, "loss": 0.0001, "step": 69350 }, { "epoch": 14.96, "learning_rate": 2.977572788451627e-05, "loss": 0.0001, "step": 69400 }, { "epoch": 14.97, "learning_rate": 2.977475460250495e-05, "loss": 0.0001, "step": 69450 }, { "epoch": 14.98, "learning_rate": 2.9773779229145843e-05, "loss": 0.0001, "step": 69500 }, { "epoch": 14.99, "learning_rate": 2.977280176457701e-05, "loss": 0.0001, "step": 69550 }, { "epoch": 15.0, "learning_rate": 2.9771822208936813e-05, "loss": 0.0001, "step": 69600 }, { "epoch": 15.01, "learning_rate": 2.9770840562363902e-05, "loss": 0.0001, "step": 69650 }, { "epoch": 15.02, "learning_rate": 2.9769856824997233e-05, "loss": 0.0001, "step": 69700 }, { "epoch": 15.04, "learning_rate": 2.976887099697605e-05, "loss": 0.0001, "step": 69750 }, { "epoch": 15.05, "learning_rate": 2.9767883078439893e-05, "loss": 0.0001, "step": 69800 }, { "epoch": 15.06, "learning_rate": 2.9766893069528607e-05, "loss": 0.0001, "step": 69850 }, { "epoch": 15.07, "learning_rate": 2.976590097038232e-05, "loss": 0.0001, "step": 69900 }, { "epoch": 15.08, "learning_rate": 2.9764906781141468e-05, "loss": 0.0001, "step": 69950 }, { "epoch": 15.09, "learning_rate": 2.9763910501946773e-05, "loss": 0.0001, "step": 70000 }, { "epoch": 15.1, "learning_rate": 2.976291213293926e-05, "loss": 0.0001, "step": 70050 }, { "epoch": 15.11, "learning_rate": 2.9761911674260246e-05, "loss": 0.0001, "step": 70100 }, { "epoch": 15.12, "learning_rate": 2.976090912605134e-05, "loss": 0.0001, "step": 70150 }, { "epoch": 15.13, "learning_rate": 2.975990448845446e-05, "loss": 0.0001, "step": 70200 }, { "epoch": 15.14, "learning_rate": 2.9758897761611805e-05, "loss": 0.0001, "step": 70250 }, { "epoch": 15.15, "learning_rate": 2.975788894566588e-05, "loss": 0.0001, "step": 70300 }, { "epoch": 15.16, "learning_rate": 2.9756878040759482e-05, "loss": 0.0001, "step": 70350 }, { "epoch": 15.18, "learning_rate": 2.9755865047035694e-05, "loss": 0.0001, "step": 70400 }, { "epoch": 15.19, "learning_rate": 2.9754849964637915e-05, "loss": 0.0001, "step": 70450 }, { "epoch": 15.2, "learning_rate": 2.975383279370982e-05, "loss": 0.0001, "step": 70500 }, { "epoch": 15.21, "learning_rate": 2.9752813534395397e-05, "loss": 0.0001, "step": 70550 }, { "epoch": 15.22, "learning_rate": 2.9751792186838914e-05, "loss": 0.0001, "step": 70600 }, { "epoch": 15.23, "learning_rate": 2.9750768751184944e-05, "loss": 0.0001, "step": 70650 }, { "epoch": 15.24, "learning_rate": 2.9749743227578358e-05, "loss": 0.0001, "step": 70700 }, { "epoch": 15.25, "learning_rate": 2.974871561616431e-05, "loss": 0.0001, "step": 70750 }, { "epoch": 15.26, "learning_rate": 2.9747685917088253e-05, "loss": 0.0001, "step": 70800 }, { "epoch": 15.27, "learning_rate": 2.974665413049595e-05, "loss": 0.0001, "step": 70850 }, { "epoch": 15.28, "learning_rate": 2.974562025653345e-05, "loss": 0.0001, "step": 70900 }, { "epoch": 15.29, "learning_rate": 2.974458429534709e-05, "loss": 0.0001, "step": 70950 }, { "epoch": 15.31, "learning_rate": 2.9743546247083505e-05, "loss": 0.0001, "step": 71000 }, { "epoch": 15.32, "learning_rate": 2.974250611188964e-05, "loss": 0.0001, "step": 71050 }, { "epoch": 15.33, "learning_rate": 2.9741463889912718e-05, "loss": 0.0001, "step": 71100 }, { "epoch": 15.34, "learning_rate": 2.9740419581300266e-05, "loss": 0.0001, "step": 71150 }, { "epoch": 15.35, "learning_rate": 2.973937318620011e-05, "loss": 0.0001, "step": 71200 }, { "epoch": 15.36, "learning_rate": 2.9738324704760355e-05, "loss": 0.0001, "step": 71250 }, { "epoch": 15.37, "learning_rate": 2.973727413712942e-05, "loss": 0.0001, "step": 71300 }, { "epoch": 15.38, "learning_rate": 2.9736221483456007e-05, "loss": 0.0001, "step": 71350 }, { "epoch": 15.39, "learning_rate": 2.9735166743889125e-05, "loss": 0.0001, "step": 71400 }, { "epoch": 15.4, "learning_rate": 2.973410991857806e-05, "loss": 0.0001, "step": 71450 }, { "epoch": 15.41, "learning_rate": 2.9733051007672415e-05, "loss": 0.0001, "step": 71500 }, { "epoch": 15.42, "learning_rate": 2.973199001132207e-05, "loss": 0.0001, "step": 71550 }, { "epoch": 15.43, "learning_rate": 2.9730926929677218e-05, "loss": 0.0001, "step": 71600 }, { "epoch": 15.45, "learning_rate": 2.9729861762888323e-05, "loss": 0.0001, "step": 71650 }, { "epoch": 15.46, "learning_rate": 2.9728794511106168e-05, "loss": 0.0001, "step": 71700 }, { "epoch": 15.47, "learning_rate": 2.9727725174481816e-05, "loss": 0.0001, "step": 71750 }, { "epoch": 15.48, "learning_rate": 2.9726653753166632e-05, "loss": 0.0001, "step": 71800 }, { "epoch": 15.49, "learning_rate": 2.9725580247312273e-05, "loss": 0.0001, "step": 71850 }, { "epoch": 15.5, "learning_rate": 2.97245046570707e-05, "loss": 0.0001, "step": 71900 }, { "epoch": 15.51, "learning_rate": 2.9723426982594145e-05, "loss": 0.0001, "step": 71950 }, { "epoch": 15.52, "learning_rate": 2.972234722403517e-05, "loss": 0.0001, "step": 72000 }, { "epoch": 15.53, "learning_rate": 2.9721265381546595e-05, "loss": 0.0001, "step": 72050 }, { "epoch": 15.54, "learning_rate": 2.9720181455281573e-05, "loss": 0.0001, "step": 72100 }, { "epoch": 15.55, "learning_rate": 2.971909544539352e-05, "loss": 0.0001, "step": 72150 }, { "epoch": 15.56, "learning_rate": 2.9718007352036164e-05, "loss": 0.0002, "step": 72200 }, { "epoch": 15.57, "learning_rate": 2.9716917175363518e-05, "loss": 0.0001, "step": 72250 }, { "epoch": 15.59, "learning_rate": 2.9715824915529903e-05, "loss": 0.0002, "step": 72300 }, { "epoch": 15.6, "learning_rate": 2.9714730572689917e-05, "loss": 0.0001, "step": 72350 }, { "epoch": 15.61, "learning_rate": 2.971363414699847e-05, "loss": 0.0001, "step": 72400 }, { "epoch": 15.62, "learning_rate": 2.9712535638610766e-05, "loss": 0.0001, "step": 72450 }, { "epoch": 15.63, "learning_rate": 2.9711435047682277e-05, "loss": 0.0001, "step": 72500 }, { "epoch": 15.64, "learning_rate": 2.9710332374368813e-05, "loss": 0.0001, "step": 72550 }, { "epoch": 15.65, "learning_rate": 2.970922761882644e-05, "loss": 0.0001, "step": 72600 }, { "epoch": 15.66, "learning_rate": 2.970812078121154e-05, "loss": 0.0001, "step": 72650 }, { "epoch": 15.67, "learning_rate": 2.9707011861680792e-05, "loss": 0.0002, "step": 72700 }, { "epoch": 15.68, "learning_rate": 2.9705900860391148e-05, "loss": 0.0001, "step": 72750 }, { "epoch": 15.69, "learning_rate": 2.970478777749988e-05, "loss": 0.0001, "step": 72800 }, { "epoch": 15.7, "learning_rate": 2.9703672613164537e-05, "loss": 0.0001, "step": 72850 }, { "epoch": 15.71, "learning_rate": 2.970255536754297e-05, "loss": 0.0001, "step": 72900 }, { "epoch": 15.73, "learning_rate": 2.9701436040793325e-05, "loss": 0.0001, "step": 72950 }, { "epoch": 15.74, "learning_rate": 2.9700314633074047e-05, "loss": 0.0001, "step": 73000 }, { "epoch": 15.75, "learning_rate": 2.969919114454386e-05, "loss": 0.0001, "step": 73050 }, { "epoch": 15.76, "learning_rate": 2.96980655753618e-05, "loss": 0.0001, "step": 73100 }, { "epoch": 15.77, "learning_rate": 2.9696937925687185e-05, "loss": 0.0001, "step": 73150 }, { "epoch": 15.78, "learning_rate": 2.9695808195679634e-05, "loss": 0.0001, "step": 73200 }, { "epoch": 15.79, "learning_rate": 2.9694676385499056e-05, "loss": 0.0001, "step": 73250 }, { "epoch": 15.8, "learning_rate": 2.9693542495305665e-05, "loss": 0.0002, "step": 73300 }, { "epoch": 15.81, "learning_rate": 2.9692406525259957e-05, "loss": 0.0001, "step": 73350 }, { "epoch": 15.82, "learning_rate": 2.9691268475522725e-05, "loss": 0.0001, "step": 73400 }, { "epoch": 15.83, "learning_rate": 2.969012834625506e-05, "loss": 0.0001, "step": 73450 }, { "epoch": 15.84, "learning_rate": 2.9688986137618346e-05, "loss": 0.0001, "step": 73500 }, { "epoch": 15.85, "learning_rate": 2.9687841849774264e-05, "loss": 0.0001, "step": 73550 }, { "epoch": 15.87, "learning_rate": 2.968669548288478e-05, "loss": 0.0001, "step": 73600 }, { "epoch": 15.88, "learning_rate": 2.968554703711217e-05, "loss": 0.0001, "step": 73650 }, { "epoch": 15.89, "learning_rate": 2.968439651261899e-05, "loss": 0.0001, "step": 73700 }, { "epoch": 15.9, "learning_rate": 2.9683243909568087e-05, "loss": 0.0001, "step": 73750 }, { "epoch": 15.91, "learning_rate": 2.9682089228122627e-05, "loss": 0.0001, "step": 73800 }, { "epoch": 15.92, "learning_rate": 2.9680932468446042e-05, "loss": 0.0001, "step": 73850 }, { "epoch": 15.93, "learning_rate": 2.9679773630702074e-05, "loss": 0.0001, "step": 73900 }, { "epoch": 15.94, "learning_rate": 2.9678612715054752e-05, "loss": 0.0001, "step": 73950 }, { "epoch": 15.95, "learning_rate": 2.9677449721668407e-05, "loss": 0.0001, "step": 74000 }, { "epoch": 15.96, "learning_rate": 2.9676284650707656e-05, "loss": 0.0001, "step": 74050 }, { "epoch": 15.97, "learning_rate": 2.9675117502337412e-05, "loss": 0.0001, "step": 74100 }, { "epoch": 15.98, "learning_rate": 2.9673948276722888e-05, "loss": 0.0001, "step": 74150 }, { "epoch": 15.99, "learning_rate": 2.9672776974029586e-05, "loss": 0.0001, "step": 74200 }, { "epoch": 16.01, "learning_rate": 2.9671603594423294e-05, "loss": 0.0001, "step": 74250 }, { "epoch": 16.02, "learning_rate": 2.967042813807011e-05, "loss": 0.0001, "step": 74300 }, { "epoch": 16.03, "learning_rate": 2.9669250605136418e-05, "loss": 0.0001, "step": 74350 }, { "epoch": 16.04, "learning_rate": 2.96680709957889e-05, "loss": 0.0001, "step": 74400 }, { "epoch": 16.05, "learning_rate": 2.9666889310194516e-05, "loss": 0.0001, "step": 74450 }, { "epoch": 16.06, "learning_rate": 2.9665705548520542e-05, "loss": 0.0001, "step": 74500 }, { "epoch": 16.07, "learning_rate": 2.9664519710934535e-05, "loss": 0.0001, "step": 74550 }, { "epoch": 16.08, "learning_rate": 2.966333179760435e-05, "loss": 0.0001, "step": 74600 }, { "epoch": 16.09, "learning_rate": 2.9662141808698134e-05, "loss": 0.0001, "step": 74650 }, { "epoch": 16.1, "learning_rate": 2.9660949744384328e-05, "loss": 0.0001, "step": 74700 }, { "epoch": 16.11, "learning_rate": 2.965975560483167e-05, "loss": 0.0001, "step": 74750 }, { "epoch": 16.12, "learning_rate": 2.9658559390209182e-05, "loss": 0.0001, "step": 74800 }, { "epoch": 16.13, "learning_rate": 2.9657361100686195e-05, "loss": 0.0001, "step": 74850 }, { "epoch": 16.15, "learning_rate": 2.9656160736432318e-05, "loss": 0.0001, "step": 74900 }, { "epoch": 16.16, "learning_rate": 2.9654958297617465e-05, "loss": 0.0001, "step": 74950 }, { "epoch": 16.17, "learning_rate": 2.965375378441184e-05, "loss": 0.0001, "step": 75000 }, { "epoch": 16.18, "learning_rate": 2.965254719698594e-05, "loss": 0.0001, "step": 75050 }, { "epoch": 16.19, "learning_rate": 2.9651338535510556e-05, "loss": 0.0001, "step": 75100 }, { "epoch": 16.2, "learning_rate": 2.965012780015677e-05, "loss": 0.0001, "step": 75150 }, { "epoch": 16.21, "learning_rate": 2.9648914991095964e-05, "loss": 0.0001, "step": 75200 }, { "epoch": 16.22, "learning_rate": 2.9647700108499805e-05, "loss": 0.0001, "step": 75250 }, { "epoch": 16.23, "learning_rate": 2.9646483152540264e-05, "loss": 0.0001, "step": 75300 }, { "epoch": 16.24, "learning_rate": 2.9645264123389597e-05, "loss": 0.0001, "step": 75350 }, { "epoch": 16.25, "learning_rate": 2.9644043021220354e-05, "loss": 0.0001, "step": 75400 }, { "epoch": 16.26, "learning_rate": 2.9642819846205378e-05, "loss": 0.0001, "step": 75450 }, { "epoch": 16.28, "learning_rate": 2.9641594598517815e-05, "loss": 0.0001, "step": 75500 }, { "epoch": 16.29, "learning_rate": 2.964036727833109e-05, "loss": 0.0001, "step": 75550 }, { "epoch": 16.3, "learning_rate": 2.9639137885818935e-05, "loss": 0.0001, "step": 75600 }, { "epoch": 16.31, "learning_rate": 2.963790642115537e-05, "loss": 0.0001, "step": 75650 }, { "epoch": 16.32, "learning_rate": 2.96366728845147e-05, "loss": 0.0001, "step": 75700 }, { "epoch": 16.33, "learning_rate": 2.9635437276071533e-05, "loss": 0.0001, "step": 75750 }, { "epoch": 16.34, "learning_rate": 2.9634199596000776e-05, "loss": 0.0001, "step": 75800 }, { "epoch": 16.35, "learning_rate": 2.963295984447761e-05, "loss": 0.0001, "step": 75850 }, { "epoch": 16.36, "learning_rate": 2.963171802167752e-05, "loss": 0.0001, "step": 75900 }, { "epoch": 16.37, "learning_rate": 2.963047412777629e-05, "loss": 0.0001, "step": 75950 }, { "epoch": 16.38, "learning_rate": 2.9629228162949993e-05, "loss": 0.0001, "step": 76000 }, { "epoch": 16.39, "learning_rate": 2.962798012737499e-05, "loss": 0.0001, "step": 76050 }, { "epoch": 16.4, "learning_rate": 2.9626730021227944e-05, "loss": 0.0001, "step": 76100 }, { "epoch": 16.42, "learning_rate": 2.9625477844685795e-05, "loss": 0.0001, "step": 76150 }, { "epoch": 16.43, "learning_rate": 2.96242235979258e-05, "loss": 0.0001, "step": 76200 }, { "epoch": 16.44, "learning_rate": 2.9622967281125486e-05, "loss": 0.0001, "step": 76250 }, { "epoch": 16.45, "learning_rate": 2.9621708894462686e-05, "loss": 0.0001, "step": 76300 }, { "epoch": 16.46, "learning_rate": 2.962044843811553e-05, "loss": 0.0001, "step": 76350 }, { "epoch": 16.47, "learning_rate": 2.9619185912262424e-05, "loss": 0.0001, "step": 76400 }, { "epoch": 16.48, "learning_rate": 2.961792131708208e-05, "loss": 0.0001, "step": 76450 }, { "epoch": 16.49, "learning_rate": 2.9616654652753504e-05, "loss": 0.0001, "step": 76500 }, { "epoch": 16.5, "learning_rate": 2.9615385919455985e-05, "loss": 0.0002, "step": 76550 }, { "epoch": 16.51, "learning_rate": 2.9614115117369113e-05, "loss": 0.0001, "step": 76600 }, { "epoch": 16.52, "learning_rate": 2.9612842246672773e-05, "loss": 0.0001, "step": 76650 }, { "epoch": 16.53, "learning_rate": 2.961156730754713e-05, "loss": 0.0001, "step": 76700 }, { "epoch": 16.54, "learning_rate": 2.961029030017265e-05, "loss": 0.0001, "step": 76750 }, { "epoch": 16.56, "learning_rate": 2.9609011224730103e-05, "loss": 0.0001, "step": 76800 }, { "epoch": 16.57, "learning_rate": 2.9607730081400534e-05, "loss": 0.0001, "step": 76850 }, { "epoch": 16.58, "learning_rate": 2.960644687036528e-05, "loss": 0.0001, "step": 76900 }, { "epoch": 16.59, "learning_rate": 2.9605161591805988e-05, "loss": 0.0001, "step": 76950 }, { "epoch": 16.6, "learning_rate": 2.960387424590458e-05, "loss": 0.0001, "step": 77000 }, { "epoch": 16.61, "learning_rate": 2.9602584832843285e-05, "loss": 0.0001, "step": 77050 }, { "epoch": 16.62, "learning_rate": 2.9601293352804613e-05, "loss": 0.0001, "step": 77100 }, { "epoch": 16.63, "learning_rate": 2.9599999805971374e-05, "loss": 0.0001, "step": 77150 }, { "epoch": 16.64, "learning_rate": 2.9598704192526668e-05, "loss": 0.0001, "step": 77200 }, { "epoch": 16.65, "learning_rate": 2.9597406512653883e-05, "loss": 0.0001, "step": 77250 }, { "epoch": 16.66, "learning_rate": 2.9596106766536712e-05, "loss": 0.0001, "step": 77300 }, { "epoch": 16.67, "learning_rate": 2.9594804954359128e-05, "loss": 0.0001, "step": 77350 }, { "epoch": 16.68, "learning_rate": 2.9593501076305397e-05, "loss": 0.0001, "step": 77400 }, { "epoch": 16.7, "learning_rate": 2.9592195132560084e-05, "loss": 0.0001, "step": 77450 }, { "epoch": 16.71, "learning_rate": 2.9590887123308054e-05, "loss": 0.0001, "step": 77500 }, { "epoch": 16.72, "learning_rate": 2.9589577048734436e-05, "loss": 0.0001, "step": 77550 }, { "epoch": 16.73, "learning_rate": 2.9588264909024677e-05, "loss": 0.0001, "step": 77600 }, { "epoch": 16.74, "learning_rate": 2.958695070436451e-05, "loss": 0.0001, "step": 77650 }, { "epoch": 16.75, "learning_rate": 2.9585634434939963e-05, "loss": 0.0001, "step": 77700 }, { "epoch": 16.76, "learning_rate": 2.958431610093735e-05, "loss": 0.0001, "step": 77750 }, { "epoch": 16.77, "learning_rate": 2.9582995702543273e-05, "loss": 0.0001, "step": 77800 }, { "epoch": 16.78, "learning_rate": 2.9581673239944638e-05, "loss": 0.0001, "step": 77850 }, { "epoch": 16.79, "learning_rate": 2.958034871332864e-05, "loss": 0.0001, "step": 77900 }, { "epoch": 16.8, "learning_rate": 2.957902212288276e-05, "loss": 0.0001, "step": 77950 }, { "epoch": 16.81, "learning_rate": 2.9577693468794776e-05, "loss": 0.0002, "step": 78000 }, { "epoch": 16.82, "learning_rate": 2.957636275125276e-05, "loss": 0.0001, "step": 78050 }, { "epoch": 16.84, "learning_rate": 2.9575029970445065e-05, "loss": 0.0001, "step": 78100 }, { "epoch": 16.85, "learning_rate": 2.9573695126560354e-05, "loss": 0.0001, "step": 78150 }, { "epoch": 16.86, "learning_rate": 2.9572358219787567e-05, "loss": 0.0001, "step": 78200 }, { "epoch": 16.87, "learning_rate": 2.9571019250315946e-05, "loss": 0.0001, "step": 78250 }, { "epoch": 16.88, "learning_rate": 2.9569678218335016e-05, "loss": 0.0001, "step": 78300 }, { "epoch": 16.89, "learning_rate": 2.9568335124034604e-05, "loss": 0.0001, "step": 78350 }, { "epoch": 16.9, "learning_rate": 2.9566989967604818e-05, "loss": 0.0001, "step": 78400 }, { "epoch": 16.91, "learning_rate": 2.9565642749236063e-05, "loss": 0.0001, "step": 78450 }, { "epoch": 16.92, "learning_rate": 2.9564293469119042e-05, "loss": 0.0001, "step": 78500 }, { "epoch": 16.93, "learning_rate": 2.9562942127444736e-05, "loss": 0.0001, "step": 78550 }, { "epoch": 16.94, "learning_rate": 2.956158872440443e-05, "loss": 0.0001, "step": 78600 }, { "epoch": 16.95, "learning_rate": 2.9560233260189696e-05, "loss": 0.0001, "step": 78650 }, { "epoch": 16.96, "learning_rate": 2.95588757349924e-05, "loss": 0.0001, "step": 78700 }, { "epoch": 16.98, "learning_rate": 2.9557516149004693e-05, "loss": 0.0001, "step": 78750 }, { "epoch": 16.99, "learning_rate": 2.9556154502419027e-05, "loss": 0.0002, "step": 78800 }, { "epoch": 17.0, "learning_rate": 2.955479079542814e-05, "loss": 0.0001, "step": 78850 }, { "epoch": 17.01, "learning_rate": 2.9553425028225065e-05, "loss": 0.0001, "step": 78900 }, { "epoch": 17.02, "learning_rate": 2.955205720100312e-05, "loss": 0.0001, "step": 78950 }, { "epoch": 17.03, "learning_rate": 2.9550687313955926e-05, "loss": 0.0001, "step": 79000 }, { "epoch": 17.04, "learning_rate": 2.9549315367277382e-05, "loss": 0.0001, "step": 79050 }, { "epoch": 17.05, "learning_rate": 2.9547941361161687e-05, "loss": 0.0001, "step": 79100 }, { "epoch": 17.06, "learning_rate": 2.9546565295803333e-05, "loss": 0.0001, "step": 79150 }, { "epoch": 17.07, "learning_rate": 2.9545187171397097e-05, "loss": 0.0001, "step": 79200 }, { "epoch": 17.08, "learning_rate": 2.954380698813805e-05, "loss": 0.0001, "step": 79250 }, { "epoch": 17.09, "learning_rate": 2.9542424746221562e-05, "loss": 0.0001, "step": 79300 }, { "epoch": 17.1, "learning_rate": 2.954104044584328e-05, "loss": 0.0001, "step": 79350 }, { "epoch": 17.12, "learning_rate": 2.9539654087199154e-05, "loss": 0.0001, "step": 79400 }, { "epoch": 17.13, "learning_rate": 2.953826567048542e-05, "loss": 0.0001, "step": 79450 }, { "epoch": 17.14, "learning_rate": 2.9536875195898605e-05, "loss": 0.0001, "step": 79500 }, { "epoch": 17.15, "learning_rate": 2.953548266363553e-05, "loss": 0.0001, "step": 79550 }, { "epoch": 17.16, "learning_rate": 2.953408807389331e-05, "loss": 0.0001, "step": 79600 }, { "epoch": 17.17, "learning_rate": 2.9532691426869343e-05, "loss": 0.0001, "step": 79650 }, { "epoch": 17.18, "learning_rate": 2.9531292722761323e-05, "loss": 0.0001, "step": 79700 }, { "epoch": 17.19, "learning_rate": 2.952989196176723e-05, "loss": 0.0001, "step": 79750 }, { "epoch": 17.2, "learning_rate": 2.9528489144085356e-05, "loss": 0.0001, "step": 79800 }, { "epoch": 17.21, "learning_rate": 2.9527084269914252e-05, "loss": 0.0001, "step": 79850 }, { "epoch": 17.22, "learning_rate": 2.952567733945278e-05, "loss": 0.0001, "step": 79900 }, { "epoch": 17.23, "learning_rate": 2.9524268352900096e-05, "loss": 0.0, "step": 79950 }, { "epoch": 17.25, "learning_rate": 2.9522857310455634e-05, "loss": 0.0001, "step": 80000 }, { "epoch": 17.26, "learning_rate": 2.952144421231912e-05, "loss": 0.0001, "step": 80050 }, { "epoch": 17.27, "learning_rate": 2.9520029058690596e-05, "loss": 0.0001, "step": 80100 }, { "epoch": 17.28, "learning_rate": 2.951861184977035e-05, "loss": 0.0001, "step": 80150 }, { "epoch": 17.29, "learning_rate": 2.9517192585759e-05, "loss": 0.0001, "step": 80200 }, { "epoch": 17.3, "learning_rate": 2.9515771266857442e-05, "loss": 0.0001, "step": 80250 }, { "epoch": 17.31, "learning_rate": 2.951434789326686e-05, "loss": 0.0001, "step": 80300 }, { "epoch": 17.32, "learning_rate": 2.9512922465188726e-05, "loss": 0.0001, "step": 80350 }, { "epoch": 17.33, "learning_rate": 2.951149498282481e-05, "loss": 0.0002, "step": 80400 }, { "epoch": 17.34, "learning_rate": 2.951006544637718e-05, "loss": 0.0001, "step": 80450 }, { "epoch": 17.35, "learning_rate": 2.950863385604817e-05, "loss": 0.0001, "step": 80500 }, { "epoch": 17.36, "learning_rate": 2.9507200212040425e-05, "loss": 0.0001, "step": 80550 }, { "epoch": 17.37, "learning_rate": 2.950576451455688e-05, "loss": 0.0001, "step": 80600 }, { "epoch": 17.39, "learning_rate": 2.950432676380075e-05, "loss": 0.0001, "step": 80650 }, { "epoch": 17.4, "learning_rate": 2.9502886959975555e-05, "loss": 0.0001, "step": 80700 }, { "epoch": 17.41, "learning_rate": 2.9501445103285092e-05, "loss": 0.0001, "step": 80750 }, { "epoch": 17.42, "learning_rate": 2.9500001193933446e-05, "loss": 0.0001, "step": 80800 }, { "epoch": 17.43, "learning_rate": 2.9498555232125018e-05, "loss": 0.0001, "step": 80850 }, { "epoch": 17.44, "learning_rate": 2.949710721806447e-05, "loss": 0.0, "step": 80900 }, { "epoch": 17.45, "learning_rate": 2.9495657151956767e-05, "loss": 0.0001, "step": 80950 }, { "epoch": 17.46, "learning_rate": 2.949420503400717e-05, "loss": 0.0001, "step": 81000 }, { "epoch": 17.47, "learning_rate": 2.949275086442122e-05, "loss": 0.0001, "step": 81050 }, { "epoch": 17.48, "learning_rate": 2.949129464340475e-05, "loss": 0.0001, "step": 81100 }, { "epoch": 17.49, "learning_rate": 2.9489836371163897e-05, "loss": 0.0001, "step": 81150 }, { "epoch": 17.5, "learning_rate": 2.9488376047905064e-05, "loss": 0.0001, "step": 81200 }, { "epoch": 17.51, "learning_rate": 2.9486913673834967e-05, "loss": 0.0001, "step": 81250 }, { "epoch": 17.53, "learning_rate": 2.9485449249160604e-05, "loss": 0.0001, "step": 81300 }, { "epoch": 17.54, "learning_rate": 2.9483982774089253e-05, "loss": 0.0001, "step": 81350 }, { "epoch": 17.55, "learning_rate": 2.9482514248828504e-05, "loss": 0.0001, "step": 81400 }, { "epoch": 17.56, "learning_rate": 2.9481043673586216e-05, "loss": 0.0001, "step": 81450 }, { "epoch": 17.57, "learning_rate": 2.947957104857055e-05, "loss": 0.0001, "step": 81500 }, { "epoch": 17.58, "learning_rate": 2.9478096373989953e-05, "loss": 0.0001, "step": 81550 }, { "epoch": 17.59, "learning_rate": 2.9476619650053162e-05, "loss": 0.0001, "step": 81600 }, { "epoch": 17.6, "learning_rate": 2.9475140876969214e-05, "loss": 0.0001, "step": 81650 }, { "epoch": 17.61, "learning_rate": 2.9473660054947416e-05, "loss": 0.0001, "step": 81700 }, { "epoch": 17.62, "learning_rate": 2.9472177184197383e-05, "loss": 0.0001, "step": 81750 }, { "epoch": 17.63, "learning_rate": 2.9470692264929013e-05, "loss": 0.0001, "step": 81800 }, { "epoch": 17.64, "learning_rate": 2.946920529735249e-05, "loss": 0.0001, "step": 81850 }, { "epoch": 17.65, "learning_rate": 2.9467716281678302e-05, "loss": 0.0001, "step": 81900 }, { "epoch": 17.67, "learning_rate": 2.946622521811721e-05, "loss": 0.0001, "step": 81950 }, { "epoch": 17.68, "learning_rate": 2.9464732106880267e-05, "loss": 0.0001, "step": 82000 }, { "epoch": 17.69, "learning_rate": 2.9463236948178833e-05, "loss": 0.0001, "step": 82050 }, { "epoch": 17.7, "learning_rate": 2.946173974222454e-05, "loss": 0.0001, "step": 82100 }, { "epoch": 17.71, "learning_rate": 2.946024048922932e-05, "loss": 0.0001, "step": 82150 }, { "epoch": 17.72, "learning_rate": 2.9458739189405376e-05, "loss": 0.0001, "step": 82200 }, { "epoch": 17.73, "learning_rate": 2.945723584296523e-05, "loss": 0.0001, "step": 82250 }, { "epoch": 17.74, "learning_rate": 2.9455730450121675e-05, "loss": 0.0001, "step": 82300 }, { "epoch": 17.75, "learning_rate": 2.9454223011087796e-05, "loss": 0.0001, "step": 82350 }, { "epoch": 17.76, "learning_rate": 2.945271352607697e-05, "loss": 0.0001, "step": 82400 }, { "epoch": 17.77, "learning_rate": 2.945120199530286e-05, "loss": 0.0002, "step": 82450 }, { "epoch": 17.78, "learning_rate": 2.9449688418979427e-05, "loss": 0.0001, "step": 82500 }, { "epoch": 17.79, "learning_rate": 2.944817279732091e-05, "loss": 0.0001, "step": 82550 }, { "epoch": 17.81, "learning_rate": 2.9446655130541845e-05, "loss": 0.0001, "step": 82600 }, { "epoch": 17.82, "learning_rate": 2.9445135418857057e-05, "loss": 0.0001, "step": 82650 }, { "epoch": 17.83, "learning_rate": 2.9443613662481662e-05, "loss": 0.0001, "step": 82700 }, { "epoch": 17.84, "learning_rate": 2.944208986163106e-05, "loss": 0.0001, "step": 82750 }, { "epoch": 17.85, "learning_rate": 2.9440564016520936e-05, "loss": 0.0001, "step": 82800 }, { "epoch": 17.86, "learning_rate": 2.943903612736728e-05, "loss": 0.0001, "step": 82850 }, { "epoch": 17.87, "learning_rate": 2.9437506194386364e-05, "loss": 0.0001, "step": 82900 }, { "epoch": 17.88, "learning_rate": 2.9435974217794742e-05, "loss": 0.0001, "step": 82950 }, { "epoch": 17.89, "learning_rate": 2.9434440197809268e-05, "loss": 0.0001, "step": 83000 }, { "epoch": 17.9, "learning_rate": 2.9432904134647073e-05, "loss": 0.0002, "step": 83050 }, { "epoch": 17.91, "learning_rate": 2.94313660285256e-05, "loss": 0.0001, "step": 83100 }, { "epoch": 17.92, "learning_rate": 2.942982587966255e-05, "loss": 0.0001, "step": 83150 }, { "epoch": 17.93, "learning_rate": 2.9428283688275938e-05, "loss": 0.0001, "step": 83200 }, { "epoch": 17.95, "learning_rate": 2.9426739454584054e-05, "loss": 0.0001, "step": 83250 }, { "epoch": 17.96, "learning_rate": 2.942519317880549e-05, "loss": 0.0001, "step": 83300 }, { "epoch": 17.97, "learning_rate": 2.942364486115911e-05, "loss": 0.0001, "step": 83350 }, { "epoch": 17.98, "learning_rate": 2.9422094501864085e-05, "loss": 0.0001, "step": 83400 }, { "epoch": 17.99, "learning_rate": 2.942054210113986e-05, "loss": 0.0001, "step": 83450 }, { "epoch": 18.0, "learning_rate": 2.9418987659206178e-05, "loss": 0.0001, "step": 83500 }, { "epoch": 18.01, "learning_rate": 2.9417431176283066e-05, "loss": 0.0001, "step": 83550 }, { "epoch": 18.02, "learning_rate": 2.9415872652590846e-05, "loss": 0.0001, "step": 83600 }, { "epoch": 18.03, "learning_rate": 2.941431208835012e-05, "loss": 0.0001, "step": 83650 }, { "epoch": 18.04, "learning_rate": 2.9412749483781795e-05, "loss": 0.0001, "step": 83700 }, { "epoch": 18.05, "learning_rate": 2.941118483910704e-05, "loss": 0.0001, "step": 83750 }, { "epoch": 18.06, "learning_rate": 2.940961815454734e-05, "loss": 0.0, "step": 83800 }, { "epoch": 18.08, "learning_rate": 2.9408049430324455e-05, "loss": 0.0001, "step": 83850 }, { "epoch": 18.09, "learning_rate": 2.9406478666660432e-05, "loss": 0.0001, "step": 83900 }, { "epoch": 18.1, "learning_rate": 2.9404905863777614e-05, "loss": 0.0001, "step": 83950 }, { "epoch": 18.11, "learning_rate": 2.9403331021898626e-05, "loss": 0.0001, "step": 84000 }, { "epoch": 18.12, "learning_rate": 2.9401754141246387e-05, "loss": 0.0001, "step": 84050 }, { "epoch": 18.13, "learning_rate": 2.940017522204411e-05, "loss": 0.0001, "step": 84100 }, { "epoch": 18.14, "learning_rate": 2.939859426451527e-05, "loss": 0.0001, "step": 84150 }, { "epoch": 18.15, "learning_rate": 2.9397011268883667e-05, "loss": 0.0001, "step": 84200 }, { "epoch": 18.16, "learning_rate": 2.9395426235373367e-05, "loss": 0.0001, "step": 84250 }, { "epoch": 18.17, "learning_rate": 2.9393839164208727e-05, "loss": 0.0001, "step": 84300 }, { "epoch": 18.18, "learning_rate": 2.93922500556144e-05, "loss": 0.0001, "step": 84350 }, { "epoch": 18.19, "learning_rate": 2.9390658909815315e-05, "loss": 0.0001, "step": 84400 }, { "epoch": 18.2, "learning_rate": 2.93890657270367e-05, "loss": 0.0001, "step": 84450 }, { "epoch": 18.22, "learning_rate": 2.9387470507504074e-05, "loss": 0.0001, "step": 84500 }, { "epoch": 18.23, "learning_rate": 2.9385873251443223e-05, "loss": 0.0001, "step": 84550 }, { "epoch": 18.24, "learning_rate": 2.9384273959080256e-05, "loss": 0.0001, "step": 84600 }, { "epoch": 18.25, "learning_rate": 2.938267263064154e-05, "loss": 0.0001, "step": 84650 }, { "epoch": 18.26, "learning_rate": 2.938106926635374e-05, "loss": 0.0, "step": 84700 }, { "epoch": 18.27, "learning_rate": 2.9379463866443813e-05, "loss": 0.0001, "step": 84750 }, { "epoch": 18.28, "learning_rate": 2.9377856431138996e-05, "loss": 0.0001, "step": 84800 }, { "epoch": 18.29, "learning_rate": 2.9376246960666833e-05, "loss": 0.0001, "step": 84850 }, { "epoch": 18.3, "learning_rate": 2.9374635455255124e-05, "loss": 0.0001, "step": 84900 }, { "epoch": 18.31, "learning_rate": 2.9373021915131993e-05, "loss": 0.0, "step": 84950 }, { "epoch": 18.32, "learning_rate": 2.9371406340525823e-05, "loss": 0.0001, "step": 85000 }, { "epoch": 18.33, "learning_rate": 2.9369788731665303e-05, "loss": 0.0001, "step": 85050 }, { "epoch": 18.34, "learning_rate": 2.93681690887794e-05, "loss": 0.0001, "step": 85100 }, { "epoch": 18.36, "learning_rate": 2.936654741209737e-05, "loss": 0.0, "step": 85150 }, { "epoch": 18.37, "learning_rate": 2.9364923701848766e-05, "loss": 0.0001, "step": 85200 }, { "epoch": 18.38, "learning_rate": 2.936329795826342e-05, "loss": 0.0001, "step": 85250 }, { "epoch": 18.39, "learning_rate": 2.9361670181571448e-05, "loss": 0.0001, "step": 85300 }, { "epoch": 18.4, "learning_rate": 2.9360040372003267e-05, "loss": 0.0001, "step": 85350 }, { "epoch": 18.41, "learning_rate": 2.935840852978957e-05, "loss": 0.0001, "step": 85400 }, { "epoch": 18.42, "learning_rate": 2.9356774655161345e-05, "loss": 0.0001, "step": 85450 }, { "epoch": 18.43, "learning_rate": 2.9355138748349865e-05, "loss": 0.0001, "step": 85500 }, { "epoch": 18.44, "learning_rate": 2.9353500809586687e-05, "loss": 0.0, "step": 85550 }, { "epoch": 18.45, "learning_rate": 2.9351860839103665e-05, "loss": 0.0001, "step": 85600 }, { "epoch": 18.46, "learning_rate": 2.9350218837132926e-05, "loss": 0.0001, "step": 85650 }, { "epoch": 18.47, "learning_rate": 2.9348574803906906e-05, "loss": 0.0001, "step": 85700 }, { "epoch": 18.48, "learning_rate": 2.9346928739658303e-05, "loss": 0.0001, "step": 85750 }, { "epoch": 18.5, "learning_rate": 2.9345280644620124e-05, "loss": 0.0001, "step": 85800 }, { "epoch": 18.51, "learning_rate": 2.9343630519025646e-05, "loss": 0.0001, "step": 85850 }, { "epoch": 18.52, "learning_rate": 2.934197836310845e-05, "loss": 0.0001, "step": 85900 }, { "epoch": 18.53, "learning_rate": 2.9340324177102395e-05, "loss": 0.0001, "step": 85950 }, { "epoch": 18.54, "learning_rate": 2.9338667961241632e-05, "loss": 0.0001, "step": 86000 }, { "epoch": 18.55, "learning_rate": 2.9337009715760586e-05, "loss": 0.0001, "step": 86050 }, { "epoch": 18.56, "learning_rate": 2.933534944089399e-05, "loss": 0.0001, "step": 86100 }, { "epoch": 18.57, "learning_rate": 2.9333687136876846e-05, "loss": 0.0001, "step": 86150 }, { "epoch": 18.58, "learning_rate": 2.933202280394446e-05, "loss": 0.0001, "step": 86200 }, { "epoch": 18.59, "learning_rate": 2.933035644233241e-05, "loss": 0.0001, "step": 86250 }, { "epoch": 18.6, "learning_rate": 2.9328688052276563e-05, "loss": 0.0, "step": 86300 }, { "epoch": 18.61, "learning_rate": 2.9327017634013087e-05, "loss": 0.0001, "step": 86350 }, { "epoch": 18.62, "learning_rate": 2.9325345187778424e-05, "loss": 0.0001, "step": 86400 }, { "epoch": 18.64, "learning_rate": 2.9323670713809308e-05, "loss": 0.0001, "step": 86450 }, { "epoch": 18.65, "learning_rate": 2.9321994212342752e-05, "loss": 0.0001, "step": 86500 }, { "epoch": 18.66, "learning_rate": 2.932031568361607e-05, "loss": 0.0001, "step": 86550 }, { "epoch": 18.67, "learning_rate": 2.9318635127866854e-05, "loss": 0.0001, "step": 86600 }, { "epoch": 18.68, "learning_rate": 2.931695254533298e-05, "loss": 0.0001, "step": 86650 }, { "epoch": 18.69, "learning_rate": 2.9315267936252622e-05, "loss": 0.0001, "step": 86700 }, { "epoch": 18.7, "learning_rate": 2.9313581300864227e-05, "loss": 0.0001, "step": 86750 }, { "epoch": 18.71, "learning_rate": 2.931189263940655e-05, "loss": 0.0001, "step": 86800 }, { "epoch": 18.72, "learning_rate": 2.9310201952118598e-05, "loss": 0.0001, "step": 86850 }, { "epoch": 18.73, "learning_rate": 2.9308509239239703e-05, "loss": 0.0001, "step": 86900 }, { "epoch": 18.74, "learning_rate": 2.9306814501009458e-05, "loss": 0.0001, "step": 86950 }, { "epoch": 18.75, "learning_rate": 2.9305117737667754e-05, "loss": 0.0001, "step": 87000 }, { "epoch": 18.76, "learning_rate": 2.9303418949454766e-05, "loss": 0.0001, "step": 87050 }, { "epoch": 18.78, "learning_rate": 2.9301718136610955e-05, "loss": 0.0001, "step": 87100 }, { "epoch": 18.79, "learning_rate": 2.9300015299377068e-05, "loss": 0.0001, "step": 87150 }, { "epoch": 18.8, "learning_rate": 2.9298310437994138e-05, "loss": 0.0001, "step": 87200 }, { "epoch": 18.81, "learning_rate": 2.929660355270349e-05, "loss": 0.0001, "step": 87250 }, { "epoch": 18.82, "learning_rate": 2.9294894643746727e-05, "loss": 0.0001, "step": 87300 }, { "epoch": 18.83, "learning_rate": 2.9293183711365746e-05, "loss": 0.0001, "step": 87350 }, { "epoch": 18.84, "learning_rate": 2.9291470755802725e-05, "loss": 0.0001, "step": 87400 }, { "epoch": 18.85, "learning_rate": 2.9289755777300135e-05, "loss": 0.0001, "step": 87450 }, { "epoch": 18.86, "learning_rate": 2.9288038776100723e-05, "loss": 0.0001, "step": 87500 }, { "epoch": 18.87, "learning_rate": 2.9286319752447532e-05, "loss": 0.0001, "step": 87550 }, { "epoch": 18.88, "learning_rate": 2.928459870658389e-05, "loss": 0.0001, "step": 87600 }, { "epoch": 18.89, "learning_rate": 2.9282875638753406e-05, "loss": 0.0001, "step": 87650 }, { "epoch": 18.9, "learning_rate": 2.9281150549199975e-05, "loss": 0.0001, "step": 87700 }, { "epoch": 18.92, "learning_rate": 2.927942343816778e-05, "loss": 0.0001, "step": 87750 }, { "epoch": 18.93, "learning_rate": 2.9277694305901307e-05, "loss": 0.0001, "step": 87800 }, { "epoch": 18.94, "learning_rate": 2.9275963152645292e-05, "loss": 0.0001, "step": 87850 }, { "epoch": 18.95, "learning_rate": 2.927422997864479e-05, "loss": 0.0001, "step": 87900 }, { "epoch": 18.96, "learning_rate": 2.9272494784145123e-05, "loss": 0.0001, "step": 87950 }, { "epoch": 18.97, "learning_rate": 2.9270757569391912e-05, "loss": 0.0001, "step": 88000 }, { "epoch": 18.98, "learning_rate": 2.9269018334631055e-05, "loss": 0.0001, "step": 88050 }, { "epoch": 18.99, "learning_rate": 2.926727708010873e-05, "loss": 0.0, "step": 88100 }, { "epoch": 19.0, "learning_rate": 2.9265533806071426e-05, "loss": 0.0, "step": 88150 }, { "epoch": 19.01, "learning_rate": 2.926378851276589e-05, "loss": 0.0, "step": 88200 }, { "epoch": 19.02, "learning_rate": 2.926204120043917e-05, "loss": 0.0001, "step": 88250 }, { "epoch": 19.03, "learning_rate": 2.9260291869338595e-05, "loss": 0.0001, "step": 88300 }, { "epoch": 19.05, "learning_rate": 2.9258540519711773e-05, "loss": 0.0001, "step": 88350 }, { "epoch": 19.06, "learning_rate": 2.925678715180662e-05, "loss": 0.0001, "step": 88400 }, { "epoch": 19.07, "learning_rate": 2.925503176587131e-05, "loss": 0.0001, "step": 88450 }, { "epoch": 19.08, "learning_rate": 2.9253274362154323e-05, "loss": 0.0001, "step": 88500 }, { "epoch": 19.09, "learning_rate": 2.9251514940904412e-05, "loss": 0.0001, "step": 88550 }, { "epoch": 19.1, "learning_rate": 2.924975350237063e-05, "loss": 0.0001, "step": 88600 }, { "epoch": 19.11, "learning_rate": 2.92479900468023e-05, "loss": 0.0001, "step": 88650 }, { "epoch": 19.12, "learning_rate": 2.9246224574449027e-05, "loss": 0.0001, "step": 88700 }, { "epoch": 19.13, "learning_rate": 2.9244457085560733e-05, "loss": 0.0, "step": 88750 }, { "epoch": 19.14, "learning_rate": 2.9242687580387587e-05, "loss": 0.0001, "step": 88800 }, { "epoch": 19.15, "learning_rate": 2.9240916059180068e-05, "loss": 0.0001, "step": 88850 }, { "epoch": 19.16, "learning_rate": 2.9239142522188928e-05, "loss": 0.0001, "step": 88900 }, { "epoch": 19.17, "learning_rate": 2.9237366969665215e-05, "loss": 0.0001, "step": 88950 }, { "epoch": 19.19, "learning_rate": 2.923558940186025e-05, "loss": 0.0001, "step": 89000 }, { "epoch": 19.2, "learning_rate": 2.9233809819025646e-05, "loss": 0.0001, "step": 89050 }, { "epoch": 19.21, "learning_rate": 2.9232028221413305e-05, "loss": 0.0001, "step": 89100 }, { "epoch": 19.22, "learning_rate": 2.9230244609275407e-05, "loss": 0.0001, "step": 89150 }, { "epoch": 19.23, "learning_rate": 2.922845898286442e-05, "loss": 0.0001, "step": 89200 }, { "epoch": 19.24, "learning_rate": 2.92266713424331e-05, "loss": 0.0001, "step": 89250 }, { "epoch": 19.25, "learning_rate": 2.922488168823448e-05, "loss": 0.0001, "step": 89300 }, { "epoch": 19.26, "learning_rate": 2.9223090020521888e-05, "loss": 0.0001, "step": 89350 }, { "epoch": 19.27, "learning_rate": 2.9221296339548925e-05, "loss": 0.0001, "step": 89400 }, { "epoch": 19.28, "learning_rate": 2.9219500645569495e-05, "loss": 0.0001, "step": 89450 }, { "epoch": 19.29, "learning_rate": 2.921770293883777e-05, "loss": 0.0001, "step": 89500 }, { "epoch": 19.3, "learning_rate": 2.9215903219608214e-05, "loss": 0.0001, "step": 89550 }, { "epoch": 19.31, "learning_rate": 2.9214101488135575e-05, "loss": 0.0001, "step": 89600 }, { "epoch": 19.33, "learning_rate": 2.921229774467489e-05, "loss": 0.0001, "step": 89650 }, { "epoch": 19.34, "learning_rate": 2.9210491989481466e-05, "loss": 0.0001, "step": 89700 }, { "epoch": 19.35, "learning_rate": 2.9208684222810916e-05, "loss": 0.0001, "step": 89750 }, { "epoch": 19.36, "learning_rate": 2.9206874444919114e-05, "loss": 0.0001, "step": 89800 }, { "epoch": 19.37, "learning_rate": 2.9205062656062253e-05, "loss": 0.0001, "step": 89850 }, { "epoch": 19.38, "learning_rate": 2.920324885649677e-05, "loss": 0.0001, "step": 89900 }, { "epoch": 19.39, "learning_rate": 2.9201433046479413e-05, "loss": 0.0001, "step": 89950 }, { "epoch": 19.4, "learning_rate": 2.919961522626721e-05, "loss": 0.0001, "step": 90000 }, { "epoch": 19.41, "learning_rate": 2.9197795396117468e-05, "loss": 0.0001, "step": 90050 }, { "epoch": 19.42, "learning_rate": 2.9195973556287783e-05, "loss": 0.0001, "step": 90100 }, { "epoch": 19.43, "learning_rate": 2.9194149707036032e-05, "loss": 0.0001, "step": 90150 }, { "epoch": 19.44, "learning_rate": 2.9192323848620383e-05, "loss": 0.0, "step": 90200 }, { "epoch": 19.45, "learning_rate": 2.919049598129928e-05, "loss": 0.0001, "step": 90250 }, { "epoch": 19.47, "learning_rate": 2.9188666105331455e-05, "loss": 0.0001, "step": 90300 }, { "epoch": 19.48, "learning_rate": 2.918683422097593e-05, "loss": 0.0001, "step": 90350 }, { "epoch": 19.49, "learning_rate": 2.9185000328491997e-05, "loss": 0.0001, "step": 90400 }, { "epoch": 19.5, "learning_rate": 2.9183164428139248e-05, "loss": 0.0001, "step": 90450 }, { "epoch": 19.51, "learning_rate": 2.918132652017755e-05, "loss": 0.0001, "step": 90500 }, { "epoch": 19.52, "learning_rate": 2.9179486604867057e-05, "loss": 0.0, "step": 90550 }, { "epoch": 19.53, "learning_rate": 2.917764468246821e-05, "loss": 0.0001, "step": 90600 }, { "epoch": 19.54, "learning_rate": 2.917580075324172e-05, "loss": 0.0001, "step": 90650 }, { "epoch": 19.55, "learning_rate": 2.9173954817448603e-05, "loss": 0.0001, "step": 90700 }, { "epoch": 19.56, "learning_rate": 2.9172106875350148e-05, "loss": 0.0001, "step": 90750 }, { "epoch": 19.57, "learning_rate": 2.9170256927207926e-05, "loss": 0.0001, "step": 90800 }, { "epoch": 19.58, "learning_rate": 2.91684049732838e-05, "loss": 0.0001, "step": 90850 }, { "epoch": 19.59, "learning_rate": 2.91665510138399e-05, "loss": 0.0001, "step": 90900 }, { "epoch": 19.61, "learning_rate": 2.9164695049138667e-05, "loss": 0.0001, "step": 90950 }, { "epoch": 19.62, "learning_rate": 2.91628370794428e-05, "loss": 0.0001, "step": 91000 }, { "epoch": 19.63, "learning_rate": 2.9160977105015292e-05, "loss": 0.0001, "step": 91050 }, { "epoch": 19.64, "learning_rate": 2.9159115126119426e-05, "loss": 0.0001, "step": 91100 }, { "epoch": 19.65, "learning_rate": 2.9157251143018757e-05, "loss": 0.0001, "step": 91150 }, { "epoch": 19.66, "learning_rate": 2.9155385155977134e-05, "loss": 0.0001, "step": 91200 }, { "epoch": 19.67, "learning_rate": 2.915351716525868e-05, "loss": 0.0001, "step": 91250 }, { "epoch": 19.68, "learning_rate": 2.9151647171127817e-05, "loss": 0.0001, "step": 91300 }, { "epoch": 19.69, "learning_rate": 2.9149775173849225e-05, "loss": 0.0001, "step": 91350 }, { "epoch": 19.7, "learning_rate": 2.91479011736879e-05, "loss": 0.0001, "step": 91400 }, { "epoch": 19.71, "learning_rate": 2.9146025170909084e-05, "loss": 0.0001, "step": 91450 }, { "epoch": 19.72, "learning_rate": 2.9144147165778344e-05, "loss": 0.0, "step": 91500 }, { "epoch": 19.73, "learning_rate": 2.9142267158561495e-05, "loss": 0.0001, "step": 91550 }, { "epoch": 19.75, "learning_rate": 2.9140385149524655e-05, "loss": 0.0001, "step": 91600 }, { "epoch": 19.76, "learning_rate": 2.913850113893422e-05, "loss": 0.0001, "step": 91650 }, { "epoch": 19.77, "learning_rate": 2.9136615127056866e-05, "loss": 0.0001, "step": 91700 }, { "epoch": 19.78, "learning_rate": 2.913472711415956e-05, "loss": 0.0001, "step": 91750 }, { "epoch": 19.79, "learning_rate": 2.9132837100509547e-05, "loss": 0.0001, "step": 91800 }, { "epoch": 19.8, "learning_rate": 2.9130945086374354e-05, "loss": 0.0001, "step": 91850 }, { "epoch": 19.81, "learning_rate": 2.9129051072021797e-05, "loss": 0.0001, "step": 91900 }, { "epoch": 19.82, "learning_rate": 2.912715505771996e-05, "loss": 0.0001, "step": 91950 }, { "epoch": 19.83, "learning_rate": 2.9125257043737237e-05, "loss": 0.0001, "step": 92000 }, { "epoch": 19.84, "learning_rate": 2.912335703034228e-05, "loss": 0.0001, "step": 92050 }, { "epoch": 19.85, "learning_rate": 2.9121455017804032e-05, "loss": 0.0001, "step": 92100 }, { "epoch": 19.86, "learning_rate": 2.9119551006391732e-05, "loss": 0.0001, "step": 92150 }, { "epoch": 19.87, "learning_rate": 2.911764499637488e-05, "loss": 0.0001, "step": 92200 }, { "epoch": 19.89, "learning_rate": 2.9115736988023267e-05, "loss": 0.0001, "step": 92250 }, { "epoch": 19.9, "learning_rate": 2.911382698160698e-05, "loss": 0.0001, "step": 92300 }, { "epoch": 19.91, "learning_rate": 2.911191497739637e-05, "loss": 0.0001, "step": 92350 }, { "epoch": 19.92, "learning_rate": 2.911000097566208e-05, "loss": 0.0001, "step": 92400 }, { "epoch": 19.93, "learning_rate": 2.9108084976675036e-05, "loss": 0.0001, "step": 92450 }, { "epoch": 19.94, "learning_rate": 2.910616698070645e-05, "loss": 0.0001, "step": 92500 }, { "epoch": 19.95, "learning_rate": 2.9104246988027796e-05, "loss": 0.0001, "step": 92550 }, { "epoch": 19.96, "learning_rate": 2.9102324998910864e-05, "loss": 0.0001, "step": 92600 }, { "epoch": 19.97, "learning_rate": 2.9100401013627703e-05, "loss": 0.0001, "step": 92650 }, { "epoch": 19.98, "learning_rate": 2.9098475032450646e-05, "loss": 0.0001, "step": 92700 }, { "epoch": 19.99, "learning_rate": 2.9096547055652318e-05, "loss": 0.0001, "step": 92750 }, { "epoch": 20.0, "learning_rate": 2.909461708350562e-05, "loss": 0.0001, "step": 92800 }, { "epoch": 20.02, "learning_rate": 2.9092685116283738e-05, "loss": 0.0, "step": 92850 }, { "epoch": 20.03, "learning_rate": 2.9090751154260145e-05, "loss": 0.0, "step": 92900 }, { "epoch": 20.04, "learning_rate": 2.908881519770858e-05, "loss": 0.0, "step": 92950 }, { "epoch": 20.05, "learning_rate": 2.9086877246903082e-05, "loss": 0.0001, "step": 93000 }, { "epoch": 20.06, "learning_rate": 2.908493730211796e-05, "loss": 0.0, "step": 93050 }, { "epoch": 20.07, "learning_rate": 2.9082995363627818e-05, "loss": 0.0001, "step": 93100 }, { "epoch": 20.08, "learning_rate": 2.908105143170753e-05, "loss": 0.0, "step": 93150 }, { "epoch": 20.09, "learning_rate": 2.9079105506632264e-05, "loss": 0.0001, "step": 93200 }, { "epoch": 20.1, "learning_rate": 2.9077157588677452e-05, "loss": 0.0, "step": 93250 }, { "epoch": 20.11, "learning_rate": 2.9075207678118828e-05, "loss": 0.0001, "step": 93300 }, { "epoch": 20.12, "learning_rate": 2.9073255775232395e-05, "loss": 0.0, "step": 93350 }, { "epoch": 20.13, "learning_rate": 2.9071301880294444e-05, "loss": 0.0001, "step": 93400 }, { "epoch": 20.14, "learning_rate": 2.906934599358155e-05, "loss": 0.0001, "step": 93450 }, { "epoch": 20.16, "learning_rate": 2.906738811537056e-05, "loss": 0.0001, "step": 93500 }, { "epoch": 20.17, "learning_rate": 2.906542824593861e-05, "loss": 0.0, "step": 93550 }, { "epoch": 20.18, "learning_rate": 2.906346638556312e-05, "loss": 0.0001, "step": 93600 }, { "epoch": 20.19, "learning_rate": 2.9061502534521786e-05, "loss": 0.0001, "step": 93650 }, { "epoch": 20.2, "learning_rate": 2.9059536693092594e-05, "loss": 0.0001, "step": 93700 }, { "epoch": 20.21, "learning_rate": 2.90575688615538e-05, "loss": 0.0, "step": 93750 }, { "epoch": 20.22, "learning_rate": 2.905559904018395e-05, "loss": 0.0, "step": 93800 }, { "epoch": 20.23, "learning_rate": 2.9053627229261874e-05, "loss": 0.0001, "step": 93850 }, { "epoch": 20.24, "learning_rate": 2.9051653429066674e-05, "loss": 0.0001, "step": 93900 }, { "epoch": 20.25, "learning_rate": 2.9049677639877736e-05, "loss": 0.0001, "step": 93950 }, { "epoch": 20.26, "learning_rate": 2.904769986197474e-05, "loss": 0.0001, "step": 94000 }, { "epoch": 20.27, "learning_rate": 2.9045720095637633e-05, "loss": 0.0, "step": 94050 }, { "epoch": 20.28, "learning_rate": 2.9043738341146648e-05, "loss": 0.0001, "step": 94100 }, { "epoch": 20.3, "learning_rate": 2.90417545987823e-05, "loss": 0.0001, "step": 94150 }, { "epoch": 20.31, "learning_rate": 2.9039768868825386e-05, "loss": 0.0, "step": 94200 }, { "epoch": 20.32, "learning_rate": 2.9037781151556986e-05, "loss": 0.0001, "step": 94250 }, { "epoch": 20.33, "learning_rate": 2.903579144725845e-05, "loss": 0.0001, "step": 94300 }, { "epoch": 20.34, "learning_rate": 2.903379975621143e-05, "loss": 0.0001, "step": 94350 }, { "epoch": 20.35, "learning_rate": 2.9031806078697836e-05, "loss": 0.0001, "step": 94400 }, { "epoch": 20.36, "learning_rate": 2.902981041499988e-05, "loss": 0.0001, "step": 94450 }, { "epoch": 20.37, "learning_rate": 2.902781276540004e-05, "loss": 0.0001, "step": 94500 }, { "epoch": 20.38, "learning_rate": 2.902581313018108e-05, "loss": 0.0001, "step": 94550 }, { "epoch": 20.39, "learning_rate": 2.902381150962605e-05, "loss": 0.0001, "step": 94600 }, { "epoch": 20.4, "learning_rate": 2.902180790401828e-05, "loss": 0.0001, "step": 94650 }, { "epoch": 20.41, "learning_rate": 2.9019802313641367e-05, "loss": 0.0, "step": 94700 }, { "epoch": 20.42, "learning_rate": 2.9017794738779208e-05, "loss": 0.0, "step": 94750 }, { "epoch": 20.44, "learning_rate": 2.901578517971597e-05, "loss": 0.0001, "step": 94800 }, { "epoch": 20.45, "learning_rate": 2.9013773636736103e-05, "loss": 0.0001, "step": 94850 }, { "epoch": 20.46, "learning_rate": 2.901176011012434e-05, "loss": 0.0001, "step": 94900 }, { "epoch": 20.47, "learning_rate": 2.9009744600165694e-05, "loss": 0.0, "step": 94950 }, { "epoch": 20.48, "learning_rate": 2.9007727107145457e-05, "loss": 0.0001, "step": 95000 }, { "epoch": 20.49, "learning_rate": 2.90057076313492e-05, "loss": 0.0001, "step": 95050 }, { "epoch": 20.5, "learning_rate": 2.9003686173062773e-05, "loss": 0.0001, "step": 95100 }, { "epoch": 20.51, "learning_rate": 2.9001662732572324e-05, "loss": 0.0001, "step": 95150 }, { "epoch": 20.52, "learning_rate": 2.899963731016426e-05, "loss": 0.0001, "step": 95200 }, { "epoch": 20.53, "learning_rate": 2.8997609906125282e-05, "loss": 0.0001, "step": 95250 }, { "epoch": 20.54, "learning_rate": 2.8995580520742354e-05, "loss": 0.0001, "step": 95300 }, { "epoch": 20.55, "learning_rate": 2.899354915430275e-05, "loss": 0.0001, "step": 95350 }, { "epoch": 20.56, "learning_rate": 2.8991515807093995e-05, "loss": 0.0001, "step": 95400 }, { "epoch": 20.58, "learning_rate": 2.8989480479403917e-05, "loss": 0.0001, "step": 95450 }, { "epoch": 20.59, "learning_rate": 2.89874431715206e-05, "loss": 0.0001, "step": 95500 }, { "epoch": 20.6, "learning_rate": 2.8985403883732432e-05, "loss": 0.0001, "step": 95550 }, { "epoch": 20.61, "learning_rate": 2.898336261632807e-05, "loss": 0.0001, "step": 95600 }, { "epoch": 20.62, "learning_rate": 2.8981319369596455e-05, "loss": 0.0, "step": 95650 }, { "epoch": 20.63, "learning_rate": 2.8979274143826802e-05, "loss": 0.0001, "step": 95700 }, { "epoch": 20.64, "learning_rate": 2.8977226939308614e-05, "loss": 0.0001, "step": 95750 }, { "epoch": 20.65, "learning_rate": 2.8975177756331667e-05, "loss": 0.0001, "step": 95800 }, { "epoch": 20.66, "learning_rate": 2.897312659518602e-05, "loss": 0.0, "step": 95850 }, { "epoch": 20.67, "learning_rate": 2.8971073456162015e-05, "loss": 0.0001, "step": 95900 }, { "epoch": 20.68, "learning_rate": 2.8969018339550266e-05, "loss": 0.0001, "step": 95950 }, { "epoch": 20.69, "learning_rate": 2.8966961245641678e-05, "loss": 0.0001, "step": 96000 }, { "epoch": 20.7, "learning_rate": 2.896490217472743e-05, "loss": 0.0001, "step": 96050 }, { "epoch": 20.72, "learning_rate": 2.8962841127098976e-05, "loss": 0.0001, "step": 96100 }, { "epoch": 20.73, "learning_rate": 2.896077810304806e-05, "loss": 0.0001, "step": 96150 }, { "epoch": 20.74, "learning_rate": 2.8958713102866697e-05, "loss": 0.0001, "step": 96200 }, { "epoch": 20.75, "learning_rate": 2.8956646126847184e-05, "loss": 0.0001, "step": 96250 }, { "epoch": 20.76, "learning_rate": 2.8954577175282104e-05, "loss": 0.0001, "step": 96300 }, { "epoch": 20.77, "learning_rate": 2.895250624846431e-05, "loss": 0.0001, "step": 96350 }, { "epoch": 20.78, "learning_rate": 2.8950433346686936e-05, "loss": 0.0001, "step": 96400 }, { "epoch": 20.79, "learning_rate": 2.8948358470243413e-05, "loss": 0.0001, "step": 96450 }, { "epoch": 20.8, "learning_rate": 2.8946281619427418e-05, "loss": 0.0001, "step": 96500 }, { "epoch": 20.81, "learning_rate": 2.894420279453294e-05, "loss": 0.0001, "step": 96550 }, { "epoch": 20.82, "learning_rate": 2.894212199585423e-05, "loss": 0.0001, "step": 96600 }, { "epoch": 20.83, "learning_rate": 2.894003922368582e-05, "loss": 0.0001, "step": 96650 }, { "epoch": 20.85, "learning_rate": 2.8937954478322528e-05, "loss": 0.0001, "step": 96700 }, { "epoch": 20.86, "learning_rate": 2.8935867760059443e-05, "loss": 0.0001, "step": 96750 }, { "epoch": 20.87, "learning_rate": 2.893377906919194e-05, "loss": 0.0001, "step": 96800 }, { "epoch": 20.88, "learning_rate": 2.8931688406015672e-05, "loss": 0.0001, "step": 96850 }, { "epoch": 20.89, "learning_rate": 2.8929595770826563e-05, "loss": 0.0001, "step": 96900 }, { "epoch": 20.9, "learning_rate": 2.8927501163920835e-05, "loss": 0.0001, "step": 96950 }, { "epoch": 20.91, "learning_rate": 2.8925404585594964e-05, "loss": 0.0001, "step": 97000 }, { "epoch": 20.92, "learning_rate": 2.8923306036145726e-05, "loss": 0.0001, "step": 97050 }, { "epoch": 20.93, "learning_rate": 2.892120551587016e-05, "loss": 0.0001, "step": 97100 }, { "epoch": 20.94, "learning_rate": 2.8919103025065606e-05, "loss": 0.0001, "step": 97150 }, { "epoch": 20.95, "learning_rate": 2.8916998564029654e-05, "loss": 0.0001, "step": 97200 }, { "epoch": 20.96, "learning_rate": 2.8914892133060197e-05, "loss": 0.0001, "step": 97250 }, { "epoch": 20.97, "learning_rate": 2.8912783732455396e-05, "loss": 0.0, "step": 97300 }, { "epoch": 20.99, "learning_rate": 2.8910673362513686e-05, "loss": 0.0001, "step": 97350 }, { "epoch": 21.0, "learning_rate": 2.8908561023533797e-05, "loss": 0.0001, "step": 97400 }, { "epoch": 21.01, "learning_rate": 2.890644671581472e-05, "loss": 0.0, "step": 97450 }, { "epoch": 21.02, "learning_rate": 2.890433043965574e-05, "loss": 0.0001, "step": 97500 }, { "epoch": 21.03, "learning_rate": 2.8902212195356405e-05, "loss": 0.0001, "step": 97550 }, { "epoch": 21.04, "learning_rate": 2.8900091983216557e-05, "loss": 0.0001, "step": 97600 }, { "epoch": 21.05, "learning_rate": 2.8897969803536306e-05, "loss": 0.0001, "step": 97650 }, { "epoch": 21.06, "learning_rate": 2.8895845656616043e-05, "loss": 0.0001, "step": 97700 }, { "epoch": 21.07, "learning_rate": 2.889371954275644e-05, "loss": 0.0001, "step": 97750 }, { "epoch": 21.08, "learning_rate": 2.889159146225844e-05, "loss": 0.0001, "step": 97800 }, { "epoch": 21.09, "learning_rate": 2.888946141542328e-05, "loss": 0.0, "step": 97850 }, { "epoch": 21.1, "learning_rate": 2.8887329402552462e-05, "loss": 0.0001, "step": 97900 }, { "epoch": 21.11, "learning_rate": 2.8885195423947768e-05, "loss": 0.0001, "step": 97950 }, { "epoch": 21.13, "learning_rate": 2.888305947991126e-05, "loss": 0.0, "step": 98000 }, { "epoch": 21.14, "learning_rate": 2.8880921570745275e-05, "loss": 0.0001, "step": 98050 }, { "epoch": 21.15, "learning_rate": 2.887878169675244e-05, "loss": 0.0001, "step": 98100 }, { "epoch": 21.16, "learning_rate": 2.887663985823564e-05, "loss": 0.0001, "step": 98150 }, { "epoch": 21.17, "learning_rate": 2.8874496055498065e-05, "loss": 0.0001, "step": 98200 }, { "epoch": 21.18, "learning_rate": 2.887235028884315e-05, "loss": 0.0001, "step": 98250 }, { "epoch": 21.19, "learning_rate": 2.8870202558574636e-05, "loss": 0.0001, "step": 98300 }, { "epoch": 21.2, "learning_rate": 2.886805286499653e-05, "loss": 0.0001, "step": 98350 }, { "epoch": 21.21, "learning_rate": 2.8865901208413118e-05, "loss": 0.0001, "step": 98400 }, { "epoch": 21.22, "learning_rate": 2.8863747589128963e-05, "loss": 0.0, "step": 98450 }, { "epoch": 21.23, "learning_rate": 2.886159200744891e-05, "loss": 0.0001, "step": 98500 }, { "epoch": 21.24, "learning_rate": 2.8859434463678077e-05, "loss": 0.0001, "step": 98550 }, { "epoch": 21.25, "learning_rate": 2.885727495812186e-05, "loss": 0.0001, "step": 98600 }, { "epoch": 21.27, "learning_rate": 2.8855113491085935e-05, "loss": 0.0001, "step": 98650 }, { "epoch": 21.28, "learning_rate": 2.885295006287626e-05, "loss": 0.0001, "step": 98700 }, { "epoch": 21.29, "learning_rate": 2.8850784673799058e-05, "loss": 0.0001, "step": 98750 }, { "epoch": 21.3, "learning_rate": 2.8848617324160844e-05, "loss": 0.0001, "step": 98800 }, { "epoch": 21.31, "learning_rate": 2.88464480142684e-05, "loss": 0.0001, "step": 98850 }, { "epoch": 21.32, "learning_rate": 2.8844276744428782e-05, "loss": 0.0001, "step": 98900 }, { "epoch": 21.33, "learning_rate": 2.884210351494934e-05, "loss": 0.0001, "step": 98950 }, { "epoch": 21.34, "learning_rate": 2.8839928326137695e-05, "loss": 0.0001, "step": 99000 }, { "epoch": 21.35, "learning_rate": 2.883775117830173e-05, "loss": 0.0001, "step": 99050 }, { "epoch": 21.36, "learning_rate": 2.883557207174963e-05, "loss": 0.0, "step": 99100 }, { "epoch": 21.37, "learning_rate": 2.8833391006789836e-05, "loss": 0.0001, "step": 99150 }, { "epoch": 21.38, "learning_rate": 2.8831207983731077e-05, "loss": 0.0001, "step": 99200 }, { "epoch": 21.39, "learning_rate": 2.882902300288236e-05, "loss": 0.0001, "step": 99250 }, { "epoch": 21.41, "learning_rate": 2.8826836064552962e-05, "loss": 0.0001, "step": 99300 }, { "epoch": 21.42, "learning_rate": 2.8824647169052445e-05, "loss": 0.0001, "step": 99350 }, { "epoch": 21.43, "learning_rate": 2.8822456316690644e-05, "loss": 0.0001, "step": 99400 }, { "epoch": 21.44, "learning_rate": 2.882026350777767e-05, "loss": 0.0001, "step": 99450 }, { "epoch": 21.45, "learning_rate": 2.8818068742623912e-05, "loss": 0.0, "step": 99500 }, { "epoch": 21.46, "learning_rate": 2.8815872021540034e-05, "loss": 0.0, "step": 99550 }, { "epoch": 21.47, "learning_rate": 2.881367334483699e-05, "loss": 0.0001, "step": 99600 }, { "epoch": 21.48, "learning_rate": 2.8811472712825988e-05, "loss": 0.0001, "step": 99650 }, { "epoch": 21.49, "learning_rate": 2.8809270125818525e-05, "loss": 0.0001, "step": 99700 }, { "epoch": 21.5, "learning_rate": 2.880706558412638e-05, "loss": 0.0001, "step": 99750 }, { "epoch": 21.51, "learning_rate": 2.8804859088061605e-05, "loss": 0.0001, "step": 99800 }, { "epoch": 21.52, "learning_rate": 2.880265063793652e-05, "loss": 0.0, "step": 99850 }, { "epoch": 21.53, "learning_rate": 2.880044023406373e-05, "loss": 0.0001, "step": 99900 }, { "epoch": 21.55, "learning_rate": 2.8798227876756118e-05, "loss": 0.0001, "step": 99950 }, { "epoch": 21.56, "learning_rate": 2.8796013566326837e-05, "loss": 0.0001, "step": 100000 }, { "epoch": 21.57, "learning_rate": 2.8793797303089324e-05, "loss": 0.0001, "step": 100050 }, { "epoch": 21.58, "learning_rate": 2.879157908735728e-05, "loss": 0.0001, "step": 100100 }, { "epoch": 21.59, "learning_rate": 2.8789358919444704e-05, "loss": 0.0, "step": 100150 }, { "epoch": 21.6, "learning_rate": 2.8787136799665844e-05, "loss": 0.0, "step": 100200 }, { "epoch": 21.61, "learning_rate": 2.8784912728335248e-05, "loss": 0.0001, "step": 100250 }, { "epoch": 21.62, "learning_rate": 2.8782686705767726e-05, "loss": 0.0001, "step": 100300 }, { "epoch": 21.63, "learning_rate": 2.8780458732278365e-05, "loss": 0.0001, "step": 100350 }, { "epoch": 21.64, "learning_rate": 2.8778228808182536e-05, "loss": 0.0001, "step": 100400 }, { "epoch": 21.65, "learning_rate": 2.877599693379589e-05, "loss": 0.0001, "step": 100450 }, { "epoch": 21.66, "learning_rate": 2.8773763109434334e-05, "loss": 0.0001, "step": 100500 }, { "epoch": 21.67, "learning_rate": 2.8771527335414064e-05, "loss": 0.0001, "step": 100550 }, { "epoch": 21.69, "learning_rate": 2.8769289612051556e-05, "loss": 0.0001, "step": 100600 }, { "epoch": 21.7, "learning_rate": 2.876704993966356e-05, "loss": 0.0001, "step": 100650 }, { "epoch": 21.71, "learning_rate": 2.876480831856709e-05, "loss": 0.0001, "step": 100700 }, { "epoch": 21.72, "learning_rate": 2.8762564749079448e-05, "loss": 0.0001, "step": 100750 }, { "epoch": 21.73, "learning_rate": 2.876031923151821e-05, "loss": 0.0001, "step": 100800 }, { "epoch": 21.74, "learning_rate": 2.8758071766201226e-05, "loss": 0.0001, "step": 100850 }, { "epoch": 21.75, "learning_rate": 2.875582235344662e-05, "loss": 0.0001, "step": 100900 }, { "epoch": 21.76, "learning_rate": 2.8753570993572795e-05, "loss": 0.0001, "step": 100950 }, { "epoch": 21.77, "learning_rate": 2.8751317686898428e-05, "loss": 0.0001, "step": 101000 }, { "epoch": 21.78, "learning_rate": 2.8749062433742467e-05, "loss": 0.0001, "step": 101050 }, { "epoch": 21.79, "learning_rate": 2.8746805234424148e-05, "loss": 0.0, "step": 101100 }, { "epoch": 21.8, "learning_rate": 2.874454608926297e-05, "loss": 0.0001, "step": 101150 }, { "epoch": 21.82, "learning_rate": 2.8742284998578716e-05, "loss": 0.0001, "step": 101200 }, { "epoch": 21.83, "learning_rate": 2.8740021962691434e-05, "loss": 0.0001, "step": 101250 }, { "epoch": 21.84, "learning_rate": 2.8737756981921458e-05, "loss": 0.0001, "step": 101300 }, { "epoch": 21.85, "learning_rate": 2.87354900565894e-05, "loss": 0.0001, "step": 101350 }, { "epoch": 21.86, "learning_rate": 2.8733221187016125e-05, "loss": 0.0001, "step": 101400 }, { "epoch": 21.87, "learning_rate": 2.87309503735228e-05, "loss": 0.0001, "step": 101450 }, { "epoch": 21.88, "learning_rate": 2.8728677616430855e-05, "loss": 0.0001, "step": 101500 }, { "epoch": 21.89, "learning_rate": 2.872640291606199e-05, "loss": 0.0001, "step": 101550 }, { "epoch": 21.9, "learning_rate": 2.8724126272738195e-05, "loss": 0.0001, "step": 101600 }, { "epoch": 21.91, "learning_rate": 2.872184768678172e-05, "loss": 0.0, "step": 101650 }, { "epoch": 21.92, "learning_rate": 2.871956715851509e-05, "loss": 0.0001, "step": 101700 }, { "epoch": 21.93, "learning_rate": 2.871728468826112e-05, "loss": 0.0, "step": 101750 }, { "epoch": 21.94, "learning_rate": 2.871500027634289e-05, "loss": 0.0, "step": 101800 }, { "epoch": 21.96, "learning_rate": 2.871271392308375e-05, "loss": 0.0001, "step": 101850 }, { "epoch": 21.97, "learning_rate": 2.8710425628807337e-05, "loss": 0.0001, "step": 101900 }, { "epoch": 21.98, "learning_rate": 2.870813539383755e-05, "loss": 0.0001, "step": 101950 }, { "epoch": 21.99, "learning_rate": 2.8705843218498576e-05, "loss": 0.0002, "step": 102000 }, { "epoch": 22.0, "learning_rate": 2.870354910311487e-05, "loss": 0.0001, "step": 102050 }, { "epoch": 22.01, "learning_rate": 2.870125304801115e-05, "loss": 0.0, "step": 102100 }, { "epoch": 22.02, "learning_rate": 2.8698955053512426e-05, "loss": 0.0, "step": 102150 }, { "epoch": 22.03, "learning_rate": 2.8696655119943978e-05, "loss": 0.0, "step": 102200 }, { "epoch": 22.04, "learning_rate": 2.869435324763136e-05, "loss": 0.0001, "step": 102250 }, { "epoch": 22.05, "learning_rate": 2.869204943690039e-05, "loss": 0.0, "step": 102300 }, { "epoch": 22.06, "learning_rate": 2.8689743688077184e-05, "loss": 0.0001, "step": 102350 }, { "epoch": 22.07, "learning_rate": 2.8687436001488106e-05, "loss": 0.0, "step": 102400 }, { "epoch": 22.08, "learning_rate": 2.868512637745981e-05, "loss": 0.0001, "step": 102450 }, { "epoch": 22.1, "learning_rate": 2.8682814816319222e-05, "loss": 0.0001, "step": 102500 }, { "epoch": 22.11, "learning_rate": 2.8680501318393536e-05, "loss": 0.0, "step": 102550 }, { "epoch": 22.12, "learning_rate": 2.867818588401023e-05, "loss": 0.0001, "step": 102600 }, { "epoch": 22.13, "learning_rate": 2.8675868513497048e-05, "loss": 0.0, "step": 102650 }, { "epoch": 22.14, "learning_rate": 2.8673549207182012e-05, "loss": 0.0, "step": 102700 }, { "epoch": 22.15, "learning_rate": 2.8671227965393416e-05, "loss": 0.0001, "step": 102750 }, { "epoch": 22.16, "learning_rate": 2.8668904788459826e-05, "loss": 0.0, "step": 102800 }, { "epoch": 22.17, "learning_rate": 2.866657967671009e-05, "loss": 0.0001, "step": 102850 }, { "epoch": 22.18, "learning_rate": 2.8664252630473322e-05, "loss": 0.0001, "step": 102900 }, { "epoch": 22.19, "learning_rate": 2.866192365007891e-05, "loss": 0.0001, "step": 102950 }, { "epoch": 22.2, "learning_rate": 2.8659592735856526e-05, "loss": 0.0001, "step": 103000 }, { "epoch": 22.21, "learning_rate": 2.86572598881361e-05, "loss": 0.0001, "step": 103050 }, { "epoch": 22.22, "learning_rate": 2.8654925107247843e-05, "loss": 0.0001, "step": 103100 }, { "epoch": 22.24, "learning_rate": 2.865258839352225e-05, "loss": 0.0001, "step": 103150 }, { "epoch": 22.25, "learning_rate": 2.865024974729007e-05, "loss": 0.0, "step": 103200 }, { "epoch": 22.26, "learning_rate": 2.8647909168882342e-05, "loss": 0.0001, "step": 103250 }, { "epoch": 22.27, "learning_rate": 2.864556665863037e-05, "loss": 0.0001, "step": 103300 }, { "epoch": 22.28, "learning_rate": 2.8643222216865724e-05, "loss": 0.0001, "step": 103350 }, { "epoch": 22.29, "learning_rate": 2.8640875843920274e-05, "loss": 0.0001, "step": 103400 }, { "epoch": 22.3, "learning_rate": 2.8638527540126135e-05, "loss": 0.0001, "step": 103450 }, { "epoch": 22.31, "learning_rate": 2.8636177305815705e-05, "loss": 0.0, "step": 103500 }, { "epoch": 22.32, "learning_rate": 2.8633825141321662e-05, "loss": 0.0001, "step": 103550 }, { "epoch": 22.33, "learning_rate": 2.8631471046976948e-05, "loss": 0.0001, "step": 103600 }, { "epoch": 22.34, "learning_rate": 2.862911502311479e-05, "loss": 0.0001, "step": 103650 }, { "epoch": 22.35, "learning_rate": 2.8626757070068674e-05, "loss": 0.0001, "step": 103700 }, { "epoch": 22.36, "learning_rate": 2.8624397188172366e-05, "loss": 0.0001, "step": 103750 }, { "epoch": 22.38, "learning_rate": 2.8622035377759897e-05, "loss": 0.0001, "step": 103800 }, { "epoch": 22.39, "learning_rate": 2.861967163916559e-05, "loss": 0.0001, "step": 103850 }, { "epoch": 22.4, "learning_rate": 2.8617305972724022e-05, "loss": 0.0001, "step": 103900 }, { "epoch": 22.41, "learning_rate": 2.8614938378770057e-05, "loss": 0.0001, "step": 103950 }, { "epoch": 22.42, "learning_rate": 2.861256885763882e-05, "loss": 0.0001, "step": 104000 }, { "epoch": 22.43, "learning_rate": 2.8610197409665714e-05, "loss": 0.0001, "step": 104050 }, { "epoch": 22.44, "learning_rate": 2.8607824035186412e-05, "loss": 0.0001, "step": 104100 }, { "epoch": 22.45, "learning_rate": 2.860544873453687e-05, "loss": 0.0001, "step": 104150 }, { "epoch": 22.46, "learning_rate": 2.86030715080533e-05, "loss": 0.0001, "step": 104200 }, { "epoch": 22.47, "learning_rate": 2.8600692356072197e-05, "loss": 0.0001, "step": 104250 }, { "epoch": 22.48, "learning_rate": 2.8598311278930332e-05, "loss": 0.0001, "step": 104300 }, { "epoch": 22.49, "learning_rate": 2.859592827696474e-05, "loss": 0.0001, "step": 104350 }, { "epoch": 22.5, "learning_rate": 2.859354335051273e-05, "loss": 0.0, "step": 104400 }, { "epoch": 22.52, "learning_rate": 2.859115649991189e-05, "loss": 0.0, "step": 104450 }, { "epoch": 22.53, "learning_rate": 2.8588767725500077e-05, "loss": 0.0001, "step": 104500 }, { "epoch": 22.54, "learning_rate": 2.8586377027615407e-05, "loss": 0.0001, "step": 104550 }, { "epoch": 22.55, "learning_rate": 2.8583984406596295e-05, "loss": 0.0001, "step": 104600 }, { "epoch": 22.56, "learning_rate": 2.8581589862781405e-05, "loss": 0.0, "step": 104650 }, { "epoch": 22.57, "learning_rate": 2.8579193396509685e-05, "loss": 0.0, "step": 104700 }, { "epoch": 22.58, "learning_rate": 2.8576795008120346e-05, "loss": 0.0001, "step": 104750 }, { "epoch": 22.59, "learning_rate": 2.8574394697952884e-05, "loss": 0.0, "step": 104800 }, { "epoch": 22.6, "learning_rate": 2.8571992466347055e-05, "loss": 0.0001, "step": 104850 }, { "epoch": 22.61, "learning_rate": 2.85695883136429e-05, "loss": 0.0001, "step": 104900 }, { "epoch": 22.62, "learning_rate": 2.8567182240180716e-05, "loss": 0.0001, "step": 104950 }, { "epoch": 22.63, "learning_rate": 2.8564774246301077e-05, "loss": 0.0001, "step": 105000 }, { "epoch": 22.64, "learning_rate": 2.856236433234484e-05, "loss": 0.0001, "step": 105050 }, { "epoch": 22.66, "learning_rate": 2.855995249865312e-05, "loss": 0.0, "step": 105100 }, { "epoch": 22.67, "learning_rate": 2.855753874556731e-05, "loss": 0.0, "step": 105150 }, { "epoch": 22.68, "learning_rate": 2.8555123073429076e-05, "loss": 0.0001, "step": 105200 }, { "epoch": 22.69, "learning_rate": 2.8552705482580348e-05, "loss": 0.0001, "step": 105250 }, { "epoch": 22.7, "learning_rate": 2.8550285973363336e-05, "loss": 0.0001, "step": 105300 }, { "epoch": 22.71, "learning_rate": 2.854786454612052e-05, "loss": 0.0001, "step": 105350 }, { "epoch": 22.72, "learning_rate": 2.854544120119465e-05, "loss": 0.0001, "step": 105400 }, { "epoch": 22.73, "learning_rate": 2.8543015938928746e-05, "loss": 0.0001, "step": 105450 }, { "epoch": 22.74, "learning_rate": 2.8540588759666096e-05, "loss": 0.0001, "step": 105500 }, { "epoch": 22.75, "learning_rate": 2.8538159663750273e-05, "loss": 0.0001, "step": 105550 }, { "epoch": 22.76, "learning_rate": 2.8535728651525107e-05, "loss": 0.0001, "step": 105600 }, { "epoch": 22.77, "learning_rate": 2.853329572333471e-05, "loss": 0.0001, "step": 105650 }, { "epoch": 22.79, "learning_rate": 2.853086087952345e-05, "loss": 0.0001, "step": 105700 }, { "epoch": 22.8, "learning_rate": 2.8528424120435985e-05, "loss": 0.0, "step": 105750 }, { "epoch": 22.81, "learning_rate": 2.8525985446417232e-05, "loss": 0.0, "step": 105800 }, { "epoch": 22.82, "learning_rate": 2.8523544857812383e-05, "loss": 0.0, "step": 105850 }, { "epoch": 22.83, "learning_rate": 2.8521102354966896e-05, "loss": 0.0, "step": 105900 }, { "epoch": 22.84, "learning_rate": 2.851865793822651e-05, "loss": 0.0, "step": 105950 }, { "epoch": 22.85, "learning_rate": 2.8516211607937227e-05, "loss": 0.0, "step": 106000 }, { "epoch": 22.86, "learning_rate": 2.8513763364445324e-05, "loss": 0.0001, "step": 106050 }, { "epoch": 22.87, "learning_rate": 2.851131320809734e-05, "loss": 0.0001, "step": 106100 }, { "epoch": 22.88, "learning_rate": 2.85088611392401e-05, "loss": 0.0, "step": 106150 }, { "epoch": 22.89, "learning_rate": 2.8506407158220684e-05, "loss": 0.0001, "step": 106200 }, { "epoch": 22.9, "learning_rate": 2.8503951265386457e-05, "loss": 0.0001, "step": 106250 }, { "epoch": 22.91, "learning_rate": 2.850149346108504e-05, "loss": 0.0001, "step": 106300 }, { "epoch": 22.93, "learning_rate": 2.849903374566434e-05, "loss": 0.0, "step": 106350 }, { "epoch": 22.94, "learning_rate": 2.849657211947252e-05, "loss": 0.0001, "step": 106400 }, { "epoch": 22.95, "learning_rate": 2.8494108582858025e-05, "loss": 0.0, "step": 106450 }, { "epoch": 22.96, "learning_rate": 2.849164313616956e-05, "loss": 0.0001, "step": 106500 }, { "epoch": 22.97, "learning_rate": 2.848917577975611e-05, "loss": 0.0001, "step": 106550 }, { "epoch": 22.98, "learning_rate": 2.8486706513966923e-05, "loss": 0.0, "step": 106600 }, { "epoch": 22.99, "learning_rate": 2.8484235339151524e-05, "loss": 0.0, "step": 106650 }, { "epoch": 23.0, "learning_rate": 2.8481762255659702e-05, "loss": 0.0, "step": 106700 }, { "epoch": 23.01, "learning_rate": 2.847928726384152e-05, "loss": 0.0001, "step": 106750 }, { "epoch": 23.02, "learning_rate": 2.8476810364047316e-05, "loss": 0.0001, "step": 106800 }, { "epoch": 23.03, "learning_rate": 2.8474331556627682e-05, "loss": 0.0001, "step": 106850 }, { "epoch": 23.04, "learning_rate": 2.847185084193349e-05, "loss": 0.0001, "step": 106900 }, { "epoch": 23.05, "learning_rate": 2.8469368220315887e-05, "loss": 0.0001, "step": 106950 }, { "epoch": 23.07, "learning_rate": 2.846688369212629e-05, "loss": 0.0001, "step": 107000 }, { "epoch": 23.08, "learning_rate": 2.8464397257716374e-05, "loss": 0.0, "step": 107050 }, { "epoch": 23.09, "learning_rate": 2.8461908917438083e-05, "loss": 0.0, "step": 107100 }, { "epoch": 23.1, "learning_rate": 2.8459418671643652e-05, "loss": 0.0001, "step": 107150 }, { "epoch": 23.11, "learning_rate": 2.8456926520685572e-05, "loss": 0.0001, "step": 107200 }, { "epoch": 23.12, "learning_rate": 2.845443246491659e-05, "loss": 0.0, "step": 107250 }, { "epoch": 23.13, "learning_rate": 2.8451936504689748e-05, "loss": 0.0, "step": 107300 }, { "epoch": 23.14, "learning_rate": 2.8449438640358348e-05, "loss": 0.0, "step": 107350 }, { "epoch": 23.15, "learning_rate": 2.844693887227595e-05, "loss": 0.0001, "step": 107400 }, { "epoch": 23.16, "learning_rate": 2.8444437200796397e-05, "loss": 0.0001, "step": 107450 }, { "epoch": 23.17, "learning_rate": 2.8441933626273803e-05, "loss": 0.0001, "step": 107500 }, { "epoch": 23.18, "learning_rate": 2.8439428149062536e-05, "loss": 0.0001, "step": 107550 }, { "epoch": 23.19, "learning_rate": 2.8436920769517252e-05, "loss": 0.0001, "step": 107600 }, { "epoch": 23.21, "learning_rate": 2.843441148799286e-05, "loss": 0.0001, "step": 107650 }, { "epoch": 23.22, "learning_rate": 2.8431900304844554e-05, "loss": 0.0001, "step": 107700 }, { "epoch": 23.23, "learning_rate": 2.8429387220427775e-05, "loss": 0.0001, "step": 107750 }, { "epoch": 23.24, "learning_rate": 2.8426872235098264e-05, "loss": 0.0001, "step": 107800 }, { "epoch": 23.25, "learning_rate": 2.8424355349212005e-05, "loss": 0.0, "step": 107850 }, { "epoch": 23.26, "learning_rate": 2.8421836563125262e-05, "loss": 0.0001, "step": 107900 }, { "epoch": 23.27, "learning_rate": 2.841931587719456e-05, "loss": 0.0001, "step": 107950 }, { "epoch": 23.28, "learning_rate": 2.8416793291776704e-05, "loss": 0.0001, "step": 108000 }, { "epoch": 23.29, "learning_rate": 2.8414268807228764e-05, "loss": 0.0001, "step": 108050 }, { "epoch": 23.3, "learning_rate": 2.8411742423908077e-05, "loss": 0.0001, "step": 108100 }, { "epoch": 23.31, "learning_rate": 2.8409214142172248e-05, "loss": 0.0001, "step": 108150 }, { "epoch": 23.32, "learning_rate": 2.840668396237915e-05, "loss": 0.0, "step": 108200 }, { "epoch": 23.33, "learning_rate": 2.8404151884886932e-05, "loss": 0.0, "step": 108250 }, { "epoch": 23.35, "learning_rate": 2.8401617910054003e-05, "loss": 0.0001, "step": 108300 }, { "epoch": 23.36, "learning_rate": 2.8399082038239046e-05, "loss": 0.0001, "step": 108350 }, { "epoch": 23.37, "learning_rate": 2.8396544269801006e-05, "loss": 0.0001, "step": 108400 }, { "epoch": 23.38, "learning_rate": 2.8394004605099108e-05, "loss": 0.0001, "step": 108450 }, { "epoch": 23.39, "learning_rate": 2.8391463044492828e-05, "loss": 0.0, "step": 108500 }, { "epoch": 23.4, "learning_rate": 2.8388919588341933e-05, "loss": 0.0001, "step": 108550 }, { "epoch": 23.41, "learning_rate": 2.8386374237006438e-05, "loss": 0.0001, "step": 108600 }, { "epoch": 23.42, "learning_rate": 2.8383826990846635e-05, "loss": 0.0001, "step": 108650 }, { "epoch": 23.43, "learning_rate": 2.8381277850223083e-05, "loss": 0.0001, "step": 108700 }, { "epoch": 23.44, "learning_rate": 2.8378726815496614e-05, "loss": 0.0, "step": 108750 }, { "epoch": 23.45, "learning_rate": 2.837617388702832e-05, "loss": 0.0001, "step": 108800 }, { "epoch": 23.46, "learning_rate": 2.8373619065179566e-05, "loss": 0.0001, "step": 108850 }, { "epoch": 23.47, "learning_rate": 2.837106235031198e-05, "loss": 0.0, "step": 108900 }, { "epoch": 23.49, "learning_rate": 2.836850374278747e-05, "loss": 0.0001, "step": 108950 }, { "epoch": 23.5, "learning_rate": 2.83659432429682e-05, "loss": 0.0, "step": 109000 }, { "epoch": 23.51, "learning_rate": 2.8363380851216598e-05, "loss": 0.0001, "step": 109050 }, { "epoch": 23.52, "learning_rate": 2.8360816567895383e-05, "loss": 0.0001, "step": 109100 }, { "epoch": 23.53, "learning_rate": 2.8358250393367507e-05, "loss": 0.0001, "step": 109150 }, { "epoch": 23.54, "learning_rate": 2.8355682327996227e-05, "loss": 0.0001, "step": 109200 }, { "epoch": 23.55, "learning_rate": 2.8353112372145034e-05, "loss": 0.0001, "step": 109250 }, { "epoch": 23.56, "learning_rate": 2.8350540526177712e-05, "loss": 0.0001, "step": 109300 }, { "epoch": 23.57, "learning_rate": 2.83479667904583e-05, "loss": 0.0001, "step": 109350 }, { "epoch": 23.58, "learning_rate": 2.8345391165351108e-05, "loss": 0.0001, "step": 109400 }, { "epoch": 23.59, "learning_rate": 2.8342813651220707e-05, "loss": 0.0001, "step": 109450 }, { "epoch": 23.6, "learning_rate": 2.8340234248431944e-05, "loss": 0.0001, "step": 109500 }, { "epoch": 23.62, "learning_rate": 2.8337652957349932e-05, "loss": 0.0, "step": 109550 }, { "epoch": 23.63, "learning_rate": 2.833506977834005e-05, "loss": 0.0, "step": 109600 }, { "epoch": 23.64, "learning_rate": 2.8332484711767945e-05, "loss": 0.0001, "step": 109650 }, { "epoch": 23.65, "learning_rate": 2.8329897757999525e-05, "loss": 0.0001, "step": 109700 }, { "epoch": 23.66, "learning_rate": 2.8327308917400973e-05, "loss": 0.0001, "step": 109750 }, { "epoch": 23.67, "learning_rate": 2.832471819033873e-05, "loss": 0.0, "step": 109800 }, { "epoch": 23.68, "learning_rate": 2.8322125577179522e-05, "loss": 0.0001, "step": 109850 }, { "epoch": 23.69, "learning_rate": 2.831953107829032e-05, "loss": 0.0001, "step": 109900 }, { "epoch": 23.7, "learning_rate": 2.8316934694038373e-05, "loss": 0.0001, "step": 109950 }, { "epoch": 23.71, "learning_rate": 2.8314336424791203e-05, "loss": 0.0001, "step": 110000 }, { "epoch": 23.72, "learning_rate": 2.8311736270916587e-05, "loss": 0.0, "step": 110050 }, { "epoch": 23.73, "learning_rate": 2.8309134232782573e-05, "loss": 0.0, "step": 110100 }, { "epoch": 23.74, "learning_rate": 2.8306530310757475e-05, "loss": 0.0001, "step": 110150 }, { "epoch": 23.76, "learning_rate": 2.8303924505209878e-05, "loss": 0.0, "step": 110200 }, { "epoch": 23.77, "learning_rate": 2.8301316816508628e-05, "loss": 0.0001, "step": 110250 }, { "epoch": 23.78, "learning_rate": 2.829870724502284e-05, "loss": 0.0, "step": 110300 }, { "epoch": 23.79, "learning_rate": 2.8296095791121903e-05, "loss": 0.0001, "step": 110350 }, { "epoch": 23.8, "learning_rate": 2.829348245517545e-05, "loss": 0.0001, "step": 110400 }, { "epoch": 23.81, "learning_rate": 2.8290867237553412e-05, "loss": 0.0001, "step": 110450 }, { "epoch": 23.82, "learning_rate": 2.828825013862596e-05, "loss": 0.0, "step": 110500 }, { "epoch": 23.83, "learning_rate": 2.8285631158763543e-05, "loss": 0.0001, "step": 110550 }, { "epoch": 23.84, "learning_rate": 2.828301029833687e-05, "loss": 0.0001, "step": 110600 }, { "epoch": 23.85, "learning_rate": 2.828038755771693e-05, "loss": 0.0001, "step": 110650 }, { "epoch": 23.86, "learning_rate": 2.827776293727496e-05, "loss": 0.0, "step": 110700 }, { "epoch": 23.87, "learning_rate": 2.8275136437382472e-05, "loss": 0.0001, "step": 110750 }, { "epoch": 23.88, "learning_rate": 2.8272508058411252e-05, "loss": 0.0001, "step": 110800 }, { "epoch": 23.9, "learning_rate": 2.8269877800733336e-05, "loss": 0.0001, "step": 110850 }, { "epoch": 23.91, "learning_rate": 2.8267245664721037e-05, "loss": 0.0001, "step": 110900 }, { "epoch": 23.92, "learning_rate": 2.826461165074693e-05, "loss": 0.0, "step": 110950 }, { "epoch": 23.93, "learning_rate": 2.826197575918385e-05, "loss": 0.0, "step": 111000 }, { "epoch": 23.94, "learning_rate": 2.8259337990404915e-05, "loss": 0.0, "step": 111050 }, { "epoch": 23.95, "learning_rate": 2.8256698344783495e-05, "loss": 0.0001, "step": 111100 }, { "epoch": 23.96, "learning_rate": 2.8254056822693222e-05, "loss": 0.0001, "step": 111150 }, { "epoch": 23.97, "learning_rate": 2.8251413424508008e-05, "loss": 0.0001, "step": 111200 }, { "epoch": 23.98, "learning_rate": 2.8248768150602016e-05, "loss": 0.0001, "step": 111250 }, { "epoch": 23.99, "learning_rate": 2.8246121001349687e-05, "loss": 0.0, "step": 111300 }, { "epoch": 24.0, "learning_rate": 2.8243471977125718e-05, "loss": 0.0, "step": 111350 }, { "epoch": 24.01, "learning_rate": 2.8240821078305075e-05, "loss": 0.0001, "step": 111400 }, { "epoch": 24.02, "learning_rate": 2.8238168305262993e-05, "loss": 0.0, "step": 111450 }, { "epoch": 24.04, "learning_rate": 2.8235513658374962e-05, "loss": 0.0001, "step": 111500 }, { "epoch": 24.05, "learning_rate": 2.823285713801675e-05, "loss": 0.0001, "step": 111550 }, { "epoch": 24.06, "learning_rate": 2.823019874456438e-05, "loss": 0.0001, "step": 111600 }, { "epoch": 24.07, "learning_rate": 2.8227538478394147e-05, "loss": 0.0, "step": 111650 }, { "epoch": 24.08, "learning_rate": 2.8224876339882608e-05, "loss": 0.0001, "step": 111700 }, { "epoch": 24.09, "learning_rate": 2.8222212329406585e-05, "loss": 0.0, "step": 111750 }, { "epoch": 24.1, "learning_rate": 2.8219546447343164e-05, "loss": 0.0, "step": 111800 }, { "epoch": 24.11, "learning_rate": 2.8216878694069697e-05, "loss": 0.0, "step": 111850 }, { "epoch": 24.12, "learning_rate": 2.8214209069963806e-05, "loss": 0.0001, "step": 111900 }, { "epoch": 24.13, "learning_rate": 2.821153757540337e-05, "loss": 0.0001, "step": 111950 }, { "epoch": 24.14, "learning_rate": 2.820886421076653e-05, "loss": 0.0, "step": 112000 }, { "epoch": 24.15, "learning_rate": 2.8206188976431706e-05, "loss": 0.0001, "step": 112050 }, { "epoch": 24.16, "learning_rate": 2.820351187277757e-05, "loss": 0.0001, "step": 112100 }, { "epoch": 24.18, "learning_rate": 2.8200832900183062e-05, "loss": 0.0, "step": 112150 }, { "epoch": 24.19, "learning_rate": 2.8198152059027387e-05, "loss": 0.0001, "step": 112200 }, { "epoch": 24.2, "learning_rate": 2.819546934969002e-05, "loss": 0.0, "step": 112250 }, { "epoch": 24.21, "learning_rate": 2.8192784772550686e-05, "loss": 0.0001, "step": 112300 }, { "epoch": 24.22, "learning_rate": 2.819009832798939e-05, "loss": 0.0001, "step": 112350 }, { "epoch": 24.23, "learning_rate": 2.8187410016386396e-05, "loss": 0.0001, "step": 112400 }, { "epoch": 24.24, "learning_rate": 2.8184719838122225e-05, "loss": 0.0001, "step": 112450 }, { "epoch": 24.25, "learning_rate": 2.8182027793577673e-05, "loss": 0.0, "step": 112500 }, { "epoch": 24.26, "learning_rate": 2.8179333883133796e-05, "loss": 0.0, "step": 112550 }, { "epoch": 24.27, "learning_rate": 2.817663810717191e-05, "loss": 0.0, "step": 112600 }, { "epoch": 24.28, "learning_rate": 2.81739404660736e-05, "loss": 0.0, "step": 112650 }, { "epoch": 24.29, "learning_rate": 2.817124096022071e-05, "loss": 0.0001, "step": 112700 }, { "epoch": 24.3, "learning_rate": 2.8168539589995365e-05, "loss": 0.0001, "step": 112750 }, { "epoch": 24.32, "learning_rate": 2.816583635577993e-05, "loss": 0.0, "step": 112800 }, { "epoch": 24.33, "learning_rate": 2.816313125795704e-05, "loss": 0.0, "step": 112850 }, { "epoch": 24.34, "learning_rate": 2.8160424296909603e-05, "loss": 0.0, "step": 112900 }, { "epoch": 24.35, "learning_rate": 2.8157715473020793e-05, "loss": 0.0, "step": 112950 }, { "epoch": 24.36, "learning_rate": 2.815500478667403e-05, "loss": 0.0, "step": 113000 }, { "epoch": 24.37, "learning_rate": 2.8152292238253014e-05, "loss": 0.0001, "step": 113050 }, { "epoch": 24.38, "learning_rate": 2.8149577828141703e-05, "loss": 0.0, "step": 113100 }, { "epoch": 24.39, "learning_rate": 2.8146861556724314e-05, "loss": 0.0001, "step": 113150 }, { "epoch": 24.4, "learning_rate": 2.8144143424385334e-05, "loss": 0.0, "step": 113200 }, { "epoch": 24.41, "learning_rate": 2.8141423431509514e-05, "loss": 0.0001, "step": 113250 }, { "epoch": 24.42, "learning_rate": 2.813870157848186e-05, "loss": 0.0001, "step": 113300 }, { "epoch": 24.43, "learning_rate": 2.8135977865687655e-05, "loss": 0.0001, "step": 113350 }, { "epoch": 24.44, "learning_rate": 2.813325229351243e-05, "loss": 0.0001, "step": 113400 }, { "epoch": 24.46, "learning_rate": 2.8130524862341987e-05, "loss": 0.0001, "step": 113450 }, { "epoch": 24.47, "learning_rate": 2.812779557256239e-05, "loss": 0.0001, "step": 113500 }, { "epoch": 24.48, "learning_rate": 2.8125064424559975e-05, "loss": 0.0, "step": 113550 }, { "epoch": 24.49, "learning_rate": 2.812233141872132e-05, "loss": 0.0001, "step": 113600 }, { "epoch": 24.5, "learning_rate": 2.8119596555433287e-05, "loss": 0.0001, "step": 113650 }, { "epoch": 24.51, "learning_rate": 2.811685983508299e-05, "loss": 0.0001, "step": 113700 }, { "epoch": 24.52, "learning_rate": 2.8114121258057806e-05, "loss": 0.0, "step": 113750 }, { "epoch": 24.53, "learning_rate": 2.811138082474538e-05, "loss": 0.0001, "step": 113800 }, { "epoch": 24.54, "learning_rate": 2.8108638535533617e-05, "loss": 0.0, "step": 113850 }, { "epoch": 24.55, "learning_rate": 2.8105894390810683e-05, "loss": 0.0001, "step": 113900 }, { "epoch": 24.56, "learning_rate": 2.8103148390965008e-05, "loss": 0.0001, "step": 113950 }, { "epoch": 24.57, "learning_rate": 2.810040053638529e-05, "loss": 0.0001, "step": 114000 }, { "epoch": 24.59, "learning_rate": 2.8097650827460476e-05, "loss": 0.0001, "step": 114050 }, { "epoch": 24.6, "learning_rate": 2.8094899264579786e-05, "loss": 0.0, "step": 114100 }, { "epoch": 24.61, "learning_rate": 2.8092145848132706e-05, "loss": 0.0001, "step": 114150 }, { "epoch": 24.62, "learning_rate": 2.8089390578508973e-05, "loss": 0.0, "step": 114200 }, { "epoch": 24.63, "learning_rate": 2.8086633456098595e-05, "loss": 0.0, "step": 114250 }, { "epoch": 24.64, "learning_rate": 2.8083874481291835e-05, "loss": 0.0, "step": 114300 }, { "epoch": 24.65, "learning_rate": 2.8081113654479228e-05, "loss": 0.0001, "step": 114350 }, { "epoch": 24.66, "learning_rate": 2.8078350976051558e-05, "loss": 0.0001, "step": 114400 }, { "epoch": 24.67, "learning_rate": 2.8075586446399886e-05, "loss": 0.0001, "step": 114450 }, { "epoch": 24.68, "learning_rate": 2.8072820065915526e-05, "loss": 0.0001, "step": 114500 }, { "epoch": 24.69, "learning_rate": 2.8070051834990053e-05, "loss": 0.0, "step": 114550 }, { "epoch": 24.7, "learning_rate": 2.8067281754015312e-05, "loss": 0.0001, "step": 114600 }, { "epoch": 24.71, "learning_rate": 2.8064509823383398e-05, "loss": 0.0001, "step": 114650 }, { "epoch": 24.73, "learning_rate": 2.8061736043486673e-05, "loss": 0.0001, "step": 114700 }, { "epoch": 24.74, "learning_rate": 2.8058960414717774e-05, "loss": 0.0, "step": 114750 }, { "epoch": 24.75, "learning_rate": 2.8056182937469576e-05, "loss": 0.0, "step": 114800 }, { "epoch": 24.76, "learning_rate": 2.805340361213523e-05, "loss": 0.0, "step": 114850 }, { "epoch": 24.77, "learning_rate": 2.8050622439108143e-05, "loss": 0.0001, "step": 114900 }, { "epoch": 24.78, "learning_rate": 2.8047839418782e-05, "loss": 0.0001, "step": 114950 }, { "epoch": 24.79, "learning_rate": 2.804505455155072e-05, "loss": 0.0001, "step": 115000 }, { "epoch": 24.8, "learning_rate": 2.80422678378085e-05, "loss": 0.0001, "step": 115050 }, { "epoch": 24.81, "learning_rate": 2.8039479277949797e-05, "loss": 0.0, "step": 115100 }, { "epoch": 24.82, "learning_rate": 2.803668887236933e-05, "loss": 0.0, "step": 115150 }, { "epoch": 24.83, "learning_rate": 2.803389662146208e-05, "loss": 0.0001, "step": 115200 }, { "epoch": 24.84, "learning_rate": 2.8031102525623282e-05, "loss": 0.0001, "step": 115250 }, { "epoch": 24.85, "learning_rate": 2.8028306585248437e-05, "loss": 0.0001, "step": 115300 }, { "epoch": 24.87, "learning_rate": 2.8025508800733303e-05, "loss": 0.0001, "step": 115350 }, { "epoch": 24.88, "learning_rate": 2.8022709172473912e-05, "loss": 0.0001, "step": 115400 }, { "epoch": 24.89, "learning_rate": 2.801990770086654e-05, "loss": 0.0001, "step": 115450 }, { "epoch": 24.9, "learning_rate": 2.801710438630774e-05, "loss": 0.0001, "step": 115500 }, { "epoch": 24.91, "learning_rate": 2.801429922919431e-05, "loss": 0.0, "step": 115550 }, { "epoch": 24.92, "learning_rate": 2.801149222992332e-05, "loss": 0.0001, "step": 115600 }, { "epoch": 24.93, "learning_rate": 2.8008683388892093e-05, "loss": 0.0001, "step": 115650 }, { "epoch": 24.94, "learning_rate": 2.800587270649822e-05, "loss": 0.0, "step": 115700 }, { "epoch": 24.95, "learning_rate": 2.800306018313955e-05, "loss": 0.0001, "step": 115750 }, { "epoch": 24.96, "learning_rate": 2.8000245819214193e-05, "loss": 0.0, "step": 115800 }, { "epoch": 24.97, "learning_rate": 2.7997429615120513e-05, "loss": 0.0, "step": 115850 }, { "epoch": 24.98, "learning_rate": 2.799461157125715e-05, "loss": 0.0, "step": 115900 }, { "epoch": 24.99, "learning_rate": 2.7991791688022987e-05, "loss": 0.0001, "step": 115950 }, { "epoch": 25.01, "learning_rate": 2.7988969965817176e-05, "loss": 0.0001, "step": 116000 }, { "epoch": 25.02, "learning_rate": 2.798614640503913e-05, "loss": 0.0001, "step": 116050 }, { "epoch": 25.03, "learning_rate": 2.798332100608852e-05, "loss": 0.0001, "step": 116100 }, { "epoch": 25.04, "learning_rate": 2.7980493769365276e-05, "loss": 0.0, "step": 116150 }, { "epoch": 25.05, "learning_rate": 2.797766469526959e-05, "loss": 0.0001, "step": 116200 }, { "epoch": 25.06, "learning_rate": 2.797483378420192e-05, "loss": 0.0, "step": 116250 }, { "epoch": 25.07, "learning_rate": 2.7972001036562965e-05, "loss": 0.0001, "step": 116300 }, { "epoch": 25.08, "learning_rate": 2.796916645275371e-05, "loss": 0.0, "step": 116350 }, { "epoch": 25.09, "learning_rate": 2.7966330033175376e-05, "loss": 0.0, "step": 116400 }, { "epoch": 25.1, "learning_rate": 2.7963491778229463e-05, "loss": 0.0001, "step": 116450 }, { "epoch": 25.11, "learning_rate": 2.7960651688317724e-05, "loss": 0.0001, "step": 116500 }, { "epoch": 25.12, "learning_rate": 2.7957809763842155e-05, "loss": 0.0001, "step": 116550 }, { "epoch": 25.13, "learning_rate": 2.795496600520504e-05, "loss": 0.0, "step": 116600 }, { "epoch": 25.15, "learning_rate": 2.795212041280891e-05, "loss": 0.0, "step": 116650 }, { "epoch": 25.16, "learning_rate": 2.7949272987056554e-05, "loss": 0.0, "step": 116700 }, { "epoch": 25.17, "learning_rate": 2.7946423728351014e-05, "loss": 0.0001, "step": 116750 }, { "epoch": 25.18, "learning_rate": 2.7943572637095606e-05, "loss": 0.0, "step": 116800 }, { "epoch": 25.19, "learning_rate": 2.7940719713693895e-05, "loss": 0.0, "step": 116850 }, { "epoch": 25.2, "learning_rate": 2.7937864958549714e-05, "loss": 0.0, "step": 116900 }, { "epoch": 25.21, "learning_rate": 2.793500837206714e-05, "loss": 0.0001, "step": 116950 }, { "epoch": 25.22, "learning_rate": 2.793214995465053e-05, "loss": 0.0001, "step": 117000 }, { "epoch": 25.23, "learning_rate": 2.792928970670448e-05, "loss": 0.0, "step": 117050 }, { "epoch": 25.24, "learning_rate": 2.792642762863386e-05, "loss": 0.0, "step": 117100 }, { "epoch": 25.25, "learning_rate": 2.79235637208438e-05, "loss": 0.0, "step": 117150 }, { "epoch": 25.26, "learning_rate": 2.792069798373967e-05, "loss": 0.0, "step": 117200 }, { "epoch": 25.27, "learning_rate": 2.7917830417727113e-05, "loss": 0.0001, "step": 117250 }, { "epoch": 25.29, "learning_rate": 2.7914961023212036e-05, "loss": 0.0001, "step": 117300 }, { "epoch": 25.3, "learning_rate": 2.7912089800600593e-05, "loss": 0.0, "step": 117350 }, { "epoch": 25.31, "learning_rate": 2.7909216750299207e-05, "loss": 0.0001, "step": 117400 }, { "epoch": 25.32, "learning_rate": 2.790634187271455e-05, "loss": 0.0, "step": 117450 }, { "epoch": 25.33, "learning_rate": 2.790346516825356e-05, "loss": 0.0001, "step": 117500 }, { "epoch": 25.34, "learning_rate": 2.790058663732343e-05, "loss": 0.0001, "step": 117550 }, { "epoch": 25.35, "learning_rate": 2.789770628033161e-05, "loss": 0.0001, "step": 117600 }, { "epoch": 25.36, "learning_rate": 2.7894824097685818e-05, "loss": 0.0001, "step": 117650 }, { "epoch": 25.37, "learning_rate": 2.789194008979401e-05, "loss": 0.0001, "step": 117700 }, { "epoch": 25.38, "learning_rate": 2.7889054257064428e-05, "loss": 0.0001, "step": 117750 }, { "epoch": 25.39, "learning_rate": 2.7886166599905553e-05, "loss": 0.0001, "step": 117800 }, { "epoch": 25.4, "learning_rate": 2.7883277118726123e-05, "loss": 0.0, "step": 117850 }, { "epoch": 25.41, "learning_rate": 2.7880385813935152e-05, "loss": 0.0, "step": 117900 }, { "epoch": 25.43, "learning_rate": 2.787749268594189e-05, "loss": 0.0, "step": 117950 }, { "epoch": 25.44, "learning_rate": 2.787459773515586e-05, "loss": 0.0001, "step": 118000 }, { "epoch": 25.45, "learning_rate": 2.7871700961986837e-05, "loss": 0.0001, "step": 118050 }, { "epoch": 25.46, "learning_rate": 2.786880236684486e-05, "loss": 0.0001, "step": 118100 }, { "epoch": 25.47, "learning_rate": 2.7865901950140214e-05, "loss": 0.0001, "step": 118150 }, { "epoch": 25.48, "learning_rate": 2.7862999712283458e-05, "loss": 0.0, "step": 118200 }, { "epoch": 25.49, "learning_rate": 2.7860095653685392e-05, "loss": 0.0001, "step": 118250 }, { "epoch": 25.5, "learning_rate": 2.7857189774757088e-05, "loss": 0.0, "step": 118300 }, { "epoch": 25.51, "learning_rate": 2.7854282075909866e-05, "loss": 0.0001, "step": 118350 }, { "epoch": 25.52, "learning_rate": 2.785137255755531e-05, "loss": 0.0001, "step": 118400 }, { "epoch": 25.53, "learning_rate": 2.7848461220105255e-05, "loss": 0.0001, "step": 118450 }, { "epoch": 25.54, "learning_rate": 2.7845548063971802e-05, "loss": 0.0001, "step": 118500 }, { "epoch": 25.56, "learning_rate": 2.7842633089567297e-05, "loss": 0.0001, "step": 118550 }, { "epoch": 25.57, "learning_rate": 2.7839716297304357e-05, "loss": 0.0, "step": 118600 }, { "epoch": 25.58, "learning_rate": 2.783679768759585e-05, "loss": 0.0001, "step": 118650 }, { "epoch": 25.59, "learning_rate": 2.7833877260854897e-05, "loss": 0.0, "step": 118700 }, { "epoch": 25.6, "learning_rate": 2.7830955017494885e-05, "loss": 0.0001, "step": 118750 }, { "epoch": 25.61, "learning_rate": 2.7828030957929452e-05, "loss": 0.0, "step": 118800 }, { "epoch": 25.62, "learning_rate": 2.78251050825725e-05, "loss": 0.0, "step": 118850 }, { "epoch": 25.63, "learning_rate": 2.7822177391838176e-05, "loss": 0.0, "step": 118900 }, { "epoch": 25.64, "learning_rate": 2.7819247886140894e-05, "loss": 0.0001, "step": 118950 }, { "epoch": 25.65, "learning_rate": 2.781631656589532e-05, "loss": 0.0, "step": 119000 }, { "epoch": 25.66, "learning_rate": 2.781338343151638e-05, "loss": 0.0, "step": 119050 }, { "epoch": 25.67, "learning_rate": 2.781044848341926e-05, "loss": 0.0001, "step": 119100 }, { "epoch": 25.68, "learning_rate": 2.7807511722019386e-05, "loss": 0.0001, "step": 119150 }, { "epoch": 25.7, "learning_rate": 2.7804573147732464e-05, "loss": 0.0001, "step": 119200 }, { "epoch": 25.71, "learning_rate": 2.780163276097444e-05, "loss": 0.0, "step": 119250 }, { "epoch": 25.72, "learning_rate": 2.7798690562161528e-05, "loss": 0.0001, "step": 119300 }, { "epoch": 25.73, "learning_rate": 2.7795746551710185e-05, "loss": 0.0, "step": 119350 }, { "epoch": 25.74, "learning_rate": 2.7792800730037137e-05, "loss": 0.0001, "step": 119400 }, { "epoch": 25.75, "learning_rate": 2.778985309755936e-05, "loss": 0.0001, "step": 119450 }, { "epoch": 25.76, "learning_rate": 2.7786903654694084e-05, "loss": 0.0001, "step": 119500 }, { "epoch": 25.77, "learning_rate": 2.7783952401858806e-05, "loss": 0.0001, "step": 119550 }, { "epoch": 25.78, "learning_rate": 2.7780999339471267e-05, "loss": 0.0001, "step": 119600 }, { "epoch": 25.79, "learning_rate": 2.7778044467949472e-05, "loss": 0.0, "step": 119650 }, { "epoch": 25.8, "learning_rate": 2.7775087787711675e-05, "loss": 0.0, "step": 119700 }, { "epoch": 25.81, "learning_rate": 2.7772129299176395e-05, "loss": 0.0, "step": 119750 }, { "epoch": 25.82, "learning_rate": 2.77691690027624e-05, "loss": 0.0, "step": 119800 }, { "epoch": 25.84, "learning_rate": 2.776620689888872e-05, "loss": 0.0, "step": 119850 }, { "epoch": 25.85, "learning_rate": 2.776324298797463e-05, "loss": 0.0, "step": 119900 }, { "epoch": 25.86, "learning_rate": 2.776027727043967e-05, "loss": 0.0001, "step": 119950 }, { "epoch": 25.87, "learning_rate": 2.775730974670364e-05, "loss": 0.0001, "step": 120000 }, { "epoch": 25.88, "learning_rate": 2.7754340417186582e-05, "loss": 0.0, "step": 120050 }, { "epoch": 25.89, "learning_rate": 2.7751369282308805e-05, "loss": 0.0, "step": 120100 }, { "epoch": 25.9, "learning_rate": 2.7748396342490863e-05, "loss": 0.0, "step": 120150 }, { "epoch": 25.91, "learning_rate": 2.774542159815358e-05, "loss": 0.0001, "step": 120200 }, { "epoch": 25.92, "learning_rate": 2.7742445049718022e-05, "loss": 0.0001, "step": 120250 }, { "epoch": 25.93, "learning_rate": 2.7739466697605522e-05, "loss": 0.0001, "step": 120300 }, { "epoch": 25.94, "learning_rate": 2.7736486542237655e-05, "loss": 0.0001, "step": 120350 }, { "epoch": 25.95, "learning_rate": 2.7733504584036257e-05, "loss": 0.0001, "step": 120400 }, { "epoch": 25.96, "learning_rate": 2.7730520823423428e-05, "loss": 0.0, "step": 120450 }, { "epoch": 25.98, "learning_rate": 2.772753526082151e-05, "loss": 0.0, "step": 120500 }, { "epoch": 25.99, "learning_rate": 2.772454789665311e-05, "loss": 0.0, "step": 120550 }, { "epoch": 26.0, "learning_rate": 2.772155873134108e-05, "loss": 0.0001, "step": 120600 }, { "epoch": 26.01, "learning_rate": 2.7718567765308534e-05, "loss": 0.0001, "step": 120650 }, { "epoch": 26.02, "learning_rate": 2.7715574998978842e-05, "loss": 0.0, "step": 120700 }, { "epoch": 26.03, "learning_rate": 2.7712580432775626e-05, "loss": 0.0, "step": 120750 }, { "epoch": 26.04, "learning_rate": 2.770958406712276e-05, "loss": 0.0001, "step": 120800 }, { "epoch": 26.05, "learning_rate": 2.770658590244438e-05, "loss": 0.0001, "step": 120850 }, { "epoch": 26.06, "learning_rate": 2.7703585939164873e-05, "loss": 0.0, "step": 120900 }, { "epoch": 26.07, "learning_rate": 2.7700584177708874e-05, "loss": 0.0001, "step": 120950 }, { "epoch": 26.08, "learning_rate": 2.769758061850128e-05, "loss": 0.0, "step": 121000 }, { "epoch": 26.09, "learning_rate": 2.7694575261967242e-05, "loss": 0.0, "step": 121050 }, { "epoch": 26.1, "learning_rate": 2.769156810853217e-05, "loss": 0.0, "step": 121100 }, { "epoch": 26.12, "learning_rate": 2.7688559158621718e-05, "loss": 0.0001, "step": 121150 }, { "epoch": 26.13, "learning_rate": 2.76855484126618e-05, "loss": 0.0001, "step": 121200 }, { "epoch": 26.14, "learning_rate": 2.7682535871078573e-05, "loss": 0.0, "step": 121250 }, { "epoch": 26.15, "learning_rate": 2.767952153429848e-05, "loss": 0.0001, "step": 121300 }, { "epoch": 26.16, "learning_rate": 2.7676505402748177e-05, "loss": 0.0, "step": 121350 }, { "epoch": 26.17, "learning_rate": 2.76734874768546e-05, "loss": 0.0001, "step": 121400 }, { "epoch": 26.18, "learning_rate": 2.7670467757044935e-05, "loss": 0.0001, "step": 121450 }, { "epoch": 26.19, "learning_rate": 2.766744624374662e-05, "loss": 0.0001, "step": 121500 }, { "epoch": 26.2, "learning_rate": 2.7664422937387343e-05, "loss": 0.0, "step": 121550 }, { "epoch": 26.21, "learning_rate": 2.766139783839505e-05, "loss": 0.0, "step": 121600 }, { "epoch": 26.22, "learning_rate": 2.7658370947197938e-05, "loss": 0.0, "step": 121650 }, { "epoch": 26.23, "learning_rate": 2.7655342264224468e-05, "loss": 0.0001, "step": 121700 }, { "epoch": 26.24, "learning_rate": 2.7652311789903336e-05, "loss": 0.0, "step": 121750 }, { "epoch": 26.26, "learning_rate": 2.764927952466351e-05, "loss": 0.0, "step": 121800 }, { "epoch": 26.27, "learning_rate": 2.7646245468934197e-05, "loss": 0.0001, "step": 121850 }, { "epoch": 26.28, "learning_rate": 2.7643209623144864e-05, "loss": 0.0001, "step": 121900 }, { "epoch": 26.29, "learning_rate": 2.7640171987725236e-05, "loss": 0.0, "step": 121950 }, { "epoch": 26.3, "learning_rate": 2.7637132563105285e-05, "loss": 0.0, "step": 122000 }, { "epoch": 26.31, "learning_rate": 2.7634091349715234e-05, "loss": 0.0, "step": 122050 }, { "epoch": 26.32, "learning_rate": 2.7631048347985565e-05, "loss": 0.0001, "step": 122100 }, { "epoch": 26.33, "learning_rate": 2.7628003558347014e-05, "loss": 0.0, "step": 122150 }, { "epoch": 26.34, "learning_rate": 2.7624956981230563e-05, "loss": 0.0, "step": 122200 }, { "epoch": 26.35, "learning_rate": 2.762190861706745e-05, "loss": 0.0001, "step": 122250 }, { "epoch": 26.36, "learning_rate": 2.7618858466289174e-05, "loss": 0.0, "step": 122300 }, { "epoch": 26.37, "learning_rate": 2.7615806529327474e-05, "loss": 0.0001, "step": 122350 }, { "epoch": 26.38, "learning_rate": 2.761275280661435e-05, "loss": 0.0, "step": 122400 }, { "epoch": 26.4, "learning_rate": 2.760969729858205e-05, "loss": 0.0, "step": 122450 }, { "epoch": 26.41, "learning_rate": 2.7606640005663076e-05, "loss": 0.0, "step": 122500 }, { "epoch": 26.42, "learning_rate": 2.7603580928290197e-05, "loss": 0.0, "step": 122550 }, { "epoch": 26.43, "learning_rate": 2.7600520066896405e-05, "loss": 0.0001, "step": 122600 }, { "epoch": 26.44, "learning_rate": 2.7597457421914972e-05, "loss": 0.0, "step": 122650 }, { "epoch": 26.45, "learning_rate": 2.7594392993779406e-05, "loss": 0.0, "step": 122700 }, { "epoch": 26.46, "learning_rate": 2.7591326782923472e-05, "loss": 0.0, "step": 122750 }, { "epoch": 26.47, "learning_rate": 2.758825878978119e-05, "loss": 0.0, "step": 122800 }, { "epoch": 26.48, "learning_rate": 2.7585189014786835e-05, "loss": 0.0001, "step": 122850 }, { "epoch": 26.49, "learning_rate": 2.7582117458374923e-05, "loss": 0.0001, "step": 122900 }, { "epoch": 26.5, "learning_rate": 2.7579044120980234e-05, "loss": 0.0, "step": 122950 }, { "epoch": 26.51, "learning_rate": 2.7575969003037792e-05, "loss": 0.0, "step": 123000 }, { "epoch": 26.53, "learning_rate": 2.7572892104982873e-05, "loss": 0.0001, "step": 123050 }, { "epoch": 26.54, "learning_rate": 2.756981342725102e-05, "loss": 0.0, "step": 123100 }, { "epoch": 26.55, "learning_rate": 2.7566732970278002e-05, "loss": 0.0, "step": 123150 }, { "epoch": 26.56, "learning_rate": 2.756365073449986e-05, "loss": 0.0, "step": 123200 }, { "epoch": 26.57, "learning_rate": 2.7560566720352884e-05, "loss": 0.0, "step": 123250 }, { "epoch": 26.58, "learning_rate": 2.7557480928273602e-05, "loss": 0.0, "step": 123300 }, { "epoch": 26.59, "learning_rate": 2.7554393358698815e-05, "loss": 0.0001, "step": 123350 }, { "epoch": 26.6, "learning_rate": 2.755130401206556e-05, "loss": 0.0001, "step": 123400 }, { "epoch": 26.61, "learning_rate": 2.7548212888811124e-05, "loss": 0.0, "step": 123450 }, { "epoch": 26.62, "learning_rate": 2.7545119989373063e-05, "loss": 0.0001, "step": 123500 }, { "epoch": 26.63, "learning_rate": 2.754202531418917e-05, "loss": 0.0, "step": 123550 }, { "epoch": 26.64, "learning_rate": 2.7538928863697482e-05, "loss": 0.0, "step": 123600 }, { "epoch": 26.65, "learning_rate": 2.753583063833631e-05, "loss": 0.0001, "step": 123650 }, { "epoch": 26.67, "learning_rate": 2.7532730638544207e-05, "loss": 0.0, "step": 123700 }, { "epoch": 26.68, "learning_rate": 2.752962886475996e-05, "loss": 0.0001, "step": 123750 }, { "epoch": 26.69, "learning_rate": 2.7526525317422622e-05, "loss": 0.0001, "step": 123800 }, { "epoch": 26.7, "learning_rate": 2.7523419996971508e-05, "loss": 0.0, "step": 123850 }, { "epoch": 26.71, "learning_rate": 2.752031290384617e-05, "loss": 0.0001, "step": 123900 }, { "epoch": 26.72, "learning_rate": 2.75172040384864e-05, "loss": 0.0001, "step": 123950 }, { "epoch": 26.73, "learning_rate": 2.751409340133227e-05, "loss": 0.0001, "step": 124000 }, { "epoch": 26.74, "learning_rate": 2.7510980992824075e-05, "loss": 0.0, "step": 124050 }, { "epoch": 26.75, "learning_rate": 2.7507866813402377e-05, "loss": 0.0, "step": 124100 }, { "epoch": 26.76, "learning_rate": 2.7504750863507993e-05, "loss": 0.0, "step": 124150 }, { "epoch": 26.77, "learning_rate": 2.7501633143581966e-05, "loss": 0.0001, "step": 124200 }, { "epoch": 26.78, "learning_rate": 2.749851365406561e-05, "loss": 0.0, "step": 124250 }, { "epoch": 26.79, "learning_rate": 2.7495392395400495e-05, "loss": 0.0, "step": 124300 }, { "epoch": 26.81, "learning_rate": 2.749226936802842e-05, "loss": 0.0001, "step": 124350 }, { "epoch": 26.82, "learning_rate": 2.748914457239145e-05, "loss": 0.0, "step": 124400 }, { "epoch": 26.83, "learning_rate": 2.748601800893189e-05, "loss": 0.0, "step": 124450 }, { "epoch": 26.84, "learning_rate": 2.748288967809231e-05, "loss": 0.0001, "step": 124500 }, { "epoch": 26.85, "learning_rate": 2.7479759580315518e-05, "loss": 0.0001, "step": 124550 }, { "epoch": 26.86, "learning_rate": 2.7476627716044575e-05, "loss": 0.0001, "step": 124600 }, { "epoch": 26.87, "learning_rate": 2.7473494085722788e-05, "loss": 0.0001, "step": 124650 }, { "epoch": 26.88, "learning_rate": 2.7470358689793727e-05, "loss": 0.0001, "step": 124700 }, { "epoch": 26.89, "learning_rate": 2.7467221528701195e-05, "loss": 0.0, "step": 124750 }, { "epoch": 26.9, "learning_rate": 2.7464082602889264e-05, "loss": 0.0, "step": 124800 }, { "epoch": 26.91, "learning_rate": 2.7460941912802232e-05, "loss": 0.0, "step": 124850 }, { "epoch": 26.92, "learning_rate": 2.7457799458884667e-05, "loss": 0.0, "step": 124900 }, { "epoch": 26.93, "learning_rate": 2.745465524158138e-05, "loss": 0.0001, "step": 124950 }, { "epoch": 26.95, "learning_rate": 2.7451509261337424e-05, "loss": 0.0001, "step": 125000 }, { "epoch": 26.96, "learning_rate": 2.744836151859812e-05, "loss": 0.0, "step": 125050 }, { "epoch": 26.97, "learning_rate": 2.7445212013809014e-05, "loss": 0.0001, "step": 125100 }, { "epoch": 26.98, "learning_rate": 2.7442060747415926e-05, "loss": 0.0001, "step": 125150 }, { "epoch": 26.99, "learning_rate": 2.7438907719864908e-05, "loss": 0.0001, "step": 125200 }, { "epoch": 27.0, "learning_rate": 2.7435752931602263e-05, "loss": 0.0, "step": 125250 }, { "epoch": 27.01, "learning_rate": 2.7432596383074555e-05, "loss": 0.0001, "step": 125300 }, { "epoch": 27.02, "learning_rate": 2.742943807472859e-05, "loss": 0.0, "step": 125350 }, { "epoch": 27.03, "learning_rate": 2.7426278007011415e-05, "loss": 0.0, "step": 125400 }, { "epoch": 27.04, "learning_rate": 2.7423116180370342e-05, "loss": 0.0, "step": 125450 }, { "epoch": 27.05, "learning_rate": 2.7419952595252912e-05, "loss": 0.0, "step": 125500 }, { "epoch": 27.06, "learning_rate": 2.7416787252106935e-05, "loss": 0.0, "step": 125550 }, { "epoch": 27.07, "learning_rate": 2.7413620151380467e-05, "loss": 0.0, "step": 125600 }, { "epoch": 27.09, "learning_rate": 2.7410451293521796e-05, "loss": 0.0, "step": 125650 }, { "epoch": 27.1, "learning_rate": 2.7407280678979472e-05, "loss": 0.0001, "step": 125700 }, { "epoch": 27.11, "learning_rate": 2.7404108308202295e-05, "loss": 0.0, "step": 125750 }, { "epoch": 27.12, "learning_rate": 2.7400934181639304e-05, "loss": 0.0, "step": 125800 }, { "epoch": 27.13, "learning_rate": 2.7397758299739805e-05, "loss": 0.0, "step": 125850 }, { "epoch": 27.14, "learning_rate": 2.7394580662953327e-05, "loss": 0.0, "step": 125900 }, { "epoch": 27.15, "learning_rate": 2.7391401271729666e-05, "loss": 0.0001, "step": 125950 }, { "epoch": 27.16, "learning_rate": 2.7388220126518863e-05, "loss": 0.0001, "step": 126000 }, { "epoch": 27.17, "learning_rate": 2.7385037227771202e-05, "loss": 0.0, "step": 126050 }, { "epoch": 27.18, "learning_rate": 2.7381852575937215e-05, "loss": 0.0, "step": 126100 }, { "epoch": 27.19, "learning_rate": 2.737866617146769e-05, "loss": 0.0, "step": 126150 }, { "epoch": 27.2, "learning_rate": 2.7375478014813663e-05, "loss": 0.0001, "step": 126200 }, { "epoch": 27.21, "learning_rate": 2.73722881064264e-05, "loss": 0.0, "step": 126250 }, { "epoch": 27.23, "learning_rate": 2.7369096446757443e-05, "loss": 0.0001, "step": 126300 }, { "epoch": 27.24, "learning_rate": 2.7365903036258556e-05, "loss": 0.0001, "step": 126350 }, { "epoch": 27.25, "learning_rate": 2.736270787538177e-05, "loss": 0.0, "step": 126400 }, { "epoch": 27.26, "learning_rate": 2.7359510964579354e-05, "loss": 0.0, "step": 126450 }, { "epoch": 27.27, "learning_rate": 2.7356312304303824e-05, "loss": 0.0, "step": 126500 }, { "epoch": 27.28, "learning_rate": 2.7353111895007946e-05, "loss": 0.0, "step": 126550 }, { "epoch": 27.29, "learning_rate": 2.734990973714474e-05, "loss": 0.0, "step": 126600 }, { "epoch": 27.3, "learning_rate": 2.7346705831167462e-05, "loss": 0.0, "step": 126650 }, { "epoch": 27.31, "learning_rate": 2.7343500177529622e-05, "loss": 0.0, "step": 126700 }, { "epoch": 27.32, "learning_rate": 2.7340292776684975e-05, "loss": 0.0, "step": 126750 }, { "epoch": 27.33, "learning_rate": 2.733708362908753e-05, "loss": 0.0001, "step": 126800 }, { "epoch": 27.34, "learning_rate": 2.733387273519153e-05, "loss": 0.0001, "step": 126850 }, { "epoch": 27.36, "learning_rate": 2.7330660095451482e-05, "loss": 0.0001, "step": 126900 }, { "epoch": 27.37, "learning_rate": 2.732744571032212e-05, "loss": 0.0, "step": 126950 }, { "epoch": 27.38, "learning_rate": 2.7324229580258444e-05, "loss": 0.0, "step": 127000 }, { "epoch": 27.39, "learning_rate": 2.732101170571569e-05, "loss": 0.0001, "step": 127050 }, { "epoch": 27.4, "learning_rate": 2.7317792087149346e-05, "loss": 0.0001, "step": 127100 }, { "epoch": 27.41, "learning_rate": 2.7314570725015143e-05, "loss": 0.0001, "step": 127150 }, { "epoch": 27.42, "learning_rate": 2.7311347619769063e-05, "loss": 0.0, "step": 127200 }, { "epoch": 27.43, "learning_rate": 2.730812277186733e-05, "loss": 0.0001, "step": 127250 }, { "epoch": 27.44, "learning_rate": 2.7304896181766422e-05, "loss": 0.0, "step": 127300 }, { "epoch": 27.45, "learning_rate": 2.7301667849923054e-05, "loss": 0.0, "step": 127350 }, { "epoch": 27.46, "learning_rate": 2.729843777679419e-05, "loss": 0.0001, "step": 127400 }, { "epoch": 27.47, "learning_rate": 2.7295205962837052e-05, "loss": 0.0, "step": 127450 }, { "epoch": 27.48, "learning_rate": 2.729197240850909e-05, "loss": 0.0, "step": 127500 }, { "epoch": 27.5, "learning_rate": 2.728873711426801e-05, "loss": 0.0001, "step": 127550 }, { "epoch": 27.51, "learning_rate": 2.7285500080571768e-05, "loss": 0.0001, "step": 127600 }, { "epoch": 27.52, "learning_rate": 2.728226130787856e-05, "loss": 0.0001, "step": 127650 }, { "epoch": 27.53, "learning_rate": 2.7279020796646836e-05, "loss": 0.0, "step": 127700 }, { "epoch": 27.54, "learning_rate": 2.727577854733528e-05, "loss": 0.0, "step": 127750 }, { "epoch": 27.55, "learning_rate": 2.7272534560402825e-05, "loss": 0.0, "step": 127800 }, { "epoch": 27.56, "learning_rate": 2.726928883630866e-05, "loss": 0.0, "step": 127850 }, { "epoch": 27.57, "learning_rate": 2.726604137551221e-05, "loss": 0.0, "step": 127900 }, { "epoch": 27.58, "learning_rate": 2.7262792178473153e-05, "loss": 0.0, "step": 127950 }, { "epoch": 27.59, "learning_rate": 2.72595412456514e-05, "loss": 0.0, "step": 128000 }, { "epoch": 27.6, "learning_rate": 2.7256288577507127e-05, "loss": 0.0, "step": 128050 }, { "epoch": 27.61, "learning_rate": 2.7253034174500735e-05, "loss": 0.0001, "step": 128100 }, { "epoch": 27.62, "learning_rate": 2.724977803709289e-05, "loss": 0.0, "step": 128150 }, { "epoch": 27.64, "learning_rate": 2.724652016574449e-05, "loss": 0.0, "step": 128200 }, { "epoch": 27.65, "learning_rate": 2.7243260560916677e-05, "loss": 0.0001, "step": 128250 }, { "epoch": 27.66, "learning_rate": 2.723999922307085e-05, "loss": 0.0001, "step": 128300 }, { "epoch": 27.67, "learning_rate": 2.7236736152668647e-05, "loss": 0.0001, "step": 128350 }, { "epoch": 27.68, "learning_rate": 2.7233471350171952e-05, "loss": 0.0, "step": 128400 }, { "epoch": 27.69, "learning_rate": 2.7230204816042895e-05, "loss": 0.0, "step": 128450 }, { "epoch": 27.7, "learning_rate": 2.7226936550743845e-05, "loss": 0.0, "step": 128500 }, { "epoch": 27.71, "learning_rate": 2.7223666554737424e-05, "loss": 0.0001, "step": 128550 }, { "epoch": 27.72, "learning_rate": 2.722039482848649e-05, "loss": 0.0, "step": 128600 }, { "epoch": 27.73, "learning_rate": 2.7217121372454165e-05, "loss": 0.0001, "step": 128650 }, { "epoch": 27.74, "learning_rate": 2.7213846187103787e-05, "loss": 0.0, "step": 128700 }, { "epoch": 27.75, "learning_rate": 2.7210569272898967e-05, "loss": 0.0, "step": 128750 }, { "epoch": 27.76, "learning_rate": 2.7207290630303537e-05, "loss": 0.0, "step": 128800 }, { "epoch": 27.78, "learning_rate": 2.7204010259781593e-05, "loss": 0.0001, "step": 128850 }, { "epoch": 27.79, "learning_rate": 2.7200728161797467e-05, "loss": 0.0, "step": 128900 }, { "epoch": 27.8, "learning_rate": 2.719744433681573e-05, "loss": 0.0001, "step": 128950 }, { "epoch": 27.81, "learning_rate": 2.719415878530121e-05, "loss": 0.0001, "step": 129000 }, { "epoch": 27.82, "learning_rate": 2.7190871507718967e-05, "loss": 0.0001, "step": 129050 }, { "epoch": 27.83, "learning_rate": 2.7187582504534312e-05, "loss": 0.0, "step": 129100 }, { "epoch": 27.84, "learning_rate": 2.7184291776212802e-05, "loss": 0.0001, "step": 129150 }, { "epoch": 27.85, "learning_rate": 2.7180999323220234e-05, "loss": 0.0001, "step": 129200 }, { "epoch": 27.86, "learning_rate": 2.717770514602265e-05, "loss": 0.0001, "step": 129250 }, { "epoch": 27.87, "learning_rate": 2.7174409245086334e-05, "loss": 0.0, "step": 129300 }, { "epoch": 27.88, "learning_rate": 2.7171111620877825e-05, "loss": 0.0, "step": 129350 }, { "epoch": 27.89, "learning_rate": 2.716781227386389e-05, "loss": 0.0, "step": 129400 }, { "epoch": 27.9, "learning_rate": 2.716451120451155e-05, "loss": 0.0001, "step": 129450 }, { "epoch": 27.92, "learning_rate": 2.716120841328806e-05, "loss": 0.0001, "step": 129500 }, { "epoch": 27.93, "learning_rate": 2.7157903900660942e-05, "loss": 0.0001, "step": 129550 }, { "epoch": 27.94, "learning_rate": 2.7154597667097932e-05, "loss": 0.0, "step": 129600 }, { "epoch": 27.95, "learning_rate": 2.7151289713067028e-05, "loss": 0.0, "step": 129650 }, { "epoch": 27.96, "learning_rate": 2.7147980039036466e-05, "loss": 0.0, "step": 129700 }, { "epoch": 27.97, "learning_rate": 2.7144668645474726e-05, "loss": 0.0, "step": 129750 }, { "epoch": 27.98, "learning_rate": 2.7141355532850532e-05, "loss": 0.0001, "step": 129800 }, { "epoch": 27.99, "learning_rate": 2.7138040701632853e-05, "loss": 0.0, "step": 129850 }, { "epoch": 28.0, "learning_rate": 2.7134724152290895e-05, "loss": 0.0001, "step": 129900 }, { "epoch": 28.01, "learning_rate": 2.7131405885294113e-05, "loss": 0.0, "step": 129950 }, { "epoch": 28.02, "learning_rate": 2.71280859011122e-05, "loss": 0.0, "step": 130000 }, { "epoch": 28.03, "learning_rate": 2.7124764200215105e-05, "loss": 0.0, "step": 130050 }, { "epoch": 28.04, "learning_rate": 2.7121440783073002e-05, "loss": 0.0, "step": 130100 }, { "epoch": 28.06, "learning_rate": 2.7118115650156322e-05, "loss": 0.0001, "step": 130150 }, { "epoch": 28.07, "learning_rate": 2.7114788801935728e-05, "loss": 0.0, "step": 130200 }, { "epoch": 28.08, "learning_rate": 2.7111460238882133e-05, "loss": 0.0, "step": 130250 }, { "epoch": 28.09, "learning_rate": 2.7108129961466693e-05, "loss": 0.0, "step": 130300 }, { "epoch": 28.1, "learning_rate": 2.7104797970160804e-05, "loss": 0.0, "step": 130350 }, { "epoch": 28.11, "learning_rate": 2.7101464265436107e-05, "loss": 0.0001, "step": 130400 }, { "epoch": 28.12, "learning_rate": 2.709812884776448e-05, "loss": 0.0, "step": 130450 }, { "epoch": 28.13, "learning_rate": 2.709479171761804e-05, "loss": 0.0001, "step": 130500 }, { "epoch": 28.14, "learning_rate": 2.709145287546917e-05, "loss": 0.0, "step": 130550 }, { "epoch": 28.15, "learning_rate": 2.708811232179047e-05, "loss": 0.0, "step": 130600 }, { "epoch": 28.16, "learning_rate": 2.708477005705479e-05, "loss": 0.0001, "step": 130650 }, { "epoch": 28.17, "learning_rate": 2.7081426081735223e-05, "loss": 0.0, "step": 130700 }, { "epoch": 28.18, "learning_rate": 2.707808039630511e-05, "loss": 0.0, "step": 130750 }, { "epoch": 28.2, "learning_rate": 2.7074733001238024e-05, "loss": 0.0001, "step": 130800 }, { "epoch": 28.21, "learning_rate": 2.7071383897007786e-05, "loss": 0.0001, "step": 130850 }, { "epoch": 28.22, "learning_rate": 2.7068033084088457e-05, "loss": 0.0, "step": 130900 }, { "epoch": 28.23, "learning_rate": 2.706468056295434e-05, "loss": 0.0, "step": 130950 }, { "epoch": 28.24, "learning_rate": 2.706132633407998e-05, "loss": 0.0, "step": 131000 }, { "epoch": 28.25, "learning_rate": 2.705797039794016e-05, "loss": 0.0001, "step": 131050 }, { "epoch": 28.26, "learning_rate": 2.7054612755009925e-05, "loss": 0.0, "step": 131100 }, { "epoch": 28.27, "learning_rate": 2.7051253405764524e-05, "loss": 0.0001, "step": 131150 }, { "epoch": 28.28, "learning_rate": 2.7047892350679477e-05, "loss": 0.0001, "step": 131200 }, { "epoch": 28.29, "learning_rate": 2.7044529590230543e-05, "loss": 0.0001, "step": 131250 }, { "epoch": 28.3, "learning_rate": 2.7041165124893706e-05, "loss": 0.0, "step": 131300 }, { "epoch": 28.31, "learning_rate": 2.703779895514521e-05, "loss": 0.0, "step": 131350 }, { "epoch": 28.33, "learning_rate": 2.7034431081461528e-05, "loss": 0.0, "step": 131400 }, { "epoch": 28.34, "learning_rate": 2.7031061504319378e-05, "loss": 0.0, "step": 131450 }, { "epoch": 28.35, "learning_rate": 2.7027690224195724e-05, "loss": 0.0, "step": 131500 }, { "epoch": 28.36, "learning_rate": 2.702431724156776e-05, "loss": 0.0, "step": 131550 }, { "epoch": 28.37, "learning_rate": 2.7020942556912934e-05, "loss": 0.0001, "step": 131600 }, { "epoch": 28.38, "learning_rate": 2.701756617070892e-05, "loss": 0.0, "step": 131650 }, { "epoch": 28.39, "learning_rate": 2.701418808343365e-05, "loss": 0.0, "step": 131700 }, { "epoch": 28.4, "learning_rate": 2.7010808295565283e-05, "loss": 0.0001, "step": 131750 }, { "epoch": 28.41, "learning_rate": 2.7007426807582225e-05, "loss": 0.0, "step": 131800 }, { "epoch": 28.42, "learning_rate": 2.7004043619963118e-05, "loss": 0.0, "step": 131850 }, { "epoch": 28.43, "learning_rate": 2.7000658733186853e-05, "loss": 0.0, "step": 131900 }, { "epoch": 28.44, "learning_rate": 2.6997272147732554e-05, "loss": 0.0, "step": 131950 }, { "epoch": 28.45, "learning_rate": 2.6993883864079593e-05, "loss": 0.0001, "step": 132000 }, { "epoch": 28.47, "learning_rate": 2.6990493882707562e-05, "loss": 0.0, "step": 132050 }, { "epoch": 28.48, "learning_rate": 2.698710220409633e-05, "loss": 0.0, "step": 132100 }, { "epoch": 28.49, "learning_rate": 2.698370882872597e-05, "loss": 0.0, "step": 132150 }, { "epoch": 28.5, "learning_rate": 2.698031375707681e-05, "loss": 0.0, "step": 132200 }, { "epoch": 28.51, "learning_rate": 2.6976916989629426e-05, "loss": 0.0001, "step": 132250 }, { "epoch": 28.52, "learning_rate": 2.697351852686462e-05, "loss": 0.0001, "step": 132300 }, { "epoch": 28.53, "learning_rate": 2.6970118369263445e-05, "loss": 0.0, "step": 132350 }, { "epoch": 28.54, "learning_rate": 2.6966716517307186e-05, "loss": 0.0, "step": 132400 }, { "epoch": 28.55, "learning_rate": 2.696331297147737e-05, "loss": 0.0, "step": 132450 }, { "epoch": 28.56, "learning_rate": 2.6959907732255763e-05, "loss": 0.0001, "step": 132500 }, { "epoch": 28.57, "learning_rate": 2.6956500800124376e-05, "loss": 0.0, "step": 132550 }, { "epoch": 28.58, "learning_rate": 2.6953092175565457e-05, "loss": 0.0001, "step": 132600 }, { "epoch": 28.59, "learning_rate": 2.694968185906149e-05, "loss": 0.0, "step": 132650 }, { "epoch": 28.61, "learning_rate": 2.69462698510952e-05, "loss": 0.0, "step": 132700 }, { "epoch": 28.62, "learning_rate": 2.6942856152149552e-05, "loss": 0.0001, "step": 132750 }, { "epoch": 28.63, "learning_rate": 2.6939440762707756e-05, "loss": 0.0, "step": 132800 }, { "epoch": 28.64, "learning_rate": 2.693602368325325e-05, "loss": 0.0, "step": 132850 }, { "epoch": 28.65, "learning_rate": 2.6932604914269723e-05, "loss": 0.0001, "step": 132900 }, { "epoch": 28.66, "learning_rate": 2.6929184456241093e-05, "loss": 0.0001, "step": 132950 }, { "epoch": 28.67, "learning_rate": 2.6925762309651524e-05, "loss": 0.0, "step": 133000 }, { "epoch": 28.68, "learning_rate": 2.692233847498541e-05, "loss": 0.0, "step": 133050 }, { "epoch": 28.69, "learning_rate": 2.6918912952727396e-05, "loss": 0.0, "step": 133100 }, { "epoch": 28.7, "learning_rate": 2.6915485743362368e-05, "loss": 0.0001, "step": 133150 }, { "epoch": 28.71, "learning_rate": 2.691205684737543e-05, "loss": 0.0001, "step": 133200 }, { "epoch": 28.72, "learning_rate": 2.6908626265251938e-05, "loss": 0.0, "step": 133250 }, { "epoch": 28.73, "learning_rate": 2.6905193997477497e-05, "loss": 0.0, "step": 133300 }, { "epoch": 28.75, "learning_rate": 2.6901760044537935e-05, "loss": 0.0, "step": 133350 }, { "epoch": 28.76, "learning_rate": 2.6898324406919322e-05, "loss": 0.0, "step": 133400 }, { "epoch": 28.77, "learning_rate": 2.6894887085107973e-05, "loss": 0.0001, "step": 133450 }, { "epoch": 28.78, "learning_rate": 2.6891448079590427e-05, "loss": 0.0, "step": 133500 }, { "epoch": 28.79, "learning_rate": 2.6888007390853487e-05, "loss": 0.0001, "step": 133550 }, { "epoch": 28.8, "learning_rate": 2.6884565019384157e-05, "loss": 0.0001, "step": 133600 }, { "epoch": 28.81, "learning_rate": 2.6881120965669714e-05, "loss": 0.0, "step": 133650 }, { "epoch": 28.82, "learning_rate": 2.687767523019766e-05, "loss": 0.0, "step": 133700 }, { "epoch": 28.83, "learning_rate": 2.6874227813455733e-05, "loss": 0.0, "step": 133750 }, { "epoch": 28.84, "learning_rate": 2.68707787159319e-05, "loss": 0.0, "step": 133800 }, { "epoch": 28.85, "learning_rate": 2.6867327938114397e-05, "loss": 0.0, "step": 133850 }, { "epoch": 28.86, "learning_rate": 2.6863875480491654e-05, "loss": 0.0, "step": 133900 }, { "epoch": 28.87, "learning_rate": 2.686042134355238e-05, "loss": 0.0, "step": 133950 }, { "epoch": 28.89, "learning_rate": 2.68569655277855e-05, "loss": 0.0001, "step": 134000 }, { "epoch": 28.9, "learning_rate": 2.6853508033680173e-05, "loss": 0.0, "step": 134050 }, { "epoch": 28.91, "learning_rate": 2.6850048861725808e-05, "loss": 0.0, "step": 134100 }, { "epoch": 28.92, "learning_rate": 2.6846588012412043e-05, "loss": 0.0, "step": 134150 }, { "epoch": 28.93, "learning_rate": 2.6843125486228762e-05, "loss": 0.0001, "step": 134200 }, { "epoch": 28.94, "learning_rate": 2.6839661283666082e-05, "loss": 0.0001, "step": 134250 }, { "epoch": 28.95, "learning_rate": 2.683619540521435e-05, "loss": 0.0, "step": 134300 }, { "epoch": 28.96, "learning_rate": 2.683272785136416e-05, "loss": 0.0001, "step": 134350 }, { "epoch": 28.97, "learning_rate": 2.682925862260634e-05, "loss": 0.0, "step": 134400 }, { "epoch": 28.98, "learning_rate": 2.6825787719431955e-05, "loss": 0.0, "step": 134450 }, { "epoch": 28.99, "learning_rate": 2.6822315142332306e-05, "loss": 0.0, "step": 134500 }, { "epoch": 29.0, "learning_rate": 2.6818840891798934e-05, "loss": 0.0, "step": 134550 }, { "epoch": 29.01, "learning_rate": 2.6815364968323616e-05, "loss": 0.0, "step": 134600 }, { "epoch": 29.03, "learning_rate": 2.6811887372398355e-05, "loss": 0.0, "step": 134650 }, { "epoch": 29.04, "learning_rate": 2.6808408104515413e-05, "loss": 0.0, "step": 134700 }, { "epoch": 29.05, "learning_rate": 2.680492716516727e-05, "loss": 0.0, "step": 134750 }, { "epoch": 29.06, "learning_rate": 2.6801444554846643e-05, "loss": 0.0001, "step": 134800 }, { "epoch": 29.07, "learning_rate": 2.67979602740465e-05, "loss": 0.0, "step": 134850 }, { "epoch": 29.08, "learning_rate": 2.6794474323260036e-05, "loss": 0.0, "step": 134900 }, { "epoch": 29.09, "learning_rate": 2.6790986702980676e-05, "loss": 0.0001, "step": 134950 }, { "epoch": 29.1, "learning_rate": 2.6787497413702095e-05, "loss": 0.0, "step": 135000 }, { "epoch": 29.11, "learning_rate": 2.6784006455918194e-05, "loss": 0.0001, "step": 135050 }, { "epoch": 29.12, "learning_rate": 2.6780513830123118e-05, "loss": 0.0, "step": 135100 }, { "epoch": 29.13, "learning_rate": 2.6777019536811237e-05, "loss": 0.0, "step": 135150 }, { "epoch": 29.14, "learning_rate": 2.677352357647717e-05, "loss": 0.0, "step": 135200 }, { "epoch": 29.15, "learning_rate": 2.677002594961576e-05, "loss": 0.0, "step": 135250 }, { "epoch": 29.17, "learning_rate": 2.6766526656722096e-05, "loss": 0.0001, "step": 135300 }, { "epoch": 29.18, "learning_rate": 2.6763025698291495e-05, "loss": 0.0001, "step": 135350 }, { "epoch": 29.19, "learning_rate": 2.6759523074819518e-05, "loss": 0.0001, "step": 135400 }, { "epoch": 29.2, "learning_rate": 2.675601878680195e-05, "loss": 0.0001, "step": 135450 }, { "epoch": 29.21, "learning_rate": 2.675251283473483e-05, "loss": 0.0, "step": 135500 }, { "epoch": 29.22, "learning_rate": 2.674900521911441e-05, "loss": 0.0, "step": 135550 }, { "epoch": 29.23, "learning_rate": 2.674549594043719e-05, "loss": 0.0, "step": 135600 }, { "epoch": 29.24, "learning_rate": 2.6741984999199908e-05, "loss": 0.0, "step": 135650 }, { "epoch": 29.25, "learning_rate": 2.6738472395899532e-05, "loss": 0.0001, "step": 135700 }, { "epoch": 29.26, "learning_rate": 2.673495813103327e-05, "loss": 0.0, "step": 135750 }, { "epoch": 29.27, "learning_rate": 2.673144220509855e-05, "loss": 0.0001, "step": 135800 }, { "epoch": 29.28, "learning_rate": 2.6727924618593058e-05, "loss": 0.0, "step": 135850 }, { "epoch": 29.3, "learning_rate": 2.6724405372014704e-05, "loss": 0.0001, "step": 135900 }, { "epoch": 29.31, "learning_rate": 2.6720884465861626e-05, "loss": 0.0, "step": 135950 }, { "epoch": 29.32, "learning_rate": 2.6717361900632206e-05, "loss": 0.0, "step": 136000 }, { "epoch": 29.33, "learning_rate": 2.671383767682506e-05, "loss": 0.0, "step": 136050 }, { "epoch": 29.34, "learning_rate": 2.671031179493904e-05, "loss": 0.0001, "step": 136100 }, { "epoch": 29.35, "learning_rate": 2.6706784255473227e-05, "loss": 0.0, "step": 136150 }, { "epoch": 29.36, "learning_rate": 2.6703255058926943e-05, "loss": 0.0, "step": 136200 }, { "epoch": 29.37, "learning_rate": 2.6699724205799733e-05, "loss": 0.0, "step": 136250 }, { "epoch": 29.38, "learning_rate": 2.669619169659139e-05, "loss": 0.0, "step": 136300 }, { "epoch": 29.39, "learning_rate": 2.6692657531801947e-05, "loss": 0.0, "step": 136350 }, { "epoch": 29.4, "learning_rate": 2.668912171193164e-05, "loss": 0.0, "step": 136400 }, { "epoch": 29.41, "learning_rate": 2.668558423748097e-05, "loss": 0.0, "step": 136450 }, { "epoch": 29.42, "learning_rate": 2.6682045108950666e-05, "loss": 0.0001, "step": 136500 }, { "epoch": 29.44, "learning_rate": 2.6678504326841683e-05, "loss": 0.0, "step": 136550 }, { "epoch": 29.45, "learning_rate": 2.667496189165521e-05, "loss": 0.0, "step": 136600 }, { "epoch": 29.46, "learning_rate": 2.6671417803892685e-05, "loss": 0.0, "step": 136650 }, { "epoch": 29.47, "learning_rate": 2.6667872064055755e-05, "loss": 0.0, "step": 136700 }, { "epoch": 29.48, "learning_rate": 2.6664324672646326e-05, "loss": 0.0, "step": 136750 }, { "epoch": 29.49, "learning_rate": 2.666077563016652e-05, "loss": 0.0001, "step": 136800 }, { "epoch": 29.5, "learning_rate": 2.665722493711871e-05, "loss": 0.0, "step": 136850 }, { "epoch": 29.51, "learning_rate": 2.6653672594005482e-05, "loss": 0.0001, "step": 136900 }, { "epoch": 29.52, "learning_rate": 2.665011860132967e-05, "loss": 0.0, "step": 136950 }, { "epoch": 29.53, "learning_rate": 2.664656295959433e-05, "loss": 0.0, "step": 137000 }, { "epoch": 29.54, "learning_rate": 2.664300566930277e-05, "loss": 0.0001, "step": 137050 }, { "epoch": 29.55, "learning_rate": 2.663944673095851e-05, "loss": 0.0, "step": 137100 }, { "epoch": 29.56, "learning_rate": 2.663588614506532e-05, "loss": 0.0, "step": 137150 }, { "epoch": 29.58, "learning_rate": 2.6632323912127194e-05, "loss": 0.0001, "step": 137200 }, { "epoch": 29.59, "learning_rate": 2.6628760032648362e-05, "loss": 0.0, "step": 137250 }, { "epoch": 29.6, "learning_rate": 2.6625194507133287e-05, "loss": 0.0001, "step": 137300 }, { "epoch": 29.61, "learning_rate": 2.662162733608666e-05, "loss": 0.0, "step": 137350 }, { "epoch": 29.62, "learning_rate": 2.6618058520013415e-05, "loss": 0.0, "step": 137400 }, { "epoch": 29.63, "learning_rate": 2.6614488059418712e-05, "loss": 0.0, "step": 137450 }, { "epoch": 29.64, "learning_rate": 2.6610915954807946e-05, "loss": 0.0, "step": 137500 }, { "epoch": 29.65, "learning_rate": 2.6607342206686742e-05, "loss": 0.0001, "step": 137550 }, { "epoch": 29.66, "learning_rate": 2.6603766815560958e-05, "loss": 0.0, "step": 137600 }, { "epoch": 29.67, "learning_rate": 2.660018978193669e-05, "loss": 0.0001, "step": 137650 }, { "epoch": 29.68, "learning_rate": 2.6596611106320262e-05, "loss": 0.0, "step": 137700 }, { "epoch": 29.69, "learning_rate": 2.659303078921823e-05, "loss": 0.0, "step": 137750 }, { "epoch": 29.7, "learning_rate": 2.658944883113738e-05, "loss": 0.0, "step": 137800 }, { "epoch": 29.72, "learning_rate": 2.6585865232584737e-05, "loss": 0.0001, "step": 137850 }, { "epoch": 29.73, "learning_rate": 2.6582279994067554e-05, "loss": 0.0, "step": 137900 }, { "epoch": 29.74, "learning_rate": 2.6578693116093324e-05, "loss": 0.0001, "step": 137950 }, { "epoch": 29.75, "learning_rate": 2.6575104599169755e-05, "loss": 0.0, "step": 138000 }, { "epoch": 29.76, "learning_rate": 2.65715144438048e-05, "loss": 0.0001, "step": 138050 }, { "epoch": 29.77, "learning_rate": 2.6567922650506643e-05, "loss": 0.0001, "step": 138100 }, { "epoch": 29.78, "learning_rate": 2.6564329219783693e-05, "loss": 0.0, "step": 138150 }, { "epoch": 29.79, "learning_rate": 2.65607341521446e-05, "loss": 0.0001, "step": 138200 }, { "epoch": 29.8, "learning_rate": 2.6557137448098242e-05, "loss": 0.0001, "step": 138250 }, { "epoch": 29.81, "learning_rate": 2.6553539108153726e-05, "loss": 0.0, "step": 138300 }, { "epoch": 29.82, "learning_rate": 2.6549939132820392e-05, "loss": 0.0, "step": 138350 }, { "epoch": 29.83, "learning_rate": 2.654633752260781e-05, "loss": 0.0001, "step": 138400 }, { "epoch": 29.84, "learning_rate": 2.654273427802579e-05, "loss": 0.0001, "step": 138450 }, { "epoch": 29.86, "learning_rate": 2.6539129399584365e-05, "loss": 0.0001, "step": 138500 }, { "epoch": 29.87, "learning_rate": 2.6535522887793793e-05, "loss": 0.0, "step": 138550 }, { "epoch": 29.88, "learning_rate": 2.6531914743164577e-05, "loss": 0.0, "step": 138600 }, { "epoch": 29.89, "learning_rate": 2.6528304966207446e-05, "loss": 0.0, "step": 138650 }, { "epoch": 29.9, "learning_rate": 2.6524693557433363e-05, "loss": 0.0, "step": 138700 }, { "epoch": 29.91, "learning_rate": 2.652108051735351e-05, "loss": 0.0, "step": 138750 }, { "epoch": 29.92, "learning_rate": 2.651746584647931e-05, "loss": 0.0, "step": 138800 }, { "epoch": 29.93, "learning_rate": 2.651384954532242e-05, "loss": 0.0, "step": 138850 }, { "epoch": 29.94, "learning_rate": 2.651023161439472e-05, "loss": 0.0001, "step": 138900 }, { "epoch": 29.95, "learning_rate": 2.650661205420832e-05, "loss": 0.0001, "step": 138950 }, { "epoch": 29.96, "learning_rate": 2.650299086527557e-05, "loss": 0.0, "step": 139000 }, { "epoch": 29.97, "learning_rate": 2.6499368048109044e-05, "loss": 0.0, "step": 139050 }, { "epoch": 29.98, "learning_rate": 2.6495743603221542e-05, "loss": 0.0, "step": 139100 }, { "epoch": 30.0, "learning_rate": 2.6492117531126104e-05, "loss": 0.0, "step": 139150 }, { "epoch": 30.01, "learning_rate": 2.6488489832335996e-05, "loss": 0.0, "step": 139200 }, { "epoch": 30.02, "learning_rate": 2.648486050736471e-05, "loss": 0.0, "step": 139250 }, { "epoch": 30.03, "learning_rate": 2.6481229556725983e-05, "loss": 0.0001, "step": 139300 }, { "epoch": 30.04, "learning_rate": 2.647759698093376e-05, "loss": 0.0001, "step": 139350 }, { "epoch": 30.05, "learning_rate": 2.6473962780502233e-05, "loss": 0.0, "step": 139400 }, { "epoch": 30.06, "learning_rate": 2.6470326955945813e-05, "loss": 0.0001, "step": 139450 }, { "epoch": 30.07, "learning_rate": 2.646668950777915e-05, "loss": 0.0, "step": 139500 }, { "epoch": 30.08, "learning_rate": 2.646305043651712e-05, "loss": 0.0001, "step": 139550 }, { "epoch": 30.09, "learning_rate": 2.6459409742674833e-05, "loss": 0.0001, "step": 139600 }, { "epoch": 30.1, "learning_rate": 2.645576742676762e-05, "loss": 0.0001, "step": 139650 }, { "epoch": 30.11, "learning_rate": 2.645212348931105e-05, "loss": 0.0, "step": 139700 }, { "epoch": 30.13, "learning_rate": 2.644847793082091e-05, "loss": 0.0001, "step": 139750 }, { "epoch": 30.14, "learning_rate": 2.644483075181323e-05, "loss": 0.0, "step": 139800 }, { "epoch": 30.15, "learning_rate": 2.6441181952804264e-05, "loss": 0.0, "step": 139850 }, { "epoch": 30.16, "learning_rate": 2.6437531534310494e-05, "loss": 0.0, "step": 139900 }, { "epoch": 30.17, "learning_rate": 2.6433879496848633e-05, "loss": 0.0, "step": 139950 }, { "epoch": 30.18, "learning_rate": 2.6430225840935616e-05, "loss": 0.0, "step": 140000 }, { "epoch": 30.19, "learning_rate": 2.6426570567088623e-05, "loss": 0.0001, "step": 140050 }, { "epoch": 30.2, "learning_rate": 2.6422913675825047e-05, "loss": 0.0, "step": 140100 }, { "epoch": 30.21, "learning_rate": 2.6419255167662518e-05, "loss": 0.0, "step": 140150 }, { "epoch": 30.22, "learning_rate": 2.6415595043118895e-05, "loss": 0.0, "step": 140200 }, { "epoch": 30.23, "learning_rate": 2.6411933302712256e-05, "loss": 0.0001, "step": 140250 }, { "epoch": 30.24, "learning_rate": 2.640826994696093e-05, "loss": 0.0, "step": 140300 }, { "epoch": 30.25, "learning_rate": 2.6404604976383446e-05, "loss": 0.0, "step": 140350 }, { "epoch": 30.27, "learning_rate": 2.6400938391498585e-05, "loss": 0.0001, "step": 140400 }, { "epoch": 30.28, "learning_rate": 2.6397270192825343e-05, "loss": 0.0, "step": 140450 }, { "epoch": 30.29, "learning_rate": 2.6393600380882955e-05, "loss": 0.0001, "step": 140500 }, { "epoch": 30.3, "learning_rate": 2.638992895619087e-05, "loss": 0.0, "step": 140550 }, { "epoch": 30.31, "learning_rate": 2.6386255919268776e-05, "loss": 0.0, "step": 140600 }, { "epoch": 30.32, "learning_rate": 2.638258127063659e-05, "loss": 0.0, "step": 140650 }, { "epoch": 30.33, "learning_rate": 2.637890501081445e-05, "loss": 0.0001, "step": 140700 }, { "epoch": 30.34, "learning_rate": 2.6375227140322732e-05, "loss": 0.0001, "step": 140750 }, { "epoch": 30.35, "learning_rate": 2.6371547659682024e-05, "loss": 0.0, "step": 140800 }, { "epoch": 30.36, "learning_rate": 2.636786656941316e-05, "loss": 0.0, "step": 140850 }, { "epoch": 30.37, "learning_rate": 2.636418387003719e-05, "loss": 0.0, "step": 140900 }, { "epoch": 30.38, "learning_rate": 2.63604995620754e-05, "loss": 0.0, "step": 140950 }, { "epoch": 30.39, "learning_rate": 2.6356813646049295e-05, "loss": 0.0001, "step": 141000 }, { "epoch": 30.41, "learning_rate": 2.635312612248061e-05, "loss": 0.0, "step": 141050 }, { "epoch": 30.42, "learning_rate": 2.6349436991891313e-05, "loss": 0.0, "step": 141100 }, { "epoch": 30.43, "learning_rate": 2.6345746254803594e-05, "loss": 0.0, "step": 141150 }, { "epoch": 30.44, "learning_rate": 2.6342053911739872e-05, "loss": 0.0, "step": 141200 }, { "epoch": 30.45, "learning_rate": 2.6338359963222797e-05, "loss": 0.0, "step": 141250 }, { "epoch": 30.46, "learning_rate": 2.6334664409775236e-05, "loss": 0.0, "step": 141300 }, { "epoch": 30.47, "learning_rate": 2.6330967251920296e-05, "loss": 0.0, "step": 141350 }, { "epoch": 30.48, "learning_rate": 2.6327268490181304e-05, "loss": 0.0, "step": 141400 }, { "epoch": 30.49, "learning_rate": 2.632356812508181e-05, "loss": 0.0, "step": 141450 }, { "epoch": 30.5, "learning_rate": 2.6319866157145603e-05, "loss": 0.0, "step": 141500 }, { "epoch": 30.51, "learning_rate": 2.6316162586896684e-05, "loss": 0.0, "step": 141550 }, { "epoch": 30.52, "learning_rate": 2.6312457414859304e-05, "loss": 0.0, "step": 141600 }, { "epoch": 30.53, "learning_rate": 2.6308750641557907e-05, "loss": 0.0, "step": 141650 }, { "epoch": 30.55, "learning_rate": 2.6305042267517187e-05, "loss": 0.0, "step": 141700 }, { "epoch": 30.56, "learning_rate": 2.6301332293262067e-05, "loss": 0.0, "step": 141750 }, { "epoch": 30.57, "learning_rate": 2.6297620719317684e-05, "loss": 0.0, "step": 141800 }, { "epoch": 30.58, "learning_rate": 2.6293907546209404e-05, "loss": 0.0, "step": 141850 }, { "epoch": 30.59, "learning_rate": 2.629019277446283e-05, "loss": 0.0, "step": 141900 }, { "epoch": 30.6, "learning_rate": 2.6286476404603777e-05, "loss": 0.0001, "step": 141950 }, { "epoch": 30.61, "learning_rate": 2.6282758437158295e-05, "loss": 0.0, "step": 142000 }, { "epoch": 30.62, "learning_rate": 2.6279038872652652e-05, "loss": 0.0, "step": 142050 }, { "epoch": 30.63, "learning_rate": 2.6275317711613357e-05, "loss": 0.0, "step": 142100 }, { "epoch": 30.64, "learning_rate": 2.6271594954567125e-05, "loss": 0.0, "step": 142150 }, { "epoch": 30.65, "learning_rate": 2.6267870602040916e-05, "loss": 0.0001, "step": 142200 }, { "epoch": 30.66, "learning_rate": 2.6264144654561907e-05, "loss": 0.0001, "step": 142250 }, { "epoch": 30.67, "learning_rate": 2.6260417112657493e-05, "loss": 0.0, "step": 142300 }, { "epoch": 30.69, "learning_rate": 2.6256687976855312e-05, "loss": 0.0, "step": 142350 }, { "epoch": 30.7, "learning_rate": 2.625295724768322e-05, "loss": 0.0, "step": 142400 }, { "epoch": 30.71, "learning_rate": 2.6249224925669285e-05, "loss": 0.0, "step": 142450 }, { "epoch": 30.72, "learning_rate": 2.624549101134182e-05, "loss": 0.0001, "step": 142500 }, { "epoch": 30.73, "learning_rate": 2.624175550522936e-05, "loss": 0.0001, "step": 142550 }, { "epoch": 30.74, "learning_rate": 2.623801840786065e-05, "loss": 0.0001, "step": 142600 }, { "epoch": 30.75, "learning_rate": 2.6234279719764685e-05, "loss": 0.0001, "step": 142650 }, { "epoch": 30.76, "learning_rate": 2.6230539441470657e-05, "loss": 0.0, "step": 142700 }, { "epoch": 30.77, "learning_rate": 2.6226797573508007e-05, "loss": 0.0, "step": 142750 }, { "epoch": 30.78, "learning_rate": 2.6223054116406393e-05, "loss": 0.0001, "step": 142800 }, { "epoch": 30.79, "learning_rate": 2.6219309070695693e-05, "loss": 0.0, "step": 142850 }, { "epoch": 30.8, "learning_rate": 2.6215562436906014e-05, "loss": 0.0, "step": 142900 }, { "epoch": 30.81, "learning_rate": 2.6211814215567683e-05, "loss": 0.0, "step": 142950 }, { "epoch": 30.83, "learning_rate": 2.6208064407211264e-05, "loss": 0.0, "step": 143000 }, { "epoch": 30.84, "learning_rate": 2.6204313012367533e-05, "loss": 0.0, "step": 143050 }, { "epoch": 30.85, "learning_rate": 2.6200560031567493e-05, "loss": 0.0, "step": 143100 }, { "epoch": 30.86, "learning_rate": 2.6196805465342378e-05, "loss": 0.0, "step": 143150 }, { "epoch": 30.87, "learning_rate": 2.619304931422364e-05, "loss": 0.0, "step": 143200 }, { "epoch": 30.88, "learning_rate": 2.618929157874296e-05, "loss": 0.0, "step": 143250 }, { "epoch": 30.89, "learning_rate": 2.6185532259432235e-05, "loss": 0.0001, "step": 143300 }, { "epoch": 30.9, "learning_rate": 2.61817713568236e-05, "loss": 0.0001, "step": 143350 }, { "epoch": 30.91, "learning_rate": 2.6178008871449405e-05, "loss": 0.0001, "step": 143400 }, { "epoch": 30.92, "learning_rate": 2.6174244803842213e-05, "loss": 0.0, "step": 143450 }, { "epoch": 30.93, "learning_rate": 2.6170479154534834e-05, "loss": 0.0, "step": 143500 }, { "epoch": 30.94, "learning_rate": 2.6166711924060292e-05, "loss": 0.0001, "step": 143550 }, { "epoch": 30.95, "learning_rate": 2.6162943112951827e-05, "loss": 0.0001, "step": 143600 }, { "epoch": 30.97, "learning_rate": 2.6159172721742915e-05, "loss": 0.0, "step": 143650 }, { "epoch": 30.98, "learning_rate": 2.6155400750967245e-05, "loss": 0.0, "step": 143700 }, { "epoch": 30.99, "learning_rate": 2.615162720115874e-05, "loss": 0.0001, "step": 143750 }, { "epoch": 31.0, "learning_rate": 2.6147852072851538e-05, "loss": 0.0001, "step": 143800 }, { "epoch": 31.01, "learning_rate": 2.6144075366580004e-05, "loss": 0.0001, "step": 143850 }, { "epoch": 31.02, "learning_rate": 2.6140297082878728e-05, "loss": 0.0, "step": 143900 }, { "epoch": 31.03, "learning_rate": 2.6136517222282516e-05, "loss": 0.0, "step": 143950 }, { "epoch": 31.04, "learning_rate": 2.6132735785326408e-05, "loss": 0.0, "step": 144000 }, { "epoch": 31.05, "learning_rate": 2.612895277254566e-05, "loss": 0.0001, "step": 144050 }, { "epoch": 31.06, "learning_rate": 2.612516818447575e-05, "loss": 0.0001, "step": 144100 }, { "epoch": 31.07, "learning_rate": 2.6121382021652386e-05, "loss": 0.0, "step": 144150 }, { "epoch": 31.08, "learning_rate": 2.611759428461149e-05, "loss": 0.0, "step": 144200 }, { "epoch": 31.1, "learning_rate": 2.6113804973889218e-05, "loss": 0.0, "step": 144250 }, { "epoch": 31.11, "learning_rate": 2.6110014090021934e-05, "loss": 0.0, "step": 144300 }, { "epoch": 31.12, "learning_rate": 2.610622163354624e-05, "loss": 0.0, "step": 144350 }, { "epoch": 31.13, "learning_rate": 2.610242760499895e-05, "loss": 0.0, "step": 144400 }, { "epoch": 31.14, "learning_rate": 2.6098632004917106e-05, "loss": 0.0, "step": 144450 }, { "epoch": 31.15, "learning_rate": 2.6094834833837962e-05, "loss": 0.0, "step": 144500 }, { "epoch": 31.16, "learning_rate": 2.6091036092299012e-05, "loss": 0.0, "step": 144550 }, { "epoch": 31.17, "learning_rate": 2.608723578083796e-05, "loss": 0.0001, "step": 144600 }, { "epoch": 31.18, "learning_rate": 2.608343389999274e-05, "loss": 0.0001, "step": 144650 }, { "epoch": 31.19, "learning_rate": 2.6079630450301494e-05, "loss": 0.0, "step": 144700 }, { "epoch": 31.2, "learning_rate": 2.60758254323026e-05, "loss": 0.0, "step": 144750 }, { "epoch": 31.21, "learning_rate": 2.6072018846534658e-05, "loss": 0.0, "step": 144800 }, { "epoch": 31.22, "learning_rate": 2.606821069353648e-05, "loss": 0.0, "step": 144850 }, { "epoch": 31.24, "learning_rate": 2.6064400973847106e-05, "loss": 0.0, "step": 144900 }, { "epoch": 31.25, "learning_rate": 2.6060589688005797e-05, "loss": 0.0001, "step": 144950 }, { "epoch": 31.26, "learning_rate": 2.6056776836552035e-05, "loss": 0.0, "step": 145000 }, { "epoch": 31.27, "learning_rate": 2.6052962420025528e-05, "loss": 0.0001, "step": 145050 }, { "epoch": 31.28, "learning_rate": 2.6049146438966196e-05, "loss": 0.0, "step": 145100 }, { "epoch": 31.29, "learning_rate": 2.6045328893914194e-05, "loss": 0.0001, "step": 145150 }, { "epoch": 31.3, "learning_rate": 2.6041509785409883e-05, "loss": 0.0, "step": 145200 }, { "epoch": 31.31, "learning_rate": 2.603768911399386e-05, "loss": 0.0, "step": 145250 }, { "epoch": 31.32, "learning_rate": 2.603386688020693e-05, "loss": 0.0, "step": 145300 }, { "epoch": 31.33, "learning_rate": 2.603004308459013e-05, "loss": 0.0, "step": 145350 }, { "epoch": 31.34, "learning_rate": 2.6026217727684705e-05, "loss": 0.0, "step": 145400 }, { "epoch": 31.35, "learning_rate": 2.6022390810032142e-05, "loss": 0.0, "step": 145450 }, { "epoch": 31.36, "learning_rate": 2.601856233217413e-05, "loss": 0.0001, "step": 145500 }, { "epoch": 31.38, "learning_rate": 2.6014732294652582e-05, "loss": 0.0, "step": 145550 }, { "epoch": 31.39, "learning_rate": 2.601090069800964e-05, "loss": 0.0, "step": 145600 }, { "epoch": 31.4, "learning_rate": 2.6007067542787657e-05, "loss": 0.0, "step": 145650 }, { "epoch": 31.41, "learning_rate": 2.6003232829529218e-05, "loss": 0.0001, "step": 145700 }, { "epoch": 31.42, "learning_rate": 2.5999396558777118e-05, "loss": 0.0001, "step": 145750 }, { "epoch": 31.43, "learning_rate": 2.599555873107437e-05, "loss": 0.0, "step": 145800 }, { "epoch": 31.44, "learning_rate": 2.5991719346964224e-05, "loss": 0.0, "step": 145850 }, { "epoch": 31.45, "learning_rate": 2.5987878406990135e-05, "loss": 0.0, "step": 145900 }, { "epoch": 31.46, "learning_rate": 2.5984035911695788e-05, "loss": 0.0, "step": 145950 }, { "epoch": 31.47, "learning_rate": 2.5980191861625073e-05, "loss": 0.0, "step": 146000 }, { "epoch": 31.48, "learning_rate": 2.597634625732212e-05, "loss": 0.0, "step": 146050 }, { "epoch": 31.49, "learning_rate": 2.597249909933127e-05, "loss": 0.0, "step": 146100 }, { "epoch": 31.5, "learning_rate": 2.596865038819707e-05, "loss": 0.0, "step": 146150 }, { "epoch": 31.52, "learning_rate": 2.5964800124464315e-05, "loss": 0.0001, "step": 146200 }, { "epoch": 31.53, "learning_rate": 2.5960948308678002e-05, "loss": 0.0, "step": 146250 }, { "epoch": 31.54, "learning_rate": 2.5957094941383344e-05, "loss": 0.0, "step": 146300 }, { "epoch": 31.55, "learning_rate": 2.595324002312579e-05, "loss": 0.0, "step": 146350 }, { "epoch": 31.56, "learning_rate": 2.594938355445099e-05, "loss": 0.0, "step": 146400 }, { "epoch": 31.57, "learning_rate": 2.5945525535904823e-05, "loss": 0.0, "step": 146450 }, { "epoch": 31.58, "learning_rate": 2.5941665968033386e-05, "loss": 0.0, "step": 146500 }, { "epoch": 31.59, "learning_rate": 2.5937804851383005e-05, "loss": 0.0, "step": 146550 }, { "epoch": 31.6, "learning_rate": 2.5933942186500206e-05, "loss": 0.0, "step": 146600 }, { "epoch": 31.61, "learning_rate": 2.5930077973931755e-05, "loss": 0.0001, "step": 146650 }, { "epoch": 31.62, "learning_rate": 2.5926212214224616e-05, "loss": 0.0, "step": 146700 }, { "epoch": 31.63, "learning_rate": 2.592234490792599e-05, "loss": 0.0, "step": 146750 }, { "epoch": 31.64, "learning_rate": 2.591847605558328e-05, "loss": 0.0, "step": 146800 }, { "epoch": 31.66, "learning_rate": 2.5914605657744127e-05, "loss": 0.0, "step": 146850 }, { "epoch": 31.67, "learning_rate": 2.5910733714956376e-05, "loss": 0.0, "step": 146900 }, { "epoch": 31.68, "learning_rate": 2.5906860227768093e-05, "loss": 0.0, "step": 146950 }, { "epoch": 31.69, "learning_rate": 2.590298519672757e-05, "loss": 0.0001, "step": 147000 }, { "epoch": 31.7, "learning_rate": 2.5899108622383316e-05, "loss": 0.0001, "step": 147050 }, { "epoch": 31.71, "learning_rate": 2.5895230505284045e-05, "loss": 0.0001, "step": 147100 }, { "epoch": 31.72, "learning_rate": 2.5891350845978707e-05, "loss": 0.0001, "step": 147150 }, { "epoch": 31.73, "learning_rate": 2.5887469645016458e-05, "loss": 0.0, "step": 147200 }, { "epoch": 31.74, "learning_rate": 2.5883586902946682e-05, "loss": 0.0001, "step": 147250 }, { "epoch": 31.75, "learning_rate": 2.5879702620318976e-05, "loss": 0.0001, "step": 147300 }, { "epoch": 31.76, "learning_rate": 2.5875816797683152e-05, "loss": 0.0001, "step": 147350 }, { "epoch": 31.77, "learning_rate": 2.5871929435589246e-05, "loss": 0.0, "step": 147400 }, { "epoch": 31.78, "learning_rate": 2.5868040534587504e-05, "loss": 0.0, "step": 147450 }, { "epoch": 31.8, "learning_rate": 2.5864150095228403e-05, "loss": 0.0, "step": 147500 }, { "epoch": 31.81, "learning_rate": 2.5860258118062626e-05, "loss": 0.0, "step": 147550 }, { "epoch": 31.82, "learning_rate": 2.585636460364107e-05, "loss": 0.0001, "step": 147600 }, { "epoch": 31.83, "learning_rate": 2.585246955251487e-05, "loss": 0.0, "step": 147650 }, { "epoch": 31.84, "learning_rate": 2.584857296523536e-05, "loss": 0.0, "step": 147700 }, { "epoch": 31.85, "learning_rate": 2.5844674842354095e-05, "loss": 0.0, "step": 147750 }, { "epoch": 31.86, "learning_rate": 2.584077518442285e-05, "loss": 0.0, "step": 147800 }, { "epoch": 31.87, "learning_rate": 2.583687399199362e-05, "loss": 0.0, "step": 147850 }, { "epoch": 31.88, "learning_rate": 2.5832971265618608e-05, "loss": 0.0, "step": 147900 }, { "epoch": 31.89, "learning_rate": 2.582906700585025e-05, "loss": 0.0, "step": 147950 }, { "epoch": 31.9, "learning_rate": 2.5825161213241176e-05, "loss": 0.0001, "step": 148000 }, { "epoch": 31.91, "learning_rate": 2.5821253888344256e-05, "loss": 0.0001, "step": 148050 }, { "epoch": 31.92, "learning_rate": 2.5817345031712562e-05, "loss": 0.0001, "step": 148100 }, { "epoch": 31.94, "learning_rate": 2.581343464389939e-05, "loss": 0.0, "step": 148150 }, { "epoch": 31.95, "learning_rate": 2.580952272545825e-05, "loss": 0.0, "step": 148200 }, { "epoch": 31.96, "learning_rate": 2.5805609276942864e-05, "loss": 0.0, "step": 148250 }, { "epoch": 31.97, "learning_rate": 2.5801694298907184e-05, "loss": 0.0001, "step": 148300 }, { "epoch": 31.98, "learning_rate": 2.5797777791905366e-05, "loss": 0.0, "step": 148350 }, { "epoch": 31.99, "learning_rate": 2.5793859756491785e-05, "loss": 0.0, "step": 148400 }, { "epoch": 32.0, "learning_rate": 2.5789940193221037e-05, "loss": 0.0, "step": 148450 }, { "epoch": 32.01, "learning_rate": 2.5786019102647934e-05, "loss": 0.0, "step": 148500 }, { "epoch": 32.02, "learning_rate": 2.5782096485327494e-05, "loss": 0.0, "step": 148550 }, { "epoch": 32.03, "learning_rate": 2.5778172341814965e-05, "loss": 0.0001, "step": 148600 }, { "epoch": 32.04, "learning_rate": 2.57742466726658e-05, "loss": 0.0, "step": 148650 }, { "epoch": 32.05, "learning_rate": 2.577031947843567e-05, "loss": 0.0, "step": 148700 }, { "epoch": 32.07, "learning_rate": 2.5766390759680473e-05, "loss": 0.0, "step": 148750 }, { "epoch": 32.08, "learning_rate": 2.576246051695631e-05, "loss": 0.0, "step": 148800 }, { "epoch": 32.09, "learning_rate": 2.57585287508195e-05, "loss": 0.0, "step": 148850 }, { "epoch": 32.1, "learning_rate": 2.5754595461826583e-05, "loss": 0.0, "step": 148900 }, { "epoch": 32.11, "learning_rate": 2.575066065053431e-05, "loss": 0.0, "step": 148950 }, { "epoch": 32.12, "learning_rate": 2.5746724317499642e-05, "loss": 0.0, "step": 149000 }, { "epoch": 32.13, "learning_rate": 2.574278646327977e-05, "loss": 0.0001, "step": 149050 }, { "epoch": 32.14, "learning_rate": 2.573884708843209e-05, "loss": 0.0001, "step": 149100 }, { "epoch": 32.15, "learning_rate": 2.5734906193514217e-05, "loss": 0.0001, "step": 149150 }, { "epoch": 32.16, "learning_rate": 2.5730963779083978e-05, "loss": 0.0001, "step": 149200 }, { "epoch": 32.17, "learning_rate": 2.5727019845699414e-05, "loss": 0.0, "step": 149250 }, { "epoch": 32.18, "learning_rate": 2.572307439391879e-05, "loss": 0.0, "step": 149300 }, { "epoch": 32.19, "learning_rate": 2.571912742430058e-05, "loss": 0.0, "step": 149350 }, { "epoch": 32.21, "learning_rate": 2.5715178937403462e-05, "loss": 0.0, "step": 149400 }, { "epoch": 32.22, "learning_rate": 2.571122893378635e-05, "loss": 0.0, "step": 149450 }, { "epoch": 32.23, "learning_rate": 2.5707277414008353e-05, "loss": 0.0, "step": 149500 }, { "epoch": 32.24, "learning_rate": 2.570332437862881e-05, "loss": 0.0, "step": 149550 }, { "epoch": 32.25, "learning_rate": 2.569936982820727e-05, "loss": 0.0, "step": 149600 }, { "epoch": 32.26, "learning_rate": 2.569541376330349e-05, "loss": 0.0, "step": 149650 }, { "epoch": 32.27, "learning_rate": 2.569145618447745e-05, "loss": 0.0, "step": 149700 }, { "epoch": 32.28, "learning_rate": 2.5687497092289338e-05, "loss": 0.0, "step": 149750 }, { "epoch": 32.29, "learning_rate": 2.5683536487299556e-05, "loss": 0.0, "step": 149800 }, { "epoch": 32.3, "learning_rate": 2.5679574370068725e-05, "loss": 0.0, "step": 149850 }, { "epoch": 32.31, "learning_rate": 2.5675610741157677e-05, "loss": 0.0, "step": 149900 }, { "epoch": 32.32, "learning_rate": 2.567164560112746e-05, "loss": 0.0, "step": 149950 }, { "epoch": 32.33, "learning_rate": 2.5667678950539335e-05, "loss": 0.0, "step": 150000 }, { "epoch": 32.35, "learning_rate": 2.5663710789954772e-05, "loss": 0.0, "step": 150050 }, { "epoch": 32.36, "learning_rate": 2.5659741119935467e-05, "loss": 0.0, "step": 150100 }, { "epoch": 32.37, "learning_rate": 2.5655769941043316e-05, "loss": 0.0, "step": 150150 }, { "epoch": 32.38, "learning_rate": 2.5651797253840432e-05, "loss": 0.0, "step": 150200 }, { "epoch": 32.39, "learning_rate": 2.564782305888915e-05, "loss": 0.0, "step": 150250 }, { "epoch": 32.4, "learning_rate": 2.5643847356752007e-05, "loss": 0.0001, "step": 150300 }, { "epoch": 32.41, "learning_rate": 2.5639870147991763e-05, "loss": 0.0, "step": 150350 }, { "epoch": 32.42, "learning_rate": 2.563589143317138e-05, "loss": 0.0001, "step": 150400 }, { "epoch": 32.43, "learning_rate": 2.5631911212854055e-05, "loss": 0.0, "step": 150450 }, { "epoch": 32.44, "learning_rate": 2.5627929487603164e-05, "loss": 0.0, "step": 150500 }, { "epoch": 32.45, "learning_rate": 2.5623946257982325e-05, "loss": 0.0, "step": 150550 }, { "epoch": 32.46, "learning_rate": 2.5619961524555367e-05, "loss": 0.0, "step": 150600 }, { "epoch": 32.47, "learning_rate": 2.5615975287886307e-05, "loss": 0.0, "step": 150650 }, { "epoch": 32.49, "learning_rate": 2.5611987548539403e-05, "loss": 0.0, "step": 150700 }, { "epoch": 32.5, "learning_rate": 2.560799830707911e-05, "loss": 0.0, "step": 150750 }, { "epoch": 32.51, "learning_rate": 2.5604007564070102e-05, "loss": 0.0, "step": 150800 }, { "epoch": 32.52, "learning_rate": 2.5600015320077267e-05, "loss": 0.0, "step": 150850 }, { "epoch": 32.53, "learning_rate": 2.5596021575665693e-05, "loss": 0.0, "step": 150900 }, { "epoch": 32.54, "learning_rate": 2.5592026331400695e-05, "loss": 0.0, "step": 150950 }, { "epoch": 32.55, "learning_rate": 2.55880295878478e-05, "loss": 0.0, "step": 151000 }, { "epoch": 32.56, "learning_rate": 2.558403134557273e-05, "loss": 0.0, "step": 151050 }, { "epoch": 32.57, "learning_rate": 2.5580031605141443e-05, "loss": 0.0, "step": 151100 }, { "epoch": 32.58, "learning_rate": 2.5576030367120087e-05, "loss": 0.0, "step": 151150 }, { "epoch": 32.59, "learning_rate": 2.5572027632075036e-05, "loss": 0.0, "step": 151200 }, { "epoch": 32.6, "learning_rate": 2.5568023400572875e-05, "loss": 0.0, "step": 151250 }, { "epoch": 32.61, "learning_rate": 2.55640176731804e-05, "loss": 0.0, "step": 151300 }, { "epoch": 32.63, "learning_rate": 2.556001045046461e-05, "loss": 0.0, "step": 151350 }, { "epoch": 32.64, "learning_rate": 2.5556001732992727e-05, "loss": 0.0, "step": 151400 }, { "epoch": 32.65, "learning_rate": 2.5551991521332175e-05, "loss": 0.0, "step": 151450 }, { "epoch": 32.66, "learning_rate": 2.5547979816050592e-05, "loss": 0.0, "step": 151500 }, { "epoch": 32.67, "learning_rate": 2.554396661771584e-05, "loss": 0.0, "step": 151550 }, { "epoch": 32.68, "learning_rate": 2.5539951926895983e-05, "loss": 0.0001, "step": 151600 }, { "epoch": 32.69, "learning_rate": 2.5535935744159278e-05, "loss": 0.0, "step": 151650 }, { "epoch": 32.7, "learning_rate": 2.553191807007423e-05, "loss": 0.0, "step": 151700 }, { "epoch": 32.71, "learning_rate": 2.552789890520953e-05, "loss": 0.0, "step": 151750 }, { "epoch": 32.72, "learning_rate": 2.5523878250134075e-05, "loss": 0.0, "step": 151800 }, { "epoch": 32.73, "learning_rate": 2.5519856105417e-05, "loss": 0.0, "step": 151850 }, { "epoch": 32.74, "learning_rate": 2.551583247162762e-05, "loss": 0.0, "step": 151900 }, { "epoch": 32.75, "learning_rate": 2.5511807349335484e-05, "loss": 0.0, "step": 151950 }, { "epoch": 32.77, "learning_rate": 2.5507780739110337e-05, "loss": 0.0, "step": 152000 }, { "epoch": 32.78, "learning_rate": 2.5503752641522148e-05, "loss": 0.0001, "step": 152050 }, { "epoch": 32.79, "learning_rate": 2.549972305714109e-05, "loss": 0.0, "step": 152100 }, { "epoch": 32.8, "learning_rate": 2.5495691986537535e-05, "loss": 0.0, "step": 152150 }, { "epoch": 32.81, "learning_rate": 2.5491659430282085e-05, "loss": 0.0, "step": 152200 }, { "epoch": 32.82, "learning_rate": 2.5487625388945537e-05, "loss": 0.0, "step": 152250 }, { "epoch": 32.83, "learning_rate": 2.548358986309891e-05, "loss": 0.0, "step": 152300 }, { "epoch": 32.84, "learning_rate": 2.5479552853313423e-05, "loss": 0.0, "step": 152350 }, { "epoch": 32.85, "learning_rate": 2.5475514360160512e-05, "loss": 0.0001, "step": 152400 }, { "epoch": 32.86, "learning_rate": 2.5471474384211825e-05, "loss": 0.0001, "step": 152450 }, { "epoch": 32.87, "learning_rate": 2.5467432926039204e-05, "loss": 0.0, "step": 152500 }, { "epoch": 32.88, "learning_rate": 2.5463389986214725e-05, "loss": 0.0, "step": 152550 }, { "epoch": 32.9, "learning_rate": 2.545934556531065e-05, "loss": 0.0001, "step": 152600 }, { "epoch": 32.91, "learning_rate": 2.5455299663899468e-05, "loss": 0.0001, "step": 152650 }, { "epoch": 32.92, "learning_rate": 2.5451252282553877e-05, "loss": 0.0001, "step": 152700 }, { "epoch": 32.93, "learning_rate": 2.5447203421846762e-05, "loss": 0.0, "step": 152750 }, { "epoch": 32.94, "learning_rate": 2.5443153082351252e-05, "loss": 0.0, "step": 152800 }, { "epoch": 32.95, "learning_rate": 2.5439101264640653e-05, "loss": 0.0, "step": 152850 }, { "epoch": 32.96, "learning_rate": 2.5435047969288504e-05, "loss": 0.0, "step": 152900 }, { "epoch": 32.97, "learning_rate": 2.5430993196868538e-05, "loss": 0.0, "step": 152950 }, { "epoch": 32.98, "learning_rate": 2.5426936947954708e-05, "loss": 0.0, "step": 153000 }, { "epoch": 32.99, "learning_rate": 2.5422879223121166e-05, "loss": 0.0, "step": 153050 }, { "epoch": 33.0, "learning_rate": 2.5418820022942287e-05, "loss": 0.0, "step": 153100 }, { "epoch": 33.01, "learning_rate": 2.541475934799263e-05, "loss": 0.0, "step": 153150 }, { "epoch": 33.02, "learning_rate": 2.541069719884699e-05, "loss": 0.0001, "step": 153200 }, { "epoch": 33.04, "learning_rate": 2.540663357608036e-05, "loss": 0.0, "step": 153250 }, { "epoch": 33.05, "learning_rate": 2.5402568480267934e-05, "loss": 0.0001, "step": 153300 }, { "epoch": 33.06, "learning_rate": 2.5398501911985125e-05, "loss": 0.0, "step": 153350 }, { "epoch": 33.07, "learning_rate": 2.539443387180755e-05, "loss": 0.0, "step": 153400 }, { "epoch": 33.08, "learning_rate": 2.5390364360311036e-05, "loss": 0.0, "step": 153450 }, { "epoch": 33.09, "learning_rate": 2.5386293378071613e-05, "loss": 0.0001, "step": 153500 }, { "epoch": 33.1, "learning_rate": 2.5382220925665527e-05, "loss": 0.0, "step": 153550 }, { "epoch": 33.11, "learning_rate": 2.5378147003669227e-05, "loss": 0.0, "step": 153600 }, { "epoch": 33.12, "learning_rate": 2.5374071612659377e-05, "loss": 0.0, "step": 153650 }, { "epoch": 33.13, "learning_rate": 2.5369994753212833e-05, "loss": 0.0, "step": 153700 }, { "epoch": 33.14, "learning_rate": 2.536591642590668e-05, "loss": 0.0, "step": 153750 }, { "epoch": 33.15, "learning_rate": 2.536183663131819e-05, "loss": 0.0, "step": 153800 }, { "epoch": 33.16, "learning_rate": 2.5357755370024862e-05, "loss": 0.0001, "step": 153850 }, { "epoch": 33.18, "learning_rate": 2.5353672642604386e-05, "loss": 0.0, "step": 153900 }, { "epoch": 33.19, "learning_rate": 2.5349588449634673e-05, "loss": 0.0, "step": 153950 }, { "epoch": 33.2, "learning_rate": 2.5345502791693827e-05, "loss": 0.0001, "step": 154000 }, { "epoch": 33.21, "learning_rate": 2.534141566936018e-05, "loss": 0.0001, "step": 154050 }, { "epoch": 33.22, "learning_rate": 2.5337327083212245e-05, "loss": 0.0, "step": 154100 }, { "epoch": 33.23, "learning_rate": 2.533323703382877e-05, "loss": 0.0, "step": 154150 }, { "epoch": 33.24, "learning_rate": 2.5329145521788688e-05, "loss": 0.0, "step": 154200 }, { "epoch": 33.25, "learning_rate": 2.5325052547671146e-05, "loss": 0.0, "step": 154250 }, { "epoch": 33.26, "learning_rate": 2.5320958112055502e-05, "loss": 0.0, "step": 154300 }, { "epoch": 33.27, "learning_rate": 2.5316862215521324e-05, "loss": 0.0, "step": 154350 }, { "epoch": 33.28, "learning_rate": 2.5312764858648372e-05, "loss": 0.0, "step": 154400 }, { "epoch": 33.29, "learning_rate": 2.5308666042016623e-05, "loss": 0.0, "step": 154450 }, { "epoch": 33.3, "learning_rate": 2.5304565766206264e-05, "loss": 0.0, "step": 154500 }, { "epoch": 33.32, "learning_rate": 2.530046403179768e-05, "loss": 0.0, "step": 154550 }, { "epoch": 33.33, "learning_rate": 2.5296360839371468e-05, "loss": 0.0, "step": 154600 }, { "epoch": 33.34, "learning_rate": 2.5292256189508428e-05, "loss": 0.0, "step": 154650 }, { "epoch": 33.35, "learning_rate": 2.528815008278957e-05, "loss": 0.0, "step": 154700 }, { "epoch": 33.36, "learning_rate": 2.5284042519796105e-05, "loss": 0.0, "step": 154750 }, { "epoch": 33.37, "learning_rate": 2.527993350110946e-05, "loss": 0.0, "step": 154800 }, { "epoch": 33.38, "learning_rate": 2.5275823027311255e-05, "loss": 0.0, "step": 154850 }, { "epoch": 33.39, "learning_rate": 2.5271711098983326e-05, "loss": 0.0, "step": 154900 }, { "epoch": 33.4, "learning_rate": 2.5267597716707703e-05, "loss": 0.0, "step": 154950 }, { "epoch": 33.41, "learning_rate": 2.5263482881066647e-05, "loss": 0.0001, "step": 155000 }, { "epoch": 33.42, "learning_rate": 2.5259366592642588e-05, "loss": 0.0001, "step": 155050 }, { "epoch": 33.43, "learning_rate": 2.5255248852018197e-05, "loss": 0.0, "step": 155100 }, { "epoch": 33.44, "learning_rate": 2.5251129659776327e-05, "loss": 0.0, "step": 155150 }, { "epoch": 33.46, "learning_rate": 2.5247009016500045e-05, "loss": 0.0, "step": 155200 }, { "epoch": 33.47, "learning_rate": 2.5242886922772628e-05, "loss": 0.0001, "step": 155250 }, { "epoch": 33.48, "learning_rate": 2.5238763379177543e-05, "loss": 0.0, "step": 155300 }, { "epoch": 33.49, "learning_rate": 2.5234638386298488e-05, "loss": 0.0, "step": 155350 }, { "epoch": 33.5, "learning_rate": 2.523051194471933e-05, "loss": 0.0, "step": 155400 }, { "epoch": 33.51, "learning_rate": 2.5226384055024184e-05, "loss": 0.0, "step": 155450 }, { "epoch": 33.52, "learning_rate": 2.5222254717797327e-05, "loss": 0.0, "step": 155500 }, { "epoch": 33.53, "learning_rate": 2.5218123933623273e-05, "loss": 0.0, "step": 155550 }, { "epoch": 33.54, "learning_rate": 2.5213991703086733e-05, "loss": 0.0, "step": 155600 }, { "epoch": 33.55, "learning_rate": 2.5209858026772607e-05, "loss": 0.0, "step": 155650 }, { "epoch": 33.56, "learning_rate": 2.5205722905266017e-05, "loss": 0.0001, "step": 155700 }, { "epoch": 33.57, "learning_rate": 2.5201586339152288e-05, "loss": 0.0, "step": 155750 }, { "epoch": 33.58, "learning_rate": 2.519744832901694e-05, "loss": 0.0, "step": 155800 }, { "epoch": 33.6, "learning_rate": 2.519330887544571e-05, "loss": 0.0, "step": 155850 }, { "epoch": 33.61, "learning_rate": 2.5189167979024522e-05, "loss": 0.0, "step": 155900 }, { "epoch": 33.62, "learning_rate": 2.5185025640339522e-05, "loss": 0.0, "step": 155950 }, { "epoch": 33.63, "learning_rate": 2.5180881859977053e-05, "loss": 0.0, "step": 156000 }, { "epoch": 33.64, "learning_rate": 2.5176736638523662e-05, "loss": 0.0, "step": 156050 }, { "epoch": 33.65, "learning_rate": 2.51725899765661e-05, "loss": 0.0, "step": 156100 }, { "epoch": 33.66, "learning_rate": 2.5168441874691314e-05, "loss": 0.0, "step": 156150 }, { "epoch": 33.67, "learning_rate": 2.516429233348647e-05, "loss": 0.0, "step": 156200 }, { "epoch": 33.68, "learning_rate": 2.516014135353893e-05, "loss": 0.0, "step": 156250 }, { "epoch": 33.69, "learning_rate": 2.515598893543626e-05, "loss": 0.0001, "step": 156300 }, { "epoch": 33.7, "learning_rate": 2.5151835079766225e-05, "loss": 0.0, "step": 156350 }, { "epoch": 33.71, "learning_rate": 2.5147679787116807e-05, "loss": 0.0, "step": 156400 }, { "epoch": 33.72, "learning_rate": 2.5143523058076167e-05, "loss": 0.0, "step": 156450 }, { "epoch": 33.74, "learning_rate": 2.5139364893232706e-05, "loss": 0.0, "step": 156500 }, { "epoch": 33.75, "learning_rate": 2.5135205293174986e-05, "loss": 0.0, "step": 156550 }, { "epoch": 33.76, "learning_rate": 2.513104425849181e-05, "loss": 0.0, "step": 156600 }, { "epoch": 33.77, "learning_rate": 2.512688178977215e-05, "loss": 0.0, "step": 156650 }, { "epoch": 33.78, "learning_rate": 2.5122717887605218e-05, "loss": 0.0, "step": 156700 }, { "epoch": 33.79, "learning_rate": 2.5118552552580394e-05, "loss": 0.0, "step": 156750 }, { "epoch": 33.8, "learning_rate": 2.511438578528728e-05, "loss": 0.0001, "step": 156800 }, { "epoch": 33.81, "learning_rate": 2.511021758631568e-05, "loss": 0.0, "step": 156850 }, { "epoch": 33.82, "learning_rate": 2.5106047956255594e-05, "loss": 0.0, "step": 156900 }, { "epoch": 33.83, "learning_rate": 2.5101876895697228e-05, "loss": 0.0001, "step": 156950 }, { "epoch": 33.84, "learning_rate": 2.509770440523099e-05, "loss": 0.0, "step": 157000 }, { "epoch": 33.85, "learning_rate": 2.5093530485447492e-05, "loss": 0.0, "step": 157050 }, { "epoch": 33.87, "learning_rate": 2.5089355136937547e-05, "loss": 0.0, "step": 157100 }, { "epoch": 33.88, "learning_rate": 2.508517836029217e-05, "loss": 0.0, "step": 157150 }, { "epoch": 33.89, "learning_rate": 2.5081000156102578e-05, "loss": 0.0, "step": 157200 }, { "epoch": 33.9, "learning_rate": 2.5076820524960194e-05, "loss": 0.0, "step": 157250 }, { "epoch": 33.91, "learning_rate": 2.507263946745663e-05, "loss": 0.0, "step": 157300 }, { "epoch": 33.92, "learning_rate": 2.5068456984183718e-05, "loss": 0.0001, "step": 157350 }, { "epoch": 33.93, "learning_rate": 2.5064273075733483e-05, "loss": 0.0, "step": 157400 }, { "epoch": 33.94, "learning_rate": 2.506008774269815e-05, "loss": 0.0001, "step": 157450 }, { "epoch": 33.95, "learning_rate": 2.505590098567015e-05, "loss": 0.0, "step": 157500 }, { "epoch": 33.96, "learning_rate": 2.5051712805242106e-05, "loss": 0.0, "step": 157550 }, { "epoch": 33.97, "learning_rate": 2.5047523202006855e-05, "loss": 0.0, "step": 157600 }, { "epoch": 33.98, "learning_rate": 2.504333217655743e-05, "loss": 0.0, "step": 157650 }, { "epoch": 33.99, "learning_rate": 2.5039139729487067e-05, "loss": 0.0, "step": 157700 }, { "epoch": 34.01, "learning_rate": 2.5034945861389193e-05, "loss": 0.0, "step": 157750 }, { "epoch": 34.02, "learning_rate": 2.503075057285746e-05, "loss": 0.0, "step": 157800 }, { "epoch": 34.03, "learning_rate": 2.5026553864485692e-05, "loss": 0.0, "step": 157850 }, { "epoch": 34.04, "learning_rate": 2.5022355736867933e-05, "loss": 0.0, "step": 157900 }, { "epoch": 34.05, "learning_rate": 2.501815619059842e-05, "loss": 0.0, "step": 157950 }, { "epoch": 34.06, "learning_rate": 2.50139552262716e-05, "loss": 0.0, "step": 158000 }, { "epoch": 34.07, "learning_rate": 2.5009752844482105e-05, "loss": 0.0, "step": 158050 }, { "epoch": 34.08, "learning_rate": 2.500554904582478e-05, "loss": 0.0, "step": 158100 }, { "epoch": 34.09, "learning_rate": 2.5001343830894674e-05, "loss": 0.0, "step": 158150 }, { "epoch": 34.1, "learning_rate": 2.4997137200287025e-05, "loss": 0.0, "step": 158200 }, { "epoch": 34.11, "learning_rate": 2.499292915459727e-05, "loss": 0.0, "step": 158250 }, { "epoch": 34.12, "learning_rate": 2.4988719694421065e-05, "loss": 0.0, "step": 158300 }, { "epoch": 34.13, "learning_rate": 2.4984508820354248e-05, "loss": 0.0, "step": 158350 }, { "epoch": 34.15, "learning_rate": 2.4980296532992857e-05, "loss": 0.0, "step": 158400 }, { "epoch": 34.16, "learning_rate": 2.4976082832933146e-05, "loss": 0.0, "step": 158450 }, { "epoch": 34.17, "learning_rate": 2.497186772077155e-05, "loss": 0.0001, "step": 158500 }, { "epoch": 34.18, "learning_rate": 2.4967651197104717e-05, "loss": 0.0, "step": 158550 }, { "epoch": 34.19, "learning_rate": 2.4963433262529492e-05, "loss": 0.0001, "step": 158600 }, { "epoch": 34.2, "learning_rate": 2.495921391764292e-05, "loss": 0.0, "step": 158650 }, { "epoch": 34.21, "learning_rate": 2.4954993163042236e-05, "loss": 0.0, "step": 158700 }, { "epoch": 34.22, "learning_rate": 2.4950770999324887e-05, "loss": 0.0, "step": 158750 }, { "epoch": 34.23, "learning_rate": 2.4946547427088518e-05, "loss": 0.0, "step": 158800 }, { "epoch": 34.24, "learning_rate": 2.4942322446930965e-05, "loss": 0.0, "step": 158850 }, { "epoch": 34.25, "learning_rate": 2.4938096059450268e-05, "loss": 0.0001, "step": 158900 }, { "epoch": 34.26, "learning_rate": 2.4933868265244675e-05, "loss": 0.0001, "step": 158950 }, { "epoch": 34.27, "learning_rate": 2.4929639064912617e-05, "loss": 0.0, "step": 159000 }, { "epoch": 34.29, "learning_rate": 2.4925408459052733e-05, "loss": 0.0, "step": 159050 }, { "epoch": 34.3, "learning_rate": 2.4921176448263868e-05, "loss": 0.0, "step": 159100 }, { "epoch": 34.31, "learning_rate": 2.491694303314505e-05, "loss": 0.0, "step": 159150 }, { "epoch": 34.32, "learning_rate": 2.491270821429551e-05, "loss": 0.0, "step": 159200 }, { "epoch": 34.33, "learning_rate": 2.4908471992314684e-05, "loss": 0.0, "step": 159250 }, { "epoch": 34.34, "learning_rate": 2.4904234367802208e-05, "loss": 0.0001, "step": 159300 }, { "epoch": 34.35, "learning_rate": 2.489999534135791e-05, "loss": 0.0, "step": 159350 }, { "epoch": 34.36, "learning_rate": 2.4895754913581823e-05, "loss": 0.0, "step": 159400 }, { "epoch": 34.37, "learning_rate": 2.489151308507417e-05, "loss": 0.0, "step": 159450 }, { "epoch": 34.38, "learning_rate": 2.4887269856435375e-05, "loss": 0.0, "step": 159500 }, { "epoch": 34.39, "learning_rate": 2.4883025228266062e-05, "loss": 0.0, "step": 159550 }, { "epoch": 34.4, "learning_rate": 2.4878779201167054e-05, "loss": 0.0, "step": 159600 }, { "epoch": 34.41, "learning_rate": 2.487453177573937e-05, "loss": 0.0, "step": 159650 }, { "epoch": 34.43, "learning_rate": 2.4870282952584232e-05, "loss": 0.0, "step": 159700 }, { "epoch": 34.44, "learning_rate": 2.486603273230305e-05, "loss": 0.0001, "step": 159750 }, { "epoch": 34.45, "learning_rate": 2.486178111549744e-05, "loss": 0.0, "step": 159800 }, { "epoch": 34.46, "learning_rate": 2.485752810276921e-05, "loss": 0.0, "step": 159850 }, { "epoch": 34.47, "learning_rate": 2.4853273694720375e-05, "loss": 0.0, "step": 159900 }, { "epoch": 34.48, "learning_rate": 2.484901789195313e-05, "loss": 0.0, "step": 159950 }, { "epoch": 34.49, "learning_rate": 2.484476069506989e-05, "loss": 0.0, "step": 160000 }, { "epoch": 34.5, "learning_rate": 2.4840502104673254e-05, "loss": 0.0, "step": 160050 }, { "epoch": 34.51, "learning_rate": 2.4836242121366014e-05, "loss": 0.0, "step": 160100 }, { "epoch": 34.52, "learning_rate": 2.483198074575117e-05, "loss": 0.0, "step": 160150 }, { "epoch": 34.53, "learning_rate": 2.482771797843191e-05, "loss": 0.0, "step": 160200 }, { "epoch": 34.54, "learning_rate": 2.482345382001163e-05, "loss": 0.0, "step": 160250 }, { "epoch": 34.55, "learning_rate": 2.4819188271093906e-05, "loss": 0.0, "step": 160300 }, { "epoch": 34.57, "learning_rate": 2.4814921332282534e-05, "loss": 0.0, "step": 160350 }, { "epoch": 34.58, "learning_rate": 2.4810653004181484e-05, "loss": 0.0, "step": 160400 }, { "epoch": 34.59, "learning_rate": 2.4806383287394935e-05, "loss": 0.0, "step": 160450 }, { "epoch": 34.6, "learning_rate": 2.4802112182527262e-05, "loss": 0.0001, "step": 160500 }, { "epoch": 34.61, "learning_rate": 2.4797839690183033e-05, "loss": 0.0, "step": 160550 }, { "epoch": 34.62, "learning_rate": 2.4793565810967013e-05, "loss": 0.0, "step": 160600 }, { "epoch": 34.63, "learning_rate": 2.478929054548417e-05, "loss": 0.0, "step": 160650 }, { "epoch": 34.64, "learning_rate": 2.4785013894339656e-05, "loss": 0.0, "step": 160700 }, { "epoch": 34.65, "learning_rate": 2.4780735858138824e-05, "loss": 0.0, "step": 160750 }, { "epoch": 34.66, "learning_rate": 2.477645643748723e-05, "loss": 0.0, "step": 160800 }, { "epoch": 34.67, "learning_rate": 2.4772175632990613e-05, "loss": 0.0, "step": 160850 }, { "epoch": 34.68, "learning_rate": 2.4767893445254925e-05, "loss": 0.0001, "step": 160900 }, { "epoch": 34.69, "learning_rate": 2.47636098748863e-05, "loss": 0.0, "step": 160950 }, { "epoch": 34.71, "learning_rate": 2.475932492249107e-05, "loss": 0.0, "step": 161000 }, { "epoch": 34.72, "learning_rate": 2.475503858867577e-05, "loss": 0.0, "step": 161050 }, { "epoch": 34.73, "learning_rate": 2.475075087404712e-05, "loss": 0.0, "step": 161100 }, { "epoch": 34.74, "learning_rate": 2.4746461779212043e-05, "loss": 0.0, "step": 161150 }, { "epoch": 34.75, "learning_rate": 2.4742171304777653e-05, "loss": 0.0, "step": 161200 }, { "epoch": 34.76, "learning_rate": 2.4737879451351265e-05, "loss": 0.0001, "step": 161250 }, { "epoch": 34.77, "learning_rate": 2.4733586219540377e-05, "loss": 0.0, "step": 161300 }, { "epoch": 34.78, "learning_rate": 2.47292916099527e-05, "loss": 0.0, "step": 161350 }, { "epoch": 34.79, "learning_rate": 2.472499562319613e-05, "loss": 0.0, "step": 161400 }, { "epoch": 34.8, "learning_rate": 2.4720698259878748e-05, "loss": 0.0, "step": 161450 }, { "epoch": 34.81, "learning_rate": 2.471639952060885e-05, "loss": 0.0, "step": 161500 }, { "epoch": 34.82, "learning_rate": 2.471209940599492e-05, "loss": 0.0, "step": 161550 }, { "epoch": 34.84, "learning_rate": 2.4707797916645624e-05, "loss": 0.0, "step": 161600 }, { "epoch": 34.85, "learning_rate": 2.4703495053169838e-05, "loss": 0.0, "step": 161650 }, { "epoch": 34.86, "learning_rate": 2.4699190816176623e-05, "loss": 0.0, "step": 161700 }, { "epoch": 34.87, "learning_rate": 2.4694885206275245e-05, "loss": 0.0001, "step": 161750 }, { "epoch": 34.88, "learning_rate": 2.4690578224075153e-05, "loss": 0.0, "step": 161800 }, { "epoch": 34.89, "learning_rate": 2.4686269870186e-05, "loss": 0.0, "step": 161850 }, { "epoch": 34.9, "learning_rate": 2.468196014521761e-05, "loss": 0.0001, "step": 161900 }, { "epoch": 34.91, "learning_rate": 2.4677649049780045e-05, "loss": 0.0, "step": 161950 }, { "epoch": 34.92, "learning_rate": 2.467333658448352e-05, "loss": 0.0, "step": 162000 }, { "epoch": 34.93, "learning_rate": 2.4669022749938462e-05, "loss": 0.0, "step": 162050 }, { "epoch": 34.94, "learning_rate": 2.466470754675549e-05, "loss": 0.0001, "step": 162100 }, { "epoch": 34.95, "learning_rate": 2.4660390975545407e-05, "loss": 0.0, "step": 162150 }, { "epoch": 34.96, "learning_rate": 2.4656073036919227e-05, "loss": 0.0, "step": 162200 }, { "epoch": 34.98, "learning_rate": 2.4651753731488154e-05, "loss": 0.0, "step": 162250 }, { "epoch": 34.99, "learning_rate": 2.4647433059863566e-05, "loss": 0.0, "step": 162300 }, { "epoch": 35.0, "learning_rate": 2.4643111022657062e-05, "loss": 0.0, "step": 162350 }, { "epoch": 35.01, "learning_rate": 2.4638787620480412e-05, "loss": 0.0, "step": 162400 }, { "epoch": 35.02, "learning_rate": 2.463446285394559e-05, "loss": 0.0, "step": 162450 }, { "epoch": 35.03, "learning_rate": 2.4630136723664762e-05, "loss": 0.0, "step": 162500 }, { "epoch": 35.04, "learning_rate": 2.462580923025029e-05, "loss": 0.0, "step": 162550 }, { "epoch": 35.05, "learning_rate": 2.4621480374314718e-05, "loss": 0.0, "step": 162600 }, { "epoch": 35.06, "learning_rate": 2.4617150156470796e-05, "loss": 0.0, "step": 162650 }, { "epoch": 35.07, "learning_rate": 2.461281857733145e-05, "loss": 0.0, "step": 162700 }, { "epoch": 35.08, "learning_rate": 2.460848563750983e-05, "loss": 0.0, "step": 162750 }, { "epoch": 35.09, "learning_rate": 2.4604151337619242e-05, "loss": 0.0, "step": 162800 }, { "epoch": 35.1, "learning_rate": 2.4599815678273202e-05, "loss": 0.0, "step": 162850 }, { "epoch": 35.12, "learning_rate": 2.4595478660085425e-05, "loss": 0.0, "step": 162900 }, { "epoch": 35.13, "learning_rate": 2.4591140283669798e-05, "loss": 0.0, "step": 162950 }, { "epoch": 35.14, "learning_rate": 2.458680054964042e-05, "loss": 0.0, "step": 163000 }, { "epoch": 35.15, "learning_rate": 2.458245945861158e-05, "loss": 0.0, "step": 163050 }, { "epoch": 35.16, "learning_rate": 2.457811701119775e-05, "loss": 0.0, "step": 163100 }, { "epoch": 35.17, "learning_rate": 2.4573773208013594e-05, "loss": 0.0, "step": 163150 }, { "epoch": 35.18, "learning_rate": 2.456942804967398e-05, "loss": 0.0, "step": 163200 }, { "epoch": 35.19, "learning_rate": 2.4565081536793945e-05, "loss": 0.0, "step": 163250 }, { "epoch": 35.2, "learning_rate": 2.4560733669988747e-05, "loss": 0.0, "step": 163300 }, { "epoch": 35.21, "learning_rate": 2.4556384449873815e-05, "loss": 0.0, "step": 163350 }, { "epoch": 35.22, "learning_rate": 2.4552033877064773e-05, "loss": 0.0, "step": 163400 }, { "epoch": 35.23, "learning_rate": 2.4547681952177448e-05, "loss": 0.0, "step": 163450 }, { "epoch": 35.24, "learning_rate": 2.4543328675827835e-05, "loss": 0.0, "step": 163500 }, { "epoch": 35.26, "learning_rate": 2.4538974048632155e-05, "loss": 0.0, "step": 163550 }, { "epoch": 35.27, "learning_rate": 2.4534618071206777e-05, "loss": 0.0001, "step": 163600 }, { "epoch": 35.28, "learning_rate": 2.4530260744168306e-05, "loss": 0.0, "step": 163650 }, { "epoch": 35.29, "learning_rate": 2.4525902068133498e-05, "loss": 0.0, "step": 163700 }, { "epoch": 35.3, "learning_rate": 2.452154204371933e-05, "loss": 0.0, "step": 163750 }, { "epoch": 35.31, "learning_rate": 2.4517180671542952e-05, "loss": 0.0, "step": 163800 }, { "epoch": 35.32, "learning_rate": 2.4512817952221713e-05, "loss": 0.0, "step": 163850 }, { "epoch": 35.33, "learning_rate": 2.4508453886373144e-05, "loss": 0.0, "step": 163900 }, { "epoch": 35.34, "learning_rate": 2.450408847461499e-05, "loss": 0.0001, "step": 163950 }, { "epoch": 35.35, "learning_rate": 2.4499721717565146e-05, "loss": 0.0, "step": 164000 }, { "epoch": 35.36, "learning_rate": 2.4495353615841742e-05, "loss": 0.0, "step": 164050 }, { "epoch": 35.37, "learning_rate": 2.4490984170063066e-05, "loss": 0.0, "step": 164100 }, { "epoch": 35.38, "learning_rate": 2.4486613380847608e-05, "loss": 0.0, "step": 164150 }, { "epoch": 35.4, "learning_rate": 2.4482241248814054e-05, "loss": 0.0, "step": 164200 }, { "epoch": 35.41, "learning_rate": 2.4477867774581264e-05, "loss": 0.0001, "step": 164250 }, { "epoch": 35.42, "learning_rate": 2.447349295876831e-05, "loss": 0.0001, "step": 164300 }, { "epoch": 35.43, "learning_rate": 2.4469116801994426e-05, "loss": 0.0, "step": 164350 }, { "epoch": 35.44, "learning_rate": 2.4464739304879067e-05, "loss": 0.0, "step": 164400 }, { "epoch": 35.45, "learning_rate": 2.4460360468041853e-05, "loss": 0.0, "step": 164450 }, { "epoch": 35.46, "learning_rate": 2.4455980292102605e-05, "loss": 0.0001, "step": 164500 }, { "epoch": 35.47, "learning_rate": 2.445159877768133e-05, "loss": 0.0, "step": 164550 }, { "epoch": 35.48, "learning_rate": 2.444721592539823e-05, "loss": 0.0, "step": 164600 }, { "epoch": 35.49, "learning_rate": 2.4442831735873687e-05, "loss": 0.0, "step": 164650 }, { "epoch": 35.5, "learning_rate": 2.443844620972828e-05, "loss": 0.0, "step": 164700 }, { "epoch": 35.51, "learning_rate": 2.4434059347582777e-05, "loss": 0.0, "step": 164750 }, { "epoch": 35.52, "learning_rate": 2.442967115005813e-05, "loss": 0.0, "step": 164800 }, { "epoch": 35.54, "learning_rate": 2.442528161777548e-05, "loss": 0.0, "step": 164850 }, { "epoch": 35.55, "learning_rate": 2.442089075135617e-05, "loss": 0.0, "step": 164900 }, { "epoch": 35.56, "learning_rate": 2.441649855142171e-05, "loss": 0.0, "step": 164950 }, { "epoch": 35.57, "learning_rate": 2.441210501859381e-05, "loss": 0.0001, "step": 165000 }, { "epoch": 35.58, "learning_rate": 2.4407710153494385e-05, "loss": 0.0, "step": 165050 }, { "epoch": 35.59, "learning_rate": 2.4403313956745503e-05, "loss": 0.0, "step": 165100 }, { "epoch": 35.6, "learning_rate": 2.439891642896945e-05, "loss": 0.0, "step": 165150 }, { "epoch": 35.61, "learning_rate": 2.4394517570788688e-05, "loss": 0.0001, "step": 165200 }, { "epoch": 35.62, "learning_rate": 2.4390117382825873e-05, "loss": 0.0, "step": 165250 }, { "epoch": 35.63, "learning_rate": 2.438571586570384e-05, "loss": 0.0, "step": 165300 }, { "epoch": 35.64, "learning_rate": 2.438131302004563e-05, "loss": 0.0001, "step": 165350 }, { "epoch": 35.65, "learning_rate": 2.4376908846474446e-05, "loss": 0.0, "step": 165400 }, { "epoch": 35.67, "learning_rate": 2.4372503345613704e-05, "loss": 0.0, "step": 165450 }, { "epoch": 35.68, "learning_rate": 2.4368096518086988e-05, "loss": 0.0001, "step": 165500 }, { "epoch": 35.69, "learning_rate": 2.4363688364518083e-05, "loss": 0.0001, "step": 165550 }, { "epoch": 35.7, "learning_rate": 2.435927888553096e-05, "loss": 0.0, "step": 165600 }, { "epoch": 35.71, "learning_rate": 2.4354868081749774e-05, "loss": 0.0, "step": 165650 }, { "epoch": 35.72, "learning_rate": 2.4350455953798865e-05, "loss": 0.0, "step": 165700 }, { "epoch": 35.73, "learning_rate": 2.4346042502302767e-05, "loss": 0.0, "step": 165750 }, { "epoch": 35.74, "learning_rate": 2.43416277278862e-05, "loss": 0.0, "step": 165800 }, { "epoch": 35.75, "learning_rate": 2.433721163117407e-05, "loss": 0.0001, "step": 165850 }, { "epoch": 35.76, "learning_rate": 2.4332794212791463e-05, "loss": 0.0, "step": 165900 }, { "epoch": 35.77, "learning_rate": 2.4328375473363665e-05, "loss": 0.0, "step": 165950 }, { "epoch": 35.78, "learning_rate": 2.432395541351615e-05, "loss": 0.0, "step": 166000 }, { "epoch": 35.79, "learning_rate": 2.431953403387455e-05, "loss": 0.0, "step": 166050 }, { "epoch": 35.81, "learning_rate": 2.4315111335064727e-05, "loss": 0.0, "step": 166100 }, { "epoch": 35.82, "learning_rate": 2.43106873177127e-05, "loss": 0.0, "step": 166150 }, { "epoch": 35.83, "learning_rate": 2.4306261982444687e-05, "loss": 0.0, "step": 166200 }, { "epoch": 35.84, "learning_rate": 2.4301835329887082e-05, "loss": 0.0, "step": 166250 }, { "epoch": 35.85, "learning_rate": 2.4297407360666474e-05, "loss": 0.0, "step": 166300 }, { "epoch": 35.86, "learning_rate": 2.4292978075409638e-05, "loss": 0.0, "step": 166350 }, { "epoch": 35.87, "learning_rate": 2.4288547474743537e-05, "loss": 0.0, "step": 166400 }, { "epoch": 35.88, "learning_rate": 2.4284115559295314e-05, "loss": 0.0001, "step": 166450 }, { "epoch": 35.89, "learning_rate": 2.42796823296923e-05, "loss": 0.0, "step": 166500 }, { "epoch": 35.9, "learning_rate": 2.4275247786562014e-05, "loss": 0.0, "step": 166550 }, { "epoch": 35.91, "learning_rate": 2.4270811930532152e-05, "loss": 0.0, "step": 166600 }, { "epoch": 35.92, "learning_rate": 2.426637476223062e-05, "loss": 0.0, "step": 166650 }, { "epoch": 35.93, "learning_rate": 2.426193628228548e-05, "loss": 0.0, "step": 166700 }, { "epoch": 35.95, "learning_rate": 2.4257496491325007e-05, "loss": 0.0, "step": 166750 }, { "epoch": 35.96, "learning_rate": 2.4253055389977628e-05, "loss": 0.0, "step": 166800 }, { "epoch": 35.97, "learning_rate": 2.4248612978871992e-05, "loss": 0.0, "step": 166850 }, { "epoch": 35.98, "learning_rate": 2.424416925863691e-05, "loss": 0.0, "step": 166900 }, { "epoch": 35.99, "learning_rate": 2.4239724229901384e-05, "loss": 0.0, "step": 166950 }, { "epoch": 36.0, "learning_rate": 2.4235277893294606e-05, "loss": 0.0, "step": 167000 }, { "epoch": 36.01, "learning_rate": 2.4230830249445945e-05, "loss": 0.0, "step": 167050 }, { "epoch": 36.02, "learning_rate": 2.422638129898496e-05, "loss": 0.0, "step": 167100 }, { "epoch": 36.03, "learning_rate": 2.4221931042541404e-05, "loss": 0.0, "step": 167150 }, { "epoch": 36.04, "learning_rate": 2.4217479480745193e-05, "loss": 0.0, "step": 167200 }, { "epoch": 36.05, "learning_rate": 2.421302661422644e-05, "loss": 0.0, "step": 167250 }, { "epoch": 36.06, "learning_rate": 2.420857244361545e-05, "loss": 0.0, "step": 167300 }, { "epoch": 36.07, "learning_rate": 2.4204116969542705e-05, "loss": 0.0, "step": 167350 }, { "epoch": 36.09, "learning_rate": 2.419966019263887e-05, "loss": 0.0, "step": 167400 }, { "epoch": 36.1, "learning_rate": 2.4195202113534786e-05, "loss": 0.0, "step": 167450 }, { "epoch": 36.11, "learning_rate": 2.4190742732861502e-05, "loss": 0.0, "step": 167500 }, { "epoch": 36.12, "learning_rate": 2.418628205125023e-05, "loss": 0.0, "step": 167550 }, { "epoch": 36.13, "learning_rate": 2.4181820069332382e-05, "loss": 0.0, "step": 167600 }, { "epoch": 36.14, "learning_rate": 2.4177356787739536e-05, "loss": 0.0, "step": 167650 }, { "epoch": 36.15, "learning_rate": 2.417289220710347e-05, "loss": 0.0, "step": 167700 }, { "epoch": 36.16, "learning_rate": 2.4168426328056137e-05, "loss": 0.0, "step": 167750 }, { "epoch": 36.17, "learning_rate": 2.416395915122968e-05, "loss": 0.0, "step": 167800 }, { "epoch": 36.18, "learning_rate": 2.415949067725642e-05, "loss": 0.0001, "step": 167850 }, { "epoch": 36.19, "learning_rate": 2.415502090676886e-05, "loss": 0.0, "step": 167900 }, { "epoch": 36.2, "learning_rate": 2.4150549840399696e-05, "loss": 0.0, "step": 167950 }, { "epoch": 36.21, "learning_rate": 2.4146077478781802e-05, "loss": 0.0, "step": 168000 }, { "epoch": 36.23, "learning_rate": 2.4141603822548228e-05, "loss": 0.0, "step": 168050 }, { "epoch": 36.24, "learning_rate": 2.4137128872332225e-05, "loss": 0.0, "step": 168100 }, { "epoch": 36.25, "learning_rate": 2.413265262876721e-05, "loss": 0.0, "step": 168150 }, { "epoch": 36.26, "learning_rate": 2.4128175092486788e-05, "loss": 0.0, "step": 168200 }, { "epoch": 36.27, "learning_rate": 2.412369626412475e-05, "loss": 0.0, "step": 168250 }, { "epoch": 36.28, "learning_rate": 2.4119216144315073e-05, "loss": 0.0, "step": 168300 }, { "epoch": 36.29, "learning_rate": 2.4114734733691907e-05, "loss": 0.0001, "step": 168350 }, { "epoch": 36.3, "learning_rate": 2.411025203288959e-05, "loss": 0.0, "step": 168400 }, { "epoch": 36.31, "learning_rate": 2.4105768042542648e-05, "loss": 0.0, "step": 168450 }, { "epoch": 36.32, "learning_rate": 2.4101282763285776e-05, "loss": 0.0, "step": 168500 }, { "epoch": 36.33, "learning_rate": 2.409679619575387e-05, "loss": 0.0, "step": 168550 }, { "epoch": 36.34, "learning_rate": 2.4092308340581985e-05, "loss": 0.0001, "step": 168600 }, { "epoch": 36.35, "learning_rate": 2.4087819198405384e-05, "loss": 0.0, "step": 168650 }, { "epoch": 36.37, "learning_rate": 2.408332876985949e-05, "loss": 0.0, "step": 168700 }, { "epoch": 36.38, "learning_rate": 2.4078837055579925e-05, "loss": 0.0, "step": 168750 }, { "epoch": 36.39, "learning_rate": 2.407434405620248e-05, "loss": 0.0, "step": 168800 }, { "epoch": 36.4, "learning_rate": 2.4069849772363132e-05, "loss": 0.0, "step": 168850 }, { "epoch": 36.41, "learning_rate": 2.4065354204698047e-05, "loss": 0.0, "step": 168900 }, { "epoch": 36.42, "learning_rate": 2.406085735384357e-05, "loss": 0.0, "step": 168950 }, { "epoch": 36.43, "learning_rate": 2.405635922043621e-05, "loss": 0.0, "step": 169000 }, { "epoch": 36.44, "learning_rate": 2.4051859805112684e-05, "loss": 0.0, "step": 169050 }, { "epoch": 36.45, "learning_rate": 2.404735910850987e-05, "loss": 0.0, "step": 169100 }, { "epoch": 36.46, "learning_rate": 2.4042857131264847e-05, "loss": 0.0, "step": 169150 }, { "epoch": 36.47, "learning_rate": 2.4038353874014858e-05, "loss": 0.0, "step": 169200 }, { "epoch": 36.48, "learning_rate": 2.403384933739733e-05, "loss": 0.0, "step": 169250 }, { "epoch": 36.49, "learning_rate": 2.4029343522049883e-05, "loss": 0.0, "step": 169300 }, { "epoch": 36.51, "learning_rate": 2.4024836428610302e-05, "loss": 0.0, "step": 169350 }, { "epoch": 36.52, "learning_rate": 2.4020328057716565e-05, "loss": 0.0, "step": 169400 }, { "epoch": 36.53, "learning_rate": 2.401581841000682e-05, "loss": 0.0, "step": 169450 }, { "epoch": 36.54, "learning_rate": 2.401130748611941e-05, "loss": 0.0, "step": 169500 }, { "epoch": 36.55, "learning_rate": 2.4006795286692844e-05, "loss": 0.0, "step": 169550 }, { "epoch": 36.56, "learning_rate": 2.4002281812365824e-05, "loss": 0.0, "step": 169600 }, { "epoch": 36.57, "learning_rate": 2.399776706377722e-05, "loss": 0.0001, "step": 169650 }, { "epoch": 36.58, "learning_rate": 2.399325104156609e-05, "loss": 0.0, "step": 169700 }, { "epoch": 36.59, "learning_rate": 2.3988733746371677e-05, "loss": 0.0001, "step": 169750 }, { "epoch": 36.6, "learning_rate": 2.3984215178833396e-05, "loss": 0.0, "step": 169800 }, { "epoch": 36.61, "learning_rate": 2.3979695339590843e-05, "loss": 0.0, "step": 169850 }, { "epoch": 36.62, "learning_rate": 2.3975174229283795e-05, "loss": 0.0, "step": 169900 }, { "epoch": 36.64, "learning_rate": 2.3970651848552206e-05, "loss": 0.0, "step": 169950 }, { "epoch": 36.65, "learning_rate": 2.3966128198036226e-05, "loss": 0.0, "step": 170000 }, { "epoch": 36.66, "learning_rate": 2.396160327837616e-05, "loss": 0.0, "step": 170050 }, { "epoch": 36.67, "learning_rate": 2.3957077090212507e-05, "loss": 0.0, "step": 170100 }, { "epoch": 36.68, "learning_rate": 2.3952549634185947e-05, "loss": 0.0, "step": 170150 }, { "epoch": 36.69, "learning_rate": 2.3948020910937327e-05, "loss": 0.0001, "step": 170200 }, { "epoch": 36.7, "learning_rate": 2.394349092110769e-05, "loss": 0.0, "step": 170250 }, { "epoch": 36.71, "learning_rate": 2.393895966533825e-05, "loss": 0.0, "step": 170300 }, { "epoch": 36.72, "learning_rate": 2.39344271442704e-05, "loss": 0.0, "step": 170350 }, { "epoch": 36.73, "learning_rate": 2.392989335854571e-05, "loss": 0.0, "step": 170400 }, { "epoch": 36.74, "learning_rate": 2.3925358308805937e-05, "loss": 0.0, "step": 170450 }, { "epoch": 36.75, "learning_rate": 2.3920821995693003e-05, "loss": 0.0, "step": 170500 }, { "epoch": 36.76, "learning_rate": 2.3916284419849024e-05, "loss": 0.0, "step": 170550 }, { "epoch": 36.78, "learning_rate": 2.3911745581916283e-05, "loss": 0.0, "step": 170600 }, { "epoch": 36.79, "learning_rate": 2.3907205482537256e-05, "loss": 0.0, "step": 170650 }, { "epoch": 36.8, "learning_rate": 2.3902664122354576e-05, "loss": 0.0001, "step": 170700 }, { "epoch": 36.81, "learning_rate": 2.389812150201108e-05, "loss": 0.0, "step": 170750 }, { "epoch": 36.82, "learning_rate": 2.3893577622149757e-05, "loss": 0.0, "step": 170800 }, { "epoch": 36.83, "learning_rate": 2.3889032483413794e-05, "loss": 0.0, "step": 170850 }, { "epoch": 36.84, "learning_rate": 2.3884486086446556e-05, "loss": 0.0, "step": 170900 }, { "epoch": 36.85, "learning_rate": 2.3879938431891565e-05, "loss": 0.0001, "step": 170950 }, { "epoch": 36.86, "learning_rate": 2.3875389520392545e-05, "loss": 0.0, "step": 171000 }, { "epoch": 36.87, "learning_rate": 2.387083935259339e-05, "loss": 0.0001, "step": 171050 }, { "epoch": 36.88, "learning_rate": 2.3866287929138168e-05, "loss": 0.0, "step": 171100 }, { "epoch": 36.89, "learning_rate": 2.3861735250671125e-05, "loss": 0.0, "step": 171150 }, { "epoch": 36.9, "learning_rate": 2.385718131783669e-05, "loss": 0.0, "step": 171200 }, { "epoch": 36.92, "learning_rate": 2.3852626131279462e-05, "loss": 0.0, "step": 171250 }, { "epoch": 36.93, "learning_rate": 2.3848069691644233e-05, "loss": 0.0, "step": 171300 }, { "epoch": 36.94, "learning_rate": 2.3843511999575947e-05, "loss": 0.0, "step": 171350 }, { "epoch": 36.95, "learning_rate": 2.383895305571975e-05, "loss": 0.0, "step": 171400 }, { "epoch": 36.96, "learning_rate": 2.3834392860720947e-05, "loss": 0.0, "step": 171450 }, { "epoch": 36.97, "learning_rate": 2.3829831415225036e-05, "loss": 0.0, "step": 171500 }, { "epoch": 36.98, "learning_rate": 2.382526871987768e-05, "loss": 0.0001, "step": 171550 }, { "epoch": 36.99, "learning_rate": 2.3820704775324715e-05, "loss": 0.0, "step": 171600 }, { "epoch": 37.0, "learning_rate": 2.381613958221217e-05, "loss": 0.0, "step": 171650 }, { "epoch": 37.01, "learning_rate": 2.3811573141186247e-05, "loss": 0.0, "step": 171700 }, { "epoch": 37.02, "learning_rate": 2.380700545289331e-05, "loss": 0.0, "step": 171750 }, { "epoch": 37.03, "learning_rate": 2.380243651797991e-05, "loss": 0.0, "step": 171800 }, { "epoch": 37.04, "learning_rate": 2.379786633709279e-05, "loss": 0.0, "step": 171850 }, { "epoch": 37.06, "learning_rate": 2.379329491087883e-05, "loss": 0.0, "step": 171900 }, { "epoch": 37.07, "learning_rate": 2.378872223998512e-05, "loss": 0.0, "step": 171950 }, { "epoch": 37.08, "learning_rate": 2.378414832505892e-05, "loss": 0.0, "step": 172000 }, { "epoch": 37.09, "learning_rate": 2.3779573166747656e-05, "loss": 0.0, "step": 172050 }, { "epoch": 37.1, "learning_rate": 2.377499676569894e-05, "loss": 0.0, "step": 172100 }, { "epoch": 37.11, "learning_rate": 2.3770419122560552e-05, "loss": 0.0, "step": 172150 }, { "epoch": 37.12, "learning_rate": 2.376584023798045e-05, "loss": 0.0, "step": 172200 }, { "epoch": 37.13, "learning_rate": 2.3761260112606775e-05, "loss": 0.0, "step": 172250 }, { "epoch": 37.14, "learning_rate": 2.3756678747087836e-05, "loss": 0.0, "step": 172300 }, { "epoch": 37.15, "learning_rate": 2.375209614207212e-05, "loss": 0.0, "step": 172350 }, { "epoch": 37.16, "learning_rate": 2.374751229820829e-05, "loss": 0.0, "step": 172400 }, { "epoch": 37.17, "learning_rate": 2.3742927216145173e-05, "loss": 0.0, "step": 172450 }, { "epoch": 37.18, "learning_rate": 2.3738340896531796e-05, "loss": 0.0, "step": 172500 }, { "epoch": 37.2, "learning_rate": 2.373375334001734e-05, "loss": 0.0, "step": 172550 }, { "epoch": 37.21, "learning_rate": 2.3729164547251166e-05, "loss": 0.0, "step": 172600 }, { "epoch": 37.22, "learning_rate": 2.3724574518882814e-05, "loss": 0.0, "step": 172650 }, { "epoch": 37.23, "learning_rate": 2.3719983255562e-05, "loss": 0.0, "step": 172700 }, { "epoch": 37.24, "learning_rate": 2.3715390757938606e-05, "loss": 0.0, "step": 172750 }, { "epoch": 37.25, "learning_rate": 2.3710797026662695e-05, "loss": 0.0001, "step": 172800 }, { "epoch": 37.26, "learning_rate": 2.3706202062384504e-05, "loss": 0.0, "step": 172850 }, { "epoch": 37.27, "learning_rate": 2.370160586575445e-05, "loss": 0.0, "step": 172900 }, { "epoch": 37.28, "learning_rate": 2.3697008437423114e-05, "loss": 0.0001, "step": 172950 }, { "epoch": 37.29, "learning_rate": 2.3692409778041255e-05, "loss": 0.0, "step": 173000 }, { "epoch": 37.3, "learning_rate": 2.3687809888259804e-05, "loss": 0.0, "step": 173050 }, { "epoch": 37.31, "learning_rate": 2.368320876872988e-05, "loss": 0.0, "step": 173100 }, { "epoch": 37.32, "learning_rate": 2.3678606420102765e-05, "loss": 0.0, "step": 173150 }, { "epoch": 37.34, "learning_rate": 2.3674002843029903e-05, "loss": 0.0, "step": 173200 }, { "epoch": 37.35, "learning_rate": 2.3669398038162936e-05, "loss": 0.0, "step": 173250 }, { "epoch": 37.36, "learning_rate": 2.3664792006153663e-05, "loss": 0.0001, "step": 173300 }, { "epoch": 37.37, "learning_rate": 2.3660184747654065e-05, "loss": 0.0, "step": 173350 }, { "epoch": 37.38, "learning_rate": 2.365557626331629e-05, "loss": 0.0, "step": 173400 }, { "epoch": 37.39, "learning_rate": 2.365096655379267e-05, "loss": 0.0, "step": 173450 }, { "epoch": 37.4, "learning_rate": 2.3646355619735693e-05, "loss": 0.0, "step": 173500 }, { "epoch": 37.41, "learning_rate": 2.364174346179804e-05, "loss": 0.0, "step": 173550 }, { "epoch": 37.42, "learning_rate": 2.3637130080632552e-05, "loss": 0.0, "step": 173600 }, { "epoch": 37.43, "learning_rate": 2.3632515476892247e-05, "loss": 0.0, "step": 173650 }, { "epoch": 37.44, "learning_rate": 2.3627899651230316e-05, "loss": 0.0, "step": 173700 }, { "epoch": 37.45, "learning_rate": 2.362328260430013e-05, "loss": 0.0, "step": 173750 }, { "epoch": 37.46, "learning_rate": 2.3618664336755217e-05, "loss": 0.0, "step": 173800 }, { "epoch": 37.48, "learning_rate": 2.3614044849249292e-05, "loss": 0.0, "step": 173850 }, { "epoch": 37.49, "learning_rate": 2.3609424142436237e-05, "loss": 0.0, "step": 173900 }, { "epoch": 37.5, "learning_rate": 2.3604802216970105e-05, "loss": 0.0, "step": 173950 }, { "epoch": 37.51, "learning_rate": 2.3600179073505127e-05, "loss": 0.0, "step": 174000 }, { "epoch": 37.52, "learning_rate": 2.3595554712695696e-05, "loss": 0.0, "step": 174050 }, { "epoch": 37.53, "learning_rate": 2.3590929135196397e-05, "loss": 0.0, "step": 174100 }, { "epoch": 37.54, "learning_rate": 2.358630234166196e-05, "loss": 0.0, "step": 174150 }, { "epoch": 37.55, "learning_rate": 2.3581674332747316e-05, "loss": 0.0, "step": 174200 }, { "epoch": 37.56, "learning_rate": 2.3577045109107545e-05, "loss": 0.0, "step": 174250 }, { "epoch": 37.57, "learning_rate": 2.3572414671397907e-05, "loss": 0.0, "step": 174300 }, { "epoch": 37.58, "learning_rate": 2.3567783020273844e-05, "loss": 0.0, "step": 174350 }, { "epoch": 37.59, "learning_rate": 2.3563150156390944e-05, "loss": 0.0001, "step": 174400 }, { "epoch": 37.61, "learning_rate": 2.3558516080404998e-05, "loss": 0.0, "step": 174450 }, { "epoch": 37.62, "learning_rate": 2.355388079297195e-05, "loss": 0.0, "step": 174500 }, { "epoch": 37.63, "learning_rate": 2.354924429474791e-05, "loss": 0.0, "step": 174550 }, { "epoch": 37.64, "learning_rate": 2.3544606586389184e-05, "loss": 0.0, "step": 174600 }, { "epoch": 37.65, "learning_rate": 2.353996766855222e-05, "loss": 0.0, "step": 174650 }, { "epoch": 37.66, "learning_rate": 2.353532754189366e-05, "loss": 0.0, "step": 174700 }, { "epoch": 37.67, "learning_rate": 2.3530686207070302e-05, "loss": 0.0, "step": 174750 }, { "epoch": 37.68, "learning_rate": 2.3526043664739127e-05, "loss": 0.0, "step": 174800 }, { "epoch": 37.69, "learning_rate": 2.3521399915557274e-05, "loss": 0.0, "step": 174850 }, { "epoch": 37.7, "learning_rate": 2.3516754960182064e-05, "loss": 0.0, "step": 174900 }, { "epoch": 37.71, "learning_rate": 2.3512108799270982e-05, "loss": 0.0, "step": 174950 }, { "epoch": 37.72, "learning_rate": 2.350746143348169e-05, "loss": 0.0, "step": 175000 }, { "epoch": 37.73, "learning_rate": 2.3502812863472018e-05, "loss": 0.0, "step": 175050 }, { "epoch": 37.75, "learning_rate": 2.3498163089899954e-05, "loss": 0.0, "step": 175100 }, { "epoch": 37.76, "learning_rate": 2.3493512113423682e-05, "loss": 0.0, "step": 175150 }, { "epoch": 37.77, "learning_rate": 2.348885993470153e-05, "loss": 0.0001, "step": 175200 }, { "epoch": 37.78, "learning_rate": 2.348420655439202e-05, "loss": 0.0, "step": 175250 }, { "epoch": 37.79, "learning_rate": 2.347955197315382e-05, "loss": 0.0, "step": 175300 }, { "epoch": 37.8, "learning_rate": 2.347489619164579e-05, "loss": 0.0, "step": 175350 }, { "epoch": 37.81, "learning_rate": 2.347023921052694e-05, "loss": 0.0, "step": 175400 }, { "epoch": 37.82, "learning_rate": 2.3465581030456468e-05, "loss": 0.0, "step": 175450 }, { "epoch": 37.83, "learning_rate": 2.346092165209373e-05, "loss": 0.0, "step": 175500 }, { "epoch": 37.84, "learning_rate": 2.3456261076098258e-05, "loss": 0.0, "step": 175550 }, { "epoch": 37.85, "learning_rate": 2.3451599303129753e-05, "loss": 0.0, "step": 175600 }, { "epoch": 37.86, "learning_rate": 2.3446936333848064e-05, "loss": 0.0001, "step": 175650 }, { "epoch": 37.87, "learning_rate": 2.3442272168913258e-05, "loss": 0.0, "step": 175700 }, { "epoch": 37.89, "learning_rate": 2.3437606808985515e-05, "loss": 0.0, "step": 175750 }, { "epoch": 37.9, "learning_rate": 2.343294025472523e-05, "loss": 0.0, "step": 175800 }, { "epoch": 37.91, "learning_rate": 2.3428272506792935e-05, "loss": 0.0, "step": 175850 }, { "epoch": 37.92, "learning_rate": 2.342360356584935e-05, "loss": 0.0, "step": 175900 }, { "epoch": 37.93, "learning_rate": 2.3418933432555363e-05, "loss": 0.0, "step": 175950 }, { "epoch": 37.94, "learning_rate": 2.3414262107572012e-05, "loss": 0.0, "step": 176000 }, { "epoch": 37.95, "learning_rate": 2.3409589591560526e-05, "loss": 0.0, "step": 176050 }, { "epoch": 37.96, "learning_rate": 2.3404915885182293e-05, "loss": 0.0, "step": 176100 }, { "epoch": 37.97, "learning_rate": 2.3400240989098868e-05, "loss": 0.0, "step": 176150 }, { "epoch": 37.98, "learning_rate": 2.3395564903971977e-05, "loss": 0.0, "step": 176200 }, { "epoch": 37.99, "learning_rate": 2.3390887630463517e-05, "loss": 0.0, "step": 176250 }, { "epoch": 38.0, "learning_rate": 2.3386209169235548e-05, "loss": 0.0, "step": 176300 }, { "epoch": 38.01, "learning_rate": 2.3381529520950296e-05, "loss": 0.0, "step": 176350 }, { "epoch": 38.03, "learning_rate": 2.3376848686270167e-05, "loss": 0.0, "step": 176400 }, { "epoch": 38.04, "learning_rate": 2.3372166665857726e-05, "loss": 0.0, "step": 176450 }, { "epoch": 38.05, "learning_rate": 2.3367483460375698e-05, "loss": 0.0, "step": 176500 }, { "epoch": 38.06, "learning_rate": 2.3362799070487e-05, "loss": 0.0, "step": 176550 }, { "epoch": 38.07, "learning_rate": 2.3358113496854685e-05, "loss": 0.0, "step": 176600 }, { "epoch": 38.08, "learning_rate": 2.3353426740142e-05, "loss": 0.0, "step": 176650 }, { "epoch": 38.09, "learning_rate": 2.334873880101235e-05, "loss": 0.0, "step": 176700 }, { "epoch": 38.1, "learning_rate": 2.33440496801293e-05, "loss": 0.0, "step": 176750 }, { "epoch": 38.11, "learning_rate": 2.3339359378156596e-05, "loss": 0.0, "step": 176800 }, { "epoch": 38.12, "learning_rate": 2.3334667895758143e-05, "loss": 0.0, "step": 176850 }, { "epoch": 38.13, "learning_rate": 2.3329975233598006e-05, "loss": 0.0, "step": 176900 }, { "epoch": 38.14, "learning_rate": 2.332528139234044e-05, "loss": 0.0, "step": 176950 }, { "epoch": 38.15, "learning_rate": 2.3320586372649838e-05, "loss": 0.0, "step": 177000 }, { "epoch": 38.17, "learning_rate": 2.3315890175190792e-05, "loss": 0.0, "step": 177050 }, { "epoch": 38.18, "learning_rate": 2.331119280062802e-05, "loss": 0.0, "step": 177100 }, { "epoch": 38.19, "learning_rate": 2.3306494249626445e-05, "loss": 0.0, "step": 177150 }, { "epoch": 38.2, "learning_rate": 2.3301794522851136e-05, "loss": 0.0, "step": 177200 }, { "epoch": 38.21, "learning_rate": 2.3297093620967336e-05, "loss": 0.0, "step": 177250 }, { "epoch": 38.22, "learning_rate": 2.329239154464045e-05, "loss": 0.0001, "step": 177300 }, { "epoch": 38.23, "learning_rate": 2.3287688294536054e-05, "loss": 0.0, "step": 177350 }, { "epoch": 38.24, "learning_rate": 2.328298387131988e-05, "loss": 0.0, "step": 177400 }, { "epoch": 38.25, "learning_rate": 2.3278278275657833e-05, "loss": 0.0, "step": 177450 }, { "epoch": 38.26, "learning_rate": 2.3273571508215995e-05, "loss": 0.0, "step": 177500 }, { "epoch": 38.27, "learning_rate": 2.3268863569660593e-05, "loss": 0.0001, "step": 177550 }, { "epoch": 38.28, "learning_rate": 2.3264154460658032e-05, "loss": 0.0, "step": 177600 }, { "epoch": 38.29, "learning_rate": 2.3259444181874884e-05, "loss": 0.0, "step": 177650 }, { "epoch": 38.31, "learning_rate": 2.3254732733977885e-05, "loss": 0.0, "step": 177700 }, { "epoch": 38.32, "learning_rate": 2.3250020117633924e-05, "loss": 0.0, "step": 177750 }, { "epoch": 38.33, "learning_rate": 2.324530633351007e-05, "loss": 0.0, "step": 177800 }, { "epoch": 38.34, "learning_rate": 2.3240591382273556e-05, "loss": 0.0, "step": 177850 }, { "epoch": 38.35, "learning_rate": 2.3235875264591782e-05, "loss": 0.0, "step": 177900 }, { "epoch": 38.36, "learning_rate": 2.3231157981132298e-05, "loss": 0.0, "step": 177950 }, { "epoch": 38.37, "learning_rate": 2.3226439532562833e-05, "loss": 0.0, "step": 178000 }, { "epoch": 38.38, "learning_rate": 2.322171991955128e-05, "loss": 0.0, "step": 178050 }, { "epoch": 38.39, "learning_rate": 2.321699914276569e-05, "loss": 0.0, "step": 178100 }, { "epoch": 38.4, "learning_rate": 2.321227720287429e-05, "loss": 0.0, "step": 178150 }, { "epoch": 38.41, "learning_rate": 2.3207554100545462e-05, "loss": 0.0, "step": 178200 }, { "epoch": 38.42, "learning_rate": 2.320282983644775e-05, "loss": 0.0, "step": 178250 }, { "epoch": 38.44, "learning_rate": 2.319810441124987e-05, "loss": 0.0, "step": 178300 }, { "epoch": 38.45, "learning_rate": 2.3193377825620708e-05, "loss": 0.0, "step": 178350 }, { "epoch": 38.46, "learning_rate": 2.318865008022929e-05, "loss": 0.0001, "step": 178400 }, { "epoch": 38.47, "learning_rate": 2.318392117574484e-05, "loss": 0.0, "step": 178450 }, { "epoch": 38.48, "learning_rate": 2.3179191112836713e-05, "loss": 0.0, "step": 178500 }, { "epoch": 38.49, "learning_rate": 2.3174459892174454e-05, "loss": 0.0, "step": 178550 }, { "epoch": 38.5, "learning_rate": 2.316972751442776e-05, "loss": 0.0, "step": 178600 }, { "epoch": 38.51, "learning_rate": 2.316499398026649e-05, "loss": 0.0, "step": 178650 }, { "epoch": 38.52, "learning_rate": 2.3160259290360665e-05, "loss": 0.0, "step": 178700 }, { "epoch": 38.53, "learning_rate": 2.315552344538049e-05, "loss": 0.0, "step": 178750 }, { "epoch": 38.54, "learning_rate": 2.31507864459963e-05, "loss": 0.0, "step": 178800 }, { "epoch": 38.55, "learning_rate": 2.3146048292878626e-05, "loss": 0.0, "step": 178850 }, { "epoch": 38.56, "learning_rate": 2.3141308986698133e-05, "loss": 0.0, "step": 178900 }, { "epoch": 38.58, "learning_rate": 2.3136568528125673e-05, "loss": 0.0, "step": 178950 }, { "epoch": 38.59, "learning_rate": 2.3131826917832256e-05, "loss": 0.0, "step": 179000 }, { "epoch": 38.6, "learning_rate": 2.3127084156489043e-05, "loss": 0.0, "step": 179050 }, { "epoch": 38.61, "learning_rate": 2.3122340244767365e-05, "loss": 0.0, "step": 179100 }, { "epoch": 38.62, "learning_rate": 2.311759518333872e-05, "loss": 0.0, "step": 179150 }, { "epoch": 38.63, "learning_rate": 2.3112848972874767e-05, "loss": 0.0, "step": 179200 }, { "epoch": 38.64, "learning_rate": 2.3108101614047322e-05, "loss": 0.0, "step": 179250 }, { "epoch": 38.65, "learning_rate": 2.3103353107528377e-05, "loss": 0.0001, "step": 179300 }, { "epoch": 38.66, "learning_rate": 2.3098603453990064e-05, "loss": 0.0, "step": 179350 }, { "epoch": 38.67, "learning_rate": 2.30938526541047e-05, "loss": 0.0, "step": 179400 }, { "epoch": 38.68, "learning_rate": 2.3089100708544747e-05, "loss": 0.0, "step": 179450 }, { "epoch": 38.69, "learning_rate": 2.308434761798285e-05, "loss": 0.0, "step": 179500 }, { "epoch": 38.7, "learning_rate": 2.3079593383091778e-05, "loss": 0.0, "step": 179550 }, { "epoch": 38.72, "learning_rate": 2.3074838004544517e-05, "loss": 0.0, "step": 179600 }, { "epoch": 38.73, "learning_rate": 2.3070081483014168e-05, "loss": 0.0, "step": 179650 }, { "epoch": 38.74, "learning_rate": 2.306532381917401e-05, "loss": 0.0, "step": 179700 }, { "epoch": 38.75, "learning_rate": 2.3060565013697493e-05, "loss": 0.0, "step": 179750 }, { "epoch": 38.76, "learning_rate": 2.3055805067258208e-05, "loss": 0.0, "step": 179800 }, { "epoch": 38.77, "learning_rate": 2.305104398052993e-05, "loss": 0.0, "step": 179850 }, { "epoch": 38.78, "learning_rate": 2.3046281754186577e-05, "loss": 0.0, "step": 179900 }, { "epoch": 38.79, "learning_rate": 2.3041518388902244e-05, "loss": 0.0, "step": 179950 }, { "epoch": 38.8, "learning_rate": 2.3036753885351174e-05, "loss": 0.0, "step": 180000 }, { "epoch": 38.81, "learning_rate": 2.303198824420778e-05, "loss": 0.0, "step": 180050 }, { "epoch": 38.82, "learning_rate": 2.3027221466146625e-05, "loss": 0.0001, "step": 180100 }, { "epoch": 38.83, "learning_rate": 2.302245355184245e-05, "loss": 0.0, "step": 180150 }, { "epoch": 38.84, "learning_rate": 2.3017684501970137e-05, "loss": 0.0, "step": 180200 }, { "epoch": 38.86, "learning_rate": 2.301291431720475e-05, "loss": 0.0, "step": 180250 }, { "epoch": 38.87, "learning_rate": 2.3008142998221497e-05, "loss": 0.0, "step": 180300 }, { "epoch": 38.88, "learning_rate": 2.3003370545695754e-05, "loss": 0.0, "step": 180350 }, { "epoch": 38.89, "learning_rate": 2.2998596960303048e-05, "loss": 0.0, "step": 180400 }, { "epoch": 38.9, "learning_rate": 2.2993822242719077e-05, "loss": 0.0, "step": 180450 }, { "epoch": 38.91, "learning_rate": 2.2989046393619713e-05, "loss": 0.0, "step": 180500 }, { "epoch": 38.92, "learning_rate": 2.2984269413680945e-05, "loss": 0.0, "step": 180550 }, { "epoch": 38.93, "learning_rate": 2.2979491303578968e-05, "loss": 0.0, "step": 180600 }, { "epoch": 38.94, "learning_rate": 2.2974712063990105e-05, "loss": 0.0, "step": 180650 }, { "epoch": 38.95, "learning_rate": 2.296993169559086e-05, "loss": 0.0, "step": 180700 }, { "epoch": 38.96, "learning_rate": 2.2965150199057887e-05, "loss": 0.0, "step": 180750 }, { "epoch": 38.97, "learning_rate": 2.2960367575067992e-05, "loss": 0.0, "step": 180800 }, { "epoch": 38.98, "learning_rate": 2.2955583824298164e-05, "loss": 0.0, "step": 180850 }, { "epoch": 39.0, "learning_rate": 2.2950798947425527e-05, "loss": 0.0, "step": 180900 }, { "epoch": 39.01, "learning_rate": 2.2946012945127375e-05, "loss": 0.0, "step": 180950 }, { "epoch": 39.02, "learning_rate": 2.294122581808116e-05, "loss": 0.0, "step": 181000 }, { "epoch": 39.03, "learning_rate": 2.2936437566964496e-05, "loss": 0.0, "step": 181050 }, { "epoch": 39.04, "learning_rate": 2.2931648192455153e-05, "loss": 0.0, "step": 181100 }, { "epoch": 39.05, "learning_rate": 2.2926857695231058e-05, "loss": 0.0, "step": 181150 }, { "epoch": 39.06, "learning_rate": 2.292206607597031e-05, "loss": 0.0, "step": 181200 }, { "epoch": 39.07, "learning_rate": 2.2917273335351146e-05, "loss": 0.0, "step": 181250 }, { "epoch": 39.08, "learning_rate": 2.2912479474051978e-05, "loss": 0.0, "step": 181300 }, { "epoch": 39.09, "learning_rate": 2.290768449275137e-05, "loss": 0.0, "step": 181350 }, { "epoch": 39.1, "learning_rate": 2.290288839212804e-05, "loss": 0.0, "step": 181400 }, { "epoch": 39.11, "learning_rate": 2.2898091172860877e-05, "loss": 0.0, "step": 181450 }, { "epoch": 39.12, "learning_rate": 2.2893292835628917e-05, "loss": 0.0, "step": 181500 }, { "epoch": 39.14, "learning_rate": 2.2888493381111365e-05, "loss": 0.0, "step": 181550 }, { "epoch": 39.15, "learning_rate": 2.288369280998757e-05, "loss": 0.0, "step": 181600 }, { "epoch": 39.16, "learning_rate": 2.2878891122937048e-05, "loss": 0.0, "step": 181650 }, { "epoch": 39.17, "learning_rate": 2.2874088320639476e-05, "loss": 0.0, "step": 181700 }, { "epoch": 39.18, "learning_rate": 2.286928440377468e-05, "loss": 0.0, "step": 181750 }, { "epoch": 39.19, "learning_rate": 2.286447937302265e-05, "loss": 0.0, "step": 181800 }, { "epoch": 39.2, "learning_rate": 2.2859673229063537e-05, "loss": 0.0, "step": 181850 }, { "epoch": 39.21, "learning_rate": 2.2854865972577634e-05, "loss": 0.0, "step": 181900 }, { "epoch": 39.22, "learning_rate": 2.2850057604245416e-05, "loss": 0.0, "step": 181950 }, { "epoch": 39.23, "learning_rate": 2.2845248124747483e-05, "loss": 0.0, "step": 182000 }, { "epoch": 39.24, "learning_rate": 2.2840437534764633e-05, "loss": 0.0, "step": 182050 }, { "epoch": 39.25, "learning_rate": 2.283562583497778e-05, "loss": 0.0001, "step": 182100 }, { "epoch": 39.26, "learning_rate": 2.283081302606802e-05, "loss": 0.0, "step": 182150 }, { "epoch": 39.28, "learning_rate": 2.2825999108716606e-05, "loss": 0.0, "step": 182200 }, { "epoch": 39.29, "learning_rate": 2.2821184083604934e-05, "loss": 0.0, "step": 182250 }, { "epoch": 39.3, "learning_rate": 2.281636795141457e-05, "loss": 0.0, "step": 182300 }, { "epoch": 39.31, "learning_rate": 2.2811550712827228e-05, "loss": 0.0, "step": 182350 }, { "epoch": 39.32, "learning_rate": 2.280673236852478e-05, "loss": 0.0, "step": 182400 }, { "epoch": 39.33, "learning_rate": 2.2801912919189264e-05, "loss": 0.0, "step": 182450 }, { "epoch": 39.34, "learning_rate": 2.2797092365502866e-05, "loss": 0.0001, "step": 182500 }, { "epoch": 39.35, "learning_rate": 2.279227070814792e-05, "loss": 0.0001, "step": 182550 }, { "epoch": 39.36, "learning_rate": 2.2787447947806934e-05, "loss": 0.0, "step": 182600 }, { "epoch": 39.37, "learning_rate": 2.278262408516256e-05, "loss": 0.0, "step": 182650 }, { "epoch": 39.38, "learning_rate": 2.277779912089761e-05, "loss": 0.0, "step": 182700 }, { "epoch": 39.39, "learning_rate": 2.277297305569505e-05, "loss": 0.0, "step": 182750 }, { "epoch": 39.41, "learning_rate": 2.2768145890238014e-05, "loss": 0.0, "step": 182800 }, { "epoch": 39.42, "learning_rate": 2.2763317625209766e-05, "loss": 0.0, "step": 182850 }, { "epoch": 39.43, "learning_rate": 2.2758488261293753e-05, "loss": 0.0, "step": 182900 }, { "epoch": 39.44, "learning_rate": 2.2753657799173554e-05, "loss": 0.0, "step": 182950 }, { "epoch": 39.45, "learning_rate": 2.274882623953292e-05, "loss": 0.0, "step": 183000 }, { "epoch": 39.46, "learning_rate": 2.274399358305576e-05, "loss": 0.0, "step": 183050 }, { "epoch": 39.47, "learning_rate": 2.2739159830426116e-05, "loss": 0.0, "step": 183100 }, { "epoch": 39.48, "learning_rate": 2.2734324982328215e-05, "loss": 0.0, "step": 183150 }, { "epoch": 39.49, "learning_rate": 2.272948903944641e-05, "loss": 0.0, "step": 183200 }, { "epoch": 39.5, "learning_rate": 2.2724652002465227e-05, "loss": 0.0, "step": 183250 }, { "epoch": 39.51, "learning_rate": 2.271981387206935e-05, "loss": 0.0, "step": 183300 }, { "epoch": 39.52, "learning_rate": 2.2714974648943597e-05, "loss": 0.0, "step": 183350 }, { "epoch": 39.53, "learning_rate": 2.271013433377296e-05, "loss": 0.0, "step": 183400 }, { "epoch": 39.55, "learning_rate": 2.2705292927242586e-05, "loss": 0.0, "step": 183450 }, { "epoch": 39.56, "learning_rate": 2.270045043003776e-05, "loss": 0.0, "step": 183500 }, { "epoch": 39.57, "learning_rate": 2.269560684284394e-05, "loss": 0.0, "step": 183550 }, { "epoch": 39.58, "learning_rate": 2.2690762166346716e-05, "loss": 0.0, "step": 183600 }, { "epoch": 39.59, "learning_rate": 2.2685916401231866e-05, "loss": 0.0, "step": 183650 }, { "epoch": 39.6, "learning_rate": 2.2681069548185284e-05, "loss": 0.0, "step": 183700 }, { "epoch": 39.61, "learning_rate": 2.2676221607893045e-05, "loss": 0.0, "step": 183750 }, { "epoch": 39.62, "learning_rate": 2.2671372581041367e-05, "loss": 0.0001, "step": 183800 }, { "epoch": 39.63, "learning_rate": 2.266652246831662e-05, "loss": 0.0, "step": 183850 }, { "epoch": 39.64, "learning_rate": 2.266167127040534e-05, "loss": 0.0, "step": 183900 }, { "epoch": 39.65, "learning_rate": 2.2656818987994194e-05, "loss": 0.0, "step": 183950 }, { "epoch": 39.66, "learning_rate": 2.265196562177003e-05, "loss": 0.0, "step": 184000 }, { "epoch": 39.67, "learning_rate": 2.2647111172419827e-05, "loss": 0.0, "step": 184050 }, { "epoch": 39.69, "learning_rate": 2.2642255640630736e-05, "loss": 0.0, "step": 184100 }, { "epoch": 39.7, "learning_rate": 2.2637399027090043e-05, "loss": 0.0, "step": 184150 }, { "epoch": 39.71, "learning_rate": 2.2632541332485195e-05, "loss": 0.0, "step": 184200 }, { "epoch": 39.72, "learning_rate": 2.26276825575038e-05, "loss": 0.0, "step": 184250 }, { "epoch": 39.73, "learning_rate": 2.2622822702833597e-05, "loss": 0.0, "step": 184300 }, { "epoch": 39.74, "learning_rate": 2.261796176916251e-05, "loss": 0.0, "step": 184350 }, { "epoch": 39.75, "learning_rate": 2.261309975717859e-05, "loss": 0.0, "step": 184400 }, { "epoch": 39.76, "learning_rate": 2.2608236667570044e-05, "loss": 0.0, "step": 184450 }, { "epoch": 39.77, "learning_rate": 2.2603372501025245e-05, "loss": 0.0001, "step": 184500 }, { "epoch": 39.78, "learning_rate": 2.25985072582327e-05, "loss": 0.0, "step": 184550 }, { "epoch": 39.79, "learning_rate": 2.2593640939881093e-05, "loss": 0.0, "step": 184600 }, { "epoch": 39.8, "learning_rate": 2.2588773546659224e-05, "loss": 0.0, "step": 184650 }, { "epoch": 39.81, "learning_rate": 2.2583905079256083e-05, "loss": 0.0, "step": 184700 }, { "epoch": 39.83, "learning_rate": 2.2579035538360793e-05, "loss": 0.0, "step": 184750 }, { "epoch": 39.84, "learning_rate": 2.2574164924662622e-05, "loss": 0.0, "step": 184800 }, { "epoch": 39.85, "learning_rate": 2.2569293238851014e-05, "loss": 0.0, "step": 184850 }, { "epoch": 39.86, "learning_rate": 2.2564420481615532e-05, "loss": 0.0, "step": 184900 }, { "epoch": 39.87, "learning_rate": 2.2559546653645922e-05, "loss": 0.0, "step": 184950 }, { "epoch": 39.88, "learning_rate": 2.2554671755632068e-05, "loss": 0.0, "step": 185000 }, { "epoch": 39.89, "learning_rate": 2.2549795788264e-05, "loss": 0.0, "step": 185050 }, { "epoch": 39.9, "learning_rate": 2.2544918752231906e-05, "loss": 0.0, "step": 185100 }, { "epoch": 39.91, "learning_rate": 2.2540040648226126e-05, "loss": 0.0, "step": 185150 }, { "epoch": 39.92, "learning_rate": 2.2535161476937145e-05, "loss": 0.0, "step": 185200 }, { "epoch": 39.93, "learning_rate": 2.2530281239055613e-05, "loss": 0.0, "step": 185250 }, { "epoch": 39.94, "learning_rate": 2.2525399935272313e-05, "loss": 0.0, "step": 185300 }, { "epoch": 39.95, "learning_rate": 2.2520517566278194e-05, "loss": 0.0, "step": 185350 }, { "epoch": 39.97, "learning_rate": 2.251563413276434e-05, "loss": 0.0, "step": 185400 }, { "epoch": 39.98, "learning_rate": 2.2510749635422005e-05, "loss": 0.0, "step": 185450 }, { "epoch": 39.99, "learning_rate": 2.250586407494257e-05, "loss": 0.0, "step": 185500 }, { "epoch": 40.0, "learning_rate": 2.250097745201759e-05, "loss": 0.0, "step": 185550 }, { "epoch": 40.01, "learning_rate": 2.249608976733877e-05, "loss": 0.0, "step": 185600 }, { "epoch": 40.02, "learning_rate": 2.2491201021597933e-05, "loss": 0.0, "step": 185650 }, { "epoch": 40.03, "learning_rate": 2.248631121548709e-05, "loss": 0.0, "step": 185700 }, { "epoch": 40.04, "learning_rate": 2.2481420349698385e-05, "loss": 0.0, "step": 185750 }, { "epoch": 40.05, "learning_rate": 2.2476528424924107e-05, "loss": 0.0, "step": 185800 }, { "epoch": 40.06, "learning_rate": 2.247163544185671e-05, "loss": 0.0, "step": 185850 }, { "epoch": 40.07, "learning_rate": 2.2466741401188785e-05, "loss": 0.0001, "step": 185900 }, { "epoch": 40.08, "learning_rate": 2.2461846303613076e-05, "loss": 0.0, "step": 185950 }, { "epoch": 40.09, "learning_rate": 2.2456950149822485e-05, "loss": 0.0, "step": 186000 }, { "epoch": 40.11, "learning_rate": 2.245205294051005e-05, "loss": 0.0, "step": 186050 }, { "epoch": 40.12, "learning_rate": 2.2447154676368963e-05, "loss": 0.0, "step": 186100 }, { "epoch": 40.13, "learning_rate": 2.244225535809257e-05, "loss": 0.0, "step": 186150 }, { "epoch": 40.14, "learning_rate": 2.243735498637437e-05, "loss": 0.0, "step": 186200 }, { "epoch": 40.15, "learning_rate": 2.2432453561907993e-05, "loss": 0.0, "step": 186250 }, { "epoch": 40.16, "learning_rate": 2.2427551085387234e-05, "loss": 0.0, "step": 186300 }, { "epoch": 40.17, "learning_rate": 2.2422647557506037e-05, "loss": 0.0, "step": 186350 }, { "epoch": 40.18, "learning_rate": 2.241774297895848e-05, "loss": 0.0, "step": 186400 }, { "epoch": 40.19, "learning_rate": 2.2412837350438812e-05, "loss": 0.0, "step": 186450 }, { "epoch": 40.2, "learning_rate": 2.2407930672641408e-05, "loss": 0.0, "step": 186500 }, { "epoch": 40.21, "learning_rate": 2.240302294626081e-05, "loss": 0.0, "step": 186550 }, { "epoch": 40.22, "learning_rate": 2.2398114171991693e-05, "loss": 0.0, "step": 186600 }, { "epoch": 40.23, "learning_rate": 2.239320435052889e-05, "loss": 0.0, "step": 186650 }, { "epoch": 40.25, "learning_rate": 2.2388293482567387e-05, "loss": 0.0, "step": 186700 }, { "epoch": 40.26, "learning_rate": 2.2383381568802304e-05, "loss": 0.0, "step": 186750 }, { "epoch": 40.27, "learning_rate": 2.2378468609928918e-05, "loss": 0.0, "step": 186800 }, { "epoch": 40.28, "learning_rate": 2.2373554606642652e-05, "loss": 0.0, "step": 186850 }, { "epoch": 40.29, "learning_rate": 2.2368639559639078e-05, "loss": 0.0, "step": 186900 }, { "epoch": 40.3, "learning_rate": 2.2363723469613908e-05, "loss": 0.0, "step": 186950 }, { "epoch": 40.31, "learning_rate": 2.2358806337263024e-05, "loss": 0.0, "step": 187000 }, { "epoch": 40.32, "learning_rate": 2.235388816328242e-05, "loss": 0.0, "step": 187050 }, { "epoch": 40.33, "learning_rate": 2.234896894836827e-05, "loss": 0.0, "step": 187100 }, { "epoch": 40.34, "learning_rate": 2.234404869321688e-05, "loss": 0.0, "step": 187150 }, { "epoch": 40.35, "learning_rate": 2.2339127398524707e-05, "loss": 0.0001, "step": 187200 }, { "epoch": 40.36, "learning_rate": 2.233420506498835e-05, "loss": 0.0, "step": 187250 }, { "epoch": 40.38, "learning_rate": 2.2329281693304566e-05, "loss": 0.0, "step": 187300 }, { "epoch": 40.39, "learning_rate": 2.232435728417024e-05, "loss": 0.0, "step": 187350 }, { "epoch": 40.4, "learning_rate": 2.231943183828243e-05, "loss": 0.0, "step": 187400 }, { "epoch": 40.41, "learning_rate": 2.231450535633832e-05, "loss": 0.0, "step": 187450 }, { "epoch": 40.42, "learning_rate": 2.2309577839035246e-05, "loss": 0.0, "step": 187500 }, { "epoch": 40.43, "learning_rate": 2.2304649287070695e-05, "loss": 0.0, "step": 187550 }, { "epoch": 40.44, "learning_rate": 2.2299719701142294e-05, "loss": 0.0, "step": 187600 }, { "epoch": 40.45, "learning_rate": 2.229478908194782e-05, "loss": 0.0, "step": 187650 }, { "epoch": 40.46, "learning_rate": 2.22898574301852e-05, "loss": 0.0, "step": 187700 }, { "epoch": 40.47, "learning_rate": 2.2284924746552498e-05, "loss": 0.0, "step": 187750 }, { "epoch": 40.48, "learning_rate": 2.2279991031747935e-05, "loss": 0.0001, "step": 187800 }, { "epoch": 40.49, "learning_rate": 2.2275056286469862e-05, "loss": 0.0, "step": 187850 }, { "epoch": 40.5, "learning_rate": 2.22701205114168e-05, "loss": 0.0001, "step": 187900 }, { "epoch": 40.52, "learning_rate": 2.226518370728739e-05, "loss": 0.0, "step": 187950 }, { "epoch": 40.53, "learning_rate": 2.226024587478043e-05, "loss": 0.0, "step": 188000 }, { "epoch": 40.54, "learning_rate": 2.2255307014594875e-05, "loss": 0.0, "step": 188050 }, { "epoch": 40.55, "learning_rate": 2.22503671274298e-05, "loss": 0.0, "step": 188100 }, { "epoch": 40.56, "learning_rate": 2.2245426213984456e-05, "loss": 0.0, "step": 188150 }, { "epoch": 40.57, "learning_rate": 2.224048427495821e-05, "loss": 0.0, "step": 188200 }, { "epoch": 40.58, "learning_rate": 2.2235541311050593e-05, "loss": 0.0, "step": 188250 }, { "epoch": 40.59, "learning_rate": 2.2230597322961274e-05, "loss": 0.0, "step": 188300 }, { "epoch": 40.6, "learning_rate": 2.222565231139007e-05, "loss": 0.0, "step": 188350 }, { "epoch": 40.61, "learning_rate": 2.2220706277036934e-05, "loss": 0.0, "step": 188400 }, { "epoch": 40.62, "learning_rate": 2.221575922060198e-05, "loss": 0.0, "step": 188450 }, { "epoch": 40.63, "learning_rate": 2.2210811142785456e-05, "loss": 0.0, "step": 188500 }, { "epoch": 40.64, "learning_rate": 2.2205862044287747e-05, "loss": 0.0, "step": 188550 }, { "epoch": 40.66, "learning_rate": 2.2200911925809406e-05, "loss": 0.0, "step": 188600 }, { "epoch": 40.67, "learning_rate": 2.2195960788051105e-05, "loss": 0.0, "step": 188650 }, { "epoch": 40.68, "learning_rate": 2.219100863171368e-05, "loss": 0.0, "step": 188700 }, { "epoch": 40.69, "learning_rate": 2.2186055457498088e-05, "loss": 0.0, "step": 188750 }, { "epoch": 40.7, "learning_rate": 2.218110126610546e-05, "loss": 0.0, "step": 188800 }, { "epoch": 40.71, "learning_rate": 2.2176146058237047e-05, "loss": 0.0, "step": 188850 }, { "epoch": 40.72, "learning_rate": 2.2171189834594257e-05, "loss": 0.0, "step": 188900 }, { "epoch": 40.73, "learning_rate": 2.216623259587863e-05, "loss": 0.0, "step": 188950 }, { "epoch": 40.74, "learning_rate": 2.2161274342791872e-05, "loss": 0.0, "step": 189000 }, { "epoch": 40.75, "learning_rate": 2.2156315076035797e-05, "loss": 0.0, "step": 189050 }, { "epoch": 40.76, "learning_rate": 2.2151354796312394e-05, "loss": 0.0, "step": 189100 }, { "epoch": 40.77, "learning_rate": 2.214639350432379e-05, "loss": 0.0, "step": 189150 }, { "epoch": 40.78, "learning_rate": 2.214143120077224e-05, "loss": 0.0, "step": 189200 }, { "epoch": 40.8, "learning_rate": 2.2136467886360152e-05, "loss": 0.0, "step": 189250 }, { "epoch": 40.81, "learning_rate": 2.2131503561790084e-05, "loss": 0.0, "step": 189300 }, { "epoch": 40.82, "learning_rate": 2.2126538227764726e-05, "loss": 0.0, "step": 189350 }, { "epoch": 40.83, "learning_rate": 2.2121571884986907e-05, "loss": 0.0, "step": 189400 }, { "epoch": 40.84, "learning_rate": 2.2116604534159622e-05, "loss": 0.0001, "step": 189450 }, { "epoch": 40.85, "learning_rate": 2.211163617598598e-05, "loss": 0.0, "step": 189500 }, { "epoch": 40.86, "learning_rate": 2.2106666811169254e-05, "loss": 0.0001, "step": 189550 }, { "epoch": 40.87, "learning_rate": 2.2101696440412848e-05, "loss": 0.0, "step": 189600 }, { "epoch": 40.88, "learning_rate": 2.2096725064420308e-05, "loss": 0.0, "step": 189650 }, { "epoch": 40.89, "learning_rate": 2.209175268389533e-05, "loss": 0.0001, "step": 189700 }, { "epoch": 40.9, "learning_rate": 2.2086779299541753e-05, "loss": 0.0, "step": 189750 }, { "epoch": 40.91, "learning_rate": 2.2081804912063546e-05, "loss": 0.0, "step": 189800 }, { "epoch": 40.92, "learning_rate": 2.207682952216483e-05, "loss": 0.0, "step": 189850 }, { "epoch": 40.94, "learning_rate": 2.207185313054986e-05, "loss": 0.0, "step": 189900 }, { "epoch": 40.95, "learning_rate": 2.2066875737923047e-05, "loss": 0.0, "step": 189950 }, { "epoch": 40.96, "learning_rate": 2.2061897344988928e-05, "loss": 0.0, "step": 190000 }, { "epoch": 40.97, "learning_rate": 2.205691795245219e-05, "loss": 0.0, "step": 190050 }, { "epoch": 40.98, "learning_rate": 2.2051937561017662e-05, "loss": 0.0, "step": 190100 }, { "epoch": 40.99, "learning_rate": 2.204695617139031e-05, "loss": 0.0, "step": 190150 }, { "epoch": 41.0, "learning_rate": 2.2041973784275242e-05, "loss": 0.0, "step": 190200 }, { "epoch": 41.01, "learning_rate": 2.203699040037771e-05, "loss": 0.0, "step": 190250 }, { "epoch": 41.02, "learning_rate": 2.2032006020403108e-05, "loss": 0.0, "step": 190300 }, { "epoch": 41.03, "learning_rate": 2.2027020645056968e-05, "loss": 0.0, "step": 190350 }, { "epoch": 41.04, "learning_rate": 2.2022034275044954e-05, "loss": 0.0, "step": 190400 }, { "epoch": 41.05, "learning_rate": 2.2017046911072896e-05, "loss": 0.0, "step": 190450 }, { "epoch": 41.06, "learning_rate": 2.201205855384674e-05, "loss": 0.0, "step": 190500 }, { "epoch": 41.08, "learning_rate": 2.200706920407258e-05, "loss": 0.0, "step": 190550 }, { "epoch": 41.09, "learning_rate": 2.2002078862456662e-05, "loss": 0.0, "step": 190600 }, { "epoch": 41.1, "learning_rate": 2.1997087529705346e-05, "loss": 0.0, "step": 190650 }, { "epoch": 41.11, "learning_rate": 2.1992095206525168e-05, "loss": 0.0, "step": 190700 }, { "epoch": 41.12, "learning_rate": 2.1987101893622766e-05, "loss": 0.0, "step": 190750 }, { "epoch": 41.13, "learning_rate": 2.198210759170495e-05, "loss": 0.0, "step": 190800 }, { "epoch": 41.14, "learning_rate": 2.1977112301478656e-05, "loss": 0.0, "step": 190850 }, { "epoch": 41.15, "learning_rate": 2.197211602365096e-05, "loss": 0.0, "step": 190900 }, { "epoch": 41.16, "learning_rate": 2.1967118758929074e-05, "loss": 0.0, "step": 190950 }, { "epoch": 41.17, "learning_rate": 2.196212050802036e-05, "loss": 0.0, "step": 191000 }, { "epoch": 41.18, "learning_rate": 2.1957121271632313e-05, "loss": 0.0, "step": 191050 }, { "epoch": 41.19, "learning_rate": 2.1952121050472566e-05, "loss": 0.0, "step": 191100 }, { "epoch": 41.21, "learning_rate": 2.1947119845248897e-05, "loss": 0.0, "step": 191150 }, { "epoch": 41.22, "learning_rate": 2.1942117656669215e-05, "loss": 0.0, "step": 191200 }, { "epoch": 41.23, "learning_rate": 2.1937114485441584e-05, "loss": 0.0, "step": 191250 }, { "epoch": 41.24, "learning_rate": 2.1932110332274187e-05, "loss": 0.0, "step": 191300 }, { "epoch": 41.25, "learning_rate": 2.1927105197875356e-05, "loss": 0.0, "step": 191350 }, { "epoch": 41.26, "learning_rate": 2.1922099082953564e-05, "loss": 0.0, "step": 191400 }, { "epoch": 41.27, "learning_rate": 2.1917091988217425e-05, "loss": 0.0, "step": 191450 }, { "epoch": 41.28, "learning_rate": 2.1912083914375675e-05, "loss": 0.0, "step": 191500 }, { "epoch": 41.29, "learning_rate": 2.1907074862137216e-05, "loss": 0.0, "step": 191550 }, { "epoch": 41.3, "learning_rate": 2.1902064832211056e-05, "loss": 0.0, "step": 191600 }, { "epoch": 41.31, "learning_rate": 2.189705382530637e-05, "loss": 0.0, "step": 191650 }, { "epoch": 41.32, "learning_rate": 2.189204184213246e-05, "loss": 0.0, "step": 191700 }, { "epoch": 41.33, "learning_rate": 2.1887028883398757e-05, "loss": 0.0, "step": 191750 }, { "epoch": 41.35, "learning_rate": 2.188201494981485e-05, "loss": 0.0, "step": 191800 }, { "epoch": 41.36, "learning_rate": 2.1877000042090447e-05, "loss": 0.0001, "step": 191850 }, { "epoch": 41.37, "learning_rate": 2.1871984160935402e-05, "loss": 0.0, "step": 191900 }, { "epoch": 41.38, "learning_rate": 2.1866967307059712e-05, "loss": 0.0, "step": 191950 }, { "epoch": 41.39, "learning_rate": 2.1861949481173504e-05, "loss": 0.0, "step": 192000 }, { "epoch": 41.4, "learning_rate": 2.185693068398704e-05, "loss": 0.0, "step": 192050 }, { "epoch": 41.41, "learning_rate": 2.185191091621073e-05, "loss": 0.0, "step": 192100 }, { "epoch": 41.42, "learning_rate": 2.1846890178555113e-05, "loss": 0.0, "step": 192150 }, { "epoch": 41.43, "learning_rate": 2.184186847173087e-05, "loss": 0.0, "step": 192200 }, { "epoch": 41.44, "learning_rate": 2.1836845796448814e-05, "loss": 0.0, "step": 192250 }, { "epoch": 41.45, "learning_rate": 2.183182215341991e-05, "loss": 0.0, "step": 192300 }, { "epoch": 41.46, "learning_rate": 2.1826797543355228e-05, "loss": 0.0, "step": 192350 }, { "epoch": 41.47, "learning_rate": 2.1821771966966015e-05, "loss": 0.0, "step": 192400 }, { "epoch": 41.49, "learning_rate": 2.181674542496362e-05, "loss": 0.0, "step": 192450 }, { "epoch": 41.5, "learning_rate": 2.1811717918059553e-05, "loss": 0.0, "step": 192500 }, { "epoch": 41.51, "learning_rate": 2.1806689446965448e-05, "loss": 0.0, "step": 192550 }, { "epoch": 41.52, "learning_rate": 2.1801660012393077e-05, "loss": 0.0, "step": 192600 }, { "epoch": 41.53, "learning_rate": 2.1796629615054357e-05, "loss": 0.0, "step": 192650 }, { "epoch": 41.54, "learning_rate": 2.179159825566132e-05, "loss": 0.0, "step": 192700 }, { "epoch": 41.55, "learning_rate": 2.1786565934926168e-05, "loss": 0.0, "step": 192750 }, { "epoch": 41.56, "learning_rate": 2.1781532653561205e-05, "loss": 0.0, "step": 192800 }, { "epoch": 41.57, "learning_rate": 2.177649841227889e-05, "loss": 0.0, "step": 192850 }, { "epoch": 41.58, "learning_rate": 2.1771463211791812e-05, "loss": 0.0, "step": 192900 }, { "epoch": 41.59, "learning_rate": 2.17664270528127e-05, "loss": 0.0, "step": 192950 }, { "epoch": 41.6, "learning_rate": 2.1761389936054414e-05, "loss": 0.0, "step": 193000 }, { "epoch": 41.61, "learning_rate": 2.1756351862229955e-05, "loss": 0.0, "step": 193050 }, { "epoch": 41.63, "learning_rate": 2.1751312832052447e-05, "loss": 0.0, "step": 193100 }, { "epoch": 41.64, "learning_rate": 2.1746272846235173e-05, "loss": 0.0, "step": 193150 }, { "epoch": 41.65, "learning_rate": 2.174123190549152e-05, "loss": 0.0, "step": 193200 }, { "epoch": 41.66, "learning_rate": 2.173619001053504e-05, "loss": 0.0, "step": 193250 }, { "epoch": 41.67, "learning_rate": 2.1731147162079397e-05, "loss": 0.0, "step": 193300 }, { "epoch": 41.68, "learning_rate": 2.1726103360838407e-05, "loss": 0.0, "step": 193350 }, { "epoch": 41.69, "learning_rate": 2.1721058607526012e-05, "loss": 0.0, "step": 193400 }, { "epoch": 41.7, "learning_rate": 2.171601290285629e-05, "loss": 0.0, "step": 193450 }, { "epoch": 41.71, "learning_rate": 2.1710966247543454e-05, "loss": 0.0, "step": 193500 }, { "epoch": 41.72, "learning_rate": 2.170591864230185e-05, "loss": 0.0, "step": 193550 }, { "epoch": 41.73, "learning_rate": 2.1700870087845965e-05, "loss": 0.0, "step": 193600 }, { "epoch": 41.74, "learning_rate": 2.1695820584890407e-05, "loss": 0.0, "step": 193650 }, { "epoch": 41.75, "learning_rate": 2.1690770134149934e-05, "loss": 0.0, "step": 193700 }, { "epoch": 41.77, "learning_rate": 2.168571873633943e-05, "loss": 0.0, "step": 193750 }, { "epoch": 41.78, "learning_rate": 2.168066639217391e-05, "loss": 0.0, "step": 193800 }, { "epoch": 41.79, "learning_rate": 2.167561310236853e-05, "loss": 0.0, "step": 193850 }, { "epoch": 41.8, "learning_rate": 2.167055886763858e-05, "loss": 0.0, "step": 193900 }, { "epoch": 41.81, "learning_rate": 2.166550368869947e-05, "loss": 0.0, "step": 193950 }, { "epoch": 41.82, "learning_rate": 2.1660447566266766e-05, "loss": 0.0001, "step": 194000 }, { "epoch": 41.83, "learning_rate": 2.1655390501056145e-05, "loss": 0.0, "step": 194050 }, { "epoch": 41.84, "learning_rate": 2.165033249378344e-05, "loss": 0.0, "step": 194100 }, { "epoch": 41.85, "learning_rate": 2.1645273545164595e-05, "loss": 0.0, "step": 194150 }, { "epoch": 41.86, "learning_rate": 2.1640213655915706e-05, "loss": 0.0, "step": 194200 }, { "epoch": 41.87, "learning_rate": 2.1635152826752987e-05, "loss": 0.0, "step": 194250 }, { "epoch": 41.88, "learning_rate": 2.1630091058392798e-05, "loss": 0.0, "step": 194300 }, { "epoch": 41.89, "learning_rate": 2.1625028351551623e-05, "loss": 0.0, "step": 194350 }, { "epoch": 41.91, "learning_rate": 2.161996470694608e-05, "loss": 0.0, "step": 194400 }, { "epoch": 41.92, "learning_rate": 2.161490012529292e-05, "loss": 0.0, "step": 194450 }, { "epoch": 41.93, "learning_rate": 2.1609834607309037e-05, "loss": 0.0, "step": 194500 }, { "epoch": 41.94, "learning_rate": 2.1604768153711438e-05, "loss": 0.0, "step": 194550 }, { "epoch": 41.95, "learning_rate": 2.1599700765217282e-05, "loss": 0.0, "step": 194600 }, { "epoch": 41.96, "learning_rate": 2.1594632442543845e-05, "loss": 0.0, "step": 194650 }, { "epoch": 41.97, "learning_rate": 2.1589563186408546e-05, "loss": 0.0, "step": 194700 }, { "epoch": 41.98, "learning_rate": 2.1584492997528928e-05, "loss": 0.0, "step": 194750 }, { "epoch": 41.99, "learning_rate": 2.157942187662267e-05, "loss": 0.0, "step": 194800 }, { "epoch": 42.0, "learning_rate": 2.157434982440759e-05, "loss": 0.0, "step": 194850 }, { "epoch": 42.01, "learning_rate": 2.156927684160162e-05, "loss": 0.0, "step": 194900 }, { "epoch": 42.02, "learning_rate": 2.156420292892284e-05, "loss": 0.0, "step": 194950 }, { "epoch": 42.03, "learning_rate": 2.1559128087089453e-05, "loss": 0.0, "step": 195000 }, { "epoch": 42.05, "learning_rate": 2.1554052316819802e-05, "loss": 0.0, "step": 195050 }, { "epoch": 42.06, "learning_rate": 2.1548975618832352e-05, "loss": 0.0, "step": 195100 }, { "epoch": 42.07, "learning_rate": 2.1543897993845703e-05, "loss": 0.0, "step": 195150 }, { "epoch": 42.08, "learning_rate": 2.1538819442578587e-05, "loss": 0.0001, "step": 195200 }, { "epoch": 42.09, "learning_rate": 2.1533739965749867e-05, "loss": 0.0, "step": 195250 }, { "epoch": 42.1, "learning_rate": 2.1528659564078536e-05, "loss": 0.0, "step": 195300 }, { "epoch": 42.11, "learning_rate": 2.1523578238283715e-05, "loss": 0.0, "step": 195350 }, { "epoch": 42.12, "learning_rate": 2.151849598908467e-05, "loss": 0.0, "step": 195400 }, { "epoch": 42.13, "learning_rate": 2.151341281720077e-05, "loss": 0.0, "step": 195450 }, { "epoch": 42.14, "learning_rate": 2.150832872335155e-05, "loss": 0.0, "step": 195500 }, { "epoch": 42.15, "learning_rate": 2.150324370825664e-05, "loss": 0.0, "step": 195550 }, { "epoch": 42.16, "learning_rate": 2.1498157772635834e-05, "loss": 0.0, "step": 195600 }, { "epoch": 42.18, "learning_rate": 2.1493070917209024e-05, "loss": 0.0001, "step": 195650 }, { "epoch": 42.19, "learning_rate": 2.1487983142696258e-05, "loss": 0.0, "step": 195700 }, { "epoch": 42.2, "learning_rate": 2.14828944498177e-05, "loss": 0.0, "step": 195750 }, { "epoch": 42.21, "learning_rate": 2.1477804839293652e-05, "loss": 0.0, "step": 195800 }, { "epoch": 42.22, "learning_rate": 2.1472714311844538e-05, "loss": 0.0, "step": 195850 }, { "epoch": 42.23, "learning_rate": 2.1467622868190917e-05, "loss": 0.0, "step": 195900 }, { "epoch": 42.24, "learning_rate": 2.1462530509053478e-05, "loss": 0.0, "step": 195950 }, { "epoch": 42.25, "learning_rate": 2.1457437235153033e-05, "loss": 0.0, "step": 196000 }, { "epoch": 42.26, "learning_rate": 2.1452343047210536e-05, "loss": 0.0, "step": 196050 }, { "epoch": 42.27, "learning_rate": 2.144724794594705e-05, "loss": 0.0, "step": 196100 }, { "epoch": 42.28, "learning_rate": 2.14421519320838e-05, "loss": 0.0, "step": 196150 }, { "epoch": 42.29, "learning_rate": 2.14370550063421e-05, "loss": 0.0, "step": 196200 }, { "epoch": 42.3, "learning_rate": 2.143195716944343e-05, "loss": 0.0, "step": 196250 }, { "epoch": 42.32, "learning_rate": 2.1426858422109367e-05, "loss": 0.0, "step": 196300 }, { "epoch": 42.33, "learning_rate": 2.1421758765061643e-05, "loss": 0.0, "step": 196350 }, { "epoch": 42.34, "learning_rate": 2.1416658199022103e-05, "loss": 0.0, "step": 196400 }, { "epoch": 42.35, "learning_rate": 2.1411556724712733e-05, "loss": 0.0, "step": 196450 }, { "epoch": 42.36, "learning_rate": 2.140645434285563e-05, "loss": 0.0, "step": 196500 }, { "epoch": 42.37, "learning_rate": 2.140135105417304e-05, "loss": 0.0, "step": 196550 }, { "epoch": 42.38, "learning_rate": 2.1396246859387315e-05, "loss": 0.0, "step": 196600 }, { "epoch": 42.39, "learning_rate": 2.139114175922096e-05, "loss": 0.0001, "step": 196650 }, { "epoch": 42.4, "learning_rate": 2.1386035754396583e-05, "loss": 0.0, "step": 196700 }, { "epoch": 42.41, "learning_rate": 2.1380928845636943e-05, "loss": 0.0, "step": 196750 }, { "epoch": 42.42, "learning_rate": 2.137582103366491e-05, "loss": 0.0, "step": 196800 }, { "epoch": 42.43, "learning_rate": 2.1370712319203493e-05, "loss": 0.0, "step": 196850 }, { "epoch": 42.44, "learning_rate": 2.136560270297582e-05, "loss": 0.0, "step": 196900 }, { "epoch": 42.46, "learning_rate": 2.1360492185705155e-05, "loss": 0.0, "step": 196950 }, { "epoch": 42.47, "learning_rate": 2.135538076811488e-05, "loss": 0.0, "step": 197000 }, { "epoch": 42.48, "learning_rate": 2.1350268450928513e-05, "loss": 0.0, "step": 197050 }, { "epoch": 42.49, "learning_rate": 2.1345155234869697e-05, "loss": 0.0, "step": 197100 }, { "epoch": 42.5, "learning_rate": 2.1340041120662196e-05, "loss": 0.0, "step": 197150 }, { "epoch": 42.51, "learning_rate": 2.1334926109029914e-05, "loss": 0.0, "step": 197200 }, { "epoch": 42.52, "learning_rate": 2.1329810200696865e-05, "loss": 0.0, "step": 197250 }, { "epoch": 42.53, "learning_rate": 2.132469339638721e-05, "loss": 0.0, "step": 197300 }, { "epoch": 42.54, "learning_rate": 2.131957569682522e-05, "loss": 0.0, "step": 197350 }, { "epoch": 42.55, "learning_rate": 2.13144571027353e-05, "loss": 0.0, "step": 197400 }, { "epoch": 42.56, "learning_rate": 2.1309337614841976e-05, "loss": 0.0, "step": 197450 }, { "epoch": 42.57, "learning_rate": 2.130421723386991e-05, "loss": 0.0, "step": 197500 }, { "epoch": 42.58, "learning_rate": 2.129909596054389e-05, "loss": 0.0, "step": 197550 }, { "epoch": 42.6, "learning_rate": 2.1293973795588817e-05, "loss": 0.0, "step": 197600 }, { "epoch": 42.61, "learning_rate": 2.1288850739729735e-05, "loss": 0.0, "step": 197650 }, { "epoch": 42.62, "learning_rate": 2.1283726793691795e-05, "loss": 0.0, "step": 197700 }, { "epoch": 42.63, "learning_rate": 2.1278601958200295e-05, "loss": 0.0, "step": 197750 }, { "epoch": 42.64, "learning_rate": 2.1273476233980644e-05, "loss": 0.0, "step": 197800 }, { "epoch": 42.65, "learning_rate": 2.126834962175838e-05, "loss": 0.0, "step": 197850 }, { "epoch": 42.66, "learning_rate": 2.1263222122259177e-05, "loss": 0.0, "step": 197900 }, { "epoch": 42.67, "learning_rate": 2.125809373620882e-05, "loss": 0.0, "step": 197950 }, { "epoch": 42.68, "learning_rate": 2.1252964464333226e-05, "loss": 0.0, "step": 198000 }, { "epoch": 42.69, "learning_rate": 2.1247834307358437e-05, "loss": 0.0, "step": 198050 }, { "epoch": 42.7, "learning_rate": 2.124270326601062e-05, "loss": 0.0, "step": 198100 }, { "epoch": 42.71, "learning_rate": 2.123757134101607e-05, "loss": 0.0, "step": 198150 }, { "epoch": 42.72, "learning_rate": 2.12324385331012e-05, "loss": 0.0, "step": 198200 }, { "epoch": 42.74, "learning_rate": 2.1227304842992556e-05, "loss": 0.0, "step": 198250 }, { "epoch": 42.75, "learning_rate": 2.12221702714168e-05, "loss": 0.0001, "step": 198300 }, { "epoch": 42.76, "learning_rate": 2.1217034819100735e-05, "loss": 0.0, "step": 198350 }, { "epoch": 42.77, "learning_rate": 2.1211898486771272e-05, "loss": 0.0, "step": 198400 }, { "epoch": 42.78, "learning_rate": 2.1206761275155443e-05, "loss": 0.0, "step": 198450 }, { "epoch": 42.79, "learning_rate": 2.120162318498043e-05, "loss": 0.0, "step": 198500 }, { "epoch": 42.8, "learning_rate": 2.119648421697351e-05, "loss": 0.0, "step": 198550 }, { "epoch": 42.81, "learning_rate": 2.1191344371862106e-05, "loss": 0.0, "step": 198600 }, { "epoch": 42.82, "learning_rate": 2.118620365037375e-05, "loss": 0.0, "step": 198650 }, { "epoch": 42.83, "learning_rate": 2.118106205323611e-05, "loss": 0.0, "step": 198700 }, { "epoch": 42.84, "learning_rate": 2.1175919581176973e-05, "loss": 0.0, "step": 198750 }, { "epoch": 42.85, "learning_rate": 2.1170776234924244e-05, "loss": 0.0, "step": 198800 }, { "epoch": 42.86, "learning_rate": 2.1165632015205963e-05, "loss": 0.0, "step": 198850 }, { "epoch": 42.88, "learning_rate": 2.1160486922750283e-05, "loss": 0.0, "step": 198900 }, { "epoch": 42.89, "learning_rate": 2.115534095828548e-05, "loss": 0.0, "step": 198950 }, { "epoch": 42.9, "learning_rate": 2.1150194122539975e-05, "loss": 0.0, "step": 199000 }, { "epoch": 42.91, "learning_rate": 2.1145046416242284e-05, "loss": 0.0, "step": 199050 }, { "epoch": 42.92, "learning_rate": 2.113989784012106e-05, "loss": 0.0, "step": 199100 }, { "epoch": 42.93, "learning_rate": 2.113474839490508e-05, "loss": 0.0, "step": 199150 }, { "epoch": 42.94, "learning_rate": 2.1129598081323236e-05, "loss": 0.0, "step": 199200 }, { "epoch": 42.95, "learning_rate": 2.1124446900104565e-05, "loss": 0.0, "step": 199250 }, { "epoch": 42.96, "learning_rate": 2.1119294851978186e-05, "loss": 0.0, "step": 199300 }, { "epoch": 42.97, "learning_rate": 2.111414193767338e-05, "loss": 0.0, "step": 199350 }, { "epoch": 42.98, "learning_rate": 2.110898815791953e-05, "loss": 0.0, "step": 199400 }, { "epoch": 42.99, "learning_rate": 2.110383351344615e-05, "loss": 0.0, "step": 199450 }, { "epoch": 43.0, "learning_rate": 2.1098678004982876e-05, "loss": 0.0, "step": 199500 }, { "epoch": 43.02, "learning_rate": 2.1093521633259454e-05, "loss": 0.0, "step": 199550 }, { "epoch": 43.03, "learning_rate": 2.1088364399005774e-05, "loss": 0.0, "step": 199600 }, { "epoch": 43.04, "learning_rate": 2.1083206302951825e-05, "loss": 0.0, "step": 199650 }, { "epoch": 43.05, "learning_rate": 2.1078047345827736e-05, "loss": 0.0, "step": 199700 }, { "epoch": 43.06, "learning_rate": 2.107288752836375e-05, "loss": 0.0, "step": 199750 }, { "epoch": 43.07, "learning_rate": 2.106772685129023e-05, "loss": 0.0, "step": 199800 }, { "epoch": 43.08, "learning_rate": 2.106256531533767e-05, "loss": 0.0, "step": 199850 }, { "epoch": 43.09, "learning_rate": 2.105740292123666e-05, "loss": 0.0, "step": 199900 }, { "epoch": 43.1, "learning_rate": 2.1052239669717955e-05, "loss": 0.0, "step": 199950 }, { "epoch": 43.11, "learning_rate": 2.1047075561512393e-05, "loss": 0.0, "step": 200000 }, { "epoch": 43.12, "learning_rate": 2.1041910597350948e-05, "loss": 0.0, "step": 200050 }, { "epoch": 43.13, "learning_rate": 2.103674477796472e-05, "loss": 0.0, "step": 200100 }, { "epoch": 43.15, "learning_rate": 2.1031578104084913e-05, "loss": 0.0, "step": 200150 }, { "epoch": 43.16, "learning_rate": 2.1026410576442878e-05, "loss": 0.0, "step": 200200 }, { "epoch": 43.17, "learning_rate": 2.1021242195770056e-05, "loss": 0.0, "step": 200250 }, { "epoch": 43.18, "learning_rate": 2.1016072962798035e-05, "loss": 0.0, "step": 200300 }, { "epoch": 43.19, "learning_rate": 2.1010902878258517e-05, "loss": 0.0, "step": 200350 }, { "epoch": 43.2, "learning_rate": 2.1005731942883307e-05, "loss": 0.0, "step": 200400 }, { "epoch": 43.21, "learning_rate": 2.1000560157404357e-05, "loss": 0.0, "step": 200450 }, { "epoch": 43.22, "learning_rate": 2.099538752255372e-05, "loss": 0.0, "step": 200500 }, { "epoch": 43.23, "learning_rate": 2.099021403906358e-05, "loss": 0.0, "step": 200550 }, { "epoch": 43.24, "learning_rate": 2.0985039707666232e-05, "loss": 0.0, "step": 200600 }, { "epoch": 43.25, "learning_rate": 2.09798645290941e-05, "loss": 0.0, "step": 200650 }, { "epoch": 43.26, "learning_rate": 2.0974688504079724e-05, "loss": 0.0, "step": 200700 }, { "epoch": 43.27, "learning_rate": 2.096951163335576e-05, "loss": 0.0, "step": 200750 }, { "epoch": 43.29, "learning_rate": 2.0964333917654996e-05, "loss": 0.0, "step": 200800 }, { "epoch": 43.3, "learning_rate": 2.0959155357710322e-05, "loss": 0.0, "step": 200850 }, { "epoch": 43.31, "learning_rate": 2.095397595425476e-05, "loss": 0.0, "step": 200900 }, { "epoch": 43.32, "learning_rate": 2.0948795708021452e-05, "loss": 0.0, "step": 200950 }, { "epoch": 43.33, "learning_rate": 2.0943614619743645e-05, "loss": 0.0, "step": 201000 }, { "epoch": 43.34, "learning_rate": 2.0938432690154724e-05, "loss": 0.0001, "step": 201050 }, { "epoch": 43.35, "learning_rate": 2.0933249919988182e-05, "loss": 0.0, "step": 201100 }, { "epoch": 43.36, "learning_rate": 2.0928066309977637e-05, "loss": 0.0, "step": 201150 }, { "epoch": 43.37, "learning_rate": 2.0922881860856814e-05, "loss": 0.0, "step": 201200 }, { "epoch": 43.38, "learning_rate": 2.0917696573359574e-05, "loss": 0.0, "step": 201250 }, { "epoch": 43.39, "learning_rate": 2.0912510448219882e-05, "loss": 0.0, "step": 201300 }, { "epoch": 43.4, "learning_rate": 2.090732348617183e-05, "loss": 0.0, "step": 201350 }, { "epoch": 43.41, "learning_rate": 2.090213568794963e-05, "loss": 0.0, "step": 201400 }, { "epoch": 43.43, "learning_rate": 2.08969470542876e-05, "loss": 0.0, "step": 201450 }, { "epoch": 43.44, "learning_rate": 2.089175758592019e-05, "loss": 0.0, "step": 201500 }, { "epoch": 43.45, "learning_rate": 2.0886567283581963e-05, "loss": 0.0, "step": 201550 }, { "epoch": 43.46, "learning_rate": 2.0881376148007587e-05, "loss": 0.0, "step": 201600 }, { "epoch": 43.47, "learning_rate": 2.0876184179931887e-05, "loss": 0.0, "step": 201650 }, { "epoch": 43.48, "learning_rate": 2.0870991380089754e-05, "loss": 0.0, "step": 201700 }, { "epoch": 43.49, "learning_rate": 2.0865797749216232e-05, "loss": 0.0, "step": 201750 }, { "epoch": 43.5, "learning_rate": 2.086060328804648e-05, "loss": 0.0, "step": 201800 }, { "epoch": 43.51, "learning_rate": 2.0855407997315755e-05, "loss": 0.0, "step": 201850 }, { "epoch": 43.52, "learning_rate": 2.0850211877759453e-05, "loss": 0.0, "step": 201900 }, { "epoch": 43.53, "learning_rate": 2.084501493011307e-05, "loss": 0.0, "step": 201950 }, { "epoch": 43.54, "learning_rate": 2.0839817155112234e-05, "loss": 0.0, "step": 202000 }, { "epoch": 43.55, "learning_rate": 2.0834618553492685e-05, "loss": 0.0, "step": 202050 }, { "epoch": 43.57, "learning_rate": 2.082941912599027e-05, "loss": 0.0001, "step": 202100 }, { "epoch": 43.58, "learning_rate": 2.0824218873340966e-05, "loss": 0.0, "step": 202150 }, { "epoch": 43.59, "learning_rate": 2.0819017796280866e-05, "loss": 0.0, "step": 202200 }, { "epoch": 43.6, "learning_rate": 2.0813815895546168e-05, "loss": 0.0, "step": 202250 }, { "epoch": 43.61, "learning_rate": 2.08086131718732e-05, "loss": 0.0, "step": 202300 }, { "epoch": 43.62, "learning_rate": 2.0803409625998398e-05, "loss": 0.0, "step": 202350 }, { "epoch": 43.63, "learning_rate": 2.0798205258658326e-05, "loss": 0.0, "step": 202400 }, { "epoch": 43.64, "learning_rate": 2.0793000070589644e-05, "loss": 0.0, "step": 202450 }, { "epoch": 43.65, "learning_rate": 2.078779406252914e-05, "loss": 0.0, "step": 202500 }, { "epoch": 43.66, "learning_rate": 2.0782587235213725e-05, "loss": 0.0, "step": 202550 }, { "epoch": 43.67, "learning_rate": 2.0777379589380417e-05, "loss": 0.0, "step": 202600 }, { "epoch": 43.68, "learning_rate": 2.0772171125766352e-05, "loss": 0.0, "step": 202650 }, { "epoch": 43.69, "learning_rate": 2.0766961845108776e-05, "loss": 0.0, "step": 202700 }, { "epoch": 43.71, "learning_rate": 2.0761751748145064e-05, "loss": 0.0, "step": 202750 }, { "epoch": 43.72, "learning_rate": 2.075654083561269e-05, "loss": 0.0, "step": 202800 }, { "epoch": 43.73, "learning_rate": 2.075132910824926e-05, "loss": 0.0, "step": 202850 }, { "epoch": 43.74, "learning_rate": 2.0746116566792484e-05, "loss": 0.0, "step": 202900 }, { "epoch": 43.75, "learning_rate": 2.074090321198019e-05, "loss": 0.0, "step": 202950 }, { "epoch": 43.76, "learning_rate": 2.0735689044550327e-05, "loss": 0.0, "step": 203000 }, { "epoch": 43.77, "learning_rate": 2.0730474065240944e-05, "loss": 0.0, "step": 203050 }, { "epoch": 43.78, "learning_rate": 2.0725258274790225e-05, "loss": 0.0, "step": 203100 }, { "epoch": 43.79, "learning_rate": 2.072004167393645e-05, "loss": 0.0, "step": 203150 }, { "epoch": 43.8, "learning_rate": 2.071482426341802e-05, "loss": 0.0, "step": 203200 }, { "epoch": 43.81, "learning_rate": 2.070960604397347e-05, "loss": 0.0, "step": 203250 }, { "epoch": 43.82, "learning_rate": 2.0704387016341416e-05, "loss": 0.0, "step": 203300 }, { "epoch": 43.83, "learning_rate": 2.069916718126061e-05, "loss": 0.0, "step": 203350 }, { "epoch": 43.85, "learning_rate": 2.0693946539469915e-05, "loss": 0.0, "step": 203400 }, { "epoch": 43.86, "learning_rate": 2.0688725091708297e-05, "loss": 0.0, "step": 203450 }, { "epoch": 43.87, "learning_rate": 2.068350283871486e-05, "loss": 0.0, "step": 203500 }, { "epoch": 43.88, "learning_rate": 2.067827978122879e-05, "loss": 0.0, "step": 203550 }, { "epoch": 43.89, "learning_rate": 2.0673055919989424e-05, "loss": 0.0, "step": 203600 }, { "epoch": 43.9, "learning_rate": 2.066783125573618e-05, "loss": 0.0, "step": 203650 }, { "epoch": 43.91, "learning_rate": 2.06626057892086e-05, "loss": 0.0, "step": 203700 }, { "epoch": 43.92, "learning_rate": 2.0657379521146357e-05, "loss": 0.0, "step": 203750 }, { "epoch": 43.93, "learning_rate": 2.0652152452289208e-05, "loss": 0.0, "step": 203800 }, { "epoch": 43.94, "learning_rate": 2.0646924583377042e-05, "loss": 0.0, "step": 203850 }, { "epoch": 43.95, "learning_rate": 2.064169591514986e-05, "loss": 0.0, "step": 203900 }, { "epoch": 43.96, "learning_rate": 2.063646644834777e-05, "loss": 0.0, "step": 203950 }, { "epoch": 43.97, "learning_rate": 2.0631236183711002e-05, "loss": 0.0, "step": 204000 }, { "epoch": 43.99, "learning_rate": 2.0626005121979886e-05, "loss": 0.0, "step": 204050 }, { "epoch": 44.0, "learning_rate": 2.0620773263894877e-05, "loss": 0.0, "step": 204100 }, { "epoch": 44.01, "learning_rate": 2.0615540610196537e-05, "loss": 0.0, "step": 204150 }, { "epoch": 44.02, "learning_rate": 2.0610307161625535e-05, "loss": 0.0, "step": 204200 }, { "epoch": 44.03, "learning_rate": 2.0605072918922674e-05, "loss": 0.0, "step": 204250 }, { "epoch": 44.04, "learning_rate": 2.0599837882828838e-05, "loss": 0.0, "step": 204300 }, { "epoch": 44.05, "learning_rate": 2.0594602054085052e-05, "loss": 0.0, "step": 204350 }, { "epoch": 44.06, "learning_rate": 2.058936543343243e-05, "loss": 0.0, "step": 204400 }, { "epoch": 44.07, "learning_rate": 2.0584128021612223e-05, "loss": 0.0, "step": 204450 }, { "epoch": 44.08, "learning_rate": 2.057888981936577e-05, "loss": 0.0, "step": 204500 }, { "epoch": 44.09, "learning_rate": 2.0573650827434532e-05, "loss": 0.0001, "step": 204550 }, { "epoch": 44.1, "learning_rate": 2.0568411046560084e-05, "loss": 0.0, "step": 204600 }, { "epoch": 44.12, "learning_rate": 2.056317047748411e-05, "loss": 0.0, "step": 204650 }, { "epoch": 44.13, "learning_rate": 2.0557929120948408e-05, "loss": 0.0, "step": 204700 }, { "epoch": 44.14, "learning_rate": 2.055268697769488e-05, "loss": 0.0, "step": 204750 }, { "epoch": 44.15, "learning_rate": 2.0547444048465548e-05, "loss": 0.0, "step": 204800 }, { "epoch": 44.16, "learning_rate": 2.0542200334002545e-05, "loss": 0.0, "step": 204850 }, { "epoch": 44.17, "learning_rate": 2.0536955835048106e-05, "loss": 0.0, "step": 204900 }, { "epoch": 44.18, "learning_rate": 2.0531710552344587e-05, "loss": 0.0, "step": 204950 }, { "epoch": 44.19, "learning_rate": 2.0526464486634454e-05, "loss": 0.0, "step": 205000 }, { "epoch": 44.2, "learning_rate": 2.052121763866027e-05, "loss": 0.0, "step": 205050 }, { "epoch": 44.21, "learning_rate": 2.0515970009164733e-05, "loss": 0.0, "step": 205100 }, { "epoch": 44.22, "learning_rate": 2.0510721598890628e-05, "loss": 0.0, "step": 205150 }, { "epoch": 44.23, "learning_rate": 2.050547240858087e-05, "loss": 0.0, "step": 205200 }, { "epoch": 44.24, "learning_rate": 2.0500222438978465e-05, "loss": 0.0, "step": 205250 }, { "epoch": 44.26, "learning_rate": 2.049497169082655e-05, "loss": 0.0, "step": 205300 }, { "epoch": 44.27, "learning_rate": 2.0489720164868356e-05, "loss": 0.0, "step": 205350 }, { "epoch": 44.28, "learning_rate": 2.0484467861847228e-05, "loss": 0.0, "step": 205400 }, { "epoch": 44.29, "learning_rate": 2.0479214782506628e-05, "loss": 0.0, "step": 205450 }, { "epoch": 44.3, "learning_rate": 2.0473960927590115e-05, "loss": 0.0, "step": 205500 }, { "epoch": 44.31, "learning_rate": 2.0468706297841378e-05, "loss": 0.0, "step": 205550 }, { "epoch": 44.32, "learning_rate": 2.0463450894004192e-05, "loss": 0.0, "step": 205600 }, { "epoch": 44.33, "learning_rate": 2.0458194716822452e-05, "loss": 0.0, "step": 205650 }, { "epoch": 44.34, "learning_rate": 2.045293776704018e-05, "loss": 0.0001, "step": 205700 }, { "epoch": 44.35, "learning_rate": 2.0447680045401467e-05, "loss": 0.0, "step": 205750 }, { "epoch": 44.36, "learning_rate": 2.0442421552650553e-05, "loss": 0.0, "step": 205800 }, { "epoch": 44.37, "learning_rate": 2.0437162289531766e-05, "loss": 0.0, "step": 205850 }, { "epoch": 44.38, "learning_rate": 2.0431902256789543e-05, "loss": 0.0, "step": 205900 }, { "epoch": 44.4, "learning_rate": 2.042664145516845e-05, "loss": 0.0, "step": 205950 }, { "epoch": 44.41, "learning_rate": 2.0421379885413125e-05, "loss": 0.0, "step": 206000 }, { "epoch": 44.42, "learning_rate": 2.041611754826836e-05, "loss": 0.0, "step": 206050 }, { "epoch": 44.43, "learning_rate": 2.041085444447901e-05, "loss": 0.0, "step": 206100 }, { "epoch": 44.44, "learning_rate": 2.0405590574790076e-05, "loss": 0.0, "step": 206150 }, { "epoch": 44.45, "learning_rate": 2.040032593994665e-05, "loss": 0.0, "step": 206200 }, { "epoch": 44.46, "learning_rate": 2.0395060540693932e-05, "loss": 0.0, "step": 206250 }, { "epoch": 44.47, "learning_rate": 2.038979437777723e-05, "loss": 0.0, "step": 206300 }, { "epoch": 44.48, "learning_rate": 2.0384527451941962e-05, "loss": 0.0, "step": 206350 }, { "epoch": 44.49, "learning_rate": 2.0379259763933663e-05, "loss": 0.0, "step": 206400 }, { "epoch": 44.5, "learning_rate": 2.037399131449796e-05, "loss": 0.0, "step": 206450 }, { "epoch": 44.51, "learning_rate": 2.0368722104380596e-05, "loss": 0.0, "step": 206500 }, { "epoch": 44.52, "learning_rate": 2.0363452134327426e-05, "loss": 0.0, "step": 206550 }, { "epoch": 44.54, "learning_rate": 2.0358181405084406e-05, "loss": 0.0, "step": 206600 }, { "epoch": 44.55, "learning_rate": 2.0352909917397595e-05, "loss": 0.0, "step": 206650 }, { "epoch": 44.56, "learning_rate": 2.034763767201317e-05, "loss": 0.0, "step": 206700 }, { "epoch": 44.57, "learning_rate": 2.0342364669677406e-05, "loss": 0.0, "step": 206750 }, { "epoch": 44.58, "learning_rate": 2.0337090911136704e-05, "loss": 0.0, "step": 206800 }, { "epoch": 44.59, "learning_rate": 2.0331816397137538e-05, "loss": 0.0, "step": 206850 }, { "epoch": 44.6, "learning_rate": 2.0326541128426527e-05, "loss": 0.0, "step": 206900 }, { "epoch": 44.61, "learning_rate": 2.0321265105750365e-05, "loss": 0.0, "step": 206950 }, { "epoch": 44.62, "learning_rate": 2.0315988329855874e-05, "loss": 0.0, "step": 207000 }, { "epoch": 44.63, "learning_rate": 2.0310710801489974e-05, "loss": 0.0, "step": 207050 }, { "epoch": 44.64, "learning_rate": 2.0305432521399687e-05, "loss": 0.0, "step": 207100 }, { "epoch": 44.65, "learning_rate": 2.0300153490332154e-05, "loss": 0.0001, "step": 207150 }, { "epoch": 44.66, "learning_rate": 2.029487370903461e-05, "loss": 0.0, "step": 207200 }, { "epoch": 44.68, "learning_rate": 2.0289593178254405e-05, "loss": 0.0, "step": 207250 }, { "epoch": 44.69, "learning_rate": 2.028431189873899e-05, "loss": 0.0, "step": 207300 }, { "epoch": 44.7, "learning_rate": 2.0279029871235924e-05, "loss": 0.0, "step": 207350 }, { "epoch": 44.71, "learning_rate": 2.027374709649287e-05, "loss": 0.0, "step": 207400 }, { "epoch": 44.72, "learning_rate": 2.0268463575257598e-05, "loss": 0.0, "step": 207450 }, { "epoch": 44.73, "learning_rate": 2.0263179308277982e-05, "loss": 0.0, "step": 207500 }, { "epoch": 44.74, "learning_rate": 2.025789429630201e-05, "loss": 0.0, "step": 207550 }, { "epoch": 44.75, "learning_rate": 2.0252608540077757e-05, "loss": 0.0, "step": 207600 }, { "epoch": 44.76, "learning_rate": 2.0247322040353434e-05, "loss": 0.0, "step": 207650 }, { "epoch": 44.77, "learning_rate": 2.0242034797877315e-05, "loss": 0.0, "step": 207700 }, { "epoch": 44.78, "learning_rate": 2.0236746813397823e-05, "loss": 0.0, "step": 207750 }, { "epoch": 44.79, "learning_rate": 2.0231458087663454e-05, "loss": 0.0, "step": 207800 }, { "epoch": 44.8, "learning_rate": 2.022616862142282e-05, "loss": 0.0, "step": 207850 }, { "epoch": 44.82, "learning_rate": 2.0220878415424644e-05, "loss": 0.0, "step": 207900 }, { "epoch": 44.83, "learning_rate": 2.0215587470417746e-05, "loss": 0.0, "step": 207950 }, { "epoch": 44.84, "learning_rate": 2.0210295787151053e-05, "loss": 0.0, "step": 208000 }, { "epoch": 44.85, "learning_rate": 2.0205003366373593e-05, "loss": 0.0, "step": 208050 }, { "epoch": 44.86, "learning_rate": 2.0199710208834502e-05, "loss": 0.0, "step": 208100 }, { "epoch": 44.87, "learning_rate": 2.019441631528303e-05, "loss": 0.0, "step": 208150 }, { "epoch": 44.88, "learning_rate": 2.0189121686468505e-05, "loss": 0.0, "step": 208200 }, { "epoch": 44.89, "learning_rate": 2.0183826323140382e-05, "loss": 0.0, "step": 208250 }, { "epoch": 44.9, "learning_rate": 2.017853022604821e-05, "loss": 0.0, "step": 208300 }, { "epoch": 44.91, "learning_rate": 2.0173233395941654e-05, "loss": 0.0, "step": 208350 }, { "epoch": 44.92, "learning_rate": 2.016793583357047e-05, "loss": 0.0, "step": 208400 }, { "epoch": 44.93, "learning_rate": 2.0162637539684505e-05, "loss": 0.0, "step": 208450 }, { "epoch": 44.95, "learning_rate": 2.0157338515033757e-05, "loss": 0.0, "step": 208500 }, { "epoch": 44.96, "learning_rate": 2.0152038760368263e-05, "loss": 0.0, "step": 208550 }, { "epoch": 44.97, "learning_rate": 2.0146738276438224e-05, "loss": 0.0, "step": 208600 }, { "epoch": 44.98, "learning_rate": 2.01414370639939e-05, "loss": 0.0, "step": 208650 }, { "epoch": 44.99, "learning_rate": 2.0136135123785674e-05, "loss": 0.0, "step": 208700 }, { "epoch": 45.0, "learning_rate": 2.0130832456564038e-05, "loss": 0.0, "step": 208750 }, { "epoch": 45.01, "learning_rate": 2.0125529063079565e-05, "loss": 0.0, "step": 208800 }, { "epoch": 45.02, "learning_rate": 2.0120224944082953e-05, "loss": 0.0, "step": 208850 }, { "epoch": 45.03, "learning_rate": 2.011492010032499e-05, "loss": 0.0, "step": 208900 }, { "epoch": 45.04, "learning_rate": 2.0109614532556566e-05, "loss": 0.0, "step": 208950 }, { "epoch": 45.05, "learning_rate": 2.0104308241528682e-05, "loss": 0.0, "step": 209000 }, { "epoch": 45.06, "learning_rate": 2.0099001227992438e-05, "loss": 0.0, "step": 209050 }, { "epoch": 45.07, "learning_rate": 2.009369349269903e-05, "loss": 0.0, "step": 209100 }, { "epoch": 45.09, "learning_rate": 2.0088385036399765e-05, "loss": 0.0, "step": 209150 }, { "epoch": 45.1, "learning_rate": 2.0083075859846048e-05, "loss": 0.0, "step": 209200 }, { "epoch": 45.11, "learning_rate": 2.0077765963789382e-05, "loss": 0.0, "step": 209250 }, { "epoch": 45.12, "learning_rate": 2.007245534898138e-05, "loss": 0.0, "step": 209300 }, { "epoch": 45.13, "learning_rate": 2.0067144016173762e-05, "loss": 0.0, "step": 209350 }, { "epoch": 45.14, "learning_rate": 2.006183196611832e-05, "loss": 0.0, "step": 209400 }, { "epoch": 45.15, "learning_rate": 2.0056519199566985e-05, "loss": 0.0, "step": 209450 }, { "epoch": 45.16, "learning_rate": 2.005120571727176e-05, "loss": 0.0, "step": 209500 }, { "epoch": 45.17, "learning_rate": 2.0045891519984774e-05, "loss": 0.0, "step": 209550 }, { "epoch": 45.18, "learning_rate": 2.0040576608458234e-05, "loss": 0.0, "step": 209600 }, { "epoch": 45.19, "learning_rate": 2.0035260983444468e-05, "loss": 0.0, "step": 209650 }, { "epoch": 45.2, "learning_rate": 2.002994464569589e-05, "loss": 0.0, "step": 209700 }, { "epoch": 45.21, "learning_rate": 2.002462759596502e-05, "loss": 0.0, "step": 209750 }, { "epoch": 45.23, "learning_rate": 2.0019309835004488e-05, "loss": 0.0, "step": 209800 }, { "epoch": 45.24, "learning_rate": 2.0013991363567002e-05, "loss": 0.0, "step": 209850 }, { "epoch": 45.25, "learning_rate": 2.0008672182405398e-05, "loss": 0.0, "step": 209900 }, { "epoch": 45.26, "learning_rate": 2.0003352292272595e-05, "loss": 0.0, "step": 209950 }, { "epoch": 45.27, "learning_rate": 1.9998031693921617e-05, "loss": 0.0, "step": 210000 }, { "epoch": 45.28, "learning_rate": 1.9992710388105587e-05, "loss": 0.0001, "step": 210050 }, { "epoch": 45.29, "learning_rate": 1.9987388375577727e-05, "loss": 0.0, "step": 210100 }, { "epoch": 45.3, "learning_rate": 1.9982065657091366e-05, "loss": 0.0, "step": 210150 }, { "epoch": 45.31, "learning_rate": 1.9976742233399928e-05, "loss": 0.0, "step": 210200 }, { "epoch": 45.32, "learning_rate": 1.9971418105256928e-05, "loss": 0.0, "step": 210250 }, { "epoch": 45.33, "learning_rate": 1.9966093273416004e-05, "loss": 0.0, "step": 210300 }, { "epoch": 45.34, "learning_rate": 1.9960767738630867e-05, "loss": 0.0, "step": 210350 }, { "epoch": 45.35, "learning_rate": 1.9955441501655342e-05, "loss": 0.0, "step": 210400 }, { "epoch": 45.37, "learning_rate": 1.9950114563243356e-05, "loss": 0.0, "step": 210450 }, { "epoch": 45.38, "learning_rate": 1.994478692414893e-05, "loss": 0.0, "step": 210500 }, { "epoch": 45.39, "learning_rate": 1.993945858512618e-05, "loss": 0.0, "step": 210550 }, { "epoch": 45.4, "learning_rate": 1.993412954692933e-05, "loss": 0.0, "step": 210600 }, { "epoch": 45.41, "learning_rate": 1.992879981031269e-05, "loss": 0.0, "step": 210650 }, { "epoch": 45.42, "learning_rate": 1.9923469376030687e-05, "loss": 0.0, "step": 210700 }, { "epoch": 45.43, "learning_rate": 1.9918138244837837e-05, "loss": 0.0, "step": 210750 }, { "epoch": 45.44, "learning_rate": 1.9912806417488743e-05, "loss": 0.0, "step": 210800 }, { "epoch": 45.45, "learning_rate": 1.9907473894738136e-05, "loss": 0.0, "step": 210850 }, { "epoch": 45.46, "learning_rate": 1.9902140677340816e-05, "loss": 0.0, "step": 210900 }, { "epoch": 45.47, "learning_rate": 1.9896806766051697e-05, "loss": 0.0, "step": 210950 }, { "epoch": 45.48, "learning_rate": 1.9891472161625786e-05, "loss": 0.0, "step": 211000 }, { "epoch": 45.49, "learning_rate": 1.9886136864818194e-05, "loss": 0.0, "step": 211050 }, { "epoch": 45.51, "learning_rate": 1.9880800876384114e-05, "loss": 0.0, "step": 211100 }, { "epoch": 45.52, "learning_rate": 1.987546419707886e-05, "loss": 0.0, "step": 211150 }, { "epoch": 45.53, "learning_rate": 1.9870126827657827e-05, "loss": 0.0, "step": 211200 }, { "epoch": 45.54, "learning_rate": 1.986478876887652e-05, "loss": 0.0, "step": 211250 }, { "epoch": 45.55, "learning_rate": 1.9859450021490522e-05, "loss": 0.0, "step": 211300 }, { "epoch": 45.56, "learning_rate": 1.9854110586255532e-05, "loss": 0.0, "step": 211350 }, { "epoch": 45.57, "learning_rate": 1.9848770463927345e-05, "loss": 0.0, "step": 211400 }, { "epoch": 45.58, "learning_rate": 1.9843429655261842e-05, "loss": 0.0, "step": 211450 }, { "epoch": 45.59, "learning_rate": 1.9838088161015013e-05, "loss": 0.0, "step": 211500 }, { "epoch": 45.6, "learning_rate": 1.983274598194293e-05, "loss": 0.0, "step": 211550 }, { "epoch": 45.61, "learning_rate": 1.982740311880179e-05, "loss": 0.0, "step": 211600 }, { "epoch": 45.62, "learning_rate": 1.9822059572347846e-05, "loss": 0.0, "step": 211650 }, { "epoch": 45.63, "learning_rate": 1.9816715343337484e-05, "loss": 0.0, "step": 211700 }, { "epoch": 45.65, "learning_rate": 1.9811370432527172e-05, "loss": 0.0, "step": 211750 }, { "epoch": 45.66, "learning_rate": 1.9806024840673477e-05, "loss": 0.0, "step": 211800 }, { "epoch": 45.67, "learning_rate": 1.980067856853305e-05, "loss": 0.0, "step": 211850 }, { "epoch": 45.68, "learning_rate": 1.979533161686266e-05, "loss": 0.0, "step": 211900 }, { "epoch": 45.69, "learning_rate": 1.978998398641915e-05, "loss": 0.0, "step": 211950 }, { "epoch": 45.7, "learning_rate": 1.978463567795949e-05, "loss": 0.0, "step": 212000 }, { "epoch": 45.71, "learning_rate": 1.9779286692240706e-05, "loss": 0.0, "step": 212050 }, { "epoch": 45.72, "learning_rate": 1.9773937030019947e-05, "loss": 0.0, "step": 212100 }, { "epoch": 45.73, "learning_rate": 1.9768586692054457e-05, "loss": 0.0, "step": 212150 }, { "epoch": 45.74, "learning_rate": 1.9763235679101564e-05, "loss": 0.0, "step": 212200 }, { "epoch": 45.75, "learning_rate": 1.9757883991918695e-05, "loss": 0.0, "step": 212250 }, { "epoch": 45.76, "learning_rate": 1.975253163126338e-05, "loss": 0.0, "step": 212300 }, { "epoch": 45.77, "learning_rate": 1.9747178597893235e-05, "loss": 0.0, "step": 212350 }, { "epoch": 45.79, "learning_rate": 1.9741824892565976e-05, "loss": 0.0, "step": 212400 }, { "epoch": 45.8, "learning_rate": 1.973647051603941e-05, "loss": 0.0, "step": 212450 }, { "epoch": 45.81, "learning_rate": 1.973111546907145e-05, "loss": 0.0, "step": 212500 }, { "epoch": 45.82, "learning_rate": 1.9725759752420093e-05, "loss": 0.0, "step": 212550 }, { "epoch": 45.83, "learning_rate": 1.9720403366843427e-05, "loss": 0.0, "step": 212600 }, { "epoch": 45.84, "learning_rate": 1.9715046313099657e-05, "loss": 0.0, "step": 212650 }, { "epoch": 45.85, "learning_rate": 1.9709688591947043e-05, "loss": 0.0, "step": 212700 }, { "epoch": 45.86, "learning_rate": 1.9704330204143992e-05, "loss": 0.0, "step": 212750 }, { "epoch": 45.87, "learning_rate": 1.9698971150448955e-05, "loss": 0.0, "step": 212800 }, { "epoch": 45.88, "learning_rate": 1.969361143162051e-05, "loss": 0.0, "step": 212850 }, { "epoch": 45.89, "learning_rate": 1.9688251048417317e-05, "loss": 0.0, "step": 212900 }, { "epoch": 45.9, "learning_rate": 1.9682890001598124e-05, "loss": 0.0, "step": 212950 }, { "epoch": 45.92, "learning_rate": 1.9677528291921793e-05, "loss": 0.0, "step": 213000 }, { "epoch": 45.93, "learning_rate": 1.967216592014726e-05, "loss": 0.0, "step": 213050 }, { "epoch": 45.94, "learning_rate": 1.966680288703356e-05, "loss": 0.0, "step": 213100 }, { "epoch": 45.95, "learning_rate": 1.9661439193339825e-05, "loss": 0.0, "step": 213150 }, { "epoch": 45.96, "learning_rate": 1.9656074839825285e-05, "loss": 0.0, "step": 213200 }, { "epoch": 45.97, "learning_rate": 1.965070982724925e-05, "loss": 0.0, "step": 213250 }, { "epoch": 45.98, "learning_rate": 1.9645344156371135e-05, "loss": 0.0, "step": 213300 }, { "epoch": 45.99, "learning_rate": 1.963997782795044e-05, "loss": 0.0, "step": 213350 }, { "epoch": 46.0, "learning_rate": 1.9634610842746768e-05, "loss": 0.0, "step": 213400 }, { "epoch": 46.01, "learning_rate": 1.96292432015198e-05, "loss": 0.0, "step": 213450 }, { "epoch": 46.02, "learning_rate": 1.962387490502933e-05, "loss": 0.0, "step": 213500 }, { "epoch": 46.03, "learning_rate": 1.961850595403522e-05, "loss": 0.0, "step": 213550 }, { "epoch": 46.04, "learning_rate": 1.9613136349297453e-05, "loss": 0.0, "step": 213600 }, { "epoch": 46.06, "learning_rate": 1.960776609157608e-05, "loss": 0.0, "step": 213650 }, { "epoch": 46.07, "learning_rate": 1.9602395181631254e-05, "loss": 0.0, "step": 213700 }, { "epoch": 46.08, "learning_rate": 1.9597023620223228e-05, "loss": 0.0, "step": 213750 }, { "epoch": 46.09, "learning_rate": 1.9591651408112333e-05, "loss": 0.0, "step": 213800 }, { "epoch": 46.1, "learning_rate": 1.9586278546059e-05, "loss": 0.0, "step": 213850 }, { "epoch": 46.11, "learning_rate": 1.958090503482375e-05, "loss": 0.0, "step": 213900 }, { "epoch": 46.12, "learning_rate": 1.9575530875167202e-05, "loss": 0.0, "step": 213950 }, { "epoch": 46.13, "learning_rate": 1.957015606785006e-05, "loss": 0.0, "step": 214000 }, { "epoch": 46.14, "learning_rate": 1.9564780613633115e-05, "loss": 0.0, "step": 214050 }, { "epoch": 46.15, "learning_rate": 1.955940451327726e-05, "loss": 0.0, "step": 214100 }, { "epoch": 46.16, "learning_rate": 1.9554027767543476e-05, "loss": 0.0, "step": 214150 }, { "epoch": 46.17, "learning_rate": 1.9548650377192833e-05, "loss": 0.0, "step": 214200 }, { "epoch": 46.18, "learning_rate": 1.9543272342986493e-05, "loss": 0.0, "step": 214250 }, { "epoch": 46.2, "learning_rate": 1.953789366568571e-05, "loss": 0.0, "step": 214300 }, { "epoch": 46.21, "learning_rate": 1.9532514346051832e-05, "loss": 0.0, "step": 214350 }, { "epoch": 46.22, "learning_rate": 1.9527134384846293e-05, "loss": 0.0, "step": 214400 }, { "epoch": 46.23, "learning_rate": 1.9521753782830623e-05, "loss": 0.0, "step": 214450 }, { "epoch": 46.24, "learning_rate": 1.9516372540766432e-05, "loss": 0.0, "step": 214500 }, { "epoch": 46.25, "learning_rate": 1.951099065941543e-05, "loss": 0.0, "step": 214550 }, { "epoch": 46.26, "learning_rate": 1.9505608139539422e-05, "loss": 0.0, "step": 214600 }, { "epoch": 46.27, "learning_rate": 1.950022498190029e-05, "loss": 0.0, "step": 214650 }, { "epoch": 46.28, "learning_rate": 1.9494841187260015e-05, "loss": 0.0, "step": 214700 }, { "epoch": 46.29, "learning_rate": 1.948945675638067e-05, "loss": 0.0, "step": 214750 }, { "epoch": 46.3, "learning_rate": 1.948407169002441e-05, "loss": 0.0, "step": 214800 }, { "epoch": 46.31, "learning_rate": 1.9478685988953486e-05, "loss": 0.0, "step": 214850 }, { "epoch": 46.32, "learning_rate": 1.9473299653930238e-05, "loss": 0.0, "step": 214900 }, { "epoch": 46.34, "learning_rate": 1.946791268571709e-05, "loss": 0.0, "step": 214950 }, { "epoch": 46.35, "learning_rate": 1.9462525085076566e-05, "loss": 0.0, "step": 215000 }, { "epoch": 46.36, "learning_rate": 1.9457136852771272e-05, "loss": 0.0, "step": 215050 }, { "epoch": 46.37, "learning_rate": 1.9451747989563904e-05, "loss": 0.0, "step": 215100 }, { "epoch": 46.38, "learning_rate": 1.944635849621725e-05, "loss": 0.0, "step": 215150 }, { "epoch": 46.39, "learning_rate": 1.944096837349419e-05, "loss": 0.0, "step": 215200 }, { "epoch": 46.4, "learning_rate": 1.943557762215768e-05, "loss": 0.0, "step": 215250 }, { "epoch": 46.41, "learning_rate": 1.9430186242970785e-05, "loss": 0.0, "step": 215300 }, { "epoch": 46.42, "learning_rate": 1.9424794236696636e-05, "loss": 0.0, "step": 215350 }, { "epoch": 46.43, "learning_rate": 1.9419401604098475e-05, "loss": 0.0, "step": 215400 }, { "epoch": 46.44, "learning_rate": 1.941400834593962e-05, "loss": 0.0, "step": 215450 }, { "epoch": 46.45, "learning_rate": 1.9408614462983477e-05, "loss": 0.0, "step": 215500 }, { "epoch": 46.46, "learning_rate": 1.9403219955993547e-05, "loss": 0.0, "step": 215550 }, { "epoch": 46.48, "learning_rate": 1.9397824825733407e-05, "loss": 0.0, "step": 215600 }, { "epoch": 46.49, "learning_rate": 1.939242907296674e-05, "loss": 0.0, "step": 215650 }, { "epoch": 46.5, "learning_rate": 1.9387032698457307e-05, "loss": 0.0, "step": 215700 }, { "epoch": 46.51, "learning_rate": 1.938163570296895e-05, "loss": 0.0, "step": 215750 }, { "epoch": 46.52, "learning_rate": 1.9376238087265617e-05, "loss": 0.0, "step": 215800 }, { "epoch": 46.53, "learning_rate": 1.9370839852111327e-05, "loss": 0.0, "step": 215850 }, { "epoch": 46.54, "learning_rate": 1.93654409982702e-05, "loss": 0.0, "step": 215900 }, { "epoch": 46.55, "learning_rate": 1.9360041526506427e-05, "loss": 0.0, "step": 215950 }, { "epoch": 46.56, "learning_rate": 1.9354641437584303e-05, "loss": 0.0, "step": 216000 }, { "epoch": 46.57, "learning_rate": 1.9349240732268206e-05, "loss": 0.0, "step": 216050 }, { "epoch": 46.58, "learning_rate": 1.9343839411322593e-05, "loss": 0.0, "step": 216100 }, { "epoch": 46.59, "learning_rate": 1.933843747551202e-05, "loss": 0.0, "step": 216150 }, { "epoch": 46.6, "learning_rate": 1.9333034925601117e-05, "loss": 0.0, "step": 216200 }, { "epoch": 46.62, "learning_rate": 1.9327631762354615e-05, "loss": 0.0, "step": 216250 }, { "epoch": 46.63, "learning_rate": 1.932222798653732e-05, "loss": 0.0, "step": 216300 }, { "epoch": 46.64, "learning_rate": 1.931682359891413e-05, "loss": 0.0, "step": 216350 }, { "epoch": 46.65, "learning_rate": 1.9311418600250042e-05, "loss": 0.0, "step": 216400 }, { "epoch": 46.66, "learning_rate": 1.9306012991310103e-05, "loss": 0.0, "step": 216450 }, { "epoch": 46.67, "learning_rate": 1.9300606772859488e-05, "loss": 0.0, "step": 216500 }, { "epoch": 46.68, "learning_rate": 1.9295199945663434e-05, "loss": 0.0, "step": 216550 }, { "epoch": 46.69, "learning_rate": 1.928979251048727e-05, "loss": 0.0, "step": 216600 }, { "epoch": 46.7, "learning_rate": 1.9284384468096415e-05, "loss": 0.0, "step": 216650 }, { "epoch": 46.71, "learning_rate": 1.9278975819256366e-05, "loss": 0.0, "step": 216700 }, { "epoch": 46.72, "learning_rate": 1.9273566564732717e-05, "loss": 0.0, "step": 216750 }, { "epoch": 46.73, "learning_rate": 1.926815670529114e-05, "loss": 0.0, "step": 216800 }, { "epoch": 46.74, "learning_rate": 1.926274624169738e-05, "loss": 0.0, "step": 216850 }, { "epoch": 46.76, "learning_rate": 1.9257335174717305e-05, "loss": 0.0, "step": 216900 }, { "epoch": 46.77, "learning_rate": 1.925192350511682e-05, "loss": 0.0, "step": 216950 }, { "epoch": 46.78, "learning_rate": 1.9246511233661963e-05, "loss": 0.0, "step": 217000 }, { "epoch": 46.79, "learning_rate": 1.9241098361118815e-05, "loss": 0.0, "step": 217050 }, { "epoch": 46.8, "learning_rate": 1.923568488825357e-05, "loss": 0.0, "step": 217100 }, { "epoch": 46.81, "learning_rate": 1.92302708158325e-05, "loss": 0.0, "step": 217150 }, { "epoch": 46.82, "learning_rate": 1.9224856144621957e-05, "loss": 0.0, "step": 217200 }, { "epoch": 46.83, "learning_rate": 1.921944087538839e-05, "loss": 0.0, "step": 217250 }, { "epoch": 46.84, "learning_rate": 1.9214025008898302e-05, "loss": 0.0, "step": 217300 }, { "epoch": 46.85, "learning_rate": 1.920860854591832e-05, "loss": 0.0, "step": 217350 }, { "epoch": 46.86, "learning_rate": 1.920319148721513e-05, "loss": 0.0, "step": 217400 }, { "epoch": 46.87, "learning_rate": 1.9197773833555515e-05, "loss": 0.0, "step": 217450 }, { "epoch": 46.89, "learning_rate": 1.9192355585706335e-05, "loss": 0.0, "step": 217500 }, { "epoch": 46.9, "learning_rate": 1.9186936744434532e-05, "loss": 0.0, "step": 217550 }, { "epoch": 46.91, "learning_rate": 1.9181517310507143e-05, "loss": 0.0, "step": 217600 }, { "epoch": 46.92, "learning_rate": 1.9176097284691276e-05, "loss": 0.0, "step": 217650 }, { "epoch": 46.93, "learning_rate": 1.917067666775413e-05, "loss": 0.0, "step": 217700 }, { "epoch": 46.94, "learning_rate": 1.9165255460462995e-05, "loss": 0.0, "step": 217750 }, { "epoch": 46.95, "learning_rate": 1.915983366358522e-05, "loss": 0.0, "step": 217800 }, { "epoch": 46.96, "learning_rate": 1.9154411277888268e-05, "loss": 0.0, "step": 217850 }, { "epoch": 46.97, "learning_rate": 1.9148988304139665e-05, "loss": 0.0, "step": 217900 }, { "epoch": 46.98, "learning_rate": 1.914356474310702e-05, "loss": 0.0, "step": 217950 }, { "epoch": 46.99, "learning_rate": 1.9138140595558045e-05, "loss": 0.0, "step": 218000 }, { "epoch": 47.0, "learning_rate": 1.9132715862260507e-05, "loss": 0.0001, "step": 218050 }, { "epoch": 47.01, "learning_rate": 1.9127290543982284e-05, "loss": 0.0, "step": 218100 }, { "epoch": 47.03, "learning_rate": 1.9121864641491308e-05, "loss": 0.0, "step": 218150 }, { "epoch": 47.04, "learning_rate": 1.911643815555562e-05, "loss": 0.0, "step": 218200 }, { "epoch": 47.05, "learning_rate": 1.911101108694333e-05, "loss": 0.0, "step": 218250 }, { "epoch": 47.06, "learning_rate": 1.910558343642263e-05, "loss": 0.0, "step": 218300 }, { "epoch": 47.07, "learning_rate": 1.9100155204761793e-05, "loss": 0.0, "step": 218350 }, { "epoch": 47.08, "learning_rate": 1.909472639272919e-05, "loss": 0.0, "step": 218400 }, { "epoch": 47.09, "learning_rate": 1.9089297001093254e-05, "loss": 0.0, "step": 218450 }, { "epoch": 47.1, "learning_rate": 1.908386703062251e-05, "loss": 0.0, "step": 218500 }, { "epoch": 47.11, "learning_rate": 1.9078436482085558e-05, "loss": 0.0, "step": 218550 }, { "epoch": 47.12, "learning_rate": 1.90730053562511e-05, "loss": 0.0, "step": 218600 }, { "epoch": 47.13, "learning_rate": 1.906757365388789e-05, "loss": 0.0, "step": 218650 }, { "epoch": 47.14, "learning_rate": 1.9062141375764795e-05, "loss": 0.0, "step": 218700 }, { "epoch": 47.15, "learning_rate": 1.9056708522650725e-05, "loss": 0.0, "step": 218750 }, { "epoch": 47.17, "learning_rate": 1.905127509531471e-05, "loss": 0.0, "step": 218800 }, { "epoch": 47.18, "learning_rate": 1.9045841094525843e-05, "loss": 0.0, "step": 218850 }, { "epoch": 47.19, "learning_rate": 1.9040406521053288e-05, "loss": 0.0, "step": 218900 }, { "epoch": 47.2, "learning_rate": 1.903497137566632e-05, "loss": 0.0, "step": 218950 }, { "epoch": 47.21, "learning_rate": 1.9029535659134264e-05, "loss": 0.0, "step": 219000 }, { "epoch": 47.22, "learning_rate": 1.9024099372226545e-05, "loss": 0.0, "step": 219050 }, { "epoch": 47.23, "learning_rate": 1.9018662515712656e-05, "loss": 0.0, "step": 219100 }, { "epoch": 47.24, "learning_rate": 1.9013225090362184e-05, "loss": 0.0, "step": 219150 }, { "epoch": 47.25, "learning_rate": 1.9007787096944787e-05, "loss": 0.0, "step": 219200 }, { "epoch": 47.26, "learning_rate": 1.9002348536230203e-05, "loss": 0.0, "step": 219250 }, { "epoch": 47.27, "learning_rate": 1.8996909408988262e-05, "loss": 0.0, "step": 219300 }, { "epoch": 47.28, "learning_rate": 1.8991469715988854e-05, "loss": 0.0, "step": 219350 }, { "epoch": 47.29, "learning_rate": 1.898602945800197e-05, "loss": 0.0, "step": 219400 }, { "epoch": 47.31, "learning_rate": 1.8980588635797668e-05, "loss": 0.0, "step": 219450 }, { "epoch": 47.32, "learning_rate": 1.8975147250146084e-05, "loss": 0.0, "step": 219500 }, { "epoch": 47.33, "learning_rate": 1.8969705301817456e-05, "loss": 0.0, "step": 219550 }, { "epoch": 47.34, "learning_rate": 1.8964262791582064e-05, "loss": 0.0, "step": 219600 }, { "epoch": 47.35, "learning_rate": 1.8958819720210303e-05, "loss": 0.0, "step": 219650 }, { "epoch": 47.36, "learning_rate": 1.8953376088472634e-05, "loss": 0.0, "step": 219700 }, { "epoch": 47.37, "learning_rate": 1.894793189713958e-05, "loss": 0.0, "step": 219750 }, { "epoch": 47.38, "learning_rate": 1.8942487146981773e-05, "loss": 0.0, "step": 219800 }, { "epoch": 47.39, "learning_rate": 1.8937041838769908e-05, "loss": 0.0, "step": 219850 }, { "epoch": 47.4, "learning_rate": 1.893159597327476e-05, "loss": 0.0, "step": 219900 }, { "epoch": 47.41, "learning_rate": 1.8926149551267188e-05, "loss": 0.0, "step": 219950 }, { "epoch": 47.42, "learning_rate": 1.892070257351812e-05, "loss": 0.0, "step": 220000 }, { "epoch": 47.43, "learning_rate": 1.8915255040798574e-05, "loss": 0.0, "step": 220050 }, { "epoch": 47.45, "learning_rate": 1.8909806953879638e-05, "loss": 0.0, "step": 220100 }, { "epoch": 47.46, "learning_rate": 1.890435831353248e-05, "loss": 0.0, "step": 220150 }, { "epoch": 47.47, "learning_rate": 1.8898909120528358e-05, "loss": 0.0, "step": 220200 }, { "epoch": 47.48, "learning_rate": 1.8893459375638583e-05, "loss": 0.0, "step": 220250 }, { "epoch": 47.49, "learning_rate": 1.8888009079634576e-05, "loss": 0.0, "step": 220300 }, { "epoch": 47.5, "learning_rate": 1.8882558233287808e-05, "loss": 0.0, "step": 220350 }, { "epoch": 47.51, "learning_rate": 1.8877106837369845e-05, "loss": 0.0, "step": 220400 }, { "epoch": 47.52, "learning_rate": 1.887165489265232e-05, "loss": 0.0, "step": 220450 }, { "epoch": 47.53, "learning_rate": 1.8866202399906956e-05, "loss": 0.0, "step": 220500 }, { "epoch": 47.54, "learning_rate": 1.8860749359905544e-05, "loss": 0.0, "step": 220550 }, { "epoch": 47.55, "learning_rate": 1.8855295773419945e-05, "loss": 0.0, "step": 220600 }, { "epoch": 47.56, "learning_rate": 1.8849841641222118e-05, "loss": 0.0, "step": 220650 }, { "epoch": 47.57, "learning_rate": 1.884438696408409e-05, "loss": 0.0, "step": 220700 }, { "epoch": 47.59, "learning_rate": 1.8838931742777955e-05, "loss": 0.0, "step": 220750 }, { "epoch": 47.6, "learning_rate": 1.8833475978075904e-05, "loss": 0.0, "step": 220800 }, { "epoch": 47.61, "learning_rate": 1.8828019670750184e-05, "loss": 0.0, "step": 220850 }, { "epoch": 47.62, "learning_rate": 1.882256282157313e-05, "loss": 0.0, "step": 220900 }, { "epoch": 47.63, "learning_rate": 1.8817105431317156e-05, "loss": 0.0, "step": 220950 }, { "epoch": 47.64, "learning_rate": 1.881164750075474e-05, "loss": 0.0, "step": 221000 }, { "epoch": 47.65, "learning_rate": 1.8806189030658462e-05, "loss": 0.0, "step": 221050 }, { "epoch": 47.66, "learning_rate": 1.8800730021800937e-05, "loss": 0.0, "step": 221100 }, { "epoch": 47.67, "learning_rate": 1.8795270474954908e-05, "loss": 0.0, "step": 221150 }, { "epoch": 47.68, "learning_rate": 1.8789810390893144e-05, "loss": 0.0, "step": 221200 }, { "epoch": 47.69, "learning_rate": 1.8784349770388527e-05, "loss": 0.0, "step": 221250 }, { "epoch": 47.7, "learning_rate": 1.8778888614214e-05, "loss": 0.0, "step": 221300 }, { "epoch": 47.72, "learning_rate": 1.877342692314257e-05, "loss": 0.0, "step": 221350 }, { "epoch": 47.73, "learning_rate": 1.876796469794735e-05, "loss": 0.0, "step": 221400 }, { "epoch": 47.74, "learning_rate": 1.8762501939401495e-05, "loss": 0.0, "step": 221450 }, { "epoch": 47.75, "learning_rate": 1.8757038648278263e-05, "loss": 0.0, "step": 221500 }, { "epoch": 47.76, "learning_rate": 1.875157482535097e-05, "loss": 0.0, "step": 221550 }, { "epoch": 47.77, "learning_rate": 1.874611047139302e-05, "loss": 0.0, "step": 221600 }, { "epoch": 47.78, "learning_rate": 1.8740645587177877e-05, "loss": 0.0, "step": 221650 }, { "epoch": 47.79, "learning_rate": 1.8735180173479095e-05, "loss": 0.0, "step": 221700 }, { "epoch": 47.8, "learning_rate": 1.8729714231070293e-05, "loss": 0.0, "step": 221750 }, { "epoch": 47.81, "learning_rate": 1.872424776072517e-05, "loss": 0.0, "step": 221800 }, { "epoch": 47.82, "learning_rate": 1.871878076321749e-05, "loss": 0.0, "step": 221850 }, { "epoch": 47.83, "learning_rate": 1.8713313239321114e-05, "loss": 0.0, "step": 221900 }, { "epoch": 47.84, "learning_rate": 1.870784518980995e-05, "loss": 0.0, "step": 221950 }, { "epoch": 47.86, "learning_rate": 1.8702376615458005e-05, "loss": 0.0, "step": 222000 }, { "epoch": 47.87, "learning_rate": 1.8696907517039333e-05, "loss": 0.0, "step": 222050 }, { "epoch": 47.88, "learning_rate": 1.8691437895328096e-05, "loss": 0.0, "step": 222100 }, { "epoch": 47.89, "learning_rate": 1.8685967751098504e-05, "loss": 0.0, "step": 222150 }, { "epoch": 47.9, "learning_rate": 1.868049708512484e-05, "loss": 0.0, "step": 222200 }, { "epoch": 47.91, "learning_rate": 1.8675025898181487e-05, "loss": 0.0, "step": 222250 }, { "epoch": 47.92, "learning_rate": 1.8669554191042873e-05, "loss": 0.0, "step": 222300 }, { "epoch": 47.93, "learning_rate": 1.8664081964483513e-05, "loss": 0.0, "step": 222350 }, { "epoch": 47.94, "learning_rate": 1.8658609219277995e-05, "loss": 0.0, "step": 222400 }, { "epoch": 47.95, "learning_rate": 1.865313595620098e-05, "loss": 0.0, "step": 222450 }, { "epoch": 47.96, "learning_rate": 1.8647662176027203e-05, "loss": 0.0, "step": 222500 }, { "epoch": 47.97, "learning_rate": 1.8642187879531464e-05, "loss": 0.0, "step": 222550 }, { "epoch": 47.98, "learning_rate": 1.8636713067488647e-05, "loss": 0.0, "step": 222600 }, { "epoch": 48.0, "learning_rate": 1.8631237740673707e-05, "loss": 0.0, "step": 222650 }, { "epoch": 48.01, "learning_rate": 1.862576189986166e-05, "loss": 0.0, "step": 222700 }, { "epoch": 48.02, "learning_rate": 1.8620285545827626e-05, "loss": 0.0001, "step": 222750 }, { "epoch": 48.03, "learning_rate": 1.8614808679346748e-05, "loss": 0.0, "step": 222800 }, { "epoch": 48.04, "learning_rate": 1.860933130119429e-05, "loss": 0.0, "step": 222850 }, { "epoch": 48.05, "learning_rate": 1.8603853412145562e-05, "loss": 0.0, "step": 222900 }, { "epoch": 48.06, "learning_rate": 1.8598375012975945e-05, "loss": 0.0, "step": 222950 }, { "epoch": 48.07, "learning_rate": 1.8592896104460913e-05, "loss": 0.0, "step": 223000 }, { "epoch": 48.08, "learning_rate": 1.8587416687375983e-05, "loss": 0.0, "step": 223050 }, { "epoch": 48.09, "learning_rate": 1.8581936762496776e-05, "loss": 0.0, "step": 223100 }, { "epoch": 48.1, "learning_rate": 1.8576456330598952e-05, "loss": 0.0, "step": 223150 }, { "epoch": 48.11, "learning_rate": 1.857097539245827e-05, "loss": 0.0, "step": 223200 }, { "epoch": 48.12, "learning_rate": 1.856549394885055e-05, "loss": 0.0, "step": 223250 }, { "epoch": 48.14, "learning_rate": 1.856001200055168e-05, "loss": 0.0, "step": 223300 }, { "epoch": 48.15, "learning_rate": 1.8554529548337623e-05, "loss": 0.0, "step": 223350 }, { "epoch": 48.16, "learning_rate": 1.854904659298441e-05, "loss": 0.0, "step": 223400 }, { "epoch": 48.17, "learning_rate": 1.8543563135268155e-05, "loss": 0.0, "step": 223450 }, { "epoch": 48.18, "learning_rate": 1.8538079175965027e-05, "loss": 0.0, "step": 223500 }, { "epoch": 48.19, "learning_rate": 1.8532594715851274e-05, "loss": 0.0, "step": 223550 }, { "epoch": 48.2, "learning_rate": 1.852710975570322e-05, "loss": 0.0, "step": 223600 }, { "epoch": 48.21, "learning_rate": 1.8521624296297246e-05, "loss": 0.0, "step": 223650 }, { "epoch": 48.22, "learning_rate": 1.8516138338409823e-05, "loss": 0.0, "step": 223700 }, { "epoch": 48.23, "learning_rate": 1.8510651882817472e-05, "loss": 0.0, "step": 223750 }, { "epoch": 48.24, "learning_rate": 1.850516493029679e-05, "loss": 0.0, "step": 223800 }, { "epoch": 48.25, "learning_rate": 1.8499677481624464e-05, "loss": 0.0, "step": 223850 }, { "epoch": 48.26, "learning_rate": 1.8494189537577222e-05, "loss": 0.0, "step": 223900 }, { "epoch": 48.28, "learning_rate": 1.8488701098931884e-05, "loss": 0.0, "step": 223950 }, { "epoch": 48.29, "learning_rate": 1.848321216646532e-05, "loss": 0.0, "step": 224000 }, { "epoch": 48.3, "learning_rate": 1.847772274095449e-05, "loss": 0.0, "step": 224050 }, { "epoch": 48.31, "learning_rate": 1.8472232823176417e-05, "loss": 0.0, "step": 224100 }, { "epoch": 48.32, "learning_rate": 1.8466742413908185e-05, "loss": 0.0, "step": 224150 }, { "epoch": 48.33, "learning_rate": 1.8461251513926963e-05, "loss": 0.0, "step": 224200 }, { "epoch": 48.34, "learning_rate": 1.845576012400997e-05, "loss": 0.0001, "step": 224250 }, { "epoch": 48.35, "learning_rate": 1.845026824493452e-05, "loss": 0.0, "step": 224300 }, { "epoch": 48.36, "learning_rate": 1.8444775877477966e-05, "loss": 0.0, "step": 224350 }, { "epoch": 48.37, "learning_rate": 1.843928302241775e-05, "loss": 0.0, "step": 224400 }, { "epoch": 48.38, "learning_rate": 1.843378968053139e-05, "loss": 0.0, "step": 224450 }, { "epoch": 48.39, "learning_rate": 1.8428295852596443e-05, "loss": 0.0, "step": 224500 }, { "epoch": 48.4, "learning_rate": 1.8422801539390573e-05, "loss": 0.0, "step": 224550 }, { "epoch": 48.42, "learning_rate": 1.841730674169148e-05, "loss": 0.0, "step": 224600 }, { "epoch": 48.43, "learning_rate": 1.8411811460276943e-05, "loss": 0.0, "step": 224650 }, { "epoch": 48.44, "learning_rate": 1.8406315695924825e-05, "loss": 0.0, "step": 224700 }, { "epoch": 48.45, "learning_rate": 1.8400819449413036e-05, "loss": 0.0, "step": 224750 }, { "epoch": 48.46, "learning_rate": 1.8395322721519568e-05, "loss": 0.0, "step": 224800 }, { "epoch": 48.47, "learning_rate": 1.8389825513022462e-05, "loss": 0.0, "step": 224850 }, { "epoch": 48.48, "learning_rate": 1.8384327824699856e-05, "loss": 0.0, "step": 224900 }, { "epoch": 48.49, "learning_rate": 1.8378829657329938e-05, "loss": 0.0, "step": 224950 }, { "epoch": 48.5, "learning_rate": 1.8373331011690963e-05, "loss": 0.0, "step": 225000 }, { "epoch": 48.51, "learning_rate": 1.836783188856126e-05, "loss": 0.0, "step": 225050 }, { "epoch": 48.52, "learning_rate": 1.836233228871922e-05, "loss": 0.0, "step": 225100 }, { "epoch": 48.53, "learning_rate": 1.8356832212943305e-05, "loss": 0.0, "step": 225150 }, { "epoch": 48.54, "learning_rate": 1.8351331662012047e-05, "loss": 0.0, "step": 225200 }, { "epoch": 48.56, "learning_rate": 1.8345830636704036e-05, "loss": 0.0, "step": 225250 }, { "epoch": 48.57, "learning_rate": 1.834032913779795e-05, "loss": 0.0, "step": 225300 }, { "epoch": 48.58, "learning_rate": 1.8334827166072495e-05, "loss": 0.0, "step": 225350 }, { "epoch": 48.59, "learning_rate": 1.8329324722306492e-05, "loss": 0.0, "step": 225400 }, { "epoch": 48.6, "learning_rate": 1.8323821807278787e-05, "loss": 0.0, "step": 225450 }, { "epoch": 48.61, "learning_rate": 1.8318318421768316e-05, "loss": 0.0, "step": 225500 }, { "epoch": 48.62, "learning_rate": 1.8312814566554087e-05, "loss": 0.0, "step": 225550 }, { "epoch": 48.63, "learning_rate": 1.830731024241515e-05, "loss": 0.0, "step": 225600 }, { "epoch": 48.64, "learning_rate": 1.830180545013064e-05, "loss": 0.0, "step": 225650 }, { "epoch": 48.65, "learning_rate": 1.8296300190479752e-05, "loss": 0.0, "step": 225700 }, { "epoch": 48.66, "learning_rate": 1.8290794464241748e-05, "loss": 0.0, "step": 225750 }, { "epoch": 48.67, "learning_rate": 1.8285288272195962e-05, "loss": 0.0, "step": 225800 }, { "epoch": 48.69, "learning_rate": 1.8279781615121783e-05, "loss": 0.0, "step": 225850 }, { "epoch": 48.7, "learning_rate": 1.8274274493798678e-05, "loss": 0.0, "step": 225900 }, { "epoch": 48.71, "learning_rate": 1.8268766909006164e-05, "loss": 0.0, "step": 225950 }, { "epoch": 48.72, "learning_rate": 1.826325886152384e-05, "loss": 0.0, "step": 226000 }, { "epoch": 48.73, "learning_rate": 1.8257750352131356e-05, "loss": 0.0, "step": 226050 }, { "epoch": 48.74, "learning_rate": 1.825224138160844e-05, "loss": 0.0, "step": 226100 }, { "epoch": 48.75, "learning_rate": 1.8246731950734883e-05, "loss": 0.0, "step": 226150 }, { "epoch": 48.76, "learning_rate": 1.824122206029053e-05, "loss": 0.0, "step": 226200 }, { "epoch": 48.77, "learning_rate": 1.82357117110553e-05, "loss": 0.0, "step": 226250 }, { "epoch": 48.78, "learning_rate": 1.8230200903809183e-05, "loss": 0.0, "step": 226300 }, { "epoch": 48.79, "learning_rate": 1.8224689639332217e-05, "loss": 0.0, "step": 226350 }, { "epoch": 48.8, "learning_rate": 1.8219177918404523e-05, "loss": 0.0, "step": 226400 }, { "epoch": 48.81, "learning_rate": 1.8213665741806274e-05, "loss": 0.0, "step": 226450 }, { "epoch": 48.83, "learning_rate": 1.8208153110317714e-05, "loss": 0.0, "step": 226500 }, { "epoch": 48.84, "learning_rate": 1.8202640024719142e-05, "loss": 0.0, "step": 226550 }, { "epoch": 48.85, "learning_rate": 1.819712648579094e-05, "loss": 0.0, "step": 226600 }, { "epoch": 48.86, "learning_rate": 1.8191612494313534e-05, "loss": 0.0, "step": 226650 }, { "epoch": 48.87, "learning_rate": 1.8186098051067423e-05, "loss": 0.0, "step": 226700 }, { "epoch": 48.88, "learning_rate": 1.818058315683317e-05, "loss": 0.0, "step": 226750 }, { "epoch": 48.89, "learning_rate": 1.8175067812391405e-05, "loss": 0.0, "step": 226800 }, { "epoch": 48.9, "learning_rate": 1.8169552018522817e-05, "loss": 0.0, "step": 226850 }, { "epoch": 48.91, "learning_rate": 1.8164035776008157e-05, "loss": 0.0, "step": 226900 }, { "epoch": 48.92, "learning_rate": 1.815851908562824e-05, "loss": 0.0, "step": 226950 }, { "epoch": 48.93, "learning_rate": 1.8153001948163955e-05, "loss": 0.0, "step": 227000 }, { "epoch": 48.94, "learning_rate": 1.8147484364396235e-05, "loss": 0.0, "step": 227050 }, { "epoch": 48.95, "learning_rate": 1.8141966335106096e-05, "loss": 0.0, "step": 227100 }, { "epoch": 48.97, "learning_rate": 1.8136447861074602e-05, "loss": 0.0, "step": 227150 }, { "epoch": 48.98, "learning_rate": 1.8130928943082888e-05, "loss": 0.0, "step": 227200 }, { "epoch": 48.99, "learning_rate": 1.8125409581912155e-05, "loss": 0.0, "step": 227250 }, { "epoch": 49.0, "learning_rate": 1.811988977834365e-05, "loss": 0.0, "step": 227300 }, { "epoch": 49.01, "learning_rate": 1.811436953315871e-05, "loss": 0.0, "step": 227350 }, { "epoch": 49.02, "learning_rate": 1.81088488471387e-05, "loss": 0.0, "step": 227400 }, { "epoch": 49.03, "learning_rate": 1.810332772106508e-05, "loss": 0.0, "step": 227450 }, { "epoch": 49.04, "learning_rate": 1.8097806155719355e-05, "loss": 0.0, "step": 227500 }, { "epoch": 49.05, "learning_rate": 1.8092284151883096e-05, "loss": 0.0, "step": 227550 }, { "epoch": 49.06, "learning_rate": 1.8086761710337934e-05, "loss": 0.0, "step": 227600 }, { "epoch": 49.07, "learning_rate": 1.8081238831865565e-05, "loss": 0.0, "step": 227650 }, { "epoch": 49.08, "learning_rate": 1.8075715517247742e-05, "loss": 0.0, "step": 227700 }, { "epoch": 49.09, "learning_rate": 1.807019176726629e-05, "loss": 0.0, "step": 227750 }, { "epoch": 49.11, "learning_rate": 1.8064667582703082e-05, "loss": 0.0, "step": 227800 }, { "epoch": 49.12, "learning_rate": 1.805914296434006e-05, "loss": 0.0, "step": 227850 }, { "epoch": 49.13, "learning_rate": 1.8053617912959232e-05, "loss": 0.0, "step": 227900 }, { "epoch": 49.14, "learning_rate": 1.8048092429342657e-05, "loss": 0.0, "step": 227950 }, { "epoch": 49.15, "learning_rate": 1.8042566514272462e-05, "loss": 0.0, "step": 228000 }, { "epoch": 49.16, "learning_rate": 1.803704016853083e-05, "loss": 0.0, "step": 228050 }, { "epoch": 49.17, "learning_rate": 1.8031513392900018e-05, "loss": 0.0, "step": 228100 }, { "epoch": 49.18, "learning_rate": 1.802598618816232e-05, "loss": 0.0, "step": 228150 }, { "epoch": 49.19, "learning_rate": 1.8020458555100114e-05, "loss": 0.0, "step": 228200 }, { "epoch": 49.2, "learning_rate": 1.8014930494495834e-05, "loss": 0.0, "step": 228250 }, { "epoch": 49.21, "learning_rate": 1.800940200713196e-05, "loss": 0.0, "step": 228300 }, { "epoch": 49.22, "learning_rate": 1.8003873093791045e-05, "loss": 0.0, "step": 228350 }, { "epoch": 49.23, "learning_rate": 1.7998343755255705e-05, "loss": 0.0, "step": 228400 }, { "epoch": 49.25, "learning_rate": 1.7992813992308602e-05, "loss": 0.0, "step": 228450 }, { "epoch": 49.26, "learning_rate": 1.7987283805732474e-05, "loss": 0.0, "step": 228500 }, { "epoch": 49.27, "learning_rate": 1.7981753196310107e-05, "loss": 0.0, "step": 228550 }, { "epoch": 49.28, "learning_rate": 1.7976222164824357e-05, "loss": 0.0, "step": 228600 }, { "epoch": 49.29, "learning_rate": 1.797069071205813e-05, "loss": 0.0, "step": 228650 }, { "epoch": 49.3, "learning_rate": 1.7965158838794397e-05, "loss": 0.0, "step": 228700 }, { "epoch": 49.31, "learning_rate": 1.795962654581619e-05, "loss": 0.0, "step": 228750 }, { "epoch": 49.32, "learning_rate": 1.7954093833906592e-05, "loss": 0.0, "step": 228800 }, { "epoch": 49.33, "learning_rate": 1.7948560703848767e-05, "loss": 0.0, "step": 228850 }, { "epoch": 49.34, "learning_rate": 1.7943027156425902e-05, "loss": 0.0, "step": 228900 }, { "epoch": 49.35, "learning_rate": 1.793749319242128e-05, "loss": 0.0, "step": 228950 }, { "epoch": 49.36, "learning_rate": 1.7931958812618208e-05, "loss": 0.0, "step": 229000 }, { "epoch": 49.37, "learning_rate": 1.7926424017800093e-05, "loss": 0.0, "step": 229050 }, { "epoch": 49.39, "learning_rate": 1.7920888808750365e-05, "loss": 0.0, "step": 229100 }, { "epoch": 49.4, "learning_rate": 1.7915353186252526e-05, "loss": 0.0, "step": 229150 }, { "epoch": 49.41, "learning_rate": 1.790981715109014e-05, "loss": 0.0, "step": 229200 }, { "epoch": 49.42, "learning_rate": 1.7904280704046822e-05, "loss": 0.0, "step": 229250 }, { "epoch": 49.43, "learning_rate": 1.7898743845906254e-05, "loss": 0.0, "step": 229300 }, { "epoch": 49.44, "learning_rate": 1.789320657745217e-05, "loss": 0.0, "step": 229350 }, { "epoch": 49.45, "learning_rate": 1.788766889946836e-05, "loss": 0.0, "step": 229400 }, { "epoch": 49.46, "learning_rate": 1.788213081273868e-05, "loss": 0.0, "step": 229450 }, { "epoch": 49.47, "learning_rate": 1.787659231804703e-05, "loss": 0.0, "step": 229500 }, { "epoch": 49.48, "learning_rate": 1.7871053416177388e-05, "loss": 0.0, "step": 229550 }, { "epoch": 49.49, "learning_rate": 1.786551410791377e-05, "loss": 0.0, "step": 229600 }, { "epoch": 49.5, "learning_rate": 1.7859974394040263e-05, "loss": 0.0, "step": 229650 }, { "epoch": 49.51, "learning_rate": 1.785443427534101e-05, "loss": 0.0, "step": 229700 }, { "epoch": 49.53, "learning_rate": 1.7848893752600197e-05, "loss": 0.0, "step": 229750 }, { "epoch": 49.54, "learning_rate": 1.784335282660209e-05, "loss": 0.0, "step": 229800 }, { "epoch": 49.55, "learning_rate": 1.7837811498130987e-05, "loss": 0.0, "step": 229850 }, { "epoch": 49.56, "learning_rate": 1.783226976797127e-05, "loss": 0.0, "step": 229900 }, { "epoch": 49.57, "learning_rate": 1.7826727636907353e-05, "loss": 0.0, "step": 229950 }, { "epoch": 49.58, "learning_rate": 1.7821185105723726e-05, "loss": 0.0001, "step": 230000 }, { "epoch": 49.59, "learning_rate": 1.781564217520492e-05, "loss": 0.0, "step": 230050 }, { "epoch": 49.6, "learning_rate": 1.781009884613553e-05, "loss": 0.0, "step": 230100 }, { "epoch": 49.61, "learning_rate": 1.7804555119300216e-05, "loss": 0.0, "step": 230150 }, { "epoch": 49.62, "learning_rate": 1.7799010995483675e-05, "loss": 0.0, "step": 230200 }, { "epoch": 49.63, "learning_rate": 1.7793466475470678e-05, "loss": 0.0, "step": 230250 }, { "epoch": 49.64, "learning_rate": 1.778792156004604e-05, "loss": 0.0, "step": 230300 }, { "epoch": 49.66, "learning_rate": 1.7782376249994638e-05, "loss": 0.0001, "step": 230350 }, { "epoch": 49.67, "learning_rate": 1.7776830546101404e-05, "loss": 0.0, "step": 230400 }, { "epoch": 49.68, "learning_rate": 1.777128444915133e-05, "loss": 0.0, "step": 230450 }, { "epoch": 49.69, "learning_rate": 1.7765737959929445e-05, "loss": 0.0, "step": 230500 }, { "epoch": 49.7, "learning_rate": 1.776019107922087e-05, "loss": 0.0, "step": 230550 }, { "epoch": 49.71, "learning_rate": 1.7754643807810736e-05, "loss": 0.0, "step": 230600 }, { "epoch": 49.72, "learning_rate": 1.7749096146484267e-05, "loss": 0.0, "step": 230650 }, { "epoch": 49.73, "learning_rate": 1.7743548096026724e-05, "loss": 0.0, "step": 230700 }, { "epoch": 49.74, "learning_rate": 1.773799965722342e-05, "loss": 0.0, "step": 230750 }, { "epoch": 49.75, "learning_rate": 1.773245083085974e-05, "loss": 0.0, "step": 230800 }, { "epoch": 49.76, "learning_rate": 1.7726901617721105e-05, "loss": 0.0, "step": 230850 }, { "epoch": 49.77, "learning_rate": 1.7721352018593e-05, "loss": 0.0, "step": 230900 }, { "epoch": 49.78, "learning_rate": 1.771580203426097e-05, "loss": 0.0, "step": 230950 }, { "epoch": 49.8, "learning_rate": 1.7710251665510607e-05, "loss": 0.0, "step": 231000 }, { "epoch": 49.81, "learning_rate": 1.770470091312755e-05, "loss": 0.0, "step": 231050 }, { "epoch": 49.82, "learning_rate": 1.769914977789751e-05, "loss": 0.0, "step": 231100 }, { "epoch": 49.83, "learning_rate": 1.7693598260606238e-05, "loss": 0.0, "step": 231150 }, { "epoch": 49.84, "learning_rate": 1.7688046362039547e-05, "loss": 0.0, "step": 231200 }, { "epoch": 49.85, "learning_rate": 1.7682494082983302e-05, "loss": 0.0, "step": 231250 }, { "epoch": 49.86, "learning_rate": 1.767694142422342e-05, "loss": 0.0, "step": 231300 }, { "epoch": 49.87, "learning_rate": 1.7671388386545863e-05, "loss": 0.0, "step": 231350 }, { "epoch": 49.88, "learning_rate": 1.7665834970736677e-05, "loss": 0.0, "step": 231400 }, { "epoch": 49.89, "learning_rate": 1.7660281177581923e-05, "loss": 0.0, "step": 231450 }, { "epoch": 49.9, "learning_rate": 1.765472700786775e-05, "loss": 0.0, "step": 231500 }, { "epoch": 49.91, "learning_rate": 1.7649172462380318e-05, "loss": 0.0, "step": 231550 }, { "epoch": 49.92, "learning_rate": 1.764361754190589e-05, "loss": 0.0, "step": 231600 }, { "epoch": 49.94, "learning_rate": 1.7638062247230753e-05, "loss": 0.0, "step": 231650 }, { "epoch": 49.95, "learning_rate": 1.7632506579141245e-05, "loss": 0.0, "step": 231700 }, { "epoch": 49.96, "learning_rate": 1.7626950538423766e-05, "loss": 0.0, "step": 231750 }, { "epoch": 49.97, "learning_rate": 1.762139412586477e-05, "loss": 0.0, "step": 231800 }, { "epoch": 49.98, "learning_rate": 1.7615837342250757e-05, "loss": 0.0, "step": 231850 }, { "epoch": 49.99, "learning_rate": 1.7610280188368284e-05, "loss": 0.0, "step": 231900 }, { "epoch": 50.0, "learning_rate": 1.760472266500396e-05, "loss": 0.0, "step": 231950 }, { "epoch": 50.01, "learning_rate": 1.759916477294444e-05, "loss": 0.0, "step": 232000 }, { "epoch": 50.02, "learning_rate": 1.7593606512976445e-05, "loss": 0.0, "step": 232050 }, { "epoch": 50.03, "learning_rate": 1.758804788588673e-05, "loss": 0.0, "step": 232100 }, { "epoch": 50.04, "learning_rate": 1.758248889246212e-05, "loss": 0.0, "step": 232150 }, { "epoch": 50.05, "learning_rate": 1.757692953348948e-05, "loss": 0.0, "step": 232200 }, { "epoch": 50.06, "learning_rate": 1.7571369809755733e-05, "loss": 0.0, "step": 232250 }, { "epoch": 50.08, "learning_rate": 1.7565809722047844e-05, "loss": 0.0, "step": 232300 }, { "epoch": 50.09, "learning_rate": 1.7560249271152847e-05, "loss": 0.0, "step": 232350 }, { "epoch": 50.1, "learning_rate": 1.7554688457857805e-05, "loss": 0.0, "step": 232400 }, { "epoch": 50.11, "learning_rate": 1.754912728294985e-05, "loss": 0.0, "step": 232450 }, { "epoch": 50.12, "learning_rate": 1.7543565747216162e-05, "loss": 0.0, "step": 232500 }, { "epoch": 50.13, "learning_rate": 1.7538003851443965e-05, "loss": 0.0, "step": 232550 }, { "epoch": 50.14, "learning_rate": 1.7532441596420537e-05, "loss": 0.0, "step": 232600 }, { "epoch": 50.15, "learning_rate": 1.7526878982933214e-05, "loss": 0.0, "step": 232650 }, { "epoch": 50.16, "learning_rate": 1.7521316011769375e-05, "loss": 0.0, "step": 232700 }, { "epoch": 50.17, "learning_rate": 1.7515752683716448e-05, "loss": 0.0, "step": 232750 }, { "epoch": 50.18, "learning_rate": 1.751018899956192e-05, "loss": 0.0, "step": 232800 }, { "epoch": 50.19, "learning_rate": 1.7504624960093317e-05, "loss": 0.0, "step": 232850 }, { "epoch": 50.2, "learning_rate": 1.7499060566098225e-05, "loss": 0.0, "step": 232900 }, { "epoch": 50.22, "learning_rate": 1.7493495818364278e-05, "loss": 0.0, "step": 232950 }, { "epoch": 50.23, "learning_rate": 1.748793071767916e-05, "loss": 0.0, "step": 233000 }, { "epoch": 50.24, "learning_rate": 1.74823652648306e-05, "loss": 0.0, "step": 233050 }, { "epoch": 50.25, "learning_rate": 1.7476799460606385e-05, "loss": 0.0, "step": 233100 }, { "epoch": 50.26, "learning_rate": 1.747123330579434e-05, "loss": 0.0, "step": 233150 }, { "epoch": 50.27, "learning_rate": 1.746566680118236e-05, "loss": 0.0, "step": 233200 }, { "epoch": 50.28, "learning_rate": 1.7460099947558366e-05, "loss": 0.0, "step": 233250 }, { "epoch": 50.29, "learning_rate": 1.7454532745710342e-05, "loss": 0.0, "step": 233300 }, { "epoch": 50.3, "learning_rate": 1.744896519642632e-05, "loss": 0.0, "step": 233350 }, { "epoch": 50.31, "learning_rate": 1.744339730049438e-05, "loss": 0.0, "step": 233400 }, { "epoch": 50.32, "learning_rate": 1.7437829058702646e-05, "loss": 0.0, "step": 233450 }, { "epoch": 50.33, "learning_rate": 1.74322604718393e-05, "loss": 0.0, "step": 233500 }, { "epoch": 50.34, "learning_rate": 1.7426691540692568e-05, "loss": 0.0, "step": 233550 }, { "epoch": 50.36, "learning_rate": 1.742112226605072e-05, "loss": 0.0, "step": 233600 }, { "epoch": 50.37, "learning_rate": 1.741555264870209e-05, "loss": 0.0, "step": 233650 }, { "epoch": 50.38, "learning_rate": 1.740998268943504e-05, "loss": 0.0, "step": 233700 }, { "epoch": 50.39, "learning_rate": 1.7404412389037993e-05, "loss": 0.0, "step": 233750 }, { "epoch": 50.4, "learning_rate": 1.7398841748299425e-05, "loss": 0.0, "step": 233800 }, { "epoch": 50.41, "learning_rate": 1.7393270768007846e-05, "loss": 0.0, "step": 233850 }, { "epoch": 50.42, "learning_rate": 1.738769944895182e-05, "loss": 0.0, "step": 233900 }, { "epoch": 50.43, "learning_rate": 1.7382127791919967e-05, "loss": 0.0, "step": 233950 }, { "epoch": 50.44, "learning_rate": 1.737655579770094e-05, "loss": 0.0, "step": 234000 }, { "epoch": 50.45, "learning_rate": 1.7370983467083458e-05, "loss": 0.0, "step": 234050 }, { "epoch": 50.46, "learning_rate": 1.7365410800856264e-05, "loss": 0.0, "step": 234100 }, { "epoch": 50.47, "learning_rate": 1.7359837799808174e-05, "loss": 0.0, "step": 234150 }, { "epoch": 50.49, "learning_rate": 1.7354264464728034e-05, "loss": 0.0, "step": 234200 }, { "epoch": 50.5, "learning_rate": 1.7348690796404742e-05, "loss": 0.0, "step": 234250 }, { "epoch": 50.51, "learning_rate": 1.7343116795627247e-05, "loss": 0.0, "step": 234300 }, { "epoch": 50.52, "learning_rate": 1.7337542463184537e-05, "loss": 0.0, "step": 234350 }, { "epoch": 50.53, "learning_rate": 1.7331967799865656e-05, "loss": 0.0, "step": 234400 }, { "epoch": 50.54, "learning_rate": 1.7326392806459685e-05, "loss": 0.0, "step": 234450 }, { "epoch": 50.55, "learning_rate": 1.7320817483755765e-05, "loss": 0.0, "step": 234500 }, { "epoch": 50.56, "learning_rate": 1.731524183254307e-05, "loss": 0.0, "step": 234550 }, { "epoch": 50.57, "learning_rate": 1.7309665853610824e-05, "loss": 0.0, "step": 234600 }, { "epoch": 50.58, "learning_rate": 1.7304089547748308e-05, "loss": 0.0, "step": 234650 }, { "epoch": 50.59, "learning_rate": 1.7298512915744836e-05, "loss": 0.0, "step": 234700 }, { "epoch": 50.6, "learning_rate": 1.7292935958389774e-05, "loss": 0.0, "step": 234750 }, { "epoch": 50.61, "learning_rate": 1.7287358676472535e-05, "loss": 0.0, "step": 234800 }, { "epoch": 50.63, "learning_rate": 1.7281781070782568e-05, "loss": 0.0, "step": 234850 }, { "epoch": 50.64, "learning_rate": 1.727620314210939e-05, "loss": 0.0, "step": 234900 }, { "epoch": 50.65, "learning_rate": 1.7270624891242537e-05, "loss": 0.0, "step": 234950 }, { "epoch": 50.66, "learning_rate": 1.7265046318971612e-05, "loss": 0.0, "step": 235000 }, { "epoch": 50.67, "learning_rate": 1.725946742608625e-05, "loss": 0.0, "step": 235050 }, { "epoch": 50.68, "learning_rate": 1.725388821337614e-05, "loss": 0.0, "step": 235100 }, { "epoch": 50.69, "learning_rate": 1.724830868163101e-05, "loss": 0.0, "step": 235150 }, { "epoch": 50.7, "learning_rate": 1.724272883164063e-05, "loss": 0.0, "step": 235200 }, { "epoch": 50.71, "learning_rate": 1.7237148664194827e-05, "loss": 0.0, "step": 235250 }, { "epoch": 50.72, "learning_rate": 1.723156818008347e-05, "loss": 0.0, "step": 235300 }, { "epoch": 50.73, "learning_rate": 1.7225987380096465e-05, "loss": 0.0, "step": 235350 }, { "epoch": 50.74, "learning_rate": 1.7220406265023767e-05, "loss": 0.0, "step": 235400 }, { "epoch": 50.75, "learning_rate": 1.7214824835655376e-05, "loss": 0.0, "step": 235450 }, { "epoch": 50.77, "learning_rate": 1.7209243092781338e-05, "loss": 0.0, "step": 235500 }, { "epoch": 50.78, "learning_rate": 1.720366103719174e-05, "loss": 0.0, "step": 235550 }, { "epoch": 50.79, "learning_rate": 1.719807866967671e-05, "loss": 0.0, "step": 235600 }, { "epoch": 50.8, "learning_rate": 1.719249599102644e-05, "loss": 0.0, "step": 235650 }, { "epoch": 50.81, "learning_rate": 1.7186913002031133e-05, "loss": 0.0, "step": 235700 }, { "epoch": 50.82, "learning_rate": 1.718132970348107e-05, "loss": 0.0, "step": 235750 }, { "epoch": 50.83, "learning_rate": 1.7175746096166542e-05, "loss": 0.0, "step": 235800 }, { "epoch": 50.84, "learning_rate": 1.717016218087792e-05, "loss": 0.0, "step": 235850 }, { "epoch": 50.85, "learning_rate": 1.7164577958405588e-05, "loss": 0.0, "step": 235900 }, { "epoch": 50.86, "learning_rate": 1.715899342953999e-05, "loss": 0.0, "step": 235950 }, { "epoch": 50.87, "learning_rate": 1.715340859507161e-05, "loss": 0.0, "step": 236000 }, { "epoch": 50.88, "learning_rate": 1.714782345579097e-05, "loss": 0.0, "step": 236050 }, { "epoch": 50.89, "learning_rate": 1.7142238012488644e-05, "loss": 0.0, "step": 236100 }, { "epoch": 50.91, "learning_rate": 1.7136652265955247e-05, "loss": 0.0, "step": 236150 }, { "epoch": 50.92, "learning_rate": 1.7131066216981424e-05, "loss": 0.0, "step": 236200 }, { "epoch": 50.93, "learning_rate": 1.712547986635788e-05, "loss": 0.0, "step": 236250 }, { "epoch": 50.94, "learning_rate": 1.7119893214875355e-05, "loss": 0.0, "step": 236300 }, { "epoch": 50.95, "learning_rate": 1.7114306263324633e-05, "loss": 0.0, "step": 236350 }, { "epoch": 50.96, "learning_rate": 1.7108719012496538e-05, "loss": 0.0, "step": 236400 }, { "epoch": 50.97, "learning_rate": 1.7103131463181936e-05, "loss": 0.0, "step": 236450 }, { "epoch": 50.98, "learning_rate": 1.7097543616171747e-05, "loss": 0.0, "step": 236500 }, { "epoch": 50.99, "learning_rate": 1.7091955472256913e-05, "loss": 0.0, "step": 236550 }, { "epoch": 51.0, "learning_rate": 1.7086367032228434e-05, "loss": 0.0, "step": 236600 }, { "epoch": 51.01, "learning_rate": 1.708077829687735e-05, "loss": 0.0, "step": 236650 }, { "epoch": 51.02, "learning_rate": 1.707518926699473e-05, "loss": 0.0001, "step": 236700 }, { "epoch": 51.03, "learning_rate": 1.7069599943371704e-05, "loss": 0.0, "step": 236750 }, { "epoch": 51.05, "learning_rate": 1.7064010326799428e-05, "loss": 0.0, "step": 236800 }, { "epoch": 51.06, "learning_rate": 1.7058420418069104e-05, "loss": 0.0, "step": 236850 }, { "epoch": 51.07, "learning_rate": 1.7052830217971985e-05, "loss": 0.0, "step": 236900 }, { "epoch": 51.08, "learning_rate": 1.7047239727299343e-05, "loss": 0.0, "step": 236950 }, { "epoch": 51.09, "learning_rate": 1.7041648946842516e-05, "loss": 0.0, "step": 237000 }, { "epoch": 51.1, "learning_rate": 1.7036057877392868e-05, "loss": 0.0, "step": 237050 }, { "epoch": 51.11, "learning_rate": 1.7030466519741808e-05, "loss": 0.0, "step": 237100 }, { "epoch": 51.12, "learning_rate": 1.7024874874680786e-05, "loss": 0.0, "step": 237150 }, { "epoch": 51.13, "learning_rate": 1.701928294300129e-05, "loss": 0.0, "step": 237200 }, { "epoch": 51.14, "learning_rate": 1.701369072549485e-05, "loss": 0.0, "step": 237250 }, { "epoch": 51.15, "learning_rate": 1.7008098222953043e-05, "loss": 0.0, "step": 237300 }, { "epoch": 51.16, "learning_rate": 1.7002505436167482e-05, "loss": 0.0, "step": 237350 }, { "epoch": 51.17, "learning_rate": 1.699691236592981e-05, "loss": 0.0, "step": 237400 }, { "epoch": 51.19, "learning_rate": 1.6991319013031727e-05, "loss": 0.0, "step": 237450 }, { "epoch": 51.2, "learning_rate": 1.6985725378264956e-05, "loss": 0.0, "step": 237500 }, { "epoch": 51.21, "learning_rate": 1.6980131462421278e-05, "loss": 0.0, "step": 237550 }, { "epoch": 51.22, "learning_rate": 1.6974537266292504e-05, "loss": 0.0, "step": 237600 }, { "epoch": 51.23, "learning_rate": 1.6968942790670475e-05, "loss": 0.0, "step": 237650 }, { "epoch": 51.24, "learning_rate": 1.6963348036347097e-05, "loss": 0.0, "step": 237700 }, { "epoch": 51.25, "learning_rate": 1.6957753004114287e-05, "loss": 0.0, "step": 237750 }, { "epoch": 51.26, "learning_rate": 1.6952157694764026e-05, "loss": 0.0, "step": 237800 }, { "epoch": 51.27, "learning_rate": 1.6946562109088314e-05, "loss": 0.0, "step": 237850 }, { "epoch": 51.28, "learning_rate": 1.69409662478792e-05, "loss": 0.0, "step": 237900 }, { "epoch": 51.29, "learning_rate": 1.693537011192878e-05, "loss": 0.0, "step": 237950 }, { "epoch": 51.3, "learning_rate": 1.692977370202917e-05, "loss": 0.0, "step": 238000 }, { "epoch": 51.31, "learning_rate": 1.6924177018972535e-05, "loss": 0.0, "step": 238050 }, { "epoch": 51.33, "learning_rate": 1.6918580063551093e-05, "loss": 0.0, "step": 238100 }, { "epoch": 51.34, "learning_rate": 1.6912982836557064e-05, "loss": 0.0, "step": 238150 }, { "epoch": 51.35, "learning_rate": 1.6907385338782746e-05, "loss": 0.0, "step": 238200 }, { "epoch": 51.36, "learning_rate": 1.6901787571020444e-05, "loss": 0.0, "step": 238250 }, { "epoch": 51.37, "learning_rate": 1.6896189534062523e-05, "loss": 0.0, "step": 238300 }, { "epoch": 51.38, "learning_rate": 1.689059122870138e-05, "loss": 0.0, "step": 238350 }, { "epoch": 51.39, "learning_rate": 1.6884992655729445e-05, "loss": 0.0, "step": 238400 }, { "epoch": 51.4, "learning_rate": 1.6879393815939188e-05, "loss": 0.0, "step": 238450 }, { "epoch": 51.41, "learning_rate": 1.6873794710123117e-05, "loss": 0.0, "step": 238500 }, { "epoch": 51.42, "learning_rate": 1.686819533907378e-05, "loss": 0.0, "step": 238550 }, { "epoch": 51.43, "learning_rate": 1.686259570358376e-05, "loss": 0.0, "step": 238600 }, { "epoch": 51.44, "learning_rate": 1.685699580444568e-05, "loss": 0.0, "step": 238650 }, { "epoch": 51.46, "learning_rate": 1.6851395642452196e-05, "loss": 0.0, "step": 238700 }, { "epoch": 51.47, "learning_rate": 1.684579521839601e-05, "loss": 0.0, "step": 238750 }, { "epoch": 51.48, "learning_rate": 1.6840194533069843e-05, "loss": 0.0, "step": 238800 }, { "epoch": 51.49, "learning_rate": 1.683459358726648e-05, "loss": 0.0, "step": 238850 }, { "epoch": 51.5, "learning_rate": 1.682899238177871e-05, "loss": 0.0, "step": 238900 }, { "epoch": 51.51, "learning_rate": 1.6823390917399398e-05, "loss": 0.0, "step": 238950 }, { "epoch": 51.52, "learning_rate": 1.6817789194921404e-05, "loss": 0.0, "step": 239000 }, { "epoch": 51.53, "learning_rate": 1.681218721513766e-05, "loss": 0.0, "step": 239050 }, { "epoch": 51.54, "learning_rate": 1.6806584978841105e-05, "loss": 0.0, "step": 239100 }, { "epoch": 51.55, "learning_rate": 1.6800982486824737e-05, "loss": 0.0, "step": 239150 }, { "epoch": 51.56, "learning_rate": 1.6795379739881586e-05, "loss": 0.0, "step": 239200 }, { "epoch": 51.57, "learning_rate": 1.678977673880471e-05, "loss": 0.0, "step": 239250 }, { "epoch": 51.58, "learning_rate": 1.6784173484387203e-05, "loss": 0.0, "step": 239300 }, { "epoch": 51.6, "learning_rate": 1.6778569977422195e-05, "loss": 0.0, "step": 239350 }, { "epoch": 51.61, "learning_rate": 1.6772966218702864e-05, "loss": 0.0, "step": 239400 }, { "epoch": 51.62, "learning_rate": 1.6767362209022413e-05, "loss": 0.0, "step": 239450 }, { "epoch": 51.63, "learning_rate": 1.676175794917408e-05, "loss": 0.0, "step": 239500 }, { "epoch": 51.64, "learning_rate": 1.6756153439951144e-05, "loss": 0.0, "step": 239550 }, { "epoch": 51.65, "learning_rate": 1.6750548682146915e-05, "loss": 0.0, "step": 239600 }, { "epoch": 51.66, "learning_rate": 1.6744943676554733e-05, "loss": 0.0, "step": 239650 }, { "epoch": 51.67, "learning_rate": 1.673933842396799e-05, "loss": 0.0, "step": 239700 }, { "epoch": 51.68, "learning_rate": 1.6733732925180095e-05, "loss": 0.0, "step": 239750 }, { "epoch": 51.69, "learning_rate": 1.6728127180984507e-05, "loss": 0.0, "step": 239800 }, { "epoch": 51.7, "learning_rate": 1.67225211921747e-05, "loss": 0.0, "step": 239850 }, { "epoch": 51.71, "learning_rate": 1.671691495954421e-05, "loss": 0.0, "step": 239900 }, { "epoch": 51.72, "learning_rate": 1.6711308483886574e-05, "loss": 0.0, "step": 239950 }, { "epoch": 51.74, "learning_rate": 1.6705701765995395e-05, "loss": 0.0, "step": 240000 }, { "epoch": 51.75, "learning_rate": 1.67000948066643e-05, "loss": 0.0, "step": 240050 }, { "epoch": 51.76, "learning_rate": 1.6694487606686927e-05, "loss": 0.0, "step": 240100 }, { "epoch": 51.77, "learning_rate": 1.6688880166856994e-05, "loss": 0.0, "step": 240150 }, { "epoch": 51.78, "learning_rate": 1.6683272487968203e-05, "loss": 0.0, "step": 240200 }, { "epoch": 51.79, "learning_rate": 1.6677664570814327e-05, "loss": 0.0, "step": 240250 }, { "epoch": 51.8, "learning_rate": 1.6672056416189163e-05, "loss": 0.0, "step": 240300 }, { "epoch": 51.81, "learning_rate": 1.6666448024886527e-05, "loss": 0.0, "step": 240350 }, { "epoch": 51.82, "learning_rate": 1.666083939770029e-05, "loss": 0.0, "step": 240400 }, { "epoch": 51.83, "learning_rate": 1.6655230535424334e-05, "loss": 0.0, "step": 240450 }, { "epoch": 51.84, "learning_rate": 1.66496214388526e-05, "loss": 0.0, "step": 240500 }, { "epoch": 51.85, "learning_rate": 1.664401210877904e-05, "loss": 0.0, "step": 240550 }, { "epoch": 51.86, "learning_rate": 1.663840254599765e-05, "loss": 0.0, "step": 240600 }, { "epoch": 51.88, "learning_rate": 1.663279275130246e-05, "loss": 0.0, "step": 240650 }, { "epoch": 51.89, "learning_rate": 1.6627182725487514e-05, "loss": 0.0, "step": 240700 }, { "epoch": 51.9, "learning_rate": 1.662157246934693e-05, "loss": 0.0, "step": 240750 }, { "epoch": 51.91, "learning_rate": 1.6615961983674803e-05, "loss": 0.0, "step": 240800 }, { "epoch": 51.92, "learning_rate": 1.6610351269265315e-05, "loss": 0.0, "step": 240850 }, { "epoch": 51.93, "learning_rate": 1.6604740326912647e-05, "loss": 0.0, "step": 240900 }, { "epoch": 51.94, "learning_rate": 1.6599129157411015e-05, "loss": 0.0, "step": 240950 }, { "epoch": 51.95, "learning_rate": 1.659351776155468e-05, "loss": 0.0, "step": 241000 }, { "epoch": 51.96, "learning_rate": 1.6587906140137926e-05, "loss": 0.0, "step": 241050 }, { "epoch": 51.97, "learning_rate": 1.658229429395507e-05, "loss": 0.0, "step": 241100 }, { "epoch": 51.98, "learning_rate": 1.6576682223800458e-05, "loss": 0.0, "step": 241150 }, { "epoch": 51.99, "learning_rate": 1.6571069930468484e-05, "loss": 0.0, "step": 241200 }, { "epoch": 52.0, "learning_rate": 1.6565457414753547e-05, "loss": 0.0, "step": 241250 }, { "epoch": 52.02, "learning_rate": 1.6559844677450098e-05, "loss": 0.0, "step": 241300 }, { "epoch": 52.03, "learning_rate": 1.6554231719352613e-05, "loss": 0.0, "step": 241350 }, { "epoch": 52.04, "learning_rate": 1.6548618541255594e-05, "loss": 0.0, "step": 241400 }, { "epoch": 52.05, "learning_rate": 1.6543005143953582e-05, "loss": 0.0, "step": 241450 }, { "epoch": 52.06, "learning_rate": 1.6537391528241155e-05, "loss": 0.0, "step": 241500 }, { "epoch": 52.07, "learning_rate": 1.65317776949129e-05, "loss": 0.0001, "step": 241550 }, { "epoch": 52.08, "learning_rate": 1.6526163644763458e-05, "loss": 0.0, "step": 241600 }, { "epoch": 52.09, "learning_rate": 1.652054937858748e-05, "loss": 0.0, "step": 241650 }, { "epoch": 52.1, "learning_rate": 1.6514934897179673e-05, "loss": 0.0, "step": 241700 }, { "epoch": 52.11, "learning_rate": 1.650932020133475e-05, "loss": 0.0, "step": 241750 }, { "epoch": 52.12, "learning_rate": 1.6503705291847458e-05, "loss": 0.0, "step": 241800 }, { "epoch": 52.13, "learning_rate": 1.6498090169512597e-05, "loss": 0.0, "step": 241850 }, { "epoch": 52.14, "learning_rate": 1.6492474835124966e-05, "loss": 0.0, "step": 241900 }, { "epoch": 52.16, "learning_rate": 1.6486859289479416e-05, "loss": 0.0, "step": 241950 }, { "epoch": 52.17, "learning_rate": 1.6481243533370814e-05, "loss": 0.0, "step": 242000 }, { "epoch": 52.18, "learning_rate": 1.647562756759407e-05, "loss": 0.0, "step": 242050 }, { "epoch": 52.19, "learning_rate": 1.6470011392944117e-05, "loss": 0.0, "step": 242100 }, { "epoch": 52.2, "learning_rate": 1.646439501021591e-05, "loss": 0.0, "step": 242150 }, { "epoch": 52.21, "learning_rate": 1.6458778420204448e-05, "loss": 0.0, "step": 242200 }, { "epoch": 52.22, "learning_rate": 1.6453161623704745e-05, "loss": 0.0, "step": 242250 }, { "epoch": 52.23, "learning_rate": 1.644754462151186e-05, "loss": 0.0, "step": 242300 }, { "epoch": 52.24, "learning_rate": 1.644192741442087e-05, "loss": 0.0, "step": 242350 }, { "epoch": 52.25, "learning_rate": 1.6436310003226878e-05, "loss": 0.0, "step": 242400 }, { "epoch": 52.26, "learning_rate": 1.643069238872503e-05, "loss": 0.0, "step": 242450 }, { "epoch": 52.27, "learning_rate": 1.6425074571710483e-05, "loss": 0.0, "step": 242500 }, { "epoch": 52.28, "learning_rate": 1.6419456552978437e-05, "loss": 0.0, "step": 242550 }, { "epoch": 52.3, "learning_rate": 1.6413838333324117e-05, "loss": 0.0, "step": 242600 }, { "epoch": 52.31, "learning_rate": 1.6408219913542772e-05, "loss": 0.0, "step": 242650 }, { "epoch": 52.32, "learning_rate": 1.6402601294429686e-05, "loss": 0.0, "step": 242700 }, { "epoch": 52.33, "learning_rate": 1.6396982476780158e-05, "loss": 0.0, "step": 242750 }, { "epoch": 52.34, "learning_rate": 1.6391363461389535e-05, "loss": 0.0, "step": 242800 }, { "epoch": 52.35, "learning_rate": 1.638574424905318e-05, "loss": 0.0, "step": 242850 }, { "epoch": 52.36, "learning_rate": 1.638012484056648e-05, "loss": 0.0, "step": 242900 }, { "epoch": 52.37, "learning_rate": 1.637450523672486e-05, "loss": 0.0, "step": 242950 }, { "epoch": 52.38, "learning_rate": 1.6368885438323766e-05, "loss": 0.0, "step": 243000 }, { "epoch": 52.39, "learning_rate": 1.6363265446158668e-05, "loss": 0.0, "step": 243050 }, { "epoch": 52.4, "learning_rate": 1.6357645261025075e-05, "loss": 0.0, "step": 243100 }, { "epoch": 52.41, "learning_rate": 1.6352024883718518e-05, "loss": 0.0, "step": 243150 }, { "epoch": 52.43, "learning_rate": 1.6346404315034556e-05, "loss": 0.0, "step": 243200 }, { "epoch": 52.44, "learning_rate": 1.6340783555768764e-05, "loss": 0.0, "step": 243250 }, { "epoch": 52.45, "learning_rate": 1.6335162606716766e-05, "loss": 0.0, "step": 243300 }, { "epoch": 52.46, "learning_rate": 1.6329541468674184e-05, "loss": 0.0, "step": 243350 }, { "epoch": 52.47, "learning_rate": 1.6323920142436697e-05, "loss": 0.0, "step": 243400 }, { "epoch": 52.48, "learning_rate": 1.6318298628799994e-05, "loss": 0.0, "step": 243450 }, { "epoch": 52.49, "learning_rate": 1.6312676928559785e-05, "loss": 0.0, "step": 243500 }, { "epoch": 52.5, "learning_rate": 1.6307055042511827e-05, "loss": 0.0, "step": 243550 }, { "epoch": 52.51, "learning_rate": 1.6301432971451883e-05, "loss": 0.0, "step": 243600 }, { "epoch": 52.52, "learning_rate": 1.6295810716175747e-05, "loss": 0.0, "step": 243650 }, { "epoch": 52.53, "learning_rate": 1.6290188277479248e-05, "loss": 0.0, "step": 243700 }, { "epoch": 52.54, "learning_rate": 1.6284565656158234e-05, "loss": 0.0, "step": 243750 }, { "epoch": 52.55, "learning_rate": 1.6278942853008584e-05, "loss": 0.0, "step": 243800 }, { "epoch": 52.57, "learning_rate": 1.627331986882619e-05, "loss": 0.0, "step": 243850 }, { "epoch": 52.58, "learning_rate": 1.6267696704406984e-05, "loss": 0.0, "step": 243900 }, { "epoch": 52.59, "learning_rate": 1.6262073360546914e-05, "loss": 0.0, "step": 243950 }, { "epoch": 52.6, "learning_rate": 1.6256449838041957e-05, "loss": 0.0, "step": 244000 }, { "epoch": 52.61, "learning_rate": 1.6250826137688124e-05, "loss": 0.0, "step": 244050 }, { "epoch": 52.62, "learning_rate": 1.6245202260281434e-05, "loss": 0.0, "step": 244100 }, { "epoch": 52.63, "learning_rate": 1.6239578206617946e-05, "loss": 0.0, "step": 244150 }, { "epoch": 52.64, "learning_rate": 1.623395397749373e-05, "loss": 0.0, "step": 244200 }, { "epoch": 52.65, "learning_rate": 1.622832957370489e-05, "loss": 0.0, "step": 244250 }, { "epoch": 52.66, "learning_rate": 1.6222704996047558e-05, "loss": 0.0, "step": 244300 }, { "epoch": 52.67, "learning_rate": 1.6217080245317885e-05, "loss": 0.0, "step": 244350 }, { "epoch": 52.68, "learning_rate": 1.621145532231204e-05, "loss": 0.0, "step": 244400 }, { "epoch": 52.69, "learning_rate": 1.6205830227826234e-05, "loss": 0.0, "step": 244450 }, { "epoch": 52.71, "learning_rate": 1.620020496265669e-05, "loss": 0.0, "step": 244500 }, { "epoch": 52.72, "learning_rate": 1.6194579527599653e-05, "loss": 0.0, "step": 244550 }, { "epoch": 52.73, "learning_rate": 1.61889539234514e-05, "loss": 0.0, "step": 244600 }, { "epoch": 52.74, "learning_rate": 1.6183328151008222e-05, "loss": 0.0, "step": 244650 }, { "epoch": 52.75, "learning_rate": 1.6177702211066446e-05, "loss": 0.0, "step": 244700 }, { "epoch": 52.76, "learning_rate": 1.6172076104422413e-05, "loss": 0.0, "step": 244750 }, { "epoch": 52.77, "learning_rate": 1.6166449831872496e-05, "loss": 0.0, "step": 244800 }, { "epoch": 52.78, "learning_rate": 1.6160823394213077e-05, "loss": 0.0, "step": 244850 }, { "epoch": 52.79, "learning_rate": 1.6155196792240588e-05, "loss": 0.0, "step": 244900 }, { "epoch": 52.8, "learning_rate": 1.6149570026751452e-05, "loss": 0.0, "step": 244950 }, { "epoch": 52.81, "learning_rate": 1.6143943098542137e-05, "loss": 0.0, "step": 245000 }, { "epoch": 52.82, "learning_rate": 1.613831600840912e-05, "loss": 0.0, "step": 245050 }, { "epoch": 52.83, "learning_rate": 1.613268875714892e-05, "loss": 0.0, "step": 245100 }, { "epoch": 52.85, "learning_rate": 1.6127061345558062e-05, "loss": 0.0, "step": 245150 }, { "epoch": 52.86, "learning_rate": 1.6121433774433098e-05, "loss": 0.0, "step": 245200 }, { "epoch": 52.87, "learning_rate": 1.6115806044570605e-05, "loss": 0.0, "step": 245250 }, { "epoch": 52.88, "learning_rate": 1.611017815676718e-05, "loss": 0.0, "step": 245300 }, { "epoch": 52.89, "learning_rate": 1.6104550111819447e-05, "loss": 0.0, "step": 245350 }, { "epoch": 52.9, "learning_rate": 1.6098921910524043e-05, "loss": 0.0, "step": 245400 }, { "epoch": 52.91, "learning_rate": 1.6093293553677636e-05, "loss": 0.0, "step": 245450 }, { "epoch": 52.92, "learning_rate": 1.6087665042076913e-05, "loss": 0.0, "step": 245500 }, { "epoch": 52.93, "learning_rate": 1.6082036376518577e-05, "loss": 0.0, "step": 245550 }, { "epoch": 52.94, "learning_rate": 1.607640755779937e-05, "loss": 0.0, "step": 245600 }, { "epoch": 52.95, "learning_rate": 1.6070778586716035e-05, "loss": 0.0, "step": 245650 }, { "epoch": 52.96, "learning_rate": 1.6065149464065344e-05, "loss": 0.0, "step": 245700 }, { "epoch": 52.97, "learning_rate": 1.6059520190644103e-05, "loss": 0.0, "step": 245750 }, { "epoch": 52.99, "learning_rate": 1.605389076724912e-05, "loss": 0.0, "step": 245800 }, { "epoch": 53.0, "learning_rate": 1.604826119467723e-05, "loss": 0.0, "step": 245850 }, { "epoch": 53.01, "learning_rate": 1.6042631473725306e-05, "loss": 0.0, "step": 245900 }, { "epoch": 53.02, "learning_rate": 1.6037001605190206e-05, "loss": 0.0, "step": 245950 }, { "epoch": 53.03, "learning_rate": 1.6031371589868856e-05, "loss": 0.0, "step": 246000 }, { "epoch": 53.04, "learning_rate": 1.6025741428558153e-05, "loss": 0.0, "step": 246050 }, { "epoch": 53.05, "learning_rate": 1.6020111122055055e-05, "loss": 0.0, "step": 246100 }, { "epoch": 53.06, "learning_rate": 1.6014480671156523e-05, "loss": 0.0, "step": 246150 }, { "epoch": 53.07, "learning_rate": 1.6008850076659536e-05, "loss": 0.0, "step": 246200 }, { "epoch": 53.08, "learning_rate": 1.60032193393611e-05, "loss": 0.0, "step": 246250 }, { "epoch": 53.09, "learning_rate": 1.599758846005824e-05, "loss": 0.0, "step": 246300 }, { "epoch": 53.1, "learning_rate": 1.5991957439547996e-05, "loss": 0.0, "step": 246350 }, { "epoch": 53.11, "learning_rate": 1.5986326278627433e-05, "loss": 0.0, "step": 246400 }, { "epoch": 53.13, "learning_rate": 1.598069497809363e-05, "loss": 0.0, "step": 246450 }, { "epoch": 53.14, "learning_rate": 1.5975063538743706e-05, "loss": 0.0, "step": 246500 }, { "epoch": 53.15, "learning_rate": 1.596943196137477e-05, "loss": 0.0, "step": 246550 }, { "epoch": 53.16, "learning_rate": 1.596380024678397e-05, "loss": 0.0, "step": 246600 }, { "epoch": 53.17, "learning_rate": 1.5958168395768465e-05, "loss": 0.0, "step": 246650 }, { "epoch": 53.18, "learning_rate": 1.5952536409125433e-05, "loss": 0.0, "step": 246700 }, { "epoch": 53.19, "learning_rate": 1.594690428765209e-05, "loss": 0.0, "step": 246750 }, { "epoch": 53.2, "learning_rate": 1.5941272032145635e-05, "loss": 0.0, "step": 246800 }, { "epoch": 53.21, "learning_rate": 1.5935639643403323e-05, "loss": 0.0, "step": 246850 }, { "epoch": 53.22, "learning_rate": 1.5930007122222403e-05, "loss": 0.0, "step": 246900 }, { "epoch": 53.23, "learning_rate": 1.5924374469400155e-05, "loss": 0.0, "step": 246950 }, { "epoch": 53.24, "learning_rate": 1.5918741685733872e-05, "loss": 0.0, "step": 247000 }, { "epoch": 53.26, "learning_rate": 1.5913108772020865e-05, "loss": 0.0, "step": 247050 }, { "epoch": 53.27, "learning_rate": 1.5907475729058474e-05, "loss": 0.0, "step": 247100 }, { "epoch": 53.28, "learning_rate": 1.5901842557644038e-05, "loss": 0.0, "step": 247150 }, { "epoch": 53.29, "learning_rate": 1.5896209258574937e-05, "loss": 0.0, "step": 247200 }, { "epoch": 53.3, "learning_rate": 1.5890575832648547e-05, "loss": 0.0, "step": 247250 }, { "epoch": 53.31, "learning_rate": 1.5884942280662272e-05, "loss": 0.0, "step": 247300 }, { "epoch": 53.32, "learning_rate": 1.587930860341355e-05, "loss": 0.0, "step": 247350 }, { "epoch": 53.33, "learning_rate": 1.5873674801699807e-05, "loss": 0.0, "step": 247400 }, { "epoch": 53.34, "learning_rate": 1.58680408763185e-05, "loss": 0.0, "step": 247450 }, { "epoch": 53.35, "learning_rate": 1.5862406828067106e-05, "loss": 0.0, "step": 247500 }, { "epoch": 53.36, "learning_rate": 1.5856772657743117e-05, "loss": 0.0, "step": 247550 }, { "epoch": 53.37, "learning_rate": 1.5851138366144047e-05, "loss": 0.0, "step": 247600 }, { "epoch": 53.38, "learning_rate": 1.584550395406742e-05, "loss": 0.0, "step": 247650 }, { "epoch": 53.4, "learning_rate": 1.5839869422310782e-05, "loss": 0.0, "step": 247700 }, { "epoch": 53.41, "learning_rate": 1.583423477167169e-05, "loss": 0.0, "step": 247750 }, { "epoch": 53.42, "learning_rate": 1.582860000294772e-05, "loss": 0.0, "step": 247800 }, { "epoch": 53.43, "learning_rate": 1.5822965116936475e-05, "loss": 0.0, "step": 247850 }, { "epoch": 53.44, "learning_rate": 1.581733011443556e-05, "loss": 0.0, "step": 247900 }, { "epoch": 53.45, "learning_rate": 1.5811694996242605e-05, "loss": 0.0, "step": 247950 }, { "epoch": 53.46, "learning_rate": 1.5806059763155248e-05, "loss": 0.0, "step": 248000 }, { "epoch": 53.47, "learning_rate": 1.5800424415971153e-05, "loss": 0.0, "step": 248050 }, { "epoch": 53.48, "learning_rate": 1.5794788955488e-05, "loss": 0.0, "step": 248100 }, { "epoch": 53.49, "learning_rate": 1.5789153382503473e-05, "loss": 0.0, "step": 248150 }, { "epoch": 53.5, "learning_rate": 1.578351769781529e-05, "loss": 0.0, "step": 248200 }, { "epoch": 53.51, "learning_rate": 1.5777881902221164e-05, "loss": 0.0, "step": 248250 }, { "epoch": 53.52, "learning_rate": 1.5772245996518847e-05, "loss": 0.0, "step": 248300 }, { "epoch": 53.54, "learning_rate": 1.5766609981506086e-05, "loss": 0.0, "step": 248350 }, { "epoch": 53.55, "learning_rate": 1.5760973857980648e-05, "loss": 0.0, "step": 248400 }, { "epoch": 53.56, "learning_rate": 1.575533762674033e-05, "loss": 0.0, "step": 248450 }, { "epoch": 53.57, "learning_rate": 1.5749701288582924e-05, "loss": 0.0, "step": 248500 }, { "epoch": 53.58, "learning_rate": 1.574406484430626e-05, "loss": 0.0, "step": 248550 }, { "epoch": 53.59, "learning_rate": 1.573842829470815e-05, "loss": 0.0, "step": 248600 }, { "epoch": 53.6, "learning_rate": 1.573279164058646e-05, "loss": 0.0, "step": 248650 }, { "epoch": 53.61, "learning_rate": 1.5727154882739037e-05, "loss": 0.0, "step": 248700 }, { "epoch": 53.62, "learning_rate": 1.5721518021963762e-05, "loss": 0.0, "step": 248750 }, { "epoch": 53.63, "learning_rate": 1.5715881059058526e-05, "loss": 0.0, "step": 248800 }, { "epoch": 53.64, "learning_rate": 1.5710243994821235e-05, "loss": 0.0, "step": 248850 }, { "epoch": 53.65, "learning_rate": 1.5704606830049806e-05, "loss": 0.0, "step": 248900 }, { "epoch": 53.66, "learning_rate": 1.5698969565542173e-05, "loss": 0.0, "step": 248950 }, { "epoch": 53.68, "learning_rate": 1.569333220209628e-05, "loss": 0.0, "step": 249000 }, { "epoch": 53.69, "learning_rate": 1.56876947405101e-05, "loss": 0.0, "step": 249050 }, { "epoch": 53.7, "learning_rate": 1.5682057181581596e-05, "loss": 0.0, "step": 249100 }, { "epoch": 53.71, "learning_rate": 1.5676419526108765e-05, "loss": 0.0, "step": 249150 }, { "epoch": 53.72, "learning_rate": 1.5670781774889603e-05, "loss": 0.0, "step": 249200 }, { "epoch": 53.73, "learning_rate": 1.5665143928722134e-05, "loss": 0.0, "step": 249250 }, { "epoch": 53.74, "learning_rate": 1.5659505988404388e-05, "loss": 0.0, "step": 249300 }, { "epoch": 53.75, "learning_rate": 1.5653867954734395e-05, "loss": 0.0, "step": 249350 }, { "epoch": 53.76, "learning_rate": 1.5648229828510234e-05, "loss": 0.0, "step": 249400 }, { "epoch": 53.77, "learning_rate": 1.5642591610529954e-05, "loss": 0.0, "step": 249450 }, { "epoch": 53.78, "learning_rate": 1.563695330159165e-05, "loss": 0.0, "step": 249500 }, { "epoch": 53.79, "learning_rate": 1.563131490249341e-05, "loss": 0.0, "step": 249550 }, { "epoch": 53.8, "learning_rate": 1.5625676414033353e-05, "loss": 0.0, "step": 249600 }, { "epoch": 53.82, "learning_rate": 1.5620037837009587e-05, "loss": 0.0, "step": 249650 }, { "epoch": 53.83, "learning_rate": 1.561439917222025e-05, "loss": 0.0, "step": 249700 }, { "epoch": 53.84, "learning_rate": 1.5608760420463495e-05, "loss": 0.0, "step": 249750 }, { "epoch": 53.85, "learning_rate": 1.5603121582537467e-05, "loss": 0.0, "step": 249800 }, { "epoch": 53.86, "learning_rate": 1.5597482659240347e-05, "loss": 0.0, "step": 249850 }, { "epoch": 53.87, "learning_rate": 1.559184365137031e-05, "loss": 0.0, "step": 249900 }, { "epoch": 53.88, "learning_rate": 1.5586204559725563e-05, "loss": 0.0, "step": 249950 }, { "epoch": 53.89, "learning_rate": 1.5580565385104297e-05, "loss": 0.0, "step": 250000 }, { "epoch": 53.9, "learning_rate": 1.5574926128304736e-05, "loss": 0.0, "step": 250050 }, { "epoch": 53.91, "learning_rate": 1.556928679012511e-05, "loss": 0.0, "step": 250100 }, { "epoch": 53.92, "learning_rate": 1.5563647371363665e-05, "loss": 0.0, "step": 250150 }, { "epoch": 53.93, "learning_rate": 1.555800787281864e-05, "loss": 0.0, "step": 250200 }, { "epoch": 53.94, "learning_rate": 1.555236829528832e-05, "loss": 0.0, "step": 250250 }, { "epoch": 53.96, "learning_rate": 1.554672863957096e-05, "loss": 0.0, "step": 250300 }, { "epoch": 53.97, "learning_rate": 1.5541088906464855e-05, "loss": 0.0, "step": 250350 }, { "epoch": 53.98, "learning_rate": 1.55354490967683e-05, "loss": 0.0, "step": 250400 }, { "epoch": 53.99, "learning_rate": 1.5529809211279605e-05, "loss": 0.0, "step": 250450 }, { "epoch": 54.0, "learning_rate": 1.552416925079709e-05, "loss": 0.0, "step": 250500 }, { "epoch": 54.01, "learning_rate": 1.551852921611908e-05, "loss": 0.0, "step": 250550 }, { "epoch": 54.02, "learning_rate": 1.551288910804392e-05, "loss": 0.0, "step": 250600 }, { "epoch": 54.03, "learning_rate": 1.5507248927369954e-05, "loss": 0.0, "step": 250650 }, { "epoch": 54.04, "learning_rate": 1.5501608674895543e-05, "loss": 0.0, "step": 250700 }, { "epoch": 54.05, "learning_rate": 1.5495968351419063e-05, "loss": 0.0, "step": 250750 }, { "epoch": 54.06, "learning_rate": 1.5490327957738893e-05, "loss": 0.0, "step": 250800 }, { "epoch": 54.07, "learning_rate": 1.5484687494653416e-05, "loss": 0.0, "step": 250850 }, { "epoch": 54.08, "learning_rate": 1.5479046962961046e-05, "loss": 0.0, "step": 250900 }, { "epoch": 54.1, "learning_rate": 1.5473406363460177e-05, "loss": 0.0, "step": 250950 }, { "epoch": 54.11, "learning_rate": 1.5467765696949245e-05, "loss": 0.0, "step": 251000 }, { "epoch": 54.12, "learning_rate": 1.5462124964226665e-05, "loss": 0.0, "step": 251050 }, { "epoch": 54.13, "learning_rate": 1.545648416609089e-05, "loss": 0.0, "step": 251100 }, { "epoch": 54.14, "learning_rate": 1.545084330334035e-05, "loss": 0.0, "step": 251150 }, { "epoch": 54.15, "learning_rate": 1.5445202376773517e-05, "loss": 0.0, "step": 251200 }, { "epoch": 54.16, "learning_rate": 1.543956138718885e-05, "loss": 0.0, "step": 251250 }, { "epoch": 54.17, "learning_rate": 1.5433920335384833e-05, "loss": 0.0, "step": 251300 }, { "epoch": 54.18, "learning_rate": 1.5428279222159932e-05, "loss": 0.0, "step": 251350 }, { "epoch": 54.19, "learning_rate": 1.542263804831266e-05, "loss": 0.0, "step": 251400 }, { "epoch": 54.2, "learning_rate": 1.54169968146415e-05, "loss": 0.0, "step": 251450 }, { "epoch": 54.21, "learning_rate": 1.541135552194497e-05, "loss": 0.0, "step": 251500 }, { "epoch": 54.23, "learning_rate": 1.5405714171021592e-05, "loss": 0.0, "step": 251550 }, { "epoch": 54.24, "learning_rate": 1.5400072762669887e-05, "loss": 0.0, "step": 251600 }, { "epoch": 54.25, "learning_rate": 1.539443129768839e-05, "loss": 0.0, "step": 251650 }, { "epoch": 54.26, "learning_rate": 1.5388789776875635e-05, "loss": 0.0, "step": 251700 }, { "epoch": 54.27, "learning_rate": 1.5383148201030185e-05, "loss": 0.0, "step": 251750 }, { "epoch": 54.28, "learning_rate": 1.5377506570950587e-05, "loss": 0.0, "step": 251800 }, { "epoch": 54.29, "learning_rate": 1.5371864887435425e-05, "loss": 0.0, "step": 251850 }, { "epoch": 54.3, "learning_rate": 1.5366223151283247e-05, "loss": 0.0, "step": 251900 }, { "epoch": 54.31, "learning_rate": 1.5360581363292656e-05, "loss": 0.0, "step": 251950 }, { "epoch": 54.32, "learning_rate": 1.5354939524262224e-05, "loss": 0.0, "step": 252000 }, { "epoch": 54.33, "learning_rate": 1.5349297634990552e-05, "loss": 0.0, "step": 252050 }, { "epoch": 54.34, "learning_rate": 1.534365569627625e-05, "loss": 0.0, "step": 252100 }, { "epoch": 54.35, "learning_rate": 1.533801370891791e-05, "loss": 0.0, "step": 252150 }, { "epoch": 54.37, "learning_rate": 1.5332371673714163e-05, "loss": 0.0, "step": 252200 }, { "epoch": 54.38, "learning_rate": 1.5326729591463625e-05, "loss": 0.0, "step": 252250 }, { "epoch": 54.39, "learning_rate": 1.5321087462964928e-05, "loss": 0.0, "step": 252300 }, { "epoch": 54.4, "learning_rate": 1.5315445289016707e-05, "loss": 0.0, "step": 252350 }, { "epoch": 54.41, "learning_rate": 1.5309803070417608e-05, "loss": 0.0, "step": 252400 }, { "epoch": 54.42, "learning_rate": 1.5304160807966275e-05, "loss": 0.0, "step": 252450 }, { "epoch": 54.43, "learning_rate": 1.5298518502461362e-05, "loss": 0.0, "step": 252500 }, { "epoch": 54.44, "learning_rate": 1.5292876154701537e-05, "loss": 0.0, "step": 252550 }, { "epoch": 54.45, "learning_rate": 1.5287233765485465e-05, "loss": 0.0, "step": 252600 }, { "epoch": 54.46, "learning_rate": 1.5281591335611813e-05, "loss": 0.0, "step": 252650 }, { "epoch": 54.47, "learning_rate": 1.5275948865879273e-05, "loss": 0.0, "step": 252700 }, { "epoch": 54.48, "learning_rate": 1.527030635708651e-05, "loss": 0.0, "step": 252750 }, { "epoch": 54.49, "learning_rate": 1.5264663810032233e-05, "loss": 0.0, "step": 252800 }, { "epoch": 54.51, "learning_rate": 1.525902122551512e-05, "loss": 0.0, "step": 252850 }, { "epoch": 54.52, "learning_rate": 1.525337860433389e-05, "loss": 0.0, "step": 252900 }, { "epoch": 54.53, "learning_rate": 1.5247735947287235e-05, "loss": 0.0, "step": 252950 }, { "epoch": 54.54, "learning_rate": 1.524209325517387e-05, "loss": 0.0, "step": 253000 }, { "epoch": 54.55, "learning_rate": 1.5236450528792514e-05, "loss": 0.0, "step": 253050 }, { "epoch": 54.56, "learning_rate": 1.5230807768941887e-05, "loss": 0.0, "step": 253100 }, { "epoch": 54.57, "learning_rate": 1.522516497642071e-05, "loss": 0.0, "step": 253150 }, { "epoch": 54.58, "learning_rate": 1.5219522152027715e-05, "loss": 0.0, "step": 253200 }, { "epoch": 54.59, "learning_rate": 1.5213879296561638e-05, "loss": 0.0, "step": 253250 }, { "epoch": 54.6, "learning_rate": 1.520823641082122e-05, "loss": 0.0, "step": 253300 }, { "epoch": 54.61, "learning_rate": 1.5202593495605203e-05, "loss": 0.0, "step": 253350 }, { "epoch": 54.62, "learning_rate": 1.519695055171233e-05, "loss": 0.0, "step": 253400 }, { "epoch": 54.63, "learning_rate": 1.5191307579941355e-05, "loss": 0.0, "step": 253450 }, { "epoch": 54.65, "learning_rate": 1.5185664581091035e-05, "loss": 0.0, "step": 253500 }, { "epoch": 54.66, "learning_rate": 1.5180021555960134e-05, "loss": 0.0, "step": 253550 }, { "epoch": 54.67, "learning_rate": 1.5174378505347403e-05, "loss": 0.0, "step": 253600 }, { "epoch": 54.68, "learning_rate": 1.516873543005162e-05, "loss": 0.0, "step": 253650 }, { "epoch": 54.69, "learning_rate": 1.516309233087155e-05, "loss": 0.0, "step": 253700 }, { "epoch": 54.7, "learning_rate": 1.5157449208605965e-05, "loss": 0.0, "step": 253750 }, { "epoch": 54.71, "learning_rate": 1.5151806064053649e-05, "loss": 0.0, "step": 253800 }, { "epoch": 54.72, "learning_rate": 1.5146162898013372e-05, "loss": 0.0, "step": 253850 }, { "epoch": 54.73, "learning_rate": 1.5140519711283927e-05, "loss": 0.0, "step": 253900 }, { "epoch": 54.74, "learning_rate": 1.513487650466409e-05, "loss": 0.0, "step": 253950 }, { "epoch": 54.75, "learning_rate": 1.5129233278952659e-05, "loss": 0.0, "step": 254000 }, { "epoch": 54.76, "learning_rate": 1.512359003494842e-05, "loss": 0.0, "step": 254050 }, { "epoch": 54.77, "learning_rate": 1.5117946773450166e-05, "loss": 0.0, "step": 254100 }, { "epoch": 54.79, "learning_rate": 1.5112303495256697e-05, "loss": 0.0, "step": 254150 }, { "epoch": 54.8, "learning_rate": 1.5106660201166811e-05, "loss": 0.0, "step": 254200 }, { "epoch": 54.81, "learning_rate": 1.5101016891979304e-05, "loss": 0.0, "step": 254250 }, { "epoch": 54.82, "learning_rate": 1.5095373568492992e-05, "loss": 0.0, "step": 254300 }, { "epoch": 54.83, "learning_rate": 1.5089730231506667e-05, "loss": 0.0, "step": 254350 }, { "epoch": 54.84, "learning_rate": 1.5084086881819147e-05, "loss": 0.0, "step": 254400 }, { "epoch": 54.85, "learning_rate": 1.5078443520229226e-05, "loss": 0.0, "step": 254450 }, { "epoch": 54.86, "learning_rate": 1.5072800147535732e-05, "loss": 0.0, "step": 254500 }, { "epoch": 54.87, "learning_rate": 1.5067156764537464e-05, "loss": 0.0, "step": 254550 }, { "epoch": 54.88, "learning_rate": 1.5061513372033244e-05, "loss": 0.0, "step": 254600 }, { "epoch": 54.89, "learning_rate": 1.5055869970821886e-05, "loss": 0.0, "step": 254650 }, { "epoch": 54.9, "learning_rate": 1.5050226561702204e-05, "loss": 0.0, "step": 254700 }, { "epoch": 54.91, "learning_rate": 1.5044583145473011e-05, "loss": 0.0, "step": 254750 }, { "epoch": 54.93, "learning_rate": 1.5038939722933135e-05, "loss": 0.0, "step": 254800 }, { "epoch": 54.94, "learning_rate": 1.503329629488139e-05, "loss": 0.0, "step": 254850 }, { "epoch": 54.95, "learning_rate": 1.5027652862116595e-05, "loss": 0.0, "step": 254900 }, { "epoch": 54.96, "learning_rate": 1.5022009425437576e-05, "loss": 0.0, "step": 254950 }, { "epoch": 54.97, "learning_rate": 1.5016365985643147e-05, "loss": 0.0001, "step": 255000 }, { "epoch": 54.98, "learning_rate": 1.5010722543532135e-05, "loss": 0.0, "step": 255050 }, { "epoch": 54.99, "learning_rate": 1.5005079099903357e-05, "loss": 0.0, "step": 255100 }, { "epoch": 55.0, "learning_rate": 1.4999435655555644e-05, "loss": 0.0, "step": 255150 }, { "epoch": 55.01, "learning_rate": 1.4993792211287813e-05, "loss": 0.0, "step": 255200 }, { "epoch": 55.02, "learning_rate": 1.4988148767898682e-05, "loss": 0.0, "step": 255250 }, { "epoch": 55.03, "learning_rate": 1.498250532618708e-05, "loss": 0.0, "step": 255300 }, { "epoch": 55.04, "learning_rate": 1.4976861886951828e-05, "loss": 0.0, "step": 255350 }, { "epoch": 55.05, "learning_rate": 1.497121845099175e-05, "loss": 0.0, "step": 255400 }, { "epoch": 55.07, "learning_rate": 1.4965575019105656e-05, "loss": 0.0, "step": 255450 }, { "epoch": 55.08, "learning_rate": 1.495993159209238e-05, "loss": 0.0, "step": 255500 }, { "epoch": 55.09, "learning_rate": 1.4954288170750733e-05, "loss": 0.0, "step": 255550 }, { "epoch": 55.1, "learning_rate": 1.494864475587954e-05, "loss": 0.0, "step": 255600 }, { "epoch": 55.11, "learning_rate": 1.494300134827762e-05, "loss": 0.0, "step": 255650 }, { "epoch": 55.12, "learning_rate": 1.4937357948743777e-05, "loss": 0.0, "step": 255700 }, { "epoch": 55.13, "learning_rate": 1.4931714558076846e-05, "loss": 0.0, "step": 255750 }, { "epoch": 55.14, "learning_rate": 1.4926071177075629e-05, "loss": 0.0, "step": 255800 }, { "epoch": 55.15, "learning_rate": 1.4920427806538944e-05, "loss": 0.0, "step": 255850 }, { "epoch": 55.16, "learning_rate": 1.4914784447265598e-05, "loss": 0.0, "step": 255900 }, { "epoch": 55.17, "learning_rate": 1.4909141100054411e-05, "loss": 0.0, "step": 255950 }, { "epoch": 55.18, "learning_rate": 1.4903497765704188e-05, "loss": 0.0, "step": 256000 }, { "epoch": 55.2, "learning_rate": 1.4897854445013729e-05, "loss": 0.0, "step": 256050 }, { "epoch": 55.21, "learning_rate": 1.4892211138781843e-05, "loss": 0.0, "step": 256100 }, { "epoch": 55.22, "learning_rate": 1.4886567847807337e-05, "loss": 0.0, "step": 256150 }, { "epoch": 55.23, "learning_rate": 1.4880924572889007e-05, "loss": 0.0, "step": 256200 }, { "epoch": 55.24, "learning_rate": 1.4875281314825656e-05, "loss": 0.0, "step": 256250 }, { "epoch": 55.25, "learning_rate": 1.4869638074416067e-05, "loss": 0.0, "step": 256300 }, { "epoch": 55.26, "learning_rate": 1.4863994852459054e-05, "loss": 0.0, "step": 256350 }, { "epoch": 55.27, "learning_rate": 1.4858351649753394e-05, "loss": 0.0, "step": 256400 }, { "epoch": 55.28, "learning_rate": 1.4852708467097878e-05, "loss": 0.0, "step": 256450 }, { "epoch": 55.29, "learning_rate": 1.4847065305291296e-05, "loss": 0.0, "step": 256500 }, { "epoch": 55.3, "learning_rate": 1.4841422165132416e-05, "loss": 0.0, "step": 256550 }, { "epoch": 55.31, "learning_rate": 1.4835779047420035e-05, "loss": 0.0, "step": 256600 }, { "epoch": 55.32, "learning_rate": 1.4830135952952921e-05, "loss": 0.0, "step": 256650 }, { "epoch": 55.34, "learning_rate": 1.4824492882529848e-05, "loss": 0.0, "step": 256700 }, { "epoch": 55.35, "learning_rate": 1.4818849836949581e-05, "loss": 0.0, "step": 256750 }, { "epoch": 55.36, "learning_rate": 1.4813206817010893e-05, "loss": 0.0, "step": 256800 }, { "epoch": 55.37, "learning_rate": 1.4807563823512547e-05, "loss": 0.0, "step": 256850 }, { "epoch": 55.38, "learning_rate": 1.4801920857253292e-05, "loss": 0.0, "step": 256900 }, { "epoch": 55.39, "learning_rate": 1.479627791903189e-05, "loss": 0.0, "step": 256950 }, { "epoch": 55.4, "learning_rate": 1.479063500964709e-05, "loss": 0.0, "step": 257000 }, { "epoch": 55.41, "learning_rate": 1.4784992129897639e-05, "loss": 0.0, "step": 257050 }, { "epoch": 55.42, "learning_rate": 1.4779349280582284e-05, "loss": 0.0, "step": 257100 }, { "epoch": 55.43, "learning_rate": 1.477370646249975e-05, "loss": 0.0, "step": 257150 }, { "epoch": 55.44, "learning_rate": 1.4768063676448788e-05, "loss": 0.0, "step": 257200 }, { "epoch": 55.45, "learning_rate": 1.4762420923228115e-05, "loss": 0.0, "step": 257250 }, { "epoch": 55.46, "learning_rate": 1.475677820363646e-05, "loss": 0.0, "step": 257300 }, { "epoch": 55.48, "learning_rate": 1.4751135518472546e-05, "loss": 0.0, "step": 257350 }, { "epoch": 55.49, "learning_rate": 1.4745492868535071e-05, "loss": 0.0, "step": 257400 }, { "epoch": 55.5, "learning_rate": 1.473985025462277e-05, "loss": 0.0, "step": 257450 }, { "epoch": 55.51, "learning_rate": 1.4734207677534332e-05, "loss": 0.0, "step": 257500 }, { "epoch": 55.52, "learning_rate": 1.4728565138068462e-05, "loss": 0.0, "step": 257550 }, { "epoch": 55.53, "learning_rate": 1.4722922637023848e-05, "loss": 0.0, "step": 257600 }, { "epoch": 55.54, "learning_rate": 1.4717280175199186e-05, "loss": 0.0, "step": 257650 }, { "epoch": 55.55, "learning_rate": 1.4711637753393161e-05, "loss": 0.0, "step": 257700 }, { "epoch": 55.56, "learning_rate": 1.4705995372404443e-05, "loss": 0.0, "step": 257750 }, { "epoch": 55.57, "learning_rate": 1.4700353033031705e-05, "loss": 0.0, "step": 257800 }, { "epoch": 55.58, "learning_rate": 1.4694710736073617e-05, "loss": 0.0, "step": 257850 }, { "epoch": 55.59, "learning_rate": 1.468906848232884e-05, "loss": 0.0, "step": 257900 }, { "epoch": 55.6, "learning_rate": 1.4683426272596024e-05, "loss": 0.0, "step": 257950 }, { "epoch": 55.62, "learning_rate": 1.4677784107673813e-05, "loss": 0.0, "step": 258000 }, { "epoch": 55.63, "learning_rate": 1.467214198836086e-05, "loss": 0.0, "step": 258050 }, { "epoch": 55.64, "learning_rate": 1.4666499915455791e-05, "loss": 0.0, "step": 258100 }, { "epoch": 55.65, "learning_rate": 1.4660857889757236e-05, "loss": 0.0, "step": 258150 }, { "epoch": 55.66, "learning_rate": 1.465521591206382e-05, "loss": 0.0, "step": 258200 }, { "epoch": 55.67, "learning_rate": 1.4649573983174145e-05, "loss": 0.0, "step": 258250 }, { "epoch": 55.68, "learning_rate": 1.4643932103886837e-05, "loss": 0.0, "step": 258300 }, { "epoch": 55.69, "learning_rate": 1.4638290275000486e-05, "loss": 0.0, "step": 258350 }, { "epoch": 55.7, "learning_rate": 1.463264849731369e-05, "loss": 0.0, "step": 258400 }, { "epoch": 55.71, "learning_rate": 1.4627006771625028e-05, "loss": 0.0, "step": 258450 }, { "epoch": 55.72, "learning_rate": 1.462136509873309e-05, "loss": 0.0, "step": 258500 }, { "epoch": 55.73, "learning_rate": 1.4615723479436445e-05, "loss": 0.0, "step": 258550 }, { "epoch": 55.74, "learning_rate": 1.4610081914533649e-05, "loss": 0.0, "step": 258600 }, { "epoch": 55.76, "learning_rate": 1.4604440404823263e-05, "loss": 0.0, "step": 258650 }, { "epoch": 55.77, "learning_rate": 1.4598798951103838e-05, "loss": 0.0, "step": 258700 }, { "epoch": 55.78, "learning_rate": 1.4593157554173915e-05, "loss": 0.0, "step": 258750 }, { "epoch": 55.79, "learning_rate": 1.4587516214832025e-05, "loss": 0.0, "step": 258800 }, { "epoch": 55.8, "learning_rate": 1.4581874933876691e-05, "loss": 0.0, "step": 258850 }, { "epoch": 55.81, "learning_rate": 1.4576233712106426e-05, "loss": 0.0, "step": 258900 }, { "epoch": 55.82, "learning_rate": 1.4570592550319746e-05, "loss": 0.0, "step": 258950 }, { "epoch": 55.83, "learning_rate": 1.4564951449315145e-05, "loss": 0.0, "step": 259000 }, { "epoch": 55.84, "learning_rate": 1.455931040989112e-05, "loss": 0.0, "step": 259050 }, { "epoch": 55.85, "learning_rate": 1.4553669432846136e-05, "loss": 0.0, "step": 259100 }, { "epoch": 55.86, "learning_rate": 1.4548028518978691e-05, "loss": 0.0, "step": 259150 }, { "epoch": 55.87, "learning_rate": 1.454238766908723e-05, "loss": 0.0, "step": 259200 }, { "epoch": 55.88, "learning_rate": 1.4536746883970216e-05, "loss": 0.0, "step": 259250 }, { "epoch": 55.9, "learning_rate": 1.4531106164426088e-05, "loss": 0.0, "step": 259300 }, { "epoch": 55.91, "learning_rate": 1.4525465511253294e-05, "loss": 0.0, "step": 259350 }, { "epoch": 55.92, "learning_rate": 1.4519824925250257e-05, "loss": 0.0, "step": 259400 }, { "epoch": 55.93, "learning_rate": 1.4514184407215391e-05, "loss": 0.0, "step": 259450 }, { "epoch": 55.94, "learning_rate": 1.4508543957947103e-05, "loss": 0.0, "step": 259500 }, { "epoch": 55.95, "learning_rate": 1.4502903578243799e-05, "loss": 0.0, "step": 259550 }, { "epoch": 55.96, "learning_rate": 1.4497263268903861e-05, "loss": 0.0, "step": 259600 }, { "epoch": 55.97, "learning_rate": 1.4491623030725674e-05, "loss": 0.0, "step": 259650 }, { "epoch": 55.98, "learning_rate": 1.4485982864507602e-05, "loss": 0.0, "step": 259700 }, { "epoch": 55.99, "learning_rate": 1.4480342771047997e-05, "loss": 0.0, "step": 259750 }, { "epoch": 56.0, "learning_rate": 1.447470275114522e-05, "loss": 0.0, "step": 259800 }, { "epoch": 56.01, "learning_rate": 1.4469062805597603e-05, "loss": 0.0, "step": 259850 }, { "epoch": 56.03, "learning_rate": 1.4463422935203474e-05, "loss": 0.0, "step": 259900 }, { "epoch": 56.04, "learning_rate": 1.4457783140761142e-05, "loss": 0.0, "step": 259950 }, { "epoch": 56.05, "learning_rate": 1.4452143423068926e-05, "loss": 0.0, "step": 260000 }, { "epoch": 56.06, "learning_rate": 1.4446503782925112e-05, "loss": 0.0, "step": 260050 }, { "epoch": 56.07, "learning_rate": 1.4440864221127986e-05, "loss": 0.0, "step": 260100 }, { "epoch": 56.08, "learning_rate": 1.4435224738475818e-05, "loss": 0.0, "step": 260150 }, { "epoch": 56.09, "learning_rate": 1.4429585335766875e-05, "loss": 0.0, "step": 260200 }, { "epoch": 56.1, "learning_rate": 1.4423946013799407e-05, "loss": 0.0, "step": 260250 }, { "epoch": 56.11, "learning_rate": 1.441830677337165e-05, "loss": 0.0, "step": 260300 }, { "epoch": 56.12, "learning_rate": 1.4412667615281826e-05, "loss": 0.0, "step": 260350 }, { "epoch": 56.13, "learning_rate": 1.4407028540328161e-05, "loss": 0.0, "step": 260400 }, { "epoch": 56.14, "learning_rate": 1.4401389549308854e-05, "loss": 0.0, "step": 260450 }, { "epoch": 56.15, "learning_rate": 1.4395750643022103e-05, "loss": 0.0, "step": 260500 }, { "epoch": 56.17, "learning_rate": 1.4390111822266078e-05, "loss": 0.0, "step": 260550 }, { "epoch": 56.18, "learning_rate": 1.4384473087838948e-05, "loss": 0.0, "step": 260600 }, { "epoch": 56.19, "learning_rate": 1.4378834440538876e-05, "loss": 0.0, "step": 260650 }, { "epoch": 56.2, "learning_rate": 1.4373195881164002e-05, "loss": 0.0, "step": 260700 }, { "epoch": 56.21, "learning_rate": 1.436755741051246e-05, "loss": 0.0, "step": 260750 }, { "epoch": 56.22, "learning_rate": 1.4361919029382354e-05, "loss": 0.0, "step": 260800 }, { "epoch": 56.23, "learning_rate": 1.435628073857181e-05, "loss": 0.0, "step": 260850 }, { "epoch": 56.24, "learning_rate": 1.4350642538878908e-05, "loss": 0.0, "step": 260900 }, { "epoch": 56.25, "learning_rate": 1.4345004431101735e-05, "loss": 0.0, "step": 260950 }, { "epoch": 56.26, "learning_rate": 1.4339366416038347e-05, "loss": 0.0, "step": 261000 }, { "epoch": 56.27, "learning_rate": 1.4333728494486813e-05, "loss": 0.0, "step": 261050 }, { "epoch": 56.28, "learning_rate": 1.4328090667245166e-05, "loss": 0.0, "step": 261100 }, { "epoch": 56.29, "learning_rate": 1.4322452935111432e-05, "loss": 0.0, "step": 261150 }, { "epoch": 56.31, "learning_rate": 1.4316815298883623e-05, "loss": 0.0, "step": 261200 }, { "epoch": 56.32, "learning_rate": 1.4311177759359745e-05, "loss": 0.0, "step": 261250 }, { "epoch": 56.33, "learning_rate": 1.4305540317337782e-05, "loss": 0.0, "step": 261300 }, { "epoch": 56.34, "learning_rate": 1.429990297361571e-05, "loss": 0.0, "step": 261350 }, { "epoch": 56.35, "learning_rate": 1.4294265728991483e-05, "loss": 0.0, "step": 261400 }, { "epoch": 56.36, "learning_rate": 1.4288628584263045e-05, "loss": 0.0, "step": 261450 }, { "epoch": 56.37, "learning_rate": 1.428299154022833e-05, "loss": 0.0, "step": 261500 }, { "epoch": 56.38, "learning_rate": 1.4277354597685254e-05, "loss": 0.0, "step": 261550 }, { "epoch": 56.39, "learning_rate": 1.4271717757431724e-05, "loss": 0.0, "step": 261600 }, { "epoch": 56.4, "learning_rate": 1.426608102026561e-05, "loss": 0.0, "step": 261650 }, { "epoch": 56.41, "learning_rate": 1.4260444386984809e-05, "loss": 0.0, "step": 261700 }, { "epoch": 56.42, "learning_rate": 1.4254807858387163e-05, "loss": 0.0, "step": 261750 }, { "epoch": 56.43, "learning_rate": 1.4249171435270523e-05, "loss": 0.0, "step": 261800 }, { "epoch": 56.45, "learning_rate": 1.4243535118432708e-05, "loss": 0.0, "step": 261850 }, { "epoch": 56.46, "learning_rate": 1.4237898908671544e-05, "loss": 0.0, "step": 261900 }, { "epoch": 56.47, "learning_rate": 1.4232262806784825e-05, "loss": 0.0, "step": 261950 }, { "epoch": 56.48, "learning_rate": 1.4226626813570333e-05, "loss": 0.0, "step": 262000 }, { "epoch": 56.49, "learning_rate": 1.4220990929825828e-05, "loss": 0.0, "step": 262050 }, { "epoch": 56.5, "learning_rate": 1.4215355156349073e-05, "loss": 0.0, "step": 262100 }, { "epoch": 56.51, "learning_rate": 1.4209719493937803e-05, "loss": 0.0, "step": 262150 }, { "epoch": 56.52, "learning_rate": 1.420408394338974e-05, "loss": 0.0, "step": 262200 }, { "epoch": 56.53, "learning_rate": 1.4198448505502582e-05, "loss": 0.0, "step": 262250 }, { "epoch": 56.54, "learning_rate": 1.419281318107402e-05, "loss": 0.0, "step": 262300 }, { "epoch": 56.55, "learning_rate": 1.418717797090173e-05, "loss": 0.0, "step": 262350 }, { "epoch": 56.56, "learning_rate": 1.4181542875783371e-05, "loss": 0.0, "step": 262400 }, { "epoch": 56.57, "learning_rate": 1.4175907896516583e-05, "loss": 0.0, "step": 262450 }, { "epoch": 56.59, "learning_rate": 1.4170273033898977e-05, "loss": 0.0, "step": 262500 }, { "epoch": 56.6, "learning_rate": 1.416463828872818e-05, "loss": 0.0, "step": 262550 }, { "epoch": 56.61, "learning_rate": 1.4159003661801774e-05, "loss": 0.0, "step": 262600 }, { "epoch": 56.62, "learning_rate": 1.4153369153917333e-05, "loss": 0.0, "step": 262650 }, { "epoch": 56.63, "learning_rate": 1.4147734765872412e-05, "loss": 0.0, "step": 262700 }, { "epoch": 56.64, "learning_rate": 1.4142100498464557e-05, "loss": 0.0, "step": 262750 }, { "epoch": 56.65, "learning_rate": 1.4136466352491294e-05, "loss": 0.0, "step": 262800 }, { "epoch": 56.66, "learning_rate": 1.413083232875012e-05, "loss": 0.0, "step": 262850 }, { "epoch": 56.67, "learning_rate": 1.4125198428038524e-05, "loss": 0.0, "step": 262900 }, { "epoch": 56.68, "learning_rate": 1.4119564651153986e-05, "loss": 0.0, "step": 262950 }, { "epoch": 56.69, "learning_rate": 1.4113930998893957e-05, "loss": 0.0, "step": 263000 }, { "epoch": 56.7, "learning_rate": 1.4108297472055873e-05, "loss": 0.0, "step": 263050 }, { "epoch": 56.71, "learning_rate": 1.4102664071437149e-05, "loss": 0.0, "step": 263100 }, { "epoch": 56.73, "learning_rate": 1.4097030797835188e-05, "loss": 0.0, "step": 263150 }, { "epoch": 56.74, "learning_rate": 1.4091397652047373e-05, "loss": 0.0, "step": 263200 }, { "epoch": 56.75, "learning_rate": 1.408576463487107e-05, "loss": 0.0, "step": 263250 }, { "epoch": 56.76, "learning_rate": 1.4080131747103626e-05, "loss": 0.0, "step": 263300 }, { "epoch": 56.77, "learning_rate": 1.4074498989542358e-05, "loss": 0.0, "step": 263350 }, { "epoch": 56.78, "learning_rate": 1.4068866362984596e-05, "loss": 0.0, "step": 263400 }, { "epoch": 56.79, "learning_rate": 1.4063233868227617e-05, "loss": 0.0, "step": 263450 }, { "epoch": 56.8, "learning_rate": 1.4057601506068696e-05, "loss": 0.0, "step": 263500 }, { "epoch": 56.81, "learning_rate": 1.4051969277305084e-05, "loss": 0.0, "step": 263550 }, { "epoch": 56.82, "learning_rate": 1.4046337182734022e-05, "loss": 0.0, "step": 263600 }, { "epoch": 56.83, "learning_rate": 1.4040705223152729e-05, "loss": 0.0, "step": 263650 }, { "epoch": 56.84, "learning_rate": 1.403507339935839e-05, "loss": 0.0, "step": 263700 }, { "epoch": 56.85, "learning_rate": 1.4029441712148189e-05, "loss": 0.0, "step": 263750 }, { "epoch": 56.87, "learning_rate": 1.4023810162319279e-05, "loss": 0.0, "step": 263800 }, { "epoch": 56.88, "learning_rate": 1.4018178750668808e-05, "loss": 0.0, "step": 263850 }, { "epoch": 56.89, "learning_rate": 1.4012547477993894e-05, "loss": 0.0, "step": 263900 }, { "epoch": 56.9, "learning_rate": 1.400691634509163e-05, "loss": 0.0, "step": 263950 }, { "epoch": 56.91, "learning_rate": 1.4001285352759093e-05, "loss": 0.0, "step": 264000 }, { "epoch": 56.92, "learning_rate": 1.3995654501793353e-05, "loss": 0.0, "step": 264050 }, { "epoch": 56.93, "learning_rate": 1.3990023792991447e-05, "loss": 0.0, "step": 264100 }, { "epoch": 56.94, "learning_rate": 1.3984393227150396e-05, "loss": 0.0, "step": 264150 }, { "epoch": 56.95, "learning_rate": 1.3978762805067184e-05, "loss": 0.0, "step": 264200 }, { "epoch": 56.96, "learning_rate": 1.3973132527538815e-05, "loss": 0.0, "step": 264250 }, { "epoch": 56.97, "learning_rate": 1.396750239536223e-05, "loss": 0.0, "step": 264300 }, { "epoch": 56.98, "learning_rate": 1.3961872409334373e-05, "loss": 0.0, "step": 264350 }, { "epoch": 57.0, "learning_rate": 1.3956242570252159e-05, "loss": 0.0, "step": 264400 }, { "epoch": 57.01, "learning_rate": 1.3950612878912488e-05, "loss": 0.0, "step": 264450 }, { "epoch": 57.02, "learning_rate": 1.3944983336112238e-05, "loss": 0.0, "step": 264500 }, { "epoch": 57.03, "learning_rate": 1.3939353942648256e-05, "loss": 0.0, "step": 264550 }, { "epoch": 57.04, "learning_rate": 1.3933724699317378e-05, "loss": 0.0, "step": 264600 }, { "epoch": 57.05, "learning_rate": 1.3928095606916416e-05, "loss": 0.0, "step": 264650 }, { "epoch": 57.06, "learning_rate": 1.3922466666242163e-05, "loss": 0.0, "step": 264700 }, { "epoch": 57.07, "learning_rate": 1.3916837878091394e-05, "loss": 0.0, "step": 264750 }, { "epoch": 57.08, "learning_rate": 1.3911209243260845e-05, "loss": 0.0, "step": 264800 }, { "epoch": 57.09, "learning_rate": 1.3905580762547247e-05, "loss": 0.0, "step": 264850 }, { "epoch": 57.1, "learning_rate": 1.3899952436747306e-05, "loss": 0.0, "step": 264900 }, { "epoch": 57.11, "learning_rate": 1.3894324266657702e-05, "loss": 0.0, "step": 264950 }, { "epoch": 57.12, "learning_rate": 1.3888696253075102e-05, "loss": 0.0, "step": 265000 }, { "epoch": 57.14, "learning_rate": 1.388306839679613e-05, "loss": 0.0, "step": 265050 }, { "epoch": 57.15, "learning_rate": 1.387744069861742e-05, "loss": 0.0, "step": 265100 }, { "epoch": 57.16, "learning_rate": 1.3871813159335551e-05, "loss": 0.0, "step": 265150 }, { "epoch": 57.17, "learning_rate": 1.3866185779747101e-05, "loss": 0.0, "step": 265200 }, { "epoch": 57.18, "learning_rate": 1.3860558560648613e-05, "loss": 0.0, "step": 265250 }, { "epoch": 57.19, "learning_rate": 1.385493150283662e-05, "loss": 0.0, "step": 265300 }, { "epoch": 57.2, "learning_rate": 1.3849304607107627e-05, "loss": 0.0, "step": 265350 }, { "epoch": 57.21, "learning_rate": 1.3843677874258102e-05, "loss": 0.0, "step": 265400 }, { "epoch": 57.22, "learning_rate": 1.383805130508451e-05, "loss": 0.0, "step": 265450 }, { "epoch": 57.23, "learning_rate": 1.3832424900383279e-05, "loss": 0.0, "step": 265500 }, { "epoch": 57.24, "learning_rate": 1.3826798660950826e-05, "loss": 0.0, "step": 265550 }, { "epoch": 57.25, "learning_rate": 1.3821172587583541e-05, "loss": 0.0, "step": 265600 }, { "epoch": 57.26, "learning_rate": 1.3815546681077777e-05, "loss": 0.0, "step": 265650 }, { "epoch": 57.28, "learning_rate": 1.3809920942229878e-05, "loss": 0.0, "step": 265700 }, { "epoch": 57.29, "learning_rate": 1.3804295371836162e-05, "loss": 0.0, "step": 265750 }, { "epoch": 57.3, "learning_rate": 1.3798669970692923e-05, "loss": 0.0, "step": 265800 }, { "epoch": 57.31, "learning_rate": 1.3793044739596429e-05, "loss": 0.0, "step": 265850 }, { "epoch": 57.32, "learning_rate": 1.3787419679342912e-05, "loss": 0.0, "step": 265900 }, { "epoch": 57.33, "learning_rate": 1.3781794790728613e-05, "loss": 0.0, "step": 265950 }, { "epoch": 57.34, "learning_rate": 1.3776170074549716e-05, "loss": 0.0, "step": 266000 }, { "epoch": 57.35, "learning_rate": 1.377054553160239e-05, "loss": 0.0, "step": 266050 }, { "epoch": 57.36, "learning_rate": 1.3764921162682786e-05, "loss": 0.0, "step": 266100 }, { "epoch": 57.37, "learning_rate": 1.3759296968587027e-05, "loss": 0.0, "step": 266150 }, { "epoch": 57.38, "learning_rate": 1.3753672950111214e-05, "loss": 0.0, "step": 266200 }, { "epoch": 57.39, "learning_rate": 1.374804910805141e-05, "loss": 0.0, "step": 266250 }, { "epoch": 57.4, "learning_rate": 1.374242544320367e-05, "loss": 0.0, "step": 266300 }, { "epoch": 57.42, "learning_rate": 1.3736801956364013e-05, "loss": 0.0, "step": 266350 }, { "epoch": 57.43, "learning_rate": 1.3731178648328437e-05, "loss": 0.0, "step": 266400 }, { "epoch": 57.44, "learning_rate": 1.3725555519892922e-05, "loss": 0.0, "step": 266450 }, { "epoch": 57.45, "learning_rate": 1.3719932571853403e-05, "loss": 0.0, "step": 266500 }, { "epoch": 57.46, "learning_rate": 1.3714309805005803e-05, "loss": 0.0, "step": 266550 }, { "epoch": 57.47, "learning_rate": 1.3708687220146024e-05, "loss": 0.0, "step": 266600 }, { "epoch": 57.48, "learning_rate": 1.3703064818069934e-05, "loss": 0.0, "step": 266650 }, { "epoch": 57.49, "learning_rate": 1.3697442599573375e-05, "loss": 0.0, "step": 266700 }, { "epoch": 57.5, "learning_rate": 1.3691820565452158e-05, "loss": 0.0, "step": 266750 }, { "epoch": 57.51, "learning_rate": 1.3686198716502091e-05, "loss": 0.0, "step": 266800 }, { "epoch": 57.52, "learning_rate": 1.3680577053518926e-05, "loss": 0.0, "step": 266850 }, { "epoch": 57.53, "learning_rate": 1.3674955577298408e-05, "loss": 0.0, "step": 266900 }, { "epoch": 57.54, "learning_rate": 1.3669334288636246e-05, "loss": 0.0, "step": 266950 }, { "epoch": 57.56, "learning_rate": 1.366371318832813e-05, "loss": 0.0, "step": 267000 }, { "epoch": 57.57, "learning_rate": 1.3658092277169723e-05, "loss": 0.0, "step": 267050 }, { "epoch": 57.58, "learning_rate": 1.3652471555956652e-05, "loss": 0.0, "step": 267100 }, { "epoch": 57.59, "learning_rate": 1.3646851025484522e-05, "loss": 0.0, "step": 267150 }, { "epoch": 57.6, "learning_rate": 1.3641230686548911e-05, "loss": 0.0, "step": 267200 }, { "epoch": 57.61, "learning_rate": 1.363561053994538e-05, "loss": 0.0, "step": 267250 }, { "epoch": 57.62, "learning_rate": 1.3629990586469447e-05, "loss": 0.0, "step": 267300 }, { "epoch": 57.63, "learning_rate": 1.362437082691661e-05, "loss": 0.0, "step": 267350 }, { "epoch": 57.64, "learning_rate": 1.3618751262082335e-05, "loss": 0.0, "step": 267400 }, { "epoch": 57.65, "learning_rate": 1.3613131892762072e-05, "loss": 0.0, "step": 267450 }, { "epoch": 57.66, "learning_rate": 1.360751271975123e-05, "loss": 0.0, "step": 267500 }, { "epoch": 57.67, "learning_rate": 1.3601893743845203e-05, "loss": 0.0, "step": 267550 }, { "epoch": 57.68, "learning_rate": 1.3596274965839333e-05, "loss": 0.0, "step": 267600 }, { "epoch": 57.7, "learning_rate": 1.3590656386528972e-05, "loss": 0.0, "step": 267650 }, { "epoch": 57.71, "learning_rate": 1.358503800670941e-05, "loss": 0.0, "step": 267700 }, { "epoch": 57.72, "learning_rate": 1.3579419827175928e-05, "loss": 0.0, "step": 267750 }, { "epoch": 57.73, "learning_rate": 1.3573801848723762e-05, "loss": 0.0, "step": 267800 }, { "epoch": 57.74, "learning_rate": 1.3568184072148142e-05, "loss": 0.0, "step": 267850 }, { "epoch": 57.75, "learning_rate": 1.3562566498244254e-05, "loss": 0.0, "step": 267900 }, { "epoch": 57.76, "learning_rate": 1.3556949127807254e-05, "loss": 0.0, "step": 267950 }, { "epoch": 57.77, "learning_rate": 1.3551331961632275e-05, "loss": 0.0, "step": 268000 }, { "epoch": 57.78, "learning_rate": 1.3545715000514417e-05, "loss": 0.0, "step": 268050 }, { "epoch": 57.79, "learning_rate": 1.3540098245248763e-05, "loss": 0.0, "step": 268100 }, { "epoch": 57.8, "learning_rate": 1.3534481696630354e-05, "loss": 0.0, "step": 268150 }, { "epoch": 57.81, "learning_rate": 1.35288653554542e-05, "loss": 0.0, "step": 268200 }, { "epoch": 57.82, "learning_rate": 1.3523249222515286e-05, "loss": 0.0, "step": 268250 }, { "epoch": 57.84, "learning_rate": 1.3517633298608578e-05, "loss": 0.0, "step": 268300 }, { "epoch": 57.85, "learning_rate": 1.3512017584528997e-05, "loss": 0.0, "step": 268350 }, { "epoch": 57.86, "learning_rate": 1.3506402081071445e-05, "loss": 0.0, "step": 268400 }, { "epoch": 57.87, "learning_rate": 1.3500786789030775e-05, "loss": 0.0, "step": 268450 }, { "epoch": 57.88, "learning_rate": 1.3495171709201845e-05, "loss": 0.0, "step": 268500 }, { "epoch": 57.89, "learning_rate": 1.3489556842379448e-05, "loss": 0.0, "step": 268550 }, { "epoch": 57.9, "learning_rate": 1.3483942189358367e-05, "loss": 0.0, "step": 268600 }, { "epoch": 57.91, "learning_rate": 1.3478327750933346e-05, "loss": 0.0, "step": 268650 }, { "epoch": 57.92, "learning_rate": 1.3472713527899101e-05, "loss": 0.0, "step": 268700 }, { "epoch": 57.93, "learning_rate": 1.3467099521050326e-05, "loss": 0.0, "step": 268750 }, { "epoch": 57.94, "learning_rate": 1.346148573118167e-05, "loss": 0.0, "step": 268800 }, { "epoch": 57.95, "learning_rate": 1.3455872159087756e-05, "loss": 0.0, "step": 268850 }, { "epoch": 57.97, "learning_rate": 1.345025880556318e-05, "loss": 0.0, "step": 268900 }, { "epoch": 57.98, "learning_rate": 1.3444645671402507e-05, "loss": 0.0, "step": 268950 }, { "epoch": 57.99, "learning_rate": 1.3439032757400272e-05, "loss": 0.0, "step": 269000 }, { "epoch": 58.0, "learning_rate": 1.3433420064350967e-05, "loss": 0.0, "step": 269050 }, { "epoch": 58.01, "learning_rate": 1.3427807593049067e-05, "loss": 0.0, "step": 269100 }, { "epoch": 58.02, "learning_rate": 1.342219534428901e-05, "loss": 0.0, "step": 269150 }, { "epoch": 58.03, "learning_rate": 1.3416583318865202e-05, "loss": 0.0, "step": 269200 }, { "epoch": 58.04, "learning_rate": 1.341097151757202e-05, "loss": 0.0, "step": 269250 }, { "epoch": 58.05, "learning_rate": 1.34053599412038e-05, "loss": 0.0, "step": 269300 }, { "epoch": 58.06, "learning_rate": 1.3399748590554868e-05, "loss": 0.0, "step": 269350 }, { "epoch": 58.07, "learning_rate": 1.339413746641949e-05, "loss": 0.0, "step": 269400 }, { "epoch": 58.08, "learning_rate": 1.338852656959192e-05, "loss": 0.0, "step": 269450 }, { "epoch": 58.09, "learning_rate": 1.338291590086637e-05, "loss": 0.0, "step": 269500 }, { "epoch": 58.11, "learning_rate": 1.3377305461037021e-05, "loss": 0.0, "step": 269550 }, { "epoch": 58.12, "learning_rate": 1.3371695250898036e-05, "loss": 0.0, "step": 269600 }, { "epoch": 58.13, "learning_rate": 1.336608527124352e-05, "loss": 0.0, "step": 269650 }, { "epoch": 58.14, "learning_rate": 1.3360475522867564e-05, "loss": 0.0, "step": 269700 }, { "epoch": 58.15, "learning_rate": 1.3354866006564214e-05, "loss": 0.0, "step": 269750 }, { "epoch": 58.16, "learning_rate": 1.3349256723127501e-05, "loss": 0.0, "step": 269800 }, { "epoch": 58.17, "learning_rate": 1.3343647673351407e-05, "loss": 0.0, "step": 269850 }, { "epoch": 58.18, "learning_rate": 1.3338038858029887e-05, "loss": 0.0, "step": 269900 }, { "epoch": 58.19, "learning_rate": 1.3332430277956855e-05, "loss": 0.0, "step": 269950 }, { "epoch": 58.2, "learning_rate": 1.3326821933926207e-05, "loss": 0.0, "step": 270000 }, { "epoch": 58.21, "learning_rate": 1.3321213826731793e-05, "loss": 0.0, "step": 270050 }, { "epoch": 58.22, "learning_rate": 1.3315605957167434e-05, "loss": 0.0, "step": 270100 }, { "epoch": 58.23, "learning_rate": 1.3309998326026912e-05, "loss": 0.0, "step": 270150 }, { "epoch": 58.25, "learning_rate": 1.3304390934103996e-05, "loss": 0.0, "step": 270200 }, { "epoch": 58.26, "learning_rate": 1.3298783782192388e-05, "loss": 0.0, "step": 270250 }, { "epoch": 58.27, "learning_rate": 1.3293176871085779e-05, "loss": 0.0, "step": 270300 }, { "epoch": 58.28, "learning_rate": 1.3287570201577822e-05, "loss": 0.0, "step": 270350 }, { "epoch": 58.29, "learning_rate": 1.3281963774462129e-05, "loss": 0.0, "step": 270400 }, { "epoch": 58.3, "learning_rate": 1.3276357590532292e-05, "loss": 0.0, "step": 270450 }, { "epoch": 58.31, "learning_rate": 1.327075165058185e-05, "loss": 0.0, "step": 270500 }, { "epoch": 58.32, "learning_rate": 1.326514595540432e-05, "loss": 0.0, "step": 270550 }, { "epoch": 58.33, "learning_rate": 1.325954050579318e-05, "loss": 0.0, "step": 270600 }, { "epoch": 58.34, "learning_rate": 1.3253935302541877e-05, "loss": 0.0, "step": 270650 }, { "epoch": 58.35, "learning_rate": 1.324833034644382e-05, "loss": 0.0, "step": 270700 }, { "epoch": 58.36, "learning_rate": 1.3242725638292385e-05, "loss": 0.0, "step": 270750 }, { "epoch": 58.37, "learning_rate": 1.3237121178880902e-05, "loss": 0.0, "step": 270800 }, { "epoch": 58.39, "learning_rate": 1.3231516969002685e-05, "loss": 0.0, "step": 270850 }, { "epoch": 58.4, "learning_rate": 1.3225913009450998e-05, "loss": 0.0, "step": 270900 }, { "epoch": 58.41, "learning_rate": 1.3220309301019078e-05, "loss": 0.0, "step": 270950 }, { "epoch": 58.42, "learning_rate": 1.3214705844500117e-05, "loss": 0.0, "step": 271000 }, { "epoch": 58.43, "learning_rate": 1.3209102640687289e-05, "loss": 0.0, "step": 271050 }, { "epoch": 58.44, "learning_rate": 1.3203499690373708e-05, "loss": 0.0, "step": 271100 }, { "epoch": 58.45, "learning_rate": 1.3197896994352467e-05, "loss": 0.0, "step": 271150 }, { "epoch": 58.46, "learning_rate": 1.3192294553416626e-05, "loss": 0.0, "step": 271200 }, { "epoch": 58.47, "learning_rate": 1.3186692368359194e-05, "loss": 0.0, "step": 271250 }, { "epoch": 58.48, "learning_rate": 1.3181090439973163e-05, "loss": 0.0, "step": 271300 }, { "epoch": 58.49, "learning_rate": 1.3175488769051479e-05, "loss": 0.0, "step": 271350 }, { "epoch": 58.5, "learning_rate": 1.3169887356387041e-05, "loss": 0.0, "step": 271400 }, { "epoch": 58.51, "learning_rate": 1.3164286202772727e-05, "loss": 0.0, "step": 271450 }, { "epoch": 58.53, "learning_rate": 1.3158685309001379e-05, "loss": 0.0, "step": 271500 }, { "epoch": 58.54, "learning_rate": 1.315308467586579e-05, "loss": 0.0, "step": 271550 }, { "epoch": 58.55, "learning_rate": 1.3147484304158729e-05, "loss": 0.0, "step": 271600 }, { "epoch": 58.56, "learning_rate": 1.3141884194672908e-05, "loss": 0.0, "step": 271650 }, { "epoch": 58.57, "learning_rate": 1.313628434820103e-05, "loss": 0.0, "step": 271700 }, { "epoch": 58.58, "learning_rate": 1.3130684765535737e-05, "loss": 0.0, "step": 271750 }, { "epoch": 58.59, "learning_rate": 1.3125085447469649e-05, "loss": 0.0, "step": 271800 }, { "epoch": 58.6, "learning_rate": 1.3119486394795335e-05, "loss": 0.0, "step": 271850 }, { "epoch": 58.61, "learning_rate": 1.3113887608305342e-05, "loss": 0.0, "step": 271900 }, { "epoch": 58.62, "learning_rate": 1.3108289088792171e-05, "loss": 0.0, "step": 271950 }, { "epoch": 58.63, "learning_rate": 1.310269083704828e-05, "loss": 0.0, "step": 272000 }, { "epoch": 58.64, "learning_rate": 1.3097092853866094e-05, "loss": 0.0, "step": 272050 }, { "epoch": 58.65, "learning_rate": 1.3091495140038002e-05, "loss": 0.0, "step": 272100 }, { "epoch": 58.67, "learning_rate": 1.3085897696356355e-05, "loss": 0.0, "step": 272150 }, { "epoch": 58.68, "learning_rate": 1.3080300523613466e-05, "loss": 0.0, "step": 272200 }, { "epoch": 58.69, "learning_rate": 1.3074703622601602e-05, "loss": 0.0, "step": 272250 }, { "epoch": 58.7, "learning_rate": 1.3069106994112998e-05, "loss": 0.0, "step": 272300 }, { "epoch": 58.71, "learning_rate": 1.3063510638939853e-05, "loss": 0.0, "step": 272350 }, { "epoch": 58.72, "learning_rate": 1.3057914557874324e-05, "loss": 0.0, "step": 272400 }, { "epoch": 58.73, "learning_rate": 1.3052318751708532e-05, "loss": 0.0, "step": 272450 }, { "epoch": 58.74, "learning_rate": 1.3046723221234541e-05, "loss": 0.0, "step": 272500 }, { "epoch": 58.75, "learning_rate": 1.3041127967244412e-05, "loss": 0.0, "step": 272550 }, { "epoch": 58.76, "learning_rate": 1.3035532990530134e-05, "loss": 0.0, "step": 272600 }, { "epoch": 58.77, "learning_rate": 1.302993829188367e-05, "loss": 0.0, "step": 272650 }, { "epoch": 58.78, "learning_rate": 1.3024343872096943e-05, "loss": 0.0, "step": 272700 }, { "epoch": 58.79, "learning_rate": 1.301874973196184e-05, "loss": 0.0, "step": 272750 }, { "epoch": 58.81, "learning_rate": 1.3013155872270202e-05, "loss": 0.0, "step": 272800 }, { "epoch": 58.82, "learning_rate": 1.300756229381383e-05, "loss": 0.0, "step": 272850 }, { "epoch": 58.83, "learning_rate": 1.3001968997384491e-05, "loss": 0.0, "step": 272900 }, { "epoch": 58.84, "learning_rate": 1.2996375983773904e-05, "loss": 0.0, "step": 272950 }, { "epoch": 58.85, "learning_rate": 1.299078325377376e-05, "loss": 0.0, "step": 273000 }, { "epoch": 58.86, "learning_rate": 1.2985190808175704e-05, "loss": 0.0, "step": 273050 }, { "epoch": 58.87, "learning_rate": 1.2979598647771331e-05, "loss": 0.0, "step": 273100 }, { "epoch": 58.88, "learning_rate": 1.2974006773352205e-05, "loss": 0.0, "step": 273150 }, { "epoch": 58.89, "learning_rate": 1.2968415185709858e-05, "loss": 0.0, "step": 273200 }, { "epoch": 58.9, "learning_rate": 1.2962823885635766e-05, "loss": 0.0, "step": 273250 }, { "epoch": 58.91, "learning_rate": 1.2957232873921374e-05, "loss": 0.0, "step": 273300 }, { "epoch": 58.92, "learning_rate": 1.2951642151358069e-05, "loss": 0.0, "step": 273350 }, { "epoch": 58.94, "learning_rate": 1.2946051718737231e-05, "loss": 0.0, "step": 273400 }, { "epoch": 58.95, "learning_rate": 1.2940461576850167e-05, "loss": 0.0, "step": 273450 }, { "epoch": 58.96, "learning_rate": 1.2934871726488156e-05, "loss": 0.0, "step": 273500 }, { "epoch": 58.97, "learning_rate": 1.2929282168442436e-05, "loss": 0.0, "step": 273550 }, { "epoch": 58.98, "learning_rate": 1.2923692903504194e-05, "loss": 0.0, "step": 273600 }, { "epoch": 58.99, "learning_rate": 1.2918103932464598e-05, "loss": 0.0, "step": 273650 }, { "epoch": 59.0, "learning_rate": 1.2912515256114748e-05, "loss": 0.0, "step": 273700 }, { "epoch": 59.01, "learning_rate": 1.2906926875245719e-05, "loss": 0.0, "step": 273750 }, { "epoch": 59.02, "learning_rate": 1.2901338790648534e-05, "loss": 0.0, "step": 273800 }, { "epoch": 59.03, "learning_rate": 1.2895751003114186e-05, "loss": 0.0, "step": 273850 }, { "epoch": 59.04, "learning_rate": 1.2890163513433622e-05, "loss": 0.0, "step": 273900 }, { "epoch": 59.05, "learning_rate": 1.2884576322397733e-05, "loss": 0.0, "step": 273950 }, { "epoch": 59.06, "learning_rate": 1.287898943079738e-05, "loss": 0.0, "step": 274000 }, { "epoch": 59.08, "learning_rate": 1.2873402839423393e-05, "loss": 0.0, "step": 274050 }, { "epoch": 59.09, "learning_rate": 1.2867816549066534e-05, "loss": 0.0, "step": 274100 }, { "epoch": 59.1, "learning_rate": 1.2862230560517545e-05, "loss": 0.0, "step": 274150 }, { "epoch": 59.11, "learning_rate": 1.2856644874567102e-05, "loss": 0.0, "step": 274200 }, { "epoch": 59.12, "learning_rate": 1.285105949200587e-05, "loss": 0.0, "step": 274250 }, { "epoch": 59.13, "learning_rate": 1.2845474413624438e-05, "loss": 0.0, "step": 274300 }, { "epoch": 59.14, "learning_rate": 1.2839889640213372e-05, "loss": 0.0, "step": 274350 }, { "epoch": 59.15, "learning_rate": 1.2834305172563195e-05, "loss": 0.0, "step": 274400 }, { "epoch": 59.16, "learning_rate": 1.2828721011464366e-05, "loss": 0.0, "step": 274450 }, { "epoch": 59.17, "learning_rate": 1.2823137157707334e-05, "loss": 0.0, "step": 274500 }, { "epoch": 59.18, "learning_rate": 1.2817553612082476e-05, "loss": 0.0, "step": 274550 }, { "epoch": 59.19, "learning_rate": 1.2811970375380138e-05, "loss": 0.0, "step": 274600 }, { "epoch": 59.2, "learning_rate": 1.2806387448390617e-05, "loss": 0.0, "step": 274650 }, { "epoch": 59.22, "learning_rate": 1.2800804831904175e-05, "loss": 0.0, "step": 274700 }, { "epoch": 59.23, "learning_rate": 1.2795222526711025e-05, "loss": 0.0, "step": 274750 }, { "epoch": 59.24, "learning_rate": 1.278964053360133e-05, "loss": 0.0, "step": 274800 }, { "epoch": 59.25, "learning_rate": 1.2784058853365213e-05, "loss": 0.0, "step": 274850 }, { "epoch": 59.26, "learning_rate": 1.2778477486792759e-05, "loss": 0.0, "step": 274900 }, { "epoch": 59.27, "learning_rate": 1.2772896434674e-05, "loss": 0.0, "step": 274950 }, { "epoch": 59.28, "learning_rate": 1.2767315697798932e-05, "loss": 0.0, "step": 275000 }, { "epoch": 59.29, "learning_rate": 1.276173527695749e-05, "loss": 0.0, "step": 275050 }, { "epoch": 59.3, "learning_rate": 1.275615517293959e-05, "loss": 0.0, "step": 275100 }, { "epoch": 59.31, "learning_rate": 1.2750575386535076e-05, "loss": 0.0, "step": 275150 }, { "epoch": 59.32, "learning_rate": 1.2744995918533767e-05, "loss": 0.0, "step": 275200 }, { "epoch": 59.33, "learning_rate": 1.2739416769725432e-05, "loss": 0.0, "step": 275250 }, { "epoch": 59.34, "learning_rate": 1.2733837940899778e-05, "loss": 0.0, "step": 275300 }, { "epoch": 59.36, "learning_rate": 1.27282594328465e-05, "loss": 0.0, "step": 275350 }, { "epoch": 59.37, "learning_rate": 1.2722681246355217e-05, "loss": 0.0, "step": 275400 }, { "epoch": 59.38, "learning_rate": 1.2717103382215519e-05, "loss": 0.0, "step": 275450 }, { "epoch": 59.39, "learning_rate": 1.2711525841216938e-05, "loss": 0.0, "step": 275500 }, { "epoch": 59.4, "learning_rate": 1.2705948624148978e-05, "loss": 0.0, "step": 275550 }, { "epoch": 59.41, "learning_rate": 1.2700371731801086e-05, "loss": 0.0, "step": 275600 }, { "epoch": 59.42, "learning_rate": 1.2694795164962657e-05, "loss": 0.0, "step": 275650 }, { "epoch": 59.43, "learning_rate": 1.2689218924423049e-05, "loss": 0.0, "step": 275700 }, { "epoch": 59.44, "learning_rate": 1.2683643010971576e-05, "loss": 0.0, "step": 275750 }, { "epoch": 59.45, "learning_rate": 1.2678067425397498e-05, "loss": 0.0, "step": 275800 }, { "epoch": 59.46, "learning_rate": 1.2672492168490036e-05, "loss": 0.0, "step": 275850 }, { "epoch": 59.47, "learning_rate": 1.266691724103835e-05, "loss": 0.0, "step": 275900 }, { "epoch": 59.48, "learning_rate": 1.266134264383158e-05, "loss": 0.0, "step": 275950 }, { "epoch": 59.5, "learning_rate": 1.2655768377658792e-05, "loss": 0.0, "step": 276000 }, { "epoch": 59.51, "learning_rate": 1.2650194443309016e-05, "loss": 0.0, "step": 276050 }, { "epoch": 59.52, "learning_rate": 1.2644620841571246e-05, "loss": 0.0, "step": 276100 }, { "epoch": 59.53, "learning_rate": 1.26390475732344e-05, "loss": 0.0, "step": 276150 }, { "epoch": 59.54, "learning_rate": 1.2633474639087384e-05, "loss": 0.0, "step": 276200 }, { "epoch": 59.55, "learning_rate": 1.2627902039919033e-05, "loss": 0.0, "step": 276250 }, { "epoch": 59.56, "learning_rate": 1.2622329776518141e-05, "loss": 0.0, "step": 276300 }, { "epoch": 59.57, "learning_rate": 1.2616757849673452e-05, "loss": 0.0, "step": 276350 }, { "epoch": 59.58, "learning_rate": 1.2611186260173673e-05, "loss": 0.0, "step": 276400 }, { "epoch": 59.59, "learning_rate": 1.2605615008807453e-05, "loss": 0.0, "step": 276450 }, { "epoch": 59.6, "learning_rate": 1.2600044096363393e-05, "loss": 0.0, "step": 276500 }, { "epoch": 59.61, "learning_rate": 1.2594473523630044e-05, "loss": 0.0, "step": 276550 }, { "epoch": 59.62, "learning_rate": 1.2588903291395923e-05, "loss": 0.0, "step": 276600 }, { "epoch": 59.64, "learning_rate": 1.2583333400449485e-05, "loss": 0.0, "step": 276650 }, { "epoch": 59.65, "learning_rate": 1.2577763851579145e-05, "loss": 0.0, "step": 276700 }, { "epoch": 59.66, "learning_rate": 1.257219464557325e-05, "loss": 0.0, "step": 276750 }, { "epoch": 59.67, "learning_rate": 1.2566625783220139e-05, "loss": 0.0, "step": 276800 }, { "epoch": 59.68, "learning_rate": 1.2561057265308058e-05, "loss": 0.0, "step": 276850 }, { "epoch": 59.69, "learning_rate": 1.255548909262523e-05, "loss": 0.0, "step": 276900 }, { "epoch": 59.7, "learning_rate": 1.2549921265959827e-05, "loss": 0.0, "step": 276950 }, { "epoch": 59.71, "learning_rate": 1.2544353786099953e-05, "loss": 0.0, "step": 277000 }, { "epoch": 59.72, "learning_rate": 1.2538786653833698e-05, "loss": 0.0, "step": 277050 }, { "epoch": 59.73, "learning_rate": 1.2533219869949072e-05, "loss": 0.0, "step": 277100 }, { "epoch": 59.74, "learning_rate": 1.2527653435234045e-05, "loss": 0.0, "step": 277150 }, { "epoch": 59.75, "learning_rate": 1.2522087350476536e-05, "loss": 0.0, "step": 277200 }, { "epoch": 59.77, "learning_rate": 1.2516521616464426e-05, "loss": 0.0, "step": 277250 }, { "epoch": 59.78, "learning_rate": 1.2510956233985539e-05, "loss": 0.0, "step": 277300 }, { "epoch": 59.79, "learning_rate": 1.2505391203827638e-05, "loss": 0.0, "step": 277350 }, { "epoch": 59.8, "learning_rate": 1.2499826526778444e-05, "loss": 0.0, "step": 277400 }, { "epoch": 59.81, "learning_rate": 1.2494262203625642e-05, "loss": 0.0, "step": 277450 }, { "epoch": 59.82, "learning_rate": 1.248869823515685e-05, "loss": 0.0, "step": 277500 }, { "epoch": 59.83, "learning_rate": 1.2483134622159642e-05, "loss": 0.0, "step": 277550 }, { "epoch": 59.84, "learning_rate": 1.2477571365421531e-05, "loss": 0.0, "step": 277600 }, { "epoch": 59.85, "learning_rate": 1.2472008465730003e-05, "loss": 0.0, "step": 277650 }, { "epoch": 59.86, "learning_rate": 1.246644592387247e-05, "loss": 0.0, "step": 277700 }, { "epoch": 59.87, "learning_rate": 1.2460883740636305e-05, "loss": 0.0, "step": 277750 }, { "epoch": 59.88, "learning_rate": 1.2455321916808835e-05, "loss": 0.0, "step": 277800 }, { "epoch": 59.89, "learning_rate": 1.2449760453177309e-05, "loss": 0.0, "step": 277850 }, { "epoch": 59.91, "learning_rate": 1.2444199350528968e-05, "loss": 0.0, "step": 277900 }, { "epoch": 59.92, "learning_rate": 1.2438638609650968e-05, "loss": 0.0, "step": 277950 }, { "epoch": 59.93, "learning_rate": 1.2433078231330427e-05, "loss": 0.0, "step": 278000 }, { "epoch": 59.94, "learning_rate": 1.2427518216354404e-05, "loss": 0.0, "step": 278050 }, { "epoch": 59.95, "learning_rate": 1.2421958565509922e-05, "loss": 0.0, "step": 278100 }, { "epoch": 59.96, "learning_rate": 1.2416399279583938e-05, "loss": 0.0, "step": 278150 }, { "epoch": 59.97, "learning_rate": 1.2410840359363363e-05, "loss": 0.0, "step": 278200 }, { "epoch": 59.98, "learning_rate": 1.2405281805635045e-05, "loss": 0.0, "step": 278250 }, { "epoch": 59.99, "learning_rate": 1.2399723619185806e-05, "loss": 0.0, "step": 278300 }, { "epoch": 60.0, "learning_rate": 1.2394165800802392e-05, "loss": 0.0, "step": 278350 }, { "epoch": 60.01, "learning_rate": 1.2388608351271508e-05, "loss": 0.0, "step": 278400 }, { "epoch": 60.02, "learning_rate": 1.23830512713798e-05, "loss": 0.0, "step": 278450 }, { "epoch": 60.03, "learning_rate": 1.2377494561913863e-05, "loss": 0.0, "step": 278500 }, { "epoch": 60.05, "learning_rate": 1.237193822366025e-05, "loss": 0.0, "step": 278550 }, { "epoch": 60.06, "learning_rate": 1.2366382257405454e-05, "loss": 0.0, "step": 278600 }, { "epoch": 60.07, "learning_rate": 1.2360826663935913e-05, "loss": 0.0, "step": 278650 }, { "epoch": 60.08, "learning_rate": 1.2355271444038002e-05, "loss": 0.0, "step": 278700 }, { "epoch": 60.09, "learning_rate": 1.2349716598498077e-05, "loss": 0.0, "step": 278750 }, { "epoch": 60.1, "learning_rate": 1.2344162128102404e-05, "loss": 0.0, "step": 278800 }, { "epoch": 60.11, "learning_rate": 1.2338608033637216e-05, "loss": 0.0, "step": 278850 }, { "epoch": 60.12, "learning_rate": 1.2333054315888684e-05, "loss": 0.0, "step": 278900 }, { "epoch": 60.13, "learning_rate": 1.2327500975642939e-05, "loss": 0.0, "step": 278950 }, { "epoch": 60.14, "learning_rate": 1.2321948013686046e-05, "loss": 0.0, "step": 279000 }, { "epoch": 60.15, "learning_rate": 1.2316395430804012e-05, "loss": 0.0, "step": 279050 }, { "epoch": 60.16, "learning_rate": 1.2310843227782803e-05, "loss": 0.0, "step": 279100 }, { "epoch": 60.17, "learning_rate": 1.2305291405408328e-05, "loss": 0.0, "step": 279150 }, { "epoch": 60.19, "learning_rate": 1.2299739964466439e-05, "loss": 0.0, "step": 279200 }, { "epoch": 60.2, "learning_rate": 1.2294188905742941e-05, "loss": 0.0, "step": 279250 }, { "epoch": 60.21, "learning_rate": 1.228863823002357e-05, "loss": 0.0, "step": 279300 }, { "epoch": 60.22, "learning_rate": 1.2283087938094016e-05, "loss": 0.0, "step": 279350 }, { "epoch": 60.23, "learning_rate": 1.2277538030739926e-05, "loss": 0.0, "step": 279400 }, { "epoch": 60.24, "learning_rate": 1.2271988508746876e-05, "loss": 0.0, "step": 279450 }, { "epoch": 60.25, "learning_rate": 1.2266439372900399e-05, "loss": 0.0, "step": 279500 }, { "epoch": 60.26, "learning_rate": 1.2260890623985954e-05, "loss": 0.0, "step": 279550 }, { "epoch": 60.27, "learning_rate": 1.225534226278898e-05, "loss": 0.0, "step": 279600 }, { "epoch": 60.28, "learning_rate": 1.2249794290094826e-05, "loss": 0.0, "step": 279650 }, { "epoch": 60.29, "learning_rate": 1.2244246706688805e-05, "loss": 0.0, "step": 279700 }, { "epoch": 60.3, "learning_rate": 1.2238699513356167e-05, "loss": 0.0, "step": 279750 }, { "epoch": 60.31, "learning_rate": 1.2233152710882115e-05, "loss": 0.0, "step": 279800 }, { "epoch": 60.33, "learning_rate": 1.2227606300051795e-05, "loss": 0.0, "step": 279850 }, { "epoch": 60.34, "learning_rate": 1.2222060281650285e-05, "loss": 0.0, "step": 279900 }, { "epoch": 60.35, "learning_rate": 1.2216514656462619e-05, "loss": 0.0, "step": 279950 }, { "epoch": 60.36, "learning_rate": 1.2210969425273776e-05, "loss": 0.0, "step": 280000 }, { "epoch": 60.37, "learning_rate": 1.220542458886868e-05, "loss": 0.0, "step": 280050 }, { "epoch": 60.38, "learning_rate": 1.2199880148032193e-05, "loss": 0.0, "step": 280100 }, { "epoch": 60.39, "learning_rate": 1.2194336103549118e-05, "loss": 0.0, "step": 280150 }, { "epoch": 60.4, "learning_rate": 1.218879245620421e-05, "loss": 0.0, "step": 280200 }, { "epoch": 60.41, "learning_rate": 1.2183249206782169e-05, "loss": 0.0, "step": 280250 }, { "epoch": 60.42, "learning_rate": 1.2177706356067631e-05, "loss": 0.0, "step": 280300 }, { "epoch": 60.43, "learning_rate": 1.2172163904845186e-05, "loss": 0.0, "step": 280350 }, { "epoch": 60.44, "learning_rate": 1.2166621853899347e-05, "loss": 0.0, "step": 280400 }, { "epoch": 60.45, "learning_rate": 1.2161080204014604e-05, "loss": 0.0, "step": 280450 }, { "epoch": 60.47, "learning_rate": 1.2155538955975355e-05, "loss": 0.0, "step": 280500 }, { "epoch": 60.48, "learning_rate": 1.2149998110565962e-05, "loss": 0.0, "step": 280550 }, { "epoch": 60.49, "learning_rate": 1.214445766857072e-05, "loss": 0.0, "step": 280600 }, { "epoch": 60.5, "learning_rate": 1.213891763077388e-05, "loss": 0.0, "step": 280650 }, { "epoch": 60.51, "learning_rate": 1.2133377997959628e-05, "loss": 0.0, "step": 280700 }, { "epoch": 60.52, "learning_rate": 1.2127838770912082e-05, "loss": 0.0, "step": 280750 }, { "epoch": 60.53, "learning_rate": 1.2122299950415315e-05, "loss": 0.0, "step": 280800 }, { "epoch": 60.54, "learning_rate": 1.2116761537253346e-05, "loss": 0.0, "step": 280850 }, { "epoch": 60.55, "learning_rate": 1.2111223532210126e-05, "loss": 0.0, "step": 280900 }, { "epoch": 60.56, "learning_rate": 1.2105685936069558e-05, "loss": 0.0, "step": 280950 }, { "epoch": 60.57, "learning_rate": 1.2100148749615475e-05, "loss": 0.0, "step": 281000 }, { "epoch": 60.58, "learning_rate": 1.2094611973631657e-05, "loss": 0.0, "step": 281050 }, { "epoch": 60.59, "learning_rate": 1.2089075608901834e-05, "loss": 0.0, "step": 281100 }, { "epoch": 60.61, "learning_rate": 1.2083539656209669e-05, "loss": 0.0, "step": 281150 }, { "epoch": 60.62, "learning_rate": 1.2078004116338774e-05, "loss": 0.0, "step": 281200 }, { "epoch": 60.63, "learning_rate": 1.2072468990072684e-05, "loss": 0.0, "step": 281250 }, { "epoch": 60.64, "learning_rate": 1.2066934278194906e-05, "loss": 0.0, "step": 281300 }, { "epoch": 60.65, "learning_rate": 1.2061399981488862e-05, "loss": 0.0, "step": 281350 }, { "epoch": 60.66, "learning_rate": 1.2055866100737923e-05, "loss": 0.0, "step": 281400 }, { "epoch": 60.67, "learning_rate": 1.2050332636725405e-05, "loss": 0.0, "step": 281450 }, { "epoch": 60.68, "learning_rate": 1.2044799590234566e-05, "loss": 0.0, "step": 281500 }, { "epoch": 60.69, "learning_rate": 1.2039266962048602e-05, "loss": 0.0, "step": 281550 }, { "epoch": 60.7, "learning_rate": 1.2033734752950646e-05, "loss": 0.0, "step": 281600 }, { "epoch": 60.71, "learning_rate": 1.2028202963723772e-05, "loss": 0.0, "step": 281650 }, { "epoch": 60.72, "learning_rate": 1.2022671595151004e-05, "loss": 0.0, "step": 281700 }, { "epoch": 60.74, "learning_rate": 1.20171406480153e-05, "loss": 0.0, "step": 281750 }, { "epoch": 60.75, "learning_rate": 1.2011610123099558e-05, "loss": 0.0, "step": 281800 }, { "epoch": 60.76, "learning_rate": 1.2006080021186614e-05, "loss": 0.0, "step": 281850 }, { "epoch": 60.77, "learning_rate": 1.2000550343059245e-05, "loss": 0.0, "step": 281900 }, { "epoch": 60.78, "learning_rate": 1.1995021089500175e-05, "loss": 0.0, "step": 281950 }, { "epoch": 60.79, "learning_rate": 1.1989492261292062e-05, "loss": 0.0, "step": 282000 }, { "epoch": 60.8, "learning_rate": 1.1983963859217507e-05, "loss": 0.0, "step": 282050 }, { "epoch": 60.81, "learning_rate": 1.1978435884059037e-05, "loss": 0.0, "step": 282100 }, { "epoch": 60.82, "learning_rate": 1.1972908336599144e-05, "loss": 0.0, "step": 282150 }, { "epoch": 60.83, "learning_rate": 1.196738121762024e-05, "loss": 0.0, "step": 282200 }, { "epoch": 60.84, "learning_rate": 1.1961854527904679e-05, "loss": 0.0, "step": 282250 }, { "epoch": 60.85, "learning_rate": 1.1956328268234756e-05, "loss": 0.0, "step": 282300 }, { "epoch": 60.86, "learning_rate": 1.1950802439392713e-05, "loss": 0.0, "step": 282350 }, { "epoch": 60.88, "learning_rate": 1.1945277042160722e-05, "loss": 0.0, "step": 282400 }, { "epoch": 60.89, "learning_rate": 1.1939752077320892e-05, "loss": 0.0, "step": 282450 }, { "epoch": 60.9, "learning_rate": 1.1934227545655271e-05, "loss": 0.0, "step": 282500 }, { "epoch": 60.91, "learning_rate": 1.192870344794586e-05, "loss": 0.0, "step": 282550 }, { "epoch": 60.92, "learning_rate": 1.1923179784974584e-05, "loss": 0.0, "step": 282600 }, { "epoch": 60.93, "learning_rate": 1.1917656557523314e-05, "loss": 0.0, "step": 282650 }, { "epoch": 60.94, "learning_rate": 1.1912133766373846e-05, "loss": 0.0, "step": 282700 }, { "epoch": 60.95, "learning_rate": 1.1906611412307928e-05, "loss": 0.0, "step": 282750 }, { "epoch": 60.96, "learning_rate": 1.1901089496107247e-05, "loss": 0.0, "step": 282800 }, { "epoch": 60.97, "learning_rate": 1.1895568018553419e-05, "loss": 0.0, "step": 282850 }, { "epoch": 60.98, "learning_rate": 1.1890046980428008e-05, "loss": 0.0, "step": 282900 }, { "epoch": 60.99, "learning_rate": 1.1884526382512495e-05, "loss": 0.0, "step": 282950 }, { "epoch": 61.0, "learning_rate": 1.1879006225588332e-05, "loss": 0.0, "step": 283000 }, { "epoch": 61.02, "learning_rate": 1.1873486510436881e-05, "loss": 0.0, "step": 283050 }, { "epoch": 61.03, "learning_rate": 1.186796723783945e-05, "loss": 0.0, "step": 283100 }, { "epoch": 61.04, "learning_rate": 1.1862448408577286e-05, "loss": 0.0, "step": 283150 }, { "epoch": 61.05, "learning_rate": 1.1856930023431576e-05, "loss": 0.0, "step": 283200 }, { "epoch": 61.06, "learning_rate": 1.1851412083183439e-05, "loss": 0.0, "step": 283250 }, { "epoch": 61.07, "learning_rate": 1.1845894588613928e-05, "loss": 0.0, "step": 283300 }, { "epoch": 61.08, "learning_rate": 1.1840377540504041e-05, "loss": 0.0, "step": 283350 }, { "epoch": 61.09, "learning_rate": 1.1834860939634707e-05, "loss": 0.0, "step": 283400 }, { "epoch": 61.1, "learning_rate": 1.1829344786786796e-05, "loss": 0.0, "step": 283450 }, { "epoch": 61.11, "learning_rate": 1.1823829082741115e-05, "loss": 0.0, "step": 283500 }, { "epoch": 61.12, "learning_rate": 1.18183138282784e-05, "loss": 0.0, "step": 283550 }, { "epoch": 61.13, "learning_rate": 1.1812799024179327e-05, "loss": 0.0, "step": 283600 }, { "epoch": 61.14, "learning_rate": 1.1807284671224516e-05, "loss": 0.0, "step": 283650 }, { "epoch": 61.16, "learning_rate": 1.180177077019451e-05, "loss": 0.0, "step": 283700 }, { "epoch": 61.17, "learning_rate": 1.1796257321869804e-05, "loss": 0.0, "step": 283750 }, { "epoch": 61.18, "learning_rate": 1.1790744327030802e-05, "loss": 0.0, "step": 283800 }, { "epoch": 61.19, "learning_rate": 1.1785231786457886e-05, "loss": 0.0, "step": 283850 }, { "epoch": 61.2, "learning_rate": 1.177971970093133e-05, "loss": 0.0, "step": 283900 }, { "epoch": 61.21, "learning_rate": 1.177420807123137e-05, "loss": 0.0, "step": 283950 }, { "epoch": 61.22, "learning_rate": 1.1768696898138166e-05, "loss": 0.0, "step": 284000 }, { "epoch": 61.23, "learning_rate": 1.1763186182431823e-05, "loss": 0.0, "step": 284050 }, { "epoch": 61.24, "learning_rate": 1.1757675924892376e-05, "loss": 0.0, "step": 284100 }, { "epoch": 61.25, "learning_rate": 1.1752166126299791e-05, "loss": 0.0, "step": 284150 }, { "epoch": 61.26, "learning_rate": 1.1746656787433973e-05, "loss": 0.0, "step": 284200 }, { "epoch": 61.27, "learning_rate": 1.174114790907476e-05, "loss": 0.0, "step": 284250 }, { "epoch": 61.28, "learning_rate": 1.1735639492001935e-05, "loss": 0.0, "step": 284300 }, { "epoch": 61.3, "learning_rate": 1.1730131536995205e-05, "loss": 0.0, "step": 284350 }, { "epoch": 61.31, "learning_rate": 1.1724624044834208e-05, "loss": 0.0, "step": 284400 }, { "epoch": 61.32, "learning_rate": 1.1719117016298522e-05, "loss": 0.0, "step": 284450 }, { "epoch": 61.33, "learning_rate": 1.171361045216767e-05, "loss": 0.0, "step": 284500 }, { "epoch": 61.34, "learning_rate": 1.170810435322109e-05, "loss": 0.0, "step": 284550 }, { "epoch": 61.35, "learning_rate": 1.1702598720238171e-05, "loss": 0.0, "step": 284600 }, { "epoch": 61.36, "learning_rate": 1.1697093553998216e-05, "loss": 0.0, "step": 284650 }, { "epoch": 61.37, "learning_rate": 1.169158885528049e-05, "loss": 0.0, "step": 284700 }, { "epoch": 61.38, "learning_rate": 1.1686084624864165e-05, "loss": 0.0, "step": 284750 }, { "epoch": 61.39, "learning_rate": 1.1680580863528362e-05, "loss": 0.0, "step": 284800 }, { "epoch": 61.4, "learning_rate": 1.1675077572052127e-05, "loss": 0.0, "step": 284850 }, { "epoch": 61.41, "learning_rate": 1.166957475121445e-05, "loss": 0.0, "step": 284900 }, { "epoch": 61.42, "learning_rate": 1.1664072401794252e-05, "loss": 0.0, "step": 284950 }, { "epoch": 61.44, "learning_rate": 1.165857052457037e-05, "loss": 0.0, "step": 285000 }, { "epoch": 61.45, "learning_rate": 1.1653069120321598e-05, "loss": 0.0, "step": 285050 }, { "epoch": 61.46, "learning_rate": 1.1647568189826645e-05, "loss": 0.0, "step": 285100 }, { "epoch": 61.47, "learning_rate": 1.1642067733864171e-05, "loss": 0.0, "step": 285150 }, { "epoch": 61.48, "learning_rate": 1.1636567753212757e-05, "loss": 0.0, "step": 285200 }, { "epoch": 61.49, "learning_rate": 1.163106824865091e-05, "loss": 0.0, "step": 285250 }, { "epoch": 61.5, "learning_rate": 1.1625569220957078e-05, "loss": 0.0, "step": 285300 }, { "epoch": 61.51, "learning_rate": 1.1620070670909651e-05, "loss": 0.0, "step": 285350 }, { "epoch": 61.52, "learning_rate": 1.1614572599286936e-05, "loss": 0.0, "step": 285400 }, { "epoch": 61.53, "learning_rate": 1.1609075006867178e-05, "loss": 0.0, "step": 285450 }, { "epoch": 61.54, "learning_rate": 1.1603577894428553e-05, "loss": 0.0, "step": 285500 }, { "epoch": 61.55, "learning_rate": 1.1598081262749178e-05, "loss": 0.0, "step": 285550 }, { "epoch": 61.56, "learning_rate": 1.1592585112607086e-05, "loss": 0.0, "step": 285600 }, { "epoch": 61.58, "learning_rate": 1.1587089444780253e-05, "loss": 0.0, "step": 285650 }, { "epoch": 61.59, "learning_rate": 1.1581594260046578e-05, "loss": 0.0, "step": 285700 }, { "epoch": 61.6, "learning_rate": 1.1576099559183908e-05, "loss": 0.0, "step": 285750 }, { "epoch": 61.61, "learning_rate": 1.157060534297001e-05, "loss": 0.0, "step": 285800 }, { "epoch": 61.62, "learning_rate": 1.1565111612182574e-05, "loss": 0.0, "step": 285850 }, { "epoch": 61.63, "learning_rate": 1.1559618367599237e-05, "loss": 0.0, "step": 285900 }, { "epoch": 61.64, "learning_rate": 1.1554125609997555e-05, "loss": 0.0, "step": 285950 }, { "epoch": 61.65, "learning_rate": 1.154863334015503e-05, "loss": 0.0, "step": 286000 }, { "epoch": 61.66, "learning_rate": 1.1543141558849084e-05, "loss": 0.0, "step": 286050 }, { "epoch": 61.67, "learning_rate": 1.1537650266857066e-05, "loss": 0.0, "step": 286100 }, { "epoch": 61.68, "learning_rate": 1.153215946495626e-05, "loss": 0.0, "step": 286150 }, { "epoch": 61.69, "learning_rate": 1.152666915392389e-05, "loss": 0.0, "step": 286200 }, { "epoch": 61.71, "learning_rate": 1.15211793345371e-05, "loss": 0.0, "step": 286250 }, { "epoch": 61.72, "learning_rate": 1.1515690007572966e-05, "loss": 0.0, "step": 286300 }, { "epoch": 61.73, "learning_rate": 1.1510201173808494e-05, "loss": 0.0, "step": 286350 }, { "epoch": 61.74, "learning_rate": 1.1504712834020627e-05, "loss": 0.0, "step": 286400 }, { "epoch": 61.75, "learning_rate": 1.1499224988986224e-05, "loss": 0.0, "step": 286450 }, { "epoch": 61.76, "learning_rate": 1.1493737639482088e-05, "loss": 0.0, "step": 286500 }, { "epoch": 61.77, "learning_rate": 1.1488250786284942e-05, "loss": 0.0, "step": 286550 }, { "epoch": 61.78, "learning_rate": 1.1482764430171448e-05, "loss": 0.0, "step": 286600 }, { "epoch": 61.79, "learning_rate": 1.1477278571918197e-05, "loss": 0.0, "step": 286650 }, { "epoch": 61.8, "learning_rate": 1.1471793212301695e-05, "loss": 0.0, "step": 286700 }, { "epoch": 61.81, "learning_rate": 1.1466308352098392e-05, "loss": 0.0, "step": 286750 }, { "epoch": 61.82, "learning_rate": 1.1460823992084658e-05, "loss": 0.0, "step": 286800 }, { "epoch": 61.83, "learning_rate": 1.1455340133036809e-05, "loss": 0.0, "step": 286850 }, { "epoch": 61.85, "learning_rate": 1.1449856775731069e-05, "loss": 0.0, "step": 286900 }, { "epoch": 61.86, "learning_rate": 1.1444373920943607e-05, "loss": 0.0, "step": 286950 }, { "epoch": 61.87, "learning_rate": 1.1438891569450501e-05, "loss": 0.0, "step": 287000 }, { "epoch": 61.88, "learning_rate": 1.1433409722027785e-05, "loss": 0.0, "step": 287050 }, { "epoch": 61.89, "learning_rate": 1.1427928379451403e-05, "loss": 0.0, "step": 287100 }, { "epoch": 61.9, "learning_rate": 1.1422447542497233e-05, "loss": 0.0, "step": 287150 }, { "epoch": 61.91, "learning_rate": 1.1416967211941075e-05, "loss": 0.0, "step": 287200 }, { "epoch": 61.92, "learning_rate": 1.1411487388558672e-05, "loss": 0.0, "step": 287250 }, { "epoch": 61.93, "learning_rate": 1.1406008073125678e-05, "loss": 0.0, "step": 287300 }, { "epoch": 61.94, "learning_rate": 1.1400529266417684e-05, "loss": 0.0, "step": 287350 }, { "epoch": 61.95, "learning_rate": 1.139505096921021e-05, "loss": 0.0, "step": 287400 }, { "epoch": 61.96, "learning_rate": 1.1389573182278705e-05, "loss": 0.0, "step": 287450 }, { "epoch": 61.97, "learning_rate": 1.138409590639854e-05, "loss": 0.0, "step": 287500 }, { "epoch": 61.99, "learning_rate": 1.1378619142345019e-05, "loss": 0.0, "step": 287550 }, { "epoch": 62.0, "learning_rate": 1.1373142890893366e-05, "loss": 0.0, "step": 287600 }, { "epoch": 62.01, "learning_rate": 1.1367667152818738e-05, "loss": 0.0, "step": 287650 }, { "epoch": 62.02, "learning_rate": 1.136219192889622e-05, "loss": 0.0, "step": 287700 }, { "epoch": 62.03, "learning_rate": 1.1356717219900826e-05, "loss": 0.0, "step": 287750 }, { "epoch": 62.04, "learning_rate": 1.1351243026607497e-05, "loss": 0.0, "step": 287800 }, { "epoch": 62.05, "learning_rate": 1.1345769349791084e-05, "loss": 0.0, "step": 287850 }, { "epoch": 62.06, "learning_rate": 1.1340296190226393e-05, "loss": 0.0, "step": 287900 }, { "epoch": 62.07, "learning_rate": 1.1334823548688138e-05, "loss": 0.0, "step": 287950 }, { "epoch": 62.08, "learning_rate": 1.1329351425950963e-05, "loss": 0.0, "step": 288000 }, { "epoch": 62.09, "learning_rate": 1.1323879822789439e-05, "loss": 0.0, "step": 288050 }, { "epoch": 62.1, "learning_rate": 1.1318408739978071e-05, "loss": 0.0, "step": 288100 }, { "epoch": 62.11, "learning_rate": 1.131293817829128e-05, "loss": 0.0, "step": 288150 }, { "epoch": 62.13, "learning_rate": 1.1307468138503415e-05, "loss": 0.0, "step": 288200 }, { "epoch": 62.14, "learning_rate": 1.1301998621388755e-05, "loss": 0.0, "step": 288250 }, { "epoch": 62.15, "learning_rate": 1.12965296277215e-05, "loss": 0.0, "step": 288300 }, { "epoch": 62.16, "learning_rate": 1.1291061158275785e-05, "loss": 0.0, "step": 288350 }, { "epoch": 62.17, "learning_rate": 1.1285593213825666e-05, "loss": 0.0, "step": 288400 }, { "epoch": 62.18, "learning_rate": 1.1280125795145116e-05, "loss": 0.0, "step": 288450 }, { "epoch": 62.19, "learning_rate": 1.1274658903008043e-05, "loss": 0.0, "step": 288500 }, { "epoch": 62.2, "learning_rate": 1.1269192538188282e-05, "loss": 0.0, "step": 288550 }, { "epoch": 62.21, "learning_rate": 1.126372670145959e-05, "loss": 0.0, "step": 288600 }, { "epoch": 62.22, "learning_rate": 1.1258261393595652e-05, "loss": 0.0, "step": 288650 }, { "epoch": 62.23, "learning_rate": 1.1252796615370062e-05, "loss": 0.0, "step": 288700 }, { "epoch": 62.24, "learning_rate": 1.1247332367556367e-05, "loss": 0.0, "step": 288750 }, { "epoch": 62.25, "learning_rate": 1.1241868650928019e-05, "loss": 0.0, "step": 288800 }, { "epoch": 62.27, "learning_rate": 1.12364054662584e-05, "loss": 0.0, "step": 288850 }, { "epoch": 62.28, "learning_rate": 1.1230942814320815e-05, "loss": 0.0, "step": 288900 }, { "epoch": 62.29, "learning_rate": 1.12254806958885e-05, "loss": 0.0, "step": 288950 }, { "epoch": 62.3, "learning_rate": 1.1220019111734611e-05, "loss": 0.0, "step": 289000 }, { "epoch": 62.31, "learning_rate": 1.1214558062632223e-05, "loss": 0.0, "step": 289050 }, { "epoch": 62.32, "learning_rate": 1.1209097549354346e-05, "loss": 0.0, "step": 289100 }, { "epoch": 62.33, "learning_rate": 1.1203637572673899e-05, "loss": 0.0, "step": 289150 }, { "epoch": 62.34, "learning_rate": 1.1198178133363747e-05, "loss": 0.0, "step": 289200 }, { "epoch": 62.35, "learning_rate": 1.1192719232196665e-05, "loss": 0.0, "step": 289250 }, { "epoch": 62.36, "learning_rate": 1.1187260869945345e-05, "loss": 0.0, "step": 289300 }, { "epoch": 62.37, "learning_rate": 1.1181803047382417e-05, "loss": 0.0, "step": 289350 }, { "epoch": 62.38, "learning_rate": 1.1176345765280429e-05, "loss": 0.0, "step": 289400 }, { "epoch": 62.39, "learning_rate": 1.1170889024411854e-05, "loss": 0.0, "step": 289450 }, { "epoch": 62.41, "learning_rate": 1.1165432825549084e-05, "loss": 0.0, "step": 289500 }, { "epoch": 62.42, "learning_rate": 1.115997716946443e-05, "loss": 0.0, "step": 289550 }, { "epoch": 62.43, "learning_rate": 1.115452205693015e-05, "loss": 0.0, "step": 289600 }, { "epoch": 62.44, "learning_rate": 1.1149067488718394e-05, "loss": 0.0, "step": 289650 }, { "epoch": 62.45, "learning_rate": 1.1143613465601257e-05, "loss": 0.0, "step": 289700 }, { "epoch": 62.46, "learning_rate": 1.1138159988350742e-05, "loss": 0.0, "step": 289750 }, { "epoch": 62.47, "learning_rate": 1.1132707057738789e-05, "loss": 0.0, "step": 289800 }, { "epoch": 62.48, "learning_rate": 1.1127254674537253e-05, "loss": 0.0, "step": 289850 }, { "epoch": 62.49, "learning_rate": 1.1121802839517908e-05, "loss": 0.0, "step": 289900 }, { "epoch": 62.5, "learning_rate": 1.1116351553452454e-05, "loss": 0.0, "step": 289950 }, { "epoch": 62.51, "learning_rate": 1.1110900817112512e-05, "loss": 0.0, "step": 290000 }, { "epoch": 62.52, "learning_rate": 1.1105450631269636e-05, "loss": 0.0, "step": 290050 }, { "epoch": 62.54, "learning_rate": 1.1100000996695289e-05, "loss": 0.0, "step": 290100 }, { "epoch": 62.55, "learning_rate": 1.1094551914160852e-05, "loss": 0.0, "step": 290150 }, { "epoch": 62.56, "learning_rate": 1.1089103384437641e-05, "loss": 0.0, "step": 290200 }, { "epoch": 62.57, "learning_rate": 1.1083655408296893e-05, "loss": 0.0, "step": 290250 }, { "epoch": 62.58, "learning_rate": 1.1078207986509757e-05, "loss": 0.0, "step": 290300 }, { "epoch": 62.59, "learning_rate": 1.1072761119847314e-05, "loss": 0.0, "step": 290350 }, { "epoch": 62.6, "learning_rate": 1.1067314809080545e-05, "loss": 0.0, "step": 290400 }, { "epoch": 62.61, "learning_rate": 1.1061869054980392e-05, "loss": 0.0, "step": 290450 }, { "epoch": 62.62, "learning_rate": 1.1056423858317679e-05, "loss": 0.0, "step": 290500 }, { "epoch": 62.63, "learning_rate": 1.105097921986317e-05, "loss": 0.0, "step": 290550 }, { "epoch": 62.64, "learning_rate": 1.1045535140387546e-05, "loss": 0.0, "step": 290600 }, { "epoch": 62.65, "learning_rate": 1.1040091620661413e-05, "loss": 0.0, "step": 290650 }, { "epoch": 62.66, "learning_rate": 1.1034648661455295e-05, "loss": 0.0, "step": 290700 }, { "epoch": 62.68, "learning_rate": 1.102920626353963e-05, "loss": 0.0, "step": 290750 }, { "epoch": 62.69, "learning_rate": 1.1023764427684786e-05, "loss": 0.0, "step": 290800 }, { "epoch": 62.7, "learning_rate": 1.1018323154661044e-05, "loss": 0.0, "step": 290850 }, { "epoch": 62.71, "learning_rate": 1.1012882445238618e-05, "loss": 0.0, "step": 290900 }, { "epoch": 62.72, "learning_rate": 1.1007442300187628e-05, "loss": 0.0, "step": 290950 }, { "epoch": 62.73, "learning_rate": 1.100200272027812e-05, "loss": 0.0, "step": 291000 }, { "epoch": 62.74, "learning_rate": 1.0996563706280056e-05, "loss": 0.0, "step": 291050 }, { "epoch": 62.75, "learning_rate": 1.099112525896333e-05, "loss": 0.0, "step": 291100 }, { "epoch": 62.76, "learning_rate": 1.098568737909774e-05, "loss": 0.0, "step": 291150 }, { "epoch": 62.77, "learning_rate": 1.0980250067453019e-05, "loss": 0.0, "step": 291200 }, { "epoch": 62.78, "learning_rate": 1.0974813324798794e-05, "loss": 0.0, "step": 291250 }, { "epoch": 62.79, "learning_rate": 1.0969377151904652e-05, "loss": 0.0, "step": 291300 }, { "epoch": 62.8, "learning_rate": 1.096394154954006e-05, "loss": 0.0, "step": 291350 }, { "epoch": 62.82, "learning_rate": 1.0958506518474429e-05, "loss": 0.0, "step": 291400 }, { "epoch": 62.83, "learning_rate": 1.0953072059477075e-05, "loss": 0.0, "step": 291450 }, { "epoch": 62.84, "learning_rate": 1.0947638173317245e-05, "loss": 0.0, "step": 291500 }, { "epoch": 62.85, "learning_rate": 1.0942204860764096e-05, "loss": 0.0, "step": 291550 }, { "epoch": 62.86, "learning_rate": 1.0936772122586706e-05, "loss": 0.0, "step": 291600 }, { "epoch": 62.87, "learning_rate": 1.0931339959554071e-05, "loss": 0.0, "step": 291650 }, { "epoch": 62.88, "learning_rate": 1.0925908372435105e-05, "loss": 0.0, "step": 291700 }, { "epoch": 62.89, "learning_rate": 1.0920477361998648e-05, "loss": 0.0, "step": 291750 }, { "epoch": 62.9, "learning_rate": 1.0915046929013452e-05, "loss": 0.0, "step": 291800 }, { "epoch": 62.91, "learning_rate": 1.0909617074248185e-05, "loss": 0.0, "step": 291850 }, { "epoch": 62.92, "learning_rate": 1.0904187798471433e-05, "loss": 0.0, "step": 291900 }, { "epoch": 62.93, "learning_rate": 1.0898759102451709e-05, "loss": 0.0, "step": 291950 }, { "epoch": 62.94, "learning_rate": 1.0893330986957439e-05, "loss": 0.0, "step": 292000 }, { "epoch": 62.96, "learning_rate": 1.0887903452756962e-05, "loss": 0.0, "step": 292050 }, { "epoch": 62.97, "learning_rate": 1.0882476500618531e-05, "loss": 0.0, "step": 292100 }, { "epoch": 62.98, "learning_rate": 1.0877050131310342e-05, "loss": 0.0, "step": 292150 }, { "epoch": 62.99, "learning_rate": 1.087162434560048e-05, "loss": 0.0, "step": 292200 }, { "epoch": 63.0, "learning_rate": 1.0866199144256958e-05, "loss": 0.0, "step": 292250 }, { "epoch": 63.01, "learning_rate": 1.0860774528047705e-05, "loss": 0.0, "step": 292300 }, { "epoch": 63.02, "learning_rate": 1.0855350497740574e-05, "loss": 0.0, "step": 292350 }, { "epoch": 63.03, "learning_rate": 1.0849927054103331e-05, "loss": 0.0, "step": 292400 }, { "epoch": 63.04, "learning_rate": 1.084450419790365e-05, "loss": 0.0, "step": 292450 }, { "epoch": 63.05, "learning_rate": 1.0839081929909133e-05, "loss": 0.0, "step": 292500 }, { "epoch": 63.06, "learning_rate": 1.0833660250887293e-05, "loss": 0.0, "step": 292550 }, { "epoch": 63.07, "learning_rate": 1.0828239161605564e-05, "loss": 0.0, "step": 292600 }, { "epoch": 63.08, "learning_rate": 1.08228186628313e-05, "loss": 0.0, "step": 292650 }, { "epoch": 63.1, "learning_rate": 1.0817398755331755e-05, "loss": 0.0, "step": 292700 }, { "epoch": 63.11, "learning_rate": 1.0811979439874113e-05, "loss": 0.0, "step": 292750 }, { "epoch": 63.12, "learning_rate": 1.0806560717225476e-05, "loss": 0.0, "step": 292800 }, { "epoch": 63.13, "learning_rate": 1.0801142588152857e-05, "loss": 0.0, "step": 292850 }, { "epoch": 63.14, "learning_rate": 1.0795725053423184e-05, "loss": 0.0, "step": 292900 }, { "epoch": 63.15, "learning_rate": 1.0790308113803292e-05, "loss": 0.0, "step": 292950 }, { "epoch": 63.16, "learning_rate": 1.078489177005996e-05, "loss": 0.0, "step": 293000 }, { "epoch": 63.17, "learning_rate": 1.0779476022959851e-05, "loss": 0.0, "step": 293050 }, { "epoch": 63.18, "learning_rate": 1.0774060873269564e-05, "loss": 0.0, "step": 293100 }, { "epoch": 63.19, "learning_rate": 1.0768646321755607e-05, "loss": 0.0, "step": 293150 }, { "epoch": 63.2, "learning_rate": 1.076323236918439e-05, "loss": 0.0, "step": 293200 }, { "epoch": 63.21, "learning_rate": 1.075781901632227e-05, "loss": 0.0, "step": 293250 }, { "epoch": 63.22, "learning_rate": 1.0752406263935492e-05, "loss": 0.0, "step": 293300 }, { "epoch": 63.24, "learning_rate": 1.0746994112790219e-05, "loss": 0.0, "step": 293350 }, { "epoch": 63.25, "learning_rate": 1.0741582563652538e-05, "loss": 0.0, "step": 293400 }, { "epoch": 63.26, "learning_rate": 1.0736171617288452e-05, "loss": 0.0, "step": 293450 }, { "epoch": 63.27, "learning_rate": 1.0730761274463871e-05, "loss": 0.0, "step": 293500 }, { "epoch": 63.28, "learning_rate": 1.0725351535944619e-05, "loss": 0.0, "step": 293550 }, { "epoch": 63.29, "learning_rate": 1.0719942402496435e-05, "loss": 0.0, "step": 293600 }, { "epoch": 63.3, "learning_rate": 1.0714533874884984e-05, "loss": 0.0, "step": 293650 }, { "epoch": 63.31, "learning_rate": 1.0709125953875829e-05, "loss": 0.0, "step": 293700 }, { "epoch": 63.32, "learning_rate": 1.070371864023446e-05, "loss": 0.0, "step": 293750 }, { "epoch": 63.33, "learning_rate": 1.0698311934726264e-05, "loss": 0.0, "step": 293800 }, { "epoch": 63.34, "learning_rate": 1.0692905838116571e-05, "loss": 0.0, "step": 293850 }, { "epoch": 63.35, "learning_rate": 1.0687500351170596e-05, "loss": 0.0, "step": 293900 }, { "epoch": 63.36, "learning_rate": 1.0682095474653476e-05, "loss": 0.0, "step": 293950 }, { "epoch": 63.38, "learning_rate": 1.0676691209330275e-05, "loss": 0.0, "step": 294000 }, { "epoch": 63.39, "learning_rate": 1.0671287555965947e-05, "loss": 0.0, "step": 294050 }, { "epoch": 63.4, "learning_rate": 1.0665884515325384e-05, "loss": 0.0, "step": 294100 }, { "epoch": 63.41, "learning_rate": 1.0660482088173373e-05, "loss": 0.0, "step": 294150 }, { "epoch": 63.42, "learning_rate": 1.0655080275274623e-05, "loss": 0.0, "step": 294200 }, { "epoch": 63.43, "learning_rate": 1.064967907739375e-05, "loss": 0.0, "step": 294250 }, { "epoch": 63.44, "learning_rate": 1.0644278495295291e-05, "loss": 0.0, "step": 294300 }, { "epoch": 63.45, "learning_rate": 1.0638878529743696e-05, "loss": 0.0, "step": 294350 }, { "epoch": 63.46, "learning_rate": 1.0633479181503313e-05, "loss": 0.0, "step": 294400 }, { "epoch": 63.47, "learning_rate": 1.0628080451338415e-05, "loss": 0.0, "step": 294450 }, { "epoch": 63.48, "learning_rate": 1.062268234001319e-05, "loss": 0.0, "step": 294500 }, { "epoch": 63.49, "learning_rate": 1.0617284848291733e-05, "loss": 0.0, "step": 294550 }, { "epoch": 63.51, "learning_rate": 1.0611887976938056e-05, "loss": 0.0, "step": 294600 }, { "epoch": 63.52, "learning_rate": 1.0606491726716064e-05, "loss": 0.0, "step": 294650 }, { "epoch": 63.53, "learning_rate": 1.0601096098389606e-05, "loss": 0.0, "step": 294700 }, { "epoch": 63.54, "learning_rate": 1.059570109272242e-05, "loss": 0.0, "step": 294750 }, { "epoch": 63.55, "learning_rate": 1.0590306710478159e-05, "loss": 0.0, "step": 294800 }, { "epoch": 63.56, "learning_rate": 1.0584912952420399e-05, "loss": 0.0, "step": 294850 }, { "epoch": 63.57, "learning_rate": 1.0579519819312606e-05, "loss": 0.0, "step": 294900 }, { "epoch": 63.58, "learning_rate": 1.0574127311918189e-05, "loss": 0.0, "step": 294950 }, { "epoch": 63.59, "learning_rate": 1.0568735431000437e-05, "loss": 0.0, "step": 295000 }, { "epoch": 63.6, "learning_rate": 1.056334417732257e-05, "loss": 0.0, "step": 295050 }, { "epoch": 63.61, "learning_rate": 1.055795355164771e-05, "loss": 0.0, "step": 295100 }, { "epoch": 63.62, "learning_rate": 1.0552563554738894e-05, "loss": 0.0, "step": 295150 }, { "epoch": 63.63, "learning_rate": 1.0547174187359075e-05, "loss": 0.0, "step": 295200 }, { "epoch": 63.65, "learning_rate": 1.0541785450271102e-05, "loss": 0.0, "step": 295250 }, { "epoch": 63.66, "learning_rate": 1.0536397344237747e-05, "loss": 0.0, "step": 295300 }, { "epoch": 63.67, "learning_rate": 1.0531009870021693e-05, "loss": 0.0, "step": 295350 }, { "epoch": 63.68, "learning_rate": 1.0525623028385525e-05, "loss": 0.0, "step": 295400 }, { "epoch": 63.69, "learning_rate": 1.0520236820091753e-05, "loss": 0.0, "step": 295450 }, { "epoch": 63.7, "learning_rate": 1.0514851245902771e-05, "loss": 0.0, "step": 295500 }, { "epoch": 63.71, "learning_rate": 1.050946630658092e-05, "loss": 0.0, "step": 295550 }, { "epoch": 63.72, "learning_rate": 1.0504082002888417e-05, "loss": 0.0, "step": 295600 }, { "epoch": 63.73, "learning_rate": 1.049869833558741e-05, "loss": 0.0, "step": 295650 }, { "epoch": 63.74, "learning_rate": 1.0493315305439951e-05, "loss": 0.0, "step": 295700 }, { "epoch": 63.75, "learning_rate": 1.048793291320799e-05, "loss": 0.0, "step": 295750 }, { "epoch": 63.76, "learning_rate": 1.0482551159653413e-05, "loss": 0.0, "step": 295800 }, { "epoch": 63.77, "learning_rate": 1.0477170045537992e-05, "loss": 0.0, "step": 295850 }, { "epoch": 63.79, "learning_rate": 1.047178957162342e-05, "loss": 0.0, "step": 295900 }, { "epoch": 63.8, "learning_rate": 1.0466409738671292e-05, "loss": 0.0, "step": 295950 }, { "epoch": 63.81, "learning_rate": 1.0461030547443122e-05, "loss": 0.0, "step": 296000 }, { "epoch": 63.82, "learning_rate": 1.0455651998700327e-05, "loss": 0.0, "step": 296050 }, { "epoch": 63.83, "learning_rate": 1.045027409320423e-05, "loss": 0.0, "step": 296100 }, { "epoch": 63.84, "learning_rate": 1.0444896831716065e-05, "loss": 0.0, "step": 296150 }, { "epoch": 63.85, "learning_rate": 1.0439520214996983e-05, "loss": 0.0, "step": 296200 }, { "epoch": 63.86, "learning_rate": 1.0434144243808035e-05, "loss": 0.0, "step": 296250 }, { "epoch": 63.87, "learning_rate": 1.0428768918910186e-05, "loss": 0.0, "step": 296300 }, { "epoch": 63.88, "learning_rate": 1.0423394241064292e-05, "loss": 0.0, "step": 296350 }, { "epoch": 63.89, "learning_rate": 1.0418020211031154e-05, "loss": 0.0, "step": 296400 }, { "epoch": 63.9, "learning_rate": 1.0412646829571444e-05, "loss": 0.0, "step": 296450 }, { "epoch": 63.91, "learning_rate": 1.040727409744576e-05, "loss": 0.0, "step": 296500 }, { "epoch": 63.93, "learning_rate": 1.0401902015414612e-05, "loss": 0.0, "step": 296550 }, { "epoch": 63.94, "learning_rate": 1.0396530584238393e-05, "loss": 0.0, "step": 296600 }, { "epoch": 63.95, "learning_rate": 1.0391159804677445e-05, "loss": 0.0, "step": 296650 }, { "epoch": 63.96, "learning_rate": 1.0385789677491986e-05, "loss": 0.0, "step": 296700 }, { "epoch": 63.97, "learning_rate": 1.0380420203442148e-05, "loss": 0.0, "step": 296750 }, { "epoch": 63.98, "learning_rate": 1.0375051383287971e-05, "loss": 0.0, "step": 296800 }, { "epoch": 63.99, "learning_rate": 1.0369683217789411e-05, "loss": 0.0, "step": 296850 }, { "epoch": 64.0, "learning_rate": 1.0364315707706328e-05, "loss": 0.0, "step": 296900 }, { "epoch": 64.01, "learning_rate": 1.0358948853798479e-05, "loss": 0.0, "step": 296950 }, { "epoch": 64.02, "learning_rate": 1.0353582656825534e-05, "loss": 0.0, "step": 297000 }, { "epoch": 64.03, "learning_rate": 1.0348217117547074e-05, "loss": 0.0, "step": 297050 }, { "epoch": 64.04, "learning_rate": 1.0342852236722589e-05, "loss": 0.0, "step": 297100 }, { "epoch": 64.05, "learning_rate": 1.0337488015111469e-05, "loss": 0.0, "step": 297150 }, { "epoch": 64.07, "learning_rate": 1.0332124453473001e-05, "loss": 0.0, "step": 297200 }, { "epoch": 64.08, "learning_rate": 1.0326761552566415e-05, "loss": 0.0, "step": 297250 }, { "epoch": 64.09, "learning_rate": 1.03213993131508e-05, "loss": 0.0, "step": 297300 }, { "epoch": 64.1, "learning_rate": 1.0316037735985183e-05, "loss": 0.0, "step": 297350 }, { "epoch": 64.11, "learning_rate": 1.031067682182849e-05, "loss": 0.0, "step": 297400 }, { "epoch": 64.12, "learning_rate": 1.0305316571439543e-05, "loss": 0.0, "step": 297450 }, { "epoch": 64.13, "learning_rate": 1.0299956985577091e-05, "loss": 0.0, "step": 297500 }, { "epoch": 64.14, "learning_rate": 1.0294598064999771e-05, "loss": 0.0, "step": 297550 }, { "epoch": 64.15, "learning_rate": 1.028923981046613e-05, "loss": 0.0, "step": 297600 }, { "epoch": 64.16, "learning_rate": 1.0283882222734619e-05, "loss": 0.0, "step": 297650 }, { "epoch": 64.17, "learning_rate": 1.0278525302563604e-05, "loss": 0.0, "step": 297700 }, { "epoch": 64.18, "learning_rate": 1.0273169050711356e-05, "loss": 0.0, "step": 297750 }, { "epoch": 64.19, "learning_rate": 1.0267813467936032e-05, "loss": 0.0, "step": 297800 }, { "epoch": 64.21, "learning_rate": 1.026245855499571e-05, "loss": 0.0, "step": 297850 }, { "epoch": 64.22, "learning_rate": 1.0257104312648378e-05, "loss": 0.0, "step": 297900 }, { "epoch": 64.23, "learning_rate": 1.0251750741651922e-05, "loss": 0.0, "step": 297950 }, { "epoch": 64.24, "learning_rate": 1.024639784276413e-05, "loss": 0.0, "step": 298000 }, { "epoch": 64.25, "learning_rate": 1.0241045616742697e-05, "loss": 0.0, "step": 298050 }, { "epoch": 64.26, "learning_rate": 1.0235694064345222e-05, "loss": 0.0, "step": 298100 }, { "epoch": 64.27, "learning_rate": 1.0230343186329219e-05, "loss": 0.0, "step": 298150 }, { "epoch": 64.28, "learning_rate": 1.022499298345209e-05, "loss": 0.0, "step": 298200 }, { "epoch": 64.29, "learning_rate": 1.0219643456471157e-05, "loss": 0.0, "step": 298250 }, { "epoch": 64.3, "learning_rate": 1.0214294606143623e-05, "loss": 0.0, "step": 298300 }, { "epoch": 64.31, "learning_rate": 1.0208946433226631e-05, "loss": 0.0, "step": 298350 }, { "epoch": 64.32, "learning_rate": 1.0203598938477199e-05, "loss": 0.0, "step": 298400 }, { "epoch": 64.33, "learning_rate": 1.0198252122652255e-05, "loss": 0.0, "step": 298450 }, { "epoch": 64.35, "learning_rate": 1.0192905986508635e-05, "loss": 0.0, "step": 298500 }, { "epoch": 64.36, "learning_rate": 1.0187560530803084e-05, "loss": 0.0, "step": 298550 }, { "epoch": 64.37, "learning_rate": 1.0182215756292242e-05, "loss": 0.0, "step": 298600 }, { "epoch": 64.38, "learning_rate": 1.0176871663732652e-05, "loss": 0.0, "step": 298650 }, { "epoch": 64.39, "learning_rate": 1.017152825388076e-05, "loss": 0.0, "step": 298700 }, { "epoch": 64.4, "learning_rate": 1.0166185527492929e-05, "loss": 0.0, "step": 298750 }, { "epoch": 64.41, "learning_rate": 1.0160843485325408e-05, "loss": 0.0, "step": 298800 }, { "epoch": 64.42, "learning_rate": 1.0155502128134364e-05, "loss": 0.0, "step": 298850 }, { "epoch": 64.43, "learning_rate": 1.0150161456675847e-05, "loss": 0.0, "step": 298900 }, { "epoch": 64.44, "learning_rate": 1.0144821471705827e-05, "loss": 0.0, "step": 298950 }, { "epoch": 64.45, "learning_rate": 1.0139482173980179e-05, "loss": 0.0, "step": 299000 }, { "epoch": 64.46, "learning_rate": 1.0134143564254668e-05, "loss": 0.0, "step": 299050 }, { "epoch": 64.48, "learning_rate": 1.0128805643284971e-05, "loss": 0.0, "step": 299100 }, { "epoch": 64.49, "learning_rate": 1.0123468411826653e-05, "loss": 0.0, "step": 299150 }, { "epoch": 64.5, "learning_rate": 1.0118131870635209e-05, "loss": 0.0, "step": 299200 }, { "epoch": 64.51, "learning_rate": 1.0112796020466008e-05, "loss": 0.0, "step": 299250 }, { "epoch": 64.52, "learning_rate": 1.0107460862074336e-05, "loss": 0.0, "step": 299300 }, { "epoch": 64.53, "learning_rate": 1.0102126396215374e-05, "loss": 0.0, "step": 299350 }, { "epoch": 64.54, "learning_rate": 1.0096792623644219e-05, "loss": 0.0, "step": 299400 }, { "epoch": 64.55, "learning_rate": 1.0091459545115855e-05, "loss": 0.0, "step": 299450 }, { "epoch": 64.56, "learning_rate": 1.0086127161385168e-05, "loss": 0.0, "step": 299500 }, { "epoch": 64.57, "learning_rate": 1.0080795473206951e-05, "loss": 0.0, "step": 299550 }, { "epoch": 64.58, "learning_rate": 1.00754644813359e-05, "loss": 0.0, "step": 299600 }, { "epoch": 64.59, "learning_rate": 1.0070134186526614e-05, "loss": 0.0, "step": 299650 }, { "epoch": 64.6, "learning_rate": 1.0064804589533586e-05, "loss": 0.0, "step": 299700 }, { "epoch": 64.62, "learning_rate": 1.0059475691111213e-05, "loss": 0.0, "step": 299750 }, { "epoch": 64.63, "learning_rate": 1.0054147492013788e-05, "loss": 0.0, "step": 299800 }, { "epoch": 64.64, "learning_rate": 1.0048819992995521e-05, "loss": 0.0, "step": 299850 }, { "epoch": 64.65, "learning_rate": 1.0043493194810509e-05, "loss": 0.0, "step": 299900 }, { "epoch": 64.66, "learning_rate": 1.0038167098212757e-05, "loss": 0.0, "step": 299950 }, { "epoch": 64.67, "learning_rate": 1.0032841703956155e-05, "loss": 0.0, "step": 300000 }, { "epoch": 64.68, "learning_rate": 1.002751701279452e-05, "loss": 0.0, "step": 300050 }, { "epoch": 64.69, "learning_rate": 1.002219302548155e-05, "loss": 0.0, "step": 300100 }, { "epoch": 64.7, "learning_rate": 1.0016869742770846e-05, "loss": 0.0, "step": 300150 }, { "epoch": 64.71, "learning_rate": 1.0011547165415913e-05, "loss": 0.0, "step": 300200 }, { "epoch": 64.72, "learning_rate": 1.0006225294170158e-05, "loss": 0.0, "step": 300250 }, { "epoch": 64.73, "learning_rate": 1.0000904129786886e-05, "loss": 0.0, "step": 300300 }, { "epoch": 64.74, "learning_rate": 9.995583673019295e-06, "loss": 0.0, "step": 300350 }, { "epoch": 64.76, "learning_rate": 9.990263924620491e-06, "loss": 0.0, "step": 300400 }, { "epoch": 64.77, "learning_rate": 9.98494488534348e-06, "loss": 0.0, "step": 300450 }, { "epoch": 64.78, "learning_rate": 9.979626555941165e-06, "loss": 0.0, "step": 300500 }, { "epoch": 64.79, "learning_rate": 9.974308937166354e-06, "loss": 0.0, "step": 300550 }, { "epoch": 64.8, "learning_rate": 9.968992029771736e-06, "loss": 0.0, "step": 300600 }, { "epoch": 64.81, "learning_rate": 9.963675834509918e-06, "loss": 0.0, "step": 300650 }, { "epoch": 64.82, "learning_rate": 9.958360352133406e-06, "loss": 0.0, "step": 300700 }, { "epoch": 64.83, "learning_rate": 9.953045583394595e-06, "loss": 0.0, "step": 300750 }, { "epoch": 64.84, "learning_rate": 9.947731529045789e-06, "loss": 0.0, "step": 300800 }, { "epoch": 64.85, "learning_rate": 9.942418189839171e-06, "loss": 0.0, "step": 300850 }, { "epoch": 64.86, "learning_rate": 9.93710556652686e-06, "loss": 0.0, "step": 300900 }, { "epoch": 64.87, "learning_rate": 9.931793659860834e-06, "loss": 0.0, "step": 300950 }, { "epoch": 64.88, "learning_rate": 9.926482470592994e-06, "loss": 0.0, "step": 301000 }, { "epoch": 64.9, "learning_rate": 9.921171999475125e-06, "loss": 0.0, "step": 301050 }, { "epoch": 64.91, "learning_rate": 9.915862247258925e-06, "loss": 0.0, "step": 301100 }, { "epoch": 64.92, "learning_rate": 9.910553214695984e-06, "loss": 0.0, "step": 301150 }, { "epoch": 64.93, "learning_rate": 9.905244902537781e-06, "loss": 0.0, "step": 301200 }, { "epoch": 64.94, "learning_rate": 9.899937311535702e-06, "loss": 0.0, "step": 301250 }, { "epoch": 64.95, "learning_rate": 9.894630442441038e-06, "loss": 0.0, "step": 301300 }, { "epoch": 64.96, "learning_rate": 9.889324296004961e-06, "loss": 0.0, "step": 301350 }, { "epoch": 64.97, "learning_rate": 9.884018872978557e-06, "loss": 0.0, "step": 301400 }, { "epoch": 64.98, "learning_rate": 9.878714174112797e-06, "loss": 0.0, "step": 301450 }, { "epoch": 64.99, "learning_rate": 9.873410200158546e-06, "loss": 0.0, "step": 301500 }, { "epoch": 65.0, "learning_rate": 9.868106951866591e-06, "loss": 0.0, "step": 301550 }, { "epoch": 65.01, "learning_rate": 9.862804429987591e-06, "loss": 0.0, "step": 301600 }, { "epoch": 65.02, "learning_rate": 9.857502635272114e-06, "loss": 0.0, "step": 301650 }, { "epoch": 65.04, "learning_rate": 9.852201568470616e-06, "loss": 0.0, "step": 301700 }, { "epoch": 65.05, "learning_rate": 9.846901230333468e-06, "loss": 0.0, "step": 301750 }, { "epoch": 65.06, "learning_rate": 9.841601621610919e-06, "loss": 0.0, "step": 301800 }, { "epoch": 65.07, "learning_rate": 9.836302743053119e-06, "loss": 0.0, "step": 301850 }, { "epoch": 65.08, "learning_rate": 9.831004595410118e-06, "loss": 0.0, "step": 301900 }, { "epoch": 65.09, "learning_rate": 9.82570717943187e-06, "loss": 0.0, "step": 301950 }, { "epoch": 65.1, "learning_rate": 9.820410495868218e-06, "loss": 0.0, "step": 302000 }, { "epoch": 65.11, "learning_rate": 9.815114545468888e-06, "loss": 0.0, "step": 302050 }, { "epoch": 65.12, "learning_rate": 9.809819328983523e-06, "loss": 0.0, "step": 302100 }, { "epoch": 65.13, "learning_rate": 9.804524847161654e-06, "loss": 0.0, "step": 302150 }, { "epoch": 65.14, "learning_rate": 9.799231100752711e-06, "loss": 0.0, "step": 302200 }, { "epoch": 65.15, "learning_rate": 9.793938090506016e-06, "loss": 0.0, "step": 302250 }, { "epoch": 65.16, "learning_rate": 9.788645817170783e-06, "loss": 0.0, "step": 302300 }, { "epoch": 65.18, "learning_rate": 9.783354281496126e-06, "loss": 0.0, "step": 302350 }, { "epoch": 65.19, "learning_rate": 9.778063484231061e-06, "loss": 0.0, "step": 302400 }, { "epoch": 65.2, "learning_rate": 9.772773426124492e-06, "loss": 0.0, "step": 302450 }, { "epoch": 65.21, "learning_rate": 9.767484107925217e-06, "loss": 0.0, "step": 302500 }, { "epoch": 65.22, "learning_rate": 9.762195530381937e-06, "loss": 0.0, "step": 302550 }, { "epoch": 65.23, "learning_rate": 9.756907694243244e-06, "loss": 0.0, "step": 302600 }, { "epoch": 65.24, "learning_rate": 9.751620600257618e-06, "loss": 0.0, "step": 302650 }, { "epoch": 65.25, "learning_rate": 9.746334249173445e-06, "loss": 0.0, "step": 302700 }, { "epoch": 65.26, "learning_rate": 9.741048641738997e-06, "loss": 0.0, "step": 302750 }, { "epoch": 65.27, "learning_rate": 9.735763778702452e-06, "loss": 0.0, "step": 302800 }, { "epoch": 65.28, "learning_rate": 9.730479660811873e-06, "loss": 0.0, "step": 302850 }, { "epoch": 65.29, "learning_rate": 9.725196288815218e-06, "loss": 0.0, "step": 302900 }, { "epoch": 65.31, "learning_rate": 9.719913663460344e-06, "loss": 0.0, "step": 302950 }, { "epoch": 65.32, "learning_rate": 9.714631785494992e-06, "loss": 0.0, "step": 303000 }, { "epoch": 65.33, "learning_rate": 9.709350655666819e-06, "loss": 0.0, "step": 303050 }, { "epoch": 65.34, "learning_rate": 9.704070274723353e-06, "loss": 0.0, "step": 303100 }, { "epoch": 65.35, "learning_rate": 9.698790643412031e-06, "loss": 0.0, "step": 303150 }, { "epoch": 65.36, "learning_rate": 9.693511762480168e-06, "loss": 0.0, "step": 303200 }, { "epoch": 65.37, "learning_rate": 9.688233632674992e-06, "loss": 0.0, "step": 303250 }, { "epoch": 65.38, "learning_rate": 9.682956254743614e-06, "loss": 0.0, "step": 303300 }, { "epoch": 65.39, "learning_rate": 9.677679629433039e-06, "loss": 0.0, "step": 303350 }, { "epoch": 65.4, "learning_rate": 9.672403757490163e-06, "loss": 0.0, "step": 303400 }, { "epoch": 65.41, "learning_rate": 9.66712863966179e-06, "loss": 0.0, "step": 303450 }, { "epoch": 65.42, "learning_rate": 9.661854276694594e-06, "loss": 0.0, "step": 303500 }, { "epoch": 65.43, "learning_rate": 9.656580669335161e-06, "loss": 0.0, "step": 303550 }, { "epoch": 65.45, "learning_rate": 9.651307818329958e-06, "loss": 0.0, "step": 303600 }, { "epoch": 65.46, "learning_rate": 9.646035724425358e-06, "loss": 0.0, "step": 303650 }, { "epoch": 65.47, "learning_rate": 9.64076438836762e-06, "loss": 0.0, "step": 303700 }, { "epoch": 65.48, "learning_rate": 9.635493810902883e-06, "loss": 0.0, "step": 303750 }, { "epoch": 65.49, "learning_rate": 9.630223992777201e-06, "loss": 0.0, "step": 303800 }, { "epoch": 65.5, "learning_rate": 9.624954934736503e-06, "loss": 0.0, "step": 303850 }, { "epoch": 65.51, "learning_rate": 9.619686637526626e-06, "loss": 0.0, "step": 303900 }, { "epoch": 65.52, "learning_rate": 9.614419101893284e-06, "loss": 0.0, "step": 303950 }, { "epoch": 65.53, "learning_rate": 9.609152328582096e-06, "loss": 0.0, "step": 304000 }, { "epoch": 65.54, "learning_rate": 9.603886318338558e-06, "loss": 0.0, "step": 304050 }, { "epoch": 65.55, "learning_rate": 9.598621071908075e-06, "loss": 0.0, "step": 304100 }, { "epoch": 65.56, "learning_rate": 9.593356590035933e-06, "loss": 0.0, "step": 304150 }, { "epoch": 65.57, "learning_rate": 9.588092873467313e-06, "loss": 0.0, "step": 304200 }, { "epoch": 65.59, "learning_rate": 9.582829922947286e-06, "loss": 0.0, "step": 304250 }, { "epoch": 65.6, "learning_rate": 9.577567739220823e-06, "loss": 0.0, "step": 304300 }, { "epoch": 65.61, "learning_rate": 9.57230632303277e-06, "loss": 0.0, "step": 304350 }, { "epoch": 65.62, "learning_rate": 9.567045675127878e-06, "loss": 0.0, "step": 304400 }, { "epoch": 65.63, "learning_rate": 9.561785796250782e-06, "loss": 0.0, "step": 304450 }, { "epoch": 65.64, "learning_rate": 9.556526687146017e-06, "loss": 0.0, "step": 304500 }, { "epoch": 65.65, "learning_rate": 9.551268348557999e-06, "loss": 0.0, "step": 304550 }, { "epoch": 65.66, "learning_rate": 9.546010781231046e-06, "loss": 0.0, "step": 304600 }, { "epoch": 65.67, "learning_rate": 9.540753985909349e-06, "loss": 0.0, "step": 304650 }, { "epoch": 65.68, "learning_rate": 9.535497963337004e-06, "loss": 0.0, "step": 304700 }, { "epoch": 65.69, "learning_rate": 9.530242714257999e-06, "loss": 0.0, "step": 304750 }, { "epoch": 65.7, "learning_rate": 9.524988239416204e-06, "loss": 0.0, "step": 304800 }, { "epoch": 65.71, "learning_rate": 9.519734539555389e-06, "loss": 0.0, "step": 304850 }, { "epoch": 65.73, "learning_rate": 9.5144816154192e-06, "loss": 0.0, "step": 304900 }, { "epoch": 65.74, "learning_rate": 9.509229467751187e-06, "loss": 0.0, "step": 304950 }, { "epoch": 65.75, "learning_rate": 9.503978097294782e-06, "loss": 0.0, "step": 305000 }, { "epoch": 65.76, "learning_rate": 9.498727504793313e-06, "loss": 0.0, "step": 305050 }, { "epoch": 65.77, "learning_rate": 9.49347769098999e-06, "loss": 0.0, "step": 305100 }, { "epoch": 65.78, "learning_rate": 9.488228656627923e-06, "loss": 0.0, "step": 305150 }, { "epoch": 65.79, "learning_rate": 9.48298040245011e-06, "loss": 0.0, "step": 305200 }, { "epoch": 65.8, "learning_rate": 9.47773292919942e-06, "loss": 0.0, "step": 305250 }, { "epoch": 65.81, "learning_rate": 9.472486237618635e-06, "loss": 0.0, "step": 305300 }, { "epoch": 65.82, "learning_rate": 9.467240328450421e-06, "loss": 0.0, "step": 305350 }, { "epoch": 65.83, "learning_rate": 9.461995202437325e-06, "loss": 0.0, "step": 305400 }, { "epoch": 65.84, "learning_rate": 9.456750860321794e-06, "loss": 0.0, "step": 305450 }, { "epoch": 65.85, "learning_rate": 9.451507302846146e-06, "loss": 0.0, "step": 305500 }, { "epoch": 65.87, "learning_rate": 9.446264530752606e-06, "loss": 0.0, "step": 305550 }, { "epoch": 65.88, "learning_rate": 9.441022544783287e-06, "loss": 0.0, "step": 305600 }, { "epoch": 65.89, "learning_rate": 9.43578134568018e-06, "loss": 0.0, "step": 305650 }, { "epoch": 65.9, "learning_rate": 9.430540934185175e-06, "loss": 0.0, "step": 305700 }, { "epoch": 65.91, "learning_rate": 9.425301311040031e-06, "loss": 0.0, "step": 305750 }, { "epoch": 65.92, "learning_rate": 9.42006247698643e-06, "loss": 0.0, "step": 305800 }, { "epoch": 65.93, "learning_rate": 9.414824432765911e-06, "loss": 0.0, "step": 305850 }, { "epoch": 65.94, "learning_rate": 9.409587179119916e-06, "loss": 0.0, "step": 305900 }, { "epoch": 65.95, "learning_rate": 9.404350716789767e-06, "loss": 0.0, "step": 305950 }, { "epoch": 65.96, "learning_rate": 9.399115046516683e-06, "loss": 0.0, "step": 306000 }, { "epoch": 65.97, "learning_rate": 9.393880169041771e-06, "loss": 0.0, "step": 306050 }, { "epoch": 65.98, "learning_rate": 9.38864608510601e-06, "loss": 0.0, "step": 306100 }, { "epoch": 65.99, "learning_rate": 9.38341279545028e-06, "loss": 0.0, "step": 306150 }, { "epoch": 66.01, "learning_rate": 9.378180300815359e-06, "loss": 0.0, "step": 306200 }, { "epoch": 66.02, "learning_rate": 9.372948601941885e-06, "loss": 0.0, "step": 306250 }, { "epoch": 66.03, "learning_rate": 9.36771769957041e-06, "loss": 0.0, "step": 306300 }, { "epoch": 66.04, "learning_rate": 9.362487594441351e-06, "loss": 0.0, "step": 306350 }, { "epoch": 66.05, "learning_rate": 9.357258287295027e-06, "loss": 0.0, "step": 306400 }, { "epoch": 66.06, "learning_rate": 9.352029778871644e-06, "loss": 0.0, "step": 306450 }, { "epoch": 66.07, "learning_rate": 9.346802069911286e-06, "loss": 0.0, "step": 306500 }, { "epoch": 66.08, "learning_rate": 9.341575161153934e-06, "loss": 0.0, "step": 306550 }, { "epoch": 66.09, "learning_rate": 9.336349053339436e-06, "loss": 0.0, "step": 306600 }, { "epoch": 66.1, "learning_rate": 9.33112374720756e-06, "loss": 0.0, "step": 306650 }, { "epoch": 66.11, "learning_rate": 9.32589924349793e-06, "loss": 0.0, "step": 306700 }, { "epoch": 66.12, "learning_rate": 9.320675542950073e-06, "loss": 0.0, "step": 306750 }, { "epoch": 66.13, "learning_rate": 9.315452646303389e-06, "loss": 0.0, "step": 306800 }, { "epoch": 66.15, "learning_rate": 9.310230554297181e-06, "loss": 0.0, "step": 306850 }, { "epoch": 66.16, "learning_rate": 9.305009267670632e-06, "loss": 0.0, "step": 306900 }, { "epoch": 66.17, "learning_rate": 9.2997887871628e-06, "loss": 0.0, "step": 306950 }, { "epoch": 66.18, "learning_rate": 9.294569113512635e-06, "loss": 0.0, "step": 307000 }, { "epoch": 66.19, "learning_rate": 9.289350247458984e-06, "loss": 0.0, "step": 307050 }, { "epoch": 66.2, "learning_rate": 9.284132189740568e-06, "loss": 0.0, "step": 307100 }, { "epoch": 66.21, "learning_rate": 9.278914941095999e-06, "loss": 0.0, "step": 307150 }, { "epoch": 66.22, "learning_rate": 9.273698502263768e-06, "loss": 0.0, "step": 307200 }, { "epoch": 66.23, "learning_rate": 9.268482873982251e-06, "loss": 0.0, "step": 307250 }, { "epoch": 66.24, "learning_rate": 9.26326805698972e-06, "loss": 0.0, "step": 307300 }, { "epoch": 66.25, "learning_rate": 9.25805405202433e-06, "loss": 0.0, "step": 307350 }, { "epoch": 66.26, "learning_rate": 9.25284085982411e-06, "loss": 0.0, "step": 307400 }, { "epoch": 66.28, "learning_rate": 9.247628481126975e-06, "loss": 0.0, "step": 307450 }, { "epoch": 66.29, "learning_rate": 9.242416916670747e-06, "loss": 0.0, "step": 307500 }, { "epoch": 66.3, "learning_rate": 9.237206167193103e-06, "loss": 0.0, "step": 307550 }, { "epoch": 66.31, "learning_rate": 9.231996233431623e-06, "loss": 0.0, "step": 307600 }, { "epoch": 66.32, "learning_rate": 9.226787116123762e-06, "loss": 0.0, "step": 307650 }, { "epoch": 66.33, "learning_rate": 9.221578816006872e-06, "loss": 0.0, "step": 307700 }, { "epoch": 66.34, "learning_rate": 9.216371333818178e-06, "loss": 0.0, "step": 307750 }, { "epoch": 66.35, "learning_rate": 9.211164670294789e-06, "loss": 0.0, "step": 307800 }, { "epoch": 66.36, "learning_rate": 9.205958826173704e-06, "loss": 0.0, "step": 307850 }, { "epoch": 66.37, "learning_rate": 9.200753802191799e-06, "loss": 0.0, "step": 307900 }, { "epoch": 66.38, "learning_rate": 9.195549599085849e-06, "loss": 0.0, "step": 307950 }, { "epoch": 66.39, "learning_rate": 9.190346217592496e-06, "loss": 0.0, "step": 308000 }, { "epoch": 66.4, "learning_rate": 9.185143658448272e-06, "loss": 0.0, "step": 308050 }, { "epoch": 66.42, "learning_rate": 9.179941922389591e-06, "loss": 0.0, "step": 308100 }, { "epoch": 66.43, "learning_rate": 9.174741010152755e-06, "loss": 0.0, "step": 308150 }, { "epoch": 66.44, "learning_rate": 9.169540922473947e-06, "loss": 0.0, "step": 308200 }, { "epoch": 66.45, "learning_rate": 9.164341660089234e-06, "loss": 0.0, "step": 308250 }, { "epoch": 66.46, "learning_rate": 9.159143223734555e-06, "loss": 0.0, "step": 308300 }, { "epoch": 66.47, "learning_rate": 9.153945614145759e-06, "loss": 0.0, "step": 308350 }, { "epoch": 66.48, "learning_rate": 9.148748832058547e-06, "loss": 0.0, "step": 308400 }, { "epoch": 66.49, "learning_rate": 9.143552878208522e-06, "loss": 0.0, "step": 308450 }, { "epoch": 66.5, "learning_rate": 9.138357753331162e-06, "loss": 0.0, "step": 308500 }, { "epoch": 66.51, "learning_rate": 9.133163458161837e-06, "loss": 0.0, "step": 308550 }, { "epoch": 66.52, "learning_rate": 9.12796999343579e-06, "loss": 0.0, "step": 308600 }, { "epoch": 66.53, "learning_rate": 9.122777359888145e-06, "loss": 0.0, "step": 308650 }, { "epoch": 66.54, "learning_rate": 9.117585558253918e-06, "loss": 0.0, "step": 308700 }, { "epoch": 66.56, "learning_rate": 9.112394589267994e-06, "loss": 0.0, "step": 308750 }, { "epoch": 66.57, "learning_rate": 9.107204453665157e-06, "loss": 0.0, "step": 308800 }, { "epoch": 66.58, "learning_rate": 9.102015152180065e-06, "loss": 0.0, "step": 308850 }, { "epoch": 66.59, "learning_rate": 9.096826685547248e-06, "loss": 0.0, "step": 308900 }, { "epoch": 66.6, "learning_rate": 9.091639054501132e-06, "loss": 0.0, "step": 308950 }, { "epoch": 66.61, "learning_rate": 9.08645225977602e-06, "loss": 0.0, "step": 309000 }, { "epoch": 66.62, "learning_rate": 9.081266302106096e-06, "loss": 0.0, "step": 309050 }, { "epoch": 66.63, "learning_rate": 9.07608118222543e-06, "loss": 0.0, "step": 309100 }, { "epoch": 66.64, "learning_rate": 9.070896900867951e-06, "loss": 0.0, "step": 309150 }, { "epoch": 66.65, "learning_rate": 9.065713458767512e-06, "loss": 0.0, "step": 309200 }, { "epoch": 66.66, "learning_rate": 9.060530856657811e-06, "loss": 0.0, "step": 309250 }, { "epoch": 66.67, "learning_rate": 9.055349095272438e-06, "loss": 0.0, "step": 309300 }, { "epoch": 66.68, "learning_rate": 9.050168175344862e-06, "loss": 0.0, "step": 309350 }, { "epoch": 66.7, "learning_rate": 9.044988097608444e-06, "loss": 0.0, "step": 309400 }, { "epoch": 66.71, "learning_rate": 9.039808862796416e-06, "loss": 0.0, "step": 309450 }, { "epoch": 66.72, "learning_rate": 9.034630471641886e-06, "loss": 0.0, "step": 309500 }, { "epoch": 66.73, "learning_rate": 9.029452924877851e-06, "loss": 0.0, "step": 309550 }, { "epoch": 66.74, "learning_rate": 9.024276223237185e-06, "loss": 0.0, "step": 309600 }, { "epoch": 66.75, "learning_rate": 9.019100367452652e-06, "loss": 0.0, "step": 309650 }, { "epoch": 66.76, "learning_rate": 9.01392535825688e-06, "loss": 0.0, "step": 309700 }, { "epoch": 66.77, "learning_rate": 9.008751196382385e-06, "loss": 0.0, "step": 309750 }, { "epoch": 66.78, "learning_rate": 9.003577882561563e-06, "loss": 0.0, "step": 309800 }, { "epoch": 66.79, "learning_rate": 8.998405417526694e-06, "loss": 0.0, "step": 309850 }, { "epoch": 66.8, "learning_rate": 8.99323380200993e-06, "loss": 0.0, "step": 309900 }, { "epoch": 66.81, "learning_rate": 8.988063036743312e-06, "loss": 0.0, "step": 309950 }, { "epoch": 66.82, "learning_rate": 8.982893122458742e-06, "loss": 0.0, "step": 310000 }, { "epoch": 66.84, "learning_rate": 8.977724059888034e-06, "loss": 0.0, "step": 310050 }, { "epoch": 66.85, "learning_rate": 8.97255584976285e-06, "loss": 0.0, "step": 310100 }, { "epoch": 66.86, "learning_rate": 8.967388492814745e-06, "loss": 0.0, "step": 310150 }, { "epoch": 66.87, "learning_rate": 8.96222198977515e-06, "loss": 0.0, "step": 310200 }, { "epoch": 66.88, "learning_rate": 8.957056341375383e-06, "loss": 0.0, "step": 310250 }, { "epoch": 66.89, "learning_rate": 8.951891548346636e-06, "loss": 0.0, "step": 310300 }, { "epoch": 66.9, "learning_rate": 8.94672761141997e-06, "loss": 0.0, "step": 310350 }, { "epoch": 66.91, "learning_rate": 8.941564531326338e-06, "loss": 0.0, "step": 310400 }, { "epoch": 66.92, "learning_rate": 8.93640230879657e-06, "loss": 0.0, "step": 310450 }, { "epoch": 66.93, "learning_rate": 8.931240944561366e-06, "loss": 0.0, "step": 310500 }, { "epoch": 66.94, "learning_rate": 8.926080439351322e-06, "loss": 0.0, "step": 310550 }, { "epoch": 66.95, "learning_rate": 8.920920793896888e-06, "loss": 0.0, "step": 310600 }, { "epoch": 66.96, "learning_rate": 8.91576200892841e-06, "loss": 0.0, "step": 310650 }, { "epoch": 66.98, "learning_rate": 8.91060408517611e-06, "loss": 0.0, "step": 310700 }, { "epoch": 66.99, "learning_rate": 8.905447023370083e-06, "loss": 0.0, "step": 310750 }, { "epoch": 67.0, "learning_rate": 8.90029082424031e-06, "loss": 0.0, "step": 310800 }, { "epoch": 67.01, "learning_rate": 8.89513548851663e-06, "loss": 0.0, "step": 310850 }, { "epoch": 67.02, "learning_rate": 8.889981016928791e-06, "loss": 0.0, "step": 310900 }, { "epoch": 67.03, "learning_rate": 8.884827410206397e-06, "loss": 0.0, "step": 310950 }, { "epoch": 67.04, "learning_rate": 8.87967466907893e-06, "loss": 0.0, "step": 311000 }, { "epoch": 67.05, "learning_rate": 8.874522794275752e-06, "loss": 0.0, "step": 311050 }, { "epoch": 67.06, "learning_rate": 8.869371786526113e-06, "loss": 0.0, "step": 311100 }, { "epoch": 67.07, "learning_rate": 8.86422164655913e-06, "loss": 0.0, "step": 311150 }, { "epoch": 67.08, "learning_rate": 8.859072375103794e-06, "loss": 0.0, "step": 311200 }, { "epoch": 67.09, "learning_rate": 8.853923972888982e-06, "loss": 0.0, "step": 311250 }, { "epoch": 67.1, "learning_rate": 8.848776440643434e-06, "loss": 0.0, "step": 311300 }, { "epoch": 67.12, "learning_rate": 8.843629779095793e-06, "loss": 0.0, "step": 311350 }, { "epoch": 67.13, "learning_rate": 8.838483988974553e-06, "loss": 0.0, "step": 311400 }, { "epoch": 67.14, "learning_rate": 8.833339071008092e-06, "loss": 0.0, "step": 311450 }, { "epoch": 67.15, "learning_rate": 8.828195025924669e-06, "loss": 0.0, "step": 311500 }, { "epoch": 67.16, "learning_rate": 8.82305185445242e-06, "loss": 0.0, "step": 311550 }, { "epoch": 67.17, "learning_rate": 8.81790955731935e-06, "loss": 0.0, "step": 311600 }, { "epoch": 67.18, "learning_rate": 8.812768135253352e-06, "loss": 0.0, "step": 311650 }, { "epoch": 67.19, "learning_rate": 8.807627588982172e-06, "loss": 0.0, "step": 311700 }, { "epoch": 67.2, "learning_rate": 8.80248791923347e-06, "loss": 0.0, "step": 311750 }, { "epoch": 67.21, "learning_rate": 8.797349126734744e-06, "loss": 0.0, "step": 311800 }, { "epoch": 67.22, "learning_rate": 8.792211212213387e-06, "loss": 0.0, "step": 311850 }, { "epoch": 67.23, "learning_rate": 8.787074176396661e-06, "loss": 0.0, "step": 311900 }, { "epoch": 67.25, "learning_rate": 8.781938020011716e-06, "loss": 0.0, "step": 311950 }, { "epoch": 67.26, "learning_rate": 8.776802743785563e-06, "loss": 0.0, "step": 312000 }, { "epoch": 67.27, "learning_rate": 8.771668348445094e-06, "loss": 0.0, "step": 312050 }, { "epoch": 67.28, "learning_rate": 8.766534834717076e-06, "loss": 0.0, "step": 312100 }, { "epoch": 67.29, "learning_rate": 8.76140220332815e-06, "loss": 0.0, "step": 312150 }, { "epoch": 67.3, "learning_rate": 8.756270455004834e-06, "loss": 0.0, "step": 312200 }, { "epoch": 67.31, "learning_rate": 8.751139590473525e-06, "loss": 0.0, "step": 312250 }, { "epoch": 67.32, "learning_rate": 8.746009610460487e-06, "loss": 0.0, "step": 312300 }, { "epoch": 67.33, "learning_rate": 8.74088051569186e-06, "loss": 0.0, "step": 312350 }, { "epoch": 67.34, "learning_rate": 8.735752306893663e-06, "loss": 0.0, "step": 312400 }, { "epoch": 67.35, "learning_rate": 8.730624984791789e-06, "loss": 0.0, "step": 312450 }, { "epoch": 67.36, "learning_rate": 8.725498550112003e-06, "loss": 0.0, "step": 312500 }, { "epoch": 67.37, "learning_rate": 8.720373003579942e-06, "loss": 0.0, "step": 312550 }, { "epoch": 67.39, "learning_rate": 8.715248345921124e-06, "loss": 0.0, "step": 312600 }, { "epoch": 67.4, "learning_rate": 8.710124577860939e-06, "loss": 0.0, "step": 312650 }, { "epoch": 67.41, "learning_rate": 8.705001700124647e-06, "loss": 0.0, "step": 312700 }, { "epoch": 67.42, "learning_rate": 8.699879713437385e-06, "loss": 0.0, "step": 312750 }, { "epoch": 67.43, "learning_rate": 8.694758618524165e-06, "loss": 0.0, "step": 312800 }, { "epoch": 67.44, "learning_rate": 8.689638416109873e-06, "loss": 0.0, "step": 312850 }, { "epoch": 67.45, "learning_rate": 8.684519106919263e-06, "loss": 0.0, "step": 312900 }, { "epoch": 67.46, "learning_rate": 8.67940069167697e-06, "loss": 0.0, "step": 312950 }, { "epoch": 67.47, "learning_rate": 8.674283171107497e-06, "loss": 0.0, "step": 313000 }, { "epoch": 67.48, "learning_rate": 8.669166545935224e-06, "loss": 0.0, "step": 313050 }, { "epoch": 67.49, "learning_rate": 8.664050816884402e-06, "loss": 0.0, "step": 313100 }, { "epoch": 67.5, "learning_rate": 8.658935984679159e-06, "loss": 0.0, "step": 313150 }, { "epoch": 67.51, "learning_rate": 8.653822050043483e-06, "loss": 0.0, "step": 313200 }, { "epoch": 67.53, "learning_rate": 8.648709013701256e-06, "loss": 0.0, "step": 313250 }, { "epoch": 67.54, "learning_rate": 8.643596876376229e-06, "loss": 0.0, "step": 313300 }, { "epoch": 67.55, "learning_rate": 8.638485638791999e-06, "loss": 0.0, "step": 313350 }, { "epoch": 67.56, "learning_rate": 8.633375301672059e-06, "loss": 0.0, "step": 313400 }, { "epoch": 67.57, "learning_rate": 8.62826586573978e-06, "loss": 0.0, "step": 313450 }, { "epoch": 67.58, "learning_rate": 8.623157331718393e-06, "loss": 0.0, "step": 313500 }, { "epoch": 67.59, "learning_rate": 8.61804970033101e-06, "loss": 0.0, "step": 313550 }, { "epoch": 67.6, "learning_rate": 8.612942972300594e-06, "loss": 0.0, "step": 313600 }, { "epoch": 67.61, "learning_rate": 8.607837148350003e-06, "loss": 0.0, "step": 313650 }, { "epoch": 67.62, "learning_rate": 8.602732229201968e-06, "loss": 0.0, "step": 313700 }, { "epoch": 67.63, "learning_rate": 8.597628215579076e-06, "loss": 0.0, "step": 313750 }, { "epoch": 67.64, "learning_rate": 8.592525108203803e-06, "loss": 0.0, "step": 313800 }, { "epoch": 67.65, "learning_rate": 8.587422907798466e-06, "loss": 0.0, "step": 313850 }, { "epoch": 67.67, "learning_rate": 8.582321615085296e-06, "loss": 0.0, "step": 313900 }, { "epoch": 67.68, "learning_rate": 8.577221230786367e-06, "loss": 0.0, "step": 313950 }, { "epoch": 67.69, "learning_rate": 8.572121755623631e-06, "loss": 0.0, "step": 314000 }, { "epoch": 67.7, "learning_rate": 8.567023190318917e-06, "loss": 0.0, "step": 314050 }, { "epoch": 67.71, "learning_rate": 8.561925535593915e-06, "loss": 0.0, "step": 314100 }, { "epoch": 67.72, "learning_rate": 8.556828792170194e-06, "loss": 0.0, "step": 314150 }, { "epoch": 67.73, "learning_rate": 8.55173296076919e-06, "loss": 0.0, "step": 314200 }, { "epoch": 67.74, "learning_rate": 8.546638042112214e-06, "loss": 0.0, "step": 314250 }, { "epoch": 67.75, "learning_rate": 8.541544036920443e-06, "loss": 0.0, "step": 314300 }, { "epoch": 67.76, "learning_rate": 8.536450945914928e-06, "loss": 0.0, "step": 314350 }, { "epoch": 67.77, "learning_rate": 8.531358769816591e-06, "loss": 0.0, "step": 314400 }, { "epoch": 67.78, "learning_rate": 8.526267509346222e-06, "loss": 0.0, "step": 314450 }, { "epoch": 67.79, "learning_rate": 8.521177165224479e-06, "loss": 0.0, "step": 314500 }, { "epoch": 67.81, "learning_rate": 8.5160877381719e-06, "loss": 0.0, "step": 314550 }, { "epoch": 67.82, "learning_rate": 8.510999228908881e-06, "loss": 0.0, "step": 314600 }, { "epoch": 67.83, "learning_rate": 8.505911638155698e-06, "loss": 0.0, "step": 314650 }, { "epoch": 67.84, "learning_rate": 8.500824966632484e-06, "loss": 0.0, "step": 314700 }, { "epoch": 67.85, "learning_rate": 8.495739215059272e-06, "loss": 0.0, "step": 314750 }, { "epoch": 67.86, "learning_rate": 8.490654384155926e-06, "loss": 0.0, "step": 314800 }, { "epoch": 67.87, "learning_rate": 8.4855704746422e-06, "loss": 0.0, "step": 314850 }, { "epoch": 67.88, "learning_rate": 8.480487487237711e-06, "loss": 0.0, "step": 314900 }, { "epoch": 67.89, "learning_rate": 8.475405422661962e-06, "loss": 0.0, "step": 314950 }, { "epoch": 67.9, "learning_rate": 8.47032428163431e-06, "loss": 0.0, "step": 315000 }, { "epoch": 67.91, "learning_rate": 8.465244064873974e-06, "loss": 0.0, "step": 315050 }, { "epoch": 67.92, "learning_rate": 8.460164773100055e-06, "loss": 0.0, "step": 315100 }, { "epoch": 67.93, "learning_rate": 8.455086407031528e-06, "loss": 0.0, "step": 315150 }, { "epoch": 67.95, "learning_rate": 8.450008967387224e-06, "loss": 0.0, "step": 315200 }, { "epoch": 67.96, "learning_rate": 8.444932454885855e-06, "loss": 0.0, "step": 315250 }, { "epoch": 67.97, "learning_rate": 8.439856870245985e-06, "loss": 0.0, "step": 315300 }, { "epoch": 67.98, "learning_rate": 8.43478221418605e-06, "loss": 0.0, "step": 315350 }, { "epoch": 67.99, "learning_rate": 8.429708487424383e-06, "loss": 0.0, "step": 315400 }, { "epoch": 68.0, "learning_rate": 8.424635690679149e-06, "loss": 0.0, "step": 315450 }, { "epoch": 68.01, "learning_rate": 8.419563824668406e-06, "loss": 0.0, "step": 315500 }, { "epoch": 68.02, "learning_rate": 8.414492890110052e-06, "loss": 0.0, "step": 315550 }, { "epoch": 68.03, "learning_rate": 8.409422887721887e-06, "loss": 0.0, "step": 315600 }, { "epoch": 68.04, "learning_rate": 8.40435381822156e-06, "loss": 0.0, "step": 315650 }, { "epoch": 68.05, "learning_rate": 8.39928568232659e-06, "loss": 0.0, "step": 315700 }, { "epoch": 68.06, "learning_rate": 8.394218480754367e-06, "loss": 0.0, "step": 315750 }, { "epoch": 68.08, "learning_rate": 8.389152214222145e-06, "loss": 0.0, "step": 315800 }, { "epoch": 68.09, "learning_rate": 8.384086883447048e-06, "loss": 0.0, "step": 315850 }, { "epoch": 68.1, "learning_rate": 8.379022489146068e-06, "loss": 0.0, "step": 315900 }, { "epoch": 68.11, "learning_rate": 8.373959032036063e-06, "loss": 0.0, "step": 315950 }, { "epoch": 68.12, "learning_rate": 8.368896512833758e-06, "loss": 0.0, "step": 316000 }, { "epoch": 68.13, "learning_rate": 8.363834932255747e-06, "loss": 0.0, "step": 316050 }, { "epoch": 68.14, "learning_rate": 8.35877429101849e-06, "loss": 0.0, "step": 316100 }, { "epoch": 68.15, "learning_rate": 8.353714589838317e-06, "loss": 0.0, "step": 316150 }, { "epoch": 68.16, "learning_rate": 8.348655829431418e-06, "loss": 0.0, "step": 316200 }, { "epoch": 68.17, "learning_rate": 8.34359801051386e-06, "loss": 0.0, "step": 316250 }, { "epoch": 68.18, "learning_rate": 8.338541133801563e-06, "loss": 0.0, "step": 316300 }, { "epoch": 68.19, "learning_rate": 8.333485200010325e-06, "loss": 0.0, "step": 316350 }, { "epoch": 68.2, "learning_rate": 8.328430209855803e-06, "loss": 0.0, "step": 316400 }, { "epoch": 68.22, "learning_rate": 8.323376164053543e-06, "loss": 0.0, "step": 316450 }, { "epoch": 68.23, "learning_rate": 8.318323063318917e-06, "loss": 0.0, "step": 316500 }, { "epoch": 68.24, "learning_rate": 8.313270908367191e-06, "loss": 0.0, "step": 316550 }, { "epoch": 68.25, "learning_rate": 8.30821969991349e-06, "loss": 0.0, "step": 316600 }, { "epoch": 68.26, "learning_rate": 8.303169438672814e-06, "loss": 0.0, "step": 316650 }, { "epoch": 68.27, "learning_rate": 8.298120125360021e-06, "loss": 0.0, "step": 316700 }, { "epoch": 68.28, "learning_rate": 8.293071760689825e-06, "loss": 0.0, "step": 316750 }, { "epoch": 68.29, "learning_rate": 8.288024345376815e-06, "loss": 0.0, "step": 316800 }, { "epoch": 68.3, "learning_rate": 8.282977880135458e-06, "loss": 0.0, "step": 316850 }, { "epoch": 68.31, "learning_rate": 8.277932365680066e-06, "loss": 0.0, "step": 316900 }, { "epoch": 68.32, "learning_rate": 8.272887802724838e-06, "loss": 0.0, "step": 316950 }, { "epoch": 68.33, "learning_rate": 8.267844191983806e-06, "loss": 0.0, "step": 317000 }, { "epoch": 68.34, "learning_rate": 8.262801534170892e-06, "loss": 0.0, "step": 317050 }, { "epoch": 68.36, "learning_rate": 8.25775982999989e-06, "loss": 0.0, "step": 317100 }, { "epoch": 68.37, "learning_rate": 8.252719080184435e-06, "loss": 0.0, "step": 317150 }, { "epoch": 68.38, "learning_rate": 8.24767928543805e-06, "loss": 0.0, "step": 317200 }, { "epoch": 68.39, "learning_rate": 8.242640446474091e-06, "loss": 0.0, "step": 317250 }, { "epoch": 68.4, "learning_rate": 8.237602564005817e-06, "loss": 0.0, "step": 317300 }, { "epoch": 68.41, "learning_rate": 8.232565638746333e-06, "loss": 0.0, "step": 317350 }, { "epoch": 68.42, "learning_rate": 8.227529671408604e-06, "loss": 0.0, "step": 317400 }, { "epoch": 68.43, "learning_rate": 8.222494662705464e-06, "loss": 0.0, "step": 317450 }, { "epoch": 68.44, "learning_rate": 8.217460613349616e-06, "loss": 0.0, "step": 317500 }, { "epoch": 68.45, "learning_rate": 8.212427524053625e-06, "loss": 0.0, "step": 317550 }, { "epoch": 68.46, "learning_rate": 8.20739539552991e-06, "loss": 0.0, "step": 317600 }, { "epoch": 68.47, "learning_rate": 8.20236422849077e-06, "loss": 0.0, "step": 317650 }, { "epoch": 68.48, "learning_rate": 8.197334023648357e-06, "loss": 0.0, "step": 317700 }, { "epoch": 68.5, "learning_rate": 8.192304781714693e-06, "loss": 0.0, "step": 317750 }, { "epoch": 68.51, "learning_rate": 8.187276503401658e-06, "loss": 0.0, "step": 317800 }, { "epoch": 68.52, "learning_rate": 8.182249189420999e-06, "loss": 0.0, "step": 317850 }, { "epoch": 68.53, "learning_rate": 8.177222840484326e-06, "loss": 0.0, "step": 317900 }, { "epoch": 68.54, "learning_rate": 8.172197457303111e-06, "loss": 0.0, "step": 317950 }, { "epoch": 68.55, "learning_rate": 8.167173040588695e-06, "loss": 0.0, "step": 318000 }, { "epoch": 68.56, "learning_rate": 8.162149591052272e-06, "loss": 0.0, "step": 318050 }, { "epoch": 68.57, "learning_rate": 8.157127109404906e-06, "loss": 0.0, "step": 318100 }, { "epoch": 68.58, "learning_rate": 8.152105596357532e-06, "loss": 0.0, "step": 318150 }, { "epoch": 68.59, "learning_rate": 8.147085052620928e-06, "loss": 0.0, "step": 318200 }, { "epoch": 68.6, "learning_rate": 8.14206547890575e-06, "loss": 0.0, "step": 318250 }, { "epoch": 68.61, "learning_rate": 8.137046875922503e-06, "loss": 0.0, "step": 318300 }, { "epoch": 68.62, "learning_rate": 8.13202924438158e-06, "loss": 0.0, "step": 318350 }, { "epoch": 68.64, "learning_rate": 8.127012584993218e-06, "loss": 0.0, "step": 318400 }, { "epoch": 68.65, "learning_rate": 8.12199689846751e-06, "loss": 0.0, "step": 318450 }, { "epoch": 68.66, "learning_rate": 8.116982185514421e-06, "loss": 0.0, "step": 318500 }, { "epoch": 68.67, "learning_rate": 8.111968446843775e-06, "loss": 0.0, "step": 318550 }, { "epoch": 68.68, "learning_rate": 8.106955683165271e-06, "loss": 0.0, "step": 318600 }, { "epoch": 68.69, "learning_rate": 8.10194389518846e-06, "loss": 0.0, "step": 318650 }, { "epoch": 68.7, "learning_rate": 8.09693308362274e-06, "loss": 0.0, "step": 318700 }, { "epoch": 68.71, "learning_rate": 8.091923249177387e-06, "loss": 0.0, "step": 318750 }, { "epoch": 68.72, "learning_rate": 8.08691439256155e-06, "loss": 0.0, "step": 318800 }, { "epoch": 68.73, "learning_rate": 8.081906514484218e-06, "loss": 0.0, "step": 318850 }, { "epoch": 68.74, "learning_rate": 8.076899615654257e-06, "loss": 0.0, "step": 318900 }, { "epoch": 68.75, "learning_rate": 8.07189369678037e-06, "loss": 0.0, "step": 318950 }, { "epoch": 68.76, "learning_rate": 8.066888758571155e-06, "loss": 0.0, "step": 319000 }, { "epoch": 68.78, "learning_rate": 8.061884801735048e-06, "loss": 0.0, "step": 319050 }, { "epoch": 68.79, "learning_rate": 8.056881826980352e-06, "loss": 0.0, "step": 319100 }, { "epoch": 68.8, "learning_rate": 8.051879835015236e-06, "loss": 0.0, "step": 319150 }, { "epoch": 68.81, "learning_rate": 8.04687882654772e-06, "loss": 0.0, "step": 319200 }, { "epoch": 68.82, "learning_rate": 8.041878802285697e-06, "loss": 0.0, "step": 319250 }, { "epoch": 68.83, "learning_rate": 8.036879762936911e-06, "loss": 0.0, "step": 319300 }, { "epoch": 68.84, "learning_rate": 8.031881709208965e-06, "loss": 0.0, "step": 319350 }, { "epoch": 68.85, "learning_rate": 8.026884641809335e-06, "loss": 0.0, "step": 319400 }, { "epoch": 68.86, "learning_rate": 8.021888561445346e-06, "loss": 0.0, "step": 319450 }, { "epoch": 68.87, "learning_rate": 8.016893468824185e-06, "loss": 0.0, "step": 319500 }, { "epoch": 68.88, "learning_rate": 8.011899364652904e-06, "loss": 0.0, "step": 319550 }, { "epoch": 68.89, "learning_rate": 8.00690624963841e-06, "loss": 0.0, "step": 319600 }, { "epoch": 68.9, "learning_rate": 8.001914124487474e-06, "loss": 0.0, "step": 319650 }, { "epoch": 68.92, "learning_rate": 7.996922989906722e-06, "loss": 0.0, "step": 319700 }, { "epoch": 68.93, "learning_rate": 7.991932846602647e-06, "loss": 0.0, "step": 319750 }, { "epoch": 68.94, "learning_rate": 7.98694369528159e-06, "loss": 0.0, "step": 319800 }, { "epoch": 68.95, "learning_rate": 7.981955536649775e-06, "loss": 0.0, "step": 319850 }, { "epoch": 68.96, "learning_rate": 7.976968371413254e-06, "loss": 0.0, "step": 319900 }, { "epoch": 68.97, "learning_rate": 7.971982200277957e-06, "loss": 0.0, "step": 319950 }, { "epoch": 68.98, "learning_rate": 7.966997023949665e-06, "loss": 0.0, "step": 320000 }, { "epoch": 68.99, "learning_rate": 7.96201284313404e-06, "loss": 0.0, "step": 320050 }, { "epoch": 69.0, "learning_rate": 7.95702965853658e-06, "loss": 0.0, "step": 320100 }, { "epoch": 69.01, "learning_rate": 7.952047470862642e-06, "loss": 0.0, "step": 320150 }, { "epoch": 69.02, "learning_rate": 7.947066280817449e-06, "loss": 0.0, "step": 320200 }, { "epoch": 69.03, "learning_rate": 7.94208608910608e-06, "loss": 0.0, "step": 320250 }, { "epoch": 69.05, "learning_rate": 7.937106896433486e-06, "loss": 0.0, "step": 320300 }, { "epoch": 69.06, "learning_rate": 7.932128703504463e-06, "loss": 0.0, "step": 320350 }, { "epoch": 69.07, "learning_rate": 7.927151511023661e-06, "loss": 0.0, "step": 320400 }, { "epoch": 69.08, "learning_rate": 7.922175319695591e-06, "loss": 0.0, "step": 320450 }, { "epoch": 69.09, "learning_rate": 7.91720013022464e-06, "loss": 0.0, "step": 320500 }, { "epoch": 69.1, "learning_rate": 7.912225943315032e-06, "loss": 0.0, "step": 320550 }, { "epoch": 69.11, "learning_rate": 7.907252759670865e-06, "loss": 0.0, "step": 320600 }, { "epoch": 69.12, "learning_rate": 7.902280579996069e-06, "loss": 0.0, "step": 320650 }, { "epoch": 69.13, "learning_rate": 7.897309404994468e-06, "loss": 0.0, "step": 320700 }, { "epoch": 69.14, "learning_rate": 7.892339235369718e-06, "loss": 0.0, "step": 320750 }, { "epoch": 69.15, "learning_rate": 7.887370071825338e-06, "loss": 0.0, "step": 320800 }, { "epoch": 69.16, "learning_rate": 7.882401915064713e-06, "loss": 0.0, "step": 320850 }, { "epoch": 69.17, "learning_rate": 7.877434765791075e-06, "loss": 0.0, "step": 320900 }, { "epoch": 69.19, "learning_rate": 7.872468624707517e-06, "loss": 0.0, "step": 320950 }, { "epoch": 69.2, "learning_rate": 7.867503492516993e-06, "loss": 0.0, "step": 321000 }, { "epoch": 69.21, "learning_rate": 7.862539369922308e-06, "loss": 0.0, "step": 321050 }, { "epoch": 69.22, "learning_rate": 7.857576257626132e-06, "loss": 0.0, "step": 321100 }, { "epoch": 69.23, "learning_rate": 7.85261415633098e-06, "loss": 0.0, "step": 321150 }, { "epoch": 69.24, "learning_rate": 7.84765306673924e-06, "loss": 0.0, "step": 321200 }, { "epoch": 69.25, "learning_rate": 7.842692989553139e-06, "loss": 0.0, "step": 321250 }, { "epoch": 69.26, "learning_rate": 7.837733925474776e-06, "loss": 0.0, "step": 321300 }, { "epoch": 69.27, "learning_rate": 7.832775875206097e-06, "loss": 0.0, "step": 321350 }, { "epoch": 69.28, "learning_rate": 7.82781883944891e-06, "loss": 0.0, "step": 321400 }, { "epoch": 69.29, "learning_rate": 7.822862818904876e-06, "loss": 0.0, "step": 321450 }, { "epoch": 69.3, "learning_rate": 7.817907814275508e-06, "loss": 0.0, "step": 321500 }, { "epoch": 69.31, "learning_rate": 7.812953826262196e-06, "loss": 0.0, "step": 321550 }, { "epoch": 69.33, "learning_rate": 7.808000855566156e-06, "loss": 0.0, "step": 321600 }, { "epoch": 69.34, "learning_rate": 7.80304890288848e-06, "loss": 0.0, "step": 321650 }, { "epoch": 69.35, "learning_rate": 7.798097968930102e-06, "loss": 0.0, "step": 321700 }, { "epoch": 69.36, "learning_rate": 7.793148054391837e-06, "loss": 0.0, "step": 321750 }, { "epoch": 69.37, "learning_rate": 7.788199159974335e-06, "loss": 0.0, "step": 321800 }, { "epoch": 69.38, "learning_rate": 7.783251286378099e-06, "loss": 0.0, "step": 321850 }, { "epoch": 69.39, "learning_rate": 7.778304434303493e-06, "loss": 0.0, "step": 321900 }, { "epoch": 69.4, "learning_rate": 7.773358604450733e-06, "loss": 0.0, "step": 321950 }, { "epoch": 69.41, "learning_rate": 7.768413797519915e-06, "loss": 0.0, "step": 322000 }, { "epoch": 69.42, "learning_rate": 7.763470014210962e-06, "loss": 0.0, "step": 322050 }, { "epoch": 69.43, "learning_rate": 7.75852725522365e-06, "loss": 0.0, "step": 322100 }, { "epoch": 69.44, "learning_rate": 7.753585521257622e-06, "loss": 0.0, "step": 322150 }, { "epoch": 69.45, "learning_rate": 7.748644813012385e-06, "loss": 0.0, "step": 322200 }, { "epoch": 69.47, "learning_rate": 7.743705131187283e-06, "loss": 0.0, "step": 322250 }, { "epoch": 69.48, "learning_rate": 7.738766476481528e-06, "loss": 0.0, "step": 322300 }, { "epoch": 69.49, "learning_rate": 7.733828849594167e-06, "loss": 0.0, "step": 322350 }, { "epoch": 69.5, "learning_rate": 7.728892251224126e-06, "loss": 0.0, "step": 322400 }, { "epoch": 69.51, "learning_rate": 7.723956682070171e-06, "loss": 0.0, "step": 322450 }, { "epoch": 69.52, "learning_rate": 7.719022142830924e-06, "loss": 0.0, "step": 322500 }, { "epoch": 69.53, "learning_rate": 7.714088634204869e-06, "loss": 0.0, "step": 322550 }, { "epoch": 69.54, "learning_rate": 7.709156156890331e-06, "loss": 0.0, "step": 322600 }, { "epoch": 69.55, "learning_rate": 7.704224711585498e-06, "loss": 0.0, "step": 322650 }, { "epoch": 69.56, "learning_rate": 7.69929429898841e-06, "loss": 0.0, "step": 322700 }, { "epoch": 69.57, "learning_rate": 7.694364919796962e-06, "loss": 0.0, "step": 322750 }, { "epoch": 69.58, "learning_rate": 7.6894365747089e-06, "loss": 0.0, "step": 322800 }, { "epoch": 69.59, "learning_rate": 7.684509264421826e-06, "loss": 0.0, "step": 322850 }, { "epoch": 69.61, "learning_rate": 7.679582989633193e-06, "loss": 0.0, "step": 322900 }, { "epoch": 69.62, "learning_rate": 7.674657751040308e-06, "loss": 0.0, "step": 322950 }, { "epoch": 69.63, "learning_rate": 7.669733549340335e-06, "loss": 0.0, "step": 323000 }, { "epoch": 69.64, "learning_rate": 7.664810385230289e-06, "loss": 0.0, "step": 323050 }, { "epoch": 69.65, "learning_rate": 7.659888259407035e-06, "loss": 0.0, "step": 323100 }, { "epoch": 69.66, "learning_rate": 7.654967172567295e-06, "loss": 0.0, "step": 323150 }, { "epoch": 69.67, "learning_rate": 7.650047125407636e-06, "loss": 0.0, "step": 323200 }, { "epoch": 69.68, "learning_rate": 7.6451281186245e-06, "loss": 0.0, "step": 323250 }, { "epoch": 69.69, "learning_rate": 7.640210152914156e-06, "loss": 0.0, "step": 323300 }, { "epoch": 69.7, "learning_rate": 7.635293228972733e-06, "loss": 0.0, "step": 323350 }, { "epoch": 69.71, "learning_rate": 7.630377347496221e-06, "loss": 0.0, "step": 323400 }, { "epoch": 69.72, "learning_rate": 7.625462509180447e-06, "loss": 0.0, "step": 323450 }, { "epoch": 69.73, "learning_rate": 7.620548714721124e-06, "loss": 0.0, "step": 323500 }, { "epoch": 69.75, "learning_rate": 7.615635964813768e-06, "loss": 0.0, "step": 323550 }, { "epoch": 69.76, "learning_rate": 7.610724260153783e-06, "loss": 0.0, "step": 323600 }, { "epoch": 69.77, "learning_rate": 7.605813601436408e-06, "loss": 0.0, "step": 323650 }, { "epoch": 69.78, "learning_rate": 7.600903989356752e-06, "loss": 0.0, "step": 323700 }, { "epoch": 69.79, "learning_rate": 7.595995424609763e-06, "loss": 0.0, "step": 323750 }, { "epoch": 69.8, "learning_rate": 7.591087907890234e-06, "loss": 0.0, "step": 323800 }, { "epoch": 69.81, "learning_rate": 7.586181439892814e-06, "loss": 0.0, "step": 323850 }, { "epoch": 69.82, "learning_rate": 7.581276021312021e-06, "loss": 0.0, "step": 323900 }, { "epoch": 69.83, "learning_rate": 7.576371652842203e-06, "loss": 0.0, "step": 323950 }, { "epoch": 69.84, "learning_rate": 7.571468335177575e-06, "loss": 0.0, "step": 324000 }, { "epoch": 69.85, "learning_rate": 7.566566069012179e-06, "loss": 0.0, "step": 324050 }, { "epoch": 69.86, "learning_rate": 7.561664855039939e-06, "loss": 0.0, "step": 324100 }, { "epoch": 69.87, "learning_rate": 7.556764693954612e-06, "loss": 0.0, "step": 324150 }, { "epoch": 69.89, "learning_rate": 7.5518655864498075e-06, "loss": 0.0, "step": 324200 }, { "epoch": 69.9, "learning_rate": 7.546967533218995e-06, "loss": 0.0, "step": 324250 }, { "epoch": 69.91, "learning_rate": 7.542070534955472e-06, "loss": 0.0, "step": 324300 }, { "epoch": 69.92, "learning_rate": 7.537174592352416e-06, "loss": 0.0, "step": 324350 }, { "epoch": 69.93, "learning_rate": 7.532279706102839e-06, "loss": 0.0, "step": 324400 }, { "epoch": 69.94, "learning_rate": 7.527385876899604e-06, "loss": 0.0, "step": 324450 }, { "epoch": 69.95, "learning_rate": 7.5224931054354286e-06, "loss": 0.0, "step": 324500 }, { "epoch": 69.96, "learning_rate": 7.517601392402875e-06, "loss": 0.0, "step": 324550 }, { "epoch": 69.97, "learning_rate": 7.512710738494362e-06, "loss": 0.0, "step": 324600 }, { "epoch": 69.98, "learning_rate": 7.5078211444021534e-06, "loss": 0.0, "step": 324650 }, { "epoch": 69.99, "learning_rate": 7.502932610818369e-06, "loss": 0.0, "step": 324700 }, { "epoch": 70.0, "learning_rate": 7.4980451384349684e-06, "loss": 0.0, "step": 324750 }, { "epoch": 70.02, "learning_rate": 7.493158727943771e-06, "loss": 0.0, "step": 324800 }, { "epoch": 70.03, "learning_rate": 7.48827338003644e-06, "loss": 0.0, "step": 324850 }, { "epoch": 70.04, "learning_rate": 7.483389095404488e-06, "loss": 0.0, "step": 324900 }, { "epoch": 70.05, "learning_rate": 7.4785058747392925e-06, "loss": 0.0, "step": 324950 }, { "epoch": 70.06, "learning_rate": 7.473623718732052e-06, "loss": 0.0, "step": 325000 }, { "epoch": 70.07, "learning_rate": 7.468742628073836e-06, "loss": 0.0, "step": 325050 }, { "epoch": 70.08, "learning_rate": 7.4638626034555544e-06, "loss": 0.0, "step": 325100 }, { "epoch": 70.09, "learning_rate": 7.458983645567964e-06, "loss": 0.0, "step": 325150 }, { "epoch": 70.1, "learning_rate": 7.454105755101692e-06, "loss": 0.0, "step": 325200 }, { "epoch": 70.11, "learning_rate": 7.449228932747182e-06, "loss": 0.0, "step": 325250 }, { "epoch": 70.12, "learning_rate": 7.444353179194746e-06, "loss": 0.0, "step": 325300 }, { "epoch": 70.13, "learning_rate": 7.439478495134535e-06, "loss": 0.0, "step": 325350 }, { "epoch": 70.14, "learning_rate": 7.434604881256567e-06, "loss": 0.0, "step": 325400 }, { "epoch": 70.16, "learning_rate": 7.429732338250695e-06, "loss": 0.0, "step": 325450 }, { "epoch": 70.17, "learning_rate": 7.424860866806608e-06, "loss": 0.0, "step": 325500 }, { "epoch": 70.18, "learning_rate": 7.419990467613862e-06, "loss": 0.0, "step": 325550 }, { "epoch": 70.19, "learning_rate": 7.415121141361862e-06, "loss": 0.0, "step": 325600 }, { "epoch": 70.2, "learning_rate": 7.4102528887398505e-06, "loss": 0.0, "step": 325650 }, { "epoch": 70.21, "learning_rate": 7.40538571043693e-06, "loss": 0.0, "step": 325700 }, { "epoch": 70.22, "learning_rate": 7.400519607142025e-06, "loss": 0.0, "step": 325750 }, { "epoch": 70.23, "learning_rate": 7.395654579543943e-06, "loss": 0.0, "step": 325800 }, { "epoch": 70.24, "learning_rate": 7.390790628331317e-06, "loss": 0.0, "step": 325850 }, { "epoch": 70.25, "learning_rate": 7.385927754192633e-06, "loss": 0.0, "step": 325900 }, { "epoch": 70.26, "learning_rate": 7.381065957816232e-06, "loss": 0.0, "step": 325950 }, { "epoch": 70.27, "learning_rate": 7.376205239890277e-06, "loss": 0.0, "step": 326000 }, { "epoch": 70.28, "learning_rate": 7.371345601102812e-06, "loss": 0.0, "step": 326050 }, { "epoch": 70.3, "learning_rate": 7.3664870421417085e-06, "loss": 0.0, "step": 326100 }, { "epoch": 70.31, "learning_rate": 7.36162956369469e-06, "loss": 0.0, "step": 326150 }, { "epoch": 70.32, "learning_rate": 7.356773166449325e-06, "loss": 0.0, "step": 326200 }, { "epoch": 70.33, "learning_rate": 7.351917851093032e-06, "loss": 0.0, "step": 326250 }, { "epoch": 70.34, "learning_rate": 7.347063618313071e-06, "loss": 0.0, "step": 326300 }, { "epoch": 70.35, "learning_rate": 7.342210468796558e-06, "loss": 0.0, "step": 326350 }, { "epoch": 70.36, "learning_rate": 7.3373584032304466e-06, "loss": 0.0, "step": 326400 }, { "epoch": 70.37, "learning_rate": 7.33250742230154e-06, "loss": 0.0, "step": 326450 }, { "epoch": 70.38, "learning_rate": 7.327657526696491e-06, "loss": 0.0, "step": 326500 }, { "epoch": 70.39, "learning_rate": 7.322808717101795e-06, "loss": 0.0, "step": 326550 }, { "epoch": 70.4, "learning_rate": 7.317960994203789e-06, "loss": 0.0, "step": 326600 }, { "epoch": 70.41, "learning_rate": 7.313114358688679e-06, "loss": 0.0, "step": 326650 }, { "epoch": 70.42, "learning_rate": 7.308268811242483e-06, "loss": 0.0, "step": 326700 }, { "epoch": 70.44, "learning_rate": 7.30342435255109e-06, "loss": 0.0, "step": 326750 }, { "epoch": 70.45, "learning_rate": 7.298580983300223e-06, "loss": 0.0, "step": 326800 }, { "epoch": 70.46, "learning_rate": 7.29373870417545e-06, "loss": 0.0, "step": 326850 }, { "epoch": 70.47, "learning_rate": 7.28889751586221e-06, "loss": 0.0, "step": 326900 }, { "epoch": 70.48, "learning_rate": 7.2840574190457475e-06, "loss": 0.0, "step": 326950 }, { "epoch": 70.49, "learning_rate": 7.279218414411177e-06, "loss": 0.0, "step": 327000 }, { "epoch": 70.5, "learning_rate": 7.274380502643447e-06, "loss": 0.0, "step": 327050 }, { "epoch": 70.51, "learning_rate": 7.269543684427373e-06, "loss": 0.0, "step": 327100 }, { "epoch": 70.52, "learning_rate": 7.264707960447599e-06, "loss": 0.0, "step": 327150 }, { "epoch": 70.53, "learning_rate": 7.2598733313886014e-06, "loss": 0.0, "step": 327200 }, { "epoch": 70.54, "learning_rate": 7.255039797934718e-06, "loss": 0.0, "step": 327250 }, { "epoch": 70.55, "learning_rate": 7.250207360770143e-06, "loss": 0.0, "step": 327300 }, { "epoch": 70.56, "learning_rate": 7.24537602057889e-06, "loss": 0.0, "step": 327350 }, { "epoch": 70.58, "learning_rate": 7.240545778044842e-06, "loss": 0.0, "step": 327400 }, { "epoch": 70.59, "learning_rate": 7.235716633851691e-06, "loss": 0.0, "step": 327450 }, { "epoch": 70.6, "learning_rate": 7.230888588683016e-06, "loss": 0.0, "step": 327500 }, { "epoch": 70.61, "learning_rate": 7.226061643222216e-06, "loss": 0.0, "step": 327550 }, { "epoch": 70.62, "learning_rate": 7.221235798152536e-06, "loss": 0.0, "step": 327600 }, { "epoch": 70.63, "learning_rate": 7.216411054157068e-06, "loss": 0.0, "step": 327650 }, { "epoch": 70.64, "learning_rate": 7.211587411918754e-06, "loss": 0.0, "step": 327700 }, { "epoch": 70.65, "learning_rate": 7.2067648721203674e-06, "loss": 0.0, "step": 327750 }, { "epoch": 70.66, "learning_rate": 7.201943435444537e-06, "loss": 0.0, "step": 327800 }, { "epoch": 70.67, "learning_rate": 7.1971231025737295e-06, "loss": 0.0, "step": 327850 }, { "epoch": 70.68, "learning_rate": 7.192303874190256e-06, "loss": 0.0, "step": 327900 }, { "epoch": 70.69, "learning_rate": 7.187485750976275e-06, "loss": 0.0, "step": 327950 }, { "epoch": 70.7, "learning_rate": 7.1826687336137845e-06, "loss": 0.0, "step": 328000 }, { "epoch": 70.72, "learning_rate": 7.177852822784626e-06, "loss": 0.0, "step": 328050 }, { "epoch": 70.73, "learning_rate": 7.173038019170486e-06, "loss": 0.0, "step": 328100 }, { "epoch": 70.74, "learning_rate": 7.168224323452895e-06, "loss": 0.0, "step": 328150 }, { "epoch": 70.75, "learning_rate": 7.163411736313224e-06, "loss": 0.0, "step": 328200 }, { "epoch": 70.76, "learning_rate": 7.15860025843269e-06, "loss": 0.0, "step": 328250 }, { "epoch": 70.77, "learning_rate": 7.153789890492351e-06, "loss": 0.0, "step": 328300 }, { "epoch": 70.78, "learning_rate": 7.1489806331731065e-06, "loss": 0.0, "step": 328350 }, { "epoch": 70.79, "learning_rate": 7.144172487155705e-06, "loss": 0.0, "step": 328400 }, { "epoch": 70.8, "learning_rate": 7.139365453120731e-06, "loss": 0.0, "step": 328450 }, { "epoch": 70.81, "learning_rate": 7.134559531748613e-06, "loss": 0.0, "step": 328500 }, { "epoch": 70.82, "learning_rate": 7.12975472371962e-06, "loss": 0.0, "step": 328550 }, { "epoch": 70.83, "learning_rate": 7.124951029713881e-06, "loss": 0.0, "step": 328600 }, { "epoch": 70.85, "learning_rate": 7.120148450411337e-06, "loss": 0.0, "step": 328650 }, { "epoch": 70.86, "learning_rate": 7.1153469864917935e-06, "loss": 0.0, "step": 328700 }, { "epoch": 70.87, "learning_rate": 7.110546638634886e-06, "loss": 0.0, "step": 328750 }, { "epoch": 70.88, "learning_rate": 7.105747407520106e-06, "loss": 0.0, "step": 328800 }, { "epoch": 70.89, "learning_rate": 7.100949293826782e-06, "loss": 0.0, "step": 328850 }, { "epoch": 70.9, "learning_rate": 7.096152298234067e-06, "loss": 0.0, "step": 328900 }, { "epoch": 70.91, "learning_rate": 7.091356421420973e-06, "loss": 0.0, "step": 328950 }, { "epoch": 70.92, "learning_rate": 7.086561664066357e-06, "loss": 0.0, "step": 329000 }, { "epoch": 70.93, "learning_rate": 7.0817680268489095e-06, "loss": 0.0, "step": 329050 }, { "epoch": 70.94, "learning_rate": 7.076975510447161e-06, "loss": 0.0, "step": 329100 }, { "epoch": 70.95, "learning_rate": 7.072184115539494e-06, "loss": 0.0, "step": 329150 }, { "epoch": 70.96, "learning_rate": 7.067393842804104e-06, "loss": 0.0, "step": 329200 }, { "epoch": 70.97, "learning_rate": 7.062604692919069e-06, "loss": 0.0, "step": 329250 }, { "epoch": 70.99, "learning_rate": 7.057816666562278e-06, "loss": 0.0, "step": 329300 }, { "epoch": 71.0, "learning_rate": 7.053029764411473e-06, "loss": 0.0, "step": 329350 }, { "epoch": 71.01, "learning_rate": 7.048243987144232e-06, "loss": 0.0, "step": 329400 }, { "epoch": 71.02, "learning_rate": 7.043459335437976e-06, "loss": 0.0, "step": 329450 }, { "epoch": 71.03, "learning_rate": 7.038675809969969e-06, "loss": 0.0, "step": 329500 }, { "epoch": 71.04, "learning_rate": 7.033893411417308e-06, "loss": 0.0, "step": 329550 }, { "epoch": 71.05, "learning_rate": 7.029112140456939e-06, "loss": 0.0, "step": 329600 }, { "epoch": 71.06, "learning_rate": 7.024331997765643e-06, "loss": 0.0, "step": 329650 }, { "epoch": 71.07, "learning_rate": 7.019552984020044e-06, "loss": 0.0, "step": 329700 }, { "epoch": 71.08, "learning_rate": 7.014775099896608e-06, "loss": 0.0, "step": 329750 }, { "epoch": 71.09, "learning_rate": 7.009998346071633e-06, "loss": 0.0, "step": 329800 }, { "epoch": 71.1, "learning_rate": 7.005222723221268e-06, "loss": 0.0, "step": 329850 }, { "epoch": 71.11, "learning_rate": 7.000448232021491e-06, "loss": 0.0, "step": 329900 }, { "epoch": 71.13, "learning_rate": 6.9956748731481296e-06, "loss": 0.0, "step": 329950 }, { "epoch": 71.14, "learning_rate": 6.990902647276844e-06, "loss": 0.0, "step": 330000 }, { "epoch": 71.15, "learning_rate": 6.986131555083137e-06, "loss": 0.0, "step": 330050 }, { "epoch": 71.16, "learning_rate": 6.981361597242353e-06, "loss": 0.0, "step": 330100 }, { "epoch": 71.17, "learning_rate": 6.9765927744296715e-06, "loss": 0.0, "step": 330150 }, { "epoch": 71.18, "learning_rate": 6.9718250873201115e-06, "loss": 0.0, "step": 330200 }, { "epoch": 71.19, "learning_rate": 6.967058536588532e-06, "loss": 0.0, "step": 330250 }, { "epoch": 71.2, "learning_rate": 6.962293122909641e-06, "loss": 0.0, "step": 330300 }, { "epoch": 71.21, "learning_rate": 6.957528846957975e-06, "loss": 0.0, "step": 330350 }, { "epoch": 71.22, "learning_rate": 6.952765709407904e-06, "loss": 0.0, "step": 330400 }, { "epoch": 71.23, "learning_rate": 6.94800371093364e-06, "loss": 0.0, "step": 330450 }, { "epoch": 71.24, "learning_rate": 6.943242852209253e-06, "loss": 0.0, "step": 330500 }, { "epoch": 71.25, "learning_rate": 6.938483133908627e-06, "loss": 0.0, "step": 330550 }, { "epoch": 71.27, "learning_rate": 6.933724556705501e-06, "loss": 0.0, "step": 330600 }, { "epoch": 71.28, "learning_rate": 6.92896712127343e-06, "loss": 0.0, "step": 330650 }, { "epoch": 71.29, "learning_rate": 6.924210828285839e-06, "loss": 0.0, "step": 330700 }, { "epoch": 71.3, "learning_rate": 6.919455678415971e-06, "loss": 0.0, "step": 330750 }, { "epoch": 71.31, "learning_rate": 6.91470167233691e-06, "loss": 0.0, "step": 330800 }, { "epoch": 71.32, "learning_rate": 6.909948810721584e-06, "loss": 0.0, "step": 330850 }, { "epoch": 71.33, "learning_rate": 6.90519709424274e-06, "loss": 0.0, "step": 330900 }, { "epoch": 71.34, "learning_rate": 6.900446523572993e-06, "loss": 0.0, "step": 330950 }, { "epoch": 71.35, "learning_rate": 6.895697099384775e-06, "loss": 0.0, "step": 331000 }, { "epoch": 71.36, "learning_rate": 6.890948822350361e-06, "loss": 0.0, "step": 331050 }, { "epoch": 71.37, "learning_rate": 6.886201693141862e-06, "loss": 0.0, "step": 331100 }, { "epoch": 71.38, "learning_rate": 6.8814557124312325e-06, "loss": 0.0, "step": 331150 }, { "epoch": 71.39, "learning_rate": 6.876710880890256e-06, "loss": 0.0, "step": 331200 }, { "epoch": 71.41, "learning_rate": 6.8719671991905576e-06, "loss": 0.0, "step": 331250 }, { "epoch": 71.42, "learning_rate": 6.867224668003601e-06, "loss": 0.0, "step": 331300 }, { "epoch": 71.43, "learning_rate": 6.862483288000685e-06, "loss": 0.0, "step": 331350 }, { "epoch": 71.44, "learning_rate": 6.857743059852945e-06, "loss": 0.0, "step": 331400 }, { "epoch": 71.45, "learning_rate": 6.853003984231355e-06, "loss": 0.0, "step": 331450 }, { "epoch": 71.46, "learning_rate": 6.848266061806727e-06, "loss": 0.0, "step": 331500 }, { "epoch": 71.47, "learning_rate": 6.843529293249705e-06, "loss": 0.0, "step": 331550 }, { "epoch": 71.48, "learning_rate": 6.838793679230773e-06, "loss": 0.0, "step": 331600 }, { "epoch": 71.49, "learning_rate": 6.834059220420253e-06, "loss": 0.0, "step": 331650 }, { "epoch": 71.5, "learning_rate": 6.8293259174883e-06, "loss": 0.0, "step": 331700 }, { "epoch": 71.51, "learning_rate": 6.824593771104902e-06, "loss": 0.0, "step": 331750 }, { "epoch": 71.52, "learning_rate": 6.819862781939906e-06, "loss": 0.0, "step": 331800 }, { "epoch": 71.53, "learning_rate": 6.81513295066296e-06, "loss": 0.0, "step": 331850 }, { "epoch": 71.55, "learning_rate": 6.8104042779435695e-06, "loss": 0.0, "step": 331900 }, { "epoch": 71.56, "learning_rate": 6.805676764451068e-06, "loss": 0.0, "step": 331950 }, { "epoch": 71.57, "learning_rate": 6.800950410854642e-06, "loss": 0.0, "step": 332000 }, { "epoch": 71.58, "learning_rate": 6.796225217823299e-06, "loss": 0.0, "step": 332050 }, { "epoch": 71.59, "learning_rate": 6.791501186025871e-06, "loss": 0.0, "step": 332100 }, { "epoch": 71.6, "learning_rate": 6.7867783161310424e-06, "loss": 0.0, "step": 332150 }, { "epoch": 71.61, "learning_rate": 6.782056608807339e-06, "loss": 0.0, "step": 332200 }, { "epoch": 71.62, "learning_rate": 6.777336064723107e-06, "loss": 0.0, "step": 332250 }, { "epoch": 71.63, "learning_rate": 6.772616684546539e-06, "loss": 0.0, "step": 332300 }, { "epoch": 71.64, "learning_rate": 6.76789846894564e-06, "loss": 0.0, "step": 332350 }, { "epoch": 71.65, "learning_rate": 6.763181418588283e-06, "loss": 0.0, "step": 332400 }, { "epoch": 71.66, "learning_rate": 6.7584655341421595e-06, "loss": 0.0, "step": 332450 }, { "epoch": 71.67, "learning_rate": 6.7537508162747935e-06, "loss": 0.0, "step": 332500 }, { "epoch": 71.69, "learning_rate": 6.749037265653553e-06, "loss": 0.0, "step": 332550 }, { "epoch": 71.7, "learning_rate": 6.744324882945618e-06, "loss": 0.0, "step": 332600 }, { "epoch": 71.71, "learning_rate": 6.739613668818039e-06, "loss": 0.0, "step": 332650 }, { "epoch": 71.72, "learning_rate": 6.734903623937674e-06, "loss": 0.0, "step": 332700 }, { "epoch": 71.73, "learning_rate": 6.730194748971226e-06, "loss": 0.0, "step": 332750 }, { "epoch": 71.74, "learning_rate": 6.725487044585231e-06, "loss": 0.0, "step": 332800 }, { "epoch": 71.75, "learning_rate": 6.720780511446057e-06, "loss": 0.0, "step": 332850 }, { "epoch": 71.76, "learning_rate": 6.716075150219908e-06, "loss": 0.0, "step": 332900 }, { "epoch": 71.77, "learning_rate": 6.711370961572822e-06, "loss": 0.0, "step": 332950 }, { "epoch": 71.78, "learning_rate": 6.706667946170671e-06, "loss": 0.0, "step": 333000 }, { "epoch": 71.79, "learning_rate": 6.701966104679163e-06, "loss": 0.0, "step": 333050 }, { "epoch": 71.8, "learning_rate": 6.697265437763834e-06, "loss": 0.0, "step": 333100 }, { "epoch": 71.82, "learning_rate": 6.692565946090058e-06, "loss": 0.0, "step": 333150 }, { "epoch": 71.83, "learning_rate": 6.687867630323045e-06, "loss": 0.0, "step": 333200 }, { "epoch": 71.84, "learning_rate": 6.683170491127832e-06, "loss": 0.0, "step": 333250 }, { "epoch": 71.85, "learning_rate": 6.678474529169298e-06, "loss": 0.0, "step": 333300 }, { "epoch": 71.86, "learning_rate": 6.673779745112145e-06, "loss": 0.0, "step": 333350 }, { "epoch": 71.87, "learning_rate": 6.669086139620917e-06, "loss": 0.0, "step": 333400 }, { "epoch": 71.88, "learning_rate": 6.664393713359981e-06, "loss": 0.0, "step": 333450 }, { "epoch": 71.89, "learning_rate": 6.6597024669935605e-06, "loss": 0.0, "step": 333500 }, { "epoch": 71.9, "learning_rate": 6.65501240118568e-06, "loss": 0.0, "step": 333550 }, { "epoch": 71.91, "learning_rate": 6.6503235166002176e-06, "loss": 0.0, "step": 333600 }, { "epoch": 71.92, "learning_rate": 6.6456358139008735e-06, "loss": 0.0, "step": 333650 }, { "epoch": 71.93, "learning_rate": 6.6409492937511965e-06, "loss": 0.0, "step": 333700 }, { "epoch": 71.94, "learning_rate": 6.636263956814558e-06, "loss": 0.0, "step": 333750 }, { "epoch": 71.96, "learning_rate": 6.6315798037541506e-06, "loss": 0.0, "step": 333800 }, { "epoch": 71.97, "learning_rate": 6.626896835233011e-06, "loss": 0.0, "step": 333850 }, { "epoch": 71.98, "learning_rate": 6.622215051914015e-06, "loss": 0.0, "step": 333900 }, { "epoch": 71.99, "learning_rate": 6.617534454459862e-06, "loss": 0.0, "step": 333950 }, { "epoch": 72.0, "learning_rate": 6.612855043533088e-06, "loss": 0.0, "step": 334000 }, { "epoch": 72.01, "learning_rate": 6.608176819796047e-06, "loss": 0.0, "step": 334050 }, { "epoch": 72.02, "learning_rate": 6.603499783910934e-06, "loss": 0.0, "step": 334100 }, { "epoch": 72.03, "learning_rate": 6.598823936539792e-06, "loss": 0.0, "step": 334150 }, { "epoch": 72.04, "learning_rate": 6.5941492783444726e-06, "loss": 0.0, "step": 334200 }, { "epoch": 72.05, "learning_rate": 6.589475809986675e-06, "loss": 0.0, "step": 334250 }, { "epoch": 72.06, "learning_rate": 6.584803532127907e-06, "loss": 0.0, "step": 334300 }, { "epoch": 72.07, "learning_rate": 6.580132445429537e-06, "loss": 0.0, "step": 334350 }, { "epoch": 72.08, "learning_rate": 6.575462550552747e-06, "loss": 0.0, "step": 334400 }, { "epoch": 72.1, "learning_rate": 6.5707938481585565e-06, "loss": 0.0, "step": 334450 }, { "epoch": 72.11, "learning_rate": 6.566126338907814e-06, "loss": 0.0, "step": 334500 }, { "epoch": 72.12, "learning_rate": 6.561460023461198e-06, "loss": 0.0, "step": 334550 }, { "epoch": 72.13, "learning_rate": 6.556794902479221e-06, "loss": 0.0, "step": 334600 }, { "epoch": 72.14, "learning_rate": 6.5521309766222245e-06, "loss": 0.0, "step": 334650 }, { "epoch": 72.15, "learning_rate": 6.547468246550381e-06, "loss": 0.0, "step": 334700 }, { "epoch": 72.16, "learning_rate": 6.542806712923693e-06, "loss": 0.0, "step": 334750 }, { "epoch": 72.17, "learning_rate": 6.538146376401997e-06, "loss": 0.0, "step": 334800 }, { "epoch": 72.18, "learning_rate": 6.533487237644956e-06, "loss": 0.0, "step": 334850 }, { "epoch": 72.19, "learning_rate": 6.5288292973120655e-06, "loss": 0.0, "step": 334900 }, { "epoch": 72.2, "learning_rate": 6.524172556062652e-06, "loss": 0.0, "step": 334950 }, { "epoch": 72.21, "learning_rate": 6.5195170145558706e-06, "loss": 0.0, "step": 335000 }, { "epoch": 72.22, "learning_rate": 6.514862673450707e-06, "loss": 0.0, "step": 335050 }, { "epoch": 72.24, "learning_rate": 6.510209533405977e-06, "loss": 0.0, "step": 335100 }, { "epoch": 72.25, "learning_rate": 6.505557595080323e-06, "loss": 0.0, "step": 335150 }, { "epoch": 72.26, "learning_rate": 6.500906859132235e-06, "loss": 0.0, "step": 335200 }, { "epoch": 72.27, "learning_rate": 6.4962573262200045e-06, "loss": 0.0, "step": 335250 }, { "epoch": 72.28, "learning_rate": 6.491608997001772e-06, "loss": 0.0, "step": 335300 }, { "epoch": 72.29, "learning_rate": 6.486961872135496e-06, "loss": 0.0, "step": 335350 }, { "epoch": 72.3, "learning_rate": 6.482315952278982e-06, "loss": 0.0, "step": 335400 }, { "epoch": 72.31, "learning_rate": 6.477671238089856e-06, "loss": 0.0, "step": 335450 }, { "epoch": 72.32, "learning_rate": 6.473027730225559e-06, "loss": 0.0, "step": 335500 }, { "epoch": 72.33, "learning_rate": 6.468385429343374e-06, "loss": 0.0, "step": 335550 }, { "epoch": 72.34, "learning_rate": 6.463744336100424e-06, "loss": 0.0, "step": 335600 }, { "epoch": 72.35, "learning_rate": 6.459104451153643e-06, "loss": 0.0, "step": 335650 }, { "epoch": 72.36, "learning_rate": 6.454465775159807e-06, "loss": 0.0, "step": 335700 }, { "epoch": 72.38, "learning_rate": 6.449828308775505e-06, "loss": 0.0, "step": 335750 }, { "epoch": 72.39, "learning_rate": 6.445192052657164e-06, "loss": 0.0, "step": 335800 }, { "epoch": 72.4, "learning_rate": 6.44055700746105e-06, "loss": 0.0, "step": 335850 }, { "epoch": 72.41, "learning_rate": 6.435923173843245e-06, "loss": 0.0, "step": 335900 }, { "epoch": 72.42, "learning_rate": 6.431290552459663e-06, "loss": 0.0, "step": 335950 }, { "epoch": 72.43, "learning_rate": 6.426659143966034e-06, "loss": 0.0, "step": 336000 }, { "epoch": 72.44, "learning_rate": 6.422028949017941e-06, "loss": 0.0, "step": 336050 }, { "epoch": 72.45, "learning_rate": 6.41739996827078e-06, "loss": 0.0, "step": 336100 }, { "epoch": 72.46, "learning_rate": 6.412772202379774e-06, "loss": 0.0, "step": 336150 }, { "epoch": 72.47, "learning_rate": 6.40814565199998e-06, "loss": 0.0, "step": 336200 }, { "epoch": 72.48, "learning_rate": 6.403520317786279e-06, "loss": 0.0, "step": 336250 }, { "epoch": 72.49, "learning_rate": 6.398896200393381e-06, "loss": 0.0, "step": 336300 }, { "epoch": 72.5, "learning_rate": 6.394273300475826e-06, "loss": 0.0, "step": 336350 }, { "epoch": 72.52, "learning_rate": 6.3896516186879776e-06, "loss": 0.0, "step": 336400 }, { "epoch": 72.53, "learning_rate": 6.385031155684028e-06, "loss": 0.0, "step": 336450 }, { "epoch": 72.54, "learning_rate": 6.380411912118e-06, "loss": 0.0, "step": 336500 }, { "epoch": 72.55, "learning_rate": 6.3757938886437415e-06, "loss": 0.0, "step": 336550 }, { "epoch": 72.56, "learning_rate": 6.3711770859149284e-06, "loss": 0.0, "step": 336600 }, { "epoch": 72.57, "learning_rate": 6.366561504585062e-06, "loss": 0.0, "step": 336650 }, { "epoch": 72.58, "learning_rate": 6.3619471453074715e-06, "loss": 0.0, "step": 336700 }, { "epoch": 72.59, "learning_rate": 6.357334008735314e-06, "loss": 0.0, "step": 336750 }, { "epoch": 72.6, "learning_rate": 6.352722095521574e-06, "loss": 0.0, "step": 336800 }, { "epoch": 72.61, "learning_rate": 6.348111406319056e-06, "loss": 0.0, "step": 336850 }, { "epoch": 72.62, "learning_rate": 6.343501941780414e-06, "loss": 0.0, "step": 336900 }, { "epoch": 72.63, "learning_rate": 6.3388937025580946e-06, "loss": 0.0, "step": 336950 }, { "epoch": 72.64, "learning_rate": 6.334286689304395e-06, "loss": 0.0, "step": 337000 }, { "epoch": 72.66, "learning_rate": 6.329680902671425e-06, "loss": 0.0, "step": 337050 }, { "epoch": 72.67, "learning_rate": 6.325076343311141e-06, "loss": 0.0, "step": 337100 }, { "epoch": 72.68, "learning_rate": 6.32047301187531e-06, "loss": 0.0, "step": 337150 }, { "epoch": 72.69, "learning_rate": 6.315870909015517e-06, "loss": 0.0, "step": 337200 }, { "epoch": 72.7, "learning_rate": 6.311270035383187e-06, "loss": 0.0, "step": 337250 }, { "epoch": 72.71, "learning_rate": 6.306670391629574e-06, "loss": 0.0, "step": 337300 }, { "epoch": 72.72, "learning_rate": 6.302071978405752e-06, "loss": 0.0, "step": 337350 }, { "epoch": 72.73, "learning_rate": 6.297474796362621e-06, "loss": 0.0, "step": 337400 }, { "epoch": 72.74, "learning_rate": 6.2928788461508964e-06, "loss": 0.0, "step": 337450 }, { "epoch": 72.75, "learning_rate": 6.288284128421131e-06, "loss": 0.0, "step": 337500 }, { "epoch": 72.76, "learning_rate": 6.283690643823712e-06, "loss": 0.0, "step": 337550 }, { "epoch": 72.77, "learning_rate": 6.279098393008834e-06, "loss": 0.0, "step": 337600 }, { "epoch": 72.79, "learning_rate": 6.27450737662653e-06, "loss": 0.0, "step": 337650 }, { "epoch": 72.8, "learning_rate": 6.269917595326639e-06, "loss": 0.0, "step": 337700 }, { "epoch": 72.81, "learning_rate": 6.2653290497588515e-06, "loss": 0.0, "step": 337750 }, { "epoch": 72.82, "learning_rate": 6.2607417405726665e-06, "loss": 0.0, "step": 337800 }, { "epoch": 72.83, "learning_rate": 6.256155668417412e-06, "loss": 0.0, "step": 337850 }, { "epoch": 72.84, "learning_rate": 6.251570833942239e-06, "loss": 0.0, "step": 337900 }, { "epoch": 72.85, "learning_rate": 6.2469872377961295e-06, "loss": 0.0, "step": 337950 }, { "epoch": 72.86, "learning_rate": 6.24240488062788e-06, "loss": 0.0, "step": 338000 }, { "epoch": 72.87, "learning_rate": 6.23782376308612e-06, "loss": 0.0, "step": 338050 }, { "epoch": 72.88, "learning_rate": 6.233243885819303e-06, "loss": 0.0, "step": 338100 }, { "epoch": 72.89, "learning_rate": 6.2286652494757e-06, "loss": 0.0, "step": 338150 }, { "epoch": 72.9, "learning_rate": 6.2240878547034164e-06, "loss": 0.0, "step": 338200 }, { "epoch": 72.91, "learning_rate": 6.219511702150372e-06, "loss": 0.0, "step": 338250 }, { "epoch": 72.93, "learning_rate": 6.214936792464318e-06, "loss": 0.0, "step": 338300 }, { "epoch": 72.94, "learning_rate": 6.210363126292827e-06, "loss": 0.0, "step": 338350 }, { "epoch": 72.95, "learning_rate": 6.205790704283295e-06, "loss": 0.0, "step": 338400 }, { "epoch": 72.96, "learning_rate": 6.201219527082942e-06, "loss": 0.0, "step": 338450 }, { "epoch": 72.97, "learning_rate": 6.196649595338814e-06, "loss": 0.0, "step": 338500 }, { "epoch": 72.98, "learning_rate": 6.192080909697773e-06, "loss": 0.0, "step": 338550 }, { "epoch": 72.99, "learning_rate": 6.187513470806525e-06, "loss": 0.0, "step": 338600 }, { "epoch": 73.0, "learning_rate": 6.182947279311572e-06, "loss": 0.0, "step": 338650 }, { "epoch": 73.01, "learning_rate": 6.1783823358592565e-06, "loss": 0.0, "step": 338700 }, { "epoch": 73.02, "learning_rate": 6.173818641095734e-06, "loss": 0.0, "step": 338750 }, { "epoch": 73.03, "learning_rate": 6.169256195667003e-06, "loss": 0.0, "step": 338800 }, { "epoch": 73.04, "learning_rate": 6.164695000218869e-06, "loss": 0.0, "step": 338850 }, { "epoch": 73.05, "learning_rate": 6.1601350553969555e-06, "loss": 0.0, "step": 338900 }, { "epoch": 73.07, "learning_rate": 6.1555763618467205e-06, "loss": 0.0, "step": 338950 }, { "epoch": 73.08, "learning_rate": 6.151018920213436e-06, "loss": 0.0, "step": 339000 }, { "epoch": 73.09, "learning_rate": 6.1464627311422125e-06, "loss": 0.0, "step": 339050 }, { "epoch": 73.1, "learning_rate": 6.141907795277976e-06, "loss": 0.0, "step": 339100 }, { "epoch": 73.11, "learning_rate": 6.137354113265456e-06, "loss": 0.0, "step": 339150 }, { "epoch": 73.12, "learning_rate": 6.132801685749224e-06, "loss": 0.0, "step": 339200 }, { "epoch": 73.13, "learning_rate": 6.128250513373682e-06, "loss": 0.0, "step": 339250 }, { "epoch": 73.14, "learning_rate": 6.123700596783033e-06, "loss": 0.0, "step": 339300 }, { "epoch": 73.15, "learning_rate": 6.11915193662132e-06, "loss": 0.0, "step": 339350 }, { "epoch": 73.16, "learning_rate": 6.114604533532385e-06, "loss": 0.0, "step": 339400 }, { "epoch": 73.17, "learning_rate": 6.110058388159922e-06, "loss": 0.0, "step": 339450 }, { "epoch": 73.18, "learning_rate": 6.105513501147425e-06, "loss": 0.0, "step": 339500 }, { "epoch": 73.19, "learning_rate": 6.1009698731382205e-06, "loss": 0.0, "step": 339550 }, { "epoch": 73.21, "learning_rate": 6.09642750477545e-06, "loss": 0.0, "step": 339600 }, { "epoch": 73.22, "learning_rate": 6.091886396702083e-06, "loss": 0.0, "step": 339650 }, { "epoch": 73.23, "learning_rate": 6.0873465495609056e-06, "loss": 0.0, "step": 339700 }, { "epoch": 73.24, "learning_rate": 6.0828079639945274e-06, "loss": 0.0, "step": 339750 }, { "epoch": 73.25, "learning_rate": 6.07827064064538e-06, "loss": 0.0, "step": 339800 }, { "epoch": 73.26, "learning_rate": 6.073734580155716e-06, "loss": 0.0, "step": 339850 }, { "epoch": 73.27, "learning_rate": 6.069199783167609e-06, "loss": 0.0, "step": 339900 }, { "epoch": 73.28, "learning_rate": 6.0646662503229515e-06, "loss": 0.0, "step": 339950 }, { "epoch": 73.29, "learning_rate": 6.060133982263464e-06, "loss": 0.0, "step": 340000 }, { "epoch": 73.3, "learning_rate": 6.055602979630679e-06, "loss": 0.0, "step": 340050 }, { "epoch": 73.31, "learning_rate": 6.051073243065957e-06, "loss": 0.0, "step": 340100 }, { "epoch": 73.32, "learning_rate": 6.046544773210474e-06, "loss": 0.0, "step": 340150 }, { "epoch": 73.33, "learning_rate": 6.042017570705232e-06, "loss": 0.0, "step": 340200 }, { "epoch": 73.35, "learning_rate": 6.037491636191042e-06, "loss": 0.0, "step": 340250 }, { "epoch": 73.36, "learning_rate": 6.032966970308563e-06, "loss": 0.0, "step": 340300 }, { "epoch": 73.37, "learning_rate": 6.02844357369824e-06, "loss": 0.0, "step": 340350 }, { "epoch": 73.38, "learning_rate": 6.023921447000356e-06, "loss": 0.0, "step": 340400 }, { "epoch": 73.39, "learning_rate": 6.019400590855011e-06, "loss": 0.0, "step": 340450 }, { "epoch": 73.4, "learning_rate": 6.0148810059021335e-06, "loss": 0.0, "step": 340500 }, { "epoch": 73.41, "learning_rate": 6.010362692781469e-06, "loss": 0.0, "step": 340550 }, { "epoch": 73.42, "learning_rate": 6.005845652132566e-06, "loss": 0.0, "step": 340600 }, { "epoch": 73.43, "learning_rate": 6.0013298845948094e-06, "loss": 0.0, "step": 340650 }, { "epoch": 73.44, "learning_rate": 5.9968153908073995e-06, "loss": 0.0, "step": 340700 }, { "epoch": 73.45, "learning_rate": 5.992302171409364e-06, "loss": 0.0, "step": 340750 }, { "epoch": 73.46, "learning_rate": 5.9877902270395445e-06, "loss": 0.0, "step": 340800 }, { "epoch": 73.47, "learning_rate": 5.983279558336592e-06, "loss": 0.0, "step": 340850 }, { "epoch": 73.49, "learning_rate": 5.978770165938985e-06, "loss": 0.0, "step": 340900 }, { "epoch": 73.5, "learning_rate": 5.97426205048503e-06, "loss": 0.0, "step": 340950 }, { "epoch": 73.51, "learning_rate": 5.969755212612843e-06, "loss": 0.0, "step": 341000 }, { "epoch": 73.52, "learning_rate": 5.965249652960367e-06, "loss": 0.0, "step": 341050 }, { "epoch": 73.53, "learning_rate": 5.960745372165341e-06, "loss": 0.0, "step": 341100 }, { "epoch": 73.54, "learning_rate": 5.956242370865356e-06, "loss": 0.0, "step": 341150 }, { "epoch": 73.55, "learning_rate": 5.9517406496978e-06, "loss": 0.0, "step": 341200 }, { "epoch": 73.56, "learning_rate": 5.947240209299887e-06, "loss": 0.0, "step": 341250 }, { "epoch": 73.57, "learning_rate": 5.942741050308651e-06, "loss": 0.0, "step": 341300 }, { "epoch": 73.58, "learning_rate": 5.938243173360937e-06, "loss": 0.0, "step": 341350 }, { "epoch": 73.59, "learning_rate": 5.933746579093418e-06, "loss": 0.0, "step": 341400 }, { "epoch": 73.6, "learning_rate": 5.929251268142579e-06, "loss": 0.0, "step": 341450 }, { "epoch": 73.62, "learning_rate": 5.924757241144727e-06, "loss": 0.0, "step": 341500 }, { "epoch": 73.63, "learning_rate": 5.920264498735984e-06, "loss": 0.0, "step": 341550 }, { "epoch": 73.64, "learning_rate": 5.915773041552296e-06, "loss": 0.0, "step": 341600 }, { "epoch": 73.65, "learning_rate": 5.911282870229418e-06, "loss": 0.0, "step": 341650 }, { "epoch": 73.66, "learning_rate": 5.906793985402931e-06, "loss": 0.0, "step": 341700 }, { "epoch": 73.67, "learning_rate": 5.90230638770823e-06, "loss": 0.0, "step": 341750 }, { "epoch": 73.68, "learning_rate": 5.89782007778053e-06, "loss": 0.0, "step": 341800 }, { "epoch": 73.69, "learning_rate": 5.89333505625486e-06, "loss": 0.0, "step": 341850 }, { "epoch": 73.7, "learning_rate": 5.888851323766072e-06, "loss": 0.0, "step": 341900 }, { "epoch": 73.71, "learning_rate": 5.8843688809488265e-06, "loss": 0.0, "step": 341950 }, { "epoch": 73.72, "learning_rate": 5.879887728437622e-06, "loss": 0.0, "step": 342000 }, { "epoch": 73.73, "learning_rate": 5.875407866866745e-06, "loss": 0.0, "step": 342050 }, { "epoch": 73.74, "learning_rate": 5.870929296870321e-06, "loss": 0.0, "step": 342100 }, { "epoch": 73.76, "learning_rate": 5.866452019082276e-06, "loss": 0.0, "step": 342150 }, { "epoch": 73.77, "learning_rate": 5.86197603413638e-06, "loss": 0.0, "step": 342200 }, { "epoch": 73.78, "learning_rate": 5.857501342666199e-06, "loss": 0.0, "step": 342250 }, { "epoch": 73.79, "learning_rate": 5.85302794530511e-06, "loss": 0.0, "step": 342300 }, { "epoch": 73.8, "learning_rate": 5.848555842686322e-06, "loss": 0.0, "step": 342350 }, { "epoch": 73.81, "learning_rate": 5.84408503544285e-06, "loss": 0.0, "step": 342400 }, { "epoch": 73.82, "learning_rate": 5.839615524207542e-06, "loss": 0.0, "step": 342450 }, { "epoch": 73.83, "learning_rate": 5.835147309613052e-06, "loss": 0.0, "step": 342500 }, { "epoch": 73.84, "learning_rate": 5.83068039229184e-06, "loss": 0.0, "step": 342550 }, { "epoch": 73.85, "learning_rate": 5.82621477287619e-06, "loss": 0.0, "step": 342600 }, { "epoch": 73.86, "learning_rate": 5.821750451998217e-06, "loss": 0.0, "step": 342650 }, { "epoch": 73.87, "learning_rate": 5.817287430289837e-06, "loss": 0.0, "step": 342700 }, { "epoch": 73.88, "learning_rate": 5.812825708382785e-06, "loss": 0.0, "step": 342750 }, { "epoch": 73.9, "learning_rate": 5.808365286908601e-06, "loss": 0.0, "step": 342800 }, { "epoch": 73.91, "learning_rate": 5.803906166498666e-06, "loss": 0.0, "step": 342850 }, { "epoch": 73.92, "learning_rate": 5.799448347784158e-06, "loss": 0.0, "step": 342900 }, { "epoch": 73.93, "learning_rate": 5.794991831396074e-06, "loss": 0.0, "step": 342950 }, { "epoch": 73.94, "learning_rate": 5.790536617965237e-06, "loss": 0.0, "step": 343000 }, { "epoch": 73.95, "learning_rate": 5.786082708122256e-06, "loss": 0.0, "step": 343050 }, { "epoch": 73.96, "learning_rate": 5.781630102497598e-06, "loss": 0.0, "step": 343100 }, { "epoch": 73.97, "learning_rate": 5.777178801721514e-06, "loss": 0.0, "step": 343150 }, { "epoch": 73.98, "learning_rate": 5.772728806424079e-06, "loss": 0.0, "step": 343200 }, { "epoch": 73.99, "learning_rate": 5.76828011723519e-06, "loss": 0.0, "step": 343250 }, { "epoch": 74.0, "learning_rate": 5.763832734784548e-06, "loss": 0.0, "step": 343300 }, { "epoch": 74.01, "learning_rate": 5.759386659701676e-06, "loss": 0.0, "step": 343350 }, { "epoch": 74.02, "learning_rate": 5.754941892615911e-06, "loss": 0.0, "step": 343400 }, { "epoch": 74.04, "learning_rate": 5.750498434156403e-06, "loss": 0.0, "step": 343450 }, { "epoch": 74.05, "learning_rate": 5.746056284952119e-06, "loss": 0.0, "step": 343500 }, { "epoch": 74.06, "learning_rate": 5.74161544563184e-06, "loss": 0.0, "step": 343550 }, { "epoch": 74.07, "learning_rate": 5.737175916824159e-06, "loss": 0.0, "step": 343600 }, { "epoch": 74.08, "learning_rate": 5.73273769915748e-06, "loss": 0.0, "step": 343650 }, { "epoch": 74.09, "learning_rate": 5.728300793260048e-06, "loss": 0.0, "step": 343700 }, { "epoch": 74.1, "learning_rate": 5.723865199759881e-06, "loss": 0.0, "step": 343750 }, { "epoch": 74.11, "learning_rate": 5.719430919284837e-06, "loss": 0.0, "step": 343800 }, { "epoch": 74.12, "learning_rate": 5.714997952462584e-06, "loss": 0.0, "step": 343850 }, { "epoch": 74.13, "learning_rate": 5.710566299920599e-06, "loss": 0.0, "step": 343900 }, { "epoch": 74.14, "learning_rate": 5.706135962286189e-06, "loss": 0.0, "step": 343950 }, { "epoch": 74.15, "learning_rate": 5.70170694018645e-06, "loss": 0.0, "step": 344000 }, { "epoch": 74.16, "learning_rate": 5.69727923424831e-06, "loss": 0.0, "step": 344050 }, { "epoch": 74.18, "learning_rate": 5.6928528450984965e-06, "loss": 0.0, "step": 344100 }, { "epoch": 74.19, "learning_rate": 5.688427773363571e-06, "loss": 0.0, "step": 344150 }, { "epoch": 74.2, "learning_rate": 5.684004019669901e-06, "loss": 0.0, "step": 344200 }, { "epoch": 74.21, "learning_rate": 5.679581584643647e-06, "loss": 0.0, "step": 344250 }, { "epoch": 74.22, "learning_rate": 5.675160468910801e-06, "loss": 0.0, "step": 344300 }, { "epoch": 74.23, "learning_rate": 5.670740673097179e-06, "loss": 0.0, "step": 344350 }, { "epoch": 74.24, "learning_rate": 5.666322197828388e-06, "loss": 0.0, "step": 344400 }, { "epoch": 74.25, "learning_rate": 5.661905043729868e-06, "loss": 0.0, "step": 344450 }, { "epoch": 74.26, "learning_rate": 5.6574892114268415e-06, "loss": 0.0, "step": 344500 }, { "epoch": 74.27, "learning_rate": 5.653074701544383e-06, "loss": 0.0, "step": 344550 }, { "epoch": 74.28, "learning_rate": 5.648661514707351e-06, "loss": 0.0, "step": 344600 }, { "epoch": 74.29, "learning_rate": 5.6442496515404315e-06, "loss": 0.0, "step": 344650 }, { "epoch": 74.3, "learning_rate": 5.639839112668115e-06, "loss": 0.0, "step": 344700 }, { "epoch": 74.32, "learning_rate": 5.635429898714709e-06, "loss": 0.0, "step": 344750 }, { "epoch": 74.33, "learning_rate": 5.631022010304334e-06, "loss": 0.0, "step": 344800 }, { "epoch": 74.34, "learning_rate": 5.626615448060916e-06, "loss": 0.0, "step": 344850 }, { "epoch": 74.35, "learning_rate": 5.622210212608202e-06, "loss": 0.0, "step": 344900 }, { "epoch": 74.36, "learning_rate": 5.617806304569747e-06, "loss": 0.0, "step": 344950 }, { "epoch": 74.37, "learning_rate": 5.613403724568917e-06, "loss": 0.0, "step": 345000 }, { "epoch": 74.38, "learning_rate": 5.609002473228895e-06, "loss": 0.0, "step": 345050 }, { "epoch": 74.39, "learning_rate": 5.6046025511726695e-06, "loss": 0.0, "step": 345100 }, { "epoch": 74.4, "learning_rate": 5.600203959023043e-06, "loss": 0.0, "step": 345150 }, { "epoch": 74.41, "learning_rate": 5.595806697402636e-06, "loss": 0.0, "step": 345200 }, { "epoch": 74.42, "learning_rate": 5.59141076693387e-06, "loss": 0.0, "step": 345250 }, { "epoch": 74.43, "learning_rate": 5.5870161682389855e-06, "loss": 0.0, "step": 345300 }, { "epoch": 74.44, "learning_rate": 5.582622901940025e-06, "loss": 0.0, "step": 345350 }, { "epoch": 74.46, "learning_rate": 5.578230968658869e-06, "loss": 0.0, "step": 345400 }, { "epoch": 74.47, "learning_rate": 5.573840369017172e-06, "loss": 0.0, "step": 345450 }, { "epoch": 74.48, "learning_rate": 5.569451103636425e-06, "loss": 0.0, "step": 345500 }, { "epoch": 74.49, "learning_rate": 5.565063173137921e-06, "loss": 0.0, "step": 345550 }, { "epoch": 74.5, "learning_rate": 5.560676578142762e-06, "loss": 0.0, "step": 345600 }, { "epoch": 74.51, "learning_rate": 5.556291319271878e-06, "loss": 0.0, "step": 345650 }, { "epoch": 74.52, "learning_rate": 5.551907397145985e-06, "loss": 0.0, "step": 345700 }, { "epoch": 74.53, "learning_rate": 5.547524812385625e-06, "loss": 0.0, "step": 345750 }, { "epoch": 74.54, "learning_rate": 5.543143565611142e-06, "loss": 0.0, "step": 345800 }, { "epoch": 74.55, "learning_rate": 5.538763657442705e-06, "loss": 0.0, "step": 345850 }, { "epoch": 74.56, "learning_rate": 5.534385088500283e-06, "loss": 0.0, "step": 345900 }, { "epoch": 74.57, "learning_rate": 5.530007859403659e-06, "loss": 0.0, "step": 345950 }, { "epoch": 74.59, "learning_rate": 5.525631970772409e-06, "loss": 0.0, "step": 346000 }, { "epoch": 74.6, "learning_rate": 5.5212574232259505e-06, "loss": 0.0, "step": 346050 }, { "epoch": 74.61, "learning_rate": 5.516884217383491e-06, "loss": 0.0, "step": 346100 }, { "epoch": 74.62, "learning_rate": 5.512512353864051e-06, "loss": 0.0, "step": 346150 }, { "epoch": 74.63, "learning_rate": 5.5081418332864634e-06, "loss": 0.0, "step": 346200 }, { "epoch": 74.64, "learning_rate": 5.503772656269367e-06, "loss": 0.0, "step": 346250 }, { "epoch": 74.65, "learning_rate": 5.4994048234312165e-06, "loss": 0.0, "step": 346300 }, { "epoch": 74.66, "learning_rate": 5.495038335390272e-06, "loss": 0.0, "step": 346350 }, { "epoch": 74.67, "learning_rate": 5.490673192764603e-06, "loss": 0.0, "step": 346400 }, { "epoch": 74.68, "learning_rate": 5.486309396172093e-06, "loss": 0.0, "step": 346450 }, { "epoch": 74.69, "learning_rate": 5.481946946230428e-06, "loss": 0.0, "step": 346500 }, { "epoch": 74.7, "learning_rate": 5.477585843557111e-06, "loss": 0.0, "step": 346550 }, { "epoch": 74.71, "learning_rate": 5.473226088769448e-06, "loss": 0.0, "step": 346600 }, { "epoch": 74.73, "learning_rate": 5.468867682484558e-06, "loss": 0.0, "step": 346650 }, { "epoch": 74.74, "learning_rate": 5.464510625319367e-06, "loss": 0.0, "step": 346700 }, { "epoch": 74.75, "learning_rate": 5.460154917890611e-06, "loss": 0.0, "step": 346750 }, { "epoch": 74.76, "learning_rate": 5.455800560814835e-06, "loss": 0.0, "step": 346800 }, { "epoch": 74.77, "learning_rate": 5.451447554708394e-06, "loss": 0.0, "step": 346850 }, { "epoch": 74.78, "learning_rate": 5.447095900187449e-06, "loss": 0.0, "step": 346900 }, { "epoch": 74.79, "learning_rate": 5.442745597867971e-06, "loss": 0.0, "step": 346950 }, { "epoch": 74.8, "learning_rate": 5.438396648365741e-06, "loss": 0.0, "step": 347000 }, { "epoch": 74.81, "learning_rate": 5.4340490522963426e-06, "loss": 0.0, "step": 347050 }, { "epoch": 74.82, "learning_rate": 5.4297028102751854e-06, "loss": 0.0, "step": 347100 }, { "epoch": 74.83, "learning_rate": 5.425357922917461e-06, "loss": 0.0, "step": 347150 }, { "epoch": 74.84, "learning_rate": 5.421014390838188e-06, "loss": 0.0, "step": 347200 }, { "epoch": 74.85, "learning_rate": 5.416672214652187e-06, "loss": 0.0, "step": 347250 }, { "epoch": 74.87, "learning_rate": 5.412331394974083e-06, "loss": 0.0, "step": 347300 }, { "epoch": 74.88, "learning_rate": 5.407991932418321e-06, "loss": 0.0, "step": 347350 }, { "epoch": 74.89, "learning_rate": 5.403653827599151e-06, "loss": 0.0, "step": 347400 }, { "epoch": 74.9, "learning_rate": 5.399317081130614e-06, "loss": 0.0, "step": 347450 }, { "epoch": 74.91, "learning_rate": 5.39498169362657e-06, "loss": 0.0, "step": 347500 }, { "epoch": 74.92, "learning_rate": 5.3906476657007e-06, "loss": 0.0, "step": 347550 }, { "epoch": 74.93, "learning_rate": 5.386314997966472e-06, "loss": 0.0, "step": 347600 }, { "epoch": 74.94, "learning_rate": 5.381983691037177e-06, "loss": 0.0, "step": 347650 }, { "epoch": 74.95, "learning_rate": 5.377653745525889e-06, "loss": 0.0, "step": 347700 }, { "epoch": 74.96, "learning_rate": 5.373325162045526e-06, "loss": 0.0, "step": 347750 }, { "epoch": 74.97, "learning_rate": 5.368997941208782e-06, "loss": 0.0, "step": 347800 }, { "epoch": 74.98, "learning_rate": 5.3646720836281725e-06, "loss": 0.0, "step": 347850 }, { "epoch": 74.99, "learning_rate": 5.3603475899160245e-06, "loss": 0.0, "step": 347900 }, { "epoch": 75.01, "learning_rate": 5.356024460684446e-06, "loss": 0.0, "step": 347950 }, { "epoch": 75.02, "learning_rate": 5.351702696545389e-06, "loss": 0.0, "step": 348000 }, { "epoch": 75.03, "learning_rate": 5.347382298110584e-06, "loss": 0.0, "step": 348050 }, { "epoch": 75.04, "learning_rate": 5.343063265991581e-06, "loss": 0.0, "step": 348100 }, { "epoch": 75.05, "learning_rate": 5.338745600799736e-06, "loss": 0.0, "step": 348150 }, { "epoch": 75.06, "learning_rate": 5.334429303146204e-06, "loss": 0.0, "step": 348200 }, { "epoch": 75.07, "learning_rate": 5.330114373641953e-06, "loss": 0.0, "step": 348250 }, { "epoch": 75.08, "learning_rate": 5.325800812897757e-06, "loss": 0.0, "step": 348300 }, { "epoch": 75.09, "learning_rate": 5.321488621524197e-06, "loss": 0.0, "step": 348350 }, { "epoch": 75.1, "learning_rate": 5.317177800131653e-06, "loss": 0.0, "step": 348400 }, { "epoch": 75.11, "learning_rate": 5.312868349330321e-06, "loss": 0.0, "step": 348450 }, { "epoch": 75.12, "learning_rate": 5.308560269730193e-06, "loss": 0.0, "step": 348500 }, { "epoch": 75.13, "learning_rate": 5.304253561941073e-06, "loss": 0.0, "step": 348550 }, { "epoch": 75.15, "learning_rate": 5.2999482265725825e-06, "loss": 0.0, "step": 348600 }, { "epoch": 75.16, "learning_rate": 5.29564426423412e-06, "loss": 0.0, "step": 348650 }, { "epoch": 75.17, "learning_rate": 5.291341675534913e-06, "loss": 0.0, "step": 348700 }, { "epoch": 75.18, "learning_rate": 5.287040461083987e-06, "loss": 0.0, "step": 348750 }, { "epoch": 75.19, "learning_rate": 5.282740621490169e-06, "loss": 0.0, "step": 348800 }, { "epoch": 75.2, "learning_rate": 5.278442157362109e-06, "loss": 0.0, "step": 348850 }, { "epoch": 75.21, "learning_rate": 5.274145069308237e-06, "loss": 0.0, "step": 348900 }, { "epoch": 75.22, "learning_rate": 5.269849357936801e-06, "loss": 0.0, "step": 348950 }, { "epoch": 75.23, "learning_rate": 5.2655550238558514e-06, "loss": 0.0, "step": 349000 }, { "epoch": 75.24, "learning_rate": 5.261262067673256e-06, "loss": 0.0, "step": 349050 }, { "epoch": 75.25, "learning_rate": 5.256970489996676e-06, "loss": 0.0, "step": 349100 }, { "epoch": 75.26, "learning_rate": 5.252680291433572e-06, "loss": 0.0, "step": 349150 }, { "epoch": 75.27, "learning_rate": 5.24839147259121e-06, "loss": 0.0, "step": 349200 }, { "epoch": 75.29, "learning_rate": 5.244104034076684e-06, "loss": 0.0, "step": 349250 }, { "epoch": 75.3, "learning_rate": 5.239817976496864e-06, "loss": 0.0, "step": 349300 }, { "epoch": 75.31, "learning_rate": 5.235533300458446e-06, "loss": 0.0, "step": 349350 }, { "epoch": 75.32, "learning_rate": 5.231250006567903e-06, "loss": 0.0, "step": 349400 }, { "epoch": 75.33, "learning_rate": 5.226968095431546e-06, "loss": 0.0, "step": 349450 }, { "epoch": 75.34, "learning_rate": 5.222687567655467e-06, "loss": 0.0, "step": 349500 }, { "epoch": 75.35, "learning_rate": 5.218408423845571e-06, "loss": 0.0, "step": 349550 }, { "epoch": 75.36, "learning_rate": 5.214130664607571e-06, "loss": 0.0, "step": 349600 }, { "epoch": 75.37, "learning_rate": 5.209854290546961e-06, "loss": 0.0, "step": 349650 }, { "epoch": 75.38, "learning_rate": 5.205579302269073e-06, "loss": 0.0, "step": 349700 }, { "epoch": 75.39, "learning_rate": 5.201305700379022e-06, "loss": 0.0, "step": 349750 }, { "epoch": 75.4, "learning_rate": 5.1970334854817294e-06, "loss": 0.0, "step": 349800 }, { "epoch": 75.41, "learning_rate": 5.192762658181922e-06, "loss": 0.0, "step": 349850 }, { "epoch": 75.43, "learning_rate": 5.188493219084131e-06, "loss": 0.0, "step": 349900 }, { "epoch": 75.44, "learning_rate": 5.184225168792688e-06, "loss": 0.0, "step": 349950 }, { "epoch": 75.45, "learning_rate": 5.179958507911734e-06, "loss": 0.0, "step": 350000 }, { "epoch": 75.46, "learning_rate": 5.175693237045205e-06, "loss": 0.0, "step": 350050 }, { "epoch": 75.47, "learning_rate": 5.171429356796848e-06, "loss": 0.0, "step": 350100 }, { "epoch": 75.48, "learning_rate": 5.1671668677702075e-06, "loss": 0.0, "step": 350150 }, { "epoch": 75.49, "learning_rate": 5.162905770568633e-06, "loss": 0.0, "step": 350200 }, { "epoch": 75.5, "learning_rate": 5.158646065795276e-06, "loss": 0.0, "step": 350250 }, { "epoch": 75.51, "learning_rate": 5.154387754053105e-06, "loss": 0.0, "step": 350300 }, { "epoch": 75.52, "learning_rate": 5.150130835944862e-06, "loss": 0.0, "step": 350350 }, { "epoch": 75.53, "learning_rate": 5.145875312073116e-06, "loss": 0.0, "step": 350400 }, { "epoch": 75.54, "learning_rate": 5.141621183040231e-06, "loss": 0.0, "step": 350450 }, { "epoch": 75.56, "learning_rate": 5.137368449448367e-06, "loss": 0.0, "step": 350500 }, { "epoch": 75.57, "learning_rate": 5.133117111899509e-06, "loss": 0.0, "step": 350550 }, { "epoch": 75.58, "learning_rate": 5.128867170995413e-06, "loss": 0.0, "step": 350600 }, { "epoch": 75.59, "learning_rate": 5.124618627337659e-06, "loss": 0.0, "step": 350650 }, { "epoch": 75.6, "learning_rate": 5.120371481527615e-06, "loss": 0.0, "step": 350700 }, { "epoch": 75.61, "learning_rate": 5.116125734166471e-06, "loss": 0.0, "step": 350750 }, { "epoch": 75.62, "learning_rate": 5.1118813858552075e-06, "loss": 0.0, "step": 350800 }, { "epoch": 75.63, "learning_rate": 5.107638437194597e-06, "loss": 0.0, "step": 350850 }, { "epoch": 75.64, "learning_rate": 5.103396888785223e-06, "loss": 0.0, "step": 350900 }, { "epoch": 75.65, "learning_rate": 5.09915674122748e-06, "loss": 0.0, "step": 350950 }, { "epoch": 75.66, "learning_rate": 5.094917995121552e-06, "loss": 0.0, "step": 351000 }, { "epoch": 75.67, "learning_rate": 5.090680651067433e-06, "loss": 0.0, "step": 351050 }, { "epoch": 75.68, "learning_rate": 5.0864447096648976e-06, "loss": 0.0, "step": 351100 }, { "epoch": 75.7, "learning_rate": 5.082210171513554e-06, "loss": 0.0, "step": 351150 }, { "epoch": 75.71, "learning_rate": 5.077977037212792e-06, "loss": 0.0, "step": 351200 }, { "epoch": 75.72, "learning_rate": 5.0737453073618046e-06, "loss": 0.0, "step": 351250 }, { "epoch": 75.73, "learning_rate": 5.069514982559591e-06, "loss": 0.0, "step": 351300 }, { "epoch": 75.74, "learning_rate": 5.065286063404937e-06, "loss": 0.0, "step": 351350 }, { "epoch": 75.75, "learning_rate": 5.061058550496453e-06, "loss": 0.0, "step": 351400 }, { "epoch": 75.76, "learning_rate": 5.056832444432536e-06, "loss": 0.0, "step": 351450 }, { "epoch": 75.77, "learning_rate": 5.052607745811383e-06, "loss": 0.0, "step": 351500 }, { "epoch": 75.78, "learning_rate": 5.048384455230996e-06, "loss": 0.0, "step": 351550 }, { "epoch": 75.79, "learning_rate": 5.044162573289177e-06, "loss": 0.0, "step": 351600 }, { "epoch": 75.8, "learning_rate": 5.0399421005835266e-06, "loss": 0.0, "step": 351650 }, { "epoch": 75.81, "learning_rate": 5.035723037711448e-06, "loss": 0.0, "step": 351700 }, { "epoch": 75.82, "learning_rate": 5.031505385270144e-06, "loss": 0.0, "step": 351750 }, { "epoch": 75.84, "learning_rate": 5.027289143856619e-06, "loss": 0.0, "step": 351800 }, { "epoch": 75.85, "learning_rate": 5.023074314067676e-06, "loss": 0.0, "step": 351850 }, { "epoch": 75.86, "learning_rate": 5.018860896499918e-06, "loss": 0.0, "step": 351900 }, { "epoch": 75.87, "learning_rate": 5.014648891749745e-06, "loss": 0.0, "step": 351950 }, { "epoch": 75.88, "learning_rate": 5.010438300413377e-06, "loss": 0.0, "step": 352000 }, { "epoch": 75.89, "learning_rate": 5.0062291230868e-06, "loss": 0.0, "step": 352050 }, { "epoch": 75.9, "learning_rate": 5.0020213603658235e-06, "loss": 0.0, "step": 352100 }, { "epoch": 75.91, "learning_rate": 4.9978150128460526e-06, "loss": 0.0, "step": 352150 }, { "epoch": 75.92, "learning_rate": 4.993610081122885e-06, "loss": 0.0, "step": 352200 }, { "epoch": 75.93, "learning_rate": 4.989406565791536e-06, "loss": 0.0, "step": 352250 }, { "epoch": 75.94, "learning_rate": 4.9852044674469974e-06, "loss": 0.0, "step": 352300 }, { "epoch": 75.95, "learning_rate": 4.9810037866840725e-06, "loss": 0.0, "step": 352350 }, { "epoch": 75.96, "learning_rate": 4.976804524097357e-06, "loss": 0.0, "step": 352400 }, { "epoch": 75.98, "learning_rate": 4.972606680281264e-06, "loss": 0.0, "step": 352450 }, { "epoch": 75.99, "learning_rate": 4.968410255829991e-06, "loss": 0.0, "step": 352500 }, { "epoch": 76.0, "learning_rate": 4.964215251337526e-06, "loss": 0.0, "step": 352550 }, { "epoch": 76.01, "learning_rate": 4.9600216673976674e-06, "loss": 0.0, "step": 352600 }, { "epoch": 76.02, "learning_rate": 4.955829504604022e-06, "loss": 0.0, "step": 352650 }, { "epoch": 76.03, "learning_rate": 4.95163876354998e-06, "loss": 0.0, "step": 352700 }, { "epoch": 76.04, "learning_rate": 4.947449444828741e-06, "loss": 0.0, "step": 352750 }, { "epoch": 76.05, "learning_rate": 4.943261549033287e-06, "loss": 0.0, "step": 352800 }, { "epoch": 76.06, "learning_rate": 4.939075076756409e-06, "loss": 0.0, "step": 352850 }, { "epoch": 76.07, "learning_rate": 4.934890028590707e-06, "loss": 0.0, "step": 352900 }, { "epoch": 76.08, "learning_rate": 4.9307064051285655e-06, "loss": 0.0, "step": 352950 }, { "epoch": 76.09, "learning_rate": 4.926524206962174e-06, "loss": 0.0, "step": 353000 }, { "epoch": 76.1, "learning_rate": 4.922343434683505e-06, "loss": 0.0, "step": 353050 }, { "epoch": 76.12, "learning_rate": 4.918164088884353e-06, "loss": 0.0, "step": 353100 }, { "epoch": 76.13, "learning_rate": 4.913986170156298e-06, "loss": 0.0, "step": 353150 }, { "epoch": 76.14, "learning_rate": 4.9098096790907164e-06, "loss": 0.0, "step": 353200 }, { "epoch": 76.15, "learning_rate": 4.905634616278787e-06, "loss": 0.0, "step": 353250 }, { "epoch": 76.16, "learning_rate": 4.901460982311483e-06, "loss": 0.0, "step": 353300 }, { "epoch": 76.17, "learning_rate": 4.897288777779578e-06, "loss": 0.0, "step": 353350 }, { "epoch": 76.18, "learning_rate": 4.893118003273642e-06, "loss": 0.0, "step": 353400 }, { "epoch": 76.19, "learning_rate": 4.888948659384041e-06, "loss": 0.0, "step": 353450 }, { "epoch": 76.2, "learning_rate": 4.884780746700944e-06, "loss": 0.0, "step": 353500 }, { "epoch": 76.21, "learning_rate": 4.880614265814313e-06, "loss": 0.0, "step": 353550 }, { "epoch": 76.22, "learning_rate": 4.876449217313905e-06, "loss": 0.0, "step": 353600 }, { "epoch": 76.23, "learning_rate": 4.872285601789279e-06, "loss": 0.0, "step": 353650 }, { "epoch": 76.24, "learning_rate": 4.868123419829791e-06, "loss": 0.0, "step": 353700 }, { "epoch": 76.26, "learning_rate": 4.8639626720245915e-06, "loss": 0.0, "step": 353750 }, { "epoch": 76.27, "learning_rate": 4.859803358962628e-06, "loss": 0.0, "step": 353800 }, { "epoch": 76.28, "learning_rate": 4.855645481232649e-06, "loss": 0.0, "step": 353850 }, { "epoch": 76.29, "learning_rate": 4.851489039423188e-06, "loss": 0.0, "step": 353900 }, { "epoch": 76.3, "learning_rate": 4.847334034122601e-06, "loss": 0.0, "step": 353950 }, { "epoch": 76.31, "learning_rate": 4.84318046591901e-06, "loss": 0.0, "step": 354000 }, { "epoch": 76.32, "learning_rate": 4.83902833540035e-06, "loss": 0.0, "step": 354050 }, { "epoch": 76.33, "learning_rate": 4.834877643154345e-06, "loss": 0.0, "step": 354100 }, { "epoch": 76.34, "learning_rate": 4.830728389768531e-06, "loss": 0.0, "step": 354150 }, { "epoch": 76.35, "learning_rate": 4.826580575830231e-06, "loss": 0.0, "step": 354200 }, { "epoch": 76.36, "learning_rate": 4.82243420192655e-06, "loss": 0.0, "step": 354250 }, { "epoch": 76.37, "learning_rate": 4.818289268644404e-06, "loss": 0.0, "step": 354300 }, { "epoch": 76.38, "learning_rate": 4.8141457765705125e-06, "loss": 0.0, "step": 354350 }, { "epoch": 76.4, "learning_rate": 4.810003726291376e-06, "loss": 0.0, "step": 354400 }, { "epoch": 76.41, "learning_rate": 4.805863118393303e-06, "loss": 0.0, "step": 354450 }, { "epoch": 76.42, "learning_rate": 4.801723953462379e-06, "loss": 0.0, "step": 354500 }, { "epoch": 76.43, "learning_rate": 4.797586232084498e-06, "loss": 0.0, "step": 354550 }, { "epoch": 76.44, "learning_rate": 4.793449954845361e-06, "loss": 0.0, "step": 354600 }, { "epoch": 76.45, "learning_rate": 4.7893151223304475e-06, "loss": 0.0, "step": 354650 }, { "epoch": 76.46, "learning_rate": 4.78518173512504e-06, "loss": 0.0, "step": 354700 }, { "epoch": 76.47, "learning_rate": 4.781049793814204e-06, "loss": 0.0, "step": 354750 }, { "epoch": 76.48, "learning_rate": 4.776919298982819e-06, "loss": 0.0, "step": 354800 }, { "epoch": 76.49, "learning_rate": 4.772790251215554e-06, "loss": 0.0, "step": 354850 }, { "epoch": 76.5, "learning_rate": 4.768662651096866e-06, "loss": 0.0, "step": 354900 }, { "epoch": 76.51, "learning_rate": 4.7645364992110105e-06, "loss": 0.0, "step": 354950 }, { "epoch": 76.53, "learning_rate": 4.760411796142041e-06, "loss": 0.0, "step": 355000 }, { "epoch": 76.54, "learning_rate": 4.756288542473805e-06, "loss": 0.0, "step": 355050 }, { "epoch": 76.55, "learning_rate": 4.7521667387899415e-06, "loss": 0.0, "step": 355100 }, { "epoch": 76.56, "learning_rate": 4.748046385673887e-06, "loss": 0.0, "step": 355150 }, { "epoch": 76.57, "learning_rate": 4.743927483708874e-06, "loss": 0.0, "step": 355200 }, { "epoch": 76.58, "learning_rate": 4.739810033477927e-06, "loss": 0.0, "step": 355250 }, { "epoch": 76.59, "learning_rate": 4.735694035563866e-06, "loss": 0.0, "step": 355300 }, { "epoch": 76.6, "learning_rate": 4.731579490549303e-06, "loss": 0.0, "step": 355350 }, { "epoch": 76.61, "learning_rate": 4.72746639901665e-06, "loss": 0.0, "step": 355400 }, { "epoch": 76.62, "learning_rate": 4.723354761548109e-06, "loss": 0.0, "step": 355450 }, { "epoch": 76.63, "learning_rate": 4.719244578725675e-06, "loss": 0.0, "step": 355500 }, { "epoch": 76.64, "learning_rate": 4.7151358511311416e-06, "loss": 0.0, "step": 355550 }, { "epoch": 76.65, "learning_rate": 4.711028579346087e-06, "loss": 0.0, "step": 355600 }, { "epoch": 76.67, "learning_rate": 4.706922763951908e-06, "loss": 0.0, "step": 355650 }, { "epoch": 76.68, "learning_rate": 4.7028184055297605e-06, "loss": 0.0, "step": 355700 }, { "epoch": 76.69, "learning_rate": 4.698715504660615e-06, "loss": 0.0, "step": 355750 }, { "epoch": 76.7, "learning_rate": 4.694614061925231e-06, "loss": 0.0, "step": 355800 }, { "epoch": 76.71, "learning_rate": 4.690514077904169e-06, "loss": 0.0, "step": 355850 }, { "epoch": 76.72, "learning_rate": 4.686415553177778e-06, "loss": 0.0, "step": 355900 }, { "epoch": 76.73, "learning_rate": 4.682318488326189e-06, "loss": 0.0, "step": 355950 }, { "epoch": 76.74, "learning_rate": 4.678222883929335e-06, "loss": 0.0, "step": 356000 }, { "epoch": 76.75, "learning_rate": 4.674128740566955e-06, "loss": 0.0, "step": 356050 }, { "epoch": 76.76, "learning_rate": 4.6700360588185645e-06, "loss": 0.0, "step": 356100 }, { "epoch": 76.77, "learning_rate": 4.6659448392634825e-06, "loss": 0.0, "step": 356150 }, { "epoch": 76.78, "learning_rate": 4.661855082480805e-06, "loss": 0.0, "step": 356200 }, { "epoch": 76.79, "learning_rate": 4.657766789049434e-06, "loss": 0.0, "step": 356250 }, { "epoch": 76.81, "learning_rate": 4.653679959548069e-06, "loss": 0.0, "step": 356300 }, { "epoch": 76.82, "learning_rate": 4.649594594555193e-06, "loss": 0.0, "step": 356350 }, { "epoch": 76.83, "learning_rate": 4.64551069464909e-06, "loss": 0.0, "step": 356400 }, { "epoch": 76.84, "learning_rate": 4.641428260407813e-06, "loss": 0.0, "step": 356450 }, { "epoch": 76.85, "learning_rate": 4.637347292409243e-06, "loss": 0.0, "step": 356500 }, { "epoch": 76.86, "learning_rate": 4.63326779123103e-06, "loss": 0.0, "step": 356550 }, { "epoch": 76.87, "learning_rate": 4.629189757450621e-06, "loss": 0.0, "step": 356600 }, { "epoch": 76.88, "learning_rate": 4.625113191645257e-06, "loss": 0.0, "step": 356650 }, { "epoch": 76.89, "learning_rate": 4.621038094391974e-06, "loss": 0.0, "step": 356700 }, { "epoch": 76.9, "learning_rate": 4.616964466267592e-06, "loss": 0.0, "step": 356750 }, { "epoch": 76.91, "learning_rate": 4.61289230784873e-06, "loss": 0.0, "step": 356800 }, { "epoch": 76.92, "learning_rate": 4.608821619711797e-06, "loss": 0.0, "step": 356850 }, { "epoch": 76.93, "learning_rate": 4.604752402432993e-06, "loss": 0.0, "step": 356900 }, { "epoch": 76.95, "learning_rate": 4.600684656588312e-06, "loss": 0.0, "step": 356950 }, { "epoch": 76.96, "learning_rate": 4.596618382753537e-06, "loss": 0.0, "step": 357000 }, { "epoch": 76.97, "learning_rate": 4.592553581504244e-06, "loss": 0.0, "step": 357050 }, { "epoch": 76.98, "learning_rate": 4.588490253415801e-06, "loss": 0.0, "step": 357100 }, { "epoch": 76.99, "learning_rate": 4.584428399063366e-06, "loss": 0.0, "step": 357150 }, { "epoch": 77.0, "learning_rate": 4.580368019021892e-06, "loss": 0.0, "step": 357200 }, { "epoch": 77.01, "learning_rate": 4.5763091138661175e-06, "loss": 0.0, "step": 357250 }, { "epoch": 77.02, "learning_rate": 4.572251684170573e-06, "loss": 0.0, "step": 357300 }, { "epoch": 77.03, "learning_rate": 4.568195730509594e-06, "loss": 0.0, "step": 357350 }, { "epoch": 77.04, "learning_rate": 4.564141253457283e-06, "loss": 0.0, "step": 357400 }, { "epoch": 77.05, "learning_rate": 4.5600882535875506e-06, "loss": 0.0, "step": 357450 }, { "epoch": 77.06, "learning_rate": 4.55603673147409e-06, "loss": 0.0, "step": 357500 }, { "epoch": 77.07, "learning_rate": 4.551986687690399e-06, "loss": 0.0, "step": 357550 }, { "epoch": 77.09, "learning_rate": 4.547938122809753e-06, "loss": 0.0, "step": 357600 }, { "epoch": 77.1, "learning_rate": 4.543891037405214e-06, "loss": 0.0, "step": 357650 }, { "epoch": 77.11, "learning_rate": 4.539845432049648e-06, "loss": 0.0, "step": 357700 }, { "epoch": 77.12, "learning_rate": 4.535801307315696e-06, "loss": 0.0, "step": 357750 }, { "epoch": 77.13, "learning_rate": 4.531758663775813e-06, "loss": 0.0, "step": 357800 }, { "epoch": 77.14, "learning_rate": 4.527717502002227e-06, "loss": 0.0, "step": 357850 }, { "epoch": 77.15, "learning_rate": 4.523677822566951e-06, "loss": 0.0, "step": 357900 }, { "epoch": 77.16, "learning_rate": 4.519639626041797e-06, "loss": 0.0, "step": 357950 }, { "epoch": 77.17, "learning_rate": 4.515602912998374e-06, "loss": 0.0, "step": 358000 }, { "epoch": 77.18, "learning_rate": 4.5115676840080725e-06, "loss": 0.0, "step": 358050 }, { "epoch": 77.19, "learning_rate": 4.507533939642075e-06, "loss": 0.0, "step": 358100 }, { "epoch": 77.2, "learning_rate": 4.503501680471341e-06, "loss": 0.0, "step": 358150 }, { "epoch": 77.21, "learning_rate": 4.499470907066645e-06, "loss": 0.0, "step": 358200 }, { "epoch": 77.23, "learning_rate": 4.495441619998535e-06, "loss": 0.0, "step": 358250 }, { "epoch": 77.24, "learning_rate": 4.491413819837347e-06, "loss": 0.0, "step": 358300 }, { "epoch": 77.25, "learning_rate": 4.4873875071532165e-06, "loss": 0.0, "step": 358350 }, { "epoch": 77.26, "learning_rate": 4.483362682516059e-06, "loss": 0.0, "step": 358400 }, { "epoch": 77.27, "learning_rate": 4.479339346495585e-06, "loss": 0.0, "step": 358450 }, { "epoch": 77.28, "learning_rate": 4.475317499661292e-06, "loss": 0.0, "step": 358500 }, { "epoch": 77.29, "learning_rate": 4.471297142582466e-06, "loss": 0.0, "step": 358550 }, { "epoch": 77.3, "learning_rate": 4.467278275828189e-06, "loss": 0.0, "step": 358600 }, { "epoch": 77.31, "learning_rate": 4.463260899967318e-06, "loss": 0.0, "step": 358650 }, { "epoch": 77.32, "learning_rate": 4.459245015568514e-06, "loss": 0.0, "step": 358700 }, { "epoch": 77.33, "learning_rate": 4.455230623200219e-06, "loss": 0.0, "step": 358750 }, { "epoch": 77.34, "learning_rate": 4.451217723430664e-06, "loss": 0.0, "step": 358800 }, { "epoch": 77.36, "learning_rate": 4.4472063168278705e-06, "loss": 0.0, "step": 358850 }, { "epoch": 77.37, "learning_rate": 4.4431964039596475e-06, "loss": 0.0, "step": 358900 }, { "epoch": 77.38, "learning_rate": 4.439187985393594e-06, "loss": 0.0, "step": 358950 }, { "epoch": 77.39, "learning_rate": 4.43518106169709e-06, "loss": 0.0, "step": 359000 }, { "epoch": 77.4, "learning_rate": 4.431175633437327e-06, "loss": 0.0, "step": 359050 }, { "epoch": 77.41, "learning_rate": 4.4271717011812535e-06, "loss": 0.0, "step": 359100 }, { "epoch": 77.42, "learning_rate": 4.4231692654956244e-06, "loss": 0.0, "step": 359150 }, { "epoch": 77.43, "learning_rate": 4.419168326946976e-06, "loss": 0.0, "step": 359200 }, { "epoch": 77.44, "learning_rate": 4.415168886101645e-06, "loss": 0.0, "step": 359250 }, { "epoch": 77.45, "learning_rate": 4.411170943525747e-06, "loss": 0.0, "step": 359300 }, { "epoch": 77.46, "learning_rate": 4.407174499785177e-06, "loss": 0.0, "step": 359350 }, { "epoch": 77.47, "learning_rate": 4.40317955544563e-06, "loss": 0.0, "step": 359400 }, { "epoch": 77.48, "learning_rate": 4.399186111072583e-06, "loss": 0.0, "step": 359450 }, { "epoch": 77.5, "learning_rate": 4.395194167231309e-06, "loss": 0.0, "step": 359500 }, { "epoch": 77.51, "learning_rate": 4.391203724486866e-06, "loss": 0.0, "step": 359550 }, { "epoch": 77.52, "learning_rate": 4.387214783404086e-06, "loss": 0.0, "step": 359600 }, { "epoch": 77.53, "learning_rate": 4.383227344547595e-06, "loss": 0.0, "step": 359650 }, { "epoch": 77.54, "learning_rate": 4.3792414084818235e-06, "loss": 0.0, "step": 359700 }, { "epoch": 77.55, "learning_rate": 4.375256975770969e-06, "loss": 0.0, "step": 359750 }, { "epoch": 77.56, "learning_rate": 4.371274046979028e-06, "loss": 0.0, "step": 359800 }, { "epoch": 77.57, "learning_rate": 4.367292622669766e-06, "loss": 0.0, "step": 359850 }, { "epoch": 77.58, "learning_rate": 4.363312703406761e-06, "loss": 0.0, "step": 359900 }, { "epoch": 77.59, "learning_rate": 4.359334289753361e-06, "loss": 0.0, "step": 359950 }, { "epoch": 77.6, "learning_rate": 4.355357382272707e-06, "loss": 0.0, "step": 360000 }, { "epoch": 77.61, "learning_rate": 4.351381981527724e-06, "loss": 0.0, "step": 360050 }, { "epoch": 77.62, "learning_rate": 4.3474080880811235e-06, "loss": 0.0, "step": 360100 }, { "epoch": 77.64, "learning_rate": 4.343435702495409e-06, "loss": 0.0, "step": 360150 }, { "epoch": 77.65, "learning_rate": 4.339464825332863e-06, "loss": 0.0, "step": 360200 }, { "epoch": 77.66, "learning_rate": 4.33549545715556e-06, "loss": 0.0, "step": 360250 }, { "epoch": 77.67, "learning_rate": 4.331527598525358e-06, "loss": 0.0, "step": 360300 }, { "epoch": 77.68, "learning_rate": 4.327561250003902e-06, "loss": 0.0, "step": 360350 }, { "epoch": 77.69, "learning_rate": 4.323596412152625e-06, "loss": 0.0, "step": 360400 }, { "epoch": 77.7, "learning_rate": 4.3196330855327445e-06, "loss": 0.0, "step": 360450 }, { "epoch": 77.71, "learning_rate": 4.315671270705264e-06, "loss": 0.0, "step": 360500 }, { "epoch": 77.72, "learning_rate": 4.311710968230975e-06, "loss": 0.0, "step": 360550 }, { "epoch": 77.73, "learning_rate": 4.307752178670449e-06, "loss": 0.0, "step": 360600 }, { "epoch": 77.74, "learning_rate": 4.303794902584052e-06, "loss": 0.0, "step": 360650 }, { "epoch": 77.75, "learning_rate": 4.299839140531926e-06, "loss": 0.0, "step": 360700 }, { "epoch": 77.76, "learning_rate": 4.295884893074015e-06, "loss": 0.0, "step": 360750 }, { "epoch": 77.78, "learning_rate": 4.291932160770027e-06, "loss": 0.0, "step": 360800 }, { "epoch": 77.79, "learning_rate": 4.287980944179471e-06, "loss": 0.0, "step": 360850 }, { "epoch": 77.8, "learning_rate": 4.284031243861629e-06, "loss": 0.0, "step": 360900 }, { "epoch": 77.81, "learning_rate": 4.280083060375588e-06, "loss": 0.0, "step": 360950 }, { "epoch": 77.82, "learning_rate": 4.276136394280207e-06, "loss": 0.0, "step": 361000 }, { "epoch": 77.83, "learning_rate": 4.272191246134125e-06, "loss": 0.0, "step": 361050 }, { "epoch": 77.84, "learning_rate": 4.268247616495774e-06, "loss": 0.0, "step": 361100 }, { "epoch": 77.85, "learning_rate": 4.264305505923366e-06, "loss": 0.0, "step": 361150 }, { "epoch": 77.86, "learning_rate": 4.2603649149749115e-06, "loss": 0.0, "step": 361200 }, { "epoch": 77.87, "learning_rate": 4.256425844208197e-06, "loss": 0.0, "step": 361250 }, { "epoch": 77.88, "learning_rate": 4.252488294180783e-06, "loss": 0.0, "step": 361300 }, { "epoch": 77.89, "learning_rate": 4.248552265450023e-06, "loss": 0.0, "step": 361350 }, { "epoch": 77.9, "learning_rate": 4.244617758573068e-06, "loss": 0.0, "step": 361400 }, { "epoch": 77.92, "learning_rate": 4.240684774106838e-06, "loss": 0.0, "step": 361450 }, { "epoch": 77.93, "learning_rate": 4.2367533126080425e-06, "loss": 0.0, "step": 361500 }, { "epoch": 77.94, "learning_rate": 4.232823374633173e-06, "loss": 0.0, "step": 361550 }, { "epoch": 77.95, "learning_rate": 4.228894960738507e-06, "loss": 0.0, "step": 361600 }, { "epoch": 77.96, "learning_rate": 4.224968071480109e-06, "loss": 0.0, "step": 361650 }, { "epoch": 77.97, "learning_rate": 4.221042707413824e-06, "loss": 0.0, "step": 361700 }, { "epoch": 77.98, "learning_rate": 4.217118869095281e-06, "loss": 0.0, "step": 361750 }, { "epoch": 77.99, "learning_rate": 4.213196557079896e-06, "loss": 0.0, "step": 361800 }, { "epoch": 78.0, "learning_rate": 4.209275771922867e-06, "loss": 0.0, "step": 361850 }, { "epoch": 78.01, "learning_rate": 4.205356514179176e-06, "loss": 0.0, "step": 361900 }, { "epoch": 78.02, "learning_rate": 4.2014387844035885e-06, "loss": 0.0, "step": 361950 }, { "epoch": 78.03, "learning_rate": 4.197522583150655e-06, "loss": 0.0, "step": 362000 }, { "epoch": 78.04, "learning_rate": 4.193607910974709e-06, "loss": 0.0, "step": 362050 }, { "epoch": 78.06, "learning_rate": 4.189694768429866e-06, "loss": 0.0, "step": 362100 }, { "epoch": 78.07, "learning_rate": 4.185783156070027e-06, "loss": 0.0, "step": 362150 }, { "epoch": 78.08, "learning_rate": 4.181873074448877e-06, "loss": 0.0, "step": 362200 }, { "epoch": 78.09, "learning_rate": 4.177964524119881e-06, "loss": 0.0, "step": 362250 }, { "epoch": 78.1, "learning_rate": 4.174057505636289e-06, "loss": 0.0, "step": 362300 }, { "epoch": 78.11, "learning_rate": 4.170152019551138e-06, "loss": 0.0, "step": 362350 }, { "epoch": 78.12, "learning_rate": 4.166248066417237e-06, "loss": 0.0, "step": 362400 }, { "epoch": 78.13, "learning_rate": 4.162345646787202e-06, "loss": 0.0, "step": 362450 }, { "epoch": 78.14, "learning_rate": 4.158444761213399e-06, "loss": 0.0, "step": 362500 }, { "epoch": 78.15, "learning_rate": 4.154545410247998e-06, "loss": 0.0, "step": 362550 }, { "epoch": 78.16, "learning_rate": 4.1506475944429484e-06, "loss": 0.0, "step": 362600 }, { "epoch": 78.17, "learning_rate": 4.146751314349979e-06, "loss": 0.0, "step": 362650 }, { "epoch": 78.18, "learning_rate": 4.142856570520611e-06, "loss": 0.0, "step": 362700 }, { "epoch": 78.2, "learning_rate": 4.138963363506132e-06, "loss": 0.0, "step": 362750 }, { "epoch": 78.21, "learning_rate": 4.1350716938576235e-06, "loss": 0.0, "step": 362800 }, { "epoch": 78.22, "learning_rate": 4.1311815621259404e-06, "loss": 0.0, "step": 362850 }, { "epoch": 78.23, "learning_rate": 4.127292968861738e-06, "loss": 0.0, "step": 362900 }, { "epoch": 78.24, "learning_rate": 4.123405914615432e-06, "loss": 0.0, "step": 362950 }, { "epoch": 78.25, "learning_rate": 4.11952039993724e-06, "loss": 0.0, "step": 363000 }, { "epoch": 78.26, "learning_rate": 4.115636425377137e-06, "loss": 0.0, "step": 363050 }, { "epoch": 78.27, "learning_rate": 4.111753991484906e-06, "loss": 0.0, "step": 363100 }, { "epoch": 78.28, "learning_rate": 4.107873098810098e-06, "loss": 0.0, "step": 363150 }, { "epoch": 78.29, "learning_rate": 4.103993747902048e-06, "loss": 0.0, "step": 363200 }, { "epoch": 78.3, "learning_rate": 4.100115939309873e-06, "loss": 0.0, "step": 363250 }, { "epoch": 78.31, "learning_rate": 4.096239673582471e-06, "loss": 0.0, "step": 363300 }, { "epoch": 78.33, "learning_rate": 4.092364951268524e-06, "loss": 0.0, "step": 363350 }, { "epoch": 78.34, "learning_rate": 4.088491772916493e-06, "loss": 0.0, "step": 363400 }, { "epoch": 78.35, "learning_rate": 4.084620139074622e-06, "loss": 0.0, "step": 363450 }, { "epoch": 78.36, "learning_rate": 4.080750050290937e-06, "loss": 0.0, "step": 363500 }, { "epoch": 78.37, "learning_rate": 4.076881507113241e-06, "loss": 0.0, "step": 363550 }, { "epoch": 78.38, "learning_rate": 4.073014510089125e-06, "loss": 0.0, "step": 363600 }, { "epoch": 78.39, "learning_rate": 4.069149059765956e-06, "loss": 0.0, "step": 363650 }, { "epoch": 78.4, "learning_rate": 4.0652851566908805e-06, "loss": 0.0, "step": 363700 }, { "epoch": 78.41, "learning_rate": 4.0614228014108345e-06, "loss": 0.0, "step": 363750 }, { "epoch": 78.42, "learning_rate": 4.0575619944725265e-06, "loss": 0.0, "step": 363800 }, { "epoch": 78.43, "learning_rate": 4.05370273642245e-06, "loss": 0.0, "step": 363850 }, { "epoch": 78.44, "learning_rate": 4.049845027806877e-06, "loss": 0.0, "step": 363900 }, { "epoch": 78.45, "learning_rate": 4.045988869171861e-06, "loss": 0.0, "step": 363950 }, { "epoch": 78.47, "learning_rate": 4.042134261063238e-06, "loss": 0.0, "step": 364000 }, { "epoch": 78.48, "learning_rate": 4.038281204026621e-06, "loss": 0.0, "step": 364050 }, { "epoch": 78.49, "learning_rate": 4.034429698607403e-06, "loss": 0.0, "step": 364100 }, { "epoch": 78.5, "learning_rate": 4.030579745350767e-06, "loss": 0.0, "step": 364150 }, { "epoch": 78.51, "learning_rate": 4.0267313448016685e-06, "loss": 0.0, "step": 364200 }, { "epoch": 78.52, "learning_rate": 4.022884497504837e-06, "loss": 0.0, "step": 364250 }, { "epoch": 78.53, "learning_rate": 4.019039204004793e-06, "loss": 0.0, "step": 364300 }, { "epoch": 78.54, "learning_rate": 4.015195464845825e-06, "loss": 0.0, "step": 364350 }, { "epoch": 78.55, "learning_rate": 4.011353280572023e-06, "loss": 0.0, "step": 364400 }, { "epoch": 78.56, "learning_rate": 4.007512651727241e-06, "loss": 0.0, "step": 364450 }, { "epoch": 78.57, "learning_rate": 4.003673578855105e-06, "loss": 0.0, "step": 364500 }, { "epoch": 78.58, "learning_rate": 3.999836062499033e-06, "loss": 0.0, "step": 364550 }, { "epoch": 78.59, "learning_rate": 3.9960001032022295e-06, "loss": 0.0, "step": 364600 }, { "epoch": 78.61, "learning_rate": 3.992165701507664e-06, "loss": 0.0, "step": 364650 }, { "epoch": 78.62, "learning_rate": 3.9883328579580956e-06, "loss": 0.0, "step": 364700 }, { "epoch": 78.63, "learning_rate": 3.984501573096046e-06, "loss": 0.0, "step": 364750 }, { "epoch": 78.64, "learning_rate": 3.980671847463842e-06, "loss": 0.0, "step": 364800 }, { "epoch": 78.65, "learning_rate": 3.976843681603571e-06, "loss": 0.0, "step": 364850 }, { "epoch": 78.66, "learning_rate": 3.973017076057107e-06, "loss": 0.0, "step": 364900 }, { "epoch": 78.67, "learning_rate": 3.969192031366099e-06, "loss": 0.0, "step": 364950 }, { "epoch": 78.68, "learning_rate": 3.965368548071978e-06, "loss": 0.0, "step": 365000 }, { "epoch": 78.69, "learning_rate": 3.961546626715955e-06, "loss": 0.0, "step": 365050 }, { "epoch": 78.7, "learning_rate": 3.957726267839016e-06, "loss": 0.0, "step": 365100 }, { "epoch": 78.71, "learning_rate": 3.953907471981927e-06, "loss": 0.0, "step": 365150 }, { "epoch": 78.72, "learning_rate": 3.950090239685236e-06, "loss": 0.0, "step": 365200 }, { "epoch": 78.73, "learning_rate": 3.9462745714892666e-06, "loss": 0.0, "step": 365250 }, { "epoch": 78.75, "learning_rate": 3.942460467934121e-06, "loss": 0.0, "step": 365300 }, { "epoch": 78.76, "learning_rate": 3.938647929559681e-06, "loss": 0.0, "step": 365350 }, { "epoch": 78.77, "learning_rate": 3.934836956905609e-06, "loss": 0.0, "step": 365400 }, { "epoch": 78.78, "learning_rate": 3.931027550511341e-06, "loss": 0.0, "step": 365450 }, { "epoch": 78.79, "learning_rate": 3.927219710916093e-06, "loss": 0.0, "step": 365500 }, { "epoch": 78.8, "learning_rate": 3.9234134386588614e-06, "loss": 0.0, "step": 365550 }, { "epoch": 78.81, "learning_rate": 3.919608734278412e-06, "loss": 0.0, "step": 365600 }, { "epoch": 78.82, "learning_rate": 3.915805598313315e-06, "loss": 0.0, "step": 365650 }, { "epoch": 78.83, "learning_rate": 3.912004031301882e-06, "loss": 0.0, "step": 365700 }, { "epoch": 78.84, "learning_rate": 3.908204033782224e-06, "loss": 0.0, "step": 365750 }, { "epoch": 78.85, "learning_rate": 3.904405606292221e-06, "loss": 0.0, "step": 365800 }, { "epoch": 78.86, "learning_rate": 3.900608749369548e-06, "loss": 0.0, "step": 365850 }, { "epoch": 78.87, "learning_rate": 3.896813463551643e-06, "loss": 0.0, "step": 365900 }, { "epoch": 78.89, "learning_rate": 3.893019749375716e-06, "loss": 0.0, "step": 365950 }, { "epoch": 78.9, "learning_rate": 3.889227607378764e-06, "loss": 0.0, "step": 366000 }, { "epoch": 78.91, "learning_rate": 3.885437038097559e-06, "loss": 0.0, "step": 366050 }, { "epoch": 78.92, "learning_rate": 3.881648042068658e-06, "loss": 0.0, "step": 366100 }, { "epoch": 78.93, "learning_rate": 3.87786061982839e-06, "loss": 0.0, "step": 366150 }, { "epoch": 78.94, "learning_rate": 3.874074771912849e-06, "loss": 0.0, "step": 366200 }, { "epoch": 78.95, "learning_rate": 3.870290498857921e-06, "loss": 0.0, "step": 366250 }, { "epoch": 78.96, "learning_rate": 3.866507801199269e-06, "loss": 0.0, "step": 366300 }, { "epoch": 78.97, "learning_rate": 3.862726679472325e-06, "loss": 0.0, "step": 366350 }, { "epoch": 78.98, "learning_rate": 3.85894713421231e-06, "loss": 0.0, "step": 366400 }, { "epoch": 78.99, "learning_rate": 3.8551691659542e-06, "loss": 0.0, "step": 366450 }, { "epoch": 79.0, "learning_rate": 3.851392775232772e-06, "loss": 0.0, "step": 366500 }, { "epoch": 79.01, "learning_rate": 3.847617962582568e-06, "loss": 0.0, "step": 366550 }, { "epoch": 79.03, "learning_rate": 3.843844728537905e-06, "loss": 0.0, "step": 366600 }, { "epoch": 79.04, "learning_rate": 3.840073073632882e-06, "loss": 0.0, "step": 366650 }, { "epoch": 79.05, "learning_rate": 3.83630299840137e-06, "loss": 0.0, "step": 366700 }, { "epoch": 79.06, "learning_rate": 3.83253450337702e-06, "loss": 0.0, "step": 366750 }, { "epoch": 79.07, "learning_rate": 3.828767589093256e-06, "loss": 0.0, "step": 366800 }, { "epoch": 79.08, "learning_rate": 3.825002256083282e-06, "loss": 0.0, "step": 366850 }, { "epoch": 79.09, "learning_rate": 3.821238504880075e-06, "loss": 0.0, "step": 366900 }, { "epoch": 79.1, "learning_rate": 3.81747633601639e-06, "loss": 0.0, "step": 366950 }, { "epoch": 79.11, "learning_rate": 3.8137157500247556e-06, "loss": 0.0, "step": 367000 }, { "epoch": 79.12, "learning_rate": 3.8099567474374784e-06, "loss": 0.0, "step": 367050 }, { "epoch": 79.13, "learning_rate": 3.806199328786642e-06, "loss": 0.0, "step": 367100 }, { "epoch": 79.14, "learning_rate": 3.802443494604102e-06, "loss": 0.0, "step": 367150 }, { "epoch": 79.15, "learning_rate": 3.7986892454214933e-06, "loss": 0.0, "step": 367200 }, { "epoch": 79.17, "learning_rate": 3.794936581770226e-06, "loss": 0.0, "step": 367250 }, { "epoch": 79.18, "learning_rate": 3.791185504181478e-06, "loss": 0.0, "step": 367300 }, { "epoch": 79.19, "learning_rate": 3.7874360131862253e-06, "loss": 0.0, "step": 367350 }, { "epoch": 79.2, "learning_rate": 3.7836881093151893e-06, "loss": 0.0, "step": 367400 }, { "epoch": 79.21, "learning_rate": 3.779941793098886e-06, "loss": 0.0, "step": 367450 }, { "epoch": 79.22, "learning_rate": 3.7761970650675998e-06, "loss": 0.0, "step": 367500 }, { "epoch": 79.23, "learning_rate": 3.7724539257513897e-06, "loss": 0.0, "step": 367550 }, { "epoch": 79.24, "learning_rate": 3.768712375680105e-06, "loss": 0.0, "step": 367600 }, { "epoch": 79.25, "learning_rate": 3.764972415383346e-06, "loss": 0.0, "step": 367650 }, { "epoch": 79.26, "learning_rate": 3.7612340453905002e-06, "loss": 0.0, "step": 367700 }, { "epoch": 79.27, "learning_rate": 3.7574972662307268e-06, "loss": 0.0, "step": 367750 }, { "epoch": 79.28, "learning_rate": 3.75376207843297e-06, "loss": 0.0, "step": 367800 }, { "epoch": 79.3, "learning_rate": 3.7500284825259413e-06, "loss": 0.0, "step": 367850 }, { "epoch": 79.31, "learning_rate": 3.7462964790381187e-06, "loss": 0.0, "step": 367900 }, { "epoch": 79.32, "learning_rate": 3.742566068497762e-06, "loss": 0.0, "step": 367950 }, { "epoch": 79.33, "learning_rate": 3.7388372514329133e-06, "loss": 0.0, "step": 368000 }, { "epoch": 79.34, "learning_rate": 3.7351100283713807e-06, "loss": 0.0, "step": 368050 }, { "epoch": 79.35, "learning_rate": 3.7313843998407486e-06, "loss": 0.0, "step": 368100 }, { "epoch": 79.36, "learning_rate": 3.7276603663683656e-06, "loss": 0.0, "step": 368150 }, { "epoch": 79.37, "learning_rate": 3.7239379284813747e-06, "loss": 0.0, "step": 368200 }, { "epoch": 79.38, "learning_rate": 3.720217086706678e-06, "loss": 0.0, "step": 368250 }, { "epoch": 79.39, "learning_rate": 3.716497841570955e-06, "loss": 0.0, "step": 368300 }, { "epoch": 79.4, "learning_rate": 3.7127801936006674e-06, "loss": 0.0, "step": 368350 }, { "epoch": 79.41, "learning_rate": 3.7090641433220286e-06, "loss": 0.0, "step": 368400 }, { "epoch": 79.42, "learning_rate": 3.705349691261053e-06, "loss": 0.0, "step": 368450 }, { "epoch": 79.44, "learning_rate": 3.701636837943515e-06, "loss": 0.0, "step": 368500 }, { "epoch": 79.45, "learning_rate": 3.697925583894961e-06, "loss": 0.0, "step": 368550 }, { "epoch": 79.46, "learning_rate": 3.6942159296407175e-06, "loss": 0.0, "step": 368600 }, { "epoch": 79.47, "learning_rate": 3.690507875705879e-06, "loss": 0.0, "step": 368650 }, { "epoch": 79.48, "learning_rate": 3.6868014226153164e-06, "loss": 0.0, "step": 368700 }, { "epoch": 79.49, "learning_rate": 3.6830965708936744e-06, "loss": 0.0, "step": 368750 }, { "epoch": 79.5, "learning_rate": 3.6793933210653686e-06, "loss": 0.0, "step": 368800 }, { "epoch": 79.51, "learning_rate": 3.6756916736545913e-06, "loss": 0.0, "step": 368850 }, { "epoch": 79.52, "learning_rate": 3.671991629185304e-06, "loss": 0.0, "step": 368900 }, { "epoch": 79.53, "learning_rate": 3.668293188181244e-06, "loss": 0.0, "step": 368950 }, { "epoch": 79.54, "learning_rate": 3.664596351165917e-06, "loss": 0.0, "step": 369000 }, { "epoch": 79.55, "learning_rate": 3.660901118662617e-06, "loss": 0.0, "step": 369050 }, { "epoch": 79.56, "learning_rate": 3.65720749119439e-06, "loss": 0.0, "step": 369100 }, { "epoch": 79.58, "learning_rate": 3.6535154692840643e-06, "loss": 0.0, "step": 369150 }, { "epoch": 79.59, "learning_rate": 3.649825053454244e-06, "loss": 0.0, "step": 369200 }, { "epoch": 79.6, "learning_rate": 3.6461362442272965e-06, "loss": 0.0, "step": 369250 }, { "epoch": 79.61, "learning_rate": 3.6424490421253836e-06, "loss": 0.0, "step": 369300 }, { "epoch": 79.62, "learning_rate": 3.638763447670408e-06, "loss": 0.0, "step": 369350 }, { "epoch": 79.63, "learning_rate": 3.6350794613840673e-06, "loss": 0.0, "step": 369400 }, { "epoch": 79.64, "learning_rate": 3.631397083787821e-06, "loss": 0.0, "step": 369450 }, { "epoch": 79.65, "learning_rate": 3.627716315402912e-06, "loss": 0.0, "step": 369500 }, { "epoch": 79.66, "learning_rate": 3.624037156750349e-06, "loss": 0.0, "step": 369550 }, { "epoch": 79.67, "learning_rate": 3.620359608350905e-06, "loss": 0.0, "step": 369600 }, { "epoch": 79.68, "learning_rate": 3.6166836707251304e-06, "loss": 0.0, "step": 369650 }, { "epoch": 79.69, "learning_rate": 3.6130093443933574e-06, "loss": 0.0, "step": 369700 }, { "epoch": 79.7, "learning_rate": 3.609336629875682e-06, "loss": 0.0, "step": 369750 }, { "epoch": 79.72, "learning_rate": 3.6056655276919726e-06, "loss": 0.0, "step": 369800 }, { "epoch": 79.73, "learning_rate": 3.601996038361859e-06, "loss": 0.0, "step": 369850 }, { "epoch": 79.74, "learning_rate": 3.598328162404762e-06, "loss": 0.0, "step": 369900 }, { "epoch": 79.75, "learning_rate": 3.5946619003398634e-06, "loss": 0.0, "step": 369950 }, { "epoch": 79.76, "learning_rate": 3.590997252686118e-06, "loss": 0.0, "step": 370000 }, { "epoch": 79.77, "learning_rate": 3.5873342199622553e-06, "loss": 0.0, "step": 370050 }, { "epoch": 79.78, "learning_rate": 3.5836728026867615e-06, "loss": 0.0, "step": 370100 }, { "epoch": 79.79, "learning_rate": 3.5800130013779165e-06, "loss": 0.0, "step": 370150 }, { "epoch": 79.8, "learning_rate": 3.5763548165537556e-06, "loss": 0.0, "step": 370200 }, { "epoch": 79.81, "learning_rate": 3.5726982487320937e-06, "loss": 0.0, "step": 370250 }, { "epoch": 79.82, "learning_rate": 3.5690432984305094e-06, "loss": 0.0, "step": 370300 }, { "epoch": 79.83, "learning_rate": 3.5653899661663607e-06, "loss": 0.0, "step": 370350 }, { "epoch": 79.84, "learning_rate": 3.5617382524567676e-06, "loss": 0.0, "step": 370400 }, { "epoch": 79.86, "learning_rate": 3.558088157818628e-06, "loss": 0.0, "step": 370450 }, { "epoch": 79.87, "learning_rate": 3.5544396827686087e-06, "loss": 0.0, "step": 370500 }, { "epoch": 79.88, "learning_rate": 3.550792827823145e-06, "loss": 0.0, "step": 370550 }, { "epoch": 79.89, "learning_rate": 3.5471475934984455e-06, "loss": 0.0, "step": 370600 }, { "epoch": 79.9, "learning_rate": 3.543503980310489e-06, "loss": 0.0, "step": 370650 }, { "epoch": 79.91, "learning_rate": 3.539861988775018e-06, "loss": 0.0, "step": 370700 }, { "epoch": 79.92, "learning_rate": 3.5362216194075676e-06, "loss": 0.0, "step": 370750 }, { "epoch": 79.93, "learning_rate": 3.5325828727234127e-06, "loss": 0.0, "step": 370800 }, { "epoch": 79.94, "learning_rate": 3.528945749237617e-06, "loss": 0.0, "step": 370850 }, { "epoch": 79.95, "learning_rate": 3.525310249465012e-06, "loss": 0.0, "step": 370900 }, { "epoch": 79.96, "learning_rate": 3.521676373920194e-06, "loss": 0.0, "step": 370950 }, { "epoch": 79.97, "learning_rate": 3.5180441231175454e-06, "loss": 0.0, "step": 371000 }, { "epoch": 79.98, "learning_rate": 3.5144134975711945e-06, "loss": 0.0, "step": 371050 }, { "epoch": 80.0, "learning_rate": 3.510784497795057e-06, "loss": 0.0, "step": 371100 }, { "epoch": 80.01, "learning_rate": 3.5071571243028053e-06, "loss": 0.0, "step": 371150 }, { "epoch": 80.02, "learning_rate": 3.5035313776079026e-06, "loss": 0.0, "step": 371200 }, { "epoch": 80.03, "learning_rate": 3.499907258223566e-06, "loss": 0.0, "step": 371250 }, { "epoch": 80.04, "learning_rate": 3.496284766662777e-06, "loss": 0.0, "step": 371300 }, { "epoch": 80.05, "learning_rate": 3.4926639034382935e-06, "loss": 0.0, "step": 371350 }, { "epoch": 80.06, "learning_rate": 3.489044669062656e-06, "loss": 0.0, "step": 371400 }, { "epoch": 80.07, "learning_rate": 3.4854270640481543e-06, "loss": 0.0, "step": 371450 }, { "epoch": 80.08, "learning_rate": 3.4818110889068625e-06, "loss": 0.0, "step": 371500 }, { "epoch": 80.09, "learning_rate": 3.4781967441506046e-06, "loss": 0.0, "step": 371550 }, { "epoch": 80.1, "learning_rate": 3.474584030290999e-06, "loss": 0.0, "step": 371600 }, { "epoch": 80.11, "learning_rate": 3.4709729478394157e-06, "loss": 0.0, "step": 371650 }, { "epoch": 80.13, "learning_rate": 3.467363497306999e-06, "loss": 0.0, "step": 371700 }, { "epoch": 80.14, "learning_rate": 3.463755679204668e-06, "loss": 0.0, "step": 371750 }, { "epoch": 80.15, "learning_rate": 3.46014949404309e-06, "loss": 0.0, "step": 371800 }, { "epoch": 80.16, "learning_rate": 3.4565449423327277e-06, "loss": 0.0, "step": 371850 }, { "epoch": 80.17, "learning_rate": 3.4529420245838e-06, "loss": 0.0, "step": 371900 }, { "epoch": 80.18, "learning_rate": 3.449340741306294e-06, "loss": 0.0, "step": 371950 }, { "epoch": 80.19, "learning_rate": 3.4457410930099647e-06, "loss": 0.0, "step": 372000 }, { "epoch": 80.2, "learning_rate": 3.4421430802043404e-06, "loss": 0.0, "step": 372050 }, { "epoch": 80.21, "learning_rate": 3.4385467033987144e-06, "loss": 0.0, "step": 372100 }, { "epoch": 80.22, "learning_rate": 3.434951963102149e-06, "loss": 0.0, "step": 372150 }, { "epoch": 80.23, "learning_rate": 3.4313588598234742e-06, "loss": 0.0, "step": 372200 }, { "epoch": 80.24, "learning_rate": 3.4277673940712907e-06, "loss": 0.0, "step": 372250 }, { "epoch": 80.25, "learning_rate": 3.4241775663539665e-06, "loss": 0.0, "step": 372300 }, { "epoch": 80.27, "learning_rate": 3.420589377179635e-06, "loss": 0.0, "step": 372350 }, { "epoch": 80.28, "learning_rate": 3.4170028270562026e-06, "loss": 0.0, "step": 372400 }, { "epoch": 80.29, "learning_rate": 3.413417916491337e-06, "loss": 0.0, "step": 372450 }, { "epoch": 80.3, "learning_rate": 3.4098346459924826e-06, "loss": 0.0, "step": 372500 }, { "epoch": 80.31, "learning_rate": 3.4062530160668425e-06, "loss": 0.0, "step": 372550 }, { "epoch": 80.32, "learning_rate": 3.4026730272213942e-06, "loss": 0.0, "step": 372600 }, { "epoch": 80.33, "learning_rate": 3.399094679962877e-06, "loss": 0.0, "step": 372650 }, { "epoch": 80.34, "learning_rate": 3.395517974797811e-06, "loss": 0.0, "step": 372700 }, { "epoch": 80.35, "learning_rate": 3.3919429122324647e-06, "loss": 0.0, "step": 372750 }, { "epoch": 80.36, "learning_rate": 3.3883694927728866e-06, "loss": 0.0, "step": 372800 }, { "epoch": 80.37, "learning_rate": 3.384797716924885e-06, "loss": 0.0, "step": 372850 }, { "epoch": 80.38, "learning_rate": 3.3812275851940492e-06, "loss": 0.0, "step": 372900 }, { "epoch": 80.39, "learning_rate": 3.377659098085727e-06, "loss": 0.0, "step": 372950 }, { "epoch": 80.41, "learning_rate": 3.374092256105025e-06, "loss": 0.0, "step": 373000 }, { "epoch": 80.42, "learning_rate": 3.3705270597568254e-06, "loss": 0.0, "step": 373050 }, { "epoch": 80.43, "learning_rate": 3.366963509545784e-06, "loss": 0.0, "step": 373100 }, { "epoch": 80.44, "learning_rate": 3.363401605976314e-06, "loss": 0.0, "step": 373150 }, { "epoch": 80.45, "learning_rate": 3.359841349552602e-06, "loss": 0.0, "step": 373200 }, { "epoch": 80.46, "learning_rate": 3.35628274077859e-06, "loss": 0.0, "step": 373250 }, { "epoch": 80.47, "learning_rate": 3.3527257801579954e-06, "loss": 0.0, "step": 373300 }, { "epoch": 80.48, "learning_rate": 3.3491704681943065e-06, "loss": 0.0, "step": 373350 }, { "epoch": 80.49, "learning_rate": 3.345616805390773e-06, "loss": 0.0, "step": 373400 }, { "epoch": 80.5, "learning_rate": 3.3420647922504133e-06, "loss": 0.0, "step": 373450 }, { "epoch": 80.51, "learning_rate": 3.338514429275996e-06, "loss": 0.0, "step": 373500 }, { "epoch": 80.52, "learning_rate": 3.3349657169700878e-06, "loss": 0.0, "step": 373550 }, { "epoch": 80.53, "learning_rate": 3.3314186558349975e-06, "loss": 0.0, "step": 373600 }, { "epoch": 80.55, "learning_rate": 3.3278732463728055e-06, "loss": 0.0, "step": 373650 }, { "epoch": 80.56, "learning_rate": 3.324329489085364e-06, "loss": 0.0, "step": 373700 }, { "epoch": 80.57, "learning_rate": 3.320787384474284e-06, "loss": 0.0, "step": 373750 }, { "epoch": 80.58, "learning_rate": 3.3172469330409478e-06, "loss": 0.0, "step": 373800 }, { "epoch": 80.59, "learning_rate": 3.3137081352865005e-06, "loss": 0.0, "step": 373850 }, { "epoch": 80.6, "learning_rate": 3.310170991711855e-06, "loss": 0.0, "step": 373900 }, { "epoch": 80.61, "learning_rate": 3.306635502817689e-06, "loss": 0.0, "step": 373950 }, { "epoch": 80.62, "learning_rate": 3.303101669104448e-06, "loss": 0.0, "step": 374000 }, { "epoch": 80.63, "learning_rate": 3.2995694910723397e-06, "loss": 0.0, "step": 374050 }, { "epoch": 80.64, "learning_rate": 3.2960389692213417e-06, "loss": 0.0, "step": 374100 }, { "epoch": 80.65, "learning_rate": 3.292510104051193e-06, "loss": 0.0, "step": 374150 }, { "epoch": 80.66, "learning_rate": 3.288982896061399e-06, "loss": 0.0, "step": 374200 }, { "epoch": 80.67, "learning_rate": 3.285457345751235e-06, "loss": 0.0, "step": 374250 }, { "epoch": 80.69, "learning_rate": 3.2819334536197355e-06, "loss": 0.0, "step": 374300 }, { "epoch": 80.7, "learning_rate": 3.2784112201656994e-06, "loss": 0.0, "step": 374350 }, { "epoch": 80.71, "learning_rate": 3.2748906458877088e-06, "loss": 0.0, "step": 374400 }, { "epoch": 80.72, "learning_rate": 3.2713717312840806e-06, "loss": 0.0, "step": 374450 }, { "epoch": 80.73, "learning_rate": 3.267854476852919e-06, "loss": 0.0, "step": 374500 }, { "epoch": 80.74, "learning_rate": 3.2643388830920817e-06, "loss": 0.0, "step": 374550 }, { "epoch": 80.75, "learning_rate": 3.260824950499206e-06, "loss": 0.0, "step": 374600 }, { "epoch": 80.76, "learning_rate": 3.2573126795716855e-06, "loss": 0.0, "step": 374650 }, { "epoch": 80.77, "learning_rate": 3.253802070806666e-06, "loss": 0.0, "step": 374700 }, { "epoch": 80.78, "learning_rate": 3.250293124701072e-06, "loss": 0.0, "step": 374750 }, { "epoch": 80.79, "learning_rate": 3.2467858417515988e-06, "loss": 0.0, "step": 374800 }, { "epoch": 80.8, "learning_rate": 3.243280222454693e-06, "loss": 0.0, "step": 374850 }, { "epoch": 80.81, "learning_rate": 3.2397762673065746e-06, "loss": 0.0, "step": 374900 }, { "epoch": 80.83, "learning_rate": 3.236273976803218e-06, "loss": 0.0, "step": 374950 }, { "epoch": 80.84, "learning_rate": 3.2327733514403633e-06, "loss": 0.0, "step": 375000 }, { "epoch": 80.85, "learning_rate": 3.229274391713533e-06, "loss": 0.0, "step": 375050 }, { "epoch": 80.86, "learning_rate": 3.2257770981179924e-06, "loss": 0.0, "step": 375100 }, { "epoch": 80.87, "learning_rate": 3.2222814711487854e-06, "loss": 0.0, "step": 375150 }, { "epoch": 80.88, "learning_rate": 3.2187875113006994e-06, "loss": 0.0, "step": 375200 }, { "epoch": 80.89, "learning_rate": 3.2152952190683154e-06, "loss": 0.0, "step": 375250 }, { "epoch": 80.9, "learning_rate": 3.2118045949459552e-06, "loss": 0.0, "step": 375300 }, { "epoch": 80.91, "learning_rate": 3.208315639427713e-06, "loss": 0.0, "step": 375350 }, { "epoch": 80.92, "learning_rate": 3.204828353007449e-06, "loss": 0.0, "step": 375400 }, { "epoch": 80.93, "learning_rate": 3.20134273617878e-06, "loss": 0.0, "step": 375450 }, { "epoch": 80.94, "learning_rate": 3.197858789435095e-06, "loss": 0.0, "step": 375500 }, { "epoch": 80.95, "learning_rate": 3.1943765132695387e-06, "loss": 0.0, "step": 375550 }, { "epoch": 80.97, "learning_rate": 3.190895908175024e-06, "loss": 0.0, "step": 375600 }, { "epoch": 80.98, "learning_rate": 3.187416974644225e-06, "loss": 0.0, "step": 375650 }, { "epoch": 80.99, "learning_rate": 3.183939713169583e-06, "loss": 0.0, "step": 375700 }, { "epoch": 81.0, "learning_rate": 3.1804641242432973e-06, "loss": 0.0, "step": 375750 }, { "epoch": 81.01, "learning_rate": 3.1769902083573364e-06, "loss": 0.0, "step": 375800 }, { "epoch": 81.02, "learning_rate": 3.1735179660034253e-06, "loss": 0.0, "step": 375850 }, { "epoch": 81.03, "learning_rate": 3.170047397673057e-06, "loss": 0.0, "step": 375900 }, { "epoch": 81.04, "learning_rate": 3.1665785038574857e-06, "loss": 0.0, "step": 375950 }, { "epoch": 81.05, "learning_rate": 3.163111285047728e-06, "loss": 0.0, "step": 376000 }, { "epoch": 81.06, "learning_rate": 3.1596457417345643e-06, "loss": 0.0, "step": 376050 }, { "epoch": 81.07, "learning_rate": 3.156181874408544e-06, "loss": 0.0, "step": 376100 }, { "epoch": 81.08, "learning_rate": 3.1527196835599657e-06, "loss": 0.0, "step": 376150 }, { "epoch": 81.1, "learning_rate": 3.149259169678901e-06, "loss": 0.0, "step": 376200 }, { "epoch": 81.11, "learning_rate": 3.1458003332551753e-06, "loss": 0.0, "step": 376250 }, { "epoch": 81.12, "learning_rate": 3.1423431747783933e-06, "loss": 0.0, "step": 376300 }, { "epoch": 81.13, "learning_rate": 3.138887694737909e-06, "loss": 0.0, "step": 376350 }, { "epoch": 81.14, "learning_rate": 3.1354338936228353e-06, "loss": 0.0, "step": 376400 }, { "epoch": 81.15, "learning_rate": 3.131981771922051e-06, "loss": 0.0, "step": 376450 }, { "epoch": 81.16, "learning_rate": 3.1285313301242103e-06, "loss": 0.0, "step": 376500 }, { "epoch": 81.17, "learning_rate": 3.1250825687177137e-06, "loss": 0.0, "step": 376550 }, { "epoch": 81.18, "learning_rate": 3.121635488190731e-06, "loss": 0.0, "step": 376600 }, { "epoch": 81.19, "learning_rate": 3.118190089031187e-06, "loss": 0.0, "step": 376650 }, { "epoch": 81.2, "learning_rate": 3.114746371726773e-06, "loss": 0.0, "step": 376700 }, { "epoch": 81.21, "learning_rate": 3.1113043367649498e-06, "loss": 0.0, "step": 376750 }, { "epoch": 81.22, "learning_rate": 3.10786398463293e-06, "loss": 0.0, "step": 376800 }, { "epoch": 81.24, "learning_rate": 3.1044253158176946e-06, "loss": 0.0, "step": 376850 }, { "epoch": 81.25, "learning_rate": 3.1009883308059708e-06, "loss": 0.0, "step": 376900 }, { "epoch": 81.26, "learning_rate": 3.0975530300842708e-06, "loss": 0.0, "step": 376950 }, { "epoch": 81.27, "learning_rate": 3.094119414138853e-06, "loss": 0.0, "step": 377000 }, { "epoch": 81.28, "learning_rate": 3.0906874834557447e-06, "loss": 0.0, "step": 377050 }, { "epoch": 81.29, "learning_rate": 3.0872572385207277e-06, "loss": 0.0, "step": 377100 }, { "epoch": 81.3, "learning_rate": 3.0838286798193498e-06, "loss": 0.0, "step": 377150 }, { "epoch": 81.31, "learning_rate": 3.080401807836918e-06, "loss": 0.0, "step": 377200 }, { "epoch": 81.32, "learning_rate": 3.076976623058505e-06, "loss": 0.0, "step": 377250 }, { "epoch": 81.33, "learning_rate": 3.0735531259689386e-06, "loss": 0.0, "step": 377300 }, { "epoch": 81.34, "learning_rate": 3.0701313170528117e-06, "loss": 0.0, "step": 377350 }, { "epoch": 81.35, "learning_rate": 3.066711196794477e-06, "loss": 0.0, "step": 377400 }, { "epoch": 81.36, "learning_rate": 3.0632927656780464e-06, "loss": 0.0, "step": 377450 }, { "epoch": 81.38, "learning_rate": 3.059876024187395e-06, "loss": 0.0, "step": 377500 }, { "epoch": 81.39, "learning_rate": 3.0564609728061617e-06, "loss": 0.0, "step": 377550 }, { "epoch": 81.4, "learning_rate": 3.053047612017737e-06, "loss": 0.0, "step": 377600 }, { "epoch": 81.41, "learning_rate": 3.0496359423052835e-06, "loss": 0.0, "step": 377650 }, { "epoch": 81.42, "learning_rate": 3.046225964151714e-06, "loss": 0.0, "step": 377700 }, { "epoch": 81.43, "learning_rate": 3.042817678039707e-06, "loss": 0.0, "step": 377750 }, { "epoch": 81.44, "learning_rate": 3.039411084451709e-06, "loss": 0.0, "step": 377800 }, { "epoch": 81.45, "learning_rate": 3.0360061838699082e-06, "loss": 0.0, "step": 377850 }, { "epoch": 81.46, "learning_rate": 3.0326029767762703e-06, "loss": 0.0, "step": 377900 }, { "epoch": 81.47, "learning_rate": 3.0292014636525085e-06, "loss": 0.0, "step": 377950 }, { "epoch": 81.48, "learning_rate": 3.0258016449801106e-06, "loss": 0.0, "step": 378000 }, { "epoch": 81.49, "learning_rate": 3.0224035212403172e-06, "loss": 0.0, "step": 378050 }, { "epoch": 81.5, "learning_rate": 3.019007092914122e-06, "loss": 0.0, "step": 378100 }, { "epoch": 81.52, "learning_rate": 3.015612360482286e-06, "loss": 0.0, "step": 378150 }, { "epoch": 81.53, "learning_rate": 3.012219324425328e-06, "loss": 0.0, "step": 378200 }, { "epoch": 81.54, "learning_rate": 3.0088279852235338e-06, "loss": 0.0, "step": 378250 }, { "epoch": 81.55, "learning_rate": 3.005438343356946e-06, "loss": 0.0, "step": 378300 }, { "epoch": 81.56, "learning_rate": 3.0020503993053535e-06, "loss": 0.0, "step": 378350 }, { "epoch": 81.57, "learning_rate": 2.9986641535483154e-06, "loss": 0.0, "step": 378400 }, { "epoch": 81.58, "learning_rate": 2.995279606565161e-06, "loss": 0.0, "step": 378450 }, { "epoch": 81.59, "learning_rate": 2.9918967588349614e-06, "loss": 0.0, "step": 378500 }, { "epoch": 81.6, "learning_rate": 2.9885156108365576e-06, "loss": 0.0, "step": 378550 }, { "epoch": 81.61, "learning_rate": 2.9851361630485456e-06, "loss": 0.0, "step": 378600 }, { "epoch": 81.62, "learning_rate": 2.9817584159492804e-06, "loss": 0.0, "step": 378650 }, { "epoch": 81.63, "learning_rate": 2.9783823700168796e-06, "loss": 0.0, "step": 378700 }, { "epoch": 81.64, "learning_rate": 2.9750080257292184e-06, "loss": 0.0, "step": 378750 }, { "epoch": 81.66, "learning_rate": 2.971635383563929e-06, "loss": 0.0, "step": 378800 }, { "epoch": 81.67, "learning_rate": 2.968264443998407e-06, "loss": 0.0, "step": 378850 }, { "epoch": 81.68, "learning_rate": 2.9648952075098027e-06, "loss": 0.0, "step": 378900 }, { "epoch": 81.69, "learning_rate": 2.961527674575027e-06, "loss": 0.0, "step": 378950 }, { "epoch": 81.7, "learning_rate": 2.9581618456707526e-06, "loss": 0.0, "step": 379000 }, { "epoch": 81.71, "learning_rate": 2.954797721273407e-06, "loss": 0.0, "step": 379050 }, { "epoch": 81.72, "learning_rate": 2.9514353018591767e-06, "loss": 0.0, "step": 379100 }, { "epoch": 81.73, "learning_rate": 2.9480745879040082e-06, "loss": 0.0, "step": 379150 }, { "epoch": 81.74, "learning_rate": 2.944715579883607e-06, "loss": 0.0, "step": 379200 }, { "epoch": 81.75, "learning_rate": 2.9413582782734366e-06, "loss": 0.0, "step": 379250 }, { "epoch": 81.76, "learning_rate": 2.938002683548718e-06, "loss": 0.0, "step": 379300 }, { "epoch": 81.77, "learning_rate": 2.934648796184432e-06, "loss": 0.0, "step": 379350 }, { "epoch": 81.78, "learning_rate": 2.931296616655318e-06, "loss": 0.0, "step": 379400 }, { "epoch": 81.8, "learning_rate": 2.9279461454358657e-06, "loss": 0.0, "step": 379450 }, { "epoch": 81.81, "learning_rate": 2.9245973830003446e-06, "loss": 0.0, "step": 379500 }, { "epoch": 81.82, "learning_rate": 2.9212503298227557e-06, "loss": 0.0, "step": 379550 }, { "epoch": 81.83, "learning_rate": 2.9179049863768726e-06, "loss": 0.0, "step": 379600 }, { "epoch": 81.84, "learning_rate": 2.914561353136222e-06, "loss": 0.0, "step": 379650 }, { "epoch": 81.85, "learning_rate": 2.9112194305740988e-06, "loss": 0.0, "step": 379700 }, { "epoch": 81.86, "learning_rate": 2.9078792191635424e-06, "loss": 0.0, "step": 379750 }, { "epoch": 81.87, "learning_rate": 2.90454071937736e-06, "loss": 0.0, "step": 379800 }, { "epoch": 81.88, "learning_rate": 2.9012039316881065e-06, "loss": 0.0, "step": 379850 }, { "epoch": 81.89, "learning_rate": 2.8978688565680964e-06, "loss": 0.0, "step": 379900 }, { "epoch": 81.9, "learning_rate": 2.894535494489417e-06, "loss": 0.0, "step": 379950 }, { "epoch": 81.91, "learning_rate": 2.891203845923894e-06, "loss": 0.0, "step": 380000 } ], "max_steps": 463900, "num_train_epochs": 100, "total_flos": 2.952231674688737e+17, "trial_name": null, "trial_params": null }