{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.999959474245883, "global_step": 92533, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.5000000000000004e-07, "loss": 4.9964, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.5e-06, "loss": 4.5118, "step": 10 }, { "epoch": 0.0, "learning_rate": 5e-06, "loss": 3.7526, "step": 20 }, { "epoch": 0.0, "learning_rate": 7.5e-06, "loss": 3.0438, "step": 30 }, { "epoch": 0.0, "learning_rate": 1e-05, "loss": 2.7736, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.25e-05, "loss": 2.7086, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.5e-05, "loss": 2.7405, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.75e-05, "loss": 2.6797, "step": 70 }, { "epoch": 0.0, "learning_rate": 2e-05, "loss": 2.7344, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.25e-05, "loss": 2.5937, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.5e-05, "loss": 2.5583, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.7500000000000004e-05, "loss": 2.5879, "step": 110 }, { "epoch": 0.01, "learning_rate": 3e-05, "loss": 2.5027, "step": 120 }, { "epoch": 0.01, "learning_rate": 3.2500000000000004e-05, "loss": 2.5664, "step": 130 }, { "epoch": 0.01, "learning_rate": 3.5e-05, "loss": 2.5603, "step": 140 }, { "epoch": 0.01, "learning_rate": 3.7500000000000003e-05, "loss": 2.5154, "step": 150 }, { "epoch": 0.01, "learning_rate": 4e-05, "loss": 2.5236, "step": 160 }, { "epoch": 0.01, "learning_rate": 4.25e-05, "loss": 2.5099, "step": 170 }, { "epoch": 0.01, "learning_rate": 4.5e-05, "loss": 2.5236, "step": 180 }, { "epoch": 0.01, "learning_rate": 4.75e-05, "loss": 2.5865, "step": 190 }, { "epoch": 0.01, "learning_rate": 5e-05, "loss": 2.5606, "step": 200 }, { "epoch": 0.01, "learning_rate": 4.9997295250459805e-05, "loss": 2.5648, "step": 210 }, { "epoch": 0.01, "learning_rate": 4.999459050091962e-05, "loss": 2.4755, "step": 220 }, { "epoch": 0.01, "learning_rate": 4.9991885751379425e-05, "loss": 2.5267, "step": 230 }, { "epoch": 0.01, "learning_rate": 4.998918100183923e-05, "loss": 2.536, "step": 240 }, { "epoch": 0.01, "learning_rate": 4.9986476252299044e-05, "loss": 2.5486, "step": 250 }, { "epoch": 0.01, "learning_rate": 4.998377150275885e-05, "loss": 2.5042, "step": 260 }, { "epoch": 0.01, "learning_rate": 4.998106675321865e-05, "loss": 2.4587, "step": 270 }, { "epoch": 0.02, "learning_rate": 4.9978362003678466e-05, "loss": 2.5262, "step": 280 }, { "epoch": 0.02, "learning_rate": 4.997565725413827e-05, "loss": 2.5196, "step": 290 }, { "epoch": 0.02, "learning_rate": 4.997295250459807e-05, "loss": 2.4566, "step": 300 }, { "epoch": 0.02, "learning_rate": 4.997024775505789e-05, "loss": 2.4334, "step": 310 }, { "epoch": 0.02, "learning_rate": 4.996754300551769e-05, "loss": 2.4275, "step": 320 }, { "epoch": 0.02, "learning_rate": 4.99648382559775e-05, "loss": 2.3684, "step": 330 }, { "epoch": 0.02, "learning_rate": 4.996213350643731e-05, "loss": 2.4606, "step": 340 }, { "epoch": 0.02, "learning_rate": 4.995942875689711e-05, "loss": 2.4487, "step": 350 }, { "epoch": 0.02, "learning_rate": 4.995672400735692e-05, "loss": 2.4389, "step": 360 }, { "epoch": 0.02, "learning_rate": 4.995401925781673e-05, "loss": 2.4623, "step": 370 }, { "epoch": 0.02, "learning_rate": 4.9951314508276535e-05, "loss": 2.4587, "step": 380 }, { "epoch": 0.02, "learning_rate": 4.9948609758736345e-05, "loss": 2.5758, "step": 390 }, { "epoch": 0.02, "learning_rate": 4.9945905009196155e-05, "loss": 2.4081, "step": 400 }, { "epoch": 0.02, "learning_rate": 4.994320025965596e-05, "loss": 2.4782, "step": 410 }, { "epoch": 0.02, "learning_rate": 4.994049551011577e-05, "loss": 2.4133, "step": 420 }, { "epoch": 0.02, "learning_rate": 4.993779076057558e-05, "loss": 2.4387, "step": 430 }, { "epoch": 0.02, "learning_rate": 4.993508601103538e-05, "loss": 2.518, "step": 440 }, { "epoch": 0.02, "learning_rate": 4.993238126149519e-05, "loss": 2.5378, "step": 450 }, { "epoch": 0.02, "learning_rate": 4.9929676511955e-05, "loss": 2.4558, "step": 460 }, { "epoch": 0.03, "learning_rate": 4.99269717624148e-05, "loss": 2.3859, "step": 470 }, { "epoch": 0.03, "learning_rate": 4.992426701287461e-05, "loss": 2.3808, "step": 480 }, { "epoch": 0.03, "learning_rate": 4.9921562263334414e-05, "loss": 2.4066, "step": 490 }, { "epoch": 0.03, "learning_rate": 4.9918857513794224e-05, "loss": 2.445, "step": 500 }, { "epoch": 0.03, "learning_rate": 4.9916152764254034e-05, "loss": 2.4505, "step": 510 }, { "epoch": 0.03, "learning_rate": 4.9913448014713837e-05, "loss": 2.385, "step": 520 }, { "epoch": 0.03, "learning_rate": 4.9910743265173646e-05, "loss": 2.3818, "step": 530 }, { "epoch": 0.03, "learning_rate": 4.9908038515633456e-05, "loss": 2.4285, "step": 540 }, { "epoch": 0.03, "learning_rate": 4.990533376609326e-05, "loss": 2.4128, "step": 550 }, { "epoch": 0.03, "learning_rate": 4.990262901655307e-05, "loss": 2.368, "step": 560 }, { "epoch": 0.03, "learning_rate": 4.989992426701288e-05, "loss": 2.3208, "step": 570 }, { "epoch": 0.03, "learning_rate": 4.989721951747268e-05, "loss": 2.5087, "step": 580 }, { "epoch": 0.03, "learning_rate": 4.989451476793249e-05, "loss": 2.4243, "step": 590 }, { "epoch": 0.03, "learning_rate": 4.98918100183923e-05, "loss": 2.3673, "step": 600 }, { "epoch": 0.03, "learning_rate": 4.98891052688521e-05, "loss": 2.4617, "step": 610 }, { "epoch": 0.03, "learning_rate": 4.988640051931191e-05, "loss": 2.3699, "step": 620 }, { "epoch": 0.03, "learning_rate": 4.988369576977172e-05, "loss": 2.3196, "step": 630 }, { "epoch": 0.03, "learning_rate": 4.9880991020231525e-05, "loss": 2.4384, "step": 640 }, { "epoch": 0.04, "learning_rate": 4.9878286270691335e-05, "loss": 2.4522, "step": 650 }, { "epoch": 0.04, "learning_rate": 4.9875581521151145e-05, "loss": 2.3065, "step": 660 }, { "epoch": 0.04, "learning_rate": 4.987287677161095e-05, "loss": 2.4238, "step": 670 }, { "epoch": 0.04, "learning_rate": 4.987017202207076e-05, "loss": 2.4371, "step": 680 }, { "epoch": 0.04, "learning_rate": 4.986746727253057e-05, "loss": 2.3898, "step": 690 }, { "epoch": 0.04, "learning_rate": 4.986476252299037e-05, "loss": 2.4442, "step": 700 }, { "epoch": 0.04, "learning_rate": 4.986205777345018e-05, "loss": 2.4145, "step": 710 }, { "epoch": 0.04, "learning_rate": 4.985935302390999e-05, "loss": 2.3808, "step": 720 }, { "epoch": 0.04, "learning_rate": 4.985664827436979e-05, "loss": 2.3753, "step": 730 }, { "epoch": 0.04, "learning_rate": 4.985394352482961e-05, "loss": 2.3327, "step": 740 }, { "epoch": 0.04, "learning_rate": 4.985123877528941e-05, "loss": 2.3, "step": 750 }, { "epoch": 0.04, "learning_rate": 4.9848534025749214e-05, "loss": 2.4109, "step": 760 }, { "epoch": 0.04, "learning_rate": 4.984582927620903e-05, "loss": 2.3665, "step": 770 }, { "epoch": 0.04, "learning_rate": 4.984312452666883e-05, "loss": 2.3479, "step": 780 }, { "epoch": 0.04, "learning_rate": 4.9840419777128636e-05, "loss": 2.4957, "step": 790 }, { "epoch": 0.04, "learning_rate": 4.983771502758845e-05, "loss": 2.4682, "step": 800 }, { "epoch": 0.04, "learning_rate": 4.9835010278048255e-05, "loss": 2.3642, "step": 810 }, { "epoch": 0.04, "learning_rate": 4.983230552850806e-05, "loss": 2.3265, "step": 820 }, { "epoch": 0.04, "learning_rate": 4.9829600778967875e-05, "loss": 2.3292, "step": 830 }, { "epoch": 0.05, "learning_rate": 4.982689602942768e-05, "loss": 2.2998, "step": 840 }, { "epoch": 0.05, "learning_rate": 4.982419127988748e-05, "loss": 2.3412, "step": 850 }, { "epoch": 0.05, "learning_rate": 4.98214865303473e-05, "loss": 2.4435, "step": 860 }, { "epoch": 0.05, "learning_rate": 4.98187817808071e-05, "loss": 2.3583, "step": 870 }, { "epoch": 0.05, "learning_rate": 4.98160770312669e-05, "loss": 2.4743, "step": 880 }, { "epoch": 0.05, "learning_rate": 4.981337228172672e-05, "loss": 2.2989, "step": 890 }, { "epoch": 0.05, "learning_rate": 4.981066753218652e-05, "loss": 2.3116, "step": 900 }, { "epoch": 0.05, "learning_rate": 4.9807962782646325e-05, "loss": 2.2918, "step": 910 }, { "epoch": 0.05, "learning_rate": 4.980525803310614e-05, "loss": 2.2938, "step": 920 }, { "epoch": 0.05, "learning_rate": 4.9802553283565944e-05, "loss": 2.2815, "step": 930 }, { "epoch": 0.05, "learning_rate": 4.979984853402575e-05, "loss": 2.4071, "step": 940 }, { "epoch": 0.05, "learning_rate": 4.9797143784485563e-05, "loss": 2.3791, "step": 950 }, { "epoch": 0.05, "learning_rate": 4.9794439034945366e-05, "loss": 2.3402, "step": 960 }, { "epoch": 0.05, "learning_rate": 4.9791734285405176e-05, "loss": 2.3781, "step": 970 }, { "epoch": 0.05, "learning_rate": 4.9789029535864986e-05, "loss": 2.3451, "step": 980 }, { "epoch": 0.05, "learning_rate": 4.978632478632479e-05, "loss": 2.2885, "step": 990 }, { "epoch": 0.05, "learning_rate": 4.97836200367846e-05, "loss": 2.3021, "step": 1000 }, { "epoch": 0.05, "learning_rate": 4.978091528724441e-05, "loss": 2.3877, "step": 1010 }, { "epoch": 0.06, "learning_rate": 4.977821053770421e-05, "loss": 2.3151, "step": 1020 }, { "epoch": 0.06, "learning_rate": 4.977550578816402e-05, "loss": 2.3931, "step": 1030 }, { "epoch": 0.06, "learning_rate": 4.977280103862383e-05, "loss": 2.3296, "step": 1040 }, { "epoch": 0.06, "learning_rate": 4.977009628908363e-05, "loss": 2.3231, "step": 1050 }, { "epoch": 0.06, "learning_rate": 4.976739153954344e-05, "loss": 2.2743, "step": 1060 }, { "epoch": 0.06, "learning_rate": 4.9764686790003245e-05, "loss": 2.2891, "step": 1070 }, { "epoch": 0.06, "learning_rate": 4.9761982040463055e-05, "loss": 2.285, "step": 1080 }, { "epoch": 0.06, "learning_rate": 4.9759277290922865e-05, "loss": 2.2899, "step": 1090 }, { "epoch": 0.06, "learning_rate": 4.975657254138267e-05, "loss": 2.3141, "step": 1100 }, { "epoch": 0.06, "learning_rate": 4.975386779184248e-05, "loss": 2.3264, "step": 1110 }, { "epoch": 0.06, "learning_rate": 4.975116304230229e-05, "loss": 2.3372, "step": 1120 }, { "epoch": 0.06, "learning_rate": 4.974845829276209e-05, "loss": 2.4023, "step": 1130 }, { "epoch": 0.06, "learning_rate": 4.97457535432219e-05, "loss": 2.2754, "step": 1140 }, { "epoch": 0.06, "learning_rate": 4.974304879368171e-05, "loss": 2.2779, "step": 1150 }, { "epoch": 0.06, "learning_rate": 4.974034404414151e-05, "loss": 2.2921, "step": 1160 }, { "epoch": 0.06, "learning_rate": 4.973763929460132e-05, "loss": 2.3031, "step": 1170 }, { "epoch": 0.06, "learning_rate": 4.973493454506113e-05, "loss": 2.3194, "step": 1180 }, { "epoch": 0.06, "learning_rate": 4.9732229795520934e-05, "loss": 2.3449, "step": 1190 }, { "epoch": 0.06, "learning_rate": 4.9729525045980744e-05, "loss": 2.3568, "step": 1200 }, { "epoch": 0.07, "learning_rate": 4.972682029644055e-05, "loss": 2.2834, "step": 1210 }, { "epoch": 0.07, "learning_rate": 4.9724115546900356e-05, "loss": 2.3349, "step": 1220 }, { "epoch": 0.07, "learning_rate": 4.9721410797360166e-05, "loss": 2.3737, "step": 1230 }, { "epoch": 0.07, "learning_rate": 4.9718706047819975e-05, "loss": 2.2718, "step": 1240 }, { "epoch": 0.07, "learning_rate": 4.971600129827978e-05, "loss": 2.3483, "step": 1250 }, { "epoch": 0.07, "learning_rate": 4.971329654873959e-05, "loss": 2.2749, "step": 1260 }, { "epoch": 0.07, "learning_rate": 4.97105917991994e-05, "loss": 2.3068, "step": 1270 }, { "epoch": 0.07, "learning_rate": 4.97078870496592e-05, "loss": 2.3303, "step": 1280 }, { "epoch": 0.07, "learning_rate": 4.970518230011901e-05, "loss": 2.2087, "step": 1290 }, { "epoch": 0.07, "learning_rate": 4.970247755057882e-05, "loss": 2.3, "step": 1300 }, { "epoch": 0.07, "learning_rate": 4.969977280103862e-05, "loss": 2.3792, "step": 1310 }, { "epoch": 0.07, "learning_rate": 4.969706805149843e-05, "loss": 2.3288, "step": 1320 }, { "epoch": 0.07, "learning_rate": 4.969436330195824e-05, "loss": 2.3004, "step": 1330 }, { "epoch": 0.07, "learning_rate": 4.9691658552418045e-05, "loss": 2.3269, "step": 1340 }, { "epoch": 0.07, "learning_rate": 4.9688953802877854e-05, "loss": 2.2713, "step": 1350 }, { "epoch": 0.07, "learning_rate": 4.9686249053337664e-05, "loss": 2.2668, "step": 1360 }, { "epoch": 0.07, "learning_rate": 4.968354430379747e-05, "loss": 2.3268, "step": 1370 }, { "epoch": 0.07, "learning_rate": 4.9680839554257283e-05, "loss": 2.2619, "step": 1380 }, { "epoch": 0.08, "learning_rate": 4.9678134804717086e-05, "loss": 2.3143, "step": 1390 }, { "epoch": 0.08, "learning_rate": 4.967543005517689e-05, "loss": 2.3761, "step": 1400 }, { "epoch": 0.08, "learning_rate": 4.9672725305636706e-05, "loss": 2.3771, "step": 1410 }, { "epoch": 0.08, "learning_rate": 4.967002055609651e-05, "loss": 2.3782, "step": 1420 }, { "epoch": 0.08, "learning_rate": 4.966731580655631e-05, "loss": 2.2408, "step": 1430 }, { "epoch": 0.08, "learning_rate": 4.966461105701613e-05, "loss": 2.266, "step": 1440 }, { "epoch": 0.08, "learning_rate": 4.966190630747593e-05, "loss": 2.3044, "step": 1450 }, { "epoch": 0.08, "learning_rate": 4.9659201557935734e-05, "loss": 2.2969, "step": 1460 }, { "epoch": 0.08, "learning_rate": 4.965649680839555e-05, "loss": 2.3389, "step": 1470 }, { "epoch": 0.08, "learning_rate": 4.965379205885535e-05, "loss": 2.3102, "step": 1480 }, { "epoch": 0.08, "learning_rate": 4.9651087309315156e-05, "loss": 2.2465, "step": 1490 }, { "epoch": 0.08, "learning_rate": 4.964838255977497e-05, "loss": 2.2942, "step": 1500 }, { "epoch": 0.08, "learning_rate": 4.9645677810234775e-05, "loss": 2.2684, "step": 1510 }, { "epoch": 0.08, "learning_rate": 4.964297306069458e-05, "loss": 2.2857, "step": 1520 }, { "epoch": 0.08, "learning_rate": 4.9640268311154394e-05, "loss": 2.3909, "step": 1530 }, { "epoch": 0.08, "learning_rate": 4.96375635616142e-05, "loss": 2.3169, "step": 1540 }, { "epoch": 0.08, "learning_rate": 4.9634858812074e-05, "loss": 2.3132, "step": 1550 }, { "epoch": 0.08, "learning_rate": 4.9632154062533816e-05, "loss": 2.3025, "step": 1560 }, { "epoch": 0.08, "learning_rate": 4.962944931299362e-05, "loss": 2.228, "step": 1570 }, { "epoch": 0.09, "learning_rate": 4.962674456345343e-05, "loss": 2.2536, "step": 1580 }, { "epoch": 0.09, "learning_rate": 4.962403981391324e-05, "loss": 2.3005, "step": 1590 }, { "epoch": 0.09, "learning_rate": 4.962133506437304e-05, "loss": 2.3427, "step": 1600 }, { "epoch": 0.09, "learning_rate": 4.961863031483285e-05, "loss": 2.2771, "step": 1610 }, { "epoch": 0.09, "learning_rate": 4.9615925565292654e-05, "loss": 2.3221, "step": 1620 }, { "epoch": 0.09, "learning_rate": 4.9613220815752464e-05, "loss": 2.2261, "step": 1630 }, { "epoch": 0.09, "learning_rate": 4.961051606621227e-05, "loss": 2.2667, "step": 1640 }, { "epoch": 0.09, "learning_rate": 4.9607811316672076e-05, "loss": 2.2536, "step": 1650 }, { "epoch": 0.09, "learning_rate": 4.9605106567131886e-05, "loss": 2.2601, "step": 1660 }, { "epoch": 0.09, "learning_rate": 4.9602401817591695e-05, "loss": 2.2748, "step": 1670 }, { "epoch": 0.09, "learning_rate": 4.95996970680515e-05, "loss": 2.1729, "step": 1680 }, { "epoch": 0.09, "learning_rate": 4.959699231851131e-05, "loss": 2.3266, "step": 1690 }, { "epoch": 0.09, "learning_rate": 4.959428756897112e-05, "loss": 2.2304, "step": 1700 }, { "epoch": 0.09, "learning_rate": 4.959158281943092e-05, "loss": 2.3254, "step": 1710 }, { "epoch": 0.09, "learning_rate": 4.958887806989073e-05, "loss": 2.374, "step": 1720 }, { "epoch": 0.09, "learning_rate": 4.958617332035054e-05, "loss": 2.2244, "step": 1730 }, { "epoch": 0.09, "learning_rate": 4.958346857081034e-05, "loss": 2.2105, "step": 1740 }, { "epoch": 0.09, "learning_rate": 4.958076382127015e-05, "loss": 2.4058, "step": 1750 }, { "epoch": 0.1, "learning_rate": 4.957805907172996e-05, "loss": 2.2942, "step": 1760 }, { "epoch": 0.1, "learning_rate": 4.9575354322189765e-05, "loss": 2.2506, "step": 1770 }, { "epoch": 0.1, "learning_rate": 4.9572649572649575e-05, "loss": 2.3281, "step": 1780 }, { "epoch": 0.1, "learning_rate": 4.9569944823109384e-05, "loss": 2.2814, "step": 1790 }, { "epoch": 0.1, "learning_rate": 4.956724007356919e-05, "loss": 2.3081, "step": 1800 }, { "epoch": 0.1, "learning_rate": 4.9564535324029e-05, "loss": 2.2059, "step": 1810 }, { "epoch": 0.1, "learning_rate": 4.9561830574488806e-05, "loss": 2.2145, "step": 1820 }, { "epoch": 0.1, "learning_rate": 4.955912582494861e-05, "loss": 2.3385, "step": 1830 }, { "epoch": 0.1, "learning_rate": 4.955642107540842e-05, "loss": 2.2641, "step": 1840 }, { "epoch": 0.1, "learning_rate": 4.955371632586823e-05, "loss": 2.287, "step": 1850 }, { "epoch": 0.1, "learning_rate": 4.955101157632803e-05, "loss": 2.2525, "step": 1860 }, { "epoch": 0.1, "learning_rate": 4.954830682678784e-05, "loss": 2.1847, "step": 1870 }, { "epoch": 0.1, "learning_rate": 4.954560207724765e-05, "loss": 2.2099, "step": 1880 }, { "epoch": 0.1, "learning_rate": 4.9542897327707454e-05, "loss": 2.336, "step": 1890 }, { "epoch": 0.1, "learning_rate": 4.954019257816726e-05, "loss": 2.2327, "step": 1900 }, { "epoch": 0.1, "learning_rate": 4.953748782862707e-05, "loss": 2.2394, "step": 1910 }, { "epoch": 0.1, "learning_rate": 4.9534783079086876e-05, "loss": 2.2039, "step": 1920 }, { "epoch": 0.1, "learning_rate": 4.9532078329546685e-05, "loss": 2.24, "step": 1930 }, { "epoch": 0.1, "learning_rate": 4.9529373580006495e-05, "loss": 2.2389, "step": 1940 }, { "epoch": 0.11, "learning_rate": 4.95266688304663e-05, "loss": 2.3995, "step": 1950 }, { "epoch": 0.11, "learning_rate": 4.952396408092611e-05, "loss": 2.2585, "step": 1960 }, { "epoch": 0.11, "learning_rate": 4.952125933138592e-05, "loss": 2.2, "step": 1970 }, { "epoch": 0.11, "learning_rate": 4.951855458184572e-05, "loss": 2.3263, "step": 1980 }, { "epoch": 0.11, "learning_rate": 4.9515849832305536e-05, "loss": 2.3269, "step": 1990 }, { "epoch": 0.11, "learning_rate": 4.951314508276534e-05, "loss": 2.3041, "step": 2000 }, { "epoch": 0.11, "learning_rate": 4.951044033322514e-05, "loss": 2.3077, "step": 2010 }, { "epoch": 0.11, "learning_rate": 4.950773558368496e-05, "loss": 2.0697, "step": 2020 }, { "epoch": 0.11, "learning_rate": 4.950503083414476e-05, "loss": 2.219, "step": 2030 }, { "epoch": 0.11, "learning_rate": 4.9502326084604564e-05, "loss": 2.2175, "step": 2040 }, { "epoch": 0.11, "learning_rate": 4.949962133506438e-05, "loss": 2.2032, "step": 2050 }, { "epoch": 0.11, "learning_rate": 4.9496916585524184e-05, "loss": 2.3012, "step": 2060 }, { "epoch": 0.11, "learning_rate": 4.9494211835983987e-05, "loss": 2.2117, "step": 2070 }, { "epoch": 0.11, "learning_rate": 4.94915070864438e-05, "loss": 2.197, "step": 2080 }, { "epoch": 0.11, "learning_rate": 4.9488802336903606e-05, "loss": 2.235, "step": 2090 }, { "epoch": 0.11, "learning_rate": 4.948609758736341e-05, "loss": 2.3396, "step": 2100 }, { "epoch": 0.11, "learning_rate": 4.9483392837823225e-05, "loss": 2.1886, "step": 2110 }, { "epoch": 0.11, "learning_rate": 4.948068808828303e-05, "loss": 2.2429, "step": 2120 }, { "epoch": 0.12, "learning_rate": 4.947798333874283e-05, "loss": 2.1844, "step": 2130 }, { "epoch": 0.12, "learning_rate": 4.947527858920265e-05, "loss": 2.2821, "step": 2140 }, { "epoch": 0.12, "learning_rate": 4.947257383966245e-05, "loss": 2.2655, "step": 2150 }, { "epoch": 0.12, "learning_rate": 4.946986909012225e-05, "loss": 2.2174, "step": 2160 }, { "epoch": 0.12, "learning_rate": 4.946716434058206e-05, "loss": 2.2452, "step": 2170 }, { "epoch": 0.12, "learning_rate": 4.946445959104187e-05, "loss": 2.344, "step": 2180 }, { "epoch": 0.12, "learning_rate": 4.9461754841501675e-05, "loss": 2.2585, "step": 2190 }, { "epoch": 0.12, "learning_rate": 4.9459050091961485e-05, "loss": 2.21, "step": 2200 }, { "epoch": 0.12, "learning_rate": 4.9456345342421295e-05, "loss": 2.2857, "step": 2210 }, { "epoch": 0.12, "learning_rate": 4.9453640592881104e-05, "loss": 2.3012, "step": 2220 }, { "epoch": 0.12, "learning_rate": 4.945093584334091e-05, "loss": 2.2224, "step": 2230 }, { "epoch": 0.12, "learning_rate": 4.944823109380072e-05, "loss": 2.156, "step": 2240 }, { "epoch": 0.12, "learning_rate": 4.9445526344260526e-05, "loss": 2.1973, "step": 2250 }, { "epoch": 0.12, "learning_rate": 4.944282159472033e-05, "loss": 2.2781, "step": 2260 }, { "epoch": 0.12, "learning_rate": 4.944011684518014e-05, "loss": 2.274, "step": 2270 }, { "epoch": 0.12, "learning_rate": 4.943741209563995e-05, "loss": 2.2558, "step": 2280 }, { "epoch": 0.12, "learning_rate": 4.943470734609975e-05, "loss": 2.1752, "step": 2290 }, { "epoch": 0.12, "learning_rate": 4.943200259655956e-05, "loss": 2.1442, "step": 2300 }, { "epoch": 0.12, "learning_rate": 4.942929784701937e-05, "loss": 2.2182, "step": 2310 }, { "epoch": 0.13, "learning_rate": 4.9426593097479174e-05, "loss": 2.2071, "step": 2320 }, { "epoch": 0.13, "learning_rate": 4.942388834793898e-05, "loss": 2.1908, "step": 2330 }, { "epoch": 0.13, "learning_rate": 4.942118359839879e-05, "loss": 2.2235, "step": 2340 }, { "epoch": 0.13, "learning_rate": 4.9418478848858596e-05, "loss": 2.2581, "step": 2350 }, { "epoch": 0.13, "learning_rate": 4.9415774099318405e-05, "loss": 2.1659, "step": 2360 }, { "epoch": 0.13, "learning_rate": 4.9413069349778215e-05, "loss": 2.1731, "step": 2370 }, { "epoch": 0.13, "learning_rate": 4.941036460023802e-05, "loss": 2.2329, "step": 2380 }, { "epoch": 0.13, "learning_rate": 4.940765985069783e-05, "loss": 2.1759, "step": 2390 }, { "epoch": 0.13, "learning_rate": 4.940495510115764e-05, "loss": 2.2921, "step": 2400 }, { "epoch": 0.13, "learning_rate": 4.940225035161744e-05, "loss": 2.2203, "step": 2410 }, { "epoch": 0.13, "learning_rate": 4.939954560207725e-05, "loss": 2.2249, "step": 2420 }, { "epoch": 0.13, "learning_rate": 4.939684085253706e-05, "loss": 2.2723, "step": 2430 }, { "epoch": 0.13, "learning_rate": 4.939413610299686e-05, "loss": 2.314, "step": 2440 }, { "epoch": 0.13, "learning_rate": 4.939143135345667e-05, "loss": 2.2228, "step": 2450 }, { "epoch": 0.13, "learning_rate": 4.938872660391648e-05, "loss": 2.1519, "step": 2460 }, { "epoch": 0.13, "learning_rate": 4.9386021854376284e-05, "loss": 2.1572, "step": 2470 }, { "epoch": 0.13, "learning_rate": 4.9383317104836094e-05, "loss": 2.277, "step": 2480 }, { "epoch": 0.13, "learning_rate": 4.9380612355295904e-05, "loss": 2.1876, "step": 2490 }, { "epoch": 0.14, "learning_rate": 4.9377907605755707e-05, "loss": 2.2588, "step": 2500 }, { "epoch": 0.14, "learning_rate": 4.9375202856215516e-05, "loss": 2.1396, "step": 2510 }, { "epoch": 0.14, "learning_rate": 4.9372498106675326e-05, "loss": 2.2556, "step": 2520 }, { "epoch": 0.14, "learning_rate": 4.936979335713513e-05, "loss": 2.1822, "step": 2530 }, { "epoch": 0.14, "learning_rate": 4.936708860759494e-05, "loss": 2.2173, "step": 2540 }, { "epoch": 0.14, "learning_rate": 4.936438385805475e-05, "loss": 2.156, "step": 2550 }, { "epoch": 0.14, "learning_rate": 4.936167910851455e-05, "loss": 2.2922, "step": 2560 }, { "epoch": 0.14, "learning_rate": 4.935897435897436e-05, "loss": 2.3035, "step": 2570 }, { "epoch": 0.14, "learning_rate": 4.935626960943417e-05, "loss": 2.154, "step": 2580 }, { "epoch": 0.14, "learning_rate": 4.935356485989397e-05, "loss": 2.1863, "step": 2590 }, { "epoch": 0.14, "learning_rate": 4.935086011035378e-05, "loss": 2.2932, "step": 2600 }, { "epoch": 0.14, "learning_rate": 4.934815536081359e-05, "loss": 2.0817, "step": 2610 }, { "epoch": 0.14, "learning_rate": 4.9345450611273395e-05, "loss": 2.2851, "step": 2620 }, { "epoch": 0.14, "learning_rate": 4.934274586173321e-05, "loss": 2.2903, "step": 2630 }, { "epoch": 0.14, "learning_rate": 4.9340041112193015e-05, "loss": 2.1853, "step": 2640 }, { "epoch": 0.14, "learning_rate": 4.933733636265282e-05, "loss": 2.1486, "step": 2650 }, { "epoch": 0.14, "learning_rate": 4.9334631613112634e-05, "loss": 2.1917, "step": 2660 }, { "epoch": 0.14, "learning_rate": 4.933192686357244e-05, "loss": 2.1511, "step": 2670 }, { "epoch": 0.14, "learning_rate": 4.932922211403224e-05, "loss": 2.2124, "step": 2680 }, { "epoch": 0.15, "learning_rate": 4.9326517364492056e-05, "loss": 2.2817, "step": 2690 }, { "epoch": 0.15, "learning_rate": 4.932381261495186e-05, "loss": 2.327, "step": 2700 }, { "epoch": 0.15, "learning_rate": 4.932110786541166e-05, "loss": 2.224, "step": 2710 }, { "epoch": 0.15, "learning_rate": 4.931840311587148e-05, "loss": 2.2294, "step": 2720 }, { "epoch": 0.15, "learning_rate": 4.931569836633128e-05, "loss": 2.2373, "step": 2730 }, { "epoch": 0.15, "learning_rate": 4.9312993616791084e-05, "loss": 2.1423, "step": 2740 }, { "epoch": 0.15, "learning_rate": 4.9310288867250894e-05, "loss": 2.1715, "step": 2750 }, { "epoch": 0.15, "learning_rate": 4.93075841177107e-05, "loss": 2.2182, "step": 2760 }, { "epoch": 0.15, "learning_rate": 4.9304879368170506e-05, "loss": 2.1761, "step": 2770 }, { "epoch": 0.15, "learning_rate": 4.9302174618630316e-05, "loss": 2.1994, "step": 2780 }, { "epoch": 0.15, "learning_rate": 4.9299469869090125e-05, "loss": 2.2308, "step": 2790 }, { "epoch": 0.15, "learning_rate": 4.929676511954993e-05, "loss": 2.1424, "step": 2800 }, { "epoch": 0.15, "learning_rate": 4.929406037000974e-05, "loss": 2.2646, "step": 2810 }, { "epoch": 0.15, "learning_rate": 4.929135562046955e-05, "loss": 2.1702, "step": 2820 }, { "epoch": 0.15, "learning_rate": 4.928865087092935e-05, "loss": 2.2244, "step": 2830 }, { "epoch": 0.15, "learning_rate": 4.928594612138916e-05, "loss": 2.3073, "step": 2840 }, { "epoch": 0.15, "learning_rate": 4.928324137184897e-05, "loss": 2.2129, "step": 2850 }, { "epoch": 0.15, "learning_rate": 4.928053662230878e-05, "loss": 2.2436, "step": 2860 }, { "epoch": 0.16, "learning_rate": 4.927783187276858e-05, "loss": 2.1745, "step": 2870 }, { "epoch": 0.16, "learning_rate": 4.927512712322839e-05, "loss": 2.2685, "step": 2880 }, { "epoch": 0.16, "learning_rate": 4.92724223736882e-05, "loss": 2.2411, "step": 2890 }, { "epoch": 0.16, "learning_rate": 4.9269717624148004e-05, "loss": 2.2232, "step": 2900 }, { "epoch": 0.16, "learning_rate": 4.9267012874607814e-05, "loss": 2.1313, "step": 2910 }, { "epoch": 0.16, "learning_rate": 4.9264308125067624e-05, "loss": 2.2953, "step": 2920 }, { "epoch": 0.16, "learning_rate": 4.9261603375527427e-05, "loss": 2.1986, "step": 2930 }, { "epoch": 0.16, "learning_rate": 4.9258898625987236e-05, "loss": 2.2235, "step": 2940 }, { "epoch": 0.16, "learning_rate": 4.9256193876447046e-05, "loss": 2.2274, "step": 2950 }, { "epoch": 0.16, "learning_rate": 4.925348912690685e-05, "loss": 2.2546, "step": 2960 }, { "epoch": 0.16, "learning_rate": 4.925078437736666e-05, "loss": 2.1838, "step": 2970 }, { "epoch": 0.16, "learning_rate": 4.924807962782647e-05, "loss": 2.2368, "step": 2980 }, { "epoch": 0.16, "learning_rate": 4.924537487828627e-05, "loss": 2.1892, "step": 2990 }, { "epoch": 0.16, "learning_rate": 4.924267012874608e-05, "loss": 2.245, "step": 3000 }, { "epoch": 0.16, "learning_rate": 4.923996537920589e-05, "loss": 2.1876, "step": 3010 }, { "epoch": 0.16, "learning_rate": 4.923726062966569e-05, "loss": 2.1988, "step": 3020 }, { "epoch": 0.16, "learning_rate": 4.92345558801255e-05, "loss": 2.2589, "step": 3030 }, { "epoch": 0.16, "learning_rate": 4.923185113058531e-05, "loss": 2.1951, "step": 3040 }, { "epoch": 0.16, "learning_rate": 4.9229146381045115e-05, "loss": 2.2083, "step": 3050 }, { "epoch": 0.17, "learning_rate": 4.9226441631504925e-05, "loss": 2.1858, "step": 3060 }, { "epoch": 0.17, "learning_rate": 4.9223736881964735e-05, "loss": 2.1636, "step": 3070 }, { "epoch": 0.17, "learning_rate": 4.922103213242454e-05, "loss": 2.1094, "step": 3080 }, { "epoch": 0.17, "learning_rate": 4.921832738288435e-05, "loss": 2.2163, "step": 3090 }, { "epoch": 0.17, "learning_rate": 4.921562263334416e-05, "loss": 2.1934, "step": 3100 }, { "epoch": 0.17, "learning_rate": 4.921291788380396e-05, "loss": 2.2257, "step": 3110 }, { "epoch": 0.17, "learning_rate": 4.921021313426377e-05, "loss": 2.1992, "step": 3120 }, { "epoch": 0.17, "learning_rate": 4.920750838472358e-05, "loss": 2.0981, "step": 3130 }, { "epoch": 0.17, "learning_rate": 4.920480363518338e-05, "loss": 2.169, "step": 3140 }, { "epoch": 0.17, "learning_rate": 4.920209888564319e-05, "loss": 2.1987, "step": 3150 }, { "epoch": 0.17, "learning_rate": 4.9199394136103e-05, "loss": 2.1939, "step": 3160 }, { "epoch": 0.17, "learning_rate": 4.9196689386562804e-05, "loss": 2.2754, "step": 3170 }, { "epoch": 0.17, "learning_rate": 4.9193984637022614e-05, "loss": 2.1393, "step": 3180 }, { "epoch": 0.17, "learning_rate": 4.919127988748242e-05, "loss": 2.2281, "step": 3190 }, { "epoch": 0.17, "learning_rate": 4.9188575137942226e-05, "loss": 2.2154, "step": 3200 }, { "epoch": 0.17, "learning_rate": 4.9185870388402036e-05, "loss": 2.2656, "step": 3210 }, { "epoch": 0.17, "learning_rate": 4.9183165638861845e-05, "loss": 2.2272, "step": 3220 }, { "epoch": 0.17, "learning_rate": 4.918046088932165e-05, "loss": 2.1755, "step": 3230 }, { "epoch": 0.18, "learning_rate": 4.917775613978146e-05, "loss": 2.2705, "step": 3240 }, { "epoch": 0.18, "learning_rate": 4.917505139024127e-05, "loss": 2.2852, "step": 3250 }, { "epoch": 0.18, "learning_rate": 4.917234664070107e-05, "loss": 2.1994, "step": 3260 }, { "epoch": 0.18, "learning_rate": 4.916964189116089e-05, "loss": 2.1301, "step": 3270 }, { "epoch": 0.18, "learning_rate": 4.916693714162069e-05, "loss": 2.2607, "step": 3280 }, { "epoch": 0.18, "learning_rate": 4.916423239208049e-05, "loss": 2.2228, "step": 3290 }, { "epoch": 0.18, "learning_rate": 4.91615276425403e-05, "loss": 2.2238, "step": 3300 }, { "epoch": 0.18, "learning_rate": 4.915882289300011e-05, "loss": 2.1411, "step": 3310 }, { "epoch": 0.18, "learning_rate": 4.9156118143459915e-05, "loss": 2.1145, "step": 3320 }, { "epoch": 0.18, "learning_rate": 4.9153413393919724e-05, "loss": 2.279, "step": 3330 }, { "epoch": 0.18, "learning_rate": 4.9150708644379534e-05, "loss": 2.1528, "step": 3340 }, { "epoch": 0.18, "learning_rate": 4.914800389483934e-05, "loss": 2.2984, "step": 3350 }, { "epoch": 0.18, "learning_rate": 4.9145299145299147e-05, "loss": 2.1857, "step": 3360 }, { "epoch": 0.18, "learning_rate": 4.9142594395758956e-05, "loss": 2.1755, "step": 3370 }, { "epoch": 0.18, "learning_rate": 4.913988964621876e-05, "loss": 2.273, "step": 3380 }, { "epoch": 0.18, "learning_rate": 4.913718489667857e-05, "loss": 2.136, "step": 3390 }, { "epoch": 0.18, "learning_rate": 4.913448014713838e-05, "loss": 2.1313, "step": 3400 }, { "epoch": 0.18, "learning_rate": 4.913177539759818e-05, "loss": 2.1539, "step": 3410 }, { "epoch": 0.18, "learning_rate": 4.912907064805799e-05, "loss": 2.1286, "step": 3420 }, { "epoch": 0.19, "learning_rate": 4.91263658985178e-05, "loss": 2.0857, "step": 3430 }, { "epoch": 0.19, "learning_rate": 4.9123661148977603e-05, "loss": 2.2199, "step": 3440 }, { "epoch": 0.19, "learning_rate": 4.912095639943741e-05, "loss": 2.1829, "step": 3450 }, { "epoch": 0.19, "learning_rate": 4.911825164989722e-05, "loss": 2.166, "step": 3460 }, { "epoch": 0.19, "learning_rate": 4.9115546900357026e-05, "loss": 2.235, "step": 3470 }, { "epoch": 0.19, "learning_rate": 4.9112842150816835e-05, "loss": 2.1729, "step": 3480 }, { "epoch": 0.19, "learning_rate": 4.9110137401276645e-05, "loss": 2.136, "step": 3490 }, { "epoch": 0.19, "learning_rate": 4.9107432651736455e-05, "loss": 2.1778, "step": 3500 }, { "epoch": 0.19, "learning_rate": 4.910472790219626e-05, "loss": 2.1762, "step": 3510 }, { "epoch": 0.19, "learning_rate": 4.910202315265607e-05, "loss": 2.169, "step": 3520 }, { "epoch": 0.19, "learning_rate": 4.909931840311588e-05, "loss": 2.2395, "step": 3530 }, { "epoch": 0.19, "learning_rate": 4.909661365357568e-05, "loss": 2.207, "step": 3540 }, { "epoch": 0.19, "learning_rate": 4.909390890403549e-05, "loss": 2.2039, "step": 3550 }, { "epoch": 0.19, "learning_rate": 4.90912041544953e-05, "loss": 2.1966, "step": 3560 }, { "epoch": 0.19, "learning_rate": 4.90884994049551e-05, "loss": 2.2251, "step": 3570 }, { "epoch": 0.19, "learning_rate": 4.908579465541491e-05, "loss": 2.1842, "step": 3580 }, { "epoch": 0.19, "learning_rate": 4.908308990587472e-05, "loss": 2.1581, "step": 3590 }, { "epoch": 0.19, "learning_rate": 4.9080385156334524e-05, "loss": 2.1879, "step": 3600 }, { "epoch": 0.2, "learning_rate": 4.9077680406794334e-05, "loss": 2.1926, "step": 3610 }, { "epoch": 0.2, "learning_rate": 4.907497565725414e-05, "loss": 2.2294, "step": 3620 }, { "epoch": 0.2, "learning_rate": 4.9072270907713946e-05, "loss": 2.1543, "step": 3630 }, { "epoch": 0.2, "learning_rate": 4.9069566158173756e-05, "loss": 2.1729, "step": 3640 }, { "epoch": 0.2, "learning_rate": 4.9066861408633565e-05, "loss": 2.129, "step": 3650 }, { "epoch": 0.2, "learning_rate": 4.906415665909337e-05, "loss": 2.2007, "step": 3660 }, { "epoch": 0.2, "learning_rate": 4.906145190955318e-05, "loss": 2.2141, "step": 3670 }, { "epoch": 0.2, "learning_rate": 4.905874716001299e-05, "loss": 2.2148, "step": 3680 }, { "epoch": 0.2, "learning_rate": 4.905604241047279e-05, "loss": 2.196, "step": 3690 }, { "epoch": 0.2, "learning_rate": 4.90533376609326e-05, "loss": 2.1947, "step": 3700 }, { "epoch": 0.2, "learning_rate": 4.905063291139241e-05, "loss": 2.1857, "step": 3710 }, { "epoch": 0.2, "learning_rate": 4.904792816185221e-05, "loss": 2.1929, "step": 3720 }, { "epoch": 0.2, "learning_rate": 4.904522341231202e-05, "loss": 2.1385, "step": 3730 }, { "epoch": 0.2, "learning_rate": 4.904251866277183e-05, "loss": 2.1149, "step": 3740 }, { "epoch": 0.2, "learning_rate": 4.9039813913231635e-05, "loss": 2.1619, "step": 3750 }, { "epoch": 0.2, "learning_rate": 4.9037109163691444e-05, "loss": 2.1847, "step": 3760 }, { "epoch": 0.2, "learning_rate": 4.9034404414151254e-05, "loss": 2.1822, "step": 3770 }, { "epoch": 0.2, "learning_rate": 4.903169966461106e-05, "loss": 2.2085, "step": 3780 }, { "epoch": 0.2, "learning_rate": 4.902899491507087e-05, "loss": 2.1385, "step": 3790 }, { "epoch": 0.21, "learning_rate": 4.9026290165530676e-05, "loss": 2.0758, "step": 3800 }, { "epoch": 0.21, "learning_rate": 4.902358541599048e-05, "loss": 2.1987, "step": 3810 }, { "epoch": 0.21, "learning_rate": 4.902088066645029e-05, "loss": 2.128, "step": 3820 }, { "epoch": 0.21, "learning_rate": 4.90181759169101e-05, "loss": 2.1373, "step": 3830 }, { "epoch": 0.21, "learning_rate": 4.90154711673699e-05, "loss": 2.1146, "step": 3840 }, { "epoch": 0.21, "learning_rate": 4.901276641782971e-05, "loss": 2.1235, "step": 3850 }, { "epoch": 0.21, "learning_rate": 4.901006166828952e-05, "loss": 2.0737, "step": 3860 }, { "epoch": 0.21, "learning_rate": 4.9007356918749323e-05, "loss": 2.1604, "step": 3870 }, { "epoch": 0.21, "learning_rate": 4.900465216920913e-05, "loss": 2.1459, "step": 3880 }, { "epoch": 0.21, "learning_rate": 4.900194741966894e-05, "loss": 2.05, "step": 3890 }, { "epoch": 0.21, "learning_rate": 4.8999242670128746e-05, "loss": 2.2853, "step": 3900 }, { "epoch": 0.21, "learning_rate": 4.8996537920588555e-05, "loss": 2.1384, "step": 3910 }, { "epoch": 0.21, "learning_rate": 4.8993833171048365e-05, "loss": 2.1456, "step": 3920 }, { "epoch": 0.21, "learning_rate": 4.899112842150817e-05, "loss": 2.0723, "step": 3930 }, { "epoch": 0.21, "learning_rate": 4.898842367196798e-05, "loss": 2.2023, "step": 3940 }, { "epoch": 0.21, "learning_rate": 4.898571892242779e-05, "loss": 2.2038, "step": 3950 }, { "epoch": 0.21, "learning_rate": 4.898301417288759e-05, "loss": 2.0899, "step": 3960 }, { "epoch": 0.21, "learning_rate": 4.89803094233474e-05, "loss": 2.1205, "step": 3970 }, { "epoch": 0.22, "learning_rate": 4.897760467380721e-05, "loss": 2.1213, "step": 3980 }, { "epoch": 0.22, "learning_rate": 4.897489992426701e-05, "loss": 2.1685, "step": 3990 }, { "epoch": 0.22, "learning_rate": 4.897219517472682e-05, "loss": 2.2794, "step": 4000 }, { "epoch": 0.22, "learning_rate": 4.896949042518663e-05, "loss": 2.1567, "step": 4010 }, { "epoch": 0.22, "learning_rate": 4.8966785675646434e-05, "loss": 2.1086, "step": 4020 }, { "epoch": 0.22, "learning_rate": 4.8964080926106244e-05, "loss": 2.1684, "step": 4030 }, { "epoch": 0.22, "learning_rate": 4.8961376176566054e-05, "loss": 2.0768, "step": 4040 }, { "epoch": 0.22, "learning_rate": 4.8958671427025857e-05, "loss": 2.0707, "step": 4050 }, { "epoch": 0.22, "learning_rate": 4.8955966677485666e-05, "loss": 2.0717, "step": 4060 }, { "epoch": 0.22, "learning_rate": 4.8953261927945476e-05, "loss": 2.1114, "step": 4070 }, { "epoch": 0.22, "learning_rate": 4.895055717840528e-05, "loss": 2.2103, "step": 4080 }, { "epoch": 0.22, "learning_rate": 4.894785242886509e-05, "loss": 2.0926, "step": 4090 }, { "epoch": 0.22, "learning_rate": 4.89451476793249e-05, "loss": 2.0938, "step": 4100 }, { "epoch": 0.22, "learning_rate": 4.894244292978471e-05, "loss": 2.2068, "step": 4110 }, { "epoch": 0.22, "learning_rate": 4.893973818024451e-05, "loss": 2.1906, "step": 4120 }, { "epoch": 0.22, "learning_rate": 4.893703343070432e-05, "loss": 2.165, "step": 4130 }, { "epoch": 0.22, "learning_rate": 4.893432868116413e-05, "loss": 2.0793, "step": 4140 }, { "epoch": 0.22, "learning_rate": 4.893162393162393e-05, "loss": 2.0799, "step": 4150 }, { "epoch": 0.22, "learning_rate": 4.892891918208374e-05, "loss": 2.2392, "step": 4160 }, { "epoch": 0.23, "learning_rate": 4.892621443254355e-05, "loss": 2.1097, "step": 4170 }, { "epoch": 0.23, "learning_rate": 4.8923509683003355e-05, "loss": 2.1154, "step": 4180 }, { "epoch": 0.23, "learning_rate": 4.8920804933463164e-05, "loss": 2.0298, "step": 4190 }, { "epoch": 0.23, "learning_rate": 4.8918100183922974e-05, "loss": 2.0672, "step": 4200 }, { "epoch": 0.23, "learning_rate": 4.891539543438278e-05, "loss": 2.2863, "step": 4210 }, { "epoch": 0.23, "learning_rate": 4.891269068484259e-05, "loss": 2.2196, "step": 4220 }, { "epoch": 0.23, "learning_rate": 4.8909985935302396e-05, "loss": 2.1059, "step": 4230 }, { "epoch": 0.23, "learning_rate": 4.89072811857622e-05, "loss": 2.1494, "step": 4240 }, { "epoch": 0.23, "learning_rate": 4.890457643622201e-05, "loss": 2.0783, "step": 4250 }, { "epoch": 0.23, "learning_rate": 4.890187168668182e-05, "loss": 2.1305, "step": 4260 }, { "epoch": 0.23, "learning_rate": 4.889916693714162e-05, "loss": 2.1234, "step": 4270 }, { "epoch": 0.23, "learning_rate": 4.889646218760143e-05, "loss": 2.1472, "step": 4280 }, { "epoch": 0.23, "learning_rate": 4.889375743806124e-05, "loss": 2.1995, "step": 4290 }, { "epoch": 0.23, "learning_rate": 4.8891052688521044e-05, "loss": 2.193, "step": 4300 }, { "epoch": 0.23, "learning_rate": 4.888834793898085e-05, "loss": 2.152, "step": 4310 }, { "epoch": 0.23, "learning_rate": 4.888564318944066e-05, "loss": 2.1803, "step": 4320 }, { "epoch": 0.23, "learning_rate": 4.8882938439900466e-05, "loss": 2.128, "step": 4330 }, { "epoch": 0.23, "learning_rate": 4.8880233690360275e-05, "loss": 2.1841, "step": 4340 }, { "epoch": 0.24, "learning_rate": 4.8877528940820085e-05, "loss": 2.1476, "step": 4350 }, { "epoch": 0.24, "learning_rate": 4.887482419127989e-05, "loss": 2.1595, "step": 4360 }, { "epoch": 0.24, "learning_rate": 4.88721194417397e-05, "loss": 2.0705, "step": 4370 }, { "epoch": 0.24, "learning_rate": 4.886941469219951e-05, "loss": 2.1818, "step": 4380 }, { "epoch": 0.24, "learning_rate": 4.886670994265931e-05, "loss": 2.1307, "step": 4390 }, { "epoch": 0.24, "learning_rate": 4.886400519311912e-05, "loss": 2.1351, "step": 4400 }, { "epoch": 0.24, "learning_rate": 4.886130044357893e-05, "loss": 2.1485, "step": 4410 }, { "epoch": 0.24, "learning_rate": 4.885859569403873e-05, "loss": 2.1805, "step": 4420 }, { "epoch": 0.24, "learning_rate": 4.885589094449854e-05, "loss": 2.169, "step": 4430 }, { "epoch": 0.24, "learning_rate": 4.885318619495835e-05, "loss": 2.1482, "step": 4440 }, { "epoch": 0.24, "learning_rate": 4.8850481445418154e-05, "loss": 2.109, "step": 4450 }, { "epoch": 0.24, "learning_rate": 4.8847776695877964e-05, "loss": 2.1289, "step": 4460 }, { "epoch": 0.24, "learning_rate": 4.8845071946337774e-05, "loss": 2.1478, "step": 4470 }, { "epoch": 0.24, "learning_rate": 4.8842367196797577e-05, "loss": 2.1419, "step": 4480 }, { "epoch": 0.24, "learning_rate": 4.8839662447257386e-05, "loss": 2.1215, "step": 4490 }, { "epoch": 0.24, "learning_rate": 4.8836957697717196e-05, "loss": 2.0913, "step": 4500 }, { "epoch": 0.24, "learning_rate": 4.8834252948177e-05, "loss": 2.1563, "step": 4510 }, { "epoch": 0.24, "learning_rate": 4.883154819863681e-05, "loss": 2.1591, "step": 4520 }, { "epoch": 0.24, "learning_rate": 4.882884344909662e-05, "loss": 2.2, "step": 4530 }, { "epoch": 0.25, "learning_rate": 4.882613869955642e-05, "loss": 2.1788, "step": 4540 }, { "epoch": 0.25, "learning_rate": 4.882343395001623e-05, "loss": 2.1693, "step": 4550 }, { "epoch": 0.25, "learning_rate": 4.882072920047604e-05, "loss": 2.1675, "step": 4560 }, { "epoch": 0.25, "learning_rate": 4.881802445093584e-05, "loss": 2.0414, "step": 4570 }, { "epoch": 0.25, "learning_rate": 4.881531970139565e-05, "loss": 2.1366, "step": 4580 }, { "epoch": 0.25, "learning_rate": 4.881261495185546e-05, "loss": 2.1731, "step": 4590 }, { "epoch": 0.25, "learning_rate": 4.8809910202315265e-05, "loss": 2.199, "step": 4600 }, { "epoch": 0.25, "learning_rate": 4.8807205452775075e-05, "loss": 2.1227, "step": 4610 }, { "epoch": 0.25, "learning_rate": 4.8804500703234885e-05, "loss": 2.2154, "step": 4620 }, { "epoch": 0.25, "learning_rate": 4.880179595369469e-05, "loss": 2.1539, "step": 4630 }, { "epoch": 0.25, "learning_rate": 4.87990912041545e-05, "loss": 2.1191, "step": 4640 }, { "epoch": 0.25, "learning_rate": 4.879638645461431e-05, "loss": 2.142, "step": 4650 }, { "epoch": 0.25, "learning_rate": 4.879368170507411e-05, "loss": 2.2152, "step": 4660 }, { "epoch": 0.25, "learning_rate": 4.879097695553392e-05, "loss": 2.1402, "step": 4670 }, { "epoch": 0.25, "learning_rate": 4.878827220599373e-05, "loss": 2.2118, "step": 4680 }, { "epoch": 0.25, "learning_rate": 4.878556745645353e-05, "loss": 2.1556, "step": 4690 }, { "epoch": 0.25, "learning_rate": 4.878286270691334e-05, "loss": 2.087, "step": 4700 }, { "epoch": 0.25, "learning_rate": 4.878015795737315e-05, "loss": 2.1782, "step": 4710 }, { "epoch": 0.26, "learning_rate": 4.8777453207832954e-05, "loss": 2.2166, "step": 4720 }, { "epoch": 0.26, "learning_rate": 4.8774748458292764e-05, "loss": 2.1329, "step": 4730 }, { "epoch": 0.26, "learning_rate": 4.877204370875257e-05, "loss": 2.1564, "step": 4740 }, { "epoch": 0.26, "learning_rate": 4.876933895921238e-05, "loss": 2.1415, "step": 4750 }, { "epoch": 0.26, "learning_rate": 4.8766634209672186e-05, "loss": 2.2028, "step": 4760 }, { "epoch": 0.26, "learning_rate": 4.8763929460131995e-05, "loss": 2.1602, "step": 4770 }, { "epoch": 0.26, "learning_rate": 4.8761224710591805e-05, "loss": 2.0192, "step": 4780 }, { "epoch": 0.26, "learning_rate": 4.875851996105161e-05, "loss": 2.111, "step": 4790 }, { "epoch": 0.26, "learning_rate": 4.875581521151142e-05, "loss": 2.0504, "step": 4800 }, { "epoch": 0.26, "learning_rate": 4.875311046197123e-05, "loss": 2.1953, "step": 4810 }, { "epoch": 0.26, "learning_rate": 4.875040571243103e-05, "loss": 2.0584, "step": 4820 }, { "epoch": 0.26, "learning_rate": 4.874770096289084e-05, "loss": 2.1499, "step": 4830 }, { "epoch": 0.26, "learning_rate": 4.874499621335065e-05, "loss": 2.1209, "step": 4840 }, { "epoch": 0.26, "learning_rate": 4.874229146381045e-05, "loss": 2.0557, "step": 4850 }, { "epoch": 0.26, "learning_rate": 4.873958671427026e-05, "loss": 2.203, "step": 4860 }, { "epoch": 0.26, "learning_rate": 4.873688196473007e-05, "loss": 2.2086, "step": 4870 }, { "epoch": 0.26, "learning_rate": 4.8734177215189874e-05, "loss": 2.1598, "step": 4880 }, { "epoch": 0.26, "learning_rate": 4.8731472465649684e-05, "loss": 2.1035, "step": 4890 }, { "epoch": 0.26, "learning_rate": 4.8728767716109494e-05, "loss": 2.1245, "step": 4900 }, { "epoch": 0.27, "learning_rate": 4.8726062966569297e-05, "loss": 2.0773, "step": 4910 }, { "epoch": 0.27, "learning_rate": 4.8723358217029106e-05, "loss": 2.1491, "step": 4920 }, { "epoch": 0.27, "learning_rate": 4.8720653467488916e-05, "loss": 2.1825, "step": 4930 }, { "epoch": 0.27, "learning_rate": 4.871794871794872e-05, "loss": 2.2472, "step": 4940 }, { "epoch": 0.27, "learning_rate": 4.871524396840853e-05, "loss": 2.0781, "step": 4950 }, { "epoch": 0.27, "learning_rate": 4.871253921886834e-05, "loss": 2.1632, "step": 4960 }, { "epoch": 0.27, "learning_rate": 4.870983446932814e-05, "loss": 2.1474, "step": 4970 }, { "epoch": 0.27, "learning_rate": 4.870712971978795e-05, "loss": 2.1621, "step": 4980 }, { "epoch": 0.27, "learning_rate": 4.870442497024776e-05, "loss": 2.1332, "step": 4990 }, { "epoch": 0.27, "learning_rate": 4.870172022070756e-05, "loss": 2.1885, "step": 5000 }, { "epoch": 0.27, "learning_rate": 4.869901547116737e-05, "loss": 2.0893, "step": 5010 }, { "epoch": 0.27, "learning_rate": 4.869631072162718e-05, "loss": 2.1262, "step": 5020 }, { "epoch": 0.27, "learning_rate": 4.8693605972086985e-05, "loss": 2.0766, "step": 5030 }, { "epoch": 0.27, "learning_rate": 4.8690901222546795e-05, "loss": 2.158, "step": 5040 }, { "epoch": 0.27, "learning_rate": 4.8688196473006605e-05, "loss": 2.1305, "step": 5050 }, { "epoch": 0.27, "learning_rate": 4.868549172346641e-05, "loss": 2.1856, "step": 5060 }, { "epoch": 0.27, "learning_rate": 4.868278697392622e-05, "loss": 2.1239, "step": 5070 }, { "epoch": 0.27, "learning_rate": 4.868008222438603e-05, "loss": 2.1659, "step": 5080 }, { "epoch": 0.28, "learning_rate": 4.867737747484583e-05, "loss": 2.1423, "step": 5090 }, { "epoch": 0.28, "learning_rate": 4.867467272530564e-05, "loss": 2.0823, "step": 5100 }, { "epoch": 0.28, "learning_rate": 4.867196797576545e-05, "loss": 2.0916, "step": 5110 }, { "epoch": 0.28, "learning_rate": 4.866926322622525e-05, "loss": 2.1181, "step": 5120 }, { "epoch": 0.28, "learning_rate": 4.866655847668506e-05, "loss": 2.1198, "step": 5130 }, { "epoch": 0.28, "learning_rate": 4.866385372714487e-05, "loss": 2.0338, "step": 5140 }, { "epoch": 0.28, "learning_rate": 4.8661148977604674e-05, "loss": 2.0937, "step": 5150 }, { "epoch": 0.28, "learning_rate": 4.8658444228064484e-05, "loss": 2.0794, "step": 5160 }, { "epoch": 0.28, "learning_rate": 4.865573947852429e-05, "loss": 2.1517, "step": 5170 }, { "epoch": 0.28, "learning_rate": 4.8653034728984096e-05, "loss": 2.1549, "step": 5180 }, { "epoch": 0.28, "learning_rate": 4.8650329979443906e-05, "loss": 2.1174, "step": 5190 }, { "epoch": 0.28, "learning_rate": 4.8647625229903715e-05, "loss": 2.0992, "step": 5200 }, { "epoch": 0.28, "learning_rate": 4.864492048036352e-05, "loss": 2.1452, "step": 5210 }, { "epoch": 0.28, "learning_rate": 4.864221573082333e-05, "loss": 2.0618, "step": 5220 }, { "epoch": 0.28, "learning_rate": 4.863951098128314e-05, "loss": 2.0445, "step": 5230 }, { "epoch": 0.28, "learning_rate": 4.863680623174294e-05, "loss": 2.0831, "step": 5240 }, { "epoch": 0.28, "learning_rate": 4.863410148220275e-05, "loss": 2.1118, "step": 5250 }, { "epoch": 0.28, "learning_rate": 4.863139673266256e-05, "loss": 2.1932, "step": 5260 }, { "epoch": 0.28, "learning_rate": 4.862869198312236e-05, "loss": 2.0984, "step": 5270 }, { "epoch": 0.29, "learning_rate": 4.862598723358217e-05, "loss": 2.2012, "step": 5280 }, { "epoch": 0.29, "learning_rate": 4.862328248404198e-05, "loss": 2.1743, "step": 5290 }, { "epoch": 0.29, "learning_rate": 4.8620577734501785e-05, "loss": 2.0929, "step": 5300 }, { "epoch": 0.29, "learning_rate": 4.8617872984961594e-05, "loss": 2.1927, "step": 5310 }, { "epoch": 0.29, "learning_rate": 4.8615168235421404e-05, "loss": 2.0931, "step": 5320 }, { "epoch": 0.29, "learning_rate": 4.861246348588121e-05, "loss": 2.1771, "step": 5330 }, { "epoch": 0.29, "learning_rate": 4.8609758736341017e-05, "loss": 2.1156, "step": 5340 }, { "epoch": 0.29, "learning_rate": 4.8607053986800826e-05, "loss": 2.0911, "step": 5350 }, { "epoch": 0.29, "learning_rate": 4.860434923726063e-05, "loss": 2.1872, "step": 5360 }, { "epoch": 0.29, "learning_rate": 4.860164448772044e-05, "loss": 2.0721, "step": 5370 }, { "epoch": 0.29, "learning_rate": 4.859893973818025e-05, "loss": 1.9995, "step": 5380 }, { "epoch": 0.29, "learning_rate": 4.859623498864006e-05, "loss": 2.2122, "step": 5390 }, { "epoch": 0.29, "learning_rate": 4.859353023909986e-05, "loss": 2.1734, "step": 5400 }, { "epoch": 0.29, "learning_rate": 4.859082548955967e-05, "loss": 2.1543, "step": 5410 }, { "epoch": 0.29, "learning_rate": 4.858812074001948e-05, "loss": 2.1649, "step": 5420 }, { "epoch": 0.29, "learning_rate": 4.858541599047928e-05, "loss": 2.1118, "step": 5430 }, { "epoch": 0.29, "learning_rate": 4.858271124093909e-05, "loss": 1.9781, "step": 5440 }, { "epoch": 0.29, "learning_rate": 4.85800064913989e-05, "loss": 2.14, "step": 5450 }, { "epoch": 0.3, "learning_rate": 4.8577301741858705e-05, "loss": 2.0918, "step": 5460 }, { "epoch": 0.3, "learning_rate": 4.8574596992318515e-05, "loss": 2.014, "step": 5470 }, { "epoch": 0.3, "learning_rate": 4.8571892242778325e-05, "loss": 2.0508, "step": 5480 }, { "epoch": 0.3, "learning_rate": 4.856918749323813e-05, "loss": 2.1018, "step": 5490 }, { "epoch": 0.3, "learning_rate": 4.856648274369794e-05, "loss": 2.1366, "step": 5500 }, { "epoch": 0.3, "learning_rate": 4.856377799415775e-05, "loss": 2.0481, "step": 5510 }, { "epoch": 0.3, "learning_rate": 4.856107324461755e-05, "loss": 2.0503, "step": 5520 }, { "epoch": 0.3, "learning_rate": 4.855836849507735e-05, "loss": 2.1387, "step": 5530 }, { "epoch": 0.3, "learning_rate": 4.855566374553717e-05, "loss": 2.0415, "step": 5540 }, { "epoch": 0.3, "learning_rate": 4.855295899599697e-05, "loss": 2.0969, "step": 5550 }, { "epoch": 0.3, "learning_rate": 4.8550254246456775e-05, "loss": 2.1755, "step": 5560 }, { "epoch": 0.3, "learning_rate": 4.854754949691659e-05, "loss": 2.0534, "step": 5570 }, { "epoch": 0.3, "learning_rate": 4.8544844747376394e-05, "loss": 2.1393, "step": 5580 }, { "epoch": 0.3, "learning_rate": 4.85421399978362e-05, "loss": 2.0886, "step": 5590 }, { "epoch": 0.3, "learning_rate": 4.853943524829601e-05, "loss": 2.0897, "step": 5600 }, { "epoch": 0.3, "learning_rate": 4.8536730498755816e-05, "loss": 2.1263, "step": 5610 }, { "epoch": 0.3, "learning_rate": 4.8534025749215626e-05, "loss": 2.0236, "step": 5620 }, { "epoch": 0.3, "learning_rate": 4.8531320999675435e-05, "loss": 2.0302, "step": 5630 }, { "epoch": 0.3, "learning_rate": 4.852861625013524e-05, "loss": 2.1676, "step": 5640 }, { "epoch": 0.31, "learning_rate": 4.852591150059505e-05, "loss": 2.0892, "step": 5650 }, { "epoch": 0.31, "learning_rate": 4.852320675105486e-05, "loss": 2.1052, "step": 5660 }, { "epoch": 0.31, "learning_rate": 4.852050200151466e-05, "loss": 2.1084, "step": 5670 }, { "epoch": 0.31, "learning_rate": 4.851779725197447e-05, "loss": 2.1005, "step": 5680 }, { "epoch": 0.31, "learning_rate": 4.851509250243428e-05, "loss": 2.1331, "step": 5690 }, { "epoch": 0.31, "learning_rate": 4.851238775289408e-05, "loss": 2.1225, "step": 5700 }, { "epoch": 0.31, "learning_rate": 4.850968300335389e-05, "loss": 2.1193, "step": 5710 }, { "epoch": 0.31, "learning_rate": 4.85069782538137e-05, "loss": 2.1501, "step": 5720 }, { "epoch": 0.31, "learning_rate": 4.8504273504273505e-05, "loss": 2.0745, "step": 5730 }, { "epoch": 0.31, "learning_rate": 4.8501568754733314e-05, "loss": 2.304, "step": 5740 }, { "epoch": 0.31, "learning_rate": 4.8498864005193124e-05, "loss": 2.1034, "step": 5750 }, { "epoch": 0.31, "learning_rate": 4.849615925565293e-05, "loss": 2.0864, "step": 5760 }, { "epoch": 0.31, "learning_rate": 4.8493454506112737e-05, "loss": 2.0248, "step": 5770 }, { "epoch": 0.31, "learning_rate": 4.8490749756572546e-05, "loss": 2.0726, "step": 5780 }, { "epoch": 0.31, "learning_rate": 4.848804500703235e-05, "loss": 2.0804, "step": 5790 }, { "epoch": 0.31, "learning_rate": 4.848534025749216e-05, "loss": 2.095, "step": 5800 }, { "epoch": 0.31, "learning_rate": 4.848263550795197e-05, "loss": 2.0683, "step": 5810 }, { "epoch": 0.31, "learning_rate": 4.847993075841177e-05, "loss": 2.1429, "step": 5820 }, { "epoch": 0.32, "learning_rate": 4.847722600887158e-05, "loss": 2.0943, "step": 5830 }, { "epoch": 0.32, "learning_rate": 4.847452125933139e-05, "loss": 2.2089, "step": 5840 }, { "epoch": 0.32, "learning_rate": 4.8471816509791193e-05, "loss": 2.1073, "step": 5850 }, { "epoch": 0.32, "learning_rate": 4.8469111760251e-05, "loss": 2.0883, "step": 5860 }, { "epoch": 0.32, "learning_rate": 4.846640701071081e-05, "loss": 2.1735, "step": 5870 }, { "epoch": 0.32, "learning_rate": 4.8463702261170616e-05, "loss": 2.1653, "step": 5880 }, { "epoch": 0.32, "learning_rate": 4.8460997511630425e-05, "loss": 2.1176, "step": 5890 }, { "epoch": 0.32, "learning_rate": 4.8458292762090235e-05, "loss": 2.0641, "step": 5900 }, { "epoch": 0.32, "learning_rate": 4.845558801255004e-05, "loss": 2.1243, "step": 5910 }, { "epoch": 0.32, "learning_rate": 4.845288326300985e-05, "loss": 2.153, "step": 5920 }, { "epoch": 0.32, "learning_rate": 4.845017851346966e-05, "loss": 1.9643, "step": 5930 }, { "epoch": 0.32, "learning_rate": 4.844747376392946e-05, "loss": 2.1161, "step": 5940 }, { "epoch": 0.32, "learning_rate": 4.844476901438927e-05, "loss": 2.0689, "step": 5950 }, { "epoch": 0.32, "learning_rate": 4.844206426484908e-05, "loss": 2.1078, "step": 5960 }, { "epoch": 0.32, "learning_rate": 4.843935951530888e-05, "loss": 2.1318, "step": 5970 }, { "epoch": 0.32, "learning_rate": 4.843665476576869e-05, "loss": 2.1226, "step": 5980 }, { "epoch": 0.32, "learning_rate": 4.84339500162285e-05, "loss": 2.102, "step": 5990 }, { "epoch": 0.32, "learning_rate": 4.8431245266688304e-05, "loss": 2.0168, "step": 6000 }, { "epoch": 0.32, "learning_rate": 4.8428540517148114e-05, "loss": 2.1381, "step": 6010 }, { "epoch": 0.33, "learning_rate": 4.8425835767607924e-05, "loss": 2.1414, "step": 6020 }, { "epoch": 0.33, "learning_rate": 4.842313101806773e-05, "loss": 2.0751, "step": 6030 }, { "epoch": 0.33, "learning_rate": 4.8420426268527536e-05, "loss": 2.1685, "step": 6040 }, { "epoch": 0.33, "learning_rate": 4.8417721518987346e-05, "loss": 2.1224, "step": 6050 }, { "epoch": 0.33, "learning_rate": 4.8415016769447155e-05, "loss": 2.12, "step": 6060 }, { "epoch": 0.33, "learning_rate": 4.841231201990696e-05, "loss": 2.1069, "step": 6070 }, { "epoch": 0.33, "learning_rate": 4.840960727036677e-05, "loss": 1.9921, "step": 6080 }, { "epoch": 0.33, "learning_rate": 4.840690252082658e-05, "loss": 2.1098, "step": 6090 }, { "epoch": 0.33, "learning_rate": 4.840419777128638e-05, "loss": 2.0476, "step": 6100 }, { "epoch": 0.33, "learning_rate": 4.840149302174618e-05, "loss": 2.1331, "step": 6110 }, { "epoch": 0.33, "learning_rate": 4.8398788272206e-05, "loss": 2.0428, "step": 6120 }, { "epoch": 0.33, "learning_rate": 4.83960835226658e-05, "loss": 2.1373, "step": 6130 }, { "epoch": 0.33, "learning_rate": 4.8393378773125605e-05, "loss": 2.0874, "step": 6140 }, { "epoch": 0.33, "learning_rate": 4.839067402358542e-05, "loss": 2.1961, "step": 6150 }, { "epoch": 0.33, "learning_rate": 4.8387969274045225e-05, "loss": 2.114, "step": 6160 }, { "epoch": 0.33, "learning_rate": 4.838526452450503e-05, "loss": 2.1379, "step": 6170 }, { "epoch": 0.33, "learning_rate": 4.8382559774964844e-05, "loss": 2.0529, "step": 6180 }, { "epoch": 0.33, "learning_rate": 4.837985502542465e-05, "loss": 2.1002, "step": 6190 }, { "epoch": 0.34, "learning_rate": 4.837715027588445e-05, "loss": 2.0881, "step": 6200 }, { "epoch": 0.34, "learning_rate": 4.8374445526344266e-05, "loss": 2.1108, "step": 6210 }, { "epoch": 0.34, "learning_rate": 4.837174077680407e-05, "loss": 2.0584, "step": 6220 }, { "epoch": 0.34, "learning_rate": 4.836903602726388e-05, "loss": 2.1558, "step": 6230 }, { "epoch": 0.34, "learning_rate": 4.836633127772369e-05, "loss": 2.1448, "step": 6240 }, { "epoch": 0.34, "learning_rate": 4.836362652818349e-05, "loss": 2.1046, "step": 6250 }, { "epoch": 0.34, "learning_rate": 4.83609217786433e-05, "loss": 2.0926, "step": 6260 }, { "epoch": 0.34, "learning_rate": 4.835821702910311e-05, "loss": 2.0141, "step": 6270 }, { "epoch": 0.34, "learning_rate": 4.8355512279562913e-05, "loss": 2.0952, "step": 6280 }, { "epoch": 0.34, "learning_rate": 4.835280753002272e-05, "loss": 2.1004, "step": 6290 }, { "epoch": 0.34, "learning_rate": 4.835010278048253e-05, "loss": 2.0649, "step": 6300 }, { "epoch": 0.34, "learning_rate": 4.8347398030942336e-05, "loss": 2.0904, "step": 6310 }, { "epoch": 0.34, "learning_rate": 4.8344693281402145e-05, "loss": 2.0793, "step": 6320 }, { "epoch": 0.34, "learning_rate": 4.8341988531861955e-05, "loss": 2.0586, "step": 6330 }, { "epoch": 0.34, "learning_rate": 4.833928378232176e-05, "loss": 2.1601, "step": 6340 }, { "epoch": 0.34, "learning_rate": 4.833657903278157e-05, "loss": 2.1123, "step": 6350 }, { "epoch": 0.34, "learning_rate": 4.833387428324138e-05, "loss": 2.1017, "step": 6360 }, { "epoch": 0.34, "learning_rate": 4.833116953370118e-05, "loss": 2.1596, "step": 6370 }, { "epoch": 0.34, "learning_rate": 4.832846478416099e-05, "loss": 2.1159, "step": 6380 }, { "epoch": 0.35, "learning_rate": 4.83257600346208e-05, "loss": 2.171, "step": 6390 }, { "epoch": 0.35, "learning_rate": 4.83230552850806e-05, "loss": 2.1348, "step": 6400 }, { "epoch": 0.35, "learning_rate": 4.832035053554041e-05, "loss": 2.1296, "step": 6410 }, { "epoch": 0.35, "learning_rate": 4.831764578600022e-05, "loss": 2.1226, "step": 6420 }, { "epoch": 0.35, "learning_rate": 4.8314941036460024e-05, "loss": 2.1222, "step": 6430 }, { "epoch": 0.35, "learning_rate": 4.8312236286919834e-05, "loss": 2.065, "step": 6440 }, { "epoch": 0.35, "learning_rate": 4.8309531537379644e-05, "loss": 2.0582, "step": 6450 }, { "epoch": 0.35, "learning_rate": 4.8306826787839446e-05, "loss": 2.0788, "step": 6460 }, { "epoch": 0.35, "learning_rate": 4.8304122038299256e-05, "loss": 2.1553, "step": 6470 }, { "epoch": 0.35, "learning_rate": 4.8301417288759066e-05, "loss": 2.0549, "step": 6480 }, { "epoch": 0.35, "learning_rate": 4.829871253921887e-05, "loss": 2.1092, "step": 6490 }, { "epoch": 0.35, "learning_rate": 4.829600778967868e-05, "loss": 2.0254, "step": 6500 }, { "epoch": 0.35, "learning_rate": 4.829330304013849e-05, "loss": 2.1021, "step": 6510 }, { "epoch": 0.35, "learning_rate": 4.829059829059829e-05, "loss": 2.1233, "step": 6520 }, { "epoch": 0.35, "learning_rate": 4.82878935410581e-05, "loss": 2.1018, "step": 6530 }, { "epoch": 0.35, "learning_rate": 4.828518879151791e-05, "loss": 2.1134, "step": 6540 }, { "epoch": 0.35, "learning_rate": 4.828248404197771e-05, "loss": 2.1016, "step": 6550 }, { "epoch": 0.35, "learning_rate": 4.827977929243752e-05, "loss": 2.1259, "step": 6560 }, { "epoch": 0.36, "learning_rate": 4.827707454289733e-05, "loss": 2.0486, "step": 6570 }, { "epoch": 0.36, "learning_rate": 4.8274369793357135e-05, "loss": 2.0023, "step": 6580 }, { "epoch": 0.36, "learning_rate": 4.8271665043816945e-05, "loss": 2.0508, "step": 6590 }, { "epoch": 0.36, "learning_rate": 4.8268960294276754e-05, "loss": 2.027, "step": 6600 }, { "epoch": 0.36, "learning_rate": 4.826625554473656e-05, "loss": 2.1374, "step": 6610 }, { "epoch": 0.36, "learning_rate": 4.826355079519637e-05, "loss": 2.1038, "step": 6620 }, { "epoch": 0.36, "learning_rate": 4.826084604565618e-05, "loss": 2.0817, "step": 6630 }, { "epoch": 0.36, "learning_rate": 4.825814129611598e-05, "loss": 2.2052, "step": 6640 }, { "epoch": 0.36, "learning_rate": 4.825543654657579e-05, "loss": 2.099, "step": 6650 }, { "epoch": 0.36, "learning_rate": 4.825273179703559e-05, "loss": 2.1241, "step": 6660 }, { "epoch": 0.36, "learning_rate": 4.825002704749541e-05, "loss": 2.1416, "step": 6670 }, { "epoch": 0.36, "learning_rate": 4.824732229795521e-05, "loss": 2.0978, "step": 6680 }, { "epoch": 0.36, "learning_rate": 4.8244617548415014e-05, "loss": 2.0922, "step": 6690 }, { "epoch": 0.36, "learning_rate": 4.824191279887483e-05, "loss": 2.0611, "step": 6700 }, { "epoch": 0.36, "learning_rate": 4.8239208049334633e-05, "loss": 2.0971, "step": 6710 }, { "epoch": 0.36, "learning_rate": 4.8236503299794436e-05, "loss": 2.0229, "step": 6720 }, { "epoch": 0.36, "learning_rate": 4.823379855025425e-05, "loss": 2.0801, "step": 6730 }, { "epoch": 0.36, "learning_rate": 4.8231093800714056e-05, "loss": 2.1185, "step": 6740 }, { "epoch": 0.36, "learning_rate": 4.822838905117386e-05, "loss": 2.0782, "step": 6750 }, { "epoch": 0.37, "learning_rate": 4.8225684301633675e-05, "loss": 2.1115, "step": 6760 }, { "epoch": 0.37, "learning_rate": 4.822297955209348e-05, "loss": 2.0077, "step": 6770 }, { "epoch": 0.37, "learning_rate": 4.822027480255328e-05, "loss": 2.1124, "step": 6780 }, { "epoch": 0.37, "learning_rate": 4.82175700530131e-05, "loss": 2.0364, "step": 6790 }, { "epoch": 0.37, "learning_rate": 4.82148653034729e-05, "loss": 2.0268, "step": 6800 }, { "epoch": 0.37, "learning_rate": 4.82121605539327e-05, "loss": 2.1407, "step": 6810 }, { "epoch": 0.37, "learning_rate": 4.820945580439252e-05, "loss": 2.1124, "step": 6820 }, { "epoch": 0.37, "learning_rate": 4.820675105485232e-05, "loss": 2.0911, "step": 6830 }, { "epoch": 0.37, "learning_rate": 4.8204046305312125e-05, "loss": 2.126, "step": 6840 }, { "epoch": 0.37, "learning_rate": 4.820134155577194e-05, "loss": 2.0309, "step": 6850 }, { "epoch": 0.37, "learning_rate": 4.8198636806231744e-05, "loss": 2.1095, "step": 6860 }, { "epoch": 0.37, "learning_rate": 4.8195932056691554e-05, "loss": 2.103, "step": 6870 }, { "epoch": 0.37, "learning_rate": 4.8193227307151364e-05, "loss": 2.1151, "step": 6880 }, { "epoch": 0.37, "learning_rate": 4.8190522557611167e-05, "loss": 2.099, "step": 6890 }, { "epoch": 0.37, "learning_rate": 4.8187817808070976e-05, "loss": 2.0917, "step": 6900 }, { "epoch": 0.37, "learning_rate": 4.8185113058530786e-05, "loss": 2.0778, "step": 6910 }, { "epoch": 0.37, "learning_rate": 4.818240830899059e-05, "loss": 2.1163, "step": 6920 }, { "epoch": 0.37, "learning_rate": 4.81797035594504e-05, "loss": 2.1312, "step": 6930 }, { "epoch": 0.37, "learning_rate": 4.817699880991021e-05, "loss": 2.0743, "step": 6940 }, { "epoch": 0.38, "learning_rate": 4.817429406037001e-05, "loss": 2.0442, "step": 6950 }, { "epoch": 0.38, "learning_rate": 4.817158931082982e-05, "loss": 2.2059, "step": 6960 }, { "epoch": 0.38, "learning_rate": 4.816888456128963e-05, "loss": 2.1682, "step": 6970 }, { "epoch": 0.38, "learning_rate": 4.816617981174943e-05, "loss": 2.1025, "step": 6980 }, { "epoch": 0.38, "learning_rate": 4.816347506220924e-05, "loss": 2.0286, "step": 6990 }, { "epoch": 0.38, "learning_rate": 4.816077031266905e-05, "loss": 2.0424, "step": 7000 }, { "epoch": 0.38, "learning_rate": 4.8158065563128855e-05, "loss": 2.122, "step": 7010 }, { "epoch": 0.38, "learning_rate": 4.8155360813588665e-05, "loss": 2.0345, "step": 7020 }, { "epoch": 0.38, "learning_rate": 4.8152656064048474e-05, "loss": 2.0534, "step": 7030 }, { "epoch": 0.38, "learning_rate": 4.814995131450828e-05, "loss": 2.0746, "step": 7040 }, { "epoch": 0.38, "learning_rate": 4.814724656496809e-05, "loss": 2.1172, "step": 7050 }, { "epoch": 0.38, "learning_rate": 4.81445418154279e-05, "loss": 2.1125, "step": 7060 }, { "epoch": 0.38, "learning_rate": 4.81418370658877e-05, "loss": 2.0522, "step": 7070 }, { "epoch": 0.38, "learning_rate": 4.813913231634751e-05, "loss": 2.0783, "step": 7080 }, { "epoch": 0.38, "learning_rate": 4.813642756680732e-05, "loss": 2.0832, "step": 7090 }, { "epoch": 0.38, "learning_rate": 4.813372281726712e-05, "loss": 2.0372, "step": 7100 }, { "epoch": 0.38, "learning_rate": 4.813101806772693e-05, "loss": 2.1366, "step": 7110 }, { "epoch": 0.38, "learning_rate": 4.812831331818674e-05, "loss": 2.0432, "step": 7120 }, { "epoch": 0.39, "learning_rate": 4.8125608568646544e-05, "loss": 2.1027, "step": 7130 }, { "epoch": 0.39, "learning_rate": 4.8122903819106354e-05, "loss": 2.1157, "step": 7140 }, { "epoch": 0.39, "learning_rate": 4.812019906956616e-05, "loss": 2.0857, "step": 7150 }, { "epoch": 0.39, "learning_rate": 4.8117494320025966e-05, "loss": 2.014, "step": 7160 }, { "epoch": 0.39, "learning_rate": 4.8114789570485776e-05, "loss": 2.1186, "step": 7170 }, { "epoch": 0.39, "learning_rate": 4.8112084820945585e-05, "loss": 2.0234, "step": 7180 }, { "epoch": 0.39, "learning_rate": 4.810938007140539e-05, "loss": 2.1308, "step": 7190 }, { "epoch": 0.39, "learning_rate": 4.81066753218652e-05, "loss": 2.1128, "step": 7200 }, { "epoch": 0.39, "learning_rate": 4.8103970572325e-05, "loss": 2.0832, "step": 7210 }, { "epoch": 0.39, "learning_rate": 4.810126582278481e-05, "loss": 2.1033, "step": 7220 }, { "epoch": 0.39, "learning_rate": 4.809856107324462e-05, "loss": 2.0836, "step": 7230 }, { "epoch": 0.39, "learning_rate": 4.809585632370442e-05, "loss": 2.0607, "step": 7240 }, { "epoch": 0.39, "learning_rate": 4.809315157416423e-05, "loss": 2.1112, "step": 7250 }, { "epoch": 0.39, "learning_rate": 4.809044682462404e-05, "loss": 1.9418, "step": 7260 }, { "epoch": 0.39, "learning_rate": 4.8087742075083845e-05, "loss": 2.1436, "step": 7270 }, { "epoch": 0.39, "learning_rate": 4.808503732554366e-05, "loss": 2.1087, "step": 7280 }, { "epoch": 0.39, "learning_rate": 4.8082332576003464e-05, "loss": 2.0366, "step": 7290 }, { "epoch": 0.39, "learning_rate": 4.807962782646327e-05, "loss": 2.0307, "step": 7300 }, { "epoch": 0.39, "learning_rate": 4.8076923076923084e-05, "loss": 1.9857, "step": 7310 }, { "epoch": 0.4, "learning_rate": 4.8074218327382887e-05, "loss": 2.0303, "step": 7320 }, { "epoch": 0.4, "learning_rate": 4.807151357784269e-05, "loss": 1.9889, "step": 7330 }, { "epoch": 0.4, "learning_rate": 4.8068808828302506e-05, "loss": 2.0596, "step": 7340 }, { "epoch": 0.4, "learning_rate": 4.806610407876231e-05, "loss": 2.0327, "step": 7350 }, { "epoch": 0.4, "learning_rate": 4.806339932922211e-05, "loss": 2.1242, "step": 7360 }, { "epoch": 0.4, "learning_rate": 4.806069457968193e-05, "loss": 2.1426, "step": 7370 }, { "epoch": 0.4, "learning_rate": 4.805798983014173e-05, "loss": 2.0535, "step": 7380 }, { "epoch": 0.4, "learning_rate": 4.8055285080601534e-05, "loss": 2.1334, "step": 7390 }, { "epoch": 0.4, "learning_rate": 4.805258033106135e-05, "loss": 2.0761, "step": 7400 }, { "epoch": 0.4, "learning_rate": 4.804987558152115e-05, "loss": 2.0537, "step": 7410 }, { "epoch": 0.4, "learning_rate": 4.8047170831980956e-05, "loss": 2.0145, "step": 7420 }, { "epoch": 0.4, "learning_rate": 4.804446608244077e-05, "loss": 2.0372, "step": 7430 }, { "epoch": 0.4, "learning_rate": 4.8041761332900575e-05, "loss": 2.0378, "step": 7440 }, { "epoch": 0.4, "learning_rate": 4.803905658336038e-05, "loss": 2.0431, "step": 7450 }, { "epoch": 0.4, "learning_rate": 4.8036351833820195e-05, "loss": 2.0751, "step": 7460 }, { "epoch": 0.4, "learning_rate": 4.803364708428e-05, "loss": 2.0546, "step": 7470 }, { "epoch": 0.4, "learning_rate": 4.80309423347398e-05, "loss": 2.0669, "step": 7480 }, { "epoch": 0.4, "learning_rate": 4.802823758519962e-05, "loss": 2.0371, "step": 7490 }, { "epoch": 0.41, "learning_rate": 4.802553283565942e-05, "loss": 2.0085, "step": 7500 }, { "epoch": 0.41, "learning_rate": 4.802282808611923e-05, "loss": 2.0412, "step": 7510 }, { "epoch": 0.41, "learning_rate": 4.802012333657904e-05, "loss": 2.1066, "step": 7520 }, { "epoch": 0.41, "learning_rate": 4.801741858703884e-05, "loss": 2.0666, "step": 7530 }, { "epoch": 0.41, "learning_rate": 4.801471383749865e-05, "loss": 1.9891, "step": 7540 }, { "epoch": 0.41, "learning_rate": 4.801200908795846e-05, "loss": 2.0983, "step": 7550 }, { "epoch": 0.41, "learning_rate": 4.8009304338418264e-05, "loss": 2.1094, "step": 7560 }, { "epoch": 0.41, "learning_rate": 4.8006599588878074e-05, "loss": 2.0901, "step": 7570 }, { "epoch": 0.41, "learning_rate": 4.800389483933788e-05, "loss": 2.061, "step": 7580 }, { "epoch": 0.41, "learning_rate": 4.8001190089797686e-05, "loss": 2.0782, "step": 7590 }, { "epoch": 0.41, "learning_rate": 4.7998485340257496e-05, "loss": 2.1625, "step": 7600 }, { "epoch": 0.41, "learning_rate": 4.7995780590717305e-05, "loss": 2.0023, "step": 7610 }, { "epoch": 0.41, "learning_rate": 4.799307584117711e-05, "loss": 2.0693, "step": 7620 }, { "epoch": 0.41, "learning_rate": 4.799037109163692e-05, "loss": 2.0473, "step": 7630 }, { "epoch": 0.41, "learning_rate": 4.798766634209673e-05, "loss": 2.1, "step": 7640 }, { "epoch": 0.41, "learning_rate": 4.798496159255653e-05, "loss": 2.1694, "step": 7650 }, { "epoch": 0.41, "learning_rate": 4.798225684301634e-05, "loss": 2.0591, "step": 7660 }, { "epoch": 0.41, "learning_rate": 4.797955209347615e-05, "loss": 2.0808, "step": 7670 }, { "epoch": 0.41, "learning_rate": 4.797684734393595e-05, "loss": 2.1263, "step": 7680 }, { "epoch": 0.42, "learning_rate": 4.797414259439576e-05, "loss": 2.0523, "step": 7690 }, { "epoch": 0.42, "learning_rate": 4.797143784485557e-05, "loss": 1.9046, "step": 7700 }, { "epoch": 0.42, "learning_rate": 4.7968733095315375e-05, "loss": 2.0654, "step": 7710 }, { "epoch": 0.42, "learning_rate": 4.7966028345775184e-05, "loss": 2.0108, "step": 7720 }, { "epoch": 0.42, "learning_rate": 4.7963323596234994e-05, "loss": 2.1146, "step": 7730 }, { "epoch": 0.42, "learning_rate": 4.79606188466948e-05, "loss": 2.0678, "step": 7740 }, { "epoch": 0.42, "learning_rate": 4.7957914097154607e-05, "loss": 2.0849, "step": 7750 }, { "epoch": 0.42, "learning_rate": 4.7955209347614416e-05, "loss": 2.0737, "step": 7760 }, { "epoch": 0.42, "learning_rate": 4.795250459807422e-05, "loss": 2.0362, "step": 7770 }, { "epoch": 0.42, "learning_rate": 4.794979984853403e-05, "loss": 2.0527, "step": 7780 }, { "epoch": 0.42, "learning_rate": 4.794709509899383e-05, "loss": 2.1155, "step": 7790 }, { "epoch": 0.42, "learning_rate": 4.794439034945364e-05, "loss": 2.1273, "step": 7800 }, { "epoch": 0.42, "learning_rate": 4.794168559991345e-05, "loss": 1.9762, "step": 7810 }, { "epoch": 0.42, "learning_rate": 4.7938980850373254e-05, "loss": 2.0193, "step": 7820 }, { "epoch": 0.42, "learning_rate": 4.7936276100833063e-05, "loss": 1.9861, "step": 7830 }, { "epoch": 0.42, "learning_rate": 4.793357135129287e-05, "loss": 2.0374, "step": 7840 }, { "epoch": 0.42, "learning_rate": 4.7930866601752676e-05, "loss": 2.0228, "step": 7850 }, { "epoch": 0.42, "learning_rate": 4.7928161852212486e-05, "loss": 2.0516, "step": 7860 }, { "epoch": 0.43, "learning_rate": 4.7925457102672295e-05, "loss": 2.0542, "step": 7870 }, { "epoch": 0.43, "learning_rate": 4.79227523531321e-05, "loss": 2.1633, "step": 7880 }, { "epoch": 0.43, "learning_rate": 4.792004760359191e-05, "loss": 2.0536, "step": 7890 }, { "epoch": 0.43, "learning_rate": 4.791734285405172e-05, "loss": 2.093, "step": 7900 }, { "epoch": 0.43, "learning_rate": 4.791463810451152e-05, "loss": 2.0988, "step": 7910 }, { "epoch": 0.43, "learning_rate": 4.791193335497134e-05, "loss": 2.0529, "step": 7920 }, { "epoch": 0.43, "learning_rate": 4.790922860543114e-05, "loss": 2.0661, "step": 7930 }, { "epoch": 0.43, "learning_rate": 4.790652385589094e-05, "loss": 2.0555, "step": 7940 }, { "epoch": 0.43, "learning_rate": 4.790381910635076e-05, "loss": 2.1093, "step": 7950 }, { "epoch": 0.43, "learning_rate": 4.790111435681056e-05, "loss": 2.0502, "step": 7960 }, { "epoch": 0.43, "learning_rate": 4.7898409607270365e-05, "loss": 2.0479, "step": 7970 }, { "epoch": 0.43, "learning_rate": 4.789570485773018e-05, "loss": 2.0475, "step": 7980 }, { "epoch": 0.43, "learning_rate": 4.7893000108189984e-05, "loss": 2.0214, "step": 7990 }, { "epoch": 0.43, "learning_rate": 4.789029535864979e-05, "loss": 2.1644, "step": 8000 }, { "epoch": 0.43, "learning_rate": 4.78875906091096e-05, "loss": 2.0794, "step": 8010 }, { "epoch": 0.43, "learning_rate": 4.7884885859569406e-05, "loss": 2.1384, "step": 8020 }, { "epoch": 0.43, "learning_rate": 4.788218111002921e-05, "loss": 2.1498, "step": 8030 }, { "epoch": 0.43, "learning_rate": 4.7879476360489025e-05, "loss": 2.1235, "step": 8040 }, { "epoch": 0.43, "learning_rate": 4.787677161094883e-05, "loss": 2.1588, "step": 8050 }, { "epoch": 0.44, "learning_rate": 4.787406686140863e-05, "loss": 1.9969, "step": 8060 }, { "epoch": 0.44, "learning_rate": 4.787136211186845e-05, "loss": 2.0681, "step": 8070 }, { "epoch": 0.44, "learning_rate": 4.786865736232825e-05, "loss": 2.0892, "step": 8080 }, { "epoch": 0.44, "learning_rate": 4.786595261278805e-05, "loss": 2.0454, "step": 8090 }, { "epoch": 0.44, "learning_rate": 4.786324786324787e-05, "loss": 2.0739, "step": 8100 }, { "epoch": 0.44, "learning_rate": 4.786054311370767e-05, "loss": 2.0319, "step": 8110 }, { "epoch": 0.44, "learning_rate": 4.7857838364167475e-05, "loss": 2.0293, "step": 8120 }, { "epoch": 0.44, "learning_rate": 4.785513361462729e-05, "loss": 2.0864, "step": 8130 }, { "epoch": 0.44, "learning_rate": 4.7852428865087095e-05, "loss": 2.1603, "step": 8140 }, { "epoch": 0.44, "learning_rate": 4.7849724115546904e-05, "loss": 2.037, "step": 8150 }, { "epoch": 0.44, "learning_rate": 4.7847019366006714e-05, "loss": 2.0788, "step": 8160 }, { "epoch": 0.44, "learning_rate": 4.784431461646652e-05, "loss": 2.0082, "step": 8170 }, { "epoch": 0.44, "learning_rate": 4.7841609866926327e-05, "loss": 2.0523, "step": 8180 }, { "epoch": 0.44, "learning_rate": 4.7838905117386136e-05, "loss": 2.1493, "step": 8190 }, { "epoch": 0.44, "learning_rate": 4.783620036784594e-05, "loss": 2.0041, "step": 8200 }, { "epoch": 0.44, "learning_rate": 4.783349561830575e-05, "loss": 2.1332, "step": 8210 }, { "epoch": 0.44, "learning_rate": 4.783079086876556e-05, "loss": 2.0501, "step": 8220 }, { "epoch": 0.44, "learning_rate": 4.782808611922536e-05, "loss": 2.0767, "step": 8230 }, { "epoch": 0.45, "learning_rate": 4.782538136968517e-05, "loss": 2.107, "step": 8240 }, { "epoch": 0.45, "learning_rate": 4.782267662014498e-05, "loss": 2.054, "step": 8250 }, { "epoch": 0.45, "learning_rate": 4.7819971870604783e-05, "loss": 2.1155, "step": 8260 }, { "epoch": 0.45, "learning_rate": 4.781726712106459e-05, "loss": 2.1182, "step": 8270 }, { "epoch": 0.45, "learning_rate": 4.78145623715244e-05, "loss": 1.9992, "step": 8280 }, { "epoch": 0.45, "learning_rate": 4.7811857621984206e-05, "loss": 2.1491, "step": 8290 }, { "epoch": 0.45, "learning_rate": 4.7809152872444015e-05, "loss": 2.139, "step": 8300 }, { "epoch": 0.45, "learning_rate": 4.7806448122903825e-05, "loss": 2.0448, "step": 8310 }, { "epoch": 0.45, "learning_rate": 4.780374337336363e-05, "loss": 2.1137, "step": 8320 }, { "epoch": 0.45, "learning_rate": 4.780103862382344e-05, "loss": 2.0272, "step": 8330 }, { "epoch": 0.45, "learning_rate": 4.779833387428324e-05, "loss": 2.0726, "step": 8340 }, { "epoch": 0.45, "learning_rate": 4.779562912474305e-05, "loss": 2.0395, "step": 8350 }, { "epoch": 0.45, "learning_rate": 4.779292437520286e-05, "loss": 2.0345, "step": 8360 }, { "epoch": 0.45, "learning_rate": 4.779021962566266e-05, "loss": 2.0747, "step": 8370 }, { "epoch": 0.45, "learning_rate": 4.778751487612247e-05, "loss": 2.013, "step": 8380 }, { "epoch": 0.45, "learning_rate": 4.778481012658228e-05, "loss": 2.0634, "step": 8390 }, { "epoch": 0.45, "learning_rate": 4.7782105377042085e-05, "loss": 2.1131, "step": 8400 }, { "epoch": 0.45, "learning_rate": 4.7779400627501894e-05, "loss": 2.0721, "step": 8410 }, { "epoch": 0.45, "learning_rate": 4.7776695877961704e-05, "loss": 2.0528, "step": 8420 }, { "epoch": 0.46, "learning_rate": 4.777399112842151e-05, "loss": 2.0833, "step": 8430 }, { "epoch": 0.46, "learning_rate": 4.7771286378881316e-05, "loss": 2.0631, "step": 8440 }, { "epoch": 0.46, "learning_rate": 4.7768581629341126e-05, "loss": 2.0645, "step": 8450 }, { "epoch": 0.46, "learning_rate": 4.776587687980093e-05, "loss": 2.0129, "step": 8460 }, { "epoch": 0.46, "learning_rate": 4.776317213026074e-05, "loss": 2.0344, "step": 8470 }, { "epoch": 0.46, "learning_rate": 4.776046738072055e-05, "loss": 1.9849, "step": 8480 }, { "epoch": 0.46, "learning_rate": 4.775776263118035e-05, "loss": 2.0415, "step": 8490 }, { "epoch": 0.46, "learning_rate": 4.775505788164016e-05, "loss": 2.0843, "step": 8500 }, { "epoch": 0.46, "learning_rate": 4.775235313209997e-05, "loss": 2.096, "step": 8510 }, { "epoch": 0.46, "learning_rate": 4.774964838255977e-05, "loss": 2.1046, "step": 8520 }, { "epoch": 0.46, "learning_rate": 4.774694363301958e-05, "loss": 2.07, "step": 8530 }, { "epoch": 0.46, "learning_rate": 4.774423888347939e-05, "loss": 2.0388, "step": 8540 }, { "epoch": 0.46, "learning_rate": 4.7741534133939195e-05, "loss": 2.0943, "step": 8550 }, { "epoch": 0.46, "learning_rate": 4.773882938439901e-05, "loss": 2.1133, "step": 8560 }, { "epoch": 0.46, "learning_rate": 4.7736124634858815e-05, "loss": 2.0801, "step": 8570 }, { "epoch": 0.46, "learning_rate": 4.773341988531862e-05, "loss": 2.1373, "step": 8580 }, { "epoch": 0.46, "learning_rate": 4.7730715135778434e-05, "loss": 2.064, "step": 8590 }, { "epoch": 0.46, "learning_rate": 4.772801038623824e-05, "loss": 2.0687, "step": 8600 }, { "epoch": 0.47, "learning_rate": 4.772530563669804e-05, "loss": 1.9777, "step": 8610 }, { "epoch": 0.47, "learning_rate": 4.7722600887157856e-05, "loss": 1.9211, "step": 8620 }, { "epoch": 0.47, "learning_rate": 4.771989613761766e-05, "loss": 2.0391, "step": 8630 }, { "epoch": 0.47, "learning_rate": 4.771719138807746e-05, "loss": 2.0094, "step": 8640 }, { "epoch": 0.47, "learning_rate": 4.771448663853728e-05, "loss": 2.1458, "step": 8650 }, { "epoch": 0.47, "learning_rate": 4.771178188899708e-05, "loss": 1.9533, "step": 8660 }, { "epoch": 0.47, "learning_rate": 4.7709077139456884e-05, "loss": 2.0308, "step": 8670 }, { "epoch": 0.47, "learning_rate": 4.77063723899167e-05, "loss": 2.0532, "step": 8680 }, { "epoch": 0.47, "learning_rate": 4.7703667640376503e-05, "loss": 2.0491, "step": 8690 }, { "epoch": 0.47, "learning_rate": 4.7700962890836306e-05, "loss": 2.0602, "step": 8700 }, { "epoch": 0.47, "learning_rate": 4.769825814129612e-05, "loss": 2.1349, "step": 8710 }, { "epoch": 0.47, "learning_rate": 4.7695553391755926e-05, "loss": 2.0361, "step": 8720 }, { "epoch": 0.47, "learning_rate": 4.769284864221573e-05, "loss": 2.1011, "step": 8730 }, { "epoch": 0.47, "learning_rate": 4.7690143892675545e-05, "loss": 2.1606, "step": 8740 }, { "epoch": 0.47, "learning_rate": 4.768743914313535e-05, "loss": 1.9794, "step": 8750 }, { "epoch": 0.47, "learning_rate": 4.768473439359515e-05, "loss": 1.9929, "step": 8760 }, { "epoch": 0.47, "learning_rate": 4.768202964405497e-05, "loss": 2.0818, "step": 8770 }, { "epoch": 0.47, "learning_rate": 4.767932489451477e-05, "loss": 2.0549, "step": 8780 }, { "epoch": 0.47, "learning_rate": 4.767662014497458e-05, "loss": 1.9329, "step": 8790 }, { "epoch": 0.48, "learning_rate": 4.767391539543439e-05, "loss": 2.0327, "step": 8800 }, { "epoch": 0.48, "learning_rate": 4.767121064589419e-05, "loss": 2.0493, "step": 8810 }, { "epoch": 0.48, "learning_rate": 4.7668505896354e-05, "loss": 2.0311, "step": 8820 }, { "epoch": 0.48, "learning_rate": 4.766580114681381e-05, "loss": 2.132, "step": 8830 }, { "epoch": 0.48, "learning_rate": 4.7663096397273614e-05, "loss": 2.0606, "step": 8840 }, { "epoch": 0.48, "learning_rate": 4.7660391647733424e-05, "loss": 2.0445, "step": 8850 }, { "epoch": 0.48, "learning_rate": 4.7657686898193234e-05, "loss": 1.9724, "step": 8860 }, { "epoch": 0.48, "learning_rate": 4.7654982148653036e-05, "loss": 2.1019, "step": 8870 }, { "epoch": 0.48, "learning_rate": 4.7652277399112846e-05, "loss": 2.0469, "step": 8880 }, { "epoch": 0.48, "learning_rate": 4.764957264957265e-05, "loss": 2.0993, "step": 8890 }, { "epoch": 0.48, "learning_rate": 4.764686790003246e-05, "loss": 1.9853, "step": 8900 }, { "epoch": 0.48, "learning_rate": 4.764416315049227e-05, "loss": 1.9066, "step": 8910 }, { "epoch": 0.48, "learning_rate": 4.764145840095207e-05, "loss": 2.0471, "step": 8920 }, { "epoch": 0.48, "learning_rate": 4.763875365141188e-05, "loss": 2.1091, "step": 8930 }, { "epoch": 0.48, "learning_rate": 4.763604890187169e-05, "loss": 2.0639, "step": 8940 }, { "epoch": 0.48, "learning_rate": 4.763334415233149e-05, "loss": 2.0842, "step": 8950 }, { "epoch": 0.48, "learning_rate": 4.76306394027913e-05, "loss": 2.0812, "step": 8960 }, { "epoch": 0.48, "learning_rate": 4.762793465325111e-05, "loss": 2.0466, "step": 8970 }, { "epoch": 0.49, "learning_rate": 4.7625229903710915e-05, "loss": 2.1558, "step": 8980 }, { "epoch": 0.49, "learning_rate": 4.7622525154170725e-05, "loss": 2.0903, "step": 8990 }, { "epoch": 0.49, "learning_rate": 4.7619820404630535e-05, "loss": 2.0617, "step": 9000 }, { "epoch": 0.49, "learning_rate": 4.761711565509034e-05, "loss": 2.018, "step": 9010 }, { "epoch": 0.49, "learning_rate": 4.761441090555015e-05, "loss": 2.0246, "step": 9020 }, { "epoch": 0.49, "learning_rate": 4.761170615600996e-05, "loss": 2.0211, "step": 9030 }, { "epoch": 0.49, "learning_rate": 4.760900140646976e-05, "loss": 2.0221, "step": 9040 }, { "epoch": 0.49, "learning_rate": 4.760629665692957e-05, "loss": 2.0744, "step": 9050 }, { "epoch": 0.49, "learning_rate": 4.760359190738938e-05, "loss": 2.01, "step": 9060 }, { "epoch": 0.49, "learning_rate": 4.760088715784918e-05, "loss": 2.0603, "step": 9070 }, { "epoch": 0.49, "learning_rate": 4.759818240830899e-05, "loss": 1.9935, "step": 9080 }, { "epoch": 0.49, "learning_rate": 4.75954776587688e-05, "loss": 2.0012, "step": 9090 }, { "epoch": 0.49, "learning_rate": 4.7592772909228604e-05, "loss": 2.0779, "step": 9100 }, { "epoch": 0.49, "learning_rate": 4.7590068159688414e-05, "loss": 2.0231, "step": 9110 }, { "epoch": 0.49, "learning_rate": 4.7587363410148223e-05, "loss": 1.9978, "step": 9120 }, { "epoch": 0.49, "learning_rate": 4.7584658660608026e-05, "loss": 2.0785, "step": 9130 }, { "epoch": 0.49, "learning_rate": 4.7581953911067836e-05, "loss": 2.0525, "step": 9140 }, { "epoch": 0.49, "learning_rate": 4.7579249161527646e-05, "loss": 2.1215, "step": 9150 }, { "epoch": 0.49, "learning_rate": 4.757654441198745e-05, "loss": 2.0308, "step": 9160 }, { "epoch": 0.5, "learning_rate": 4.757383966244726e-05, "loss": 2.0003, "step": 9170 }, { "epoch": 0.5, "learning_rate": 4.757113491290707e-05, "loss": 2.092, "step": 9180 }, { "epoch": 0.5, "learning_rate": 4.756843016336687e-05, "loss": 1.9986, "step": 9190 }, { "epoch": 0.5, "learning_rate": 4.756572541382669e-05, "loss": 2.0459, "step": 9200 }, { "epoch": 0.5, "learning_rate": 4.756302066428649e-05, "loss": 2.0266, "step": 9210 }, { "epoch": 0.5, "learning_rate": 4.756031591474629e-05, "loss": 1.9934, "step": 9220 }, { "epoch": 0.5, "learning_rate": 4.755761116520611e-05, "loss": 2.0032, "step": 9230 }, { "epoch": 0.5, "learning_rate": 4.755490641566591e-05, "loss": 2.0237, "step": 9240 }, { "epoch": 0.5, "learning_rate": 4.7552201666125715e-05, "loss": 1.9647, "step": 9250 }, { "epoch": 0.5, "learning_rate": 4.754949691658553e-05, "loss": 2.0454, "step": 9260 }, { "epoch": 0.5, "learning_rate": 4.7546792167045334e-05, "loss": 2.0622, "step": 9270 }, { "epoch": 0.5, "learning_rate": 4.754408741750514e-05, "loss": 2.1082, "step": 9280 }, { "epoch": 0.5, "learning_rate": 4.7541382667964954e-05, "loss": 2.0114, "step": 9290 }, { "epoch": 0.5, "learning_rate": 4.7538677918424756e-05, "loss": 2.0147, "step": 9300 }, { "epoch": 0.5, "learning_rate": 4.753597316888456e-05, "loss": 2.0018, "step": 9310 }, { "epoch": 0.5, "learning_rate": 4.7533268419344376e-05, "loss": 2.0682, "step": 9320 }, { "epoch": 0.5, "learning_rate": 4.753056366980418e-05, "loss": 2.0035, "step": 9330 }, { "epoch": 0.5, "learning_rate": 4.752785892026398e-05, "loss": 1.9684, "step": 9340 }, { "epoch": 0.51, "learning_rate": 4.75251541707238e-05, "loss": 2.1051, "step": 9350 }, { "epoch": 0.51, "learning_rate": 4.75224494211836e-05, "loss": 2.0515, "step": 9360 }, { "epoch": 0.51, "learning_rate": 4.7519744671643404e-05, "loss": 2.0352, "step": 9370 }, { "epoch": 0.51, "learning_rate": 4.751703992210322e-05, "loss": 1.9358, "step": 9380 }, { "epoch": 0.51, "learning_rate": 4.751433517256302e-05, "loss": 1.9975, "step": 9390 }, { "epoch": 0.51, "learning_rate": 4.751163042302283e-05, "loss": 1.9447, "step": 9400 }, { "epoch": 0.51, "learning_rate": 4.750892567348264e-05, "loss": 1.9507, "step": 9410 }, { "epoch": 0.51, "learning_rate": 4.7506220923942445e-05, "loss": 1.9846, "step": 9420 }, { "epoch": 0.51, "learning_rate": 4.7503516174402255e-05, "loss": 2.0913, "step": 9430 }, { "epoch": 0.51, "learning_rate": 4.7500811424862064e-05, "loss": 2.0237, "step": 9440 }, { "epoch": 0.51, "learning_rate": 4.749810667532187e-05, "loss": 2.0277, "step": 9450 }, { "epoch": 0.51, "learning_rate": 4.749540192578168e-05, "loss": 2.0229, "step": 9460 }, { "epoch": 0.51, "learning_rate": 4.749269717624148e-05, "loss": 2.0706, "step": 9470 }, { "epoch": 0.51, "learning_rate": 4.748999242670129e-05, "loss": 1.936, "step": 9480 }, { "epoch": 0.51, "learning_rate": 4.74872876771611e-05, "loss": 2.071, "step": 9490 }, { "epoch": 0.51, "learning_rate": 4.74845829276209e-05, "loss": 2.0435, "step": 9500 }, { "epoch": 0.51, "learning_rate": 4.748187817808071e-05, "loss": 1.9748, "step": 9510 }, { "epoch": 0.51, "learning_rate": 4.747917342854052e-05, "loss": 2.0897, "step": 9520 }, { "epoch": 0.51, "learning_rate": 4.7476468679000324e-05, "loss": 2.1553, "step": 9530 }, { "epoch": 0.52, "learning_rate": 4.7473763929460134e-05, "loss": 2.0131, "step": 9540 }, { "epoch": 0.52, "learning_rate": 4.7471059179919943e-05, "loss": 2.0306, "step": 9550 }, { "epoch": 0.52, "learning_rate": 4.7468354430379746e-05, "loss": 2.074, "step": 9560 }, { "epoch": 0.52, "learning_rate": 4.7465649680839556e-05, "loss": 2.0265, "step": 9570 }, { "epoch": 0.52, "learning_rate": 4.7462944931299366e-05, "loss": 1.9422, "step": 9580 }, { "epoch": 0.52, "learning_rate": 4.746024018175917e-05, "loss": 2.0663, "step": 9590 }, { "epoch": 0.52, "learning_rate": 4.745753543221898e-05, "loss": 1.9385, "step": 9600 }, { "epoch": 0.52, "learning_rate": 4.745483068267879e-05, "loss": 2.1486, "step": 9610 }, { "epoch": 0.52, "learning_rate": 4.745212593313859e-05, "loss": 2.1181, "step": 9620 }, { "epoch": 0.52, "learning_rate": 4.74494211835984e-05, "loss": 2.028, "step": 9630 }, { "epoch": 0.52, "learning_rate": 4.744671643405821e-05, "loss": 2.081, "step": 9640 }, { "epoch": 0.52, "learning_rate": 4.744401168451801e-05, "loss": 1.9565, "step": 9650 }, { "epoch": 0.52, "learning_rate": 4.744130693497782e-05, "loss": 2.0188, "step": 9660 }, { "epoch": 0.52, "learning_rate": 4.743860218543763e-05, "loss": 2.1093, "step": 9670 }, { "epoch": 0.52, "learning_rate": 4.7435897435897435e-05, "loss": 2.0069, "step": 9680 }, { "epoch": 0.52, "learning_rate": 4.7433192686357245e-05, "loss": 2.0698, "step": 9690 }, { "epoch": 0.52, "learning_rate": 4.7430487936817054e-05, "loss": 2.0366, "step": 9700 }, { "epoch": 0.52, "learning_rate": 4.742778318727686e-05, "loss": 2.0603, "step": 9710 }, { "epoch": 0.53, "learning_rate": 4.742507843773667e-05, "loss": 2.0522, "step": 9720 }, { "epoch": 0.53, "learning_rate": 4.7422373688196477e-05, "loss": 2.1171, "step": 9730 }, { "epoch": 0.53, "learning_rate": 4.741966893865628e-05, "loss": 2.062, "step": 9740 }, { "epoch": 0.53, "learning_rate": 4.741696418911609e-05, "loss": 2.0264, "step": 9750 }, { "epoch": 0.53, "learning_rate": 4.74142594395759e-05, "loss": 2.035, "step": 9760 }, { "epoch": 0.53, "learning_rate": 4.74115546900357e-05, "loss": 2.0933, "step": 9770 }, { "epoch": 0.53, "learning_rate": 4.740884994049551e-05, "loss": 1.9805, "step": 9780 }, { "epoch": 0.53, "learning_rate": 4.740614519095532e-05, "loss": 1.9706, "step": 9790 }, { "epoch": 0.53, "learning_rate": 4.7403440441415124e-05, "loss": 2.0386, "step": 9800 }, { "epoch": 0.53, "learning_rate": 4.740073569187494e-05, "loss": 1.9998, "step": 9810 }, { "epoch": 0.53, "learning_rate": 4.739803094233474e-05, "loss": 1.9846, "step": 9820 }, { "epoch": 0.53, "learning_rate": 4.7395326192794546e-05, "loss": 2.0532, "step": 9830 }, { "epoch": 0.53, "learning_rate": 4.739262144325436e-05, "loss": 2.0592, "step": 9840 }, { "epoch": 0.53, "learning_rate": 4.7389916693714165e-05, "loss": 2.0503, "step": 9850 }, { "epoch": 0.53, "learning_rate": 4.738721194417397e-05, "loss": 2.074, "step": 9860 }, { "epoch": 0.53, "learning_rate": 4.7384507194633784e-05, "loss": 1.9517, "step": 9870 }, { "epoch": 0.53, "learning_rate": 4.738180244509359e-05, "loss": 2.0516, "step": 9880 }, { "epoch": 0.53, "learning_rate": 4.737909769555339e-05, "loss": 2.0256, "step": 9890 }, { "epoch": 0.53, "learning_rate": 4.737639294601321e-05, "loss": 2.0007, "step": 9900 }, { "epoch": 0.54, "learning_rate": 4.737368819647301e-05, "loss": 1.9803, "step": 9910 }, { "epoch": 0.54, "learning_rate": 4.737098344693281e-05, "loss": 2.0128, "step": 9920 }, { "epoch": 0.54, "learning_rate": 4.736827869739263e-05, "loss": 2.0327, "step": 9930 }, { "epoch": 0.54, "learning_rate": 4.736557394785243e-05, "loss": 2.0247, "step": 9940 }, { "epoch": 0.54, "learning_rate": 4.7362869198312235e-05, "loss": 2.0718, "step": 9950 }, { "epoch": 0.54, "learning_rate": 4.736016444877205e-05, "loss": 1.9632, "step": 9960 }, { "epoch": 0.54, "learning_rate": 4.7357459699231854e-05, "loss": 2.0498, "step": 9970 }, { "epoch": 0.54, "learning_rate": 4.735475494969166e-05, "loss": 2.0346, "step": 9980 }, { "epoch": 0.54, "learning_rate": 4.735205020015147e-05, "loss": 2.068, "step": 9990 }, { "epoch": 0.54, "learning_rate": 4.7349345450611276e-05, "loss": 2.028, "step": 10000 }, { "epoch": 0.54, "learning_rate": 4.734664070107108e-05, "loss": 2.0537, "step": 10010 }, { "epoch": 0.54, "learning_rate": 4.734393595153089e-05, "loss": 1.9852, "step": 10020 }, { "epoch": 0.54, "learning_rate": 4.73412312019907e-05, "loss": 2.0547, "step": 10030 }, { "epoch": 0.54, "learning_rate": 4.733852645245051e-05, "loss": 1.9707, "step": 10040 }, { "epoch": 0.54, "learning_rate": 4.733582170291031e-05, "loss": 1.8892, "step": 10050 }, { "epoch": 0.54, "learning_rate": 4.733311695337012e-05, "loss": 2.0548, "step": 10060 }, { "epoch": 0.54, "learning_rate": 4.733041220382993e-05, "loss": 2.0872, "step": 10070 }, { "epoch": 0.54, "learning_rate": 4.732770745428973e-05, "loss": 1.9218, "step": 10080 }, { "epoch": 0.55, "learning_rate": 4.732500270474954e-05, "loss": 1.9968, "step": 10090 }, { "epoch": 0.55, "learning_rate": 4.732229795520935e-05, "loss": 2.0443, "step": 10100 }, { "epoch": 0.55, "learning_rate": 4.7319593205669155e-05, "loss": 2.0608, "step": 10110 }, { "epoch": 0.55, "learning_rate": 4.7316888456128965e-05, "loss": 2.035, "step": 10120 }, { "epoch": 0.55, "learning_rate": 4.7314183706588774e-05, "loss": 2.0234, "step": 10130 }, { "epoch": 0.55, "learning_rate": 4.731147895704858e-05, "loss": 1.9899, "step": 10140 }, { "epoch": 0.55, "learning_rate": 4.730877420750839e-05, "loss": 2.0227, "step": 10150 }, { "epoch": 0.55, "learning_rate": 4.7306069457968197e-05, "loss": 2.0822, "step": 10160 }, { "epoch": 0.55, "learning_rate": 4.7303364708428e-05, "loss": 2.0295, "step": 10170 }, { "epoch": 0.55, "learning_rate": 4.730065995888781e-05, "loss": 2.0423, "step": 10180 }, { "epoch": 0.55, "learning_rate": 4.729795520934762e-05, "loss": 2.0018, "step": 10190 }, { "epoch": 0.55, "learning_rate": 4.729525045980742e-05, "loss": 2.0548, "step": 10200 }, { "epoch": 0.55, "learning_rate": 4.729254571026723e-05, "loss": 2.0058, "step": 10210 }, { "epoch": 0.55, "learning_rate": 4.728984096072704e-05, "loss": 2.0512, "step": 10220 }, { "epoch": 0.55, "learning_rate": 4.7287136211186844e-05, "loss": 2.0152, "step": 10230 }, { "epoch": 0.55, "learning_rate": 4.7284431461646653e-05, "loss": 1.9959, "step": 10240 }, { "epoch": 0.55, "learning_rate": 4.728172671210646e-05, "loss": 2.0351, "step": 10250 }, { "epoch": 0.55, "learning_rate": 4.7279021962566266e-05, "loss": 2.0155, "step": 10260 }, { "epoch": 0.55, "learning_rate": 4.7276317213026076e-05, "loss": 1.968, "step": 10270 }, { "epoch": 0.56, "learning_rate": 4.7273612463485885e-05, "loss": 1.9948, "step": 10280 }, { "epoch": 0.56, "learning_rate": 4.727090771394569e-05, "loss": 2.0008, "step": 10290 }, { "epoch": 0.56, "learning_rate": 4.72682029644055e-05, "loss": 2.0428, "step": 10300 }, { "epoch": 0.56, "learning_rate": 4.726549821486531e-05, "loss": 2.025, "step": 10310 }, { "epoch": 0.56, "learning_rate": 4.726279346532511e-05, "loss": 1.979, "step": 10320 }, { "epoch": 0.56, "learning_rate": 4.726008871578492e-05, "loss": 2.0706, "step": 10330 }, { "epoch": 0.56, "learning_rate": 4.725738396624473e-05, "loss": 2.0615, "step": 10340 }, { "epoch": 0.56, "learning_rate": 4.725467921670453e-05, "loss": 2.0019, "step": 10350 }, { "epoch": 0.56, "learning_rate": 4.725197446716434e-05, "loss": 2.0208, "step": 10360 }, { "epoch": 0.56, "learning_rate": 4.724926971762415e-05, "loss": 1.9969, "step": 10370 }, { "epoch": 0.56, "learning_rate": 4.7246564968083955e-05, "loss": 1.9933, "step": 10380 }, { "epoch": 0.56, "learning_rate": 4.7243860218543764e-05, "loss": 2.0776, "step": 10390 }, { "epoch": 0.56, "learning_rate": 4.7241155469003574e-05, "loss": 2.0903, "step": 10400 }, { "epoch": 0.56, "learning_rate": 4.723845071946338e-05, "loss": 2.077, "step": 10410 }, { "epoch": 0.56, "learning_rate": 4.7235745969923186e-05, "loss": 2.0351, "step": 10420 }, { "epoch": 0.56, "learning_rate": 4.7233041220382996e-05, "loss": 2.0759, "step": 10430 }, { "epoch": 0.56, "learning_rate": 4.72303364708428e-05, "loss": 1.9787, "step": 10440 }, { "epoch": 0.56, "learning_rate": 4.7227631721302615e-05, "loss": 2.1055, "step": 10450 }, { "epoch": 0.57, "learning_rate": 4.722492697176242e-05, "loss": 2.1121, "step": 10460 }, { "epoch": 0.57, "learning_rate": 4.722222222222222e-05, "loss": 1.9822, "step": 10470 }, { "epoch": 0.57, "learning_rate": 4.721951747268204e-05, "loss": 1.9373, "step": 10480 }, { "epoch": 0.57, "learning_rate": 4.721681272314184e-05, "loss": 2.0994, "step": 10490 }, { "epoch": 0.57, "learning_rate": 4.721410797360164e-05, "loss": 1.9989, "step": 10500 }, { "epoch": 0.57, "learning_rate": 4.721140322406146e-05, "loss": 1.96, "step": 10510 }, { "epoch": 0.57, "learning_rate": 4.720869847452126e-05, "loss": 2.1242, "step": 10520 }, { "epoch": 0.57, "learning_rate": 4.7205993724981065e-05, "loss": 1.9576, "step": 10530 }, { "epoch": 0.57, "learning_rate": 4.720328897544088e-05, "loss": 1.9961, "step": 10540 }, { "epoch": 0.57, "learning_rate": 4.7200584225900685e-05, "loss": 1.9726, "step": 10550 }, { "epoch": 0.57, "learning_rate": 4.719787947636049e-05, "loss": 2.0333, "step": 10560 }, { "epoch": 0.57, "learning_rate": 4.71951747268203e-05, "loss": 2.0265, "step": 10570 }, { "epoch": 0.57, "learning_rate": 4.719246997728011e-05, "loss": 2.0764, "step": 10580 }, { "epoch": 0.57, "learning_rate": 4.718976522773991e-05, "loss": 1.9802, "step": 10590 }, { "epoch": 0.57, "learning_rate": 4.718706047819972e-05, "loss": 1.9911, "step": 10600 }, { "epoch": 0.57, "learning_rate": 4.718435572865953e-05, "loss": 1.9743, "step": 10610 }, { "epoch": 0.57, "learning_rate": 4.718165097911933e-05, "loss": 2.045, "step": 10620 }, { "epoch": 0.57, "learning_rate": 4.717894622957914e-05, "loss": 2.0374, "step": 10630 }, { "epoch": 0.57, "learning_rate": 4.717624148003895e-05, "loss": 1.9863, "step": 10640 }, { "epoch": 0.58, "learning_rate": 4.7173536730498754e-05, "loss": 2.0171, "step": 10650 }, { "epoch": 0.58, "learning_rate": 4.7170831980958564e-05, "loss": 1.955, "step": 10660 }, { "epoch": 0.58, "learning_rate": 4.7168127231418373e-05, "loss": 1.9789, "step": 10670 }, { "epoch": 0.58, "learning_rate": 4.716542248187818e-05, "loss": 2.0487, "step": 10680 }, { "epoch": 0.58, "learning_rate": 4.7162717732337986e-05, "loss": 2.0577, "step": 10690 }, { "epoch": 0.58, "learning_rate": 4.7160012982797796e-05, "loss": 1.9839, "step": 10700 }, { "epoch": 0.58, "learning_rate": 4.7157308233257605e-05, "loss": 2.032, "step": 10710 }, { "epoch": 0.58, "learning_rate": 4.715460348371741e-05, "loss": 1.9894, "step": 10720 }, { "epoch": 0.58, "learning_rate": 4.715189873417722e-05, "loss": 2.0444, "step": 10730 }, { "epoch": 0.58, "learning_rate": 4.714919398463703e-05, "loss": 1.9831, "step": 10740 }, { "epoch": 0.58, "learning_rate": 4.714648923509683e-05, "loss": 2.0789, "step": 10750 }, { "epoch": 0.58, "learning_rate": 4.714378448555664e-05, "loss": 2.027, "step": 10760 }, { "epoch": 0.58, "learning_rate": 4.714107973601645e-05, "loss": 1.9746, "step": 10770 }, { "epoch": 0.58, "learning_rate": 4.713837498647625e-05, "loss": 2.046, "step": 10780 }, { "epoch": 0.58, "learning_rate": 4.713567023693606e-05, "loss": 2.004, "step": 10790 }, { "epoch": 0.58, "learning_rate": 4.713296548739587e-05, "loss": 2.029, "step": 10800 }, { "epoch": 0.58, "learning_rate": 4.7130260737855675e-05, "loss": 1.9807, "step": 10810 }, { "epoch": 0.58, "learning_rate": 4.7127555988315484e-05, "loss": 1.9846, "step": 10820 }, { "epoch": 0.59, "learning_rate": 4.7124851238775294e-05, "loss": 2.0337, "step": 10830 }, { "epoch": 0.59, "learning_rate": 4.71221464892351e-05, "loss": 1.9849, "step": 10840 }, { "epoch": 0.59, "learning_rate": 4.7119441739694906e-05, "loss": 2.0304, "step": 10850 }, { "epoch": 0.59, "learning_rate": 4.7116736990154716e-05, "loss": 2.0274, "step": 10860 }, { "epoch": 0.59, "learning_rate": 4.711403224061452e-05, "loss": 1.9902, "step": 10870 }, { "epoch": 0.59, "learning_rate": 4.711132749107433e-05, "loss": 2.0399, "step": 10880 }, { "epoch": 0.59, "learning_rate": 4.710862274153414e-05, "loss": 2.0101, "step": 10890 }, { "epoch": 0.59, "learning_rate": 4.710591799199394e-05, "loss": 2.0229, "step": 10900 }, { "epoch": 0.59, "learning_rate": 4.710321324245375e-05, "loss": 2.0818, "step": 10910 }, { "epoch": 0.59, "learning_rate": 4.710050849291356e-05, "loss": 1.962, "step": 10920 }, { "epoch": 0.59, "learning_rate": 4.709780374337336e-05, "loss": 2.0181, "step": 10930 }, { "epoch": 0.59, "learning_rate": 4.709509899383317e-05, "loss": 2.0493, "step": 10940 }, { "epoch": 0.59, "learning_rate": 4.709239424429298e-05, "loss": 2.0125, "step": 10950 }, { "epoch": 0.59, "learning_rate": 4.7089689494752785e-05, "loss": 2.0497, "step": 10960 }, { "epoch": 0.59, "learning_rate": 4.7086984745212595e-05, "loss": 2.0308, "step": 10970 }, { "epoch": 0.59, "learning_rate": 4.7084279995672405e-05, "loss": 1.9861, "step": 10980 }, { "epoch": 0.59, "learning_rate": 4.708157524613221e-05, "loss": 2.1276, "step": 10990 }, { "epoch": 0.59, "learning_rate": 4.707887049659202e-05, "loss": 2.0325, "step": 11000 }, { "epoch": 0.59, "learning_rate": 4.707616574705183e-05, "loss": 2.075, "step": 11010 }, { "epoch": 0.6, "learning_rate": 4.707346099751163e-05, "loss": 2.0679, "step": 11020 }, { "epoch": 0.6, "learning_rate": 4.707075624797144e-05, "loss": 2.0252, "step": 11030 }, { "epoch": 0.6, "learning_rate": 4.706805149843125e-05, "loss": 1.9244, "step": 11040 }, { "epoch": 0.6, "learning_rate": 4.706534674889105e-05, "loss": 1.9163, "step": 11050 }, { "epoch": 0.6, "learning_rate": 4.706264199935086e-05, "loss": 2.0002, "step": 11060 }, { "epoch": 0.6, "learning_rate": 4.705993724981067e-05, "loss": 1.998, "step": 11070 }, { "epoch": 0.6, "learning_rate": 4.7057232500270474e-05, "loss": 1.9824, "step": 11080 }, { "epoch": 0.6, "learning_rate": 4.705452775073029e-05, "loss": 2.0427, "step": 11090 }, { "epoch": 0.6, "learning_rate": 4.7051823001190093e-05, "loss": 2.0855, "step": 11100 }, { "epoch": 0.6, "learning_rate": 4.7049118251649896e-05, "loss": 1.9815, "step": 11110 }, { "epoch": 0.6, "learning_rate": 4.704641350210971e-05, "loss": 1.9419, "step": 11120 }, { "epoch": 0.6, "learning_rate": 4.7043708752569516e-05, "loss": 2.0973, "step": 11130 }, { "epoch": 0.6, "learning_rate": 4.704100400302932e-05, "loss": 1.9972, "step": 11140 }, { "epoch": 0.6, "learning_rate": 4.703829925348913e-05, "loss": 1.9456, "step": 11150 }, { "epoch": 0.6, "learning_rate": 4.703559450394894e-05, "loss": 2.0072, "step": 11160 }, { "epoch": 0.6, "learning_rate": 4.703288975440874e-05, "loss": 2.0735, "step": 11170 }, { "epoch": 0.6, "learning_rate": 4.703018500486855e-05, "loss": 2.0699, "step": 11180 }, { "epoch": 0.6, "learning_rate": 4.702748025532836e-05, "loss": 1.9315, "step": 11190 }, { "epoch": 0.61, "learning_rate": 4.702477550578816e-05, "loss": 1.9904, "step": 11200 }, { "epoch": 0.61, "learning_rate": 4.702207075624797e-05, "loss": 1.9329, "step": 11210 }, { "epoch": 0.61, "learning_rate": 4.701936600670778e-05, "loss": 2.0042, "step": 11220 }, { "epoch": 0.61, "learning_rate": 4.7016661257167585e-05, "loss": 2.1943, "step": 11230 }, { "epoch": 0.61, "learning_rate": 4.7013956507627395e-05, "loss": 2.0125, "step": 11240 }, { "epoch": 0.61, "learning_rate": 4.7011251758087204e-05, "loss": 1.9748, "step": 11250 }, { "epoch": 0.61, "learning_rate": 4.700854700854701e-05, "loss": 2.0313, "step": 11260 }, { "epoch": 0.61, "learning_rate": 4.700584225900682e-05, "loss": 2.0229, "step": 11270 }, { "epoch": 0.61, "learning_rate": 4.7003137509466626e-05, "loss": 2.0011, "step": 11280 }, { "epoch": 0.61, "learning_rate": 4.700043275992643e-05, "loss": 2.025, "step": 11290 }, { "epoch": 0.61, "learning_rate": 4.699772801038624e-05, "loss": 1.9807, "step": 11300 }, { "epoch": 0.61, "learning_rate": 4.699502326084605e-05, "loss": 1.9153, "step": 11310 }, { "epoch": 0.61, "learning_rate": 4.699231851130586e-05, "loss": 2.1053, "step": 11320 }, { "epoch": 0.61, "learning_rate": 4.698961376176566e-05, "loss": 2.063, "step": 11330 }, { "epoch": 0.61, "learning_rate": 4.698690901222547e-05, "loss": 2.0107, "step": 11340 }, { "epoch": 0.61, "learning_rate": 4.698420426268528e-05, "loss": 2.0246, "step": 11350 }, { "epoch": 0.61, "learning_rate": 4.698149951314508e-05, "loss": 2.0653, "step": 11360 }, { "epoch": 0.61, "learning_rate": 4.697879476360489e-05, "loss": 2.0084, "step": 11370 }, { "epoch": 0.61, "learning_rate": 4.69760900140647e-05, "loss": 2.0577, "step": 11380 }, { "epoch": 0.62, "learning_rate": 4.6973385264524505e-05, "loss": 2.0183, "step": 11390 }, { "epoch": 0.62, "learning_rate": 4.6970680514984315e-05, "loss": 1.9773, "step": 11400 }, { "epoch": 0.62, "learning_rate": 4.6967975765444125e-05, "loss": 1.9712, "step": 11410 }, { "epoch": 0.62, "learning_rate": 4.696527101590393e-05, "loss": 2.0239, "step": 11420 }, { "epoch": 0.62, "learning_rate": 4.696256626636374e-05, "loss": 2.0211, "step": 11430 }, { "epoch": 0.62, "learning_rate": 4.695986151682355e-05, "loss": 2.0465, "step": 11440 }, { "epoch": 0.62, "learning_rate": 4.695715676728335e-05, "loss": 2.0756, "step": 11450 }, { "epoch": 0.62, "learning_rate": 4.695445201774316e-05, "loss": 2.0099, "step": 11460 }, { "epoch": 0.62, "learning_rate": 4.695174726820297e-05, "loss": 2.0457, "step": 11470 }, { "epoch": 0.62, "learning_rate": 4.694904251866277e-05, "loss": 1.9559, "step": 11480 }, { "epoch": 0.62, "learning_rate": 4.694633776912258e-05, "loss": 2.0099, "step": 11490 }, { "epoch": 0.62, "learning_rate": 4.694363301958239e-05, "loss": 1.9039, "step": 11500 }, { "epoch": 0.62, "learning_rate": 4.6940928270042194e-05, "loss": 1.9989, "step": 11510 }, { "epoch": 0.62, "learning_rate": 4.6938223520502004e-05, "loss": 2.0522, "step": 11520 }, { "epoch": 0.62, "learning_rate": 4.6935518770961813e-05, "loss": 1.9534, "step": 11530 }, { "epoch": 0.62, "learning_rate": 4.6932814021421616e-05, "loss": 1.9293, "step": 11540 }, { "epoch": 0.62, "learning_rate": 4.6930109271881426e-05, "loss": 2.0551, "step": 11550 }, { "epoch": 0.62, "learning_rate": 4.6927404522341236e-05, "loss": 2.0393, "step": 11560 }, { "epoch": 0.63, "learning_rate": 4.692469977280104e-05, "loss": 1.9358, "step": 11570 }, { "epoch": 0.63, "learning_rate": 4.692199502326085e-05, "loss": 1.9273, "step": 11580 }, { "epoch": 0.63, "learning_rate": 4.691929027372066e-05, "loss": 2.0029, "step": 11590 }, { "epoch": 0.63, "learning_rate": 4.691658552418046e-05, "loss": 1.9469, "step": 11600 }, { "epoch": 0.63, "learning_rate": 4.691388077464027e-05, "loss": 2.0316, "step": 11610 }, { "epoch": 0.63, "learning_rate": 4.691117602510008e-05, "loss": 2.0294, "step": 11620 }, { "epoch": 0.63, "learning_rate": 4.690847127555988e-05, "loss": 2.0328, "step": 11630 }, { "epoch": 0.63, "learning_rate": 4.690576652601969e-05, "loss": 2.0016, "step": 11640 }, { "epoch": 0.63, "learning_rate": 4.69030617764795e-05, "loss": 2.0557, "step": 11650 }, { "epoch": 0.63, "learning_rate": 4.6900357026939305e-05, "loss": 1.9889, "step": 11660 }, { "epoch": 0.63, "learning_rate": 4.6897652277399115e-05, "loss": 1.9842, "step": 11670 }, { "epoch": 0.63, "learning_rate": 4.6894947527858924e-05, "loss": 2.0232, "step": 11680 }, { "epoch": 0.63, "learning_rate": 4.689224277831873e-05, "loss": 2.0858, "step": 11690 }, { "epoch": 0.63, "learning_rate": 4.688953802877854e-05, "loss": 2.0648, "step": 11700 }, { "epoch": 0.63, "learning_rate": 4.6886833279238346e-05, "loss": 2.0419, "step": 11710 }, { "epoch": 0.63, "learning_rate": 4.688412852969815e-05, "loss": 2.0219, "step": 11720 }, { "epoch": 0.63, "learning_rate": 4.688142378015796e-05, "loss": 1.8824, "step": 11730 }, { "epoch": 0.63, "learning_rate": 4.687871903061777e-05, "loss": 1.9771, "step": 11740 }, { "epoch": 0.63, "learning_rate": 4.687601428107757e-05, "loss": 1.9494, "step": 11750 }, { "epoch": 0.64, "learning_rate": 4.687330953153738e-05, "loss": 2.0859, "step": 11760 }, { "epoch": 0.64, "learning_rate": 4.687060478199719e-05, "loss": 2.0756, "step": 11770 }, { "epoch": 0.64, "learning_rate": 4.6867900032456994e-05, "loss": 1.9563, "step": 11780 }, { "epoch": 0.64, "learning_rate": 4.68651952829168e-05, "loss": 2.0631, "step": 11790 }, { "epoch": 0.64, "learning_rate": 4.686249053337661e-05, "loss": 1.9899, "step": 11800 }, { "epoch": 0.64, "learning_rate": 4.6859785783836416e-05, "loss": 1.9812, "step": 11810 }, { "epoch": 0.64, "learning_rate": 4.6857081034296225e-05, "loss": 2.0242, "step": 11820 }, { "epoch": 0.64, "learning_rate": 4.6854376284756035e-05, "loss": 1.8974, "step": 11830 }, { "epoch": 0.64, "learning_rate": 4.685167153521584e-05, "loss": 1.9637, "step": 11840 }, { "epoch": 0.64, "learning_rate": 4.684896678567565e-05, "loss": 1.9813, "step": 11850 }, { "epoch": 0.64, "learning_rate": 4.684626203613546e-05, "loss": 2.0516, "step": 11860 }, { "epoch": 0.64, "learning_rate": 4.684355728659526e-05, "loss": 2.0519, "step": 11870 }, { "epoch": 0.64, "learning_rate": 4.684085253705507e-05, "loss": 2.0246, "step": 11880 }, { "epoch": 0.64, "learning_rate": 4.683814778751488e-05, "loss": 1.8794, "step": 11890 }, { "epoch": 0.64, "learning_rate": 4.683544303797468e-05, "loss": 2.0736, "step": 11900 }, { "epoch": 0.64, "learning_rate": 4.683273828843449e-05, "loss": 2.039, "step": 11910 }, { "epoch": 0.64, "learning_rate": 4.68300335388943e-05, "loss": 1.956, "step": 11920 }, { "epoch": 0.64, "learning_rate": 4.682732878935411e-05, "loss": 2.0906, "step": 11930 }, { "epoch": 0.65, "learning_rate": 4.6824624039813914e-05, "loss": 2.0031, "step": 11940 }, { "epoch": 0.65, "learning_rate": 4.6821919290273724e-05, "loss": 2.0692, "step": 11950 }, { "epoch": 0.65, "learning_rate": 4.6819214540733533e-05, "loss": 2.0357, "step": 11960 }, { "epoch": 0.65, "learning_rate": 4.6816509791193336e-05, "loss": 1.9601, "step": 11970 }, { "epoch": 0.65, "learning_rate": 4.6813805041653146e-05, "loss": 2.0344, "step": 11980 }, { "epoch": 0.65, "learning_rate": 4.6811100292112956e-05, "loss": 2.0176, "step": 11990 }, { "epoch": 0.65, "learning_rate": 4.680839554257276e-05, "loss": 2.0127, "step": 12000 }, { "epoch": 0.65, "learning_rate": 4.680569079303257e-05, "loss": 1.9252, "step": 12010 }, { "epoch": 0.65, "learning_rate": 4.680298604349238e-05, "loss": 2.093, "step": 12020 }, { "epoch": 0.65, "learning_rate": 4.680028129395218e-05, "loss": 2.0971, "step": 12030 }, { "epoch": 0.65, "learning_rate": 4.679757654441199e-05, "loss": 1.9587, "step": 12040 }, { "epoch": 0.65, "learning_rate": 4.67948717948718e-05, "loss": 1.9995, "step": 12050 }, { "epoch": 0.65, "learning_rate": 4.67921670453316e-05, "loss": 2.084, "step": 12060 }, { "epoch": 0.65, "learning_rate": 4.678946229579141e-05, "loss": 1.8575, "step": 12070 }, { "epoch": 0.65, "learning_rate": 4.678675754625122e-05, "loss": 1.9833, "step": 12080 }, { "epoch": 0.65, "learning_rate": 4.6784052796711025e-05, "loss": 1.9727, "step": 12090 }, { "epoch": 0.65, "learning_rate": 4.6781348047170835e-05, "loss": 1.9595, "step": 12100 }, { "epoch": 0.65, "learning_rate": 4.6778643297630644e-05, "loss": 2.0029, "step": 12110 }, { "epoch": 0.65, "learning_rate": 4.677593854809045e-05, "loss": 2.0567, "step": 12120 }, { "epoch": 0.66, "learning_rate": 4.677323379855026e-05, "loss": 2.0207, "step": 12130 }, { "epoch": 0.66, "learning_rate": 4.6770529049010066e-05, "loss": 1.935, "step": 12140 }, { "epoch": 0.66, "learning_rate": 4.676782429946987e-05, "loss": 1.9935, "step": 12150 }, { "epoch": 0.66, "learning_rate": 4.676511954992968e-05, "loss": 1.9997, "step": 12160 }, { "epoch": 0.66, "learning_rate": 4.676241480038949e-05, "loss": 2.0024, "step": 12170 }, { "epoch": 0.66, "learning_rate": 4.675971005084929e-05, "loss": 1.9983, "step": 12180 }, { "epoch": 0.66, "learning_rate": 4.67570053013091e-05, "loss": 2.0302, "step": 12190 }, { "epoch": 0.66, "learning_rate": 4.675430055176891e-05, "loss": 2.0889, "step": 12200 }, { "epoch": 0.66, "learning_rate": 4.6751595802228714e-05, "loss": 1.9845, "step": 12210 }, { "epoch": 0.66, "learning_rate": 4.674889105268852e-05, "loss": 2.0971, "step": 12220 }, { "epoch": 0.66, "learning_rate": 4.674618630314833e-05, "loss": 1.9707, "step": 12230 }, { "epoch": 0.66, "learning_rate": 4.6743481553608136e-05, "loss": 1.9131, "step": 12240 }, { "epoch": 0.66, "learning_rate": 4.6740776804067946e-05, "loss": 1.967, "step": 12250 }, { "epoch": 0.66, "learning_rate": 4.6738072054527755e-05, "loss": 1.9749, "step": 12260 }, { "epoch": 0.66, "learning_rate": 4.673536730498756e-05, "loss": 2.0202, "step": 12270 }, { "epoch": 0.66, "learning_rate": 4.673266255544737e-05, "loss": 1.9641, "step": 12280 }, { "epoch": 0.66, "learning_rate": 4.672995780590718e-05, "loss": 2.0283, "step": 12290 }, { "epoch": 0.66, "learning_rate": 4.672725305636698e-05, "loss": 1.9484, "step": 12300 }, { "epoch": 0.67, "learning_rate": 4.672454830682679e-05, "loss": 2.0054, "step": 12310 }, { "epoch": 0.67, "learning_rate": 4.67218435572866e-05, "loss": 1.9791, "step": 12320 }, { "epoch": 0.67, "learning_rate": 4.67191388077464e-05, "loss": 2.0773, "step": 12330 }, { "epoch": 0.67, "learning_rate": 4.671643405820621e-05, "loss": 2.0302, "step": 12340 }, { "epoch": 0.67, "learning_rate": 4.671372930866602e-05, "loss": 2.0193, "step": 12350 }, { "epoch": 0.67, "learning_rate": 4.6711024559125825e-05, "loss": 2.0345, "step": 12360 }, { "epoch": 0.67, "learning_rate": 4.6708319809585634e-05, "loss": 1.9268, "step": 12370 }, { "epoch": 0.67, "learning_rate": 4.6705615060045444e-05, "loss": 2.0105, "step": 12380 }, { "epoch": 0.67, "learning_rate": 4.670291031050525e-05, "loss": 2.0474, "step": 12390 }, { "epoch": 0.67, "learning_rate": 4.6700205560965056e-05, "loss": 2.0194, "step": 12400 }, { "epoch": 0.67, "learning_rate": 4.6697500811424866e-05, "loss": 1.9469, "step": 12410 }, { "epoch": 0.67, "learning_rate": 4.669479606188467e-05, "loss": 2.0846, "step": 12420 }, { "epoch": 0.67, "learning_rate": 4.669209131234448e-05, "loss": 2.0114, "step": 12430 }, { "epoch": 0.67, "learning_rate": 4.668938656280429e-05, "loss": 2.0116, "step": 12440 }, { "epoch": 0.67, "learning_rate": 4.668668181326409e-05, "loss": 2.0241, "step": 12450 }, { "epoch": 0.67, "learning_rate": 4.66839770637239e-05, "loss": 2.0393, "step": 12460 }, { "epoch": 0.67, "learning_rate": 4.668127231418371e-05, "loss": 1.9844, "step": 12470 }, { "epoch": 0.67, "learning_rate": 4.667856756464351e-05, "loss": 2.0415, "step": 12480 }, { "epoch": 0.67, "learning_rate": 4.667586281510332e-05, "loss": 2.0031, "step": 12490 }, { "epoch": 0.68, "learning_rate": 4.667315806556313e-05, "loss": 2.0576, "step": 12500 }, { "epoch": 0.68, "learning_rate": 4.6670453316022935e-05, "loss": 1.8986, "step": 12510 }, { "epoch": 0.68, "learning_rate": 4.6667748566482745e-05, "loss": 2.0434, "step": 12520 }, { "epoch": 0.68, "learning_rate": 4.6665043816942555e-05, "loss": 2.0192, "step": 12530 }, { "epoch": 0.68, "learning_rate": 4.666233906740236e-05, "loss": 1.9554, "step": 12540 }, { "epoch": 0.68, "learning_rate": 4.665963431786217e-05, "loss": 2.0936, "step": 12550 }, { "epoch": 0.68, "learning_rate": 4.665692956832198e-05, "loss": 2.0361, "step": 12560 }, { "epoch": 0.68, "learning_rate": 4.6654224818781787e-05, "loss": 2.0187, "step": 12570 }, { "epoch": 0.68, "learning_rate": 4.665152006924159e-05, "loss": 1.9209, "step": 12580 }, { "epoch": 0.68, "learning_rate": 4.66488153197014e-05, "loss": 1.9916, "step": 12590 }, { "epoch": 0.68, "learning_rate": 4.664611057016121e-05, "loss": 1.93, "step": 12600 }, { "epoch": 0.68, "learning_rate": 4.664340582062101e-05, "loss": 1.9744, "step": 12610 }, { "epoch": 0.68, "learning_rate": 4.664070107108082e-05, "loss": 2.0329, "step": 12620 }, { "epoch": 0.68, "learning_rate": 4.663799632154063e-05, "loss": 2.0184, "step": 12630 }, { "epoch": 0.68, "learning_rate": 4.6635291572000434e-05, "loss": 2.0137, "step": 12640 }, { "epoch": 0.68, "learning_rate": 4.663258682246024e-05, "loss": 2.1003, "step": 12650 }, { "epoch": 0.68, "learning_rate": 4.662988207292005e-05, "loss": 2.0014, "step": 12660 }, { "epoch": 0.68, "learning_rate": 4.6627177323379856e-05, "loss": 2.0161, "step": 12670 }, { "epoch": 0.69, "learning_rate": 4.6624472573839666e-05, "loss": 1.932, "step": 12680 }, { "epoch": 0.69, "learning_rate": 4.6621767824299475e-05, "loss": 2.0527, "step": 12690 }, { "epoch": 0.69, "learning_rate": 4.661906307475928e-05, "loss": 2.0175, "step": 12700 }, { "epoch": 0.69, "learning_rate": 4.661635832521909e-05, "loss": 1.9941, "step": 12710 }, { "epoch": 0.69, "learning_rate": 4.66136535756789e-05, "loss": 1.9945, "step": 12720 }, { "epoch": 0.69, "learning_rate": 4.66109488261387e-05, "loss": 2.0861, "step": 12730 }, { "epoch": 0.69, "learning_rate": 4.660824407659851e-05, "loss": 1.9694, "step": 12740 }, { "epoch": 0.69, "learning_rate": 4.660553932705832e-05, "loss": 1.9925, "step": 12750 }, { "epoch": 0.69, "learning_rate": 4.660283457751812e-05, "loss": 2.0725, "step": 12760 }, { "epoch": 0.69, "learning_rate": 4.660012982797793e-05, "loss": 2.0315, "step": 12770 }, { "epoch": 0.69, "learning_rate": 4.659742507843774e-05, "loss": 1.9864, "step": 12780 }, { "epoch": 0.69, "learning_rate": 4.6594720328897545e-05, "loss": 2.0197, "step": 12790 }, { "epoch": 0.69, "learning_rate": 4.6592015579357354e-05, "loss": 1.9359, "step": 12800 }, { "epoch": 0.69, "learning_rate": 4.6589310829817164e-05, "loss": 2.0037, "step": 12810 }, { "epoch": 0.69, "learning_rate": 4.658660608027697e-05, "loss": 2.0273, "step": 12820 }, { "epoch": 0.69, "learning_rate": 4.6583901330736776e-05, "loss": 1.9948, "step": 12830 }, { "epoch": 0.69, "learning_rate": 4.6581196581196586e-05, "loss": 1.9808, "step": 12840 }, { "epoch": 0.69, "learning_rate": 4.657849183165639e-05, "loss": 1.9857, "step": 12850 }, { "epoch": 0.69, "learning_rate": 4.65757870821162e-05, "loss": 2.0068, "step": 12860 }, { "epoch": 0.7, "learning_rate": 4.657308233257601e-05, "loss": 1.9807, "step": 12870 }, { "epoch": 0.7, "learning_rate": 4.657037758303581e-05, "loss": 2.0137, "step": 12880 }, { "epoch": 0.7, "learning_rate": 4.656767283349562e-05, "loss": 2.063, "step": 12890 }, { "epoch": 0.7, "learning_rate": 4.656496808395543e-05, "loss": 1.9156, "step": 12900 }, { "epoch": 0.7, "learning_rate": 4.656226333441523e-05, "loss": 2.0549, "step": 12910 }, { "epoch": 0.7, "learning_rate": 4.655955858487504e-05, "loss": 1.9536, "step": 12920 }, { "epoch": 0.7, "learning_rate": 4.655685383533485e-05, "loss": 1.9401, "step": 12930 }, { "epoch": 0.7, "learning_rate": 4.6554149085794655e-05, "loss": 1.9928, "step": 12940 }, { "epoch": 0.7, "learning_rate": 4.6551444336254465e-05, "loss": 1.9421, "step": 12950 }, { "epoch": 0.7, "learning_rate": 4.6548739586714275e-05, "loss": 1.9839, "step": 12960 }, { "epoch": 0.7, "learning_rate": 4.654603483717408e-05, "loss": 1.9792, "step": 12970 }, { "epoch": 0.7, "learning_rate": 4.654333008763389e-05, "loss": 1.9388, "step": 12980 }, { "epoch": 0.7, "learning_rate": 4.65406253380937e-05, "loss": 1.9999, "step": 12990 }, { "epoch": 0.7, "learning_rate": 4.65379205885535e-05, "loss": 1.9087, "step": 13000 }, { "epoch": 0.7, "learning_rate": 4.653521583901331e-05, "loss": 2.0134, "step": 13010 }, { "epoch": 0.7, "learning_rate": 4.653251108947312e-05, "loss": 1.9831, "step": 13020 }, { "epoch": 0.7, "learning_rate": 4.652980633993292e-05, "loss": 1.9945, "step": 13030 }, { "epoch": 0.7, "learning_rate": 4.652710159039273e-05, "loss": 1.9451, "step": 13040 }, { "epoch": 0.71, "learning_rate": 4.652439684085254e-05, "loss": 2.0707, "step": 13050 }, { "epoch": 0.71, "learning_rate": 4.6521692091312344e-05, "loss": 2.0008, "step": 13060 }, { "epoch": 0.71, "learning_rate": 4.6518987341772154e-05, "loss": 2.0751, "step": 13070 }, { "epoch": 0.71, "learning_rate": 4.6516282592231963e-05, "loss": 1.9867, "step": 13080 }, { "epoch": 0.71, "learning_rate": 4.6513577842691766e-05, "loss": 2.0105, "step": 13090 }, { "epoch": 0.71, "learning_rate": 4.6510873093151576e-05, "loss": 1.9785, "step": 13100 }, { "epoch": 0.71, "learning_rate": 4.6508168343611386e-05, "loss": 1.9946, "step": 13110 }, { "epoch": 0.71, "learning_rate": 4.650546359407119e-05, "loss": 1.9838, "step": 13120 }, { "epoch": 0.71, "learning_rate": 4.6502758844531e-05, "loss": 2.0448, "step": 13130 }, { "epoch": 0.71, "learning_rate": 4.650005409499081e-05, "loss": 1.9416, "step": 13140 }, { "epoch": 0.71, "learning_rate": 4.649734934545061e-05, "loss": 1.9854, "step": 13150 }, { "epoch": 0.71, "learning_rate": 4.649464459591042e-05, "loss": 2.0025, "step": 13160 }, { "epoch": 0.71, "learning_rate": 4.649193984637023e-05, "loss": 2.0349, "step": 13170 }, { "epoch": 0.71, "learning_rate": 4.648923509683003e-05, "loss": 1.9372, "step": 13180 }, { "epoch": 0.71, "learning_rate": 4.648653034728984e-05, "loss": 1.985, "step": 13190 }, { "epoch": 0.71, "learning_rate": 4.648382559774965e-05, "loss": 1.9315, "step": 13200 }, { "epoch": 0.71, "learning_rate": 4.648112084820946e-05, "loss": 2.0433, "step": 13210 }, { "epoch": 0.71, "learning_rate": 4.6478416098669265e-05, "loss": 1.9636, "step": 13220 }, { "epoch": 0.71, "learning_rate": 4.6475711349129074e-05, "loss": 1.9657, "step": 13230 }, { "epoch": 0.72, "learning_rate": 4.6473006599588884e-05, "loss": 1.9984, "step": 13240 }, { "epoch": 0.72, "learning_rate": 4.647030185004869e-05, "loss": 1.9208, "step": 13250 }, { "epoch": 0.72, "learning_rate": 4.6467597100508496e-05, "loss": 1.9468, "step": 13260 }, { "epoch": 0.72, "learning_rate": 4.6464892350968306e-05, "loss": 1.8933, "step": 13270 }, { "epoch": 0.72, "learning_rate": 4.646218760142811e-05, "loss": 2.0033, "step": 13280 }, { "epoch": 0.72, "learning_rate": 4.645948285188792e-05, "loss": 2.0107, "step": 13290 }, { "epoch": 0.72, "learning_rate": 4.645677810234773e-05, "loss": 2.0662, "step": 13300 }, { "epoch": 0.72, "learning_rate": 4.645407335280753e-05, "loss": 1.9396, "step": 13310 }, { "epoch": 0.72, "learning_rate": 4.645136860326734e-05, "loss": 1.9533, "step": 13320 }, { "epoch": 0.72, "learning_rate": 4.644866385372715e-05, "loss": 1.9657, "step": 13330 }, { "epoch": 0.72, "learning_rate": 4.644595910418695e-05, "loss": 1.9927, "step": 13340 }, { "epoch": 0.72, "learning_rate": 4.644325435464676e-05, "loss": 1.9316, "step": 13350 }, { "epoch": 0.72, "learning_rate": 4.644054960510657e-05, "loss": 2.0491, "step": 13360 }, { "epoch": 0.72, "learning_rate": 4.6437844855566375e-05, "loss": 1.9046, "step": 13370 }, { "epoch": 0.72, "learning_rate": 4.643514010602618e-05, "loss": 2.0203, "step": 13380 }, { "epoch": 0.72, "learning_rate": 4.6432435356485995e-05, "loss": 1.954, "step": 13390 }, { "epoch": 0.72, "learning_rate": 4.64297306069458e-05, "loss": 2.0286, "step": 13400 }, { "epoch": 0.72, "learning_rate": 4.64270258574056e-05, "loss": 2.0113, "step": 13410 }, { "epoch": 0.73, "learning_rate": 4.642432110786542e-05, "loss": 1.9783, "step": 13420 }, { "epoch": 0.73, "learning_rate": 4.642161635832522e-05, "loss": 1.9691, "step": 13430 }, { "epoch": 0.73, "learning_rate": 4.641891160878503e-05, "loss": 1.91, "step": 13440 }, { "epoch": 0.73, "learning_rate": 4.641620685924484e-05, "loss": 2.005, "step": 13450 }, { "epoch": 0.73, "learning_rate": 4.641350210970464e-05, "loss": 1.9939, "step": 13460 }, { "epoch": 0.73, "learning_rate": 4.641079736016445e-05, "loss": 2.0328, "step": 13470 }, { "epoch": 0.73, "learning_rate": 4.640809261062426e-05, "loss": 2.0387, "step": 13480 }, { "epoch": 0.73, "learning_rate": 4.6405387861084064e-05, "loss": 1.9268, "step": 13490 }, { "epoch": 0.73, "learning_rate": 4.6402683111543874e-05, "loss": 2.0077, "step": 13500 }, { "epoch": 0.73, "learning_rate": 4.6399978362003683e-05, "loss": 1.8793, "step": 13510 }, { "epoch": 0.73, "learning_rate": 4.6397273612463486e-05, "loss": 1.9635, "step": 13520 }, { "epoch": 0.73, "learning_rate": 4.6394568862923296e-05, "loss": 1.9268, "step": 13530 }, { "epoch": 0.73, "learning_rate": 4.6391864113383106e-05, "loss": 1.9829, "step": 13540 }, { "epoch": 0.73, "learning_rate": 4.638915936384291e-05, "loss": 2.0173, "step": 13550 }, { "epoch": 0.73, "learning_rate": 4.638645461430272e-05, "loss": 2.0108, "step": 13560 }, { "epoch": 0.73, "learning_rate": 4.638374986476253e-05, "loss": 2.0771, "step": 13570 }, { "epoch": 0.73, "learning_rate": 4.638104511522233e-05, "loss": 2.0418, "step": 13580 }, { "epoch": 0.73, "learning_rate": 4.637834036568214e-05, "loss": 1.9477, "step": 13590 }, { "epoch": 0.73, "learning_rate": 4.637563561614195e-05, "loss": 2.0991, "step": 13600 }, { "epoch": 0.74, "learning_rate": 4.637293086660175e-05, "loss": 1.8851, "step": 13610 }, { "epoch": 0.74, "learning_rate": 4.637022611706156e-05, "loss": 2.0056, "step": 13620 }, { "epoch": 0.74, "learning_rate": 4.636752136752137e-05, "loss": 1.953, "step": 13630 }, { "epoch": 0.74, "learning_rate": 4.6364816617981175e-05, "loss": 2.0294, "step": 13640 }, { "epoch": 0.74, "learning_rate": 4.6362111868440985e-05, "loss": 1.9925, "step": 13650 }, { "epoch": 0.74, "learning_rate": 4.6359407118900794e-05, "loss": 1.9726, "step": 13660 }, { "epoch": 0.74, "learning_rate": 4.63567023693606e-05, "loss": 1.9474, "step": 13670 }, { "epoch": 0.74, "learning_rate": 4.635399761982041e-05, "loss": 1.9254, "step": 13680 }, { "epoch": 0.74, "learning_rate": 4.6351292870280216e-05, "loss": 1.9757, "step": 13690 }, { "epoch": 0.74, "learning_rate": 4.634858812074002e-05, "loss": 2.0156, "step": 13700 }, { "epoch": 0.74, "learning_rate": 4.634588337119983e-05, "loss": 1.9882, "step": 13710 }, { "epoch": 0.74, "learning_rate": 4.634317862165964e-05, "loss": 1.9934, "step": 13720 }, { "epoch": 0.74, "learning_rate": 4.634047387211944e-05, "loss": 2.0715, "step": 13730 }, { "epoch": 0.74, "learning_rate": 4.633776912257925e-05, "loss": 2.0108, "step": 13740 }, { "epoch": 0.74, "learning_rate": 4.633506437303906e-05, "loss": 1.9933, "step": 13750 }, { "epoch": 0.74, "learning_rate": 4.6332359623498864e-05, "loss": 1.9106, "step": 13760 }, { "epoch": 0.74, "learning_rate": 4.632965487395867e-05, "loss": 1.9795, "step": 13770 }, { "epoch": 0.74, "learning_rate": 4.632695012441848e-05, "loss": 1.9506, "step": 13780 }, { "epoch": 0.75, "learning_rate": 4.6324245374878286e-05, "loss": 1.9323, "step": 13790 }, { "epoch": 0.75, "learning_rate": 4.6321540625338095e-05, "loss": 1.9183, "step": 13800 }, { "epoch": 0.75, "learning_rate": 4.6318835875797905e-05, "loss": 1.9712, "step": 13810 }, { "epoch": 0.75, "learning_rate": 4.631613112625771e-05, "loss": 2.006, "step": 13820 }, { "epoch": 0.75, "learning_rate": 4.631342637671752e-05, "loss": 1.985, "step": 13830 }, { "epoch": 0.75, "learning_rate": 4.631072162717733e-05, "loss": 1.9784, "step": 13840 }, { "epoch": 0.75, "learning_rate": 4.630801687763714e-05, "loss": 1.8951, "step": 13850 }, { "epoch": 0.75, "learning_rate": 4.630531212809694e-05, "loss": 2.0001, "step": 13860 }, { "epoch": 0.75, "learning_rate": 4.630260737855675e-05, "loss": 2.0276, "step": 13870 }, { "epoch": 0.75, "learning_rate": 4.629990262901656e-05, "loss": 2.0668, "step": 13880 }, { "epoch": 0.75, "learning_rate": 4.629719787947636e-05, "loss": 2.0283, "step": 13890 }, { "epoch": 0.75, "learning_rate": 4.629449312993617e-05, "loss": 1.9975, "step": 13900 }, { "epoch": 0.75, "learning_rate": 4.629178838039598e-05, "loss": 2.0354, "step": 13910 }, { "epoch": 0.75, "learning_rate": 4.6289083630855784e-05, "loss": 2.0259, "step": 13920 }, { "epoch": 0.75, "learning_rate": 4.628637888131559e-05, "loss": 1.9959, "step": 13930 }, { "epoch": 0.75, "learning_rate": 4.6283674131775403e-05, "loss": 1.9627, "step": 13940 }, { "epoch": 0.75, "learning_rate": 4.6280969382235206e-05, "loss": 1.9954, "step": 13950 }, { "epoch": 0.75, "learning_rate": 4.627826463269501e-05, "loss": 1.959, "step": 13960 }, { "epoch": 0.75, "learning_rate": 4.6275559883154826e-05, "loss": 1.9865, "step": 13970 }, { "epoch": 0.76, "learning_rate": 4.627285513361463e-05, "loss": 2.038, "step": 13980 }, { "epoch": 0.76, "learning_rate": 4.627015038407443e-05, "loss": 1.9576, "step": 13990 }, { "epoch": 0.76, "learning_rate": 4.626744563453425e-05, "loss": 1.9939, "step": 14000 }, { "epoch": 0.76, "learning_rate": 4.626474088499405e-05, "loss": 1.971, "step": 14010 }, { "epoch": 0.76, "learning_rate": 4.6262036135453854e-05, "loss": 2.0907, "step": 14020 }, { "epoch": 0.76, "learning_rate": 4.625933138591367e-05, "loss": 1.9214, "step": 14030 }, { "epoch": 0.76, "learning_rate": 4.625662663637347e-05, "loss": 1.9242, "step": 14040 }, { "epoch": 0.76, "learning_rate": 4.625392188683328e-05, "loss": 2.0305, "step": 14050 }, { "epoch": 0.76, "learning_rate": 4.625121713729309e-05, "loss": 2.042, "step": 14060 }, { "epoch": 0.76, "learning_rate": 4.6248512387752895e-05, "loss": 2.0411, "step": 14070 }, { "epoch": 0.76, "learning_rate": 4.6245807638212705e-05, "loss": 1.9385, "step": 14080 }, { "epoch": 0.76, "learning_rate": 4.6243102888672514e-05, "loss": 2.0294, "step": 14090 }, { "epoch": 0.76, "learning_rate": 4.624039813913232e-05, "loss": 1.9861, "step": 14100 }, { "epoch": 0.76, "learning_rate": 4.623769338959213e-05, "loss": 1.8696, "step": 14110 }, { "epoch": 0.76, "learning_rate": 4.6234988640051936e-05, "loss": 1.9727, "step": 14120 }, { "epoch": 0.76, "learning_rate": 4.623228389051174e-05, "loss": 1.9255, "step": 14130 }, { "epoch": 0.76, "learning_rate": 4.622957914097155e-05, "loss": 1.9833, "step": 14140 }, { "epoch": 0.76, "learning_rate": 4.622687439143136e-05, "loss": 1.9207, "step": 14150 }, { "epoch": 0.77, "learning_rate": 4.622416964189116e-05, "loss": 2.0222, "step": 14160 }, { "epoch": 0.77, "learning_rate": 4.622146489235097e-05, "loss": 1.9672, "step": 14170 }, { "epoch": 0.77, "learning_rate": 4.621876014281078e-05, "loss": 1.984, "step": 14180 }, { "epoch": 0.77, "learning_rate": 4.6216055393270584e-05, "loss": 1.942, "step": 14190 }, { "epoch": 0.77, "learning_rate": 4.621335064373039e-05, "loss": 1.8892, "step": 14200 }, { "epoch": 0.77, "learning_rate": 4.62106458941902e-05, "loss": 1.9877, "step": 14210 }, { "epoch": 0.77, "learning_rate": 4.6207941144650006e-05, "loss": 1.9639, "step": 14220 }, { "epoch": 0.77, "learning_rate": 4.6205236395109815e-05, "loss": 1.938, "step": 14230 }, { "epoch": 0.77, "learning_rate": 4.6202531645569625e-05, "loss": 1.9713, "step": 14240 }, { "epoch": 0.77, "learning_rate": 4.619982689602943e-05, "loss": 2.0544, "step": 14250 }, { "epoch": 0.77, "learning_rate": 4.619712214648924e-05, "loss": 1.8695, "step": 14260 }, { "epoch": 0.77, "learning_rate": 4.619441739694905e-05, "loss": 1.965, "step": 14270 }, { "epoch": 0.77, "learning_rate": 4.619171264740885e-05, "loss": 1.946, "step": 14280 }, { "epoch": 0.77, "learning_rate": 4.618900789786866e-05, "loss": 2.0172, "step": 14290 }, { "epoch": 0.77, "learning_rate": 4.618630314832847e-05, "loss": 1.9525, "step": 14300 }, { "epoch": 0.77, "learning_rate": 4.618359839878827e-05, "loss": 2.0299, "step": 14310 }, { "epoch": 0.77, "learning_rate": 4.618089364924808e-05, "loss": 1.9242, "step": 14320 }, { "epoch": 0.77, "learning_rate": 4.617818889970789e-05, "loss": 2.0166, "step": 14330 }, { "epoch": 0.77, "learning_rate": 4.6175484150167695e-05, "loss": 1.9954, "step": 14340 }, { "epoch": 0.78, "learning_rate": 4.6172779400627504e-05, "loss": 2.0241, "step": 14350 }, { "epoch": 0.78, "learning_rate": 4.6170074651087314e-05, "loss": 2.0001, "step": 14360 }, { "epoch": 0.78, "learning_rate": 4.616736990154712e-05, "loss": 1.9957, "step": 14370 }, { "epoch": 0.78, "learning_rate": 4.6164665152006926e-05, "loss": 1.9291, "step": 14380 }, { "epoch": 0.78, "learning_rate": 4.6161960402466736e-05, "loss": 1.9161, "step": 14390 }, { "epoch": 0.78, "learning_rate": 4.615925565292654e-05, "loss": 1.9766, "step": 14400 }, { "epoch": 0.78, "learning_rate": 4.615655090338635e-05, "loss": 1.9283, "step": 14410 }, { "epoch": 0.78, "learning_rate": 4.615384615384616e-05, "loss": 1.9828, "step": 14420 }, { "epoch": 0.78, "learning_rate": 4.615114140430596e-05, "loss": 1.9461, "step": 14430 }, { "epoch": 0.78, "learning_rate": 4.614843665476577e-05, "loss": 1.9625, "step": 14440 }, { "epoch": 0.78, "learning_rate": 4.614573190522558e-05, "loss": 2.0829, "step": 14450 }, { "epoch": 0.78, "learning_rate": 4.614302715568539e-05, "loss": 1.998, "step": 14460 }, { "epoch": 0.78, "learning_rate": 4.614032240614519e-05, "loss": 1.9971, "step": 14470 }, { "epoch": 0.78, "learning_rate": 4.6137617656605e-05, "loss": 1.9338, "step": 14480 }, { "epoch": 0.78, "learning_rate": 4.613491290706481e-05, "loss": 1.9582, "step": 14490 }, { "epoch": 0.78, "learning_rate": 4.6132208157524615e-05, "loss": 1.9579, "step": 14500 }, { "epoch": 0.78, "learning_rate": 4.612950340798442e-05, "loss": 2.0072, "step": 14510 }, { "epoch": 0.78, "learning_rate": 4.6126798658444234e-05, "loss": 2.0091, "step": 14520 }, { "epoch": 0.79, "learning_rate": 4.612409390890404e-05, "loss": 2.0684, "step": 14530 }, { "epoch": 0.79, "learning_rate": 4.612138915936384e-05, "loss": 2.0149, "step": 14540 }, { "epoch": 0.79, "learning_rate": 4.6118684409823656e-05, "loss": 1.9658, "step": 14550 }, { "epoch": 0.79, "learning_rate": 4.611597966028346e-05, "loss": 2.0085, "step": 14560 }, { "epoch": 0.79, "learning_rate": 4.611327491074326e-05, "loss": 2.0511, "step": 14570 }, { "epoch": 0.79, "learning_rate": 4.611057016120308e-05, "loss": 1.9818, "step": 14580 }, { "epoch": 0.79, "learning_rate": 4.610786541166288e-05, "loss": 2.0129, "step": 14590 }, { "epoch": 0.79, "learning_rate": 4.6105160662122684e-05, "loss": 1.9404, "step": 14600 }, { "epoch": 0.79, "learning_rate": 4.61024559125825e-05, "loss": 1.9482, "step": 14610 }, { "epoch": 0.79, "learning_rate": 4.6099751163042304e-05, "loss": 1.9652, "step": 14620 }, { "epoch": 0.79, "learning_rate": 4.6097046413502107e-05, "loss": 1.9984, "step": 14630 }, { "epoch": 0.79, "learning_rate": 4.609434166396192e-05, "loss": 2.0287, "step": 14640 }, { "epoch": 0.79, "learning_rate": 4.6091636914421726e-05, "loss": 1.9965, "step": 14650 }, { "epoch": 0.79, "learning_rate": 4.608893216488153e-05, "loss": 1.895, "step": 14660 }, { "epoch": 0.79, "learning_rate": 4.6086227415341345e-05, "loss": 1.95, "step": 14670 }, { "epoch": 0.79, "learning_rate": 4.608352266580115e-05, "loss": 1.9567, "step": 14680 }, { "epoch": 0.79, "learning_rate": 4.608081791626096e-05, "loss": 1.8837, "step": 14690 }, { "epoch": 0.79, "learning_rate": 4.607811316672077e-05, "loss": 1.9583, "step": 14700 }, { "epoch": 0.79, "learning_rate": 4.607540841718057e-05, "loss": 1.9406, "step": 14710 }, { "epoch": 0.8, "learning_rate": 4.607270366764038e-05, "loss": 2.038, "step": 14720 }, { "epoch": 0.8, "learning_rate": 4.606999891810019e-05, "loss": 1.9262, "step": 14730 }, { "epoch": 0.8, "learning_rate": 4.606729416855999e-05, "loss": 2.0263, "step": 14740 }, { "epoch": 0.8, "learning_rate": 4.60645894190198e-05, "loss": 1.9232, "step": 14750 }, { "epoch": 0.8, "learning_rate": 4.606188466947961e-05, "loss": 1.9212, "step": 14760 }, { "epoch": 0.8, "learning_rate": 4.6059179919939415e-05, "loss": 1.9893, "step": 14770 }, { "epoch": 0.8, "learning_rate": 4.6056475170399224e-05, "loss": 1.9457, "step": 14780 }, { "epoch": 0.8, "learning_rate": 4.6053770420859034e-05, "loss": 1.8671, "step": 14790 }, { "epoch": 0.8, "learning_rate": 4.605106567131884e-05, "loss": 1.9841, "step": 14800 }, { "epoch": 0.8, "learning_rate": 4.6048360921778646e-05, "loss": 2.0175, "step": 14810 }, { "epoch": 0.8, "learning_rate": 4.6045656172238456e-05, "loss": 2.004, "step": 14820 }, { "epoch": 0.8, "learning_rate": 4.604295142269826e-05, "loss": 1.9894, "step": 14830 }, { "epoch": 0.8, "learning_rate": 4.604024667315807e-05, "loss": 1.9739, "step": 14840 }, { "epoch": 0.8, "learning_rate": 4.603754192361788e-05, "loss": 1.9203, "step": 14850 }, { "epoch": 0.8, "learning_rate": 4.603483717407768e-05, "loss": 1.9077, "step": 14860 }, { "epoch": 0.8, "learning_rate": 4.603213242453749e-05, "loss": 2.0439, "step": 14870 }, { "epoch": 0.8, "learning_rate": 4.60294276749973e-05, "loss": 1.9536, "step": 14880 }, { "epoch": 0.8, "learning_rate": 4.60267229254571e-05, "loss": 1.9689, "step": 14890 }, { "epoch": 0.81, "learning_rate": 4.602401817591691e-05, "loss": 1.9704, "step": 14900 }, { "epoch": 0.81, "learning_rate": 4.602131342637672e-05, "loss": 1.9273, "step": 14910 }, { "epoch": 0.81, "learning_rate": 4.6018608676836525e-05, "loss": 2.0787, "step": 14920 }, { "epoch": 0.81, "learning_rate": 4.6015903927296335e-05, "loss": 2.0135, "step": 14930 }, { "epoch": 0.81, "learning_rate": 4.6013199177756145e-05, "loss": 1.9239, "step": 14940 }, { "epoch": 0.81, "learning_rate": 4.601049442821595e-05, "loss": 1.9523, "step": 14950 }, { "epoch": 0.81, "learning_rate": 4.600778967867576e-05, "loss": 1.9749, "step": 14960 }, { "epoch": 0.81, "learning_rate": 4.600508492913557e-05, "loss": 1.9978, "step": 14970 }, { "epoch": 0.81, "learning_rate": 4.600238017959537e-05, "loss": 2.0273, "step": 14980 }, { "epoch": 0.81, "learning_rate": 4.599967543005518e-05, "loss": 1.9603, "step": 14990 }, { "epoch": 0.81, "learning_rate": 4.599697068051499e-05, "loss": 2.0403, "step": 15000 }, { "epoch": 0.81, "learning_rate": 4.599426593097479e-05, "loss": 1.9454, "step": 15010 }, { "epoch": 0.81, "learning_rate": 4.59915611814346e-05, "loss": 1.9146, "step": 15020 }, { "epoch": 0.81, "learning_rate": 4.598885643189441e-05, "loss": 1.9015, "step": 15030 }, { "epoch": 0.81, "learning_rate": 4.5986151682354214e-05, "loss": 1.9455, "step": 15040 }, { "epoch": 0.81, "learning_rate": 4.5983446932814024e-05, "loss": 2.0138, "step": 15050 }, { "epoch": 0.81, "learning_rate": 4.5980742183273827e-05, "loss": 1.956, "step": 15060 }, { "epoch": 0.81, "learning_rate": 4.5978037433733636e-05, "loss": 1.9209, "step": 15070 }, { "epoch": 0.81, "learning_rate": 4.5975332684193446e-05, "loss": 1.9573, "step": 15080 }, { "epoch": 0.82, "learning_rate": 4.597262793465325e-05, "loss": 1.9583, "step": 15090 }, { "epoch": 0.82, "learning_rate": 4.5969923185113065e-05, "loss": 2.0081, "step": 15100 }, { "epoch": 0.82, "learning_rate": 4.596721843557287e-05, "loss": 1.8874, "step": 15110 }, { "epoch": 0.82, "learning_rate": 4.596451368603267e-05, "loss": 1.9207, "step": 15120 }, { "epoch": 0.82, "learning_rate": 4.596180893649249e-05, "loss": 1.9339, "step": 15130 }, { "epoch": 0.82, "learning_rate": 4.595910418695229e-05, "loss": 1.8885, "step": 15140 }, { "epoch": 0.82, "learning_rate": 4.595639943741209e-05, "loss": 1.9743, "step": 15150 }, { "epoch": 0.82, "learning_rate": 4.595369468787191e-05, "loss": 1.936, "step": 15160 }, { "epoch": 0.82, "learning_rate": 4.595098993833171e-05, "loss": 1.9671, "step": 15170 }, { "epoch": 0.82, "learning_rate": 4.5948285188791515e-05, "loss": 1.9623, "step": 15180 }, { "epoch": 0.82, "learning_rate": 4.594558043925133e-05, "loss": 2.016, "step": 15190 }, { "epoch": 0.82, "learning_rate": 4.5942875689711135e-05, "loss": 1.8852, "step": 15200 }, { "epoch": 0.82, "learning_rate": 4.594017094017094e-05, "loss": 2.047, "step": 15210 }, { "epoch": 0.82, "learning_rate": 4.5937466190630754e-05, "loss": 1.9926, "step": 15220 }, { "epoch": 0.82, "learning_rate": 4.593476144109056e-05, "loss": 2.0672, "step": 15230 }, { "epoch": 0.82, "learning_rate": 4.593205669155036e-05, "loss": 1.9362, "step": 15240 }, { "epoch": 0.82, "learning_rate": 4.5929351942010176e-05, "loss": 1.9971, "step": 15250 }, { "epoch": 0.82, "learning_rate": 4.592664719246998e-05, "loss": 2.0239, "step": 15260 }, { "epoch": 0.83, "learning_rate": 4.592394244292978e-05, "loss": 2.014, "step": 15270 }, { "epoch": 0.83, "learning_rate": 4.59212376933896e-05, "loss": 1.8622, "step": 15280 }, { "epoch": 0.83, "learning_rate": 4.59185329438494e-05, "loss": 1.9031, "step": 15290 }, { "epoch": 0.83, "learning_rate": 4.5915828194309204e-05, "loss": 2.0207, "step": 15300 }, { "epoch": 0.83, "learning_rate": 4.591312344476902e-05, "loss": 2.0296, "step": 15310 }, { "epoch": 0.83, "learning_rate": 4.591041869522882e-05, "loss": 1.9226, "step": 15320 }, { "epoch": 0.83, "learning_rate": 4.590771394568863e-05, "loss": 1.9077, "step": 15330 }, { "epoch": 0.83, "learning_rate": 4.590500919614844e-05, "loss": 1.9602, "step": 15340 }, { "epoch": 0.83, "learning_rate": 4.5902304446608245e-05, "loss": 2.0039, "step": 15350 }, { "epoch": 0.83, "learning_rate": 4.5899599697068055e-05, "loss": 1.8976, "step": 15360 }, { "epoch": 0.83, "learning_rate": 4.5896894947527865e-05, "loss": 1.9512, "step": 15370 }, { "epoch": 0.83, "learning_rate": 4.589419019798767e-05, "loss": 1.9634, "step": 15380 }, { "epoch": 0.83, "learning_rate": 4.589148544844748e-05, "loss": 1.9268, "step": 15390 }, { "epoch": 0.83, "learning_rate": 4.588878069890729e-05, "loss": 1.8819, "step": 15400 }, { "epoch": 0.83, "learning_rate": 4.588607594936709e-05, "loss": 1.9413, "step": 15410 }, { "epoch": 0.83, "learning_rate": 4.58833711998269e-05, "loss": 1.9294, "step": 15420 }, { "epoch": 0.83, "learning_rate": 4.588066645028671e-05, "loss": 2.0603, "step": 15430 }, { "epoch": 0.83, "learning_rate": 4.587796170074651e-05, "loss": 2.0302, "step": 15440 }, { "epoch": 0.83, "learning_rate": 4.587525695120632e-05, "loss": 1.9176, "step": 15450 }, { "epoch": 0.84, "learning_rate": 4.587255220166613e-05, "loss": 1.9653, "step": 15460 }, { "epoch": 0.84, "learning_rate": 4.5869847452125934e-05, "loss": 2.0198, "step": 15470 }, { "epoch": 0.84, "learning_rate": 4.5867142702585744e-05, "loss": 1.9094, "step": 15480 }, { "epoch": 0.84, "learning_rate": 4.586443795304555e-05, "loss": 1.9816, "step": 15490 }, { "epoch": 0.84, "learning_rate": 4.5861733203505356e-05, "loss": 2.1126, "step": 15500 }, { "epoch": 0.84, "learning_rate": 4.5859028453965166e-05, "loss": 1.8255, "step": 15510 }, { "epoch": 0.84, "learning_rate": 4.5856323704424976e-05, "loss": 1.9623, "step": 15520 }, { "epoch": 0.84, "learning_rate": 4.585361895488478e-05, "loss": 2.0066, "step": 15530 }, { "epoch": 0.84, "learning_rate": 4.585091420534459e-05, "loss": 1.9528, "step": 15540 }, { "epoch": 0.84, "learning_rate": 4.58482094558044e-05, "loss": 2.0029, "step": 15550 }, { "epoch": 0.84, "learning_rate": 4.58455047062642e-05, "loss": 1.9332, "step": 15560 }, { "epoch": 0.84, "learning_rate": 4.584279995672401e-05, "loss": 1.9532, "step": 15570 }, { "epoch": 0.84, "learning_rate": 4.584009520718382e-05, "loss": 1.9594, "step": 15580 }, { "epoch": 0.84, "learning_rate": 4.583739045764362e-05, "loss": 1.9538, "step": 15590 }, { "epoch": 0.84, "learning_rate": 4.583468570810343e-05, "loss": 1.9802, "step": 15600 }, { "epoch": 0.84, "learning_rate": 4.5831980958563235e-05, "loss": 2.0162, "step": 15610 }, { "epoch": 0.84, "learning_rate": 4.5829276209023045e-05, "loss": 1.9703, "step": 15620 }, { "epoch": 0.84, "learning_rate": 4.5826571459482855e-05, "loss": 2.062, "step": 15630 }, { "epoch": 0.85, "learning_rate": 4.582386670994266e-05, "loss": 2.0178, "step": 15640 }, { "epoch": 0.85, "learning_rate": 4.582116196040247e-05, "loss": 1.9599, "step": 15650 }, { "epoch": 0.85, "learning_rate": 4.581845721086228e-05, "loss": 1.9869, "step": 15660 }, { "epoch": 0.85, "learning_rate": 4.581575246132208e-05, "loss": 1.867, "step": 15670 }, { "epoch": 0.85, "learning_rate": 4.581304771178189e-05, "loss": 2.0041, "step": 15680 }, { "epoch": 0.85, "learning_rate": 4.58103429622417e-05, "loss": 1.882, "step": 15690 }, { "epoch": 0.85, "learning_rate": 4.58076382127015e-05, "loss": 1.9737, "step": 15700 }, { "epoch": 0.85, "learning_rate": 4.580493346316131e-05, "loss": 1.9893, "step": 15710 }, { "epoch": 0.85, "learning_rate": 4.580222871362112e-05, "loss": 1.9448, "step": 15720 }, { "epoch": 0.85, "learning_rate": 4.5799523964080924e-05, "loss": 1.9816, "step": 15730 }, { "epoch": 0.85, "learning_rate": 4.579681921454074e-05, "loss": 2.0193, "step": 15740 }, { "epoch": 0.85, "learning_rate": 4.579411446500054e-05, "loss": 1.9769, "step": 15750 }, { "epoch": 0.85, "learning_rate": 4.5791409715460346e-05, "loss": 2.0723, "step": 15760 }, { "epoch": 0.85, "learning_rate": 4.578870496592016e-05, "loss": 1.9558, "step": 15770 }, { "epoch": 0.85, "learning_rate": 4.5786000216379965e-05, "loss": 1.9215, "step": 15780 }, { "epoch": 0.85, "learning_rate": 4.578329546683977e-05, "loss": 2.0179, "step": 15790 }, { "epoch": 0.85, "learning_rate": 4.5780590717299585e-05, "loss": 1.9344, "step": 15800 }, { "epoch": 0.85, "learning_rate": 4.577788596775939e-05, "loss": 1.9587, "step": 15810 }, { "epoch": 0.85, "learning_rate": 4.577518121821919e-05, "loss": 1.9522, "step": 15820 }, { "epoch": 0.86, "learning_rate": 4.577247646867901e-05, "loss": 1.9557, "step": 15830 }, { "epoch": 0.86, "learning_rate": 4.576977171913881e-05, "loss": 1.9262, "step": 15840 }, { "epoch": 0.86, "learning_rate": 4.576706696959861e-05, "loss": 1.9752, "step": 15850 }, { "epoch": 0.86, "learning_rate": 4.576436222005843e-05, "loss": 1.9993, "step": 15860 }, { "epoch": 0.86, "learning_rate": 4.576165747051823e-05, "loss": 1.9407, "step": 15870 }, { "epoch": 0.86, "learning_rate": 4.5758952720978035e-05, "loss": 1.8329, "step": 15880 }, { "epoch": 0.86, "learning_rate": 4.575624797143785e-05, "loss": 1.9875, "step": 15890 }, { "epoch": 0.86, "learning_rate": 4.5753543221897654e-05, "loss": 1.9179, "step": 15900 }, { "epoch": 0.86, "learning_rate": 4.575083847235746e-05, "loss": 1.8991, "step": 15910 }, { "epoch": 0.86, "learning_rate": 4.5748133722817273e-05, "loss": 1.9234, "step": 15920 }, { "epoch": 0.86, "learning_rate": 4.5745428973277076e-05, "loss": 1.9105, "step": 15930 }, { "epoch": 0.86, "learning_rate": 4.574272422373688e-05, "loss": 1.9439, "step": 15940 }, { "epoch": 0.86, "learning_rate": 4.5740019474196696e-05, "loss": 1.8994, "step": 15950 }, { "epoch": 0.86, "learning_rate": 4.57373147246565e-05, "loss": 1.93, "step": 15960 }, { "epoch": 0.86, "learning_rate": 4.573460997511631e-05, "loss": 2.0237, "step": 15970 }, { "epoch": 0.86, "learning_rate": 4.573190522557612e-05, "loss": 1.8954, "step": 15980 }, { "epoch": 0.86, "learning_rate": 4.572920047603592e-05, "loss": 2.0557, "step": 15990 }, { "epoch": 0.86, "learning_rate": 4.572649572649573e-05, "loss": 1.9402, "step": 16000 }, { "epoch": 0.87, "learning_rate": 4.572379097695554e-05, "loss": 1.9914, "step": 16010 }, { "epoch": 0.87, "learning_rate": 4.572108622741534e-05, "loss": 1.9892, "step": 16020 }, { "epoch": 0.87, "learning_rate": 4.571838147787515e-05, "loss": 2.0475, "step": 16030 }, { "epoch": 0.87, "learning_rate": 4.571567672833496e-05, "loss": 2.0561, "step": 16040 }, { "epoch": 0.87, "learning_rate": 4.5712971978794765e-05, "loss": 1.9266, "step": 16050 }, { "epoch": 0.87, "learning_rate": 4.5710267229254575e-05, "loss": 1.9608, "step": 16060 }, { "epoch": 0.87, "learning_rate": 4.5707562479714384e-05, "loss": 1.9319, "step": 16070 }, { "epoch": 0.87, "learning_rate": 4.570485773017419e-05, "loss": 1.9572, "step": 16080 }, { "epoch": 0.87, "learning_rate": 4.5702152980634e-05, "loss": 1.9899, "step": 16090 }, { "epoch": 0.87, "learning_rate": 4.5699448231093806e-05, "loss": 2.0091, "step": 16100 }, { "epoch": 0.87, "learning_rate": 4.569674348155361e-05, "loss": 1.9038, "step": 16110 }, { "epoch": 0.87, "learning_rate": 4.569403873201342e-05, "loss": 1.9155, "step": 16120 }, { "epoch": 0.87, "learning_rate": 4.569133398247323e-05, "loss": 1.9509, "step": 16130 }, { "epoch": 0.87, "learning_rate": 4.568862923293303e-05, "loss": 1.9307, "step": 16140 }, { "epoch": 0.87, "learning_rate": 4.568592448339284e-05, "loss": 1.9049, "step": 16150 }, { "epoch": 0.87, "learning_rate": 4.568321973385265e-05, "loss": 1.9745, "step": 16160 }, { "epoch": 0.87, "learning_rate": 4.5680514984312454e-05, "loss": 1.9082, "step": 16170 }, { "epoch": 0.87, "learning_rate": 4.567781023477226e-05, "loss": 1.9926, "step": 16180 }, { "epoch": 0.87, "learning_rate": 4.5675105485232066e-05, "loss": 1.8714, "step": 16190 }, { "epoch": 0.88, "learning_rate": 4.5672400735691876e-05, "loss": 1.9929, "step": 16200 }, { "epoch": 0.88, "learning_rate": 4.5669695986151685e-05, "loss": 1.9219, "step": 16210 }, { "epoch": 0.88, "learning_rate": 4.566699123661149e-05, "loss": 1.8967, "step": 16220 }, { "epoch": 0.88, "learning_rate": 4.56642864870713e-05, "loss": 2.0029, "step": 16230 }, { "epoch": 0.88, "learning_rate": 4.566158173753111e-05, "loss": 2.0425, "step": 16240 }, { "epoch": 0.88, "learning_rate": 4.565887698799091e-05, "loss": 1.9807, "step": 16250 }, { "epoch": 0.88, "learning_rate": 4.565617223845072e-05, "loss": 2.0048, "step": 16260 }, { "epoch": 0.88, "learning_rate": 4.565346748891053e-05, "loss": 2.0499, "step": 16270 }, { "epoch": 0.88, "learning_rate": 4.565076273937033e-05, "loss": 1.9343, "step": 16280 }, { "epoch": 0.88, "learning_rate": 4.564805798983014e-05, "loss": 2.0119, "step": 16290 }, { "epoch": 0.88, "learning_rate": 4.564535324028995e-05, "loss": 1.9637, "step": 16300 }, { "epoch": 0.88, "learning_rate": 4.5642648490749755e-05, "loss": 1.9578, "step": 16310 }, { "epoch": 0.88, "learning_rate": 4.5639943741209564e-05, "loss": 1.9957, "step": 16320 }, { "epoch": 0.88, "learning_rate": 4.5637238991669374e-05, "loss": 1.953, "step": 16330 }, { "epoch": 0.88, "learning_rate": 4.563453424212918e-05, "loss": 1.8744, "step": 16340 }, { "epoch": 0.88, "learning_rate": 4.563182949258899e-05, "loss": 1.9131, "step": 16350 }, { "epoch": 0.88, "learning_rate": 4.5629124743048796e-05, "loss": 1.9752, "step": 16360 }, { "epoch": 0.88, "learning_rate": 4.56264199935086e-05, "loss": 1.9213, "step": 16370 }, { "epoch": 0.89, "learning_rate": 4.5623715243968416e-05, "loss": 1.992, "step": 16380 }, { "epoch": 0.89, "learning_rate": 4.562101049442822e-05, "loss": 1.9559, "step": 16390 }, { "epoch": 0.89, "learning_rate": 4.561830574488802e-05, "loss": 1.8728, "step": 16400 }, { "epoch": 0.89, "learning_rate": 4.561560099534784e-05, "loss": 1.9552, "step": 16410 }, { "epoch": 0.89, "learning_rate": 4.561289624580764e-05, "loss": 1.9268, "step": 16420 }, { "epoch": 0.89, "learning_rate": 4.5610191496267443e-05, "loss": 1.9138, "step": 16430 }, { "epoch": 0.89, "learning_rate": 4.560748674672726e-05, "loss": 2.0122, "step": 16440 }, { "epoch": 0.89, "learning_rate": 4.560478199718706e-05, "loss": 1.9293, "step": 16450 }, { "epoch": 0.89, "learning_rate": 4.5602077247646866e-05, "loss": 1.9635, "step": 16460 }, { "epoch": 0.89, "learning_rate": 4.559937249810668e-05, "loss": 1.9991, "step": 16470 }, { "epoch": 0.89, "learning_rate": 4.5596667748566485e-05, "loss": 1.9563, "step": 16480 }, { "epoch": 0.89, "learning_rate": 4.559396299902629e-05, "loss": 1.8981, "step": 16490 }, { "epoch": 0.89, "learning_rate": 4.5591258249486104e-05, "loss": 1.9183, "step": 16500 }, { "epoch": 0.89, "learning_rate": 4.558855349994591e-05, "loss": 1.9677, "step": 16510 }, { "epoch": 0.89, "learning_rate": 4.558584875040571e-05, "loss": 1.9394, "step": 16520 }, { "epoch": 0.89, "learning_rate": 4.5583144000865526e-05, "loss": 1.9085, "step": 16530 }, { "epoch": 0.89, "learning_rate": 4.558043925132533e-05, "loss": 1.9772, "step": 16540 }, { "epoch": 0.89, "learning_rate": 4.557773450178513e-05, "loss": 1.9695, "step": 16550 }, { "epoch": 0.89, "learning_rate": 4.557502975224495e-05, "loss": 1.9586, "step": 16560 }, { "epoch": 0.9, "learning_rate": 4.557232500270475e-05, "loss": 1.8691, "step": 16570 }, { "epoch": 0.9, "learning_rate": 4.556962025316456e-05, "loss": 1.9153, "step": 16580 }, { "epoch": 0.9, "learning_rate": 4.556691550362437e-05, "loss": 2.0234, "step": 16590 }, { "epoch": 0.9, "learning_rate": 4.5564210754084174e-05, "loss": 1.8935, "step": 16600 }, { "epoch": 0.9, "learning_rate": 4.556150600454398e-05, "loss": 1.8853, "step": 16610 }, { "epoch": 0.9, "learning_rate": 4.555880125500379e-05, "loss": 2.0745, "step": 16620 }, { "epoch": 0.9, "learning_rate": 4.5556096505463596e-05, "loss": 1.958, "step": 16630 }, { "epoch": 0.9, "learning_rate": 4.5553391755923405e-05, "loss": 1.9039, "step": 16640 }, { "epoch": 0.9, "learning_rate": 4.5550687006383215e-05, "loss": 1.9825, "step": 16650 }, { "epoch": 0.9, "learning_rate": 4.554798225684302e-05, "loss": 2.0014, "step": 16660 }, { "epoch": 0.9, "learning_rate": 4.554527750730283e-05, "loss": 1.9644, "step": 16670 }, { "epoch": 0.9, "learning_rate": 4.554257275776264e-05, "loss": 2.0194, "step": 16680 }, { "epoch": 0.9, "learning_rate": 4.553986800822244e-05, "loss": 1.9803, "step": 16690 }, { "epoch": 0.9, "learning_rate": 4.553716325868225e-05, "loss": 1.9175, "step": 16700 }, { "epoch": 0.9, "learning_rate": 4.553445850914206e-05, "loss": 1.8915, "step": 16710 }, { "epoch": 0.9, "learning_rate": 4.553175375960186e-05, "loss": 1.9288, "step": 16720 }, { "epoch": 0.9, "learning_rate": 4.552904901006167e-05, "loss": 1.9865, "step": 16730 }, { "epoch": 0.9, "learning_rate": 4.5526344260521475e-05, "loss": 1.8859, "step": 16740 }, { "epoch": 0.91, "learning_rate": 4.5523639510981284e-05, "loss": 1.9802, "step": 16750 }, { "epoch": 0.91, "learning_rate": 4.5520934761441094e-05, "loss": 1.9387, "step": 16760 }, { "epoch": 0.91, "learning_rate": 4.55182300119009e-05, "loss": 1.9045, "step": 16770 }, { "epoch": 0.91, "learning_rate": 4.551552526236071e-05, "loss": 2.0729, "step": 16780 }, { "epoch": 0.91, "learning_rate": 4.5512820512820516e-05, "loss": 1.9151, "step": 16790 }, { "epoch": 0.91, "learning_rate": 4.551011576328032e-05, "loss": 1.9388, "step": 16800 }, { "epoch": 0.91, "learning_rate": 4.550741101374013e-05, "loss": 1.925, "step": 16810 }, { "epoch": 0.91, "learning_rate": 4.550470626419994e-05, "loss": 1.9501, "step": 16820 }, { "epoch": 0.91, "learning_rate": 4.550200151465974e-05, "loss": 1.9055, "step": 16830 }, { "epoch": 0.91, "learning_rate": 4.549929676511955e-05, "loss": 1.9414, "step": 16840 }, { "epoch": 0.91, "learning_rate": 4.549659201557936e-05, "loss": 1.9917, "step": 16850 }, { "epoch": 0.91, "learning_rate": 4.5493887266039164e-05, "loss": 1.8802, "step": 16860 }, { "epoch": 0.91, "learning_rate": 4.549118251649897e-05, "loss": 1.9904, "step": 16870 }, { "epoch": 0.91, "learning_rate": 4.548847776695878e-05, "loss": 1.9728, "step": 16880 }, { "epoch": 0.91, "learning_rate": 4.5485773017418586e-05, "loss": 1.9797, "step": 16890 }, { "epoch": 0.91, "learning_rate": 4.5483068267878395e-05, "loss": 1.9624, "step": 16900 }, { "epoch": 0.91, "learning_rate": 4.5480363518338205e-05, "loss": 1.9442, "step": 16910 }, { "epoch": 0.91, "learning_rate": 4.547765876879801e-05, "loss": 2.0015, "step": 16920 }, { "epoch": 0.91, "learning_rate": 4.547495401925782e-05, "loss": 1.9188, "step": 16930 }, { "epoch": 0.92, "learning_rate": 4.547224926971763e-05, "loss": 1.9221, "step": 16940 }, { "epoch": 0.92, "learning_rate": 4.546954452017743e-05, "loss": 1.9747, "step": 16950 }, { "epoch": 0.92, "learning_rate": 4.546683977063724e-05, "loss": 1.8407, "step": 16960 }, { "epoch": 0.92, "learning_rate": 4.546413502109705e-05, "loss": 1.9095, "step": 16970 }, { "epoch": 0.92, "learning_rate": 4.546143027155685e-05, "loss": 1.9147, "step": 16980 }, { "epoch": 0.92, "learning_rate": 4.545872552201667e-05, "loss": 2.0142, "step": 16990 }, { "epoch": 0.92, "learning_rate": 4.545602077247647e-05, "loss": 1.934, "step": 17000 }, { "epoch": 0.92, "learning_rate": 4.5453316022936274e-05, "loss": 1.9811, "step": 17010 }, { "epoch": 0.92, "learning_rate": 4.545061127339609e-05, "loss": 1.9291, "step": 17020 }, { "epoch": 0.92, "learning_rate": 4.5447906523855894e-05, "loss": 1.9469, "step": 17030 }, { "epoch": 0.92, "learning_rate": 4.5445201774315697e-05, "loss": 1.9546, "step": 17040 }, { "epoch": 0.92, "learning_rate": 4.544249702477551e-05, "loss": 1.904, "step": 17050 }, { "epoch": 0.92, "learning_rate": 4.5439792275235316e-05, "loss": 1.9801, "step": 17060 }, { "epoch": 0.92, "learning_rate": 4.543708752569512e-05, "loss": 1.9093, "step": 17070 }, { "epoch": 0.92, "learning_rate": 4.5434382776154935e-05, "loss": 1.947, "step": 17080 }, { "epoch": 0.92, "learning_rate": 4.543167802661474e-05, "loss": 1.8087, "step": 17090 }, { "epoch": 0.92, "learning_rate": 4.542897327707454e-05, "loss": 1.9737, "step": 17100 }, { "epoch": 0.92, "learning_rate": 4.542626852753436e-05, "loss": 1.8966, "step": 17110 }, { "epoch": 0.93, "learning_rate": 4.542356377799416e-05, "loss": 1.9356, "step": 17120 }, { "epoch": 0.93, "learning_rate": 4.542085902845396e-05, "loss": 1.8929, "step": 17130 }, { "epoch": 0.93, "learning_rate": 4.541815427891378e-05, "loss": 2.0565, "step": 17140 }, { "epoch": 0.93, "learning_rate": 4.541544952937358e-05, "loss": 1.9703, "step": 17150 }, { "epoch": 0.93, "learning_rate": 4.5412744779833385e-05, "loss": 1.8859, "step": 17160 }, { "epoch": 0.93, "learning_rate": 4.54100400302932e-05, "loss": 1.9173, "step": 17170 }, { "epoch": 0.93, "learning_rate": 4.5407335280753005e-05, "loss": 1.9019, "step": 17180 }, { "epoch": 0.93, "learning_rate": 4.540463053121281e-05, "loss": 1.9707, "step": 17190 }, { "epoch": 0.93, "learning_rate": 4.5401925781672624e-05, "loss": 1.9327, "step": 17200 }, { "epoch": 0.93, "learning_rate": 4.539922103213243e-05, "loss": 1.9166, "step": 17210 }, { "epoch": 0.93, "learning_rate": 4.5396516282592236e-05, "loss": 1.9528, "step": 17220 }, { "epoch": 0.93, "learning_rate": 4.5393811533052046e-05, "loss": 1.9508, "step": 17230 }, { "epoch": 0.93, "learning_rate": 4.539110678351185e-05, "loss": 1.9443, "step": 17240 }, { "epoch": 0.93, "learning_rate": 4.538840203397166e-05, "loss": 1.9555, "step": 17250 }, { "epoch": 0.93, "learning_rate": 4.538569728443147e-05, "loss": 1.9308, "step": 17260 }, { "epoch": 0.93, "learning_rate": 4.538299253489127e-05, "loss": 2.0043, "step": 17270 }, { "epoch": 0.93, "learning_rate": 4.538028778535108e-05, "loss": 1.9492, "step": 17280 }, { "epoch": 0.93, "learning_rate": 4.5377583035810884e-05, "loss": 1.8934, "step": 17290 }, { "epoch": 0.93, "learning_rate": 4.537487828627069e-05, "loss": 2.0439, "step": 17300 }, { "epoch": 0.94, "learning_rate": 4.53721735367305e-05, "loss": 1.9354, "step": 17310 }, { "epoch": 0.94, "learning_rate": 4.5369468787190306e-05, "loss": 1.9992, "step": 17320 }, { "epoch": 0.94, "learning_rate": 4.5366764037650115e-05, "loss": 1.9503, "step": 17330 }, { "epoch": 0.94, "learning_rate": 4.5364059288109925e-05, "loss": 1.9267, "step": 17340 }, { "epoch": 0.94, "learning_rate": 4.536135453856973e-05, "loss": 1.8733, "step": 17350 }, { "epoch": 0.94, "learning_rate": 4.535864978902954e-05, "loss": 1.9322, "step": 17360 }, { "epoch": 0.94, "learning_rate": 4.535594503948935e-05, "loss": 1.9552, "step": 17370 }, { "epoch": 0.94, "learning_rate": 4.535324028994915e-05, "loss": 1.9275, "step": 17380 }, { "epoch": 0.94, "learning_rate": 4.535053554040896e-05, "loss": 1.9487, "step": 17390 }, { "epoch": 0.94, "learning_rate": 4.534783079086877e-05, "loss": 1.8453, "step": 17400 }, { "epoch": 0.94, "learning_rate": 4.534512604132857e-05, "loss": 2.0054, "step": 17410 }, { "epoch": 0.94, "learning_rate": 4.534242129178838e-05, "loss": 1.9934, "step": 17420 }, { "epoch": 0.94, "learning_rate": 4.533971654224819e-05, "loss": 1.9405, "step": 17430 }, { "epoch": 0.94, "learning_rate": 4.5337011792707994e-05, "loss": 1.9796, "step": 17440 }, { "epoch": 0.94, "learning_rate": 4.5334307043167804e-05, "loss": 1.8591, "step": 17450 }, { "epoch": 0.94, "learning_rate": 4.5331602293627614e-05, "loss": 1.9473, "step": 17460 }, { "epoch": 0.94, "learning_rate": 4.5328897544087417e-05, "loss": 1.9433, "step": 17470 }, { "epoch": 0.94, "learning_rate": 4.5326192794547226e-05, "loss": 1.9578, "step": 17480 }, { "epoch": 0.95, "learning_rate": 4.5323488045007036e-05, "loss": 1.9003, "step": 17490 }, { "epoch": 0.95, "learning_rate": 4.532078329546684e-05, "loss": 2.0097, "step": 17500 }, { "epoch": 0.95, "learning_rate": 4.531807854592665e-05, "loss": 2.125, "step": 17510 }, { "epoch": 0.95, "learning_rate": 4.531537379638646e-05, "loss": 1.933, "step": 17520 }, { "epoch": 0.95, "learning_rate": 4.531266904684626e-05, "loss": 1.9138, "step": 17530 }, { "epoch": 0.95, "learning_rate": 4.530996429730607e-05, "loss": 1.9293, "step": 17540 }, { "epoch": 0.95, "learning_rate": 4.530725954776588e-05, "loss": 1.9238, "step": 17550 }, { "epoch": 0.95, "learning_rate": 4.530455479822568e-05, "loss": 1.9382, "step": 17560 }, { "epoch": 0.95, "learning_rate": 4.530185004868549e-05, "loss": 1.9525, "step": 17570 }, { "epoch": 0.95, "learning_rate": 4.52991452991453e-05, "loss": 1.9221, "step": 17580 }, { "epoch": 0.95, "learning_rate": 4.5296440549605105e-05, "loss": 1.9436, "step": 17590 }, { "epoch": 0.95, "learning_rate": 4.5293735800064915e-05, "loss": 2.0036, "step": 17600 }, { "epoch": 0.95, "learning_rate": 4.5291031050524725e-05, "loss": 1.9734, "step": 17610 }, { "epoch": 0.95, "learning_rate": 4.528832630098453e-05, "loss": 2.0382, "step": 17620 }, { "epoch": 0.95, "learning_rate": 4.5285621551444344e-05, "loss": 2.0979, "step": 17630 }, { "epoch": 0.95, "learning_rate": 4.528291680190415e-05, "loss": 1.9428, "step": 17640 }, { "epoch": 0.95, "learning_rate": 4.528021205236395e-05, "loss": 1.9264, "step": 17650 }, { "epoch": 0.95, "learning_rate": 4.5277507302823766e-05, "loss": 1.9271, "step": 17660 }, { "epoch": 0.95, "learning_rate": 4.527480255328357e-05, "loss": 1.7556, "step": 17670 }, { "epoch": 0.96, "learning_rate": 4.527209780374337e-05, "loss": 1.8956, "step": 17680 }, { "epoch": 0.96, "learning_rate": 4.526939305420319e-05, "loss": 2.0136, "step": 17690 }, { "epoch": 0.96, "learning_rate": 4.526668830466299e-05, "loss": 1.9132, "step": 17700 }, { "epoch": 0.96, "learning_rate": 4.5263983555122794e-05, "loss": 1.9814, "step": 17710 }, { "epoch": 0.96, "learning_rate": 4.526127880558261e-05, "loss": 1.9586, "step": 17720 }, { "epoch": 0.96, "learning_rate": 4.525857405604241e-05, "loss": 1.9713, "step": 17730 }, { "epoch": 0.96, "learning_rate": 4.5255869306502216e-05, "loss": 1.9245, "step": 17740 }, { "epoch": 0.96, "learning_rate": 4.525316455696203e-05, "loss": 2.0424, "step": 17750 }, { "epoch": 0.96, "learning_rate": 4.5250459807421835e-05, "loss": 1.918, "step": 17760 }, { "epoch": 0.96, "learning_rate": 4.524775505788164e-05, "loss": 1.9007, "step": 17770 }, { "epoch": 0.96, "learning_rate": 4.5245050308341455e-05, "loss": 1.9881, "step": 17780 }, { "epoch": 0.96, "learning_rate": 4.524234555880126e-05, "loss": 1.9397, "step": 17790 }, { "epoch": 0.96, "learning_rate": 4.523964080926106e-05, "loss": 1.9692, "step": 17800 }, { "epoch": 0.96, "learning_rate": 4.523693605972088e-05, "loss": 1.8722, "step": 17810 }, { "epoch": 0.96, "learning_rate": 4.523423131018068e-05, "loss": 1.9354, "step": 17820 }, { "epoch": 0.96, "learning_rate": 4.523152656064048e-05, "loss": 1.8906, "step": 17830 }, { "epoch": 0.96, "learning_rate": 4.52288218111003e-05, "loss": 1.84, "step": 17840 }, { "epoch": 0.96, "learning_rate": 4.52261170615601e-05, "loss": 1.9893, "step": 17850 }, { "epoch": 0.97, "learning_rate": 4.522341231201991e-05, "loss": 1.9059, "step": 17860 }, { "epoch": 0.97, "learning_rate": 4.5220707562479714e-05, "loss": 1.9457, "step": 17870 }, { "epoch": 0.97, "learning_rate": 4.5218002812939524e-05, "loss": 2.037, "step": 17880 }, { "epoch": 0.97, "learning_rate": 4.5215298063399334e-05, "loss": 1.9713, "step": 17890 }, { "epoch": 0.97, "learning_rate": 4.5212593313859137e-05, "loss": 1.8712, "step": 17900 }, { "epoch": 0.97, "learning_rate": 4.5209888564318946e-05, "loss": 1.9539, "step": 17910 }, { "epoch": 0.97, "learning_rate": 4.5207183814778756e-05, "loss": 1.9619, "step": 17920 }, { "epoch": 0.97, "learning_rate": 4.520447906523856e-05, "loss": 1.877, "step": 17930 }, { "epoch": 0.97, "learning_rate": 4.520177431569837e-05, "loss": 1.8799, "step": 17940 }, { "epoch": 0.97, "learning_rate": 4.519906956615818e-05, "loss": 1.8922, "step": 17950 }, { "epoch": 0.97, "learning_rate": 4.519636481661798e-05, "loss": 1.9545, "step": 17960 }, { "epoch": 0.97, "learning_rate": 4.519366006707779e-05, "loss": 1.9022, "step": 17970 }, { "epoch": 0.97, "learning_rate": 4.51909553175376e-05, "loss": 1.9111, "step": 17980 }, { "epoch": 0.97, "learning_rate": 4.51882505679974e-05, "loss": 1.9474, "step": 17990 }, { "epoch": 0.97, "learning_rate": 4.518554581845721e-05, "loss": 1.9274, "step": 18000 }, { "epoch": 0.97, "learning_rate": 4.518284106891702e-05, "loss": 1.9254, "step": 18010 }, { "epoch": 0.97, "learning_rate": 4.5180136319376825e-05, "loss": 1.9074, "step": 18020 }, { "epoch": 0.97, "learning_rate": 4.5177431569836635e-05, "loss": 1.9335, "step": 18030 }, { "epoch": 0.97, "learning_rate": 4.5174726820296445e-05, "loss": 1.8718, "step": 18040 }, { "epoch": 0.98, "learning_rate": 4.517202207075625e-05, "loss": 1.8774, "step": 18050 }, { "epoch": 0.98, "learning_rate": 4.516931732121606e-05, "loss": 1.9216, "step": 18060 }, { "epoch": 0.98, "learning_rate": 4.516661257167587e-05, "loss": 1.898, "step": 18070 }, { "epoch": 0.98, "learning_rate": 4.516390782213567e-05, "loss": 1.9483, "step": 18080 }, { "epoch": 0.98, "learning_rate": 4.516120307259548e-05, "loss": 1.8437, "step": 18090 }, { "epoch": 0.98, "learning_rate": 4.515849832305529e-05, "loss": 1.8732, "step": 18100 }, { "epoch": 0.98, "learning_rate": 4.515579357351509e-05, "loss": 1.9257, "step": 18110 }, { "epoch": 0.98, "learning_rate": 4.51530888239749e-05, "loss": 1.9844, "step": 18120 }, { "epoch": 0.98, "learning_rate": 4.515038407443471e-05, "loss": 1.9263, "step": 18130 }, { "epoch": 0.98, "learning_rate": 4.5147679324894514e-05, "loss": 1.9873, "step": 18140 }, { "epoch": 0.98, "learning_rate": 4.5144974575354324e-05, "loss": 1.9261, "step": 18150 }, { "epoch": 0.98, "learning_rate": 4.514226982581413e-05, "loss": 1.9864, "step": 18160 }, { "epoch": 0.98, "learning_rate": 4.5139565076273936e-05, "loss": 1.9172, "step": 18170 }, { "epoch": 0.98, "learning_rate": 4.5136860326733746e-05, "loss": 1.9402, "step": 18180 }, { "epoch": 0.98, "learning_rate": 4.5134155577193555e-05, "loss": 1.969, "step": 18190 }, { "epoch": 0.98, "learning_rate": 4.513145082765336e-05, "loss": 1.9916, "step": 18200 }, { "epoch": 0.98, "learning_rate": 4.512874607811317e-05, "loss": 1.9146, "step": 18210 }, { "epoch": 0.98, "learning_rate": 4.512604132857298e-05, "loss": 1.9191, "step": 18220 }, { "epoch": 0.99, "learning_rate": 4.512333657903278e-05, "loss": 1.9874, "step": 18230 }, { "epoch": 0.99, "learning_rate": 4.512063182949259e-05, "loss": 1.9114, "step": 18240 }, { "epoch": 0.99, "learning_rate": 4.51179270799524e-05, "loss": 1.8951, "step": 18250 }, { "epoch": 0.99, "learning_rate": 4.51152223304122e-05, "loss": 1.9616, "step": 18260 }, { "epoch": 0.99, "learning_rate": 4.511251758087202e-05, "loss": 1.9032, "step": 18270 }, { "epoch": 0.99, "learning_rate": 4.510981283133182e-05, "loss": 1.9123, "step": 18280 }, { "epoch": 0.99, "learning_rate": 4.5107108081791625e-05, "loss": 1.9637, "step": 18290 }, { "epoch": 0.99, "learning_rate": 4.510440333225144e-05, "loss": 1.965, "step": 18300 }, { "epoch": 0.99, "learning_rate": 4.5101698582711244e-05, "loss": 1.9317, "step": 18310 }, { "epoch": 0.99, "learning_rate": 4.509899383317105e-05, "loss": 1.965, "step": 18320 }, { "epoch": 0.99, "learning_rate": 4.509628908363086e-05, "loss": 1.9827, "step": 18330 }, { "epoch": 0.99, "learning_rate": 4.5093584334090666e-05, "loss": 1.9591, "step": 18340 }, { "epoch": 0.99, "learning_rate": 4.509087958455047e-05, "loss": 1.9344, "step": 18350 }, { "epoch": 0.99, "learning_rate": 4.5088174835010286e-05, "loss": 1.9081, "step": 18360 }, { "epoch": 0.99, "learning_rate": 4.508547008547009e-05, "loss": 1.9475, "step": 18370 }, { "epoch": 0.99, "learning_rate": 4.508276533592989e-05, "loss": 1.901, "step": 18380 }, { "epoch": 0.99, "learning_rate": 4.508006058638971e-05, "loss": 1.9452, "step": 18390 }, { "epoch": 0.99, "learning_rate": 4.507735583684951e-05, "loss": 1.8156, "step": 18400 }, { "epoch": 0.99, "learning_rate": 4.5074651087309313e-05, "loss": 1.9836, "step": 18410 }, { "epoch": 1.0, "learning_rate": 4.507194633776912e-05, "loss": 1.907, "step": 18420 }, { "epoch": 1.0, "learning_rate": 4.506924158822893e-05, "loss": 1.9647, "step": 18430 }, { "epoch": 1.0, "learning_rate": 4.5066536838688736e-05, "loss": 1.9975, "step": 18440 }, { "epoch": 1.0, "learning_rate": 4.5063832089148545e-05, "loss": 1.7868, "step": 18450 }, { "epoch": 1.0, "learning_rate": 4.5061127339608355e-05, "loss": 1.9532, "step": 18460 }, { "epoch": 1.0, "learning_rate": 4.505842259006816e-05, "loss": 1.9492, "step": 18470 }, { "epoch": 1.0, "learning_rate": 4.505571784052797e-05, "loss": 1.9895, "step": 18480 }, { "epoch": 1.0, "learning_rate": 4.505301309098778e-05, "loss": 1.9893, "step": 18490 }, { "epoch": 1.0, "learning_rate": 4.505030834144759e-05, "loss": 1.9241, "step": 18500 }, { "epoch": 1.0, "learning_rate": 4.504760359190739e-05, "loss": 1.8807, "step": 18510 }, { "epoch": 1.0, "learning_rate": 4.50448988423672e-05, "loss": 1.8362, "step": 18520 }, { "epoch": 1.0, "learning_rate": 4.504219409282701e-05, "loss": 1.8825, "step": 18530 }, { "epoch": 1.0, "learning_rate": 4.503948934328681e-05, "loss": 1.8112, "step": 18540 }, { "epoch": 1.0, "learning_rate": 4.503678459374662e-05, "loss": 1.8811, "step": 18550 }, { "epoch": 1.0, "learning_rate": 4.503407984420643e-05, "loss": 1.8107, "step": 18560 }, { "epoch": 1.0, "learning_rate": 4.5031375094666234e-05, "loss": 1.8488, "step": 18570 }, { "epoch": 1.0, "learning_rate": 4.5028670345126044e-05, "loss": 1.8668, "step": 18580 }, { "epoch": 1.0, "learning_rate": 4.502596559558585e-05, "loss": 1.8626, "step": 18590 }, { "epoch": 1.01, "learning_rate": 4.5023260846045656e-05, "loss": 1.8615, "step": 18600 }, { "epoch": 1.01, "learning_rate": 4.5020556096505466e-05, "loss": 1.8034, "step": 18610 }, { "epoch": 1.01, "learning_rate": 4.5017851346965275e-05, "loss": 1.7783, "step": 18620 }, { "epoch": 1.01, "learning_rate": 4.501514659742508e-05, "loss": 1.7775, "step": 18630 }, { "epoch": 1.01, "learning_rate": 4.501244184788489e-05, "loss": 1.8298, "step": 18640 }, { "epoch": 1.01, "learning_rate": 4.50097370983447e-05, "loss": 1.6898, "step": 18650 }, { "epoch": 1.01, "learning_rate": 4.50070323488045e-05, "loss": 1.7836, "step": 18660 }, { "epoch": 1.01, "learning_rate": 4.500432759926431e-05, "loss": 1.8265, "step": 18670 }, { "epoch": 1.01, "learning_rate": 4.500162284972412e-05, "loss": 1.7368, "step": 18680 }, { "epoch": 1.01, "learning_rate": 4.499891810018392e-05, "loss": 1.8255, "step": 18690 }, { "epoch": 1.01, "learning_rate": 4.499621335064373e-05, "loss": 1.8622, "step": 18700 }, { "epoch": 1.01, "learning_rate": 4.499350860110354e-05, "loss": 1.7371, "step": 18710 }, { "epoch": 1.01, "learning_rate": 4.4990803851563345e-05, "loss": 1.7246, "step": 18720 }, { "epoch": 1.01, "learning_rate": 4.4988099102023154e-05, "loss": 1.8238, "step": 18730 }, { "epoch": 1.01, "learning_rate": 4.4985394352482964e-05, "loss": 1.8102, "step": 18740 }, { "epoch": 1.01, "learning_rate": 4.498268960294277e-05, "loss": 1.8032, "step": 18750 }, { "epoch": 1.01, "learning_rate": 4.497998485340258e-05, "loss": 1.8689, "step": 18760 }, { "epoch": 1.01, "learning_rate": 4.4977280103862386e-05, "loss": 1.8427, "step": 18770 }, { "epoch": 1.01, "learning_rate": 4.497457535432219e-05, "loss": 1.8554, "step": 18780 }, { "epoch": 1.02, "learning_rate": 4.4971870604782e-05, "loss": 1.8074, "step": 18790 }, { "epoch": 1.02, "learning_rate": 4.496916585524181e-05, "loss": 1.8409, "step": 18800 }, { "epoch": 1.02, "learning_rate": 4.496646110570161e-05, "loss": 1.7885, "step": 18810 }, { "epoch": 1.02, "learning_rate": 4.496375635616142e-05, "loss": 1.8272, "step": 18820 }, { "epoch": 1.02, "learning_rate": 4.496105160662123e-05, "loss": 1.7679, "step": 18830 }, { "epoch": 1.02, "learning_rate": 4.4958346857081033e-05, "loss": 1.7831, "step": 18840 }, { "epoch": 1.02, "learning_rate": 4.495564210754084e-05, "loss": 1.8356, "step": 18850 }, { "epoch": 1.02, "learning_rate": 4.495293735800065e-05, "loss": 1.8586, "step": 18860 }, { "epoch": 1.02, "learning_rate": 4.4950232608460456e-05, "loss": 1.7788, "step": 18870 }, { "epoch": 1.02, "learning_rate": 4.4947527858920265e-05, "loss": 1.8098, "step": 18880 }, { "epoch": 1.02, "learning_rate": 4.4944823109380075e-05, "loss": 1.7887, "step": 18890 }, { "epoch": 1.02, "learning_rate": 4.494211835983988e-05, "loss": 1.7758, "step": 18900 }, { "epoch": 1.02, "learning_rate": 4.4939413610299694e-05, "loss": 1.8476, "step": 18910 }, { "epoch": 1.02, "learning_rate": 4.49367088607595e-05, "loss": 1.8146, "step": 18920 }, { "epoch": 1.02, "learning_rate": 4.49340041112193e-05, "loss": 1.766, "step": 18930 }, { "epoch": 1.02, "learning_rate": 4.4931299361679116e-05, "loss": 1.8189, "step": 18940 }, { "epoch": 1.02, "learning_rate": 4.492859461213892e-05, "loss": 1.8354, "step": 18950 }, { "epoch": 1.02, "learning_rate": 4.492588986259872e-05, "loss": 1.8498, "step": 18960 }, { "epoch": 1.03, "learning_rate": 4.492318511305853e-05, "loss": 1.8316, "step": 18970 }, { "epoch": 1.03, "learning_rate": 4.492048036351834e-05, "loss": 1.82, "step": 18980 }, { "epoch": 1.03, "learning_rate": 4.4917775613978144e-05, "loss": 1.7939, "step": 18990 }, { "epoch": 1.03, "learning_rate": 4.4915070864437954e-05, "loss": 1.827, "step": 19000 }, { "epoch": 1.03, "learning_rate": 4.4912366114897764e-05, "loss": 1.7677, "step": 19010 }, { "epoch": 1.03, "learning_rate": 4.4909661365357566e-05, "loss": 1.8206, "step": 19020 }, { "epoch": 1.03, "learning_rate": 4.4906956615817376e-05, "loss": 1.7894, "step": 19030 }, { "epoch": 1.03, "learning_rate": 4.4904251866277186e-05, "loss": 1.8486, "step": 19040 }, { "epoch": 1.03, "learning_rate": 4.490154711673699e-05, "loss": 1.8516, "step": 19050 }, { "epoch": 1.03, "learning_rate": 4.48988423671968e-05, "loss": 1.7595, "step": 19060 }, { "epoch": 1.03, "learning_rate": 4.489613761765661e-05, "loss": 1.7833, "step": 19070 }, { "epoch": 1.03, "learning_rate": 4.489343286811641e-05, "loss": 1.8995, "step": 19080 }, { "epoch": 1.03, "learning_rate": 4.489072811857622e-05, "loss": 1.8841, "step": 19090 }, { "epoch": 1.03, "learning_rate": 4.488802336903603e-05, "loss": 1.8325, "step": 19100 }, { "epoch": 1.03, "learning_rate": 4.488531861949583e-05, "loss": 1.8447, "step": 19110 }, { "epoch": 1.03, "learning_rate": 4.488261386995564e-05, "loss": 1.8005, "step": 19120 }, { "epoch": 1.03, "learning_rate": 4.487990912041545e-05, "loss": 1.9072, "step": 19130 }, { "epoch": 1.03, "learning_rate": 4.487720437087526e-05, "loss": 1.8254, "step": 19140 }, { "epoch": 1.03, "learning_rate": 4.4874499621335065e-05, "loss": 1.7406, "step": 19150 }, { "epoch": 1.04, "learning_rate": 4.4871794871794874e-05, "loss": 1.7973, "step": 19160 }, { "epoch": 1.04, "learning_rate": 4.4869090122254684e-05, "loss": 1.8561, "step": 19170 }, { "epoch": 1.04, "learning_rate": 4.486638537271449e-05, "loss": 1.8613, "step": 19180 }, { "epoch": 1.04, "learning_rate": 4.48636806231743e-05, "loss": 1.8069, "step": 19190 }, { "epoch": 1.04, "learning_rate": 4.4860975873634106e-05, "loss": 1.7901, "step": 19200 }, { "epoch": 1.04, "learning_rate": 4.485827112409391e-05, "loss": 1.8253, "step": 19210 }, { "epoch": 1.04, "learning_rate": 4.485556637455372e-05, "loss": 1.9134, "step": 19220 }, { "epoch": 1.04, "learning_rate": 4.485286162501353e-05, "loss": 1.7534, "step": 19230 }, { "epoch": 1.04, "learning_rate": 4.485015687547333e-05, "loss": 1.76, "step": 19240 }, { "epoch": 1.04, "learning_rate": 4.484745212593314e-05, "loss": 1.8106, "step": 19250 }, { "epoch": 1.04, "learning_rate": 4.484474737639295e-05, "loss": 1.8741, "step": 19260 }, { "epoch": 1.04, "learning_rate": 4.4842042626852753e-05, "loss": 1.8256, "step": 19270 }, { "epoch": 1.04, "learning_rate": 4.483933787731256e-05, "loss": 1.7701, "step": 19280 }, { "epoch": 1.04, "learning_rate": 4.483663312777237e-05, "loss": 1.7323, "step": 19290 }, { "epoch": 1.04, "learning_rate": 4.4833928378232176e-05, "loss": 1.8478, "step": 19300 }, { "epoch": 1.04, "learning_rate": 4.4831223628691985e-05, "loss": 1.7717, "step": 19310 }, { "epoch": 1.04, "learning_rate": 4.4828518879151795e-05, "loss": 1.801, "step": 19320 }, { "epoch": 1.04, "learning_rate": 4.48258141296116e-05, "loss": 1.8182, "step": 19330 }, { "epoch": 1.05, "learning_rate": 4.482310938007141e-05, "loss": 1.8834, "step": 19340 }, { "epoch": 1.05, "learning_rate": 4.482040463053122e-05, "loss": 1.8901, "step": 19350 }, { "epoch": 1.05, "learning_rate": 4.481769988099102e-05, "loss": 1.795, "step": 19360 }, { "epoch": 1.05, "learning_rate": 4.481499513145083e-05, "loss": 1.8563, "step": 19370 }, { "epoch": 1.05, "learning_rate": 4.481229038191064e-05, "loss": 1.7881, "step": 19380 }, { "epoch": 1.05, "learning_rate": 4.480958563237044e-05, "loss": 1.8373, "step": 19390 }, { "epoch": 1.05, "learning_rate": 4.480688088283025e-05, "loss": 1.7478, "step": 19400 }, { "epoch": 1.05, "learning_rate": 4.480417613329006e-05, "loss": 1.8308, "step": 19410 }, { "epoch": 1.05, "learning_rate": 4.4801471383749864e-05, "loss": 1.771, "step": 19420 }, { "epoch": 1.05, "learning_rate": 4.4798766634209674e-05, "loss": 1.7643, "step": 19430 }, { "epoch": 1.05, "learning_rate": 4.4796061884669484e-05, "loss": 1.8085, "step": 19440 }, { "epoch": 1.05, "learning_rate": 4.4793357135129287e-05, "loss": 1.8412, "step": 19450 }, { "epoch": 1.05, "learning_rate": 4.4790652385589096e-05, "loss": 1.8734, "step": 19460 }, { "epoch": 1.05, "learning_rate": 4.4787947636048906e-05, "loss": 1.7653, "step": 19470 }, { "epoch": 1.05, "learning_rate": 4.478524288650871e-05, "loss": 1.8065, "step": 19480 }, { "epoch": 1.05, "learning_rate": 4.478253813696852e-05, "loss": 1.8564, "step": 19490 }, { "epoch": 1.05, "learning_rate": 4.477983338742833e-05, "loss": 1.777, "step": 19500 }, { "epoch": 1.05, "learning_rate": 4.477712863788813e-05, "loss": 1.8227, "step": 19510 }, { "epoch": 1.05, "learning_rate": 4.477442388834794e-05, "loss": 1.8438, "step": 19520 }, { "epoch": 1.06, "learning_rate": 4.477171913880775e-05, "loss": 1.8307, "step": 19530 }, { "epoch": 1.06, "learning_rate": 4.476901438926755e-05, "loss": 1.8023, "step": 19540 }, { "epoch": 1.06, "learning_rate": 4.476630963972736e-05, "loss": 1.7896, "step": 19550 }, { "epoch": 1.06, "learning_rate": 4.476360489018717e-05, "loss": 1.8445, "step": 19560 }, { "epoch": 1.06, "learning_rate": 4.4760900140646975e-05, "loss": 1.8033, "step": 19570 }, { "epoch": 1.06, "learning_rate": 4.4758195391106785e-05, "loss": 1.8709, "step": 19580 }, { "epoch": 1.06, "learning_rate": 4.4755490641566594e-05, "loss": 1.8009, "step": 19590 }, { "epoch": 1.06, "learning_rate": 4.47527858920264e-05, "loss": 1.8566, "step": 19600 }, { "epoch": 1.06, "learning_rate": 4.475008114248621e-05, "loss": 1.8159, "step": 19610 }, { "epoch": 1.06, "learning_rate": 4.474737639294602e-05, "loss": 1.7841, "step": 19620 }, { "epoch": 1.06, "learning_rate": 4.474467164340582e-05, "loss": 1.9033, "step": 19630 }, { "epoch": 1.06, "learning_rate": 4.474196689386563e-05, "loss": 1.8588, "step": 19640 }, { "epoch": 1.06, "learning_rate": 4.473926214432544e-05, "loss": 1.842, "step": 19650 }, { "epoch": 1.06, "learning_rate": 4.473655739478524e-05, "loss": 1.8025, "step": 19660 }, { "epoch": 1.06, "learning_rate": 4.473385264524505e-05, "loss": 1.7426, "step": 19670 }, { "epoch": 1.06, "learning_rate": 4.473114789570486e-05, "loss": 1.7946, "step": 19680 }, { "epoch": 1.06, "learning_rate": 4.4728443146164664e-05, "loss": 1.9306, "step": 19690 }, { "epoch": 1.06, "learning_rate": 4.4725738396624474e-05, "loss": 1.7744, "step": 19700 }, { "epoch": 1.07, "learning_rate": 4.472303364708428e-05, "loss": 1.7998, "step": 19710 }, { "epoch": 1.07, "learning_rate": 4.4720328897544086e-05, "loss": 1.7036, "step": 19720 }, { "epoch": 1.07, "learning_rate": 4.4717624148003896e-05, "loss": 1.8696, "step": 19730 }, { "epoch": 1.07, "learning_rate": 4.4714919398463705e-05, "loss": 1.6871, "step": 19740 }, { "epoch": 1.07, "learning_rate": 4.4712214648923515e-05, "loss": 1.8052, "step": 19750 }, { "epoch": 1.07, "learning_rate": 4.470950989938332e-05, "loss": 1.7387, "step": 19760 }, { "epoch": 1.07, "learning_rate": 4.470680514984313e-05, "loss": 1.7134, "step": 19770 }, { "epoch": 1.07, "learning_rate": 4.470410040030294e-05, "loss": 1.8646, "step": 19780 }, { "epoch": 1.07, "learning_rate": 4.470139565076274e-05, "loss": 1.8342, "step": 19790 }, { "epoch": 1.07, "learning_rate": 4.469869090122255e-05, "loss": 1.8525, "step": 19800 }, { "epoch": 1.07, "learning_rate": 4.469598615168236e-05, "loss": 1.818, "step": 19810 }, { "epoch": 1.07, "learning_rate": 4.469328140214216e-05, "loss": 1.7771, "step": 19820 }, { "epoch": 1.07, "learning_rate": 4.469057665260197e-05, "loss": 1.7872, "step": 19830 }, { "epoch": 1.07, "learning_rate": 4.468787190306178e-05, "loss": 1.8435, "step": 19840 }, { "epoch": 1.07, "learning_rate": 4.4685167153521584e-05, "loss": 1.8139, "step": 19850 }, { "epoch": 1.07, "learning_rate": 4.4682462403981394e-05, "loss": 1.7552, "step": 19860 }, { "epoch": 1.07, "learning_rate": 4.4679757654441204e-05, "loss": 1.8849, "step": 19870 }, { "epoch": 1.07, "learning_rate": 4.4677052904901007e-05, "loss": 1.9557, "step": 19880 }, { "epoch": 1.07, "learning_rate": 4.4674348155360816e-05, "loss": 1.7683, "step": 19890 }, { "epoch": 1.08, "learning_rate": 4.4671643405820626e-05, "loss": 1.8423, "step": 19900 }, { "epoch": 1.08, "learning_rate": 4.466893865628043e-05, "loss": 1.7962, "step": 19910 }, { "epoch": 1.08, "learning_rate": 4.466623390674024e-05, "loss": 1.8185, "step": 19920 }, { "epoch": 1.08, "learning_rate": 4.466352915720005e-05, "loss": 1.7904, "step": 19930 }, { "epoch": 1.08, "learning_rate": 4.466082440765985e-05, "loss": 1.8293, "step": 19940 }, { "epoch": 1.08, "learning_rate": 4.465811965811966e-05, "loss": 1.7785, "step": 19950 }, { "epoch": 1.08, "learning_rate": 4.465541490857947e-05, "loss": 1.7702, "step": 19960 }, { "epoch": 1.08, "learning_rate": 4.465271015903927e-05, "loss": 1.7504, "step": 19970 }, { "epoch": 1.08, "learning_rate": 4.465000540949908e-05, "loss": 1.8643, "step": 19980 }, { "epoch": 1.08, "learning_rate": 4.464730065995889e-05, "loss": 1.851, "step": 19990 }, { "epoch": 1.08, "learning_rate": 4.4644595910418695e-05, "loss": 1.816, "step": 20000 }, { "epoch": 1.08, "learning_rate": 4.4641891160878505e-05, "loss": 1.7733, "step": 20010 }, { "epoch": 1.08, "learning_rate": 4.4639186411338315e-05, "loss": 1.8154, "step": 20020 }, { "epoch": 1.08, "learning_rate": 4.463648166179812e-05, "loss": 1.7889, "step": 20030 }, { "epoch": 1.08, "learning_rate": 4.463377691225793e-05, "loss": 1.735, "step": 20040 }, { "epoch": 1.08, "learning_rate": 4.463107216271774e-05, "loss": 1.8302, "step": 20050 }, { "epoch": 1.08, "learning_rate": 4.462836741317754e-05, "loss": 1.8284, "step": 20060 }, { "epoch": 1.08, "learning_rate": 4.462566266363735e-05, "loss": 1.7962, "step": 20070 }, { "epoch": 1.09, "learning_rate": 4.462295791409716e-05, "loss": 1.808, "step": 20080 }, { "epoch": 1.09, "learning_rate": 4.462025316455696e-05, "loss": 1.7062, "step": 20090 }, { "epoch": 1.09, "learning_rate": 4.461754841501677e-05, "loss": 1.792, "step": 20100 }, { "epoch": 1.09, "learning_rate": 4.461484366547658e-05, "loss": 1.8218, "step": 20110 }, { "epoch": 1.09, "learning_rate": 4.4612138915936384e-05, "loss": 1.8719, "step": 20120 }, { "epoch": 1.09, "learning_rate": 4.4609434166396194e-05, "loss": 1.8438, "step": 20130 }, { "epoch": 1.09, "learning_rate": 4.4606729416856e-05, "loss": 1.8462, "step": 20140 }, { "epoch": 1.09, "learning_rate": 4.4604024667315806e-05, "loss": 1.7888, "step": 20150 }, { "epoch": 1.09, "learning_rate": 4.4601319917775616e-05, "loss": 1.8133, "step": 20160 }, { "epoch": 1.09, "learning_rate": 4.4598615168235425e-05, "loss": 1.7944, "step": 20170 }, { "epoch": 1.09, "learning_rate": 4.459591041869523e-05, "loss": 1.8035, "step": 20180 }, { "epoch": 1.09, "learning_rate": 4.459320566915504e-05, "loss": 1.7607, "step": 20190 }, { "epoch": 1.09, "learning_rate": 4.459050091961485e-05, "loss": 1.8174, "step": 20200 }, { "epoch": 1.09, "learning_rate": 4.458779617007465e-05, "loss": 1.8373, "step": 20210 }, { "epoch": 1.09, "learning_rate": 4.458509142053446e-05, "loss": 1.9213, "step": 20220 }, { "epoch": 1.09, "learning_rate": 4.458238667099427e-05, "loss": 1.8046, "step": 20230 }, { "epoch": 1.09, "learning_rate": 4.457968192145407e-05, "loss": 1.8292, "step": 20240 }, { "epoch": 1.09, "learning_rate": 4.457697717191388e-05, "loss": 1.8626, "step": 20250 }, { "epoch": 1.09, "learning_rate": 4.457427242237369e-05, "loss": 1.873, "step": 20260 }, { "epoch": 1.1, "learning_rate": 4.4571567672833495e-05, "loss": 1.9173, "step": 20270 }, { "epoch": 1.1, "learning_rate": 4.4568862923293304e-05, "loss": 1.7487, "step": 20280 }, { "epoch": 1.1, "learning_rate": 4.4566158173753114e-05, "loss": 1.7455, "step": 20290 }, { "epoch": 1.1, "learning_rate": 4.456345342421292e-05, "loss": 1.8277, "step": 20300 }, { "epoch": 1.1, "learning_rate": 4.4560748674672727e-05, "loss": 1.7833, "step": 20310 }, { "epoch": 1.1, "learning_rate": 4.4558043925132536e-05, "loss": 1.8113, "step": 20320 }, { "epoch": 1.1, "learning_rate": 4.455533917559234e-05, "loss": 1.8721, "step": 20330 }, { "epoch": 1.1, "learning_rate": 4.455263442605215e-05, "loss": 1.7327, "step": 20340 }, { "epoch": 1.1, "learning_rate": 4.454992967651196e-05, "loss": 1.7722, "step": 20350 }, { "epoch": 1.1, "learning_rate": 4.454722492697176e-05, "loss": 1.8445, "step": 20360 }, { "epoch": 1.1, "learning_rate": 4.454452017743157e-05, "loss": 1.8152, "step": 20370 }, { "epoch": 1.1, "learning_rate": 4.454181542789138e-05, "loss": 1.8463, "step": 20380 }, { "epoch": 1.1, "learning_rate": 4.453911067835119e-05, "loss": 1.7962, "step": 20390 }, { "epoch": 1.1, "learning_rate": 4.453640592881099e-05, "loss": 1.8449, "step": 20400 }, { "epoch": 1.1, "learning_rate": 4.45337011792708e-05, "loss": 1.77, "step": 20410 }, { "epoch": 1.1, "learning_rate": 4.453099642973061e-05, "loss": 1.8606, "step": 20420 }, { "epoch": 1.1, "learning_rate": 4.4528291680190415e-05, "loss": 1.8065, "step": 20430 }, { "epoch": 1.1, "learning_rate": 4.4525586930650225e-05, "loss": 1.8109, "step": 20440 }, { "epoch": 1.11, "learning_rate": 4.4522882181110035e-05, "loss": 1.82, "step": 20450 }, { "epoch": 1.11, "learning_rate": 4.452017743156984e-05, "loss": 1.8235, "step": 20460 }, { "epoch": 1.11, "learning_rate": 4.451747268202965e-05, "loss": 1.7613, "step": 20470 }, { "epoch": 1.11, "learning_rate": 4.451476793248946e-05, "loss": 1.8065, "step": 20480 }, { "epoch": 1.11, "learning_rate": 4.451206318294926e-05, "loss": 1.7842, "step": 20490 }, { "epoch": 1.11, "learning_rate": 4.450935843340907e-05, "loss": 1.8105, "step": 20500 }, { "epoch": 1.11, "learning_rate": 4.450665368386888e-05, "loss": 1.8172, "step": 20510 }, { "epoch": 1.11, "learning_rate": 4.450394893432868e-05, "loss": 1.7747, "step": 20520 }, { "epoch": 1.11, "learning_rate": 4.450124418478849e-05, "loss": 1.7535, "step": 20530 }, { "epoch": 1.11, "learning_rate": 4.44985394352483e-05, "loss": 1.7801, "step": 20540 }, { "epoch": 1.11, "learning_rate": 4.4495834685708104e-05, "loss": 1.8153, "step": 20550 }, { "epoch": 1.11, "learning_rate": 4.4493129936167914e-05, "loss": 1.8409, "step": 20560 }, { "epoch": 1.11, "learning_rate": 4.449042518662772e-05, "loss": 1.8247, "step": 20570 }, { "epoch": 1.11, "learning_rate": 4.4487720437087526e-05, "loss": 1.8143, "step": 20580 }, { "epoch": 1.11, "learning_rate": 4.4485015687547336e-05, "loss": 1.7374, "step": 20590 }, { "epoch": 1.11, "learning_rate": 4.4482310938007145e-05, "loss": 1.8212, "step": 20600 }, { "epoch": 1.11, "learning_rate": 4.447960618846695e-05, "loss": 1.8139, "step": 20610 }, { "epoch": 1.11, "learning_rate": 4.447690143892676e-05, "loss": 1.8572, "step": 20620 }, { "epoch": 1.11, "learning_rate": 4.447419668938657e-05, "loss": 1.77, "step": 20630 }, { "epoch": 1.12, "learning_rate": 4.447149193984637e-05, "loss": 1.8652, "step": 20640 }, { "epoch": 1.12, "learning_rate": 4.446878719030618e-05, "loss": 1.8229, "step": 20650 }, { "epoch": 1.12, "learning_rate": 4.446608244076599e-05, "loss": 1.8229, "step": 20660 }, { "epoch": 1.12, "learning_rate": 4.446337769122579e-05, "loss": 1.7568, "step": 20670 }, { "epoch": 1.12, "learning_rate": 4.44606729416856e-05, "loss": 1.8468, "step": 20680 }, { "epoch": 1.12, "learning_rate": 4.445796819214541e-05, "loss": 1.9363, "step": 20690 }, { "epoch": 1.12, "learning_rate": 4.4455263442605215e-05, "loss": 1.798, "step": 20700 }, { "epoch": 1.12, "learning_rate": 4.4452558693065024e-05, "loss": 1.785, "step": 20710 }, { "epoch": 1.12, "learning_rate": 4.4449853943524834e-05, "loss": 1.8838, "step": 20720 }, { "epoch": 1.12, "learning_rate": 4.444714919398464e-05, "loss": 1.7865, "step": 20730 }, { "epoch": 1.12, "learning_rate": 4.4444444444444447e-05, "loss": 1.7661, "step": 20740 }, { "epoch": 1.12, "learning_rate": 4.4441739694904256e-05, "loss": 1.905, "step": 20750 }, { "epoch": 1.12, "learning_rate": 4.443903494536406e-05, "loss": 1.7792, "step": 20760 }, { "epoch": 1.12, "learning_rate": 4.443633019582387e-05, "loss": 1.7433, "step": 20770 }, { "epoch": 1.12, "learning_rate": 4.443362544628368e-05, "loss": 1.804, "step": 20780 }, { "epoch": 1.12, "learning_rate": 4.443092069674348e-05, "loss": 1.7248, "step": 20790 }, { "epoch": 1.12, "learning_rate": 4.442821594720329e-05, "loss": 1.8028, "step": 20800 }, { "epoch": 1.12, "learning_rate": 4.44255111976631e-05, "loss": 1.7288, "step": 20810 }, { "epoch": 1.12, "learning_rate": 4.4422806448122903e-05, "loss": 1.7939, "step": 20820 }, { "epoch": 1.13, "learning_rate": 4.442010169858271e-05, "loss": 1.7634, "step": 20830 }, { "epoch": 1.13, "learning_rate": 4.441739694904252e-05, "loss": 1.7425, "step": 20840 }, { "epoch": 1.13, "learning_rate": 4.4414692199502326e-05, "loss": 1.7739, "step": 20850 }, { "epoch": 1.13, "learning_rate": 4.4411987449962135e-05, "loss": 1.8818, "step": 20860 }, { "epoch": 1.13, "learning_rate": 4.4409282700421945e-05, "loss": 1.8147, "step": 20870 }, { "epoch": 1.13, "learning_rate": 4.440657795088175e-05, "loss": 1.8492, "step": 20880 }, { "epoch": 1.13, "learning_rate": 4.440387320134156e-05, "loss": 1.7333, "step": 20890 }, { "epoch": 1.13, "learning_rate": 4.440116845180137e-05, "loss": 1.7546, "step": 20900 }, { "epoch": 1.13, "learning_rate": 4.439846370226117e-05, "loss": 1.7809, "step": 20910 }, { "epoch": 1.13, "learning_rate": 4.439575895272098e-05, "loss": 1.8403, "step": 20920 }, { "epoch": 1.13, "learning_rate": 4.439305420318079e-05, "loss": 1.7901, "step": 20930 }, { "epoch": 1.13, "learning_rate": 4.439034945364059e-05, "loss": 1.847, "step": 20940 }, { "epoch": 1.13, "learning_rate": 4.43876447041004e-05, "loss": 1.7952, "step": 20950 }, { "epoch": 1.13, "learning_rate": 4.438493995456021e-05, "loss": 1.8004, "step": 20960 }, { "epoch": 1.13, "learning_rate": 4.4382235205020014e-05, "loss": 1.8626, "step": 20970 }, { "epoch": 1.13, "learning_rate": 4.4379530455479824e-05, "loss": 1.8341, "step": 20980 }, { "epoch": 1.13, "learning_rate": 4.4376825705939634e-05, "loss": 1.7773, "step": 20990 }, { "epoch": 1.13, "learning_rate": 4.4374120956399436e-05, "loss": 1.7493, "step": 21000 }, { "epoch": 1.14, "learning_rate": 4.4371416206859246e-05, "loss": 1.7589, "step": 21010 }, { "epoch": 1.14, "learning_rate": 4.4368711457319056e-05, "loss": 1.8354, "step": 21020 }, { "epoch": 1.14, "learning_rate": 4.4366006707778865e-05, "loss": 1.7458, "step": 21030 }, { "epoch": 1.14, "learning_rate": 4.436330195823867e-05, "loss": 1.7949, "step": 21040 }, { "epoch": 1.14, "learning_rate": 4.436059720869848e-05, "loss": 1.764, "step": 21050 }, { "epoch": 1.14, "learning_rate": 4.435789245915829e-05, "loss": 1.8106, "step": 21060 }, { "epoch": 1.14, "learning_rate": 4.435518770961809e-05, "loss": 1.74, "step": 21070 }, { "epoch": 1.14, "learning_rate": 4.43524829600779e-05, "loss": 1.7969, "step": 21080 }, { "epoch": 1.14, "learning_rate": 4.434977821053771e-05, "loss": 1.8315, "step": 21090 }, { "epoch": 1.14, "learning_rate": 4.434707346099751e-05, "loss": 1.7269, "step": 21100 }, { "epoch": 1.14, "learning_rate": 4.434436871145732e-05, "loss": 1.8182, "step": 21110 }, { "epoch": 1.14, "learning_rate": 4.434166396191713e-05, "loss": 1.7454, "step": 21120 }, { "epoch": 1.14, "learning_rate": 4.4338959212376935e-05, "loss": 1.7983, "step": 21130 }, { "epoch": 1.14, "learning_rate": 4.4336254462836744e-05, "loss": 1.8326, "step": 21140 }, { "epoch": 1.14, "learning_rate": 4.4333549713296554e-05, "loss": 1.8491, "step": 21150 }, { "epoch": 1.14, "learning_rate": 4.433084496375636e-05, "loss": 1.7847, "step": 21160 }, { "epoch": 1.14, "learning_rate": 4.4328140214216167e-05, "loss": 1.8452, "step": 21170 }, { "epoch": 1.14, "learning_rate": 4.4325435464675976e-05, "loss": 1.9286, "step": 21180 }, { "epoch": 1.14, "learning_rate": 4.432273071513578e-05, "loss": 1.8509, "step": 21190 }, { "epoch": 1.15, "learning_rate": 4.432002596559559e-05, "loss": 1.7588, "step": 21200 }, { "epoch": 1.15, "learning_rate": 4.43173212160554e-05, "loss": 1.7909, "step": 21210 }, { "epoch": 1.15, "learning_rate": 4.43146164665152e-05, "loss": 1.7804, "step": 21220 }, { "epoch": 1.15, "learning_rate": 4.4311911716975004e-05, "loss": 1.8199, "step": 21230 }, { "epoch": 1.15, "learning_rate": 4.430920696743482e-05, "loss": 1.7597, "step": 21240 }, { "epoch": 1.15, "learning_rate": 4.4306502217894623e-05, "loss": 1.7412, "step": 21250 }, { "epoch": 1.15, "learning_rate": 4.430379746835443e-05, "loss": 1.7614, "step": 21260 }, { "epoch": 1.15, "learning_rate": 4.430109271881424e-05, "loss": 1.8446, "step": 21270 }, { "epoch": 1.15, "learning_rate": 4.4298387969274046e-05, "loss": 1.8277, "step": 21280 }, { "epoch": 1.15, "learning_rate": 4.4295683219733855e-05, "loss": 1.8739, "step": 21290 }, { "epoch": 1.15, "learning_rate": 4.4292978470193665e-05, "loss": 1.885, "step": 21300 }, { "epoch": 1.15, "learning_rate": 4.429027372065347e-05, "loss": 1.8132, "step": 21310 }, { "epoch": 1.15, "learning_rate": 4.428756897111328e-05, "loss": 1.8215, "step": 21320 }, { "epoch": 1.15, "learning_rate": 4.428486422157309e-05, "loss": 1.7791, "step": 21330 }, { "epoch": 1.15, "learning_rate": 4.428215947203289e-05, "loss": 1.7767, "step": 21340 }, { "epoch": 1.15, "learning_rate": 4.42794547224927e-05, "loss": 1.8255, "step": 21350 }, { "epoch": 1.15, "learning_rate": 4.427674997295251e-05, "loss": 1.7871, "step": 21360 }, { "epoch": 1.15, "learning_rate": 4.427404522341231e-05, "loss": 1.8902, "step": 21370 }, { "epoch": 1.16, "learning_rate": 4.427134047387212e-05, "loss": 1.8823, "step": 21380 }, { "epoch": 1.16, "learning_rate": 4.426863572433193e-05, "loss": 1.869, "step": 21390 }, { "epoch": 1.16, "learning_rate": 4.4265930974791734e-05, "loss": 1.7757, "step": 21400 }, { "epoch": 1.16, "learning_rate": 4.4263226225251544e-05, "loss": 1.8034, "step": 21410 }, { "epoch": 1.16, "learning_rate": 4.4260521475711354e-05, "loss": 1.8099, "step": 21420 }, { "epoch": 1.16, "learning_rate": 4.4257816726171156e-05, "loss": 1.7413, "step": 21430 }, { "epoch": 1.16, "learning_rate": 4.4255111976630966e-05, "loss": 1.8672, "step": 21440 }, { "epoch": 1.16, "learning_rate": 4.4252407227090776e-05, "loss": 1.6496, "step": 21450 }, { "epoch": 1.16, "learning_rate": 4.424970247755058e-05, "loss": 1.7798, "step": 21460 }, { "epoch": 1.16, "learning_rate": 4.424699772801039e-05, "loss": 1.7907, "step": 21470 }, { "epoch": 1.16, "learning_rate": 4.42442929784702e-05, "loss": 1.8192, "step": 21480 }, { "epoch": 1.16, "learning_rate": 4.424158822893e-05, "loss": 1.8413, "step": 21490 }, { "epoch": 1.16, "learning_rate": 4.423888347938981e-05, "loss": 1.8409, "step": 21500 }, { "epoch": 1.16, "learning_rate": 4.423617872984962e-05, "loss": 1.816, "step": 21510 }, { "epoch": 1.16, "learning_rate": 4.423347398030942e-05, "loss": 1.7612, "step": 21520 }, { "epoch": 1.16, "learning_rate": 4.423076923076923e-05, "loss": 1.8133, "step": 21530 }, { "epoch": 1.16, "learning_rate": 4.422806448122904e-05, "loss": 1.8676, "step": 21540 }, { "epoch": 1.16, "learning_rate": 4.4225359731688845e-05, "loss": 1.8066, "step": 21550 }, { "epoch": 1.16, "learning_rate": 4.4222654982148655e-05, "loss": 1.7588, "step": 21560 }, { "epoch": 1.17, "learning_rate": 4.4219950232608464e-05, "loss": 1.8434, "step": 21570 }, { "epoch": 1.17, "learning_rate": 4.421724548306827e-05, "loss": 1.7753, "step": 21580 }, { "epoch": 1.17, "learning_rate": 4.421454073352808e-05, "loss": 1.7584, "step": 21590 }, { "epoch": 1.17, "learning_rate": 4.421183598398789e-05, "loss": 1.8078, "step": 21600 }, { "epoch": 1.17, "learning_rate": 4.420913123444769e-05, "loss": 1.8104, "step": 21610 }, { "epoch": 1.17, "learning_rate": 4.42064264849075e-05, "loss": 1.9282, "step": 21620 }, { "epoch": 1.17, "learning_rate": 4.420372173536731e-05, "loss": 1.8494, "step": 21630 }, { "epoch": 1.17, "learning_rate": 4.420101698582711e-05, "loss": 1.7486, "step": 21640 }, { "epoch": 1.17, "learning_rate": 4.419831223628692e-05, "loss": 1.7671, "step": 21650 }, { "epoch": 1.17, "learning_rate": 4.419560748674673e-05, "loss": 1.8331, "step": 21660 }, { "epoch": 1.17, "learning_rate": 4.419290273720654e-05, "loss": 1.8805, "step": 21670 }, { "epoch": 1.17, "learning_rate": 4.4190197987666343e-05, "loss": 1.8353, "step": 21680 }, { "epoch": 1.17, "learning_rate": 4.418749323812615e-05, "loss": 1.8322, "step": 21690 }, { "epoch": 1.17, "learning_rate": 4.418478848858596e-05, "loss": 1.7745, "step": 21700 }, { "epoch": 1.17, "learning_rate": 4.4182083739045766e-05, "loss": 1.8317, "step": 21710 }, { "epoch": 1.17, "learning_rate": 4.4179378989505575e-05, "loss": 1.8365, "step": 21720 }, { "epoch": 1.17, "learning_rate": 4.4176674239965385e-05, "loss": 1.8966, "step": 21730 }, { "epoch": 1.17, "learning_rate": 4.417396949042519e-05, "loss": 1.7472, "step": 21740 }, { "epoch": 1.18, "learning_rate": 4.4171264740885e-05, "loss": 1.74, "step": 21750 }, { "epoch": 1.18, "learning_rate": 4.416855999134481e-05, "loss": 1.8624, "step": 21760 }, { "epoch": 1.18, "learning_rate": 4.416585524180461e-05, "loss": 1.82, "step": 21770 }, { "epoch": 1.18, "learning_rate": 4.416315049226441e-05, "loss": 1.8054, "step": 21780 }, { "epoch": 1.18, "learning_rate": 4.416044574272423e-05, "loss": 1.7845, "step": 21790 }, { "epoch": 1.18, "learning_rate": 4.415774099318403e-05, "loss": 1.8718, "step": 21800 }, { "epoch": 1.18, "learning_rate": 4.4155036243643835e-05, "loss": 1.9106, "step": 21810 }, { "epoch": 1.18, "learning_rate": 4.415233149410365e-05, "loss": 1.8265, "step": 21820 }, { "epoch": 1.18, "learning_rate": 4.4149626744563454e-05, "loss": 1.857, "step": 21830 }, { "epoch": 1.18, "learning_rate": 4.414692199502326e-05, "loss": 1.8164, "step": 21840 }, { "epoch": 1.18, "learning_rate": 4.4144217245483074e-05, "loss": 1.7338, "step": 21850 }, { "epoch": 1.18, "learning_rate": 4.4141512495942876e-05, "loss": 1.7778, "step": 21860 }, { "epoch": 1.18, "learning_rate": 4.4138807746402686e-05, "loss": 1.7988, "step": 21870 }, { "epoch": 1.18, "learning_rate": 4.4136102996862496e-05, "loss": 1.8551, "step": 21880 }, { "epoch": 1.18, "learning_rate": 4.41333982473223e-05, "loss": 1.832, "step": 21890 }, { "epoch": 1.18, "learning_rate": 4.413069349778211e-05, "loss": 1.7303, "step": 21900 }, { "epoch": 1.18, "learning_rate": 4.412798874824192e-05, "loss": 1.7992, "step": 21910 }, { "epoch": 1.18, "learning_rate": 4.412528399870172e-05, "loss": 1.7701, "step": 21920 }, { "epoch": 1.18, "learning_rate": 4.412257924916153e-05, "loss": 1.8114, "step": 21930 }, { "epoch": 1.19, "learning_rate": 4.411987449962134e-05, "loss": 1.7702, "step": 21940 }, { "epoch": 1.19, "learning_rate": 4.411716975008114e-05, "loss": 1.8061, "step": 21950 }, { "epoch": 1.19, "learning_rate": 4.411446500054095e-05, "loss": 1.8105, "step": 21960 }, { "epoch": 1.19, "learning_rate": 4.411176025100076e-05, "loss": 1.784, "step": 21970 }, { "epoch": 1.19, "learning_rate": 4.4109055501460565e-05, "loss": 1.7884, "step": 21980 }, { "epoch": 1.19, "learning_rate": 4.4106350751920375e-05, "loss": 1.8091, "step": 21990 }, { "epoch": 1.19, "learning_rate": 4.4103646002380184e-05, "loss": 1.7698, "step": 22000 }, { "epoch": 1.19, "learning_rate": 4.410094125283999e-05, "loss": 1.8581, "step": 22010 }, { "epoch": 1.19, "learning_rate": 4.40982365032998e-05, "loss": 1.7749, "step": 22020 }, { "epoch": 1.19, "learning_rate": 4.409553175375961e-05, "loss": 1.8215, "step": 22030 }, { "epoch": 1.19, "learning_rate": 4.409282700421941e-05, "loss": 1.7617, "step": 22040 }, { "epoch": 1.19, "learning_rate": 4.409012225467922e-05, "loss": 1.8353, "step": 22050 }, { "epoch": 1.19, "learning_rate": 4.408741750513903e-05, "loss": 1.8278, "step": 22060 }, { "epoch": 1.19, "learning_rate": 4.408471275559883e-05, "loss": 1.771, "step": 22070 }, { "epoch": 1.19, "learning_rate": 4.408200800605864e-05, "loss": 1.811, "step": 22080 }, { "epoch": 1.19, "learning_rate": 4.407930325651845e-05, "loss": 1.8899, "step": 22090 }, { "epoch": 1.19, "learning_rate": 4.4076598506978254e-05, "loss": 1.8669, "step": 22100 }, { "epoch": 1.19, "learning_rate": 4.4073893757438063e-05, "loss": 1.7712, "step": 22110 }, { "epoch": 1.2, "learning_rate": 4.407118900789787e-05, "loss": 1.7926, "step": 22120 }, { "epoch": 1.2, "learning_rate": 4.4068484258357676e-05, "loss": 1.7452, "step": 22130 }, { "epoch": 1.2, "learning_rate": 4.4065779508817486e-05, "loss": 1.8666, "step": 22140 }, { "epoch": 1.2, "learning_rate": 4.4063074759277295e-05, "loss": 1.7327, "step": 22150 }, { "epoch": 1.2, "learning_rate": 4.40603700097371e-05, "loss": 1.7373, "step": 22160 }, { "epoch": 1.2, "learning_rate": 4.405766526019691e-05, "loss": 1.7797, "step": 22170 }, { "epoch": 1.2, "learning_rate": 4.405496051065672e-05, "loss": 1.7855, "step": 22180 }, { "epoch": 1.2, "learning_rate": 4.405225576111652e-05, "loss": 1.8416, "step": 22190 }, { "epoch": 1.2, "learning_rate": 4.404955101157633e-05, "loss": 1.8375, "step": 22200 }, { "epoch": 1.2, "learning_rate": 4.404684626203614e-05, "loss": 1.7795, "step": 22210 }, { "epoch": 1.2, "learning_rate": 4.404414151249594e-05, "loss": 1.8052, "step": 22220 }, { "epoch": 1.2, "learning_rate": 4.404143676295575e-05, "loss": 1.784, "step": 22230 }, { "epoch": 1.2, "learning_rate": 4.403873201341556e-05, "loss": 1.8332, "step": 22240 }, { "epoch": 1.2, "learning_rate": 4.4036027263875365e-05, "loss": 1.7849, "step": 22250 }, { "epoch": 1.2, "learning_rate": 4.4033322514335174e-05, "loss": 1.7869, "step": 22260 }, { "epoch": 1.2, "learning_rate": 4.4030617764794984e-05, "loss": 1.8484, "step": 22270 }, { "epoch": 1.2, "learning_rate": 4.4027913015254794e-05, "loss": 1.8615, "step": 22280 }, { "epoch": 1.2, "learning_rate": 4.4025208265714597e-05, "loss": 1.8389, "step": 22290 }, { "epoch": 1.2, "learning_rate": 4.4022503516174406e-05, "loss": 1.815, "step": 22300 }, { "epoch": 1.21, "learning_rate": 4.4019798766634216e-05, "loss": 1.8303, "step": 22310 }, { "epoch": 1.21, "learning_rate": 4.401709401709402e-05, "loss": 1.7929, "step": 22320 }, { "epoch": 1.21, "learning_rate": 4.401438926755382e-05, "loss": 1.845, "step": 22330 }, { "epoch": 1.21, "learning_rate": 4.401168451801364e-05, "loss": 1.8188, "step": 22340 }, { "epoch": 1.21, "learning_rate": 4.400897976847344e-05, "loss": 1.8386, "step": 22350 }, { "epoch": 1.21, "learning_rate": 4.4006275018933244e-05, "loss": 1.724, "step": 22360 }, { "epoch": 1.21, "learning_rate": 4.400357026939306e-05, "loss": 1.7503, "step": 22370 }, { "epoch": 1.21, "learning_rate": 4.400086551985286e-05, "loss": 1.7758, "step": 22380 }, { "epoch": 1.21, "learning_rate": 4.3998160770312666e-05, "loss": 1.8612, "step": 22390 }, { "epoch": 1.21, "learning_rate": 4.399545602077248e-05, "loss": 1.8483, "step": 22400 }, { "epoch": 1.21, "learning_rate": 4.3992751271232285e-05, "loss": 1.7923, "step": 22410 }, { "epoch": 1.21, "learning_rate": 4.399004652169209e-05, "loss": 1.8348, "step": 22420 }, { "epoch": 1.21, "learning_rate": 4.3987341772151904e-05, "loss": 1.8021, "step": 22430 }, { "epoch": 1.21, "learning_rate": 4.398463702261171e-05, "loss": 1.8384, "step": 22440 }, { "epoch": 1.21, "learning_rate": 4.398193227307151e-05, "loss": 1.851, "step": 22450 }, { "epoch": 1.21, "learning_rate": 4.397922752353133e-05, "loss": 1.8397, "step": 22460 }, { "epoch": 1.21, "learning_rate": 4.397652277399113e-05, "loss": 1.7805, "step": 22470 }, { "epoch": 1.21, "learning_rate": 4.397381802445093e-05, "loss": 1.8279, "step": 22480 }, { "epoch": 1.22, "learning_rate": 4.397111327491075e-05, "loss": 1.8464, "step": 22490 }, { "epoch": 1.22, "learning_rate": 4.396840852537055e-05, "loss": 1.81, "step": 22500 }, { "epoch": 1.22, "learning_rate": 4.396570377583036e-05, "loss": 1.7966, "step": 22510 }, { "epoch": 1.22, "learning_rate": 4.396299902629017e-05, "loss": 1.8523, "step": 22520 }, { "epoch": 1.22, "learning_rate": 4.3960294276749974e-05, "loss": 1.7697, "step": 22530 }, { "epoch": 1.22, "learning_rate": 4.3957589527209784e-05, "loss": 1.7885, "step": 22540 }, { "epoch": 1.22, "learning_rate": 4.395488477766959e-05, "loss": 1.8462, "step": 22550 }, { "epoch": 1.22, "learning_rate": 4.3952180028129396e-05, "loss": 1.878, "step": 22560 }, { "epoch": 1.22, "learning_rate": 4.3949475278589206e-05, "loss": 1.8492, "step": 22570 }, { "epoch": 1.22, "learning_rate": 4.3946770529049015e-05, "loss": 1.851, "step": 22580 }, { "epoch": 1.22, "learning_rate": 4.394406577950882e-05, "loss": 1.8166, "step": 22590 }, { "epoch": 1.22, "learning_rate": 4.394136102996863e-05, "loss": 1.8143, "step": 22600 }, { "epoch": 1.22, "learning_rate": 4.393865628042844e-05, "loss": 1.7854, "step": 22610 }, { "epoch": 1.22, "learning_rate": 4.393595153088824e-05, "loss": 1.9286, "step": 22620 }, { "epoch": 1.22, "learning_rate": 4.393324678134805e-05, "loss": 1.7289, "step": 22630 }, { "epoch": 1.22, "learning_rate": 4.393054203180786e-05, "loss": 1.7976, "step": 22640 }, { "epoch": 1.22, "learning_rate": 4.392783728226766e-05, "loss": 1.8335, "step": 22650 }, { "epoch": 1.22, "learning_rate": 4.392513253272747e-05, "loss": 1.7735, "step": 22660 }, { "epoch": 1.22, "learning_rate": 4.392242778318728e-05, "loss": 1.8073, "step": 22670 }, { "epoch": 1.23, "learning_rate": 4.3919723033647085e-05, "loss": 1.744, "step": 22680 }, { "epoch": 1.23, "learning_rate": 4.3917018284106894e-05, "loss": 1.8786, "step": 22690 }, { "epoch": 1.23, "learning_rate": 4.3914313534566704e-05, "loss": 1.7906, "step": 22700 }, { "epoch": 1.23, "learning_rate": 4.391160878502651e-05, "loss": 1.7794, "step": 22710 }, { "epoch": 1.23, "learning_rate": 4.3908904035486317e-05, "loss": 1.8414, "step": 22720 }, { "epoch": 1.23, "learning_rate": 4.3906199285946126e-05, "loss": 1.8365, "step": 22730 }, { "epoch": 1.23, "learning_rate": 4.390349453640593e-05, "loss": 1.7735, "step": 22740 }, { "epoch": 1.23, "learning_rate": 4.390078978686574e-05, "loss": 1.8431, "step": 22750 }, { "epoch": 1.23, "learning_rate": 4.389808503732555e-05, "loss": 1.8803, "step": 22760 }, { "epoch": 1.23, "learning_rate": 4.389538028778535e-05, "loss": 1.8202, "step": 22770 }, { "epoch": 1.23, "learning_rate": 4.389267553824516e-05, "loss": 1.7894, "step": 22780 }, { "epoch": 1.23, "learning_rate": 4.388997078870497e-05, "loss": 1.8897, "step": 22790 }, { "epoch": 1.23, "learning_rate": 4.3887266039164773e-05, "loss": 1.7632, "step": 22800 }, { "epoch": 1.23, "learning_rate": 4.388456128962458e-05, "loss": 1.8673, "step": 22810 }, { "epoch": 1.23, "learning_rate": 4.388185654008439e-05, "loss": 1.7817, "step": 22820 }, { "epoch": 1.23, "learning_rate": 4.3879151790544196e-05, "loss": 1.7624, "step": 22830 }, { "epoch": 1.23, "learning_rate": 4.3876447041004005e-05, "loss": 1.8189, "step": 22840 }, { "epoch": 1.23, "learning_rate": 4.3873742291463815e-05, "loss": 1.7661, "step": 22850 }, { "epoch": 1.24, "learning_rate": 4.387103754192362e-05, "loss": 1.8501, "step": 22860 }, { "epoch": 1.24, "learning_rate": 4.386833279238343e-05, "loss": 1.775, "step": 22870 }, { "epoch": 1.24, "learning_rate": 4.386562804284324e-05, "loss": 1.8345, "step": 22880 }, { "epoch": 1.24, "learning_rate": 4.386292329330304e-05, "loss": 1.8633, "step": 22890 }, { "epoch": 1.24, "learning_rate": 4.386021854376285e-05, "loss": 1.8627, "step": 22900 }, { "epoch": 1.24, "learning_rate": 4.385751379422265e-05, "loss": 1.7799, "step": 22910 }, { "epoch": 1.24, "learning_rate": 4.385480904468247e-05, "loss": 1.8569, "step": 22920 }, { "epoch": 1.24, "learning_rate": 4.385210429514227e-05, "loss": 1.8165, "step": 22930 }, { "epoch": 1.24, "learning_rate": 4.3849399545602075e-05, "loss": 1.8408, "step": 22940 }, { "epoch": 1.24, "learning_rate": 4.384669479606189e-05, "loss": 1.7669, "step": 22950 }, { "epoch": 1.24, "learning_rate": 4.3843990046521694e-05, "loss": 1.818, "step": 22960 }, { "epoch": 1.24, "learning_rate": 4.38412852969815e-05, "loss": 1.8397, "step": 22970 }, { "epoch": 1.24, "learning_rate": 4.383858054744131e-05, "loss": 1.8742, "step": 22980 }, { "epoch": 1.24, "learning_rate": 4.3835875797901116e-05, "loss": 1.8137, "step": 22990 }, { "epoch": 1.24, "learning_rate": 4.383317104836092e-05, "loss": 1.7425, "step": 23000 }, { "epoch": 1.24, "learning_rate": 4.3830466298820735e-05, "loss": 1.7851, "step": 23010 }, { "epoch": 1.24, "learning_rate": 4.382776154928054e-05, "loss": 1.7474, "step": 23020 }, { "epoch": 1.24, "learning_rate": 4.382505679974034e-05, "loss": 1.8329, "step": 23030 }, { "epoch": 1.24, "learning_rate": 4.382235205020016e-05, "loss": 1.7755, "step": 23040 }, { "epoch": 1.25, "learning_rate": 4.381964730065996e-05, "loss": 1.7688, "step": 23050 }, { "epoch": 1.25, "learning_rate": 4.381694255111976e-05, "loss": 1.7887, "step": 23060 }, { "epoch": 1.25, "learning_rate": 4.381423780157958e-05, "loss": 1.853, "step": 23070 }, { "epoch": 1.25, "learning_rate": 4.381153305203938e-05, "loss": 1.8037, "step": 23080 }, { "epoch": 1.25, "learning_rate": 4.3808828302499185e-05, "loss": 1.8125, "step": 23090 }, { "epoch": 1.25, "learning_rate": 4.3806123552959e-05, "loss": 1.8435, "step": 23100 }, { "epoch": 1.25, "learning_rate": 4.3803418803418805e-05, "loss": 1.7925, "step": 23110 }, { "epoch": 1.25, "learning_rate": 4.380071405387861e-05, "loss": 1.7668, "step": 23120 }, { "epoch": 1.25, "learning_rate": 4.3798009304338424e-05, "loss": 1.804, "step": 23130 }, { "epoch": 1.25, "learning_rate": 4.379530455479823e-05, "loss": 1.8117, "step": 23140 }, { "epoch": 1.25, "learning_rate": 4.3792599805258037e-05, "loss": 1.7817, "step": 23150 }, { "epoch": 1.25, "learning_rate": 4.3789895055717846e-05, "loss": 1.8118, "step": 23160 }, { "epoch": 1.25, "learning_rate": 4.378719030617765e-05, "loss": 1.7504, "step": 23170 }, { "epoch": 1.25, "learning_rate": 4.378448555663746e-05, "loss": 1.8296, "step": 23180 }, { "epoch": 1.25, "learning_rate": 4.378178080709727e-05, "loss": 1.865, "step": 23190 }, { "epoch": 1.25, "learning_rate": 4.377907605755707e-05, "loss": 1.8527, "step": 23200 }, { "epoch": 1.25, "learning_rate": 4.377637130801688e-05, "loss": 1.8395, "step": 23210 }, { "epoch": 1.25, "learning_rate": 4.377366655847669e-05, "loss": 1.8956, "step": 23220 }, { "epoch": 1.26, "learning_rate": 4.3770961808936493e-05, "loss": 1.8504, "step": 23230 }, { "epoch": 1.26, "learning_rate": 4.37682570593963e-05, "loss": 1.8861, "step": 23240 }, { "epoch": 1.26, "learning_rate": 4.376555230985611e-05, "loss": 1.7995, "step": 23250 }, { "epoch": 1.26, "learning_rate": 4.3762847560315916e-05, "loss": 1.6933, "step": 23260 }, { "epoch": 1.26, "learning_rate": 4.3760142810775725e-05, "loss": 1.7975, "step": 23270 }, { "epoch": 1.26, "learning_rate": 4.3757438061235535e-05, "loss": 1.7908, "step": 23280 }, { "epoch": 1.26, "learning_rate": 4.375473331169534e-05, "loss": 1.7616, "step": 23290 }, { "epoch": 1.26, "learning_rate": 4.375202856215515e-05, "loss": 1.7895, "step": 23300 }, { "epoch": 1.26, "learning_rate": 4.374932381261496e-05, "loss": 1.7665, "step": 23310 }, { "epoch": 1.26, "learning_rate": 4.374661906307476e-05, "loss": 1.8317, "step": 23320 }, { "epoch": 1.26, "learning_rate": 4.374391431353457e-05, "loss": 1.8636, "step": 23330 }, { "epoch": 1.26, "learning_rate": 4.374120956399438e-05, "loss": 1.799, "step": 23340 }, { "epoch": 1.26, "learning_rate": 4.373850481445418e-05, "loss": 1.7337, "step": 23350 }, { "epoch": 1.26, "learning_rate": 4.373580006491399e-05, "loss": 1.858, "step": 23360 }, { "epoch": 1.26, "learning_rate": 4.37330953153738e-05, "loss": 1.7693, "step": 23370 }, { "epoch": 1.26, "learning_rate": 4.3730390565833604e-05, "loss": 1.8327, "step": 23380 }, { "epoch": 1.26, "learning_rate": 4.3727685816293414e-05, "loss": 1.8656, "step": 23390 }, { "epoch": 1.26, "learning_rate": 4.3724981066753224e-05, "loss": 1.8011, "step": 23400 }, { "epoch": 1.26, "learning_rate": 4.3722276317213026e-05, "loss": 1.7735, "step": 23410 }, { "epoch": 1.27, "learning_rate": 4.3719571567672836e-05, "loss": 1.7938, "step": 23420 }, { "epoch": 1.27, "learning_rate": 4.3716866818132646e-05, "loss": 1.8166, "step": 23430 }, { "epoch": 1.27, "learning_rate": 4.371416206859245e-05, "loss": 1.8009, "step": 23440 }, { "epoch": 1.27, "learning_rate": 4.371145731905226e-05, "loss": 1.7542, "step": 23450 }, { "epoch": 1.27, "learning_rate": 4.370875256951206e-05, "loss": 1.7976, "step": 23460 }, { "epoch": 1.27, "learning_rate": 4.370604781997187e-05, "loss": 1.7668, "step": 23470 }, { "epoch": 1.27, "learning_rate": 4.370334307043168e-05, "loss": 1.7749, "step": 23480 }, { "epoch": 1.27, "learning_rate": 4.370063832089148e-05, "loss": 1.8007, "step": 23490 }, { "epoch": 1.27, "learning_rate": 4.369793357135129e-05, "loss": 1.7803, "step": 23500 }, { "epoch": 1.27, "learning_rate": 4.36952288218111e-05, "loss": 1.8381, "step": 23510 }, { "epoch": 1.27, "learning_rate": 4.3692524072270905e-05, "loss": 1.6825, "step": 23520 }, { "epoch": 1.27, "learning_rate": 4.3689819322730715e-05, "loss": 1.7675, "step": 23530 }, { "epoch": 1.27, "learning_rate": 4.3687114573190525e-05, "loss": 1.8406, "step": 23540 }, { "epoch": 1.27, "learning_rate": 4.368440982365033e-05, "loss": 1.7761, "step": 23550 }, { "epoch": 1.27, "learning_rate": 4.3681705074110144e-05, "loss": 1.778, "step": 23560 }, { "epoch": 1.27, "learning_rate": 4.367900032456995e-05, "loss": 1.8829, "step": 23570 }, { "epoch": 1.27, "learning_rate": 4.367629557502975e-05, "loss": 1.7295, "step": 23580 }, { "epoch": 1.27, "learning_rate": 4.3673590825489566e-05, "loss": 1.7931, "step": 23590 }, { "epoch": 1.28, "learning_rate": 4.367088607594937e-05, "loss": 1.8311, "step": 23600 }, { "epoch": 1.28, "learning_rate": 4.366818132640917e-05, "loss": 1.8189, "step": 23610 }, { "epoch": 1.28, "learning_rate": 4.366547657686899e-05, "loss": 1.831, "step": 23620 }, { "epoch": 1.28, "learning_rate": 4.366277182732879e-05, "loss": 1.8025, "step": 23630 }, { "epoch": 1.28, "learning_rate": 4.3660067077788594e-05, "loss": 1.8284, "step": 23640 }, { "epoch": 1.28, "learning_rate": 4.365736232824841e-05, "loss": 1.8453, "step": 23650 }, { "epoch": 1.28, "learning_rate": 4.3654657578708213e-05, "loss": 1.8814, "step": 23660 }, { "epoch": 1.28, "learning_rate": 4.3651952829168016e-05, "loss": 1.6939, "step": 23670 }, { "epoch": 1.28, "learning_rate": 4.364924807962783e-05, "loss": 1.8169, "step": 23680 }, { "epoch": 1.28, "learning_rate": 4.3646543330087636e-05, "loss": 1.8146, "step": 23690 }, { "epoch": 1.28, "learning_rate": 4.364383858054744e-05, "loss": 1.8693, "step": 23700 }, { "epoch": 1.28, "learning_rate": 4.3641133831007255e-05, "loss": 1.7558, "step": 23710 }, { "epoch": 1.28, "learning_rate": 4.363842908146706e-05, "loss": 1.8462, "step": 23720 }, { "epoch": 1.28, "learning_rate": 4.363572433192686e-05, "loss": 1.7897, "step": 23730 }, { "epoch": 1.28, "learning_rate": 4.363301958238668e-05, "loss": 1.7769, "step": 23740 }, { "epoch": 1.28, "learning_rate": 4.363031483284648e-05, "loss": 1.8088, "step": 23750 }, { "epoch": 1.28, "learning_rate": 4.362761008330628e-05, "loss": 1.8924, "step": 23760 }, { "epoch": 1.28, "learning_rate": 4.36249053337661e-05, "loss": 1.7379, "step": 23770 }, { "epoch": 1.28, "learning_rate": 4.36222005842259e-05, "loss": 1.8929, "step": 23780 }, { "epoch": 1.29, "learning_rate": 4.361949583468571e-05, "loss": 1.8455, "step": 23790 }, { "epoch": 1.29, "learning_rate": 4.361679108514552e-05, "loss": 1.854, "step": 23800 }, { "epoch": 1.29, "learning_rate": 4.3614086335605324e-05, "loss": 1.752, "step": 23810 }, { "epoch": 1.29, "learning_rate": 4.3611381586065134e-05, "loss": 1.851, "step": 23820 }, { "epoch": 1.29, "learning_rate": 4.3608676836524944e-05, "loss": 1.7646, "step": 23830 }, { "epoch": 1.29, "learning_rate": 4.3605972086984746e-05, "loss": 1.7855, "step": 23840 }, { "epoch": 1.29, "learning_rate": 4.3603267337444556e-05, "loss": 1.7663, "step": 23850 }, { "epoch": 1.29, "learning_rate": 4.3600562587904366e-05, "loss": 1.7701, "step": 23860 }, { "epoch": 1.29, "learning_rate": 4.359785783836417e-05, "loss": 1.8562, "step": 23870 }, { "epoch": 1.29, "learning_rate": 4.359515308882398e-05, "loss": 1.7464, "step": 23880 }, { "epoch": 1.29, "learning_rate": 4.359244833928379e-05, "loss": 1.8438, "step": 23890 }, { "epoch": 1.29, "learning_rate": 4.358974358974359e-05, "loss": 1.8029, "step": 23900 }, { "epoch": 1.29, "learning_rate": 4.35870388402034e-05, "loss": 1.8612, "step": 23910 }, { "epoch": 1.29, "learning_rate": 4.358433409066321e-05, "loss": 1.7337, "step": 23920 }, { "epoch": 1.29, "learning_rate": 4.358162934112301e-05, "loss": 1.7976, "step": 23930 }, { "epoch": 1.29, "learning_rate": 4.357892459158282e-05, "loss": 1.817, "step": 23940 }, { "epoch": 1.29, "learning_rate": 4.357621984204263e-05, "loss": 1.8067, "step": 23950 }, { "epoch": 1.29, "learning_rate": 4.3573515092502435e-05, "loss": 1.7831, "step": 23960 }, { "epoch": 1.3, "learning_rate": 4.3570810342962245e-05, "loss": 1.7319, "step": 23970 }, { "epoch": 1.3, "learning_rate": 4.3568105593422054e-05, "loss": 1.7706, "step": 23980 }, { "epoch": 1.3, "learning_rate": 4.356540084388186e-05, "loss": 1.7583, "step": 23990 }, { "epoch": 1.3, "learning_rate": 4.356269609434167e-05, "loss": 1.8216, "step": 24000 }, { "epoch": 1.3, "learning_rate": 4.355999134480147e-05, "loss": 1.8178, "step": 24010 }, { "epoch": 1.3, "learning_rate": 4.355728659526128e-05, "loss": 1.7334, "step": 24020 }, { "epoch": 1.3, "learning_rate": 4.355458184572109e-05, "loss": 1.7956, "step": 24030 }, { "epoch": 1.3, "learning_rate": 4.355187709618089e-05, "loss": 1.7527, "step": 24040 }, { "epoch": 1.3, "learning_rate": 4.35491723466407e-05, "loss": 1.8166, "step": 24050 }, { "epoch": 1.3, "learning_rate": 4.354646759710051e-05, "loss": 1.7734, "step": 24060 }, { "epoch": 1.3, "learning_rate": 4.3543762847560314e-05, "loss": 1.8581, "step": 24070 }, { "epoch": 1.3, "learning_rate": 4.3541058098020124e-05, "loss": 1.7346, "step": 24080 }, { "epoch": 1.3, "learning_rate": 4.3538353348479933e-05, "loss": 1.8191, "step": 24090 }, { "epoch": 1.3, "learning_rate": 4.3535648598939736e-05, "loss": 1.8652, "step": 24100 }, { "epoch": 1.3, "learning_rate": 4.3532943849399546e-05, "loss": 1.7433, "step": 24110 }, { "epoch": 1.3, "learning_rate": 4.3530239099859356e-05, "loss": 1.8244, "step": 24120 }, { "epoch": 1.3, "learning_rate": 4.352753435031916e-05, "loss": 1.7928, "step": 24130 }, { "epoch": 1.3, "learning_rate": 4.352482960077897e-05, "loss": 1.7942, "step": 24140 }, { "epoch": 1.3, "learning_rate": 4.352212485123878e-05, "loss": 1.8878, "step": 24150 }, { "epoch": 1.31, "learning_rate": 4.351942010169858e-05, "loss": 1.8106, "step": 24160 }, { "epoch": 1.31, "learning_rate": 4.351671535215839e-05, "loss": 1.8735, "step": 24170 }, { "epoch": 1.31, "learning_rate": 4.35140106026182e-05, "loss": 1.7761, "step": 24180 }, { "epoch": 1.31, "learning_rate": 4.3511305853078e-05, "loss": 1.7976, "step": 24190 }, { "epoch": 1.31, "learning_rate": 4.350860110353782e-05, "loss": 1.8226, "step": 24200 }, { "epoch": 1.31, "learning_rate": 4.350589635399762e-05, "loss": 1.88, "step": 24210 }, { "epoch": 1.31, "learning_rate": 4.3503191604457425e-05, "loss": 1.7793, "step": 24220 }, { "epoch": 1.31, "learning_rate": 4.350048685491724e-05, "loss": 1.7201, "step": 24230 }, { "epoch": 1.31, "learning_rate": 4.3497782105377044e-05, "loss": 1.8146, "step": 24240 }, { "epoch": 1.31, "learning_rate": 4.349507735583685e-05, "loss": 1.7137, "step": 24250 }, { "epoch": 1.31, "learning_rate": 4.3492372606296664e-05, "loss": 1.8426, "step": 24260 }, { "epoch": 1.31, "learning_rate": 4.3489667856756466e-05, "loss": 1.8555, "step": 24270 }, { "epoch": 1.31, "learning_rate": 4.348696310721627e-05, "loss": 1.8224, "step": 24280 }, { "epoch": 1.31, "learning_rate": 4.3484258357676086e-05, "loss": 1.8249, "step": 24290 }, { "epoch": 1.31, "learning_rate": 4.348155360813589e-05, "loss": 1.8661, "step": 24300 }, { "epoch": 1.31, "learning_rate": 4.347884885859569e-05, "loss": 1.734, "step": 24310 }, { "epoch": 1.31, "learning_rate": 4.347614410905551e-05, "loss": 1.9118, "step": 24320 }, { "epoch": 1.31, "learning_rate": 4.347343935951531e-05, "loss": 1.8555, "step": 24330 }, { "epoch": 1.32, "learning_rate": 4.3470734609975114e-05, "loss": 1.8482, "step": 24340 }, { "epoch": 1.32, "learning_rate": 4.346802986043493e-05, "loss": 1.8773, "step": 24350 }, { "epoch": 1.32, "learning_rate": 4.346532511089473e-05, "loss": 1.8154, "step": 24360 }, { "epoch": 1.32, "learning_rate": 4.3462620361354536e-05, "loss": 1.8408, "step": 24370 }, { "epoch": 1.32, "learning_rate": 4.345991561181435e-05, "loss": 1.7493, "step": 24380 }, { "epoch": 1.32, "learning_rate": 4.3457210862274155e-05, "loss": 1.8035, "step": 24390 }, { "epoch": 1.32, "learning_rate": 4.3454506112733965e-05, "loss": 1.7767, "step": 24400 }, { "epoch": 1.32, "learning_rate": 4.3451801363193774e-05, "loss": 1.7691, "step": 24410 }, { "epoch": 1.32, "learning_rate": 4.344909661365358e-05, "loss": 1.8642, "step": 24420 }, { "epoch": 1.32, "learning_rate": 4.344639186411339e-05, "loss": 1.92, "step": 24430 }, { "epoch": 1.32, "learning_rate": 4.34436871145732e-05, "loss": 1.8031, "step": 24440 }, { "epoch": 1.32, "learning_rate": 4.3440982365033e-05, "loss": 1.8128, "step": 24450 }, { "epoch": 1.32, "learning_rate": 4.343827761549281e-05, "loss": 1.7425, "step": 24460 }, { "epoch": 1.32, "learning_rate": 4.343557286595262e-05, "loss": 1.7492, "step": 24470 }, { "epoch": 1.32, "learning_rate": 4.343286811641242e-05, "loss": 1.7822, "step": 24480 }, { "epoch": 1.32, "learning_rate": 4.343016336687223e-05, "loss": 1.804, "step": 24490 }, { "epoch": 1.32, "learning_rate": 4.342745861733204e-05, "loss": 1.7828, "step": 24500 }, { "epoch": 1.32, "learning_rate": 4.3424753867791844e-05, "loss": 1.7855, "step": 24510 }, { "epoch": 1.32, "learning_rate": 4.3422049118251653e-05, "loss": 1.7675, "step": 24520 }, { "epoch": 1.33, "learning_rate": 4.341934436871146e-05, "loss": 1.7507, "step": 24530 }, { "epoch": 1.33, "learning_rate": 4.3416639619171266e-05, "loss": 1.7328, "step": 24540 }, { "epoch": 1.33, "learning_rate": 4.3413934869631076e-05, "loss": 1.8149, "step": 24550 }, { "epoch": 1.33, "learning_rate": 4.3411230120090885e-05, "loss": 1.798, "step": 24560 }, { "epoch": 1.33, "learning_rate": 4.340852537055069e-05, "loss": 1.7661, "step": 24570 }, { "epoch": 1.33, "learning_rate": 4.34058206210105e-05, "loss": 1.7553, "step": 24580 }, { "epoch": 1.33, "learning_rate": 4.34031158714703e-05, "loss": 1.8192, "step": 24590 }, { "epoch": 1.33, "learning_rate": 4.340041112193011e-05, "loss": 1.8349, "step": 24600 }, { "epoch": 1.33, "learning_rate": 4.339770637238992e-05, "loss": 1.7822, "step": 24610 }, { "epoch": 1.33, "learning_rate": 4.339500162284972e-05, "loss": 1.8179, "step": 24620 }, { "epoch": 1.33, "learning_rate": 4.339229687330953e-05, "loss": 1.7958, "step": 24630 }, { "epoch": 1.33, "learning_rate": 4.338959212376934e-05, "loss": 1.8146, "step": 24640 }, { "epoch": 1.33, "learning_rate": 4.3386887374229145e-05, "loss": 1.8324, "step": 24650 }, { "epoch": 1.33, "learning_rate": 4.3384182624688955e-05, "loss": 1.7938, "step": 24660 }, { "epoch": 1.33, "learning_rate": 4.3381477875148764e-05, "loss": 1.9384, "step": 24670 }, { "epoch": 1.33, "learning_rate": 4.337877312560857e-05, "loss": 1.7745, "step": 24680 }, { "epoch": 1.33, "learning_rate": 4.337606837606838e-05, "loss": 1.8018, "step": 24690 }, { "epoch": 1.33, "learning_rate": 4.3373363626528186e-05, "loss": 1.8216, "step": 24700 }, { "epoch": 1.34, "learning_rate": 4.337065887698799e-05, "loss": 1.8115, "step": 24710 }, { "epoch": 1.34, "learning_rate": 4.33679541274478e-05, "loss": 1.8314, "step": 24720 }, { "epoch": 1.34, "learning_rate": 4.336524937790761e-05, "loss": 1.8647, "step": 24730 }, { "epoch": 1.34, "learning_rate": 4.336254462836741e-05, "loss": 1.8299, "step": 24740 }, { "epoch": 1.34, "learning_rate": 4.335983987882722e-05, "loss": 1.828, "step": 24750 }, { "epoch": 1.34, "learning_rate": 4.335713512928703e-05, "loss": 1.8672, "step": 24760 }, { "epoch": 1.34, "learning_rate": 4.3354430379746834e-05, "loss": 1.8257, "step": 24770 }, { "epoch": 1.34, "learning_rate": 4.335172563020664e-05, "loss": 1.7895, "step": 24780 }, { "epoch": 1.34, "learning_rate": 4.334902088066645e-05, "loss": 1.8396, "step": 24790 }, { "epoch": 1.34, "learning_rate": 4.3346316131126256e-05, "loss": 1.7717, "step": 24800 }, { "epoch": 1.34, "learning_rate": 4.334361138158607e-05, "loss": 1.9056, "step": 24810 }, { "epoch": 1.34, "learning_rate": 4.3340906632045875e-05, "loss": 1.8798, "step": 24820 }, { "epoch": 1.34, "learning_rate": 4.333820188250568e-05, "loss": 1.8686, "step": 24830 }, { "epoch": 1.34, "learning_rate": 4.3335497132965494e-05, "loss": 1.7664, "step": 24840 }, { "epoch": 1.34, "learning_rate": 4.33327923834253e-05, "loss": 1.7958, "step": 24850 }, { "epoch": 1.34, "learning_rate": 4.33300876338851e-05, "loss": 1.8454, "step": 24860 }, { "epoch": 1.34, "learning_rate": 4.332738288434492e-05, "loss": 1.8595, "step": 24870 }, { "epoch": 1.34, "learning_rate": 4.332467813480472e-05, "loss": 1.9051, "step": 24880 }, { "epoch": 1.34, "learning_rate": 4.332197338526452e-05, "loss": 1.7837, "step": 24890 }, { "epoch": 1.35, "learning_rate": 4.331926863572434e-05, "loss": 1.8028, "step": 24900 }, { "epoch": 1.35, "learning_rate": 4.331656388618414e-05, "loss": 1.8868, "step": 24910 }, { "epoch": 1.35, "learning_rate": 4.3313859136643945e-05, "loss": 1.8832, "step": 24920 }, { "epoch": 1.35, "learning_rate": 4.331115438710376e-05, "loss": 1.7885, "step": 24930 }, { "epoch": 1.35, "learning_rate": 4.3308449637563564e-05, "loss": 1.7477, "step": 24940 }, { "epoch": 1.35, "learning_rate": 4.330574488802337e-05, "loss": 1.7608, "step": 24950 }, { "epoch": 1.35, "learning_rate": 4.330304013848318e-05, "loss": 1.8392, "step": 24960 }, { "epoch": 1.35, "learning_rate": 4.3300335388942986e-05, "loss": 1.7523, "step": 24970 }, { "epoch": 1.35, "learning_rate": 4.329763063940279e-05, "loss": 1.8073, "step": 24980 }, { "epoch": 1.35, "learning_rate": 4.3294925889862605e-05, "loss": 1.8947, "step": 24990 }, { "epoch": 1.35, "learning_rate": 4.329222114032241e-05, "loss": 1.8582, "step": 25000 }, { "epoch": 1.35, "learning_rate": 4.328951639078221e-05, "loss": 1.843, "step": 25010 }, { "epoch": 1.35, "learning_rate": 4.328681164124203e-05, "loss": 1.7566, "step": 25020 }, { "epoch": 1.35, "learning_rate": 4.328410689170183e-05, "loss": 1.8487, "step": 25030 }, { "epoch": 1.35, "learning_rate": 4.328140214216164e-05, "loss": 1.8676, "step": 25040 }, { "epoch": 1.35, "learning_rate": 4.327869739262145e-05, "loss": 1.7488, "step": 25050 }, { "epoch": 1.35, "learning_rate": 4.327599264308125e-05, "loss": 1.7983, "step": 25060 }, { "epoch": 1.35, "learning_rate": 4.327328789354106e-05, "loss": 1.7987, "step": 25070 }, { "epoch": 1.36, "learning_rate": 4.327058314400087e-05, "loss": 1.7695, "step": 25080 }, { "epoch": 1.36, "learning_rate": 4.3267878394460675e-05, "loss": 1.7883, "step": 25090 }, { "epoch": 1.36, "learning_rate": 4.3265173644920484e-05, "loss": 1.8739, "step": 25100 }, { "epoch": 1.36, "learning_rate": 4.3262468895380294e-05, "loss": 1.7775, "step": 25110 }, { "epoch": 1.36, "learning_rate": 4.32597641458401e-05, "loss": 1.8418, "step": 25120 }, { "epoch": 1.36, "learning_rate": 4.3257059396299907e-05, "loss": 1.8251, "step": 25130 }, { "epoch": 1.36, "learning_rate": 4.325435464675971e-05, "loss": 1.7773, "step": 25140 }, { "epoch": 1.36, "learning_rate": 4.325164989721952e-05, "loss": 1.799, "step": 25150 }, { "epoch": 1.36, "learning_rate": 4.324894514767933e-05, "loss": 1.7339, "step": 25160 }, { "epoch": 1.36, "learning_rate": 4.324624039813913e-05, "loss": 1.7933, "step": 25170 }, { "epoch": 1.36, "learning_rate": 4.324353564859894e-05, "loss": 1.7551, "step": 25180 }, { "epoch": 1.36, "learning_rate": 4.324083089905875e-05, "loss": 1.7609, "step": 25190 }, { "epoch": 1.36, "learning_rate": 4.3238126149518554e-05, "loss": 1.9142, "step": 25200 }, { "epoch": 1.36, "learning_rate": 4.323542139997836e-05, "loss": 1.7543, "step": 25210 }, { "epoch": 1.36, "learning_rate": 4.323271665043817e-05, "loss": 1.7523, "step": 25220 }, { "epoch": 1.36, "learning_rate": 4.3230011900897976e-05, "loss": 1.847, "step": 25230 }, { "epoch": 1.36, "learning_rate": 4.3227307151357786e-05, "loss": 1.8315, "step": 25240 }, { "epoch": 1.36, "learning_rate": 4.3224602401817595e-05, "loss": 1.9295, "step": 25250 }, { "epoch": 1.36, "learning_rate": 4.32218976522774e-05, "loss": 1.781, "step": 25260 }, { "epoch": 1.37, "learning_rate": 4.321919290273721e-05, "loss": 1.7896, "step": 25270 }, { "epoch": 1.37, "learning_rate": 4.321648815319702e-05, "loss": 1.7818, "step": 25280 }, { "epoch": 1.37, "learning_rate": 4.321378340365682e-05, "loss": 1.7644, "step": 25290 }, { "epoch": 1.37, "learning_rate": 4.321107865411663e-05, "loss": 1.8138, "step": 25300 }, { "epoch": 1.37, "learning_rate": 4.320837390457644e-05, "loss": 1.8716, "step": 25310 }, { "epoch": 1.37, "learning_rate": 4.320566915503624e-05, "loss": 1.7514, "step": 25320 }, { "epoch": 1.37, "learning_rate": 4.320296440549605e-05, "loss": 1.752, "step": 25330 }, { "epoch": 1.37, "learning_rate": 4.320025965595586e-05, "loss": 1.8523, "step": 25340 }, { "epoch": 1.37, "learning_rate": 4.3197554906415665e-05, "loss": 1.8373, "step": 25350 }, { "epoch": 1.37, "learning_rate": 4.3194850156875474e-05, "loss": 1.8276, "step": 25360 }, { "epoch": 1.37, "learning_rate": 4.3192145407335284e-05, "loss": 1.795, "step": 25370 }, { "epoch": 1.37, "learning_rate": 4.318944065779509e-05, "loss": 1.8357, "step": 25380 }, { "epoch": 1.37, "learning_rate": 4.3186735908254896e-05, "loss": 1.7688, "step": 25390 }, { "epoch": 1.37, "learning_rate": 4.3184031158714706e-05, "loss": 1.7735, "step": 25400 }, { "epoch": 1.37, "learning_rate": 4.318132640917451e-05, "loss": 1.7455, "step": 25410 }, { "epoch": 1.37, "learning_rate": 4.317862165963432e-05, "loss": 1.8202, "step": 25420 }, { "epoch": 1.37, "learning_rate": 4.317591691009413e-05, "loss": 1.7983, "step": 25430 }, { "epoch": 1.37, "learning_rate": 4.317321216055393e-05, "loss": 1.8653, "step": 25440 }, { "epoch": 1.38, "learning_rate": 4.317050741101375e-05, "loss": 1.7553, "step": 25450 }, { "epoch": 1.38, "learning_rate": 4.316780266147355e-05, "loss": 1.7919, "step": 25460 }, { "epoch": 1.38, "learning_rate": 4.316509791193335e-05, "loss": 1.8439, "step": 25470 }, { "epoch": 1.38, "learning_rate": 4.316239316239317e-05, "loss": 1.7773, "step": 25480 }, { "epoch": 1.38, "learning_rate": 4.315968841285297e-05, "loss": 1.7992, "step": 25490 }, { "epoch": 1.38, "learning_rate": 4.3156983663312775e-05, "loss": 1.7905, "step": 25500 }, { "epoch": 1.38, "learning_rate": 4.315427891377259e-05, "loss": 1.796, "step": 25510 }, { "epoch": 1.38, "learning_rate": 4.3151574164232395e-05, "loss": 1.8219, "step": 25520 }, { "epoch": 1.38, "learning_rate": 4.31488694146922e-05, "loss": 1.7309, "step": 25530 }, { "epoch": 1.38, "learning_rate": 4.3146164665152014e-05, "loss": 1.7995, "step": 25540 }, { "epoch": 1.38, "learning_rate": 4.314345991561182e-05, "loss": 1.7754, "step": 25550 }, { "epoch": 1.38, "learning_rate": 4.314075516607162e-05, "loss": 1.8212, "step": 25560 }, { "epoch": 1.38, "learning_rate": 4.3138050416531436e-05, "loss": 1.8179, "step": 25570 }, { "epoch": 1.38, "learning_rate": 4.313534566699124e-05, "loss": 1.8265, "step": 25580 }, { "epoch": 1.38, "learning_rate": 4.313264091745104e-05, "loss": 1.8062, "step": 25590 }, { "epoch": 1.38, "learning_rate": 4.312993616791086e-05, "loss": 1.7621, "step": 25600 }, { "epoch": 1.38, "learning_rate": 4.312723141837066e-05, "loss": 1.8153, "step": 25610 }, { "epoch": 1.38, "learning_rate": 4.3124526668830464e-05, "loss": 1.7862, "step": 25620 }, { "epoch": 1.38, "learning_rate": 4.312182191929028e-05, "loss": 1.7973, "step": 25630 }, { "epoch": 1.39, "learning_rate": 4.3119117169750083e-05, "loss": 1.8647, "step": 25640 }, { "epoch": 1.39, "learning_rate": 4.3116412420209886e-05, "loss": 1.8226, "step": 25650 }, { "epoch": 1.39, "learning_rate": 4.31137076706697e-05, "loss": 1.7139, "step": 25660 }, { "epoch": 1.39, "learning_rate": 4.3111002921129506e-05, "loss": 1.7783, "step": 25670 }, { "epoch": 1.39, "learning_rate": 4.3108298171589315e-05, "loss": 1.7471, "step": 25680 }, { "epoch": 1.39, "learning_rate": 4.310559342204912e-05, "loss": 1.7997, "step": 25690 }, { "epoch": 1.39, "learning_rate": 4.310288867250893e-05, "loss": 1.7781, "step": 25700 }, { "epoch": 1.39, "learning_rate": 4.310018392296874e-05, "loss": 1.8957, "step": 25710 }, { "epoch": 1.39, "learning_rate": 4.309747917342854e-05, "loss": 1.7679, "step": 25720 }, { "epoch": 1.39, "learning_rate": 4.309477442388835e-05, "loss": 1.8367, "step": 25730 }, { "epoch": 1.39, "learning_rate": 4.309206967434816e-05, "loss": 1.8079, "step": 25740 }, { "epoch": 1.39, "learning_rate": 4.308936492480796e-05, "loss": 1.7752, "step": 25750 }, { "epoch": 1.39, "learning_rate": 4.308666017526777e-05, "loss": 1.7553, "step": 25760 }, { "epoch": 1.39, "learning_rate": 4.308395542572758e-05, "loss": 1.7124, "step": 25770 }, { "epoch": 1.39, "learning_rate": 4.3081250676187385e-05, "loss": 1.716, "step": 25780 }, { "epoch": 1.39, "learning_rate": 4.3078545926647194e-05, "loss": 1.8705, "step": 25790 }, { "epoch": 1.39, "learning_rate": 4.3075841177107004e-05, "loss": 1.798, "step": 25800 }, { "epoch": 1.39, "learning_rate": 4.307313642756681e-05, "loss": 1.7675, "step": 25810 }, { "epoch": 1.4, "learning_rate": 4.3070431678026616e-05, "loss": 1.7797, "step": 25820 }, { "epoch": 1.4, "learning_rate": 4.3067726928486426e-05, "loss": 1.8122, "step": 25830 }, { "epoch": 1.4, "learning_rate": 4.306502217894623e-05, "loss": 1.8831, "step": 25840 }, { "epoch": 1.4, "learning_rate": 4.306231742940604e-05, "loss": 1.6971, "step": 25850 }, { "epoch": 1.4, "learning_rate": 4.305961267986585e-05, "loss": 1.7913, "step": 25860 }, { "epoch": 1.4, "learning_rate": 4.305690793032565e-05, "loss": 1.7214, "step": 25870 }, { "epoch": 1.4, "learning_rate": 4.305420318078546e-05, "loss": 1.8441, "step": 25880 }, { "epoch": 1.4, "learning_rate": 4.305149843124527e-05, "loss": 1.7268, "step": 25890 }, { "epoch": 1.4, "learning_rate": 4.304879368170507e-05, "loss": 1.8744, "step": 25900 }, { "epoch": 1.4, "learning_rate": 4.304608893216488e-05, "loss": 1.8715, "step": 25910 }, { "epoch": 1.4, "learning_rate": 4.304338418262469e-05, "loss": 1.7676, "step": 25920 }, { "epoch": 1.4, "learning_rate": 4.3040679433084495e-05, "loss": 1.8104, "step": 25930 }, { "epoch": 1.4, "learning_rate": 4.3037974683544305e-05, "loss": 1.8181, "step": 25940 }, { "epoch": 1.4, "learning_rate": 4.3035269934004115e-05, "loss": 1.7893, "step": 25950 }, { "epoch": 1.4, "learning_rate": 4.303256518446392e-05, "loss": 1.8328, "step": 25960 }, { "epoch": 1.4, "learning_rate": 4.302986043492373e-05, "loss": 1.8109, "step": 25970 }, { "epoch": 1.4, "learning_rate": 4.302715568538354e-05, "loss": 1.8302, "step": 25980 }, { "epoch": 1.4, "learning_rate": 4.302445093584334e-05, "loss": 1.8894, "step": 25990 }, { "epoch": 1.4, "learning_rate": 4.302174618630315e-05, "loss": 1.7739, "step": 26000 }, { "epoch": 1.41, "learning_rate": 4.301904143676296e-05, "loss": 1.8503, "step": 26010 }, { "epoch": 1.41, "learning_rate": 4.301633668722276e-05, "loss": 1.8454, "step": 26020 }, { "epoch": 1.41, "learning_rate": 4.301363193768257e-05, "loss": 1.8047, "step": 26030 }, { "epoch": 1.41, "learning_rate": 4.301092718814238e-05, "loss": 1.7573, "step": 26040 }, { "epoch": 1.41, "learning_rate": 4.3008222438602184e-05, "loss": 1.8293, "step": 26050 }, { "epoch": 1.41, "learning_rate": 4.3005517689061994e-05, "loss": 1.8464, "step": 26060 }, { "epoch": 1.41, "learning_rate": 4.3002812939521803e-05, "loss": 1.7654, "step": 26070 }, { "epoch": 1.41, "learning_rate": 4.3000108189981606e-05, "loss": 1.7958, "step": 26080 }, { "epoch": 1.41, "learning_rate": 4.299740344044142e-05, "loss": 1.8084, "step": 26090 }, { "epoch": 1.41, "learning_rate": 4.2994698690901226e-05, "loss": 1.7992, "step": 26100 }, { "epoch": 1.41, "learning_rate": 4.299199394136103e-05, "loss": 1.8003, "step": 26110 }, { "epoch": 1.41, "learning_rate": 4.2989289191820845e-05, "loss": 1.7218, "step": 26120 }, { "epoch": 1.41, "learning_rate": 4.298658444228065e-05, "loss": 1.7372, "step": 26130 }, { "epoch": 1.41, "learning_rate": 4.298387969274045e-05, "loss": 1.8166, "step": 26140 }, { "epoch": 1.41, "learning_rate": 4.298117494320027e-05, "loss": 1.8532, "step": 26150 }, { "epoch": 1.41, "learning_rate": 4.297847019366007e-05, "loss": 1.7899, "step": 26160 }, { "epoch": 1.41, "learning_rate": 4.297576544411987e-05, "loss": 1.7847, "step": 26170 }, { "epoch": 1.41, "learning_rate": 4.297306069457969e-05, "loss": 1.8337, "step": 26180 }, { "epoch": 1.42, "learning_rate": 4.297035594503949e-05, "loss": 1.8357, "step": 26190 }, { "epoch": 1.42, "learning_rate": 4.2967651195499295e-05, "loss": 1.858, "step": 26200 }, { "epoch": 1.42, "learning_rate": 4.296494644595911e-05, "loss": 1.8589, "step": 26210 }, { "epoch": 1.42, "learning_rate": 4.2962241696418914e-05, "loss": 1.8201, "step": 26220 }, { "epoch": 1.42, "learning_rate": 4.295953694687872e-05, "loss": 1.791, "step": 26230 }, { "epoch": 1.42, "learning_rate": 4.2956832197338534e-05, "loss": 1.8443, "step": 26240 }, { "epoch": 1.42, "learning_rate": 4.2954127447798336e-05, "loss": 1.7427, "step": 26250 }, { "epoch": 1.42, "learning_rate": 4.295142269825814e-05, "loss": 1.7925, "step": 26260 }, { "epoch": 1.42, "learning_rate": 4.294871794871795e-05, "loss": 1.6772, "step": 26270 }, { "epoch": 1.42, "learning_rate": 4.294601319917776e-05, "loss": 1.8306, "step": 26280 }, { "epoch": 1.42, "learning_rate": 4.294330844963756e-05, "loss": 1.7728, "step": 26290 }, { "epoch": 1.42, "learning_rate": 4.294060370009737e-05, "loss": 1.8274, "step": 26300 }, { "epoch": 1.42, "learning_rate": 4.293789895055718e-05, "loss": 1.7942, "step": 26310 }, { "epoch": 1.42, "learning_rate": 4.293519420101699e-05, "loss": 1.8183, "step": 26320 }, { "epoch": 1.42, "learning_rate": 4.293248945147679e-05, "loss": 1.7674, "step": 26330 }, { "epoch": 1.42, "learning_rate": 4.29297847019366e-05, "loss": 1.8147, "step": 26340 }, { "epoch": 1.42, "learning_rate": 4.292707995239641e-05, "loss": 1.7387, "step": 26350 }, { "epoch": 1.42, "learning_rate": 4.2924375202856215e-05, "loss": 1.783, "step": 26360 }, { "epoch": 1.42, "learning_rate": 4.2921670453316025e-05, "loss": 1.7813, "step": 26370 }, { "epoch": 1.43, "learning_rate": 4.2918965703775835e-05, "loss": 1.8189, "step": 26380 }, { "epoch": 1.43, "learning_rate": 4.291626095423564e-05, "loss": 1.733, "step": 26390 }, { "epoch": 1.43, "learning_rate": 4.291355620469545e-05, "loss": 1.8097, "step": 26400 }, { "epoch": 1.43, "learning_rate": 4.291085145515526e-05, "loss": 1.7548, "step": 26410 }, { "epoch": 1.43, "learning_rate": 4.290814670561506e-05, "loss": 1.7726, "step": 26420 }, { "epoch": 1.43, "learning_rate": 4.290544195607487e-05, "loss": 1.7407, "step": 26430 }, { "epoch": 1.43, "learning_rate": 4.290273720653468e-05, "loss": 1.8468, "step": 26440 }, { "epoch": 1.43, "learning_rate": 4.290003245699448e-05, "loss": 1.7331, "step": 26450 }, { "epoch": 1.43, "learning_rate": 4.289732770745429e-05, "loss": 1.7366, "step": 26460 }, { "epoch": 1.43, "learning_rate": 4.28946229579141e-05, "loss": 1.789, "step": 26470 }, { "epoch": 1.43, "learning_rate": 4.2891918208373904e-05, "loss": 1.8605, "step": 26480 }, { "epoch": 1.43, "learning_rate": 4.2889213458833714e-05, "loss": 1.7322, "step": 26490 }, { "epoch": 1.43, "learning_rate": 4.2886508709293523e-05, "loss": 1.8343, "step": 26500 }, { "epoch": 1.43, "learning_rate": 4.2883803959753326e-05, "loss": 1.7829, "step": 26510 }, { "epoch": 1.43, "learning_rate": 4.2881099210213136e-05, "loss": 1.816, "step": 26520 }, { "epoch": 1.43, "learning_rate": 4.2878394460672946e-05, "loss": 1.7982, "step": 26530 }, { "epoch": 1.43, "learning_rate": 4.287568971113275e-05, "loss": 1.6933, "step": 26540 }, { "epoch": 1.43, "learning_rate": 4.287298496159256e-05, "loss": 1.8138, "step": 26550 }, { "epoch": 1.44, "learning_rate": 4.287028021205237e-05, "loss": 1.7416, "step": 26560 }, { "epoch": 1.44, "learning_rate": 4.286757546251217e-05, "loss": 1.7415, "step": 26570 }, { "epoch": 1.44, "learning_rate": 4.286487071297198e-05, "loss": 1.8758, "step": 26580 }, { "epoch": 1.44, "learning_rate": 4.286216596343179e-05, "loss": 1.8158, "step": 26590 }, { "epoch": 1.44, "learning_rate": 4.285946121389159e-05, "loss": 1.705, "step": 26600 }, { "epoch": 1.44, "learning_rate": 4.28567564643514e-05, "loss": 1.7453, "step": 26610 }, { "epoch": 1.44, "learning_rate": 4.285405171481121e-05, "loss": 1.8709, "step": 26620 }, { "epoch": 1.44, "learning_rate": 4.2851346965271015e-05, "loss": 1.8695, "step": 26630 }, { "epoch": 1.44, "learning_rate": 4.2848642215730825e-05, "loss": 1.8249, "step": 26640 }, { "epoch": 1.44, "learning_rate": 4.2845937466190634e-05, "loss": 1.8233, "step": 26650 }, { "epoch": 1.44, "learning_rate": 4.284323271665044e-05, "loss": 1.7202, "step": 26660 }, { "epoch": 1.44, "learning_rate": 4.284052796711025e-05, "loss": 1.8777, "step": 26670 }, { "epoch": 1.44, "learning_rate": 4.2837823217570056e-05, "loss": 1.7974, "step": 26680 }, { "epoch": 1.44, "learning_rate": 4.283511846802986e-05, "loss": 1.8409, "step": 26690 }, { "epoch": 1.44, "learning_rate": 4.283241371848967e-05, "loss": 1.7757, "step": 26700 }, { "epoch": 1.44, "learning_rate": 4.282970896894948e-05, "loss": 1.7308, "step": 26710 }, { "epoch": 1.44, "learning_rate": 4.282700421940928e-05, "loss": 1.7625, "step": 26720 }, { "epoch": 1.44, "learning_rate": 4.28242994698691e-05, "loss": 1.8427, "step": 26730 }, { "epoch": 1.44, "learning_rate": 4.28215947203289e-05, "loss": 1.8164, "step": 26740 }, { "epoch": 1.45, "learning_rate": 4.2818889970788704e-05, "loss": 1.8214, "step": 26750 }, { "epoch": 1.45, "learning_rate": 4.281618522124852e-05, "loss": 1.8153, "step": 26760 }, { "epoch": 1.45, "learning_rate": 4.281348047170832e-05, "loss": 1.8552, "step": 26770 }, { "epoch": 1.45, "learning_rate": 4.2810775722168126e-05, "loss": 1.7684, "step": 26780 }, { "epoch": 1.45, "learning_rate": 4.280807097262794e-05, "loss": 1.7613, "step": 26790 }, { "epoch": 1.45, "learning_rate": 4.2805366223087745e-05, "loss": 1.7649, "step": 26800 }, { "epoch": 1.45, "learning_rate": 4.280266147354755e-05, "loss": 1.8773, "step": 26810 }, { "epoch": 1.45, "learning_rate": 4.279995672400736e-05, "loss": 1.8163, "step": 26820 }, { "epoch": 1.45, "learning_rate": 4.279725197446717e-05, "loss": 1.7888, "step": 26830 }, { "epoch": 1.45, "learning_rate": 4.279454722492697e-05, "loss": 1.7926, "step": 26840 }, { "epoch": 1.45, "learning_rate": 4.279184247538678e-05, "loss": 1.7831, "step": 26850 }, { "epoch": 1.45, "learning_rate": 4.278913772584659e-05, "loss": 1.7228, "step": 26860 }, { "epoch": 1.45, "learning_rate": 4.278643297630639e-05, "loss": 1.7922, "step": 26870 }, { "epoch": 1.45, "learning_rate": 4.27837282267662e-05, "loss": 1.73, "step": 26880 }, { "epoch": 1.45, "learning_rate": 4.278102347722601e-05, "loss": 1.7505, "step": 26890 }, { "epoch": 1.45, "learning_rate": 4.2778318727685815e-05, "loss": 1.8451, "step": 26900 }, { "epoch": 1.45, "learning_rate": 4.2775613978145624e-05, "loss": 1.7861, "step": 26910 }, { "epoch": 1.45, "learning_rate": 4.2772909228605434e-05, "loss": 1.8586, "step": 26920 }, { "epoch": 1.46, "learning_rate": 4.2770204479065243e-05, "loss": 1.7875, "step": 26930 }, { "epoch": 1.46, "learning_rate": 4.2767499729525046e-05, "loss": 1.7762, "step": 26940 }, { "epoch": 1.46, "learning_rate": 4.2764794979984856e-05, "loss": 1.8029, "step": 26950 }, { "epoch": 1.46, "learning_rate": 4.2762090230444666e-05, "loss": 1.8208, "step": 26960 }, { "epoch": 1.46, "learning_rate": 4.275938548090447e-05, "loss": 1.7545, "step": 26970 }, { "epoch": 1.46, "learning_rate": 4.275668073136428e-05, "loss": 1.8021, "step": 26980 }, { "epoch": 1.46, "learning_rate": 4.275397598182409e-05, "loss": 1.8775, "step": 26990 }, { "epoch": 1.46, "learning_rate": 4.275127123228389e-05, "loss": 1.8322, "step": 27000 }, { "epoch": 1.46, "learning_rate": 4.27485664827437e-05, "loss": 1.7976, "step": 27010 }, { "epoch": 1.46, "learning_rate": 4.274586173320351e-05, "loss": 1.7897, "step": 27020 }, { "epoch": 1.46, "learning_rate": 4.274315698366331e-05, "loss": 1.8077, "step": 27030 }, { "epoch": 1.46, "learning_rate": 4.274045223412312e-05, "loss": 1.7893, "step": 27040 }, { "epoch": 1.46, "learning_rate": 4.273774748458293e-05, "loss": 1.7707, "step": 27050 }, { "epoch": 1.46, "learning_rate": 4.2735042735042735e-05, "loss": 1.8413, "step": 27060 }, { "epoch": 1.46, "learning_rate": 4.2732337985502545e-05, "loss": 1.7753, "step": 27070 }, { "epoch": 1.46, "learning_rate": 4.2729633235962354e-05, "loss": 1.7352, "step": 27080 }, { "epoch": 1.46, "learning_rate": 4.272692848642216e-05, "loss": 1.8222, "step": 27090 }, { "epoch": 1.46, "learning_rate": 4.272422373688197e-05, "loss": 1.8695, "step": 27100 }, { "epoch": 1.46, "learning_rate": 4.2721518987341776e-05, "loss": 1.8694, "step": 27110 }, { "epoch": 1.47, "learning_rate": 4.271881423780158e-05, "loss": 1.8157, "step": 27120 }, { "epoch": 1.47, "learning_rate": 4.271610948826139e-05, "loss": 1.8206, "step": 27130 }, { "epoch": 1.47, "learning_rate": 4.27134047387212e-05, "loss": 1.7927, "step": 27140 }, { "epoch": 1.47, "learning_rate": 4.2710699989181e-05, "loss": 1.8053, "step": 27150 }, { "epoch": 1.47, "learning_rate": 4.270799523964081e-05, "loss": 1.7921, "step": 27160 }, { "epoch": 1.47, "learning_rate": 4.270529049010062e-05, "loss": 1.7874, "step": 27170 }, { "epoch": 1.47, "learning_rate": 4.2702585740560424e-05, "loss": 1.7665, "step": 27180 }, { "epoch": 1.47, "learning_rate": 4.269988099102023e-05, "loss": 1.8533, "step": 27190 }, { "epoch": 1.47, "learning_rate": 4.269717624148004e-05, "loss": 1.7208, "step": 27200 }, { "epoch": 1.47, "learning_rate": 4.2694471491939846e-05, "loss": 1.8544, "step": 27210 }, { "epoch": 1.47, "learning_rate": 4.2691766742399656e-05, "loss": 1.844, "step": 27220 }, { "epoch": 1.47, "learning_rate": 4.2689061992859465e-05, "loss": 1.7775, "step": 27230 }, { "epoch": 1.47, "learning_rate": 4.268635724331927e-05, "loss": 1.7674, "step": 27240 }, { "epoch": 1.47, "learning_rate": 4.268365249377908e-05, "loss": 1.7677, "step": 27250 }, { "epoch": 1.47, "learning_rate": 4.268094774423889e-05, "loss": 1.729, "step": 27260 }, { "epoch": 1.47, "learning_rate": 4.267824299469869e-05, "loss": 1.8071, "step": 27270 }, { "epoch": 1.47, "learning_rate": 4.26755382451585e-05, "loss": 1.7918, "step": 27280 }, { "epoch": 1.47, "learning_rate": 4.267283349561831e-05, "loss": 1.7105, "step": 27290 }, { "epoch": 1.48, "learning_rate": 4.267012874607811e-05, "loss": 1.7716, "step": 27300 }, { "epoch": 1.48, "learning_rate": 4.266742399653792e-05, "loss": 1.7892, "step": 27310 }, { "epoch": 1.48, "learning_rate": 4.266471924699773e-05, "loss": 1.8183, "step": 27320 }, { "epoch": 1.48, "learning_rate": 4.2662014497457535e-05, "loss": 1.768, "step": 27330 }, { "epoch": 1.48, "learning_rate": 4.265930974791735e-05, "loss": 1.8902, "step": 27340 }, { "epoch": 1.48, "learning_rate": 4.2656604998377154e-05, "loss": 1.7738, "step": 27350 }, { "epoch": 1.48, "learning_rate": 4.265390024883696e-05, "loss": 1.7503, "step": 27360 }, { "epoch": 1.48, "learning_rate": 4.2651195499296766e-05, "loss": 1.8892, "step": 27370 }, { "epoch": 1.48, "learning_rate": 4.2648490749756576e-05, "loss": 1.8247, "step": 27380 }, { "epoch": 1.48, "learning_rate": 4.264578600021638e-05, "loss": 1.8035, "step": 27390 }, { "epoch": 1.48, "learning_rate": 4.264308125067619e-05, "loss": 1.8029, "step": 27400 }, { "epoch": 1.48, "learning_rate": 4.2640376501136e-05, "loss": 1.68, "step": 27410 }, { "epoch": 1.48, "learning_rate": 4.26376717515958e-05, "loss": 1.7616, "step": 27420 }, { "epoch": 1.48, "learning_rate": 4.263496700205561e-05, "loss": 1.8569, "step": 27430 }, { "epoch": 1.48, "learning_rate": 4.263226225251542e-05, "loss": 1.7772, "step": 27440 }, { "epoch": 1.48, "learning_rate": 4.262955750297522e-05, "loss": 1.8629, "step": 27450 }, { "epoch": 1.48, "learning_rate": 4.262685275343503e-05, "loss": 1.7858, "step": 27460 }, { "epoch": 1.48, "learning_rate": 4.262414800389484e-05, "loss": 1.8101, "step": 27470 }, { "epoch": 1.48, "learning_rate": 4.2621443254354645e-05, "loss": 1.8441, "step": 27480 }, { "epoch": 1.49, "learning_rate": 4.2618738504814455e-05, "loss": 1.7696, "step": 27490 }, { "epoch": 1.49, "learning_rate": 4.2616033755274265e-05, "loss": 1.7238, "step": 27500 }, { "epoch": 1.49, "learning_rate": 4.261332900573407e-05, "loss": 1.7991, "step": 27510 }, { "epoch": 1.49, "learning_rate": 4.261062425619388e-05, "loss": 1.882, "step": 27520 }, { "epoch": 1.49, "learning_rate": 4.260791950665369e-05, "loss": 1.7181, "step": 27530 }, { "epoch": 1.49, "learning_rate": 4.260521475711349e-05, "loss": 1.7205, "step": 27540 }, { "epoch": 1.49, "learning_rate": 4.26025100075733e-05, "loss": 1.7924, "step": 27550 }, { "epoch": 1.49, "learning_rate": 4.259980525803311e-05, "loss": 1.8226, "step": 27560 }, { "epoch": 1.49, "learning_rate": 4.259710050849292e-05, "loss": 1.782, "step": 27570 }, { "epoch": 1.49, "learning_rate": 4.259439575895272e-05, "loss": 1.7809, "step": 27580 }, { "epoch": 1.49, "learning_rate": 4.259169100941253e-05, "loss": 1.7077, "step": 27590 }, { "epoch": 1.49, "learning_rate": 4.258898625987234e-05, "loss": 1.802, "step": 27600 }, { "epoch": 1.49, "learning_rate": 4.2586281510332144e-05, "loss": 1.846, "step": 27610 }, { "epoch": 1.49, "learning_rate": 4.258357676079195e-05, "loss": 1.7514, "step": 27620 }, { "epoch": 1.49, "learning_rate": 4.258087201125176e-05, "loss": 1.8047, "step": 27630 }, { "epoch": 1.49, "learning_rate": 4.2578167261711566e-05, "loss": 1.7161, "step": 27640 }, { "epoch": 1.49, "learning_rate": 4.2575462512171376e-05, "loss": 1.7841, "step": 27650 }, { "epoch": 1.49, "learning_rate": 4.2572757762631185e-05, "loss": 1.7708, "step": 27660 }, { "epoch": 1.5, "learning_rate": 4.257005301309099e-05, "loss": 1.8593, "step": 27670 }, { "epoch": 1.5, "learning_rate": 4.25673482635508e-05, "loss": 1.8387, "step": 27680 }, { "epoch": 1.5, "learning_rate": 4.256464351401061e-05, "loss": 1.8095, "step": 27690 }, { "epoch": 1.5, "learning_rate": 4.256193876447041e-05, "loss": 1.7955, "step": 27700 }, { "epoch": 1.5, "learning_rate": 4.255923401493022e-05, "loss": 1.6801, "step": 27710 }, { "epoch": 1.5, "learning_rate": 4.255652926539003e-05, "loss": 1.7577, "step": 27720 }, { "epoch": 1.5, "learning_rate": 4.255382451584983e-05, "loss": 1.7137, "step": 27730 }, { "epoch": 1.5, "learning_rate": 4.255111976630964e-05, "loss": 1.8108, "step": 27740 }, { "epoch": 1.5, "learning_rate": 4.254841501676945e-05, "loss": 1.7622, "step": 27750 }, { "epoch": 1.5, "learning_rate": 4.2545710267229255e-05, "loss": 1.7202, "step": 27760 }, { "epoch": 1.5, "learning_rate": 4.2543005517689064e-05, "loss": 1.7973, "step": 27770 }, { "epoch": 1.5, "learning_rate": 4.2540300768148874e-05, "loss": 1.7683, "step": 27780 }, { "epoch": 1.5, "learning_rate": 4.253759601860868e-05, "loss": 1.7631, "step": 27790 }, { "epoch": 1.5, "learning_rate": 4.2534891269068486e-05, "loss": 1.7742, "step": 27800 }, { "epoch": 1.5, "learning_rate": 4.2532186519528296e-05, "loss": 1.7437, "step": 27810 }, { "epoch": 1.5, "learning_rate": 4.25294817699881e-05, "loss": 1.7582, "step": 27820 }, { "epoch": 1.5, "learning_rate": 4.252677702044791e-05, "loss": 1.6821, "step": 27830 }, { "epoch": 1.5, "learning_rate": 4.252407227090772e-05, "loss": 1.8361, "step": 27840 }, { "epoch": 1.5, "learning_rate": 4.252136752136752e-05, "loss": 1.8548, "step": 27850 }, { "epoch": 1.51, "learning_rate": 4.251866277182733e-05, "loss": 1.7743, "step": 27860 }, { "epoch": 1.51, "learning_rate": 4.251595802228714e-05, "loss": 1.8148, "step": 27870 }, { "epoch": 1.51, "learning_rate": 4.251325327274694e-05, "loss": 1.7761, "step": 27880 }, { "epoch": 1.51, "learning_rate": 4.251054852320675e-05, "loss": 1.8671, "step": 27890 }, { "epoch": 1.51, "learning_rate": 4.250784377366656e-05, "loss": 1.6718, "step": 27900 }, { "epoch": 1.51, "learning_rate": 4.2505139024126365e-05, "loss": 1.7979, "step": 27910 }, { "epoch": 1.51, "learning_rate": 4.2502434274586175e-05, "loss": 1.7503, "step": 27920 }, { "epoch": 1.51, "learning_rate": 4.2499729525045985e-05, "loss": 1.8495, "step": 27930 }, { "epoch": 1.51, "learning_rate": 4.249702477550579e-05, "loss": 1.7682, "step": 27940 }, { "epoch": 1.51, "learning_rate": 4.24943200259656e-05, "loss": 1.8143, "step": 27950 }, { "epoch": 1.51, "learning_rate": 4.249161527642541e-05, "loss": 1.776, "step": 27960 }, { "epoch": 1.51, "learning_rate": 4.248891052688521e-05, "loss": 1.6605, "step": 27970 }, { "epoch": 1.51, "learning_rate": 4.248620577734502e-05, "loss": 1.8048, "step": 27980 }, { "epoch": 1.51, "learning_rate": 4.248350102780483e-05, "loss": 1.7777, "step": 27990 }, { "epoch": 1.51, "learning_rate": 4.248079627826463e-05, "loss": 1.7773, "step": 28000 }, { "epoch": 1.51, "learning_rate": 4.247809152872444e-05, "loss": 1.7414, "step": 28010 }, { "epoch": 1.51, "learning_rate": 4.247538677918425e-05, "loss": 1.7561, "step": 28020 }, { "epoch": 1.51, "learning_rate": 4.2472682029644054e-05, "loss": 1.8295, "step": 28030 }, { "epoch": 1.52, "learning_rate": 4.2469977280103864e-05, "loss": 1.76, "step": 28040 }, { "epoch": 1.52, "learning_rate": 4.246727253056367e-05, "loss": 1.7838, "step": 28050 }, { "epoch": 1.52, "learning_rate": 4.2464567781023476e-05, "loss": 1.7358, "step": 28060 }, { "epoch": 1.52, "learning_rate": 4.2461863031483286e-05, "loss": 1.7886, "step": 28070 }, { "epoch": 1.52, "learning_rate": 4.2459158281943096e-05, "loss": 1.7522, "step": 28080 }, { "epoch": 1.52, "learning_rate": 4.24564535324029e-05, "loss": 1.739, "step": 28090 }, { "epoch": 1.52, "learning_rate": 4.245374878286271e-05, "loss": 1.7946, "step": 28100 }, { "epoch": 1.52, "learning_rate": 4.245104403332252e-05, "loss": 1.7182, "step": 28110 }, { "epoch": 1.52, "learning_rate": 4.244833928378232e-05, "loss": 1.7587, "step": 28120 }, { "epoch": 1.52, "learning_rate": 4.244563453424213e-05, "loss": 1.7582, "step": 28130 }, { "epoch": 1.52, "learning_rate": 4.244292978470194e-05, "loss": 1.8424, "step": 28140 }, { "epoch": 1.52, "learning_rate": 4.244022503516174e-05, "loss": 1.812, "step": 28150 }, { "epoch": 1.52, "learning_rate": 4.243752028562155e-05, "loss": 1.7853, "step": 28160 }, { "epoch": 1.52, "learning_rate": 4.243481553608136e-05, "loss": 1.8224, "step": 28170 }, { "epoch": 1.52, "learning_rate": 4.2432110786541165e-05, "loss": 1.8037, "step": 28180 }, { "epoch": 1.52, "learning_rate": 4.2429406037000975e-05, "loss": 1.7633, "step": 28190 }, { "epoch": 1.52, "learning_rate": 4.2426701287460784e-05, "loss": 1.699, "step": 28200 }, { "epoch": 1.52, "learning_rate": 4.2423996537920594e-05, "loss": 1.9076, "step": 28210 }, { "epoch": 1.52, "learning_rate": 4.24212917883804e-05, "loss": 1.8331, "step": 28220 }, { "epoch": 1.53, "learning_rate": 4.2418587038840206e-05, "loss": 1.8286, "step": 28230 }, { "epoch": 1.53, "learning_rate": 4.2415882289300016e-05, "loss": 1.8147, "step": 28240 }, { "epoch": 1.53, "learning_rate": 4.241317753975982e-05, "loss": 1.6939, "step": 28250 }, { "epoch": 1.53, "learning_rate": 4.241047279021963e-05, "loss": 1.7745, "step": 28260 }, { "epoch": 1.53, "learning_rate": 4.240776804067944e-05, "loss": 1.7367, "step": 28270 }, { "epoch": 1.53, "learning_rate": 4.240506329113924e-05, "loss": 1.8044, "step": 28280 }, { "epoch": 1.53, "learning_rate": 4.240235854159905e-05, "loss": 1.7487, "step": 28290 }, { "epoch": 1.53, "learning_rate": 4.239965379205886e-05, "loss": 1.7694, "step": 28300 }, { "epoch": 1.53, "learning_rate": 4.239694904251866e-05, "loss": 1.7725, "step": 28310 }, { "epoch": 1.53, "learning_rate": 4.239424429297847e-05, "loss": 1.8865, "step": 28320 }, { "epoch": 1.53, "learning_rate": 4.239153954343828e-05, "loss": 1.792, "step": 28330 }, { "epoch": 1.53, "learning_rate": 4.2388834793898085e-05, "loss": 1.8165, "step": 28340 }, { "epoch": 1.53, "learning_rate": 4.2386130044357895e-05, "loss": 1.72, "step": 28350 }, { "epoch": 1.53, "learning_rate": 4.2383425294817705e-05, "loss": 1.7557, "step": 28360 }, { "epoch": 1.53, "learning_rate": 4.238072054527751e-05, "loss": 1.7722, "step": 28370 }, { "epoch": 1.53, "learning_rate": 4.237801579573732e-05, "loss": 1.7411, "step": 28380 }, { "epoch": 1.53, "learning_rate": 4.237531104619713e-05, "loss": 1.7064, "step": 28390 }, { "epoch": 1.53, "learning_rate": 4.237260629665693e-05, "loss": 1.7327, "step": 28400 }, { "epoch": 1.54, "learning_rate": 4.236990154711674e-05, "loss": 1.7348, "step": 28410 }, { "epoch": 1.54, "learning_rate": 4.236719679757655e-05, "loss": 1.7585, "step": 28420 }, { "epoch": 1.54, "learning_rate": 4.236449204803635e-05, "loss": 1.7173, "step": 28430 }, { "epoch": 1.54, "learning_rate": 4.236178729849616e-05, "loss": 1.7351, "step": 28440 }, { "epoch": 1.54, "learning_rate": 4.235908254895597e-05, "loss": 1.8263, "step": 28450 }, { "epoch": 1.54, "learning_rate": 4.2356377799415774e-05, "loss": 1.7592, "step": 28460 }, { "epoch": 1.54, "learning_rate": 4.2353673049875584e-05, "loss": 1.8248, "step": 28470 }, { "epoch": 1.54, "learning_rate": 4.2350968300335393e-05, "loss": 1.6761, "step": 28480 }, { "epoch": 1.54, "learning_rate": 4.2348263550795196e-05, "loss": 1.7476, "step": 28490 }, { "epoch": 1.54, "learning_rate": 4.2345558801255006e-05, "loss": 1.7517, "step": 28500 }, { "epoch": 1.54, "learning_rate": 4.2342854051714816e-05, "loss": 1.7797, "step": 28510 }, { "epoch": 1.54, "learning_rate": 4.234014930217462e-05, "loss": 1.793, "step": 28520 }, { "epoch": 1.54, "learning_rate": 4.233744455263443e-05, "loss": 1.8254, "step": 28530 }, { "epoch": 1.54, "learning_rate": 4.233473980309424e-05, "loss": 1.749, "step": 28540 }, { "epoch": 1.54, "learning_rate": 4.233203505355404e-05, "loss": 1.8162, "step": 28550 }, { "epoch": 1.54, "learning_rate": 4.232933030401385e-05, "loss": 1.8266, "step": 28560 }, { "epoch": 1.54, "learning_rate": 4.232662555447366e-05, "loss": 1.8235, "step": 28570 }, { "epoch": 1.54, "learning_rate": 4.232392080493346e-05, "loss": 1.8078, "step": 28580 }, { "epoch": 1.54, "learning_rate": 4.232121605539327e-05, "loss": 1.7657, "step": 28590 }, { "epoch": 1.55, "learning_rate": 4.231851130585308e-05, "loss": 1.8063, "step": 28600 }, { "epoch": 1.55, "learning_rate": 4.2315806556312885e-05, "loss": 1.7958, "step": 28610 }, { "epoch": 1.55, "learning_rate": 4.2313101806772695e-05, "loss": 1.853, "step": 28620 }, { "epoch": 1.55, "learning_rate": 4.2310397057232504e-05, "loss": 1.8214, "step": 28630 }, { "epoch": 1.55, "learning_rate": 4.230769230769231e-05, "loss": 1.6991, "step": 28640 }, { "epoch": 1.55, "learning_rate": 4.230498755815212e-05, "loss": 1.6887, "step": 28650 }, { "epoch": 1.55, "learning_rate": 4.2302282808611926e-05, "loss": 1.7517, "step": 28660 }, { "epoch": 1.55, "learning_rate": 4.229957805907173e-05, "loss": 1.8486, "step": 28670 }, { "epoch": 1.55, "learning_rate": 4.229687330953154e-05, "loss": 1.7885, "step": 28680 }, { "epoch": 1.55, "learning_rate": 4.229416855999135e-05, "loss": 1.882, "step": 28690 }, { "epoch": 1.55, "learning_rate": 4.229146381045115e-05, "loss": 1.7448, "step": 28700 }, { "epoch": 1.55, "learning_rate": 4.228875906091096e-05, "loss": 1.7248, "step": 28710 }, { "epoch": 1.55, "learning_rate": 4.228605431137077e-05, "loss": 1.7654, "step": 28720 }, { "epoch": 1.55, "learning_rate": 4.2283349561830574e-05, "loss": 1.708, "step": 28730 }, { "epoch": 1.55, "learning_rate": 4.228064481229038e-05, "loss": 1.7294, "step": 28740 }, { "epoch": 1.55, "learning_rate": 4.227794006275019e-05, "loss": 1.8229, "step": 28750 }, { "epoch": 1.55, "learning_rate": 4.2275235313209996e-05, "loss": 1.8324, "step": 28760 }, { "epoch": 1.55, "learning_rate": 4.2272530563669805e-05, "loss": 1.7189, "step": 28770 }, { "epoch": 1.56, "learning_rate": 4.2269825814129615e-05, "loss": 1.796, "step": 28780 }, { "epoch": 1.56, "learning_rate": 4.226712106458942e-05, "loss": 1.7253, "step": 28790 }, { "epoch": 1.56, "learning_rate": 4.226441631504923e-05, "loss": 1.8317, "step": 28800 }, { "epoch": 1.56, "learning_rate": 4.226171156550904e-05, "loss": 1.7804, "step": 28810 }, { "epoch": 1.56, "learning_rate": 4.225900681596884e-05, "loss": 1.8229, "step": 28820 }, { "epoch": 1.56, "learning_rate": 4.225630206642865e-05, "loss": 1.8242, "step": 28830 }, { "epoch": 1.56, "learning_rate": 4.225359731688846e-05, "loss": 1.655, "step": 28840 }, { "epoch": 1.56, "learning_rate": 4.225089256734827e-05, "loss": 1.8228, "step": 28850 }, { "epoch": 1.56, "learning_rate": 4.224818781780807e-05, "loss": 1.8296, "step": 28860 }, { "epoch": 1.56, "learning_rate": 4.224548306826788e-05, "loss": 1.7826, "step": 28870 }, { "epoch": 1.56, "learning_rate": 4.224277831872769e-05, "loss": 1.7789, "step": 28880 }, { "epoch": 1.56, "learning_rate": 4.2240073569187494e-05, "loss": 1.7799, "step": 28890 }, { "epoch": 1.56, "learning_rate": 4.2237368819647304e-05, "loss": 1.7935, "step": 28900 }, { "epoch": 1.56, "learning_rate": 4.2234664070107113e-05, "loss": 1.8269, "step": 28910 }, { "epoch": 1.56, "learning_rate": 4.2231959320566916e-05, "loss": 1.8344, "step": 28920 }, { "epoch": 1.56, "learning_rate": 4.2229254571026726e-05, "loss": 1.7595, "step": 28930 }, { "epoch": 1.56, "learning_rate": 4.2226549821486536e-05, "loss": 1.7909, "step": 28940 }, { "epoch": 1.56, "learning_rate": 4.222384507194634e-05, "loss": 1.8305, "step": 28950 }, { "epoch": 1.56, "learning_rate": 4.222114032240615e-05, "loss": 1.703, "step": 28960 }, { "epoch": 1.57, "learning_rate": 4.221843557286596e-05, "loss": 1.7775, "step": 28970 }, { "epoch": 1.57, "learning_rate": 4.221573082332576e-05, "loss": 1.7562, "step": 28980 }, { "epoch": 1.57, "learning_rate": 4.221302607378557e-05, "loss": 1.8558, "step": 28990 }, { "epoch": 1.57, "learning_rate": 4.221032132424538e-05, "loss": 1.6389, "step": 29000 }, { "epoch": 1.57, "learning_rate": 4.220761657470518e-05, "loss": 1.7387, "step": 29010 }, { "epoch": 1.57, "learning_rate": 4.220491182516499e-05, "loss": 1.8345, "step": 29020 }, { "epoch": 1.57, "learning_rate": 4.22022070756248e-05, "loss": 1.7552, "step": 29030 }, { "epoch": 1.57, "learning_rate": 4.2199502326084605e-05, "loss": 1.762, "step": 29040 }, { "epoch": 1.57, "learning_rate": 4.2196797576544415e-05, "loss": 1.7369, "step": 29050 }, { "epoch": 1.57, "learning_rate": 4.2194092827004224e-05, "loss": 1.8207, "step": 29060 }, { "epoch": 1.57, "learning_rate": 4.219138807746403e-05, "loss": 1.7252, "step": 29070 }, { "epoch": 1.57, "learning_rate": 4.218868332792384e-05, "loss": 1.8033, "step": 29080 }, { "epoch": 1.57, "learning_rate": 4.2185978578383646e-05, "loss": 1.8167, "step": 29090 }, { "epoch": 1.57, "learning_rate": 4.218327382884345e-05, "loss": 1.6849, "step": 29100 }, { "epoch": 1.57, "learning_rate": 4.218056907930326e-05, "loss": 1.7982, "step": 29110 }, { "epoch": 1.57, "learning_rate": 4.217786432976307e-05, "loss": 1.7859, "step": 29120 }, { "epoch": 1.57, "learning_rate": 4.217515958022287e-05, "loss": 1.6943, "step": 29130 }, { "epoch": 1.57, "learning_rate": 4.217245483068268e-05, "loss": 1.7107, "step": 29140 }, { "epoch": 1.58, "learning_rate": 4.216975008114249e-05, "loss": 1.8137, "step": 29150 }, { "epoch": 1.58, "learning_rate": 4.2167045331602294e-05, "loss": 1.6837, "step": 29160 }, { "epoch": 1.58, "learning_rate": 4.21643405820621e-05, "loss": 1.8436, "step": 29170 }, { "epoch": 1.58, "learning_rate": 4.216163583252191e-05, "loss": 1.7358, "step": 29180 }, { "epoch": 1.58, "learning_rate": 4.2158931082981716e-05, "loss": 1.7869, "step": 29190 }, { "epoch": 1.58, "learning_rate": 4.2156226333441525e-05, "loss": 1.8586, "step": 29200 }, { "epoch": 1.58, "learning_rate": 4.2153521583901335e-05, "loss": 1.7037, "step": 29210 }, { "epoch": 1.58, "learning_rate": 4.215081683436114e-05, "loss": 1.7384, "step": 29220 }, { "epoch": 1.58, "learning_rate": 4.214811208482095e-05, "loss": 1.8018, "step": 29230 }, { "epoch": 1.58, "learning_rate": 4.214540733528076e-05, "loss": 1.8084, "step": 29240 }, { "epoch": 1.58, "learning_rate": 4.214270258574056e-05, "loss": 1.7864, "step": 29250 }, { "epoch": 1.58, "learning_rate": 4.213999783620037e-05, "loss": 1.7943, "step": 29260 }, { "epoch": 1.58, "learning_rate": 4.213729308666018e-05, "loss": 1.785, "step": 29270 }, { "epoch": 1.58, "learning_rate": 4.213458833711998e-05, "loss": 1.838, "step": 29280 }, { "epoch": 1.58, "learning_rate": 4.213188358757979e-05, "loss": 1.7848, "step": 29290 }, { "epoch": 1.58, "learning_rate": 4.21291788380396e-05, "loss": 1.8272, "step": 29300 }, { "epoch": 1.58, "learning_rate": 4.2126474088499404e-05, "loss": 1.8353, "step": 29310 }, { "epoch": 1.58, "learning_rate": 4.2123769338959214e-05, "loss": 1.8381, "step": 29320 }, { "epoch": 1.58, "learning_rate": 4.2121064589419024e-05, "loss": 1.7778, "step": 29330 }, { "epoch": 1.59, "learning_rate": 4.211835983987883e-05, "loss": 1.8361, "step": 29340 }, { "epoch": 1.59, "learning_rate": 4.2115655090338636e-05, "loss": 1.7655, "step": 29350 }, { "epoch": 1.59, "learning_rate": 4.2112950340798446e-05, "loss": 1.7083, "step": 29360 }, { "epoch": 1.59, "learning_rate": 4.211024559125825e-05, "loss": 1.8188, "step": 29370 }, { "epoch": 1.59, "learning_rate": 4.210754084171806e-05, "loss": 1.8296, "step": 29380 }, { "epoch": 1.59, "learning_rate": 4.210483609217787e-05, "loss": 1.7621, "step": 29390 }, { "epoch": 1.59, "learning_rate": 4.210213134263767e-05, "loss": 1.8236, "step": 29400 }, { "epoch": 1.59, "learning_rate": 4.209942659309748e-05, "loss": 1.7411, "step": 29410 }, { "epoch": 1.59, "learning_rate": 4.209672184355729e-05, "loss": 1.8222, "step": 29420 }, { "epoch": 1.59, "learning_rate": 4.209401709401709e-05, "loss": 1.8495, "step": 29430 }, { "epoch": 1.59, "learning_rate": 4.20913123444769e-05, "loss": 1.804, "step": 29440 }, { "epoch": 1.59, "learning_rate": 4.208860759493671e-05, "loss": 1.8248, "step": 29450 }, { "epoch": 1.59, "learning_rate": 4.208590284539652e-05, "loss": 1.7167, "step": 29460 }, { "epoch": 1.59, "learning_rate": 4.2083198095856325e-05, "loss": 1.8455, "step": 29470 }, { "epoch": 1.59, "learning_rate": 4.2080493346316135e-05, "loss": 1.8632, "step": 29480 }, { "epoch": 1.59, "learning_rate": 4.2077788596775944e-05, "loss": 1.8277, "step": 29490 }, { "epoch": 1.59, "learning_rate": 4.207508384723575e-05, "loss": 1.7232, "step": 29500 }, { "epoch": 1.59, "learning_rate": 4.207237909769556e-05, "loss": 1.8079, "step": 29510 }, { "epoch": 1.6, "learning_rate": 4.2069674348155366e-05, "loss": 1.8186, "step": 29520 }, { "epoch": 1.6, "learning_rate": 4.206696959861517e-05, "loss": 1.7737, "step": 29530 }, { "epoch": 1.6, "learning_rate": 4.206426484907498e-05, "loss": 1.802, "step": 29540 }, { "epoch": 1.6, "learning_rate": 4.206156009953479e-05, "loss": 1.7939, "step": 29550 }, { "epoch": 1.6, "learning_rate": 4.205885534999459e-05, "loss": 1.7545, "step": 29560 }, { "epoch": 1.6, "learning_rate": 4.20561506004544e-05, "loss": 1.7883, "step": 29570 }, { "epoch": 1.6, "learning_rate": 4.205344585091421e-05, "loss": 1.8511, "step": 29580 }, { "epoch": 1.6, "learning_rate": 4.2050741101374014e-05, "loss": 1.7731, "step": 29590 }, { "epoch": 1.6, "learning_rate": 4.204803635183382e-05, "loss": 1.8436, "step": 29600 }, { "epoch": 1.6, "learning_rate": 4.204533160229363e-05, "loss": 1.8254, "step": 29610 }, { "epoch": 1.6, "learning_rate": 4.2042626852753436e-05, "loss": 1.7647, "step": 29620 }, { "epoch": 1.6, "learning_rate": 4.203992210321324e-05, "loss": 1.7909, "step": 29630 }, { "epoch": 1.6, "learning_rate": 4.2037217353673055e-05, "loss": 1.7965, "step": 29640 }, { "epoch": 1.6, "learning_rate": 4.203451260413286e-05, "loss": 1.7317, "step": 29650 }, { "epoch": 1.6, "learning_rate": 4.203180785459266e-05, "loss": 1.8247, "step": 29660 }, { "epoch": 1.6, "learning_rate": 4.202910310505248e-05, "loss": 1.6848, "step": 29670 }, { "epoch": 1.6, "learning_rate": 4.202639835551228e-05, "loss": 1.7834, "step": 29680 }, { "epoch": 1.6, "learning_rate": 4.202369360597209e-05, "loss": 1.8186, "step": 29690 }, { "epoch": 1.6, "learning_rate": 4.20209888564319e-05, "loss": 1.7769, "step": 29700 }, { "epoch": 1.61, "learning_rate": 4.20182841068917e-05, "loss": 1.7194, "step": 29710 }, { "epoch": 1.61, "learning_rate": 4.201557935735151e-05, "loss": 1.7704, "step": 29720 }, { "epoch": 1.61, "learning_rate": 4.201287460781132e-05, "loss": 1.8361, "step": 29730 }, { "epoch": 1.61, "learning_rate": 4.2010169858271125e-05, "loss": 1.6742, "step": 29740 }, { "epoch": 1.61, "learning_rate": 4.2007465108730934e-05, "loss": 1.8223, "step": 29750 }, { "epoch": 1.61, "learning_rate": 4.2004760359190744e-05, "loss": 1.8537, "step": 29760 }, { "epoch": 1.61, "learning_rate": 4.200205560965055e-05, "loss": 1.8135, "step": 29770 }, { "epoch": 1.61, "learning_rate": 4.1999350860110356e-05, "loss": 1.8009, "step": 29780 }, { "epoch": 1.61, "learning_rate": 4.1996646110570166e-05, "loss": 1.7125, "step": 29790 }, { "epoch": 1.61, "learning_rate": 4.199394136102997e-05, "loss": 1.7715, "step": 29800 }, { "epoch": 1.61, "learning_rate": 4.199123661148978e-05, "loss": 1.8967, "step": 29810 }, { "epoch": 1.61, "learning_rate": 4.198853186194959e-05, "loss": 1.8412, "step": 29820 }, { "epoch": 1.61, "learning_rate": 4.198582711240939e-05, "loss": 1.8363, "step": 29830 }, { "epoch": 1.61, "learning_rate": 4.19831223628692e-05, "loss": 1.7251, "step": 29840 }, { "epoch": 1.61, "learning_rate": 4.198041761332901e-05, "loss": 1.7675, "step": 29850 }, { "epoch": 1.61, "learning_rate": 4.197771286378881e-05, "loss": 1.7911, "step": 29860 }, { "epoch": 1.61, "learning_rate": 4.197500811424862e-05, "loss": 1.6857, "step": 29870 }, { "epoch": 1.61, "learning_rate": 4.197230336470843e-05, "loss": 1.8326, "step": 29880 }, { "epoch": 1.62, "learning_rate": 4.1969598615168235e-05, "loss": 1.8147, "step": 29890 }, { "epoch": 1.62, "learning_rate": 4.1966893865628045e-05, "loss": 1.7684, "step": 29900 }, { "epoch": 1.62, "learning_rate": 4.1964189116087855e-05, "loss": 1.7491, "step": 29910 }, { "epoch": 1.62, "learning_rate": 4.196148436654766e-05, "loss": 1.7836, "step": 29920 }, { "epoch": 1.62, "learning_rate": 4.195877961700747e-05, "loss": 1.7772, "step": 29930 }, { "epoch": 1.62, "learning_rate": 4.195607486746728e-05, "loss": 1.8432, "step": 29940 }, { "epoch": 1.62, "learning_rate": 4.195337011792708e-05, "loss": 1.7372, "step": 29950 }, { "epoch": 1.62, "learning_rate": 4.195066536838689e-05, "loss": 1.7503, "step": 29960 }, { "epoch": 1.62, "learning_rate": 4.19479606188467e-05, "loss": 1.7356, "step": 29970 }, { "epoch": 1.62, "learning_rate": 4.19452558693065e-05, "loss": 1.7719, "step": 29980 }, { "epoch": 1.62, "learning_rate": 4.194255111976631e-05, "loss": 1.7169, "step": 29990 }, { "epoch": 1.62, "learning_rate": 4.193984637022612e-05, "loss": 1.9235, "step": 30000 }, { "epoch": 1.62, "learning_rate": 4.1937141620685924e-05, "loss": 1.7416, "step": 30010 }, { "epoch": 1.62, "learning_rate": 4.1934436871145734e-05, "loss": 1.7291, "step": 30020 }, { "epoch": 1.62, "learning_rate": 4.193173212160554e-05, "loss": 1.7424, "step": 30030 }, { "epoch": 1.62, "learning_rate": 4.1929027372065346e-05, "loss": 1.8218, "step": 30040 }, { "epoch": 1.62, "learning_rate": 4.1926322622525156e-05, "loss": 1.8118, "step": 30050 }, { "epoch": 1.62, "learning_rate": 4.1923617872984966e-05, "loss": 1.7212, "step": 30060 }, { "epoch": 1.62, "learning_rate": 4.192091312344477e-05, "loss": 1.8016, "step": 30070 }, { "epoch": 1.63, "learning_rate": 4.191820837390458e-05, "loss": 1.7213, "step": 30080 }, { "epoch": 1.63, "learning_rate": 4.191550362436439e-05, "loss": 1.8216, "step": 30090 }, { "epoch": 1.63, "learning_rate": 4.19127988748242e-05, "loss": 1.7096, "step": 30100 }, { "epoch": 1.63, "learning_rate": 4.1910094125284e-05, "loss": 1.7464, "step": 30110 }, { "epoch": 1.63, "learning_rate": 4.190738937574381e-05, "loss": 1.8155, "step": 30120 }, { "epoch": 1.63, "learning_rate": 4.190468462620362e-05, "loss": 1.6517, "step": 30130 }, { "epoch": 1.63, "learning_rate": 4.190197987666342e-05, "loss": 1.8182, "step": 30140 }, { "epoch": 1.63, "learning_rate": 4.189927512712323e-05, "loss": 1.6959, "step": 30150 }, { "epoch": 1.63, "learning_rate": 4.189657037758304e-05, "loss": 1.7459, "step": 30160 }, { "epoch": 1.63, "learning_rate": 4.1893865628042845e-05, "loss": 1.8393, "step": 30170 }, { "epoch": 1.63, "learning_rate": 4.189116087850265e-05, "loss": 1.7622, "step": 30180 }, { "epoch": 1.63, "learning_rate": 4.1888456128962464e-05, "loss": 1.7874, "step": 30190 }, { "epoch": 1.63, "learning_rate": 4.188575137942227e-05, "loss": 1.7517, "step": 30200 }, { "epoch": 1.63, "learning_rate": 4.188304662988207e-05, "loss": 1.8498, "step": 30210 }, { "epoch": 1.63, "learning_rate": 4.1880341880341886e-05, "loss": 1.7124, "step": 30220 }, { "epoch": 1.63, "learning_rate": 4.187763713080169e-05, "loss": 1.8011, "step": 30230 }, { "epoch": 1.63, "learning_rate": 4.187493238126149e-05, "loss": 1.8344, "step": 30240 }, { "epoch": 1.63, "learning_rate": 4.187222763172131e-05, "loss": 1.7756, "step": 30250 }, { "epoch": 1.64, "learning_rate": 4.186952288218111e-05, "loss": 1.7974, "step": 30260 }, { "epoch": 1.64, "learning_rate": 4.1866818132640914e-05, "loss": 1.851, "step": 30270 }, { "epoch": 1.64, "learning_rate": 4.186411338310073e-05, "loss": 1.687, "step": 30280 }, { "epoch": 1.64, "learning_rate": 4.186140863356053e-05, "loss": 1.8427, "step": 30290 }, { "epoch": 1.64, "learning_rate": 4.1858703884020336e-05, "loss": 1.6794, "step": 30300 }, { "epoch": 1.64, "learning_rate": 4.185599913448015e-05, "loss": 1.8251, "step": 30310 }, { "epoch": 1.64, "learning_rate": 4.1853294384939955e-05, "loss": 1.8488, "step": 30320 }, { "epoch": 1.64, "learning_rate": 4.1850589635399765e-05, "loss": 1.7294, "step": 30330 }, { "epoch": 1.64, "learning_rate": 4.1847884885859575e-05, "loss": 1.7672, "step": 30340 }, { "epoch": 1.64, "learning_rate": 4.184518013631938e-05, "loss": 1.8108, "step": 30350 }, { "epoch": 1.64, "learning_rate": 4.184247538677919e-05, "loss": 1.6966, "step": 30360 }, { "epoch": 1.64, "learning_rate": 4.1839770637239e-05, "loss": 1.779, "step": 30370 }, { "epoch": 1.64, "learning_rate": 4.18370658876988e-05, "loss": 1.7361, "step": 30380 }, { "epoch": 1.64, "learning_rate": 4.183436113815861e-05, "loss": 1.7004, "step": 30390 }, { "epoch": 1.64, "learning_rate": 4.183165638861842e-05, "loss": 1.9005, "step": 30400 }, { "epoch": 1.64, "learning_rate": 4.182895163907822e-05, "loss": 1.7133, "step": 30410 }, { "epoch": 1.64, "learning_rate": 4.182624688953803e-05, "loss": 1.8297, "step": 30420 }, { "epoch": 1.64, "learning_rate": 4.182354213999784e-05, "loss": 1.8066, "step": 30430 }, { "epoch": 1.64, "learning_rate": 4.1820837390457644e-05, "loss": 1.7384, "step": 30440 }, { "epoch": 1.65, "learning_rate": 4.1818132640917454e-05, "loss": 1.7491, "step": 30450 }, { "epoch": 1.65, "learning_rate": 4.181542789137726e-05, "loss": 1.8107, "step": 30460 }, { "epoch": 1.65, "learning_rate": 4.1812723141837066e-05, "loss": 1.8046, "step": 30470 }, { "epoch": 1.65, "learning_rate": 4.1810018392296876e-05, "loss": 1.7516, "step": 30480 }, { "epoch": 1.65, "learning_rate": 4.1807313642756686e-05, "loss": 1.7865, "step": 30490 }, { "epoch": 1.65, "learning_rate": 4.180460889321649e-05, "loss": 1.7944, "step": 30500 }, { "epoch": 1.65, "learning_rate": 4.18019041436763e-05, "loss": 1.6689, "step": 30510 }, { "epoch": 1.65, "learning_rate": 4.179919939413611e-05, "loss": 1.7762, "step": 30520 }, { "epoch": 1.65, "learning_rate": 4.179649464459591e-05, "loss": 1.7457, "step": 30530 }, { "epoch": 1.65, "learning_rate": 4.179378989505572e-05, "loss": 1.7011, "step": 30540 }, { "epoch": 1.65, "learning_rate": 4.179108514551553e-05, "loss": 1.8064, "step": 30550 }, { "epoch": 1.65, "learning_rate": 4.178838039597533e-05, "loss": 1.7861, "step": 30560 }, { "epoch": 1.65, "learning_rate": 4.178567564643514e-05, "loss": 1.7786, "step": 30570 }, { "epoch": 1.65, "learning_rate": 4.178297089689495e-05, "loss": 1.7798, "step": 30580 }, { "epoch": 1.65, "learning_rate": 4.1780266147354755e-05, "loss": 1.7106, "step": 30590 }, { "epoch": 1.65, "learning_rate": 4.1777561397814565e-05, "loss": 1.806, "step": 30600 }, { "epoch": 1.65, "learning_rate": 4.1774856648274374e-05, "loss": 1.7653, "step": 30610 }, { "epoch": 1.65, "learning_rate": 4.177215189873418e-05, "loss": 1.7347, "step": 30620 }, { "epoch": 1.66, "learning_rate": 4.176944714919399e-05, "loss": 1.7691, "step": 30630 }, { "epoch": 1.66, "learning_rate": 4.1766742399653796e-05, "loss": 1.763, "step": 30640 }, { "epoch": 1.66, "learning_rate": 4.17640376501136e-05, "loss": 1.7698, "step": 30650 }, { "epoch": 1.66, "learning_rate": 4.176133290057341e-05, "loss": 1.7091, "step": 30660 }, { "epoch": 1.66, "learning_rate": 4.175862815103322e-05, "loss": 1.7431, "step": 30670 }, { "epoch": 1.66, "learning_rate": 4.175592340149302e-05, "loss": 1.797, "step": 30680 }, { "epoch": 1.66, "learning_rate": 4.175321865195283e-05, "loss": 1.8358, "step": 30690 }, { "epoch": 1.66, "learning_rate": 4.175051390241264e-05, "loss": 1.7728, "step": 30700 }, { "epoch": 1.66, "learning_rate": 4.1747809152872444e-05, "loss": 1.7916, "step": 30710 }, { "epoch": 1.66, "learning_rate": 4.174510440333225e-05, "loss": 1.8196, "step": 30720 }, { "epoch": 1.66, "learning_rate": 4.1742399653792056e-05, "loss": 1.7242, "step": 30730 }, { "epoch": 1.66, "learning_rate": 4.173969490425187e-05, "loss": 1.8173, "step": 30740 }, { "epoch": 1.66, "learning_rate": 4.1736990154711675e-05, "loss": 1.8235, "step": 30750 }, { "epoch": 1.66, "learning_rate": 4.173428540517148e-05, "loss": 1.8611, "step": 30760 }, { "epoch": 1.66, "learning_rate": 4.1731580655631295e-05, "loss": 1.8091, "step": 30770 }, { "epoch": 1.66, "learning_rate": 4.17288759060911e-05, "loss": 1.7983, "step": 30780 }, { "epoch": 1.66, "learning_rate": 4.17261711565509e-05, "loss": 1.7803, "step": 30790 }, { "epoch": 1.66, "learning_rate": 4.172346640701072e-05, "loss": 1.6926, "step": 30800 }, { "epoch": 1.66, "learning_rate": 4.172076165747052e-05, "loss": 1.7216, "step": 30810 }, { "epoch": 1.67, "learning_rate": 4.171805690793032e-05, "loss": 1.7085, "step": 30820 }, { "epoch": 1.67, "learning_rate": 4.171535215839014e-05, "loss": 1.7957, "step": 30830 }, { "epoch": 1.67, "learning_rate": 4.171264740884994e-05, "loss": 1.8633, "step": 30840 }, { "epoch": 1.67, "learning_rate": 4.1709942659309745e-05, "loss": 1.7355, "step": 30850 }, { "epoch": 1.67, "learning_rate": 4.170723790976956e-05, "loss": 1.6777, "step": 30860 }, { "epoch": 1.67, "learning_rate": 4.1704533160229364e-05, "loss": 1.7653, "step": 30870 }, { "epoch": 1.67, "learning_rate": 4.170182841068917e-05, "loss": 1.7798, "step": 30880 }, { "epoch": 1.67, "learning_rate": 4.169912366114898e-05, "loss": 1.7378, "step": 30890 }, { "epoch": 1.67, "learning_rate": 4.1696418911608786e-05, "loss": 1.7491, "step": 30900 }, { "epoch": 1.67, "learning_rate": 4.169371416206859e-05, "loss": 1.7917, "step": 30910 }, { "epoch": 1.67, "learning_rate": 4.1691009412528406e-05, "loss": 1.7217, "step": 30920 }, { "epoch": 1.67, "learning_rate": 4.168830466298821e-05, "loss": 1.8379, "step": 30930 }, { "epoch": 1.67, "learning_rate": 4.168559991344801e-05, "loss": 1.7323, "step": 30940 }, { "epoch": 1.67, "learning_rate": 4.168289516390783e-05, "loss": 1.7458, "step": 30950 }, { "epoch": 1.67, "learning_rate": 4.168019041436763e-05, "loss": 1.7874, "step": 30960 }, { "epoch": 1.67, "learning_rate": 4.167748566482744e-05, "loss": 1.7647, "step": 30970 }, { "epoch": 1.67, "learning_rate": 4.167478091528725e-05, "loss": 1.8838, "step": 30980 }, { "epoch": 1.67, "learning_rate": 4.167207616574705e-05, "loss": 1.7985, "step": 30990 }, { "epoch": 1.68, "learning_rate": 4.166937141620686e-05, "loss": 1.7563, "step": 31000 }, { "epoch": 1.68, "learning_rate": 4.166666666666667e-05, "loss": 1.8228, "step": 31010 }, { "epoch": 1.68, "learning_rate": 4.1663961917126475e-05, "loss": 1.7865, "step": 31020 }, { "epoch": 1.68, "learning_rate": 4.1661257167586285e-05, "loss": 1.7612, "step": 31030 }, { "epoch": 1.68, "learning_rate": 4.1658552418046094e-05, "loss": 1.7887, "step": 31040 }, { "epoch": 1.68, "learning_rate": 4.16558476685059e-05, "loss": 1.7813, "step": 31050 }, { "epoch": 1.68, "learning_rate": 4.165314291896571e-05, "loss": 1.9193, "step": 31060 }, { "epoch": 1.68, "learning_rate": 4.1650438169425516e-05, "loss": 1.8218, "step": 31070 }, { "epoch": 1.68, "learning_rate": 4.164773341988532e-05, "loss": 1.7355, "step": 31080 }, { "epoch": 1.68, "learning_rate": 4.164502867034513e-05, "loss": 1.7734, "step": 31090 }, { "epoch": 1.68, "learning_rate": 4.164232392080494e-05, "loss": 1.8306, "step": 31100 }, { "epoch": 1.68, "learning_rate": 4.163961917126474e-05, "loss": 1.6743, "step": 31110 }, { "epoch": 1.68, "learning_rate": 4.163691442172455e-05, "loss": 1.7052, "step": 31120 }, { "epoch": 1.68, "learning_rate": 4.163420967218436e-05, "loss": 1.7687, "step": 31130 }, { "epoch": 1.68, "learning_rate": 4.1631504922644164e-05, "loss": 1.6673, "step": 31140 }, { "epoch": 1.68, "learning_rate": 4.162880017310397e-05, "loss": 1.9127, "step": 31150 }, { "epoch": 1.68, "learning_rate": 4.162609542356378e-05, "loss": 1.7611, "step": 31160 }, { "epoch": 1.68, "learning_rate": 4.1623390674023586e-05, "loss": 1.7559, "step": 31170 }, { "epoch": 1.68, "learning_rate": 4.1620685924483395e-05, "loss": 1.8335, "step": 31180 }, { "epoch": 1.69, "learning_rate": 4.1617981174943205e-05, "loss": 1.8046, "step": 31190 }, { "epoch": 1.69, "learning_rate": 4.161527642540301e-05, "loss": 1.8207, "step": 31200 }, { "epoch": 1.69, "learning_rate": 4.161257167586282e-05, "loss": 1.8518, "step": 31210 }, { "epoch": 1.69, "learning_rate": 4.160986692632263e-05, "loss": 1.798, "step": 31220 }, { "epoch": 1.69, "learning_rate": 4.160716217678243e-05, "loss": 1.7738, "step": 31230 }, { "epoch": 1.69, "learning_rate": 4.160445742724224e-05, "loss": 1.778, "step": 31240 }, { "epoch": 1.69, "learning_rate": 4.160175267770205e-05, "loss": 1.7635, "step": 31250 }, { "epoch": 1.69, "learning_rate": 4.159904792816185e-05, "loss": 1.8607, "step": 31260 }, { "epoch": 1.69, "learning_rate": 4.159634317862166e-05, "loss": 1.8198, "step": 31270 }, { "epoch": 1.69, "learning_rate": 4.159363842908147e-05, "loss": 1.889, "step": 31280 }, { "epoch": 1.69, "learning_rate": 4.1590933679541274e-05, "loss": 1.7612, "step": 31290 }, { "epoch": 1.69, "learning_rate": 4.1588228930001084e-05, "loss": 1.7738, "step": 31300 }, { "epoch": 1.69, "learning_rate": 4.158552418046089e-05, "loss": 1.7913, "step": 31310 }, { "epoch": 1.69, "learning_rate": 4.15828194309207e-05, "loss": 1.7981, "step": 31320 }, { "epoch": 1.69, "learning_rate": 4.1580114681380506e-05, "loss": 1.7235, "step": 31330 }, { "epoch": 1.69, "learning_rate": 4.157740993184031e-05, "loss": 1.7156, "step": 31340 }, { "epoch": 1.69, "learning_rate": 4.157470518230012e-05, "loss": 1.7991, "step": 31350 }, { "epoch": 1.69, "learning_rate": 4.157200043275993e-05, "loss": 1.7662, "step": 31360 }, { "epoch": 1.7, "learning_rate": 4.156929568321973e-05, "loss": 1.7674, "step": 31370 }, { "epoch": 1.7, "learning_rate": 4.156659093367955e-05, "loss": 1.7639, "step": 31380 }, { "epoch": 1.7, "learning_rate": 4.156388618413935e-05, "loss": 1.6977, "step": 31390 }, { "epoch": 1.7, "learning_rate": 4.1561181434599153e-05, "loss": 1.8024, "step": 31400 }, { "epoch": 1.7, "learning_rate": 4.155847668505897e-05, "loss": 1.7894, "step": 31410 }, { "epoch": 1.7, "learning_rate": 4.155577193551877e-05, "loss": 1.8386, "step": 31420 }, { "epoch": 1.7, "learning_rate": 4.1553067185978576e-05, "loss": 1.8431, "step": 31430 }, { "epoch": 1.7, "learning_rate": 4.155036243643839e-05, "loss": 1.7635, "step": 31440 }, { "epoch": 1.7, "learning_rate": 4.1547657686898195e-05, "loss": 1.748, "step": 31450 }, { "epoch": 1.7, "learning_rate": 4.1544952937358e-05, "loss": 1.7615, "step": 31460 }, { "epoch": 1.7, "learning_rate": 4.1542248187817814e-05, "loss": 1.7095, "step": 31470 }, { "epoch": 1.7, "learning_rate": 4.153954343827762e-05, "loss": 1.6956, "step": 31480 }, { "epoch": 1.7, "learning_rate": 4.153683868873742e-05, "loss": 1.7742, "step": 31490 }, { "epoch": 1.7, "learning_rate": 4.1534133939197236e-05, "loss": 1.7408, "step": 31500 }, { "epoch": 1.7, "learning_rate": 4.153142918965704e-05, "loss": 1.7981, "step": 31510 }, { "epoch": 1.7, "learning_rate": 4.152872444011684e-05, "loss": 1.8211, "step": 31520 }, { "epoch": 1.7, "learning_rate": 4.152601969057666e-05, "loss": 1.8189, "step": 31530 }, { "epoch": 1.7, "learning_rate": 4.152331494103646e-05, "loss": 1.782, "step": 31540 }, { "epoch": 1.7, "learning_rate": 4.1520610191496264e-05, "loss": 1.6806, "step": 31550 }, { "epoch": 1.71, "learning_rate": 4.151790544195608e-05, "loss": 1.7582, "step": 31560 }, { "epoch": 1.71, "learning_rate": 4.1515200692415884e-05, "loss": 1.7697, "step": 31570 }, { "epoch": 1.71, "learning_rate": 4.1512495942875686e-05, "loss": 1.7702, "step": 31580 }, { "epoch": 1.71, "learning_rate": 4.15097911933355e-05, "loss": 1.7807, "step": 31590 }, { "epoch": 1.71, "learning_rate": 4.1507086443795306e-05, "loss": 1.6737, "step": 31600 }, { "epoch": 1.71, "learning_rate": 4.1504381694255115e-05, "loss": 1.8135, "step": 31610 }, { "epoch": 1.71, "learning_rate": 4.1501676944714925e-05, "loss": 1.8494, "step": 31620 }, { "epoch": 1.71, "learning_rate": 4.149897219517473e-05, "loss": 1.7931, "step": 31630 }, { "epoch": 1.71, "learning_rate": 4.149626744563454e-05, "loss": 1.7802, "step": 31640 }, { "epoch": 1.71, "learning_rate": 4.149356269609435e-05, "loss": 1.7727, "step": 31650 }, { "epoch": 1.71, "learning_rate": 4.149085794655415e-05, "loss": 1.7904, "step": 31660 }, { "epoch": 1.71, "learning_rate": 4.148815319701396e-05, "loss": 1.8456, "step": 31670 }, { "epoch": 1.71, "learning_rate": 4.148544844747377e-05, "loss": 1.8129, "step": 31680 }, { "epoch": 1.71, "learning_rate": 4.148274369793357e-05, "loss": 1.8329, "step": 31690 }, { "epoch": 1.71, "learning_rate": 4.148003894839338e-05, "loss": 1.7476, "step": 31700 }, { "epoch": 1.71, "learning_rate": 4.147733419885319e-05, "loss": 1.7139, "step": 31710 }, { "epoch": 1.71, "learning_rate": 4.1474629449312994e-05, "loss": 1.756, "step": 31720 }, { "epoch": 1.71, "learning_rate": 4.1471924699772804e-05, "loss": 1.8238, "step": 31730 }, { "epoch": 1.72, "learning_rate": 4.1469219950232614e-05, "loss": 1.7451, "step": 31740 }, { "epoch": 1.72, "learning_rate": 4.146651520069242e-05, "loss": 1.7373, "step": 31750 }, { "epoch": 1.72, "learning_rate": 4.1463810451152226e-05, "loss": 1.7775, "step": 31760 }, { "epoch": 1.72, "learning_rate": 4.1461105701612036e-05, "loss": 1.8104, "step": 31770 }, { "epoch": 1.72, "learning_rate": 4.145840095207184e-05, "loss": 1.7726, "step": 31780 }, { "epoch": 1.72, "learning_rate": 4.145569620253165e-05, "loss": 1.7208, "step": 31790 }, { "epoch": 1.72, "learning_rate": 4.145299145299146e-05, "loss": 1.7303, "step": 31800 }, { "epoch": 1.72, "learning_rate": 4.145028670345126e-05, "loss": 1.786, "step": 31810 }, { "epoch": 1.72, "learning_rate": 4.144758195391107e-05, "loss": 1.7113, "step": 31820 }, { "epoch": 1.72, "learning_rate": 4.144487720437088e-05, "loss": 1.7596, "step": 31830 }, { "epoch": 1.72, "learning_rate": 4.144217245483068e-05, "loss": 1.7537, "step": 31840 }, { "epoch": 1.72, "learning_rate": 4.143946770529049e-05, "loss": 1.705, "step": 31850 }, { "epoch": 1.72, "learning_rate": 4.1436762955750296e-05, "loss": 1.7269, "step": 31860 }, { "epoch": 1.72, "learning_rate": 4.1434058206210105e-05, "loss": 1.7454, "step": 31870 }, { "epoch": 1.72, "learning_rate": 4.1431353456669915e-05, "loss": 1.8024, "step": 31880 }, { "epoch": 1.72, "learning_rate": 4.142864870712972e-05, "loss": 1.7936, "step": 31890 }, { "epoch": 1.72, "learning_rate": 4.142594395758953e-05, "loss": 1.7452, "step": 31900 }, { "epoch": 1.72, "learning_rate": 4.142323920804934e-05, "loss": 1.7686, "step": 31910 }, { "epoch": 1.72, "learning_rate": 4.142053445850914e-05, "loss": 1.7316, "step": 31920 }, { "epoch": 1.73, "learning_rate": 4.141782970896895e-05, "loss": 1.8006, "step": 31930 }, { "epoch": 1.73, "learning_rate": 4.141512495942876e-05, "loss": 1.7356, "step": 31940 }, { "epoch": 1.73, "learning_rate": 4.141242020988856e-05, "loss": 1.8324, "step": 31950 }, { "epoch": 1.73, "learning_rate": 4.140971546034837e-05, "loss": 1.7825, "step": 31960 }, { "epoch": 1.73, "learning_rate": 4.140701071080818e-05, "loss": 1.706, "step": 31970 }, { "epoch": 1.73, "learning_rate": 4.1404305961267984e-05, "loss": 1.792, "step": 31980 }, { "epoch": 1.73, "learning_rate": 4.1401601211727794e-05, "loss": 1.7973, "step": 31990 }, { "epoch": 1.73, "learning_rate": 4.1398896462187604e-05, "loss": 1.7413, "step": 32000 }, { "epoch": 1.73, "learning_rate": 4.1396191712647407e-05, "loss": 1.6928, "step": 32010 }, { "epoch": 1.73, "learning_rate": 4.139348696310722e-05, "loss": 1.8455, "step": 32020 }, { "epoch": 1.73, "learning_rate": 4.1390782213567026e-05, "loss": 1.8033, "step": 32030 }, { "epoch": 1.73, "learning_rate": 4.138807746402683e-05, "loss": 1.7846, "step": 32040 }, { "epoch": 1.73, "learning_rate": 4.1385372714486645e-05, "loss": 1.7638, "step": 32050 }, { "epoch": 1.73, "learning_rate": 4.138266796494645e-05, "loss": 1.756, "step": 32060 }, { "epoch": 1.73, "learning_rate": 4.137996321540625e-05, "loss": 1.7433, "step": 32070 }, { "epoch": 1.73, "learning_rate": 4.137725846586607e-05, "loss": 1.7368, "step": 32080 }, { "epoch": 1.73, "learning_rate": 4.137455371632587e-05, "loss": 1.7818, "step": 32090 }, { "epoch": 1.73, "learning_rate": 4.137184896678567e-05, "loss": 1.7216, "step": 32100 }, { "epoch": 1.74, "learning_rate": 4.136914421724549e-05, "loss": 1.7312, "step": 32110 }, { "epoch": 1.74, "learning_rate": 4.136643946770529e-05, "loss": 1.7422, "step": 32120 }, { "epoch": 1.74, "learning_rate": 4.1363734718165095e-05, "loss": 1.8134, "step": 32130 }, { "epoch": 1.74, "learning_rate": 4.136102996862491e-05, "loss": 1.8194, "step": 32140 }, { "epoch": 1.74, "learning_rate": 4.1358325219084714e-05, "loss": 1.844, "step": 32150 }, { "epoch": 1.74, "learning_rate": 4.135562046954452e-05, "loss": 1.8354, "step": 32160 }, { "epoch": 1.74, "learning_rate": 4.1352915720004334e-05, "loss": 1.7905, "step": 32170 }, { "epoch": 1.74, "learning_rate": 4.135021097046414e-05, "loss": 1.8444, "step": 32180 }, { "epoch": 1.74, "learning_rate": 4.134750622092394e-05, "loss": 1.6947, "step": 32190 }, { "epoch": 1.74, "learning_rate": 4.1344801471383756e-05, "loss": 1.74, "step": 32200 }, { "epoch": 1.74, "learning_rate": 4.134209672184356e-05, "loss": 1.7552, "step": 32210 }, { "epoch": 1.74, "learning_rate": 4.133939197230337e-05, "loss": 1.7216, "step": 32220 }, { "epoch": 1.74, "learning_rate": 4.133668722276318e-05, "loss": 1.6958, "step": 32230 }, { "epoch": 1.74, "learning_rate": 4.133398247322298e-05, "loss": 1.8348, "step": 32240 }, { "epoch": 1.74, "learning_rate": 4.133127772368279e-05, "loss": 1.7296, "step": 32250 }, { "epoch": 1.74, "learning_rate": 4.13285729741426e-05, "loss": 1.7793, "step": 32260 }, { "epoch": 1.74, "learning_rate": 4.13258682246024e-05, "loss": 1.7479, "step": 32270 }, { "epoch": 1.74, "learning_rate": 4.132316347506221e-05, "loss": 1.86, "step": 32280 }, { "epoch": 1.74, "learning_rate": 4.132045872552202e-05, "loss": 1.7786, "step": 32290 }, { "epoch": 1.75, "learning_rate": 4.1317753975981825e-05, "loss": 1.7913, "step": 32300 }, { "epoch": 1.75, "learning_rate": 4.1315049226441635e-05, "loss": 1.706, "step": 32310 }, { "epoch": 1.75, "learning_rate": 4.1312344476901445e-05, "loss": 1.7215, "step": 32320 }, { "epoch": 1.75, "learning_rate": 4.130963972736125e-05, "loss": 1.7522, "step": 32330 }, { "epoch": 1.75, "learning_rate": 4.130693497782106e-05, "loss": 1.7657, "step": 32340 }, { "epoch": 1.75, "learning_rate": 4.130423022828087e-05, "loss": 1.7285, "step": 32350 }, { "epoch": 1.75, "learning_rate": 4.130152547874067e-05, "loss": 1.7892, "step": 32360 }, { "epoch": 1.75, "learning_rate": 4.129882072920048e-05, "loss": 1.7816, "step": 32370 }, { "epoch": 1.75, "learning_rate": 4.129611597966029e-05, "loss": 1.7829, "step": 32380 }, { "epoch": 1.75, "learning_rate": 4.129341123012009e-05, "loss": 1.6789, "step": 32390 }, { "epoch": 1.75, "learning_rate": 4.12907064805799e-05, "loss": 1.7488, "step": 32400 }, { "epoch": 1.75, "learning_rate": 4.128800173103971e-05, "loss": 1.8124, "step": 32410 }, { "epoch": 1.75, "learning_rate": 4.1285296981499514e-05, "loss": 1.8243, "step": 32420 }, { "epoch": 1.75, "learning_rate": 4.1282592231959324e-05, "loss": 1.8369, "step": 32430 }, { "epoch": 1.75, "learning_rate": 4.1279887482419127e-05, "loss": 1.7555, "step": 32440 }, { "epoch": 1.75, "learning_rate": 4.1277182732878936e-05, "loss": 1.7942, "step": 32450 }, { "epoch": 1.75, "learning_rate": 4.1274477983338746e-05, "loss": 1.7517, "step": 32460 }, { "epoch": 1.75, "learning_rate": 4.127177323379855e-05, "loss": 1.7385, "step": 32470 }, { "epoch": 1.76, "learning_rate": 4.126906848425836e-05, "loss": 1.8625, "step": 32480 }, { "epoch": 1.76, "learning_rate": 4.126636373471817e-05, "loss": 1.7899, "step": 32490 }, { "epoch": 1.76, "learning_rate": 4.126365898517797e-05, "loss": 1.7248, "step": 32500 }, { "epoch": 1.76, "learning_rate": 4.126095423563778e-05, "loss": 1.7663, "step": 32510 }, { "epoch": 1.76, "learning_rate": 4.125824948609759e-05, "loss": 1.7091, "step": 32520 }, { "epoch": 1.76, "learning_rate": 4.125554473655739e-05, "loss": 1.7569, "step": 32530 }, { "epoch": 1.76, "learning_rate": 4.12528399870172e-05, "loss": 1.751, "step": 32540 }, { "epoch": 1.76, "learning_rate": 4.125013523747701e-05, "loss": 1.7505, "step": 32550 }, { "epoch": 1.76, "learning_rate": 4.1247430487936815e-05, "loss": 1.7621, "step": 32560 }, { "epoch": 1.76, "learning_rate": 4.1244725738396625e-05, "loss": 1.7861, "step": 32570 }, { "epoch": 1.76, "learning_rate": 4.1242020988856435e-05, "loss": 1.7313, "step": 32580 }, { "epoch": 1.76, "learning_rate": 4.123931623931624e-05, "loss": 1.7525, "step": 32590 }, { "epoch": 1.76, "learning_rate": 4.123661148977605e-05, "loss": 1.7181, "step": 32600 }, { "epoch": 1.76, "learning_rate": 4.123390674023586e-05, "loss": 1.8265, "step": 32610 }, { "epoch": 1.76, "learning_rate": 4.123120199069566e-05, "loss": 1.7713, "step": 32620 }, { "epoch": 1.76, "learning_rate": 4.1228497241155476e-05, "loss": 1.7805, "step": 32630 }, { "epoch": 1.76, "learning_rate": 4.122579249161528e-05, "loss": 1.7135, "step": 32640 }, { "epoch": 1.76, "learning_rate": 4.122308774207508e-05, "loss": 1.6435, "step": 32650 }, { "epoch": 1.76, "learning_rate": 4.12203829925349e-05, "loss": 1.81, "step": 32660 }, { "epoch": 1.77, "learning_rate": 4.12176782429947e-05, "loss": 1.7809, "step": 32670 }, { "epoch": 1.77, "learning_rate": 4.1214973493454504e-05, "loss": 1.6973, "step": 32680 }, { "epoch": 1.77, "learning_rate": 4.121226874391432e-05, "loss": 1.8232, "step": 32690 }, { "epoch": 1.77, "learning_rate": 4.120956399437412e-05, "loss": 1.7943, "step": 32700 }, { "epoch": 1.77, "learning_rate": 4.1206859244833926e-05, "loss": 1.9048, "step": 32710 }, { "epoch": 1.77, "learning_rate": 4.120415449529374e-05, "loss": 1.8364, "step": 32720 }, { "epoch": 1.77, "learning_rate": 4.1201449745753545e-05, "loss": 1.8091, "step": 32730 }, { "epoch": 1.77, "learning_rate": 4.119874499621335e-05, "loss": 1.7565, "step": 32740 }, { "epoch": 1.77, "learning_rate": 4.1196040246673165e-05, "loss": 1.7527, "step": 32750 }, { "epoch": 1.77, "learning_rate": 4.119333549713297e-05, "loss": 1.7981, "step": 32760 }, { "epoch": 1.77, "learning_rate": 4.119063074759277e-05, "loss": 1.8435, "step": 32770 }, { "epoch": 1.77, "learning_rate": 4.118792599805259e-05, "loss": 1.7177, "step": 32780 }, { "epoch": 1.77, "learning_rate": 4.118522124851239e-05, "loss": 1.8555, "step": 32790 }, { "epoch": 1.77, "learning_rate": 4.118251649897219e-05, "loss": 1.8112, "step": 32800 }, { "epoch": 1.77, "learning_rate": 4.117981174943201e-05, "loss": 1.7918, "step": 32810 }, { "epoch": 1.77, "learning_rate": 4.117710699989181e-05, "loss": 1.7148, "step": 32820 }, { "epoch": 1.77, "learning_rate": 4.1174402250351615e-05, "loss": 1.7561, "step": 32830 }, { "epoch": 1.77, "learning_rate": 4.117169750081143e-05, "loss": 1.7724, "step": 32840 }, { "epoch": 1.78, "learning_rate": 4.1168992751271234e-05, "loss": 1.7972, "step": 32850 }, { "epoch": 1.78, "learning_rate": 4.1166288001731044e-05, "loss": 1.6741, "step": 32860 }, { "epoch": 1.78, "learning_rate": 4.116358325219085e-05, "loss": 1.7302, "step": 32870 }, { "epoch": 1.78, "learning_rate": 4.1160878502650656e-05, "loss": 1.8317, "step": 32880 }, { "epoch": 1.78, "learning_rate": 4.1158173753110466e-05, "loss": 1.826, "step": 32890 }, { "epoch": 1.78, "learning_rate": 4.1155469003570276e-05, "loss": 1.757, "step": 32900 }, { "epoch": 1.78, "learning_rate": 4.115276425403008e-05, "loss": 1.7152, "step": 32910 }, { "epoch": 1.78, "learning_rate": 4.115005950448989e-05, "loss": 1.7829, "step": 32920 }, { "epoch": 1.78, "learning_rate": 4.11473547549497e-05, "loss": 1.8613, "step": 32930 }, { "epoch": 1.78, "learning_rate": 4.11446500054095e-05, "loss": 1.8218, "step": 32940 }, { "epoch": 1.78, "learning_rate": 4.114194525586931e-05, "loss": 1.6799, "step": 32950 }, { "epoch": 1.78, "learning_rate": 4.113924050632912e-05, "loss": 1.7658, "step": 32960 }, { "epoch": 1.78, "learning_rate": 4.113653575678892e-05, "loss": 1.7906, "step": 32970 }, { "epoch": 1.78, "learning_rate": 4.113383100724873e-05, "loss": 1.7534, "step": 32980 }, { "epoch": 1.78, "learning_rate": 4.1131126257708535e-05, "loss": 1.7744, "step": 32990 }, { "epoch": 1.78, "learning_rate": 4.1128421508168345e-05, "loss": 1.7873, "step": 33000 }, { "epoch": 1.78, "learning_rate": 4.1125716758628155e-05, "loss": 1.7913, "step": 33010 }, { "epoch": 1.78, "learning_rate": 4.112301200908796e-05, "loss": 1.7447, "step": 33020 }, { "epoch": 1.78, "learning_rate": 4.112030725954777e-05, "loss": 1.7174, "step": 33030 }, { "epoch": 1.79, "learning_rate": 4.111760251000758e-05, "loss": 1.7296, "step": 33040 }, { "epoch": 1.79, "learning_rate": 4.111489776046738e-05, "loss": 1.8705, "step": 33050 }, { "epoch": 1.79, "learning_rate": 4.111219301092719e-05, "loss": 1.8289, "step": 33060 }, { "epoch": 1.79, "learning_rate": 4.1109488261387e-05, "loss": 1.7272, "step": 33070 }, { "epoch": 1.79, "learning_rate": 4.11067835118468e-05, "loss": 1.748, "step": 33080 }, { "epoch": 1.79, "learning_rate": 4.110407876230661e-05, "loss": 1.8643, "step": 33090 }, { "epoch": 1.79, "learning_rate": 4.110137401276642e-05, "loss": 1.752, "step": 33100 }, { "epoch": 1.79, "learning_rate": 4.1098669263226224e-05, "loss": 1.7818, "step": 33110 }, { "epoch": 1.79, "learning_rate": 4.1095964513686034e-05, "loss": 1.8322, "step": 33120 }, { "epoch": 1.79, "learning_rate": 4.109325976414584e-05, "loss": 1.7905, "step": 33130 }, { "epoch": 1.79, "learning_rate": 4.1090555014605646e-05, "loss": 1.6881, "step": 33140 }, { "epoch": 1.79, "learning_rate": 4.1087850265065456e-05, "loss": 1.8146, "step": 33150 }, { "epoch": 1.79, "learning_rate": 4.1085145515525265e-05, "loss": 1.7591, "step": 33160 }, { "epoch": 1.79, "learning_rate": 4.108244076598507e-05, "loss": 1.855, "step": 33170 }, { "epoch": 1.79, "learning_rate": 4.107973601644488e-05, "loss": 1.8659, "step": 33180 }, { "epoch": 1.79, "learning_rate": 4.107703126690469e-05, "loss": 1.7495, "step": 33190 }, { "epoch": 1.79, "learning_rate": 4.107432651736449e-05, "loss": 1.751, "step": 33200 }, { "epoch": 1.79, "learning_rate": 4.10716217678243e-05, "loss": 1.7464, "step": 33210 }, { "epoch": 1.8, "learning_rate": 4.106891701828411e-05, "loss": 1.7624, "step": 33220 }, { "epoch": 1.8, "learning_rate": 4.106621226874391e-05, "loss": 1.7326, "step": 33230 }, { "epoch": 1.8, "learning_rate": 4.106350751920372e-05, "loss": 1.7038, "step": 33240 }, { "epoch": 1.8, "learning_rate": 4.106080276966353e-05, "loss": 1.733, "step": 33250 }, { "epoch": 1.8, "learning_rate": 4.1058098020123335e-05, "loss": 1.8018, "step": 33260 }, { "epoch": 1.8, "learning_rate": 4.105539327058315e-05, "loss": 1.7512, "step": 33270 }, { "epoch": 1.8, "learning_rate": 4.1052688521042954e-05, "loss": 1.7644, "step": 33280 }, { "epoch": 1.8, "learning_rate": 4.104998377150276e-05, "loss": 1.7681, "step": 33290 }, { "epoch": 1.8, "learning_rate": 4.104727902196257e-05, "loss": 1.784, "step": 33300 }, { "epoch": 1.8, "learning_rate": 4.1044574272422376e-05, "loss": 1.7858, "step": 33310 }, { "epoch": 1.8, "learning_rate": 4.104186952288218e-05, "loss": 1.6988, "step": 33320 }, { "epoch": 1.8, "learning_rate": 4.1039164773341996e-05, "loss": 1.828, "step": 33330 }, { "epoch": 1.8, "learning_rate": 4.10364600238018e-05, "loss": 1.7259, "step": 33340 }, { "epoch": 1.8, "learning_rate": 4.10337552742616e-05, "loss": 1.7309, "step": 33350 }, { "epoch": 1.8, "learning_rate": 4.103105052472142e-05, "loss": 1.7142, "step": 33360 }, { "epoch": 1.8, "learning_rate": 4.102834577518122e-05, "loss": 1.7932, "step": 33370 }, { "epoch": 1.8, "learning_rate": 4.1025641025641023e-05, "loss": 1.7412, "step": 33380 }, { "epoch": 1.8, "learning_rate": 4.102293627610084e-05, "loss": 1.7779, "step": 33390 }, { "epoch": 1.8, "learning_rate": 4.102023152656064e-05, "loss": 1.7643, "step": 33400 }, { "epoch": 1.81, "learning_rate": 4.1017526777020446e-05, "loss": 1.7599, "step": 33410 }, { "epoch": 1.81, "learning_rate": 4.101482202748026e-05, "loss": 1.8607, "step": 33420 }, { "epoch": 1.81, "learning_rate": 4.1012117277940065e-05, "loss": 1.799, "step": 33430 }, { "epoch": 1.81, "learning_rate": 4.100941252839987e-05, "loss": 1.7943, "step": 33440 }, { "epoch": 1.81, "learning_rate": 4.1006707778859684e-05, "loss": 1.8424, "step": 33450 }, { "epoch": 1.81, "learning_rate": 4.100400302931949e-05, "loss": 1.7309, "step": 33460 }, { "epoch": 1.81, "learning_rate": 4.100129827977929e-05, "loss": 1.7433, "step": 33470 }, { "epoch": 1.81, "learning_rate": 4.0998593530239106e-05, "loss": 1.8025, "step": 33480 }, { "epoch": 1.81, "learning_rate": 4.099588878069891e-05, "loss": 1.7542, "step": 33490 }, { "epoch": 1.81, "learning_rate": 4.099318403115872e-05, "loss": 1.7398, "step": 33500 }, { "epoch": 1.81, "learning_rate": 4.099047928161853e-05, "loss": 1.7668, "step": 33510 }, { "epoch": 1.81, "learning_rate": 4.098777453207833e-05, "loss": 1.7252, "step": 33520 }, { "epoch": 1.81, "learning_rate": 4.098506978253814e-05, "loss": 1.7312, "step": 33530 }, { "epoch": 1.81, "learning_rate": 4.0982365032997944e-05, "loss": 1.773, "step": 33540 }, { "epoch": 1.81, "learning_rate": 4.0979660283457754e-05, "loss": 1.724, "step": 33550 }, { "epoch": 1.81, "learning_rate": 4.097695553391756e-05, "loss": 1.7826, "step": 33560 }, { "epoch": 1.81, "learning_rate": 4.0974250784377366e-05, "loss": 1.8553, "step": 33570 }, { "epoch": 1.81, "learning_rate": 4.0971546034837176e-05, "loss": 1.7819, "step": 33580 }, { "epoch": 1.82, "learning_rate": 4.0968841285296985e-05, "loss": 1.8379, "step": 33590 }, { "epoch": 1.82, "learning_rate": 4.096613653575679e-05, "loss": 1.7221, "step": 33600 }, { "epoch": 1.82, "learning_rate": 4.09634317862166e-05, "loss": 1.8114, "step": 33610 }, { "epoch": 1.82, "learning_rate": 4.096072703667641e-05, "loss": 1.7602, "step": 33620 }, { "epoch": 1.82, "learning_rate": 4.095802228713621e-05, "loss": 1.777, "step": 33630 }, { "epoch": 1.82, "learning_rate": 4.095531753759602e-05, "loss": 1.7701, "step": 33640 }, { "epoch": 1.82, "learning_rate": 4.095261278805583e-05, "loss": 1.8082, "step": 33650 }, { "epoch": 1.82, "learning_rate": 4.094990803851563e-05, "loss": 1.795, "step": 33660 }, { "epoch": 1.82, "learning_rate": 4.094720328897544e-05, "loss": 1.7756, "step": 33670 }, { "epoch": 1.82, "learning_rate": 4.094449853943525e-05, "loss": 1.7531, "step": 33680 }, { "epoch": 1.82, "learning_rate": 4.0941793789895055e-05, "loss": 1.7561, "step": 33690 }, { "epoch": 1.82, "learning_rate": 4.0939089040354864e-05, "loss": 1.6618, "step": 33700 }, { "epoch": 1.82, "learning_rate": 4.0936384290814674e-05, "loss": 1.7997, "step": 33710 }, { "epoch": 1.82, "learning_rate": 4.093367954127448e-05, "loss": 1.7245, "step": 33720 }, { "epoch": 1.82, "learning_rate": 4.0930974791734287e-05, "loss": 1.6676, "step": 33730 }, { "epoch": 1.82, "learning_rate": 4.0928270042194096e-05, "loss": 1.7535, "step": 33740 }, { "epoch": 1.82, "learning_rate": 4.09255652926539e-05, "loss": 1.7644, "step": 33750 }, { "epoch": 1.82, "learning_rate": 4.092286054311371e-05, "loss": 1.7874, "step": 33760 }, { "epoch": 1.82, "learning_rate": 4.092015579357352e-05, "loss": 1.7701, "step": 33770 }, { "epoch": 1.83, "learning_rate": 4.091745104403332e-05, "loss": 1.7428, "step": 33780 }, { "epoch": 1.83, "learning_rate": 4.091474629449313e-05, "loss": 1.7246, "step": 33790 }, { "epoch": 1.83, "learning_rate": 4.091204154495294e-05, "loss": 1.7743, "step": 33800 }, { "epoch": 1.83, "learning_rate": 4.0909336795412743e-05, "loss": 1.6498, "step": 33810 }, { "epoch": 1.83, "learning_rate": 4.090663204587255e-05, "loss": 1.7839, "step": 33820 }, { "epoch": 1.83, "learning_rate": 4.090392729633236e-05, "loss": 1.7716, "step": 33830 }, { "epoch": 1.83, "learning_rate": 4.0901222546792166e-05, "loss": 1.7409, "step": 33840 }, { "epoch": 1.83, "learning_rate": 4.0898517797251975e-05, "loss": 1.7009, "step": 33850 }, { "epoch": 1.83, "learning_rate": 4.0895813047711785e-05, "loss": 1.7868, "step": 33860 }, { "epoch": 1.83, "learning_rate": 4.089310829817159e-05, "loss": 1.7478, "step": 33870 }, { "epoch": 1.83, "learning_rate": 4.08904035486314e-05, "loss": 1.7396, "step": 33880 }, { "epoch": 1.83, "learning_rate": 4.088769879909121e-05, "loss": 1.7929, "step": 33890 }, { "epoch": 1.83, "learning_rate": 4.088499404955101e-05, "loss": 1.7937, "step": 33900 }, { "epoch": 1.83, "learning_rate": 4.0882289300010826e-05, "loss": 1.6996, "step": 33910 }, { "epoch": 1.83, "learning_rate": 4.087958455047063e-05, "loss": 1.7712, "step": 33920 }, { "epoch": 1.83, "learning_rate": 4.087687980093043e-05, "loss": 1.7272, "step": 33930 }, { "epoch": 1.83, "learning_rate": 4.087417505139025e-05, "loss": 1.7627, "step": 33940 }, { "epoch": 1.83, "learning_rate": 4.087147030185005e-05, "loss": 1.8535, "step": 33950 }, { "epoch": 1.84, "learning_rate": 4.0868765552309854e-05, "loss": 1.827, "step": 33960 }, { "epoch": 1.84, "learning_rate": 4.086606080276967e-05, "loss": 1.7476, "step": 33970 }, { "epoch": 1.84, "learning_rate": 4.0863356053229474e-05, "loss": 1.7838, "step": 33980 }, { "epoch": 1.84, "learning_rate": 4.0860651303689276e-05, "loss": 1.6792, "step": 33990 }, { "epoch": 1.84, "learning_rate": 4.085794655414909e-05, "loss": 1.6933, "step": 34000 }, { "epoch": 1.84, "learning_rate": 4.0855241804608896e-05, "loss": 1.8164, "step": 34010 }, { "epoch": 1.84, "learning_rate": 4.08525370550687e-05, "loss": 1.7291, "step": 34020 }, { "epoch": 1.84, "learning_rate": 4.0849832305528515e-05, "loss": 1.772, "step": 34030 }, { "epoch": 1.84, "learning_rate": 4.084712755598832e-05, "loss": 1.7733, "step": 34040 }, { "epoch": 1.84, "learning_rate": 4.084442280644812e-05, "loss": 1.7448, "step": 34050 }, { "epoch": 1.84, "learning_rate": 4.084171805690794e-05, "loss": 1.848, "step": 34060 }, { "epoch": 1.84, "learning_rate": 4.083901330736774e-05, "loss": 1.7908, "step": 34070 }, { "epoch": 1.84, "learning_rate": 4.083630855782754e-05, "loss": 1.7208, "step": 34080 }, { "epoch": 1.84, "learning_rate": 4.083360380828736e-05, "loss": 1.716, "step": 34090 }, { "epoch": 1.84, "learning_rate": 4.083089905874716e-05, "loss": 1.754, "step": 34100 }, { "epoch": 1.84, "learning_rate": 4.0828194309206965e-05, "loss": 1.6908, "step": 34110 }, { "epoch": 1.84, "learning_rate": 4.0825489559666775e-05, "loss": 1.7674, "step": 34120 }, { "epoch": 1.84, "learning_rate": 4.0822784810126584e-05, "loss": 1.6876, "step": 34130 }, { "epoch": 1.84, "learning_rate": 4.0820080060586394e-05, "loss": 1.7171, "step": 34140 }, { "epoch": 1.85, "learning_rate": 4.08173753110462e-05, "loss": 1.6891, "step": 34150 }, { "epoch": 1.85, "learning_rate": 4.081467056150601e-05, "loss": 1.6954, "step": 34160 }, { "epoch": 1.85, "learning_rate": 4.0811965811965816e-05, "loss": 1.7752, "step": 34170 }, { "epoch": 1.85, "learning_rate": 4.080926106242562e-05, "loss": 1.8065, "step": 34180 }, { "epoch": 1.85, "learning_rate": 4.080655631288543e-05, "loss": 1.7291, "step": 34190 }, { "epoch": 1.85, "learning_rate": 4.080385156334524e-05, "loss": 1.7182, "step": 34200 }, { "epoch": 1.85, "learning_rate": 4.080114681380504e-05, "loss": 1.8034, "step": 34210 }, { "epoch": 1.85, "learning_rate": 4.079844206426485e-05, "loss": 1.746, "step": 34220 }, { "epoch": 1.85, "learning_rate": 4.079573731472466e-05, "loss": 1.8036, "step": 34230 }, { "epoch": 1.85, "learning_rate": 4.0793032565184463e-05, "loss": 1.7271, "step": 34240 }, { "epoch": 1.85, "learning_rate": 4.079032781564427e-05, "loss": 1.7079, "step": 34250 }, { "epoch": 1.85, "learning_rate": 4.078762306610408e-05, "loss": 1.7023, "step": 34260 }, { "epoch": 1.85, "learning_rate": 4.0784918316563886e-05, "loss": 1.6387, "step": 34270 }, { "epoch": 1.85, "learning_rate": 4.0782213567023695e-05, "loss": 1.7839, "step": 34280 }, { "epoch": 1.85, "learning_rate": 4.0779508817483505e-05, "loss": 1.7547, "step": 34290 }, { "epoch": 1.85, "learning_rate": 4.077680406794331e-05, "loss": 1.7597, "step": 34300 }, { "epoch": 1.85, "learning_rate": 4.077409931840312e-05, "loss": 1.7716, "step": 34310 }, { "epoch": 1.85, "learning_rate": 4.077139456886293e-05, "loss": 1.7395, "step": 34320 }, { "epoch": 1.85, "learning_rate": 4.076868981932273e-05, "loss": 1.7651, "step": 34330 }, { "epoch": 1.86, "learning_rate": 4.076598506978254e-05, "loss": 1.7316, "step": 34340 }, { "epoch": 1.86, "learning_rate": 4.076328032024235e-05, "loss": 1.7582, "step": 34350 }, { "epoch": 1.86, "learning_rate": 4.076057557070215e-05, "loss": 1.7414, "step": 34360 }, { "epoch": 1.86, "learning_rate": 4.075787082116196e-05, "loss": 1.7186, "step": 34370 }, { "epoch": 1.86, "learning_rate": 4.075516607162177e-05, "loss": 1.7185, "step": 34380 }, { "epoch": 1.86, "learning_rate": 4.0752461322081574e-05, "loss": 1.7829, "step": 34390 }, { "epoch": 1.86, "learning_rate": 4.0749756572541384e-05, "loss": 1.7666, "step": 34400 }, { "epoch": 1.86, "learning_rate": 4.0747051823001194e-05, "loss": 1.7586, "step": 34410 }, { "epoch": 1.86, "learning_rate": 4.0744347073460996e-05, "loss": 1.7145, "step": 34420 }, { "epoch": 1.86, "learning_rate": 4.0741642323920806e-05, "loss": 1.7239, "step": 34430 }, { "epoch": 1.86, "learning_rate": 4.0738937574380616e-05, "loss": 1.7595, "step": 34440 }, { "epoch": 1.86, "learning_rate": 4.073623282484042e-05, "loss": 1.7768, "step": 34450 }, { "epoch": 1.86, "learning_rate": 4.073352807530023e-05, "loss": 1.8187, "step": 34460 }, { "epoch": 1.86, "learning_rate": 4.073082332576004e-05, "loss": 1.8027, "step": 34470 }, { "epoch": 1.86, "learning_rate": 4.072811857621984e-05, "loss": 1.7716, "step": 34480 }, { "epoch": 1.86, "learning_rate": 4.072541382667965e-05, "loss": 1.7609, "step": 34490 }, { "epoch": 1.86, "learning_rate": 4.072270907713946e-05, "loss": 1.6947, "step": 34500 }, { "epoch": 1.86, "learning_rate": 4.072000432759926e-05, "loss": 1.6814, "step": 34510 }, { "epoch": 1.87, "learning_rate": 4.071729957805907e-05, "loss": 1.7065, "step": 34520 }, { "epoch": 1.87, "learning_rate": 4.071459482851888e-05, "loss": 1.752, "step": 34530 }, { "epoch": 1.87, "learning_rate": 4.0711890078978685e-05, "loss": 1.7662, "step": 34540 }, { "epoch": 1.87, "learning_rate": 4.07091853294385e-05, "loss": 1.7582, "step": 34550 }, { "epoch": 1.87, "learning_rate": 4.0706480579898304e-05, "loss": 1.7258, "step": 34560 }, { "epoch": 1.87, "learning_rate": 4.070377583035811e-05, "loss": 1.7865, "step": 34570 }, { "epoch": 1.87, "learning_rate": 4.0701071080817924e-05, "loss": 1.8106, "step": 34580 }, { "epoch": 1.87, "learning_rate": 4.069836633127773e-05, "loss": 1.7269, "step": 34590 }, { "epoch": 1.87, "learning_rate": 4.069566158173753e-05, "loss": 1.7156, "step": 34600 }, { "epoch": 1.87, "learning_rate": 4.0692956832197346e-05, "loss": 1.7396, "step": 34610 }, { "epoch": 1.87, "learning_rate": 4.069025208265715e-05, "loss": 1.7546, "step": 34620 }, { "epoch": 1.87, "learning_rate": 4.068754733311695e-05, "loss": 1.8109, "step": 34630 }, { "epoch": 1.87, "learning_rate": 4.068484258357677e-05, "loss": 1.7338, "step": 34640 }, { "epoch": 1.87, "learning_rate": 4.068213783403657e-05, "loss": 1.761, "step": 34650 }, { "epoch": 1.87, "learning_rate": 4.0679433084496374e-05, "loss": 1.7171, "step": 34660 }, { "epoch": 1.87, "learning_rate": 4.0676728334956183e-05, "loss": 1.7631, "step": 34670 }, { "epoch": 1.87, "learning_rate": 4.067402358541599e-05, "loss": 1.7661, "step": 34680 }, { "epoch": 1.87, "learning_rate": 4.0671318835875796e-05, "loss": 1.7063, "step": 34690 }, { "epoch": 1.87, "learning_rate": 4.0668614086335606e-05, "loss": 1.7268, "step": 34700 }, { "epoch": 1.88, "learning_rate": 4.0665909336795415e-05, "loss": 1.804, "step": 34710 }, { "epoch": 1.88, "learning_rate": 4.066320458725522e-05, "loss": 1.7544, "step": 34720 }, { "epoch": 1.88, "learning_rate": 4.066049983771503e-05, "loss": 1.6939, "step": 34730 }, { "epoch": 1.88, "learning_rate": 4.065779508817484e-05, "loss": 1.6853, "step": 34740 }, { "epoch": 1.88, "learning_rate": 4.065509033863465e-05, "loss": 1.7596, "step": 34750 }, { "epoch": 1.88, "learning_rate": 4.065238558909445e-05, "loss": 1.8127, "step": 34760 }, { "epoch": 1.88, "learning_rate": 4.064968083955426e-05, "loss": 1.7763, "step": 34770 }, { "epoch": 1.88, "learning_rate": 4.064697609001407e-05, "loss": 1.7815, "step": 34780 }, { "epoch": 1.88, "learning_rate": 4.064427134047387e-05, "loss": 1.745, "step": 34790 }, { "epoch": 1.88, "learning_rate": 4.064156659093368e-05, "loss": 1.6241, "step": 34800 }, { "epoch": 1.88, "learning_rate": 4.063886184139349e-05, "loss": 1.6717, "step": 34810 }, { "epoch": 1.88, "learning_rate": 4.0636157091853294e-05, "loss": 1.7361, "step": 34820 }, { "epoch": 1.88, "learning_rate": 4.0633452342313104e-05, "loss": 1.7912, "step": 34830 }, { "epoch": 1.88, "learning_rate": 4.0630747592772914e-05, "loss": 1.7797, "step": 34840 }, { "epoch": 1.88, "learning_rate": 4.0628042843232717e-05, "loss": 1.7273, "step": 34850 }, { "epoch": 1.88, "learning_rate": 4.0625338093692526e-05, "loss": 1.6965, "step": 34860 }, { "epoch": 1.88, "learning_rate": 4.0622633344152336e-05, "loss": 1.7568, "step": 34870 }, { "epoch": 1.88, "learning_rate": 4.061992859461214e-05, "loss": 1.7586, "step": 34880 }, { "epoch": 1.89, "learning_rate": 4.061722384507195e-05, "loss": 1.7725, "step": 34890 }, { "epoch": 1.89, "learning_rate": 4.061451909553176e-05, "loss": 1.8616, "step": 34900 }, { "epoch": 1.89, "learning_rate": 4.061181434599156e-05, "loss": 1.7351, "step": 34910 }, { "epoch": 1.89, "learning_rate": 4.060910959645137e-05, "loss": 1.8423, "step": 34920 }, { "epoch": 1.89, "learning_rate": 4.060640484691118e-05, "loss": 1.7908, "step": 34930 }, { "epoch": 1.89, "learning_rate": 4.060370009737098e-05, "loss": 1.7878, "step": 34940 }, { "epoch": 1.89, "learning_rate": 4.060099534783079e-05, "loss": 1.8371, "step": 34950 }, { "epoch": 1.89, "learning_rate": 4.05982905982906e-05, "loss": 1.6494, "step": 34960 }, { "epoch": 1.89, "learning_rate": 4.0595585848750405e-05, "loss": 1.7195, "step": 34970 }, { "epoch": 1.89, "learning_rate": 4.0592881099210215e-05, "loss": 1.7657, "step": 34980 }, { "epoch": 1.89, "learning_rate": 4.0590176349670024e-05, "loss": 1.7159, "step": 34990 }, { "epoch": 1.89, "learning_rate": 4.058747160012983e-05, "loss": 1.6493, "step": 35000 }, { "epoch": 1.89, "learning_rate": 4.058476685058964e-05, "loss": 1.726, "step": 35010 }, { "epoch": 1.89, "learning_rate": 4.058206210104945e-05, "loss": 1.6953, "step": 35020 }, { "epoch": 1.89, "learning_rate": 4.057935735150925e-05, "loss": 1.7613, "step": 35030 }, { "epoch": 1.89, "learning_rate": 4.057665260196906e-05, "loss": 1.7314, "step": 35040 }, { "epoch": 1.89, "learning_rate": 4.057394785242887e-05, "loss": 1.7674, "step": 35050 }, { "epoch": 1.89, "learning_rate": 4.057124310288867e-05, "loss": 1.746, "step": 35060 }, { "epoch": 1.89, "learning_rate": 4.056853835334848e-05, "loss": 1.7463, "step": 35070 }, { "epoch": 1.9, "learning_rate": 4.056583360380829e-05, "loss": 1.7349, "step": 35080 }, { "epoch": 1.9, "learning_rate": 4.0563128854268094e-05, "loss": 1.7614, "step": 35090 }, { "epoch": 1.9, "learning_rate": 4.0560424104727904e-05, "loss": 1.7861, "step": 35100 }, { "epoch": 1.9, "learning_rate": 4.055771935518771e-05, "loss": 1.8582, "step": 35110 }, { "epoch": 1.9, "learning_rate": 4.0555014605647516e-05, "loss": 1.8391, "step": 35120 }, { "epoch": 1.9, "learning_rate": 4.0552309856107326e-05, "loss": 1.7828, "step": 35130 }, { "epoch": 1.9, "learning_rate": 4.0549605106567135e-05, "loss": 1.7916, "step": 35140 }, { "epoch": 1.9, "learning_rate": 4.054690035702694e-05, "loss": 1.8231, "step": 35150 }, { "epoch": 1.9, "learning_rate": 4.0544195607486755e-05, "loss": 1.7696, "step": 35160 }, { "epoch": 1.9, "learning_rate": 4.054149085794656e-05, "loss": 1.74, "step": 35170 }, { "epoch": 1.9, "learning_rate": 4.053878610840636e-05, "loss": 1.8273, "step": 35180 }, { "epoch": 1.9, "learning_rate": 4.053608135886618e-05, "loss": 1.73, "step": 35190 }, { "epoch": 1.9, "learning_rate": 4.053337660932598e-05, "loss": 1.8021, "step": 35200 }, { "epoch": 1.9, "learning_rate": 4.053067185978578e-05, "loss": 1.6863, "step": 35210 }, { "epoch": 1.9, "learning_rate": 4.052796711024559e-05, "loss": 1.7835, "step": 35220 }, { "epoch": 1.9, "learning_rate": 4.05252623607054e-05, "loss": 1.8032, "step": 35230 }, { "epoch": 1.9, "learning_rate": 4.0522557611165205e-05, "loss": 1.8051, "step": 35240 }, { "epoch": 1.9, "learning_rate": 4.0519852861625014e-05, "loss": 1.7258, "step": 35250 }, { "epoch": 1.91, "learning_rate": 4.0517148112084824e-05, "loss": 1.839, "step": 35260 }, { "epoch": 1.91, "learning_rate": 4.051444336254463e-05, "loss": 1.7655, "step": 35270 }, { "epoch": 1.91, "learning_rate": 4.0511738613004437e-05, "loss": 1.7148, "step": 35280 }, { "epoch": 1.91, "learning_rate": 4.0509033863464246e-05, "loss": 1.6483, "step": 35290 }, { "epoch": 1.91, "learning_rate": 4.050632911392405e-05, "loss": 1.6989, "step": 35300 }, { "epoch": 1.91, "learning_rate": 4.050362436438386e-05, "loss": 1.8304, "step": 35310 }, { "epoch": 1.91, "learning_rate": 4.050091961484367e-05, "loss": 1.7927, "step": 35320 }, { "epoch": 1.91, "learning_rate": 4.049821486530347e-05, "loss": 1.7422, "step": 35330 }, { "epoch": 1.91, "learning_rate": 4.049551011576328e-05, "loss": 1.7668, "step": 35340 }, { "epoch": 1.91, "learning_rate": 4.049280536622309e-05, "loss": 1.764, "step": 35350 }, { "epoch": 1.91, "learning_rate": 4.0490100616682893e-05, "loss": 1.7244, "step": 35360 }, { "epoch": 1.91, "learning_rate": 4.04873958671427e-05, "loss": 1.7469, "step": 35370 }, { "epoch": 1.91, "learning_rate": 4.048469111760251e-05, "loss": 1.7073, "step": 35380 }, { "epoch": 1.91, "learning_rate": 4.048198636806232e-05, "loss": 1.8231, "step": 35390 }, { "epoch": 1.91, "learning_rate": 4.0479281618522125e-05, "loss": 1.7535, "step": 35400 }, { "epoch": 1.91, "learning_rate": 4.0476576868981935e-05, "loss": 1.8546, "step": 35410 }, { "epoch": 1.91, "learning_rate": 4.0473872119441745e-05, "loss": 1.7537, "step": 35420 }, { "epoch": 1.91, "learning_rate": 4.047116736990155e-05, "loss": 1.7342, "step": 35430 }, { "epoch": 1.91, "learning_rate": 4.046846262036136e-05, "loss": 1.74, "step": 35440 }, { "epoch": 1.92, "learning_rate": 4.046575787082117e-05, "loss": 1.7782, "step": 35450 }, { "epoch": 1.92, "learning_rate": 4.046305312128097e-05, "loss": 1.8022, "step": 35460 }, { "epoch": 1.92, "learning_rate": 4.046034837174078e-05, "loss": 1.8507, "step": 35470 }, { "epoch": 1.92, "learning_rate": 4.045764362220059e-05, "loss": 1.739, "step": 35480 }, { "epoch": 1.92, "learning_rate": 4.045493887266039e-05, "loss": 1.7573, "step": 35490 }, { "epoch": 1.92, "learning_rate": 4.04522341231202e-05, "loss": 1.7632, "step": 35500 }, { "epoch": 1.92, "learning_rate": 4.044952937358001e-05, "loss": 1.8174, "step": 35510 }, { "epoch": 1.92, "learning_rate": 4.0446824624039814e-05, "loss": 1.7468, "step": 35520 }, { "epoch": 1.92, "learning_rate": 4.0444119874499624e-05, "loss": 1.7591, "step": 35530 }, { "epoch": 1.92, "learning_rate": 4.044141512495943e-05, "loss": 1.7992, "step": 35540 }, { "epoch": 1.92, "learning_rate": 4.0438710375419236e-05, "loss": 1.8177, "step": 35550 }, { "epoch": 1.92, "learning_rate": 4.0436005625879046e-05, "loss": 1.8453, "step": 35560 }, { "epoch": 1.92, "learning_rate": 4.0433300876338855e-05, "loss": 1.7482, "step": 35570 }, { "epoch": 1.92, "learning_rate": 4.043059612679866e-05, "loss": 1.8081, "step": 35580 }, { "epoch": 1.92, "learning_rate": 4.042789137725847e-05, "loss": 1.7582, "step": 35590 }, { "epoch": 1.92, "learning_rate": 4.042518662771828e-05, "loss": 1.7359, "step": 35600 }, { "epoch": 1.92, "learning_rate": 4.042248187817808e-05, "loss": 1.7218, "step": 35610 }, { "epoch": 1.92, "learning_rate": 4.041977712863789e-05, "loss": 1.6995, "step": 35620 }, { "epoch": 1.93, "learning_rate": 4.04170723790977e-05, "loss": 1.7527, "step": 35630 }, { "epoch": 1.93, "learning_rate": 4.04143676295575e-05, "loss": 1.8261, "step": 35640 }, { "epoch": 1.93, "learning_rate": 4.041166288001731e-05, "loss": 1.7513, "step": 35650 }, { "epoch": 1.93, "learning_rate": 4.040895813047712e-05, "loss": 1.7648, "step": 35660 }, { "epoch": 1.93, "learning_rate": 4.0406253380936925e-05, "loss": 1.7292, "step": 35670 }, { "epoch": 1.93, "learning_rate": 4.0403548631396734e-05, "loss": 1.7235, "step": 35680 }, { "epoch": 1.93, "learning_rate": 4.0400843881856544e-05, "loss": 1.7749, "step": 35690 }, { "epoch": 1.93, "learning_rate": 4.039813913231635e-05, "loss": 1.7954, "step": 35700 }, { "epoch": 1.93, "learning_rate": 4.0395434382776157e-05, "loss": 1.6911, "step": 35710 }, { "epoch": 1.93, "learning_rate": 4.0392729633235966e-05, "loss": 1.6618, "step": 35720 }, { "epoch": 1.93, "learning_rate": 4.039002488369577e-05, "loss": 1.6867, "step": 35730 }, { "epoch": 1.93, "learning_rate": 4.038732013415558e-05, "loss": 1.7291, "step": 35740 }, { "epoch": 1.93, "learning_rate": 4.038461538461539e-05, "loss": 1.8566, "step": 35750 }, { "epoch": 1.93, "learning_rate": 4.038191063507519e-05, "loss": 1.726, "step": 35760 }, { "epoch": 1.93, "learning_rate": 4.0379205885535e-05, "loss": 1.7714, "step": 35770 }, { "epoch": 1.93, "learning_rate": 4.037650113599481e-05, "loss": 1.7794, "step": 35780 }, { "epoch": 1.93, "learning_rate": 4.0373796386454613e-05, "loss": 1.7808, "step": 35790 }, { "epoch": 1.93, "learning_rate": 4.037109163691442e-05, "loss": 1.6972, "step": 35800 }, { "epoch": 1.93, "learning_rate": 4.036838688737423e-05, "loss": 1.8761, "step": 35810 }, { "epoch": 1.94, "learning_rate": 4.0365682137834036e-05, "loss": 1.7574, "step": 35820 }, { "epoch": 1.94, "learning_rate": 4.0362977388293845e-05, "loss": 1.7098, "step": 35830 }, { "epoch": 1.94, "learning_rate": 4.0360272638753655e-05, "loss": 1.7476, "step": 35840 }, { "epoch": 1.94, "learning_rate": 4.035756788921346e-05, "loss": 1.6935, "step": 35850 }, { "epoch": 1.94, "learning_rate": 4.035486313967327e-05, "loss": 1.7666, "step": 35860 }, { "epoch": 1.94, "learning_rate": 4.035215839013308e-05, "loss": 1.7442, "step": 35870 }, { "epoch": 1.94, "learning_rate": 4.034945364059288e-05, "loss": 1.7123, "step": 35880 }, { "epoch": 1.94, "learning_rate": 4.034674889105269e-05, "loss": 1.7671, "step": 35890 }, { "epoch": 1.94, "learning_rate": 4.03440441415125e-05, "loss": 1.7241, "step": 35900 }, { "epoch": 1.94, "learning_rate": 4.03413393919723e-05, "loss": 1.7426, "step": 35910 }, { "epoch": 1.94, "learning_rate": 4.033863464243211e-05, "loss": 1.7217, "step": 35920 }, { "epoch": 1.94, "learning_rate": 4.033592989289192e-05, "loss": 1.7386, "step": 35930 }, { "epoch": 1.94, "learning_rate": 4.0333225143351724e-05, "loss": 1.6907, "step": 35940 }, { "epoch": 1.94, "learning_rate": 4.0330520393811534e-05, "loss": 1.8518, "step": 35950 }, { "epoch": 1.94, "learning_rate": 4.0327815644271344e-05, "loss": 1.7826, "step": 35960 }, { "epoch": 1.94, "learning_rate": 4.0325110894731146e-05, "loss": 1.8788, "step": 35970 }, { "epoch": 1.94, "learning_rate": 4.0322406145190956e-05, "loss": 1.7058, "step": 35980 }, { "epoch": 1.94, "learning_rate": 4.0319701395650766e-05, "loss": 1.8237, "step": 35990 }, { "epoch": 1.95, "learning_rate": 4.031699664611057e-05, "loss": 1.7754, "step": 36000 }, { "epoch": 1.95, "learning_rate": 4.031429189657038e-05, "loss": 1.7676, "step": 36010 }, { "epoch": 1.95, "learning_rate": 4.031158714703019e-05, "loss": 1.8186, "step": 36020 }, { "epoch": 1.95, "learning_rate": 4.030888239749e-05, "loss": 1.7517, "step": 36030 }, { "epoch": 1.95, "learning_rate": 4.03061776479498e-05, "loss": 1.762, "step": 36040 }, { "epoch": 1.95, "learning_rate": 4.030347289840961e-05, "loss": 1.7784, "step": 36050 }, { "epoch": 1.95, "learning_rate": 4.030076814886942e-05, "loss": 1.7188, "step": 36060 }, { "epoch": 1.95, "learning_rate": 4.029806339932922e-05, "loss": 1.8534, "step": 36070 }, { "epoch": 1.95, "learning_rate": 4.029535864978903e-05, "loss": 1.7861, "step": 36080 }, { "epoch": 1.95, "learning_rate": 4.029265390024884e-05, "loss": 1.7248, "step": 36090 }, { "epoch": 1.95, "learning_rate": 4.0289949150708645e-05, "loss": 1.6338, "step": 36100 }, { "epoch": 1.95, "learning_rate": 4.0287244401168454e-05, "loss": 1.7429, "step": 36110 }, { "epoch": 1.95, "learning_rate": 4.0284539651628264e-05, "loss": 1.7703, "step": 36120 }, { "epoch": 1.95, "learning_rate": 4.028183490208807e-05, "loss": 1.742, "step": 36130 }, { "epoch": 1.95, "learning_rate": 4.0279130152547877e-05, "loss": 1.7758, "step": 36140 }, { "epoch": 1.95, "learning_rate": 4.0276425403007686e-05, "loss": 1.7421, "step": 36150 }, { "epoch": 1.95, "learning_rate": 4.027372065346749e-05, "loss": 1.8459, "step": 36160 }, { "epoch": 1.95, "learning_rate": 4.02710159039273e-05, "loss": 1.829, "step": 36170 }, { "epoch": 1.95, "learning_rate": 4.026831115438711e-05, "loss": 1.7447, "step": 36180 }, { "epoch": 1.96, "learning_rate": 4.026560640484691e-05, "loss": 1.8048, "step": 36190 }, { "epoch": 1.96, "learning_rate": 4.026290165530672e-05, "loss": 1.8022, "step": 36200 }, { "epoch": 1.96, "learning_rate": 4.026019690576653e-05, "loss": 1.7131, "step": 36210 }, { "epoch": 1.96, "learning_rate": 4.0257492156226333e-05, "loss": 1.6966, "step": 36220 }, { "epoch": 1.96, "learning_rate": 4.025478740668614e-05, "loss": 1.7956, "step": 36230 }, { "epoch": 1.96, "learning_rate": 4.025208265714595e-05, "loss": 1.7057, "step": 36240 }, { "epoch": 1.96, "learning_rate": 4.0249377907605756e-05, "loss": 1.8452, "step": 36250 }, { "epoch": 1.96, "learning_rate": 4.0246673158065565e-05, "loss": 1.7975, "step": 36260 }, { "epoch": 1.96, "learning_rate": 4.0243968408525375e-05, "loss": 1.7877, "step": 36270 }, { "epoch": 1.96, "learning_rate": 4.024126365898518e-05, "loss": 1.7403, "step": 36280 }, { "epoch": 1.96, "learning_rate": 4.023855890944499e-05, "loss": 1.7384, "step": 36290 }, { "epoch": 1.96, "learning_rate": 4.02358541599048e-05, "loss": 1.7757, "step": 36300 }, { "epoch": 1.96, "learning_rate": 4.02331494103646e-05, "loss": 1.856, "step": 36310 }, { "epoch": 1.96, "learning_rate": 4.023044466082441e-05, "loss": 1.7038, "step": 36320 }, { "epoch": 1.96, "learning_rate": 4.022773991128422e-05, "loss": 1.8588, "step": 36330 }, { "epoch": 1.96, "learning_rate": 4.022503516174402e-05, "loss": 1.685, "step": 36340 }, { "epoch": 1.96, "learning_rate": 4.022233041220383e-05, "loss": 1.7199, "step": 36350 }, { "epoch": 1.96, "learning_rate": 4.021962566266364e-05, "loss": 1.7903, "step": 36360 }, { "epoch": 1.97, "learning_rate": 4.0216920913123444e-05, "loss": 1.7309, "step": 36370 }, { "epoch": 1.97, "learning_rate": 4.0214216163583254e-05, "loss": 1.6965, "step": 36380 }, { "epoch": 1.97, "learning_rate": 4.0211511414043064e-05, "loss": 1.7773, "step": 36390 }, { "epoch": 1.97, "learning_rate": 4.0208806664502866e-05, "loss": 1.7642, "step": 36400 }, { "epoch": 1.97, "learning_rate": 4.0206101914962676e-05, "loss": 1.8507, "step": 36410 }, { "epoch": 1.97, "learning_rate": 4.0203397165422486e-05, "loss": 1.768, "step": 36420 }, { "epoch": 1.97, "learning_rate": 4.020069241588229e-05, "loss": 1.7526, "step": 36430 }, { "epoch": 1.97, "learning_rate": 4.01979876663421e-05, "loss": 1.7668, "step": 36440 }, { "epoch": 1.97, "learning_rate": 4.019528291680191e-05, "loss": 1.7699, "step": 36450 }, { "epoch": 1.97, "learning_rate": 4.019257816726171e-05, "loss": 1.778, "step": 36460 }, { "epoch": 1.97, "learning_rate": 4.018987341772152e-05, "loss": 1.7948, "step": 36470 }, { "epoch": 1.97, "learning_rate": 4.018716866818133e-05, "loss": 1.7643, "step": 36480 }, { "epoch": 1.97, "learning_rate": 4.018446391864113e-05, "loss": 1.7393, "step": 36490 }, { "epoch": 1.97, "learning_rate": 4.018175916910094e-05, "loss": 1.7657, "step": 36500 }, { "epoch": 1.97, "learning_rate": 4.017905441956075e-05, "loss": 1.6695, "step": 36510 }, { "epoch": 1.97, "learning_rate": 4.0176349670020555e-05, "loss": 1.8323, "step": 36520 }, { "epoch": 1.97, "learning_rate": 4.0173644920480365e-05, "loss": 1.7249, "step": 36530 }, { "epoch": 1.97, "learning_rate": 4.0170940170940174e-05, "loss": 1.8068, "step": 36540 }, { "epoch": 1.97, "learning_rate": 4.016823542139998e-05, "loss": 1.8279, "step": 36550 }, { "epoch": 1.98, "learning_rate": 4.016553067185979e-05, "loss": 1.7579, "step": 36560 }, { "epoch": 1.98, "learning_rate": 4.0162825922319597e-05, "loss": 1.6853, "step": 36570 }, { "epoch": 1.98, "learning_rate": 4.01601211727794e-05, "loss": 1.7991, "step": 36580 }, { "epoch": 1.98, "learning_rate": 4.015741642323921e-05, "loss": 1.7153, "step": 36590 }, { "epoch": 1.98, "learning_rate": 4.015471167369902e-05, "loss": 1.8432, "step": 36600 }, { "epoch": 1.98, "learning_rate": 4.015200692415882e-05, "loss": 1.6749, "step": 36610 }, { "epoch": 1.98, "learning_rate": 4.014930217461863e-05, "loss": 1.7342, "step": 36620 }, { "epoch": 1.98, "learning_rate": 4.014659742507844e-05, "loss": 1.7712, "step": 36630 }, { "epoch": 1.98, "learning_rate": 4.0143892675538244e-05, "loss": 1.7105, "step": 36640 }, { "epoch": 1.98, "learning_rate": 4.0141187925998053e-05, "loss": 1.7287, "step": 36650 }, { "epoch": 1.98, "learning_rate": 4.013848317645786e-05, "loss": 1.7055, "step": 36660 }, { "epoch": 1.98, "learning_rate": 4.013577842691767e-05, "loss": 1.7753, "step": 36670 }, { "epoch": 1.98, "learning_rate": 4.0133073677377476e-05, "loss": 1.7245, "step": 36680 }, { "epoch": 1.98, "learning_rate": 4.0130368927837285e-05, "loss": 1.7397, "step": 36690 }, { "epoch": 1.98, "learning_rate": 4.0127664178297095e-05, "loss": 1.7792, "step": 36700 }, { "epoch": 1.98, "learning_rate": 4.01249594287569e-05, "loss": 1.7441, "step": 36710 }, { "epoch": 1.98, "learning_rate": 4.012225467921671e-05, "loss": 1.7932, "step": 36720 }, { "epoch": 1.98, "learning_rate": 4.011954992967652e-05, "loss": 1.759, "step": 36730 }, { "epoch": 1.99, "learning_rate": 4.011684518013632e-05, "loss": 1.7578, "step": 36740 }, { "epoch": 1.99, "learning_rate": 4.011414043059613e-05, "loss": 1.7554, "step": 36750 }, { "epoch": 1.99, "learning_rate": 4.011143568105594e-05, "loss": 1.7138, "step": 36760 }, { "epoch": 1.99, "learning_rate": 4.010873093151574e-05, "loss": 1.7687, "step": 36770 }, { "epoch": 1.99, "learning_rate": 4.010602618197555e-05, "loss": 1.6629, "step": 36780 }, { "epoch": 1.99, "learning_rate": 4.010332143243536e-05, "loss": 1.7122, "step": 36790 }, { "epoch": 1.99, "learning_rate": 4.0100616682895164e-05, "loss": 1.7896, "step": 36800 }, { "epoch": 1.99, "learning_rate": 4.0097911933354974e-05, "loss": 1.7654, "step": 36810 }, { "epoch": 1.99, "learning_rate": 4.0095207183814784e-05, "loss": 1.7216, "step": 36820 }, { "epoch": 1.99, "learning_rate": 4.0092502434274586e-05, "loss": 1.7287, "step": 36830 }, { "epoch": 1.99, "learning_rate": 4.0089797684734396e-05, "loss": 1.7113, "step": 36840 }, { "epoch": 1.99, "learning_rate": 4.0087092935194206e-05, "loss": 1.8978, "step": 36850 }, { "epoch": 1.99, "learning_rate": 4.008438818565401e-05, "loss": 1.6583, "step": 36860 }, { "epoch": 1.99, "learning_rate": 4.008168343611382e-05, "loss": 1.7215, "step": 36870 }, { "epoch": 1.99, "learning_rate": 4.007897868657363e-05, "loss": 1.7431, "step": 36880 }, { "epoch": 1.99, "learning_rate": 4.007627393703343e-05, "loss": 1.7612, "step": 36890 }, { "epoch": 1.99, "learning_rate": 4.007356918749324e-05, "loss": 1.7903, "step": 36900 }, { "epoch": 1.99, "learning_rate": 4.007086443795305e-05, "loss": 1.7088, "step": 36910 }, { "epoch": 1.99, "learning_rate": 4.006815968841285e-05, "loss": 1.8296, "step": 36920 }, { "epoch": 2.0, "learning_rate": 4.006545493887266e-05, "loss": 1.7674, "step": 36930 }, { "epoch": 2.0, "learning_rate": 4.006275018933247e-05, "loss": 1.7156, "step": 36940 }, { "epoch": 2.0, "learning_rate": 4.0060045439792275e-05, "loss": 1.6553, "step": 36950 }, { "epoch": 2.0, "learning_rate": 4.0057340690252085e-05, "loss": 1.7181, "step": 36960 }, { "epoch": 2.0, "learning_rate": 4.0054635940711894e-05, "loss": 1.6879, "step": 36970 }, { "epoch": 2.0, "learning_rate": 4.00519311911717e-05, "loss": 1.7932, "step": 36980 }, { "epoch": 2.0, "learning_rate": 4.004922644163151e-05, "loss": 1.7597, "step": 36990 }, { "epoch": 2.0, "learning_rate": 4.004652169209132e-05, "loss": 1.8302, "step": 37000 }, { "epoch": 2.0, "learning_rate": 4.004381694255112e-05, "loss": 1.7425, "step": 37010 }, { "epoch": 2.0, "learning_rate": 4.004111219301093e-05, "loss": 1.7067, "step": 37020 }, { "epoch": 2.0, "learning_rate": 4.003840744347074e-05, "loss": 1.5585, "step": 37030 }, { "epoch": 2.0, "learning_rate": 4.003570269393054e-05, "loss": 1.614, "step": 37040 }, { "epoch": 2.0, "learning_rate": 4.003299794439035e-05, "loss": 1.5801, "step": 37050 }, { "epoch": 2.0, "learning_rate": 4.003029319485016e-05, "loss": 1.581, "step": 37060 }, { "epoch": 2.0, "learning_rate": 4.0027588445309964e-05, "loss": 1.7214, "step": 37070 }, { "epoch": 2.0, "learning_rate": 4.0024883695769773e-05, "loss": 1.6976, "step": 37080 }, { "epoch": 2.0, "learning_rate": 4.002217894622958e-05, "loss": 1.5085, "step": 37090 }, { "epoch": 2.0, "learning_rate": 4.0019474196689386e-05, "loss": 1.6595, "step": 37100 }, { "epoch": 2.01, "learning_rate": 4.0016769447149196e-05, "loss": 1.5913, "step": 37110 }, { "epoch": 2.01, "learning_rate": 4.0014064697609005e-05, "loss": 1.5792, "step": 37120 }, { "epoch": 2.01, "learning_rate": 4.001135994806881e-05, "loss": 1.5803, "step": 37130 }, { "epoch": 2.01, "learning_rate": 4.000865519852862e-05, "loss": 1.6591, "step": 37140 }, { "epoch": 2.01, "learning_rate": 4.000595044898843e-05, "loss": 1.6158, "step": 37150 }, { "epoch": 2.01, "learning_rate": 4.000324569944823e-05, "loss": 1.6286, "step": 37160 }, { "epoch": 2.01, "learning_rate": 4.000054094990804e-05, "loss": 1.6413, "step": 37170 }, { "epoch": 2.01, "learning_rate": 3.999783620036785e-05, "loss": 1.6406, "step": 37180 }, { "epoch": 2.01, "learning_rate": 3.999513145082765e-05, "loss": 1.5825, "step": 37190 }, { "epoch": 2.01, "learning_rate": 3.999242670128746e-05, "loss": 1.5976, "step": 37200 }, { "epoch": 2.01, "learning_rate": 3.998972195174727e-05, "loss": 1.6144, "step": 37210 }, { "epoch": 2.01, "learning_rate": 3.9987017202207075e-05, "loss": 1.6355, "step": 37220 }, { "epoch": 2.01, "learning_rate": 3.9984312452666884e-05, "loss": 1.6897, "step": 37230 }, { "epoch": 2.01, "learning_rate": 3.9981607703126694e-05, "loss": 1.6035, "step": 37240 }, { "epoch": 2.01, "learning_rate": 3.99789029535865e-05, "loss": 1.5388, "step": 37250 }, { "epoch": 2.01, "learning_rate": 3.9976198204046306e-05, "loss": 1.5718, "step": 37260 }, { "epoch": 2.01, "learning_rate": 3.9973493454506116e-05, "loss": 1.5884, "step": 37270 }, { "epoch": 2.01, "learning_rate": 3.9970788704965926e-05, "loss": 1.636, "step": 37280 }, { "epoch": 2.01, "learning_rate": 3.996808395542573e-05, "loss": 1.5514, "step": 37290 }, { "epoch": 2.02, "learning_rate": 3.996537920588554e-05, "loss": 1.6336, "step": 37300 }, { "epoch": 2.02, "learning_rate": 3.996267445634535e-05, "loss": 1.5562, "step": 37310 }, { "epoch": 2.02, "learning_rate": 3.995996970680515e-05, "loss": 1.6454, "step": 37320 }, { "epoch": 2.02, "learning_rate": 3.995726495726496e-05, "loss": 1.6097, "step": 37330 }, { "epoch": 2.02, "learning_rate": 3.995456020772477e-05, "loss": 1.6231, "step": 37340 }, { "epoch": 2.02, "learning_rate": 3.995185545818457e-05, "loss": 1.6514, "step": 37350 }, { "epoch": 2.02, "learning_rate": 3.994915070864438e-05, "loss": 1.6265, "step": 37360 }, { "epoch": 2.02, "learning_rate": 3.994644595910419e-05, "loss": 1.6607, "step": 37370 }, { "epoch": 2.02, "learning_rate": 3.9943741209563995e-05, "loss": 1.6222, "step": 37380 }, { "epoch": 2.02, "learning_rate": 3.9941036460023805e-05, "loss": 1.5602, "step": 37390 }, { "epoch": 2.02, "learning_rate": 3.9938331710483614e-05, "loss": 1.6411, "step": 37400 }, { "epoch": 2.02, "learning_rate": 3.993562696094342e-05, "loss": 1.6409, "step": 37410 }, { "epoch": 2.02, "learning_rate": 3.993292221140323e-05, "loss": 1.5553, "step": 37420 }, { "epoch": 2.02, "learning_rate": 3.993021746186304e-05, "loss": 1.6366, "step": 37430 }, { "epoch": 2.02, "learning_rate": 3.992751271232284e-05, "loss": 1.6341, "step": 37440 }, { "epoch": 2.02, "learning_rate": 3.992480796278265e-05, "loss": 1.636, "step": 37450 }, { "epoch": 2.02, "learning_rate": 3.992210321324246e-05, "loss": 1.626, "step": 37460 }, { "epoch": 2.02, "learning_rate": 3.991939846370226e-05, "loss": 1.6715, "step": 37470 }, { "epoch": 2.03, "learning_rate": 3.9916693714162065e-05, "loss": 1.6709, "step": 37480 }, { "epoch": 2.03, "learning_rate": 3.991398896462188e-05, "loss": 1.6502, "step": 37490 }, { "epoch": 2.03, "learning_rate": 3.9911284215081684e-05, "loss": 1.6644, "step": 37500 }, { "epoch": 2.03, "learning_rate": 3.9908579465541493e-05, "loss": 1.694, "step": 37510 }, { "epoch": 2.03, "learning_rate": 3.99058747160013e-05, "loss": 1.6873, "step": 37520 }, { "epoch": 2.03, "learning_rate": 3.9903169966461106e-05, "loss": 1.6632, "step": 37530 }, { "epoch": 2.03, "learning_rate": 3.9900465216920916e-05, "loss": 1.5566, "step": 37540 }, { "epoch": 2.03, "learning_rate": 3.9897760467380725e-05, "loss": 1.6293, "step": 37550 }, { "epoch": 2.03, "learning_rate": 3.989505571784053e-05, "loss": 1.5663, "step": 37560 }, { "epoch": 2.03, "learning_rate": 3.989235096830034e-05, "loss": 1.6965, "step": 37570 }, { "epoch": 2.03, "learning_rate": 3.988964621876015e-05, "loss": 1.6469, "step": 37580 }, { "epoch": 2.03, "learning_rate": 3.988694146921995e-05, "loss": 1.6202, "step": 37590 }, { "epoch": 2.03, "learning_rate": 3.988423671967976e-05, "loss": 1.5616, "step": 37600 }, { "epoch": 2.03, "learning_rate": 3.988153197013957e-05, "loss": 1.6693, "step": 37610 }, { "epoch": 2.03, "learning_rate": 3.987882722059937e-05, "loss": 1.7277, "step": 37620 }, { "epoch": 2.03, "learning_rate": 3.987612247105918e-05, "loss": 1.5934, "step": 37630 }, { "epoch": 2.03, "learning_rate": 3.987341772151899e-05, "loss": 1.7176, "step": 37640 }, { "epoch": 2.03, "learning_rate": 3.9870712971978795e-05, "loss": 1.6924, "step": 37650 }, { "epoch": 2.03, "learning_rate": 3.9868008222438604e-05, "loss": 1.7019, "step": 37660 }, { "epoch": 2.04, "learning_rate": 3.9865303472898414e-05, "loss": 1.6494, "step": 37670 }, { "epoch": 2.04, "learning_rate": 3.986259872335822e-05, "loss": 1.6089, "step": 37680 }, { "epoch": 2.04, "learning_rate": 3.9859893973818027e-05, "loss": 1.596, "step": 37690 }, { "epoch": 2.04, "learning_rate": 3.9857189224277836e-05, "loss": 1.6299, "step": 37700 }, { "epoch": 2.04, "learning_rate": 3.985448447473764e-05, "loss": 1.6363, "step": 37710 }, { "epoch": 2.04, "learning_rate": 3.985177972519745e-05, "loss": 1.6542, "step": 37720 }, { "epoch": 2.04, "learning_rate": 3.984907497565726e-05, "loss": 1.6359, "step": 37730 }, { "epoch": 2.04, "learning_rate": 3.984637022611706e-05, "loss": 1.6356, "step": 37740 }, { "epoch": 2.04, "learning_rate": 3.984366547657687e-05, "loss": 1.656, "step": 37750 }, { "epoch": 2.04, "learning_rate": 3.984096072703668e-05, "loss": 1.6797, "step": 37760 }, { "epoch": 2.04, "learning_rate": 3.983825597749648e-05, "loss": 1.6778, "step": 37770 }, { "epoch": 2.04, "learning_rate": 3.983555122795629e-05, "loss": 1.5832, "step": 37780 }, { "epoch": 2.04, "learning_rate": 3.98328464784161e-05, "loss": 1.7241, "step": 37790 }, { "epoch": 2.04, "learning_rate": 3.9830141728875906e-05, "loss": 1.5294, "step": 37800 }, { "epoch": 2.04, "learning_rate": 3.9827436979335715e-05, "loss": 1.5761, "step": 37810 }, { "epoch": 2.04, "learning_rate": 3.9824732229795525e-05, "loss": 1.6678, "step": 37820 }, { "epoch": 2.04, "learning_rate": 3.982202748025533e-05, "loss": 1.6363, "step": 37830 }, { "epoch": 2.04, "learning_rate": 3.981932273071514e-05, "loss": 1.6257, "step": 37840 }, { "epoch": 2.05, "learning_rate": 3.981661798117495e-05, "loss": 1.6087, "step": 37850 }, { "epoch": 2.05, "learning_rate": 3.981391323163475e-05, "loss": 1.6536, "step": 37860 }, { "epoch": 2.05, "learning_rate": 3.981120848209456e-05, "loss": 1.665, "step": 37870 }, { "epoch": 2.05, "learning_rate": 3.980850373255437e-05, "loss": 1.5669, "step": 37880 }, { "epoch": 2.05, "learning_rate": 3.980579898301417e-05, "loss": 1.591, "step": 37890 }, { "epoch": 2.05, "learning_rate": 3.980309423347398e-05, "loss": 1.6531, "step": 37900 }, { "epoch": 2.05, "learning_rate": 3.980038948393379e-05, "loss": 1.627, "step": 37910 }, { "epoch": 2.05, "learning_rate": 3.97976847343936e-05, "loss": 1.589, "step": 37920 }, { "epoch": 2.05, "learning_rate": 3.9794979984853404e-05, "loss": 1.6409, "step": 37930 }, { "epoch": 2.05, "learning_rate": 3.9792275235313214e-05, "loss": 1.6368, "step": 37940 }, { "epoch": 2.05, "learning_rate": 3.978957048577302e-05, "loss": 1.6147, "step": 37950 }, { "epoch": 2.05, "learning_rate": 3.9786865736232826e-05, "loss": 1.5356, "step": 37960 }, { "epoch": 2.05, "learning_rate": 3.9784160986692636e-05, "loss": 1.6213, "step": 37970 }, { "epoch": 2.05, "learning_rate": 3.9781456237152445e-05, "loss": 1.6018, "step": 37980 }, { "epoch": 2.05, "learning_rate": 3.977875148761225e-05, "loss": 1.6416, "step": 37990 }, { "epoch": 2.05, "learning_rate": 3.977604673807206e-05, "loss": 1.569, "step": 38000 }, { "epoch": 2.05, "learning_rate": 3.977334198853187e-05, "loss": 1.586, "step": 38010 }, { "epoch": 2.05, "learning_rate": 3.977063723899167e-05, "loss": 1.5635, "step": 38020 }, { "epoch": 2.05, "learning_rate": 3.976793248945147e-05, "loss": 1.5714, "step": 38030 }, { "epoch": 2.06, "learning_rate": 3.976522773991129e-05, "loss": 1.6732, "step": 38040 }, { "epoch": 2.06, "learning_rate": 3.976252299037109e-05, "loss": 1.5987, "step": 38050 }, { "epoch": 2.06, "learning_rate": 3.9759818240830895e-05, "loss": 1.6376, "step": 38060 }, { "epoch": 2.06, "learning_rate": 3.975711349129071e-05, "loss": 1.5972, "step": 38070 }, { "epoch": 2.06, "learning_rate": 3.9754408741750515e-05, "loss": 1.6228, "step": 38080 }, { "epoch": 2.06, "learning_rate": 3.975170399221032e-05, "loss": 1.7167, "step": 38090 }, { "epoch": 2.06, "learning_rate": 3.9748999242670134e-05, "loss": 1.6197, "step": 38100 }, { "epoch": 2.06, "learning_rate": 3.974629449312994e-05, "loss": 1.5234, "step": 38110 }, { "epoch": 2.06, "learning_rate": 3.974358974358974e-05, "loss": 1.6266, "step": 38120 }, { "epoch": 2.06, "learning_rate": 3.9740884994049556e-05, "loss": 1.6371, "step": 38130 }, { "epoch": 2.06, "learning_rate": 3.973818024450936e-05, "loss": 1.5504, "step": 38140 }, { "epoch": 2.06, "learning_rate": 3.973547549496917e-05, "loss": 1.6623, "step": 38150 }, { "epoch": 2.06, "learning_rate": 3.973277074542898e-05, "loss": 1.5904, "step": 38160 }, { "epoch": 2.06, "learning_rate": 3.973006599588878e-05, "loss": 1.6431, "step": 38170 }, { "epoch": 2.06, "learning_rate": 3.972736124634859e-05, "loss": 1.6201, "step": 38180 }, { "epoch": 2.06, "learning_rate": 3.97246564968084e-05, "loss": 1.5779, "step": 38190 }, { "epoch": 2.06, "learning_rate": 3.9721951747268203e-05, "loss": 1.6214, "step": 38200 }, { "epoch": 2.06, "learning_rate": 3.971924699772801e-05, "loss": 1.5934, "step": 38210 }, { "epoch": 2.07, "learning_rate": 3.971654224818782e-05, "loss": 1.5903, "step": 38220 }, { "epoch": 2.07, "learning_rate": 3.9713837498647626e-05, "loss": 1.5957, "step": 38230 }, { "epoch": 2.07, "learning_rate": 3.9711132749107435e-05, "loss": 1.6529, "step": 38240 }, { "epoch": 2.07, "learning_rate": 3.9708427999567245e-05, "loss": 1.643, "step": 38250 }, { "epoch": 2.07, "learning_rate": 3.970572325002705e-05, "loss": 1.6003, "step": 38260 }, { "epoch": 2.07, "learning_rate": 3.970301850048686e-05, "loss": 1.6888, "step": 38270 }, { "epoch": 2.07, "learning_rate": 3.970031375094667e-05, "loss": 1.6613, "step": 38280 }, { "epoch": 2.07, "learning_rate": 3.969760900140647e-05, "loss": 1.6708, "step": 38290 }, { "epoch": 2.07, "learning_rate": 3.969490425186628e-05, "loss": 1.58, "step": 38300 }, { "epoch": 2.07, "learning_rate": 3.969219950232609e-05, "loss": 1.6179, "step": 38310 }, { "epoch": 2.07, "learning_rate": 3.968949475278589e-05, "loss": 1.5963, "step": 38320 }, { "epoch": 2.07, "learning_rate": 3.96867900032457e-05, "loss": 1.6114, "step": 38330 }, { "epoch": 2.07, "learning_rate": 3.968408525370551e-05, "loss": 1.6206, "step": 38340 }, { "epoch": 2.07, "learning_rate": 3.9681380504165314e-05, "loss": 1.654, "step": 38350 }, { "epoch": 2.07, "learning_rate": 3.9678675754625124e-05, "loss": 1.6404, "step": 38360 }, { "epoch": 2.07, "learning_rate": 3.9675971005084934e-05, "loss": 1.6559, "step": 38370 }, { "epoch": 2.07, "learning_rate": 3.9673266255544736e-05, "loss": 1.7075, "step": 38380 }, { "epoch": 2.07, "learning_rate": 3.9670561506004546e-05, "loss": 1.669, "step": 38390 }, { "epoch": 2.07, "learning_rate": 3.9667856756464356e-05, "loss": 1.6556, "step": 38400 }, { "epoch": 2.08, "learning_rate": 3.966515200692416e-05, "loss": 1.6334, "step": 38410 }, { "epoch": 2.08, "learning_rate": 3.966244725738397e-05, "loss": 1.7072, "step": 38420 }, { "epoch": 2.08, "learning_rate": 3.965974250784378e-05, "loss": 1.5927, "step": 38430 }, { "epoch": 2.08, "learning_rate": 3.965703775830358e-05, "loss": 1.6135, "step": 38440 }, { "epoch": 2.08, "learning_rate": 3.965433300876339e-05, "loss": 1.7003, "step": 38450 }, { "epoch": 2.08, "learning_rate": 3.96516282592232e-05, "loss": 1.6204, "step": 38460 }, { "epoch": 2.08, "learning_rate": 3.9648923509683e-05, "loss": 1.6827, "step": 38470 }, { "epoch": 2.08, "learning_rate": 3.964621876014281e-05, "loss": 1.6084, "step": 38480 }, { "epoch": 2.08, "learning_rate": 3.964351401060262e-05, "loss": 1.655, "step": 38490 }, { "epoch": 2.08, "learning_rate": 3.9640809261062425e-05, "loss": 1.6123, "step": 38500 }, { "epoch": 2.08, "learning_rate": 3.9638104511522235e-05, "loss": 1.5838, "step": 38510 }, { "epoch": 2.08, "learning_rate": 3.9635399761982044e-05, "loss": 1.6858, "step": 38520 }, { "epoch": 2.08, "learning_rate": 3.963269501244185e-05, "loss": 1.7517, "step": 38530 }, { "epoch": 2.08, "learning_rate": 3.962999026290166e-05, "loss": 1.6165, "step": 38540 }, { "epoch": 2.08, "learning_rate": 3.9627285513361467e-05, "loss": 1.6126, "step": 38550 }, { "epoch": 2.08, "learning_rate": 3.9624580763821276e-05, "loss": 1.6034, "step": 38560 }, { "epoch": 2.08, "learning_rate": 3.962187601428108e-05, "loss": 1.598, "step": 38570 }, { "epoch": 2.08, "learning_rate": 3.961917126474088e-05, "loss": 1.6118, "step": 38580 }, { "epoch": 2.09, "learning_rate": 3.96164665152007e-05, "loss": 1.6597, "step": 38590 }, { "epoch": 2.09, "learning_rate": 3.96137617656605e-05, "loss": 1.6822, "step": 38600 }, { "epoch": 2.09, "learning_rate": 3.9611057016120304e-05, "loss": 1.5813, "step": 38610 }, { "epoch": 2.09, "learning_rate": 3.960835226658012e-05, "loss": 1.6256, "step": 38620 }, { "epoch": 2.09, "learning_rate": 3.9605647517039923e-05, "loss": 1.6767, "step": 38630 }, { "epoch": 2.09, "learning_rate": 3.9602942767499726e-05, "loss": 1.5746, "step": 38640 }, { "epoch": 2.09, "learning_rate": 3.960023801795954e-05, "loss": 1.5688, "step": 38650 }, { "epoch": 2.09, "learning_rate": 3.9597533268419346e-05, "loss": 1.6181, "step": 38660 }, { "epoch": 2.09, "learning_rate": 3.959482851887915e-05, "loss": 1.6243, "step": 38670 }, { "epoch": 2.09, "learning_rate": 3.9592123769338965e-05, "loss": 1.6084, "step": 38680 }, { "epoch": 2.09, "learning_rate": 3.958941901979877e-05, "loss": 1.671, "step": 38690 }, { "epoch": 2.09, "learning_rate": 3.958671427025857e-05, "loss": 1.6351, "step": 38700 }, { "epoch": 2.09, "learning_rate": 3.958400952071839e-05, "loss": 1.5834, "step": 38710 }, { "epoch": 2.09, "learning_rate": 3.958130477117819e-05, "loss": 1.6385, "step": 38720 }, { "epoch": 2.09, "learning_rate": 3.957860002163799e-05, "loss": 1.6538, "step": 38730 }, { "epoch": 2.09, "learning_rate": 3.957589527209781e-05, "loss": 1.6492, "step": 38740 }, { "epoch": 2.09, "learning_rate": 3.957319052255761e-05, "loss": 1.6214, "step": 38750 }, { "epoch": 2.09, "learning_rate": 3.9570485773017415e-05, "loss": 1.5603, "step": 38760 }, { "epoch": 2.09, "learning_rate": 3.956778102347723e-05, "loss": 1.5126, "step": 38770 }, { "epoch": 2.1, "learning_rate": 3.9565076273937034e-05, "loss": 1.6538, "step": 38780 }, { "epoch": 2.1, "learning_rate": 3.9562371524396844e-05, "loss": 1.6195, "step": 38790 }, { "epoch": 2.1, "learning_rate": 3.9559666774856654e-05, "loss": 1.6627, "step": 38800 }, { "epoch": 2.1, "learning_rate": 3.9556962025316456e-05, "loss": 1.588, "step": 38810 }, { "epoch": 2.1, "learning_rate": 3.9554257275776266e-05, "loss": 1.6331, "step": 38820 }, { "epoch": 2.1, "learning_rate": 3.9551552526236076e-05, "loss": 1.6442, "step": 38830 }, { "epoch": 2.1, "learning_rate": 3.954884777669588e-05, "loss": 1.5968, "step": 38840 }, { "epoch": 2.1, "learning_rate": 3.954614302715569e-05, "loss": 1.6272, "step": 38850 }, { "epoch": 2.1, "learning_rate": 3.95434382776155e-05, "loss": 1.6568, "step": 38860 }, { "epoch": 2.1, "learning_rate": 3.95407335280753e-05, "loss": 1.625, "step": 38870 }, { "epoch": 2.1, "learning_rate": 3.953802877853511e-05, "loss": 1.635, "step": 38880 }, { "epoch": 2.1, "learning_rate": 3.953532402899492e-05, "loss": 1.7207, "step": 38890 }, { "epoch": 2.1, "learning_rate": 3.953261927945472e-05, "loss": 1.5645, "step": 38900 }, { "epoch": 2.1, "learning_rate": 3.952991452991453e-05, "loss": 1.6266, "step": 38910 }, { "epoch": 2.1, "learning_rate": 3.952720978037434e-05, "loss": 1.7219, "step": 38920 }, { "epoch": 2.1, "learning_rate": 3.9524505030834145e-05, "loss": 1.633, "step": 38930 }, { "epoch": 2.1, "learning_rate": 3.9521800281293955e-05, "loss": 1.6705, "step": 38940 }, { "epoch": 2.1, "learning_rate": 3.9519095531753764e-05, "loss": 1.6216, "step": 38950 }, { "epoch": 2.11, "learning_rate": 3.951639078221357e-05, "loss": 1.6205, "step": 38960 }, { "epoch": 2.11, "learning_rate": 3.951368603267338e-05, "loss": 1.6169, "step": 38970 }, { "epoch": 2.11, "learning_rate": 3.9510981283133187e-05, "loss": 1.6518, "step": 38980 }, { "epoch": 2.11, "learning_rate": 3.950827653359299e-05, "loss": 1.6831, "step": 38990 }, { "epoch": 2.11, "learning_rate": 3.95055717840528e-05, "loss": 1.6368, "step": 39000 }, { "epoch": 2.11, "learning_rate": 3.950286703451261e-05, "loss": 1.5671, "step": 39010 }, { "epoch": 2.11, "learning_rate": 3.950016228497241e-05, "loss": 1.7436, "step": 39020 }, { "epoch": 2.11, "learning_rate": 3.949745753543222e-05, "loss": 1.6993, "step": 39030 }, { "epoch": 2.11, "learning_rate": 3.949475278589203e-05, "loss": 1.6628, "step": 39040 }, { "epoch": 2.11, "learning_rate": 3.9492048036351834e-05, "loss": 1.5337, "step": 39050 }, { "epoch": 2.11, "learning_rate": 3.9489343286811643e-05, "loss": 1.5263, "step": 39060 }, { "epoch": 2.11, "learning_rate": 3.948663853727145e-05, "loss": 1.6449, "step": 39070 }, { "epoch": 2.11, "learning_rate": 3.9483933787731256e-05, "loss": 1.6606, "step": 39080 }, { "epoch": 2.11, "learning_rate": 3.9481229038191066e-05, "loss": 1.6129, "step": 39090 }, { "epoch": 2.11, "learning_rate": 3.9478524288650875e-05, "loss": 1.6468, "step": 39100 }, { "epoch": 2.11, "learning_rate": 3.947581953911068e-05, "loss": 1.6084, "step": 39110 }, { "epoch": 2.11, "learning_rate": 3.947311478957049e-05, "loss": 1.6748, "step": 39120 }, { "epoch": 2.11, "learning_rate": 3.94704100400303e-05, "loss": 1.6152, "step": 39130 }, { "epoch": 2.11, "learning_rate": 3.94677052904901e-05, "loss": 1.5728, "step": 39140 }, { "epoch": 2.12, "learning_rate": 3.946500054094991e-05, "loss": 1.6155, "step": 39150 }, { "epoch": 2.12, "learning_rate": 3.946229579140971e-05, "loss": 1.5892, "step": 39160 }, { "epoch": 2.12, "learning_rate": 3.945959104186952e-05, "loss": 1.5978, "step": 39170 }, { "epoch": 2.12, "learning_rate": 3.945688629232933e-05, "loss": 1.5258, "step": 39180 }, { "epoch": 2.12, "learning_rate": 3.9454181542789135e-05, "loss": 1.5724, "step": 39190 }, { "epoch": 2.12, "learning_rate": 3.945147679324895e-05, "loss": 1.5596, "step": 39200 }, { "epoch": 2.12, "learning_rate": 3.9448772043708754e-05, "loss": 1.6789, "step": 39210 }, { "epoch": 2.12, "learning_rate": 3.944606729416856e-05, "loss": 1.6298, "step": 39220 }, { "epoch": 2.12, "learning_rate": 3.9443362544628374e-05, "loss": 1.6412, "step": 39230 }, { "epoch": 2.12, "learning_rate": 3.9440657795088176e-05, "loss": 1.4981, "step": 39240 }, { "epoch": 2.12, "learning_rate": 3.943795304554798e-05, "loss": 1.6397, "step": 39250 }, { "epoch": 2.12, "learning_rate": 3.9435248296007796e-05, "loss": 1.5789, "step": 39260 }, { "epoch": 2.12, "learning_rate": 3.94325435464676e-05, "loss": 1.6206, "step": 39270 }, { "epoch": 2.12, "learning_rate": 3.94298387969274e-05, "loss": 1.6748, "step": 39280 }, { "epoch": 2.12, "learning_rate": 3.942713404738722e-05, "loss": 1.6061, "step": 39290 }, { "epoch": 2.12, "learning_rate": 3.942442929784702e-05, "loss": 1.6561, "step": 39300 }, { "epoch": 2.12, "learning_rate": 3.9421724548306824e-05, "loss": 1.6272, "step": 39310 }, { "epoch": 2.12, "learning_rate": 3.941901979876664e-05, "loss": 1.5823, "step": 39320 }, { "epoch": 2.13, "learning_rate": 3.941631504922644e-05, "loss": 1.5935, "step": 39330 }, { "epoch": 2.13, "learning_rate": 3.9413610299686246e-05, "loss": 1.543, "step": 39340 }, { "epoch": 2.13, "learning_rate": 3.941090555014606e-05, "loss": 1.5741, "step": 39350 }, { "epoch": 2.13, "learning_rate": 3.9408200800605865e-05, "loss": 1.651, "step": 39360 }, { "epoch": 2.13, "learning_rate": 3.940549605106567e-05, "loss": 1.5894, "step": 39370 }, { "epoch": 2.13, "learning_rate": 3.9402791301525484e-05, "loss": 1.6258, "step": 39380 }, { "epoch": 2.13, "learning_rate": 3.940008655198529e-05, "loss": 1.7344, "step": 39390 }, { "epoch": 2.13, "learning_rate": 3.93973818024451e-05, "loss": 1.5469, "step": 39400 }, { "epoch": 2.13, "learning_rate": 3.9394677052904907e-05, "loss": 1.6197, "step": 39410 }, { "epoch": 2.13, "learning_rate": 3.939197230336471e-05, "loss": 1.689, "step": 39420 }, { "epoch": 2.13, "learning_rate": 3.938926755382452e-05, "loss": 1.6412, "step": 39430 }, { "epoch": 2.13, "learning_rate": 3.938656280428433e-05, "loss": 1.6917, "step": 39440 }, { "epoch": 2.13, "learning_rate": 3.938385805474413e-05, "loss": 1.5436, "step": 39450 }, { "epoch": 2.13, "learning_rate": 3.938115330520394e-05, "loss": 1.6824, "step": 39460 }, { "epoch": 2.13, "learning_rate": 3.937844855566375e-05, "loss": 1.6653, "step": 39470 }, { "epoch": 2.13, "learning_rate": 3.9375743806123554e-05, "loss": 1.6442, "step": 39480 }, { "epoch": 2.13, "learning_rate": 3.9373039056583363e-05, "loss": 1.6813, "step": 39490 }, { "epoch": 2.13, "learning_rate": 3.937033430704317e-05, "loss": 1.6765, "step": 39500 }, { "epoch": 2.13, "learning_rate": 3.9367629557502976e-05, "loss": 1.5388, "step": 39510 }, { "epoch": 2.14, "learning_rate": 3.9364924807962786e-05, "loss": 1.5627, "step": 39520 }, { "epoch": 2.14, "learning_rate": 3.9362220058422595e-05, "loss": 1.6021, "step": 39530 }, { "epoch": 2.14, "learning_rate": 3.93595153088824e-05, "loss": 1.6117, "step": 39540 }, { "epoch": 2.14, "learning_rate": 3.935681055934221e-05, "loss": 1.6168, "step": 39550 }, { "epoch": 2.14, "learning_rate": 3.935410580980202e-05, "loss": 1.5674, "step": 39560 }, { "epoch": 2.14, "learning_rate": 3.935140106026182e-05, "loss": 1.6597, "step": 39570 }, { "epoch": 2.14, "learning_rate": 3.934869631072163e-05, "loss": 1.6837, "step": 39580 }, { "epoch": 2.14, "learning_rate": 3.934599156118144e-05, "loss": 1.4885, "step": 39590 }, { "epoch": 2.14, "learning_rate": 3.934328681164124e-05, "loss": 1.5817, "step": 39600 }, { "epoch": 2.14, "learning_rate": 3.934058206210105e-05, "loss": 1.555, "step": 39610 }, { "epoch": 2.14, "learning_rate": 3.933787731256086e-05, "loss": 1.6416, "step": 39620 }, { "epoch": 2.14, "learning_rate": 3.9335172563020665e-05, "loss": 1.6296, "step": 39630 }, { "epoch": 2.14, "learning_rate": 3.9332467813480474e-05, "loss": 1.6401, "step": 39640 }, { "epoch": 2.14, "learning_rate": 3.9329763063940284e-05, "loss": 1.6659, "step": 39650 }, { "epoch": 2.14, "learning_rate": 3.932705831440009e-05, "loss": 1.6269, "step": 39660 }, { "epoch": 2.14, "learning_rate": 3.9324353564859896e-05, "loss": 1.6021, "step": 39670 }, { "epoch": 2.14, "learning_rate": 3.9321648815319706e-05, "loss": 1.7047, "step": 39680 }, { "epoch": 2.14, "learning_rate": 3.931894406577951e-05, "loss": 1.6287, "step": 39690 }, { "epoch": 2.15, "learning_rate": 3.931623931623932e-05, "loss": 1.5615, "step": 39700 }, { "epoch": 2.15, "learning_rate": 3.931353456669912e-05, "loss": 1.5868, "step": 39710 }, { "epoch": 2.15, "learning_rate": 3.931082981715893e-05, "loss": 1.677, "step": 39720 }, { "epoch": 2.15, "learning_rate": 3.930812506761874e-05, "loss": 1.5926, "step": 39730 }, { "epoch": 2.15, "learning_rate": 3.9305420318078544e-05, "loss": 1.6664, "step": 39740 }, { "epoch": 2.15, "learning_rate": 3.930271556853835e-05, "loss": 1.6489, "step": 39750 }, { "epoch": 2.15, "learning_rate": 3.930001081899816e-05, "loss": 1.6142, "step": 39760 }, { "epoch": 2.15, "learning_rate": 3.9297306069457966e-05, "loss": 1.6061, "step": 39770 }, { "epoch": 2.15, "learning_rate": 3.9294601319917776e-05, "loss": 1.6388, "step": 39780 }, { "epoch": 2.15, "learning_rate": 3.9291896570377585e-05, "loss": 1.6288, "step": 39790 }, { "epoch": 2.15, "learning_rate": 3.928919182083739e-05, "loss": 1.6916, "step": 39800 }, { "epoch": 2.15, "learning_rate": 3.9286487071297204e-05, "loss": 1.6417, "step": 39810 }, { "epoch": 2.15, "learning_rate": 3.928378232175701e-05, "loss": 1.6835, "step": 39820 }, { "epoch": 2.15, "learning_rate": 3.928107757221681e-05, "loss": 1.5599, "step": 39830 }, { "epoch": 2.15, "learning_rate": 3.927837282267663e-05, "loss": 1.5527, "step": 39840 }, { "epoch": 2.15, "learning_rate": 3.927566807313643e-05, "loss": 1.6261, "step": 39850 }, { "epoch": 2.15, "learning_rate": 3.927296332359623e-05, "loss": 1.5963, "step": 39860 }, { "epoch": 2.15, "learning_rate": 3.927025857405605e-05, "loss": 1.6619, "step": 39870 }, { "epoch": 2.15, "learning_rate": 3.926755382451585e-05, "loss": 1.627, "step": 39880 }, { "epoch": 2.16, "learning_rate": 3.9264849074975655e-05, "loss": 1.6078, "step": 39890 }, { "epoch": 2.16, "learning_rate": 3.926214432543547e-05, "loss": 1.6041, "step": 39900 }, { "epoch": 2.16, "learning_rate": 3.9259439575895274e-05, "loss": 1.6502, "step": 39910 }, { "epoch": 2.16, "learning_rate": 3.925673482635508e-05, "loss": 1.6379, "step": 39920 }, { "epoch": 2.16, "learning_rate": 3.925403007681489e-05, "loss": 1.6294, "step": 39930 }, { "epoch": 2.16, "learning_rate": 3.9251325327274696e-05, "loss": 1.5853, "step": 39940 }, { "epoch": 2.16, "learning_rate": 3.92486205777345e-05, "loss": 1.6434, "step": 39950 }, { "epoch": 2.16, "learning_rate": 3.9245915828194315e-05, "loss": 1.6888, "step": 39960 }, { "epoch": 2.16, "learning_rate": 3.924321107865412e-05, "loss": 1.6013, "step": 39970 }, { "epoch": 2.16, "learning_rate": 3.924050632911392e-05, "loss": 1.5767, "step": 39980 }, { "epoch": 2.16, "learning_rate": 3.923780157957374e-05, "loss": 1.599, "step": 39990 }, { "epoch": 2.16, "learning_rate": 3.923509683003354e-05, "loss": 1.5433, "step": 40000 }, { "epoch": 2.16, "learning_rate": 3.923239208049334e-05, "loss": 1.6264, "step": 40010 }, { "epoch": 2.16, "learning_rate": 3.922968733095316e-05, "loss": 1.6118, "step": 40020 }, { "epoch": 2.16, "learning_rate": 3.922698258141296e-05, "loss": 1.699, "step": 40030 }, { "epoch": 2.16, "learning_rate": 3.922427783187277e-05, "loss": 1.693, "step": 40040 }, { "epoch": 2.16, "learning_rate": 3.922157308233258e-05, "loss": 1.5836, "step": 40050 }, { "epoch": 2.16, "learning_rate": 3.9218868332792385e-05, "loss": 1.691, "step": 40060 }, { "epoch": 2.17, "learning_rate": 3.9216163583252194e-05, "loss": 1.7399, "step": 40070 }, { "epoch": 2.17, "learning_rate": 3.9213458833712004e-05, "loss": 1.6158, "step": 40080 }, { "epoch": 2.17, "learning_rate": 3.921075408417181e-05, "loss": 1.6899, "step": 40090 }, { "epoch": 2.17, "learning_rate": 3.9208049334631616e-05, "loss": 1.6712, "step": 40100 }, { "epoch": 2.17, "learning_rate": 3.9205344585091426e-05, "loss": 1.5977, "step": 40110 }, { "epoch": 2.17, "learning_rate": 3.920263983555123e-05, "loss": 1.6665, "step": 40120 }, { "epoch": 2.17, "learning_rate": 3.919993508601104e-05, "loss": 1.5828, "step": 40130 }, { "epoch": 2.17, "learning_rate": 3.919723033647085e-05, "loss": 1.5853, "step": 40140 }, { "epoch": 2.17, "learning_rate": 3.919452558693065e-05, "loss": 1.6155, "step": 40150 }, { "epoch": 2.17, "learning_rate": 3.919182083739046e-05, "loss": 1.6347, "step": 40160 }, { "epoch": 2.17, "learning_rate": 3.918911608785027e-05, "loss": 1.7045, "step": 40170 }, { "epoch": 2.17, "learning_rate": 3.918641133831007e-05, "loss": 1.6171, "step": 40180 }, { "epoch": 2.17, "learning_rate": 3.918370658876988e-05, "loss": 1.5927, "step": 40190 }, { "epoch": 2.17, "learning_rate": 3.918100183922969e-05, "loss": 1.6691, "step": 40200 }, { "epoch": 2.17, "learning_rate": 3.9178297089689496e-05, "loss": 1.6576, "step": 40210 }, { "epoch": 2.17, "learning_rate": 3.9175592340149305e-05, "loss": 1.6393, "step": 40220 }, { "epoch": 2.17, "learning_rate": 3.9172887590609115e-05, "loss": 1.6473, "step": 40230 }, { "epoch": 2.17, "learning_rate": 3.917018284106892e-05, "loss": 1.571, "step": 40240 }, { "epoch": 2.17, "learning_rate": 3.916747809152873e-05, "loss": 1.667, "step": 40250 }, { "epoch": 2.18, "learning_rate": 3.916477334198853e-05, "loss": 1.5744, "step": 40260 }, { "epoch": 2.18, "learning_rate": 3.916206859244834e-05, "loss": 1.5938, "step": 40270 }, { "epoch": 2.18, "learning_rate": 3.915936384290815e-05, "loss": 1.6513, "step": 40280 }, { "epoch": 2.18, "learning_rate": 3.915665909336795e-05, "loss": 1.6529, "step": 40290 }, { "epoch": 2.18, "learning_rate": 3.915395434382776e-05, "loss": 1.5072, "step": 40300 }, { "epoch": 2.18, "learning_rate": 3.915124959428757e-05, "loss": 1.5666, "step": 40310 }, { "epoch": 2.18, "learning_rate": 3.9148544844747375e-05, "loss": 1.5847, "step": 40320 }, { "epoch": 2.18, "learning_rate": 3.9145840095207184e-05, "loss": 1.5937, "step": 40330 }, { "epoch": 2.18, "learning_rate": 3.9143135345666994e-05, "loss": 1.6573, "step": 40340 }, { "epoch": 2.18, "learning_rate": 3.91404305961268e-05, "loss": 1.5877, "step": 40350 }, { "epoch": 2.18, "learning_rate": 3.9137725846586606e-05, "loss": 1.6641, "step": 40360 }, { "epoch": 2.18, "learning_rate": 3.9135021097046416e-05, "loss": 1.588, "step": 40370 }, { "epoch": 2.18, "learning_rate": 3.913231634750622e-05, "loss": 1.649, "step": 40380 }, { "epoch": 2.18, "learning_rate": 3.912961159796603e-05, "loss": 1.6445, "step": 40390 }, { "epoch": 2.18, "learning_rate": 3.912690684842584e-05, "loss": 1.6196, "step": 40400 }, { "epoch": 2.18, "learning_rate": 3.912420209888564e-05, "loss": 1.5935, "step": 40410 }, { "epoch": 2.18, "learning_rate": 3.912149734934545e-05, "loss": 1.6558, "step": 40420 }, { "epoch": 2.18, "learning_rate": 3.911879259980526e-05, "loss": 1.6212, "step": 40430 }, { "epoch": 2.19, "learning_rate": 3.911608785026506e-05, "loss": 1.5943, "step": 40440 }, { "epoch": 2.19, "learning_rate": 3.911338310072488e-05, "loss": 1.6181, "step": 40450 }, { "epoch": 2.19, "learning_rate": 3.911067835118468e-05, "loss": 1.6495, "step": 40460 }, { "epoch": 2.19, "learning_rate": 3.9107973601644485e-05, "loss": 1.6227, "step": 40470 }, { "epoch": 2.19, "learning_rate": 3.91052688521043e-05, "loss": 1.5991, "step": 40480 }, { "epoch": 2.19, "learning_rate": 3.9102564102564105e-05, "loss": 1.5944, "step": 40490 }, { "epoch": 2.19, "learning_rate": 3.909985935302391e-05, "loss": 1.6995, "step": 40500 }, { "epoch": 2.19, "learning_rate": 3.9097154603483724e-05, "loss": 1.6731, "step": 40510 }, { "epoch": 2.19, "learning_rate": 3.909444985394353e-05, "loss": 1.6779, "step": 40520 }, { "epoch": 2.19, "learning_rate": 3.909174510440333e-05, "loss": 1.5911, "step": 40530 }, { "epoch": 2.19, "learning_rate": 3.9089040354863146e-05, "loss": 1.6325, "step": 40540 }, { "epoch": 2.19, "learning_rate": 3.908633560532295e-05, "loss": 1.6217, "step": 40550 }, { "epoch": 2.19, "learning_rate": 3.908363085578275e-05, "loss": 1.6131, "step": 40560 }, { "epoch": 2.19, "learning_rate": 3.908092610624257e-05, "loss": 1.5846, "step": 40570 }, { "epoch": 2.19, "learning_rate": 3.907822135670237e-05, "loss": 1.6582, "step": 40580 }, { "epoch": 2.19, "learning_rate": 3.9075516607162174e-05, "loss": 1.5956, "step": 40590 }, { "epoch": 2.19, "learning_rate": 3.907281185762199e-05, "loss": 1.6509, "step": 40600 }, { "epoch": 2.19, "learning_rate": 3.907010710808179e-05, "loss": 1.6055, "step": 40610 }, { "epoch": 2.19, "learning_rate": 3.9067402358541596e-05, "loss": 1.5789, "step": 40620 }, { "epoch": 2.2, "learning_rate": 3.906469760900141e-05, "loss": 1.5498, "step": 40630 }, { "epoch": 2.2, "learning_rate": 3.9061992859461216e-05, "loss": 1.5929, "step": 40640 }, { "epoch": 2.2, "learning_rate": 3.905928810992102e-05, "loss": 1.5741, "step": 40650 }, { "epoch": 2.2, "learning_rate": 3.9056583360380835e-05, "loss": 1.6844, "step": 40660 }, { "epoch": 2.2, "learning_rate": 3.905387861084064e-05, "loss": 1.6093, "step": 40670 }, { "epoch": 2.2, "learning_rate": 3.905117386130045e-05, "loss": 1.6083, "step": 40680 }, { "epoch": 2.2, "learning_rate": 3.904846911176026e-05, "loss": 1.5505, "step": 40690 }, { "epoch": 2.2, "learning_rate": 3.904576436222006e-05, "loss": 1.6067, "step": 40700 }, { "epoch": 2.2, "learning_rate": 3.904305961267987e-05, "loss": 1.641, "step": 40710 }, { "epoch": 2.2, "learning_rate": 3.904035486313968e-05, "loss": 1.5925, "step": 40720 }, { "epoch": 2.2, "learning_rate": 3.903765011359948e-05, "loss": 1.6493, "step": 40730 }, { "epoch": 2.2, "learning_rate": 3.903494536405929e-05, "loss": 1.6918, "step": 40740 }, { "epoch": 2.2, "learning_rate": 3.90322406145191e-05, "loss": 1.588, "step": 40750 }, { "epoch": 2.2, "learning_rate": 3.9029535864978904e-05, "loss": 1.7248, "step": 40760 }, { "epoch": 2.2, "learning_rate": 3.9026831115438714e-05, "loss": 1.5751, "step": 40770 }, { "epoch": 2.2, "learning_rate": 3.9024126365898524e-05, "loss": 1.7092, "step": 40780 }, { "epoch": 2.2, "learning_rate": 3.9021421616358326e-05, "loss": 1.6292, "step": 40790 }, { "epoch": 2.2, "learning_rate": 3.9018716866818136e-05, "loss": 1.5895, "step": 40800 }, { "epoch": 2.21, "learning_rate": 3.9016012117277946e-05, "loss": 1.6015, "step": 40810 }, { "epoch": 2.21, "learning_rate": 3.901330736773775e-05, "loss": 1.5009, "step": 40820 }, { "epoch": 2.21, "learning_rate": 3.901060261819756e-05, "loss": 1.6025, "step": 40830 }, { "epoch": 2.21, "learning_rate": 3.900789786865736e-05, "loss": 1.688, "step": 40840 }, { "epoch": 2.21, "learning_rate": 3.900519311911717e-05, "loss": 1.6431, "step": 40850 }, { "epoch": 2.21, "learning_rate": 3.900248836957698e-05, "loss": 1.5957, "step": 40860 }, { "epoch": 2.21, "learning_rate": 3.899978362003678e-05, "loss": 1.6074, "step": 40870 }, { "epoch": 2.21, "learning_rate": 3.899707887049659e-05, "loss": 1.6523, "step": 40880 }, { "epoch": 2.21, "learning_rate": 3.89943741209564e-05, "loss": 1.6899, "step": 40890 }, { "epoch": 2.21, "learning_rate": 3.8991669371416205e-05, "loss": 1.647, "step": 40900 }, { "epoch": 2.21, "learning_rate": 3.8988964621876015e-05, "loss": 1.6355, "step": 40910 }, { "epoch": 2.21, "learning_rate": 3.8986259872335825e-05, "loss": 1.6494, "step": 40920 }, { "epoch": 2.21, "learning_rate": 3.898355512279563e-05, "loss": 1.6639, "step": 40930 }, { "epoch": 2.21, "learning_rate": 3.898085037325544e-05, "loss": 1.6929, "step": 40940 }, { "epoch": 2.21, "learning_rate": 3.897814562371525e-05, "loss": 1.6429, "step": 40950 }, { "epoch": 2.21, "learning_rate": 3.897544087417505e-05, "loss": 1.683, "step": 40960 }, { "epoch": 2.21, "learning_rate": 3.897273612463486e-05, "loss": 1.6801, "step": 40970 }, { "epoch": 2.21, "learning_rate": 3.897003137509467e-05, "loss": 1.6337, "step": 40980 }, { "epoch": 2.21, "learning_rate": 3.896732662555447e-05, "loss": 1.7451, "step": 40990 }, { "epoch": 2.22, "learning_rate": 3.896462187601428e-05, "loss": 1.6138, "step": 41000 }, { "epoch": 2.22, "learning_rate": 3.896191712647409e-05, "loss": 1.629, "step": 41010 }, { "epoch": 2.22, "learning_rate": 3.8959212376933894e-05, "loss": 1.5337, "step": 41020 }, { "epoch": 2.22, "learning_rate": 3.8956507627393704e-05, "loss": 1.5371, "step": 41030 }, { "epoch": 2.22, "learning_rate": 3.8953802877853513e-05, "loss": 1.5935, "step": 41040 }, { "epoch": 2.22, "learning_rate": 3.8951098128313316e-05, "loss": 1.6789, "step": 41050 }, { "epoch": 2.22, "learning_rate": 3.8948393378773126e-05, "loss": 1.6807, "step": 41060 }, { "epoch": 2.22, "learning_rate": 3.8945688629232936e-05, "loss": 1.6373, "step": 41070 }, { "epoch": 2.22, "learning_rate": 3.894298387969274e-05, "loss": 1.5884, "step": 41080 }, { "epoch": 2.22, "learning_rate": 3.8940279130152555e-05, "loss": 1.6344, "step": 41090 }, { "epoch": 2.22, "learning_rate": 3.893757438061236e-05, "loss": 1.6667, "step": 41100 }, { "epoch": 2.22, "learning_rate": 3.893486963107216e-05, "loss": 1.5763, "step": 41110 }, { "epoch": 2.22, "learning_rate": 3.893216488153198e-05, "loss": 1.6465, "step": 41120 }, { "epoch": 2.22, "learning_rate": 3.892946013199178e-05, "loss": 1.5916, "step": 41130 }, { "epoch": 2.22, "learning_rate": 3.892675538245158e-05, "loss": 1.6302, "step": 41140 }, { "epoch": 2.22, "learning_rate": 3.89240506329114e-05, "loss": 1.6467, "step": 41150 }, { "epoch": 2.22, "learning_rate": 3.89213458833712e-05, "loss": 1.5434, "step": 41160 }, { "epoch": 2.22, "learning_rate": 3.8918641133831005e-05, "loss": 1.6047, "step": 41170 }, { "epoch": 2.23, "learning_rate": 3.891593638429082e-05, "loss": 1.6582, "step": 41180 }, { "epoch": 2.23, "learning_rate": 3.8913231634750624e-05, "loss": 1.6857, "step": 41190 }, { "epoch": 2.23, "learning_rate": 3.891052688521043e-05, "loss": 1.7303, "step": 41200 }, { "epoch": 2.23, "learning_rate": 3.8907822135670244e-05, "loss": 1.6597, "step": 41210 }, { "epoch": 2.23, "learning_rate": 3.8905117386130046e-05, "loss": 1.6329, "step": 41220 }, { "epoch": 2.23, "learning_rate": 3.890241263658985e-05, "loss": 1.6514, "step": 41230 }, { "epoch": 2.23, "learning_rate": 3.8899707887049666e-05, "loss": 1.7029, "step": 41240 }, { "epoch": 2.23, "learning_rate": 3.889700313750947e-05, "loss": 1.6044, "step": 41250 }, { "epoch": 2.23, "learning_rate": 3.889429838796927e-05, "loss": 1.6707, "step": 41260 }, { "epoch": 2.23, "learning_rate": 3.889159363842909e-05, "loss": 1.5881, "step": 41270 }, { "epoch": 2.23, "learning_rate": 3.888888888888889e-05, "loss": 1.6423, "step": 41280 }, { "epoch": 2.23, "learning_rate": 3.8886184139348694e-05, "loss": 1.7114, "step": 41290 }, { "epoch": 2.23, "learning_rate": 3.888347938980851e-05, "loss": 1.6343, "step": 41300 }, { "epoch": 2.23, "learning_rate": 3.888077464026831e-05, "loss": 1.6609, "step": 41310 }, { "epoch": 2.23, "learning_rate": 3.887806989072812e-05, "loss": 1.6525, "step": 41320 }, { "epoch": 2.23, "learning_rate": 3.887536514118793e-05, "loss": 1.6108, "step": 41330 }, { "epoch": 2.23, "learning_rate": 3.8872660391647735e-05, "loss": 1.559, "step": 41340 }, { "epoch": 2.23, "learning_rate": 3.8869955642107545e-05, "loss": 1.5004, "step": 41350 }, { "epoch": 2.23, "learning_rate": 3.8867250892567354e-05, "loss": 1.6133, "step": 41360 }, { "epoch": 2.24, "learning_rate": 3.886454614302716e-05, "loss": 1.7001, "step": 41370 }, { "epoch": 2.24, "learning_rate": 3.886184139348697e-05, "loss": 1.6254, "step": 41380 }, { "epoch": 2.24, "learning_rate": 3.885913664394677e-05, "loss": 1.5396, "step": 41390 }, { "epoch": 2.24, "learning_rate": 3.885643189440658e-05, "loss": 1.5605, "step": 41400 }, { "epoch": 2.24, "learning_rate": 3.885372714486639e-05, "loss": 1.5667, "step": 41410 }, { "epoch": 2.24, "learning_rate": 3.885102239532619e-05, "loss": 1.6725, "step": 41420 }, { "epoch": 2.24, "learning_rate": 3.8848317645786e-05, "loss": 1.6393, "step": 41430 }, { "epoch": 2.24, "learning_rate": 3.884561289624581e-05, "loss": 1.6715, "step": 41440 }, { "epoch": 2.24, "learning_rate": 3.8842908146705614e-05, "loss": 1.6782, "step": 41450 }, { "epoch": 2.24, "learning_rate": 3.8840203397165424e-05, "loss": 1.6677, "step": 41460 }, { "epoch": 2.24, "learning_rate": 3.8837498647625233e-05, "loss": 1.6608, "step": 41470 }, { "epoch": 2.24, "learning_rate": 3.8834793898085036e-05, "loss": 1.6159, "step": 41480 }, { "epoch": 2.24, "learning_rate": 3.8832089148544846e-05, "loss": 1.7043, "step": 41490 }, { "epoch": 2.24, "learning_rate": 3.8829384399004656e-05, "loss": 1.6127, "step": 41500 }, { "epoch": 2.24, "learning_rate": 3.882667964946446e-05, "loss": 1.644, "step": 41510 }, { "epoch": 2.24, "learning_rate": 3.882397489992427e-05, "loss": 1.6119, "step": 41520 }, { "epoch": 2.24, "learning_rate": 3.882127015038408e-05, "loss": 1.7362, "step": 41530 }, { "epoch": 2.24, "learning_rate": 3.881856540084388e-05, "loss": 1.6516, "step": 41540 }, { "epoch": 2.25, "learning_rate": 3.881586065130369e-05, "loss": 1.6094, "step": 41550 }, { "epoch": 2.25, "learning_rate": 3.88131559017635e-05, "loss": 1.675, "step": 41560 }, { "epoch": 2.25, "learning_rate": 3.88104511522233e-05, "loss": 1.6534, "step": 41570 }, { "epoch": 2.25, "learning_rate": 3.880774640268311e-05, "loss": 1.6147, "step": 41580 }, { "epoch": 2.25, "learning_rate": 3.880504165314292e-05, "loss": 1.6865, "step": 41590 }, { "epoch": 2.25, "learning_rate": 3.8802336903602725e-05, "loss": 1.7353, "step": 41600 }, { "epoch": 2.25, "learning_rate": 3.8799632154062535e-05, "loss": 1.5949, "step": 41610 }, { "epoch": 2.25, "learning_rate": 3.8796927404522344e-05, "loss": 1.5566, "step": 41620 }, { "epoch": 2.25, "learning_rate": 3.879422265498215e-05, "loss": 1.4868, "step": 41630 }, { "epoch": 2.25, "learning_rate": 3.879151790544196e-05, "loss": 1.6668, "step": 41640 }, { "epoch": 2.25, "learning_rate": 3.8788813155901766e-05, "loss": 1.6006, "step": 41650 }, { "epoch": 2.25, "learning_rate": 3.878610840636157e-05, "loss": 1.6693, "step": 41660 }, { "epoch": 2.25, "learning_rate": 3.878340365682138e-05, "loss": 1.58, "step": 41670 }, { "epoch": 2.25, "learning_rate": 3.878069890728119e-05, "loss": 1.6652, "step": 41680 }, { "epoch": 2.25, "learning_rate": 3.877799415774099e-05, "loss": 1.6046, "step": 41690 }, { "epoch": 2.25, "learning_rate": 3.87752894082008e-05, "loss": 1.6039, "step": 41700 }, { "epoch": 2.25, "learning_rate": 3.877258465866061e-05, "loss": 1.5922, "step": 41710 }, { "epoch": 2.25, "learning_rate": 3.8769879909120414e-05, "loss": 1.6283, "step": 41720 }, { "epoch": 2.25, "learning_rate": 3.876717515958023e-05, "loss": 1.6908, "step": 41730 }, { "epoch": 2.26, "learning_rate": 3.876447041004003e-05, "loss": 1.5713, "step": 41740 }, { "epoch": 2.26, "learning_rate": 3.8761765660499836e-05, "loss": 1.5191, "step": 41750 }, { "epoch": 2.26, "learning_rate": 3.875906091095965e-05, "loss": 1.5627, "step": 41760 }, { "epoch": 2.26, "learning_rate": 3.8756356161419455e-05, "loss": 1.6384, "step": 41770 }, { "epoch": 2.26, "learning_rate": 3.875365141187926e-05, "loss": 1.7833, "step": 41780 }, { "epoch": 2.26, "learning_rate": 3.8750946662339074e-05, "loss": 1.6922, "step": 41790 }, { "epoch": 2.26, "learning_rate": 3.874824191279888e-05, "loss": 1.5487, "step": 41800 }, { "epoch": 2.26, "learning_rate": 3.874553716325868e-05, "loss": 1.5683, "step": 41810 }, { "epoch": 2.26, "learning_rate": 3.8742832413718497e-05, "loss": 1.6524, "step": 41820 }, { "epoch": 2.26, "learning_rate": 3.87401276641783e-05, "loss": 1.5963, "step": 41830 }, { "epoch": 2.26, "learning_rate": 3.87374229146381e-05, "loss": 1.7677, "step": 41840 }, { "epoch": 2.26, "learning_rate": 3.873471816509792e-05, "loss": 1.6466, "step": 41850 }, { "epoch": 2.26, "learning_rate": 3.873201341555772e-05, "loss": 1.6202, "step": 41860 }, { "epoch": 2.26, "learning_rate": 3.8729308666017524e-05, "loss": 1.6745, "step": 41870 }, { "epoch": 2.26, "learning_rate": 3.872660391647734e-05, "loss": 1.5738, "step": 41880 }, { "epoch": 2.26, "learning_rate": 3.8723899166937144e-05, "loss": 1.657, "step": 41890 }, { "epoch": 2.26, "learning_rate": 3.872119441739695e-05, "loss": 1.6229, "step": 41900 }, { "epoch": 2.26, "learning_rate": 3.871848966785676e-05, "loss": 1.7259, "step": 41910 }, { "epoch": 2.27, "learning_rate": 3.8715784918316566e-05, "loss": 1.5065, "step": 41920 }, { "epoch": 2.27, "learning_rate": 3.8713080168776376e-05, "loss": 1.5708, "step": 41930 }, { "epoch": 2.27, "learning_rate": 3.871037541923618e-05, "loss": 1.6049, "step": 41940 }, { "epoch": 2.27, "learning_rate": 3.870767066969599e-05, "loss": 1.6661, "step": 41950 }, { "epoch": 2.27, "learning_rate": 3.87049659201558e-05, "loss": 1.6584, "step": 41960 }, { "epoch": 2.27, "learning_rate": 3.87022611706156e-05, "loss": 1.6512, "step": 41970 }, { "epoch": 2.27, "learning_rate": 3.869955642107541e-05, "loss": 1.624, "step": 41980 }, { "epoch": 2.27, "learning_rate": 3.869685167153522e-05, "loss": 1.6737, "step": 41990 }, { "epoch": 2.27, "learning_rate": 3.869414692199502e-05, "loss": 1.523, "step": 42000 }, { "epoch": 2.27, "learning_rate": 3.869144217245483e-05, "loss": 1.6946, "step": 42010 }, { "epoch": 2.27, "learning_rate": 3.868873742291464e-05, "loss": 1.6567, "step": 42020 }, { "epoch": 2.27, "learning_rate": 3.8686032673374445e-05, "loss": 1.6769, "step": 42030 }, { "epoch": 2.27, "learning_rate": 3.8683327923834255e-05, "loss": 1.6448, "step": 42040 }, { "epoch": 2.27, "learning_rate": 3.8680623174294064e-05, "loss": 1.5503, "step": 42050 }, { "epoch": 2.27, "learning_rate": 3.867791842475387e-05, "loss": 1.6086, "step": 42060 }, { "epoch": 2.27, "learning_rate": 3.867521367521368e-05, "loss": 1.667, "step": 42070 }, { "epoch": 2.27, "learning_rate": 3.8672508925673486e-05, "loss": 1.6039, "step": 42080 }, { "epoch": 2.27, "learning_rate": 3.866980417613329e-05, "loss": 1.6356, "step": 42090 }, { "epoch": 2.27, "learning_rate": 3.86670994265931e-05, "loss": 1.6517, "step": 42100 }, { "epoch": 2.28, "learning_rate": 3.866439467705291e-05, "loss": 1.6761, "step": 42110 }, { "epoch": 2.28, "learning_rate": 3.866168992751271e-05, "loss": 1.6493, "step": 42120 }, { "epoch": 2.28, "learning_rate": 3.865898517797252e-05, "loss": 1.5843, "step": 42130 }, { "epoch": 2.28, "learning_rate": 3.865628042843233e-05, "loss": 1.6101, "step": 42140 }, { "epoch": 2.28, "learning_rate": 3.8653575678892134e-05, "loss": 1.6203, "step": 42150 }, { "epoch": 2.28, "learning_rate": 3.865087092935194e-05, "loss": 1.6209, "step": 42160 }, { "epoch": 2.28, "learning_rate": 3.864816617981175e-05, "loss": 1.6028, "step": 42170 }, { "epoch": 2.28, "learning_rate": 3.8645461430271556e-05, "loss": 1.6859, "step": 42180 }, { "epoch": 2.28, "learning_rate": 3.8642756680731365e-05, "loss": 1.6367, "step": 42190 }, { "epoch": 2.28, "learning_rate": 3.8640051931191175e-05, "loss": 1.6851, "step": 42200 }, { "epoch": 2.28, "learning_rate": 3.863734718165098e-05, "loss": 1.6627, "step": 42210 }, { "epoch": 2.28, "learning_rate": 3.863464243211079e-05, "loss": 1.654, "step": 42220 }, { "epoch": 2.28, "learning_rate": 3.86319376825706e-05, "loss": 1.5871, "step": 42230 }, { "epoch": 2.28, "learning_rate": 3.86292329330304e-05, "loss": 1.6151, "step": 42240 }, { "epoch": 2.28, "learning_rate": 3.862652818349021e-05, "loss": 1.6187, "step": 42250 }, { "epoch": 2.28, "learning_rate": 3.862382343395002e-05, "loss": 1.5852, "step": 42260 }, { "epoch": 2.28, "learning_rate": 3.862111868440982e-05, "loss": 1.6597, "step": 42270 }, { "epoch": 2.28, "learning_rate": 3.861841393486963e-05, "loss": 1.5813, "step": 42280 }, { "epoch": 2.29, "learning_rate": 3.861570918532944e-05, "loss": 1.6979, "step": 42290 }, { "epoch": 2.29, "learning_rate": 3.8613004435789245e-05, "loss": 1.595, "step": 42300 }, { "epoch": 2.29, "learning_rate": 3.8610299686249054e-05, "loss": 1.5925, "step": 42310 }, { "epoch": 2.29, "learning_rate": 3.8607594936708864e-05, "loss": 1.6201, "step": 42320 }, { "epoch": 2.29, "learning_rate": 3.860489018716867e-05, "loss": 1.5556, "step": 42330 }, { "epoch": 2.29, "learning_rate": 3.860218543762848e-05, "loss": 1.5433, "step": 42340 }, { "epoch": 2.29, "learning_rate": 3.8599480688088286e-05, "loss": 1.5794, "step": 42350 }, { "epoch": 2.29, "learning_rate": 3.859677593854809e-05, "loss": 1.6943, "step": 42360 }, { "epoch": 2.29, "learning_rate": 3.8594071189007905e-05, "loss": 1.6182, "step": 42370 }, { "epoch": 2.29, "learning_rate": 3.859136643946771e-05, "loss": 1.5455, "step": 42380 }, { "epoch": 2.29, "learning_rate": 3.858866168992751e-05, "loss": 1.596, "step": 42390 }, { "epoch": 2.29, "learning_rate": 3.858595694038733e-05, "loss": 1.6037, "step": 42400 }, { "epoch": 2.29, "learning_rate": 3.858325219084713e-05, "loss": 1.5873, "step": 42410 }, { "epoch": 2.29, "learning_rate": 3.858054744130693e-05, "loss": 1.5778, "step": 42420 }, { "epoch": 2.29, "learning_rate": 3.857784269176675e-05, "loss": 1.6619, "step": 42430 }, { "epoch": 2.29, "learning_rate": 3.857513794222655e-05, "loss": 1.5625, "step": 42440 }, { "epoch": 2.29, "learning_rate": 3.8572433192686355e-05, "loss": 1.6709, "step": 42450 }, { "epoch": 2.29, "learning_rate": 3.856972844314617e-05, "loss": 1.6394, "step": 42460 }, { "epoch": 2.29, "learning_rate": 3.8567023693605975e-05, "loss": 1.6139, "step": 42470 }, { "epoch": 2.3, "learning_rate": 3.856431894406578e-05, "loss": 1.6312, "step": 42480 }, { "epoch": 2.3, "learning_rate": 3.8561614194525594e-05, "loss": 1.5964, "step": 42490 }, { "epoch": 2.3, "learning_rate": 3.85589094449854e-05, "loss": 1.644, "step": 42500 }, { "epoch": 2.3, "learning_rate": 3.85562046954452e-05, "loss": 1.648, "step": 42510 }, { "epoch": 2.3, "learning_rate": 3.855349994590501e-05, "loss": 1.7004, "step": 42520 }, { "epoch": 2.3, "learning_rate": 3.855079519636482e-05, "loss": 1.5718, "step": 42530 }, { "epoch": 2.3, "learning_rate": 3.854809044682462e-05, "loss": 1.6487, "step": 42540 }, { "epoch": 2.3, "learning_rate": 3.854538569728443e-05, "loss": 1.6508, "step": 42550 }, { "epoch": 2.3, "learning_rate": 3.854268094774424e-05, "loss": 1.6667, "step": 42560 }, { "epoch": 2.3, "learning_rate": 3.853997619820405e-05, "loss": 1.5796, "step": 42570 }, { "epoch": 2.3, "learning_rate": 3.8537271448663854e-05, "loss": 1.5701, "step": 42580 }, { "epoch": 2.3, "learning_rate": 3.853456669912366e-05, "loss": 1.6659, "step": 42590 }, { "epoch": 2.3, "learning_rate": 3.853186194958347e-05, "loss": 1.5542, "step": 42600 }, { "epoch": 2.3, "learning_rate": 3.8529157200043276e-05, "loss": 1.5812, "step": 42610 }, { "epoch": 2.3, "learning_rate": 3.8526452450503086e-05, "loss": 1.5852, "step": 42620 }, { "epoch": 2.3, "learning_rate": 3.8523747700962895e-05, "loss": 1.6587, "step": 42630 }, { "epoch": 2.3, "learning_rate": 3.85210429514227e-05, "loss": 1.6185, "step": 42640 }, { "epoch": 2.3, "learning_rate": 3.851833820188251e-05, "loss": 1.6925, "step": 42650 }, { "epoch": 2.31, "learning_rate": 3.851563345234232e-05, "loss": 1.6619, "step": 42660 }, { "epoch": 2.31, "learning_rate": 3.851292870280212e-05, "loss": 1.649, "step": 42670 }, { "epoch": 2.31, "learning_rate": 3.851022395326193e-05, "loss": 1.6681, "step": 42680 }, { "epoch": 2.31, "learning_rate": 3.850751920372174e-05, "loss": 1.6488, "step": 42690 }, { "epoch": 2.31, "learning_rate": 3.850481445418154e-05, "loss": 1.659, "step": 42700 }, { "epoch": 2.31, "learning_rate": 3.850210970464135e-05, "loss": 1.5891, "step": 42710 }, { "epoch": 2.31, "learning_rate": 3.849940495510116e-05, "loss": 1.6347, "step": 42720 }, { "epoch": 2.31, "learning_rate": 3.8496700205560965e-05, "loss": 1.6032, "step": 42730 }, { "epoch": 2.31, "learning_rate": 3.8493995456020774e-05, "loss": 1.6699, "step": 42740 }, { "epoch": 2.31, "learning_rate": 3.8491290706480584e-05, "loss": 1.6497, "step": 42750 }, { "epoch": 2.31, "learning_rate": 3.848858595694039e-05, "loss": 1.6218, "step": 42760 }, { "epoch": 2.31, "learning_rate": 3.8485881207400196e-05, "loss": 1.631, "step": 42770 }, { "epoch": 2.31, "learning_rate": 3.8483176457860006e-05, "loss": 1.6638, "step": 42780 }, { "epoch": 2.31, "learning_rate": 3.848047170831981e-05, "loss": 1.5988, "step": 42790 }, { "epoch": 2.31, "learning_rate": 3.847776695877962e-05, "loss": 1.6042, "step": 42800 }, { "epoch": 2.31, "learning_rate": 3.847506220923943e-05, "loss": 1.6149, "step": 42810 }, { "epoch": 2.31, "learning_rate": 3.847235745969923e-05, "loss": 1.627, "step": 42820 }, { "epoch": 2.31, "learning_rate": 3.846965271015904e-05, "loss": 1.6541, "step": 42830 }, { "epoch": 2.31, "learning_rate": 3.846694796061885e-05, "loss": 1.5941, "step": 42840 }, { "epoch": 2.32, "learning_rate": 3.846424321107865e-05, "loss": 1.7156, "step": 42850 }, { "epoch": 2.32, "learning_rate": 3.846153846153846e-05, "loss": 1.7507, "step": 42860 }, { "epoch": 2.32, "learning_rate": 3.845883371199827e-05, "loss": 1.6452, "step": 42870 }, { "epoch": 2.32, "learning_rate": 3.8456128962458075e-05, "loss": 1.6237, "step": 42880 }, { "epoch": 2.32, "learning_rate": 3.8453424212917885e-05, "loss": 1.5787, "step": 42890 }, { "epoch": 2.32, "learning_rate": 3.8450719463377695e-05, "loss": 1.6268, "step": 42900 }, { "epoch": 2.32, "learning_rate": 3.84480147138375e-05, "loss": 1.6575, "step": 42910 }, { "epoch": 2.32, "learning_rate": 3.844530996429731e-05, "loss": 1.5809, "step": 42920 }, { "epoch": 2.32, "learning_rate": 3.844260521475712e-05, "loss": 1.5953, "step": 42930 }, { "epoch": 2.32, "learning_rate": 3.843990046521692e-05, "loss": 1.5933, "step": 42940 }, { "epoch": 2.32, "learning_rate": 3.843719571567673e-05, "loss": 1.6576, "step": 42950 }, { "epoch": 2.32, "learning_rate": 3.843449096613654e-05, "loss": 1.5731, "step": 42960 }, { "epoch": 2.32, "learning_rate": 3.843178621659634e-05, "loss": 1.6213, "step": 42970 }, { "epoch": 2.32, "learning_rate": 3.842908146705616e-05, "loss": 1.6088, "step": 42980 }, { "epoch": 2.32, "learning_rate": 3.842637671751596e-05, "loss": 1.6434, "step": 42990 }, { "epoch": 2.32, "learning_rate": 3.8423671967975764e-05, "loss": 1.5831, "step": 43000 }, { "epoch": 2.32, "learning_rate": 3.842096721843558e-05, "loss": 1.5916, "step": 43010 }, { "epoch": 2.32, "learning_rate": 3.841826246889538e-05, "loss": 1.6611, "step": 43020 }, { "epoch": 2.33, "learning_rate": 3.8415557719355186e-05, "loss": 1.6468, "step": 43030 }, { "epoch": 2.33, "learning_rate": 3.8412852969815e-05, "loss": 1.6642, "step": 43040 }, { "epoch": 2.33, "learning_rate": 3.8410148220274806e-05, "loss": 1.6642, "step": 43050 }, { "epoch": 2.33, "learning_rate": 3.840744347073461e-05, "loss": 1.6299, "step": 43060 }, { "epoch": 2.33, "learning_rate": 3.840473872119442e-05, "loss": 1.6052, "step": 43070 }, { "epoch": 2.33, "learning_rate": 3.840203397165423e-05, "loss": 1.6431, "step": 43080 }, { "epoch": 2.33, "learning_rate": 3.839932922211403e-05, "loss": 1.6809, "step": 43090 }, { "epoch": 2.33, "learning_rate": 3.839662447257384e-05, "loss": 1.5976, "step": 43100 }, { "epoch": 2.33, "learning_rate": 3.839391972303365e-05, "loss": 1.6841, "step": 43110 }, { "epoch": 2.33, "learning_rate": 3.839121497349345e-05, "loss": 1.5493, "step": 43120 }, { "epoch": 2.33, "learning_rate": 3.838851022395326e-05, "loss": 1.6235, "step": 43130 }, { "epoch": 2.33, "learning_rate": 3.838580547441307e-05, "loss": 1.6404, "step": 43140 }, { "epoch": 2.33, "learning_rate": 3.8383100724872875e-05, "loss": 1.6464, "step": 43150 }, { "epoch": 2.33, "learning_rate": 3.8380395975332685e-05, "loss": 1.6316, "step": 43160 }, { "epoch": 2.33, "learning_rate": 3.8377691225792494e-05, "loss": 1.6469, "step": 43170 }, { "epoch": 2.33, "learning_rate": 3.83749864762523e-05, "loss": 1.6116, "step": 43180 }, { "epoch": 2.33, "learning_rate": 3.837228172671211e-05, "loss": 1.6494, "step": 43190 }, { "epoch": 2.33, "learning_rate": 3.8369576977171916e-05, "loss": 1.5984, "step": 43200 }, { "epoch": 2.33, "learning_rate": 3.8366872227631726e-05, "loss": 1.707, "step": 43210 }, { "epoch": 2.34, "learning_rate": 3.836416747809153e-05, "loss": 1.5615, "step": 43220 }, { "epoch": 2.34, "learning_rate": 3.836146272855134e-05, "loss": 1.6118, "step": 43230 }, { "epoch": 2.34, "learning_rate": 3.835875797901115e-05, "loss": 1.6005, "step": 43240 }, { "epoch": 2.34, "learning_rate": 3.835605322947095e-05, "loss": 1.5827, "step": 43250 }, { "epoch": 2.34, "learning_rate": 3.835334847993076e-05, "loss": 1.5427, "step": 43260 }, { "epoch": 2.34, "learning_rate": 3.835064373039057e-05, "loss": 1.6003, "step": 43270 }, { "epoch": 2.34, "learning_rate": 3.834793898085037e-05, "loss": 1.6052, "step": 43280 }, { "epoch": 2.34, "learning_rate": 3.834523423131018e-05, "loss": 1.6744, "step": 43290 }, { "epoch": 2.34, "learning_rate": 3.834252948176999e-05, "loss": 1.6877, "step": 43300 }, { "epoch": 2.34, "learning_rate": 3.8339824732229795e-05, "loss": 1.6137, "step": 43310 }, { "epoch": 2.34, "learning_rate": 3.8337119982689605e-05, "loss": 1.5533, "step": 43320 }, { "epoch": 2.34, "learning_rate": 3.8334415233149415e-05, "loss": 1.7072, "step": 43330 }, { "epoch": 2.34, "learning_rate": 3.833171048360922e-05, "loss": 1.5582, "step": 43340 }, { "epoch": 2.34, "learning_rate": 3.832900573406903e-05, "loss": 1.6109, "step": 43350 }, { "epoch": 2.34, "learning_rate": 3.832630098452884e-05, "loss": 1.6904, "step": 43360 }, { "epoch": 2.34, "learning_rate": 3.832359623498864e-05, "loss": 1.6665, "step": 43370 }, { "epoch": 2.34, "learning_rate": 3.832089148544845e-05, "loss": 1.6494, "step": 43380 }, { "epoch": 2.34, "learning_rate": 3.831818673590826e-05, "loss": 1.6704, "step": 43390 }, { "epoch": 2.35, "learning_rate": 3.831548198636806e-05, "loss": 1.5499, "step": 43400 }, { "epoch": 2.35, "learning_rate": 3.831277723682787e-05, "loss": 1.6344, "step": 43410 }, { "epoch": 2.35, "learning_rate": 3.831007248728768e-05, "loss": 1.6348, "step": 43420 }, { "epoch": 2.35, "learning_rate": 3.8307367737747484e-05, "loss": 1.623, "step": 43430 }, { "epoch": 2.35, "learning_rate": 3.8304662988207294e-05, "loss": 1.5923, "step": 43440 }, { "epoch": 2.35, "learning_rate": 3.83019582386671e-05, "loss": 1.7457, "step": 43450 }, { "epoch": 2.35, "learning_rate": 3.8299253489126906e-05, "loss": 1.5908, "step": 43460 }, { "epoch": 2.35, "learning_rate": 3.8296548739586716e-05, "loss": 1.6272, "step": 43470 }, { "epoch": 2.35, "learning_rate": 3.8293843990046526e-05, "loss": 1.5475, "step": 43480 }, { "epoch": 2.35, "learning_rate": 3.829113924050633e-05, "loss": 1.5905, "step": 43490 }, { "epoch": 2.35, "learning_rate": 3.828843449096614e-05, "loss": 1.7126, "step": 43500 }, { "epoch": 2.35, "learning_rate": 3.828572974142595e-05, "loss": 1.6312, "step": 43510 }, { "epoch": 2.35, "learning_rate": 3.828302499188575e-05, "loss": 1.6394, "step": 43520 }, { "epoch": 2.35, "learning_rate": 3.828032024234556e-05, "loss": 1.5741, "step": 43530 }, { "epoch": 2.35, "learning_rate": 3.827761549280537e-05, "loss": 1.5952, "step": 43540 }, { "epoch": 2.35, "learning_rate": 3.827491074326517e-05, "loss": 1.6437, "step": 43550 }, { "epoch": 2.35, "learning_rate": 3.827220599372498e-05, "loss": 1.6691, "step": 43560 }, { "epoch": 2.35, "learning_rate": 3.826950124418479e-05, "loss": 1.6465, "step": 43570 }, { "epoch": 2.35, "learning_rate": 3.8266796494644595e-05, "loss": 1.6528, "step": 43580 }, { "epoch": 2.36, "learning_rate": 3.8264091745104405e-05, "loss": 1.6198, "step": 43590 }, { "epoch": 2.36, "learning_rate": 3.8261386995564214e-05, "loss": 1.6439, "step": 43600 }, { "epoch": 2.36, "learning_rate": 3.825868224602402e-05, "loss": 1.6234, "step": 43610 }, { "epoch": 2.36, "learning_rate": 3.825597749648383e-05, "loss": 1.6319, "step": 43620 }, { "epoch": 2.36, "learning_rate": 3.8253272746943636e-05, "loss": 1.5428, "step": 43630 }, { "epoch": 2.36, "learning_rate": 3.825056799740344e-05, "loss": 1.6143, "step": 43640 }, { "epoch": 2.36, "learning_rate": 3.824786324786325e-05, "loss": 1.666, "step": 43650 }, { "epoch": 2.36, "learning_rate": 3.824515849832306e-05, "loss": 1.6846, "step": 43660 }, { "epoch": 2.36, "learning_rate": 3.824245374878286e-05, "loss": 1.5791, "step": 43670 }, { "epoch": 2.36, "learning_rate": 3.823974899924267e-05, "loss": 1.5871, "step": 43680 }, { "epoch": 2.36, "learning_rate": 3.823704424970248e-05, "loss": 1.5328, "step": 43690 }, { "epoch": 2.36, "learning_rate": 3.8234339500162284e-05, "loss": 1.6139, "step": 43700 }, { "epoch": 2.36, "learning_rate": 3.823163475062209e-05, "loss": 1.6322, "step": 43710 }, { "epoch": 2.36, "learning_rate": 3.82289300010819e-05, "loss": 1.5963, "step": 43720 }, { "epoch": 2.36, "learning_rate": 3.8226225251541706e-05, "loss": 1.6294, "step": 43730 }, { "epoch": 2.36, "learning_rate": 3.8223520502001515e-05, "loss": 1.675, "step": 43740 }, { "epoch": 2.36, "learning_rate": 3.8220815752461325e-05, "loss": 1.6254, "step": 43750 }, { "epoch": 2.36, "learning_rate": 3.821811100292113e-05, "loss": 1.6108, "step": 43760 }, { "epoch": 2.37, "learning_rate": 3.821540625338094e-05, "loss": 1.5154, "step": 43770 }, { "epoch": 2.37, "learning_rate": 3.821270150384075e-05, "loss": 1.6221, "step": 43780 }, { "epoch": 2.37, "learning_rate": 3.820999675430055e-05, "loss": 1.5812, "step": 43790 }, { "epoch": 2.37, "learning_rate": 3.820729200476036e-05, "loss": 1.585, "step": 43800 }, { "epoch": 2.37, "learning_rate": 3.820458725522017e-05, "loss": 1.6093, "step": 43810 }, { "epoch": 2.37, "learning_rate": 3.820188250567997e-05, "loss": 1.5307, "step": 43820 }, { "epoch": 2.37, "learning_rate": 3.819917775613978e-05, "loss": 1.6325, "step": 43830 }, { "epoch": 2.37, "learning_rate": 3.819647300659959e-05, "loss": 1.5926, "step": 43840 }, { "epoch": 2.37, "learning_rate": 3.81937682570594e-05, "loss": 1.6278, "step": 43850 }, { "epoch": 2.37, "learning_rate": 3.8191063507519204e-05, "loss": 1.6461, "step": 43860 }, { "epoch": 2.37, "learning_rate": 3.8188358757979014e-05, "loss": 1.678, "step": 43870 }, { "epoch": 2.37, "learning_rate": 3.8185654008438823e-05, "loss": 1.6146, "step": 43880 }, { "epoch": 2.37, "learning_rate": 3.8182949258898626e-05, "loss": 1.6164, "step": 43890 }, { "epoch": 2.37, "learning_rate": 3.8180244509358436e-05, "loss": 1.6319, "step": 43900 }, { "epoch": 2.37, "learning_rate": 3.8177539759818246e-05, "loss": 1.6677, "step": 43910 }, { "epoch": 2.37, "learning_rate": 3.817483501027805e-05, "loss": 1.5526, "step": 43920 }, { "epoch": 2.37, "learning_rate": 3.817213026073786e-05, "loss": 1.6365, "step": 43930 }, { "epoch": 2.37, "learning_rate": 3.816942551119767e-05, "loss": 1.6329, "step": 43940 }, { "epoch": 2.37, "learning_rate": 3.816672076165747e-05, "loss": 1.6116, "step": 43950 }, { "epoch": 2.38, "learning_rate": 3.816401601211728e-05, "loss": 1.6088, "step": 43960 }, { "epoch": 2.38, "learning_rate": 3.816131126257709e-05, "loss": 1.6168, "step": 43970 }, { "epoch": 2.38, "learning_rate": 3.815860651303689e-05, "loss": 1.6093, "step": 43980 }, { "epoch": 2.38, "learning_rate": 3.81559017634967e-05, "loss": 1.6458, "step": 43990 }, { "epoch": 2.38, "learning_rate": 3.815319701395651e-05, "loss": 1.6689, "step": 44000 }, { "epoch": 2.38, "learning_rate": 3.8150492264416315e-05, "loss": 1.6308, "step": 44010 }, { "epoch": 2.38, "learning_rate": 3.8147787514876125e-05, "loss": 1.6657, "step": 44020 }, { "epoch": 2.38, "learning_rate": 3.8145082765335934e-05, "loss": 1.6041, "step": 44030 }, { "epoch": 2.38, "learning_rate": 3.814237801579574e-05, "loss": 1.6846, "step": 44040 }, { "epoch": 2.38, "learning_rate": 3.813967326625555e-05, "loss": 1.643, "step": 44050 }, { "epoch": 2.38, "learning_rate": 3.8136968516715356e-05, "loss": 1.5715, "step": 44060 }, { "epoch": 2.38, "learning_rate": 3.813426376717516e-05, "loss": 1.6055, "step": 44070 }, { "epoch": 2.38, "learning_rate": 3.813155901763497e-05, "loss": 1.6029, "step": 44080 }, { "epoch": 2.38, "learning_rate": 3.812885426809478e-05, "loss": 1.6841, "step": 44090 }, { "epoch": 2.38, "learning_rate": 3.812614951855458e-05, "loss": 1.5937, "step": 44100 }, { "epoch": 2.38, "learning_rate": 3.812344476901439e-05, "loss": 1.6346, "step": 44110 }, { "epoch": 2.38, "learning_rate": 3.81207400194742e-05, "loss": 1.5984, "step": 44120 }, { "epoch": 2.38, "learning_rate": 3.8118035269934004e-05, "loss": 1.6524, "step": 44130 }, { "epoch": 2.39, "learning_rate": 3.811533052039381e-05, "loss": 1.5685, "step": 44140 }, { "epoch": 2.39, "learning_rate": 3.811262577085362e-05, "loss": 1.6126, "step": 44150 }, { "epoch": 2.39, "learning_rate": 3.8109921021313426e-05, "loss": 1.6363, "step": 44160 }, { "epoch": 2.39, "learning_rate": 3.8107216271773235e-05, "loss": 1.7244, "step": 44170 }, { "epoch": 2.39, "learning_rate": 3.8104511522233045e-05, "loss": 1.6533, "step": 44180 }, { "epoch": 2.39, "learning_rate": 3.810180677269285e-05, "loss": 1.6149, "step": 44190 }, { "epoch": 2.39, "learning_rate": 3.809910202315266e-05, "loss": 1.6435, "step": 44200 }, { "epoch": 2.39, "learning_rate": 3.809639727361247e-05, "loss": 1.5855, "step": 44210 }, { "epoch": 2.39, "learning_rate": 3.809369252407227e-05, "loss": 1.7161, "step": 44220 }, { "epoch": 2.39, "learning_rate": 3.809098777453208e-05, "loss": 1.5647, "step": 44230 }, { "epoch": 2.39, "learning_rate": 3.808828302499189e-05, "loss": 1.6116, "step": 44240 }, { "epoch": 2.39, "learning_rate": 3.808557827545169e-05, "loss": 1.6357, "step": 44250 }, { "epoch": 2.39, "learning_rate": 3.80828735259115e-05, "loss": 1.6094, "step": 44260 }, { "epoch": 2.39, "learning_rate": 3.808016877637131e-05, "loss": 1.5905, "step": 44270 }, { "epoch": 2.39, "learning_rate": 3.8077464026831114e-05, "loss": 1.6165, "step": 44280 }, { "epoch": 2.39, "learning_rate": 3.8074759277290924e-05, "loss": 1.536, "step": 44290 }, { "epoch": 2.39, "learning_rate": 3.8072054527750734e-05, "loss": 1.6435, "step": 44300 }, { "epoch": 2.39, "learning_rate": 3.806934977821054e-05, "loss": 1.6771, "step": 44310 }, { "epoch": 2.39, "learning_rate": 3.8066645028670346e-05, "loss": 1.5668, "step": 44320 }, { "epoch": 2.4, "learning_rate": 3.8063940279130156e-05, "loss": 1.6538, "step": 44330 }, { "epoch": 2.4, "learning_rate": 3.806123552958996e-05, "loss": 1.5739, "step": 44340 }, { "epoch": 2.4, "learning_rate": 3.805853078004977e-05, "loss": 1.5623, "step": 44350 }, { "epoch": 2.4, "learning_rate": 3.805582603050958e-05, "loss": 1.6015, "step": 44360 }, { "epoch": 2.4, "learning_rate": 3.805312128096938e-05, "loss": 1.7665, "step": 44370 }, { "epoch": 2.4, "learning_rate": 3.805041653142919e-05, "loss": 1.6267, "step": 44380 }, { "epoch": 2.4, "learning_rate": 3.8047711781889e-05, "loss": 1.6096, "step": 44390 }, { "epoch": 2.4, "learning_rate": 3.80450070323488e-05, "loss": 1.638, "step": 44400 }, { "epoch": 2.4, "learning_rate": 3.804230228280861e-05, "loss": 1.6362, "step": 44410 }, { "epoch": 2.4, "learning_rate": 3.803959753326842e-05, "loss": 1.624, "step": 44420 }, { "epoch": 2.4, "learning_rate": 3.8036892783728225e-05, "loss": 1.6316, "step": 44430 }, { "epoch": 2.4, "learning_rate": 3.8034188034188035e-05, "loss": 1.6119, "step": 44440 }, { "epoch": 2.4, "learning_rate": 3.8031483284647845e-05, "loss": 1.6564, "step": 44450 }, { "epoch": 2.4, "learning_rate": 3.802877853510765e-05, "loss": 1.6147, "step": 44460 }, { "epoch": 2.4, "learning_rate": 3.802607378556746e-05, "loss": 1.6778, "step": 44470 }, { "epoch": 2.4, "learning_rate": 3.802336903602727e-05, "loss": 1.5899, "step": 44480 }, { "epoch": 2.4, "learning_rate": 3.8020664286487076e-05, "loss": 1.5464, "step": 44490 }, { "epoch": 2.4, "learning_rate": 3.801795953694688e-05, "loss": 1.6063, "step": 44500 }, { "epoch": 2.41, "learning_rate": 3.801525478740669e-05, "loss": 1.6826, "step": 44510 }, { "epoch": 2.41, "learning_rate": 3.80125500378665e-05, "loss": 1.596, "step": 44520 }, { "epoch": 2.41, "learning_rate": 3.80098452883263e-05, "loss": 1.5827, "step": 44530 }, { "epoch": 2.41, "learning_rate": 3.800714053878611e-05, "loss": 1.6272, "step": 44540 }, { "epoch": 2.41, "learning_rate": 3.800443578924592e-05, "loss": 1.669, "step": 44550 }, { "epoch": 2.41, "learning_rate": 3.8001731039705724e-05, "loss": 1.6425, "step": 44560 }, { "epoch": 2.41, "learning_rate": 3.799902629016553e-05, "loss": 1.5868, "step": 44570 }, { "epoch": 2.41, "learning_rate": 3.799632154062534e-05, "loss": 1.5633, "step": 44580 }, { "epoch": 2.41, "learning_rate": 3.7993616791085146e-05, "loss": 1.6111, "step": 44590 }, { "epoch": 2.41, "learning_rate": 3.7990912041544955e-05, "loss": 1.6755, "step": 44600 }, { "epoch": 2.41, "learning_rate": 3.7988207292004765e-05, "loss": 1.6227, "step": 44610 }, { "epoch": 2.41, "learning_rate": 3.798550254246457e-05, "loss": 1.6366, "step": 44620 }, { "epoch": 2.41, "learning_rate": 3.798279779292438e-05, "loss": 1.6239, "step": 44630 }, { "epoch": 2.41, "learning_rate": 3.798009304338419e-05, "loss": 1.5571, "step": 44640 }, { "epoch": 2.41, "learning_rate": 3.797738829384399e-05, "loss": 1.6158, "step": 44650 }, { "epoch": 2.41, "learning_rate": 3.79746835443038e-05, "loss": 1.6231, "step": 44660 }, { "epoch": 2.41, "learning_rate": 3.797197879476361e-05, "loss": 1.5988, "step": 44670 }, { "epoch": 2.41, "learning_rate": 3.796927404522341e-05, "loss": 1.6156, "step": 44680 }, { "epoch": 2.41, "learning_rate": 3.796656929568322e-05, "loss": 1.5533, "step": 44690 }, { "epoch": 2.42, "learning_rate": 3.796386454614303e-05, "loss": 1.6791, "step": 44700 }, { "epoch": 2.42, "learning_rate": 3.7961159796602834e-05, "loss": 1.5603, "step": 44710 }, { "epoch": 2.42, "learning_rate": 3.7958455047062644e-05, "loss": 1.6746, "step": 44720 }, { "epoch": 2.42, "learning_rate": 3.7955750297522454e-05, "loss": 1.5216, "step": 44730 }, { "epoch": 2.42, "learning_rate": 3.795304554798226e-05, "loss": 1.6477, "step": 44740 }, { "epoch": 2.42, "learning_rate": 3.7950340798442066e-05, "loss": 1.6087, "step": 44750 }, { "epoch": 2.42, "learning_rate": 3.7947636048901876e-05, "loss": 1.5991, "step": 44760 }, { "epoch": 2.42, "learning_rate": 3.794493129936168e-05, "loss": 1.6302, "step": 44770 }, { "epoch": 2.42, "learning_rate": 3.794222654982149e-05, "loss": 1.6489, "step": 44780 }, { "epoch": 2.42, "learning_rate": 3.79395218002813e-05, "loss": 1.6685, "step": 44790 }, { "epoch": 2.42, "learning_rate": 3.79368170507411e-05, "loss": 1.6737, "step": 44800 }, { "epoch": 2.42, "learning_rate": 3.793411230120091e-05, "loss": 1.6294, "step": 44810 }, { "epoch": 2.42, "learning_rate": 3.793140755166072e-05, "loss": 1.691, "step": 44820 }, { "epoch": 2.42, "learning_rate": 3.792870280212052e-05, "loss": 1.7313, "step": 44830 }, { "epoch": 2.42, "learning_rate": 3.792599805258033e-05, "loss": 1.6422, "step": 44840 }, { "epoch": 2.42, "learning_rate": 3.792329330304014e-05, "loss": 1.5808, "step": 44850 }, { "epoch": 2.42, "learning_rate": 3.7920588553499945e-05, "loss": 1.655, "step": 44860 }, { "epoch": 2.42, "learning_rate": 3.7917883803959755e-05, "loss": 1.5905, "step": 44870 }, { "epoch": 2.43, "learning_rate": 3.7915179054419565e-05, "loss": 1.7206, "step": 44880 }, { "epoch": 2.43, "learning_rate": 3.791247430487937e-05, "loss": 1.6184, "step": 44890 }, { "epoch": 2.43, "learning_rate": 3.790976955533918e-05, "loss": 1.6569, "step": 44900 }, { "epoch": 2.43, "learning_rate": 3.790706480579899e-05, "loss": 1.6065, "step": 44910 }, { "epoch": 2.43, "learning_rate": 3.790436005625879e-05, "loss": 1.6593, "step": 44920 }, { "epoch": 2.43, "learning_rate": 3.79016553067186e-05, "loss": 1.6454, "step": 44930 }, { "epoch": 2.43, "learning_rate": 3.789895055717841e-05, "loss": 1.6303, "step": 44940 }, { "epoch": 2.43, "learning_rate": 3.789624580763821e-05, "loss": 1.6165, "step": 44950 }, { "epoch": 2.43, "learning_rate": 3.789354105809802e-05, "loss": 1.6653, "step": 44960 }, { "epoch": 2.43, "learning_rate": 3.789083630855783e-05, "loss": 1.5805, "step": 44970 }, { "epoch": 2.43, "learning_rate": 3.7888131559017634e-05, "loss": 1.6253, "step": 44980 }, { "epoch": 2.43, "learning_rate": 3.7885426809477444e-05, "loss": 1.5661, "step": 44990 }, { "epoch": 2.43, "learning_rate": 3.788272205993725e-05, "loss": 1.5937, "step": 45000 }, { "epoch": 2.43, "learning_rate": 3.7880017310397056e-05, "loss": 1.5897, "step": 45010 }, { "epoch": 2.43, "learning_rate": 3.7877312560856866e-05, "loss": 1.6189, "step": 45020 }, { "epoch": 2.43, "learning_rate": 3.7874607811316675e-05, "loss": 1.5738, "step": 45030 }, { "epoch": 2.43, "learning_rate": 3.787190306177648e-05, "loss": 1.6419, "step": 45040 }, { "epoch": 2.43, "learning_rate": 3.786919831223629e-05, "loss": 1.552, "step": 45050 }, { "epoch": 2.43, "learning_rate": 3.78664935626961e-05, "loss": 1.7354, "step": 45060 }, { "epoch": 2.44, "learning_rate": 3.78637888131559e-05, "loss": 1.7191, "step": 45070 }, { "epoch": 2.44, "learning_rate": 3.786108406361571e-05, "loss": 1.602, "step": 45080 }, { "epoch": 2.44, "learning_rate": 3.785837931407552e-05, "loss": 1.5641, "step": 45090 }, { "epoch": 2.44, "learning_rate": 3.785567456453533e-05, "loss": 1.5523, "step": 45100 }, { "epoch": 2.44, "learning_rate": 3.785296981499513e-05, "loss": 1.626, "step": 45110 }, { "epoch": 2.44, "learning_rate": 3.785026506545494e-05, "loss": 1.6163, "step": 45120 }, { "epoch": 2.44, "learning_rate": 3.784756031591475e-05, "loss": 1.6465, "step": 45130 }, { "epoch": 2.44, "learning_rate": 3.7844855566374555e-05, "loss": 1.7197, "step": 45140 }, { "epoch": 2.44, "learning_rate": 3.7842150816834364e-05, "loss": 1.6313, "step": 45150 }, { "epoch": 2.44, "learning_rate": 3.7839446067294174e-05, "loss": 1.5716, "step": 45160 }, { "epoch": 2.44, "learning_rate": 3.783674131775398e-05, "loss": 1.6029, "step": 45170 }, { "epoch": 2.44, "learning_rate": 3.7834036568213786e-05, "loss": 1.6232, "step": 45180 }, { "epoch": 2.44, "learning_rate": 3.7831331818673596e-05, "loss": 1.6555, "step": 45190 }, { "epoch": 2.44, "learning_rate": 3.78286270691334e-05, "loss": 1.6688, "step": 45200 }, { "epoch": 2.44, "learning_rate": 3.782592231959321e-05, "loss": 1.6298, "step": 45210 }, { "epoch": 2.44, "learning_rate": 3.782321757005302e-05, "loss": 1.5735, "step": 45220 }, { "epoch": 2.44, "learning_rate": 3.782051282051282e-05, "loss": 1.6639, "step": 45230 }, { "epoch": 2.44, "learning_rate": 3.781780807097263e-05, "loss": 1.6084, "step": 45240 }, { "epoch": 2.45, "learning_rate": 3.781510332143244e-05, "loss": 1.6272, "step": 45250 }, { "epoch": 2.45, "learning_rate": 3.781239857189224e-05, "loss": 1.625, "step": 45260 }, { "epoch": 2.45, "learning_rate": 3.780969382235205e-05, "loss": 1.6256, "step": 45270 }, { "epoch": 2.45, "learning_rate": 3.780698907281186e-05, "loss": 1.6083, "step": 45280 }, { "epoch": 2.45, "learning_rate": 3.7804284323271665e-05, "loss": 1.6302, "step": 45290 }, { "epoch": 2.45, "learning_rate": 3.780157957373147e-05, "loss": 1.611, "step": 45300 }, { "epoch": 2.45, "learning_rate": 3.7798874824191285e-05, "loss": 1.605, "step": 45310 }, { "epoch": 2.45, "learning_rate": 3.779617007465109e-05, "loss": 1.6645, "step": 45320 }, { "epoch": 2.45, "learning_rate": 3.77934653251109e-05, "loss": 1.5678, "step": 45330 }, { "epoch": 2.45, "learning_rate": 3.779076057557071e-05, "loss": 1.6109, "step": 45340 }, { "epoch": 2.45, "learning_rate": 3.778805582603051e-05, "loss": 1.6379, "step": 45350 }, { "epoch": 2.45, "learning_rate": 3.778535107649032e-05, "loss": 1.656, "step": 45360 }, { "epoch": 2.45, "learning_rate": 3.778264632695013e-05, "loss": 1.6101, "step": 45370 }, { "epoch": 2.45, "learning_rate": 3.777994157740993e-05, "loss": 1.6132, "step": 45380 }, { "epoch": 2.45, "learning_rate": 3.777723682786974e-05, "loss": 1.6343, "step": 45390 }, { "epoch": 2.45, "learning_rate": 3.777453207832955e-05, "loss": 1.5871, "step": 45400 }, { "epoch": 2.45, "learning_rate": 3.7771827328789354e-05, "loss": 1.6187, "step": 45410 }, { "epoch": 2.45, "learning_rate": 3.7769122579249164e-05, "loss": 1.6307, "step": 45420 }, { "epoch": 2.45, "learning_rate": 3.776641782970897e-05, "loss": 1.6187, "step": 45430 }, { "epoch": 2.46, "learning_rate": 3.7763713080168776e-05, "loss": 1.6128, "step": 45440 }, { "epoch": 2.46, "learning_rate": 3.7761008330628586e-05, "loss": 1.5631, "step": 45450 }, { "epoch": 2.46, "learning_rate": 3.7758303581088396e-05, "loss": 1.6188, "step": 45460 }, { "epoch": 2.46, "learning_rate": 3.77555988315482e-05, "loss": 1.6567, "step": 45470 }, { "epoch": 2.46, "learning_rate": 3.775289408200801e-05, "loss": 1.7345, "step": 45480 }, { "epoch": 2.46, "learning_rate": 3.775018933246782e-05, "loss": 1.6386, "step": 45490 }, { "epoch": 2.46, "learning_rate": 3.774748458292762e-05, "loss": 1.602, "step": 45500 }, { "epoch": 2.46, "learning_rate": 3.774477983338743e-05, "loss": 1.6041, "step": 45510 }, { "epoch": 2.46, "learning_rate": 3.774207508384724e-05, "loss": 1.617, "step": 45520 }, { "epoch": 2.46, "learning_rate": 3.773937033430704e-05, "loss": 1.6033, "step": 45530 }, { "epoch": 2.46, "learning_rate": 3.773666558476685e-05, "loss": 1.6389, "step": 45540 }, { "epoch": 2.46, "learning_rate": 3.773396083522666e-05, "loss": 1.6696, "step": 45550 }, { "epoch": 2.46, "learning_rate": 3.7731256085686465e-05, "loss": 1.592, "step": 45560 }, { "epoch": 2.46, "learning_rate": 3.7728551336146275e-05, "loss": 1.7334, "step": 45570 }, { "epoch": 2.46, "learning_rate": 3.7725846586606084e-05, "loss": 1.6593, "step": 45580 }, { "epoch": 2.46, "learning_rate": 3.772314183706589e-05, "loss": 1.558, "step": 45590 }, { "epoch": 2.46, "learning_rate": 3.77204370875257e-05, "loss": 1.6045, "step": 45600 }, { "epoch": 2.46, "learning_rate": 3.7717732337985506e-05, "loss": 1.6287, "step": 45610 }, { "epoch": 2.47, "learning_rate": 3.771502758844531e-05, "loss": 1.5885, "step": 45620 }, { "epoch": 2.47, "learning_rate": 3.771232283890512e-05, "loss": 1.6962, "step": 45630 }, { "epoch": 2.47, "learning_rate": 3.770961808936493e-05, "loss": 1.6837, "step": 45640 }, { "epoch": 2.47, "learning_rate": 3.770691333982473e-05, "loss": 1.6234, "step": 45650 }, { "epoch": 2.47, "learning_rate": 3.770420859028454e-05, "loss": 1.6259, "step": 45660 }, { "epoch": 2.47, "learning_rate": 3.770150384074435e-05, "loss": 1.6792, "step": 45670 }, { "epoch": 2.47, "learning_rate": 3.7698799091204154e-05, "loss": 1.6393, "step": 45680 }, { "epoch": 2.47, "learning_rate": 3.769609434166396e-05, "loss": 1.6354, "step": 45690 }, { "epoch": 2.47, "learning_rate": 3.769338959212377e-05, "loss": 1.6427, "step": 45700 }, { "epoch": 2.47, "learning_rate": 3.7690684842583576e-05, "loss": 1.6656, "step": 45710 }, { "epoch": 2.47, "learning_rate": 3.7687980093043385e-05, "loss": 1.663, "step": 45720 }, { "epoch": 2.47, "learning_rate": 3.7685275343503195e-05, "loss": 1.615, "step": 45730 }, { "epoch": 2.47, "learning_rate": 3.7682570593963005e-05, "loss": 1.6659, "step": 45740 }, { "epoch": 2.47, "learning_rate": 3.767986584442281e-05, "loss": 1.5559, "step": 45750 }, { "epoch": 2.47, "learning_rate": 3.767716109488262e-05, "loss": 1.5944, "step": 45760 }, { "epoch": 2.47, "learning_rate": 3.767445634534243e-05, "loss": 1.646, "step": 45770 }, { "epoch": 2.47, "learning_rate": 3.767175159580223e-05, "loss": 1.4954, "step": 45780 }, { "epoch": 2.47, "learning_rate": 3.766904684626204e-05, "loss": 1.6699, "step": 45790 }, { "epoch": 2.47, "learning_rate": 3.766634209672185e-05, "loss": 1.5993, "step": 45800 }, { "epoch": 2.48, "learning_rate": 3.766363734718165e-05, "loss": 1.663, "step": 45810 }, { "epoch": 2.48, "learning_rate": 3.766093259764146e-05, "loss": 1.5706, "step": 45820 }, { "epoch": 2.48, "learning_rate": 3.765822784810127e-05, "loss": 1.6417, "step": 45830 }, { "epoch": 2.48, "learning_rate": 3.7655523098561074e-05, "loss": 1.59, "step": 45840 }, { "epoch": 2.48, "learning_rate": 3.7652818349020884e-05, "loss": 1.5835, "step": 45850 }, { "epoch": 2.48, "learning_rate": 3.765011359948069e-05, "loss": 1.611, "step": 45860 }, { "epoch": 2.48, "learning_rate": 3.7647408849940496e-05, "loss": 1.5587, "step": 45870 }, { "epoch": 2.48, "learning_rate": 3.76447041004003e-05, "loss": 1.5513, "step": 45880 }, { "epoch": 2.48, "learning_rate": 3.7641999350860116e-05, "loss": 1.5916, "step": 45890 }, { "epoch": 2.48, "learning_rate": 3.763929460131992e-05, "loss": 1.6451, "step": 45900 }, { "epoch": 2.48, "learning_rate": 3.763658985177972e-05, "loss": 1.6014, "step": 45910 }, { "epoch": 2.48, "learning_rate": 3.763388510223954e-05, "loss": 1.5215, "step": 45920 }, { "epoch": 2.48, "learning_rate": 3.763118035269934e-05, "loss": 1.7008, "step": 45930 }, { "epoch": 2.48, "learning_rate": 3.7628475603159143e-05, "loss": 1.5092, "step": 45940 }, { "epoch": 2.48, "learning_rate": 3.762577085361896e-05, "loss": 1.6365, "step": 45950 }, { "epoch": 2.48, "learning_rate": 3.762306610407876e-05, "loss": 1.5764, "step": 45960 }, { "epoch": 2.48, "learning_rate": 3.762036135453857e-05, "loss": 1.6276, "step": 45970 }, { "epoch": 2.48, "learning_rate": 3.761765660499838e-05, "loss": 1.5852, "step": 45980 }, { "epoch": 2.49, "learning_rate": 3.7614951855458185e-05, "loss": 1.5575, "step": 45990 }, { "epoch": 2.49, "learning_rate": 3.7612247105917995e-05, "loss": 1.7025, "step": 46000 }, { "epoch": 2.49, "learning_rate": 3.7609542356377804e-05, "loss": 1.5954, "step": 46010 }, { "epoch": 2.49, "learning_rate": 3.760683760683761e-05, "loss": 1.7727, "step": 46020 }, { "epoch": 2.49, "learning_rate": 3.760413285729742e-05, "loss": 1.5901, "step": 46030 }, { "epoch": 2.49, "learning_rate": 3.7601428107757226e-05, "loss": 1.7332, "step": 46040 }, { "epoch": 2.49, "learning_rate": 3.759872335821703e-05, "loss": 1.5749, "step": 46050 }, { "epoch": 2.49, "learning_rate": 3.759601860867684e-05, "loss": 1.647, "step": 46060 }, { "epoch": 2.49, "learning_rate": 3.759331385913665e-05, "loss": 1.6087, "step": 46070 }, { "epoch": 2.49, "learning_rate": 3.759060910959645e-05, "loss": 1.6786, "step": 46080 }, { "epoch": 2.49, "learning_rate": 3.758790436005626e-05, "loss": 1.6537, "step": 46090 }, { "epoch": 2.49, "learning_rate": 3.758519961051607e-05, "loss": 1.6303, "step": 46100 }, { "epoch": 2.49, "learning_rate": 3.7582494860975874e-05, "loss": 1.666, "step": 46110 }, { "epoch": 2.49, "learning_rate": 3.757979011143568e-05, "loss": 1.609, "step": 46120 }, { "epoch": 2.49, "learning_rate": 3.757708536189549e-05, "loss": 1.6054, "step": 46130 }, { "epoch": 2.49, "learning_rate": 3.7574380612355296e-05, "loss": 1.6845, "step": 46140 }, { "epoch": 2.49, "learning_rate": 3.7571675862815105e-05, "loss": 1.6486, "step": 46150 }, { "epoch": 2.49, "learning_rate": 3.7568971113274915e-05, "loss": 1.6731, "step": 46160 }, { "epoch": 2.49, "learning_rate": 3.756626636373472e-05, "loss": 1.6032, "step": 46170 }, { "epoch": 2.5, "learning_rate": 3.756356161419453e-05, "loss": 1.6412, "step": 46180 }, { "epoch": 2.5, "learning_rate": 3.756085686465434e-05, "loss": 1.5872, "step": 46190 }, { "epoch": 2.5, "learning_rate": 3.755815211511414e-05, "loss": 1.6023, "step": 46200 }, { "epoch": 2.5, "learning_rate": 3.755544736557395e-05, "loss": 1.5982, "step": 46210 }, { "epoch": 2.5, "learning_rate": 3.755274261603376e-05, "loss": 1.5769, "step": 46220 }, { "epoch": 2.5, "learning_rate": 3.755003786649356e-05, "loss": 1.6713, "step": 46230 }, { "epoch": 2.5, "learning_rate": 3.754733311695337e-05, "loss": 1.644, "step": 46240 }, { "epoch": 2.5, "learning_rate": 3.754462836741318e-05, "loss": 1.6419, "step": 46250 }, { "epoch": 2.5, "learning_rate": 3.7541923617872984e-05, "loss": 1.6783, "step": 46260 }, { "epoch": 2.5, "learning_rate": 3.7539218868332794e-05, "loss": 1.6028, "step": 46270 }, { "epoch": 2.5, "learning_rate": 3.7536514118792604e-05, "loss": 1.5199, "step": 46280 }, { "epoch": 2.5, "learning_rate": 3.7533809369252407e-05, "loss": 1.6535, "step": 46290 }, { "epoch": 2.5, "learning_rate": 3.7531104619712216e-05, "loss": 1.6326, "step": 46300 }, { "epoch": 2.5, "learning_rate": 3.7528399870172026e-05, "loss": 1.596, "step": 46310 }, { "epoch": 2.5, "learning_rate": 3.752569512063183e-05, "loss": 1.6971, "step": 46320 }, { "epoch": 2.5, "learning_rate": 3.752299037109164e-05, "loss": 1.6672, "step": 46330 }, { "epoch": 2.5, "learning_rate": 3.752028562155145e-05, "loss": 1.6075, "step": 46340 }, { "epoch": 2.5, "learning_rate": 3.751758087201125e-05, "loss": 1.6175, "step": 46350 }, { "epoch": 2.51, "learning_rate": 3.751487612247106e-05, "loss": 1.5597, "step": 46360 }, { "epoch": 2.51, "learning_rate": 3.751217137293087e-05, "loss": 1.5532, "step": 46370 }, { "epoch": 2.51, "learning_rate": 3.750946662339068e-05, "loss": 1.6135, "step": 46380 }, { "epoch": 2.51, "learning_rate": 3.750676187385048e-05, "loss": 1.688, "step": 46390 }, { "epoch": 2.51, "learning_rate": 3.750405712431029e-05, "loss": 1.6502, "step": 46400 }, { "epoch": 2.51, "learning_rate": 3.75013523747701e-05, "loss": 1.6242, "step": 46410 }, { "epoch": 2.51, "learning_rate": 3.7498647625229905e-05, "loss": 1.6294, "step": 46420 }, { "epoch": 2.51, "learning_rate": 3.749594287568971e-05, "loss": 1.6863, "step": 46430 }, { "epoch": 2.51, "learning_rate": 3.7493238126149524e-05, "loss": 1.5601, "step": 46440 }, { "epoch": 2.51, "learning_rate": 3.749053337660933e-05, "loss": 1.666, "step": 46450 }, { "epoch": 2.51, "learning_rate": 3.748782862706913e-05, "loss": 1.5948, "step": 46460 }, { "epoch": 2.51, "learning_rate": 3.7485123877528946e-05, "loss": 1.5155, "step": 46470 }, { "epoch": 2.51, "learning_rate": 3.748241912798875e-05, "loss": 1.6395, "step": 46480 }, { "epoch": 2.51, "learning_rate": 3.747971437844855e-05, "loss": 1.693, "step": 46490 }, { "epoch": 2.51, "learning_rate": 3.747700962890837e-05, "loss": 1.6849, "step": 46500 }, { "epoch": 2.51, "learning_rate": 3.747430487936817e-05, "loss": 1.5796, "step": 46510 }, { "epoch": 2.51, "learning_rate": 3.7471600129827974e-05, "loss": 1.6937, "step": 46520 }, { "epoch": 2.51, "learning_rate": 3.746889538028779e-05, "loss": 1.6111, "step": 46530 }, { "epoch": 2.51, "learning_rate": 3.7466190630747594e-05, "loss": 1.7095, "step": 46540 }, { "epoch": 2.52, "learning_rate": 3.7463485881207396e-05, "loss": 1.6245, "step": 46550 }, { "epoch": 2.52, "learning_rate": 3.746078113166721e-05, "loss": 1.6608, "step": 46560 }, { "epoch": 2.52, "learning_rate": 3.7458076382127016e-05, "loss": 1.6203, "step": 46570 }, { "epoch": 2.52, "learning_rate": 3.745537163258682e-05, "loss": 1.6095, "step": 46580 }, { "epoch": 2.52, "learning_rate": 3.7452666883046635e-05, "loss": 1.6805, "step": 46590 }, { "epoch": 2.52, "learning_rate": 3.744996213350644e-05, "loss": 1.5844, "step": 46600 }, { "epoch": 2.52, "learning_rate": 3.744725738396625e-05, "loss": 1.6378, "step": 46610 }, { "epoch": 2.52, "learning_rate": 3.744455263442606e-05, "loss": 1.6529, "step": 46620 }, { "epoch": 2.52, "learning_rate": 3.744184788488586e-05, "loss": 1.5945, "step": 46630 }, { "epoch": 2.52, "learning_rate": 3.743914313534567e-05, "loss": 1.604, "step": 46640 }, { "epoch": 2.52, "learning_rate": 3.743643838580548e-05, "loss": 1.6498, "step": 46650 }, { "epoch": 2.52, "learning_rate": 3.743373363626528e-05, "loss": 1.7333, "step": 46660 }, { "epoch": 2.52, "learning_rate": 3.743102888672509e-05, "loss": 1.7058, "step": 46670 }, { "epoch": 2.52, "learning_rate": 3.74283241371849e-05, "loss": 1.7334, "step": 46680 }, { "epoch": 2.52, "learning_rate": 3.7425619387644704e-05, "loss": 1.6765, "step": 46690 }, { "epoch": 2.52, "learning_rate": 3.7422914638104514e-05, "loss": 1.6672, "step": 46700 }, { "epoch": 2.52, "learning_rate": 3.7420209888564324e-05, "loss": 1.6696, "step": 46710 }, { "epoch": 2.52, "learning_rate": 3.741750513902413e-05, "loss": 1.5529, "step": 46720 }, { "epoch": 2.53, "learning_rate": 3.7414800389483936e-05, "loss": 1.5772, "step": 46730 }, { "epoch": 2.53, "learning_rate": 3.7412095639943746e-05, "loss": 1.6151, "step": 46740 }, { "epoch": 2.53, "learning_rate": 3.740939089040355e-05, "loss": 1.6017, "step": 46750 }, { "epoch": 2.53, "learning_rate": 3.740668614086336e-05, "loss": 1.5138, "step": 46760 }, { "epoch": 2.53, "learning_rate": 3.740398139132317e-05, "loss": 1.6752, "step": 46770 }, { "epoch": 2.53, "learning_rate": 3.740127664178297e-05, "loss": 1.6154, "step": 46780 }, { "epoch": 2.53, "learning_rate": 3.739857189224278e-05, "loss": 1.5802, "step": 46790 }, { "epoch": 2.53, "learning_rate": 3.739586714270259e-05, "loss": 1.6169, "step": 46800 }, { "epoch": 2.53, "learning_rate": 3.739316239316239e-05, "loss": 1.6, "step": 46810 }, { "epoch": 2.53, "learning_rate": 3.73904576436222e-05, "loss": 1.6073, "step": 46820 }, { "epoch": 2.53, "learning_rate": 3.738775289408201e-05, "loss": 1.5123, "step": 46830 }, { "epoch": 2.53, "learning_rate": 3.7385048144541815e-05, "loss": 1.59, "step": 46840 }, { "epoch": 2.53, "learning_rate": 3.7382343395001625e-05, "loss": 1.6414, "step": 46850 }, { "epoch": 2.53, "learning_rate": 3.7379638645461435e-05, "loss": 1.6822, "step": 46860 }, { "epoch": 2.53, "learning_rate": 3.737693389592124e-05, "loss": 1.6505, "step": 46870 }, { "epoch": 2.53, "learning_rate": 3.737422914638105e-05, "loss": 1.5966, "step": 46880 }, { "epoch": 2.53, "learning_rate": 3.737152439684086e-05, "loss": 1.6322, "step": 46890 }, { "epoch": 2.53, "learning_rate": 3.736881964730066e-05, "loss": 1.6651, "step": 46900 }, { "epoch": 2.53, "learning_rate": 3.736611489776047e-05, "loss": 1.6076, "step": 46910 }, { "epoch": 2.54, "learning_rate": 3.736341014822028e-05, "loss": 1.6571, "step": 46920 }, { "epoch": 2.54, "learning_rate": 3.736070539868008e-05, "loss": 1.6275, "step": 46930 }, { "epoch": 2.54, "learning_rate": 3.735800064913989e-05, "loss": 1.6049, "step": 46940 }, { "epoch": 2.54, "learning_rate": 3.73552958995997e-05, "loss": 1.6352, "step": 46950 }, { "epoch": 2.54, "learning_rate": 3.7352591150059504e-05, "loss": 1.5979, "step": 46960 }, { "epoch": 2.54, "learning_rate": 3.7349886400519314e-05, "loss": 1.5957, "step": 46970 }, { "epoch": 2.54, "learning_rate": 3.7347181650979116e-05, "loss": 1.6819, "step": 46980 }, { "epoch": 2.54, "learning_rate": 3.7344476901438926e-05, "loss": 1.5851, "step": 46990 }, { "epoch": 2.54, "learning_rate": 3.7341772151898736e-05, "loss": 1.6632, "step": 47000 }, { "epoch": 2.54, "learning_rate": 3.733906740235854e-05, "loss": 1.6658, "step": 47010 }, { "epoch": 2.54, "learning_rate": 3.7336362652818355e-05, "loss": 1.6167, "step": 47020 }, { "epoch": 2.54, "learning_rate": 3.733365790327816e-05, "loss": 1.5815, "step": 47030 }, { "epoch": 2.54, "learning_rate": 3.733095315373796e-05, "loss": 1.5484, "step": 47040 }, { "epoch": 2.54, "learning_rate": 3.732824840419778e-05, "loss": 1.6705, "step": 47050 }, { "epoch": 2.54, "learning_rate": 3.732554365465758e-05, "loss": 1.6388, "step": 47060 }, { "epoch": 2.54, "learning_rate": 3.732283890511738e-05, "loss": 1.6066, "step": 47070 }, { "epoch": 2.54, "learning_rate": 3.73201341555772e-05, "loss": 1.6059, "step": 47080 }, { "epoch": 2.54, "learning_rate": 3.7317429406037e-05, "loss": 1.5836, "step": 47090 }, { "epoch": 2.55, "learning_rate": 3.7314724656496805e-05, "loss": 1.6437, "step": 47100 }, { "epoch": 2.55, "learning_rate": 3.731201990695662e-05, "loss": 1.6351, "step": 47110 }, { "epoch": 2.55, "learning_rate": 3.7309315157416424e-05, "loss": 1.612, "step": 47120 }, { "epoch": 2.55, "learning_rate": 3.730661040787623e-05, "loss": 1.6204, "step": 47130 }, { "epoch": 2.55, "learning_rate": 3.7303905658336044e-05, "loss": 1.7406, "step": 47140 }, { "epoch": 2.55, "learning_rate": 3.730120090879585e-05, "loss": 1.6404, "step": 47150 }, { "epoch": 2.55, "learning_rate": 3.729849615925565e-05, "loss": 1.6844, "step": 47160 }, { "epoch": 2.55, "learning_rate": 3.7295791409715466e-05, "loss": 1.6613, "step": 47170 }, { "epoch": 2.55, "learning_rate": 3.729308666017527e-05, "loss": 1.6406, "step": 47180 }, { "epoch": 2.55, "learning_rate": 3.729038191063507e-05, "loss": 1.6668, "step": 47190 }, { "epoch": 2.55, "learning_rate": 3.728767716109489e-05, "loss": 1.5622, "step": 47200 }, { "epoch": 2.55, "learning_rate": 3.728497241155469e-05, "loss": 1.6033, "step": 47210 }, { "epoch": 2.55, "learning_rate": 3.72822676620145e-05, "loss": 1.6267, "step": 47220 }, { "epoch": 2.55, "learning_rate": 3.727956291247431e-05, "loss": 1.586, "step": 47230 }, { "epoch": 2.55, "learning_rate": 3.727685816293411e-05, "loss": 1.6185, "step": 47240 }, { "epoch": 2.55, "learning_rate": 3.727415341339392e-05, "loss": 1.6321, "step": 47250 }, { "epoch": 2.55, "learning_rate": 3.727144866385373e-05, "loss": 1.6378, "step": 47260 }, { "epoch": 2.55, "learning_rate": 3.7268743914313535e-05, "loss": 1.5923, "step": 47270 }, { "epoch": 2.55, "learning_rate": 3.7266039164773345e-05, "loss": 1.6289, "step": 47280 }, { "epoch": 2.56, "learning_rate": 3.7263334415233155e-05, "loss": 1.6171, "step": 47290 }, { "epoch": 2.56, "learning_rate": 3.726062966569296e-05, "loss": 1.6389, "step": 47300 }, { "epoch": 2.56, "learning_rate": 3.725792491615277e-05, "loss": 1.6467, "step": 47310 }, { "epoch": 2.56, "learning_rate": 3.725522016661258e-05, "loss": 1.5704, "step": 47320 }, { "epoch": 2.56, "learning_rate": 3.725251541707238e-05, "loss": 1.6244, "step": 47330 }, { "epoch": 2.56, "learning_rate": 3.724981066753219e-05, "loss": 1.509, "step": 47340 }, { "epoch": 2.56, "learning_rate": 3.7247105917992e-05, "loss": 1.5656, "step": 47350 }, { "epoch": 2.56, "learning_rate": 3.72444011684518e-05, "loss": 1.596, "step": 47360 }, { "epoch": 2.56, "learning_rate": 3.724169641891161e-05, "loss": 1.6157, "step": 47370 }, { "epoch": 2.56, "learning_rate": 3.723899166937142e-05, "loss": 1.5967, "step": 47380 }, { "epoch": 2.56, "learning_rate": 3.7236286919831224e-05, "loss": 1.6192, "step": 47390 }, { "epoch": 2.56, "learning_rate": 3.7233582170291034e-05, "loss": 1.7636, "step": 47400 }, { "epoch": 2.56, "learning_rate": 3.723087742075084e-05, "loss": 1.5667, "step": 47410 }, { "epoch": 2.56, "learning_rate": 3.7228172671210646e-05, "loss": 1.6091, "step": 47420 }, { "epoch": 2.56, "learning_rate": 3.7225467921670456e-05, "loss": 1.7043, "step": 47430 }, { "epoch": 2.56, "learning_rate": 3.7222763172130265e-05, "loss": 1.6275, "step": 47440 }, { "epoch": 2.56, "learning_rate": 3.722005842259007e-05, "loss": 1.6375, "step": 47450 }, { "epoch": 2.56, "learning_rate": 3.721735367304988e-05, "loss": 1.6497, "step": 47460 }, { "epoch": 2.57, "learning_rate": 3.721464892350969e-05, "loss": 1.7524, "step": 47470 }, { "epoch": 2.57, "learning_rate": 3.721194417396949e-05, "loss": 1.6612, "step": 47480 }, { "epoch": 2.57, "learning_rate": 3.72092394244293e-05, "loss": 1.5568, "step": 47490 }, { "epoch": 2.57, "learning_rate": 3.720653467488911e-05, "loss": 1.6222, "step": 47500 }, { "epoch": 2.57, "learning_rate": 3.720382992534891e-05, "loss": 1.4606, "step": 47510 }, { "epoch": 2.57, "learning_rate": 3.720112517580872e-05, "loss": 1.6303, "step": 47520 }, { "epoch": 2.57, "learning_rate": 3.719842042626853e-05, "loss": 1.6295, "step": 47530 }, { "epoch": 2.57, "learning_rate": 3.7195715676728335e-05, "loss": 1.6588, "step": 47540 }, { "epoch": 2.57, "learning_rate": 3.7193010927188144e-05, "loss": 1.6554, "step": 47550 }, { "epoch": 2.57, "learning_rate": 3.719030617764795e-05, "loss": 1.5977, "step": 47560 }, { "epoch": 2.57, "learning_rate": 3.718760142810776e-05, "loss": 1.7371, "step": 47570 }, { "epoch": 2.57, "learning_rate": 3.718489667856757e-05, "loss": 1.6277, "step": 47580 }, { "epoch": 2.57, "learning_rate": 3.718219192902737e-05, "loss": 1.6908, "step": 47590 }, { "epoch": 2.57, "learning_rate": 3.717948717948718e-05, "loss": 1.6722, "step": 47600 }, { "epoch": 2.57, "learning_rate": 3.717678242994699e-05, "loss": 1.6323, "step": 47610 }, { "epoch": 2.57, "learning_rate": 3.717407768040679e-05, "loss": 1.6147, "step": 47620 }, { "epoch": 2.57, "learning_rate": 3.717137293086661e-05, "loss": 1.5944, "step": 47630 }, { "epoch": 2.57, "learning_rate": 3.716866818132641e-05, "loss": 1.6059, "step": 47640 }, { "epoch": 2.57, "learning_rate": 3.7165963431786214e-05, "loss": 1.6304, "step": 47650 }, { "epoch": 2.58, "learning_rate": 3.716325868224603e-05, "loss": 1.6586, "step": 47660 }, { "epoch": 2.58, "learning_rate": 3.716055393270583e-05, "loss": 1.6352, "step": 47670 }, { "epoch": 2.58, "learning_rate": 3.7157849183165636e-05, "loss": 1.6423, "step": 47680 }, { "epoch": 2.58, "learning_rate": 3.715514443362545e-05, "loss": 1.6923, "step": 47690 }, { "epoch": 2.58, "learning_rate": 3.7152439684085255e-05, "loss": 1.5361, "step": 47700 }, { "epoch": 2.58, "learning_rate": 3.714973493454506e-05, "loss": 1.5615, "step": 47710 }, { "epoch": 2.58, "learning_rate": 3.7147030185004875e-05, "loss": 1.7253, "step": 47720 }, { "epoch": 2.58, "learning_rate": 3.714432543546468e-05, "loss": 1.6414, "step": 47730 }, { "epoch": 2.58, "learning_rate": 3.714162068592448e-05, "loss": 1.6084, "step": 47740 }, { "epoch": 2.58, "learning_rate": 3.71389159363843e-05, "loss": 1.6826, "step": 47750 }, { "epoch": 2.58, "learning_rate": 3.71362111868441e-05, "loss": 1.6584, "step": 47760 }, { "epoch": 2.58, "learning_rate": 3.71335064373039e-05, "loss": 1.6371, "step": 47770 }, { "epoch": 2.58, "learning_rate": 3.713080168776372e-05, "loss": 1.5711, "step": 47780 }, { "epoch": 2.58, "learning_rate": 3.712809693822352e-05, "loss": 1.4955, "step": 47790 }, { "epoch": 2.58, "learning_rate": 3.7125392188683325e-05, "loss": 1.5549, "step": 47800 }, { "epoch": 2.58, "learning_rate": 3.712268743914314e-05, "loss": 1.6008, "step": 47810 }, { "epoch": 2.58, "learning_rate": 3.7119982689602944e-05, "loss": 1.5453, "step": 47820 }, { "epoch": 2.58, "learning_rate": 3.711727794006275e-05, "loss": 1.5853, "step": 47830 }, { "epoch": 2.59, "learning_rate": 3.711457319052256e-05, "loss": 1.5757, "step": 47840 }, { "epoch": 2.59, "learning_rate": 3.7111868440982366e-05, "loss": 1.5837, "step": 47850 }, { "epoch": 2.59, "learning_rate": 3.7109163691442176e-05, "loss": 1.5375, "step": 47860 }, { "epoch": 2.59, "learning_rate": 3.7106458941901985e-05, "loss": 1.6396, "step": 47870 }, { "epoch": 2.59, "learning_rate": 3.710375419236179e-05, "loss": 1.6013, "step": 47880 }, { "epoch": 2.59, "learning_rate": 3.71010494428216e-05, "loss": 1.6494, "step": 47890 }, { "epoch": 2.59, "learning_rate": 3.709834469328141e-05, "loss": 1.5941, "step": 47900 }, { "epoch": 2.59, "learning_rate": 3.709563994374121e-05, "loss": 1.5899, "step": 47910 }, { "epoch": 2.59, "learning_rate": 3.709293519420102e-05, "loss": 1.6174, "step": 47920 }, { "epoch": 2.59, "learning_rate": 3.709023044466083e-05, "loss": 1.6122, "step": 47930 }, { "epoch": 2.59, "learning_rate": 3.708752569512063e-05, "loss": 1.6007, "step": 47940 }, { "epoch": 2.59, "learning_rate": 3.708482094558044e-05, "loss": 1.5347, "step": 47950 }, { "epoch": 2.59, "learning_rate": 3.708211619604025e-05, "loss": 1.6578, "step": 47960 }, { "epoch": 2.59, "learning_rate": 3.7079411446500055e-05, "loss": 1.6284, "step": 47970 }, { "epoch": 2.59, "learning_rate": 3.7076706696959865e-05, "loss": 1.5983, "step": 47980 }, { "epoch": 2.59, "learning_rate": 3.7074001947419674e-05, "loss": 1.5425, "step": 47990 }, { "epoch": 2.59, "learning_rate": 3.707129719787948e-05, "loss": 1.6543, "step": 48000 }, { "epoch": 2.59, "learning_rate": 3.706859244833929e-05, "loss": 1.5388, "step": 48010 }, { "epoch": 2.59, "learning_rate": 3.7065887698799096e-05, "loss": 1.668, "step": 48020 }, { "epoch": 2.6, "learning_rate": 3.70631829492589e-05, "loss": 1.5586, "step": 48030 }, { "epoch": 2.6, "learning_rate": 3.706047819971871e-05, "loss": 1.605, "step": 48040 }, { "epoch": 2.6, "learning_rate": 3.705777345017852e-05, "loss": 1.6018, "step": 48050 }, { "epoch": 2.6, "learning_rate": 3.705506870063832e-05, "loss": 1.6803, "step": 48060 }, { "epoch": 2.6, "learning_rate": 3.705236395109813e-05, "loss": 1.6466, "step": 48070 }, { "epoch": 2.6, "learning_rate": 3.704965920155794e-05, "loss": 1.6655, "step": 48080 }, { "epoch": 2.6, "learning_rate": 3.7046954452017744e-05, "loss": 1.6241, "step": 48090 }, { "epoch": 2.6, "learning_rate": 3.704424970247755e-05, "loss": 1.6463, "step": 48100 }, { "epoch": 2.6, "learning_rate": 3.7041544952937356e-05, "loss": 1.6734, "step": 48110 }, { "epoch": 2.6, "learning_rate": 3.7038840203397166e-05, "loss": 1.6036, "step": 48120 }, { "epoch": 2.6, "learning_rate": 3.7036135453856975e-05, "loss": 1.5698, "step": 48130 }, { "epoch": 2.6, "learning_rate": 3.703343070431678e-05, "loss": 1.6187, "step": 48140 }, { "epoch": 2.6, "learning_rate": 3.703072595477659e-05, "loss": 1.6781, "step": 48150 }, { "epoch": 2.6, "learning_rate": 3.70280212052364e-05, "loss": 1.6643, "step": 48160 }, { "epoch": 2.6, "learning_rate": 3.70253164556962e-05, "loss": 1.6254, "step": 48170 }, { "epoch": 2.6, "learning_rate": 3.702261170615601e-05, "loss": 1.6343, "step": 48180 }, { "epoch": 2.6, "learning_rate": 3.701990695661582e-05, "loss": 1.6096, "step": 48190 }, { "epoch": 2.6, "learning_rate": 3.701720220707562e-05, "loss": 1.6216, "step": 48200 }, { "epoch": 2.6, "learning_rate": 3.701449745753543e-05, "loss": 1.6152, "step": 48210 }, { "epoch": 2.61, "learning_rate": 3.701179270799524e-05, "loss": 1.6421, "step": 48220 }, { "epoch": 2.61, "learning_rate": 3.7009087958455045e-05, "loss": 1.6985, "step": 48230 }, { "epoch": 2.61, "learning_rate": 3.7006383208914854e-05, "loss": 1.6039, "step": 48240 }, { "epoch": 2.61, "learning_rate": 3.7003678459374664e-05, "loss": 1.5442, "step": 48250 }, { "epoch": 2.61, "learning_rate": 3.700097370983447e-05, "loss": 1.6202, "step": 48260 }, { "epoch": 2.61, "learning_rate": 3.699826896029428e-05, "loss": 1.6979, "step": 48270 }, { "epoch": 2.61, "learning_rate": 3.6995564210754086e-05, "loss": 1.6641, "step": 48280 }, { "epoch": 2.61, "learning_rate": 3.699285946121389e-05, "loss": 1.6232, "step": 48290 }, { "epoch": 2.61, "learning_rate": 3.6990154711673706e-05, "loss": 1.6047, "step": 48300 }, { "epoch": 2.61, "learning_rate": 3.698744996213351e-05, "loss": 1.6766, "step": 48310 }, { "epoch": 2.61, "learning_rate": 3.698474521259331e-05, "loss": 1.601, "step": 48320 }, { "epoch": 2.61, "learning_rate": 3.698204046305313e-05, "loss": 1.6273, "step": 48330 }, { "epoch": 2.61, "learning_rate": 3.697933571351293e-05, "loss": 1.6373, "step": 48340 }, { "epoch": 2.61, "learning_rate": 3.6976630963972733e-05, "loss": 1.6288, "step": 48350 }, { "epoch": 2.61, "learning_rate": 3.697392621443255e-05, "loss": 1.5841, "step": 48360 }, { "epoch": 2.61, "learning_rate": 3.697122146489235e-05, "loss": 1.5953, "step": 48370 }, { "epoch": 2.61, "learning_rate": 3.6968516715352156e-05, "loss": 1.6706, "step": 48380 }, { "epoch": 2.61, "learning_rate": 3.696581196581197e-05, "loss": 1.6758, "step": 48390 }, { "epoch": 2.62, "learning_rate": 3.6963107216271775e-05, "loss": 1.5308, "step": 48400 }, { "epoch": 2.62, "learning_rate": 3.696040246673158e-05, "loss": 1.6838, "step": 48410 }, { "epoch": 2.62, "learning_rate": 3.6957697717191394e-05, "loss": 1.5968, "step": 48420 }, { "epoch": 2.62, "learning_rate": 3.69549929676512e-05, "loss": 1.7128, "step": 48430 }, { "epoch": 2.62, "learning_rate": 3.6952288218111e-05, "loss": 1.6252, "step": 48440 }, { "epoch": 2.62, "learning_rate": 3.6949583468570816e-05, "loss": 1.5902, "step": 48450 }, { "epoch": 2.62, "learning_rate": 3.694687871903062e-05, "loss": 1.6709, "step": 48460 }, { "epoch": 2.62, "learning_rate": 3.694417396949042e-05, "loss": 1.6394, "step": 48470 }, { "epoch": 2.62, "learning_rate": 3.694146921995024e-05, "loss": 1.5983, "step": 48480 }, { "epoch": 2.62, "learning_rate": 3.693876447041004e-05, "loss": 1.6448, "step": 48490 }, { "epoch": 2.62, "learning_rate": 3.693605972086985e-05, "loss": 1.6554, "step": 48500 }, { "epoch": 2.62, "learning_rate": 3.693335497132966e-05, "loss": 1.5903, "step": 48510 }, { "epoch": 2.62, "learning_rate": 3.6930650221789464e-05, "loss": 1.6726, "step": 48520 }, { "epoch": 2.62, "learning_rate": 3.692794547224927e-05, "loss": 1.6026, "step": 48530 }, { "epoch": 2.62, "learning_rate": 3.692524072270908e-05, "loss": 1.6626, "step": 48540 }, { "epoch": 2.62, "learning_rate": 3.6922535973168886e-05, "loss": 1.676, "step": 48550 }, { "epoch": 2.62, "learning_rate": 3.6919831223628695e-05, "loss": 1.6417, "step": 48560 }, { "epoch": 2.62, "learning_rate": 3.6917126474088505e-05, "loss": 1.5519, "step": 48570 }, { "epoch": 2.62, "learning_rate": 3.691442172454831e-05, "loss": 1.6642, "step": 48580 }, { "epoch": 2.63, "learning_rate": 3.691171697500812e-05, "loss": 1.6795, "step": 48590 }, { "epoch": 2.63, "learning_rate": 3.690901222546793e-05, "loss": 1.6715, "step": 48600 }, { "epoch": 2.63, "learning_rate": 3.690630747592773e-05, "loss": 1.6514, "step": 48610 }, { "epoch": 2.63, "learning_rate": 3.690360272638754e-05, "loss": 1.6786, "step": 48620 }, { "epoch": 2.63, "learning_rate": 3.690089797684735e-05, "loss": 1.6157, "step": 48630 }, { "epoch": 2.63, "learning_rate": 3.689819322730715e-05, "loss": 1.5218, "step": 48640 }, { "epoch": 2.63, "learning_rate": 3.689548847776696e-05, "loss": 1.6687, "step": 48650 }, { "epoch": 2.63, "learning_rate": 3.6892783728226765e-05, "loss": 1.6123, "step": 48660 }, { "epoch": 2.63, "learning_rate": 3.6890078978686574e-05, "loss": 1.7121, "step": 48670 }, { "epoch": 2.63, "learning_rate": 3.6887374229146384e-05, "loss": 1.6008, "step": 48680 }, { "epoch": 2.63, "learning_rate": 3.688466947960619e-05, "loss": 1.6604, "step": 48690 }, { "epoch": 2.63, "learning_rate": 3.6881964730065997e-05, "loss": 1.6072, "step": 48700 }, { "epoch": 2.63, "learning_rate": 3.6879259980525806e-05, "loss": 1.6297, "step": 48710 }, { "epoch": 2.63, "learning_rate": 3.687655523098561e-05, "loss": 1.5682, "step": 48720 }, { "epoch": 2.63, "learning_rate": 3.687385048144542e-05, "loss": 1.5572, "step": 48730 }, { "epoch": 2.63, "learning_rate": 3.687114573190523e-05, "loss": 1.6813, "step": 48740 }, { "epoch": 2.63, "learning_rate": 3.686844098236503e-05, "loss": 1.6471, "step": 48750 }, { "epoch": 2.63, "learning_rate": 3.686573623282484e-05, "loss": 1.6492, "step": 48760 }, { "epoch": 2.64, "learning_rate": 3.686303148328465e-05, "loss": 1.6933, "step": 48770 }, { "epoch": 2.64, "learning_rate": 3.6860326733744453e-05, "loss": 1.5944, "step": 48780 }, { "epoch": 2.64, "learning_rate": 3.685762198420426e-05, "loss": 1.5808, "step": 48790 }, { "epoch": 2.64, "learning_rate": 3.685491723466407e-05, "loss": 1.5712, "step": 48800 }, { "epoch": 2.64, "learning_rate": 3.6852212485123876e-05, "loss": 1.6293, "step": 48810 }, { "epoch": 2.64, "learning_rate": 3.6849507735583685e-05, "loss": 1.5977, "step": 48820 }, { "epoch": 2.64, "learning_rate": 3.6846802986043495e-05, "loss": 1.5279, "step": 48830 }, { "epoch": 2.64, "learning_rate": 3.68440982365033e-05, "loss": 1.6113, "step": 48840 }, { "epoch": 2.64, "learning_rate": 3.684139348696311e-05, "loss": 1.6407, "step": 48850 }, { "epoch": 2.64, "learning_rate": 3.683868873742292e-05, "loss": 1.6197, "step": 48860 }, { "epoch": 2.64, "learning_rate": 3.683598398788272e-05, "loss": 1.6224, "step": 48870 }, { "epoch": 2.64, "learning_rate": 3.683327923834253e-05, "loss": 1.6379, "step": 48880 }, { "epoch": 2.64, "learning_rate": 3.683057448880234e-05, "loss": 1.6578, "step": 48890 }, { "epoch": 2.64, "learning_rate": 3.682786973926214e-05, "loss": 1.5661, "step": 48900 }, { "epoch": 2.64, "learning_rate": 3.682516498972196e-05, "loss": 1.6036, "step": 48910 }, { "epoch": 2.64, "learning_rate": 3.682246024018176e-05, "loss": 1.6869, "step": 48920 }, { "epoch": 2.64, "learning_rate": 3.6819755490641564e-05, "loss": 1.5575, "step": 48930 }, { "epoch": 2.64, "learning_rate": 3.681705074110138e-05, "loss": 1.6252, "step": 48940 }, { "epoch": 2.64, "learning_rate": 3.6814345991561184e-05, "loss": 1.5854, "step": 48950 }, { "epoch": 2.65, "learning_rate": 3.6811641242020986e-05, "loss": 1.6168, "step": 48960 }, { "epoch": 2.65, "learning_rate": 3.68089364924808e-05, "loss": 1.6925, "step": 48970 }, { "epoch": 2.65, "learning_rate": 3.6806231742940606e-05, "loss": 1.5719, "step": 48980 }, { "epoch": 2.65, "learning_rate": 3.680352699340041e-05, "loss": 1.6077, "step": 48990 }, { "epoch": 2.65, "learning_rate": 3.6800822243860225e-05, "loss": 1.661, "step": 49000 }, { "epoch": 2.65, "learning_rate": 3.679811749432003e-05, "loss": 1.6449, "step": 49010 }, { "epoch": 2.65, "learning_rate": 3.679541274477983e-05, "loss": 1.5832, "step": 49020 }, { "epoch": 2.65, "learning_rate": 3.679270799523965e-05, "loss": 1.6615, "step": 49030 }, { "epoch": 2.65, "learning_rate": 3.679000324569945e-05, "loss": 1.5236, "step": 49040 }, { "epoch": 2.65, "learning_rate": 3.678729849615925e-05, "loss": 1.7139, "step": 49050 }, { "epoch": 2.65, "learning_rate": 3.678459374661907e-05, "loss": 1.6503, "step": 49060 }, { "epoch": 2.65, "learning_rate": 3.678188899707887e-05, "loss": 1.6797, "step": 49070 }, { "epoch": 2.65, "learning_rate": 3.6779184247538675e-05, "loss": 1.6168, "step": 49080 }, { "epoch": 2.65, "learning_rate": 3.677647949799849e-05, "loss": 1.6167, "step": 49090 }, { "epoch": 2.65, "learning_rate": 3.6773774748458294e-05, "loss": 1.5345, "step": 49100 }, { "epoch": 2.65, "learning_rate": 3.67710699989181e-05, "loss": 1.6364, "step": 49110 }, { "epoch": 2.65, "learning_rate": 3.6768365249377914e-05, "loss": 1.6916, "step": 49120 }, { "epoch": 2.65, "learning_rate": 3.6765660499837717e-05, "loss": 1.6527, "step": 49130 }, { "epoch": 2.66, "learning_rate": 3.6762955750297526e-05, "loss": 1.5868, "step": 49140 }, { "epoch": 2.66, "learning_rate": 3.6760251000757336e-05, "loss": 1.6273, "step": 49150 }, { "epoch": 2.66, "learning_rate": 3.675754625121714e-05, "loss": 1.6748, "step": 49160 }, { "epoch": 2.66, "learning_rate": 3.675484150167695e-05, "loss": 1.6614, "step": 49170 }, { "epoch": 2.66, "learning_rate": 3.675213675213676e-05, "loss": 1.6211, "step": 49180 }, { "epoch": 2.66, "learning_rate": 3.674943200259656e-05, "loss": 1.6037, "step": 49190 }, { "epoch": 2.66, "learning_rate": 3.674672725305637e-05, "loss": 1.5175, "step": 49200 }, { "epoch": 2.66, "learning_rate": 3.674402250351618e-05, "loss": 1.6285, "step": 49210 }, { "epoch": 2.66, "learning_rate": 3.674131775397598e-05, "loss": 1.6386, "step": 49220 }, { "epoch": 2.66, "learning_rate": 3.673861300443579e-05, "loss": 1.5289, "step": 49230 }, { "epoch": 2.66, "learning_rate": 3.6735908254895596e-05, "loss": 1.6802, "step": 49240 }, { "epoch": 2.66, "learning_rate": 3.6733203505355405e-05, "loss": 1.5482, "step": 49250 }, { "epoch": 2.66, "learning_rate": 3.6730498755815215e-05, "loss": 1.6246, "step": 49260 }, { "epoch": 2.66, "learning_rate": 3.672779400627502e-05, "loss": 1.6144, "step": 49270 }, { "epoch": 2.66, "learning_rate": 3.672508925673483e-05, "loss": 1.6023, "step": 49280 }, { "epoch": 2.66, "learning_rate": 3.672238450719464e-05, "loss": 1.6764, "step": 49290 }, { "epoch": 2.66, "learning_rate": 3.671967975765444e-05, "loss": 1.6006, "step": 49300 }, { "epoch": 2.66, "learning_rate": 3.671697500811425e-05, "loss": 1.621, "step": 49310 }, { "epoch": 2.66, "learning_rate": 3.671427025857406e-05, "loss": 1.5988, "step": 49320 }, { "epoch": 2.67, "learning_rate": 3.671156550903386e-05, "loss": 1.6777, "step": 49330 }, { "epoch": 2.67, "learning_rate": 3.670886075949367e-05, "loss": 1.5923, "step": 49340 }, { "epoch": 2.67, "learning_rate": 3.670615600995348e-05, "loss": 1.637, "step": 49350 }, { "epoch": 2.67, "learning_rate": 3.6703451260413284e-05, "loss": 1.6167, "step": 49360 }, { "epoch": 2.67, "learning_rate": 3.6700746510873094e-05, "loss": 1.621, "step": 49370 }, { "epoch": 2.67, "learning_rate": 3.6698041761332904e-05, "loss": 1.6283, "step": 49380 }, { "epoch": 2.67, "learning_rate": 3.6695337011792706e-05, "loss": 1.6267, "step": 49390 }, { "epoch": 2.67, "learning_rate": 3.6692632262252516e-05, "loss": 1.6414, "step": 49400 }, { "epoch": 2.67, "learning_rate": 3.6689927512712326e-05, "loss": 1.5864, "step": 49410 }, { "epoch": 2.67, "learning_rate": 3.668722276317213e-05, "loss": 1.6424, "step": 49420 }, { "epoch": 2.67, "learning_rate": 3.668451801363194e-05, "loss": 1.69, "step": 49430 }, { "epoch": 2.67, "learning_rate": 3.668181326409175e-05, "loss": 1.6957, "step": 49440 }, { "epoch": 2.67, "learning_rate": 3.667910851455155e-05, "loss": 1.6286, "step": 49450 }, { "epoch": 2.67, "learning_rate": 3.667640376501136e-05, "loss": 1.6443, "step": 49460 }, { "epoch": 2.67, "learning_rate": 3.667369901547117e-05, "loss": 1.5396, "step": 49470 }, { "epoch": 2.67, "learning_rate": 3.667099426593097e-05, "loss": 1.6024, "step": 49480 }, { "epoch": 2.67, "learning_rate": 3.666828951639078e-05, "loss": 1.6279, "step": 49490 }, { "epoch": 2.67, "learning_rate": 3.666558476685059e-05, "loss": 1.7265, "step": 49500 }, { "epoch": 2.68, "learning_rate": 3.6662880017310395e-05, "loss": 1.5623, "step": 49510 }, { "epoch": 2.68, "learning_rate": 3.6660175267770205e-05, "loss": 1.6883, "step": 49520 }, { "epoch": 2.68, "learning_rate": 3.6657470518230014e-05, "loss": 1.6254, "step": 49530 }, { "epoch": 2.68, "learning_rate": 3.665476576868982e-05, "loss": 1.5581, "step": 49540 }, { "epoch": 2.68, "learning_rate": 3.6652061019149634e-05, "loss": 1.6188, "step": 49550 }, { "epoch": 2.68, "learning_rate": 3.664935626960944e-05, "loss": 1.6219, "step": 49560 }, { "epoch": 2.68, "learning_rate": 3.664665152006924e-05, "loss": 1.6274, "step": 49570 }, { "epoch": 2.68, "learning_rate": 3.6643946770529056e-05, "loss": 1.6183, "step": 49580 }, { "epoch": 2.68, "learning_rate": 3.664124202098886e-05, "loss": 1.6481, "step": 49590 }, { "epoch": 2.68, "learning_rate": 3.663853727144866e-05, "loss": 1.6622, "step": 49600 }, { "epoch": 2.68, "learning_rate": 3.663583252190848e-05, "loss": 1.6208, "step": 49610 }, { "epoch": 2.68, "learning_rate": 3.663312777236828e-05, "loss": 1.6119, "step": 49620 }, { "epoch": 2.68, "learning_rate": 3.6630423022828084e-05, "loss": 1.6396, "step": 49630 }, { "epoch": 2.68, "learning_rate": 3.66277182732879e-05, "loss": 1.7014, "step": 49640 }, { "epoch": 2.68, "learning_rate": 3.66250135237477e-05, "loss": 1.6581, "step": 49650 }, { "epoch": 2.68, "learning_rate": 3.6622308774207506e-05, "loss": 1.5779, "step": 49660 }, { "epoch": 2.68, "learning_rate": 3.661960402466732e-05, "loss": 1.6157, "step": 49670 }, { "epoch": 2.68, "learning_rate": 3.6616899275127125e-05, "loss": 1.6629, "step": 49680 }, { "epoch": 2.68, "learning_rate": 3.661419452558693e-05, "loss": 1.6556, "step": 49690 }, { "epoch": 2.69, "learning_rate": 3.6611489776046745e-05, "loss": 1.6325, "step": 49700 }, { "epoch": 2.69, "learning_rate": 3.660878502650655e-05, "loss": 1.6003, "step": 49710 }, { "epoch": 2.69, "learning_rate": 3.660608027696635e-05, "loss": 1.5628, "step": 49720 }, { "epoch": 2.69, "learning_rate": 3.660337552742617e-05, "loss": 1.6706, "step": 49730 }, { "epoch": 2.69, "learning_rate": 3.660067077788597e-05, "loss": 1.7102, "step": 49740 }, { "epoch": 2.69, "learning_rate": 3.659796602834578e-05, "loss": 1.6472, "step": 49750 }, { "epoch": 2.69, "learning_rate": 3.659526127880559e-05, "loss": 1.5837, "step": 49760 }, { "epoch": 2.69, "learning_rate": 3.659255652926539e-05, "loss": 1.5598, "step": 49770 }, { "epoch": 2.69, "learning_rate": 3.65898517797252e-05, "loss": 1.5386, "step": 49780 }, { "epoch": 2.69, "learning_rate": 3.6587147030185004e-05, "loss": 1.5962, "step": 49790 }, { "epoch": 2.69, "learning_rate": 3.6584442280644814e-05, "loss": 1.5713, "step": 49800 }, { "epoch": 2.69, "learning_rate": 3.6581737531104624e-05, "loss": 1.5594, "step": 49810 }, { "epoch": 2.69, "learning_rate": 3.6579032781564426e-05, "loss": 1.7019, "step": 49820 }, { "epoch": 2.69, "learning_rate": 3.6576328032024236e-05, "loss": 1.5745, "step": 49830 }, { "epoch": 2.69, "learning_rate": 3.6573623282484046e-05, "loss": 1.6832, "step": 49840 }, { "epoch": 2.69, "learning_rate": 3.657091853294385e-05, "loss": 1.6012, "step": 49850 }, { "epoch": 2.69, "learning_rate": 3.656821378340366e-05, "loss": 1.6218, "step": 49860 }, { "epoch": 2.69, "learning_rate": 3.656550903386347e-05, "loss": 1.6716, "step": 49870 }, { "epoch": 2.7, "learning_rate": 3.656280428432327e-05, "loss": 1.6113, "step": 49880 }, { "epoch": 2.7, "learning_rate": 3.656009953478308e-05, "loss": 1.5244, "step": 49890 }, { "epoch": 2.7, "learning_rate": 3.655739478524289e-05, "loss": 1.6269, "step": 49900 }, { "epoch": 2.7, "learning_rate": 3.655469003570269e-05, "loss": 1.5092, "step": 49910 }, { "epoch": 2.7, "learning_rate": 3.65519852861625e-05, "loss": 1.657, "step": 49920 }, { "epoch": 2.7, "learning_rate": 3.654928053662231e-05, "loss": 1.6076, "step": 49930 }, { "epoch": 2.7, "learning_rate": 3.6546575787082115e-05, "loss": 1.6393, "step": 49940 }, { "epoch": 2.7, "learning_rate": 3.6543871037541925e-05, "loss": 1.6169, "step": 49950 }, { "epoch": 2.7, "learning_rate": 3.6541166288001734e-05, "loss": 1.5899, "step": 49960 }, { "epoch": 2.7, "learning_rate": 3.653846153846154e-05, "loss": 1.5593, "step": 49970 }, { "epoch": 2.7, "learning_rate": 3.653575678892135e-05, "loss": 1.6489, "step": 49980 }, { "epoch": 2.7, "learning_rate": 3.653305203938116e-05, "loss": 1.6465, "step": 49990 }, { "epoch": 2.7, "learning_rate": 3.653034728984096e-05, "loss": 1.5997, "step": 50000 }, { "epoch": 2.7, "learning_rate": 3.652764254030077e-05, "loss": 1.5631, "step": 50010 }, { "epoch": 2.7, "learning_rate": 3.652493779076058e-05, "loss": 1.6117, "step": 50020 }, { "epoch": 2.7, "learning_rate": 3.652223304122038e-05, "loss": 1.5998, "step": 50030 }, { "epoch": 2.7, "learning_rate": 3.651952829168019e-05, "loss": 1.6767, "step": 50040 }, { "epoch": 2.7, "learning_rate": 3.651682354214e-05, "loss": 1.6398, "step": 50050 }, { "epoch": 2.7, "learning_rate": 3.6514118792599804e-05, "loss": 1.6264, "step": 50060 }, { "epoch": 2.71, "learning_rate": 3.6511414043059614e-05, "loss": 1.6636, "step": 50070 }, { "epoch": 2.71, "learning_rate": 3.650870929351942e-05, "loss": 1.6387, "step": 50080 }, { "epoch": 2.71, "learning_rate": 3.6506004543979226e-05, "loss": 1.5765, "step": 50090 }, { "epoch": 2.71, "learning_rate": 3.6503299794439036e-05, "loss": 1.4963, "step": 50100 }, { "epoch": 2.71, "learning_rate": 3.6500595044898845e-05, "loss": 1.6802, "step": 50110 }, { "epoch": 2.71, "learning_rate": 3.649789029535865e-05, "loss": 1.6516, "step": 50120 }, { "epoch": 2.71, "learning_rate": 3.649518554581846e-05, "loss": 1.6515, "step": 50130 }, { "epoch": 2.71, "learning_rate": 3.649248079627827e-05, "loss": 1.5801, "step": 50140 }, { "epoch": 2.71, "learning_rate": 3.648977604673807e-05, "loss": 1.6223, "step": 50150 }, { "epoch": 2.71, "learning_rate": 3.648707129719789e-05, "loss": 1.5725, "step": 50160 }, { "epoch": 2.71, "learning_rate": 3.648436654765769e-05, "loss": 1.584, "step": 50170 }, { "epoch": 2.71, "learning_rate": 3.648166179811749e-05, "loss": 1.6539, "step": 50180 }, { "epoch": 2.71, "learning_rate": 3.647895704857731e-05, "loss": 1.6528, "step": 50190 }, { "epoch": 2.71, "learning_rate": 3.647625229903711e-05, "loss": 1.5483, "step": 50200 }, { "epoch": 2.71, "learning_rate": 3.6473547549496915e-05, "loss": 1.5829, "step": 50210 }, { "epoch": 2.71, "learning_rate": 3.647084279995673e-05, "loss": 1.635, "step": 50220 }, { "epoch": 2.71, "learning_rate": 3.6468138050416534e-05, "loss": 1.5894, "step": 50230 }, { "epoch": 2.71, "learning_rate": 3.646543330087634e-05, "loss": 1.6817, "step": 50240 }, { "epoch": 2.72, "learning_rate": 3.646272855133615e-05, "loss": 1.6065, "step": 50250 }, { "epoch": 2.72, "learning_rate": 3.6460023801795956e-05, "loss": 1.564, "step": 50260 }, { "epoch": 2.72, "learning_rate": 3.645731905225576e-05, "loss": 1.6198, "step": 50270 }, { "epoch": 2.72, "learning_rate": 3.6454614302715575e-05, "loss": 1.5431, "step": 50280 }, { "epoch": 2.72, "learning_rate": 3.645190955317538e-05, "loss": 1.5824, "step": 50290 }, { "epoch": 2.72, "learning_rate": 3.644920480363518e-05, "loss": 1.6119, "step": 50300 }, { "epoch": 2.72, "learning_rate": 3.6446500054095e-05, "loss": 1.7201, "step": 50310 }, { "epoch": 2.72, "learning_rate": 3.64437953045548e-05, "loss": 1.6011, "step": 50320 }, { "epoch": 2.72, "learning_rate": 3.64410905550146e-05, "loss": 1.6246, "step": 50330 }, { "epoch": 2.72, "learning_rate": 3.643838580547441e-05, "loss": 1.5817, "step": 50340 }, { "epoch": 2.72, "learning_rate": 3.643568105593422e-05, "loss": 1.5946, "step": 50350 }, { "epoch": 2.72, "learning_rate": 3.6432976306394026e-05, "loss": 1.6539, "step": 50360 }, { "epoch": 2.72, "learning_rate": 3.6430271556853835e-05, "loss": 1.6555, "step": 50370 }, { "epoch": 2.72, "learning_rate": 3.6427566807313645e-05, "loss": 1.6382, "step": 50380 }, { "epoch": 2.72, "learning_rate": 3.6424862057773454e-05, "loss": 1.5981, "step": 50390 }, { "epoch": 2.72, "learning_rate": 3.642215730823326e-05, "loss": 1.6379, "step": 50400 }, { "epoch": 2.72, "learning_rate": 3.641945255869307e-05, "loss": 1.5436, "step": 50410 }, { "epoch": 2.72, "learning_rate": 3.641674780915288e-05, "loss": 1.6755, "step": 50420 }, { "epoch": 2.72, "learning_rate": 3.641404305961268e-05, "loss": 1.6664, "step": 50430 }, { "epoch": 2.73, "learning_rate": 3.641133831007249e-05, "loss": 1.7419, "step": 50440 }, { "epoch": 2.73, "learning_rate": 3.64086335605323e-05, "loss": 1.6685, "step": 50450 }, { "epoch": 2.73, "learning_rate": 3.64059288109921e-05, "loss": 1.5929, "step": 50460 }, { "epoch": 2.73, "learning_rate": 3.640322406145191e-05, "loss": 1.6224, "step": 50470 }, { "epoch": 2.73, "learning_rate": 3.640051931191172e-05, "loss": 1.6699, "step": 50480 }, { "epoch": 2.73, "learning_rate": 3.6397814562371524e-05, "loss": 1.5867, "step": 50490 }, { "epoch": 2.73, "learning_rate": 3.6395109812831334e-05, "loss": 1.491, "step": 50500 }, { "epoch": 2.73, "learning_rate": 3.639240506329114e-05, "loss": 1.6182, "step": 50510 }, { "epoch": 2.73, "learning_rate": 3.6389700313750946e-05, "loss": 1.6048, "step": 50520 }, { "epoch": 2.73, "learning_rate": 3.6386995564210756e-05, "loss": 1.661, "step": 50530 }, { "epoch": 2.73, "learning_rate": 3.6384290814670565e-05, "loss": 1.5954, "step": 50540 }, { "epoch": 2.73, "learning_rate": 3.638158606513037e-05, "loss": 1.5407, "step": 50550 }, { "epoch": 2.73, "learning_rate": 3.637888131559018e-05, "loss": 1.5156, "step": 50560 }, { "epoch": 2.73, "learning_rate": 3.637617656604999e-05, "loss": 1.6959, "step": 50570 }, { "epoch": 2.73, "learning_rate": 3.637347181650979e-05, "loss": 1.5696, "step": 50580 }, { "epoch": 2.73, "learning_rate": 3.63707670669696e-05, "loss": 1.616, "step": 50590 }, { "epoch": 2.73, "learning_rate": 3.636806231742941e-05, "loss": 1.5713, "step": 50600 }, { "epoch": 2.73, "learning_rate": 3.636535756788921e-05, "loss": 1.6293, "step": 50610 }, { "epoch": 2.74, "learning_rate": 3.636265281834902e-05, "loss": 1.572, "step": 50620 }, { "epoch": 2.74, "learning_rate": 3.635994806880883e-05, "loss": 1.5246, "step": 50630 }, { "epoch": 2.74, "learning_rate": 3.6357243319268635e-05, "loss": 1.6566, "step": 50640 }, { "epoch": 2.74, "learning_rate": 3.6354538569728444e-05, "loss": 1.6156, "step": 50650 }, { "epoch": 2.74, "learning_rate": 3.6351833820188254e-05, "loss": 1.5874, "step": 50660 }, { "epoch": 2.74, "learning_rate": 3.634912907064806e-05, "loss": 1.6368, "step": 50670 }, { "epoch": 2.74, "learning_rate": 3.6346424321107867e-05, "loss": 1.5998, "step": 50680 }, { "epoch": 2.74, "learning_rate": 3.6343719571567676e-05, "loss": 1.6567, "step": 50690 }, { "epoch": 2.74, "learning_rate": 3.634101482202748e-05, "loss": 1.642, "step": 50700 }, { "epoch": 2.74, "learning_rate": 3.633831007248729e-05, "loss": 1.5672, "step": 50710 }, { "epoch": 2.74, "learning_rate": 3.63356053229471e-05, "loss": 1.5197, "step": 50720 }, { "epoch": 2.74, "learning_rate": 3.63329005734069e-05, "loss": 1.5652, "step": 50730 }, { "epoch": 2.74, "learning_rate": 3.633019582386671e-05, "loss": 1.6376, "step": 50740 }, { "epoch": 2.74, "learning_rate": 3.632749107432652e-05, "loss": 1.6326, "step": 50750 }, { "epoch": 2.74, "learning_rate": 3.6324786324786323e-05, "loss": 1.6043, "step": 50760 }, { "epoch": 2.74, "learning_rate": 3.632208157524613e-05, "loss": 1.6723, "step": 50770 }, { "epoch": 2.74, "learning_rate": 3.631937682570594e-05, "loss": 1.5752, "step": 50780 }, { "epoch": 2.74, "learning_rate": 3.6316672076165746e-05, "loss": 1.705, "step": 50790 }, { "epoch": 2.74, "learning_rate": 3.631396732662556e-05, "loss": 1.6212, "step": 50800 }, { "epoch": 2.75, "learning_rate": 3.6311262577085365e-05, "loss": 1.4982, "step": 50810 }, { "epoch": 2.75, "learning_rate": 3.630855782754517e-05, "loss": 1.6439, "step": 50820 }, { "epoch": 2.75, "learning_rate": 3.6305853078004984e-05, "loss": 1.6169, "step": 50830 }, { "epoch": 2.75, "learning_rate": 3.630314832846479e-05, "loss": 1.6202, "step": 50840 }, { "epoch": 2.75, "learning_rate": 3.630044357892459e-05, "loss": 1.5694, "step": 50850 }, { "epoch": 2.75, "learning_rate": 3.6297738829384406e-05, "loss": 1.684, "step": 50860 }, { "epoch": 2.75, "learning_rate": 3.629503407984421e-05, "loss": 1.7078, "step": 50870 }, { "epoch": 2.75, "learning_rate": 3.629232933030401e-05, "loss": 1.6065, "step": 50880 }, { "epoch": 2.75, "learning_rate": 3.628962458076383e-05, "loss": 1.6171, "step": 50890 }, { "epoch": 2.75, "learning_rate": 3.628691983122363e-05, "loss": 1.6443, "step": 50900 }, { "epoch": 2.75, "learning_rate": 3.6284215081683434e-05, "loss": 1.6102, "step": 50910 }, { "epoch": 2.75, "learning_rate": 3.6281510332143244e-05, "loss": 1.61, "step": 50920 }, { "epoch": 2.75, "learning_rate": 3.6278805582603054e-05, "loss": 1.5561, "step": 50930 }, { "epoch": 2.75, "learning_rate": 3.6276100833062856e-05, "loss": 1.5788, "step": 50940 }, { "epoch": 2.75, "learning_rate": 3.6273396083522666e-05, "loss": 1.6547, "step": 50950 }, { "epoch": 2.75, "learning_rate": 3.6270691333982476e-05, "loss": 1.5671, "step": 50960 }, { "epoch": 2.75, "learning_rate": 3.626798658444228e-05, "loss": 1.6545, "step": 50970 }, { "epoch": 2.75, "learning_rate": 3.626528183490209e-05, "loss": 1.6256, "step": 50980 }, { "epoch": 2.76, "learning_rate": 3.62625770853619e-05, "loss": 1.5788, "step": 50990 }, { "epoch": 2.76, "learning_rate": 3.62598723358217e-05, "loss": 1.5899, "step": 51000 }, { "epoch": 2.76, "learning_rate": 3.625716758628151e-05, "loss": 1.7055, "step": 51010 }, { "epoch": 2.76, "learning_rate": 3.625446283674132e-05, "loss": 1.6594, "step": 51020 }, { "epoch": 2.76, "learning_rate": 3.625175808720113e-05, "loss": 1.6203, "step": 51030 }, { "epoch": 2.76, "learning_rate": 3.624905333766093e-05, "loss": 1.6679, "step": 51040 }, { "epoch": 2.76, "learning_rate": 3.624634858812074e-05, "loss": 1.5861, "step": 51050 }, { "epoch": 2.76, "learning_rate": 3.624364383858055e-05, "loss": 1.5859, "step": 51060 }, { "epoch": 2.76, "learning_rate": 3.6240939089040355e-05, "loss": 1.5829, "step": 51070 }, { "epoch": 2.76, "learning_rate": 3.6238234339500164e-05, "loss": 1.6098, "step": 51080 }, { "epoch": 2.76, "learning_rate": 3.6235529589959974e-05, "loss": 1.5597, "step": 51090 }, { "epoch": 2.76, "learning_rate": 3.623282484041978e-05, "loss": 1.6094, "step": 51100 }, { "epoch": 2.76, "learning_rate": 3.6230120090879587e-05, "loss": 1.5708, "step": 51110 }, { "epoch": 2.76, "learning_rate": 3.6227415341339396e-05, "loss": 1.5849, "step": 51120 }, { "epoch": 2.76, "learning_rate": 3.62247105917992e-05, "loss": 1.7096, "step": 51130 }, { "epoch": 2.76, "learning_rate": 3.622200584225901e-05, "loss": 1.624, "step": 51140 }, { "epoch": 2.76, "learning_rate": 3.621930109271882e-05, "loss": 1.6544, "step": 51150 }, { "epoch": 2.76, "learning_rate": 3.621659634317862e-05, "loss": 1.5201, "step": 51160 }, { "epoch": 2.76, "learning_rate": 3.621389159363843e-05, "loss": 1.6818, "step": 51170 }, { "epoch": 2.77, "learning_rate": 3.621118684409824e-05, "loss": 1.5894, "step": 51180 }, { "epoch": 2.77, "learning_rate": 3.6208482094558043e-05, "loss": 1.6579, "step": 51190 }, { "epoch": 2.77, "learning_rate": 3.620577734501785e-05, "loss": 1.6689, "step": 51200 }, { "epoch": 2.77, "learning_rate": 3.620307259547766e-05, "loss": 1.6251, "step": 51210 }, { "epoch": 2.77, "learning_rate": 3.6200367845937466e-05, "loss": 1.6207, "step": 51220 }, { "epoch": 2.77, "learning_rate": 3.6197663096397275e-05, "loss": 1.6301, "step": 51230 }, { "epoch": 2.77, "learning_rate": 3.6194958346857085e-05, "loss": 1.6308, "step": 51240 }, { "epoch": 2.77, "learning_rate": 3.619225359731689e-05, "loss": 1.6283, "step": 51250 }, { "epoch": 2.77, "learning_rate": 3.61895488477767e-05, "loss": 1.6047, "step": 51260 }, { "epoch": 2.77, "learning_rate": 3.618684409823651e-05, "loss": 1.6308, "step": 51270 }, { "epoch": 2.77, "learning_rate": 3.618413934869631e-05, "loss": 1.7249, "step": 51280 }, { "epoch": 2.77, "learning_rate": 3.618143459915612e-05, "loss": 1.5976, "step": 51290 }, { "epoch": 2.77, "learning_rate": 3.617872984961593e-05, "loss": 1.6488, "step": 51300 }, { "epoch": 2.77, "learning_rate": 3.617602510007573e-05, "loss": 1.5816, "step": 51310 }, { "epoch": 2.77, "learning_rate": 3.617332035053554e-05, "loss": 1.5389, "step": 51320 }, { "epoch": 2.77, "learning_rate": 3.617061560099535e-05, "loss": 1.5568, "step": 51330 }, { "epoch": 2.77, "learning_rate": 3.6167910851455154e-05, "loss": 1.6321, "step": 51340 }, { "epoch": 2.77, "learning_rate": 3.6165206101914964e-05, "loss": 1.6185, "step": 51350 }, { "epoch": 2.78, "learning_rate": 3.6162501352374774e-05, "loss": 1.6131, "step": 51360 }, { "epoch": 2.78, "learning_rate": 3.6159796602834576e-05, "loss": 1.6334, "step": 51370 }, { "epoch": 2.78, "learning_rate": 3.6157091853294386e-05, "loss": 1.5993, "step": 51380 }, { "epoch": 2.78, "learning_rate": 3.6154387103754196e-05, "loss": 1.6102, "step": 51390 }, { "epoch": 2.78, "learning_rate": 3.6151682354214e-05, "loss": 1.6304, "step": 51400 }, { "epoch": 2.78, "learning_rate": 3.614897760467381e-05, "loss": 1.5935, "step": 51410 }, { "epoch": 2.78, "learning_rate": 3.614627285513362e-05, "loss": 1.6291, "step": 51420 }, { "epoch": 2.78, "learning_rate": 3.614356810559342e-05, "loss": 1.6831, "step": 51430 }, { "epoch": 2.78, "learning_rate": 3.614086335605324e-05, "loss": 1.7077, "step": 51440 }, { "epoch": 2.78, "learning_rate": 3.613815860651304e-05, "loss": 1.6454, "step": 51450 }, { "epoch": 2.78, "learning_rate": 3.613545385697284e-05, "loss": 1.5975, "step": 51460 }, { "epoch": 2.78, "learning_rate": 3.613274910743265e-05, "loss": 1.6416, "step": 51470 }, { "epoch": 2.78, "learning_rate": 3.613004435789246e-05, "loss": 1.6466, "step": 51480 }, { "epoch": 2.78, "learning_rate": 3.6127339608352265e-05, "loss": 1.7198, "step": 51490 }, { "epoch": 2.78, "learning_rate": 3.6124634858812075e-05, "loss": 1.5497, "step": 51500 }, { "epoch": 2.78, "learning_rate": 3.6121930109271884e-05, "loss": 1.5881, "step": 51510 }, { "epoch": 2.78, "learning_rate": 3.611922535973169e-05, "loss": 1.6264, "step": 51520 }, { "epoch": 2.78, "learning_rate": 3.61165206101915e-05, "loss": 1.5806, "step": 51530 }, { "epoch": 2.78, "learning_rate": 3.6113815860651307e-05, "loss": 1.6832, "step": 51540 }, { "epoch": 2.79, "learning_rate": 3.611111111111111e-05, "loss": 1.5275, "step": 51550 }, { "epoch": 2.79, "learning_rate": 3.610840636157092e-05, "loss": 1.5353, "step": 51560 }, { "epoch": 2.79, "learning_rate": 3.610570161203073e-05, "loss": 1.5741, "step": 51570 }, { "epoch": 2.79, "learning_rate": 3.610299686249053e-05, "loss": 1.6716, "step": 51580 }, { "epoch": 2.79, "learning_rate": 3.610029211295034e-05, "loss": 1.637, "step": 51590 }, { "epoch": 2.79, "learning_rate": 3.609758736341015e-05, "loss": 1.5944, "step": 51600 }, { "epoch": 2.79, "learning_rate": 3.6094882613869954e-05, "loss": 1.6104, "step": 51610 }, { "epoch": 2.79, "learning_rate": 3.6092177864329763e-05, "loss": 1.66, "step": 51620 }, { "epoch": 2.79, "learning_rate": 3.608947311478957e-05, "loss": 1.6266, "step": 51630 }, { "epoch": 2.79, "learning_rate": 3.6086768365249376e-05, "loss": 1.6235, "step": 51640 }, { "epoch": 2.79, "learning_rate": 3.6084063615709186e-05, "loss": 1.6621, "step": 51650 }, { "epoch": 2.79, "learning_rate": 3.6081358866168995e-05, "loss": 1.6374, "step": 51660 }, { "epoch": 2.79, "learning_rate": 3.6078654116628805e-05, "loss": 1.6914, "step": 51670 }, { "epoch": 2.79, "learning_rate": 3.607594936708861e-05, "loss": 1.6675, "step": 51680 }, { "epoch": 2.79, "learning_rate": 3.607324461754842e-05, "loss": 1.6843, "step": 51690 }, { "epoch": 2.79, "learning_rate": 3.607053986800823e-05, "loss": 1.5108, "step": 51700 }, { "epoch": 2.79, "learning_rate": 3.606783511846803e-05, "loss": 1.6489, "step": 51710 }, { "epoch": 2.79, "learning_rate": 3.606513036892784e-05, "loss": 1.5757, "step": 51720 }, { "epoch": 2.8, "learning_rate": 3.606242561938765e-05, "loss": 1.6355, "step": 51730 }, { "epoch": 2.8, "learning_rate": 3.605972086984745e-05, "loss": 1.6414, "step": 51740 }, { "epoch": 2.8, "learning_rate": 3.605701612030726e-05, "loss": 1.6276, "step": 51750 }, { "epoch": 2.8, "learning_rate": 3.605431137076707e-05, "loss": 1.586, "step": 51760 }, { "epoch": 2.8, "learning_rate": 3.6051606621226874e-05, "loss": 1.6152, "step": 51770 }, { "epoch": 2.8, "learning_rate": 3.6048901871686684e-05, "loss": 1.6379, "step": 51780 }, { "epoch": 2.8, "learning_rate": 3.6046197122146494e-05, "loss": 1.6141, "step": 51790 }, { "epoch": 2.8, "learning_rate": 3.6043492372606296e-05, "loss": 1.6407, "step": 51800 }, { "epoch": 2.8, "learning_rate": 3.6040787623066106e-05, "loss": 1.6562, "step": 51810 }, { "epoch": 2.8, "learning_rate": 3.6038082873525916e-05, "loss": 1.6167, "step": 51820 }, { "epoch": 2.8, "learning_rate": 3.603537812398572e-05, "loss": 1.6433, "step": 51830 }, { "epoch": 2.8, "learning_rate": 3.603267337444553e-05, "loss": 1.6466, "step": 51840 }, { "epoch": 2.8, "learning_rate": 3.602996862490534e-05, "loss": 1.6544, "step": 51850 }, { "epoch": 2.8, "learning_rate": 3.602726387536514e-05, "loss": 1.6637, "step": 51860 }, { "epoch": 2.8, "learning_rate": 3.602455912582495e-05, "loss": 1.6682, "step": 51870 }, { "epoch": 2.8, "learning_rate": 3.602185437628476e-05, "loss": 1.6266, "step": 51880 }, { "epoch": 2.8, "learning_rate": 3.601914962674456e-05, "loss": 1.5783, "step": 51890 }, { "epoch": 2.8, "learning_rate": 3.601644487720437e-05, "loss": 1.6384, "step": 51900 }, { "epoch": 2.8, "learning_rate": 3.601374012766418e-05, "loss": 1.6231, "step": 51910 }, { "epoch": 2.81, "learning_rate": 3.6011035378123985e-05, "loss": 1.6464, "step": 51920 }, { "epoch": 2.81, "learning_rate": 3.6008330628583795e-05, "loss": 1.6682, "step": 51930 }, { "epoch": 2.81, "learning_rate": 3.6005625879043604e-05, "loss": 1.5417, "step": 51940 }, { "epoch": 2.81, "learning_rate": 3.600292112950341e-05, "loss": 1.6081, "step": 51950 }, { "epoch": 2.81, "learning_rate": 3.600021637996322e-05, "loss": 1.5765, "step": 51960 }, { "epoch": 2.81, "learning_rate": 3.5997511630423027e-05, "loss": 1.6268, "step": 51970 }, { "epoch": 2.81, "learning_rate": 3.599480688088283e-05, "loss": 1.5844, "step": 51980 }, { "epoch": 2.81, "learning_rate": 3.599210213134264e-05, "loss": 1.6173, "step": 51990 }, { "epoch": 2.81, "learning_rate": 3.598939738180245e-05, "loss": 1.635, "step": 52000 }, { "epoch": 2.81, "learning_rate": 3.598669263226225e-05, "loss": 1.6415, "step": 52010 }, { "epoch": 2.81, "learning_rate": 3.598398788272206e-05, "loss": 1.6224, "step": 52020 }, { "epoch": 2.81, "learning_rate": 3.598128313318187e-05, "loss": 1.6351, "step": 52030 }, { "epoch": 2.81, "learning_rate": 3.5978578383641674e-05, "loss": 1.6167, "step": 52040 }, { "epoch": 2.81, "learning_rate": 3.5975873634101483e-05, "loss": 1.6169, "step": 52050 }, { "epoch": 2.81, "learning_rate": 3.597316888456129e-05, "loss": 1.694, "step": 52060 }, { "epoch": 2.81, "learning_rate": 3.5970464135021096e-05, "loss": 1.6893, "step": 52070 }, { "epoch": 2.81, "learning_rate": 3.5967759385480906e-05, "loss": 1.7116, "step": 52080 }, { "epoch": 2.81, "learning_rate": 3.5965054635940715e-05, "loss": 1.6136, "step": 52090 }, { "epoch": 2.82, "learning_rate": 3.596234988640052e-05, "loss": 1.6054, "step": 52100 }, { "epoch": 2.82, "learning_rate": 3.595964513686033e-05, "loss": 1.6744, "step": 52110 }, { "epoch": 2.82, "learning_rate": 3.595694038732014e-05, "loss": 1.561, "step": 52120 }, { "epoch": 2.82, "learning_rate": 3.595423563777994e-05, "loss": 1.585, "step": 52130 }, { "epoch": 2.82, "learning_rate": 3.595153088823975e-05, "loss": 1.6343, "step": 52140 }, { "epoch": 2.82, "learning_rate": 3.594882613869956e-05, "loss": 1.5895, "step": 52150 }, { "epoch": 2.82, "learning_rate": 3.594612138915936e-05, "loss": 1.6225, "step": 52160 }, { "epoch": 2.82, "learning_rate": 3.594341663961917e-05, "loss": 1.5097, "step": 52170 }, { "epoch": 2.82, "learning_rate": 3.594071189007898e-05, "loss": 1.6441, "step": 52180 }, { "epoch": 2.82, "learning_rate": 3.5938007140538785e-05, "loss": 1.6556, "step": 52190 }, { "epoch": 2.82, "learning_rate": 3.5935302390998594e-05, "loss": 1.6175, "step": 52200 }, { "epoch": 2.82, "learning_rate": 3.5932597641458404e-05, "loss": 1.6517, "step": 52210 }, { "epoch": 2.82, "learning_rate": 3.592989289191821e-05, "loss": 1.6098, "step": 52220 }, { "epoch": 2.82, "learning_rate": 3.5927188142378016e-05, "loss": 1.6619, "step": 52230 }, { "epoch": 2.82, "learning_rate": 3.5924483392837826e-05, "loss": 1.5457, "step": 52240 }, { "epoch": 2.82, "learning_rate": 3.592177864329763e-05, "loss": 1.6455, "step": 52250 }, { "epoch": 2.82, "learning_rate": 3.591907389375744e-05, "loss": 1.5909, "step": 52260 }, { "epoch": 2.82, "learning_rate": 3.591636914421725e-05, "loss": 1.61, "step": 52270 }, { "epoch": 2.82, "learning_rate": 3.591366439467706e-05, "loss": 1.6405, "step": 52280 }, { "epoch": 2.83, "learning_rate": 3.591095964513686e-05, "loss": 1.7225, "step": 52290 }, { "epoch": 2.83, "learning_rate": 3.590825489559667e-05, "loss": 1.6776, "step": 52300 }, { "epoch": 2.83, "learning_rate": 3.590555014605648e-05, "loss": 1.6905, "step": 52310 }, { "epoch": 2.83, "learning_rate": 3.590284539651628e-05, "loss": 1.527, "step": 52320 }, { "epoch": 2.83, "learning_rate": 3.590014064697609e-05, "loss": 1.6625, "step": 52330 }, { "epoch": 2.83, "learning_rate": 3.58974358974359e-05, "loss": 1.5845, "step": 52340 }, { "epoch": 2.83, "learning_rate": 3.5894731147895705e-05, "loss": 1.5605, "step": 52350 }, { "epoch": 2.83, "learning_rate": 3.5892026398355515e-05, "loss": 1.5705, "step": 52360 }, { "epoch": 2.83, "learning_rate": 3.5889321648815324e-05, "loss": 1.6241, "step": 52370 }, { "epoch": 2.83, "learning_rate": 3.588661689927513e-05, "loss": 1.6371, "step": 52380 }, { "epoch": 2.83, "learning_rate": 3.588391214973494e-05, "loss": 1.5661, "step": 52390 }, { "epoch": 2.83, "learning_rate": 3.588120740019475e-05, "loss": 1.7196, "step": 52400 }, { "epoch": 2.83, "learning_rate": 3.587850265065455e-05, "loss": 1.6656, "step": 52410 }, { "epoch": 2.83, "learning_rate": 3.587579790111436e-05, "loss": 1.6004, "step": 52420 }, { "epoch": 2.83, "learning_rate": 3.587309315157417e-05, "loss": 1.4913, "step": 52430 }, { "epoch": 2.83, "learning_rate": 3.587038840203397e-05, "loss": 1.5637, "step": 52440 }, { "epoch": 2.83, "learning_rate": 3.586768365249378e-05, "loss": 1.5613, "step": 52450 }, { "epoch": 2.83, "learning_rate": 3.586497890295359e-05, "loss": 1.5827, "step": 52460 }, { "epoch": 2.84, "learning_rate": 3.5862274153413394e-05, "loss": 1.5954, "step": 52470 }, { "epoch": 2.84, "learning_rate": 3.5859569403873203e-05, "loss": 1.589, "step": 52480 }, { "epoch": 2.84, "learning_rate": 3.585686465433301e-05, "loss": 1.6492, "step": 52490 }, { "epoch": 2.84, "learning_rate": 3.5854159904792816e-05, "loss": 1.6515, "step": 52500 }, { "epoch": 2.84, "learning_rate": 3.5851455155252626e-05, "loss": 1.6406, "step": 52510 }, { "epoch": 2.84, "learning_rate": 3.5848750405712435e-05, "loss": 1.571, "step": 52520 }, { "epoch": 2.84, "learning_rate": 3.584604565617224e-05, "loss": 1.5586, "step": 52530 }, { "epoch": 2.84, "learning_rate": 3.584334090663205e-05, "loss": 1.6225, "step": 52540 }, { "epoch": 2.84, "learning_rate": 3.584063615709186e-05, "loss": 1.6142, "step": 52550 }, { "epoch": 2.84, "learning_rate": 3.583793140755166e-05, "loss": 1.6488, "step": 52560 }, { "epoch": 2.84, "learning_rate": 3.583522665801147e-05, "loss": 1.6154, "step": 52570 }, { "epoch": 2.84, "learning_rate": 3.583252190847128e-05, "loss": 1.6263, "step": 52580 }, { "epoch": 2.84, "learning_rate": 3.582981715893108e-05, "loss": 1.695, "step": 52590 }, { "epoch": 2.84, "learning_rate": 3.582711240939089e-05, "loss": 1.6261, "step": 52600 }, { "epoch": 2.84, "learning_rate": 3.58244076598507e-05, "loss": 1.6753, "step": 52610 }, { "epoch": 2.84, "learning_rate": 3.5821702910310505e-05, "loss": 1.6105, "step": 52620 }, { "epoch": 2.84, "learning_rate": 3.5818998160770314e-05, "loss": 1.6387, "step": 52630 }, { "epoch": 2.84, "learning_rate": 3.5816293411230124e-05, "loss": 1.5904, "step": 52640 }, { "epoch": 2.84, "learning_rate": 3.581358866168993e-05, "loss": 1.6373, "step": 52650 }, { "epoch": 2.85, "learning_rate": 3.5810883912149736e-05, "loss": 1.6431, "step": 52660 }, { "epoch": 2.85, "learning_rate": 3.5808179162609546e-05, "loss": 1.5435, "step": 52670 }, { "epoch": 2.85, "learning_rate": 3.580547441306935e-05, "loss": 1.6322, "step": 52680 }, { "epoch": 2.85, "learning_rate": 3.580276966352916e-05, "loss": 1.5671, "step": 52690 }, { "epoch": 2.85, "learning_rate": 3.580006491398897e-05, "loss": 1.6389, "step": 52700 }, { "epoch": 2.85, "learning_rate": 3.579736016444877e-05, "loss": 1.6383, "step": 52710 }, { "epoch": 2.85, "learning_rate": 3.579465541490858e-05, "loss": 1.633, "step": 52720 }, { "epoch": 2.85, "learning_rate": 3.579195066536839e-05, "loss": 1.7, "step": 52730 }, { "epoch": 2.85, "learning_rate": 3.578924591582819e-05, "loss": 1.5987, "step": 52740 }, { "epoch": 2.85, "learning_rate": 3.5786541166288e-05, "loss": 1.6593, "step": 52750 }, { "epoch": 2.85, "learning_rate": 3.578383641674781e-05, "loss": 1.7356, "step": 52760 }, { "epoch": 2.85, "learning_rate": 3.5781131667207616e-05, "loss": 1.5879, "step": 52770 }, { "epoch": 2.85, "learning_rate": 3.5778426917667425e-05, "loss": 1.6094, "step": 52780 }, { "epoch": 2.85, "learning_rate": 3.5775722168127235e-05, "loss": 1.6218, "step": 52790 }, { "epoch": 2.85, "learning_rate": 3.577301741858704e-05, "loss": 1.5705, "step": 52800 }, { "epoch": 2.85, "learning_rate": 3.577031266904685e-05, "loss": 1.6852, "step": 52810 }, { "epoch": 2.85, "learning_rate": 3.576760791950666e-05, "loss": 1.6561, "step": 52820 }, { "epoch": 2.85, "learning_rate": 3.576490316996646e-05, "loss": 1.5605, "step": 52830 }, { "epoch": 2.86, "learning_rate": 3.576219842042627e-05, "loss": 1.5646, "step": 52840 }, { "epoch": 2.86, "learning_rate": 3.575949367088608e-05, "loss": 1.6682, "step": 52850 }, { "epoch": 2.86, "learning_rate": 3.575678892134588e-05, "loss": 1.5839, "step": 52860 }, { "epoch": 2.86, "learning_rate": 3.575408417180569e-05, "loss": 1.582, "step": 52870 }, { "epoch": 2.86, "learning_rate": 3.57513794222655e-05, "loss": 1.6739, "step": 52880 }, { "epoch": 2.86, "learning_rate": 3.5748674672725304e-05, "loss": 1.67, "step": 52890 }, { "epoch": 2.86, "learning_rate": 3.5745969923185114e-05, "loss": 1.559, "step": 52900 }, { "epoch": 2.86, "learning_rate": 3.5743265173644924e-05, "loss": 1.6499, "step": 52910 }, { "epoch": 2.86, "learning_rate": 3.574056042410473e-05, "loss": 1.728, "step": 52920 }, { "epoch": 2.86, "learning_rate": 3.5737855674564536e-05, "loss": 1.6364, "step": 52930 }, { "epoch": 2.86, "learning_rate": 3.5735150925024346e-05, "loss": 1.6394, "step": 52940 }, { "epoch": 2.86, "learning_rate": 3.5732446175484155e-05, "loss": 1.5882, "step": 52950 }, { "epoch": 2.86, "learning_rate": 3.572974142594396e-05, "loss": 1.5645, "step": 52960 }, { "epoch": 2.86, "learning_rate": 3.572703667640377e-05, "loss": 1.6224, "step": 52970 }, { "epoch": 2.86, "learning_rate": 3.572433192686358e-05, "loss": 1.6694, "step": 52980 }, { "epoch": 2.86, "learning_rate": 3.572162717732338e-05, "loss": 1.6204, "step": 52990 }, { "epoch": 2.86, "learning_rate": 3.571892242778319e-05, "loss": 1.7153, "step": 53000 }, { "epoch": 2.86, "learning_rate": 3.5716217678243e-05, "loss": 1.5594, "step": 53010 }, { "epoch": 2.86, "learning_rate": 3.57135129287028e-05, "loss": 1.6259, "step": 53020 }, { "epoch": 2.87, "learning_rate": 3.571080817916261e-05, "loss": 1.6318, "step": 53030 }, { "epoch": 2.87, "learning_rate": 3.570810342962242e-05, "loss": 1.6894, "step": 53040 }, { "epoch": 2.87, "learning_rate": 3.5705398680082225e-05, "loss": 1.6977, "step": 53050 }, { "epoch": 2.87, "learning_rate": 3.5702693930542034e-05, "loss": 1.6054, "step": 53060 }, { "epoch": 2.87, "learning_rate": 3.5699989181001844e-05, "loss": 1.5755, "step": 53070 }, { "epoch": 2.87, "learning_rate": 3.569728443146165e-05, "loss": 1.6405, "step": 53080 }, { "epoch": 2.87, "learning_rate": 3.5694579681921457e-05, "loss": 1.6487, "step": 53090 }, { "epoch": 2.87, "learning_rate": 3.5691874932381266e-05, "loss": 1.5837, "step": 53100 }, { "epoch": 2.87, "learning_rate": 3.568917018284107e-05, "loss": 1.5619, "step": 53110 }, { "epoch": 2.87, "learning_rate": 3.568646543330088e-05, "loss": 1.5319, "step": 53120 }, { "epoch": 2.87, "learning_rate": 3.568376068376069e-05, "loss": 1.6462, "step": 53130 }, { "epoch": 2.87, "learning_rate": 3.568105593422049e-05, "loss": 1.5482, "step": 53140 }, { "epoch": 2.87, "learning_rate": 3.56783511846803e-05, "loss": 1.5304, "step": 53150 }, { "epoch": 2.87, "learning_rate": 3.567564643514011e-05, "loss": 1.541, "step": 53160 }, { "epoch": 2.87, "learning_rate": 3.567294168559991e-05, "loss": 1.7015, "step": 53170 }, { "epoch": 2.87, "learning_rate": 3.567023693605972e-05, "loss": 1.6582, "step": 53180 }, { "epoch": 2.87, "learning_rate": 3.566753218651953e-05, "loss": 1.6278, "step": 53190 }, { "epoch": 2.87, "learning_rate": 3.5664827436979336e-05, "loss": 1.5543, "step": 53200 }, { "epoch": 2.88, "learning_rate": 3.5662122687439145e-05, "loss": 1.6242, "step": 53210 }, { "epoch": 2.88, "learning_rate": 3.5659417937898955e-05, "loss": 1.6686, "step": 53220 }, { "epoch": 2.88, "learning_rate": 3.565671318835876e-05, "loss": 1.6253, "step": 53230 }, { "epoch": 2.88, "learning_rate": 3.565400843881857e-05, "loss": 1.561, "step": 53240 }, { "epoch": 2.88, "learning_rate": 3.565130368927838e-05, "loss": 1.574, "step": 53250 }, { "epoch": 2.88, "learning_rate": 3.564859893973818e-05, "loss": 1.603, "step": 53260 }, { "epoch": 2.88, "learning_rate": 3.564589419019799e-05, "loss": 1.6924, "step": 53270 }, { "epoch": 2.88, "learning_rate": 3.56431894406578e-05, "loss": 1.5834, "step": 53280 }, { "epoch": 2.88, "learning_rate": 3.56404846911176e-05, "loss": 1.5671, "step": 53290 }, { "epoch": 2.88, "learning_rate": 3.563777994157741e-05, "loss": 1.7144, "step": 53300 }, { "epoch": 2.88, "learning_rate": 3.563507519203722e-05, "loss": 1.5609, "step": 53310 }, { "epoch": 2.88, "learning_rate": 3.5632370442497024e-05, "loss": 1.6225, "step": 53320 }, { "epoch": 2.88, "learning_rate": 3.5629665692956834e-05, "loss": 1.5828, "step": 53330 }, { "epoch": 2.88, "learning_rate": 3.5626960943416644e-05, "loss": 1.65, "step": 53340 }, { "epoch": 2.88, "learning_rate": 3.5624256193876446e-05, "loss": 1.6099, "step": 53350 }, { "epoch": 2.88, "learning_rate": 3.5621551444336256e-05, "loss": 1.6207, "step": 53360 }, { "epoch": 2.88, "learning_rate": 3.5618846694796066e-05, "loss": 1.6119, "step": 53370 }, { "epoch": 2.88, "learning_rate": 3.561614194525587e-05, "loss": 1.6206, "step": 53380 }, { "epoch": 2.88, "learning_rate": 3.561343719571568e-05, "loss": 1.6878, "step": 53390 }, { "epoch": 2.89, "learning_rate": 3.561073244617549e-05, "loss": 1.6355, "step": 53400 }, { "epoch": 2.89, "learning_rate": 3.560802769663529e-05, "loss": 1.6161, "step": 53410 }, { "epoch": 2.89, "learning_rate": 3.56053229470951e-05, "loss": 1.5251, "step": 53420 }, { "epoch": 2.89, "learning_rate": 3.560261819755491e-05, "loss": 1.6422, "step": 53430 }, { "epoch": 2.89, "learning_rate": 3.559991344801471e-05, "loss": 1.5169, "step": 53440 }, { "epoch": 2.89, "learning_rate": 3.559720869847452e-05, "loss": 1.5836, "step": 53450 }, { "epoch": 2.89, "learning_rate": 3.559450394893433e-05, "loss": 1.6148, "step": 53460 }, { "epoch": 2.89, "learning_rate": 3.5591799199394135e-05, "loss": 1.5862, "step": 53470 }, { "epoch": 2.89, "learning_rate": 3.5589094449853945e-05, "loss": 1.7251, "step": 53480 }, { "epoch": 2.89, "learning_rate": 3.5586389700313754e-05, "loss": 1.727, "step": 53490 }, { "epoch": 2.89, "learning_rate": 3.558368495077356e-05, "loss": 1.5355, "step": 53500 }, { "epoch": 2.89, "learning_rate": 3.558098020123337e-05, "loss": 1.5644, "step": 53510 }, { "epoch": 2.89, "learning_rate": 3.5578275451693177e-05, "loss": 1.6991, "step": 53520 }, { "epoch": 2.89, "learning_rate": 3.557557070215298e-05, "loss": 1.6095, "step": 53530 }, { "epoch": 2.89, "learning_rate": 3.557286595261279e-05, "loss": 1.6116, "step": 53540 }, { "epoch": 2.89, "learning_rate": 3.55701612030726e-05, "loss": 1.7004, "step": 53550 }, { "epoch": 2.89, "learning_rate": 3.556745645353241e-05, "loss": 1.6605, "step": 53560 }, { "epoch": 2.89, "learning_rate": 3.556475170399221e-05, "loss": 1.5883, "step": 53570 }, { "epoch": 2.9, "learning_rate": 3.556204695445202e-05, "loss": 1.5288, "step": 53580 }, { "epoch": 2.9, "learning_rate": 3.555934220491183e-05, "loss": 1.6039, "step": 53590 }, { "epoch": 2.9, "learning_rate": 3.5556637455371633e-05, "loss": 1.6172, "step": 53600 }, { "epoch": 2.9, "learning_rate": 3.555393270583144e-05, "loss": 1.5946, "step": 53610 }, { "epoch": 2.9, "learning_rate": 3.555122795629125e-05, "loss": 1.6054, "step": 53620 }, { "epoch": 2.9, "learning_rate": 3.5548523206751056e-05, "loss": 1.6252, "step": 53630 }, { "epoch": 2.9, "learning_rate": 3.5545818457210865e-05, "loss": 1.6223, "step": 53640 }, { "epoch": 2.9, "learning_rate": 3.5543113707670675e-05, "loss": 1.5318, "step": 53650 }, { "epoch": 2.9, "learning_rate": 3.554040895813048e-05, "loss": 1.7142, "step": 53660 }, { "epoch": 2.9, "learning_rate": 3.553770420859029e-05, "loss": 1.5937, "step": 53670 }, { "epoch": 2.9, "learning_rate": 3.55349994590501e-05, "loss": 1.5748, "step": 53680 }, { "epoch": 2.9, "learning_rate": 3.55322947095099e-05, "loss": 1.5949, "step": 53690 }, { "epoch": 2.9, "learning_rate": 3.55295899599697e-05, "loss": 1.5414, "step": 53700 }, { "epoch": 2.9, "learning_rate": 3.552688521042952e-05, "loss": 1.5789, "step": 53710 }, { "epoch": 2.9, "learning_rate": 3.552418046088932e-05, "loss": 1.5838, "step": 53720 }, { "epoch": 2.9, "learning_rate": 3.5521475711349125e-05, "loss": 1.6241, "step": 53730 }, { "epoch": 2.9, "learning_rate": 3.551877096180894e-05, "loss": 1.6003, "step": 53740 }, { "epoch": 2.9, "learning_rate": 3.5516066212268744e-05, "loss": 1.5718, "step": 53750 }, { "epoch": 2.9, "learning_rate": 3.551336146272855e-05, "loss": 1.5456, "step": 53760 }, { "epoch": 2.91, "learning_rate": 3.5510656713188364e-05, "loss": 1.6003, "step": 53770 }, { "epoch": 2.91, "learning_rate": 3.5507951963648166e-05, "loss": 1.646, "step": 53780 }, { "epoch": 2.91, "learning_rate": 3.5505247214107976e-05, "loss": 1.6628, "step": 53790 }, { "epoch": 2.91, "learning_rate": 3.5502542464567786e-05, "loss": 1.6253, "step": 53800 }, { "epoch": 2.91, "learning_rate": 3.549983771502759e-05, "loss": 1.5519, "step": 53810 }, { "epoch": 2.91, "learning_rate": 3.54971329654874e-05, "loss": 1.6583, "step": 53820 }, { "epoch": 2.91, "learning_rate": 3.549442821594721e-05, "loss": 1.5568, "step": 53830 }, { "epoch": 2.91, "learning_rate": 3.549172346640701e-05, "loss": 1.5738, "step": 53840 }, { "epoch": 2.91, "learning_rate": 3.548901871686682e-05, "loss": 1.5403, "step": 53850 }, { "epoch": 2.91, "learning_rate": 3.548631396732663e-05, "loss": 1.6507, "step": 53860 }, { "epoch": 2.91, "learning_rate": 3.548360921778643e-05, "loss": 1.5747, "step": 53870 }, { "epoch": 2.91, "learning_rate": 3.548090446824624e-05, "loss": 1.5687, "step": 53880 }, { "epoch": 2.91, "learning_rate": 3.547819971870605e-05, "loss": 1.6351, "step": 53890 }, { "epoch": 2.91, "learning_rate": 3.5475494969165855e-05, "loss": 1.6475, "step": 53900 }, { "epoch": 2.91, "learning_rate": 3.5472790219625665e-05, "loss": 1.6552, "step": 53910 }, { "epoch": 2.91, "learning_rate": 3.5470085470085474e-05, "loss": 1.6506, "step": 53920 }, { "epoch": 2.91, "learning_rate": 3.546738072054528e-05, "loss": 1.6307, "step": 53930 }, { "epoch": 2.91, "learning_rate": 3.546467597100509e-05, "loss": 1.6952, "step": 53940 }, { "epoch": 2.92, "learning_rate": 3.5461971221464897e-05, "loss": 1.6198, "step": 53950 }, { "epoch": 2.92, "learning_rate": 3.54592664719247e-05, "loss": 1.7108, "step": 53960 }, { "epoch": 2.92, "learning_rate": 3.545656172238451e-05, "loss": 1.5907, "step": 53970 }, { "epoch": 2.92, "learning_rate": 3.545385697284432e-05, "loss": 1.7114, "step": 53980 }, { "epoch": 2.92, "learning_rate": 3.545115222330412e-05, "loss": 1.6086, "step": 53990 }, { "epoch": 2.92, "learning_rate": 3.544844747376393e-05, "loss": 1.5845, "step": 54000 }, { "epoch": 2.92, "learning_rate": 3.544574272422374e-05, "loss": 1.5866, "step": 54010 }, { "epoch": 2.92, "learning_rate": 3.5443037974683544e-05, "loss": 1.6293, "step": 54020 }, { "epoch": 2.92, "learning_rate": 3.5440333225143353e-05, "loss": 1.5742, "step": 54030 }, { "epoch": 2.92, "learning_rate": 3.543762847560316e-05, "loss": 1.563, "step": 54040 }, { "epoch": 2.92, "learning_rate": 3.5434923726062966e-05, "loss": 1.592, "step": 54050 }, { "epoch": 2.92, "learning_rate": 3.5432218976522776e-05, "loss": 1.5966, "step": 54060 }, { "epoch": 2.92, "learning_rate": 3.5429514226982585e-05, "loss": 1.6214, "step": 54070 }, { "epoch": 2.92, "learning_rate": 3.542680947744239e-05, "loss": 1.6034, "step": 54080 }, { "epoch": 2.92, "learning_rate": 3.54241047279022e-05, "loss": 1.6224, "step": 54090 }, { "epoch": 2.92, "learning_rate": 3.542139997836201e-05, "loss": 1.6565, "step": 54100 }, { "epoch": 2.92, "learning_rate": 3.541869522882181e-05, "loss": 1.6222, "step": 54110 }, { "epoch": 2.92, "learning_rate": 3.541599047928162e-05, "loss": 1.6275, "step": 54120 }, { "epoch": 2.92, "learning_rate": 3.541328572974143e-05, "loss": 1.6224, "step": 54130 }, { "epoch": 2.93, "learning_rate": 3.541058098020123e-05, "loss": 1.7328, "step": 54140 }, { "epoch": 2.93, "learning_rate": 3.540787623066104e-05, "loss": 1.6099, "step": 54150 }, { "epoch": 2.93, "learning_rate": 3.540517148112085e-05, "loss": 1.6385, "step": 54160 }, { "epoch": 2.93, "learning_rate": 3.5402466731580655e-05, "loss": 1.5708, "step": 54170 }, { "epoch": 2.93, "learning_rate": 3.5399761982040464e-05, "loss": 1.5637, "step": 54180 }, { "epoch": 2.93, "learning_rate": 3.5397057232500274e-05, "loss": 1.6547, "step": 54190 }, { "epoch": 2.93, "learning_rate": 3.5394352482960084e-05, "loss": 1.6256, "step": 54200 }, { "epoch": 2.93, "learning_rate": 3.5391647733419886e-05, "loss": 1.665, "step": 54210 }, { "epoch": 2.93, "learning_rate": 3.5388942983879696e-05, "loss": 1.5603, "step": 54220 }, { "epoch": 2.93, "learning_rate": 3.5386238234339506e-05, "loss": 1.5646, "step": 54230 }, { "epoch": 2.93, "learning_rate": 3.538353348479931e-05, "loss": 1.6687, "step": 54240 }, { "epoch": 2.93, "learning_rate": 3.538082873525912e-05, "loss": 1.5622, "step": 54250 }, { "epoch": 2.93, "learning_rate": 3.537812398571893e-05, "loss": 1.6402, "step": 54260 }, { "epoch": 2.93, "learning_rate": 3.537541923617873e-05, "loss": 1.5812, "step": 54270 }, { "epoch": 2.93, "learning_rate": 3.5372714486638534e-05, "loss": 1.574, "step": 54280 }, { "epoch": 2.93, "learning_rate": 3.537000973709835e-05, "loss": 1.5233, "step": 54290 }, { "epoch": 2.93, "learning_rate": 3.536730498755815e-05, "loss": 1.6338, "step": 54300 }, { "epoch": 2.93, "learning_rate": 3.5364600238017956e-05, "loss": 1.5761, "step": 54310 }, { "epoch": 2.94, "learning_rate": 3.536189548847777e-05, "loss": 1.5552, "step": 54320 }, { "epoch": 2.94, "learning_rate": 3.5359190738937575e-05, "loss": 1.6154, "step": 54330 }, { "epoch": 2.94, "learning_rate": 3.535648598939738e-05, "loss": 1.5879, "step": 54340 }, { "epoch": 2.94, "learning_rate": 3.5353781239857194e-05, "loss": 1.6347, "step": 54350 }, { "epoch": 2.94, "learning_rate": 3.5351076490317e-05, "loss": 1.6735, "step": 54360 }, { "epoch": 2.94, "learning_rate": 3.53483717407768e-05, "loss": 1.55, "step": 54370 }, { "epoch": 2.94, "learning_rate": 3.5345666991236617e-05, "loss": 1.5817, "step": 54380 }, { "epoch": 2.94, "learning_rate": 3.534296224169642e-05, "loss": 1.5803, "step": 54390 }, { "epoch": 2.94, "learning_rate": 3.534025749215623e-05, "loss": 1.6, "step": 54400 }, { "epoch": 2.94, "learning_rate": 3.533755274261604e-05, "loss": 1.5513, "step": 54410 }, { "epoch": 2.94, "learning_rate": 3.533484799307584e-05, "loss": 1.5925, "step": 54420 }, { "epoch": 2.94, "learning_rate": 3.533214324353565e-05, "loss": 1.6035, "step": 54430 }, { "epoch": 2.94, "learning_rate": 3.532943849399546e-05, "loss": 1.6074, "step": 54440 }, { "epoch": 2.94, "learning_rate": 3.5326733744455264e-05, "loss": 1.5798, "step": 54450 }, { "epoch": 2.94, "learning_rate": 3.5324028994915073e-05, "loss": 1.5461, "step": 54460 }, { "epoch": 2.94, "learning_rate": 3.532132424537488e-05, "loss": 1.6307, "step": 54470 }, { "epoch": 2.94, "learning_rate": 3.5318619495834686e-05, "loss": 1.5201, "step": 54480 }, { "epoch": 2.94, "learning_rate": 3.5315914746294496e-05, "loss": 1.6403, "step": 54490 }, { "epoch": 2.94, "learning_rate": 3.5313209996754305e-05, "loss": 1.5262, "step": 54500 }, { "epoch": 2.95, "learning_rate": 3.531050524721411e-05, "loss": 1.6621, "step": 54510 }, { "epoch": 2.95, "learning_rate": 3.530780049767392e-05, "loss": 1.654, "step": 54520 }, { "epoch": 2.95, "learning_rate": 3.530509574813373e-05, "loss": 1.5934, "step": 54530 }, { "epoch": 2.95, "learning_rate": 3.530239099859353e-05, "loss": 1.5985, "step": 54540 }, { "epoch": 2.95, "learning_rate": 3.529968624905334e-05, "loss": 1.5985, "step": 54550 }, { "epoch": 2.95, "learning_rate": 3.529698149951315e-05, "loss": 1.5552, "step": 54560 }, { "epoch": 2.95, "learning_rate": 3.529427674997295e-05, "loss": 1.5855, "step": 54570 }, { "epoch": 2.95, "learning_rate": 3.529157200043276e-05, "loss": 1.5308, "step": 54580 }, { "epoch": 2.95, "learning_rate": 3.528886725089257e-05, "loss": 1.674, "step": 54590 }, { "epoch": 2.95, "learning_rate": 3.5286162501352375e-05, "loss": 1.5576, "step": 54600 }, { "epoch": 2.95, "learning_rate": 3.5283457751812184e-05, "loss": 1.5395, "step": 54610 }, { "epoch": 2.95, "learning_rate": 3.5280753002271994e-05, "loss": 1.6311, "step": 54620 }, { "epoch": 2.95, "learning_rate": 3.52780482527318e-05, "loss": 1.6333, "step": 54630 }, { "epoch": 2.95, "learning_rate": 3.5275343503191606e-05, "loss": 1.5199, "step": 54640 }, { "epoch": 2.95, "learning_rate": 3.5272638753651416e-05, "loss": 1.584, "step": 54650 }, { "epoch": 2.95, "learning_rate": 3.526993400411122e-05, "loss": 1.6523, "step": 54660 }, { "epoch": 2.95, "learning_rate": 3.526722925457103e-05, "loss": 1.6518, "step": 54670 }, { "epoch": 2.95, "learning_rate": 3.526452450503084e-05, "loss": 1.6092, "step": 54680 }, { "epoch": 2.96, "learning_rate": 3.526181975549064e-05, "loss": 1.6377, "step": 54690 }, { "epoch": 2.96, "learning_rate": 3.525911500595045e-05, "loss": 1.5974, "step": 54700 }, { "epoch": 2.96, "learning_rate": 3.525641025641026e-05, "loss": 1.5746, "step": 54710 }, { "epoch": 2.96, "learning_rate": 3.525370550687006e-05, "loss": 1.6616, "step": 54720 }, { "epoch": 2.96, "learning_rate": 3.525100075732987e-05, "loss": 1.6211, "step": 54730 }, { "epoch": 2.96, "learning_rate": 3.524829600778968e-05, "loss": 1.5837, "step": 54740 }, { "epoch": 2.96, "learning_rate": 3.5245591258249485e-05, "loss": 1.6601, "step": 54750 }, { "epoch": 2.96, "learning_rate": 3.5242886508709295e-05, "loss": 1.6019, "step": 54760 }, { "epoch": 2.96, "learning_rate": 3.5240181759169105e-05, "loss": 1.5529, "step": 54770 }, { "epoch": 2.96, "learning_rate": 3.523747700962891e-05, "loss": 1.6496, "step": 54780 }, { "epoch": 2.96, "learning_rate": 3.523477226008872e-05, "loss": 1.6298, "step": 54790 }, { "epoch": 2.96, "learning_rate": 3.523206751054853e-05, "loss": 1.5674, "step": 54800 }, { "epoch": 2.96, "learning_rate": 3.5229362761008337e-05, "loss": 1.5417, "step": 54810 }, { "epoch": 2.96, "learning_rate": 3.522665801146814e-05, "loss": 1.5431, "step": 54820 }, { "epoch": 2.96, "learning_rate": 3.522395326192794e-05, "loss": 1.6347, "step": 54830 }, { "epoch": 2.96, "learning_rate": 3.522124851238776e-05, "loss": 1.5236, "step": 54840 }, { "epoch": 2.96, "learning_rate": 3.521854376284756e-05, "loss": 1.6186, "step": 54850 }, { "epoch": 2.96, "learning_rate": 3.5215839013307365e-05, "loss": 1.5844, "step": 54860 }, { "epoch": 2.96, "learning_rate": 3.521313426376718e-05, "loss": 1.6128, "step": 54870 }, { "epoch": 2.97, "learning_rate": 3.5210429514226984e-05, "loss": 1.6383, "step": 54880 }, { "epoch": 2.97, "learning_rate": 3.520772476468679e-05, "loss": 1.6719, "step": 54890 }, { "epoch": 2.97, "learning_rate": 3.52050200151466e-05, "loss": 1.5922, "step": 54900 }, { "epoch": 2.97, "learning_rate": 3.5202315265606406e-05, "loss": 1.5643, "step": 54910 }, { "epoch": 2.97, "learning_rate": 3.519961051606621e-05, "loss": 1.6465, "step": 54920 }, { "epoch": 2.97, "learning_rate": 3.5196905766526025e-05, "loss": 1.6933, "step": 54930 }, { "epoch": 2.97, "learning_rate": 3.519420101698583e-05, "loss": 1.6656, "step": 54940 }, { "epoch": 2.97, "learning_rate": 3.519149626744563e-05, "loss": 1.6059, "step": 54950 }, { "epoch": 2.97, "learning_rate": 3.518879151790545e-05, "loss": 1.6475, "step": 54960 }, { "epoch": 2.97, "learning_rate": 3.518608676836525e-05, "loss": 1.577, "step": 54970 }, { "epoch": 2.97, "learning_rate": 3.518338201882505e-05, "loss": 1.5975, "step": 54980 }, { "epoch": 2.97, "learning_rate": 3.518067726928487e-05, "loss": 1.565, "step": 54990 }, { "epoch": 2.97, "learning_rate": 3.517797251974467e-05, "loss": 1.5447, "step": 55000 }, { "epoch": 2.97, "learning_rate": 3.5175267770204475e-05, "loss": 1.6294, "step": 55010 }, { "epoch": 2.97, "learning_rate": 3.517256302066429e-05, "loss": 1.5896, "step": 55020 }, { "epoch": 2.97, "learning_rate": 3.5169858271124095e-05, "loss": 1.5708, "step": 55030 }, { "epoch": 2.97, "learning_rate": 3.5167153521583904e-05, "loss": 1.7041, "step": 55040 }, { "epoch": 2.97, "learning_rate": 3.5164448772043714e-05, "loss": 1.5667, "step": 55050 }, { "epoch": 2.98, "learning_rate": 3.516174402250352e-05, "loss": 1.6178, "step": 55060 }, { "epoch": 2.98, "learning_rate": 3.5159039272963326e-05, "loss": 1.5816, "step": 55070 }, { "epoch": 2.98, "learning_rate": 3.5156334523423136e-05, "loss": 1.5408, "step": 55080 }, { "epoch": 2.98, "learning_rate": 3.515362977388294e-05, "loss": 1.6049, "step": 55090 }, { "epoch": 2.98, "learning_rate": 3.515092502434275e-05, "loss": 1.6833, "step": 55100 }, { "epoch": 2.98, "learning_rate": 3.514822027480256e-05, "loss": 1.6438, "step": 55110 }, { "epoch": 2.98, "learning_rate": 3.514551552526236e-05, "loss": 1.6184, "step": 55120 }, { "epoch": 2.98, "learning_rate": 3.514281077572217e-05, "loss": 1.5446, "step": 55130 }, { "epoch": 2.98, "learning_rate": 3.514010602618198e-05, "loss": 1.618, "step": 55140 }, { "epoch": 2.98, "learning_rate": 3.513740127664178e-05, "loss": 1.54, "step": 55150 }, { "epoch": 2.98, "learning_rate": 3.513469652710159e-05, "loss": 1.6405, "step": 55160 }, { "epoch": 2.98, "learning_rate": 3.51319917775614e-05, "loss": 1.5976, "step": 55170 }, { "epoch": 2.98, "learning_rate": 3.5129287028021206e-05, "loss": 1.6034, "step": 55180 }, { "epoch": 2.98, "learning_rate": 3.5126582278481015e-05, "loss": 1.5257, "step": 55190 }, { "epoch": 2.98, "learning_rate": 3.5123877528940825e-05, "loss": 1.6494, "step": 55200 }, { "epoch": 2.98, "learning_rate": 3.512117277940063e-05, "loss": 1.6169, "step": 55210 }, { "epoch": 2.98, "learning_rate": 3.511846802986044e-05, "loss": 1.513, "step": 55220 }, { "epoch": 2.98, "learning_rate": 3.511576328032025e-05, "loss": 1.6211, "step": 55230 }, { "epoch": 2.98, "learning_rate": 3.511305853078005e-05, "loss": 1.6031, "step": 55240 }, { "epoch": 2.99, "learning_rate": 3.511035378123986e-05, "loss": 1.6521, "step": 55250 }, { "epoch": 2.99, "learning_rate": 3.510764903169967e-05, "loss": 1.6013, "step": 55260 }, { "epoch": 2.99, "learning_rate": 3.510494428215947e-05, "loss": 1.6332, "step": 55270 }, { "epoch": 2.99, "learning_rate": 3.510223953261928e-05, "loss": 1.5775, "step": 55280 }, { "epoch": 2.99, "learning_rate": 3.509953478307909e-05, "loss": 1.6138, "step": 55290 }, { "epoch": 2.99, "learning_rate": 3.5096830033538894e-05, "loss": 1.5105, "step": 55300 }, { "epoch": 2.99, "learning_rate": 3.5094125283998704e-05, "loss": 1.6359, "step": 55310 }, { "epoch": 2.99, "learning_rate": 3.5091420534458513e-05, "loss": 1.5533, "step": 55320 }, { "epoch": 2.99, "learning_rate": 3.5088715784918316e-05, "loss": 1.594, "step": 55330 }, { "epoch": 2.99, "learning_rate": 3.5086011035378126e-05, "loss": 1.6039, "step": 55340 }, { "epoch": 2.99, "learning_rate": 3.5083306285837936e-05, "loss": 1.6338, "step": 55350 }, { "epoch": 2.99, "learning_rate": 3.508060153629774e-05, "loss": 1.5537, "step": 55360 }, { "epoch": 2.99, "learning_rate": 3.507789678675755e-05, "loss": 1.572, "step": 55370 }, { "epoch": 2.99, "learning_rate": 3.507519203721735e-05, "loss": 1.6414, "step": 55380 }, { "epoch": 2.99, "learning_rate": 3.507248728767716e-05, "loss": 1.5923, "step": 55390 }, { "epoch": 2.99, "learning_rate": 3.506978253813697e-05, "loss": 1.6481, "step": 55400 }, { "epoch": 2.99, "learning_rate": 3.506707778859677e-05, "loss": 1.6745, "step": 55410 }, { "epoch": 2.99, "learning_rate": 3.506437303905658e-05, "loss": 1.6528, "step": 55420 }, { "epoch": 3.0, "learning_rate": 3.506166828951639e-05, "loss": 1.5807, "step": 55430 }, { "epoch": 3.0, "learning_rate": 3.5058963539976195e-05, "loss": 1.6258, "step": 55440 }, { "epoch": 3.0, "learning_rate": 3.505625879043601e-05, "loss": 1.6347, "step": 55450 }, { "epoch": 3.0, "learning_rate": 3.5053554040895815e-05, "loss": 1.5552, "step": 55460 }, { "epoch": 3.0, "learning_rate": 3.505084929135562e-05, "loss": 1.694, "step": 55470 }, { "epoch": 3.0, "learning_rate": 3.5048144541815434e-05, "loss": 1.6293, "step": 55480 }, { "epoch": 3.0, "learning_rate": 3.504543979227524e-05, "loss": 1.66, "step": 55490 }, { "epoch": 3.0, "learning_rate": 3.504273504273504e-05, "loss": 1.6243, "step": 55500 }, { "epoch": 3.0, "learning_rate": 3.5040030293194856e-05, "loss": 1.6726, "step": 55510 }, { "epoch": 3.0, "learning_rate": 3.503732554365466e-05, "loss": 1.6745, "step": 55520 }, { "epoch": 3.0, "learning_rate": 3.503462079411446e-05, "loss": 1.6135, "step": 55530 }, { "epoch": 3.0, "learning_rate": 3.503191604457428e-05, "loss": 1.4242, "step": 55540 }, { "epoch": 3.0, "learning_rate": 3.502921129503408e-05, "loss": 1.4723, "step": 55550 }, { "epoch": 3.0, "learning_rate": 3.5026506545493884e-05, "loss": 1.5245, "step": 55560 }, { "epoch": 3.0, "learning_rate": 3.50238017959537e-05, "loss": 1.4952, "step": 55570 }, { "epoch": 3.0, "learning_rate": 3.50210970464135e-05, "loss": 1.5049, "step": 55580 }, { "epoch": 3.0, "learning_rate": 3.5018392296873306e-05, "loss": 1.5456, "step": 55590 }, { "epoch": 3.0, "learning_rate": 3.501568754733312e-05, "loss": 1.5377, "step": 55600 }, { "epoch": 3.0, "learning_rate": 3.5012982797792926e-05, "loss": 1.5101, "step": 55610 }, { "epoch": 3.01, "learning_rate": 3.501027804825273e-05, "loss": 1.4685, "step": 55620 }, { "epoch": 3.01, "learning_rate": 3.5007573298712545e-05, "loss": 1.4428, "step": 55630 }, { "epoch": 3.01, "learning_rate": 3.500486854917235e-05, "loss": 1.4446, "step": 55640 }, { "epoch": 3.01, "learning_rate": 3.500216379963215e-05, "loss": 1.4349, "step": 55650 }, { "epoch": 3.01, "learning_rate": 3.499945905009197e-05, "loss": 1.4281, "step": 55660 }, { "epoch": 3.01, "learning_rate": 3.499675430055177e-05, "loss": 1.4552, "step": 55670 }, { "epoch": 3.01, "learning_rate": 3.499404955101158e-05, "loss": 1.4563, "step": 55680 }, { "epoch": 3.01, "learning_rate": 3.499134480147139e-05, "loss": 1.4734, "step": 55690 }, { "epoch": 3.01, "learning_rate": 3.498864005193119e-05, "loss": 1.4474, "step": 55700 }, { "epoch": 3.01, "learning_rate": 3.4985935302391e-05, "loss": 1.5156, "step": 55710 }, { "epoch": 3.01, "learning_rate": 3.498323055285081e-05, "loss": 1.4465, "step": 55720 }, { "epoch": 3.01, "learning_rate": 3.4980525803310614e-05, "loss": 1.5856, "step": 55730 }, { "epoch": 3.01, "learning_rate": 3.4977821053770424e-05, "loss": 1.4405, "step": 55740 }, { "epoch": 3.01, "learning_rate": 3.4975116304230234e-05, "loss": 1.47, "step": 55750 }, { "epoch": 3.01, "learning_rate": 3.4972411554690036e-05, "loss": 1.4388, "step": 55760 }, { "epoch": 3.01, "learning_rate": 3.4969706805149846e-05, "loss": 1.4755, "step": 55770 }, { "epoch": 3.01, "learning_rate": 3.4967002055609656e-05, "loss": 1.443, "step": 55780 }, { "epoch": 3.01, "learning_rate": 3.496429730606946e-05, "loss": 1.5346, "step": 55790 }, { "epoch": 3.02, "learning_rate": 3.496159255652927e-05, "loss": 1.4202, "step": 55800 }, { "epoch": 3.02, "learning_rate": 3.495888780698908e-05, "loss": 1.4932, "step": 55810 }, { "epoch": 3.02, "learning_rate": 3.495618305744888e-05, "loss": 1.5401, "step": 55820 }, { "epoch": 3.02, "learning_rate": 3.495347830790869e-05, "loss": 1.507, "step": 55830 }, { "epoch": 3.02, "learning_rate": 3.49507735583685e-05, "loss": 1.5675, "step": 55840 }, { "epoch": 3.02, "learning_rate": 3.49480688088283e-05, "loss": 1.5219, "step": 55850 }, { "epoch": 3.02, "learning_rate": 3.494536405928811e-05, "loss": 1.4926, "step": 55860 }, { "epoch": 3.02, "learning_rate": 3.494265930974792e-05, "loss": 1.4421, "step": 55870 }, { "epoch": 3.02, "learning_rate": 3.4939954560207725e-05, "loss": 1.4523, "step": 55880 }, { "epoch": 3.02, "learning_rate": 3.4937249810667535e-05, "loss": 1.4913, "step": 55890 }, { "epoch": 3.02, "learning_rate": 3.4934545061127344e-05, "loss": 1.4803, "step": 55900 }, { "epoch": 3.02, "learning_rate": 3.493184031158715e-05, "loss": 1.4612, "step": 55910 }, { "epoch": 3.02, "learning_rate": 3.492913556204696e-05, "loss": 1.474, "step": 55920 }, { "epoch": 3.02, "learning_rate": 3.4926430812506767e-05, "loss": 1.4925, "step": 55930 }, { "epoch": 3.02, "learning_rate": 3.492372606296657e-05, "loss": 1.5303, "step": 55940 }, { "epoch": 3.02, "learning_rate": 3.492102131342638e-05, "loss": 1.4763, "step": 55950 }, { "epoch": 3.02, "learning_rate": 3.491831656388618e-05, "loss": 1.4787, "step": 55960 }, { "epoch": 3.02, "learning_rate": 3.491561181434599e-05, "loss": 1.437, "step": 55970 }, { "epoch": 3.02, "learning_rate": 3.49129070648058e-05, "loss": 1.521, "step": 55980 }, { "epoch": 3.03, "learning_rate": 3.4910202315265604e-05, "loss": 1.4255, "step": 55990 }, { "epoch": 3.03, "learning_rate": 3.4907497565725414e-05, "loss": 1.5455, "step": 56000 }, { "epoch": 3.03, "learning_rate": 3.490479281618522e-05, "loss": 1.4206, "step": 56010 }, { "epoch": 3.03, "learning_rate": 3.4902088066645026e-05, "loss": 1.4864, "step": 56020 }, { "epoch": 3.03, "learning_rate": 3.4899383317104836e-05, "loss": 1.5135, "step": 56030 }, { "epoch": 3.03, "learning_rate": 3.4896678567564646e-05, "loss": 1.5049, "step": 56040 }, { "epoch": 3.03, "learning_rate": 3.489397381802445e-05, "loss": 1.5002, "step": 56050 }, { "epoch": 3.03, "learning_rate": 3.489126906848426e-05, "loss": 1.4918, "step": 56060 }, { "epoch": 3.03, "learning_rate": 3.488856431894407e-05, "loss": 1.5226, "step": 56070 }, { "epoch": 3.03, "learning_rate": 3.488585956940387e-05, "loss": 1.5217, "step": 56080 }, { "epoch": 3.03, "learning_rate": 3.488315481986369e-05, "loss": 1.6005, "step": 56090 }, { "epoch": 3.03, "learning_rate": 3.488045007032349e-05, "loss": 1.4866, "step": 56100 }, { "epoch": 3.03, "learning_rate": 3.487774532078329e-05, "loss": 1.5221, "step": 56110 }, { "epoch": 3.03, "learning_rate": 3.487504057124311e-05, "loss": 1.406, "step": 56120 }, { "epoch": 3.03, "learning_rate": 3.487233582170291e-05, "loss": 1.5339, "step": 56130 }, { "epoch": 3.03, "learning_rate": 3.4869631072162715e-05, "loss": 1.5371, "step": 56140 }, { "epoch": 3.03, "learning_rate": 3.486692632262253e-05, "loss": 1.5128, "step": 56150 }, { "epoch": 3.03, "learning_rate": 3.4864221573082334e-05, "loss": 1.5771, "step": 56160 }, { "epoch": 3.04, "learning_rate": 3.486151682354214e-05, "loss": 1.4855, "step": 56170 }, { "epoch": 3.04, "learning_rate": 3.4858812074001954e-05, "loss": 1.5064, "step": 56180 }, { "epoch": 3.04, "learning_rate": 3.4856107324461756e-05, "loss": 1.5252, "step": 56190 }, { "epoch": 3.04, "learning_rate": 3.485340257492156e-05, "loss": 1.4133, "step": 56200 }, { "epoch": 3.04, "learning_rate": 3.4850697825381376e-05, "loss": 1.4408, "step": 56210 }, { "epoch": 3.04, "learning_rate": 3.484799307584118e-05, "loss": 1.5182, "step": 56220 }, { "epoch": 3.04, "learning_rate": 3.484528832630098e-05, "loss": 1.4368, "step": 56230 }, { "epoch": 3.04, "learning_rate": 3.48425835767608e-05, "loss": 1.4905, "step": 56240 }, { "epoch": 3.04, "learning_rate": 3.48398788272206e-05, "loss": 1.4739, "step": 56250 }, { "epoch": 3.04, "learning_rate": 3.4837174077680404e-05, "loss": 1.4499, "step": 56260 }, { "epoch": 3.04, "learning_rate": 3.483446932814022e-05, "loss": 1.4293, "step": 56270 }, { "epoch": 3.04, "learning_rate": 3.483176457860002e-05, "loss": 1.5148, "step": 56280 }, { "epoch": 3.04, "learning_rate": 3.4829059829059826e-05, "loss": 1.5497, "step": 56290 }, { "epoch": 3.04, "learning_rate": 3.482635507951964e-05, "loss": 1.4957, "step": 56300 }, { "epoch": 3.04, "learning_rate": 3.4823650329979445e-05, "loss": 1.4734, "step": 56310 }, { "epoch": 3.04, "learning_rate": 3.4820945580439255e-05, "loss": 1.3828, "step": 56320 }, { "epoch": 3.04, "learning_rate": 3.4818240830899064e-05, "loss": 1.4462, "step": 56330 }, { "epoch": 3.04, "learning_rate": 3.481553608135887e-05, "loss": 1.4374, "step": 56340 }, { "epoch": 3.04, "learning_rate": 3.481283133181868e-05, "loss": 1.4758, "step": 56350 }, { "epoch": 3.05, "learning_rate": 3.4810126582278487e-05, "loss": 1.5122, "step": 56360 }, { "epoch": 3.05, "learning_rate": 3.480742183273829e-05, "loss": 1.5069, "step": 56370 }, { "epoch": 3.05, "learning_rate": 3.48047170831981e-05, "loss": 1.5134, "step": 56380 }, { "epoch": 3.05, "learning_rate": 3.480201233365791e-05, "loss": 1.4079, "step": 56390 }, { "epoch": 3.05, "learning_rate": 3.479930758411771e-05, "loss": 1.5039, "step": 56400 }, { "epoch": 3.05, "learning_rate": 3.479660283457752e-05, "loss": 1.4648, "step": 56410 }, { "epoch": 3.05, "learning_rate": 3.479389808503733e-05, "loss": 1.4033, "step": 56420 }, { "epoch": 3.05, "learning_rate": 3.4791193335497134e-05, "loss": 1.4727, "step": 56430 }, { "epoch": 3.05, "learning_rate": 3.4788488585956943e-05, "loss": 1.4345, "step": 56440 }, { "epoch": 3.05, "learning_rate": 3.478578383641675e-05, "loss": 1.4942, "step": 56450 }, { "epoch": 3.05, "learning_rate": 3.4783079086876556e-05, "loss": 1.5176, "step": 56460 }, { "epoch": 3.05, "learning_rate": 3.4780374337336366e-05, "loss": 1.5405, "step": 56470 }, { "epoch": 3.05, "learning_rate": 3.4777669587796175e-05, "loss": 1.5085, "step": 56480 }, { "epoch": 3.05, "learning_rate": 3.477496483825598e-05, "loss": 1.4823, "step": 56490 }, { "epoch": 3.05, "learning_rate": 3.477226008871579e-05, "loss": 1.5277, "step": 56500 }, { "epoch": 3.05, "learning_rate": 3.476955533917559e-05, "loss": 1.5093, "step": 56510 }, { "epoch": 3.05, "learning_rate": 3.47668505896354e-05, "loss": 1.4658, "step": 56520 }, { "epoch": 3.05, "learning_rate": 3.476414584009521e-05, "loss": 1.486, "step": 56530 }, { "epoch": 3.06, "learning_rate": 3.476144109055501e-05, "loss": 1.4953, "step": 56540 }, { "epoch": 3.06, "learning_rate": 3.475873634101482e-05, "loss": 1.4664, "step": 56550 }, { "epoch": 3.06, "learning_rate": 3.475603159147463e-05, "loss": 1.548, "step": 56560 }, { "epoch": 3.06, "learning_rate": 3.4753326841934435e-05, "loss": 1.5615, "step": 56570 }, { "epoch": 3.06, "learning_rate": 3.4750622092394245e-05, "loss": 1.4559, "step": 56580 }, { "epoch": 3.06, "learning_rate": 3.4747917342854054e-05, "loss": 1.3688, "step": 56590 }, { "epoch": 3.06, "learning_rate": 3.474521259331386e-05, "loss": 1.5166, "step": 56600 }, { "epoch": 3.06, "learning_rate": 3.474250784377367e-05, "loss": 1.4928, "step": 56610 }, { "epoch": 3.06, "learning_rate": 3.4739803094233476e-05, "loss": 1.5384, "step": 56620 }, { "epoch": 3.06, "learning_rate": 3.473709834469328e-05, "loss": 1.4531, "step": 56630 }, { "epoch": 3.06, "learning_rate": 3.473439359515309e-05, "loss": 1.458, "step": 56640 }, { "epoch": 3.06, "learning_rate": 3.47316888456129e-05, "loss": 1.5362, "step": 56650 }, { "epoch": 3.06, "learning_rate": 3.47289840960727e-05, "loss": 1.4774, "step": 56660 }, { "epoch": 3.06, "learning_rate": 3.472627934653251e-05, "loss": 1.4577, "step": 56670 }, { "epoch": 3.06, "learning_rate": 3.472357459699232e-05, "loss": 1.4864, "step": 56680 }, { "epoch": 3.06, "learning_rate": 3.4720869847452124e-05, "loss": 1.5324, "step": 56690 }, { "epoch": 3.06, "learning_rate": 3.471816509791193e-05, "loss": 1.5319, "step": 56700 }, { "epoch": 3.06, "learning_rate": 3.471546034837174e-05, "loss": 1.4864, "step": 56710 }, { "epoch": 3.06, "learning_rate": 3.4712755598831546e-05, "loss": 1.4889, "step": 56720 }, { "epoch": 3.07, "learning_rate": 3.471005084929136e-05, "loss": 1.512, "step": 56730 }, { "epoch": 3.07, "learning_rate": 3.4707346099751165e-05, "loss": 1.4442, "step": 56740 }, { "epoch": 3.07, "learning_rate": 3.470464135021097e-05, "loss": 1.5602, "step": 56750 }, { "epoch": 3.07, "learning_rate": 3.4701936600670784e-05, "loss": 1.5723, "step": 56760 }, { "epoch": 3.07, "learning_rate": 3.469923185113059e-05, "loss": 1.5666, "step": 56770 }, { "epoch": 3.07, "learning_rate": 3.469652710159039e-05, "loss": 1.4582, "step": 56780 }, { "epoch": 3.07, "learning_rate": 3.4693822352050207e-05, "loss": 1.5452, "step": 56790 }, { "epoch": 3.07, "learning_rate": 3.469111760251001e-05, "loss": 1.5229, "step": 56800 }, { "epoch": 3.07, "learning_rate": 3.468841285296981e-05, "loss": 1.4261, "step": 56810 }, { "epoch": 3.07, "learning_rate": 3.468570810342963e-05, "loss": 1.5096, "step": 56820 }, { "epoch": 3.07, "learning_rate": 3.468300335388943e-05, "loss": 1.4762, "step": 56830 }, { "epoch": 3.07, "learning_rate": 3.4680298604349234e-05, "loss": 1.426, "step": 56840 }, { "epoch": 3.07, "learning_rate": 3.467759385480905e-05, "loss": 1.5534, "step": 56850 }, { "epoch": 3.07, "learning_rate": 3.4674889105268854e-05, "loss": 1.4938, "step": 56860 }, { "epoch": 3.07, "learning_rate": 3.467218435572866e-05, "loss": 1.4619, "step": 56870 }, { "epoch": 3.07, "learning_rate": 3.466947960618847e-05, "loss": 1.4662, "step": 56880 }, { "epoch": 3.07, "learning_rate": 3.4666774856648276e-05, "loss": 1.3947, "step": 56890 }, { "epoch": 3.07, "learning_rate": 3.466407010710808e-05, "loss": 1.4322, "step": 56900 }, { "epoch": 3.08, "learning_rate": 3.4661365357567895e-05, "loss": 1.5062, "step": 56910 }, { "epoch": 3.08, "learning_rate": 3.46586606080277e-05, "loss": 1.4651, "step": 56920 }, { "epoch": 3.08, "learning_rate": 3.46559558584875e-05, "loss": 1.489, "step": 56930 }, { "epoch": 3.08, "learning_rate": 3.465325110894732e-05, "loss": 1.4918, "step": 56940 }, { "epoch": 3.08, "learning_rate": 3.465054635940712e-05, "loss": 1.4833, "step": 56950 }, { "epoch": 3.08, "learning_rate": 3.464784160986693e-05, "loss": 1.4642, "step": 56960 }, { "epoch": 3.08, "learning_rate": 3.464513686032674e-05, "loss": 1.4983, "step": 56970 }, { "epoch": 3.08, "learning_rate": 3.464243211078654e-05, "loss": 1.4782, "step": 56980 }, { "epoch": 3.08, "learning_rate": 3.463972736124635e-05, "loss": 1.4984, "step": 56990 }, { "epoch": 3.08, "learning_rate": 3.463702261170616e-05, "loss": 1.4888, "step": 57000 }, { "epoch": 3.08, "learning_rate": 3.4634317862165965e-05, "loss": 1.4569, "step": 57010 }, { "epoch": 3.08, "learning_rate": 3.4631613112625774e-05, "loss": 1.5474, "step": 57020 }, { "epoch": 3.08, "learning_rate": 3.4628908363085584e-05, "loss": 1.5202, "step": 57030 }, { "epoch": 3.08, "learning_rate": 3.462620361354539e-05, "loss": 1.4643, "step": 57040 }, { "epoch": 3.08, "learning_rate": 3.4623498864005196e-05, "loss": 1.5166, "step": 57050 }, { "epoch": 3.08, "learning_rate": 3.4620794114465e-05, "loss": 1.471, "step": 57060 }, { "epoch": 3.08, "learning_rate": 3.461808936492481e-05, "loss": 1.5091, "step": 57070 }, { "epoch": 3.08, "learning_rate": 3.461538461538462e-05, "loss": 1.5163, "step": 57080 }, { "epoch": 3.08, "learning_rate": 3.461267986584442e-05, "loss": 1.4812, "step": 57090 }, { "epoch": 3.09, "learning_rate": 3.460997511630423e-05, "loss": 1.4914, "step": 57100 }, { "epoch": 3.09, "learning_rate": 3.460727036676404e-05, "loss": 1.417, "step": 57110 }, { "epoch": 3.09, "learning_rate": 3.4604565617223844e-05, "loss": 1.4818, "step": 57120 }, { "epoch": 3.09, "learning_rate": 3.460186086768365e-05, "loss": 1.4496, "step": 57130 }, { "epoch": 3.09, "learning_rate": 3.459915611814346e-05, "loss": 1.5453, "step": 57140 }, { "epoch": 3.09, "learning_rate": 3.4596451368603266e-05, "loss": 1.4909, "step": 57150 }, { "epoch": 3.09, "learning_rate": 3.4593746619063075e-05, "loss": 1.4409, "step": 57160 }, { "epoch": 3.09, "learning_rate": 3.4591041869522885e-05, "loss": 1.453, "step": 57170 }, { "epoch": 3.09, "learning_rate": 3.458833711998269e-05, "loss": 1.4748, "step": 57180 }, { "epoch": 3.09, "learning_rate": 3.45856323704425e-05, "loss": 1.5244, "step": 57190 }, { "epoch": 3.09, "learning_rate": 3.458292762090231e-05, "loss": 1.5716, "step": 57200 }, { "epoch": 3.09, "learning_rate": 3.458022287136211e-05, "loss": 1.4788, "step": 57210 }, { "epoch": 3.09, "learning_rate": 3.457751812182192e-05, "loss": 1.4556, "step": 57220 }, { "epoch": 3.09, "learning_rate": 3.457481337228173e-05, "loss": 1.5326, "step": 57230 }, { "epoch": 3.09, "learning_rate": 3.457210862274153e-05, "loss": 1.4121, "step": 57240 }, { "epoch": 3.09, "learning_rate": 3.456940387320134e-05, "loss": 1.4646, "step": 57250 }, { "epoch": 3.09, "learning_rate": 3.456669912366115e-05, "loss": 1.5497, "step": 57260 }, { "epoch": 3.09, "learning_rate": 3.4563994374120954e-05, "loss": 1.4722, "step": 57270 }, { "epoch": 3.1, "learning_rate": 3.4561289624580764e-05, "loss": 1.4443, "step": 57280 }, { "epoch": 3.1, "learning_rate": 3.4558584875040574e-05, "loss": 1.4988, "step": 57290 }, { "epoch": 3.1, "learning_rate": 3.455588012550038e-05, "loss": 1.4758, "step": 57300 }, { "epoch": 3.1, "learning_rate": 3.4553175375960186e-05, "loss": 1.4402, "step": 57310 }, { "epoch": 3.1, "learning_rate": 3.4550470626419996e-05, "loss": 1.4388, "step": 57320 }, { "epoch": 3.1, "learning_rate": 3.45477658768798e-05, "loss": 1.5161, "step": 57330 }, { "epoch": 3.1, "learning_rate": 3.454506112733961e-05, "loss": 1.4988, "step": 57340 }, { "epoch": 3.1, "learning_rate": 3.454235637779942e-05, "loss": 1.4626, "step": 57350 }, { "epoch": 3.1, "learning_rate": 3.453965162825922e-05, "loss": 1.4642, "step": 57360 }, { "epoch": 3.1, "learning_rate": 3.453694687871904e-05, "loss": 1.5182, "step": 57370 }, { "epoch": 3.1, "learning_rate": 3.453424212917884e-05, "loss": 1.4649, "step": 57380 }, { "epoch": 3.1, "learning_rate": 3.453153737963864e-05, "loss": 1.5404, "step": 57390 }, { "epoch": 3.1, "learning_rate": 3.452883263009846e-05, "loss": 1.4582, "step": 57400 }, { "epoch": 3.1, "learning_rate": 3.452612788055826e-05, "loss": 1.472, "step": 57410 }, { "epoch": 3.1, "learning_rate": 3.4523423131018065e-05, "loss": 1.4925, "step": 57420 }, { "epoch": 3.1, "learning_rate": 3.452071838147788e-05, "loss": 1.5064, "step": 57430 }, { "epoch": 3.1, "learning_rate": 3.4518013631937685e-05, "loss": 1.5029, "step": 57440 }, { "epoch": 3.1, "learning_rate": 3.451530888239749e-05, "loss": 1.4012, "step": 57450 }, { "epoch": 3.1, "learning_rate": 3.4512604132857304e-05, "loss": 1.514, "step": 57460 }, { "epoch": 3.11, "learning_rate": 3.450989938331711e-05, "loss": 1.4974, "step": 57470 }, { "epoch": 3.11, "learning_rate": 3.450719463377691e-05, "loss": 1.5056, "step": 57480 }, { "epoch": 3.11, "learning_rate": 3.4504489884236726e-05, "loss": 1.5127, "step": 57490 }, { "epoch": 3.11, "learning_rate": 3.450178513469653e-05, "loss": 1.5309, "step": 57500 }, { "epoch": 3.11, "learning_rate": 3.449908038515633e-05, "loss": 1.4693, "step": 57510 }, { "epoch": 3.11, "learning_rate": 3.449637563561615e-05, "loss": 1.4054, "step": 57520 }, { "epoch": 3.11, "learning_rate": 3.449367088607595e-05, "loss": 1.5119, "step": 57530 }, { "epoch": 3.11, "learning_rate": 3.4490966136535754e-05, "loss": 1.5136, "step": 57540 }, { "epoch": 3.11, "learning_rate": 3.448826138699557e-05, "loss": 1.463, "step": 57550 }, { "epoch": 3.11, "learning_rate": 3.448555663745537e-05, "loss": 1.4287, "step": 57560 }, { "epoch": 3.11, "learning_rate": 3.448285188791518e-05, "loss": 1.4721, "step": 57570 }, { "epoch": 3.11, "learning_rate": 3.448014713837499e-05, "loss": 1.5326, "step": 57580 }, { "epoch": 3.11, "learning_rate": 3.4477442388834795e-05, "loss": 1.472, "step": 57590 }, { "epoch": 3.11, "learning_rate": 3.4474737639294605e-05, "loss": 1.5146, "step": 57600 }, { "epoch": 3.11, "learning_rate": 3.4472032889754415e-05, "loss": 1.52, "step": 57610 }, { "epoch": 3.11, "learning_rate": 3.446932814021422e-05, "loss": 1.4578, "step": 57620 }, { "epoch": 3.11, "learning_rate": 3.446662339067403e-05, "loss": 1.3878, "step": 57630 }, { "epoch": 3.11, "learning_rate": 3.446391864113383e-05, "loss": 1.5488, "step": 57640 }, { "epoch": 3.12, "learning_rate": 3.446121389159364e-05, "loss": 1.5446, "step": 57650 }, { "epoch": 3.12, "learning_rate": 3.445850914205345e-05, "loss": 1.5469, "step": 57660 }, { "epoch": 3.12, "learning_rate": 3.445580439251325e-05, "loss": 1.4377, "step": 57670 }, { "epoch": 3.12, "learning_rate": 3.445309964297306e-05, "loss": 1.5362, "step": 57680 }, { "epoch": 3.12, "learning_rate": 3.445039489343287e-05, "loss": 1.4786, "step": 57690 }, { "epoch": 3.12, "learning_rate": 3.4447690143892675e-05, "loss": 1.5051, "step": 57700 }, { "epoch": 3.12, "learning_rate": 3.4444985394352484e-05, "loss": 1.5092, "step": 57710 }, { "epoch": 3.12, "learning_rate": 3.4442280644812294e-05, "loss": 1.446, "step": 57720 }, { "epoch": 3.12, "learning_rate": 3.44395758952721e-05, "loss": 1.3859, "step": 57730 }, { "epoch": 3.12, "learning_rate": 3.4436871145731906e-05, "loss": 1.5115, "step": 57740 }, { "epoch": 3.12, "learning_rate": 3.4434166396191716e-05, "loss": 1.4779, "step": 57750 }, { "epoch": 3.12, "learning_rate": 3.443146164665152e-05, "loss": 1.5334, "step": 57760 }, { "epoch": 3.12, "learning_rate": 3.442875689711133e-05, "loss": 1.5004, "step": 57770 }, { "epoch": 3.12, "learning_rate": 3.442605214757114e-05, "loss": 1.5223, "step": 57780 }, { "epoch": 3.12, "learning_rate": 3.442334739803094e-05, "loss": 1.3964, "step": 57790 }, { "epoch": 3.12, "learning_rate": 3.442064264849075e-05, "loss": 1.3378, "step": 57800 }, { "epoch": 3.12, "learning_rate": 3.441793789895056e-05, "loss": 1.5079, "step": 57810 }, { "epoch": 3.12, "learning_rate": 3.441523314941036e-05, "loss": 1.4624, "step": 57820 }, { "epoch": 3.12, "learning_rate": 3.441252839987017e-05, "loss": 1.4531, "step": 57830 }, { "epoch": 3.13, "learning_rate": 3.440982365032998e-05, "loss": 1.5363, "step": 57840 }, { "epoch": 3.13, "learning_rate": 3.4407118900789785e-05, "loss": 1.4333, "step": 57850 }, { "epoch": 3.13, "learning_rate": 3.4404414151249595e-05, "loss": 1.4608, "step": 57860 }, { "epoch": 3.13, "learning_rate": 3.4401709401709405e-05, "loss": 1.53, "step": 57870 }, { "epoch": 3.13, "learning_rate": 3.439900465216921e-05, "loss": 1.4453, "step": 57880 }, { "epoch": 3.13, "learning_rate": 3.439629990262902e-05, "loss": 1.5076, "step": 57890 }, { "epoch": 3.13, "learning_rate": 3.439359515308883e-05, "loss": 1.5668, "step": 57900 }, { "epoch": 3.13, "learning_rate": 3.439089040354863e-05, "loss": 1.453, "step": 57910 }, { "epoch": 3.13, "learning_rate": 3.438818565400844e-05, "loss": 1.4788, "step": 57920 }, { "epoch": 3.13, "learning_rate": 3.438548090446825e-05, "loss": 1.4715, "step": 57930 }, { "epoch": 3.13, "learning_rate": 3.438277615492805e-05, "loss": 1.4482, "step": 57940 }, { "epoch": 3.13, "learning_rate": 3.438007140538786e-05, "loss": 1.5655, "step": 57950 }, { "epoch": 3.13, "learning_rate": 3.437736665584767e-05, "loss": 1.4945, "step": 57960 }, { "epoch": 3.13, "learning_rate": 3.4374661906307474e-05, "loss": 1.4408, "step": 57970 }, { "epoch": 3.13, "learning_rate": 3.437195715676729e-05, "loss": 1.4072, "step": 57980 }, { "epoch": 3.13, "learning_rate": 3.436925240722709e-05, "loss": 1.3889, "step": 57990 }, { "epoch": 3.13, "learning_rate": 3.4366547657686896e-05, "loss": 1.471, "step": 58000 }, { "epoch": 3.13, "learning_rate": 3.436384290814671e-05, "loss": 1.4687, "step": 58010 }, { "epoch": 3.14, "learning_rate": 3.4361138158606516e-05, "loss": 1.3934, "step": 58020 }, { "epoch": 3.14, "learning_rate": 3.435843340906632e-05, "loss": 1.4711, "step": 58030 }, { "epoch": 3.14, "learning_rate": 3.4355728659526135e-05, "loss": 1.4559, "step": 58040 }, { "epoch": 3.14, "learning_rate": 3.435302390998594e-05, "loss": 1.4996, "step": 58050 }, { "epoch": 3.14, "learning_rate": 3.435031916044574e-05, "loss": 1.4682, "step": 58060 }, { "epoch": 3.14, "learning_rate": 3.434761441090556e-05, "loss": 1.4631, "step": 58070 }, { "epoch": 3.14, "learning_rate": 3.434490966136536e-05, "loss": 1.4645, "step": 58080 }, { "epoch": 3.14, "learning_rate": 3.434220491182516e-05, "loss": 1.5176, "step": 58090 }, { "epoch": 3.14, "learning_rate": 3.433950016228498e-05, "loss": 1.409, "step": 58100 }, { "epoch": 3.14, "learning_rate": 3.433679541274478e-05, "loss": 1.4474, "step": 58110 }, { "epoch": 3.14, "learning_rate": 3.4334090663204585e-05, "loss": 1.4133, "step": 58120 }, { "epoch": 3.14, "learning_rate": 3.43313859136644e-05, "loss": 1.4248, "step": 58130 }, { "epoch": 3.14, "learning_rate": 3.4328681164124204e-05, "loss": 1.5461, "step": 58140 }, { "epoch": 3.14, "learning_rate": 3.432597641458401e-05, "loss": 1.4838, "step": 58150 }, { "epoch": 3.14, "learning_rate": 3.4323271665043823e-05, "loss": 1.4868, "step": 58160 }, { "epoch": 3.14, "learning_rate": 3.4320566915503626e-05, "loss": 1.5175, "step": 58170 }, { "epoch": 3.14, "learning_rate": 3.431786216596343e-05, "loss": 1.5357, "step": 58180 }, { "epoch": 3.14, "learning_rate": 3.431515741642324e-05, "loss": 1.5055, "step": 58190 }, { "epoch": 3.14, "learning_rate": 3.431245266688305e-05, "loss": 1.5356, "step": 58200 }, { "epoch": 3.15, "learning_rate": 3.430974791734286e-05, "loss": 1.5315, "step": 58210 }, { "epoch": 3.15, "learning_rate": 3.430704316780266e-05, "loss": 1.5234, "step": 58220 }, { "epoch": 3.15, "learning_rate": 3.430433841826247e-05, "loss": 1.3971, "step": 58230 }, { "epoch": 3.15, "learning_rate": 3.430163366872228e-05, "loss": 1.4716, "step": 58240 }, { "epoch": 3.15, "learning_rate": 3.429892891918208e-05, "loss": 1.4454, "step": 58250 }, { "epoch": 3.15, "learning_rate": 3.429622416964189e-05, "loss": 1.4962, "step": 58260 }, { "epoch": 3.15, "learning_rate": 3.42935194201017e-05, "loss": 1.5109, "step": 58270 }, { "epoch": 3.15, "learning_rate": 3.4290814670561505e-05, "loss": 1.4824, "step": 58280 }, { "epoch": 3.15, "learning_rate": 3.4288109921021315e-05, "loss": 1.4439, "step": 58290 }, { "epoch": 3.15, "learning_rate": 3.4285405171481125e-05, "loss": 1.4862, "step": 58300 }, { "epoch": 3.15, "learning_rate": 3.428270042194093e-05, "loss": 1.5436, "step": 58310 }, { "epoch": 3.15, "learning_rate": 3.427999567240074e-05, "loss": 1.4581, "step": 58320 }, { "epoch": 3.15, "learning_rate": 3.427729092286055e-05, "loss": 1.4736, "step": 58330 }, { "epoch": 3.15, "learning_rate": 3.427458617332035e-05, "loss": 1.4831, "step": 58340 }, { "epoch": 3.15, "learning_rate": 3.427188142378016e-05, "loss": 1.4465, "step": 58350 }, { "epoch": 3.15, "learning_rate": 3.426917667423997e-05, "loss": 1.4588, "step": 58360 }, { "epoch": 3.15, "learning_rate": 3.426647192469977e-05, "loss": 1.5168, "step": 58370 }, { "epoch": 3.15, "learning_rate": 3.426376717515958e-05, "loss": 1.5342, "step": 58380 }, { "epoch": 3.16, "learning_rate": 3.426106242561939e-05, "loss": 1.4652, "step": 58390 }, { "epoch": 3.16, "learning_rate": 3.4258357676079194e-05, "loss": 1.477, "step": 58400 }, { "epoch": 3.16, "learning_rate": 3.4255652926539004e-05, "loss": 1.4733, "step": 58410 }, { "epoch": 3.16, "learning_rate": 3.425294817699881e-05, "loss": 1.5243, "step": 58420 }, { "epoch": 3.16, "learning_rate": 3.4250243427458616e-05, "loss": 1.4519, "step": 58430 }, { "epoch": 3.16, "learning_rate": 3.4247538677918426e-05, "loss": 1.5186, "step": 58440 }, { "epoch": 3.16, "learning_rate": 3.4244833928378236e-05, "loss": 1.4351, "step": 58450 }, { "epoch": 3.16, "learning_rate": 3.424212917883804e-05, "loss": 1.511, "step": 58460 }, { "epoch": 3.16, "learning_rate": 3.423942442929785e-05, "loss": 1.5442, "step": 58470 }, { "epoch": 3.16, "learning_rate": 3.423671967975766e-05, "loss": 1.469, "step": 58480 }, { "epoch": 3.16, "learning_rate": 3.423401493021746e-05, "loss": 1.4867, "step": 58490 }, { "epoch": 3.16, "learning_rate": 3.423131018067727e-05, "loss": 1.4838, "step": 58500 }, { "epoch": 3.16, "learning_rate": 3.422860543113708e-05, "loss": 1.5309, "step": 58510 }, { "epoch": 3.16, "learning_rate": 3.422590068159688e-05, "loss": 1.395, "step": 58520 }, { "epoch": 3.16, "learning_rate": 3.422319593205669e-05, "loss": 1.4488, "step": 58530 }, { "epoch": 3.16, "learning_rate": 3.42204911825165e-05, "loss": 1.4232, "step": 58540 }, { "epoch": 3.16, "learning_rate": 3.4217786432976305e-05, "loss": 1.4975, "step": 58550 }, { "epoch": 3.16, "learning_rate": 3.4215081683436115e-05, "loss": 1.555, "step": 58560 }, { "epoch": 3.16, "learning_rate": 3.4212376933895924e-05, "loss": 1.5538, "step": 58570 }, { "epoch": 3.17, "learning_rate": 3.420967218435573e-05, "loss": 1.4202, "step": 58580 }, { "epoch": 3.17, "learning_rate": 3.420696743481554e-05, "loss": 1.4428, "step": 58590 }, { "epoch": 3.17, "learning_rate": 3.4204262685275346e-05, "loss": 1.5115, "step": 58600 }, { "epoch": 3.17, "learning_rate": 3.420155793573515e-05, "loss": 1.5005, "step": 58610 }, { "epoch": 3.17, "learning_rate": 3.4198853186194966e-05, "loss": 1.4788, "step": 58620 }, { "epoch": 3.17, "learning_rate": 3.419614843665477e-05, "loss": 1.4134, "step": 58630 }, { "epoch": 3.17, "learning_rate": 3.419344368711457e-05, "loss": 1.5027, "step": 58640 }, { "epoch": 3.17, "learning_rate": 3.419073893757439e-05, "loss": 1.5562, "step": 58650 }, { "epoch": 3.17, "learning_rate": 3.418803418803419e-05, "loss": 1.6159, "step": 58660 }, { "epoch": 3.17, "learning_rate": 3.4185329438493994e-05, "loss": 1.4476, "step": 58670 }, { "epoch": 3.17, "learning_rate": 3.418262468895381e-05, "loss": 1.4946, "step": 58680 }, { "epoch": 3.17, "learning_rate": 3.417991993941361e-05, "loss": 1.4488, "step": 58690 }, { "epoch": 3.17, "learning_rate": 3.4177215189873416e-05, "loss": 1.5047, "step": 58700 }, { "epoch": 3.17, "learning_rate": 3.417451044033323e-05, "loss": 1.5481, "step": 58710 }, { "epoch": 3.17, "learning_rate": 3.4171805690793035e-05, "loss": 1.5051, "step": 58720 }, { "epoch": 3.17, "learning_rate": 3.416910094125284e-05, "loss": 1.4336, "step": 58730 }, { "epoch": 3.17, "learning_rate": 3.416639619171265e-05, "loss": 1.5413, "step": 58740 }, { "epoch": 3.17, "learning_rate": 3.416369144217246e-05, "loss": 1.5486, "step": 58750 }, { "epoch": 3.18, "learning_rate": 3.416098669263226e-05, "loss": 1.4812, "step": 58760 }, { "epoch": 3.18, "learning_rate": 3.415828194309207e-05, "loss": 1.4665, "step": 58770 }, { "epoch": 3.18, "learning_rate": 3.415557719355188e-05, "loss": 1.5184, "step": 58780 }, { "epoch": 3.18, "learning_rate": 3.415287244401168e-05, "loss": 1.4539, "step": 58790 }, { "epoch": 3.18, "learning_rate": 3.415016769447149e-05, "loss": 1.43, "step": 58800 }, { "epoch": 3.18, "learning_rate": 3.41474629449313e-05, "loss": 1.4862, "step": 58810 }, { "epoch": 3.18, "learning_rate": 3.4144758195391104e-05, "loss": 1.4815, "step": 58820 }, { "epoch": 3.18, "learning_rate": 3.4142053445850914e-05, "loss": 1.4813, "step": 58830 }, { "epoch": 3.18, "learning_rate": 3.4139348696310724e-05, "loss": 1.4889, "step": 58840 }, { "epoch": 3.18, "learning_rate": 3.413664394677053e-05, "loss": 1.5232, "step": 58850 }, { "epoch": 3.18, "learning_rate": 3.4133939197230336e-05, "loss": 1.5352, "step": 58860 }, { "epoch": 3.18, "learning_rate": 3.4131234447690146e-05, "loss": 1.5786, "step": 58870 }, { "epoch": 3.18, "learning_rate": 3.4128529698149956e-05, "loss": 1.4808, "step": 58880 }, { "epoch": 3.18, "learning_rate": 3.412582494860976e-05, "loss": 1.5374, "step": 58890 }, { "epoch": 3.18, "learning_rate": 3.412312019906957e-05, "loss": 1.5326, "step": 58900 }, { "epoch": 3.18, "learning_rate": 3.412041544952938e-05, "loss": 1.5168, "step": 58910 }, { "epoch": 3.18, "learning_rate": 3.411771069998918e-05, "loss": 1.5501, "step": 58920 }, { "epoch": 3.18, "learning_rate": 3.411500595044899e-05, "loss": 1.6041, "step": 58930 }, { "epoch": 3.18, "learning_rate": 3.41123012009088e-05, "loss": 1.5425, "step": 58940 }, { "epoch": 3.19, "learning_rate": 3.41095964513686e-05, "loss": 1.4982, "step": 58950 }, { "epoch": 3.19, "learning_rate": 3.410689170182841e-05, "loss": 1.4878, "step": 58960 }, { "epoch": 3.19, "learning_rate": 3.410418695228822e-05, "loss": 1.4818, "step": 58970 }, { "epoch": 3.19, "learning_rate": 3.4101482202748025e-05, "loss": 1.4617, "step": 58980 }, { "epoch": 3.19, "learning_rate": 3.4098777453207835e-05, "loss": 1.5121, "step": 58990 }, { "epoch": 3.19, "learning_rate": 3.4096072703667644e-05, "loss": 1.4769, "step": 59000 }, { "epoch": 3.19, "learning_rate": 3.409336795412745e-05, "loss": 1.4476, "step": 59010 }, { "epoch": 3.19, "learning_rate": 3.409066320458726e-05, "loss": 1.4397, "step": 59020 }, { "epoch": 3.19, "learning_rate": 3.4087958455047066e-05, "loss": 1.447, "step": 59030 }, { "epoch": 3.19, "learning_rate": 3.408525370550687e-05, "loss": 1.465, "step": 59040 }, { "epoch": 3.19, "learning_rate": 3.408254895596668e-05, "loss": 1.5196, "step": 59050 }, { "epoch": 3.19, "learning_rate": 3.407984420642649e-05, "loss": 1.492, "step": 59060 }, { "epoch": 3.19, "learning_rate": 3.407713945688629e-05, "loss": 1.4737, "step": 59070 }, { "epoch": 3.19, "learning_rate": 3.40744347073461e-05, "loss": 1.4556, "step": 59080 }, { "epoch": 3.19, "learning_rate": 3.407172995780591e-05, "loss": 1.4827, "step": 59090 }, { "epoch": 3.19, "learning_rate": 3.4069025208265714e-05, "loss": 1.4707, "step": 59100 }, { "epoch": 3.19, "learning_rate": 3.406632045872552e-05, "loss": 1.4973, "step": 59110 }, { "epoch": 3.19, "learning_rate": 3.406361570918533e-05, "loss": 1.526, "step": 59120 }, { "epoch": 3.2, "learning_rate": 3.4060910959645136e-05, "loss": 1.5177, "step": 59130 }, { "epoch": 3.2, "learning_rate": 3.4058206210104945e-05, "loss": 1.4519, "step": 59140 }, { "epoch": 3.2, "learning_rate": 3.4055501460564755e-05, "loss": 1.5461, "step": 59150 }, { "epoch": 3.2, "learning_rate": 3.405279671102456e-05, "loss": 1.4585, "step": 59160 }, { "epoch": 3.2, "learning_rate": 3.405009196148437e-05, "loss": 1.5032, "step": 59170 }, { "epoch": 3.2, "learning_rate": 3.404738721194418e-05, "loss": 1.4924, "step": 59180 }, { "epoch": 3.2, "learning_rate": 3.404468246240398e-05, "loss": 1.4671, "step": 59190 }, { "epoch": 3.2, "learning_rate": 3.404197771286379e-05, "loss": 1.4856, "step": 59200 }, { "epoch": 3.2, "learning_rate": 3.40392729633236e-05, "loss": 1.4365, "step": 59210 }, { "epoch": 3.2, "learning_rate": 3.40365682137834e-05, "loss": 1.5292, "step": 59220 }, { "epoch": 3.2, "learning_rate": 3.403386346424321e-05, "loss": 1.4263, "step": 59230 }, { "epoch": 3.2, "learning_rate": 3.403115871470302e-05, "loss": 1.455, "step": 59240 }, { "epoch": 3.2, "learning_rate": 3.4028453965162824e-05, "loss": 1.4529, "step": 59250 }, { "epoch": 3.2, "learning_rate": 3.402574921562264e-05, "loss": 1.5298, "step": 59260 }, { "epoch": 3.2, "learning_rate": 3.4023044466082444e-05, "loss": 1.4737, "step": 59270 }, { "epoch": 3.2, "learning_rate": 3.402033971654225e-05, "loss": 1.4341, "step": 59280 }, { "epoch": 3.2, "learning_rate": 3.401763496700206e-05, "loss": 1.4494, "step": 59290 }, { "epoch": 3.2, "learning_rate": 3.4014930217461866e-05, "loss": 1.4644, "step": 59300 }, { "epoch": 3.2, "learning_rate": 3.401222546792167e-05, "loss": 1.5064, "step": 59310 }, { "epoch": 3.21, "learning_rate": 3.400952071838148e-05, "loss": 1.432, "step": 59320 }, { "epoch": 3.21, "learning_rate": 3.400681596884129e-05, "loss": 1.4947, "step": 59330 }, { "epoch": 3.21, "learning_rate": 3.400411121930109e-05, "loss": 1.4589, "step": 59340 }, { "epoch": 3.21, "learning_rate": 3.40014064697609e-05, "loss": 1.5395, "step": 59350 }, { "epoch": 3.21, "learning_rate": 3.399870172022071e-05, "loss": 1.4796, "step": 59360 }, { "epoch": 3.21, "learning_rate": 3.399599697068051e-05, "loss": 1.4947, "step": 59370 }, { "epoch": 3.21, "learning_rate": 3.399329222114032e-05, "loss": 1.4416, "step": 59380 }, { "epoch": 3.21, "learning_rate": 3.399058747160013e-05, "loss": 1.5677, "step": 59390 }, { "epoch": 3.21, "learning_rate": 3.3987882722059935e-05, "loss": 1.5591, "step": 59400 }, { "epoch": 3.21, "learning_rate": 3.3985177972519745e-05, "loss": 1.5392, "step": 59410 }, { "epoch": 3.21, "learning_rate": 3.3982473222979555e-05, "loss": 1.5627, "step": 59420 }, { "epoch": 3.21, "learning_rate": 3.397976847343936e-05, "loss": 1.4865, "step": 59430 }, { "epoch": 3.21, "learning_rate": 3.397706372389917e-05, "loss": 1.5171, "step": 59440 }, { "epoch": 3.21, "learning_rate": 3.397435897435898e-05, "loss": 1.4789, "step": 59450 }, { "epoch": 3.21, "learning_rate": 3.397165422481878e-05, "loss": 1.4522, "step": 59460 }, { "epoch": 3.21, "learning_rate": 3.396894947527859e-05, "loss": 1.4727, "step": 59470 }, { "epoch": 3.21, "learning_rate": 3.39662447257384e-05, "loss": 1.4807, "step": 59480 }, { "epoch": 3.21, "learning_rate": 3.396353997619821e-05, "loss": 1.519, "step": 59490 }, { "epoch": 3.22, "learning_rate": 3.396083522665801e-05, "loss": 1.4856, "step": 59500 }, { "epoch": 3.22, "learning_rate": 3.395813047711782e-05, "loss": 1.5255, "step": 59510 }, { "epoch": 3.22, "learning_rate": 3.395542572757763e-05, "loss": 1.5228, "step": 59520 }, { "epoch": 3.22, "learning_rate": 3.3952720978037434e-05, "loss": 1.5775, "step": 59530 }, { "epoch": 3.22, "learning_rate": 3.395001622849724e-05, "loss": 1.4347, "step": 59540 }, { "epoch": 3.22, "learning_rate": 3.394731147895705e-05, "loss": 1.4973, "step": 59550 }, { "epoch": 3.22, "learning_rate": 3.3944606729416856e-05, "loss": 1.5018, "step": 59560 }, { "epoch": 3.22, "learning_rate": 3.3941901979876665e-05, "loss": 1.5867, "step": 59570 }, { "epoch": 3.22, "learning_rate": 3.3939197230336475e-05, "loss": 1.4291, "step": 59580 }, { "epoch": 3.22, "learning_rate": 3.393649248079628e-05, "loss": 1.4631, "step": 59590 }, { "epoch": 3.22, "learning_rate": 3.393378773125609e-05, "loss": 1.5323, "step": 59600 }, { "epoch": 3.22, "learning_rate": 3.39310829817159e-05, "loss": 1.5158, "step": 59610 }, { "epoch": 3.22, "learning_rate": 3.39283782321757e-05, "loss": 1.5199, "step": 59620 }, { "epoch": 3.22, "learning_rate": 3.392567348263551e-05, "loss": 1.4458, "step": 59630 }, { "epoch": 3.22, "learning_rate": 3.392296873309532e-05, "loss": 1.4911, "step": 59640 }, { "epoch": 3.22, "learning_rate": 3.392026398355512e-05, "loss": 1.5734, "step": 59650 }, { "epoch": 3.22, "learning_rate": 3.391755923401493e-05, "loss": 1.5365, "step": 59660 }, { "epoch": 3.22, "learning_rate": 3.391485448447474e-05, "loss": 1.4711, "step": 59670 }, { "epoch": 3.22, "learning_rate": 3.3912149734934544e-05, "loss": 1.5321, "step": 59680 }, { "epoch": 3.23, "learning_rate": 3.3909444985394354e-05, "loss": 1.4705, "step": 59690 }, { "epoch": 3.23, "learning_rate": 3.3906740235854164e-05, "loss": 1.4753, "step": 59700 }, { "epoch": 3.23, "learning_rate": 3.390403548631397e-05, "loss": 1.5302, "step": 59710 }, { "epoch": 3.23, "learning_rate": 3.3901330736773776e-05, "loss": 1.4882, "step": 59720 }, { "epoch": 3.23, "learning_rate": 3.3898625987233586e-05, "loss": 1.4349, "step": 59730 }, { "epoch": 3.23, "learning_rate": 3.389592123769339e-05, "loss": 1.4966, "step": 59740 }, { "epoch": 3.23, "learning_rate": 3.38932164881532e-05, "loss": 1.4524, "step": 59750 }, { "epoch": 3.23, "learning_rate": 3.389051173861301e-05, "loss": 1.4405, "step": 59760 }, { "epoch": 3.23, "learning_rate": 3.388780698907281e-05, "loss": 1.4583, "step": 59770 }, { "epoch": 3.23, "learning_rate": 3.388510223953262e-05, "loss": 1.4778, "step": 59780 }, { "epoch": 3.23, "learning_rate": 3.388239748999243e-05, "loss": 1.4702, "step": 59790 }, { "epoch": 3.23, "learning_rate": 3.387969274045223e-05, "loss": 1.5133, "step": 59800 }, { "epoch": 3.23, "learning_rate": 3.387698799091204e-05, "loss": 1.4785, "step": 59810 }, { "epoch": 3.23, "learning_rate": 3.387428324137185e-05, "loss": 1.544, "step": 59820 }, { "epoch": 3.23, "learning_rate": 3.3871578491831655e-05, "loss": 1.491, "step": 59830 }, { "epoch": 3.23, "learning_rate": 3.3868873742291465e-05, "loss": 1.5659, "step": 59840 }, { "epoch": 3.23, "learning_rate": 3.3866168992751275e-05, "loss": 1.4794, "step": 59850 }, { "epoch": 3.23, "learning_rate": 3.386346424321108e-05, "loss": 1.4312, "step": 59860 }, { "epoch": 3.24, "learning_rate": 3.386075949367089e-05, "loss": 1.5416, "step": 59870 }, { "epoch": 3.24, "learning_rate": 3.38580547441307e-05, "loss": 1.4592, "step": 59880 }, { "epoch": 3.24, "learning_rate": 3.38553499945905e-05, "loss": 1.469, "step": 59890 }, { "epoch": 3.24, "learning_rate": 3.385264524505031e-05, "loss": 1.5501, "step": 59900 }, { "epoch": 3.24, "learning_rate": 3.384994049551012e-05, "loss": 1.5579, "step": 59910 }, { "epoch": 3.24, "learning_rate": 3.384723574596992e-05, "loss": 1.3452, "step": 59920 }, { "epoch": 3.24, "learning_rate": 3.384453099642973e-05, "loss": 1.4622, "step": 59930 }, { "epoch": 3.24, "learning_rate": 3.384182624688954e-05, "loss": 1.4165, "step": 59940 }, { "epoch": 3.24, "learning_rate": 3.3839121497349344e-05, "loss": 1.5226, "step": 59950 }, { "epoch": 3.24, "learning_rate": 3.3836416747809154e-05, "loss": 1.4712, "step": 59960 }, { "epoch": 3.24, "learning_rate": 3.383371199826896e-05, "loss": 1.5111, "step": 59970 }, { "epoch": 3.24, "learning_rate": 3.3831007248728766e-05, "loss": 1.496, "step": 59980 }, { "epoch": 3.24, "learning_rate": 3.3828302499188576e-05, "loss": 1.4827, "step": 59990 }, { "epoch": 3.24, "learning_rate": 3.3825597749648385e-05, "loss": 1.5109, "step": 60000 }, { "epoch": 3.24, "learning_rate": 3.382289300010819e-05, "loss": 1.5993, "step": 60010 }, { "epoch": 3.24, "learning_rate": 3.3820188250568e-05, "loss": 1.4728, "step": 60020 }, { "epoch": 3.24, "learning_rate": 3.381748350102781e-05, "loss": 1.4891, "step": 60030 }, { "epoch": 3.24, "learning_rate": 3.381477875148761e-05, "loss": 1.495, "step": 60040 }, { "epoch": 3.24, "learning_rate": 3.381207400194742e-05, "loss": 1.5281, "step": 60050 }, { "epoch": 3.25, "learning_rate": 3.380936925240723e-05, "loss": 1.5189, "step": 60060 }, { "epoch": 3.25, "learning_rate": 3.380666450286703e-05, "loss": 1.4951, "step": 60070 }, { "epoch": 3.25, "learning_rate": 3.380395975332684e-05, "loss": 1.4745, "step": 60080 }, { "epoch": 3.25, "learning_rate": 3.380125500378665e-05, "loss": 1.4739, "step": 60090 }, { "epoch": 3.25, "learning_rate": 3.379855025424646e-05, "loss": 1.5325, "step": 60100 }, { "epoch": 3.25, "learning_rate": 3.3795845504706264e-05, "loss": 1.4764, "step": 60110 }, { "epoch": 3.25, "learning_rate": 3.3793140755166074e-05, "loss": 1.5046, "step": 60120 }, { "epoch": 3.25, "learning_rate": 3.3790436005625884e-05, "loss": 1.5463, "step": 60130 }, { "epoch": 3.25, "learning_rate": 3.378773125608569e-05, "loss": 1.5144, "step": 60140 }, { "epoch": 3.25, "learning_rate": 3.3785026506545496e-05, "loss": 1.4052, "step": 60150 }, { "epoch": 3.25, "learning_rate": 3.3782321757005306e-05, "loss": 1.4614, "step": 60160 }, { "epoch": 3.25, "learning_rate": 3.377961700746511e-05, "loss": 1.5383, "step": 60170 }, { "epoch": 3.25, "learning_rate": 3.377691225792492e-05, "loss": 1.5421, "step": 60180 }, { "epoch": 3.25, "learning_rate": 3.377420750838473e-05, "loss": 1.4778, "step": 60190 }, { "epoch": 3.25, "learning_rate": 3.377150275884453e-05, "loss": 1.529, "step": 60200 }, { "epoch": 3.25, "learning_rate": 3.376879800930434e-05, "loss": 1.4964, "step": 60210 }, { "epoch": 3.25, "learning_rate": 3.376609325976415e-05, "loss": 1.4757, "step": 60220 }, { "epoch": 3.25, "learning_rate": 3.376338851022395e-05, "loss": 1.5408, "step": 60230 }, { "epoch": 3.26, "learning_rate": 3.376068376068376e-05, "loss": 1.4489, "step": 60240 }, { "epoch": 3.26, "learning_rate": 3.375797901114357e-05, "loss": 1.4731, "step": 60250 }, { "epoch": 3.26, "learning_rate": 3.3755274261603375e-05, "loss": 1.5138, "step": 60260 }, { "epoch": 3.26, "learning_rate": 3.3752569512063185e-05, "loss": 1.4233, "step": 60270 }, { "epoch": 3.26, "learning_rate": 3.3749864762522995e-05, "loss": 1.5103, "step": 60280 }, { "epoch": 3.26, "learning_rate": 3.37471600129828e-05, "loss": 1.519, "step": 60290 }, { "epoch": 3.26, "learning_rate": 3.374445526344261e-05, "loss": 1.4833, "step": 60300 }, { "epoch": 3.26, "learning_rate": 3.374175051390242e-05, "loss": 1.4525, "step": 60310 }, { "epoch": 3.26, "learning_rate": 3.373904576436222e-05, "loss": 1.4743, "step": 60320 }, { "epoch": 3.26, "learning_rate": 3.373634101482203e-05, "loss": 1.4788, "step": 60330 }, { "epoch": 3.26, "learning_rate": 3.373363626528184e-05, "loss": 1.4341, "step": 60340 }, { "epoch": 3.26, "learning_rate": 3.373093151574164e-05, "loss": 1.5336, "step": 60350 }, { "epoch": 3.26, "learning_rate": 3.372822676620145e-05, "loss": 1.4881, "step": 60360 }, { "epoch": 3.26, "learning_rate": 3.372552201666126e-05, "loss": 1.4445, "step": 60370 }, { "epoch": 3.26, "learning_rate": 3.3722817267121064e-05, "loss": 1.5268, "step": 60380 }, { "epoch": 3.26, "learning_rate": 3.3720112517580874e-05, "loss": 1.367, "step": 60390 }, { "epoch": 3.26, "learning_rate": 3.371740776804068e-05, "loss": 1.4748, "step": 60400 }, { "epoch": 3.26, "learning_rate": 3.3714703018500486e-05, "loss": 1.4959, "step": 60410 }, { "epoch": 3.26, "learning_rate": 3.3711998268960296e-05, "loss": 1.4888, "step": 60420 }, { "epoch": 3.27, "learning_rate": 3.3709293519420105e-05, "loss": 1.5028, "step": 60430 }, { "epoch": 3.27, "learning_rate": 3.370658876987991e-05, "loss": 1.445, "step": 60440 }, { "epoch": 3.27, "learning_rate": 3.370388402033972e-05, "loss": 1.6161, "step": 60450 }, { "epoch": 3.27, "learning_rate": 3.370117927079953e-05, "loss": 1.5703, "step": 60460 }, { "epoch": 3.27, "learning_rate": 3.369847452125933e-05, "loss": 1.4627, "step": 60470 }, { "epoch": 3.27, "learning_rate": 3.369576977171914e-05, "loss": 1.549, "step": 60480 }, { "epoch": 3.27, "learning_rate": 3.369306502217895e-05, "loss": 1.4925, "step": 60490 }, { "epoch": 3.27, "learning_rate": 3.369036027263875e-05, "loss": 1.5436, "step": 60500 }, { "epoch": 3.27, "learning_rate": 3.368765552309856e-05, "loss": 1.491, "step": 60510 }, { "epoch": 3.27, "learning_rate": 3.368495077355837e-05, "loss": 1.4953, "step": 60520 }, { "epoch": 3.27, "learning_rate": 3.3682246024018175e-05, "loss": 1.5591, "step": 60530 }, { "epoch": 3.27, "learning_rate": 3.3679541274477985e-05, "loss": 1.538, "step": 60540 }, { "epoch": 3.27, "learning_rate": 3.3676836524937794e-05, "loss": 1.3806, "step": 60550 }, { "epoch": 3.27, "learning_rate": 3.36741317753976e-05, "loss": 1.5182, "step": 60560 }, { "epoch": 3.27, "learning_rate": 3.367142702585741e-05, "loss": 1.4855, "step": 60570 }, { "epoch": 3.27, "learning_rate": 3.3668722276317216e-05, "loss": 1.4971, "step": 60580 }, { "epoch": 3.27, "learning_rate": 3.366601752677702e-05, "loss": 1.4256, "step": 60590 }, { "epoch": 3.27, "learning_rate": 3.366331277723683e-05, "loss": 1.4416, "step": 60600 }, { "epoch": 3.28, "learning_rate": 3.366060802769664e-05, "loss": 1.4645, "step": 60610 }, { "epoch": 3.28, "learning_rate": 3.365790327815644e-05, "loss": 1.4584, "step": 60620 }, { "epoch": 3.28, "learning_rate": 3.365519852861625e-05, "loss": 1.4606, "step": 60630 }, { "epoch": 3.28, "learning_rate": 3.365249377907606e-05, "loss": 1.5633, "step": 60640 }, { "epoch": 3.28, "learning_rate": 3.3649789029535864e-05, "loss": 1.5384, "step": 60650 }, { "epoch": 3.28, "learning_rate": 3.364708427999567e-05, "loss": 1.4592, "step": 60660 }, { "epoch": 3.28, "learning_rate": 3.364437953045548e-05, "loss": 1.4806, "step": 60670 }, { "epoch": 3.28, "learning_rate": 3.3641674780915286e-05, "loss": 1.4508, "step": 60680 }, { "epoch": 3.28, "learning_rate": 3.3638970031375095e-05, "loss": 1.5536, "step": 60690 }, { "epoch": 3.28, "learning_rate": 3.3636265281834905e-05, "loss": 1.4381, "step": 60700 }, { "epoch": 3.28, "learning_rate": 3.363356053229471e-05, "loss": 1.5172, "step": 60710 }, { "epoch": 3.28, "learning_rate": 3.363085578275452e-05, "loss": 1.4455, "step": 60720 }, { "epoch": 3.28, "learning_rate": 3.362815103321433e-05, "loss": 1.5582, "step": 60730 }, { "epoch": 3.28, "learning_rate": 3.362544628367414e-05, "loss": 1.3897, "step": 60740 }, { "epoch": 3.28, "learning_rate": 3.362274153413394e-05, "loss": 1.4995, "step": 60750 }, { "epoch": 3.28, "learning_rate": 3.362003678459375e-05, "loss": 1.4769, "step": 60760 }, { "epoch": 3.28, "learning_rate": 3.361733203505356e-05, "loss": 1.5178, "step": 60770 }, { "epoch": 3.28, "learning_rate": 3.361462728551336e-05, "loss": 1.5619, "step": 60780 }, { "epoch": 3.28, "learning_rate": 3.361192253597317e-05, "loss": 1.5003, "step": 60790 }, { "epoch": 3.29, "learning_rate": 3.360921778643298e-05, "loss": 1.5474, "step": 60800 }, { "epoch": 3.29, "learning_rate": 3.3606513036892784e-05, "loss": 1.5352, "step": 60810 }, { "epoch": 3.29, "learning_rate": 3.3603808287352594e-05, "loss": 1.5331, "step": 60820 }, { "epoch": 3.29, "learning_rate": 3.36011035378124e-05, "loss": 1.4458, "step": 60830 }, { "epoch": 3.29, "learning_rate": 3.3598398788272206e-05, "loss": 1.4789, "step": 60840 }, { "epoch": 3.29, "learning_rate": 3.3595694038732016e-05, "loss": 1.56, "step": 60850 }, { "epoch": 3.29, "learning_rate": 3.3592989289191826e-05, "loss": 1.5975, "step": 60860 }, { "epoch": 3.29, "learning_rate": 3.359028453965163e-05, "loss": 1.4375, "step": 60870 }, { "epoch": 3.29, "learning_rate": 3.358757979011144e-05, "loss": 1.5114, "step": 60880 }, { "epoch": 3.29, "learning_rate": 3.358487504057125e-05, "loss": 1.4863, "step": 60890 }, { "epoch": 3.29, "learning_rate": 3.358217029103105e-05, "loss": 1.5104, "step": 60900 }, { "epoch": 3.29, "learning_rate": 3.357946554149086e-05, "loss": 1.5376, "step": 60910 }, { "epoch": 3.29, "learning_rate": 3.357676079195067e-05, "loss": 1.4806, "step": 60920 }, { "epoch": 3.29, "learning_rate": 3.357405604241047e-05, "loss": 1.5552, "step": 60930 }, { "epoch": 3.29, "learning_rate": 3.357135129287028e-05, "loss": 1.4857, "step": 60940 }, { "epoch": 3.29, "learning_rate": 3.356864654333009e-05, "loss": 1.4671, "step": 60950 }, { "epoch": 3.29, "learning_rate": 3.3565941793789895e-05, "loss": 1.4765, "step": 60960 }, { "epoch": 3.29, "learning_rate": 3.3563237044249705e-05, "loss": 1.4287, "step": 60970 }, { "epoch": 3.3, "learning_rate": 3.3560532294709514e-05, "loss": 1.6297, "step": 60980 }, { "epoch": 3.3, "learning_rate": 3.355782754516932e-05, "loss": 1.4476, "step": 60990 }, { "epoch": 3.3, "learning_rate": 3.355512279562913e-05, "loss": 1.4357, "step": 61000 }, { "epoch": 3.3, "learning_rate": 3.3552418046088936e-05, "loss": 1.4444, "step": 61010 }, { "epoch": 3.3, "learning_rate": 3.354971329654874e-05, "loss": 1.4749, "step": 61020 }, { "epoch": 3.3, "learning_rate": 3.354700854700855e-05, "loss": 1.5253, "step": 61030 }, { "epoch": 3.3, "learning_rate": 3.354430379746836e-05, "loss": 1.478, "step": 61040 }, { "epoch": 3.3, "learning_rate": 3.354159904792816e-05, "loss": 1.5138, "step": 61050 }, { "epoch": 3.3, "learning_rate": 3.353889429838797e-05, "loss": 1.4406, "step": 61060 }, { "epoch": 3.3, "learning_rate": 3.353618954884778e-05, "loss": 1.5119, "step": 61070 }, { "epoch": 3.3, "learning_rate": 3.3533484799307584e-05, "loss": 1.4899, "step": 61080 }, { "epoch": 3.3, "learning_rate": 3.353078004976739e-05, "loss": 1.457, "step": 61090 }, { "epoch": 3.3, "learning_rate": 3.35280753002272e-05, "loss": 1.549, "step": 61100 }, { "epoch": 3.3, "learning_rate": 3.3525370550687006e-05, "loss": 1.4613, "step": 61110 }, { "epoch": 3.3, "learning_rate": 3.3522665801146815e-05, "loss": 1.4787, "step": 61120 }, { "epoch": 3.3, "learning_rate": 3.3519961051606625e-05, "loss": 1.5144, "step": 61130 }, { "epoch": 3.3, "learning_rate": 3.351725630206643e-05, "loss": 1.4654, "step": 61140 }, { "epoch": 3.3, "learning_rate": 3.351455155252624e-05, "loss": 1.5048, "step": 61150 }, { "epoch": 3.3, "learning_rate": 3.351184680298605e-05, "loss": 1.505, "step": 61160 }, { "epoch": 3.31, "learning_rate": 3.350914205344585e-05, "loss": 1.4742, "step": 61170 }, { "epoch": 3.31, "learning_rate": 3.350643730390566e-05, "loss": 1.5388, "step": 61180 }, { "epoch": 3.31, "learning_rate": 3.350373255436547e-05, "loss": 1.5446, "step": 61190 }, { "epoch": 3.31, "learning_rate": 3.350102780482527e-05, "loss": 1.5247, "step": 61200 }, { "epoch": 3.31, "learning_rate": 3.349832305528508e-05, "loss": 1.4469, "step": 61210 }, { "epoch": 3.31, "learning_rate": 3.349561830574489e-05, "loss": 1.4225, "step": 61220 }, { "epoch": 3.31, "learning_rate": 3.3492913556204694e-05, "loss": 1.5279, "step": 61230 }, { "epoch": 3.31, "learning_rate": 3.3490208806664504e-05, "loss": 1.4942, "step": 61240 }, { "epoch": 3.31, "learning_rate": 3.3487504057124314e-05, "loss": 1.4472, "step": 61250 }, { "epoch": 3.31, "learning_rate": 3.3484799307584117e-05, "loss": 1.4592, "step": 61260 }, { "epoch": 3.31, "learning_rate": 3.3482094558043926e-05, "loss": 1.5009, "step": 61270 }, { "epoch": 3.31, "learning_rate": 3.3479389808503736e-05, "loss": 1.4312, "step": 61280 }, { "epoch": 3.31, "learning_rate": 3.347668505896354e-05, "loss": 1.485, "step": 61290 }, { "epoch": 3.31, "learning_rate": 3.347398030942335e-05, "loss": 1.5492, "step": 61300 }, { "epoch": 3.31, "learning_rate": 3.347127555988316e-05, "loss": 1.5471, "step": 61310 }, { "epoch": 3.31, "learning_rate": 3.346857081034296e-05, "loss": 1.5286, "step": 61320 }, { "epoch": 3.31, "learning_rate": 3.346586606080277e-05, "loss": 1.4752, "step": 61330 }, { "epoch": 3.31, "learning_rate": 3.346316131126258e-05, "loss": 1.4444, "step": 61340 }, { "epoch": 3.32, "learning_rate": 3.346045656172238e-05, "loss": 1.5136, "step": 61350 }, { "epoch": 3.32, "learning_rate": 3.345775181218219e-05, "loss": 1.4954, "step": 61360 }, { "epoch": 3.32, "learning_rate": 3.3455047062642e-05, "loss": 1.4911, "step": 61370 }, { "epoch": 3.32, "learning_rate": 3.345234231310181e-05, "loss": 1.5225, "step": 61380 }, { "epoch": 3.32, "learning_rate": 3.3449637563561615e-05, "loss": 1.642, "step": 61390 }, { "epoch": 3.32, "learning_rate": 3.3446932814021425e-05, "loss": 1.4049, "step": 61400 }, { "epoch": 3.32, "learning_rate": 3.3444228064481234e-05, "loss": 1.4124, "step": 61410 }, { "epoch": 3.32, "learning_rate": 3.344152331494104e-05, "loss": 1.5212, "step": 61420 }, { "epoch": 3.32, "learning_rate": 3.343881856540085e-05, "loss": 1.4629, "step": 61430 }, { "epoch": 3.32, "learning_rate": 3.3436113815860656e-05, "loss": 1.4743, "step": 61440 }, { "epoch": 3.32, "learning_rate": 3.343340906632046e-05, "loss": 1.5452, "step": 61450 }, { "epoch": 3.32, "learning_rate": 3.343070431678027e-05, "loss": 1.4768, "step": 61460 }, { "epoch": 3.32, "learning_rate": 3.342799956724008e-05, "loss": 1.5542, "step": 61470 }, { "epoch": 3.32, "learning_rate": 3.342529481769988e-05, "loss": 1.5243, "step": 61480 }, { "epoch": 3.32, "learning_rate": 3.342259006815969e-05, "loss": 1.4626, "step": 61490 }, { "epoch": 3.32, "learning_rate": 3.34198853186195e-05, "loss": 1.474, "step": 61500 }, { "epoch": 3.32, "learning_rate": 3.3417180569079304e-05, "loss": 1.4852, "step": 61510 }, { "epoch": 3.32, "learning_rate": 3.341447581953911e-05, "loss": 1.4792, "step": 61520 }, { "epoch": 3.32, "learning_rate": 3.341177106999892e-05, "loss": 1.5157, "step": 61530 }, { "epoch": 3.33, "learning_rate": 3.3409066320458726e-05, "loss": 1.5157, "step": 61540 }, { "epoch": 3.33, "learning_rate": 3.340636157091853e-05, "loss": 1.582, "step": 61550 }, { "epoch": 3.33, "learning_rate": 3.3403656821378345e-05, "loss": 1.4307, "step": 61560 }, { "epoch": 3.33, "learning_rate": 3.340095207183815e-05, "loss": 1.4948, "step": 61570 }, { "epoch": 3.33, "learning_rate": 3.339824732229795e-05, "loss": 1.5073, "step": 61580 }, { "epoch": 3.33, "learning_rate": 3.339554257275777e-05, "loss": 1.5235, "step": 61590 }, { "epoch": 3.33, "learning_rate": 3.339283782321757e-05, "loss": 1.5406, "step": 61600 }, { "epoch": 3.33, "learning_rate": 3.339013307367738e-05, "loss": 1.5471, "step": 61610 }, { "epoch": 3.33, "learning_rate": 3.338742832413719e-05, "loss": 1.4274, "step": 61620 }, { "epoch": 3.33, "learning_rate": 3.338472357459699e-05, "loss": 1.539, "step": 61630 }, { "epoch": 3.33, "learning_rate": 3.33820188250568e-05, "loss": 1.4438, "step": 61640 }, { "epoch": 3.33, "learning_rate": 3.337931407551661e-05, "loss": 1.461, "step": 61650 }, { "epoch": 3.33, "learning_rate": 3.3376609325976414e-05, "loss": 1.4605, "step": 61660 }, { "epoch": 3.33, "learning_rate": 3.3373904576436224e-05, "loss": 1.379, "step": 61670 }, { "epoch": 3.33, "learning_rate": 3.3371199826896034e-05, "loss": 1.5988, "step": 61680 }, { "epoch": 3.33, "learning_rate": 3.3368495077355837e-05, "loss": 1.4875, "step": 61690 }, { "epoch": 3.33, "learning_rate": 3.3365790327815646e-05, "loss": 1.4995, "step": 61700 }, { "epoch": 3.33, "learning_rate": 3.3363085578275456e-05, "loss": 1.5322, "step": 61710 }, { "epoch": 3.33, "learning_rate": 3.336038082873526e-05, "loss": 1.4557, "step": 61720 }, { "epoch": 3.34, "learning_rate": 3.335767607919507e-05, "loss": 1.484, "step": 61730 }, { "epoch": 3.34, "learning_rate": 3.335497132965488e-05, "loss": 1.5079, "step": 61740 }, { "epoch": 3.34, "learning_rate": 3.335226658011468e-05, "loss": 1.4573, "step": 61750 }, { "epoch": 3.34, "learning_rate": 3.334956183057449e-05, "loss": 1.4974, "step": 61760 }, { "epoch": 3.34, "learning_rate": 3.33468570810343e-05, "loss": 1.5296, "step": 61770 }, { "epoch": 3.34, "learning_rate": 3.33441523314941e-05, "loss": 1.4387, "step": 61780 }, { "epoch": 3.34, "learning_rate": 3.334144758195391e-05, "loss": 1.5194, "step": 61790 }, { "epoch": 3.34, "learning_rate": 3.333874283241372e-05, "loss": 1.4875, "step": 61800 }, { "epoch": 3.34, "learning_rate": 3.3336038082873525e-05, "loss": 1.5132, "step": 61810 }, { "epoch": 3.34, "learning_rate": 3.3333333333333335e-05, "loss": 1.4838, "step": 61820 }, { "epoch": 3.34, "learning_rate": 3.3330628583793145e-05, "loss": 1.5483, "step": 61830 }, { "epoch": 3.34, "learning_rate": 3.332792383425295e-05, "loss": 1.53, "step": 61840 }, { "epoch": 3.34, "learning_rate": 3.332521908471276e-05, "loss": 1.4926, "step": 61850 }, { "epoch": 3.34, "learning_rate": 3.332251433517257e-05, "loss": 1.5106, "step": 61860 }, { "epoch": 3.34, "learning_rate": 3.331980958563237e-05, "loss": 1.511, "step": 61870 }, { "epoch": 3.34, "learning_rate": 3.331710483609218e-05, "loss": 1.511, "step": 61880 }, { "epoch": 3.34, "learning_rate": 3.331440008655199e-05, "loss": 1.5161, "step": 61890 }, { "epoch": 3.34, "learning_rate": 3.331169533701179e-05, "loss": 1.5332, "step": 61900 }, { "epoch": 3.35, "learning_rate": 3.33089905874716e-05, "loss": 1.4937, "step": 61910 }, { "epoch": 3.35, "learning_rate": 3.330628583793141e-05, "loss": 1.5334, "step": 61920 }, { "epoch": 3.35, "learning_rate": 3.3303581088391214e-05, "loss": 1.455, "step": 61930 }, { "epoch": 3.35, "learning_rate": 3.3300876338851024e-05, "loss": 1.4324, "step": 61940 }, { "epoch": 3.35, "learning_rate": 3.329817158931083e-05, "loss": 1.488, "step": 61950 }, { "epoch": 3.35, "learning_rate": 3.3295466839770636e-05, "loss": 1.4365, "step": 61960 }, { "epoch": 3.35, "learning_rate": 3.3292762090230446e-05, "loss": 1.4953, "step": 61970 }, { "epoch": 3.35, "learning_rate": 3.3290057340690255e-05, "loss": 1.4818, "step": 61980 }, { "epoch": 3.35, "learning_rate": 3.328735259115006e-05, "loss": 1.5011, "step": 61990 }, { "epoch": 3.35, "learning_rate": 3.328464784160987e-05, "loss": 1.5356, "step": 62000 }, { "epoch": 3.35, "learning_rate": 3.328194309206968e-05, "loss": 1.5292, "step": 62010 }, { "epoch": 3.35, "learning_rate": 3.327923834252949e-05, "loss": 1.5816, "step": 62020 }, { "epoch": 3.35, "learning_rate": 3.327653359298929e-05, "loss": 1.4527, "step": 62030 }, { "epoch": 3.35, "learning_rate": 3.32738288434491e-05, "loss": 1.5382, "step": 62040 }, { "epoch": 3.35, "learning_rate": 3.327112409390891e-05, "loss": 1.5556, "step": 62050 }, { "epoch": 3.35, "learning_rate": 3.326841934436871e-05, "loss": 1.4472, "step": 62060 }, { "epoch": 3.35, "learning_rate": 3.326571459482852e-05, "loss": 1.4526, "step": 62070 }, { "epoch": 3.35, "learning_rate": 3.326300984528833e-05, "loss": 1.5224, "step": 62080 }, { "epoch": 3.35, "learning_rate": 3.3260305095748134e-05, "loss": 1.4537, "step": 62090 }, { "epoch": 3.36, "learning_rate": 3.3257600346207944e-05, "loss": 1.539, "step": 62100 }, { "epoch": 3.36, "learning_rate": 3.3254895596667754e-05, "loss": 1.4713, "step": 62110 }, { "epoch": 3.36, "learning_rate": 3.325219084712756e-05, "loss": 1.4828, "step": 62120 }, { "epoch": 3.36, "learning_rate": 3.324948609758736e-05, "loss": 1.468, "step": 62130 }, { "epoch": 3.36, "learning_rate": 3.3246781348047176e-05, "loss": 1.553, "step": 62140 }, { "epoch": 3.36, "learning_rate": 3.324407659850698e-05, "loss": 1.4856, "step": 62150 }, { "epoch": 3.36, "learning_rate": 3.324137184896678e-05, "loss": 1.5504, "step": 62160 }, { "epoch": 3.36, "learning_rate": 3.32386670994266e-05, "loss": 1.5385, "step": 62170 }, { "epoch": 3.36, "learning_rate": 3.32359623498864e-05, "loss": 1.3939, "step": 62180 }, { "epoch": 3.36, "learning_rate": 3.3233257600346204e-05, "loss": 1.5683, "step": 62190 }, { "epoch": 3.36, "learning_rate": 3.323055285080602e-05, "loss": 1.4907, "step": 62200 }, { "epoch": 3.36, "learning_rate": 3.322784810126582e-05, "loss": 1.5849, "step": 62210 }, { "epoch": 3.36, "learning_rate": 3.322514335172563e-05, "loss": 1.5972, "step": 62220 }, { "epoch": 3.36, "learning_rate": 3.322243860218544e-05, "loss": 1.4623, "step": 62230 }, { "epoch": 3.36, "learning_rate": 3.3219733852645245e-05, "loss": 1.4825, "step": 62240 }, { "epoch": 3.36, "learning_rate": 3.3217029103105055e-05, "loss": 1.4079, "step": 62250 }, { "epoch": 3.36, "learning_rate": 3.3214324353564865e-05, "loss": 1.4803, "step": 62260 }, { "epoch": 3.36, "learning_rate": 3.321161960402467e-05, "loss": 1.412, "step": 62270 }, { "epoch": 3.37, "learning_rate": 3.320891485448448e-05, "loss": 1.442, "step": 62280 }, { "epoch": 3.37, "learning_rate": 3.320621010494429e-05, "loss": 1.4426, "step": 62290 }, { "epoch": 3.37, "learning_rate": 3.320350535540409e-05, "loss": 1.5464, "step": 62300 }, { "epoch": 3.37, "learning_rate": 3.32008006058639e-05, "loss": 1.5218, "step": 62310 }, { "epoch": 3.37, "learning_rate": 3.319809585632371e-05, "loss": 1.5238, "step": 62320 }, { "epoch": 3.37, "learning_rate": 3.319539110678351e-05, "loss": 1.5457, "step": 62330 }, { "epoch": 3.37, "learning_rate": 3.319268635724332e-05, "loss": 1.5153, "step": 62340 }, { "epoch": 3.37, "learning_rate": 3.318998160770313e-05, "loss": 1.4671, "step": 62350 }, { "epoch": 3.37, "learning_rate": 3.3187276858162934e-05, "loss": 1.5517, "step": 62360 }, { "epoch": 3.37, "learning_rate": 3.3184572108622744e-05, "loss": 1.4925, "step": 62370 }, { "epoch": 3.37, "learning_rate": 3.318186735908255e-05, "loss": 1.4517, "step": 62380 }, { "epoch": 3.37, "learning_rate": 3.3179162609542356e-05, "loss": 1.5277, "step": 62390 }, { "epoch": 3.37, "learning_rate": 3.3176457860002166e-05, "loss": 1.5433, "step": 62400 }, { "epoch": 3.37, "learning_rate": 3.3173753110461975e-05, "loss": 1.5066, "step": 62410 }, { "epoch": 3.37, "learning_rate": 3.317104836092178e-05, "loss": 1.5169, "step": 62420 }, { "epoch": 3.37, "learning_rate": 3.316834361138159e-05, "loss": 1.4969, "step": 62430 }, { "epoch": 3.37, "learning_rate": 3.31656388618414e-05, "loss": 1.5062, "step": 62440 }, { "epoch": 3.37, "learning_rate": 3.31629341123012e-05, "loss": 1.4614, "step": 62450 }, { "epoch": 3.37, "learning_rate": 3.316022936276101e-05, "loss": 1.5408, "step": 62460 }, { "epoch": 3.38, "learning_rate": 3.315752461322082e-05, "loss": 1.5044, "step": 62470 }, { "epoch": 3.38, "learning_rate": 3.315481986368062e-05, "loss": 1.5072, "step": 62480 }, { "epoch": 3.38, "learning_rate": 3.315211511414043e-05, "loss": 1.5276, "step": 62490 }, { "epoch": 3.38, "learning_rate": 3.314941036460024e-05, "loss": 1.5229, "step": 62500 }, { "epoch": 3.38, "learning_rate": 3.3146705615060045e-05, "loss": 1.4514, "step": 62510 }, { "epoch": 3.38, "learning_rate": 3.3144000865519854e-05, "loss": 1.5035, "step": 62520 }, { "epoch": 3.38, "learning_rate": 3.3141296115979664e-05, "loss": 1.4652, "step": 62530 }, { "epoch": 3.38, "learning_rate": 3.313859136643947e-05, "loss": 1.5031, "step": 62540 }, { "epoch": 3.38, "learning_rate": 3.313588661689928e-05, "loss": 1.4125, "step": 62550 }, { "epoch": 3.38, "learning_rate": 3.3133181867359086e-05, "loss": 1.5148, "step": 62560 }, { "epoch": 3.38, "learning_rate": 3.313047711781889e-05, "loss": 1.4794, "step": 62570 }, { "epoch": 3.38, "learning_rate": 3.31277723682787e-05, "loss": 1.4547, "step": 62580 }, { "epoch": 3.38, "learning_rate": 3.312506761873851e-05, "loss": 1.5197, "step": 62590 }, { "epoch": 3.38, "learning_rate": 3.312236286919831e-05, "loss": 1.6089, "step": 62600 }, { "epoch": 3.38, "learning_rate": 3.311965811965812e-05, "loss": 1.5497, "step": 62610 }, { "epoch": 3.38, "learning_rate": 3.311695337011793e-05, "loss": 1.4228, "step": 62620 }, { "epoch": 3.38, "learning_rate": 3.311424862057774e-05, "loss": 1.4319, "step": 62630 }, { "epoch": 3.38, "learning_rate": 3.311154387103754e-05, "loss": 1.5305, "step": 62640 }, { "epoch": 3.39, "learning_rate": 3.310883912149735e-05, "loss": 1.5373, "step": 62650 }, { "epoch": 3.39, "learning_rate": 3.310613437195716e-05, "loss": 1.485, "step": 62660 }, { "epoch": 3.39, "learning_rate": 3.3103429622416965e-05, "loss": 1.4674, "step": 62670 }, { "epoch": 3.39, "learning_rate": 3.310072487287677e-05, "loss": 1.5193, "step": 62680 }, { "epoch": 3.39, "learning_rate": 3.3098020123336585e-05, "loss": 1.508, "step": 62690 }, { "epoch": 3.39, "learning_rate": 3.309531537379639e-05, "loss": 1.4559, "step": 62700 }, { "epoch": 3.39, "learning_rate": 3.309261062425619e-05, "loss": 1.4521, "step": 62710 }, { "epoch": 3.39, "learning_rate": 3.308990587471601e-05, "loss": 1.4499, "step": 62720 }, { "epoch": 3.39, "learning_rate": 3.308720112517581e-05, "loss": 1.5149, "step": 62730 }, { "epoch": 3.39, "learning_rate": 3.308449637563561e-05, "loss": 1.4169, "step": 62740 }, { "epoch": 3.39, "learning_rate": 3.308179162609543e-05, "loss": 1.4931, "step": 62750 }, { "epoch": 3.39, "learning_rate": 3.307908687655523e-05, "loss": 1.4625, "step": 62760 }, { "epoch": 3.39, "learning_rate": 3.3076382127015035e-05, "loss": 1.5189, "step": 62770 }, { "epoch": 3.39, "learning_rate": 3.307367737747485e-05, "loss": 1.4787, "step": 62780 }, { "epoch": 3.39, "learning_rate": 3.3070972627934654e-05, "loss": 1.4949, "step": 62790 }, { "epoch": 3.39, "learning_rate": 3.306826787839446e-05, "loss": 1.502, "step": 62800 }, { "epoch": 3.39, "learning_rate": 3.306556312885427e-05, "loss": 1.5242, "step": 62810 }, { "epoch": 3.39, "learning_rate": 3.3062858379314076e-05, "loss": 1.4385, "step": 62820 }, { "epoch": 3.39, "learning_rate": 3.306015362977388e-05, "loss": 1.4981, "step": 62830 }, { "epoch": 3.4, "learning_rate": 3.3057448880233695e-05, "loss": 1.4306, "step": 62840 }, { "epoch": 3.4, "learning_rate": 3.30547441306935e-05, "loss": 1.4421, "step": 62850 }, { "epoch": 3.4, "learning_rate": 3.305203938115331e-05, "loss": 1.4443, "step": 62860 }, { "epoch": 3.4, "learning_rate": 3.304933463161312e-05, "loss": 1.5678, "step": 62870 }, { "epoch": 3.4, "learning_rate": 3.304662988207292e-05, "loss": 1.4799, "step": 62880 }, { "epoch": 3.4, "learning_rate": 3.304392513253273e-05, "loss": 1.5335, "step": 62890 }, { "epoch": 3.4, "learning_rate": 3.304122038299254e-05, "loss": 1.4657, "step": 62900 }, { "epoch": 3.4, "learning_rate": 3.303851563345234e-05, "loss": 1.4664, "step": 62910 }, { "epoch": 3.4, "learning_rate": 3.303581088391215e-05, "loss": 1.3975, "step": 62920 }, { "epoch": 3.4, "learning_rate": 3.303310613437196e-05, "loss": 1.3719, "step": 62930 }, { "epoch": 3.4, "learning_rate": 3.3030401384831765e-05, "loss": 1.5249, "step": 62940 }, { "epoch": 3.4, "learning_rate": 3.3027696635291574e-05, "loss": 1.5309, "step": 62950 }, { "epoch": 3.4, "learning_rate": 3.3024991885751384e-05, "loss": 1.4956, "step": 62960 }, { "epoch": 3.4, "learning_rate": 3.302228713621119e-05, "loss": 1.5561, "step": 62970 }, { "epoch": 3.4, "learning_rate": 3.3019582386671e-05, "loss": 1.5398, "step": 62980 }, { "epoch": 3.4, "learning_rate": 3.3016877637130806e-05, "loss": 1.537, "step": 62990 }, { "epoch": 3.4, "learning_rate": 3.301417288759061e-05, "loss": 1.443, "step": 63000 }, { "epoch": 3.4, "learning_rate": 3.301146813805042e-05, "loss": 1.4975, "step": 63010 }, { "epoch": 3.41, "learning_rate": 3.300876338851023e-05, "loss": 1.4092, "step": 63020 }, { "epoch": 3.41, "learning_rate": 3.300605863897003e-05, "loss": 1.5095, "step": 63030 }, { "epoch": 3.41, "learning_rate": 3.300335388942984e-05, "loss": 1.515, "step": 63040 }, { "epoch": 3.41, "learning_rate": 3.300064913988965e-05, "loss": 1.5112, "step": 63050 }, { "epoch": 3.41, "learning_rate": 3.2997944390349454e-05, "loss": 1.4825, "step": 63060 }, { "epoch": 3.41, "learning_rate": 3.299523964080926e-05, "loss": 1.457, "step": 63070 }, { "epoch": 3.41, "learning_rate": 3.299253489126907e-05, "loss": 1.5215, "step": 63080 }, { "epoch": 3.41, "learning_rate": 3.2989830141728876e-05, "loss": 1.3897, "step": 63090 }, { "epoch": 3.41, "learning_rate": 3.2987125392188685e-05, "loss": 1.5265, "step": 63100 }, { "epoch": 3.41, "learning_rate": 3.2984420642648495e-05, "loss": 1.562, "step": 63110 }, { "epoch": 3.41, "learning_rate": 3.29817158931083e-05, "loss": 1.4803, "step": 63120 }, { "epoch": 3.41, "learning_rate": 3.297901114356811e-05, "loss": 1.4681, "step": 63130 }, { "epoch": 3.41, "learning_rate": 3.297630639402792e-05, "loss": 1.5291, "step": 63140 }, { "epoch": 3.41, "learning_rate": 3.297360164448772e-05, "loss": 1.4869, "step": 63150 }, { "epoch": 3.41, "learning_rate": 3.297089689494753e-05, "loss": 1.474, "step": 63160 }, { "epoch": 3.41, "learning_rate": 3.296819214540734e-05, "loss": 1.4336, "step": 63170 }, { "epoch": 3.41, "learning_rate": 3.296548739586714e-05, "loss": 1.4654, "step": 63180 }, { "epoch": 3.41, "learning_rate": 3.296278264632695e-05, "loss": 1.4223, "step": 63190 }, { "epoch": 3.41, "learning_rate": 3.296007789678676e-05, "loss": 1.5295, "step": 63200 }, { "epoch": 3.42, "learning_rate": 3.2957373147246564e-05, "loss": 1.4196, "step": 63210 }, { "epoch": 3.42, "learning_rate": 3.2954668397706374e-05, "loss": 1.4736, "step": 63220 }, { "epoch": 3.42, "learning_rate": 3.295196364816618e-05, "loss": 1.5464, "step": 63230 }, { "epoch": 3.42, "learning_rate": 3.2949258898625987e-05, "loss": 1.5431, "step": 63240 }, { "epoch": 3.42, "learning_rate": 3.2946554149085796e-05, "loss": 1.5302, "step": 63250 }, { "epoch": 3.42, "learning_rate": 3.29438493995456e-05, "loss": 1.4899, "step": 63260 }, { "epoch": 3.42, "learning_rate": 3.2941144650005415e-05, "loss": 1.4833, "step": 63270 }, { "epoch": 3.42, "learning_rate": 3.293843990046522e-05, "loss": 1.4571, "step": 63280 }, { "epoch": 3.42, "learning_rate": 3.293573515092502e-05, "loss": 1.4919, "step": 63290 }, { "epoch": 3.42, "learning_rate": 3.293303040138484e-05, "loss": 1.4411, "step": 63300 }, { "epoch": 3.42, "learning_rate": 3.293032565184464e-05, "loss": 1.4206, "step": 63310 }, { "epoch": 3.42, "learning_rate": 3.2927620902304443e-05, "loss": 1.4964, "step": 63320 }, { "epoch": 3.42, "learning_rate": 3.292491615276426e-05, "loss": 1.5202, "step": 63330 }, { "epoch": 3.42, "learning_rate": 3.292221140322406e-05, "loss": 1.4915, "step": 63340 }, { "epoch": 3.42, "learning_rate": 3.2919506653683866e-05, "loss": 1.5448, "step": 63350 }, { "epoch": 3.42, "learning_rate": 3.291680190414368e-05, "loss": 1.5311, "step": 63360 }, { "epoch": 3.42, "learning_rate": 3.2914097154603485e-05, "loss": 1.5417, "step": 63370 }, { "epoch": 3.42, "learning_rate": 3.291139240506329e-05, "loss": 1.6183, "step": 63380 }, { "epoch": 3.43, "learning_rate": 3.2908687655523104e-05, "loss": 1.5029, "step": 63390 }, { "epoch": 3.43, "learning_rate": 3.290598290598291e-05, "loss": 1.5095, "step": 63400 }, { "epoch": 3.43, "learning_rate": 3.290327815644271e-05, "loss": 1.4263, "step": 63410 }, { "epoch": 3.43, "learning_rate": 3.2900573406902526e-05, "loss": 1.4247, "step": 63420 }, { "epoch": 3.43, "learning_rate": 3.289786865736233e-05, "loss": 1.396, "step": 63430 }, { "epoch": 3.43, "learning_rate": 3.289516390782213e-05, "loss": 1.481, "step": 63440 }, { "epoch": 3.43, "learning_rate": 3.289245915828195e-05, "loss": 1.5669, "step": 63450 }, { "epoch": 3.43, "learning_rate": 3.288975440874175e-05, "loss": 1.5424, "step": 63460 }, { "epoch": 3.43, "learning_rate": 3.2887049659201554e-05, "loss": 1.4863, "step": 63470 }, { "epoch": 3.43, "learning_rate": 3.288434490966137e-05, "loss": 1.4721, "step": 63480 }, { "epoch": 3.43, "learning_rate": 3.2881640160121174e-05, "loss": 1.451, "step": 63490 }, { "epoch": 3.43, "learning_rate": 3.287893541058098e-05, "loss": 1.525, "step": 63500 }, { "epoch": 3.43, "learning_rate": 3.287623066104079e-05, "loss": 1.5088, "step": 63510 }, { "epoch": 3.43, "learning_rate": 3.2873525911500596e-05, "loss": 1.5047, "step": 63520 }, { "epoch": 3.43, "learning_rate": 3.2870821161960405e-05, "loss": 1.4845, "step": 63530 }, { "epoch": 3.43, "learning_rate": 3.2868116412420215e-05, "loss": 1.4907, "step": 63540 }, { "epoch": 3.43, "learning_rate": 3.286541166288002e-05, "loss": 1.533, "step": 63550 }, { "epoch": 3.43, "learning_rate": 3.286270691333983e-05, "loss": 1.4577, "step": 63560 }, { "epoch": 3.43, "learning_rate": 3.286000216379964e-05, "loss": 1.4812, "step": 63570 }, { "epoch": 3.44, "learning_rate": 3.285729741425944e-05, "loss": 1.5221, "step": 63580 }, { "epoch": 3.44, "learning_rate": 3.285459266471925e-05, "loss": 1.4601, "step": 63590 }, { "epoch": 3.44, "learning_rate": 3.285188791517906e-05, "loss": 1.4679, "step": 63600 }, { "epoch": 3.44, "learning_rate": 3.284918316563886e-05, "loss": 1.4671, "step": 63610 }, { "epoch": 3.44, "learning_rate": 3.284647841609867e-05, "loss": 1.5542, "step": 63620 }, { "epoch": 3.44, "learning_rate": 3.284377366655848e-05, "loss": 1.5033, "step": 63630 }, { "epoch": 3.44, "learning_rate": 3.2841068917018284e-05, "loss": 1.5238, "step": 63640 }, { "epoch": 3.44, "learning_rate": 3.2838364167478094e-05, "loss": 1.4508, "step": 63650 }, { "epoch": 3.44, "learning_rate": 3.2835659417937904e-05, "loss": 1.5552, "step": 63660 }, { "epoch": 3.44, "learning_rate": 3.2832954668397707e-05, "loss": 1.5076, "step": 63670 }, { "epoch": 3.44, "learning_rate": 3.2830249918857516e-05, "loss": 1.5237, "step": 63680 }, { "epoch": 3.44, "learning_rate": 3.2827545169317326e-05, "loss": 1.5091, "step": 63690 }, { "epoch": 3.44, "learning_rate": 3.282484041977713e-05, "loss": 1.5031, "step": 63700 }, { "epoch": 3.44, "learning_rate": 3.282213567023694e-05, "loss": 1.4883, "step": 63710 }, { "epoch": 3.44, "learning_rate": 3.281943092069675e-05, "loss": 1.4792, "step": 63720 }, { "epoch": 3.44, "learning_rate": 3.281672617115655e-05, "loss": 1.4968, "step": 63730 }, { "epoch": 3.44, "learning_rate": 3.281402142161636e-05, "loss": 1.5223, "step": 63740 }, { "epoch": 3.44, "learning_rate": 3.281131667207617e-05, "loss": 1.4849, "step": 63750 }, { "epoch": 3.45, "learning_rate": 3.280861192253597e-05, "loss": 1.5518, "step": 63760 }, { "epoch": 3.45, "learning_rate": 3.280590717299578e-05, "loss": 1.498, "step": 63770 }, { "epoch": 3.45, "learning_rate": 3.280320242345559e-05, "loss": 1.3847, "step": 63780 }, { "epoch": 3.45, "learning_rate": 3.2800497673915395e-05, "loss": 1.3624, "step": 63790 }, { "epoch": 3.45, "learning_rate": 3.2797792924375205e-05, "loss": 1.4921, "step": 63800 }, { "epoch": 3.45, "learning_rate": 3.279508817483501e-05, "loss": 1.4684, "step": 63810 }, { "epoch": 3.45, "learning_rate": 3.279238342529482e-05, "loss": 1.3739, "step": 63820 }, { "epoch": 3.45, "learning_rate": 3.278967867575463e-05, "loss": 1.4563, "step": 63830 }, { "epoch": 3.45, "learning_rate": 3.278697392621443e-05, "loss": 1.4626, "step": 63840 }, { "epoch": 3.45, "learning_rate": 3.278426917667424e-05, "loss": 1.6123, "step": 63850 }, { "epoch": 3.45, "learning_rate": 3.278156442713405e-05, "loss": 1.4195, "step": 63860 }, { "epoch": 3.45, "learning_rate": 3.277885967759385e-05, "loss": 1.4714, "step": 63870 }, { "epoch": 3.45, "learning_rate": 3.277615492805366e-05, "loss": 1.5447, "step": 63880 }, { "epoch": 3.45, "learning_rate": 3.277345017851347e-05, "loss": 1.5432, "step": 63890 }, { "epoch": 3.45, "learning_rate": 3.2770745428973274e-05, "loss": 1.537, "step": 63900 }, { "epoch": 3.45, "learning_rate": 3.276804067943309e-05, "loss": 1.4954, "step": 63910 }, { "epoch": 3.45, "learning_rate": 3.2765335929892894e-05, "loss": 1.5046, "step": 63920 }, { "epoch": 3.45, "learning_rate": 3.2762631180352696e-05, "loss": 1.5515, "step": 63930 }, { "epoch": 3.45, "learning_rate": 3.275992643081251e-05, "loss": 1.5295, "step": 63940 }, { "epoch": 3.46, "learning_rate": 3.2757221681272316e-05, "loss": 1.5132, "step": 63950 }, { "epoch": 3.46, "learning_rate": 3.275451693173212e-05, "loss": 1.4004, "step": 63960 }, { "epoch": 3.46, "learning_rate": 3.2751812182191935e-05, "loss": 1.5018, "step": 63970 }, { "epoch": 3.46, "learning_rate": 3.274910743265174e-05, "loss": 1.5034, "step": 63980 }, { "epoch": 3.46, "learning_rate": 3.274640268311154e-05, "loss": 1.4855, "step": 63990 }, { "epoch": 3.46, "learning_rate": 3.274369793357136e-05, "loss": 1.4585, "step": 64000 }, { "epoch": 3.46, "learning_rate": 3.274099318403116e-05, "loss": 1.5104, "step": 64010 }, { "epoch": 3.46, "learning_rate": 3.273828843449096e-05, "loss": 1.4341, "step": 64020 }, { "epoch": 3.46, "learning_rate": 3.273558368495078e-05, "loss": 1.5401, "step": 64030 }, { "epoch": 3.46, "learning_rate": 3.273287893541058e-05, "loss": 1.5054, "step": 64040 }, { "epoch": 3.46, "learning_rate": 3.2730174185870385e-05, "loss": 1.4912, "step": 64050 }, { "epoch": 3.46, "learning_rate": 3.27274694363302e-05, "loss": 1.58, "step": 64060 }, { "epoch": 3.46, "learning_rate": 3.2724764686790004e-05, "loss": 1.4958, "step": 64070 }, { "epoch": 3.46, "learning_rate": 3.272205993724981e-05, "loss": 1.4636, "step": 64080 }, { "epoch": 3.46, "learning_rate": 3.2719355187709624e-05, "loss": 1.5404, "step": 64090 }, { "epoch": 3.46, "learning_rate": 3.2716650438169427e-05, "loss": 1.4221, "step": 64100 }, { "epoch": 3.46, "learning_rate": 3.271394568862923e-05, "loss": 1.533, "step": 64110 }, { "epoch": 3.46, "learning_rate": 3.2711240939089046e-05, "loss": 1.5197, "step": 64120 }, { "epoch": 3.47, "learning_rate": 3.270853618954885e-05, "loss": 1.4165, "step": 64130 }, { "epoch": 3.47, "learning_rate": 3.270583144000866e-05, "loss": 1.4487, "step": 64140 }, { "epoch": 3.47, "learning_rate": 3.270312669046847e-05, "loss": 1.5389, "step": 64150 }, { "epoch": 3.47, "learning_rate": 3.270042194092827e-05, "loss": 1.4959, "step": 64160 }, { "epoch": 3.47, "learning_rate": 3.269771719138808e-05, "loss": 1.5042, "step": 64170 }, { "epoch": 3.47, "learning_rate": 3.269501244184789e-05, "loss": 1.4694, "step": 64180 }, { "epoch": 3.47, "learning_rate": 3.269230769230769e-05, "loss": 1.5927, "step": 64190 }, { "epoch": 3.47, "learning_rate": 3.26896029427675e-05, "loss": 1.5279, "step": 64200 }, { "epoch": 3.47, "learning_rate": 3.268689819322731e-05, "loss": 1.5281, "step": 64210 }, { "epoch": 3.47, "learning_rate": 3.2684193443687115e-05, "loss": 1.5127, "step": 64220 }, { "epoch": 3.47, "learning_rate": 3.2681488694146925e-05, "loss": 1.5168, "step": 64230 }, { "epoch": 3.47, "learning_rate": 3.2678783944606735e-05, "loss": 1.5107, "step": 64240 }, { "epoch": 3.47, "learning_rate": 3.267607919506654e-05, "loss": 1.4441, "step": 64250 }, { "epoch": 3.47, "learning_rate": 3.267337444552635e-05, "loss": 1.4374, "step": 64260 }, { "epoch": 3.47, "learning_rate": 3.267066969598616e-05, "loss": 1.5237, "step": 64270 }, { "epoch": 3.47, "learning_rate": 3.266796494644596e-05, "loss": 1.5269, "step": 64280 }, { "epoch": 3.47, "learning_rate": 3.266526019690577e-05, "loss": 1.443, "step": 64290 }, { "epoch": 3.47, "learning_rate": 3.266255544736558e-05, "loss": 1.4755, "step": 64300 }, { "epoch": 3.47, "learning_rate": 3.265985069782538e-05, "loss": 1.4691, "step": 64310 }, { "epoch": 3.48, "learning_rate": 3.265714594828519e-05, "loss": 1.4335, "step": 64320 }, { "epoch": 3.48, "learning_rate": 3.2654441198745e-05, "loss": 1.4672, "step": 64330 }, { "epoch": 3.48, "learning_rate": 3.2651736449204804e-05, "loss": 1.6134, "step": 64340 }, { "epoch": 3.48, "learning_rate": 3.2649031699664614e-05, "loss": 1.5247, "step": 64350 }, { "epoch": 3.48, "learning_rate": 3.2646326950124416e-05, "loss": 1.457, "step": 64360 }, { "epoch": 3.48, "learning_rate": 3.2643622200584226e-05, "loss": 1.5736, "step": 64370 }, { "epoch": 3.48, "learning_rate": 3.2640917451044036e-05, "loss": 1.5022, "step": 64380 }, { "epoch": 3.48, "learning_rate": 3.263821270150384e-05, "loss": 1.5111, "step": 64390 }, { "epoch": 3.48, "learning_rate": 3.263550795196365e-05, "loss": 1.5272, "step": 64400 }, { "epoch": 3.48, "learning_rate": 3.263280320242346e-05, "loss": 1.5488, "step": 64410 }, { "epoch": 3.48, "learning_rate": 3.263009845288326e-05, "loss": 1.396, "step": 64420 }, { "epoch": 3.48, "learning_rate": 3.262739370334307e-05, "loss": 1.463, "step": 64430 }, { "epoch": 3.48, "learning_rate": 3.262468895380288e-05, "loss": 1.5691, "step": 64440 }, { "epoch": 3.48, "learning_rate": 3.262198420426268e-05, "loss": 1.515, "step": 64450 }, { "epoch": 3.48, "learning_rate": 3.261927945472249e-05, "loss": 1.5261, "step": 64460 }, { "epoch": 3.48, "learning_rate": 3.26165747051823e-05, "loss": 1.5379, "step": 64470 }, { "epoch": 3.48, "learning_rate": 3.2613869955642105e-05, "loss": 1.5273, "step": 64480 }, { "epoch": 3.48, "learning_rate": 3.2611165206101915e-05, "loss": 1.4239, "step": 64490 }, { "epoch": 3.49, "learning_rate": 3.2608460456561724e-05, "loss": 1.5548, "step": 64500 }, { "epoch": 3.49, "learning_rate": 3.260575570702153e-05, "loss": 1.4589, "step": 64510 }, { "epoch": 3.49, "learning_rate": 3.260305095748134e-05, "loss": 1.4672, "step": 64520 }, { "epoch": 3.49, "learning_rate": 3.2600346207941147e-05, "loss": 1.541, "step": 64530 }, { "epoch": 3.49, "learning_rate": 3.259764145840095e-05, "loss": 1.4938, "step": 64540 }, { "epoch": 3.49, "learning_rate": 3.2594936708860766e-05, "loss": 1.4709, "step": 64550 }, { "epoch": 3.49, "learning_rate": 3.259223195932057e-05, "loss": 1.4018, "step": 64560 }, { "epoch": 3.49, "learning_rate": 3.258952720978037e-05, "loss": 1.4796, "step": 64570 }, { "epoch": 3.49, "learning_rate": 3.258682246024019e-05, "loss": 1.4713, "step": 64580 }, { "epoch": 3.49, "learning_rate": 3.258411771069999e-05, "loss": 1.4703, "step": 64590 }, { "epoch": 3.49, "learning_rate": 3.2581412961159794e-05, "loss": 1.478, "step": 64600 }, { "epoch": 3.49, "learning_rate": 3.257870821161961e-05, "loss": 1.4299, "step": 64610 }, { "epoch": 3.49, "learning_rate": 3.257600346207941e-05, "loss": 1.5771, "step": 64620 }, { "epoch": 3.49, "learning_rate": 3.2573298712539216e-05, "loss": 1.504, "step": 64630 }, { "epoch": 3.49, "learning_rate": 3.257059396299903e-05, "loss": 1.5301, "step": 64640 }, { "epoch": 3.49, "learning_rate": 3.2567889213458835e-05, "loss": 1.5346, "step": 64650 }, { "epoch": 3.49, "learning_rate": 3.256518446391864e-05, "loss": 1.5078, "step": 64660 }, { "epoch": 3.49, "learning_rate": 3.2562479714378455e-05, "loss": 1.4296, "step": 64670 }, { "epoch": 3.49, "learning_rate": 3.255977496483826e-05, "loss": 1.4115, "step": 64680 }, { "epoch": 3.5, "learning_rate": 3.255707021529806e-05, "loss": 1.5585, "step": 64690 }, { "epoch": 3.5, "learning_rate": 3.255436546575788e-05, "loss": 1.5188, "step": 64700 }, { "epoch": 3.5, "learning_rate": 3.255166071621768e-05, "loss": 1.4691, "step": 64710 }, { "epoch": 3.5, "learning_rate": 3.254895596667748e-05, "loss": 1.5045, "step": 64720 }, { "epoch": 3.5, "learning_rate": 3.25462512171373e-05, "loss": 1.4979, "step": 64730 }, { "epoch": 3.5, "learning_rate": 3.25435464675971e-05, "loss": 1.5063, "step": 64740 }, { "epoch": 3.5, "learning_rate": 3.254084171805691e-05, "loss": 1.4387, "step": 64750 }, { "epoch": 3.5, "learning_rate": 3.253813696851672e-05, "loss": 1.502, "step": 64760 }, { "epoch": 3.5, "learning_rate": 3.2535432218976524e-05, "loss": 1.4753, "step": 64770 }, { "epoch": 3.5, "learning_rate": 3.2532727469436334e-05, "loss": 1.5742, "step": 64780 }, { "epoch": 3.5, "learning_rate": 3.253002271989614e-05, "loss": 1.425, "step": 64790 }, { "epoch": 3.5, "learning_rate": 3.2527317970355946e-05, "loss": 1.4399, "step": 64800 }, { "epoch": 3.5, "learning_rate": 3.2524613220815756e-05, "loss": 1.487, "step": 64810 }, { "epoch": 3.5, "learning_rate": 3.2521908471275565e-05, "loss": 1.5212, "step": 64820 }, { "epoch": 3.5, "learning_rate": 3.251920372173537e-05, "loss": 1.6097, "step": 64830 }, { "epoch": 3.5, "learning_rate": 3.251649897219518e-05, "loss": 1.484, "step": 64840 }, { "epoch": 3.5, "learning_rate": 3.251379422265499e-05, "loss": 1.5601, "step": 64850 }, { "epoch": 3.5, "learning_rate": 3.251108947311479e-05, "loss": 1.4323, "step": 64860 }, { "epoch": 3.51, "learning_rate": 3.25083847235746e-05, "loss": 1.5188, "step": 64870 }, { "epoch": 3.51, "learning_rate": 3.250567997403441e-05, "loss": 1.5291, "step": 64880 }, { "epoch": 3.51, "learning_rate": 3.250297522449421e-05, "loss": 1.4962, "step": 64890 }, { "epoch": 3.51, "learning_rate": 3.250027047495402e-05, "loss": 1.4829, "step": 64900 }, { "epoch": 3.51, "learning_rate": 3.2497565725413825e-05, "loss": 1.4975, "step": 64910 }, { "epoch": 3.51, "learning_rate": 3.2494860975873635e-05, "loss": 1.5965, "step": 64920 }, { "epoch": 3.51, "learning_rate": 3.2492156226333444e-05, "loss": 1.5131, "step": 64930 }, { "epoch": 3.51, "learning_rate": 3.248945147679325e-05, "loss": 1.502, "step": 64940 }, { "epoch": 3.51, "learning_rate": 3.248674672725306e-05, "loss": 1.5112, "step": 64950 }, { "epoch": 3.51, "learning_rate": 3.248404197771287e-05, "loss": 1.4283, "step": 64960 }, { "epoch": 3.51, "learning_rate": 3.248133722817267e-05, "loss": 1.4434, "step": 64970 }, { "epoch": 3.51, "learning_rate": 3.247863247863248e-05, "loss": 1.486, "step": 64980 }, { "epoch": 3.51, "learning_rate": 3.247592772909229e-05, "loss": 1.4669, "step": 64990 }, { "epoch": 3.51, "learning_rate": 3.247322297955209e-05, "loss": 1.4085, "step": 65000 }, { "epoch": 3.51, "learning_rate": 3.24705182300119e-05, "loss": 1.4906, "step": 65010 }, { "epoch": 3.51, "learning_rate": 3.246781348047171e-05, "loss": 1.4417, "step": 65020 }, { "epoch": 3.51, "learning_rate": 3.2465108730931514e-05, "loss": 1.5141, "step": 65030 }, { "epoch": 3.51, "learning_rate": 3.2462403981391323e-05, "loss": 1.5224, "step": 65040 }, { "epoch": 3.51, "learning_rate": 3.245969923185113e-05, "loss": 1.4587, "step": 65050 }, { "epoch": 3.52, "learning_rate": 3.2456994482310936e-05, "loss": 1.539, "step": 65060 }, { "epoch": 3.52, "learning_rate": 3.2454289732770746e-05, "loss": 1.551, "step": 65070 }, { "epoch": 3.52, "learning_rate": 3.2451584983230555e-05, "loss": 1.5063, "step": 65080 }, { "epoch": 3.52, "learning_rate": 3.244888023369036e-05, "loss": 1.4733, "step": 65090 }, { "epoch": 3.52, "learning_rate": 3.244617548415017e-05, "loss": 1.5041, "step": 65100 }, { "epoch": 3.52, "learning_rate": 3.244347073460998e-05, "loss": 1.4421, "step": 65110 }, { "epoch": 3.52, "learning_rate": 3.244076598506978e-05, "loss": 1.5042, "step": 65120 }, { "epoch": 3.52, "learning_rate": 3.243806123552959e-05, "loss": 1.5864, "step": 65130 }, { "epoch": 3.52, "learning_rate": 3.24353564859894e-05, "loss": 1.441, "step": 65140 }, { "epoch": 3.52, "learning_rate": 3.24326517364492e-05, "loss": 1.5013, "step": 65150 }, { "epoch": 3.52, "learning_rate": 3.242994698690902e-05, "loss": 1.5353, "step": 65160 }, { "epoch": 3.52, "learning_rate": 3.242724223736882e-05, "loss": 1.5253, "step": 65170 }, { "epoch": 3.52, "learning_rate": 3.2424537487828625e-05, "loss": 1.516, "step": 65180 }, { "epoch": 3.52, "learning_rate": 3.242183273828844e-05, "loss": 1.4318, "step": 65190 }, { "epoch": 3.52, "learning_rate": 3.2419127988748244e-05, "loss": 1.5271, "step": 65200 }, { "epoch": 3.52, "learning_rate": 3.241642323920805e-05, "loss": 1.4063, "step": 65210 }, { "epoch": 3.52, "learning_rate": 3.241371848966786e-05, "loss": 1.4393, "step": 65220 }, { "epoch": 3.52, "learning_rate": 3.2411013740127666e-05, "loss": 1.5409, "step": 65230 }, { "epoch": 3.53, "learning_rate": 3.240830899058747e-05, "loss": 1.5184, "step": 65240 }, { "epoch": 3.53, "learning_rate": 3.2405604241047285e-05, "loss": 1.506, "step": 65250 }, { "epoch": 3.53, "learning_rate": 3.240289949150709e-05, "loss": 1.5015, "step": 65260 }, { "epoch": 3.53, "learning_rate": 3.240019474196689e-05, "loss": 1.4178, "step": 65270 }, { "epoch": 3.53, "learning_rate": 3.239748999242671e-05, "loss": 1.4657, "step": 65280 }, { "epoch": 3.53, "learning_rate": 3.239478524288651e-05, "loss": 1.4363, "step": 65290 }, { "epoch": 3.53, "learning_rate": 3.239208049334631e-05, "loss": 1.4074, "step": 65300 }, { "epoch": 3.53, "learning_rate": 3.238937574380613e-05, "loss": 1.5739, "step": 65310 }, { "epoch": 3.53, "learning_rate": 3.238667099426593e-05, "loss": 1.5366, "step": 65320 }, { "epoch": 3.53, "learning_rate": 3.2383966244725736e-05, "loss": 1.492, "step": 65330 }, { "epoch": 3.53, "learning_rate": 3.238126149518555e-05, "loss": 1.3938, "step": 65340 }, { "epoch": 3.53, "learning_rate": 3.2378556745645355e-05, "loss": 1.4684, "step": 65350 }, { "epoch": 3.53, "learning_rate": 3.237585199610516e-05, "loss": 1.4554, "step": 65360 }, { "epoch": 3.53, "learning_rate": 3.2373147246564974e-05, "loss": 1.4826, "step": 65370 }, { "epoch": 3.53, "learning_rate": 3.237044249702478e-05, "loss": 1.4561, "step": 65380 }, { "epoch": 3.53, "learning_rate": 3.236773774748459e-05, "loss": 1.4532, "step": 65390 }, { "epoch": 3.53, "learning_rate": 3.2365032997944396e-05, "loss": 1.5305, "step": 65400 }, { "epoch": 3.53, "learning_rate": 3.23623282484042e-05, "loss": 1.4935, "step": 65410 }, { "epoch": 3.53, "learning_rate": 3.235962349886401e-05, "loss": 1.4838, "step": 65420 }, { "epoch": 3.54, "learning_rate": 3.235691874932382e-05, "loss": 1.4719, "step": 65430 }, { "epoch": 3.54, "learning_rate": 3.235421399978362e-05, "loss": 1.5289, "step": 65440 }, { "epoch": 3.54, "learning_rate": 3.235150925024343e-05, "loss": 1.4469, "step": 65450 }, { "epoch": 3.54, "learning_rate": 3.234880450070324e-05, "loss": 1.4524, "step": 65460 }, { "epoch": 3.54, "learning_rate": 3.2346099751163044e-05, "loss": 1.518, "step": 65470 }, { "epoch": 3.54, "learning_rate": 3.234339500162285e-05, "loss": 1.4168, "step": 65480 }, { "epoch": 3.54, "learning_rate": 3.2340690252082656e-05, "loss": 1.4622, "step": 65490 }, { "epoch": 3.54, "learning_rate": 3.2337985502542466e-05, "loss": 1.4457, "step": 65500 }, { "epoch": 3.54, "learning_rate": 3.2335280753002275e-05, "loss": 1.4555, "step": 65510 }, { "epoch": 3.54, "learning_rate": 3.233257600346208e-05, "loss": 1.4445, "step": 65520 }, { "epoch": 3.54, "learning_rate": 3.232987125392189e-05, "loss": 1.529, "step": 65530 }, { "epoch": 3.54, "learning_rate": 3.23271665043817e-05, "loss": 1.5256, "step": 65540 }, { "epoch": 3.54, "learning_rate": 3.23244617548415e-05, "loss": 1.5144, "step": 65550 }, { "epoch": 3.54, "learning_rate": 3.232175700530131e-05, "loss": 1.4786, "step": 65560 }, { "epoch": 3.54, "learning_rate": 3.231905225576112e-05, "loss": 1.4935, "step": 65570 }, { "epoch": 3.54, "learning_rate": 3.231634750622092e-05, "loss": 1.4902, "step": 65580 }, { "epoch": 3.54, "learning_rate": 3.231364275668073e-05, "loss": 1.4835, "step": 65590 }, { "epoch": 3.54, "learning_rate": 3.231093800714054e-05, "loss": 1.506, "step": 65600 }, { "epoch": 3.55, "learning_rate": 3.2308233257600345e-05, "loss": 1.3915, "step": 65610 }, { "epoch": 3.55, "learning_rate": 3.2305528508060154e-05, "loss": 1.4568, "step": 65620 }, { "epoch": 3.55, "learning_rate": 3.2302823758519964e-05, "loss": 1.5373, "step": 65630 }, { "epoch": 3.55, "learning_rate": 3.230011900897977e-05, "loss": 1.4442, "step": 65640 }, { "epoch": 3.55, "learning_rate": 3.2297414259439577e-05, "loss": 1.4976, "step": 65650 }, { "epoch": 3.55, "learning_rate": 3.2294709509899386e-05, "loss": 1.5037, "step": 65660 }, { "epoch": 3.55, "learning_rate": 3.229200476035919e-05, "loss": 1.4755, "step": 65670 }, { "epoch": 3.55, "learning_rate": 3.2289300010819e-05, "loss": 1.5493, "step": 65680 }, { "epoch": 3.55, "learning_rate": 3.228659526127881e-05, "loss": 1.4898, "step": 65690 }, { "epoch": 3.55, "learning_rate": 3.228389051173861e-05, "loss": 1.4676, "step": 65700 }, { "epoch": 3.55, "learning_rate": 3.228118576219842e-05, "loss": 1.4812, "step": 65710 }, { "epoch": 3.55, "learning_rate": 3.227848101265823e-05, "loss": 1.4751, "step": 65720 }, { "epoch": 3.55, "learning_rate": 3.227577626311803e-05, "loss": 1.4597, "step": 65730 }, { "epoch": 3.55, "learning_rate": 3.227307151357784e-05, "loss": 1.4936, "step": 65740 }, { "epoch": 3.55, "learning_rate": 3.227036676403765e-05, "loss": 1.4868, "step": 65750 }, { "epoch": 3.55, "learning_rate": 3.2267662014497456e-05, "loss": 1.4924, "step": 65760 }, { "epoch": 3.55, "learning_rate": 3.2264957264957265e-05, "loss": 1.4922, "step": 65770 }, { "epoch": 3.55, "learning_rate": 3.2262252515417075e-05, "loss": 1.4767, "step": 65780 }, { "epoch": 3.55, "learning_rate": 3.225954776587688e-05, "loss": 1.4049, "step": 65790 }, { "epoch": 3.56, "learning_rate": 3.2256843016336694e-05, "loss": 1.5044, "step": 65800 }, { "epoch": 3.56, "learning_rate": 3.22541382667965e-05, "loss": 1.494, "step": 65810 }, { "epoch": 3.56, "learning_rate": 3.22514335172563e-05, "loss": 1.5436, "step": 65820 }, { "epoch": 3.56, "learning_rate": 3.2248728767716116e-05, "loss": 1.4913, "step": 65830 }, { "epoch": 3.56, "learning_rate": 3.224602401817592e-05, "loss": 1.4646, "step": 65840 }, { "epoch": 3.56, "learning_rate": 3.224331926863572e-05, "loss": 1.4548, "step": 65850 }, { "epoch": 3.56, "learning_rate": 3.224061451909554e-05, "loss": 1.4559, "step": 65860 }, { "epoch": 3.56, "learning_rate": 3.223790976955534e-05, "loss": 1.5566, "step": 65870 }, { "epoch": 3.56, "learning_rate": 3.2235205020015144e-05, "loss": 1.5002, "step": 65880 }, { "epoch": 3.56, "learning_rate": 3.223250027047496e-05, "loss": 1.4693, "step": 65890 }, { "epoch": 3.56, "learning_rate": 3.2229795520934764e-05, "loss": 1.4543, "step": 65900 }, { "epoch": 3.56, "learning_rate": 3.2227090771394566e-05, "loss": 1.4928, "step": 65910 }, { "epoch": 3.56, "learning_rate": 3.222438602185438e-05, "loss": 1.483, "step": 65920 }, { "epoch": 3.56, "learning_rate": 3.2221681272314186e-05, "loss": 1.4769, "step": 65930 }, { "epoch": 3.56, "learning_rate": 3.221897652277399e-05, "loss": 1.4141, "step": 65940 }, { "epoch": 3.56, "learning_rate": 3.2216271773233805e-05, "loss": 1.4704, "step": 65950 }, { "epoch": 3.56, "learning_rate": 3.221356702369361e-05, "loss": 1.5057, "step": 65960 }, { "epoch": 3.56, "learning_rate": 3.221086227415341e-05, "loss": 1.4306, "step": 65970 }, { "epoch": 3.57, "learning_rate": 3.220815752461323e-05, "loss": 1.531, "step": 65980 }, { "epoch": 3.57, "learning_rate": 3.220545277507303e-05, "loss": 1.421, "step": 65990 }, { "epoch": 3.57, "learning_rate": 3.220274802553283e-05, "loss": 1.4736, "step": 66000 }, { "epoch": 3.57, "learning_rate": 3.220004327599265e-05, "loss": 1.4783, "step": 66010 }, { "epoch": 3.57, "learning_rate": 3.219733852645245e-05, "loss": 1.5341, "step": 66020 }, { "epoch": 3.57, "learning_rate": 3.219463377691226e-05, "loss": 1.4583, "step": 66030 }, { "epoch": 3.57, "learning_rate": 3.2191929027372065e-05, "loss": 1.463, "step": 66040 }, { "epoch": 3.57, "learning_rate": 3.2189224277831874e-05, "loss": 1.4576, "step": 66050 }, { "epoch": 3.57, "learning_rate": 3.2186519528291684e-05, "loss": 1.3795, "step": 66060 }, { "epoch": 3.57, "learning_rate": 3.218381477875149e-05, "loss": 1.4197, "step": 66070 }, { "epoch": 3.57, "learning_rate": 3.2181110029211297e-05, "loss": 1.5054, "step": 66080 }, { "epoch": 3.57, "learning_rate": 3.2178405279671106e-05, "loss": 1.4893, "step": 66090 }, { "epoch": 3.57, "learning_rate": 3.217570053013091e-05, "loss": 1.489, "step": 66100 }, { "epoch": 3.57, "learning_rate": 3.217299578059072e-05, "loss": 1.4232, "step": 66110 }, { "epoch": 3.57, "learning_rate": 3.217029103105053e-05, "loss": 1.5384, "step": 66120 }, { "epoch": 3.57, "learning_rate": 3.216758628151033e-05, "loss": 1.5346, "step": 66130 }, { "epoch": 3.57, "learning_rate": 3.216488153197014e-05, "loss": 1.5067, "step": 66140 }, { "epoch": 3.57, "learning_rate": 3.216217678242995e-05, "loss": 1.4695, "step": 66150 }, { "epoch": 3.57, "learning_rate": 3.2159472032889753e-05, "loss": 1.3996, "step": 66160 }, { "epoch": 3.58, "learning_rate": 3.215676728334956e-05, "loss": 1.5068, "step": 66170 }, { "epoch": 3.58, "learning_rate": 3.215406253380937e-05, "loss": 1.4492, "step": 66180 }, { "epoch": 3.58, "learning_rate": 3.2151357784269176e-05, "loss": 1.4461, "step": 66190 }, { "epoch": 3.58, "learning_rate": 3.2148653034728985e-05, "loss": 1.5285, "step": 66200 }, { "epoch": 3.58, "learning_rate": 3.2145948285188795e-05, "loss": 1.4685, "step": 66210 }, { "epoch": 3.58, "learning_rate": 3.21432435356486e-05, "loss": 1.5343, "step": 66220 }, { "epoch": 3.58, "learning_rate": 3.214053878610841e-05, "loss": 1.5291, "step": 66230 }, { "epoch": 3.58, "learning_rate": 3.213783403656822e-05, "loss": 1.4652, "step": 66240 }, { "epoch": 3.58, "learning_rate": 3.213512928702802e-05, "loss": 1.5424, "step": 66250 }, { "epoch": 3.58, "learning_rate": 3.213242453748783e-05, "loss": 1.5557, "step": 66260 }, { "epoch": 3.58, "learning_rate": 3.212971978794764e-05, "loss": 1.4087, "step": 66270 }, { "epoch": 3.58, "learning_rate": 3.212701503840744e-05, "loss": 1.5122, "step": 66280 }, { "epoch": 3.58, "learning_rate": 3.212431028886725e-05, "loss": 1.5304, "step": 66290 }, { "epoch": 3.58, "learning_rate": 3.212160553932706e-05, "loss": 1.4769, "step": 66300 }, { "epoch": 3.58, "learning_rate": 3.2118900789786864e-05, "loss": 1.5245, "step": 66310 }, { "epoch": 3.58, "learning_rate": 3.2116196040246674e-05, "loss": 1.5417, "step": 66320 }, { "epoch": 3.58, "learning_rate": 3.2113491290706484e-05, "loss": 1.5161, "step": 66330 }, { "epoch": 3.58, "learning_rate": 3.2110786541166286e-05, "loss": 1.5114, "step": 66340 }, { "epoch": 3.59, "learning_rate": 3.2108081791626096e-05, "loss": 1.5292, "step": 66350 }, { "epoch": 3.59, "learning_rate": 3.2105377042085906e-05, "loss": 1.5193, "step": 66360 }, { "epoch": 3.59, "learning_rate": 3.210267229254571e-05, "loss": 1.4517, "step": 66370 }, { "epoch": 3.59, "learning_rate": 3.209996754300552e-05, "loss": 1.4883, "step": 66380 }, { "epoch": 3.59, "learning_rate": 3.209726279346533e-05, "loss": 1.624, "step": 66390 }, { "epoch": 3.59, "learning_rate": 3.209455804392513e-05, "loss": 1.4978, "step": 66400 }, { "epoch": 3.59, "learning_rate": 3.209185329438494e-05, "loss": 1.4125, "step": 66410 }, { "epoch": 3.59, "learning_rate": 3.208914854484475e-05, "loss": 1.5485, "step": 66420 }, { "epoch": 3.59, "learning_rate": 3.208644379530455e-05, "loss": 1.5669, "step": 66430 }, { "epoch": 3.59, "learning_rate": 3.208373904576437e-05, "loss": 1.4725, "step": 66440 }, { "epoch": 3.59, "learning_rate": 3.208103429622417e-05, "loss": 1.4911, "step": 66450 }, { "epoch": 3.59, "learning_rate": 3.2078329546683975e-05, "loss": 1.4625, "step": 66460 }, { "epoch": 3.59, "learning_rate": 3.207562479714379e-05, "loss": 1.497, "step": 66470 }, { "epoch": 3.59, "learning_rate": 3.2072920047603594e-05, "loss": 1.5228, "step": 66480 }, { "epoch": 3.59, "learning_rate": 3.20702152980634e-05, "loss": 1.5116, "step": 66490 }, { "epoch": 3.59, "learning_rate": 3.2067510548523214e-05, "loss": 1.4958, "step": 66500 }, { "epoch": 3.59, "learning_rate": 3.2064805798983017e-05, "loss": 1.4433, "step": 66510 }, { "epoch": 3.59, "learning_rate": 3.206210104944282e-05, "loss": 1.5153, "step": 66520 }, { "epoch": 3.59, "learning_rate": 3.2059396299902636e-05, "loss": 1.5307, "step": 66530 }, { "epoch": 3.6, "learning_rate": 3.205669155036244e-05, "loss": 1.5205, "step": 66540 }, { "epoch": 3.6, "learning_rate": 3.205398680082224e-05, "loss": 1.4889, "step": 66550 }, { "epoch": 3.6, "learning_rate": 3.205128205128206e-05, "loss": 1.4788, "step": 66560 }, { "epoch": 3.6, "learning_rate": 3.204857730174186e-05, "loss": 1.5222, "step": 66570 }, { "epoch": 3.6, "learning_rate": 3.2045872552201664e-05, "loss": 1.4843, "step": 66580 }, { "epoch": 3.6, "learning_rate": 3.2043167802661473e-05, "loss": 1.5235, "step": 66590 }, { "epoch": 3.6, "learning_rate": 3.204046305312128e-05, "loss": 1.5176, "step": 66600 }, { "epoch": 3.6, "learning_rate": 3.2037758303581086e-05, "loss": 1.5666, "step": 66610 }, { "epoch": 3.6, "learning_rate": 3.2035053554040896e-05, "loss": 1.4455, "step": 66620 }, { "epoch": 3.6, "learning_rate": 3.2032348804500705e-05, "loss": 1.4391, "step": 66630 }, { "epoch": 3.6, "learning_rate": 3.202964405496051e-05, "loss": 1.4256, "step": 66640 }, { "epoch": 3.6, "learning_rate": 3.202693930542032e-05, "loss": 1.5645, "step": 66650 }, { "epoch": 3.6, "learning_rate": 3.202423455588013e-05, "loss": 1.4527, "step": 66660 }, { "epoch": 3.6, "learning_rate": 3.202152980633994e-05, "loss": 1.5739, "step": 66670 }, { "epoch": 3.6, "learning_rate": 3.201882505679974e-05, "loss": 1.5062, "step": 66680 }, { "epoch": 3.6, "learning_rate": 3.201612030725955e-05, "loss": 1.5242, "step": 66690 }, { "epoch": 3.6, "learning_rate": 3.201341555771936e-05, "loss": 1.4317, "step": 66700 }, { "epoch": 3.6, "learning_rate": 3.201071080817916e-05, "loss": 1.5003, "step": 66710 }, { "epoch": 3.61, "learning_rate": 3.200800605863897e-05, "loss": 1.587, "step": 66720 }, { "epoch": 3.61, "learning_rate": 3.200530130909878e-05, "loss": 1.4051, "step": 66730 }, { "epoch": 3.61, "learning_rate": 3.2002596559558584e-05, "loss": 1.5024, "step": 66740 }, { "epoch": 3.61, "learning_rate": 3.1999891810018394e-05, "loss": 1.557, "step": 66750 }, { "epoch": 3.61, "learning_rate": 3.1997187060478204e-05, "loss": 1.4993, "step": 66760 }, { "epoch": 3.61, "learning_rate": 3.1994482310938006e-05, "loss": 1.4938, "step": 66770 }, { "epoch": 3.61, "learning_rate": 3.1991777561397816e-05, "loss": 1.5198, "step": 66780 }, { "epoch": 3.61, "learning_rate": 3.1989072811857626e-05, "loss": 1.5083, "step": 66790 }, { "epoch": 3.61, "learning_rate": 3.198636806231743e-05, "loss": 1.5381, "step": 66800 }, { "epoch": 3.61, "learning_rate": 3.198366331277724e-05, "loss": 1.4907, "step": 66810 }, { "epoch": 3.61, "learning_rate": 3.198095856323705e-05, "loss": 1.4904, "step": 66820 }, { "epoch": 3.61, "learning_rate": 3.197825381369685e-05, "loss": 1.4664, "step": 66830 }, { "epoch": 3.61, "learning_rate": 3.197554906415666e-05, "loss": 1.5418, "step": 66840 }, { "epoch": 3.61, "learning_rate": 3.197284431461647e-05, "loss": 1.4765, "step": 66850 }, { "epoch": 3.61, "learning_rate": 3.197013956507627e-05, "loss": 1.4668, "step": 66860 }, { "epoch": 3.61, "learning_rate": 3.196743481553608e-05, "loss": 1.5179, "step": 66870 }, { "epoch": 3.61, "learning_rate": 3.196473006599589e-05, "loss": 1.5184, "step": 66880 }, { "epoch": 3.61, "learning_rate": 3.1962025316455695e-05, "loss": 1.4759, "step": 66890 }, { "epoch": 3.61, "learning_rate": 3.1959320566915505e-05, "loss": 1.4779, "step": 66900 }, { "epoch": 3.62, "learning_rate": 3.1956615817375314e-05, "loss": 1.4048, "step": 66910 }, { "epoch": 3.62, "learning_rate": 3.195391106783512e-05, "loss": 1.5581, "step": 66920 }, { "epoch": 3.62, "learning_rate": 3.195120631829493e-05, "loss": 1.5407, "step": 66930 }, { "epoch": 3.62, "learning_rate": 3.1948501568754737e-05, "loss": 1.4618, "step": 66940 }, { "epoch": 3.62, "learning_rate": 3.194579681921454e-05, "loss": 1.4849, "step": 66950 }, { "epoch": 3.62, "learning_rate": 3.194309206967435e-05, "loss": 1.4884, "step": 66960 }, { "epoch": 3.62, "learning_rate": 3.194038732013416e-05, "loss": 1.4646, "step": 66970 }, { "epoch": 3.62, "learning_rate": 3.193768257059396e-05, "loss": 1.4939, "step": 66980 }, { "epoch": 3.62, "learning_rate": 3.193497782105377e-05, "loss": 1.4951, "step": 66990 }, { "epoch": 3.62, "learning_rate": 3.193227307151358e-05, "loss": 1.5976, "step": 67000 }, { "epoch": 3.62, "learning_rate": 3.1929568321973384e-05, "loss": 1.5332, "step": 67010 }, { "epoch": 3.62, "learning_rate": 3.1926863572433193e-05, "loss": 1.4841, "step": 67020 }, { "epoch": 3.62, "learning_rate": 3.1924158822893e-05, "loss": 1.4814, "step": 67030 }, { "epoch": 3.62, "learning_rate": 3.1921454073352806e-05, "loss": 1.5337, "step": 67040 }, { "epoch": 3.62, "learning_rate": 3.1918749323812616e-05, "loss": 1.5131, "step": 67050 }, { "epoch": 3.62, "learning_rate": 3.1916044574272425e-05, "loss": 1.5487, "step": 67060 }, { "epoch": 3.62, "learning_rate": 3.191333982473223e-05, "loss": 1.4865, "step": 67070 }, { "epoch": 3.62, "learning_rate": 3.1910635075192045e-05, "loss": 1.5407, "step": 67080 }, { "epoch": 3.63, "learning_rate": 3.190793032565185e-05, "loss": 1.5015, "step": 67090 }, { "epoch": 3.63, "learning_rate": 3.190522557611165e-05, "loss": 1.4845, "step": 67100 }, { "epoch": 3.63, "learning_rate": 3.190252082657147e-05, "loss": 1.5752, "step": 67110 }, { "epoch": 3.63, "learning_rate": 3.189981607703127e-05, "loss": 1.4476, "step": 67120 }, { "epoch": 3.63, "learning_rate": 3.189711132749107e-05, "loss": 1.4832, "step": 67130 }, { "epoch": 3.63, "learning_rate": 3.189440657795089e-05, "loss": 1.4704, "step": 67140 }, { "epoch": 3.63, "learning_rate": 3.189170182841069e-05, "loss": 1.4746, "step": 67150 }, { "epoch": 3.63, "learning_rate": 3.1888997078870495e-05, "loss": 1.5326, "step": 67160 }, { "epoch": 3.63, "learning_rate": 3.1886292329330304e-05, "loss": 1.5052, "step": 67170 }, { "epoch": 3.63, "learning_rate": 3.1883587579790114e-05, "loss": 1.5577, "step": 67180 }, { "epoch": 3.63, "learning_rate": 3.188088283024992e-05, "loss": 1.5122, "step": 67190 }, { "epoch": 3.63, "learning_rate": 3.1878178080709726e-05, "loss": 1.3963, "step": 67200 }, { "epoch": 3.63, "learning_rate": 3.1875473331169536e-05, "loss": 1.4962, "step": 67210 }, { "epoch": 3.63, "learning_rate": 3.187276858162934e-05, "loss": 1.4737, "step": 67220 }, { "epoch": 3.63, "learning_rate": 3.187006383208915e-05, "loss": 1.5034, "step": 67230 }, { "epoch": 3.63, "learning_rate": 3.186735908254896e-05, "loss": 1.6083, "step": 67240 }, { "epoch": 3.63, "learning_rate": 3.186465433300876e-05, "loss": 1.5238, "step": 67250 }, { "epoch": 3.63, "learning_rate": 3.186194958346857e-05, "loss": 1.3825, "step": 67260 }, { "epoch": 3.63, "learning_rate": 3.185924483392838e-05, "loss": 1.4857, "step": 67270 }, { "epoch": 3.64, "learning_rate": 3.185654008438819e-05, "loss": 1.4439, "step": 67280 }, { "epoch": 3.64, "learning_rate": 3.185383533484799e-05, "loss": 1.5465, "step": 67290 }, { "epoch": 3.64, "learning_rate": 3.18511305853078e-05, "loss": 1.4666, "step": 67300 }, { "epoch": 3.64, "learning_rate": 3.184842583576761e-05, "loss": 1.522, "step": 67310 }, { "epoch": 3.64, "learning_rate": 3.1845721086227415e-05, "loss": 1.5103, "step": 67320 }, { "epoch": 3.64, "learning_rate": 3.1843016336687225e-05, "loss": 1.5068, "step": 67330 }, { "epoch": 3.64, "learning_rate": 3.1840311587147034e-05, "loss": 1.5178, "step": 67340 }, { "epoch": 3.64, "learning_rate": 3.183760683760684e-05, "loss": 1.5092, "step": 67350 }, { "epoch": 3.64, "learning_rate": 3.183490208806665e-05, "loss": 1.4227, "step": 67360 }, { "epoch": 3.64, "learning_rate": 3.1832197338526457e-05, "loss": 1.5459, "step": 67370 }, { "epoch": 3.64, "learning_rate": 3.182949258898626e-05, "loss": 1.413, "step": 67380 }, { "epoch": 3.64, "learning_rate": 3.182678783944607e-05, "loss": 1.4445, "step": 67390 }, { "epoch": 3.64, "learning_rate": 3.182408308990588e-05, "loss": 1.5328, "step": 67400 }, { "epoch": 3.64, "learning_rate": 3.182137834036568e-05, "loss": 1.5684, "step": 67410 }, { "epoch": 3.64, "learning_rate": 3.181867359082549e-05, "loss": 1.4375, "step": 67420 }, { "epoch": 3.64, "learning_rate": 3.18159688412853e-05, "loss": 1.4933, "step": 67430 }, { "epoch": 3.64, "learning_rate": 3.1813264091745104e-05, "loss": 1.5011, "step": 67440 }, { "epoch": 3.64, "learning_rate": 3.1810559342204913e-05, "loss": 1.4395, "step": 67450 }, { "epoch": 3.65, "learning_rate": 3.180785459266472e-05, "loss": 1.454, "step": 67460 }, { "epoch": 3.65, "learning_rate": 3.1805149843124526e-05, "loss": 1.4231, "step": 67470 }, { "epoch": 3.65, "learning_rate": 3.1802445093584336e-05, "loss": 1.4868, "step": 67480 }, { "epoch": 3.65, "learning_rate": 3.1799740344044145e-05, "loss": 1.5363, "step": 67490 }, { "epoch": 3.65, "learning_rate": 3.179703559450395e-05, "loss": 1.5703, "step": 67500 }, { "epoch": 3.65, "learning_rate": 3.179433084496376e-05, "loss": 1.5071, "step": 67510 }, { "epoch": 3.65, "learning_rate": 3.179162609542357e-05, "loss": 1.5125, "step": 67520 }, { "epoch": 3.65, "learning_rate": 3.178892134588337e-05, "loss": 1.4923, "step": 67530 }, { "epoch": 3.65, "learning_rate": 3.178621659634318e-05, "loss": 1.4991, "step": 67540 }, { "epoch": 3.65, "learning_rate": 3.178351184680299e-05, "loss": 1.5268, "step": 67550 }, { "epoch": 3.65, "learning_rate": 3.178080709726279e-05, "loss": 1.438, "step": 67560 }, { "epoch": 3.65, "learning_rate": 3.17781023477226e-05, "loss": 1.4684, "step": 67570 }, { "epoch": 3.65, "learning_rate": 3.177539759818241e-05, "loss": 1.4249, "step": 67580 }, { "epoch": 3.65, "learning_rate": 3.1772692848642215e-05, "loss": 1.5058, "step": 67590 }, { "epoch": 3.65, "learning_rate": 3.1769988099102024e-05, "loss": 1.5599, "step": 67600 }, { "epoch": 3.65, "learning_rate": 3.1767283349561834e-05, "loss": 1.439, "step": 67610 }, { "epoch": 3.65, "learning_rate": 3.176457860002164e-05, "loss": 1.5398, "step": 67620 }, { "epoch": 3.65, "learning_rate": 3.1761873850481446e-05, "loss": 1.5354, "step": 67630 }, { "epoch": 3.65, "learning_rate": 3.1759169100941256e-05, "loss": 1.4563, "step": 67640 }, { "epoch": 3.66, "learning_rate": 3.175646435140106e-05, "loss": 1.4557, "step": 67650 }, { "epoch": 3.66, "learning_rate": 3.175375960186087e-05, "loss": 1.5615, "step": 67660 }, { "epoch": 3.66, "learning_rate": 3.175105485232068e-05, "loss": 1.3994, "step": 67670 }, { "epoch": 3.66, "learning_rate": 3.174835010278048e-05, "loss": 1.4917, "step": 67680 }, { "epoch": 3.66, "learning_rate": 3.174564535324029e-05, "loss": 1.4288, "step": 67690 }, { "epoch": 3.66, "learning_rate": 3.17429406037001e-05, "loss": 1.5538, "step": 67700 }, { "epoch": 3.66, "learning_rate": 3.17402358541599e-05, "loss": 1.5164, "step": 67710 }, { "epoch": 3.66, "learning_rate": 3.173753110461971e-05, "loss": 1.4282, "step": 67720 }, { "epoch": 3.66, "learning_rate": 3.173482635507952e-05, "loss": 1.4008, "step": 67730 }, { "epoch": 3.66, "learning_rate": 3.1732121605539326e-05, "loss": 1.5098, "step": 67740 }, { "epoch": 3.66, "learning_rate": 3.1729416855999135e-05, "loss": 1.4716, "step": 67750 }, { "epoch": 3.66, "learning_rate": 3.1726712106458945e-05, "loss": 1.5371, "step": 67760 }, { "epoch": 3.66, "learning_rate": 3.172400735691875e-05, "loss": 1.435, "step": 67770 }, { "epoch": 3.66, "learning_rate": 3.172130260737856e-05, "loss": 1.5425, "step": 67780 }, { "epoch": 3.66, "learning_rate": 3.171859785783837e-05, "loss": 1.5022, "step": 67790 }, { "epoch": 3.66, "learning_rate": 3.171589310829817e-05, "loss": 1.5404, "step": 67800 }, { "epoch": 3.66, "learning_rate": 3.171318835875798e-05, "loss": 1.5327, "step": 67810 }, { "epoch": 3.66, "learning_rate": 3.171048360921779e-05, "loss": 1.501, "step": 67820 }, { "epoch": 3.67, "learning_rate": 3.170777885967759e-05, "loss": 1.4921, "step": 67830 }, { "epoch": 3.67, "learning_rate": 3.17050741101374e-05, "loss": 1.5369, "step": 67840 }, { "epoch": 3.67, "learning_rate": 3.170236936059721e-05, "loss": 1.5351, "step": 67850 }, { "epoch": 3.67, "learning_rate": 3.1699664611057014e-05, "loss": 1.5037, "step": 67860 }, { "epoch": 3.67, "learning_rate": 3.1696959861516824e-05, "loss": 1.5002, "step": 67870 }, { "epoch": 3.67, "learning_rate": 3.1694255111976633e-05, "loss": 1.3837, "step": 67880 }, { "epoch": 3.67, "learning_rate": 3.1691550362436436e-05, "loss": 1.4571, "step": 67890 }, { "epoch": 3.67, "learning_rate": 3.1688845612896246e-05, "loss": 1.4949, "step": 67900 }, { "epoch": 3.67, "learning_rate": 3.1686140863356056e-05, "loss": 1.3489, "step": 67910 }, { "epoch": 3.67, "learning_rate": 3.1683436113815865e-05, "loss": 1.5891, "step": 67920 }, { "epoch": 3.67, "learning_rate": 3.168073136427567e-05, "loss": 1.4765, "step": 67930 }, { "epoch": 3.67, "learning_rate": 3.167802661473548e-05, "loss": 1.4965, "step": 67940 }, { "epoch": 3.67, "learning_rate": 3.167532186519529e-05, "loss": 1.5795, "step": 67950 }, { "epoch": 3.67, "learning_rate": 3.167261711565509e-05, "loss": 1.504, "step": 67960 }, { "epoch": 3.67, "learning_rate": 3.16699123661149e-05, "loss": 1.5588, "step": 67970 }, { "epoch": 3.67, "learning_rate": 3.166720761657471e-05, "loss": 1.4769, "step": 67980 }, { "epoch": 3.67, "learning_rate": 3.166450286703451e-05, "loss": 1.5402, "step": 67990 }, { "epoch": 3.67, "learning_rate": 3.166179811749432e-05, "loss": 1.5437, "step": 68000 }, { "epoch": 3.67, "learning_rate": 3.165909336795413e-05, "loss": 1.5155, "step": 68010 }, { "epoch": 3.68, "learning_rate": 3.1656388618413935e-05, "loss": 1.4025, "step": 68020 }, { "epoch": 3.68, "learning_rate": 3.1653683868873744e-05, "loss": 1.4623, "step": 68030 }, { "epoch": 3.68, "learning_rate": 3.1650979119333554e-05, "loss": 1.5008, "step": 68040 }, { "epoch": 3.68, "learning_rate": 3.164827436979336e-05, "loss": 1.5526, "step": 68050 }, { "epoch": 3.68, "learning_rate": 3.1645569620253167e-05, "loss": 1.4025, "step": 68060 }, { "epoch": 3.68, "learning_rate": 3.1642864870712976e-05, "loss": 1.5101, "step": 68070 }, { "epoch": 3.68, "learning_rate": 3.164016012117278e-05, "loss": 1.5822, "step": 68080 }, { "epoch": 3.68, "learning_rate": 3.163745537163259e-05, "loss": 1.4741, "step": 68090 }, { "epoch": 3.68, "learning_rate": 3.16347506220924e-05, "loss": 1.4605, "step": 68100 }, { "epoch": 3.68, "learning_rate": 3.16320458725522e-05, "loss": 1.4692, "step": 68110 }, { "epoch": 3.68, "learning_rate": 3.162934112301201e-05, "loss": 1.5253, "step": 68120 }, { "epoch": 3.68, "learning_rate": 3.162663637347182e-05, "loss": 1.5082, "step": 68130 }, { "epoch": 3.68, "learning_rate": 3.162393162393162e-05, "loss": 1.4702, "step": 68140 }, { "epoch": 3.68, "learning_rate": 3.162122687439143e-05, "loss": 1.5169, "step": 68150 }, { "epoch": 3.68, "learning_rate": 3.161852212485124e-05, "loss": 1.4744, "step": 68160 }, { "epoch": 3.68, "learning_rate": 3.1615817375311046e-05, "loss": 1.4213, "step": 68170 }, { "epoch": 3.68, "learning_rate": 3.1613112625770855e-05, "loss": 1.5089, "step": 68180 }, { "epoch": 3.68, "learning_rate": 3.1610407876230665e-05, "loss": 1.5555, "step": 68190 }, { "epoch": 3.69, "learning_rate": 3.160770312669047e-05, "loss": 1.4764, "step": 68200 }, { "epoch": 3.69, "learning_rate": 3.160499837715028e-05, "loss": 1.485, "step": 68210 }, { "epoch": 3.69, "learning_rate": 3.160229362761009e-05, "loss": 1.4705, "step": 68220 }, { "epoch": 3.69, "learning_rate": 3.159958887806989e-05, "loss": 1.4881, "step": 68230 }, { "epoch": 3.69, "learning_rate": 3.15968841285297e-05, "loss": 1.52, "step": 68240 }, { "epoch": 3.69, "learning_rate": 3.159417937898951e-05, "loss": 1.551, "step": 68250 }, { "epoch": 3.69, "learning_rate": 3.159147462944931e-05, "loss": 1.4438, "step": 68260 }, { "epoch": 3.69, "learning_rate": 3.158876987990912e-05, "loss": 1.4703, "step": 68270 }, { "epoch": 3.69, "learning_rate": 3.158606513036893e-05, "loss": 1.568, "step": 68280 }, { "epoch": 3.69, "learning_rate": 3.1583360380828734e-05, "loss": 1.5429, "step": 68290 }, { "epoch": 3.69, "learning_rate": 3.1580655631288544e-05, "loss": 1.4362, "step": 68300 }, { "epoch": 3.69, "learning_rate": 3.1577950881748354e-05, "loss": 1.4657, "step": 68310 }, { "epoch": 3.69, "learning_rate": 3.1575246132208156e-05, "loss": 1.5437, "step": 68320 }, { "epoch": 3.69, "learning_rate": 3.1572541382667966e-05, "loss": 1.5003, "step": 68330 }, { "epoch": 3.69, "learning_rate": 3.1569836633127776e-05, "loss": 1.5062, "step": 68340 }, { "epoch": 3.69, "learning_rate": 3.156713188358758e-05, "loss": 1.5266, "step": 68350 }, { "epoch": 3.69, "learning_rate": 3.156442713404739e-05, "loss": 1.4742, "step": 68360 }, { "epoch": 3.69, "learning_rate": 3.15617223845072e-05, "loss": 1.5767, "step": 68370 }, { "epoch": 3.69, "learning_rate": 3.1559017634967e-05, "loss": 1.447, "step": 68380 }, { "epoch": 3.7, "learning_rate": 3.155631288542681e-05, "loss": 1.5475, "step": 68390 }, { "epoch": 3.7, "learning_rate": 3.155360813588662e-05, "loss": 1.4166, "step": 68400 }, { "epoch": 3.7, "learning_rate": 3.155090338634642e-05, "loss": 1.5351, "step": 68410 }, { "epoch": 3.7, "learning_rate": 3.154819863680623e-05, "loss": 1.4924, "step": 68420 }, { "epoch": 3.7, "learning_rate": 3.154549388726604e-05, "loss": 1.4987, "step": 68430 }, { "epoch": 3.7, "learning_rate": 3.1542789137725845e-05, "loss": 1.5203, "step": 68440 }, { "epoch": 3.7, "learning_rate": 3.1540084388185655e-05, "loss": 1.5449, "step": 68450 }, { "epoch": 3.7, "learning_rate": 3.1537379638645464e-05, "loss": 1.5296, "step": 68460 }, { "epoch": 3.7, "learning_rate": 3.153467488910527e-05, "loss": 1.4446, "step": 68470 }, { "epoch": 3.7, "learning_rate": 3.153197013956508e-05, "loss": 1.472, "step": 68480 }, { "epoch": 3.7, "learning_rate": 3.1529265390024887e-05, "loss": 1.5223, "step": 68490 }, { "epoch": 3.7, "learning_rate": 3.152656064048469e-05, "loss": 1.4624, "step": 68500 }, { "epoch": 3.7, "learning_rate": 3.15238558909445e-05, "loss": 1.4235, "step": 68510 }, { "epoch": 3.7, "learning_rate": 3.152115114140431e-05, "loss": 1.5354, "step": 68520 }, { "epoch": 3.7, "learning_rate": 3.151844639186411e-05, "loss": 1.4716, "step": 68530 }, { "epoch": 3.7, "learning_rate": 3.151574164232392e-05, "loss": 1.48, "step": 68540 }, { "epoch": 3.7, "learning_rate": 3.151303689278373e-05, "loss": 1.5037, "step": 68550 }, { "epoch": 3.7, "learning_rate": 3.151033214324354e-05, "loss": 1.4458, "step": 68560 }, { "epoch": 3.71, "learning_rate": 3.150762739370334e-05, "loss": 1.5208, "step": 68570 }, { "epoch": 3.71, "learning_rate": 3.150492264416315e-05, "loss": 1.5058, "step": 68580 }, { "epoch": 3.71, "learning_rate": 3.150221789462296e-05, "loss": 1.4919, "step": 68590 }, { "epoch": 3.71, "learning_rate": 3.1499513145082766e-05, "loss": 1.4529, "step": 68600 }, { "epoch": 3.71, "learning_rate": 3.1496808395542575e-05, "loss": 1.5398, "step": 68610 }, { "epoch": 3.71, "learning_rate": 3.1494103646002385e-05, "loss": 1.5021, "step": 68620 }, { "epoch": 3.71, "learning_rate": 3.149139889646219e-05, "loss": 1.4758, "step": 68630 }, { "epoch": 3.71, "learning_rate": 3.1488694146922e-05, "loss": 1.522, "step": 68640 }, { "epoch": 3.71, "learning_rate": 3.148598939738181e-05, "loss": 1.4408, "step": 68650 }, { "epoch": 3.71, "learning_rate": 3.148328464784161e-05, "loss": 1.5499, "step": 68660 }, { "epoch": 3.71, "learning_rate": 3.148057989830142e-05, "loss": 1.4905, "step": 68670 }, { "epoch": 3.71, "learning_rate": 3.147787514876123e-05, "loss": 1.4716, "step": 68680 }, { "epoch": 3.71, "learning_rate": 3.147517039922103e-05, "loss": 1.4454, "step": 68690 }, { "epoch": 3.71, "learning_rate": 3.147246564968084e-05, "loss": 1.48, "step": 68700 }, { "epoch": 3.71, "learning_rate": 3.146976090014065e-05, "loss": 1.5102, "step": 68710 }, { "epoch": 3.71, "learning_rate": 3.1467056150600454e-05, "loss": 1.4651, "step": 68720 }, { "epoch": 3.71, "learning_rate": 3.1464351401060264e-05, "loss": 1.462, "step": 68730 }, { "epoch": 3.71, "learning_rate": 3.1461646651520074e-05, "loss": 1.481, "step": 68740 }, { "epoch": 3.71, "learning_rate": 3.1458941901979876e-05, "loss": 1.5413, "step": 68750 }, { "epoch": 3.72, "learning_rate": 3.1456237152439686e-05, "loss": 1.5411, "step": 68760 }, { "epoch": 3.72, "learning_rate": 3.1453532402899496e-05, "loss": 1.442, "step": 68770 }, { "epoch": 3.72, "learning_rate": 3.14508276533593e-05, "loss": 1.5014, "step": 68780 }, { "epoch": 3.72, "learning_rate": 3.144812290381911e-05, "loss": 1.4347, "step": 68790 }, { "epoch": 3.72, "learning_rate": 3.144541815427892e-05, "loss": 1.5401, "step": 68800 }, { "epoch": 3.72, "learning_rate": 3.144271340473872e-05, "loss": 1.4549, "step": 68810 }, { "epoch": 3.72, "learning_rate": 3.144000865519853e-05, "loss": 1.5039, "step": 68820 }, { "epoch": 3.72, "learning_rate": 3.143730390565834e-05, "loss": 1.4046, "step": 68830 }, { "epoch": 3.72, "learning_rate": 3.143459915611814e-05, "loss": 1.5782, "step": 68840 }, { "epoch": 3.72, "learning_rate": 3.143189440657795e-05, "loss": 1.5592, "step": 68850 }, { "epoch": 3.72, "learning_rate": 3.142918965703776e-05, "loss": 1.4535, "step": 68860 }, { "epoch": 3.72, "learning_rate": 3.1426484907497565e-05, "loss": 1.5157, "step": 68870 }, { "epoch": 3.72, "learning_rate": 3.1423780157957375e-05, "loss": 1.5373, "step": 68880 }, { "epoch": 3.72, "learning_rate": 3.1421075408417184e-05, "loss": 1.4352, "step": 68890 }, { "epoch": 3.72, "learning_rate": 3.141837065887699e-05, "loss": 1.5151, "step": 68900 }, { "epoch": 3.72, "learning_rate": 3.14156659093368e-05, "loss": 1.4903, "step": 68910 }, { "epoch": 3.72, "learning_rate": 3.1412961159796607e-05, "loss": 1.4695, "step": 68920 }, { "epoch": 3.72, "learning_rate": 3.141025641025641e-05, "loss": 1.4791, "step": 68930 }, { "epoch": 3.73, "learning_rate": 3.140755166071622e-05, "loss": 1.4963, "step": 68940 }, { "epoch": 3.73, "learning_rate": 3.140484691117603e-05, "loss": 1.4813, "step": 68950 }, { "epoch": 3.73, "learning_rate": 3.140214216163583e-05, "loss": 1.4946, "step": 68960 }, { "epoch": 3.73, "learning_rate": 3.139943741209564e-05, "loss": 1.4986, "step": 68970 }, { "epoch": 3.73, "learning_rate": 3.139673266255545e-05, "loss": 1.4518, "step": 68980 }, { "epoch": 3.73, "learning_rate": 3.1394027913015254e-05, "loss": 1.4742, "step": 68990 }, { "epoch": 3.73, "learning_rate": 3.1391323163475063e-05, "loss": 1.5059, "step": 69000 }, { "epoch": 3.73, "learning_rate": 3.138861841393487e-05, "loss": 1.4535, "step": 69010 }, { "epoch": 3.73, "learning_rate": 3.1385913664394676e-05, "loss": 1.4672, "step": 69020 }, { "epoch": 3.73, "learning_rate": 3.1383208914854486e-05, "loss": 1.5013, "step": 69030 }, { "epoch": 3.73, "learning_rate": 3.1380504165314295e-05, "loss": 1.5154, "step": 69040 }, { "epoch": 3.73, "learning_rate": 3.13777994157741e-05, "loss": 1.5298, "step": 69050 }, { "epoch": 3.73, "learning_rate": 3.137509466623391e-05, "loss": 1.4774, "step": 69060 }, { "epoch": 3.73, "learning_rate": 3.137238991669372e-05, "loss": 1.5157, "step": 69070 }, { "epoch": 3.73, "learning_rate": 3.136968516715352e-05, "loss": 1.4604, "step": 69080 }, { "epoch": 3.73, "learning_rate": 3.136698041761333e-05, "loss": 1.5457, "step": 69090 }, { "epoch": 3.73, "learning_rate": 3.136427566807314e-05, "loss": 1.4897, "step": 69100 }, { "epoch": 3.73, "learning_rate": 3.136157091853294e-05, "loss": 1.5115, "step": 69110 }, { "epoch": 3.73, "learning_rate": 3.135886616899275e-05, "loss": 1.5205, "step": 69120 }, { "epoch": 3.74, "learning_rate": 3.135616141945256e-05, "loss": 1.4934, "step": 69130 }, { "epoch": 3.74, "learning_rate": 3.1353456669912365e-05, "loss": 1.4758, "step": 69140 }, { "epoch": 3.74, "learning_rate": 3.1350751920372174e-05, "loss": 1.5531, "step": 69150 }, { "epoch": 3.74, "learning_rate": 3.1348047170831984e-05, "loss": 1.5245, "step": 69160 }, { "epoch": 3.74, "learning_rate": 3.134534242129179e-05, "loss": 1.5336, "step": 69170 }, { "epoch": 3.74, "learning_rate": 3.1342637671751596e-05, "loss": 1.532, "step": 69180 }, { "epoch": 3.74, "learning_rate": 3.1339932922211406e-05, "loss": 1.5384, "step": 69190 }, { "epoch": 3.74, "learning_rate": 3.1337228172671216e-05, "loss": 1.4676, "step": 69200 }, { "epoch": 3.74, "learning_rate": 3.133452342313102e-05, "loss": 1.5203, "step": 69210 }, { "epoch": 3.74, "learning_rate": 3.133181867359083e-05, "loss": 1.5373, "step": 69220 }, { "epoch": 3.74, "learning_rate": 3.132911392405064e-05, "loss": 1.5204, "step": 69230 }, { "epoch": 3.74, "learning_rate": 3.132640917451044e-05, "loss": 1.5238, "step": 69240 }, { "epoch": 3.74, "learning_rate": 3.132370442497025e-05, "loss": 1.4753, "step": 69250 }, { "epoch": 3.74, "learning_rate": 3.132099967543006e-05, "loss": 1.5144, "step": 69260 }, { "epoch": 3.74, "learning_rate": 3.131829492588986e-05, "loss": 1.4872, "step": 69270 }, { "epoch": 3.74, "learning_rate": 3.131559017634967e-05, "loss": 1.5568, "step": 69280 }, { "epoch": 3.74, "learning_rate": 3.131288542680948e-05, "loss": 1.5448, "step": 69290 }, { "epoch": 3.74, "learning_rate": 3.1310180677269285e-05, "loss": 1.4791, "step": 69300 }, { "epoch": 3.75, "learning_rate": 3.1307475927729095e-05, "loss": 1.4955, "step": 69310 }, { "epoch": 3.75, "learning_rate": 3.1304771178188904e-05, "loss": 1.4881, "step": 69320 }, { "epoch": 3.75, "learning_rate": 3.130206642864871e-05, "loss": 1.52, "step": 69330 }, { "epoch": 3.75, "learning_rate": 3.129936167910852e-05, "loss": 1.5254, "step": 69340 }, { "epoch": 3.75, "learning_rate": 3.1296656929568327e-05, "loss": 1.4635, "step": 69350 }, { "epoch": 3.75, "learning_rate": 3.129395218002813e-05, "loss": 1.4545, "step": 69360 }, { "epoch": 3.75, "learning_rate": 3.129124743048794e-05, "loss": 1.4711, "step": 69370 }, { "epoch": 3.75, "learning_rate": 3.128854268094775e-05, "loss": 1.5295, "step": 69380 }, { "epoch": 3.75, "learning_rate": 3.128583793140755e-05, "loss": 1.5119, "step": 69390 }, { "epoch": 3.75, "learning_rate": 3.1283133181867354e-05, "loss": 1.5672, "step": 69400 }, { "epoch": 3.75, "learning_rate": 3.128042843232717e-05, "loss": 1.4657, "step": 69410 }, { "epoch": 3.75, "learning_rate": 3.1277723682786974e-05, "loss": 1.4554, "step": 69420 }, { "epoch": 3.75, "learning_rate": 3.1275018933246783e-05, "loss": 1.5152, "step": 69430 }, { "epoch": 3.75, "learning_rate": 3.127231418370659e-05, "loss": 1.5222, "step": 69440 }, { "epoch": 3.75, "learning_rate": 3.1269609434166396e-05, "loss": 1.4251, "step": 69450 }, { "epoch": 3.75, "learning_rate": 3.1266904684626206e-05, "loss": 1.5723, "step": 69460 }, { "epoch": 3.75, "learning_rate": 3.1264199935086015e-05, "loss": 1.455, "step": 69470 }, { "epoch": 3.75, "learning_rate": 3.126149518554582e-05, "loss": 1.4935, "step": 69480 }, { "epoch": 3.75, "learning_rate": 3.125879043600563e-05, "loss": 1.528, "step": 69490 }, { "epoch": 3.76, "learning_rate": 3.125608568646544e-05, "loss": 1.4263, "step": 69500 }, { "epoch": 3.76, "learning_rate": 3.125338093692524e-05, "loss": 1.41, "step": 69510 }, { "epoch": 3.76, "learning_rate": 3.125067618738505e-05, "loss": 1.5082, "step": 69520 }, { "epoch": 3.76, "learning_rate": 3.124797143784486e-05, "loss": 1.5119, "step": 69530 }, { "epoch": 3.76, "learning_rate": 3.124526668830466e-05, "loss": 1.4717, "step": 69540 }, { "epoch": 3.76, "learning_rate": 3.124256193876447e-05, "loss": 1.5133, "step": 69550 }, { "epoch": 3.76, "learning_rate": 3.123985718922428e-05, "loss": 1.4221, "step": 69560 }, { "epoch": 3.76, "learning_rate": 3.1237152439684085e-05, "loss": 1.4285, "step": 69570 }, { "epoch": 3.76, "learning_rate": 3.1234447690143894e-05, "loss": 1.5029, "step": 69580 }, { "epoch": 3.76, "learning_rate": 3.1231742940603704e-05, "loss": 1.4564, "step": 69590 }, { "epoch": 3.76, "learning_rate": 3.122903819106351e-05, "loss": 1.5041, "step": 69600 }, { "epoch": 3.76, "learning_rate": 3.1226333441523316e-05, "loss": 1.3561, "step": 69610 }, { "epoch": 3.76, "learning_rate": 3.1223628691983126e-05, "loss": 1.5074, "step": 69620 }, { "epoch": 3.76, "learning_rate": 3.122092394244293e-05, "loss": 1.4664, "step": 69630 }, { "epoch": 3.76, "learning_rate": 3.121821919290274e-05, "loss": 1.4775, "step": 69640 }, { "epoch": 3.76, "learning_rate": 3.121551444336255e-05, "loss": 1.5073, "step": 69650 }, { "epoch": 3.76, "learning_rate": 3.121280969382235e-05, "loss": 1.4817, "step": 69660 }, { "epoch": 3.76, "learning_rate": 3.121010494428216e-05, "loss": 1.497, "step": 69670 }, { "epoch": 3.77, "learning_rate": 3.120740019474197e-05, "loss": 1.4387, "step": 69680 }, { "epoch": 3.77, "learning_rate": 3.120469544520177e-05, "loss": 1.6199, "step": 69690 }, { "epoch": 3.77, "learning_rate": 3.120199069566158e-05, "loss": 1.5671, "step": 69700 }, { "epoch": 3.77, "learning_rate": 3.119928594612139e-05, "loss": 1.4373, "step": 69710 }, { "epoch": 3.77, "learning_rate": 3.1196581196581195e-05, "loss": 1.4824, "step": 69720 }, { "epoch": 3.77, "learning_rate": 3.1193876447041005e-05, "loss": 1.4969, "step": 69730 }, { "epoch": 3.77, "learning_rate": 3.1191171697500815e-05, "loss": 1.4602, "step": 69740 }, { "epoch": 3.77, "learning_rate": 3.118846694796062e-05, "loss": 1.6094, "step": 69750 }, { "epoch": 3.77, "learning_rate": 3.118576219842043e-05, "loss": 1.5512, "step": 69760 }, { "epoch": 3.77, "learning_rate": 3.118305744888024e-05, "loss": 1.4534, "step": 69770 }, { "epoch": 3.77, "learning_rate": 3.118035269934004e-05, "loss": 1.4772, "step": 69780 }, { "epoch": 3.77, "learning_rate": 3.117764794979985e-05, "loss": 1.5152, "step": 69790 }, { "epoch": 3.77, "learning_rate": 3.117494320025966e-05, "loss": 1.5163, "step": 69800 }, { "epoch": 3.77, "learning_rate": 3.117223845071946e-05, "loss": 1.5901, "step": 69810 }, { "epoch": 3.77, "learning_rate": 3.116953370117927e-05, "loss": 1.543, "step": 69820 }, { "epoch": 3.77, "learning_rate": 3.116682895163908e-05, "loss": 1.5115, "step": 69830 }, { "epoch": 3.77, "learning_rate": 3.116412420209889e-05, "loss": 1.5486, "step": 69840 }, { "epoch": 3.77, "learning_rate": 3.1161419452558694e-05, "loss": 1.5537, "step": 69850 }, { "epoch": 3.77, "learning_rate": 3.1158714703018503e-05, "loss": 1.5804, "step": 69860 }, { "epoch": 3.78, "learning_rate": 3.115600995347831e-05, "loss": 1.4388, "step": 69870 }, { "epoch": 3.78, "learning_rate": 3.1153305203938116e-05, "loss": 1.4405, "step": 69880 }, { "epoch": 3.78, "learning_rate": 3.1150600454397926e-05, "loss": 1.529, "step": 69890 }, { "epoch": 3.78, "learning_rate": 3.1147895704857735e-05, "loss": 1.5122, "step": 69900 }, { "epoch": 3.78, "learning_rate": 3.114519095531754e-05, "loss": 1.5604, "step": 69910 }, { "epoch": 3.78, "learning_rate": 3.114248620577735e-05, "loss": 1.4247, "step": 69920 }, { "epoch": 3.78, "learning_rate": 3.113978145623716e-05, "loss": 1.4859, "step": 69930 }, { "epoch": 3.78, "learning_rate": 3.113707670669696e-05, "loss": 1.5227, "step": 69940 }, { "epoch": 3.78, "learning_rate": 3.113437195715676e-05, "loss": 1.4929, "step": 69950 }, { "epoch": 3.78, "learning_rate": 3.113166720761658e-05, "loss": 1.6252, "step": 69960 }, { "epoch": 3.78, "learning_rate": 3.112896245807638e-05, "loss": 1.4747, "step": 69970 }, { "epoch": 3.78, "learning_rate": 3.1126257708536185e-05, "loss": 1.5318, "step": 69980 }, { "epoch": 3.78, "learning_rate": 3.1123552958996e-05, "loss": 1.4591, "step": 69990 }, { "epoch": 3.78, "learning_rate": 3.1120848209455805e-05, "loss": 1.5546, "step": 70000 }, { "epoch": 3.78, "learning_rate": 3.111814345991561e-05, "loss": 1.4081, "step": 70010 }, { "epoch": 3.78, "learning_rate": 3.1115438710375424e-05, "loss": 1.5571, "step": 70020 }, { "epoch": 3.78, "learning_rate": 3.111273396083523e-05, "loss": 1.4282, "step": 70030 }, { "epoch": 3.78, "learning_rate": 3.1110029211295036e-05, "loss": 1.4565, "step": 70040 }, { "epoch": 3.79, "learning_rate": 3.1107324461754846e-05, "loss": 1.444, "step": 70050 }, { "epoch": 3.79, "learning_rate": 3.110461971221465e-05, "loss": 1.5506, "step": 70060 }, { "epoch": 3.79, "learning_rate": 3.110191496267446e-05, "loss": 1.5375, "step": 70070 }, { "epoch": 3.79, "learning_rate": 3.109921021313427e-05, "loss": 1.4504, "step": 70080 }, { "epoch": 3.79, "learning_rate": 3.109650546359407e-05, "loss": 1.4167, "step": 70090 }, { "epoch": 3.79, "learning_rate": 3.109380071405388e-05, "loss": 1.4995, "step": 70100 }, { "epoch": 3.79, "learning_rate": 3.109109596451369e-05, "loss": 1.4949, "step": 70110 }, { "epoch": 3.79, "learning_rate": 3.108839121497349e-05, "loss": 1.4315, "step": 70120 }, { "epoch": 3.79, "learning_rate": 3.10856864654333e-05, "loss": 1.4912, "step": 70130 }, { "epoch": 3.79, "learning_rate": 3.108298171589311e-05, "loss": 1.4419, "step": 70140 }, { "epoch": 3.79, "learning_rate": 3.1080276966352915e-05, "loss": 1.4944, "step": 70150 }, { "epoch": 3.79, "learning_rate": 3.1077572216812725e-05, "loss": 1.5023, "step": 70160 }, { "epoch": 3.79, "learning_rate": 3.1074867467272535e-05, "loss": 1.4397, "step": 70170 }, { "epoch": 3.79, "learning_rate": 3.107216271773234e-05, "loss": 1.5088, "step": 70180 }, { "epoch": 3.79, "learning_rate": 3.106945796819215e-05, "loss": 1.485, "step": 70190 }, { "epoch": 3.79, "learning_rate": 3.106675321865196e-05, "loss": 1.404, "step": 70200 }, { "epoch": 3.79, "learning_rate": 3.106404846911176e-05, "loss": 1.4972, "step": 70210 }, { "epoch": 3.79, "learning_rate": 3.106134371957157e-05, "loss": 1.4316, "step": 70220 }, { "epoch": 3.79, "learning_rate": 3.105863897003138e-05, "loss": 1.5621, "step": 70230 }, { "epoch": 3.8, "learning_rate": 3.105593422049118e-05, "loss": 1.4904, "step": 70240 }, { "epoch": 3.8, "learning_rate": 3.105322947095099e-05, "loss": 1.5532, "step": 70250 }, { "epoch": 3.8, "learning_rate": 3.10505247214108e-05, "loss": 1.6097, "step": 70260 }, { "epoch": 3.8, "learning_rate": 3.1047819971870604e-05, "loss": 1.5056, "step": 70270 }, { "epoch": 3.8, "learning_rate": 3.1045115222330414e-05, "loss": 1.5032, "step": 70280 }, { "epoch": 3.8, "learning_rate": 3.1042410472790223e-05, "loss": 1.5316, "step": 70290 }, { "epoch": 3.8, "learning_rate": 3.1039705723250026e-05, "loss": 1.459, "step": 70300 }, { "epoch": 3.8, "learning_rate": 3.1037000973709836e-05, "loss": 1.5511, "step": 70310 }, { "epoch": 3.8, "learning_rate": 3.1034296224169646e-05, "loss": 1.5204, "step": 70320 }, { "epoch": 3.8, "learning_rate": 3.103159147462945e-05, "loss": 1.486, "step": 70330 }, { "epoch": 3.8, "learning_rate": 3.102888672508926e-05, "loss": 1.5024, "step": 70340 }, { "epoch": 3.8, "learning_rate": 3.102618197554907e-05, "loss": 1.6013, "step": 70350 }, { "epoch": 3.8, "learning_rate": 3.102347722600887e-05, "loss": 1.4772, "step": 70360 }, { "epoch": 3.8, "learning_rate": 3.102077247646868e-05, "loss": 1.5863, "step": 70370 }, { "epoch": 3.8, "learning_rate": 3.101806772692849e-05, "loss": 1.4254, "step": 70380 }, { "epoch": 3.8, "learning_rate": 3.101536297738829e-05, "loss": 1.4591, "step": 70390 }, { "epoch": 3.8, "learning_rate": 3.10126582278481e-05, "loss": 1.399, "step": 70400 }, { "epoch": 3.8, "learning_rate": 3.100995347830791e-05, "loss": 1.5173, "step": 70410 }, { "epoch": 3.81, "learning_rate": 3.1007248728767715e-05, "loss": 1.5447, "step": 70420 }, { "epoch": 3.81, "learning_rate": 3.1004543979227525e-05, "loss": 1.4248, "step": 70430 }, { "epoch": 3.81, "learning_rate": 3.1001839229687334e-05, "loss": 1.5456, "step": 70440 }, { "epoch": 3.81, "learning_rate": 3.0999134480147144e-05, "loss": 1.5073, "step": 70450 }, { "epoch": 3.81, "learning_rate": 3.099642973060695e-05, "loss": 1.4771, "step": 70460 }, { "epoch": 3.81, "learning_rate": 3.0993724981066756e-05, "loss": 1.3921, "step": 70470 }, { "epoch": 3.81, "learning_rate": 3.0991020231526566e-05, "loss": 1.4867, "step": 70480 }, { "epoch": 3.81, "learning_rate": 3.098831548198637e-05, "loss": 1.5337, "step": 70490 }, { "epoch": 3.81, "learning_rate": 3.098561073244618e-05, "loss": 1.5506, "step": 70500 }, { "epoch": 3.81, "learning_rate": 3.098290598290599e-05, "loss": 1.5177, "step": 70510 }, { "epoch": 3.81, "learning_rate": 3.098020123336579e-05, "loss": 1.5014, "step": 70520 }, { "epoch": 3.81, "learning_rate": 3.0977496483825594e-05, "loss": 1.4226, "step": 70530 }, { "epoch": 3.81, "learning_rate": 3.097479173428541e-05, "loss": 1.4699, "step": 70540 }, { "epoch": 3.81, "learning_rate": 3.097208698474521e-05, "loss": 1.5014, "step": 70550 }, { "epoch": 3.81, "learning_rate": 3.0969382235205016e-05, "loss": 1.495, "step": 70560 }, { "epoch": 3.81, "learning_rate": 3.096667748566483e-05, "loss": 1.6199, "step": 70570 }, { "epoch": 3.81, "learning_rate": 3.0963972736124636e-05, "loss": 1.4817, "step": 70580 }, { "epoch": 3.81, "learning_rate": 3.096126798658444e-05, "loss": 1.5912, "step": 70590 }, { "epoch": 3.81, "learning_rate": 3.0958563237044255e-05, "loss": 1.4712, "step": 70600 }, { "epoch": 3.82, "learning_rate": 3.095585848750406e-05, "loss": 1.4978, "step": 70610 }, { "epoch": 3.82, "learning_rate": 3.095315373796386e-05, "loss": 1.5675, "step": 70620 }, { "epoch": 3.82, "learning_rate": 3.095044898842368e-05, "loss": 1.5221, "step": 70630 }, { "epoch": 3.82, "learning_rate": 3.094774423888348e-05, "loss": 1.5362, "step": 70640 }, { "epoch": 3.82, "learning_rate": 3.094503948934328e-05, "loss": 1.4819, "step": 70650 }, { "epoch": 3.82, "learning_rate": 3.09423347398031e-05, "loss": 1.5028, "step": 70660 }, { "epoch": 3.82, "learning_rate": 3.09396299902629e-05, "loss": 1.5698, "step": 70670 }, { "epoch": 3.82, "learning_rate": 3.093692524072271e-05, "loss": 1.5227, "step": 70680 }, { "epoch": 3.82, "learning_rate": 3.093422049118252e-05, "loss": 1.457, "step": 70690 }, { "epoch": 3.82, "learning_rate": 3.0931515741642324e-05, "loss": 1.5054, "step": 70700 }, { "epoch": 3.82, "learning_rate": 3.0928810992102134e-05, "loss": 1.4651, "step": 70710 }, { "epoch": 3.82, "learning_rate": 3.0926106242561943e-05, "loss": 1.4874, "step": 70720 }, { "epoch": 3.82, "learning_rate": 3.0923401493021746e-05, "loss": 1.4655, "step": 70730 }, { "epoch": 3.82, "learning_rate": 3.0920696743481556e-05, "loss": 1.4527, "step": 70740 }, { "epoch": 3.82, "learning_rate": 3.0917991993941366e-05, "loss": 1.5041, "step": 70750 }, { "epoch": 3.82, "learning_rate": 3.091528724440117e-05, "loss": 1.4247, "step": 70760 }, { "epoch": 3.82, "learning_rate": 3.091258249486098e-05, "loss": 1.486, "step": 70770 }, { "epoch": 3.82, "learning_rate": 3.090987774532079e-05, "loss": 1.4421, "step": 70780 }, { "epoch": 3.83, "learning_rate": 3.090717299578059e-05, "loss": 1.4743, "step": 70790 }, { "epoch": 3.83, "learning_rate": 3.09044682462404e-05, "loss": 1.4996, "step": 70800 }, { "epoch": 3.83, "learning_rate": 3.090176349670021e-05, "loss": 1.436, "step": 70810 }, { "epoch": 3.83, "learning_rate": 3.089905874716001e-05, "loss": 1.51, "step": 70820 }, { "epoch": 3.83, "learning_rate": 3.089635399761982e-05, "loss": 1.4754, "step": 70830 }, { "epoch": 3.83, "learning_rate": 3.089364924807963e-05, "loss": 1.541, "step": 70840 }, { "epoch": 3.83, "learning_rate": 3.0890944498539435e-05, "loss": 1.5247, "step": 70850 }, { "epoch": 3.83, "learning_rate": 3.0888239748999245e-05, "loss": 1.5286, "step": 70860 }, { "epoch": 3.83, "learning_rate": 3.0885534999459054e-05, "loss": 1.495, "step": 70870 }, { "epoch": 3.83, "learning_rate": 3.088283024991886e-05, "loss": 1.5437, "step": 70880 }, { "epoch": 3.83, "learning_rate": 3.088012550037867e-05, "loss": 1.4534, "step": 70890 }, { "epoch": 3.83, "learning_rate": 3.0877420750838477e-05, "loss": 1.4234, "step": 70900 }, { "epoch": 3.83, "learning_rate": 3.087471600129828e-05, "loss": 1.4024, "step": 70910 }, { "epoch": 3.83, "learning_rate": 3.087201125175809e-05, "loss": 1.5703, "step": 70920 }, { "epoch": 3.83, "learning_rate": 3.08693065022179e-05, "loss": 1.4828, "step": 70930 }, { "epoch": 3.83, "learning_rate": 3.08666017526777e-05, "loss": 1.4956, "step": 70940 }, { "epoch": 3.83, "learning_rate": 3.086389700313751e-05, "loss": 1.5316, "step": 70950 }, { "epoch": 3.83, "learning_rate": 3.086119225359732e-05, "loss": 1.5862, "step": 70960 }, { "epoch": 3.83, "learning_rate": 3.0858487504057124e-05, "loss": 1.5609, "step": 70970 }, { "epoch": 3.84, "learning_rate": 3.085578275451693e-05, "loss": 1.4958, "step": 70980 }, { "epoch": 3.84, "learning_rate": 3.085307800497674e-05, "loss": 1.4909, "step": 70990 }, { "epoch": 3.84, "learning_rate": 3.0850373255436546e-05, "loss": 1.3792, "step": 71000 }, { "epoch": 3.84, "learning_rate": 3.0847668505896356e-05, "loss": 1.5082, "step": 71010 }, { "epoch": 3.84, "learning_rate": 3.0844963756356165e-05, "loss": 1.525, "step": 71020 }, { "epoch": 3.84, "learning_rate": 3.084225900681597e-05, "loss": 1.4893, "step": 71030 }, { "epoch": 3.84, "learning_rate": 3.083955425727578e-05, "loss": 1.4868, "step": 71040 }, { "epoch": 3.84, "learning_rate": 3.083684950773559e-05, "loss": 1.5646, "step": 71050 }, { "epoch": 3.84, "learning_rate": 3.083414475819539e-05, "loss": 1.4769, "step": 71060 }, { "epoch": 3.84, "learning_rate": 3.08314400086552e-05, "loss": 1.4986, "step": 71070 }, { "epoch": 3.84, "learning_rate": 3.0828735259115e-05, "loss": 1.6007, "step": 71080 }, { "epoch": 3.84, "learning_rate": 3.082603050957482e-05, "loss": 1.4618, "step": 71090 }, { "epoch": 3.84, "learning_rate": 3.082332576003462e-05, "loss": 1.4778, "step": 71100 }, { "epoch": 3.84, "learning_rate": 3.0820621010494425e-05, "loss": 1.5277, "step": 71110 }, { "epoch": 3.84, "learning_rate": 3.081791626095424e-05, "loss": 1.3957, "step": 71120 }, { "epoch": 3.84, "learning_rate": 3.0815211511414044e-05, "loss": 1.4547, "step": 71130 }, { "epoch": 3.84, "learning_rate": 3.081250676187385e-05, "loss": 1.5302, "step": 71140 }, { "epoch": 3.84, "learning_rate": 3.0809802012333664e-05, "loss": 1.5557, "step": 71150 }, { "epoch": 3.85, "learning_rate": 3.0807097262793466e-05, "loss": 1.4722, "step": 71160 }, { "epoch": 3.85, "learning_rate": 3.080439251325327e-05, "loss": 1.5691, "step": 71170 }, { "epoch": 3.85, "learning_rate": 3.0801687763713086e-05, "loss": 1.4866, "step": 71180 }, { "epoch": 3.85, "learning_rate": 3.079898301417289e-05, "loss": 1.5107, "step": 71190 }, { "epoch": 3.85, "learning_rate": 3.079627826463269e-05, "loss": 1.4696, "step": 71200 }, { "epoch": 3.85, "learning_rate": 3.079357351509251e-05, "loss": 1.4427, "step": 71210 }, { "epoch": 3.85, "learning_rate": 3.079086876555231e-05, "loss": 1.5463, "step": 71220 }, { "epoch": 3.85, "learning_rate": 3.0788164016012114e-05, "loss": 1.5151, "step": 71230 }, { "epoch": 3.85, "learning_rate": 3.078545926647193e-05, "loss": 1.522, "step": 71240 }, { "epoch": 3.85, "learning_rate": 3.078275451693173e-05, "loss": 1.5416, "step": 71250 }, { "epoch": 3.85, "learning_rate": 3.0780049767391536e-05, "loss": 1.5066, "step": 71260 }, { "epoch": 3.85, "learning_rate": 3.077734501785135e-05, "loss": 1.5139, "step": 71270 }, { "epoch": 3.85, "learning_rate": 3.0774640268311155e-05, "loss": 1.5692, "step": 71280 }, { "epoch": 3.85, "learning_rate": 3.077193551877096e-05, "loss": 1.5235, "step": 71290 }, { "epoch": 3.85, "learning_rate": 3.0769230769230774e-05, "loss": 1.4654, "step": 71300 }, { "epoch": 3.85, "learning_rate": 3.076652601969058e-05, "loss": 1.4834, "step": 71310 }, { "epoch": 3.85, "learning_rate": 3.076382127015039e-05, "loss": 1.5527, "step": 71320 }, { "epoch": 3.85, "learning_rate": 3.0761116520610197e-05, "loss": 1.4464, "step": 71330 }, { "epoch": 3.85, "learning_rate": 3.075841177107e-05, "loss": 1.5295, "step": 71340 }, { "epoch": 3.86, "learning_rate": 3.075570702152981e-05, "loss": 1.4191, "step": 71350 }, { "epoch": 3.86, "learning_rate": 3.075300227198962e-05, "loss": 1.4769, "step": 71360 }, { "epoch": 3.86, "learning_rate": 3.075029752244942e-05, "loss": 1.5325, "step": 71370 }, { "epoch": 3.86, "learning_rate": 3.074759277290923e-05, "loss": 1.5476, "step": 71380 }, { "epoch": 3.86, "learning_rate": 3.074488802336904e-05, "loss": 1.4725, "step": 71390 }, { "epoch": 3.86, "learning_rate": 3.0742183273828844e-05, "loss": 1.4951, "step": 71400 }, { "epoch": 3.86, "learning_rate": 3.073947852428865e-05, "loss": 1.4115, "step": 71410 }, { "epoch": 3.86, "learning_rate": 3.073677377474846e-05, "loss": 1.57, "step": 71420 }, { "epoch": 3.86, "learning_rate": 3.0734069025208266e-05, "loss": 1.4425, "step": 71430 }, { "epoch": 3.86, "learning_rate": 3.0731364275668076e-05, "loss": 1.4748, "step": 71440 }, { "epoch": 3.86, "learning_rate": 3.0728659526127885e-05, "loss": 1.4934, "step": 71450 }, { "epoch": 3.86, "learning_rate": 3.072595477658769e-05, "loss": 1.4718, "step": 71460 }, { "epoch": 3.86, "learning_rate": 3.07232500270475e-05, "loss": 1.5893, "step": 71470 }, { "epoch": 3.86, "learning_rate": 3.072054527750731e-05, "loss": 1.459, "step": 71480 }, { "epoch": 3.86, "learning_rate": 3.071784052796711e-05, "loss": 1.4717, "step": 71490 }, { "epoch": 3.86, "learning_rate": 3.071513577842692e-05, "loss": 1.4889, "step": 71500 }, { "epoch": 3.86, "learning_rate": 3.071243102888673e-05, "loss": 1.4109, "step": 71510 }, { "epoch": 3.86, "learning_rate": 3.070972627934653e-05, "loss": 1.5185, "step": 71520 }, { "epoch": 3.87, "learning_rate": 3.070702152980634e-05, "loss": 1.4558, "step": 71530 }, { "epoch": 3.87, "learning_rate": 3.070431678026615e-05, "loss": 1.4343, "step": 71540 }, { "epoch": 3.87, "learning_rate": 3.0701612030725955e-05, "loss": 1.4936, "step": 71550 }, { "epoch": 3.87, "learning_rate": 3.0698907281185764e-05, "loss": 1.5121, "step": 71560 }, { "epoch": 3.87, "learning_rate": 3.0696202531645574e-05, "loss": 1.4496, "step": 71570 }, { "epoch": 3.87, "learning_rate": 3.069349778210538e-05, "loss": 1.459, "step": 71580 }, { "epoch": 3.87, "learning_rate": 3.0690793032565186e-05, "loss": 1.5117, "step": 71590 }, { "epoch": 3.87, "learning_rate": 3.0688088283024996e-05, "loss": 1.5035, "step": 71600 }, { "epoch": 3.87, "learning_rate": 3.06853835334848e-05, "loss": 1.4395, "step": 71610 }, { "epoch": 3.87, "learning_rate": 3.068267878394461e-05, "loss": 1.4547, "step": 71620 }, { "epoch": 3.87, "learning_rate": 3.067997403440441e-05, "loss": 1.4501, "step": 71630 }, { "epoch": 3.87, "learning_rate": 3.067726928486422e-05, "loss": 1.5247, "step": 71640 }, { "epoch": 3.87, "learning_rate": 3.067456453532403e-05, "loss": 1.4735, "step": 71650 }, { "epoch": 3.87, "learning_rate": 3.0671859785783834e-05, "loss": 1.5553, "step": 71660 }, { "epoch": 3.87, "learning_rate": 3.066915503624364e-05, "loss": 1.45, "step": 71670 }, { "epoch": 3.87, "learning_rate": 3.066645028670345e-05, "loss": 1.5321, "step": 71680 }, { "epoch": 3.87, "learning_rate": 3.0663745537163256e-05, "loss": 1.589, "step": 71690 }, { "epoch": 3.87, "learning_rate": 3.0661040787623065e-05, "loss": 1.5043, "step": 71700 }, { "epoch": 3.87, "learning_rate": 3.0658336038082875e-05, "loss": 1.473, "step": 71710 }, { "epoch": 3.88, "learning_rate": 3.065563128854268e-05, "loss": 1.5204, "step": 71720 }, { "epoch": 3.88, "learning_rate": 3.0652926539002494e-05, "loss": 1.5694, "step": 71730 }, { "epoch": 3.88, "learning_rate": 3.06502217894623e-05, "loss": 1.5322, "step": 71740 }, { "epoch": 3.88, "learning_rate": 3.06475170399221e-05, "loss": 1.553, "step": 71750 }, { "epoch": 3.88, "learning_rate": 3.0644812290381917e-05, "loss": 1.5077, "step": 71760 }, { "epoch": 3.88, "learning_rate": 3.064210754084172e-05, "loss": 1.4913, "step": 71770 }, { "epoch": 3.88, "learning_rate": 3.063940279130152e-05, "loss": 1.5409, "step": 71780 }, { "epoch": 3.88, "learning_rate": 3.063669804176134e-05, "loss": 1.5426, "step": 71790 }, { "epoch": 3.88, "learning_rate": 3.063399329222114e-05, "loss": 1.4963, "step": 71800 }, { "epoch": 3.88, "learning_rate": 3.0631288542680944e-05, "loss": 1.5289, "step": 71810 }, { "epoch": 3.88, "learning_rate": 3.062858379314076e-05, "loss": 1.5027, "step": 71820 }, { "epoch": 3.88, "learning_rate": 3.0625879043600564e-05, "loss": 1.4906, "step": 71830 }, { "epoch": 3.88, "learning_rate": 3.062317429406037e-05, "loss": 1.5557, "step": 71840 }, { "epoch": 3.88, "learning_rate": 3.062046954452018e-05, "loss": 1.6278, "step": 71850 }, { "epoch": 3.88, "learning_rate": 3.0617764794979986e-05, "loss": 1.4809, "step": 71860 }, { "epoch": 3.88, "learning_rate": 3.061506004543979e-05, "loss": 1.5495, "step": 71870 }, { "epoch": 3.88, "learning_rate": 3.0612355295899605e-05, "loss": 1.4615, "step": 71880 }, { "epoch": 3.88, "learning_rate": 3.060965054635941e-05, "loss": 1.5099, "step": 71890 }, { "epoch": 3.89, "learning_rate": 3.060694579681921e-05, "loss": 1.4767, "step": 71900 }, { "epoch": 3.89, "learning_rate": 3.060424104727903e-05, "loss": 1.4935, "step": 71910 }, { "epoch": 3.89, "learning_rate": 3.060153629773883e-05, "loss": 1.5499, "step": 71920 }, { "epoch": 3.89, "learning_rate": 3.059883154819863e-05, "loss": 1.5068, "step": 71930 }, { "epoch": 3.89, "learning_rate": 3.059612679865845e-05, "loss": 1.4047, "step": 71940 }, { "epoch": 3.89, "learning_rate": 3.059342204911825e-05, "loss": 1.4807, "step": 71950 }, { "epoch": 3.89, "learning_rate": 3.059071729957806e-05, "loss": 1.52, "step": 71960 }, { "epoch": 3.89, "learning_rate": 3.058801255003787e-05, "loss": 1.4901, "step": 71970 }, { "epoch": 3.89, "learning_rate": 3.0585307800497675e-05, "loss": 1.4994, "step": 71980 }, { "epoch": 3.89, "learning_rate": 3.0582603050957484e-05, "loss": 1.518, "step": 71990 }, { "epoch": 3.89, "learning_rate": 3.0579898301417294e-05, "loss": 1.4553, "step": 72000 }, { "epoch": 3.89, "learning_rate": 3.05771935518771e-05, "loss": 1.4933, "step": 72010 }, { "epoch": 3.89, "learning_rate": 3.0574488802336906e-05, "loss": 1.5275, "step": 72020 }, { "epoch": 3.89, "learning_rate": 3.0571784052796716e-05, "loss": 1.5281, "step": 72030 }, { "epoch": 3.89, "learning_rate": 3.056907930325652e-05, "loss": 1.5704, "step": 72040 }, { "epoch": 3.89, "learning_rate": 3.056637455371633e-05, "loss": 1.583, "step": 72050 }, { "epoch": 3.89, "learning_rate": 3.056366980417614e-05, "loss": 1.5058, "step": 72060 }, { "epoch": 3.89, "learning_rate": 3.056096505463594e-05, "loss": 1.6202, "step": 72070 }, { "epoch": 3.89, "learning_rate": 3.055826030509575e-05, "loss": 1.4898, "step": 72080 }, { "epoch": 3.9, "learning_rate": 3.055555555555556e-05, "loss": 1.4841, "step": 72090 }, { "epoch": 3.9, "learning_rate": 3.055285080601536e-05, "loss": 1.5772, "step": 72100 }, { "epoch": 3.9, "learning_rate": 3.055014605647517e-05, "loss": 1.4744, "step": 72110 }, { "epoch": 3.9, "learning_rate": 3.054744130693498e-05, "loss": 1.4252, "step": 72120 }, { "epoch": 3.9, "learning_rate": 3.0544736557394785e-05, "loss": 1.5699, "step": 72130 }, { "epoch": 3.9, "learning_rate": 3.0542031807854595e-05, "loss": 1.5014, "step": 72140 }, { "epoch": 3.9, "learning_rate": 3.0539327058314405e-05, "loss": 1.5706, "step": 72150 }, { "epoch": 3.9, "learning_rate": 3.053662230877421e-05, "loss": 1.4568, "step": 72160 }, { "epoch": 3.9, "learning_rate": 3.053391755923402e-05, "loss": 1.5291, "step": 72170 }, { "epoch": 3.9, "learning_rate": 3.053121280969383e-05, "loss": 1.4861, "step": 72180 }, { "epoch": 3.9, "learning_rate": 3.052850806015363e-05, "loss": 1.4438, "step": 72190 }, { "epoch": 3.9, "learning_rate": 3.052580331061344e-05, "loss": 1.4181, "step": 72200 }, { "epoch": 3.9, "learning_rate": 3.052309856107324e-05, "loss": 1.5487, "step": 72210 }, { "epoch": 3.9, "learning_rate": 3.052039381153305e-05, "loss": 1.4426, "step": 72220 }, { "epoch": 3.9, "learning_rate": 3.051768906199286e-05, "loss": 1.4832, "step": 72230 }, { "epoch": 3.9, "learning_rate": 3.0514984312452664e-05, "loss": 1.4303, "step": 72240 }, { "epoch": 3.9, "learning_rate": 3.0512279562912478e-05, "loss": 1.4734, "step": 72250 }, { "epoch": 3.9, "learning_rate": 3.0509574813372284e-05, "loss": 1.5481, "step": 72260 }, { "epoch": 3.91, "learning_rate": 3.0506870063832087e-05, "loss": 1.5402, "step": 72270 }, { "epoch": 3.91, "learning_rate": 3.05041653142919e-05, "loss": 1.5277, "step": 72280 }, { "epoch": 3.91, "learning_rate": 3.0501460564751706e-05, "loss": 1.4794, "step": 72290 }, { "epoch": 3.91, "learning_rate": 3.049875581521151e-05, "loss": 1.4035, "step": 72300 }, { "epoch": 3.91, "learning_rate": 3.0496051065671322e-05, "loss": 1.5043, "step": 72310 }, { "epoch": 3.91, "learning_rate": 3.0493346316131128e-05, "loss": 1.4833, "step": 72320 }, { "epoch": 3.91, "learning_rate": 3.0490641566590934e-05, "loss": 1.5586, "step": 72330 }, { "epoch": 3.91, "learning_rate": 3.0487936817050744e-05, "loss": 1.5579, "step": 72340 }, { "epoch": 3.91, "learning_rate": 3.048523206751055e-05, "loss": 1.499, "step": 72350 }, { "epoch": 3.91, "learning_rate": 3.0482527317970357e-05, "loss": 1.5296, "step": 72360 }, { "epoch": 3.91, "learning_rate": 3.0479822568430166e-05, "loss": 1.3595, "step": 72370 }, { "epoch": 3.91, "learning_rate": 3.0477117818889972e-05, "loss": 1.4295, "step": 72380 }, { "epoch": 3.91, "learning_rate": 3.047441306934978e-05, "loss": 1.5406, "step": 72390 }, { "epoch": 3.91, "learning_rate": 3.047170831980959e-05, "loss": 1.4606, "step": 72400 }, { "epoch": 3.91, "learning_rate": 3.0469003570269395e-05, "loss": 1.521, "step": 72410 }, { "epoch": 3.91, "learning_rate": 3.04662988207292e-05, "loss": 1.4443, "step": 72420 }, { "epoch": 3.91, "learning_rate": 3.046359407118901e-05, "loss": 1.4492, "step": 72430 }, { "epoch": 3.91, "learning_rate": 3.0460889321648817e-05, "loss": 1.4687, "step": 72440 }, { "epoch": 3.91, "learning_rate": 3.0458184572108623e-05, "loss": 1.4615, "step": 72450 }, { "epoch": 3.92, "learning_rate": 3.0455479822568433e-05, "loss": 1.4972, "step": 72460 }, { "epoch": 3.92, "learning_rate": 3.045277507302824e-05, "loss": 1.4001, "step": 72470 }, { "epoch": 3.92, "learning_rate": 3.0450070323488045e-05, "loss": 1.5202, "step": 72480 }, { "epoch": 3.92, "learning_rate": 3.0447365573947855e-05, "loss": 1.4444, "step": 72490 }, { "epoch": 3.92, "learning_rate": 3.044466082440766e-05, "loss": 1.4548, "step": 72500 }, { "epoch": 3.92, "learning_rate": 3.0441956074867467e-05, "loss": 1.5045, "step": 72510 }, { "epoch": 3.92, "learning_rate": 3.0439251325327277e-05, "loss": 1.5223, "step": 72520 }, { "epoch": 3.92, "learning_rate": 3.0436546575787083e-05, "loss": 1.5291, "step": 72530 }, { "epoch": 3.92, "learning_rate": 3.043384182624689e-05, "loss": 1.4945, "step": 72540 }, { "epoch": 3.92, "learning_rate": 3.04311370767067e-05, "loss": 1.5481, "step": 72550 }, { "epoch": 3.92, "learning_rate": 3.0428432327166505e-05, "loss": 1.5306, "step": 72560 }, { "epoch": 3.92, "learning_rate": 3.0425727577626312e-05, "loss": 1.4915, "step": 72570 }, { "epoch": 3.92, "learning_rate": 3.042302282808612e-05, "loss": 1.458, "step": 72580 }, { "epoch": 3.92, "learning_rate": 3.0420318078545928e-05, "loss": 1.526, "step": 72590 }, { "epoch": 3.92, "learning_rate": 3.0417613329005734e-05, "loss": 1.462, "step": 72600 }, { "epoch": 3.92, "learning_rate": 3.0414908579465547e-05, "loss": 1.4366, "step": 72610 }, { "epoch": 3.92, "learning_rate": 3.041220382992535e-05, "loss": 1.5117, "step": 72620 }, { "epoch": 3.92, "learning_rate": 3.0409499080385156e-05, "loss": 1.4751, "step": 72630 }, { "epoch": 3.93, "learning_rate": 3.040679433084497e-05, "loss": 1.5644, "step": 72640 }, { "epoch": 3.93, "learning_rate": 3.0404089581304772e-05, "loss": 1.5822, "step": 72650 }, { "epoch": 3.93, "learning_rate": 3.0401384831764578e-05, "loss": 1.5939, "step": 72660 }, { "epoch": 3.93, "learning_rate": 3.039868008222439e-05, "loss": 1.4613, "step": 72670 }, { "epoch": 3.93, "learning_rate": 3.0395975332684194e-05, "loss": 1.4936, "step": 72680 }, { "epoch": 3.93, "learning_rate": 3.0393270583144e-05, "loss": 1.5153, "step": 72690 }, { "epoch": 3.93, "learning_rate": 3.0390565833603813e-05, "loss": 1.4974, "step": 72700 }, { "epoch": 3.93, "learning_rate": 3.0387861084063616e-05, "loss": 1.4311, "step": 72710 }, { "epoch": 3.93, "learning_rate": 3.0385156334523423e-05, "loss": 1.5568, "step": 72720 }, { "epoch": 3.93, "learning_rate": 3.0382451584983236e-05, "loss": 1.4259, "step": 72730 }, { "epoch": 3.93, "learning_rate": 3.0379746835443042e-05, "loss": 1.44, "step": 72740 }, { "epoch": 3.93, "learning_rate": 3.0377042085902845e-05, "loss": 1.4753, "step": 72750 }, { "epoch": 3.93, "learning_rate": 3.037433733636265e-05, "loss": 1.5461, "step": 72760 }, { "epoch": 3.93, "learning_rate": 3.0371632586822464e-05, "loss": 1.4717, "step": 72770 }, { "epoch": 3.93, "learning_rate": 3.0368927837282267e-05, "loss": 1.4495, "step": 72780 }, { "epoch": 3.93, "learning_rate": 3.0366223087742073e-05, "loss": 1.4382, "step": 72790 }, { "epoch": 3.93, "learning_rate": 3.0363518338201886e-05, "loss": 1.4795, "step": 72800 }, { "epoch": 3.93, "learning_rate": 3.036081358866169e-05, "loss": 1.4506, "step": 72810 }, { "epoch": 3.93, "learning_rate": 3.0358108839121495e-05, "loss": 1.4599, "step": 72820 }, { "epoch": 3.94, "learning_rate": 3.035540408958131e-05, "loss": 1.485, "step": 72830 }, { "epoch": 3.94, "learning_rate": 3.0352699340041115e-05, "loss": 1.487, "step": 72840 }, { "epoch": 3.94, "learning_rate": 3.0349994590500918e-05, "loss": 1.5219, "step": 72850 }, { "epoch": 3.94, "learning_rate": 3.034728984096073e-05, "loss": 1.4985, "step": 72860 }, { "epoch": 3.94, "learning_rate": 3.0344585091420537e-05, "loss": 1.436, "step": 72870 }, { "epoch": 3.94, "learning_rate": 3.034188034188034e-05, "loss": 1.4806, "step": 72880 }, { "epoch": 3.94, "learning_rate": 3.0339175592340153e-05, "loss": 1.5173, "step": 72890 }, { "epoch": 3.94, "learning_rate": 3.033647084279996e-05, "loss": 1.4783, "step": 72900 }, { "epoch": 3.94, "learning_rate": 3.0333766093259762e-05, "loss": 1.5735, "step": 72910 }, { "epoch": 3.94, "learning_rate": 3.0331061343719575e-05, "loss": 1.4863, "step": 72920 }, { "epoch": 3.94, "learning_rate": 3.032835659417938e-05, "loss": 1.538, "step": 72930 }, { "epoch": 3.94, "learning_rate": 3.0325651844639187e-05, "loss": 1.5092, "step": 72940 }, { "epoch": 3.94, "learning_rate": 3.0322947095098997e-05, "loss": 1.5021, "step": 72950 }, { "epoch": 3.94, "learning_rate": 3.0320242345558803e-05, "loss": 1.4863, "step": 72960 }, { "epoch": 3.94, "learning_rate": 3.031753759601861e-05, "loss": 1.4648, "step": 72970 }, { "epoch": 3.94, "learning_rate": 3.031483284647842e-05, "loss": 1.4814, "step": 72980 }, { "epoch": 3.94, "learning_rate": 3.0312128096938225e-05, "loss": 1.5388, "step": 72990 }, { "epoch": 3.94, "learning_rate": 3.0309423347398032e-05, "loss": 1.5072, "step": 73000 }, { "epoch": 3.95, "learning_rate": 3.030671859785784e-05, "loss": 1.529, "step": 73010 }, { "epoch": 3.95, "learning_rate": 3.0304013848317648e-05, "loss": 1.4538, "step": 73020 }, { "epoch": 3.95, "learning_rate": 3.0301309098777454e-05, "loss": 1.4242, "step": 73030 }, { "epoch": 3.95, "learning_rate": 3.0298604349237264e-05, "loss": 1.5063, "step": 73040 }, { "epoch": 3.95, "learning_rate": 3.029589959969707e-05, "loss": 1.4977, "step": 73050 }, { "epoch": 3.95, "learning_rate": 3.0293194850156876e-05, "loss": 1.4183, "step": 73060 }, { "epoch": 3.95, "learning_rate": 3.0290490100616686e-05, "loss": 1.5251, "step": 73070 }, { "epoch": 3.95, "learning_rate": 3.0287785351076492e-05, "loss": 1.5836, "step": 73080 }, { "epoch": 3.95, "learning_rate": 3.0285080601536298e-05, "loss": 1.4677, "step": 73090 }, { "epoch": 3.95, "learning_rate": 3.0282375851996108e-05, "loss": 1.5465, "step": 73100 }, { "epoch": 3.95, "learning_rate": 3.0279671102455914e-05, "loss": 1.5309, "step": 73110 }, { "epoch": 3.95, "learning_rate": 3.027696635291572e-05, "loss": 1.5405, "step": 73120 }, { "epoch": 3.95, "learning_rate": 3.027426160337553e-05, "loss": 1.4684, "step": 73130 }, { "epoch": 3.95, "learning_rate": 3.0271556853835336e-05, "loss": 1.4611, "step": 73140 }, { "epoch": 3.95, "learning_rate": 3.0268852104295143e-05, "loss": 1.4577, "step": 73150 }, { "epoch": 3.95, "learning_rate": 3.0266147354754952e-05, "loss": 1.6054, "step": 73160 }, { "epoch": 3.95, "learning_rate": 3.026344260521476e-05, "loss": 1.5092, "step": 73170 }, { "epoch": 3.95, "learning_rate": 3.0260737855674565e-05, "loss": 1.4413, "step": 73180 }, { "epoch": 3.95, "learning_rate": 3.0258033106134374e-05, "loss": 1.5407, "step": 73190 }, { "epoch": 3.96, "learning_rate": 3.025532835659418e-05, "loss": 1.5662, "step": 73200 }, { "epoch": 3.96, "learning_rate": 3.0252623607053987e-05, "loss": 1.4089, "step": 73210 }, { "epoch": 3.96, "learning_rate": 3.0249918857513797e-05, "loss": 1.4988, "step": 73220 }, { "epoch": 3.96, "learning_rate": 3.0247214107973603e-05, "loss": 1.4908, "step": 73230 }, { "epoch": 3.96, "learning_rate": 3.024450935843341e-05, "loss": 1.5088, "step": 73240 }, { "epoch": 3.96, "learning_rate": 3.0241804608893222e-05, "loss": 1.5274, "step": 73250 }, { "epoch": 3.96, "learning_rate": 3.0239099859353025e-05, "loss": 1.5027, "step": 73260 }, { "epoch": 3.96, "learning_rate": 3.023639510981283e-05, "loss": 1.5122, "step": 73270 }, { "epoch": 3.96, "learning_rate": 3.0233690360272644e-05, "loss": 1.5157, "step": 73280 }, { "epoch": 3.96, "learning_rate": 3.0230985610732447e-05, "loss": 1.3899, "step": 73290 }, { "epoch": 3.96, "learning_rate": 3.0228280861192253e-05, "loss": 1.4714, "step": 73300 }, { "epoch": 3.96, "learning_rate": 3.022557611165206e-05, "loss": 1.4932, "step": 73310 }, { "epoch": 3.96, "learning_rate": 3.022287136211187e-05, "loss": 1.4746, "step": 73320 }, { "epoch": 3.96, "learning_rate": 3.0220166612571676e-05, "loss": 1.4579, "step": 73330 }, { "epoch": 3.96, "learning_rate": 3.0217461863031482e-05, "loss": 1.518, "step": 73340 }, { "epoch": 3.96, "learning_rate": 3.0214757113491295e-05, "loss": 1.5137, "step": 73350 }, { "epoch": 3.96, "learning_rate": 3.0212052363951098e-05, "loss": 1.5368, "step": 73360 }, { "epoch": 3.96, "learning_rate": 3.0209347614410904e-05, "loss": 1.4513, "step": 73370 }, { "epoch": 3.97, "learning_rate": 3.0206642864870717e-05, "loss": 1.4124, "step": 73380 }, { "epoch": 3.97, "learning_rate": 3.020393811533052e-05, "loss": 1.445, "step": 73390 }, { "epoch": 3.97, "learning_rate": 3.0201233365790326e-05, "loss": 1.424, "step": 73400 }, { "epoch": 3.97, "learning_rate": 3.019852861625014e-05, "loss": 1.4291, "step": 73410 }, { "epoch": 3.97, "learning_rate": 3.0195823866709942e-05, "loss": 1.4568, "step": 73420 }, { "epoch": 3.97, "learning_rate": 3.019311911716975e-05, "loss": 1.5047, "step": 73430 }, { "epoch": 3.97, "learning_rate": 3.019041436762956e-05, "loss": 1.4195, "step": 73440 }, { "epoch": 3.97, "learning_rate": 3.0187709618089364e-05, "loss": 1.511, "step": 73450 }, { "epoch": 3.97, "learning_rate": 3.018500486854917e-05, "loss": 1.5456, "step": 73460 }, { "epoch": 3.97, "learning_rate": 3.0182300119008984e-05, "loss": 1.4193, "step": 73470 }, { "epoch": 3.97, "learning_rate": 3.017959536946879e-05, "loss": 1.5088, "step": 73480 }, { "epoch": 3.97, "learning_rate": 3.0176890619928593e-05, "loss": 1.5145, "step": 73490 }, { "epoch": 3.97, "learning_rate": 3.0174185870388406e-05, "loss": 1.5335, "step": 73500 }, { "epoch": 3.97, "learning_rate": 3.0171481120848212e-05, "loss": 1.4474, "step": 73510 }, { "epoch": 3.97, "learning_rate": 3.0168776371308015e-05, "loss": 1.5728, "step": 73520 }, { "epoch": 3.97, "learning_rate": 3.0166071621767828e-05, "loss": 1.5134, "step": 73530 }, { "epoch": 3.97, "learning_rate": 3.0163366872227634e-05, "loss": 1.4835, "step": 73540 }, { "epoch": 3.97, "learning_rate": 3.0160662122687437e-05, "loss": 1.4707, "step": 73550 }, { "epoch": 3.97, "learning_rate": 3.015795737314725e-05, "loss": 1.4404, "step": 73560 }, { "epoch": 3.98, "learning_rate": 3.0155252623607056e-05, "loss": 1.5203, "step": 73570 }, { "epoch": 3.98, "learning_rate": 3.0152547874066863e-05, "loss": 1.4831, "step": 73580 }, { "epoch": 3.98, "learning_rate": 3.0149843124526672e-05, "loss": 1.4311, "step": 73590 }, { "epoch": 3.98, "learning_rate": 3.014713837498648e-05, "loss": 1.4197, "step": 73600 }, { "epoch": 3.98, "learning_rate": 3.0144433625446285e-05, "loss": 1.4916, "step": 73610 }, { "epoch": 3.98, "learning_rate": 3.0141728875906094e-05, "loss": 1.4602, "step": 73620 }, { "epoch": 3.98, "learning_rate": 3.01390241263659e-05, "loss": 1.4312, "step": 73630 }, { "epoch": 3.98, "learning_rate": 3.0136319376825707e-05, "loss": 1.5007, "step": 73640 }, { "epoch": 3.98, "learning_rate": 3.0133614627285517e-05, "loss": 1.4615, "step": 73650 }, { "epoch": 3.98, "learning_rate": 3.0130909877745323e-05, "loss": 1.4248, "step": 73660 }, { "epoch": 3.98, "learning_rate": 3.012820512820513e-05, "loss": 1.5182, "step": 73670 }, { "epoch": 3.98, "learning_rate": 3.012550037866494e-05, "loss": 1.5278, "step": 73680 }, { "epoch": 3.98, "learning_rate": 3.0122795629124745e-05, "loss": 1.544, "step": 73690 }, { "epoch": 3.98, "learning_rate": 3.012009087958455e-05, "loss": 1.5125, "step": 73700 }, { "epoch": 3.98, "learning_rate": 3.011738613004436e-05, "loss": 1.4389, "step": 73710 }, { "epoch": 3.98, "learning_rate": 3.0114681380504167e-05, "loss": 1.5021, "step": 73720 }, { "epoch": 3.98, "learning_rate": 3.0111976630963973e-05, "loss": 1.4668, "step": 73730 }, { "epoch": 3.98, "learning_rate": 3.0109271881423783e-05, "loss": 1.505, "step": 73740 }, { "epoch": 3.99, "learning_rate": 3.010656713188359e-05, "loss": 1.4668, "step": 73750 }, { "epoch": 3.99, "learning_rate": 3.0103862382343396e-05, "loss": 1.588, "step": 73760 }, { "epoch": 3.99, "learning_rate": 3.0101157632803205e-05, "loss": 1.4762, "step": 73770 }, { "epoch": 3.99, "learning_rate": 3.009845288326301e-05, "loss": 1.5444, "step": 73780 }, { "epoch": 3.99, "learning_rate": 3.0095748133722818e-05, "loss": 1.5251, "step": 73790 }, { "epoch": 3.99, "learning_rate": 3.0093043384182627e-05, "loss": 1.4177, "step": 73800 }, { "epoch": 3.99, "learning_rate": 3.0090338634642434e-05, "loss": 1.5053, "step": 73810 }, { "epoch": 3.99, "learning_rate": 3.008763388510224e-05, "loss": 1.5594, "step": 73820 }, { "epoch": 3.99, "learning_rate": 3.008492913556205e-05, "loss": 1.4487, "step": 73830 }, { "epoch": 3.99, "learning_rate": 3.0082224386021856e-05, "loss": 1.427, "step": 73840 }, { "epoch": 3.99, "learning_rate": 3.0079519636481662e-05, "loss": 1.4145, "step": 73850 }, { "epoch": 3.99, "learning_rate": 3.0076814886941472e-05, "loss": 1.499, "step": 73860 }, { "epoch": 3.99, "learning_rate": 3.0074110137401278e-05, "loss": 1.5002, "step": 73870 }, { "epoch": 3.99, "learning_rate": 3.0071405387861084e-05, "loss": 1.3829, "step": 73880 }, { "epoch": 3.99, "learning_rate": 3.006870063832089e-05, "loss": 1.4805, "step": 73890 }, { "epoch": 3.99, "learning_rate": 3.00659958887807e-05, "loss": 1.5672, "step": 73900 }, { "epoch": 3.99, "learning_rate": 3.0063291139240506e-05, "loss": 1.4601, "step": 73910 }, { "epoch": 3.99, "learning_rate": 3.0060586389700313e-05, "loss": 1.5785, "step": 73920 }, { "epoch": 3.99, "learning_rate": 3.0057881640160122e-05, "loss": 1.4496, "step": 73930 }, { "epoch": 4.0, "learning_rate": 3.005517689061993e-05, "loss": 1.4753, "step": 73940 }, { "epoch": 4.0, "learning_rate": 3.0052472141079735e-05, "loss": 1.4885, "step": 73950 }, { "epoch": 4.0, "learning_rate": 3.0049767391539545e-05, "loss": 1.4817, "step": 73960 }, { "epoch": 4.0, "learning_rate": 3.004706264199935e-05, "loss": 1.4946, "step": 73970 }, { "epoch": 4.0, "learning_rate": 3.0044357892459157e-05, "loss": 1.4787, "step": 73980 }, { "epoch": 4.0, "learning_rate": 3.004165314291897e-05, "loss": 1.5246, "step": 73990 }, { "epoch": 4.0, "learning_rate": 3.0038948393378773e-05, "loss": 1.4977, "step": 74000 }, { "epoch": 4.0, "learning_rate": 3.003624364383858e-05, "loss": 1.5379, "step": 74010 }, { "epoch": 4.0, "learning_rate": 3.0033538894298392e-05, "loss": 1.4585, "step": 74020 }, { "epoch": 4.0, "learning_rate": 3.0030834144758195e-05, "loss": 1.4754, "step": 74030 }, { "epoch": 4.0, "learning_rate": 3.0028129395218e-05, "loss": 1.3844, "step": 74040 }, { "epoch": 4.0, "learning_rate": 3.0025424645677814e-05, "loss": 1.3405, "step": 74050 }, { "epoch": 4.0, "learning_rate": 3.0022719896137617e-05, "loss": 1.361, "step": 74060 }, { "epoch": 4.0, "learning_rate": 3.0020015146597424e-05, "loss": 1.3576, "step": 74070 }, { "epoch": 4.0, "learning_rate": 3.0017310397057237e-05, "loss": 1.4186, "step": 74080 }, { "epoch": 4.0, "learning_rate": 3.001460564751704e-05, "loss": 1.3517, "step": 74090 }, { "epoch": 4.0, "learning_rate": 3.0011900897976846e-05, "loss": 1.4682, "step": 74100 }, { "epoch": 4.0, "learning_rate": 3.000919614843666e-05, "loss": 1.407, "step": 74110 }, { "epoch": 4.01, "learning_rate": 3.0006491398896465e-05, "loss": 1.355, "step": 74120 }, { "epoch": 4.01, "learning_rate": 3.0003786649356268e-05, "loss": 1.3817, "step": 74130 }, { "epoch": 4.01, "learning_rate": 3.000108189981608e-05, "loss": 1.357, "step": 74140 }, { "epoch": 4.01, "learning_rate": 2.9998377150275887e-05, "loss": 1.3427, "step": 74150 }, { "epoch": 4.01, "learning_rate": 2.999567240073569e-05, "loss": 1.3778, "step": 74160 }, { "epoch": 4.01, "learning_rate": 2.9992967651195503e-05, "loss": 1.3739, "step": 74170 }, { "epoch": 4.01, "learning_rate": 2.999026290165531e-05, "loss": 1.4037, "step": 74180 }, { "epoch": 4.01, "learning_rate": 2.9987558152115112e-05, "loss": 1.39, "step": 74190 }, { "epoch": 4.01, "learning_rate": 2.9984853402574925e-05, "loss": 1.2788, "step": 74200 }, { "epoch": 4.01, "learning_rate": 2.998214865303473e-05, "loss": 1.3402, "step": 74210 }, { "epoch": 4.01, "learning_rate": 2.9979443903494538e-05, "loss": 1.2905, "step": 74220 }, { "epoch": 4.01, "learning_rate": 2.9976739153954347e-05, "loss": 1.348, "step": 74230 }, { "epoch": 4.01, "learning_rate": 2.9974034404414154e-05, "loss": 1.378, "step": 74240 }, { "epoch": 4.01, "learning_rate": 2.997132965487396e-05, "loss": 1.2596, "step": 74250 }, { "epoch": 4.01, "learning_rate": 2.996862490533377e-05, "loss": 1.3956, "step": 74260 }, { "epoch": 4.01, "learning_rate": 2.9965920155793576e-05, "loss": 1.3819, "step": 74270 }, { "epoch": 4.01, "learning_rate": 2.9963215406253382e-05, "loss": 1.3692, "step": 74280 }, { "epoch": 4.01, "learning_rate": 2.9960510656713192e-05, "loss": 1.3616, "step": 74290 }, { "epoch": 4.01, "learning_rate": 2.9957805907172998e-05, "loss": 1.3694, "step": 74300 }, { "epoch": 4.02, "learning_rate": 2.9955101157632804e-05, "loss": 1.2922, "step": 74310 }, { "epoch": 4.02, "learning_rate": 2.9952396408092614e-05, "loss": 1.2703, "step": 74320 }, { "epoch": 4.02, "learning_rate": 2.994969165855242e-05, "loss": 1.3646, "step": 74330 }, { "epoch": 4.02, "learning_rate": 2.9946986909012227e-05, "loss": 1.3552, "step": 74340 }, { "epoch": 4.02, "learning_rate": 2.9944282159472036e-05, "loss": 1.307, "step": 74350 }, { "epoch": 4.02, "learning_rate": 2.9941577409931842e-05, "loss": 1.3229, "step": 74360 }, { "epoch": 4.02, "learning_rate": 2.993887266039165e-05, "loss": 1.3326, "step": 74370 }, { "epoch": 4.02, "learning_rate": 2.993616791085146e-05, "loss": 1.374, "step": 74380 }, { "epoch": 4.02, "learning_rate": 2.9933463161311265e-05, "loss": 1.3908, "step": 74390 }, { "epoch": 4.02, "learning_rate": 2.993075841177107e-05, "loss": 1.3188, "step": 74400 }, { "epoch": 4.02, "learning_rate": 2.992805366223088e-05, "loss": 1.3738, "step": 74410 }, { "epoch": 4.02, "learning_rate": 2.9925348912690687e-05, "loss": 1.3797, "step": 74420 }, { "epoch": 4.02, "learning_rate": 2.9922644163150493e-05, "loss": 1.3617, "step": 74430 }, { "epoch": 4.02, "learning_rate": 2.99199394136103e-05, "loss": 1.4038, "step": 74440 }, { "epoch": 4.02, "learning_rate": 2.991723466407011e-05, "loss": 1.4162, "step": 74450 }, { "epoch": 4.02, "learning_rate": 2.9914529914529915e-05, "loss": 1.4277, "step": 74460 }, { "epoch": 4.02, "learning_rate": 2.991182516498972e-05, "loss": 1.3881, "step": 74470 }, { "epoch": 4.02, "learning_rate": 2.990912041544953e-05, "loss": 1.3091, "step": 74480 }, { "epoch": 4.03, "learning_rate": 2.9906415665909337e-05, "loss": 1.2935, "step": 74490 }, { "epoch": 4.03, "learning_rate": 2.9903710916369144e-05, "loss": 1.3063, "step": 74500 }, { "epoch": 4.03, "learning_rate": 2.9901006166828953e-05, "loss": 1.4068, "step": 74510 }, { "epoch": 4.03, "learning_rate": 2.989830141728876e-05, "loss": 1.3732, "step": 74520 }, { "epoch": 4.03, "learning_rate": 2.9895596667748566e-05, "loss": 1.3072, "step": 74530 }, { "epoch": 4.03, "learning_rate": 2.9892891918208375e-05, "loss": 1.3711, "step": 74540 }, { "epoch": 4.03, "learning_rate": 2.989018716866818e-05, "loss": 1.2912, "step": 74550 }, { "epoch": 4.03, "learning_rate": 2.9887482419127988e-05, "loss": 1.377, "step": 74560 }, { "epoch": 4.03, "learning_rate": 2.9884777669587798e-05, "loss": 1.3186, "step": 74570 }, { "epoch": 4.03, "learning_rate": 2.9882072920047604e-05, "loss": 1.3916, "step": 74580 }, { "epoch": 4.03, "learning_rate": 2.987936817050741e-05, "loss": 1.3279, "step": 74590 }, { "epoch": 4.03, "learning_rate": 2.987666342096722e-05, "loss": 1.3063, "step": 74600 }, { "epoch": 4.03, "learning_rate": 2.9873958671427026e-05, "loss": 1.3542, "step": 74610 }, { "epoch": 4.03, "learning_rate": 2.9871253921886832e-05, "loss": 1.3252, "step": 74620 }, { "epoch": 4.03, "learning_rate": 2.9868549172346645e-05, "loss": 1.3397, "step": 74630 }, { "epoch": 4.03, "learning_rate": 2.9865844422806448e-05, "loss": 1.4095, "step": 74640 }, { "epoch": 4.03, "learning_rate": 2.9863139673266254e-05, "loss": 1.3971, "step": 74650 }, { "epoch": 4.03, "learning_rate": 2.9860434923726067e-05, "loss": 1.3631, "step": 74660 }, { "epoch": 4.03, "learning_rate": 2.985773017418587e-05, "loss": 1.3282, "step": 74670 }, { "epoch": 4.04, "learning_rate": 2.9855025424645677e-05, "loss": 1.3507, "step": 74680 }, { "epoch": 4.04, "learning_rate": 2.985232067510549e-05, "loss": 1.3863, "step": 74690 }, { "epoch": 4.04, "learning_rate": 2.9849615925565293e-05, "loss": 1.3383, "step": 74700 }, { "epoch": 4.04, "learning_rate": 2.98469111760251e-05, "loss": 1.4188, "step": 74710 }, { "epoch": 4.04, "learning_rate": 2.9844206426484912e-05, "loss": 1.3802, "step": 74720 }, { "epoch": 4.04, "learning_rate": 2.9841501676944718e-05, "loss": 1.4234, "step": 74730 }, { "epoch": 4.04, "learning_rate": 2.983879692740452e-05, "loss": 1.4021, "step": 74740 }, { "epoch": 4.04, "learning_rate": 2.9836092177864334e-05, "loss": 1.4185, "step": 74750 }, { "epoch": 4.04, "learning_rate": 2.983338742832414e-05, "loss": 1.4164, "step": 74760 }, { "epoch": 4.04, "learning_rate": 2.9830682678783943e-05, "loss": 1.308, "step": 74770 }, { "epoch": 4.04, "learning_rate": 2.9827977929243756e-05, "loss": 1.3529, "step": 74780 }, { "epoch": 4.04, "learning_rate": 2.9825273179703562e-05, "loss": 1.4176, "step": 74790 }, { "epoch": 4.04, "learning_rate": 2.9822568430163365e-05, "loss": 1.3499, "step": 74800 }, { "epoch": 4.04, "learning_rate": 2.981986368062318e-05, "loss": 1.3279, "step": 74810 }, { "epoch": 4.04, "learning_rate": 2.9817158931082985e-05, "loss": 1.3568, "step": 74820 }, { "epoch": 4.04, "learning_rate": 2.9814454181542787e-05, "loss": 1.2808, "step": 74830 }, { "epoch": 4.04, "learning_rate": 2.98117494320026e-05, "loss": 1.318, "step": 74840 }, { "epoch": 4.04, "learning_rate": 2.9809044682462407e-05, "loss": 1.352, "step": 74850 }, { "epoch": 4.05, "learning_rate": 2.9806339932922213e-05, "loss": 1.278, "step": 74860 }, { "epoch": 4.05, "learning_rate": 2.9803635183382023e-05, "loss": 1.4278, "step": 74870 }, { "epoch": 4.05, "learning_rate": 2.980093043384183e-05, "loss": 1.4082, "step": 74880 }, { "epoch": 4.05, "learning_rate": 2.9798225684301635e-05, "loss": 1.4028, "step": 74890 }, { "epoch": 4.05, "learning_rate": 2.9795520934761445e-05, "loss": 1.4191, "step": 74900 }, { "epoch": 4.05, "learning_rate": 2.979281618522125e-05, "loss": 1.3738, "step": 74910 }, { "epoch": 4.05, "learning_rate": 2.9790111435681057e-05, "loss": 1.3982, "step": 74920 }, { "epoch": 4.05, "learning_rate": 2.9787406686140867e-05, "loss": 1.3914, "step": 74930 }, { "epoch": 4.05, "learning_rate": 2.9784701936600673e-05, "loss": 1.3597, "step": 74940 }, { "epoch": 4.05, "learning_rate": 2.978199718706048e-05, "loss": 1.3632, "step": 74950 }, { "epoch": 4.05, "learning_rate": 2.977929243752029e-05, "loss": 1.3718, "step": 74960 }, { "epoch": 4.05, "learning_rate": 2.9776587687980095e-05, "loss": 1.2619, "step": 74970 }, { "epoch": 4.05, "learning_rate": 2.9773882938439902e-05, "loss": 1.402, "step": 74980 }, { "epoch": 4.05, "learning_rate": 2.9771178188899708e-05, "loss": 1.2979, "step": 74990 }, { "epoch": 4.05, "learning_rate": 2.9768473439359518e-05, "loss": 1.2832, "step": 75000 }, { "epoch": 4.05, "learning_rate": 2.9765768689819324e-05, "loss": 1.3993, "step": 75010 }, { "epoch": 4.05, "learning_rate": 2.976306394027913e-05, "loss": 1.4021, "step": 75020 }, { "epoch": 4.05, "learning_rate": 2.976035919073894e-05, "loss": 1.3132, "step": 75030 }, { "epoch": 4.05, "learning_rate": 2.9757654441198746e-05, "loss": 1.3779, "step": 75040 }, { "epoch": 4.06, "learning_rate": 2.9754949691658552e-05, "loss": 1.3272, "step": 75050 }, { "epoch": 4.06, "learning_rate": 2.9752244942118362e-05, "loss": 1.3591, "step": 75060 }, { "epoch": 4.06, "learning_rate": 2.9749540192578168e-05, "loss": 1.324, "step": 75070 }, { "epoch": 4.06, "learning_rate": 2.9746835443037974e-05, "loss": 1.399, "step": 75080 }, { "epoch": 4.06, "learning_rate": 2.9744130693497784e-05, "loss": 1.3402, "step": 75090 }, { "epoch": 4.06, "learning_rate": 2.974142594395759e-05, "loss": 1.2851, "step": 75100 }, { "epoch": 4.06, "learning_rate": 2.9738721194417397e-05, "loss": 1.3818, "step": 75110 }, { "epoch": 4.06, "learning_rate": 2.9736016444877206e-05, "loss": 1.3448, "step": 75120 }, { "epoch": 4.06, "learning_rate": 2.9733311695337013e-05, "loss": 1.4305, "step": 75130 }, { "epoch": 4.06, "learning_rate": 2.973060694579682e-05, "loss": 1.3578, "step": 75140 }, { "epoch": 4.06, "learning_rate": 2.972790219625663e-05, "loss": 1.3995, "step": 75150 }, { "epoch": 4.06, "learning_rate": 2.9725197446716435e-05, "loss": 1.3829, "step": 75160 }, { "epoch": 4.06, "learning_rate": 2.972249269717624e-05, "loss": 1.3586, "step": 75170 }, { "epoch": 4.06, "learning_rate": 2.971978794763605e-05, "loss": 1.3959, "step": 75180 }, { "epoch": 4.06, "learning_rate": 2.9717083198095857e-05, "loss": 1.3318, "step": 75190 }, { "epoch": 4.06, "learning_rate": 2.9714378448555663e-05, "loss": 1.3652, "step": 75200 }, { "epoch": 4.06, "learning_rate": 2.9711673699015473e-05, "loss": 1.3487, "step": 75210 }, { "epoch": 4.06, "learning_rate": 2.970896894947528e-05, "loss": 1.3764, "step": 75220 }, { "epoch": 4.07, "learning_rate": 2.9706264199935085e-05, "loss": 1.2892, "step": 75230 }, { "epoch": 4.07, "learning_rate": 2.9703559450394895e-05, "loss": 1.3635, "step": 75240 }, { "epoch": 4.07, "learning_rate": 2.97008547008547e-05, "loss": 1.3844, "step": 75250 }, { "epoch": 4.07, "learning_rate": 2.9698149951314507e-05, "loss": 1.4019, "step": 75260 }, { "epoch": 4.07, "learning_rate": 2.969544520177432e-05, "loss": 1.3205, "step": 75270 }, { "epoch": 4.07, "learning_rate": 2.9692740452234123e-05, "loss": 1.4028, "step": 75280 }, { "epoch": 4.07, "learning_rate": 2.969003570269393e-05, "loss": 1.3297, "step": 75290 }, { "epoch": 4.07, "learning_rate": 2.9687330953153743e-05, "loss": 1.3446, "step": 75300 }, { "epoch": 4.07, "learning_rate": 2.9684626203613546e-05, "loss": 1.2742, "step": 75310 }, { "epoch": 4.07, "learning_rate": 2.9681921454073352e-05, "loss": 1.3084, "step": 75320 }, { "epoch": 4.07, "learning_rate": 2.9679216704533165e-05, "loss": 1.3948, "step": 75330 }, { "epoch": 4.07, "learning_rate": 2.9676511954992968e-05, "loss": 1.3396, "step": 75340 }, { "epoch": 4.07, "learning_rate": 2.9673807205452774e-05, "loss": 1.4303, "step": 75350 }, { "epoch": 4.07, "learning_rate": 2.9671102455912587e-05, "loss": 1.4274, "step": 75360 }, { "epoch": 4.07, "learning_rate": 2.9668397706372393e-05, "loss": 1.3287, "step": 75370 }, { "epoch": 4.07, "learning_rate": 2.9665692956832196e-05, "loss": 1.3873, "step": 75380 }, { "epoch": 4.07, "learning_rate": 2.966298820729201e-05, "loss": 1.3311, "step": 75390 }, { "epoch": 4.07, "learning_rate": 2.9660283457751815e-05, "loss": 1.407, "step": 75400 }, { "epoch": 4.07, "learning_rate": 2.965757870821162e-05, "loss": 1.3862, "step": 75410 }, { "epoch": 4.08, "learning_rate": 2.965487395867143e-05, "loss": 1.346, "step": 75420 }, { "epoch": 4.08, "learning_rate": 2.9652169209131238e-05, "loss": 1.3828, "step": 75430 }, { "epoch": 4.08, "learning_rate": 2.964946445959104e-05, "loss": 1.4372, "step": 75440 }, { "epoch": 4.08, "learning_rate": 2.9646759710050854e-05, "loss": 1.3572, "step": 75450 }, { "epoch": 4.08, "learning_rate": 2.964405496051066e-05, "loss": 1.3831, "step": 75460 }, { "epoch": 4.08, "learning_rate": 2.9641350210970466e-05, "loss": 1.3909, "step": 75470 }, { "epoch": 4.08, "learning_rate": 2.9638645461430276e-05, "loss": 1.3554, "step": 75480 }, { "epoch": 4.08, "learning_rate": 2.9635940711890082e-05, "loss": 1.3318, "step": 75490 }, { "epoch": 4.08, "learning_rate": 2.9633235962349888e-05, "loss": 1.4036, "step": 75500 }, { "epoch": 4.08, "learning_rate": 2.9630531212809698e-05, "loss": 1.3334, "step": 75510 }, { "epoch": 4.08, "learning_rate": 2.9627826463269504e-05, "loss": 1.3966, "step": 75520 }, { "epoch": 4.08, "learning_rate": 2.962512171372931e-05, "loss": 1.365, "step": 75530 }, { "epoch": 4.08, "learning_rate": 2.962241696418912e-05, "loss": 1.386, "step": 75540 }, { "epoch": 4.08, "learning_rate": 2.9619712214648926e-05, "loss": 1.3415, "step": 75550 }, { "epoch": 4.08, "learning_rate": 2.9617007465108733e-05, "loss": 1.3266, "step": 75560 }, { "epoch": 4.08, "learning_rate": 2.9614302715568535e-05, "loss": 1.3458, "step": 75570 }, { "epoch": 4.08, "learning_rate": 2.961159796602835e-05, "loss": 1.3718, "step": 75580 }, { "epoch": 4.08, "learning_rate": 2.9608893216488155e-05, "loss": 1.3493, "step": 75590 }, { "epoch": 4.08, "learning_rate": 2.960618846694796e-05, "loss": 1.4556, "step": 75600 }, { "epoch": 4.09, "learning_rate": 2.960348371740777e-05, "loss": 1.3458, "step": 75610 }, { "epoch": 4.09, "learning_rate": 2.9600778967867577e-05, "loss": 1.3737, "step": 75620 }, { "epoch": 4.09, "learning_rate": 2.9598074218327383e-05, "loss": 1.3004, "step": 75630 }, { "epoch": 4.09, "learning_rate": 2.9595369468787193e-05, "loss": 1.2995, "step": 75640 }, { "epoch": 4.09, "learning_rate": 2.9592664719247e-05, "loss": 1.4052, "step": 75650 }, { "epoch": 4.09, "learning_rate": 2.9589959969706805e-05, "loss": 1.4063, "step": 75660 }, { "epoch": 4.09, "learning_rate": 2.9587255220166615e-05, "loss": 1.4006, "step": 75670 }, { "epoch": 4.09, "learning_rate": 2.958455047062642e-05, "loss": 1.3826, "step": 75680 }, { "epoch": 4.09, "learning_rate": 2.9581845721086228e-05, "loss": 1.3317, "step": 75690 }, { "epoch": 4.09, "learning_rate": 2.9579140971546037e-05, "loss": 1.3476, "step": 75700 }, { "epoch": 4.09, "learning_rate": 2.9576436222005843e-05, "loss": 1.334, "step": 75710 }, { "epoch": 4.09, "learning_rate": 2.957373147246565e-05, "loss": 1.327, "step": 75720 }, { "epoch": 4.09, "learning_rate": 2.957102672292546e-05, "loss": 1.3626, "step": 75730 }, { "epoch": 4.09, "learning_rate": 2.9568321973385266e-05, "loss": 1.4114, "step": 75740 }, { "epoch": 4.09, "learning_rate": 2.9565617223845072e-05, "loss": 1.4075, "step": 75750 }, { "epoch": 4.09, "learning_rate": 2.956291247430488e-05, "loss": 1.3547, "step": 75760 }, { "epoch": 4.09, "learning_rate": 2.9560207724764688e-05, "loss": 1.3893, "step": 75770 }, { "epoch": 4.09, "learning_rate": 2.9557502975224494e-05, "loss": 1.3193, "step": 75780 }, { "epoch": 4.1, "learning_rate": 2.9554798225684304e-05, "loss": 1.4426, "step": 75790 }, { "epoch": 4.1, "learning_rate": 2.955209347614411e-05, "loss": 1.3033, "step": 75800 }, { "epoch": 4.1, "learning_rate": 2.9549388726603916e-05, "loss": 1.3915, "step": 75810 }, { "epoch": 4.1, "learning_rate": 2.9546683977063726e-05, "loss": 1.3352, "step": 75820 }, { "epoch": 4.1, "learning_rate": 2.9543979227523532e-05, "loss": 1.4713, "step": 75830 }, { "epoch": 4.1, "learning_rate": 2.954127447798334e-05, "loss": 1.3184, "step": 75840 }, { "epoch": 4.1, "learning_rate": 2.9538569728443148e-05, "loss": 1.3127, "step": 75850 }, { "epoch": 4.1, "learning_rate": 2.9535864978902954e-05, "loss": 1.4285, "step": 75860 }, { "epoch": 4.1, "learning_rate": 2.953316022936276e-05, "loss": 1.3941, "step": 75870 }, { "epoch": 4.1, "learning_rate": 2.9530455479822574e-05, "loss": 1.3845, "step": 75880 }, { "epoch": 4.1, "learning_rate": 2.9527750730282376e-05, "loss": 1.4561, "step": 75890 }, { "epoch": 4.1, "learning_rate": 2.9525045980742183e-05, "loss": 1.334, "step": 75900 }, { "epoch": 4.1, "learning_rate": 2.9522341231201996e-05, "loss": 1.3461, "step": 75910 }, { "epoch": 4.1, "learning_rate": 2.95196364816618e-05, "loss": 1.3362, "step": 75920 }, { "epoch": 4.1, "learning_rate": 2.9516931732121605e-05, "loss": 1.3548, "step": 75930 }, { "epoch": 4.1, "learning_rate": 2.9514226982581418e-05, "loss": 1.3198, "step": 75940 }, { "epoch": 4.1, "learning_rate": 2.951152223304122e-05, "loss": 1.3625, "step": 75950 }, { "epoch": 4.1, "learning_rate": 2.9508817483501027e-05, "loss": 1.3605, "step": 75960 }, { "epoch": 4.1, "learning_rate": 2.950611273396084e-05, "loss": 1.4066, "step": 75970 }, { "epoch": 4.11, "learning_rate": 2.9503407984420643e-05, "loss": 1.3233, "step": 75980 }, { "epoch": 4.11, "learning_rate": 2.950070323488045e-05, "loss": 1.3741, "step": 75990 }, { "epoch": 4.11, "learning_rate": 2.9497998485340262e-05, "loss": 1.3873, "step": 76000 }, { "epoch": 4.11, "learning_rate": 2.949529373580007e-05, "loss": 1.4067, "step": 76010 }, { "epoch": 4.11, "learning_rate": 2.949258898625987e-05, "loss": 1.3576, "step": 76020 }, { "epoch": 4.11, "learning_rate": 2.9489884236719684e-05, "loss": 1.4019, "step": 76030 }, { "epoch": 4.11, "learning_rate": 2.948717948717949e-05, "loss": 1.297, "step": 76040 }, { "epoch": 4.11, "learning_rate": 2.9484474737639294e-05, "loss": 1.4262, "step": 76050 }, { "epoch": 4.11, "learning_rate": 2.9481769988099107e-05, "loss": 1.3318, "step": 76060 }, { "epoch": 4.11, "learning_rate": 2.9479065238558913e-05, "loss": 1.3895, "step": 76070 }, { "epoch": 4.11, "learning_rate": 2.9476360489018716e-05, "loss": 1.3475, "step": 76080 }, { "epoch": 4.11, "learning_rate": 2.947365573947853e-05, "loss": 1.336, "step": 76090 }, { "epoch": 4.11, "learning_rate": 2.9470950989938335e-05, "loss": 1.4056, "step": 76100 }, { "epoch": 4.11, "learning_rate": 2.946824624039814e-05, "loss": 1.265, "step": 76110 }, { "epoch": 4.11, "learning_rate": 2.9465541490857944e-05, "loss": 1.4618, "step": 76120 }, { "epoch": 4.11, "learning_rate": 2.9462836741317757e-05, "loss": 1.395, "step": 76130 }, { "epoch": 4.11, "learning_rate": 2.9460131991777563e-05, "loss": 1.4386, "step": 76140 }, { "epoch": 4.11, "learning_rate": 2.9457427242237366e-05, "loss": 1.4113, "step": 76150 }, { "epoch": 4.12, "learning_rate": 2.945472249269718e-05, "loss": 1.4173, "step": 76160 }, { "epoch": 4.12, "learning_rate": 2.9452017743156986e-05, "loss": 1.3555, "step": 76170 }, { "epoch": 4.12, "learning_rate": 2.944931299361679e-05, "loss": 1.2979, "step": 76180 }, { "epoch": 4.12, "learning_rate": 2.94466082440766e-05, "loss": 1.2964, "step": 76190 }, { "epoch": 4.12, "learning_rate": 2.9443903494536408e-05, "loss": 1.3278, "step": 76200 }, { "epoch": 4.12, "learning_rate": 2.944119874499621e-05, "loss": 1.367, "step": 76210 }, { "epoch": 4.12, "learning_rate": 2.9438493995456024e-05, "loss": 1.3213, "step": 76220 }, { "epoch": 4.12, "learning_rate": 2.943578924591583e-05, "loss": 1.4339, "step": 76230 }, { "epoch": 4.12, "learning_rate": 2.9433084496375636e-05, "loss": 1.3893, "step": 76240 }, { "epoch": 4.12, "learning_rate": 2.9430379746835446e-05, "loss": 1.3662, "step": 76250 }, { "epoch": 4.12, "learning_rate": 2.9427674997295252e-05, "loss": 1.363, "step": 76260 }, { "epoch": 4.12, "learning_rate": 2.942497024775506e-05, "loss": 1.3679, "step": 76270 }, { "epoch": 4.12, "learning_rate": 2.9422265498214868e-05, "loss": 1.3828, "step": 76280 }, { "epoch": 4.12, "learning_rate": 2.9419560748674674e-05, "loss": 1.3664, "step": 76290 }, { "epoch": 4.12, "learning_rate": 2.941685599913448e-05, "loss": 1.4165, "step": 76300 }, { "epoch": 4.12, "learning_rate": 2.941415124959429e-05, "loss": 1.441, "step": 76310 }, { "epoch": 4.12, "learning_rate": 2.9411446500054096e-05, "loss": 1.4097, "step": 76320 }, { "epoch": 4.12, "learning_rate": 2.9408741750513903e-05, "loss": 1.3199, "step": 76330 }, { "epoch": 4.12, "learning_rate": 2.9406037000973712e-05, "loss": 1.4036, "step": 76340 }, { "epoch": 4.13, "learning_rate": 2.940333225143352e-05, "loss": 1.3707, "step": 76350 }, { "epoch": 4.13, "learning_rate": 2.9400627501893325e-05, "loss": 1.3298, "step": 76360 }, { "epoch": 4.13, "learning_rate": 2.9397922752353135e-05, "loss": 1.4103, "step": 76370 }, { "epoch": 4.13, "learning_rate": 2.939521800281294e-05, "loss": 1.3675, "step": 76380 }, { "epoch": 4.13, "learning_rate": 2.9392513253272747e-05, "loss": 1.4379, "step": 76390 }, { "epoch": 4.13, "learning_rate": 2.9389808503732557e-05, "loss": 1.3561, "step": 76400 }, { "epoch": 4.13, "learning_rate": 2.9387103754192363e-05, "loss": 1.309, "step": 76410 }, { "epoch": 4.13, "learning_rate": 2.938439900465217e-05, "loss": 1.3674, "step": 76420 }, { "epoch": 4.13, "learning_rate": 2.938169425511198e-05, "loss": 1.4137, "step": 76430 }, { "epoch": 4.13, "learning_rate": 2.9378989505571785e-05, "loss": 1.3968, "step": 76440 }, { "epoch": 4.13, "learning_rate": 2.937628475603159e-05, "loss": 1.3275, "step": 76450 }, { "epoch": 4.13, "learning_rate": 2.93735800064914e-05, "loss": 1.3433, "step": 76460 }, { "epoch": 4.13, "learning_rate": 2.9370875256951207e-05, "loss": 1.3446, "step": 76470 }, { "epoch": 4.13, "learning_rate": 2.9368170507411014e-05, "loss": 1.3942, "step": 76480 }, { "epoch": 4.13, "learning_rate": 2.9365465757870823e-05, "loss": 1.4039, "step": 76490 }, { "epoch": 4.13, "learning_rate": 2.936276100833063e-05, "loss": 1.4111, "step": 76500 }, { "epoch": 4.13, "learning_rate": 2.9360056258790436e-05, "loss": 1.3671, "step": 76510 }, { "epoch": 4.13, "learning_rate": 2.935735150925025e-05, "loss": 1.3567, "step": 76520 }, { "epoch": 4.14, "learning_rate": 2.935464675971005e-05, "loss": 1.3403, "step": 76530 }, { "epoch": 4.14, "learning_rate": 2.9351942010169858e-05, "loss": 1.3734, "step": 76540 }, { "epoch": 4.14, "learning_rate": 2.934923726062967e-05, "loss": 1.3374, "step": 76550 }, { "epoch": 4.14, "learning_rate": 2.9346532511089474e-05, "loss": 1.4447, "step": 76560 }, { "epoch": 4.14, "learning_rate": 2.934382776154928e-05, "loss": 1.3487, "step": 76570 }, { "epoch": 4.14, "learning_rate": 2.9341123012009093e-05, "loss": 1.4015, "step": 76580 }, { "epoch": 4.14, "learning_rate": 2.9338418262468896e-05, "loss": 1.3802, "step": 76590 }, { "epoch": 4.14, "learning_rate": 2.9335713512928702e-05, "loss": 1.3047, "step": 76600 }, { "epoch": 4.14, "learning_rate": 2.9333008763388515e-05, "loss": 1.2549, "step": 76610 }, { "epoch": 4.14, "learning_rate": 2.9330304013848318e-05, "loss": 1.3924, "step": 76620 }, { "epoch": 4.14, "learning_rate": 2.9327599264308124e-05, "loss": 1.3986, "step": 76630 }, { "epoch": 4.14, "learning_rate": 2.9324894514767937e-05, "loss": 1.2959, "step": 76640 }, { "epoch": 4.14, "learning_rate": 2.9322189765227744e-05, "loss": 1.3541, "step": 76650 }, { "epoch": 4.14, "learning_rate": 2.9319485015687547e-05, "loss": 1.4183, "step": 76660 }, { "epoch": 4.14, "learning_rate": 2.9316780266147353e-05, "loss": 1.4627, "step": 76670 }, { "epoch": 4.14, "learning_rate": 2.9314075516607166e-05, "loss": 1.4753, "step": 76680 }, { "epoch": 4.14, "learning_rate": 2.931137076706697e-05, "loss": 1.3999, "step": 76690 }, { "epoch": 4.14, "learning_rate": 2.9308666017526775e-05, "loss": 1.371, "step": 76700 }, { "epoch": 4.14, "learning_rate": 2.9305961267986588e-05, "loss": 1.3587, "step": 76710 }, { "epoch": 4.15, "learning_rate": 2.930325651844639e-05, "loss": 1.3922, "step": 76720 }, { "epoch": 4.15, "learning_rate": 2.9300551768906197e-05, "loss": 1.3834, "step": 76730 }, { "epoch": 4.15, "learning_rate": 2.929784701936601e-05, "loss": 1.3875, "step": 76740 }, { "epoch": 4.15, "learning_rate": 2.9295142269825816e-05, "loss": 1.4469, "step": 76750 }, { "epoch": 4.15, "learning_rate": 2.929243752028562e-05, "loss": 1.3741, "step": 76760 }, { "epoch": 4.15, "learning_rate": 2.9289732770745432e-05, "loss": 1.3113, "step": 76770 }, { "epoch": 4.15, "learning_rate": 2.928702802120524e-05, "loss": 1.4217, "step": 76780 }, { "epoch": 4.15, "learning_rate": 2.928432327166504e-05, "loss": 1.4312, "step": 76790 }, { "epoch": 4.15, "learning_rate": 2.9281618522124855e-05, "loss": 1.3411, "step": 76800 }, { "epoch": 4.15, "learning_rate": 2.927891377258466e-05, "loss": 1.4574, "step": 76810 }, { "epoch": 4.15, "learning_rate": 2.9276209023044464e-05, "loss": 1.3938, "step": 76820 }, { "epoch": 4.15, "learning_rate": 2.9273504273504277e-05, "loss": 1.3655, "step": 76830 }, { "epoch": 4.15, "learning_rate": 2.9270799523964083e-05, "loss": 1.3328, "step": 76840 }, { "epoch": 4.15, "learning_rate": 2.926809477442389e-05, "loss": 1.3754, "step": 76850 }, { "epoch": 4.15, "learning_rate": 2.92653900248837e-05, "loss": 1.4114, "step": 76860 }, { "epoch": 4.15, "learning_rate": 2.9262685275343505e-05, "loss": 1.3183, "step": 76870 }, { "epoch": 4.15, "learning_rate": 2.925998052580331e-05, "loss": 1.3489, "step": 76880 }, { "epoch": 4.15, "learning_rate": 2.925727577626312e-05, "loss": 1.3778, "step": 76890 }, { "epoch": 4.16, "learning_rate": 2.9254571026722927e-05, "loss": 1.3313, "step": 76900 }, { "epoch": 4.16, "learning_rate": 2.9251866277182734e-05, "loss": 1.3256, "step": 76910 }, { "epoch": 4.16, "learning_rate": 2.9249161527642543e-05, "loss": 1.3486, "step": 76920 }, { "epoch": 4.16, "learning_rate": 2.924645677810235e-05, "loss": 1.3754, "step": 76930 }, { "epoch": 4.16, "learning_rate": 2.9243752028562156e-05, "loss": 1.3558, "step": 76940 }, { "epoch": 4.16, "learning_rate": 2.9241047279021965e-05, "loss": 1.4108, "step": 76950 }, { "epoch": 4.16, "learning_rate": 2.923834252948177e-05, "loss": 1.498, "step": 76960 }, { "epoch": 4.16, "learning_rate": 2.9235637779941578e-05, "loss": 1.3894, "step": 76970 }, { "epoch": 4.16, "learning_rate": 2.9232933030401388e-05, "loss": 1.3283, "step": 76980 }, { "epoch": 4.16, "learning_rate": 2.9230228280861194e-05, "loss": 1.3869, "step": 76990 }, { "epoch": 4.16, "learning_rate": 2.9227523531321e-05, "loss": 1.3335, "step": 77000 }, { "epoch": 4.16, "learning_rate": 2.922481878178081e-05, "loss": 1.3668, "step": 77010 }, { "epoch": 4.16, "learning_rate": 2.9222114032240616e-05, "loss": 1.3952, "step": 77020 }, { "epoch": 4.16, "learning_rate": 2.9219409282700422e-05, "loss": 1.3488, "step": 77030 }, { "epoch": 4.16, "learning_rate": 2.9216704533160232e-05, "loss": 1.3604, "step": 77040 }, { "epoch": 4.16, "learning_rate": 2.9213999783620038e-05, "loss": 1.3548, "step": 77050 }, { "epoch": 4.16, "learning_rate": 2.9211295034079844e-05, "loss": 1.3927, "step": 77060 }, { "epoch": 4.16, "learning_rate": 2.9208590284539654e-05, "loss": 1.3916, "step": 77070 }, { "epoch": 4.16, "learning_rate": 2.920588553499946e-05, "loss": 1.32, "step": 77080 }, { "epoch": 4.17, "learning_rate": 2.9203180785459267e-05, "loss": 1.3347, "step": 77090 }, { "epoch": 4.17, "learning_rate": 2.9200476035919076e-05, "loss": 1.3076, "step": 77100 }, { "epoch": 4.17, "learning_rate": 2.9197771286378883e-05, "loss": 1.3428, "step": 77110 }, { "epoch": 4.17, "learning_rate": 2.919506653683869e-05, "loss": 1.433, "step": 77120 }, { "epoch": 4.17, "learning_rate": 2.91923617872985e-05, "loss": 1.3532, "step": 77130 }, { "epoch": 4.17, "learning_rate": 2.9189657037758305e-05, "loss": 1.3967, "step": 77140 }, { "epoch": 4.17, "learning_rate": 2.918695228821811e-05, "loss": 1.4237, "step": 77150 }, { "epoch": 4.17, "learning_rate": 2.9184247538677924e-05, "loss": 1.3471, "step": 77160 }, { "epoch": 4.17, "learning_rate": 2.9181542789137727e-05, "loss": 1.3804, "step": 77170 }, { "epoch": 4.17, "learning_rate": 2.9178838039597533e-05, "loss": 1.439, "step": 77180 }, { "epoch": 4.17, "learning_rate": 2.9176133290057346e-05, "loss": 1.3081, "step": 77190 }, { "epoch": 4.17, "learning_rate": 2.917342854051715e-05, "loss": 1.4268, "step": 77200 }, { "epoch": 4.17, "learning_rate": 2.9170723790976955e-05, "loss": 1.4172, "step": 77210 }, { "epoch": 4.17, "learning_rate": 2.916801904143677e-05, "loss": 1.3995, "step": 77220 }, { "epoch": 4.17, "learning_rate": 2.916531429189657e-05, "loss": 1.3109, "step": 77230 }, { "epoch": 4.17, "learning_rate": 2.9162609542356377e-05, "loss": 1.354, "step": 77240 }, { "epoch": 4.17, "learning_rate": 2.9159904792816184e-05, "loss": 1.3302, "step": 77250 }, { "epoch": 4.17, "learning_rate": 2.9157200043275997e-05, "loss": 1.3902, "step": 77260 }, { "epoch": 4.18, "learning_rate": 2.91544952937358e-05, "loss": 1.4349, "step": 77270 }, { "epoch": 4.18, "learning_rate": 2.9151790544195606e-05, "loss": 1.3895, "step": 77280 }, { "epoch": 4.18, "learning_rate": 2.914908579465542e-05, "loss": 1.3662, "step": 77290 }, { "epoch": 4.18, "learning_rate": 2.9146381045115222e-05, "loss": 1.3657, "step": 77300 }, { "epoch": 4.18, "learning_rate": 2.9143676295575028e-05, "loss": 1.3968, "step": 77310 }, { "epoch": 4.18, "learning_rate": 2.914097154603484e-05, "loss": 1.3167, "step": 77320 }, { "epoch": 4.18, "learning_rate": 2.9138266796494644e-05, "loss": 1.4075, "step": 77330 }, { "epoch": 4.18, "learning_rate": 2.913556204695445e-05, "loss": 1.3443, "step": 77340 }, { "epoch": 4.18, "learning_rate": 2.9132857297414263e-05, "loss": 1.3417, "step": 77350 }, { "epoch": 4.18, "learning_rate": 2.9130152547874066e-05, "loss": 1.4252, "step": 77360 }, { "epoch": 4.18, "learning_rate": 2.9127447798333872e-05, "loss": 1.4106, "step": 77370 }, { "epoch": 4.18, "learning_rate": 2.9124743048793685e-05, "loss": 1.4012, "step": 77380 }, { "epoch": 4.18, "learning_rate": 2.912203829925349e-05, "loss": 1.3742, "step": 77390 }, { "epoch": 4.18, "learning_rate": 2.9119333549713295e-05, "loss": 1.4346, "step": 77400 }, { "epoch": 4.18, "learning_rate": 2.9116628800173108e-05, "loss": 1.3653, "step": 77410 }, { "epoch": 4.18, "learning_rate": 2.9113924050632914e-05, "loss": 1.3625, "step": 77420 }, { "epoch": 4.18, "learning_rate": 2.9111219301092717e-05, "loss": 1.326, "step": 77430 }, { "epoch": 4.18, "learning_rate": 2.910851455155253e-05, "loss": 1.3359, "step": 77440 }, { "epoch": 4.18, "learning_rate": 2.9105809802012336e-05, "loss": 1.3659, "step": 77450 }, { "epoch": 4.19, "learning_rate": 2.910310505247214e-05, "loss": 1.3921, "step": 77460 }, { "epoch": 4.19, "learning_rate": 2.9100400302931952e-05, "loss": 1.3477, "step": 77470 }, { "epoch": 4.19, "learning_rate": 2.9097695553391758e-05, "loss": 1.3238, "step": 77480 }, { "epoch": 4.19, "learning_rate": 2.9094990803851564e-05, "loss": 1.377, "step": 77490 }, { "epoch": 4.19, "learning_rate": 2.9092286054311374e-05, "loss": 1.3206, "step": 77500 }, { "epoch": 4.19, "learning_rate": 2.908958130477118e-05, "loss": 1.424, "step": 77510 }, { "epoch": 4.19, "learning_rate": 2.9086876555230987e-05, "loss": 1.4323, "step": 77520 }, { "epoch": 4.19, "learning_rate": 2.9084171805690796e-05, "loss": 1.4046, "step": 77530 }, { "epoch": 4.19, "learning_rate": 2.9081467056150603e-05, "loss": 1.3809, "step": 77540 }, { "epoch": 4.19, "learning_rate": 2.907876230661041e-05, "loss": 1.3213, "step": 77550 }, { "epoch": 4.19, "learning_rate": 2.907605755707022e-05, "loss": 1.3483, "step": 77560 }, { "epoch": 4.19, "learning_rate": 2.9073352807530025e-05, "loss": 1.3957, "step": 77570 }, { "epoch": 4.19, "learning_rate": 2.907064805798983e-05, "loss": 1.4203, "step": 77580 }, { "epoch": 4.19, "learning_rate": 2.906794330844964e-05, "loss": 1.3752, "step": 77590 }, { "epoch": 4.19, "learning_rate": 2.9065238558909447e-05, "loss": 1.3366, "step": 77600 }, { "epoch": 4.19, "learning_rate": 2.9062533809369253e-05, "loss": 1.373, "step": 77610 }, { "epoch": 4.19, "learning_rate": 2.9059829059829063e-05, "loss": 1.3293, "step": 77620 }, { "epoch": 4.19, "learning_rate": 2.905712431028887e-05, "loss": 1.3939, "step": 77630 }, { "epoch": 4.2, "learning_rate": 2.9054419560748675e-05, "loss": 1.4223, "step": 77640 }, { "epoch": 4.2, "learning_rate": 2.9051714811208485e-05, "loss": 1.3586, "step": 77650 }, { "epoch": 4.2, "learning_rate": 2.904901006166829e-05, "loss": 1.3167, "step": 77660 }, { "epoch": 4.2, "learning_rate": 2.9046305312128097e-05, "loss": 1.3556, "step": 77670 }, { "epoch": 4.2, "learning_rate": 2.9043600562587907e-05, "loss": 1.4201, "step": 77680 }, { "epoch": 4.2, "learning_rate": 2.9040895813047713e-05, "loss": 1.3468, "step": 77690 }, { "epoch": 4.2, "learning_rate": 2.903819106350752e-05, "loss": 1.3805, "step": 77700 }, { "epoch": 4.2, "learning_rate": 2.903548631396733e-05, "loss": 1.3822, "step": 77710 }, { "epoch": 4.2, "learning_rate": 2.9032781564427136e-05, "loss": 1.3923, "step": 77720 }, { "epoch": 4.2, "learning_rate": 2.9030076814886942e-05, "loss": 1.3826, "step": 77730 }, { "epoch": 4.2, "learning_rate": 2.902737206534675e-05, "loss": 1.366, "step": 77740 }, { "epoch": 4.2, "learning_rate": 2.9024667315806558e-05, "loss": 1.3435, "step": 77750 }, { "epoch": 4.2, "learning_rate": 2.9021962566266364e-05, "loss": 1.4541, "step": 77760 }, { "epoch": 4.2, "learning_rate": 2.9019257816726174e-05, "loss": 1.3316, "step": 77770 }, { "epoch": 4.2, "learning_rate": 2.901655306718598e-05, "loss": 1.3682, "step": 77780 }, { "epoch": 4.2, "learning_rate": 2.9013848317645786e-05, "loss": 1.3391, "step": 77790 }, { "epoch": 4.2, "learning_rate": 2.9011143568105592e-05, "loss": 1.3684, "step": 77800 }, { "epoch": 4.2, "learning_rate": 2.9008438818565402e-05, "loss": 1.3212, "step": 77810 }, { "epoch": 4.2, "learning_rate": 2.900573406902521e-05, "loss": 1.3448, "step": 77820 }, { "epoch": 4.21, "learning_rate": 2.9003029319485015e-05, "loss": 1.3277, "step": 77830 }, { "epoch": 4.21, "learning_rate": 2.9000324569944824e-05, "loss": 1.3233, "step": 77840 }, { "epoch": 4.21, "learning_rate": 2.899761982040463e-05, "loss": 1.3368, "step": 77850 }, { "epoch": 4.21, "learning_rate": 2.8994915070864437e-05, "loss": 1.2975, "step": 77860 }, { "epoch": 4.21, "learning_rate": 2.8992210321324246e-05, "loss": 1.3959, "step": 77870 }, { "epoch": 4.21, "learning_rate": 2.8989505571784053e-05, "loss": 1.3704, "step": 77880 }, { "epoch": 4.21, "learning_rate": 2.898680082224386e-05, "loss": 1.3337, "step": 77890 }, { "epoch": 4.21, "learning_rate": 2.8984096072703672e-05, "loss": 1.3908, "step": 77900 }, { "epoch": 4.21, "learning_rate": 2.8981391323163475e-05, "loss": 1.3765, "step": 77910 }, { "epoch": 4.21, "learning_rate": 2.897868657362328e-05, "loss": 1.3591, "step": 77920 }, { "epoch": 4.21, "learning_rate": 2.8975981824083094e-05, "loss": 1.4088, "step": 77930 }, { "epoch": 4.21, "learning_rate": 2.8973277074542897e-05, "loss": 1.4094, "step": 77940 }, { "epoch": 4.21, "learning_rate": 2.8970572325002703e-05, "loss": 1.445, "step": 77950 }, { "epoch": 4.21, "learning_rate": 2.8967867575462516e-05, "loss": 1.4433, "step": 77960 }, { "epoch": 4.21, "learning_rate": 2.896516282592232e-05, "loss": 1.4031, "step": 77970 }, { "epoch": 4.21, "learning_rate": 2.8962458076382125e-05, "loss": 1.383, "step": 77980 }, { "epoch": 4.21, "learning_rate": 2.895975332684194e-05, "loss": 1.4042, "step": 77990 }, { "epoch": 4.21, "learning_rate": 2.895704857730174e-05, "loss": 1.3593, "step": 78000 }, { "epoch": 4.22, "learning_rate": 2.8954343827761548e-05, "loss": 1.3915, "step": 78010 }, { "epoch": 4.22, "learning_rate": 2.895163907822136e-05, "loss": 1.4146, "step": 78020 }, { "epoch": 4.22, "learning_rate": 2.8948934328681167e-05, "loss": 1.4921, "step": 78030 }, { "epoch": 4.22, "learning_rate": 2.894622957914097e-05, "loss": 1.3305, "step": 78040 }, { "epoch": 4.22, "learning_rate": 2.8943524829600783e-05, "loss": 1.4469, "step": 78050 }, { "epoch": 4.22, "learning_rate": 2.894082008006059e-05, "loss": 1.4136, "step": 78060 }, { "epoch": 4.22, "learning_rate": 2.8938115330520392e-05, "loss": 1.3628, "step": 78070 }, { "epoch": 4.22, "learning_rate": 2.8935410580980205e-05, "loss": 1.3877, "step": 78080 }, { "epoch": 4.22, "learning_rate": 2.893270583144001e-05, "loss": 1.3278, "step": 78090 }, { "epoch": 4.22, "learning_rate": 2.8930001081899814e-05, "loss": 1.3974, "step": 78100 }, { "epoch": 4.22, "learning_rate": 2.8927296332359627e-05, "loss": 1.3707, "step": 78110 }, { "epoch": 4.22, "learning_rate": 2.8924591582819433e-05, "loss": 1.4041, "step": 78120 }, { "epoch": 4.22, "learning_rate": 2.892188683327924e-05, "loss": 1.4098, "step": 78130 }, { "epoch": 4.22, "learning_rate": 2.891918208373905e-05, "loss": 1.4029, "step": 78140 }, { "epoch": 4.22, "learning_rate": 2.8916477334198856e-05, "loss": 1.3397, "step": 78150 }, { "epoch": 4.22, "learning_rate": 2.8913772584658662e-05, "loss": 1.327, "step": 78160 }, { "epoch": 4.22, "learning_rate": 2.891106783511847e-05, "loss": 1.3729, "step": 78170 }, { "epoch": 4.22, "learning_rate": 2.8908363085578278e-05, "loss": 1.3249, "step": 78180 }, { "epoch": 4.22, "learning_rate": 2.8905658336038084e-05, "loss": 1.4328, "step": 78190 }, { "epoch": 4.23, "learning_rate": 2.8902953586497894e-05, "loss": 1.3481, "step": 78200 }, { "epoch": 4.23, "learning_rate": 2.89002488369577e-05, "loss": 1.3292, "step": 78210 }, { "epoch": 4.23, "learning_rate": 2.8897544087417506e-05, "loss": 1.4524, "step": 78220 }, { "epoch": 4.23, "learning_rate": 2.8894839337877316e-05, "loss": 1.4008, "step": 78230 }, { "epoch": 4.23, "learning_rate": 2.8892134588337122e-05, "loss": 1.3471, "step": 78240 }, { "epoch": 4.23, "learning_rate": 2.888942983879693e-05, "loss": 1.4288, "step": 78250 }, { "epoch": 4.23, "learning_rate": 2.8886725089256738e-05, "loss": 1.3329, "step": 78260 }, { "epoch": 4.23, "learning_rate": 2.8884020339716544e-05, "loss": 1.3557, "step": 78270 }, { "epoch": 4.23, "learning_rate": 2.888131559017635e-05, "loss": 1.3828, "step": 78280 }, { "epoch": 4.23, "learning_rate": 2.887861084063616e-05, "loss": 1.2954, "step": 78290 }, { "epoch": 4.23, "learning_rate": 2.8875906091095966e-05, "loss": 1.4153, "step": 78300 }, { "epoch": 4.23, "learning_rate": 2.8873201341555773e-05, "loss": 1.4125, "step": 78310 }, { "epoch": 4.23, "learning_rate": 2.8870496592015582e-05, "loss": 1.3334, "step": 78320 }, { "epoch": 4.23, "learning_rate": 2.886779184247539e-05, "loss": 1.3958, "step": 78330 }, { "epoch": 4.23, "learning_rate": 2.8865087092935195e-05, "loss": 1.3653, "step": 78340 }, { "epoch": 4.23, "learning_rate": 2.8862382343395e-05, "loss": 1.3062, "step": 78350 }, { "epoch": 4.23, "learning_rate": 2.885967759385481e-05, "loss": 1.3166, "step": 78360 }, { "epoch": 4.23, "learning_rate": 2.8856972844314617e-05, "loss": 1.3406, "step": 78370 }, { "epoch": 4.24, "learning_rate": 2.8854268094774423e-05, "loss": 1.4585, "step": 78380 }, { "epoch": 4.24, "learning_rate": 2.8851563345234233e-05, "loss": 1.4185, "step": 78390 }, { "epoch": 4.24, "learning_rate": 2.884885859569404e-05, "loss": 1.3865, "step": 78400 }, { "epoch": 4.24, "learning_rate": 2.8846153846153845e-05, "loss": 1.3319, "step": 78410 }, { "epoch": 4.24, "learning_rate": 2.8843449096613655e-05, "loss": 1.4581, "step": 78420 }, { "epoch": 4.24, "learning_rate": 2.884074434707346e-05, "loss": 1.3794, "step": 78430 }, { "epoch": 4.24, "learning_rate": 2.8838039597533268e-05, "loss": 1.3615, "step": 78440 }, { "epoch": 4.24, "learning_rate": 2.8835334847993077e-05, "loss": 1.3564, "step": 78450 }, { "epoch": 4.24, "learning_rate": 2.8832630098452884e-05, "loss": 1.3901, "step": 78460 }, { "epoch": 4.24, "learning_rate": 2.882992534891269e-05, "loss": 1.3518, "step": 78470 }, { "epoch": 4.24, "learning_rate": 2.88272205993725e-05, "loss": 1.4053, "step": 78480 }, { "epoch": 4.24, "learning_rate": 2.8824515849832306e-05, "loss": 1.4196, "step": 78490 }, { "epoch": 4.24, "learning_rate": 2.8821811100292112e-05, "loss": 1.3795, "step": 78500 }, { "epoch": 4.24, "learning_rate": 2.881910635075192e-05, "loss": 1.3373, "step": 78510 }, { "epoch": 4.24, "learning_rate": 2.8816401601211728e-05, "loss": 1.3656, "step": 78520 }, { "epoch": 4.24, "learning_rate": 2.8813696851671534e-05, "loss": 1.4177, "step": 78530 }, { "epoch": 4.24, "learning_rate": 2.8810992102131347e-05, "loss": 1.3291, "step": 78540 }, { "epoch": 4.24, "learning_rate": 2.880828735259115e-05, "loss": 1.3144, "step": 78550 }, { "epoch": 4.24, "learning_rate": 2.8805582603050956e-05, "loss": 1.3251, "step": 78560 }, { "epoch": 4.25, "learning_rate": 2.880287785351077e-05, "loss": 1.3429, "step": 78570 }, { "epoch": 4.25, "learning_rate": 2.8800173103970572e-05, "loss": 1.4478, "step": 78580 }, { "epoch": 4.25, "learning_rate": 2.879746835443038e-05, "loss": 1.4476, "step": 78590 }, { "epoch": 4.25, "learning_rate": 2.879476360489019e-05, "loss": 1.3933, "step": 78600 }, { "epoch": 4.25, "learning_rate": 2.8792058855349994e-05, "loss": 1.2993, "step": 78610 }, { "epoch": 4.25, "learning_rate": 2.87893541058098e-05, "loss": 1.3604, "step": 78620 }, { "epoch": 4.25, "learning_rate": 2.8786649356269614e-05, "loss": 1.4126, "step": 78630 }, { "epoch": 4.25, "learning_rate": 2.878394460672942e-05, "loss": 1.4364, "step": 78640 }, { "epoch": 4.25, "learning_rate": 2.8781239857189223e-05, "loss": 1.3876, "step": 78650 }, { "epoch": 4.25, "learning_rate": 2.8778535107649036e-05, "loss": 1.3945, "step": 78660 }, { "epoch": 4.25, "learning_rate": 2.8775830358108842e-05, "loss": 1.3723, "step": 78670 }, { "epoch": 4.25, "learning_rate": 2.8773125608568645e-05, "loss": 1.382, "step": 78680 }, { "epoch": 4.25, "learning_rate": 2.8770420859028458e-05, "loss": 1.4148, "step": 78690 }, { "epoch": 4.25, "learning_rate": 2.8767716109488264e-05, "loss": 1.3388, "step": 78700 }, { "epoch": 4.25, "learning_rate": 2.8765011359948067e-05, "loss": 1.4274, "step": 78710 }, { "epoch": 4.25, "learning_rate": 2.876230661040788e-05, "loss": 1.328, "step": 78720 }, { "epoch": 4.25, "learning_rate": 2.8759601860867686e-05, "loss": 1.4328, "step": 78730 }, { "epoch": 4.25, "learning_rate": 2.875689711132749e-05, "loss": 1.3331, "step": 78740 }, { "epoch": 4.26, "learning_rate": 2.8754192361787302e-05, "loss": 1.2896, "step": 78750 }, { "epoch": 4.26, "learning_rate": 2.875148761224711e-05, "loss": 1.3411, "step": 78760 }, { "epoch": 4.26, "learning_rate": 2.8748782862706915e-05, "loss": 1.3403, "step": 78770 }, { "epoch": 4.26, "learning_rate": 2.8746078113166725e-05, "loss": 1.3473, "step": 78780 }, { "epoch": 4.26, "learning_rate": 2.874337336362653e-05, "loss": 1.4169, "step": 78790 }, { "epoch": 4.26, "learning_rate": 2.8740668614086337e-05, "loss": 1.3958, "step": 78800 }, { "epoch": 4.26, "learning_rate": 2.8737963864546147e-05, "loss": 1.3865, "step": 78810 }, { "epoch": 4.26, "learning_rate": 2.8735259115005953e-05, "loss": 1.4391, "step": 78820 }, { "epoch": 4.26, "learning_rate": 2.873255436546576e-05, "loss": 1.3341, "step": 78830 }, { "epoch": 4.26, "learning_rate": 2.872984961592557e-05, "loss": 1.4025, "step": 78840 }, { "epoch": 4.26, "learning_rate": 2.8727144866385375e-05, "loss": 1.354, "step": 78850 }, { "epoch": 4.26, "learning_rate": 2.872444011684518e-05, "loss": 1.3603, "step": 78860 }, { "epoch": 4.26, "learning_rate": 2.872173536730499e-05, "loss": 1.3326, "step": 78870 }, { "epoch": 4.26, "learning_rate": 2.8719030617764797e-05, "loss": 1.4404, "step": 78880 }, { "epoch": 4.26, "learning_rate": 2.8716325868224604e-05, "loss": 1.4234, "step": 78890 }, { "epoch": 4.26, "learning_rate": 2.8713621118684413e-05, "loss": 1.4011, "step": 78900 }, { "epoch": 4.26, "learning_rate": 2.871091636914422e-05, "loss": 1.3975, "step": 78910 }, { "epoch": 4.26, "learning_rate": 2.8708211619604026e-05, "loss": 1.3793, "step": 78920 }, { "epoch": 4.26, "learning_rate": 2.8705506870063832e-05, "loss": 1.4021, "step": 78930 }, { "epoch": 4.27, "learning_rate": 2.870280212052364e-05, "loss": 1.3191, "step": 78940 }, { "epoch": 4.27, "learning_rate": 2.8700097370983448e-05, "loss": 1.3644, "step": 78950 }, { "epoch": 4.27, "learning_rate": 2.8697392621443254e-05, "loss": 1.4344, "step": 78960 }, { "epoch": 4.27, "learning_rate": 2.8694687871903064e-05, "loss": 1.2925, "step": 78970 }, { "epoch": 4.27, "learning_rate": 2.869198312236287e-05, "loss": 1.4089, "step": 78980 }, { "epoch": 4.27, "learning_rate": 2.8689278372822676e-05, "loss": 1.3763, "step": 78990 }, { "epoch": 4.27, "learning_rate": 2.8686573623282486e-05, "loss": 1.3886, "step": 79000 }, { "epoch": 4.27, "learning_rate": 2.8683868873742292e-05, "loss": 1.3324, "step": 79010 }, { "epoch": 4.27, "learning_rate": 2.86811641242021e-05, "loss": 1.3796, "step": 79020 }, { "epoch": 4.27, "learning_rate": 2.8678459374661908e-05, "loss": 1.3905, "step": 79030 }, { "epoch": 4.27, "learning_rate": 2.8675754625121714e-05, "loss": 1.347, "step": 79040 }, { "epoch": 4.27, "learning_rate": 2.867304987558152e-05, "loss": 1.3745, "step": 79050 }, { "epoch": 4.27, "learning_rate": 2.867034512604133e-05, "loss": 1.3885, "step": 79060 }, { "epoch": 4.27, "learning_rate": 2.8667640376501137e-05, "loss": 1.4426, "step": 79070 }, { "epoch": 4.27, "learning_rate": 2.8664935626960943e-05, "loss": 1.4035, "step": 79080 }, { "epoch": 4.27, "learning_rate": 2.8662230877420752e-05, "loss": 1.3781, "step": 79090 }, { "epoch": 4.27, "learning_rate": 2.865952612788056e-05, "loss": 1.3896, "step": 79100 }, { "epoch": 4.27, "learning_rate": 2.8656821378340365e-05, "loss": 1.3087, "step": 79110 }, { "epoch": 4.28, "learning_rate": 2.8654116628800175e-05, "loss": 1.4659, "step": 79120 }, { "epoch": 4.28, "learning_rate": 2.865141187925998e-05, "loss": 1.3202, "step": 79130 }, { "epoch": 4.28, "learning_rate": 2.8648707129719787e-05, "loss": 1.3887, "step": 79140 }, { "epoch": 4.28, "learning_rate": 2.8646002380179597e-05, "loss": 1.4553, "step": 79150 }, { "epoch": 4.28, "learning_rate": 2.8643297630639403e-05, "loss": 1.4282, "step": 79160 }, { "epoch": 4.28, "learning_rate": 2.864059288109921e-05, "loss": 1.4256, "step": 79170 }, { "epoch": 4.28, "learning_rate": 2.8637888131559022e-05, "loss": 1.3939, "step": 79180 }, { "epoch": 4.28, "learning_rate": 2.8635183382018825e-05, "loss": 1.4211, "step": 79190 }, { "epoch": 4.28, "learning_rate": 2.863247863247863e-05, "loss": 1.4009, "step": 79200 }, { "epoch": 4.28, "learning_rate": 2.8629773882938445e-05, "loss": 1.3566, "step": 79210 }, { "epoch": 4.28, "learning_rate": 2.8627069133398247e-05, "loss": 1.4605, "step": 79220 }, { "epoch": 4.28, "learning_rate": 2.8624364383858054e-05, "loss": 1.3695, "step": 79230 }, { "epoch": 4.28, "learning_rate": 2.8621659634317867e-05, "loss": 1.3285, "step": 79240 }, { "epoch": 4.28, "learning_rate": 2.861895488477767e-05, "loss": 1.3771, "step": 79250 }, { "epoch": 4.28, "learning_rate": 2.8616250135237476e-05, "loss": 1.4324, "step": 79260 }, { "epoch": 4.28, "learning_rate": 2.861354538569729e-05, "loss": 1.3617, "step": 79270 }, { "epoch": 4.28, "learning_rate": 2.8610840636157095e-05, "loss": 1.3586, "step": 79280 }, { "epoch": 4.28, "learning_rate": 2.8608135886616898e-05, "loss": 1.4247, "step": 79290 }, { "epoch": 4.28, "learning_rate": 2.860543113707671e-05, "loss": 1.449, "step": 79300 }, { "epoch": 4.29, "learning_rate": 2.8602726387536517e-05, "loss": 1.3775, "step": 79310 }, { "epoch": 4.29, "learning_rate": 2.860002163799632e-05, "loss": 1.3741, "step": 79320 }, { "epoch": 4.29, "learning_rate": 2.8597316888456133e-05, "loss": 1.3587, "step": 79330 }, { "epoch": 4.29, "learning_rate": 2.859461213891594e-05, "loss": 1.4022, "step": 79340 }, { "epoch": 4.29, "learning_rate": 2.8591907389375742e-05, "loss": 1.3848, "step": 79350 }, { "epoch": 4.29, "learning_rate": 2.8589202639835555e-05, "loss": 1.4516, "step": 79360 }, { "epoch": 4.29, "learning_rate": 2.858649789029536e-05, "loss": 1.3972, "step": 79370 }, { "epoch": 4.29, "learning_rate": 2.8583793140755168e-05, "loss": 1.4024, "step": 79380 }, { "epoch": 4.29, "learning_rate": 2.8581088391214978e-05, "loss": 1.3345, "step": 79390 }, { "epoch": 4.29, "learning_rate": 2.8578383641674784e-05, "loss": 1.4237, "step": 79400 }, { "epoch": 4.29, "learning_rate": 2.857567889213459e-05, "loss": 1.4446, "step": 79410 }, { "epoch": 4.29, "learning_rate": 2.85729741425944e-05, "loss": 1.377, "step": 79420 }, { "epoch": 4.29, "learning_rate": 2.8570269393054206e-05, "loss": 1.4614, "step": 79430 }, { "epoch": 4.29, "learning_rate": 2.8567564643514012e-05, "loss": 1.3621, "step": 79440 }, { "epoch": 4.29, "learning_rate": 2.8564859893973822e-05, "loss": 1.4711, "step": 79450 }, { "epoch": 4.29, "learning_rate": 2.8562155144433628e-05, "loss": 1.3718, "step": 79460 }, { "epoch": 4.29, "learning_rate": 2.8559450394893434e-05, "loss": 1.3229, "step": 79470 }, { "epoch": 4.29, "learning_rate": 2.8556745645353237e-05, "loss": 1.393, "step": 79480 }, { "epoch": 4.3, "learning_rate": 2.855404089581305e-05, "loss": 1.409, "step": 79490 }, { "epoch": 4.3, "learning_rate": 2.8551336146272857e-05, "loss": 1.3964, "step": 79500 }, { "epoch": 4.3, "learning_rate": 2.8548631396732663e-05, "loss": 1.3474, "step": 79510 }, { "epoch": 4.3, "learning_rate": 2.8545926647192472e-05, "loss": 1.3664, "step": 79520 }, { "epoch": 4.3, "learning_rate": 2.854322189765228e-05, "loss": 1.4318, "step": 79530 }, { "epoch": 4.3, "learning_rate": 2.8540517148112085e-05, "loss": 1.4213, "step": 79540 }, { "epoch": 4.3, "learning_rate": 2.8537812398571895e-05, "loss": 1.4005, "step": 79550 }, { "epoch": 4.3, "learning_rate": 2.85351076490317e-05, "loss": 1.3607, "step": 79560 }, { "epoch": 4.3, "learning_rate": 2.8532402899491507e-05, "loss": 1.4093, "step": 79570 }, { "epoch": 4.3, "learning_rate": 2.8529698149951317e-05, "loss": 1.3887, "step": 79580 }, { "epoch": 4.3, "learning_rate": 2.8526993400411123e-05, "loss": 1.3891, "step": 79590 }, { "epoch": 4.3, "learning_rate": 2.852428865087093e-05, "loss": 1.2892, "step": 79600 }, { "epoch": 4.3, "learning_rate": 2.852158390133074e-05, "loss": 1.4261, "step": 79610 }, { "epoch": 4.3, "learning_rate": 2.8518879151790545e-05, "loss": 1.4319, "step": 79620 }, { "epoch": 4.3, "learning_rate": 2.851617440225035e-05, "loss": 1.414, "step": 79630 }, { "epoch": 4.3, "learning_rate": 2.851346965271016e-05, "loss": 1.3806, "step": 79640 }, { "epoch": 4.3, "learning_rate": 2.8510764903169967e-05, "loss": 1.4178, "step": 79650 }, { "epoch": 4.3, "learning_rate": 2.8508060153629774e-05, "loss": 1.4024, "step": 79660 }, { "epoch": 4.3, "learning_rate": 2.8505355404089583e-05, "loss": 1.4387, "step": 79670 }, { "epoch": 4.31, "learning_rate": 2.850265065454939e-05, "loss": 1.2905, "step": 79680 }, { "epoch": 4.31, "learning_rate": 2.8499945905009196e-05, "loss": 1.3049, "step": 79690 }, { "epoch": 4.31, "learning_rate": 2.8497241155469006e-05, "loss": 1.355, "step": 79700 }, { "epoch": 4.31, "learning_rate": 2.8494536405928812e-05, "loss": 1.3417, "step": 79710 }, { "epoch": 4.31, "learning_rate": 2.8491831656388618e-05, "loss": 1.4381, "step": 79720 }, { "epoch": 4.31, "learning_rate": 2.8489126906848428e-05, "loss": 1.3943, "step": 79730 }, { "epoch": 4.31, "learning_rate": 2.8486422157308234e-05, "loss": 1.3759, "step": 79740 }, { "epoch": 4.31, "learning_rate": 2.848371740776804e-05, "loss": 1.3619, "step": 79750 }, { "epoch": 4.31, "learning_rate": 2.848101265822785e-05, "loss": 1.4492, "step": 79760 }, { "epoch": 4.31, "learning_rate": 2.8478307908687656e-05, "loss": 1.3603, "step": 79770 }, { "epoch": 4.31, "learning_rate": 2.8475603159147462e-05, "loss": 1.3175, "step": 79780 }, { "epoch": 4.31, "learning_rate": 2.8472898409607275e-05, "loss": 1.3378, "step": 79790 }, { "epoch": 4.31, "learning_rate": 2.8470193660067078e-05, "loss": 1.3328, "step": 79800 }, { "epoch": 4.31, "learning_rate": 2.8467488910526885e-05, "loss": 1.4119, "step": 79810 }, { "epoch": 4.31, "learning_rate": 2.8464784160986698e-05, "loss": 1.323, "step": 79820 }, { "epoch": 4.31, "learning_rate": 2.84620794114465e-05, "loss": 1.345, "step": 79830 }, { "epoch": 4.31, "learning_rate": 2.8459374661906307e-05, "loss": 1.3667, "step": 79840 }, { "epoch": 4.31, "learning_rate": 2.845666991236612e-05, "loss": 1.3609, "step": 79850 }, { "epoch": 4.32, "learning_rate": 2.8453965162825923e-05, "loss": 1.4239, "step": 79860 }, { "epoch": 4.32, "learning_rate": 2.845126041328573e-05, "loss": 1.456, "step": 79870 }, { "epoch": 4.32, "learning_rate": 2.8448555663745542e-05, "loss": 1.3385, "step": 79880 }, { "epoch": 4.32, "learning_rate": 2.8445850914205345e-05, "loss": 1.4007, "step": 79890 }, { "epoch": 4.32, "learning_rate": 2.844314616466515e-05, "loss": 1.347, "step": 79900 }, { "epoch": 4.32, "learning_rate": 2.8440441415124964e-05, "loss": 1.3573, "step": 79910 }, { "epoch": 4.32, "learning_rate": 2.843773666558477e-05, "loss": 1.3921, "step": 79920 }, { "epoch": 4.32, "learning_rate": 2.8435031916044573e-05, "loss": 1.4668, "step": 79930 }, { "epoch": 4.32, "learning_rate": 2.8432327166504386e-05, "loss": 1.3486, "step": 79940 }, { "epoch": 4.32, "learning_rate": 2.8429622416964193e-05, "loss": 1.3948, "step": 79950 }, { "epoch": 4.32, "learning_rate": 2.8426917667423995e-05, "loss": 1.3736, "step": 79960 }, { "epoch": 4.32, "learning_rate": 2.842421291788381e-05, "loss": 1.389, "step": 79970 }, { "epoch": 4.32, "learning_rate": 2.8421508168343615e-05, "loss": 1.3408, "step": 79980 }, { "epoch": 4.32, "learning_rate": 2.8418803418803418e-05, "loss": 1.3928, "step": 79990 }, { "epoch": 4.32, "learning_rate": 2.841609866926323e-05, "loss": 1.3378, "step": 80000 }, { "epoch": 4.32, "learning_rate": 2.8413393919723037e-05, "loss": 1.3531, "step": 80010 }, { "epoch": 4.32, "learning_rate": 2.8410689170182843e-05, "loss": 1.4122, "step": 80020 }, { "epoch": 4.32, "learning_rate": 2.8407984420642646e-05, "loss": 1.356, "step": 80030 }, { "epoch": 4.32, "learning_rate": 2.840527967110246e-05, "loss": 1.3441, "step": 80040 }, { "epoch": 4.33, "learning_rate": 2.8402574921562265e-05, "loss": 1.3146, "step": 80050 }, { "epoch": 4.33, "learning_rate": 2.8399870172022068e-05, "loss": 1.4395, "step": 80060 }, { "epoch": 4.33, "learning_rate": 2.839716542248188e-05, "loss": 1.2943, "step": 80070 }, { "epoch": 4.33, "learning_rate": 2.8394460672941687e-05, "loss": 1.3439, "step": 80080 }, { "epoch": 4.33, "learning_rate": 2.839175592340149e-05, "loss": 1.4329, "step": 80090 }, { "epoch": 4.33, "learning_rate": 2.8389051173861303e-05, "loss": 1.4178, "step": 80100 }, { "epoch": 4.33, "learning_rate": 2.838634642432111e-05, "loss": 1.3026, "step": 80110 }, { "epoch": 4.33, "learning_rate": 2.8383641674780912e-05, "loss": 1.3576, "step": 80120 }, { "epoch": 4.33, "learning_rate": 2.8380936925240726e-05, "loss": 1.4198, "step": 80130 }, { "epoch": 4.33, "learning_rate": 2.8378232175700532e-05, "loss": 1.3757, "step": 80140 }, { "epoch": 4.33, "learning_rate": 2.8375527426160338e-05, "loss": 1.4113, "step": 80150 }, { "epoch": 4.33, "learning_rate": 2.8372822676620148e-05, "loss": 1.4365, "step": 80160 }, { "epoch": 4.33, "learning_rate": 2.8370117927079954e-05, "loss": 1.3885, "step": 80170 }, { "epoch": 4.33, "learning_rate": 2.836741317753976e-05, "loss": 1.2858, "step": 80180 }, { "epoch": 4.33, "learning_rate": 2.836470842799957e-05, "loss": 1.3923, "step": 80190 }, { "epoch": 4.33, "learning_rate": 2.8362003678459376e-05, "loss": 1.4881, "step": 80200 }, { "epoch": 4.33, "learning_rate": 2.8359298928919182e-05, "loss": 1.3432, "step": 80210 }, { "epoch": 4.33, "learning_rate": 2.8356594179378992e-05, "loss": 1.3901, "step": 80220 }, { "epoch": 4.34, "learning_rate": 2.8353889429838798e-05, "loss": 1.3501, "step": 80230 }, { "epoch": 4.34, "learning_rate": 2.8351184680298605e-05, "loss": 1.4073, "step": 80240 }, { "epoch": 4.34, "learning_rate": 2.8348479930758414e-05, "loss": 1.4088, "step": 80250 }, { "epoch": 4.34, "learning_rate": 2.834577518121822e-05, "loss": 1.4189, "step": 80260 }, { "epoch": 4.34, "learning_rate": 2.8343070431678027e-05, "loss": 1.3516, "step": 80270 }, { "epoch": 4.34, "learning_rate": 2.8340365682137836e-05, "loss": 1.4156, "step": 80280 }, { "epoch": 4.34, "learning_rate": 2.8337660932597643e-05, "loss": 1.3507, "step": 80290 }, { "epoch": 4.34, "learning_rate": 2.833495618305745e-05, "loss": 1.3461, "step": 80300 }, { "epoch": 4.34, "learning_rate": 2.833225143351726e-05, "loss": 1.4081, "step": 80310 }, { "epoch": 4.34, "learning_rate": 2.8329546683977065e-05, "loss": 1.3402, "step": 80320 }, { "epoch": 4.34, "learning_rate": 2.832684193443687e-05, "loss": 1.4666, "step": 80330 }, { "epoch": 4.34, "learning_rate": 2.832413718489668e-05, "loss": 1.3592, "step": 80340 }, { "epoch": 4.34, "learning_rate": 2.8321432435356487e-05, "loss": 1.3661, "step": 80350 }, { "epoch": 4.34, "learning_rate": 2.8318727685816293e-05, "loss": 1.3537, "step": 80360 }, { "epoch": 4.34, "learning_rate": 2.8316022936276103e-05, "loss": 1.4275, "step": 80370 }, { "epoch": 4.34, "learning_rate": 2.831331818673591e-05, "loss": 1.3623, "step": 80380 }, { "epoch": 4.34, "learning_rate": 2.8310613437195715e-05, "loss": 1.4857, "step": 80390 }, { "epoch": 4.34, "learning_rate": 2.8307908687655525e-05, "loss": 1.3832, "step": 80400 }, { "epoch": 4.34, "learning_rate": 2.830520393811533e-05, "loss": 1.3358, "step": 80410 }, { "epoch": 4.35, "learning_rate": 2.8302499188575138e-05, "loss": 1.357, "step": 80420 }, { "epoch": 4.35, "learning_rate": 2.829979443903495e-05, "loss": 1.3605, "step": 80430 }, { "epoch": 4.35, "learning_rate": 2.8297089689494753e-05, "loss": 1.3899, "step": 80440 }, { "epoch": 4.35, "learning_rate": 2.829438493995456e-05, "loss": 1.366, "step": 80450 }, { "epoch": 4.35, "learning_rate": 2.8291680190414373e-05, "loss": 1.4089, "step": 80460 }, { "epoch": 4.35, "learning_rate": 2.8288975440874176e-05, "loss": 1.3662, "step": 80470 }, { "epoch": 4.35, "learning_rate": 2.8286270691333982e-05, "loss": 1.3754, "step": 80480 }, { "epoch": 4.35, "learning_rate": 2.8283565941793795e-05, "loss": 1.3842, "step": 80490 }, { "epoch": 4.35, "learning_rate": 2.8280861192253598e-05, "loss": 1.4211, "step": 80500 }, { "epoch": 4.35, "learning_rate": 2.8278156442713404e-05, "loss": 1.4426, "step": 80510 }, { "epoch": 4.35, "learning_rate": 2.8275451693173217e-05, "loss": 1.4528, "step": 80520 }, { "epoch": 4.35, "learning_rate": 2.827274694363302e-05, "loss": 1.3802, "step": 80530 }, { "epoch": 4.35, "learning_rate": 2.8270042194092826e-05, "loss": 1.3572, "step": 80540 }, { "epoch": 4.35, "learning_rate": 2.826733744455264e-05, "loss": 1.4232, "step": 80550 }, { "epoch": 4.35, "learning_rate": 2.8264632695012446e-05, "loss": 1.4341, "step": 80560 }, { "epoch": 4.35, "learning_rate": 2.826192794547225e-05, "loss": 1.3644, "step": 80570 }, { "epoch": 4.35, "learning_rate": 2.825922319593206e-05, "loss": 1.4063, "step": 80580 }, { "epoch": 4.35, "learning_rate": 2.8256518446391868e-05, "loss": 1.3296, "step": 80590 }, { "epoch": 4.36, "learning_rate": 2.825381369685167e-05, "loss": 1.3697, "step": 80600 }, { "epoch": 4.36, "learning_rate": 2.8251108947311477e-05, "loss": 1.3579, "step": 80610 }, { "epoch": 4.36, "learning_rate": 2.824840419777129e-05, "loss": 1.3934, "step": 80620 }, { "epoch": 4.36, "learning_rate": 2.8245699448231093e-05, "loss": 1.4601, "step": 80630 }, { "epoch": 4.36, "learning_rate": 2.82429946986909e-05, "loss": 1.3903, "step": 80640 }, { "epoch": 4.36, "learning_rate": 2.8240289949150712e-05, "loss": 1.3398, "step": 80650 }, { "epoch": 4.36, "learning_rate": 2.823758519961052e-05, "loss": 1.4082, "step": 80660 }, { "epoch": 4.36, "learning_rate": 2.823488045007032e-05, "loss": 1.3011, "step": 80670 }, { "epoch": 4.36, "learning_rate": 2.8232175700530134e-05, "loss": 1.4097, "step": 80680 }, { "epoch": 4.36, "learning_rate": 2.822947095098994e-05, "loss": 1.2872, "step": 80690 }, { "epoch": 4.36, "learning_rate": 2.8226766201449743e-05, "loss": 1.3727, "step": 80700 }, { "epoch": 4.36, "learning_rate": 2.8224061451909556e-05, "loss": 1.4872, "step": 80710 }, { "epoch": 4.36, "learning_rate": 2.8221356702369363e-05, "loss": 1.3521, "step": 80720 }, { "epoch": 4.36, "learning_rate": 2.8218651952829166e-05, "loss": 1.3256, "step": 80730 }, { "epoch": 4.36, "learning_rate": 2.821594720328898e-05, "loss": 1.4302, "step": 80740 }, { "epoch": 4.36, "learning_rate": 2.8213242453748785e-05, "loss": 1.3751, "step": 80750 }, { "epoch": 4.36, "learning_rate": 2.821053770420859e-05, "loss": 1.4481, "step": 80760 }, { "epoch": 4.36, "learning_rate": 2.82078329546684e-05, "loss": 1.4294, "step": 80770 }, { "epoch": 4.36, "learning_rate": 2.8205128205128207e-05, "loss": 1.3879, "step": 80780 }, { "epoch": 4.37, "learning_rate": 2.8202423455588013e-05, "loss": 1.3909, "step": 80790 }, { "epoch": 4.37, "learning_rate": 2.8199718706047823e-05, "loss": 1.3654, "step": 80800 }, { "epoch": 4.37, "learning_rate": 2.819701395650763e-05, "loss": 1.3679, "step": 80810 }, { "epoch": 4.37, "learning_rate": 2.8194309206967435e-05, "loss": 1.3568, "step": 80820 }, { "epoch": 4.37, "learning_rate": 2.8191604457427245e-05, "loss": 1.4155, "step": 80830 }, { "epoch": 4.37, "learning_rate": 2.818889970788705e-05, "loss": 1.3424, "step": 80840 }, { "epoch": 4.37, "learning_rate": 2.8186194958346858e-05, "loss": 1.3099, "step": 80850 }, { "epoch": 4.37, "learning_rate": 2.8183490208806667e-05, "loss": 1.3765, "step": 80860 }, { "epoch": 4.37, "learning_rate": 2.8180785459266474e-05, "loss": 1.3087, "step": 80870 }, { "epoch": 4.37, "learning_rate": 2.817808070972628e-05, "loss": 1.392, "step": 80880 }, { "epoch": 4.37, "learning_rate": 2.817537596018609e-05, "loss": 1.3834, "step": 80890 }, { "epoch": 4.37, "learning_rate": 2.8172671210645896e-05, "loss": 1.3871, "step": 80900 }, { "epoch": 4.37, "learning_rate": 2.8169966461105702e-05, "loss": 1.4668, "step": 80910 }, { "epoch": 4.37, "learning_rate": 2.816726171156551e-05, "loss": 1.4064, "step": 80920 }, { "epoch": 4.37, "learning_rate": 2.8164556962025318e-05, "loss": 1.3796, "step": 80930 }, { "epoch": 4.37, "learning_rate": 2.8161852212485124e-05, "loss": 1.3534, "step": 80940 }, { "epoch": 4.37, "learning_rate": 2.8159147462944934e-05, "loss": 1.3677, "step": 80950 }, { "epoch": 4.37, "learning_rate": 2.815644271340474e-05, "loss": 1.3322, "step": 80960 }, { "epoch": 4.38, "learning_rate": 2.8153737963864546e-05, "loss": 1.4325, "step": 80970 }, { "epoch": 4.38, "learning_rate": 2.8151033214324356e-05, "loss": 1.42, "step": 80980 }, { "epoch": 4.38, "learning_rate": 2.8148328464784162e-05, "loss": 1.3468, "step": 80990 }, { "epoch": 4.38, "learning_rate": 2.814562371524397e-05, "loss": 1.3052, "step": 81000 }, { "epoch": 4.38, "learning_rate": 2.8142918965703778e-05, "loss": 1.3699, "step": 81010 }, { "epoch": 4.38, "learning_rate": 2.8140214216163584e-05, "loss": 1.3019, "step": 81020 }, { "epoch": 4.38, "learning_rate": 2.813750946662339e-05, "loss": 1.3719, "step": 81030 }, { "epoch": 4.38, "learning_rate": 2.81348047170832e-05, "loss": 1.354, "step": 81040 }, { "epoch": 4.38, "learning_rate": 2.8132099967543007e-05, "loss": 1.3546, "step": 81050 }, { "epoch": 4.38, "learning_rate": 2.8129395218002813e-05, "loss": 1.3437, "step": 81060 }, { "epoch": 4.38, "learning_rate": 2.8126690468462626e-05, "loss": 1.4279, "step": 81070 }, { "epoch": 4.38, "learning_rate": 2.812398571892243e-05, "loss": 1.3701, "step": 81080 }, { "epoch": 4.38, "learning_rate": 2.8121280969382235e-05, "loss": 1.3449, "step": 81090 }, { "epoch": 4.38, "learning_rate": 2.8118576219842048e-05, "loss": 1.3114, "step": 81100 }, { "epoch": 4.38, "learning_rate": 2.811587147030185e-05, "loss": 1.4189, "step": 81110 }, { "epoch": 4.38, "learning_rate": 2.8113166720761657e-05, "loss": 1.4636, "step": 81120 }, { "epoch": 4.38, "learning_rate": 2.811046197122147e-05, "loss": 1.3493, "step": 81130 }, { "epoch": 4.38, "learning_rate": 2.8107757221681273e-05, "loss": 1.4181, "step": 81140 }, { "epoch": 4.38, "learning_rate": 2.810505247214108e-05, "loss": 1.3381, "step": 81150 }, { "epoch": 4.39, "learning_rate": 2.8102347722600886e-05, "loss": 1.3933, "step": 81160 }, { "epoch": 4.39, "learning_rate": 2.80996429730607e-05, "loss": 1.3396, "step": 81170 }, { "epoch": 4.39, "learning_rate": 2.80969382235205e-05, "loss": 1.4374, "step": 81180 }, { "epoch": 4.39, "learning_rate": 2.8094233473980308e-05, "loss": 1.3609, "step": 81190 }, { "epoch": 4.39, "learning_rate": 2.809152872444012e-05, "loss": 1.4089, "step": 81200 }, { "epoch": 4.39, "learning_rate": 2.8088823974899924e-05, "loss": 1.4465, "step": 81210 }, { "epoch": 4.39, "learning_rate": 2.808611922535973e-05, "loss": 1.3886, "step": 81220 }, { "epoch": 4.39, "learning_rate": 2.8083414475819543e-05, "loss": 1.3632, "step": 81230 }, { "epoch": 4.39, "learning_rate": 2.8080709726279346e-05, "loss": 1.4567, "step": 81240 }, { "epoch": 4.39, "learning_rate": 2.8078004976739152e-05, "loss": 1.3581, "step": 81250 }, { "epoch": 4.39, "learning_rate": 2.8075300227198965e-05, "loss": 1.3789, "step": 81260 }, { "epoch": 4.39, "learning_rate": 2.8072595477658768e-05, "loss": 1.3525, "step": 81270 }, { "epoch": 4.39, "learning_rate": 2.8069890728118574e-05, "loss": 1.2351, "step": 81280 }, { "epoch": 4.39, "learning_rate": 2.8067185978578387e-05, "loss": 1.2954, "step": 81290 }, { "epoch": 4.39, "learning_rate": 2.8064481229038194e-05, "loss": 1.4242, "step": 81300 }, { "epoch": 4.39, "learning_rate": 2.8061776479497996e-05, "loss": 1.4093, "step": 81310 }, { "epoch": 4.39, "learning_rate": 2.805907172995781e-05, "loss": 1.3596, "step": 81320 }, { "epoch": 4.39, "learning_rate": 2.8056366980417616e-05, "loss": 1.378, "step": 81330 }, { "epoch": 4.4, "learning_rate": 2.805366223087742e-05, "loss": 1.3689, "step": 81340 }, { "epoch": 4.4, "learning_rate": 2.805095748133723e-05, "loss": 1.3523, "step": 81350 }, { "epoch": 4.4, "learning_rate": 2.8048252731797038e-05, "loss": 1.398, "step": 81360 }, { "epoch": 4.4, "learning_rate": 2.804554798225684e-05, "loss": 1.376, "step": 81370 }, { "epoch": 4.4, "learning_rate": 2.8042843232716654e-05, "loss": 1.3364, "step": 81380 }, { "epoch": 4.4, "learning_rate": 2.804013848317646e-05, "loss": 1.4647, "step": 81390 }, { "epoch": 4.4, "learning_rate": 2.8037433733636266e-05, "loss": 1.3954, "step": 81400 }, { "epoch": 4.4, "learning_rate": 2.8034728984096076e-05, "loss": 1.3405, "step": 81410 }, { "epoch": 4.4, "learning_rate": 2.8032024234555882e-05, "loss": 1.3188, "step": 81420 }, { "epoch": 4.4, "learning_rate": 2.802931948501569e-05, "loss": 1.3608, "step": 81430 }, { "epoch": 4.4, "learning_rate": 2.8026614735475498e-05, "loss": 1.38, "step": 81440 }, { "epoch": 4.4, "learning_rate": 2.8023909985935304e-05, "loss": 1.3964, "step": 81450 }, { "epoch": 4.4, "learning_rate": 2.802120523639511e-05, "loss": 1.3316, "step": 81460 }, { "epoch": 4.4, "learning_rate": 2.801850048685492e-05, "loss": 1.4116, "step": 81470 }, { "epoch": 4.4, "learning_rate": 2.8015795737314727e-05, "loss": 1.3358, "step": 81480 }, { "epoch": 4.4, "learning_rate": 2.8013090987774533e-05, "loss": 1.3585, "step": 81490 }, { "epoch": 4.4, "learning_rate": 2.8010386238234342e-05, "loss": 1.3225, "step": 81500 }, { "epoch": 4.4, "learning_rate": 2.800768148869415e-05, "loss": 1.3339, "step": 81510 }, { "epoch": 4.4, "learning_rate": 2.8004976739153955e-05, "loss": 1.3931, "step": 81520 }, { "epoch": 4.41, "learning_rate": 2.8002271989613765e-05, "loss": 1.4068, "step": 81530 }, { "epoch": 4.41, "learning_rate": 2.799956724007357e-05, "loss": 1.3662, "step": 81540 }, { "epoch": 4.41, "learning_rate": 2.7996862490533377e-05, "loss": 1.3399, "step": 81550 }, { "epoch": 4.41, "learning_rate": 2.7994157740993187e-05, "loss": 1.3978, "step": 81560 }, { "epoch": 4.41, "learning_rate": 2.7991452991452993e-05, "loss": 1.3299, "step": 81570 }, { "epoch": 4.41, "learning_rate": 2.79887482419128e-05, "loss": 1.4173, "step": 81580 }, { "epoch": 4.41, "learning_rate": 2.798604349237261e-05, "loss": 1.3148, "step": 81590 }, { "epoch": 4.41, "learning_rate": 2.7983338742832415e-05, "loss": 1.3684, "step": 81600 }, { "epoch": 4.41, "learning_rate": 2.798063399329222e-05, "loss": 1.4161, "step": 81610 }, { "epoch": 4.41, "learning_rate": 2.797792924375203e-05, "loss": 1.4572, "step": 81620 }, { "epoch": 4.41, "learning_rate": 2.7975224494211837e-05, "loss": 1.3343, "step": 81630 }, { "epoch": 4.41, "learning_rate": 2.7972519744671644e-05, "loss": 1.3941, "step": 81640 }, { "epoch": 4.41, "learning_rate": 2.7969814995131453e-05, "loss": 1.4071, "step": 81650 }, { "epoch": 4.41, "learning_rate": 2.796711024559126e-05, "loss": 1.411, "step": 81660 }, { "epoch": 4.41, "learning_rate": 2.7964405496051066e-05, "loss": 1.4845, "step": 81670 }, { "epoch": 4.41, "learning_rate": 2.7961700746510875e-05, "loss": 1.4109, "step": 81680 }, { "epoch": 4.41, "learning_rate": 2.7958995996970682e-05, "loss": 1.3693, "step": 81690 }, { "epoch": 4.41, "learning_rate": 2.7956291247430488e-05, "loss": 1.3522, "step": 81700 }, { "epoch": 4.42, "learning_rate": 2.7953586497890294e-05, "loss": 1.43, "step": 81710 }, { "epoch": 4.42, "learning_rate": 2.7950881748350104e-05, "loss": 1.3549, "step": 81720 }, { "epoch": 4.42, "learning_rate": 2.794817699880991e-05, "loss": 1.2778, "step": 81730 }, { "epoch": 4.42, "learning_rate": 2.7945472249269716e-05, "loss": 1.4058, "step": 81740 }, { "epoch": 4.42, "learning_rate": 2.7942767499729526e-05, "loss": 1.3796, "step": 81750 }, { "epoch": 4.42, "learning_rate": 2.7940062750189332e-05, "loss": 1.3737, "step": 81760 }, { "epoch": 4.42, "learning_rate": 2.793735800064914e-05, "loss": 1.3649, "step": 81770 }, { "epoch": 4.42, "learning_rate": 2.7934653251108948e-05, "loss": 1.333, "step": 81780 }, { "epoch": 4.42, "learning_rate": 2.7931948501568755e-05, "loss": 1.3578, "step": 81790 }, { "epoch": 4.42, "learning_rate": 2.792924375202856e-05, "loss": 1.4007, "step": 81800 }, { "epoch": 4.42, "learning_rate": 2.7926539002488374e-05, "loss": 1.4411, "step": 81810 }, { "epoch": 4.42, "learning_rate": 2.7923834252948177e-05, "loss": 1.3833, "step": 81820 }, { "epoch": 4.42, "learning_rate": 2.7921129503407983e-05, "loss": 1.4167, "step": 81830 }, { "epoch": 4.42, "learning_rate": 2.7918424753867796e-05, "loss": 1.3747, "step": 81840 }, { "epoch": 4.42, "learning_rate": 2.79157200043276e-05, "loss": 1.3506, "step": 81850 }, { "epoch": 4.42, "learning_rate": 2.7913015254787405e-05, "loss": 1.4357, "step": 81860 }, { "epoch": 4.42, "learning_rate": 2.7910310505247218e-05, "loss": 1.3819, "step": 81870 }, { "epoch": 4.42, "learning_rate": 2.790760575570702e-05, "loss": 1.3447, "step": 81880 }, { "epoch": 4.42, "learning_rate": 2.7904901006166827e-05, "loss": 1.3171, "step": 81890 }, { "epoch": 4.43, "learning_rate": 2.790219625662664e-05, "loss": 1.3105, "step": 81900 }, { "epoch": 4.43, "learning_rate": 2.7899491507086447e-05, "loss": 1.3993, "step": 81910 }, { "epoch": 4.43, "learning_rate": 2.789678675754625e-05, "loss": 1.4147, "step": 81920 }, { "epoch": 4.43, "learning_rate": 2.7894082008006062e-05, "loss": 1.3492, "step": 81930 }, { "epoch": 4.43, "learning_rate": 2.789137725846587e-05, "loss": 1.363, "step": 81940 }, { "epoch": 4.43, "learning_rate": 2.788867250892567e-05, "loss": 1.4421, "step": 81950 }, { "epoch": 4.43, "learning_rate": 2.7885967759385485e-05, "loss": 1.3059, "step": 81960 }, { "epoch": 4.43, "learning_rate": 2.788326300984529e-05, "loss": 1.3892, "step": 81970 }, { "epoch": 4.43, "learning_rate": 2.7880558260305094e-05, "loss": 1.3882, "step": 81980 }, { "epoch": 4.43, "learning_rate": 2.7877853510764907e-05, "loss": 1.3908, "step": 81990 }, { "epoch": 4.43, "learning_rate": 2.7875148761224713e-05, "loss": 1.2973, "step": 82000 }, { "epoch": 4.43, "learning_rate": 2.7872444011684516e-05, "loss": 1.3117, "step": 82010 }, { "epoch": 4.43, "learning_rate": 2.786973926214433e-05, "loss": 1.3597, "step": 82020 }, { "epoch": 4.43, "learning_rate": 2.7867034512604135e-05, "loss": 1.3988, "step": 82030 }, { "epoch": 4.43, "learning_rate": 2.786432976306394e-05, "loss": 1.4048, "step": 82040 }, { "epoch": 4.43, "learning_rate": 2.786162501352375e-05, "loss": 1.316, "step": 82050 }, { "epoch": 4.43, "learning_rate": 2.7858920263983557e-05, "loss": 1.392, "step": 82060 }, { "epoch": 4.43, "learning_rate": 2.7856215514443364e-05, "loss": 1.3771, "step": 82070 }, { "epoch": 4.44, "learning_rate": 2.7853510764903173e-05, "loss": 1.3731, "step": 82080 }, { "epoch": 4.44, "learning_rate": 2.785080601536298e-05, "loss": 1.3306, "step": 82090 }, { "epoch": 4.44, "learning_rate": 2.7848101265822786e-05, "loss": 1.3628, "step": 82100 }, { "epoch": 4.44, "learning_rate": 2.7845396516282595e-05, "loss": 1.3272, "step": 82110 }, { "epoch": 4.44, "learning_rate": 2.7842691766742402e-05, "loss": 1.391, "step": 82120 }, { "epoch": 4.44, "learning_rate": 2.7839987017202208e-05, "loss": 1.3421, "step": 82130 }, { "epoch": 4.44, "learning_rate": 2.7837282267662018e-05, "loss": 1.4079, "step": 82140 }, { "epoch": 4.44, "learning_rate": 2.7834577518121824e-05, "loss": 1.3657, "step": 82150 }, { "epoch": 4.44, "learning_rate": 2.783187276858163e-05, "loss": 1.4516, "step": 82160 }, { "epoch": 4.44, "learning_rate": 2.782916801904144e-05, "loss": 1.3279, "step": 82170 }, { "epoch": 4.44, "learning_rate": 2.7826463269501246e-05, "loss": 1.3812, "step": 82180 }, { "epoch": 4.44, "learning_rate": 2.7823758519961052e-05, "loss": 1.428, "step": 82190 }, { "epoch": 4.44, "learning_rate": 2.7821053770420862e-05, "loss": 1.3175, "step": 82200 }, { "epoch": 4.44, "learning_rate": 2.7818349020880668e-05, "loss": 1.3892, "step": 82210 }, { "epoch": 4.44, "learning_rate": 2.7815644271340475e-05, "loss": 1.3378, "step": 82220 }, { "epoch": 4.44, "learning_rate": 2.7812939521800284e-05, "loss": 1.3266, "step": 82230 }, { "epoch": 4.44, "learning_rate": 2.781023477226009e-05, "loss": 1.4312, "step": 82240 }, { "epoch": 4.44, "learning_rate": 2.7807530022719897e-05, "loss": 1.3549, "step": 82250 }, { "epoch": 4.44, "learning_rate": 2.7804825273179706e-05, "loss": 1.4741, "step": 82260 }, { "epoch": 4.45, "learning_rate": 2.7802120523639513e-05, "loss": 1.2741, "step": 82270 }, { "epoch": 4.45, "learning_rate": 2.779941577409932e-05, "loss": 1.4071, "step": 82280 }, { "epoch": 4.45, "learning_rate": 2.7796711024559125e-05, "loss": 1.3126, "step": 82290 }, { "epoch": 4.45, "learning_rate": 2.7794006275018935e-05, "loss": 1.3148, "step": 82300 }, { "epoch": 4.45, "learning_rate": 2.779130152547874e-05, "loss": 1.416, "step": 82310 }, { "epoch": 4.45, "learning_rate": 2.7788596775938547e-05, "loss": 1.3317, "step": 82320 }, { "epoch": 4.45, "learning_rate": 2.7785892026398357e-05, "loss": 1.4154, "step": 82330 }, { "epoch": 4.45, "learning_rate": 2.7783187276858163e-05, "loss": 1.3049, "step": 82340 }, { "epoch": 4.45, "learning_rate": 2.778048252731797e-05, "loss": 1.3661, "step": 82350 }, { "epoch": 4.45, "learning_rate": 2.777777777777778e-05, "loss": 1.399, "step": 82360 }, { "epoch": 4.45, "learning_rate": 2.7775073028237585e-05, "loss": 1.3752, "step": 82370 }, { "epoch": 4.45, "learning_rate": 2.777236827869739e-05, "loss": 1.3929, "step": 82380 }, { "epoch": 4.45, "learning_rate": 2.77696635291572e-05, "loss": 1.3638, "step": 82390 }, { "epoch": 4.45, "learning_rate": 2.7766958779617008e-05, "loss": 1.3618, "step": 82400 }, { "epoch": 4.45, "learning_rate": 2.7764254030076814e-05, "loss": 1.4258, "step": 82410 }, { "epoch": 4.45, "learning_rate": 2.7761549280536623e-05, "loss": 1.3098, "step": 82420 }, { "epoch": 4.45, "learning_rate": 2.775884453099643e-05, "loss": 1.4241, "step": 82430 }, { "epoch": 4.45, "learning_rate": 2.7756139781456236e-05, "loss": 1.3716, "step": 82440 }, { "epoch": 4.46, "learning_rate": 2.775343503191605e-05, "loss": 1.4221, "step": 82450 }, { "epoch": 4.46, "learning_rate": 2.7750730282375852e-05, "loss": 1.3818, "step": 82460 }, { "epoch": 4.46, "learning_rate": 2.7748025532835658e-05, "loss": 1.3917, "step": 82470 }, { "epoch": 4.46, "learning_rate": 2.774532078329547e-05, "loss": 1.4018, "step": 82480 }, { "epoch": 4.46, "learning_rate": 2.7742616033755274e-05, "loss": 1.4146, "step": 82490 }, { "epoch": 4.46, "learning_rate": 2.773991128421508e-05, "loss": 1.3713, "step": 82500 }, { "epoch": 4.46, "learning_rate": 2.7737206534674893e-05, "loss": 1.4511, "step": 82510 }, { "epoch": 4.46, "learning_rate": 2.7734501785134696e-05, "loss": 1.3994, "step": 82520 }, { "epoch": 4.46, "learning_rate": 2.7731797035594502e-05, "loss": 1.4318, "step": 82530 }, { "epoch": 4.46, "learning_rate": 2.7729092286054316e-05, "loss": 1.3148, "step": 82540 }, { "epoch": 4.46, "learning_rate": 2.7726387536514122e-05, "loss": 1.374, "step": 82550 }, { "epoch": 4.46, "learning_rate": 2.7723682786973925e-05, "loss": 1.3664, "step": 82560 }, { "epoch": 4.46, "learning_rate": 2.7720978037433738e-05, "loss": 1.4421, "step": 82570 }, { "epoch": 4.46, "learning_rate": 2.7718273287893544e-05, "loss": 1.3918, "step": 82580 }, { "epoch": 4.46, "learning_rate": 2.7715568538353347e-05, "loss": 1.312, "step": 82590 }, { "epoch": 4.46, "learning_rate": 2.771286378881316e-05, "loss": 1.3087, "step": 82600 }, { "epoch": 4.46, "learning_rate": 2.7710159039272966e-05, "loss": 1.4296, "step": 82610 }, { "epoch": 4.46, "learning_rate": 2.770745428973277e-05, "loss": 1.4324, "step": 82620 }, { "epoch": 4.46, "learning_rate": 2.7704749540192582e-05, "loss": 1.3803, "step": 82630 }, { "epoch": 4.47, "learning_rate": 2.7702044790652388e-05, "loss": 1.4111, "step": 82640 }, { "epoch": 4.47, "learning_rate": 2.769934004111219e-05, "loss": 1.45, "step": 82650 }, { "epoch": 4.47, "learning_rate": 2.7696635291572004e-05, "loss": 1.4383, "step": 82660 }, { "epoch": 4.47, "learning_rate": 2.769393054203181e-05, "loss": 1.342, "step": 82670 }, { "epoch": 4.47, "learning_rate": 2.7691225792491617e-05, "loss": 1.28, "step": 82680 }, { "epoch": 4.47, "learning_rate": 2.7688521042951426e-05, "loss": 1.3522, "step": 82690 }, { "epoch": 4.47, "learning_rate": 2.7685816293411233e-05, "loss": 1.3816, "step": 82700 }, { "epoch": 4.47, "learning_rate": 2.768311154387104e-05, "loss": 1.3267, "step": 82710 }, { "epoch": 4.47, "learning_rate": 2.768040679433085e-05, "loss": 1.4258, "step": 82720 }, { "epoch": 4.47, "learning_rate": 2.7677702044790655e-05, "loss": 1.3703, "step": 82730 }, { "epoch": 4.47, "learning_rate": 2.767499729525046e-05, "loss": 1.3557, "step": 82740 }, { "epoch": 4.47, "learning_rate": 2.767229254571027e-05, "loss": 1.3838, "step": 82750 }, { "epoch": 4.47, "learning_rate": 2.7669587796170077e-05, "loss": 1.342, "step": 82760 }, { "epoch": 4.47, "learning_rate": 2.7666883046629883e-05, "loss": 1.316, "step": 82770 }, { "epoch": 4.47, "learning_rate": 2.7664178297089693e-05, "loss": 1.3852, "step": 82780 }, { "epoch": 4.47, "learning_rate": 2.76614735475495e-05, "loss": 1.42, "step": 82790 }, { "epoch": 4.47, "learning_rate": 2.7658768798009305e-05, "loss": 1.3488, "step": 82800 }, { "epoch": 4.47, "learning_rate": 2.7656064048469115e-05, "loss": 1.3539, "step": 82810 }, { "epoch": 4.48, "learning_rate": 2.765335929892892e-05, "loss": 1.3429, "step": 82820 }, { "epoch": 4.48, "learning_rate": 2.7650654549388728e-05, "loss": 1.4392, "step": 82830 }, { "epoch": 4.48, "learning_rate": 2.7647949799848534e-05, "loss": 1.3743, "step": 82840 }, { "epoch": 4.48, "learning_rate": 2.7645245050308343e-05, "loss": 1.3292, "step": 82850 }, { "epoch": 4.48, "learning_rate": 2.764254030076815e-05, "loss": 1.3788, "step": 82860 }, { "epoch": 4.48, "learning_rate": 2.7639835551227956e-05, "loss": 1.4081, "step": 82870 }, { "epoch": 4.48, "learning_rate": 2.7637130801687766e-05, "loss": 1.3622, "step": 82880 }, { "epoch": 4.48, "learning_rate": 2.7634426052147572e-05, "loss": 1.2852, "step": 82890 }, { "epoch": 4.48, "learning_rate": 2.7631721302607378e-05, "loss": 1.3773, "step": 82900 }, { "epoch": 4.48, "learning_rate": 2.7629016553067188e-05, "loss": 1.3964, "step": 82910 }, { "epoch": 4.48, "learning_rate": 2.7626311803526994e-05, "loss": 1.4359, "step": 82920 }, { "epoch": 4.48, "learning_rate": 2.76236070539868e-05, "loss": 1.3779, "step": 82930 }, { "epoch": 4.48, "learning_rate": 2.762090230444661e-05, "loss": 1.4159, "step": 82940 }, { "epoch": 4.48, "learning_rate": 2.7618197554906416e-05, "loss": 1.3664, "step": 82950 }, { "epoch": 4.48, "learning_rate": 2.7615492805366222e-05, "loss": 1.3826, "step": 82960 }, { "epoch": 4.48, "learning_rate": 2.7612788055826032e-05, "loss": 1.4489, "step": 82970 }, { "epoch": 4.48, "learning_rate": 2.761008330628584e-05, "loss": 1.3296, "step": 82980 }, { "epoch": 4.48, "learning_rate": 2.7607378556745645e-05, "loss": 1.4247, "step": 82990 }, { "epoch": 4.48, "learning_rate": 2.7604673807205454e-05, "loss": 1.3983, "step": 83000 }, { "epoch": 4.49, "learning_rate": 2.760196905766526e-05, "loss": 1.5101, "step": 83010 }, { "epoch": 4.49, "learning_rate": 2.7599264308125067e-05, "loss": 1.4483, "step": 83020 }, { "epoch": 4.49, "learning_rate": 2.7596559558584876e-05, "loss": 1.4037, "step": 83030 }, { "epoch": 4.49, "learning_rate": 2.7593854809044683e-05, "loss": 1.3799, "step": 83040 }, { "epoch": 4.49, "learning_rate": 2.759115005950449e-05, "loss": 1.3827, "step": 83050 }, { "epoch": 4.49, "learning_rate": 2.75884453099643e-05, "loss": 1.4561, "step": 83060 }, { "epoch": 4.49, "learning_rate": 2.7585740560424105e-05, "loss": 1.4178, "step": 83070 }, { "epoch": 4.49, "learning_rate": 2.758303581088391e-05, "loss": 1.3876, "step": 83080 }, { "epoch": 4.49, "learning_rate": 2.7580331061343724e-05, "loss": 1.3722, "step": 83090 }, { "epoch": 4.49, "learning_rate": 2.7577626311803527e-05, "loss": 1.4121, "step": 83100 }, { "epoch": 4.49, "learning_rate": 2.7574921562263333e-05, "loss": 1.3988, "step": 83110 }, { "epoch": 4.49, "learning_rate": 2.7572216812723146e-05, "loss": 1.4362, "step": 83120 }, { "epoch": 4.49, "learning_rate": 2.756951206318295e-05, "loss": 1.3289, "step": 83130 }, { "epoch": 4.49, "learning_rate": 2.7566807313642756e-05, "loss": 1.3848, "step": 83140 }, { "epoch": 4.49, "learning_rate": 2.756410256410257e-05, "loss": 1.3426, "step": 83150 }, { "epoch": 4.49, "learning_rate": 2.756139781456237e-05, "loss": 1.4095, "step": 83160 }, { "epoch": 4.49, "learning_rate": 2.7558693065022178e-05, "loss": 1.3428, "step": 83170 }, { "epoch": 4.49, "learning_rate": 2.755598831548199e-05, "loss": 1.457, "step": 83180 }, { "epoch": 4.5, "learning_rate": 2.7553283565941797e-05, "loss": 1.3349, "step": 83190 }, { "epoch": 4.5, "learning_rate": 2.75505788164016e-05, "loss": 1.3861, "step": 83200 }, { "epoch": 4.5, "learning_rate": 2.7547874066861413e-05, "loss": 1.3534, "step": 83210 }, { "epoch": 4.5, "learning_rate": 2.754516931732122e-05, "loss": 1.3026, "step": 83220 }, { "epoch": 4.5, "learning_rate": 2.7542464567781022e-05, "loss": 1.3439, "step": 83230 }, { "epoch": 4.5, "learning_rate": 2.7539759818240835e-05, "loss": 1.39, "step": 83240 }, { "epoch": 4.5, "learning_rate": 2.753705506870064e-05, "loss": 1.3904, "step": 83250 }, { "epoch": 4.5, "learning_rate": 2.7534350319160444e-05, "loss": 1.3129, "step": 83260 }, { "epoch": 4.5, "learning_rate": 2.7531645569620257e-05, "loss": 1.3498, "step": 83270 }, { "epoch": 4.5, "learning_rate": 2.7528940820080063e-05, "loss": 1.3557, "step": 83280 }, { "epoch": 4.5, "learning_rate": 2.752623607053987e-05, "loss": 1.4496, "step": 83290 }, { "epoch": 4.5, "learning_rate": 2.752353132099968e-05, "loss": 1.4231, "step": 83300 }, { "epoch": 4.5, "learning_rate": 2.7520826571459486e-05, "loss": 1.3787, "step": 83310 }, { "epoch": 4.5, "learning_rate": 2.7518121821919292e-05, "loss": 1.3912, "step": 83320 }, { "epoch": 4.5, "learning_rate": 2.75154170723791e-05, "loss": 1.3801, "step": 83330 }, { "epoch": 4.5, "learning_rate": 2.7512712322838908e-05, "loss": 1.3968, "step": 83340 }, { "epoch": 4.5, "learning_rate": 2.7510007573298714e-05, "loss": 1.3509, "step": 83350 }, { "epoch": 4.5, "learning_rate": 2.7507302823758524e-05, "loss": 1.3842, "step": 83360 }, { "epoch": 4.5, "learning_rate": 2.750459807421833e-05, "loss": 1.3324, "step": 83370 }, { "epoch": 4.51, "learning_rate": 2.7501893324678136e-05, "loss": 1.3944, "step": 83380 }, { "epoch": 4.51, "learning_rate": 2.749918857513794e-05, "loss": 1.3561, "step": 83390 }, { "epoch": 4.51, "learning_rate": 2.7496483825597752e-05, "loss": 1.4422, "step": 83400 }, { "epoch": 4.51, "learning_rate": 2.749377907605756e-05, "loss": 1.3593, "step": 83410 }, { "epoch": 4.51, "learning_rate": 2.7491074326517365e-05, "loss": 1.3505, "step": 83420 }, { "epoch": 4.51, "learning_rate": 2.7488369576977174e-05, "loss": 1.4161, "step": 83430 }, { "epoch": 4.51, "learning_rate": 2.748566482743698e-05, "loss": 1.4087, "step": 83440 }, { "epoch": 4.51, "learning_rate": 2.7482960077896787e-05, "loss": 1.45, "step": 83450 }, { "epoch": 4.51, "learning_rate": 2.7480255328356597e-05, "loss": 1.3383, "step": 83460 }, { "epoch": 4.51, "learning_rate": 2.7477550578816403e-05, "loss": 1.4321, "step": 83470 }, { "epoch": 4.51, "learning_rate": 2.747484582927621e-05, "loss": 1.4548, "step": 83480 }, { "epoch": 4.51, "learning_rate": 2.747214107973602e-05, "loss": 1.3739, "step": 83490 }, { "epoch": 4.51, "learning_rate": 2.7469436330195825e-05, "loss": 1.3584, "step": 83500 }, { "epoch": 4.51, "learning_rate": 2.746673158065563e-05, "loss": 1.3531, "step": 83510 }, { "epoch": 4.51, "learning_rate": 2.746402683111544e-05, "loss": 1.3334, "step": 83520 }, { "epoch": 4.51, "learning_rate": 2.7461322081575247e-05, "loss": 1.37, "step": 83530 }, { "epoch": 4.51, "learning_rate": 2.7458617332035053e-05, "loss": 1.3635, "step": 83540 }, { "epoch": 4.51, "learning_rate": 2.7455912582494863e-05, "loss": 1.3133, "step": 83550 }, { "epoch": 4.52, "learning_rate": 2.745320783295467e-05, "loss": 1.3776, "step": 83560 }, { "epoch": 4.52, "learning_rate": 2.7450503083414476e-05, "loss": 1.3041, "step": 83570 }, { "epoch": 4.52, "learning_rate": 2.7447798333874285e-05, "loss": 1.4155, "step": 83580 }, { "epoch": 4.52, "learning_rate": 2.744509358433409e-05, "loss": 1.4553, "step": 83590 }, { "epoch": 4.52, "learning_rate": 2.7442388834793898e-05, "loss": 1.3066, "step": 83600 }, { "epoch": 4.52, "learning_rate": 2.7439684085253707e-05, "loss": 1.3698, "step": 83610 }, { "epoch": 4.52, "learning_rate": 2.7436979335713514e-05, "loss": 1.4048, "step": 83620 }, { "epoch": 4.52, "learning_rate": 2.743427458617332e-05, "loss": 1.4807, "step": 83630 }, { "epoch": 4.52, "learning_rate": 2.743156983663313e-05, "loss": 1.3379, "step": 83640 }, { "epoch": 4.52, "learning_rate": 2.7428865087092936e-05, "loss": 1.2494, "step": 83650 }, { "epoch": 4.52, "learning_rate": 2.7426160337552742e-05, "loss": 1.3492, "step": 83660 }, { "epoch": 4.52, "learning_rate": 2.742345558801255e-05, "loss": 1.3618, "step": 83670 }, { "epoch": 4.52, "learning_rate": 2.7420750838472358e-05, "loss": 1.4824, "step": 83680 }, { "epoch": 4.52, "learning_rate": 2.7418046088932164e-05, "loss": 1.3915, "step": 83690 }, { "epoch": 4.52, "learning_rate": 2.7415341339391977e-05, "loss": 1.3338, "step": 83700 }, { "epoch": 4.52, "learning_rate": 2.741263658985178e-05, "loss": 1.3893, "step": 83710 }, { "epoch": 4.52, "learning_rate": 2.7409931840311586e-05, "loss": 1.3914, "step": 83720 }, { "epoch": 4.52, "learning_rate": 2.74072270907714e-05, "loss": 1.3403, "step": 83730 }, { "epoch": 4.52, "learning_rate": 2.7404522341231202e-05, "loss": 1.3622, "step": 83740 }, { "epoch": 4.53, "learning_rate": 2.740181759169101e-05, "loss": 1.4038, "step": 83750 }, { "epoch": 4.53, "learning_rate": 2.739911284215082e-05, "loss": 1.397, "step": 83760 }, { "epoch": 4.53, "learning_rate": 2.7396408092610624e-05, "loss": 1.3158, "step": 83770 }, { "epoch": 4.53, "learning_rate": 2.739370334307043e-05, "loss": 1.4017, "step": 83780 }, { "epoch": 4.53, "learning_rate": 2.7390998593530244e-05, "loss": 1.2808, "step": 83790 }, { "epoch": 4.53, "learning_rate": 2.7388293843990047e-05, "loss": 1.4178, "step": 83800 }, { "epoch": 4.53, "learning_rate": 2.7385589094449853e-05, "loss": 1.3472, "step": 83810 }, { "epoch": 4.53, "learning_rate": 2.7382884344909666e-05, "loss": 1.3289, "step": 83820 }, { "epoch": 4.53, "learning_rate": 2.7380179595369472e-05, "loss": 1.3525, "step": 83830 }, { "epoch": 4.53, "learning_rate": 2.7377474845829275e-05, "loss": 1.4079, "step": 83840 }, { "epoch": 4.53, "learning_rate": 2.7374770096289088e-05, "loss": 1.2977, "step": 83850 }, { "epoch": 4.53, "learning_rate": 2.7372065346748894e-05, "loss": 1.3421, "step": 83860 }, { "epoch": 4.53, "learning_rate": 2.7369360597208697e-05, "loss": 1.3755, "step": 83870 }, { "epoch": 4.53, "learning_rate": 2.736665584766851e-05, "loss": 1.376, "step": 83880 }, { "epoch": 4.53, "learning_rate": 2.7363951098128317e-05, "loss": 1.3197, "step": 83890 }, { "epoch": 4.53, "learning_rate": 2.736124634858812e-05, "loss": 1.3524, "step": 83900 }, { "epoch": 4.53, "learning_rate": 2.7358541599047932e-05, "loss": 1.4129, "step": 83910 }, { "epoch": 4.53, "learning_rate": 2.735583684950774e-05, "loss": 1.3756, "step": 83920 }, { "epoch": 4.54, "learning_rate": 2.7353132099967545e-05, "loss": 1.3174, "step": 83930 }, { "epoch": 4.54, "learning_rate": 2.7350427350427355e-05, "loss": 1.4261, "step": 83940 }, { "epoch": 4.54, "learning_rate": 2.734772260088716e-05, "loss": 1.345, "step": 83950 }, { "epoch": 4.54, "learning_rate": 2.7345017851346967e-05, "loss": 1.3746, "step": 83960 }, { "epoch": 4.54, "learning_rate": 2.734231310180677e-05, "loss": 1.4398, "step": 83970 }, { "epoch": 4.54, "learning_rate": 2.7339608352266583e-05, "loss": 1.3429, "step": 83980 }, { "epoch": 4.54, "learning_rate": 2.733690360272639e-05, "loss": 1.4081, "step": 83990 }, { "epoch": 4.54, "learning_rate": 2.7334198853186192e-05, "loss": 1.4482, "step": 84000 }, { "epoch": 4.54, "learning_rate": 2.7331494103646005e-05, "loss": 1.393, "step": 84010 }, { "epoch": 4.54, "learning_rate": 2.732878935410581e-05, "loss": 1.4567, "step": 84020 }, { "epoch": 4.54, "learning_rate": 2.7326084604565618e-05, "loss": 1.4124, "step": 84030 }, { "epoch": 4.54, "learning_rate": 2.7323379855025427e-05, "loss": 1.4739, "step": 84040 }, { "epoch": 4.54, "learning_rate": 2.7320675105485234e-05, "loss": 1.3402, "step": 84050 }, { "epoch": 4.54, "learning_rate": 2.731797035594504e-05, "loss": 1.355, "step": 84060 }, { "epoch": 4.54, "learning_rate": 2.731526560640485e-05, "loss": 1.4174, "step": 84070 }, { "epoch": 4.54, "learning_rate": 2.7312560856864656e-05, "loss": 1.3444, "step": 84080 }, { "epoch": 4.54, "learning_rate": 2.7309856107324462e-05, "loss": 1.3379, "step": 84090 }, { "epoch": 4.54, "learning_rate": 2.7307151357784272e-05, "loss": 1.3984, "step": 84100 }, { "epoch": 4.54, "learning_rate": 2.7304446608244078e-05, "loss": 1.3296, "step": 84110 }, { "epoch": 4.55, "learning_rate": 2.7301741858703884e-05, "loss": 1.321, "step": 84120 }, { "epoch": 4.55, "learning_rate": 2.7299037109163694e-05, "loss": 1.377, "step": 84130 }, { "epoch": 4.55, "learning_rate": 2.72963323596235e-05, "loss": 1.4089, "step": 84140 }, { "epoch": 4.55, "learning_rate": 2.7293627610083306e-05, "loss": 1.3747, "step": 84150 }, { "epoch": 4.55, "learning_rate": 2.7290922860543116e-05, "loss": 1.3419, "step": 84160 }, { "epoch": 4.55, "learning_rate": 2.7288218111002922e-05, "loss": 1.3166, "step": 84170 }, { "epoch": 4.55, "learning_rate": 2.728551336146273e-05, "loss": 1.3632, "step": 84180 }, { "epoch": 4.55, "learning_rate": 2.7282808611922538e-05, "loss": 1.3986, "step": 84190 }, { "epoch": 4.55, "learning_rate": 2.7280103862382344e-05, "loss": 1.2987, "step": 84200 }, { "epoch": 4.55, "learning_rate": 2.727739911284215e-05, "loss": 1.4136, "step": 84210 }, { "epoch": 4.55, "learning_rate": 2.727469436330196e-05, "loss": 1.3846, "step": 84220 }, { "epoch": 4.55, "learning_rate": 2.7271989613761767e-05, "loss": 1.4212, "step": 84230 }, { "epoch": 4.55, "learning_rate": 2.7269284864221573e-05, "loss": 1.398, "step": 84240 }, { "epoch": 4.55, "learning_rate": 2.7266580114681383e-05, "loss": 1.4686, "step": 84250 }, { "epoch": 4.55, "learning_rate": 2.726387536514119e-05, "loss": 1.3284, "step": 84260 }, { "epoch": 4.55, "learning_rate": 2.7261170615600995e-05, "loss": 1.3338, "step": 84270 }, { "epoch": 4.55, "learning_rate": 2.7258465866060805e-05, "loss": 1.3972, "step": 84280 }, { "epoch": 4.55, "learning_rate": 2.725576111652061e-05, "loss": 1.4378, "step": 84290 }, { "epoch": 4.56, "learning_rate": 2.7253056366980417e-05, "loss": 1.3258, "step": 84300 }, { "epoch": 4.56, "learning_rate": 2.7250351617440227e-05, "loss": 1.4268, "step": 84310 }, { "epoch": 4.56, "learning_rate": 2.7247646867900033e-05, "loss": 1.4814, "step": 84320 }, { "epoch": 4.56, "learning_rate": 2.724494211835984e-05, "loss": 1.3802, "step": 84330 }, { "epoch": 4.56, "learning_rate": 2.7242237368819652e-05, "loss": 1.4436, "step": 84340 }, { "epoch": 4.56, "learning_rate": 2.7239532619279455e-05, "loss": 1.4269, "step": 84350 }, { "epoch": 4.56, "learning_rate": 2.723682786973926e-05, "loss": 1.4263, "step": 84360 }, { "epoch": 4.56, "learning_rate": 2.7234123120199075e-05, "loss": 1.4218, "step": 84370 }, { "epoch": 4.56, "learning_rate": 2.7231418370658877e-05, "loss": 1.4193, "step": 84380 }, { "epoch": 4.56, "learning_rate": 2.7228713621118684e-05, "loss": 1.3495, "step": 84390 }, { "epoch": 4.56, "learning_rate": 2.7226008871578497e-05, "loss": 1.3768, "step": 84400 }, { "epoch": 4.56, "learning_rate": 2.72233041220383e-05, "loss": 1.3792, "step": 84410 }, { "epoch": 4.56, "learning_rate": 2.7220599372498106e-05, "loss": 1.4091, "step": 84420 }, { "epoch": 4.56, "learning_rate": 2.721789462295792e-05, "loss": 1.3298, "step": 84430 }, { "epoch": 4.56, "learning_rate": 2.7215189873417722e-05, "loss": 1.4467, "step": 84440 }, { "epoch": 4.56, "learning_rate": 2.7212485123877528e-05, "loss": 1.4788, "step": 84450 }, { "epoch": 4.56, "learning_rate": 2.720978037433734e-05, "loss": 1.3957, "step": 84460 }, { "epoch": 4.56, "learning_rate": 2.7207075624797147e-05, "loss": 1.4217, "step": 84470 }, { "epoch": 4.56, "learning_rate": 2.720437087525695e-05, "loss": 1.3469, "step": 84480 }, { "epoch": 4.57, "learning_rate": 2.7201666125716763e-05, "loss": 1.3991, "step": 84490 }, { "epoch": 4.57, "learning_rate": 2.719896137617657e-05, "loss": 1.3592, "step": 84500 }, { "epoch": 4.57, "learning_rate": 2.7196256626636372e-05, "loss": 1.4076, "step": 84510 }, { "epoch": 4.57, "learning_rate": 2.719355187709618e-05, "loss": 1.399, "step": 84520 }, { "epoch": 4.57, "learning_rate": 2.7190847127555992e-05, "loss": 1.4411, "step": 84530 }, { "epoch": 4.57, "learning_rate": 2.7188142378015795e-05, "loss": 1.4177, "step": 84540 }, { "epoch": 4.57, "learning_rate": 2.71854376284756e-05, "loss": 1.3685, "step": 84550 }, { "epoch": 4.57, "learning_rate": 2.7182732878935414e-05, "loss": 1.5369, "step": 84560 }, { "epoch": 4.57, "learning_rate": 2.718002812939522e-05, "loss": 1.3907, "step": 84570 }, { "epoch": 4.57, "learning_rate": 2.7177323379855023e-05, "loss": 1.3792, "step": 84580 }, { "epoch": 4.57, "learning_rate": 2.7174618630314836e-05, "loss": 1.3374, "step": 84590 }, { "epoch": 4.57, "learning_rate": 2.7171913880774642e-05, "loss": 1.4052, "step": 84600 }, { "epoch": 4.57, "learning_rate": 2.7169209131234445e-05, "loss": 1.4106, "step": 84610 }, { "epoch": 4.57, "learning_rate": 2.7166504381694258e-05, "loss": 1.3264, "step": 84620 }, { "epoch": 4.57, "learning_rate": 2.7163799632154065e-05, "loss": 1.4043, "step": 84630 }, { "epoch": 4.57, "learning_rate": 2.7161094882613867e-05, "loss": 1.4448, "step": 84640 }, { "epoch": 4.57, "learning_rate": 2.715839013307368e-05, "loss": 1.359, "step": 84650 }, { "epoch": 4.57, "learning_rate": 2.7155685383533487e-05, "loss": 1.3319, "step": 84660 }, { "epoch": 4.58, "learning_rate": 2.7152980633993293e-05, "loss": 1.3807, "step": 84670 }, { "epoch": 4.58, "learning_rate": 2.7150275884453103e-05, "loss": 1.3443, "step": 84680 }, { "epoch": 4.58, "learning_rate": 2.714757113491291e-05, "loss": 1.3094, "step": 84690 }, { "epoch": 4.58, "learning_rate": 2.7144866385372715e-05, "loss": 1.3965, "step": 84700 }, { "epoch": 4.58, "learning_rate": 2.7142161635832525e-05, "loss": 1.3747, "step": 84710 }, { "epoch": 4.58, "learning_rate": 2.713945688629233e-05, "loss": 1.4373, "step": 84720 }, { "epoch": 4.58, "learning_rate": 2.7136752136752137e-05, "loss": 1.4087, "step": 84730 }, { "epoch": 4.58, "learning_rate": 2.7134047387211947e-05, "loss": 1.4165, "step": 84740 }, { "epoch": 4.58, "learning_rate": 2.7131342637671753e-05, "loss": 1.3572, "step": 84750 }, { "epoch": 4.58, "learning_rate": 2.712863788813156e-05, "loss": 1.4052, "step": 84760 }, { "epoch": 4.58, "learning_rate": 2.712593313859137e-05, "loss": 1.4172, "step": 84770 }, { "epoch": 4.58, "learning_rate": 2.7123228389051175e-05, "loss": 1.3789, "step": 84780 }, { "epoch": 4.58, "learning_rate": 2.712052363951098e-05, "loss": 1.3513, "step": 84790 }, { "epoch": 4.58, "learning_rate": 2.711781888997079e-05, "loss": 1.3886, "step": 84800 }, { "epoch": 4.58, "learning_rate": 2.7115114140430598e-05, "loss": 1.4202, "step": 84810 }, { "epoch": 4.58, "learning_rate": 2.7112409390890404e-05, "loss": 1.3669, "step": 84820 }, { "epoch": 4.58, "learning_rate": 2.7109704641350213e-05, "loss": 1.3585, "step": 84830 }, { "epoch": 4.58, "learning_rate": 2.710699989181002e-05, "loss": 1.3972, "step": 84840 }, { "epoch": 4.58, "learning_rate": 2.7104295142269826e-05, "loss": 1.3065, "step": 84850 }, { "epoch": 4.59, "learning_rate": 2.7101590392729636e-05, "loss": 1.3894, "step": 84860 }, { "epoch": 4.59, "learning_rate": 2.7098885643189442e-05, "loss": 1.3658, "step": 84870 }, { "epoch": 4.59, "learning_rate": 2.7096180893649248e-05, "loss": 1.4706, "step": 84880 }, { "epoch": 4.59, "learning_rate": 2.7093476144109058e-05, "loss": 1.416, "step": 84890 }, { "epoch": 4.59, "learning_rate": 2.7090771394568864e-05, "loss": 1.3887, "step": 84900 }, { "epoch": 4.59, "learning_rate": 2.708806664502867e-05, "loss": 1.3827, "step": 84910 }, { "epoch": 4.59, "learning_rate": 2.708536189548848e-05, "loss": 1.3734, "step": 84920 }, { "epoch": 4.59, "learning_rate": 2.7082657145948286e-05, "loss": 1.3464, "step": 84930 }, { "epoch": 4.59, "learning_rate": 2.7079952396408092e-05, "loss": 1.3742, "step": 84940 }, { "epoch": 4.59, "learning_rate": 2.7077247646867902e-05, "loss": 1.3607, "step": 84950 }, { "epoch": 4.59, "learning_rate": 2.707454289732771e-05, "loss": 1.4401, "step": 84960 }, { "epoch": 4.59, "learning_rate": 2.7071838147787515e-05, "loss": 1.364, "step": 84970 }, { "epoch": 4.59, "learning_rate": 2.7069133398247328e-05, "loss": 1.3989, "step": 84980 }, { "epoch": 4.59, "learning_rate": 2.706642864870713e-05, "loss": 1.3478, "step": 84990 }, { "epoch": 4.59, "learning_rate": 2.7063723899166937e-05, "loss": 1.3332, "step": 85000 }, { "epoch": 4.59, "learning_rate": 2.706101914962675e-05, "loss": 1.5115, "step": 85010 }, { "epoch": 4.59, "learning_rate": 2.7058314400086553e-05, "loss": 1.374, "step": 85020 }, { "epoch": 4.59, "learning_rate": 2.705560965054636e-05, "loss": 1.4118, "step": 85030 }, { "epoch": 4.6, "learning_rate": 2.7052904901006172e-05, "loss": 1.3899, "step": 85040 }, { "epoch": 4.6, "learning_rate": 2.7050200151465975e-05, "loss": 1.3651, "step": 85050 }, { "epoch": 4.6, "learning_rate": 2.704749540192578e-05, "loss": 1.3305, "step": 85060 }, { "epoch": 4.6, "learning_rate": 2.7044790652385587e-05, "loss": 1.4083, "step": 85070 }, { "epoch": 4.6, "learning_rate": 2.70420859028454e-05, "loss": 1.4037, "step": 85080 }, { "epoch": 4.6, "learning_rate": 2.7039381153305203e-05, "loss": 1.3641, "step": 85090 }, { "epoch": 4.6, "learning_rate": 2.703667640376501e-05, "loss": 1.3372, "step": 85100 }, { "epoch": 4.6, "learning_rate": 2.7033971654224823e-05, "loss": 1.4016, "step": 85110 }, { "epoch": 4.6, "learning_rate": 2.7031266904684625e-05, "loss": 1.3652, "step": 85120 }, { "epoch": 4.6, "learning_rate": 2.7028562155144432e-05, "loss": 1.3087, "step": 85130 }, { "epoch": 4.6, "learning_rate": 2.7025857405604245e-05, "loss": 1.3347, "step": 85140 }, { "epoch": 4.6, "learning_rate": 2.7023152656064048e-05, "loss": 1.3704, "step": 85150 }, { "epoch": 4.6, "learning_rate": 2.7020447906523854e-05, "loss": 1.3866, "step": 85160 }, { "epoch": 4.6, "learning_rate": 2.7017743156983667e-05, "loss": 1.3441, "step": 85170 }, { "epoch": 4.6, "learning_rate": 2.701503840744347e-05, "loss": 1.4077, "step": 85180 }, { "epoch": 4.6, "learning_rate": 2.7012333657903276e-05, "loss": 1.4051, "step": 85190 }, { "epoch": 4.6, "learning_rate": 2.700962890836309e-05, "loss": 1.4042, "step": 85200 }, { "epoch": 4.6, "learning_rate": 2.7006924158822895e-05, "loss": 1.3811, "step": 85210 }, { "epoch": 4.6, "learning_rate": 2.7004219409282698e-05, "loss": 1.4342, "step": 85220 }, { "epoch": 4.61, "learning_rate": 2.700151465974251e-05, "loss": 1.3817, "step": 85230 }, { "epoch": 4.61, "learning_rate": 2.6998809910202318e-05, "loss": 1.3773, "step": 85240 }, { "epoch": 4.61, "learning_rate": 2.699610516066212e-05, "loss": 1.3874, "step": 85250 }, { "epoch": 4.61, "learning_rate": 2.6993400411121933e-05, "loss": 1.3826, "step": 85260 }, { "epoch": 4.61, "learning_rate": 2.699069566158174e-05, "loss": 1.3724, "step": 85270 }, { "epoch": 4.61, "learning_rate": 2.6987990912041543e-05, "loss": 1.3584, "step": 85280 }, { "epoch": 4.61, "learning_rate": 2.6985286162501356e-05, "loss": 1.3736, "step": 85290 }, { "epoch": 4.61, "learning_rate": 2.6982581412961162e-05, "loss": 1.4726, "step": 85300 }, { "epoch": 4.61, "learning_rate": 2.6979876663420968e-05, "loss": 1.3555, "step": 85310 }, { "epoch": 4.61, "learning_rate": 2.6977171913880778e-05, "loss": 1.3704, "step": 85320 }, { "epoch": 4.61, "learning_rate": 2.6974467164340584e-05, "loss": 1.3768, "step": 85330 }, { "epoch": 4.61, "learning_rate": 2.697176241480039e-05, "loss": 1.3899, "step": 85340 }, { "epoch": 4.61, "learning_rate": 2.69690576652602e-05, "loss": 1.4015, "step": 85350 }, { "epoch": 4.61, "learning_rate": 2.6966352915720006e-05, "loss": 1.3628, "step": 85360 }, { "epoch": 4.61, "learning_rate": 2.6963648166179812e-05, "loss": 1.3786, "step": 85370 }, { "epoch": 4.61, "learning_rate": 2.6960943416639622e-05, "loss": 1.3268, "step": 85380 }, { "epoch": 4.61, "learning_rate": 2.695823866709943e-05, "loss": 1.3324, "step": 85390 }, { "epoch": 4.61, "learning_rate": 2.6955533917559235e-05, "loss": 1.4362, "step": 85400 }, { "epoch": 4.62, "learning_rate": 2.6952829168019044e-05, "loss": 1.3181, "step": 85410 }, { "epoch": 4.62, "learning_rate": 2.695012441847885e-05, "loss": 1.4036, "step": 85420 }, { "epoch": 4.62, "learning_rate": 2.6947419668938657e-05, "loss": 1.4789, "step": 85430 }, { "epoch": 4.62, "learning_rate": 2.6944714919398466e-05, "loss": 1.4329, "step": 85440 }, { "epoch": 4.62, "learning_rate": 2.6942010169858273e-05, "loss": 1.4114, "step": 85450 }, { "epoch": 4.62, "learning_rate": 2.693930542031808e-05, "loss": 1.3604, "step": 85460 }, { "epoch": 4.62, "learning_rate": 2.693660067077789e-05, "loss": 1.4612, "step": 85470 }, { "epoch": 4.62, "learning_rate": 2.6933895921237695e-05, "loss": 1.4485, "step": 85480 }, { "epoch": 4.62, "learning_rate": 2.69311911716975e-05, "loss": 1.3642, "step": 85490 }, { "epoch": 4.62, "learning_rate": 2.692848642215731e-05, "loss": 1.3936, "step": 85500 }, { "epoch": 4.62, "learning_rate": 2.6925781672617117e-05, "loss": 1.4103, "step": 85510 }, { "epoch": 4.62, "learning_rate": 2.6923076923076923e-05, "loss": 1.2921, "step": 85520 }, { "epoch": 4.62, "learning_rate": 2.6920372173536733e-05, "loss": 1.3599, "step": 85530 }, { "epoch": 4.62, "learning_rate": 2.691766742399654e-05, "loss": 1.3646, "step": 85540 }, { "epoch": 4.62, "learning_rate": 2.6914962674456345e-05, "loss": 1.4089, "step": 85550 }, { "epoch": 4.62, "learning_rate": 2.6912257924916155e-05, "loss": 1.4013, "step": 85560 }, { "epoch": 4.62, "learning_rate": 2.690955317537596e-05, "loss": 1.3045, "step": 85570 }, { "epoch": 4.62, "learning_rate": 2.6906848425835768e-05, "loss": 1.4478, "step": 85580 }, { "epoch": 4.62, "learning_rate": 2.6904143676295577e-05, "loss": 1.4301, "step": 85590 }, { "epoch": 4.63, "learning_rate": 2.6901438926755384e-05, "loss": 1.3472, "step": 85600 }, { "epoch": 4.63, "learning_rate": 2.689873417721519e-05, "loss": 1.4546, "step": 85610 }, { "epoch": 4.63, "learning_rate": 2.6896029427675003e-05, "loss": 1.3611, "step": 85620 }, { "epoch": 4.63, "learning_rate": 2.6893324678134806e-05, "loss": 1.3729, "step": 85630 }, { "epoch": 4.63, "learning_rate": 2.6890619928594612e-05, "loss": 1.3632, "step": 85640 }, { "epoch": 4.63, "learning_rate": 2.6887915179054418e-05, "loss": 1.3831, "step": 85650 }, { "epoch": 4.63, "learning_rate": 2.6885210429514228e-05, "loss": 1.4235, "step": 85660 }, { "epoch": 4.63, "learning_rate": 2.6882505679974034e-05, "loss": 1.3933, "step": 85670 }, { "epoch": 4.63, "learning_rate": 2.687980093043384e-05, "loss": 1.4039, "step": 85680 }, { "epoch": 4.63, "learning_rate": 2.687709618089365e-05, "loss": 1.3606, "step": 85690 }, { "epoch": 4.63, "learning_rate": 2.6874391431353456e-05, "loss": 1.3695, "step": 85700 }, { "epoch": 4.63, "learning_rate": 2.6871686681813263e-05, "loss": 1.4233, "step": 85710 }, { "epoch": 4.63, "learning_rate": 2.6868981932273076e-05, "loss": 1.3859, "step": 85720 }, { "epoch": 4.63, "learning_rate": 2.686627718273288e-05, "loss": 1.4081, "step": 85730 }, { "epoch": 4.63, "learning_rate": 2.6863572433192685e-05, "loss": 1.3333, "step": 85740 }, { "epoch": 4.63, "learning_rate": 2.6860867683652498e-05, "loss": 1.3764, "step": 85750 }, { "epoch": 4.63, "learning_rate": 2.68581629341123e-05, "loss": 1.3228, "step": 85760 }, { "epoch": 4.63, "learning_rate": 2.6855458184572107e-05, "loss": 1.4086, "step": 85770 }, { "epoch": 4.64, "learning_rate": 2.685275343503192e-05, "loss": 1.4066, "step": 85780 }, { "epoch": 4.64, "learning_rate": 2.6850048685491723e-05, "loss": 1.3864, "step": 85790 }, { "epoch": 4.64, "learning_rate": 2.684734393595153e-05, "loss": 1.3746, "step": 85800 }, { "epoch": 4.64, "learning_rate": 2.6844639186411342e-05, "loss": 1.3685, "step": 85810 }, { "epoch": 4.64, "learning_rate": 2.684193443687115e-05, "loss": 1.4114, "step": 85820 }, { "epoch": 4.64, "learning_rate": 2.683922968733095e-05, "loss": 1.3341, "step": 85830 }, { "epoch": 4.64, "learning_rate": 2.6836524937790764e-05, "loss": 1.3669, "step": 85840 }, { "epoch": 4.64, "learning_rate": 2.683382018825057e-05, "loss": 1.4191, "step": 85850 }, { "epoch": 4.64, "learning_rate": 2.6831115438710373e-05, "loss": 1.3984, "step": 85860 }, { "epoch": 4.64, "learning_rate": 2.6828410689170186e-05, "loss": 1.274, "step": 85870 }, { "epoch": 4.64, "learning_rate": 2.6825705939629993e-05, "loss": 1.3989, "step": 85880 }, { "epoch": 4.64, "learning_rate": 2.6823001190089796e-05, "loss": 1.3728, "step": 85890 }, { "epoch": 4.64, "learning_rate": 2.682029644054961e-05, "loss": 1.4042, "step": 85900 }, { "epoch": 4.64, "learning_rate": 2.6817591691009415e-05, "loss": 1.2852, "step": 85910 }, { "epoch": 4.64, "learning_rate": 2.6814886941469218e-05, "loss": 1.3676, "step": 85920 }, { "epoch": 4.64, "learning_rate": 2.681218219192903e-05, "loss": 1.4235, "step": 85930 }, { "epoch": 4.64, "learning_rate": 2.6809477442388837e-05, "loss": 1.3418, "step": 85940 }, { "epoch": 4.64, "learning_rate": 2.6806772692848643e-05, "loss": 1.3986, "step": 85950 }, { "epoch": 4.64, "learning_rate": 2.6804067943308453e-05, "loss": 1.3695, "step": 85960 }, { "epoch": 4.65, "learning_rate": 2.680136319376826e-05, "loss": 1.3928, "step": 85970 }, { "epoch": 4.65, "learning_rate": 2.6798658444228066e-05, "loss": 1.4186, "step": 85980 }, { "epoch": 4.65, "learning_rate": 2.6795953694687875e-05, "loss": 1.4183, "step": 85990 }, { "epoch": 4.65, "learning_rate": 2.679324894514768e-05, "loss": 1.4032, "step": 86000 }, { "epoch": 4.65, "learning_rate": 2.6790544195607488e-05, "loss": 1.3842, "step": 86010 }, { "epoch": 4.65, "learning_rate": 2.6787839446067297e-05, "loss": 1.4168, "step": 86020 }, { "epoch": 4.65, "learning_rate": 2.6785134696527104e-05, "loss": 1.3997, "step": 86030 }, { "epoch": 4.65, "learning_rate": 2.678242994698691e-05, "loss": 1.3191, "step": 86040 }, { "epoch": 4.65, "learning_rate": 2.677972519744672e-05, "loss": 1.4244, "step": 86050 }, { "epoch": 4.65, "learning_rate": 2.6777020447906526e-05, "loss": 1.3652, "step": 86060 }, { "epoch": 4.65, "learning_rate": 2.6774315698366332e-05, "loss": 1.3662, "step": 86070 }, { "epoch": 4.65, "learning_rate": 2.677161094882614e-05, "loss": 1.3842, "step": 86080 }, { "epoch": 4.65, "learning_rate": 2.6768906199285948e-05, "loss": 1.3652, "step": 86090 }, { "epoch": 4.65, "learning_rate": 2.6766201449745754e-05, "loss": 1.3805, "step": 86100 }, { "epoch": 4.65, "learning_rate": 2.6763496700205564e-05, "loss": 1.3573, "step": 86110 }, { "epoch": 4.65, "learning_rate": 2.676079195066537e-05, "loss": 1.3875, "step": 86120 }, { "epoch": 4.65, "learning_rate": 2.6758087201125176e-05, "loss": 1.3962, "step": 86130 }, { "epoch": 4.65, "learning_rate": 2.6755382451584986e-05, "loss": 1.342, "step": 86140 }, { "epoch": 4.66, "learning_rate": 2.6752677702044792e-05, "loss": 1.2865, "step": 86150 }, { "epoch": 4.66, "learning_rate": 2.67499729525046e-05, "loss": 1.3282, "step": 86160 }, { "epoch": 4.66, "learning_rate": 2.6747268202964408e-05, "loss": 1.4249, "step": 86170 }, { "epoch": 4.66, "learning_rate": 2.6744563453424214e-05, "loss": 1.344, "step": 86180 }, { "epoch": 4.66, "learning_rate": 2.674185870388402e-05, "loss": 1.3621, "step": 86190 }, { "epoch": 4.66, "learning_rate": 2.6739153954343827e-05, "loss": 1.3715, "step": 86200 }, { "epoch": 4.66, "learning_rate": 2.6736449204803637e-05, "loss": 1.3226, "step": 86210 }, { "epoch": 4.66, "learning_rate": 2.6733744455263443e-05, "loss": 1.44, "step": 86220 }, { "epoch": 4.66, "learning_rate": 2.673103970572325e-05, "loss": 1.3573, "step": 86230 }, { "epoch": 4.66, "learning_rate": 2.672833495618306e-05, "loss": 1.3787, "step": 86240 }, { "epoch": 4.66, "learning_rate": 2.6725630206642865e-05, "loss": 1.4078, "step": 86250 }, { "epoch": 4.66, "learning_rate": 2.672292545710267e-05, "loss": 1.3619, "step": 86260 }, { "epoch": 4.66, "learning_rate": 2.672022070756248e-05, "loss": 1.3807, "step": 86270 }, { "epoch": 4.66, "learning_rate": 2.6717515958022287e-05, "loss": 1.3553, "step": 86280 }, { "epoch": 4.66, "learning_rate": 2.6714811208482093e-05, "loss": 1.3301, "step": 86290 }, { "epoch": 4.66, "learning_rate": 2.6712106458941903e-05, "loss": 1.3428, "step": 86300 }, { "epoch": 4.66, "learning_rate": 2.670940170940171e-05, "loss": 1.3378, "step": 86310 }, { "epoch": 4.66, "learning_rate": 2.6706696959861516e-05, "loss": 1.4063, "step": 86320 }, { "epoch": 4.66, "learning_rate": 2.6703992210321325e-05, "loss": 1.39, "step": 86330 }, { "epoch": 4.67, "learning_rate": 2.670128746078113e-05, "loss": 1.4644, "step": 86340 }, { "epoch": 4.67, "learning_rate": 2.6698582711240938e-05, "loss": 1.4407, "step": 86350 }, { "epoch": 4.67, "learning_rate": 2.669587796170075e-05, "loss": 1.3206, "step": 86360 }, { "epoch": 4.67, "learning_rate": 2.6693173212160554e-05, "loss": 1.3897, "step": 86370 }, { "epoch": 4.67, "learning_rate": 2.669046846262036e-05, "loss": 1.3376, "step": 86380 }, { "epoch": 4.67, "learning_rate": 2.6687763713080173e-05, "loss": 1.4291, "step": 86390 }, { "epoch": 4.67, "learning_rate": 2.6685058963539976e-05, "loss": 1.3843, "step": 86400 }, { "epoch": 4.67, "learning_rate": 2.6682354213999782e-05, "loss": 1.3904, "step": 86410 }, { "epoch": 4.67, "learning_rate": 2.6679649464459595e-05, "loss": 1.373, "step": 86420 }, { "epoch": 4.67, "learning_rate": 2.6676944714919398e-05, "loss": 1.3881, "step": 86430 }, { "epoch": 4.67, "learning_rate": 2.6674239965379204e-05, "loss": 1.4045, "step": 86440 }, { "epoch": 4.67, "learning_rate": 2.6671535215839017e-05, "loss": 1.4617, "step": 86450 }, { "epoch": 4.67, "learning_rate": 2.6668830466298824e-05, "loss": 1.4173, "step": 86460 }, { "epoch": 4.67, "learning_rate": 2.6666125716758626e-05, "loss": 1.3727, "step": 86470 }, { "epoch": 4.67, "learning_rate": 2.666342096721844e-05, "loss": 1.327, "step": 86480 }, { "epoch": 4.67, "learning_rate": 2.6660716217678246e-05, "loss": 1.4218, "step": 86490 }, { "epoch": 4.67, "learning_rate": 2.665801146813805e-05, "loss": 1.3566, "step": 86500 }, { "epoch": 4.67, "learning_rate": 2.665530671859786e-05, "loss": 1.3885, "step": 86510 }, { "epoch": 4.68, "learning_rate": 2.6652601969057668e-05, "loss": 1.3974, "step": 86520 }, { "epoch": 4.68, "learning_rate": 2.664989721951747e-05, "loss": 1.4042, "step": 86530 }, { "epoch": 4.68, "learning_rate": 2.6647192469977284e-05, "loss": 1.3528, "step": 86540 }, { "epoch": 4.68, "learning_rate": 2.664448772043709e-05, "loss": 1.4181, "step": 86550 }, { "epoch": 4.68, "learning_rate": 2.6641782970896893e-05, "loss": 1.3385, "step": 86560 }, { "epoch": 4.68, "learning_rate": 2.6639078221356706e-05, "loss": 1.4244, "step": 86570 }, { "epoch": 4.68, "learning_rate": 2.6636373471816512e-05, "loss": 1.385, "step": 86580 }, { "epoch": 4.68, "learning_rate": 2.663366872227632e-05, "loss": 1.3965, "step": 86590 }, { "epoch": 4.68, "learning_rate": 2.6630963972736128e-05, "loss": 1.3862, "step": 86600 }, { "epoch": 4.68, "learning_rate": 2.6628259223195934e-05, "loss": 1.4889, "step": 86610 }, { "epoch": 4.68, "learning_rate": 2.662555447365574e-05, "loss": 1.3501, "step": 86620 }, { "epoch": 4.68, "learning_rate": 2.662284972411555e-05, "loss": 1.3692, "step": 86630 }, { "epoch": 4.68, "learning_rate": 2.6620144974575357e-05, "loss": 1.3644, "step": 86640 }, { "epoch": 4.68, "learning_rate": 2.6617440225035163e-05, "loss": 1.3513, "step": 86650 }, { "epoch": 4.68, "learning_rate": 2.6614735475494973e-05, "loss": 1.4144, "step": 86660 }, { "epoch": 4.68, "learning_rate": 2.661203072595478e-05, "loss": 1.3723, "step": 86670 }, { "epoch": 4.68, "learning_rate": 2.6609325976414585e-05, "loss": 1.4354, "step": 86680 }, { "epoch": 4.68, "learning_rate": 2.6606621226874395e-05, "loss": 1.3538, "step": 86690 }, { "epoch": 4.68, "learning_rate": 2.66039164773342e-05, "loss": 1.3386, "step": 86700 }, { "epoch": 4.69, "learning_rate": 2.6601211727794007e-05, "loss": 1.3384, "step": 86710 }, { "epoch": 4.69, "learning_rate": 2.6598506978253817e-05, "loss": 1.3708, "step": 86720 }, { "epoch": 4.69, "learning_rate": 2.6595802228713623e-05, "loss": 1.3558, "step": 86730 }, { "epoch": 4.69, "learning_rate": 2.659309747917343e-05, "loss": 1.4016, "step": 86740 }, { "epoch": 4.69, "learning_rate": 2.6590392729633236e-05, "loss": 1.3553, "step": 86750 }, { "epoch": 4.69, "learning_rate": 2.6587687980093045e-05, "loss": 1.3506, "step": 86760 }, { "epoch": 4.69, "learning_rate": 2.658498323055285e-05, "loss": 1.3985, "step": 86770 }, { "epoch": 4.69, "learning_rate": 2.6582278481012658e-05, "loss": 1.3287, "step": 86780 }, { "epoch": 4.69, "learning_rate": 2.6579573731472467e-05, "loss": 1.405, "step": 86790 }, { "epoch": 4.69, "learning_rate": 2.6576868981932274e-05, "loss": 1.4305, "step": 86800 }, { "epoch": 4.69, "learning_rate": 2.657416423239208e-05, "loss": 1.3632, "step": 86810 }, { "epoch": 4.69, "learning_rate": 2.657145948285189e-05, "loss": 1.4036, "step": 86820 }, { "epoch": 4.69, "learning_rate": 2.6568754733311696e-05, "loss": 1.4132, "step": 86830 }, { "epoch": 4.69, "learning_rate": 2.6566049983771502e-05, "loss": 1.3812, "step": 86840 }, { "epoch": 4.69, "learning_rate": 2.6563345234231312e-05, "loss": 1.3796, "step": 86850 }, { "epoch": 4.69, "learning_rate": 2.6560640484691118e-05, "loss": 1.365, "step": 86860 }, { "epoch": 4.69, "learning_rate": 2.6557935735150924e-05, "loss": 1.3671, "step": 86870 }, { "epoch": 4.69, "learning_rate": 2.6555230985610734e-05, "loss": 1.3562, "step": 86880 }, { "epoch": 4.7, "learning_rate": 2.655252623607054e-05, "loss": 1.3323, "step": 86890 }, { "epoch": 4.7, "learning_rate": 2.6549821486530347e-05, "loss": 1.3913, "step": 86900 }, { "epoch": 4.7, "learning_rate": 2.6547116736990156e-05, "loss": 1.3473, "step": 86910 }, { "epoch": 4.7, "learning_rate": 2.6544411987449962e-05, "loss": 1.4142, "step": 86920 }, { "epoch": 4.7, "learning_rate": 2.654170723790977e-05, "loss": 1.3439, "step": 86930 }, { "epoch": 4.7, "learning_rate": 2.653900248836958e-05, "loss": 1.308, "step": 86940 }, { "epoch": 4.7, "learning_rate": 2.6536297738829385e-05, "loss": 1.4084, "step": 86950 }, { "epoch": 4.7, "learning_rate": 2.653359298928919e-05, "loss": 1.3028, "step": 86960 }, { "epoch": 4.7, "learning_rate": 2.6530888239749e-05, "loss": 1.412, "step": 86970 }, { "epoch": 4.7, "learning_rate": 2.6528183490208807e-05, "loss": 1.3748, "step": 86980 }, { "epoch": 4.7, "learning_rate": 2.6525478740668613e-05, "loss": 1.4314, "step": 86990 }, { "epoch": 4.7, "learning_rate": 2.6522773991128426e-05, "loss": 1.3936, "step": 87000 }, { "epoch": 4.7, "learning_rate": 2.652006924158823e-05, "loss": 1.3677, "step": 87010 }, { "epoch": 4.7, "learning_rate": 2.6517364492048035e-05, "loss": 1.3288, "step": 87020 }, { "epoch": 4.7, "learning_rate": 2.6514659742507848e-05, "loss": 1.3867, "step": 87030 }, { "epoch": 4.7, "learning_rate": 2.651195499296765e-05, "loss": 1.3797, "step": 87040 }, { "epoch": 4.7, "learning_rate": 2.6509250243427457e-05, "loss": 1.4082, "step": 87050 }, { "epoch": 4.7, "learning_rate": 2.650654549388727e-05, "loss": 1.3913, "step": 87060 }, { "epoch": 4.7, "learning_rate": 2.6503840744347073e-05, "loss": 1.309, "step": 87070 }, { "epoch": 4.71, "learning_rate": 2.650113599480688e-05, "loss": 1.391, "step": 87080 }, { "epoch": 4.71, "learning_rate": 2.6498431245266693e-05, "loss": 1.352, "step": 87090 }, { "epoch": 4.71, "learning_rate": 2.64957264957265e-05, "loss": 1.3627, "step": 87100 }, { "epoch": 4.71, "learning_rate": 2.64930217461863e-05, "loss": 1.4099, "step": 87110 }, { "epoch": 4.71, "learning_rate": 2.6490316996646115e-05, "loss": 1.4136, "step": 87120 }, { "epoch": 4.71, "learning_rate": 2.648761224710592e-05, "loss": 1.3471, "step": 87130 }, { "epoch": 4.71, "learning_rate": 2.6484907497565724e-05, "loss": 1.3923, "step": 87140 }, { "epoch": 4.71, "learning_rate": 2.6482202748025537e-05, "loss": 1.3859, "step": 87150 }, { "epoch": 4.71, "learning_rate": 2.6479497998485343e-05, "loss": 1.3971, "step": 87160 }, { "epoch": 4.71, "learning_rate": 2.6476793248945146e-05, "loss": 1.3915, "step": 87170 }, { "epoch": 4.71, "learning_rate": 2.647408849940496e-05, "loss": 1.2924, "step": 87180 }, { "epoch": 4.71, "learning_rate": 2.6471383749864765e-05, "loss": 1.3721, "step": 87190 }, { "epoch": 4.71, "learning_rate": 2.646867900032457e-05, "loss": 1.3577, "step": 87200 }, { "epoch": 4.71, "learning_rate": 2.646597425078438e-05, "loss": 1.4409, "step": 87210 }, { "epoch": 4.71, "learning_rate": 2.6463269501244187e-05, "loss": 1.3603, "step": 87220 }, { "epoch": 4.71, "learning_rate": 2.6460564751703994e-05, "loss": 1.419, "step": 87230 }, { "epoch": 4.71, "learning_rate": 2.6457860002163803e-05, "loss": 1.3894, "step": 87240 }, { "epoch": 4.71, "learning_rate": 2.645515525262361e-05, "loss": 1.3037, "step": 87250 }, { "epoch": 4.72, "learning_rate": 2.6452450503083416e-05, "loss": 1.4556, "step": 87260 }, { "epoch": 4.72, "learning_rate": 2.6449745753543226e-05, "loss": 1.4185, "step": 87270 }, { "epoch": 4.72, "learning_rate": 2.6447041004003032e-05, "loss": 1.3879, "step": 87280 }, { "epoch": 4.72, "learning_rate": 2.6444336254462838e-05, "loss": 1.418, "step": 87290 }, { "epoch": 4.72, "learning_rate": 2.6441631504922648e-05, "loss": 1.3874, "step": 87300 }, { "epoch": 4.72, "learning_rate": 2.6438926755382454e-05, "loss": 1.4108, "step": 87310 }, { "epoch": 4.72, "learning_rate": 2.643622200584226e-05, "loss": 1.3965, "step": 87320 }, { "epoch": 4.72, "learning_rate": 2.6433517256302067e-05, "loss": 1.4253, "step": 87330 }, { "epoch": 4.72, "learning_rate": 2.6430812506761876e-05, "loss": 1.4343, "step": 87340 }, { "epoch": 4.72, "learning_rate": 2.6428107757221682e-05, "loss": 1.4071, "step": 87350 }, { "epoch": 4.72, "learning_rate": 2.642540300768149e-05, "loss": 1.4438, "step": 87360 }, { "epoch": 4.72, "learning_rate": 2.64226982581413e-05, "loss": 1.443, "step": 87370 }, { "epoch": 4.72, "learning_rate": 2.6419993508601105e-05, "loss": 1.4076, "step": 87380 }, { "epoch": 4.72, "learning_rate": 2.641728875906091e-05, "loss": 1.4578, "step": 87390 }, { "epoch": 4.72, "learning_rate": 2.641458400952072e-05, "loss": 1.2766, "step": 87400 }, { "epoch": 4.72, "learning_rate": 2.6411879259980527e-05, "loss": 1.3125, "step": 87410 }, { "epoch": 4.72, "learning_rate": 2.6409174510440333e-05, "loss": 1.3617, "step": 87420 }, { "epoch": 4.72, "learning_rate": 2.6406469760900143e-05, "loss": 1.4309, "step": 87430 }, { "epoch": 4.72, "learning_rate": 2.640376501135995e-05, "loss": 1.357, "step": 87440 }, { "epoch": 4.73, "learning_rate": 2.6401060261819755e-05, "loss": 1.5217, "step": 87450 }, { "epoch": 4.73, "learning_rate": 2.6398355512279565e-05, "loss": 1.3959, "step": 87460 }, { "epoch": 4.73, "learning_rate": 2.639565076273937e-05, "loss": 1.3307, "step": 87470 }, { "epoch": 4.73, "learning_rate": 2.6392946013199177e-05, "loss": 1.3944, "step": 87480 }, { "epoch": 4.73, "learning_rate": 2.6390241263658987e-05, "loss": 1.3582, "step": 87490 }, { "epoch": 4.73, "learning_rate": 2.6387536514118793e-05, "loss": 1.3461, "step": 87500 }, { "epoch": 4.73, "learning_rate": 2.63848317645786e-05, "loss": 1.3829, "step": 87510 }, { "epoch": 4.73, "learning_rate": 2.638212701503841e-05, "loss": 1.3507, "step": 87520 }, { "epoch": 4.73, "learning_rate": 2.6379422265498215e-05, "loss": 1.3438, "step": 87530 }, { "epoch": 4.73, "learning_rate": 2.6376717515958022e-05, "loss": 1.3313, "step": 87540 }, { "epoch": 4.73, "learning_rate": 2.637401276641783e-05, "loss": 1.3917, "step": 87550 }, { "epoch": 4.73, "learning_rate": 2.6371308016877638e-05, "loss": 1.4285, "step": 87560 }, { "epoch": 4.73, "learning_rate": 2.6368603267337444e-05, "loss": 1.3613, "step": 87570 }, { "epoch": 4.73, "learning_rate": 2.6365898517797254e-05, "loss": 1.3815, "step": 87580 }, { "epoch": 4.73, "learning_rate": 2.636319376825706e-05, "loss": 1.331, "step": 87590 }, { "epoch": 4.73, "learning_rate": 2.6360489018716866e-05, "loss": 1.3407, "step": 87600 }, { "epoch": 4.73, "learning_rate": 2.635778426917668e-05, "loss": 1.3762, "step": 87610 }, { "epoch": 4.73, "learning_rate": 2.6355079519636482e-05, "loss": 1.315, "step": 87620 }, { "epoch": 4.74, "learning_rate": 2.6352374770096288e-05, "loss": 1.429, "step": 87630 }, { "epoch": 4.74, "learning_rate": 2.63496700205561e-05, "loss": 1.3684, "step": 87640 }, { "epoch": 4.74, "learning_rate": 2.6346965271015904e-05, "loss": 1.4265, "step": 87650 }, { "epoch": 4.74, "learning_rate": 2.634426052147571e-05, "loss": 1.2843, "step": 87660 }, { "epoch": 4.74, "learning_rate": 2.6341555771935523e-05, "loss": 1.3786, "step": 87670 }, { "epoch": 4.74, "learning_rate": 2.6338851022395326e-05, "loss": 1.4373, "step": 87680 }, { "epoch": 4.74, "learning_rate": 2.6336146272855133e-05, "loss": 1.3727, "step": 87690 }, { "epoch": 4.74, "learning_rate": 2.6333441523314946e-05, "loss": 1.3346, "step": 87700 }, { "epoch": 4.74, "learning_rate": 2.633073677377475e-05, "loss": 1.3017, "step": 87710 }, { "epoch": 4.74, "learning_rate": 2.6328032024234555e-05, "loss": 1.4062, "step": 87720 }, { "epoch": 4.74, "learning_rate": 2.6325327274694368e-05, "loss": 1.5152, "step": 87730 }, { "epoch": 4.74, "learning_rate": 2.6322622525154174e-05, "loss": 1.45, "step": 87740 }, { "epoch": 4.74, "learning_rate": 2.6319917775613977e-05, "loss": 1.441, "step": 87750 }, { "epoch": 4.74, "learning_rate": 2.631721302607379e-05, "loss": 1.4307, "step": 87760 }, { "epoch": 4.74, "learning_rate": 2.6314508276533596e-05, "loss": 1.3969, "step": 87770 }, { "epoch": 4.74, "learning_rate": 2.63118035269934e-05, "loss": 1.3722, "step": 87780 }, { "epoch": 4.74, "learning_rate": 2.6309098777453212e-05, "loss": 1.4327, "step": 87790 }, { "epoch": 4.74, "learning_rate": 2.630639402791302e-05, "loss": 1.3591, "step": 87800 }, { "epoch": 4.74, "learning_rate": 2.630368927837282e-05, "loss": 1.4519, "step": 87810 }, { "epoch": 4.75, "learning_rate": 2.6300984528832634e-05, "loss": 1.3643, "step": 87820 }, { "epoch": 4.75, "learning_rate": 2.629827977929244e-05, "loss": 1.4115, "step": 87830 }, { "epoch": 4.75, "learning_rate": 2.6295575029752247e-05, "loss": 1.4116, "step": 87840 }, { "epoch": 4.75, "learning_rate": 2.6292870280212056e-05, "loss": 1.4113, "step": 87850 }, { "epoch": 4.75, "learning_rate": 2.6290165530671863e-05, "loss": 1.3671, "step": 87860 }, { "epoch": 4.75, "learning_rate": 2.628746078113167e-05, "loss": 1.3391, "step": 87870 }, { "epoch": 4.75, "learning_rate": 2.6284756031591472e-05, "loss": 1.3462, "step": 87880 }, { "epoch": 4.75, "learning_rate": 2.6282051282051285e-05, "loss": 1.4381, "step": 87890 }, { "epoch": 4.75, "learning_rate": 2.627934653251109e-05, "loss": 1.3617, "step": 87900 }, { "epoch": 4.75, "learning_rate": 2.6276641782970894e-05, "loss": 1.3943, "step": 87910 }, { "epoch": 4.75, "learning_rate": 2.6273937033430707e-05, "loss": 1.4069, "step": 87920 }, { "epoch": 4.75, "learning_rate": 2.6271232283890513e-05, "loss": 1.3068, "step": 87930 }, { "epoch": 4.75, "learning_rate": 2.626852753435032e-05, "loss": 1.3353, "step": 87940 }, { "epoch": 4.75, "learning_rate": 2.626582278481013e-05, "loss": 1.3876, "step": 87950 }, { "epoch": 4.75, "learning_rate": 2.6263118035269935e-05, "loss": 1.4589, "step": 87960 }, { "epoch": 4.75, "learning_rate": 2.6260413285729742e-05, "loss": 1.4144, "step": 87970 }, { "epoch": 4.75, "learning_rate": 2.625770853618955e-05, "loss": 1.4391, "step": 87980 }, { "epoch": 4.75, "learning_rate": 2.6255003786649358e-05, "loss": 1.3674, "step": 87990 }, { "epoch": 4.76, "learning_rate": 2.6252299037109164e-05, "loss": 1.3577, "step": 88000 }, { "epoch": 4.76, "learning_rate": 2.6249594287568974e-05, "loss": 1.3548, "step": 88010 }, { "epoch": 4.76, "learning_rate": 2.624688953802878e-05, "loss": 1.39, "step": 88020 }, { "epoch": 4.76, "learning_rate": 2.6244184788488586e-05, "loss": 1.3888, "step": 88030 }, { "epoch": 4.76, "learning_rate": 2.6241480038948396e-05, "loss": 1.4458, "step": 88040 }, { "epoch": 4.76, "learning_rate": 2.6238775289408202e-05, "loss": 1.3662, "step": 88050 }, { "epoch": 4.76, "learning_rate": 2.6236070539868008e-05, "loss": 1.4348, "step": 88060 }, { "epoch": 4.76, "learning_rate": 2.6233365790327818e-05, "loss": 1.4249, "step": 88070 }, { "epoch": 4.76, "learning_rate": 2.6230661040787624e-05, "loss": 1.2966, "step": 88080 }, { "epoch": 4.76, "learning_rate": 2.622795629124743e-05, "loss": 1.3695, "step": 88090 }, { "epoch": 4.76, "learning_rate": 2.622525154170724e-05, "loss": 1.382, "step": 88100 }, { "epoch": 4.76, "learning_rate": 2.6222546792167046e-05, "loss": 1.41, "step": 88110 }, { "epoch": 4.76, "learning_rate": 2.6219842042626853e-05, "loss": 1.4242, "step": 88120 }, { "epoch": 4.76, "learning_rate": 2.6217137293086662e-05, "loss": 1.2567, "step": 88130 }, { "epoch": 4.76, "learning_rate": 2.621443254354647e-05, "loss": 1.3331, "step": 88140 }, { "epoch": 4.76, "learning_rate": 2.6211727794006275e-05, "loss": 1.4195, "step": 88150 }, { "epoch": 4.76, "learning_rate": 2.6209023044466084e-05, "loss": 1.3688, "step": 88160 }, { "epoch": 4.76, "learning_rate": 2.620631829492589e-05, "loss": 1.3447, "step": 88170 }, { "epoch": 4.76, "learning_rate": 2.6203613545385697e-05, "loss": 1.3334, "step": 88180 }, { "epoch": 4.77, "learning_rate": 2.6200908795845507e-05, "loss": 1.3972, "step": 88190 }, { "epoch": 4.77, "learning_rate": 2.6198204046305313e-05, "loss": 1.4691, "step": 88200 }, { "epoch": 4.77, "learning_rate": 2.619549929676512e-05, "loss": 1.3801, "step": 88210 }, { "epoch": 4.77, "learning_rate": 2.619279454722493e-05, "loss": 1.3916, "step": 88220 }, { "epoch": 4.77, "learning_rate": 2.6190089797684735e-05, "loss": 1.3812, "step": 88230 }, { "epoch": 4.77, "learning_rate": 2.618738504814454e-05, "loss": 1.3646, "step": 88240 }, { "epoch": 4.77, "learning_rate": 2.6184680298604354e-05, "loss": 1.3331, "step": 88250 }, { "epoch": 4.77, "learning_rate": 2.6181975549064157e-05, "loss": 1.3976, "step": 88260 }, { "epoch": 4.77, "learning_rate": 2.6179270799523963e-05, "loss": 1.3967, "step": 88270 }, { "epoch": 4.77, "learning_rate": 2.6176566049983776e-05, "loss": 1.3627, "step": 88280 }, { "epoch": 4.77, "learning_rate": 2.617386130044358e-05, "loss": 1.3338, "step": 88290 }, { "epoch": 4.77, "learning_rate": 2.6171156550903386e-05, "loss": 1.3836, "step": 88300 }, { "epoch": 4.77, "learning_rate": 2.61684518013632e-05, "loss": 1.3945, "step": 88310 }, { "epoch": 4.77, "learning_rate": 2.6165747051823e-05, "loss": 1.3649, "step": 88320 }, { "epoch": 4.77, "learning_rate": 2.6163042302282808e-05, "loss": 1.3905, "step": 88330 }, { "epoch": 4.77, "learning_rate": 2.616033755274262e-05, "loss": 1.3631, "step": 88340 }, { "epoch": 4.77, "learning_rate": 2.6157632803202427e-05, "loss": 1.407, "step": 88350 }, { "epoch": 4.77, "learning_rate": 2.615492805366223e-05, "loss": 1.4083, "step": 88360 }, { "epoch": 4.78, "learning_rate": 2.6152223304122043e-05, "loss": 1.4369, "step": 88370 }, { "epoch": 4.78, "learning_rate": 2.614951855458185e-05, "loss": 1.3767, "step": 88380 }, { "epoch": 4.78, "learning_rate": 2.6146813805041652e-05, "loss": 1.4477, "step": 88390 }, { "epoch": 4.78, "learning_rate": 2.6144109055501465e-05, "loss": 1.4187, "step": 88400 }, { "epoch": 4.78, "learning_rate": 2.614140430596127e-05, "loss": 1.4882, "step": 88410 }, { "epoch": 4.78, "learning_rate": 2.6138699556421074e-05, "loss": 1.4308, "step": 88420 }, { "epoch": 4.78, "learning_rate": 2.613599480688088e-05, "loss": 1.3949, "step": 88430 }, { "epoch": 4.78, "learning_rate": 2.6133290057340694e-05, "loss": 1.424, "step": 88440 }, { "epoch": 4.78, "learning_rate": 2.6130585307800496e-05, "loss": 1.3403, "step": 88450 }, { "epoch": 4.78, "learning_rate": 2.6127880558260303e-05, "loss": 1.4179, "step": 88460 }, { "epoch": 4.78, "learning_rate": 2.6125175808720116e-05, "loss": 1.3899, "step": 88470 }, { "epoch": 4.78, "learning_rate": 2.6122471059179922e-05, "loss": 1.3891, "step": 88480 }, { "epoch": 4.78, "learning_rate": 2.6119766309639725e-05, "loss": 1.3798, "step": 88490 }, { "epoch": 4.78, "learning_rate": 2.6117061560099538e-05, "loss": 1.4014, "step": 88500 }, { "epoch": 4.78, "learning_rate": 2.6114356810559344e-05, "loss": 1.3898, "step": 88510 }, { "epoch": 4.78, "learning_rate": 2.6111652061019147e-05, "loss": 1.4142, "step": 88520 }, { "epoch": 4.78, "learning_rate": 2.610894731147896e-05, "loss": 1.3384, "step": 88530 }, { "epoch": 4.78, "learning_rate": 2.6106242561938766e-05, "loss": 1.3579, "step": 88540 }, { "epoch": 4.78, "learning_rate": 2.610353781239857e-05, "loss": 1.393, "step": 88550 }, { "epoch": 4.79, "learning_rate": 2.6100833062858382e-05, "loss": 1.3624, "step": 88560 }, { "epoch": 4.79, "learning_rate": 2.609812831331819e-05, "loss": 1.3759, "step": 88570 }, { "epoch": 4.79, "learning_rate": 2.6095423563777995e-05, "loss": 1.3856, "step": 88580 }, { "epoch": 4.79, "learning_rate": 2.6092718814237804e-05, "loss": 1.421, "step": 88590 }, { "epoch": 4.79, "learning_rate": 2.609001406469761e-05, "loss": 1.4155, "step": 88600 }, { "epoch": 4.79, "learning_rate": 2.6087309315157417e-05, "loss": 1.4011, "step": 88610 }, { "epoch": 4.79, "learning_rate": 2.6084604565617227e-05, "loss": 1.3117, "step": 88620 }, { "epoch": 4.79, "learning_rate": 2.6081899816077033e-05, "loss": 1.4676, "step": 88630 }, { "epoch": 4.79, "learning_rate": 2.607919506653684e-05, "loss": 1.3103, "step": 88640 }, { "epoch": 4.79, "learning_rate": 2.607649031699665e-05, "loss": 1.3456, "step": 88650 }, { "epoch": 4.79, "learning_rate": 2.6073785567456455e-05, "loss": 1.3892, "step": 88660 }, { "epoch": 4.79, "learning_rate": 2.607108081791626e-05, "loss": 1.3109, "step": 88670 }, { "epoch": 4.79, "learning_rate": 2.606837606837607e-05, "loss": 1.413, "step": 88680 }, { "epoch": 4.79, "learning_rate": 2.6065671318835877e-05, "loss": 1.3926, "step": 88690 }, { "epoch": 4.79, "learning_rate": 2.6062966569295683e-05, "loss": 1.3897, "step": 88700 }, { "epoch": 4.79, "learning_rate": 2.6060261819755493e-05, "loss": 1.3253, "step": 88710 }, { "epoch": 4.79, "learning_rate": 2.60575570702153e-05, "loss": 1.3735, "step": 88720 }, { "epoch": 4.79, "learning_rate": 2.6054852320675106e-05, "loss": 1.4734, "step": 88730 }, { "epoch": 4.8, "learning_rate": 2.6052147571134915e-05, "loss": 1.3657, "step": 88740 }, { "epoch": 4.8, "learning_rate": 2.604944282159472e-05, "loss": 1.3775, "step": 88750 }, { "epoch": 4.8, "learning_rate": 2.6046738072054528e-05, "loss": 1.3838, "step": 88760 }, { "epoch": 4.8, "learning_rate": 2.6044033322514337e-05, "loss": 1.4277, "step": 88770 }, { "epoch": 4.8, "learning_rate": 2.6041328572974144e-05, "loss": 1.3847, "step": 88780 }, { "epoch": 4.8, "learning_rate": 2.603862382343395e-05, "loss": 1.3589, "step": 88790 }, { "epoch": 4.8, "learning_rate": 2.603591907389376e-05, "loss": 1.4058, "step": 88800 }, { "epoch": 4.8, "learning_rate": 2.6033214324353566e-05, "loss": 1.375, "step": 88810 }, { "epoch": 4.8, "learning_rate": 2.6030509574813372e-05, "loss": 1.4341, "step": 88820 }, { "epoch": 4.8, "learning_rate": 2.6027804825273182e-05, "loss": 1.3603, "step": 88830 }, { "epoch": 4.8, "learning_rate": 2.6025100075732988e-05, "loss": 1.432, "step": 88840 }, { "epoch": 4.8, "learning_rate": 2.6022395326192794e-05, "loss": 1.3822, "step": 88850 }, { "epoch": 4.8, "learning_rate": 2.6019690576652604e-05, "loss": 1.3956, "step": 88860 }, { "epoch": 4.8, "learning_rate": 2.601698582711241e-05, "loss": 1.46, "step": 88870 }, { "epoch": 4.8, "learning_rate": 2.6014281077572216e-05, "loss": 1.4424, "step": 88880 }, { "epoch": 4.8, "learning_rate": 2.601157632803203e-05, "loss": 1.4212, "step": 88890 }, { "epoch": 4.8, "learning_rate": 2.6008871578491832e-05, "loss": 1.4153, "step": 88900 }, { "epoch": 4.8, "learning_rate": 2.600616682895164e-05, "loss": 1.4003, "step": 88910 }, { "epoch": 4.8, "learning_rate": 2.600346207941145e-05, "loss": 1.3819, "step": 88920 }, { "epoch": 4.81, "learning_rate": 2.6000757329871255e-05, "loss": 1.4359, "step": 88930 }, { "epoch": 4.81, "learning_rate": 2.599805258033106e-05, "loss": 1.3901, "step": 88940 }, { "epoch": 4.81, "learning_rate": 2.5995347830790874e-05, "loss": 1.3421, "step": 88950 }, { "epoch": 4.81, "learning_rate": 2.5992643081250677e-05, "loss": 1.3824, "step": 88960 }, { "epoch": 4.81, "learning_rate": 2.5989938331710483e-05, "loss": 1.3197, "step": 88970 }, { "epoch": 4.81, "learning_rate": 2.5987233582170296e-05, "loss": 1.4669, "step": 88980 }, { "epoch": 4.81, "learning_rate": 2.5984528832630102e-05, "loss": 1.3765, "step": 88990 }, { "epoch": 4.81, "learning_rate": 2.5981824083089905e-05, "loss": 1.3403, "step": 89000 }, { "epoch": 4.81, "learning_rate": 2.597911933354971e-05, "loss": 1.3217, "step": 89010 }, { "epoch": 4.81, "learning_rate": 2.5976414584009524e-05, "loss": 1.3108, "step": 89020 }, { "epoch": 4.81, "learning_rate": 2.5973709834469327e-05, "loss": 1.444, "step": 89030 }, { "epoch": 4.81, "learning_rate": 2.5971005084929134e-05, "loss": 1.364, "step": 89040 }, { "epoch": 4.81, "learning_rate": 2.5968300335388947e-05, "loss": 1.3291, "step": 89050 }, { "epoch": 4.81, "learning_rate": 2.596559558584875e-05, "loss": 1.4171, "step": 89060 }, { "epoch": 4.81, "learning_rate": 2.5962890836308556e-05, "loss": 1.3996, "step": 89070 }, { "epoch": 4.81, "learning_rate": 2.596018608676837e-05, "loss": 1.4254, "step": 89080 }, { "epoch": 4.81, "learning_rate": 2.595748133722817e-05, "loss": 1.4912, "step": 89090 }, { "epoch": 4.81, "learning_rate": 2.5954776587687978e-05, "loss": 1.4263, "step": 89100 }, { "epoch": 4.81, "learning_rate": 2.595207183814779e-05, "loss": 1.414, "step": 89110 }, { "epoch": 4.82, "learning_rate": 2.5949367088607597e-05, "loss": 1.3851, "step": 89120 }, { "epoch": 4.82, "learning_rate": 2.59466623390674e-05, "loss": 1.3241, "step": 89130 }, { "epoch": 4.82, "learning_rate": 2.5943957589527213e-05, "loss": 1.3294, "step": 89140 }, { "epoch": 4.82, "learning_rate": 2.594125283998702e-05, "loss": 1.4164, "step": 89150 }, { "epoch": 4.82, "learning_rate": 2.5938548090446822e-05, "loss": 1.3238, "step": 89160 }, { "epoch": 4.82, "learning_rate": 2.5935843340906635e-05, "loss": 1.3795, "step": 89170 }, { "epoch": 4.82, "learning_rate": 2.593313859136644e-05, "loss": 1.3837, "step": 89180 }, { "epoch": 4.82, "learning_rate": 2.5930433841826244e-05, "loss": 1.3356, "step": 89190 }, { "epoch": 4.82, "learning_rate": 2.5927729092286057e-05, "loss": 1.4022, "step": 89200 }, { "epoch": 4.82, "learning_rate": 2.5925024342745864e-05, "loss": 1.3626, "step": 89210 }, { "epoch": 4.82, "learning_rate": 2.592231959320567e-05, "loss": 1.371, "step": 89220 }, { "epoch": 4.82, "learning_rate": 2.591961484366548e-05, "loss": 1.3916, "step": 89230 }, { "epoch": 4.82, "learning_rate": 2.5916910094125286e-05, "loss": 1.3512, "step": 89240 }, { "epoch": 4.82, "learning_rate": 2.5914205344585092e-05, "loss": 1.4202, "step": 89250 }, { "epoch": 4.82, "learning_rate": 2.5911500595044902e-05, "loss": 1.3846, "step": 89260 }, { "epoch": 4.82, "learning_rate": 2.5908795845504708e-05, "loss": 1.3357, "step": 89270 }, { "epoch": 4.82, "learning_rate": 2.5906091095964514e-05, "loss": 1.4034, "step": 89280 }, { "epoch": 4.82, "learning_rate": 2.5903386346424324e-05, "loss": 1.3794, "step": 89290 }, { "epoch": 4.83, "learning_rate": 2.590068159688413e-05, "loss": 1.3345, "step": 89300 }, { "epoch": 4.83, "learning_rate": 2.5897976847343936e-05, "loss": 1.3871, "step": 89310 }, { "epoch": 4.83, "learning_rate": 2.5895272097803746e-05, "loss": 1.3384, "step": 89320 }, { "epoch": 4.83, "learning_rate": 2.5892567348263552e-05, "loss": 1.4526, "step": 89330 }, { "epoch": 4.83, "learning_rate": 2.588986259872336e-05, "loss": 1.3103, "step": 89340 }, { "epoch": 4.83, "learning_rate": 2.5887157849183168e-05, "loss": 1.328, "step": 89350 }, { "epoch": 4.83, "learning_rate": 2.5884453099642975e-05, "loss": 1.4107, "step": 89360 }, { "epoch": 4.83, "learning_rate": 2.588174835010278e-05, "loss": 1.3026, "step": 89370 }, { "epoch": 4.83, "learning_rate": 2.587904360056259e-05, "loss": 1.3814, "step": 89380 }, { "epoch": 4.83, "learning_rate": 2.5876338851022397e-05, "loss": 1.3258, "step": 89390 }, { "epoch": 4.83, "learning_rate": 2.5873634101482203e-05, "loss": 1.4124, "step": 89400 }, { "epoch": 4.83, "learning_rate": 2.5870929351942013e-05, "loss": 1.3964, "step": 89410 }, { "epoch": 4.83, "learning_rate": 2.586822460240182e-05, "loss": 1.372, "step": 89420 }, { "epoch": 4.83, "learning_rate": 2.5865519852861625e-05, "loss": 1.4268, "step": 89430 }, { "epoch": 4.83, "learning_rate": 2.5862815103321435e-05, "loss": 1.3971, "step": 89440 }, { "epoch": 4.83, "learning_rate": 2.586011035378124e-05, "loss": 1.4274, "step": 89450 }, { "epoch": 4.83, "learning_rate": 2.5857405604241047e-05, "loss": 1.4491, "step": 89460 }, { "epoch": 4.83, "learning_rate": 2.5854700854700857e-05, "loss": 1.3152, "step": 89470 }, { "epoch": 4.83, "learning_rate": 2.5851996105160663e-05, "loss": 1.3621, "step": 89480 }, { "epoch": 4.84, "learning_rate": 2.584929135562047e-05, "loss": 1.3489, "step": 89490 }, { "epoch": 4.84, "learning_rate": 2.584658660608028e-05, "loss": 1.3395, "step": 89500 }, { "epoch": 4.84, "learning_rate": 2.5843881856540085e-05, "loss": 1.3813, "step": 89510 }, { "epoch": 4.84, "learning_rate": 2.584117710699989e-05, "loss": 1.358, "step": 89520 }, { "epoch": 4.84, "learning_rate": 2.5838472357459705e-05, "loss": 1.3838, "step": 89530 }, { "epoch": 4.84, "learning_rate": 2.5835767607919508e-05, "loss": 1.3476, "step": 89540 }, { "epoch": 4.84, "learning_rate": 2.5833062858379314e-05, "loss": 1.3747, "step": 89550 }, { "epoch": 4.84, "learning_rate": 2.583035810883912e-05, "loss": 1.3694, "step": 89560 }, { "epoch": 4.84, "learning_rate": 2.582765335929893e-05, "loss": 1.417, "step": 89570 }, { "epoch": 4.84, "learning_rate": 2.5824948609758736e-05, "loss": 1.3195, "step": 89580 }, { "epoch": 4.84, "learning_rate": 2.5822243860218542e-05, "loss": 1.3985, "step": 89590 }, { "epoch": 4.84, "learning_rate": 2.5819539110678352e-05, "loss": 1.4951, "step": 89600 }, { "epoch": 4.84, "learning_rate": 2.5816834361138158e-05, "loss": 1.3441, "step": 89610 }, { "epoch": 4.84, "learning_rate": 2.5814129611597964e-05, "loss": 1.3686, "step": 89620 }, { "epoch": 4.84, "learning_rate": 2.5811424862057777e-05, "loss": 1.4357, "step": 89630 }, { "epoch": 4.84, "learning_rate": 2.580872011251758e-05, "loss": 1.3489, "step": 89640 }, { "epoch": 4.84, "learning_rate": 2.5806015362977387e-05, "loss": 1.3462, "step": 89650 }, { "epoch": 4.84, "learning_rate": 2.58033106134372e-05, "loss": 1.4521, "step": 89660 }, { "epoch": 4.85, "learning_rate": 2.5800605863897003e-05, "loss": 1.4016, "step": 89670 }, { "epoch": 4.85, "learning_rate": 2.579790111435681e-05, "loss": 1.4493, "step": 89680 }, { "epoch": 4.85, "learning_rate": 2.5795196364816622e-05, "loss": 1.3362, "step": 89690 }, { "epoch": 4.85, "learning_rate": 2.5792491615276425e-05, "loss": 1.3977, "step": 89700 }, { "epoch": 4.85, "learning_rate": 2.578978686573623e-05, "loss": 1.3002, "step": 89710 }, { "epoch": 4.85, "learning_rate": 2.5787082116196044e-05, "loss": 1.4066, "step": 89720 }, { "epoch": 4.85, "learning_rate": 2.578437736665585e-05, "loss": 1.3927, "step": 89730 }, { "epoch": 4.85, "learning_rate": 2.5781672617115653e-05, "loss": 1.3936, "step": 89740 }, { "epoch": 4.85, "learning_rate": 2.5778967867575466e-05, "loss": 1.3975, "step": 89750 }, { "epoch": 4.85, "learning_rate": 2.5776263118035272e-05, "loss": 1.3172, "step": 89760 }, { "epoch": 4.85, "learning_rate": 2.5773558368495075e-05, "loss": 1.3447, "step": 89770 }, { "epoch": 4.85, "learning_rate": 2.577085361895489e-05, "loss": 1.3743, "step": 89780 }, { "epoch": 4.85, "learning_rate": 2.5768148869414695e-05, "loss": 1.4705, "step": 89790 }, { "epoch": 4.85, "learning_rate": 2.5765444119874497e-05, "loss": 1.4161, "step": 89800 }, { "epoch": 4.85, "learning_rate": 2.576273937033431e-05, "loss": 1.2942, "step": 89810 }, { "epoch": 4.85, "learning_rate": 2.5760034620794117e-05, "loss": 1.3445, "step": 89820 }, { "epoch": 4.85, "learning_rate": 2.575732987125392e-05, "loss": 1.3379, "step": 89830 }, { "epoch": 4.85, "learning_rate": 2.5754625121713733e-05, "loss": 1.3779, "step": 89840 }, { "epoch": 4.85, "learning_rate": 2.575192037217354e-05, "loss": 1.3967, "step": 89850 }, { "epoch": 4.86, "learning_rate": 2.5749215622633345e-05, "loss": 1.3735, "step": 89860 }, { "epoch": 4.86, "learning_rate": 2.5746510873093155e-05, "loss": 1.3865, "step": 89870 }, { "epoch": 4.86, "learning_rate": 2.574380612355296e-05, "loss": 1.3268, "step": 89880 }, { "epoch": 4.86, "learning_rate": 2.5741101374012767e-05, "loss": 1.3545, "step": 89890 }, { "epoch": 4.86, "learning_rate": 2.5738396624472577e-05, "loss": 1.4247, "step": 89900 }, { "epoch": 4.86, "learning_rate": 2.5735691874932383e-05, "loss": 1.4283, "step": 89910 }, { "epoch": 4.86, "learning_rate": 2.573298712539219e-05, "loss": 1.4239, "step": 89920 }, { "epoch": 4.86, "learning_rate": 2.5730282375852e-05, "loss": 1.3465, "step": 89930 }, { "epoch": 4.86, "learning_rate": 2.5727577626311805e-05, "loss": 1.3742, "step": 89940 }, { "epoch": 4.86, "learning_rate": 2.572487287677161e-05, "loss": 1.3811, "step": 89950 }, { "epoch": 4.86, "learning_rate": 2.572216812723142e-05, "loss": 1.446, "step": 89960 }, { "epoch": 4.86, "learning_rate": 2.5719463377691228e-05, "loss": 1.3903, "step": 89970 }, { "epoch": 4.86, "learning_rate": 2.5716758628151034e-05, "loss": 1.3317, "step": 89980 }, { "epoch": 4.86, "learning_rate": 2.5714053878610844e-05, "loss": 1.3724, "step": 89990 }, { "epoch": 4.86, "learning_rate": 2.571134912907065e-05, "loss": 1.4531, "step": 90000 }, { "epoch": 4.86, "learning_rate": 2.5708644379530456e-05, "loss": 1.3646, "step": 90010 }, { "epoch": 4.86, "learning_rate": 2.5705939629990266e-05, "loss": 1.4086, "step": 90020 }, { "epoch": 4.86, "learning_rate": 2.5703234880450072e-05, "loss": 1.3495, "step": 90030 }, { "epoch": 4.87, "learning_rate": 2.5700530130909878e-05, "loss": 1.3317, "step": 90040 }, { "epoch": 4.87, "learning_rate": 2.5697825381369688e-05, "loss": 1.2927, "step": 90050 }, { "epoch": 4.87, "learning_rate": 2.5695120631829494e-05, "loss": 1.3712, "step": 90060 }, { "epoch": 4.87, "learning_rate": 2.56924158822893e-05, "loss": 1.4063, "step": 90070 }, { "epoch": 4.87, "learning_rate": 2.568971113274911e-05, "loss": 1.5134, "step": 90080 }, { "epoch": 4.87, "learning_rate": 2.5687006383208916e-05, "loss": 1.4106, "step": 90090 }, { "epoch": 4.87, "learning_rate": 2.5684301633668723e-05, "loss": 1.3878, "step": 90100 }, { "epoch": 4.87, "learning_rate": 2.568159688412853e-05, "loss": 1.4382, "step": 90110 }, { "epoch": 4.87, "learning_rate": 2.567889213458834e-05, "loss": 1.4173, "step": 90120 }, { "epoch": 4.87, "learning_rate": 2.5676187385048145e-05, "loss": 1.38, "step": 90130 }, { "epoch": 4.87, "learning_rate": 2.567348263550795e-05, "loss": 1.3709, "step": 90140 }, { "epoch": 4.87, "learning_rate": 2.567077788596776e-05, "loss": 1.3119, "step": 90150 }, { "epoch": 4.87, "learning_rate": 2.5668073136427567e-05, "loss": 1.4057, "step": 90160 }, { "epoch": 4.87, "learning_rate": 2.5665368386887373e-05, "loss": 1.4098, "step": 90170 }, { "epoch": 4.87, "learning_rate": 2.5662663637347183e-05, "loss": 1.4152, "step": 90180 }, { "epoch": 4.87, "learning_rate": 2.565995888780699e-05, "loss": 1.3344, "step": 90190 }, { "epoch": 4.87, "learning_rate": 2.5657254138266795e-05, "loss": 1.4066, "step": 90200 }, { "epoch": 4.87, "learning_rate": 2.5654549388726605e-05, "loss": 1.4089, "step": 90210 }, { "epoch": 4.87, "learning_rate": 2.565184463918641e-05, "loss": 1.2552, "step": 90220 }, { "epoch": 4.88, "learning_rate": 2.5649139889646217e-05, "loss": 1.4827, "step": 90230 }, { "epoch": 4.88, "learning_rate": 2.5646435140106027e-05, "loss": 1.4321, "step": 90240 }, { "epoch": 4.88, "learning_rate": 2.5643730390565833e-05, "loss": 1.4751, "step": 90250 }, { "epoch": 4.88, "learning_rate": 2.564102564102564e-05, "loss": 1.3572, "step": 90260 }, { "epoch": 4.88, "learning_rate": 2.5638320891485453e-05, "loss": 1.3914, "step": 90270 }, { "epoch": 4.88, "learning_rate": 2.5635616141945256e-05, "loss": 1.4109, "step": 90280 }, { "epoch": 4.88, "learning_rate": 2.5632911392405062e-05, "loss": 1.3259, "step": 90290 }, { "epoch": 4.88, "learning_rate": 2.5630206642864875e-05, "loss": 1.3492, "step": 90300 }, { "epoch": 4.88, "learning_rate": 2.5627501893324678e-05, "loss": 1.3583, "step": 90310 }, { "epoch": 4.88, "learning_rate": 2.5624797143784484e-05, "loss": 1.3969, "step": 90320 }, { "epoch": 4.88, "learning_rate": 2.5622092394244297e-05, "loss": 1.4264, "step": 90330 }, { "epoch": 4.88, "learning_rate": 2.56193876447041e-05, "loss": 1.3946, "step": 90340 }, { "epoch": 4.88, "learning_rate": 2.5616682895163906e-05, "loss": 1.3864, "step": 90350 }, { "epoch": 4.88, "learning_rate": 2.561397814562372e-05, "loss": 1.4146, "step": 90360 }, { "epoch": 4.88, "learning_rate": 2.5611273396083525e-05, "loss": 1.366, "step": 90370 }, { "epoch": 4.88, "learning_rate": 2.560856864654333e-05, "loss": 1.4234, "step": 90380 }, { "epoch": 4.88, "learning_rate": 2.560586389700314e-05, "loss": 1.4162, "step": 90390 }, { "epoch": 4.88, "learning_rate": 2.5603159147462948e-05, "loss": 1.3863, "step": 90400 }, { "epoch": 4.89, "learning_rate": 2.560045439792275e-05, "loss": 1.4113, "step": 90410 }, { "epoch": 4.89, "learning_rate": 2.5597749648382564e-05, "loss": 1.3219, "step": 90420 }, { "epoch": 4.89, "learning_rate": 2.559504489884237e-05, "loss": 1.3114, "step": 90430 }, { "epoch": 4.89, "learning_rate": 2.5592340149302173e-05, "loss": 1.3214, "step": 90440 }, { "epoch": 4.89, "learning_rate": 2.5589635399761986e-05, "loss": 1.4344, "step": 90450 }, { "epoch": 4.89, "learning_rate": 2.5586930650221792e-05, "loss": 1.3309, "step": 90460 }, { "epoch": 4.89, "learning_rate": 2.5584225900681595e-05, "loss": 1.3239, "step": 90470 }, { "epoch": 4.89, "learning_rate": 2.5581521151141408e-05, "loss": 1.3857, "step": 90480 }, { "epoch": 4.89, "learning_rate": 2.5578816401601214e-05, "loss": 1.3455, "step": 90490 }, { "epoch": 4.89, "learning_rate": 2.557611165206102e-05, "loss": 1.3858, "step": 90500 }, { "epoch": 4.89, "learning_rate": 2.557340690252083e-05, "loss": 1.4198, "step": 90510 }, { "epoch": 4.89, "learning_rate": 2.5570702152980636e-05, "loss": 1.3589, "step": 90520 }, { "epoch": 4.89, "learning_rate": 2.5567997403440443e-05, "loss": 1.3756, "step": 90530 }, { "epoch": 4.89, "learning_rate": 2.5565292653900252e-05, "loss": 1.3773, "step": 90540 }, { "epoch": 4.89, "learning_rate": 2.556258790436006e-05, "loss": 1.3679, "step": 90550 }, { "epoch": 4.89, "learning_rate": 2.5559883154819865e-05, "loss": 1.4028, "step": 90560 }, { "epoch": 4.89, "learning_rate": 2.5557178405279674e-05, "loss": 1.3895, "step": 90570 }, { "epoch": 4.89, "learning_rate": 2.555447365573948e-05, "loss": 1.3341, "step": 90580 }, { "epoch": 4.89, "learning_rate": 2.5551768906199287e-05, "loss": 1.3928, "step": 90590 }, { "epoch": 4.9, "learning_rate": 2.5549064156659097e-05, "loss": 1.3881, "step": 90600 }, { "epoch": 4.9, "learning_rate": 2.5546359407118903e-05, "loss": 1.3648, "step": 90610 }, { "epoch": 4.9, "learning_rate": 2.554365465757871e-05, "loss": 1.3786, "step": 90620 }, { "epoch": 4.9, "learning_rate": 2.554094990803852e-05, "loss": 1.4091, "step": 90630 }, { "epoch": 4.9, "learning_rate": 2.5538245158498325e-05, "loss": 1.4025, "step": 90640 }, { "epoch": 4.9, "learning_rate": 2.553554040895813e-05, "loss": 1.3722, "step": 90650 }, { "epoch": 4.9, "learning_rate": 2.553283565941794e-05, "loss": 1.3113, "step": 90660 }, { "epoch": 4.9, "learning_rate": 2.5530130909877747e-05, "loss": 1.3309, "step": 90670 }, { "epoch": 4.9, "learning_rate": 2.5527426160337553e-05, "loss": 1.447, "step": 90680 }, { "epoch": 4.9, "learning_rate": 2.552472141079736e-05, "loss": 1.398, "step": 90690 }, { "epoch": 4.9, "learning_rate": 2.552201666125717e-05, "loss": 1.326, "step": 90700 }, { "epoch": 4.9, "learning_rate": 2.5519311911716976e-05, "loss": 1.465, "step": 90710 }, { "epoch": 4.9, "learning_rate": 2.5516607162176782e-05, "loss": 1.3879, "step": 90720 }, { "epoch": 4.9, "learning_rate": 2.551390241263659e-05, "loss": 1.3663, "step": 90730 }, { "epoch": 4.9, "learning_rate": 2.5511197663096398e-05, "loss": 1.3854, "step": 90740 }, { "epoch": 4.9, "learning_rate": 2.5508492913556204e-05, "loss": 1.3848, "step": 90750 }, { "epoch": 4.9, "learning_rate": 2.5505788164016014e-05, "loss": 1.4286, "step": 90760 }, { "epoch": 4.9, "learning_rate": 2.550308341447582e-05, "loss": 1.4156, "step": 90770 }, { "epoch": 4.91, "learning_rate": 2.5500378664935626e-05, "loss": 1.3677, "step": 90780 }, { "epoch": 4.91, "learning_rate": 2.5497673915395436e-05, "loss": 1.3661, "step": 90790 }, { "epoch": 4.91, "learning_rate": 2.5494969165855242e-05, "loss": 1.3434, "step": 90800 }, { "epoch": 4.91, "learning_rate": 2.549226441631505e-05, "loss": 1.4813, "step": 90810 }, { "epoch": 4.91, "learning_rate": 2.5489559666774858e-05, "loss": 1.4011, "step": 90820 }, { "epoch": 4.91, "learning_rate": 2.5486854917234664e-05, "loss": 1.3098, "step": 90830 }, { "epoch": 4.91, "learning_rate": 2.548415016769447e-05, "loss": 1.396, "step": 90840 }, { "epoch": 4.91, "learning_rate": 2.548144541815428e-05, "loss": 1.3458, "step": 90850 }, { "epoch": 4.91, "learning_rate": 2.5478740668614086e-05, "loss": 1.4563, "step": 90860 }, { "epoch": 4.91, "learning_rate": 2.5476035919073893e-05, "loss": 1.3942, "step": 90870 }, { "epoch": 4.91, "learning_rate": 2.5473331169533702e-05, "loss": 1.3758, "step": 90880 }, { "epoch": 4.91, "learning_rate": 2.547062641999351e-05, "loss": 1.3723, "step": 90890 }, { "epoch": 4.91, "learning_rate": 2.5467921670453315e-05, "loss": 1.4374, "step": 90900 }, { "epoch": 4.91, "learning_rate": 2.5465216920913128e-05, "loss": 1.4538, "step": 90910 }, { "epoch": 4.91, "learning_rate": 2.546251217137293e-05, "loss": 1.3714, "step": 90920 }, { "epoch": 4.91, "learning_rate": 2.5459807421832737e-05, "loss": 1.5035, "step": 90930 }, { "epoch": 4.91, "learning_rate": 2.545710267229255e-05, "loss": 1.4169, "step": 90940 }, { "epoch": 4.91, "learning_rate": 2.5454397922752353e-05, "loss": 1.384, "step": 90950 }, { "epoch": 4.91, "learning_rate": 2.545169317321216e-05, "loss": 1.3596, "step": 90960 }, { "epoch": 4.92, "learning_rate": 2.5448988423671972e-05, "loss": 1.4346, "step": 90970 }, { "epoch": 4.92, "learning_rate": 2.5446283674131775e-05, "loss": 1.3236, "step": 90980 }, { "epoch": 4.92, "learning_rate": 2.544357892459158e-05, "loss": 1.4832, "step": 90990 }, { "epoch": 4.92, "learning_rate": 2.5440874175051394e-05, "loss": 1.4405, "step": 91000 }, { "epoch": 4.92, "learning_rate": 2.54381694255112e-05, "loss": 1.4651, "step": 91010 }, { "epoch": 4.92, "learning_rate": 2.5435464675971004e-05, "loss": 1.3394, "step": 91020 }, { "epoch": 4.92, "learning_rate": 2.5432759926430817e-05, "loss": 1.3894, "step": 91030 }, { "epoch": 4.92, "learning_rate": 2.5430055176890623e-05, "loss": 1.3246, "step": 91040 }, { "epoch": 4.92, "learning_rate": 2.5427350427350426e-05, "loss": 1.3912, "step": 91050 }, { "epoch": 4.92, "learning_rate": 2.542464567781024e-05, "loss": 1.3938, "step": 91060 }, { "epoch": 4.92, "learning_rate": 2.5421940928270045e-05, "loss": 1.4007, "step": 91070 }, { "epoch": 4.92, "learning_rate": 2.5419236178729848e-05, "loss": 1.4256, "step": 91080 }, { "epoch": 4.92, "learning_rate": 2.541653142918966e-05, "loss": 1.369, "step": 91090 }, { "epoch": 4.92, "learning_rate": 2.5413826679649467e-05, "loss": 1.3391, "step": 91100 }, { "epoch": 4.92, "learning_rate": 2.5411121930109273e-05, "loss": 1.3699, "step": 91110 }, { "epoch": 4.92, "learning_rate": 2.5408417180569083e-05, "loss": 1.3523, "step": 91120 }, { "epoch": 4.92, "learning_rate": 2.540571243102889e-05, "loss": 1.3293, "step": 91130 }, { "epoch": 4.92, "learning_rate": 2.5403007681488696e-05, "loss": 1.3227, "step": 91140 }, { "epoch": 4.93, "learning_rate": 2.5400302931948505e-05, "loss": 1.3701, "step": 91150 }, { "epoch": 4.93, "learning_rate": 2.539759818240831e-05, "loss": 1.349, "step": 91160 }, { "epoch": 4.93, "learning_rate": 2.5394893432868118e-05, "loss": 1.3259, "step": 91170 }, { "epoch": 4.93, "learning_rate": 2.5392188683327927e-05, "loss": 1.4773, "step": 91180 }, { "epoch": 4.93, "learning_rate": 2.5389483933787734e-05, "loss": 1.4603, "step": 91190 }, { "epoch": 4.93, "learning_rate": 2.538677918424754e-05, "loss": 1.4163, "step": 91200 }, { "epoch": 4.93, "learning_rate": 2.538407443470735e-05, "loss": 1.28, "step": 91210 }, { "epoch": 4.93, "learning_rate": 2.5381369685167156e-05, "loss": 1.4569, "step": 91220 }, { "epoch": 4.93, "learning_rate": 2.5378664935626962e-05, "loss": 1.4138, "step": 91230 }, { "epoch": 4.93, "learning_rate": 2.537596018608677e-05, "loss": 1.3701, "step": 91240 }, { "epoch": 4.93, "learning_rate": 2.5373255436546578e-05, "loss": 1.4304, "step": 91250 }, { "epoch": 4.93, "learning_rate": 2.5370550687006384e-05, "loss": 1.359, "step": 91260 }, { "epoch": 4.93, "learning_rate": 2.536784593746619e-05, "loss": 1.4219, "step": 91270 }, { "epoch": 4.93, "learning_rate": 2.5365141187926e-05, "loss": 1.3835, "step": 91280 }, { "epoch": 4.93, "learning_rate": 2.5362436438385806e-05, "loss": 1.3293, "step": 91290 }, { "epoch": 4.93, "learning_rate": 2.5359731688845613e-05, "loss": 1.4255, "step": 91300 }, { "epoch": 4.93, "learning_rate": 2.5357026939305422e-05, "loss": 1.3452, "step": 91310 }, { "epoch": 4.93, "learning_rate": 2.535432218976523e-05, "loss": 1.3528, "step": 91320 }, { "epoch": 4.93, "learning_rate": 2.5351617440225035e-05, "loss": 1.4486, "step": 91330 }, { "epoch": 4.94, "learning_rate": 2.5348912690684845e-05, "loss": 1.4088, "step": 91340 }, { "epoch": 4.94, "learning_rate": 2.534620794114465e-05, "loss": 1.434, "step": 91350 }, { "epoch": 4.94, "learning_rate": 2.5343503191604457e-05, "loss": 1.357, "step": 91360 }, { "epoch": 4.94, "learning_rate": 2.5340798442064267e-05, "loss": 1.3711, "step": 91370 }, { "epoch": 4.94, "learning_rate": 2.5338093692524073e-05, "loss": 1.4407, "step": 91380 }, { "epoch": 4.94, "learning_rate": 2.533538894298388e-05, "loss": 1.4542, "step": 91390 }, { "epoch": 4.94, "learning_rate": 2.533268419344369e-05, "loss": 1.3743, "step": 91400 }, { "epoch": 4.94, "learning_rate": 2.5329979443903495e-05, "loss": 1.4119, "step": 91410 }, { "epoch": 4.94, "learning_rate": 2.53272746943633e-05, "loss": 1.4267, "step": 91420 }, { "epoch": 4.94, "learning_rate": 2.532456994482311e-05, "loss": 1.392, "step": 91430 }, { "epoch": 4.94, "learning_rate": 2.5321865195282917e-05, "loss": 1.4112, "step": 91440 }, { "epoch": 4.94, "learning_rate": 2.5319160445742724e-05, "loss": 1.3301, "step": 91450 }, { "epoch": 4.94, "learning_rate": 2.5316455696202533e-05, "loss": 1.3551, "step": 91460 }, { "epoch": 4.94, "learning_rate": 2.531375094666234e-05, "loss": 1.3469, "step": 91470 }, { "epoch": 4.94, "learning_rate": 2.5311046197122146e-05, "loss": 1.4563, "step": 91480 }, { "epoch": 4.94, "learning_rate": 2.5308341447581955e-05, "loss": 1.3566, "step": 91490 }, { "epoch": 4.94, "learning_rate": 2.530563669804176e-05, "loss": 1.4148, "step": 91500 }, { "epoch": 4.94, "learning_rate": 2.5302931948501568e-05, "loss": 1.298, "step": 91510 }, { "epoch": 4.95, "learning_rate": 2.530022719896138e-05, "loss": 1.4038, "step": 91520 }, { "epoch": 4.95, "learning_rate": 2.5297522449421184e-05, "loss": 1.4339, "step": 91530 }, { "epoch": 4.95, "learning_rate": 2.529481769988099e-05, "loss": 1.3358, "step": 91540 }, { "epoch": 4.95, "learning_rate": 2.5292112950340803e-05, "loss": 1.328, "step": 91550 }, { "epoch": 4.95, "learning_rate": 2.5289408200800606e-05, "loss": 1.3931, "step": 91560 }, { "epoch": 4.95, "learning_rate": 2.5286703451260412e-05, "loss": 1.3741, "step": 91570 }, { "epoch": 4.95, "learning_rate": 2.5283998701720225e-05, "loss": 1.3512, "step": 91580 }, { "epoch": 4.95, "learning_rate": 2.5281293952180028e-05, "loss": 1.3038, "step": 91590 }, { "epoch": 4.95, "learning_rate": 2.5278589202639834e-05, "loss": 1.4122, "step": 91600 }, { "epoch": 4.95, "learning_rate": 2.5275884453099647e-05, "loss": 1.3937, "step": 91610 }, { "epoch": 4.95, "learning_rate": 2.527317970355945e-05, "loss": 1.4084, "step": 91620 }, { "epoch": 4.95, "learning_rate": 2.5270474954019257e-05, "loss": 1.3949, "step": 91630 }, { "epoch": 4.95, "learning_rate": 2.526777020447907e-05, "loss": 1.3526, "step": 91640 }, { "epoch": 4.95, "learning_rate": 2.5265065454938876e-05, "loss": 1.4019, "step": 91650 }, { "epoch": 4.95, "learning_rate": 2.526236070539868e-05, "loss": 1.3958, "step": 91660 }, { "epoch": 4.95, "learning_rate": 2.5259655955858492e-05, "loss": 1.4041, "step": 91670 }, { "epoch": 4.95, "learning_rate": 2.5256951206318298e-05, "loss": 1.3687, "step": 91680 }, { "epoch": 4.95, "learning_rate": 2.52542464567781e-05, "loss": 1.4466, "step": 91690 }, { "epoch": 4.95, "learning_rate": 2.5251541707237914e-05, "loss": 1.3832, "step": 91700 }, { "epoch": 4.96, "learning_rate": 2.524883695769772e-05, "loss": 1.4427, "step": 91710 }, { "epoch": 4.96, "learning_rate": 2.5246132208157523e-05, "loss": 1.3771, "step": 91720 }, { "epoch": 4.96, "learning_rate": 2.5243427458617336e-05, "loss": 1.3415, "step": 91730 }, { "epoch": 4.96, "learning_rate": 2.5240722709077142e-05, "loss": 1.4607, "step": 91740 }, { "epoch": 4.96, "learning_rate": 2.523801795953695e-05, "loss": 1.4067, "step": 91750 }, { "epoch": 4.96, "learning_rate": 2.5235313209996758e-05, "loss": 1.3857, "step": 91760 }, { "epoch": 4.96, "learning_rate": 2.5232608460456565e-05, "loss": 1.4312, "step": 91770 }, { "epoch": 4.96, "learning_rate": 2.522990371091637e-05, "loss": 1.4228, "step": 91780 }, { "epoch": 4.96, "learning_rate": 2.5227198961376174e-05, "loss": 1.3647, "step": 91790 }, { "epoch": 4.96, "learning_rate": 2.5224494211835987e-05, "loss": 1.3364, "step": 91800 }, { "epoch": 4.96, "learning_rate": 2.5221789462295793e-05, "loss": 1.412, "step": 91810 }, { "epoch": 4.96, "learning_rate": 2.5219084712755596e-05, "loss": 1.3512, "step": 91820 }, { "epoch": 4.96, "learning_rate": 2.521637996321541e-05, "loss": 1.4115, "step": 91830 }, { "epoch": 4.96, "learning_rate": 2.5213675213675215e-05, "loss": 1.3555, "step": 91840 }, { "epoch": 4.96, "learning_rate": 2.521097046413502e-05, "loss": 1.3934, "step": 91850 }, { "epoch": 4.96, "learning_rate": 2.520826571459483e-05, "loss": 1.3466, "step": 91860 }, { "epoch": 4.96, "learning_rate": 2.5205560965054637e-05, "loss": 1.3287, "step": 91870 }, { "epoch": 4.96, "learning_rate": 2.5202856215514444e-05, "loss": 1.3766, "step": 91880 }, { "epoch": 4.97, "learning_rate": 2.5200151465974253e-05, "loss": 1.4107, "step": 91890 }, { "epoch": 4.97, "learning_rate": 2.519744671643406e-05, "loss": 1.4719, "step": 91900 }, { "epoch": 4.97, "learning_rate": 2.5194741966893866e-05, "loss": 1.4189, "step": 91910 }, { "epoch": 4.97, "learning_rate": 2.5192037217353675e-05, "loss": 1.4127, "step": 91920 }, { "epoch": 4.97, "learning_rate": 2.518933246781348e-05, "loss": 1.398, "step": 91930 }, { "epoch": 4.97, "learning_rate": 2.5186627718273288e-05, "loss": 1.3523, "step": 91940 }, { "epoch": 4.97, "learning_rate": 2.5183922968733098e-05, "loss": 1.438, "step": 91950 }, { "epoch": 4.97, "learning_rate": 2.5181218219192904e-05, "loss": 1.3437, "step": 91960 }, { "epoch": 4.97, "learning_rate": 2.517851346965271e-05, "loss": 1.3385, "step": 91970 }, { "epoch": 4.97, "learning_rate": 2.517580872011252e-05, "loss": 1.42, "step": 91980 }, { "epoch": 4.97, "learning_rate": 2.5173103970572326e-05, "loss": 1.3772, "step": 91990 }, { "epoch": 4.97, "learning_rate": 2.5170399221032132e-05, "loss": 1.3935, "step": 92000 }, { "epoch": 4.97, "learning_rate": 2.5167694471491942e-05, "loss": 1.4158, "step": 92010 }, { "epoch": 4.97, "learning_rate": 2.5164989721951748e-05, "loss": 1.3702, "step": 92020 }, { "epoch": 4.97, "learning_rate": 2.5162284972411554e-05, "loss": 1.4074, "step": 92030 }, { "epoch": 4.97, "learning_rate": 2.5159580222871364e-05, "loss": 1.4196, "step": 92040 }, { "epoch": 4.97, "learning_rate": 2.515687547333117e-05, "loss": 1.355, "step": 92050 }, { "epoch": 4.97, "learning_rate": 2.5154170723790977e-05, "loss": 1.3193, "step": 92060 }, { "epoch": 4.97, "learning_rate": 2.5151465974250786e-05, "loss": 1.3568, "step": 92070 }, { "epoch": 4.98, "learning_rate": 2.5148761224710592e-05, "loss": 1.3922, "step": 92080 }, { "epoch": 4.98, "learning_rate": 2.51460564751704e-05, "loss": 1.3964, "step": 92090 }, { "epoch": 4.98, "learning_rate": 2.514335172563021e-05, "loss": 1.3631, "step": 92100 }, { "epoch": 4.98, "learning_rate": 2.5140646976090015e-05, "loss": 1.3473, "step": 92110 }, { "epoch": 4.98, "learning_rate": 2.513794222654982e-05, "loss": 1.3562, "step": 92120 }, { "epoch": 4.98, "learning_rate": 2.513523747700963e-05, "loss": 1.3811, "step": 92130 }, { "epoch": 4.98, "learning_rate": 2.5132532727469437e-05, "loss": 1.3098, "step": 92140 }, { "epoch": 4.98, "learning_rate": 2.5129827977929243e-05, "loss": 1.4229, "step": 92150 }, { "epoch": 4.98, "learning_rate": 2.5127123228389056e-05, "loss": 1.3779, "step": 92160 }, { "epoch": 4.98, "learning_rate": 2.512441847884886e-05, "loss": 1.3837, "step": 92170 }, { "epoch": 4.98, "learning_rate": 2.5121713729308665e-05, "loss": 1.2975, "step": 92180 }, { "epoch": 4.98, "learning_rate": 2.5119008979768478e-05, "loss": 1.3348, "step": 92190 }, { "epoch": 4.98, "learning_rate": 2.511630423022828e-05, "loss": 1.3619, "step": 92200 }, { "epoch": 4.98, "learning_rate": 2.5113599480688087e-05, "loss": 1.3987, "step": 92210 }, { "epoch": 4.98, "learning_rate": 2.51108947311479e-05, "loss": 1.3564, "step": 92220 }, { "epoch": 4.98, "learning_rate": 2.5108189981607703e-05, "loss": 1.3892, "step": 92230 }, { "epoch": 4.98, "learning_rate": 2.510548523206751e-05, "loss": 1.4076, "step": 92240 }, { "epoch": 4.98, "learning_rate": 2.5102780482527323e-05, "loss": 1.3656, "step": 92250 }, { "epoch": 4.99, "learning_rate": 2.510007573298713e-05, "loss": 1.3378, "step": 92260 }, { "epoch": 4.99, "learning_rate": 2.5097370983446932e-05, "loss": 1.3972, "step": 92270 }, { "epoch": 4.99, "learning_rate": 2.5094666233906745e-05, "loss": 1.3644, "step": 92280 }, { "epoch": 4.99, "learning_rate": 2.509196148436655e-05, "loss": 1.3397, "step": 92290 }, { "epoch": 4.99, "learning_rate": 2.5089256734826354e-05, "loss": 1.4203, "step": 92300 }, { "epoch": 4.99, "learning_rate": 2.5086551985286167e-05, "loss": 1.3854, "step": 92310 }, { "epoch": 4.99, "learning_rate": 2.5083847235745973e-05, "loss": 1.4289, "step": 92320 }, { "epoch": 4.99, "learning_rate": 2.5081142486205776e-05, "loss": 1.4909, "step": 92330 }, { "epoch": 4.99, "learning_rate": 2.507843773666559e-05, "loss": 1.3362, "step": 92340 }, { "epoch": 4.99, "learning_rate": 2.5075732987125395e-05, "loss": 1.4391, "step": 92350 }, { "epoch": 4.99, "learning_rate": 2.5073028237585198e-05, "loss": 1.3887, "step": 92360 }, { "epoch": 4.99, "learning_rate": 2.5070323488045005e-05, "loss": 1.3101, "step": 92370 }, { "epoch": 4.99, "learning_rate": 2.5067618738504818e-05, "loss": 1.4253, "step": 92380 }, { "epoch": 4.99, "learning_rate": 2.5064913988964624e-05, "loss": 1.3765, "step": 92390 }, { "epoch": 4.99, "learning_rate": 2.5062209239424427e-05, "loss": 1.3808, "step": 92400 }, { "epoch": 4.99, "learning_rate": 2.505950448988424e-05, "loss": 1.3613, "step": 92410 }, { "epoch": 4.99, "learning_rate": 2.5056799740344046e-05, "loss": 1.3899, "step": 92420 }, { "epoch": 4.99, "learning_rate": 2.505409499080385e-05, "loss": 1.429, "step": 92430 }, { "epoch": 4.99, "learning_rate": 2.5051390241263662e-05, "loss": 1.4397, "step": 92440 }, { "epoch": 5.0, "learning_rate": 2.5048685491723468e-05, "loss": 1.36, "step": 92450 }, { "epoch": 5.0, "learning_rate": 2.504598074218327e-05, "loss": 1.3786, "step": 92460 }, { "epoch": 5.0, "learning_rate": 2.5043275992643084e-05, "loss": 1.4025, "step": 92470 }, { "epoch": 5.0, "learning_rate": 2.504057124310289e-05, "loss": 1.3778, "step": 92480 }, { "epoch": 5.0, "learning_rate": 2.5037866493562697e-05, "loss": 1.4109, "step": 92490 }, { "epoch": 5.0, "learning_rate": 2.5035161744022506e-05, "loss": 1.3791, "step": 92500 }, { "epoch": 5.0, "learning_rate": 2.5032456994482313e-05, "loss": 1.4438, "step": 92510 }, { "epoch": 5.0, "learning_rate": 2.502975224494212e-05, "loss": 1.4141, "step": 92520 }, { "epoch": 5.0, "learning_rate": 2.502704749540193e-05, "loss": 1.3398, "step": 92530 } ], "max_steps": 185060, "num_train_epochs": 10, "total_flos": 2.253966288224256e+18, "trial_name": null, "trial_params": null }