diff --git "a/checkpoint-82000/trainer_state.json" "b/checkpoint-82000/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-82000/trainer_state.json" @@ -0,0 +1,49954 @@ +{ + "best_metric": 0.9234974441114349, + "best_model_checkpoint": "./models/bart-base-spelling-nl/checkpoint-82000", + "epoch": 1.752115693053903, + "global_step": 82000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 0.0002999679487179487, + "loss": 2.6278, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002999358974358974, + "loss": 1.157, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002999038461538461, + "loss": 1.0065, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029987179487179485, + "loss": 0.908, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002998397435897436, + "loss": 0.8208, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029980769230769226, + "loss": 0.7505, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 0.000299775641025641, + "loss": 0.6984, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029974358974358973, + "loss": 0.6874, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002997115384615384, + "loss": 0.6423, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029967948717948714, + "loss": 0.6271, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002996474358974359, + "loss": 0.6416, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002996153846153846, + "loss": 0.6166, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002995833333333333, + "loss": 0.5545, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 0.000299551282051282, + "loss": 0.5499, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029951923076923075, + "loss": 0.5659, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002994871794871795, + "loss": 0.5363, + "step": 160 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029945512820512817, + "loss": 0.5235, + "step": 170 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002994230769230769, + "loss": 0.5101, + "step": 180 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029939102564102563, + "loss": 0.4834, + "step": 190 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002993589743589743, + "loss": 0.4867, + "step": 200 + }, + { + "epoch": 0.0, + "learning_rate": 0.00029932692307692304, + "loss": 0.4869, + "step": 210 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002992948717948718, + "loss": 0.4692, + "step": 220 + }, + { + "epoch": 0.0, + "learning_rate": 0.0002992628205128205, + "loss": 0.4768, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002992307692307692, + "loss": 0.4735, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002991987179487179, + "loss": 0.4807, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029916666666666666, + "loss": 0.4424, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029913461538461534, + "loss": 0.4424, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029910256410256407, + "loss": 0.422, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002990705128205128, + "loss": 0.4265, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029903846153846153, + "loss": 0.4281, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002990064102564102, + "loss": 0.4362, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029897435897435895, + "loss": 0.3967, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002989423076923077, + "loss": 0.4068, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002989102564102564, + "loss": 0.413, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002988782051282051, + "loss": 0.4433, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002988461538461538, + "loss": 0.4325, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029881410256410256, + "loss": 0.3948, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029878205128205124, + "loss": 0.3787, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029874999999999997, + "loss": 0.387, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002987179487179487, + "loss": 0.3859, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029868589743589744, + "loss": 0.3685, + "step": 410 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002986538461538461, + "loss": 0.3572, + "step": 420 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029862179487179485, + "loss": 0.3605, + "step": 430 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002985897435897436, + "loss": 0.36, + "step": 440 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029855769230769226, + "loss": 0.3731, + "step": 450 + }, + { + "epoch": 0.01, + "learning_rate": 0.000298525641025641, + "loss": 0.3668, + "step": 460 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029849358974358973, + "loss": 0.3404, + "step": 470 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029846153846153846, + "loss": 0.3389, + "step": 480 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029842948717948714, + "loss": 0.3334, + "step": 490 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029839743589743587, + "loss": 0.3436, + "step": 500 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002983653846153846, + "loss": 0.349, + "step": 510 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029833333333333334, + "loss": 0.3206, + "step": 520 + }, + { + "epoch": 0.01, + "learning_rate": 0.000298301282051282, + "loss": 0.3259, + "step": 530 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029826923076923075, + "loss": 0.3408, + "step": 540 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002982371794871795, + "loss": 0.3446, + "step": 550 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029820512820512816, + "loss": 0.329, + "step": 560 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002981730769230769, + "loss": 0.3219, + "step": 570 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029814102564102563, + "loss": 0.3335, + "step": 580 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029810897435897436, + "loss": 0.3155, + "step": 590 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029807692307692304, + "loss": 0.2962, + "step": 600 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002980448717948718, + "loss": 0.3307, + "step": 610 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002980128205128205, + "loss": 0.3046, + "step": 620 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002979807692307692, + "loss": 0.313, + "step": 630 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002979487179487179, + "loss": 0.3025, + "step": 640 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029791666666666665, + "loss": 0.2968, + "step": 650 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002978846153846154, + "loss": 0.2886, + "step": 660 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029785256410256407, + "loss": 0.3015, + "step": 670 + }, + { + "epoch": 0.01, + "learning_rate": 0.0002978205128205128, + "loss": 0.292, + "step": 680 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029778846153846153, + "loss": 0.2799, + "step": 690 + }, + { + "epoch": 0.01, + "learning_rate": 0.00029775641025641026, + "loss": 0.2982, + "step": 700 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029772435897435894, + "loss": 0.305, + "step": 710 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002976923076923077, + "loss": 0.2928, + "step": 720 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002976602564102564, + "loss": 0.2849, + "step": 730 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002976282051282051, + "loss": 0.2759, + "step": 740 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002975961538461538, + "loss": 0.2719, + "step": 750 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029756410256410256, + "loss": 0.2828, + "step": 760 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002975320512820513, + "loss": 0.2684, + "step": 770 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029749999999999997, + "loss": 0.2789, + "step": 780 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002974679487179487, + "loss": 0.2608, + "step": 790 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029743589743589743, + "loss": 0.2614, + "step": 800 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002974038461538461, + "loss": 0.2596, + "step": 810 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029737179487179485, + "loss": 0.2645, + "step": 820 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002973397435897436, + "loss": 0.2627, + "step": 830 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002973076923076923, + "loss": 0.269, + "step": 840 + }, + { + "epoch": 0.02, + "learning_rate": 0.000297275641025641, + "loss": 0.2682, + "step": 850 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002972435897435897, + "loss": 0.2615, + "step": 860 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029721153846153846, + "loss": 0.2508, + "step": 870 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002971794871794872, + "loss": 0.2564, + "step": 880 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029714743589743587, + "loss": 0.254, + "step": 890 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002971153846153846, + "loss": 0.2636, + "step": 900 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029708333333333334, + "loss": 0.2538, + "step": 910 + }, + { + "epoch": 0.02, + "learning_rate": 0.000297051282051282, + "loss": 0.2799, + "step": 920 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029701923076923075, + "loss": 0.2552, + "step": 930 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002969871794871795, + "loss": 0.2507, + "step": 940 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002969551282051282, + "loss": 0.2411, + "step": 950 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002969230769230769, + "loss": 0.2524, + "step": 960 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002968910256410256, + "loss": 0.245, + "step": 970 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029685897435897436, + "loss": 0.251, + "step": 980 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029682692307692304, + "loss": 0.2361, + "step": 990 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029679487179487177, + "loss": 0.2546, + "step": 1000 + }, + { + "epoch": 0.02, + "eval_cer": 0.9244766564442448, + "eval_loss": 0.18007953464984894, + "eval_runtime": 120.7757, + "eval_samples_per_second": 16.56, + "eval_steps_per_second": 4.14, + "step": 1000 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002967628205128205, + "loss": 0.2559, + "step": 1010 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029673076923076924, + "loss": 0.2355, + "step": 1020 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002966987179487179, + "loss": 0.2397, + "step": 1030 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029666666666666665, + "loss": 0.2283, + "step": 1040 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002966346153846154, + "loss": 0.2424, + "step": 1050 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029660256410256406, + "loss": 0.2362, + "step": 1060 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002965705128205128, + "loss": 0.23, + "step": 1070 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029653846153846153, + "loss": 0.2307, + "step": 1080 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029650641025641026, + "loss": 0.2254, + "step": 1090 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029647435897435894, + "loss": 0.2376, + "step": 1100 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002964423076923077, + "loss": 0.2317, + "step": 1110 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002964102564102564, + "loss": 0.239, + "step": 1120 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029637820512820514, + "loss": 0.2349, + "step": 1130 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002963461538461538, + "loss": 0.2191, + "step": 1140 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029631410256410255, + "loss": 0.2209, + "step": 1150 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002962820512820513, + "loss": 0.2181, + "step": 1160 + }, + { + "epoch": 0.02, + "learning_rate": 0.00029624999999999996, + "loss": 0.2076, + "step": 1170 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002962179487179487, + "loss": 0.2231, + "step": 1180 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029618589743589743, + "loss": 0.2185, + "step": 1190 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029615384615384616, + "loss": 0.2099, + "step": 1200 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029612179487179484, + "loss": 0.208, + "step": 1210 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002960897435897436, + "loss": 0.2063, + "step": 1220 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002960576923076923, + "loss": 0.2101, + "step": 1230 + }, + { + "epoch": 0.03, + "learning_rate": 0.000296025641025641, + "loss": 0.2132, + "step": 1240 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002959935897435897, + "loss": 0.213, + "step": 1250 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029596153846153845, + "loss": 0.217, + "step": 1260 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002959294871794872, + "loss": 0.2207, + "step": 1270 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029589743589743587, + "loss": 0.2056, + "step": 1280 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002958653846153846, + "loss": 0.2072, + "step": 1290 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029583333333333333, + "loss": 0.2003, + "step": 1300 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029580128205128207, + "loss": 0.2253, + "step": 1310 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029576923076923074, + "loss": 0.2112, + "step": 1320 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002957371794871795, + "loss": 0.2121, + "step": 1330 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002957051282051282, + "loss": 0.2227, + "step": 1340 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002956730769230769, + "loss": 0.2015, + "step": 1350 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002956410256410256, + "loss": 0.2103, + "step": 1360 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029560897435897436, + "loss": 0.2088, + "step": 1370 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002955769230769231, + "loss": 0.2029, + "step": 1380 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029554487179487177, + "loss": 0.2012, + "step": 1390 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002955128205128205, + "loss": 0.213, + "step": 1400 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029548076923076923, + "loss": 0.1936, + "step": 1410 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002954487179487179, + "loss": 0.2083, + "step": 1420 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029541666666666665, + "loss": 0.2121, + "step": 1430 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002953846153846154, + "loss": 0.1961, + "step": 1440 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002953525641025641, + "loss": 0.1893, + "step": 1450 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002953205128205128, + "loss": 0.1932, + "step": 1460 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029528846153846147, + "loss": 0.1894, + "step": 1470 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002952564102564102, + "loss": 0.1844, + "step": 1480 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029522435897435894, + "loss": 0.204, + "step": 1490 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029519230769230767, + "loss": 0.2059, + "step": 1500 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029516025641025635, + "loss": 0.1928, + "step": 1510 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002951282051282051, + "loss": 0.2018, + "step": 1520 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002950961538461538, + "loss": 0.2026, + "step": 1530 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029506410256410255, + "loss": 0.1834, + "step": 1540 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029503205128205123, + "loss": 0.1921, + "step": 1550 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029499999999999996, + "loss": 0.1885, + "step": 1560 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002949679487179487, + "loss": 0.2139, + "step": 1570 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029493589743589737, + "loss": 0.2085, + "step": 1580 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002949038461538461, + "loss": 0.1918, + "step": 1590 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029487179487179484, + "loss": 0.1946, + "step": 1600 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029483974358974357, + "loss": 0.1914, + "step": 1610 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029480769230769225, + "loss": 0.19, + "step": 1620 + }, + { + "epoch": 0.03, + "learning_rate": 0.000294775641025641, + "loss": 0.1965, + "step": 1630 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002947435897435897, + "loss": 0.1877, + "step": 1640 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002947115384615384, + "loss": 0.1976, + "step": 1650 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029467948717948713, + "loss": 0.2044, + "step": 1660 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029464743589743586, + "loss": 0.1795, + "step": 1670 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002946153846153846, + "loss": 0.1911, + "step": 1680 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002945833333333333, + "loss": 0.1903, + "step": 1690 + }, + { + "epoch": 0.04, + "learning_rate": 0.000294551282051282, + "loss": 0.1913, + "step": 1700 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029451923076923074, + "loss": 0.1844, + "step": 1710 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002944871794871795, + "loss": 0.1958, + "step": 1720 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029445512820512815, + "loss": 0.1946, + "step": 1730 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002944230769230769, + "loss": 0.1922, + "step": 1740 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002943910256410256, + "loss": 0.1798, + "step": 1750 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002943589743589743, + "loss": 0.1881, + "step": 1760 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029432692307692303, + "loss": 0.1807, + "step": 1770 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029429487179487176, + "loss": 0.1712, + "step": 1780 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002942628205128205, + "loss": 0.1813, + "step": 1790 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002942307692307692, + "loss": 0.1724, + "step": 1800 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002941987179487179, + "loss": 0.1716, + "step": 1810 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029416666666666664, + "loss": 0.179, + "step": 1820 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002941346153846153, + "loss": 0.1872, + "step": 1830 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029410256410256406, + "loss": 0.1861, + "step": 1840 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002940705128205128, + "loss": 0.1931, + "step": 1850 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002940384615384615, + "loss": 0.1892, + "step": 1860 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002940064102564102, + "loss": 0.1741, + "step": 1870 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029397435897435893, + "loss": 0.1776, + "step": 1880 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029394230769230767, + "loss": 0.1815, + "step": 1890 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029391025641025635, + "loss": 0.1748, + "step": 1900 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002938782051282051, + "loss": 0.171, + "step": 1910 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002938461538461538, + "loss": 0.1735, + "step": 1920 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029381410256410255, + "loss": 0.1847, + "step": 1930 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002937820512820512, + "loss": 0.183, + "step": 1940 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029374999999999996, + "loss": 0.17, + "step": 1950 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002937179487179487, + "loss": 0.1717, + "step": 1960 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002936858974358974, + "loss": 0.1767, + "step": 1970 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002936538461538461, + "loss": 0.1688, + "step": 1980 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029362179487179484, + "loss": 0.1699, + "step": 1990 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029358974358974357, + "loss": 0.1646, + "step": 2000 + }, + { + "epoch": 0.04, + "eval_cer": 0.9243467176004287, + "eval_loss": 0.12032673507928848, + "eval_runtime": 120.9209, + "eval_samples_per_second": 16.54, + "eval_steps_per_second": 4.135, + "step": 2000 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029355769230769225, + "loss": 0.1733, + "step": 2010 + }, + { + "epoch": 0.04, + "learning_rate": 0.000293525641025641, + "loss": 0.166, + "step": 2020 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002934935897435897, + "loss": 0.1667, + "step": 2030 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029346153846153845, + "loss": 0.1686, + "step": 2040 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002934294871794871, + "loss": 0.1667, + "step": 2050 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029339743589743586, + "loss": 0.1617, + "step": 2060 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002933653846153846, + "loss": 0.1722, + "step": 2070 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029333333333333327, + "loss": 0.1801, + "step": 2080 + }, + { + "epoch": 0.04, + "learning_rate": 0.000293301282051282, + "loss": 0.1676, + "step": 2090 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029326923076923074, + "loss": 0.1623, + "step": 2100 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029323717948717947, + "loss": 0.179, + "step": 2110 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029320512820512815, + "loss": 0.1712, + "step": 2120 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002931730769230769, + "loss": 0.1586, + "step": 2130 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002931410256410256, + "loss": 0.1627, + "step": 2140 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029310897435897435, + "loss": 0.1705, + "step": 2150 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029307692307692303, + "loss": 0.177, + "step": 2160 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029304487179487176, + "loss": 0.1771, + "step": 2170 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002930128205128205, + "loss": 0.1657, + "step": 2180 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002929807692307692, + "loss": 0.1526, + "step": 2190 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002929487179487179, + "loss": 0.1587, + "step": 2200 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029291666666666664, + "loss": 0.1664, + "step": 2210 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002928846153846154, + "loss": 0.1522, + "step": 2220 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029285256410256405, + "loss": 0.1667, + "step": 2230 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002928205128205128, + "loss": 0.1534, + "step": 2240 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002927884615384615, + "loss": 0.1534, + "step": 2250 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002927564102564102, + "loss": 0.1811, + "step": 2260 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029272435897435893, + "loss": 0.1591, + "step": 2270 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029269230769230766, + "loss": 0.1628, + "step": 2280 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002926602564102564, + "loss": 0.1531, + "step": 2290 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002926282051282051, + "loss": 0.1756, + "step": 2300 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002925961538461538, + "loss": 0.156, + "step": 2310 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029256410256410254, + "loss": 0.1561, + "step": 2320 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002925320512820513, + "loss": 0.1701, + "step": 2330 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029249999999999995, + "loss": 0.1603, + "step": 2340 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002924679487179487, + "loss": 0.1617, + "step": 2350 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002924358974358974, + "loss": 0.1564, + "step": 2360 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002924038461538461, + "loss": 0.1563, + "step": 2370 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029237179487179483, + "loss": 0.1608, + "step": 2380 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029233974358974357, + "loss": 0.1668, + "step": 2390 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002923076923076923, + "loss": 0.1669, + "step": 2400 + }, + { + "epoch": 0.05, + "learning_rate": 0.000292275641025641, + "loss": 0.1666, + "step": 2410 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002922435897435897, + "loss": 0.1652, + "step": 2420 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029221153846153844, + "loss": 0.1611, + "step": 2430 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002921794871794871, + "loss": 0.1526, + "step": 2440 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029214743589743586, + "loss": 0.1654, + "step": 2450 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002921153846153846, + "loss": 0.1521, + "step": 2460 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002920833333333333, + "loss": 0.1506, + "step": 2470 + }, + { + "epoch": 0.05, + "learning_rate": 0.000292051282051282, + "loss": 0.1451, + "step": 2480 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029201923076923073, + "loss": 0.1481, + "step": 2490 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029198717948717947, + "loss": 0.1476, + "step": 2500 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002919551282051282, + "loss": 0.1602, + "step": 2510 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002919230769230769, + "loss": 0.148, + "step": 2520 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002918910256410256, + "loss": 0.1572, + "step": 2530 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029185897435897435, + "loss": 0.1535, + "step": 2540 + }, + { + "epoch": 0.05, + "learning_rate": 0.000291826923076923, + "loss": 0.1559, + "step": 2550 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029179487179487176, + "loss": 0.1476, + "step": 2560 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002917628205128205, + "loss": 0.1577, + "step": 2570 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002917307692307692, + "loss": 0.1479, + "step": 2580 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002916987179487179, + "loss": 0.1602, + "step": 2590 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029166666666666664, + "loss": 0.1626, + "step": 2600 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029163461538461537, + "loss": 0.1541, + "step": 2610 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029160256410256405, + "loss": 0.1705, + "step": 2620 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002915705128205128, + "loss": 0.1526, + "step": 2630 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002915384615384615, + "loss": 0.1562, + "step": 2640 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029150641025641025, + "loss": 0.1431, + "step": 2650 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029147435897435893, + "loss": 0.1519, + "step": 2660 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029144230769230766, + "loss": 0.1596, + "step": 2670 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002914102564102564, + "loss": 0.1527, + "step": 2680 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002913782051282051, + "loss": 0.1551, + "step": 2690 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002913461538461538, + "loss": 0.1425, + "step": 2700 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029131410256410254, + "loss": 0.1483, + "step": 2710 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029128205128205127, + "loss": 0.1541, + "step": 2720 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029124999999999995, + "loss": 0.1504, + "step": 2730 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002912179487179487, + "loss": 0.1592, + "step": 2740 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002911858974358974, + "loss": 0.1679, + "step": 2750 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029115384615384615, + "loss": 0.1543, + "step": 2760 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029112179487179483, + "loss": 0.1557, + "step": 2770 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029108974358974356, + "loss": 0.1565, + "step": 2780 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002910576923076923, + "loss": 0.1535, + "step": 2790 + }, + { + "epoch": 0.06, + "learning_rate": 0.000291025641025641, + "loss": 0.1543, + "step": 2800 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002909935897435897, + "loss": 0.1473, + "step": 2810 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029096153846153844, + "loss": 0.1367, + "step": 2820 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002909294871794872, + "loss": 0.1521, + "step": 2830 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029089743589743585, + "loss": 0.149, + "step": 2840 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002908653846153846, + "loss": 0.1465, + "step": 2850 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002908333333333333, + "loss": 0.1421, + "step": 2860 + }, + { + "epoch": 0.06, + "learning_rate": 0.000290801282051282, + "loss": 0.1551, + "step": 2870 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029076923076923073, + "loss": 0.1561, + "step": 2880 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029073717948717946, + "loss": 0.1474, + "step": 2890 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002907051282051282, + "loss": 0.1498, + "step": 2900 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002906730769230769, + "loss": 0.1522, + "step": 2910 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002906410256410256, + "loss": 0.1382, + "step": 2920 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029060897435897434, + "loss": 0.1571, + "step": 2930 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002905769230769231, + "loss": 0.1455, + "step": 2940 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029054487179487175, + "loss": 0.155, + "step": 2950 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002905128205128205, + "loss": 0.1432, + "step": 2960 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002904807692307692, + "loss": 0.1503, + "step": 2970 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002904487179487179, + "loss": 0.1506, + "step": 2980 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029041666666666663, + "loss": 0.1474, + "step": 2990 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029038461538461537, + "loss": 0.1456, + "step": 3000 + }, + { + "epoch": 0.06, + "eval_cer": 0.9242091352952118, + "eval_loss": 0.1016402468085289, + "eval_runtime": 120.6917, + "eval_samples_per_second": 16.571, + "eval_steps_per_second": 4.143, + "step": 3000 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002903525641025641, + "loss": 0.1443, + "step": 3010 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002903205128205128, + "loss": 0.1382, + "step": 3020 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002902884615384615, + "loss": 0.1381, + "step": 3030 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029025641025641024, + "loss": 0.1392, + "step": 3040 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002902243589743589, + "loss": 0.1431, + "step": 3050 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029019230769230766, + "loss": 0.1517, + "step": 3060 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002901602564102564, + "loss": 0.1446, + "step": 3070 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002901282051282051, + "loss": 0.1528, + "step": 3080 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002900961538461538, + "loss": 0.1424, + "step": 3090 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029006410256410254, + "loss": 0.1375, + "step": 3100 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029003205128205127, + "loss": 0.1491, + "step": 3110 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029, + "loss": 0.1278, + "step": 3120 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002899679487179487, + "loss": 0.143, + "step": 3130 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002899358974358974, + "loss": 0.132, + "step": 3140 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028990384615384615, + "loss": 0.1347, + "step": 3150 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002898717948717948, + "loss": 0.1465, + "step": 3160 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028983974358974356, + "loss": 0.1404, + "step": 3170 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002898076923076923, + "loss": 0.135, + "step": 3180 + }, + { + "epoch": 0.07, + "learning_rate": 0.000289775641025641, + "loss": 0.1392, + "step": 3190 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002897435897435897, + "loss": 0.1323, + "step": 3200 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028971153846153844, + "loss": 0.1389, + "step": 3210 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028967948717948717, + "loss": 0.1374, + "step": 3220 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028964743589743585, + "loss": 0.1351, + "step": 3230 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002896153846153846, + "loss": 0.1539, + "step": 3240 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002895833333333333, + "loss": 0.1331, + "step": 3250 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028955128205128205, + "loss": 0.1315, + "step": 3260 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028951923076923073, + "loss": 0.1351, + "step": 3270 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028948717948717946, + "loss": 0.1382, + "step": 3280 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002894551282051282, + "loss": 0.1322, + "step": 3290 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028942307692307693, + "loss": 0.1238, + "step": 3300 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002893910256410256, + "loss": 0.1405, + "step": 3310 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028935897435897434, + "loss": 0.1403, + "step": 3320 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028932692307692307, + "loss": 0.1266, + "step": 3330 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028929487179487175, + "loss": 0.1374, + "step": 3340 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002892628205128205, + "loss": 0.1355, + "step": 3350 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002892307692307692, + "loss": 0.1335, + "step": 3360 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028919871794871795, + "loss": 0.1348, + "step": 3370 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028916666666666663, + "loss": 0.134, + "step": 3380 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028913461538461536, + "loss": 0.1574, + "step": 3390 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002891025641025641, + "loss": 0.1397, + "step": 3400 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002890705128205128, + "loss": 0.1343, + "step": 3410 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002890384615384615, + "loss": 0.1466, + "step": 3420 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028900641025641024, + "loss": 0.1326, + "step": 3430 + }, + { + "epoch": 0.07, + "learning_rate": 0.000288974358974359, + "loss": 0.1483, + "step": 3440 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028894230769230765, + "loss": 0.1357, + "step": 3450 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002889102564102564, + "loss": 0.1434, + "step": 3460 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002888782051282051, + "loss": 0.1321, + "step": 3470 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028884615384615385, + "loss": 0.1263, + "step": 3480 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028881410256410253, + "loss": 0.1223, + "step": 3490 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028878205128205127, + "loss": 0.1306, + "step": 3500 + }, + { + "epoch": 0.07, + "learning_rate": 0.00028875, + "loss": 0.1338, + "step": 3510 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002887179487179487, + "loss": 0.1225, + "step": 3520 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002886858974358974, + "loss": 0.133, + "step": 3530 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028865384615384614, + "loss": 0.1297, + "step": 3540 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002886217948717949, + "loss": 0.1364, + "step": 3550 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028858974358974356, + "loss": 0.1297, + "step": 3560 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002885576923076923, + "loss": 0.1275, + "step": 3570 + }, + { + "epoch": 0.08, + "learning_rate": 0.000288525641025641, + "loss": 0.138, + "step": 3580 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002884935897435897, + "loss": 0.1316, + "step": 3590 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028846153846153843, + "loss": 0.1388, + "step": 3600 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028842948717948717, + "loss": 0.1275, + "step": 3610 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002883974358974359, + "loss": 0.1278, + "step": 3620 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002883653846153846, + "loss": 0.1242, + "step": 3630 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002883333333333333, + "loss": 0.1226, + "step": 3640 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028830128205128205, + "loss": 0.1389, + "step": 3650 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002882692307692308, + "loss": 0.1262, + "step": 3660 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028823717948717946, + "loss": 0.128, + "step": 3670 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002882051282051282, + "loss": 0.1406, + "step": 3680 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002881730769230769, + "loss": 0.1339, + "step": 3690 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002881410256410256, + "loss": 0.1362, + "step": 3700 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028810897435897434, + "loss": 0.1336, + "step": 3710 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028807692307692307, + "loss": 0.1347, + "step": 3720 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002880448717948718, + "loss": 0.1384, + "step": 3730 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002880128205128205, + "loss": 0.1281, + "step": 3740 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002879807692307692, + "loss": 0.1552, + "step": 3750 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028794871794871795, + "loss": 0.1394, + "step": 3760 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002879166666666666, + "loss": 0.129, + "step": 3770 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028788461538461536, + "loss": 0.1322, + "step": 3780 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002878525641025641, + "loss": 0.1218, + "step": 3790 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002878205128205128, + "loss": 0.1232, + "step": 3800 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002877884615384615, + "loss": 0.1322, + "step": 3810 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028775641025641024, + "loss": 0.1407, + "step": 3820 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028772435897435897, + "loss": 0.1352, + "step": 3830 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028769230769230765, + "loss": 0.1304, + "step": 3840 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002876602564102564, + "loss": 0.1333, + "step": 3850 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002876282051282051, + "loss": 0.1271, + "step": 3860 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028759615384615385, + "loss": 0.1322, + "step": 3870 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028756410256410253, + "loss": 0.1331, + "step": 3880 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028753205128205126, + "loss": 0.123, + "step": 3890 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002875, + "loss": 0.1334, + "step": 3900 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028746794871794873, + "loss": 0.1405, + "step": 3910 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002874358974358974, + "loss": 0.1253, + "step": 3920 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028740384615384614, + "loss": 0.1235, + "step": 3930 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002873717948717949, + "loss": 0.1227, + "step": 3940 + }, + { + "epoch": 0.08, + "learning_rate": 0.00028733974358974355, + "loss": 0.1259, + "step": 3950 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002873076923076923, + "loss": 0.126, + "step": 3960 + }, + { + "epoch": 0.08, + "learning_rate": 0.000287275641025641, + "loss": 0.1211, + "step": 3970 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028724358974358975, + "loss": 0.1285, + "step": 3980 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028721153846153843, + "loss": 0.1408, + "step": 3990 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028717948717948716, + "loss": 0.1204, + "step": 4000 + }, + { + "epoch": 0.09, + "eval_cer": 0.9241607267063391, + "eval_loss": 0.08493528515100479, + "eval_runtime": 120.7235, + "eval_samples_per_second": 16.567, + "eval_steps_per_second": 4.142, + "step": 4000 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002871474358974359, + "loss": 0.1177, + "step": 4010 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002871153846153846, + "loss": 0.1175, + "step": 4020 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002870833333333333, + "loss": 0.1325, + "step": 4030 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028705128205128204, + "loss": 0.1194, + "step": 4040 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002870192307692308, + "loss": 0.1292, + "step": 4050 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028698717948717945, + "loss": 0.1283, + "step": 4060 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002869551282051282, + "loss": 0.1168, + "step": 4070 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002869230769230769, + "loss": 0.1144, + "step": 4080 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028689102564102565, + "loss": 0.122, + "step": 4090 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028685897435897433, + "loss": 0.1266, + "step": 4100 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028682692307692307, + "loss": 0.1289, + "step": 4110 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002867948717948718, + "loss": 0.1197, + "step": 4120 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002867628205128205, + "loss": 0.1273, + "step": 4130 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002867307692307692, + "loss": 0.1481, + "step": 4140 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028669871794871794, + "loss": 0.1274, + "step": 4150 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002866666666666667, + "loss": 0.1134, + "step": 4160 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028663461538461536, + "loss": 0.1217, + "step": 4170 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002866025641025641, + "loss": 0.1224, + "step": 4180 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002865705128205128, + "loss": 0.1191, + "step": 4190 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002865384615384615, + "loss": 0.1321, + "step": 4200 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028650641025641024, + "loss": 0.1261, + "step": 4210 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028647435897435897, + "loss": 0.1301, + "step": 4220 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002864423076923077, + "loss": 0.1253, + "step": 4230 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002864102564102564, + "loss": 0.1238, + "step": 4240 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002863782051282051, + "loss": 0.1178, + "step": 4250 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028634615384615385, + "loss": 0.1298, + "step": 4260 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002863141025641026, + "loss": 0.1219, + "step": 4270 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028628205128205126, + "loss": 0.1167, + "step": 4280 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028625, + "loss": 0.1227, + "step": 4290 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002862179487179487, + "loss": 0.116, + "step": 4300 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002861858974358974, + "loss": 0.1136, + "step": 4310 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028615384615384614, + "loss": 0.1157, + "step": 4320 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028612179487179487, + "loss": 0.1223, + "step": 4330 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002860897435897436, + "loss": 0.1206, + "step": 4340 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002860576923076923, + "loss": 0.1142, + "step": 4350 + }, + { + "epoch": 0.09, + "learning_rate": 0.000286025641025641, + "loss": 0.1193, + "step": 4360 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028599358974358975, + "loss": 0.1179, + "step": 4370 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028596153846153843, + "loss": 0.1276, + "step": 4380 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028592948717948716, + "loss": 0.1178, + "step": 4390 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028589743589743584, + "loss": 0.1198, + "step": 4400 + }, + { + "epoch": 0.09, + "learning_rate": 0.00028586538461538457, + "loss": 0.116, + "step": 4410 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002858333333333333, + "loss": 0.1148, + "step": 4420 + }, + { + "epoch": 0.09, + "learning_rate": 0.000285801282051282, + "loss": 0.1139, + "step": 4430 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002857692307692307, + "loss": 0.1098, + "step": 4440 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028573717948717945, + "loss": 0.1153, + "step": 4450 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002857051282051282, + "loss": 0.129, + "step": 4460 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028567307692307686, + "loss": 0.1233, + "step": 4470 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002856410256410256, + "loss": 0.121, + "step": 4480 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028560897435897433, + "loss": 0.1289, + "step": 4490 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028557692307692306, + "loss": 0.1177, + "step": 4500 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028554487179487174, + "loss": 0.1118, + "step": 4510 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002855128205128205, + "loss": 0.1073, + "step": 4520 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002854807692307692, + "loss": 0.1192, + "step": 4530 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002854487179487179, + "loss": 0.1046, + "step": 4540 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002854166666666666, + "loss": 0.1136, + "step": 4550 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028538461538461535, + "loss": 0.117, + "step": 4560 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002853525641025641, + "loss": 0.1186, + "step": 4570 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028532051282051277, + "loss": 0.1157, + "step": 4580 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002852884615384615, + "loss": 0.114, + "step": 4590 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028525641025641023, + "loss": 0.1098, + "step": 4600 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002852243589743589, + "loss": 0.1109, + "step": 4610 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028519230769230764, + "loss": 0.1207, + "step": 4620 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002851602564102564, + "loss": 0.1128, + "step": 4630 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002851282051282051, + "loss": 0.1069, + "step": 4640 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002850961538461538, + "loss": 0.1123, + "step": 4650 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002850641025641025, + "loss": 0.1239, + "step": 4660 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028503205128205126, + "loss": 0.1202, + "step": 4670 + }, + { + "epoch": 0.1, + "learning_rate": 0.000285, + "loss": 0.118, + "step": 4680 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028496794871794867, + "loss": 0.1193, + "step": 4690 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002849358974358974, + "loss": 0.1178, + "step": 4700 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028490384615384613, + "loss": 0.1163, + "step": 4710 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002848717948717948, + "loss": 0.119, + "step": 4720 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028483974358974355, + "loss": 0.1123, + "step": 4730 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002848076923076923, + "loss": 0.1134, + "step": 4740 + }, + { + "epoch": 0.1, + "learning_rate": 0.000284775641025641, + "loss": 0.134, + "step": 4750 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002847435897435897, + "loss": 0.1201, + "step": 4760 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002847115384615384, + "loss": 0.1136, + "step": 4770 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028467948717948716, + "loss": 0.1218, + "step": 4780 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028464743589743584, + "loss": 0.1155, + "step": 4790 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028461538461538457, + "loss": 0.1138, + "step": 4800 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002845833333333333, + "loss": 0.1225, + "step": 4810 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028455128205128204, + "loss": 0.1173, + "step": 4820 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002845192307692307, + "loss": 0.1097, + "step": 4830 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028448717948717945, + "loss": 0.1213, + "step": 4840 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002844551282051282, + "loss": 0.1087, + "step": 4850 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028442307692307686, + "loss": 0.1106, + "step": 4860 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002843910256410256, + "loss": 0.1243, + "step": 4870 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002843589743589743, + "loss": 0.1245, + "step": 4880 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028432692307692306, + "loss": 0.113, + "step": 4890 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028429487179487174, + "loss": 0.1167, + "step": 4900 + }, + { + "epoch": 0.1, + "learning_rate": 0.00028426282051282047, + "loss": 0.1074, + "step": 4910 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002842307692307692, + "loss": 0.1169, + "step": 4920 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028419871794871794, + "loss": 0.1092, + "step": 4930 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002841666666666666, + "loss": 0.1032, + "step": 4940 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028413461538461535, + "loss": 0.1087, + "step": 4950 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002841025641025641, + "loss": 0.105, + "step": 4960 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028407051282051276, + "loss": 0.1154, + "step": 4970 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002840384615384615, + "loss": 0.1153, + "step": 4980 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028400641025641023, + "loss": 0.1081, + "step": 4990 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028397435897435896, + "loss": 0.1226, + "step": 5000 + }, + { + "epoch": 0.11, + "eval_cer": 0.9241029761090874, + "eval_loss": 0.07362516224384308, + "eval_runtime": 120.7851, + "eval_samples_per_second": 16.558, + "eval_steps_per_second": 4.14, + "step": 5000 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028394230769230764, + "loss": 0.1132, + "step": 5010 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002839102564102564, + "loss": 0.1129, + "step": 5020 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002838782051282051, + "loss": 0.1206, + "step": 5030 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002838461538461538, + "loss": 0.1108, + "step": 5040 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002838141025641025, + "loss": 0.1186, + "step": 5050 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028378205128205125, + "loss": 0.1053, + "step": 5060 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028375, + "loss": 0.1047, + "step": 5070 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028371794871794866, + "loss": 0.111, + "step": 5080 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002836858974358974, + "loss": 0.1148, + "step": 5090 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028365384615384613, + "loss": 0.1101, + "step": 5100 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028362179487179486, + "loss": 0.1032, + "step": 5110 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028358974358974354, + "loss": 0.1159, + "step": 5120 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002835576923076923, + "loss": 0.0991, + "step": 5130 + }, + { + "epoch": 0.11, + "learning_rate": 0.000283525641025641, + "loss": 0.1036, + "step": 5140 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002834935897435897, + "loss": 0.1113, + "step": 5150 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002834615384615384, + "loss": 0.1191, + "step": 5160 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028342948717948715, + "loss": 0.1214, + "step": 5170 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002833974358974359, + "loss": 0.1119, + "step": 5180 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028336538461538457, + "loss": 0.1297, + "step": 5190 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002833333333333333, + "loss": 0.1116, + "step": 5200 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028330128205128203, + "loss": 0.1193, + "step": 5210 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002832692307692307, + "loss": 0.1217, + "step": 5220 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028323717948717944, + "loss": 0.1151, + "step": 5230 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002832051282051282, + "loss": 0.1042, + "step": 5240 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002831730769230769, + "loss": 0.1038, + "step": 5250 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002831410256410256, + "loss": 0.1012, + "step": 5260 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002831089743589743, + "loss": 0.1179, + "step": 5270 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028307692307692306, + "loss": 0.0992, + "step": 5280 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002830448717948718, + "loss": 0.107, + "step": 5290 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028301282051282047, + "loss": 0.115, + "step": 5300 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002829807692307692, + "loss": 0.1121, + "step": 5310 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028294871794871793, + "loss": 0.1178, + "step": 5320 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002829166666666666, + "loss": 0.114, + "step": 5330 + }, + { + "epoch": 0.11, + "learning_rate": 0.00028288461538461535, + "loss": 0.1009, + "step": 5340 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002828525641025641, + "loss": 0.1223, + "step": 5350 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002828205128205128, + "loss": 0.1103, + "step": 5360 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002827884615384615, + "loss": 0.1102, + "step": 5370 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002827564102564102, + "loss": 0.1115, + "step": 5380 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028272435897435896, + "loss": 0.1045, + "step": 5390 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028269230769230764, + "loss": 0.103, + "step": 5400 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028266025641025637, + "loss": 0.1198, + "step": 5410 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002826282051282051, + "loss": 0.1028, + "step": 5420 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028259615384615384, + "loss": 0.1032, + "step": 5430 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002825641025641025, + "loss": 0.107, + "step": 5440 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028253205128205125, + "loss": 0.1196, + "step": 5450 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002825, + "loss": 0.1112, + "step": 5460 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002824679487179487, + "loss": 0.1148, + "step": 5470 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002824358974358974, + "loss": 0.116, + "step": 5480 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028240384615384613, + "loss": 0.1095, + "step": 5490 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028237179487179486, + "loss": 0.1079, + "step": 5500 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028233974358974354, + "loss": 0.1059, + "step": 5510 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028230769230769227, + "loss": 0.1045, + "step": 5520 + }, + { + "epoch": 0.12, + "learning_rate": 0.000282275641025641, + "loss": 0.1043, + "step": 5530 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028224358974358974, + "loss": 0.105, + "step": 5540 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002822115384615384, + "loss": 0.1148, + "step": 5550 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028217948717948715, + "loss": 0.1002, + "step": 5560 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002821474358974359, + "loss": 0.1169, + "step": 5570 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028211538461538456, + "loss": 0.1129, + "step": 5580 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002820833333333333, + "loss": 0.1072, + "step": 5590 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028205128205128203, + "loss": 0.1027, + "step": 5600 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028201923076923076, + "loss": 0.1105, + "step": 5610 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028198717948717944, + "loss": 0.1151, + "step": 5620 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002819551282051282, + "loss": 0.107, + "step": 5630 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002819230769230769, + "loss": 0.104, + "step": 5640 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028189102564102564, + "loss": 0.1107, + "step": 5650 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002818589743589743, + "loss": 0.1085, + "step": 5660 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028182692307692305, + "loss": 0.1154, + "step": 5670 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002817948717948718, + "loss": 0.1086, + "step": 5680 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028176282051282047, + "loss": 0.1092, + "step": 5690 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002817307692307692, + "loss": 0.1157, + "step": 5700 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028169871794871793, + "loss": 0.1021, + "step": 5710 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028166666666666666, + "loss": 0.1098, + "step": 5720 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028163461538461534, + "loss": 0.1053, + "step": 5730 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002816025641025641, + "loss": 0.103, + "step": 5740 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002815705128205128, + "loss": 0.1044, + "step": 5750 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002815384615384615, + "loss": 0.1093, + "step": 5760 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002815064102564102, + "loss": 0.0954, + "step": 5770 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028147435897435896, + "loss": 0.1024, + "step": 5780 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002814423076923077, + "loss": 0.1009, + "step": 5790 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028141025641025637, + "loss": 0.1055, + "step": 5800 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002813782051282051, + "loss": 0.1016, + "step": 5810 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028134615384615383, + "loss": 0.1093, + "step": 5820 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002813141025641025, + "loss": 0.1053, + "step": 5830 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028128205128205125, + "loss": 0.1146, + "step": 5840 + }, + { + "epoch": 0.12, + "learning_rate": 0.00028125, + "loss": 0.1191, + "step": 5850 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002812179487179487, + "loss": 0.1072, + "step": 5860 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002811858974358974, + "loss": 0.1002, + "step": 5870 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002811538461538461, + "loss": 0.1113, + "step": 5880 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028112179487179486, + "loss": 0.1102, + "step": 5890 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002810897435897436, + "loss": 0.1001, + "step": 5900 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028105769230769227, + "loss": 0.1052, + "step": 5910 + }, + { + "epoch": 0.13, + "learning_rate": 0.000281025641025641, + "loss": 0.1008, + "step": 5920 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028099358974358974, + "loss": 0.0966, + "step": 5930 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002809615384615384, + "loss": 0.0967, + "step": 5940 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028092948717948715, + "loss": 0.1029, + "step": 5950 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002808974358974359, + "loss": 0.1139, + "step": 5960 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002808653846153846, + "loss": 0.0923, + "step": 5970 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002808333333333333, + "loss": 0.1078, + "step": 5980 + }, + { + "epoch": 0.13, + "learning_rate": 0.000280801282051282, + "loss": 0.1012, + "step": 5990 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028076923076923076, + "loss": 0.1049, + "step": 6000 + }, + { + "epoch": 0.13, + "eval_cer": 0.9239993647434303, + "eval_loss": 0.06804803013801575, + "eval_runtime": 120.9217, + "eval_samples_per_second": 16.54, + "eval_steps_per_second": 4.135, + "step": 6000 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028073717948717944, + "loss": 0.1127, + "step": 6010 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028070512820512817, + "loss": 0.1032, + "step": 6020 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002806730769230769, + "loss": 0.1056, + "step": 6030 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028064102564102564, + "loss": 0.109, + "step": 6040 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002806089743589743, + "loss": 0.1024, + "step": 6050 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028057692307692305, + "loss": 0.0995, + "step": 6060 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002805448717948718, + "loss": 0.1101, + "step": 6070 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002805128205128205, + "loss": 0.1071, + "step": 6080 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002804807692307692, + "loss": 0.1043, + "step": 6090 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028044871794871793, + "loss": 0.1137, + "step": 6100 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028041666666666666, + "loss": 0.1, + "step": 6110 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028038461538461534, + "loss": 0.1095, + "step": 6120 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002803525641025641, + "loss": 0.104, + "step": 6130 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002803205128205128, + "loss": 0.1091, + "step": 6140 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028028846153846154, + "loss": 0.1015, + "step": 6150 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002802564102564102, + "loss": 0.1019, + "step": 6160 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028022435897435895, + "loss": 0.0987, + "step": 6170 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002801923076923077, + "loss": 0.1036, + "step": 6180 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028016025641025636, + "loss": 0.1028, + "step": 6190 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002801282051282051, + "loss": 0.0961, + "step": 6200 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028009615384615383, + "loss": 0.0934, + "step": 6210 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028006410256410256, + "loss": 0.0967, + "step": 6220 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028003205128205124, + "loss": 0.1017, + "step": 6230 + }, + { + "epoch": 0.13, + "learning_rate": 0.00028, + "loss": 0.0958, + "step": 6240 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002799679487179487, + "loss": 0.0969, + "step": 6250 + }, + { + "epoch": 0.13, + "learning_rate": 0.00027993589743589744, + "loss": 0.0998, + "step": 6260 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002799038461538461, + "loss": 0.0941, + "step": 6270 + }, + { + "epoch": 0.13, + "learning_rate": 0.00027987179487179485, + "loss": 0.1064, + "step": 6280 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002798397435897436, + "loss": 0.1022, + "step": 6290 + }, + { + "epoch": 0.13, + "learning_rate": 0.00027980769230769227, + "loss": 0.1025, + "step": 6300 + }, + { + "epoch": 0.13, + "learning_rate": 0.000279775641025641, + "loss": 0.1061, + "step": 6310 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027974358974358973, + "loss": 0.0987, + "step": 6320 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027971153846153847, + "loss": 0.1152, + "step": 6330 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027967948717948714, + "loss": 0.1073, + "step": 6340 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002796474358974359, + "loss": 0.1013, + "step": 6350 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002796153846153846, + "loss": 0.097, + "step": 6360 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002795833333333333, + "loss": 0.1134, + "step": 6370 + }, + { + "epoch": 0.14, + "learning_rate": 0.000279551282051282, + "loss": 0.1152, + "step": 6380 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027951923076923076, + "loss": 0.1018, + "step": 6390 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002794871794871795, + "loss": 0.1076, + "step": 6400 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027945512820512817, + "loss": 0.105, + "step": 6410 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002794230769230769, + "loss": 0.1001, + "step": 6420 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027939102564102563, + "loss": 0.1051, + "step": 6430 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027935897435897437, + "loss": 0.1067, + "step": 6440 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027932692307692305, + "loss": 0.1068, + "step": 6450 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002792948717948718, + "loss": 0.1091, + "step": 6460 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002792628205128205, + "loss": 0.1051, + "step": 6470 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002792307692307692, + "loss": 0.1019, + "step": 6480 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002791987179487179, + "loss": 0.1235, + "step": 6490 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027916666666666666, + "loss": 0.1066, + "step": 6500 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002791346153846154, + "loss": 0.1047, + "step": 6510 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027910256410256407, + "loss": 0.1064, + "step": 6520 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002790705128205128, + "loss": 0.1114, + "step": 6530 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027903846153846154, + "loss": 0.0973, + "step": 6540 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002790064102564102, + "loss": 0.1047, + "step": 6550 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027897435897435895, + "loss": 0.1182, + "step": 6560 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002789423076923077, + "loss": 0.1329, + "step": 6570 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002789102564102564, + "loss": 0.1087, + "step": 6580 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002788782051282051, + "loss": 0.0962, + "step": 6590 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002788461538461538, + "loss": 0.1056, + "step": 6600 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027881410256410256, + "loss": 0.1092, + "step": 6610 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002787820512820513, + "loss": 0.0998, + "step": 6620 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027874999999999997, + "loss": 0.1022, + "step": 6630 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002787179487179487, + "loss": 0.1024, + "step": 6640 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027868589743589744, + "loss": 0.1019, + "step": 6650 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002786538461538461, + "loss": 0.1027, + "step": 6660 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027862179487179485, + "loss": 0.0946, + "step": 6670 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002785897435897436, + "loss": 0.1142, + "step": 6680 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002785576923076923, + "loss": 0.0941, + "step": 6690 + }, + { + "epoch": 0.14, + "learning_rate": 0.000278525641025641, + "loss": 0.11, + "step": 6700 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027849358974358973, + "loss": 0.1105, + "step": 6710 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027846153846153846, + "loss": 0.1218, + "step": 6720 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027842948717948714, + "loss": 0.1046, + "step": 6730 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002783974358974359, + "loss": 0.1059, + "step": 6740 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002783653846153846, + "loss": 0.1025, + "step": 6750 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027833333333333334, + "loss": 0.1057, + "step": 6760 + }, + { + "epoch": 0.14, + "learning_rate": 0.000278301282051282, + "loss": 0.1143, + "step": 6770 + }, + { + "epoch": 0.14, + "learning_rate": 0.00027826923076923075, + "loss": 0.1022, + "step": 6780 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002782371794871795, + "loss": 0.0969, + "step": 6790 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027820512820512816, + "loss": 0.0991, + "step": 6800 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002781730769230769, + "loss": 0.1034, + "step": 6810 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027814102564102563, + "loss": 0.0933, + "step": 6820 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027810897435897436, + "loss": 0.0967, + "step": 6830 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027807692307692304, + "loss": 0.0931, + "step": 6840 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002780448717948718, + "loss": 0.1012, + "step": 6850 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002780128205128205, + "loss": 0.0996, + "step": 6860 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027798076923076924, + "loss": 0.1042, + "step": 6870 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002779487179487179, + "loss": 0.0988, + "step": 6880 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027791666666666665, + "loss": 0.0944, + "step": 6890 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002778846153846154, + "loss": 0.1016, + "step": 6900 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027785256410256407, + "loss": 0.1067, + "step": 6910 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002778205128205128, + "loss": 0.0987, + "step": 6920 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027778846153846153, + "loss": 0.1006, + "step": 6930 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027775641025641027, + "loss": 0.1028, + "step": 6940 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027772435897435895, + "loss": 0.103, + "step": 6950 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002776923076923077, + "loss": 0.0982, + "step": 6960 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002776602564102564, + "loss": 0.0977, + "step": 6970 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002776282051282051, + "loss": 0.0993, + "step": 6980 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002775961538461538, + "loss": 0.1105, + "step": 6990 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027756410256410256, + "loss": 0.1071, + "step": 7000 + }, + { + "epoch": 0.15, + "eval_cer": 0.9240571153406818, + "eval_loss": 0.06707759946584702, + "eval_runtime": 120.9953, + "eval_samples_per_second": 16.53, + "eval_steps_per_second": 4.132, + "step": 7000 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002775320512820513, + "loss": 0.0966, + "step": 7010 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027749999999999997, + "loss": 0.1027, + "step": 7020 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002774679487179487, + "loss": 0.0946, + "step": 7030 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027743589743589744, + "loss": 0.0965, + "step": 7040 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027740384615384617, + "loss": 0.0872, + "step": 7050 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027737179487179485, + "loss": 0.0945, + "step": 7060 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002773397435897436, + "loss": 0.0976, + "step": 7070 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002773076923076923, + "loss": 0.0946, + "step": 7080 + }, + { + "epoch": 0.15, + "learning_rate": 0.000277275641025641, + "loss": 0.096, + "step": 7090 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002772435897435897, + "loss": 0.0983, + "step": 7100 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027721153846153846, + "loss": 0.0982, + "step": 7110 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002771794871794872, + "loss": 0.0969, + "step": 7120 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027714743589743587, + "loss": 0.0927, + "step": 7130 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002771153846153846, + "loss": 0.0924, + "step": 7140 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027708333333333334, + "loss": 0.0933, + "step": 7150 + }, + { + "epoch": 0.15, + "learning_rate": 0.000277051282051282, + "loss": 0.1071, + "step": 7160 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027701923076923075, + "loss": 0.0922, + "step": 7170 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002769871794871795, + "loss": 0.1024, + "step": 7180 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002769551282051282, + "loss": 0.0947, + "step": 7190 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002769230769230769, + "loss": 0.0942, + "step": 7200 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027689102564102563, + "loss": 0.1059, + "step": 7210 + }, + { + "epoch": 0.15, + "learning_rate": 0.00027685897435897436, + "loss": 0.1031, + "step": 7220 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002768269230769231, + "loss": 0.0973, + "step": 7230 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002767948717948718, + "loss": 0.0943, + "step": 7240 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002767628205128205, + "loss": 0.1041, + "step": 7250 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027673076923076924, + "loss": 0.1025, + "step": 7260 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002766987179487179, + "loss": 0.093, + "step": 7270 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027666666666666665, + "loss": 0.0867, + "step": 7280 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002766346153846154, + "loss": 0.0933, + "step": 7290 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002766025641025641, + "loss": 0.0991, + "step": 7300 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002765705128205128, + "loss": 0.1, + "step": 7310 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002765384615384615, + "loss": 0.0954, + "step": 7320 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002765064102564102, + "loss": 0.0907, + "step": 7330 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027647435897435894, + "loss": 0.1009, + "step": 7340 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002764423076923077, + "loss": 0.099, + "step": 7350 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027641025641025635, + "loss": 0.1016, + "step": 7360 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002763782051282051, + "loss": 0.0926, + "step": 7370 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002763461538461538, + "loss": 0.0988, + "step": 7380 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002763141025641025, + "loss": 0.0943, + "step": 7390 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027628205128205123, + "loss": 0.0967, + "step": 7400 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027624999999999997, + "loss": 0.0876, + "step": 7410 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002762179487179487, + "loss": 0.0913, + "step": 7420 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002761858974358974, + "loss": 0.0913, + "step": 7430 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002761538461538461, + "loss": 0.0957, + "step": 7440 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027612179487179484, + "loss": 0.0832, + "step": 7450 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002760897435897436, + "loss": 0.0886, + "step": 7460 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027605769230769226, + "loss": 0.0846, + "step": 7470 + }, + { + "epoch": 0.16, + "learning_rate": 0.000276025641025641, + "loss": 0.0919, + "step": 7480 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002759935897435897, + "loss": 0.0886, + "step": 7490 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002759615384615384, + "loss": 0.0934, + "step": 7500 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027592948717948713, + "loss": 0.1058, + "step": 7510 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027589743589743587, + "loss": 0.1076, + "step": 7520 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002758653846153846, + "loss": 0.1056, + "step": 7530 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002758333333333333, + "loss": 0.1118, + "step": 7540 + }, + { + "epoch": 0.16, + "learning_rate": 0.000275801282051282, + "loss": 0.1025, + "step": 7550 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027576923076923075, + "loss": 0.0936, + "step": 7560 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002757371794871794, + "loss": 0.0922, + "step": 7570 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027570512820512816, + "loss": 0.1047, + "step": 7580 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002756730769230769, + "loss": 0.0993, + "step": 7590 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002756410256410256, + "loss": 0.0943, + "step": 7600 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002756089743589743, + "loss": 0.094, + "step": 7610 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027557692307692304, + "loss": 0.0925, + "step": 7620 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027554487179487177, + "loss": 0.0913, + "step": 7630 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027551282051282045, + "loss": 0.1011, + "step": 7640 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002754807692307692, + "loss": 0.1103, + "step": 7650 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002754487179487179, + "loss": 0.0992, + "step": 7660 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027541666666666665, + "loss": 0.0886, + "step": 7670 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027538461538461533, + "loss": 0.0998, + "step": 7680 + }, + { + "epoch": 0.16, + "learning_rate": 0.00027535256410256406, + "loss": 0.0872, + "step": 7690 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002753205128205128, + "loss": 0.1016, + "step": 7700 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002752884615384615, + "loss": 0.1054, + "step": 7710 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002752564102564102, + "loss": 0.0956, + "step": 7720 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027522435897435894, + "loss": 0.096, + "step": 7730 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027519230769230767, + "loss": 0.0898, + "step": 7740 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027516025641025635, + "loss": 0.088, + "step": 7750 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002751282051282051, + "loss": 0.0895, + "step": 7760 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002750961538461538, + "loss": 0.1, + "step": 7770 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027506410256410255, + "loss": 0.0947, + "step": 7780 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027503205128205123, + "loss": 0.0886, + "step": 7790 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027499999999999996, + "loss": 0.0926, + "step": 7800 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002749679487179487, + "loss": 0.1006, + "step": 7810 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002749358974358974, + "loss": 0.0981, + "step": 7820 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002749038461538461, + "loss": 0.0965, + "step": 7830 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027487179487179484, + "loss": 0.0941, + "step": 7840 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002748397435897436, + "loss": 0.0956, + "step": 7850 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027480769230769225, + "loss": 0.0943, + "step": 7860 + }, + { + "epoch": 0.17, + "learning_rate": 0.000274775641025641, + "loss": 0.0903, + "step": 7870 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002747435897435897, + "loss": 0.0977, + "step": 7880 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027471153846153845, + "loss": 0.0907, + "step": 7890 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027467948717948713, + "loss": 0.0922, + "step": 7900 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027464743589743586, + "loss": 0.0837, + "step": 7910 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002746153846153846, + "loss": 0.111, + "step": 7920 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002745833333333333, + "loss": 0.0949, + "step": 7930 + }, + { + "epoch": 0.17, + "learning_rate": 0.000274551282051282, + "loss": 0.0923, + "step": 7940 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027451923076923074, + "loss": 0.0913, + "step": 7950 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002744871794871795, + "loss": 0.0933, + "step": 7960 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027445512820512815, + "loss": 0.1059, + "step": 7970 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002744230769230769, + "loss": 0.0945, + "step": 7980 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002743910256410256, + "loss": 0.0951, + "step": 7990 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002743589743589743, + "loss": 0.1038, + "step": 8000 + }, + { + "epoch": 0.17, + "eval_cer": 0.9239619967099145, + "eval_loss": 0.061513788998126984, + "eval_runtime": 120.6477, + "eval_samples_per_second": 16.577, + "eval_steps_per_second": 4.144, + "step": 8000 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027432692307692303, + "loss": 0.0983, + "step": 8010 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027429487179487177, + "loss": 0.0991, + "step": 8020 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002742628205128205, + "loss": 0.0967, + "step": 8030 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002742307692307692, + "loss": 0.1007, + "step": 8040 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002741987179487179, + "loss": 0.0913, + "step": 8050 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027416666666666664, + "loss": 0.0888, + "step": 8060 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002741346153846154, + "loss": 0.0848, + "step": 8070 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027410256410256406, + "loss": 0.0899, + "step": 8080 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002740705128205128, + "loss": 0.0994, + "step": 8090 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002740384615384615, + "loss": 0.0931, + "step": 8100 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002740064102564102, + "loss": 0.0956, + "step": 8110 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027397435897435894, + "loss": 0.0977, + "step": 8120 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027394230769230767, + "loss": 0.0942, + "step": 8130 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002739102564102564, + "loss": 0.0934, + "step": 8140 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002738782051282051, + "loss": 0.0958, + "step": 8150 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002738461538461538, + "loss": 0.1003, + "step": 8160 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027381410256410255, + "loss": 0.1116, + "step": 8170 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002737820512820512, + "loss": 0.0952, + "step": 8180 + }, + { + "epoch": 0.17, + "learning_rate": 0.00027374999999999996, + "loss": 0.0969, + "step": 8190 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002737179487179487, + "loss": 0.0979, + "step": 8200 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002736858974358974, + "loss": 0.1008, + "step": 8210 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002736538461538461, + "loss": 0.0963, + "step": 8220 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027362179487179484, + "loss": 0.0913, + "step": 8230 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027358974358974357, + "loss": 0.0957, + "step": 8240 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002735576923076923, + "loss": 0.1048, + "step": 8250 + }, + { + "epoch": 0.18, + "learning_rate": 0.000273525641025641, + "loss": 0.0961, + "step": 8260 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002734935897435897, + "loss": 0.0891, + "step": 8270 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027346153846153845, + "loss": 0.0892, + "step": 8280 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027342948717948713, + "loss": 0.1004, + "step": 8290 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027339743589743586, + "loss": 0.0915, + "step": 8300 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002733653846153846, + "loss": 0.0976, + "step": 8310 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027333333333333333, + "loss": 0.1027, + "step": 8320 + }, + { + "epoch": 0.18, + "learning_rate": 0.000273301282051282, + "loss": 0.1025, + "step": 8330 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027326923076923074, + "loss": 0.1037, + "step": 8340 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027323717948717947, + "loss": 0.1022, + "step": 8350 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027320512820512815, + "loss": 0.0912, + "step": 8360 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002731730769230769, + "loss": 0.0968, + "step": 8370 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002731410256410256, + "loss": 0.1133, + "step": 8380 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027310897435897435, + "loss": 0.0967, + "step": 8390 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027307692307692303, + "loss": 0.0978, + "step": 8400 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027304487179487176, + "loss": 0.0985, + "step": 8410 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002730128205128205, + "loss": 0.0932, + "step": 8420 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027298076923076923, + "loss": 0.1028, + "step": 8430 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002729487179487179, + "loss": 0.1011, + "step": 8440 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027291666666666664, + "loss": 0.0899, + "step": 8450 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002728846153846154, + "loss": 0.0923, + "step": 8460 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027285256410256405, + "loss": 0.0958, + "step": 8470 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002728205128205128, + "loss": 0.0992, + "step": 8480 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002727884615384615, + "loss": 0.0873, + "step": 8490 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027275641025641025, + "loss": 0.0919, + "step": 8500 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027272435897435893, + "loss": 0.0925, + "step": 8510 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027269230769230767, + "loss": 0.0979, + "step": 8520 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002726602564102564, + "loss": 0.105, + "step": 8530 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002726282051282051, + "loss": 0.0852, + "step": 8540 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002725961538461538, + "loss": 0.0964, + "step": 8550 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027256410256410254, + "loss": 0.0919, + "step": 8560 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002725320512820513, + "loss": 0.0884, + "step": 8570 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027249999999999996, + "loss": 0.0975, + "step": 8580 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002724679487179487, + "loss": 0.0956, + "step": 8590 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002724358974358974, + "loss": 0.0767, + "step": 8600 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002724038461538461, + "loss": 0.0908, + "step": 8610 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027237179487179483, + "loss": 0.1017, + "step": 8620 + }, + { + "epoch": 0.18, + "learning_rate": 0.00027233974358974357, + "loss": 0.0916, + "step": 8630 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002723076923076923, + "loss": 0.1033, + "step": 8640 + }, + { + "epoch": 0.18, + "learning_rate": 0.000272275641025641, + "loss": 0.0944, + "step": 8650 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002722435897435897, + "loss": 0.092, + "step": 8660 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027221153846153845, + "loss": 0.092, + "step": 8670 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002721794871794872, + "loss": 0.097, + "step": 8680 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027214743589743586, + "loss": 0.1019, + "step": 8690 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002721153846153846, + "loss": 0.096, + "step": 8700 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002720833333333333, + "loss": 0.0975, + "step": 8710 + }, + { + "epoch": 0.19, + "learning_rate": 0.000272051282051282, + "loss": 0.0969, + "step": 8720 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027201923076923074, + "loss": 0.0937, + "step": 8730 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027198717948717947, + "loss": 0.0947, + "step": 8740 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002719551282051282, + "loss": 0.0982, + "step": 8750 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002719230769230769, + "loss": 0.0823, + "step": 8760 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002718910256410256, + "loss": 0.0923, + "step": 8770 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027185897435897435, + "loss": 0.0836, + "step": 8780 + }, + { + "epoch": 0.19, + "learning_rate": 0.000271826923076923, + "loss": 0.0875, + "step": 8790 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027179487179487176, + "loss": 0.091, + "step": 8800 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002717628205128205, + "loss": 0.0913, + "step": 8810 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002717307692307692, + "loss": 0.0937, + "step": 8820 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002716987179487179, + "loss": 0.0976, + "step": 8830 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027166666666666664, + "loss": 0.1068, + "step": 8840 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027163461538461537, + "loss": 0.0896, + "step": 8850 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002716025641025641, + "loss": 0.0933, + "step": 8860 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002715705128205128, + "loss": 0.1008, + "step": 8870 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002715384615384615, + "loss": 0.0984, + "step": 8880 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027150641025641025, + "loss": 0.1047, + "step": 8890 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027147435897435893, + "loss": 0.087, + "step": 8900 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027144230769230766, + "loss": 0.0852, + "step": 8910 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002714102564102564, + "loss": 0.0906, + "step": 8920 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027137820512820513, + "loss": 0.1001, + "step": 8930 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002713461538461538, + "loss": 0.0955, + "step": 8940 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027131410256410254, + "loss": 0.0885, + "step": 8950 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002712820512820513, + "loss": 0.0885, + "step": 8960 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027124999999999995, + "loss": 0.0843, + "step": 8970 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002712179487179487, + "loss": 0.0907, + "step": 8980 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002711858974358974, + "loss": 0.0862, + "step": 8990 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027115384615384615, + "loss": 0.0815, + "step": 9000 + }, + { + "epoch": 0.19, + "eval_cer": 0.9240053096578532, + "eval_loss": 0.05747230350971222, + "eval_runtime": 120.6526, + "eval_samples_per_second": 16.577, + "eval_steps_per_second": 4.144, + "step": 9000 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027112179487179483, + "loss": 0.0805, + "step": 9010 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027108974358974356, + "loss": 0.0912, + "step": 9020 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002710576923076923, + "loss": 0.0824, + "step": 9030 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027102564102564103, + "loss": 0.0958, + "step": 9040 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002709935897435897, + "loss": 0.0785, + "step": 9050 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027096153846153844, + "loss": 0.0861, + "step": 9060 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002709294871794872, + "loss": 0.0838, + "step": 9070 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027089743589743585, + "loss": 0.0797, + "step": 9080 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002708653846153846, + "loss": 0.0953, + "step": 9090 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002708333333333333, + "loss": 0.0892, + "step": 9100 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027080128205128205, + "loss": 0.0964, + "step": 9110 + }, + { + "epoch": 0.19, + "learning_rate": 0.00027076923076923073, + "loss": 0.0864, + "step": 9120 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027073717948717947, + "loss": 0.079, + "step": 9130 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002707051282051282, + "loss": 0.0897, + "step": 9140 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002706730769230769, + "loss": 0.0879, + "step": 9150 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002706410256410256, + "loss": 0.0831, + "step": 9160 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027060897435897434, + "loss": 0.0936, + "step": 9170 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002705769230769231, + "loss": 0.0919, + "step": 9180 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027054487179487176, + "loss": 0.095, + "step": 9190 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002705128205128205, + "loss": 0.0912, + "step": 9200 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002704807692307692, + "loss": 0.0887, + "step": 9210 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027044871794871796, + "loss": 0.0897, + "step": 9220 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027041666666666664, + "loss": 0.0892, + "step": 9230 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027038461538461537, + "loss": 0.0934, + "step": 9240 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002703525641025641, + "loss": 0.0844, + "step": 9250 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002703205128205128, + "loss": 0.1071, + "step": 9260 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002702884615384615, + "loss": 0.0918, + "step": 9270 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027025641025641025, + "loss": 0.0839, + "step": 9280 + }, + { + "epoch": 0.2, + "learning_rate": 0.000270224358974359, + "loss": 0.0823, + "step": 9290 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027019230769230766, + "loss": 0.0866, + "step": 9300 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002701602564102564, + "loss": 0.0965, + "step": 9310 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002701282051282051, + "loss": 0.0898, + "step": 9320 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002700961538461538, + "loss": 0.108, + "step": 9330 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027006410256410254, + "loss": 0.0825, + "step": 9340 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027003205128205127, + "loss": 0.0894, + "step": 9350 + }, + { + "epoch": 0.2, + "learning_rate": 0.00027, + "loss": 0.0898, + "step": 9360 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002699679487179487, + "loss": 0.101, + "step": 9370 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002699358974358974, + "loss": 0.097, + "step": 9380 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026990384615384615, + "loss": 0.0951, + "step": 9390 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002698717948717949, + "loss": 0.0885, + "step": 9400 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026983974358974356, + "loss": 0.0818, + "step": 9410 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002698076923076923, + "loss": 0.0902, + "step": 9420 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026977564102564103, + "loss": 0.0906, + "step": 9430 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002697435897435897, + "loss": 0.0882, + "step": 9440 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026971153846153844, + "loss": 0.0892, + "step": 9450 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026967948717948717, + "loss": 0.0911, + "step": 9460 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002696474358974359, + "loss": 0.1027, + "step": 9470 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002696153846153846, + "loss": 0.0897, + "step": 9480 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002695833333333333, + "loss": 0.0938, + "step": 9490 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026955128205128205, + "loss": 0.1043, + "step": 9500 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026951923076923073, + "loss": 0.0854, + "step": 9510 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026948717948717946, + "loss": 0.0866, + "step": 9520 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002694551282051282, + "loss": 0.1041, + "step": 9530 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026942307692307693, + "loss": 0.0949, + "step": 9540 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002693910256410256, + "loss": 0.0949, + "step": 9550 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026935897435897434, + "loss": 0.0936, + "step": 9560 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002693269230769231, + "loss": 0.0905, + "step": 9570 + }, + { + "epoch": 0.2, + "learning_rate": 0.00026929487179487175, + "loss": 0.0808, + "step": 9580 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002692628205128205, + "loss": 0.0949, + "step": 9590 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002692307692307692, + "loss": 0.0903, + "step": 9600 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026919871794871795, + "loss": 0.0788, + "step": 9610 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026916666666666663, + "loss": 0.0879, + "step": 9620 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026913461538461536, + "loss": 0.0892, + "step": 9630 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002691025641025641, + "loss": 0.0882, + "step": 9640 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026907051282051283, + "loss": 0.0928, + "step": 9650 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002690384615384615, + "loss": 0.0966, + "step": 9660 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026900641025641024, + "loss": 0.0865, + "step": 9670 + }, + { + "epoch": 0.21, + "learning_rate": 0.000268974358974359, + "loss": 0.0929, + "step": 9680 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026894230769230766, + "loss": 0.0797, + "step": 9690 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002689102564102564, + "loss": 0.0803, + "step": 9700 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002688782051282051, + "loss": 0.0948, + "step": 9710 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026884615384615385, + "loss": 0.0818, + "step": 9720 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026881410256410253, + "loss": 0.0906, + "step": 9730 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026878205128205127, + "loss": 0.0825, + "step": 9740 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026875, + "loss": 0.0931, + "step": 9750 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002687179487179487, + "loss": 0.092, + "step": 9760 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002686858974358974, + "loss": 0.0908, + "step": 9770 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026865384615384615, + "loss": 0.078, + "step": 9780 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002686217948717949, + "loss": 0.0864, + "step": 9790 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026858974358974356, + "loss": 0.0868, + "step": 9800 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002685576923076923, + "loss": 0.0898, + "step": 9810 + }, + { + "epoch": 0.21, + "learning_rate": 0.000268525641025641, + "loss": 0.0818, + "step": 9820 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026849358974358976, + "loss": 0.0881, + "step": 9830 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026846153846153844, + "loss": 0.0888, + "step": 9840 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026842948717948717, + "loss": 0.0862, + "step": 9850 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002683974358974359, + "loss": 0.0833, + "step": 9860 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002683653846153846, + "loss": 0.0934, + "step": 9870 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002683333333333333, + "loss": 0.0841, + "step": 9880 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026830128205128205, + "loss": 0.082, + "step": 9890 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002682692307692308, + "loss": 0.0876, + "step": 9900 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026823717948717946, + "loss": 0.0847, + "step": 9910 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002682051282051282, + "loss": 0.0839, + "step": 9920 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002681730769230769, + "loss": 0.0865, + "step": 9930 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002681410256410256, + "loss": 0.0814, + "step": 9940 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026810897435897434, + "loss": 0.0872, + "step": 9950 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026807692307692307, + "loss": 0.0857, + "step": 9960 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002680448717948718, + "loss": 0.0876, + "step": 9970 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002680128205128205, + "loss": 0.1015, + "step": 9980 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002679807692307692, + "loss": 0.0863, + "step": 9990 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026794871794871795, + "loss": 0.0828, + "step": 10000 + }, + { + "epoch": 0.21, + "eval_cer": 0.9240622109816158, + "eval_loss": 0.057162459939718246, + "eval_runtime": 120.4524, + "eval_samples_per_second": 16.604, + "eval_steps_per_second": 4.151, + "step": 10000 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002679166666666667, + "loss": 0.0945, + "step": 10010 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026788461538461536, + "loss": 0.0774, + "step": 10020 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002678525641025641, + "loss": 0.0933, + "step": 10030 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026782051282051283, + "loss": 0.0829, + "step": 10040 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002677884615384615, + "loss": 0.0785, + "step": 10050 + }, + { + "epoch": 0.21, + "learning_rate": 0.00026775641025641024, + "loss": 0.0869, + "step": 10060 + }, + { + "epoch": 0.22, + "learning_rate": 0.000267724358974359, + "loss": 0.0875, + "step": 10070 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002676923076923077, + "loss": 0.0854, + "step": 10080 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002676602564102564, + "loss": 0.0928, + "step": 10090 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002676282051282051, + "loss": 0.0861, + "step": 10100 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026759615384615385, + "loss": 0.0954, + "step": 10110 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026756410256410253, + "loss": 0.0877, + "step": 10120 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026753205128205126, + "loss": 0.0852, + "step": 10130 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002675, + "loss": 0.0757, + "step": 10140 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026746794871794873, + "loss": 0.0839, + "step": 10150 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002674358974358974, + "loss": 0.0911, + "step": 10160 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026740384615384614, + "loss": 0.091, + "step": 10170 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002673717948717949, + "loss": 0.0913, + "step": 10180 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002673397435897436, + "loss": 0.1012, + "step": 10190 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002673076923076923, + "loss": 0.093, + "step": 10200 + }, + { + "epoch": 0.22, + "learning_rate": 0.000267275641025641, + "loss": 0.0968, + "step": 10210 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026724358974358975, + "loss": 0.0889, + "step": 10220 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026721153846153843, + "loss": 0.0835, + "step": 10230 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026717948717948717, + "loss": 0.0784, + "step": 10240 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026714743589743584, + "loss": 0.0822, + "step": 10250 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002671153846153846, + "loss": 0.0765, + "step": 10260 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002670833333333333, + "loss": 0.07, + "step": 10270 + }, + { + "epoch": 0.22, + "learning_rate": 0.000267051282051282, + "loss": 0.081, + "step": 10280 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002670192307692307, + "loss": 0.0914, + "step": 10290 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026698717948717946, + "loss": 0.0797, + "step": 10300 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002669551282051282, + "loss": 0.09, + "step": 10310 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026692307692307687, + "loss": 0.0924, + "step": 10320 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002668910256410256, + "loss": 0.0852, + "step": 10330 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026685897435897433, + "loss": 0.0897, + "step": 10340 + }, + { + "epoch": 0.22, + "learning_rate": 0.000266826923076923, + "loss": 0.0808, + "step": 10350 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026679487179487175, + "loss": 0.0823, + "step": 10360 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002667628205128205, + "loss": 0.0829, + "step": 10370 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002667307692307692, + "loss": 0.0931, + "step": 10380 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002666987179487179, + "loss": 0.0782, + "step": 10390 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002666666666666666, + "loss": 0.0804, + "step": 10400 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026663461538461536, + "loss": 0.0887, + "step": 10410 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002666025641025641, + "loss": 0.0903, + "step": 10420 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026657051282051277, + "loss": 0.0847, + "step": 10430 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002665384615384615, + "loss": 0.0899, + "step": 10440 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026650641025641024, + "loss": 0.0814, + "step": 10450 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002664743589743589, + "loss": 0.0899, + "step": 10460 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026644230769230765, + "loss": 0.0828, + "step": 10470 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002664102564102564, + "loss": 0.0819, + "step": 10480 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002663782051282051, + "loss": 0.0893, + "step": 10490 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002663461538461538, + "loss": 0.0863, + "step": 10500 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026631410256410253, + "loss": 0.0866, + "step": 10510 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026628205128205126, + "loss": 0.0742, + "step": 10520 + }, + { + "epoch": 0.22, + "learning_rate": 0.00026624999999999994, + "loss": 0.0807, + "step": 10530 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026621794871794867, + "loss": 0.0847, + "step": 10540 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002661858974358974, + "loss": 0.081, + "step": 10550 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026615384615384614, + "loss": 0.0801, + "step": 10560 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002661217948717948, + "loss": 0.0851, + "step": 10570 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026608974358974355, + "loss": 0.0792, + "step": 10580 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002660576923076923, + "loss": 0.0805, + "step": 10590 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026602564102564096, + "loss": 0.0878, + "step": 10600 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002659935897435897, + "loss": 0.092, + "step": 10610 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026596153846153843, + "loss": 0.0845, + "step": 10620 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026592948717948716, + "loss": 0.0993, + "step": 10630 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026589743589743584, + "loss": 0.0772, + "step": 10640 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002658653846153846, + "loss": 0.0704, + "step": 10650 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002658333333333333, + "loss": 0.0925, + "step": 10660 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026580128205128204, + "loss": 0.091, + "step": 10670 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002657692307692307, + "loss": 0.078, + "step": 10680 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026573717948717945, + "loss": 0.0785, + "step": 10690 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002657051282051282, + "loss": 0.085, + "step": 10700 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026567307692307687, + "loss": 0.0809, + "step": 10710 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002656410256410256, + "loss": 0.0987, + "step": 10720 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026560897435897433, + "loss": 0.081, + "step": 10730 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026557692307692306, + "loss": 0.0868, + "step": 10740 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026554487179487174, + "loss": 0.0811, + "step": 10750 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002655128205128205, + "loss": 0.0876, + "step": 10760 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002654807692307692, + "loss": 0.0801, + "step": 10770 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002654487179487179, + "loss": 0.0903, + "step": 10780 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002654166666666666, + "loss": 0.0825, + "step": 10790 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026538461538461536, + "loss": 0.0879, + "step": 10800 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002653525641025641, + "loss": 0.0893, + "step": 10810 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026532051282051277, + "loss": 0.0945, + "step": 10820 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002652884615384615, + "loss": 0.0822, + "step": 10830 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026525641025641023, + "loss": 0.0899, + "step": 10840 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026522435897435897, + "loss": 0.0865, + "step": 10850 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026519230769230765, + "loss": 0.0802, + "step": 10860 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002651602564102564, + "loss": 0.081, + "step": 10870 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002651282051282051, + "loss": 0.086, + "step": 10880 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002650961538461538, + "loss": 0.073, + "step": 10890 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002650641025641025, + "loss": 0.0922, + "step": 10900 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026503205128205126, + "loss": 0.0867, + "step": 10910 + }, + { + "epoch": 0.23, + "learning_rate": 0.000265, + "loss": 0.0811, + "step": 10920 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026496794871794867, + "loss": 0.0904, + "step": 10930 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002649358974358974, + "loss": 0.0734, + "step": 10940 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026490384615384614, + "loss": 0.0812, + "step": 10950 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002648717948717948, + "loss": 0.0878, + "step": 10960 + }, + { + "epoch": 0.23, + "learning_rate": 0.00026483974358974355, + "loss": 0.0836, + "step": 10970 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002648076923076923, + "loss": 0.0858, + "step": 10980 + }, + { + "epoch": 0.23, + "learning_rate": 0.000264775641025641, + "loss": 0.0904, + "step": 10990 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002647435897435897, + "loss": 0.0851, + "step": 11000 + }, + { + "epoch": 0.24, + "eval_cer": 0.9240995790151315, + "eval_loss": 0.05332302674651146, + "eval_runtime": 120.6387, + "eval_samples_per_second": 16.578, + "eval_steps_per_second": 4.145, + "step": 11000 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002647115384615384, + "loss": 0.0844, + "step": 11010 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026467948717948716, + "loss": 0.0774, + "step": 11020 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002646474358974359, + "loss": 0.0749, + "step": 11030 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026461538461538457, + "loss": 0.0903, + "step": 11040 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002645833333333333, + "loss": 0.0999, + "step": 11050 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026455128205128204, + "loss": 0.0881, + "step": 11060 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002645192307692307, + "loss": 0.0809, + "step": 11070 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026448717948717945, + "loss": 0.0869, + "step": 11080 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002644551282051282, + "loss": 0.0902, + "step": 11090 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002644230769230769, + "loss": 0.0787, + "step": 11100 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002643910256410256, + "loss": 0.0894, + "step": 11110 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026435897435897433, + "loss": 0.0905, + "step": 11120 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026432692307692306, + "loss": 0.0763, + "step": 11130 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026429487179487174, + "loss": 0.0779, + "step": 11140 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002642628205128205, + "loss": 0.0812, + "step": 11150 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002642307692307692, + "loss": 0.0886, + "step": 11160 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026419871794871794, + "loss": 0.0874, + "step": 11170 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002641666666666666, + "loss": 0.0832, + "step": 11180 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026413461538461535, + "loss": 0.0836, + "step": 11190 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002641025641025641, + "loss": 0.0865, + "step": 11200 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002640705128205128, + "loss": 0.084, + "step": 11210 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002640384615384615, + "loss": 0.0912, + "step": 11220 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026400641025641023, + "loss": 0.0886, + "step": 11230 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026397435897435896, + "loss": 0.0883, + "step": 11240 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026394230769230764, + "loss": 0.0881, + "step": 11250 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002639102564102564, + "loss": 0.0959, + "step": 11260 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002638782051282051, + "loss": 0.0967, + "step": 11270 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026384615384615384, + "loss": 0.0772, + "step": 11280 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002638141025641025, + "loss": 0.0857, + "step": 11290 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026378205128205125, + "loss": 0.0861, + "step": 11300 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026375, + "loss": 0.081, + "step": 11310 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026371794871794867, + "loss": 0.0801, + "step": 11320 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002636858974358974, + "loss": 0.0806, + "step": 11330 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026365384615384613, + "loss": 0.0874, + "step": 11340 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026362179487179487, + "loss": 0.0831, + "step": 11350 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026358974358974354, + "loss": 0.0902, + "step": 11360 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002635576923076923, + "loss": 0.1005, + "step": 11370 + }, + { + "epoch": 0.24, + "learning_rate": 0.000263525641025641, + "loss": 0.1001, + "step": 11380 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026349358974358974, + "loss": 0.0887, + "step": 11390 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002634615384615384, + "loss": 0.0842, + "step": 11400 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026342948717948716, + "loss": 0.0849, + "step": 11410 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002633974358974359, + "loss": 0.0746, + "step": 11420 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026336538461538457, + "loss": 0.0878, + "step": 11430 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002633333333333333, + "loss": 0.0894, + "step": 11440 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026330128205128203, + "loss": 0.0948, + "step": 11450 + }, + { + "epoch": 0.24, + "learning_rate": 0.00026326923076923077, + "loss": 0.0806, + "step": 11460 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026323717948717945, + "loss": 0.0853, + "step": 11470 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002632051282051282, + "loss": 0.0865, + "step": 11480 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002631730769230769, + "loss": 0.0859, + "step": 11490 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002631410256410256, + "loss": 0.0864, + "step": 11500 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002631089743589743, + "loss": 0.0914, + "step": 11510 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026307692307692306, + "loss": 0.0849, + "step": 11520 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002630448717948718, + "loss": 0.0856, + "step": 11530 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026301282051282047, + "loss": 0.0857, + "step": 11540 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002629807692307692, + "loss": 0.0979, + "step": 11550 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026294871794871794, + "loss": 0.0831, + "step": 11560 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002629166666666666, + "loss": 0.0835, + "step": 11570 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026288461538461535, + "loss": 0.0852, + "step": 11580 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002628525641025641, + "loss": 0.0774, + "step": 11590 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002628205128205128, + "loss": 0.0808, + "step": 11600 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002627884615384615, + "loss": 0.081, + "step": 11610 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002627564102564102, + "loss": 0.0982, + "step": 11620 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026272435897435896, + "loss": 0.081, + "step": 11630 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002626923076923077, + "loss": 0.0834, + "step": 11640 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026266025641025637, + "loss": 0.0885, + "step": 11650 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002626282051282051, + "loss": 0.0812, + "step": 11660 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026259615384615384, + "loss": 0.0763, + "step": 11670 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002625641025641025, + "loss": 0.0824, + "step": 11680 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026253205128205125, + "loss": 0.0819, + "step": 11690 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002625, + "loss": 0.0775, + "step": 11700 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002624679487179487, + "loss": 0.079, + "step": 11710 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002624358974358974, + "loss": 0.0782, + "step": 11720 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026240384615384613, + "loss": 0.0819, + "step": 11730 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026237179487179486, + "loss": 0.0913, + "step": 11740 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026233974358974354, + "loss": 0.0989, + "step": 11750 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002623076923076923, + "loss": 0.0853, + "step": 11760 + }, + { + "epoch": 0.25, + "learning_rate": 0.000262275641025641, + "loss": 0.0745, + "step": 11770 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026224358974358974, + "loss": 0.0765, + "step": 11780 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002622115384615384, + "loss": 0.0881, + "step": 11790 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026217948717948715, + "loss": 0.074, + "step": 11800 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002621474358974359, + "loss": 0.0758, + "step": 11810 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002621153846153846, + "loss": 0.0989, + "step": 11820 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002620833333333333, + "loss": 0.0774, + "step": 11830 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026205128205128203, + "loss": 0.0814, + "step": 11840 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026201923076923076, + "loss": 0.086, + "step": 11850 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026198717948717944, + "loss": 0.0845, + "step": 11860 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002619551282051282, + "loss": 0.0783, + "step": 11870 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002619230769230769, + "loss": 0.0802, + "step": 11880 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026189102564102564, + "loss": 0.084, + "step": 11890 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002618589743589743, + "loss": 0.0891, + "step": 11900 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026182692307692305, + "loss": 0.0852, + "step": 11910 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002617948717948718, + "loss": 0.0817, + "step": 11920 + }, + { + "epoch": 0.25, + "learning_rate": 0.00026176282051282047, + "loss": 0.079, + "step": 11930 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002617307692307692, + "loss": 0.0893, + "step": 11940 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026169871794871793, + "loss": 0.0713, + "step": 11950 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026166666666666667, + "loss": 0.084, + "step": 11960 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026163461538461535, + "loss": 0.089, + "step": 11970 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002616025641025641, + "loss": 0.0939, + "step": 11980 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002615705128205128, + "loss": 0.0855, + "step": 11990 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026153846153846154, + "loss": 0.0787, + "step": 12000 + }, + { + "epoch": 0.26, + "eval_cer": 0.9240961819211755, + "eval_loss": 0.05290180444717407, + "eval_runtime": 120.6826, + "eval_samples_per_second": 16.572, + "eval_steps_per_second": 4.143, + "step": 12000 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002615064102564102, + "loss": 0.0835, + "step": 12010 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026147435897435896, + "loss": 0.0776, + "step": 12020 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002614423076923077, + "loss": 0.0762, + "step": 12030 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026141025641025637, + "loss": 0.0916, + "step": 12040 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002613782051282051, + "loss": 0.0824, + "step": 12050 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026134615384615384, + "loss": 0.0754, + "step": 12060 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026131410256410257, + "loss": 0.0814, + "step": 12070 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026128205128205125, + "loss": 0.0717, + "step": 12080 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026125, + "loss": 0.0776, + "step": 12090 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002612179487179487, + "loss": 0.0806, + "step": 12100 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002611858974358974, + "loss": 0.0837, + "step": 12110 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002611538461538461, + "loss": 0.082, + "step": 12120 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026112179487179486, + "loss": 0.0905, + "step": 12130 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002610897435897436, + "loss": 0.0696, + "step": 12140 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026105769230769227, + "loss": 0.0922, + "step": 12150 + }, + { + "epoch": 0.26, + "learning_rate": 0.000261025641025641, + "loss": 0.0753, + "step": 12160 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026099358974358974, + "loss": 0.083, + "step": 12170 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026096153846153847, + "loss": 0.0801, + "step": 12180 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026092948717948715, + "loss": 0.0799, + "step": 12190 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002608974358974359, + "loss": 0.075, + "step": 12200 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002608653846153846, + "loss": 0.0725, + "step": 12210 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002608333333333333, + "loss": 0.0856, + "step": 12220 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026080128205128203, + "loss": 0.0784, + "step": 12230 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026076923076923076, + "loss": 0.0741, + "step": 12240 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002607371794871795, + "loss": 0.0887, + "step": 12250 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002607051282051282, + "loss": 0.0796, + "step": 12260 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002606730769230769, + "loss": 0.0855, + "step": 12270 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026064102564102564, + "loss": 0.0787, + "step": 12280 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002606089743589743, + "loss": 0.083, + "step": 12290 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026057692307692305, + "loss": 0.0863, + "step": 12300 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002605448717948718, + "loss": 0.0807, + "step": 12310 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002605128205128205, + "loss": 0.0771, + "step": 12320 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002604807692307692, + "loss": 0.069, + "step": 12330 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026044871794871793, + "loss": 0.0816, + "step": 12340 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026041666666666666, + "loss": 0.0692, + "step": 12350 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002603846153846154, + "loss": 0.0763, + "step": 12360 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002603525641025641, + "loss": 0.076, + "step": 12370 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002603205128205128, + "loss": 0.0897, + "step": 12380 + }, + { + "epoch": 0.26, + "learning_rate": 0.00026028846153846154, + "loss": 0.0789, + "step": 12390 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002602564102564102, + "loss": 0.0743, + "step": 12400 + }, + { + "epoch": 0.27, + "learning_rate": 0.00026022435897435895, + "loss": 0.0787, + "step": 12410 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002601923076923077, + "loss": 0.0778, + "step": 12420 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002601602564102564, + "loss": 0.0802, + "step": 12430 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002601282051282051, + "loss": 0.084, + "step": 12440 + }, + { + "epoch": 0.27, + "learning_rate": 0.00026009615384615383, + "loss": 0.0845, + "step": 12450 + }, + { + "epoch": 0.27, + "learning_rate": 0.00026006410256410257, + "loss": 0.0838, + "step": 12460 + }, + { + "epoch": 0.27, + "learning_rate": 0.00026003205128205124, + "loss": 0.0742, + "step": 12470 + }, + { + "epoch": 0.27, + "learning_rate": 0.00026, + "loss": 0.0783, + "step": 12480 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002599679487179487, + "loss": 0.0673, + "step": 12490 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025993589743589744, + "loss": 0.0771, + "step": 12500 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002599038461538461, + "loss": 0.0813, + "step": 12510 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025987179487179486, + "loss": 0.0826, + "step": 12520 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002598397435897436, + "loss": 0.0736, + "step": 12530 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025980769230769227, + "loss": 0.0761, + "step": 12540 + }, + { + "epoch": 0.27, + "learning_rate": 0.000259775641025641, + "loss": 0.0978, + "step": 12550 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025974358974358973, + "loss": 0.0763, + "step": 12560 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025971153846153847, + "loss": 0.0843, + "step": 12570 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025967948717948715, + "loss": 0.0831, + "step": 12580 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002596474358974359, + "loss": 0.0711, + "step": 12590 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002596153846153846, + "loss": 0.073, + "step": 12600 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025958333333333335, + "loss": 0.0794, + "step": 12610 + }, + { + "epoch": 0.27, + "learning_rate": 0.000259551282051282, + "loss": 0.0817, + "step": 12620 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025951923076923076, + "loss": 0.0803, + "step": 12630 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002594871794871795, + "loss": 0.0751, + "step": 12640 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025945512820512817, + "loss": 0.0748, + "step": 12650 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002594230769230769, + "loss": 0.079, + "step": 12660 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025939102564102564, + "loss": 0.0797, + "step": 12670 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025935897435897437, + "loss": 0.0813, + "step": 12680 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025932692307692305, + "loss": 0.0826, + "step": 12690 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002592948717948718, + "loss": 0.0762, + "step": 12700 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002592628205128205, + "loss": 0.0756, + "step": 12710 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002592307692307692, + "loss": 0.0883, + "step": 12720 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002591987179487179, + "loss": 0.0755, + "step": 12730 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025916666666666666, + "loss": 0.0782, + "step": 12740 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002591346153846154, + "loss": 0.0782, + "step": 12750 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025910256410256407, + "loss": 0.0835, + "step": 12760 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002590705128205128, + "loss": 0.0907, + "step": 12770 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025903846153846154, + "loss": 0.0903, + "step": 12780 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025900641025641027, + "loss": 0.0759, + "step": 12790 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025897435897435895, + "loss": 0.0978, + "step": 12800 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002589423076923077, + "loss": 0.0871, + "step": 12810 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002589102564102564, + "loss": 0.0865, + "step": 12820 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002588782051282051, + "loss": 0.0789, + "step": 12830 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025884615384615383, + "loss": 0.0761, + "step": 12840 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025881410256410256, + "loss": 0.0765, + "step": 12850 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002587820512820513, + "loss": 0.0822, + "step": 12860 + }, + { + "epoch": 0.27, + "learning_rate": 0.00025875, + "loss": 0.0708, + "step": 12870 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002587179487179487, + "loss": 0.075, + "step": 12880 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025868589743589744, + "loss": 0.0759, + "step": 12890 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002586538461538461, + "loss": 0.0832, + "step": 12900 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025862179487179485, + "loss": 0.0773, + "step": 12910 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002585897435897436, + "loss": 0.0855, + "step": 12920 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002585576923076923, + "loss": 0.0895, + "step": 12930 + }, + { + "epoch": 0.28, + "learning_rate": 0.000258525641025641, + "loss": 0.0819, + "step": 12940 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025849358974358973, + "loss": 0.0749, + "step": 12950 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025846153846153846, + "loss": 0.0844, + "step": 12960 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002584294871794872, + "loss": 0.0817, + "step": 12970 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002583974358974359, + "loss": 0.0858, + "step": 12980 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002583653846153846, + "loss": 0.0765, + "step": 12990 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025833333333333334, + "loss": 0.0795, + "step": 13000 + }, + { + "epoch": 0.28, + "eval_cer": 0.9239297243173328, + "eval_loss": 0.051773346960544586, + "eval_runtime": 120.5124, + "eval_samples_per_second": 16.596, + "eval_steps_per_second": 4.149, + "step": 13000 + }, + { + "epoch": 0.28, + "learning_rate": 0.000258301282051282, + "loss": 0.0741, + "step": 13010 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025826923076923075, + "loss": 0.086, + "step": 13020 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002582371794871795, + "loss": 0.08, + "step": 13030 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002582051282051282, + "loss": 0.0843, + "step": 13040 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002581730769230769, + "loss": 0.0733, + "step": 13050 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025814102564102563, + "loss": 0.0823, + "step": 13060 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025810897435897437, + "loss": 0.0771, + "step": 13070 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025807692307692304, + "loss": 0.0804, + "step": 13080 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002580448717948718, + "loss": 0.0692, + "step": 13090 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002580128205128205, + "loss": 0.0859, + "step": 13100 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025798076923076924, + "loss": 0.0848, + "step": 13110 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002579487179487179, + "loss": 0.0826, + "step": 13120 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025791666666666666, + "loss": 0.0945, + "step": 13130 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002578846153846154, + "loss": 0.0871, + "step": 13140 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002578525641025641, + "loss": 0.0913, + "step": 13150 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002578205128205128, + "loss": 0.0825, + "step": 13160 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002577884615384615, + "loss": 0.0727, + "step": 13170 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002577564102564102, + "loss": 0.0628, + "step": 13180 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025772435897435895, + "loss": 0.0879, + "step": 13190 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002576923076923077, + "loss": 0.0821, + "step": 13200 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025766025641025636, + "loss": 0.0749, + "step": 13210 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002576282051282051, + "loss": 0.0853, + "step": 13220 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002575961538461538, + "loss": 0.0864, + "step": 13230 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002575641025641025, + "loss": 0.0941, + "step": 13240 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025753205128205124, + "loss": 0.081, + "step": 13250 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025749999999999997, + "loss": 0.0765, + "step": 13260 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002574679487179487, + "loss": 0.074, + "step": 13270 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002574358974358974, + "loss": 0.077, + "step": 13280 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002574038461538461, + "loss": 0.0844, + "step": 13290 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025737179487179485, + "loss": 0.0804, + "step": 13300 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025733974358974353, + "loss": 0.0784, + "step": 13310 + }, + { + "epoch": 0.28, + "learning_rate": 0.00025730769230769226, + "loss": 0.0857, + "step": 13320 + }, + { + "epoch": 0.28, + "learning_rate": 0.000257275641025641, + "loss": 0.0727, + "step": 13330 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025724358974358973, + "loss": 0.0784, + "step": 13340 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002572115384615384, + "loss": 0.0899, + "step": 13350 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025717948717948714, + "loss": 0.073, + "step": 13360 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025714743589743587, + "loss": 0.0835, + "step": 13370 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025711538461538455, + "loss": 0.0799, + "step": 13380 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002570833333333333, + "loss": 0.0816, + "step": 13390 + }, + { + "epoch": 0.29, + "learning_rate": 0.000257051282051282, + "loss": 0.0753, + "step": 13400 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025701923076923075, + "loss": 0.0796, + "step": 13410 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025698717948717943, + "loss": 0.0866, + "step": 13420 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025695512820512816, + "loss": 0.0775, + "step": 13430 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002569230769230769, + "loss": 0.0806, + "step": 13440 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025689102564102563, + "loss": 0.0834, + "step": 13450 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002568589743589743, + "loss": 0.0901, + "step": 13460 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025682692307692304, + "loss": 0.0868, + "step": 13470 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002567948717948718, + "loss": 0.1011, + "step": 13480 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025676282051282045, + "loss": 0.0813, + "step": 13490 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002567307692307692, + "loss": 0.0891, + "step": 13500 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002566987179487179, + "loss": 0.0766, + "step": 13510 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025666666666666665, + "loss": 0.0793, + "step": 13520 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025663461538461533, + "loss": 0.0767, + "step": 13530 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025660256410256407, + "loss": 0.072, + "step": 13540 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002565705128205128, + "loss": 0.0764, + "step": 13550 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002565384615384615, + "loss": 0.0719, + "step": 13560 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002565064102564102, + "loss": 0.071, + "step": 13570 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025647435897435894, + "loss": 0.0804, + "step": 13580 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002564423076923077, + "loss": 0.0735, + "step": 13590 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025641025641025636, + "loss": 0.0798, + "step": 13600 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002563782051282051, + "loss": 0.0818, + "step": 13610 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002563461538461538, + "loss": 0.0618, + "step": 13620 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025631410256410256, + "loss": 0.0849, + "step": 13630 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025628205128205123, + "loss": 0.0874, + "step": 13640 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025624999999999997, + "loss": 0.0777, + "step": 13650 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002562179487179487, + "loss": 0.0772, + "step": 13660 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002561858974358974, + "loss": 0.0828, + "step": 13670 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002561538461538461, + "loss": 0.0851, + "step": 13680 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025612179487179485, + "loss": 0.0744, + "step": 13690 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002560897435897436, + "loss": 0.0761, + "step": 13700 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025605769230769226, + "loss": 0.0738, + "step": 13710 + }, + { + "epoch": 0.29, + "learning_rate": 0.000256025641025641, + "loss": 0.0715, + "step": 13720 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002559935897435897, + "loss": 0.0735, + "step": 13730 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002559615384615384, + "loss": 0.076, + "step": 13740 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025592948717948714, + "loss": 0.0689, + "step": 13750 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025589743589743587, + "loss": 0.0742, + "step": 13760 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002558653846153846, + "loss": 0.0671, + "step": 13770 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002558333333333333, + "loss": 0.0753, + "step": 13780 + }, + { + "epoch": 0.29, + "learning_rate": 0.000255801282051282, + "loss": 0.0754, + "step": 13790 + }, + { + "epoch": 0.29, + "learning_rate": 0.00025576923076923075, + "loss": 0.0684, + "step": 13800 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002557371794871795, + "loss": 0.0774, + "step": 13810 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025570512820512816, + "loss": 0.0669, + "step": 13820 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002556730769230769, + "loss": 0.0679, + "step": 13830 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002556410256410256, + "loss": 0.0828, + "step": 13840 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002556089743589743, + "loss": 0.0759, + "step": 13850 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025557692307692304, + "loss": 0.0822, + "step": 13860 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025554487179487177, + "loss": 0.0837, + "step": 13870 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002555128205128205, + "loss": 0.0729, + "step": 13880 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002554807692307692, + "loss": 0.073, + "step": 13890 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002554487179487179, + "loss": 0.0869, + "step": 13900 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025541666666666665, + "loss": 0.0761, + "step": 13910 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025538461538461533, + "loss": 0.0754, + "step": 13920 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025535256410256406, + "loss": 0.0812, + "step": 13930 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002553205128205128, + "loss": 0.065, + "step": 13940 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025528846153846153, + "loss": 0.0727, + "step": 13950 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002552564102564102, + "loss": 0.072, + "step": 13960 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025522435897435894, + "loss": 0.07, + "step": 13970 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002551923076923077, + "loss": 0.0797, + "step": 13980 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002551602564102564, + "loss": 0.0863, + "step": 13990 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002551282051282051, + "loss": 0.0864, + "step": 14000 + }, + { + "epoch": 0.3, + "eval_cer": 0.9239135881210419, + "eval_loss": 0.049238067120313644, + "eval_runtime": 121.0334, + "eval_samples_per_second": 16.524, + "eval_steps_per_second": 4.131, + "step": 14000 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002550961538461538, + "loss": 0.0746, + "step": 14010 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025506410256410255, + "loss": 0.0814, + "step": 14020 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025503205128205123, + "loss": 0.0801, + "step": 14030 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025499999999999996, + "loss": 0.0832, + "step": 14040 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002549679487179487, + "loss": 0.0786, + "step": 14050 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025493589743589743, + "loss": 0.0792, + "step": 14060 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002549038461538461, + "loss": 0.0956, + "step": 14070 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025487179487179484, + "loss": 0.0812, + "step": 14080 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002548397435897436, + "loss": 0.0794, + "step": 14090 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025480769230769225, + "loss": 0.0764, + "step": 14100 + }, + { + "epoch": 0.3, + "learning_rate": 0.000254775641025641, + "loss": 0.0795, + "step": 14110 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002547435897435897, + "loss": 0.0782, + "step": 14120 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025471153846153845, + "loss": 0.0842, + "step": 14130 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025467948717948713, + "loss": 0.0674, + "step": 14140 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025464743589743587, + "loss": 0.0858, + "step": 14150 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002546153846153846, + "loss": 0.0764, + "step": 14160 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025458333333333333, + "loss": 0.0849, + "step": 14170 + }, + { + "epoch": 0.3, + "learning_rate": 0.000254551282051282, + "loss": 0.0857, + "step": 14180 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025451923076923074, + "loss": 0.0727, + "step": 14190 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002544871794871795, + "loss": 0.0724, + "step": 14200 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025445512820512816, + "loss": 0.0747, + "step": 14210 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002544230769230769, + "loss": 0.0659, + "step": 14220 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002543910256410256, + "loss": 0.0817, + "step": 14230 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025435897435897436, + "loss": 0.0771, + "step": 14240 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025432692307692304, + "loss": 0.0776, + "step": 14250 + }, + { + "epoch": 0.3, + "learning_rate": 0.00025429487179487177, + "loss": 0.0864, + "step": 14260 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002542628205128205, + "loss": 0.0678, + "step": 14270 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002542307692307692, + "loss": 0.0739, + "step": 14280 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002541987179487179, + "loss": 0.0765, + "step": 14290 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025416666666666665, + "loss": 0.0699, + "step": 14300 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002541346153846154, + "loss": 0.0753, + "step": 14310 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025410256410256406, + "loss": 0.0751, + "step": 14320 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002540705128205128, + "loss": 0.0727, + "step": 14330 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002540384615384615, + "loss": 0.0786, + "step": 14340 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002540064102564102, + "loss": 0.0707, + "step": 14350 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025397435897435894, + "loss": 0.0728, + "step": 14360 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025394230769230767, + "loss": 0.0766, + "step": 14370 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002539102564102564, + "loss": 0.0796, + "step": 14380 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002538782051282051, + "loss": 0.0832, + "step": 14390 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002538461538461538, + "loss": 0.0737, + "step": 14400 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025381410256410255, + "loss": 0.0731, + "step": 14410 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002537820512820513, + "loss": 0.0785, + "step": 14420 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025374999999999996, + "loss": 0.0686, + "step": 14430 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002537179487179487, + "loss": 0.0812, + "step": 14440 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025368589743589743, + "loss": 0.085, + "step": 14450 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002536538461538461, + "loss": 0.0745, + "step": 14460 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025362179487179484, + "loss": 0.0725, + "step": 14470 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025358974358974357, + "loss": 0.0806, + "step": 14480 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002535576923076923, + "loss": 0.0687, + "step": 14490 + }, + { + "epoch": 0.31, + "learning_rate": 0.000253525641025641, + "loss": 0.088, + "step": 14500 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002534935897435897, + "loss": 0.0814, + "step": 14510 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025346153846153845, + "loss": 0.0725, + "step": 14520 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025342948717948713, + "loss": 0.076, + "step": 14530 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025339743589743586, + "loss": 0.0798, + "step": 14540 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002533653846153846, + "loss": 0.0728, + "step": 14550 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025333333333333333, + "loss": 0.0692, + "step": 14560 + }, + { + "epoch": 0.31, + "learning_rate": 0.000253301282051282, + "loss": 0.0723, + "step": 14570 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025326923076923074, + "loss": 0.0734, + "step": 14580 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002532371794871795, + "loss": 0.0771, + "step": 14590 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002532051282051282, + "loss": 0.0837, + "step": 14600 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002531730769230769, + "loss": 0.0811, + "step": 14610 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002531410256410256, + "loss": 0.0754, + "step": 14620 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025310897435897435, + "loss": 0.0854, + "step": 14630 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025307692307692303, + "loss": 0.0749, + "step": 14640 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025304487179487176, + "loss": 0.0805, + "step": 14650 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002530128205128205, + "loss": 0.0774, + "step": 14660 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025298076923076923, + "loss": 0.07, + "step": 14670 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002529487179487179, + "loss": 0.075, + "step": 14680 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025291666666666664, + "loss": 0.0741, + "step": 14690 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002528846153846154, + "loss": 0.0643, + "step": 14700 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025285256410256406, + "loss": 0.0686, + "step": 14710 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002528205128205128, + "loss": 0.0747, + "step": 14720 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002527884615384615, + "loss": 0.0699, + "step": 14730 + }, + { + "epoch": 0.31, + "learning_rate": 0.00025275641025641025, + "loss": 0.0732, + "step": 14740 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025272435897435893, + "loss": 0.0765, + "step": 14750 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025269230769230767, + "loss": 0.0751, + "step": 14760 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002526602564102564, + "loss": 0.0738, + "step": 14770 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025262820512820513, + "loss": 0.078, + "step": 14780 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002525961538461538, + "loss": 0.0762, + "step": 14790 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025256410256410255, + "loss": 0.0784, + "step": 14800 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002525320512820513, + "loss": 0.0763, + "step": 14810 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025249999999999996, + "loss": 0.085, + "step": 14820 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002524679487179487, + "loss": 0.0696, + "step": 14830 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002524358974358974, + "loss": 0.0686, + "step": 14840 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025240384615384616, + "loss": 0.0712, + "step": 14850 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025237179487179484, + "loss": 0.0834, + "step": 14860 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025233974358974357, + "loss": 0.081, + "step": 14870 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002523076923076923, + "loss": 0.0862, + "step": 14880 + }, + { + "epoch": 0.32, + "learning_rate": 0.000252275641025641, + "loss": 0.0806, + "step": 14890 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002522435897435897, + "loss": 0.0737, + "step": 14900 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025221153846153845, + "loss": 0.0727, + "step": 14910 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002521794871794872, + "loss": 0.0685, + "step": 14920 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025214743589743586, + "loss": 0.067, + "step": 14930 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002521153846153846, + "loss": 0.0712, + "step": 14940 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002520833333333333, + "loss": 0.0685, + "step": 14950 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025205128205128206, + "loss": 0.0684, + "step": 14960 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025201923076923074, + "loss": 0.0743, + "step": 14970 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025198717948717947, + "loss": 0.077, + "step": 14980 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002519551282051282, + "loss": 0.0728, + "step": 14990 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002519230769230769, + "loss": 0.0806, + "step": 15000 + }, + { + "epoch": 0.32, + "eval_cer": 0.9239297243173328, + "eval_loss": 0.04714610427618027, + "eval_runtime": 120.5789, + "eval_samples_per_second": 16.587, + "eval_steps_per_second": 4.147, + "step": 15000 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002518910256410256, + "loss": 0.0669, + "step": 15010 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025185897435897435, + "loss": 0.0741, + "step": 15020 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002518269230769231, + "loss": 0.0814, + "step": 15030 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025179487179487176, + "loss": 0.0704, + "step": 15040 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002517628205128205, + "loss": 0.0749, + "step": 15050 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025173076923076923, + "loss": 0.0828, + "step": 15060 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002516987179487179, + "loss": 0.066, + "step": 15070 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025166666666666664, + "loss": 0.0818, + "step": 15080 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002516346153846154, + "loss": 0.0739, + "step": 15090 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002516025641025641, + "loss": 0.0707, + "step": 15100 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002515705128205128, + "loss": 0.0875, + "step": 15110 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002515384615384615, + "loss": 0.0798, + "step": 15120 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025150641025641025, + "loss": 0.0733, + "step": 15130 + }, + { + "epoch": 0.32, + "learning_rate": 0.000251474358974359, + "loss": 0.0714, + "step": 15140 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025144230769230766, + "loss": 0.0717, + "step": 15150 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002514102564102564, + "loss": 0.0634, + "step": 15160 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025137820512820513, + "loss": 0.0704, + "step": 15170 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002513461538461538, + "loss": 0.069, + "step": 15180 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025131410256410254, + "loss": 0.0747, + "step": 15190 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002512820512820513, + "loss": 0.0786, + "step": 15200 + }, + { + "epoch": 0.32, + "learning_rate": 0.00025125, + "loss": 0.0786, + "step": 15210 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002512179487179487, + "loss": 0.0733, + "step": 15220 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002511858974358974, + "loss": 0.0724, + "step": 15230 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025115384615384615, + "loss": 0.0728, + "step": 15240 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025112179487179483, + "loss": 0.0765, + "step": 15250 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025108974358974357, + "loss": 0.0722, + "step": 15260 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002510576923076923, + "loss": 0.0599, + "step": 15270 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025102564102564103, + "loss": 0.0859, + "step": 15280 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002509935897435897, + "loss": 0.0687, + "step": 15290 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025096153846153844, + "loss": 0.0707, + "step": 15300 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002509294871794872, + "loss": 0.0671, + "step": 15310 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025089743589743586, + "loss": 0.0747, + "step": 15320 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002508653846153846, + "loss": 0.0672, + "step": 15330 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002508333333333333, + "loss": 0.0787, + "step": 15340 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025080128205128206, + "loss": 0.095, + "step": 15350 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025076923076923073, + "loss": 0.0677, + "step": 15360 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025073717948717947, + "loss": 0.0759, + "step": 15370 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002507051282051282, + "loss": 0.0804, + "step": 15380 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025067307692307693, + "loss": 0.0741, + "step": 15390 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002506410256410256, + "loss": 0.0699, + "step": 15400 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025060897435897435, + "loss": 0.0797, + "step": 15410 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002505769230769231, + "loss": 0.0721, + "step": 15420 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025054487179487176, + "loss": 0.0687, + "step": 15430 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002505128205128205, + "loss": 0.0773, + "step": 15440 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002504807692307692, + "loss": 0.0806, + "step": 15450 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025044871794871796, + "loss": 0.0771, + "step": 15460 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025041666666666664, + "loss": 0.0721, + "step": 15470 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025038461538461537, + "loss": 0.0838, + "step": 15480 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002503525641025641, + "loss": 0.0713, + "step": 15490 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002503205128205128, + "loss": 0.0781, + "step": 15500 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002502884615384615, + "loss": 0.0676, + "step": 15510 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025025641025641025, + "loss": 0.0722, + "step": 15520 + }, + { + "epoch": 0.33, + "learning_rate": 0.000250224358974359, + "loss": 0.0761, + "step": 15530 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025019230769230766, + "loss": 0.0662, + "step": 15540 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002501602564102564, + "loss": 0.0871, + "step": 15550 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002501282051282051, + "loss": 0.0945, + "step": 15560 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025009615384615386, + "loss": 0.0761, + "step": 15570 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025006410256410254, + "loss": 0.0749, + "step": 15580 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025003205128205127, + "loss": 0.0768, + "step": 15590 + }, + { + "epoch": 0.33, + "learning_rate": 0.00025, + "loss": 0.0759, + "step": 15600 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002499679487179487, + "loss": 0.0671, + "step": 15610 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002499358974358974, + "loss": 0.0712, + "step": 15620 + }, + { + "epoch": 0.33, + "learning_rate": 0.00024990384615384615, + "loss": 0.0711, + "step": 15630 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002498717948717949, + "loss": 0.0719, + "step": 15640 + }, + { + "epoch": 0.33, + "learning_rate": 0.00024983974358974356, + "loss": 0.0706, + "step": 15650 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002498076923076923, + "loss": 0.0773, + "step": 15660 + }, + { + "epoch": 0.33, + "learning_rate": 0.00024977564102564103, + "loss": 0.0686, + "step": 15670 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002497435897435897, + "loss": 0.0847, + "step": 15680 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024971153846153844, + "loss": 0.0775, + "step": 15690 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002496794871794872, + "loss": 0.0744, + "step": 15700 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002496474358974359, + "loss": 0.0758, + "step": 15710 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002496153846153846, + "loss": 0.0735, + "step": 15720 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002495833333333333, + "loss": 0.0758, + "step": 15730 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024955128205128205, + "loss": 0.0714, + "step": 15740 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002495192307692308, + "loss": 0.0722, + "step": 15750 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024948717948717946, + "loss": 0.0813, + "step": 15760 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002494551282051282, + "loss": 0.0735, + "step": 15770 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024942307692307693, + "loss": 0.0775, + "step": 15780 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002493910256410256, + "loss": 0.0782, + "step": 15790 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024935897435897434, + "loss": 0.0761, + "step": 15800 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002493269230769231, + "loss": 0.0732, + "step": 15810 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002492948717948718, + "loss": 0.0819, + "step": 15820 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002492628205128205, + "loss": 0.0701, + "step": 15830 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002492307692307692, + "loss": 0.0751, + "step": 15840 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024919871794871795, + "loss": 0.0755, + "step": 15850 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024916666666666663, + "loss": 0.0729, + "step": 15860 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024913461538461537, + "loss": 0.0746, + "step": 15870 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002491025641025641, + "loss": 0.0692, + "step": 15880 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024907051282051283, + "loss": 0.073, + "step": 15890 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002490384615384615, + "loss": 0.072, + "step": 15900 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024900641025641025, + "loss": 0.0782, + "step": 15910 + }, + { + "epoch": 0.34, + "learning_rate": 0.000248974358974359, + "loss": 0.0845, + "step": 15920 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002489423076923077, + "loss": 0.0658, + "step": 15930 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002489102564102564, + "loss": 0.0763, + "step": 15940 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002488782051282051, + "loss": 0.074, + "step": 15950 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024884615384615386, + "loss": 0.0697, + "step": 15960 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024881410256410254, + "loss": 0.0765, + "step": 15970 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024878205128205127, + "loss": 0.0874, + "step": 15980 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024875, + "loss": 0.0761, + "step": 15990 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024871794871794874, + "loss": 0.0808, + "step": 16000 + }, + { + "epoch": 0.34, + "eval_cer": 0.9238159216698075, + "eval_loss": 0.04827723652124405, + "eval_runtime": 120.6289, + "eval_samples_per_second": 16.58, + "eval_steps_per_second": 4.145, + "step": 16000 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002486858974358974, + "loss": 0.0714, + "step": 16010 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024865384615384615, + "loss": 0.0766, + "step": 16020 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002486217948717949, + "loss": 0.0788, + "step": 16030 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024858974358974356, + "loss": 0.085, + "step": 16040 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002485576923076923, + "loss": 0.071, + "step": 16050 + }, + { + "epoch": 0.34, + "learning_rate": 0.000248525641025641, + "loss": 0.069, + "step": 16060 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024849358974358976, + "loss": 0.0787, + "step": 16070 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024846153846153844, + "loss": 0.0729, + "step": 16080 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002484294871794871, + "loss": 0.0741, + "step": 16090 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024839743589743585, + "loss": 0.0732, + "step": 16100 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002483653846153846, + "loss": 0.074, + "step": 16110 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002483333333333333, + "loss": 0.078, + "step": 16120 + }, + { + "epoch": 0.34, + "learning_rate": 0.000248301282051282, + "loss": 0.0835, + "step": 16130 + }, + { + "epoch": 0.34, + "learning_rate": 0.00024826923076923073, + "loss": 0.0729, + "step": 16140 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024823717948717946, + "loss": 0.0826, + "step": 16150 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002482051282051282, + "loss": 0.0687, + "step": 16160 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002481730769230769, + "loss": 0.0691, + "step": 16170 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002481410256410256, + "loss": 0.0823, + "step": 16180 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024810897435897434, + "loss": 0.0658, + "step": 16190 + }, + { + "epoch": 0.35, + "learning_rate": 0.000248076923076923, + "loss": 0.0741, + "step": 16200 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024804487179487175, + "loss": 0.075, + "step": 16210 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002480128205128205, + "loss": 0.0704, + "step": 16220 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002479807692307692, + "loss": 0.079, + "step": 16230 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002479487179487179, + "loss": 0.0791, + "step": 16240 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024791666666666663, + "loss": 0.0692, + "step": 16250 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024788461538461536, + "loss": 0.0751, + "step": 16260 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024785256410256404, + "loss": 0.0694, + "step": 16270 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002478205128205128, + "loss": 0.0846, + "step": 16280 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002477884615384615, + "loss": 0.078, + "step": 16290 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024775641025641024, + "loss": 0.0687, + "step": 16300 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002477243589743589, + "loss": 0.0693, + "step": 16310 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024769230769230765, + "loss": 0.0784, + "step": 16320 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002476602564102564, + "loss": 0.074, + "step": 16330 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024762820512820507, + "loss": 0.0723, + "step": 16340 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002475961538461538, + "loss": 0.0767, + "step": 16350 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024756410256410253, + "loss": 0.0733, + "step": 16360 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024753205128205127, + "loss": 0.0754, + "step": 16370 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024749999999999994, + "loss": 0.0707, + "step": 16380 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002474679487179487, + "loss": 0.0706, + "step": 16390 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002474358974358974, + "loss": 0.0774, + "step": 16400 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024740384615384614, + "loss": 0.0682, + "step": 16410 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002473717948717948, + "loss": 0.0739, + "step": 16420 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024733974358974356, + "loss": 0.0821, + "step": 16430 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002473076923076923, + "loss": 0.0748, + "step": 16440 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024727564102564097, + "loss": 0.0693, + "step": 16450 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002472435897435897, + "loss": 0.0769, + "step": 16460 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024721153846153843, + "loss": 0.076, + "step": 16470 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024717948717948717, + "loss": 0.0717, + "step": 16480 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024714743589743585, + "loss": 0.0678, + "step": 16490 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002471153846153846, + "loss": 0.0711, + "step": 16500 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002470833333333333, + "loss": 0.066, + "step": 16510 + }, + { + "epoch": 0.35, + "learning_rate": 0.000247051282051282, + "loss": 0.0676, + "step": 16520 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002470192307692307, + "loss": 0.0751, + "step": 16530 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024698717948717946, + "loss": 0.0724, + "step": 16540 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002469551282051282, + "loss": 0.0712, + "step": 16550 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024692307692307687, + "loss": 0.071, + "step": 16560 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002468910256410256, + "loss": 0.0708, + "step": 16570 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024685897435897434, + "loss": 0.0738, + "step": 16580 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024682692307692307, + "loss": 0.0725, + "step": 16590 + }, + { + "epoch": 0.35, + "learning_rate": 0.00024679487179487175, + "loss": 0.0728, + "step": 16600 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002467628205128205, + "loss": 0.0659, + "step": 16610 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002467307692307692, + "loss": 0.0641, + "step": 16620 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002466987179487179, + "loss": 0.0699, + "step": 16630 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002466666666666666, + "loss": 0.0696, + "step": 16640 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024663461538461536, + "loss": 0.0727, + "step": 16650 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002466025641025641, + "loss": 0.0731, + "step": 16660 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024657051282051277, + "loss": 0.0787, + "step": 16670 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002465384615384615, + "loss": 0.0757, + "step": 16680 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024650641025641024, + "loss": 0.0669, + "step": 16690 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002464743589743589, + "loss": 0.0686, + "step": 16700 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024644230769230765, + "loss": 0.0769, + "step": 16710 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002464102564102564, + "loss": 0.0718, + "step": 16720 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002463782051282051, + "loss": 0.0755, + "step": 16730 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002463461538461538, + "loss": 0.0824, + "step": 16740 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024631410256410253, + "loss": 0.0668, + "step": 16750 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024628205128205126, + "loss": 0.0727, + "step": 16760 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024625, + "loss": 0.0677, + "step": 16770 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002462179487179487, + "loss": 0.0685, + "step": 16780 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002461858974358974, + "loss": 0.0771, + "step": 16790 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024615384615384614, + "loss": 0.0668, + "step": 16800 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002461217948717948, + "loss": 0.0657, + "step": 16810 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024608974358974355, + "loss": 0.0679, + "step": 16820 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002460576923076923, + "loss": 0.0734, + "step": 16830 + }, + { + "epoch": 0.36, + "learning_rate": 0.000246025641025641, + "loss": 0.0752, + "step": 16840 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002459935897435897, + "loss": 0.0731, + "step": 16850 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024596153846153843, + "loss": 0.0783, + "step": 16860 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024592948717948716, + "loss": 0.0622, + "step": 16870 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024589743589743584, + "loss": 0.066, + "step": 16880 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002458653846153846, + "loss": 0.073, + "step": 16890 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002458333333333333, + "loss": 0.0685, + "step": 16900 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024580128205128204, + "loss": 0.072, + "step": 16910 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002457692307692307, + "loss": 0.0655, + "step": 16920 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024573717948717945, + "loss": 0.0682, + "step": 16930 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002457051282051282, + "loss": 0.0696, + "step": 16940 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002456730769230769, + "loss": 0.0836, + "step": 16950 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002456410256410256, + "loss": 0.0739, + "step": 16960 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024560897435897433, + "loss": 0.0846, + "step": 16970 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024557692307692307, + "loss": 0.0673, + "step": 16980 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024554487179487175, + "loss": 0.0806, + "step": 16990 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002455128205128205, + "loss": 0.071, + "step": 17000 + }, + { + "epoch": 0.36, + "eval_cer": 0.9239348199582667, + "eval_loss": 0.046869438141584396, + "eval_runtime": 120.9846, + "eval_samples_per_second": 16.531, + "eval_steps_per_second": 4.133, + "step": 17000 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002454807692307692, + "loss": 0.0648, + "step": 17010 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024544871794871794, + "loss": 0.073, + "step": 17020 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002454166666666666, + "loss": 0.0797, + "step": 17030 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024538461538461536, + "loss": 0.0685, + "step": 17040 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002453525641025641, + "loss": 0.0718, + "step": 17050 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024532051282051277, + "loss": 0.0738, + "step": 17060 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002452884615384615, + "loss": 0.0712, + "step": 17070 + }, + { + "epoch": 0.36, + "learning_rate": 0.00024525641025641024, + "loss": 0.0647, + "step": 17080 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024522435897435897, + "loss": 0.0739, + "step": 17090 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024519230769230765, + "loss": 0.076, + "step": 17100 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002451602564102564, + "loss": 0.0738, + "step": 17110 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002451282051282051, + "loss": 0.0783, + "step": 17120 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024509615384615385, + "loss": 0.07, + "step": 17130 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002450641025641025, + "loss": 0.0732, + "step": 17140 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024503205128205126, + "loss": 0.0639, + "step": 17150 + }, + { + "epoch": 0.37, + "learning_rate": 0.000245, + "loss": 0.0746, + "step": 17160 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024496794871794867, + "loss": 0.0741, + "step": 17170 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002449358974358974, + "loss": 0.0704, + "step": 17180 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024490384615384614, + "loss": 0.0691, + "step": 17190 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024487179487179487, + "loss": 0.0655, + "step": 17200 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024483974358974355, + "loss": 0.0686, + "step": 17210 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002448076923076923, + "loss": 0.0711, + "step": 17220 + }, + { + "epoch": 0.37, + "learning_rate": 0.000244775641025641, + "loss": 0.0594, + "step": 17230 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002447435897435897, + "loss": 0.0679, + "step": 17240 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024471153846153843, + "loss": 0.0667, + "step": 17250 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024467948717948716, + "loss": 0.0627, + "step": 17260 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002446474358974359, + "loss": 0.0715, + "step": 17270 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002446153846153846, + "loss": 0.066, + "step": 17280 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002445833333333333, + "loss": 0.072, + "step": 17290 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024455128205128204, + "loss": 0.0681, + "step": 17300 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002445192307692307, + "loss": 0.0834, + "step": 17310 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024448717948717945, + "loss": 0.0645, + "step": 17320 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002444551282051282, + "loss": 0.0722, + "step": 17330 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002444230769230769, + "loss": 0.0663, + "step": 17340 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002443910256410256, + "loss": 0.0689, + "step": 17350 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024435897435897433, + "loss": 0.0648, + "step": 17360 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024432692307692306, + "loss": 0.0614, + "step": 17370 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002442948717948718, + "loss": 0.0685, + "step": 17380 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002442628205128205, + "loss": 0.0663, + "step": 17390 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002442307692307692, + "loss": 0.0689, + "step": 17400 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024419871794871794, + "loss": 0.0735, + "step": 17410 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002441666666666666, + "loss": 0.0718, + "step": 17420 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024413461538461535, + "loss": 0.0774, + "step": 17430 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002441025641025641, + "loss": 0.0685, + "step": 17440 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002440705128205128, + "loss": 0.0695, + "step": 17450 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024403846153846153, + "loss": 0.0643, + "step": 17460 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024400641025641023, + "loss": 0.0662, + "step": 17470 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024397435897435897, + "loss": 0.0663, + "step": 17480 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024394230769230767, + "loss": 0.0718, + "step": 17490 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024391025641025638, + "loss": 0.0783, + "step": 17500 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002438782051282051, + "loss": 0.074, + "step": 17510 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024384615384615382, + "loss": 0.0703, + "step": 17520 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024381410256410255, + "loss": 0.0781, + "step": 17530 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024378205128205126, + "loss": 0.0803, + "step": 17540 + }, + { + "epoch": 0.37, + "learning_rate": 0.00024375, + "loss": 0.0821, + "step": 17550 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002437179487179487, + "loss": 0.0868, + "step": 17560 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024368589743589743, + "loss": 0.0719, + "step": 17570 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024365384615384613, + "loss": 0.0726, + "step": 17580 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024362179487179484, + "loss": 0.0663, + "step": 17590 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024358974358974357, + "loss": 0.0726, + "step": 17600 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024355769230769228, + "loss": 0.0692, + "step": 17610 + }, + { + "epoch": 0.38, + "learning_rate": 0.000243525641025641, + "loss": 0.0726, + "step": 17620 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024349358974358972, + "loss": 0.0659, + "step": 17630 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024346153846153845, + "loss": 0.0714, + "step": 17640 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024342948717948716, + "loss": 0.0616, + "step": 17650 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002433974358974359, + "loss": 0.0654, + "step": 17660 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002433653846153846, + "loss": 0.0645, + "step": 17670 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002433333333333333, + "loss": 0.0585, + "step": 17680 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024330128205128204, + "loss": 0.0717, + "step": 17690 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024326923076923074, + "loss": 0.0696, + "step": 17700 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024323717948717948, + "loss": 0.0683, + "step": 17710 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024320512820512818, + "loss": 0.0741, + "step": 17720 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024317307692307691, + "loss": 0.0733, + "step": 17730 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024314102564102562, + "loss": 0.0737, + "step": 17740 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024310897435897433, + "loss": 0.0621, + "step": 17750 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024307692307692306, + "loss": 0.0673, + "step": 17760 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024304487179487177, + "loss": 0.0644, + "step": 17770 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002430128205128205, + "loss": 0.0667, + "step": 17780 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002429807692307692, + "loss": 0.065, + "step": 17790 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024294871794871794, + "loss": 0.0729, + "step": 17800 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024291666666666664, + "loss": 0.0694, + "step": 17810 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024288461538461538, + "loss": 0.0694, + "step": 17820 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024285256410256408, + "loss": 0.0757, + "step": 17830 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002428205128205128, + "loss": 0.0684, + "step": 17840 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024278846153846152, + "loss": 0.0786, + "step": 17850 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024275641025641023, + "loss": 0.073, + "step": 17860 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024272435897435896, + "loss": 0.0625, + "step": 17870 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024269230769230767, + "loss": 0.0614, + "step": 17880 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002426602564102564, + "loss": 0.0694, + "step": 17890 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002426282051282051, + "loss": 0.0666, + "step": 17900 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024259615384615384, + "loss": 0.0804, + "step": 17910 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024256410256410255, + "loss": 0.0666, + "step": 17920 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024253205128205125, + "loss": 0.073, + "step": 17930 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024249999999999999, + "loss": 0.0655, + "step": 17940 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002424679487179487, + "loss": 0.0786, + "step": 17950 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024243589743589742, + "loss": 0.069, + "step": 17960 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024240384615384613, + "loss": 0.068, + "step": 17970 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024237179487179486, + "loss": 0.0705, + "step": 17980 + }, + { + "epoch": 0.38, + "learning_rate": 0.00024233974358974357, + "loss": 0.0705, + "step": 17990 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002423076923076923, + "loss": 0.0661, + "step": 18000 + }, + { + "epoch": 0.38, + "eval_cer": 0.9239382170522227, + "eval_loss": 0.044636666774749756, + "eval_runtime": 120.8374, + "eval_samples_per_second": 16.551, + "eval_steps_per_second": 4.138, + "step": 18000 + }, + { + "epoch": 0.38, + "learning_rate": 0.000242275641025641, + "loss": 0.0696, + "step": 18010 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024224358974358972, + "loss": 0.0628, + "step": 18020 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024221153846153845, + "loss": 0.0554, + "step": 18030 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024217948717948715, + "loss": 0.0744, + "step": 18040 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002421474358974359, + "loss": 0.0732, + "step": 18050 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002421153846153846, + "loss": 0.0759, + "step": 18060 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024208333333333333, + "loss": 0.0667, + "step": 18070 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024205128205128203, + "loss": 0.0762, + "step": 18080 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024201923076923077, + "loss": 0.0683, + "step": 18090 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024198717948717947, + "loss": 0.0739, + "step": 18100 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024195512820512818, + "loss": 0.0772, + "step": 18110 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002419230769230769, + "loss": 0.0695, + "step": 18120 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024189102564102562, + "loss": 0.0724, + "step": 18130 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024185897435897435, + "loss": 0.0686, + "step": 18140 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024182692307692306, + "loss": 0.0754, + "step": 18150 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002417948717948718, + "loss": 0.0666, + "step": 18160 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002417628205128205, + "loss": 0.0663, + "step": 18170 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024173076923076923, + "loss": 0.0716, + "step": 18180 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024169871794871793, + "loss": 0.0631, + "step": 18190 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024166666666666664, + "loss": 0.0678, + "step": 18200 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024163461538461537, + "loss": 0.0763, + "step": 18210 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024160256410256408, + "loss": 0.076, + "step": 18220 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002415705128205128, + "loss": 0.0599, + "step": 18230 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024153846153846152, + "loss": 0.0747, + "step": 18240 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024150641025641025, + "loss": 0.0735, + "step": 18250 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024147435897435896, + "loss": 0.0774, + "step": 18260 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002414423076923077, + "loss": 0.065, + "step": 18270 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002414102564102564, + "loss": 0.0657, + "step": 18280 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002413782051282051, + "loss": 0.0663, + "step": 18290 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024134615384615384, + "loss": 0.0779, + "step": 18300 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024131410256410254, + "loss": 0.063, + "step": 18310 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024128205128205128, + "loss": 0.0733, + "step": 18320 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024124999999999998, + "loss": 0.0676, + "step": 18330 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024121794871794872, + "loss": 0.065, + "step": 18340 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024118589743589742, + "loss": 0.0675, + "step": 18350 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024115384615384615, + "loss": 0.067, + "step": 18360 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024112179487179486, + "loss": 0.075, + "step": 18370 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024108974358974357, + "loss": 0.0707, + "step": 18380 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002410576923076923, + "loss": 0.0709, + "step": 18390 + }, + { + "epoch": 0.39, + "learning_rate": 0.000241025641025641, + "loss": 0.07, + "step": 18400 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024099358974358974, + "loss": 0.0629, + "step": 18410 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024096153846153845, + "loss": 0.0625, + "step": 18420 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024092948717948718, + "loss": 0.0615, + "step": 18430 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024089743589743588, + "loss": 0.0627, + "step": 18440 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024086538461538462, + "loss": 0.065, + "step": 18450 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024083333333333332, + "loss": 0.0623, + "step": 18460 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024080128205128203, + "loss": 0.079, + "step": 18470 + }, + { + "epoch": 0.39, + "learning_rate": 0.00024076923076923076, + "loss": 0.0681, + "step": 18480 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024073717948717947, + "loss": 0.0708, + "step": 18490 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002407051282051282, + "loss": 0.0721, + "step": 18500 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002406730769230769, + "loss": 0.0666, + "step": 18510 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024064102564102564, + "loss": 0.07, + "step": 18520 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024060897435897435, + "loss": 0.0726, + "step": 18530 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024057692307692308, + "loss": 0.0643, + "step": 18540 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024054487179487179, + "loss": 0.0692, + "step": 18550 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002405128205128205, + "loss": 0.0662, + "step": 18560 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024048076923076923, + "loss": 0.0667, + "step": 18570 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024044871794871793, + "loss": 0.0644, + "step": 18580 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024041666666666666, + "loss": 0.0709, + "step": 18590 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024038461538461537, + "loss": 0.0626, + "step": 18600 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002403525641025641, + "loss": 0.0659, + "step": 18610 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002403205128205128, + "loss": 0.0771, + "step": 18620 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024028846153846154, + "loss": 0.0713, + "step": 18630 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024025641025641025, + "loss": 0.0837, + "step": 18640 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024022435897435896, + "loss": 0.0683, + "step": 18650 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002401923076923077, + "loss": 0.0826, + "step": 18660 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002401602564102564, + "loss": 0.0737, + "step": 18670 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024012820512820513, + "loss": 0.0718, + "step": 18680 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024009615384615383, + "loss": 0.0753, + "step": 18690 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024006410256410257, + "loss": 0.0784, + "step": 18700 + }, + { + "epoch": 0.4, + "learning_rate": 0.00024003205128205127, + "loss": 0.0664, + "step": 18710 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023999999999999998, + "loss": 0.067, + "step": 18720 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002399679487179487, + "loss": 0.0718, + "step": 18730 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023993589743589742, + "loss": 0.0648, + "step": 18740 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023990384615384615, + "loss": 0.0705, + "step": 18750 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023987179487179486, + "loss": 0.067, + "step": 18760 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002398397435897436, + "loss": 0.0803, + "step": 18770 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002398076923076923, + "loss": 0.0771, + "step": 18780 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023977564102564103, + "loss": 0.0694, + "step": 18790 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023974358974358974, + "loss": 0.0754, + "step": 18800 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023971153846153844, + "loss": 0.0649, + "step": 18810 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023967948717948717, + "loss": 0.0821, + "step": 18820 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023964743589743588, + "loss": 0.0754, + "step": 18830 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023961538461538461, + "loss": 0.0689, + "step": 18840 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023958333333333332, + "loss": 0.0685, + "step": 18850 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023955128205128205, + "loss": 0.0706, + "step": 18860 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023951923076923076, + "loss": 0.0686, + "step": 18870 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002394871794871795, + "loss": 0.0629, + "step": 18880 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002394551282051282, + "loss": 0.0644, + "step": 18890 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002394230769230769, + "loss": 0.0633, + "step": 18900 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023939102564102564, + "loss": 0.0725, + "step": 18910 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023935897435897434, + "loss": 0.0621, + "step": 18920 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023932692307692308, + "loss": 0.0715, + "step": 18930 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023929487179487178, + "loss": 0.0652, + "step": 18940 + }, + { + "epoch": 0.4, + "learning_rate": 0.00023926282051282052, + "loss": 0.0651, + "step": 18950 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023923076923076922, + "loss": 0.0779, + "step": 18960 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023919871794871796, + "loss": 0.0658, + "step": 18970 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023916666666666666, + "loss": 0.0789, + "step": 18980 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023913461538461537, + "loss": 0.0656, + "step": 18990 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002391025641025641, + "loss": 0.0641, + "step": 19000 + }, + { + "epoch": 0.41, + "eval_cer": 0.9239110403005749, + "eval_loss": 0.04373706504702568, + "eval_runtime": 120.813, + "eval_samples_per_second": 16.555, + "eval_steps_per_second": 4.139, + "step": 19000 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002390705128205128, + "loss": 0.0703, + "step": 19010 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002390384615384615, + "loss": 0.0681, + "step": 19020 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023900641025641022, + "loss": 0.0602, + "step": 19030 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023897435897435892, + "loss": 0.0686, + "step": 19040 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023894230769230766, + "loss": 0.0655, + "step": 19050 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023891025641025636, + "loss": 0.0599, + "step": 19060 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002388782051282051, + "loss": 0.0683, + "step": 19070 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002388461538461538, + "loss": 0.0658, + "step": 19080 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023881410256410254, + "loss": 0.0637, + "step": 19090 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023878205128205124, + "loss": 0.0779, + "step": 19100 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023874999999999998, + "loss": 0.0679, + "step": 19110 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023871794871794868, + "loss": 0.0718, + "step": 19120 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002386858974358974, + "loss": 0.0716, + "step": 19130 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023865384615384612, + "loss": 0.0692, + "step": 19140 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023862179487179483, + "loss": 0.0706, + "step": 19150 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023858974358974356, + "loss": 0.0763, + "step": 19160 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023855769230769227, + "loss": 0.0679, + "step": 19170 + }, + { + "epoch": 0.41, + "learning_rate": 0.000238525641025641, + "loss": 0.0658, + "step": 19180 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002384935897435897, + "loss": 0.0703, + "step": 19190 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023846153846153844, + "loss": 0.0741, + "step": 19200 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023842948717948714, + "loss": 0.0728, + "step": 19210 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023839743589743585, + "loss": 0.0709, + "step": 19220 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023836538461538458, + "loss": 0.0656, + "step": 19230 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002383333333333333, + "loss": 0.0694, + "step": 19240 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023830128205128202, + "loss": 0.0656, + "step": 19250 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023826923076923073, + "loss": 0.0602, + "step": 19260 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023823717948717946, + "loss": 0.0641, + "step": 19270 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023820512820512817, + "loss": 0.0718, + "step": 19280 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002381730769230769, + "loss": 0.0734, + "step": 19290 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002381410256410256, + "loss": 0.0805, + "step": 19300 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002381089743589743, + "loss": 0.0659, + "step": 19310 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023807692307692305, + "loss": 0.0643, + "step": 19320 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023804487179487175, + "loss": 0.0732, + "step": 19330 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023801282051282049, + "loss": 0.0655, + "step": 19340 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002379807692307692, + "loss": 0.0722, + "step": 19350 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023794871794871793, + "loss": 0.0661, + "step": 19360 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023791666666666663, + "loss": 0.0634, + "step": 19370 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023788461538461536, + "loss": 0.0788, + "step": 19380 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023785256410256407, + "loss": 0.0707, + "step": 19390 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023782051282051278, + "loss": 0.0669, + "step": 19400 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002377884615384615, + "loss": 0.0706, + "step": 19410 + }, + { + "epoch": 0.41, + "learning_rate": 0.00023775641025641022, + "loss": 0.0822, + "step": 19420 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023772435897435895, + "loss": 0.0736, + "step": 19430 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023769230769230765, + "loss": 0.0677, + "step": 19440 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002376602564102564, + "loss": 0.0693, + "step": 19450 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002376282051282051, + "loss": 0.076, + "step": 19460 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023759615384615383, + "loss": 0.0769, + "step": 19470 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023756410256410253, + "loss": 0.0694, + "step": 19480 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023753205128205124, + "loss": 0.0653, + "step": 19490 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023749999999999997, + "loss": 0.0752, + "step": 19500 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023746794871794868, + "loss": 0.0701, + "step": 19510 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002374358974358974, + "loss": 0.0681, + "step": 19520 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023740384615384612, + "loss": 0.0636, + "step": 19530 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023737179487179485, + "loss": 0.073, + "step": 19540 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023733974358974356, + "loss": 0.0697, + "step": 19550 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002373076923076923, + "loss": 0.0698, + "step": 19560 + }, + { + "epoch": 0.42, + "learning_rate": 0.000237275641025641, + "loss": 0.0679, + "step": 19570 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002372435897435897, + "loss": 0.0671, + "step": 19580 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023721153846153844, + "loss": 0.0604, + "step": 19590 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023717948717948714, + "loss": 0.0673, + "step": 19600 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023714743589743587, + "loss": 0.0697, + "step": 19610 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023711538461538458, + "loss": 0.0726, + "step": 19620 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023708333333333331, + "loss": 0.0818, + "step": 19630 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023705128205128202, + "loss": 0.0802, + "step": 19640 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023701923076923073, + "loss": 0.0741, + "step": 19650 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023698717948717946, + "loss": 0.0726, + "step": 19660 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023695512820512816, + "loss": 0.0711, + "step": 19670 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002369230769230769, + "loss": 0.0745, + "step": 19680 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002368910256410256, + "loss": 0.0613, + "step": 19690 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023685897435897434, + "loss": 0.0678, + "step": 19700 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023682692307692304, + "loss": 0.0686, + "step": 19710 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023679487179487178, + "loss": 0.0656, + "step": 19720 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023676282051282048, + "loss": 0.0711, + "step": 19730 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002367307692307692, + "loss": 0.0675, + "step": 19740 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023669871794871792, + "loss": 0.0668, + "step": 19750 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023666666666666663, + "loss": 0.0807, + "step": 19760 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023663461538461536, + "loss": 0.0679, + "step": 19770 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023660256410256407, + "loss": 0.0695, + "step": 19780 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002365705128205128, + "loss": 0.0652, + "step": 19790 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002365384615384615, + "loss": 0.0649, + "step": 19800 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023650641025641024, + "loss": 0.074, + "step": 19810 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023647435897435895, + "loss": 0.0657, + "step": 19820 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023644230769230765, + "loss": 0.0668, + "step": 19830 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023641025641025638, + "loss": 0.0677, + "step": 19840 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002363782051282051, + "loss": 0.0696, + "step": 19850 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023634615384615382, + "loss": 0.076, + "step": 19860 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023631410256410253, + "loss": 0.0634, + "step": 19870 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023628205128205126, + "loss": 0.0658, + "step": 19880 + }, + { + "epoch": 0.42, + "learning_rate": 0.00023624999999999997, + "loss": 0.0731, + "step": 19890 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002362179487179487, + "loss": 0.0607, + "step": 19900 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002361858974358974, + "loss": 0.0671, + "step": 19910 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023615384615384611, + "loss": 0.0647, + "step": 19920 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023612179487179485, + "loss": 0.0626, + "step": 19930 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023608974358974355, + "loss": 0.0651, + "step": 19940 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002360576923076923, + "loss": 0.0672, + "step": 19950 + }, + { + "epoch": 0.43, + "learning_rate": 0.000236025641025641, + "loss": 0.0646, + "step": 19960 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023599358974358973, + "loss": 0.0628, + "step": 19970 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023596153846153843, + "loss": 0.0616, + "step": 19980 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023592948717948717, + "loss": 0.078, + "step": 19990 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023589743589743587, + "loss": 0.0686, + "step": 20000 + }, + { + "epoch": 0.43, + "eval_cer": 0.9238329071395874, + "eval_loss": 0.042833201587200165, + "eval_runtime": 120.4102, + "eval_samples_per_second": 16.61, + "eval_steps_per_second": 4.152, + "step": 20000 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023586538461538458, + "loss": 0.0643, + "step": 20010 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002358333333333333, + "loss": 0.071, + "step": 20020 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023580128205128202, + "loss": 0.0749, + "step": 20030 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023576923076923075, + "loss": 0.0732, + "step": 20040 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023573717948717946, + "loss": 0.0748, + "step": 20050 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002357051282051282, + "loss": 0.0607, + "step": 20060 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002356730769230769, + "loss": 0.0716, + "step": 20070 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023564102564102563, + "loss": 0.0728, + "step": 20080 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023560897435897433, + "loss": 0.0687, + "step": 20090 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023557692307692304, + "loss": 0.0642, + "step": 20100 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023554487179487177, + "loss": 0.0672, + "step": 20110 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023551282051282048, + "loss": 0.0688, + "step": 20120 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002354807692307692, + "loss": 0.0707, + "step": 20130 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023544871794871792, + "loss": 0.0679, + "step": 20140 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023541666666666665, + "loss": 0.0652, + "step": 20150 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023538461538461536, + "loss": 0.0788, + "step": 20160 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002353525641025641, + "loss": 0.0755, + "step": 20170 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002353205128205128, + "loss": 0.0678, + "step": 20180 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002352884615384615, + "loss": 0.0661, + "step": 20190 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023525641025641024, + "loss": 0.0608, + "step": 20200 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023522435897435894, + "loss": 0.0696, + "step": 20210 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023519230769230768, + "loss": 0.0663, + "step": 20220 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023516025641025638, + "loss": 0.0721, + "step": 20230 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023512820512820511, + "loss": 0.0625, + "step": 20240 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023509615384615382, + "loss": 0.0631, + "step": 20250 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023506410256410255, + "loss": 0.0677, + "step": 20260 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023503205128205126, + "loss": 0.0639, + "step": 20270 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023499999999999997, + "loss": 0.0639, + "step": 20280 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002349679487179487, + "loss": 0.0647, + "step": 20290 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002349358974358974, + "loss": 0.0581, + "step": 20300 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023490384615384614, + "loss": 0.0639, + "step": 20310 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023487179487179484, + "loss": 0.0634, + "step": 20320 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023483974358974358, + "loss": 0.0716, + "step": 20330 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023480769230769228, + "loss": 0.0618, + "step": 20340 + }, + { + "epoch": 0.43, + "learning_rate": 0.00023477564102564102, + "loss": 0.06, + "step": 20350 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023474358974358972, + "loss": 0.0669, + "step": 20360 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023471153846153843, + "loss": 0.0618, + "step": 20370 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023467948717948716, + "loss": 0.0714, + "step": 20380 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023464743589743587, + "loss": 0.0596, + "step": 20390 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002346153846153846, + "loss": 0.0725, + "step": 20400 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002345833333333333, + "loss": 0.0678, + "step": 20410 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023455128205128204, + "loss": 0.0639, + "step": 20420 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023451923076923075, + "loss": 0.0655, + "step": 20430 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023448717948717948, + "loss": 0.062, + "step": 20440 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023445512820512819, + "loss": 0.065, + "step": 20450 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002344230769230769, + "loss": 0.0665, + "step": 20460 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023439102564102562, + "loss": 0.0629, + "step": 20470 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023435897435897433, + "loss": 0.07, + "step": 20480 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023432692307692306, + "loss": 0.0611, + "step": 20490 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023429487179487177, + "loss": 0.062, + "step": 20500 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002342628205128205, + "loss": 0.0623, + "step": 20510 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002342307692307692, + "loss": 0.0623, + "step": 20520 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023419871794871794, + "loss": 0.068, + "step": 20530 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023416666666666665, + "loss": 0.0615, + "step": 20540 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023413461538461535, + "loss": 0.0654, + "step": 20550 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002341025641025641, + "loss": 0.0674, + "step": 20560 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002340705128205128, + "loss": 0.0734, + "step": 20570 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023403846153846153, + "loss": 0.0598, + "step": 20580 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023400641025641023, + "loss": 0.0689, + "step": 20590 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023397435897435897, + "loss": 0.0728, + "step": 20600 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023394230769230767, + "loss": 0.0691, + "step": 20610 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023391025641025638, + "loss": 0.0711, + "step": 20620 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002338782051282051, + "loss": 0.0746, + "step": 20630 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023384615384615382, + "loss": 0.0669, + "step": 20640 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023381410256410255, + "loss": 0.0625, + "step": 20650 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023378205128205126, + "loss": 0.0583, + "step": 20660 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023375, + "loss": 0.0681, + "step": 20670 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002337179487179487, + "loss": 0.0686, + "step": 20680 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023368589743589743, + "loss": 0.0596, + "step": 20690 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023365384615384613, + "loss": 0.0734, + "step": 20700 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023362179487179484, + "loss": 0.0771, + "step": 20710 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023358974358974357, + "loss": 0.0668, + "step": 20720 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023355769230769228, + "loss": 0.0647, + "step": 20730 + }, + { + "epoch": 0.44, + "learning_rate": 0.000233525641025641, + "loss": 0.0613, + "step": 20740 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023349358974358972, + "loss": 0.07, + "step": 20750 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023346153846153845, + "loss": 0.068, + "step": 20760 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023342948717948716, + "loss": 0.0627, + "step": 20770 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002333974358974359, + "loss": 0.0699, + "step": 20780 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002333653846153846, + "loss": 0.059, + "step": 20790 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002333333333333333, + "loss": 0.0775, + "step": 20800 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023330128205128204, + "loss": 0.0684, + "step": 20810 + }, + { + "epoch": 0.44, + "learning_rate": 0.00023326923076923074, + "loss": 0.0582, + "step": 20820 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023323717948717948, + "loss": 0.0693, + "step": 20830 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023320512820512818, + "loss": 0.0637, + "step": 20840 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023317307692307692, + "loss": 0.0595, + "step": 20850 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023314102564102562, + "loss": 0.0647, + "step": 20860 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023310897435897435, + "loss": 0.0715, + "step": 20870 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023307692307692306, + "loss": 0.0664, + "step": 20880 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023304487179487177, + "loss": 0.0679, + "step": 20890 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002330128205128205, + "loss": 0.0702, + "step": 20900 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002329807692307692, + "loss": 0.0597, + "step": 20910 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023294871794871794, + "loss": 0.0636, + "step": 20920 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023291666666666665, + "loss": 0.0648, + "step": 20930 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023288461538461538, + "loss": 0.0637, + "step": 20940 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023285256410256408, + "loss": 0.0656, + "step": 20950 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023282051282051282, + "loss": 0.0538, + "step": 20960 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023278846153846152, + "loss": 0.0633, + "step": 20970 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023275641025641023, + "loss": 0.0588, + "step": 20980 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023272435897435896, + "loss": 0.0654, + "step": 20990 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023269230769230767, + "loss": 0.0597, + "step": 21000 + }, + { + "epoch": 0.45, + "eval_cer": 0.9238193187637636, + "eval_loss": 0.043088994920253754, + "eval_runtime": 120.9689, + "eval_samples_per_second": 16.533, + "eval_steps_per_second": 4.133, + "step": 21000 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002326602564102564, + "loss": 0.0638, + "step": 21010 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002326282051282051, + "loss": 0.0657, + "step": 21020 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023259615384615384, + "loss": 0.0628, + "step": 21030 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023256410256410255, + "loss": 0.0636, + "step": 21040 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023253205128205128, + "loss": 0.0693, + "step": 21050 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023249999999999999, + "loss": 0.068, + "step": 21060 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002324679487179487, + "loss": 0.064, + "step": 21070 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023243589743589743, + "loss": 0.0641, + "step": 21080 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023240384615384613, + "loss": 0.0605, + "step": 21090 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023237179487179486, + "loss": 0.0571, + "step": 21100 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023233974358974357, + "loss": 0.0707, + "step": 21110 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002323076923076923, + "loss": 0.0775, + "step": 21120 + }, + { + "epoch": 0.45, + "learning_rate": 0.000232275641025641, + "loss": 0.0827, + "step": 21130 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023224358974358974, + "loss": 0.0721, + "step": 21140 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023221153846153845, + "loss": 0.0692, + "step": 21150 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023217948717948716, + "loss": 0.0653, + "step": 21160 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002321474358974359, + "loss": 0.0652, + "step": 21170 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002321153846153846, + "loss": 0.0565, + "step": 21180 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023208333333333333, + "loss": 0.0708, + "step": 21190 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023205128205128203, + "loss": 0.064, + "step": 21200 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023201923076923077, + "loss": 0.0644, + "step": 21210 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023198717948717947, + "loss": 0.0695, + "step": 21220 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002319551282051282, + "loss": 0.0619, + "step": 21230 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002319230769230769, + "loss": 0.074, + "step": 21240 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023189102564102562, + "loss": 0.0597, + "step": 21250 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023185897435897435, + "loss": 0.066, + "step": 21260 + }, + { + "epoch": 0.45, + "learning_rate": 0.00023182692307692306, + "loss": 0.0708, + "step": 21270 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002317948717948718, + "loss": 0.0634, + "step": 21280 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002317628205128205, + "loss": 0.0615, + "step": 21290 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023173076923076923, + "loss": 0.0669, + "step": 21300 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023169871794871794, + "loss": 0.0599, + "step": 21310 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023166666666666667, + "loss": 0.0636, + "step": 21320 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023163461538461538, + "loss": 0.0661, + "step": 21330 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023160256410256408, + "loss": 0.055, + "step": 21340 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023157051282051281, + "loss": 0.0577, + "step": 21350 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023153846153846152, + "loss": 0.0639, + "step": 21360 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023150641025641025, + "loss": 0.0619, + "step": 21370 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023147435897435896, + "loss": 0.0642, + "step": 21380 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002314423076923077, + "loss": 0.0656, + "step": 21390 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002314102564102564, + "loss": 0.0621, + "step": 21400 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023137820512820513, + "loss": 0.0669, + "step": 21410 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023134615384615384, + "loss": 0.0678, + "step": 21420 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023131410256410254, + "loss": 0.0663, + "step": 21430 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023128205128205128, + "loss": 0.0625, + "step": 21440 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023124999999999998, + "loss": 0.0653, + "step": 21450 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023121794871794872, + "loss": 0.0584, + "step": 21460 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023118589743589742, + "loss": 0.0668, + "step": 21470 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023115384615384616, + "loss": 0.0621, + "step": 21480 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023112179487179486, + "loss": 0.0601, + "step": 21490 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002310897435897436, + "loss": 0.0746, + "step": 21500 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002310576923076923, + "loss": 0.0651, + "step": 21510 + }, + { + "epoch": 0.46, + "learning_rate": 0.000231025641025641, + "loss": 0.0651, + "step": 21520 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023099358974358974, + "loss": 0.0669, + "step": 21530 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023096153846153845, + "loss": 0.0645, + "step": 21540 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023092948717948718, + "loss": 0.0681, + "step": 21550 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023089743589743589, + "loss": 0.0614, + "step": 21560 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023086538461538462, + "loss": 0.058, + "step": 21570 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023083333333333332, + "loss": 0.0588, + "step": 21580 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023080128205128203, + "loss": 0.0757, + "step": 21590 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023076923076923076, + "loss": 0.0615, + "step": 21600 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023073717948717947, + "loss": 0.0653, + "step": 21610 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002307051282051282, + "loss": 0.0638, + "step": 21620 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002306730769230769, + "loss": 0.0548, + "step": 21630 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023064102564102564, + "loss": 0.0622, + "step": 21640 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023060897435897435, + "loss": 0.0653, + "step": 21650 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023057692307692308, + "loss": 0.0677, + "step": 21660 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002305448717948718, + "loss": 0.0704, + "step": 21670 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002305128205128205, + "loss": 0.056, + "step": 21680 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023048076923076923, + "loss": 0.0631, + "step": 21690 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023044871794871793, + "loss": 0.0605, + "step": 21700 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023041666666666667, + "loss": 0.0652, + "step": 21710 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023038461538461537, + "loss": 0.0616, + "step": 21720 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002303525641025641, + "loss": 0.0721, + "step": 21730 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002303205128205128, + "loss": 0.0628, + "step": 21740 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023028846153846154, + "loss": 0.0647, + "step": 21750 + }, + { + "epoch": 0.46, + "learning_rate": 0.00023025641025641025, + "loss": 0.0557, + "step": 21760 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023022435897435896, + "loss": 0.0556, + "step": 21770 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002301923076923077, + "loss": 0.0612, + "step": 21780 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002301602564102564, + "loss": 0.0638, + "step": 21790 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023012820512820513, + "loss": 0.0623, + "step": 21800 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023009615384615383, + "loss": 0.0578, + "step": 21810 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023006410256410257, + "loss": 0.0593, + "step": 21820 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023003205128205127, + "loss": 0.0658, + "step": 21830 + }, + { + "epoch": 0.47, + "learning_rate": 0.00023, + "loss": 0.0753, + "step": 21840 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002299679487179487, + "loss": 0.0646, + "step": 21850 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022993589743589742, + "loss": 0.0696, + "step": 21860 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022990384615384615, + "loss": 0.06, + "step": 21870 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022987179487179486, + "loss": 0.057, + "step": 21880 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002298397435897436, + "loss": 0.0835, + "step": 21890 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002298076923076923, + "loss": 0.0673, + "step": 21900 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022977564102564103, + "loss": 0.0631, + "step": 21910 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022974358974358974, + "loss": 0.0626, + "step": 21920 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022971153846153847, + "loss": 0.0708, + "step": 21930 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022967948717948715, + "loss": 0.0566, + "step": 21940 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022964743589743585, + "loss": 0.0696, + "step": 21950 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002296153846153846, + "loss": 0.0621, + "step": 21960 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002295833333333333, + "loss": 0.0625, + "step": 21970 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022955128205128203, + "loss": 0.0676, + "step": 21980 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022951923076923073, + "loss": 0.0798, + "step": 21990 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022948717948717944, + "loss": 0.0585, + "step": 22000 + }, + { + "epoch": 0.47, + "eval_cer": 0.9238286607721424, + "eval_loss": 0.041744619607925415, + "eval_runtime": 121.0049, + "eval_samples_per_second": 16.528, + "eval_steps_per_second": 4.132, + "step": 22000 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022945512820512817, + "loss": 0.0659, + "step": 22010 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022942307692307688, + "loss": 0.0775, + "step": 22020 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002293910256410256, + "loss": 0.0718, + "step": 22030 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022935897435897432, + "loss": 0.0659, + "step": 22040 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022932692307692305, + "loss": 0.0677, + "step": 22050 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022929487179487176, + "loss": 0.0668, + "step": 22060 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002292628205128205, + "loss": 0.0674, + "step": 22070 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002292307692307692, + "loss": 0.067, + "step": 22080 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002291987179487179, + "loss": 0.0683, + "step": 22090 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022916666666666664, + "loss": 0.0668, + "step": 22100 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022913461538461534, + "loss": 0.0664, + "step": 22110 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022910256410256407, + "loss": 0.0656, + "step": 22120 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022907051282051278, + "loss": 0.0557, + "step": 22130 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022903846153846151, + "loss": 0.0673, + "step": 22140 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022900641025641022, + "loss": 0.0726, + "step": 22150 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022897435897435895, + "loss": 0.0818, + "step": 22160 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022894230769230766, + "loss": 0.0655, + "step": 22170 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022891025641025637, + "loss": 0.066, + "step": 22180 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002288782051282051, + "loss": 0.067, + "step": 22190 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002288461538461538, + "loss": 0.0681, + "step": 22200 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022881410256410254, + "loss": 0.0609, + "step": 22210 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022878205128205124, + "loss": 0.0677, + "step": 22220 + }, + { + "epoch": 0.47, + "learning_rate": 0.00022874999999999998, + "loss": 0.0642, + "step": 22230 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022871794871794868, + "loss": 0.0656, + "step": 22240 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022868589743589742, + "loss": 0.0666, + "step": 22250 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022865384615384612, + "loss": 0.0717, + "step": 22260 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022862179487179483, + "loss": 0.0752, + "step": 22270 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022858974358974356, + "loss": 0.065, + "step": 22280 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022855769230769227, + "loss": 0.0637, + "step": 22290 + }, + { + "epoch": 0.48, + "learning_rate": 0.000228525641025641, + "loss": 0.061, + "step": 22300 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002284935897435897, + "loss": 0.0729, + "step": 22310 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022846153846153844, + "loss": 0.0689, + "step": 22320 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022842948717948715, + "loss": 0.0734, + "step": 22330 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022839743589743588, + "loss": 0.0593, + "step": 22340 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022836538461538458, + "loss": 0.075, + "step": 22350 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002283333333333333, + "loss": 0.0699, + "step": 22360 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022830128205128202, + "loss": 0.0714, + "step": 22370 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022826923076923073, + "loss": 0.0596, + "step": 22380 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022823717948717946, + "loss": 0.0685, + "step": 22390 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022820512820512817, + "loss": 0.06, + "step": 22400 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002281730769230769, + "loss": 0.0578, + "step": 22410 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002281410256410256, + "loss": 0.0735, + "step": 22420 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022810897435897431, + "loss": 0.0632, + "step": 22430 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022807692307692305, + "loss": 0.0648, + "step": 22440 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022804487179487175, + "loss": 0.0575, + "step": 22450 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002280128205128205, + "loss": 0.0709, + "step": 22460 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002279807692307692, + "loss": 0.066, + "step": 22470 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022794871794871793, + "loss": 0.0648, + "step": 22480 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022791666666666663, + "loss": 0.0695, + "step": 22490 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022788461538461537, + "loss": 0.0573, + "step": 22500 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022785256410256407, + "loss": 0.0622, + "step": 22510 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022782051282051278, + "loss": 0.0548, + "step": 22520 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002277884615384615, + "loss": 0.0584, + "step": 22530 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022775641025641022, + "loss": 0.065, + "step": 22540 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022772435897435895, + "loss": 0.0691, + "step": 22550 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022769230769230766, + "loss": 0.0709, + "step": 22560 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002276602564102564, + "loss": 0.0594, + "step": 22570 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002276282051282051, + "loss": 0.0611, + "step": 22580 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022759615384615383, + "loss": 0.062, + "step": 22590 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022756410256410253, + "loss": 0.0619, + "step": 22600 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022753205128205124, + "loss": 0.0648, + "step": 22610 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022749999999999997, + "loss": 0.0661, + "step": 22620 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022746794871794868, + "loss": 0.0665, + "step": 22630 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002274358974358974, + "loss": 0.0675, + "step": 22640 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022740384615384612, + "loss": 0.0891, + "step": 22650 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022737179487179485, + "loss": 0.0694, + "step": 22660 + }, + { + "epoch": 0.48, + "learning_rate": 0.00022733974358974356, + "loss": 0.062, + "step": 22670 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002273076923076923, + "loss": 0.0639, + "step": 22680 + }, + { + "epoch": 0.48, + "learning_rate": 0.000227275641025641, + "loss": 0.0626, + "step": 22690 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002272435897435897, + "loss": 0.0762, + "step": 22700 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022721153846153844, + "loss": 0.069, + "step": 22710 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022717948717948714, + "loss": 0.0708, + "step": 22720 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022714743589743588, + "loss": 0.0602, + "step": 22730 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022711538461538458, + "loss": 0.063, + "step": 22740 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022708333333333331, + "loss": 0.0673, + "step": 22750 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022705128205128202, + "loss": 0.0692, + "step": 22760 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022701923076923075, + "loss": 0.0656, + "step": 22770 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022698717948717946, + "loss": 0.0666, + "step": 22780 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022695512820512817, + "loss": 0.0707, + "step": 22790 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002269230769230769, + "loss": 0.0658, + "step": 22800 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002268910256410256, + "loss": 0.0626, + "step": 22810 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022685897435897434, + "loss": 0.0571, + "step": 22820 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022682692307692304, + "loss": 0.0634, + "step": 22830 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022679487179487178, + "loss": 0.0649, + "step": 22840 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022676282051282048, + "loss": 0.0661, + "step": 22850 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022673076923076922, + "loss": 0.0692, + "step": 22860 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022669871794871792, + "loss": 0.0593, + "step": 22870 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022666666666666663, + "loss": 0.0681, + "step": 22880 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022663461538461536, + "loss": 0.0621, + "step": 22890 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022660256410256407, + "loss": 0.0645, + "step": 22900 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002265705128205128, + "loss": 0.0654, + "step": 22910 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002265384615384615, + "loss": 0.0601, + "step": 22920 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022650641025641024, + "loss": 0.0631, + "step": 22930 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022647435897435895, + "loss": 0.0692, + "step": 22940 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022644230769230768, + "loss": 0.0702, + "step": 22950 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022641025641025639, + "loss": 0.0602, + "step": 22960 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002263782051282051, + "loss": 0.0695, + "step": 22970 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022634615384615382, + "loss": 0.076, + "step": 22980 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022631410256410253, + "loss": 0.0707, + "step": 22990 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022628205128205126, + "loss": 0.0675, + "step": 23000 + }, + { + "epoch": 0.49, + "eval_cer": 0.9238337564130764, + "eval_loss": 0.040570374578237534, + "eval_runtime": 120.8568, + "eval_samples_per_second": 16.549, + "eval_steps_per_second": 4.137, + "step": 23000 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022624999999999997, + "loss": 0.0648, + "step": 23010 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002262179487179487, + "loss": 0.0635, + "step": 23020 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002261858974358974, + "loss": 0.0578, + "step": 23030 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022615384615384614, + "loss": 0.0654, + "step": 23040 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022612179487179485, + "loss": 0.0604, + "step": 23050 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022608974358974355, + "loss": 0.0611, + "step": 23060 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002260576923076923, + "loss": 0.0573, + "step": 23070 + }, + { + "epoch": 0.49, + "learning_rate": 0.000226025641025641, + "loss": 0.061, + "step": 23080 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022599358974358973, + "loss": 0.0686, + "step": 23090 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022596153846153843, + "loss": 0.0766, + "step": 23100 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022592948717948717, + "loss": 0.0649, + "step": 23110 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022589743589743587, + "loss": 0.0655, + "step": 23120 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002258653846153846, + "loss": 0.072, + "step": 23130 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002258333333333333, + "loss": 0.0726, + "step": 23140 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022580128205128202, + "loss": 0.0625, + "step": 23150 + }, + { + "epoch": 0.49, + "learning_rate": 0.00022576923076923075, + "loss": 0.0623, + "step": 23160 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022573717948717946, + "loss": 0.0597, + "step": 23170 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002257051282051282, + "loss": 0.0668, + "step": 23180 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002256730769230769, + "loss": 0.0663, + "step": 23190 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022564102564102563, + "loss": 0.0657, + "step": 23200 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022560897435897433, + "loss": 0.0609, + "step": 23210 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022557692307692307, + "loss": 0.0621, + "step": 23220 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022554487179487177, + "loss": 0.0566, + "step": 23230 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022551282051282048, + "loss": 0.0556, + "step": 23240 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002254807692307692, + "loss": 0.0595, + "step": 23250 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022544871794871792, + "loss": 0.0674, + "step": 23260 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022541666666666665, + "loss": 0.0611, + "step": 23270 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022538461538461536, + "loss": 0.0709, + "step": 23280 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002253525641025641, + "loss": 0.0614, + "step": 23290 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002253205128205128, + "loss": 0.0688, + "step": 23300 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022528846153846153, + "loss": 0.0741, + "step": 23310 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022525641025641024, + "loss": 0.0559, + "step": 23320 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022522435897435894, + "loss": 0.0772, + "step": 23330 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022519230769230768, + "loss": 0.0646, + "step": 23340 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022516025641025638, + "loss": 0.0673, + "step": 23350 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022512820512820512, + "loss": 0.0641, + "step": 23360 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022509615384615382, + "loss": 0.0568, + "step": 23370 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022506410256410255, + "loss": 0.0638, + "step": 23380 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022503205128205126, + "loss": 0.0687, + "step": 23390 + }, + { + "epoch": 0.5, + "learning_rate": 0.000225, + "loss": 0.0652, + "step": 23400 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002249679487179487, + "loss": 0.0585, + "step": 23410 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002249358974358974, + "loss": 0.0543, + "step": 23420 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022490384615384614, + "loss": 0.0608, + "step": 23430 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022487179487179485, + "loss": 0.0687, + "step": 23440 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022483974358974358, + "loss": 0.0608, + "step": 23450 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022480769230769228, + "loss": 0.0679, + "step": 23460 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022477564102564102, + "loss": 0.063, + "step": 23470 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022474358974358972, + "loss": 0.0677, + "step": 23480 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022471153846153843, + "loss": 0.0603, + "step": 23490 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022467948717948716, + "loss": 0.0584, + "step": 23500 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022464743589743587, + "loss": 0.0611, + "step": 23510 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002246153846153846, + "loss": 0.0621, + "step": 23520 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002245833333333333, + "loss": 0.061, + "step": 23530 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022455128205128204, + "loss": 0.0604, + "step": 23540 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022451923076923075, + "loss": 0.0521, + "step": 23550 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022448717948717948, + "loss": 0.0577, + "step": 23560 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022445512820512819, + "loss": 0.0641, + "step": 23570 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002244230769230769, + "loss": 0.068, + "step": 23580 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022439102564102563, + "loss": 0.065, + "step": 23590 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022435897435897433, + "loss": 0.0587, + "step": 23600 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022432692307692306, + "loss": 0.0572, + "step": 23610 + }, + { + "epoch": 0.5, + "learning_rate": 0.00022429487179487177, + "loss": 0.0635, + "step": 23620 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002242628205128205, + "loss": 0.0636, + "step": 23630 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002242307692307692, + "loss": 0.0605, + "step": 23640 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022419871794871794, + "loss": 0.0596, + "step": 23650 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022416666666666665, + "loss": 0.0652, + "step": 23660 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022413461538461536, + "loss": 0.059, + "step": 23670 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002241025641025641, + "loss": 0.0619, + "step": 23680 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002240705128205128, + "loss": 0.0643, + "step": 23690 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022403846153846153, + "loss": 0.0598, + "step": 23700 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022400641025641023, + "loss": 0.0623, + "step": 23710 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022397435897435897, + "loss": 0.0711, + "step": 23720 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022394230769230767, + "loss": 0.0573, + "step": 23730 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002239102564102564, + "loss": 0.0589, + "step": 23740 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002238782051282051, + "loss": 0.0629, + "step": 23750 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022384615384615382, + "loss": 0.0676, + "step": 23760 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022381410256410255, + "loss": 0.0633, + "step": 23770 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022378205128205126, + "loss": 0.0691, + "step": 23780 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022375, + "loss": 0.065, + "step": 23790 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002237179487179487, + "loss": 0.0574, + "step": 23800 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022368589743589743, + "loss": 0.0623, + "step": 23810 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022365384615384614, + "loss": 0.0654, + "step": 23820 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022362179487179487, + "loss": 0.069, + "step": 23830 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022358974358974358, + "loss": 0.063, + "step": 23840 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022355769230769228, + "loss": 0.0706, + "step": 23850 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022352564102564101, + "loss": 0.0699, + "step": 23860 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022349358974358972, + "loss": 0.067, + "step": 23870 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022346153846153845, + "loss": 0.0727, + "step": 23880 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022342948717948716, + "loss": 0.0802, + "step": 23890 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002233974358974359, + "loss": 0.0681, + "step": 23900 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002233653846153846, + "loss": 0.068, + "step": 23910 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022333333333333333, + "loss": 0.0653, + "step": 23920 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022330128205128204, + "loss": 0.0617, + "step": 23930 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022326923076923074, + "loss": 0.0619, + "step": 23940 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022323717948717948, + "loss": 0.0533, + "step": 23950 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022320512820512818, + "loss": 0.0591, + "step": 23960 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022317307692307692, + "loss": 0.0659, + "step": 23970 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022314102564102562, + "loss": 0.0654, + "step": 23980 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022310897435897436, + "loss": 0.0677, + "step": 23990 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022307692307692306, + "loss": 0.0678, + "step": 24000 + }, + { + "epoch": 0.51, + "eval_cer": 0.9237760058158249, + "eval_loss": 0.03954209387302399, + "eval_runtime": 121.0225, + "eval_samples_per_second": 16.526, + "eval_steps_per_second": 4.131, + "step": 24000 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002230448717948718, + "loss": 0.0548, + "step": 24010 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002230128205128205, + "loss": 0.0571, + "step": 24020 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002229807692307692, + "loss": 0.0597, + "step": 24030 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022294871794871794, + "loss": 0.0576, + "step": 24040 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022291666666666665, + "loss": 0.0657, + "step": 24050 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022288461538461538, + "loss": 0.0616, + "step": 24060 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022285256410256409, + "loss": 0.0568, + "step": 24070 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022282051282051282, + "loss": 0.0611, + "step": 24080 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022278846153846152, + "loss": 0.0652, + "step": 24090 + }, + { + "epoch": 0.51, + "learning_rate": 0.00022275641025641026, + "loss": 0.0656, + "step": 24100 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022272435897435896, + "loss": 0.0597, + "step": 24110 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022269230769230767, + "loss": 0.0632, + "step": 24120 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002226602564102564, + "loss": 0.0688, + "step": 24130 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002226282051282051, + "loss": 0.0658, + "step": 24140 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022259615384615384, + "loss": 0.0652, + "step": 24150 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022256410256410255, + "loss": 0.055, + "step": 24160 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022253205128205128, + "loss": 0.0585, + "step": 24170 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002225, + "loss": 0.0637, + "step": 24180 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022246794871794872, + "loss": 0.0613, + "step": 24190 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022243589743589743, + "loss": 0.057, + "step": 24200 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022240384615384613, + "loss": 0.055, + "step": 24210 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022237179487179487, + "loss": 0.0618, + "step": 24220 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022233974358974357, + "loss": 0.0602, + "step": 24230 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002223076923076923, + "loss": 0.0602, + "step": 24240 + }, + { + "epoch": 0.52, + "learning_rate": 0.000222275641025641, + "loss": 0.0524, + "step": 24250 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022224358974358974, + "loss": 0.0598, + "step": 24260 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022221153846153845, + "loss": 0.0491, + "step": 24270 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022217948717948718, + "loss": 0.0647, + "step": 24280 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002221474358974359, + "loss": 0.0553, + "step": 24290 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002221153846153846, + "loss": 0.0555, + "step": 24300 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022208333333333333, + "loss": 0.0639, + "step": 24310 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022205128205128203, + "loss": 0.0605, + "step": 24320 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022201923076923077, + "loss": 0.0655, + "step": 24330 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022198717948717947, + "loss": 0.0671, + "step": 24340 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002219551282051282, + "loss": 0.0671, + "step": 24350 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002219230769230769, + "loss": 0.0649, + "step": 24360 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022189102564102565, + "loss": 0.0537, + "step": 24370 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022185897435897435, + "loss": 0.0675, + "step": 24380 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022182692307692306, + "loss": 0.0547, + "step": 24390 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002217948717948718, + "loss": 0.0584, + "step": 24400 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002217628205128205, + "loss": 0.064, + "step": 24410 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022173076923076923, + "loss": 0.0553, + "step": 24420 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022169871794871794, + "loss": 0.0665, + "step": 24430 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022166666666666667, + "loss": 0.0637, + "step": 24440 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022163461538461538, + "loss": 0.0649, + "step": 24450 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022160256410256408, + "loss": 0.0592, + "step": 24460 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022157051282051282, + "loss": 0.0553, + "step": 24470 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022153846153846152, + "loss": 0.0584, + "step": 24480 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022150641025641025, + "loss": 0.0606, + "step": 24490 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022147435897435896, + "loss": 0.0594, + "step": 24500 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002214423076923077, + "loss": 0.0544, + "step": 24510 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002214102564102564, + "loss": 0.0562, + "step": 24520 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022137820512820513, + "loss": 0.0647, + "step": 24530 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022134615384615384, + "loss": 0.061, + "step": 24540 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022131410256410254, + "loss": 0.06, + "step": 24550 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022128205128205128, + "loss": 0.0648, + "step": 24560 + }, + { + "epoch": 0.52, + "learning_rate": 0.00022124999999999998, + "loss": 0.0631, + "step": 24570 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022121794871794872, + "loss": 0.0534, + "step": 24580 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022118589743589742, + "loss": 0.0618, + "step": 24590 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022115384615384616, + "loss": 0.0589, + "step": 24600 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022112179487179486, + "loss": 0.0611, + "step": 24610 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002210897435897436, + "loss": 0.0562, + "step": 24620 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002210576923076923, + "loss": 0.0589, + "step": 24630 + }, + { + "epoch": 0.53, + "learning_rate": 0.000221025641025641, + "loss": 0.0544, + "step": 24640 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022099358974358974, + "loss": 0.0647, + "step": 24650 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022096153846153845, + "loss": 0.0524, + "step": 24660 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022092948717948718, + "loss": 0.0614, + "step": 24670 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022089743589743589, + "loss": 0.0573, + "step": 24680 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022086538461538462, + "loss": 0.0717, + "step": 24690 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022083333333333333, + "loss": 0.0608, + "step": 24700 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022080128205128206, + "loss": 0.0717, + "step": 24710 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022076923076923076, + "loss": 0.0658, + "step": 24720 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022073717948717947, + "loss": 0.0605, + "step": 24730 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002207051282051282, + "loss": 0.0616, + "step": 24740 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002206730769230769, + "loss": 0.0683, + "step": 24750 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022064102564102564, + "loss": 0.063, + "step": 24760 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022060897435897435, + "loss": 0.0671, + "step": 24770 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022057692307692308, + "loss": 0.0616, + "step": 24780 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002205448717948718, + "loss": 0.0744, + "step": 24790 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022051282051282052, + "loss": 0.0563, + "step": 24800 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022048076923076923, + "loss": 0.0635, + "step": 24810 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022044871794871793, + "loss": 0.0621, + "step": 24820 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022041666666666667, + "loss": 0.0562, + "step": 24830 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022038461538461537, + "loss": 0.0653, + "step": 24840 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002203525641025641, + "loss": 0.0625, + "step": 24850 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002203205128205128, + "loss": 0.0583, + "step": 24860 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002202884615384615, + "loss": 0.06, + "step": 24870 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022025641025641022, + "loss": 0.0637, + "step": 24880 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022022435897435893, + "loss": 0.0684, + "step": 24890 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022019230769230766, + "loss": 0.0617, + "step": 24900 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022016025641025637, + "loss": 0.0664, + "step": 24910 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002201282051282051, + "loss": 0.0546, + "step": 24920 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002200961538461538, + "loss": 0.0571, + "step": 24930 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022006410256410254, + "loss": 0.0652, + "step": 24940 + }, + { + "epoch": 0.53, + "learning_rate": 0.00022003205128205125, + "loss": 0.0562, + "step": 24950 + }, + { + "epoch": 0.53, + "learning_rate": 0.00021999999999999995, + "loss": 0.0687, + "step": 24960 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002199679487179487, + "loss": 0.0618, + "step": 24970 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002199358974358974, + "loss": 0.0604, + "step": 24980 + }, + { + "epoch": 0.53, + "learning_rate": 0.00021990384615384613, + "loss": 0.0598, + "step": 24990 + }, + { + "epoch": 0.53, + "learning_rate": 0.00021987179487179483, + "loss": 0.0581, + "step": 25000 + }, + { + "epoch": 0.53, + "eval_cer": 0.9238261129516755, + "eval_loss": 0.03929557651281357, + "eval_runtime": 120.8941, + "eval_samples_per_second": 16.543, + "eval_steps_per_second": 4.136, + "step": 25000 + }, + { + "epoch": 0.53, + "learning_rate": 0.00021983974358974357, + "loss": 0.0635, + "step": 25010 + }, + { + "epoch": 0.53, + "learning_rate": 0.00021980769230769227, + "loss": 0.0613, + "step": 25020 + }, + { + "epoch": 0.53, + "learning_rate": 0.000219775641025641, + "loss": 0.0605, + "step": 25030 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002197435897435897, + "loss": 0.059, + "step": 25040 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021971153846153842, + "loss": 0.0688, + "step": 25050 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021967948717948715, + "loss": 0.0593, + "step": 25060 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021964743589743586, + "loss": 0.058, + "step": 25070 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002196153846153846, + "loss": 0.0569, + "step": 25080 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002195833333333333, + "loss": 0.0592, + "step": 25090 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021955128205128203, + "loss": 0.0562, + "step": 25100 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021951923076923073, + "loss": 0.0595, + "step": 25110 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021948717948717947, + "loss": 0.0566, + "step": 25120 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021945512820512817, + "loss": 0.0502, + "step": 25130 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021942307692307688, + "loss": 0.0632, + "step": 25140 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002193910256410256, + "loss": 0.0647, + "step": 25150 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021935897435897432, + "loss": 0.0576, + "step": 25160 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021932692307692305, + "loss": 0.058, + "step": 25170 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021929487179487176, + "loss": 0.0596, + "step": 25180 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002192628205128205, + "loss": 0.0613, + "step": 25190 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002192307692307692, + "loss": 0.0601, + "step": 25200 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021919871794871793, + "loss": 0.0605, + "step": 25210 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021916666666666664, + "loss": 0.0579, + "step": 25220 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021913461538461534, + "loss": 0.0611, + "step": 25230 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021910256410256408, + "loss": 0.0563, + "step": 25240 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021907051282051278, + "loss": 0.0685, + "step": 25250 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021903846153846151, + "loss": 0.0561, + "step": 25260 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021900641025641022, + "loss": 0.0616, + "step": 25270 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021897435897435895, + "loss": 0.064, + "step": 25280 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021894230769230766, + "loss": 0.0612, + "step": 25290 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021891025641025637, + "loss": 0.0659, + "step": 25300 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002188782051282051, + "loss": 0.0682, + "step": 25310 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002188461538461538, + "loss": 0.058, + "step": 25320 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021881410256410254, + "loss": 0.0718, + "step": 25330 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021878205128205124, + "loss": 0.0665, + "step": 25340 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021874999999999998, + "loss": 0.0644, + "step": 25350 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021871794871794868, + "loss": 0.0667, + "step": 25360 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021868589743589742, + "loss": 0.0572, + "step": 25370 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021865384615384612, + "loss": 0.0509, + "step": 25380 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021862179487179483, + "loss": 0.0622, + "step": 25390 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021858974358974356, + "loss": 0.0541, + "step": 25400 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021855769230769227, + "loss": 0.061, + "step": 25410 + }, + { + "epoch": 0.54, + "learning_rate": 0.000218525641025641, + "loss": 0.0608, + "step": 25420 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002184935897435897, + "loss": 0.0613, + "step": 25430 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021846153846153844, + "loss": 0.054, + "step": 25440 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021842948717948715, + "loss": 0.0607, + "step": 25450 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021839743589743588, + "loss": 0.0615, + "step": 25460 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021836538461538459, + "loss": 0.0627, + "step": 25470 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002183333333333333, + "loss": 0.0594, + "step": 25480 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021830128205128202, + "loss": 0.0638, + "step": 25490 + }, + { + "epoch": 0.54, + "learning_rate": 0.00021826923076923073, + "loss": 0.0618, + "step": 25500 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021823717948717946, + "loss": 0.0585, + "step": 25510 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021820512820512817, + "loss": 0.0675, + "step": 25520 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002181730769230769, + "loss": 0.0589, + "step": 25530 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002181410256410256, + "loss": 0.0642, + "step": 25540 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021810897435897434, + "loss": 0.0637, + "step": 25550 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021807692307692305, + "loss": 0.0521, + "step": 25560 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021804487179487175, + "loss": 0.0551, + "step": 25570 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002180128205128205, + "loss": 0.0619, + "step": 25580 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002179807692307692, + "loss": 0.0592, + "step": 25590 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021794871794871793, + "loss": 0.0595, + "step": 25600 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021791666666666663, + "loss": 0.0605, + "step": 25610 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021788461538461537, + "loss": 0.0639, + "step": 25620 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021785256410256407, + "loss": 0.0548, + "step": 25630 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002178205128205128, + "loss": 0.0658, + "step": 25640 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002177884615384615, + "loss": 0.0584, + "step": 25650 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021775641025641022, + "loss": 0.0603, + "step": 25660 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021772435897435895, + "loss": 0.0502, + "step": 25670 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021769230769230766, + "loss": 0.0515, + "step": 25680 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002176602564102564, + "loss": 0.0556, + "step": 25690 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002176282051282051, + "loss": 0.0565, + "step": 25700 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021759615384615383, + "loss": 0.0712, + "step": 25710 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021756410256410253, + "loss": 0.0596, + "step": 25720 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021753205128205127, + "loss": 0.0669, + "step": 25730 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021749999999999997, + "loss": 0.0594, + "step": 25740 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021746794871794868, + "loss": 0.0576, + "step": 25750 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002174358974358974, + "loss": 0.0595, + "step": 25760 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021740384615384612, + "loss": 0.0593, + "step": 25770 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021737179487179485, + "loss": 0.0604, + "step": 25780 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021733974358974356, + "loss": 0.0737, + "step": 25790 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002173076923076923, + "loss": 0.0597, + "step": 25800 + }, + { + "epoch": 0.55, + "learning_rate": 0.000217275641025641, + "loss": 0.0654, + "step": 25810 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021724358974358973, + "loss": 0.0551, + "step": 25820 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021721153846153844, + "loss": 0.0574, + "step": 25830 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021717948717948714, + "loss": 0.057, + "step": 25840 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021714743589743588, + "loss": 0.0588, + "step": 25850 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021711538461538458, + "loss": 0.0565, + "step": 25860 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021708333333333332, + "loss": 0.0819, + "step": 25870 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021705128205128202, + "loss": 0.0642, + "step": 25880 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021701923076923075, + "loss": 0.0586, + "step": 25890 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021698717948717946, + "loss": 0.0587, + "step": 25900 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002169551282051282, + "loss": 0.0677, + "step": 25910 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002169230769230769, + "loss": 0.0631, + "step": 25920 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002168910256410256, + "loss": 0.0649, + "step": 25930 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021685897435897434, + "loss": 0.0631, + "step": 25940 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021682692307692305, + "loss": 0.0623, + "step": 25950 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021679487179487178, + "loss": 0.0592, + "step": 25960 + }, + { + "epoch": 0.55, + "learning_rate": 0.00021676282051282048, + "loss": 0.0616, + "step": 25970 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021673076923076922, + "loss": 0.0588, + "step": 25980 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021669871794871792, + "loss": 0.0554, + "step": 25990 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021666666666666666, + "loss": 0.0569, + "step": 26000 + }, + { + "epoch": 0.56, + "eval_cer": 0.9238702751731032, + "eval_loss": 0.03714323416352272, + "eval_runtime": 120.625, + "eval_samples_per_second": 16.58, + "eval_steps_per_second": 4.145, + "step": 26000 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021663461538461536, + "loss": 0.0623, + "step": 26010 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021660256410256407, + "loss": 0.0647, + "step": 26020 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002165705128205128, + "loss": 0.0731, + "step": 26030 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002165384615384615, + "loss": 0.063, + "step": 26040 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021650641025641024, + "loss": 0.0638, + "step": 26050 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021647435897435895, + "loss": 0.0541, + "step": 26060 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021644230769230768, + "loss": 0.0632, + "step": 26070 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021641025641025639, + "loss": 0.0551, + "step": 26080 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021637820512820512, + "loss": 0.0576, + "step": 26090 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021634615384615383, + "loss": 0.0609, + "step": 26100 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021631410256410253, + "loss": 0.0664, + "step": 26110 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021628205128205126, + "loss": 0.0634, + "step": 26120 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021624999999999997, + "loss": 0.0616, + "step": 26130 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002162179487179487, + "loss": 0.0519, + "step": 26140 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002161858974358974, + "loss": 0.0579, + "step": 26150 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021615384615384614, + "loss": 0.0569, + "step": 26160 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021612179487179485, + "loss": 0.061, + "step": 26170 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021608974358974358, + "loss": 0.0695, + "step": 26180 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002160576923076923, + "loss": 0.0624, + "step": 26190 + }, + { + "epoch": 0.56, + "learning_rate": 0.000216025641025641, + "loss": 0.0545, + "step": 26200 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021599358974358973, + "loss": 0.0658, + "step": 26210 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021596153846153843, + "loss": 0.0558, + "step": 26220 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021592948717948717, + "loss": 0.0563, + "step": 26230 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021589743589743587, + "loss": 0.0633, + "step": 26240 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002158653846153846, + "loss": 0.0657, + "step": 26250 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002158333333333333, + "loss": 0.0535, + "step": 26260 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021580128205128202, + "loss": 0.0521, + "step": 26270 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021576923076923075, + "loss": 0.0643, + "step": 26280 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021573717948717946, + "loss": 0.0586, + "step": 26290 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002157051282051282, + "loss": 0.0575, + "step": 26300 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002156730769230769, + "loss": 0.0691, + "step": 26310 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021564102564102563, + "loss": 0.0561, + "step": 26320 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021560897435897434, + "loss": 0.0616, + "step": 26330 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021557692307692307, + "loss": 0.0608, + "step": 26340 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021554487179487178, + "loss": 0.0591, + "step": 26350 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021551282051282048, + "loss": 0.0523, + "step": 26360 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021548076923076921, + "loss": 0.0653, + "step": 26370 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021544871794871792, + "loss": 0.0586, + "step": 26380 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021541666666666665, + "loss": 0.0711, + "step": 26390 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021538461538461536, + "loss": 0.0681, + "step": 26400 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002153525641025641, + "loss": 0.0636, + "step": 26410 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002153205128205128, + "loss": 0.0575, + "step": 26420 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021528846153846153, + "loss": 0.0646, + "step": 26430 + }, + { + "epoch": 0.56, + "learning_rate": 0.00021525641025641024, + "loss": 0.0526, + "step": 26440 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021522435897435894, + "loss": 0.0624, + "step": 26450 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021519230769230768, + "loss": 0.0557, + "step": 26460 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021516025641025638, + "loss": 0.0546, + "step": 26470 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021512820512820512, + "loss": 0.0509, + "step": 26480 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021509615384615382, + "loss": 0.0553, + "step": 26490 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021506410256410256, + "loss": 0.0602, + "step": 26500 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021503205128205126, + "loss": 0.0588, + "step": 26510 + }, + { + "epoch": 0.57, + "learning_rate": 0.000215, + "loss": 0.0639, + "step": 26520 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002149679487179487, + "loss": 0.0548, + "step": 26530 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002149358974358974, + "loss": 0.0574, + "step": 26540 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021490384615384614, + "loss": 0.0486, + "step": 26550 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021487179487179485, + "loss": 0.0526, + "step": 26560 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021483974358974358, + "loss": 0.0534, + "step": 26570 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021480769230769229, + "loss": 0.0666, + "step": 26580 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021477564102564102, + "loss": 0.0598, + "step": 26590 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021474358974358972, + "loss": 0.0563, + "step": 26600 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021471153846153846, + "loss": 0.0621, + "step": 26610 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021467948717948716, + "loss": 0.0558, + "step": 26620 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021464743589743587, + "loss": 0.0619, + "step": 26630 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002146153846153846, + "loss": 0.0621, + "step": 26640 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002145833333333333, + "loss": 0.0579, + "step": 26650 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021455128205128204, + "loss": 0.0577, + "step": 26660 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021451923076923075, + "loss": 0.0503, + "step": 26670 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021448717948717948, + "loss": 0.0578, + "step": 26680 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002144551282051282, + "loss": 0.0538, + "step": 26690 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021442307692307692, + "loss": 0.0595, + "step": 26700 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021439102564102563, + "loss": 0.0612, + "step": 26710 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021435897435897433, + "loss": 0.0664, + "step": 26720 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021432692307692307, + "loss": 0.0662, + "step": 26730 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021429487179487177, + "loss": 0.0628, + "step": 26740 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002142628205128205, + "loss": 0.0562, + "step": 26750 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002142307692307692, + "loss": 0.0619, + "step": 26760 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021419871794871794, + "loss": 0.0518, + "step": 26770 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021416666666666665, + "loss": 0.0596, + "step": 26780 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021413461538461538, + "loss": 0.0573, + "step": 26790 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002141025641025641, + "loss": 0.0597, + "step": 26800 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002140705128205128, + "loss": 0.0619, + "step": 26810 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021403846153846153, + "loss": 0.0662, + "step": 26820 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021400641025641023, + "loss": 0.0626, + "step": 26830 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021397435897435897, + "loss": 0.0627, + "step": 26840 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021394230769230767, + "loss": 0.0561, + "step": 26850 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002139102564102564, + "loss": 0.0709, + "step": 26860 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002138782051282051, + "loss": 0.0588, + "step": 26870 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021384615384615385, + "loss": 0.0566, + "step": 26880 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021381410256410255, + "loss": 0.0597, + "step": 26890 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021378205128205126, + "loss": 0.053, + "step": 26900 + }, + { + "epoch": 0.57, + "learning_rate": 0.00021375, + "loss": 0.0544, + "step": 26910 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002137179487179487, + "loss": 0.0631, + "step": 26920 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021368589743589743, + "loss": 0.0502, + "step": 26930 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021365384615384614, + "loss": 0.0514, + "step": 26940 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021362179487179487, + "loss": 0.0621, + "step": 26950 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021358974358974358, + "loss": 0.0614, + "step": 26960 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002135576923076923, + "loss": 0.0527, + "step": 26970 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021352564102564102, + "loss": 0.0534, + "step": 26980 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021349358974358972, + "loss": 0.0538, + "step": 26990 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021346153846153845, + "loss": 0.0632, + "step": 27000 + }, + { + "epoch": 0.58, + "eval_cer": 0.9238091274818956, + "eval_loss": 0.037807222455739975, + "eval_runtime": 120.7292, + "eval_samples_per_second": 16.566, + "eval_steps_per_second": 4.142, + "step": 27000 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021342948717948716, + "loss": 0.0562, + "step": 27010 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002133974358974359, + "loss": 0.0546, + "step": 27020 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002133653846153846, + "loss": 0.0573, + "step": 27030 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021333333333333333, + "loss": 0.0595, + "step": 27040 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021330128205128204, + "loss": 0.0612, + "step": 27050 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021326923076923077, + "loss": 0.0618, + "step": 27060 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021323717948717948, + "loss": 0.0497, + "step": 27070 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021320512820512818, + "loss": 0.0608, + "step": 27080 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021317307692307692, + "loss": 0.0566, + "step": 27090 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021314102564102562, + "loss": 0.0553, + "step": 27100 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021310897435897436, + "loss": 0.0549, + "step": 27110 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021307692307692306, + "loss": 0.0505, + "step": 27120 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002130448717948718, + "loss": 0.0606, + "step": 27130 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002130128205128205, + "loss": 0.0558, + "step": 27140 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021298076923076923, + "loss": 0.0539, + "step": 27150 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021294871794871794, + "loss": 0.0541, + "step": 27160 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021291666666666665, + "loss": 0.0608, + "step": 27170 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021288461538461538, + "loss": 0.0611, + "step": 27180 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021285256410256409, + "loss": 0.0541, + "step": 27190 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021282051282051282, + "loss": 0.0536, + "step": 27200 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021278846153846153, + "loss": 0.0495, + "step": 27210 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021275641025641026, + "loss": 0.065, + "step": 27220 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021272435897435896, + "loss": 0.0572, + "step": 27230 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021269230769230767, + "loss": 0.0626, + "step": 27240 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002126602564102564, + "loss": 0.0606, + "step": 27250 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002126282051282051, + "loss": 0.0572, + "step": 27260 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021259615384615384, + "loss": 0.0651, + "step": 27270 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021256410256410255, + "loss": 0.05, + "step": 27280 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021253205128205128, + "loss": 0.0645, + "step": 27290 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002125, + "loss": 0.0515, + "step": 27300 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021246794871794872, + "loss": 0.0661, + "step": 27310 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021243589743589743, + "loss": 0.0671, + "step": 27320 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021240384615384613, + "loss": 0.055, + "step": 27330 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021237179487179487, + "loss": 0.0608, + "step": 27340 + }, + { + "epoch": 0.58, + "learning_rate": 0.00021233974358974357, + "loss": 0.056, + "step": 27350 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002123076923076923, + "loss": 0.0559, + "step": 27360 + }, + { + "epoch": 0.58, + "learning_rate": 0.000212275641025641, + "loss": 0.063, + "step": 27370 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021224358974358974, + "loss": 0.0613, + "step": 27380 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021221153846153845, + "loss": 0.0544, + "step": 27390 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021217948717948718, + "loss": 0.0601, + "step": 27400 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002121474358974359, + "loss": 0.0601, + "step": 27410 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002121153846153846, + "loss": 0.0613, + "step": 27420 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021208333333333333, + "loss": 0.0615, + "step": 27430 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021205128205128204, + "loss": 0.0647, + "step": 27440 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021201923076923077, + "loss": 0.0565, + "step": 27450 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021198717948717947, + "loss": 0.068, + "step": 27460 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002119551282051282, + "loss": 0.071, + "step": 27470 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021192307692307691, + "loss": 0.0665, + "step": 27480 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021189102564102565, + "loss": 0.0605, + "step": 27490 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021185897435897435, + "loss": 0.0566, + "step": 27500 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021182692307692306, + "loss": 0.0578, + "step": 27510 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002117948717948718, + "loss": 0.0633, + "step": 27520 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002117628205128205, + "loss": 0.0637, + "step": 27530 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021173076923076923, + "loss": 0.06, + "step": 27540 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021169871794871794, + "loss": 0.0675, + "step": 27550 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021166666666666667, + "loss": 0.0605, + "step": 27560 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021163461538461538, + "loss": 0.0649, + "step": 27570 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002116025641025641, + "loss": 0.0652, + "step": 27580 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021157051282051282, + "loss": 0.0682, + "step": 27590 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021153846153846152, + "loss": 0.0547, + "step": 27600 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021150641025641026, + "loss": 0.0596, + "step": 27610 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021147435897435896, + "loss": 0.0535, + "step": 27620 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002114423076923077, + "loss": 0.0696, + "step": 27630 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002114102564102564, + "loss": 0.0638, + "step": 27640 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021137820512820513, + "loss": 0.0566, + "step": 27650 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021134615384615384, + "loss": 0.0561, + "step": 27660 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021131410256410257, + "loss": 0.057, + "step": 27670 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021128205128205128, + "loss": 0.0622, + "step": 27680 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021124999999999998, + "loss": 0.0559, + "step": 27690 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021121794871794872, + "loss": 0.0634, + "step": 27700 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021118589743589742, + "loss": 0.0598, + "step": 27710 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021115384615384616, + "loss": 0.0676, + "step": 27720 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021112179487179486, + "loss": 0.0636, + "step": 27730 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002110897435897436, + "loss": 0.0518, + "step": 27740 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002110576923076923, + "loss": 0.0589, + "step": 27750 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021102564102564104, + "loss": 0.0617, + "step": 27760 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021099358974358974, + "loss": 0.0569, + "step": 27770 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021096153846153845, + "loss": 0.0604, + "step": 27780 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021092948717948715, + "loss": 0.0595, + "step": 27790 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021089743589743586, + "loss": 0.056, + "step": 27800 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002108653846153846, + "loss": 0.057, + "step": 27810 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002108333333333333, + "loss": 0.0634, + "step": 27820 + }, + { + "epoch": 0.59, + "learning_rate": 0.000210801282051282, + "loss": 0.0649, + "step": 27830 + }, + { + "epoch": 0.59, + "learning_rate": 0.00021076923076923074, + "loss": 0.0633, + "step": 27840 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021073717948717944, + "loss": 0.0542, + "step": 27850 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021070512820512818, + "loss": 0.0579, + "step": 27860 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021067307692307688, + "loss": 0.0575, + "step": 27870 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021064102564102562, + "loss": 0.0576, + "step": 27880 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021060897435897432, + "loss": 0.0622, + "step": 27890 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021057692307692306, + "loss": 0.0566, + "step": 27900 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021054487179487176, + "loss": 0.059, + "step": 27910 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021051282051282047, + "loss": 0.0531, + "step": 27920 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002104807692307692, + "loss": 0.0529, + "step": 27930 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002104487179487179, + "loss": 0.0555, + "step": 27940 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021041666666666664, + "loss": 0.0557, + "step": 27950 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021038461538461535, + "loss": 0.0677, + "step": 27960 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021035256410256408, + "loss": 0.0643, + "step": 27970 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021032051282051279, + "loss": 0.0617, + "step": 27980 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021028846153846152, + "loss": 0.0632, + "step": 27990 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021025641025641022, + "loss": 0.0589, + "step": 28000 + }, + { + "epoch": 0.6, + "eval_cer": 0.9238006347470057, + "eval_loss": 0.03770715370774269, + "eval_runtime": 120.8596, + "eval_samples_per_second": 16.548, + "eval_steps_per_second": 4.137, + "step": 28000 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021022435897435893, + "loss": 0.06, + "step": 28010 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021019230769230766, + "loss": 0.0564, + "step": 28020 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021016025641025637, + "loss": 0.0598, + "step": 28030 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002101282051282051, + "loss": 0.061, + "step": 28040 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002100961538461538, + "loss": 0.0636, + "step": 28050 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021006410256410254, + "loss": 0.0556, + "step": 28060 + }, + { + "epoch": 0.6, + "learning_rate": 0.00021003205128205125, + "loss": 0.0561, + "step": 28070 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020999999999999998, + "loss": 0.0537, + "step": 28080 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002099679487179487, + "loss": 0.0556, + "step": 28090 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002099358974358974, + "loss": 0.0549, + "step": 28100 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020990384615384613, + "loss": 0.0547, + "step": 28110 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020987179487179483, + "loss": 0.0534, + "step": 28120 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020983974358974357, + "loss": 0.0581, + "step": 28130 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020980769230769227, + "loss": 0.0694, + "step": 28140 + }, + { + "epoch": 0.6, + "learning_rate": 0.000209775641025641, + "loss": 0.0531, + "step": 28150 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002097435897435897, + "loss": 0.0556, + "step": 28160 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020971153846153842, + "loss": 0.0568, + "step": 28170 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020967948717948715, + "loss": 0.0616, + "step": 28180 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020964743589743586, + "loss": 0.0601, + "step": 28190 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002096153846153846, + "loss": 0.0595, + "step": 28200 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002095833333333333, + "loss": 0.0605, + "step": 28210 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020955128205128203, + "loss": 0.0587, + "step": 28220 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020951923076923073, + "loss": 0.0654, + "step": 28230 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020948717948717947, + "loss": 0.0617, + "step": 28240 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020945512820512817, + "loss": 0.0666, + "step": 28250 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020942307692307688, + "loss": 0.0615, + "step": 28260 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002093910256410256, + "loss": 0.0625, + "step": 28270 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020935897435897432, + "loss": 0.0619, + "step": 28280 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020932692307692305, + "loss": 0.0636, + "step": 28290 + }, + { + "epoch": 0.6, + "learning_rate": 0.00020929487179487176, + "loss": 0.0552, + "step": 28300 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002092628205128205, + "loss": 0.0511, + "step": 28310 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002092307692307692, + "loss": 0.0618, + "step": 28320 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020919871794871793, + "loss": 0.0632, + "step": 28330 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020916666666666664, + "loss": 0.0518, + "step": 28340 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020913461538461534, + "loss": 0.049, + "step": 28350 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020910256410256408, + "loss": 0.0634, + "step": 28360 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020907051282051278, + "loss": 0.0566, + "step": 28370 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020903846153846152, + "loss": 0.0629, + "step": 28380 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020900641025641022, + "loss": 0.0586, + "step": 28390 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020897435897435895, + "loss": 0.0575, + "step": 28400 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020894230769230766, + "loss": 0.0604, + "step": 28410 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002089102564102564, + "loss": 0.0575, + "step": 28420 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002088782051282051, + "loss": 0.0602, + "step": 28430 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002088461538461538, + "loss": 0.0598, + "step": 28440 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020881410256410254, + "loss": 0.061, + "step": 28450 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020878205128205125, + "loss": 0.06, + "step": 28460 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020874999999999998, + "loss": 0.0516, + "step": 28470 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020871794871794868, + "loss": 0.0594, + "step": 28480 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020868589743589742, + "loss": 0.0506, + "step": 28490 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020865384615384612, + "loss": 0.0582, + "step": 28500 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020862179487179486, + "loss": 0.0639, + "step": 28510 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020858974358974356, + "loss": 0.0583, + "step": 28520 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020855769230769227, + "loss": 0.0699, + "step": 28530 + }, + { + "epoch": 0.61, + "learning_rate": 0.000208525641025641, + "loss": 0.0613, + "step": 28540 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002084935897435897, + "loss": 0.0598, + "step": 28550 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020846153846153844, + "loss": 0.0579, + "step": 28560 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020842948717948715, + "loss": 0.0587, + "step": 28570 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020839743589743588, + "loss": 0.0614, + "step": 28580 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020836538461538459, + "loss": 0.0525, + "step": 28590 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020833333333333332, + "loss": 0.0525, + "step": 28600 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020830128205128203, + "loss": 0.0529, + "step": 28610 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020826923076923073, + "loss": 0.0533, + "step": 28620 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020823717948717946, + "loss": 0.0769, + "step": 28630 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020820512820512817, + "loss": 0.0606, + "step": 28640 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002081730769230769, + "loss": 0.059, + "step": 28650 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002081410256410256, + "loss": 0.0616, + "step": 28660 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020810897435897434, + "loss": 0.0522, + "step": 28670 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020807692307692305, + "loss": 0.0508, + "step": 28680 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020804487179487178, + "loss": 0.058, + "step": 28690 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002080128205128205, + "loss": 0.0546, + "step": 28700 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002079807692307692, + "loss": 0.0504, + "step": 28710 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020794871794871793, + "loss": 0.0525, + "step": 28720 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020791666666666663, + "loss": 0.0597, + "step": 28730 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020788461538461537, + "loss": 0.0525, + "step": 28740 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020785256410256407, + "loss": 0.0609, + "step": 28750 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002078205128205128, + "loss": 0.0491, + "step": 28760 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002077884615384615, + "loss": 0.0557, + "step": 28770 + }, + { + "epoch": 0.61, + "learning_rate": 0.00020775641025641025, + "loss": 0.0525, + "step": 28780 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020772435897435895, + "loss": 0.0628, + "step": 28790 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020769230769230766, + "loss": 0.0511, + "step": 28800 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002076602564102564, + "loss": 0.0617, + "step": 28810 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002076282051282051, + "loss": 0.0548, + "step": 28820 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020759615384615383, + "loss": 0.0489, + "step": 28830 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020756410256410254, + "loss": 0.0502, + "step": 28840 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020753205128205127, + "loss": 0.0526, + "step": 28850 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020749999999999998, + "loss": 0.0576, + "step": 28860 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002074679487179487, + "loss": 0.0593, + "step": 28870 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020743589743589741, + "loss": 0.0577, + "step": 28880 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020740384615384612, + "loss": 0.0609, + "step": 28890 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020737179487179485, + "loss": 0.0648, + "step": 28900 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020733974358974356, + "loss": 0.0636, + "step": 28910 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002073076923076923, + "loss": 0.0539, + "step": 28920 + }, + { + "epoch": 0.62, + "learning_rate": 0.000207275641025641, + "loss": 0.0697, + "step": 28930 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020724358974358973, + "loss": 0.0593, + "step": 28940 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020721153846153844, + "loss": 0.0555, + "step": 28950 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020717948717948717, + "loss": 0.0583, + "step": 28960 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020714743589743588, + "loss": 0.0591, + "step": 28970 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020711538461538458, + "loss": 0.0566, + "step": 28980 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020708333333333332, + "loss": 0.0656, + "step": 28990 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020705128205128202, + "loss": 0.0511, + "step": 29000 + }, + { + "epoch": 0.62, + "eval_cer": 0.9237369392353312, + "eval_loss": 0.036585815250873566, + "eval_runtime": 120.6121, + "eval_samples_per_second": 16.582, + "eval_steps_per_second": 4.146, + "step": 29000 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020701923076923076, + "loss": 0.053, + "step": 29010 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020698717948717946, + "loss": 0.0561, + "step": 29020 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002069551282051282, + "loss": 0.0574, + "step": 29030 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002069230769230769, + "loss": 0.0538, + "step": 29040 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020689102564102563, + "loss": 0.0604, + "step": 29050 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020685897435897434, + "loss": 0.0581, + "step": 29060 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020682692307692305, + "loss": 0.0528, + "step": 29070 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020679487179487178, + "loss": 0.052, + "step": 29080 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020676282051282049, + "loss": 0.0539, + "step": 29090 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020673076923076922, + "loss": 0.0615, + "step": 29100 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020669871794871792, + "loss": 0.0554, + "step": 29110 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020666666666666666, + "loss": 0.0531, + "step": 29120 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020663461538461536, + "loss": 0.0611, + "step": 29130 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020660256410256407, + "loss": 0.0626, + "step": 29140 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002065705128205128, + "loss": 0.0736, + "step": 29150 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002065384615384615, + "loss": 0.0629, + "step": 29160 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020650641025641024, + "loss": 0.0548, + "step": 29170 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020647435897435895, + "loss": 0.0516, + "step": 29180 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020644230769230768, + "loss": 0.0541, + "step": 29190 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002064102564102564, + "loss": 0.0542, + "step": 29200 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020637820512820512, + "loss": 0.066, + "step": 29210 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020634615384615383, + "loss": 0.0676, + "step": 29220 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020631410256410253, + "loss": 0.0592, + "step": 29230 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020628205128205127, + "loss": 0.0743, + "step": 29240 + }, + { + "epoch": 0.62, + "learning_rate": 0.00020624999999999997, + "loss": 0.0644, + "step": 29250 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002062179487179487, + "loss": 0.0568, + "step": 29260 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002061858974358974, + "loss": 0.0543, + "step": 29270 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020615384615384614, + "loss": 0.0646, + "step": 29280 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020612179487179485, + "loss": 0.0637, + "step": 29290 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020608974358974358, + "loss": 0.0515, + "step": 29300 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002060576923076923, + "loss": 0.068, + "step": 29310 + }, + { + "epoch": 0.63, + "learning_rate": 0.000206025641025641, + "loss": 0.0618, + "step": 29320 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020599358974358973, + "loss": 0.0513, + "step": 29330 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020596153846153843, + "loss": 0.0594, + "step": 29340 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020592948717948717, + "loss": 0.0462, + "step": 29350 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020589743589743587, + "loss": 0.0556, + "step": 29360 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002058653846153846, + "loss": 0.0568, + "step": 29370 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002058333333333333, + "loss": 0.0654, + "step": 29380 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020580128205128205, + "loss": 0.0631, + "step": 29390 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020576923076923075, + "loss": 0.054, + "step": 29400 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020573717948717946, + "loss": 0.0495, + "step": 29410 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002057051282051282, + "loss": 0.0567, + "step": 29420 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002056730769230769, + "loss": 0.0558, + "step": 29430 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020564102564102563, + "loss": 0.0633, + "step": 29440 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020560897435897434, + "loss": 0.0629, + "step": 29450 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020557692307692307, + "loss": 0.0686, + "step": 29460 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020554487179487178, + "loss": 0.054, + "step": 29470 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002055128205128205, + "loss": 0.0506, + "step": 29480 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020548076923076922, + "loss": 0.0476, + "step": 29490 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020544871794871792, + "loss": 0.052, + "step": 29500 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020541666666666665, + "loss": 0.065, + "step": 29510 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020538461538461536, + "loss": 0.0556, + "step": 29520 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002053525641025641, + "loss": 0.0641, + "step": 29530 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002053205128205128, + "loss": 0.0677, + "step": 29540 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020528846153846153, + "loss": 0.0596, + "step": 29550 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020525641025641024, + "loss": 0.0619, + "step": 29560 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020522435897435897, + "loss": 0.0528, + "step": 29570 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020519230769230768, + "loss": 0.0606, + "step": 29580 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020516025641025638, + "loss": 0.0516, + "step": 29590 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020512820512820512, + "loss": 0.0579, + "step": 29600 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020509615384615382, + "loss": 0.0541, + "step": 29610 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020506410256410256, + "loss": 0.0502, + "step": 29620 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020503205128205126, + "loss": 0.0521, + "step": 29630 + }, + { + "epoch": 0.63, + "learning_rate": 0.000205, + "loss": 0.0575, + "step": 29640 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002049679487179487, + "loss": 0.0516, + "step": 29650 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020493589743589743, + "loss": 0.0563, + "step": 29660 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020490384615384614, + "loss": 0.0655, + "step": 29670 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020487179487179485, + "loss": 0.0603, + "step": 29680 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020483974358974358, + "loss": 0.0521, + "step": 29690 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020480769230769229, + "loss": 0.0627, + "step": 29700 + }, + { + "epoch": 0.63, + "learning_rate": 0.00020477564102564102, + "loss": 0.0517, + "step": 29710 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020474358974358973, + "loss": 0.0566, + "step": 29720 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020471153846153846, + "loss": 0.0618, + "step": 29730 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020467948717948716, + "loss": 0.0553, + "step": 29740 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002046474358974359, + "loss": 0.0607, + "step": 29750 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002046153846153846, + "loss": 0.0612, + "step": 29760 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002045833333333333, + "loss": 0.0531, + "step": 29770 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020455128205128204, + "loss": 0.0633, + "step": 29780 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020451923076923075, + "loss": 0.0552, + "step": 29790 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020448717948717948, + "loss": 0.0551, + "step": 29800 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002044551282051282, + "loss": 0.071, + "step": 29810 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020442307692307692, + "loss": 0.0572, + "step": 29820 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020439102564102563, + "loss": 0.0617, + "step": 29830 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020435897435897436, + "loss": 0.0694, + "step": 29840 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020432692307692307, + "loss": 0.0582, + "step": 29850 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020429487179487177, + "loss": 0.0542, + "step": 29860 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002042628205128205, + "loss": 0.0588, + "step": 29870 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002042307692307692, + "loss": 0.0514, + "step": 29880 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020419871794871794, + "loss": 0.0604, + "step": 29890 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020416666666666665, + "loss": 0.0481, + "step": 29900 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020413461538461538, + "loss": 0.0635, + "step": 29910 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002041025641025641, + "loss": 0.0631, + "step": 29920 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020407051282051282, + "loss": 0.0553, + "step": 29930 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020403846153846153, + "loss": 0.0544, + "step": 29940 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020400641025641024, + "loss": 0.0552, + "step": 29950 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020397435897435897, + "loss": 0.0614, + "step": 29960 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020394230769230767, + "loss": 0.0496, + "step": 29970 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002039102564102564, + "loss": 0.0538, + "step": 29980 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020387820512820511, + "loss": 0.0665, + "step": 29990 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020384615384615385, + "loss": 0.0651, + "step": 30000 + }, + { + "epoch": 0.64, + "eval_cer": 0.9239212315824428, + "eval_loss": 0.035759903490543365, + "eval_runtime": 120.8656, + "eval_samples_per_second": 16.547, + "eval_steps_per_second": 4.137, + "step": 30000 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020381410256410255, + "loss": 0.0555, + "step": 30010 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020378205128205129, + "loss": 0.059, + "step": 30020 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020375, + "loss": 0.0544, + "step": 30030 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002037179487179487, + "loss": 0.0597, + "step": 30040 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020368589743589743, + "loss": 0.0519, + "step": 30050 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020365384615384614, + "loss": 0.0546, + "step": 30060 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020362179487179487, + "loss": 0.0535, + "step": 30070 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020358974358974358, + "loss": 0.0609, + "step": 30080 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002035576923076923, + "loss": 0.0503, + "step": 30090 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020352564102564102, + "loss": 0.0558, + "step": 30100 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020349358974358972, + "loss": 0.0531, + "step": 30110 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020346153846153846, + "loss": 0.0454, + "step": 30120 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020342948717948716, + "loss": 0.0578, + "step": 30130 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002033974358974359, + "loss": 0.0638, + "step": 30140 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002033653846153846, + "loss": 0.0571, + "step": 30150 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020333333333333333, + "loss": 0.0518, + "step": 30160 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020330128205128204, + "loss": 0.0596, + "step": 30170 + }, + { + "epoch": 0.64, + "learning_rate": 0.00020326923076923077, + "loss": 0.0535, + "step": 30180 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020323717948717948, + "loss": 0.0499, + "step": 30190 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020320512820512818, + "loss": 0.0503, + "step": 30200 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020317307692307692, + "loss": 0.0453, + "step": 30210 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020314102564102562, + "loss": 0.0487, + "step": 30220 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020310897435897436, + "loss": 0.071, + "step": 30230 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020307692307692306, + "loss": 0.049, + "step": 30240 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002030448717948718, + "loss": 0.0577, + "step": 30250 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002030128205128205, + "loss": 0.0517, + "step": 30260 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020298076923076924, + "loss": 0.0589, + "step": 30270 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020294871794871794, + "loss": 0.0651, + "step": 30280 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020291666666666665, + "loss": 0.0527, + "step": 30290 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020288461538461538, + "loss": 0.0575, + "step": 30300 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002028525641025641, + "loss": 0.0556, + "step": 30310 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020282051282051282, + "loss": 0.0532, + "step": 30320 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020278846153846153, + "loss": 0.0598, + "step": 30330 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020275641025641026, + "loss": 0.0638, + "step": 30340 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020272435897435897, + "loss": 0.0588, + "step": 30350 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002026923076923077, + "loss": 0.057, + "step": 30360 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002026602564102564, + "loss": 0.0521, + "step": 30370 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002026282051282051, + "loss": 0.0494, + "step": 30380 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020259615384615384, + "loss": 0.0541, + "step": 30390 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020256410256410255, + "loss": 0.0613, + "step": 30400 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020253205128205128, + "loss": 0.0545, + "step": 30410 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002025, + "loss": 0.0627, + "step": 30420 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020246794871794872, + "loss": 0.0546, + "step": 30430 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020243589743589743, + "loss": 0.0624, + "step": 30440 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020240384615384616, + "loss": 0.0559, + "step": 30450 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020237179487179487, + "loss": 0.0561, + "step": 30460 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020233974358974357, + "loss": 0.0525, + "step": 30470 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002023076923076923, + "loss": 0.0491, + "step": 30480 + }, + { + "epoch": 0.65, + "learning_rate": 0.000202275641025641, + "loss": 0.0583, + "step": 30490 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020224358974358975, + "loss": 0.0541, + "step": 30500 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020221153846153845, + "loss": 0.0561, + "step": 30510 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020217948717948719, + "loss": 0.0591, + "step": 30520 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002021474358974359, + "loss": 0.052, + "step": 30530 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020211538461538462, + "loss": 0.0513, + "step": 30540 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020208333333333333, + "loss": 0.0514, + "step": 30550 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020205128205128204, + "loss": 0.0525, + "step": 30560 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020201923076923077, + "loss": 0.0519, + "step": 30570 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020198717948717948, + "loss": 0.0593, + "step": 30580 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002019551282051282, + "loss": 0.0629, + "step": 30590 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020192307692307691, + "loss": 0.0595, + "step": 30600 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020189102564102565, + "loss": 0.0506, + "step": 30610 + }, + { + "epoch": 0.65, + "learning_rate": 0.00020185897435897435, + "loss": 0.0524, + "step": 30620 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002018269230769231, + "loss": 0.0472, + "step": 30630 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002017948717948718, + "loss": 0.0484, + "step": 30640 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002017628205128205, + "loss": 0.0516, + "step": 30650 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020173076923076923, + "loss": 0.0554, + "step": 30660 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020169871794871794, + "loss": 0.064, + "step": 30670 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020166666666666667, + "loss": 0.0591, + "step": 30680 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020163461538461538, + "loss": 0.0597, + "step": 30690 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002016025641025641, + "loss": 0.0544, + "step": 30700 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020157051282051282, + "loss": 0.0636, + "step": 30710 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002015384615384615, + "loss": 0.0665, + "step": 30720 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020150641025641023, + "loss": 0.0652, + "step": 30730 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020147435897435893, + "loss": 0.0528, + "step": 30740 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020144230769230767, + "loss": 0.0603, + "step": 30750 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020141025641025637, + "loss": 0.0563, + "step": 30760 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002013782051282051, + "loss": 0.0546, + "step": 30770 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002013461538461538, + "loss": 0.0656, + "step": 30780 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020131410256410252, + "loss": 0.0689, + "step": 30790 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020128205128205125, + "loss": 0.0593, + "step": 30800 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020124999999999996, + "loss": 0.0569, + "step": 30810 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002012179487179487, + "loss": 0.056, + "step": 30820 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002011858974358974, + "loss": 0.0583, + "step": 30830 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020115384615384613, + "loss": 0.0555, + "step": 30840 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020112179487179484, + "loss": 0.0577, + "step": 30850 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020108974358974357, + "loss": 0.0487, + "step": 30860 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020105769230769228, + "loss": 0.0555, + "step": 30870 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020102564102564098, + "loss": 0.0512, + "step": 30880 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020099358974358972, + "loss": 0.0544, + "step": 30890 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020096153846153842, + "loss": 0.0527, + "step": 30900 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020092948717948715, + "loss": 0.0569, + "step": 30910 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020089743589743586, + "loss": 0.0489, + "step": 30920 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002008653846153846, + "loss": 0.049, + "step": 30930 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002008333333333333, + "loss": 0.0508, + "step": 30940 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020080128205128203, + "loss": 0.0609, + "step": 30950 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020076923076923074, + "loss": 0.0519, + "step": 30960 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020073717948717945, + "loss": 0.0636, + "step": 30970 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020070512820512818, + "loss": 0.0478, + "step": 30980 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020067307692307688, + "loss": 0.0566, + "step": 30990 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020064102564102562, + "loss": 0.0594, + "step": 31000 + }, + { + "epoch": 0.66, + "eval_cer": 0.9237955391060717, + "eval_loss": 0.035617340356111526, + "eval_runtime": 120.9283, + "eval_samples_per_second": 16.539, + "eval_steps_per_second": 4.135, + "step": 31000 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020060897435897432, + "loss": 0.0565, + "step": 31010 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020057692307692306, + "loss": 0.0506, + "step": 31020 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020054487179487176, + "loss": 0.0537, + "step": 31030 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020051282051282047, + "loss": 0.0558, + "step": 31040 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002004807692307692, + "loss": 0.0514, + "step": 31050 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002004487179487179, + "loss": 0.0627, + "step": 31060 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020041666666666664, + "loss": 0.0553, + "step": 31070 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020038461538461535, + "loss": 0.0582, + "step": 31080 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020035256410256408, + "loss": 0.0466, + "step": 31090 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002003205128205128, + "loss": 0.0496, + "step": 31100 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020028846153846152, + "loss": 0.0515, + "step": 31110 + }, + { + "epoch": 0.66, + "learning_rate": 0.00020025641025641023, + "loss": 0.0573, + "step": 31120 + }, + { + "epoch": 0.67, + "learning_rate": 0.00020022435897435893, + "loss": 0.0557, + "step": 31130 + }, + { + "epoch": 0.67, + "learning_rate": 0.00020019230769230766, + "loss": 0.0603, + "step": 31140 + }, + { + "epoch": 0.67, + "learning_rate": 0.00020016025641025637, + "loss": 0.0463, + "step": 31150 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002001282051282051, + "loss": 0.0568, + "step": 31160 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002000961538461538, + "loss": 0.0545, + "step": 31170 + }, + { + "epoch": 0.67, + "learning_rate": 0.00020006410256410254, + "loss": 0.0529, + "step": 31180 + }, + { + "epoch": 0.67, + "learning_rate": 0.00020003205128205125, + "loss": 0.0719, + "step": 31190 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019999999999999998, + "loss": 0.0537, + "step": 31200 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001999679487179487, + "loss": 0.056, + "step": 31210 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001999358974358974, + "loss": 0.0524, + "step": 31220 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019990384615384613, + "loss": 0.0546, + "step": 31230 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019987179487179483, + "loss": 0.0517, + "step": 31240 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019983974358974357, + "loss": 0.0518, + "step": 31250 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019980769230769227, + "loss": 0.0616, + "step": 31260 + }, + { + "epoch": 0.67, + "learning_rate": 0.000199775641025641, + "loss": 0.051, + "step": 31270 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001997435897435897, + "loss": 0.0488, + "step": 31280 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019971153846153845, + "loss": 0.0532, + "step": 31290 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019967948717948715, + "loss": 0.0528, + "step": 31300 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019964743589743586, + "loss": 0.0575, + "step": 31310 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001996153846153846, + "loss": 0.0618, + "step": 31320 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001995833333333333, + "loss": 0.0565, + "step": 31330 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019955128205128203, + "loss": 0.0691, + "step": 31340 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019951923076923074, + "loss": 0.0518, + "step": 31350 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019948717948717947, + "loss": 0.0584, + "step": 31360 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019945512820512818, + "loss": 0.0589, + "step": 31370 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001994230769230769, + "loss": 0.0543, + "step": 31380 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019939102564102561, + "loss": 0.0629, + "step": 31390 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019935897435897432, + "loss": 0.0532, + "step": 31400 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019932692307692305, + "loss": 0.0559, + "step": 31410 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019929487179487176, + "loss": 0.0535, + "step": 31420 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001992628205128205, + "loss": 0.0519, + "step": 31430 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001992307692307692, + "loss": 0.0698, + "step": 31440 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019919871794871793, + "loss": 0.0549, + "step": 31450 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019916666666666664, + "loss": 0.0629, + "step": 31460 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019913461538461537, + "loss": 0.0534, + "step": 31470 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019910256410256408, + "loss": 0.0531, + "step": 31480 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019907051282051278, + "loss": 0.0568, + "step": 31490 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019903846153846152, + "loss": 0.0483, + "step": 31500 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019900641025641022, + "loss": 0.0567, + "step": 31510 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019897435897435896, + "loss": 0.0558, + "step": 31520 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019894230769230766, + "loss": 0.051, + "step": 31530 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001989102564102564, + "loss": 0.0605, + "step": 31540 + }, + { + "epoch": 0.67, + "learning_rate": 0.0001988782051282051, + "loss": 0.0621, + "step": 31550 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019884615384615383, + "loss": 0.0543, + "step": 31560 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019881410256410254, + "loss": 0.0493, + "step": 31570 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019878205128205125, + "loss": 0.0602, + "step": 31580 + }, + { + "epoch": 0.67, + "learning_rate": 0.00019874999999999998, + "loss": 0.0539, + "step": 31590 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019871794871794869, + "loss": 0.0584, + "step": 31600 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019868589743589742, + "loss": 0.0561, + "step": 31610 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019865384615384612, + "loss": 0.0551, + "step": 31620 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019862179487179486, + "loss": 0.065, + "step": 31630 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019858974358974356, + "loss": 0.0624, + "step": 31640 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001985576923076923, + "loss": 0.0538, + "step": 31650 + }, + { + "epoch": 0.68, + "learning_rate": 0.000198525641025641, + "loss": 0.0476, + "step": 31660 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001984935897435897, + "loss": 0.0567, + "step": 31670 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019846153846153844, + "loss": 0.0566, + "step": 31680 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019842948717948715, + "loss": 0.0697, + "step": 31690 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019839743589743588, + "loss": 0.0589, + "step": 31700 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001983653846153846, + "loss": 0.0495, + "step": 31710 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019833333333333332, + "loss": 0.0557, + "step": 31720 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019830128205128203, + "loss": 0.0593, + "step": 31730 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019826923076923076, + "loss": 0.0501, + "step": 31740 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019823717948717947, + "loss": 0.0534, + "step": 31750 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019820512820512817, + "loss": 0.0518, + "step": 31760 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001981730769230769, + "loss": 0.0547, + "step": 31770 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001981410256410256, + "loss": 0.055, + "step": 31780 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019810897435897434, + "loss": 0.0513, + "step": 31790 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019807692307692305, + "loss": 0.0614, + "step": 31800 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019804487179487178, + "loss": 0.0508, + "step": 31810 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001980128205128205, + "loss": 0.0546, + "step": 31820 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019798076923076922, + "loss": 0.0548, + "step": 31830 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019794871794871793, + "loss": 0.0602, + "step": 31840 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019791666666666663, + "loss": 0.0586, + "step": 31850 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019788461538461537, + "loss": 0.0729, + "step": 31860 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019785256410256407, + "loss": 0.059, + "step": 31870 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001978205128205128, + "loss": 0.0542, + "step": 31880 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001977884615384615, + "loss": 0.0535, + "step": 31890 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019775641025641025, + "loss": 0.0591, + "step": 31900 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019772435897435895, + "loss": 0.052, + "step": 31910 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019769230769230769, + "loss": 0.0561, + "step": 31920 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001976602564102564, + "loss": 0.0523, + "step": 31930 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001976282051282051, + "loss": 0.0516, + "step": 31940 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019759615384615383, + "loss": 0.0612, + "step": 31950 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019756410256410254, + "loss": 0.0576, + "step": 31960 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019753205128205127, + "loss": 0.0603, + "step": 31970 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019749999999999998, + "loss": 0.0589, + "step": 31980 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001974679487179487, + "loss": 0.059, + "step": 31990 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019743589743589742, + "loss": 0.054, + "step": 32000 + }, + { + "epoch": 0.68, + "eval_cer": 0.9238099767553846, + "eval_loss": 0.03680818900465965, + "eval_runtime": 120.7161, + "eval_samples_per_second": 16.568, + "eval_steps_per_second": 4.142, + "step": 32000 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019740384615384612, + "loss": 0.0551, + "step": 32010 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019737179487179485, + "loss": 0.049, + "step": 32020 + }, + { + "epoch": 0.68, + "learning_rate": 0.00019733974358974356, + "loss": 0.059, + "step": 32030 + }, + { + "epoch": 0.68, + "learning_rate": 0.0001973076923076923, + "loss": 0.0548, + "step": 32040 + }, + { + "epoch": 0.68, + "learning_rate": 0.000197275641025641, + "loss": 0.0556, + "step": 32050 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019724358974358973, + "loss": 0.0498, + "step": 32060 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019721153846153844, + "loss": 0.0524, + "step": 32070 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019717948717948717, + "loss": 0.0512, + "step": 32080 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019714743589743588, + "loss": 0.0534, + "step": 32090 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019711538461538458, + "loss": 0.0576, + "step": 32100 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019708333333333332, + "loss": 0.0497, + "step": 32110 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019705128205128202, + "loss": 0.0514, + "step": 32120 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019701923076923076, + "loss": 0.0499, + "step": 32130 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019698717948717946, + "loss": 0.0534, + "step": 32140 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001969551282051282, + "loss": 0.0501, + "step": 32150 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001969230769230769, + "loss": 0.0578, + "step": 32160 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019689102564102563, + "loss": 0.0527, + "step": 32170 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019685897435897434, + "loss": 0.0479, + "step": 32180 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019682692307692305, + "loss": 0.0561, + "step": 32190 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019679487179487178, + "loss": 0.068, + "step": 32200 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019676282051282049, + "loss": 0.051, + "step": 32210 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019673076923076922, + "loss": 0.0583, + "step": 32220 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019669871794871793, + "loss": 0.0529, + "step": 32230 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019666666666666666, + "loss": 0.0536, + "step": 32240 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019663461538461536, + "loss": 0.0587, + "step": 32250 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001966025641025641, + "loss": 0.0601, + "step": 32260 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001965705128205128, + "loss": 0.0574, + "step": 32270 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001965384615384615, + "loss": 0.1439, + "step": 32280 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019650641025641024, + "loss": 0.0564, + "step": 32290 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019647435897435895, + "loss": 0.0489, + "step": 32300 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019644230769230768, + "loss": 0.0647, + "step": 32310 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001964102564102564, + "loss": 0.062, + "step": 32320 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019637820512820512, + "loss": 0.0606, + "step": 32330 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019634615384615383, + "loss": 0.0638, + "step": 32340 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019631410256410256, + "loss": 0.0549, + "step": 32350 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019628205128205127, + "loss": 0.0526, + "step": 32360 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019624999999999997, + "loss": 0.0608, + "step": 32370 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001962179487179487, + "loss": 0.0559, + "step": 32380 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001961858974358974, + "loss": 0.0543, + "step": 32390 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019615384615384615, + "loss": 0.0525, + "step": 32400 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019612179487179485, + "loss": 0.048, + "step": 32410 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019608974358974358, + "loss": 0.0469, + "step": 32420 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001960576923076923, + "loss": 0.0557, + "step": 32430 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019602564102564102, + "loss": 0.0555, + "step": 32440 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019599358974358973, + "loss": 0.0511, + "step": 32450 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019596153846153844, + "loss": 0.0538, + "step": 32460 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019592948717948717, + "loss": 0.056, + "step": 32470 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019589743589743587, + "loss": 0.0441, + "step": 32480 + }, + { + "epoch": 0.69, + "learning_rate": 0.0001958653846153846, + "loss": 0.052, + "step": 32490 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019583333333333331, + "loss": 0.0615, + "step": 32500 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019580128205128205, + "loss": 0.0577, + "step": 32510 + }, + { + "epoch": 0.69, + "learning_rate": 0.00019576923076923075, + "loss": 0.0487, + "step": 32520 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019573717948717949, + "loss": 0.0588, + "step": 32530 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001957051282051282, + "loss": 0.0482, + "step": 32540 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001956730769230769, + "loss": 0.0545, + "step": 32550 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019564102564102563, + "loss": 0.0546, + "step": 32560 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019560897435897434, + "loss": 0.0522, + "step": 32570 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019557692307692307, + "loss": 0.0589, + "step": 32580 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019554487179487178, + "loss": 0.0501, + "step": 32590 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001955128205128205, + "loss": 0.0522, + "step": 32600 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019548076923076922, + "loss": 0.0556, + "step": 32610 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019544871794871795, + "loss": 0.0591, + "step": 32620 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019541666666666666, + "loss": 0.0491, + "step": 32630 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019538461538461536, + "loss": 0.049, + "step": 32640 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001953525641025641, + "loss": 0.0511, + "step": 32650 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001953205128205128, + "loss": 0.0493, + "step": 32660 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019528846153846153, + "loss": 0.0499, + "step": 32670 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019525641025641024, + "loss": 0.0562, + "step": 32680 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019522435897435897, + "loss": 0.0623, + "step": 32690 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019519230769230768, + "loss": 0.0464, + "step": 32700 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001951602564102564, + "loss": 0.0565, + "step": 32710 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019512820512820512, + "loss": 0.049, + "step": 32720 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019509615384615382, + "loss": 0.0508, + "step": 32730 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019506410256410256, + "loss": 0.0473, + "step": 32740 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019503205128205126, + "loss": 0.0535, + "step": 32750 + }, + { + "epoch": 0.7, + "learning_rate": 0.000195, + "loss": 0.0535, + "step": 32760 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001949679487179487, + "loss": 0.0542, + "step": 32770 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019493589743589744, + "loss": 0.0548, + "step": 32780 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019490384615384614, + "loss": 0.0564, + "step": 32790 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019487179487179487, + "loss": 0.0553, + "step": 32800 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019483974358974358, + "loss": 0.0512, + "step": 32810 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001948076923076923, + "loss": 0.05, + "step": 32820 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019477564102564102, + "loss": 0.0594, + "step": 32830 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019474358974358973, + "loss": 0.0553, + "step": 32840 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019471153846153846, + "loss": 0.0476, + "step": 32850 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019467948717948717, + "loss": 0.0557, + "step": 32860 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001946474358974359, + "loss": 0.0555, + "step": 32870 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001946153846153846, + "loss": 0.0496, + "step": 32880 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019458333333333334, + "loss": 0.0591, + "step": 32890 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019455128205128204, + "loss": 0.0436, + "step": 32900 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019451923076923075, + "loss": 0.0618, + "step": 32910 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019448717948717948, + "loss": 0.0531, + "step": 32920 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001944551282051282, + "loss": 0.0477, + "step": 32930 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019442307692307692, + "loss": 0.0576, + "step": 32940 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019439102564102563, + "loss": 0.0547, + "step": 32950 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019435897435897436, + "loss": 0.0467, + "step": 32960 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019432692307692307, + "loss": 0.0582, + "step": 32970 + }, + { + "epoch": 0.7, + "learning_rate": 0.00019429487179487177, + "loss": 0.0543, + "step": 32980 + }, + { + "epoch": 0.7, + "learning_rate": 0.0001942628205128205, + "loss": 0.0594, + "step": 32990 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001942307692307692, + "loss": 0.0498, + "step": 33000 + }, + { + "epoch": 0.71, + "eval_cer": 0.923753075431622, + "eval_loss": 0.035282835364341736, + "eval_runtime": 120.7556, + "eval_samples_per_second": 16.562, + "eval_steps_per_second": 4.141, + "step": 33000 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019419871794871795, + "loss": 0.0542, + "step": 33010 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019416666666666665, + "loss": 0.0533, + "step": 33020 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019413461538461539, + "loss": 0.0508, + "step": 33030 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001941025641025641, + "loss": 0.0466, + "step": 33040 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019407051282051282, + "loss": 0.0549, + "step": 33050 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019403846153846153, + "loss": 0.0604, + "step": 33060 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019400641025641024, + "loss": 0.0526, + "step": 33070 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019397435897435897, + "loss": 0.0511, + "step": 33080 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019394230769230768, + "loss": 0.059, + "step": 33090 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001939102564102564, + "loss": 0.0538, + "step": 33100 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019387820512820511, + "loss": 0.0568, + "step": 33110 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019384615384615385, + "loss": 0.0555, + "step": 33120 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019381410256410255, + "loss": 0.0512, + "step": 33130 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001937820512820513, + "loss": 0.0573, + "step": 33140 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019375, + "loss": 0.0517, + "step": 33150 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001937179487179487, + "loss": 0.0518, + "step": 33160 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019368589743589743, + "loss": 0.0619, + "step": 33170 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019365384615384614, + "loss": 0.0576, + "step": 33180 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019362179487179487, + "loss": 0.058, + "step": 33190 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019358974358974358, + "loss": 0.066, + "step": 33200 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001935576923076923, + "loss": 0.053, + "step": 33210 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019352564102564102, + "loss": 0.0526, + "step": 33220 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019349358974358975, + "loss": 0.0605, + "step": 33230 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019346153846153846, + "loss": 0.0591, + "step": 33240 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019342948717948716, + "loss": 0.0521, + "step": 33250 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001933974358974359, + "loss": 0.0544, + "step": 33260 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001933653846153846, + "loss": 0.0535, + "step": 33270 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019333333333333333, + "loss": 0.0491, + "step": 33280 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019330128205128204, + "loss": 0.0553, + "step": 33290 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019326923076923077, + "loss": 0.056, + "step": 33300 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019323717948717948, + "loss": 0.0524, + "step": 33310 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001932051282051282, + "loss": 0.0546, + "step": 33320 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019317307692307692, + "loss": 0.0489, + "step": 33330 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019314102564102562, + "loss": 0.0556, + "step": 33340 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019310897435897436, + "loss": 0.055, + "step": 33350 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019307692307692306, + "loss": 0.0484, + "step": 33360 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001930448717948718, + "loss": 0.0536, + "step": 33370 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001930128205128205, + "loss": 0.0536, + "step": 33380 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019298076923076924, + "loss": 0.0556, + "step": 33390 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019294871794871794, + "loss": 0.0538, + "step": 33400 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019291666666666668, + "loss": 0.0506, + "step": 33410 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019288461538461538, + "loss": 0.0601, + "step": 33420 + }, + { + "epoch": 0.71, + "learning_rate": 0.0001928525641025641, + "loss": 0.0609, + "step": 33430 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019282051282051282, + "loss": 0.0572, + "step": 33440 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019278846153846153, + "loss": 0.0605, + "step": 33450 + }, + { + "epoch": 0.71, + "learning_rate": 0.00019275641025641026, + "loss": 0.0528, + "step": 33460 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019272435897435897, + "loss": 0.0566, + "step": 33470 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001926923076923077, + "loss": 0.0497, + "step": 33480 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001926602564102564, + "loss": 0.0535, + "step": 33490 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019262820512820514, + "loss": 0.0585, + "step": 33500 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019259615384615384, + "loss": 0.0533, + "step": 33510 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019256410256410255, + "loss": 0.0551, + "step": 33520 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019253205128205128, + "loss": 0.0516, + "step": 33530 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001925, + "loss": 0.0527, + "step": 33540 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019246794871794872, + "loss": 0.0504, + "step": 33550 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019243589743589743, + "loss": 0.0558, + "step": 33560 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019240384615384616, + "loss": 0.0554, + "step": 33570 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019237179487179487, + "loss": 0.0535, + "step": 33580 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001923397435897436, + "loss": 0.0528, + "step": 33590 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001923076923076923, + "loss": 0.0508, + "step": 33600 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019227564102564101, + "loss": 0.0535, + "step": 33610 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019224358974358975, + "loss": 0.0473, + "step": 33620 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019221153846153845, + "loss": 0.0513, + "step": 33630 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019217948717948716, + "loss": 0.0498, + "step": 33640 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019214743589743586, + "loss": 0.056, + "step": 33650 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019211538461538457, + "loss": 0.0427, + "step": 33660 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001920833333333333, + "loss": 0.0511, + "step": 33670 + }, + { + "epoch": 0.72, + "learning_rate": 0.000192051282051282, + "loss": 0.0463, + "step": 33680 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019201923076923074, + "loss": 0.0506, + "step": 33690 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019198717948717945, + "loss": 0.049, + "step": 33700 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019195512820512818, + "loss": 0.0538, + "step": 33710 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001919230769230769, + "loss": 0.0527, + "step": 33720 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019189102564102562, + "loss": 0.0547, + "step": 33730 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019185897435897433, + "loss": 0.0431, + "step": 33740 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019182692307692303, + "loss": 0.0504, + "step": 33750 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019179487179487177, + "loss": 0.0575, + "step": 33760 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019176282051282047, + "loss": 0.0568, + "step": 33770 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001917307692307692, + "loss": 0.0529, + "step": 33780 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001916987179487179, + "loss": 0.0521, + "step": 33790 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019166666666666665, + "loss": 0.0569, + "step": 33800 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019163461538461535, + "loss": 0.0602, + "step": 33810 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019160256410256408, + "loss": 0.055, + "step": 33820 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001915705128205128, + "loss": 0.0485, + "step": 33830 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001915384615384615, + "loss": 0.0547, + "step": 33840 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019150641025641023, + "loss": 0.0524, + "step": 33850 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019147435897435894, + "loss": 0.0572, + "step": 33860 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019144230769230767, + "loss": 0.0471, + "step": 33870 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019141025641025638, + "loss": 0.0486, + "step": 33880 + }, + { + "epoch": 0.72, + "learning_rate": 0.0001913782051282051, + "loss": 0.0516, + "step": 33890 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019134615384615381, + "loss": 0.0456, + "step": 33900 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019131410256410252, + "loss": 0.0532, + "step": 33910 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019128205128205125, + "loss": 0.0574, + "step": 33920 + }, + { + "epoch": 0.72, + "learning_rate": 0.00019124999999999996, + "loss": 0.0491, + "step": 33930 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001912179487179487, + "loss": 0.0607, + "step": 33940 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001911858974358974, + "loss": 0.0503, + "step": 33950 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019115384615384613, + "loss": 0.0551, + "step": 33960 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019112179487179484, + "loss": 0.0502, + "step": 33970 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019108974358974357, + "loss": 0.0531, + "step": 33980 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019105769230769228, + "loss": 0.055, + "step": 33990 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019102564102564098, + "loss": 0.0559, + "step": 34000 + }, + { + "epoch": 0.73, + "eval_cer": 0.9237709101748909, + "eval_loss": 0.033693794161081314, + "eval_runtime": 120.6634, + "eval_samples_per_second": 16.575, + "eval_steps_per_second": 4.144, + "step": 34000 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019099358974358972, + "loss": 0.0524, + "step": 34010 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019096153846153842, + "loss": 0.0583, + "step": 34020 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019092948717948716, + "loss": 0.05, + "step": 34030 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019089743589743586, + "loss": 0.0699, + "step": 34040 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001908653846153846, + "loss": 0.0563, + "step": 34050 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001908333333333333, + "loss": 0.0515, + "step": 34060 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019080128205128203, + "loss": 0.0699, + "step": 34070 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019076923076923074, + "loss": 0.0642, + "step": 34080 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019073717948717945, + "loss": 0.0547, + "step": 34090 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019070512820512818, + "loss": 0.0541, + "step": 34100 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019067307692307689, + "loss": 0.0551, + "step": 34110 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019064102564102562, + "loss": 0.0526, + "step": 34120 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019060897435897432, + "loss": 0.0468, + "step": 34130 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019057692307692306, + "loss": 0.0469, + "step": 34140 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019054487179487176, + "loss": 0.0462, + "step": 34150 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001905128205128205, + "loss": 0.0543, + "step": 34160 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001904807692307692, + "loss": 0.0544, + "step": 34170 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001904487179487179, + "loss": 0.0491, + "step": 34180 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019041666666666664, + "loss": 0.0522, + "step": 34190 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019038461538461535, + "loss": 0.0447, + "step": 34200 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019035256410256408, + "loss": 0.0577, + "step": 34210 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001903205128205128, + "loss": 0.0517, + "step": 34220 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019028846153846152, + "loss": 0.0442, + "step": 34230 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019025641025641023, + "loss": 0.0564, + "step": 34240 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019022435897435896, + "loss": 0.0518, + "step": 34250 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019019230769230767, + "loss": 0.0461, + "step": 34260 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019016025641025637, + "loss": 0.0532, + "step": 34270 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001901282051282051, + "loss": 0.0583, + "step": 34280 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001900961538461538, + "loss": 0.0603, + "step": 34290 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019006410256410254, + "loss": 0.0554, + "step": 34300 + }, + { + "epoch": 0.73, + "learning_rate": 0.00019003205128205125, + "loss": 0.0442, + "step": 34310 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018999999999999998, + "loss": 0.0507, + "step": 34320 + }, + { + "epoch": 0.73, + "learning_rate": 0.0001899679487179487, + "loss": 0.0489, + "step": 34330 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018993589743589742, + "loss": 0.0437, + "step": 34340 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018990384615384613, + "loss": 0.0454, + "step": 34350 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018987179487179483, + "loss": 0.0608, + "step": 34360 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018983974358974357, + "loss": 0.0563, + "step": 34370 + }, + { + "epoch": 0.73, + "learning_rate": 0.00018980769230769227, + "loss": 0.0512, + "step": 34380 + }, + { + "epoch": 0.73, + "learning_rate": 0.000189775641025641, + "loss": 0.0597, + "step": 34390 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001897435897435897, + "loss": 0.0571, + "step": 34400 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018971153846153845, + "loss": 0.0618, + "step": 34410 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018967948717948715, + "loss": 0.0476, + "step": 34420 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018964743589743589, + "loss": 0.0458, + "step": 34430 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001896153846153846, + "loss": 0.0484, + "step": 34440 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001895833333333333, + "loss": 0.0658, + "step": 34450 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018955128205128203, + "loss": 0.0511, + "step": 34460 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018951923076923074, + "loss": 0.0477, + "step": 34470 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018948717948717947, + "loss": 0.0477, + "step": 34480 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018945512820512818, + "loss": 0.0504, + "step": 34490 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001894230769230769, + "loss": 0.0453, + "step": 34500 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018939102564102562, + "loss": 0.0502, + "step": 34510 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018935897435897435, + "loss": 0.052, + "step": 34520 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018932692307692305, + "loss": 0.0513, + "step": 34530 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018929487179487176, + "loss": 0.0529, + "step": 34540 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001892628205128205, + "loss": 0.0635, + "step": 34550 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001892307692307692, + "loss": 0.0548, + "step": 34560 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018919871794871793, + "loss": 0.0572, + "step": 34570 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018916666666666664, + "loss": 0.0552, + "step": 34580 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018913461538461537, + "loss": 0.0507, + "step": 34590 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018910256410256408, + "loss": 0.0457, + "step": 34600 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001890705128205128, + "loss": 0.0467, + "step": 34610 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018903846153846152, + "loss": 0.048, + "step": 34620 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018900641025641022, + "loss": 0.057, + "step": 34630 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018897435897435896, + "loss": 0.0409, + "step": 34640 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018894230769230766, + "loss": 0.0526, + "step": 34650 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001889102564102564, + "loss": 0.0584, + "step": 34660 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001888782051282051, + "loss": 0.0477, + "step": 34670 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018884615384615383, + "loss": 0.05, + "step": 34680 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018881410256410254, + "loss": 0.0504, + "step": 34690 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018878205128205127, + "loss": 0.059, + "step": 34700 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018874999999999998, + "loss": 0.0563, + "step": 34710 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018871794871794869, + "loss": 0.0469, + "step": 34720 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018868589743589742, + "loss": 0.0502, + "step": 34730 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018865384615384613, + "loss": 0.0485, + "step": 34740 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018862179487179486, + "loss": 0.0531, + "step": 34750 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018858974358974356, + "loss": 0.0496, + "step": 34760 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001885576923076923, + "loss": 0.0501, + "step": 34770 + }, + { + "epoch": 0.74, + "learning_rate": 0.000188525641025641, + "loss": 0.0549, + "step": 34780 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018849358974358974, + "loss": 0.0506, + "step": 34790 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018846153846153844, + "loss": 0.0569, + "step": 34800 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018842948717948715, + "loss": 0.0564, + "step": 34810 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018839743589743588, + "loss": 0.0493, + "step": 34820 + }, + { + "epoch": 0.74, + "learning_rate": 0.0001883653846153846, + "loss": 0.0484, + "step": 34830 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018833333333333332, + "loss": 0.049, + "step": 34840 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018830128205128203, + "loss": 0.0503, + "step": 34850 + }, + { + "epoch": 0.74, + "learning_rate": 0.00018826923076923076, + "loss": 0.0495, + "step": 34860 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018823717948717947, + "loss": 0.0434, + "step": 34870 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018820512820512817, + "loss": 0.0482, + "step": 34880 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001881730769230769, + "loss": 0.0441, + "step": 34890 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001881410256410256, + "loss": 0.0538, + "step": 34900 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018810897435897435, + "loss": 0.0565, + "step": 34910 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018807692307692305, + "loss": 0.0572, + "step": 34920 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018804487179487178, + "loss": 0.0654, + "step": 34930 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001880128205128205, + "loss": 0.054, + "step": 34940 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018798076923076922, + "loss": 0.0506, + "step": 34950 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018794871794871793, + "loss": 0.0566, + "step": 34960 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018791666666666664, + "loss": 0.0499, + "step": 34970 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018788461538461537, + "loss": 0.0615, + "step": 34980 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018785256410256407, + "loss": 0.0532, + "step": 34990 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001878205128205128, + "loss": 0.0502, + "step": 35000 + }, + { + "epoch": 0.75, + "eval_cer": 0.9237649652604679, + "eval_loss": 0.034136708825826645, + "eval_runtime": 120.4185, + "eval_samples_per_second": 16.609, + "eval_steps_per_second": 4.152, + "step": 35000 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018778846153846151, + "loss": 0.053, + "step": 35010 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018775641025641025, + "loss": 0.0555, + "step": 35020 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018772435897435895, + "loss": 0.0524, + "step": 35030 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018769230769230769, + "loss": 0.0553, + "step": 35040 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001876602564102564, + "loss": 0.0631, + "step": 35050 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001876282051282051, + "loss": 0.0546, + "step": 35060 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018759615384615383, + "loss": 0.0541, + "step": 35070 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018756410256410254, + "loss": 0.0495, + "step": 35080 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018753205128205127, + "loss": 0.0557, + "step": 35090 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018749999999999998, + "loss": 0.0484, + "step": 35100 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001874679487179487, + "loss": 0.0485, + "step": 35110 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018743589743589742, + "loss": 0.0571, + "step": 35120 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018740384615384615, + "loss": 0.0442, + "step": 35130 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018737179487179486, + "loss": 0.0514, + "step": 35140 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018733974358974356, + "loss": 0.0461, + "step": 35150 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001873076923076923, + "loss": 0.0474, + "step": 35160 + }, + { + "epoch": 0.75, + "learning_rate": 0.000187275641025641, + "loss": 0.0514, + "step": 35170 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018724358974358973, + "loss": 0.054, + "step": 35180 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018721153846153844, + "loss": 0.0536, + "step": 35190 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018717948717948717, + "loss": 0.0559, + "step": 35200 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018714743589743588, + "loss": 0.0523, + "step": 35210 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001871153846153846, + "loss": 0.046, + "step": 35220 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018708333333333332, + "loss": 0.0463, + "step": 35230 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018705128205128202, + "loss": 0.0574, + "step": 35240 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018701923076923076, + "loss": 0.0566, + "step": 35250 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018698717948717946, + "loss": 0.0491, + "step": 35260 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001869551282051282, + "loss": 0.0496, + "step": 35270 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001869230769230769, + "loss": 0.0711, + "step": 35280 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018689102564102564, + "loss": 0.0597, + "step": 35290 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018685897435897434, + "loss": 0.0533, + "step": 35300 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018682692307692307, + "loss": 0.0536, + "step": 35310 + }, + { + "epoch": 0.75, + "learning_rate": 0.00018679487179487178, + "loss": 0.0475, + "step": 35320 + }, + { + "epoch": 0.75, + "learning_rate": 0.0001867628205128205, + "loss": 0.0566, + "step": 35330 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018673076923076922, + "loss": 0.0511, + "step": 35340 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018669871794871793, + "loss": 0.05, + "step": 35350 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018666666666666666, + "loss": 0.0541, + "step": 35360 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018663461538461537, + "loss": 0.0474, + "step": 35370 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001866025641025641, + "loss": 0.0468, + "step": 35380 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001865705128205128, + "loss": 0.055, + "step": 35390 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018653846153846154, + "loss": 0.0489, + "step": 35400 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018650641025641024, + "loss": 0.0516, + "step": 35410 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018647435897435895, + "loss": 0.0521, + "step": 35420 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018644230769230768, + "loss": 0.0544, + "step": 35430 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001864102564102564, + "loss": 0.0547, + "step": 35440 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018637820512820512, + "loss": 0.0515, + "step": 35450 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018634615384615383, + "loss": 0.0491, + "step": 35460 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018631410256410256, + "loss": 0.0547, + "step": 35470 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018628205128205127, + "loss": 0.0538, + "step": 35480 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018625, + "loss": 0.0423, + "step": 35490 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001862179487179487, + "loss": 0.0482, + "step": 35500 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001861858974358974, + "loss": 0.0542, + "step": 35510 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018615384615384615, + "loss": 0.0445, + "step": 35520 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018612179487179485, + "loss": 0.0465, + "step": 35530 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018608974358974359, + "loss": 0.0517, + "step": 35540 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001860576923076923, + "loss": 0.0534, + "step": 35550 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018602564102564102, + "loss": 0.0471, + "step": 35560 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018599358974358973, + "loss": 0.0521, + "step": 35570 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018596153846153846, + "loss": 0.0554, + "step": 35580 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018592948717948717, + "loss": 0.043, + "step": 35590 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018589743589743588, + "loss": 0.0475, + "step": 35600 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001858653846153846, + "loss": 0.0555, + "step": 35610 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018583333333333331, + "loss": 0.055, + "step": 35620 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018580128205128205, + "loss": 0.0508, + "step": 35630 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018576923076923075, + "loss": 0.0442, + "step": 35640 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001857371794871795, + "loss": 0.0534, + "step": 35650 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001857051282051282, + "loss": 0.0524, + "step": 35660 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018567307692307693, + "loss": 0.0439, + "step": 35670 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018564102564102563, + "loss": 0.055, + "step": 35680 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018560897435897434, + "loss": 0.0577, + "step": 35690 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018557692307692307, + "loss": 0.0526, + "step": 35700 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018554487179487178, + "loss": 0.05, + "step": 35710 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001855128205128205, + "loss": 0.0518, + "step": 35720 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018548076923076922, + "loss": 0.064, + "step": 35730 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018544871794871795, + "loss": 0.0517, + "step": 35740 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018541666666666666, + "loss": 0.047, + "step": 35750 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001853846153846154, + "loss": 0.0447, + "step": 35760 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001853525641025641, + "loss": 0.0469, + "step": 35770 + }, + { + "epoch": 0.76, + "learning_rate": 0.0001853205128205128, + "loss": 0.0638, + "step": 35780 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018528846153846153, + "loss": 0.0488, + "step": 35790 + }, + { + "epoch": 0.76, + "learning_rate": 0.00018525641025641024, + "loss": 0.0638, + "step": 35800 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018522435897435897, + "loss": 0.0514, + "step": 35810 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018519230769230768, + "loss": 0.0453, + "step": 35820 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001851602564102564, + "loss": 0.0514, + "step": 35830 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018512820512820512, + "loss": 0.049, + "step": 35840 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018509615384615382, + "loss": 0.0499, + "step": 35850 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018506410256410256, + "loss": 0.0552, + "step": 35860 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018503205128205126, + "loss": 0.0635, + "step": 35870 + }, + { + "epoch": 0.77, + "learning_rate": 0.000185, + "loss": 0.0484, + "step": 35880 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001849679487179487, + "loss": 0.0455, + "step": 35890 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018493589743589744, + "loss": 0.0572, + "step": 35900 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018490384615384614, + "loss": 0.0532, + "step": 35910 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018487179487179488, + "loss": 0.0467, + "step": 35920 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018483974358974358, + "loss": 0.0474, + "step": 35930 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001848076923076923, + "loss": 0.0484, + "step": 35940 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018477564102564102, + "loss": 0.0515, + "step": 35950 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018474358974358973, + "loss": 0.0509, + "step": 35960 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018471153846153846, + "loss": 0.0504, + "step": 35970 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018467948717948717, + "loss": 0.0531, + "step": 35980 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001846474358974359, + "loss": 0.0503, + "step": 35990 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001846153846153846, + "loss": 0.0588, + "step": 36000 + }, + { + "epoch": 0.77, + "eval_cer": 0.9238617824382133, + "eval_loss": 0.0338614396750927, + "eval_runtime": 120.9412, + "eval_samples_per_second": 16.537, + "eval_steps_per_second": 4.134, + "step": 36000 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018458333333333334, + "loss": 0.052, + "step": 36010 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018455128205128204, + "loss": 0.0587, + "step": 36020 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018451923076923075, + "loss": 0.0541, + "step": 36030 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018448717948717948, + "loss": 0.0507, + "step": 36040 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001844551282051282, + "loss": 0.0482, + "step": 36050 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018442307692307692, + "loss": 0.0532, + "step": 36060 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018439102564102563, + "loss": 0.0585, + "step": 36070 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018435897435897436, + "loss": 0.05, + "step": 36080 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018432692307692307, + "loss": 0.0495, + "step": 36090 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001842948717948718, + "loss": 0.0625, + "step": 36100 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001842628205128205, + "loss": 0.051, + "step": 36110 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018423076923076921, + "loss": 0.0593, + "step": 36120 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018419871794871795, + "loss": 0.05, + "step": 36130 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018416666666666665, + "loss": 0.0498, + "step": 36140 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018413461538461539, + "loss": 0.0538, + "step": 36150 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001841025641025641, + "loss": 0.0429, + "step": 36160 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018407051282051283, + "loss": 0.0535, + "step": 36170 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018403846153846153, + "loss": 0.0525, + "step": 36180 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018400641025641026, + "loss": 0.0475, + "step": 36190 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018397435897435897, + "loss": 0.0461, + "step": 36200 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018394230769230768, + "loss": 0.0478, + "step": 36210 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001839102564102564, + "loss": 0.0645, + "step": 36220 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018387820512820512, + "loss": 0.0506, + "step": 36230 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018384615384615385, + "loss": 0.0542, + "step": 36240 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018381410256410255, + "loss": 0.0509, + "step": 36250 + }, + { + "epoch": 0.77, + "learning_rate": 0.0001837820512820513, + "loss": 0.0533, + "step": 36260 + }, + { + "epoch": 0.77, + "learning_rate": 0.00018375, + "loss": 0.0536, + "step": 36270 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018371794871794873, + "loss": 0.0436, + "step": 36280 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018368589743589743, + "loss": 0.047, + "step": 36290 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018365384615384614, + "loss": 0.0495, + "step": 36300 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018362179487179487, + "loss": 0.0507, + "step": 36310 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018358974358974358, + "loss": 0.0544, + "step": 36320 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001835576923076923, + "loss": 0.0609, + "step": 36330 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018352564102564102, + "loss": 0.0486, + "step": 36340 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018349358974358975, + "loss": 0.0501, + "step": 36350 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018346153846153846, + "loss": 0.0474, + "step": 36360 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001834294871794872, + "loss": 0.0478, + "step": 36370 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001833974358974359, + "loss": 0.0505, + "step": 36380 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001833653846153846, + "loss": 0.0468, + "step": 36390 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018333333333333334, + "loss": 0.0555, + "step": 36400 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018330128205128204, + "loss": 0.0413, + "step": 36410 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018326923076923077, + "loss": 0.0512, + "step": 36420 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018323717948717948, + "loss": 0.0483, + "step": 36430 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018320512820512821, + "loss": 0.0527, + "step": 36440 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018317307692307692, + "loss": 0.0523, + "step": 36450 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018314102564102565, + "loss": 0.0442, + "step": 36460 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018310897435897436, + "loss": 0.0474, + "step": 36470 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018307692307692307, + "loss": 0.0481, + "step": 36480 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001830448717948718, + "loss": 0.0449, + "step": 36490 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001830128205128205, + "loss": 0.0434, + "step": 36500 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018298076923076924, + "loss": 0.0493, + "step": 36510 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018294871794871794, + "loss": 0.0507, + "step": 36520 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018291666666666668, + "loss": 0.0527, + "step": 36530 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018288461538461538, + "loss": 0.0462, + "step": 36540 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018285256410256412, + "loss": 0.0429, + "step": 36550 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018282051282051282, + "loss": 0.0411, + "step": 36560 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001827884615384615, + "loss": 0.0477, + "step": 36570 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018275641025641023, + "loss": 0.0489, + "step": 36580 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018272435897435894, + "loss": 0.0538, + "step": 36590 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018269230769230767, + "loss": 0.0549, + "step": 36600 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018266025641025638, + "loss": 0.0558, + "step": 36610 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018262820512820509, + "loss": 0.0502, + "step": 36620 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018259615384615382, + "loss": 0.0539, + "step": 36630 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018256410256410252, + "loss": 0.0521, + "step": 36640 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018253205128205126, + "loss": 0.0552, + "step": 36650 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018249999999999996, + "loss": 0.0573, + "step": 36660 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001824679487179487, + "loss": 0.0646, + "step": 36670 + }, + { + "epoch": 0.78, + "learning_rate": 0.0001824358974358974, + "loss": 0.0507, + "step": 36680 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018240384615384614, + "loss": 0.0505, + "step": 36690 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018237179487179484, + "loss": 0.0494, + "step": 36700 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018233974358974355, + "loss": 0.0507, + "step": 36710 + }, + { + "epoch": 0.78, + "learning_rate": 0.00018230769230769228, + "loss": 0.0488, + "step": 36720 + }, + { + "epoch": 0.78, + "learning_rate": 0.000182275641025641, + "loss": 0.057, + "step": 36730 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018224358974358972, + "loss": 0.0612, + "step": 36740 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018221153846153843, + "loss": 0.0507, + "step": 36750 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018217948717948716, + "loss": 0.0559, + "step": 36760 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018214743589743587, + "loss": 0.0563, + "step": 36770 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018211538461538457, + "loss": 0.0576, + "step": 36780 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001820833333333333, + "loss": 0.047, + "step": 36790 + }, + { + "epoch": 0.79, + "learning_rate": 0.000182051282051282, + "loss": 0.0564, + "step": 36800 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018201923076923074, + "loss": 0.0521, + "step": 36810 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018198717948717945, + "loss": 0.0578, + "step": 36820 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018195512820512818, + "loss": 0.051, + "step": 36830 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001819230769230769, + "loss": 0.0481, + "step": 36840 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018189102564102562, + "loss": 0.0521, + "step": 36850 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018185897435897433, + "loss": 0.0506, + "step": 36860 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018182692307692303, + "loss": 0.0479, + "step": 36870 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018179487179487177, + "loss": 0.0527, + "step": 36880 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018176282051282047, + "loss": 0.0519, + "step": 36890 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001817307692307692, + "loss": 0.0572, + "step": 36900 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001816987179487179, + "loss": 0.0453, + "step": 36910 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018166666666666665, + "loss": 0.0539, + "step": 36920 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018163461538461535, + "loss": 0.0578, + "step": 36930 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018160256410256409, + "loss": 0.0495, + "step": 36940 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001815705128205128, + "loss": 0.0575, + "step": 36950 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001815384615384615, + "loss": 0.0463, + "step": 36960 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018150641025641023, + "loss": 0.0459, + "step": 36970 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018147435897435894, + "loss": 0.0514, + "step": 36980 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018144230769230767, + "loss": 0.0474, + "step": 36990 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018141025641025638, + "loss": 0.0487, + "step": 37000 + }, + { + "epoch": 0.79, + "eval_cer": 0.9237021190222824, + "eval_loss": 0.033780451864004135, + "eval_runtime": 120.6259, + "eval_samples_per_second": 16.58, + "eval_steps_per_second": 4.145, + "step": 37000 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001813782051282051, + "loss": 0.0505, + "step": 37010 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018134615384615382, + "loss": 0.0515, + "step": 37020 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018131410256410255, + "loss": 0.0475, + "step": 37030 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018128205128205125, + "loss": 0.046, + "step": 37040 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018124999999999996, + "loss": 0.0499, + "step": 37050 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001812179487179487, + "loss": 0.0553, + "step": 37060 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001811858974358974, + "loss": 0.045, + "step": 37070 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018115384615384613, + "loss": 0.0544, + "step": 37080 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018112179487179484, + "loss": 0.0517, + "step": 37090 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018108974358974357, + "loss": 0.0469, + "step": 37100 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018105769230769228, + "loss": 0.0511, + "step": 37110 + }, + { + "epoch": 0.79, + "learning_rate": 0.000181025641025641, + "loss": 0.0509, + "step": 37120 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018099358974358972, + "loss": 0.05, + "step": 37130 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018096153846153842, + "loss": 0.0477, + "step": 37140 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018092948717948716, + "loss": 0.0589, + "step": 37150 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018089743589743586, + "loss": 0.0513, + "step": 37160 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001808653846153846, + "loss": 0.0525, + "step": 37170 + }, + { + "epoch": 0.79, + "learning_rate": 0.0001808333333333333, + "loss": 0.0522, + "step": 37180 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018080128205128203, + "loss": 0.0466, + "step": 37190 + }, + { + "epoch": 0.79, + "learning_rate": 0.00018076923076923074, + "loss": 0.0471, + "step": 37200 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018073717948717947, + "loss": 0.0556, + "step": 37210 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018070512820512818, + "loss": 0.0537, + "step": 37220 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018067307692307689, + "loss": 0.0471, + "step": 37230 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018064102564102562, + "loss": 0.0479, + "step": 37240 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018060897435897433, + "loss": 0.0629, + "step": 37250 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018057692307692306, + "loss": 0.0535, + "step": 37260 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018054487179487176, + "loss": 0.06, + "step": 37270 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001805128205128205, + "loss": 0.0563, + "step": 37280 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001804807692307692, + "loss": 0.0513, + "step": 37290 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018044871794871794, + "loss": 0.0466, + "step": 37300 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018041666666666664, + "loss": 0.0573, + "step": 37310 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018038461538461535, + "loss": 0.0554, + "step": 37320 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018035256410256408, + "loss": 0.0565, + "step": 37330 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001803205128205128, + "loss": 0.0539, + "step": 37340 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018028846153846152, + "loss": 0.0467, + "step": 37350 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018025641025641023, + "loss": 0.0526, + "step": 37360 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018022435897435896, + "loss": 0.0548, + "step": 37370 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018019230769230767, + "loss": 0.0524, + "step": 37380 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001801602564102564, + "loss": 0.0446, + "step": 37390 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001801282051282051, + "loss": 0.0481, + "step": 37400 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001800961538461538, + "loss": 0.0413, + "step": 37410 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018006410256410255, + "loss": 0.0482, + "step": 37420 + }, + { + "epoch": 0.8, + "learning_rate": 0.00018003205128205125, + "loss": 0.0564, + "step": 37430 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017999999999999998, + "loss": 0.0469, + "step": 37440 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001799679487179487, + "loss": 0.046, + "step": 37450 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017993589743589742, + "loss": 0.0528, + "step": 37460 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017990384615384613, + "loss": 0.0479, + "step": 37470 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017987179487179486, + "loss": 0.0521, + "step": 37480 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017983974358974357, + "loss": 0.0474, + "step": 37490 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017980769230769227, + "loss": 0.0517, + "step": 37500 + }, + { + "epoch": 0.8, + "learning_rate": 0.000179775641025641, + "loss": 0.0528, + "step": 37510 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017974358974358971, + "loss": 0.0506, + "step": 37520 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017971153846153845, + "loss": 0.0523, + "step": 37530 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017967948717948715, + "loss": 0.0461, + "step": 37540 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017964743589743589, + "loss": 0.0577, + "step": 37550 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001796153846153846, + "loss": 0.0502, + "step": 37560 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017958333333333333, + "loss": 0.0497, + "step": 37570 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017955128205128203, + "loss": 0.0505, + "step": 37580 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017951923076923074, + "loss": 0.0515, + "step": 37590 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017948717948717947, + "loss": 0.0478, + "step": 37600 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017945512820512818, + "loss": 0.0519, + "step": 37610 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001794230769230769, + "loss": 0.0559, + "step": 37620 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017939102564102562, + "loss": 0.0448, + "step": 37630 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017935897435897435, + "loss": 0.0455, + "step": 37640 + }, + { + "epoch": 0.8, + "learning_rate": 0.00017932692307692306, + "loss": 0.0465, + "step": 37650 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001792948717948718, + "loss": 0.0493, + "step": 37660 + }, + { + "epoch": 0.8, + "learning_rate": 0.0001792628205128205, + "loss": 0.0513, + "step": 37670 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001792307692307692, + "loss": 0.0527, + "step": 37680 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017919871794871793, + "loss": 0.0571, + "step": 37690 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017916666666666664, + "loss": 0.0568, + "step": 37700 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017913461538461537, + "loss": 0.0433, + "step": 37710 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017910256410256408, + "loss": 0.0458, + "step": 37720 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001790705128205128, + "loss": 0.0506, + "step": 37730 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017903846153846152, + "loss": 0.0537, + "step": 37740 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017900641025641022, + "loss": 0.0497, + "step": 37750 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017897435897435896, + "loss": 0.0504, + "step": 37760 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017894230769230766, + "loss": 0.0415, + "step": 37770 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001789102564102564, + "loss": 0.0524, + "step": 37780 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001788782051282051, + "loss": 0.0515, + "step": 37790 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017884615384615384, + "loss": 0.0535, + "step": 37800 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017881410256410254, + "loss": 0.0525, + "step": 37810 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017878205128205127, + "loss": 0.0623, + "step": 37820 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017874999999999998, + "loss": 0.0474, + "step": 37830 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001787179487179487, + "loss": 0.0469, + "step": 37840 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017868589743589742, + "loss": 0.0551, + "step": 37850 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017865384615384613, + "loss": 0.0493, + "step": 37860 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017862179487179486, + "loss": 0.0537, + "step": 37870 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017858974358974357, + "loss": 0.0564, + "step": 37880 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001785576923076923, + "loss": 0.0491, + "step": 37890 + }, + { + "epoch": 0.81, + "learning_rate": 0.000178525641025641, + "loss": 0.058, + "step": 37900 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017849358974358974, + "loss": 0.0524, + "step": 37910 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017846153846153844, + "loss": 0.0555, + "step": 37920 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017842948717948715, + "loss": 0.0549, + "step": 37930 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017839743589743588, + "loss": 0.053, + "step": 37940 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001783653846153846, + "loss": 0.0507, + "step": 37950 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017833333333333332, + "loss": 0.0472, + "step": 37960 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017830128205128203, + "loss": 0.0568, + "step": 37970 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017826923076923076, + "loss": 0.0507, + "step": 37980 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017823717948717947, + "loss": 0.0495, + "step": 37990 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001782051282051282, + "loss": 0.0489, + "step": 38000 + }, + { + "epoch": 0.81, + "eval_cer": 0.923620588767339, + "eval_loss": 0.03334152698516846, + "eval_runtime": 120.6129, + "eval_samples_per_second": 16.582, + "eval_steps_per_second": 4.145, + "step": 38000 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001781730769230769, + "loss": 0.0496, + "step": 38010 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001781410256410256, + "loss": 0.0465, + "step": 38020 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017810897435897435, + "loss": 0.0445, + "step": 38030 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017807692307692305, + "loss": 0.0502, + "step": 38040 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017804487179487179, + "loss": 0.0594, + "step": 38050 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001780128205128205, + "loss": 0.0555, + "step": 38060 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017798076923076922, + "loss": 0.0537, + "step": 38070 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017794871794871793, + "loss": 0.0475, + "step": 38080 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017791666666666666, + "loss": 0.0451, + "step": 38090 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017788461538461537, + "loss": 0.0497, + "step": 38100 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017785256410256408, + "loss": 0.0545, + "step": 38110 + }, + { + "epoch": 0.81, + "learning_rate": 0.0001778205128205128, + "loss": 0.0523, + "step": 38120 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017778846153846151, + "loss": 0.0459, + "step": 38130 + }, + { + "epoch": 0.81, + "learning_rate": 0.00017775641025641025, + "loss": 0.0496, + "step": 38140 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017772435897435895, + "loss": 0.0552, + "step": 38150 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001776923076923077, + "loss": 0.0476, + "step": 38160 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001776602564102564, + "loss": 0.0442, + "step": 38170 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017762820512820513, + "loss": 0.0497, + "step": 38180 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017759615384615383, + "loss": 0.0527, + "step": 38190 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017756410256410254, + "loss": 0.0483, + "step": 38200 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017753205128205127, + "loss": 0.0615, + "step": 38210 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017749999999999998, + "loss": 0.0492, + "step": 38220 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001774679487179487, + "loss": 0.0478, + "step": 38230 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017743589743589742, + "loss": 0.0465, + "step": 38240 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017740384615384615, + "loss": 0.0481, + "step": 38250 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017737179487179486, + "loss": 0.0543, + "step": 38260 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001773397435897436, + "loss": 0.0466, + "step": 38270 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001773076923076923, + "loss": 0.058, + "step": 38280 + }, + { + "epoch": 0.82, + "learning_rate": 0.000177275641025641, + "loss": 0.0485, + "step": 38290 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017724358974358973, + "loss": 0.0518, + "step": 38300 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017721153846153844, + "loss": 0.0486, + "step": 38310 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017717948717948717, + "loss": 0.0502, + "step": 38320 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017714743589743588, + "loss": 0.062, + "step": 38330 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001771153846153846, + "loss": 0.0477, + "step": 38340 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017708333333333332, + "loss": 0.0481, + "step": 38350 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017705128205128205, + "loss": 0.0468, + "step": 38360 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017701923076923076, + "loss": 0.0498, + "step": 38370 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017698717948717946, + "loss": 0.063, + "step": 38380 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001769551282051282, + "loss": 0.0539, + "step": 38390 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001769230769230769, + "loss": 0.0531, + "step": 38400 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017689102564102564, + "loss": 0.0531, + "step": 38410 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017685897435897434, + "loss": 0.0521, + "step": 38420 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017682692307692308, + "loss": 0.0481, + "step": 38430 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017679487179487178, + "loss": 0.0533, + "step": 38440 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017676282051282051, + "loss": 0.0541, + "step": 38450 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017673076923076922, + "loss": 0.0565, + "step": 38460 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017669871794871793, + "loss": 0.0543, + "step": 38470 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017666666666666666, + "loss": 0.0466, + "step": 38480 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017663461538461537, + "loss": 0.0445, + "step": 38490 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001766025641025641, + "loss": 0.0429, + "step": 38500 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001765705128205128, + "loss": 0.0436, + "step": 38510 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017653846153846154, + "loss": 0.0452, + "step": 38520 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017650641025641024, + "loss": 0.0514, + "step": 38530 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017647435897435898, + "loss": 0.0459, + "step": 38540 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017644230769230768, + "loss": 0.0452, + "step": 38550 + }, + { + "epoch": 0.82, + "learning_rate": 0.0001764102564102564, + "loss": 0.0403, + "step": 38560 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017637820512820512, + "loss": 0.0472, + "step": 38570 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017634615384615383, + "loss": 0.0463, + "step": 38580 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017631410256410256, + "loss": 0.0527, + "step": 38590 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017628205128205127, + "loss": 0.0457, + "step": 38600 + }, + { + "epoch": 0.82, + "learning_rate": 0.00017625, + "loss": 0.0572, + "step": 38610 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001762179487179487, + "loss": 0.0509, + "step": 38620 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017618589743589744, + "loss": 0.0495, + "step": 38630 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017615384615384615, + "loss": 0.0453, + "step": 38640 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017612179487179485, + "loss": 0.0486, + "step": 38650 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017608974358974359, + "loss": 0.0488, + "step": 38660 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001760576923076923, + "loss": 0.0605, + "step": 38670 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017602564102564103, + "loss": 0.0475, + "step": 38680 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017599358974358973, + "loss": 0.0497, + "step": 38690 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017596153846153846, + "loss": 0.0552, + "step": 38700 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017592948717948717, + "loss": 0.0475, + "step": 38710 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017589743589743588, + "loss": 0.0517, + "step": 38720 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001758653846153846, + "loss": 0.0445, + "step": 38730 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017583333333333332, + "loss": 0.0479, + "step": 38740 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017580128205128205, + "loss": 0.0513, + "step": 38750 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017576923076923075, + "loss": 0.0457, + "step": 38760 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001757371794871795, + "loss": 0.0495, + "step": 38770 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001757051282051282, + "loss": 0.0569, + "step": 38780 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017567307692307693, + "loss": 0.0509, + "step": 38790 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017564102564102563, + "loss": 0.0538, + "step": 38800 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017560897435897434, + "loss": 0.045, + "step": 38810 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017557692307692307, + "loss": 0.0439, + "step": 38820 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017554487179487178, + "loss": 0.0406, + "step": 38830 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001755128205128205, + "loss": 0.0464, + "step": 38840 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017548076923076922, + "loss": 0.061, + "step": 38850 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017544871794871795, + "loss": 0.0522, + "step": 38860 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017541666666666666, + "loss": 0.0458, + "step": 38870 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001753846153846154, + "loss": 0.0562, + "step": 38880 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001753525641025641, + "loss": 0.0511, + "step": 38890 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001753205128205128, + "loss": 0.0428, + "step": 38900 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017528846153846154, + "loss": 0.0412, + "step": 38910 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017525641025641024, + "loss": 0.0514, + "step": 38920 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017522435897435897, + "loss": 0.0482, + "step": 38930 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017519230769230768, + "loss": 0.0532, + "step": 38940 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017516025641025641, + "loss": 0.0452, + "step": 38950 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017512820512820512, + "loss": 0.0552, + "step": 38960 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017509615384615385, + "loss": 0.0507, + "step": 38970 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017506410256410256, + "loss": 0.0435, + "step": 38980 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017503205128205127, + "loss": 0.0483, + "step": 38990 + }, + { + "epoch": 0.83, + "learning_rate": 0.000175, + "loss": 0.0493, + "step": 39000 + }, + { + "epoch": 0.83, + "eval_cer": 0.9236562582538768, + "eval_loss": 0.033102501183748245, + "eval_runtime": 120.5659, + "eval_samples_per_second": 16.588, + "eval_steps_per_second": 4.147, + "step": 39000 + }, + { + "epoch": 0.83, + "learning_rate": 0.0001749679487179487, + "loss": 0.0489, + "step": 39010 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017493589743589744, + "loss": 0.0495, + "step": 39020 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017490384615384614, + "loss": 0.0459, + "step": 39030 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017487179487179488, + "loss": 0.0419, + "step": 39040 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017483974358974358, + "loss": 0.0453, + "step": 39050 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017480769230769232, + "loss": 0.0479, + "step": 39060 + }, + { + "epoch": 0.83, + "learning_rate": 0.00017477564102564102, + "loss": 0.0584, + "step": 39070 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017474358974358973, + "loss": 0.0408, + "step": 39080 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017471153846153846, + "loss": 0.0431, + "step": 39090 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017467948717948717, + "loss": 0.0434, + "step": 39100 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001746474358974359, + "loss": 0.0516, + "step": 39110 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001746153846153846, + "loss": 0.0491, + "step": 39120 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017458333333333334, + "loss": 0.054, + "step": 39130 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017455128205128205, + "loss": 0.0555, + "step": 39140 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017451923076923078, + "loss": 0.0532, + "step": 39150 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017448717948717948, + "loss": 0.0475, + "step": 39160 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001744551282051282, + "loss": 0.0426, + "step": 39170 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017442307692307692, + "loss": 0.0486, + "step": 39180 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017439102564102563, + "loss": 0.0571, + "step": 39190 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017435897435897436, + "loss": 0.0463, + "step": 39200 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017432692307692307, + "loss": 0.0564, + "step": 39210 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001742948717948718, + "loss": 0.0514, + "step": 39220 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001742628205128205, + "loss": 0.0433, + "step": 39230 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017423076923076924, + "loss": 0.0426, + "step": 39240 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017419871794871795, + "loss": 0.05, + "step": 39250 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017416666666666665, + "loss": 0.0453, + "step": 39260 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001741346153846154, + "loss": 0.0434, + "step": 39270 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001741025641025641, + "loss": 0.0515, + "step": 39280 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017407051282051283, + "loss": 0.0502, + "step": 39290 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017403846153846153, + "loss": 0.054, + "step": 39300 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017400641025641027, + "loss": 0.059, + "step": 39310 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017397435897435897, + "loss": 0.0507, + "step": 39320 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001739423076923077, + "loss": 0.0477, + "step": 39330 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001739102564102564, + "loss": 0.0548, + "step": 39340 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017387820512820512, + "loss": 0.0467, + "step": 39350 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017384615384615385, + "loss": 0.0508, + "step": 39360 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017381410256410256, + "loss": 0.0519, + "step": 39370 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001737820512820513, + "loss": 0.0431, + "step": 39380 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017375, + "loss": 0.0449, + "step": 39390 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017371794871794873, + "loss": 0.0513, + "step": 39400 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017368589743589743, + "loss": 0.0408, + "step": 39410 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017365384615384617, + "loss": 0.0489, + "step": 39420 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017362179487179487, + "loss": 0.0462, + "step": 39430 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017358974358974358, + "loss": 0.0531, + "step": 39440 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001735576923076923, + "loss": 0.0471, + "step": 39450 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017352564102564102, + "loss": 0.0552, + "step": 39460 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017349358974358975, + "loss": 0.052, + "step": 39470 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017346153846153846, + "loss": 0.0539, + "step": 39480 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017342948717948714, + "loss": 0.0498, + "step": 39490 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017339743589743587, + "loss": 0.0371, + "step": 39500 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017336538461538458, + "loss": 0.0469, + "step": 39510 + }, + { + "epoch": 0.84, + "learning_rate": 0.0001733333333333333, + "loss": 0.0467, + "step": 39520 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017330128205128202, + "loss": 0.052, + "step": 39530 + }, + { + "epoch": 0.84, + "learning_rate": 0.00017326923076923075, + "loss": 0.0502, + "step": 39540 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017323717948717945, + "loss": 0.0429, + "step": 39550 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001732051282051282, + "loss": 0.0503, + "step": 39560 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001731730769230769, + "loss": 0.0517, + "step": 39570 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001731410256410256, + "loss": 0.0458, + "step": 39580 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017310897435897433, + "loss": 0.0496, + "step": 39590 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017307692307692304, + "loss": 0.0429, + "step": 39600 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017304487179487177, + "loss": 0.0595, + "step": 39610 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017301282051282048, + "loss": 0.0561, + "step": 39620 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001729807692307692, + "loss": 0.0529, + "step": 39630 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017294871794871792, + "loss": 0.0484, + "step": 39640 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017291666666666662, + "loss": 0.0443, + "step": 39650 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017288461538461536, + "loss": 0.0512, + "step": 39660 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017285256410256406, + "loss": 0.0434, + "step": 39670 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001728205128205128, + "loss": 0.0466, + "step": 39680 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001727884615384615, + "loss": 0.0449, + "step": 39690 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017275641025641023, + "loss": 0.0506, + "step": 39700 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017272435897435894, + "loss": 0.0476, + "step": 39710 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017269230769230767, + "loss": 0.0464, + "step": 39720 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017266025641025638, + "loss": 0.0445, + "step": 39730 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017262820512820509, + "loss": 0.0435, + "step": 39740 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017259615384615382, + "loss": 0.0482, + "step": 39750 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017256410256410253, + "loss": 0.0412, + "step": 39760 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017253205128205126, + "loss": 0.0493, + "step": 39770 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017249999999999996, + "loss": 0.053, + "step": 39780 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001724679487179487, + "loss": 0.0592, + "step": 39790 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001724358974358974, + "loss": 0.0503, + "step": 39800 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017240384615384614, + "loss": 0.0471, + "step": 39810 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017237179487179484, + "loss": 0.0481, + "step": 39820 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017233974358974355, + "loss": 0.0512, + "step": 39830 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017230769230769228, + "loss": 0.0542, + "step": 39840 + }, + { + "epoch": 0.85, + "learning_rate": 0.000172275641025641, + "loss": 0.0434, + "step": 39850 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017224358974358972, + "loss": 0.0449, + "step": 39860 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017221153846153843, + "loss": 0.0449, + "step": 39870 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017217948717948716, + "loss": 0.0478, + "step": 39880 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017214743589743587, + "loss": 0.0547, + "step": 39890 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001721153846153846, + "loss": 0.0559, + "step": 39900 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001720833333333333, + "loss": 0.0412, + "step": 39910 + }, + { + "epoch": 0.85, + "learning_rate": 0.000172051282051282, + "loss": 0.0491, + "step": 39920 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017201923076923075, + "loss": 0.0549, + "step": 39930 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017198717948717945, + "loss": 0.0425, + "step": 39940 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017195512820512818, + "loss": 0.046, + "step": 39950 + }, + { + "epoch": 0.85, + "learning_rate": 0.0001719230769230769, + "loss": 0.0509, + "step": 39960 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017189102564102562, + "loss": 0.0501, + "step": 39970 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017185897435897433, + "loss": 0.0593, + "step": 39980 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017182692307692306, + "loss": 0.0499, + "step": 39990 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017179487179487177, + "loss": 0.0481, + "step": 40000 + }, + { + "epoch": 0.85, + "eval_cer": 0.9236919277404144, + "eval_loss": 0.03232666105031967, + "eval_runtime": 120.7408, + "eval_samples_per_second": 16.564, + "eval_steps_per_second": 4.141, + "step": 40000 + }, + { + "epoch": 0.85, + "learning_rate": 0.00017176282051282047, + "loss": 0.044, + "step": 40010 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001717307692307692, + "loss": 0.0492, + "step": 40020 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017169871794871791, + "loss": 0.0472, + "step": 40030 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017166666666666665, + "loss": 0.0505, + "step": 40040 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017163461538461535, + "loss": 0.0437, + "step": 40050 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017160256410256409, + "loss": 0.0485, + "step": 40060 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001715705128205128, + "loss": 0.0497, + "step": 40070 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017153846153846153, + "loss": 0.0458, + "step": 40080 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017150641025641023, + "loss": 0.0492, + "step": 40090 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017147435897435894, + "loss": 0.0444, + "step": 40100 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017144230769230767, + "loss": 0.0507, + "step": 40110 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017141025641025638, + "loss": 0.0491, + "step": 40120 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001713782051282051, + "loss": 0.0513, + "step": 40130 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017134615384615382, + "loss": 0.0515, + "step": 40140 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017131410256410255, + "loss": 0.0454, + "step": 40150 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017128205128205126, + "loss": 0.0523, + "step": 40160 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017125, + "loss": 0.0477, + "step": 40170 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001712179487179487, + "loss": 0.0501, + "step": 40180 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001711858974358974, + "loss": 0.0464, + "step": 40190 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017115384615384613, + "loss": 0.047, + "step": 40200 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017112179487179484, + "loss": 0.0508, + "step": 40210 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017108974358974357, + "loss": 0.05, + "step": 40220 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017105769230769228, + "loss": 0.0594, + "step": 40230 + }, + { + "epoch": 0.86, + "learning_rate": 0.000171025641025641, + "loss": 0.0445, + "step": 40240 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017099358974358972, + "loss": 0.0473, + "step": 40250 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017096153846153845, + "loss": 0.0505, + "step": 40260 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017092948717948716, + "loss": 0.0647, + "step": 40270 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017089743589743586, + "loss": 0.0505, + "step": 40280 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001708653846153846, + "loss": 0.0536, + "step": 40290 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001708333333333333, + "loss": 0.0467, + "step": 40300 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017080128205128204, + "loss": 0.0501, + "step": 40310 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017076923076923074, + "loss": 0.0561, + "step": 40320 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017073717948717947, + "loss": 0.054, + "step": 40330 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017070512820512818, + "loss": 0.0471, + "step": 40340 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017067307692307691, + "loss": 0.0505, + "step": 40350 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017064102564102562, + "loss": 0.0467, + "step": 40360 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017060897435897433, + "loss": 0.043, + "step": 40370 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017057692307692306, + "loss": 0.0468, + "step": 40380 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017054487179487177, + "loss": 0.0442, + "step": 40390 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001705128205128205, + "loss": 0.0437, + "step": 40400 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001704807692307692, + "loss": 0.0427, + "step": 40410 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017044871794871794, + "loss": 0.0512, + "step": 40420 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017041666666666664, + "loss": 0.049, + "step": 40430 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017038461538461538, + "loss": 0.0472, + "step": 40440 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017035256410256408, + "loss": 0.0465, + "step": 40450 + }, + { + "epoch": 0.86, + "learning_rate": 0.0001703205128205128, + "loss": 0.0526, + "step": 40460 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017028846153846152, + "loss": 0.0426, + "step": 40470 + }, + { + "epoch": 0.86, + "learning_rate": 0.00017025641025641023, + "loss": 0.0428, + "step": 40480 + }, + { + "epoch": 0.87, + "learning_rate": 0.00017022435897435896, + "loss": 0.0614, + "step": 40490 + }, + { + "epoch": 0.87, + "learning_rate": 0.00017019230769230767, + "loss": 0.0473, + "step": 40500 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001701602564102564, + "loss": 0.0498, + "step": 40510 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001701282051282051, + "loss": 0.0487, + "step": 40520 + }, + { + "epoch": 0.87, + "learning_rate": 0.00017009615384615384, + "loss": 0.0491, + "step": 40530 + }, + { + "epoch": 0.87, + "learning_rate": 0.00017006410256410255, + "loss": 0.0537, + "step": 40540 + }, + { + "epoch": 0.87, + "learning_rate": 0.00017003205128205125, + "loss": 0.0489, + "step": 40550 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016999999999999999, + "loss": 0.0566, + "step": 40560 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001699679487179487, + "loss": 0.0543, + "step": 40570 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016993589743589742, + "loss": 0.0487, + "step": 40580 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016990384615384613, + "loss": 0.041, + "step": 40590 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016987179487179486, + "loss": 0.0497, + "step": 40600 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016983974358974357, + "loss": 0.0469, + "step": 40610 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016980769230769228, + "loss": 0.0489, + "step": 40620 + }, + { + "epoch": 0.87, + "learning_rate": 0.000169775641025641, + "loss": 0.0587, + "step": 40630 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016974358974358971, + "loss": 0.048, + "step": 40640 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016971153846153845, + "loss": 0.0488, + "step": 40650 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016967948717948715, + "loss": 0.0461, + "step": 40660 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001696474358974359, + "loss": 0.0468, + "step": 40670 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001696153846153846, + "loss": 0.0449, + "step": 40680 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016958333333333333, + "loss": 0.0472, + "step": 40690 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016955128205128203, + "loss": 0.0452, + "step": 40700 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016951923076923074, + "loss": 0.045, + "step": 40710 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016948717948717947, + "loss": 0.0433, + "step": 40720 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016945512820512818, + "loss": 0.0488, + "step": 40730 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001694230769230769, + "loss": 0.0405, + "step": 40740 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016939102564102562, + "loss": 0.0585, + "step": 40750 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016935897435897435, + "loss": 0.0458, + "step": 40760 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016932692307692306, + "loss": 0.0417, + "step": 40770 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001692948717948718, + "loss": 0.0409, + "step": 40780 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001692628205128205, + "loss": 0.0443, + "step": 40790 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001692307692307692, + "loss": 0.0418, + "step": 40800 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016919871794871793, + "loss": 0.0428, + "step": 40810 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016916666666666664, + "loss": 0.0467, + "step": 40820 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016913461538461537, + "loss": 0.0512, + "step": 40830 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016910256410256408, + "loss": 0.0467, + "step": 40840 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001690705128205128, + "loss": 0.0441, + "step": 40850 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016903846153846152, + "loss": 0.0494, + "step": 40860 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016900641025641025, + "loss": 0.049, + "step": 40870 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016897435897435896, + "loss": 0.0534, + "step": 40880 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016894230769230766, + "loss": 0.0518, + "step": 40890 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001689102564102564, + "loss": 0.0458, + "step": 40900 + }, + { + "epoch": 0.87, + "learning_rate": 0.0001688782051282051, + "loss": 0.0488, + "step": 40910 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016884615384615384, + "loss": 0.0463, + "step": 40920 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016881410256410254, + "loss": 0.0499, + "step": 40930 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016878205128205128, + "loss": 0.054, + "step": 40940 + }, + { + "epoch": 0.87, + "learning_rate": 0.00016874999999999998, + "loss": 0.052, + "step": 40950 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016871794871794871, + "loss": 0.0551, + "step": 40960 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016868589743589742, + "loss": 0.0441, + "step": 40970 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016865384615384613, + "loss": 0.0583, + "step": 40980 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016862179487179486, + "loss": 0.0542, + "step": 40990 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016858974358974357, + "loss": 0.0444, + "step": 41000 + }, + { + "epoch": 0.88, + "eval_cer": 0.9237369392353312, + "eval_loss": 0.03184425085783005, + "eval_runtime": 120.4545, + "eval_samples_per_second": 16.604, + "eval_steps_per_second": 4.151, + "step": 41000 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001685576923076923, + "loss": 0.0457, + "step": 41010 + }, + { + "epoch": 0.88, + "learning_rate": 0.000168525641025641, + "loss": 0.056, + "step": 41020 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016849358974358974, + "loss": 0.0571, + "step": 41030 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016846153846153844, + "loss": 0.0484, + "step": 41040 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016842948717948718, + "loss": 0.0505, + "step": 41050 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016839743589743588, + "loss": 0.0479, + "step": 41060 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001683653846153846, + "loss": 0.0511, + "step": 41070 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016833333333333332, + "loss": 0.0526, + "step": 41080 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016830128205128203, + "loss": 0.0579, + "step": 41090 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016826923076923076, + "loss": 0.046, + "step": 41100 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016823717948717947, + "loss": 0.0458, + "step": 41110 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001682051282051282, + "loss": 0.0451, + "step": 41120 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001681730769230769, + "loss": 0.051, + "step": 41130 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016814102564102564, + "loss": 0.0514, + "step": 41140 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016810897435897435, + "loss": 0.0515, + "step": 41150 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016807692307692305, + "loss": 0.0469, + "step": 41160 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016804487179487179, + "loss": 0.0448, + "step": 41170 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001680128205128205, + "loss": 0.0597, + "step": 41180 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016798076923076923, + "loss": 0.049, + "step": 41190 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016794871794871793, + "loss": 0.0599, + "step": 41200 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016791666666666666, + "loss": 0.0481, + "step": 41210 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016788461538461537, + "loss": 0.0476, + "step": 41220 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001678525641025641, + "loss": 0.0471, + "step": 41230 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001678205128205128, + "loss": 0.0513, + "step": 41240 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016778846153846152, + "loss": 0.0524, + "step": 41250 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016775641025641025, + "loss": 0.0417, + "step": 41260 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016772435897435895, + "loss": 0.0611, + "step": 41270 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001676923076923077, + "loss": 0.0462, + "step": 41280 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001676602564102564, + "loss": 0.0526, + "step": 41290 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016762820512820513, + "loss": 0.0529, + "step": 41300 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016759615384615383, + "loss": 0.0477, + "step": 41310 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016756410256410257, + "loss": 0.0506, + "step": 41320 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016753205128205127, + "loss": 0.0537, + "step": 41330 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016749999999999998, + "loss": 0.0579, + "step": 41340 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001674679487179487, + "loss": 0.0525, + "step": 41350 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016743589743589742, + "loss": 0.0447, + "step": 41360 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016740384615384615, + "loss": 0.0467, + "step": 41370 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016737179487179486, + "loss": 0.0534, + "step": 41380 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001673397435897436, + "loss": 0.0492, + "step": 41390 + }, + { + "epoch": 0.88, + "learning_rate": 0.0001673076923076923, + "loss": 0.0466, + "step": 41400 + }, + { + "epoch": 0.88, + "learning_rate": 0.00016727564102564103, + "loss": 0.0495, + "step": 41410 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016724358974358974, + "loss": 0.0448, + "step": 41420 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016721153846153844, + "loss": 0.046, + "step": 41430 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016717948717948717, + "loss": 0.0423, + "step": 41440 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016714743589743588, + "loss": 0.0445, + "step": 41450 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016711538461538461, + "loss": 0.0493, + "step": 41460 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016708333333333332, + "loss": 0.051, + "step": 41470 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016705128205128205, + "loss": 0.0415, + "step": 41480 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016701923076923076, + "loss": 0.0509, + "step": 41490 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001669871794871795, + "loss": 0.0472, + "step": 41500 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001669551282051282, + "loss": 0.0534, + "step": 41510 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001669230769230769, + "loss": 0.0474, + "step": 41520 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016689102564102564, + "loss": 0.0527, + "step": 41530 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016685897435897434, + "loss": 0.0463, + "step": 41540 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016682692307692308, + "loss": 0.0459, + "step": 41550 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016679487179487178, + "loss": 0.0445, + "step": 41560 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016676282051282052, + "loss": 0.0468, + "step": 41570 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016673076923076922, + "loss": 0.0474, + "step": 41580 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016669871794871793, + "loss": 0.0462, + "step": 41590 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016666666666666666, + "loss": 0.0565, + "step": 41600 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016663461538461537, + "loss": 0.0448, + "step": 41610 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001666025641025641, + "loss": 0.046, + "step": 41620 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001665705128205128, + "loss": 0.0467, + "step": 41630 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016653846153846154, + "loss": 0.0497, + "step": 41640 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016650641025641025, + "loss": 0.0542, + "step": 41650 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016647435897435898, + "loss": 0.0542, + "step": 41660 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016644230769230768, + "loss": 0.0464, + "step": 41670 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001664102564102564, + "loss": 0.0503, + "step": 41680 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016637820512820512, + "loss": 0.0479, + "step": 41690 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016634615384615383, + "loss": 0.0587, + "step": 41700 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016631410256410256, + "loss": 0.0572, + "step": 41710 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016628205128205127, + "loss": 0.0497, + "step": 41720 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016625, + "loss": 0.0401, + "step": 41730 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001662179487179487, + "loss": 0.0486, + "step": 41740 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016618589743589744, + "loss": 0.0501, + "step": 41750 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016615384615384615, + "loss": 0.0421, + "step": 41760 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016612179487179485, + "loss": 0.0465, + "step": 41770 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001660897435897436, + "loss": 0.0529, + "step": 41780 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001660576923076923, + "loss": 0.0479, + "step": 41790 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016602564102564103, + "loss": 0.0437, + "step": 41800 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016599358974358973, + "loss": 0.0434, + "step": 41810 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016596153846153847, + "loss": 0.0539, + "step": 41820 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016592948717948717, + "loss": 0.0501, + "step": 41830 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001658974358974359, + "loss": 0.0521, + "step": 41840 + }, + { + "epoch": 0.89, + "learning_rate": 0.0001658653846153846, + "loss": 0.0449, + "step": 41850 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016583333333333332, + "loss": 0.0474, + "step": 41860 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016580128205128205, + "loss": 0.05, + "step": 41870 + }, + { + "epoch": 0.89, + "learning_rate": 0.00016576923076923076, + "loss": 0.0543, + "step": 41880 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001657371794871795, + "loss": 0.0499, + "step": 41890 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001657051282051282, + "loss": 0.0457, + "step": 41900 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016567307692307693, + "loss": 0.0463, + "step": 41910 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016564102564102563, + "loss": 0.0479, + "step": 41920 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016560897435897437, + "loss": 0.0482, + "step": 41930 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016557692307692307, + "loss": 0.0509, + "step": 41940 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016554487179487178, + "loss": 0.0454, + "step": 41950 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001655128205128205, + "loss": 0.0509, + "step": 41960 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016548076923076922, + "loss": 0.0467, + "step": 41970 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016544871794871795, + "loss": 0.0399, + "step": 41980 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016541666666666666, + "loss": 0.0426, + "step": 41990 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001653846153846154, + "loss": 0.0446, + "step": 42000 + }, + { + "epoch": 0.9, + "eval_cer": 0.9238227158577195, + "eval_loss": 0.031112128868699074, + "eval_runtime": 120.6436, + "eval_samples_per_second": 16.578, + "eval_steps_per_second": 4.144, + "step": 42000 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001653525641025641, + "loss": 0.0517, + "step": 42010 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016532051282051283, + "loss": 0.0519, + "step": 42020 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016528846153846154, + "loss": 0.0451, + "step": 42030 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016525641025641024, + "loss": 0.0455, + "step": 42040 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016522435897435898, + "loss": 0.0538, + "step": 42050 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016519230769230768, + "loss": 0.0551, + "step": 42060 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016516025641025641, + "loss": 0.0469, + "step": 42070 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016512820512820512, + "loss": 0.059, + "step": 42080 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016509615384615385, + "loss": 0.0416, + "step": 42090 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016506410256410256, + "loss": 0.0477, + "step": 42100 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001650320512820513, + "loss": 0.045, + "step": 42110 + }, + { + "epoch": 0.9, + "learning_rate": 0.000165, + "loss": 0.0456, + "step": 42120 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001649679487179487, + "loss": 0.0496, + "step": 42130 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016493589743589744, + "loss": 0.0487, + "step": 42140 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016490384615384614, + "loss": 0.0487, + "step": 42150 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016487179487179488, + "loss": 0.0544, + "step": 42160 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016483974358974358, + "loss": 0.05, + "step": 42170 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016480769230769232, + "loss": 0.0427, + "step": 42180 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016477564102564102, + "loss": 0.0529, + "step": 42190 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016474358974358976, + "loss": 0.0466, + "step": 42200 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016471153846153846, + "loss": 0.0466, + "step": 42210 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016467948717948717, + "loss": 0.0479, + "step": 42220 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001646474358974359, + "loss": 0.0418, + "step": 42230 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001646153846153846, + "loss": 0.0408, + "step": 42240 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016458333333333334, + "loss": 0.0564, + "step": 42250 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016455128205128205, + "loss": 0.0411, + "step": 42260 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016451923076923078, + "loss": 0.0414, + "step": 42270 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016448717948717949, + "loss": 0.0464, + "step": 42280 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016445512820512822, + "loss": 0.0526, + "step": 42290 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016442307692307692, + "loss": 0.0402, + "step": 42300 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016439102564102563, + "loss": 0.0466, + "step": 42310 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016435897435897436, + "loss": 0.0503, + "step": 42320 + }, + { + "epoch": 0.9, + "learning_rate": 0.00016432692307692307, + "loss": 0.0496, + "step": 42330 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001642948717948718, + "loss": 0.0504, + "step": 42340 + }, + { + "epoch": 0.9, + "learning_rate": 0.0001642628205128205, + "loss": 0.0508, + "step": 42350 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016423076923076924, + "loss": 0.0656, + "step": 42360 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016419871794871795, + "loss": 0.0473, + "step": 42370 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016416666666666668, + "loss": 0.0481, + "step": 42380 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001641346153846154, + "loss": 0.0523, + "step": 42390 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001641025641025641, + "loss": 0.0415, + "step": 42400 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016407051282051283, + "loss": 0.0511, + "step": 42410 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001640384615384615, + "loss": 0.0837, + "step": 42420 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016400641025641024, + "loss": 0.0443, + "step": 42430 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016397435897435895, + "loss": 0.0463, + "step": 42440 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016394230769230765, + "loss": 0.0565, + "step": 42450 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016391025641025638, + "loss": 0.0555, + "step": 42460 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001638782051282051, + "loss": 0.0448, + "step": 42470 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016384615384615382, + "loss": 0.0495, + "step": 42480 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016381410256410253, + "loss": 0.0475, + "step": 42490 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016378205128205126, + "loss": 0.0398, + "step": 42500 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016374999999999997, + "loss": 0.0526, + "step": 42510 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016371794871794867, + "loss": 0.0459, + "step": 42520 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001636858974358974, + "loss": 0.0435, + "step": 42530 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016365384615384611, + "loss": 0.0393, + "step": 42540 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016362179487179485, + "loss": 0.0579, + "step": 42550 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016358974358974355, + "loss": 0.046, + "step": 42560 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016355769230769229, + "loss": 0.0691, + "step": 42570 + }, + { + "epoch": 0.91, + "learning_rate": 0.000163525641025641, + "loss": 0.0534, + "step": 42580 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016349358974358973, + "loss": 0.0469, + "step": 42590 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016346153846153843, + "loss": 0.0463, + "step": 42600 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016342948717948714, + "loss": 0.0529, + "step": 42610 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016339743589743587, + "loss": 0.0454, + "step": 42620 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016336538461538458, + "loss": 0.0514, + "step": 42630 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001633333333333333, + "loss": 0.0428, + "step": 42640 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016330128205128202, + "loss": 0.0602, + "step": 42650 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016326923076923075, + "loss": 0.0452, + "step": 42660 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016323717948717946, + "loss": 0.0521, + "step": 42670 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001632051282051282, + "loss": 0.0434, + "step": 42680 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001631730769230769, + "loss": 0.0509, + "step": 42690 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001631410256410256, + "loss": 0.056, + "step": 42700 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016310897435897433, + "loss": 0.0482, + "step": 42710 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016307692307692304, + "loss": 0.0434, + "step": 42720 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016304487179487177, + "loss": 0.0469, + "step": 42730 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016301282051282048, + "loss": 0.0453, + "step": 42740 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001629807692307692, + "loss": 0.0515, + "step": 42750 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016294871794871792, + "loss": 0.0474, + "step": 42760 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016291666666666665, + "loss": 0.0438, + "step": 42770 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016288461538461536, + "loss": 0.0465, + "step": 42780 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016285256410256406, + "loss": 0.0475, + "step": 42790 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001628205128205128, + "loss": 0.0567, + "step": 42800 + }, + { + "epoch": 0.91, + "learning_rate": 0.0001627884615384615, + "loss": 0.0432, + "step": 42810 + }, + { + "epoch": 0.91, + "learning_rate": 0.00016275641025641024, + "loss": 0.0486, + "step": 42820 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016272435897435894, + "loss": 0.0441, + "step": 42830 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016269230769230767, + "loss": 0.0515, + "step": 42840 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016266025641025638, + "loss": 0.0502, + "step": 42850 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016262820512820511, + "loss": 0.0523, + "step": 42860 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016259615384615382, + "loss": 0.0463, + "step": 42870 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016256410256410253, + "loss": 0.0505, + "step": 42880 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016253205128205126, + "loss": 0.0417, + "step": 42890 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016249999999999997, + "loss": 0.0503, + "step": 42900 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001624679487179487, + "loss": 0.0469, + "step": 42910 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001624358974358974, + "loss": 0.0605, + "step": 42920 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016240384615384614, + "loss": 0.0485, + "step": 42930 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016237179487179484, + "loss": 0.0465, + "step": 42940 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016233974358974358, + "loss": 0.0415, + "step": 42950 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016230769230769228, + "loss": 0.05, + "step": 42960 + }, + { + "epoch": 0.92, + "learning_rate": 0.000162275641025641, + "loss": 0.0571, + "step": 42970 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016224358974358972, + "loss": 0.0536, + "step": 42980 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016221153846153843, + "loss": 0.0417, + "step": 42990 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016217948717948716, + "loss": 0.0469, + "step": 43000 + }, + { + "epoch": 0.92, + "eval_cer": 0.9237174059450842, + "eval_loss": 0.03107159025967121, + "eval_runtime": 120.5835, + "eval_samples_per_second": 16.586, + "eval_steps_per_second": 4.147, + "step": 43000 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016214743589743587, + "loss": 0.0396, + "step": 43010 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001621153846153846, + "loss": 0.0421, + "step": 43020 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001620833333333333, + "loss": 0.0507, + "step": 43030 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016205128205128204, + "loss": 0.0429, + "step": 43040 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016201923076923075, + "loss": 0.0473, + "step": 43050 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016198717948717945, + "loss": 0.0538, + "step": 43060 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016195512820512819, + "loss": 0.0452, + "step": 43070 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001619230769230769, + "loss": 0.0509, + "step": 43080 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016189102564102562, + "loss": 0.048, + "step": 43090 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016185897435897433, + "loss": 0.0474, + "step": 43100 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016182692307692306, + "loss": 0.0457, + "step": 43110 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016179487179487177, + "loss": 0.0464, + "step": 43120 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001617628205128205, + "loss": 0.0469, + "step": 43130 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001617307692307692, + "loss": 0.0441, + "step": 43140 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016169871794871791, + "loss": 0.0442, + "step": 43150 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016166666666666665, + "loss": 0.0485, + "step": 43160 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016163461538461535, + "loss": 0.0483, + "step": 43170 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001616025641025641, + "loss": 0.0424, + "step": 43180 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001615705128205128, + "loss": 0.0423, + "step": 43190 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016153846153846153, + "loss": 0.0412, + "step": 43200 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016150641025641023, + "loss": 0.0396, + "step": 43210 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016147435897435897, + "loss": 0.0511, + "step": 43220 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016144230769230767, + "loss": 0.0504, + "step": 43230 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016141025641025638, + "loss": 0.0545, + "step": 43240 + }, + { + "epoch": 0.92, + "learning_rate": 0.0001613782051282051, + "loss": 0.0442, + "step": 43250 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016134615384615382, + "loss": 0.0419, + "step": 43260 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016131410256410255, + "loss": 0.0468, + "step": 43270 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016128205128205126, + "loss": 0.0455, + "step": 43280 + }, + { + "epoch": 0.92, + "learning_rate": 0.00016125, + "loss": 0.0426, + "step": 43290 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001612179487179487, + "loss": 0.0474, + "step": 43300 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016118589743589743, + "loss": 0.0397, + "step": 43310 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016115384615384613, + "loss": 0.046, + "step": 43320 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016112179487179484, + "loss": 0.0611, + "step": 43330 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016108974358974357, + "loss": 0.05, + "step": 43340 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016105769230769228, + "loss": 0.0462, + "step": 43350 + }, + { + "epoch": 0.93, + "learning_rate": 0.000161025641025641, + "loss": 0.0494, + "step": 43360 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016099358974358972, + "loss": 0.0422, + "step": 43370 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016096153846153845, + "loss": 0.046, + "step": 43380 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016092948717948716, + "loss": 0.0488, + "step": 43390 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001608974358974359, + "loss": 0.0501, + "step": 43400 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001608653846153846, + "loss": 0.0521, + "step": 43410 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001608333333333333, + "loss": 0.0497, + "step": 43420 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016080128205128204, + "loss": 0.0482, + "step": 43430 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016076923076923074, + "loss": 0.0464, + "step": 43440 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016073717948717948, + "loss": 0.0473, + "step": 43450 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016070512820512818, + "loss": 0.0486, + "step": 43460 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016067307692307691, + "loss": 0.0517, + "step": 43470 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016064102564102562, + "loss": 0.0477, + "step": 43480 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016060897435897433, + "loss": 0.0436, + "step": 43490 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016057692307692306, + "loss": 0.0481, + "step": 43500 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016054487179487177, + "loss": 0.044, + "step": 43510 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001605128205128205, + "loss": 0.0423, + "step": 43520 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001604807692307692, + "loss": 0.0507, + "step": 43530 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016044871794871794, + "loss": 0.0503, + "step": 43540 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016041666666666664, + "loss": 0.0483, + "step": 43550 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016038461538461538, + "loss": 0.0447, + "step": 43560 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016035256410256408, + "loss": 0.0442, + "step": 43570 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001603205128205128, + "loss": 0.0549, + "step": 43580 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016028846153846152, + "loss": 0.0537, + "step": 43590 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016025641025641023, + "loss": 0.0543, + "step": 43600 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016022435897435896, + "loss": 0.0483, + "step": 43610 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016019230769230767, + "loss": 0.0446, + "step": 43620 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001601602564102564, + "loss": 0.047, + "step": 43630 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001601282051282051, + "loss": 0.0403, + "step": 43640 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016009615384615384, + "loss": 0.0417, + "step": 43650 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016006410256410255, + "loss": 0.0532, + "step": 43660 + }, + { + "epoch": 0.93, + "learning_rate": 0.00016003205128205125, + "loss": 0.0438, + "step": 43670 + }, + { + "epoch": 0.93, + "learning_rate": 0.00015999999999999999, + "loss": 0.0419, + "step": 43680 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001599679487179487, + "loss": 0.0533, + "step": 43690 + }, + { + "epoch": 0.93, + "learning_rate": 0.00015993589743589743, + "loss": 0.0452, + "step": 43700 + }, + { + "epoch": 0.93, + "learning_rate": 0.00015990384615384613, + "loss": 0.0593, + "step": 43710 + }, + { + "epoch": 0.93, + "learning_rate": 0.00015987179487179486, + "loss": 0.0431, + "step": 43720 + }, + { + "epoch": 0.93, + "learning_rate": 0.00015983974358974357, + "loss": 0.0443, + "step": 43730 + }, + { + "epoch": 0.93, + "learning_rate": 0.0001598076923076923, + "loss": 0.0419, + "step": 43740 + }, + { + "epoch": 0.93, + "learning_rate": 0.000159775641025641, + "loss": 0.0559, + "step": 43750 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015974358974358972, + "loss": 0.0509, + "step": 43760 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015971153846153845, + "loss": 0.0442, + "step": 43770 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015967948717948715, + "loss": 0.0423, + "step": 43780 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001596474358974359, + "loss": 0.0478, + "step": 43790 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001596153846153846, + "loss": 0.0406, + "step": 43800 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015958333333333333, + "loss": 0.0412, + "step": 43810 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015955128205128203, + "loss": 0.0482, + "step": 43820 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015951923076923077, + "loss": 0.0503, + "step": 43830 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015948717948717947, + "loss": 0.0472, + "step": 43840 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015945512820512818, + "loss": 0.0474, + "step": 43850 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001594230769230769, + "loss": 0.0438, + "step": 43860 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015939102564102562, + "loss": 0.0403, + "step": 43870 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015935897435897435, + "loss": 0.0538, + "step": 43880 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015932692307692306, + "loss": 0.0481, + "step": 43890 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001592948717948718, + "loss": 0.0445, + "step": 43900 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001592628205128205, + "loss": 0.0468, + "step": 43910 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015923076923076923, + "loss": 0.0472, + "step": 43920 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015919871794871794, + "loss": 0.0459, + "step": 43930 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015916666666666664, + "loss": 0.0529, + "step": 43940 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015913461538461537, + "loss": 0.0401, + "step": 43950 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015910256410256408, + "loss": 0.0384, + "step": 43960 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015907051282051281, + "loss": 0.0442, + "step": 43970 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015903846153846152, + "loss": 0.0447, + "step": 43980 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015900641025641025, + "loss": 0.0425, + "step": 43990 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015897435897435896, + "loss": 0.0525, + "step": 44000 + }, + { + "epoch": 0.94, + "eval_cer": 0.923747979790688, + "eval_loss": 0.03118530660867691, + "eval_runtime": 120.7999, + "eval_samples_per_second": 16.556, + "eval_steps_per_second": 4.139, + "step": 44000 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001589423076923077, + "loss": 0.0423, + "step": 44010 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001589102564102564, + "loss": 0.0483, + "step": 44020 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001588782051282051, + "loss": 0.053, + "step": 44030 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015884615384615384, + "loss": 0.0555, + "step": 44040 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015881410256410254, + "loss": 0.0482, + "step": 44050 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015878205128205128, + "loss": 0.0427, + "step": 44060 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015874999999999998, + "loss": 0.0485, + "step": 44070 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015871794871794872, + "loss": 0.0456, + "step": 44080 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015868589743589742, + "loss": 0.0483, + "step": 44090 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015865384615384616, + "loss": 0.0438, + "step": 44100 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015862179487179486, + "loss": 0.0493, + "step": 44110 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015858974358974357, + "loss": 0.0446, + "step": 44120 + }, + { + "epoch": 0.94, + "learning_rate": 0.0001585576923076923, + "loss": 0.0483, + "step": 44130 + }, + { + "epoch": 0.94, + "learning_rate": 0.000158525641025641, + "loss": 0.045, + "step": 44140 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015849358974358974, + "loss": 0.0486, + "step": 44150 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015846153846153845, + "loss": 0.0428, + "step": 44160 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015842948717948718, + "loss": 0.0567, + "step": 44170 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015839743589743588, + "loss": 0.041, + "step": 44180 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015836538461538462, + "loss": 0.0447, + "step": 44190 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015833333333333332, + "loss": 0.0453, + "step": 44200 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015830128205128203, + "loss": 0.0556, + "step": 44210 + }, + { + "epoch": 0.94, + "learning_rate": 0.00015826923076923076, + "loss": 0.0418, + "step": 44220 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015823717948717947, + "loss": 0.0431, + "step": 44230 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001582051282051282, + "loss": 0.0454, + "step": 44240 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001581730769230769, + "loss": 0.0465, + "step": 44250 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015814102564102564, + "loss": 0.0498, + "step": 44260 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015810897435897435, + "loss": 0.0483, + "step": 44270 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015807692307692308, + "loss": 0.0441, + "step": 44280 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001580448717948718, + "loss": 0.0481, + "step": 44290 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001580128205128205, + "loss": 0.0468, + "step": 44300 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015798076923076923, + "loss": 0.0538, + "step": 44310 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015794871794871793, + "loss": 0.0517, + "step": 44320 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015791666666666667, + "loss": 0.0485, + "step": 44330 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015788461538461537, + "loss": 0.0436, + "step": 44340 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001578525641025641, + "loss": 0.048, + "step": 44350 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001578205128205128, + "loss": 0.0382, + "step": 44360 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015778846153846154, + "loss": 0.0494, + "step": 44370 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015775641025641025, + "loss": 0.044, + "step": 44380 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015772435897435896, + "loss": 0.0453, + "step": 44390 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001576923076923077, + "loss": 0.0486, + "step": 44400 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001576602564102564, + "loss": 0.0447, + "step": 44410 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015762820512820513, + "loss": 0.0517, + "step": 44420 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015759615384615383, + "loss": 0.0464, + "step": 44430 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015756410256410257, + "loss": 0.0536, + "step": 44440 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015753205128205127, + "loss": 0.0521, + "step": 44450 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015749999999999998, + "loss": 0.047, + "step": 44460 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001574679487179487, + "loss": 0.051, + "step": 44470 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015743589743589742, + "loss": 0.0459, + "step": 44480 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015740384615384615, + "loss": 0.0537, + "step": 44490 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015737179487179486, + "loss": 0.0519, + "step": 44500 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001573397435897436, + "loss": 0.0408, + "step": 44510 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001573076923076923, + "loss": 0.052, + "step": 44520 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015727564102564103, + "loss": 0.0429, + "step": 44530 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015724358974358974, + "loss": 0.0476, + "step": 44540 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015721153846153844, + "loss": 0.045, + "step": 44550 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015717948717948718, + "loss": 0.0486, + "step": 44560 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015714743589743588, + "loss": 0.0483, + "step": 44570 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015711538461538461, + "loss": 0.0466, + "step": 44580 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015708333333333332, + "loss": 0.043, + "step": 44590 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015705128205128205, + "loss": 0.045, + "step": 44600 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015701923076923076, + "loss": 0.0492, + "step": 44610 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001569871794871795, + "loss": 0.0431, + "step": 44620 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001569551282051282, + "loss": 0.0406, + "step": 44630 + }, + { + "epoch": 0.95, + "learning_rate": 0.0001569230769230769, + "loss": 0.0406, + "step": 44640 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015689102564102564, + "loss": 0.0495, + "step": 44650 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015685897435897434, + "loss": 0.0385, + "step": 44660 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015682692307692308, + "loss": 0.0452, + "step": 44670 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015679487179487178, + "loss": 0.0416, + "step": 44680 + }, + { + "epoch": 0.95, + "learning_rate": 0.00015676282051282052, + "loss": 0.0421, + "step": 44690 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015673076923076922, + "loss": 0.0388, + "step": 44700 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015669871794871796, + "loss": 0.0483, + "step": 44710 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015666666666666666, + "loss": 0.0472, + "step": 44720 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015663461538461537, + "loss": 0.0455, + "step": 44730 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001566025641025641, + "loss": 0.0437, + "step": 44740 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001565705128205128, + "loss": 0.0433, + "step": 44750 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015653846153846154, + "loss": 0.0359, + "step": 44760 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015650641025641025, + "loss": 0.0483, + "step": 44770 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015647435897435898, + "loss": 0.0414, + "step": 44780 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015644230769230769, + "loss": 0.0455, + "step": 44790 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015641025641025642, + "loss": 0.0469, + "step": 44800 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015637820512820512, + "loss": 0.0467, + "step": 44810 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015634615384615383, + "loss": 0.0473, + "step": 44820 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015631410256410256, + "loss": 0.0408, + "step": 44830 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015628205128205127, + "loss": 0.0442, + "step": 44840 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015625, + "loss": 0.0624, + "step": 44850 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001562179487179487, + "loss": 0.049, + "step": 44860 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015618589743589744, + "loss": 0.0451, + "step": 44870 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015615384615384615, + "loss": 0.0439, + "step": 44880 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015612179487179488, + "loss": 0.0457, + "step": 44890 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001560897435897436, + "loss": 0.0472, + "step": 44900 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001560576923076923, + "loss": 0.0433, + "step": 44910 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015602564102564103, + "loss": 0.05, + "step": 44920 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015599358974358973, + "loss": 0.0394, + "step": 44930 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015596153846153847, + "loss": 0.0485, + "step": 44940 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015592948717948717, + "loss": 0.0535, + "step": 44950 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001558974358974359, + "loss": 0.0445, + "step": 44960 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001558653846153846, + "loss": 0.0479, + "step": 44970 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015583333333333334, + "loss": 0.0453, + "step": 44980 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015580128205128205, + "loss": 0.05, + "step": 44990 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015576923076923076, + "loss": 0.042, + "step": 45000 + }, + { + "epoch": 0.96, + "eval_cer": 0.923617191673383, + "eval_loss": 0.031189823523163795, + "eval_runtime": 120.6483, + "eval_samples_per_second": 16.577, + "eval_steps_per_second": 4.144, + "step": 45000 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001557371794871795, + "loss": 0.0449, + "step": 45010 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001557051282051282, + "loss": 0.0509, + "step": 45020 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015567307692307693, + "loss": 0.0408, + "step": 45030 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015564102564102564, + "loss": 0.0517, + "step": 45040 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015560897435897437, + "loss": 0.0391, + "step": 45050 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015557692307692307, + "loss": 0.0461, + "step": 45060 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001555448717948718, + "loss": 0.0479, + "step": 45070 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001555128205128205, + "loss": 0.047, + "step": 45080 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015548076923076922, + "loss": 0.047, + "step": 45090 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015544871794871795, + "loss": 0.0484, + "step": 45100 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015541666666666666, + "loss": 0.0497, + "step": 45110 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001553846153846154, + "loss": 0.0554, + "step": 45120 + }, + { + "epoch": 0.96, + "learning_rate": 0.0001553525641025641, + "loss": 0.0403, + "step": 45130 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015532051282051283, + "loss": 0.0508, + "step": 45140 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015528846153846154, + "loss": 0.0526, + "step": 45150 + }, + { + "epoch": 0.96, + "learning_rate": 0.00015525641025641027, + "loss": 0.052, + "step": 45160 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015522435897435898, + "loss": 0.0417, + "step": 45170 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015519230769230768, + "loss": 0.0457, + "step": 45180 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015516025641025642, + "loss": 0.0438, + "step": 45190 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015512820512820512, + "loss": 0.0504, + "step": 45200 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015509615384615385, + "loss": 0.0434, + "step": 45210 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015506410256410256, + "loss": 0.0437, + "step": 45220 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001550320512820513, + "loss": 0.04, + "step": 45230 + }, + { + "epoch": 0.97, + "learning_rate": 0.000155, + "loss": 0.0371, + "step": 45240 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015496794871794873, + "loss": 0.0447, + "step": 45250 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015493589743589744, + "loss": 0.0478, + "step": 45260 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015490384615384615, + "loss": 0.0452, + "step": 45270 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015487179487179488, + "loss": 0.0547, + "step": 45280 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015483974358974358, + "loss": 0.0435, + "step": 45290 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015480769230769232, + "loss": 0.0502, + "step": 45300 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015477564102564102, + "loss": 0.0326, + "step": 45310 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015474358974358976, + "loss": 0.0496, + "step": 45320 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015471153846153846, + "loss": 0.0477, + "step": 45330 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015467948717948714, + "loss": 0.0507, + "step": 45340 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015464743589743587, + "loss": 0.047, + "step": 45350 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015461538461538458, + "loss": 0.0402, + "step": 45360 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015458333333333331, + "loss": 0.0419, + "step": 45370 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015455128205128202, + "loss": 0.0418, + "step": 45380 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015451923076923073, + "loss": 0.048, + "step": 45390 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015448717948717946, + "loss": 0.0435, + "step": 45400 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015445512820512817, + "loss": 0.0469, + "step": 45410 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001544230769230769, + "loss": 0.049, + "step": 45420 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001543910256410256, + "loss": 0.0393, + "step": 45430 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015435897435897434, + "loss": 0.041, + "step": 45440 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015432692307692304, + "loss": 0.0503, + "step": 45450 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015429487179487178, + "loss": 0.0469, + "step": 45460 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015426282051282048, + "loss": 0.0416, + "step": 45470 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001542307692307692, + "loss": 0.0482, + "step": 45480 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015419871794871792, + "loss": 0.0488, + "step": 45490 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015416666666666663, + "loss": 0.0511, + "step": 45500 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015413461538461536, + "loss": 0.0464, + "step": 45510 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015410256410256407, + "loss": 0.0459, + "step": 45520 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001540705128205128, + "loss": 0.0422, + "step": 45530 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001540384615384615, + "loss": 0.0406, + "step": 45540 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015400641025641024, + "loss": 0.0421, + "step": 45550 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015397435897435895, + "loss": 0.0391, + "step": 45560 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015394230769230765, + "loss": 0.0507, + "step": 45570 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015391025641025639, + "loss": 0.0438, + "step": 45580 + }, + { + "epoch": 0.97, + "learning_rate": 0.0001538782051282051, + "loss": 0.0459, + "step": 45590 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015384615384615382, + "loss": 0.0405, + "step": 45600 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015381410256410253, + "loss": 0.0361, + "step": 45610 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015378205128205126, + "loss": 0.0463, + "step": 45620 + }, + { + "epoch": 0.97, + "learning_rate": 0.00015374999999999997, + "loss": 0.0451, + "step": 45630 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001537179487179487, + "loss": 0.0424, + "step": 45640 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001536858974358974, + "loss": 0.0409, + "step": 45650 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015365384615384611, + "loss": 0.0403, + "step": 45660 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015362179487179485, + "loss": 0.0408, + "step": 45670 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015358974358974355, + "loss": 0.0496, + "step": 45680 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001535576923076923, + "loss": 0.042, + "step": 45690 + }, + { + "epoch": 0.98, + "learning_rate": 0.000153525641025641, + "loss": 0.0567, + "step": 45700 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015349358974358973, + "loss": 0.0475, + "step": 45710 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015346153846153843, + "loss": 0.0415, + "step": 45720 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015342948717948717, + "loss": 0.043, + "step": 45730 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015339743589743587, + "loss": 0.0426, + "step": 45740 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015336538461538458, + "loss": 0.0478, + "step": 45750 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001533333333333333, + "loss": 0.0522, + "step": 45760 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015330128205128202, + "loss": 0.0448, + "step": 45770 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015326923076923075, + "loss": 0.0403, + "step": 45780 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015323717948717946, + "loss": 0.0446, + "step": 45790 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001532051282051282, + "loss": 0.0408, + "step": 45800 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001531730769230769, + "loss": 0.0449, + "step": 45810 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015314102564102563, + "loss": 0.0454, + "step": 45820 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015310897435897433, + "loss": 0.0487, + "step": 45830 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015307692307692304, + "loss": 0.0426, + "step": 45840 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015304487179487177, + "loss": 0.0499, + "step": 45850 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015301282051282048, + "loss": 0.0454, + "step": 45860 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001529807692307692, + "loss": 0.0573, + "step": 45870 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015294871794871792, + "loss": 0.0406, + "step": 45880 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015291666666666665, + "loss": 0.0408, + "step": 45890 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015288461538461536, + "loss": 0.0515, + "step": 45900 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001528525641025641, + "loss": 0.0464, + "step": 45910 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001528205128205128, + "loss": 0.0434, + "step": 45920 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001527884615384615, + "loss": 0.0485, + "step": 45930 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015275641025641024, + "loss": 0.0468, + "step": 45940 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015272435897435894, + "loss": 0.0422, + "step": 45950 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015269230769230768, + "loss": 0.044, + "step": 45960 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015266025641025638, + "loss": 0.0522, + "step": 45970 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015262820512820512, + "loss": 0.0578, + "step": 45980 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015259615384615382, + "loss": 0.0462, + "step": 45990 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015256410256410255, + "loss": 0.0541, + "step": 46000 + }, + { + "epoch": 0.98, + "eval_cer": 0.9237275972269522, + "eval_loss": 0.03036876767873764, + "eval_runtime": 120.516, + "eval_samples_per_second": 16.595, + "eval_steps_per_second": 4.149, + "step": 46000 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015253205128205126, + "loss": 0.0493, + "step": 46010 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015249999999999997, + "loss": 0.0465, + "step": 46020 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001524679487179487, + "loss": 0.0461, + "step": 46030 + }, + { + "epoch": 0.98, + "learning_rate": 0.0001524358974358974, + "loss": 0.0432, + "step": 46040 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015240384615384614, + "loss": 0.0471, + "step": 46050 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015237179487179484, + "loss": 0.0463, + "step": 46060 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015233974358974358, + "loss": 0.0482, + "step": 46070 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015230769230769228, + "loss": 0.0495, + "step": 46080 + }, + { + "epoch": 0.98, + "learning_rate": 0.00015227564102564102, + "loss": 0.0504, + "step": 46090 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015224358974358972, + "loss": 0.037, + "step": 46100 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015221153846153843, + "loss": 0.0446, + "step": 46110 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015217948717948716, + "loss": 0.0465, + "step": 46120 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015214743589743587, + "loss": 0.0531, + "step": 46130 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001521153846153846, + "loss": 0.0449, + "step": 46140 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001520833333333333, + "loss": 0.0509, + "step": 46150 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015205128205128204, + "loss": 0.0402, + "step": 46160 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015201923076923075, + "loss": 0.048, + "step": 46170 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015198717948717948, + "loss": 0.0425, + "step": 46180 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015195512820512819, + "loss": 0.0438, + "step": 46190 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001519230769230769, + "loss": 0.0464, + "step": 46200 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015189102564102563, + "loss": 0.044, + "step": 46210 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015185897435897433, + "loss": 0.0468, + "step": 46220 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015182692307692306, + "loss": 0.0513, + "step": 46230 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015179487179487177, + "loss": 0.0364, + "step": 46240 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001517628205128205, + "loss": 0.0433, + "step": 46250 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001517307692307692, + "loss": 0.049, + "step": 46260 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015169871794871794, + "loss": 0.0461, + "step": 46270 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015166666666666665, + "loss": 0.044, + "step": 46280 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015163461538461535, + "loss": 0.0443, + "step": 46290 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001516025641025641, + "loss": 0.0444, + "step": 46300 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001515705128205128, + "loss": 0.041, + "step": 46310 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015153846153846153, + "loss": 0.0438, + "step": 46320 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015150641025641023, + "loss": 0.0441, + "step": 46330 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015147435897435897, + "loss": 0.0488, + "step": 46340 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015144230769230767, + "loss": 0.0408, + "step": 46350 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015141025641025638, + "loss": 0.0416, + "step": 46360 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001513782051282051, + "loss": 0.0428, + "step": 46370 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015134615384615382, + "loss": 0.0473, + "step": 46380 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015131410256410255, + "loss": 0.0431, + "step": 46390 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015128205128205126, + "loss": 0.0413, + "step": 46400 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015125, + "loss": 0.0421, + "step": 46410 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001512179487179487, + "loss": 0.0426, + "step": 46420 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015118589743589743, + "loss": 0.0427, + "step": 46430 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015115384615384614, + "loss": 0.0399, + "step": 46440 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015112179487179484, + "loss": 0.0398, + "step": 46450 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015108974358974357, + "loss": 0.048, + "step": 46460 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015105769230769228, + "loss": 0.0433, + "step": 46470 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015102564102564101, + "loss": 0.0465, + "step": 46480 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015099358974358972, + "loss": 0.0379, + "step": 46490 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015096153846153845, + "loss": 0.0484, + "step": 46500 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015092948717948716, + "loss": 0.0412, + "step": 46510 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001508974358974359, + "loss": 0.0543, + "step": 46520 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001508653846153846, + "loss": 0.0415, + "step": 46530 + }, + { + "epoch": 0.99, + "learning_rate": 0.0001508333333333333, + "loss": 0.0422, + "step": 46540 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015080128205128204, + "loss": 0.0449, + "step": 46550 + }, + { + "epoch": 0.99, + "learning_rate": 0.00015076923076923074, + "loss": 0.0465, + "step": 46560 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015073717948717948, + "loss": 0.0502, + "step": 46570 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015070512820512818, + "loss": 0.0433, + "step": 46580 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015067307692307692, + "loss": 0.0467, + "step": 46590 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015064102564102562, + "loss": 0.0563, + "step": 46600 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015060897435897436, + "loss": 0.0506, + "step": 46610 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015057692307692306, + "loss": 0.0424, + "step": 46620 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015054487179487177, + "loss": 0.0459, + "step": 46630 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001505128205128205, + "loss": 0.0416, + "step": 46640 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001504807692307692, + "loss": 0.0501, + "step": 46650 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015044871794871794, + "loss": 0.0385, + "step": 46660 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015041666666666665, + "loss": 0.0429, + "step": 46670 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015038461538461538, + "loss": 0.0426, + "step": 46680 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015035256410256408, + "loss": 0.0473, + "step": 46690 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015032051282051282, + "loss": 0.0411, + "step": 46700 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015028846153846152, + "loss": 0.0502, + "step": 46710 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015025641025641023, + "loss": 0.0459, + "step": 46720 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015022435897435896, + "loss": 0.0451, + "step": 46730 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015019230769230767, + "loss": 0.0458, + "step": 46740 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001501602564102564, + "loss": 0.0382, + "step": 46750 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001501282051282051, + "loss": 0.046, + "step": 46760 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015009615384615384, + "loss": 0.0461, + "step": 46770 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015006410256410255, + "loss": 0.0472, + "step": 46780 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015003205128205128, + "loss": 0.0394, + "step": 46790 + }, + { + "epoch": 1.0, + "learning_rate": 0.00015, + "loss": 0.0409, + "step": 46800 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001499679487179487, + "loss": 0.0407, + "step": 46810 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014993589743589743, + "loss": 0.0451, + "step": 46820 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014990384615384613, + "loss": 0.0447, + "step": 46830 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014987179487179487, + "loss": 0.0448, + "step": 46840 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014983974358974357, + "loss": 0.0449, + "step": 46850 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001498076923076923, + "loss": 0.0519, + "step": 46860 + }, + { + "epoch": 1.0, + "learning_rate": 0.000149775641025641, + "loss": 0.0382, + "step": 46870 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014974358974358974, + "loss": 0.0371, + "step": 46880 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014971153846153845, + "loss": 0.0363, + "step": 46890 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014967948717948716, + "loss": 0.0356, + "step": 46900 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001496474358974359, + "loss": 0.0404, + "step": 46910 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001496153846153846, + "loss": 0.0425, + "step": 46920 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014958333333333333, + "loss": 0.0386, + "step": 46930 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014955128205128203, + "loss": 0.0388, + "step": 46940 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014951923076923077, + "loss": 0.0422, + "step": 46950 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014948717948717947, + "loss": 0.0361, + "step": 46960 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001494551282051282, + "loss": 0.0423, + "step": 46970 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001494230769230769, + "loss": 0.0479, + "step": 46980 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014939102564102562, + "loss": 0.0394, + "step": 46990 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014935897435897435, + "loss": 0.0417, + "step": 47000 + }, + { + "epoch": 1.0, + "eval_cer": 0.9237955391060717, + "eval_loss": 0.0292538870126009, + "eval_runtime": 121.0123, + "eval_samples_per_second": 16.527, + "eval_steps_per_second": 4.132, + "step": 47000 + }, + { + "epoch": 1.0, + "learning_rate": 0.00014932692307692306, + "loss": 0.0387, + "step": 47010 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001492948717948718, + "loss": 0.0428, + "step": 47020 + }, + { + "epoch": 1.0, + "learning_rate": 0.0001492628205128205, + "loss": 0.0394, + "step": 47030 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014923076923076923, + "loss": 0.0374, + "step": 47040 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014919871794871794, + "loss": 0.0432, + "step": 47050 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014916666666666667, + "loss": 0.0347, + "step": 47060 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014913461538461538, + "loss": 0.042, + "step": 47070 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014910256410256408, + "loss": 0.0421, + "step": 47080 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014907051282051281, + "loss": 0.0391, + "step": 47090 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014903846153846152, + "loss": 0.042, + "step": 47100 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014900641025641025, + "loss": 0.0424, + "step": 47110 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014897435897435896, + "loss": 0.0458, + "step": 47120 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001489423076923077, + "loss": 0.0481, + "step": 47130 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001489102564102564, + "loss": 0.032, + "step": 47140 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014887820512820513, + "loss": 0.0422, + "step": 47150 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014884615384615384, + "loss": 0.0479, + "step": 47160 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014881410256410254, + "loss": 0.0429, + "step": 47170 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014878205128205128, + "loss": 0.0528, + "step": 47180 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014874999999999998, + "loss": 0.0416, + "step": 47190 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014871794871794872, + "loss": 0.041, + "step": 47200 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014868589743589742, + "loss": 0.0383, + "step": 47210 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014865384615384616, + "loss": 0.0483, + "step": 47220 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014862179487179486, + "loss": 0.0383, + "step": 47230 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001485897435897436, + "loss": 0.0405, + "step": 47240 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001485576923076923, + "loss": 0.0347, + "step": 47250 + }, + { + "epoch": 1.01, + "learning_rate": 0.000148525641025641, + "loss": 0.039, + "step": 47260 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014849358974358974, + "loss": 0.0359, + "step": 47270 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014846153846153845, + "loss": 0.0364, + "step": 47280 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014842948717948718, + "loss": 0.0422, + "step": 47290 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014839743589743589, + "loss": 0.0392, + "step": 47300 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014836538461538462, + "loss": 0.0417, + "step": 47310 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014833333333333332, + "loss": 0.046, + "step": 47320 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014830128205128203, + "loss": 0.0414, + "step": 47330 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014826923076923076, + "loss": 0.0407, + "step": 47340 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014823717948717947, + "loss": 0.0396, + "step": 47350 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001482051282051282, + "loss": 0.0426, + "step": 47360 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001481730769230769, + "loss": 0.048, + "step": 47370 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014814102564102564, + "loss": 0.0355, + "step": 47380 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014810897435897435, + "loss": 0.0356, + "step": 47390 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014807692307692308, + "loss": 0.0424, + "step": 47400 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001480448717948718, + "loss": 0.039, + "step": 47410 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001480128205128205, + "loss": 0.0423, + "step": 47420 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014798076923076923, + "loss": 0.0387, + "step": 47430 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014794871794871793, + "loss": 0.038, + "step": 47440 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014791666666666667, + "loss": 0.0415, + "step": 47450 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014788461538461537, + "loss": 0.0406, + "step": 47460 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001478525641025641, + "loss": 0.0384, + "step": 47470 + }, + { + "epoch": 1.01, + "learning_rate": 0.0001478205128205128, + "loss": 0.0506, + "step": 47480 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014778846153846154, + "loss": 0.0469, + "step": 47490 + }, + { + "epoch": 1.01, + "learning_rate": 0.00014775641025641025, + "loss": 0.0473, + "step": 47500 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014772435897435896, + "loss": 0.0434, + "step": 47510 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001476923076923077, + "loss": 0.0484, + "step": 47520 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001476602564102564, + "loss": 0.0403, + "step": 47530 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001476282051282051, + "loss": 0.0391, + "step": 47540 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014759615384615384, + "loss": 0.0394, + "step": 47550 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014756410256410254, + "loss": 0.0382, + "step": 47560 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014753205128205127, + "loss": 0.0429, + "step": 47570 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014749999999999998, + "loss": 0.0402, + "step": 47580 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014746794871794869, + "loss": 0.0445, + "step": 47590 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014743589743589742, + "loss": 0.0504, + "step": 47600 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014740384615384613, + "loss": 0.0462, + "step": 47610 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014737179487179486, + "loss": 0.0428, + "step": 47620 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014733974358974356, + "loss": 0.0391, + "step": 47630 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001473076923076923, + "loss": 0.0444, + "step": 47640 + }, + { + "epoch": 1.02, + "learning_rate": 0.000147275641025641, + "loss": 0.0379, + "step": 47650 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014724358974358974, + "loss": 0.041, + "step": 47660 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014721153846153844, + "loss": 0.0387, + "step": 47670 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014717948717948715, + "loss": 0.0364, + "step": 47680 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014714743589743588, + "loss": 0.0488, + "step": 47690 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001471153846153846, + "loss": 0.0384, + "step": 47700 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014708333333333332, + "loss": 0.0365, + "step": 47710 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014705128205128203, + "loss": 0.0614, + "step": 47720 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014701923076923076, + "loss": 0.0455, + "step": 47730 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014698717948717947, + "loss": 0.0389, + "step": 47740 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014695512820512817, + "loss": 0.0445, + "step": 47750 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001469230769230769, + "loss": 0.0367, + "step": 47760 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001468910256410256, + "loss": 0.0344, + "step": 47770 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014685897435897435, + "loss": 0.0367, + "step": 47780 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014682692307692305, + "loss": 0.0404, + "step": 47790 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014679487179487178, + "loss": 0.034, + "step": 47800 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001467628205128205, + "loss": 0.0358, + "step": 47810 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014673076923076922, + "loss": 0.04, + "step": 47820 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014669871794871793, + "loss": 0.0431, + "step": 47830 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014666666666666664, + "loss": 0.0439, + "step": 47840 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014663461538461537, + "loss": 0.0424, + "step": 47850 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014660256410256407, + "loss": 0.0428, + "step": 47860 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001465705128205128, + "loss": 0.0431, + "step": 47870 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014653846153846151, + "loss": 0.0383, + "step": 47880 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014650641025641025, + "loss": 0.0404, + "step": 47890 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014647435897435895, + "loss": 0.0375, + "step": 47900 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001464423076923077, + "loss": 0.0379, + "step": 47910 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001464102564102564, + "loss": 0.0464, + "step": 47920 + }, + { + "epoch": 1.02, + "learning_rate": 0.0001463782051282051, + "loss": 0.0356, + "step": 47930 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014634615384615383, + "loss": 0.0429, + "step": 47940 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014631410256410254, + "loss": 0.0392, + "step": 47950 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014628205128205127, + "loss": 0.04, + "step": 47960 + }, + { + "epoch": 1.02, + "learning_rate": 0.00014624999999999998, + "loss": 0.0401, + "step": 47970 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001462179487179487, + "loss": 0.0418, + "step": 47980 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014618589743589742, + "loss": 0.0456, + "step": 47990 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014615384615384615, + "loss": 0.0369, + "step": 48000 + }, + { + "epoch": 1.03, + "eval_cer": 0.9237089132101943, + "eval_loss": 0.03051287680864334, + "eval_runtime": 120.7492, + "eval_samples_per_second": 16.563, + "eval_steps_per_second": 4.141, + "step": 48000 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014612179487179486, + "loss": 0.0383, + "step": 48010 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014608974358974356, + "loss": 0.041, + "step": 48020 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001460576923076923, + "loss": 0.0375, + "step": 48030 + }, + { + "epoch": 1.03, + "learning_rate": 0.000146025641025641, + "loss": 0.041, + "step": 48040 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014599358974358973, + "loss": 0.0376, + "step": 48050 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014596153846153844, + "loss": 0.037, + "step": 48060 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014592948717948717, + "loss": 0.0407, + "step": 48070 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014589743589743588, + "loss": 0.0394, + "step": 48080 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001458653846153846, + "loss": 0.0466, + "step": 48090 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014583333333333332, + "loss": 0.0468, + "step": 48100 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014580128205128202, + "loss": 0.0448, + "step": 48110 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014576923076923076, + "loss": 0.0375, + "step": 48120 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014573717948717946, + "loss": 0.0394, + "step": 48130 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001457051282051282, + "loss": 0.0383, + "step": 48140 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001456730769230769, + "loss": 0.0399, + "step": 48150 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014564102564102564, + "loss": 0.0353, + "step": 48160 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014560897435897434, + "loss": 0.0352, + "step": 48170 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014557692307692308, + "loss": 0.0474, + "step": 48180 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014554487179487178, + "loss": 0.0386, + "step": 48190 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001455128205128205, + "loss": 0.0497, + "step": 48200 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014548076923076922, + "loss": 0.0374, + "step": 48210 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014544871794871793, + "loss": 0.0399, + "step": 48220 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014541666666666666, + "loss": 0.0432, + "step": 48230 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014538461538461537, + "loss": 0.0417, + "step": 48240 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001453525641025641, + "loss": 0.0354, + "step": 48250 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001453205128205128, + "loss": 0.04, + "step": 48260 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014528846153846154, + "loss": 0.0454, + "step": 48270 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014525641025641024, + "loss": 0.0461, + "step": 48280 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014522435897435895, + "loss": 0.0404, + "step": 48290 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014519230769230768, + "loss": 0.0404, + "step": 48300 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001451602564102564, + "loss": 0.05, + "step": 48310 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014512820512820512, + "loss": 0.0349, + "step": 48320 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014509615384615383, + "loss": 0.0378, + "step": 48330 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014506410256410256, + "loss": 0.0368, + "step": 48340 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014503205128205127, + "loss": 0.0383, + "step": 48350 + }, + { + "epoch": 1.03, + "learning_rate": 0.000145, + "loss": 0.0385, + "step": 48360 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001449679487179487, + "loss": 0.0389, + "step": 48370 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001449358974358974, + "loss": 0.0361, + "step": 48380 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014490384615384615, + "loss": 0.0452, + "step": 48390 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014487179487179485, + "loss": 0.0413, + "step": 48400 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014483974358974359, + "loss": 0.0443, + "step": 48410 + }, + { + "epoch": 1.03, + "learning_rate": 0.0001448076923076923, + "loss": 0.0423, + "step": 48420 + }, + { + "epoch": 1.03, + "learning_rate": 0.00014477564102564102, + "loss": 0.0397, + "step": 48430 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014474358974358973, + "loss": 0.0358, + "step": 48440 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014471153846153846, + "loss": 0.0365, + "step": 48450 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014467948717948717, + "loss": 0.0425, + "step": 48460 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014464743589743588, + "loss": 0.0363, + "step": 48470 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001446153846153846, + "loss": 0.0341, + "step": 48480 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014458333333333332, + "loss": 0.0399, + "step": 48490 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014455128205128205, + "loss": 0.0394, + "step": 48500 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014451923076923075, + "loss": 0.0371, + "step": 48510 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001444871794871795, + "loss": 0.0445, + "step": 48520 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001444551282051282, + "loss": 0.0466, + "step": 48530 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014442307692307693, + "loss": 0.0491, + "step": 48540 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014439102564102563, + "loss": 0.0394, + "step": 48550 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014435897435897434, + "loss": 0.0409, + "step": 48560 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014432692307692307, + "loss": 0.0369, + "step": 48570 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014429487179487178, + "loss": 0.0415, + "step": 48580 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001442628205128205, + "loss": 0.0414, + "step": 48590 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014423076923076922, + "loss": 0.0374, + "step": 48600 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014419871794871795, + "loss": 0.0465, + "step": 48610 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014416666666666666, + "loss": 0.0393, + "step": 48620 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001441346153846154, + "loss": 0.0366, + "step": 48630 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001441025641025641, + "loss": 0.0416, + "step": 48640 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001440705128205128, + "loss": 0.0367, + "step": 48650 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014403846153846153, + "loss": 0.0367, + "step": 48660 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014400641025641024, + "loss": 0.0394, + "step": 48670 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014397435897435897, + "loss": 0.0424, + "step": 48680 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014394230769230768, + "loss": 0.0402, + "step": 48690 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001439102564102564, + "loss": 0.0332, + "step": 48700 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014387820512820512, + "loss": 0.0378, + "step": 48710 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014384615384615383, + "loss": 0.0355, + "step": 48720 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014381410256410256, + "loss": 0.0409, + "step": 48730 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014378205128205126, + "loss": 0.0442, + "step": 48740 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014375, + "loss": 0.0417, + "step": 48750 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001437179487179487, + "loss": 0.0389, + "step": 48760 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014368589743589744, + "loss": 0.0392, + "step": 48770 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014365384615384614, + "loss": 0.0374, + "step": 48780 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014362179487179488, + "loss": 0.0372, + "step": 48790 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014358974358974358, + "loss": 0.0355, + "step": 48800 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001435576923076923, + "loss": 0.0347, + "step": 48810 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014352564102564102, + "loss": 0.047, + "step": 48820 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014349358974358973, + "loss": 0.0348, + "step": 48830 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014346153846153846, + "loss": 0.0384, + "step": 48840 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014342948717948717, + "loss": 0.0437, + "step": 48850 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001433974358974359, + "loss": 0.0361, + "step": 48860 + }, + { + "epoch": 1.04, + "learning_rate": 0.0001433653846153846, + "loss": 0.0395, + "step": 48870 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014333333333333334, + "loss": 0.0342, + "step": 48880 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014330128205128204, + "loss": 0.0422, + "step": 48890 + }, + { + "epoch": 1.04, + "learning_rate": 0.00014326923076923075, + "loss": 0.0387, + "step": 48900 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014323717948717948, + "loss": 0.0425, + "step": 48910 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001432051282051282, + "loss": 0.0367, + "step": 48920 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014317307692307692, + "loss": 0.0415, + "step": 48930 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014314102564102563, + "loss": 0.045, + "step": 48940 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014310897435897436, + "loss": 0.04, + "step": 48950 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014307692307692307, + "loss": 0.0382, + "step": 48960 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001430448717948718, + "loss": 0.0389, + "step": 48970 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001430128205128205, + "loss": 0.0428, + "step": 48980 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014298076923076921, + "loss": 0.0377, + "step": 48990 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014294871794871792, + "loss": 0.0357, + "step": 49000 + }, + { + "epoch": 1.05, + "eval_cer": 0.9236732437236566, + "eval_loss": 0.029686592519283295, + "eval_runtime": 120.5312, + "eval_samples_per_second": 16.593, + "eval_steps_per_second": 4.148, + "step": 49000 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014291666666666665, + "loss": 0.0404, + "step": 49010 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014288461538461536, + "loss": 0.0364, + "step": 49020 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001428525641025641, + "loss": 0.0422, + "step": 49030 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001428205128205128, + "loss": 0.0377, + "step": 49040 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014278846153846153, + "loss": 0.0385, + "step": 49050 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014275641025641024, + "loss": 0.036, + "step": 49060 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014272435897435894, + "loss": 0.0371, + "step": 49070 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014269230769230768, + "loss": 0.04, + "step": 49080 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014266025641025638, + "loss": 0.0326, + "step": 49090 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014262820512820512, + "loss": 0.0401, + "step": 49100 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014259615384615382, + "loss": 0.0417, + "step": 49110 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014256410256410256, + "loss": 0.0413, + "step": 49120 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014253205128205126, + "loss": 0.0338, + "step": 49130 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001425, + "loss": 0.0411, + "step": 49140 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001424679487179487, + "loss": 0.0401, + "step": 49150 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001424358974358974, + "loss": 0.0398, + "step": 49160 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014240384615384614, + "loss": 0.0349, + "step": 49170 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014237179487179485, + "loss": 0.0428, + "step": 49180 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014233974358974358, + "loss": 0.0431, + "step": 49190 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014230769230769228, + "loss": 0.0428, + "step": 49200 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014227564102564102, + "loss": 0.0357, + "step": 49210 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014224358974358972, + "loss": 0.0367, + "step": 49220 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014221153846153843, + "loss": 0.0393, + "step": 49230 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014217948717948716, + "loss": 0.0365, + "step": 49240 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014214743589743587, + "loss": 0.0407, + "step": 49250 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001421153846153846, + "loss": 0.0342, + "step": 49260 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001420833333333333, + "loss": 0.0488, + "step": 49270 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014205128205128204, + "loss": 0.0371, + "step": 49280 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014201923076923075, + "loss": 0.038, + "step": 49290 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014198717948717948, + "loss": 0.0373, + "step": 49300 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001419551282051282, + "loss": 0.0419, + "step": 49310 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001419230769230769, + "loss": 0.0488, + "step": 49320 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014189102564102563, + "loss": 0.0407, + "step": 49330 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014185897435897433, + "loss": 0.0396, + "step": 49340 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014182692307692307, + "loss": 0.0414, + "step": 49350 + }, + { + "epoch": 1.05, + "learning_rate": 0.00014179487179487177, + "loss": 0.0461, + "step": 49360 + }, + { + "epoch": 1.05, + "learning_rate": 0.0001417628205128205, + "loss": 0.0432, + "step": 49370 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001417307692307692, + "loss": 0.0375, + "step": 49380 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014169871794871794, + "loss": 0.0366, + "step": 49390 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014166666666666665, + "loss": 0.0328, + "step": 49400 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014163461538461536, + "loss": 0.0409, + "step": 49410 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001416025641025641, + "loss": 0.0377, + "step": 49420 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001415705128205128, + "loss": 0.0377, + "step": 49430 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014153846153846153, + "loss": 0.0411, + "step": 49440 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014150641025641023, + "loss": 0.0382, + "step": 49450 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014147435897435897, + "loss": 0.0377, + "step": 49460 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014144230769230767, + "loss": 0.0389, + "step": 49470 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001414102564102564, + "loss": 0.0426, + "step": 49480 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001413782051282051, + "loss": 0.0403, + "step": 49490 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014134615384615382, + "loss": 0.0389, + "step": 49500 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014131410256410255, + "loss": 0.0415, + "step": 49510 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014128205128205126, + "loss": 0.0348, + "step": 49520 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014125, + "loss": 0.0353, + "step": 49530 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001412179487179487, + "loss": 0.0366, + "step": 49540 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014118589743589743, + "loss": 0.0321, + "step": 49550 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014115384615384614, + "loss": 0.0371, + "step": 49560 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014112179487179487, + "loss": 0.0463, + "step": 49570 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014108974358974358, + "loss": 0.0397, + "step": 49580 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014105769230769228, + "loss": 0.0367, + "step": 49590 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014102564102564101, + "loss": 0.0481, + "step": 49600 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014099358974358972, + "loss": 0.0453, + "step": 49610 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014096153846153845, + "loss": 0.0406, + "step": 49620 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014092948717948716, + "loss": 0.0417, + "step": 49630 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001408974358974359, + "loss": 0.0359, + "step": 49640 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001408653846153846, + "loss": 0.0381, + "step": 49650 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014083333333333333, + "loss": 0.0309, + "step": 49660 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014080128205128204, + "loss": 0.0353, + "step": 49670 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014076923076923074, + "loss": 0.0396, + "step": 49680 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014073717948717948, + "loss": 0.0378, + "step": 49690 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014070512820512818, + "loss": 0.038, + "step": 49700 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014067307692307692, + "loss": 0.0386, + "step": 49710 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014064102564102562, + "loss": 0.0447, + "step": 49720 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014060897435897436, + "loss": 0.035, + "step": 49730 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014057692307692306, + "loss": 0.0361, + "step": 49740 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001405448717948718, + "loss": 0.0375, + "step": 49750 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001405128205128205, + "loss": 0.0359, + "step": 49760 + }, + { + "epoch": 1.06, + "learning_rate": 0.0001404807692307692, + "loss": 0.0362, + "step": 49770 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014044871794871794, + "loss": 0.037, + "step": 49780 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014041666666666665, + "loss": 0.0349, + "step": 49790 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014038461538461538, + "loss": 0.0349, + "step": 49800 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014035256410256409, + "loss": 0.0375, + "step": 49810 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014032051282051282, + "loss": 0.0486, + "step": 49820 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014028846153846152, + "loss": 0.0383, + "step": 49830 + }, + { + "epoch": 1.06, + "learning_rate": 0.00014025641025641026, + "loss": 0.0386, + "step": 49840 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014022435897435896, + "loss": 0.0416, + "step": 49850 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014019230769230767, + "loss": 0.0395, + "step": 49860 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001401602564102564, + "loss": 0.0345, + "step": 49870 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001401282051282051, + "loss": 0.0397, + "step": 49880 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014009615384615384, + "loss": 0.0356, + "step": 49890 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014006410256410255, + "loss": 0.0351, + "step": 49900 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014003205128205128, + "loss": 0.037, + "step": 49910 + }, + { + "epoch": 1.07, + "learning_rate": 0.00014, + "loss": 0.0351, + "step": 49920 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013996794871794872, + "loss": 0.035, + "step": 49930 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013993589743589743, + "loss": 0.0492, + "step": 49940 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013990384615384613, + "loss": 0.0435, + "step": 49950 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013987179487179487, + "loss": 0.034, + "step": 49960 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013983974358974357, + "loss": 0.0329, + "step": 49970 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001398076923076923, + "loss": 0.041, + "step": 49980 + }, + { + "epoch": 1.07, + "learning_rate": 0.000139775641025641, + "loss": 0.0422, + "step": 49990 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013974358974358974, + "loss": 0.0394, + "step": 50000 + }, + { + "epoch": 1.07, + "eval_cer": 0.9236783393645905, + "eval_loss": 0.02962096408009529, + "eval_runtime": 120.8994, + "eval_samples_per_second": 16.543, + "eval_steps_per_second": 4.136, + "step": 50000 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013971153846153845, + "loss": 0.0398, + "step": 50010 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013967948717948718, + "loss": 0.039, + "step": 50020 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001396474358974359, + "loss": 0.0423, + "step": 50030 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001396153846153846, + "loss": 0.0423, + "step": 50040 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013958333333333333, + "loss": 0.0421, + "step": 50050 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013955128205128204, + "loss": 0.0472, + "step": 50060 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013951923076923077, + "loss": 0.0349, + "step": 50070 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013948717948717947, + "loss": 0.0462, + "step": 50080 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001394551282051282, + "loss": 0.0415, + "step": 50090 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001394230769230769, + "loss": 0.0376, + "step": 50100 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013939102564102565, + "loss": 0.0393, + "step": 50110 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013935897435897435, + "loss": 0.0372, + "step": 50120 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013932692307692306, + "loss": 0.0344, + "step": 50130 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001392948717948718, + "loss": 0.0359, + "step": 50140 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001392628205128205, + "loss": 0.0378, + "step": 50150 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013923076923076923, + "loss": 0.032, + "step": 50160 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013919871794871794, + "loss": 0.0356, + "step": 50170 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013916666666666667, + "loss": 0.0344, + "step": 50180 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013913461538461538, + "loss": 0.0514, + "step": 50190 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013910256410256408, + "loss": 0.0379, + "step": 50200 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013907051282051282, + "loss": 0.042, + "step": 50210 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013903846153846152, + "loss": 0.0392, + "step": 50220 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013900641025641025, + "loss": 0.0377, + "step": 50230 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013897435897435896, + "loss": 0.0473, + "step": 50240 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001389423076923077, + "loss": 0.0382, + "step": 50250 + }, + { + "epoch": 1.07, + "learning_rate": 0.0001389102564102564, + "loss": 0.0409, + "step": 50260 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013887820512820513, + "loss": 0.0381, + "step": 50270 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013884615384615384, + "loss": 0.0403, + "step": 50280 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013881410256410255, + "loss": 0.0429, + "step": 50290 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013878205128205128, + "loss": 0.0343, + "step": 50300 + }, + { + "epoch": 1.07, + "learning_rate": 0.00013874999999999998, + "loss": 0.0407, + "step": 50310 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013871794871794872, + "loss": 0.0404, + "step": 50320 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013868589743589742, + "loss": 0.0435, + "step": 50330 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013865384615384616, + "loss": 0.0384, + "step": 50340 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013862179487179486, + "loss": 0.0417, + "step": 50350 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001385897435897436, + "loss": 0.0443, + "step": 50360 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001385576923076923, + "loss": 0.0366, + "step": 50370 + }, + { + "epoch": 1.08, + "learning_rate": 0.000138525641025641, + "loss": 0.0379, + "step": 50380 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013849358974358974, + "loss": 0.0558, + "step": 50390 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013846153846153845, + "loss": 0.0445, + "step": 50400 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013842948717948718, + "loss": 0.0426, + "step": 50410 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001383974358974359, + "loss": 0.0444, + "step": 50420 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013836538461538462, + "loss": 0.0367, + "step": 50430 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013833333333333333, + "loss": 0.0316, + "step": 50440 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013830128205128206, + "loss": 0.0388, + "step": 50450 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013826923076923074, + "loss": 0.039, + "step": 50460 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013823717948717947, + "loss": 0.0365, + "step": 50470 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013820512820512818, + "loss": 0.0395, + "step": 50480 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001381730769230769, + "loss": 0.0381, + "step": 50490 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013814102564102562, + "loss": 0.0414, + "step": 50500 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013810897435897435, + "loss": 0.0364, + "step": 50510 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013807692307692306, + "loss": 0.0362, + "step": 50520 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001380448717948718, + "loss": 0.0412, + "step": 50530 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001380128205128205, + "loss": 0.0362, + "step": 50540 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001379807692307692, + "loss": 0.0381, + "step": 50550 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013794871794871793, + "loss": 0.0379, + "step": 50560 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013791666666666664, + "loss": 0.0428, + "step": 50570 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013788461538461537, + "loss": 0.0396, + "step": 50580 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013785256410256408, + "loss": 0.0432, + "step": 50590 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001378205128205128, + "loss": 0.0399, + "step": 50600 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013778846153846152, + "loss": 0.0359, + "step": 50610 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013775641025641022, + "loss": 0.0419, + "step": 50620 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013772435897435896, + "loss": 0.0379, + "step": 50630 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013769230769230766, + "loss": 0.0377, + "step": 50640 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001376602564102564, + "loss": 0.0368, + "step": 50650 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001376282051282051, + "loss": 0.0423, + "step": 50660 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013759615384615384, + "loss": 0.046, + "step": 50670 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013756410256410254, + "loss": 0.0371, + "step": 50680 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013753205128205128, + "loss": 0.0385, + "step": 50690 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013749999999999998, + "loss": 0.0313, + "step": 50700 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001374679487179487, + "loss": 0.0319, + "step": 50710 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013743589743589742, + "loss": 0.0354, + "step": 50720 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013740384615384613, + "loss": 0.0413, + "step": 50730 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013737179487179486, + "loss": 0.0395, + "step": 50740 + }, + { + "epoch": 1.08, + "learning_rate": 0.00013733974358974357, + "loss": 0.0379, + "step": 50750 + }, + { + "epoch": 1.08, + "learning_rate": 0.0001373076923076923, + "loss": 0.0399, + "step": 50760 + }, + { + "epoch": 1.08, + "learning_rate": 0.000137275641025641, + "loss": 0.038, + "step": 50770 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013724358974358974, + "loss": 0.0414, + "step": 50780 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013721153846153844, + "loss": 0.0385, + "step": 50790 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013717948717948715, + "loss": 0.0392, + "step": 50800 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013714743589743588, + "loss": 0.0476, + "step": 50810 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001371153846153846, + "loss": 0.0415, + "step": 50820 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013708333333333332, + "loss": 0.0365, + "step": 50830 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013705128205128203, + "loss": 0.0327, + "step": 50840 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013701923076923076, + "loss": 0.0433, + "step": 50850 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013698717948717947, + "loss": 0.0321, + "step": 50860 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001369551282051282, + "loss": 0.0351, + "step": 50870 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001369230769230769, + "loss": 0.0384, + "step": 50880 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001368910256410256, + "loss": 0.0393, + "step": 50890 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013685897435897435, + "loss": 0.0355, + "step": 50900 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013682692307692305, + "loss": 0.0517, + "step": 50910 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013679487179487179, + "loss": 0.037, + "step": 50920 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001367628205128205, + "loss": 0.0392, + "step": 50930 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013673076923076922, + "loss": 0.0359, + "step": 50940 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013669871794871793, + "loss": 0.0385, + "step": 50950 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013666666666666666, + "loss": 0.0362, + "step": 50960 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013663461538461537, + "loss": 0.0373, + "step": 50970 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013660256410256408, + "loss": 0.0367, + "step": 50980 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001365705128205128, + "loss": 0.0344, + "step": 50990 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013653846153846152, + "loss": 0.0343, + "step": 51000 + }, + { + "epoch": 1.09, + "eval_cer": 0.9236868320994805, + "eval_loss": 0.02877923659980297, + "eval_runtime": 120.6412, + "eval_samples_per_second": 16.578, + "eval_steps_per_second": 4.145, + "step": 51000 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013650641025641025, + "loss": 0.0316, + "step": 51010 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013647435897435895, + "loss": 0.0494, + "step": 51020 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001364423076923077, + "loss": 0.0427, + "step": 51030 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001364102564102564, + "loss": 0.0346, + "step": 51040 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013637820512820513, + "loss": 0.0385, + "step": 51050 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013634615384615383, + "loss": 0.0418, + "step": 51060 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013631410256410254, + "loss": 0.0451, + "step": 51070 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013628205128205127, + "loss": 0.0378, + "step": 51080 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013624999999999998, + "loss": 0.0377, + "step": 51090 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001362179487179487, + "loss": 0.0377, + "step": 51100 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013618589743589742, + "loss": 0.0356, + "step": 51110 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013615384615384615, + "loss": 0.0452, + "step": 51120 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013612179487179486, + "loss": 0.0396, + "step": 51130 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001360897435897436, + "loss": 0.0431, + "step": 51140 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001360576923076923, + "loss": 0.0429, + "step": 51150 + }, + { + "epoch": 1.09, + "learning_rate": 0.000136025641025641, + "loss": 0.0363, + "step": 51160 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013599358974358973, + "loss": 0.0438, + "step": 51170 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013596153846153844, + "loss": 0.0423, + "step": 51180 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013592948717948717, + "loss": 0.0479, + "step": 51190 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013589743589743588, + "loss": 0.0381, + "step": 51200 + }, + { + "epoch": 1.09, + "learning_rate": 0.0001358653846153846, + "loss": 0.0341, + "step": 51210 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013583333333333332, + "loss": 0.0397, + "step": 51220 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013580128205128205, + "loss": 0.0447, + "step": 51230 + }, + { + "epoch": 1.09, + "learning_rate": 0.00013576923076923076, + "loss": 0.044, + "step": 51240 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013573717948717946, + "loss": 0.0367, + "step": 51250 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001357051282051282, + "loss": 0.0402, + "step": 51260 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001356730769230769, + "loss": 0.0356, + "step": 51270 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013564102564102564, + "loss": 0.0388, + "step": 51280 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013560897435897434, + "loss": 0.0359, + "step": 51290 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013557692307692308, + "loss": 0.0356, + "step": 51300 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013554487179487178, + "loss": 0.0354, + "step": 51310 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013551282051282052, + "loss": 0.0336, + "step": 51320 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013548076923076922, + "loss": 0.0365, + "step": 51330 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013544871794871793, + "loss": 0.0393, + "step": 51340 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013541666666666666, + "loss": 0.0385, + "step": 51350 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013538461538461537, + "loss": 0.0471, + "step": 51360 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001353525641025641, + "loss": 0.0385, + "step": 51370 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001353205128205128, + "loss": 0.0372, + "step": 51380 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013528846153846154, + "loss": 0.0323, + "step": 51390 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013525641025641024, + "loss": 0.0309, + "step": 51400 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013522435897435898, + "loss": 0.0415, + "step": 51410 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013519230769230768, + "loss": 0.0355, + "step": 51420 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001351602564102564, + "loss": 0.03, + "step": 51430 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013512820512820512, + "loss": 0.0361, + "step": 51440 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013509615384615383, + "loss": 0.0325, + "step": 51450 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013506410256410256, + "loss": 0.0448, + "step": 51460 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013503205128205127, + "loss": 0.0398, + "step": 51470 + }, + { + "epoch": 1.1, + "learning_rate": 0.000135, + "loss": 0.0382, + "step": 51480 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001349679487179487, + "loss": 0.0375, + "step": 51490 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013493589743589744, + "loss": 0.0409, + "step": 51500 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013490384615384615, + "loss": 0.0378, + "step": 51510 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013487179487179485, + "loss": 0.0356, + "step": 51520 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013483974358974359, + "loss": 0.0354, + "step": 51530 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001348076923076923, + "loss": 0.0374, + "step": 51540 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013477564102564103, + "loss": 0.0406, + "step": 51550 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013474358974358973, + "loss": 0.0342, + "step": 51560 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013471153846153846, + "loss": 0.0315, + "step": 51570 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013467948717948717, + "loss": 0.0423, + "step": 51580 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013464743589743588, + "loss": 0.0418, + "step": 51590 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001346153846153846, + "loss": 0.038, + "step": 51600 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013458333333333332, + "loss": 0.0358, + "step": 51610 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013455128205128205, + "loss": 0.0405, + "step": 51620 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013451923076923076, + "loss": 0.039, + "step": 51630 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001344871794871795, + "loss": 0.0339, + "step": 51640 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001344551282051282, + "loss": 0.0353, + "step": 51650 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013442307692307693, + "loss": 0.0358, + "step": 51660 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013439102564102563, + "loss": 0.0345, + "step": 51670 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013435897435897434, + "loss": 0.0391, + "step": 51680 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013432692307692307, + "loss": 0.0365, + "step": 51690 + }, + { + "epoch": 1.1, + "learning_rate": 0.00013429487179487178, + "loss": 0.0408, + "step": 51700 + }, + { + "epoch": 1.1, + "learning_rate": 0.0001342628205128205, + "loss": 0.0382, + "step": 51710 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013423076923076922, + "loss": 0.0386, + "step": 51720 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013419871794871795, + "loss": 0.0317, + "step": 51730 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013416666666666666, + "loss": 0.0326, + "step": 51740 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001341346153846154, + "loss": 0.0398, + "step": 51750 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001341025641025641, + "loss": 0.0353, + "step": 51760 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001340705128205128, + "loss": 0.0431, + "step": 51770 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013403846153846154, + "loss": 0.0382, + "step": 51780 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013400641025641024, + "loss": 0.0404, + "step": 51790 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013397435897435897, + "loss": 0.0365, + "step": 51800 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013394230769230768, + "loss": 0.0493, + "step": 51810 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013391025641025641, + "loss": 0.0378, + "step": 51820 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013387820512820512, + "loss": 0.0429, + "step": 51830 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013384615384615385, + "loss": 0.0375, + "step": 51840 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013381410256410256, + "loss": 0.0393, + "step": 51850 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013378205128205127, + "loss": 0.0343, + "step": 51860 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013375, + "loss": 0.0362, + "step": 51870 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001337179487179487, + "loss": 0.0384, + "step": 51880 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013368589743589744, + "loss": 0.0397, + "step": 51890 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013365384615384614, + "loss": 0.0322, + "step": 51900 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013362179487179488, + "loss": 0.0375, + "step": 51910 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013358974358974358, + "loss": 0.0392, + "step": 51920 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001335576923076923, + "loss": 0.0322, + "step": 51930 + }, + { + "epoch": 1.11, + "learning_rate": 0.000133525641025641, + "loss": 0.0422, + "step": 51940 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013349358974358973, + "loss": 0.0356, + "step": 51950 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013346153846153843, + "loss": 0.0375, + "step": 51960 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013342948717948717, + "loss": 0.0417, + "step": 51970 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013339743589743587, + "loss": 0.0293, + "step": 51980 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001333653846153846, + "loss": 0.0436, + "step": 51990 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001333333333333333, + "loss": 0.037, + "step": 52000 + }, + { + "epoch": 1.11, + "eval_cer": 0.9237089132101943, + "eval_loss": 0.028605736792087555, + "eval_runtime": 120.8443, + "eval_samples_per_second": 16.55, + "eval_steps_per_second": 4.138, + "step": 52000 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013330128205128205, + "loss": 0.04, + "step": 52010 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013326923076923075, + "loss": 0.0391, + "step": 52020 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013323717948717946, + "loss": 0.0495, + "step": 52030 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001332051282051282, + "loss": 0.0397, + "step": 52040 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001331730769230769, + "loss": 0.0352, + "step": 52050 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013314102564102563, + "loss": 0.0404, + "step": 52060 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013310897435897434, + "loss": 0.0417, + "step": 52070 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013307692307692307, + "loss": 0.0436, + "step": 52080 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013304487179487178, + "loss": 0.0387, + "step": 52090 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013301282051282048, + "loss": 0.0437, + "step": 52100 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013298076923076921, + "loss": 0.0496, + "step": 52110 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013294871794871792, + "loss": 0.0393, + "step": 52120 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013291666666666665, + "loss": 0.0358, + "step": 52130 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013288461538461536, + "loss": 0.0383, + "step": 52140 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001328525641025641, + "loss": 0.0346, + "step": 52150 + }, + { + "epoch": 1.11, + "learning_rate": 0.0001328205128205128, + "loss": 0.044, + "step": 52160 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013278846153846153, + "loss": 0.0361, + "step": 52170 + }, + { + "epoch": 1.11, + "learning_rate": 0.00013275641025641024, + "loss": 0.0403, + "step": 52180 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013272435897435894, + "loss": 0.0388, + "step": 52190 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013269230769230768, + "loss": 0.037, + "step": 52200 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013266025641025638, + "loss": 0.036, + "step": 52210 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013262820512820512, + "loss": 0.0425, + "step": 52220 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013259615384615382, + "loss": 0.0399, + "step": 52230 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013256410256410256, + "loss": 0.0411, + "step": 52240 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013253205128205126, + "loss": 0.037, + "step": 52250 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001325, + "loss": 0.0465, + "step": 52260 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001324679487179487, + "loss": 0.039, + "step": 52270 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001324358974358974, + "loss": 0.0491, + "step": 52280 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013240384615384614, + "loss": 0.0341, + "step": 52290 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013237179487179485, + "loss": 0.0379, + "step": 52300 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013233974358974358, + "loss": 0.04, + "step": 52310 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013230769230769229, + "loss": 0.0384, + "step": 52320 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013227564102564102, + "loss": 0.042, + "step": 52330 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013224358974358972, + "loss": 0.0394, + "step": 52340 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013221153846153846, + "loss": 0.0425, + "step": 52350 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013217948717948716, + "loss": 0.0416, + "step": 52360 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013214743589743587, + "loss": 0.0392, + "step": 52370 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001321153846153846, + "loss": 0.035, + "step": 52380 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001320833333333333, + "loss": 0.0382, + "step": 52390 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013205128205128204, + "loss": 0.0371, + "step": 52400 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013201923076923075, + "loss": 0.041, + "step": 52410 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013198717948717948, + "loss": 0.0389, + "step": 52420 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001319551282051282, + "loss": 0.0392, + "step": 52430 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013192307692307692, + "loss": 0.0421, + "step": 52440 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013189102564102563, + "loss": 0.0347, + "step": 52450 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013185897435897433, + "loss": 0.0364, + "step": 52460 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013182692307692307, + "loss": 0.0312, + "step": 52470 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013179487179487177, + "loss": 0.036, + "step": 52480 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001317628205128205, + "loss": 0.0464, + "step": 52490 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001317307692307692, + "loss": 0.0466, + "step": 52500 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013169871794871794, + "loss": 0.0369, + "step": 52510 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013166666666666665, + "loss": 0.0421, + "step": 52520 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013163461538461538, + "loss": 0.0433, + "step": 52530 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001316025641025641, + "loss": 0.0369, + "step": 52540 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001315705128205128, + "loss": 0.0411, + "step": 52550 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013153846153846153, + "loss": 0.0354, + "step": 52560 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013150641025641024, + "loss": 0.0425, + "step": 52570 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013147435897435897, + "loss": 0.0402, + "step": 52580 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013144230769230767, + "loss": 0.037, + "step": 52590 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001314102564102564, + "loss": 0.0374, + "step": 52600 + }, + { + "epoch": 1.12, + "learning_rate": 0.0001313782051282051, + "loss": 0.0342, + "step": 52610 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013134615384615385, + "loss": 0.0368, + "step": 52620 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013131410256410255, + "loss": 0.0414, + "step": 52630 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013128205128205126, + "loss": 0.0397, + "step": 52640 + }, + { + "epoch": 1.12, + "learning_rate": 0.00013125, + "loss": 0.0403, + "step": 52650 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001312179487179487, + "loss": 0.0378, + "step": 52660 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013118589743589743, + "loss": 0.0419, + "step": 52670 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013115384615384614, + "loss": 0.0332, + "step": 52680 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013112179487179487, + "loss": 0.0338, + "step": 52690 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013108974358974358, + "loss": 0.0394, + "step": 52700 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001310576923076923, + "loss": 0.0313, + "step": 52710 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013102564102564102, + "loss": 0.0402, + "step": 52720 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013099358974358972, + "loss": 0.0426, + "step": 52730 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013096153846153845, + "loss": 0.0351, + "step": 52740 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013092948717948716, + "loss": 0.0416, + "step": 52750 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001308974358974359, + "loss": 0.0361, + "step": 52760 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001308653846153846, + "loss": 0.0321, + "step": 52770 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013083333333333333, + "loss": 0.0424, + "step": 52780 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013080128205128204, + "loss": 0.0371, + "step": 52790 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013076923076923077, + "loss": 0.0401, + "step": 52800 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013073717948717948, + "loss": 0.0459, + "step": 52810 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013070512820512818, + "loss": 0.0365, + "step": 52820 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013067307692307692, + "loss": 0.0421, + "step": 52830 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013064102564102562, + "loss": 0.037, + "step": 52840 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013060897435897436, + "loss": 0.0374, + "step": 52850 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013057692307692306, + "loss": 0.0349, + "step": 52860 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001305448717948718, + "loss": 0.0333, + "step": 52870 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001305128205128205, + "loss": 0.0349, + "step": 52880 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013048076923076924, + "loss": 0.0345, + "step": 52890 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013044871794871794, + "loss": 0.0366, + "step": 52900 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013041666666666665, + "loss": 0.0317, + "step": 52910 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013038461538461538, + "loss": 0.0349, + "step": 52920 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001303525641025641, + "loss": 0.0418, + "step": 52930 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013032051282051282, + "loss": 0.0383, + "step": 52940 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013028846153846153, + "loss": 0.0357, + "step": 52950 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013025641025641026, + "loss": 0.0438, + "step": 52960 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013022435897435896, + "loss": 0.0356, + "step": 52970 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001301923076923077, + "loss": 0.0383, + "step": 52980 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001301602564102564, + "loss": 0.0341, + "step": 52990 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001301282051282051, + "loss": 0.0367, + "step": 53000 + }, + { + "epoch": 1.13, + "eval_cer": 0.9236579568008547, + "eval_loss": 0.02814340963959694, + "eval_runtime": 120.4403, + "eval_samples_per_second": 16.606, + "eval_steps_per_second": 4.151, + "step": 53000 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013009615384615384, + "loss": 0.0338, + "step": 53010 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013006410256410255, + "loss": 0.0353, + "step": 53020 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013003205128205128, + "loss": 0.0424, + "step": 53030 + }, + { + "epoch": 1.13, + "learning_rate": 0.00013, + "loss": 0.0336, + "step": 53040 + }, + { + "epoch": 1.13, + "learning_rate": 0.00012996794871794872, + "loss": 0.0331, + "step": 53050 + }, + { + "epoch": 1.13, + "learning_rate": 0.00012993589743589743, + "loss": 0.0342, + "step": 53060 + }, + { + "epoch": 1.13, + "learning_rate": 0.00012990384615384613, + "loss": 0.0384, + "step": 53070 + }, + { + "epoch": 1.13, + "learning_rate": 0.00012987179487179487, + "loss": 0.0376, + "step": 53080 + }, + { + "epoch": 1.13, + "learning_rate": 0.00012983974358974357, + "loss": 0.0409, + "step": 53090 + }, + { + "epoch": 1.13, + "learning_rate": 0.0001298076923076923, + "loss": 0.0426, + "step": 53100 + }, + { + "epoch": 1.13, + "learning_rate": 0.000129775641025641, + "loss": 0.0327, + "step": 53110 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012974358974358975, + "loss": 0.0414, + "step": 53120 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012971153846153845, + "loss": 0.0304, + "step": 53130 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012967948717948718, + "loss": 0.0389, + "step": 53140 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001296474358974359, + "loss": 0.0371, + "step": 53150 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001296153846153846, + "loss": 0.0328, + "step": 53160 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012958333333333333, + "loss": 0.0313, + "step": 53170 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012955128205128204, + "loss": 0.0355, + "step": 53180 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012951923076923077, + "loss": 0.0318, + "step": 53190 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012948717948717948, + "loss": 0.0313, + "step": 53200 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001294551282051282, + "loss": 0.0377, + "step": 53210 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012942307692307691, + "loss": 0.0383, + "step": 53220 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012939102564102565, + "loss": 0.0382, + "step": 53230 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012935897435897435, + "loss": 0.0351, + "step": 53240 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012932692307692306, + "loss": 0.041, + "step": 53250 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001292948717948718, + "loss": 0.0372, + "step": 53260 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001292628205128205, + "loss": 0.0388, + "step": 53270 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012923076923076923, + "loss": 0.0436, + "step": 53280 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012919871794871794, + "loss": 0.0419, + "step": 53290 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012916666666666667, + "loss": 0.0383, + "step": 53300 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012913461538461538, + "loss": 0.0367, + "step": 53310 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001291025641025641, + "loss": 0.0359, + "step": 53320 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012907051282051282, + "loss": 0.0314, + "step": 53330 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012903846153846152, + "loss": 0.031, + "step": 53340 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012900641025641026, + "loss": 0.0424, + "step": 53350 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012897435897435896, + "loss": 0.0368, + "step": 53360 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001289423076923077, + "loss": 0.0377, + "step": 53370 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001289102564102564, + "loss": 0.0376, + "step": 53380 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001288782051282051, + "loss": 0.0419, + "step": 53390 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012884615384615384, + "loss": 0.0385, + "step": 53400 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012881410256410255, + "loss": 0.038, + "step": 53410 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012878205128205125, + "loss": 0.0406, + "step": 53420 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012874999999999999, + "loss": 0.0416, + "step": 53430 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001287179487179487, + "loss": 0.0335, + "step": 53440 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012868589743589742, + "loss": 0.0375, + "step": 53450 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012865384615384613, + "loss": 0.033, + "step": 53460 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012862179487179486, + "loss": 0.039, + "step": 53470 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012858974358974357, + "loss": 0.0396, + "step": 53480 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012855769230769228, + "loss": 0.0427, + "step": 53490 + }, + { + "epoch": 1.14, + "learning_rate": 0.000128525641025641, + "loss": 0.0324, + "step": 53500 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012849358974358972, + "loss": 0.0361, + "step": 53510 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012846153846153845, + "loss": 0.0369, + "step": 53520 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012842948717948715, + "loss": 0.0428, + "step": 53530 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001283974358974359, + "loss": 0.0361, + "step": 53540 + }, + { + "epoch": 1.14, + "learning_rate": 0.0001283653846153846, + "loss": 0.0441, + "step": 53550 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012833333333333333, + "loss": 0.0346, + "step": 53560 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012830128205128203, + "loss": 0.0383, + "step": 53570 + }, + { + "epoch": 1.14, + "learning_rate": 0.00012826923076923074, + "loss": 0.0317, + "step": 53580 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012823717948717947, + "loss": 0.0384, + "step": 53590 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012820512820512818, + "loss": 0.0351, + "step": 53600 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001281730769230769, + "loss": 0.0348, + "step": 53610 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012814102564102562, + "loss": 0.0303, + "step": 53620 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012810897435897435, + "loss": 0.0344, + "step": 53630 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012807692307692306, + "loss": 0.0417, + "step": 53640 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001280448717948718, + "loss": 0.0303, + "step": 53650 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001280128205128205, + "loss": 0.0374, + "step": 53660 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001279807692307692, + "loss": 0.0369, + "step": 53670 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012794871794871793, + "loss": 0.0399, + "step": 53680 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012791666666666664, + "loss": 0.0471, + "step": 53690 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012788461538461537, + "loss": 0.0381, + "step": 53700 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012785256410256408, + "loss": 0.0414, + "step": 53710 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001278205128205128, + "loss": 0.0362, + "step": 53720 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012778846153846152, + "loss": 0.0353, + "step": 53730 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012775641025641025, + "loss": 0.0347, + "step": 53740 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012772435897435896, + "loss": 0.0414, + "step": 53750 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012769230769230766, + "loss": 0.046, + "step": 53760 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001276602564102564, + "loss": 0.0368, + "step": 53770 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001276282051282051, + "loss": 0.038, + "step": 53780 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012759615384615384, + "loss": 0.0391, + "step": 53790 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012756410256410254, + "loss": 0.0333, + "step": 53800 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012753205128205128, + "loss": 0.0334, + "step": 53810 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012749999999999998, + "loss": 0.0351, + "step": 53820 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012746794871794872, + "loss": 0.0377, + "step": 53830 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012743589743589742, + "loss": 0.0402, + "step": 53840 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012740384615384613, + "loss": 0.0377, + "step": 53850 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012737179487179486, + "loss": 0.0443, + "step": 53860 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012733974358974357, + "loss": 0.0367, + "step": 53870 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001273076923076923, + "loss": 0.0389, + "step": 53880 + }, + { + "epoch": 1.15, + "learning_rate": 0.000127275641025641, + "loss": 0.0374, + "step": 53890 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012724358974358974, + "loss": 0.0428, + "step": 53900 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012721153846153844, + "loss": 0.0326, + "step": 53910 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012717948717948718, + "loss": 0.0406, + "step": 53920 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012714743589743588, + "loss": 0.032, + "step": 53930 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001271153846153846, + "loss": 0.0359, + "step": 53940 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012708333333333332, + "loss": 0.0403, + "step": 53950 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012705128205128203, + "loss": 0.0334, + "step": 53960 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012701923076923076, + "loss": 0.0379, + "step": 53970 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012698717948717947, + "loss": 0.0348, + "step": 53980 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001269551282051282, + "loss": 0.039, + "step": 53990 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001269230769230769, + "loss": 0.0336, + "step": 54000 + }, + { + "epoch": 1.15, + "eval_cer": 0.9236120960324491, + "eval_loss": 0.02865663170814514, + "eval_runtime": 120.4744, + "eval_samples_per_second": 16.601, + "eval_steps_per_second": 4.15, + "step": 54000 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012689102564102564, + "loss": 0.0372, + "step": 54010 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012685897435897435, + "loss": 0.0377, + "step": 54020 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012682692307692305, + "loss": 0.0371, + "step": 54030 + }, + { + "epoch": 1.15, + "learning_rate": 0.00012679487179487179, + "loss": 0.0396, + "step": 54040 + }, + { + "epoch": 1.15, + "learning_rate": 0.0001267628205128205, + "loss": 0.0403, + "step": 54050 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012673076923076923, + "loss": 0.0387, + "step": 54060 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012669871794871793, + "loss": 0.0351, + "step": 54070 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012666666666666666, + "loss": 0.0346, + "step": 54080 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012663461538461537, + "loss": 0.0442, + "step": 54090 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001266025641025641, + "loss": 0.0373, + "step": 54100 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001265705128205128, + "loss": 0.036, + "step": 54110 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012653846153846152, + "loss": 0.0355, + "step": 54120 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012650641025641025, + "loss": 0.0339, + "step": 54130 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012647435897435896, + "loss": 0.0349, + "step": 54140 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001264423076923077, + "loss": 0.0392, + "step": 54150 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001264102564102564, + "loss": 0.0435, + "step": 54160 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012637820512820513, + "loss": 0.0344, + "step": 54170 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012634615384615383, + "loss": 0.0381, + "step": 54180 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012631410256410257, + "loss": 0.0409, + "step": 54190 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012628205128205127, + "loss": 0.0366, + "step": 54200 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012624999999999998, + "loss": 0.0358, + "step": 54210 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001262179487179487, + "loss": 0.0516, + "step": 54220 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012618589743589742, + "loss": 0.0411, + "step": 54230 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012615384615384615, + "loss": 0.0404, + "step": 54240 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012612179487179486, + "loss": 0.0393, + "step": 54250 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001260897435897436, + "loss": 0.0394, + "step": 54260 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001260576923076923, + "loss": 0.0338, + "step": 54270 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012602564102564103, + "loss": 0.0371, + "step": 54280 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012599358974358974, + "loss": 0.0346, + "step": 54290 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012596153846153844, + "loss": 0.0368, + "step": 54300 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012592948717948717, + "loss": 0.0395, + "step": 54310 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012589743589743588, + "loss": 0.0446, + "step": 54320 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012586538461538461, + "loss": 0.0338, + "step": 54330 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012583333333333332, + "loss": 0.036, + "step": 54340 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012580128205128205, + "loss": 0.0367, + "step": 54350 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012576923076923076, + "loss": 0.0357, + "step": 54360 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001257371794871795, + "loss": 0.0366, + "step": 54370 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001257051282051282, + "loss": 0.0349, + "step": 54380 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001256730769230769, + "loss": 0.0348, + "step": 54390 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012564102564102564, + "loss": 0.0403, + "step": 54400 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012560897435897434, + "loss": 0.0414, + "step": 54410 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012557692307692308, + "loss": 0.0358, + "step": 54420 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012554487179487178, + "loss": 0.0346, + "step": 54430 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012551282051282052, + "loss": 0.0349, + "step": 54440 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012548076923076922, + "loss": 0.0364, + "step": 54450 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012544871794871793, + "loss": 0.04, + "step": 54460 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012541666666666666, + "loss": 0.0425, + "step": 54470 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012538461538461537, + "loss": 0.0401, + "step": 54480 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001253525641025641, + "loss": 0.0382, + "step": 54490 + }, + { + "epoch": 1.16, + "learning_rate": 0.0001253205128205128, + "loss": 0.037, + "step": 54500 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012528846153846154, + "loss": 0.0364, + "step": 54510 + }, + { + "epoch": 1.16, + "learning_rate": 0.00012525641025641025, + "loss": 0.0378, + "step": 54520 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012522435897435898, + "loss": 0.0383, + "step": 54530 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012519230769230768, + "loss": 0.0324, + "step": 54540 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001251602564102564, + "loss": 0.0396, + "step": 54550 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012512820512820512, + "loss": 0.0419, + "step": 54560 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012509615384615383, + "loss": 0.0336, + "step": 54570 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012506410256410256, + "loss": 0.0329, + "step": 54580 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012503205128205127, + "loss": 0.0392, + "step": 54590 + }, + { + "epoch": 1.17, + "learning_rate": 0.000125, + "loss": 0.0319, + "step": 54600 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001249679487179487, + "loss": 0.0395, + "step": 54610 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012493589743589744, + "loss": 0.0425, + "step": 54620 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012490384615384615, + "loss": 0.0351, + "step": 54630 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012487179487179485, + "loss": 0.0394, + "step": 54640 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001248397435897436, + "loss": 0.043, + "step": 54650 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001248076923076923, + "loss": 0.0405, + "step": 54660 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012477564102564103, + "loss": 0.0429, + "step": 54670 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012474358974358973, + "loss": 0.035, + "step": 54680 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012471153846153847, + "loss": 0.0352, + "step": 54690 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012467948717948717, + "loss": 0.035, + "step": 54700 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001246474358974359, + "loss": 0.0391, + "step": 54710 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001246153846153846, + "loss": 0.0353, + "step": 54720 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012458333333333332, + "loss": 0.0395, + "step": 54730 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012455128205128205, + "loss": 0.0436, + "step": 54740 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012451923076923076, + "loss": 0.0363, + "step": 54750 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001244871794871795, + "loss": 0.0357, + "step": 54760 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001244551282051282, + "loss": 0.0378, + "step": 54770 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012442307692307693, + "loss": 0.0335, + "step": 54780 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012439102564102563, + "loss": 0.0394, + "step": 54790 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012435897435897437, + "loss": 0.0337, + "step": 54800 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012432692307692307, + "loss": 0.0377, + "step": 54810 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012429487179487178, + "loss": 0.0401, + "step": 54820 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001242628205128205, + "loss": 0.039, + "step": 54830 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012423076923076922, + "loss": 0.035, + "step": 54840 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012419871794871792, + "loss": 0.034, + "step": 54850 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012416666666666666, + "loss": 0.0395, + "step": 54860 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012413461538461536, + "loss": 0.0366, + "step": 54870 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001241025641025641, + "loss": 0.0402, + "step": 54880 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001240705128205128, + "loss": 0.0359, + "step": 54890 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001240384615384615, + "loss": 0.0466, + "step": 54900 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012400641025641024, + "loss": 0.0453, + "step": 54910 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012397435897435895, + "loss": 0.0439, + "step": 54920 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012394230769230768, + "loss": 0.0456, + "step": 54930 + }, + { + "epoch": 1.17, + "learning_rate": 0.0001239102564102564, + "loss": 0.0353, + "step": 54940 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012387820512820512, + "loss": 0.037, + "step": 54950 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012384615384615383, + "loss": 0.0416, + "step": 54960 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012381410256410253, + "loss": 0.0339, + "step": 54970 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012378205128205127, + "loss": 0.0371, + "step": 54980 + }, + { + "epoch": 1.17, + "learning_rate": 0.00012374999999999997, + "loss": 0.036, + "step": 54990 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001237179487179487, + "loss": 0.0331, + "step": 55000 + }, + { + "epoch": 1.18, + "eval_cer": 0.9236851335525025, + "eval_loss": 0.027525708079338074, + "eval_runtime": 120.3492, + "eval_samples_per_second": 16.618, + "eval_steps_per_second": 4.155, + "step": 55000 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001236858974358974, + "loss": 0.0391, + "step": 55010 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012365384615384614, + "loss": 0.0364, + "step": 55020 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012362179487179485, + "loss": 0.0392, + "step": 55030 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012358974358974358, + "loss": 0.0346, + "step": 55040 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001235576923076923, + "loss": 0.0395, + "step": 55050 + }, + { + "epoch": 1.18, + "learning_rate": 0.000123525641025641, + "loss": 0.0384, + "step": 55060 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012349358974358973, + "loss": 0.0385, + "step": 55070 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012346153846153844, + "loss": 0.0366, + "step": 55080 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012342948717948717, + "loss": 0.0364, + "step": 55090 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012339743589743587, + "loss": 0.0352, + "step": 55100 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001233653846153846, + "loss": 0.0388, + "step": 55110 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001233333333333333, + "loss": 0.0358, + "step": 55120 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012330128205128205, + "loss": 0.0366, + "step": 55130 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012326923076923075, + "loss": 0.0372, + "step": 55140 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012323717948717946, + "loss": 0.039, + "step": 55150 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001232051282051282, + "loss": 0.0321, + "step": 55160 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001231730769230769, + "loss": 0.0391, + "step": 55170 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012314102564102563, + "loss": 0.0334, + "step": 55180 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012310897435897434, + "loss": 0.0381, + "step": 55190 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012307692307692307, + "loss": 0.0361, + "step": 55200 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012304487179487178, + "loss": 0.0363, + "step": 55210 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001230128205128205, + "loss": 0.0363, + "step": 55220 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012298076923076922, + "loss": 0.0395, + "step": 55230 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012294871794871792, + "loss": 0.0425, + "step": 55240 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012291666666666665, + "loss": 0.0335, + "step": 55250 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012288461538461536, + "loss": 0.0359, + "step": 55260 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001228525641025641, + "loss": 0.0441, + "step": 55270 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001228205128205128, + "loss": 0.0436, + "step": 55280 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012278846153846153, + "loss": 0.0379, + "step": 55290 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012275641025641024, + "loss": 0.033, + "step": 55300 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012272435897435897, + "loss": 0.0347, + "step": 55310 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012269230769230768, + "loss": 0.0391, + "step": 55320 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012266025641025638, + "loss": 0.0399, + "step": 55330 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012262820512820512, + "loss": 0.0333, + "step": 55340 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012259615384615382, + "loss": 0.0375, + "step": 55350 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012256410256410256, + "loss": 0.0339, + "step": 55360 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012253205128205126, + "loss": 0.0374, + "step": 55370 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001225, + "loss": 0.0323, + "step": 55380 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001224679487179487, + "loss": 0.0333, + "step": 55390 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012243589743589744, + "loss": 0.0442, + "step": 55400 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012240384615384614, + "loss": 0.0401, + "step": 55410 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012237179487179485, + "loss": 0.0372, + "step": 55420 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012233974358974358, + "loss": 0.0361, + "step": 55430 + }, + { + "epoch": 1.18, + "learning_rate": 0.0001223076923076923, + "loss": 0.0389, + "step": 55440 + }, + { + "epoch": 1.18, + "learning_rate": 0.00012227564102564102, + "loss": 0.0407, + "step": 55450 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012224358974358973, + "loss": 0.0398, + "step": 55460 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012221153846153846, + "loss": 0.0397, + "step": 55470 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012217948717948716, + "loss": 0.0404, + "step": 55480 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001221474358974359, + "loss": 0.0334, + "step": 55490 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001221153846153846, + "loss": 0.0461, + "step": 55500 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001220833333333333, + "loss": 0.0359, + "step": 55510 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012205128205128204, + "loss": 0.0372, + "step": 55520 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012201923076923076, + "loss": 0.0325, + "step": 55530 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012198717948717948, + "loss": 0.0347, + "step": 55540 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012195512820512819, + "loss": 0.0389, + "step": 55550 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012192307692307691, + "loss": 0.0303, + "step": 55560 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012189102564102563, + "loss": 0.0423, + "step": 55570 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012185897435897435, + "loss": 0.0375, + "step": 55580 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012182692307692307, + "loss": 0.0315, + "step": 55590 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012179487179487179, + "loss": 0.047, + "step": 55600 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001217628205128205, + "loss": 0.0436, + "step": 55610 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012173076923076923, + "loss": 0.0386, + "step": 55620 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012169871794871795, + "loss": 0.0407, + "step": 55630 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012166666666666665, + "loss": 0.0357, + "step": 55640 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012163461538461537, + "loss": 0.0367, + "step": 55650 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012160256410256409, + "loss": 0.0322, + "step": 55660 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012157051282051281, + "loss": 0.0416, + "step": 55670 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012153846153846153, + "loss": 0.0407, + "step": 55680 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012150641025641025, + "loss": 0.0347, + "step": 55690 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012147435897435897, + "loss": 0.0411, + "step": 55700 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012144230769230769, + "loss": 0.0343, + "step": 55710 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001214102564102564, + "loss": 0.0386, + "step": 55720 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012137820512820511, + "loss": 0.0369, + "step": 55730 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012134615384615383, + "loss": 0.0366, + "step": 55740 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012131410256410255, + "loss": 0.0423, + "step": 55750 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012128205128205127, + "loss": 0.0313, + "step": 55760 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012124999999999999, + "loss": 0.0326, + "step": 55770 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012121794871794871, + "loss": 0.0332, + "step": 55780 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012118589743589743, + "loss": 0.0338, + "step": 55790 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012115384615384615, + "loss": 0.0409, + "step": 55800 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012112179487179486, + "loss": 0.037, + "step": 55810 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012108974358974358, + "loss": 0.0346, + "step": 55820 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001210576923076923, + "loss": 0.039, + "step": 55830 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012102564102564102, + "loss": 0.035, + "step": 55840 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012099358974358974, + "loss": 0.0351, + "step": 55850 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012096153846153846, + "loss": 0.0355, + "step": 55860 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012092948717948718, + "loss": 0.0376, + "step": 55870 + }, + { + "epoch": 1.19, + "learning_rate": 0.0001208974358974359, + "loss": 0.039, + "step": 55880 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012086538461538461, + "loss": 0.0396, + "step": 55890 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012083333333333332, + "loss": 0.0366, + "step": 55900 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012080128205128204, + "loss": 0.0337, + "step": 55910 + }, + { + "epoch": 1.19, + "learning_rate": 0.00012076923076923076, + "loss": 0.0364, + "step": 55920 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012073717948717948, + "loss": 0.0383, + "step": 55930 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001207051282051282, + "loss": 0.0358, + "step": 55940 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012067307692307692, + "loss": 0.0348, + "step": 55950 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012064102564102564, + "loss": 0.0343, + "step": 55960 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012060897435897436, + "loss": 0.0342, + "step": 55970 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012057692307692308, + "loss": 0.039, + "step": 55980 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012054487179487178, + "loss": 0.039, + "step": 55990 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001205128205128205, + "loss": 0.0381, + "step": 56000 + }, + { + "epoch": 1.2, + "eval_cer": 0.9236554089803878, + "eval_loss": 0.027558309957385063, + "eval_runtime": 120.4611, + "eval_samples_per_second": 16.603, + "eval_steps_per_second": 4.151, + "step": 56000 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012048076923076922, + "loss": 0.0353, + "step": 56010 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012044871794871794, + "loss": 0.0306, + "step": 56020 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012041666666666666, + "loss": 0.0313, + "step": 56030 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012038461538461538, + "loss": 0.0296, + "step": 56040 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001203525641025641, + "loss": 0.0366, + "step": 56050 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012032051282051282, + "loss": 0.0542, + "step": 56060 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012028846153846154, + "loss": 0.0411, + "step": 56070 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012025641025641025, + "loss": 0.0363, + "step": 56080 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012022435897435897, + "loss": 0.0329, + "step": 56090 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012019230769230769, + "loss": 0.0325, + "step": 56100 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001201602564102564, + "loss": 0.0348, + "step": 56110 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012012820512820512, + "loss": 0.0347, + "step": 56120 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012009615384615384, + "loss": 0.0347, + "step": 56130 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012006410256410256, + "loss": 0.0383, + "step": 56140 + }, + { + "epoch": 1.2, + "learning_rate": 0.00012003205128205128, + "loss": 0.0362, + "step": 56150 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011999999999999999, + "loss": 0.0378, + "step": 56160 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011996794871794871, + "loss": 0.0387, + "step": 56170 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011993589743589743, + "loss": 0.0312, + "step": 56180 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011990384615384615, + "loss": 0.0587, + "step": 56190 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011987179487179487, + "loss": 0.0392, + "step": 56200 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011983974358974359, + "loss": 0.0371, + "step": 56210 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011980769230769231, + "loss": 0.0372, + "step": 56220 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011977564102564103, + "loss": 0.0358, + "step": 56230 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011974358974358975, + "loss": 0.0382, + "step": 56240 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011971153846153845, + "loss": 0.0365, + "step": 56250 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011967948717948717, + "loss": 0.0369, + "step": 56260 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011964743589743589, + "loss": 0.0298, + "step": 56270 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011961538461538461, + "loss": 0.0404, + "step": 56280 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011958333333333333, + "loss": 0.0346, + "step": 56290 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011955128205128205, + "loss": 0.0412, + "step": 56300 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011951923076923076, + "loss": 0.0405, + "step": 56310 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011948717948717946, + "loss": 0.0445, + "step": 56320 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011945512820512818, + "loss": 0.0386, + "step": 56330 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001194230769230769, + "loss": 0.0405, + "step": 56340 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011939102564102562, + "loss": 0.0378, + "step": 56350 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011935897435897434, + "loss": 0.04, + "step": 56360 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011932692307692306, + "loss": 0.0309, + "step": 56370 + }, + { + "epoch": 1.2, + "learning_rate": 0.00011929487179487178, + "loss": 0.0412, + "step": 56380 + }, + { + "epoch": 1.2, + "learning_rate": 0.0001192628205128205, + "loss": 0.0451, + "step": 56390 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011923076923076922, + "loss": 0.0371, + "step": 56400 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011919871794871793, + "loss": 0.0309, + "step": 56410 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011916666666666664, + "loss": 0.0358, + "step": 56420 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011913461538461536, + "loss": 0.0367, + "step": 56430 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011910256410256408, + "loss": 0.0356, + "step": 56440 + }, + { + "epoch": 1.21, + "learning_rate": 0.0001190705128205128, + "loss": 0.0397, + "step": 56450 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011903846153846152, + "loss": 0.0404, + "step": 56460 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011900641025641024, + "loss": 0.0381, + "step": 56470 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011897435897435896, + "loss": 0.0363, + "step": 56480 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011894230769230768, + "loss": 0.0395, + "step": 56490 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011891025641025639, + "loss": 0.0363, + "step": 56500 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011887820512820511, + "loss": 0.0407, + "step": 56510 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011884615384615383, + "loss": 0.0327, + "step": 56520 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011881410256410255, + "loss": 0.0277, + "step": 56530 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011878205128205127, + "loss": 0.0361, + "step": 56540 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011874999999999999, + "loss": 0.045, + "step": 56550 + }, + { + "epoch": 1.21, + "learning_rate": 0.0001187179487179487, + "loss": 0.0323, + "step": 56560 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011868589743589743, + "loss": 0.0352, + "step": 56570 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011865384615384615, + "loss": 0.0391, + "step": 56580 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011862179487179485, + "loss": 0.0367, + "step": 56590 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011858974358974357, + "loss": 0.0333, + "step": 56600 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011855769230769229, + "loss": 0.0363, + "step": 56610 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011852564102564101, + "loss": 0.0351, + "step": 56620 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011849358974358973, + "loss": 0.0378, + "step": 56630 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011846153846153845, + "loss": 0.0377, + "step": 56640 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011842948717948717, + "loss": 0.0374, + "step": 56650 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011839743589743589, + "loss": 0.0325, + "step": 56660 + }, + { + "epoch": 1.21, + "learning_rate": 0.0001183653846153846, + "loss": 0.0298, + "step": 56670 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011833333333333331, + "loss": 0.0374, + "step": 56680 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011830128205128203, + "loss": 0.0338, + "step": 56690 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011826923076923075, + "loss": 0.0386, + "step": 56700 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011823717948717947, + "loss": 0.036, + "step": 56710 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011820512820512819, + "loss": 0.0333, + "step": 56720 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011817307692307691, + "loss": 0.0334, + "step": 56730 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011814102564102563, + "loss": 0.031, + "step": 56740 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011810897435897435, + "loss": 0.0368, + "step": 56750 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011807692307692306, + "loss": 0.0343, + "step": 56760 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011804487179487178, + "loss": 0.0374, + "step": 56770 + }, + { + "epoch": 1.21, + "learning_rate": 0.0001180128205128205, + "loss": 0.0346, + "step": 56780 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011798076923076922, + "loss": 0.0361, + "step": 56790 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011794871794871794, + "loss": 0.0351, + "step": 56800 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011791666666666666, + "loss": 0.036, + "step": 56810 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011788461538461537, + "loss": 0.0388, + "step": 56820 + }, + { + "epoch": 1.21, + "learning_rate": 0.0001178525641025641, + "loss": 0.0436, + "step": 56830 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011782051282051281, + "loss": 0.0321, + "step": 56840 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011778846153846152, + "loss": 0.0351, + "step": 56850 + }, + { + "epoch": 1.21, + "learning_rate": 0.00011775641025641024, + "loss": 0.0429, + "step": 56860 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011772435897435896, + "loss": 0.0391, + "step": 56870 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011769230769230768, + "loss": 0.0348, + "step": 56880 + }, + { + "epoch": 1.22, + "learning_rate": 0.0001176602564102564, + "loss": 0.0333, + "step": 56890 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011762820512820512, + "loss": 0.0443, + "step": 56900 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011759615384615384, + "loss": 0.036, + "step": 56910 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011756410256410256, + "loss": 0.0323, + "step": 56920 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011753205128205128, + "loss": 0.0315, + "step": 56930 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011749999999999998, + "loss": 0.0403, + "step": 56940 + }, + { + "epoch": 1.22, + "learning_rate": 0.0001174679487179487, + "loss": 0.0345, + "step": 56950 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011743589743589742, + "loss": 0.0411, + "step": 56960 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011740384615384614, + "loss": 0.0355, + "step": 56970 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011737179487179486, + "loss": 0.0329, + "step": 56980 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011733974358974358, + "loss": 0.0337, + "step": 56990 + }, + { + "epoch": 1.22, + "learning_rate": 0.0001173076923076923, + "loss": 0.0438, + "step": 57000 + }, + { + "epoch": 1.22, + "eval_cer": 0.9236817364585466, + "eval_loss": 0.026895621791481972, + "eval_runtime": 120.5871, + "eval_samples_per_second": 16.586, + "eval_steps_per_second": 4.146, + "step": 57000 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011727564102564102, + "loss": 0.0326, + "step": 57010 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011724358974358974, + "loss": 0.0463, + "step": 57020 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011721153846153845, + "loss": 0.0326, + "step": 57030 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011717948717948717, + "loss": 0.0336, + "step": 57040 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011714743589743589, + "loss": 0.04, + "step": 57050 + }, + { + "epoch": 1.22, + "learning_rate": 0.0001171153846153846, + "loss": 0.0365, + "step": 57060 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011708333333333332, + "loss": 0.0358, + "step": 57070 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011705128205128204, + "loss": 0.0358, + "step": 57080 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011701923076923076, + "loss": 0.0381, + "step": 57090 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011698717948717948, + "loss": 0.0325, + "step": 57100 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011695512820512819, + "loss": 0.0348, + "step": 57110 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011692307692307691, + "loss": 0.0382, + "step": 57120 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011689102564102563, + "loss": 0.035, + "step": 57130 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011685897435897435, + "loss": 0.0389, + "step": 57140 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011682692307692307, + "loss": 0.0393, + "step": 57150 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011679487179487179, + "loss": 0.0371, + "step": 57160 + }, + { + "epoch": 1.22, + "learning_rate": 0.0001167628205128205, + "loss": 0.0352, + "step": 57170 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011673076923076923, + "loss": 0.0397, + "step": 57180 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011669871794871795, + "loss": 0.0372, + "step": 57190 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011666666666666665, + "loss": 0.0416, + "step": 57200 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011663461538461537, + "loss": 0.037, + "step": 57210 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011660256410256409, + "loss": 0.0379, + "step": 57220 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011657051282051281, + "loss": 0.0321, + "step": 57230 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011653846153846153, + "loss": 0.0388, + "step": 57240 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011650641025641025, + "loss": 0.0338, + "step": 57250 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011647435897435897, + "loss": 0.0362, + "step": 57260 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011644230769230769, + "loss": 0.0297, + "step": 57270 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011641025641025641, + "loss": 0.0383, + "step": 57280 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011637820512820511, + "loss": 0.0382, + "step": 57290 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011634615384615383, + "loss": 0.0328, + "step": 57300 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011631410256410255, + "loss": 0.0373, + "step": 57310 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011628205128205127, + "loss": 0.0387, + "step": 57320 + }, + { + "epoch": 1.22, + "learning_rate": 0.00011624999999999999, + "loss": 0.0304, + "step": 57330 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011621794871794871, + "loss": 0.0375, + "step": 57340 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011618589743589743, + "loss": 0.0299, + "step": 57350 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011615384615384615, + "loss": 0.0335, + "step": 57360 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011612179487179487, + "loss": 0.0345, + "step": 57370 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011608974358974358, + "loss": 0.034, + "step": 57380 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001160576923076923, + "loss": 0.0363, + "step": 57390 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011602564102564102, + "loss": 0.0326, + "step": 57400 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011599358974358974, + "loss": 0.0398, + "step": 57410 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011596153846153846, + "loss": 0.0331, + "step": 57420 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011592948717948718, + "loss": 0.0333, + "step": 57430 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001158974358974359, + "loss": 0.0369, + "step": 57440 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011586538461538461, + "loss": 0.0357, + "step": 57450 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011583333333333333, + "loss": 0.0312, + "step": 57460 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011580128205128204, + "loss": 0.0357, + "step": 57470 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011576923076923076, + "loss": 0.034, + "step": 57480 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011573717948717948, + "loss": 0.0335, + "step": 57490 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001157051282051282, + "loss": 0.0345, + "step": 57500 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011567307692307692, + "loss": 0.0305, + "step": 57510 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011564102564102564, + "loss": 0.0397, + "step": 57520 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011560897435897436, + "loss": 0.0412, + "step": 57530 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011557692307692308, + "loss": 0.0351, + "step": 57540 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001155448717948718, + "loss": 0.0357, + "step": 57550 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001155128205128205, + "loss": 0.037, + "step": 57560 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011548076923076922, + "loss": 0.0347, + "step": 57570 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011544871794871794, + "loss": 0.03, + "step": 57580 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011541666666666666, + "loss": 0.0421, + "step": 57590 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011538461538461538, + "loss": 0.0329, + "step": 57600 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001153525641025641, + "loss": 0.0412, + "step": 57610 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011532051282051282, + "loss": 0.0373, + "step": 57620 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011528846153846154, + "loss": 0.0364, + "step": 57630 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011525641025641025, + "loss": 0.0384, + "step": 57640 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011522435897435897, + "loss": 0.0357, + "step": 57650 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011519230769230769, + "loss": 0.0368, + "step": 57660 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001151602564102564, + "loss": 0.0321, + "step": 57670 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011512820512820513, + "loss": 0.0373, + "step": 57680 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011509615384615384, + "loss": 0.0318, + "step": 57690 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011506410256410256, + "loss": 0.0358, + "step": 57700 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011503205128205128, + "loss": 0.0374, + "step": 57710 + }, + { + "epoch": 1.23, + "learning_rate": 0.000115, + "loss": 0.0343, + "step": 57720 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011496794871794871, + "loss": 0.0407, + "step": 57730 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011493589743589743, + "loss": 0.0401, + "step": 57740 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011490384615384615, + "loss": 0.0362, + "step": 57750 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011487179487179487, + "loss": 0.0441, + "step": 57760 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011483974358974357, + "loss": 0.0333, + "step": 57770 + }, + { + "epoch": 1.23, + "learning_rate": 0.0001148076923076923, + "loss": 0.0377, + "step": 57780 + }, + { + "epoch": 1.23, + "learning_rate": 0.00011477564102564101, + "loss": 0.0413, + "step": 57790 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011474358974358972, + "loss": 0.0386, + "step": 57800 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011471153846153844, + "loss": 0.0408, + "step": 57810 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011467948717948716, + "loss": 0.0411, + "step": 57820 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011464743589743588, + "loss": 0.0343, + "step": 57830 + }, + { + "epoch": 1.24, + "learning_rate": 0.0001146153846153846, + "loss": 0.036, + "step": 57840 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011458333333333332, + "loss": 0.0318, + "step": 57850 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011455128205128204, + "loss": 0.0344, + "step": 57860 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011451923076923076, + "loss": 0.0351, + "step": 57870 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011448717948717948, + "loss": 0.0335, + "step": 57880 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011445512820512818, + "loss": 0.0433, + "step": 57890 + }, + { + "epoch": 1.24, + "learning_rate": 0.0001144230769230769, + "loss": 0.0344, + "step": 57900 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011439102564102562, + "loss": 0.0319, + "step": 57910 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011435897435897434, + "loss": 0.0388, + "step": 57920 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011432692307692306, + "loss": 0.0411, + "step": 57930 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011429487179487178, + "loss": 0.0375, + "step": 57940 + }, + { + "epoch": 1.24, + "learning_rate": 0.0001142628205128205, + "loss": 0.0327, + "step": 57950 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011423076923076922, + "loss": 0.0372, + "step": 57960 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011419871794871794, + "loss": 0.0387, + "step": 57970 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011416666666666665, + "loss": 0.0403, + "step": 57980 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011413461538461536, + "loss": 0.0395, + "step": 57990 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011410256410256408, + "loss": 0.0319, + "step": 58000 + }, + { + "epoch": 1.24, + "eval_cer": 0.9236358756901408, + "eval_loss": 0.027382049709558487, + "eval_runtime": 120.3758, + "eval_samples_per_second": 16.615, + "eval_steps_per_second": 4.154, + "step": 58000 + }, + { + "epoch": 1.24, + "learning_rate": 0.0001140705128205128, + "loss": 0.0423, + "step": 58010 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011403846153846152, + "loss": 0.0381, + "step": 58020 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011400641025641024, + "loss": 0.0358, + "step": 58030 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011397435897435896, + "loss": 0.0396, + "step": 58040 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011394230769230768, + "loss": 0.0319, + "step": 58050 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011391025641025639, + "loss": 0.0391, + "step": 58060 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011387820512820511, + "loss": 0.0338, + "step": 58070 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011384615384615383, + "loss": 0.0352, + "step": 58080 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011381410256410255, + "loss": 0.0409, + "step": 58090 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011378205128205127, + "loss": 0.0368, + "step": 58100 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011374999999999999, + "loss": 0.0325, + "step": 58110 + }, + { + "epoch": 1.24, + "learning_rate": 0.0001137179487179487, + "loss": 0.0342, + "step": 58120 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011368589743589743, + "loss": 0.037, + "step": 58130 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011365384615384615, + "loss": 0.048, + "step": 58140 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011362179487179485, + "loss": 0.0344, + "step": 58150 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011358974358974357, + "loss": 0.0332, + "step": 58160 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011355769230769229, + "loss": 0.0347, + "step": 58170 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011352564102564101, + "loss": 0.0391, + "step": 58180 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011349358974358973, + "loss": 0.0336, + "step": 58190 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011346153846153845, + "loss": 0.0355, + "step": 58200 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011342948717948717, + "loss": 0.0343, + "step": 58210 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011339743589743589, + "loss": 0.0378, + "step": 58220 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011336538461538461, + "loss": 0.0404, + "step": 58230 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011333333333333331, + "loss": 0.0372, + "step": 58240 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011330128205128203, + "loss": 0.0385, + "step": 58250 + }, + { + "epoch": 1.24, + "learning_rate": 0.00011326923076923075, + "loss": 0.0367, + "step": 58260 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011323717948717947, + "loss": 0.0325, + "step": 58270 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011320512820512819, + "loss": 0.0337, + "step": 58280 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011317307692307691, + "loss": 0.0381, + "step": 58290 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011314102564102563, + "loss": 0.0443, + "step": 58300 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011310897435897435, + "loss": 0.0312, + "step": 58310 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011307692307692307, + "loss": 0.0377, + "step": 58320 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011304487179487178, + "loss": 0.0376, + "step": 58330 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001130128205128205, + "loss": 0.0369, + "step": 58340 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011298076923076922, + "loss": 0.0333, + "step": 58350 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011294871794871794, + "loss": 0.0384, + "step": 58360 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011291666666666666, + "loss": 0.0314, + "step": 58370 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011288461538461538, + "loss": 0.0369, + "step": 58380 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001128525641025641, + "loss": 0.0348, + "step": 58390 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011282051282051281, + "loss": 0.0417, + "step": 58400 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011278846153846153, + "loss": 0.0311, + "step": 58410 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011275641025641024, + "loss": 0.0316, + "step": 58420 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011272435897435896, + "loss": 0.0332, + "step": 58430 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011269230769230768, + "loss": 0.0381, + "step": 58440 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001126602564102564, + "loss": 0.0442, + "step": 58450 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011262820512820512, + "loss": 0.0325, + "step": 58460 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011259615384615384, + "loss": 0.0407, + "step": 58470 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011256410256410256, + "loss": 0.036, + "step": 58480 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011253205128205128, + "loss": 0.0342, + "step": 58490 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001125, + "loss": 0.037, + "step": 58500 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001124679487179487, + "loss": 0.0376, + "step": 58510 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011243589743589742, + "loss": 0.0311, + "step": 58520 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011240384615384614, + "loss": 0.04, + "step": 58530 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011237179487179486, + "loss": 0.033, + "step": 58540 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011233974358974358, + "loss": 0.0354, + "step": 58550 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001123076923076923, + "loss": 0.0347, + "step": 58560 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011227564102564102, + "loss": 0.0372, + "step": 58570 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011224358974358974, + "loss": 0.0331, + "step": 58580 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011221153846153845, + "loss": 0.0386, + "step": 58590 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011217948717948717, + "loss": 0.0439, + "step": 58600 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011214743589743589, + "loss": 0.0407, + "step": 58610 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001121153846153846, + "loss": 0.0318, + "step": 58620 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011208333333333332, + "loss": 0.0304, + "step": 58630 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011205128205128204, + "loss": 0.036, + "step": 58640 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011201923076923076, + "loss": 0.0324, + "step": 58650 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011198717948717948, + "loss": 0.0373, + "step": 58660 + }, + { + "epoch": 1.25, + "learning_rate": 0.0001119551282051282, + "loss": 0.033, + "step": 58670 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011192307692307691, + "loss": 0.0349, + "step": 58680 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011189102564102563, + "loss": 0.0348, + "step": 58690 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011185897435897435, + "loss": 0.034, + "step": 58700 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011182692307692307, + "loss": 0.0339, + "step": 58710 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011179487179487179, + "loss": 0.0311, + "step": 58720 + }, + { + "epoch": 1.25, + "learning_rate": 0.00011176282051282051, + "loss": 0.0339, + "step": 58730 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011173076923076923, + "loss": 0.0371, + "step": 58740 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011169871794871795, + "loss": 0.0313, + "step": 58750 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011166666666666667, + "loss": 0.0342, + "step": 58760 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011163461538461537, + "loss": 0.0387, + "step": 58770 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011160256410256409, + "loss": 0.0385, + "step": 58780 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011157051282051281, + "loss": 0.0326, + "step": 58790 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011153846153846153, + "loss": 0.037, + "step": 58800 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011150641025641025, + "loss": 0.0339, + "step": 58810 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011147435897435897, + "loss": 0.0349, + "step": 58820 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011144230769230769, + "loss": 0.0342, + "step": 58830 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011141025641025641, + "loss": 0.04, + "step": 58840 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011137820512820513, + "loss": 0.0301, + "step": 58850 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011134615384615383, + "loss": 0.0339, + "step": 58860 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011131410256410255, + "loss": 0.0352, + "step": 58870 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011128205128205127, + "loss": 0.0398, + "step": 58880 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011125, + "loss": 0.041, + "step": 58890 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011121794871794871, + "loss": 0.0361, + "step": 58900 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011118589743589743, + "loss": 0.0339, + "step": 58910 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011115384615384615, + "loss": 0.0345, + "step": 58920 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011112179487179487, + "loss": 0.0326, + "step": 58930 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011108974358974359, + "loss": 0.0308, + "step": 58940 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001110576923076923, + "loss": 0.0267, + "step": 58950 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011102564102564102, + "loss": 0.0341, + "step": 58960 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011099358974358974, + "loss": 0.0373, + "step": 58970 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011096153846153846, + "loss": 0.0361, + "step": 58980 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011092948717948718, + "loss": 0.0415, + "step": 58990 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001108974358974359, + "loss": 0.0364, + "step": 59000 + }, + { + "epoch": 1.26, + "eval_cer": 0.9236774900911016, + "eval_loss": 0.026541907340288162, + "eval_runtime": 120.4726, + "eval_samples_per_second": 16.601, + "eval_steps_per_second": 4.15, + "step": 59000 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011086538461538462, + "loss": 0.0358, + "step": 59010 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011083333333333333, + "loss": 0.0386, + "step": 59020 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011080128205128204, + "loss": 0.0318, + "step": 59030 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011076923076923076, + "loss": 0.03, + "step": 59040 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011073717948717948, + "loss": 0.0339, + "step": 59050 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001107051282051282, + "loss": 0.0287, + "step": 59060 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011067307692307692, + "loss": 0.0337, + "step": 59070 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011064102564102564, + "loss": 0.0358, + "step": 59080 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011060897435897436, + "loss": 0.0419, + "step": 59090 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011057692307692308, + "loss": 0.0376, + "step": 59100 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001105448717948718, + "loss": 0.031, + "step": 59110 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001105128205128205, + "loss": 0.0334, + "step": 59120 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011048076923076922, + "loss": 0.0367, + "step": 59130 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011044871794871794, + "loss": 0.0307, + "step": 59140 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011041666666666666, + "loss": 0.0336, + "step": 59150 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011038461538461538, + "loss": 0.0369, + "step": 59160 + }, + { + "epoch": 1.26, + "learning_rate": 0.0001103525641025641, + "loss": 0.0387, + "step": 59170 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011032051282051282, + "loss": 0.0337, + "step": 59180 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011028846153846154, + "loss": 0.0299, + "step": 59190 + }, + { + "epoch": 1.26, + "learning_rate": 0.00011025641025641026, + "loss": 0.0389, + "step": 59200 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011022435897435897, + "loss": 0.0383, + "step": 59210 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011019230769230769, + "loss": 0.0359, + "step": 59220 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001101602564102564, + "loss": 0.0353, + "step": 59230 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011012820512820511, + "loss": 0.0347, + "step": 59240 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011009615384615383, + "loss": 0.0349, + "step": 59250 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011006410256410255, + "loss": 0.0344, + "step": 59260 + }, + { + "epoch": 1.27, + "learning_rate": 0.00011003205128205127, + "loss": 0.0314, + "step": 59270 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010999999999999998, + "loss": 0.0401, + "step": 59280 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001099679487179487, + "loss": 0.0363, + "step": 59290 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010993589743589742, + "loss": 0.0329, + "step": 59300 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010990384615384614, + "loss": 0.033, + "step": 59310 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010987179487179486, + "loss": 0.0374, + "step": 59320 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010983974358974357, + "loss": 0.0395, + "step": 59330 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001098076923076923, + "loss": 0.0357, + "step": 59340 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010977564102564101, + "loss": 0.0297, + "step": 59350 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010974358974358973, + "loss": 0.0344, + "step": 59360 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010971153846153844, + "loss": 0.0329, + "step": 59370 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010967948717948716, + "loss": 0.0346, + "step": 59380 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010964743589743588, + "loss": 0.0351, + "step": 59390 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001096153846153846, + "loss": 0.035, + "step": 59400 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010958333333333332, + "loss": 0.0317, + "step": 59410 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010955128205128204, + "loss": 0.0303, + "step": 59420 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010951923076923076, + "loss": 0.0309, + "step": 59430 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010948717948717948, + "loss": 0.0311, + "step": 59440 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010945512820512818, + "loss": 0.0354, + "step": 59450 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001094230769230769, + "loss": 0.031, + "step": 59460 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010939102564102562, + "loss": 0.0331, + "step": 59470 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010935897435897434, + "loss": 0.0369, + "step": 59480 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010932692307692306, + "loss": 0.0315, + "step": 59490 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010929487179487178, + "loss": 0.051, + "step": 59500 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001092628205128205, + "loss": 0.0375, + "step": 59510 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010923076923076922, + "loss": 0.0335, + "step": 59520 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010919871794871794, + "loss": 0.0359, + "step": 59530 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010916666666666665, + "loss": 0.038, + "step": 59540 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010913461538461537, + "loss": 0.0329, + "step": 59550 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010910256410256409, + "loss": 0.0387, + "step": 59560 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001090705128205128, + "loss": 0.0369, + "step": 59570 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010903846153846152, + "loss": 0.0319, + "step": 59580 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010900641025641024, + "loss": 0.0404, + "step": 59590 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010897435897435896, + "loss": 0.0327, + "step": 59600 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010894230769230768, + "loss": 0.0308, + "step": 59610 + }, + { + "epoch": 1.27, + "learning_rate": 0.0001089102564102564, + "loss": 0.0335, + "step": 59620 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010887820512820511, + "loss": 0.0426, + "step": 59630 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010884615384615383, + "loss": 0.0333, + "step": 59640 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010881410256410255, + "loss": 0.0368, + "step": 59650 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010878205128205127, + "loss": 0.0338, + "step": 59660 + }, + { + "epoch": 1.27, + "learning_rate": 0.00010874999999999999, + "loss": 0.0374, + "step": 59670 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001087179487179487, + "loss": 0.0388, + "step": 59680 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010868589743589743, + "loss": 0.0342, + "step": 59690 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010865384615384615, + "loss": 0.0352, + "step": 59700 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010862179487179487, + "loss": 0.0413, + "step": 59710 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010858974358974357, + "loss": 0.0364, + "step": 59720 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010855769230769229, + "loss": 0.0362, + "step": 59730 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010852564102564101, + "loss": 0.0305, + "step": 59740 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010849358974358973, + "loss": 0.0339, + "step": 59750 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010846153846153845, + "loss": 0.0328, + "step": 59760 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010842948717948717, + "loss": 0.034, + "step": 59770 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010839743589743589, + "loss": 0.0343, + "step": 59780 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010836538461538461, + "loss": 0.0298, + "step": 59790 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010833333333333333, + "loss": 0.0319, + "step": 59800 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010830128205128203, + "loss": 0.0344, + "step": 59810 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010826923076923075, + "loss": 0.0345, + "step": 59820 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010823717948717947, + "loss": 0.0347, + "step": 59830 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010820512820512819, + "loss": 0.0385, + "step": 59840 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010817307692307691, + "loss": 0.0323, + "step": 59850 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010814102564102563, + "loss": 0.0386, + "step": 59860 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010810897435897435, + "loss": 0.04, + "step": 59870 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010807692307692307, + "loss": 0.0351, + "step": 59880 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010804487179487179, + "loss": 0.0334, + "step": 59890 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001080128205128205, + "loss": 0.0343, + "step": 59900 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010798076923076922, + "loss": 0.0391, + "step": 59910 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010794871794871794, + "loss": 0.0318, + "step": 59920 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010791666666666666, + "loss": 0.036, + "step": 59930 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010788461538461538, + "loss": 0.0307, + "step": 59940 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001078525641025641, + "loss": 0.0351, + "step": 59950 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010782051282051281, + "loss": 0.036, + "step": 59960 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010778846153846153, + "loss": 0.0332, + "step": 59970 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010775641025641024, + "loss": 0.0295, + "step": 59980 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010772435897435896, + "loss": 0.0308, + "step": 59990 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010769230769230768, + "loss": 0.0402, + "step": 60000 + }, + { + "epoch": 1.28, + "eval_cer": 0.9236893799199475, + "eval_loss": 0.026246914640069008, + "eval_runtime": 120.4556, + "eval_samples_per_second": 16.604, + "eval_steps_per_second": 4.151, + "step": 60000 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001076602564102564, + "loss": 0.032, + "step": 60010 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010762820512820512, + "loss": 0.0315, + "step": 60020 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010759615384615384, + "loss": 0.0371, + "step": 60030 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010756410256410256, + "loss": 0.0365, + "step": 60040 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010753205128205128, + "loss": 0.0402, + "step": 60050 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001075, + "loss": 0.0376, + "step": 60060 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001074679487179487, + "loss": 0.0372, + "step": 60070 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010743589743589742, + "loss": 0.0343, + "step": 60080 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010740384615384614, + "loss": 0.0342, + "step": 60090 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010737179487179486, + "loss": 0.0351, + "step": 60100 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010733974358974358, + "loss": 0.0329, + "step": 60110 + }, + { + "epoch": 1.28, + "learning_rate": 0.0001073076923076923, + "loss": 0.0333, + "step": 60120 + }, + { + "epoch": 1.28, + "learning_rate": 0.00010727564102564102, + "loss": 0.0334, + "step": 60130 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010724358974358974, + "loss": 0.0342, + "step": 60140 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010721153846153846, + "loss": 0.0343, + "step": 60150 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010717948717948717, + "loss": 0.0298, + "step": 60160 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010714743589743589, + "loss": 0.0271, + "step": 60170 + }, + { + "epoch": 1.29, + "learning_rate": 0.0001071153846153846, + "loss": 0.0291, + "step": 60180 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010708333333333333, + "loss": 0.0357, + "step": 60190 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010705128205128204, + "loss": 0.0379, + "step": 60200 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010701923076923076, + "loss": 0.0335, + "step": 60210 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010698717948717948, + "loss": 0.0312, + "step": 60220 + }, + { + "epoch": 1.29, + "learning_rate": 0.0001069551282051282, + "loss": 0.0338, + "step": 60230 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010692307692307692, + "loss": 0.0394, + "step": 60240 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010689102564102563, + "loss": 0.0329, + "step": 60250 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010685897435897435, + "loss": 0.0388, + "step": 60260 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010682692307692307, + "loss": 0.0286, + "step": 60270 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010679487179487179, + "loss": 0.0363, + "step": 60280 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010676282051282051, + "loss": 0.0376, + "step": 60290 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010673076923076923, + "loss": 0.0342, + "step": 60300 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010669871794871795, + "loss": 0.0396, + "step": 60310 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010666666666666667, + "loss": 0.033, + "step": 60320 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010663461538461539, + "loss": 0.0382, + "step": 60330 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010660256410256409, + "loss": 0.0318, + "step": 60340 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010657051282051281, + "loss": 0.0301, + "step": 60350 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010653846153846153, + "loss": 0.03, + "step": 60360 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010650641025641025, + "loss": 0.0358, + "step": 60370 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010647435897435897, + "loss": 0.0361, + "step": 60380 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010644230769230769, + "loss": 0.0342, + "step": 60390 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010641025641025641, + "loss": 0.035, + "step": 60400 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010637820512820513, + "loss": 0.0379, + "step": 60410 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010634615384615384, + "loss": 0.0276, + "step": 60420 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010631410256410255, + "loss": 0.0357, + "step": 60430 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010628205128205127, + "loss": 0.0383, + "step": 60440 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010625, + "loss": 0.038, + "step": 60450 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010621794871794871, + "loss": 0.0385, + "step": 60460 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010618589743589743, + "loss": 0.0388, + "step": 60470 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010615384615384615, + "loss": 0.0297, + "step": 60480 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010612179487179487, + "loss": 0.031, + "step": 60490 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010608974358974359, + "loss": 0.0363, + "step": 60500 + }, + { + "epoch": 1.29, + "learning_rate": 0.0001060576923076923, + "loss": 0.0322, + "step": 60510 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010602564102564102, + "loss": 0.0348, + "step": 60520 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010599358974358974, + "loss": 0.0371, + "step": 60530 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010596153846153846, + "loss": 0.0391, + "step": 60540 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010592948717948718, + "loss": 0.0337, + "step": 60550 + }, + { + "epoch": 1.29, + "learning_rate": 0.0001058974358974359, + "loss": 0.0385, + "step": 60560 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010586538461538462, + "loss": 0.0395, + "step": 60570 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010583333333333334, + "loss": 0.0368, + "step": 60580 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010580128205128205, + "loss": 0.0366, + "step": 60590 + }, + { + "epoch": 1.29, + "learning_rate": 0.00010576923076923076, + "loss": 0.0318, + "step": 60600 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010573717948717948, + "loss": 0.0348, + "step": 60610 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001057051282051282, + "loss": 0.0349, + "step": 60620 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010567307692307692, + "loss": 0.0342, + "step": 60630 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010564102564102564, + "loss": 0.0335, + "step": 60640 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010560897435897436, + "loss": 0.0423, + "step": 60650 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010557692307692308, + "loss": 0.0362, + "step": 60660 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001055448717948718, + "loss": 0.0349, + "step": 60670 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010551282051282052, + "loss": 0.033, + "step": 60680 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010548076923076922, + "loss": 0.0383, + "step": 60690 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010544871794871793, + "loss": 0.0346, + "step": 60700 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010541666666666665, + "loss": 0.0312, + "step": 60710 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010538461538461537, + "loss": 0.0304, + "step": 60720 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010535256410256409, + "loss": 0.0427, + "step": 60730 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010532051282051281, + "loss": 0.0383, + "step": 60740 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010528846153846153, + "loss": 0.0337, + "step": 60750 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010525641025641023, + "loss": 0.0365, + "step": 60760 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010522435897435895, + "loss": 0.0379, + "step": 60770 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010519230769230767, + "loss": 0.0321, + "step": 60780 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010516025641025639, + "loss": 0.0386, + "step": 60790 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010512820512820511, + "loss": 0.0325, + "step": 60800 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010509615384615383, + "loss": 0.0301, + "step": 60810 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010506410256410255, + "loss": 0.0315, + "step": 60820 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010503205128205127, + "loss": 0.0341, + "step": 60830 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010499999999999999, + "loss": 0.0352, + "step": 60840 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001049679487179487, + "loss": 0.0305, + "step": 60850 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010493589743589742, + "loss": 0.0319, + "step": 60860 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010490384615384614, + "loss": 0.0332, + "step": 60870 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010487179487179486, + "loss": 0.0302, + "step": 60880 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010483974358974358, + "loss": 0.0377, + "step": 60890 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001048076923076923, + "loss": 0.0361, + "step": 60900 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010477564102564101, + "loss": 0.0445, + "step": 60910 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010474358974358973, + "loss": 0.0295, + "step": 60920 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010471153846153844, + "loss": 0.0302, + "step": 60930 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010467948717948716, + "loss": 0.032, + "step": 60940 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010464743589743588, + "loss": 0.0337, + "step": 60950 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001046153846153846, + "loss": 0.0326, + "step": 60960 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010458333333333332, + "loss": 0.0374, + "step": 60970 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010455128205128204, + "loss": 0.0378, + "step": 60980 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010451923076923076, + "loss": 0.0362, + "step": 60990 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010448717948717948, + "loss": 0.0341, + "step": 61000 + }, + { + "epoch": 1.3, + "eval_cer": 0.9236664495357446, + "eval_loss": 0.025926979258656502, + "eval_runtime": 120.4904, + "eval_samples_per_second": 16.599, + "eval_steps_per_second": 4.15, + "step": 61000 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001044551282051282, + "loss": 0.0374, + "step": 61010 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001044230769230769, + "loss": 0.0289, + "step": 61020 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010439102564102562, + "loss": 0.037, + "step": 61030 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010435897435897434, + "loss": 0.0321, + "step": 61040 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010432692307692306, + "loss": 0.0389, + "step": 61050 + }, + { + "epoch": 1.3, + "learning_rate": 0.00010429487179487178, + "loss": 0.0321, + "step": 61060 + }, + { + "epoch": 1.3, + "learning_rate": 0.0001042628205128205, + "loss": 0.0331, + "step": 61070 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010423076923076922, + "loss": 0.0309, + "step": 61080 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010419871794871794, + "loss": 0.0401, + "step": 61090 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010416666666666666, + "loss": 0.0351, + "step": 61100 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010413461538461537, + "loss": 0.0368, + "step": 61110 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010410256410256409, + "loss": 0.0385, + "step": 61120 + }, + { + "epoch": 1.31, + "learning_rate": 0.0001040705128205128, + "loss": 0.0366, + "step": 61130 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010403846153846152, + "loss": 0.0394, + "step": 61140 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010400641025641024, + "loss": 0.0344, + "step": 61150 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010397435897435896, + "loss": 0.0408, + "step": 61160 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010394230769230768, + "loss": 0.0341, + "step": 61170 + }, + { + "epoch": 1.31, + "learning_rate": 0.0001039102564102564, + "loss": 0.0346, + "step": 61180 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010387820512820512, + "loss": 0.0377, + "step": 61190 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010384615384615383, + "loss": 0.0292, + "step": 61200 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010381410256410255, + "loss": 0.0337, + "step": 61210 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010378205128205127, + "loss": 0.0338, + "step": 61220 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010374999999999999, + "loss": 0.0349, + "step": 61230 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010371794871794871, + "loss": 0.0306, + "step": 61240 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010368589743589743, + "loss": 0.0256, + "step": 61250 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010365384615384615, + "loss": 0.0337, + "step": 61260 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010362179487179487, + "loss": 0.0322, + "step": 61270 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010358974358974359, + "loss": 0.0333, + "step": 61280 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010355769230769229, + "loss": 0.0328, + "step": 61290 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010352564102564101, + "loss": 0.0304, + "step": 61300 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010349358974358973, + "loss": 0.0425, + "step": 61310 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010346153846153845, + "loss": 0.0317, + "step": 61320 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010342948717948717, + "loss": 0.0307, + "step": 61330 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010339743589743589, + "loss": 0.0349, + "step": 61340 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010336538461538461, + "loss": 0.0356, + "step": 61350 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010333333333333333, + "loss": 0.0368, + "step": 61360 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010330128205128203, + "loss": 0.0334, + "step": 61370 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010326923076923075, + "loss": 0.0286, + "step": 61380 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010323717948717947, + "loss": 0.031, + "step": 61390 + }, + { + "epoch": 1.31, + "learning_rate": 0.0001032051282051282, + "loss": 0.0292, + "step": 61400 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010317307692307691, + "loss": 0.0424, + "step": 61410 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010314102564102563, + "loss": 0.0374, + "step": 61420 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010310897435897435, + "loss": 0.0324, + "step": 61430 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010307692307692307, + "loss": 0.0379, + "step": 61440 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010304487179487179, + "loss": 0.0353, + "step": 61450 + }, + { + "epoch": 1.31, + "learning_rate": 0.0001030128205128205, + "loss": 0.0443, + "step": 61460 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010298076923076922, + "loss": 0.0392, + "step": 61470 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010294871794871794, + "loss": 0.037, + "step": 61480 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010291666666666666, + "loss": 0.0332, + "step": 61490 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010288461538461538, + "loss": 0.0406, + "step": 61500 + }, + { + "epoch": 1.31, + "learning_rate": 0.0001028525641025641, + "loss": 0.0343, + "step": 61510 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010282051282051282, + "loss": 0.0346, + "step": 61520 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010278846153846153, + "loss": 0.0295, + "step": 61530 + }, + { + "epoch": 1.31, + "learning_rate": 0.00010275641025641025, + "loss": 0.0383, + "step": 61540 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010272435897435896, + "loss": 0.0355, + "step": 61550 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010269230769230768, + "loss": 0.0335, + "step": 61560 + }, + { + "epoch": 1.32, + "learning_rate": 0.0001026602564102564, + "loss": 0.0386, + "step": 61570 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010262820512820512, + "loss": 0.026, + "step": 61580 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010259615384615384, + "loss": 0.0299, + "step": 61590 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010256410256410256, + "loss": 0.037, + "step": 61600 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010253205128205128, + "loss": 0.0435, + "step": 61610 + }, + { + "epoch": 1.32, + "learning_rate": 0.0001025, + "loss": 0.0352, + "step": 61620 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010246794871794872, + "loss": 0.0372, + "step": 61630 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010243589743589742, + "loss": 0.0318, + "step": 61640 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010240384615384614, + "loss": 0.0335, + "step": 61650 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010237179487179486, + "loss": 0.0327, + "step": 61660 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010233974358974358, + "loss": 0.0353, + "step": 61670 + }, + { + "epoch": 1.32, + "learning_rate": 0.0001023076923076923, + "loss": 0.033, + "step": 61680 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010227564102564102, + "loss": 0.0361, + "step": 61690 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010224358974358974, + "loss": 0.0389, + "step": 61700 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010221153846153846, + "loss": 0.0432, + "step": 61710 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010217948717948718, + "loss": 0.0358, + "step": 61720 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010214743589743589, + "loss": 0.0328, + "step": 61730 + }, + { + "epoch": 1.32, + "learning_rate": 0.0001021153846153846, + "loss": 0.0324, + "step": 61740 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010208333333333333, + "loss": 0.0328, + "step": 61750 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010205128205128205, + "loss": 0.0398, + "step": 61760 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010201923076923076, + "loss": 0.0391, + "step": 61770 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010198717948717948, + "loss": 0.0351, + "step": 61780 + }, + { + "epoch": 1.32, + "learning_rate": 0.0001019551282051282, + "loss": 0.0316, + "step": 61790 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010192307692307692, + "loss": 0.0335, + "step": 61800 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010189102564102564, + "loss": 0.0314, + "step": 61810 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010185897435897435, + "loss": 0.0297, + "step": 61820 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010182692307692307, + "loss": 0.039, + "step": 61830 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010179487179487179, + "loss": 0.0353, + "step": 61840 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010176282051282051, + "loss": 0.0354, + "step": 61850 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010173076923076923, + "loss": 0.0271, + "step": 61860 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010169871794871795, + "loss": 0.0305, + "step": 61870 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010166666666666667, + "loss": 0.04, + "step": 61880 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010163461538461539, + "loss": 0.033, + "step": 61890 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010160256410256409, + "loss": 0.0344, + "step": 61900 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010157051282051281, + "loss": 0.0379, + "step": 61910 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010153846153846153, + "loss": 0.0312, + "step": 61920 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010150641025641025, + "loss": 0.0424, + "step": 61930 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010147435897435897, + "loss": 0.0426, + "step": 61940 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010144230769230769, + "loss": 0.042, + "step": 61950 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010141025641025641, + "loss": 0.0403, + "step": 61960 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010137820512820513, + "loss": 0.0348, + "step": 61970 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010134615384615385, + "loss": 0.0361, + "step": 61980 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010131410256410256, + "loss": 0.0351, + "step": 61990 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010128205128205127, + "loss": 0.0346, + "step": 62000 + }, + { + "epoch": 1.32, + "eval_cer": 0.9236596553478327, + "eval_loss": 0.025804216042160988, + "eval_runtime": 120.3519, + "eval_samples_per_second": 16.618, + "eval_steps_per_second": 4.154, + "step": 62000 + }, + { + "epoch": 1.32, + "learning_rate": 0.00010125, + "loss": 0.0394, + "step": 62010 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010121794871794871, + "loss": 0.0306, + "step": 62020 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010118589743589743, + "loss": 0.0371, + "step": 62030 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010115384615384615, + "loss": 0.0319, + "step": 62040 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010112179487179487, + "loss": 0.0327, + "step": 62050 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010108974358974359, + "loss": 0.0371, + "step": 62060 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010105769230769231, + "loss": 0.0343, + "step": 62070 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010102564102564102, + "loss": 0.0368, + "step": 62080 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010099358974358974, + "loss": 0.0475, + "step": 62090 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010096153846153846, + "loss": 0.0278, + "step": 62100 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010092948717948718, + "loss": 0.0333, + "step": 62110 + }, + { + "epoch": 1.33, + "learning_rate": 0.0001008974358974359, + "loss": 0.0321, + "step": 62120 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010086538461538462, + "loss": 0.0334, + "step": 62130 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010083333333333334, + "loss": 0.0358, + "step": 62140 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010080128205128206, + "loss": 0.0351, + "step": 62150 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010076923076923075, + "loss": 0.04, + "step": 62160 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010073717948717947, + "loss": 0.0312, + "step": 62170 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010070512820512819, + "loss": 0.0371, + "step": 62180 + }, + { + "epoch": 1.33, + "learning_rate": 0.0001006730769230769, + "loss": 0.0319, + "step": 62190 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010064102564102563, + "loss": 0.0342, + "step": 62200 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010060897435897435, + "loss": 0.0289, + "step": 62210 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010057692307692307, + "loss": 0.0389, + "step": 62220 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010054487179487179, + "loss": 0.0391, + "step": 62230 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010051282051282049, + "loss": 0.0308, + "step": 62240 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010048076923076921, + "loss": 0.0434, + "step": 62250 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010044871794871793, + "loss": 0.0334, + "step": 62260 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010041666666666665, + "loss": 0.0314, + "step": 62270 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010038461538461537, + "loss": 0.0394, + "step": 62280 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010035256410256409, + "loss": 0.0341, + "step": 62290 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010032051282051281, + "loss": 0.0298, + "step": 62300 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010028846153846153, + "loss": 0.0352, + "step": 62310 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010025641025641023, + "loss": 0.0338, + "step": 62320 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010022435897435895, + "loss": 0.0353, + "step": 62330 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010019230769230767, + "loss": 0.0337, + "step": 62340 + }, + { + "epoch": 1.33, + "learning_rate": 0.0001001602564102564, + "loss": 0.0295, + "step": 62350 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010012820512820511, + "loss": 0.0328, + "step": 62360 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010009615384615383, + "loss": 0.0339, + "step": 62370 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010006410256410255, + "loss": 0.0367, + "step": 62380 + }, + { + "epoch": 1.33, + "learning_rate": 0.00010003205128205127, + "loss": 0.0349, + "step": 62390 + }, + { + "epoch": 1.33, + "learning_rate": 9.999999999999999e-05, + "loss": 0.0355, + "step": 62400 + }, + { + "epoch": 1.33, + "learning_rate": 9.99679487179487e-05, + "loss": 0.0307, + "step": 62410 + }, + { + "epoch": 1.33, + "learning_rate": 9.993589743589742e-05, + "loss": 0.0334, + "step": 62420 + }, + { + "epoch": 1.33, + "learning_rate": 9.990384615384614e-05, + "loss": 0.0365, + "step": 62430 + }, + { + "epoch": 1.33, + "learning_rate": 9.987179487179486e-05, + "loss": 0.035, + "step": 62440 + }, + { + "epoch": 1.33, + "learning_rate": 9.983974358974358e-05, + "loss": 0.0291, + "step": 62450 + }, + { + "epoch": 1.33, + "learning_rate": 9.98076923076923e-05, + "loss": 0.0273, + "step": 62460 + }, + { + "epoch": 1.33, + "learning_rate": 9.977564102564101e-05, + "loss": 0.0302, + "step": 62470 + }, + { + "epoch": 1.34, + "learning_rate": 9.974358974358973e-05, + "loss": 0.0386, + "step": 62480 + }, + { + "epoch": 1.34, + "learning_rate": 9.971153846153845e-05, + "loss": 0.0439, + "step": 62490 + }, + { + "epoch": 1.34, + "learning_rate": 9.967948717948716e-05, + "loss": 0.0337, + "step": 62500 + }, + { + "epoch": 1.34, + "learning_rate": 9.964743589743588e-05, + "loss": 0.0361, + "step": 62510 + }, + { + "epoch": 1.34, + "learning_rate": 9.96153846153846e-05, + "loss": 0.0356, + "step": 62520 + }, + { + "epoch": 1.34, + "learning_rate": 9.958333333333332e-05, + "loss": 0.0321, + "step": 62530 + }, + { + "epoch": 1.34, + "learning_rate": 9.955128205128204e-05, + "loss": 0.0401, + "step": 62540 + }, + { + "epoch": 1.34, + "learning_rate": 9.951923076923076e-05, + "loss": 0.0335, + "step": 62550 + }, + { + "epoch": 1.34, + "learning_rate": 9.948717948717948e-05, + "loss": 0.0329, + "step": 62560 + }, + { + "epoch": 1.34, + "learning_rate": 9.94551282051282e-05, + "loss": 0.0311, + "step": 62570 + }, + { + "epoch": 1.34, + "learning_rate": 9.942307692307692e-05, + "loss": 0.0393, + "step": 62580 + }, + { + "epoch": 1.34, + "learning_rate": 9.939102564102562e-05, + "loss": 0.0327, + "step": 62590 + }, + { + "epoch": 1.34, + "learning_rate": 9.935897435897434e-05, + "loss": 0.0322, + "step": 62600 + }, + { + "epoch": 1.34, + "learning_rate": 9.932692307692306e-05, + "loss": 0.031, + "step": 62610 + }, + { + "epoch": 1.34, + "learning_rate": 9.929487179487178e-05, + "loss": 0.036, + "step": 62620 + }, + { + "epoch": 1.34, + "learning_rate": 9.92628205128205e-05, + "loss": 0.0407, + "step": 62630 + }, + { + "epoch": 1.34, + "learning_rate": 9.923076923076922e-05, + "loss": 0.0288, + "step": 62640 + }, + { + "epoch": 1.34, + "learning_rate": 9.919871794871794e-05, + "loss": 0.0296, + "step": 62650 + }, + { + "epoch": 1.34, + "learning_rate": 9.916666666666666e-05, + "loss": 0.0369, + "step": 62660 + }, + { + "epoch": 1.34, + "learning_rate": 9.913461538461538e-05, + "loss": 0.0358, + "step": 62670 + }, + { + "epoch": 1.34, + "learning_rate": 9.910256410256409e-05, + "loss": 0.0341, + "step": 62680 + }, + { + "epoch": 1.34, + "learning_rate": 9.90705128205128e-05, + "loss": 0.0345, + "step": 62690 + }, + { + "epoch": 1.34, + "learning_rate": 9.903846153846153e-05, + "loss": 0.038, + "step": 62700 + }, + { + "epoch": 1.34, + "learning_rate": 9.900641025641024e-05, + "loss": 0.0315, + "step": 62710 + }, + { + "epoch": 1.34, + "learning_rate": 9.897435897435896e-05, + "loss": 0.0361, + "step": 62720 + }, + { + "epoch": 1.34, + "learning_rate": 9.894230769230768e-05, + "loss": 0.0328, + "step": 62730 + }, + { + "epoch": 1.34, + "learning_rate": 9.89102564102564e-05, + "loss": 0.0317, + "step": 62740 + }, + { + "epoch": 1.34, + "learning_rate": 9.887820512820512e-05, + "loss": 0.0324, + "step": 62750 + }, + { + "epoch": 1.34, + "learning_rate": 9.884615384615384e-05, + "loss": 0.0378, + "step": 62760 + }, + { + "epoch": 1.34, + "learning_rate": 9.881410256410255e-05, + "loss": 0.0336, + "step": 62770 + }, + { + "epoch": 1.34, + "learning_rate": 9.878205128205127e-05, + "loss": 0.0298, + "step": 62780 + }, + { + "epoch": 1.34, + "learning_rate": 9.874999999999999e-05, + "loss": 0.0359, + "step": 62790 + }, + { + "epoch": 1.34, + "learning_rate": 9.871794871794871e-05, + "loss": 0.0301, + "step": 62800 + }, + { + "epoch": 1.34, + "learning_rate": 9.868589743589743e-05, + "loss": 0.0326, + "step": 62810 + }, + { + "epoch": 1.34, + "learning_rate": 9.865384615384615e-05, + "loss": 0.032, + "step": 62820 + }, + { + "epoch": 1.34, + "learning_rate": 9.862179487179487e-05, + "loss": 0.0345, + "step": 62830 + }, + { + "epoch": 1.34, + "learning_rate": 9.858974358974359e-05, + "loss": 0.0337, + "step": 62840 + }, + { + "epoch": 1.34, + "learning_rate": 9.855769230769229e-05, + "loss": 0.0326, + "step": 62850 + }, + { + "epoch": 1.34, + "learning_rate": 9.852564102564101e-05, + "loss": 0.0344, + "step": 62860 + }, + { + "epoch": 1.34, + "learning_rate": 9.849358974358973e-05, + "loss": 0.0303, + "step": 62870 + }, + { + "epoch": 1.34, + "learning_rate": 9.846153846153845e-05, + "loss": 0.0373, + "step": 62880 + }, + { + "epoch": 1.34, + "learning_rate": 9.842948717948717e-05, + "loss": 0.0299, + "step": 62890 + }, + { + "epoch": 1.34, + "learning_rate": 9.839743589743589e-05, + "loss": 0.0292, + "step": 62900 + }, + { + "epoch": 1.34, + "learning_rate": 9.836538461538461e-05, + "loss": 0.033, + "step": 62910 + }, + { + "epoch": 1.34, + "learning_rate": 9.833333333333333e-05, + "loss": 0.0335, + "step": 62920 + }, + { + "epoch": 1.34, + "learning_rate": 9.830128205128205e-05, + "loss": 0.0356, + "step": 62930 + }, + { + "epoch": 1.34, + "learning_rate": 9.826923076923075e-05, + "loss": 0.0339, + "step": 62940 + }, + { + "epoch": 1.35, + "learning_rate": 9.823717948717947e-05, + "loss": 0.0383, + "step": 62950 + }, + { + "epoch": 1.35, + "learning_rate": 9.82051282051282e-05, + "loss": 0.0344, + "step": 62960 + }, + { + "epoch": 1.35, + "learning_rate": 9.817307692307691e-05, + "loss": 0.0366, + "step": 62970 + }, + { + "epoch": 1.35, + "learning_rate": 9.814102564102563e-05, + "loss": 0.032, + "step": 62980 + }, + { + "epoch": 1.35, + "learning_rate": 9.810897435897435e-05, + "loss": 0.038, + "step": 62990 + }, + { + "epoch": 1.35, + "learning_rate": 9.807692307692307e-05, + "loss": 0.0378, + "step": 63000 + }, + { + "epoch": 1.35, + "eval_cer": 0.9235815221868453, + "eval_loss": 0.025751110166311264, + "eval_runtime": 120.3856, + "eval_samples_per_second": 16.613, + "eval_steps_per_second": 4.153, + "step": 63000 + }, + { + "epoch": 1.35, + "learning_rate": 9.804487179487179e-05, + "loss": 0.0351, + "step": 63010 + }, + { + "epoch": 1.35, + "learning_rate": 9.801282051282051e-05, + "loss": 0.0334, + "step": 63020 + }, + { + "epoch": 1.35, + "learning_rate": 9.798076923076922e-05, + "loss": 0.0322, + "step": 63030 + }, + { + "epoch": 1.35, + "learning_rate": 9.794871794871794e-05, + "loss": 0.0389, + "step": 63040 + }, + { + "epoch": 1.35, + "learning_rate": 9.791666666666666e-05, + "loss": 0.0338, + "step": 63050 + }, + { + "epoch": 1.35, + "learning_rate": 9.788461538461538e-05, + "loss": 0.0269, + "step": 63060 + }, + { + "epoch": 1.35, + "learning_rate": 9.78525641025641e-05, + "loss": 0.036, + "step": 63070 + }, + { + "epoch": 1.35, + "learning_rate": 9.782051282051282e-05, + "loss": 0.0409, + "step": 63080 + }, + { + "epoch": 1.35, + "learning_rate": 9.778846153846154e-05, + "loss": 0.0373, + "step": 63090 + }, + { + "epoch": 1.35, + "learning_rate": 9.775641025641025e-05, + "loss": 0.0306, + "step": 63100 + }, + { + "epoch": 1.35, + "learning_rate": 9.772435897435897e-05, + "loss": 0.0448, + "step": 63110 + }, + { + "epoch": 1.35, + "learning_rate": 9.769230769230768e-05, + "loss": 0.0333, + "step": 63120 + }, + { + "epoch": 1.35, + "learning_rate": 9.76602564102564e-05, + "loss": 0.0369, + "step": 63130 + }, + { + "epoch": 1.35, + "learning_rate": 9.762820512820512e-05, + "loss": 0.0374, + "step": 63140 + }, + { + "epoch": 1.35, + "learning_rate": 9.759615384615384e-05, + "loss": 0.035, + "step": 63150 + }, + { + "epoch": 1.35, + "learning_rate": 9.756410256410256e-05, + "loss": 0.0333, + "step": 63160 + }, + { + "epoch": 1.35, + "learning_rate": 9.753205128205128e-05, + "loss": 0.0307, + "step": 63170 + }, + { + "epoch": 1.35, + "learning_rate": 9.75e-05, + "loss": 0.0302, + "step": 63180 + }, + { + "epoch": 1.35, + "learning_rate": 9.746794871794872e-05, + "loss": 0.0334, + "step": 63190 + }, + { + "epoch": 1.35, + "learning_rate": 9.743589743589744e-05, + "loss": 0.0342, + "step": 63200 + }, + { + "epoch": 1.35, + "learning_rate": 9.740384615384614e-05, + "loss": 0.0368, + "step": 63210 + }, + { + "epoch": 1.35, + "learning_rate": 9.737179487179486e-05, + "loss": 0.0376, + "step": 63220 + }, + { + "epoch": 1.35, + "learning_rate": 9.733974358974358e-05, + "loss": 0.0354, + "step": 63230 + }, + { + "epoch": 1.35, + "learning_rate": 9.73076923076923e-05, + "loss": 0.0276, + "step": 63240 + }, + { + "epoch": 1.35, + "learning_rate": 9.727564102564102e-05, + "loss": 0.038, + "step": 63250 + }, + { + "epoch": 1.35, + "learning_rate": 9.724358974358974e-05, + "loss": 0.0347, + "step": 63260 + }, + { + "epoch": 1.35, + "learning_rate": 9.721153846153846e-05, + "loss": 0.0341, + "step": 63270 + }, + { + "epoch": 1.35, + "learning_rate": 9.717948717948718e-05, + "loss": 0.0344, + "step": 63280 + }, + { + "epoch": 1.35, + "learning_rate": 9.714743589743589e-05, + "loss": 0.0339, + "step": 63290 + }, + { + "epoch": 1.35, + "learning_rate": 9.71153846153846e-05, + "loss": 0.03, + "step": 63300 + }, + { + "epoch": 1.35, + "learning_rate": 9.708333333333333e-05, + "loss": 0.0263, + "step": 63310 + }, + { + "epoch": 1.35, + "learning_rate": 9.705128205128205e-05, + "loss": 0.0314, + "step": 63320 + }, + { + "epoch": 1.35, + "learning_rate": 9.701923076923077e-05, + "loss": 0.0311, + "step": 63330 + }, + { + "epoch": 1.35, + "learning_rate": 9.698717948717948e-05, + "loss": 0.0339, + "step": 63340 + }, + { + "epoch": 1.35, + "learning_rate": 9.69551282051282e-05, + "loss": 0.0337, + "step": 63350 + }, + { + "epoch": 1.35, + "learning_rate": 9.692307692307692e-05, + "loss": 0.0342, + "step": 63360 + }, + { + "epoch": 1.35, + "learning_rate": 9.689102564102564e-05, + "loss": 0.0379, + "step": 63370 + }, + { + "epoch": 1.35, + "learning_rate": 9.685897435897435e-05, + "loss": 0.0366, + "step": 63380 + }, + { + "epoch": 1.35, + "learning_rate": 9.682692307692307e-05, + "loss": 0.0335, + "step": 63390 + }, + { + "epoch": 1.35, + "learning_rate": 9.679487179487179e-05, + "loss": 0.0378, + "step": 63400 + }, + { + "epoch": 1.35, + "learning_rate": 9.676282051282051e-05, + "loss": 0.0304, + "step": 63410 + }, + { + "epoch": 1.36, + "learning_rate": 9.673076923076923e-05, + "loss": 0.0355, + "step": 63420 + }, + { + "epoch": 1.36, + "learning_rate": 9.669871794871795e-05, + "loss": 0.0303, + "step": 63430 + }, + { + "epoch": 1.36, + "learning_rate": 9.666666666666667e-05, + "loss": 0.0337, + "step": 63440 + }, + { + "epoch": 1.36, + "learning_rate": 9.663461538461539e-05, + "loss": 0.035, + "step": 63450 + }, + { + "epoch": 1.36, + "learning_rate": 9.66025641025641e-05, + "loss": 0.0383, + "step": 63460 + }, + { + "epoch": 1.36, + "learning_rate": 9.657051282051281e-05, + "loss": 0.0342, + "step": 63470 + }, + { + "epoch": 1.36, + "learning_rate": 9.653846153846153e-05, + "loss": 0.0322, + "step": 63480 + }, + { + "epoch": 1.36, + "learning_rate": 9.650641025641025e-05, + "loss": 0.0351, + "step": 63490 + }, + { + "epoch": 1.36, + "learning_rate": 9.647435897435897e-05, + "loss": 0.0343, + "step": 63500 + }, + { + "epoch": 1.36, + "learning_rate": 9.644230769230769e-05, + "loss": 0.033, + "step": 63510 + }, + { + "epoch": 1.36, + "learning_rate": 9.641025641025641e-05, + "loss": 0.0306, + "step": 63520 + }, + { + "epoch": 1.36, + "learning_rate": 9.637820512820513e-05, + "loss": 0.028, + "step": 63530 + }, + { + "epoch": 1.36, + "learning_rate": 9.634615384615385e-05, + "loss": 0.0342, + "step": 63540 + }, + { + "epoch": 1.36, + "learning_rate": 9.631410256410257e-05, + "loss": 0.0267, + "step": 63550 + }, + { + "epoch": 1.36, + "learning_rate": 9.628205128205128e-05, + "loss": 0.0296, + "step": 63560 + }, + { + "epoch": 1.36, + "learning_rate": 9.625e-05, + "loss": 0.0291, + "step": 63570 + }, + { + "epoch": 1.36, + "learning_rate": 9.621794871794871e-05, + "loss": 0.0372, + "step": 63580 + }, + { + "epoch": 1.36, + "learning_rate": 9.618589743589743e-05, + "loss": 0.0304, + "step": 63590 + }, + { + "epoch": 1.36, + "learning_rate": 9.615384615384615e-05, + "loss": 0.0351, + "step": 63600 + }, + { + "epoch": 1.36, + "learning_rate": 9.612179487179487e-05, + "loss": 0.0377, + "step": 63610 + }, + { + "epoch": 1.36, + "learning_rate": 9.608974358974358e-05, + "loss": 0.0314, + "step": 63620 + }, + { + "epoch": 1.36, + "learning_rate": 9.605769230769229e-05, + "loss": 0.0316, + "step": 63630 + }, + { + "epoch": 1.36, + "learning_rate": 9.6025641025641e-05, + "loss": 0.0308, + "step": 63640 + }, + { + "epoch": 1.36, + "learning_rate": 9.599358974358972e-05, + "loss": 0.0322, + "step": 63650 + }, + { + "epoch": 1.36, + "learning_rate": 9.596153846153844e-05, + "loss": 0.0337, + "step": 63660 + }, + { + "epoch": 1.36, + "learning_rate": 9.592948717948716e-05, + "loss": 0.0321, + "step": 63670 + }, + { + "epoch": 1.36, + "learning_rate": 9.589743589743588e-05, + "loss": 0.029, + "step": 63680 + }, + { + "epoch": 1.36, + "learning_rate": 9.58653846153846e-05, + "loss": 0.029, + "step": 63690 + }, + { + "epoch": 1.36, + "learning_rate": 9.583333333333332e-05, + "loss": 0.0375, + "step": 63700 + }, + { + "epoch": 1.36, + "learning_rate": 9.580128205128204e-05, + "loss": 0.0389, + "step": 63710 + }, + { + "epoch": 1.36, + "learning_rate": 9.576923076923075e-05, + "loss": 0.0363, + "step": 63720 + }, + { + "epoch": 1.36, + "learning_rate": 9.573717948717947e-05, + "loss": 0.0335, + "step": 63730 + }, + { + "epoch": 1.36, + "learning_rate": 9.570512820512819e-05, + "loss": 0.0334, + "step": 63740 + }, + { + "epoch": 1.36, + "learning_rate": 9.567307692307691e-05, + "loss": 0.0365, + "step": 63750 + }, + { + "epoch": 1.36, + "learning_rate": 9.564102564102563e-05, + "loss": 0.0348, + "step": 63760 + }, + { + "epoch": 1.36, + "learning_rate": 9.560897435897435e-05, + "loss": 0.0356, + "step": 63770 + }, + { + "epoch": 1.36, + "learning_rate": 9.557692307692307e-05, + "loss": 0.0319, + "step": 63780 + }, + { + "epoch": 1.36, + "learning_rate": 9.554487179487179e-05, + "loss": 0.0301, + "step": 63790 + }, + { + "epoch": 1.36, + "learning_rate": 9.551282051282049e-05, + "loss": 0.0374, + "step": 63800 + }, + { + "epoch": 1.36, + "learning_rate": 9.548076923076921e-05, + "loss": 0.0435, + "step": 63810 + }, + { + "epoch": 1.36, + "learning_rate": 9.544871794871793e-05, + "loss": 0.0362, + "step": 63820 + }, + { + "epoch": 1.36, + "learning_rate": 9.541666666666665e-05, + "loss": 0.0364, + "step": 63830 + }, + { + "epoch": 1.36, + "learning_rate": 9.538461538461537e-05, + "loss": 0.0343, + "step": 63840 + }, + { + "epoch": 1.36, + "learning_rate": 9.535256410256409e-05, + "loss": 0.0281, + "step": 63850 + }, + { + "epoch": 1.36, + "learning_rate": 9.532051282051281e-05, + "loss": 0.0368, + "step": 63860 + }, + { + "epoch": 1.36, + "learning_rate": 9.528846153846153e-05, + "loss": 0.036, + "step": 63870 + }, + { + "epoch": 1.36, + "learning_rate": 9.525641025641025e-05, + "loss": 0.0365, + "step": 63880 + }, + { + "epoch": 1.37, + "learning_rate": 9.522435897435895e-05, + "loss": 0.0288, + "step": 63890 + }, + { + "epoch": 1.37, + "learning_rate": 9.519230769230767e-05, + "loss": 0.036, + "step": 63900 + }, + { + "epoch": 1.37, + "learning_rate": 9.51602564102564e-05, + "loss": 0.0376, + "step": 63910 + }, + { + "epoch": 1.37, + "learning_rate": 9.512820512820511e-05, + "loss": 0.0317, + "step": 63920 + }, + { + "epoch": 1.37, + "learning_rate": 9.509615384615383e-05, + "loss": 0.0327, + "step": 63930 + }, + { + "epoch": 1.37, + "learning_rate": 9.506410256410255e-05, + "loss": 0.0353, + "step": 63940 + }, + { + "epoch": 1.37, + "learning_rate": 9.503205128205127e-05, + "loss": 0.0392, + "step": 63950 + }, + { + "epoch": 1.37, + "learning_rate": 9.499999999999999e-05, + "loss": 0.0315, + "step": 63960 + }, + { + "epoch": 1.37, + "learning_rate": 9.496794871794871e-05, + "loss": 0.0334, + "step": 63970 + }, + { + "epoch": 1.37, + "learning_rate": 9.493589743589742e-05, + "loss": 0.0373, + "step": 63980 + }, + { + "epoch": 1.37, + "learning_rate": 9.490384615384614e-05, + "loss": 0.034, + "step": 63990 + }, + { + "epoch": 1.37, + "learning_rate": 9.487179487179486e-05, + "loss": 0.0372, + "step": 64000 + }, + { + "epoch": 1.37, + "eval_cer": 0.9236740929971456, + "eval_loss": 0.02532467246055603, + "eval_runtime": 120.7173, + "eval_samples_per_second": 16.568, + "eval_steps_per_second": 4.142, + "step": 64000 + }, + { + "epoch": 1.37, + "learning_rate": 9.483974358974358e-05, + "loss": 0.0359, + "step": 64010 + }, + { + "epoch": 1.37, + "learning_rate": 9.48076923076923e-05, + "loss": 0.0381, + "step": 64020 + }, + { + "epoch": 1.37, + "learning_rate": 9.477564102564102e-05, + "loss": 0.0323, + "step": 64030 + }, + { + "epoch": 1.37, + "learning_rate": 9.474358974358973e-05, + "loss": 0.0298, + "step": 64040 + }, + { + "epoch": 1.37, + "learning_rate": 9.471153846153845e-05, + "loss": 0.0351, + "step": 64050 + }, + { + "epoch": 1.37, + "learning_rate": 9.467948717948717e-05, + "loss": 0.0333, + "step": 64060 + }, + { + "epoch": 1.37, + "learning_rate": 9.464743589743588e-05, + "loss": 0.0361, + "step": 64070 + }, + { + "epoch": 1.37, + "learning_rate": 9.46153846153846e-05, + "loss": 0.0343, + "step": 64080 + }, + { + "epoch": 1.37, + "learning_rate": 9.458333333333332e-05, + "loss": 0.029, + "step": 64090 + }, + { + "epoch": 1.37, + "learning_rate": 9.455128205128204e-05, + "loss": 0.0355, + "step": 64100 + }, + { + "epoch": 1.37, + "learning_rate": 9.451923076923076e-05, + "loss": 0.0371, + "step": 64110 + }, + { + "epoch": 1.37, + "learning_rate": 9.448717948717948e-05, + "loss": 0.0349, + "step": 64120 + }, + { + "epoch": 1.37, + "learning_rate": 9.44551282051282e-05, + "loss": 0.0387, + "step": 64130 + }, + { + "epoch": 1.37, + "learning_rate": 9.442307692307692e-05, + "loss": 0.0392, + "step": 64140 + }, + { + "epoch": 1.37, + "learning_rate": 9.439102564102564e-05, + "loss": 0.0295, + "step": 64150 + }, + { + "epoch": 1.37, + "learning_rate": 9.435897435897434e-05, + "loss": 0.0304, + "step": 64160 + }, + { + "epoch": 1.37, + "learning_rate": 9.432692307692306e-05, + "loss": 0.0299, + "step": 64170 + }, + { + "epoch": 1.37, + "learning_rate": 9.429487179487178e-05, + "loss": 0.0296, + "step": 64180 + }, + { + "epoch": 1.37, + "learning_rate": 9.42628205128205e-05, + "loss": 0.0352, + "step": 64190 + }, + { + "epoch": 1.37, + "learning_rate": 9.423076923076922e-05, + "loss": 0.0396, + "step": 64200 + }, + { + "epoch": 1.37, + "learning_rate": 9.419871794871794e-05, + "loss": 0.0449, + "step": 64210 + }, + { + "epoch": 1.37, + "learning_rate": 9.416666666666666e-05, + "loss": 0.0365, + "step": 64220 + }, + { + "epoch": 1.37, + "learning_rate": 9.413461538461538e-05, + "loss": 0.0329, + "step": 64230 + }, + { + "epoch": 1.37, + "learning_rate": 9.410256410256409e-05, + "loss": 0.027, + "step": 64240 + }, + { + "epoch": 1.37, + "learning_rate": 9.40705128205128e-05, + "loss": 0.0301, + "step": 64250 + }, + { + "epoch": 1.37, + "learning_rate": 9.403846153846153e-05, + "loss": 0.0322, + "step": 64260 + }, + { + "epoch": 1.37, + "learning_rate": 9.400641025641025e-05, + "loss": 0.0369, + "step": 64270 + }, + { + "epoch": 1.37, + "learning_rate": 9.397435897435896e-05, + "loss": 0.0301, + "step": 64280 + }, + { + "epoch": 1.37, + "learning_rate": 9.394230769230768e-05, + "loss": 0.0348, + "step": 64290 + }, + { + "epoch": 1.37, + "learning_rate": 9.39102564102564e-05, + "loss": 0.0325, + "step": 64300 + }, + { + "epoch": 1.37, + "learning_rate": 9.387820512820512e-05, + "loss": 0.0389, + "step": 64310 + }, + { + "epoch": 1.37, + "learning_rate": 9.384615384615384e-05, + "loss": 0.0365, + "step": 64320 + }, + { + "epoch": 1.37, + "learning_rate": 9.381410256410255e-05, + "loss": 0.0305, + "step": 64330 + }, + { + "epoch": 1.37, + "learning_rate": 9.378205128205127e-05, + "loss": 0.0322, + "step": 64340 + }, + { + "epoch": 1.37, + "learning_rate": 9.374999999999999e-05, + "loss": 0.0456, + "step": 64350 + }, + { + "epoch": 1.38, + "learning_rate": 9.371794871794871e-05, + "loss": 0.0351, + "step": 64360 + }, + { + "epoch": 1.38, + "learning_rate": 9.368589743589743e-05, + "loss": 0.0392, + "step": 64370 + }, + { + "epoch": 1.38, + "learning_rate": 9.365384615384615e-05, + "loss": 0.0302, + "step": 64380 + }, + { + "epoch": 1.38, + "learning_rate": 9.362179487179487e-05, + "loss": 0.0341, + "step": 64390 + }, + { + "epoch": 1.38, + "learning_rate": 9.358974358974359e-05, + "loss": 0.0289, + "step": 64400 + }, + { + "epoch": 1.38, + "learning_rate": 9.35576923076923e-05, + "loss": 0.0412, + "step": 64410 + }, + { + "epoch": 1.38, + "learning_rate": 9.352564102564101e-05, + "loss": 0.0387, + "step": 64420 + }, + { + "epoch": 1.38, + "learning_rate": 9.349358974358973e-05, + "loss": 0.0316, + "step": 64430 + }, + { + "epoch": 1.38, + "learning_rate": 9.346153846153845e-05, + "loss": 0.0348, + "step": 64440 + }, + { + "epoch": 1.38, + "learning_rate": 9.342948717948717e-05, + "loss": 0.0388, + "step": 64450 + }, + { + "epoch": 1.38, + "learning_rate": 9.339743589743589e-05, + "loss": 0.0285, + "step": 64460 + }, + { + "epoch": 1.38, + "learning_rate": 9.336538461538461e-05, + "loss": 0.0322, + "step": 64470 + }, + { + "epoch": 1.38, + "learning_rate": 9.333333333333333e-05, + "loss": 0.032, + "step": 64480 + }, + { + "epoch": 1.38, + "learning_rate": 9.330128205128205e-05, + "loss": 0.0374, + "step": 64490 + }, + { + "epoch": 1.38, + "learning_rate": 9.326923076923077e-05, + "loss": 0.0354, + "step": 64500 + }, + { + "epoch": 1.38, + "learning_rate": 9.323717948717947e-05, + "loss": 0.0342, + "step": 64510 + }, + { + "epoch": 1.38, + "learning_rate": 9.32051282051282e-05, + "loss": 0.0334, + "step": 64520 + }, + { + "epoch": 1.38, + "learning_rate": 9.317307692307691e-05, + "loss": 0.0383, + "step": 64530 + }, + { + "epoch": 1.38, + "learning_rate": 9.314102564102563e-05, + "loss": 0.0342, + "step": 64540 + }, + { + "epoch": 1.38, + "learning_rate": 9.310897435897435e-05, + "loss": 0.0256, + "step": 64550 + }, + { + "epoch": 1.38, + "learning_rate": 9.307692307692307e-05, + "loss": 0.0331, + "step": 64560 + }, + { + "epoch": 1.38, + "learning_rate": 9.304487179487179e-05, + "loss": 0.0301, + "step": 64570 + }, + { + "epoch": 1.38, + "learning_rate": 9.301282051282051e-05, + "loss": 0.0365, + "step": 64580 + }, + { + "epoch": 1.38, + "learning_rate": 9.298076923076923e-05, + "loss": 0.0308, + "step": 64590 + }, + { + "epoch": 1.38, + "learning_rate": 9.294871794871794e-05, + "loss": 0.0313, + "step": 64600 + }, + { + "epoch": 1.38, + "learning_rate": 9.291666666666666e-05, + "loss": 0.0349, + "step": 64610 + }, + { + "epoch": 1.38, + "learning_rate": 9.288461538461538e-05, + "loss": 0.0344, + "step": 64620 + }, + { + "epoch": 1.38, + "learning_rate": 9.28525641025641e-05, + "loss": 0.0385, + "step": 64630 + }, + { + "epoch": 1.38, + "learning_rate": 9.282051282051282e-05, + "loss": 0.0296, + "step": 64640 + }, + { + "epoch": 1.38, + "learning_rate": 9.278846153846154e-05, + "loss": 0.0311, + "step": 64650 + }, + { + "epoch": 1.38, + "learning_rate": 9.275641025641026e-05, + "loss": 0.0309, + "step": 64660 + }, + { + "epoch": 1.38, + "learning_rate": 9.272435897435898e-05, + "loss": 0.0328, + "step": 64670 + }, + { + "epoch": 1.38, + "learning_rate": 9.26923076923077e-05, + "loss": 0.0343, + "step": 64680 + }, + { + "epoch": 1.38, + "learning_rate": 9.26602564102564e-05, + "loss": 0.0347, + "step": 64690 + }, + { + "epoch": 1.38, + "learning_rate": 9.262820512820512e-05, + "loss": 0.0335, + "step": 64700 + }, + { + "epoch": 1.38, + "learning_rate": 9.259615384615384e-05, + "loss": 0.033, + "step": 64710 + }, + { + "epoch": 1.38, + "learning_rate": 9.256410256410256e-05, + "loss": 0.0378, + "step": 64720 + }, + { + "epoch": 1.38, + "learning_rate": 9.253205128205128e-05, + "loss": 0.028, + "step": 64730 + }, + { + "epoch": 1.38, + "learning_rate": 9.25e-05, + "loss": 0.0406, + "step": 64740 + }, + { + "epoch": 1.38, + "learning_rate": 9.246794871794872e-05, + "loss": 0.0306, + "step": 64750 + }, + { + "epoch": 1.38, + "learning_rate": 9.243589743589744e-05, + "loss": 0.0317, + "step": 64760 + }, + { + "epoch": 1.38, + "learning_rate": 9.240384615384614e-05, + "loss": 0.03, + "step": 64770 + }, + { + "epoch": 1.38, + "learning_rate": 9.237179487179486e-05, + "loss": 0.0402, + "step": 64780 + }, + { + "epoch": 1.38, + "learning_rate": 9.233974358974358e-05, + "loss": 0.0299, + "step": 64790 + }, + { + "epoch": 1.38, + "learning_rate": 9.23076923076923e-05, + "loss": 0.0355, + "step": 64800 + }, + { + "epoch": 1.38, + "learning_rate": 9.227564102564102e-05, + "loss": 0.029, + "step": 64810 + }, + { + "epoch": 1.39, + "learning_rate": 9.224358974358974e-05, + "loss": 0.0258, + "step": 64820 + }, + { + "epoch": 1.39, + "learning_rate": 9.221153846153846e-05, + "loss": 0.0326, + "step": 64830 + }, + { + "epoch": 1.39, + "learning_rate": 9.217948717948718e-05, + "loss": 0.0386, + "step": 64840 + }, + { + "epoch": 1.39, + "learning_rate": 9.21474358974359e-05, + "loss": 0.0361, + "step": 64850 + }, + { + "epoch": 1.39, + "learning_rate": 9.211538461538461e-05, + "loss": 0.036, + "step": 64860 + }, + { + "epoch": 1.39, + "learning_rate": 9.208333333333333e-05, + "loss": 0.0344, + "step": 64870 + }, + { + "epoch": 1.39, + "learning_rate": 9.205128205128205e-05, + "loss": 0.0379, + "step": 64880 + }, + { + "epoch": 1.39, + "learning_rate": 9.201923076923077e-05, + "loss": 0.0334, + "step": 64890 + }, + { + "epoch": 1.39, + "learning_rate": 9.198717948717949e-05, + "loss": 0.0368, + "step": 64900 + }, + { + "epoch": 1.39, + "learning_rate": 9.19551282051282e-05, + "loss": 0.0301, + "step": 64910 + }, + { + "epoch": 1.39, + "learning_rate": 9.192307692307692e-05, + "loss": 0.0319, + "step": 64920 + }, + { + "epoch": 1.39, + "learning_rate": 9.189102564102564e-05, + "loss": 0.0361, + "step": 64930 + }, + { + "epoch": 1.39, + "learning_rate": 9.185897435897436e-05, + "loss": 0.0311, + "step": 64940 + }, + { + "epoch": 1.39, + "learning_rate": 9.182692307692307e-05, + "loss": 0.029, + "step": 64950 + }, + { + "epoch": 1.39, + "learning_rate": 9.179487179487179e-05, + "loss": 0.0359, + "step": 64960 + }, + { + "epoch": 1.39, + "learning_rate": 9.176282051282051e-05, + "loss": 0.0321, + "step": 64970 + }, + { + "epoch": 1.39, + "learning_rate": 9.173076923076923e-05, + "loss": 0.0329, + "step": 64980 + }, + { + "epoch": 1.39, + "learning_rate": 9.169871794871795e-05, + "loss": 0.0386, + "step": 64990 + }, + { + "epoch": 1.39, + "learning_rate": 9.166666666666667e-05, + "loss": 0.0375, + "step": 65000 + }, + { + "epoch": 1.39, + "eval_cer": 0.9236978726548374, + "eval_loss": 0.024814415723085403, + "eval_runtime": 120.5192, + "eval_samples_per_second": 16.595, + "eval_steps_per_second": 4.149, + "step": 65000 + }, + { + "epoch": 1.39, + "learning_rate": 9.163461538461539e-05, + "loss": 0.0266, + "step": 65010 + }, + { + "epoch": 1.39, + "learning_rate": 9.160256410256411e-05, + "loss": 0.0345, + "step": 65020 + }, + { + "epoch": 1.39, + "learning_rate": 9.157051282051283e-05, + "loss": 0.0369, + "step": 65030 + }, + { + "epoch": 1.39, + "learning_rate": 9.153846153846153e-05, + "loss": 0.0345, + "step": 65040 + }, + { + "epoch": 1.39, + "learning_rate": 9.150641025641025e-05, + "loss": 0.0309, + "step": 65050 + }, + { + "epoch": 1.39, + "learning_rate": 9.147435897435897e-05, + "loss": 0.0293, + "step": 65060 + }, + { + "epoch": 1.39, + "learning_rate": 9.144230769230769e-05, + "loss": 0.0332, + "step": 65070 + }, + { + "epoch": 1.39, + "learning_rate": 9.141025641025641e-05, + "loss": 0.0338, + "step": 65080 + }, + { + "epoch": 1.39, + "learning_rate": 9.137820512820512e-05, + "loss": 0.0378, + "step": 65090 + }, + { + "epoch": 1.39, + "learning_rate": 9.134615384615384e-05, + "loss": 0.0377, + "step": 65100 + }, + { + "epoch": 1.39, + "learning_rate": 9.131410256410254e-05, + "loss": 0.0334, + "step": 65110 + }, + { + "epoch": 1.39, + "learning_rate": 9.128205128205126e-05, + "loss": 0.0395, + "step": 65120 + }, + { + "epoch": 1.39, + "learning_rate": 9.124999999999998e-05, + "loss": 0.0352, + "step": 65130 + }, + { + "epoch": 1.39, + "learning_rate": 9.12179487179487e-05, + "loss": 0.0308, + "step": 65140 + }, + { + "epoch": 1.39, + "learning_rate": 9.118589743589742e-05, + "loss": 0.0338, + "step": 65150 + }, + { + "epoch": 1.39, + "learning_rate": 9.115384615384614e-05, + "loss": 0.0324, + "step": 65160 + }, + { + "epoch": 1.39, + "learning_rate": 9.112179487179486e-05, + "loss": 0.033, + "step": 65170 + }, + { + "epoch": 1.39, + "learning_rate": 9.108974358974358e-05, + "loss": 0.0425, + "step": 65180 + }, + { + "epoch": 1.39, + "learning_rate": 9.105769230769229e-05, + "loss": 0.037, + "step": 65190 + }, + { + "epoch": 1.39, + "learning_rate": 9.1025641025641e-05, + "loss": 0.0337, + "step": 65200 + }, + { + "epoch": 1.39, + "learning_rate": 9.099358974358973e-05, + "loss": 0.0422, + "step": 65210 + }, + { + "epoch": 1.39, + "learning_rate": 9.096153846153844e-05, + "loss": 0.0317, + "step": 65220 + }, + { + "epoch": 1.39, + "learning_rate": 9.092948717948716e-05, + "loss": 0.0327, + "step": 65230 + }, + { + "epoch": 1.39, + "learning_rate": 9.089743589743588e-05, + "loss": 0.0373, + "step": 65240 + }, + { + "epoch": 1.39, + "learning_rate": 9.08653846153846e-05, + "loss": 0.0332, + "step": 65250 + }, + { + "epoch": 1.39, + "learning_rate": 9.083333333333332e-05, + "loss": 0.0414, + "step": 65260 + }, + { + "epoch": 1.39, + "learning_rate": 9.080128205128204e-05, + "loss": 0.0322, + "step": 65270 + }, + { + "epoch": 1.39, + "learning_rate": 9.076923076923075e-05, + "loss": 0.0357, + "step": 65280 + }, + { + "epoch": 1.4, + "learning_rate": 9.073717948717947e-05, + "loss": 0.0312, + "step": 65290 + }, + { + "epoch": 1.4, + "learning_rate": 9.070512820512819e-05, + "loss": 0.0338, + "step": 65300 + }, + { + "epoch": 1.4, + "learning_rate": 9.067307692307691e-05, + "loss": 0.0292, + "step": 65310 + }, + { + "epoch": 1.4, + "learning_rate": 9.064102564102563e-05, + "loss": 0.0343, + "step": 65320 + }, + { + "epoch": 1.4, + "learning_rate": 9.060897435897435e-05, + "loss": 0.0314, + "step": 65330 + }, + { + "epoch": 1.4, + "learning_rate": 9.057692307692307e-05, + "loss": 0.0317, + "step": 65340 + }, + { + "epoch": 1.4, + "learning_rate": 9.054487179487179e-05, + "loss": 0.0341, + "step": 65350 + }, + { + "epoch": 1.4, + "learning_rate": 9.05128205128205e-05, + "loss": 0.0365, + "step": 65360 + }, + { + "epoch": 1.4, + "learning_rate": 9.048076923076921e-05, + "loss": 0.0383, + "step": 65370 + }, + { + "epoch": 1.4, + "learning_rate": 9.044871794871793e-05, + "loss": 0.0309, + "step": 65380 + }, + { + "epoch": 1.4, + "learning_rate": 9.041666666666665e-05, + "loss": 0.0344, + "step": 65390 + }, + { + "epoch": 1.4, + "learning_rate": 9.038461538461537e-05, + "loss": 0.0349, + "step": 65400 + }, + { + "epoch": 1.4, + "learning_rate": 9.035256410256409e-05, + "loss": 0.0321, + "step": 65410 + }, + { + "epoch": 1.4, + "learning_rate": 9.032051282051281e-05, + "loss": 0.0307, + "step": 65420 + }, + { + "epoch": 1.4, + "learning_rate": 9.028846153846153e-05, + "loss": 0.0362, + "step": 65430 + }, + { + "epoch": 1.4, + "learning_rate": 9.025641025641025e-05, + "loss": 0.0328, + "step": 65440 + }, + { + "epoch": 1.4, + "learning_rate": 9.022435897435897e-05, + "loss": 0.0324, + "step": 65450 + }, + { + "epoch": 1.4, + "learning_rate": 9.019230769230767e-05, + "loss": 0.0363, + "step": 65460 + }, + { + "epoch": 1.4, + "learning_rate": 9.01602564102564e-05, + "loss": 0.0297, + "step": 65470 + }, + { + "epoch": 1.4, + "learning_rate": 9.012820512820511e-05, + "loss": 0.0365, + "step": 65480 + }, + { + "epoch": 1.4, + "learning_rate": 9.009615384615383e-05, + "loss": 0.0388, + "step": 65490 + }, + { + "epoch": 1.4, + "learning_rate": 9.006410256410255e-05, + "loss": 0.0323, + "step": 65500 + }, + { + "epoch": 1.4, + "learning_rate": 9.003205128205127e-05, + "loss": 0.0297, + "step": 65510 + }, + { + "epoch": 1.4, + "learning_rate": 8.999999999999999e-05, + "loss": 0.0341, + "step": 65520 + }, + { + "epoch": 1.4, + "learning_rate": 8.996794871794871e-05, + "loss": 0.0331, + "step": 65530 + }, + { + "epoch": 1.4, + "learning_rate": 8.993589743589743e-05, + "loss": 0.036, + "step": 65540 + }, + { + "epoch": 1.4, + "learning_rate": 8.990384615384614e-05, + "loss": 0.0352, + "step": 65550 + }, + { + "epoch": 1.4, + "learning_rate": 8.987179487179486e-05, + "loss": 0.0361, + "step": 65560 + }, + { + "epoch": 1.4, + "learning_rate": 8.983974358974358e-05, + "loss": 0.0344, + "step": 65570 + }, + { + "epoch": 1.4, + "learning_rate": 8.98076923076923e-05, + "loss": 0.0392, + "step": 65580 + }, + { + "epoch": 1.4, + "learning_rate": 8.977564102564102e-05, + "loss": 0.0411, + "step": 65590 + }, + { + "epoch": 1.4, + "learning_rate": 8.974358974358974e-05, + "loss": 0.0334, + "step": 65600 + }, + { + "epoch": 1.4, + "learning_rate": 8.971153846153845e-05, + "loss": 0.0343, + "step": 65610 + }, + { + "epoch": 1.4, + "learning_rate": 8.967948717948717e-05, + "loss": 0.0367, + "step": 65620 + }, + { + "epoch": 1.4, + "learning_rate": 8.96474358974359e-05, + "loss": 0.0334, + "step": 65630 + }, + { + "epoch": 1.4, + "learning_rate": 8.96153846153846e-05, + "loss": 0.0293, + "step": 65640 + }, + { + "epoch": 1.4, + "learning_rate": 8.958333333333332e-05, + "loss": 0.0335, + "step": 65650 + }, + { + "epoch": 1.4, + "learning_rate": 8.955128205128204e-05, + "loss": 0.0316, + "step": 65660 + }, + { + "epoch": 1.4, + "learning_rate": 8.951923076923076e-05, + "loss": 0.0361, + "step": 65670 + }, + { + "epoch": 1.4, + "learning_rate": 8.948717948717948e-05, + "loss": 0.0288, + "step": 65680 + }, + { + "epoch": 1.4, + "learning_rate": 8.94551282051282e-05, + "loss": 0.0318, + "step": 65690 + }, + { + "epoch": 1.4, + "learning_rate": 8.942307692307692e-05, + "loss": 0.036, + "step": 65700 + }, + { + "epoch": 1.4, + "learning_rate": 8.939102564102564e-05, + "loss": 0.0353, + "step": 65710 + }, + { + "epoch": 1.4, + "learning_rate": 8.935897435897434e-05, + "loss": 0.0315, + "step": 65720 + }, + { + "epoch": 1.4, + "learning_rate": 8.932692307692306e-05, + "loss": 0.03, + "step": 65730 + }, + { + "epoch": 1.4, + "learning_rate": 8.929487179487178e-05, + "loss": 0.0321, + "step": 65740 + }, + { + "epoch": 1.4, + "learning_rate": 8.92628205128205e-05, + "loss": 0.0367, + "step": 65750 + }, + { + "epoch": 1.41, + "learning_rate": 8.923076923076922e-05, + "loss": 0.0301, + "step": 65760 + }, + { + "epoch": 1.41, + "learning_rate": 8.919871794871794e-05, + "loss": 0.0329, + "step": 65770 + }, + { + "epoch": 1.41, + "learning_rate": 8.916666666666666e-05, + "loss": 0.039, + "step": 65780 + }, + { + "epoch": 1.41, + "learning_rate": 8.913461538461538e-05, + "loss": 0.0335, + "step": 65790 + }, + { + "epoch": 1.41, + "learning_rate": 8.91025641025641e-05, + "loss": 0.0251, + "step": 65800 + }, + { + "epoch": 1.41, + "learning_rate": 8.90705128205128e-05, + "loss": 0.0331, + "step": 65810 + }, + { + "epoch": 1.41, + "learning_rate": 8.903846153846153e-05, + "loss": 0.0316, + "step": 65820 + }, + { + "epoch": 1.41, + "learning_rate": 8.900641025641025e-05, + "loss": 0.0341, + "step": 65830 + }, + { + "epoch": 1.41, + "learning_rate": 8.897435897435897e-05, + "loss": 0.0307, + "step": 65840 + }, + { + "epoch": 1.41, + "learning_rate": 8.894230769230768e-05, + "loss": 0.0318, + "step": 65850 + }, + { + "epoch": 1.41, + "learning_rate": 8.89102564102564e-05, + "loss": 0.0357, + "step": 65860 + }, + { + "epoch": 1.41, + "learning_rate": 8.887820512820512e-05, + "loss": 0.0387, + "step": 65870 + }, + { + "epoch": 1.41, + "learning_rate": 8.884615384615384e-05, + "loss": 0.0298, + "step": 65880 + }, + { + "epoch": 1.41, + "learning_rate": 8.881410256410256e-05, + "loss": 0.0328, + "step": 65890 + }, + { + "epoch": 1.41, + "learning_rate": 8.878205128205127e-05, + "loss": 0.027, + "step": 65900 + }, + { + "epoch": 1.41, + "learning_rate": 8.874999999999999e-05, + "loss": 0.0368, + "step": 65910 + }, + { + "epoch": 1.41, + "learning_rate": 8.871794871794871e-05, + "loss": 0.0285, + "step": 65920 + }, + { + "epoch": 1.41, + "learning_rate": 8.868589743589743e-05, + "loss": 0.0335, + "step": 65930 + }, + { + "epoch": 1.41, + "learning_rate": 8.865384615384615e-05, + "loss": 0.0352, + "step": 65940 + }, + { + "epoch": 1.41, + "learning_rate": 8.862179487179487e-05, + "loss": 0.0274, + "step": 65950 + }, + { + "epoch": 1.41, + "learning_rate": 8.858974358974359e-05, + "loss": 0.0363, + "step": 65960 + }, + { + "epoch": 1.41, + "learning_rate": 8.85576923076923e-05, + "loss": 0.0298, + "step": 65970 + }, + { + "epoch": 1.41, + "learning_rate": 8.852564102564103e-05, + "loss": 0.0333, + "step": 65980 + }, + { + "epoch": 1.41, + "learning_rate": 8.849358974358973e-05, + "loss": 0.0408, + "step": 65990 + }, + { + "epoch": 1.41, + "learning_rate": 8.846153846153845e-05, + "loss": 0.0336, + "step": 66000 + }, + { + "epoch": 1.41, + "eval_cer": 0.923625684408273, + "eval_loss": 0.02462809719145298, + "eval_runtime": 120.2399, + "eval_samples_per_second": 16.633, + "eval_steps_per_second": 4.158, + "step": 66000 + }, + { + "epoch": 1.41, + "learning_rate": 8.842948717948717e-05, + "loss": 0.0335, + "step": 66010 + }, + { + "epoch": 1.41, + "learning_rate": 8.839743589743589e-05, + "loss": 0.0378, + "step": 66020 + }, + { + "epoch": 1.41, + "learning_rate": 8.836538461538461e-05, + "loss": 0.0312, + "step": 66030 + }, + { + "epoch": 1.41, + "learning_rate": 8.833333333333333e-05, + "loss": 0.0305, + "step": 66040 + }, + { + "epoch": 1.41, + "learning_rate": 8.830128205128205e-05, + "loss": 0.0313, + "step": 66050 + }, + { + "epoch": 1.41, + "learning_rate": 8.826923076923077e-05, + "loss": 0.0327, + "step": 66060 + }, + { + "epoch": 1.41, + "learning_rate": 8.823717948717949e-05, + "loss": 0.0303, + "step": 66070 + }, + { + "epoch": 1.41, + "learning_rate": 8.82051282051282e-05, + "loss": 0.0363, + "step": 66080 + }, + { + "epoch": 1.41, + "learning_rate": 8.817307692307691e-05, + "loss": 0.0322, + "step": 66090 + }, + { + "epoch": 1.41, + "learning_rate": 8.814102564102563e-05, + "loss": 0.0301, + "step": 66100 + }, + { + "epoch": 1.41, + "learning_rate": 8.810897435897435e-05, + "loss": 0.0357, + "step": 66110 + }, + { + "epoch": 1.41, + "learning_rate": 8.807692307692307e-05, + "loss": 0.0379, + "step": 66120 + }, + { + "epoch": 1.41, + "learning_rate": 8.804487179487179e-05, + "loss": 0.0382, + "step": 66130 + }, + { + "epoch": 1.41, + "learning_rate": 8.801282051282051e-05, + "loss": 0.0334, + "step": 66140 + }, + { + "epoch": 1.41, + "learning_rate": 8.798076923076923e-05, + "loss": 0.0332, + "step": 66150 + }, + { + "epoch": 1.41, + "learning_rate": 8.794871794871794e-05, + "loss": 0.0308, + "step": 66160 + }, + { + "epoch": 1.41, + "learning_rate": 8.791666666666666e-05, + "loss": 0.0344, + "step": 66170 + }, + { + "epoch": 1.41, + "learning_rate": 8.788461538461538e-05, + "loss": 0.0292, + "step": 66180 + }, + { + "epoch": 1.41, + "learning_rate": 8.78525641025641e-05, + "loss": 0.0329, + "step": 66190 + }, + { + "epoch": 1.41, + "learning_rate": 8.782051282051282e-05, + "loss": 0.0353, + "step": 66200 + }, + { + "epoch": 1.41, + "learning_rate": 8.778846153846154e-05, + "loss": 0.0326, + "step": 66210 + }, + { + "epoch": 1.41, + "learning_rate": 8.775641025641026e-05, + "loss": 0.0301, + "step": 66220 + }, + { + "epoch": 1.42, + "learning_rate": 8.772435897435898e-05, + "loss": 0.031, + "step": 66230 + }, + { + "epoch": 1.42, + "learning_rate": 8.76923076923077e-05, + "loss": 0.0336, + "step": 66240 + }, + { + "epoch": 1.42, + "learning_rate": 8.76602564102564e-05, + "loss": 0.0331, + "step": 66250 + }, + { + "epoch": 1.42, + "learning_rate": 8.762820512820512e-05, + "loss": 0.0324, + "step": 66260 + }, + { + "epoch": 1.42, + "learning_rate": 8.759615384615384e-05, + "loss": 0.0396, + "step": 66270 + }, + { + "epoch": 1.42, + "learning_rate": 8.756410256410256e-05, + "loss": 0.0305, + "step": 66280 + }, + { + "epoch": 1.42, + "learning_rate": 8.753205128205128e-05, + "loss": 0.0292, + "step": 66290 + }, + { + "epoch": 1.42, + "learning_rate": 8.75e-05, + "loss": 0.0313, + "step": 66300 + }, + { + "epoch": 1.42, + "learning_rate": 8.746794871794872e-05, + "loss": 0.0355, + "step": 66310 + }, + { + "epoch": 1.42, + "learning_rate": 8.743589743589744e-05, + "loss": 0.0336, + "step": 66320 + }, + { + "epoch": 1.42, + "learning_rate": 8.740384615384616e-05, + "loss": 0.0316, + "step": 66330 + }, + { + "epoch": 1.42, + "learning_rate": 8.737179487179486e-05, + "loss": 0.032, + "step": 66340 + }, + { + "epoch": 1.42, + "learning_rate": 8.733974358974358e-05, + "loss": 0.0257, + "step": 66350 + }, + { + "epoch": 1.42, + "learning_rate": 8.73076923076923e-05, + "loss": 0.0381, + "step": 66360 + }, + { + "epoch": 1.42, + "learning_rate": 8.727564102564102e-05, + "loss": 0.0268, + "step": 66370 + }, + { + "epoch": 1.42, + "learning_rate": 8.724358974358974e-05, + "loss": 0.0293, + "step": 66380 + }, + { + "epoch": 1.42, + "learning_rate": 8.721153846153846e-05, + "loss": 0.0282, + "step": 66390 + }, + { + "epoch": 1.42, + "learning_rate": 8.717948717948718e-05, + "loss": 0.0331, + "step": 66400 + }, + { + "epoch": 1.42, + "learning_rate": 8.71474358974359e-05, + "loss": 0.0305, + "step": 66410 + }, + { + "epoch": 1.42, + "learning_rate": 8.711538461538462e-05, + "loss": 0.0309, + "step": 66420 + }, + { + "epoch": 1.42, + "learning_rate": 8.708333333333333e-05, + "loss": 0.0313, + "step": 66430 + }, + { + "epoch": 1.42, + "learning_rate": 8.705128205128205e-05, + "loss": 0.0351, + "step": 66440 + }, + { + "epoch": 1.42, + "learning_rate": 8.701923076923077e-05, + "loss": 0.0277, + "step": 66450 + }, + { + "epoch": 1.42, + "learning_rate": 8.698717948717949e-05, + "loss": 0.0332, + "step": 66460 + }, + { + "epoch": 1.42, + "learning_rate": 8.69551282051282e-05, + "loss": 0.0308, + "step": 66470 + }, + { + "epoch": 1.42, + "learning_rate": 8.692307692307692e-05, + "loss": 0.0308, + "step": 66480 + }, + { + "epoch": 1.42, + "learning_rate": 8.689102564102564e-05, + "loss": 0.0335, + "step": 66490 + }, + { + "epoch": 1.42, + "learning_rate": 8.685897435897436e-05, + "loss": 0.0325, + "step": 66500 + }, + { + "epoch": 1.42, + "learning_rate": 8.682692307692308e-05, + "loss": 0.0317, + "step": 66510 + }, + { + "epoch": 1.42, + "learning_rate": 8.679487179487179e-05, + "loss": 0.0375, + "step": 66520 + }, + { + "epoch": 1.42, + "learning_rate": 8.676282051282051e-05, + "loss": 0.0322, + "step": 66530 + }, + { + "epoch": 1.42, + "learning_rate": 8.673076923076923e-05, + "loss": 0.039, + "step": 66540 + }, + { + "epoch": 1.42, + "learning_rate": 8.669871794871793e-05, + "loss": 0.0295, + "step": 66550 + }, + { + "epoch": 1.42, + "learning_rate": 8.666666666666665e-05, + "loss": 0.0331, + "step": 66560 + }, + { + "epoch": 1.42, + "learning_rate": 8.663461538461537e-05, + "loss": 0.0317, + "step": 66570 + }, + { + "epoch": 1.42, + "learning_rate": 8.66025641025641e-05, + "loss": 0.0307, + "step": 66580 + }, + { + "epoch": 1.42, + "learning_rate": 8.65705128205128e-05, + "loss": 0.0342, + "step": 66590 + }, + { + "epoch": 1.42, + "learning_rate": 8.653846153846152e-05, + "loss": 0.0386, + "step": 66600 + }, + { + "epoch": 1.42, + "learning_rate": 8.650641025641024e-05, + "loss": 0.0304, + "step": 66610 + }, + { + "epoch": 1.42, + "learning_rate": 8.647435897435896e-05, + "loss": 0.0304, + "step": 66620 + }, + { + "epoch": 1.42, + "learning_rate": 8.644230769230768e-05, + "loss": 0.0349, + "step": 66630 + }, + { + "epoch": 1.42, + "learning_rate": 8.64102564102564e-05, + "loss": 0.0335, + "step": 66640 + }, + { + "epoch": 1.42, + "learning_rate": 8.637820512820512e-05, + "loss": 0.0325, + "step": 66650 + }, + { + "epoch": 1.42, + "learning_rate": 8.634615384615384e-05, + "loss": 0.0289, + "step": 66660 + }, + { + "epoch": 1.42, + "learning_rate": 8.631410256410254e-05, + "loss": 0.035, + "step": 66670 + }, + { + "epoch": 1.42, + "learning_rate": 8.628205128205126e-05, + "loss": 0.0385, + "step": 66680 + }, + { + "epoch": 1.42, + "learning_rate": 8.624999999999998e-05, + "loss": 0.0352, + "step": 66690 + }, + { + "epoch": 1.43, + "learning_rate": 8.62179487179487e-05, + "loss": 0.0293, + "step": 66700 + }, + { + "epoch": 1.43, + "learning_rate": 8.618589743589742e-05, + "loss": 0.0343, + "step": 66710 + }, + { + "epoch": 1.43, + "learning_rate": 8.615384615384614e-05, + "loss": 0.0291, + "step": 66720 + }, + { + "epoch": 1.43, + "learning_rate": 8.612179487179486e-05, + "loss": 0.0346, + "step": 66730 + }, + { + "epoch": 1.43, + "learning_rate": 8.608974358974358e-05, + "loss": 0.0285, + "step": 66740 + }, + { + "epoch": 1.43, + "learning_rate": 8.60576923076923e-05, + "loss": 0.0362, + "step": 66750 + }, + { + "epoch": 1.43, + "learning_rate": 8.6025641025641e-05, + "loss": 0.0353, + "step": 66760 + }, + { + "epoch": 1.43, + "learning_rate": 8.599358974358973e-05, + "loss": 0.0344, + "step": 66770 + }, + { + "epoch": 1.43, + "learning_rate": 8.596153846153845e-05, + "loss": 0.0321, + "step": 66780 + }, + { + "epoch": 1.43, + "learning_rate": 8.592948717948716e-05, + "loss": 0.0347, + "step": 66790 + }, + { + "epoch": 1.43, + "learning_rate": 8.589743589743588e-05, + "loss": 0.0307, + "step": 66800 + }, + { + "epoch": 1.43, + "learning_rate": 8.58653846153846e-05, + "loss": 0.0312, + "step": 66810 + }, + { + "epoch": 1.43, + "learning_rate": 8.583333333333332e-05, + "loss": 0.0339, + "step": 66820 + }, + { + "epoch": 1.43, + "learning_rate": 8.580128205128204e-05, + "loss": 0.0322, + "step": 66830 + }, + { + "epoch": 1.43, + "learning_rate": 8.576923076923076e-05, + "loss": 0.0307, + "step": 66840 + }, + { + "epoch": 1.43, + "learning_rate": 8.573717948717947e-05, + "loss": 0.0302, + "step": 66850 + }, + { + "epoch": 1.43, + "learning_rate": 8.570512820512819e-05, + "loss": 0.033, + "step": 66860 + }, + { + "epoch": 1.43, + "learning_rate": 8.567307692307691e-05, + "loss": 0.033, + "step": 66870 + }, + { + "epoch": 1.43, + "learning_rate": 8.564102564102563e-05, + "loss": 0.0373, + "step": 66880 + }, + { + "epoch": 1.43, + "learning_rate": 8.560897435897435e-05, + "loss": 0.0321, + "step": 66890 + }, + { + "epoch": 1.43, + "learning_rate": 8.557692307692307e-05, + "loss": 0.029, + "step": 66900 + }, + { + "epoch": 1.43, + "learning_rate": 8.554487179487179e-05, + "loss": 0.0274, + "step": 66910 + }, + { + "epoch": 1.43, + "learning_rate": 8.55128205128205e-05, + "loss": 0.038, + "step": 66920 + }, + { + "epoch": 1.43, + "learning_rate": 8.548076923076923e-05, + "loss": 0.0307, + "step": 66930 + }, + { + "epoch": 1.43, + "learning_rate": 8.544871794871793e-05, + "loss": 0.032, + "step": 66940 + }, + { + "epoch": 1.43, + "learning_rate": 8.541666666666665e-05, + "loss": 0.0339, + "step": 66950 + }, + { + "epoch": 1.43, + "learning_rate": 8.538461538461537e-05, + "loss": 0.0319, + "step": 66960 + }, + { + "epoch": 1.43, + "learning_rate": 8.535256410256409e-05, + "loss": 0.0342, + "step": 66970 + }, + { + "epoch": 1.43, + "learning_rate": 8.532051282051281e-05, + "loss": 0.0331, + "step": 66980 + }, + { + "epoch": 1.43, + "learning_rate": 8.528846153846153e-05, + "loss": 0.0301, + "step": 66990 + }, + { + "epoch": 1.43, + "learning_rate": 8.525641025641025e-05, + "loss": 0.031, + "step": 67000 + }, + { + "epoch": 1.43, + "eval_cer": 0.9236622031682997, + "eval_loss": 0.024645386263728142, + "eval_runtime": 120.4862, + "eval_samples_per_second": 16.599, + "eval_steps_per_second": 4.15, + "step": 67000 + }, + { + "epoch": 1.43, + "learning_rate": 8.522435897435897e-05, + "loss": 0.0326, + "step": 67010 + }, + { + "epoch": 1.43, + "learning_rate": 8.519230769230769e-05, + "loss": 0.0287, + "step": 67020 + }, + { + "epoch": 1.43, + "learning_rate": 8.51602564102564e-05, + "loss": 0.0316, + "step": 67030 + }, + { + "epoch": 1.43, + "learning_rate": 8.512820512820511e-05, + "loss": 0.0317, + "step": 67040 + }, + { + "epoch": 1.43, + "learning_rate": 8.509615384615383e-05, + "loss": 0.0308, + "step": 67050 + }, + { + "epoch": 1.43, + "learning_rate": 8.506410256410255e-05, + "loss": 0.0351, + "step": 67060 + }, + { + "epoch": 1.43, + "learning_rate": 8.503205128205127e-05, + "loss": 0.0311, + "step": 67070 + }, + { + "epoch": 1.43, + "learning_rate": 8.499999999999999e-05, + "loss": 0.0324, + "step": 67080 + }, + { + "epoch": 1.43, + "learning_rate": 8.496794871794871e-05, + "loss": 0.0299, + "step": 67090 + }, + { + "epoch": 1.43, + "learning_rate": 8.493589743589743e-05, + "loss": 0.0346, + "step": 67100 + }, + { + "epoch": 1.43, + "learning_rate": 8.490384615384614e-05, + "loss": 0.0347, + "step": 67110 + }, + { + "epoch": 1.43, + "learning_rate": 8.487179487179486e-05, + "loss": 0.0321, + "step": 67120 + }, + { + "epoch": 1.43, + "learning_rate": 8.483974358974358e-05, + "loss": 0.0303, + "step": 67130 + }, + { + "epoch": 1.43, + "learning_rate": 8.48076923076923e-05, + "loss": 0.0328, + "step": 67140 + }, + { + "epoch": 1.43, + "learning_rate": 8.477564102564102e-05, + "loss": 0.0299, + "step": 67150 + }, + { + "epoch": 1.44, + "learning_rate": 8.474358974358974e-05, + "loss": 0.0384, + "step": 67160 + }, + { + "epoch": 1.44, + "learning_rate": 8.471153846153846e-05, + "loss": 0.0316, + "step": 67170 + }, + { + "epoch": 1.44, + "learning_rate": 8.467948717948718e-05, + "loss": 0.0328, + "step": 67180 + }, + { + "epoch": 1.44, + "learning_rate": 8.46474358974359e-05, + "loss": 0.0299, + "step": 67190 + }, + { + "epoch": 1.44, + "learning_rate": 8.46153846153846e-05, + "loss": 0.0324, + "step": 67200 + }, + { + "epoch": 1.44, + "learning_rate": 8.458333333333332e-05, + "loss": 0.0296, + "step": 67210 + }, + { + "epoch": 1.44, + "learning_rate": 8.455128205128204e-05, + "loss": 0.0325, + "step": 67220 + }, + { + "epoch": 1.44, + "learning_rate": 8.451923076923076e-05, + "loss": 0.0326, + "step": 67230 + }, + { + "epoch": 1.44, + "learning_rate": 8.448717948717948e-05, + "loss": 0.0281, + "step": 67240 + }, + { + "epoch": 1.44, + "learning_rate": 8.44551282051282e-05, + "loss": 0.0313, + "step": 67250 + }, + { + "epoch": 1.44, + "learning_rate": 8.442307692307692e-05, + "loss": 0.0339, + "step": 67260 + }, + { + "epoch": 1.44, + "learning_rate": 8.439102564102564e-05, + "loss": 0.0302, + "step": 67270 + }, + { + "epoch": 1.44, + "learning_rate": 8.435897435897436e-05, + "loss": 0.0373, + "step": 67280 + }, + { + "epoch": 1.44, + "learning_rate": 8.432692307692306e-05, + "loss": 0.0289, + "step": 67290 + }, + { + "epoch": 1.44, + "learning_rate": 8.429487179487178e-05, + "loss": 0.0301, + "step": 67300 + }, + { + "epoch": 1.44, + "learning_rate": 8.42628205128205e-05, + "loss": 0.0338, + "step": 67310 + }, + { + "epoch": 1.44, + "learning_rate": 8.423076923076922e-05, + "loss": 0.0291, + "step": 67320 + }, + { + "epoch": 1.44, + "learning_rate": 8.419871794871794e-05, + "loss": 0.0353, + "step": 67330 + }, + { + "epoch": 1.44, + "learning_rate": 8.416666666666666e-05, + "loss": 0.029, + "step": 67340 + }, + { + "epoch": 1.44, + "learning_rate": 8.413461538461538e-05, + "loss": 0.0322, + "step": 67350 + }, + { + "epoch": 1.44, + "learning_rate": 8.41025641025641e-05, + "loss": 0.0341, + "step": 67360 + }, + { + "epoch": 1.44, + "learning_rate": 8.407051282051282e-05, + "loss": 0.0323, + "step": 67370 + }, + { + "epoch": 1.44, + "learning_rate": 8.403846153846153e-05, + "loss": 0.0309, + "step": 67380 + }, + { + "epoch": 1.44, + "learning_rate": 8.400641025641025e-05, + "loss": 0.0334, + "step": 67390 + }, + { + "epoch": 1.44, + "learning_rate": 8.397435897435897e-05, + "loss": 0.0328, + "step": 67400 + }, + { + "epoch": 1.44, + "learning_rate": 8.394230769230769e-05, + "loss": 0.0306, + "step": 67410 + }, + { + "epoch": 1.44, + "learning_rate": 8.39102564102564e-05, + "loss": 0.0339, + "step": 67420 + }, + { + "epoch": 1.44, + "learning_rate": 8.387820512820512e-05, + "loss": 0.0335, + "step": 67430 + }, + { + "epoch": 1.44, + "learning_rate": 8.384615384615384e-05, + "loss": 0.0321, + "step": 67440 + }, + { + "epoch": 1.44, + "learning_rate": 8.381410256410256e-05, + "loss": 0.0382, + "step": 67450 + }, + { + "epoch": 1.44, + "learning_rate": 8.378205128205128e-05, + "loss": 0.0327, + "step": 67460 + }, + { + "epoch": 1.44, + "learning_rate": 8.374999999999999e-05, + "loss": 0.0363, + "step": 67470 + }, + { + "epoch": 1.44, + "learning_rate": 8.371794871794871e-05, + "loss": 0.0334, + "step": 67480 + }, + { + "epoch": 1.44, + "learning_rate": 8.368589743589743e-05, + "loss": 0.0307, + "step": 67490 + }, + { + "epoch": 1.44, + "learning_rate": 8.365384615384615e-05, + "loss": 0.0385, + "step": 67500 + }, + { + "epoch": 1.44, + "learning_rate": 8.362179487179487e-05, + "loss": 0.0376, + "step": 67510 + }, + { + "epoch": 1.44, + "learning_rate": 8.358974358974359e-05, + "loss": 0.0322, + "step": 67520 + }, + { + "epoch": 1.44, + "learning_rate": 8.355769230769231e-05, + "loss": 0.0357, + "step": 67530 + }, + { + "epoch": 1.44, + "learning_rate": 8.352564102564103e-05, + "loss": 0.0274, + "step": 67540 + }, + { + "epoch": 1.44, + "learning_rate": 8.349358974358975e-05, + "loss": 0.0293, + "step": 67550 + }, + { + "epoch": 1.44, + "learning_rate": 8.346153846153845e-05, + "loss": 0.0328, + "step": 67560 + }, + { + "epoch": 1.44, + "learning_rate": 8.342948717948717e-05, + "loss": 0.0331, + "step": 67570 + }, + { + "epoch": 1.44, + "learning_rate": 8.339743589743589e-05, + "loss": 0.037, + "step": 67580 + }, + { + "epoch": 1.44, + "learning_rate": 8.336538461538461e-05, + "loss": 0.032, + "step": 67590 + }, + { + "epoch": 1.44, + "learning_rate": 8.333333333333333e-05, + "loss": 0.0284, + "step": 67600 + }, + { + "epoch": 1.44, + "learning_rate": 8.330128205128205e-05, + "loss": 0.0335, + "step": 67610 + }, + { + "epoch": 1.44, + "learning_rate": 8.326923076923077e-05, + "loss": 0.0337, + "step": 67620 + }, + { + "epoch": 1.45, + "learning_rate": 8.323717948717949e-05, + "loss": 0.0324, + "step": 67630 + }, + { + "epoch": 1.45, + "learning_rate": 8.32051282051282e-05, + "loss": 0.0364, + "step": 67640 + }, + { + "epoch": 1.45, + "learning_rate": 8.317307692307692e-05, + "loss": 0.0342, + "step": 67650 + }, + { + "epoch": 1.45, + "learning_rate": 8.314102564102563e-05, + "loss": 0.0351, + "step": 67660 + }, + { + "epoch": 1.45, + "learning_rate": 8.310897435897435e-05, + "loss": 0.0343, + "step": 67670 + }, + { + "epoch": 1.45, + "learning_rate": 8.307692307692307e-05, + "loss": 0.0315, + "step": 67680 + }, + { + "epoch": 1.45, + "learning_rate": 8.30448717948718e-05, + "loss": 0.032, + "step": 67690 + }, + { + "epoch": 1.45, + "learning_rate": 8.301282051282051e-05, + "loss": 0.0324, + "step": 67700 + }, + { + "epoch": 1.45, + "learning_rate": 8.298076923076923e-05, + "loss": 0.0299, + "step": 67710 + }, + { + "epoch": 1.45, + "learning_rate": 8.294871794871795e-05, + "loss": 0.0326, + "step": 67720 + }, + { + "epoch": 1.45, + "learning_rate": 8.291666666666666e-05, + "loss": 0.0259, + "step": 67730 + }, + { + "epoch": 1.45, + "learning_rate": 8.288461538461538e-05, + "loss": 0.0326, + "step": 67740 + }, + { + "epoch": 1.45, + "learning_rate": 8.28525641025641e-05, + "loss": 0.0291, + "step": 67750 + }, + { + "epoch": 1.45, + "learning_rate": 8.282051282051282e-05, + "loss": 0.0292, + "step": 67760 + }, + { + "epoch": 1.45, + "learning_rate": 8.278846153846154e-05, + "loss": 0.0326, + "step": 67770 + }, + { + "epoch": 1.45, + "learning_rate": 8.275641025641026e-05, + "loss": 0.0324, + "step": 67780 + }, + { + "epoch": 1.45, + "learning_rate": 8.272435897435898e-05, + "loss": 0.0296, + "step": 67790 + }, + { + "epoch": 1.45, + "learning_rate": 8.26923076923077e-05, + "loss": 0.0329, + "step": 67800 + }, + { + "epoch": 1.45, + "learning_rate": 8.266025641025642e-05, + "loss": 0.0326, + "step": 67810 + }, + { + "epoch": 1.45, + "learning_rate": 8.262820512820512e-05, + "loss": 0.0303, + "step": 67820 + }, + { + "epoch": 1.45, + "learning_rate": 8.259615384615384e-05, + "loss": 0.0337, + "step": 67830 + }, + { + "epoch": 1.45, + "learning_rate": 8.256410256410256e-05, + "loss": 0.0325, + "step": 67840 + }, + { + "epoch": 1.45, + "learning_rate": 8.253205128205128e-05, + "loss": 0.0342, + "step": 67850 + }, + { + "epoch": 1.45, + "learning_rate": 8.25e-05, + "loss": 0.0301, + "step": 67860 + }, + { + "epoch": 1.45, + "learning_rate": 8.246794871794872e-05, + "loss": 0.0349, + "step": 67870 + }, + { + "epoch": 1.45, + "learning_rate": 8.243589743589744e-05, + "loss": 0.0307, + "step": 67880 + }, + { + "epoch": 1.45, + "learning_rate": 8.240384615384616e-05, + "loss": 0.0346, + "step": 67890 + }, + { + "epoch": 1.45, + "learning_rate": 8.237179487179488e-05, + "loss": 0.0282, + "step": 67900 + }, + { + "epoch": 1.45, + "learning_rate": 8.233974358974358e-05, + "loss": 0.0304, + "step": 67910 + }, + { + "epoch": 1.45, + "learning_rate": 8.23076923076923e-05, + "loss": 0.0361, + "step": 67920 + }, + { + "epoch": 1.45, + "learning_rate": 8.227564102564102e-05, + "loss": 0.0263, + "step": 67930 + }, + { + "epoch": 1.45, + "learning_rate": 8.224358974358974e-05, + "loss": 0.0342, + "step": 67940 + }, + { + "epoch": 1.45, + "learning_rate": 8.221153846153846e-05, + "loss": 0.0336, + "step": 67950 + }, + { + "epoch": 1.45, + "learning_rate": 8.217948717948718e-05, + "loss": 0.0335, + "step": 67960 + }, + { + "epoch": 1.45, + "learning_rate": 8.21474358974359e-05, + "loss": 0.0404, + "step": 67970 + }, + { + "epoch": 1.45, + "learning_rate": 8.211538461538462e-05, + "loss": 0.0292, + "step": 67980 + }, + { + "epoch": 1.45, + "learning_rate": 8.208333333333334e-05, + "loss": 0.0328, + "step": 67990 + }, + { + "epoch": 1.45, + "learning_rate": 8.205128205128205e-05, + "loss": 0.0344, + "step": 68000 + }, + { + "epoch": 1.45, + "eval_cer": 0.9235815221868453, + "eval_loss": 0.024751950055360794, + "eval_runtime": 120.4637, + "eval_samples_per_second": 16.603, + "eval_steps_per_second": 4.151, + "step": 68000 + }, + { + "epoch": 1.45, + "learning_rate": 8.201923076923075e-05, + "loss": 0.0268, + "step": 68010 + }, + { + "epoch": 1.45, + "learning_rate": 8.198717948717947e-05, + "loss": 0.0261, + "step": 68020 + }, + { + "epoch": 1.45, + "learning_rate": 8.195512820512819e-05, + "loss": 0.0359, + "step": 68030 + }, + { + "epoch": 1.45, + "learning_rate": 8.192307692307691e-05, + "loss": 0.0313, + "step": 68040 + }, + { + "epoch": 1.45, + "learning_rate": 8.189102564102563e-05, + "loss": 0.0317, + "step": 68050 + }, + { + "epoch": 1.45, + "learning_rate": 8.185897435897434e-05, + "loss": 0.0311, + "step": 68060 + }, + { + "epoch": 1.45, + "learning_rate": 8.182692307692306e-05, + "loss": 0.0326, + "step": 68070 + }, + { + "epoch": 1.45, + "learning_rate": 8.179487179487178e-05, + "loss": 0.0304, + "step": 68080 + }, + { + "epoch": 1.45, + "learning_rate": 8.17628205128205e-05, + "loss": 0.0382, + "step": 68090 + }, + { + "epoch": 1.46, + "learning_rate": 8.173076923076922e-05, + "loss": 0.032, + "step": 68100 + }, + { + "epoch": 1.46, + "learning_rate": 8.169871794871794e-05, + "loss": 0.033, + "step": 68110 + }, + { + "epoch": 1.46, + "learning_rate": 8.166666666666665e-05, + "loss": 0.0313, + "step": 68120 + }, + { + "epoch": 1.46, + "learning_rate": 8.163461538461537e-05, + "loss": 0.0311, + "step": 68130 + }, + { + "epoch": 1.46, + "learning_rate": 8.16025641025641e-05, + "loss": 0.032, + "step": 68140 + }, + { + "epoch": 1.46, + "learning_rate": 8.15705128205128e-05, + "loss": 0.0362, + "step": 68150 + }, + { + "epoch": 1.46, + "learning_rate": 8.153846153846152e-05, + "loss": 0.0374, + "step": 68160 + }, + { + "epoch": 1.46, + "learning_rate": 8.150641025641024e-05, + "loss": 0.0322, + "step": 68170 + }, + { + "epoch": 1.46, + "learning_rate": 8.147435897435896e-05, + "loss": 0.0356, + "step": 68180 + }, + { + "epoch": 1.46, + "learning_rate": 8.144230769230768e-05, + "loss": 0.0304, + "step": 68190 + }, + { + "epoch": 1.46, + "learning_rate": 8.14102564102564e-05, + "loss": 0.0268, + "step": 68200 + }, + { + "epoch": 1.46, + "learning_rate": 8.137820512820512e-05, + "loss": 0.0412, + "step": 68210 + }, + { + "epoch": 1.46, + "learning_rate": 8.134615384615384e-05, + "loss": 0.0309, + "step": 68220 + }, + { + "epoch": 1.46, + "learning_rate": 8.131410256410256e-05, + "loss": 0.0346, + "step": 68230 + }, + { + "epoch": 1.46, + "learning_rate": 8.128205128205126e-05, + "loss": 0.0352, + "step": 68240 + }, + { + "epoch": 1.46, + "learning_rate": 8.124999999999998e-05, + "loss": 0.0326, + "step": 68250 + }, + { + "epoch": 1.46, + "learning_rate": 8.12179487179487e-05, + "loss": 0.0302, + "step": 68260 + }, + { + "epoch": 1.46, + "learning_rate": 8.118589743589742e-05, + "loss": 0.0327, + "step": 68270 + }, + { + "epoch": 1.46, + "learning_rate": 8.115384615384614e-05, + "loss": 0.0325, + "step": 68280 + }, + { + "epoch": 1.46, + "learning_rate": 8.112179487179486e-05, + "loss": 0.0258, + "step": 68290 + }, + { + "epoch": 1.46, + "learning_rate": 8.108974358974358e-05, + "loss": 0.0383, + "step": 68300 + }, + { + "epoch": 1.46, + "learning_rate": 8.10576923076923e-05, + "loss": 0.0355, + "step": 68310 + }, + { + "epoch": 1.46, + "learning_rate": 8.102564102564102e-05, + "loss": 0.033, + "step": 68320 + }, + { + "epoch": 1.46, + "learning_rate": 8.099358974358973e-05, + "loss": 0.0376, + "step": 68330 + }, + { + "epoch": 1.46, + "learning_rate": 8.096153846153845e-05, + "loss": 0.027, + "step": 68340 + }, + { + "epoch": 1.46, + "learning_rate": 8.092948717948717e-05, + "loss": 0.0342, + "step": 68350 + }, + { + "epoch": 1.46, + "learning_rate": 8.089743589743588e-05, + "loss": 0.0343, + "step": 68360 + }, + { + "epoch": 1.46, + "learning_rate": 8.08653846153846e-05, + "loss": 0.0334, + "step": 68370 + }, + { + "epoch": 1.46, + "learning_rate": 8.083333333333332e-05, + "loss": 0.0407, + "step": 68380 + }, + { + "epoch": 1.46, + "learning_rate": 8.080128205128204e-05, + "loss": 0.0368, + "step": 68390 + }, + { + "epoch": 1.46, + "learning_rate": 8.076923076923076e-05, + "loss": 0.0365, + "step": 68400 + }, + { + "epoch": 1.46, + "learning_rate": 8.073717948717948e-05, + "loss": 0.031, + "step": 68410 + }, + { + "epoch": 1.46, + "learning_rate": 8.070512820512819e-05, + "loss": 0.0303, + "step": 68420 + }, + { + "epoch": 1.46, + "learning_rate": 8.067307692307691e-05, + "loss": 0.0325, + "step": 68430 + }, + { + "epoch": 1.46, + "learning_rate": 8.064102564102563e-05, + "loss": 0.0298, + "step": 68440 + }, + { + "epoch": 1.46, + "learning_rate": 8.060897435897435e-05, + "loss": 0.0332, + "step": 68450 + }, + { + "epoch": 1.46, + "learning_rate": 8.057692307692307e-05, + "loss": 0.0316, + "step": 68460 + }, + { + "epoch": 1.46, + "learning_rate": 8.054487179487179e-05, + "loss": 0.0327, + "step": 68470 + }, + { + "epoch": 1.46, + "learning_rate": 8.05128205128205e-05, + "loss": 0.028, + "step": 68480 + }, + { + "epoch": 1.46, + "learning_rate": 8.048076923076923e-05, + "loss": 0.0303, + "step": 68490 + }, + { + "epoch": 1.46, + "learning_rate": 8.044871794871795e-05, + "loss": 0.039, + "step": 68500 + }, + { + "epoch": 1.46, + "learning_rate": 8.041666666666665e-05, + "loss": 0.031, + "step": 68510 + }, + { + "epoch": 1.46, + "learning_rate": 8.038461538461537e-05, + "loss": 0.0325, + "step": 68520 + }, + { + "epoch": 1.46, + "learning_rate": 8.035256410256409e-05, + "loss": 0.0341, + "step": 68530 + }, + { + "epoch": 1.46, + "learning_rate": 8.032051282051281e-05, + "loss": 0.0284, + "step": 68540 + }, + { + "epoch": 1.46, + "learning_rate": 8.028846153846153e-05, + "loss": 0.0366, + "step": 68550 + }, + { + "epoch": 1.46, + "learning_rate": 8.025641025641025e-05, + "loss": 0.0328, + "step": 68560 + }, + { + "epoch": 1.47, + "learning_rate": 8.022435897435897e-05, + "loss": 0.0304, + "step": 68570 + }, + { + "epoch": 1.47, + "learning_rate": 8.019230769230769e-05, + "loss": 0.0322, + "step": 68580 + }, + { + "epoch": 1.47, + "learning_rate": 8.01602564102564e-05, + "loss": 0.0316, + "step": 68590 + }, + { + "epoch": 1.47, + "learning_rate": 8.012820512820511e-05, + "loss": 0.0264, + "step": 68600 + }, + { + "epoch": 1.47, + "learning_rate": 8.009615384615383e-05, + "loss": 0.0326, + "step": 68610 + }, + { + "epoch": 1.47, + "learning_rate": 8.006410256410255e-05, + "loss": 0.0284, + "step": 68620 + }, + { + "epoch": 1.47, + "learning_rate": 8.003205128205127e-05, + "loss": 0.0362, + "step": 68630 + }, + { + "epoch": 1.47, + "learning_rate": 7.999999999999999e-05, + "loss": 0.0344, + "step": 68640 + }, + { + "epoch": 1.47, + "learning_rate": 7.996794871794871e-05, + "loss": 0.0302, + "step": 68650 + }, + { + "epoch": 1.47, + "learning_rate": 7.993589743589743e-05, + "loss": 0.0344, + "step": 68660 + }, + { + "epoch": 1.47, + "learning_rate": 7.990384615384615e-05, + "loss": 0.0349, + "step": 68670 + }, + { + "epoch": 1.47, + "learning_rate": 7.987179487179486e-05, + "loss": 0.0337, + "step": 68680 + }, + { + "epoch": 1.47, + "learning_rate": 7.983974358974358e-05, + "loss": 0.0345, + "step": 68690 + }, + { + "epoch": 1.47, + "learning_rate": 7.98076923076923e-05, + "loss": 0.0264, + "step": 68700 + }, + { + "epoch": 1.47, + "learning_rate": 7.977564102564102e-05, + "loss": 0.0332, + "step": 68710 + }, + { + "epoch": 1.47, + "learning_rate": 7.974358974358974e-05, + "loss": 0.0326, + "step": 68720 + }, + { + "epoch": 1.47, + "learning_rate": 7.971153846153846e-05, + "loss": 0.0316, + "step": 68730 + }, + { + "epoch": 1.47, + "learning_rate": 7.967948717948718e-05, + "loss": 0.0357, + "step": 68740 + }, + { + "epoch": 1.47, + "learning_rate": 7.96474358974359e-05, + "loss": 0.0317, + "step": 68750 + }, + { + "epoch": 1.47, + "learning_rate": 7.961538461538461e-05, + "loss": 0.0318, + "step": 68760 + }, + { + "epoch": 1.47, + "learning_rate": 7.958333333333332e-05, + "loss": 0.0303, + "step": 68770 + }, + { + "epoch": 1.47, + "learning_rate": 7.955128205128204e-05, + "loss": 0.031, + "step": 68780 + }, + { + "epoch": 1.47, + "learning_rate": 7.951923076923076e-05, + "loss": 0.0318, + "step": 68790 + }, + { + "epoch": 1.47, + "learning_rate": 7.948717948717948e-05, + "loss": 0.0309, + "step": 68800 + }, + { + "epoch": 1.47, + "learning_rate": 7.94551282051282e-05, + "loss": 0.0319, + "step": 68810 + }, + { + "epoch": 1.47, + "learning_rate": 7.942307692307692e-05, + "loss": 0.0316, + "step": 68820 + }, + { + "epoch": 1.47, + "learning_rate": 7.939102564102564e-05, + "loss": 0.0292, + "step": 68830 + }, + { + "epoch": 1.47, + "learning_rate": 7.935897435897436e-05, + "loss": 0.0309, + "step": 68840 + }, + { + "epoch": 1.47, + "learning_rate": 7.932692307692308e-05, + "loss": 0.0277, + "step": 68850 + }, + { + "epoch": 1.47, + "learning_rate": 7.929487179487178e-05, + "loss": 0.0299, + "step": 68860 + }, + { + "epoch": 1.47, + "learning_rate": 7.92628205128205e-05, + "loss": 0.0301, + "step": 68870 + }, + { + "epoch": 1.47, + "learning_rate": 7.923076923076922e-05, + "loss": 0.0308, + "step": 68880 + }, + { + "epoch": 1.47, + "learning_rate": 7.919871794871794e-05, + "loss": 0.027, + "step": 68890 + }, + { + "epoch": 1.47, + "learning_rate": 7.916666666666666e-05, + "loss": 0.0318, + "step": 68900 + }, + { + "epoch": 1.47, + "learning_rate": 7.913461538461538e-05, + "loss": 0.0325, + "step": 68910 + }, + { + "epoch": 1.47, + "learning_rate": 7.91025641025641e-05, + "loss": 0.0286, + "step": 68920 + }, + { + "epoch": 1.47, + "learning_rate": 7.907051282051282e-05, + "loss": 0.0309, + "step": 68930 + }, + { + "epoch": 1.47, + "learning_rate": 7.903846153846154e-05, + "loss": 0.0292, + "step": 68940 + }, + { + "epoch": 1.47, + "learning_rate": 7.900641025641025e-05, + "loss": 0.0257, + "step": 68950 + }, + { + "epoch": 1.47, + "learning_rate": 7.897435897435897e-05, + "loss": 0.0309, + "step": 68960 + }, + { + "epoch": 1.47, + "learning_rate": 7.894230769230769e-05, + "loss": 0.0346, + "step": 68970 + }, + { + "epoch": 1.47, + "learning_rate": 7.89102564102564e-05, + "loss": 0.036, + "step": 68980 + }, + { + "epoch": 1.47, + "learning_rate": 7.887820512820512e-05, + "loss": 0.0295, + "step": 68990 + }, + { + "epoch": 1.47, + "learning_rate": 7.884615384615384e-05, + "loss": 0.0307, + "step": 69000 + }, + { + "epoch": 1.47, + "eval_cer": 0.9235534961617085, + "eval_loss": 0.024392802268266678, + "eval_runtime": 120.5099, + "eval_samples_per_second": 16.596, + "eval_steps_per_second": 4.149, + "step": 69000 + }, + { + "epoch": 1.47, + "learning_rate": 7.881410256410256e-05, + "loss": 0.0341, + "step": 69010 + }, + { + "epoch": 1.47, + "learning_rate": 7.878205128205128e-05, + "loss": 0.0359, + "step": 69020 + }, + { + "epoch": 1.47, + "learning_rate": 7.874999999999999e-05, + "loss": 0.0347, + "step": 69030 + }, + { + "epoch": 1.48, + "learning_rate": 7.871794871794871e-05, + "loss": 0.0295, + "step": 69040 + }, + { + "epoch": 1.48, + "learning_rate": 7.868589743589743e-05, + "loss": 0.0328, + "step": 69050 + }, + { + "epoch": 1.48, + "learning_rate": 7.865384615384615e-05, + "loss": 0.0264, + "step": 69060 + }, + { + "epoch": 1.48, + "learning_rate": 7.862179487179487e-05, + "loss": 0.0352, + "step": 69070 + }, + { + "epoch": 1.48, + "learning_rate": 7.858974358974359e-05, + "loss": 0.031, + "step": 69080 + }, + { + "epoch": 1.48, + "learning_rate": 7.855769230769231e-05, + "loss": 0.0387, + "step": 69090 + }, + { + "epoch": 1.48, + "learning_rate": 7.852564102564103e-05, + "loss": 0.0272, + "step": 69100 + }, + { + "epoch": 1.48, + "learning_rate": 7.849358974358975e-05, + "loss": 0.0292, + "step": 69110 + }, + { + "epoch": 1.48, + "learning_rate": 7.846153846153845e-05, + "loss": 0.0349, + "step": 69120 + }, + { + "epoch": 1.48, + "learning_rate": 7.842948717948717e-05, + "loss": 0.0286, + "step": 69130 + }, + { + "epoch": 1.48, + "learning_rate": 7.839743589743589e-05, + "loss": 0.0365, + "step": 69140 + }, + { + "epoch": 1.48, + "learning_rate": 7.836538461538461e-05, + "loss": 0.0279, + "step": 69150 + }, + { + "epoch": 1.48, + "learning_rate": 7.833333333333333e-05, + "loss": 0.0318, + "step": 69160 + }, + { + "epoch": 1.48, + "learning_rate": 7.830128205128205e-05, + "loss": 0.0335, + "step": 69170 + }, + { + "epoch": 1.48, + "learning_rate": 7.826923076923077e-05, + "loss": 0.0337, + "step": 69180 + }, + { + "epoch": 1.48, + "learning_rate": 7.823717948717949e-05, + "loss": 0.0238, + "step": 69190 + }, + { + "epoch": 1.48, + "learning_rate": 7.820512820512821e-05, + "loss": 0.0295, + "step": 69200 + }, + { + "epoch": 1.48, + "learning_rate": 7.817307692307692e-05, + "loss": 0.0311, + "step": 69210 + }, + { + "epoch": 1.48, + "learning_rate": 7.814102564102564e-05, + "loss": 0.0349, + "step": 69220 + }, + { + "epoch": 1.48, + "learning_rate": 7.810897435897435e-05, + "loss": 0.0304, + "step": 69230 + }, + { + "epoch": 1.48, + "learning_rate": 7.807692307692307e-05, + "loss": 0.0299, + "step": 69240 + }, + { + "epoch": 1.48, + "learning_rate": 7.80448717948718e-05, + "loss": 0.0313, + "step": 69250 + }, + { + "epoch": 1.48, + "learning_rate": 7.801282051282051e-05, + "loss": 0.0301, + "step": 69260 + }, + { + "epoch": 1.48, + "learning_rate": 7.798076923076923e-05, + "loss": 0.0359, + "step": 69270 + }, + { + "epoch": 1.48, + "learning_rate": 7.794871794871795e-05, + "loss": 0.0351, + "step": 69280 + }, + { + "epoch": 1.48, + "learning_rate": 7.791666666666667e-05, + "loss": 0.0325, + "step": 69290 + }, + { + "epoch": 1.48, + "learning_rate": 7.788461538461538e-05, + "loss": 0.0347, + "step": 69300 + }, + { + "epoch": 1.48, + "learning_rate": 7.78525641025641e-05, + "loss": 0.0302, + "step": 69310 + }, + { + "epoch": 1.48, + "learning_rate": 7.782051282051282e-05, + "loss": 0.0331, + "step": 69320 + }, + { + "epoch": 1.48, + "learning_rate": 7.778846153846154e-05, + "loss": 0.0262, + "step": 69330 + }, + { + "epoch": 1.48, + "learning_rate": 7.775641025641026e-05, + "loss": 0.0302, + "step": 69340 + }, + { + "epoch": 1.48, + "learning_rate": 7.772435897435898e-05, + "loss": 0.0341, + "step": 69350 + }, + { + "epoch": 1.48, + "learning_rate": 7.76923076923077e-05, + "loss": 0.0327, + "step": 69360 + }, + { + "epoch": 1.48, + "learning_rate": 7.766025641025642e-05, + "loss": 0.0386, + "step": 69370 + }, + { + "epoch": 1.48, + "learning_rate": 7.762820512820514e-05, + "loss": 0.0355, + "step": 69380 + }, + { + "epoch": 1.48, + "learning_rate": 7.759615384615384e-05, + "loss": 0.0372, + "step": 69390 + }, + { + "epoch": 1.48, + "learning_rate": 7.756410256410256e-05, + "loss": 0.0314, + "step": 69400 + }, + { + "epoch": 1.48, + "learning_rate": 7.753205128205128e-05, + "loss": 0.0314, + "step": 69410 + }, + { + "epoch": 1.48, + "learning_rate": 7.75e-05, + "loss": 0.039, + "step": 69420 + }, + { + "epoch": 1.48, + "learning_rate": 7.746794871794872e-05, + "loss": 0.0229, + "step": 69430 + }, + { + "epoch": 1.48, + "learning_rate": 7.743589743589744e-05, + "loss": 0.033, + "step": 69440 + }, + { + "epoch": 1.48, + "learning_rate": 7.740384615384616e-05, + "loss": 0.0309, + "step": 69450 + }, + { + "epoch": 1.48, + "learning_rate": 7.737179487179488e-05, + "loss": 0.0299, + "step": 69460 + }, + { + "epoch": 1.48, + "learning_rate": 7.733974358974357e-05, + "loss": 0.0321, + "step": 69470 + }, + { + "epoch": 1.48, + "learning_rate": 7.730769230769229e-05, + "loss": 0.0261, + "step": 69480 + }, + { + "epoch": 1.48, + "learning_rate": 7.727564102564101e-05, + "loss": 0.0409, + "step": 69490 + }, + { + "epoch": 1.49, + "learning_rate": 7.724358974358973e-05, + "loss": 0.0314, + "step": 69500 + }, + { + "epoch": 1.49, + "learning_rate": 7.721153846153845e-05, + "loss": 0.0375, + "step": 69510 + }, + { + "epoch": 1.49, + "learning_rate": 7.717948717948717e-05, + "loss": 0.0344, + "step": 69520 + }, + { + "epoch": 1.49, + "learning_rate": 7.714743589743589e-05, + "loss": 0.0297, + "step": 69530 + }, + { + "epoch": 1.49, + "learning_rate": 7.71153846153846e-05, + "loss": 0.0323, + "step": 69540 + }, + { + "epoch": 1.49, + "learning_rate": 7.708333333333331e-05, + "loss": 0.0371, + "step": 69550 + }, + { + "epoch": 1.49, + "learning_rate": 7.705128205128203e-05, + "loss": 0.0272, + "step": 69560 + }, + { + "epoch": 1.49, + "learning_rate": 7.701923076923075e-05, + "loss": 0.0332, + "step": 69570 + }, + { + "epoch": 1.49, + "learning_rate": 7.698717948717947e-05, + "loss": 0.0298, + "step": 69580 + }, + { + "epoch": 1.49, + "learning_rate": 7.695512820512819e-05, + "loss": 0.0291, + "step": 69590 + }, + { + "epoch": 1.49, + "learning_rate": 7.692307692307691e-05, + "loss": 0.0319, + "step": 69600 + }, + { + "epoch": 1.49, + "learning_rate": 7.689102564102563e-05, + "loss": 0.0329, + "step": 69610 + }, + { + "epoch": 1.49, + "learning_rate": 7.685897435897435e-05, + "loss": 0.0266, + "step": 69620 + }, + { + "epoch": 1.49, + "learning_rate": 7.682692307692306e-05, + "loss": 0.0259, + "step": 69630 + }, + { + "epoch": 1.49, + "learning_rate": 7.679487179487178e-05, + "loss": 0.0356, + "step": 69640 + }, + { + "epoch": 1.49, + "learning_rate": 7.67628205128205e-05, + "loss": 0.0237, + "step": 69650 + }, + { + "epoch": 1.49, + "learning_rate": 7.673076923076922e-05, + "loss": 0.032, + "step": 69660 + }, + { + "epoch": 1.49, + "learning_rate": 7.669871794871794e-05, + "loss": 0.0266, + "step": 69670 + }, + { + "epoch": 1.49, + "learning_rate": 7.666666666666666e-05, + "loss": 0.0283, + "step": 69680 + }, + { + "epoch": 1.49, + "learning_rate": 7.663461538461538e-05, + "loss": 0.0296, + "step": 69690 + }, + { + "epoch": 1.49, + "learning_rate": 7.66025641025641e-05, + "loss": 0.0332, + "step": 69700 + }, + { + "epoch": 1.49, + "learning_rate": 7.657051282051281e-05, + "loss": 0.0333, + "step": 69710 + }, + { + "epoch": 1.49, + "learning_rate": 7.653846153846152e-05, + "loss": 0.0351, + "step": 69720 + }, + { + "epoch": 1.49, + "learning_rate": 7.650641025641024e-05, + "loss": 0.0294, + "step": 69730 + }, + { + "epoch": 1.49, + "learning_rate": 7.647435897435896e-05, + "loss": 0.0269, + "step": 69740 + }, + { + "epoch": 1.49, + "learning_rate": 7.644230769230768e-05, + "loss": 0.0315, + "step": 69750 + }, + { + "epoch": 1.49, + "learning_rate": 7.64102564102564e-05, + "loss": 0.0337, + "step": 69760 + }, + { + "epoch": 1.49, + "learning_rate": 7.637820512820512e-05, + "loss": 0.033, + "step": 69770 + }, + { + "epoch": 1.49, + "learning_rate": 7.634615384615384e-05, + "loss": 0.0311, + "step": 69780 + }, + { + "epoch": 1.49, + "learning_rate": 7.631410256410256e-05, + "loss": 0.0269, + "step": 69790 + }, + { + "epoch": 1.49, + "learning_rate": 7.628205128205128e-05, + "loss": 0.0261, + "step": 69800 + }, + { + "epoch": 1.49, + "learning_rate": 7.624999999999998e-05, + "loss": 0.032, + "step": 69810 + }, + { + "epoch": 1.49, + "learning_rate": 7.62179487179487e-05, + "loss": 0.0311, + "step": 69820 + }, + { + "epoch": 1.49, + "learning_rate": 7.618589743589742e-05, + "loss": 0.0297, + "step": 69830 + }, + { + "epoch": 1.49, + "learning_rate": 7.615384615384614e-05, + "loss": 0.0236, + "step": 69840 + }, + { + "epoch": 1.49, + "learning_rate": 7.612179487179486e-05, + "loss": 0.0301, + "step": 69850 + }, + { + "epoch": 1.49, + "learning_rate": 7.608974358974358e-05, + "loss": 0.0324, + "step": 69860 + }, + { + "epoch": 1.49, + "learning_rate": 7.60576923076923e-05, + "loss": 0.0365, + "step": 69870 + }, + { + "epoch": 1.49, + "learning_rate": 7.602564102564102e-05, + "loss": 0.0338, + "step": 69880 + }, + { + "epoch": 1.49, + "learning_rate": 7.599358974358974e-05, + "loss": 0.0296, + "step": 69890 + }, + { + "epoch": 1.49, + "learning_rate": 7.596153846153845e-05, + "loss": 0.0326, + "step": 69900 + }, + { + "epoch": 1.49, + "learning_rate": 7.592948717948717e-05, + "loss": 0.0339, + "step": 69910 + }, + { + "epoch": 1.49, + "learning_rate": 7.589743589743589e-05, + "loss": 0.0337, + "step": 69920 + }, + { + "epoch": 1.49, + "learning_rate": 7.58653846153846e-05, + "loss": 0.0284, + "step": 69930 + }, + { + "epoch": 1.49, + "learning_rate": 7.583333333333332e-05, + "loss": 0.0344, + "step": 69940 + }, + { + "epoch": 1.49, + "learning_rate": 7.580128205128204e-05, + "loss": 0.0339, + "step": 69950 + }, + { + "epoch": 1.49, + "learning_rate": 7.576923076923076e-05, + "loss": 0.0288, + "step": 69960 + }, + { + "epoch": 1.5, + "learning_rate": 7.573717948717948e-05, + "loss": 0.0352, + "step": 69970 + }, + { + "epoch": 1.5, + "learning_rate": 7.570512820512819e-05, + "loss": 0.0333, + "step": 69980 + }, + { + "epoch": 1.5, + "learning_rate": 7.567307692307691e-05, + "loss": 0.0347, + "step": 69990 + }, + { + "epoch": 1.5, + "learning_rate": 7.564102564102563e-05, + "loss": 0.0293, + "step": 70000 + }, + { + "epoch": 1.5, + "eval_cer": 0.9236639017152777, + "eval_loss": 0.02389584109187126, + "eval_runtime": 120.2898, + "eval_samples_per_second": 16.627, + "eval_steps_per_second": 4.157, + "step": 70000 + }, + { + "epoch": 1.5, + "learning_rate": 7.560897435897435e-05, + "loss": 0.0318, + "step": 70010 + }, + { + "epoch": 1.5, + "learning_rate": 7.557692307692307e-05, + "loss": 0.0323, + "step": 70020 + }, + { + "epoch": 1.5, + "learning_rate": 7.554487179487179e-05, + "loss": 0.0367, + "step": 70030 + }, + { + "epoch": 1.5, + "learning_rate": 7.551282051282051e-05, + "loss": 0.0349, + "step": 70040 + }, + { + "epoch": 1.5, + "learning_rate": 7.548076923076923e-05, + "loss": 0.0342, + "step": 70050 + }, + { + "epoch": 1.5, + "learning_rate": 7.544871794871795e-05, + "loss": 0.0325, + "step": 70060 + }, + { + "epoch": 1.5, + "learning_rate": 7.541666666666665e-05, + "loss": 0.032, + "step": 70070 + }, + { + "epoch": 1.5, + "learning_rate": 7.538461538461537e-05, + "loss": 0.0267, + "step": 70080 + }, + { + "epoch": 1.5, + "learning_rate": 7.535256410256409e-05, + "loss": 0.034, + "step": 70090 + }, + { + "epoch": 1.5, + "learning_rate": 7.532051282051281e-05, + "loss": 0.0301, + "step": 70100 + }, + { + "epoch": 1.5, + "learning_rate": 7.528846153846153e-05, + "loss": 0.0327, + "step": 70110 + }, + { + "epoch": 1.5, + "learning_rate": 7.525641025641025e-05, + "loss": 0.0334, + "step": 70120 + }, + { + "epoch": 1.5, + "learning_rate": 7.522435897435897e-05, + "loss": 0.0331, + "step": 70130 + }, + { + "epoch": 1.5, + "learning_rate": 7.519230769230769e-05, + "loss": 0.0333, + "step": 70140 + }, + { + "epoch": 1.5, + "learning_rate": 7.516025641025641e-05, + "loss": 0.0256, + "step": 70150 + }, + { + "epoch": 1.5, + "learning_rate": 7.512820512820512e-05, + "loss": 0.0315, + "step": 70160 + }, + { + "epoch": 1.5, + "learning_rate": 7.509615384615383e-05, + "loss": 0.0318, + "step": 70170 + }, + { + "epoch": 1.5, + "learning_rate": 7.506410256410255e-05, + "loss": 0.0306, + "step": 70180 + }, + { + "epoch": 1.5, + "learning_rate": 7.503205128205127e-05, + "loss": 0.0281, + "step": 70190 + }, + { + "epoch": 1.5, + "learning_rate": 7.5e-05, + "loss": 0.0274, + "step": 70200 + }, + { + "epoch": 1.5, + "learning_rate": 7.496794871794871e-05, + "loss": 0.0293, + "step": 70210 + }, + { + "epoch": 1.5, + "learning_rate": 7.493589743589743e-05, + "loss": 0.0338, + "step": 70220 + }, + { + "epoch": 1.5, + "learning_rate": 7.490384615384615e-05, + "loss": 0.0325, + "step": 70230 + }, + { + "epoch": 1.5, + "learning_rate": 7.487179487179487e-05, + "loss": 0.0332, + "step": 70240 + }, + { + "epoch": 1.5, + "learning_rate": 7.483974358974358e-05, + "loss": 0.0337, + "step": 70250 + }, + { + "epoch": 1.5, + "learning_rate": 7.48076923076923e-05, + "loss": 0.0334, + "step": 70260 + }, + { + "epoch": 1.5, + "learning_rate": 7.477564102564102e-05, + "loss": 0.0396, + "step": 70270 + }, + { + "epoch": 1.5, + "learning_rate": 7.474358974358974e-05, + "loss": 0.0322, + "step": 70280 + }, + { + "epoch": 1.5, + "learning_rate": 7.471153846153846e-05, + "loss": 0.0326, + "step": 70290 + }, + { + "epoch": 1.5, + "learning_rate": 7.467948717948718e-05, + "loss": 0.0369, + "step": 70300 + }, + { + "epoch": 1.5, + "learning_rate": 7.46474358974359e-05, + "loss": 0.0304, + "step": 70310 + }, + { + "epoch": 1.5, + "learning_rate": 7.461538461538462e-05, + "loss": 0.0344, + "step": 70320 + }, + { + "epoch": 1.5, + "learning_rate": 7.458333333333333e-05, + "loss": 0.0301, + "step": 70330 + }, + { + "epoch": 1.5, + "learning_rate": 7.455128205128204e-05, + "loss": 0.0269, + "step": 70340 + }, + { + "epoch": 1.5, + "learning_rate": 7.451923076923076e-05, + "loss": 0.0333, + "step": 70350 + }, + { + "epoch": 1.5, + "learning_rate": 7.448717948717948e-05, + "loss": 0.037, + "step": 70360 + }, + { + "epoch": 1.5, + "learning_rate": 7.44551282051282e-05, + "loss": 0.0347, + "step": 70370 + }, + { + "epoch": 1.5, + "learning_rate": 7.442307692307692e-05, + "loss": 0.0347, + "step": 70380 + }, + { + "epoch": 1.5, + "learning_rate": 7.439102564102564e-05, + "loss": 0.0268, + "step": 70390 + }, + { + "epoch": 1.5, + "learning_rate": 7.435897435897436e-05, + "loss": 0.0327, + "step": 70400 + }, + { + "epoch": 1.5, + "learning_rate": 7.432692307692308e-05, + "loss": 0.0333, + "step": 70410 + }, + { + "epoch": 1.5, + "learning_rate": 7.42948717948718e-05, + "loss": 0.0439, + "step": 70420 + }, + { + "epoch": 1.5, + "learning_rate": 7.42628205128205e-05, + "loss": 0.0317, + "step": 70430 + }, + { + "epoch": 1.51, + "learning_rate": 7.423076923076922e-05, + "loss": 0.0361, + "step": 70440 + }, + { + "epoch": 1.51, + "learning_rate": 7.419871794871794e-05, + "loss": 0.0299, + "step": 70450 + }, + { + "epoch": 1.51, + "learning_rate": 7.416666666666666e-05, + "loss": 0.0318, + "step": 70460 + }, + { + "epoch": 1.51, + "learning_rate": 7.413461538461538e-05, + "loss": 0.0361, + "step": 70470 + }, + { + "epoch": 1.51, + "learning_rate": 7.41025641025641e-05, + "loss": 0.0283, + "step": 70480 + }, + { + "epoch": 1.51, + "learning_rate": 7.407051282051282e-05, + "loss": 0.0341, + "step": 70490 + }, + { + "epoch": 1.51, + "learning_rate": 7.403846153846154e-05, + "loss": 0.0299, + "step": 70500 + }, + { + "epoch": 1.51, + "learning_rate": 7.400641025641025e-05, + "loss": 0.0401, + "step": 70510 + }, + { + "epoch": 1.51, + "learning_rate": 7.397435897435897e-05, + "loss": 0.0269, + "step": 70520 + }, + { + "epoch": 1.51, + "learning_rate": 7.394230769230769e-05, + "loss": 0.0299, + "step": 70530 + }, + { + "epoch": 1.51, + "learning_rate": 7.39102564102564e-05, + "loss": 0.033, + "step": 70540 + }, + { + "epoch": 1.51, + "learning_rate": 7.387820512820513e-05, + "loss": 0.0297, + "step": 70550 + }, + { + "epoch": 1.51, + "learning_rate": 7.384615384615384e-05, + "loss": 0.0336, + "step": 70560 + }, + { + "epoch": 1.51, + "learning_rate": 7.381410256410255e-05, + "loss": 0.0307, + "step": 70570 + }, + { + "epoch": 1.51, + "learning_rate": 7.378205128205127e-05, + "loss": 0.0308, + "step": 70580 + }, + { + "epoch": 1.51, + "learning_rate": 7.374999999999999e-05, + "loss": 0.0388, + "step": 70590 + }, + { + "epoch": 1.51, + "learning_rate": 7.371794871794871e-05, + "loss": 0.031, + "step": 70600 + }, + { + "epoch": 1.51, + "learning_rate": 7.368589743589743e-05, + "loss": 0.0331, + "step": 70610 + }, + { + "epoch": 1.51, + "learning_rate": 7.365384615384615e-05, + "loss": 0.0312, + "step": 70620 + }, + { + "epoch": 1.51, + "learning_rate": 7.362179487179487e-05, + "loss": 0.0296, + "step": 70630 + }, + { + "epoch": 1.51, + "learning_rate": 7.358974358974357e-05, + "loss": 0.033, + "step": 70640 + }, + { + "epoch": 1.51, + "learning_rate": 7.35576923076923e-05, + "loss": 0.038, + "step": 70650 + }, + { + "epoch": 1.51, + "learning_rate": 7.352564102564101e-05, + "loss": 0.0376, + "step": 70660 + }, + { + "epoch": 1.51, + "learning_rate": 7.349358974358973e-05, + "loss": 0.0333, + "step": 70670 + }, + { + "epoch": 1.51, + "learning_rate": 7.346153846153845e-05, + "loss": 0.0283, + "step": 70680 + }, + { + "epoch": 1.51, + "learning_rate": 7.342948717948717e-05, + "loss": 0.0243, + "step": 70690 + }, + { + "epoch": 1.51, + "learning_rate": 7.339743589743589e-05, + "loss": 0.032, + "step": 70700 + }, + { + "epoch": 1.51, + "learning_rate": 7.336538461538461e-05, + "loss": 0.0384, + "step": 70710 + }, + { + "epoch": 1.51, + "learning_rate": 7.333333333333332e-05, + "loss": 0.0282, + "step": 70720 + }, + { + "epoch": 1.51, + "learning_rate": 7.330128205128204e-05, + "loss": 0.0351, + "step": 70730 + }, + { + "epoch": 1.51, + "learning_rate": 7.326923076923076e-05, + "loss": 0.0326, + "step": 70740 + }, + { + "epoch": 1.51, + "learning_rate": 7.323717948717948e-05, + "loss": 0.0263, + "step": 70750 + }, + { + "epoch": 1.51, + "learning_rate": 7.32051282051282e-05, + "loss": 0.0295, + "step": 70760 + }, + { + "epoch": 1.51, + "learning_rate": 7.317307692307692e-05, + "loss": 0.0352, + "step": 70770 + }, + { + "epoch": 1.51, + "learning_rate": 7.314102564102564e-05, + "loss": 0.0326, + "step": 70780 + }, + { + "epoch": 1.51, + "learning_rate": 7.310897435897436e-05, + "loss": 0.0336, + "step": 70790 + }, + { + "epoch": 1.51, + "learning_rate": 7.307692307692307e-05, + "loss": 0.0299, + "step": 70800 + }, + { + "epoch": 1.51, + "learning_rate": 7.304487179487178e-05, + "loss": 0.0306, + "step": 70810 + }, + { + "epoch": 1.51, + "learning_rate": 7.30128205128205e-05, + "loss": 0.0348, + "step": 70820 + }, + { + "epoch": 1.51, + "learning_rate": 7.298076923076922e-05, + "loss": 0.0267, + "step": 70830 + }, + { + "epoch": 1.51, + "learning_rate": 7.294871794871794e-05, + "loss": 0.0293, + "step": 70840 + }, + { + "epoch": 1.51, + "learning_rate": 7.291666666666666e-05, + "loss": 0.0302, + "step": 70850 + }, + { + "epoch": 1.51, + "learning_rate": 7.288461538461538e-05, + "loss": 0.0316, + "step": 70860 + }, + { + "epoch": 1.51, + "learning_rate": 7.28525641025641e-05, + "loss": 0.0267, + "step": 70870 + }, + { + "epoch": 1.51, + "learning_rate": 7.282051282051282e-05, + "loss": 0.0301, + "step": 70880 + }, + { + "epoch": 1.51, + "learning_rate": 7.278846153846154e-05, + "loss": 0.0313, + "step": 70890 + }, + { + "epoch": 1.51, + "learning_rate": 7.275641025641024e-05, + "loss": 0.0346, + "step": 70900 + }, + { + "epoch": 1.52, + "learning_rate": 7.272435897435896e-05, + "loss": 0.0315, + "step": 70910 + }, + { + "epoch": 1.52, + "learning_rate": 7.269230769230768e-05, + "loss": 0.0293, + "step": 70920 + }, + { + "epoch": 1.52, + "learning_rate": 7.26602564102564e-05, + "loss": 0.0336, + "step": 70930 + }, + { + "epoch": 1.52, + "learning_rate": 7.262820512820512e-05, + "loss": 0.0286, + "step": 70940 + }, + { + "epoch": 1.52, + "learning_rate": 7.259615384615384e-05, + "loss": 0.0329, + "step": 70950 + }, + { + "epoch": 1.52, + "learning_rate": 7.256410256410256e-05, + "loss": 0.0342, + "step": 70960 + }, + { + "epoch": 1.52, + "learning_rate": 7.253205128205128e-05, + "loss": 0.0348, + "step": 70970 + }, + { + "epoch": 1.52, + "learning_rate": 7.25e-05, + "loss": 0.0417, + "step": 70980 + }, + { + "epoch": 1.52, + "learning_rate": 7.24679487179487e-05, + "loss": 0.0392, + "step": 70990 + }, + { + "epoch": 1.52, + "learning_rate": 7.243589743589743e-05, + "loss": 0.0406, + "step": 71000 + }, + { + "epoch": 1.52, + "eval_cer": 0.9236341771431629, + "eval_loss": 0.023512328043580055, + "eval_runtime": 120.2851, + "eval_samples_per_second": 16.627, + "eval_steps_per_second": 4.157, + "step": 71000 + }, + { + "epoch": 1.52, + "learning_rate": 7.240384615384615e-05, + "loss": 0.0348, + "step": 71010 + }, + { + "epoch": 1.52, + "learning_rate": 7.237179487179487e-05, + "loss": 0.03, + "step": 71020 + }, + { + "epoch": 1.52, + "learning_rate": 7.233974358974358e-05, + "loss": 0.0348, + "step": 71030 + }, + { + "epoch": 1.52, + "learning_rate": 7.23076923076923e-05, + "loss": 0.0367, + "step": 71040 + }, + { + "epoch": 1.52, + "learning_rate": 7.227564102564102e-05, + "loss": 0.0343, + "step": 71050 + }, + { + "epoch": 1.52, + "learning_rate": 7.224358974358974e-05, + "loss": 0.0383, + "step": 71060 + }, + { + "epoch": 1.52, + "learning_rate": 7.221153846153846e-05, + "loss": 0.0316, + "step": 71070 + }, + { + "epoch": 1.52, + "learning_rate": 7.217948717948717e-05, + "loss": 0.0336, + "step": 71080 + }, + { + "epoch": 1.52, + "learning_rate": 7.214743589743589e-05, + "loss": 0.032, + "step": 71090 + }, + { + "epoch": 1.52, + "learning_rate": 7.211538461538461e-05, + "loss": 0.0279, + "step": 71100 + }, + { + "epoch": 1.52, + "learning_rate": 7.208333333333333e-05, + "loss": 0.0285, + "step": 71110 + }, + { + "epoch": 1.52, + "learning_rate": 7.205128205128205e-05, + "loss": 0.029, + "step": 71120 + }, + { + "epoch": 1.52, + "learning_rate": 7.201923076923077e-05, + "loss": 0.0315, + "step": 71130 + }, + { + "epoch": 1.52, + "learning_rate": 7.198717948717949e-05, + "loss": 0.0326, + "step": 71140 + }, + { + "epoch": 1.52, + "learning_rate": 7.19551282051282e-05, + "loss": 0.0324, + "step": 71150 + }, + { + "epoch": 1.52, + "learning_rate": 7.192307692307691e-05, + "loss": 0.037, + "step": 71160 + }, + { + "epoch": 1.52, + "learning_rate": 7.189102564102563e-05, + "loss": 0.0297, + "step": 71170 + }, + { + "epoch": 1.52, + "learning_rate": 7.185897435897435e-05, + "loss": 0.0289, + "step": 71180 + }, + { + "epoch": 1.52, + "learning_rate": 7.182692307692307e-05, + "loss": 0.033, + "step": 71190 + }, + { + "epoch": 1.52, + "learning_rate": 7.179487179487179e-05, + "loss": 0.0364, + "step": 71200 + }, + { + "epoch": 1.52, + "learning_rate": 7.176282051282051e-05, + "loss": 0.032, + "step": 71210 + }, + { + "epoch": 1.52, + "learning_rate": 7.173076923076923e-05, + "loss": 0.0298, + "step": 71220 + }, + { + "epoch": 1.52, + "learning_rate": 7.169871794871795e-05, + "loss": 0.0312, + "step": 71230 + }, + { + "epoch": 1.52, + "learning_rate": 7.166666666666667e-05, + "loss": 0.0356, + "step": 71240 + }, + { + "epoch": 1.52, + "learning_rate": 7.163461538461538e-05, + "loss": 0.0348, + "step": 71250 + }, + { + "epoch": 1.52, + "learning_rate": 7.16025641025641e-05, + "loss": 0.0338, + "step": 71260 + }, + { + "epoch": 1.52, + "learning_rate": 7.157051282051281e-05, + "loss": 0.0311, + "step": 71270 + }, + { + "epoch": 1.52, + "learning_rate": 7.153846153846153e-05, + "loss": 0.0277, + "step": 71280 + }, + { + "epoch": 1.52, + "learning_rate": 7.150641025641025e-05, + "loss": 0.031, + "step": 71290 + }, + { + "epoch": 1.52, + "learning_rate": 7.147435897435896e-05, + "loss": 0.0349, + "step": 71300 + }, + { + "epoch": 1.52, + "learning_rate": 7.144230769230768e-05, + "loss": 0.0306, + "step": 71310 + }, + { + "epoch": 1.52, + "learning_rate": 7.14102564102564e-05, + "loss": 0.0298, + "step": 71320 + }, + { + "epoch": 1.52, + "learning_rate": 7.137820512820512e-05, + "loss": 0.0288, + "step": 71330 + }, + { + "epoch": 1.52, + "learning_rate": 7.134615384615384e-05, + "loss": 0.029, + "step": 71340 + }, + { + "epoch": 1.52, + "learning_rate": 7.131410256410256e-05, + "loss": 0.0258, + "step": 71350 + }, + { + "epoch": 1.52, + "learning_rate": 7.128205128205128e-05, + "loss": 0.0325, + "step": 71360 + }, + { + "epoch": 1.52, + "learning_rate": 7.125e-05, + "loss": 0.03, + "step": 71370 + }, + { + "epoch": 1.53, + "learning_rate": 7.12179487179487e-05, + "loss": 0.0298, + "step": 71380 + }, + { + "epoch": 1.53, + "learning_rate": 7.118589743589742e-05, + "loss": 0.0307, + "step": 71390 + }, + { + "epoch": 1.53, + "learning_rate": 7.115384615384614e-05, + "loss": 0.0304, + "step": 71400 + }, + { + "epoch": 1.53, + "learning_rate": 7.112179487179486e-05, + "loss": 0.0339, + "step": 71410 + }, + { + "epoch": 1.53, + "learning_rate": 7.108974358974358e-05, + "loss": 0.0332, + "step": 71420 + }, + { + "epoch": 1.53, + "learning_rate": 7.10576923076923e-05, + "loss": 0.0307, + "step": 71430 + }, + { + "epoch": 1.53, + "learning_rate": 7.102564102564102e-05, + "loss": 0.0314, + "step": 71440 + }, + { + "epoch": 1.53, + "learning_rate": 7.099358974358974e-05, + "loss": 0.0293, + "step": 71450 + }, + { + "epoch": 1.53, + "learning_rate": 7.096153846153845e-05, + "loss": 0.0271, + "step": 71460 + }, + { + "epoch": 1.53, + "learning_rate": 7.092948717948717e-05, + "loss": 0.0331, + "step": 71470 + }, + { + "epoch": 1.53, + "learning_rate": 7.089743589743589e-05, + "loss": 0.0307, + "step": 71480 + }, + { + "epoch": 1.53, + "learning_rate": 7.08653846153846e-05, + "loss": 0.0294, + "step": 71490 + }, + { + "epoch": 1.53, + "learning_rate": 7.083333333333332e-05, + "loss": 0.0286, + "step": 71500 + }, + { + "epoch": 1.53, + "learning_rate": 7.080128205128204e-05, + "loss": 0.0322, + "step": 71510 + }, + { + "epoch": 1.53, + "learning_rate": 7.076923076923076e-05, + "loss": 0.0276, + "step": 71520 + }, + { + "epoch": 1.53, + "learning_rate": 7.073717948717948e-05, + "loss": 0.0305, + "step": 71530 + }, + { + "epoch": 1.53, + "learning_rate": 7.07051282051282e-05, + "loss": 0.0276, + "step": 71540 + }, + { + "epoch": 1.53, + "learning_rate": 7.067307692307691e-05, + "loss": 0.0318, + "step": 71550 + }, + { + "epoch": 1.53, + "learning_rate": 7.064102564102563e-05, + "loss": 0.0315, + "step": 71560 + }, + { + "epoch": 1.53, + "learning_rate": 7.060897435897435e-05, + "loss": 0.0249, + "step": 71570 + }, + { + "epoch": 1.53, + "learning_rate": 7.057692307692307e-05, + "loss": 0.031, + "step": 71580 + }, + { + "epoch": 1.53, + "learning_rate": 7.054487179487179e-05, + "loss": 0.0324, + "step": 71590 + }, + { + "epoch": 1.53, + "learning_rate": 7.051282051282051e-05, + "loss": 0.0337, + "step": 71600 + }, + { + "epoch": 1.53, + "learning_rate": 7.048076923076923e-05, + "loss": 0.0317, + "step": 71610 + }, + { + "epoch": 1.53, + "learning_rate": 7.044871794871795e-05, + "loss": 0.0336, + "step": 71620 + }, + { + "epoch": 1.53, + "learning_rate": 7.041666666666667e-05, + "loss": 0.0347, + "step": 71630 + }, + { + "epoch": 1.53, + "learning_rate": 7.038461538461537e-05, + "loss": 0.0455, + "step": 71640 + }, + { + "epoch": 1.53, + "learning_rate": 7.035256410256409e-05, + "loss": 0.0307, + "step": 71650 + }, + { + "epoch": 1.53, + "learning_rate": 7.032051282051281e-05, + "loss": 0.0394, + "step": 71660 + }, + { + "epoch": 1.53, + "learning_rate": 7.028846153846153e-05, + "loss": 0.0319, + "step": 71670 + }, + { + "epoch": 1.53, + "learning_rate": 7.025641025641025e-05, + "loss": 0.0299, + "step": 71680 + }, + { + "epoch": 1.53, + "learning_rate": 7.022435897435897e-05, + "loss": 0.0274, + "step": 71690 + }, + { + "epoch": 1.53, + "learning_rate": 7.019230769230769e-05, + "loss": 0.0326, + "step": 71700 + }, + { + "epoch": 1.53, + "learning_rate": 7.016025641025641e-05, + "loss": 0.031, + "step": 71710 + }, + { + "epoch": 1.53, + "learning_rate": 7.012820512820513e-05, + "loss": 0.0323, + "step": 71720 + }, + { + "epoch": 1.53, + "learning_rate": 7.009615384615384e-05, + "loss": 0.0393, + "step": 71730 + }, + { + "epoch": 1.53, + "learning_rate": 7.006410256410255e-05, + "loss": 0.0285, + "step": 71740 + }, + { + "epoch": 1.53, + "learning_rate": 7.003205128205127e-05, + "loss": 0.0342, + "step": 71750 + }, + { + "epoch": 1.53, + "learning_rate": 7e-05, + "loss": 0.0265, + "step": 71760 + }, + { + "epoch": 1.53, + "learning_rate": 6.996794871794871e-05, + "loss": 0.027, + "step": 71770 + }, + { + "epoch": 1.53, + "learning_rate": 6.993589743589743e-05, + "loss": 0.0298, + "step": 71780 + }, + { + "epoch": 1.53, + "learning_rate": 6.990384615384615e-05, + "loss": 0.0251, + "step": 71790 + }, + { + "epoch": 1.53, + "learning_rate": 6.987179487179487e-05, + "loss": 0.0297, + "step": 71800 + }, + { + "epoch": 1.53, + "learning_rate": 6.983974358974359e-05, + "loss": 0.0293, + "step": 71810 + }, + { + "epoch": 1.53, + "learning_rate": 6.98076923076923e-05, + "loss": 0.0343, + "step": 71820 + }, + { + "epoch": 1.53, + "learning_rate": 6.977564102564102e-05, + "loss": 0.0315, + "step": 71830 + }, + { + "epoch": 1.54, + "learning_rate": 6.974358974358974e-05, + "loss": 0.032, + "step": 71840 + }, + { + "epoch": 1.54, + "learning_rate": 6.971153846153846e-05, + "loss": 0.0341, + "step": 71850 + }, + { + "epoch": 1.54, + "learning_rate": 6.967948717948718e-05, + "loss": 0.0291, + "step": 71860 + }, + { + "epoch": 1.54, + "learning_rate": 6.96474358974359e-05, + "loss": 0.0235, + "step": 71870 + }, + { + "epoch": 1.54, + "learning_rate": 6.961538461538462e-05, + "loss": 0.0305, + "step": 71880 + }, + { + "epoch": 1.54, + "learning_rate": 6.958333333333334e-05, + "loss": 0.0323, + "step": 71890 + }, + { + "epoch": 1.54, + "learning_rate": 6.955128205128204e-05, + "loss": 0.0295, + "step": 71900 + }, + { + "epoch": 1.54, + "learning_rate": 6.951923076923076e-05, + "loss": 0.0312, + "step": 71910 + }, + { + "epoch": 1.54, + "learning_rate": 6.948717948717948e-05, + "loss": 0.0331, + "step": 71920 + }, + { + "epoch": 1.54, + "learning_rate": 6.94551282051282e-05, + "loss": 0.0307, + "step": 71930 + }, + { + "epoch": 1.54, + "learning_rate": 6.942307692307692e-05, + "loss": 0.0326, + "step": 71940 + }, + { + "epoch": 1.54, + "learning_rate": 6.939102564102564e-05, + "loss": 0.0319, + "step": 71950 + }, + { + "epoch": 1.54, + "learning_rate": 6.935897435897436e-05, + "loss": 0.0351, + "step": 71960 + }, + { + "epoch": 1.54, + "learning_rate": 6.932692307692308e-05, + "loss": 0.0255, + "step": 71970 + }, + { + "epoch": 1.54, + "learning_rate": 6.92948717948718e-05, + "loss": 0.0332, + "step": 71980 + }, + { + "epoch": 1.54, + "learning_rate": 6.92628205128205e-05, + "loss": 0.0276, + "step": 71990 + }, + { + "epoch": 1.54, + "learning_rate": 6.923076923076922e-05, + "loss": 0.0273, + "step": 72000 + }, + { + "epoch": 1.54, + "eval_cer": 0.9235798236398672, + "eval_loss": 0.0234876349568367, + "eval_runtime": 120.212, + "eval_samples_per_second": 16.637, + "eval_steps_per_second": 4.159, + "step": 72000 + }, + { + "epoch": 1.54, + "learning_rate": 6.919871794871794e-05, + "loss": 0.029, + "step": 72010 + }, + { + "epoch": 1.54, + "learning_rate": 6.916666666666666e-05, + "loss": 0.032, + "step": 72020 + }, + { + "epoch": 1.54, + "learning_rate": 6.913461538461537e-05, + "loss": 0.0307, + "step": 72030 + }, + { + "epoch": 1.54, + "learning_rate": 6.910256410256409e-05, + "loss": 0.0345, + "step": 72040 + }, + { + "epoch": 1.54, + "learning_rate": 6.907051282051281e-05, + "loss": 0.0277, + "step": 72050 + }, + { + "epoch": 1.54, + "learning_rate": 6.903846153846153e-05, + "loss": 0.0317, + "step": 72060 + }, + { + "epoch": 1.54, + "learning_rate": 6.900641025641025e-05, + "loss": 0.0286, + "step": 72070 + }, + { + "epoch": 1.54, + "learning_rate": 6.897435897435897e-05, + "loss": 0.0304, + "step": 72080 + }, + { + "epoch": 1.54, + "learning_rate": 6.894230769230769e-05, + "loss": 0.0286, + "step": 72090 + }, + { + "epoch": 1.54, + "learning_rate": 6.89102564102564e-05, + "loss": 0.0382, + "step": 72100 + }, + { + "epoch": 1.54, + "learning_rate": 6.887820512820511e-05, + "loss": 0.0303, + "step": 72110 + }, + { + "epoch": 1.54, + "learning_rate": 6.884615384615383e-05, + "loss": 0.0314, + "step": 72120 + }, + { + "epoch": 1.54, + "learning_rate": 6.881410256410255e-05, + "loss": 0.0323, + "step": 72130 + }, + { + "epoch": 1.54, + "learning_rate": 6.878205128205127e-05, + "loss": 0.0287, + "step": 72140 + }, + { + "epoch": 1.54, + "learning_rate": 6.874999999999999e-05, + "loss": 0.0275, + "step": 72150 + }, + { + "epoch": 1.54, + "learning_rate": 6.871794871794871e-05, + "loss": 0.0258, + "step": 72160 + }, + { + "epoch": 1.54, + "learning_rate": 6.868589743589743e-05, + "loss": 0.0337, + "step": 72170 + }, + { + "epoch": 1.54, + "learning_rate": 6.865384615384615e-05, + "loss": 0.0378, + "step": 72180 + }, + { + "epoch": 1.54, + "learning_rate": 6.862179487179487e-05, + "loss": 0.0317, + "step": 72190 + }, + { + "epoch": 1.54, + "learning_rate": 6.858974358974358e-05, + "loss": 0.0271, + "step": 72200 + }, + { + "epoch": 1.54, + "learning_rate": 6.85576923076923e-05, + "loss": 0.0442, + "step": 72210 + }, + { + "epoch": 1.54, + "learning_rate": 6.852564102564101e-05, + "loss": 0.0316, + "step": 72220 + }, + { + "epoch": 1.54, + "learning_rate": 6.849358974358973e-05, + "loss": 0.0328, + "step": 72230 + }, + { + "epoch": 1.54, + "learning_rate": 6.846153846153845e-05, + "loss": 0.0245, + "step": 72240 + }, + { + "epoch": 1.54, + "learning_rate": 6.842948717948717e-05, + "loss": 0.034, + "step": 72250 + }, + { + "epoch": 1.54, + "learning_rate": 6.839743589743589e-05, + "loss": 0.0321, + "step": 72260 + }, + { + "epoch": 1.54, + "learning_rate": 6.836538461538461e-05, + "loss": 0.0274, + "step": 72270 + }, + { + "epoch": 1.54, + "learning_rate": 6.833333333333333e-05, + "loss": 0.0318, + "step": 72280 + }, + { + "epoch": 1.54, + "learning_rate": 6.830128205128204e-05, + "loss": 0.0266, + "step": 72290 + }, + { + "epoch": 1.54, + "learning_rate": 6.826923076923076e-05, + "loss": 0.0282, + "step": 72300 + }, + { + "epoch": 1.55, + "learning_rate": 6.823717948717948e-05, + "loss": 0.0279, + "step": 72310 + }, + { + "epoch": 1.55, + "learning_rate": 6.82051282051282e-05, + "loss": 0.039, + "step": 72320 + }, + { + "epoch": 1.55, + "learning_rate": 6.817307692307692e-05, + "loss": 0.0335, + "step": 72330 + }, + { + "epoch": 1.55, + "learning_rate": 6.814102564102564e-05, + "loss": 0.0336, + "step": 72340 + }, + { + "epoch": 1.55, + "learning_rate": 6.810897435897436e-05, + "loss": 0.028, + "step": 72350 + }, + { + "epoch": 1.55, + "learning_rate": 6.807692307692308e-05, + "loss": 0.0353, + "step": 72360 + }, + { + "epoch": 1.55, + "learning_rate": 6.80448717948718e-05, + "loss": 0.0328, + "step": 72370 + }, + { + "epoch": 1.55, + "learning_rate": 6.80128205128205e-05, + "loss": 0.0295, + "step": 72380 + }, + { + "epoch": 1.55, + "learning_rate": 6.798076923076922e-05, + "loss": 0.0358, + "step": 72390 + }, + { + "epoch": 1.55, + "learning_rate": 6.794871794871794e-05, + "loss": 0.0347, + "step": 72400 + }, + { + "epoch": 1.55, + "learning_rate": 6.791666666666666e-05, + "loss": 0.0303, + "step": 72410 + }, + { + "epoch": 1.55, + "learning_rate": 6.788461538461538e-05, + "loss": 0.0402, + "step": 72420 + }, + { + "epoch": 1.55, + "learning_rate": 6.78525641025641e-05, + "loss": 0.0299, + "step": 72430 + }, + { + "epoch": 1.55, + "learning_rate": 6.782051282051282e-05, + "loss": 0.0312, + "step": 72440 + }, + { + "epoch": 1.55, + "learning_rate": 6.778846153846154e-05, + "loss": 0.0292, + "step": 72450 + }, + { + "epoch": 1.55, + "learning_rate": 6.775641025641026e-05, + "loss": 0.0248, + "step": 72460 + }, + { + "epoch": 1.55, + "learning_rate": 6.772435897435896e-05, + "loss": 0.0314, + "step": 72470 + }, + { + "epoch": 1.55, + "learning_rate": 6.769230769230768e-05, + "loss": 0.0278, + "step": 72480 + }, + { + "epoch": 1.55, + "learning_rate": 6.76602564102564e-05, + "loss": 0.0313, + "step": 72490 + }, + { + "epoch": 1.55, + "learning_rate": 6.762820512820512e-05, + "loss": 0.0318, + "step": 72500 + }, + { + "epoch": 1.55, + "learning_rate": 6.759615384615384e-05, + "loss": 0.0303, + "step": 72510 + }, + { + "epoch": 1.55, + "learning_rate": 6.756410256410256e-05, + "loss": 0.0268, + "step": 72520 + }, + { + "epoch": 1.55, + "learning_rate": 6.753205128205128e-05, + "loss": 0.0297, + "step": 72530 + }, + { + "epoch": 1.55, + "learning_rate": 6.75e-05, + "loss": 0.0328, + "step": 72540 + }, + { + "epoch": 1.55, + "learning_rate": 6.746794871794872e-05, + "loss": 0.0298, + "step": 72550 + }, + { + "epoch": 1.55, + "learning_rate": 6.743589743589743e-05, + "loss": 0.0323, + "step": 72560 + }, + { + "epoch": 1.55, + "learning_rate": 6.740384615384615e-05, + "loss": 0.0269, + "step": 72570 + }, + { + "epoch": 1.55, + "learning_rate": 6.737179487179487e-05, + "loss": 0.034, + "step": 72580 + }, + { + "epoch": 1.55, + "learning_rate": 6.733974358974359e-05, + "loss": 0.0288, + "step": 72590 + }, + { + "epoch": 1.55, + "learning_rate": 6.73076923076923e-05, + "loss": 0.0288, + "step": 72600 + }, + { + "epoch": 1.55, + "learning_rate": 6.727564102564102e-05, + "loss": 0.0307, + "step": 72610 + }, + { + "epoch": 1.55, + "learning_rate": 6.724358974358974e-05, + "loss": 0.0321, + "step": 72620 + }, + { + "epoch": 1.55, + "learning_rate": 6.721153846153846e-05, + "loss": 0.0291, + "step": 72630 + }, + { + "epoch": 1.55, + "learning_rate": 6.717948717948717e-05, + "loss": 0.027, + "step": 72640 + }, + { + "epoch": 1.55, + "learning_rate": 6.714743589743589e-05, + "loss": 0.0295, + "step": 72650 + }, + { + "epoch": 1.55, + "learning_rate": 6.711538461538461e-05, + "loss": 0.026, + "step": 72660 + }, + { + "epoch": 1.55, + "learning_rate": 6.708333333333333e-05, + "loss": 0.0315, + "step": 72670 + }, + { + "epoch": 1.55, + "learning_rate": 6.705128205128205e-05, + "loss": 0.0281, + "step": 72680 + }, + { + "epoch": 1.55, + "learning_rate": 6.701923076923077e-05, + "loss": 0.03, + "step": 72690 + }, + { + "epoch": 1.55, + "learning_rate": 6.698717948717949e-05, + "loss": 0.0343, + "step": 72700 + }, + { + "epoch": 1.55, + "learning_rate": 6.695512820512821e-05, + "loss": 0.0289, + "step": 72710 + }, + { + "epoch": 1.55, + "learning_rate": 6.692307692307693e-05, + "loss": 0.0261, + "step": 72720 + }, + { + "epoch": 1.55, + "learning_rate": 6.689102564102563e-05, + "loss": 0.029, + "step": 72730 + }, + { + "epoch": 1.55, + "learning_rate": 6.685897435897435e-05, + "loss": 0.0293, + "step": 72740 + }, + { + "epoch": 1.55, + "learning_rate": 6.682692307692307e-05, + "loss": 0.0316, + "step": 72750 + }, + { + "epoch": 1.55, + "learning_rate": 6.679487179487179e-05, + "loss": 0.0356, + "step": 72760 + }, + { + "epoch": 1.55, + "learning_rate": 6.67628205128205e-05, + "loss": 0.0324, + "step": 72770 + }, + { + "epoch": 1.56, + "learning_rate": 6.673076923076922e-05, + "loss": 0.0277, + "step": 72780 + }, + { + "epoch": 1.56, + "learning_rate": 6.669871794871794e-05, + "loss": 0.0288, + "step": 72790 + }, + { + "epoch": 1.56, + "learning_rate": 6.666666666666666e-05, + "loss": 0.0373, + "step": 72800 + }, + { + "epoch": 1.56, + "learning_rate": 6.663461538461538e-05, + "loss": 0.0287, + "step": 72810 + }, + { + "epoch": 1.56, + "learning_rate": 6.66025641025641e-05, + "loss": 0.0288, + "step": 72820 + }, + { + "epoch": 1.56, + "learning_rate": 6.657051282051282e-05, + "loss": 0.0317, + "step": 72830 + }, + { + "epoch": 1.56, + "learning_rate": 6.653846153846153e-05, + "loss": 0.0357, + "step": 72840 + }, + { + "epoch": 1.56, + "learning_rate": 6.650641025641024e-05, + "loss": 0.0282, + "step": 72850 + }, + { + "epoch": 1.56, + "learning_rate": 6.647435897435896e-05, + "loss": 0.0295, + "step": 72860 + }, + { + "epoch": 1.56, + "learning_rate": 6.644230769230768e-05, + "loss": 0.0282, + "step": 72870 + }, + { + "epoch": 1.56, + "learning_rate": 6.64102564102564e-05, + "loss": 0.0388, + "step": 72880 + }, + { + "epoch": 1.56, + "learning_rate": 6.637820512820512e-05, + "loss": 0.0268, + "step": 72890 + }, + { + "epoch": 1.56, + "learning_rate": 6.634615384615384e-05, + "loss": 0.0281, + "step": 72900 + }, + { + "epoch": 1.56, + "learning_rate": 6.631410256410256e-05, + "loss": 0.0282, + "step": 72910 + }, + { + "epoch": 1.56, + "learning_rate": 6.628205128205128e-05, + "loss": 0.0262, + "step": 72920 + }, + { + "epoch": 1.56, + "learning_rate": 6.625e-05, + "loss": 0.0304, + "step": 72930 + }, + { + "epoch": 1.56, + "learning_rate": 6.62179487179487e-05, + "loss": 0.0329, + "step": 72940 + }, + { + "epoch": 1.56, + "learning_rate": 6.618589743589742e-05, + "loss": 0.0343, + "step": 72950 + }, + { + "epoch": 1.56, + "learning_rate": 6.615384615384614e-05, + "loss": 0.0311, + "step": 72960 + }, + { + "epoch": 1.56, + "learning_rate": 6.612179487179486e-05, + "loss": 0.0295, + "step": 72970 + }, + { + "epoch": 1.56, + "learning_rate": 6.608974358974358e-05, + "loss": 0.0275, + "step": 72980 + }, + { + "epoch": 1.56, + "learning_rate": 6.60576923076923e-05, + "loss": 0.0329, + "step": 72990 + }, + { + "epoch": 1.56, + "learning_rate": 6.602564102564102e-05, + "loss": 0.0316, + "step": 73000 + }, + { + "epoch": 1.56, + "eval_cer": 0.9235246208630826, + "eval_loss": 0.02337363362312317, + "eval_runtime": 120.3831, + "eval_samples_per_second": 16.614, + "eval_steps_per_second": 4.153, + "step": 73000 + }, + { + "epoch": 1.56, + "learning_rate": 6.599358974358974e-05, + "loss": 0.0281, + "step": 73010 + }, + { + "epoch": 1.56, + "learning_rate": 6.596153846153846e-05, + "loss": 0.0341, + "step": 73020 + }, + { + "epoch": 1.56, + "learning_rate": 6.592948717948717e-05, + "loss": 0.0275, + "step": 73030 + }, + { + "epoch": 1.56, + "learning_rate": 6.589743589743589e-05, + "loss": 0.0317, + "step": 73040 + }, + { + "epoch": 1.56, + "learning_rate": 6.58653846153846e-05, + "loss": 0.0302, + "step": 73050 + }, + { + "epoch": 1.56, + "learning_rate": 6.583333333333333e-05, + "loss": 0.0275, + "step": 73060 + }, + { + "epoch": 1.56, + "learning_rate": 6.580128205128204e-05, + "loss": 0.0415, + "step": 73070 + }, + { + "epoch": 1.56, + "learning_rate": 6.576923076923076e-05, + "loss": 0.0285, + "step": 73080 + }, + { + "epoch": 1.56, + "learning_rate": 6.573717948717948e-05, + "loss": 0.0315, + "step": 73090 + }, + { + "epoch": 1.56, + "learning_rate": 6.57051282051282e-05, + "loss": 0.0324, + "step": 73100 + }, + { + "epoch": 1.56, + "learning_rate": 6.567307692307692e-05, + "loss": 0.0268, + "step": 73110 + }, + { + "epoch": 1.56, + "learning_rate": 6.564102564102563e-05, + "loss": 0.031, + "step": 73120 + }, + { + "epoch": 1.56, + "learning_rate": 6.560897435897435e-05, + "loss": 0.0304, + "step": 73130 + }, + { + "epoch": 1.56, + "learning_rate": 6.557692307692307e-05, + "loss": 0.0303, + "step": 73140 + }, + { + "epoch": 1.56, + "learning_rate": 6.554487179487179e-05, + "loss": 0.0285, + "step": 73150 + }, + { + "epoch": 1.56, + "learning_rate": 6.551282051282051e-05, + "loss": 0.0352, + "step": 73160 + }, + { + "epoch": 1.56, + "learning_rate": 6.548076923076923e-05, + "loss": 0.0269, + "step": 73170 + }, + { + "epoch": 1.56, + "learning_rate": 6.544871794871795e-05, + "loss": 0.0304, + "step": 73180 + }, + { + "epoch": 1.56, + "learning_rate": 6.541666666666667e-05, + "loss": 0.0351, + "step": 73190 + }, + { + "epoch": 1.56, + "learning_rate": 6.538461538461539e-05, + "loss": 0.0311, + "step": 73200 + }, + { + "epoch": 1.56, + "learning_rate": 6.535256410256409e-05, + "loss": 0.0376, + "step": 73210 + }, + { + "epoch": 1.56, + "learning_rate": 6.532051282051281e-05, + "loss": 0.0287, + "step": 73220 + }, + { + "epoch": 1.56, + "learning_rate": 6.528846153846153e-05, + "loss": 0.0331, + "step": 73230 + }, + { + "epoch": 1.56, + "learning_rate": 6.525641025641025e-05, + "loss": 0.0288, + "step": 73240 + }, + { + "epoch": 1.57, + "learning_rate": 6.522435897435897e-05, + "loss": 0.0263, + "step": 73250 + }, + { + "epoch": 1.57, + "learning_rate": 6.519230769230769e-05, + "loss": 0.026, + "step": 73260 + }, + { + "epoch": 1.57, + "learning_rate": 6.516025641025641e-05, + "loss": 0.0265, + "step": 73270 + }, + { + "epoch": 1.57, + "learning_rate": 6.512820512820513e-05, + "loss": 0.0325, + "step": 73280 + }, + { + "epoch": 1.57, + "learning_rate": 6.509615384615385e-05, + "loss": 0.0311, + "step": 73290 + }, + { + "epoch": 1.57, + "learning_rate": 6.506410256410256e-05, + "loss": 0.0327, + "step": 73300 + }, + { + "epoch": 1.57, + "learning_rate": 6.503205128205127e-05, + "loss": 0.0335, + "step": 73310 + }, + { + "epoch": 1.57, + "learning_rate": 6.5e-05, + "loss": 0.0273, + "step": 73320 + }, + { + "epoch": 1.57, + "learning_rate": 6.496794871794871e-05, + "loss": 0.0295, + "step": 73330 + }, + { + "epoch": 1.57, + "learning_rate": 6.493589743589743e-05, + "loss": 0.032, + "step": 73340 + }, + { + "epoch": 1.57, + "learning_rate": 6.490384615384615e-05, + "loss": 0.0259, + "step": 73350 + }, + { + "epoch": 1.57, + "learning_rate": 6.487179487179487e-05, + "loss": 0.0293, + "step": 73360 + }, + { + "epoch": 1.57, + "learning_rate": 6.483974358974359e-05, + "loss": 0.0293, + "step": 73370 + }, + { + "epoch": 1.57, + "learning_rate": 6.48076923076923e-05, + "loss": 0.0313, + "step": 73380 + }, + { + "epoch": 1.57, + "learning_rate": 6.477564102564102e-05, + "loss": 0.0424, + "step": 73390 + }, + { + "epoch": 1.57, + "learning_rate": 6.474358974358974e-05, + "loss": 0.0346, + "step": 73400 + }, + { + "epoch": 1.57, + "learning_rate": 6.471153846153846e-05, + "loss": 0.0299, + "step": 73410 + }, + { + "epoch": 1.57, + "learning_rate": 6.467948717948718e-05, + "loss": 0.0304, + "step": 73420 + }, + { + "epoch": 1.57, + "learning_rate": 6.46474358974359e-05, + "loss": 0.0348, + "step": 73430 + }, + { + "epoch": 1.57, + "learning_rate": 6.461538461538462e-05, + "loss": 0.0293, + "step": 73440 + }, + { + "epoch": 1.57, + "learning_rate": 6.458333333333334e-05, + "loss": 0.0334, + "step": 73450 + }, + { + "epoch": 1.57, + "learning_rate": 6.455128205128206e-05, + "loss": 0.0333, + "step": 73460 + }, + { + "epoch": 1.57, + "learning_rate": 6.451923076923076e-05, + "loss": 0.0304, + "step": 73470 + }, + { + "epoch": 1.57, + "learning_rate": 6.448717948717948e-05, + "loss": 0.0303, + "step": 73480 + }, + { + "epoch": 1.57, + "learning_rate": 6.44551282051282e-05, + "loss": 0.0335, + "step": 73490 + }, + { + "epoch": 1.57, + "learning_rate": 6.442307692307692e-05, + "loss": 0.0332, + "step": 73500 + }, + { + "epoch": 1.57, + "learning_rate": 6.439102564102563e-05, + "loss": 0.0313, + "step": 73510 + }, + { + "epoch": 1.57, + "learning_rate": 6.435897435897435e-05, + "loss": 0.029, + "step": 73520 + }, + { + "epoch": 1.57, + "learning_rate": 6.432692307692307e-05, + "loss": 0.0325, + "step": 73530 + }, + { + "epoch": 1.57, + "learning_rate": 6.429487179487178e-05, + "loss": 0.0317, + "step": 73540 + }, + { + "epoch": 1.57, + "learning_rate": 6.42628205128205e-05, + "loss": 0.034, + "step": 73550 + }, + { + "epoch": 1.57, + "learning_rate": 6.423076923076922e-05, + "loss": 0.0279, + "step": 73560 + }, + { + "epoch": 1.57, + "learning_rate": 6.419871794871794e-05, + "loss": 0.0303, + "step": 73570 + }, + { + "epoch": 1.57, + "learning_rate": 6.416666666666666e-05, + "loss": 0.0282, + "step": 73580 + }, + { + "epoch": 1.57, + "learning_rate": 6.413461538461537e-05, + "loss": 0.0313, + "step": 73590 + }, + { + "epoch": 1.57, + "learning_rate": 6.410256410256409e-05, + "loss": 0.0249, + "step": 73600 + }, + { + "epoch": 1.57, + "learning_rate": 6.407051282051281e-05, + "loss": 0.0307, + "step": 73610 + }, + { + "epoch": 1.57, + "learning_rate": 6.403846153846153e-05, + "loss": 0.0288, + "step": 73620 + }, + { + "epoch": 1.57, + "learning_rate": 6.400641025641025e-05, + "loss": 0.035, + "step": 73630 + }, + { + "epoch": 1.57, + "learning_rate": 6.397435897435897e-05, + "loss": 0.0301, + "step": 73640 + }, + { + "epoch": 1.57, + "learning_rate": 6.394230769230769e-05, + "loss": 0.034, + "step": 73650 + }, + { + "epoch": 1.57, + "learning_rate": 6.39102564102564e-05, + "loss": 0.029, + "step": 73660 + }, + { + "epoch": 1.57, + "learning_rate": 6.387820512820513e-05, + "loss": 0.0349, + "step": 73670 + }, + { + "epoch": 1.57, + "learning_rate": 6.384615384615383e-05, + "loss": 0.0328, + "step": 73680 + }, + { + "epoch": 1.57, + "learning_rate": 6.381410256410255e-05, + "loss": 0.0369, + "step": 73690 + }, + { + "epoch": 1.57, + "learning_rate": 6.378205128205127e-05, + "loss": 0.032, + "step": 73700 + }, + { + "epoch": 1.57, + "learning_rate": 6.374999999999999e-05, + "loss": 0.0319, + "step": 73710 + }, + { + "epoch": 1.58, + "learning_rate": 6.371794871794871e-05, + "loss": 0.0309, + "step": 73720 + }, + { + "epoch": 1.58, + "learning_rate": 6.368589743589743e-05, + "loss": 0.0322, + "step": 73730 + }, + { + "epoch": 1.58, + "learning_rate": 6.365384615384615e-05, + "loss": 0.0305, + "step": 73740 + }, + { + "epoch": 1.58, + "learning_rate": 6.362179487179487e-05, + "loss": 0.0322, + "step": 73750 + }, + { + "epoch": 1.58, + "learning_rate": 6.358974358974359e-05, + "loss": 0.0284, + "step": 73760 + }, + { + "epoch": 1.58, + "learning_rate": 6.35576923076923e-05, + "loss": 0.0287, + "step": 73770 + }, + { + "epoch": 1.58, + "learning_rate": 6.352564102564101e-05, + "loss": 0.0234, + "step": 73780 + }, + { + "epoch": 1.58, + "learning_rate": 6.349358974358973e-05, + "loss": 0.0287, + "step": 73790 + }, + { + "epoch": 1.58, + "learning_rate": 6.346153846153845e-05, + "loss": 0.0311, + "step": 73800 + }, + { + "epoch": 1.58, + "learning_rate": 6.342948717948717e-05, + "loss": 0.0304, + "step": 73810 + }, + { + "epoch": 1.58, + "learning_rate": 6.339743589743589e-05, + "loss": 0.0374, + "step": 73820 + }, + { + "epoch": 1.58, + "learning_rate": 6.336538461538461e-05, + "loss": 0.0338, + "step": 73830 + }, + { + "epoch": 1.58, + "learning_rate": 6.333333333333333e-05, + "loss": 0.0324, + "step": 73840 + }, + { + "epoch": 1.58, + "learning_rate": 6.330128205128205e-05, + "loss": 0.0357, + "step": 73850 + }, + { + "epoch": 1.58, + "learning_rate": 6.326923076923076e-05, + "loss": 0.033, + "step": 73860 + }, + { + "epoch": 1.58, + "learning_rate": 6.323717948717948e-05, + "loss": 0.027, + "step": 73870 + }, + { + "epoch": 1.58, + "learning_rate": 6.32051282051282e-05, + "loss": 0.032, + "step": 73880 + }, + { + "epoch": 1.58, + "learning_rate": 6.317307692307692e-05, + "loss": 0.0362, + "step": 73890 + }, + { + "epoch": 1.58, + "learning_rate": 6.314102564102564e-05, + "loss": 0.0277, + "step": 73900 + }, + { + "epoch": 1.58, + "learning_rate": 6.310897435897436e-05, + "loss": 0.0275, + "step": 73910 + }, + { + "epoch": 1.58, + "learning_rate": 6.307692307692308e-05, + "loss": 0.0298, + "step": 73920 + }, + { + "epoch": 1.58, + "learning_rate": 6.30448717948718e-05, + "loss": 0.0276, + "step": 73930 + }, + { + "epoch": 1.58, + "learning_rate": 6.301282051282051e-05, + "loss": 0.0267, + "step": 73940 + }, + { + "epoch": 1.58, + "learning_rate": 6.298076923076922e-05, + "loss": 0.0331, + "step": 73950 + }, + { + "epoch": 1.58, + "learning_rate": 6.294871794871794e-05, + "loss": 0.0309, + "step": 73960 + }, + { + "epoch": 1.58, + "learning_rate": 6.291666666666666e-05, + "loss": 0.0308, + "step": 73970 + }, + { + "epoch": 1.58, + "learning_rate": 6.288461538461538e-05, + "loss": 0.0377, + "step": 73980 + }, + { + "epoch": 1.58, + "learning_rate": 6.28525641025641e-05, + "loss": 0.0369, + "step": 73990 + }, + { + "epoch": 1.58, + "learning_rate": 6.282051282051282e-05, + "loss": 0.0308, + "step": 74000 + }, + { + "epoch": 1.58, + "eval_cer": 0.9235704816314884, + "eval_loss": 0.022941503673791885, + "eval_runtime": 120.8675, + "eval_samples_per_second": 16.547, + "eval_steps_per_second": 4.137, + "step": 74000 + }, + { + "epoch": 1.58, + "learning_rate": 6.278846153846154e-05, + "loss": 0.0256, + "step": 74010 + }, + { + "epoch": 1.58, + "learning_rate": 6.275641025641026e-05, + "loss": 0.0251, + "step": 74020 + }, + { + "epoch": 1.58, + "learning_rate": 6.272435897435896e-05, + "loss": 0.0291, + "step": 74030 + }, + { + "epoch": 1.58, + "learning_rate": 6.269230769230768e-05, + "loss": 0.0308, + "step": 74040 + }, + { + "epoch": 1.58, + "learning_rate": 6.26602564102564e-05, + "loss": 0.0305, + "step": 74050 + }, + { + "epoch": 1.58, + "learning_rate": 6.262820512820512e-05, + "loss": 0.0302, + "step": 74060 + }, + { + "epoch": 1.58, + "learning_rate": 6.259615384615384e-05, + "loss": 0.0243, + "step": 74070 + }, + { + "epoch": 1.58, + "learning_rate": 6.256410256410256e-05, + "loss": 0.0298, + "step": 74080 + }, + { + "epoch": 1.58, + "learning_rate": 6.253205128205128e-05, + "loss": 0.0287, + "step": 74090 + }, + { + "epoch": 1.58, + "learning_rate": 6.25e-05, + "loss": 0.0291, + "step": 74100 + }, + { + "epoch": 1.58, + "learning_rate": 6.246794871794872e-05, + "loss": 0.0308, + "step": 74110 + }, + { + "epoch": 1.58, + "learning_rate": 6.243589743589743e-05, + "loss": 0.0306, + "step": 74120 + }, + { + "epoch": 1.58, + "learning_rate": 6.240384615384615e-05, + "loss": 0.0319, + "step": 74130 + }, + { + "epoch": 1.58, + "learning_rate": 6.237179487179487e-05, + "loss": 0.0275, + "step": 74140 + }, + { + "epoch": 1.58, + "learning_rate": 6.233974358974359e-05, + "loss": 0.0312, + "step": 74150 + }, + { + "epoch": 1.58, + "learning_rate": 6.23076923076923e-05, + "loss": 0.0334, + "step": 74160 + }, + { + "epoch": 1.58, + "learning_rate": 6.227564102564102e-05, + "loss": 0.0244, + "step": 74170 + }, + { + "epoch": 1.59, + "learning_rate": 6.224358974358974e-05, + "loss": 0.0293, + "step": 74180 + }, + { + "epoch": 1.59, + "learning_rate": 6.221153846153846e-05, + "loss": 0.0311, + "step": 74190 + }, + { + "epoch": 1.59, + "learning_rate": 6.217948717948718e-05, + "loss": 0.0269, + "step": 74200 + }, + { + "epoch": 1.59, + "learning_rate": 6.214743589743589e-05, + "loss": 0.0254, + "step": 74210 + }, + { + "epoch": 1.59, + "learning_rate": 6.211538461538461e-05, + "loss": 0.0319, + "step": 74220 + }, + { + "epoch": 1.59, + "learning_rate": 6.208333333333333e-05, + "loss": 0.026, + "step": 74230 + }, + { + "epoch": 1.59, + "learning_rate": 6.205128205128205e-05, + "loss": 0.0304, + "step": 74240 + }, + { + "epoch": 1.59, + "learning_rate": 6.201923076923075e-05, + "loss": 0.029, + "step": 74250 + }, + { + "epoch": 1.59, + "learning_rate": 6.198717948717947e-05, + "loss": 0.027, + "step": 74260 + }, + { + "epoch": 1.59, + "learning_rate": 6.19551282051282e-05, + "loss": 0.0246, + "step": 74270 + }, + { + "epoch": 1.59, + "learning_rate": 6.192307692307691e-05, + "loss": 0.0356, + "step": 74280 + }, + { + "epoch": 1.59, + "learning_rate": 6.189102564102563e-05, + "loss": 0.028, + "step": 74290 + }, + { + "epoch": 1.59, + "learning_rate": 6.185897435897435e-05, + "loss": 0.0344, + "step": 74300 + }, + { + "epoch": 1.59, + "learning_rate": 6.182692307692307e-05, + "loss": 0.0298, + "step": 74310 + }, + { + "epoch": 1.59, + "learning_rate": 6.179487179487179e-05, + "loss": 0.0267, + "step": 74320 + }, + { + "epoch": 1.59, + "learning_rate": 6.17628205128205e-05, + "loss": 0.0276, + "step": 74330 + }, + { + "epoch": 1.59, + "learning_rate": 6.173076923076922e-05, + "loss": 0.0297, + "step": 74340 + }, + { + "epoch": 1.59, + "learning_rate": 6.169871794871794e-05, + "loss": 0.0352, + "step": 74350 + }, + { + "epoch": 1.59, + "learning_rate": 6.166666666666666e-05, + "loss": 0.0329, + "step": 74360 + }, + { + "epoch": 1.59, + "learning_rate": 6.163461538461538e-05, + "loss": 0.0324, + "step": 74370 + }, + { + "epoch": 1.59, + "learning_rate": 6.16025641025641e-05, + "loss": 0.0338, + "step": 74380 + }, + { + "epoch": 1.59, + "learning_rate": 6.157051282051282e-05, + "loss": 0.0348, + "step": 74390 + }, + { + "epoch": 1.59, + "learning_rate": 6.153846153846154e-05, + "loss": 0.0307, + "step": 74400 + }, + { + "epoch": 1.59, + "learning_rate": 6.150641025641025e-05, + "loss": 0.0295, + "step": 74410 + }, + { + "epoch": 1.59, + "learning_rate": 6.147435897435896e-05, + "loss": 0.0248, + "step": 74420 + }, + { + "epoch": 1.59, + "learning_rate": 6.144230769230768e-05, + "loss": 0.0277, + "step": 74430 + }, + { + "epoch": 1.59, + "learning_rate": 6.14102564102564e-05, + "loss": 0.0262, + "step": 74440 + }, + { + "epoch": 1.59, + "learning_rate": 6.137820512820512e-05, + "loss": 0.0272, + "step": 74450 + }, + { + "epoch": 1.59, + "learning_rate": 6.134615384615384e-05, + "loss": 0.0298, + "step": 74460 + }, + { + "epoch": 1.59, + "learning_rate": 6.131410256410256e-05, + "loss": 0.0282, + "step": 74470 + }, + { + "epoch": 1.59, + "learning_rate": 6.128205128205128e-05, + "loss": 0.0256, + "step": 74480 + }, + { + "epoch": 1.59, + "learning_rate": 6.125e-05, + "loss": 0.0363, + "step": 74490 + }, + { + "epoch": 1.59, + "learning_rate": 6.121794871794872e-05, + "loss": 0.0326, + "step": 74500 + }, + { + "epoch": 1.59, + "learning_rate": 6.118589743589742e-05, + "loss": 0.0277, + "step": 74510 + }, + { + "epoch": 1.59, + "learning_rate": 6.115384615384614e-05, + "loss": 0.0302, + "step": 74520 + }, + { + "epoch": 1.59, + "learning_rate": 6.112179487179486e-05, + "loss": 0.031, + "step": 74530 + }, + { + "epoch": 1.59, + "learning_rate": 6.108974358974358e-05, + "loss": 0.0341, + "step": 74540 + }, + { + "epoch": 1.59, + "learning_rate": 6.10576923076923e-05, + "loss": 0.0321, + "step": 74550 + }, + { + "epoch": 1.59, + "learning_rate": 6.102564102564102e-05, + "loss": 0.0308, + "step": 74560 + }, + { + "epoch": 1.59, + "learning_rate": 6.099358974358974e-05, + "loss": 0.0297, + "step": 74570 + }, + { + "epoch": 1.59, + "learning_rate": 6.0961538461538454e-05, + "loss": 0.0306, + "step": 74580 + }, + { + "epoch": 1.59, + "learning_rate": 6.0929487179487174e-05, + "loss": 0.0244, + "step": 74590 + }, + { + "epoch": 1.59, + "learning_rate": 6.089743589743589e-05, + "loss": 0.0274, + "step": 74600 + }, + { + "epoch": 1.59, + "learning_rate": 6.086538461538461e-05, + "loss": 0.0303, + "step": 74610 + }, + { + "epoch": 1.59, + "learning_rate": 6.0833333333333326e-05, + "loss": 0.0307, + "step": 74620 + }, + { + "epoch": 1.59, + "learning_rate": 6.0801282051282045e-05, + "loss": 0.0321, + "step": 74630 + }, + { + "epoch": 1.59, + "learning_rate": 6.0769230769230765e-05, + "loss": 0.0281, + "step": 74640 + }, + { + "epoch": 1.6, + "learning_rate": 6.0737179487179485e-05, + "loss": 0.0317, + "step": 74650 + }, + { + "epoch": 1.6, + "learning_rate": 6.07051282051282e-05, + "loss": 0.0287, + "step": 74660 + }, + { + "epoch": 1.6, + "learning_rate": 6.067307692307692e-05, + "loss": 0.0253, + "step": 74670 + }, + { + "epoch": 1.6, + "learning_rate": 6.0641025641025637e-05, + "loss": 0.0278, + "step": 74680 + }, + { + "epoch": 1.6, + "learning_rate": 6.0608974358974356e-05, + "loss": 0.0307, + "step": 74690 + }, + { + "epoch": 1.6, + "learning_rate": 6.0576923076923076e-05, + "loss": 0.0309, + "step": 74700 + }, + { + "epoch": 1.6, + "learning_rate": 6.054487179487179e-05, + "loss": 0.0295, + "step": 74710 + }, + { + "epoch": 1.6, + "learning_rate": 6.051282051282051e-05, + "loss": 0.0305, + "step": 74720 + }, + { + "epoch": 1.6, + "learning_rate": 6.048076923076923e-05, + "loss": 0.0329, + "step": 74730 + }, + { + "epoch": 1.6, + "learning_rate": 6.044871794871795e-05, + "loss": 0.0259, + "step": 74740 + }, + { + "epoch": 1.6, + "learning_rate": 6.041666666666666e-05, + "loss": 0.0304, + "step": 74750 + }, + { + "epoch": 1.6, + "learning_rate": 6.038461538461538e-05, + "loss": 0.0296, + "step": 74760 + }, + { + "epoch": 1.6, + "learning_rate": 6.03525641025641e-05, + "loss": 0.0289, + "step": 74770 + }, + { + "epoch": 1.6, + "learning_rate": 6.032051282051282e-05, + "loss": 0.0285, + "step": 74780 + }, + { + "epoch": 1.6, + "learning_rate": 6.028846153846154e-05, + "loss": 0.0321, + "step": 74790 + }, + { + "epoch": 1.6, + "learning_rate": 6.025641025641025e-05, + "loss": 0.0365, + "step": 74800 + }, + { + "epoch": 1.6, + "learning_rate": 6.022435897435897e-05, + "loss": 0.0286, + "step": 74810 + }, + { + "epoch": 1.6, + "learning_rate": 6.019230769230769e-05, + "loss": 0.0269, + "step": 74820 + }, + { + "epoch": 1.6, + "learning_rate": 6.016025641025641e-05, + "loss": 0.0297, + "step": 74830 + }, + { + "epoch": 1.6, + "learning_rate": 6.012820512820512e-05, + "loss": 0.0322, + "step": 74840 + }, + { + "epoch": 1.6, + "learning_rate": 6.009615384615384e-05, + "loss": 0.0271, + "step": 74850 + }, + { + "epoch": 1.6, + "learning_rate": 6.006410256410256e-05, + "loss": 0.0275, + "step": 74860 + }, + { + "epoch": 1.6, + "learning_rate": 6.003205128205128e-05, + "loss": 0.0329, + "step": 74870 + }, + { + "epoch": 1.6, + "learning_rate": 5.9999999999999995e-05, + "loss": 0.0263, + "step": 74880 + }, + { + "epoch": 1.6, + "learning_rate": 5.9967948717948714e-05, + "loss": 0.0297, + "step": 74890 + }, + { + "epoch": 1.6, + "learning_rate": 5.9935897435897434e-05, + "loss": 0.0346, + "step": 74900 + }, + { + "epoch": 1.6, + "learning_rate": 5.9903846153846154e-05, + "loss": 0.034, + "step": 74910 + }, + { + "epoch": 1.6, + "learning_rate": 5.987179487179487e-05, + "loss": 0.0331, + "step": 74920 + }, + { + "epoch": 1.6, + "learning_rate": 5.9839743589743586e-05, + "loss": 0.0286, + "step": 74930 + }, + { + "epoch": 1.6, + "learning_rate": 5.9807692307692306e-05, + "loss": 0.0278, + "step": 74940 + }, + { + "epoch": 1.6, + "learning_rate": 5.9775641025641025e-05, + "loss": 0.0338, + "step": 74950 + }, + { + "epoch": 1.6, + "learning_rate": 5.974358974358973e-05, + "loss": 0.0326, + "step": 74960 + }, + { + "epoch": 1.6, + "learning_rate": 5.971153846153845e-05, + "loss": 0.0302, + "step": 74970 + }, + { + "epoch": 1.6, + "learning_rate": 5.967948717948717e-05, + "loss": 0.0232, + "step": 74980 + }, + { + "epoch": 1.6, + "learning_rate": 5.964743589743589e-05, + "loss": 0.0297, + "step": 74990 + }, + { + "epoch": 1.6, + "learning_rate": 5.961538461538461e-05, + "loss": 0.0291, + "step": 75000 + }, + { + "epoch": 1.6, + "eval_cer": 0.9235755772724223, + "eval_loss": 0.02294476516544819, + "eval_runtime": 120.5102, + "eval_samples_per_second": 16.596, + "eval_steps_per_second": 4.149, + "step": 75000 + }, + { + "epoch": 1.6, + "learning_rate": 5.958333333333332e-05, + "loss": 0.0257, + "step": 75010 + }, + { + "epoch": 1.6, + "learning_rate": 5.955128205128204e-05, + "loss": 0.034, + "step": 75020 + }, + { + "epoch": 1.6, + "learning_rate": 5.951923076923076e-05, + "loss": 0.0313, + "step": 75030 + }, + { + "epoch": 1.6, + "learning_rate": 5.948717948717948e-05, + "loss": 0.0306, + "step": 75040 + }, + { + "epoch": 1.6, + "learning_rate": 5.9455128205128194e-05, + "loss": 0.0317, + "step": 75050 + }, + { + "epoch": 1.6, + "learning_rate": 5.9423076923076914e-05, + "loss": 0.0305, + "step": 75060 + }, + { + "epoch": 1.6, + "learning_rate": 5.939102564102563e-05, + "loss": 0.0299, + "step": 75070 + }, + { + "epoch": 1.6, + "learning_rate": 5.935897435897435e-05, + "loss": 0.0348, + "step": 75080 + }, + { + "epoch": 1.6, + "learning_rate": 5.932692307692307e-05, + "loss": 0.0319, + "step": 75090 + }, + { + "epoch": 1.6, + "learning_rate": 5.9294871794871785e-05, + "loss": 0.0291, + "step": 75100 + }, + { + "epoch": 1.6, + "learning_rate": 5.9262820512820505e-05, + "loss": 0.0276, + "step": 75110 + }, + { + "epoch": 1.61, + "learning_rate": 5.9230769230769225e-05, + "loss": 0.0352, + "step": 75120 + }, + { + "epoch": 1.61, + "learning_rate": 5.9198717948717944e-05, + "loss": 0.0271, + "step": 75130 + }, + { + "epoch": 1.61, + "learning_rate": 5.916666666666666e-05, + "loss": 0.0268, + "step": 75140 + }, + { + "epoch": 1.61, + "learning_rate": 5.9134615384615377e-05, + "loss": 0.031, + "step": 75150 + }, + { + "epoch": 1.61, + "learning_rate": 5.9102564102564096e-05, + "loss": 0.0279, + "step": 75160 + }, + { + "epoch": 1.61, + "learning_rate": 5.9070512820512816e-05, + "loss": 0.0301, + "step": 75170 + }, + { + "epoch": 1.61, + "learning_rate": 5.903846153846153e-05, + "loss": 0.0271, + "step": 75180 + }, + { + "epoch": 1.61, + "learning_rate": 5.900641025641025e-05, + "loss": 0.0319, + "step": 75190 + }, + { + "epoch": 1.61, + "learning_rate": 5.897435897435897e-05, + "loss": 0.0344, + "step": 75200 + }, + { + "epoch": 1.61, + "learning_rate": 5.894230769230769e-05, + "loss": 0.0247, + "step": 75210 + }, + { + "epoch": 1.61, + "learning_rate": 5.891025641025641e-05, + "loss": 0.0314, + "step": 75220 + }, + { + "epoch": 1.61, + "learning_rate": 5.887820512820512e-05, + "loss": 0.0372, + "step": 75230 + }, + { + "epoch": 1.61, + "learning_rate": 5.884615384615384e-05, + "loss": 0.0296, + "step": 75240 + }, + { + "epoch": 1.61, + "learning_rate": 5.881410256410256e-05, + "loss": 0.0286, + "step": 75250 + }, + { + "epoch": 1.61, + "learning_rate": 5.878205128205128e-05, + "loss": 0.0284, + "step": 75260 + }, + { + "epoch": 1.61, + "learning_rate": 5.874999999999999e-05, + "loss": 0.0309, + "step": 75270 + }, + { + "epoch": 1.61, + "learning_rate": 5.871794871794871e-05, + "loss": 0.026, + "step": 75280 + }, + { + "epoch": 1.61, + "learning_rate": 5.868589743589743e-05, + "loss": 0.0322, + "step": 75290 + }, + { + "epoch": 1.61, + "learning_rate": 5.865384615384615e-05, + "loss": 0.031, + "step": 75300 + }, + { + "epoch": 1.61, + "learning_rate": 5.862179487179487e-05, + "loss": 0.034, + "step": 75310 + }, + { + "epoch": 1.61, + "learning_rate": 5.858974358974358e-05, + "loss": 0.0292, + "step": 75320 + }, + { + "epoch": 1.61, + "learning_rate": 5.85576923076923e-05, + "loss": 0.0269, + "step": 75330 + }, + { + "epoch": 1.61, + "learning_rate": 5.852564102564102e-05, + "loss": 0.0313, + "step": 75340 + }, + { + "epoch": 1.61, + "learning_rate": 5.849358974358974e-05, + "loss": 0.0288, + "step": 75350 + }, + { + "epoch": 1.61, + "learning_rate": 5.8461538461538454e-05, + "loss": 0.0264, + "step": 75360 + }, + { + "epoch": 1.61, + "learning_rate": 5.8429487179487174e-05, + "loss": 0.023, + "step": 75370 + }, + { + "epoch": 1.61, + "learning_rate": 5.8397435897435894e-05, + "loss": 0.0303, + "step": 75380 + }, + { + "epoch": 1.61, + "learning_rate": 5.836538461538461e-05, + "loss": 0.0271, + "step": 75390 + }, + { + "epoch": 1.61, + "learning_rate": 5.8333333333333326e-05, + "loss": 0.0298, + "step": 75400 + }, + { + "epoch": 1.61, + "learning_rate": 5.8301282051282046e-05, + "loss": 0.0292, + "step": 75410 + }, + { + "epoch": 1.61, + "learning_rate": 5.8269230769230765e-05, + "loss": 0.0333, + "step": 75420 + }, + { + "epoch": 1.61, + "learning_rate": 5.8237179487179485e-05, + "loss": 0.0261, + "step": 75430 + }, + { + "epoch": 1.61, + "learning_rate": 5.8205128205128204e-05, + "loss": 0.0277, + "step": 75440 + }, + { + "epoch": 1.61, + "learning_rate": 5.817307692307692e-05, + "loss": 0.0302, + "step": 75450 + }, + { + "epoch": 1.61, + "learning_rate": 5.814102564102564e-05, + "loss": 0.0344, + "step": 75460 + }, + { + "epoch": 1.61, + "learning_rate": 5.8108974358974356e-05, + "loss": 0.0281, + "step": 75470 + }, + { + "epoch": 1.61, + "learning_rate": 5.8076923076923076e-05, + "loss": 0.0254, + "step": 75480 + }, + { + "epoch": 1.61, + "learning_rate": 5.804487179487179e-05, + "loss": 0.0259, + "step": 75490 + }, + { + "epoch": 1.61, + "learning_rate": 5.801282051282051e-05, + "loss": 0.0305, + "step": 75500 + }, + { + "epoch": 1.61, + "learning_rate": 5.798076923076923e-05, + "loss": 0.0326, + "step": 75510 + }, + { + "epoch": 1.61, + "learning_rate": 5.794871794871795e-05, + "loss": 0.0294, + "step": 75520 + }, + { + "epoch": 1.61, + "learning_rate": 5.791666666666667e-05, + "loss": 0.0306, + "step": 75530 + }, + { + "epoch": 1.61, + "learning_rate": 5.788461538461538e-05, + "loss": 0.0284, + "step": 75540 + }, + { + "epoch": 1.61, + "learning_rate": 5.78525641025641e-05, + "loss": 0.0377, + "step": 75550 + }, + { + "epoch": 1.61, + "learning_rate": 5.782051282051282e-05, + "loss": 0.0301, + "step": 75560 + }, + { + "epoch": 1.61, + "learning_rate": 5.778846153846154e-05, + "loss": 0.0339, + "step": 75570 + }, + { + "epoch": 1.61, + "learning_rate": 5.775641025641025e-05, + "loss": 0.0309, + "step": 75580 + }, + { + "epoch": 1.62, + "learning_rate": 5.772435897435897e-05, + "loss": 0.0289, + "step": 75590 + }, + { + "epoch": 1.62, + "learning_rate": 5.769230769230769e-05, + "loss": 0.0251, + "step": 75600 + }, + { + "epoch": 1.62, + "learning_rate": 5.766025641025641e-05, + "loss": 0.0318, + "step": 75610 + }, + { + "epoch": 1.62, + "learning_rate": 5.762820512820512e-05, + "loss": 0.0352, + "step": 75620 + }, + { + "epoch": 1.62, + "learning_rate": 5.759615384615384e-05, + "loss": 0.0299, + "step": 75630 + }, + { + "epoch": 1.62, + "learning_rate": 5.756410256410256e-05, + "loss": 0.0295, + "step": 75640 + }, + { + "epoch": 1.62, + "learning_rate": 5.753205128205128e-05, + "loss": 0.0272, + "step": 75650 + }, + { + "epoch": 1.62, + "learning_rate": 5.75e-05, + "loss": 0.0314, + "step": 75660 + }, + { + "epoch": 1.62, + "learning_rate": 5.7467948717948715e-05, + "loss": 0.0301, + "step": 75670 + }, + { + "epoch": 1.62, + "learning_rate": 5.7435897435897434e-05, + "loss": 0.0283, + "step": 75680 + }, + { + "epoch": 1.62, + "learning_rate": 5.740384615384615e-05, + "loss": 0.0361, + "step": 75690 + }, + { + "epoch": 1.62, + "learning_rate": 5.737179487179486e-05, + "loss": 0.0364, + "step": 75700 + }, + { + "epoch": 1.62, + "learning_rate": 5.733974358974358e-05, + "loss": 0.0284, + "step": 75710 + }, + { + "epoch": 1.62, + "learning_rate": 5.73076923076923e-05, + "loss": 0.0318, + "step": 75720 + }, + { + "epoch": 1.62, + "learning_rate": 5.727564102564102e-05, + "loss": 0.0302, + "step": 75730 + }, + { + "epoch": 1.62, + "learning_rate": 5.724358974358974e-05, + "loss": 0.0322, + "step": 75740 + }, + { + "epoch": 1.62, + "learning_rate": 5.721153846153845e-05, + "loss": 0.0261, + "step": 75750 + }, + { + "epoch": 1.62, + "learning_rate": 5.717948717948717e-05, + "loss": 0.0319, + "step": 75760 + }, + { + "epoch": 1.62, + "learning_rate": 5.714743589743589e-05, + "loss": 0.0294, + "step": 75770 + }, + { + "epoch": 1.62, + "learning_rate": 5.711538461538461e-05, + "loss": 0.0268, + "step": 75780 + }, + { + "epoch": 1.62, + "learning_rate": 5.708333333333332e-05, + "loss": 0.0335, + "step": 75790 + }, + { + "epoch": 1.62, + "learning_rate": 5.705128205128204e-05, + "loss": 0.029, + "step": 75800 + }, + { + "epoch": 1.62, + "learning_rate": 5.701923076923076e-05, + "loss": 0.031, + "step": 75810 + }, + { + "epoch": 1.62, + "learning_rate": 5.698717948717948e-05, + "loss": 0.0293, + "step": 75820 + }, + { + "epoch": 1.62, + "learning_rate": 5.6955128205128194e-05, + "loss": 0.029, + "step": 75830 + }, + { + "epoch": 1.62, + "learning_rate": 5.6923076923076914e-05, + "loss": 0.0262, + "step": 75840 + }, + { + "epoch": 1.62, + "learning_rate": 5.6891025641025634e-05, + "loss": 0.0322, + "step": 75850 + }, + { + "epoch": 1.62, + "learning_rate": 5.685897435897435e-05, + "loss": 0.0287, + "step": 75860 + }, + { + "epoch": 1.62, + "learning_rate": 5.682692307692307e-05, + "loss": 0.0294, + "step": 75870 + }, + { + "epoch": 1.62, + "learning_rate": 5.6794871794871786e-05, + "loss": 0.026, + "step": 75880 + }, + { + "epoch": 1.62, + "learning_rate": 5.6762820512820505e-05, + "loss": 0.0358, + "step": 75890 + }, + { + "epoch": 1.62, + "learning_rate": 5.6730769230769225e-05, + "loss": 0.0276, + "step": 75900 + }, + { + "epoch": 1.62, + "learning_rate": 5.6698717948717944e-05, + "loss": 0.0296, + "step": 75910 + }, + { + "epoch": 1.62, + "learning_rate": 5.666666666666666e-05, + "loss": 0.0318, + "step": 75920 + }, + { + "epoch": 1.62, + "learning_rate": 5.663461538461538e-05, + "loss": 0.0322, + "step": 75930 + }, + { + "epoch": 1.62, + "learning_rate": 5.6602564102564096e-05, + "loss": 0.0274, + "step": 75940 + }, + { + "epoch": 1.62, + "learning_rate": 5.6570512820512816e-05, + "loss": 0.029, + "step": 75950 + }, + { + "epoch": 1.62, + "learning_rate": 5.6538461538461536e-05, + "loss": 0.0296, + "step": 75960 + }, + { + "epoch": 1.62, + "learning_rate": 5.650641025641025e-05, + "loss": 0.0292, + "step": 75970 + }, + { + "epoch": 1.62, + "learning_rate": 5.647435897435897e-05, + "loss": 0.0347, + "step": 75980 + }, + { + "epoch": 1.62, + "learning_rate": 5.644230769230769e-05, + "loss": 0.0283, + "step": 75990 + }, + { + "epoch": 1.62, + "learning_rate": 5.641025641025641e-05, + "loss": 0.0325, + "step": 76000 + }, + { + "epoch": 1.62, + "eval_cer": 0.9235509483412415, + "eval_loss": 0.02291790582239628, + "eval_runtime": 120.5514, + "eval_samples_per_second": 16.59, + "eval_steps_per_second": 4.148, + "step": 76000 + }, + { + "epoch": 1.62, + "learning_rate": 5.637820512820512e-05, + "loss": 0.0302, + "step": 76010 + }, + { + "epoch": 1.62, + "learning_rate": 5.634615384615384e-05, + "loss": 0.029, + "step": 76020 + }, + { + "epoch": 1.62, + "learning_rate": 5.631410256410256e-05, + "loss": 0.0327, + "step": 76030 + }, + { + "epoch": 1.62, + "learning_rate": 5.628205128205128e-05, + "loss": 0.0293, + "step": 76040 + }, + { + "epoch": 1.62, + "learning_rate": 5.625e-05, + "loss": 0.0288, + "step": 76050 + }, + { + "epoch": 1.63, + "learning_rate": 5.621794871794871e-05, + "loss": 0.0318, + "step": 76060 + }, + { + "epoch": 1.63, + "learning_rate": 5.618589743589743e-05, + "loss": 0.0324, + "step": 76070 + }, + { + "epoch": 1.63, + "learning_rate": 5.615384615384615e-05, + "loss": 0.0328, + "step": 76080 + }, + { + "epoch": 1.63, + "learning_rate": 5.612179487179487e-05, + "loss": 0.0337, + "step": 76090 + }, + { + "epoch": 1.63, + "learning_rate": 5.608974358974358e-05, + "loss": 0.0288, + "step": 76100 + }, + { + "epoch": 1.63, + "learning_rate": 5.60576923076923e-05, + "loss": 0.0338, + "step": 76110 + }, + { + "epoch": 1.63, + "learning_rate": 5.602564102564102e-05, + "loss": 0.0265, + "step": 76120 + }, + { + "epoch": 1.63, + "learning_rate": 5.599358974358974e-05, + "loss": 0.0282, + "step": 76130 + }, + { + "epoch": 1.63, + "learning_rate": 5.5961538461538455e-05, + "loss": 0.0239, + "step": 76140 + }, + { + "epoch": 1.63, + "learning_rate": 5.5929487179487174e-05, + "loss": 0.0246, + "step": 76150 + }, + { + "epoch": 1.63, + "learning_rate": 5.5897435897435894e-05, + "loss": 0.0285, + "step": 76160 + }, + { + "epoch": 1.63, + "learning_rate": 5.586538461538461e-05, + "loss": 0.0346, + "step": 76170 + }, + { + "epoch": 1.63, + "learning_rate": 5.583333333333333e-05, + "loss": 0.0279, + "step": 76180 + }, + { + "epoch": 1.63, + "learning_rate": 5.5801282051282046e-05, + "loss": 0.0301, + "step": 76190 + }, + { + "epoch": 1.63, + "learning_rate": 5.5769230769230765e-05, + "loss": 0.0314, + "step": 76200 + }, + { + "epoch": 1.63, + "learning_rate": 5.5737179487179485e-05, + "loss": 0.031, + "step": 76210 + }, + { + "epoch": 1.63, + "learning_rate": 5.5705128205128205e-05, + "loss": 0.0316, + "step": 76220 + }, + { + "epoch": 1.63, + "learning_rate": 5.567307692307692e-05, + "loss": 0.0314, + "step": 76230 + }, + { + "epoch": 1.63, + "learning_rate": 5.564102564102564e-05, + "loss": 0.0275, + "step": 76240 + }, + { + "epoch": 1.63, + "learning_rate": 5.560897435897436e-05, + "loss": 0.0304, + "step": 76250 + }, + { + "epoch": 1.63, + "learning_rate": 5.5576923076923076e-05, + "loss": 0.0292, + "step": 76260 + }, + { + "epoch": 1.63, + "learning_rate": 5.5544871794871796e-05, + "loss": 0.0261, + "step": 76270 + }, + { + "epoch": 1.63, + "learning_rate": 5.551282051282051e-05, + "loss": 0.0283, + "step": 76280 + }, + { + "epoch": 1.63, + "learning_rate": 5.548076923076923e-05, + "loss": 0.0319, + "step": 76290 + }, + { + "epoch": 1.63, + "learning_rate": 5.544871794871795e-05, + "loss": 0.0336, + "step": 76300 + }, + { + "epoch": 1.63, + "learning_rate": 5.541666666666667e-05, + "loss": 0.0353, + "step": 76310 + }, + { + "epoch": 1.63, + "learning_rate": 5.538461538461538e-05, + "loss": 0.044, + "step": 76320 + }, + { + "epoch": 1.63, + "learning_rate": 5.53525641025641e-05, + "loss": 0.0301, + "step": 76330 + }, + { + "epoch": 1.63, + "learning_rate": 5.532051282051282e-05, + "loss": 0.0312, + "step": 76340 + }, + { + "epoch": 1.63, + "learning_rate": 5.528846153846154e-05, + "loss": 0.0263, + "step": 76350 + }, + { + "epoch": 1.63, + "learning_rate": 5.525641025641025e-05, + "loss": 0.0252, + "step": 76360 + }, + { + "epoch": 1.63, + "learning_rate": 5.522435897435897e-05, + "loss": 0.0293, + "step": 76370 + }, + { + "epoch": 1.63, + "learning_rate": 5.519230769230769e-05, + "loss": 0.0244, + "step": 76380 + }, + { + "epoch": 1.63, + "learning_rate": 5.516025641025641e-05, + "loss": 0.0296, + "step": 76390 + }, + { + "epoch": 1.63, + "learning_rate": 5.512820512820513e-05, + "loss": 0.032, + "step": 76400 + }, + { + "epoch": 1.63, + "learning_rate": 5.509615384615384e-05, + "loss": 0.0293, + "step": 76410 + }, + { + "epoch": 1.63, + "learning_rate": 5.5064102564102556e-05, + "loss": 0.0276, + "step": 76420 + }, + { + "epoch": 1.63, + "learning_rate": 5.5032051282051276e-05, + "loss": 0.0305, + "step": 76430 + }, + { + "epoch": 1.63, + "learning_rate": 5.499999999999999e-05, + "loss": 0.0255, + "step": 76440 + }, + { + "epoch": 1.63, + "learning_rate": 5.496794871794871e-05, + "loss": 0.0315, + "step": 76450 + }, + { + "epoch": 1.63, + "learning_rate": 5.493589743589743e-05, + "loss": 0.0271, + "step": 76460 + }, + { + "epoch": 1.63, + "learning_rate": 5.490384615384615e-05, + "loss": 0.0323, + "step": 76470 + }, + { + "epoch": 1.63, + "learning_rate": 5.487179487179487e-05, + "loss": 0.0313, + "step": 76480 + }, + { + "epoch": 1.63, + "learning_rate": 5.483974358974358e-05, + "loss": 0.0322, + "step": 76490 + }, + { + "epoch": 1.63, + "learning_rate": 5.48076923076923e-05, + "loss": 0.0337, + "step": 76500 + }, + { + "epoch": 1.63, + "learning_rate": 5.477564102564102e-05, + "loss": 0.0379, + "step": 76510 + }, + { + "epoch": 1.64, + "learning_rate": 5.474358974358974e-05, + "loss": 0.025, + "step": 76520 + }, + { + "epoch": 1.64, + "learning_rate": 5.471153846153845e-05, + "loss": 0.0285, + "step": 76530 + }, + { + "epoch": 1.64, + "learning_rate": 5.467948717948717e-05, + "loss": 0.0245, + "step": 76540 + }, + { + "epoch": 1.64, + "learning_rate": 5.464743589743589e-05, + "loss": 0.0218, + "step": 76550 + }, + { + "epoch": 1.64, + "learning_rate": 5.461538461538461e-05, + "loss": 0.0306, + "step": 76560 + }, + { + "epoch": 1.64, + "learning_rate": 5.458333333333332e-05, + "loss": 0.0324, + "step": 76570 + }, + { + "epoch": 1.64, + "learning_rate": 5.455128205128204e-05, + "loss": 0.023, + "step": 76580 + }, + { + "epoch": 1.64, + "learning_rate": 5.451923076923076e-05, + "loss": 0.0308, + "step": 76590 + }, + { + "epoch": 1.64, + "learning_rate": 5.448717948717948e-05, + "loss": 0.0276, + "step": 76600 + }, + { + "epoch": 1.64, + "learning_rate": 5.44551282051282e-05, + "loss": 0.0358, + "step": 76610 + }, + { + "epoch": 1.64, + "learning_rate": 5.4423076923076914e-05, + "loss": 0.0313, + "step": 76620 + }, + { + "epoch": 1.64, + "learning_rate": 5.4391025641025634e-05, + "loss": 0.0292, + "step": 76630 + }, + { + "epoch": 1.64, + "learning_rate": 5.435897435897435e-05, + "loss": 0.0297, + "step": 76640 + }, + { + "epoch": 1.64, + "learning_rate": 5.432692307692307e-05, + "loss": 0.025, + "step": 76650 + }, + { + "epoch": 1.64, + "learning_rate": 5.4294871794871786e-05, + "loss": 0.0273, + "step": 76660 + }, + { + "epoch": 1.64, + "learning_rate": 5.4262820512820505e-05, + "loss": 0.0347, + "step": 76670 + }, + { + "epoch": 1.64, + "learning_rate": 5.4230769230769225e-05, + "loss": 0.0336, + "step": 76680 + }, + { + "epoch": 1.64, + "learning_rate": 5.4198717948717945e-05, + "loss": 0.0314, + "step": 76690 + }, + { + "epoch": 1.64, + "learning_rate": 5.4166666666666664e-05, + "loss": 0.0309, + "step": 76700 + }, + { + "epoch": 1.64, + "learning_rate": 5.413461538461538e-05, + "loss": 0.0371, + "step": 76710 + }, + { + "epoch": 1.64, + "learning_rate": 5.4102564102564097e-05, + "loss": 0.029, + "step": 76720 + }, + { + "epoch": 1.64, + "learning_rate": 5.4070512820512816e-05, + "loss": 0.0314, + "step": 76730 + }, + { + "epoch": 1.64, + "learning_rate": 5.4038461538461536e-05, + "loss": 0.0285, + "step": 76740 + }, + { + "epoch": 1.64, + "learning_rate": 5.400641025641025e-05, + "loss": 0.026, + "step": 76750 + }, + { + "epoch": 1.64, + "learning_rate": 5.397435897435897e-05, + "loss": 0.0275, + "step": 76760 + }, + { + "epoch": 1.64, + "learning_rate": 5.394230769230769e-05, + "loss": 0.0287, + "step": 76770 + }, + { + "epoch": 1.64, + "learning_rate": 5.391025641025641e-05, + "loss": 0.0307, + "step": 76780 + }, + { + "epoch": 1.64, + "learning_rate": 5.387820512820512e-05, + "loss": 0.0286, + "step": 76790 + }, + { + "epoch": 1.64, + "learning_rate": 5.384615384615384e-05, + "loss": 0.0283, + "step": 76800 + }, + { + "epoch": 1.64, + "learning_rate": 5.381410256410256e-05, + "loss": 0.0282, + "step": 76810 + }, + { + "epoch": 1.64, + "learning_rate": 5.378205128205128e-05, + "loss": 0.028, + "step": 76820 + }, + { + "epoch": 1.64, + "learning_rate": 5.375e-05, + "loss": 0.0312, + "step": 76830 + }, + { + "epoch": 1.64, + "learning_rate": 5.371794871794871e-05, + "loss": 0.0264, + "step": 76840 + }, + { + "epoch": 1.64, + "learning_rate": 5.368589743589743e-05, + "loss": 0.0291, + "step": 76850 + }, + { + "epoch": 1.64, + "learning_rate": 5.365384615384615e-05, + "loss": 0.0315, + "step": 76860 + }, + { + "epoch": 1.64, + "learning_rate": 5.362179487179487e-05, + "loss": 0.0301, + "step": 76870 + }, + { + "epoch": 1.64, + "learning_rate": 5.358974358974358e-05, + "loss": 0.0394, + "step": 76880 + }, + { + "epoch": 1.64, + "learning_rate": 5.35576923076923e-05, + "loss": 0.0285, + "step": 76890 + }, + { + "epoch": 1.64, + "learning_rate": 5.352564102564102e-05, + "loss": 0.0332, + "step": 76900 + }, + { + "epoch": 1.64, + "learning_rate": 5.349358974358974e-05, + "loss": 0.0332, + "step": 76910 + }, + { + "epoch": 1.64, + "learning_rate": 5.346153846153846e-05, + "loss": 0.0243, + "step": 76920 + }, + { + "epoch": 1.64, + "learning_rate": 5.3429487179487174e-05, + "loss": 0.0238, + "step": 76930 + }, + { + "epoch": 1.64, + "learning_rate": 5.3397435897435894e-05, + "loss": 0.0297, + "step": 76940 + }, + { + "epoch": 1.64, + "learning_rate": 5.3365384615384614e-05, + "loss": 0.031, + "step": 76950 + }, + { + "epoch": 1.64, + "learning_rate": 5.333333333333333e-05, + "loss": 0.0285, + "step": 76960 + }, + { + "epoch": 1.64, + "learning_rate": 5.3301282051282046e-05, + "loss": 0.0296, + "step": 76970 + }, + { + "epoch": 1.64, + "learning_rate": 5.3269230769230766e-05, + "loss": 0.0302, + "step": 76980 + }, + { + "epoch": 1.65, + "learning_rate": 5.3237179487179485e-05, + "loss": 0.0336, + "step": 76990 + }, + { + "epoch": 1.65, + "learning_rate": 5.3205128205128205e-05, + "loss": 0.0347, + "step": 77000 + }, + { + "epoch": 1.65, + "eval_cer": 0.9236605046213217, + "eval_loss": 0.022434020414948463, + "eval_runtime": 120.5333, + "eval_samples_per_second": 16.593, + "eval_steps_per_second": 4.148, + "step": 77000 + }, + { + "epoch": 1.65, + "learning_rate": 5.317307692307692e-05, + "loss": 0.035, + "step": 77010 + }, + { + "epoch": 1.65, + "learning_rate": 5.314102564102564e-05, + "loss": 0.0285, + "step": 77020 + }, + { + "epoch": 1.65, + "learning_rate": 5.310897435897436e-05, + "loss": 0.0274, + "step": 77030 + }, + { + "epoch": 1.65, + "learning_rate": 5.3076923076923076e-05, + "loss": 0.0274, + "step": 77040 + }, + { + "epoch": 1.65, + "learning_rate": 5.3044871794871796e-05, + "loss": 0.0337, + "step": 77050 + }, + { + "epoch": 1.65, + "learning_rate": 5.301282051282051e-05, + "loss": 0.0271, + "step": 77060 + }, + { + "epoch": 1.65, + "learning_rate": 5.298076923076923e-05, + "loss": 0.0336, + "step": 77070 + }, + { + "epoch": 1.65, + "learning_rate": 5.294871794871795e-05, + "loss": 0.0286, + "step": 77080 + }, + { + "epoch": 1.65, + "learning_rate": 5.291666666666667e-05, + "loss": 0.0373, + "step": 77090 + }, + { + "epoch": 1.65, + "learning_rate": 5.288461538461538e-05, + "loss": 0.0282, + "step": 77100 + }, + { + "epoch": 1.65, + "learning_rate": 5.28525641025641e-05, + "loss": 0.025, + "step": 77110 + }, + { + "epoch": 1.65, + "learning_rate": 5.282051282051282e-05, + "loss": 0.028, + "step": 77120 + }, + { + "epoch": 1.65, + "learning_rate": 5.278846153846154e-05, + "loss": 0.025, + "step": 77130 + }, + { + "epoch": 1.65, + "learning_rate": 5.275641025641026e-05, + "loss": 0.0309, + "step": 77140 + }, + { + "epoch": 1.65, + "learning_rate": 5.2724358974358965e-05, + "loss": 0.0227, + "step": 77150 + }, + { + "epoch": 1.65, + "learning_rate": 5.2692307692307685e-05, + "loss": 0.0263, + "step": 77160 + }, + { + "epoch": 1.65, + "learning_rate": 5.2660256410256404e-05, + "loss": 0.0259, + "step": 77170 + }, + { + "epoch": 1.65, + "learning_rate": 5.262820512820512e-05, + "loss": 0.0311, + "step": 77180 + }, + { + "epoch": 1.65, + "learning_rate": 5.2596153846153837e-05, + "loss": 0.0308, + "step": 77190 + }, + { + "epoch": 1.65, + "learning_rate": 5.2564102564102556e-05, + "loss": 0.0276, + "step": 77200 + }, + { + "epoch": 1.65, + "learning_rate": 5.2532051282051276e-05, + "loss": 0.0286, + "step": 77210 + }, + { + "epoch": 1.65, + "learning_rate": 5.2499999999999995e-05, + "loss": 0.025, + "step": 77220 + }, + { + "epoch": 1.65, + "learning_rate": 5.246794871794871e-05, + "loss": 0.0282, + "step": 77230 + }, + { + "epoch": 1.65, + "learning_rate": 5.243589743589743e-05, + "loss": 0.0211, + "step": 77240 + }, + { + "epoch": 1.65, + "learning_rate": 5.240384615384615e-05, + "loss": 0.0274, + "step": 77250 + }, + { + "epoch": 1.65, + "learning_rate": 5.237179487179487e-05, + "loss": 0.0246, + "step": 77260 + }, + { + "epoch": 1.65, + "learning_rate": 5.233974358974358e-05, + "loss": 0.035, + "step": 77270 + }, + { + "epoch": 1.65, + "learning_rate": 5.23076923076923e-05, + "loss": 0.0311, + "step": 77280 + }, + { + "epoch": 1.65, + "learning_rate": 5.227564102564102e-05, + "loss": 0.0322, + "step": 77290 + }, + { + "epoch": 1.65, + "learning_rate": 5.224358974358974e-05, + "loss": 0.0256, + "step": 77300 + }, + { + "epoch": 1.65, + "learning_rate": 5.221153846153845e-05, + "loss": 0.0343, + "step": 77310 + }, + { + "epoch": 1.65, + "learning_rate": 5.217948717948717e-05, + "loss": 0.0285, + "step": 77320 + }, + { + "epoch": 1.65, + "learning_rate": 5.214743589743589e-05, + "loss": 0.0297, + "step": 77330 + }, + { + "epoch": 1.65, + "learning_rate": 5.211538461538461e-05, + "loss": 0.0284, + "step": 77340 + }, + { + "epoch": 1.65, + "learning_rate": 5.208333333333333e-05, + "loss": 0.0292, + "step": 77350 + }, + { + "epoch": 1.65, + "learning_rate": 5.205128205128204e-05, + "loss": 0.0325, + "step": 77360 + }, + { + "epoch": 1.65, + "learning_rate": 5.201923076923076e-05, + "loss": 0.0316, + "step": 77370 + }, + { + "epoch": 1.65, + "learning_rate": 5.198717948717948e-05, + "loss": 0.028, + "step": 77380 + }, + { + "epoch": 1.65, + "learning_rate": 5.19551282051282e-05, + "loss": 0.0336, + "step": 77390 + }, + { + "epoch": 1.65, + "learning_rate": 5.1923076923076914e-05, + "loss": 0.0319, + "step": 77400 + }, + { + "epoch": 1.65, + "learning_rate": 5.1891025641025634e-05, + "loss": 0.0296, + "step": 77410 + }, + { + "epoch": 1.65, + "learning_rate": 5.1858974358974354e-05, + "loss": 0.0323, + "step": 77420 + }, + { + "epoch": 1.65, + "learning_rate": 5.182692307692307e-05, + "loss": 0.0246, + "step": 77430 + }, + { + "epoch": 1.65, + "learning_rate": 5.179487179487179e-05, + "loss": 0.0311, + "step": 77440 + }, + { + "epoch": 1.65, + "learning_rate": 5.1762820512820506e-05, + "loss": 0.0333, + "step": 77450 + }, + { + "epoch": 1.66, + "learning_rate": 5.1730769230769225e-05, + "loss": 0.0297, + "step": 77460 + }, + { + "epoch": 1.66, + "learning_rate": 5.1698717948717945e-05, + "loss": 0.032, + "step": 77470 + }, + { + "epoch": 1.66, + "learning_rate": 5.1666666666666664e-05, + "loss": 0.031, + "step": 77480 + }, + { + "epoch": 1.66, + "learning_rate": 5.163461538461538e-05, + "loss": 0.0276, + "step": 77490 + }, + { + "epoch": 1.66, + "learning_rate": 5.16025641025641e-05, + "loss": 0.0323, + "step": 77500 + }, + { + "epoch": 1.66, + "learning_rate": 5.1570512820512816e-05, + "loss": 0.0314, + "step": 77510 + }, + { + "epoch": 1.66, + "learning_rate": 5.1538461538461536e-05, + "loss": 0.0345, + "step": 77520 + }, + { + "epoch": 1.66, + "learning_rate": 5.150641025641025e-05, + "loss": 0.0274, + "step": 77530 + }, + { + "epoch": 1.66, + "learning_rate": 5.147435897435897e-05, + "loss": 0.0303, + "step": 77540 + }, + { + "epoch": 1.66, + "learning_rate": 5.144230769230769e-05, + "loss": 0.0254, + "step": 77550 + }, + { + "epoch": 1.66, + "learning_rate": 5.141025641025641e-05, + "loss": 0.0317, + "step": 77560 + }, + { + "epoch": 1.66, + "learning_rate": 5.137820512820513e-05, + "loss": 0.025, + "step": 77570 + }, + { + "epoch": 1.66, + "learning_rate": 5.134615384615384e-05, + "loss": 0.0251, + "step": 77580 + }, + { + "epoch": 1.66, + "learning_rate": 5.131410256410256e-05, + "loss": 0.0338, + "step": 77590 + }, + { + "epoch": 1.66, + "learning_rate": 5.128205128205128e-05, + "loss": 0.0338, + "step": 77600 + }, + { + "epoch": 1.66, + "learning_rate": 5.125e-05, + "loss": 0.0278, + "step": 77610 + }, + { + "epoch": 1.66, + "learning_rate": 5.121794871794871e-05, + "loss": 0.0321, + "step": 77620 + }, + { + "epoch": 1.66, + "learning_rate": 5.118589743589743e-05, + "loss": 0.0314, + "step": 77630 + }, + { + "epoch": 1.66, + "learning_rate": 5.115384615384615e-05, + "loss": 0.0321, + "step": 77640 + }, + { + "epoch": 1.66, + "learning_rate": 5.112179487179487e-05, + "loss": 0.0268, + "step": 77650 + }, + { + "epoch": 1.66, + "learning_rate": 5.108974358974359e-05, + "loss": 0.0304, + "step": 77660 + }, + { + "epoch": 1.66, + "learning_rate": 5.10576923076923e-05, + "loss": 0.0341, + "step": 77670 + }, + { + "epoch": 1.66, + "learning_rate": 5.102564102564102e-05, + "loss": 0.0304, + "step": 77680 + }, + { + "epoch": 1.66, + "learning_rate": 5.099358974358974e-05, + "loss": 0.0307, + "step": 77690 + }, + { + "epoch": 1.66, + "learning_rate": 5.096153846153846e-05, + "loss": 0.0316, + "step": 77700 + }, + { + "epoch": 1.66, + "learning_rate": 5.0929487179487175e-05, + "loss": 0.0265, + "step": 77710 + }, + { + "epoch": 1.66, + "learning_rate": 5.0897435897435894e-05, + "loss": 0.0292, + "step": 77720 + }, + { + "epoch": 1.66, + "learning_rate": 5.0865384615384614e-05, + "loss": 0.0299, + "step": 77730 + }, + { + "epoch": 1.66, + "learning_rate": 5.0833333333333333e-05, + "loss": 0.0249, + "step": 77740 + }, + { + "epoch": 1.66, + "learning_rate": 5.0801282051282046e-05, + "loss": 0.0258, + "step": 77750 + }, + { + "epoch": 1.66, + "learning_rate": 5.0769230769230766e-05, + "loss": 0.0241, + "step": 77760 + }, + { + "epoch": 1.66, + "learning_rate": 5.0737179487179485e-05, + "loss": 0.0248, + "step": 77770 + }, + { + "epoch": 1.66, + "learning_rate": 5.0705128205128205e-05, + "loss": 0.0324, + "step": 77780 + }, + { + "epoch": 1.66, + "learning_rate": 5.0673076923076925e-05, + "loss": 0.0294, + "step": 77790 + }, + { + "epoch": 1.66, + "learning_rate": 5.064102564102564e-05, + "loss": 0.0274, + "step": 77800 + }, + { + "epoch": 1.66, + "learning_rate": 5.060897435897436e-05, + "loss": 0.0289, + "step": 77810 + }, + { + "epoch": 1.66, + "learning_rate": 5.057692307692308e-05, + "loss": 0.026, + "step": 77820 + }, + { + "epoch": 1.66, + "learning_rate": 5.0544871794871796e-05, + "loss": 0.0259, + "step": 77830 + }, + { + "epoch": 1.66, + "learning_rate": 5.051282051282051e-05, + "loss": 0.029, + "step": 77840 + }, + { + "epoch": 1.66, + "learning_rate": 5.048076923076923e-05, + "loss": 0.0341, + "step": 77850 + }, + { + "epoch": 1.66, + "learning_rate": 5.044871794871795e-05, + "loss": 0.0305, + "step": 77860 + }, + { + "epoch": 1.66, + "learning_rate": 5.041666666666667e-05, + "loss": 0.0332, + "step": 77870 + }, + { + "epoch": 1.66, + "learning_rate": 5.0384615384615374e-05, + "loss": 0.0283, + "step": 77880 + }, + { + "epoch": 1.66, + "learning_rate": 5.0352564102564094e-05, + "loss": 0.0347, + "step": 77890 + }, + { + "epoch": 1.66, + "learning_rate": 5.032051282051281e-05, + "loss": 0.0277, + "step": 77900 + }, + { + "epoch": 1.66, + "learning_rate": 5.028846153846153e-05, + "loss": 0.0328, + "step": 77910 + }, + { + "epoch": 1.66, + "learning_rate": 5.0256410256410246e-05, + "loss": 0.0278, + "step": 77920 + }, + { + "epoch": 1.67, + "learning_rate": 5.0224358974358965e-05, + "loss": 0.0289, + "step": 77930 + }, + { + "epoch": 1.67, + "learning_rate": 5.0192307692307685e-05, + "loss": 0.0284, + "step": 77940 + }, + { + "epoch": 1.67, + "learning_rate": 5.0160256410256404e-05, + "loss": 0.0329, + "step": 77950 + }, + { + "epoch": 1.67, + "learning_rate": 5.012820512820512e-05, + "loss": 0.0294, + "step": 77960 + }, + { + "epoch": 1.67, + "learning_rate": 5.009615384615384e-05, + "loss": 0.0316, + "step": 77970 + }, + { + "epoch": 1.67, + "learning_rate": 5.0064102564102556e-05, + "loss": 0.0277, + "step": 77980 + }, + { + "epoch": 1.67, + "learning_rate": 5.0032051282051276e-05, + "loss": 0.0301, + "step": 77990 + }, + { + "epoch": 1.67, + "learning_rate": 4.9999999999999996e-05, + "loss": 0.0268, + "step": 78000 + }, + { + "epoch": 1.67, + "eval_cer": 0.9236571075273657, + "eval_loss": 0.022638017311692238, + "eval_runtime": 120.643, + "eval_samples_per_second": 16.578, + "eval_steps_per_second": 4.144, + "step": 78000 + }, + { + "epoch": 1.67, + "learning_rate": 4.996794871794871e-05, + "loss": 0.0283, + "step": 78010 + }, + { + "epoch": 1.67, + "learning_rate": 4.993589743589743e-05, + "loss": 0.0276, + "step": 78020 + }, + { + "epoch": 1.67, + "learning_rate": 4.990384615384615e-05, + "loss": 0.0266, + "step": 78030 + }, + { + "epoch": 1.67, + "learning_rate": 4.987179487179487e-05, + "loss": 0.0249, + "step": 78040 + }, + { + "epoch": 1.67, + "learning_rate": 4.983974358974358e-05, + "loss": 0.029, + "step": 78050 + }, + { + "epoch": 1.67, + "learning_rate": 4.98076923076923e-05, + "loss": 0.0259, + "step": 78060 + }, + { + "epoch": 1.67, + "learning_rate": 4.977564102564102e-05, + "loss": 0.033, + "step": 78070 + }, + { + "epoch": 1.67, + "learning_rate": 4.974358974358974e-05, + "loss": 0.0306, + "step": 78080 + }, + { + "epoch": 1.67, + "learning_rate": 4.971153846153846e-05, + "loss": 0.028, + "step": 78090 + }, + { + "epoch": 1.67, + "learning_rate": 4.967948717948717e-05, + "loss": 0.0293, + "step": 78100 + }, + { + "epoch": 1.67, + "learning_rate": 4.964743589743589e-05, + "loss": 0.0284, + "step": 78110 + }, + { + "epoch": 1.67, + "learning_rate": 4.961538461538461e-05, + "loss": 0.0235, + "step": 78120 + }, + { + "epoch": 1.67, + "learning_rate": 4.958333333333333e-05, + "loss": 0.0247, + "step": 78130 + }, + { + "epoch": 1.67, + "learning_rate": 4.955128205128204e-05, + "loss": 0.0282, + "step": 78140 + }, + { + "epoch": 1.67, + "learning_rate": 4.951923076923076e-05, + "loss": 0.0329, + "step": 78150 + }, + { + "epoch": 1.67, + "learning_rate": 4.948717948717948e-05, + "loss": 0.0285, + "step": 78160 + }, + { + "epoch": 1.67, + "learning_rate": 4.94551282051282e-05, + "loss": 0.0277, + "step": 78170 + }, + { + "epoch": 1.67, + "learning_rate": 4.942307692307692e-05, + "loss": 0.0317, + "step": 78180 + }, + { + "epoch": 1.67, + "learning_rate": 4.9391025641025634e-05, + "loss": 0.0349, + "step": 78190 + }, + { + "epoch": 1.67, + "learning_rate": 4.9358974358974354e-05, + "loss": 0.024, + "step": 78200 + }, + { + "epoch": 1.67, + "learning_rate": 4.932692307692307e-05, + "loss": 0.0345, + "step": 78210 + }, + { + "epoch": 1.67, + "learning_rate": 4.929487179487179e-05, + "loss": 0.033, + "step": 78220 + }, + { + "epoch": 1.67, + "learning_rate": 4.9262820512820506e-05, + "loss": 0.0272, + "step": 78230 + }, + { + "epoch": 1.67, + "learning_rate": 4.9230769230769225e-05, + "loss": 0.0324, + "step": 78240 + }, + { + "epoch": 1.67, + "learning_rate": 4.9198717948717945e-05, + "loss": 0.0258, + "step": 78250 + }, + { + "epoch": 1.67, + "learning_rate": 4.9166666666666665e-05, + "loss": 0.0249, + "step": 78260 + }, + { + "epoch": 1.67, + "learning_rate": 4.913461538461538e-05, + "loss": 0.0266, + "step": 78270 + }, + { + "epoch": 1.67, + "learning_rate": 4.91025641025641e-05, + "loss": 0.0242, + "step": 78280 + }, + { + "epoch": 1.67, + "learning_rate": 4.907051282051282e-05, + "loss": 0.0355, + "step": 78290 + }, + { + "epoch": 1.67, + "learning_rate": 4.9038461538461536e-05, + "loss": 0.0349, + "step": 78300 + }, + { + "epoch": 1.67, + "learning_rate": 4.9006410256410256e-05, + "loss": 0.03, + "step": 78310 + }, + { + "epoch": 1.67, + "learning_rate": 4.897435897435897e-05, + "loss": 0.0253, + "step": 78320 + }, + { + "epoch": 1.67, + "learning_rate": 4.894230769230769e-05, + "loss": 0.0354, + "step": 78330 + }, + { + "epoch": 1.67, + "learning_rate": 4.891025641025641e-05, + "loss": 0.0281, + "step": 78340 + }, + { + "epoch": 1.67, + "learning_rate": 4.887820512820513e-05, + "loss": 0.0277, + "step": 78350 + }, + { + "epoch": 1.67, + "learning_rate": 4.884615384615384e-05, + "loss": 0.0262, + "step": 78360 + }, + { + "epoch": 1.67, + "learning_rate": 4.881410256410256e-05, + "loss": 0.0297, + "step": 78370 + }, + { + "epoch": 1.67, + "learning_rate": 4.878205128205128e-05, + "loss": 0.0296, + "step": 78380 + }, + { + "epoch": 1.67, + "learning_rate": 4.875e-05, + "loss": 0.0275, + "step": 78390 + }, + { + "epoch": 1.68, + "learning_rate": 4.871794871794872e-05, + "loss": 0.0322, + "step": 78400 + }, + { + "epoch": 1.68, + "learning_rate": 4.868589743589743e-05, + "loss": 0.0236, + "step": 78410 + }, + { + "epoch": 1.68, + "learning_rate": 4.865384615384615e-05, + "loss": 0.0292, + "step": 78420 + }, + { + "epoch": 1.68, + "learning_rate": 4.862179487179487e-05, + "loss": 0.0333, + "step": 78430 + }, + { + "epoch": 1.68, + "learning_rate": 4.858974358974359e-05, + "loss": 0.0297, + "step": 78440 + }, + { + "epoch": 1.68, + "learning_rate": 4.85576923076923e-05, + "loss": 0.028, + "step": 78450 + }, + { + "epoch": 1.68, + "learning_rate": 4.852564102564102e-05, + "loss": 0.0262, + "step": 78460 + }, + { + "epoch": 1.68, + "learning_rate": 4.849358974358974e-05, + "loss": 0.0257, + "step": 78470 + }, + { + "epoch": 1.68, + "learning_rate": 4.846153846153846e-05, + "loss": 0.0305, + "step": 78480 + }, + { + "epoch": 1.68, + "learning_rate": 4.8429487179487175e-05, + "loss": 0.033, + "step": 78490 + }, + { + "epoch": 1.68, + "learning_rate": 4.8397435897435894e-05, + "loss": 0.0322, + "step": 78500 + }, + { + "epoch": 1.68, + "learning_rate": 4.8365384615384614e-05, + "loss": 0.0319, + "step": 78510 + }, + { + "epoch": 1.68, + "learning_rate": 4.8333333333333334e-05, + "loss": 0.0287, + "step": 78520 + }, + { + "epoch": 1.68, + "learning_rate": 4.830128205128205e-05, + "loss": 0.0286, + "step": 78530 + }, + { + "epoch": 1.68, + "learning_rate": 4.8269230769230766e-05, + "loss": 0.0384, + "step": 78540 + }, + { + "epoch": 1.68, + "learning_rate": 4.8237179487179486e-05, + "loss": 0.0311, + "step": 78550 + }, + { + "epoch": 1.68, + "learning_rate": 4.8205128205128205e-05, + "loss": 0.0327, + "step": 78560 + }, + { + "epoch": 1.68, + "learning_rate": 4.8173076923076925e-05, + "loss": 0.0281, + "step": 78570 + }, + { + "epoch": 1.68, + "learning_rate": 4.814102564102564e-05, + "loss": 0.0294, + "step": 78580 + }, + { + "epoch": 1.68, + "learning_rate": 4.810897435897436e-05, + "loss": 0.0249, + "step": 78590 + }, + { + "epoch": 1.68, + "learning_rate": 4.807692307692308e-05, + "loss": 0.0359, + "step": 78600 + }, + { + "epoch": 1.68, + "learning_rate": 4.804487179487179e-05, + "loss": 0.0282, + "step": 78610 + }, + { + "epoch": 1.68, + "learning_rate": 4.80128205128205e-05, + "loss": 0.0254, + "step": 78620 + }, + { + "epoch": 1.68, + "learning_rate": 4.798076923076922e-05, + "loss": 0.0245, + "step": 78630 + }, + { + "epoch": 1.68, + "learning_rate": 4.794871794871794e-05, + "loss": 0.0248, + "step": 78640 + }, + { + "epoch": 1.68, + "learning_rate": 4.791666666666666e-05, + "loss": 0.0278, + "step": 78650 + }, + { + "epoch": 1.68, + "learning_rate": 4.7884615384615374e-05, + "loss": 0.0291, + "step": 78660 + }, + { + "epoch": 1.68, + "learning_rate": 4.7852564102564094e-05, + "loss": 0.0269, + "step": 78670 + }, + { + "epoch": 1.68, + "learning_rate": 4.782051282051281e-05, + "loss": 0.0343, + "step": 78680 + }, + { + "epoch": 1.68, + "learning_rate": 4.778846153846153e-05, + "loss": 0.0328, + "step": 78690 + }, + { + "epoch": 1.68, + "learning_rate": 4.7756410256410246e-05, + "loss": 0.027, + "step": 78700 + }, + { + "epoch": 1.68, + "learning_rate": 4.7724358974358965e-05, + "loss": 0.036, + "step": 78710 + }, + { + "epoch": 1.68, + "learning_rate": 4.7692307692307685e-05, + "loss": 0.0232, + "step": 78720 + }, + { + "epoch": 1.68, + "learning_rate": 4.7660256410256405e-05, + "loss": 0.0279, + "step": 78730 + }, + { + "epoch": 1.68, + "learning_rate": 4.7628205128205124e-05, + "loss": 0.0294, + "step": 78740 + }, + { + "epoch": 1.68, + "learning_rate": 4.759615384615384e-05, + "loss": 0.0286, + "step": 78750 + }, + { + "epoch": 1.68, + "learning_rate": 4.756410256410256e-05, + "loss": 0.0288, + "step": 78760 + }, + { + "epoch": 1.68, + "learning_rate": 4.7532051282051276e-05, + "loss": 0.0287, + "step": 78770 + }, + { + "epoch": 1.68, + "learning_rate": 4.7499999999999996e-05, + "loss": 0.0248, + "step": 78780 + }, + { + "epoch": 1.68, + "learning_rate": 4.746794871794871e-05, + "loss": 0.0314, + "step": 78790 + }, + { + "epoch": 1.68, + "learning_rate": 4.743589743589743e-05, + "loss": 0.0288, + "step": 78800 + }, + { + "epoch": 1.68, + "learning_rate": 4.740384615384615e-05, + "loss": 0.0316, + "step": 78810 + }, + { + "epoch": 1.68, + "learning_rate": 4.737179487179487e-05, + "loss": 0.0326, + "step": 78820 + }, + { + "epoch": 1.68, + "learning_rate": 4.733974358974359e-05, + "loss": 0.0277, + "step": 78830 + }, + { + "epoch": 1.68, + "learning_rate": 4.73076923076923e-05, + "loss": 0.031, + "step": 78840 + }, + { + "epoch": 1.68, + "learning_rate": 4.727564102564102e-05, + "loss": 0.0273, + "step": 78850 + }, + { + "epoch": 1.69, + "learning_rate": 4.724358974358974e-05, + "loss": 0.029, + "step": 78860 + }, + { + "epoch": 1.69, + "learning_rate": 4.721153846153846e-05, + "loss": 0.0293, + "step": 78870 + }, + { + "epoch": 1.69, + "learning_rate": 4.717948717948717e-05, + "loss": 0.0294, + "step": 78880 + }, + { + "epoch": 1.69, + "learning_rate": 4.714743589743589e-05, + "loss": 0.0282, + "step": 78890 + }, + { + "epoch": 1.69, + "learning_rate": 4.711538461538461e-05, + "loss": 0.0309, + "step": 78900 + }, + { + "epoch": 1.69, + "learning_rate": 4.708333333333333e-05, + "loss": 0.0304, + "step": 78910 + }, + { + "epoch": 1.69, + "learning_rate": 4.705128205128204e-05, + "loss": 0.0353, + "step": 78920 + }, + { + "epoch": 1.69, + "learning_rate": 4.701923076923076e-05, + "loss": 0.0269, + "step": 78930 + }, + { + "epoch": 1.69, + "learning_rate": 4.698717948717948e-05, + "loss": 0.0295, + "step": 78940 + }, + { + "epoch": 1.69, + "learning_rate": 4.69551282051282e-05, + "loss": 0.0256, + "step": 78950 + }, + { + "epoch": 1.69, + "learning_rate": 4.692307692307692e-05, + "loss": 0.0329, + "step": 78960 + }, + { + "epoch": 1.69, + "learning_rate": 4.6891025641025634e-05, + "loss": 0.0263, + "step": 78970 + }, + { + "epoch": 1.69, + "learning_rate": 4.6858974358974354e-05, + "loss": 0.0284, + "step": 78980 + }, + { + "epoch": 1.69, + "learning_rate": 4.6826923076923074e-05, + "loss": 0.0277, + "step": 78990 + }, + { + "epoch": 1.69, + "learning_rate": 4.679487179487179e-05, + "loss": 0.0279, + "step": 79000 + }, + { + "epoch": 1.69, + "eval_cer": 0.9236477655189868, + "eval_loss": 0.021934477612376213, + "eval_runtime": 120.4975, + "eval_samples_per_second": 16.598, + "eval_steps_per_second": 4.149, + "step": 79000 + }, + { + "epoch": 1.69, + "learning_rate": 4.6762820512820506e-05, + "loss": 0.0259, + "step": 79010 + }, + { + "epoch": 1.69, + "learning_rate": 4.6730769230769226e-05, + "loss": 0.026, + "step": 79020 + }, + { + "epoch": 1.69, + "learning_rate": 4.6698717948717945e-05, + "loss": 0.0335, + "step": 79030 + }, + { + "epoch": 1.69, + "learning_rate": 4.6666666666666665e-05, + "loss": 0.0332, + "step": 79040 + }, + { + "epoch": 1.69, + "learning_rate": 4.6634615384615384e-05, + "loss": 0.0294, + "step": 79050 + }, + { + "epoch": 1.69, + "learning_rate": 4.66025641025641e-05, + "loss": 0.0281, + "step": 79060 + }, + { + "epoch": 1.69, + "learning_rate": 4.657051282051282e-05, + "loss": 0.0311, + "step": 79070 + }, + { + "epoch": 1.69, + "learning_rate": 4.6538461538461536e-05, + "loss": 0.0273, + "step": 79080 + }, + { + "epoch": 1.69, + "learning_rate": 4.6506410256410256e-05, + "loss": 0.0322, + "step": 79090 + }, + { + "epoch": 1.69, + "learning_rate": 4.647435897435897e-05, + "loss": 0.0305, + "step": 79100 + }, + { + "epoch": 1.69, + "learning_rate": 4.644230769230769e-05, + "loss": 0.0262, + "step": 79110 + }, + { + "epoch": 1.69, + "learning_rate": 4.641025641025641e-05, + "loss": 0.0287, + "step": 79120 + }, + { + "epoch": 1.69, + "learning_rate": 4.637820512820513e-05, + "loss": 0.0248, + "step": 79130 + }, + { + "epoch": 1.69, + "learning_rate": 4.634615384615385e-05, + "loss": 0.0325, + "step": 79140 + }, + { + "epoch": 1.69, + "learning_rate": 4.631410256410256e-05, + "loss": 0.0254, + "step": 79150 + }, + { + "epoch": 1.69, + "learning_rate": 4.628205128205128e-05, + "loss": 0.0271, + "step": 79160 + }, + { + "epoch": 1.69, + "learning_rate": 4.625e-05, + "loss": 0.031, + "step": 79170 + }, + { + "epoch": 1.69, + "learning_rate": 4.621794871794872e-05, + "loss": 0.0327, + "step": 79180 + }, + { + "epoch": 1.69, + "learning_rate": 4.618589743589743e-05, + "loss": 0.0266, + "step": 79190 + }, + { + "epoch": 1.69, + "learning_rate": 4.615384615384615e-05, + "loss": 0.0294, + "step": 79200 + }, + { + "epoch": 1.69, + "learning_rate": 4.612179487179487e-05, + "loss": 0.0296, + "step": 79210 + }, + { + "epoch": 1.69, + "learning_rate": 4.608974358974359e-05, + "loss": 0.0277, + "step": 79220 + }, + { + "epoch": 1.69, + "learning_rate": 4.6057692307692303e-05, + "loss": 0.0278, + "step": 79230 + }, + { + "epoch": 1.69, + "learning_rate": 4.602564102564102e-05, + "loss": 0.029, + "step": 79240 + }, + { + "epoch": 1.69, + "learning_rate": 4.599358974358974e-05, + "loss": 0.0284, + "step": 79250 + }, + { + "epoch": 1.69, + "learning_rate": 4.596153846153846e-05, + "loss": 0.0272, + "step": 79260 + }, + { + "epoch": 1.69, + "learning_rate": 4.592948717948718e-05, + "loss": 0.0296, + "step": 79270 + }, + { + "epoch": 1.69, + "learning_rate": 4.5897435897435895e-05, + "loss": 0.0263, + "step": 79280 + }, + { + "epoch": 1.69, + "learning_rate": 4.5865384615384614e-05, + "loss": 0.0259, + "step": 79290 + }, + { + "epoch": 1.69, + "learning_rate": 4.5833333333333334e-05, + "loss": 0.0292, + "step": 79300 + }, + { + "epoch": 1.69, + "learning_rate": 4.5801282051282053e-05, + "loss": 0.032, + "step": 79310 + }, + { + "epoch": 1.69, + "learning_rate": 4.5769230769230766e-05, + "loss": 0.0328, + "step": 79320 + }, + { + "epoch": 1.7, + "learning_rate": 4.5737179487179486e-05, + "loss": 0.0248, + "step": 79330 + }, + { + "epoch": 1.7, + "learning_rate": 4.5705128205128205e-05, + "loss": 0.0276, + "step": 79340 + }, + { + "epoch": 1.7, + "learning_rate": 4.567307692307692e-05, + "loss": 0.0266, + "step": 79350 + }, + { + "epoch": 1.7, + "learning_rate": 4.564102564102563e-05, + "loss": 0.0269, + "step": 79360 + }, + { + "epoch": 1.7, + "learning_rate": 4.560897435897435e-05, + "loss": 0.0288, + "step": 79370 + }, + { + "epoch": 1.7, + "learning_rate": 4.557692307692307e-05, + "loss": 0.0238, + "step": 79380 + }, + { + "epoch": 1.7, + "learning_rate": 4.554487179487179e-05, + "loss": 0.0255, + "step": 79390 + }, + { + "epoch": 1.7, + "learning_rate": 4.55128205128205e-05, + "loss": 0.0369, + "step": 79400 + }, + { + "epoch": 1.7, + "learning_rate": 4.548076923076922e-05, + "loss": 0.0236, + "step": 79410 + }, + { + "epoch": 1.7, + "learning_rate": 4.544871794871794e-05, + "loss": 0.0263, + "step": 79420 + }, + { + "epoch": 1.7, + "learning_rate": 4.541666666666666e-05, + "loss": 0.0281, + "step": 79430 + }, + { + "epoch": 1.7, + "learning_rate": 4.5384615384615374e-05, + "loss": 0.0279, + "step": 79440 + }, + { + "epoch": 1.7, + "learning_rate": 4.5352564102564094e-05, + "loss": 0.0271, + "step": 79450 + }, + { + "epoch": 1.7, + "learning_rate": 4.5320512820512814e-05, + "loss": 0.0325, + "step": 79460 + }, + { + "epoch": 1.7, + "learning_rate": 4.528846153846153e-05, + "loss": 0.0231, + "step": 79470 + }, + { + "epoch": 1.7, + "learning_rate": 4.525641025641025e-05, + "loss": 0.0254, + "step": 79480 + }, + { + "epoch": 1.7, + "learning_rate": 4.5224358974358966e-05, + "loss": 0.0252, + "step": 79490 + }, + { + "epoch": 1.7, + "learning_rate": 4.5192307692307685e-05, + "loss": 0.0379, + "step": 79500 + }, + { + "epoch": 1.7, + "learning_rate": 4.5160256410256405e-05, + "loss": 0.032, + "step": 79510 + }, + { + "epoch": 1.7, + "learning_rate": 4.5128205128205124e-05, + "loss": 0.0332, + "step": 79520 + }, + { + "epoch": 1.7, + "learning_rate": 4.509615384615384e-05, + "loss": 0.0299, + "step": 79530 + }, + { + "epoch": 1.7, + "learning_rate": 4.506410256410256e-05, + "loss": 0.0262, + "step": 79540 + }, + { + "epoch": 1.7, + "learning_rate": 4.5032051282051276e-05, + "loss": 0.031, + "step": 79550 + }, + { + "epoch": 1.7, + "learning_rate": 4.4999999999999996e-05, + "loss": 0.0315, + "step": 79560 + }, + { + "epoch": 1.7, + "learning_rate": 4.4967948717948716e-05, + "loss": 0.0251, + "step": 79570 + }, + { + "epoch": 1.7, + "learning_rate": 4.493589743589743e-05, + "loss": 0.0329, + "step": 79580 + }, + { + "epoch": 1.7, + "learning_rate": 4.490384615384615e-05, + "loss": 0.0271, + "step": 79590 + }, + { + "epoch": 1.7, + "learning_rate": 4.487179487179487e-05, + "loss": 0.0309, + "step": 79600 + }, + { + "epoch": 1.7, + "learning_rate": 4.483974358974359e-05, + "loss": 0.0332, + "step": 79610 + }, + { + "epoch": 1.7, + "learning_rate": 4.48076923076923e-05, + "loss": 0.0328, + "step": 79620 + }, + { + "epoch": 1.7, + "learning_rate": 4.477564102564102e-05, + "loss": 0.0263, + "step": 79630 + }, + { + "epoch": 1.7, + "learning_rate": 4.474358974358974e-05, + "loss": 0.024, + "step": 79640 + }, + { + "epoch": 1.7, + "learning_rate": 4.471153846153846e-05, + "loss": 0.0287, + "step": 79650 + }, + { + "epoch": 1.7, + "learning_rate": 4.467948717948717e-05, + "loss": 0.026, + "step": 79660 + }, + { + "epoch": 1.7, + "learning_rate": 4.464743589743589e-05, + "loss": 0.0307, + "step": 79670 + }, + { + "epoch": 1.7, + "learning_rate": 4.461538461538461e-05, + "loss": 0.0273, + "step": 79680 + }, + { + "epoch": 1.7, + "learning_rate": 4.458333333333333e-05, + "loss": 0.0288, + "step": 79690 + }, + { + "epoch": 1.7, + "learning_rate": 4.455128205128205e-05, + "loss": 0.0274, + "step": 79700 + }, + { + "epoch": 1.7, + "learning_rate": 4.451923076923076e-05, + "loss": 0.0339, + "step": 79710 + }, + { + "epoch": 1.7, + "learning_rate": 4.448717948717948e-05, + "loss": 0.0381, + "step": 79720 + }, + { + "epoch": 1.7, + "learning_rate": 4.44551282051282e-05, + "loss": 0.0292, + "step": 79730 + }, + { + "epoch": 1.7, + "learning_rate": 4.442307692307692e-05, + "loss": 0.0281, + "step": 79740 + }, + { + "epoch": 1.7, + "learning_rate": 4.4391025641025635e-05, + "loss": 0.0314, + "step": 79750 + }, + { + "epoch": 1.7, + "learning_rate": 4.4358974358974354e-05, + "loss": 0.0287, + "step": 79760 + }, + { + "epoch": 1.7, + "learning_rate": 4.4326923076923074e-05, + "loss": 0.033, + "step": 79770 + }, + { + "epoch": 1.7, + "learning_rate": 4.4294871794871793e-05, + "loss": 0.026, + "step": 79780 + }, + { + "epoch": 1.7, + "learning_rate": 4.426282051282051e-05, + "loss": 0.031, + "step": 79790 + }, + { + "epoch": 1.71, + "learning_rate": 4.4230769230769226e-05, + "loss": 0.0325, + "step": 79800 + }, + { + "epoch": 1.71, + "learning_rate": 4.4198717948717945e-05, + "loss": 0.0337, + "step": 79810 + }, + { + "epoch": 1.71, + "learning_rate": 4.4166666666666665e-05, + "loss": 0.0307, + "step": 79820 + }, + { + "epoch": 1.71, + "learning_rate": 4.4134615384615385e-05, + "loss": 0.0335, + "step": 79830 + }, + { + "epoch": 1.71, + "learning_rate": 4.41025641025641e-05, + "loss": 0.0248, + "step": 79840 + }, + { + "epoch": 1.71, + "learning_rate": 4.407051282051282e-05, + "loss": 0.0277, + "step": 79850 + }, + { + "epoch": 1.71, + "learning_rate": 4.403846153846154e-05, + "loss": 0.0241, + "step": 79860 + }, + { + "epoch": 1.71, + "learning_rate": 4.4006410256410256e-05, + "loss": 0.0294, + "step": 79870 + }, + { + "epoch": 1.71, + "learning_rate": 4.397435897435897e-05, + "loss": 0.0333, + "step": 79880 + }, + { + "epoch": 1.71, + "learning_rate": 4.394230769230769e-05, + "loss": 0.0289, + "step": 79890 + }, + { + "epoch": 1.71, + "learning_rate": 4.391025641025641e-05, + "loss": 0.0273, + "step": 79900 + }, + { + "epoch": 1.71, + "learning_rate": 4.387820512820513e-05, + "loss": 0.0326, + "step": 79910 + }, + { + "epoch": 1.71, + "learning_rate": 4.384615384615385e-05, + "loss": 0.0282, + "step": 79920 + }, + { + "epoch": 1.71, + "learning_rate": 4.381410256410256e-05, + "loss": 0.0326, + "step": 79930 + }, + { + "epoch": 1.71, + "learning_rate": 4.378205128205128e-05, + "loss": 0.0325, + "step": 79940 + }, + { + "epoch": 1.71, + "learning_rate": 4.375e-05, + "loss": 0.0243, + "step": 79950 + }, + { + "epoch": 1.71, + "learning_rate": 4.371794871794872e-05, + "loss": 0.0315, + "step": 79960 + }, + { + "epoch": 1.71, + "learning_rate": 4.368589743589743e-05, + "loss": 0.0251, + "step": 79970 + }, + { + "epoch": 1.71, + "learning_rate": 4.365384615384615e-05, + "loss": 0.0281, + "step": 79980 + }, + { + "epoch": 1.71, + "learning_rate": 4.362179487179487e-05, + "loss": 0.029, + "step": 79990 + }, + { + "epoch": 1.71, + "learning_rate": 4.358974358974359e-05, + "loss": 0.0247, + "step": 80000 + }, + { + "epoch": 1.71, + "eval_cer": 0.9235118817607477, + "eval_loss": 0.021996650844812393, + "eval_runtime": 120.6532, + "eval_samples_per_second": 16.576, + "eval_steps_per_second": 4.144, + "step": 80000 + }, + { + "epoch": 1.71, + "learning_rate": 4.355769230769231e-05, + "loss": 0.0273, + "step": 80010 + }, + { + "epoch": 1.71, + "learning_rate": 4.352564102564102e-05, + "loss": 0.0256, + "step": 80020 + }, + { + "epoch": 1.71, + "learning_rate": 4.349358974358974e-05, + "loss": 0.0257, + "step": 80030 + }, + { + "epoch": 1.71, + "learning_rate": 4.346153846153846e-05, + "loss": 0.034, + "step": 80040 + }, + { + "epoch": 1.71, + "learning_rate": 4.342948717948718e-05, + "loss": 0.0275, + "step": 80050 + }, + { + "epoch": 1.71, + "learning_rate": 4.3397435897435895e-05, + "loss": 0.0268, + "step": 80060 + }, + { + "epoch": 1.71, + "learning_rate": 4.3365384615384614e-05, + "loss": 0.027, + "step": 80070 + }, + { + "epoch": 1.71, + "learning_rate": 4.333333333333333e-05, + "loss": 0.027, + "step": 80080 + }, + { + "epoch": 1.71, + "learning_rate": 4.330128205128205e-05, + "loss": 0.0289, + "step": 80090 + }, + { + "epoch": 1.71, + "learning_rate": 4.326923076923076e-05, + "loss": 0.0254, + "step": 80100 + }, + { + "epoch": 1.71, + "learning_rate": 4.323717948717948e-05, + "loss": 0.0305, + "step": 80110 + }, + { + "epoch": 1.71, + "learning_rate": 4.32051282051282e-05, + "loss": 0.0274, + "step": 80120 + }, + { + "epoch": 1.71, + "learning_rate": 4.317307692307692e-05, + "loss": 0.0263, + "step": 80130 + }, + { + "epoch": 1.71, + "learning_rate": 4.314102564102563e-05, + "loss": 0.0261, + "step": 80140 + }, + { + "epoch": 1.71, + "learning_rate": 4.310897435897435e-05, + "loss": 0.0304, + "step": 80150 + }, + { + "epoch": 1.71, + "learning_rate": 4.307692307692307e-05, + "loss": 0.026, + "step": 80160 + }, + { + "epoch": 1.71, + "learning_rate": 4.304487179487179e-05, + "loss": 0.0291, + "step": 80170 + }, + { + "epoch": 1.71, + "learning_rate": 4.30128205128205e-05, + "loss": 0.0275, + "step": 80180 + }, + { + "epoch": 1.71, + "learning_rate": 4.298076923076922e-05, + "loss": 0.0331, + "step": 80190 + }, + { + "epoch": 1.71, + "learning_rate": 4.294871794871794e-05, + "loss": 0.0305, + "step": 80200 + }, + { + "epoch": 1.71, + "learning_rate": 4.291666666666666e-05, + "loss": 0.0278, + "step": 80210 + }, + { + "epoch": 1.71, + "learning_rate": 4.288461538461538e-05, + "loss": 0.0261, + "step": 80220 + }, + { + "epoch": 1.71, + "learning_rate": 4.2852564102564094e-05, + "loss": 0.0246, + "step": 80230 + }, + { + "epoch": 1.71, + "learning_rate": 4.2820512820512814e-05, + "loss": 0.0286, + "step": 80240 + }, + { + "epoch": 1.71, + "learning_rate": 4.2788461538461533e-05, + "loss": 0.0304, + "step": 80250 + }, + { + "epoch": 1.71, + "learning_rate": 4.275641025641025e-05, + "loss": 0.0252, + "step": 80260 + }, + { + "epoch": 1.72, + "learning_rate": 4.2724358974358966e-05, + "loss": 0.0305, + "step": 80270 + }, + { + "epoch": 1.72, + "learning_rate": 4.2692307692307685e-05, + "loss": 0.0263, + "step": 80280 + }, + { + "epoch": 1.72, + "learning_rate": 4.2660256410256405e-05, + "loss": 0.0269, + "step": 80290 + }, + { + "epoch": 1.72, + "learning_rate": 4.2628205128205125e-05, + "loss": 0.0311, + "step": 80300 + }, + { + "epoch": 1.72, + "learning_rate": 4.2596153846153844e-05, + "loss": 0.0324, + "step": 80310 + }, + { + "epoch": 1.72, + "learning_rate": 4.256410256410256e-05, + "loss": 0.0307, + "step": 80320 + }, + { + "epoch": 1.72, + "learning_rate": 4.253205128205128e-05, + "loss": 0.0284, + "step": 80330 + }, + { + "epoch": 1.72, + "learning_rate": 4.2499999999999996e-05, + "loss": 0.0272, + "step": 80340 + }, + { + "epoch": 1.72, + "learning_rate": 4.2467948717948716e-05, + "loss": 0.0318, + "step": 80350 + }, + { + "epoch": 1.72, + "learning_rate": 4.243589743589743e-05, + "loss": 0.0252, + "step": 80360 + }, + { + "epoch": 1.72, + "learning_rate": 4.240384615384615e-05, + "loss": 0.0329, + "step": 80370 + }, + { + "epoch": 1.72, + "learning_rate": 4.237179487179487e-05, + "loss": 0.0261, + "step": 80380 + }, + { + "epoch": 1.72, + "learning_rate": 4.233974358974359e-05, + "loss": 0.0249, + "step": 80390 + }, + { + "epoch": 1.72, + "learning_rate": 4.23076923076923e-05, + "loss": 0.0268, + "step": 80400 + }, + { + "epoch": 1.72, + "learning_rate": 4.227564102564102e-05, + "loss": 0.0261, + "step": 80410 + }, + { + "epoch": 1.72, + "learning_rate": 4.224358974358974e-05, + "loss": 0.0323, + "step": 80420 + }, + { + "epoch": 1.72, + "learning_rate": 4.221153846153846e-05, + "loss": 0.0317, + "step": 80430 + }, + { + "epoch": 1.72, + "learning_rate": 4.217948717948718e-05, + "loss": 0.0279, + "step": 80440 + }, + { + "epoch": 1.72, + "learning_rate": 4.214743589743589e-05, + "loss": 0.0295, + "step": 80450 + }, + { + "epoch": 1.72, + "learning_rate": 4.211538461538461e-05, + "loss": 0.03, + "step": 80460 + }, + { + "epoch": 1.72, + "learning_rate": 4.208333333333333e-05, + "loss": 0.0303, + "step": 80470 + }, + { + "epoch": 1.72, + "learning_rate": 4.205128205128205e-05, + "loss": 0.0317, + "step": 80480 + }, + { + "epoch": 1.72, + "learning_rate": 4.201923076923076e-05, + "loss": 0.0279, + "step": 80490 + }, + { + "epoch": 1.72, + "learning_rate": 4.198717948717948e-05, + "loss": 0.0219, + "step": 80500 + }, + { + "epoch": 1.72, + "learning_rate": 4.19551282051282e-05, + "loss": 0.0347, + "step": 80510 + }, + { + "epoch": 1.72, + "learning_rate": 4.192307692307692e-05, + "loss": 0.0302, + "step": 80520 + }, + { + "epoch": 1.72, + "learning_rate": 4.189102564102564e-05, + "loss": 0.0283, + "step": 80530 + }, + { + "epoch": 1.72, + "learning_rate": 4.1858974358974354e-05, + "loss": 0.028, + "step": 80540 + }, + { + "epoch": 1.72, + "learning_rate": 4.1826923076923074e-05, + "loss": 0.0324, + "step": 80550 + }, + { + "epoch": 1.72, + "learning_rate": 4.1794871794871794e-05, + "loss": 0.0268, + "step": 80560 + }, + { + "epoch": 1.72, + "learning_rate": 4.176282051282051e-05, + "loss": 0.0318, + "step": 80570 + }, + { + "epoch": 1.72, + "learning_rate": 4.1730769230769226e-05, + "loss": 0.0262, + "step": 80580 + }, + { + "epoch": 1.72, + "learning_rate": 4.1698717948717946e-05, + "loss": 0.0303, + "step": 80590 + }, + { + "epoch": 1.72, + "learning_rate": 4.1666666666666665e-05, + "loss": 0.0246, + "step": 80600 + }, + { + "epoch": 1.72, + "learning_rate": 4.1634615384615385e-05, + "loss": 0.0261, + "step": 80610 + }, + { + "epoch": 1.72, + "learning_rate": 4.16025641025641e-05, + "loss": 0.0309, + "step": 80620 + }, + { + "epoch": 1.72, + "learning_rate": 4.157051282051282e-05, + "loss": 0.0303, + "step": 80630 + }, + { + "epoch": 1.72, + "learning_rate": 4.153846153846154e-05, + "loss": 0.0361, + "step": 80640 + }, + { + "epoch": 1.72, + "learning_rate": 4.1506410256410257e-05, + "loss": 0.0295, + "step": 80650 + }, + { + "epoch": 1.72, + "learning_rate": 4.1474358974358976e-05, + "loss": 0.0258, + "step": 80660 + }, + { + "epoch": 1.72, + "learning_rate": 4.144230769230769e-05, + "loss": 0.027, + "step": 80670 + }, + { + "epoch": 1.72, + "learning_rate": 4.141025641025641e-05, + "loss": 0.0268, + "step": 80680 + }, + { + "epoch": 1.72, + "learning_rate": 4.137820512820513e-05, + "loss": 0.0328, + "step": 80690 + }, + { + "epoch": 1.72, + "learning_rate": 4.134615384615385e-05, + "loss": 0.0291, + "step": 80700 + }, + { + "epoch": 1.72, + "learning_rate": 4.131410256410256e-05, + "loss": 0.0286, + "step": 80710 + }, + { + "epoch": 1.72, + "learning_rate": 4.128205128205128e-05, + "loss": 0.0305, + "step": 80720 + }, + { + "epoch": 1.72, + "learning_rate": 4.125e-05, + "loss": 0.0317, + "step": 80730 + }, + { + "epoch": 1.73, + "learning_rate": 4.121794871794872e-05, + "loss": 0.0336, + "step": 80740 + }, + { + "epoch": 1.73, + "learning_rate": 4.118589743589744e-05, + "loss": 0.0229, + "step": 80750 + }, + { + "epoch": 1.73, + "learning_rate": 4.115384615384615e-05, + "loss": 0.0315, + "step": 80760 + }, + { + "epoch": 1.73, + "learning_rate": 4.112179487179487e-05, + "loss": 0.0263, + "step": 80770 + }, + { + "epoch": 1.73, + "learning_rate": 4.108974358974359e-05, + "loss": 0.0285, + "step": 80780 + }, + { + "epoch": 1.73, + "learning_rate": 4.105769230769231e-05, + "loss": 0.0277, + "step": 80790 + }, + { + "epoch": 1.73, + "learning_rate": 4.1025641025641023e-05, + "loss": 0.0276, + "step": 80800 + }, + { + "epoch": 1.73, + "learning_rate": 4.0993589743589736e-05, + "loss": 0.0266, + "step": 80810 + }, + { + "epoch": 1.73, + "learning_rate": 4.0961538461538456e-05, + "loss": 0.0245, + "step": 80820 + }, + { + "epoch": 1.73, + "learning_rate": 4.092948717948717e-05, + "loss": 0.0267, + "step": 80830 + }, + { + "epoch": 1.73, + "learning_rate": 4.089743589743589e-05, + "loss": 0.0308, + "step": 80840 + }, + { + "epoch": 1.73, + "learning_rate": 4.086538461538461e-05, + "loss": 0.0314, + "step": 80850 + }, + { + "epoch": 1.73, + "learning_rate": 4.083333333333333e-05, + "loss": 0.0294, + "step": 80860 + }, + { + "epoch": 1.73, + "learning_rate": 4.080128205128205e-05, + "loss": 0.0241, + "step": 80870 + }, + { + "epoch": 1.73, + "learning_rate": 4.076923076923076e-05, + "loss": 0.0229, + "step": 80880 + }, + { + "epoch": 1.73, + "learning_rate": 4.073717948717948e-05, + "loss": 0.0288, + "step": 80890 + }, + { + "epoch": 1.73, + "learning_rate": 4.07051282051282e-05, + "loss": 0.0331, + "step": 80900 + }, + { + "epoch": 1.73, + "learning_rate": 4.067307692307692e-05, + "loss": 0.0288, + "step": 80910 + }, + { + "epoch": 1.73, + "learning_rate": 4.064102564102563e-05, + "loss": 0.0321, + "step": 80920 + }, + { + "epoch": 1.73, + "learning_rate": 4.060897435897435e-05, + "loss": 0.0364, + "step": 80930 + }, + { + "epoch": 1.73, + "learning_rate": 4.057692307692307e-05, + "loss": 0.0302, + "step": 80940 + }, + { + "epoch": 1.73, + "learning_rate": 4.054487179487179e-05, + "loss": 0.0239, + "step": 80950 + }, + { + "epoch": 1.73, + "learning_rate": 4.051282051282051e-05, + "loss": 0.027, + "step": 80960 + }, + { + "epoch": 1.73, + "learning_rate": 4.048076923076922e-05, + "loss": 0.0245, + "step": 80970 + }, + { + "epoch": 1.73, + "learning_rate": 4.044871794871794e-05, + "loss": 0.0345, + "step": 80980 + }, + { + "epoch": 1.73, + "learning_rate": 4.041666666666666e-05, + "loss": 0.0231, + "step": 80990 + }, + { + "epoch": 1.73, + "learning_rate": 4.038461538461538e-05, + "loss": 0.0259, + "step": 81000 + }, + { + "epoch": 1.73, + "eval_cer": 0.9235500990677525, + "eval_loss": 0.021532561630010605, + "eval_runtime": 120.3706, + "eval_samples_per_second": 16.615, + "eval_steps_per_second": 4.154, + "step": 81000 + }, + { + "epoch": 1.73, + "learning_rate": 4.0352564102564094e-05, + "loss": 0.0302, + "step": 81010 + }, + { + "epoch": 1.73, + "learning_rate": 4.0320512820512814e-05, + "loss": 0.029, + "step": 81020 + }, + { + "epoch": 1.73, + "learning_rate": 4.0288461538461534e-05, + "loss": 0.0313, + "step": 81030 + }, + { + "epoch": 1.73, + "learning_rate": 4.025641025641025e-05, + "loss": 0.0296, + "step": 81040 + }, + { + "epoch": 1.73, + "learning_rate": 4.022435897435897e-05, + "loss": 0.0276, + "step": 81050 + }, + { + "epoch": 1.73, + "learning_rate": 4.0192307692307686e-05, + "loss": 0.0322, + "step": 81060 + }, + { + "epoch": 1.73, + "learning_rate": 4.0160256410256405e-05, + "loss": 0.0332, + "step": 81070 + }, + { + "epoch": 1.73, + "learning_rate": 4.0128205128205125e-05, + "loss": 0.0271, + "step": 81080 + }, + { + "epoch": 1.73, + "learning_rate": 4.0096153846153844e-05, + "loss": 0.0321, + "step": 81090 + }, + { + "epoch": 1.73, + "learning_rate": 4.006410256410256e-05, + "loss": 0.0324, + "step": 81100 + }, + { + "epoch": 1.73, + "learning_rate": 4.003205128205128e-05, + "loss": 0.0255, + "step": 81110 + }, + { + "epoch": 1.73, + "learning_rate": 3.9999999999999996e-05, + "loss": 0.0332, + "step": 81120 + }, + { + "epoch": 1.73, + "learning_rate": 3.9967948717948716e-05, + "loss": 0.0338, + "step": 81130 + }, + { + "epoch": 1.73, + "learning_rate": 3.993589743589743e-05, + "loss": 0.0296, + "step": 81140 + }, + { + "epoch": 1.73, + "learning_rate": 3.990384615384615e-05, + "loss": 0.0314, + "step": 81150 + }, + { + "epoch": 1.73, + "learning_rate": 3.987179487179487e-05, + "loss": 0.0244, + "step": 81160 + }, + { + "epoch": 1.73, + "learning_rate": 3.983974358974359e-05, + "loss": 0.0259, + "step": 81170 + }, + { + "epoch": 1.73, + "learning_rate": 3.980769230769231e-05, + "loss": 0.0229, + "step": 81180 + }, + { + "epoch": 1.73, + "learning_rate": 3.977564102564102e-05, + "loss": 0.0286, + "step": 81190 + }, + { + "epoch": 1.74, + "learning_rate": 3.974358974358974e-05, + "loss": 0.0259, + "step": 81200 + }, + { + "epoch": 1.74, + "learning_rate": 3.971153846153846e-05, + "loss": 0.0257, + "step": 81210 + }, + { + "epoch": 1.74, + "learning_rate": 3.967948717948718e-05, + "loss": 0.0325, + "step": 81220 + }, + { + "epoch": 1.74, + "learning_rate": 3.964743589743589e-05, + "loss": 0.0282, + "step": 81230 + }, + { + "epoch": 1.74, + "learning_rate": 3.961538461538461e-05, + "loss": 0.0294, + "step": 81240 + }, + { + "epoch": 1.74, + "learning_rate": 3.958333333333333e-05, + "loss": 0.0238, + "step": 81250 + }, + { + "epoch": 1.74, + "learning_rate": 3.955128205128205e-05, + "loss": 0.0273, + "step": 81260 + }, + { + "epoch": 1.74, + "learning_rate": 3.951923076923077e-05, + "loss": 0.0252, + "step": 81270 + }, + { + "epoch": 1.74, + "learning_rate": 3.948717948717948e-05, + "loss": 0.0263, + "step": 81280 + }, + { + "epoch": 1.74, + "learning_rate": 3.94551282051282e-05, + "loss": 0.0291, + "step": 81290 + }, + { + "epoch": 1.74, + "learning_rate": 3.942307692307692e-05, + "loss": 0.0316, + "step": 81300 + }, + { + "epoch": 1.74, + "learning_rate": 3.939102564102564e-05, + "loss": 0.0265, + "step": 81310 + }, + { + "epoch": 1.74, + "learning_rate": 3.9358974358974355e-05, + "loss": 0.0292, + "step": 81320 + }, + { + "epoch": 1.74, + "learning_rate": 3.9326923076923074e-05, + "loss": 0.0286, + "step": 81330 + }, + { + "epoch": 1.74, + "learning_rate": 3.9294871794871794e-05, + "loss": 0.0293, + "step": 81340 + }, + { + "epoch": 1.74, + "learning_rate": 3.9262820512820513e-05, + "loss": 0.0268, + "step": 81350 + }, + { + "epoch": 1.74, + "learning_rate": 3.9230769230769226e-05, + "loss": 0.0291, + "step": 81360 + }, + { + "epoch": 1.74, + "learning_rate": 3.9198717948717946e-05, + "loss": 0.0369, + "step": 81370 + }, + { + "epoch": 1.74, + "learning_rate": 3.9166666666666665e-05, + "loss": 0.0217, + "step": 81380 + }, + { + "epoch": 1.74, + "learning_rate": 3.9134615384615385e-05, + "loss": 0.0237, + "step": 81390 + }, + { + "epoch": 1.74, + "learning_rate": 3.9102564102564105e-05, + "loss": 0.0268, + "step": 81400 + }, + { + "epoch": 1.74, + "learning_rate": 3.907051282051282e-05, + "loss": 0.0291, + "step": 81410 + }, + { + "epoch": 1.74, + "learning_rate": 3.903846153846154e-05, + "loss": 0.025, + "step": 81420 + }, + { + "epoch": 1.74, + "learning_rate": 3.900641025641026e-05, + "loss": 0.0279, + "step": 81430 + }, + { + "epoch": 1.74, + "learning_rate": 3.8974358974358976e-05, + "loss": 0.0277, + "step": 81440 + }, + { + "epoch": 1.74, + "learning_rate": 3.894230769230769e-05, + "loss": 0.0299, + "step": 81450 + }, + { + "epoch": 1.74, + "learning_rate": 3.891025641025641e-05, + "loss": 0.0238, + "step": 81460 + }, + { + "epoch": 1.74, + "learning_rate": 3.887820512820513e-05, + "loss": 0.0265, + "step": 81470 + }, + { + "epoch": 1.74, + "learning_rate": 3.884615384615385e-05, + "loss": 0.024, + "step": 81480 + }, + { + "epoch": 1.74, + "learning_rate": 3.881410256410257e-05, + "loss": 0.0264, + "step": 81490 + }, + { + "epoch": 1.74, + "learning_rate": 3.878205128205128e-05, + "loss": 0.0253, + "step": 81500 + }, + { + "epoch": 1.74, + "learning_rate": 3.875e-05, + "loss": 0.0278, + "step": 81510 + }, + { + "epoch": 1.74, + "learning_rate": 3.871794871794872e-05, + "loss": 0.0294, + "step": 81520 + }, + { + "epoch": 1.74, + "learning_rate": 3.868589743589744e-05, + "loss": 0.0301, + "step": 81530 + }, + { + "epoch": 1.74, + "learning_rate": 3.8653846153846145e-05, + "loss": 0.0269, + "step": 81540 + }, + { + "epoch": 1.74, + "learning_rate": 3.8621794871794865e-05, + "loss": 0.0311, + "step": 81550 + }, + { + "epoch": 1.74, + "learning_rate": 3.8589743589743584e-05, + "loss": 0.0317, + "step": 81560 + }, + { + "epoch": 1.74, + "learning_rate": 3.85576923076923e-05, + "loss": 0.028, + "step": 81570 + }, + { + "epoch": 1.74, + "learning_rate": 3.852564102564102e-05, + "loss": 0.0268, + "step": 81580 + }, + { + "epoch": 1.74, + "learning_rate": 3.8493589743589736e-05, + "loss": 0.0283, + "step": 81590 + }, + { + "epoch": 1.74, + "learning_rate": 3.8461538461538456e-05, + "loss": 0.0272, + "step": 81600 + }, + { + "epoch": 1.74, + "learning_rate": 3.8429487179487176e-05, + "loss": 0.0294, + "step": 81610 + }, + { + "epoch": 1.74, + "learning_rate": 3.839743589743589e-05, + "loss": 0.0309, + "step": 81620 + }, + { + "epoch": 1.74, + "learning_rate": 3.836538461538461e-05, + "loss": 0.0261, + "step": 81630 + }, + { + "epoch": 1.74, + "learning_rate": 3.833333333333333e-05, + "loss": 0.0271, + "step": 81640 + }, + { + "epoch": 1.74, + "learning_rate": 3.830128205128205e-05, + "loss": 0.0249, + "step": 81650 + }, + { + "epoch": 1.74, + "learning_rate": 3.826923076923076e-05, + "loss": 0.0297, + "step": 81660 + }, + { + "epoch": 1.75, + "learning_rate": 3.823717948717948e-05, + "loss": 0.0275, + "step": 81670 + }, + { + "epoch": 1.75, + "learning_rate": 3.82051282051282e-05, + "loss": 0.0253, + "step": 81680 + }, + { + "epoch": 1.75, + "learning_rate": 3.817307692307692e-05, + "loss": 0.0327, + "step": 81690 + }, + { + "epoch": 1.75, + "learning_rate": 3.814102564102564e-05, + "loss": 0.0276, + "step": 81700 + }, + { + "epoch": 1.75, + "learning_rate": 3.810897435897435e-05, + "loss": 0.0315, + "step": 81710 + }, + { + "epoch": 1.75, + "learning_rate": 3.807692307692307e-05, + "loss": 0.0305, + "step": 81720 + }, + { + "epoch": 1.75, + "learning_rate": 3.804487179487179e-05, + "loss": 0.0242, + "step": 81730 + }, + { + "epoch": 1.75, + "learning_rate": 3.801282051282051e-05, + "loss": 0.0252, + "step": 81740 + }, + { + "epoch": 1.75, + "learning_rate": 3.798076923076922e-05, + "loss": 0.0245, + "step": 81750 + }, + { + "epoch": 1.75, + "learning_rate": 3.794871794871794e-05, + "loss": 0.0296, + "step": 81760 + }, + { + "epoch": 1.75, + "learning_rate": 3.791666666666666e-05, + "loss": 0.0271, + "step": 81770 + }, + { + "epoch": 1.75, + "learning_rate": 3.788461538461538e-05, + "loss": 0.03, + "step": 81780 + }, + { + "epoch": 1.75, + "learning_rate": 3.7852564102564095e-05, + "loss": 0.0305, + "step": 81790 + }, + { + "epoch": 1.75, + "learning_rate": 3.7820512820512814e-05, + "loss": 0.0278, + "step": 81800 + }, + { + "epoch": 1.75, + "learning_rate": 3.7788461538461534e-05, + "loss": 0.0301, + "step": 81810 + }, + { + "epoch": 1.75, + "learning_rate": 3.7756410256410253e-05, + "loss": 0.032, + "step": 81820 + }, + { + "epoch": 1.75, + "learning_rate": 3.772435897435897e-05, + "loss": 0.03, + "step": 81830 + }, + { + "epoch": 1.75, + "learning_rate": 3.7692307692307686e-05, + "loss": 0.0322, + "step": 81840 + }, + { + "epoch": 1.75, + "learning_rate": 3.7660256410256405e-05, + "loss": 0.0248, + "step": 81850 + }, + { + "epoch": 1.75, + "learning_rate": 3.7628205128205125e-05, + "loss": 0.0269, + "step": 81860 + }, + { + "epoch": 1.75, + "learning_rate": 3.7596153846153845e-05, + "loss": 0.0301, + "step": 81870 + }, + { + "epoch": 1.75, + "learning_rate": 3.756410256410256e-05, + "loss": 0.0264, + "step": 81880 + }, + { + "epoch": 1.75, + "learning_rate": 3.753205128205128e-05, + "loss": 0.0288, + "step": 81890 + }, + { + "epoch": 1.75, + "learning_rate": 3.75e-05, + "loss": 0.024, + "step": 81900 + }, + { + "epoch": 1.75, + "learning_rate": 3.7467948717948716e-05, + "loss": 0.0255, + "step": 81910 + }, + { + "epoch": 1.75, + "learning_rate": 3.7435897435897436e-05, + "loss": 0.026, + "step": 81920 + }, + { + "epoch": 1.75, + "learning_rate": 3.740384615384615e-05, + "loss": 0.0275, + "step": 81930 + }, + { + "epoch": 1.75, + "learning_rate": 3.737179487179487e-05, + "loss": 0.0307, + "step": 81940 + }, + { + "epoch": 1.75, + "learning_rate": 3.733974358974359e-05, + "loss": 0.0287, + "step": 81950 + }, + { + "epoch": 1.75, + "learning_rate": 3.730769230769231e-05, + "loss": 0.0262, + "step": 81960 + }, + { + "epoch": 1.75, + "learning_rate": 3.727564102564102e-05, + "loss": 0.0266, + "step": 81970 + }, + { + "epoch": 1.75, + "learning_rate": 3.724358974358974e-05, + "loss": 0.0255, + "step": 81980 + }, + { + "epoch": 1.75, + "learning_rate": 3.721153846153846e-05, + "loss": 0.0341, + "step": 81990 + }, + { + "epoch": 1.75, + "learning_rate": 3.717948717948718e-05, + "loss": 0.0294, + "step": 82000 + }, + { + "epoch": 1.75, + "eval_cer": 0.9234974441114349, + "eval_loss": 0.02169678919017315, + "eval_runtime": 120.3833, + "eval_samples_per_second": 16.614, + "eval_steps_per_second": 4.153, + "step": 82000 + } + ], + "max_steps": 93600, + "num_train_epochs": 2, + "total_flos": 4.443123472807219e+17, + "trial_name": null, + "trial_params": null +}