diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,46909 @@ +{ + "best_metric": 0.1670677214860916, + "best_model_checkpoint": "./wav2vec2-xls-r-300m/checkpoint-77000", + "epoch": 4.997403946002077, + "global_step": 77000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.8e-06, + "loss": 16.9743, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.0799999999999998e-05, + "loss": 14.7096, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 1.68e-05, + "loss": 13.7597, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 2.28e-05, + "loss": 13.6175, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 2.8199999999999998e-05, + "loss": 12.0065, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 3.42e-05, + "loss": 10.6388, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 4.02e-05, + "loss": 7.2542, + "step": 70 + }, + { + "epoch": 0.01, + "learning_rate": 4.62e-05, + "loss": 6.4431, + "step": 80 + }, + { + "epoch": 0.01, + "learning_rate": 5.2199999999999995e-05, + "loss": 5.8211, + "step": 90 + }, + { + "epoch": 0.01, + "learning_rate": 5.82e-05, + "loss": 5.6051, + "step": 100 + }, + { + "epoch": 0.01, + "learning_rate": 6.419999999999999e-05, + "loss": 5.0519, + "step": 110 + }, + { + "epoch": 0.01, + "learning_rate": 7.02e-05, + "loss": 4.5556, + "step": 120 + }, + { + "epoch": 0.01, + "learning_rate": 7.62e-05, + "loss": 4.0521, + "step": 130 + }, + { + "epoch": 0.01, + "learning_rate": 8.22e-05, + "loss": 3.8713, + "step": 140 + }, + { + "epoch": 0.01, + "learning_rate": 8.819999999999999e-05, + "loss": 3.6606, + "step": 150 + }, + { + "epoch": 0.01, + "learning_rate": 9.419999999999999e-05, + "loss": 3.5816, + "step": 160 + }, + { + "epoch": 0.01, + "learning_rate": 0.0001002, + "loss": 3.3292, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 0.00010619999999999998, + "loss": 3.2553, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 0.00011219999999999999, + "loss": 3.2468, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 0.0001182, + "loss": 3.3054, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 0.00012419999999999998, + "loss": 3.1961, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 0.0001302, + "loss": 3.1635, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 0.0001362, + "loss": 3.2034, + "step": 230 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001422, + "loss": 3.1785, + "step": 240 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001482, + "loss": 3.2516, + "step": 250 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015419999999999998, + "loss": 3.1703, + "step": 260 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001602, + "loss": 3.1605, + "step": 270 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001662, + "loss": 3.1715, + "step": 280 + }, + { + "epoch": 0.02, + "learning_rate": 0.00017219999999999998, + "loss": 3.1869, + "step": 290 + }, + { + "epoch": 0.02, + "learning_rate": 0.00017819999999999997, + "loss": 3.2258, + "step": 300 + }, + { + "epoch": 0.02, + "learning_rate": 0.00018419999999999998, + "loss": 3.1942, + "step": 310 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001902, + "loss": 3.1084, + "step": 320 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001962, + "loss": 3.112, + "step": 330 + }, + { + "epoch": 0.02, + "learning_rate": 0.0002022, + "loss": 3.1277, + "step": 340 + }, + { + "epoch": 0.02, + "learning_rate": 0.00020819999999999996, + "loss": 3.2239, + "step": 350 + }, + { + "epoch": 0.02, + "learning_rate": 0.00021419999999999998, + "loss": 3.1243, + "step": 360 + }, + { + "epoch": 0.02, + "learning_rate": 0.00022019999999999999, + "loss": 3.093, + "step": 370 + }, + { + "epoch": 0.02, + "learning_rate": 0.00022619999999999997, + "loss": 3.1214, + "step": 380 + }, + { + "epoch": 0.03, + "learning_rate": 0.00023219999999999998, + "loss": 3.1254, + "step": 390 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002382, + "loss": 3.1427, + "step": 400 + }, + { + "epoch": 0.03, + "learning_rate": 0.00024419999999999997, + "loss": 3.1655, + "step": 410 + }, + { + "epoch": 0.03, + "learning_rate": 0.00025019999999999996, + "loss": 3.1289, + "step": 420 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002562, + "loss": 3.124, + "step": 430 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002622, + "loss": 3.1255, + "step": 440 + }, + { + "epoch": 0.03, + "learning_rate": 0.00026819999999999996, + "loss": 3.1791, + "step": 450 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002742, + "loss": 3.1565, + "step": 460 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002802, + "loss": 3.0892, + "step": 470 + }, + { + "epoch": 0.03, + "learning_rate": 0.00028619999999999996, + "loss": 3.0283, + "step": 480 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029219999999999995, + "loss": 2.9589, + "step": 490 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002982, + "loss": 2.8604, + "step": 500 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029997256336556044, + "loss": 2.7027, + "step": 510 + }, + { + "epoch": 0.03, + "learning_rate": 0.000299933368173504, + "loss": 2.383, + "step": 520 + }, + { + "epoch": 0.03, + "learning_rate": 0.00029989417298144756, + "loss": 1.9945, + "step": 530 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002998549777893911, + "loss": 1.7479, + "step": 540 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002998157825973347, + "loss": 1.7199, + "step": 550 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029977658740527825, + "loss": 1.6166, + "step": 560 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002997373922132218, + "loss": 1.3413, + "step": 570 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002996981970211654, + "loss": 1.2923, + "step": 580 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029965900182910894, + "loss": 1.2102, + "step": 590 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002996198066370525, + "loss": 1.2882, + "step": 600 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029958061144499606, + "loss": 1.2009, + "step": 610 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002995414162529396, + "loss": 1.0415, + "step": 620 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002995022210608832, + "loss": 1.0745, + "step": 630 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029946302586882675, + "loss": 0.9673, + "step": 640 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002994238306767703, + "loss": 1.2104, + "step": 650 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002993846354847139, + "loss": 0.9191, + "step": 660 + }, + { + "epoch": 0.04, + "learning_rate": 0.00029934544029265744, + "loss": 0.889, + "step": 670 + }, + { + "epoch": 0.04, + "learning_rate": 0.000299306245100601, + "loss": 0.8939, + "step": 680 + }, + { + "epoch": 0.04, + "learning_rate": 0.0002992670499085445, + "loss": 0.8803, + "step": 690 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029922785471648807, + "loss": 0.9765, + "step": 700 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029918865952443163, + "loss": 0.8623, + "step": 710 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002991494643323752, + "loss": 0.7088, + "step": 720 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029911026914031876, + "loss": 0.7844, + "step": 730 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002990710739482623, + "loss": 0.8041, + "step": 740 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002990318787562059, + "loss": 0.9041, + "step": 750 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029899268356414944, + "loss": 0.6925, + "step": 760 + }, + { + "epoch": 0.05, + "learning_rate": 0.000298953488372093, + "loss": 0.6302, + "step": 770 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029891429318003657, + "loss": 0.7329, + "step": 780 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029887509798798013, + "loss": 0.7678, + "step": 790 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002988359027959237, + "loss": 0.8388, + "step": 800 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029879670760386725, + "loss": 0.6808, + "step": 810 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002987575124118108, + "loss": 0.584, + "step": 820 + }, + { + "epoch": 0.05, + "learning_rate": 0.0002987183172197544, + "loss": 0.6117, + "step": 830 + }, + { + "epoch": 0.05, + "learning_rate": 0.00029867912202769794, + "loss": 0.5984, + "step": 840 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029863992683564145, + "loss": 0.7709, + "step": 850 + }, + { + "epoch": 0.06, + "learning_rate": 0.000298600731643585, + "loss": 0.6299, + "step": 860 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002985615364515286, + "loss": 0.6055, + "step": 870 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029852234125947214, + "loss": 0.6358, + "step": 880 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002984831460674157, + "loss": 0.6254, + "step": 890 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029844395087535926, + "loss": 0.8013, + "step": 900 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002984047556833028, + "loss": 0.6341, + "step": 910 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002983655604912464, + "loss": 0.4995, + "step": 920 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029832636529918995, + "loss": 0.5411, + "step": 930 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002982871701071335, + "loss": 0.713, + "step": 940 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029824797491507707, + "loss": 0.8155, + "step": 950 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002982087797230206, + "loss": 0.5836, + "step": 960 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029816958453096414, + "loss": 0.5819, + "step": 970 + }, + { + "epoch": 0.06, + "learning_rate": 0.0002981303893389077, + "loss": 0.5416, + "step": 980 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029809119414685127, + "loss": 0.6206, + "step": 990 + }, + { + "epoch": 0.06, + "learning_rate": 0.00029805199895479483, + "loss": 0.7448, + "step": 1000 + }, + { + "epoch": 0.06, + "eval_loss": 0.6239742040634155, + "eval_runtime": 1769.7335, + "eval_samples_per_second": 8.808, + "eval_steps_per_second": 1.101, + "eval_wer": 0.5583005216699123, + "step": 1000 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002980128037627384, + "loss": 0.5615, + "step": 1010 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029797360857068195, + "loss": 0.5542, + "step": 1020 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002979344133786255, + "loss": 0.51, + "step": 1030 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002978952181865691, + "loss": 0.5501, + "step": 1040 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029785602299451264, + "loss": 0.6871, + "step": 1050 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002978168278024562, + "loss": 0.5419, + "step": 1060 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029777763261039977, + "loss": 0.5369, + "step": 1070 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029773843741834333, + "loss": 0.5446, + "step": 1080 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002976992422262869, + "loss": 0.5413, + "step": 1090 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029766004703423045, + "loss": 0.6633, + "step": 1100 + }, + { + "epoch": 0.07, + "learning_rate": 0.000297620851842174, + "loss": 0.4938, + "step": 1110 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002975816566501176, + "loss": 0.5184, + "step": 1120 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029754246145806114, + "loss": 0.4816, + "step": 1130 + }, + { + "epoch": 0.07, + "learning_rate": 0.0002975032662660047, + "loss": 0.6463, + "step": 1140 + }, + { + "epoch": 0.07, + "learning_rate": 0.00029746407107394826, + "loss": 0.6026, + "step": 1150 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002974248758818918, + "loss": 0.4514, + "step": 1160 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029738568068983533, + "loss": 0.5025, + "step": 1170 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002973464854977789, + "loss": 0.4657, + "step": 1180 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029730729030572246, + "loss": 0.5764, + "step": 1190 + }, + { + "epoch": 0.08, + "learning_rate": 0.000297268095113666, + "loss": 0.7248, + "step": 1200 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002972288999216096, + "loss": 0.511, + "step": 1210 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029718970472955315, + "loss": 0.4752, + "step": 1220 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002971505095374967, + "loss": 0.5, + "step": 1230 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029711131434544027, + "loss": 0.5382, + "step": 1240 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029707211915338383, + "loss": 0.6556, + "step": 1250 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002970329239613274, + "loss": 0.4285, + "step": 1260 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029699372876927096, + "loss": 0.4856, + "step": 1270 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002969545335772145, + "loss": 0.5096, + "step": 1280 + }, + { + "epoch": 0.08, + "learning_rate": 0.0002969153383851581, + "loss": 0.5342, + "step": 1290 + }, + { + "epoch": 0.08, + "learning_rate": 0.00029687614319310164, + "loss": 0.6266, + "step": 1300 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002968369480010452, + "loss": 0.4646, + "step": 1310 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029679775280898877, + "loss": 0.4688, + "step": 1320 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029675855761693233, + "loss": 0.4479, + "step": 1330 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029671936242487584, + "loss": 0.5453, + "step": 1340 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002966801672328194, + "loss": 0.5819, + "step": 1350 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029664097204076296, + "loss": 0.4571, + "step": 1360 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002966017768487065, + "loss": 0.4397, + "step": 1370 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002965625816566501, + "loss": 0.4093, + "step": 1380 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029652338646459365, + "loss": 0.4869, + "step": 1390 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002964841912725372, + "loss": 0.6004, + "step": 1400 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002964449960804808, + "loss": 0.4854, + "step": 1410 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029640580088842434, + "loss": 0.4639, + "step": 1420 + }, + { + "epoch": 0.09, + "learning_rate": 0.0002963666056963679, + "loss": 0.5067, + "step": 1430 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029632741050431146, + "loss": 0.5542, + "step": 1440 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029628821531225497, + "loss": 0.6948, + "step": 1450 + }, + { + "epoch": 0.09, + "learning_rate": 0.00029624902012019853, + "loss": 0.4594, + "step": 1460 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002962098249281421, + "loss": 0.4528, + "step": 1470 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029617062973608566, + "loss": 0.4562, + "step": 1480 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002961314345440292, + "loss": 0.4138, + "step": 1490 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002960922393519728, + "loss": 0.6161, + "step": 1500 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029605304415991634, + "loss": 0.4293, + "step": 1510 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002960138489678599, + "loss": 0.4386, + "step": 1520 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029597465377580347, + "loss": 0.46, + "step": 1530 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029593545858374703, + "loss": 0.5067, + "step": 1540 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002958962633916906, + "loss": 0.6368, + "step": 1550 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029585706819963416, + "loss": 0.4602, + "step": 1560 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002958178730075777, + "loss": 0.3584, + "step": 1570 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002957786778155213, + "loss": 0.4114, + "step": 1580 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029573948262346484, + "loss": 0.5129, + "step": 1590 + }, + { + "epoch": 0.1, + "learning_rate": 0.0002957002874314084, + "loss": 0.5897, + "step": 1600 + }, + { + "epoch": 0.1, + "learning_rate": 0.00029566109223935197, + "loss": 0.4176, + "step": 1610 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029562189704729553, + "loss": 0.4691, + "step": 1620 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002955827018552391, + "loss": 0.3983, + "step": 1630 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029554350666318265, + "loss": 0.4544, + "step": 1640 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029550431147112616, + "loss": 0.5603, + "step": 1650 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002954651162790697, + "loss": 0.457, + "step": 1660 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002954259210870133, + "loss": 0.3916, + "step": 1670 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029538672589495685, + "loss": 0.4459, + "step": 1680 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002953475307029004, + "loss": 0.4676, + "step": 1690 + }, + { + "epoch": 0.11, + "learning_rate": 0.000295308335510844, + "loss": 0.561, + "step": 1700 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029526914031878754, + "loss": 0.3632, + "step": 1710 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002952299451267311, + "loss": 0.4031, + "step": 1720 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029519074993467466, + "loss": 0.4096, + "step": 1730 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002951515547426182, + "loss": 0.4542, + "step": 1740 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002951123595505618, + "loss": 0.5244, + "step": 1750 + }, + { + "epoch": 0.11, + "learning_rate": 0.00029507316435850535, + "loss": 0.3749, + "step": 1760 + }, + { + "epoch": 0.11, + "learning_rate": 0.0002950339691664489, + "loss": 0.4329, + "step": 1770 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029499477397439247, + "loss": 0.4162, + "step": 1780 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029495557878233603, + "loss": 0.4706, + "step": 1790 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002949163835902796, + "loss": 0.5679, + "step": 1800 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029487718839822316, + "loss": 0.3496, + "step": 1810 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002948379932061667, + "loss": 0.3657, + "step": 1820 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029479879801411023, + "loss": 0.377, + "step": 1830 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002947596028220538, + "loss": 0.4545, + "step": 1840 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029472040762999735, + "loss": 0.5802, + "step": 1850 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002946812124379409, + "loss": 0.3955, + "step": 1860 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002946420172458845, + "loss": 0.3586, + "step": 1870 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029460282205382804, + "loss": 0.41, + "step": 1880 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002945636268617716, + "loss": 0.4732, + "step": 1890 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029452443166971516, + "loss": 0.5313, + "step": 1900 + }, + { + "epoch": 0.12, + "learning_rate": 0.00029448523647765873, + "loss": 0.3887, + "step": 1910 + }, + { + "epoch": 0.12, + "learning_rate": 0.0002944460412856023, + "loss": 0.3236, + "step": 1920 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002944068460935458, + "loss": 0.4082, + "step": 1930 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029436765090148936, + "loss": 0.4512, + "step": 1940 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002943284557094329, + "loss": 0.5833, + "step": 1950 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002942892605173765, + "loss": 0.3868, + "step": 1960 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029425006532532005, + "loss": 0.3627, + "step": 1970 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002942108701332636, + "loss": 0.3977, + "step": 1980 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029417167494120717, + "loss": 0.4719, + "step": 1990 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029413247974915073, + "loss": 0.5027, + "step": 2000 + }, + { + "epoch": 0.13, + "eval_loss": 0.43298059701919556, + "eval_runtime": 1393.5439, + "eval_samples_per_second": 11.186, + "eval_steps_per_second": 1.399, + "eval_wer": 0.42938977413256385, + "step": 2000 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002940932845570943, + "loss": 0.3667, + "step": 2010 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029405408936503786, + "loss": 0.369, + "step": 2020 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002940148941729814, + "loss": 0.3745, + "step": 2030 + }, + { + "epoch": 0.13, + "learning_rate": 0.000293975698980925, + "loss": 0.4639, + "step": 2040 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029393650378886855, + "loss": 0.5466, + "step": 2050 + }, + { + "epoch": 0.13, + "learning_rate": 0.0002938973085968121, + "loss": 0.41, + "step": 2060 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029385811340475567, + "loss": 0.3669, + "step": 2070 + }, + { + "epoch": 0.13, + "learning_rate": 0.00029381891821269923, + "loss": 0.3843, + "step": 2080 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002937797230206428, + "loss": 0.3565, + "step": 2090 + }, + { + "epoch": 0.14, + "learning_rate": 0.00029374052782858636, + "loss": 0.5322, + "step": 2100 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002937013326365299, + "loss": 0.3334, + "step": 2110 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002936621374444735, + "loss": 0.3393, + "step": 2120 + }, + { + "epoch": 0.14, + "learning_rate": 0.000293622942252417, + "loss": 0.349, + "step": 2130 + }, + { + "epoch": 0.14, + "learning_rate": 0.00029358374706036055, + "loss": 0.4454, + "step": 2140 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002935445518683041, + "loss": 0.5646, + "step": 2150 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002935053566762477, + "loss": 0.3602, + "step": 2160 + }, + { + "epoch": 0.14, + "learning_rate": 0.00029346616148419124, + "loss": 0.3602, + "step": 2170 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002934269662921348, + "loss": 0.3942, + "step": 2180 + }, + { + "epoch": 0.14, + "learning_rate": 0.00029338777110007836, + "loss": 0.4186, + "step": 2190 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002933485759080219, + "loss": 0.518, + "step": 2200 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002933093807159655, + "loss": 0.3857, + "step": 2210 + }, + { + "epoch": 0.14, + "learning_rate": 0.00029327018552390905, + "loss": 0.3821, + "step": 2220 + }, + { + "epoch": 0.14, + "learning_rate": 0.0002932309903318526, + "loss": 0.4039, + "step": 2230 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002931917951397962, + "loss": 0.4331, + "step": 2240 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029315259994773974, + "loss": 0.4773, + "step": 2250 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002931134047556833, + "loss": 0.3371, + "step": 2260 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029307420956362686, + "loss": 0.3418, + "step": 2270 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002930350143715704, + "loss": 0.3798, + "step": 2280 + }, + { + "epoch": 0.15, + "learning_rate": 0.000292995819179514, + "loss": 0.4313, + "step": 2290 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029295662398745755, + "loss": 0.5536, + "step": 2300 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002929174287954011, + "loss": 0.3915, + "step": 2310 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002928782336033446, + "loss": 0.3319, + "step": 2320 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002928390384112882, + "loss": 0.347, + "step": 2330 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029279984321923174, + "loss": 0.3754, + "step": 2340 + }, + { + "epoch": 0.15, + "learning_rate": 0.0002927606480271753, + "loss": 0.5592, + "step": 2350 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029272145283511887, + "loss": 0.3628, + "step": 2360 + }, + { + "epoch": 0.15, + "learning_rate": 0.00029268225764306243, + "loss": 0.3683, + "step": 2370 + }, + { + "epoch": 0.15, + "learning_rate": 0.000292643062451006, + "loss": 0.3621, + "step": 2380 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029260386725894955, + "loss": 0.3917, + "step": 2390 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002925646720668931, + "loss": 0.4622, + "step": 2400 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002925254768748366, + "loss": 0.3721, + "step": 2410 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002924862816827802, + "loss": 0.3105, + "step": 2420 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029244708649072375, + "loss": 0.3904, + "step": 2430 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002924078912986673, + "loss": 0.4013, + "step": 2440 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002923686961066109, + "loss": 0.5803, + "step": 2450 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029232950091455444, + "loss": 0.3823, + "step": 2460 + }, + { + "epoch": 0.16, + "learning_rate": 0.000292290305722498, + "loss": 0.3642, + "step": 2470 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029225111053044156, + "loss": 0.3729, + "step": 2480 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002922119153383851, + "loss": 0.4141, + "step": 2490 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002921727201463287, + "loss": 0.46, + "step": 2500 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029213352495427225, + "loss": 0.3702, + "step": 2510 + }, + { + "epoch": 0.16, + "learning_rate": 0.0002920943297622158, + "loss": 0.3994, + "step": 2520 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029205513457015937, + "loss": 0.3718, + "step": 2530 + }, + { + "epoch": 0.16, + "learning_rate": 0.00029201593937810293, + "loss": 0.3716, + "step": 2540 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002919767441860465, + "loss": 0.5318, + "step": 2550 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029193754899399006, + "loss": 0.3288, + "step": 2560 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002918983538019336, + "loss": 0.3576, + "step": 2570 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002918591586098772, + "loss": 0.3847, + "step": 2580 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029181996341782075, + "loss": 0.3863, + "step": 2590 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002917807682257643, + "loss": 0.5548, + "step": 2600 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002917415730337078, + "loss": 0.3594, + "step": 2610 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002917023778416514, + "loss": 0.3157, + "step": 2620 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029166318264959494, + "loss": 0.3687, + "step": 2630 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002916239874575385, + "loss": 0.4431, + "step": 2640 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029158479226548207, + "loss": 0.4957, + "step": 2650 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029154559707342563, + "loss": 0.3238, + "step": 2660 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002915064018813692, + "loss": 0.3875, + "step": 2670 + }, + { + "epoch": 0.17, + "learning_rate": 0.00029146720668931275, + "loss": 0.353, + "step": 2680 + }, + { + "epoch": 0.17, + "learning_rate": 0.0002914280114972563, + "loss": 0.4398, + "step": 2690 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002913888163051999, + "loss": 0.5356, + "step": 2700 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029134962111314344, + "loss": 0.3455, + "step": 2710 + }, + { + "epoch": 0.18, + "learning_rate": 0.000291310425921087, + "loss": 0.3809, + "step": 2720 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029127123072903056, + "loss": 0.415, + "step": 2730 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002912320355369741, + "loss": 0.4039, + "step": 2740 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002911928403449177, + "loss": 0.4672, + "step": 2750 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029115364515286125, + "loss": 0.3827, + "step": 2760 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002911144499608048, + "loss": 0.2917, + "step": 2770 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002910752547687484, + "loss": 0.3653, + "step": 2780 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029103605957669194, + "loss": 0.3637, + "step": 2790 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002909968643846355, + "loss": 0.5101, + "step": 2800 + }, + { + "epoch": 0.18, + "learning_rate": 0.000290957669192579, + "loss": 0.3407, + "step": 2810 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029091847400052257, + "loss": 0.365, + "step": 2820 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029087927880846613, + "loss": 0.3586, + "step": 2830 + }, + { + "epoch": 0.18, + "learning_rate": 0.0002908400836164097, + "loss": 0.3468, + "step": 2840 + }, + { + "epoch": 0.18, + "learning_rate": 0.00029080088842435326, + "loss": 0.4659, + "step": 2850 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002907616932322968, + "loss": 0.3723, + "step": 2860 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002907224980402404, + "loss": 0.3827, + "step": 2870 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029068330284818394, + "loss": 0.3876, + "step": 2880 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029064410765612745, + "loss": 0.3913, + "step": 2890 + }, + { + "epoch": 0.19, + "learning_rate": 0.000290604912464071, + "loss": 0.5514, + "step": 2900 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002905657172720146, + "loss": 0.2999, + "step": 2910 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029052652207995814, + "loss": 0.4309, + "step": 2920 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002904873268879017, + "loss": 0.3425, + "step": 2930 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029044813169584526, + "loss": 0.4039, + "step": 2940 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002904089365037888, + "loss": 0.5332, + "step": 2950 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002903697413117324, + "loss": 0.3437, + "step": 2960 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029033054611967595, + "loss": 0.3525, + "step": 2970 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002902913509276195, + "loss": 0.3929, + "step": 2980 + }, + { + "epoch": 0.19, + "learning_rate": 0.0002902521557355631, + "loss": 0.3778, + "step": 2990 + }, + { + "epoch": 0.19, + "learning_rate": 0.00029021296054350664, + "loss": 0.5066, + "step": 3000 + }, + { + "epoch": 0.19, + "eval_loss": 0.4207065999507904, + "eval_runtime": 1413.7737, + "eval_samples_per_second": 11.026, + "eval_steps_per_second": 1.379, + "eval_wer": 0.40069792895623374, + "step": 3000 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002901737653514502, + "loss": 0.3298, + "step": 3010 + }, + { + "epoch": 0.2, + "learning_rate": 0.00029013457015939376, + "loss": 0.3484, + "step": 3020 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002900953749673373, + "loss": 0.3536, + "step": 3030 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002900561797752809, + "loss": 0.3651, + "step": 3040 + }, + { + "epoch": 0.2, + "learning_rate": 0.00029001698458322445, + "loss": 0.5463, + "step": 3050 + }, + { + "epoch": 0.2, + "learning_rate": 0.000289977789391168, + "loss": 0.3716, + "step": 3060 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002899385941991116, + "loss": 0.3216, + "step": 3070 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028989939900705514, + "loss": 0.3356, + "step": 3080 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028986020381499864, + "loss": 0.4041, + "step": 3090 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002898210086229422, + "loss": 0.4808, + "step": 3100 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028978181343088577, + "loss": 0.3187, + "step": 3110 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028974261823882933, + "loss": 0.3011, + "step": 3120 + }, + { + "epoch": 0.2, + "learning_rate": 0.0002897034230467729, + "loss": 0.3774, + "step": 3130 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028966422785471646, + "loss": 0.4352, + "step": 3140 + }, + { + "epoch": 0.2, + "learning_rate": 0.00028962503266266, + "loss": 0.474, + "step": 3150 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002895858374706036, + "loss": 0.3177, + "step": 3160 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028954664227854714, + "loss": 0.3381, + "step": 3170 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002895074470864907, + "loss": 0.3596, + "step": 3180 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028946825189443427, + "loss": 0.4179, + "step": 3190 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028942905670237783, + "loss": 0.4624, + "step": 3200 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002893898615103214, + "loss": 0.3397, + "step": 3210 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028935066631826495, + "loss": 0.3268, + "step": 3220 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002893114711262085, + "loss": 0.3912, + "step": 3230 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002892722759341521, + "loss": 0.3563, + "step": 3240 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028923308074209564, + "loss": 0.4438, + "step": 3250 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002891938855500392, + "loss": 0.3044, + "step": 3260 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028915469035798277, + "loss": 0.2965, + "step": 3270 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028911549516592633, + "loss": 0.3361, + "step": 3280 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028907629997386984, + "loss": 0.3644, + "step": 3290 + }, + { + "epoch": 0.21, + "learning_rate": 0.0002890371047818134, + "loss": 0.54, + "step": 3300 + }, + { + "epoch": 0.21, + "learning_rate": 0.00028899790958975696, + "loss": 0.3397, + "step": 3310 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002889587143977005, + "loss": 0.3249, + "step": 3320 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002889195192056441, + "loss": 0.3499, + "step": 3330 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028888032401358765, + "loss": 0.3792, + "step": 3340 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002888411288215312, + "loss": 0.5587, + "step": 3350 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028880193362947477, + "loss": 0.3354, + "step": 3360 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002887627384374183, + "loss": 0.3406, + "step": 3370 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028872354324536184, + "loss": 0.364, + "step": 3380 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002886843480533054, + "loss": 0.3502, + "step": 3390 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028864515286124897, + "loss": 0.438, + "step": 3400 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028860595766919253, + "loss": 0.3217, + "step": 3410 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002885667624771361, + "loss": 0.3355, + "step": 3420 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028852756728507965, + "loss": 0.3618, + "step": 3430 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002884883720930232, + "loss": 0.4404, + "step": 3440 + }, + { + "epoch": 0.22, + "learning_rate": 0.0002884491769009668, + "loss": 0.4472, + "step": 3450 + }, + { + "epoch": 0.22, + "learning_rate": 0.00028840998170891034, + "loss": 0.3557, + "step": 3460 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002883707865168539, + "loss": 0.3644, + "step": 3470 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028833159132479746, + "loss": 0.3307, + "step": 3480 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028829239613274103, + "loss": 0.336, + "step": 3490 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002882532009406846, + "loss": 0.4675, + "step": 3500 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028821400574862815, + "loss": 0.3544, + "step": 3510 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002881748105565717, + "loss": 0.3001, + "step": 3520 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002881356153645153, + "loss": 0.3863, + "step": 3530 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028809642017245884, + "loss": 0.4114, + "step": 3540 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002880572249804024, + "loss": 0.4567, + "step": 3550 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028801802978834596, + "loss": 0.3234, + "step": 3560 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028797883459628947, + "loss": 0.3064, + "step": 3570 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028793963940423303, + "loss": 0.3147, + "step": 3580 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002879004442121766, + "loss": 0.3255, + "step": 3590 + }, + { + "epoch": 0.23, + "learning_rate": 0.00028786124902012016, + "loss": 0.5227, + "step": 3600 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002878220538280637, + "loss": 0.3294, + "step": 3610 + }, + { + "epoch": 0.23, + "learning_rate": 0.0002877828586360073, + "loss": 0.3342, + "step": 3620 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028774366344395085, + "loss": 0.3445, + "step": 3630 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002877044682518944, + "loss": 0.4013, + "step": 3640 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028766527305983797, + "loss": 0.4694, + "step": 3650 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028762607786778153, + "loss": 0.3161, + "step": 3660 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002875868826757251, + "loss": 0.3281, + "step": 3670 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028754768748366866, + "loss": 0.3452, + "step": 3680 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002875084922916122, + "loss": 0.3612, + "step": 3690 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002874692970995558, + "loss": 0.4962, + "step": 3700 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028743010190749934, + "loss": 0.3374, + "step": 3710 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002873909067154429, + "loss": 0.3092, + "step": 3720 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028735171152338647, + "loss": 0.3421, + "step": 3730 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028731251633133003, + "loss": 0.4114, + "step": 3740 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002872733211392736, + "loss": 0.5185, + "step": 3750 + }, + { + "epoch": 0.24, + "learning_rate": 0.00028723412594721715, + "loss": 0.3472, + "step": 3760 + }, + { + "epoch": 0.24, + "learning_rate": 0.0002871949307551607, + "loss": 0.3191, + "step": 3770 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002871557355631042, + "loss": 0.339, + "step": 3780 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002871165403710478, + "loss": 0.3831, + "step": 3790 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028707734517899135, + "loss": 0.549, + "step": 3800 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002870381499869349, + "loss": 0.3129, + "step": 3810 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002869989547948785, + "loss": 0.3104, + "step": 3820 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028695975960282204, + "loss": 0.3641, + "step": 3830 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002869205644107656, + "loss": 0.4269, + "step": 3840 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002868813692187091, + "loss": 0.4537, + "step": 3850 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028684217402665267, + "loss": 0.3027, + "step": 3860 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028680297883459623, + "loss": 0.3383, + "step": 3870 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002867637836425398, + "loss": 0.2987, + "step": 3880 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028672458845048336, + "loss": 0.394, + "step": 3890 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002866853932584269, + "loss": 0.497, + "step": 3900 + }, + { + "epoch": 0.25, + "learning_rate": 0.0002866461980663705, + "loss": 0.3424, + "step": 3910 + }, + { + "epoch": 0.25, + "learning_rate": 0.00028660700287431404, + "loss": 0.3463, + "step": 3920 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002865678076822576, + "loss": 0.3457, + "step": 3930 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028652861249020117, + "loss": 0.3148, + "step": 3940 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028648941729814473, + "loss": 0.5063, + "step": 3950 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002864502221060883, + "loss": 0.2959, + "step": 3960 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028641102691403185, + "loss": 0.3475, + "step": 3970 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002863718317219754, + "loss": 0.3374, + "step": 3980 + }, + { + "epoch": 0.26, + "learning_rate": 0.000286332636529919, + "loss": 0.3299, + "step": 3990 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028629344133786254, + "loss": 0.4994, + "step": 4000 + }, + { + "epoch": 0.26, + "eval_loss": 0.3732546865940094, + "eval_runtime": 1389.2784, + "eval_samples_per_second": 11.22, + "eval_steps_per_second": 1.403, + "eval_wer": 0.37260309021904453, + "step": 4000 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002862542461458061, + "loss": 0.3212, + "step": 4010 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028621505095374967, + "loss": 0.2929, + "step": 4020 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028617585576169323, + "loss": 0.3651, + "step": 4030 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002861366605696368, + "loss": 0.3571, + "step": 4040 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028609746537758035, + "loss": 0.4808, + "step": 4050 + }, + { + "epoch": 0.26, + "learning_rate": 0.00028605827018552386, + "loss": 0.3389, + "step": 4060 + }, + { + "epoch": 0.26, + "learning_rate": 0.0002860190749934674, + "loss": 0.3413, + "step": 4070 + }, + { + "epoch": 0.26, + "learning_rate": 0.000285979879801411, + "loss": 0.3312, + "step": 4080 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028594068460935455, + "loss": 0.3325, + "step": 4090 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002859014894172981, + "loss": 0.4705, + "step": 4100 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028586229422524167, + "loss": 0.3748, + "step": 4110 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028582309903318523, + "loss": 0.3117, + "step": 4120 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002857839038411288, + "loss": 0.3398, + "step": 4130 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028574470864907236, + "loss": 0.4, + "step": 4140 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002857055134570159, + "loss": 0.5127, + "step": 4150 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002856663182649595, + "loss": 0.3551, + "step": 4160 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028562712307290305, + "loss": 0.3232, + "step": 4170 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002855879278808466, + "loss": 0.3485, + "step": 4180 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028554873268879017, + "loss": 0.3534, + "step": 4190 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028550953749673373, + "loss": 0.4534, + "step": 4200 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002854703423046773, + "loss": 0.3288, + "step": 4210 + }, + { + "epoch": 0.27, + "learning_rate": 0.00028543114711262086, + "loss": 0.3275, + "step": 4220 + }, + { + "epoch": 0.27, + "learning_rate": 0.0002853919519205644, + "loss": 0.2882, + "step": 4230 + }, + { + "epoch": 0.28, + "learning_rate": 0.000285352756728508, + "loss": 0.3311, + "step": 4240 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028531356153645154, + "loss": 0.4253, + "step": 4250 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028527436634439505, + "loss": 0.3743, + "step": 4260 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002852351711523386, + "loss": 0.3325, + "step": 4270 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002851959759602822, + "loss": 0.3108, + "step": 4280 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028515678076822574, + "loss": 0.3288, + "step": 4290 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002851175855761693, + "loss": 0.4143, + "step": 4300 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028507839038411286, + "loss": 0.2937, + "step": 4310 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002850391951920564, + "loss": 0.3236, + "step": 4320 + }, + { + "epoch": 0.28, + "learning_rate": 0.000285, + "loss": 0.3517, + "step": 4330 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002849608048079435, + "loss": 0.3643, + "step": 4340 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028492160961588706, + "loss": 0.4363, + "step": 4350 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002848824144238306, + "loss": 0.3076, + "step": 4360 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002848432192317742, + "loss": 0.3544, + "step": 4370 + }, + { + "epoch": 0.28, + "learning_rate": 0.00028480402403971775, + "loss": 0.3358, + "step": 4380 + }, + { + "epoch": 0.28, + "learning_rate": 0.0002847648288476613, + "loss": 0.3521, + "step": 4390 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028472563365560487, + "loss": 0.5092, + "step": 4400 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028468643846354843, + "loss": 0.307, + "step": 4410 + }, + { + "epoch": 0.29, + "learning_rate": 0.000284647243271492, + "loss": 0.3316, + "step": 4420 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028460804807943556, + "loss": 0.3301, + "step": 4430 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002845688528873791, + "loss": 0.4104, + "step": 4440 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002845296576953227, + "loss": 0.4933, + "step": 4450 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028449046250326624, + "loss": 0.3148, + "step": 4460 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002844512673112098, + "loss": 0.2993, + "step": 4470 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028441207211915337, + "loss": 0.3098, + "step": 4480 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028437287692709693, + "loss": 0.3334, + "step": 4490 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002843336817350405, + "loss": 0.5094, + "step": 4500 + }, + { + "epoch": 0.29, + "learning_rate": 0.00028429448654298406, + "loss": 0.3136, + "step": 4510 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002842552913509276, + "loss": 0.2701, + "step": 4520 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002842160961588712, + "loss": 0.3528, + "step": 4530 + }, + { + "epoch": 0.29, + "learning_rate": 0.0002841769009668147, + "loss": 0.3265, + "step": 4540 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028413770577475825, + "loss": 0.5483, + "step": 4550 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002840985105827018, + "loss": 0.2911, + "step": 4560 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002840593153906454, + "loss": 0.303, + "step": 4570 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028402012019858894, + "loss": 0.2898, + "step": 4580 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002839809250065325, + "loss": 0.3593, + "step": 4590 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028394172981447606, + "loss": 0.4172, + "step": 4600 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002839025346224196, + "loss": 0.3307, + "step": 4610 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002838633394303632, + "loss": 0.3504, + "step": 4620 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028382414423830675, + "loss": 0.3039, + "step": 4630 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002837849490462503, + "loss": 0.4089, + "step": 4640 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002837457538541939, + "loss": 0.4656, + "step": 4650 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028370655866213744, + "loss": 0.2686, + "step": 4660 + }, + { + "epoch": 0.3, + "learning_rate": 0.000283667363470081, + "loss": 0.328, + "step": 4670 + }, + { + "epoch": 0.3, + "learning_rate": 0.00028362816827802456, + "loss": 0.3386, + "step": 4680 + }, + { + "epoch": 0.3, + "learning_rate": 0.0002835889730859681, + "loss": 0.3551, + "step": 4690 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002835497778939117, + "loss": 0.4499, + "step": 4700 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028351058270185525, + "loss": 0.3139, + "step": 4710 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002834713875097988, + "loss": 0.2817, + "step": 4720 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028343219231774237, + "loss": 0.3318, + "step": 4730 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002833929971256859, + "loss": 0.3358, + "step": 4740 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028335380193362944, + "loss": 0.4955, + "step": 4750 + }, + { + "epoch": 0.31, + "learning_rate": 0.000283314606741573, + "loss": 0.2712, + "step": 4760 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028327541154951657, + "loss": 0.3247, + "step": 4770 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028323621635746013, + "loss": 0.2972, + "step": 4780 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002831970211654037, + "loss": 0.4467, + "step": 4790 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028315782597334725, + "loss": 0.503, + "step": 4800 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002831186307812908, + "loss": 0.3166, + "step": 4810 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002830794355892343, + "loss": 0.3302, + "step": 4820 + }, + { + "epoch": 0.31, + "learning_rate": 0.0002830402403971779, + "loss": 0.3159, + "step": 4830 + }, + { + "epoch": 0.31, + "learning_rate": 0.00028300104520512145, + "loss": 0.3868, + "step": 4840 + }, + { + "epoch": 0.31, + "learning_rate": 0.000282961850013065, + "loss": 0.4717, + "step": 4850 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002829226548210086, + "loss": 0.3137, + "step": 4860 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028288345962895214, + "loss": 0.3132, + "step": 4870 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002828442644368957, + "loss": 0.2978, + "step": 4880 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028280506924483926, + "loss": 0.3747, + "step": 4890 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002827658740527828, + "loss": 0.4814, + "step": 4900 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002827266788607264, + "loss": 0.3118, + "step": 4910 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028268748366866995, + "loss": 0.2766, + "step": 4920 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002826482884766135, + "loss": 0.3214, + "step": 4930 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028260909328455707, + "loss": 0.332, + "step": 4940 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028256989809250063, + "loss": 0.4176, + "step": 4950 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002825307029004442, + "loss": 0.2628, + "step": 4960 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028249150770838776, + "loss": 0.3002, + "step": 4970 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002824523125163313, + "loss": 0.3388, + "step": 4980 + }, + { + "epoch": 0.32, + "learning_rate": 0.0002824131173242749, + "loss": 0.3555, + "step": 4990 + }, + { + "epoch": 0.32, + "learning_rate": 0.00028237392213221845, + "loss": 0.4877, + "step": 5000 + }, + { + "epoch": 0.32, + "eval_loss": 0.37171265482902527, + "eval_runtime": 1669.5499, + "eval_samples_per_second": 9.337, + "eval_steps_per_second": 1.167, + "eval_wer": 0.36130971841196285, + "step": 5000 + }, + { + "epoch": 0.33, + "learning_rate": 0.000282334726940162, + "loss": 0.3077, + "step": 5010 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002822955317481055, + "loss": 0.3245, + "step": 5020 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002822563365560491, + "loss": 0.2978, + "step": 5030 + }, + { + "epoch": 0.33, + "learning_rate": 0.00028221714136399264, + "loss": 0.3321, + "step": 5040 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002821779461719362, + "loss": 0.4104, + "step": 5050 + }, + { + "epoch": 0.33, + "learning_rate": 0.00028213875097987976, + "loss": 0.2965, + "step": 5060 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002820995557878233, + "loss": 0.2953, + "step": 5070 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002820603605957669, + "loss": 0.31, + "step": 5080 + }, + { + "epoch": 0.33, + "learning_rate": 0.00028202116540371045, + "loss": 0.3281, + "step": 5090 + }, + { + "epoch": 0.33, + "learning_rate": 0.000281981970211654, + "loss": 0.4182, + "step": 5100 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002819427750195976, + "loss": 0.3064, + "step": 5110 + }, + { + "epoch": 0.33, + "learning_rate": 0.00028190357982754114, + "loss": 0.2637, + "step": 5120 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002818643846354847, + "loss": 0.3076, + "step": 5130 + }, + { + "epoch": 0.33, + "learning_rate": 0.00028182518944342826, + "loss": 0.3452, + "step": 5140 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002817859942513718, + "loss": 0.4275, + "step": 5150 + }, + { + "epoch": 0.33, + "learning_rate": 0.0002817467990593154, + "loss": 0.3079, + "step": 5160 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028170760386725895, + "loss": 0.3159, + "step": 5170 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002816684086752025, + "loss": 0.3234, + "step": 5180 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002816292134831461, + "loss": 0.3324, + "step": 5190 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028159001829108964, + "loss": 0.5642, + "step": 5200 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002815508230990332, + "loss": 0.3138, + "step": 5210 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002815116279069767, + "loss": 0.2983, + "step": 5220 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028147243271492027, + "loss": 0.3261, + "step": 5230 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028143323752286383, + "loss": 0.362, + "step": 5240 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002813940423308074, + "loss": 0.4757, + "step": 5250 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028135484713875096, + "loss": 0.269, + "step": 5260 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002813156519466945, + "loss": 0.2761, + "step": 5270 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002812764567546381, + "loss": 0.3187, + "step": 5280 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028123726156258164, + "loss": 0.288, + "step": 5290 + }, + { + "epoch": 0.34, + "learning_rate": 0.00028119806637052515, + "loss": 0.4421, + "step": 5300 + }, + { + "epoch": 0.34, + "learning_rate": 0.0002811588711784687, + "loss": 0.2711, + "step": 5310 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002811196759864123, + "loss": 0.2929, + "step": 5320 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028108048079435584, + "loss": 0.3481, + "step": 5330 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002810412856022994, + "loss": 0.3559, + "step": 5340 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028100209041024296, + "loss": 0.4442, + "step": 5350 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002809628952181865, + "loss": 0.3155, + "step": 5360 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002809237000261301, + "loss": 0.3093, + "step": 5370 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028088450483407365, + "loss": 0.3126, + "step": 5380 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002808453096420172, + "loss": 0.3667, + "step": 5390 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002808061144499608, + "loss": 0.4791, + "step": 5400 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028076691925790434, + "loss": 0.2708, + "step": 5410 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002807277240658479, + "loss": 0.2896, + "step": 5420 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028068852887379146, + "loss": 0.337, + "step": 5430 + }, + { + "epoch": 0.35, + "learning_rate": 0.000280649333681735, + "loss": 0.3713, + "step": 5440 + }, + { + "epoch": 0.35, + "learning_rate": 0.0002806101384896786, + "loss": 0.4635, + "step": 5450 + }, + { + "epoch": 0.35, + "learning_rate": 0.00028057094329762215, + "loss": 0.2561, + "step": 5460 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002805317481055657, + "loss": 0.2918, + "step": 5470 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028049255291350927, + "loss": 0.2964, + "step": 5480 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028045335772145283, + "loss": 0.3934, + "step": 5490 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028041416252939634, + "loss": 0.486, + "step": 5500 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002803749673373399, + "loss": 0.316, + "step": 5510 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028033577214528347, + "loss": 0.2586, + "step": 5520 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028029657695322703, + "loss": 0.306, + "step": 5530 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002802573817611706, + "loss": 0.3094, + "step": 5540 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028021818656911415, + "loss": 0.5131, + "step": 5550 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002801789913770577, + "loss": 0.2692, + "step": 5560 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002801397961850013, + "loss": 0.3111, + "step": 5570 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028010060099294484, + "loss": 0.3669, + "step": 5580 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002800614058008884, + "loss": 0.3268, + "step": 5590 + }, + { + "epoch": 0.36, + "learning_rate": 0.00028002221060883197, + "loss": 0.4569, + "step": 5600 + }, + { + "epoch": 0.36, + "learning_rate": 0.00027998301541677553, + "loss": 0.2858, + "step": 5610 + }, + { + "epoch": 0.36, + "learning_rate": 0.0002799438202247191, + "loss": 0.3284, + "step": 5620 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027990462503266265, + "loss": 0.3685, + "step": 5630 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002798654298406062, + "loss": 0.4039, + "step": 5640 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002798262346485498, + "loss": 0.4424, + "step": 5650 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027978703945649334, + "loss": 0.285, + "step": 5660 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002797478442644369, + "loss": 0.269, + "step": 5670 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027970864907238046, + "loss": 0.3057, + "step": 5680 + }, + { + "epoch": 0.37, + "learning_rate": 0.000279669453880324, + "loss": 0.3446, + "step": 5690 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027963025868826753, + "loss": 0.4615, + "step": 5700 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002795910634962111, + "loss": 0.3241, + "step": 5710 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027955186830415466, + "loss": 0.2815, + "step": 5720 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002795126731120982, + "loss": 0.3406, + "step": 5730 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002794734779200418, + "loss": 0.3716, + "step": 5740 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027943428272798535, + "loss": 0.4412, + "step": 5750 + }, + { + "epoch": 0.37, + "learning_rate": 0.0002793950875359289, + "loss": 0.3428, + "step": 5760 + }, + { + "epoch": 0.37, + "learning_rate": 0.00027935589234387247, + "loss": 0.2964, + "step": 5770 + }, + { + "epoch": 0.38, + "learning_rate": 0.000279316697151816, + "loss": 0.3145, + "step": 5780 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027927750195975954, + "loss": 0.3477, + "step": 5790 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002792383067677031, + "loss": 0.4081, + "step": 5800 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027919911157564667, + "loss": 0.2944, + "step": 5810 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027915991638359023, + "loss": 0.3528, + "step": 5820 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002791207211915338, + "loss": 0.3338, + "step": 5830 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027908152599947735, + "loss": 0.3684, + "step": 5840 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002790423308074209, + "loss": 0.4344, + "step": 5850 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002790031356153645, + "loss": 0.3295, + "step": 5860 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027896394042330804, + "loss": 0.3124, + "step": 5870 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002789247452312516, + "loss": 0.3123, + "step": 5880 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027888555003919516, + "loss": 0.3462, + "step": 5890 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002788463548471387, + "loss": 0.4603, + "step": 5900 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002788071596550823, + "loss": 0.2867, + "step": 5910 + }, + { + "epoch": 0.38, + "learning_rate": 0.00027876796446302585, + "loss": 0.3175, + "step": 5920 + }, + { + "epoch": 0.38, + "learning_rate": 0.0002787287692709694, + "loss": 0.3388, + "step": 5930 + }, + { + "epoch": 0.39, + "learning_rate": 0.000278689574078913, + "loss": 0.3519, + "step": 5940 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027865037888685654, + "loss": 0.4689, + "step": 5950 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002786111836948001, + "loss": 0.2766, + "step": 5960 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027857198850274366, + "loss": 0.2756, + "step": 5970 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027853279331068717, + "loss": 0.3184, + "step": 5980 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027849359811863073, + "loss": 0.4049, + "step": 5990 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002784544029265743, + "loss": 0.4966, + "step": 6000 + }, + { + "epoch": 0.39, + "eval_loss": 0.3752055764198303, + "eval_runtime": 2171.6421, + "eval_samples_per_second": 7.178, + "eval_steps_per_second": 0.897, + "eval_wer": 0.3694688064135549, + "step": 6000 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027841520773451786, + "loss": 0.2736, + "step": 6010 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002783760125424614, + "loss": 0.2942, + "step": 6020 + }, + { + "epoch": 0.39, + "learning_rate": 0.000278336817350405, + "loss": 0.3363, + "step": 6030 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027829762215834854, + "loss": 0.3495, + "step": 6040 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002782584269662921, + "loss": 0.4398, + "step": 6050 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027821923177423567, + "loss": 0.3443, + "step": 6060 + }, + { + "epoch": 0.39, + "learning_rate": 0.00027818003658217923, + "loss": 0.3313, + "step": 6070 + }, + { + "epoch": 0.39, + "learning_rate": 0.0002781408413901228, + "loss": 0.2949, + "step": 6080 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027810164619806636, + "loss": 0.3681, + "step": 6090 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002780624510060099, + "loss": 0.4029, + "step": 6100 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002780232558139535, + "loss": 0.325, + "step": 6110 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027798406062189704, + "loss": 0.2896, + "step": 6120 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002779448654298406, + "loss": 0.3116, + "step": 6130 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027790567023778417, + "loss": 0.3595, + "step": 6140 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027786647504572773, + "loss": 0.4125, + "step": 6150 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002778272798536713, + "loss": 0.2668, + "step": 6160 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027778808466161485, + "loss": 0.3145, + "step": 6170 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027774888946955836, + "loss": 0.3234, + "step": 6180 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002777096942775019, + "loss": 0.3792, + "step": 6190 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002776704990854455, + "loss": 0.3844, + "step": 6200 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027763130389338905, + "loss": 0.2977, + "step": 6210 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002775921087013326, + "loss": 0.2977, + "step": 6220 + }, + { + "epoch": 0.4, + "learning_rate": 0.0002775529135092762, + "loss": 0.3215, + "step": 6230 + }, + { + "epoch": 0.4, + "learning_rate": 0.00027751371831721974, + "loss": 0.3752, + "step": 6240 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002774745231251633, + "loss": 0.4476, + "step": 6250 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002774353279331068, + "loss": 0.3254, + "step": 6260 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027739613274105037, + "loss": 0.291, + "step": 6270 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027735693754899393, + "loss": 0.3052, + "step": 6280 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002773177423569375, + "loss": 0.3677, + "step": 6290 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027727854716488106, + "loss": 0.4331, + "step": 6300 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002772393519728246, + "loss": 0.2861, + "step": 6310 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002772001567807682, + "loss": 0.298, + "step": 6320 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027716096158871174, + "loss": 0.3236, + "step": 6330 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002771217663966553, + "loss": 0.3399, + "step": 6340 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027708257120459887, + "loss": 0.479, + "step": 6350 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027704337601254243, + "loss": 0.2635, + "step": 6360 + }, + { + "epoch": 0.41, + "learning_rate": 0.000277004180820486, + "loss": 0.2832, + "step": 6370 + }, + { + "epoch": 0.41, + "learning_rate": 0.00027696498562842955, + "loss": 0.3152, + "step": 6380 + }, + { + "epoch": 0.41, + "learning_rate": 0.0002769257904363731, + "loss": 0.3565, + "step": 6390 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002768865952443167, + "loss": 0.4582, + "step": 6400 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027684740005226024, + "loss": 0.2983, + "step": 6410 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002768082048602038, + "loss": 0.2945, + "step": 6420 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027676900966814736, + "loss": 0.343, + "step": 6430 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027672981447609093, + "loss": 0.3324, + "step": 6440 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002766906192840345, + "loss": 0.4225, + "step": 6450 + }, + { + "epoch": 0.42, + "learning_rate": 0.000276651424091978, + "loss": 0.3104, + "step": 6460 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027661222889992156, + "loss": 0.2767, + "step": 6470 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002765730337078651, + "loss": 0.3275, + "step": 6480 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002765338385158087, + "loss": 0.31, + "step": 6490 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027649464332375225, + "loss": 0.4626, + "step": 6500 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002764554481316958, + "loss": 0.3232, + "step": 6510 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027641625293963937, + "loss": 0.3007, + "step": 6520 + }, + { + "epoch": 0.42, + "learning_rate": 0.00027637705774758293, + "loss": 0.2785, + "step": 6530 + }, + { + "epoch": 0.42, + "learning_rate": 0.0002763378625555265, + "loss": 0.3881, + "step": 6540 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027629866736347006, + "loss": 0.4039, + "step": 6550 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002762594721714136, + "loss": 0.2897, + "step": 6560 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002762202769793572, + "loss": 0.3589, + "step": 6570 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027618108178730075, + "loss": 0.3608, + "step": 6580 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002761418865952443, + "loss": 0.3517, + "step": 6590 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027610269140318787, + "loss": 0.4061, + "step": 6600 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027606349621113143, + "loss": 0.2441, + "step": 6610 + }, + { + "epoch": 0.43, + "learning_rate": 0.000276024301019075, + "loss": 0.3492, + "step": 6620 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027598510582701856, + "loss": 0.32, + "step": 6630 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002759459106349621, + "loss": 0.3703, + "step": 6640 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002759067154429057, + "loss": 0.3989, + "step": 6650 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027586752025084924, + "loss": 0.2885, + "step": 6660 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027582832505879275, + "loss": 0.3087, + "step": 6670 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002757891298667363, + "loss": 0.3079, + "step": 6680 + }, + { + "epoch": 0.43, + "learning_rate": 0.0002757499346746799, + "loss": 0.3943, + "step": 6690 + }, + { + "epoch": 0.43, + "learning_rate": 0.00027571073948262344, + "loss": 0.4446, + "step": 6700 + }, + { + "epoch": 0.44, + "learning_rate": 0.000275671544290567, + "loss": 0.2932, + "step": 6710 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027563234909851056, + "loss": 0.3135, + "step": 6720 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002755931539064541, + "loss": 0.3367, + "step": 6730 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027555395871439763, + "loss": 0.3787, + "step": 6740 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002755147635223412, + "loss": 0.4106, + "step": 6750 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027547556833028476, + "loss": 0.3337, + "step": 6760 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002754363731382283, + "loss": 0.2976, + "step": 6770 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002753971779461719, + "loss": 0.3432, + "step": 6780 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027535798275411544, + "loss": 0.3794, + "step": 6790 + }, + { + "epoch": 0.44, + "learning_rate": 0.000275318787562059, + "loss": 0.4738, + "step": 6800 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027527959237000257, + "loss": 0.2745, + "step": 6810 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027524039717794613, + "loss": 0.2871, + "step": 6820 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002752012019858897, + "loss": 0.3232, + "step": 6830 + }, + { + "epoch": 0.44, + "learning_rate": 0.00027516200679383326, + "loss": 0.3259, + "step": 6840 + }, + { + "epoch": 0.44, + "learning_rate": 0.0002751228116017768, + "loss": 0.5016, + "step": 6850 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002750836164097204, + "loss": 0.2881, + "step": 6860 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027504442121766394, + "loss": 0.2774, + "step": 6870 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002750052260256075, + "loss": 0.3147, + "step": 6880 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027496603083355107, + "loss": 0.3563, + "step": 6890 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027492683564149463, + "loss": 0.3834, + "step": 6900 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002748876404494382, + "loss": 0.3038, + "step": 6910 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027484844525738175, + "loss": 0.2987, + "step": 6920 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002748092500653253, + "loss": 0.3347, + "step": 6930 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002747700548732689, + "loss": 0.3834, + "step": 6940 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002747308596812124, + "loss": 0.4511, + "step": 6950 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027469166448915595, + "loss": 0.2605, + "step": 6960 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002746524692970995, + "loss": 0.3179, + "step": 6970 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002746132741050431, + "loss": 0.3193, + "step": 6980 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027457407891298664, + "loss": 0.3176, + "step": 6990 + }, + { + "epoch": 0.45, + "learning_rate": 0.0002745348837209302, + "loss": 0.5035, + "step": 7000 + }, + { + "epoch": 0.45, + "eval_loss": 0.3448316156864166, + "eval_runtime": 1492.2585, + "eval_samples_per_second": 10.446, + "eval_steps_per_second": 1.306, + "eval_wer": 0.3543872866057341, + "step": 7000 + }, + { + "epoch": 0.45, + "learning_rate": 0.00027449568852887376, + "loss": 0.2597, + "step": 7010 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002744564933368173, + "loss": 0.3003, + "step": 7020 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002744172981447609, + "loss": 0.2838, + "step": 7030 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027437810295270445, + "loss": 0.3602, + "step": 7040 + }, + { + "epoch": 0.46, + "learning_rate": 0.000274338907760648, + "loss": 0.4495, + "step": 7050 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027429971256859157, + "loss": 0.2861, + "step": 7060 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027426051737653513, + "loss": 0.2684, + "step": 7070 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002742213221844787, + "loss": 0.3331, + "step": 7080 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027418212699242226, + "loss": 0.2959, + "step": 7090 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002741429318003658, + "loss": 0.4699, + "step": 7100 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002741037366083094, + "loss": 0.27, + "step": 7110 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027406454141625295, + "loss": 0.2639, + "step": 7120 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002740253462241965, + "loss": 0.3187, + "step": 7130 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027398615103214007, + "loss": 0.355, + "step": 7140 + }, + { + "epoch": 0.46, + "learning_rate": 0.0002739469558400836, + "loss": 0.463, + "step": 7150 + }, + { + "epoch": 0.46, + "learning_rate": 0.00027390776064802714, + "loss": 0.237, + "step": 7160 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002738685654559707, + "loss": 0.2737, + "step": 7170 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027382937026391427, + "loss": 0.312, + "step": 7180 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027379017507185783, + "loss": 0.3708, + "step": 7190 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002737509798798014, + "loss": 0.4522, + "step": 7200 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027371178468774495, + "loss": 0.2925, + "step": 7210 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002736725894956885, + "loss": 0.3192, + "step": 7220 + }, + { + "epoch": 0.47, + "learning_rate": 0.000273633394303632, + "loss": 0.3625, + "step": 7230 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002735941991115756, + "loss": 0.2908, + "step": 7240 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027355500391951915, + "loss": 0.4519, + "step": 7250 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002735158087274627, + "loss": 0.2514, + "step": 7260 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027347661353540627, + "loss": 0.2985, + "step": 7270 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027343741834334983, + "loss": 0.3839, + "step": 7280 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002733982231512934, + "loss": 0.3855, + "step": 7290 + }, + { + "epoch": 0.47, + "learning_rate": 0.00027335902795923696, + "loss": 0.423, + "step": 7300 + }, + { + "epoch": 0.47, + "learning_rate": 0.0002733198327671805, + "loss": 0.3237, + "step": 7310 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002732806375751241, + "loss": 0.3068, + "step": 7320 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027324144238306765, + "loss": 0.3065, + "step": 7330 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002732022471910112, + "loss": 0.3243, + "step": 7340 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027316305199895477, + "loss": 0.4884, + "step": 7350 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027312385680689833, + "loss": 0.3062, + "step": 7360 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002730846616148419, + "loss": 0.2859, + "step": 7370 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027304546642278546, + "loss": 0.3286, + "step": 7380 + }, + { + "epoch": 0.48, + "learning_rate": 0.000273006271230729, + "loss": 0.3014, + "step": 7390 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002729670760386726, + "loss": 0.4712, + "step": 7400 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027292788084661614, + "loss": 0.3007, + "step": 7410 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002728886856545597, + "loss": 0.2931, + "step": 7420 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002728494904625032, + "loss": 0.3216, + "step": 7430 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002728102952704468, + "loss": 0.3394, + "step": 7440 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027277110007839034, + "loss": 0.4071, + "step": 7450 + }, + { + "epoch": 0.48, + "learning_rate": 0.0002727319048863339, + "loss": 0.2728, + "step": 7460 + }, + { + "epoch": 0.48, + "learning_rate": 0.00027269270969427746, + "loss": 0.2825, + "step": 7470 + }, + { + "epoch": 0.49, + "learning_rate": 0.000272653514502221, + "loss": 0.2925, + "step": 7480 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002726143193101646, + "loss": 0.3408, + "step": 7490 + }, + { + "epoch": 0.49, + "learning_rate": 0.00027257512411810815, + "loss": 0.3633, + "step": 7500 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002725359289260517, + "loss": 0.3042, + "step": 7510 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002724967337339953, + "loss": 0.2974, + "step": 7520 + }, + { + "epoch": 0.49, + "learning_rate": 0.00027245753854193884, + "loss": 0.3318, + "step": 7530 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002724183433498824, + "loss": 0.3571, + "step": 7540 + }, + { + "epoch": 0.49, + "learning_rate": 0.00027237914815782596, + "loss": 0.4373, + "step": 7550 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002723399529657695, + "loss": 0.2981, + "step": 7560 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002723007577737131, + "loss": 0.296, + "step": 7570 + }, + { + "epoch": 0.49, + "learning_rate": 0.00027226156258165665, + "loss": 0.3641, + "step": 7580 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002722223673896002, + "loss": 0.308, + "step": 7590 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002721831721975438, + "loss": 0.4388, + "step": 7600 + }, + { + "epoch": 0.49, + "learning_rate": 0.00027214397700548734, + "loss": 0.2855, + "step": 7610 + }, + { + "epoch": 0.49, + "learning_rate": 0.0002721047818134309, + "loss": 0.2879, + "step": 7620 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002720655866213744, + "loss": 0.3245, + "step": 7630 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027202639142931797, + "loss": 0.3285, + "step": 7640 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027198719623726153, + "loss": 0.4187, + "step": 7650 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002719480010452051, + "loss": 0.2906, + "step": 7660 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027190880585314866, + "loss": 0.3205, + "step": 7670 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002718696106610922, + "loss": 0.322, + "step": 7680 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002718304154690358, + "loss": 0.3339, + "step": 7690 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027179122027697934, + "loss": 0.3904, + "step": 7700 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002717520250849229, + "loss": 0.2805, + "step": 7710 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002717128298928664, + "loss": 0.2821, + "step": 7720 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027167363470081, + "loss": 0.2898, + "step": 7730 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027163443950875354, + "loss": 0.3181, + "step": 7740 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002715952443166971, + "loss": 0.4323, + "step": 7750 + }, + { + "epoch": 0.5, + "learning_rate": 0.00027155604912464066, + "loss": 0.3144, + "step": 7760 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002715168539325842, + "loss": 0.288, + "step": 7770 + }, + { + "epoch": 0.5, + "learning_rate": 0.0002714776587405278, + "loss": 0.2862, + "step": 7780 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027143846354847135, + "loss": 0.2961, + "step": 7790 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002713992683564149, + "loss": 0.4547, + "step": 7800 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002713600731643585, + "loss": 0.2913, + "step": 7810 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027132087797230204, + "loss": 0.3206, + "step": 7820 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002712816827802456, + "loss": 0.3229, + "step": 7830 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027124248758818916, + "loss": 0.3557, + "step": 7840 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002712032923961327, + "loss": 0.4533, + "step": 7850 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002711640972040763, + "loss": 0.2965, + "step": 7860 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027112490201201985, + "loss": 0.3152, + "step": 7870 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002710857068199634, + "loss": 0.2714, + "step": 7880 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027104651162790697, + "loss": 0.3023, + "step": 7890 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027100731643585053, + "loss": 0.42, + "step": 7900 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027096812124379404, + "loss": 0.2888, + "step": 7910 + }, + { + "epoch": 0.51, + "learning_rate": 0.0002709289260517376, + "loss": 0.2963, + "step": 7920 + }, + { + "epoch": 0.51, + "learning_rate": 0.00027088973085968117, + "loss": 0.3359, + "step": 7930 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027085053566762473, + "loss": 0.3187, + "step": 7940 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002708113404755683, + "loss": 0.4253, + "step": 7950 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027077214528351185, + "loss": 0.2691, + "step": 7960 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002707329500914554, + "loss": 0.2951, + "step": 7970 + }, + { + "epoch": 0.52, + "learning_rate": 0.000270693754899399, + "loss": 0.2567, + "step": 7980 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027065455970734254, + "loss": 0.3456, + "step": 7990 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002706153645152861, + "loss": 0.4245, + "step": 8000 + }, + { + "epoch": 0.52, + "eval_loss": 0.3476312458515167, + "eval_runtime": 1263.6612, + "eval_samples_per_second": 12.336, + "eval_steps_per_second": 1.542, + "eval_wer": 0.3422979062131313, + "step": 8000 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027057616932322966, + "loss": 0.271, + "step": 8010 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027053697413117323, + "loss": 0.2869, + "step": 8020 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002704977789391168, + "loss": 0.3205, + "step": 8030 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027045858374706035, + "loss": 0.3238, + "step": 8040 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002704193885550039, + "loss": 0.4726, + "step": 8050 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002703801933629475, + "loss": 0.2816, + "step": 8060 + }, + { + "epoch": 0.52, + "learning_rate": 0.00027034099817089104, + "loss": 0.2581, + "step": 8070 + }, + { + "epoch": 0.52, + "learning_rate": 0.0002703018029788346, + "loss": 0.2671, + "step": 8080 + }, + { + "epoch": 0.53, + "learning_rate": 0.00027026260778677816, + "loss": 0.3355, + "step": 8090 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002702234125947217, + "loss": 0.42, + "step": 8100 + }, + { + "epoch": 0.53, + "learning_rate": 0.00027018421740266523, + "loss": 0.2678, + "step": 8110 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002701450222106088, + "loss": 0.282, + "step": 8120 + }, + { + "epoch": 0.53, + "learning_rate": 0.00027010582701855236, + "loss": 0.3301, + "step": 8130 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002700666318264959, + "loss": 0.3685, + "step": 8140 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002700274366344395, + "loss": 0.4231, + "step": 8150 + }, + { + "epoch": 0.53, + "learning_rate": 0.00026998824144238305, + "loss": 0.2953, + "step": 8160 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002699490462503266, + "loss": 0.2789, + "step": 8170 + }, + { + "epoch": 0.53, + "learning_rate": 0.00026990985105827017, + "loss": 0.2639, + "step": 8180 + }, + { + "epoch": 0.53, + "learning_rate": 0.00026987065586621373, + "loss": 0.2956, + "step": 8190 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002698314606741573, + "loss": 0.3968, + "step": 8200 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002697922654821008, + "loss": 0.287, + "step": 8210 + }, + { + "epoch": 0.53, + "learning_rate": 0.00026975307029004436, + "loss": 0.2752, + "step": 8220 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002697138750979879, + "loss": 0.2833, + "step": 8230 + }, + { + "epoch": 0.53, + "learning_rate": 0.0002696746799059315, + "loss": 0.3694, + "step": 8240 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026963548471387505, + "loss": 0.4739, + "step": 8250 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002695962895218186, + "loss": 0.2812, + "step": 8260 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002695570943297622, + "loss": 0.2941, + "step": 8270 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026951789913770574, + "loss": 0.265, + "step": 8280 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002694787039456493, + "loss": 0.3324, + "step": 8290 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026943950875359286, + "loss": 0.4461, + "step": 8300 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002694003135615364, + "loss": 0.2827, + "step": 8310 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026936111836948, + "loss": 0.2805, + "step": 8320 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026932192317742355, + "loss": 0.3237, + "step": 8330 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002692827279853671, + "loss": 0.3399, + "step": 8340 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002692435327933107, + "loss": 0.3966, + "step": 8350 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026920433760125424, + "loss": 0.2925, + "step": 8360 + }, + { + "epoch": 0.54, + "learning_rate": 0.0002691651424091978, + "loss": 0.26, + "step": 8370 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026912594721714136, + "loss": 0.3132, + "step": 8380 + }, + { + "epoch": 0.54, + "learning_rate": 0.00026908675202508487, + "loss": 0.3565, + "step": 8390 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026904755683302843, + "loss": 0.4054, + "step": 8400 + }, + { + "epoch": 0.55, + "learning_rate": 0.000269008361640972, + "loss": 0.2842, + "step": 8410 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026896916644891556, + "loss": 0.275, + "step": 8420 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002689299712568591, + "loss": 0.3091, + "step": 8430 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002688907760648027, + "loss": 0.3488, + "step": 8440 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026885158087274624, + "loss": 0.4289, + "step": 8450 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002688123856806898, + "loss": 0.2521, + "step": 8460 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026877319048863337, + "loss": 0.227, + "step": 8470 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026873399529657693, + "loss": 0.2706, + "step": 8480 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002686948001045205, + "loss": 0.3152, + "step": 8490 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026865560491246405, + "loss": 0.3648, + "step": 8500 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002686164097204076, + "loss": 0.2457, + "step": 8510 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002685772145283512, + "loss": 0.2568, + "step": 8520 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026853801933629474, + "loss": 0.3024, + "step": 8530 + }, + { + "epoch": 0.55, + "learning_rate": 0.0002684988241442383, + "loss": 0.3569, + "step": 8540 + }, + { + "epoch": 0.55, + "learning_rate": 0.00026845962895218187, + "loss": 0.4214, + "step": 8550 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026842043376012543, + "loss": 0.2946, + "step": 8560 + }, + { + "epoch": 0.56, + "learning_rate": 0.000268381238568069, + "loss": 0.2773, + "step": 8570 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026834204337601255, + "loss": 0.3064, + "step": 8580 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026830284818395606, + "loss": 0.3183, + "step": 8590 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002682636529918996, + "loss": 0.4401, + "step": 8600 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002682244577998432, + "loss": 0.295, + "step": 8610 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026818526260778675, + "loss": 0.3193, + "step": 8620 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002681460674157303, + "loss": 0.3438, + "step": 8630 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026810687222367387, + "loss": 0.3036, + "step": 8640 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026806767703161743, + "loss": 0.3668, + "step": 8650 + }, + { + "epoch": 0.56, + "learning_rate": 0.000268028481839561, + "loss": 0.2522, + "step": 8660 + }, + { + "epoch": 0.56, + "learning_rate": 0.00026798928664750456, + "loss": 0.2852, + "step": 8670 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002679500914554481, + "loss": 0.2795, + "step": 8680 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002679108962633917, + "loss": 0.3367, + "step": 8690 + }, + { + "epoch": 0.56, + "learning_rate": 0.0002678717010713352, + "loss": 0.3705, + "step": 8700 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026783250587927875, + "loss": 0.2231, + "step": 8710 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002677933106872223, + "loss": 0.2547, + "step": 8720 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002677541154951659, + "loss": 0.3023, + "step": 8730 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026771492030310944, + "loss": 0.3134, + "step": 8740 + }, + { + "epoch": 0.57, + "learning_rate": 0.000267675725111053, + "loss": 0.4138, + "step": 8750 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026763652991899657, + "loss": 0.2511, + "step": 8760 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026759733472694013, + "loss": 0.2795, + "step": 8770 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002675581395348837, + "loss": 0.2775, + "step": 8780 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026751894434282725, + "loss": 0.2653, + "step": 8790 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002674797491507708, + "loss": 0.4733, + "step": 8800 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002674405539587144, + "loss": 0.2629, + "step": 8810 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026740135876665794, + "loss": 0.2945, + "step": 8820 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002673621635746015, + "loss": 0.2845, + "step": 8830 + }, + { + "epoch": 0.57, + "learning_rate": 0.00026732296838254506, + "loss": 0.2883, + "step": 8840 + }, + { + "epoch": 0.57, + "learning_rate": 0.0002672837731904886, + "loss": 0.4085, + "step": 8850 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002672445779984322, + "loss": 0.2782, + "step": 8860 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002672053828063757, + "loss": 0.2648, + "step": 8870 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026716618761431926, + "loss": 0.338, + "step": 8880 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002671269924222628, + "loss": 0.3059, + "step": 8890 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002670877972302064, + "loss": 0.4468, + "step": 8900 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026704860203814995, + "loss": 0.3086, + "step": 8910 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002670094068460935, + "loss": 0.2706, + "step": 8920 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026697021165403707, + "loss": 0.2933, + "step": 8930 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026693101646198063, + "loss": 0.3263, + "step": 8940 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002668918212699242, + "loss": 0.443, + "step": 8950 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026685262607786776, + "loss": 0.2808, + "step": 8960 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002668134308858113, + "loss": 0.3031, + "step": 8970 + }, + { + "epoch": 0.58, + "learning_rate": 0.0002667742356937549, + "loss": 0.2693, + "step": 8980 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026673504050169844, + "loss": 0.3406, + "step": 8990 + }, + { + "epoch": 0.58, + "learning_rate": 0.000266695845309642, + "loss": 0.412, + "step": 9000 + }, + { + "epoch": 0.58, + "eval_loss": 0.3567412495613098, + "eval_runtime": 1121.6032, + "eval_samples_per_second": 13.898, + "eval_steps_per_second": 1.738, + "eval_wer": 0.34035763528592344, + "step": 9000 + }, + { + "epoch": 0.58, + "learning_rate": 0.00026665665011758557, + "loss": 0.2406, + "step": 9010 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026661745492552913, + "loss": 0.2531, + "step": 9020 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002665782597334727, + "loss": 0.2775, + "step": 9030 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026653906454141626, + "loss": 0.3236, + "step": 9040 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002664998693493598, + "loss": 0.3841, + "step": 9050 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002664606741573034, + "loss": 0.2929, + "step": 9060 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002664214789652469, + "loss": 0.2646, + "step": 9070 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026638228377319045, + "loss": 0.2935, + "step": 9080 + }, + { + "epoch": 0.59, + "learning_rate": 0.000266343088581134, + "loss": 0.2991, + "step": 9090 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002663038933890776, + "loss": 0.4354, + "step": 9100 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026626469819702114, + "loss": 0.2527, + "step": 9110 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002662255030049647, + "loss": 0.2891, + "step": 9120 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026618630781290826, + "loss": 0.2791, + "step": 9130 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002661471126208518, + "loss": 0.2806, + "step": 9140 + }, + { + "epoch": 0.59, + "learning_rate": 0.0002661079174287954, + "loss": 0.3874, + "step": 9150 + }, + { + "epoch": 0.59, + "learning_rate": 0.00026606872223673895, + "loss": 0.2695, + "step": 9160 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002660295270446825, + "loss": 0.3203, + "step": 9170 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002659903318526261, + "loss": 0.3446, + "step": 9180 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002659511366605696, + "loss": 0.3292, + "step": 9190 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026591194146851314, + "loss": 0.3745, + "step": 9200 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002658727462764567, + "loss": 0.2703, + "step": 9210 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026583355108440027, + "loss": 0.2841, + "step": 9220 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026579435589234383, + "loss": 0.245, + "step": 9230 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002657551607002874, + "loss": 0.3454, + "step": 9240 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026571596550823096, + "loss": 0.4138, + "step": 9250 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002656767703161745, + "loss": 0.267, + "step": 9260 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002656375751241181, + "loss": 0.2891, + "step": 9270 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026559837993206164, + "loss": 0.2901, + "step": 9280 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002655591847400052, + "loss": 0.3338, + "step": 9290 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026551998954794877, + "loss": 0.4125, + "step": 9300 + }, + { + "epoch": 0.6, + "learning_rate": 0.00026548079435589233, + "loss": 0.2318, + "step": 9310 + }, + { + "epoch": 0.6, + "learning_rate": 0.0002654415991638359, + "loss": 0.2715, + "step": 9320 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026540240397177945, + "loss": 0.2924, + "step": 9330 + }, + { + "epoch": 0.61, + "learning_rate": 0.000265363208779723, + "loss": 0.3064, + "step": 9340 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002653240135876665, + "loss": 0.5218, + "step": 9350 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002652848183956101, + "loss": 0.2552, + "step": 9360 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026524562320355365, + "loss": 0.2795, + "step": 9370 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002652064280114972, + "loss": 0.2837, + "step": 9380 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002651672328194408, + "loss": 0.3295, + "step": 9390 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026512803762738434, + "loss": 0.4641, + "step": 9400 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002650888424353279, + "loss": 0.2707, + "step": 9410 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026504964724327146, + "loss": 0.2441, + "step": 9420 + }, + { + "epoch": 0.61, + "learning_rate": 0.000265010452051215, + "loss": 0.3014, + "step": 9430 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002649712568591586, + "loss": 0.3112, + "step": 9440 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026493206166710215, + "loss": 0.345, + "step": 9450 + }, + { + "epoch": 0.61, + "learning_rate": 0.0002648928664750457, + "loss": 0.2652, + "step": 9460 + }, + { + "epoch": 0.61, + "learning_rate": 0.00026485367128298927, + "loss": 0.266, + "step": 9470 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026481447609093283, + "loss": 0.3039, + "step": 9480 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002647752808988764, + "loss": 0.3264, + "step": 9490 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026473608570681996, + "loss": 0.4203, + "step": 9500 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002646968905147635, + "loss": 0.2601, + "step": 9510 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002646576953227071, + "loss": 0.2777, + "step": 9520 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026461850013065065, + "loss": 0.2967, + "step": 9530 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002645793049385942, + "loss": 0.2921, + "step": 9540 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026454010974653777, + "loss": 0.4108, + "step": 9550 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002645009145544813, + "loss": 0.2608, + "step": 9560 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026446171936242484, + "loss": 0.2949, + "step": 9570 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002644225241703684, + "loss": 0.2803, + "step": 9580 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026438332897831196, + "loss": 0.3388, + "step": 9590 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002643441337862555, + "loss": 0.4501, + "step": 9600 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002643049385941991, + "loss": 0.2851, + "step": 9610 + }, + { + "epoch": 0.62, + "learning_rate": 0.00026426574340214265, + "loss": 0.2657, + "step": 9620 + }, + { + "epoch": 0.62, + "learning_rate": 0.0002642265482100862, + "loss": 0.2952, + "step": 9630 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002641873530180298, + "loss": 0.3209, + "step": 9640 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026414815782597334, + "loss": 0.425, + "step": 9650 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002641089626339169, + "loss": 0.2609, + "step": 9660 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026406976744186046, + "loss": 0.2823, + "step": 9670 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026403057224980397, + "loss": 0.3142, + "step": 9680 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026399137705774753, + "loss": 0.3125, + "step": 9690 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002639521818656911, + "loss": 0.3882, + "step": 9700 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026391298667363466, + "loss": 0.2697, + "step": 9710 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002638737914815782, + "loss": 0.2715, + "step": 9720 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002638345962895218, + "loss": 0.2895, + "step": 9730 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026379540109746534, + "loss": 0.3432, + "step": 9740 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002637562059054089, + "loss": 0.3942, + "step": 9750 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026371701071335247, + "loss": 0.2749, + "step": 9760 + }, + { + "epoch": 0.63, + "learning_rate": 0.00026367781552129603, + "loss": 0.3062, + "step": 9770 + }, + { + "epoch": 0.63, + "learning_rate": 0.0002636386203292396, + "loss": 0.3386, + "step": 9780 + }, + { + "epoch": 0.64, + "learning_rate": 0.00026359942513718316, + "loss": 0.3178, + "step": 9790 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002635602299451267, + "loss": 0.4299, + "step": 9800 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002635210347530703, + "loss": 0.2611, + "step": 9810 + }, + { + "epoch": 0.64, + "learning_rate": 0.00026348183956101384, + "loss": 0.2533, + "step": 9820 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002634426443689574, + "loss": 0.2686, + "step": 9830 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002634034491769009, + "loss": 0.3521, + "step": 9840 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002633642539848445, + "loss": 0.4043, + "step": 9850 + }, + { + "epoch": 0.64, + "learning_rate": 0.00026332505879278804, + "loss": 0.2551, + "step": 9860 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002632858636007316, + "loss": 0.2676, + "step": 9870 + }, + { + "epoch": 0.64, + "learning_rate": 0.00026324666840867516, + "loss": 0.3007, + "step": 9880 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002632074732166187, + "loss": 0.2834, + "step": 9890 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002631682780245623, + "loss": 0.3676, + "step": 9900 + }, + { + "epoch": 0.64, + "learning_rate": 0.00026312908283250585, + "loss": 0.2799, + "step": 9910 + }, + { + "epoch": 0.64, + "learning_rate": 0.0002630898876404494, + "loss": 0.268, + "step": 9920 + }, + { + "epoch": 0.64, + "learning_rate": 0.000263050692448393, + "loss": 0.2367, + "step": 9930 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026301149725633654, + "loss": 0.3223, + "step": 9940 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002629723020642801, + "loss": 0.3884, + "step": 9950 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026293310687222366, + "loss": 0.2609, + "step": 9960 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002628939116801672, + "loss": 0.2794, + "step": 9970 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002628547164881108, + "loss": 0.2638, + "step": 9980 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026281552129605435, + "loss": 0.327, + "step": 9990 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002627763261039979, + "loss": 0.4708, + "step": 10000 + }, + { + "epoch": 0.65, + "eval_loss": 0.32637062668800354, + "eval_runtime": 967.4635, + "eval_samples_per_second": 16.112, + "eval_steps_per_second": 2.015, + "eval_wer": 0.32370541996560104, + "step": 10000 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026273713091194147, + "loss": 0.2355, + "step": 10010 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026269793571988503, + "loss": 0.2741, + "step": 10020 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002626587405278286, + "loss": 0.279, + "step": 10030 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002626195453357721, + "loss": 0.3086, + "step": 10040 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026258035014371567, + "loss": 0.3583, + "step": 10050 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026254115495165923, + "loss": 0.2472, + "step": 10060 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002625019597596028, + "loss": 0.2822, + "step": 10070 + }, + { + "epoch": 0.65, + "learning_rate": 0.00026246276456754635, + "loss": 0.2737, + "step": 10080 + }, + { + "epoch": 0.65, + "learning_rate": 0.0002624235693754899, + "loss": 0.29, + "step": 10090 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002623843741834335, + "loss": 0.3797, + "step": 10100 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026234517899137704, + "loss": 0.3134, + "step": 10110 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002623059837993206, + "loss": 0.2648, + "step": 10120 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026226678860726417, + "loss": 0.287, + "step": 10130 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026222759341520773, + "loss": 0.274, + "step": 10140 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002621883982231513, + "loss": 0.4522, + "step": 10150 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026214920303109485, + "loss": 0.2539, + "step": 10160 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026211000783903836, + "loss": 0.261, + "step": 10170 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002620708126469819, + "loss": 0.2748, + "step": 10180 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002620316174549255, + "loss": 0.3037, + "step": 10190 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026199242226286905, + "loss": 0.4251, + "step": 10200 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002619532270708126, + "loss": 0.2757, + "step": 10210 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026191403187875617, + "loss": 0.2729, + "step": 10220 + }, + { + "epoch": 0.66, + "learning_rate": 0.00026187483668669973, + "loss": 0.2658, + "step": 10230 + }, + { + "epoch": 0.66, + "learning_rate": 0.0002618356414946433, + "loss": 0.3277, + "step": 10240 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026179644630258686, + "loss": 0.3392, + "step": 10250 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002617572511105304, + "loss": 0.2643, + "step": 10260 + }, + { + "epoch": 0.67, + "learning_rate": 0.000261718055918474, + "loss": 0.2635, + "step": 10270 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026167886072641755, + "loss": 0.2994, + "step": 10280 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002616396655343611, + "loss": 0.3208, + "step": 10290 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026160047034230467, + "loss": 0.3974, + "step": 10300 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026156127515024823, + "loss": 0.2484, + "step": 10310 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026152207995819174, + "loss": 0.2721, + "step": 10320 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002614828847661353, + "loss": 0.2685, + "step": 10330 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026144368957407887, + "loss": 0.2861, + "step": 10340 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026140449438202243, + "loss": 0.4029, + "step": 10350 + }, + { + "epoch": 0.67, + "learning_rate": 0.000261365299189966, + "loss": 0.274, + "step": 10360 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026132610399790955, + "loss": 0.2482, + "step": 10370 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002612869088058531, + "loss": 0.2894, + "step": 10380 + }, + { + "epoch": 0.67, + "learning_rate": 0.0002612477136137967, + "loss": 0.3049, + "step": 10390 + }, + { + "epoch": 0.67, + "learning_rate": 0.00026120851842174024, + "loss": 0.4552, + "step": 10400 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002611693232296838, + "loss": 0.2858, + "step": 10410 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026113012803762736, + "loss": 0.286, + "step": 10420 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002610909328455709, + "loss": 0.2781, + "step": 10430 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002610517376535145, + "loss": 0.2818, + "step": 10440 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026101254246145805, + "loss": 0.3897, + "step": 10450 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002609733472694016, + "loss": 0.2402, + "step": 10460 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002609341520773452, + "loss": 0.2437, + "step": 10470 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026089495688528874, + "loss": 0.3215, + "step": 10480 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002608557616932323, + "loss": 0.2834, + "step": 10490 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026081656650117586, + "loss": 0.4895, + "step": 10500 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002607773713091194, + "loss": 0.25, + "step": 10510 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026073817611706293, + "loss": 0.2707, + "step": 10520 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002606989809250065, + "loss": 0.282, + "step": 10530 + }, + { + "epoch": 0.68, + "learning_rate": 0.00026065978573295006, + "loss": 0.2658, + "step": 10540 + }, + { + "epoch": 0.68, + "learning_rate": 0.0002606205905408936, + "loss": 0.3727, + "step": 10550 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002605813953488372, + "loss": 0.2642, + "step": 10560 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026054220015678074, + "loss": 0.2534, + "step": 10570 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002605030049647243, + "loss": 0.2861, + "step": 10580 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026046380977266787, + "loss": 0.28, + "step": 10590 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026042461458061143, + "loss": 0.3381, + "step": 10600 + }, + { + "epoch": 0.69, + "learning_rate": 0.000260385419388555, + "loss": 0.2686, + "step": 10610 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026034622419649856, + "loss": 0.2697, + "step": 10620 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002603070290044421, + "loss": 0.2367, + "step": 10630 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002602678338123857, + "loss": 0.3366, + "step": 10640 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026022863862032924, + "loss": 0.4595, + "step": 10650 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026018944342827275, + "loss": 0.2616, + "step": 10660 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002601502482362163, + "loss": 0.3172, + "step": 10670 + }, + { + "epoch": 0.69, + "learning_rate": 0.0002601110530441599, + "loss": 0.274, + "step": 10680 + }, + { + "epoch": 0.69, + "learning_rate": 0.00026007185785210344, + "loss": 0.3128, + "step": 10690 + }, + { + "epoch": 0.69, + "learning_rate": 0.000260032662660047, + "loss": 0.387, + "step": 10700 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025999346746799056, + "loss": 0.2528, + "step": 10710 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002599542722759341, + "loss": 0.267, + "step": 10720 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002599150770838777, + "loss": 0.3083, + "step": 10730 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025987588189182125, + "loss": 0.3077, + "step": 10740 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002598406062189705, + "loss": 0.4265, + "step": 10750 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025980141102691404, + "loss": 0.2777, + "step": 10760 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025976221583485755, + "loss": 0.2196, + "step": 10770 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002597230206428011, + "loss": 0.255, + "step": 10780 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002596838254507447, + "loss": 0.3909, + "step": 10790 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025964463025868824, + "loss": 0.3783, + "step": 10800 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002596054350666318, + "loss": 0.2563, + "step": 10810 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025956623987457536, + "loss": 0.2612, + "step": 10820 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002595270446825189, + "loss": 0.2673, + "step": 10830 + }, + { + "epoch": 0.7, + "learning_rate": 0.0002594878494904625, + "loss": 0.3055, + "step": 10840 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025944865429840605, + "loss": 0.3539, + "step": 10850 + }, + { + "epoch": 0.7, + "learning_rate": 0.00025940945910634956, + "loss": 0.2517, + "step": 10860 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002593702639142931, + "loss": 0.2828, + "step": 10870 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002593310687222367, + "loss": 0.3236, + "step": 10880 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025929187353018025, + "loss": 0.2983, + "step": 10890 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002592526783381238, + "loss": 0.3658, + "step": 10900 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025921348314606737, + "loss": 0.238, + "step": 10910 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025917428795401093, + "loss": 0.214, + "step": 10920 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002591350927619545, + "loss": 0.2757, + "step": 10930 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025909589756989806, + "loss": 0.3236, + "step": 10940 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002590567023778416, + "loss": 0.4002, + "step": 10950 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002590175071857852, + "loss": 0.2345, + "step": 10960 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025897831199372874, + "loss": 0.2601, + "step": 10970 + }, + { + "epoch": 0.71, + "learning_rate": 0.0002589391168016723, + "loss": 0.2339, + "step": 10980 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025889992160961587, + "loss": 0.3096, + "step": 10990 + }, + { + "epoch": 0.71, + "learning_rate": 0.00025886072641755943, + "loss": 0.3895, + "step": 11000 + }, + { + "epoch": 0.71, + "eval_loss": 0.3404909670352936, + "eval_runtime": 993.7919, + "eval_samples_per_second": 15.685, + "eval_steps_per_second": 1.961, + "eval_wer": 0.32276015977029465, + "step": 11000 + }, + { + "epoch": 0.71, + "learning_rate": 0.000258821531225503, + "loss": 0.2663, + "step": 11010 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025878233603344656, + "loss": 0.2902, + "step": 11020 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002587431408413901, + "loss": 0.251, + "step": 11030 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002587039456493337, + "loss": 0.3552, + "step": 11040 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002586647504572772, + "loss": 0.3618, + "step": 11050 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025862555526522075, + "loss": 0.2469, + "step": 11060 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002585863600731643, + "loss": 0.27, + "step": 11070 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002585471648811079, + "loss": 0.2419, + "step": 11080 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025850796968905144, + "loss": 0.269, + "step": 11090 + }, + { + "epoch": 0.72, + "learning_rate": 0.000258468774496995, + "loss": 0.3411, + "step": 11100 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025842957930493856, + "loss": 0.2476, + "step": 11110 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002583903841128821, + "loss": 0.2771, + "step": 11120 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002583511889208257, + "loss": 0.2884, + "step": 11130 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025831199372876925, + "loss": 0.3018, + "step": 11140 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002582727985367128, + "loss": 0.3534, + "step": 11150 + }, + { + "epoch": 0.72, + "learning_rate": 0.0002582336033446564, + "loss": 0.2725, + "step": 11160 + }, + { + "epoch": 0.72, + "learning_rate": 0.00025819440815259994, + "loss": 0.3233, + "step": 11170 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002581552129605435, + "loss": 0.3047, + "step": 11180 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025811601776848706, + "loss": 0.2927, + "step": 11190 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002580768225764306, + "loss": 0.4566, + "step": 11200 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002580376273843742, + "loss": 0.3009, + "step": 11210 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025799843219231775, + "loss": 0.2853, + "step": 11220 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002579592370002613, + "loss": 0.2706, + "step": 11230 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025792004180820487, + "loss": 0.3086, + "step": 11240 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002578808466161484, + "loss": 0.4638, + "step": 11250 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025784165142409194, + "loss": 0.2319, + "step": 11260 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002578024562320355, + "loss": 0.2595, + "step": 11270 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025776326103997907, + "loss": 0.273, + "step": 11280 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025772406584792263, + "loss": 0.267, + "step": 11290 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002576848706558662, + "loss": 0.4139, + "step": 11300 + }, + { + "epoch": 0.73, + "learning_rate": 0.00025764567546380975, + "loss": 0.2376, + "step": 11310 + }, + { + "epoch": 0.73, + "learning_rate": 0.0002576064802717533, + "loss": 0.2388, + "step": 11320 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002575672850796969, + "loss": 0.3108, + "step": 11330 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025752808988764044, + "loss": 0.3006, + "step": 11340 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025748889469558395, + "loss": 0.3941, + "step": 11350 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002574496995035275, + "loss": 0.2502, + "step": 11360 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002574105043114711, + "loss": 0.2962, + "step": 11370 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025737130911941464, + "loss": 0.2832, + "step": 11380 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002573321139273582, + "loss": 0.2949, + "step": 11390 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025729291873530176, + "loss": 0.4185, + "step": 11400 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002572537235432453, + "loss": 0.244, + "step": 11410 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002572145283511889, + "loss": 0.2848, + "step": 11420 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025717533315913245, + "loss": 0.2988, + "step": 11430 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002571400574862817, + "loss": 0.2924, + "step": 11440 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025710086229422524, + "loss": 0.4397, + "step": 11450 + }, + { + "epoch": 0.74, + "learning_rate": 0.0002570616671021688, + "loss": 0.2984, + "step": 11460 + }, + { + "epoch": 0.74, + "learning_rate": 0.00025702247191011237, + "loss": 0.2416, + "step": 11470 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025698327671805593, + "loss": 0.2628, + "step": 11480 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002569440815259995, + "loss": 0.345, + "step": 11490 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025690488633394305, + "loss": 0.4512, + "step": 11500 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025686569114188656, + "loss": 0.2856, + "step": 11510 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002568264959498301, + "loss": 0.2615, + "step": 11520 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002567873007577737, + "loss": 0.3242, + "step": 11530 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025674810556571725, + "loss": 0.3101, + "step": 11540 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002567089103736608, + "loss": 0.3486, + "step": 11550 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002566697151816044, + "loss": 0.2614, + "step": 11560 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025663051998954794, + "loss": 0.2654, + "step": 11570 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002565913247974915, + "loss": 0.2909, + "step": 11580 + }, + { + "epoch": 0.75, + "learning_rate": 0.000256552129605435, + "loss": 0.3039, + "step": 11590 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025651293441337857, + "loss": 0.3766, + "step": 11600 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025647373922132213, + "loss": 0.2502, + "step": 11610 + }, + { + "epoch": 0.75, + "learning_rate": 0.0002564345440292657, + "loss": 0.2518, + "step": 11620 + }, + { + "epoch": 0.75, + "learning_rate": 0.00025639534883720925, + "loss": 0.2837, + "step": 11630 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002563561536451528, + "loss": 0.314, + "step": 11640 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002563169584530964, + "loss": 0.3593, + "step": 11650 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025627776326103994, + "loss": 0.2378, + "step": 11660 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002562385680689835, + "loss": 0.2722, + "step": 11670 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025619937287692707, + "loss": 0.2671, + "step": 11680 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025616017768487063, + "loss": 0.3482, + "step": 11690 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002561209824928142, + "loss": 0.4184, + "step": 11700 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025608178730075775, + "loss": 0.2464, + "step": 11710 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002560425921087013, + "loss": 0.2786, + "step": 11720 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002560033969166449, + "loss": 0.2874, + "step": 11730 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025596420172458844, + "loss": 0.298, + "step": 11740 + }, + { + "epoch": 0.76, + "learning_rate": 0.000255925006532532, + "loss": 0.3791, + "step": 11750 + }, + { + "epoch": 0.76, + "learning_rate": 0.00025588581134047556, + "loss": 0.2491, + "step": 11760 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002558466161484191, + "loss": 0.2631, + "step": 11770 + }, + { + "epoch": 0.76, + "learning_rate": 0.0002558074209563627, + "loss": 0.2693, + "step": 11780 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002557682257643062, + "loss": 0.3006, + "step": 11790 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025572903057224976, + "loss": 0.4116, + "step": 11800 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002556898353801933, + "loss": 0.2667, + "step": 11810 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002556506401881369, + "loss": 0.2494, + "step": 11820 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025561144499608045, + "loss": 0.2544, + "step": 11830 + }, + { + "epoch": 0.77, + "learning_rate": 0.000255572249804024, + "loss": 0.3052, + "step": 11840 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025553305461196757, + "loss": 0.436, + "step": 11850 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025549385941991113, + "loss": 0.2601, + "step": 11860 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002554546642278547, + "loss": 0.2605, + "step": 11870 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025541546903579826, + "loss": 0.2908, + "step": 11880 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002553762738437418, + "loss": 0.3008, + "step": 11890 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002553370786516854, + "loss": 0.3645, + "step": 11900 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025529788345962894, + "loss": 0.2615, + "step": 11910 + }, + { + "epoch": 0.77, + "learning_rate": 0.0002552586882675725, + "loss": 0.254, + "step": 11920 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025521949307551607, + "loss": 0.2654, + "step": 11930 + }, + { + "epoch": 0.77, + "learning_rate": 0.00025518029788345963, + "loss": 0.2841, + "step": 11940 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002551411026914032, + "loss": 0.4529, + "step": 11950 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025510190749934676, + "loss": 0.2621, + "step": 11960 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002550627123072903, + "loss": 0.2518, + "step": 11970 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002550235171152339, + "loss": 0.2883, + "step": 11980 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002549843219231774, + "loss": 0.3152, + "step": 11990 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025494512673112095, + "loss": 0.3895, + "step": 12000 + }, + { + "epoch": 0.78, + "eval_loss": 0.3215894103050232, + "eval_runtime": 1021.0447, + "eval_samples_per_second": 15.267, + "eval_steps_per_second": 1.909, + "eval_wer": 0.30797003596253075, + "step": 12000 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002549059315390645, + "loss": 0.2857, + "step": 12010 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002548667363470081, + "loss": 0.2716, + "step": 12020 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025482754115495164, + "loss": 0.2737, + "step": 12030 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002547883459628952, + "loss": 0.2821, + "step": 12040 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025474915077083876, + "loss": 0.3752, + "step": 12050 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002547099555787823, + "loss": 0.2454, + "step": 12060 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025467076038672583, + "loss": 0.256, + "step": 12070 + }, + { + "epoch": 0.78, + "learning_rate": 0.0002546315651946694, + "loss": 0.2725, + "step": 12080 + }, + { + "epoch": 0.78, + "learning_rate": 0.00025459237000261296, + "loss": 0.268, + "step": 12090 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002545531748105565, + "loss": 0.3814, + "step": 12100 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002545139796185001, + "loss": 0.2296, + "step": 12110 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025447478442644364, + "loss": 0.2404, + "step": 12120 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002544355892343872, + "loss": 0.2378, + "step": 12130 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025439639404233077, + "loss": 0.2784, + "step": 12140 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025435719885027433, + "loss": 0.388, + "step": 12150 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002543180036582179, + "loss": 0.2642, + "step": 12160 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025427880846616146, + "loss": 0.2209, + "step": 12170 + }, + { + "epoch": 0.79, + "learning_rate": 0.000254239613274105, + "loss": 0.3195, + "step": 12180 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002542004180820486, + "loss": 0.31, + "step": 12190 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025416122288999214, + "loss": 0.3943, + "step": 12200 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002541220276979357, + "loss": 0.2287, + "step": 12210 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025408283250587927, + "loss": 0.2654, + "step": 12220 + }, + { + "epoch": 0.79, + "learning_rate": 0.00025404363731382283, + "loss": 0.3163, + "step": 12230 + }, + { + "epoch": 0.79, + "learning_rate": 0.0002540044421217664, + "loss": 0.2942, + "step": 12240 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025396524692970995, + "loss": 0.3781, + "step": 12250 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002539260517376535, + "loss": 0.2531, + "step": 12260 + }, + { + "epoch": 0.8, + "learning_rate": 0.000253886856545597, + "loss": 0.2691, + "step": 12270 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002538476613535406, + "loss": 0.2862, + "step": 12280 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025380846616148415, + "loss": 0.2761, + "step": 12290 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002537692709694277, + "loss": 0.3447, + "step": 12300 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002537300757773713, + "loss": 0.2549, + "step": 12310 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025369088058531484, + "loss": 0.2517, + "step": 12320 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002536516853932584, + "loss": 0.2428, + "step": 12330 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025361249020120196, + "loss": 0.2809, + "step": 12340 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002535732950091455, + "loss": 0.3849, + "step": 12350 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002535340998170891, + "loss": 0.2966, + "step": 12360 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025349490462503265, + "loss": 0.291, + "step": 12370 + }, + { + "epoch": 0.8, + "learning_rate": 0.0002534557094329762, + "loss": 0.2842, + "step": 12380 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025341651424091977, + "loss": 0.3347, + "step": 12390 + }, + { + "epoch": 0.8, + "learning_rate": 0.00025337731904886333, + "loss": 0.3897, + "step": 12400 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002533381238568069, + "loss": 0.2625, + "step": 12410 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025329892866475046, + "loss": 0.246, + "step": 12420 + }, + { + "epoch": 0.81, + "learning_rate": 0.000253259733472694, + "loss": 0.2807, + "step": 12430 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002532205382806376, + "loss": 0.3225, + "step": 12440 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025318134308858115, + "loss": 0.4033, + "step": 12450 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002531421478965247, + "loss": 0.2184, + "step": 12460 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002531029527044682, + "loss": 0.2575, + "step": 12470 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002530637575124118, + "loss": 0.2958, + "step": 12480 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025302456232035534, + "loss": 0.2804, + "step": 12490 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002529853671282989, + "loss": 0.3656, + "step": 12500 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025294617193624247, + "loss": 0.246, + "step": 12510 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025290697674418603, + "loss": 0.2706, + "step": 12520 + }, + { + "epoch": 0.81, + "learning_rate": 0.0002528677815521296, + "loss": 0.2709, + "step": 12530 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025282858636007315, + "loss": 0.33, + "step": 12540 + }, + { + "epoch": 0.81, + "learning_rate": 0.00025278939116801666, + "loss": 0.3679, + "step": 12550 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002527501959759602, + "loss": 0.2721, + "step": 12560 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002527110007839038, + "loss": 0.2631, + "step": 12570 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025267180559184735, + "loss": 0.2804, + "step": 12580 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002526326103997909, + "loss": 0.2485, + "step": 12590 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025259341520773447, + "loss": 0.3461, + "step": 12600 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025255422001567803, + "loss": 0.2855, + "step": 12610 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002525150248236216, + "loss": 0.2686, + "step": 12620 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025247582963156516, + "loss": 0.2798, + "step": 12630 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002524366344395087, + "loss": 0.2986, + "step": 12640 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002523974392474523, + "loss": 0.3471, + "step": 12650 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025235824405539585, + "loss": 0.2054, + "step": 12660 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002523190488633394, + "loss": 0.2169, + "step": 12670 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025227985367128297, + "loss": 0.2785, + "step": 12680 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025224065847922653, + "loss": 0.3191, + "step": 12690 + }, + { + "epoch": 0.82, + "learning_rate": 0.0002522014632871701, + "loss": 0.4082, + "step": 12700 + }, + { + "epoch": 0.82, + "learning_rate": 0.00025216226809511366, + "loss": 0.2569, + "step": 12710 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002521230729030572, + "loss": 0.2615, + "step": 12720 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002520838777110008, + "loss": 0.2534, + "step": 12730 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025204468251894434, + "loss": 0.2709, + "step": 12740 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025200548732688785, + "loss": 0.3772, + "step": 12750 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002519662921348314, + "loss": 0.28, + "step": 12760 + }, + { + "epoch": 0.83, + "learning_rate": 0.000251927096942775, + "loss": 0.2648, + "step": 12770 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025188790175071854, + "loss": 0.2619, + "step": 12780 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002518487065586621, + "loss": 0.2805, + "step": 12790 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025180951136660566, + "loss": 0.4441, + "step": 12800 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002517703161745492, + "loss": 0.2578, + "step": 12810 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002517311209824928, + "loss": 0.2729, + "step": 12820 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025169192579043635, + "loss": 0.2787, + "step": 12830 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002516527305983799, + "loss": 0.2901, + "step": 12840 + }, + { + "epoch": 0.83, + "learning_rate": 0.0002516135354063235, + "loss": 0.3631, + "step": 12850 + }, + { + "epoch": 0.83, + "learning_rate": 0.00025157434021426704, + "loss": 0.2337, + "step": 12860 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002515351450222106, + "loss": 0.2246, + "step": 12870 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025149594983015416, + "loss": 0.2593, + "step": 12880 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002514567546380977, + "loss": 0.2961, + "step": 12890 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002514175594460413, + "loss": 0.3793, + "step": 12900 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025137836425398485, + "loss": 0.2137, + "step": 12910 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002513391690619284, + "loss": 0.2407, + "step": 12920 + }, + { + "epoch": 0.84, + "learning_rate": 0.000251299973869872, + "loss": 0.2812, + "step": 12930 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025126077867781554, + "loss": 0.2966, + "step": 12940 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025122158348575904, + "loss": 0.4045, + "step": 12950 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002511823882937026, + "loss": 0.2458, + "step": 12960 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025114319310164617, + "loss": 0.2343, + "step": 12970 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025110399790958973, + "loss": 0.2808, + "step": 12980 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002510648027175333, + "loss": 0.3116, + "step": 12990 + }, + { + "epoch": 0.84, + "learning_rate": 0.00025102560752547686, + "loss": 0.3489, + "step": 13000 + }, + { + "epoch": 0.84, + "eval_loss": 0.32267671823501587, + "eval_runtime": 1745.039, + "eval_samples_per_second": 8.933, + "eval_steps_per_second": 1.117, + "eval_wer": 0.3146223934272434, + "step": 13000 + }, + { + "epoch": 0.84, + "learning_rate": 0.0002509864123334204, + "loss": 0.2199, + "step": 13010 + }, + { + "epoch": 0.85, + "learning_rate": 0.000250947217141364, + "loss": 0.238, + "step": 13020 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002509080219493075, + "loss": 0.2551, + "step": 13030 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025086882675725105, + "loss": 0.2873, + "step": 13040 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002508296315651946, + "loss": 0.4606, + "step": 13050 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002507904363731382, + "loss": 0.2478, + "step": 13060 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025075124118108174, + "loss": 0.2485, + "step": 13070 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002507120459890253, + "loss": 0.22, + "step": 13080 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025067285079696886, + "loss": 0.2804, + "step": 13090 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002506336556049124, + "loss": 0.4432, + "step": 13100 + }, + { + "epoch": 0.85, + "learning_rate": 0.000250594460412856, + "loss": 0.2452, + "step": 13110 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025055526522079955, + "loss": 0.2498, + "step": 13120 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002505160700287431, + "loss": 0.2839, + "step": 13130 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002504768748366867, + "loss": 0.2925, + "step": 13140 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025043767964463024, + "loss": 0.3378, + "step": 13150 + }, + { + "epoch": 0.85, + "learning_rate": 0.0002503984844525738, + "loss": 0.2342, + "step": 13160 + }, + { + "epoch": 0.85, + "learning_rate": 0.00025035928926051736, + "loss": 0.2457, + "step": 13170 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002503200940684609, + "loss": 0.2944, + "step": 13180 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002502808988764045, + "loss": 0.3253, + "step": 13190 + }, + { + "epoch": 0.86, + "learning_rate": 0.00025024170368434805, + "loss": 0.374, + "step": 13200 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002502025084922916, + "loss": 0.21, + "step": 13210 + }, + { + "epoch": 0.86, + "learning_rate": 0.00025016331330023517, + "loss": 0.2301, + "step": 13220 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002501241181081787, + "loss": 0.289, + "step": 13230 + }, + { + "epoch": 0.86, + "learning_rate": 0.00025008492291612224, + "loss": 0.3392, + "step": 13240 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002500457277240658, + "loss": 0.4321, + "step": 13250 + }, + { + "epoch": 0.86, + "learning_rate": 0.00025000653253200937, + "loss": 0.2427, + "step": 13260 + }, + { + "epoch": 0.86, + "learning_rate": 0.00024996733733995293, + "loss": 0.2401, + "step": 13270 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002499281421478965, + "loss": 0.2779, + "step": 13280 + }, + { + "epoch": 0.86, + "learning_rate": 0.00024988894695584005, + "loss": 0.2548, + "step": 13290 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002498497517637836, + "loss": 0.3788, + "step": 13300 + }, + { + "epoch": 0.86, + "learning_rate": 0.0002498105565717272, + "loss": 0.2253, + "step": 13310 + }, + { + "epoch": 0.86, + "learning_rate": 0.00024977136137967074, + "loss": 0.2432, + "step": 13320 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002497321661876143, + "loss": 0.253, + "step": 13330 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024969297099555786, + "loss": 0.2885, + "step": 13340 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002496537758035014, + "loss": 0.418, + "step": 13350 + }, + { + "epoch": 0.87, + "learning_rate": 0.000249614580611445, + "loss": 0.2175, + "step": 13360 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024957538541938855, + "loss": 0.2743, + "step": 13370 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002495361902273321, + "loss": 0.2602, + "step": 13380 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002494969950352757, + "loss": 0.3103, + "step": 13390 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024945779984321924, + "loss": 0.3663, + "step": 13400 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002494186046511628, + "loss": 0.2586, + "step": 13410 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024937940945910636, + "loss": 0.266, + "step": 13420 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024934021426704987, + "loss": 0.2812, + "step": 13430 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024930101907499343, + "loss": 0.2833, + "step": 13440 + }, + { + "epoch": 0.87, + "learning_rate": 0.000249261823882937, + "loss": 0.4003, + "step": 13450 + }, + { + "epoch": 0.87, + "learning_rate": 0.00024922262869088056, + "loss": 0.2619, + "step": 13460 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002491834334988241, + "loss": 0.3041, + "step": 13470 + }, + { + "epoch": 0.87, + "learning_rate": 0.0002491442383067677, + "loss": 0.2629, + "step": 13480 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024910504311471124, + "loss": 0.2995, + "step": 13490 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002490658479226548, + "loss": 0.3956, + "step": 13500 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002490266527305983, + "loss": 0.2473, + "step": 13510 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002489874575385419, + "loss": 0.2752, + "step": 13520 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024894826234648544, + "loss": 0.2812, + "step": 13530 + }, + { + "epoch": 0.88, + "learning_rate": 0.000248909067154429, + "loss": 0.2634, + "step": 13540 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024886987196237256, + "loss": 0.3654, + "step": 13550 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002488306767703161, + "loss": 0.2415, + "step": 13560 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002487914815782597, + "loss": 0.2849, + "step": 13570 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024875228638620325, + "loss": 0.2402, + "step": 13580 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002487130911941468, + "loss": 0.3148, + "step": 13590 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002486738960020904, + "loss": 0.3746, + "step": 13600 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024863470081003394, + "loss": 0.2103, + "step": 13610 + }, + { + "epoch": 0.88, + "learning_rate": 0.0002485955056179775, + "loss": 0.2347, + "step": 13620 + }, + { + "epoch": 0.88, + "learning_rate": 0.00024855631042592106, + "loss": 0.2856, + "step": 13630 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002485171152338646, + "loss": 0.311, + "step": 13640 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002484779200418082, + "loss": 0.3633, + "step": 13650 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024843872484975175, + "loss": 0.2323, + "step": 13660 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002483995296576953, + "loss": 0.2715, + "step": 13670 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002483603344656389, + "loss": 0.2512, + "step": 13680 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024832113927358244, + "loss": 0.3491, + "step": 13690 + }, + { + "epoch": 0.89, + "learning_rate": 0.000248281944081526, + "loss": 0.4054, + "step": 13700 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002482427488894695, + "loss": 0.267, + "step": 13710 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024820355369741307, + "loss": 0.2757, + "step": 13720 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024816435850535663, + "loss": 0.2803, + "step": 13730 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002481251633133002, + "loss": 0.3043, + "step": 13740 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024808596812124376, + "loss": 0.378, + "step": 13750 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002480467729291873, + "loss": 0.2689, + "step": 13760 + }, + { + "epoch": 0.89, + "learning_rate": 0.0002480075777371309, + "loss": 0.2369, + "step": 13770 + }, + { + "epoch": 0.89, + "learning_rate": 0.00024796838254507444, + "loss": 0.2635, + "step": 13780 + }, + { + "epoch": 0.89, + "learning_rate": 0.000247929187353018, + "loss": 0.2874, + "step": 13790 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024788999216096157, + "loss": 0.4193, + "step": 13800 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024785079696890513, + "loss": 0.2542, + "step": 13810 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002478116017768487, + "loss": 0.278, + "step": 13820 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024777240658479225, + "loss": 0.2765, + "step": 13830 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002477332113927358, + "loss": 0.3112, + "step": 13840 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002476940162006794, + "loss": 0.3636, + "step": 13850 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024765482100862294, + "loss": 0.2846, + "step": 13860 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002476156258165665, + "loss": 0.2627, + "step": 13870 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024757643062451007, + "loss": 0.2467, + "step": 13880 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024753723543245363, + "loss": 0.3016, + "step": 13890 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002474980402403972, + "loss": 0.376, + "step": 13900 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024745884504834075, + "loss": 0.2092, + "step": 13910 + }, + { + "epoch": 0.9, + "learning_rate": 0.00024741964985628426, + "loss": 0.2548, + "step": 13920 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002473804546642278, + "loss": 0.2729, + "step": 13930 + }, + { + "epoch": 0.9, + "learning_rate": 0.0002473412594721714, + "loss": 0.3015, + "step": 13940 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024730206428011495, + "loss": 0.3442, + "step": 13950 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002472628690880585, + "loss": 0.2209, + "step": 13960 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024722367389600207, + "loss": 0.2684, + "step": 13970 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024718447870394563, + "loss": 0.2687, + "step": 13980 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002471452835118892, + "loss": 0.2885, + "step": 13990 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002471060883198327, + "loss": 0.4014, + "step": 14000 + }, + { + "epoch": 0.91, + "eval_loss": 0.3171861171722412, + "eval_runtime": 1598.5993, + "eval_samples_per_second": 9.751, + "eval_steps_per_second": 1.219, + "eval_wer": 0.3005216699122969, + "step": 14000 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024706689312777627, + "loss": 0.2345, + "step": 14010 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024702769793571983, + "loss": 0.2487, + "step": 14020 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002469885027436634, + "loss": 0.3068, + "step": 14030 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024694930755160695, + "loss": 0.3007, + "step": 14040 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002469101123595505, + "loss": 0.3492, + "step": 14050 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002468709171674941, + "loss": 0.2287, + "step": 14060 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024683172197543764, + "loss": 0.2142, + "step": 14070 + }, + { + "epoch": 0.91, + "learning_rate": 0.0002467925267833812, + "loss": 0.2944, + "step": 14080 + }, + { + "epoch": 0.91, + "learning_rate": 0.00024675333159132477, + "loss": 0.28, + "step": 14090 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024671413639926833, + "loss": 0.3794, + "step": 14100 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002466749412072119, + "loss": 0.2203, + "step": 14110 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024663574601515545, + "loss": 0.2496, + "step": 14120 + }, + { + "epoch": 0.92, + "learning_rate": 0.000246596550823099, + "loss": 0.2852, + "step": 14130 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002465573556310426, + "loss": 0.3077, + "step": 14140 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024651816043898614, + "loss": 0.3577, + "step": 14150 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002464789652469297, + "loss": 0.2553, + "step": 14160 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024643977005487326, + "loss": 0.2424, + "step": 14170 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002464005748628168, + "loss": 0.2376, + "step": 14180 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002463613796707604, + "loss": 0.2722, + "step": 14190 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002463221844787039, + "loss": 0.3894, + "step": 14200 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024628298928664746, + "loss": 0.2381, + "step": 14210 + }, + { + "epoch": 0.92, + "learning_rate": 0.000246243794094591, + "loss": 0.2459, + "step": 14220 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002462045989025346, + "loss": 0.2658, + "step": 14230 + }, + { + "epoch": 0.92, + "learning_rate": 0.00024616540371047815, + "loss": 0.3087, + "step": 14240 + }, + { + "epoch": 0.92, + "learning_rate": 0.0002461262085184217, + "loss": 0.4246, + "step": 14250 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024608701332636527, + "loss": 0.2369, + "step": 14260 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024604781813430883, + "loss": 0.2499, + "step": 14270 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002460086229422524, + "loss": 0.2647, + "step": 14280 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024596942775019596, + "loss": 0.3165, + "step": 14290 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002459302325581395, + "loss": 0.4206, + "step": 14300 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002458910373660831, + "loss": 0.2634, + "step": 14310 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024585184217402664, + "loss": 0.2731, + "step": 14320 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002458126469819702, + "loss": 0.2546, + "step": 14330 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024577345178991377, + "loss": 0.3088, + "step": 14340 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024573425659785733, + "loss": 0.4138, + "step": 14350 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002456950614058009, + "loss": 0.2226, + "step": 14360 + }, + { + "epoch": 0.93, + "learning_rate": 0.00024565586621374446, + "loss": 0.2573, + "step": 14370 + }, + { + "epoch": 0.93, + "learning_rate": 0.000245616671021688, + "loss": 0.2884, + "step": 14380 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002455774758296316, + "loss": 0.2797, + "step": 14390 + }, + { + "epoch": 0.93, + "learning_rate": 0.0002455382806375751, + "loss": 0.3238, + "step": 14400 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024549908544551865, + "loss": 0.2407, + "step": 14410 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002454598902534622, + "loss": 0.2329, + "step": 14420 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002454206950614058, + "loss": 0.2485, + "step": 14430 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024538149986934934, + "loss": 0.2775, + "step": 14440 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002453423046772929, + "loss": 0.4112, + "step": 14450 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024530310948523646, + "loss": 0.2295, + "step": 14460 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024526391429318, + "loss": 0.2387, + "step": 14470 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002452247191011236, + "loss": 0.2614, + "step": 14480 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002451855239090671, + "loss": 0.3118, + "step": 14490 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024514632871701066, + "loss": 0.353, + "step": 14500 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002451071335249542, + "loss": 0.2524, + "step": 14510 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002450679383328978, + "loss": 0.2473, + "step": 14520 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024502874314084134, + "loss": 0.2361, + "step": 14530 + }, + { + "epoch": 0.94, + "learning_rate": 0.0002449895479487849, + "loss": 0.2807, + "step": 14540 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024495035275672847, + "loss": 0.433, + "step": 14550 + }, + { + "epoch": 0.94, + "learning_rate": 0.00024491115756467203, + "loss": 0.2164, + "step": 14560 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002448719623726156, + "loss": 0.236, + "step": 14570 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024483276718055916, + "loss": 0.2785, + "step": 14580 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002447935719885027, + "loss": 0.2813, + "step": 14590 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002447543767964463, + "loss": 0.3798, + "step": 14600 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024471518160438984, + "loss": 0.2532, + "step": 14610 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002446759864123334, + "loss": 0.244, + "step": 14620 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024463679122027697, + "loss": 0.2398, + "step": 14630 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024459759602822053, + "loss": 0.3104, + "step": 14640 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002445584008361641, + "loss": 0.3818, + "step": 14650 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024451920564410765, + "loss": 0.2339, + "step": 14660 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002444800104520512, + "loss": 0.2529, + "step": 14670 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002444408152599947, + "loss": 0.2447, + "step": 14680 + }, + { + "epoch": 0.95, + "learning_rate": 0.00024440553958714396, + "loss": 0.2917, + "step": 14690 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002443663443950875, + "loss": 0.4212, + "step": 14700 + }, + { + "epoch": 0.95, + "learning_rate": 0.0002443271492030311, + "loss": 0.2153, + "step": 14710 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024428795401097464, + "loss": 0.2304, + "step": 14720 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024424875881891815, + "loss": 0.2604, + "step": 14730 + }, + { + "epoch": 0.96, + "learning_rate": 0.0002442095636268617, + "loss": 0.3425, + "step": 14740 + }, + { + "epoch": 0.96, + "learning_rate": 0.0002441703684348053, + "loss": 0.3822, + "step": 14750 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024413117324274887, + "loss": 0.2724, + "step": 14760 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024409197805069243, + "loss": 0.3051, + "step": 14770 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024405278285863596, + "loss": 0.2708, + "step": 14780 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024401358766657953, + "loss": 0.2776, + "step": 14790 + }, + { + "epoch": 0.96, + "learning_rate": 0.0002439743924745231, + "loss": 0.4292, + "step": 14800 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024393519728246665, + "loss": 0.1806, + "step": 14810 + }, + { + "epoch": 0.96, + "learning_rate": 0.0002438960020904102, + "loss": 0.2631, + "step": 14820 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024385680689835377, + "loss": 0.307, + "step": 14830 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024381761170629734, + "loss": 0.3048, + "step": 14840 + }, + { + "epoch": 0.96, + "learning_rate": 0.0002437784165142409, + "loss": 0.4102, + "step": 14850 + }, + { + "epoch": 0.96, + "learning_rate": 0.00024373922132218446, + "loss": 0.2596, + "step": 14860 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024370002613012802, + "loss": 0.2507, + "step": 14870 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024366083093807156, + "loss": 0.2685, + "step": 14880 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024362163574601512, + "loss": 0.2964, + "step": 14890 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024358244055395868, + "loss": 0.4351, + "step": 14900 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024354324536190225, + "loss": 0.244, + "step": 14910 + }, + { + "epoch": 0.97, + "learning_rate": 0.0002435040501698458, + "loss": 0.2913, + "step": 14920 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024346485497778937, + "loss": 0.2557, + "step": 14930 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024342565978573293, + "loss": 0.2635, + "step": 14940 + }, + { + "epoch": 0.97, + "learning_rate": 0.0002433864645936765, + "loss": 0.4106, + "step": 14950 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024334726940162006, + "loss": 0.2413, + "step": 14960 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024330807420956362, + "loss": 0.2444, + "step": 14970 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024326887901750715, + "loss": 0.2438, + "step": 14980 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024322968382545072, + "loss": 0.2971, + "step": 14990 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024319048863339428, + "loss": 0.352, + "step": 15000 + }, + { + "epoch": 0.97, + "eval_loss": 0.30451446771621704, + "eval_runtime": 1646.353, + "eval_samples_per_second": 9.468, + "eval_steps_per_second": 1.184, + "eval_wer": 0.29593751332603657, + "step": 15000 + }, + { + "epoch": 0.97, + "learning_rate": 0.00024315129344133784, + "loss": 0.2196, + "step": 15010 + }, + { + "epoch": 0.97, + "learning_rate": 0.0002431120982492814, + "loss": 0.27, + "step": 15020 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024307290305722497, + "loss": 0.313, + "step": 15030 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024303370786516853, + "loss": 0.2978, + "step": 15040 + }, + { + "epoch": 0.98, + "learning_rate": 0.0002429945126731121, + "loss": 0.3856, + "step": 15050 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024295531748105565, + "loss": 0.2006, + "step": 15060 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024291612228899922, + "loss": 0.2182, + "step": 15070 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024287692709694275, + "loss": 0.2793, + "step": 15080 + }, + { + "epoch": 0.98, + "learning_rate": 0.0002428377319048863, + "loss": 0.2352, + "step": 15090 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024279853671282988, + "loss": 0.372, + "step": 15100 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024275934152077344, + "loss": 0.233, + "step": 15110 + }, + { + "epoch": 0.98, + "learning_rate": 0.000242720146328717, + "loss": 0.2496, + "step": 15120 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024268095113666056, + "loss": 0.263, + "step": 15130 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024264175594460412, + "loss": 0.3158, + "step": 15140 + }, + { + "epoch": 0.98, + "learning_rate": 0.0002426025607525477, + "loss": 0.3333, + "step": 15150 + }, + { + "epoch": 0.98, + "learning_rate": 0.00024256336556049125, + "loss": 0.2481, + "step": 15160 + }, + { + "epoch": 0.98, + "learning_rate": 0.0002425241703684348, + "loss": 0.255, + "step": 15170 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024248497517637832, + "loss": 0.2815, + "step": 15180 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024244577998432188, + "loss": 0.2698, + "step": 15190 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024240658479226544, + "loss": 0.3972, + "step": 15200 + }, + { + "epoch": 0.99, + "learning_rate": 0.000242367389600209, + "loss": 0.2455, + "step": 15210 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024232819440815257, + "loss": 0.236, + "step": 15220 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024228899921609613, + "loss": 0.2088, + "step": 15230 + }, + { + "epoch": 0.99, + "learning_rate": 0.0002422498040240397, + "loss": 0.2838, + "step": 15240 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024221060883198326, + "loss": 0.3116, + "step": 15250 + }, + { + "epoch": 0.99, + "learning_rate": 0.0002421714136399268, + "loss": 0.2534, + "step": 15260 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024213221844787035, + "loss": 0.2511, + "step": 15270 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024209302325581392, + "loss": 0.2316, + "step": 15280 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024205382806375748, + "loss": 0.2509, + "step": 15290 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024201463287170104, + "loss": 0.3988, + "step": 15300 + }, + { + "epoch": 0.99, + "learning_rate": 0.0002419754376796446, + "loss": 0.2599, + "step": 15310 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024193624248758816, + "loss": 0.262, + "step": 15320 + }, + { + "epoch": 0.99, + "learning_rate": 0.00024189704729553173, + "loss": 0.2448, + "step": 15330 + }, + { + "epoch": 1.0, + "learning_rate": 0.0002418578521034753, + "loss": 0.313, + "step": 15340 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024181865691141885, + "loss": 0.3712, + "step": 15350 + }, + { + "epoch": 1.0, + "learning_rate": 0.0002417794617193624, + "loss": 0.2158, + "step": 15360 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024174026652730595, + "loss": 0.2164, + "step": 15370 + }, + { + "epoch": 1.0, + "learning_rate": 0.0002417010713352495, + "loss": 0.2946, + "step": 15380 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024166187614319307, + "loss": 0.3418, + "step": 15390 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024162268095113664, + "loss": 0.3332, + "step": 15400 + }, + { + "epoch": 1.0, + "learning_rate": 0.0002415834857590802, + "loss": 0.2904, + "step": 15410 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024154429056702376, + "loss": 0.2405, + "step": 15420 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024150509537496732, + "loss": 0.2485, + "step": 15430 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024146590018291088, + "loss": 0.2396, + "step": 15440 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024142670499085445, + "loss": 0.2802, + "step": 15450 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024138750979879798, + "loss": 0.2864, + "step": 15460 + }, + { + "epoch": 1.0, + "learning_rate": 0.00024134831460674154, + "loss": 0.2348, + "step": 15470 + }, + { + "epoch": 1.0, + "learning_rate": 0.0002413091194146851, + "loss": 0.2276, + "step": 15480 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024126992422262867, + "loss": 0.2438, + "step": 15490 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024123072903057223, + "loss": 0.272, + "step": 15500 + }, + { + "epoch": 1.01, + "learning_rate": 0.0002411915338385158, + "loss": 0.2917, + "step": 15510 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024115233864645936, + "loss": 0.1826, + "step": 15520 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024111314345440292, + "loss": 0.2278, + "step": 15530 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024107394826234648, + "loss": 0.2831, + "step": 15540 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024103475307029004, + "loss": 0.2698, + "step": 15550 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024099555787823358, + "loss": 0.2932, + "step": 15560 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024095636268617714, + "loss": 0.2291, + "step": 15570 + }, + { + "epoch": 1.01, + "learning_rate": 0.0002409171674941207, + "loss": 0.2211, + "step": 15580 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024087797230206426, + "loss": 0.2325, + "step": 15590 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024083877711000783, + "loss": 0.2858, + "step": 15600 + }, + { + "epoch": 1.01, + "learning_rate": 0.0002407995819179514, + "loss": 0.3039, + "step": 15610 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024076038672589495, + "loss": 0.2313, + "step": 15620 + }, + { + "epoch": 1.01, + "learning_rate": 0.00024072119153383851, + "loss": 0.2353, + "step": 15630 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024068199634178208, + "loss": 0.2752, + "step": 15640 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024064280114972564, + "loss": 0.241, + "step": 15650 + }, + { + "epoch": 1.02, + "learning_rate": 0.0002406036059576692, + "loss": 0.3985, + "step": 15660 + }, + { + "epoch": 1.02, + "learning_rate": 0.0002405644107656127, + "loss": 0.2348, + "step": 15670 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024052521557355627, + "loss": 0.2162, + "step": 15680 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024048602038149983, + "loss": 0.2375, + "step": 15690 + }, + { + "epoch": 1.02, + "learning_rate": 0.0002404468251894434, + "loss": 0.2803, + "step": 15700 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024040762999738696, + "loss": 0.3514, + "step": 15710 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024036843480533052, + "loss": 0.1983, + "step": 15720 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024032923961327408, + "loss": 0.2046, + "step": 15730 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024029004442121762, + "loss": 0.2646, + "step": 15740 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024025084922916118, + "loss": 0.2915, + "step": 15750 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024021165403710474, + "loss": 0.3759, + "step": 15760 + }, + { + "epoch": 1.02, + "learning_rate": 0.0002401724588450483, + "loss": 0.2307, + "step": 15770 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024013326365299187, + "loss": 0.23, + "step": 15780 + }, + { + "epoch": 1.02, + "learning_rate": 0.00024009406846093543, + "loss": 0.2783, + "step": 15790 + }, + { + "epoch": 1.03, + "learning_rate": 0.000240054873268879, + "loss": 0.2907, + "step": 15800 + }, + { + "epoch": 1.03, + "learning_rate": 0.00024001567807682255, + "loss": 0.3324, + "step": 15810 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023997648288476612, + "loss": 0.2368, + "step": 15820 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023993728769270968, + "loss": 0.2284, + "step": 15830 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023989809250065321, + "loss": 0.2174, + "step": 15840 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023985889730859678, + "loss": 0.2491, + "step": 15850 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023981970211654034, + "loss": 0.3383, + "step": 15860 + }, + { + "epoch": 1.03, + "learning_rate": 0.0002397805069244839, + "loss": 0.2133, + "step": 15870 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023974131173242746, + "loss": 0.236, + "step": 15880 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023970211654037103, + "loss": 0.2452, + "step": 15890 + }, + { + "epoch": 1.03, + "learning_rate": 0.0002396629213483146, + "loss": 0.277, + "step": 15900 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023962372615625815, + "loss": 0.3237, + "step": 15910 + }, + { + "epoch": 1.03, + "learning_rate": 0.0002395845309642017, + "loss": 0.2305, + "step": 15920 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023954533577214527, + "loss": 0.2135, + "step": 15930 + }, + { + "epoch": 1.03, + "learning_rate": 0.00023950614058008884, + "loss": 0.2653, + "step": 15940 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023946694538803237, + "loss": 0.3012, + "step": 15950 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023942775019597593, + "loss": 0.3372, + "step": 15960 + }, + { + "epoch": 1.04, + "learning_rate": 0.0002393885550039195, + "loss": 0.1852, + "step": 15970 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023934935981186306, + "loss": 0.2049, + "step": 15980 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023931016461980662, + "loss": 0.2325, + "step": 15990 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023927096942775018, + "loss": 0.3059, + "step": 16000 + }, + { + "epoch": 1.04, + "eval_loss": 0.29607099294662476, + "eval_runtime": 1669.7329, + "eval_samples_per_second": 9.336, + "eval_steps_per_second": 1.167, + "eval_wer": 0.2955821523503575, + "step": 16000 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023923177423569375, + "loss": 0.3311, + "step": 16010 + }, + { + "epoch": 1.04, + "learning_rate": 0.0002391925790436373, + "loss": 0.2124, + "step": 16020 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023915338385158087, + "loss": 0.1921, + "step": 16030 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023911418865952443, + "loss": 0.2504, + "step": 16040 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023907499346746797, + "loss": 0.2898, + "step": 16050 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023903579827541153, + "loss": 0.3481, + "step": 16060 + }, + { + "epoch": 1.04, + "learning_rate": 0.0002389966030833551, + "loss": 0.2737, + "step": 16070 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023895740789129865, + "loss": 0.1881, + "step": 16080 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023891821269924222, + "loss": 0.2458, + "step": 16090 + }, + { + "epoch": 1.04, + "learning_rate": 0.00023887901750718578, + "loss": 0.2623, + "step": 16100 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023883982231512934, + "loss": 0.3684, + "step": 16110 + }, + { + "epoch": 1.05, + "learning_rate": 0.0002388006271230729, + "loss": 0.2393, + "step": 16120 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023876143193101647, + "loss": 0.2438, + "step": 16130 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023872223673896003, + "loss": 0.2567, + "step": 16140 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023868304154690356, + "loss": 0.278, + "step": 16150 + }, + { + "epoch": 1.05, + "learning_rate": 0.0002386438463548471, + "loss": 0.3514, + "step": 16160 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023860465116279066, + "loss": 0.2224, + "step": 16170 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023856545597073422, + "loss": 0.2299, + "step": 16180 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023852626077867779, + "loss": 0.2358, + "step": 16190 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023848706558662135, + "loss": 0.2787, + "step": 16200 + }, + { + "epoch": 1.05, + "learning_rate": 0.0002384478703945649, + "loss": 0.4017, + "step": 16210 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023840867520250847, + "loss": 0.2511, + "step": 16220 + }, + { + "epoch": 1.05, + "learning_rate": 0.000238369480010452, + "loss": 0.1996, + "step": 16230 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023833028481839557, + "loss": 0.2592, + "step": 16240 + }, + { + "epoch": 1.05, + "learning_rate": 0.00023829108962633913, + "loss": 0.2329, + "step": 16250 + }, + { + "epoch": 1.06, + "learning_rate": 0.0002382518944342827, + "loss": 0.3644, + "step": 16260 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023821269924222626, + "loss": 0.2221, + "step": 16270 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023817350405016982, + "loss": 0.2525, + "step": 16280 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023813430885811338, + "loss": 0.246, + "step": 16290 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023809511366605694, + "loss": 0.2571, + "step": 16300 + }, + { + "epoch": 1.06, + "learning_rate": 0.0002380559184740005, + "loss": 0.3107, + "step": 16310 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023801672328194407, + "loss": 0.2095, + "step": 16320 + }, + { + "epoch": 1.06, + "learning_rate": 0.0002379775280898876, + "loss": 0.1932, + "step": 16330 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023793833289783117, + "loss": 0.2546, + "step": 16340 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023789913770577473, + "loss": 0.2471, + "step": 16350 + }, + { + "epoch": 1.06, + "learning_rate": 0.0002378599425137183, + "loss": 0.3276, + "step": 16360 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023782074732166185, + "loss": 0.195, + "step": 16370 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023778155212960541, + "loss": 0.2253, + "step": 16380 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023774235693754898, + "loss": 0.2184, + "step": 16390 + }, + { + "epoch": 1.06, + "learning_rate": 0.00023770316174549254, + "loss": 0.2889, + "step": 16400 + }, + { + "epoch": 1.07, + "learning_rate": 0.0002376639665534361, + "loss": 0.3341, + "step": 16410 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023762477136137966, + "loss": 0.2159, + "step": 16420 + }, + { + "epoch": 1.07, + "learning_rate": 0.0002375855761693232, + "loss": 0.2084, + "step": 16430 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023754638097726676, + "loss": 0.2445, + "step": 16440 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023750718578521032, + "loss": 0.2643, + "step": 16450 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023746799059315389, + "loss": 0.3245, + "step": 16460 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023742879540109745, + "loss": 0.2073, + "step": 16470 + }, + { + "epoch": 1.07, + "learning_rate": 0.000237389600209041, + "loss": 0.2568, + "step": 16480 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023735040501698457, + "loss": 0.2561, + "step": 16490 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023731120982492814, + "loss": 0.2521, + "step": 16500 + }, + { + "epoch": 1.07, + "learning_rate": 0.0002372720146328717, + "loss": 0.2952, + "step": 16510 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023723281944081526, + "loss": 0.2342, + "step": 16520 + }, + { + "epoch": 1.07, + "learning_rate": 0.0002371936242487588, + "loss": 0.2109, + "step": 16530 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023715442905670236, + "loss": 0.2717, + "step": 16540 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023711523386464592, + "loss": 0.2561, + "step": 16550 + }, + { + "epoch": 1.07, + "learning_rate": 0.00023707603867258948, + "loss": 0.3037, + "step": 16560 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023703684348053304, + "loss": 0.2176, + "step": 16570 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002369976482884766, + "loss": 0.2158, + "step": 16580 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023695845309642017, + "loss": 0.2575, + "step": 16590 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023691925790436373, + "loss": 0.2959, + "step": 16600 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002368800627123073, + "loss": 0.3332, + "step": 16610 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023684086752025086, + "loss": 0.21, + "step": 16620 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002368016723281944, + "loss": 0.2322, + "step": 16630 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023676247713613795, + "loss": 0.1951, + "step": 16640 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002367232819440815, + "loss": 0.2632, + "step": 16650 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023668408675202505, + "loss": 0.3375, + "step": 16660 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002366448915599686, + "loss": 0.2485, + "step": 16670 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023660569636791218, + "loss": 0.212, + "step": 16680 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023656650117585574, + "loss": 0.2353, + "step": 16690 + }, + { + "epoch": 1.08, + "learning_rate": 0.0002365273059837993, + "loss": 0.2519, + "step": 16700 + }, + { + "epoch": 1.08, + "learning_rate": 0.00023648811079174283, + "loss": 0.329, + "step": 16710 + }, + { + "epoch": 1.09, + "learning_rate": 0.0002364489155996864, + "loss": 0.1968, + "step": 16720 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023640972040762996, + "loss": 0.2269, + "step": 16730 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023637052521557352, + "loss": 0.2505, + "step": 16740 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023633133002351708, + "loss": 0.2687, + "step": 16750 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023629213483146065, + "loss": 0.3268, + "step": 16760 + }, + { + "epoch": 1.09, + "learning_rate": 0.0002362529396394042, + "loss": 0.2016, + "step": 16770 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023621374444734777, + "loss": 0.2309, + "step": 16780 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023617454925529133, + "loss": 0.2082, + "step": 16790 + }, + { + "epoch": 1.09, + "learning_rate": 0.0002361353540632349, + "loss": 0.2599, + "step": 16800 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023609615887117843, + "loss": 0.3604, + "step": 16810 + }, + { + "epoch": 1.09, + "learning_rate": 0.000236056963679122, + "loss": 0.2304, + "step": 16820 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023601776848706556, + "loss": 0.2322, + "step": 16830 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023597857329500912, + "loss": 0.2495, + "step": 16840 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023593937810295268, + "loss": 0.2924, + "step": 16850 + }, + { + "epoch": 1.09, + "learning_rate": 0.00023590018291089624, + "loss": 0.2954, + "step": 16860 + }, + { + "epoch": 1.09, + "learning_rate": 0.0002358609877188398, + "loss": 0.2179, + "step": 16870 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023582179252678337, + "loss": 0.2334, + "step": 16880 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023578259733472693, + "loss": 0.2298, + "step": 16890 + }, + { + "epoch": 1.1, + "learning_rate": 0.0002357434021426705, + "loss": 0.2766, + "step": 16900 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023570420695061403, + "loss": 0.3604, + "step": 16910 + }, + { + "epoch": 1.1, + "learning_rate": 0.0002356650117585576, + "loss": 0.2133, + "step": 16920 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023562581656650115, + "loss": 0.2231, + "step": 16930 + }, + { + "epoch": 1.1, + "learning_rate": 0.0002355866213744447, + "loss": 0.2322, + "step": 16940 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023554742618238828, + "loss": 0.2607, + "step": 16950 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023550823099033184, + "loss": 0.3535, + "step": 16960 + }, + { + "epoch": 1.1, + "learning_rate": 0.0002354690357982754, + "loss": 0.243, + "step": 16970 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023542984060621896, + "loss": 0.2132, + "step": 16980 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023539064541416252, + "loss": 0.2428, + "step": 16990 + }, + { + "epoch": 1.1, + "learning_rate": 0.0002353514502221061, + "loss": 0.2819, + "step": 17000 + }, + { + "epoch": 1.1, + "eval_loss": 0.29627758264541626, + "eval_runtime": 1828.0915, + "eval_samples_per_second": 8.527, + "eval_steps_per_second": 1.066, + "eval_wer": 0.2941820301061819, + "step": 17000 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023531225503004962, + "loss": 0.2957, + "step": 17010 + }, + { + "epoch": 1.1, + "learning_rate": 0.00023527305983799318, + "loss": 0.2191, + "step": 17020 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023523386464593675, + "loss": 0.245, + "step": 17030 + }, + { + "epoch": 1.11, + "learning_rate": 0.0002351946694538803, + "loss": 0.2687, + "step": 17040 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023515547426182387, + "loss": 0.2808, + "step": 17050 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023511627906976743, + "loss": 0.3599, + "step": 17060 + }, + { + "epoch": 1.11, + "learning_rate": 0.000235077083877711, + "loss": 0.2093, + "step": 17070 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023503788868565456, + "loss": 0.238, + "step": 17080 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023499869349359812, + "loss": 0.2621, + "step": 17090 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023495949830154168, + "loss": 0.2439, + "step": 17100 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023492030310948522, + "loss": 0.3255, + "step": 17110 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023488110791742878, + "loss": 0.2285, + "step": 17120 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023484191272537234, + "loss": 0.2454, + "step": 17130 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023480271753331588, + "loss": 0.224, + "step": 17140 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023476352234125944, + "loss": 0.2638, + "step": 17150 + }, + { + "epoch": 1.11, + "learning_rate": 0.000234724327149203, + "loss": 0.3255, + "step": 17160 + }, + { + "epoch": 1.11, + "learning_rate": 0.00023468513195714656, + "loss": 0.2168, + "step": 17170 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023464593676509013, + "loss": 0.2323, + "step": 17180 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023460674157303366, + "loss": 0.237, + "step": 17190 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023456754638097722, + "loss": 0.2245, + "step": 17200 + }, + { + "epoch": 1.12, + "learning_rate": 0.0002345283511889208, + "loss": 0.2794, + "step": 17210 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023448915599686435, + "loss": 0.2311, + "step": 17220 + }, + { + "epoch": 1.12, + "learning_rate": 0.0002344499608048079, + "loss": 0.2023, + "step": 17230 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023441076561275147, + "loss": 0.2368, + "step": 17240 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023437157042069504, + "loss": 0.2909, + "step": 17250 + }, + { + "epoch": 1.12, + "learning_rate": 0.0002343323752286386, + "loss": 0.292, + "step": 17260 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023429318003658216, + "loss": 0.229, + "step": 17270 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023425398484452572, + "loss": 0.2264, + "step": 17280 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023421478965246926, + "loss": 0.2388, + "step": 17290 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023417559446041282, + "loss": 0.2621, + "step": 17300 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023413639926835638, + "loss": 0.3067, + "step": 17310 + }, + { + "epoch": 1.12, + "learning_rate": 0.00023409720407629995, + "loss": 0.2137, + "step": 17320 + }, + { + "epoch": 1.12, + "learning_rate": 0.0002340580088842435, + "loss": 0.211, + "step": 17330 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023401881369218707, + "loss": 0.2345, + "step": 17340 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023398353801933625, + "loss": 0.3267, + "step": 17350 + }, + { + "epoch": 1.13, + "learning_rate": 0.0002339443428272798, + "loss": 0.3559, + "step": 17360 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023390514763522337, + "loss": 0.2342, + "step": 17370 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023386595244316694, + "loss": 0.2519, + "step": 17380 + }, + { + "epoch": 1.13, + "learning_rate": 0.0002338267572511105, + "loss": 0.2524, + "step": 17390 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023378756205905406, + "loss": 0.2436, + "step": 17400 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023374836686699762, + "loss": 0.3503, + "step": 17410 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023370917167494118, + "loss": 0.1997, + "step": 17420 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023366997648288475, + "loss": 0.205, + "step": 17430 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023363078129082828, + "loss": 0.2021, + "step": 17440 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023359158609877184, + "loss": 0.273, + "step": 17450 + }, + { + "epoch": 1.13, + "learning_rate": 0.0002335523909067154, + "loss": 0.2839, + "step": 17460 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023351319571465897, + "loss": 0.2351, + "step": 17470 + }, + { + "epoch": 1.13, + "learning_rate": 0.00023347400052260253, + "loss": 0.2231, + "step": 17480 + }, + { + "epoch": 1.14, + "learning_rate": 0.0002334348053305461, + "loss": 0.2575, + "step": 17490 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023339561013848966, + "loss": 0.2645, + "step": 17500 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023335641494643322, + "loss": 0.3461, + "step": 17510 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023331721975437678, + "loss": 0.2069, + "step": 17520 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023327802456232034, + "loss": 0.2476, + "step": 17530 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023323882937026388, + "loss": 0.242, + "step": 17540 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023319963417820744, + "loss": 0.2822, + "step": 17550 + }, + { + "epoch": 1.14, + "learning_rate": 0.000233160438986151, + "loss": 0.3431, + "step": 17560 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023312124379409456, + "loss": 0.2138, + "step": 17570 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023308204860203813, + "loss": 0.2514, + "step": 17580 + }, + { + "epoch": 1.14, + "learning_rate": 0.0002330428534099817, + "loss": 0.2522, + "step": 17590 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023300365821792525, + "loss": 0.2831, + "step": 17600 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023296446302586881, + "loss": 0.3466, + "step": 17610 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023292526783381238, + "loss": 0.2584, + "step": 17620 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023288607264175594, + "loss": 0.2354, + "step": 17630 + }, + { + "epoch": 1.14, + "learning_rate": 0.00023284687744969947, + "loss": 0.2348, + "step": 17640 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023280768225764304, + "loss": 0.2629, + "step": 17650 + }, + { + "epoch": 1.15, + "learning_rate": 0.0002327684870655866, + "loss": 0.3382, + "step": 17660 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023272929187353016, + "loss": 0.2501, + "step": 17670 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023269009668147372, + "loss": 0.2271, + "step": 17680 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023265090148941729, + "loss": 0.2551, + "step": 17690 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023261170629736085, + "loss": 0.2533, + "step": 17700 + }, + { + "epoch": 1.15, + "learning_rate": 0.0002325725111053044, + "loss": 0.3055, + "step": 17710 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023253331591324797, + "loss": 0.2411, + "step": 17720 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023249412072119153, + "loss": 0.2181, + "step": 17730 + }, + { + "epoch": 1.15, + "learning_rate": 0.0002324549255291351, + "loss": 0.2353, + "step": 17740 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023241573033707863, + "loss": 0.2754, + "step": 17750 + }, + { + "epoch": 1.15, + "learning_rate": 0.0002323765351450222, + "loss": 0.3454, + "step": 17760 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023233733995296576, + "loss": 0.2067, + "step": 17770 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023229814476090932, + "loss": 0.2496, + "step": 17780 + }, + { + "epoch": 1.15, + "learning_rate": 0.00023225894956885288, + "loss": 0.257, + "step": 17790 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023221975437679644, + "loss": 0.2529, + "step": 17800 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023218055918474, + "loss": 0.3392, + "step": 17810 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023214136399268357, + "loss": 0.1949, + "step": 17820 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023210216880062708, + "loss": 0.2169, + "step": 17830 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023206297360857064, + "loss": 0.2305, + "step": 17840 + }, + { + "epoch": 1.16, + "learning_rate": 0.0002320237784165142, + "loss": 0.2963, + "step": 17850 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023198458322445776, + "loss": 0.3339, + "step": 17860 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023194538803240133, + "loss": 0.2205, + "step": 17870 + }, + { + "epoch": 1.16, + "learning_rate": 0.0002319061928403449, + "loss": 0.2207, + "step": 17880 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023186699764828845, + "loss": 0.2421, + "step": 17890 + }, + { + "epoch": 1.16, + "learning_rate": 0.000231827802456232, + "loss": 0.253, + "step": 17900 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023178860726417557, + "loss": 0.3418, + "step": 17910 + }, + { + "epoch": 1.16, + "learning_rate": 0.0002317494120721191, + "loss": 0.2372, + "step": 17920 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023171021688006267, + "loss": 0.2533, + "step": 17930 + }, + { + "epoch": 1.16, + "learning_rate": 0.00023167102168800623, + "loss": 0.2564, + "step": 17940 + }, + { + "epoch": 1.16, + "learning_rate": 0.0002316318264959498, + "loss": 0.2696, + "step": 17950 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023159263130389336, + "loss": 0.257, + "step": 17960 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023155343611183692, + "loss": 0.246, + "step": 17970 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023151424091978048, + "loss": 0.2128, + "step": 17980 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023147504572772405, + "loss": 0.2417, + "step": 17990 + }, + { + "epoch": 1.17, + "learning_rate": 0.0002314358505356676, + "loss": 0.2538, + "step": 18000 + }, + { + "epoch": 1.17, + "eval_loss": 0.29100653529167175, + "eval_runtime": 1760.1597, + "eval_samples_per_second": 8.856, + "eval_steps_per_second": 1.107, + "eval_wer": 0.28448067547014255, + "step": 18000 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023139665534361117, + "loss": 0.307, + "step": 18010 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023135746015155473, + "loss": 0.2463, + "step": 18020 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023131826495949827, + "loss": 0.2343, + "step": 18030 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023127906976744183, + "loss": 0.2357, + "step": 18040 + }, + { + "epoch": 1.17, + "learning_rate": 0.0002312398745753854, + "loss": 0.2922, + "step": 18050 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023120067938332895, + "loss": 0.3159, + "step": 18060 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023116148419127252, + "loss": 0.2075, + "step": 18070 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023112228899921608, + "loss": 0.2328, + "step": 18080 + }, + { + "epoch": 1.17, + "learning_rate": 0.00023108309380715964, + "loss": 0.247, + "step": 18090 + }, + { + "epoch": 1.17, + "learning_rate": 0.0002310438986151032, + "loss": 0.2417, + "step": 18100 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023100470342304677, + "loss": 0.3105, + "step": 18110 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023096550823099033, + "loss": 0.2298, + "step": 18120 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023092631303893386, + "loss": 0.2038, + "step": 18130 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023088711784687743, + "loss": 0.2228, + "step": 18140 + }, + { + "epoch": 1.18, + "learning_rate": 0.000230847922654821, + "loss": 0.2793, + "step": 18150 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023080872746276455, + "loss": 0.3238, + "step": 18160 + }, + { + "epoch": 1.18, + "learning_rate": 0.0002307695322707081, + "loss": 0.2249, + "step": 18170 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023073033707865167, + "loss": 0.2471, + "step": 18180 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023069114188659524, + "loss": 0.2388, + "step": 18190 + }, + { + "epoch": 1.18, + "learning_rate": 0.0002306519466945388, + "loss": 0.2374, + "step": 18200 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023061275150248236, + "loss": 0.3674, + "step": 18210 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023057355631042592, + "loss": 0.1863, + "step": 18220 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023053436111836946, + "loss": 0.256, + "step": 18230 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023049516592631302, + "loss": 0.2653, + "step": 18240 + }, + { + "epoch": 1.18, + "learning_rate": 0.00023045597073425658, + "loss": 0.2837, + "step": 18250 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023041677554220015, + "loss": 0.3333, + "step": 18260 + }, + { + "epoch": 1.19, + "learning_rate": 0.0002303775803501437, + "loss": 0.192, + "step": 18270 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023033838515808727, + "loss": 0.2164, + "step": 18280 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023029918996603083, + "loss": 0.2132, + "step": 18290 + }, + { + "epoch": 1.19, + "learning_rate": 0.0002302599947739744, + "loss": 0.2504, + "step": 18300 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023022079958191796, + "loss": 0.3399, + "step": 18310 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023018160438986147, + "loss": 0.2029, + "step": 18320 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023014240919780503, + "loss": 0.2183, + "step": 18330 + }, + { + "epoch": 1.19, + "learning_rate": 0.0002301032140057486, + "loss": 0.257, + "step": 18340 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023006401881369215, + "loss": 0.2573, + "step": 18350 + }, + { + "epoch": 1.19, + "learning_rate": 0.00023002482362163571, + "loss": 0.3343, + "step": 18360 + }, + { + "epoch": 1.19, + "learning_rate": 0.00022998562842957928, + "loss": 0.2156, + "step": 18370 + }, + { + "epoch": 1.19, + "learning_rate": 0.00022994643323752284, + "loss": 0.2207, + "step": 18380 + }, + { + "epoch": 1.19, + "learning_rate": 0.0002299072380454664, + "loss": 0.2229, + "step": 18390 + }, + { + "epoch": 1.19, + "learning_rate": 0.00022986804285340996, + "loss": 0.2774, + "step": 18400 + }, + { + "epoch": 1.19, + "learning_rate": 0.0002298288476613535, + "loss": 0.3279, + "step": 18410 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022978965246929706, + "loss": 0.2375, + "step": 18420 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022975045727724062, + "loss": 0.234, + "step": 18430 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022971126208518419, + "loss": 0.2276, + "step": 18440 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022967206689312775, + "loss": 0.2646, + "step": 18450 + }, + { + "epoch": 1.2, + "learning_rate": 0.0002296328717010713, + "loss": 0.3981, + "step": 18460 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022959367650901487, + "loss": 0.217, + "step": 18470 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022955448131695844, + "loss": 0.242, + "step": 18480 + }, + { + "epoch": 1.2, + "learning_rate": 0.000229515286124902, + "loss": 0.2432, + "step": 18490 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022947609093284556, + "loss": 0.2504, + "step": 18500 + }, + { + "epoch": 1.2, + "learning_rate": 0.0002294368957407891, + "loss": 0.304, + "step": 18510 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022939770054873266, + "loss": 0.1812, + "step": 18520 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022935850535667622, + "loss": 0.2093, + "step": 18530 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022931931016461978, + "loss": 0.2697, + "step": 18540 + }, + { + "epoch": 1.2, + "learning_rate": 0.00022928011497256334, + "loss": 0.2535, + "step": 18550 + }, + { + "epoch": 1.2, + "learning_rate": 0.0002292409197805069, + "loss": 0.341, + "step": 18560 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022920172458845047, + "loss": 0.2033, + "step": 18570 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022916252939639403, + "loss": 0.2012, + "step": 18580 + }, + { + "epoch": 1.21, + "learning_rate": 0.0002291233342043376, + "loss": 0.2478, + "step": 18590 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022908413901228116, + "loss": 0.2376, + "step": 18600 + }, + { + "epoch": 1.21, + "learning_rate": 0.0002290449438202247, + "loss": 0.349, + "step": 18610 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022900574862816825, + "loss": 0.1938, + "step": 18620 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022896655343611182, + "loss": 0.2356, + "step": 18630 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022892735824405538, + "loss": 0.2438, + "step": 18640 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022888816305199894, + "loss": 0.2986, + "step": 18650 + }, + { + "epoch": 1.21, + "learning_rate": 0.0002288489678599425, + "loss": 0.3382, + "step": 18660 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022880977266788606, + "loss": 0.1947, + "step": 18670 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022877057747582963, + "loss": 0.2178, + "step": 18680 + }, + { + "epoch": 1.21, + "learning_rate": 0.0002287313822837732, + "loss": 0.2226, + "step": 18690 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022869218709171675, + "loss": 0.2758, + "step": 18700 + }, + { + "epoch": 1.21, + "learning_rate": 0.0002286529918996603, + "loss": 0.3417, + "step": 18710 + }, + { + "epoch": 1.21, + "learning_rate": 0.00022861379670760385, + "loss": 0.2475, + "step": 18720 + }, + { + "epoch": 1.22, + "learning_rate": 0.0002285746015155474, + "loss": 0.2196, + "step": 18730 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022853540632349097, + "loss": 0.2445, + "step": 18740 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022849621113143454, + "loss": 0.2332, + "step": 18750 + }, + { + "epoch": 1.22, + "learning_rate": 0.0002284570159393781, + "loss": 0.3128, + "step": 18760 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022841782074732166, + "loss": 0.2068, + "step": 18770 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022837862555526522, + "loss": 0.2342, + "step": 18780 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022833943036320878, + "loss": 0.2317, + "step": 18790 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022830023517115235, + "loss": 0.3235, + "step": 18800 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022826103997909586, + "loss": 0.3282, + "step": 18810 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022822184478703942, + "loss": 0.2261, + "step": 18820 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022818264959498298, + "loss": 0.224, + "step": 18830 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022814345440292654, + "loss": 0.234, + "step": 18840 + }, + { + "epoch": 1.22, + "learning_rate": 0.0002281042592108701, + "loss": 0.2534, + "step": 18850 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022806506401881367, + "loss": 0.3074, + "step": 18860 + }, + { + "epoch": 1.22, + "learning_rate": 0.00022802586882675723, + "loss": 0.1776, + "step": 18870 + }, + { + "epoch": 1.23, + "learning_rate": 0.0002279866736347008, + "loss": 0.2191, + "step": 18880 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022794747844264433, + "loss": 0.2308, + "step": 18890 + }, + { + "epoch": 1.23, + "learning_rate": 0.0002279082832505879, + "loss": 0.2762, + "step": 18900 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022786908805853145, + "loss": 0.3058, + "step": 18910 + }, + { + "epoch": 1.23, + "learning_rate": 0.000227829892866475, + "loss": 0.2105, + "step": 18920 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022779069767441858, + "loss": 0.2074, + "step": 18930 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022775150248236214, + "loss": 0.2659, + "step": 18940 + }, + { + "epoch": 1.23, + "learning_rate": 0.0002277123072903057, + "loss": 0.3194, + "step": 18950 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022767311209824926, + "loss": 0.3213, + "step": 18960 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022763391690619282, + "loss": 0.2309, + "step": 18970 + }, + { + "epoch": 1.23, + "learning_rate": 0.0002275947217141364, + "loss": 0.2508, + "step": 18980 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022755552652207992, + "loss": 0.2289, + "step": 18990 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022751633133002348, + "loss": 0.2405, + "step": 19000 + }, + { + "epoch": 1.23, + "eval_loss": 0.28857871890068054, + "eval_runtime": 1856.8287, + "eval_samples_per_second": 8.395, + "eval_steps_per_second": 1.05, + "eval_wer": 0.28055749029864535, + "step": 19000 + }, + { + "epoch": 1.23, + "learning_rate": 0.00022747713613796705, + "loss": 0.3035, + "step": 19010 + }, + { + "epoch": 1.23, + "learning_rate": 0.0002274379409459106, + "loss": 0.2022, + "step": 19020 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022739874575385417, + "loss": 0.2325, + "step": 19030 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022735955056179773, + "loss": 0.2491, + "step": 19040 + }, + { + "epoch": 1.24, + "learning_rate": 0.0002273203553697413, + "loss": 0.3011, + "step": 19050 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022728116017768486, + "loss": 0.3047, + "step": 19060 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022724196498562842, + "loss": 0.2133, + "step": 19070 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022720276979357198, + "loss": 0.1925, + "step": 19080 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022716357460151552, + "loss": 0.2533, + "step": 19090 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022712437940945908, + "loss": 0.2652, + "step": 19100 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022708518421740264, + "loss": 0.3707, + "step": 19110 + }, + { + "epoch": 1.24, + "learning_rate": 0.0002270459890253462, + "loss": 0.1953, + "step": 19120 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022700679383328977, + "loss": 0.2215, + "step": 19130 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022696759864123333, + "loss": 0.2537, + "step": 19140 + }, + { + "epoch": 1.24, + "learning_rate": 0.0002269284034491769, + "loss": 0.2626, + "step": 19150 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022688920825712045, + "loss": 0.3797, + "step": 19160 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022685001306506402, + "loss": 0.2018, + "step": 19170 + }, + { + "epoch": 1.24, + "learning_rate": 0.00022681081787300758, + "loss": 0.1922, + "step": 19180 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022677162268095111, + "loss": 0.2201, + "step": 19190 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022673242748889468, + "loss": 0.2706, + "step": 19200 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022669323229683824, + "loss": 0.3174, + "step": 19210 + }, + { + "epoch": 1.25, + "learning_rate": 0.0002266540371047818, + "loss": 0.2308, + "step": 19220 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022661484191272536, + "loss": 0.224, + "step": 19230 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022657564672066893, + "loss": 0.2081, + "step": 19240 + }, + { + "epoch": 1.25, + "learning_rate": 0.0002265364515286125, + "loss": 0.2417, + "step": 19250 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022649725633655605, + "loss": 0.3129, + "step": 19260 + }, + { + "epoch": 1.25, + "learning_rate": 0.0002264580611444996, + "loss": 0.1898, + "step": 19270 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022641886595244317, + "loss": 0.2433, + "step": 19280 + }, + { + "epoch": 1.25, + "learning_rate": 0.0002263796707603867, + "loss": 0.2311, + "step": 19290 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022634047556833024, + "loss": 0.242, + "step": 19300 + }, + { + "epoch": 1.25, + "learning_rate": 0.0002263012803762738, + "loss": 0.3312, + "step": 19310 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022626208518421737, + "loss": 0.2381, + "step": 19320 + }, + { + "epoch": 1.25, + "learning_rate": 0.00022622288999216093, + "loss": 0.1987, + "step": 19330 + }, + { + "epoch": 1.26, + "learning_rate": 0.0002261836948001045, + "loss": 0.206, + "step": 19340 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022614449960804806, + "loss": 0.2366, + "step": 19350 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022610530441599162, + "loss": 0.3462, + "step": 19360 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022606610922393515, + "loss": 0.2133, + "step": 19370 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022602691403187872, + "loss": 0.2006, + "step": 19380 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022598771883982228, + "loss": 0.2345, + "step": 19390 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022594852364776584, + "loss": 0.2805, + "step": 19400 + }, + { + "epoch": 1.26, + "learning_rate": 0.0002259093284557094, + "loss": 0.3569, + "step": 19410 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022587013326365297, + "loss": 0.2212, + "step": 19420 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022583093807159653, + "loss": 0.2116, + "step": 19430 + }, + { + "epoch": 1.26, + "learning_rate": 0.0002257917428795401, + "loss": 0.2204, + "step": 19440 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022575254768748365, + "loss": 0.3132, + "step": 19450 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022571335249542721, + "loss": 0.3189, + "step": 19460 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022567415730337075, + "loss": 0.2096, + "step": 19470 + }, + { + "epoch": 1.26, + "learning_rate": 0.0002256349621113143, + "loss": 0.2159, + "step": 19480 + }, + { + "epoch": 1.26, + "learning_rate": 0.00022559576691925787, + "loss": 0.2742, + "step": 19490 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022555657172720144, + "loss": 0.2537, + "step": 19500 + }, + { + "epoch": 1.27, + "learning_rate": 0.000225517376535145, + "loss": 0.3385, + "step": 19510 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022547818134308856, + "loss": 0.1892, + "step": 19520 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022543898615103212, + "loss": 0.2141, + "step": 19530 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022539979095897569, + "loss": 0.1971, + "step": 19540 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022536059576691925, + "loss": 0.2718, + "step": 19550 + }, + { + "epoch": 1.27, + "learning_rate": 0.0002253214005748628, + "loss": 0.2969, + "step": 19560 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022528220538280635, + "loss": 0.2231, + "step": 19570 + }, + { + "epoch": 1.27, + "learning_rate": 0.0002252430101907499, + "loss": 0.2156, + "step": 19580 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022520381499869347, + "loss": 0.2372, + "step": 19590 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022516461980663703, + "loss": 0.2237, + "step": 19600 + }, + { + "epoch": 1.27, + "learning_rate": 0.0002251254246145806, + "loss": 0.3106, + "step": 19610 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022508622942252416, + "loss": 0.1904, + "step": 19620 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022504703423046772, + "loss": 0.2453, + "step": 19630 + }, + { + "epoch": 1.27, + "learning_rate": 0.00022500783903841128, + "loss": 0.2488, + "step": 19640 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022496864384635484, + "loss": 0.2728, + "step": 19650 + }, + { + "epoch": 1.28, + "learning_rate": 0.0002249294486542984, + "loss": 0.2919, + "step": 19660 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022489025346224194, + "loss": 0.2366, + "step": 19670 + }, + { + "epoch": 1.28, + "learning_rate": 0.0002248510582701855, + "loss": 0.2051, + "step": 19680 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022481186307812907, + "loss": 0.2444, + "step": 19690 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022477266788607263, + "loss": 0.2867, + "step": 19700 + }, + { + "epoch": 1.28, + "learning_rate": 0.0002247334726940162, + "loss": 0.3331, + "step": 19710 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022469427750195975, + "loss": 0.216, + "step": 19720 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022465508230990331, + "loss": 0.1718, + "step": 19730 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022461588711784688, + "loss": 0.2522, + "step": 19740 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022457669192579044, + "loss": 0.2681, + "step": 19750 + }, + { + "epoch": 1.28, + "learning_rate": 0.000224537496733734, + "loss": 0.3501, + "step": 19760 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022449830154167754, + "loss": 0.226, + "step": 19770 + }, + { + "epoch": 1.28, + "learning_rate": 0.0002244591063496211, + "loss": 0.2071, + "step": 19780 + }, + { + "epoch": 1.28, + "learning_rate": 0.00022441991115756463, + "loss": 0.2268, + "step": 19790 + }, + { + "epoch": 1.29, + "learning_rate": 0.0002243807159655082, + "loss": 0.2585, + "step": 19800 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022434152077345176, + "loss": 0.3369, + "step": 19810 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022430232558139532, + "loss": 0.2378, + "step": 19820 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022426313038933888, + "loss": 0.2224, + "step": 19830 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022422393519728245, + "loss": 0.2499, + "step": 19840 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022418474000522598, + "loss": 0.2523, + "step": 19850 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022414554481316954, + "loss": 0.3291, + "step": 19860 + }, + { + "epoch": 1.29, + "learning_rate": 0.0002241063496211131, + "loss": 0.2088, + "step": 19870 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022406715442905667, + "loss": 0.2287, + "step": 19880 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022402795923700023, + "loss": 0.246, + "step": 19890 + }, + { + "epoch": 1.29, + "learning_rate": 0.0002239887640449438, + "loss": 0.2602, + "step": 19900 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022394956885288735, + "loss": 0.3576, + "step": 19910 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022391037366083092, + "loss": 0.2001, + "step": 19920 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022387117846877448, + "loss": 0.2069, + "step": 19930 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022383198327671804, + "loss": 0.2287, + "step": 19940 + }, + { + "epoch": 1.29, + "learning_rate": 0.00022379278808466158, + "loss": 0.2394, + "step": 19950 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022375359289260514, + "loss": 0.2856, + "step": 19960 + }, + { + "epoch": 1.3, + "learning_rate": 0.0002237143977005487, + "loss": 0.2103, + "step": 19970 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022367520250849226, + "loss": 0.2105, + "step": 19980 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022363600731643583, + "loss": 0.2298, + "step": 19990 + }, + { + "epoch": 1.3, + "learning_rate": 0.0002235968121243794, + "loss": 0.2734, + "step": 20000 + }, + { + "epoch": 1.3, + "eval_loss": 0.2862153947353363, + "eval_runtime": 963.0332, + "eval_samples_per_second": 16.186, + "eval_steps_per_second": 2.024, + "eval_wer": 0.2832795553723472, + "step": 20000 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022355761693232295, + "loss": 0.284, + "step": 20010 + }, + { + "epoch": 1.3, + "learning_rate": 0.0002235184217402665, + "loss": 0.2246, + "step": 20020 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022347922654821008, + "loss": 0.2335, + "step": 20030 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022344003135615364, + "loss": 0.28, + "step": 20040 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022340083616409717, + "loss": 0.2231, + "step": 20050 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022336164097204073, + "loss": 0.3393, + "step": 20060 + }, + { + "epoch": 1.3, + "learning_rate": 0.0002233224457799843, + "loss": 0.2398, + "step": 20070 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022328325058792786, + "loss": 0.2175, + "step": 20080 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022324405539587142, + "loss": 0.2557, + "step": 20090 + }, + { + "epoch": 1.3, + "learning_rate": 0.00022320486020381498, + "loss": 0.2568, + "step": 20100 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022316566501175855, + "loss": 0.3544, + "step": 20110 + }, + { + "epoch": 1.31, + "learning_rate": 0.0002231264698197021, + "loss": 0.2136, + "step": 20120 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022308727462764567, + "loss": 0.2345, + "step": 20130 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022304807943558923, + "loss": 0.2712, + "step": 20140 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022300888424353277, + "loss": 0.2415, + "step": 20150 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022296968905147633, + "loss": 0.3369, + "step": 20160 + }, + { + "epoch": 1.31, + "learning_rate": 0.0002229304938594199, + "loss": 0.1922, + "step": 20170 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022289129866736346, + "loss": 0.2451, + "step": 20180 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022285210347530702, + "loss": 0.2578, + "step": 20190 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022281290828325058, + "loss": 0.2834, + "step": 20200 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022277371309119414, + "loss": 0.4481, + "step": 20210 + }, + { + "epoch": 1.31, + "learning_rate": 0.0002227345178991377, + "loss": 0.209, + "step": 20220 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022269532270708127, + "loss": 0.2396, + "step": 20230 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022265612751502483, + "loss": 0.2321, + "step": 20240 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022261693232296836, + "loss": 0.243, + "step": 20250 + }, + { + "epoch": 1.31, + "learning_rate": 0.00022257773713091193, + "loss": 0.3397, + "step": 20260 + }, + { + "epoch": 1.32, + "learning_rate": 0.0002225385419388555, + "loss": 0.2505, + "step": 20270 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022249934674679902, + "loss": 0.1953, + "step": 20280 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022246015155474259, + "loss": 0.223, + "step": 20290 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022242095636268615, + "loss": 0.2753, + "step": 20300 + }, + { + "epoch": 1.32, + "learning_rate": 0.0002223817611706297, + "loss": 0.3296, + "step": 20310 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022234256597857327, + "loss": 0.2024, + "step": 20320 + }, + { + "epoch": 1.32, + "learning_rate": 0.0002223033707865168, + "loss": 0.245, + "step": 20330 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022226417559446037, + "loss": 0.2502, + "step": 20340 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022222498040240393, + "loss": 0.306, + "step": 20350 + }, + { + "epoch": 1.32, + "learning_rate": 0.0002221857852103475, + "loss": 0.3138, + "step": 20360 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022214659001829106, + "loss": 0.1854, + "step": 20370 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022210739482623462, + "loss": 0.22, + "step": 20380 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022206819963417818, + "loss": 0.2514, + "step": 20390 + }, + { + "epoch": 1.32, + "learning_rate": 0.00022202900444212174, + "loss": 0.2509, + "step": 20400 + }, + { + "epoch": 1.32, + "learning_rate": 0.0002219898092500653, + "loss": 0.342, + "step": 20410 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022195061405800887, + "loss": 0.2079, + "step": 20420 + }, + { + "epoch": 1.33, + "learning_rate": 0.0002219114188659524, + "loss": 0.2618, + "step": 20430 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022187222367389597, + "loss": 0.2265, + "step": 20440 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022183302848183953, + "loss": 0.2557, + "step": 20450 + }, + { + "epoch": 1.33, + "learning_rate": 0.0002217938332897831, + "loss": 0.3153, + "step": 20460 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022175463809772665, + "loss": 0.2214, + "step": 20470 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022171544290567022, + "loss": 0.1934, + "step": 20480 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022167624771361378, + "loss": 0.2182, + "step": 20490 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022163705252155734, + "loss": 0.2163, + "step": 20500 + }, + { + "epoch": 1.33, + "learning_rate": 0.0002215978573295009, + "loss": 0.2655, + "step": 20510 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022155866213744446, + "loss": 0.2062, + "step": 20520 + }, + { + "epoch": 1.33, + "learning_rate": 0.000221519466945388, + "loss": 0.2261, + "step": 20530 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022148027175333156, + "loss": 0.2586, + "step": 20540 + }, + { + "epoch": 1.33, + "learning_rate": 0.00022144107656127512, + "loss": 0.2488, + "step": 20550 + }, + { + "epoch": 1.33, + "learning_rate": 0.0002214018813692187, + "loss": 0.3238, + "step": 20560 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022136268617716225, + "loss": 0.2111, + "step": 20570 + }, + { + "epoch": 1.34, + "learning_rate": 0.0002213234909851058, + "loss": 0.2075, + "step": 20580 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022128429579304937, + "loss": 0.2344, + "step": 20590 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022124510060099294, + "loss": 0.2568, + "step": 20600 + }, + { + "epoch": 1.34, + "learning_rate": 0.0002212059054089365, + "loss": 0.312, + "step": 20610 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022116671021688006, + "loss": 0.2135, + "step": 20620 + }, + { + "epoch": 1.34, + "learning_rate": 0.0002211275150248236, + "loss": 0.234, + "step": 20630 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022108831983276716, + "loss": 0.2493, + "step": 20640 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022104912464071072, + "loss": 0.2657, + "step": 20650 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022100992944865428, + "loss": 0.2916, + "step": 20660 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022097073425659785, + "loss": 0.188, + "step": 20670 + }, + { + "epoch": 1.34, + "learning_rate": 0.0002209315390645414, + "loss": 0.218, + "step": 20680 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022089234387248497, + "loss": 0.2478, + "step": 20690 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022085314868042853, + "loss": 0.2241, + "step": 20700 + }, + { + "epoch": 1.34, + "learning_rate": 0.0002208139534883721, + "loss": 0.3791, + "step": 20710 + }, + { + "epoch": 1.34, + "learning_rate": 0.00022077475829631566, + "loss": 0.2079, + "step": 20720 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022073556310425922, + "loss": 0.205, + "step": 20730 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022069636791220275, + "loss": 0.2347, + "step": 20740 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022065717272014632, + "loss": 0.2893, + "step": 20750 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022061797752808988, + "loss": 0.3439, + "step": 20760 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022057878233603344, + "loss": 0.1931, + "step": 20770 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022053958714397698, + "loss": 0.2287, + "step": 20780 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022050039195192054, + "loss": 0.226, + "step": 20790 + }, + { + "epoch": 1.35, + "learning_rate": 0.0002204611967598641, + "loss": 0.2655, + "step": 20800 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022042200156780764, + "loss": 0.3072, + "step": 20810 + }, + { + "epoch": 1.35, + "learning_rate": 0.0002203828063757512, + "loss": 0.2154, + "step": 20820 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022034361118369476, + "loss": 0.2321, + "step": 20830 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022030441599163832, + "loss": 0.2484, + "step": 20840 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022026522079958188, + "loss": 0.2738, + "step": 20850 + }, + { + "epoch": 1.35, + "learning_rate": 0.00022022602560752545, + "loss": 0.3188, + "step": 20860 + }, + { + "epoch": 1.35, + "learning_rate": 0.000220186830415469, + "loss": 0.2043, + "step": 20870 + }, + { + "epoch": 1.36, + "learning_rate": 0.00022014763522341257, + "loss": 0.2236, + "step": 20880 + }, + { + "epoch": 1.36, + "learning_rate": 0.00022010844003135613, + "loss": 0.2451, + "step": 20890 + }, + { + "epoch": 1.36, + "learning_rate": 0.0002200692448392997, + "loss": 0.2363, + "step": 20900 + }, + { + "epoch": 1.36, + "learning_rate": 0.00022003004964724323, + "loss": 0.2984, + "step": 20910 + }, + { + "epoch": 1.36, + "learning_rate": 0.0002199908544551868, + "loss": 0.226, + "step": 20920 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021995165926313036, + "loss": 0.1887, + "step": 20930 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021991246407107392, + "loss": 0.2414, + "step": 20940 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021987326887901748, + "loss": 0.2492, + "step": 20950 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021983407368696104, + "loss": 0.2814, + "step": 20960 + }, + { + "epoch": 1.36, + "learning_rate": 0.0002197948784949046, + "loss": 0.2247, + "step": 20970 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021975568330284817, + "loss": 0.2182, + "step": 20980 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021971648811079173, + "loss": 0.2411, + "step": 20990 + }, + { + "epoch": 1.36, + "learning_rate": 0.0002196812124379409, + "loss": 0.2585, + "step": 21000 + }, + { + "epoch": 1.36, + "eval_loss": 0.2830195128917694, + "eval_runtime": 986.273, + "eval_samples_per_second": 15.805, + "eval_steps_per_second": 1.976, + "eval_wer": 0.27263294054100157, + "step": 21000 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021964201724588447, + "loss": 0.2848, + "step": 21010 + }, + { + "epoch": 1.36, + "learning_rate": 0.00021960282205382803, + "loss": 0.2121, + "step": 21020 + }, + { + "epoch": 1.36, + "learning_rate": 0.0002195636268617716, + "loss": 0.1801, + "step": 21030 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021952443166971516, + "loss": 0.2293, + "step": 21040 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021948523647765872, + "loss": 0.2731, + "step": 21050 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021944604128560228, + "loss": 0.3529, + "step": 21060 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021940684609354582, + "loss": 0.187, + "step": 21070 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021936765090148938, + "loss": 0.2206, + "step": 21080 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021932845570943294, + "loss": 0.2762, + "step": 21090 + }, + { + "epoch": 1.37, + "learning_rate": 0.0002192892605173765, + "loss": 0.2541, + "step": 21100 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021925006532532007, + "loss": 0.3229, + "step": 21110 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021921087013326363, + "loss": 0.2262, + "step": 21120 + }, + { + "epoch": 1.37, + "learning_rate": 0.0002191716749412072, + "loss": 0.2259, + "step": 21130 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021913247974915075, + "loss": 0.2485, + "step": 21140 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021909328455709432, + "loss": 0.2823, + "step": 21150 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021905408936503788, + "loss": 0.2904, + "step": 21160 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021901489417298141, + "loss": 0.1756, + "step": 21170 + }, + { + "epoch": 1.37, + "learning_rate": 0.00021897569898092498, + "loss": 0.2191, + "step": 21180 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021893650378886854, + "loss": 0.2042, + "step": 21190 + }, + { + "epoch": 1.38, + "learning_rate": 0.0002188973085968121, + "loss": 0.2791, + "step": 21200 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021885811340475566, + "loss": 0.2799, + "step": 21210 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021881891821269923, + "loss": 0.2081, + "step": 21220 + }, + { + "epoch": 1.38, + "learning_rate": 0.0002187797230206428, + "loss": 0.2216, + "step": 21230 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021874052782858635, + "loss": 0.2353, + "step": 21240 + }, + { + "epoch": 1.38, + "learning_rate": 0.0002187013326365299, + "loss": 0.2483, + "step": 21250 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021866213744447347, + "loss": 0.3458, + "step": 21260 + }, + { + "epoch": 1.38, + "learning_rate": 0.000218622942252417, + "loss": 0.2134, + "step": 21270 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021858374706036057, + "loss": 0.2119, + "step": 21280 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021854455186830413, + "loss": 0.2181, + "step": 21290 + }, + { + "epoch": 1.38, + "learning_rate": 0.0002185053566762477, + "loss": 0.2736, + "step": 21300 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021846616148419126, + "loss": 0.3367, + "step": 21310 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021842696629213482, + "loss": 0.1875, + "step": 21320 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021838777110007838, + "loss": 0.2231, + "step": 21330 + }, + { + "epoch": 1.38, + "learning_rate": 0.00021834857590802195, + "loss": 0.235, + "step": 21340 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002183093807159655, + "loss": 0.2502, + "step": 21350 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021827018552390907, + "loss": 0.2939, + "step": 21360 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002182309903318526, + "loss": 0.1797, + "step": 21370 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021819179513979617, + "loss": 0.1941, + "step": 21380 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021815259994773973, + "loss": 0.2301, + "step": 21390 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002181134047556833, + "loss": 0.2229, + "step": 21400 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021807420956362685, + "loss": 0.3238, + "step": 21410 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021803501437157042, + "loss": 0.2208, + "step": 21420 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021799581917951398, + "loss": 0.2044, + "step": 21430 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021795662398745754, + "loss": 0.2371, + "step": 21440 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002179174287954011, + "loss": 0.2368, + "step": 21450 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002178782336033446, + "loss": 0.2697, + "step": 21460 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021783903841128817, + "loss": 0.2013, + "step": 21470 + }, + { + "epoch": 1.39, + "learning_rate": 0.00021779984321923174, + "loss": 0.1946, + "step": 21480 + }, + { + "epoch": 1.39, + "learning_rate": 0.0002177606480271753, + "loss": 0.2363, + "step": 21490 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021772145283511886, + "loss": 0.2309, + "step": 21500 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021768225764306242, + "loss": 0.317, + "step": 21510 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021764306245100599, + "loss": 0.2096, + "step": 21520 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021760386725894955, + "loss": 0.2015, + "step": 21530 + }, + { + "epoch": 1.4, + "learning_rate": 0.0002175646720668931, + "loss": 0.2571, + "step": 21540 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021752547687483665, + "loss": 0.2198, + "step": 21550 + }, + { + "epoch": 1.4, + "learning_rate": 0.0002174862816827802, + "loss": 0.2928, + "step": 21560 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021744708649072377, + "loss": 0.2069, + "step": 21570 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021740789129866733, + "loss": 0.2196, + "step": 21580 + }, + { + "epoch": 1.4, + "learning_rate": 0.0002173686961066109, + "loss": 0.2367, + "step": 21590 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021732950091455446, + "loss": 0.2857, + "step": 21600 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021729030572249802, + "loss": 0.3472, + "step": 21610 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021725111053044158, + "loss": 0.2248, + "step": 21620 + }, + { + "epoch": 1.4, + "learning_rate": 0.00021721191533838514, + "loss": 0.2395, + "step": 21630 + }, + { + "epoch": 1.4, + "learning_rate": 0.0002171727201463287, + "loss": 0.2615, + "step": 21640 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021713352495427224, + "loss": 0.2525, + "step": 21650 + }, + { + "epoch": 1.41, + "learning_rate": 0.0002170943297622158, + "loss": 0.3205, + "step": 21660 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021705513457015937, + "loss": 0.2149, + "step": 21670 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021701593937810293, + "loss": 0.2073, + "step": 21680 + }, + { + "epoch": 1.41, + "learning_rate": 0.0002169767441860465, + "loss": 0.2318, + "step": 21690 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021693754899399005, + "loss": 0.2859, + "step": 21700 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021689835380193361, + "loss": 0.2984, + "step": 21710 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021685915860987718, + "loss": 0.198, + "step": 21720 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021681996341782074, + "loss": 0.2136, + "step": 21730 + }, + { + "epoch": 1.41, + "learning_rate": 0.0002167807682257643, + "loss": 0.2474, + "step": 21740 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021674157303370784, + "loss": 0.2276, + "step": 21750 + }, + { + "epoch": 1.41, + "learning_rate": 0.0002167023778416514, + "loss": 0.2987, + "step": 21760 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021666318264959496, + "loss": 0.2376, + "step": 21770 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021662398745753852, + "loss": 0.1848, + "step": 21780 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021658479226548209, + "loss": 0.239, + "step": 21790 + }, + { + "epoch": 1.41, + "learning_rate": 0.00021654559707342565, + "loss": 0.3002, + "step": 21800 + }, + { + "epoch": 1.42, + "learning_rate": 0.0002165064018813692, + "loss": 0.3567, + "step": 21810 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021646720668931277, + "loss": 0.202, + "step": 21820 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021642801149725634, + "loss": 0.2155, + "step": 21830 + }, + { + "epoch": 1.42, + "learning_rate": 0.0002163888163051999, + "loss": 0.2259, + "step": 21840 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021634962111314343, + "loss": 0.2722, + "step": 21850 + }, + { + "epoch": 1.42, + "learning_rate": 0.000216310425921087, + "loss": 0.3012, + "step": 21860 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021627123072903056, + "loss": 0.198, + "step": 21870 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021623203553697412, + "loss": 0.2024, + "step": 21880 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021619284034491768, + "loss": 0.2295, + "step": 21890 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021615364515286124, + "loss": 0.2837, + "step": 21900 + }, + { + "epoch": 1.42, + "learning_rate": 0.0002161144499608048, + "loss": 0.3115, + "step": 21910 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021607525476874837, + "loss": 0.2258, + "step": 21920 + }, + { + "epoch": 1.42, + "learning_rate": 0.00021603605957669193, + "loss": 0.2077, + "step": 21930 + }, + { + "epoch": 1.42, + "learning_rate": 0.0002159968643846355, + "loss": 0.2272, + "step": 21940 + }, + { + "epoch": 1.42, + "learning_rate": 0.000215957669192579, + "loss": 0.2435, + "step": 21950 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021591847400052256, + "loss": 0.3562, + "step": 21960 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021587927880846613, + "loss": 0.2041, + "step": 21970 + }, + { + "epoch": 1.43, + "learning_rate": 0.0002158400836164097, + "loss": 0.2195, + "step": 21980 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021580088842435325, + "loss": 0.2401, + "step": 21990 + }, + { + "epoch": 1.43, + "learning_rate": 0.0002157616932322968, + "loss": 0.2054, + "step": 22000 + }, + { + "epoch": 1.43, + "eval_loss": 0.2840675711631775, + "eval_runtime": 987.1934, + "eval_samples_per_second": 15.79, + "eval_steps_per_second": 1.974, + "eval_wer": 0.2732868047362511, + "step": 22000 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021572249804024037, + "loss": 0.3018, + "step": 22010 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021568330284818394, + "loss": 0.2105, + "step": 22020 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021564410765612747, + "loss": 0.2089, + "step": 22030 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021560491246407103, + "loss": 0.203, + "step": 22040 + }, + { + "epoch": 1.43, + "learning_rate": 0.0002155657172720146, + "loss": 0.2838, + "step": 22050 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021552652207995816, + "loss": 0.2864, + "step": 22060 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021548732688790172, + "loss": 0.2148, + "step": 22070 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021544813169584528, + "loss": 0.209, + "step": 22080 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021540893650378885, + "loss": 0.2519, + "step": 22090 + }, + { + "epoch": 1.43, + "learning_rate": 0.0002153697413117324, + "loss": 0.2675, + "step": 22100 + }, + { + "epoch": 1.43, + "learning_rate": 0.00021533054611967597, + "loss": 0.302, + "step": 22110 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021529135092761953, + "loss": 0.208, + "step": 22120 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021525215573556307, + "loss": 0.2181, + "step": 22130 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021521296054350663, + "loss": 0.2443, + "step": 22140 + }, + { + "epoch": 1.44, + "learning_rate": 0.0002151737653514502, + "loss": 0.261, + "step": 22150 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021513457015939376, + "loss": 0.3239, + "step": 22160 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021509537496733732, + "loss": 0.188, + "step": 22170 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021505617977528088, + "loss": 0.2201, + "step": 22180 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021501698458322444, + "loss": 0.2709, + "step": 22190 + }, + { + "epoch": 1.44, + "learning_rate": 0.000214977789391168, + "loss": 0.2722, + "step": 22200 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021493859419911157, + "loss": 0.3265, + "step": 22210 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021489939900705513, + "loss": 0.2075, + "step": 22220 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021486020381499866, + "loss": 0.2428, + "step": 22230 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021482100862294223, + "loss": 0.2707, + "step": 22240 + }, + { + "epoch": 1.44, + "learning_rate": 0.0002147818134308858, + "loss": 0.2246, + "step": 22250 + }, + { + "epoch": 1.44, + "learning_rate": 0.00021474261823882935, + "loss": 0.2856, + "step": 22260 + }, + { + "epoch": 1.45, + "learning_rate": 0.0002147034230467729, + "loss": 0.2354, + "step": 22270 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021466422785471648, + "loss": 0.1937, + "step": 22280 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021462503266266004, + "loss": 0.2545, + "step": 22290 + }, + { + "epoch": 1.45, + "learning_rate": 0.0002145858374706036, + "loss": 0.2273, + "step": 22300 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021454664227854716, + "loss": 0.2721, + "step": 22310 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021450744708649072, + "loss": 0.1883, + "step": 22320 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021446825189443426, + "loss": 0.1994, + "step": 22330 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021442905670237782, + "loss": 0.2338, + "step": 22340 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021438986151032138, + "loss": 0.2653, + "step": 22350 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021435066631826495, + "loss": 0.2712, + "step": 22360 + }, + { + "epoch": 1.45, + "learning_rate": 0.0002143114711262085, + "loss": 0.185, + "step": 22370 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021427227593415207, + "loss": 0.2051, + "step": 22380 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021423308074209563, + "loss": 0.2272, + "step": 22390 + }, + { + "epoch": 1.45, + "learning_rate": 0.0002141938855500392, + "loss": 0.2186, + "step": 22400 + }, + { + "epoch": 1.45, + "learning_rate": 0.00021415469035798276, + "loss": 0.2848, + "step": 22410 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021411549516592632, + "loss": 0.2172, + "step": 22420 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021407629997386986, + "loss": 0.2085, + "step": 22430 + }, + { + "epoch": 1.46, + "learning_rate": 0.0002140371047818134, + "loss": 0.2173, + "step": 22440 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021399790958975695, + "loss": 0.2411, + "step": 22450 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021395871439770052, + "loss": 0.3295, + "step": 22460 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021391951920564408, + "loss": 0.1868, + "step": 22470 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021388032401358764, + "loss": 0.2132, + "step": 22480 + }, + { + "epoch": 1.46, + "learning_rate": 0.0002138411288215312, + "loss": 0.2004, + "step": 22490 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021380193362947476, + "loss": 0.2673, + "step": 22500 + }, + { + "epoch": 1.46, + "learning_rate": 0.0002137627384374183, + "loss": 0.2756, + "step": 22510 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021372354324536186, + "loss": 0.1825, + "step": 22520 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021368434805330542, + "loss": 0.2285, + "step": 22530 + }, + { + "epoch": 1.46, + "learning_rate": 0.000213645152861249, + "loss": 0.2204, + "step": 22540 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021360595766919255, + "loss": 0.2765, + "step": 22550 + }, + { + "epoch": 1.46, + "learning_rate": 0.0002135667624771361, + "loss": 0.343, + "step": 22560 + }, + { + "epoch": 1.46, + "learning_rate": 0.00021352756728507967, + "loss": 0.1924, + "step": 22570 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021348837209302324, + "loss": 0.2173, + "step": 22580 + }, + { + "epoch": 1.47, + "learning_rate": 0.0002134491769009668, + "loss": 0.2096, + "step": 22590 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021340998170891036, + "loss": 0.2535, + "step": 22600 + }, + { + "epoch": 1.47, + "learning_rate": 0.0002133707865168539, + "loss": 0.3144, + "step": 22610 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021333159132479746, + "loss": 0.1933, + "step": 22620 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021329239613274102, + "loss": 0.1928, + "step": 22630 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021325320094068458, + "loss": 0.2476, + "step": 22640 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021321400574862814, + "loss": 0.2247, + "step": 22650 + }, + { + "epoch": 1.47, + "learning_rate": 0.0002131748105565717, + "loss": 0.3301, + "step": 22660 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021313561536451527, + "loss": 0.194, + "step": 22670 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021309642017245883, + "loss": 0.2148, + "step": 22680 + }, + { + "epoch": 1.47, + "learning_rate": 0.0002130572249804024, + "loss": 0.2092, + "step": 22690 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021301802978834596, + "loss": 0.2809, + "step": 22700 + }, + { + "epoch": 1.47, + "learning_rate": 0.0002129788345962895, + "loss": 0.2866, + "step": 22710 + }, + { + "epoch": 1.47, + "learning_rate": 0.00021293963940423305, + "loss": 0.1781, + "step": 22720 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021290044421217662, + "loss": 0.1998, + "step": 22730 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021286124902012018, + "loss": 0.2639, + "step": 22740 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021282205382806374, + "loss": 0.2587, + "step": 22750 + }, + { + "epoch": 1.48, + "learning_rate": 0.0002127828586360073, + "loss": 0.3512, + "step": 22760 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021274366344395087, + "loss": 0.2279, + "step": 22770 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021270446825189443, + "loss": 0.2144, + "step": 22780 + }, + { + "epoch": 1.48, + "learning_rate": 0.000212665273059838, + "loss": 0.2154, + "step": 22790 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021262607786778155, + "loss": 0.2626, + "step": 22800 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021258688267572511, + "loss": 0.3018, + "step": 22810 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021254768748366865, + "loss": 0.1952, + "step": 22820 + }, + { + "epoch": 1.48, + "learning_rate": 0.0002125084922916122, + "loss": 0.1863, + "step": 22830 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021246929709955577, + "loss": 0.2205, + "step": 22840 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021243010190749934, + "loss": 0.2461, + "step": 22850 + }, + { + "epoch": 1.48, + "learning_rate": 0.0002123909067154429, + "loss": 0.3044, + "step": 22860 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021235171152338646, + "loss": 0.1788, + "step": 22870 + }, + { + "epoch": 1.48, + "learning_rate": 0.00021231251633133002, + "loss": 0.2365, + "step": 22880 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021227332113927359, + "loss": 0.2559, + "step": 22890 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021223412594721715, + "loss": 0.275, + "step": 22900 + }, + { + "epoch": 1.49, + "learning_rate": 0.0002121949307551607, + "loss": 0.283, + "step": 22910 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021215573556310425, + "loss": 0.2128, + "step": 22920 + }, + { + "epoch": 1.49, + "learning_rate": 0.0002121165403710478, + "loss": 0.1974, + "step": 22930 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021207734517899134, + "loss": 0.2439, + "step": 22940 + }, + { + "epoch": 1.49, + "learning_rate": 0.0002120381499869349, + "loss": 0.2543, + "step": 22950 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021199895479487847, + "loss": 0.305, + "step": 22960 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021195975960282203, + "loss": 0.1991, + "step": 22970 + }, + { + "epoch": 1.49, + "learning_rate": 0.0002119205644107656, + "loss": 0.2342, + "step": 22980 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021188136921870913, + "loss": 0.2199, + "step": 22990 + }, + { + "epoch": 1.49, + "learning_rate": 0.0002118421740266527, + "loss": 0.2655, + "step": 23000 + }, + { + "epoch": 1.49, + "eval_loss": 0.27194133400917053, + "eval_runtime": 968.7719, + "eval_samples_per_second": 16.09, + "eval_steps_per_second": 2.012, + "eval_wer": 0.26282497761225854, + "step": 23000 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021180297883459625, + "loss": 0.3146, + "step": 23010 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021176378364253981, + "loss": 0.1764, + "step": 23020 + }, + { + "epoch": 1.49, + "learning_rate": 0.00021172458845048338, + "loss": 0.2253, + "step": 23030 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021168539325842694, + "loss": 0.2363, + "step": 23040 + }, + { + "epoch": 1.5, + "learning_rate": 0.0002116461980663705, + "loss": 0.2277, + "step": 23050 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021160700287431406, + "loss": 0.3029, + "step": 23060 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021156780768225763, + "loss": 0.186, + "step": 23070 + }, + { + "epoch": 1.5, + "learning_rate": 0.0002115286124902012, + "loss": 0.2152, + "step": 23080 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021148941729814475, + "loss": 0.2472, + "step": 23090 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021145022210608829, + "loss": 0.2586, + "step": 23100 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021141102691403185, + "loss": 0.3437, + "step": 23110 + }, + { + "epoch": 1.5, + "learning_rate": 0.0002113718317219754, + "loss": 0.211, + "step": 23120 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021133263652991897, + "loss": 0.18, + "step": 23130 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021129344133786253, + "loss": 0.2083, + "step": 23140 + }, + { + "epoch": 1.5, + "learning_rate": 0.0002112542461458061, + "loss": 0.2746, + "step": 23150 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021121505095374966, + "loss": 0.357, + "step": 23160 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021117585576169322, + "loss": 0.1925, + "step": 23170 + }, + { + "epoch": 1.5, + "learning_rate": 0.00021113666056963678, + "loss": 0.2146, + "step": 23180 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021109746537758035, + "loss": 0.231, + "step": 23190 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021105827018552388, + "loss": 0.3186, + "step": 23200 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021101907499346744, + "loss": 0.3226, + "step": 23210 + }, + { + "epoch": 1.51, + "learning_rate": 0.000210979879801411, + "loss": 0.1837, + "step": 23220 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021094068460935457, + "loss": 0.2205, + "step": 23230 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021090148941729813, + "loss": 0.2281, + "step": 23240 + }, + { + "epoch": 1.51, + "learning_rate": 0.0002108622942252417, + "loss": 0.2544, + "step": 23250 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021082309903318525, + "loss": 0.3265, + "step": 23260 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021078390384112882, + "loss": 0.2146, + "step": 23270 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021074470864907238, + "loss": 0.2225, + "step": 23280 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021070551345701594, + "loss": 0.2427, + "step": 23290 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021066631826495948, + "loss": 0.2679, + "step": 23300 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021062712307290304, + "loss": 0.3418, + "step": 23310 + }, + { + "epoch": 1.51, + "learning_rate": 0.0002105879278808466, + "loss": 0.188, + "step": 23320 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021054873268879016, + "loss": 0.2038, + "step": 23330 + }, + { + "epoch": 1.51, + "learning_rate": 0.00021050953749673373, + "loss": 0.2887, + "step": 23340 + }, + { + "epoch": 1.52, + "learning_rate": 0.0002104703423046773, + "loss": 0.216, + "step": 23350 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021043114711262085, + "loss": 0.3165, + "step": 23360 + }, + { + "epoch": 1.52, + "learning_rate": 0.0002103919519205644, + "loss": 0.1912, + "step": 23370 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021035275672850798, + "loss": 0.1915, + "step": 23380 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021031356153645154, + "loss": 0.2367, + "step": 23390 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021027436634439507, + "loss": 0.2475, + "step": 23400 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021023517115233863, + "loss": 0.3583, + "step": 23410 + }, + { + "epoch": 1.52, + "learning_rate": 0.0002101959759602822, + "loss": 0.2045, + "step": 23420 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021015678076822573, + "loss": 0.2043, + "step": 23430 + }, + { + "epoch": 1.52, + "learning_rate": 0.0002101175855761693, + "loss": 0.2325, + "step": 23440 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021007839038411286, + "loss": 0.2208, + "step": 23450 + }, + { + "epoch": 1.52, + "learning_rate": 0.00021003919519205642, + "loss": 0.3069, + "step": 23460 + }, + { + "epoch": 1.52, + "learning_rate": 0.00020999999999999998, + "loss": 0.1917, + "step": 23470 + }, + { + "epoch": 1.52, + "learning_rate": 0.00020996080480794352, + "loss": 0.2183, + "step": 23480 + }, + { + "epoch": 1.52, + "learning_rate": 0.00020992160961588708, + "loss": 0.2625, + "step": 23490 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020988241442383064, + "loss": 0.2621, + "step": 23500 + }, + { + "epoch": 1.53, + "learning_rate": 0.0002098432192317742, + "loss": 0.2774, + "step": 23510 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020980402403971777, + "loss": 0.1928, + "step": 23520 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020976482884766133, + "loss": 0.2463, + "step": 23530 + }, + { + "epoch": 1.53, + "learning_rate": 0.0002097256336556049, + "loss": 0.2539, + "step": 23540 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020968643846354845, + "loss": 0.2299, + "step": 23550 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020964724327149202, + "loss": 0.2902, + "step": 23560 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020960804807943558, + "loss": 0.1652, + "step": 23570 + }, + { + "epoch": 1.53, + "learning_rate": 0.0002095688528873791, + "loss": 0.2166, + "step": 23580 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020952965769532267, + "loss": 0.2639, + "step": 23590 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020949046250326624, + "loss": 0.2529, + "step": 23600 + }, + { + "epoch": 1.53, + "learning_rate": 0.0002094512673112098, + "loss": 0.3, + "step": 23610 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020941207211915336, + "loss": 0.2024, + "step": 23620 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020937287692709692, + "loss": 0.2183, + "step": 23630 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020933368173504049, + "loss": 0.2119, + "step": 23640 + }, + { + "epoch": 1.53, + "learning_rate": 0.00020929448654298405, + "loss": 0.2405, + "step": 23650 + }, + { + "epoch": 1.54, + "learning_rate": 0.0002092552913509276, + "loss": 0.3101, + "step": 23660 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020921609615887117, + "loss": 0.1999, + "step": 23670 + }, + { + "epoch": 1.54, + "learning_rate": 0.0002091769009668147, + "loss": 0.2097, + "step": 23680 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020913770577475827, + "loss": 0.2426, + "step": 23690 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020909851058270183, + "loss": 0.2143, + "step": 23700 + }, + { + "epoch": 1.54, + "learning_rate": 0.0002090593153906454, + "loss": 0.3419, + "step": 23710 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020902012019858896, + "loss": 0.2006, + "step": 23720 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020898092500653252, + "loss": 0.195, + "step": 23730 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020894172981447608, + "loss": 0.2591, + "step": 23740 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020890253462241964, + "loss": 0.2587, + "step": 23750 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020886725894956882, + "loss": 0.2439, + "step": 23760 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020882806375751239, + "loss": 0.2059, + "step": 23770 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020878886856545595, + "loss": 0.2362, + "step": 23780 + }, + { + "epoch": 1.54, + "learning_rate": 0.0002087496733733995, + "loss": 0.2453, + "step": 23790 + }, + { + "epoch": 1.54, + "learning_rate": 0.00020871047818134307, + "loss": 0.2576, + "step": 23800 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020867128298928663, + "loss": 0.3113, + "step": 23810 + }, + { + "epoch": 1.55, + "learning_rate": 0.0002086320877972302, + "loss": 0.21, + "step": 23820 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020859289260517373, + "loss": 0.2567, + "step": 23830 + }, + { + "epoch": 1.55, + "learning_rate": 0.0002085536974131173, + "loss": 0.2292, + "step": 23840 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020851450222106086, + "loss": 0.277, + "step": 23850 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020847530702900442, + "loss": 0.3495, + "step": 23860 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020843611183694798, + "loss": 0.2158, + "step": 23870 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020839691664489154, + "loss": 0.1963, + "step": 23880 + }, + { + "epoch": 1.55, + "learning_rate": 0.0002083577214528351, + "loss": 0.2422, + "step": 23890 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020831852626077867, + "loss": 0.2544, + "step": 23900 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020827933106872223, + "loss": 0.3036, + "step": 23910 + }, + { + "epoch": 1.55, + "learning_rate": 0.0002082401358766658, + "loss": 0.2287, + "step": 23920 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020820094068460933, + "loss": 0.2233, + "step": 23930 + }, + { + "epoch": 1.55, + "learning_rate": 0.0002081617454925529, + "loss": 0.2336, + "step": 23940 + }, + { + "epoch": 1.55, + "learning_rate": 0.00020812255030049645, + "loss": 0.2679, + "step": 23950 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020808335510844002, + "loss": 0.3454, + "step": 23960 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020804415991638358, + "loss": 0.2213, + "step": 23970 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020800496472432714, + "loss": 0.1911, + "step": 23980 + }, + { + "epoch": 1.56, + "learning_rate": 0.0002079657695322707, + "loss": 0.2012, + "step": 23990 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020792657434021426, + "loss": 0.2843, + "step": 24000 + }, + { + "epoch": 1.56, + "eval_loss": 0.27305540442466736, + "eval_runtime": 991.8968, + "eval_samples_per_second": 15.715, + "eval_steps_per_second": 1.965, + "eval_wer": 0.2666628761495928, + "step": 24000 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020788737914815783, + "loss": 0.2724, + "step": 24010 + }, + { + "epoch": 1.56, + "learning_rate": 0.0002078481839561014, + "loss": 0.2098, + "step": 24020 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020780898876404492, + "loss": 0.2091, + "step": 24030 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020776979357198849, + "loss": 0.2277, + "step": 24040 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020773059837993205, + "loss": 0.2263, + "step": 24050 + }, + { + "epoch": 1.56, + "learning_rate": 0.0002076914031878756, + "loss": 0.2876, + "step": 24060 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020765220799581917, + "loss": 0.2081, + "step": 24070 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020761301280376274, + "loss": 0.2294, + "step": 24080 + }, + { + "epoch": 1.56, + "learning_rate": 0.0002075738176117063, + "loss": 0.2261, + "step": 24090 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020753462241964986, + "loss": 0.2384, + "step": 24100 + }, + { + "epoch": 1.56, + "learning_rate": 0.00020749542722759337, + "loss": 0.3047, + "step": 24110 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020745623203553693, + "loss": 0.2003, + "step": 24120 + }, + { + "epoch": 1.57, + "learning_rate": 0.0002074170368434805, + "loss": 0.2013, + "step": 24130 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020737784165142405, + "loss": 0.1938, + "step": 24140 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020733864645936762, + "loss": 0.2355, + "step": 24150 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020729945126731118, + "loss": 0.2754, + "step": 24160 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020726025607525474, + "loss": 0.1966, + "step": 24170 + }, + { + "epoch": 1.57, + "learning_rate": 0.0002072210608831983, + "loss": 0.1902, + "step": 24180 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020718186569114187, + "loss": 0.2336, + "step": 24190 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020714267049908543, + "loss": 0.2528, + "step": 24200 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020710347530702896, + "loss": 0.2845, + "step": 24210 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020706428011497253, + "loss": 0.2108, + "step": 24220 + }, + { + "epoch": 1.57, + "learning_rate": 0.0002070250849229161, + "loss": 0.1953, + "step": 24230 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020698588973085965, + "loss": 0.235, + "step": 24240 + }, + { + "epoch": 1.57, + "learning_rate": 0.0002069466945388032, + "loss": 0.2682, + "step": 24250 + }, + { + "epoch": 1.57, + "learning_rate": 0.00020690749934674678, + "loss": 0.2979, + "step": 24260 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020686830415469034, + "loss": 0.1856, + "step": 24270 + }, + { + "epoch": 1.58, + "learning_rate": 0.0002068291089626339, + "loss": 0.2096, + "step": 24280 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020678991377057746, + "loss": 0.2328, + "step": 24290 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020675071857852102, + "loss": 0.2278, + "step": 24300 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020671152338646456, + "loss": 0.3341, + "step": 24310 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020667232819440812, + "loss": 0.1825, + "step": 24320 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020663313300235168, + "loss": 0.1938, + "step": 24330 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020659393781029525, + "loss": 0.2146, + "step": 24340 + }, + { + "epoch": 1.58, + "learning_rate": 0.0002065547426182388, + "loss": 0.237, + "step": 24350 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020651554742618237, + "loss": 0.313, + "step": 24360 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020647635223412593, + "loss": 0.2266, + "step": 24370 + }, + { + "epoch": 1.58, + "learning_rate": 0.0002064371570420695, + "loss": 0.2248, + "step": 24380 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020639796185001306, + "loss": 0.2299, + "step": 24390 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020635876665795662, + "loss": 0.2261, + "step": 24400 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020631957146590016, + "loss": 0.3033, + "step": 24410 + }, + { + "epoch": 1.58, + "learning_rate": 0.00020628037627384372, + "loss": 0.1977, + "step": 24420 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020624118108178728, + "loss": 0.2295, + "step": 24430 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020620198588973084, + "loss": 0.2212, + "step": 24440 + }, + { + "epoch": 1.59, + "learning_rate": 0.0002061627906976744, + "loss": 0.2587, + "step": 24450 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020612359550561797, + "loss": 0.293, + "step": 24460 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020608440031356153, + "loss": 0.201, + "step": 24470 + }, + { + "epoch": 1.59, + "learning_rate": 0.0002060452051215051, + "loss": 0.1713, + "step": 24480 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020600600992944865, + "loss": 0.216, + "step": 24490 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020596681473739222, + "loss": 0.25, + "step": 24500 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020592761954533575, + "loss": 0.2871, + "step": 24510 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020588842435327931, + "loss": 0.2052, + "step": 24520 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020584922916122288, + "loss": 0.2049, + "step": 24530 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020581003396916644, + "loss": 0.249, + "step": 24540 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020577083877711, + "loss": 0.2475, + "step": 24550 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020573164358505356, + "loss": 0.2998, + "step": 24560 + }, + { + "epoch": 1.59, + "learning_rate": 0.00020569244839299713, + "loss": 0.1846, + "step": 24570 + }, + { + "epoch": 1.6, + "learning_rate": 0.0002056532532009407, + "loss": 0.1913, + "step": 24580 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020561405800888425, + "loss": 0.1768, + "step": 24590 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020557486281682776, + "loss": 0.2844, + "step": 24600 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020553566762477132, + "loss": 0.319, + "step": 24610 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020549647243271488, + "loss": 0.2105, + "step": 24620 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020545727724065844, + "loss": 0.2026, + "step": 24630 + }, + { + "epoch": 1.6, + "learning_rate": 0.000205418082048602, + "loss": 0.2209, + "step": 24640 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020537888685654557, + "loss": 0.2635, + "step": 24650 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020533969166448913, + "loss": 0.2836, + "step": 24660 + }, + { + "epoch": 1.6, + "learning_rate": 0.0002053004964724327, + "loss": 0.2021, + "step": 24670 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020526130128037626, + "loss": 0.1911, + "step": 24680 + }, + { + "epoch": 1.6, + "learning_rate": 0.0002052221060883198, + "loss": 0.1963, + "step": 24690 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020518291089626335, + "loss": 0.2462, + "step": 24700 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020514371570420692, + "loss": 0.296, + "step": 24710 + }, + { + "epoch": 1.6, + "learning_rate": 0.00020510452051215048, + "loss": 0.2227, + "step": 24720 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020506532532009404, + "loss": 0.2127, + "step": 24730 + }, + { + "epoch": 1.61, + "learning_rate": 0.0002050261301280376, + "loss": 0.1957, + "step": 24740 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020498693493598116, + "loss": 0.2415, + "step": 24750 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020494773974392473, + "loss": 0.3072, + "step": 24760 + }, + { + "epoch": 1.61, + "learning_rate": 0.0002049085445518683, + "loss": 0.2114, + "step": 24770 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020486934935981185, + "loss": 0.2213, + "step": 24780 + }, + { + "epoch": 1.61, + "learning_rate": 0.0002048301541677554, + "loss": 0.2376, + "step": 24790 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020479095897569895, + "loss": 0.2658, + "step": 24800 + }, + { + "epoch": 1.61, + "learning_rate": 0.0002047517637836425, + "loss": 0.302, + "step": 24810 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020471256859158607, + "loss": 0.1981, + "step": 24820 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020467337339952964, + "loss": 0.1758, + "step": 24830 + }, + { + "epoch": 1.61, + "learning_rate": 0.0002046341782074732, + "loss": 0.2072, + "step": 24840 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020459498301541676, + "loss": 0.2186, + "step": 24850 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020455578782336032, + "loss": 0.3148, + "step": 24860 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020451659263130389, + "loss": 0.1839, + "step": 24870 + }, + { + "epoch": 1.61, + "learning_rate": 0.00020447739743924745, + "loss": 0.2314, + "step": 24880 + }, + { + "epoch": 1.62, + "learning_rate": 0.000204438202247191, + "loss": 0.2433, + "step": 24890 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020439900705513455, + "loss": 0.295, + "step": 24900 + }, + { + "epoch": 1.62, + "learning_rate": 0.0002043598118630781, + "loss": 0.2797, + "step": 24910 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020432061667102167, + "loss": 0.2141, + "step": 24920 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020428142147896523, + "loss": 0.21, + "step": 24930 + }, + { + "epoch": 1.62, + "learning_rate": 0.0002042422262869088, + "loss": 0.1924, + "step": 24940 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020420303109485236, + "loss": 0.2561, + "step": 24950 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020416383590279592, + "loss": 0.3222, + "step": 24960 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020412464071073948, + "loss": 0.2066, + "step": 24970 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020408544551868304, + "loss": 0.227, + "step": 24980 + }, + { + "epoch": 1.62, + "learning_rate": 0.0002040462503266266, + "loss": 0.2454, + "step": 24990 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020400705513457014, + "loss": 0.2253, + "step": 25000 + }, + { + "epoch": 1.62, + "eval_loss": 0.26669734716415405, + "eval_runtime": 1009.1423, + "eval_samples_per_second": 15.447, + "eval_steps_per_second": 1.931, + "eval_wer": 0.26266151156344614, + "step": 25000 + }, + { + "epoch": 1.62, + "learning_rate": 0.0002039678599425137, + "loss": 0.3107, + "step": 25010 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020392866475045727, + "loss": 0.1855, + "step": 25020 + }, + { + "epoch": 1.62, + "learning_rate": 0.00020388946955840083, + "loss": 0.1753, + "step": 25030 + }, + { + "epoch": 1.63, + "learning_rate": 0.0002038502743663444, + "loss": 0.2353, + "step": 25040 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020381107917428795, + "loss": 0.2289, + "step": 25050 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020377188398223151, + "loss": 0.3311, + "step": 25060 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020373268879017508, + "loss": 0.1946, + "step": 25070 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020369349359811864, + "loss": 0.2273, + "step": 25080 + }, + { + "epoch": 1.63, + "learning_rate": 0.0002036542984060622, + "loss": 0.2536, + "step": 25090 + }, + { + "epoch": 1.63, + "learning_rate": 0.0002036151032140057, + "loss": 0.2209, + "step": 25100 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020357590802194927, + "loss": 0.2807, + "step": 25110 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020353671282989283, + "loss": 0.2196, + "step": 25120 + }, + { + "epoch": 1.63, + "learning_rate": 0.0002034975176378364, + "loss": 0.2174, + "step": 25130 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020345832244577996, + "loss": 0.2246, + "step": 25140 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020341912725372352, + "loss": 0.2608, + "step": 25150 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020337993206166708, + "loss": 0.2852, + "step": 25160 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020334073686961062, + "loss": 0.1771, + "step": 25170 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020330154167755418, + "loss": 0.2049, + "step": 25180 + }, + { + "epoch": 1.63, + "learning_rate": 0.00020326234648549774, + "loss": 0.2021, + "step": 25190 + }, + { + "epoch": 1.64, + "learning_rate": 0.0002032231512934413, + "loss": 0.2703, + "step": 25200 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020318395610138487, + "loss": 0.2916, + "step": 25210 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020314476090932843, + "loss": 0.1897, + "step": 25220 + }, + { + "epoch": 1.64, + "learning_rate": 0.000203105565717272, + "loss": 0.2058, + "step": 25230 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020306637052521555, + "loss": 0.2519, + "step": 25240 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020302717533315912, + "loss": 0.203, + "step": 25250 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020298798014110268, + "loss": 0.317, + "step": 25260 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020294878494904624, + "loss": 0.2039, + "step": 25270 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020290958975698978, + "loss": 0.2025, + "step": 25280 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020287039456493334, + "loss": 0.2141, + "step": 25290 + }, + { + "epoch": 1.64, + "learning_rate": 0.0002028311993728769, + "loss": 0.2538, + "step": 25300 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020279200418082046, + "loss": 0.2725, + "step": 25310 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020275280898876403, + "loss": 0.2045, + "step": 25320 + }, + { + "epoch": 1.64, + "learning_rate": 0.0002027136137967076, + "loss": 0.1869, + "step": 25330 + }, + { + "epoch": 1.64, + "learning_rate": 0.00020267441860465115, + "loss": 0.2339, + "step": 25340 + }, + { + "epoch": 1.65, + "learning_rate": 0.0002026352234125947, + "loss": 0.2882, + "step": 25350 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020259602822053827, + "loss": 0.2894, + "step": 25360 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020255683302848184, + "loss": 0.2018, + "step": 25370 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020251763783642537, + "loss": 0.1852, + "step": 25380 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020247844264436893, + "loss": 0.2199, + "step": 25390 + }, + { + "epoch": 1.65, + "learning_rate": 0.0002024392474523125, + "loss": 0.2382, + "step": 25400 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020240005226025606, + "loss": 0.3315, + "step": 25410 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020236085706819962, + "loss": 0.2075, + "step": 25420 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020232166187614318, + "loss": 0.1996, + "step": 25430 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020228246668408675, + "loss": 0.2101, + "step": 25440 + }, + { + "epoch": 1.65, + "learning_rate": 0.0002022432714920303, + "loss": 0.2689, + "step": 25450 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020220407629997387, + "loss": 0.2509, + "step": 25460 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020216488110791743, + "loss": 0.1932, + "step": 25470 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020212568591586097, + "loss": 0.2444, + "step": 25480 + }, + { + "epoch": 1.65, + "learning_rate": 0.00020208649072380453, + "loss": 0.2055, + "step": 25490 + }, + { + "epoch": 1.65, + "learning_rate": 0.0002020472955317481, + "loss": 0.2312, + "step": 25500 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020200810033969166, + "loss": 0.293, + "step": 25510 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020196890514763522, + "loss": 0.224, + "step": 25520 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020192970995557878, + "loss": 0.195, + "step": 25530 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020189051476352234, + "loss": 0.238, + "step": 25540 + }, + { + "epoch": 1.66, + "learning_rate": 0.0002018513195714659, + "loss": 0.2458, + "step": 25550 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020181212437940947, + "loss": 0.2971, + "step": 25560 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020177292918735303, + "loss": 0.2001, + "step": 25570 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020173373399529656, + "loss": 0.2171, + "step": 25580 + }, + { + "epoch": 1.66, + "learning_rate": 0.0002016945388032401, + "loss": 0.2728, + "step": 25590 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020165534361118366, + "loss": 0.2522, + "step": 25600 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020161614841912722, + "loss": 0.2714, + "step": 25610 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020157695322707079, + "loss": 0.2197, + "step": 25620 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020153775803501435, + "loss": 0.1956, + "step": 25630 + }, + { + "epoch": 1.66, + "learning_rate": 0.0002014985628429579, + "loss": 0.208, + "step": 25640 + }, + { + "epoch": 1.66, + "learning_rate": 0.00020145936765090147, + "loss": 0.2553, + "step": 25650 + }, + { + "epoch": 1.67, + "learning_rate": 0.000201420172458845, + "loss": 0.3373, + "step": 25660 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020138097726678857, + "loss": 0.181, + "step": 25670 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020134178207473213, + "loss": 0.2053, + "step": 25680 + }, + { + "epoch": 1.67, + "learning_rate": 0.0002013025868826757, + "loss": 0.2213, + "step": 25690 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020126339169061926, + "loss": 0.2289, + "step": 25700 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020122419649856282, + "loss": 0.3218, + "step": 25710 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020118500130650638, + "loss": 0.1856, + "step": 25720 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020114580611444994, + "loss": 0.2087, + "step": 25730 + }, + { + "epoch": 1.67, + "learning_rate": 0.0002011066109223935, + "loss": 0.2431, + "step": 25740 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020106741573033707, + "loss": 0.2326, + "step": 25750 + }, + { + "epoch": 1.67, + "learning_rate": 0.0002010282205382806, + "loss": 0.2606, + "step": 25760 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020098902534622417, + "loss": 0.2061, + "step": 25770 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020094983015416773, + "loss": 0.2019, + "step": 25780 + }, + { + "epoch": 1.67, + "learning_rate": 0.0002009106349621113, + "loss": 0.2087, + "step": 25790 + }, + { + "epoch": 1.67, + "learning_rate": 0.00020087143977005485, + "loss": 0.2569, + "step": 25800 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020083224457799842, + "loss": 0.2848, + "step": 25810 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020079304938594198, + "loss": 0.2258, + "step": 25820 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020075385419388554, + "loss": 0.2098, + "step": 25830 + }, + { + "epoch": 1.68, + "learning_rate": 0.0002007146590018291, + "loss": 0.2511, + "step": 25840 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020067546380977266, + "loss": 0.2641, + "step": 25850 + }, + { + "epoch": 1.68, + "learning_rate": 0.0002006362686177162, + "loss": 0.2972, + "step": 25860 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020059707342565976, + "loss": 0.197, + "step": 25870 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020055787823360332, + "loss": 0.209, + "step": 25880 + }, + { + "epoch": 1.68, + "learning_rate": 0.0002005186830415469, + "loss": 0.2231, + "step": 25890 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020047948784949045, + "loss": 0.2718, + "step": 25900 + }, + { + "epoch": 1.68, + "learning_rate": 0.000200440292657434, + "loss": 0.2829, + "step": 25910 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020040109746537757, + "loss": 0.1792, + "step": 25920 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020036190227332114, + "loss": 0.2155, + "step": 25930 + }, + { + "epoch": 1.68, + "learning_rate": 0.0002003227070812647, + "loss": 0.2109, + "step": 25940 + }, + { + "epoch": 1.68, + "learning_rate": 0.00020028351188920826, + "loss": 0.2085, + "step": 25950 + }, + { + "epoch": 1.68, + "learning_rate": 0.0002002443166971518, + "loss": 0.3271, + "step": 25960 + }, + { + "epoch": 1.69, + "learning_rate": 0.00020020512150509536, + "loss": 0.1755, + "step": 25970 + }, + { + "epoch": 1.69, + "learning_rate": 0.00020016592631303892, + "loss": 0.2095, + "step": 25980 + }, + { + "epoch": 1.69, + "learning_rate": 0.00020012673112098248, + "loss": 0.2282, + "step": 25990 + }, + { + "epoch": 1.69, + "learning_rate": 0.00020008753592892604, + "loss": 0.2495, + "step": 26000 + }, + { + "epoch": 1.69, + "eval_loss": 0.2659270167350769, + "eval_runtime": 1015.4445, + "eval_samples_per_second": 15.351, + "eval_steps_per_second": 1.919, + "eval_wer": 0.2645662463930861, + "step": 26000 + }, + { + "epoch": 1.69, + "learning_rate": 0.0002000483407368696, + "loss": 0.2958, + "step": 26010 + }, + { + "epoch": 1.69, + "learning_rate": 0.00020000914554481317, + "loss": 0.2186, + "step": 26020 + }, + { + "epoch": 1.69, + "learning_rate": 0.00019996995035275673, + "loss": 0.1824, + "step": 26030 + }, + { + "epoch": 1.69, + "learning_rate": 0.0001999307551607003, + "loss": 0.2389, + "step": 26040 + }, + { + "epoch": 1.69, + "learning_rate": 0.00019989155996864386, + "loss": 0.2535, + "step": 26050 + }, + { + "epoch": 1.69, + "learning_rate": 0.0001998523647765874, + "loss": 0.3427, + "step": 26060 + }, + { + "epoch": 1.69, + "learning_rate": 0.00019981316958453095, + "loss": 0.2146, + "step": 26070 + }, + { + "epoch": 1.69, + "learning_rate": 0.0001997739743924745, + "loss": 0.1876, + "step": 26080 + }, + { + "epoch": 1.69, + "learning_rate": 0.00019973477920041805, + "loss": 0.2181, + "step": 26090 + }, + { + "epoch": 1.69, + "learning_rate": 0.0001996955840083616, + "loss": 0.2403, + "step": 26100 + }, + { + "epoch": 1.69, + "learning_rate": 0.00019965638881630518, + "loss": 0.3225, + "step": 26110 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019961719362424874, + "loss": 0.2195, + "step": 26120 + }, + { + "epoch": 1.7, + "learning_rate": 0.0001995779984321923, + "loss": 0.2026, + "step": 26130 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019953880324013584, + "loss": 0.2403, + "step": 26140 + }, + { + "epoch": 1.7, + "learning_rate": 0.0001994996080480794, + "loss": 0.249, + "step": 26150 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019946041285602296, + "loss": 0.3202, + "step": 26160 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019942121766396652, + "loss": 0.1928, + "step": 26170 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019938202247191008, + "loss": 0.1891, + "step": 26180 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019934282727985365, + "loss": 0.2496, + "step": 26190 + }, + { + "epoch": 1.7, + "learning_rate": 0.0001993036320877972, + "loss": 0.2902, + "step": 26200 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019926443689574077, + "loss": 0.261, + "step": 26210 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019922524170368433, + "loss": 0.1914, + "step": 26220 + }, + { + "epoch": 1.7, + "learning_rate": 0.0001991860465116279, + "loss": 0.1986, + "step": 26230 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019914685131957143, + "loss": 0.2227, + "step": 26240 + }, + { + "epoch": 1.7, + "learning_rate": 0.000199107656127515, + "loss": 0.249, + "step": 26250 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019906846093545856, + "loss": 0.3107, + "step": 26260 + }, + { + "epoch": 1.7, + "learning_rate": 0.00019902926574340212, + "loss": 0.1853, + "step": 26270 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019899007055134568, + "loss": 0.2032, + "step": 26280 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019895087535928924, + "loss": 0.2103, + "step": 26290 + }, + { + "epoch": 1.71, + "learning_rate": 0.0001989116801672328, + "loss": 0.2685, + "step": 26300 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019887248497517637, + "loss": 0.3214, + "step": 26310 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019883328978311993, + "loss": 0.2165, + "step": 26320 + }, + { + "epoch": 1.71, + "learning_rate": 0.0001987940945910635, + "loss": 0.2043, + "step": 26330 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019875489939900703, + "loss": 0.2539, + "step": 26340 + }, + { + "epoch": 1.71, + "learning_rate": 0.0001987157042069506, + "loss": 0.2069, + "step": 26350 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019867650901489415, + "loss": 0.2568, + "step": 26360 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019863731382283771, + "loss": 0.22, + "step": 26370 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019859811863078128, + "loss": 0.2074, + "step": 26380 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019855892343872484, + "loss": 0.2026, + "step": 26390 + }, + { + "epoch": 1.71, + "learning_rate": 0.0001985197282466684, + "loss": 0.2229, + "step": 26400 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019848053305461196, + "loss": 0.2714, + "step": 26410 + }, + { + "epoch": 1.71, + "learning_rate": 0.00019844133786255553, + "loss": 0.1945, + "step": 26420 + }, + { + "epoch": 1.72, + "learning_rate": 0.0001984021426704991, + "loss": 0.21, + "step": 26430 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019836294747844262, + "loss": 0.2316, + "step": 26440 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019832375228638619, + "loss": 0.2532, + "step": 26450 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019828455709432975, + "loss": 0.3237, + "step": 26460 + }, + { + "epoch": 1.72, + "learning_rate": 0.0001982453619022733, + "loss": 0.2198, + "step": 26470 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019820616671021687, + "loss": 0.2056, + "step": 26480 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019816697151816043, + "loss": 0.2587, + "step": 26490 + }, + { + "epoch": 1.72, + "learning_rate": 0.000198127776326104, + "loss": 0.2874, + "step": 26500 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019808858113404756, + "loss": 0.3183, + "step": 26510 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019804938594199112, + "loss": 0.192, + "step": 26520 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019801019074993468, + "loss": 0.2001, + "step": 26530 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019797099555787822, + "loss": 0.2151, + "step": 26540 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019793180036582178, + "loss": 0.2328, + "step": 26550 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019789260517376534, + "loss": 0.2795, + "step": 26560 + }, + { + "epoch": 1.72, + "learning_rate": 0.00019785340998170888, + "loss": 0.1903, + "step": 26570 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019781421478965244, + "loss": 0.1887, + "step": 26580 + }, + { + "epoch": 1.73, + "learning_rate": 0.000197775019597596, + "loss": 0.2362, + "step": 26590 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019773582440553957, + "loss": 0.2249, + "step": 26600 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019769662921348313, + "loss": 0.2853, + "step": 26610 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019765743402142666, + "loss": 0.1755, + "step": 26620 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019761823882937023, + "loss": 0.1775, + "step": 26630 + }, + { + "epoch": 1.73, + "learning_rate": 0.0001975790436373138, + "loss": 0.2207, + "step": 26640 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019753984844525735, + "loss": 0.226, + "step": 26650 + }, + { + "epoch": 1.73, + "learning_rate": 0.0001975006532532009, + "loss": 0.355, + "step": 26660 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019746145806114447, + "loss": 0.2287, + "step": 26670 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019742226286908804, + "loss": 0.2089, + "step": 26680 + }, + { + "epoch": 1.73, + "learning_rate": 0.0001973830676770316, + "loss": 0.1986, + "step": 26690 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019734387248497516, + "loss": 0.2521, + "step": 26700 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019730467729291872, + "loss": 0.361, + "step": 26710 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019726548210086226, + "loss": 0.2171, + "step": 26720 + }, + { + "epoch": 1.73, + "learning_rate": 0.00019722628690880582, + "loss": 0.1962, + "step": 26730 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019718709171674938, + "loss": 0.1926, + "step": 26740 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019715181604389862, + "loss": 0.295, + "step": 26750 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019711262085184215, + "loss": 0.2734, + "step": 26760 + }, + { + "epoch": 1.74, + "learning_rate": 0.0001970734256597857, + "loss": 0.2249, + "step": 26770 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019703423046772925, + "loss": 0.2083, + "step": 26780 + }, + { + "epoch": 1.74, + "learning_rate": 0.0001969950352756728, + "loss": 0.2076, + "step": 26790 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019695584008361637, + "loss": 0.2319, + "step": 26800 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019691664489155994, + "loss": 0.2679, + "step": 26810 + }, + { + "epoch": 1.74, + "learning_rate": 0.0001968774496995035, + "loss": 0.1765, + "step": 26820 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019683825450744706, + "loss": 0.2095, + "step": 26830 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019679905931539062, + "loss": 0.2312, + "step": 26840 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019675986412333419, + "loss": 0.2533, + "step": 26850 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019672066893127775, + "loss": 0.2847, + "step": 26860 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019668147373922128, + "loss": 0.1648, + "step": 26870 + }, + { + "epoch": 1.74, + "learning_rate": 0.00019664227854716484, + "loss": 0.2252, + "step": 26880 + }, + { + "epoch": 1.75, + "learning_rate": 0.0001966030833551084, + "loss": 0.1833, + "step": 26890 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019656388816305197, + "loss": 0.2645, + "step": 26900 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019652469297099553, + "loss": 0.29, + "step": 26910 + }, + { + "epoch": 1.75, + "learning_rate": 0.0001964854977789391, + "loss": 0.2033, + "step": 26920 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019644630258688266, + "loss": 0.2132, + "step": 26930 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019640710739482622, + "loss": 0.2044, + "step": 26940 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019636791220276978, + "loss": 0.2515, + "step": 26950 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019632871701071334, + "loss": 0.3175, + "step": 26960 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019628952181865688, + "loss": 0.2081, + "step": 26970 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019625032662660044, + "loss": 0.2243, + "step": 26980 + }, + { + "epoch": 1.75, + "learning_rate": 0.000196211131434544, + "loss": 0.1984, + "step": 26990 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019617193624248757, + "loss": 0.2186, + "step": 27000 + }, + { + "epoch": 1.75, + "eval_loss": 0.27321794629096985, + "eval_runtime": 1004.0529, + "eval_samples_per_second": 15.525, + "eval_steps_per_second": 1.941, + "eval_wer": 0.2576793506844252, + "step": 27000 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019613274105043113, + "loss": 0.297, + "step": 27010 + }, + { + "epoch": 1.75, + "learning_rate": 0.0001960935458583747, + "loss": 0.2107, + "step": 27020 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019605435066631825, + "loss": 0.198, + "step": 27030 + }, + { + "epoch": 1.75, + "learning_rate": 0.00019601515547426181, + "loss": 0.2139, + "step": 27040 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019597596028220538, + "loss": 0.2646, + "step": 27050 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019593676509014894, + "loss": 0.3519, + "step": 27060 + }, + { + "epoch": 1.76, + "learning_rate": 0.0001958975698980925, + "loss": 0.1886, + "step": 27070 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019585837470603604, + "loss": 0.1869, + "step": 27080 + }, + { + "epoch": 1.76, + "learning_rate": 0.0001958191795139796, + "loss": 0.2075, + "step": 27090 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019577998432192316, + "loss": 0.2768, + "step": 27100 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019574078912986672, + "loss": 0.2918, + "step": 27110 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019570159393781029, + "loss": 0.2025, + "step": 27120 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019566239874575385, + "loss": 0.223, + "step": 27130 + }, + { + "epoch": 1.76, + "learning_rate": 0.0001956232035536974, + "loss": 0.2047, + "step": 27140 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019558400836164097, + "loss": 0.2151, + "step": 27150 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019554481316958453, + "loss": 0.2738, + "step": 27160 + }, + { + "epoch": 1.76, + "learning_rate": 0.0001955056179775281, + "loss": 0.2016, + "step": 27170 + }, + { + "epoch": 1.76, + "learning_rate": 0.00019546642278547163, + "loss": 0.1914, + "step": 27180 + }, + { + "epoch": 1.76, + "learning_rate": 0.0001954272275934152, + "loss": 0.229, + "step": 27190 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019538803240135876, + "loss": 0.2479, + "step": 27200 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019534883720930232, + "loss": 0.2891, + "step": 27210 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019530964201724588, + "loss": 0.1971, + "step": 27220 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019527044682518944, + "loss": 0.1967, + "step": 27230 + }, + { + "epoch": 1.77, + "learning_rate": 0.000195231251633133, + "loss": 0.2186, + "step": 27240 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019519205644107651, + "loss": 0.2579, + "step": 27250 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019515286124902008, + "loss": 0.267, + "step": 27260 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019511366605696364, + "loss": 0.2067, + "step": 27270 + }, + { + "epoch": 1.77, + "learning_rate": 0.0001950744708649072, + "loss": 0.2135, + "step": 27280 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019503527567285076, + "loss": 0.2195, + "step": 27290 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019499608048079433, + "loss": 0.2508, + "step": 27300 + }, + { + "epoch": 1.77, + "learning_rate": 0.0001949568852887379, + "loss": 0.2953, + "step": 27310 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019491769009668145, + "loss": 0.1992, + "step": 27320 + }, + { + "epoch": 1.77, + "learning_rate": 0.000194878494904625, + "loss": 0.2239, + "step": 27330 + }, + { + "epoch": 1.77, + "learning_rate": 0.00019483929971256857, + "loss": 0.1817, + "step": 27340 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019480010452051214, + "loss": 0.2429, + "step": 27350 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019476090932845567, + "loss": 0.319, + "step": 27360 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019472171413639923, + "loss": 0.1895, + "step": 27370 + }, + { + "epoch": 1.78, + "learning_rate": 0.0001946825189443428, + "loss": 0.2048, + "step": 27380 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019464332375228636, + "loss": 0.2255, + "step": 27390 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019460412856022992, + "loss": 0.272, + "step": 27400 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019456493336817348, + "loss": 0.322, + "step": 27410 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019452573817611705, + "loss": 0.1851, + "step": 27420 + }, + { + "epoch": 1.78, + "learning_rate": 0.0001944865429840606, + "loss": 0.243, + "step": 27430 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019444734779200417, + "loss": 0.1678, + "step": 27440 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019440815259994773, + "loss": 0.2121, + "step": 27450 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019436895740789127, + "loss": 0.3066, + "step": 27460 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019432976221583483, + "loss": 0.1884, + "step": 27470 + }, + { + "epoch": 1.78, + "learning_rate": 0.0001942905670237784, + "loss": 0.2015, + "step": 27480 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019425137183172195, + "loss": 0.1933, + "step": 27490 + }, + { + "epoch": 1.78, + "learning_rate": 0.00019421217663966552, + "loss": 0.2647, + "step": 27500 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019417298144760908, + "loss": 0.283, + "step": 27510 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019413378625555264, + "loss": 0.2314, + "step": 27520 + }, + { + "epoch": 1.79, + "learning_rate": 0.0001940945910634962, + "loss": 0.1674, + "step": 27530 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019405539587143977, + "loss": 0.2071, + "step": 27540 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019401620067938333, + "loss": 0.2484, + "step": 27550 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019397700548732686, + "loss": 0.2862, + "step": 27560 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019393781029527043, + "loss": 0.1764, + "step": 27570 + }, + { + "epoch": 1.79, + "learning_rate": 0.000193898615103214, + "loss": 0.183, + "step": 27580 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019385941991115755, + "loss": 0.1886, + "step": 27590 + }, + { + "epoch": 1.79, + "learning_rate": 0.0001938202247191011, + "loss": 0.2119, + "step": 27600 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019378102952704468, + "loss": 0.2877, + "step": 27610 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019374183433498824, + "loss": 0.1674, + "step": 27620 + }, + { + "epoch": 1.79, + "learning_rate": 0.0001937026391429318, + "loss": 0.1985, + "step": 27630 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019366344395087536, + "loss": 0.2271, + "step": 27640 + }, + { + "epoch": 1.79, + "learning_rate": 0.00019362424875881892, + "loss": 0.244, + "step": 27650 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019358505356676246, + "loss": 0.278, + "step": 27660 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019354585837470602, + "loss": 0.1848, + "step": 27670 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019350666318264958, + "loss": 0.1985, + "step": 27680 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019346746799059315, + "loss": 0.2206, + "step": 27690 + }, + { + "epoch": 1.8, + "learning_rate": 0.0001934282727985367, + "loss": 0.2443, + "step": 27700 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019338907760648027, + "loss": 0.2805, + "step": 27710 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019334988241442383, + "loss": 0.1867, + "step": 27720 + }, + { + "epoch": 1.8, + "learning_rate": 0.0001933106872223674, + "loss": 0.1979, + "step": 27730 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019327149203031096, + "loss": 0.2449, + "step": 27740 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019323229683825447, + "loss": 0.2383, + "step": 27750 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019319310164619803, + "loss": 0.3063, + "step": 27760 + }, + { + "epoch": 1.8, + "learning_rate": 0.0001931539064541416, + "loss": 0.1945, + "step": 27770 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019311471126208515, + "loss": 0.229, + "step": 27780 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019307551607002872, + "loss": 0.2225, + "step": 27790 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019303632087797228, + "loss": 0.2673, + "step": 27800 + }, + { + "epoch": 1.8, + "learning_rate": 0.00019299712568591584, + "loss": 0.3011, + "step": 27810 + }, + { + "epoch": 1.81, + "learning_rate": 0.0001929579304938594, + "loss": 0.1782, + "step": 27820 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019291873530180296, + "loss": 0.208, + "step": 27830 + }, + { + "epoch": 1.81, + "learning_rate": 0.0001928795401097465, + "loss": 0.2074, + "step": 27840 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019284034491769006, + "loss": 0.2318, + "step": 27850 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019280114972563362, + "loss": 0.2804, + "step": 27860 + }, + { + "epoch": 1.81, + "learning_rate": 0.0001927619545335772, + "loss": 0.1814, + "step": 27870 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019272275934152075, + "loss": 0.2033, + "step": 27880 + }, + { + "epoch": 1.81, + "learning_rate": 0.0001926835641494643, + "loss": 0.1926, + "step": 27890 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019264436895740787, + "loss": 0.2016, + "step": 27900 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019260517376535144, + "loss": 0.3016, + "step": 27910 + }, + { + "epoch": 1.81, + "learning_rate": 0.000192565978573295, + "loss": 0.1932, + "step": 27920 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019252678338123856, + "loss": 0.201, + "step": 27930 + }, + { + "epoch": 1.81, + "learning_rate": 0.0001924875881891821, + "loss": 0.1876, + "step": 27940 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019244839299712566, + "loss": 0.2348, + "step": 27950 + }, + { + "epoch": 1.81, + "learning_rate": 0.00019240919780506922, + "loss": 0.2557, + "step": 27960 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019237000261301278, + "loss": 0.1747, + "step": 27970 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019233080742095634, + "loss": 0.2034, + "step": 27980 + }, + { + "epoch": 1.82, + "learning_rate": 0.0001922916122288999, + "loss": 0.212, + "step": 27990 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019225241703684347, + "loss": 0.2671, + "step": 28000 + }, + { + "epoch": 1.82, + "eval_loss": 0.2614744007587433, + "eval_runtime": 1008.1614, + "eval_samples_per_second": 15.462, + "eval_steps_per_second": 1.933, + "eval_wer": 0.261304032636352, + "step": 28000 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019221322184478703, + "loss": 0.323, + "step": 28010 + }, + { + "epoch": 1.82, + "learning_rate": 0.0001921740266527306, + "loss": 0.1733, + "step": 28020 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019213483146067416, + "loss": 0.1992, + "step": 28030 + }, + { + "epoch": 1.82, + "learning_rate": 0.0001920956362686177, + "loss": 0.18, + "step": 28040 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019205644107656125, + "loss": 0.2481, + "step": 28050 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019201724588450482, + "loss": 0.3052, + "step": 28060 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019197805069244838, + "loss": 0.1819, + "step": 28070 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019193885550039194, + "loss": 0.1797, + "step": 28080 + }, + { + "epoch": 1.82, + "learning_rate": 0.0001918996603083355, + "loss": 0.2582, + "step": 28090 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019186046511627906, + "loss": 0.2269, + "step": 28100 + }, + { + "epoch": 1.82, + "learning_rate": 0.00019182126992422263, + "loss": 0.3118, + "step": 28110 + }, + { + "epoch": 1.83, + "learning_rate": 0.0001917820747321662, + "loss": 0.1873, + "step": 28120 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019174287954010975, + "loss": 0.1947, + "step": 28130 + }, + { + "epoch": 1.83, + "learning_rate": 0.0001917036843480533, + "loss": 0.2142, + "step": 28140 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019166448915599685, + "loss": 0.1848, + "step": 28150 + }, + { + "epoch": 1.83, + "learning_rate": 0.0001916252939639404, + "loss": 0.2744, + "step": 28160 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019158609877188397, + "loss": 0.2119, + "step": 28170 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019154690357982754, + "loss": 0.215, + "step": 28180 + }, + { + "epoch": 1.83, + "learning_rate": 0.0001915077083877711, + "loss": 0.2186, + "step": 28190 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019146851319571466, + "loss": 0.2355, + "step": 28200 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019142931800365822, + "loss": 0.333, + "step": 28210 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019139012281160179, + "loss": 0.184, + "step": 28220 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019135092761954535, + "loss": 0.2094, + "step": 28230 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019131173242748886, + "loss": 0.2435, + "step": 28240 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019127253723543242, + "loss": 0.2004, + "step": 28250 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019123334204337598, + "loss": 0.2907, + "step": 28260 + }, + { + "epoch": 1.83, + "learning_rate": 0.00019119414685131954, + "loss": 0.1358, + "step": 28270 + }, + { + "epoch": 1.84, + "learning_rate": 0.0001911549516592631, + "loss": 0.2013, + "step": 28280 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019111575646720667, + "loss": 0.2131, + "step": 28290 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019107656127515023, + "loss": 0.2342, + "step": 28300 + }, + { + "epoch": 1.84, + "learning_rate": 0.0001910373660830938, + "loss": 0.2901, + "step": 28310 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019099817089103733, + "loss": 0.1789, + "step": 28320 + }, + { + "epoch": 1.84, + "learning_rate": 0.0001909589756989809, + "loss": 0.1831, + "step": 28330 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019091978050692445, + "loss": 0.207, + "step": 28340 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019088058531486801, + "loss": 0.2371, + "step": 28350 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019084139012281158, + "loss": 0.3361, + "step": 28360 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019080219493075514, + "loss": 0.1911, + "step": 28370 + }, + { + "epoch": 1.84, + "learning_rate": 0.0001907629997386987, + "loss": 0.2099, + "step": 28380 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019072380454664226, + "loss": 0.1948, + "step": 28390 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019068460935458583, + "loss": 0.2506, + "step": 28400 + }, + { + "epoch": 1.84, + "learning_rate": 0.0001906454141625294, + "loss": 0.2598, + "step": 28410 + }, + { + "epoch": 1.84, + "learning_rate": 0.00019060621897047292, + "loss": 0.1918, + "step": 28420 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019056702377841649, + "loss": 0.1785, + "step": 28430 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019052782858636005, + "loss": 0.2494, + "step": 28440 + }, + { + "epoch": 1.85, + "learning_rate": 0.0001904886333943036, + "loss": 0.2394, + "step": 28450 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019044943820224717, + "loss": 0.2797, + "step": 28460 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019041024301019073, + "loss": 0.1686, + "step": 28470 + }, + { + "epoch": 1.85, + "learning_rate": 0.0001903710478181343, + "loss": 0.2025, + "step": 28480 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019033185262607786, + "loss": 0.211, + "step": 28490 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019029265743402142, + "loss": 0.2148, + "step": 28500 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019025346224196498, + "loss": 0.2556, + "step": 28510 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019021426704990852, + "loss": 0.1896, + "step": 28520 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019017507185785208, + "loss": 0.2077, + "step": 28530 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019013587666579564, + "loss": 0.2317, + "step": 28540 + }, + { + "epoch": 1.85, + "learning_rate": 0.0001900966814737392, + "loss": 0.2424, + "step": 28550 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019005748628168277, + "loss": 0.2858, + "step": 28560 + }, + { + "epoch": 1.85, + "learning_rate": 0.00019001829108962633, + "loss": 0.1753, + "step": 28570 + }, + { + "epoch": 1.85, + "learning_rate": 0.0001899790958975699, + "loss": 0.2254, + "step": 28580 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018993990070551345, + "loss": 0.2178, + "step": 28590 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018990070551345702, + "loss": 0.2409, + "step": 28600 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018986151032140058, + "loss": 0.3128, + "step": 28610 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018982231512934411, + "loss": 0.202, + "step": 28620 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018978311993728768, + "loss": 0.2084, + "step": 28630 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018974392474523124, + "loss": 0.2252, + "step": 28640 + }, + { + "epoch": 1.86, + "learning_rate": 0.0001897047295531748, + "loss": 0.235, + "step": 28650 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018966553436111836, + "loss": 0.3178, + "step": 28660 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018962633916906193, + "loss": 0.1555, + "step": 28670 + }, + { + "epoch": 1.86, + "learning_rate": 0.0001895871439770055, + "loss": 0.209, + "step": 28680 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018954794878494905, + "loss": 0.218, + "step": 28690 + }, + { + "epoch": 1.86, + "learning_rate": 0.0001895087535928926, + "loss": 0.242, + "step": 28700 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018946955840083617, + "loss": 0.2403, + "step": 28710 + }, + { + "epoch": 1.86, + "learning_rate": 0.0001894303632087797, + "loss": 0.2045, + "step": 28720 + }, + { + "epoch": 1.86, + "learning_rate": 0.00018939116801672325, + "loss": 0.2054, + "step": 28730 + }, + { + "epoch": 1.87, + "learning_rate": 0.0001893519728246668, + "loss": 0.2089, + "step": 28740 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018931277763261037, + "loss": 0.2336, + "step": 28750 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018927358244055393, + "loss": 0.2803, + "step": 28760 + }, + { + "epoch": 1.87, + "learning_rate": 0.0001892343872484975, + "loss": 0.1741, + "step": 28770 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018919519205644106, + "loss": 0.1872, + "step": 28780 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018915599686438462, + "loss": 0.2185, + "step": 28790 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018911680167232815, + "loss": 0.2236, + "step": 28800 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018907760648027172, + "loss": 0.2782, + "step": 28810 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018903841128821528, + "loss": 0.2081, + "step": 28820 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018899921609615884, + "loss": 0.2014, + "step": 28830 + }, + { + "epoch": 1.87, + "learning_rate": 0.0001889600209041024, + "loss": 0.2149, + "step": 28840 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018892082571204597, + "loss": 0.2634, + "step": 28850 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018888163051998953, + "loss": 0.3021, + "step": 28860 + }, + { + "epoch": 1.87, + "learning_rate": 0.0001888424353279331, + "loss": 0.1795, + "step": 28870 + }, + { + "epoch": 1.87, + "learning_rate": 0.00018880324013587665, + "loss": 0.1938, + "step": 28880 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018876404494382021, + "loss": 0.2216, + "step": 28890 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018872484975176375, + "loss": 0.2066, + "step": 28900 + }, + { + "epoch": 1.88, + "learning_rate": 0.0001886856545597073, + "loss": 0.2713, + "step": 28910 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018864645936765087, + "loss": 0.1796, + "step": 28920 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018860726417559444, + "loss": 0.2114, + "step": 28930 + }, + { + "epoch": 1.88, + "learning_rate": 0.000188568068983538, + "loss": 0.2056, + "step": 28940 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018852887379148156, + "loss": 0.2357, + "step": 28950 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018848967859942512, + "loss": 0.3368, + "step": 28960 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018845048340736869, + "loss": 0.1683, + "step": 28970 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018841128821531225, + "loss": 0.2302, + "step": 28980 + }, + { + "epoch": 1.88, + "learning_rate": 0.0001883720930232558, + "loss": 0.2096, + "step": 28990 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018833289783119935, + "loss": 0.2613, + "step": 29000 + }, + { + "epoch": 1.88, + "eval_loss": 0.25850141048431396, + "eval_runtime": 1006.3053, + "eval_samples_per_second": 15.49, + "eval_steps_per_second": 1.937, + "eval_wer": 0.25517760941564444, + "step": 29000 + }, + { + "epoch": 1.88, + "learning_rate": 0.0001882937026391429, + "loss": 0.2878, + "step": 29010 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018825842696629211, + "loss": 0.1837, + "step": 29020 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018821923177423568, + "loss": 0.2316, + "step": 29030 + }, + { + "epoch": 1.88, + "learning_rate": 0.00018818003658217924, + "loss": 0.2146, + "step": 29040 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018814084139012277, + "loss": 0.2426, + "step": 29050 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018810164619806634, + "loss": 0.3411, + "step": 29060 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001880624510060099, + "loss": 0.1756, + "step": 29070 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018802325581395346, + "loss": 0.1822, + "step": 29080 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018798406062189702, + "loss": 0.1889, + "step": 29090 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018794486542984059, + "loss": 0.2387, + "step": 29100 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018790567023778415, + "loss": 0.2893, + "step": 29110 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001878664750457277, + "loss": 0.2173, + "step": 29120 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018782727985367127, + "loss": 0.2031, + "step": 29130 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018778808466161483, + "loss": 0.2353, + "step": 29140 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001877488894695584, + "loss": 0.2358, + "step": 29150 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018771361379670758, + "loss": 0.296, + "step": 29160 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018767441860465114, + "loss": 0.1973, + "step": 29170 + }, + { + "epoch": 1.89, + "learning_rate": 0.0001876352234125947, + "loss": 0.1932, + "step": 29180 + }, + { + "epoch": 1.89, + "learning_rate": 0.00018759602822053826, + "loss": 0.2183, + "step": 29190 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018755683302848183, + "loss": 0.2145, + "step": 29200 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018751763783642536, + "loss": 0.2745, + "step": 29210 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018747844264436892, + "loss": 0.1893, + "step": 29220 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018743924745231248, + "loss": 0.2253, + "step": 29230 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018740005226025605, + "loss": 0.1985, + "step": 29240 + }, + { + "epoch": 1.9, + "learning_rate": 0.0001873608570681996, + "loss": 0.2485, + "step": 29250 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018732166187614317, + "loss": 0.2788, + "step": 29260 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018728246668408673, + "loss": 0.1812, + "step": 29270 + }, + { + "epoch": 1.9, + "learning_rate": 0.0001872432714920303, + "loss": 0.2078, + "step": 29280 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018720407629997386, + "loss": 0.2031, + "step": 29290 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018716488110791742, + "loss": 0.2153, + "step": 29300 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018712568591586096, + "loss": 0.2718, + "step": 29310 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018708649072380452, + "loss": 0.1837, + "step": 29320 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018704729553174808, + "loss": 0.1745, + "step": 29330 + }, + { + "epoch": 1.9, + "learning_rate": 0.00018700810033969164, + "loss": 0.2087, + "step": 29340 + }, + { + "epoch": 1.9, + "learning_rate": 0.0001869689051476352, + "loss": 0.2368, + "step": 29350 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018692970995557877, + "loss": 0.3166, + "step": 29360 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018689051476352233, + "loss": 0.1902, + "step": 29370 + }, + { + "epoch": 1.91, + "learning_rate": 0.0001868513195714659, + "loss": 0.1902, + "step": 29380 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018681212437940945, + "loss": 0.1804, + "step": 29390 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018677292918735302, + "loss": 0.2074, + "step": 29400 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018673373399529655, + "loss": 0.2903, + "step": 29410 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018669453880324011, + "loss": 0.1794, + "step": 29420 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018665534361118368, + "loss": 0.1989, + "step": 29430 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018661614841912724, + "loss": 0.2311, + "step": 29440 + }, + { + "epoch": 1.91, + "learning_rate": 0.0001865769532270708, + "loss": 0.2688, + "step": 29450 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018653775803501436, + "loss": 0.2952, + "step": 29460 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018649856284295793, + "loss": 0.203, + "step": 29470 + }, + { + "epoch": 1.91, + "learning_rate": 0.0001864593676509015, + "loss": 0.2161, + "step": 29480 + }, + { + "epoch": 1.91, + "learning_rate": 0.00018642017245884505, + "loss": 0.1999, + "step": 29490 + }, + { + "epoch": 1.91, + "learning_rate": 0.0001863809772667886, + "loss": 0.2392, + "step": 29500 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018634178207473215, + "loss": 0.3383, + "step": 29510 + }, + { + "epoch": 1.92, + "learning_rate": 0.0001863025868826757, + "loss": 0.1893, + "step": 29520 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018626339169061927, + "loss": 0.1692, + "step": 29530 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018622419649856283, + "loss": 0.1887, + "step": 29540 + }, + { + "epoch": 1.92, + "learning_rate": 0.0001861850013065064, + "loss": 0.1979, + "step": 29550 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018614580611444996, + "loss": 0.303, + "step": 29560 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018610661092239352, + "loss": 0.1986, + "step": 29570 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018606741573033708, + "loss": 0.1964, + "step": 29580 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018602822053828065, + "loss": 0.2294, + "step": 29590 + }, + { + "epoch": 1.92, + "learning_rate": 0.0001859890253462242, + "loss": 0.2068, + "step": 29600 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018594983015416772, + "loss": 0.3099, + "step": 29610 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018591063496211128, + "loss": 0.2008, + "step": 29620 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018587143977005484, + "loss": 0.1857, + "step": 29630 + }, + { + "epoch": 1.92, + "learning_rate": 0.0001858322445779984, + "loss": 0.2026, + "step": 29640 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018579304938594197, + "loss": 0.2547, + "step": 29650 + }, + { + "epoch": 1.92, + "learning_rate": 0.00018575385419388553, + "loss": 0.2813, + "step": 29660 + }, + { + "epoch": 1.93, + "learning_rate": 0.0001857146590018291, + "loss": 0.17, + "step": 29670 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018567546380977265, + "loss": 0.2109, + "step": 29680 + }, + { + "epoch": 1.93, + "learning_rate": 0.0001856362686177162, + "loss": 0.2275, + "step": 29690 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018559707342565975, + "loss": 0.2403, + "step": 29700 + }, + { + "epoch": 1.93, + "learning_rate": 0.0001855578782336033, + "loss": 0.3231, + "step": 29710 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018551868304154687, + "loss": 0.1878, + "step": 29720 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018547948784949044, + "loss": 0.1786, + "step": 29730 + }, + { + "epoch": 1.93, + "learning_rate": 0.000185440292657434, + "loss": 0.2026, + "step": 29740 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018540109746537756, + "loss": 0.233, + "step": 29750 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018536190227332112, + "loss": 0.2729, + "step": 29760 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018532270708126469, + "loss": 0.1592, + "step": 29770 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018528351188920825, + "loss": 0.2158, + "step": 29780 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018524431669715178, + "loss": 0.2062, + "step": 29790 + }, + { + "epoch": 1.93, + "learning_rate": 0.00018520512150509535, + "loss": 0.224, + "step": 29800 + }, + { + "epoch": 1.93, + "learning_rate": 0.0001851659263130389, + "loss": 0.2949, + "step": 29810 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018512673112098247, + "loss": 0.1752, + "step": 29820 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018508753592892603, + "loss": 0.2047, + "step": 29830 + }, + { + "epoch": 1.94, + "learning_rate": 0.0001850483407368696, + "loss": 0.2215, + "step": 29840 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018500914554481316, + "loss": 0.2606, + "step": 29850 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018496995035275672, + "loss": 0.3249, + "step": 29860 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018493075516070028, + "loss": 0.1841, + "step": 29870 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018489155996864384, + "loss": 0.188, + "step": 29880 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018485236477658738, + "loss": 0.2235, + "step": 29890 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018481316958453094, + "loss": 0.251, + "step": 29900 + }, + { + "epoch": 1.94, + "learning_rate": 0.0001847739743924745, + "loss": 0.2589, + "step": 29910 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018473477920041807, + "loss": 0.1913, + "step": 29920 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018469558400836163, + "loss": 0.2305, + "step": 29930 + }, + { + "epoch": 1.94, + "learning_rate": 0.0001846563888163052, + "loss": 0.2191, + "step": 29940 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018461719362424875, + "loss": 0.24, + "step": 29950 + }, + { + "epoch": 1.94, + "learning_rate": 0.00018457799843219232, + "loss": 0.3172, + "step": 29960 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018453880324013588, + "loss": 0.1874, + "step": 29970 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018449960804807944, + "loss": 0.2055, + "step": 29980 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018446041285602297, + "loss": 0.1941, + "step": 29990 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018442121766396654, + "loss": 0.2258, + "step": 30000 + }, + { + "epoch": 1.95, + "eval_loss": 0.257784366607666, + "eval_runtime": 1012.8453, + "eval_samples_per_second": 15.39, + "eval_steps_per_second": 1.924, + "eval_wer": 0.25558272092791856, + "step": 30000 + }, + { + "epoch": 1.95, + "learning_rate": 0.0001843820224719101, + "loss": 0.2686, + "step": 30010 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018434282727985366, + "loss": 0.1421, + "step": 30020 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018430363208779722, + "loss": 0.2337, + "step": 30030 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018426443689574079, + "loss": 0.1961, + "step": 30040 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018422524170368435, + "loss": 0.2539, + "step": 30050 + }, + { + "epoch": 1.95, + "learning_rate": 0.0001841860465116279, + "loss": 0.3216, + "step": 30060 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018414685131957147, + "loss": 0.1923, + "step": 30070 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018410765612751504, + "loss": 0.1988, + "step": 30080 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018406846093545857, + "loss": 0.1976, + "step": 30090 + }, + { + "epoch": 1.95, + "learning_rate": 0.0001840292657434021, + "loss": 0.2136, + "step": 30100 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018399007055134567, + "loss": 0.3297, + "step": 30110 + }, + { + "epoch": 1.95, + "learning_rate": 0.00018395087535928923, + "loss": 0.1661, + "step": 30120 + }, + { + "epoch": 1.96, + "learning_rate": 0.0001839116801672328, + "loss": 0.1976, + "step": 30130 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018387248497517636, + "loss": 0.2104, + "step": 30140 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018383328978311992, + "loss": 0.2456, + "step": 30150 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018379409459106348, + "loss": 0.2927, + "step": 30160 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018375489939900701, + "loss": 0.2055, + "step": 30170 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018371570420695058, + "loss": 0.1833, + "step": 30180 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018367650901489414, + "loss": 0.2025, + "step": 30190 + }, + { + "epoch": 1.96, + "learning_rate": 0.0001836373138228377, + "loss": 0.2401, + "step": 30200 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018359811863078126, + "loss": 0.2653, + "step": 30210 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018355892343872483, + "loss": 0.1937, + "step": 30220 + }, + { + "epoch": 1.96, + "learning_rate": 0.0001835197282466684, + "loss": 0.1627, + "step": 30230 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018348053305461195, + "loss": 0.2089, + "step": 30240 + }, + { + "epoch": 1.96, + "learning_rate": 0.0001834413378625555, + "loss": 0.2252, + "step": 30250 + }, + { + "epoch": 1.96, + "learning_rate": 0.00018340214267049908, + "loss": 0.3172, + "step": 30260 + }, + { + "epoch": 1.96, + "learning_rate": 0.0001833629474784426, + "loss": 0.1975, + "step": 30270 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018332375228638617, + "loss": 0.2087, + "step": 30280 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018328455709432974, + "loss": 0.1979, + "step": 30290 + }, + { + "epoch": 1.97, + "learning_rate": 0.0001832453619022733, + "loss": 0.206, + "step": 30300 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018320616671021686, + "loss": 0.3072, + "step": 30310 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018316697151816042, + "loss": 0.151, + "step": 30320 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018312777632610398, + "loss": 0.1951, + "step": 30330 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018308858113404755, + "loss": 0.2166, + "step": 30340 + }, + { + "epoch": 1.97, + "learning_rate": 0.0001830493859419911, + "loss": 0.2119, + "step": 30350 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018301019074993467, + "loss": 0.2969, + "step": 30360 + }, + { + "epoch": 1.97, + "learning_rate": 0.0001829709955578782, + "loss": 0.1833, + "step": 30370 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018293180036582177, + "loss": 0.1754, + "step": 30380 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018289260517376533, + "loss": 0.2223, + "step": 30390 + }, + { + "epoch": 1.97, + "learning_rate": 0.0001828534099817089, + "loss": 0.1894, + "step": 30400 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018281421478965246, + "loss": 0.3207, + "step": 30410 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018277501959759602, + "loss": 0.1941, + "step": 30420 + }, + { + "epoch": 1.97, + "learning_rate": 0.00018273582440553958, + "loss": 0.1798, + "step": 30430 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018269662921348314, + "loss": 0.2405, + "step": 30440 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001826574340214267, + "loss": 0.1947, + "step": 30450 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018261823882937027, + "loss": 0.2696, + "step": 30460 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001825790436373138, + "loss": 0.212, + "step": 30470 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018253984844525736, + "loss": 0.1981, + "step": 30480 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018250065325320093, + "loss": 0.2247, + "step": 30490 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001824614580611445, + "loss": 0.2515, + "step": 30500 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018242226286908805, + "loss": 0.2592, + "step": 30510 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018238306767703161, + "loss": 0.2108, + "step": 30520 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018234387248497518, + "loss": 0.1847, + "step": 30530 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018230467729291874, + "loss": 0.2311, + "step": 30540 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001822654821008623, + "loss": 0.2247, + "step": 30550 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018222628690880586, + "loss": 0.286, + "step": 30560 + }, + { + "epoch": 1.98, + "learning_rate": 0.0001821870917167494, + "loss": 0.2018, + "step": 30570 + }, + { + "epoch": 1.98, + "learning_rate": 0.00018214789652469296, + "loss": 0.1842, + "step": 30580 + }, + { + "epoch": 1.99, + "learning_rate": 0.0001821087013326365, + "loss": 0.2329, + "step": 30590 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018206950614058006, + "loss": 0.236, + "step": 30600 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018203031094852362, + "loss": 0.3202, + "step": 30610 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018199111575646718, + "loss": 0.1893, + "step": 30620 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018195192056441074, + "loss": 0.2036, + "step": 30630 + }, + { + "epoch": 1.99, + "learning_rate": 0.0001819127253723543, + "loss": 0.1954, + "step": 30640 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018187353018029784, + "loss": 0.2402, + "step": 30650 + }, + { + "epoch": 1.99, + "learning_rate": 0.0001818343349882414, + "loss": 0.2419, + "step": 30660 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018179513979618497, + "loss": 0.1882, + "step": 30670 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018175594460412853, + "loss": 0.1679, + "step": 30680 + }, + { + "epoch": 1.99, + "learning_rate": 0.0001817167494120721, + "loss": 0.2112, + "step": 30690 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018167755422001565, + "loss": 0.2415, + "step": 30700 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018163835902795922, + "loss": 0.2817, + "step": 30710 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018159916383590278, + "loss": 0.2099, + "step": 30720 + }, + { + "epoch": 1.99, + "learning_rate": 0.00018155996864384634, + "loss": 0.1937, + "step": 30730 + }, + { + "epoch": 2.0, + "learning_rate": 0.0001815207734517899, + "loss": 0.2233, + "step": 30740 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018148157825973344, + "loss": 0.2472, + "step": 30750 + }, + { + "epoch": 2.0, + "learning_rate": 0.000181442383067677, + "loss": 0.3263, + "step": 30760 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018140318787562056, + "loss": 0.1781, + "step": 30770 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018136399268356412, + "loss": 0.1984, + "step": 30780 + }, + { + "epoch": 2.0, + "learning_rate": 0.0001813247974915077, + "loss": 0.2077, + "step": 30790 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018128560229945125, + "loss": 0.2365, + "step": 30800 + }, + { + "epoch": 2.0, + "learning_rate": 0.0001812464071073948, + "loss": 0.3393, + "step": 30810 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018120721191533837, + "loss": 0.21, + "step": 30820 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018116801672328194, + "loss": 0.1999, + "step": 30830 + }, + { + "epoch": 2.0, + "learning_rate": 0.0001811288215312255, + "loss": 0.1637, + "step": 30840 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018108962633916903, + "loss": 0.1574, + "step": 30850 + }, + { + "epoch": 2.0, + "learning_rate": 0.0001810504311471126, + "loss": 0.1974, + "step": 30860 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018101123595505616, + "loss": 0.2854, + "step": 30870 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018097204076299972, + "loss": 0.1569, + "step": 30880 + }, + { + "epoch": 2.0, + "learning_rate": 0.00018093284557094328, + "loss": 0.1669, + "step": 30890 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018089365037888685, + "loss": 0.234, + "step": 30900 + }, + { + "epoch": 2.01, + "learning_rate": 0.0001808544551868304, + "loss": 0.202, + "step": 30910 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018081525999477397, + "loss": 0.2041, + "step": 30920 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018077606480271753, + "loss": 0.1681, + "step": 30930 + }, + { + "epoch": 2.01, + "learning_rate": 0.0001807368696106611, + "loss": 0.1764, + "step": 30940 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018069767441860466, + "loss": 0.1824, + "step": 30950 + }, + { + "epoch": 2.01, + "learning_rate": 0.0001806584792265482, + "loss": 0.2257, + "step": 30960 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018061928403449175, + "loss": 0.2691, + "step": 30970 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018058008884243532, + "loss": 0.1603, + "step": 30980 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018054089365037888, + "loss": 0.1981, + "step": 30990 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018050169845832244, + "loss": 0.2109, + "step": 31000 + }, + { + "epoch": 2.01, + "eval_loss": 0.25705796480178833, + "eval_runtime": 990.8158, + "eval_samples_per_second": 15.732, + "eval_steps_per_second": 1.967, + "eval_wer": 0.2579778539039957, + "step": 31000 + }, + { + "epoch": 2.01, + "learning_rate": 0.000180462503266266, + "loss": 0.2304, + "step": 31010 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018042330807420957, + "loss": 0.2374, + "step": 31020 + }, + { + "epoch": 2.01, + "learning_rate": 0.00018038411288215313, + "loss": 0.1782, + "step": 31030 + }, + { + "epoch": 2.01, + "learning_rate": 0.0001803449176900967, + "loss": 0.2075, + "step": 31040 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018030572249804025, + "loss": 0.2038, + "step": 31050 + }, + { + "epoch": 2.02, + "learning_rate": 0.0001802665273059838, + "loss": 0.2185, + "step": 31060 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018022733211392735, + "loss": 0.2882, + "step": 31070 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018018813692187089, + "loss": 0.1995, + "step": 31080 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018014894172981445, + "loss": 0.1913, + "step": 31090 + }, + { + "epoch": 2.02, + "learning_rate": 0.000180109746537758, + "loss": 0.2197, + "step": 31100 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018007055134570157, + "loss": 0.2066, + "step": 31110 + }, + { + "epoch": 2.02, + "learning_rate": 0.00018003135615364513, + "loss": 0.2574, + "step": 31120 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017999216096158867, + "loss": 0.1769, + "step": 31130 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017995296576953223, + "loss": 0.1777, + "step": 31140 + }, + { + "epoch": 2.02, + "learning_rate": 0.0001799137705774758, + "loss": 0.1947, + "step": 31150 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017987457538541936, + "loss": 0.2229, + "step": 31160 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017983538019336292, + "loss": 0.2373, + "step": 31170 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017979618500130648, + "loss": 0.1951, + "step": 31180 + }, + { + "epoch": 2.02, + "learning_rate": 0.00017975698980925004, + "loss": 0.1639, + "step": 31190 + }, + { + "epoch": 2.02, + "learning_rate": 0.0001797177946171936, + "loss": 0.1998, + "step": 31200 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017967859942513717, + "loss": 0.2557, + "step": 31210 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017963940423308073, + "loss": 0.229, + "step": 31220 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017960020904102427, + "loss": 0.1693, + "step": 31230 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017956101384896783, + "loss": 0.1717, + "step": 31240 + }, + { + "epoch": 2.03, + "learning_rate": 0.0001795218186569114, + "loss": 0.186, + "step": 31250 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017948262346485495, + "loss": 0.1979, + "step": 31260 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017944342827279851, + "loss": 0.2607, + "step": 31270 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017940423308074208, + "loss": 0.1679, + "step": 31280 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017936503788868564, + "loss": 0.1826, + "step": 31290 + }, + { + "epoch": 2.03, + "learning_rate": 0.0001793258426966292, + "loss": 0.1973, + "step": 31300 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017929056702377838, + "loss": 0.2215, + "step": 31310 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017925137183172194, + "loss": 0.288, + "step": 31320 + }, + { + "epoch": 2.03, + "learning_rate": 0.0001792121766396655, + "loss": 0.1592, + "step": 31330 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017917298144760907, + "loss": 0.1646, + "step": 31340 + }, + { + "epoch": 2.03, + "learning_rate": 0.00017913378625555263, + "loss": 0.2187, + "step": 31350 + }, + { + "epoch": 2.04, + "learning_rate": 0.0001790945910634962, + "loss": 0.1993, + "step": 31360 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017905539587143975, + "loss": 0.2347, + "step": 31370 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017901620067938332, + "loss": 0.1665, + "step": 31380 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017897700548732685, + "loss": 0.1736, + "step": 31390 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017893781029527041, + "loss": 0.1868, + "step": 31400 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017889861510321398, + "loss": 0.2253, + "step": 31410 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017885941991115754, + "loss": 0.2503, + "step": 31420 + }, + { + "epoch": 2.04, + "learning_rate": 0.0001788202247191011, + "loss": 0.1598, + "step": 31430 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017878102952704466, + "loss": 0.1687, + "step": 31440 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017874183433498823, + "loss": 0.2139, + "step": 31450 + }, + { + "epoch": 2.04, + "learning_rate": 0.0001787026391429318, + "loss": 0.2171, + "step": 31460 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017866344395087535, + "loss": 0.2941, + "step": 31470 + }, + { + "epoch": 2.04, + "learning_rate": 0.0001786242487588189, + "loss": 0.1761, + "step": 31480 + }, + { + "epoch": 2.04, + "learning_rate": 0.00017858505356676245, + "loss": 0.1526, + "step": 31490 + }, + { + "epoch": 2.04, + "learning_rate": 0.000178545858374706, + "loss": 0.219, + "step": 31500 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017850666318264957, + "loss": 0.2413, + "step": 31510 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017846746799059313, + "loss": 0.2782, + "step": 31520 + }, + { + "epoch": 2.05, + "learning_rate": 0.0001784282727985367, + "loss": 0.1532, + "step": 31530 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017838907760648026, + "loss": 0.1839, + "step": 31540 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017834988241442382, + "loss": 0.2046, + "step": 31550 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017831068722236738, + "loss": 0.2255, + "step": 31560 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017827149203031095, + "loss": 0.2923, + "step": 31570 + }, + { + "epoch": 2.05, + "learning_rate": 0.0001782322968382545, + "loss": 0.1361, + "step": 31580 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017819310164619804, + "loss": 0.1938, + "step": 31590 + }, + { + "epoch": 2.05, + "learning_rate": 0.0001781539064541416, + "loss": 0.1843, + "step": 31600 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017811471126208517, + "loss": 0.2551, + "step": 31610 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017807551607002873, + "loss": 0.1985, + "step": 31620 + }, + { + "epoch": 2.05, + "learning_rate": 0.0001780363208779723, + "loss": 0.1828, + "step": 31630 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017799712568591585, + "loss": 0.1675, + "step": 31640 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017795793049385942, + "loss": 0.1667, + "step": 31650 + }, + { + "epoch": 2.05, + "learning_rate": 0.00017791873530180298, + "loss": 0.2055, + "step": 31660 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017787954010974654, + "loss": 0.2383, + "step": 31670 + }, + { + "epoch": 2.06, + "learning_rate": 0.0001778403449176901, + "loss": 0.1639, + "step": 31680 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017780114972563364, + "loss": 0.1947, + "step": 31690 + }, + { + "epoch": 2.06, + "learning_rate": 0.0001777619545335772, + "loss": 0.2236, + "step": 31700 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017772275934152076, + "loss": 0.2303, + "step": 31710 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017768356414946433, + "loss": 0.2271, + "step": 31720 + }, + { + "epoch": 2.06, + "learning_rate": 0.0001776443689574079, + "loss": 0.1554, + "step": 31730 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017760517376535145, + "loss": 0.1725, + "step": 31740 + }, + { + "epoch": 2.06, + "learning_rate": 0.000177565978573295, + "loss": 0.2313, + "step": 31750 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017752678338123858, + "loss": 0.1994, + "step": 31760 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017748758818918208, + "loss": 0.2438, + "step": 31770 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017744839299712565, + "loss": 0.206, + "step": 31780 + }, + { + "epoch": 2.06, + "learning_rate": 0.0001774091978050692, + "loss": 0.1729, + "step": 31790 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017737000261301277, + "loss": 0.205, + "step": 31800 + }, + { + "epoch": 2.06, + "learning_rate": 0.00017733080742095633, + "loss": 0.218, + "step": 31810 + }, + { + "epoch": 2.07, + "learning_rate": 0.0001772916122288999, + "loss": 0.2978, + "step": 31820 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017725241703684346, + "loss": 0.1593, + "step": 31830 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017721322184478702, + "loss": 0.1607, + "step": 31840 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017717402665273058, + "loss": 0.1926, + "step": 31850 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017713483146067414, + "loss": 0.1957, + "step": 31860 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017709563626861768, + "loss": 0.2238, + "step": 31870 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017705644107656124, + "loss": 0.161, + "step": 31880 + }, + { + "epoch": 2.07, + "learning_rate": 0.0001770172458845048, + "loss": 0.1874, + "step": 31890 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017697805069244837, + "loss": 0.1752, + "step": 31900 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017693885550039193, + "loss": 0.2157, + "step": 31910 + }, + { + "epoch": 2.07, + "learning_rate": 0.0001768996603083355, + "loss": 0.2807, + "step": 31920 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017686046511627905, + "loss": 0.1467, + "step": 31930 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017682126992422261, + "loss": 0.1664, + "step": 31940 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017678207473216618, + "loss": 0.215, + "step": 31950 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017674287954010974, + "loss": 0.2113, + "step": 31960 + }, + { + "epoch": 2.07, + "learning_rate": 0.00017670368434805327, + "loss": 0.248, + "step": 31970 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017666448915599684, + "loss": 0.1633, + "step": 31980 + }, + { + "epoch": 2.08, + "learning_rate": 0.0001766252939639404, + "loss": 0.165, + "step": 31990 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017658609877188396, + "loss": 0.1726, + "step": 32000 + }, + { + "epoch": 2.08, + "eval_loss": 0.25108110904693604, + "eval_runtime": 993.5794, + "eval_samples_per_second": 15.689, + "eval_steps_per_second": 1.962, + "eval_wer": 0.25430342141547385, + "step": 32000 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017654690357982752, + "loss": 0.2235, + "step": 32010 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017650770838777109, + "loss": 0.2513, + "step": 32020 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017646851319571465, + "loss": 0.1554, + "step": 32030 + }, + { + "epoch": 2.08, + "learning_rate": 0.0001764293180036582, + "loss": 0.1625, + "step": 32040 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017639012281160177, + "loss": 0.2025, + "step": 32050 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017635092761954534, + "loss": 0.223, + "step": 32060 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017631173242748887, + "loss": 0.193, + "step": 32070 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017627253723543243, + "loss": 0.1596, + "step": 32080 + }, + { + "epoch": 2.08, + "learning_rate": 0.000176233342043376, + "loss": 0.167, + "step": 32090 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017619414685131956, + "loss": 0.1793, + "step": 32100 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017615495165926312, + "loss": 0.2515, + "step": 32110 + }, + { + "epoch": 2.08, + "learning_rate": 0.00017611575646720668, + "loss": 0.2639, + "step": 32120 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017607656127515024, + "loss": 0.1583, + "step": 32130 + }, + { + "epoch": 2.09, + "learning_rate": 0.0001760373660830938, + "loss": 0.1831, + "step": 32140 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017599817089103737, + "loss": 0.1695, + "step": 32150 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017595897569898093, + "loss": 0.2186, + "step": 32160 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017591978050692447, + "loss": 0.201, + "step": 32170 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017588058531486803, + "loss": 0.14, + "step": 32180 + }, + { + "epoch": 2.09, + "learning_rate": 0.0001758413901228116, + "loss": 0.1643, + "step": 32190 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017580219493075515, + "loss": 0.2032, + "step": 32200 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017576299973869872, + "loss": 0.1905, + "step": 32210 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017572380454664228, + "loss": 0.2329, + "step": 32220 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017568460935458584, + "loss": 0.1916, + "step": 32230 + }, + { + "epoch": 2.09, + "learning_rate": 0.0001756454141625294, + "loss": 0.1703, + "step": 32240 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017560621897047296, + "loss": 0.1792, + "step": 32250 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017556702377841647, + "loss": 0.2435, + "step": 32260 + }, + { + "epoch": 2.09, + "learning_rate": 0.00017552782858636004, + "loss": 0.2562, + "step": 32270 + }, + { + "epoch": 2.1, + "learning_rate": 0.0001754886333943036, + "loss": 0.1562, + "step": 32280 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017544943820224716, + "loss": 0.1555, + "step": 32290 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017541024301019072, + "loss": 0.1834, + "step": 32300 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017537104781813428, + "loss": 0.2227, + "step": 32310 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017533185262607785, + "loss": 0.2447, + "step": 32320 + }, + { + "epoch": 2.1, + "learning_rate": 0.0001752926574340214, + "loss": 0.1884, + "step": 32330 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017525346224196497, + "loss": 0.1586, + "step": 32340 + }, + { + "epoch": 2.1, + "learning_rate": 0.0001752142670499085, + "loss": 0.181, + "step": 32350 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017517507185785207, + "loss": 0.2209, + "step": 32360 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017513587666579563, + "loss": 0.2506, + "step": 32370 + }, + { + "epoch": 2.1, + "learning_rate": 0.0001750966814737392, + "loss": 0.1612, + "step": 32380 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017505748628168276, + "loss": 0.1891, + "step": 32390 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017501829108962632, + "loss": 0.2118, + "step": 32400 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017497909589756988, + "loss": 0.2199, + "step": 32410 + }, + { + "epoch": 2.1, + "learning_rate": 0.00017493990070551344, + "loss": 0.2255, + "step": 32420 + }, + { + "epoch": 2.1, + "learning_rate": 0.000174900705513457, + "loss": 0.1754, + "step": 32430 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017486151032140057, + "loss": 0.1882, + "step": 32440 + }, + { + "epoch": 2.11, + "learning_rate": 0.0001748223151293441, + "loss": 0.1949, + "step": 32450 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017478311993728766, + "loss": 0.1796, + "step": 32460 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017474392474523123, + "loss": 0.2209, + "step": 32470 + }, + { + "epoch": 2.11, + "learning_rate": 0.0001747047295531748, + "loss": 0.1603, + "step": 32480 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017466553436111835, + "loss": 0.1581, + "step": 32490 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017462633916906191, + "loss": 0.1851, + "step": 32500 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017458714397700548, + "loss": 0.232, + "step": 32510 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017454794878494904, + "loss": 0.2467, + "step": 32520 + }, + { + "epoch": 2.11, + "learning_rate": 0.0001745087535928926, + "loss": 0.1581, + "step": 32530 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017446955840083616, + "loss": 0.1718, + "step": 32540 + }, + { + "epoch": 2.11, + "learning_rate": 0.0001744303632087797, + "loss": 0.1919, + "step": 32550 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017439116801672326, + "loss": 0.1929, + "step": 32560 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017435197282466682, + "loss": 0.2218, + "step": 32570 + }, + { + "epoch": 2.11, + "learning_rate": 0.00017431277763261038, + "loss": 0.1448, + "step": 32580 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017427358244055395, + "loss": 0.1514, + "step": 32590 + }, + { + "epoch": 2.12, + "learning_rate": 0.0001742343872484975, + "loss": 0.2197, + "step": 32600 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017419519205644107, + "loss": 0.2193, + "step": 32610 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017415599686438463, + "loss": 0.2906, + "step": 32620 + }, + { + "epoch": 2.12, + "learning_rate": 0.0001741168016723282, + "loss": 0.1466, + "step": 32630 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017407760648027176, + "loss": 0.1948, + "step": 32640 + }, + { + "epoch": 2.12, + "learning_rate": 0.0001740384112882153, + "loss": 0.1698, + "step": 32650 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017399921609615886, + "loss": 0.2289, + "step": 32660 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017396002090410242, + "loss": 0.2654, + "step": 32670 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017392082571204598, + "loss": 0.1557, + "step": 32680 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017388163051998954, + "loss": 0.1865, + "step": 32690 + }, + { + "epoch": 2.12, + "learning_rate": 0.0001738424353279331, + "loss": 0.1827, + "step": 32700 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017380324013587667, + "loss": 0.2144, + "step": 32710 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017376404494382023, + "loss": 0.2086, + "step": 32720 + }, + { + "epoch": 2.12, + "learning_rate": 0.0001737248497517638, + "loss": 0.1515, + "step": 32730 + }, + { + "epoch": 2.12, + "learning_rate": 0.00017368565455970735, + "loss": 0.1728, + "step": 32740 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017364645936765086, + "loss": 0.1796, + "step": 32750 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017360726417559442, + "loss": 0.2334, + "step": 32760 + }, + { + "epoch": 2.13, + "learning_rate": 0.000173568068983538, + "loss": 0.2304, + "step": 32770 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017352887379148155, + "loss": 0.1938, + "step": 32780 + }, + { + "epoch": 2.13, + "learning_rate": 0.0001734896785994251, + "loss": 0.1868, + "step": 32790 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017345048340736867, + "loss": 0.1856, + "step": 32800 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017341128821531224, + "loss": 0.185, + "step": 32810 + }, + { + "epoch": 2.13, + "learning_rate": 0.0001733720930232558, + "loss": 0.2097, + "step": 32820 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017333289783119933, + "loss": 0.1607, + "step": 32830 + }, + { + "epoch": 2.13, + "learning_rate": 0.0001732937026391429, + "loss": 0.1809, + "step": 32840 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017325450744708646, + "loss": 0.2161, + "step": 32850 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017321531225503002, + "loss": 0.2463, + "step": 32860 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017317611706297358, + "loss": 0.2351, + "step": 32870 + }, + { + "epoch": 2.13, + "learning_rate": 0.00017313692187091715, + "loss": 0.1491, + "step": 32880 + }, + { + "epoch": 2.13, + "learning_rate": 0.0001730977266788607, + "loss": 0.1798, + "step": 32890 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017305853148680427, + "loss": 0.1965, + "step": 32900 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017301933629474783, + "loss": 0.2018, + "step": 32910 + }, + { + "epoch": 2.14, + "learning_rate": 0.0001729801411026914, + "loss": 0.2506, + "step": 32920 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017294094591063493, + "loss": 0.1968, + "step": 32930 + }, + { + "epoch": 2.14, + "learning_rate": 0.0001729017507185785, + "loss": 0.1698, + "step": 32940 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017286255552652205, + "loss": 0.1808, + "step": 32950 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017282336033446562, + "loss": 0.2233, + "step": 32960 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017278416514240918, + "loss": 0.2786, + "step": 32970 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017274496995035274, + "loss": 0.1675, + "step": 32980 + }, + { + "epoch": 2.14, + "learning_rate": 0.0001727057747582963, + "loss": 0.199, + "step": 32990 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017266657956623987, + "loss": 0.2096, + "step": 33000 + }, + { + "epoch": 2.14, + "eval_loss": 0.2510358989238739, + "eval_runtime": 996.1813, + "eval_samples_per_second": 15.648, + "eval_steps_per_second": 1.956, + "eval_wer": 0.250067518585379, + "step": 33000 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017262738437418343, + "loss": 0.2216, + "step": 33010 + }, + { + "epoch": 2.14, + "learning_rate": 0.000172588189182127, + "loss": 0.1841, + "step": 33020 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017254899399007053, + "loss": 0.1588, + "step": 33030 + }, + { + "epoch": 2.14, + "learning_rate": 0.0001725097987980141, + "loss": 0.1794, + "step": 33040 + }, + { + "epoch": 2.14, + "learning_rate": 0.00017247060360595765, + "loss": 0.1897, + "step": 33050 + }, + { + "epoch": 2.15, + "learning_rate": 0.0001724314084139012, + "loss": 0.2546, + "step": 33060 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017239221322184477, + "loss": 0.2341, + "step": 33070 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017235301802978834, + "loss": 0.1778, + "step": 33080 + }, + { + "epoch": 2.15, + "learning_rate": 0.0001723138228377319, + "loss": 0.2139, + "step": 33090 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017227462764567546, + "loss": 0.2299, + "step": 33100 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017223543245361902, + "loss": 0.1773, + "step": 33110 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017219623726156259, + "loss": 0.2168, + "step": 33120 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017215704206950615, + "loss": 0.1632, + "step": 33130 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017211784687744968, + "loss": 0.184, + "step": 33140 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017207865168539325, + "loss": 0.1895, + "step": 33150 + }, + { + "epoch": 2.15, + "learning_rate": 0.0001720394564933368, + "loss": 0.2025, + "step": 33160 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017200026130128037, + "loss": 0.2921, + "step": 33170 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017196106610922393, + "loss": 0.2059, + "step": 33180 + }, + { + "epoch": 2.15, + "learning_rate": 0.0001719218709171675, + "loss": 0.1787, + "step": 33190 + }, + { + "epoch": 2.15, + "learning_rate": 0.00017188267572511106, + "loss": 0.1824, + "step": 33200 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017184348053305462, + "loss": 0.2015, + "step": 33210 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017180428534099818, + "loss": 0.2267, + "step": 33220 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017176509014894174, + "loss": 0.1837, + "step": 33230 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017172589495688525, + "loss": 0.1938, + "step": 33240 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017168669976482881, + "loss": 0.1774, + "step": 33250 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017164750457277238, + "loss": 0.206, + "step": 33260 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017160830938071594, + "loss": 0.231, + "step": 33270 + }, + { + "epoch": 2.16, + "learning_rate": 0.0001715691141886595, + "loss": 0.1669, + "step": 33280 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017152991899660306, + "loss": 0.1864, + "step": 33290 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017149072380454663, + "loss": 0.1649, + "step": 33300 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017145152861249016, + "loss": 0.199, + "step": 33310 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017141233342043372, + "loss": 0.2574, + "step": 33320 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017137313822837729, + "loss": 0.1556, + "step": 33330 + }, + { + "epoch": 2.16, + "learning_rate": 0.00017133394303632085, + "loss": 0.175, + "step": 33340 + }, + { + "epoch": 2.16, + "learning_rate": 0.0001712947478442644, + "loss": 0.2182, + "step": 33350 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017125555265220797, + "loss": 0.2393, + "step": 33360 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017121635746015153, + "loss": 0.2804, + "step": 33370 + }, + { + "epoch": 2.17, + "learning_rate": 0.0001711771622680951, + "loss": 0.1543, + "step": 33380 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017113796707603866, + "loss": 0.1845, + "step": 33390 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017109877188398222, + "loss": 0.2115, + "step": 33400 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017105957669192578, + "loss": 0.2262, + "step": 33410 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017102038149986932, + "loss": 0.2555, + "step": 33420 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017098118630781288, + "loss": 0.161, + "step": 33430 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017094199111575644, + "loss": 0.1911, + "step": 33440 + }, + { + "epoch": 2.17, + "learning_rate": 0.0001709027959237, + "loss": 0.1655, + "step": 33450 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017086360073164357, + "loss": 0.2197, + "step": 33460 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017082440553958713, + "loss": 0.2726, + "step": 33470 + }, + { + "epoch": 2.17, + "learning_rate": 0.0001707852103475307, + "loss": 0.1658, + "step": 33480 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017074601515547426, + "loss": 0.1838, + "step": 33490 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017070681996341782, + "loss": 0.1884, + "step": 33500 + }, + { + "epoch": 2.17, + "learning_rate": 0.00017066762477136138, + "loss": 0.2233, + "step": 33510 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017062842957930491, + "loss": 0.2339, + "step": 33520 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017058923438724848, + "loss": 0.1789, + "step": 33530 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017055003919519204, + "loss": 0.17, + "step": 33540 + }, + { + "epoch": 2.18, + "learning_rate": 0.0001705108440031356, + "loss": 0.2136, + "step": 33550 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017047164881107916, + "loss": 0.2013, + "step": 33560 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017043245361902273, + "loss": 0.2358, + "step": 33570 + }, + { + "epoch": 2.18, + "learning_rate": 0.0001703932584269663, + "loss": 0.1591, + "step": 33580 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017035406323490985, + "loss": 0.1923, + "step": 33590 + }, + { + "epoch": 2.18, + "learning_rate": 0.0001703148680428534, + "loss": 0.205, + "step": 33600 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017027567285079698, + "loss": 0.2145, + "step": 33610 + }, + { + "epoch": 2.18, + "learning_rate": 0.0001702364776587405, + "loss": 0.2177, + "step": 33620 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017019728246668407, + "loss": 0.1455, + "step": 33630 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017015808727462764, + "loss": 0.1509, + "step": 33640 + }, + { + "epoch": 2.18, + "learning_rate": 0.0001701188920825712, + "loss": 0.1884, + "step": 33650 + }, + { + "epoch": 2.18, + "learning_rate": 0.00017007969689051476, + "loss": 0.2275, + "step": 33660 + }, + { + "epoch": 2.19, + "learning_rate": 0.00017004050169845832, + "loss": 0.3014, + "step": 33670 + }, + { + "epoch": 2.19, + "learning_rate": 0.00017000130650640188, + "loss": 0.1594, + "step": 33680 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016996211131434545, + "loss": 0.1922, + "step": 33690 + }, + { + "epoch": 2.19, + "learning_rate": 0.000169922916122289, + "loss": 0.1884, + "step": 33700 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016988372093023257, + "loss": 0.1862, + "step": 33710 + }, + { + "epoch": 2.19, + "learning_rate": 0.0001698445257381761, + "loss": 0.2677, + "step": 33720 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016980533054611964, + "loss": 0.1722, + "step": 33730 + }, + { + "epoch": 2.19, + "learning_rate": 0.0001697661353540632, + "loss": 0.1646, + "step": 33740 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016972694016200677, + "loss": 0.1774, + "step": 33750 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016968774496995033, + "loss": 0.1946, + "step": 33760 + }, + { + "epoch": 2.19, + "learning_rate": 0.0001696485497778939, + "loss": 0.1898, + "step": 33770 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016960935458583745, + "loss": 0.1551, + "step": 33780 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016957015939378102, + "loss": 0.1827, + "step": 33790 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016953096420172455, + "loss": 0.184, + "step": 33800 + }, + { + "epoch": 2.19, + "learning_rate": 0.0001694917690096681, + "loss": 0.2725, + "step": 33810 + }, + { + "epoch": 2.19, + "learning_rate": 0.00016945257381761168, + "loss": 0.2413, + "step": 33820 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016941337862555524, + "loss": 0.1571, + "step": 33830 + }, + { + "epoch": 2.2, + "learning_rate": 0.0001693741834334988, + "loss": 0.1686, + "step": 33840 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016933498824144236, + "loss": 0.1852, + "step": 33850 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016929579304938592, + "loss": 0.2077, + "step": 33860 + }, + { + "epoch": 2.2, + "learning_rate": 0.0001692565978573295, + "loss": 0.2281, + "step": 33870 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016921740266527305, + "loss": 0.1799, + "step": 33880 + }, + { + "epoch": 2.2, + "learning_rate": 0.0001691782074732166, + "loss": 0.1652, + "step": 33890 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016913901228116015, + "loss": 0.1712, + "step": 33900 + }, + { + "epoch": 2.2, + "learning_rate": 0.0001690998170891037, + "loss": 0.1906, + "step": 33910 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016906062189704727, + "loss": 0.2245, + "step": 33920 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016902142670499083, + "loss": 0.155, + "step": 33930 + }, + { + "epoch": 2.2, + "learning_rate": 0.0001689822315129344, + "loss": 0.1761, + "step": 33940 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016894303632087796, + "loss": 0.1947, + "step": 33950 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016890384112882152, + "loss": 0.1794, + "step": 33960 + }, + { + "epoch": 2.2, + "learning_rate": 0.00016886464593676508, + "loss": 0.2497, + "step": 33970 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016882545074470864, + "loss": 0.1622, + "step": 33980 + }, + { + "epoch": 2.21, + "learning_rate": 0.0001687862555526522, + "loss": 0.1858, + "step": 33990 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016874706036059574, + "loss": 0.1808, + "step": 34000 + }, + { + "epoch": 2.21, + "eval_loss": 0.24959716200828552, + "eval_runtime": 994.8108, + "eval_samples_per_second": 15.669, + "eval_steps_per_second": 1.959, + "eval_wer": 0.2525408309761055, + "step": 34000 + }, + { + "epoch": 2.21, + "learning_rate": 0.0001687078651685393, + "loss": 0.2418, + "step": 34010 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016866866997648287, + "loss": 0.256, + "step": 34020 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016862947478442643, + "loss": 0.1642, + "step": 34030 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016859027959237, + "loss": 0.1492, + "step": 34040 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016855108440031355, + "loss": 0.168, + "step": 34050 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016851188920825712, + "loss": 0.2142, + "step": 34060 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016847269401620068, + "loss": 0.2966, + "step": 34070 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016843349882414424, + "loss": 0.1646, + "step": 34080 + }, + { + "epoch": 2.21, + "learning_rate": 0.0001683943036320878, + "loss": 0.1703, + "step": 34090 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016835510844003134, + "loss": 0.1841, + "step": 34100 + }, + { + "epoch": 2.21, + "learning_rate": 0.0001683159132479749, + "loss": 0.2043, + "step": 34110 + }, + { + "epoch": 2.21, + "learning_rate": 0.00016827671805591846, + "loss": 0.2219, + "step": 34120 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016823752286386202, + "loss": 0.1526, + "step": 34130 + }, + { + "epoch": 2.22, + "learning_rate": 0.0001681983276718056, + "loss": 0.1658, + "step": 34140 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016815913247974915, + "loss": 0.197, + "step": 34150 + }, + { + "epoch": 2.22, + "learning_rate": 0.0001681199372876927, + "loss": 0.2419, + "step": 34160 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016808074209563627, + "loss": 0.265, + "step": 34170 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016804154690357984, + "loss": 0.1663, + "step": 34180 + }, + { + "epoch": 2.22, + "learning_rate": 0.0001680023517115234, + "loss": 0.1925, + "step": 34190 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016796315651946693, + "loss": 0.1896, + "step": 34200 + }, + { + "epoch": 2.22, + "learning_rate": 0.0001679239613274105, + "loss": 0.1933, + "step": 34210 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016788476613535406, + "loss": 0.2592, + "step": 34220 + }, + { + "epoch": 2.22, + "learning_rate": 0.0001678455709432976, + "loss": 0.1744, + "step": 34230 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016780637575124116, + "loss": 0.1725, + "step": 34240 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016776718055918472, + "loss": 0.1613, + "step": 34250 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016772798536712828, + "loss": 0.1907, + "step": 34260 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016768879017507184, + "loss": 0.2424, + "step": 34270 + }, + { + "epoch": 2.22, + "learning_rate": 0.00016764959498301538, + "loss": 0.1629, + "step": 34280 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016761039979095894, + "loss": 0.1456, + "step": 34290 + }, + { + "epoch": 2.23, + "learning_rate": 0.0001675712045989025, + "loss": 0.1775, + "step": 34300 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016753200940684606, + "loss": 0.1989, + "step": 34310 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016749281421478963, + "loss": 0.1992, + "step": 34320 + }, + { + "epoch": 2.23, + "learning_rate": 0.0001674536190227332, + "loss": 0.1591, + "step": 34330 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016741442383067675, + "loss": 0.2061, + "step": 34340 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016737522863862031, + "loss": 0.1847, + "step": 34350 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016733603344656388, + "loss": 0.2025, + "step": 34360 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016729683825450744, + "loss": 0.2247, + "step": 34370 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016725764306245097, + "loss": 0.1506, + "step": 34380 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016721844787039454, + "loss": 0.2128, + "step": 34390 + }, + { + "epoch": 2.23, + "learning_rate": 0.0001671792526783381, + "loss": 0.194, + "step": 34400 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016714005748628166, + "loss": 0.2336, + "step": 34410 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016710086229422522, + "loss": 0.2342, + "step": 34420 + }, + { + "epoch": 2.23, + "learning_rate": 0.00016706166710216879, + "loss": 0.1748, + "step": 34430 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016702247191011235, + "loss": 0.1637, + "step": 34440 + }, + { + "epoch": 2.24, + "learning_rate": 0.0001669832767180559, + "loss": 0.1945, + "step": 34450 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016694408152599947, + "loss": 0.247, + "step": 34460 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016690488633394303, + "loss": 0.2656, + "step": 34470 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016686569114188657, + "loss": 0.1584, + "step": 34480 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016682649594983013, + "loss": 0.1523, + "step": 34490 + }, + { + "epoch": 2.24, + "learning_rate": 0.0001667873007577737, + "loss": 0.1843, + "step": 34500 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016674810556571726, + "loss": 0.196, + "step": 34510 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016670891037366082, + "loss": 0.2787, + "step": 34520 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016666971518160438, + "loss": 0.1852, + "step": 34530 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016663051998954794, + "loss": 0.1478, + "step": 34540 + }, + { + "epoch": 2.24, + "learning_rate": 0.0001665913247974915, + "loss": 0.1851, + "step": 34550 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016655212960543507, + "loss": 0.1896, + "step": 34560 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016651293441337863, + "loss": 0.2386, + "step": 34570 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016647373922132217, + "loss": 0.1506, + "step": 34580 + }, + { + "epoch": 2.24, + "learning_rate": 0.00016643454402926573, + "loss": 0.1543, + "step": 34590 + }, + { + "epoch": 2.25, + "learning_rate": 0.0001663953488372093, + "loss": 0.1808, + "step": 34600 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016635615364515285, + "loss": 0.2023, + "step": 34610 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016631695845309641, + "loss": 0.2465, + "step": 34620 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016627776326103998, + "loss": 0.1789, + "step": 34630 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016623856806898354, + "loss": 0.1734, + "step": 34640 + }, + { + "epoch": 2.25, + "learning_rate": 0.0001661993728769271, + "loss": 0.1753, + "step": 34650 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016616017768487066, + "loss": 0.2135, + "step": 34660 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016612098249281423, + "loss": 0.2578, + "step": 34670 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016608178730075776, + "loss": 0.1621, + "step": 34680 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016604259210870132, + "loss": 0.1888, + "step": 34690 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016600339691664489, + "loss": 0.1821, + "step": 34700 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016596420172458845, + "loss": 0.2094, + "step": 34710 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016592500653253198, + "loss": 0.247, + "step": 34720 + }, + { + "epoch": 2.25, + "learning_rate": 0.00016588581134047555, + "loss": 0.1751, + "step": 34730 + }, + { + "epoch": 2.25, + "learning_rate": 0.0001658466161484191, + "loss": 0.1657, + "step": 34740 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016580742095636267, + "loss": 0.1928, + "step": 34750 + }, + { + "epoch": 2.26, + "learning_rate": 0.0001657682257643062, + "loss": 0.2056, + "step": 34760 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016572903057224977, + "loss": 0.2829, + "step": 34770 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016568983538019333, + "loss": 0.147, + "step": 34780 + }, + { + "epoch": 2.26, + "learning_rate": 0.0001656506401881369, + "loss": 0.1594, + "step": 34790 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016561144499608045, + "loss": 0.2004, + "step": 34800 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016557224980402402, + "loss": 0.2085, + "step": 34810 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016553305461196758, + "loss": 0.2331, + "step": 34820 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016549385941991114, + "loss": 0.1647, + "step": 34830 + }, + { + "epoch": 2.26, + "learning_rate": 0.0001654546642278547, + "loss": 0.1666, + "step": 34840 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016541546903579827, + "loss": 0.1805, + "step": 34850 + }, + { + "epoch": 2.26, + "learning_rate": 0.0001653762738437418, + "loss": 0.2106, + "step": 34860 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016533707865168536, + "loss": 0.2374, + "step": 34870 + }, + { + "epoch": 2.26, + "learning_rate": 0.00016529788345962893, + "loss": 0.14, + "step": 34880 + }, + { + "epoch": 2.26, + "learning_rate": 0.0001652586882675725, + "loss": 0.1646, + "step": 34890 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016521949307551605, + "loss": 0.1918, + "step": 34900 + }, + { + "epoch": 2.27, + "learning_rate": 0.0001651802978834596, + "loss": 0.178, + "step": 34910 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016514110269140317, + "loss": 0.2301, + "step": 34920 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016510190749934674, + "loss": 0.1407, + "step": 34930 + }, + { + "epoch": 2.27, + "learning_rate": 0.0001650627123072903, + "loss": 0.2045, + "step": 34940 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016502351711523386, + "loss": 0.1992, + "step": 34950 + }, + { + "epoch": 2.27, + "learning_rate": 0.0001649843219231774, + "loss": 0.2115, + "step": 34960 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016494512673112096, + "loss": 0.2299, + "step": 34970 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016490593153906452, + "loss": 0.1782, + "step": 34980 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016486673634700808, + "loss": 0.1657, + "step": 34990 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016482754115495165, + "loss": 0.2013, + "step": 35000 + }, + { + "epoch": 2.27, + "eval_loss": 0.24518567323684692, + "eval_runtime": 999.9563, + "eval_samples_per_second": 15.589, + "eval_steps_per_second": 1.949, + "eval_wer": 0.23866753848559366, + "step": 35000 + }, + { + "epoch": 2.27, + "learning_rate": 0.0001647883459628952, + "loss": 0.2404, + "step": 35010 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016474915077083877, + "loss": 0.213, + "step": 35020 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016470995557878233, + "loss": 0.1524, + "step": 35030 + }, + { + "epoch": 2.27, + "learning_rate": 0.0001646707603867259, + "loss": 0.1582, + "step": 35040 + }, + { + "epoch": 2.27, + "learning_rate": 0.00016463156519466946, + "loss": 0.1869, + "step": 35050 + }, + { + "epoch": 2.28, + "learning_rate": 0.000164592370002613, + "loss": 0.245, + "step": 35060 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016455317481055656, + "loss": 0.2043, + "step": 35070 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016451397961850012, + "loss": 0.1621, + "step": 35080 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016447478442644368, + "loss": 0.1552, + "step": 35090 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016443558923438724, + "loss": 0.2126, + "step": 35100 + }, + { + "epoch": 2.28, + "learning_rate": 0.0001643963940423308, + "loss": 0.1911, + "step": 35110 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016435719885027437, + "loss": 0.2423, + "step": 35120 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016431800365821793, + "loss": 0.1817, + "step": 35130 + }, + { + "epoch": 2.28, + "learning_rate": 0.0001642788084661615, + "loss": 0.1659, + "step": 35140 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016423961327410505, + "loss": 0.1902, + "step": 35150 + }, + { + "epoch": 2.28, + "learning_rate": 0.0001642004180820486, + "loss": 0.1958, + "step": 35160 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016416122288999215, + "loss": 0.2236, + "step": 35170 + }, + { + "epoch": 2.28, + "learning_rate": 0.0001641220276979357, + "loss": 0.1351, + "step": 35180 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016408283250587928, + "loss": 0.1705, + "step": 35190 + }, + { + "epoch": 2.28, + "learning_rate": 0.00016404363731382284, + "loss": 0.1943, + "step": 35200 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016400444212176637, + "loss": 0.2281, + "step": 35210 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016396524692970994, + "loss": 0.2809, + "step": 35220 + }, + { + "epoch": 2.29, + "learning_rate": 0.0001639260517376535, + "loss": 0.1702, + "step": 35230 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016388685654559703, + "loss": 0.1783, + "step": 35240 + }, + { + "epoch": 2.29, + "learning_rate": 0.0001638476613535406, + "loss": 0.1971, + "step": 35250 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016380846616148416, + "loss": 0.1926, + "step": 35260 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016376927096942772, + "loss": 0.248, + "step": 35270 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016373007577737128, + "loss": 0.1672, + "step": 35280 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016369088058531484, + "loss": 0.1823, + "step": 35290 + }, + { + "epoch": 2.29, + "learning_rate": 0.0001636516853932584, + "loss": 0.1957, + "step": 35300 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016361249020120197, + "loss": 0.2195, + "step": 35310 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016357329500914553, + "loss": 0.2265, + "step": 35320 + }, + { + "epoch": 2.29, + "learning_rate": 0.0001635340998170891, + "loss": 0.1423, + "step": 35330 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016349490462503263, + "loss": 0.1775, + "step": 35340 + }, + { + "epoch": 2.29, + "learning_rate": 0.0001634557094329762, + "loss": 0.194, + "step": 35350 + }, + { + "epoch": 2.29, + "learning_rate": 0.00016341651424091975, + "loss": 0.2151, + "step": 35360 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016337731904886332, + "loss": 0.2211, + "step": 35370 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016333812385680688, + "loss": 0.1666, + "step": 35380 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016329892866475044, + "loss": 0.147, + "step": 35390 + }, + { + "epoch": 2.3, + "learning_rate": 0.000163259733472694, + "loss": 0.2062, + "step": 35400 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016322053828063756, + "loss": 0.1913, + "step": 35410 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016318134308858113, + "loss": 0.2383, + "step": 35420 + }, + { + "epoch": 2.3, + "learning_rate": 0.0001631421478965247, + "loss": 0.1803, + "step": 35430 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016310295270446822, + "loss": 0.1824, + "step": 35440 + }, + { + "epoch": 2.3, + "learning_rate": 0.0001630637575124118, + "loss": 0.1852, + "step": 35450 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016302456232035535, + "loss": 0.1963, + "step": 35460 + }, + { + "epoch": 2.3, + "learning_rate": 0.0001629853671282989, + "loss": 0.2427, + "step": 35470 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016294617193624247, + "loss": 0.1951, + "step": 35480 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016290697674418604, + "loss": 0.1905, + "step": 35490 + }, + { + "epoch": 2.3, + "learning_rate": 0.0001628677815521296, + "loss": 0.1768, + "step": 35500 + }, + { + "epoch": 2.3, + "learning_rate": 0.00016282858636007316, + "loss": 0.2145, + "step": 35510 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016278939116801672, + "loss": 0.2672, + "step": 35520 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016275019597596028, + "loss": 0.1539, + "step": 35530 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016271100078390382, + "loss": 0.1585, + "step": 35540 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016267180559184738, + "loss": 0.1852, + "step": 35550 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016263261039979094, + "loss": 0.2004, + "step": 35560 + }, + { + "epoch": 2.31, + "learning_rate": 0.0001625934152077345, + "loss": 0.2386, + "step": 35570 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016255422001567807, + "loss": 0.194, + "step": 35580 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016251502482362163, + "loss": 0.1874, + "step": 35590 + }, + { + "epoch": 2.31, + "learning_rate": 0.0001624758296315652, + "loss": 0.2147, + "step": 35600 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016243663443950876, + "loss": 0.1972, + "step": 35610 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016239743924745232, + "loss": 0.249, + "step": 35620 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016235824405539588, + "loss": 0.1615, + "step": 35630 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016231904886333942, + "loss": 0.1848, + "step": 35640 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016227985367128298, + "loss": 0.2253, + "step": 35650 + }, + { + "epoch": 2.31, + "learning_rate": 0.00016224065847922654, + "loss": 0.2146, + "step": 35660 + }, + { + "epoch": 2.32, + "learning_rate": 0.0001622014632871701, + "loss": 0.302, + "step": 35670 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016216226809511367, + "loss": 0.1535, + "step": 35680 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016212307290305723, + "loss": 0.1616, + "step": 35690 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016208387771100076, + "loss": 0.198, + "step": 35700 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016204468251894432, + "loss": 0.2018, + "step": 35710 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016200548732688786, + "loss": 0.233, + "step": 35720 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016196629213483142, + "loss": 0.1637, + "step": 35730 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016192709694277498, + "loss": 0.1728, + "step": 35740 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016188790175071855, + "loss": 0.1554, + "step": 35750 + }, + { + "epoch": 2.32, + "learning_rate": 0.0001618487065586621, + "loss": 0.2117, + "step": 35760 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016180951136660567, + "loss": 0.2054, + "step": 35770 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016177031617454923, + "loss": 0.1532, + "step": 35780 + }, + { + "epoch": 2.32, + "learning_rate": 0.0001617311209824928, + "loss": 0.1869, + "step": 35790 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016169192579043636, + "loss": 0.1868, + "step": 35800 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016165273059837992, + "loss": 0.192, + "step": 35810 + }, + { + "epoch": 2.32, + "learning_rate": 0.00016161353540632346, + "loss": 0.2326, + "step": 35820 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016157434021426702, + "loss": 0.1613, + "step": 35830 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016153514502221058, + "loss": 0.1424, + "step": 35840 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016149594983015414, + "loss": 0.1795, + "step": 35850 + }, + { + "epoch": 2.33, + "learning_rate": 0.0001614567546380977, + "loss": 0.2491, + "step": 35860 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016141755944604127, + "loss": 0.2094, + "step": 35870 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016137836425398483, + "loss": 0.1669, + "step": 35880 + }, + { + "epoch": 2.33, + "learning_rate": 0.0001613391690619284, + "loss": 0.1776, + "step": 35890 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016129997386987195, + "loss": 0.2017, + "step": 35900 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016126077867781552, + "loss": 0.2022, + "step": 35910 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016122158348575905, + "loss": 0.2199, + "step": 35920 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016118238829370261, + "loss": 0.1701, + "step": 35930 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016114319310164618, + "loss": 0.1571, + "step": 35940 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016110399790958974, + "loss": 0.1702, + "step": 35950 + }, + { + "epoch": 2.33, + "learning_rate": 0.0001610648027175333, + "loss": 0.2352, + "step": 35960 + }, + { + "epoch": 2.33, + "learning_rate": 0.00016102560752547686, + "loss": 0.2748, + "step": 35970 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016098641233342043, + "loss": 0.1726, + "step": 35980 + }, + { + "epoch": 2.34, + "learning_rate": 0.000160947217141364, + "loss": 0.1832, + "step": 35990 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016090802194930755, + "loss": 0.1996, + "step": 36000 + }, + { + "epoch": 2.34, + "eval_loss": 0.24311493337154388, + "eval_runtime": 1000.2388, + "eval_samples_per_second": 15.584, + "eval_steps_per_second": 1.949, + "eval_wer": 0.23561143409475346, + "step": 36000 + }, + { + "epoch": 2.34, + "learning_rate": 0.0001608688267572511, + "loss": 0.2388, + "step": 36010 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016082963156519467, + "loss": 0.2526, + "step": 36020 + }, + { + "epoch": 2.34, + "learning_rate": 0.0001607904363731382, + "loss": 0.1746, + "step": 36030 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016075124118108177, + "loss": 0.1586, + "step": 36040 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016071204598902533, + "loss": 0.1544, + "step": 36050 + }, + { + "epoch": 2.34, + "learning_rate": 0.0001606728507969689, + "loss": 0.2244, + "step": 36060 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016063365560491246, + "loss": 0.2291, + "step": 36070 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016059446041285602, + "loss": 0.1748, + "step": 36080 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016055526522079958, + "loss": 0.1852, + "step": 36090 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016051607002874315, + "loss": 0.1894, + "step": 36100 + }, + { + "epoch": 2.34, + "learning_rate": 0.0001604768748366867, + "loss": 0.2127, + "step": 36110 + }, + { + "epoch": 2.34, + "learning_rate": 0.00016043767964463027, + "loss": 0.2519, + "step": 36120 + }, + { + "epoch": 2.34, + "learning_rate": 0.0001603984844525738, + "loss": 0.1846, + "step": 36130 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016035928926051737, + "loss": 0.1394, + "step": 36140 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016032009406846093, + "loss": 0.1912, + "step": 36150 + }, + { + "epoch": 2.35, + "learning_rate": 0.0001602808988764045, + "loss": 0.2074, + "step": 36160 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016024170368434805, + "loss": 0.2045, + "step": 36170 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016020250849229162, + "loss": 0.1423, + "step": 36180 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016016331330023515, + "loss": 0.1798, + "step": 36190 + }, + { + "epoch": 2.35, + "learning_rate": 0.0001601241181081787, + "loss": 0.187, + "step": 36200 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016008492291612225, + "loss": 0.2197, + "step": 36210 + }, + { + "epoch": 2.35, + "learning_rate": 0.0001600457277240658, + "loss": 0.2392, + "step": 36220 + }, + { + "epoch": 2.35, + "learning_rate": 0.00016000653253200937, + "loss": 0.1732, + "step": 36230 + }, + { + "epoch": 2.35, + "learning_rate": 0.00015996733733995294, + "loss": 0.1911, + "step": 36240 + }, + { + "epoch": 2.35, + "learning_rate": 0.0001599281421478965, + "loss": 0.1716, + "step": 36250 + }, + { + "epoch": 2.35, + "learning_rate": 0.00015988894695584006, + "loss": 0.2109, + "step": 36260 + }, + { + "epoch": 2.35, + "learning_rate": 0.00015984975176378362, + "loss": 0.1985, + "step": 36270 + }, + { + "epoch": 2.35, + "learning_rate": 0.00015981055657172719, + "loss": 0.1387, + "step": 36280 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015977136137967075, + "loss": 0.1679, + "step": 36290 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015973216618761428, + "loss": 0.1923, + "step": 36300 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015969297099555785, + "loss": 0.248, + "step": 36310 + }, + { + "epoch": 2.36, + "learning_rate": 0.0001596537758035014, + "loss": 0.2276, + "step": 36320 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015961458061144497, + "loss": 0.1542, + "step": 36330 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015957538541938853, + "loss": 0.1558, + "step": 36340 + }, + { + "epoch": 2.36, + "learning_rate": 0.0001595361902273321, + "loss": 0.2137, + "step": 36350 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015949699503527566, + "loss": 0.248, + "step": 36360 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015945779984321922, + "loss": 0.2348, + "step": 36370 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015941860465116278, + "loss": 0.1636, + "step": 36380 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015937940945910634, + "loss": 0.1749, + "step": 36390 + }, + { + "epoch": 2.36, + "learning_rate": 0.0001593402142670499, + "loss": 0.1862, + "step": 36400 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015930101907499344, + "loss": 0.2228, + "step": 36410 + }, + { + "epoch": 2.36, + "learning_rate": 0.000159261823882937, + "loss": 0.1989, + "step": 36420 + }, + { + "epoch": 2.36, + "learning_rate": 0.00015922262869088057, + "loss": 0.1576, + "step": 36430 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015918343349882413, + "loss": 0.1741, + "step": 36440 + }, + { + "epoch": 2.37, + "learning_rate": 0.0001591442383067677, + "loss": 0.1757, + "step": 36450 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015910504311471125, + "loss": 0.2283, + "step": 36460 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015906584792265481, + "loss": 0.1958, + "step": 36470 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015902665273059838, + "loss": 0.1705, + "step": 36480 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015898745753854194, + "loss": 0.1678, + "step": 36490 + }, + { + "epoch": 2.37, + "learning_rate": 0.0001589482623464855, + "loss": 0.2011, + "step": 36500 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015890906715442904, + "loss": 0.191, + "step": 36510 + }, + { + "epoch": 2.37, + "learning_rate": 0.0001588698719623726, + "loss": 0.2236, + "step": 36520 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015883067677031616, + "loss": 0.1609, + "step": 36530 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015879148157825972, + "loss": 0.1861, + "step": 36540 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015875228638620329, + "loss": 0.2132, + "step": 36550 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015871701071335247, + "loss": 0.2349, + "step": 36560 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015867781552129603, + "loss": 0.2379, + "step": 36570 + }, + { + "epoch": 2.37, + "learning_rate": 0.0001586386203292396, + "loss": 0.1587, + "step": 36580 + }, + { + "epoch": 2.37, + "learning_rate": 0.00015859942513718315, + "loss": 0.1594, + "step": 36590 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015856022994512671, + "loss": 0.2052, + "step": 36600 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015852103475307028, + "loss": 0.1816, + "step": 36610 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015848183956101384, + "loss": 0.2224, + "step": 36620 + }, + { + "epoch": 2.38, + "learning_rate": 0.0001584426443689574, + "loss": 0.1773, + "step": 36630 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015840344917690096, + "loss": 0.1397, + "step": 36640 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015836425398484453, + "loss": 0.1821, + "step": 36650 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015832505879278806, + "loss": 0.213, + "step": 36660 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015828586360073162, + "loss": 0.2656, + "step": 36670 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015824666840867519, + "loss": 0.1417, + "step": 36680 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015820747321661875, + "loss": 0.1685, + "step": 36690 + }, + { + "epoch": 2.38, + "learning_rate": 0.0001581682780245623, + "loss": 0.2035, + "step": 36700 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015812908283250587, + "loss": 0.2335, + "step": 36710 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015808988764044943, + "loss": 0.2257, + "step": 36720 + }, + { + "epoch": 2.38, + "learning_rate": 0.000158050692448393, + "loss": 0.1674, + "step": 36730 + }, + { + "epoch": 2.38, + "learning_rate": 0.00015801149725633656, + "loss": 0.1889, + "step": 36740 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015797230206428012, + "loss": 0.1774, + "step": 36750 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015793310687222366, + "loss": 0.1955, + "step": 36760 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015789391168016722, + "loss": 0.2719, + "step": 36770 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015785471648811078, + "loss": 0.1886, + "step": 36780 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015781552129605434, + "loss": 0.1678, + "step": 36790 + }, + { + "epoch": 2.39, + "learning_rate": 0.0001577763261039979, + "loss": 0.1711, + "step": 36800 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015773713091194147, + "loss": 0.1994, + "step": 36810 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015769793571988503, + "loss": 0.2635, + "step": 36820 + }, + { + "epoch": 2.39, + "learning_rate": 0.0001576587405278286, + "loss": 0.1672, + "step": 36830 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015761954533577216, + "loss": 0.1888, + "step": 36840 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015758035014371572, + "loss": 0.1708, + "step": 36850 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015754115495165925, + "loss": 0.2165, + "step": 36860 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015750195975960281, + "loss": 0.2522, + "step": 36870 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015746276456754635, + "loss": 0.1573, + "step": 36880 + }, + { + "epoch": 2.39, + "learning_rate": 0.0001574235693754899, + "loss": 0.1777, + "step": 36890 + }, + { + "epoch": 2.39, + "learning_rate": 0.00015738437418343347, + "loss": 0.1683, + "step": 36900 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015734517899137704, + "loss": 0.2114, + "step": 36910 + }, + { + "epoch": 2.4, + "learning_rate": 0.0001573059837993206, + "loss": 0.2174, + "step": 36920 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015726678860726416, + "loss": 0.146, + "step": 36930 + }, + { + "epoch": 2.4, + "learning_rate": 0.0001572275934152077, + "loss": 0.1572, + "step": 36940 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015718839822315126, + "loss": 0.1929, + "step": 36950 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015714920303109482, + "loss": 0.221, + "step": 36960 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015711000783903838, + "loss": 0.2533, + "step": 36970 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015707081264698195, + "loss": 0.1474, + "step": 36980 + }, + { + "epoch": 2.4, + "learning_rate": 0.0001570316174549255, + "loss": 0.185, + "step": 36990 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015699242226286907, + "loss": 0.167, + "step": 37000 + }, + { + "epoch": 2.4, + "eval_loss": 0.2403496354818344, + "eval_runtime": 1000.4154, + "eval_samples_per_second": 15.582, + "eval_steps_per_second": 1.948, + "eval_wer": 0.23371380648462708, + "step": 37000 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015695322707081263, + "loss": 0.2165, + "step": 37010 + }, + { + "epoch": 2.4, + "learning_rate": 0.0001569140318787562, + "loss": 0.1901, + "step": 37020 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015687483668669976, + "loss": 0.1499, + "step": 37030 + }, + { + "epoch": 2.4, + "learning_rate": 0.0001568356414946433, + "loss": 0.1424, + "step": 37040 + }, + { + "epoch": 2.4, + "learning_rate": 0.00015679644630258685, + "loss": 0.1537, + "step": 37050 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015675725111053042, + "loss": 0.1653, + "step": 37060 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015671805591847398, + "loss": 0.2106, + "step": 37070 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015667886072641754, + "loss": 0.1331, + "step": 37080 + }, + { + "epoch": 2.41, + "learning_rate": 0.0001566396655343611, + "loss": 0.1756, + "step": 37090 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015660047034230467, + "loss": 0.1854, + "step": 37100 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015656127515024823, + "loss": 0.2142, + "step": 37110 + }, + { + "epoch": 2.41, + "learning_rate": 0.0001565220799581918, + "loss": 0.2685, + "step": 37120 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015648288476613535, + "loss": 0.1742, + "step": 37130 + }, + { + "epoch": 2.41, + "learning_rate": 0.0001564436895740789, + "loss": 0.192, + "step": 37140 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015640449438202245, + "loss": 0.196, + "step": 37150 + }, + { + "epoch": 2.41, + "learning_rate": 0.000156365299189966, + "loss": 0.224, + "step": 37160 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015632610399790958, + "loss": 0.2536, + "step": 37170 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015628690880585314, + "loss": 0.1745, + "step": 37180 + }, + { + "epoch": 2.41, + "learning_rate": 0.0001562477136137967, + "loss": 0.1594, + "step": 37190 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015620851842174026, + "loss": 0.171, + "step": 37200 + }, + { + "epoch": 2.41, + "learning_rate": 0.00015616932322968382, + "loss": 0.1766, + "step": 37210 + }, + { + "epoch": 2.42, + "learning_rate": 0.0001561301280376274, + "loss": 0.2148, + "step": 37220 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015609093284557095, + "loss": 0.1693, + "step": 37230 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015605173765351448, + "loss": 0.1728, + "step": 37240 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015601254246145805, + "loss": 0.2074, + "step": 37250 + }, + { + "epoch": 2.42, + "learning_rate": 0.0001559733472694016, + "loss": 0.2535, + "step": 37260 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015593415207734517, + "loss": 0.2341, + "step": 37270 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015589495688528873, + "loss": 0.1513, + "step": 37280 + }, + { + "epoch": 2.42, + "learning_rate": 0.0001558557616932323, + "loss": 0.1655, + "step": 37290 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015581656650117586, + "loss": 0.1967, + "step": 37300 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015577737130911942, + "loss": 0.1781, + "step": 37310 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015573817611706298, + "loss": 0.2281, + "step": 37320 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015569898092500654, + "loss": 0.139, + "step": 37330 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015565978573295008, + "loss": 0.1524, + "step": 37340 + }, + { + "epoch": 2.42, + "learning_rate": 0.00015562059054089364, + "loss": 0.185, + "step": 37350 + }, + { + "epoch": 2.42, + "learning_rate": 0.0001555813953488372, + "loss": 0.2039, + "step": 37360 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015554220015678074, + "loss": 0.2307, + "step": 37370 + }, + { + "epoch": 2.43, + "learning_rate": 0.0001555030049647243, + "loss": 0.1694, + "step": 37380 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015546380977266786, + "loss": 0.1732, + "step": 37390 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015542461458061143, + "loss": 0.2151, + "step": 37400 + }, + { + "epoch": 2.43, + "learning_rate": 0.000155385419388555, + "loss": 0.1848, + "step": 37410 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015534622419649852, + "loss": 0.2673, + "step": 37420 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015530702900444209, + "loss": 0.148, + "step": 37430 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015526783381238565, + "loss": 0.1451, + "step": 37440 + }, + { + "epoch": 2.43, + "learning_rate": 0.0001552286386203292, + "loss": 0.1522, + "step": 37450 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015518944342827277, + "loss": 0.2171, + "step": 37460 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015515024823621634, + "loss": 0.2259, + "step": 37470 + }, + { + "epoch": 2.43, + "learning_rate": 0.0001551110530441599, + "loss": 0.1345, + "step": 37480 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015507185785210346, + "loss": 0.1829, + "step": 37490 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015503266266004702, + "loss": 0.1591, + "step": 37500 + }, + { + "epoch": 2.43, + "learning_rate": 0.00015499346746799058, + "loss": 0.2184, + "step": 37510 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015495427227593412, + "loss": 0.2573, + "step": 37520 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015491507708387768, + "loss": 0.1658, + "step": 37530 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015487588189182124, + "loss": 0.1653, + "step": 37540 + }, + { + "epoch": 2.44, + "learning_rate": 0.0001548366866997648, + "loss": 0.1895, + "step": 37550 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015479749150770837, + "loss": 0.2297, + "step": 37560 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015475829631565193, + "loss": 0.2564, + "step": 37570 + }, + { + "epoch": 2.44, + "learning_rate": 0.0001547191011235955, + "loss": 0.1739, + "step": 37580 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015467990593153906, + "loss": 0.1799, + "step": 37590 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015464071073948262, + "loss": 0.172, + "step": 37600 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015460151554742618, + "loss": 0.168, + "step": 37610 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015456232035536972, + "loss": 0.2371, + "step": 37620 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015452312516331328, + "loss": 0.1687, + "step": 37630 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015448392997125684, + "loss": 0.1685, + "step": 37640 + }, + { + "epoch": 2.44, + "learning_rate": 0.0001544447347792004, + "loss": 0.1564, + "step": 37650 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015440553958714396, + "loss": 0.207, + "step": 37660 + }, + { + "epoch": 2.44, + "learning_rate": 0.00015436634439508753, + "loss": 0.2545, + "step": 37670 + }, + { + "epoch": 2.45, + "learning_rate": 0.0001543271492030311, + "loss": 0.1424, + "step": 37680 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015428795401097465, + "loss": 0.1692, + "step": 37690 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015424875881891821, + "loss": 0.1481, + "step": 37700 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015420956362686178, + "loss": 0.1966, + "step": 37710 + }, + { + "epoch": 2.45, + "learning_rate": 0.0001541703684348053, + "loss": 0.2985, + "step": 37720 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015413117324274887, + "loss": 0.1742, + "step": 37730 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015409197805069244, + "loss": 0.1522, + "step": 37740 + }, + { + "epoch": 2.45, + "learning_rate": 0.000154052782858636, + "loss": 0.2008, + "step": 37750 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015401358766657956, + "loss": 0.2376, + "step": 37760 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015397439247452312, + "loss": 0.2941, + "step": 37770 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015393519728246669, + "loss": 0.1624, + "step": 37780 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015389600209041025, + "loss": 0.1326, + "step": 37790 + }, + { + "epoch": 2.45, + "learning_rate": 0.0001538568068983538, + "loss": 0.1653, + "step": 37800 + }, + { + "epoch": 2.45, + "learning_rate": 0.00015381761170629737, + "loss": 0.2255, + "step": 37810 + }, + { + "epoch": 2.45, + "learning_rate": 0.0001537784165142409, + "loss": 0.2182, + "step": 37820 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015373922132218447, + "loss": 0.1707, + "step": 37830 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015370002613012803, + "loss": 0.139, + "step": 37840 + }, + { + "epoch": 2.46, + "learning_rate": 0.0001536608309380716, + "loss": 0.1827, + "step": 37850 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015362163574601513, + "loss": 0.2067, + "step": 37860 + }, + { + "epoch": 2.46, + "learning_rate": 0.0001535824405539587, + "loss": 0.2007, + "step": 37870 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015354324536190225, + "loss": 0.1578, + "step": 37880 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015350405016984582, + "loss": 0.1723, + "step": 37890 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015346485497778935, + "loss": 0.2044, + "step": 37900 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015342565978573291, + "loss": 0.191, + "step": 37910 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015338646459367648, + "loss": 0.229, + "step": 37920 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015334726940162004, + "loss": 0.141, + "step": 37930 + }, + { + "epoch": 2.46, + "learning_rate": 0.0001533080742095636, + "loss": 0.1533, + "step": 37940 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015326887901750716, + "loss": 0.2006, + "step": 37950 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015322968382545073, + "loss": 0.1767, + "step": 37960 + }, + { + "epoch": 2.46, + "learning_rate": 0.0001531904886333943, + "loss": 0.2354, + "step": 37970 + }, + { + "epoch": 2.46, + "learning_rate": 0.00015315129344133785, + "loss": 0.1816, + "step": 37980 + }, + { + "epoch": 2.47, + "learning_rate": 0.0001531120982492814, + "loss": 0.1762, + "step": 37990 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015307290305722495, + "loss": 0.2151, + "step": 38000 + }, + { + "epoch": 2.47, + "eval_loss": 0.24454525113105774, + "eval_runtime": 1001.4769, + "eval_samples_per_second": 15.565, + "eval_steps_per_second": 1.946, + "eval_wer": 0.23933561711987036, + "step": 38000 + }, + { + "epoch": 2.47, + "learning_rate": 0.0001530337078651685, + "loss": 0.1996, + "step": 38010 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015299451267311207, + "loss": 0.2235, + "step": 38020 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015295531748105563, + "loss": 0.1577, + "step": 38030 + }, + { + "epoch": 2.47, + "learning_rate": 0.0001529161222889992, + "loss": 0.1828, + "step": 38040 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015287692709694276, + "loss": 0.179, + "step": 38050 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015283773190488632, + "loss": 0.2006, + "step": 38060 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015279853671282988, + "loss": 0.2165, + "step": 38070 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015275934152077345, + "loss": 0.1692, + "step": 38080 + }, + { + "epoch": 2.47, + "learning_rate": 0.000152720146328717, + "loss": 0.1756, + "step": 38090 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015268095113666054, + "loss": 0.1599, + "step": 38100 + }, + { + "epoch": 2.47, + "learning_rate": 0.0001526417559446041, + "loss": 0.1954, + "step": 38110 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015260256075254767, + "loss": 0.2162, + "step": 38120 + }, + { + "epoch": 2.47, + "learning_rate": 0.00015256336556049123, + "loss": 0.1644, + "step": 38130 + }, + { + "epoch": 2.48, + "learning_rate": 0.0001525241703684348, + "loss": 0.1615, + "step": 38140 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015248497517637835, + "loss": 0.1856, + "step": 38150 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015244577998432192, + "loss": 0.2168, + "step": 38160 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015240658479226548, + "loss": 0.2076, + "step": 38170 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015236738960020904, + "loss": 0.144, + "step": 38180 + }, + { + "epoch": 2.48, + "learning_rate": 0.0001523281944081526, + "loss": 0.1657, + "step": 38190 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015228899921609617, + "loss": 0.1574, + "step": 38200 + }, + { + "epoch": 2.48, + "learning_rate": 0.0001522498040240397, + "loss": 0.2521, + "step": 38210 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015221060883198326, + "loss": 0.2172, + "step": 38220 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015217141363992683, + "loss": 0.1552, + "step": 38230 + }, + { + "epoch": 2.48, + "learning_rate": 0.0001521322184478704, + "loss": 0.1626, + "step": 38240 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015209302325581395, + "loss": 0.2049, + "step": 38250 + }, + { + "epoch": 2.48, + "learning_rate": 0.0001520538280637575, + "loss": 0.2078, + "step": 38260 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015201463287170107, + "loss": 0.302, + "step": 38270 + }, + { + "epoch": 2.48, + "learning_rate": 0.00015197543767964464, + "loss": 0.19, + "step": 38280 + }, + { + "epoch": 2.49, + "learning_rate": 0.0001519362424875882, + "loss": 0.1651, + "step": 38290 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015189704729553176, + "loss": 0.1763, + "step": 38300 + }, + { + "epoch": 2.49, + "learning_rate": 0.0001518578521034753, + "loss": 0.193, + "step": 38310 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015181865691141886, + "loss": 0.2164, + "step": 38320 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015177946171936242, + "loss": 0.1616, + "step": 38330 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015174026652730598, + "loss": 0.1806, + "step": 38340 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015170107133524952, + "loss": 0.1898, + "step": 38350 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015166187614319308, + "loss": 0.1948, + "step": 38360 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015162268095113664, + "loss": 0.2184, + "step": 38370 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015158348575908018, + "loss": 0.1495, + "step": 38380 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015154429056702374, + "loss": 0.1709, + "step": 38390 + }, + { + "epoch": 2.49, + "learning_rate": 0.0001515050953749673, + "loss": 0.2098, + "step": 38400 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015146590018291087, + "loss": 0.2506, + "step": 38410 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015142670499085443, + "loss": 0.2236, + "step": 38420 + }, + { + "epoch": 2.49, + "learning_rate": 0.000151387509798798, + "loss": 0.1746, + "step": 38430 + }, + { + "epoch": 2.49, + "learning_rate": 0.00015134831460674155, + "loss": 0.1521, + "step": 38440 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015130911941468511, + "loss": 0.1818, + "step": 38450 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015126992422262868, + "loss": 0.2248, + "step": 38460 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015123072903057224, + "loss": 0.248, + "step": 38470 + }, + { + "epoch": 2.5, + "learning_rate": 0.0001511915338385158, + "loss": 0.1585, + "step": 38480 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015115233864645934, + "loss": 0.1755, + "step": 38490 + }, + { + "epoch": 2.5, + "learning_rate": 0.0001511131434544029, + "loss": 0.2201, + "step": 38500 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015107394826234646, + "loss": 0.1886, + "step": 38510 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015103475307029002, + "loss": 0.2211, + "step": 38520 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015099555787823359, + "loss": 0.1655, + "step": 38530 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015095636268617715, + "loss": 0.1559, + "step": 38540 + }, + { + "epoch": 2.5, + "learning_rate": 0.0001509171674941207, + "loss": 0.2193, + "step": 38550 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015087797230206427, + "loss": 0.2245, + "step": 38560 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015083877711000784, + "loss": 0.2013, + "step": 38570 + }, + { + "epoch": 2.5, + "learning_rate": 0.0001507995819179514, + "loss": 0.1359, + "step": 38580 + }, + { + "epoch": 2.5, + "learning_rate": 0.00015076038672589493, + "loss": 0.1739, + "step": 38590 + }, + { + "epoch": 2.51, + "learning_rate": 0.0001507211915338385, + "loss": 0.2034, + "step": 38600 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015068199634178206, + "loss": 0.1979, + "step": 38610 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015064280114972562, + "loss": 0.2642, + "step": 38620 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015060360595766918, + "loss": 0.1444, + "step": 38630 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015056441076561274, + "loss": 0.1579, + "step": 38640 + }, + { + "epoch": 2.51, + "learning_rate": 0.0001505252155735563, + "loss": 0.1854, + "step": 38650 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015048602038149987, + "loss": 0.2056, + "step": 38660 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015044682518944343, + "loss": 0.2836, + "step": 38670 + }, + { + "epoch": 2.51, + "learning_rate": 0.000150407629997387, + "loss": 0.1796, + "step": 38680 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015036843480533053, + "loss": 0.1853, + "step": 38690 + }, + { + "epoch": 2.51, + "learning_rate": 0.0001503292396132741, + "loss": 0.1826, + "step": 38700 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015029004442121765, + "loss": 0.1993, + "step": 38710 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015025084922916122, + "loss": 0.2264, + "step": 38720 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015021165403710478, + "loss": 0.1684, + "step": 38730 + }, + { + "epoch": 2.51, + "learning_rate": 0.00015017245884504834, + "loss": 0.1582, + "step": 38740 + }, + { + "epoch": 2.51, + "learning_rate": 0.0001501332636529919, + "loss": 0.1698, + "step": 38750 + }, + { + "epoch": 2.52, + "learning_rate": 0.00015009406846093546, + "loss": 0.2081, + "step": 38760 + }, + { + "epoch": 2.52, + "learning_rate": 0.00015005487326887903, + "loss": 0.275, + "step": 38770 + }, + { + "epoch": 2.52, + "learning_rate": 0.0001500156780768226, + "loss": 0.1667, + "step": 38780 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014998040240397177, + "loss": 0.1703, + "step": 38790 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014994120721191533, + "loss": 0.1814, + "step": 38800 + }, + { + "epoch": 2.52, + "learning_rate": 0.0001499020120198589, + "loss": 0.2135, + "step": 38810 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014986281682780245, + "loss": 0.2285, + "step": 38820 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014982362163574602, + "loss": 0.1659, + "step": 38830 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014978442644368955, + "loss": 0.177, + "step": 38840 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014974523125163311, + "loss": 0.2145, + "step": 38850 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014970603605957668, + "loss": 0.1988, + "step": 38860 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014966684086752024, + "loss": 0.2431, + "step": 38870 + }, + { + "epoch": 2.52, + "learning_rate": 0.0001496276456754638, + "loss": 0.1702, + "step": 38880 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014958845048340736, + "loss": 0.163, + "step": 38890 + }, + { + "epoch": 2.52, + "learning_rate": 0.00014954925529135093, + "loss": 0.1683, + "step": 38900 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014951006009929446, + "loss": 0.2048, + "step": 38910 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014947086490723802, + "loss": 0.2374, + "step": 38920 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014943166971518159, + "loss": 0.1439, + "step": 38930 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014939247452312515, + "loss": 0.1777, + "step": 38940 + }, + { + "epoch": 2.53, + "learning_rate": 0.0001493532793310687, + "loss": 0.1749, + "step": 38950 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014931408413901227, + "loss": 0.1893, + "step": 38960 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014927488894695584, + "loss": 0.2212, + "step": 38970 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014923569375489937, + "loss": 0.168, + "step": 38980 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014919649856284293, + "loss": 0.168, + "step": 38990 + }, + { + "epoch": 2.53, + "learning_rate": 0.0001491573033707865, + "loss": 0.177, + "step": 39000 + }, + { + "epoch": 2.53, + "eval_loss": 0.2367456704378128, + "eval_runtime": 1004.6552, + "eval_samples_per_second": 15.516, + "eval_steps_per_second": 1.94, + "eval_wer": 0.23254822248439966, + "step": 39000 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014911810817873006, + "loss": 0.2332, + "step": 39010 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014907891298667362, + "loss": 0.2299, + "step": 39020 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014903971779461718, + "loss": 0.192, + "step": 39030 + }, + { + "epoch": 2.53, + "learning_rate": 0.00014900052260256074, + "loss": 0.1412, + "step": 39040 + }, + { + "epoch": 2.53, + "learning_rate": 0.0001489613274105043, + "loss": 0.1752, + "step": 39050 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014892213221844787, + "loss": 0.2004, + "step": 39060 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014888293702639143, + "loss": 0.2473, + "step": 39070 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014884374183433497, + "loss": 0.1882, + "step": 39080 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014880454664227853, + "loss": 0.1716, + "step": 39090 + }, + { + "epoch": 2.54, + "learning_rate": 0.0001487653514502221, + "loss": 0.172, + "step": 39100 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014872615625816565, + "loss": 0.2027, + "step": 39110 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014868696106610922, + "loss": 0.2378, + "step": 39120 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014864776587405278, + "loss": 0.1689, + "step": 39130 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014860857068199634, + "loss": 0.1787, + "step": 39140 + }, + { + "epoch": 2.54, + "learning_rate": 0.0001485693754899399, + "loss": 0.1846, + "step": 39150 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014853018029788344, + "loss": 0.2248, + "step": 39160 + }, + { + "epoch": 2.54, + "learning_rate": 0.000148490985105827, + "loss": 0.2189, + "step": 39170 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014845178991377056, + "loss": 0.1646, + "step": 39180 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014841259472171412, + "loss": 0.1605, + "step": 39190 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014837339952965769, + "loss": 0.1875, + "step": 39200 + }, + { + "epoch": 2.54, + "learning_rate": 0.00014833420433760125, + "loss": 0.2051, + "step": 39210 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014829500914554478, + "loss": 0.233, + "step": 39220 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014825581395348835, + "loss": 0.1642, + "step": 39230 + }, + { + "epoch": 2.55, + "learning_rate": 0.0001482166187614319, + "loss": 0.1908, + "step": 39240 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014817742356937547, + "loss": 0.1875, + "step": 39250 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014813822837731903, + "loss": 0.1906, + "step": 39260 + }, + { + "epoch": 2.55, + "learning_rate": 0.0001480990331852626, + "loss": 0.2317, + "step": 39270 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014805983799320616, + "loss": 0.1449, + "step": 39280 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014802064280114972, + "loss": 0.1552, + "step": 39290 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014798144760909328, + "loss": 0.171, + "step": 39300 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014794225241703684, + "loss": 0.2097, + "step": 39310 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014790305722498038, + "loss": 0.2677, + "step": 39320 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014786386203292394, + "loss": 0.1427, + "step": 39330 + }, + { + "epoch": 2.55, + "learning_rate": 0.0001478246668408675, + "loss": 0.1572, + "step": 39340 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014778547164881107, + "loss": 0.1686, + "step": 39350 + }, + { + "epoch": 2.55, + "learning_rate": 0.00014774627645675463, + "loss": 0.1847, + "step": 39360 + }, + { + "epoch": 2.56, + "learning_rate": 0.0001477070812646982, + "loss": 0.2386, + "step": 39370 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014766788607264175, + "loss": 0.1627, + "step": 39380 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014762869088058532, + "loss": 0.1651, + "step": 39390 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014758949568852885, + "loss": 0.1587, + "step": 39400 + }, + { + "epoch": 2.56, + "learning_rate": 0.0001475503004964724, + "loss": 0.2011, + "step": 39410 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014751110530441598, + "loss": 0.2249, + "step": 39420 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014747191011235954, + "loss": 0.1617, + "step": 39430 + }, + { + "epoch": 2.56, + "learning_rate": 0.0001474327149203031, + "loss": 0.1627, + "step": 39440 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014739351972824666, + "loss": 0.1907, + "step": 39450 + }, + { + "epoch": 2.56, + "learning_rate": 0.0001473543245361902, + "loss": 0.2129, + "step": 39460 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014731512934413376, + "loss": 0.2479, + "step": 39470 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014727593415207732, + "loss": 0.1524, + "step": 39480 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014723673896002088, + "loss": 0.1486, + "step": 39490 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014719754376796445, + "loss": 0.1544, + "step": 39500 + }, + { + "epoch": 2.56, + "learning_rate": 0.000147158348575908, + "loss": 0.1736, + "step": 39510 + }, + { + "epoch": 2.56, + "learning_rate": 0.00014711915338385157, + "loss": 0.1892, + "step": 39520 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014707995819179513, + "loss": 0.153, + "step": 39530 + }, + { + "epoch": 2.57, + "learning_rate": 0.0001470407629997387, + "loss": 0.1838, + "step": 39540 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014700156780768226, + "loss": 0.1732, + "step": 39550 + }, + { + "epoch": 2.57, + "learning_rate": 0.0001469623726156258, + "loss": 0.1755, + "step": 39560 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014692317742356936, + "loss": 0.2137, + "step": 39570 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014688398223151292, + "loss": 0.1522, + "step": 39580 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014684478703945648, + "loss": 0.1523, + "step": 39590 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014680559184740004, + "loss": 0.1601, + "step": 39600 + }, + { + "epoch": 2.57, + "learning_rate": 0.0001467663966553436, + "loss": 0.2, + "step": 39610 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014672720146328717, + "loss": 0.2225, + "step": 39620 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014668800627123073, + "loss": 0.1367, + "step": 39630 + }, + { + "epoch": 2.57, + "learning_rate": 0.0001466488110791743, + "loss": 0.1584, + "step": 39640 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014660961588711783, + "loss": 0.1818, + "step": 39650 + }, + { + "epoch": 2.57, + "learning_rate": 0.0001465704206950614, + "loss": 0.2021, + "step": 39660 + }, + { + "epoch": 2.57, + "learning_rate": 0.00014653122550300495, + "loss": 0.2493, + "step": 39670 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014649203031094851, + "loss": 0.1798, + "step": 39680 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014645283511889208, + "loss": 0.1496, + "step": 39690 + }, + { + "epoch": 2.58, + "learning_rate": 0.0001464136399268356, + "loss": 0.1699, + "step": 39700 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014637444473477917, + "loss": 0.1862, + "step": 39710 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014633524954272274, + "loss": 0.2138, + "step": 39720 + }, + { + "epoch": 2.58, + "learning_rate": 0.0001462960543506663, + "loss": 0.1476, + "step": 39730 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014625685915860986, + "loss": 0.1654, + "step": 39740 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014621766396655342, + "loss": 0.1589, + "step": 39750 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014617846877449698, + "loss": 0.2056, + "step": 39760 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014613927358244055, + "loss": 0.2486, + "step": 39770 + }, + { + "epoch": 2.58, + "learning_rate": 0.0001461000783903841, + "loss": 0.1349, + "step": 39780 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014606088319832767, + "loss": 0.1805, + "step": 39790 + }, + { + "epoch": 2.58, + "learning_rate": 0.0001460216880062712, + "loss": 0.1737, + "step": 39800 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014598249281421477, + "loss": 0.197, + "step": 39810 + }, + { + "epoch": 2.58, + "learning_rate": 0.00014594329762215833, + "loss": 0.2393, + "step": 39820 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001459041024301019, + "loss": 0.156, + "step": 39830 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014586490723804546, + "loss": 0.1632, + "step": 39840 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014582571204598902, + "loss": 0.1548, + "step": 39850 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014578651685393258, + "loss": 0.2017, + "step": 39860 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014574732166187614, + "loss": 0.2181, + "step": 39870 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001457081264698197, + "loss": 0.1731, + "step": 39880 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014566893127776324, + "loss": 0.1435, + "step": 39890 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001456297360857068, + "loss": 0.1798, + "step": 39900 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014559054089365037, + "loss": 0.2077, + "step": 39910 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014555134570159393, + "loss": 0.2368, + "step": 39920 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001455121505095375, + "loss": 0.1645, + "step": 39930 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014547295531748102, + "loss": 0.1653, + "step": 39940 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001454337601254246, + "loss": 0.1561, + "step": 39950 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014539456493336815, + "loss": 0.1674, + "step": 39960 + }, + { + "epoch": 2.59, + "learning_rate": 0.0001453553697413117, + "loss": 0.2389, + "step": 39970 + }, + { + "epoch": 2.59, + "learning_rate": 0.00014531617454925527, + "loss": 0.1711, + "step": 39980 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014527697935719884, + "loss": 0.1585, + "step": 39990 + }, + { + "epoch": 2.6, + "learning_rate": 0.0001452377841651424, + "loss": 0.168, + "step": 40000 + }, + { + "epoch": 2.6, + "eval_loss": 0.23083649575710297, + "eval_runtime": 1012.0817, + "eval_samples_per_second": 15.402, + "eval_steps_per_second": 1.926, + "eval_wer": 0.23200807380136743, + "step": 40000 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014519858897308596, + "loss": 0.1891, + "step": 40010 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014515939378102952, + "loss": 0.2546, + "step": 40020 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014512019858897309, + "loss": 0.1549, + "step": 40030 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014508100339691662, + "loss": 0.1421, + "step": 40040 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014504180820486018, + "loss": 0.179, + "step": 40050 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014500261301280375, + "loss": 0.2193, + "step": 40060 + }, + { + "epoch": 2.6, + "learning_rate": 0.0001449634178207473, + "loss": 0.2406, + "step": 40070 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014492422262869087, + "loss": 0.1544, + "step": 40080 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014488502743663443, + "loss": 0.1675, + "step": 40090 + }, + { + "epoch": 2.6, + "learning_rate": 0.000144845832244578, + "loss": 0.1674, + "step": 40100 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014480663705252156, + "loss": 0.1767, + "step": 40110 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014476744186046512, + "loss": 0.2366, + "step": 40120 + }, + { + "epoch": 2.6, + "learning_rate": 0.00014472824666840868, + "loss": 0.1386, + "step": 40130 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014468905147635222, + "loss": 0.1505, + "step": 40140 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014464985628429578, + "loss": 0.1685, + "step": 40150 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014461066109223934, + "loss": 0.1896, + "step": 40160 + }, + { + "epoch": 2.61, + "learning_rate": 0.0001445714659001829, + "loss": 0.2565, + "step": 40170 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014453227070812644, + "loss": 0.1313, + "step": 40180 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014449307551607, + "loss": 0.1655, + "step": 40190 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014445388032401356, + "loss": 0.169, + "step": 40200 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014441468513195713, + "loss": 0.194, + "step": 40210 + }, + { + "epoch": 2.61, + "learning_rate": 0.0001443754899399007, + "loss": 0.2843, + "step": 40220 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014433629474784425, + "loss": 0.1745, + "step": 40230 + }, + { + "epoch": 2.61, + "learning_rate": 0.0001442970995557878, + "loss": 0.1522, + "step": 40240 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014425790436373137, + "loss": 0.1652, + "step": 40250 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014421870917167494, + "loss": 0.1794, + "step": 40260 + }, + { + "epoch": 2.61, + "learning_rate": 0.0001441795139796185, + "loss": 0.266, + "step": 40270 + }, + { + "epoch": 2.61, + "learning_rate": 0.00014414031878756206, + "loss": 0.1356, + "step": 40280 + }, + { + "epoch": 2.61, + "learning_rate": 0.0001441011235955056, + "loss": 0.1924, + "step": 40290 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014406192840344916, + "loss": 0.1558, + "step": 40300 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014402273321139272, + "loss": 0.1787, + "step": 40310 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014398353801933628, + "loss": 0.1987, + "step": 40320 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014394434282727985, + "loss": 0.1583, + "step": 40330 + }, + { + "epoch": 2.62, + "learning_rate": 0.0001439051476352234, + "loss": 0.1879, + "step": 40340 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014386595244316697, + "loss": 0.1855, + "step": 40350 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014382675725111053, + "loss": 0.1937, + "step": 40360 + }, + { + "epoch": 2.62, + "learning_rate": 0.0001437875620590541, + "loss": 0.2254, + "step": 40370 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014374836686699763, + "loss": 0.1524, + "step": 40380 + }, + { + "epoch": 2.62, + "learning_rate": 0.0001437091716749412, + "loss": 0.1837, + "step": 40390 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014366997648288475, + "loss": 0.1654, + "step": 40400 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014363078129082832, + "loss": 0.2145, + "step": 40410 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014359158609877188, + "loss": 0.2266, + "step": 40420 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014355239090671541, + "loss": 0.1459, + "step": 40430 + }, + { + "epoch": 2.62, + "learning_rate": 0.00014351319571465898, + "loss": 0.1628, + "step": 40440 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014347400052260254, + "loss": 0.1458, + "step": 40450 + }, + { + "epoch": 2.63, + "learning_rate": 0.0001434348053305461, + "loss": 0.195, + "step": 40460 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014339561013848966, + "loss": 0.2174, + "step": 40470 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014335641494643323, + "loss": 0.193, + "step": 40480 + }, + { + "epoch": 2.63, + "learning_rate": 0.0001433172197543768, + "loss": 0.16, + "step": 40490 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014327802456232035, + "loss": 0.1835, + "step": 40500 + }, + { + "epoch": 2.63, + "learning_rate": 0.0001432388293702639, + "loss": 0.178, + "step": 40510 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014319963417820748, + "loss": 0.2499, + "step": 40520 + }, + { + "epoch": 2.63, + "learning_rate": 0.000143160438986151, + "loss": 0.1273, + "step": 40530 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014312124379409457, + "loss": 0.1528, + "step": 40540 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014308204860203813, + "loss": 0.1621, + "step": 40550 + }, + { + "epoch": 2.63, + "learning_rate": 0.0001430428534099817, + "loss": 0.2381, + "step": 40560 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014300365821792526, + "loss": 0.2215, + "step": 40570 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014296446302586882, + "loss": 0.147, + "step": 40580 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014292526783381238, + "loss": 0.1572, + "step": 40590 + }, + { + "epoch": 2.63, + "learning_rate": 0.00014288607264175595, + "loss": 0.1468, + "step": 40600 + }, + { + "epoch": 2.64, + "learning_rate": 0.0001428468774496995, + "loss": 0.2315, + "step": 40610 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014280768225764307, + "loss": 0.2357, + "step": 40620 + }, + { + "epoch": 2.64, + "learning_rate": 0.0001427684870655866, + "loss": 0.1453, + "step": 40630 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014272929187353017, + "loss": 0.1893, + "step": 40640 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014269009668147373, + "loss": 0.1762, + "step": 40650 + }, + { + "epoch": 2.64, + "learning_rate": 0.0001426509014894173, + "loss": 0.1747, + "step": 40660 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014261170629736083, + "loss": 0.2434, + "step": 40670 + }, + { + "epoch": 2.64, + "learning_rate": 0.0001425725111053044, + "loss": 0.1612, + "step": 40680 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014253331591324795, + "loss": 0.1723, + "step": 40690 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014249412072119152, + "loss": 0.2008, + "step": 40700 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014245492552913508, + "loss": 0.234, + "step": 40710 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014241573033707864, + "loss": 0.211, + "step": 40720 + }, + { + "epoch": 2.64, + "learning_rate": 0.0001423765351450222, + "loss": 0.1382, + "step": 40730 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014233733995296576, + "loss": 0.1683, + "step": 40740 + }, + { + "epoch": 2.64, + "learning_rate": 0.00014229814476090933, + "loss": 0.1561, + "step": 40750 + }, + { + "epoch": 2.65, + "learning_rate": 0.0001422589495688529, + "loss": 0.2128, + "step": 40760 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014221975437679642, + "loss": 0.2282, + "step": 40770 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014218055918473999, + "loss": 0.1337, + "step": 40780 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014214136399268355, + "loss": 0.1627, + "step": 40790 + }, + { + "epoch": 2.65, + "learning_rate": 0.0001421021688006271, + "loss": 0.2126, + "step": 40800 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014206297360857067, + "loss": 0.1862, + "step": 40810 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014202377841651424, + "loss": 0.2027, + "step": 40820 + }, + { + "epoch": 2.65, + "learning_rate": 0.0001419845832244578, + "loss": 0.147, + "step": 40830 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014194538803240136, + "loss": 0.1665, + "step": 40840 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014190619284034492, + "loss": 0.1953, + "step": 40850 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014186699764828848, + "loss": 0.1893, + "step": 40860 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014182780245623202, + "loss": 0.2475, + "step": 40870 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014178860726417558, + "loss": 0.1792, + "step": 40880 + }, + { + "epoch": 2.65, + "learning_rate": 0.00014174941207211914, + "loss": 0.1556, + "step": 40890 + }, + { + "epoch": 2.65, + "learning_rate": 0.0001417102168800627, + "loss": 0.1445, + "step": 40900 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014167102168800624, + "loss": 0.2291, + "step": 40910 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001416318264959498, + "loss": 0.2708, + "step": 40920 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014159263130389337, + "loss": 0.1415, + "step": 40930 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014155343611183693, + "loss": 0.1787, + "step": 40940 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001415142409197805, + "loss": 0.1694, + "step": 40950 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014147504572772405, + "loss": 0.1771, + "step": 40960 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014143585053566762, + "loss": 0.2624, + "step": 40970 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014139665534361118, + "loss": 0.1368, + "step": 40980 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014135746015155474, + "loss": 0.1709, + "step": 40990 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001413182649594983, + "loss": 0.1571, + "step": 41000 + }, + { + "epoch": 2.66, + "eval_loss": 0.22830916941165924, + "eval_runtime": 1006.3746, + "eval_samples_per_second": 15.489, + "eval_steps_per_second": 1.937, + "eval_wer": 0.22741680999559352, + "step": 41000 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014127906976744184, + "loss": 0.176, + "step": 41010 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001412398745753854, + "loss": 0.2054, + "step": 41020 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014120067938332896, + "loss": 0.1488, + "step": 41030 + }, + { + "epoch": 2.66, + "learning_rate": 0.00014116148419127252, + "loss": 0.1788, + "step": 41040 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001411222889992161, + "loss": 0.1397, + "step": 41050 + }, + { + "epoch": 2.66, + "learning_rate": 0.0001410870133263653, + "loss": 0.1749, + "step": 41060 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014104781813430883, + "loss": 0.2167, + "step": 41070 + }, + { + "epoch": 2.67, + "learning_rate": 0.0001410086229422524, + "loss": 0.1303, + "step": 41080 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014096942775019595, + "loss": 0.1755, + "step": 41090 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014093023255813951, + "loss": 0.1784, + "step": 41100 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014089103736608308, + "loss": 0.1961, + "step": 41110 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014085184217402664, + "loss": 0.2328, + "step": 41120 + }, + { + "epoch": 2.67, + "learning_rate": 0.0001408126469819702, + "loss": 0.1365, + "step": 41130 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014077345178991376, + "loss": 0.1523, + "step": 41140 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014073425659785733, + "loss": 0.1679, + "step": 41150 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014069506140580086, + "loss": 0.1765, + "step": 41160 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014065586621374442, + "loss": 0.2257, + "step": 41170 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014061667102168799, + "loss": 0.1468, + "step": 41180 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014057747582963155, + "loss": 0.1708, + "step": 41190 + }, + { + "epoch": 2.67, + "learning_rate": 0.0001405382806375751, + "loss": 0.1456, + "step": 41200 + }, + { + "epoch": 2.67, + "learning_rate": 0.00014049908544551867, + "loss": 0.2128, + "step": 41210 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014045989025346224, + "loss": 0.206, + "step": 41220 + }, + { + "epoch": 2.68, + "learning_rate": 0.0001404206950614058, + "loss": 0.1355, + "step": 41230 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014038149986934936, + "loss": 0.1733, + "step": 41240 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014034230467729292, + "loss": 0.1687, + "step": 41250 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014030310948523646, + "loss": 0.1754, + "step": 41260 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014026391429318002, + "loss": 0.2131, + "step": 41270 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014022471910112358, + "loss": 0.1457, + "step": 41280 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014018552390906714, + "loss": 0.1497, + "step": 41290 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014015024823621635, + "loss": 0.1949, + "step": 41300 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014011105304415989, + "loss": 0.2075, + "step": 41310 + }, + { + "epoch": 2.68, + "learning_rate": 0.00014007185785210345, + "loss": 0.2702, + "step": 41320 + }, + { + "epoch": 2.68, + "learning_rate": 0.000140032662660047, + "loss": 0.165, + "step": 41330 + }, + { + "epoch": 2.68, + "learning_rate": 0.00013999346746799057, + "loss": 0.1634, + "step": 41340 + }, + { + "epoch": 2.68, + "learning_rate": 0.00013995427227593413, + "loss": 0.1691, + "step": 41350 + }, + { + "epoch": 2.68, + "learning_rate": 0.0001399150770838777, + "loss": 0.1752, + "step": 41360 + }, + { + "epoch": 2.68, + "learning_rate": 0.00013987588189182126, + "loss": 0.2202, + "step": 41370 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013983668669976482, + "loss": 0.1309, + "step": 41380 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013979749150770838, + "loss": 0.1436, + "step": 41390 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013975829631565195, + "loss": 0.154, + "step": 41400 + }, + { + "epoch": 2.69, + "learning_rate": 0.0001397191011235955, + "loss": 0.1818, + "step": 41410 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013967990593153904, + "loss": 0.2321, + "step": 41420 + }, + { + "epoch": 2.69, + "learning_rate": 0.0001396407107394826, + "loss": 0.1517, + "step": 41430 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013960151554742617, + "loss": 0.138, + "step": 41440 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013956232035536973, + "loss": 0.1669, + "step": 41450 + }, + { + "epoch": 2.69, + "learning_rate": 0.0001395231251633133, + "loss": 0.2114, + "step": 41460 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013948392997125686, + "loss": 0.2213, + "step": 41470 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013944473477920042, + "loss": 0.1192, + "step": 41480 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013940553958714398, + "loss": 0.1884, + "step": 41490 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013936634439508754, + "loss": 0.1857, + "step": 41500 + }, + { + "epoch": 2.69, + "learning_rate": 0.0001393271492030311, + "loss": 0.2033, + "step": 41510 + }, + { + "epoch": 2.69, + "learning_rate": 0.00013928795401097464, + "loss": 0.2657, + "step": 41520 + }, + { + "epoch": 2.7, + "learning_rate": 0.0001392487588189182, + "loss": 0.1423, + "step": 41530 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013920956362686176, + "loss": 0.1653, + "step": 41540 + }, + { + "epoch": 2.7, + "learning_rate": 0.0001391703684348053, + "loss": 0.1777, + "step": 41550 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013913117324274886, + "loss": 0.1913, + "step": 41560 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013909197805069242, + "loss": 0.2054, + "step": 41570 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013905278285863599, + "loss": 0.1618, + "step": 41580 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013901358766657955, + "loss": 0.1693, + "step": 41590 + }, + { + "epoch": 2.7, + "learning_rate": 0.0001389743924745231, + "loss": 0.1486, + "step": 41600 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013893519728246667, + "loss": 0.2107, + "step": 41610 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013889600209041024, + "loss": 0.2425, + "step": 41620 + }, + { + "epoch": 2.7, + "learning_rate": 0.0001388568068983538, + "loss": 0.1399, + "step": 41630 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013881761170629736, + "loss": 0.1591, + "step": 41640 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013877841651424092, + "loss": 0.163, + "step": 41650 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013873922132218446, + "loss": 0.2139, + "step": 41660 + }, + { + "epoch": 2.7, + "learning_rate": 0.00013870002613012802, + "loss": 0.2215, + "step": 41670 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013866083093807158, + "loss": 0.1505, + "step": 41680 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013862163574601514, + "loss": 0.1813, + "step": 41690 + }, + { + "epoch": 2.71, + "learning_rate": 0.0001385824405539587, + "loss": 0.187, + "step": 41700 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013854324536190227, + "loss": 0.2222, + "step": 41710 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013850405016984583, + "loss": 0.2318, + "step": 41720 + }, + { + "epoch": 2.71, + "learning_rate": 0.0001384648549777894, + "loss": 0.1423, + "step": 41730 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013842565978573296, + "loss": 0.1474, + "step": 41740 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013838646459367652, + "loss": 0.2017, + "step": 41750 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013834726940162005, + "loss": 0.1868, + "step": 41760 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013830807420956362, + "loss": 0.2029, + "step": 41770 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013826887901750718, + "loss": 0.1636, + "step": 41780 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013822968382545074, + "loss": 0.1693, + "step": 41790 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013819048863339428, + "loss": 0.159, + "step": 41800 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013815129344133784, + "loss": 0.2027, + "step": 41810 + }, + { + "epoch": 2.71, + "learning_rate": 0.0001381120982492814, + "loss": 0.2369, + "step": 41820 + }, + { + "epoch": 2.71, + "learning_rate": 0.00013807290305722496, + "loss": 0.1511, + "step": 41830 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013803370786516852, + "loss": 0.1455, + "step": 41840 + }, + { + "epoch": 2.72, + "learning_rate": 0.0001379945126731121, + "loss": 0.1716, + "step": 41850 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013795531748105565, + "loss": 0.2125, + "step": 41860 + }, + { + "epoch": 2.72, + "learning_rate": 0.0001379161222889992, + "loss": 0.2367, + "step": 41870 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013787692709694277, + "loss": 0.1499, + "step": 41880 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013783773190488634, + "loss": 0.1545, + "step": 41890 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013779853671282987, + "loss": 0.1525, + "step": 41900 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013775934152077343, + "loss": 0.2024, + "step": 41910 + }, + { + "epoch": 2.72, + "learning_rate": 0.000137720146328717, + "loss": 0.225, + "step": 41920 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013768095113666056, + "loss": 0.147, + "step": 41930 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013764175594460412, + "loss": 0.193, + "step": 41940 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013760256075254768, + "loss": 0.1754, + "step": 41950 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013756336556049124, + "loss": 0.2069, + "step": 41960 + }, + { + "epoch": 2.72, + "learning_rate": 0.0001375241703684348, + "loss": 0.2108, + "step": 41970 + }, + { + "epoch": 2.72, + "learning_rate": 0.00013748497517637837, + "loss": 0.145, + "step": 41980 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013744577998432193, + "loss": 0.1317, + "step": 41990 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013740658479226547, + "loss": 0.1568, + "step": 42000 + }, + { + "epoch": 2.73, + "eval_loss": 0.230531707406044, + "eval_runtime": 1003.883, + "eval_samples_per_second": 15.528, + "eval_steps_per_second": 1.941, + "eval_wer": 0.22712541399553668, + "step": 42000 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013736738960020903, + "loss": 0.1667, + "step": 42010 + }, + { + "epoch": 2.73, + "learning_rate": 0.0001373281944081526, + "loss": 0.2096, + "step": 42020 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013728899921609615, + "loss": 0.1506, + "step": 42030 + }, + { + "epoch": 2.73, + "learning_rate": 0.0001372498040240397, + "loss": 0.1544, + "step": 42040 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013721060883198325, + "loss": 0.1816, + "step": 42050 + }, + { + "epoch": 2.73, + "learning_rate": 0.0001371714136399268, + "loss": 0.2023, + "step": 42060 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013713221844787038, + "loss": 0.2092, + "step": 42070 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013709302325581394, + "loss": 0.1383, + "step": 42080 + }, + { + "epoch": 2.73, + "learning_rate": 0.0001370538280637575, + "loss": 0.1699, + "step": 42090 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013701463287170106, + "loss": 0.1825, + "step": 42100 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013697543767964462, + "loss": 0.1801, + "step": 42110 + }, + { + "epoch": 2.73, + "learning_rate": 0.0001369362424875882, + "loss": 0.2354, + "step": 42120 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013689704729553175, + "loss": 0.1399, + "step": 42130 + }, + { + "epoch": 2.73, + "learning_rate": 0.00013685785210347528, + "loss": 0.171, + "step": 42140 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013681865691141885, + "loss": 0.1567, + "step": 42150 + }, + { + "epoch": 2.74, + "learning_rate": 0.0001367794617193624, + "loss": 0.2102, + "step": 42160 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013674026652730597, + "loss": 0.242, + "step": 42170 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013670107133524953, + "loss": 0.1537, + "step": 42180 + }, + { + "epoch": 2.74, + "learning_rate": 0.0001366618761431931, + "loss": 0.166, + "step": 42190 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013662268095113666, + "loss": 0.1881, + "step": 42200 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013658348575908022, + "loss": 0.1913, + "step": 42210 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013654429056702378, + "loss": 0.2319, + "step": 42220 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013650509537496735, + "loss": 0.1513, + "step": 42230 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013646590018291088, + "loss": 0.1731, + "step": 42240 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013642670499085444, + "loss": 0.1759, + "step": 42250 + }, + { + "epoch": 2.74, + "learning_rate": 0.000136387509798798, + "loss": 0.2096, + "step": 42260 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013634831460674157, + "loss": 0.2362, + "step": 42270 + }, + { + "epoch": 2.74, + "learning_rate": 0.0001363091194146851, + "loss": 0.1439, + "step": 42280 + }, + { + "epoch": 2.74, + "learning_rate": 0.00013626992422262866, + "loss": 0.1563, + "step": 42290 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013623072903057223, + "loss": 0.1775, + "step": 42300 + }, + { + "epoch": 2.75, + "learning_rate": 0.0001361915338385158, + "loss": 0.2066, + "step": 42310 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013615233864645935, + "loss": 0.2015, + "step": 42320 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013611314345440291, + "loss": 0.1727, + "step": 42330 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013607394826234648, + "loss": 0.1583, + "step": 42340 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013603475307029004, + "loss": 0.1836, + "step": 42350 + }, + { + "epoch": 2.75, + "learning_rate": 0.0001359955578782336, + "loss": 0.183, + "step": 42360 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013595636268617716, + "loss": 0.2433, + "step": 42370 + }, + { + "epoch": 2.75, + "learning_rate": 0.0001359171674941207, + "loss": 0.1474, + "step": 42380 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013587797230206426, + "loss": 0.1643, + "step": 42390 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013583877711000782, + "loss": 0.1754, + "step": 42400 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013579958191795139, + "loss": 0.2186, + "step": 42410 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013576038672589495, + "loss": 0.2554, + "step": 42420 + }, + { + "epoch": 2.75, + "learning_rate": 0.0001357211915338385, + "loss": 0.1392, + "step": 42430 + }, + { + "epoch": 2.75, + "learning_rate": 0.00013568199634178207, + "loss": 0.1647, + "step": 42440 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013564280114972563, + "loss": 0.1783, + "step": 42450 + }, + { + "epoch": 2.76, + "learning_rate": 0.0001356036059576692, + "loss": 0.2019, + "step": 42460 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013556441076561276, + "loss": 0.2577, + "step": 42470 + }, + { + "epoch": 2.76, + "learning_rate": 0.0001355252155735563, + "loss": 0.1538, + "step": 42480 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013548602038149986, + "loss": 0.1812, + "step": 42490 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013544682518944342, + "loss": 0.1677, + "step": 42500 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013540762999738698, + "loss": 0.1915, + "step": 42510 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013536843480533052, + "loss": 0.2017, + "step": 42520 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013532923961327408, + "loss": 0.161, + "step": 42530 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013529004442121764, + "loss": 0.165, + "step": 42540 + }, + { + "epoch": 2.76, + "learning_rate": 0.0001352508492291612, + "loss": 0.1673, + "step": 42550 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013521165403710477, + "loss": 0.1837, + "step": 42560 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013517245884504833, + "loss": 0.2462, + "step": 42570 + }, + { + "epoch": 2.76, + "learning_rate": 0.0001351332636529919, + "loss": 0.1561, + "step": 42580 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013509406846093545, + "loss": 0.1776, + "step": 42590 + }, + { + "epoch": 2.76, + "learning_rate": 0.00013505487326887901, + "loss": 0.1525, + "step": 42600 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013501567807682258, + "loss": 0.1659, + "step": 42610 + }, + { + "epoch": 2.77, + "learning_rate": 0.0001349764828847661, + "loss": 0.2186, + "step": 42620 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013493728769270967, + "loss": 0.1241, + "step": 42630 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013489809250065324, + "loss": 0.1409, + "step": 42640 + }, + { + "epoch": 2.77, + "learning_rate": 0.0001348588973085968, + "loss": 0.1744, + "step": 42650 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013481970211654036, + "loss": 0.1876, + "step": 42660 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013478050692448392, + "loss": 0.1998, + "step": 42670 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013474131173242749, + "loss": 0.1323, + "step": 42680 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013470211654037105, + "loss": 0.1645, + "step": 42690 + }, + { + "epoch": 2.77, + "learning_rate": 0.0001346629213483146, + "loss": 0.1584, + "step": 42700 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013462372615625817, + "loss": 0.2058, + "step": 42710 + }, + { + "epoch": 2.77, + "learning_rate": 0.0001345845309642017, + "loss": 0.2444, + "step": 42720 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013454533577214527, + "loss": 0.1453, + "step": 42730 + }, + { + "epoch": 2.77, + "learning_rate": 0.00013450614058008883, + "loss": 0.1546, + "step": 42740 + }, + { + "epoch": 2.77, + "learning_rate": 0.0001344669453880324, + "loss": 0.1726, + "step": 42750 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013442775019597593, + "loss": 0.1767, + "step": 42760 + }, + { + "epoch": 2.78, + "learning_rate": 0.0001343885550039195, + "loss": 0.2636, + "step": 42770 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013434935981186305, + "loss": 0.1517, + "step": 42780 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013431016461980662, + "loss": 0.1487, + "step": 42790 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013427096942775018, + "loss": 0.1639, + "step": 42800 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013423177423569374, + "loss": 0.2092, + "step": 42810 + }, + { + "epoch": 2.78, + "learning_rate": 0.0001341925790436373, + "loss": 0.206, + "step": 42820 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013415338385158087, + "loss": 0.1261, + "step": 42830 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013411418865952443, + "loss": 0.1527, + "step": 42840 + }, + { + "epoch": 2.78, + "learning_rate": 0.000134074993467468, + "loss": 0.1787, + "step": 42850 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013403579827541153, + "loss": 0.181, + "step": 42860 + }, + { + "epoch": 2.78, + "learning_rate": 0.0001339966030833551, + "loss": 0.2347, + "step": 42870 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013395740789129865, + "loss": 0.1224, + "step": 42880 + }, + { + "epoch": 2.78, + "learning_rate": 0.0001339182126992422, + "loss": 0.1396, + "step": 42890 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013387901750718577, + "loss": 0.1623, + "step": 42900 + }, + { + "epoch": 2.78, + "learning_rate": 0.00013383982231512934, + "loss": 0.1781, + "step": 42910 + }, + { + "epoch": 2.79, + "learning_rate": 0.0001338006271230729, + "loss": 0.2075, + "step": 42920 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013376143193101646, + "loss": 0.1431, + "step": 42930 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013372223673896002, + "loss": 0.1552, + "step": 42940 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013368304154690359, + "loss": 0.1866, + "step": 42950 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013364384635484712, + "loss": 0.1864, + "step": 42960 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013360465116279068, + "loss": 0.1905, + "step": 42970 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013356545597073425, + "loss": 0.1576, + "step": 42980 + }, + { + "epoch": 2.79, + "learning_rate": 0.0001335262607786778, + "loss": 0.1626, + "step": 42990 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013348706558662134, + "loss": 0.1486, + "step": 43000 + }, + { + "epoch": 2.79, + "eval_loss": 0.23158212006092072, + "eval_runtime": 1009.3969, + "eval_samples_per_second": 15.443, + "eval_steps_per_second": 1.931, + "eval_wer": 0.2277437420932183, + "step": 43000 + }, + { + "epoch": 2.79, + "learning_rate": 0.0001334478703945649, + "loss": 0.1941, + "step": 43010 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013340867520250847, + "loss": 0.2013, + "step": 43020 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013336948001045203, + "loss": 0.1428, + "step": 43030 + }, + { + "epoch": 2.79, + "learning_rate": 0.0001333302848183956, + "loss": 0.1532, + "step": 43040 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013329108962633915, + "loss": 0.1584, + "step": 43050 + }, + { + "epoch": 2.79, + "learning_rate": 0.00013325189443428272, + "loss": 0.1938, + "step": 43060 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013321269924222628, + "loss": 0.23, + "step": 43070 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013317350405016984, + "loss": 0.1287, + "step": 43080 + }, + { + "epoch": 2.8, + "learning_rate": 0.0001331343088581134, + "loss": 0.1706, + "step": 43090 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013309511366605694, + "loss": 0.1641, + "step": 43100 + }, + { + "epoch": 2.8, + "learning_rate": 0.0001330559184740005, + "loss": 0.1848, + "step": 43110 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013301672328194406, + "loss": 0.2383, + "step": 43120 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013297752808988763, + "loss": 0.1126, + "step": 43130 + }, + { + "epoch": 2.8, + "learning_rate": 0.0001329383328978312, + "loss": 0.176, + "step": 43140 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013289913770577475, + "loss": 0.1749, + "step": 43150 + }, + { + "epoch": 2.8, + "learning_rate": 0.0001328599425137183, + "loss": 0.2088, + "step": 43160 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013282074732166188, + "loss": 0.2133, + "step": 43170 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013278155212960544, + "loss": 0.1544, + "step": 43180 + }, + { + "epoch": 2.8, + "learning_rate": 0.000132742356937549, + "loss": 0.1404, + "step": 43190 + }, + { + "epoch": 2.8, + "learning_rate": 0.00013270316174549254, + "loss": 0.1547, + "step": 43200 + }, + { + "epoch": 2.8, + "learning_rate": 0.0001326639665534361, + "loss": 0.1916, + "step": 43210 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013262477136137966, + "loss": 0.2449, + "step": 43220 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013258557616932322, + "loss": 0.1154, + "step": 43230 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013254638097726676, + "loss": 0.1888, + "step": 43240 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013250718578521032, + "loss": 0.1671, + "step": 43250 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013246799059315388, + "loss": 0.1833, + "step": 43260 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013242879540109744, + "loss": 0.2063, + "step": 43270 + }, + { + "epoch": 2.81, + "learning_rate": 0.000132389600209041, + "loss": 0.1179, + "step": 43280 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013235040501698457, + "loss": 0.1458, + "step": 43290 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013231120982492813, + "loss": 0.1763, + "step": 43300 + }, + { + "epoch": 2.81, + "learning_rate": 0.0001322720146328717, + "loss": 0.1942, + "step": 43310 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013223281944081526, + "loss": 0.235, + "step": 43320 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013219362424875882, + "loss": 0.1571, + "step": 43330 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013215442905670235, + "loss": 0.1461, + "step": 43340 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013211523386464592, + "loss": 0.1685, + "step": 43350 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013207603867258948, + "loss": 0.1981, + "step": 43360 + }, + { + "epoch": 2.81, + "learning_rate": 0.00013203684348053304, + "loss": 0.2395, + "step": 43370 + }, + { + "epoch": 2.82, + "learning_rate": 0.0001319976482884766, + "loss": 0.1654, + "step": 43380 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013195845309642016, + "loss": 0.1384, + "step": 43390 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013191925790436373, + "loss": 0.2084, + "step": 43400 + }, + { + "epoch": 2.82, + "learning_rate": 0.0001318800627123073, + "loss": 0.2009, + "step": 43410 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013184086752025085, + "loss": 0.2699, + "step": 43420 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013180167232819441, + "loss": 0.1493, + "step": 43430 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013176247713613795, + "loss": 0.1656, + "step": 43440 + }, + { + "epoch": 2.82, + "learning_rate": 0.0001317232819440815, + "loss": 0.1679, + "step": 43450 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013168408675202507, + "loss": 0.1651, + "step": 43460 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013164489155996864, + "loss": 0.1931, + "step": 43470 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013160569636791217, + "loss": 0.136, + "step": 43480 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013156650117585573, + "loss": 0.147, + "step": 43490 + }, + { + "epoch": 2.82, + "learning_rate": 0.0001315273059837993, + "loss": 0.1784, + "step": 43500 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013148811079174286, + "loss": 0.1858, + "step": 43510 + }, + { + "epoch": 2.82, + "learning_rate": 0.00013144891559968642, + "loss": 0.2223, + "step": 43520 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013140972040762998, + "loss": 0.1538, + "step": 43530 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013137052521557354, + "loss": 0.1282, + "step": 43540 + }, + { + "epoch": 2.83, + "learning_rate": 0.0001313313300235171, + "loss": 0.1583, + "step": 43550 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013129213483146067, + "loss": 0.2086, + "step": 43560 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013125293963940423, + "loss": 0.2425, + "step": 43570 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013121374444734777, + "loss": 0.1507, + "step": 43580 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013117454925529133, + "loss": 0.1562, + "step": 43590 + }, + { + "epoch": 2.83, + "learning_rate": 0.0001311353540632349, + "loss": 0.1373, + "step": 43600 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013109615887117845, + "loss": 0.1699, + "step": 43610 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013105696367912202, + "loss": 0.2183, + "step": 43620 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013101776848706558, + "loss": 0.1337, + "step": 43630 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013097857329500914, + "loss": 0.1313, + "step": 43640 + }, + { + "epoch": 2.83, + "learning_rate": 0.0001309393781029527, + "loss": 0.1958, + "step": 43650 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013090018291089627, + "loss": 0.1846, + "step": 43660 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013086098771883983, + "loss": 0.2422, + "step": 43670 + }, + { + "epoch": 2.83, + "learning_rate": 0.00013082179252678336, + "loss": 0.1461, + "step": 43680 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013078259733472692, + "loss": 0.1496, + "step": 43690 + }, + { + "epoch": 2.84, + "learning_rate": 0.0001307434021426705, + "loss": 0.1734, + "step": 43700 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013070420695061405, + "loss": 0.1761, + "step": 43710 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013066501175855758, + "loss": 0.22, + "step": 43720 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013062581656650115, + "loss": 0.1614, + "step": 43730 + }, + { + "epoch": 2.84, + "learning_rate": 0.0001305866213744447, + "loss": 0.1421, + "step": 43740 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013054742618238827, + "loss": 0.1764, + "step": 43750 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013050823099033183, + "loss": 0.1834, + "step": 43760 + }, + { + "epoch": 2.84, + "learning_rate": 0.0001304690357982754, + "loss": 0.2425, + "step": 43770 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013042984060621896, + "loss": 0.1491, + "step": 43780 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013039064541416252, + "loss": 0.1421, + "step": 43790 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013035145022210608, + "loss": 0.1778, + "step": 43800 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013031225503004965, + "loss": 0.1815, + "step": 43810 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013027305983799318, + "loss": 0.2607, + "step": 43820 + }, + { + "epoch": 2.84, + "learning_rate": 0.00013023386464593674, + "loss": 0.1432, + "step": 43830 + }, + { + "epoch": 2.85, + "learning_rate": 0.0001301946694538803, + "loss": 0.1558, + "step": 43840 + }, + { + "epoch": 2.85, + "learning_rate": 0.00013015547426182387, + "loss": 0.1683, + "step": 43850 + }, + { + "epoch": 2.85, + "learning_rate": 0.00013011627906976743, + "loss": 0.1677, + "step": 43860 + }, + { + "epoch": 2.85, + "learning_rate": 0.000130077083877711, + "loss": 0.2277, + "step": 43870 + }, + { + "epoch": 2.85, + "learning_rate": 0.00013003788868565455, + "loss": 0.1219, + "step": 43880 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012999869349359812, + "loss": 0.1335, + "step": 43890 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012995949830154168, + "loss": 0.163, + "step": 43900 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012992030310948524, + "loss": 0.162, + "step": 43910 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012988110791742878, + "loss": 0.2659, + "step": 43920 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012984191272537234, + "loss": 0.1353, + "step": 43930 + }, + { + "epoch": 2.85, + "learning_rate": 0.0001298027175333159, + "loss": 0.1724, + "step": 43940 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012976352234125946, + "loss": 0.1674, + "step": 43950 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012972432714920303, + "loss": 0.2111, + "step": 43960 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012968513195714656, + "loss": 0.219, + "step": 43970 + }, + { + "epoch": 2.85, + "learning_rate": 0.00012964593676509012, + "loss": 0.1767, + "step": 43980 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012960674157303369, + "loss": 0.1479, + "step": 43990 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012956754638097725, + "loss": 0.1584, + "step": 44000 + }, + { + "epoch": 2.86, + "eval_loss": 0.22028018534183502, + "eval_runtime": 1008.1906, + "eval_samples_per_second": 15.461, + "eval_steps_per_second": 1.933, + "eval_wer": 0.2325340080453725, + "step": 44000 + }, + { + "epoch": 2.86, + "learning_rate": 0.0001295283511889208, + "loss": 0.1741, + "step": 44010 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012948915599686437, + "loss": 0.1974, + "step": 44020 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012944996080480793, + "loss": 0.1462, + "step": 44030 + }, + { + "epoch": 2.86, + "learning_rate": 0.0001294107656127515, + "loss": 0.1774, + "step": 44040 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012937157042069506, + "loss": 0.1737, + "step": 44050 + }, + { + "epoch": 2.86, + "learning_rate": 0.0001293323752286386, + "loss": 0.168, + "step": 44060 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012929318003658216, + "loss": 0.2249, + "step": 44070 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012925398484452572, + "loss": 0.1654, + "step": 44080 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012921478965246928, + "loss": 0.1348, + "step": 44090 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012917559446041284, + "loss": 0.1749, + "step": 44100 + }, + { + "epoch": 2.86, + "learning_rate": 0.0001291363992683564, + "loss": 0.1845, + "step": 44110 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012909720407629997, + "loss": 0.201, + "step": 44120 + }, + { + "epoch": 2.86, + "learning_rate": 0.00012905800888424353, + "loss": 0.1502, + "step": 44130 + }, + { + "epoch": 2.86, + "learning_rate": 0.0001290188136921871, + "loss": 0.1789, + "step": 44140 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012897961850013065, + "loss": 0.1503, + "step": 44150 + }, + { + "epoch": 2.87, + "learning_rate": 0.0001289404233080742, + "loss": 0.2042, + "step": 44160 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012890122811601775, + "loss": 0.2162, + "step": 44170 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012886203292396131, + "loss": 0.145, + "step": 44180 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012882283773190488, + "loss": 0.1403, + "step": 44190 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012878364253984844, + "loss": 0.1453, + "step": 44200 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012874444734779197, + "loss": 0.2074, + "step": 44210 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012870525215573554, + "loss": 0.2411, + "step": 44220 + }, + { + "epoch": 2.87, + "learning_rate": 0.0001286660569636791, + "loss": 0.1451, + "step": 44230 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012862686177162266, + "loss": 0.1479, + "step": 44240 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012858766657956622, + "loss": 0.1684, + "step": 44250 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012854847138750979, + "loss": 0.2354, + "step": 44260 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012850927619545335, + "loss": 0.185, + "step": 44270 + }, + { + "epoch": 2.87, + "learning_rate": 0.0001284700810033969, + "loss": 0.1482, + "step": 44280 + }, + { + "epoch": 2.87, + "learning_rate": 0.00012843088581134047, + "loss": 0.1619, + "step": 44290 + }, + { + "epoch": 2.88, + "learning_rate": 0.000128391690619284, + "loss": 0.18, + "step": 44300 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012835249542722757, + "loss": 0.1834, + "step": 44310 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012831330023517113, + "loss": 0.2077, + "step": 44320 + }, + { + "epoch": 2.88, + "learning_rate": 0.0001282741050431147, + "loss": 0.1318, + "step": 44330 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012823490985105826, + "loss": 0.1437, + "step": 44340 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012819571465900182, + "loss": 0.1782, + "step": 44350 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012815651946694538, + "loss": 0.1685, + "step": 44360 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012811732427488894, + "loss": 0.228, + "step": 44370 + }, + { + "epoch": 2.88, + "learning_rate": 0.0001280781290828325, + "loss": 0.1403, + "step": 44380 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012803893389077607, + "loss": 0.1467, + "step": 44390 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012799973869871963, + "loss": 0.1718, + "step": 44400 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012796054350666317, + "loss": 0.1705, + "step": 44410 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012792134831460673, + "loss": 0.2094, + "step": 44420 + }, + { + "epoch": 2.88, + "learning_rate": 0.0001278821531225503, + "loss": 0.1815, + "step": 44430 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012784295793049385, + "loss": 0.1795, + "step": 44440 + }, + { + "epoch": 2.88, + "learning_rate": 0.00012780376273843741, + "loss": 0.1354, + "step": 44450 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012776456754638095, + "loss": 0.1961, + "step": 44460 + }, + { + "epoch": 2.89, + "learning_rate": 0.0001277253723543245, + "loss": 0.2429, + "step": 44470 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012768617716226807, + "loss": 0.1572, + "step": 44480 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012764698197021164, + "loss": 0.128, + "step": 44490 + }, + { + "epoch": 2.89, + "learning_rate": 0.0001276077867781552, + "loss": 0.1908, + "step": 44500 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012756859158609876, + "loss": 0.1921, + "step": 44510 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012752939639404232, + "loss": 0.2046, + "step": 44520 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012749020120198589, + "loss": 0.1541, + "step": 44530 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012745100600992945, + "loss": 0.1351, + "step": 44540 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012741181081787298, + "loss": 0.1677, + "step": 44550 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012737261562581655, + "loss": 0.202, + "step": 44560 + }, + { + "epoch": 2.89, + "learning_rate": 0.0001273334204337601, + "loss": 0.2296, + "step": 44570 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012729422524170367, + "loss": 0.136, + "step": 44580 + }, + { + "epoch": 2.89, + "learning_rate": 0.00012725503004964723, + "loss": 0.1618, + "step": 44590 + }, + { + "epoch": 2.89, + "learning_rate": 0.0001272158348575908, + "loss": 0.1627, + "step": 44600 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012717663966553436, + "loss": 0.223, + "step": 44610 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012713744447347792, + "loss": 0.2026, + "step": 44620 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012709824928142148, + "loss": 0.1567, + "step": 44630 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012705905408936504, + "loss": 0.1432, + "step": 44640 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012701985889730858, + "loss": 0.1871, + "step": 44650 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012698066370525214, + "loss": 0.1889, + "step": 44660 + }, + { + "epoch": 2.9, + "learning_rate": 0.0001269414685131957, + "loss": 0.2347, + "step": 44670 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012690227332113927, + "loss": 0.1375, + "step": 44680 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012686307812908283, + "loss": 0.1558, + "step": 44690 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012682388293702636, + "loss": 0.1797, + "step": 44700 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012678468774496993, + "loss": 0.2125, + "step": 44710 + }, + { + "epoch": 2.9, + "learning_rate": 0.0001267454925529135, + "loss": 0.2213, + "step": 44720 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012670629736085705, + "loss": 0.1384, + "step": 44730 + }, + { + "epoch": 2.9, + "learning_rate": 0.0001266671021688006, + "loss": 0.1684, + "step": 44740 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012662790697674418, + "loss": 0.1698, + "step": 44750 + }, + { + "epoch": 2.9, + "learning_rate": 0.00012658871178468774, + "loss": 0.188, + "step": 44760 + }, + { + "epoch": 2.91, + "learning_rate": 0.0001265495165926313, + "loss": 0.241, + "step": 44770 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012651032140057486, + "loss": 0.1261, + "step": 44780 + }, + { + "epoch": 2.91, + "learning_rate": 0.0001264711262085184, + "loss": 0.1475, + "step": 44790 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012643193101646196, + "loss": 0.1591, + "step": 44800 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012639273582440552, + "loss": 0.2024, + "step": 44810 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012635354063234908, + "loss": 0.2283, + "step": 44820 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012631434544029265, + "loss": 0.1137, + "step": 44830 + }, + { + "epoch": 2.91, + "learning_rate": 0.0001262751502482362, + "loss": 0.1654, + "step": 44840 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012623595505617977, + "loss": 0.1711, + "step": 44850 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012619675986412333, + "loss": 0.173, + "step": 44860 + }, + { + "epoch": 2.91, + "learning_rate": 0.0001261575646720669, + "loss": 0.2265, + "step": 44870 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012611836948001046, + "loss": 0.1449, + "step": 44880 + }, + { + "epoch": 2.91, + "learning_rate": 0.000126079174287954, + "loss": 0.1564, + "step": 44890 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012603997909589756, + "loss": 0.1726, + "step": 44900 + }, + { + "epoch": 2.91, + "learning_rate": 0.00012600078390384112, + "loss": 0.2098, + "step": 44910 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012596158871178468, + "loss": 0.2436, + "step": 44920 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012592239351972824, + "loss": 0.1622, + "step": 44930 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001258831983276718, + "loss": 0.1466, + "step": 44940 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012584400313561534, + "loss": 0.1937, + "step": 44950 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001258048079435589, + "loss": 0.1983, + "step": 44960 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012576561275150246, + "loss": 0.2108, + "step": 44970 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012572641755944603, + "loss": 0.1382, + "step": 44980 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001256872223673896, + "loss": 0.1696, + "step": 44990 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012564802717533315, + "loss": 0.1534, + "step": 45000 + }, + { + "epoch": 2.92, + "eval_loss": 0.22254079580307007, + "eval_runtime": 1016.5194, + "eval_samples_per_second": 15.335, + "eval_steps_per_second": 1.917, + "eval_wer": 0.22215036033602933, + "step": 45000 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001256088319832767, + "loss": 0.1739, + "step": 45010 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012556963679122028, + "loss": 0.2074, + "step": 45020 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001255304415991638, + "loss": 0.1418, + "step": 45030 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012549124640710737, + "loss": 0.1643, + "step": 45040 + }, + { + "epoch": 2.92, + "learning_rate": 0.00012545205121505094, + "loss": 0.1544, + "step": 45050 + }, + { + "epoch": 2.92, + "learning_rate": 0.0001254128560229945, + "loss": 0.1906, + "step": 45060 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012537366083093806, + "loss": 0.1927, + "step": 45070 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012533446563888162, + "loss": 0.1288, + "step": 45080 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012529527044682518, + "loss": 0.1577, + "step": 45090 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012525607525476875, + "loss": 0.1672, + "step": 45100 + }, + { + "epoch": 2.93, + "learning_rate": 0.0001252168800627123, + "loss": 0.1747, + "step": 45110 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012517768487065587, + "loss": 0.2191, + "step": 45120 + }, + { + "epoch": 2.93, + "learning_rate": 0.0001251384896785994, + "loss": 0.1552, + "step": 45130 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012509929448654297, + "loss": 0.1543, + "step": 45140 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012506009929448653, + "loss": 0.1694, + "step": 45150 + }, + { + "epoch": 2.93, + "learning_rate": 0.0001250209041024301, + "loss": 0.18, + "step": 45160 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012498170891037366, + "loss": 0.2237, + "step": 45170 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012494251371831722, + "loss": 0.1378, + "step": 45180 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012490331852626075, + "loss": 0.149, + "step": 45190 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012486412333420432, + "loss": 0.1688, + "step": 45200 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012482492814214788, + "loss": 0.1942, + "step": 45210 + }, + { + "epoch": 2.93, + "learning_rate": 0.00012478573295009144, + "loss": 0.2236, + "step": 45220 + }, + { + "epoch": 2.94, + "learning_rate": 0.000124746537758035, + "loss": 0.1474, + "step": 45230 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012470734256597856, + "loss": 0.1588, + "step": 45240 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012466814737392213, + "loss": 0.1633, + "step": 45250 + }, + { + "epoch": 2.94, + "learning_rate": 0.0001246289521818657, + "loss": 0.2045, + "step": 45260 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012458975698980922, + "loss": 0.2638, + "step": 45270 + }, + { + "epoch": 2.94, + "learning_rate": 0.0001245505617977528, + "loss": 0.1586, + "step": 45280 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012451136660569635, + "loss": 0.1543, + "step": 45290 + }, + { + "epoch": 2.94, + "learning_rate": 0.0001244721714136399, + "loss": 0.1763, + "step": 45300 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012443297622158347, + "loss": 0.1753, + "step": 45310 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012439378102952704, + "loss": 0.2179, + "step": 45320 + }, + { + "epoch": 2.94, + "learning_rate": 0.0001243545858374706, + "loss": 0.1399, + "step": 45330 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012431539064541416, + "loss": 0.1534, + "step": 45340 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012427619545335772, + "loss": 0.1498, + "step": 45350 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012423700026130129, + "loss": 0.1751, + "step": 45360 + }, + { + "epoch": 2.94, + "learning_rate": 0.00012419780506924482, + "loss": 0.2402, + "step": 45370 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012415860987718838, + "loss": 0.1491, + "step": 45380 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012411941468513195, + "loss": 0.1633, + "step": 45390 + }, + { + "epoch": 2.95, + "learning_rate": 0.0001240802194930755, + "loss": 0.1652, + "step": 45400 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012404102430101907, + "loss": 0.2109, + "step": 45410 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012400182910896263, + "loss": 0.2363, + "step": 45420 + }, + { + "epoch": 2.95, + "learning_rate": 0.0001239626339169062, + "loss": 0.1332, + "step": 45430 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012392343872484973, + "loss": 0.1455, + "step": 45440 + }, + { + "epoch": 2.95, + "learning_rate": 0.0001238842435327933, + "loss": 0.1831, + "step": 45450 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012384504834073685, + "loss": 0.204, + "step": 45460 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012380585314868042, + "loss": 0.2115, + "step": 45470 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012376665795662398, + "loss": 0.1466, + "step": 45480 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012372746276456754, + "loss": 0.172, + "step": 45490 + }, + { + "epoch": 2.95, + "learning_rate": 0.0001236882675725111, + "loss": 0.1555, + "step": 45500 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012364907238045464, + "loss": 0.1995, + "step": 45510 + }, + { + "epoch": 2.95, + "learning_rate": 0.0001236098771883982, + "loss": 0.2536, + "step": 45520 + }, + { + "epoch": 2.95, + "learning_rate": 0.00012357068199634176, + "loss": 0.1602, + "step": 45530 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012353148680428533, + "loss": 0.1821, + "step": 45540 + }, + { + "epoch": 2.96, + "learning_rate": 0.0001234922916122289, + "loss": 0.1868, + "step": 45550 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012345309642017245, + "loss": 0.1666, + "step": 45560 + }, + { + "epoch": 2.96, + "learning_rate": 0.000123413901228116, + "loss": 0.2345, + "step": 45570 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012337470603605957, + "loss": 0.1401, + "step": 45580 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012333551084400314, + "loss": 0.1595, + "step": 45590 + }, + { + "epoch": 2.96, + "learning_rate": 0.0001232963156519467, + "loss": 0.1773, + "step": 45600 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012325712045989023, + "loss": 0.2186, + "step": 45610 + }, + { + "epoch": 2.96, + "learning_rate": 0.0001232179252678338, + "loss": 0.2434, + "step": 45620 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012317873007577736, + "loss": 0.133, + "step": 45630 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012313953488372092, + "loss": 0.1566, + "step": 45640 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012310033969166448, + "loss": 0.1635, + "step": 45650 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012306114449960805, + "loss": 0.1726, + "step": 45660 + }, + { + "epoch": 2.96, + "learning_rate": 0.0001230219493075516, + "loss": 0.2536, + "step": 45670 + }, + { + "epoch": 2.96, + "learning_rate": 0.00012298275411549517, + "loss": 0.1298, + "step": 45680 + }, + { + "epoch": 2.97, + "learning_rate": 0.0001229435589234387, + "loss": 0.1397, + "step": 45690 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012290436373138227, + "loss": 0.158, + "step": 45700 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012286516853932583, + "loss": 0.2032, + "step": 45710 + }, + { + "epoch": 2.97, + "learning_rate": 0.0001228259733472694, + "loss": 0.2062, + "step": 45720 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012278677815521295, + "loss": 0.15, + "step": 45730 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012274758296315652, + "loss": 0.1552, + "step": 45740 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012270838777110005, + "loss": 0.1525, + "step": 45750 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012266919257904361, + "loss": 0.1856, + "step": 45760 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012262999738698718, + "loss": 0.2174, + "step": 45770 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012259080219493074, + "loss": 0.1412, + "step": 45780 + }, + { + "epoch": 2.97, + "learning_rate": 0.0001225516070028743, + "loss": 0.1486, + "step": 45790 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012251241181081786, + "loss": 0.1354, + "step": 45800 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012247321661876143, + "loss": 0.2088, + "step": 45810 + }, + { + "epoch": 2.97, + "learning_rate": 0.000122434021426705, + "loss": 0.2298, + "step": 45820 + }, + { + "epoch": 2.97, + "learning_rate": 0.00012239482623464855, + "loss": 0.1397, + "step": 45830 + }, + { + "epoch": 2.98, + "learning_rate": 0.0001223556310425921, + "loss": 0.125, + "step": 45840 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012231643585053565, + "loss": 0.1713, + "step": 45850 + }, + { + "epoch": 2.98, + "learning_rate": 0.0001222772406584792, + "loss": 0.2147, + "step": 45860 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012223804546642277, + "loss": 0.2156, + "step": 45870 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012219885027436633, + "loss": 0.1325, + "step": 45880 + }, + { + "epoch": 2.98, + "learning_rate": 0.0001221596550823099, + "loss": 0.1297, + "step": 45890 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012212045989025346, + "loss": 0.1556, + "step": 45900 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012208126469819702, + "loss": 0.1878, + "step": 45910 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012204206950614058, + "loss": 0.2079, + "step": 45920 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012200287431408412, + "loss": 0.1379, + "step": 45930 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012196367912202768, + "loss": 0.1613, + "step": 45940 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012192448392997124, + "loss": 0.1687, + "step": 45950 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012188528873791479, + "loss": 0.2052, + "step": 45960 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012184609354585835, + "loss": 0.1951, + "step": 45970 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012180689835380192, + "loss": 0.1678, + "step": 45980 + }, + { + "epoch": 2.98, + "learning_rate": 0.00012176770316174548, + "loss": 0.1444, + "step": 45990 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012172850796968904, + "loss": 0.1498, + "step": 46000 + }, + { + "epoch": 2.99, + "eval_loss": 0.21941165626049042, + "eval_runtime": 1011.2551, + "eval_samples_per_second": 15.415, + "eval_steps_per_second": 1.927, + "eval_wer": 0.23125470853292776, + "step": 46000 + }, + { + "epoch": 2.99, + "learning_rate": 0.0001216893127776326, + "loss": 0.1843, + "step": 46010 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012165011758557615, + "loss": 0.2117, + "step": 46020 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012161092239351971, + "loss": 0.1401, + "step": 46030 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012157172720146328, + "loss": 0.1441, + "step": 46040 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012153253200940684, + "loss": 0.1636, + "step": 46050 + }, + { + "epoch": 2.99, + "learning_rate": 0.0001214933368173504, + "loss": 0.1676, + "step": 46060 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012145414162529395, + "loss": 0.176, + "step": 46070 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012141494643323751, + "loss": 0.1614, + "step": 46080 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012137575124118108, + "loss": 0.1275, + "step": 46090 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012133655604912464, + "loss": 0.1463, + "step": 46100 + }, + { + "epoch": 2.99, + "learning_rate": 0.0001212973608570682, + "loss": 0.2049, + "step": 46110 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012125816566501175, + "loss": 0.2182, + "step": 46120 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012121897047295531, + "loss": 0.1291, + "step": 46130 + }, + { + "epoch": 2.99, + "learning_rate": 0.00012117977528089887, + "loss": 0.1443, + "step": 46140 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012114058008884244, + "loss": 0.1599, + "step": 46150 + }, + { + "epoch": 3.0, + "learning_rate": 0.000121101384896786, + "loss": 0.1725, + "step": 46160 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012106218970472955, + "loss": 0.2406, + "step": 46170 + }, + { + "epoch": 3.0, + "learning_rate": 0.0001210229945126731, + "loss": 0.1506, + "step": 46180 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012098379932061666, + "loss": 0.1518, + "step": 46190 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012094460412856022, + "loss": 0.1807, + "step": 46200 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012090540893650377, + "loss": 0.177, + "step": 46210 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012087013326365297, + "loss": 0.232, + "step": 46220 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012083093807159654, + "loss": 0.1578, + "step": 46230 + }, + { + "epoch": 3.0, + "learning_rate": 0.0001207917428795401, + "loss": 0.1164, + "step": 46240 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012075254768748366, + "loss": 0.1768, + "step": 46250 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012071335249542722, + "loss": 0.1367, + "step": 46260 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012067415730337077, + "loss": 0.1748, + "step": 46270 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012063496211131433, + "loss": 0.1494, + "step": 46280 + }, + { + "epoch": 3.0, + "learning_rate": 0.0001205957669192579, + "loss": 0.1515, + "step": 46290 + }, + { + "epoch": 3.0, + "learning_rate": 0.00012055657172720146, + "loss": 0.1534, + "step": 46300 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012051737653514502, + "loss": 0.1681, + "step": 46310 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012047818134308857, + "loss": 0.1264, + "step": 46320 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012043898615103213, + "loss": 0.1791, + "step": 46330 + }, + { + "epoch": 3.01, + "learning_rate": 0.0001203997909589757, + "loss": 0.1068, + "step": 46340 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012036059576691926, + "loss": 0.1204, + "step": 46350 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012032140057486282, + "loss": 0.1702, + "step": 46360 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012028220538280635, + "loss": 0.165, + "step": 46370 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012024301019074992, + "loss": 0.1885, + "step": 46380 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012020381499869348, + "loss": 0.1025, + "step": 46390 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012016461980663704, + "loss": 0.1531, + "step": 46400 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012012542461458059, + "loss": 0.1494, + "step": 46410 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012008622942252415, + "loss": 0.1957, + "step": 46420 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012004703423046771, + "loss": 0.2018, + "step": 46430 + }, + { + "epoch": 3.01, + "learning_rate": 0.00012000783903841128, + "loss": 0.1264, + "step": 46440 + }, + { + "epoch": 3.01, + "learning_rate": 0.00011996864384635484, + "loss": 0.1566, + "step": 46450 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011992944865429839, + "loss": 0.1622, + "step": 46460 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011989025346224195, + "loss": 0.1711, + "step": 46470 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011985105827018551, + "loss": 0.1566, + "step": 46480 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011981186307812907, + "loss": 0.1238, + "step": 46490 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011977266788607264, + "loss": 0.1449, + "step": 46500 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011973347269401619, + "loss": 0.1393, + "step": 46510 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011969427750195975, + "loss": 0.1984, + "step": 46520 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011965508230990331, + "loss": 0.1732, + "step": 46530 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011961588711784687, + "loss": 0.1371, + "step": 46540 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011957669192579044, + "loss": 0.1381, + "step": 46550 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011953749673373398, + "loss": 0.1546, + "step": 46560 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011949830154167755, + "loss": 0.1428, + "step": 46570 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011945910634962111, + "loss": 0.1629, + "step": 46580 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011941991115756467, + "loss": 0.1563, + "step": 46590 + }, + { + "epoch": 3.02, + "learning_rate": 0.00011938071596550823, + "loss": 0.1398, + "step": 46600 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011934152077345178, + "loss": 0.122, + "step": 46610 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011930232558139533, + "loss": 0.182, + "step": 46620 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011926313038933889, + "loss": 0.1785, + "step": 46630 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011922393519728246, + "loss": 0.1149, + "step": 46640 + }, + { + "epoch": 3.03, + "learning_rate": 0.000119184740005226, + "loss": 0.1317, + "step": 46650 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011914554481316957, + "loss": 0.1388, + "step": 46660 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011911026914031877, + "loss": 0.1704, + "step": 46670 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011907107394826233, + "loss": 0.1711, + "step": 46680 + }, + { + "epoch": 3.03, + "learning_rate": 0.0001190318787562059, + "loss": 0.124, + "step": 46690 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011899268356414946, + "loss": 0.1547, + "step": 46700 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011895348837209301, + "loss": 0.1565, + "step": 46710 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011891429318003657, + "loss": 0.1949, + "step": 46720 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011887509798798013, + "loss": 0.1896, + "step": 46730 + }, + { + "epoch": 3.03, + "learning_rate": 0.0001188359027959237, + "loss": 0.1242, + "step": 46740 + }, + { + "epoch": 3.03, + "learning_rate": 0.00011879670760386726, + "loss": 0.1456, + "step": 46750 + }, + { + "epoch": 3.03, + "learning_rate": 0.0001187575124118108, + "loss": 0.1585, + "step": 46760 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011871831721975437, + "loss": 0.1503, + "step": 46770 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011867912202769793, + "loss": 0.1739, + "step": 46780 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011863992683564149, + "loss": 0.1312, + "step": 46790 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011860073164358505, + "loss": 0.1294, + "step": 46800 + }, + { + "epoch": 3.04, + "learning_rate": 0.0001185615364515286, + "loss": 0.1384, + "step": 46810 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011852234125947215, + "loss": 0.1592, + "step": 46820 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011848314606741571, + "loss": 0.1624, + "step": 46830 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011844395087535928, + "loss": 0.1485, + "step": 46840 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011840475568330283, + "loss": 0.1225, + "step": 46850 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011836556049124639, + "loss": 0.1569, + "step": 46860 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011832636529918995, + "loss": 0.1948, + "step": 46870 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011828717010713351, + "loss": 0.2116, + "step": 46880 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011824797491507707, + "loss": 0.1186, + "step": 46890 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011820877972302062, + "loss": 0.1486, + "step": 46900 + }, + { + "epoch": 3.04, + "learning_rate": 0.00011816958453096419, + "loss": 0.1478, + "step": 46910 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011813038933890775, + "loss": 0.2031, + "step": 46920 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011809119414685131, + "loss": 0.1629, + "step": 46930 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011805199895479487, + "loss": 0.1196, + "step": 46940 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011801280376273842, + "loss": 0.1577, + "step": 46950 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011797360857068198, + "loss": 0.1502, + "step": 46960 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011793441337862555, + "loss": 0.1625, + "step": 46970 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011789521818656911, + "loss": 0.1555, + "step": 46980 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011785602299451267, + "loss": 0.1061, + "step": 46990 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011781682780245623, + "loss": 0.1199, + "step": 47000 + }, + { + "epoch": 3.05, + "eval_loss": 0.22025685012340546, + "eval_runtime": 1003.0135, + "eval_samples_per_second": 15.541, + "eval_steps_per_second": 1.943, + "eval_wer": 0.21589600716407728, + "step": 47000 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011777763261039978, + "loss": 0.1165, + "step": 47010 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011773843741834334, + "loss": 0.1726, + "step": 47020 + }, + { + "epoch": 3.05, + "learning_rate": 0.0001176992422262869, + "loss": 0.158, + "step": 47030 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011766004703423047, + "loss": 0.1288, + "step": 47040 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011762085184217403, + "loss": 0.1423, + "step": 47050 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011758165665011757, + "loss": 0.1559, + "step": 47060 + }, + { + "epoch": 3.05, + "learning_rate": 0.00011754246145806113, + "loss": 0.1827, + "step": 47070 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011750326626600469, + "loss": 0.2013, + "step": 47080 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011746407107394824, + "loss": 0.1381, + "step": 47090 + }, + { + "epoch": 3.06, + "learning_rate": 0.0001174248758818918, + "loss": 0.1656, + "step": 47100 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011738568068983536, + "loss": 0.1668, + "step": 47110 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011734648549777893, + "loss": 0.1346, + "step": 47120 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011730729030572249, + "loss": 0.1965, + "step": 47130 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011726809511366605, + "loss": 0.1223, + "step": 47140 + }, + { + "epoch": 3.06, + "learning_rate": 0.0001172288999216096, + "loss": 0.1214, + "step": 47150 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011718970472955316, + "loss": 0.1797, + "step": 47160 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011715050953749672, + "loss": 0.2102, + "step": 47170 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011711131434544029, + "loss": 0.1708, + "step": 47180 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011707211915338385, + "loss": 0.1539, + "step": 47190 + }, + { + "epoch": 3.06, + "learning_rate": 0.0001170329239613274, + "loss": 0.1403, + "step": 47200 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011699372876927096, + "loss": 0.1303, + "step": 47210 + }, + { + "epoch": 3.06, + "learning_rate": 0.00011695453357721452, + "loss": 0.2022, + "step": 47220 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011691533838515808, + "loss": 0.211, + "step": 47230 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011687614319310165, + "loss": 0.1132, + "step": 47240 + }, + { + "epoch": 3.07, + "learning_rate": 0.0001168369480010452, + "loss": 0.1232, + "step": 47250 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011679775280898876, + "loss": 0.1519, + "step": 47260 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011675855761693232, + "loss": 0.1653, + "step": 47270 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011671936242487588, + "loss": 0.1975, + "step": 47280 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011668016723281944, + "loss": 0.169, + "step": 47290 + }, + { + "epoch": 3.07, + "learning_rate": 0.000116640972040763, + "loss": 0.1376, + "step": 47300 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011660177684870654, + "loss": 0.124, + "step": 47310 + }, + { + "epoch": 3.07, + "learning_rate": 0.0001165625816566501, + "loss": 0.1592, + "step": 47320 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011652338646459367, + "loss": 0.181, + "step": 47330 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011648419127253722, + "loss": 0.1205, + "step": 47340 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011644499608048078, + "loss": 0.1277, + "step": 47350 + }, + { + "epoch": 3.07, + "learning_rate": 0.00011640580088842434, + "loss": 0.1265, + "step": 47360 + }, + { + "epoch": 3.07, + "learning_rate": 0.0001163666056963679, + "loss": 0.1567, + "step": 47370 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011632741050431146, + "loss": 0.2144, + "step": 47380 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011628821531225501, + "loss": 0.1305, + "step": 47390 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011624902012019858, + "loss": 0.1221, + "step": 47400 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011620982492814214, + "loss": 0.139, + "step": 47410 + }, + { + "epoch": 3.08, + "learning_rate": 0.0001161706297360857, + "loss": 0.1672, + "step": 47420 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011613143454402926, + "loss": 0.1835, + "step": 47430 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011609223935197281, + "loss": 0.1313, + "step": 47440 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011605304415991637, + "loss": 0.1702, + "step": 47450 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011601384896785994, + "loss": 0.1309, + "step": 47460 + }, + { + "epoch": 3.08, + "learning_rate": 0.0001159746537758035, + "loss": 0.1523, + "step": 47470 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011593545858374706, + "loss": 0.1428, + "step": 47480 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011589626339169061, + "loss": 0.1293, + "step": 47490 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011585706819963417, + "loss": 0.1505, + "step": 47500 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011581787300757773, + "loss": 0.1222, + "step": 47510 + }, + { + "epoch": 3.08, + "learning_rate": 0.0001157786778155213, + "loss": 0.1844, + "step": 47520 + }, + { + "epoch": 3.08, + "learning_rate": 0.00011573948262346486, + "loss": 0.1642, + "step": 47530 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011570028743140841, + "loss": 0.1301, + "step": 47540 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011566109223935196, + "loss": 0.1528, + "step": 47550 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011562189704729552, + "loss": 0.182, + "step": 47560 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011558270185523908, + "loss": 0.1619, + "step": 47570 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011554350666318263, + "loss": 0.1741, + "step": 47580 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011550431147112619, + "loss": 0.1024, + "step": 47590 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011546511627906975, + "loss": 0.1449, + "step": 47600 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011542592108701332, + "loss": 0.1401, + "step": 47610 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011538672589495688, + "loss": 0.1448, + "step": 47620 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011534753070290043, + "loss": 0.1929, + "step": 47630 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011530833551084399, + "loss": 0.1184, + "step": 47640 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011526914031878755, + "loss": 0.1288, + "step": 47650 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011522994512673111, + "loss": 0.1645, + "step": 47660 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011519074993467468, + "loss": 0.189, + "step": 47670 + }, + { + "epoch": 3.09, + "learning_rate": 0.00011515155474261822, + "loss": 0.1875, + "step": 47680 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011511235955056179, + "loss": 0.1364, + "step": 47690 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011507316435850535, + "loss": 0.1209, + "step": 47700 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011503396916644891, + "loss": 0.1367, + "step": 47710 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011499477397439247, + "loss": 0.2066, + "step": 47720 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011495557878233602, + "loss": 0.1508, + "step": 47730 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011491638359027958, + "loss": 0.1397, + "step": 47740 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011487718839822315, + "loss": 0.1436, + "step": 47750 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011483799320616671, + "loss": 0.1325, + "step": 47760 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011479879801411027, + "loss": 0.176, + "step": 47770 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011475960282205382, + "loss": 0.155, + "step": 47780 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011472040762999738, + "loss": 0.1443, + "step": 47790 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011468121243794093, + "loss": 0.1266, + "step": 47800 + }, + { + "epoch": 3.1, + "learning_rate": 0.0001146420172458845, + "loss": 0.1338, + "step": 47810 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011460282205382804, + "loss": 0.169, + "step": 47820 + }, + { + "epoch": 3.1, + "learning_rate": 0.0001145636268617716, + "loss": 0.1718, + "step": 47830 + }, + { + "epoch": 3.1, + "learning_rate": 0.00011452443166971517, + "loss": 0.127, + "step": 47840 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011448523647765873, + "loss": 0.123, + "step": 47850 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011444604128560229, + "loss": 0.1641, + "step": 47860 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011440684609354584, + "loss": 0.1704, + "step": 47870 + }, + { + "epoch": 3.11, + "learning_rate": 0.0001143676509014894, + "loss": 0.1564, + "step": 47880 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011432845570943297, + "loss": 0.1317, + "step": 47890 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011428926051737653, + "loss": 0.1273, + "step": 47900 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011425006532532009, + "loss": 0.1313, + "step": 47910 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011421087013326364, + "loss": 0.169, + "step": 47920 + }, + { + "epoch": 3.11, + "learning_rate": 0.0001141716749412072, + "loss": 0.1717, + "step": 47930 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011413247974915076, + "loss": 0.1367, + "step": 47940 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011409328455709433, + "loss": 0.1293, + "step": 47950 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011405408936503789, + "loss": 0.13, + "step": 47960 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011401489417298144, + "loss": 0.2004, + "step": 47970 + }, + { + "epoch": 3.11, + "learning_rate": 0.000113975698980925, + "loss": 0.1728, + "step": 47980 + }, + { + "epoch": 3.11, + "learning_rate": 0.00011393650378886856, + "loss": 0.1249, + "step": 47990 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011389730859681212, + "loss": 0.1164, + "step": 48000 + }, + { + "epoch": 3.12, + "eval_loss": 0.21499036252498627, + "eval_runtime": 1008.3357, + "eval_samples_per_second": 15.459, + "eval_steps_per_second": 1.933, + "eval_wer": 0.20967008287017952, + "step": 48000 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011385811340475569, + "loss": 0.1439, + "step": 48010 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011381891821269923, + "loss": 0.2097, + "step": 48020 + }, + { + "epoch": 3.12, + "learning_rate": 0.0001137797230206428, + "loss": 0.1848, + "step": 48030 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011374052782858635, + "loss": 0.1161, + "step": 48040 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011370133263652991, + "loss": 0.1285, + "step": 48050 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011366213744447346, + "loss": 0.1409, + "step": 48060 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011362294225241702, + "loss": 0.1559, + "step": 48070 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011358374706036058, + "loss": 0.1641, + "step": 48080 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011354455186830414, + "loss": 0.1271, + "step": 48090 + }, + { + "epoch": 3.12, + "learning_rate": 0.0001135053566762477, + "loss": 0.1422, + "step": 48100 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011346616148419125, + "loss": 0.1282, + "step": 48110 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011342696629213482, + "loss": 0.1836, + "step": 48120 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011338777110007838, + "loss": 0.1867, + "step": 48130 + }, + { + "epoch": 3.12, + "learning_rate": 0.00011334857590802194, + "loss": 0.1235, + "step": 48140 + }, + { + "epoch": 3.12, + "learning_rate": 0.0001133093807159655, + "loss": 0.1366, + "step": 48150 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011327018552390905, + "loss": 0.1249, + "step": 48160 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011323099033185261, + "loss": 0.1635, + "step": 48170 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011319179513979618, + "loss": 0.1593, + "step": 48180 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011315259994773974, + "loss": 0.1193, + "step": 48190 + }, + { + "epoch": 3.13, + "learning_rate": 0.0001131134047556833, + "loss": 0.1206, + "step": 48200 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011307420956362685, + "loss": 0.1308, + "step": 48210 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011303501437157041, + "loss": 0.1379, + "step": 48220 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011299581917951397, + "loss": 0.1919, + "step": 48230 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011295662398745754, + "loss": 0.1431, + "step": 48240 + }, + { + "epoch": 3.13, + "learning_rate": 0.0001129174287954011, + "loss": 0.1296, + "step": 48250 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011287823360334465, + "loss": 0.1952, + "step": 48260 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011283903841128821, + "loss": 0.1721, + "step": 48270 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011279984321923177, + "loss": 0.1734, + "step": 48280 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011276064802717532, + "loss": 0.1319, + "step": 48290 + }, + { + "epoch": 3.13, + "learning_rate": 0.00011272145283511887, + "loss": 0.1389, + "step": 48300 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011268225764306243, + "loss": 0.1446, + "step": 48310 + }, + { + "epoch": 3.14, + "learning_rate": 0.000112643062451006, + "loss": 0.1812, + "step": 48320 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011260386725894956, + "loss": 0.148, + "step": 48330 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011256467206689312, + "loss": 0.1171, + "step": 48340 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011252547687483667, + "loss": 0.1374, + "step": 48350 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011248628168278023, + "loss": 0.139, + "step": 48360 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011244708649072379, + "loss": 0.1752, + "step": 48370 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011240789129866735, + "loss": 0.1667, + "step": 48380 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011236869610661092, + "loss": 0.1481, + "step": 48390 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011232950091455447, + "loss": 0.1498, + "step": 48400 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011229030572249803, + "loss": 0.1561, + "step": 48410 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011225111053044159, + "loss": 0.1554, + "step": 48420 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011221191533838515, + "loss": 0.2199, + "step": 48430 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011217272014632871, + "loss": 0.1228, + "step": 48440 + }, + { + "epoch": 3.14, + "learning_rate": 0.00011213352495427226, + "loss": 0.1268, + "step": 48450 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011209432976221583, + "loss": 0.1354, + "step": 48460 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011205513457015939, + "loss": 0.1723, + "step": 48470 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011201593937810295, + "loss": 0.1651, + "step": 48480 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011197674418604651, + "loss": 0.128, + "step": 48490 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011193754899399006, + "loss": 0.1416, + "step": 48500 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011189835380193362, + "loss": 0.1334, + "step": 48510 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011185915860987719, + "loss": 0.1867, + "step": 48520 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011181996341782073, + "loss": 0.2042, + "step": 48530 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011178076822576428, + "loss": 0.1562, + "step": 48540 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011174157303370785, + "loss": 0.1459, + "step": 48550 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011170237784165141, + "loss": 0.1404, + "step": 48560 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011166318264959497, + "loss": 0.1831, + "step": 48570 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011162398745753853, + "loss": 0.1612, + "step": 48580 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011158479226548208, + "loss": 0.1143, + "step": 48590 + }, + { + "epoch": 3.15, + "learning_rate": 0.00011154559707342564, + "loss": 0.1381, + "step": 48600 + }, + { + "epoch": 3.15, + "learning_rate": 0.0001115064018813692, + "loss": 0.1677, + "step": 48610 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011146720668931277, + "loss": 0.1479, + "step": 48620 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011142801149725633, + "loss": 0.18, + "step": 48630 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011138881630519988, + "loss": 0.1586, + "step": 48640 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011134962111314344, + "loss": 0.1455, + "step": 48650 + }, + { + "epoch": 3.16, + "learning_rate": 0.000111310425921087, + "loss": 0.1482, + "step": 48660 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011127123072903057, + "loss": 0.1881, + "step": 48670 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011123203553697413, + "loss": 0.1748, + "step": 48680 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011119284034491768, + "loss": 0.1324, + "step": 48690 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011115364515286124, + "loss": 0.1357, + "step": 48700 + }, + { + "epoch": 3.16, + "learning_rate": 0.0001111144499608048, + "loss": 0.1844, + "step": 48710 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011107525476874836, + "loss": 0.1335, + "step": 48720 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011103605957669193, + "loss": 0.1626, + "step": 48730 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011099686438463548, + "loss": 0.131, + "step": 48740 + }, + { + "epoch": 3.16, + "learning_rate": 0.00011095766919257904, + "loss": 0.1231, + "step": 48750 + }, + { + "epoch": 3.16, + "learning_rate": 0.0001109184740005226, + "loss": 0.1749, + "step": 48760 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011087927880846616, + "loss": 0.1349, + "step": 48770 + }, + { + "epoch": 3.17, + "learning_rate": 0.0001108400836164097, + "loss": 0.207, + "step": 48780 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011080088842435326, + "loss": 0.1489, + "step": 48790 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011076169323229682, + "loss": 0.144, + "step": 48800 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011072249804024038, + "loss": 0.1561, + "step": 48810 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011068330284818395, + "loss": 0.1486, + "step": 48820 + }, + { + "epoch": 3.17, + "learning_rate": 0.0001106441076561275, + "loss": 0.1401, + "step": 48830 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011060491246407106, + "loss": 0.1496, + "step": 48840 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011056571727201462, + "loss": 0.1329, + "step": 48850 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011052652207995818, + "loss": 0.135, + "step": 48860 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011048732688790174, + "loss": 0.1807, + "step": 48870 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011044813169584529, + "loss": 0.183, + "step": 48880 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011040893650378886, + "loss": 0.1061, + "step": 48890 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011036974131173242, + "loss": 0.1463, + "step": 48900 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011033054611967598, + "loss": 0.1483, + "step": 48910 + }, + { + "epoch": 3.17, + "learning_rate": 0.00011029135092761954, + "loss": 0.1746, + "step": 48920 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011025215573556309, + "loss": 0.1718, + "step": 48930 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011021296054350665, + "loss": 0.1154, + "step": 48940 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011017376535145022, + "loss": 0.1568, + "step": 48950 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011013457015939378, + "loss": 0.1557, + "step": 48960 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011009537496733734, + "loss": 0.1999, + "step": 48970 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011005617977528089, + "loss": 0.2035, + "step": 48980 + }, + { + "epoch": 3.18, + "learning_rate": 0.00011001698458322445, + "loss": 0.1463, + "step": 48990 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010997778939116801, + "loss": 0.146, + "step": 49000 + }, + { + "epoch": 3.18, + "eval_loss": 0.20991802215576172, + "eval_runtime": 1007.8079, + "eval_samples_per_second": 15.467, + "eval_steps_per_second": 1.934, + "eval_wer": 0.20761609643075435, + "step": 49000 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010993859419911158, + "loss": 0.1775, + "step": 49010 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010989939900705511, + "loss": 0.171, + "step": 49020 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010986020381499867, + "loss": 0.1854, + "step": 49030 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010982100862294224, + "loss": 0.1117, + "step": 49040 + }, + { + "epoch": 3.18, + "learning_rate": 0.0001097818134308858, + "loss": 0.1403, + "step": 49050 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010974261823882936, + "loss": 0.1513, + "step": 49060 + }, + { + "epoch": 3.18, + "learning_rate": 0.00010970342304677291, + "loss": 0.1616, + "step": 49070 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010966422785471647, + "loss": 0.1759, + "step": 49080 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010962503266266003, + "loss": 0.1068, + "step": 49090 + }, + { + "epoch": 3.19, + "learning_rate": 0.0001095858374706036, + "loss": 0.1286, + "step": 49100 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010954664227854716, + "loss": 0.1266, + "step": 49110 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010950744708649071, + "loss": 0.1889, + "step": 49120 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010946825189443427, + "loss": 0.1218, + "step": 49130 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010942905670237783, + "loss": 0.1358, + "step": 49140 + }, + { + "epoch": 3.19, + "learning_rate": 0.0001093898615103214, + "loss": 0.1492, + "step": 49150 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010935066631826496, + "loss": 0.1428, + "step": 49160 + }, + { + "epoch": 3.19, + "learning_rate": 0.0001093114711262085, + "loss": 0.1613, + "step": 49170 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010927227593415207, + "loss": 0.1764, + "step": 49180 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010923308074209563, + "loss": 0.1238, + "step": 49190 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010919388555003919, + "loss": 0.1474, + "step": 49200 + }, + { + "epoch": 3.19, + "learning_rate": 0.00010915469035798275, + "loss": 0.1441, + "step": 49210 + }, + { + "epoch": 3.19, + "learning_rate": 0.0001091154951659263, + "loss": 0.175, + "step": 49220 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010907629997386986, + "loss": 0.1771, + "step": 49230 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010903710478181343, + "loss": 0.1268, + "step": 49240 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010899790958975699, + "loss": 0.1509, + "step": 49250 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010895871439770055, + "loss": 0.1585, + "step": 49260 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010891951920564409, + "loss": 0.1679, + "step": 49270 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010888032401358765, + "loss": 0.1636, + "step": 49280 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010884112882153121, + "loss": 0.1111, + "step": 49290 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010880193362947477, + "loss": 0.136, + "step": 49300 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010876273843741832, + "loss": 0.1364, + "step": 49310 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010872354324536188, + "loss": 0.1552, + "step": 49320 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010868434805330545, + "loss": 0.168, + "step": 49330 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010864515286124901, + "loss": 0.1334, + "step": 49340 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010860595766919257, + "loss": 0.1288, + "step": 49350 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010856676247713612, + "loss": 0.1544, + "step": 49360 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010852756728507968, + "loss": 0.1727, + "step": 49370 + }, + { + "epoch": 3.2, + "learning_rate": 0.00010848837209302325, + "loss": 0.208, + "step": 49380 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010844917690096681, + "loss": 0.1533, + "step": 49390 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010840998170891037, + "loss": 0.1223, + "step": 49400 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010837078651685392, + "loss": 0.1185, + "step": 49410 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010833159132479748, + "loss": 0.2136, + "step": 49420 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010829239613274104, + "loss": 0.1769, + "step": 49430 + }, + { + "epoch": 3.21, + "learning_rate": 0.0001082532009406846, + "loss": 0.1139, + "step": 49440 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010821400574862817, + "loss": 0.1104, + "step": 49450 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010817481055657172, + "loss": 0.1542, + "step": 49460 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010813561536451528, + "loss": 0.1651, + "step": 49470 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010809642017245884, + "loss": 0.1882, + "step": 49480 + }, + { + "epoch": 3.21, + "learning_rate": 0.0001080572249804024, + "loss": 0.1031, + "step": 49490 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010801802978834597, + "loss": 0.1264, + "step": 49500 + }, + { + "epoch": 3.21, + "learning_rate": 0.0001079788345962895, + "loss": 0.1484, + "step": 49510 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010793963940423306, + "loss": 0.1387, + "step": 49520 + }, + { + "epoch": 3.21, + "learning_rate": 0.00010790044421217663, + "loss": 0.177, + "step": 49530 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010786124902012019, + "loss": 0.1426, + "step": 49540 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010782205382806374, + "loss": 0.1646, + "step": 49550 + }, + { + "epoch": 3.22, + "learning_rate": 0.0001077828586360073, + "loss": 0.1536, + "step": 49560 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010774366344395086, + "loss": 0.1646, + "step": 49570 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010770446825189442, + "loss": 0.2087, + "step": 49580 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010766527305983799, + "loss": 0.1059, + "step": 49590 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010762607786778153, + "loss": 0.1299, + "step": 49600 + }, + { + "epoch": 3.22, + "learning_rate": 0.0001075868826757251, + "loss": 0.1699, + "step": 49610 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010754768748366866, + "loss": 0.1847, + "step": 49620 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010750849229161222, + "loss": 0.1446, + "step": 49630 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010746929709955578, + "loss": 0.1258, + "step": 49640 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010743010190749933, + "loss": 0.1227, + "step": 49650 + }, + { + "epoch": 3.22, + "learning_rate": 0.0001073909067154429, + "loss": 0.1605, + "step": 49660 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010735171152338646, + "loss": 0.1664, + "step": 49670 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010731251633133002, + "loss": 0.146, + "step": 49680 + }, + { + "epoch": 3.22, + "learning_rate": 0.00010727332113927358, + "loss": 0.1214, + "step": 49690 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010723412594721713, + "loss": 0.127, + "step": 49700 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010719493075516069, + "loss": 0.1299, + "step": 49710 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010715573556310425, + "loss": 0.1685, + "step": 49720 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010711654037104782, + "loss": 0.213, + "step": 49730 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010707734517899138, + "loss": 0.1404, + "step": 49740 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010703814998693493, + "loss": 0.1452, + "step": 49750 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010699895479487848, + "loss": 0.1379, + "step": 49760 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010695975960282204, + "loss": 0.1702, + "step": 49770 + }, + { + "epoch": 3.23, + "learning_rate": 0.0001069205644107656, + "loss": 0.1888, + "step": 49780 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010688528873791481, + "loss": 0.1315, + "step": 49790 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010684609354585837, + "loss": 0.1337, + "step": 49800 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010680689835380192, + "loss": 0.1441, + "step": 49810 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010676770316174548, + "loss": 0.1469, + "step": 49820 + }, + { + "epoch": 3.23, + "learning_rate": 0.00010672850796968904, + "loss": 0.1594, + "step": 49830 + }, + { + "epoch": 3.23, + "learning_rate": 0.0001066893127776326, + "loss": 0.1198, + "step": 49840 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010665011758557617, + "loss": 0.1317, + "step": 49850 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010661092239351972, + "loss": 0.141, + "step": 49860 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010657172720146328, + "loss": 0.1503, + "step": 49870 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010653253200940684, + "loss": 0.1839, + "step": 49880 + }, + { + "epoch": 3.24, + "learning_rate": 0.0001064933368173504, + "loss": 0.1224, + "step": 49890 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010645414162529397, + "loss": 0.1104, + "step": 49900 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010641494643323751, + "loss": 0.1369, + "step": 49910 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010637575124118108, + "loss": 0.1505, + "step": 49920 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010633655604912464, + "loss": 0.156, + "step": 49930 + }, + { + "epoch": 3.24, + "learning_rate": 0.0001062973608570682, + "loss": 0.1176, + "step": 49940 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010625816566501176, + "loss": 0.1384, + "step": 49950 + }, + { + "epoch": 3.24, + "learning_rate": 0.0001062189704729553, + "loss": 0.1459, + "step": 49960 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010617977528089886, + "loss": 0.1981, + "step": 49970 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010614058008884242, + "loss": 0.1753, + "step": 49980 + }, + { + "epoch": 3.24, + "learning_rate": 0.00010610138489678599, + "loss": 0.1248, + "step": 49990 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010606218970472953, + "loss": 0.1329, + "step": 50000 + }, + { + "epoch": 3.25, + "eval_loss": 0.21580681204795837, + "eval_runtime": 1017.7929, + "eval_samples_per_second": 15.315, + "eval_steps_per_second": 1.915, + "eval_wer": 0.21183067760230842, + "step": 50000 + }, + { + "epoch": 3.25, + "learning_rate": 0.0001060229945126731, + "loss": 0.1545, + "step": 50010 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010598379932061666, + "loss": 0.173, + "step": 50020 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010594460412856022, + "loss": 0.1768, + "step": 50030 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010590540893650378, + "loss": 0.0984, + "step": 50040 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010586621374444733, + "loss": 0.1237, + "step": 50050 + }, + { + "epoch": 3.25, + "learning_rate": 0.0001058270185523909, + "loss": 0.1447, + "step": 50060 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010578782336033446, + "loss": 0.177, + "step": 50070 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010574862816827802, + "loss": 0.1714, + "step": 50080 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010570943297622158, + "loss": 0.1223, + "step": 50090 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010567023778416513, + "loss": 0.1507, + "step": 50100 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010563104259210869, + "loss": 0.1389, + "step": 50110 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010559184740005225, + "loss": 0.1855, + "step": 50120 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010555265220799582, + "loss": 0.1807, + "step": 50130 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010551345701593938, + "loss": 0.1403, + "step": 50140 + }, + { + "epoch": 3.25, + "learning_rate": 0.00010547426182388293, + "loss": 0.1269, + "step": 50150 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010543506663182649, + "loss": 0.1463, + "step": 50160 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010539587143977005, + "loss": 0.165, + "step": 50170 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010535667624771361, + "loss": 0.1516, + "step": 50180 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010531748105565718, + "loss": 0.1179, + "step": 50190 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010527828586360071, + "loss": 0.121, + "step": 50200 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010523909067154427, + "loss": 0.1497, + "step": 50210 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010519989547948784, + "loss": 0.1545, + "step": 50220 + }, + { + "epoch": 3.26, + "learning_rate": 0.0001051607002874314, + "loss": 0.1438, + "step": 50230 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010512150509537495, + "loss": 0.1006, + "step": 50240 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010508230990331851, + "loss": 0.1199, + "step": 50250 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010504311471126207, + "loss": 0.1509, + "step": 50260 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010500391951920563, + "loss": 0.1913, + "step": 50270 + }, + { + "epoch": 3.26, + "learning_rate": 0.0001049647243271492, + "loss": 0.1525, + "step": 50280 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010492552913509275, + "loss": 0.1188, + "step": 50290 + }, + { + "epoch": 3.26, + "learning_rate": 0.00010488633394303631, + "loss": 0.1378, + "step": 50300 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010484713875097987, + "loss": 0.1229, + "step": 50310 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010480794355892343, + "loss": 0.1542, + "step": 50320 + }, + { + "epoch": 3.27, + "learning_rate": 0.000104768748366867, + "loss": 0.1981, + "step": 50330 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010472955317481054, + "loss": 0.1216, + "step": 50340 + }, + { + "epoch": 3.27, + "learning_rate": 0.0001046903579827541, + "loss": 0.1125, + "step": 50350 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010465116279069767, + "loss": 0.133, + "step": 50360 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010461196759864123, + "loss": 0.1663, + "step": 50370 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010457277240658479, + "loss": 0.1693, + "step": 50380 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010453357721452834, + "loss": 0.1263, + "step": 50390 + }, + { + "epoch": 3.27, + "learning_rate": 0.0001044943820224719, + "loss": 0.1443, + "step": 50400 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010445518683041547, + "loss": 0.1477, + "step": 50410 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010441599163835903, + "loss": 0.194, + "step": 50420 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010437679644630259, + "loss": 0.17, + "step": 50430 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010433760125424614, + "loss": 0.1289, + "step": 50440 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010429840606218969, + "loss": 0.1367, + "step": 50450 + }, + { + "epoch": 3.27, + "learning_rate": 0.00010425921087013325, + "loss": 0.1257, + "step": 50460 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010422001567807681, + "loss": 0.1688, + "step": 50470 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010418082048602036, + "loss": 0.1944, + "step": 50480 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010414162529396392, + "loss": 0.1308, + "step": 50490 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010410243010190749, + "loss": 0.127, + "step": 50500 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010406323490985105, + "loss": 0.1691, + "step": 50510 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010402403971779461, + "loss": 0.1861, + "step": 50520 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010398484452573816, + "loss": 0.1461, + "step": 50530 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010394564933368172, + "loss": 0.1494, + "step": 50540 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010390645414162528, + "loss": 0.1535, + "step": 50550 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010386725894956885, + "loss": 0.1624, + "step": 50560 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010382806375751241, + "loss": 0.1377, + "step": 50570 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010378886856545596, + "loss": 0.1701, + "step": 50580 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010374967337339952, + "loss": 0.1347, + "step": 50590 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010371047818134308, + "loss": 0.1216, + "step": 50600 + }, + { + "epoch": 3.28, + "learning_rate": 0.00010367128298928664, + "loss": 0.1467, + "step": 50610 + }, + { + "epoch": 3.29, + "learning_rate": 0.0001036320877972302, + "loss": 0.1855, + "step": 50620 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010359289260517376, + "loss": 0.1615, + "step": 50630 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010355369741311732, + "loss": 0.1241, + "step": 50640 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010351450222106088, + "loss": 0.1407, + "step": 50650 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010347530702900444, + "loss": 0.1627, + "step": 50660 + }, + { + "epoch": 3.29, + "learning_rate": 0.000103436111836948, + "loss": 0.1654, + "step": 50670 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010339691664489155, + "loss": 0.1448, + "step": 50680 + }, + { + "epoch": 3.29, + "learning_rate": 0.0001033577214528351, + "loss": 0.1335, + "step": 50690 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010331852626077866, + "loss": 0.155, + "step": 50700 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010327933106872223, + "loss": 0.1283, + "step": 50710 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010324013587666578, + "loss": 0.2078, + "step": 50720 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010320094068460934, + "loss": 0.2031, + "step": 50730 + }, + { + "epoch": 3.29, + "learning_rate": 0.0001031617454925529, + "loss": 0.1259, + "step": 50740 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010312255030049646, + "loss": 0.1309, + "step": 50750 + }, + { + "epoch": 3.29, + "learning_rate": 0.00010308335510844002, + "loss": 0.1195, + "step": 50760 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010304415991638357, + "loss": 0.1739, + "step": 50770 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010300496472432714, + "loss": 0.2127, + "step": 50780 + }, + { + "epoch": 3.3, + "learning_rate": 0.0001029657695322707, + "loss": 0.1365, + "step": 50790 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010292657434021426, + "loss": 0.1399, + "step": 50800 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010288737914815782, + "loss": 0.158, + "step": 50810 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010284818395610137, + "loss": 0.1587, + "step": 50820 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010280898876404493, + "loss": 0.1626, + "step": 50830 + }, + { + "epoch": 3.3, + "learning_rate": 0.0001027697935719885, + "loss": 0.1244, + "step": 50840 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010273059837993206, + "loss": 0.1227, + "step": 50850 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010269140318787562, + "loss": 0.1377, + "step": 50860 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010265220799581917, + "loss": 0.1592, + "step": 50870 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010261301280376273, + "loss": 0.1636, + "step": 50880 + }, + { + "epoch": 3.3, + "learning_rate": 0.0001025738176117063, + "loss": 0.1299, + "step": 50890 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010253462241964986, + "loss": 0.1423, + "step": 50900 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010249542722759342, + "loss": 0.1813, + "step": 50910 + }, + { + "epoch": 3.3, + "learning_rate": 0.00010245623203553697, + "loss": 0.14, + "step": 50920 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010241703684348053, + "loss": 0.152, + "step": 50930 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010237784165142408, + "loss": 0.1499, + "step": 50940 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010233864645936764, + "loss": 0.1439, + "step": 50950 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010229945126731119, + "loss": 0.1774, + "step": 50960 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010226025607525475, + "loss": 0.1902, + "step": 50970 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010222106088319831, + "loss": 0.1824, + "step": 50980 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010218186569114188, + "loss": 0.1081, + "step": 50990 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010214267049908544, + "loss": 0.1335, + "step": 51000 + }, + { + "epoch": 3.31, + "eval_loss": 0.21151520311832428, + "eval_runtime": 1000.2426, + "eval_samples_per_second": 15.584, + "eval_steps_per_second": 1.949, + "eval_wer": 0.20800699350400137, + "step": 51000 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010210347530702899, + "loss": 0.1652, + "step": 51010 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010206428011497255, + "loss": 0.1658, + "step": 51020 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010202508492291611, + "loss": 0.1943, + "step": 51030 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010198588973085967, + "loss": 0.1421, + "step": 51040 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010194669453880324, + "loss": 0.1112, + "step": 51050 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010190749934674678, + "loss": 0.1339, + "step": 51060 + }, + { + "epoch": 3.31, + "learning_rate": 0.00010186830415469035, + "loss": 0.1658, + "step": 51070 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010182910896263391, + "loss": 0.1666, + "step": 51080 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010178991377057747, + "loss": 0.1299, + "step": 51090 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010175071857852103, + "loss": 0.1389, + "step": 51100 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010171152338646458, + "loss": 0.1484, + "step": 51110 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010167232819440814, + "loss": 0.1966, + "step": 51120 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010163313300235171, + "loss": 0.1948, + "step": 51130 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010159393781029527, + "loss": 0.1045, + "step": 51140 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010155474261823883, + "loss": 0.1473, + "step": 51150 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010151554742618238, + "loss": 0.1523, + "step": 51160 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010147635223412594, + "loss": 0.1412, + "step": 51170 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010143715704206949, + "loss": 0.1754, + "step": 51180 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010139796185001305, + "loss": 0.1302, + "step": 51190 + }, + { + "epoch": 3.32, + "learning_rate": 0.0001013587666579566, + "loss": 0.1339, + "step": 51200 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010131957146590016, + "loss": 0.1483, + "step": 51210 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010128037627384373, + "loss": 0.1687, + "step": 51220 + }, + { + "epoch": 3.32, + "learning_rate": 0.00010124118108178729, + "loss": 0.1866, + "step": 51230 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010120198588973085, + "loss": 0.1334, + "step": 51240 + }, + { + "epoch": 3.33, + "learning_rate": 0.0001011627906976744, + "loss": 0.1384, + "step": 51250 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010112359550561796, + "loss": 0.1471, + "step": 51260 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010108440031356152, + "loss": 0.1829, + "step": 51270 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010104520512150509, + "loss": 0.1581, + "step": 51280 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010100600992944865, + "loss": 0.1468, + "step": 51290 + }, + { + "epoch": 3.33, + "learning_rate": 0.0001009668147373922, + "loss": 0.1214, + "step": 51300 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010092761954533576, + "loss": 0.1257, + "step": 51310 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010088842435327932, + "loss": 0.1553, + "step": 51320 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010084922916122289, + "loss": 0.1752, + "step": 51330 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010081003396916645, + "loss": 0.1536, + "step": 51340 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010077083877711, + "loss": 0.1485, + "step": 51350 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010073164358505356, + "loss": 0.1587, + "step": 51360 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010069244839299712, + "loss": 0.1705, + "step": 51370 + }, + { + "epoch": 3.33, + "learning_rate": 0.00010065325320094068, + "loss": 0.1491, + "step": 51380 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010061405800888425, + "loss": 0.1302, + "step": 51390 + }, + { + "epoch": 3.34, + "learning_rate": 0.0001005748628168278, + "loss": 0.1358, + "step": 51400 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010053566762477136, + "loss": 0.1615, + "step": 51410 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010049647243271492, + "loss": 0.1668, + "step": 51420 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010045727724065847, + "loss": 0.1653, + "step": 51430 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010041808204860202, + "loss": 0.1082, + "step": 51440 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010037888685654558, + "loss": 0.1462, + "step": 51450 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010033969166448914, + "loss": 0.1262, + "step": 51460 + }, + { + "epoch": 3.34, + "learning_rate": 0.0001003004964724327, + "loss": 0.1749, + "step": 51470 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010026130128037627, + "loss": 0.1494, + "step": 51480 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010022210608831981, + "loss": 0.143, + "step": 51490 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010018291089626338, + "loss": 0.1268, + "step": 51500 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010014371570420694, + "loss": 0.1409, + "step": 51510 + }, + { + "epoch": 3.34, + "learning_rate": 0.0001001045205121505, + "loss": 0.1511, + "step": 51520 + }, + { + "epoch": 3.34, + "learning_rate": 0.00010006532532009406, + "loss": 0.1926, + "step": 51530 + }, + { + "epoch": 3.35, + "learning_rate": 0.00010002613012803761, + "loss": 0.1171, + "step": 51540 + }, + { + "epoch": 3.35, + "learning_rate": 9.998693493598117e-05, + "loss": 0.166, + "step": 51550 + }, + { + "epoch": 3.35, + "learning_rate": 9.994773974392474e-05, + "loss": 0.1236, + "step": 51560 + }, + { + "epoch": 3.35, + "learning_rate": 9.99085445518683e-05, + "loss": 0.1518, + "step": 51570 + }, + { + "epoch": 3.35, + "learning_rate": 9.986934935981186e-05, + "loss": 0.1687, + "step": 51580 + }, + { + "epoch": 3.35, + "learning_rate": 9.983015416775541e-05, + "loss": 0.1417, + "step": 51590 + }, + { + "epoch": 3.35, + "learning_rate": 9.979095897569897e-05, + "loss": 0.1237, + "step": 51600 + }, + { + "epoch": 3.35, + "learning_rate": 9.975176378364253e-05, + "loss": 0.1324, + "step": 51610 + }, + { + "epoch": 3.35, + "learning_rate": 9.97125685915861e-05, + "loss": 0.1726, + "step": 51620 + }, + { + "epoch": 3.35, + "learning_rate": 9.967337339952966e-05, + "loss": 0.1601, + "step": 51630 + }, + { + "epoch": 3.35, + "learning_rate": 9.963417820747321e-05, + "loss": 0.1341, + "step": 51640 + }, + { + "epoch": 3.35, + "learning_rate": 9.959498301541677e-05, + "loss": 0.0941, + "step": 51650 + }, + { + "epoch": 3.35, + "learning_rate": 9.955578782336033e-05, + "loss": 0.1681, + "step": 51660 + }, + { + "epoch": 3.35, + "learning_rate": 9.951659263130388e-05, + "loss": 0.1613, + "step": 51670 + }, + { + "epoch": 3.35, + "learning_rate": 9.947739743924743e-05, + "loss": 0.1866, + "step": 51680 + }, + { + "epoch": 3.35, + "learning_rate": 9.943820224719099e-05, + "loss": 0.1257, + "step": 51690 + }, + { + "epoch": 3.36, + "learning_rate": 9.939900705513455e-05, + "loss": 0.1501, + "step": 51700 + }, + { + "epoch": 3.36, + "learning_rate": 9.935981186307812e-05, + "loss": 0.1449, + "step": 51710 + }, + { + "epoch": 3.36, + "learning_rate": 9.932061667102168e-05, + "loss": 0.149, + "step": 51720 + }, + { + "epoch": 3.36, + "learning_rate": 9.928142147896523e-05, + "loss": 0.1874, + "step": 51730 + }, + { + "epoch": 3.36, + "learning_rate": 9.924222628690879e-05, + "loss": 0.1342, + "step": 51740 + }, + { + "epoch": 3.36, + "learning_rate": 9.920303109485235e-05, + "loss": 0.1454, + "step": 51750 + }, + { + "epoch": 3.36, + "learning_rate": 9.916383590279591e-05, + "loss": 0.1442, + "step": 51760 + }, + { + "epoch": 3.36, + "learning_rate": 9.912464071073948e-05, + "loss": 0.1768, + "step": 51770 + }, + { + "epoch": 3.36, + "learning_rate": 9.908544551868303e-05, + "loss": 0.1892, + "step": 51780 + }, + { + "epoch": 3.36, + "learning_rate": 9.904625032662659e-05, + "loss": 0.1157, + "step": 51790 + }, + { + "epoch": 3.36, + "learning_rate": 9.900705513457015e-05, + "loss": 0.1313, + "step": 51800 + }, + { + "epoch": 3.36, + "learning_rate": 9.896785994251371e-05, + "loss": 0.1376, + "step": 51810 + }, + { + "epoch": 3.36, + "learning_rate": 9.892866475045727e-05, + "loss": 0.1642, + "step": 51820 + }, + { + "epoch": 3.36, + "learning_rate": 9.888946955840082e-05, + "loss": 0.158, + "step": 51830 + }, + { + "epoch": 3.36, + "learning_rate": 9.885027436634439e-05, + "loss": 0.1293, + "step": 51840 + }, + { + "epoch": 3.37, + "learning_rate": 9.881107917428795e-05, + "loss": 0.1266, + "step": 51850 + }, + { + "epoch": 3.37, + "learning_rate": 9.877188398223151e-05, + "loss": 0.1321, + "step": 51860 + }, + { + "epoch": 3.37, + "learning_rate": 9.873268879017507e-05, + "loss": 0.1725, + "step": 51870 + }, + { + "epoch": 3.37, + "learning_rate": 9.869349359811862e-05, + "loss": 0.1572, + "step": 51880 + }, + { + "epoch": 3.37, + "learning_rate": 9.865429840606218e-05, + "loss": 0.1044, + "step": 51890 + }, + { + "epoch": 3.37, + "learning_rate": 9.861510321400575e-05, + "loss": 0.1401, + "step": 51900 + }, + { + "epoch": 3.37, + "learning_rate": 9.857590802194931e-05, + "loss": 0.163, + "step": 51910 + }, + { + "epoch": 3.37, + "learning_rate": 9.853671282989284e-05, + "loss": 0.1783, + "step": 51920 + }, + { + "epoch": 3.37, + "learning_rate": 9.84975176378364e-05, + "loss": 0.1697, + "step": 51930 + }, + { + "epoch": 3.37, + "learning_rate": 9.845832244577997e-05, + "loss": 0.1168, + "step": 51940 + }, + { + "epoch": 3.37, + "learning_rate": 9.841912725372353e-05, + "loss": 0.1255, + "step": 51950 + }, + { + "epoch": 3.37, + "learning_rate": 9.837993206166709e-05, + "loss": 0.1515, + "step": 51960 + }, + { + "epoch": 3.37, + "learning_rate": 9.834073686961064e-05, + "loss": 0.1918, + "step": 51970 + }, + { + "epoch": 3.37, + "learning_rate": 9.83015416775542e-05, + "loss": 0.1779, + "step": 51980 + }, + { + "epoch": 3.37, + "learning_rate": 9.826234648549777e-05, + "loss": 0.1119, + "step": 51990 + }, + { + "epoch": 3.37, + "learning_rate": 9.822315129344133e-05, + "loss": 0.1373, + "step": 52000 + }, + { + "epoch": 3.37, + "eval_loss": 0.20522776246070862, + "eval_runtime": 960.5887, + "eval_samples_per_second": 16.228, + "eval_steps_per_second": 2.029, + "eval_wer": 0.20161049594177766, + "step": 52000 + }, + { + "epoch": 3.38, + "learning_rate": 9.818395610138489e-05, + "loss": 0.1559, + "step": 52010 + }, + { + "epoch": 3.38, + "learning_rate": 9.814476090932844e-05, + "loss": 0.1766, + "step": 52020 + }, + { + "epoch": 3.38, + "learning_rate": 9.8105565717272e-05, + "loss": 0.166, + "step": 52030 + }, + { + "epoch": 3.38, + "learning_rate": 9.806637052521556e-05, + "loss": 0.1453, + "step": 52040 + }, + { + "epoch": 3.38, + "learning_rate": 9.802717533315913e-05, + "loss": 0.145, + "step": 52050 + }, + { + "epoch": 3.38, + "learning_rate": 9.798798014110269e-05, + "loss": 0.1617, + "step": 52060 + }, + { + "epoch": 3.38, + "learning_rate": 9.794878494904625e-05, + "loss": 0.1459, + "step": 52070 + }, + { + "epoch": 3.38, + "learning_rate": 9.79095897569898e-05, + "loss": 0.1716, + "step": 52080 + }, + { + "epoch": 3.38, + "learning_rate": 9.787039456493336e-05, + "loss": 0.1009, + "step": 52090 + }, + { + "epoch": 3.38, + "learning_rate": 9.783119937287692e-05, + "loss": 0.1109, + "step": 52100 + }, + { + "epoch": 3.38, + "learning_rate": 9.779200418082049e-05, + "loss": 0.1388, + "step": 52110 + }, + { + "epoch": 3.38, + "learning_rate": 9.775280898876405e-05, + "loss": 0.1612, + "step": 52120 + }, + { + "epoch": 3.38, + "learning_rate": 9.77136137967076e-05, + "loss": 0.1475, + "step": 52130 + }, + { + "epoch": 3.38, + "learning_rate": 9.767441860465116e-05, + "loss": 0.1107, + "step": 52140 + }, + { + "epoch": 3.38, + "learning_rate": 9.763522341259472e-05, + "loss": 0.1225, + "step": 52150 + }, + { + "epoch": 3.39, + "learning_rate": 9.759602822053826e-05, + "loss": 0.1267, + "step": 52160 + }, + { + "epoch": 3.39, + "learning_rate": 9.755683302848182e-05, + "loss": 0.1624, + "step": 52170 + }, + { + "epoch": 3.39, + "learning_rate": 9.751763783642538e-05, + "loss": 0.1939, + "step": 52180 + }, + { + "epoch": 3.39, + "learning_rate": 9.747844264436894e-05, + "loss": 0.1169, + "step": 52190 + }, + { + "epoch": 3.39, + "learning_rate": 9.74392474523125e-05, + "loss": 0.1459, + "step": 52200 + }, + { + "epoch": 3.39, + "learning_rate": 9.740005226025607e-05, + "loss": 0.1304, + "step": 52210 + }, + { + "epoch": 3.39, + "learning_rate": 9.736085706819962e-05, + "loss": 0.2029, + "step": 52220 + }, + { + "epoch": 3.39, + "learning_rate": 9.732166187614318e-05, + "loss": 0.184, + "step": 52230 + }, + { + "epoch": 3.39, + "learning_rate": 9.728246668408674e-05, + "loss": 0.1309, + "step": 52240 + }, + { + "epoch": 3.39, + "learning_rate": 9.72432714920303e-05, + "loss": 0.1475, + "step": 52250 + }, + { + "epoch": 3.39, + "learning_rate": 9.720407629997387e-05, + "loss": 0.1667, + "step": 52260 + }, + { + "epoch": 3.39, + "learning_rate": 9.716488110791742e-05, + "loss": 0.1591, + "step": 52270 + }, + { + "epoch": 3.39, + "learning_rate": 9.712568591586098e-05, + "loss": 0.1991, + "step": 52280 + }, + { + "epoch": 3.39, + "learning_rate": 9.708649072380454e-05, + "loss": 0.1187, + "step": 52290 + }, + { + "epoch": 3.39, + "learning_rate": 9.70472955317481e-05, + "loss": 0.1496, + "step": 52300 + }, + { + "epoch": 3.39, + "learning_rate": 9.700810033969166e-05, + "loss": 0.1453, + "step": 52310 + }, + { + "epoch": 3.4, + "learning_rate": 9.696890514763521e-05, + "loss": 0.1967, + "step": 52320 + }, + { + "epoch": 3.4, + "learning_rate": 9.692970995557878e-05, + "loss": 0.1513, + "step": 52330 + }, + { + "epoch": 3.4, + "learning_rate": 9.689051476352234e-05, + "loss": 0.1339, + "step": 52340 + }, + { + "epoch": 3.4, + "learning_rate": 9.68513195714659e-05, + "loss": 0.1294, + "step": 52350 + }, + { + "epoch": 3.4, + "learning_rate": 9.681212437940946e-05, + "loss": 0.1296, + "step": 52360 + }, + { + "epoch": 3.4, + "learning_rate": 9.677292918735301e-05, + "loss": 0.1774, + "step": 52370 + }, + { + "epoch": 3.4, + "learning_rate": 9.673373399529657e-05, + "loss": 0.1493, + "step": 52380 + }, + { + "epoch": 3.4, + "learning_rate": 9.669453880324014e-05, + "loss": 0.128, + "step": 52390 + }, + { + "epoch": 3.4, + "learning_rate": 9.66553436111837e-05, + "loss": 0.1308, + "step": 52400 + }, + { + "epoch": 3.4, + "learning_rate": 9.661614841912723e-05, + "loss": 0.1571, + "step": 52410 + }, + { + "epoch": 3.4, + "learning_rate": 9.65769532270708e-05, + "loss": 0.1719, + "step": 52420 + }, + { + "epoch": 3.4, + "learning_rate": 9.653775803501436e-05, + "loss": 0.1726, + "step": 52430 + }, + { + "epoch": 3.4, + "learning_rate": 9.649856284295792e-05, + "loss": 0.1041, + "step": 52440 + }, + { + "epoch": 3.4, + "learning_rate": 9.645936765090148e-05, + "loss": 0.1288, + "step": 52450 + }, + { + "epoch": 3.4, + "learning_rate": 9.642017245884503e-05, + "loss": 0.1405, + "step": 52460 + }, + { + "epoch": 3.41, + "learning_rate": 9.63809772667886e-05, + "loss": 0.1418, + "step": 52470 + }, + { + "epoch": 3.41, + "learning_rate": 9.634178207473216e-05, + "loss": 0.1352, + "step": 52480 + }, + { + "epoch": 3.41, + "learning_rate": 9.630258688267572e-05, + "loss": 0.1226, + "step": 52490 + }, + { + "epoch": 3.41, + "learning_rate": 9.626339169061928e-05, + "loss": 0.1361, + "step": 52500 + }, + { + "epoch": 3.41, + "learning_rate": 9.622419649856283e-05, + "loss": 0.1385, + "step": 52510 + }, + { + "epoch": 3.41, + "learning_rate": 9.618500130650639e-05, + "loss": 0.1464, + "step": 52520 + }, + { + "epoch": 3.41, + "learning_rate": 9.614580611444995e-05, + "loss": 0.1884, + "step": 52530 + }, + { + "epoch": 3.41, + "learning_rate": 9.610661092239352e-05, + "loss": 0.1271, + "step": 52540 + }, + { + "epoch": 3.41, + "learning_rate": 9.606741573033708e-05, + "loss": 0.149, + "step": 52550 + }, + { + "epoch": 3.41, + "learning_rate": 9.602822053828063e-05, + "loss": 0.1245, + "step": 52560 + }, + { + "epoch": 3.41, + "learning_rate": 9.598902534622419e-05, + "loss": 0.1524, + "step": 52570 + }, + { + "epoch": 3.41, + "learning_rate": 9.594983015416775e-05, + "loss": 0.1729, + "step": 52580 + }, + { + "epoch": 3.41, + "learning_rate": 9.591063496211131e-05, + "loss": 0.135, + "step": 52590 + }, + { + "epoch": 3.41, + "learning_rate": 9.587143977005488e-05, + "loss": 0.153, + "step": 52600 + }, + { + "epoch": 3.41, + "learning_rate": 9.583224457799842e-05, + "loss": 0.1334, + "step": 52610 + }, + { + "epoch": 3.42, + "learning_rate": 9.579304938594199e-05, + "loss": 0.1634, + "step": 52620 + }, + { + "epoch": 3.42, + "learning_rate": 9.575385419388555e-05, + "loss": 0.1395, + "step": 52630 + }, + { + "epoch": 3.42, + "learning_rate": 9.571465900182911e-05, + "loss": 0.1159, + "step": 52640 + }, + { + "epoch": 3.42, + "learning_rate": 9.567546380977267e-05, + "loss": 0.1303, + "step": 52650 + }, + { + "epoch": 3.42, + "learning_rate": 9.563626861771621e-05, + "loss": 0.1435, + "step": 52660 + }, + { + "epoch": 3.42, + "learning_rate": 9.559707342565977e-05, + "loss": 0.159, + "step": 52670 + }, + { + "epoch": 3.42, + "learning_rate": 9.555787823360333e-05, + "loss": 0.1736, + "step": 52680 + }, + { + "epoch": 3.42, + "learning_rate": 9.55186830415469e-05, + "loss": 0.1395, + "step": 52690 + }, + { + "epoch": 3.42, + "learning_rate": 9.547948784949044e-05, + "loss": 0.1116, + "step": 52700 + }, + { + "epoch": 3.42, + "learning_rate": 9.544029265743401e-05, + "loss": 0.1484, + "step": 52710 + }, + { + "epoch": 3.42, + "learning_rate": 9.540109746537757e-05, + "loss": 0.1501, + "step": 52720 + }, + { + "epoch": 3.42, + "learning_rate": 9.536582179252678e-05, + "loss": 0.172, + "step": 52730 + }, + { + "epoch": 3.42, + "learning_rate": 9.532662660047034e-05, + "loss": 0.122, + "step": 52740 + }, + { + "epoch": 3.42, + "learning_rate": 9.52874314084139e-05, + "loss": 0.102, + "step": 52750 + }, + { + "epoch": 3.42, + "learning_rate": 9.524823621635745e-05, + "loss": 0.1531, + "step": 52760 + }, + { + "epoch": 3.42, + "learning_rate": 9.520904102430101e-05, + "loss": 0.1861, + "step": 52770 + }, + { + "epoch": 3.43, + "learning_rate": 9.516984583224457e-05, + "loss": 0.1685, + "step": 52780 + }, + { + "epoch": 3.43, + "learning_rate": 9.513065064018814e-05, + "loss": 0.115, + "step": 52790 + }, + { + "epoch": 3.43, + "learning_rate": 9.50914554481317e-05, + "loss": 0.1317, + "step": 52800 + }, + { + "epoch": 3.43, + "learning_rate": 9.505226025607525e-05, + "loss": 0.1532, + "step": 52810 + }, + { + "epoch": 3.43, + "learning_rate": 9.501306506401881e-05, + "loss": 0.2076, + "step": 52820 + }, + { + "epoch": 3.43, + "learning_rate": 9.497386987196237e-05, + "loss": 0.1531, + "step": 52830 + }, + { + "epoch": 3.43, + "learning_rate": 9.493467467990593e-05, + "loss": 0.1288, + "step": 52840 + }, + { + "epoch": 3.43, + "learning_rate": 9.489547948784947e-05, + "loss": 0.1191, + "step": 52850 + }, + { + "epoch": 3.43, + "learning_rate": 9.485628429579303e-05, + "loss": 0.1394, + "step": 52860 + }, + { + "epoch": 3.43, + "learning_rate": 9.481708910373659e-05, + "loss": 0.157, + "step": 52870 + }, + { + "epoch": 3.43, + "learning_rate": 9.477789391168016e-05, + "loss": 0.151, + "step": 52880 + }, + { + "epoch": 3.43, + "learning_rate": 9.473869871962372e-05, + "loss": 0.119, + "step": 52890 + }, + { + "epoch": 3.43, + "learning_rate": 9.469950352756727e-05, + "loss": 0.1163, + "step": 52900 + }, + { + "epoch": 3.43, + "learning_rate": 9.466030833551083e-05, + "loss": 0.1294, + "step": 52910 + }, + { + "epoch": 3.43, + "learning_rate": 9.462111314345439e-05, + "loss": 0.1564, + "step": 52920 + }, + { + "epoch": 3.44, + "learning_rate": 9.458191795139795e-05, + "loss": 0.1513, + "step": 52930 + }, + { + "epoch": 3.44, + "learning_rate": 9.454272275934152e-05, + "loss": 0.1259, + "step": 52940 + }, + { + "epoch": 3.44, + "learning_rate": 9.450352756728506e-05, + "loss": 0.1394, + "step": 52950 + }, + { + "epoch": 3.44, + "learning_rate": 9.446433237522863e-05, + "loss": 0.1601, + "step": 52960 + }, + { + "epoch": 3.44, + "learning_rate": 9.442513718317219e-05, + "loss": 0.1855, + "step": 52970 + }, + { + "epoch": 3.44, + "learning_rate": 9.438594199111575e-05, + "loss": 0.1776, + "step": 52980 + }, + { + "epoch": 3.44, + "learning_rate": 9.434674679905931e-05, + "loss": 0.1304, + "step": 52990 + }, + { + "epoch": 3.44, + "learning_rate": 9.430755160700286e-05, + "loss": 0.1308, + "step": 53000 + }, + { + "epoch": 3.44, + "eval_loss": 0.20342490077018738, + "eval_runtime": 973.8611, + "eval_samples_per_second": 16.006, + "eval_steps_per_second": 2.001, + "eval_wer": 0.20323094199087433, + "step": 53000 + }, + { + "epoch": 3.44, + "learning_rate": 9.426835641494642e-05, + "loss": 0.1452, + "step": 53010 + }, + { + "epoch": 3.44, + "learning_rate": 9.422916122288999e-05, + "loss": 0.1469, + "step": 53020 + }, + { + "epoch": 3.44, + "learning_rate": 9.418996603083355e-05, + "loss": 0.1588, + "step": 53030 + }, + { + "epoch": 3.44, + "learning_rate": 9.415077083877711e-05, + "loss": 0.0891, + "step": 53040 + }, + { + "epoch": 3.44, + "learning_rate": 9.411157564672066e-05, + "loss": 0.1632, + "step": 53050 + }, + { + "epoch": 3.44, + "learning_rate": 9.407238045466422e-05, + "loss": 0.1537, + "step": 53060 + }, + { + "epoch": 3.44, + "learning_rate": 9.403318526260778e-05, + "loss": 0.1885, + "step": 53070 + }, + { + "epoch": 3.44, + "learning_rate": 9.399399007055135e-05, + "loss": 0.1843, + "step": 53080 + }, + { + "epoch": 3.45, + "learning_rate": 9.395479487849491e-05, + "loss": 0.1214, + "step": 53090 + }, + { + "epoch": 3.45, + "learning_rate": 9.391559968643844e-05, + "loss": 0.1315, + "step": 53100 + }, + { + "epoch": 3.45, + "learning_rate": 9.387640449438201e-05, + "loss": 0.1161, + "step": 53110 + }, + { + "epoch": 3.45, + "learning_rate": 9.383720930232557e-05, + "loss": 0.1642, + "step": 53120 + }, + { + "epoch": 3.45, + "learning_rate": 9.379801411026913e-05, + "loss": 0.1731, + "step": 53130 + }, + { + "epoch": 3.45, + "learning_rate": 9.375881891821268e-05, + "loss": 0.1083, + "step": 53140 + }, + { + "epoch": 3.45, + "learning_rate": 9.371962372615624e-05, + "loss": 0.1329, + "step": 53150 + }, + { + "epoch": 3.45, + "learning_rate": 9.36804285340998e-05, + "loss": 0.1456, + "step": 53160 + }, + { + "epoch": 3.45, + "learning_rate": 9.364123334204337e-05, + "loss": 0.1391, + "step": 53170 + }, + { + "epoch": 3.45, + "learning_rate": 9.360203814998693e-05, + "loss": 0.121, + "step": 53180 + }, + { + "epoch": 3.45, + "learning_rate": 9.356284295793048e-05, + "loss": 0.1378, + "step": 53190 + }, + { + "epoch": 3.45, + "learning_rate": 9.352364776587404e-05, + "loss": 0.1695, + "step": 53200 + }, + { + "epoch": 3.45, + "learning_rate": 9.34844525738176e-05, + "loss": 0.1232, + "step": 53210 + }, + { + "epoch": 3.45, + "learning_rate": 9.344525738176116e-05, + "loss": 0.1697, + "step": 53220 + }, + { + "epoch": 3.45, + "learning_rate": 9.340606218970473e-05, + "loss": 0.1684, + "step": 53230 + }, + { + "epoch": 3.46, + "learning_rate": 9.336686699764828e-05, + "loss": 0.1286, + "step": 53240 + }, + { + "epoch": 3.46, + "learning_rate": 9.332767180559184e-05, + "loss": 0.1519, + "step": 53250 + }, + { + "epoch": 3.46, + "learning_rate": 9.32884766135354e-05, + "loss": 0.1442, + "step": 53260 + }, + { + "epoch": 3.46, + "learning_rate": 9.324928142147896e-05, + "loss": 0.1947, + "step": 53270 + }, + { + "epoch": 3.46, + "learning_rate": 9.321008622942253e-05, + "loss": 0.1477, + "step": 53280 + }, + { + "epoch": 3.46, + "learning_rate": 9.317089103736607e-05, + "loss": 0.1281, + "step": 53290 + }, + { + "epoch": 3.46, + "learning_rate": 9.313169584530964e-05, + "loss": 0.1428, + "step": 53300 + }, + { + "epoch": 3.46, + "learning_rate": 9.30925006532532e-05, + "loss": 0.1689, + "step": 53310 + }, + { + "epoch": 3.46, + "learning_rate": 9.305330546119676e-05, + "loss": 0.16, + "step": 53320 + }, + { + "epoch": 3.46, + "learning_rate": 9.301411026914032e-05, + "loss": 0.1542, + "step": 53330 + }, + { + "epoch": 3.46, + "learning_rate": 9.297491507708386e-05, + "loss": 0.1133, + "step": 53340 + }, + { + "epoch": 3.46, + "learning_rate": 9.293571988502742e-05, + "loss": 0.1207, + "step": 53350 + }, + { + "epoch": 3.46, + "learning_rate": 9.289652469297098e-05, + "loss": 0.1455, + "step": 53360 + }, + { + "epoch": 3.46, + "learning_rate": 9.285732950091455e-05, + "loss": 0.1746, + "step": 53370 + }, + { + "epoch": 3.46, + "learning_rate": 9.28181343088581e-05, + "loss": 0.171, + "step": 53380 + }, + { + "epoch": 3.47, + "learning_rate": 9.277893911680166e-05, + "loss": 0.1196, + "step": 53390 + }, + { + "epoch": 3.47, + "learning_rate": 9.273974392474522e-05, + "loss": 0.1168, + "step": 53400 + }, + { + "epoch": 3.47, + "learning_rate": 9.270054873268878e-05, + "loss": 0.1386, + "step": 53410 + }, + { + "epoch": 3.47, + "learning_rate": 9.266135354063234e-05, + "loss": 0.1529, + "step": 53420 + }, + { + "epoch": 3.47, + "learning_rate": 9.262215834857589e-05, + "loss": 0.1826, + "step": 53430 + }, + { + "epoch": 3.47, + "learning_rate": 9.258296315651945e-05, + "loss": 0.137, + "step": 53440 + }, + { + "epoch": 3.47, + "learning_rate": 9.254376796446302e-05, + "loss": 0.1367, + "step": 53450 + }, + { + "epoch": 3.47, + "learning_rate": 9.250457277240658e-05, + "loss": 0.1309, + "step": 53460 + }, + { + "epoch": 3.47, + "learning_rate": 9.246537758035014e-05, + "loss": 0.1879, + "step": 53470 + }, + { + "epoch": 3.47, + "learning_rate": 9.242618238829369e-05, + "loss": 0.1544, + "step": 53480 + }, + { + "epoch": 3.47, + "learning_rate": 9.238698719623725e-05, + "loss": 0.0996, + "step": 53490 + }, + { + "epoch": 3.47, + "learning_rate": 9.234779200418081e-05, + "loss": 0.1333, + "step": 53500 + }, + { + "epoch": 3.47, + "learning_rate": 9.230859681212438e-05, + "loss": 0.133, + "step": 53510 + }, + { + "epoch": 3.47, + "learning_rate": 9.226940162006794e-05, + "loss": 0.1461, + "step": 53520 + }, + { + "epoch": 3.47, + "learning_rate": 9.223020642801149e-05, + "loss": 0.1571, + "step": 53530 + }, + { + "epoch": 3.47, + "learning_rate": 9.219101123595505e-05, + "loss": 0.1264, + "step": 53540 + }, + { + "epoch": 3.48, + "learning_rate": 9.215181604389861e-05, + "loss": 0.1351, + "step": 53550 + }, + { + "epoch": 3.48, + "learning_rate": 9.211262085184217e-05, + "loss": 0.138, + "step": 53560 + }, + { + "epoch": 3.48, + "learning_rate": 9.207342565978574e-05, + "loss": 0.1793, + "step": 53570 + }, + { + "epoch": 3.48, + "learning_rate": 9.203423046772929e-05, + "loss": 0.1643, + "step": 53580 + }, + { + "epoch": 3.48, + "learning_rate": 9.199503527567283e-05, + "loss": 0.1082, + "step": 53590 + }, + { + "epoch": 3.48, + "learning_rate": 9.19558400836164e-05, + "loss": 0.1566, + "step": 53600 + }, + { + "epoch": 3.48, + "learning_rate": 9.191664489155996e-05, + "loss": 0.15, + "step": 53610 + }, + { + "epoch": 3.48, + "learning_rate": 9.187744969950351e-05, + "loss": 0.1758, + "step": 53620 + }, + { + "epoch": 3.48, + "learning_rate": 9.183825450744707e-05, + "loss": 0.1749, + "step": 53630 + }, + { + "epoch": 3.48, + "learning_rate": 9.179905931539063e-05, + "loss": 0.1133, + "step": 53640 + }, + { + "epoch": 3.48, + "learning_rate": 9.17598641233342e-05, + "loss": 0.1241, + "step": 53650 + }, + { + "epoch": 3.48, + "learning_rate": 9.172066893127776e-05, + "loss": 0.1749, + "step": 53660 + }, + { + "epoch": 3.48, + "learning_rate": 9.16814737392213e-05, + "loss": 0.1562, + "step": 53670 + }, + { + "epoch": 3.48, + "learning_rate": 9.164227854716487e-05, + "loss": 0.1802, + "step": 53680 + }, + { + "epoch": 3.48, + "learning_rate": 9.160308335510843e-05, + "loss": 0.1201, + "step": 53690 + }, + { + "epoch": 3.49, + "learning_rate": 9.156388816305199e-05, + "loss": 0.1582, + "step": 53700 + }, + { + "epoch": 3.49, + "learning_rate": 9.152469297099555e-05, + "loss": 0.1271, + "step": 53710 + }, + { + "epoch": 3.49, + "learning_rate": 9.14854977789391e-05, + "loss": 0.1753, + "step": 53720 + }, + { + "epoch": 3.49, + "learning_rate": 9.144630258688267e-05, + "loss": 0.1479, + "step": 53730 + }, + { + "epoch": 3.49, + "learning_rate": 9.140710739482623e-05, + "loss": 0.1205, + "step": 53740 + }, + { + "epoch": 3.49, + "learning_rate": 9.136791220276979e-05, + "loss": 0.1384, + "step": 53750 + }, + { + "epoch": 3.49, + "learning_rate": 9.132871701071335e-05, + "loss": 0.1469, + "step": 53760 + }, + { + "epoch": 3.49, + "learning_rate": 9.12895218186569e-05, + "loss": 0.1602, + "step": 53770 + }, + { + "epoch": 3.49, + "learning_rate": 9.125032662660046e-05, + "loss": 0.1391, + "step": 53780 + }, + { + "epoch": 3.49, + "learning_rate": 9.121113143454403e-05, + "loss": 0.1018, + "step": 53790 + }, + { + "epoch": 3.49, + "learning_rate": 9.117193624248759e-05, + "loss": 0.1206, + "step": 53800 + }, + { + "epoch": 3.49, + "learning_rate": 9.113274105043115e-05, + "loss": 0.1463, + "step": 53810 + }, + { + "epoch": 3.49, + "learning_rate": 9.10935458583747e-05, + "loss": 0.1738, + "step": 53820 + }, + { + "epoch": 3.49, + "learning_rate": 9.105435066631825e-05, + "loss": 0.1394, + "step": 53830 + }, + { + "epoch": 3.49, + "learning_rate": 9.101515547426181e-05, + "loss": 0.1247, + "step": 53840 + }, + { + "epoch": 3.49, + "learning_rate": 9.097596028220537e-05, + "loss": 0.1185, + "step": 53850 + }, + { + "epoch": 3.5, + "learning_rate": 9.093676509014892e-05, + "loss": 0.1458, + "step": 53860 + }, + { + "epoch": 3.5, + "learning_rate": 9.089756989809248e-05, + "loss": 0.1667, + "step": 53870 + }, + { + "epoch": 3.5, + "learning_rate": 9.085837470603605e-05, + "loss": 0.1673, + "step": 53880 + }, + { + "epoch": 3.5, + "learning_rate": 9.081917951397961e-05, + "loss": 0.1452, + "step": 53890 + }, + { + "epoch": 3.5, + "learning_rate": 9.077998432192317e-05, + "loss": 0.1358, + "step": 53900 + }, + { + "epoch": 3.5, + "learning_rate": 9.074078912986672e-05, + "loss": 0.134, + "step": 53910 + }, + { + "epoch": 3.5, + "learning_rate": 9.070159393781028e-05, + "loss": 0.1626, + "step": 53920 + }, + { + "epoch": 3.5, + "learning_rate": 9.066239874575384e-05, + "loss": 0.186, + "step": 53930 + }, + { + "epoch": 3.5, + "learning_rate": 9.06232035536974e-05, + "loss": 0.1383, + "step": 53940 + }, + { + "epoch": 3.5, + "learning_rate": 9.058400836164097e-05, + "loss": 0.1418, + "step": 53950 + }, + { + "epoch": 3.5, + "learning_rate": 9.054481316958452e-05, + "loss": 0.1119, + "step": 53960 + }, + { + "epoch": 3.5, + "learning_rate": 9.050561797752808e-05, + "loss": 0.1429, + "step": 53970 + }, + { + "epoch": 3.5, + "learning_rate": 9.046642278547164e-05, + "loss": 0.1559, + "step": 53980 + }, + { + "epoch": 3.5, + "learning_rate": 9.04272275934152e-05, + "loss": 0.1227, + "step": 53990 + }, + { + "epoch": 3.5, + "learning_rate": 9.038803240135877e-05, + "loss": 0.1276, + "step": 54000 + }, + { + "epoch": 3.5, + "eval_loss": 0.20336826145648956, + "eval_runtime": 992.3926, + "eval_samples_per_second": 15.707, + "eval_steps_per_second": 1.964, + "eval_wer": 0.20279740160054582, + "step": 54000 + }, + { + "epoch": 3.51, + "learning_rate": 9.034883720930233e-05, + "loss": 0.1577, + "step": 54010 + }, + { + "epoch": 3.51, + "learning_rate": 9.030964201724588e-05, + "loss": 0.201, + "step": 54020 + }, + { + "epoch": 3.51, + "learning_rate": 9.027044682518944e-05, + "loss": 0.18, + "step": 54030 + }, + { + "epoch": 3.51, + "learning_rate": 9.0231251633133e-05, + "loss": 0.1148, + "step": 54040 + }, + { + "epoch": 3.51, + "learning_rate": 9.019205644107656e-05, + "loss": 0.1488, + "step": 54050 + }, + { + "epoch": 3.51, + "learning_rate": 9.015286124902013e-05, + "loss": 0.1164, + "step": 54060 + }, + { + "epoch": 3.51, + "learning_rate": 9.011366605696368e-05, + "loss": 0.1475, + "step": 54070 + }, + { + "epoch": 3.51, + "learning_rate": 9.007447086490722e-05, + "loss": 0.1457, + "step": 54080 + }, + { + "epoch": 3.51, + "learning_rate": 9.003527567285079e-05, + "loss": 0.1035, + "step": 54090 + }, + { + "epoch": 3.51, + "learning_rate": 8.999608048079433e-05, + "loss": 0.1126, + "step": 54100 + }, + { + "epoch": 3.51, + "learning_rate": 8.99568852887379e-05, + "loss": 0.1828, + "step": 54110 + }, + { + "epoch": 3.51, + "learning_rate": 8.991769009668146e-05, + "loss": 0.1395, + "step": 54120 + }, + { + "epoch": 3.51, + "learning_rate": 8.987849490462502e-05, + "loss": 0.1781, + "step": 54130 + }, + { + "epoch": 3.51, + "learning_rate": 8.983929971256858e-05, + "loss": 0.126, + "step": 54140 + }, + { + "epoch": 3.51, + "learning_rate": 8.980010452051213e-05, + "loss": 0.1348, + "step": 54150 + }, + { + "epoch": 3.52, + "learning_rate": 8.97609093284557e-05, + "loss": 0.143, + "step": 54160 + }, + { + "epoch": 3.52, + "learning_rate": 8.972171413639926e-05, + "loss": 0.1582, + "step": 54170 + }, + { + "epoch": 3.52, + "learning_rate": 8.968251894434282e-05, + "loss": 0.1874, + "step": 54180 + }, + { + "epoch": 3.52, + "learning_rate": 8.964332375228638e-05, + "loss": 0.1201, + "step": 54190 + }, + { + "epoch": 3.52, + "learning_rate": 8.960412856022994e-05, + "loss": 0.1386, + "step": 54200 + }, + { + "epoch": 3.52, + "learning_rate": 8.956493336817349e-05, + "loss": 0.1564, + "step": 54210 + }, + { + "epoch": 3.52, + "learning_rate": 8.952573817611706e-05, + "loss": 0.1478, + "step": 54220 + }, + { + "epoch": 3.52, + "learning_rate": 8.948654298406062e-05, + "loss": 0.1757, + "step": 54230 + }, + { + "epoch": 3.52, + "learning_rate": 8.944734779200418e-05, + "loss": 0.1157, + "step": 54240 + }, + { + "epoch": 3.52, + "learning_rate": 8.940815259994774e-05, + "loss": 0.1167, + "step": 54250 + }, + { + "epoch": 3.52, + "learning_rate": 8.936895740789129e-05, + "loss": 0.1464, + "step": 54260 + }, + { + "epoch": 3.52, + "learning_rate": 8.932976221583485e-05, + "loss": 0.1677, + "step": 54270 + }, + { + "epoch": 3.52, + "learning_rate": 8.929056702377842e-05, + "loss": 0.1889, + "step": 54280 + }, + { + "epoch": 3.52, + "learning_rate": 8.925137183172198e-05, + "loss": 0.117, + "step": 54290 + }, + { + "epoch": 3.52, + "learning_rate": 8.921217663966554e-05, + "loss": 0.1448, + "step": 54300 + }, + { + "epoch": 3.52, + "learning_rate": 8.917298144760909e-05, + "loss": 0.1371, + "step": 54310 + }, + { + "epoch": 3.53, + "learning_rate": 8.913378625555264e-05, + "loss": 0.1364, + "step": 54320 + }, + { + "epoch": 3.53, + "learning_rate": 8.90945910634962e-05, + "loss": 0.1413, + "step": 54330 + }, + { + "epoch": 3.53, + "learning_rate": 8.905539587143976e-05, + "loss": 0.1121, + "step": 54340 + }, + { + "epoch": 3.53, + "learning_rate": 8.901620067938331e-05, + "loss": 0.1225, + "step": 54350 + }, + { + "epoch": 3.53, + "learning_rate": 8.897700548732687e-05, + "loss": 0.1385, + "step": 54360 + }, + { + "epoch": 3.53, + "learning_rate": 8.893781029527044e-05, + "loss": 0.1535, + "step": 54370 + }, + { + "epoch": 3.53, + "learning_rate": 8.8898615103214e-05, + "loss": 0.146, + "step": 54380 + }, + { + "epoch": 3.53, + "learning_rate": 8.885941991115756e-05, + "loss": 0.1153, + "step": 54390 + }, + { + "epoch": 3.53, + "learning_rate": 8.882022471910111e-05, + "loss": 0.1081, + "step": 54400 + }, + { + "epoch": 3.53, + "learning_rate": 8.878102952704467e-05, + "loss": 0.1291, + "step": 54410 + }, + { + "epoch": 3.53, + "learning_rate": 8.874183433498823e-05, + "loss": 0.16, + "step": 54420 + }, + { + "epoch": 3.53, + "learning_rate": 8.87026391429318e-05, + "loss": 0.1744, + "step": 54430 + }, + { + "epoch": 3.53, + "learning_rate": 8.866344395087536e-05, + "loss": 0.116, + "step": 54440 + }, + { + "epoch": 3.53, + "learning_rate": 8.86242487588189e-05, + "loss": 0.1378, + "step": 54450 + }, + { + "epoch": 3.53, + "learning_rate": 8.858505356676247e-05, + "loss": 0.1656, + "step": 54460 + }, + { + "epoch": 3.54, + "learning_rate": 8.854585837470603e-05, + "loss": 0.1749, + "step": 54470 + }, + { + "epoch": 3.54, + "learning_rate": 8.85066631826496e-05, + "loss": 0.1646, + "step": 54480 + }, + { + "epoch": 3.54, + "learning_rate": 8.846746799059316e-05, + "loss": 0.124, + "step": 54490 + }, + { + "epoch": 3.54, + "learning_rate": 8.84282727985367e-05, + "loss": 0.1013, + "step": 54500 + }, + { + "epoch": 3.54, + "learning_rate": 8.838907760648027e-05, + "loss": 0.1551, + "step": 54510 + }, + { + "epoch": 3.54, + "learning_rate": 8.834988241442383e-05, + "loss": 0.1387, + "step": 54520 + }, + { + "epoch": 3.54, + "learning_rate": 8.831068722236739e-05, + "loss": 0.1245, + "step": 54530 + }, + { + "epoch": 3.54, + "learning_rate": 8.827149203031095e-05, + "loss": 0.1544, + "step": 54540 + }, + { + "epoch": 3.54, + "learning_rate": 8.82322968382545e-05, + "loss": 0.1101, + "step": 54550 + }, + { + "epoch": 3.54, + "learning_rate": 8.819310164619806e-05, + "loss": 0.1393, + "step": 54560 + }, + { + "epoch": 3.54, + "learning_rate": 8.815390645414161e-05, + "loss": 0.1525, + "step": 54570 + }, + { + "epoch": 3.54, + "learning_rate": 8.811471126208518e-05, + "loss": 0.155, + "step": 54580 + }, + { + "epoch": 3.54, + "learning_rate": 8.807551607002872e-05, + "loss": 0.1179, + "step": 54590 + }, + { + "epoch": 3.54, + "learning_rate": 8.803632087797229e-05, + "loss": 0.1197, + "step": 54600 + }, + { + "epoch": 3.54, + "learning_rate": 8.799712568591585e-05, + "loss": 0.1173, + "step": 54610 + }, + { + "epoch": 3.54, + "learning_rate": 8.795793049385941e-05, + "loss": 0.1437, + "step": 54620 + }, + { + "epoch": 3.55, + "learning_rate": 8.791873530180297e-05, + "loss": 0.2156, + "step": 54630 + }, + { + "epoch": 3.55, + "learning_rate": 8.787954010974652e-05, + "loss": 0.1161, + "step": 54640 + }, + { + "epoch": 3.55, + "learning_rate": 8.784034491769008e-05, + "loss": 0.1554, + "step": 54650 + }, + { + "epoch": 3.55, + "learning_rate": 8.780114972563365e-05, + "loss": 0.1347, + "step": 54660 + }, + { + "epoch": 3.55, + "learning_rate": 8.776195453357721e-05, + "loss": 0.1385, + "step": 54670 + }, + { + "epoch": 3.55, + "learning_rate": 8.772275934152077e-05, + "loss": 0.1348, + "step": 54680 + }, + { + "epoch": 3.55, + "learning_rate": 8.768356414946432e-05, + "loss": 0.1054, + "step": 54690 + }, + { + "epoch": 3.55, + "learning_rate": 8.764436895740788e-05, + "loss": 0.1463, + "step": 54700 + }, + { + "epoch": 3.55, + "learning_rate": 8.760517376535144e-05, + "loss": 0.1367, + "step": 54710 + }, + { + "epoch": 3.55, + "learning_rate": 8.756597857329501e-05, + "loss": 0.1975, + "step": 54720 + }, + { + "epoch": 3.55, + "learning_rate": 8.752678338123857e-05, + "loss": 0.1958, + "step": 54730 + }, + { + "epoch": 3.55, + "learning_rate": 8.748758818918212e-05, + "loss": 0.1272, + "step": 54740 + }, + { + "epoch": 3.55, + "learning_rate": 8.744839299712568e-05, + "loss": 0.1143, + "step": 54750 + }, + { + "epoch": 3.55, + "learning_rate": 8.740919780506924e-05, + "loss": 0.1272, + "step": 54760 + }, + { + "epoch": 3.55, + "learning_rate": 8.73700026130128e-05, + "loss": 0.1633, + "step": 54770 + }, + { + "epoch": 3.56, + "learning_rate": 8.733080742095637e-05, + "loss": 0.1483, + "step": 54780 + }, + { + "epoch": 3.56, + "learning_rate": 8.729161222889992e-05, + "loss": 0.098, + "step": 54790 + }, + { + "epoch": 3.56, + "learning_rate": 8.725241703684348e-05, + "loss": 0.1121, + "step": 54800 + }, + { + "epoch": 3.56, + "learning_rate": 8.721322184478704e-05, + "loss": 0.1383, + "step": 54810 + }, + { + "epoch": 3.56, + "learning_rate": 8.717402665273059e-05, + "loss": 0.1722, + "step": 54820 + }, + { + "epoch": 3.56, + "learning_rate": 8.713483146067414e-05, + "loss": 0.1726, + "step": 54830 + }, + { + "epoch": 3.56, + "learning_rate": 8.70956362686177e-05, + "loss": 0.1054, + "step": 54840 + }, + { + "epoch": 3.56, + "learning_rate": 8.705644107656126e-05, + "loss": 0.1344, + "step": 54850 + }, + { + "epoch": 3.56, + "learning_rate": 8.701724588450482e-05, + "loss": 0.1486, + "step": 54860 + }, + { + "epoch": 3.56, + "learning_rate": 8.697805069244839e-05, + "loss": 0.1405, + "step": 54870 + }, + { + "epoch": 3.56, + "learning_rate": 8.693885550039194e-05, + "loss": 0.1922, + "step": 54880 + }, + { + "epoch": 3.56, + "learning_rate": 8.68996603083355e-05, + "loss": 0.1016, + "step": 54890 + }, + { + "epoch": 3.56, + "learning_rate": 8.686046511627906e-05, + "loss": 0.1353, + "step": 54900 + }, + { + "epoch": 3.56, + "learning_rate": 8.682126992422262e-05, + "loss": 0.1322, + "step": 54910 + }, + { + "epoch": 3.56, + "learning_rate": 8.678207473216619e-05, + "loss": 0.151, + "step": 54920 + }, + { + "epoch": 3.57, + "learning_rate": 8.674287954010973e-05, + "loss": 0.1676, + "step": 54930 + }, + { + "epoch": 3.57, + "learning_rate": 8.67036843480533e-05, + "loss": 0.1069, + "step": 54940 + }, + { + "epoch": 3.57, + "learning_rate": 8.666448915599686e-05, + "loss": 0.1327, + "step": 54950 + }, + { + "epoch": 3.57, + "learning_rate": 8.662529396394042e-05, + "loss": 0.1224, + "step": 54960 + }, + { + "epoch": 3.57, + "learning_rate": 8.658609877188398e-05, + "loss": 0.1587, + "step": 54970 + }, + { + "epoch": 3.57, + "learning_rate": 8.654690357982753e-05, + "loss": 0.1688, + "step": 54980 + }, + { + "epoch": 3.57, + "learning_rate": 8.65077083877711e-05, + "loss": 0.1214, + "step": 54990 + }, + { + "epoch": 3.57, + "learning_rate": 8.646851319571466e-05, + "loss": 0.1342, + "step": 55000 + }, + { + "epoch": 3.57, + "eval_loss": 0.19979050755500793, + "eval_runtime": 992.4488, + "eval_samples_per_second": 15.707, + "eval_steps_per_second": 1.964, + "eval_wer": 0.1976233457946582, + "step": 55000 + }, + { + "epoch": 3.57, + "learning_rate": 8.642931800365822e-05, + "loss": 0.1375, + "step": 55010 + }, + { + "epoch": 3.57, + "learning_rate": 8.639012281160178e-05, + "loss": 0.1789, + "step": 55020 + }, + { + "epoch": 3.57, + "learning_rate": 8.635092761954533e-05, + "loss": 0.1632, + "step": 55030 + }, + { + "epoch": 3.57, + "learning_rate": 8.631173242748889e-05, + "loss": 0.1175, + "step": 55040 + }, + { + "epoch": 3.57, + "learning_rate": 8.627253723543245e-05, + "loss": 0.1194, + "step": 55050 + }, + { + "epoch": 3.57, + "learning_rate": 8.6233342043376e-05, + "loss": 0.1527, + "step": 55060 + }, + { + "epoch": 3.57, + "learning_rate": 8.619414685131955e-05, + "loss": 0.1488, + "step": 55070 + }, + { + "epoch": 3.57, + "learning_rate": 8.615495165926311e-05, + "loss": 0.1734, + "step": 55080 + }, + { + "epoch": 3.58, + "learning_rate": 8.611575646720668e-05, + "loss": 0.1213, + "step": 55090 + }, + { + "epoch": 3.58, + "learning_rate": 8.607656127515024e-05, + "loss": 0.1263, + "step": 55100 + }, + { + "epoch": 3.58, + "learning_rate": 8.60373660830938e-05, + "loss": 0.1052, + "step": 55110 + }, + { + "epoch": 3.58, + "learning_rate": 8.599817089103735e-05, + "loss": 0.181, + "step": 55120 + }, + { + "epoch": 3.58, + "learning_rate": 8.595897569898091e-05, + "loss": 0.164, + "step": 55130 + }, + { + "epoch": 3.58, + "learning_rate": 8.591978050692447e-05, + "loss": 0.1112, + "step": 55140 + }, + { + "epoch": 3.58, + "learning_rate": 8.588058531486804e-05, + "loss": 0.1304, + "step": 55150 + }, + { + "epoch": 3.58, + "learning_rate": 8.58413901228116e-05, + "loss": 0.1304, + "step": 55160 + }, + { + "epoch": 3.58, + "learning_rate": 8.580219493075515e-05, + "loss": 0.1624, + "step": 55170 + }, + { + "epoch": 3.58, + "learning_rate": 8.576299973869871e-05, + "loss": 0.1974, + "step": 55180 + }, + { + "epoch": 3.58, + "learning_rate": 8.572380454664227e-05, + "loss": 0.1019, + "step": 55190 + }, + { + "epoch": 3.58, + "learning_rate": 8.568460935458583e-05, + "loss": 0.1367, + "step": 55200 + }, + { + "epoch": 3.58, + "learning_rate": 8.56454141625294e-05, + "loss": 0.1349, + "step": 55210 + }, + { + "epoch": 3.58, + "learning_rate": 8.560621897047295e-05, + "loss": 0.1323, + "step": 55220 + }, + { + "epoch": 3.58, + "learning_rate": 8.556702377841651e-05, + "loss": 0.1588, + "step": 55230 + }, + { + "epoch": 3.59, + "learning_rate": 8.552782858636007e-05, + "loss": 0.1119, + "step": 55240 + }, + { + "epoch": 3.59, + "learning_rate": 8.548863339430363e-05, + "loss": 0.1215, + "step": 55250 + }, + { + "epoch": 3.59, + "learning_rate": 8.54494382022472e-05, + "loss": 0.1473, + "step": 55260 + }, + { + "epoch": 3.59, + "learning_rate": 8.541024301019074e-05, + "loss": 0.1712, + "step": 55270 + }, + { + "epoch": 3.59, + "learning_rate": 8.53710478181343e-05, + "loss": 0.181, + "step": 55280 + }, + { + "epoch": 3.59, + "learning_rate": 8.533185262607787e-05, + "loss": 0.115, + "step": 55290 + }, + { + "epoch": 3.59, + "learning_rate": 8.529265743402143e-05, + "loss": 0.1318, + "step": 55300 + }, + { + "epoch": 3.59, + "learning_rate": 8.525346224196497e-05, + "loss": 0.1337, + "step": 55310 + }, + { + "epoch": 3.59, + "learning_rate": 8.521426704990853e-05, + "loss": 0.1701, + "step": 55320 + }, + { + "epoch": 3.59, + "learning_rate": 8.517507185785209e-05, + "loss": 0.1775, + "step": 55330 + }, + { + "epoch": 3.59, + "learning_rate": 8.513587666579565e-05, + "loss": 0.0974, + "step": 55340 + }, + { + "epoch": 3.59, + "learning_rate": 8.509668147373921e-05, + "loss": 0.1374, + "step": 55350 + }, + { + "epoch": 3.59, + "learning_rate": 8.505748628168276e-05, + "loss": 0.1584, + "step": 55360 + }, + { + "epoch": 3.59, + "learning_rate": 8.501829108962633e-05, + "loss": 0.1602, + "step": 55370 + }, + { + "epoch": 3.59, + "learning_rate": 8.497909589756989e-05, + "loss": 0.1661, + "step": 55380 + }, + { + "epoch": 3.59, + "learning_rate": 8.493990070551345e-05, + "loss": 0.1276, + "step": 55390 + }, + { + "epoch": 3.6, + "learning_rate": 8.490070551345701e-05, + "loss": 0.1277, + "step": 55400 + }, + { + "epoch": 3.6, + "learning_rate": 8.486151032140056e-05, + "loss": 0.1169, + "step": 55410 + }, + { + "epoch": 3.6, + "learning_rate": 8.482231512934412e-05, + "loss": 0.1459, + "step": 55420 + }, + { + "epoch": 3.6, + "learning_rate": 8.478311993728769e-05, + "loss": 0.1412, + "step": 55430 + }, + { + "epoch": 3.6, + "learning_rate": 8.474392474523125e-05, + "loss": 0.1037, + "step": 55440 + }, + { + "epoch": 3.6, + "learning_rate": 8.470472955317481e-05, + "loss": 0.1232, + "step": 55450 + }, + { + "epoch": 3.6, + "learning_rate": 8.466553436111836e-05, + "loss": 0.1409, + "step": 55460 + }, + { + "epoch": 3.6, + "learning_rate": 8.462633916906192e-05, + "loss": 0.1504, + "step": 55470 + }, + { + "epoch": 3.6, + "learning_rate": 8.458714397700548e-05, + "loss": 0.1627, + "step": 55480 + }, + { + "epoch": 3.6, + "learning_rate": 8.454794878494905e-05, + "loss": 0.1143, + "step": 55490 + }, + { + "epoch": 3.6, + "learning_rate": 8.450875359289261e-05, + "loss": 0.1215, + "step": 55500 + }, + { + "epoch": 3.6, + "learning_rate": 8.446955840083616e-05, + "loss": 0.1338, + "step": 55510 + }, + { + "epoch": 3.6, + "learning_rate": 8.443036320877972e-05, + "loss": 0.1621, + "step": 55520 + }, + { + "epoch": 3.6, + "learning_rate": 8.439116801672328e-05, + "loss": 0.1461, + "step": 55530 + }, + { + "epoch": 3.6, + "learning_rate": 8.435197282466684e-05, + "loss": 0.1256, + "step": 55540 + }, + { + "epoch": 3.61, + "learning_rate": 8.431277763261038e-05, + "loss": 0.1305, + "step": 55550 + }, + { + "epoch": 3.61, + "learning_rate": 8.427358244055394e-05, + "loss": 0.1383, + "step": 55560 + }, + { + "epoch": 3.61, + "learning_rate": 8.42343872484975e-05, + "loss": 0.1591, + "step": 55570 + }, + { + "epoch": 3.61, + "learning_rate": 8.419519205644107e-05, + "loss": 0.137, + "step": 55580 + }, + { + "epoch": 3.61, + "learning_rate": 8.415599686438463e-05, + "loss": 0.089, + "step": 55590 + }, + { + "epoch": 3.61, + "learning_rate": 8.411680167232818e-05, + "loss": 0.1232, + "step": 55600 + }, + { + "epoch": 3.61, + "learning_rate": 8.407760648027174e-05, + "loss": 0.1319, + "step": 55610 + }, + { + "epoch": 3.61, + "learning_rate": 8.40384112882153e-05, + "loss": 0.159, + "step": 55620 + }, + { + "epoch": 3.61, + "learning_rate": 8.399921609615886e-05, + "loss": 0.1382, + "step": 55630 + }, + { + "epoch": 3.61, + "learning_rate": 8.396002090410243e-05, + "loss": 0.1153, + "step": 55640 + }, + { + "epoch": 3.61, + "learning_rate": 8.392082571204597e-05, + "loss": 0.1259, + "step": 55650 + }, + { + "epoch": 3.61, + "learning_rate": 8.388163051998954e-05, + "loss": 0.1142, + "step": 55660 + }, + { + "epoch": 3.61, + "learning_rate": 8.38424353279331e-05, + "loss": 0.1775, + "step": 55670 + }, + { + "epoch": 3.61, + "learning_rate": 8.380324013587666e-05, + "loss": 0.1457, + "step": 55680 + }, + { + "epoch": 3.61, + "learning_rate": 8.376404494382022e-05, + "loss": 0.1276, + "step": 55690 + }, + { + "epoch": 3.62, + "learning_rate": 8.372484975176377e-05, + "loss": 0.1369, + "step": 55700 + }, + { + "epoch": 3.62, + "learning_rate": 8.368565455970734e-05, + "loss": 0.125, + "step": 55710 + }, + { + "epoch": 3.62, + "learning_rate": 8.36464593676509e-05, + "loss": 0.1312, + "step": 55720 + }, + { + "epoch": 3.62, + "learning_rate": 8.360726417559446e-05, + "loss": 0.1544, + "step": 55730 + }, + { + "epoch": 3.62, + "learning_rate": 8.356806898353802e-05, + "loss": 0.1262, + "step": 55740 + }, + { + "epoch": 3.62, + "learning_rate": 8.352887379148157e-05, + "loss": 0.1326, + "step": 55750 + }, + { + "epoch": 3.62, + "learning_rate": 8.348967859942513e-05, + "loss": 0.1557, + "step": 55760 + }, + { + "epoch": 3.62, + "learning_rate": 8.34504834073687e-05, + "loss": 0.1657, + "step": 55770 + }, + { + "epoch": 3.62, + "learning_rate": 8.341128821531226e-05, + "loss": 0.157, + "step": 55780 + }, + { + "epoch": 3.62, + "learning_rate": 8.337209302325582e-05, + "loss": 0.1234, + "step": 55790 + }, + { + "epoch": 3.62, + "learning_rate": 8.333289783119936e-05, + "loss": 0.1251, + "step": 55800 + }, + { + "epoch": 3.62, + "learning_rate": 8.329370263914292e-05, + "loss": 0.129, + "step": 55810 + }, + { + "epoch": 3.62, + "learning_rate": 8.325450744708648e-05, + "loss": 0.1622, + "step": 55820 + }, + { + "epoch": 3.62, + "learning_rate": 8.321531225503004e-05, + "loss": 0.1429, + "step": 55830 + }, + { + "epoch": 3.62, + "learning_rate": 8.317611706297359e-05, + "loss": 0.1087, + "step": 55840 + }, + { + "epoch": 3.62, + "learning_rate": 8.313692187091715e-05, + "loss": 0.1192, + "step": 55850 + }, + { + "epoch": 3.63, + "learning_rate": 8.309772667886072e-05, + "loss": 0.1529, + "step": 55860 + }, + { + "epoch": 3.63, + "learning_rate": 8.305853148680428e-05, + "loss": 0.1664, + "step": 55870 + }, + { + "epoch": 3.63, + "learning_rate": 8.301933629474784e-05, + "loss": 0.1219, + "step": 55880 + }, + { + "epoch": 3.63, + "learning_rate": 8.298014110269139e-05, + "loss": 0.1339, + "step": 55890 + }, + { + "epoch": 3.63, + "learning_rate": 8.294094591063495e-05, + "loss": 0.1232, + "step": 55900 + }, + { + "epoch": 3.63, + "learning_rate": 8.290175071857851e-05, + "loss": 0.1337, + "step": 55910 + }, + { + "epoch": 3.63, + "learning_rate": 8.286255552652208e-05, + "loss": 0.1684, + "step": 55920 + }, + { + "epoch": 3.63, + "learning_rate": 8.282336033446564e-05, + "loss": 0.1329, + "step": 55930 + }, + { + "epoch": 3.63, + "learning_rate": 8.278416514240919e-05, + "loss": 0.1208, + "step": 55940 + }, + { + "epoch": 3.63, + "learning_rate": 8.274496995035275e-05, + "loss": 0.1164, + "step": 55950 + }, + { + "epoch": 3.63, + "learning_rate": 8.270577475829631e-05, + "loss": 0.1242, + "step": 55960 + }, + { + "epoch": 3.63, + "learning_rate": 8.266657956623987e-05, + "loss": 0.1499, + "step": 55970 + }, + { + "epoch": 3.63, + "learning_rate": 8.262738437418344e-05, + "loss": 0.1763, + "step": 55980 + }, + { + "epoch": 3.63, + "learning_rate": 8.258818918212698e-05, + "loss": 0.1146, + "step": 55990 + }, + { + "epoch": 3.63, + "learning_rate": 8.254899399007055e-05, + "loss": 0.1338, + "step": 56000 + }, + { + "epoch": 3.63, + "eval_loss": 0.19572068750858307, + "eval_runtime": 995.129, + "eval_samples_per_second": 15.664, + "eval_steps_per_second": 1.959, + "eval_wer": 0.19942857955110802, + "step": 56000 + }, + { + "epoch": 3.64, + "learning_rate": 8.250979879801411e-05, + "loss": 0.1366, + "step": 56010 + }, + { + "epoch": 3.64, + "learning_rate": 8.247060360595767e-05, + "loss": 0.1283, + "step": 56020 + }, + { + "epoch": 3.64, + "learning_rate": 8.243140841390123e-05, + "loss": 0.1988, + "step": 56030 + }, + { + "epoch": 3.64, + "learning_rate": 8.239221322184477e-05, + "loss": 0.1285, + "step": 56040 + }, + { + "epoch": 3.64, + "learning_rate": 8.235301802978833e-05, + "loss": 0.1083, + "step": 56050 + }, + { + "epoch": 3.64, + "learning_rate": 8.23138228377319e-05, + "loss": 0.1548, + "step": 56060 + }, + { + "epoch": 3.64, + "learning_rate": 8.227462764567546e-05, + "loss": 0.1921, + "step": 56070 + }, + { + "epoch": 3.64, + "learning_rate": 8.2235432453619e-05, + "loss": 0.1553, + "step": 56080 + }, + { + "epoch": 3.64, + "learning_rate": 8.219623726156257e-05, + "loss": 0.1094, + "step": 56090 + }, + { + "epoch": 3.64, + "learning_rate": 8.215704206950613e-05, + "loss": 0.1194, + "step": 56100 + }, + { + "epoch": 3.64, + "learning_rate": 8.211784687744969e-05, + "loss": 0.154, + "step": 56110 + }, + { + "epoch": 3.64, + "learning_rate": 8.207865168539325e-05, + "loss": 0.1485, + "step": 56120 + }, + { + "epoch": 3.64, + "learning_rate": 8.20394564933368e-05, + "loss": 0.1625, + "step": 56130 + }, + { + "epoch": 3.64, + "learning_rate": 8.200026130128036e-05, + "loss": 0.1183, + "step": 56140 + }, + { + "epoch": 3.64, + "learning_rate": 8.196106610922393e-05, + "loss": 0.124, + "step": 56150 + }, + { + "epoch": 3.64, + "learning_rate": 8.192187091716749e-05, + "loss": 0.1403, + "step": 56160 + }, + { + "epoch": 3.65, + "learning_rate": 8.188267572511105e-05, + "loss": 0.1719, + "step": 56170 + }, + { + "epoch": 3.65, + "learning_rate": 8.18434805330546e-05, + "loss": 0.2053, + "step": 56180 + }, + { + "epoch": 3.65, + "learning_rate": 8.180428534099816e-05, + "loss": 0.0952, + "step": 56190 + }, + { + "epoch": 3.65, + "learning_rate": 8.176509014894172e-05, + "loss": 0.1207, + "step": 56200 + }, + { + "epoch": 3.65, + "learning_rate": 8.172589495688529e-05, + "loss": 0.1234, + "step": 56210 + }, + { + "epoch": 3.65, + "learning_rate": 8.168669976482885e-05, + "loss": 0.16, + "step": 56220 + }, + { + "epoch": 3.65, + "learning_rate": 8.16475045727724e-05, + "loss": 0.154, + "step": 56230 + }, + { + "epoch": 3.65, + "learning_rate": 8.160830938071596e-05, + "loss": 0.1102, + "step": 56240 + }, + { + "epoch": 3.65, + "learning_rate": 8.156911418865952e-05, + "loss": 0.1182, + "step": 56250 + }, + { + "epoch": 3.65, + "learning_rate": 8.152991899660308e-05, + "loss": 0.1488, + "step": 56260 + }, + { + "epoch": 3.65, + "learning_rate": 8.149072380454665e-05, + "loss": 0.1598, + "step": 56270 + }, + { + "epoch": 3.65, + "learning_rate": 8.14515286124902e-05, + "loss": 0.1468, + "step": 56280 + }, + { + "epoch": 3.65, + "learning_rate": 8.141233342043374e-05, + "loss": 0.1194, + "step": 56290 + }, + { + "epoch": 3.65, + "learning_rate": 8.137313822837731e-05, + "loss": 0.1045, + "step": 56300 + }, + { + "epoch": 3.65, + "learning_rate": 8.133394303632087e-05, + "loss": 0.1261, + "step": 56310 + }, + { + "epoch": 3.66, + "learning_rate": 8.129474784426442e-05, + "loss": 0.1601, + "step": 56320 + }, + { + "epoch": 3.66, + "learning_rate": 8.125555265220798e-05, + "loss": 0.1813, + "step": 56330 + }, + { + "epoch": 3.66, + "learning_rate": 8.121635746015154e-05, + "loss": 0.1099, + "step": 56340 + }, + { + "epoch": 3.66, + "learning_rate": 8.11771622680951e-05, + "loss": 0.1222, + "step": 56350 + }, + { + "epoch": 3.66, + "learning_rate": 8.113796707603867e-05, + "loss": 0.1352, + "step": 56360 + }, + { + "epoch": 3.66, + "learning_rate": 8.109877188398222e-05, + "loss": 0.1401, + "step": 56370 + }, + { + "epoch": 3.66, + "learning_rate": 8.105957669192578e-05, + "loss": 0.1792, + "step": 56380 + }, + { + "epoch": 3.66, + "learning_rate": 8.102038149986934e-05, + "loss": 0.1213, + "step": 56390 + }, + { + "epoch": 3.66, + "learning_rate": 8.09811863078129e-05, + "loss": 0.1192, + "step": 56400 + }, + { + "epoch": 3.66, + "learning_rate": 8.094199111575647e-05, + "loss": 0.1251, + "step": 56410 + }, + { + "epoch": 3.66, + "learning_rate": 8.090279592370001e-05, + "loss": 0.1412, + "step": 56420 + }, + { + "epoch": 3.66, + "learning_rate": 8.086360073164358e-05, + "loss": 0.1717, + "step": 56430 + }, + { + "epoch": 3.66, + "learning_rate": 8.082440553958714e-05, + "loss": 0.1131, + "step": 56440 + }, + { + "epoch": 3.66, + "learning_rate": 8.07852103475307e-05, + "loss": 0.1208, + "step": 56450 + }, + { + "epoch": 3.66, + "learning_rate": 8.074601515547426e-05, + "loss": 0.1092, + "step": 56460 + }, + { + "epoch": 3.66, + "learning_rate": 8.070681996341781e-05, + "loss": 0.1701, + "step": 56470 + }, + { + "epoch": 3.67, + "learning_rate": 8.066762477136137e-05, + "loss": 0.1971, + "step": 56480 + }, + { + "epoch": 3.67, + "learning_rate": 8.062842957930494e-05, + "loss": 0.1149, + "step": 56490 + }, + { + "epoch": 3.67, + "learning_rate": 8.05892343872485e-05, + "loss": 0.1077, + "step": 56500 + }, + { + "epoch": 3.67, + "learning_rate": 8.055003919519206e-05, + "loss": 0.1282, + "step": 56510 + }, + { + "epoch": 3.67, + "learning_rate": 8.051084400313561e-05, + "loss": 0.1724, + "step": 56520 + }, + { + "epoch": 3.67, + "learning_rate": 8.047164881107916e-05, + "loss": 0.1522, + "step": 56530 + }, + { + "epoch": 3.67, + "learning_rate": 8.043245361902272e-05, + "loss": 0.1328, + "step": 56540 + }, + { + "epoch": 3.67, + "learning_rate": 8.039325842696628e-05, + "loss": 0.1247, + "step": 56550 + }, + { + "epoch": 3.67, + "learning_rate": 8.035406323490983e-05, + "loss": 0.1367, + "step": 56560 + }, + { + "epoch": 3.67, + "learning_rate": 8.03148680428534e-05, + "loss": 0.1422, + "step": 56570 + }, + { + "epoch": 3.67, + "learning_rate": 8.027567285079696e-05, + "loss": 0.1738, + "step": 56580 + }, + { + "epoch": 3.67, + "learning_rate": 8.023647765874052e-05, + "loss": 0.118, + "step": 56590 + }, + { + "epoch": 3.67, + "learning_rate": 8.019728246668408e-05, + "loss": 0.1315, + "step": 56600 + }, + { + "epoch": 3.67, + "learning_rate": 8.015808727462763e-05, + "loss": 0.12, + "step": 56610 + }, + { + "epoch": 3.67, + "learning_rate": 8.011889208257119e-05, + "loss": 0.1592, + "step": 56620 + }, + { + "epoch": 3.68, + "learning_rate": 8.007969689051475e-05, + "loss": 0.165, + "step": 56630 + }, + { + "epoch": 3.68, + "learning_rate": 8.004050169845832e-05, + "loss": 0.1128, + "step": 56640 + }, + { + "epoch": 3.68, + "learning_rate": 8.000130650640188e-05, + "loss": 0.1268, + "step": 56650 + }, + { + "epoch": 3.68, + "learning_rate": 7.996211131434543e-05, + "loss": 0.1278, + "step": 56660 + }, + { + "epoch": 3.68, + "learning_rate": 7.992291612228899e-05, + "loss": 0.1413, + "step": 56670 + }, + { + "epoch": 3.68, + "learning_rate": 7.988372093023255e-05, + "loss": 0.1761, + "step": 56680 + }, + { + "epoch": 3.68, + "learning_rate": 7.984452573817611e-05, + "loss": 0.1113, + "step": 56690 + }, + { + "epoch": 3.68, + "learning_rate": 7.980533054611968e-05, + "loss": 0.1248, + "step": 56700 + }, + { + "epoch": 3.68, + "learning_rate": 7.976613535406323e-05, + "loss": 0.1341, + "step": 56710 + }, + { + "epoch": 3.68, + "learning_rate": 7.972694016200679e-05, + "loss": 0.1669, + "step": 56720 + }, + { + "epoch": 3.68, + "learning_rate": 7.968774496995035e-05, + "loss": 0.1362, + "step": 56730 + }, + { + "epoch": 3.68, + "learning_rate": 7.964854977789391e-05, + "loss": 0.1304, + "step": 56740 + }, + { + "epoch": 3.68, + "learning_rate": 7.960935458583747e-05, + "loss": 0.1166, + "step": 56750 + }, + { + "epoch": 3.68, + "learning_rate": 7.957015939378102e-05, + "loss": 0.1345, + "step": 56760 + }, + { + "epoch": 3.68, + "learning_rate": 7.953096420172459e-05, + "loss": 0.1477, + "step": 56770 + }, + { + "epoch": 3.69, + "learning_rate": 7.949176900966813e-05, + "loss": 0.1628, + "step": 56780 + }, + { + "epoch": 3.69, + "learning_rate": 7.94525738176117e-05, + "loss": 0.1348, + "step": 56790 + }, + { + "epoch": 3.69, + "learning_rate": 7.941337862555525e-05, + "loss": 0.1241, + "step": 56800 + }, + { + "epoch": 3.69, + "learning_rate": 7.937418343349881e-05, + "loss": 0.1334, + "step": 56810 + }, + { + "epoch": 3.69, + "learning_rate": 7.933498824144237e-05, + "loss": 0.1574, + "step": 56820 + }, + { + "epoch": 3.69, + "learning_rate": 7.929579304938593e-05, + "loss": 0.1651, + "step": 56830 + }, + { + "epoch": 3.69, + "learning_rate": 7.92565978573295e-05, + "loss": 0.1237, + "step": 56840 + }, + { + "epoch": 3.69, + "learning_rate": 7.921740266527304e-05, + "loss": 0.1217, + "step": 56850 + }, + { + "epoch": 3.69, + "learning_rate": 7.91782074732166e-05, + "loss": 0.143, + "step": 56860 + }, + { + "epoch": 3.69, + "learning_rate": 7.913901228116017e-05, + "loss": 0.1383, + "step": 56870 + }, + { + "epoch": 3.69, + "learning_rate": 7.909981708910373e-05, + "loss": 0.1998, + "step": 56880 + }, + { + "epoch": 3.69, + "learning_rate": 7.906062189704729e-05, + "loss": 0.0961, + "step": 56890 + }, + { + "epoch": 3.69, + "learning_rate": 7.902142670499084e-05, + "loss": 0.1327, + "step": 56900 + }, + { + "epoch": 3.69, + "learning_rate": 7.89822315129344e-05, + "loss": 0.1429, + "step": 56910 + }, + { + "epoch": 3.69, + "learning_rate": 7.894303632087797e-05, + "loss": 0.1422, + "step": 56920 + }, + { + "epoch": 3.69, + "learning_rate": 7.890384112882153e-05, + "loss": 0.1622, + "step": 56930 + }, + { + "epoch": 3.7, + "learning_rate": 7.886464593676509e-05, + "loss": 0.1249, + "step": 56940 + }, + { + "epoch": 3.7, + "learning_rate": 7.882545074470865e-05, + "loss": 0.1118, + "step": 56950 + }, + { + "epoch": 3.7, + "learning_rate": 7.87862555526522e-05, + "loss": 0.1384, + "step": 56960 + }, + { + "epoch": 3.7, + "learning_rate": 7.874706036059576e-05, + "loss": 0.1444, + "step": 56970 + }, + { + "epoch": 3.7, + "learning_rate": 7.870786516853933e-05, + "loss": 0.1451, + "step": 56980 + }, + { + "epoch": 3.7, + "learning_rate": 7.866866997648289e-05, + "loss": 0.1214, + "step": 56990 + }, + { + "epoch": 3.7, + "learning_rate": 7.862947478442645e-05, + "loss": 0.1258, + "step": 57000 + }, + { + "epoch": 3.7, + "eval_loss": 0.19358104467391968, + "eval_runtime": 994.8459, + "eval_samples_per_second": 15.669, + "eval_steps_per_second": 1.959, + "eval_wer": 0.19561200267231454, + "step": 57000 + }, + { + "epoch": 3.7, + "learning_rate": 7.859027959237e-05, + "loss": 0.152, + "step": 57010 + }, + { + "epoch": 3.7, + "learning_rate": 7.855108440031355e-05, + "loss": 0.1559, + "step": 57020 + }, + { + "epoch": 3.7, + "learning_rate": 7.851188920825711e-05, + "loss": 0.1924, + "step": 57030 + }, + { + "epoch": 3.7, + "learning_rate": 7.847269401620066e-05, + "loss": 0.1344, + "step": 57040 + }, + { + "epoch": 3.7, + "learning_rate": 7.843349882414422e-05, + "loss": 0.1258, + "step": 57050 + }, + { + "epoch": 3.7, + "learning_rate": 7.839430363208778e-05, + "loss": 0.1388, + "step": 57060 + }, + { + "epoch": 3.7, + "learning_rate": 7.835510844003135e-05, + "loss": 0.1473, + "step": 57070 + }, + { + "epoch": 3.7, + "learning_rate": 7.831591324797491e-05, + "loss": 0.1663, + "step": 57080 + }, + { + "epoch": 3.71, + "learning_rate": 7.827671805591846e-05, + "loss": 0.0972, + "step": 57090 + }, + { + "epoch": 3.71, + "learning_rate": 7.823752286386202e-05, + "loss": 0.1385, + "step": 57100 + }, + { + "epoch": 3.71, + "learning_rate": 7.819832767180558e-05, + "loss": 0.142, + "step": 57110 + }, + { + "epoch": 3.71, + "learning_rate": 7.815913247974914e-05, + "loss": 0.1456, + "step": 57120 + }, + { + "epoch": 3.71, + "learning_rate": 7.81199372876927e-05, + "loss": 0.1451, + "step": 57130 + }, + { + "epoch": 3.71, + "learning_rate": 7.808074209563627e-05, + "loss": 0.1018, + "step": 57140 + }, + { + "epoch": 3.71, + "learning_rate": 7.804154690357982e-05, + "loss": 0.1314, + "step": 57150 + }, + { + "epoch": 3.71, + "learning_rate": 7.800235171152338e-05, + "loss": 0.1216, + "step": 57160 + }, + { + "epoch": 3.71, + "learning_rate": 7.796315651946694e-05, + "loss": 0.1746, + "step": 57170 + }, + { + "epoch": 3.71, + "learning_rate": 7.79239613274105e-05, + "loss": 0.1672, + "step": 57180 + }, + { + "epoch": 3.71, + "learning_rate": 7.788476613535407e-05, + "loss": 0.091, + "step": 57190 + }, + { + "epoch": 3.71, + "learning_rate": 7.784557094329762e-05, + "loss": 0.1185, + "step": 57200 + }, + { + "epoch": 3.71, + "learning_rate": 7.780637575124118e-05, + "loss": 0.132, + "step": 57210 + }, + { + "epoch": 3.71, + "learning_rate": 7.776718055918474e-05, + "loss": 0.1557, + "step": 57220 + }, + { + "epoch": 3.71, + "learning_rate": 7.77279853671283e-05, + "loss": 0.181, + "step": 57230 + }, + { + "epoch": 3.71, + "learning_rate": 7.768879017507186e-05, + "loss": 0.1148, + "step": 57240 + }, + { + "epoch": 3.72, + "learning_rate": 7.764959498301541e-05, + "loss": 0.1162, + "step": 57250 + }, + { + "epoch": 3.72, + "learning_rate": 7.761039979095898e-05, + "loss": 0.1385, + "step": 57260 + }, + { + "epoch": 3.72, + "learning_rate": 7.757120459890252e-05, + "loss": 0.1848, + "step": 57270 + }, + { + "epoch": 3.72, + "learning_rate": 7.753200940684609e-05, + "loss": 0.146, + "step": 57280 + }, + { + "epoch": 3.72, + "learning_rate": 7.749281421478964e-05, + "loss": 0.1041, + "step": 57290 + }, + { + "epoch": 3.72, + "learning_rate": 7.74536190227332e-05, + "loss": 0.1095, + "step": 57300 + }, + { + "epoch": 3.72, + "learning_rate": 7.741442383067676e-05, + "loss": 0.1581, + "step": 57310 + }, + { + "epoch": 3.72, + "learning_rate": 7.737522863862032e-05, + "loss": 0.1287, + "step": 57320 + }, + { + "epoch": 3.72, + "learning_rate": 7.733603344656388e-05, + "loss": 0.1731, + "step": 57330 + }, + { + "epoch": 3.72, + "learning_rate": 7.729683825450743e-05, + "loss": 0.1152, + "step": 57340 + }, + { + "epoch": 3.72, + "learning_rate": 7.7257643062451e-05, + "loss": 0.1479, + "step": 57350 + }, + { + "epoch": 3.72, + "learning_rate": 7.721844787039456e-05, + "loss": 0.1302, + "step": 57360 + }, + { + "epoch": 3.72, + "learning_rate": 7.717925267833812e-05, + "loss": 0.1264, + "step": 57370 + }, + { + "epoch": 3.72, + "learning_rate": 7.714005748628168e-05, + "loss": 0.1413, + "step": 57380 + }, + { + "epoch": 3.72, + "learning_rate": 7.710086229422523e-05, + "loss": 0.1173, + "step": 57390 + }, + { + "epoch": 3.73, + "learning_rate": 7.706166710216879e-05, + "loss": 0.107, + "step": 57400 + }, + { + "epoch": 3.73, + "learning_rate": 7.702247191011236e-05, + "loss": 0.1385, + "step": 57410 + }, + { + "epoch": 3.73, + "learning_rate": 7.698719623726156e-05, + "loss": 0.1563, + "step": 57420 + }, + { + "epoch": 3.73, + "learning_rate": 7.694800104520512e-05, + "loss": 0.1706, + "step": 57430 + }, + { + "epoch": 3.73, + "learning_rate": 7.690880585314869e-05, + "loss": 0.0895, + "step": 57440 + }, + { + "epoch": 3.73, + "learning_rate": 7.686961066109223e-05, + "loss": 0.1054, + "step": 57450 + }, + { + "epoch": 3.73, + "learning_rate": 7.68304154690358e-05, + "loss": 0.134, + "step": 57460 + }, + { + "epoch": 3.73, + "learning_rate": 7.679122027697935e-05, + "loss": 0.151, + "step": 57470 + }, + { + "epoch": 3.73, + "learning_rate": 7.675202508492291e-05, + "loss": 0.1853, + "step": 57480 + }, + { + "epoch": 3.73, + "learning_rate": 7.671282989286646e-05, + "loss": 0.1111, + "step": 57490 + }, + { + "epoch": 3.73, + "learning_rate": 7.667363470081002e-05, + "loss": 0.1258, + "step": 57500 + }, + { + "epoch": 3.73, + "learning_rate": 7.663443950875358e-05, + "loss": 0.1587, + "step": 57510 + }, + { + "epoch": 3.73, + "learning_rate": 7.659524431669714e-05, + "loss": 0.1413, + "step": 57520 + }, + { + "epoch": 3.73, + "learning_rate": 7.65560491246407e-05, + "loss": 0.1468, + "step": 57530 + }, + { + "epoch": 3.73, + "learning_rate": 7.651685393258425e-05, + "loss": 0.1029, + "step": 57540 + }, + { + "epoch": 3.74, + "learning_rate": 7.647765874052782e-05, + "loss": 0.1195, + "step": 57550 + }, + { + "epoch": 3.74, + "learning_rate": 7.643846354847138e-05, + "loss": 0.1354, + "step": 57560 + }, + { + "epoch": 3.74, + "learning_rate": 7.639926835641494e-05, + "loss": 0.1407, + "step": 57570 + }, + { + "epoch": 3.74, + "learning_rate": 7.63600731643585e-05, + "loss": 0.1891, + "step": 57580 + }, + { + "epoch": 3.74, + "learning_rate": 7.632087797230205e-05, + "loss": 0.132, + "step": 57590 + }, + { + "epoch": 3.74, + "learning_rate": 7.628168278024561e-05, + "loss": 0.1358, + "step": 57600 + }, + { + "epoch": 3.74, + "learning_rate": 7.624248758818918e-05, + "loss": 0.137, + "step": 57610 + }, + { + "epoch": 3.74, + "learning_rate": 7.620329239613274e-05, + "loss": 0.1492, + "step": 57620 + }, + { + "epoch": 3.74, + "learning_rate": 7.61640972040763e-05, + "loss": 0.1774, + "step": 57630 + }, + { + "epoch": 3.74, + "learning_rate": 7.612490201201985e-05, + "loss": 0.1332, + "step": 57640 + }, + { + "epoch": 3.74, + "learning_rate": 7.608570681996341e-05, + "loss": 0.1133, + "step": 57650 + }, + { + "epoch": 3.74, + "learning_rate": 7.604651162790698e-05, + "loss": 0.1277, + "step": 57660 + }, + { + "epoch": 3.74, + "learning_rate": 7.600731643585054e-05, + "loss": 0.1383, + "step": 57670 + }, + { + "epoch": 3.74, + "learning_rate": 7.59681212437941e-05, + "loss": 0.1543, + "step": 57680 + }, + { + "epoch": 3.74, + "learning_rate": 7.592892605173765e-05, + "loss": 0.1141, + "step": 57690 + }, + { + "epoch": 3.74, + "learning_rate": 7.588973085968121e-05, + "loss": 0.1044, + "step": 57700 + }, + { + "epoch": 3.75, + "learning_rate": 7.585053566762476e-05, + "loss": 0.1533, + "step": 57710 + }, + { + "epoch": 3.75, + "learning_rate": 7.581134047556832e-05, + "loss": 0.1776, + "step": 57720 + }, + { + "epoch": 3.75, + "learning_rate": 7.577214528351187e-05, + "loss": 0.1276, + "step": 57730 + }, + { + "epoch": 3.75, + "learning_rate": 7.573295009145543e-05, + "loss": 0.1026, + "step": 57740 + }, + { + "epoch": 3.75, + "learning_rate": 7.5693754899399e-05, + "loss": 0.1317, + "step": 57750 + }, + { + "epoch": 3.75, + "learning_rate": 7.565455970734256e-05, + "loss": 0.121, + "step": 57760 + }, + { + "epoch": 3.75, + "learning_rate": 7.561536451528612e-05, + "loss": 0.1583, + "step": 57770 + }, + { + "epoch": 3.75, + "learning_rate": 7.557616932322967e-05, + "loss": 0.1591, + "step": 57780 + }, + { + "epoch": 3.75, + "learning_rate": 7.553697413117323e-05, + "loss": 0.1065, + "step": 57790 + }, + { + "epoch": 3.75, + "learning_rate": 7.549777893911679e-05, + "loss": 0.1356, + "step": 57800 + }, + { + "epoch": 3.75, + "learning_rate": 7.545858374706036e-05, + "loss": 0.1383, + "step": 57810 + }, + { + "epoch": 3.75, + "learning_rate": 7.541938855500392e-05, + "loss": 0.1826, + "step": 57820 + }, + { + "epoch": 3.75, + "learning_rate": 7.538019336294747e-05, + "loss": 0.1441, + "step": 57830 + }, + { + "epoch": 3.75, + "learning_rate": 7.534099817089103e-05, + "loss": 0.1315, + "step": 57840 + }, + { + "epoch": 3.75, + "learning_rate": 7.530180297883459e-05, + "loss": 0.1231, + "step": 57850 + }, + { + "epoch": 3.76, + "learning_rate": 7.526260778677815e-05, + "loss": 0.1463, + "step": 57860 + }, + { + "epoch": 3.76, + "learning_rate": 7.522341259472172e-05, + "loss": 0.139, + "step": 57870 + }, + { + "epoch": 3.76, + "learning_rate": 7.518421740266526e-05, + "loss": 0.1545, + "step": 57880 + }, + { + "epoch": 3.76, + "learning_rate": 7.514502221060883e-05, + "loss": 0.1226, + "step": 57890 + }, + { + "epoch": 3.76, + "learning_rate": 7.510582701855239e-05, + "loss": 0.1145, + "step": 57900 + }, + { + "epoch": 3.76, + "learning_rate": 7.506663182649595e-05, + "loss": 0.1254, + "step": 57910 + }, + { + "epoch": 3.76, + "learning_rate": 7.502743663443951e-05, + "loss": 0.1486, + "step": 57920 + }, + { + "epoch": 3.76, + "learning_rate": 7.498824144238306e-05, + "loss": 0.171, + "step": 57930 + }, + { + "epoch": 3.76, + "learning_rate": 7.494904625032662e-05, + "loss": 0.1209, + "step": 57940 + }, + { + "epoch": 3.76, + "learning_rate": 7.490985105827017e-05, + "loss": 0.1189, + "step": 57950 + }, + { + "epoch": 3.76, + "learning_rate": 7.487065586621374e-05, + "loss": 0.1542, + "step": 57960 + }, + { + "epoch": 3.76, + "learning_rate": 7.48314606741573e-05, + "loss": 0.1752, + "step": 57970 + }, + { + "epoch": 3.76, + "learning_rate": 7.479226548210086e-05, + "loss": 0.1708, + "step": 57980 + }, + { + "epoch": 3.76, + "learning_rate": 7.475307029004442e-05, + "loss": 0.1109, + "step": 57990 + }, + { + "epoch": 3.76, + "learning_rate": 7.471387509798797e-05, + "loss": 0.1215, + "step": 58000 + }, + { + "epoch": 3.76, + "eval_loss": 0.19144190847873688, + "eval_runtime": 997.6747, + "eval_samples_per_second": 15.624, + "eval_steps_per_second": 1.954, + "eval_wer": 0.1899759775980441, + "step": 58000 + }, + { + "epoch": 3.76, + "learning_rate": 7.467467990593153e-05, + "loss": 0.1095, + "step": 58010 + }, + { + "epoch": 3.77, + "learning_rate": 7.463548471387508e-05, + "loss": 0.1749, + "step": 58020 + }, + { + "epoch": 3.77, + "learning_rate": 7.459628952181864e-05, + "loss": 0.1476, + "step": 58030 + }, + { + "epoch": 3.77, + "learning_rate": 7.45570943297622e-05, + "loss": 0.1229, + "step": 58040 + }, + { + "epoch": 3.77, + "learning_rate": 7.451789913770577e-05, + "loss": 0.142, + "step": 58050 + }, + { + "epoch": 3.77, + "learning_rate": 7.447870394564933e-05, + "loss": 0.1215, + "step": 58060 + }, + { + "epoch": 3.77, + "learning_rate": 7.443950875359288e-05, + "loss": 0.1609, + "step": 58070 + }, + { + "epoch": 3.77, + "learning_rate": 7.440031356153644e-05, + "loss": 0.156, + "step": 58080 + }, + { + "epoch": 3.77, + "learning_rate": 7.436111836948e-05, + "loss": 0.0979, + "step": 58090 + }, + { + "epoch": 3.77, + "learning_rate": 7.432192317742357e-05, + "loss": 0.1283, + "step": 58100 + }, + { + "epoch": 3.77, + "learning_rate": 7.428272798536713e-05, + "loss": 0.1553, + "step": 58110 + }, + { + "epoch": 3.77, + "learning_rate": 7.424353279331068e-05, + "loss": 0.148, + "step": 58120 + }, + { + "epoch": 3.77, + "learning_rate": 7.420433760125424e-05, + "loss": 0.1562, + "step": 58130 + }, + { + "epoch": 3.77, + "learning_rate": 7.416514240919779e-05, + "loss": 0.1137, + "step": 58140 + }, + { + "epoch": 3.77, + "learning_rate": 7.412594721714135e-05, + "loss": 0.1356, + "step": 58150 + }, + { + "epoch": 3.77, + "learning_rate": 7.408675202508491e-05, + "loss": 0.131, + "step": 58160 + }, + { + "epoch": 3.78, + "learning_rate": 7.404755683302848e-05, + "loss": 0.146, + "step": 58170 + }, + { + "epoch": 3.78, + "learning_rate": 7.400836164097204e-05, + "loss": 0.1621, + "step": 58180 + }, + { + "epoch": 3.78, + "learning_rate": 7.396916644891559e-05, + "loss": 0.1019, + "step": 58190 + }, + { + "epoch": 3.78, + "learning_rate": 7.392997125685915e-05, + "loss": 0.1189, + "step": 58200 + }, + { + "epoch": 3.78, + "learning_rate": 7.389077606480271e-05, + "loss": 0.1194, + "step": 58210 + }, + { + "epoch": 3.78, + "learning_rate": 7.385158087274627e-05, + "loss": 0.1914, + "step": 58220 + }, + { + "epoch": 3.78, + "learning_rate": 7.381238568068984e-05, + "loss": 0.1533, + "step": 58230 + }, + { + "epoch": 3.78, + "learning_rate": 7.377319048863338e-05, + "loss": 0.1074, + "step": 58240 + }, + { + "epoch": 3.78, + "learning_rate": 7.373399529657695e-05, + "loss": 0.133, + "step": 58250 + }, + { + "epoch": 3.78, + "learning_rate": 7.36948001045205e-05, + "loss": 0.1328, + "step": 58260 + }, + { + "epoch": 3.78, + "learning_rate": 7.365560491246406e-05, + "loss": 0.152, + "step": 58270 + }, + { + "epoch": 3.78, + "learning_rate": 7.361640972040762e-05, + "loss": 0.1664, + "step": 58280 + }, + { + "epoch": 3.78, + "learning_rate": 7.357721452835118e-05, + "loss": 0.0948, + "step": 58290 + }, + { + "epoch": 3.78, + "learning_rate": 7.353801933629474e-05, + "loss": 0.1298, + "step": 58300 + }, + { + "epoch": 3.78, + "learning_rate": 7.34988241442383e-05, + "loss": 0.147, + "step": 58310 + }, + { + "epoch": 3.79, + "learning_rate": 7.345962895218186e-05, + "loss": 0.1509, + "step": 58320 + }, + { + "epoch": 3.79, + "learning_rate": 7.342043376012542e-05, + "loss": 0.1516, + "step": 58330 + }, + { + "epoch": 3.79, + "learning_rate": 7.338123856806898e-05, + "loss": 0.1117, + "step": 58340 + }, + { + "epoch": 3.79, + "learning_rate": 7.334204337601254e-05, + "loss": 0.1003, + "step": 58350 + }, + { + "epoch": 3.79, + "learning_rate": 7.330284818395609e-05, + "loss": 0.112, + "step": 58360 + }, + { + "epoch": 3.79, + "learning_rate": 7.326365299189965e-05, + "loss": 0.178, + "step": 58370 + }, + { + "epoch": 3.79, + "learning_rate": 7.322445779984322e-05, + "loss": 0.1443, + "step": 58380 + }, + { + "epoch": 3.79, + "learning_rate": 7.318526260778676e-05, + "loss": 0.1158, + "step": 58390 + }, + { + "epoch": 3.79, + "learning_rate": 7.314606741573033e-05, + "loss": 0.1442, + "step": 58400 + }, + { + "epoch": 3.79, + "learning_rate": 7.310687222367389e-05, + "loss": 0.1262, + "step": 58410 + }, + { + "epoch": 3.79, + "learning_rate": 7.306767703161745e-05, + "loss": 0.1365, + "step": 58420 + }, + { + "epoch": 3.79, + "learning_rate": 7.3028481839561e-05, + "loss": 0.1427, + "step": 58430 + }, + { + "epoch": 3.79, + "learning_rate": 7.298928664750456e-05, + "loss": 0.1213, + "step": 58440 + }, + { + "epoch": 3.79, + "learning_rate": 7.295009145544813e-05, + "loss": 0.0893, + "step": 58450 + }, + { + "epoch": 3.79, + "learning_rate": 7.291089626339169e-05, + "loss": 0.1132, + "step": 58460 + }, + { + "epoch": 3.79, + "learning_rate": 7.287170107133525e-05, + "loss": 0.1637, + "step": 58470 + }, + { + "epoch": 3.8, + "learning_rate": 7.28325058792788e-05, + "loss": 0.1648, + "step": 58480 + }, + { + "epoch": 3.8, + "learning_rate": 7.279331068722236e-05, + "loss": 0.1016, + "step": 58490 + }, + { + "epoch": 3.8, + "learning_rate": 7.275411549516592e-05, + "loss": 0.0923, + "step": 58500 + }, + { + "epoch": 3.8, + "learning_rate": 7.271492030310947e-05, + "loss": 0.1443, + "step": 58510 + }, + { + "epoch": 3.8, + "learning_rate": 7.267572511105303e-05, + "loss": 0.1794, + "step": 58520 + }, + { + "epoch": 3.8, + "learning_rate": 7.26365299189966e-05, + "loss": 0.1717, + "step": 58530 + }, + { + "epoch": 3.8, + "learning_rate": 7.259733472694016e-05, + "loss": 0.1043, + "step": 58540 + }, + { + "epoch": 3.8, + "learning_rate": 7.255813953488371e-05, + "loss": 0.1592, + "step": 58550 + }, + { + "epoch": 3.8, + "learning_rate": 7.251894434282727e-05, + "loss": 0.1109, + "step": 58560 + }, + { + "epoch": 3.8, + "learning_rate": 7.247974915077083e-05, + "loss": 0.1488, + "step": 58570 + }, + { + "epoch": 3.8, + "learning_rate": 7.24405539587144e-05, + "loss": 0.1564, + "step": 58580 + }, + { + "epoch": 3.8, + "learning_rate": 7.240135876665796e-05, + "loss": 0.1112, + "step": 58590 + }, + { + "epoch": 3.8, + "learning_rate": 7.23621635746015e-05, + "loss": 0.1521, + "step": 58600 + }, + { + "epoch": 3.8, + "learning_rate": 7.232296838254507e-05, + "loss": 0.112, + "step": 58610 + }, + { + "epoch": 3.8, + "learning_rate": 7.228377319048863e-05, + "loss": 0.146, + "step": 58620 + }, + { + "epoch": 3.81, + "learning_rate": 7.224457799843218e-05, + "loss": 0.178, + "step": 58630 + }, + { + "epoch": 3.81, + "learning_rate": 7.220538280637574e-05, + "loss": 0.1037, + "step": 58640 + }, + { + "epoch": 3.81, + "learning_rate": 7.21661876143193e-05, + "loss": 0.1026, + "step": 58650 + }, + { + "epoch": 3.81, + "learning_rate": 7.212699242226287e-05, + "loss": 0.134, + "step": 58660 + }, + { + "epoch": 3.81, + "learning_rate": 7.208779723020641e-05, + "loss": 0.1709, + "step": 58670 + }, + { + "epoch": 3.81, + "learning_rate": 7.204860203814998e-05, + "loss": 0.1937, + "step": 58680 + }, + { + "epoch": 3.81, + "learning_rate": 7.200940684609354e-05, + "loss": 0.1181, + "step": 58690 + }, + { + "epoch": 3.81, + "learning_rate": 7.19702116540371e-05, + "loss": 0.0982, + "step": 58700 + }, + { + "epoch": 3.81, + "learning_rate": 7.193101646198066e-05, + "loss": 0.1558, + "step": 58710 + }, + { + "epoch": 3.81, + "learning_rate": 7.189182126992421e-05, + "loss": 0.1461, + "step": 58720 + }, + { + "epoch": 3.81, + "learning_rate": 7.185262607786777e-05, + "loss": 0.1767, + "step": 58730 + }, + { + "epoch": 3.81, + "learning_rate": 7.181343088581134e-05, + "loss": 0.0994, + "step": 58740 + }, + { + "epoch": 3.81, + "learning_rate": 7.177423569375489e-05, + "loss": 0.119, + "step": 58750 + }, + { + "epoch": 3.81, + "learning_rate": 7.173504050169845e-05, + "loss": 0.1346, + "step": 58760 + }, + { + "epoch": 3.81, + "learning_rate": 7.169584530964201e-05, + "loss": 0.1273, + "step": 58770 + }, + { + "epoch": 3.81, + "learning_rate": 7.165665011758557e-05, + "loss": 0.1374, + "step": 58780 + }, + { + "epoch": 3.82, + "learning_rate": 7.161745492552912e-05, + "loss": 0.1091, + "step": 58790 + }, + { + "epoch": 3.82, + "learning_rate": 7.157825973347268e-05, + "loss": 0.1225, + "step": 58800 + }, + { + "epoch": 3.82, + "learning_rate": 7.153906454141625e-05, + "loss": 0.1502, + "step": 58810 + }, + { + "epoch": 3.82, + "learning_rate": 7.149986934935981e-05, + "loss": 0.1527, + "step": 58820 + }, + { + "epoch": 3.82, + "learning_rate": 7.146067415730337e-05, + "loss": 0.1532, + "step": 58830 + }, + { + "epoch": 3.82, + "learning_rate": 7.142147896524692e-05, + "loss": 0.1222, + "step": 58840 + }, + { + "epoch": 3.82, + "learning_rate": 7.138228377319048e-05, + "loss": 0.1116, + "step": 58850 + }, + { + "epoch": 3.82, + "learning_rate": 7.134308858113404e-05, + "loss": 0.13, + "step": 58860 + }, + { + "epoch": 3.82, + "learning_rate": 7.13038933890776e-05, + "loss": 0.1476, + "step": 58870 + }, + { + "epoch": 3.82, + "learning_rate": 7.126469819702115e-05, + "loss": 0.1463, + "step": 58880 + }, + { + "epoch": 3.82, + "learning_rate": 7.122550300496472e-05, + "loss": 0.11, + "step": 58890 + }, + { + "epoch": 3.82, + "learning_rate": 7.118630781290828e-05, + "loss": 0.1086, + "step": 58900 + }, + { + "epoch": 3.82, + "learning_rate": 7.114711262085183e-05, + "loss": 0.1095, + "step": 58910 + }, + { + "epoch": 3.82, + "learning_rate": 7.110791742879539e-05, + "loss": 0.1492, + "step": 58920 + }, + { + "epoch": 3.82, + "learning_rate": 7.106872223673895e-05, + "loss": 0.1474, + "step": 58930 + }, + { + "epoch": 3.83, + "learning_rate": 7.102952704468251e-05, + "loss": 0.1116, + "step": 58940 + }, + { + "epoch": 3.83, + "learning_rate": 7.099033185262608e-05, + "loss": 0.1084, + "step": 58950 + }, + { + "epoch": 3.83, + "learning_rate": 7.095113666056963e-05, + "loss": 0.1297, + "step": 58960 + }, + { + "epoch": 3.83, + "learning_rate": 7.091194146851319e-05, + "loss": 0.1652, + "step": 58970 + }, + { + "epoch": 3.83, + "learning_rate": 7.087274627645675e-05, + "loss": 0.1791, + "step": 58980 + }, + { + "epoch": 3.83, + "learning_rate": 7.083355108440031e-05, + "loss": 0.0867, + "step": 58990 + }, + { + "epoch": 3.83, + "learning_rate": 7.079435589234386e-05, + "loss": 0.1095, + "step": 59000 + }, + { + "epoch": 3.83, + "eval_loss": 0.19103233516216278, + "eval_runtime": 997.2459, + "eval_samples_per_second": 15.631, + "eval_steps_per_second": 1.954, + "eval_wer": 0.1871117681340706, + "step": 59000 + }, + { + "epoch": 3.83, + "learning_rate": 7.075516070028742e-05, + "loss": 0.1472, + "step": 59010 + }, + { + "epoch": 3.83, + "learning_rate": 7.071596550823099e-05, + "loss": 0.1571, + "step": 59020 + }, + { + "epoch": 3.83, + "learning_rate": 7.067677031617453e-05, + "loss": 0.144, + "step": 59030 + }, + { + "epoch": 3.83, + "learning_rate": 7.06375751241181e-05, + "loss": 0.1185, + "step": 59040 + }, + { + "epoch": 3.83, + "learning_rate": 7.059837993206166e-05, + "loss": 0.1122, + "step": 59050 + }, + { + "epoch": 3.83, + "learning_rate": 7.055918474000522e-05, + "loss": 0.1376, + "step": 59060 + }, + { + "epoch": 3.83, + "learning_rate": 7.051998954794878e-05, + "loss": 0.158, + "step": 59070 + }, + { + "epoch": 3.83, + "learning_rate": 7.048079435589235e-05, + "loss": 0.1195, + "step": 59080 + }, + { + "epoch": 3.84, + "learning_rate": 7.04415991638359e-05, + "loss": 0.1263, + "step": 59090 + }, + { + "epoch": 3.84, + "learning_rate": 7.040240397177946e-05, + "loss": 0.1061, + "step": 59100 + }, + { + "epoch": 3.84, + "learning_rate": 7.036320877972302e-05, + "loss": 0.148, + "step": 59110 + }, + { + "epoch": 3.84, + "learning_rate": 7.032401358766657e-05, + "loss": 0.1596, + "step": 59120 + }, + { + "epoch": 3.84, + "learning_rate": 7.028481839561013e-05, + "loss": 0.1491, + "step": 59130 + }, + { + "epoch": 3.84, + "learning_rate": 7.024562320355369e-05, + "loss": 0.1054, + "step": 59140 + }, + { + "epoch": 3.84, + "learning_rate": 7.020642801149724e-05, + "loss": 0.1152, + "step": 59150 + }, + { + "epoch": 3.84, + "learning_rate": 7.01672328194408e-05, + "loss": 0.1374, + "step": 59160 + }, + { + "epoch": 3.84, + "learning_rate": 7.012803762738437e-05, + "loss": 0.1766, + "step": 59170 + }, + { + "epoch": 3.84, + "learning_rate": 7.009276195453357e-05, + "loss": 0.1655, + "step": 59180 + }, + { + "epoch": 3.84, + "learning_rate": 7.005356676247713e-05, + "loss": 0.1317, + "step": 59190 + }, + { + "epoch": 3.84, + "learning_rate": 7.001437157042068e-05, + "loss": 0.1132, + "step": 59200 + }, + { + "epoch": 3.84, + "learning_rate": 6.997517637836425e-05, + "loss": 0.1437, + "step": 59210 + }, + { + "epoch": 3.84, + "learning_rate": 6.993598118630781e-05, + "loss": 0.1595, + "step": 59220 + }, + { + "epoch": 3.84, + "learning_rate": 6.989678599425137e-05, + "loss": 0.1719, + "step": 59230 + }, + { + "epoch": 3.84, + "learning_rate": 6.985759080219492e-05, + "loss": 0.1368, + "step": 59240 + }, + { + "epoch": 3.85, + "learning_rate": 6.981839561013848e-05, + "loss": 0.1054, + "step": 59250 + }, + { + "epoch": 3.85, + "learning_rate": 6.977920041808204e-05, + "loss": 0.152, + "step": 59260 + }, + { + "epoch": 3.85, + "learning_rate": 6.97400052260256e-05, + "loss": 0.1659, + "step": 59270 + }, + { + "epoch": 3.85, + "learning_rate": 6.970081003396917e-05, + "loss": 0.1952, + "step": 59280 + }, + { + "epoch": 3.85, + "learning_rate": 6.966161484191272e-05, + "loss": 0.1019, + "step": 59290 + }, + { + "epoch": 3.85, + "learning_rate": 6.962241964985628e-05, + "loss": 0.1146, + "step": 59300 + }, + { + "epoch": 3.85, + "learning_rate": 6.958322445779984e-05, + "loss": 0.1131, + "step": 59310 + }, + { + "epoch": 3.85, + "learning_rate": 6.954402926574339e-05, + "loss": 0.1563, + "step": 59320 + }, + { + "epoch": 3.85, + "learning_rate": 6.950483407368695e-05, + "loss": 0.1593, + "step": 59330 + }, + { + "epoch": 3.85, + "learning_rate": 6.946563888163051e-05, + "loss": 0.0949, + "step": 59340 + }, + { + "epoch": 3.85, + "learning_rate": 6.942644368957408e-05, + "loss": 0.1068, + "step": 59350 + }, + { + "epoch": 3.85, + "learning_rate": 6.938724849751763e-05, + "loss": 0.1173, + "step": 59360 + }, + { + "epoch": 3.85, + "learning_rate": 6.934805330546119e-05, + "loss": 0.1413, + "step": 59370 + }, + { + "epoch": 3.85, + "learning_rate": 6.930885811340475e-05, + "loss": 0.1627, + "step": 59380 + }, + { + "epoch": 3.85, + "learning_rate": 6.926966292134831e-05, + "loss": 0.1139, + "step": 59390 + }, + { + "epoch": 3.86, + "learning_rate": 6.923046772929187e-05, + "loss": 0.1337, + "step": 59400 + }, + { + "epoch": 3.86, + "learning_rate": 6.919127253723542e-05, + "loss": 0.1247, + "step": 59410 + }, + { + "epoch": 3.86, + "learning_rate": 6.915207734517899e-05, + "loss": 0.1812, + "step": 59420 + }, + { + "epoch": 3.86, + "learning_rate": 6.911288215312255e-05, + "loss": 0.1499, + "step": 59430 + }, + { + "epoch": 3.86, + "learning_rate": 6.90736869610661e-05, + "loss": 0.1061, + "step": 59440 + }, + { + "epoch": 3.86, + "learning_rate": 6.903449176900966e-05, + "loss": 0.1341, + "step": 59450 + }, + { + "epoch": 3.86, + "learning_rate": 6.899529657695322e-05, + "loss": 0.1178, + "step": 59460 + }, + { + "epoch": 3.86, + "learning_rate": 6.895610138489678e-05, + "loss": 0.1384, + "step": 59470 + }, + { + "epoch": 3.86, + "learning_rate": 6.891690619284033e-05, + "loss": 0.1692, + "step": 59480 + }, + { + "epoch": 3.86, + "learning_rate": 6.88777110007839e-05, + "loss": 0.1222, + "step": 59490 + }, + { + "epoch": 3.86, + "learning_rate": 6.883851580872746e-05, + "loss": 0.1226, + "step": 59500 + }, + { + "epoch": 3.86, + "learning_rate": 6.879932061667102e-05, + "loss": 0.1459, + "step": 59510 + }, + { + "epoch": 3.86, + "learning_rate": 6.876012542461458e-05, + "loss": 0.166, + "step": 59520 + }, + { + "epoch": 3.86, + "learning_rate": 6.872093023255813e-05, + "loss": 0.1514, + "step": 59530 + }, + { + "epoch": 3.86, + "learning_rate": 6.868173504050169e-05, + "loss": 0.1164, + "step": 59540 + }, + { + "epoch": 3.86, + "learning_rate": 6.864253984844525e-05, + "loss": 0.1079, + "step": 59550 + }, + { + "epoch": 3.87, + "learning_rate": 6.86033446563888e-05, + "loss": 0.1359, + "step": 59560 + }, + { + "epoch": 3.87, + "learning_rate": 6.856414946433237e-05, + "loss": 0.1552, + "step": 59570 + }, + { + "epoch": 3.87, + "learning_rate": 6.852495427227593e-05, + "loss": 0.1507, + "step": 59580 + }, + { + "epoch": 3.87, + "learning_rate": 6.848575908021949e-05, + "loss": 0.1116, + "step": 59590 + }, + { + "epoch": 3.87, + "learning_rate": 6.844656388816304e-05, + "loss": 0.1206, + "step": 59600 + }, + { + "epoch": 3.87, + "learning_rate": 6.84073686961066e-05, + "loss": 0.1103, + "step": 59610 + }, + { + "epoch": 3.87, + "learning_rate": 6.836817350405016e-05, + "loss": 0.1854, + "step": 59620 + }, + { + "epoch": 3.87, + "learning_rate": 6.832897831199373e-05, + "loss": 0.1497, + "step": 59630 + }, + { + "epoch": 3.87, + "learning_rate": 6.828978311993729e-05, + "loss": 0.1176, + "step": 59640 + }, + { + "epoch": 3.87, + "learning_rate": 6.825058792788084e-05, + "loss": 0.1243, + "step": 59650 + }, + { + "epoch": 3.87, + "learning_rate": 6.82113927358244e-05, + "loss": 0.1293, + "step": 59660 + }, + { + "epoch": 3.87, + "learning_rate": 6.817219754376796e-05, + "loss": 0.1565, + "step": 59670 + }, + { + "epoch": 3.87, + "learning_rate": 6.813300235171152e-05, + "loss": 0.1602, + "step": 59680 + }, + { + "epoch": 3.87, + "learning_rate": 6.809380715965507e-05, + "loss": 0.1049, + "step": 59690 + }, + { + "epoch": 3.87, + "learning_rate": 6.805461196759864e-05, + "loss": 0.1265, + "step": 59700 + }, + { + "epoch": 3.88, + "learning_rate": 6.80154167755422e-05, + "loss": 0.1489, + "step": 59710 + }, + { + "epoch": 3.88, + "learning_rate": 6.797622158348575e-05, + "loss": 0.1739, + "step": 59720 + }, + { + "epoch": 3.88, + "learning_rate": 6.793702639142931e-05, + "loss": 0.174, + "step": 59730 + }, + { + "epoch": 3.88, + "learning_rate": 6.789783119937287e-05, + "loss": 0.1207, + "step": 59740 + }, + { + "epoch": 3.88, + "learning_rate": 6.785863600731643e-05, + "loss": 0.137, + "step": 59750 + }, + { + "epoch": 3.88, + "learning_rate": 6.781944081526e-05, + "loss": 0.1367, + "step": 59760 + }, + { + "epoch": 3.88, + "learning_rate": 6.778024562320354e-05, + "loss": 0.1631, + "step": 59770 + }, + { + "epoch": 3.88, + "learning_rate": 6.77410504311471e-05, + "loss": 0.1674, + "step": 59780 + }, + { + "epoch": 3.88, + "learning_rate": 6.770185523909067e-05, + "loss": 0.1137, + "step": 59790 + }, + { + "epoch": 3.88, + "learning_rate": 6.766266004703423e-05, + "loss": 0.12, + "step": 59800 + }, + { + "epoch": 3.88, + "learning_rate": 6.762346485497778e-05, + "loss": 0.1175, + "step": 59810 + }, + { + "epoch": 3.88, + "learning_rate": 6.758426966292134e-05, + "loss": 0.1644, + "step": 59820 + }, + { + "epoch": 3.88, + "learning_rate": 6.75450744708649e-05, + "loss": 0.1672, + "step": 59830 + }, + { + "epoch": 3.88, + "learning_rate": 6.750587927880845e-05, + "loss": 0.1077, + "step": 59840 + }, + { + "epoch": 3.88, + "learning_rate": 6.746668408675202e-05, + "loss": 0.1272, + "step": 59850 + }, + { + "epoch": 3.88, + "learning_rate": 6.742748889469558e-05, + "loss": 0.1165, + "step": 59860 + }, + { + "epoch": 3.89, + "learning_rate": 6.738829370263914e-05, + "loss": 0.1587, + "step": 59870 + }, + { + "epoch": 3.89, + "learning_rate": 6.73490985105827e-05, + "loss": 0.1864, + "step": 59880 + }, + { + "epoch": 3.89, + "learning_rate": 6.730990331852625e-05, + "loss": 0.1019, + "step": 59890 + }, + { + "epoch": 3.89, + "learning_rate": 6.727070812646981e-05, + "loss": 0.1211, + "step": 59900 + }, + { + "epoch": 3.89, + "learning_rate": 6.723151293441338e-05, + "loss": 0.1277, + "step": 59910 + }, + { + "epoch": 3.89, + "learning_rate": 6.719231774235694e-05, + "loss": 0.1652, + "step": 59920 + }, + { + "epoch": 3.89, + "learning_rate": 6.715312255030049e-05, + "loss": 0.1935, + "step": 59930 + }, + { + "epoch": 3.89, + "learning_rate": 6.711392735824405e-05, + "loss": 0.1219, + "step": 59940 + }, + { + "epoch": 3.89, + "learning_rate": 6.707473216618761e-05, + "loss": 0.1102, + "step": 59950 + }, + { + "epoch": 3.89, + "learning_rate": 6.703553697413116e-05, + "loss": 0.1255, + "step": 59960 + }, + { + "epoch": 3.89, + "learning_rate": 6.699634178207472e-05, + "loss": 0.1454, + "step": 59970 + }, + { + "epoch": 3.89, + "learning_rate": 6.695714659001828e-05, + "loss": 0.1419, + "step": 59980 + }, + { + "epoch": 3.89, + "learning_rate": 6.691795139796185e-05, + "loss": 0.1153, + "step": 59990 + }, + { + "epoch": 3.89, + "learning_rate": 6.687875620590541e-05, + "loss": 0.1159, + "step": 60000 + }, + { + "epoch": 3.89, + "eval_loss": 0.18907266855239868, + "eval_runtime": 999.7935, + "eval_samples_per_second": 15.591, + "eval_steps_per_second": 1.949, + "eval_wer": 0.1840485565237168, + "step": 60000 + }, + { + "epoch": 3.89, + "learning_rate": 6.683956101384896e-05, + "loss": 0.1171, + "step": 60010 + }, + { + "epoch": 3.9, + "learning_rate": 6.680036582179252e-05, + "loss": 0.18, + "step": 60020 + }, + { + "epoch": 3.9, + "learning_rate": 6.676117062973608e-05, + "loss": 0.1684, + "step": 60030 + }, + { + "epoch": 3.9, + "learning_rate": 6.672197543767964e-05, + "loss": 0.1057, + "step": 60040 + }, + { + "epoch": 3.9, + "learning_rate": 6.668278024562321e-05, + "loss": 0.1174, + "step": 60050 + }, + { + "epoch": 3.9, + "learning_rate": 6.664358505356676e-05, + "loss": 0.1234, + "step": 60060 + }, + { + "epoch": 3.9, + "learning_rate": 6.660438986151032e-05, + "loss": 0.1608, + "step": 60070 + }, + { + "epoch": 3.9, + "learning_rate": 6.656519466945387e-05, + "loss": 0.2077, + "step": 60080 + }, + { + "epoch": 3.9, + "learning_rate": 6.652599947739743e-05, + "loss": 0.1018, + "step": 60090 + }, + { + "epoch": 3.9, + "learning_rate": 6.648680428534099e-05, + "loss": 0.1052, + "step": 60100 + }, + { + "epoch": 3.9, + "learning_rate": 6.644760909328455e-05, + "loss": 0.1157, + "step": 60110 + }, + { + "epoch": 3.9, + "learning_rate": 6.640841390122812e-05, + "loss": 0.1354, + "step": 60120 + }, + { + "epoch": 3.9, + "learning_rate": 6.636921870917166e-05, + "loss": 0.1498, + "step": 60130 + }, + { + "epoch": 3.9, + "learning_rate": 6.633002351711523e-05, + "loss": 0.1121, + "step": 60140 + }, + { + "epoch": 3.9, + "learning_rate": 6.629082832505879e-05, + "loss": 0.1251, + "step": 60150 + }, + { + "epoch": 3.9, + "learning_rate": 6.625163313300235e-05, + "loss": 0.13, + "step": 60160 + }, + { + "epoch": 3.91, + "learning_rate": 6.621243794094591e-05, + "loss": 0.1527, + "step": 60170 + }, + { + "epoch": 3.91, + "learning_rate": 6.617324274888946e-05, + "loss": 0.1401, + "step": 60180 + }, + { + "epoch": 3.91, + "learning_rate": 6.613404755683302e-05, + "loss": 0.1241, + "step": 60190 + }, + { + "epoch": 3.91, + "learning_rate": 6.609485236477657e-05, + "loss": 0.1576, + "step": 60200 + }, + { + "epoch": 3.91, + "learning_rate": 6.605565717272014e-05, + "loss": 0.1281, + "step": 60210 + }, + { + "epoch": 3.91, + "learning_rate": 6.60164619806637e-05, + "loss": 0.1635, + "step": 60220 + }, + { + "epoch": 3.91, + "learning_rate": 6.597726678860726e-05, + "loss": 0.1265, + "step": 60230 + }, + { + "epoch": 3.91, + "learning_rate": 6.593807159655082e-05, + "loss": 0.1477, + "step": 60240 + }, + { + "epoch": 3.91, + "learning_rate": 6.589887640449437e-05, + "loss": 0.1259, + "step": 60250 + }, + { + "epoch": 3.91, + "learning_rate": 6.585968121243793e-05, + "loss": 0.1332, + "step": 60260 + }, + { + "epoch": 3.91, + "learning_rate": 6.58204860203815e-05, + "loss": 0.1712, + "step": 60270 + }, + { + "epoch": 3.91, + "learning_rate": 6.578129082832506e-05, + "loss": 0.1535, + "step": 60280 + }, + { + "epoch": 3.91, + "learning_rate": 6.574209563626862e-05, + "loss": 0.0976, + "step": 60290 + }, + { + "epoch": 3.91, + "learning_rate": 6.570290044421217e-05, + "loss": 0.1157, + "step": 60300 + }, + { + "epoch": 3.91, + "learning_rate": 6.566370525215573e-05, + "loss": 0.1318, + "step": 60310 + }, + { + "epoch": 3.91, + "learning_rate": 6.562451006009928e-05, + "loss": 0.1578, + "step": 60320 + }, + { + "epoch": 3.92, + "learning_rate": 6.558531486804284e-05, + "loss": 0.1283, + "step": 60330 + }, + { + "epoch": 3.92, + "learning_rate": 6.55461196759864e-05, + "loss": 0.1112, + "step": 60340 + }, + { + "epoch": 3.92, + "learning_rate": 6.550692448392997e-05, + "loss": 0.115, + "step": 60350 + }, + { + "epoch": 3.92, + "learning_rate": 6.546772929187353e-05, + "loss": 0.1249, + "step": 60360 + }, + { + "epoch": 3.92, + "learning_rate": 6.542853409981708e-05, + "loss": 0.1403, + "step": 60370 + }, + { + "epoch": 3.92, + "learning_rate": 6.538933890776064e-05, + "loss": 0.158, + "step": 60380 + }, + { + "epoch": 3.92, + "learning_rate": 6.53501437157042e-05, + "loss": 0.1077, + "step": 60390 + }, + { + "epoch": 3.92, + "learning_rate": 6.531094852364777e-05, + "loss": 0.1205, + "step": 60400 + }, + { + "epoch": 3.92, + "learning_rate": 6.527175333159133e-05, + "loss": 0.1364, + "step": 60410 + }, + { + "epoch": 3.92, + "learning_rate": 6.523255813953488e-05, + "loss": 0.1371, + "step": 60420 + }, + { + "epoch": 3.92, + "learning_rate": 6.519336294747844e-05, + "loss": 0.1659, + "step": 60430 + }, + { + "epoch": 3.92, + "learning_rate": 6.515416775542199e-05, + "loss": 0.1092, + "step": 60440 + }, + { + "epoch": 3.92, + "learning_rate": 6.511497256336555e-05, + "loss": 0.1121, + "step": 60450 + }, + { + "epoch": 3.92, + "learning_rate": 6.507577737130911e-05, + "loss": 0.1237, + "step": 60460 + }, + { + "epoch": 3.92, + "learning_rate": 6.503658217925267e-05, + "loss": 0.1426, + "step": 60470 + }, + { + "epoch": 3.93, + "learning_rate": 6.499738698719624e-05, + "loss": 0.136, + "step": 60480 + }, + { + "epoch": 3.93, + "learning_rate": 6.495819179513979e-05, + "loss": 0.1138, + "step": 60490 + }, + { + "epoch": 3.93, + "learning_rate": 6.491899660308335e-05, + "loss": 0.1189, + "step": 60500 + }, + { + "epoch": 3.93, + "learning_rate": 6.487980141102691e-05, + "loss": 0.1234, + "step": 60510 + }, + { + "epoch": 3.93, + "learning_rate": 6.484060621897047e-05, + "loss": 0.1068, + "step": 60520 + }, + { + "epoch": 3.93, + "learning_rate": 6.480141102691403e-05, + "loss": 0.1365, + "step": 60530 + }, + { + "epoch": 3.93, + "learning_rate": 6.476221583485758e-05, + "loss": 0.0968, + "step": 60540 + }, + { + "epoch": 3.93, + "learning_rate": 6.472302064280115e-05, + "loss": 0.1212, + "step": 60550 + }, + { + "epoch": 3.93, + "learning_rate": 6.46838254507447e-05, + "loss": 0.1392, + "step": 60560 + }, + { + "epoch": 3.93, + "learning_rate": 6.464463025868826e-05, + "loss": 0.1418, + "step": 60570 + }, + { + "epoch": 3.93, + "learning_rate": 6.460543506663182e-05, + "loss": 0.1678, + "step": 60580 + }, + { + "epoch": 3.93, + "learning_rate": 6.456623987457538e-05, + "loss": 0.0998, + "step": 60590 + }, + { + "epoch": 3.93, + "learning_rate": 6.452704468251894e-05, + "loss": 0.1065, + "step": 60600 + }, + { + "epoch": 3.93, + "learning_rate": 6.448784949046249e-05, + "loss": 0.1178, + "step": 60610 + }, + { + "epoch": 3.93, + "learning_rate": 6.444865429840605e-05, + "loss": 0.143, + "step": 60620 + }, + { + "epoch": 3.93, + "learning_rate": 6.440945910634962e-05, + "loss": 0.1563, + "step": 60630 + }, + { + "epoch": 3.94, + "learning_rate": 6.437026391429318e-05, + "loss": 0.1034, + "step": 60640 + }, + { + "epoch": 3.94, + "learning_rate": 6.433106872223674e-05, + "loss": 0.0969, + "step": 60650 + }, + { + "epoch": 3.94, + "learning_rate": 6.429187353018029e-05, + "loss": 0.1342, + "step": 60660 + }, + { + "epoch": 3.94, + "learning_rate": 6.425267833812385e-05, + "loss": 0.188, + "step": 60670 + }, + { + "epoch": 3.94, + "learning_rate": 6.42134831460674e-05, + "loss": 0.1623, + "step": 60680 + }, + { + "epoch": 3.94, + "learning_rate": 6.417428795401096e-05, + "loss": 0.0992, + "step": 60690 + }, + { + "epoch": 3.94, + "learning_rate": 6.413509276195453e-05, + "loss": 0.1157, + "step": 60700 + }, + { + "epoch": 3.94, + "learning_rate": 6.409589756989809e-05, + "loss": 0.1421, + "step": 60710 + }, + { + "epoch": 3.94, + "learning_rate": 6.405670237784165e-05, + "loss": 0.1482, + "step": 60720 + }, + { + "epoch": 3.94, + "learning_rate": 6.40175071857852e-05, + "loss": 0.1649, + "step": 60730 + }, + { + "epoch": 3.94, + "learning_rate": 6.397831199372876e-05, + "loss": 0.1095, + "step": 60740 + }, + { + "epoch": 3.94, + "learning_rate": 6.393911680167232e-05, + "loss": 0.12, + "step": 60750 + }, + { + "epoch": 3.94, + "learning_rate": 6.389992160961589e-05, + "loss": 0.1156, + "step": 60760 + }, + { + "epoch": 3.94, + "learning_rate": 6.386072641755945e-05, + "loss": 0.1454, + "step": 60770 + }, + { + "epoch": 3.94, + "learning_rate": 6.3821531225503e-05, + "loss": 0.1221, + "step": 60780 + }, + { + "epoch": 3.95, + "learning_rate": 6.378233603344656e-05, + "loss": 0.0953, + "step": 60790 + }, + { + "epoch": 3.95, + "learning_rate": 6.374314084139011e-05, + "loss": 0.125, + "step": 60800 + }, + { + "epoch": 3.95, + "learning_rate": 6.370394564933367e-05, + "loss": 0.127, + "step": 60810 + }, + { + "epoch": 3.95, + "learning_rate": 6.366475045727723e-05, + "loss": 0.1768, + "step": 60820 + }, + { + "epoch": 3.95, + "learning_rate": 6.36255552652208e-05, + "loss": 0.1522, + "step": 60830 + }, + { + "epoch": 3.95, + "learning_rate": 6.358636007316436e-05, + "loss": 0.0978, + "step": 60840 + }, + { + "epoch": 3.95, + "learning_rate": 6.35471648811079e-05, + "loss": 0.1047, + "step": 60850 + }, + { + "epoch": 3.95, + "learning_rate": 6.350796968905147e-05, + "loss": 0.1478, + "step": 60860 + }, + { + "epoch": 3.95, + "learning_rate": 6.346877449699503e-05, + "loss": 0.1523, + "step": 60870 + }, + { + "epoch": 3.95, + "learning_rate": 6.342957930493859e-05, + "loss": 0.1494, + "step": 60880 + }, + { + "epoch": 3.95, + "learning_rate": 6.339038411288215e-05, + "loss": 0.1111, + "step": 60890 + }, + { + "epoch": 3.95, + "learning_rate": 6.33511889208257e-05, + "loss": 0.1064, + "step": 60900 + }, + { + "epoch": 3.95, + "learning_rate": 6.331199372876927e-05, + "loss": 0.1022, + "step": 60910 + }, + { + "epoch": 3.95, + "learning_rate": 6.327279853671281e-05, + "loss": 0.1496, + "step": 60920 + }, + { + "epoch": 3.95, + "learning_rate": 6.323360334465638e-05, + "loss": 0.135, + "step": 60930 + }, + { + "epoch": 3.96, + "learning_rate": 6.319440815259994e-05, + "loss": 0.1109, + "step": 60940 + }, + { + "epoch": 3.96, + "learning_rate": 6.31552129605435e-05, + "loss": 0.1244, + "step": 60950 + }, + { + "epoch": 3.96, + "learning_rate": 6.311601776848706e-05, + "loss": 0.1567, + "step": 60960 + }, + { + "epoch": 3.96, + "learning_rate": 6.307682257643061e-05, + "loss": 0.1452, + "step": 60970 + }, + { + "epoch": 3.96, + "learning_rate": 6.303762738437417e-05, + "loss": 0.1549, + "step": 60980 + }, + { + "epoch": 3.96, + "learning_rate": 6.299843219231774e-05, + "loss": 0.1132, + "step": 60990 + }, + { + "epoch": 3.96, + "learning_rate": 6.29592370002613e-05, + "loss": 0.11, + "step": 61000 + }, + { + "epoch": 3.96, + "eval_loss": 0.18724286556243896, + "eval_runtime": 1003.8924, + "eval_samples_per_second": 15.528, + "eval_steps_per_second": 1.941, + "eval_wer": 0.1807863427669827, + "step": 61000 + }, + { + "epoch": 3.96, + "learning_rate": 6.292004180820486e-05, + "loss": 0.1269, + "step": 61010 + }, + { + "epoch": 3.96, + "learning_rate": 6.288084661614841e-05, + "loss": 0.146, + "step": 61020 + }, + { + "epoch": 3.96, + "learning_rate": 6.284165142409197e-05, + "loss": 0.1621, + "step": 61030 + }, + { + "epoch": 3.96, + "learning_rate": 6.280245623203552e-05, + "loss": 0.1061, + "step": 61040 + }, + { + "epoch": 3.96, + "learning_rate": 6.276326103997908e-05, + "loss": 0.1257, + "step": 61050 + }, + { + "epoch": 3.96, + "learning_rate": 6.272406584792265e-05, + "loss": 0.1219, + "step": 61060 + }, + { + "epoch": 3.96, + "learning_rate": 6.268487065586621e-05, + "loss": 0.1413, + "step": 61070 + }, + { + "epoch": 3.96, + "learning_rate": 6.264567546380977e-05, + "loss": 0.1601, + "step": 61080 + }, + { + "epoch": 3.96, + "learning_rate": 6.260648027175332e-05, + "loss": 0.1362, + "step": 61090 + }, + { + "epoch": 3.97, + "learning_rate": 6.256728507969688e-05, + "loss": 0.1347, + "step": 61100 + }, + { + "epoch": 3.97, + "learning_rate": 6.252808988764044e-05, + "loss": 0.1082, + "step": 61110 + }, + { + "epoch": 3.97, + "learning_rate": 6.2488894695584e-05, + "loss": 0.1735, + "step": 61120 + }, + { + "epoch": 3.97, + "learning_rate": 6.244969950352757e-05, + "loss": 0.1563, + "step": 61130 + }, + { + "epoch": 3.97, + "learning_rate": 6.241050431147113e-05, + "loss": 0.1127, + "step": 61140 + }, + { + "epoch": 3.97, + "learning_rate": 6.237130911941468e-05, + "loss": 0.115, + "step": 61150 + }, + { + "epoch": 3.97, + "learning_rate": 6.233211392735823e-05, + "loss": 0.1252, + "step": 61160 + }, + { + "epoch": 3.97, + "learning_rate": 6.229291873530179e-05, + "loss": 0.1505, + "step": 61170 + }, + { + "epoch": 3.97, + "learning_rate": 6.225372354324535e-05, + "loss": 0.1238, + "step": 61180 + }, + { + "epoch": 3.97, + "learning_rate": 6.221452835118892e-05, + "loss": 0.0957, + "step": 61190 + }, + { + "epoch": 3.97, + "learning_rate": 6.217533315913248e-05, + "loss": 0.1225, + "step": 61200 + }, + { + "epoch": 3.97, + "learning_rate": 6.213613796707604e-05, + "loss": 0.1154, + "step": 61210 + }, + { + "epoch": 3.97, + "learning_rate": 6.209694277501959e-05, + "loss": 0.1274, + "step": 61220 + }, + { + "epoch": 3.97, + "learning_rate": 6.205774758296315e-05, + "loss": 0.1449, + "step": 61230 + }, + { + "epoch": 3.97, + "learning_rate": 6.201855239090671e-05, + "loss": 0.1062, + "step": 61240 + }, + { + "epoch": 3.98, + "learning_rate": 6.197935719885028e-05, + "loss": 0.1059, + "step": 61250 + }, + { + "epoch": 3.98, + "learning_rate": 6.194016200679384e-05, + "loss": 0.1179, + "step": 61260 + }, + { + "epoch": 3.98, + "learning_rate": 6.190096681473739e-05, + "loss": 0.1487, + "step": 61270 + }, + { + "epoch": 3.98, + "learning_rate": 6.186177162268095e-05, + "loss": 0.177, + "step": 61280 + }, + { + "epoch": 3.98, + "learning_rate": 6.18225764306245e-05, + "loss": 0.1104, + "step": 61290 + }, + { + "epoch": 3.98, + "learning_rate": 6.178338123856806e-05, + "loss": 0.1048, + "step": 61300 + }, + { + "epoch": 3.98, + "learning_rate": 6.174418604651162e-05, + "loss": 0.142, + "step": 61310 + }, + { + "epoch": 3.98, + "learning_rate": 6.170499085445518e-05, + "loss": 0.1386, + "step": 61320 + }, + { + "epoch": 3.98, + "learning_rate": 6.166579566239875e-05, + "loss": 0.2123, + "step": 61330 + }, + { + "epoch": 3.98, + "learning_rate": 6.16266004703423e-05, + "loss": 0.1192, + "step": 61340 + }, + { + "epoch": 3.98, + "learning_rate": 6.158740527828586e-05, + "loss": 0.1138, + "step": 61350 + }, + { + "epoch": 3.98, + "learning_rate": 6.154821008622942e-05, + "loss": 0.148, + "step": 61360 + }, + { + "epoch": 3.98, + "learning_rate": 6.150901489417298e-05, + "loss": 0.1359, + "step": 61370 + }, + { + "epoch": 3.98, + "learning_rate": 6.146981970211654e-05, + "loss": 0.1484, + "step": 61380 + }, + { + "epoch": 3.98, + "learning_rate": 6.143062451006009e-05, + "loss": 0.0923, + "step": 61390 + }, + { + "epoch": 3.98, + "learning_rate": 6.139142931800366e-05, + "loss": 0.1031, + "step": 61400 + }, + { + "epoch": 3.99, + "learning_rate": 6.13522341259472e-05, + "loss": 0.1148, + "step": 61410 + }, + { + "epoch": 3.99, + "learning_rate": 6.131303893389077e-05, + "loss": 0.1252, + "step": 61420 + }, + { + "epoch": 3.99, + "learning_rate": 6.127384374183433e-05, + "loss": 0.1317, + "step": 61430 + }, + { + "epoch": 3.99, + "learning_rate": 6.123464854977789e-05, + "loss": 0.0963, + "step": 61440 + }, + { + "epoch": 3.99, + "learning_rate": 6.119545335772145e-05, + "loss": 0.1035, + "step": 61450 + }, + { + "epoch": 3.99, + "learning_rate": 6.1156258165665e-05, + "loss": 0.1332, + "step": 61460 + }, + { + "epoch": 3.99, + "learning_rate": 6.111706297360856e-05, + "loss": 0.1371, + "step": 61470 + }, + { + "epoch": 3.99, + "learning_rate": 6.107786778155213e-05, + "loss": 0.1521, + "step": 61480 + }, + { + "epoch": 3.99, + "learning_rate": 6.103867258949569e-05, + "loss": 0.1034, + "step": 61490 + }, + { + "epoch": 3.99, + "learning_rate": 6.0999477397439244e-05, + "loss": 0.1329, + "step": 61500 + }, + { + "epoch": 3.99, + "learning_rate": 6.096028220538281e-05, + "loss": 0.1365, + "step": 61510 + }, + { + "epoch": 3.99, + "learning_rate": 6.092108701332636e-05, + "loss": 0.117, + "step": 61520 + }, + { + "epoch": 3.99, + "learning_rate": 6.088189182126992e-05, + "loss": 0.1578, + "step": 61530 + }, + { + "epoch": 3.99, + "learning_rate": 6.084269662921347e-05, + "loss": 0.0928, + "step": 61540 + }, + { + "epoch": 3.99, + "learning_rate": 6.0803501437157036e-05, + "loss": 0.1345, + "step": 61550 + }, + { + "epoch": 4.0, + "learning_rate": 6.076430624510059e-05, + "loss": 0.1256, + "step": 61560 + }, + { + "epoch": 4.0, + "learning_rate": 6.072511105304415e-05, + "loss": 0.1417, + "step": 61570 + }, + { + "epoch": 4.0, + "learning_rate": 6.0685915860987716e-05, + "loss": 0.1773, + "step": 61580 + }, + { + "epoch": 4.0, + "learning_rate": 6.064672066893127e-05, + "loss": 0.1068, + "step": 61590 + }, + { + "epoch": 4.0, + "learning_rate": 6.0607525476874833e-05, + "loss": 0.1037, + "step": 61600 + }, + { + "epoch": 4.0, + "learning_rate": 6.056833028481839e-05, + "loss": 0.1142, + "step": 61610 + }, + { + "epoch": 4.0, + "learning_rate": 6.052913509276195e-05, + "loss": 0.1761, + "step": 61620 + }, + { + "epoch": 4.0, + "learning_rate": 6.0489939900705514e-05, + "loss": 0.1368, + "step": 61630 + }, + { + "epoch": 4.0, + "learning_rate": 6.045074470864907e-05, + "loss": 0.12, + "step": 61640 + }, + { + "epoch": 4.0, + "learning_rate": 6.0411549516592625e-05, + "loss": 0.1089, + "step": 61650 + }, + { + "epoch": 4.0, + "learning_rate": 6.037235432453618e-05, + "loss": 0.1024, + "step": 61660 + }, + { + "epoch": 4.0, + "learning_rate": 6.033315913247974e-05, + "loss": 0.1461, + "step": 61670 + }, + { + "epoch": 4.0, + "learning_rate": 6.02939639404233e-05, + "loss": 0.1244, + "step": 61680 + }, + { + "epoch": 4.0, + "learning_rate": 6.025476874836686e-05, + "loss": 0.1138, + "step": 61690 + }, + { + "epoch": 4.0, + "learning_rate": 6.021557355631042e-05, + "loss": 0.1036, + "step": 61700 + }, + { + "epoch": 4.01, + "learning_rate": 6.017637836425398e-05, + "loss": 0.091, + "step": 61710 + }, + { + "epoch": 4.01, + "learning_rate": 6.013718317219754e-05, + "loss": 0.1141, + "step": 61720 + }, + { + "epoch": 4.01, + "learning_rate": 6.010190749934674e-05, + "loss": 0.1288, + "step": 61730 + }, + { + "epoch": 4.01, + "learning_rate": 6.0062712307290295e-05, + "loss": 0.1413, + "step": 61740 + }, + { + "epoch": 4.01, + "learning_rate": 6.002351711523386e-05, + "loss": 0.103, + "step": 61750 + }, + { + "epoch": 4.01, + "learning_rate": 5.998432192317742e-05, + "loss": 0.1274, + "step": 61760 + }, + { + "epoch": 4.01, + "learning_rate": 5.9945126731120975e-05, + "loss": 0.1257, + "step": 61770 + }, + { + "epoch": 4.01, + "learning_rate": 5.990593153906454e-05, + "loss": 0.1393, + "step": 61780 + }, + { + "epoch": 4.01, + "learning_rate": 5.986673634700809e-05, + "loss": 0.1131, + "step": 61790 + }, + { + "epoch": 4.01, + "learning_rate": 5.9827541154951655e-05, + "loss": 0.0971, + "step": 61800 + }, + { + "epoch": 4.01, + "learning_rate": 5.978834596289522e-05, + "loss": 0.1152, + "step": 61810 + }, + { + "epoch": 4.01, + "learning_rate": 5.974915077083877e-05, + "loss": 0.1244, + "step": 61820 + }, + { + "epoch": 4.01, + "learning_rate": 5.9709955578782335e-05, + "loss": 0.1329, + "step": 61830 + }, + { + "epoch": 4.01, + "learning_rate": 5.967076038672589e-05, + "loss": 0.1396, + "step": 61840 + }, + { + "epoch": 4.01, + "learning_rate": 5.9631565194669446e-05, + "loss": 0.1122, + "step": 61850 + }, + { + "epoch": 4.01, + "learning_rate": 5.9592370002613e-05, + "loss": 0.1119, + "step": 61860 + }, + { + "epoch": 4.02, + "learning_rate": 5.9553174810556564e-05, + "loss": 0.1027, + "step": 61870 + }, + { + "epoch": 4.02, + "learning_rate": 5.9513979618500126e-05, + "loss": 0.1282, + "step": 61880 + }, + { + "epoch": 4.02, + "learning_rate": 5.947478442644368e-05, + "loss": 0.1262, + "step": 61890 + }, + { + "epoch": 4.02, + "learning_rate": 5.9435589234387244e-05, + "loss": 0.0779, + "step": 61900 + }, + { + "epoch": 4.02, + "learning_rate": 5.93963940423308e-05, + "loss": 0.1139, + "step": 61910 + }, + { + "epoch": 4.02, + "learning_rate": 5.935719885027436e-05, + "loss": 0.1149, + "step": 61920 + }, + { + "epoch": 4.02, + "learning_rate": 5.9318003658217924e-05, + "loss": 0.1484, + "step": 61930 + }, + { + "epoch": 4.02, + "learning_rate": 5.927880846616148e-05, + "loss": 0.1303, + "step": 61940 + }, + { + "epoch": 4.02, + "learning_rate": 5.923961327410504e-05, + "loss": 0.0982, + "step": 61950 + }, + { + "epoch": 4.02, + "learning_rate": 5.92004180820486e-05, + "loss": 0.0979, + "step": 61960 + }, + { + "epoch": 4.02, + "learning_rate": 5.916122288999215e-05, + "loss": 0.0943, + "step": 61970 + }, + { + "epoch": 4.02, + "learning_rate": 5.912202769793571e-05, + "loss": 0.1331, + "step": 61980 + }, + { + "epoch": 4.02, + "learning_rate": 5.908283250587927e-05, + "loss": 0.1208, + "step": 61990 + }, + { + "epoch": 4.02, + "learning_rate": 5.904363731382283e-05, + "loss": 0.1016, + "step": 62000 + }, + { + "epoch": 4.02, + "eval_loss": 0.18785810470581055, + "eval_runtime": 1000.2047, + "eval_samples_per_second": 15.585, + "eval_steps_per_second": 1.949, + "eval_wer": 0.18081477164503704, + "step": 62000 + }, + { + "epoch": 4.02, + "learning_rate": 5.900444212176639e-05, + "loss": 0.0982, + "step": 62010 + }, + { + "epoch": 4.03, + "learning_rate": 5.896524692970995e-05, + "loss": 0.1137, + "step": 62020 + }, + { + "epoch": 4.03, + "learning_rate": 5.892605173765351e-05, + "loss": 0.1224, + "step": 62030 + }, + { + "epoch": 4.03, + "learning_rate": 5.888685654559707e-05, + "loss": 0.1276, + "step": 62040 + }, + { + "epoch": 4.03, + "learning_rate": 5.884766135354063e-05, + "loss": 0.118, + "step": 62050 + }, + { + "epoch": 4.03, + "learning_rate": 5.880846616148419e-05, + "loss": 0.0987, + "step": 62060 + }, + { + "epoch": 4.03, + "learning_rate": 5.876927096942775e-05, + "loss": 0.1207, + "step": 62070 + }, + { + "epoch": 4.03, + "learning_rate": 5.8730075777371305e-05, + "loss": 0.1582, + "step": 62080 + }, + { + "epoch": 4.03, + "learning_rate": 5.869088058531486e-05, + "loss": 0.1376, + "step": 62090 + }, + { + "epoch": 4.03, + "learning_rate": 5.8651685393258416e-05, + "loss": 0.1019, + "step": 62100 + }, + { + "epoch": 4.03, + "learning_rate": 5.861249020120198e-05, + "loss": 0.1087, + "step": 62110 + }, + { + "epoch": 4.03, + "learning_rate": 5.857329500914554e-05, + "loss": 0.1213, + "step": 62120 + }, + { + "epoch": 4.03, + "learning_rate": 5.8534099817089096e-05, + "loss": 0.1293, + "step": 62130 + }, + { + "epoch": 4.03, + "learning_rate": 5.849490462503266e-05, + "loss": 0.1175, + "step": 62140 + }, + { + "epoch": 4.03, + "learning_rate": 5.8455709432976213e-05, + "loss": 0.0892, + "step": 62150 + }, + { + "epoch": 4.03, + "learning_rate": 5.8416514240919776e-05, + "loss": 0.0899, + "step": 62160 + }, + { + "epoch": 4.03, + "learning_rate": 5.837731904886334e-05, + "loss": 0.1279, + "step": 62170 + }, + { + "epoch": 4.04, + "learning_rate": 5.8338123856806894e-05, + "loss": 0.1323, + "step": 62180 + }, + { + "epoch": 4.04, + "learning_rate": 5.8298928664750456e-05, + "loss": 0.1304, + "step": 62190 + }, + { + "epoch": 4.04, + "learning_rate": 5.825973347269401e-05, + "loss": 0.1067, + "step": 62200 + }, + { + "epoch": 4.04, + "learning_rate": 5.822053828063757e-05, + "loss": 0.1208, + "step": 62210 + }, + { + "epoch": 4.04, + "learning_rate": 5.818134308858112e-05, + "loss": 0.1095, + "step": 62220 + }, + { + "epoch": 4.04, + "learning_rate": 5.8142147896524685e-05, + "loss": 0.1655, + "step": 62230 + }, + { + "epoch": 4.04, + "learning_rate": 5.810295270446825e-05, + "loss": 0.1038, + "step": 62240 + }, + { + "epoch": 4.04, + "learning_rate": 5.80637575124118e-05, + "loss": 0.107, + "step": 62250 + }, + { + "epoch": 4.04, + "learning_rate": 5.8024562320355365e-05, + "loss": 0.1103, + "step": 62260 + }, + { + "epoch": 4.04, + "learning_rate": 5.798536712829892e-05, + "loss": 0.1142, + "step": 62270 + }, + { + "epoch": 4.04, + "learning_rate": 5.794617193624248e-05, + "loss": 0.1367, + "step": 62280 + }, + { + "epoch": 4.04, + "learning_rate": 5.7906976744186045e-05, + "loss": 0.106, + "step": 62290 + }, + { + "epoch": 4.04, + "learning_rate": 5.78677815521296e-05, + "loss": 0.0882, + "step": 62300 + }, + { + "epoch": 4.04, + "learning_rate": 5.782858636007316e-05, + "loss": 0.1348, + "step": 62310 + }, + { + "epoch": 4.04, + "learning_rate": 5.778939116801672e-05, + "loss": 0.1173, + "step": 62320 + }, + { + "epoch": 4.05, + "learning_rate": 5.775019597596028e-05, + "loss": 0.1626, + "step": 62330 + }, + { + "epoch": 4.05, + "learning_rate": 5.771100078390383e-05, + "loss": 0.1173, + "step": 62340 + }, + { + "epoch": 4.05, + "learning_rate": 5.767180559184739e-05, + "loss": 0.0909, + "step": 62350 + }, + { + "epoch": 4.05, + "learning_rate": 5.7632610399790954e-05, + "loss": 0.0999, + "step": 62360 + }, + { + "epoch": 4.05, + "learning_rate": 5.759341520773451e-05, + "loss": 0.1175, + "step": 62370 + }, + { + "epoch": 4.05, + "learning_rate": 5.755422001567807e-05, + "loss": 0.1302, + "step": 62380 + }, + { + "epoch": 4.05, + "learning_rate": 5.751502482362163e-05, + "loss": 0.1397, + "step": 62390 + }, + { + "epoch": 4.05, + "learning_rate": 5.747582963156519e-05, + "loss": 0.1158, + "step": 62400 + }, + { + "epoch": 4.05, + "learning_rate": 5.743663443950875e-05, + "loss": 0.082, + "step": 62410 + }, + { + "epoch": 4.05, + "learning_rate": 5.739743924745231e-05, + "loss": 0.1123, + "step": 62420 + }, + { + "epoch": 4.05, + "learning_rate": 5.735824405539587e-05, + "loss": 0.1695, + "step": 62430 + }, + { + "epoch": 4.05, + "learning_rate": 5.731904886333943e-05, + "loss": 0.1163, + "step": 62440 + }, + { + "epoch": 4.05, + "learning_rate": 5.727985367128299e-05, + "loss": 0.1175, + "step": 62450 + }, + { + "epoch": 4.05, + "learning_rate": 5.7240658479226536e-05, + "loss": 0.1034, + "step": 62460 + }, + { + "epoch": 4.05, + "learning_rate": 5.72014632871701e-05, + "loss": 0.1224, + "step": 62470 + }, + { + "epoch": 4.06, + "learning_rate": 5.716226809511366e-05, + "loss": 0.1482, + "step": 62480 + }, + { + "epoch": 4.06, + "learning_rate": 5.7123072903057216e-05, + "loss": 0.1474, + "step": 62490 + }, + { + "epoch": 4.06, + "learning_rate": 5.708387771100078e-05, + "loss": 0.1041, + "step": 62500 + }, + { + "epoch": 4.06, + "learning_rate": 5.704468251894434e-05, + "loss": 0.1152, + "step": 62510 + }, + { + "epoch": 4.06, + "learning_rate": 5.7005487326887896e-05, + "loss": 0.1196, + "step": 62520 + }, + { + "epoch": 4.06, + "learning_rate": 5.696629213483146e-05, + "loss": 0.1248, + "step": 62530 + }, + { + "epoch": 4.06, + "learning_rate": 5.6927096942775014e-05, + "loss": 0.1302, + "step": 62540 + }, + { + "epoch": 4.06, + "learning_rate": 5.6887901750718576e-05, + "loss": 0.1092, + "step": 62550 + }, + { + "epoch": 4.06, + "learning_rate": 5.684870655866214e-05, + "loss": 0.1135, + "step": 62560 + }, + { + "epoch": 4.06, + "learning_rate": 5.6809511366605694e-05, + "loss": 0.1162, + "step": 62570 + }, + { + "epoch": 4.06, + "learning_rate": 5.677031617454924e-05, + "loss": 0.1585, + "step": 62580 + }, + { + "epoch": 4.06, + "learning_rate": 5.6731120982492805e-05, + "loss": 0.1235, + "step": 62590 + }, + { + "epoch": 4.06, + "learning_rate": 5.669192579043637e-05, + "loss": 0.1066, + "step": 62600 + }, + { + "epoch": 4.06, + "learning_rate": 5.665273059837992e-05, + "loss": 0.1048, + "step": 62610 + }, + { + "epoch": 4.06, + "learning_rate": 5.6613535406323485e-05, + "loss": 0.131, + "step": 62620 + }, + { + "epoch": 4.06, + "learning_rate": 5.657434021426705e-05, + "loss": 0.1411, + "step": 62630 + }, + { + "epoch": 4.07, + "learning_rate": 5.65351450222106e-05, + "loss": 0.0959, + "step": 62640 + }, + { + "epoch": 4.07, + "learning_rate": 5.6495949830154165e-05, + "loss": 0.0981, + "step": 62650 + }, + { + "epoch": 4.07, + "learning_rate": 5.645675463809772e-05, + "loss": 0.1079, + "step": 62660 + }, + { + "epoch": 4.07, + "learning_rate": 5.641755944604128e-05, + "loss": 0.1183, + "step": 62670 + }, + { + "epoch": 4.07, + "learning_rate": 5.6378364253984846e-05, + "loss": 0.1472, + "step": 62680 + }, + { + "epoch": 4.07, + "learning_rate": 5.63391690619284e-05, + "loss": 0.1158, + "step": 62690 + }, + { + "epoch": 4.07, + "learning_rate": 5.629997386987196e-05, + "loss": 0.1039, + "step": 62700 + }, + { + "epoch": 4.07, + "learning_rate": 5.626077867781551e-05, + "loss": 0.1064, + "step": 62710 + }, + { + "epoch": 4.07, + "learning_rate": 5.6221583485759074e-05, + "loss": 0.1144, + "step": 62720 + }, + { + "epoch": 4.07, + "learning_rate": 5.618238829370263e-05, + "loss": 0.1373, + "step": 62730 + }, + { + "epoch": 4.07, + "learning_rate": 5.614319310164619e-05, + "loss": 0.1174, + "step": 62740 + }, + { + "epoch": 4.07, + "learning_rate": 5.6103997909589754e-05, + "loss": 0.0954, + "step": 62750 + }, + { + "epoch": 4.07, + "learning_rate": 5.606480271753331e-05, + "loss": 0.1181, + "step": 62760 + }, + { + "epoch": 4.07, + "learning_rate": 5.602560752547687e-05, + "loss": 0.093, + "step": 62770 + }, + { + "epoch": 4.07, + "learning_rate": 5.598641233342043e-05, + "loss": 0.1894, + "step": 62780 + }, + { + "epoch": 4.08, + "learning_rate": 5.594721714136399e-05, + "loss": 0.1317, + "step": 62790 + }, + { + "epoch": 4.08, + "learning_rate": 5.590802194930755e-05, + "loss": 0.0957, + "step": 62800 + }, + { + "epoch": 4.08, + "learning_rate": 5.586882675725111e-05, + "loss": 0.1041, + "step": 62810 + }, + { + "epoch": 4.08, + "learning_rate": 5.582963156519467e-05, + "loss": 0.1254, + "step": 62820 + }, + { + "epoch": 4.08, + "learning_rate": 5.579043637313822e-05, + "loss": 0.1607, + "step": 62830 + }, + { + "epoch": 4.08, + "learning_rate": 5.575124118108178e-05, + "loss": 0.1354, + "step": 62840 + }, + { + "epoch": 4.08, + "learning_rate": 5.571204598902534e-05, + "loss": 0.0897, + "step": 62850 + }, + { + "epoch": 4.08, + "learning_rate": 5.56728507969689e-05, + "loss": 0.1088, + "step": 62860 + }, + { + "epoch": 4.08, + "learning_rate": 5.563365560491246e-05, + "loss": 0.111, + "step": 62870 + }, + { + "epoch": 4.08, + "learning_rate": 5.559446041285602e-05, + "loss": 0.1536, + "step": 62880 + }, + { + "epoch": 4.08, + "learning_rate": 5.555526522079958e-05, + "loss": 0.1249, + "step": 62890 + }, + { + "epoch": 4.08, + "learning_rate": 5.5516070028743135e-05, + "loss": 0.0936, + "step": 62900 + }, + { + "epoch": 4.08, + "learning_rate": 5.54768748366867e-05, + "loss": 0.0959, + "step": 62910 + }, + { + "epoch": 4.08, + "learning_rate": 5.543767964463026e-05, + "loss": 0.1114, + "step": 62920 + }, + { + "epoch": 4.08, + "learning_rate": 5.5398484452573815e-05, + "loss": 0.1713, + "step": 62930 + }, + { + "epoch": 4.08, + "learning_rate": 5.535928926051738e-05, + "loss": 0.1182, + "step": 62940 + }, + { + "epoch": 4.09, + "learning_rate": 5.5320094068460926e-05, + "loss": 0.0976, + "step": 62950 + }, + { + "epoch": 4.09, + "learning_rate": 5.528089887640449e-05, + "loss": 0.1271, + "step": 62960 + }, + { + "epoch": 4.09, + "learning_rate": 5.5241703684348044e-05, + "loss": 0.116, + "step": 62970 + }, + { + "epoch": 4.09, + "learning_rate": 5.5202508492291606e-05, + "loss": 0.1384, + "step": 62980 + }, + { + "epoch": 4.09, + "learning_rate": 5.516331330023517e-05, + "loss": 0.117, + "step": 62990 + }, + { + "epoch": 4.09, + "learning_rate": 5.5124118108178724e-05, + "loss": 0.129, + "step": 63000 + }, + { + "epoch": 4.09, + "eval_loss": 0.18426841497421265, + "eval_runtime": 1004.266, + "eval_samples_per_second": 15.522, + "eval_steps_per_second": 1.941, + "eval_wer": 0.1776449517419795, + "step": 63000 + }, + { + "epoch": 4.09, + "learning_rate": 5.5084922916122286e-05, + "loss": 0.0914, + "step": 63010 + }, + { + "epoch": 4.09, + "learning_rate": 5.504572772406584e-05, + "loss": 0.109, + "step": 63020 + }, + { + "epoch": 4.09, + "learning_rate": 5.5006532532009404e-05, + "loss": 0.1309, + "step": 63030 + }, + { + "epoch": 4.09, + "learning_rate": 5.4967337339952966e-05, + "loss": 0.1182, + "step": 63040 + }, + { + "epoch": 4.09, + "learning_rate": 5.492814214789652e-05, + "loss": 0.0998, + "step": 63050 + }, + { + "epoch": 4.09, + "learning_rate": 5.4888946955840084e-05, + "loss": 0.0983, + "step": 63060 + }, + { + "epoch": 4.09, + "learning_rate": 5.484975176378363e-05, + "loss": 0.1246, + "step": 63070 + }, + { + "epoch": 4.09, + "learning_rate": 5.4810556571727195e-05, + "loss": 0.1272, + "step": 63080 + }, + { + "epoch": 4.09, + "learning_rate": 5.477136137967075e-05, + "loss": 0.1187, + "step": 63090 + }, + { + "epoch": 4.1, + "learning_rate": 5.473216618761431e-05, + "loss": 0.1127, + "step": 63100 + }, + { + "epoch": 4.1, + "learning_rate": 5.4692970995557875e-05, + "loss": 0.119, + "step": 63110 + }, + { + "epoch": 4.1, + "learning_rate": 5.465377580350143e-05, + "loss": 0.1123, + "step": 63120 + }, + { + "epoch": 4.1, + "learning_rate": 5.461458061144499e-05, + "loss": 0.1177, + "step": 63130 + }, + { + "epoch": 4.1, + "learning_rate": 5.457538541938855e-05, + "loss": 0.1292, + "step": 63140 + }, + { + "epoch": 4.1, + "learning_rate": 5.453619022733211e-05, + "loss": 0.0998, + "step": 63150 + }, + { + "epoch": 4.1, + "learning_rate": 5.449699503527567e-05, + "loss": 0.098, + "step": 63160 + }, + { + "epoch": 4.1, + "learning_rate": 5.445779984321923e-05, + "loss": 0.1278, + "step": 63170 + }, + { + "epoch": 4.1, + "learning_rate": 5.441860465116279e-05, + "loss": 0.1263, + "step": 63180 + }, + { + "epoch": 4.1, + "learning_rate": 5.4379409459106346e-05, + "loss": 0.1144, + "step": 63190 + }, + { + "epoch": 4.1, + "learning_rate": 5.43402142670499e-05, + "loss": 0.124, + "step": 63200 + }, + { + "epoch": 4.1, + "learning_rate": 5.430101907499346e-05, + "loss": 0.1137, + "step": 63210 + }, + { + "epoch": 4.1, + "learning_rate": 5.426182388293702e-05, + "loss": 0.1177, + "step": 63220 + }, + { + "epoch": 4.1, + "learning_rate": 5.422262869088058e-05, + "loss": 0.1529, + "step": 63230 + }, + { + "epoch": 4.1, + "learning_rate": 5.418343349882414e-05, + "loss": 0.1337, + "step": 63240 + }, + { + "epoch": 4.11, + "learning_rate": 5.41442383067677e-05, + "loss": 0.0951, + "step": 63250 + }, + { + "epoch": 4.11, + "learning_rate": 5.4105043114711255e-05, + "loss": 0.1086, + "step": 63260 + }, + { + "epoch": 4.11, + "learning_rate": 5.406584792265482e-05, + "loss": 0.1234, + "step": 63270 + }, + { + "epoch": 4.11, + "learning_rate": 5.402665273059838e-05, + "loss": 0.1437, + "step": 63280 + }, + { + "epoch": 4.11, + "learning_rate": 5.3987457538541935e-05, + "loss": 0.1566, + "step": 63290 + }, + { + "epoch": 4.11, + "learning_rate": 5.39482623464855e-05, + "loss": 0.092, + "step": 63300 + }, + { + "epoch": 4.11, + "learning_rate": 5.390906715442905e-05, + "loss": 0.0851, + "step": 63310 + }, + { + "epoch": 4.11, + "learning_rate": 5.386987196237261e-05, + "loss": 0.1246, + "step": 63320 + }, + { + "epoch": 4.11, + "learning_rate": 5.3830676770316164e-05, + "loss": 0.1653, + "step": 63330 + }, + { + "epoch": 4.11, + "learning_rate": 5.3791481578259726e-05, + "loss": 0.1379, + "step": 63340 + }, + { + "epoch": 4.11, + "learning_rate": 5.375228638620329e-05, + "loss": 0.108, + "step": 63350 + }, + { + "epoch": 4.11, + "learning_rate": 5.3713091194146844e-05, + "loss": 0.109, + "step": 63360 + }, + { + "epoch": 4.11, + "learning_rate": 5.3673896002090406e-05, + "loss": 0.1327, + "step": 63370 + }, + { + "epoch": 4.11, + "learning_rate": 5.363470081003396e-05, + "loss": 0.1256, + "step": 63380 + }, + { + "epoch": 4.11, + "learning_rate": 5.3595505617977524e-05, + "loss": 0.1302, + "step": 63390 + }, + { + "epoch": 4.11, + "learning_rate": 5.3556310425921087e-05, + "loss": 0.0948, + "step": 63400 + }, + { + "epoch": 4.12, + "learning_rate": 5.351711523386464e-05, + "loss": 0.1312, + "step": 63410 + }, + { + "epoch": 4.12, + "learning_rate": 5.3477920041808204e-05, + "loss": 0.1044, + "step": 63420 + }, + { + "epoch": 4.12, + "learning_rate": 5.343872484975176e-05, + "loss": 0.1385, + "step": 63430 + }, + { + "epoch": 4.12, + "learning_rate": 5.3399529657695315e-05, + "loss": 0.1315, + "step": 63440 + }, + { + "epoch": 4.12, + "learning_rate": 5.336033446563887e-05, + "loss": 0.1052, + "step": 63450 + }, + { + "epoch": 4.12, + "learning_rate": 5.332113927358243e-05, + "loss": 0.1291, + "step": 63460 + }, + { + "epoch": 4.12, + "learning_rate": 5.3281944081525995e-05, + "loss": 0.1171, + "step": 63470 + }, + { + "epoch": 4.12, + "learning_rate": 5.324274888946955e-05, + "loss": 0.1204, + "step": 63480 + }, + { + "epoch": 4.12, + "learning_rate": 5.320355369741311e-05, + "loss": 0.1439, + "step": 63490 + }, + { + "epoch": 4.12, + "learning_rate": 5.316435850535667e-05, + "loss": 0.1018, + "step": 63500 + }, + { + "epoch": 4.12, + "learning_rate": 5.312516331330023e-05, + "loss": 0.0945, + "step": 63510 + }, + { + "epoch": 4.12, + "learning_rate": 5.308596812124379e-05, + "loss": 0.1417, + "step": 63520 + }, + { + "epoch": 4.12, + "learning_rate": 5.304677292918735e-05, + "loss": 0.1477, + "step": 63530 + }, + { + "epoch": 4.12, + "learning_rate": 5.300757773713091e-05, + "loss": 0.1235, + "step": 63540 + }, + { + "epoch": 4.12, + "learning_rate": 5.296838254507447e-05, + "loss": 0.091, + "step": 63550 + }, + { + "epoch": 4.13, + "learning_rate": 5.292918735301802e-05, + "loss": 0.0977, + "step": 63560 + }, + { + "epoch": 4.13, + "learning_rate": 5.288999216096158e-05, + "loss": 0.1117, + "step": 63570 + }, + { + "epoch": 4.13, + "learning_rate": 5.285079696890514e-05, + "loss": 0.1351, + "step": 63580 + }, + { + "epoch": 4.13, + "learning_rate": 5.28116017768487e-05, + "loss": 0.1455, + "step": 63590 + }, + { + "epoch": 4.13, + "learning_rate": 5.277240658479226e-05, + "loss": 0.0904, + "step": 63600 + }, + { + "epoch": 4.13, + "learning_rate": 5.273321139273582e-05, + "loss": 0.0913, + "step": 63610 + }, + { + "epoch": 4.13, + "learning_rate": 5.2694016200679376e-05, + "loss": 0.1086, + "step": 63620 + }, + { + "epoch": 4.13, + "learning_rate": 5.265482100862294e-05, + "loss": 0.1651, + "step": 63630 + }, + { + "epoch": 4.13, + "learning_rate": 5.26156258165665e-05, + "loss": 0.1493, + "step": 63640 + }, + { + "epoch": 4.13, + "learning_rate": 5.2576430624510056e-05, + "loss": 0.071, + "step": 63650 + }, + { + "epoch": 4.13, + "learning_rate": 5.253723543245362e-05, + "loss": 0.1198, + "step": 63660 + }, + { + "epoch": 4.13, + "learning_rate": 5.2498040240397174e-05, + "loss": 0.1259, + "step": 63670 + }, + { + "epoch": 4.13, + "learning_rate": 5.2458845048340736e-05, + "loss": 0.1372, + "step": 63680 + }, + { + "epoch": 4.13, + "learning_rate": 5.2419649856284285e-05, + "loss": 0.1204, + "step": 63690 + }, + { + "epoch": 4.13, + "learning_rate": 5.238045466422785e-05, + "loss": 0.1014, + "step": 63700 + }, + { + "epoch": 4.13, + "learning_rate": 5.234125947217141e-05, + "loss": 0.1008, + "step": 63710 + }, + { + "epoch": 4.14, + "learning_rate": 5.2302064280114965e-05, + "loss": 0.1459, + "step": 63720 + }, + { + "epoch": 4.14, + "learning_rate": 5.226286908805853e-05, + "loss": 0.1289, + "step": 63730 + }, + { + "epoch": 4.14, + "learning_rate": 5.222367389600208e-05, + "loss": 0.0992, + "step": 63740 + }, + { + "epoch": 4.14, + "learning_rate": 5.2184478703945645e-05, + "loss": 0.0848, + "step": 63750 + }, + { + "epoch": 4.14, + "learning_rate": 5.214528351188921e-05, + "loss": 0.106, + "step": 63760 + }, + { + "epoch": 4.14, + "learning_rate": 5.210608831983276e-05, + "loss": 0.0945, + "step": 63770 + }, + { + "epoch": 4.14, + "learning_rate": 5.2066893127776325e-05, + "loss": 0.1305, + "step": 63780 + }, + { + "epoch": 4.14, + "learning_rate": 5.202769793571988e-05, + "loss": 0.1204, + "step": 63790 + }, + { + "epoch": 4.14, + "learning_rate": 5.198850274366344e-05, + "loss": 0.0858, + "step": 63800 + }, + { + "epoch": 4.14, + "learning_rate": 5.194930755160699e-05, + "loss": 0.108, + "step": 63810 + }, + { + "epoch": 4.14, + "learning_rate": 5.1910112359550554e-05, + "loss": 0.1212, + "step": 63820 + }, + { + "epoch": 4.14, + "learning_rate": 5.1870917167494116e-05, + "loss": 0.1344, + "step": 63830 + }, + { + "epoch": 4.14, + "learning_rate": 5.183172197543767e-05, + "loss": 0.1341, + "step": 63840 + }, + { + "epoch": 4.14, + "learning_rate": 5.1792526783381234e-05, + "loss": 0.1008, + "step": 63850 + }, + { + "epoch": 4.14, + "learning_rate": 5.175333159132479e-05, + "loss": 0.1074, + "step": 63860 + }, + { + "epoch": 4.15, + "learning_rate": 5.171413639926835e-05, + "loss": 0.111, + "step": 63870 + }, + { + "epoch": 4.15, + "learning_rate": 5.1674941207211914e-05, + "loss": 0.104, + "step": 63880 + }, + { + "epoch": 4.15, + "learning_rate": 5.163574601515547e-05, + "loss": 0.1053, + "step": 63890 + }, + { + "epoch": 4.15, + "learning_rate": 5.159655082309903e-05, + "loss": 0.0987, + "step": 63900 + }, + { + "epoch": 4.15, + "learning_rate": 5.1557355631042594e-05, + "loss": 0.1017, + "step": 63910 + }, + { + "epoch": 4.15, + "learning_rate": 5.151816043898615e-05, + "loss": 0.1052, + "step": 63920 + }, + { + "epoch": 4.15, + "learning_rate": 5.14789652469297e-05, + "loss": 0.1567, + "step": 63930 + }, + { + "epoch": 4.15, + "learning_rate": 5.143977005487326e-05, + "loss": 0.1094, + "step": 63940 + }, + { + "epoch": 4.15, + "learning_rate": 5.140057486281682e-05, + "loss": 0.1011, + "step": 63950 + }, + { + "epoch": 4.15, + "learning_rate": 5.136137967076038e-05, + "loss": 0.1034, + "step": 63960 + }, + { + "epoch": 4.15, + "learning_rate": 5.132218447870394e-05, + "loss": 0.0997, + "step": 63970 + }, + { + "epoch": 4.15, + "learning_rate": 5.12829892866475e-05, + "loss": 0.1308, + "step": 63980 + }, + { + "epoch": 4.15, + "learning_rate": 5.124379409459106e-05, + "loss": 0.1204, + "step": 63990 + }, + { + "epoch": 4.15, + "learning_rate": 5.120459890253462e-05, + "loss": 0.0927, + "step": 64000 + }, + { + "epoch": 4.15, + "eval_loss": 0.18440814316272736, + "eval_runtime": 1000.9215, + "eval_samples_per_second": 15.574, + "eval_steps_per_second": 1.947, + "eval_wer": 0.17696265866867564, + "step": 64000 + }, + { + "epoch": 4.15, + "learning_rate": 5.1165403710478176e-05, + "loss": 0.1058, + "step": 64010 + }, + { + "epoch": 4.15, + "learning_rate": 5.112620851842174e-05, + "loss": 0.0942, + "step": 64020 + }, + { + "epoch": 4.16, + "learning_rate": 5.10870133263653e-05, + "loss": 0.1304, + "step": 64030 + }, + { + "epoch": 4.16, + "learning_rate": 5.1047818134308856e-05, + "loss": 0.1288, + "step": 64040 + }, + { + "epoch": 4.16, + "learning_rate": 5.1008622942252405e-05, + "loss": 0.1028, + "step": 64050 + }, + { + "epoch": 4.16, + "learning_rate": 5.096942775019597e-05, + "loss": 0.1075, + "step": 64060 + }, + { + "epoch": 4.16, + "learning_rate": 5.093023255813953e-05, + "loss": 0.1103, + "step": 64070 + }, + { + "epoch": 4.16, + "learning_rate": 5.0891037366083085e-05, + "loss": 0.1682, + "step": 64080 + }, + { + "epoch": 4.16, + "learning_rate": 5.085184217402665e-05, + "loss": 0.0927, + "step": 64090 + }, + { + "epoch": 4.16, + "learning_rate": 5.081264698197021e-05, + "loss": 0.1083, + "step": 64100 + }, + { + "epoch": 4.16, + "learning_rate": 5.0773451789913765e-05, + "loss": 0.1207, + "step": 64110 + }, + { + "epoch": 4.16, + "learning_rate": 5.073425659785733e-05, + "loss": 0.1113, + "step": 64120 + }, + { + "epoch": 4.16, + "learning_rate": 5.069506140580088e-05, + "loss": 0.1255, + "step": 64130 + }, + { + "epoch": 4.16, + "learning_rate": 5.0655866213744445e-05, + "loss": 0.1161, + "step": 64140 + }, + { + "epoch": 4.16, + "learning_rate": 5.061667102168801e-05, + "loss": 0.1115, + "step": 64150 + }, + { + "epoch": 4.16, + "learning_rate": 5.057747582963156e-05, + "loss": 0.1202, + "step": 64160 + }, + { + "epoch": 4.16, + "learning_rate": 5.0538280637575125e-05, + "loss": 0.1114, + "step": 64170 + }, + { + "epoch": 4.17, + "learning_rate": 5.0499085445518674e-05, + "loss": 0.1422, + "step": 64180 + }, + { + "epoch": 4.17, + "learning_rate": 5.0459890253462236e-05, + "loss": 0.1714, + "step": 64190 + }, + { + "epoch": 4.17, + "learning_rate": 5.042069506140579e-05, + "loss": 0.1022, + "step": 64200 + }, + { + "epoch": 4.17, + "learning_rate": 5.0381499869349354e-05, + "loss": 0.0958, + "step": 64210 + }, + { + "epoch": 4.17, + "learning_rate": 5.0342304677292917e-05, + "loss": 0.1, + "step": 64220 + }, + { + "epoch": 4.17, + "learning_rate": 5.030310948523647e-05, + "loss": 0.1466, + "step": 64230 + }, + { + "epoch": 4.17, + "learning_rate": 5.0263914293180034e-05, + "loss": 0.1379, + "step": 64240 + }, + { + "epoch": 4.17, + "learning_rate": 5.022471910112359e-05, + "loss": 0.0934, + "step": 64250 + }, + { + "epoch": 4.17, + "learning_rate": 5.018552390906715e-05, + "loss": 0.1021, + "step": 64260 + }, + { + "epoch": 4.17, + "learning_rate": 5.0146328717010714e-05, + "loss": 0.139, + "step": 64270 + }, + { + "epoch": 4.17, + "learning_rate": 5.010713352495427e-05, + "loss": 0.1386, + "step": 64280 + }, + { + "epoch": 4.17, + "learning_rate": 5.006793833289783e-05, + "loss": 0.1538, + "step": 64290 + }, + { + "epoch": 4.17, + "learning_rate": 5.002874314084138e-05, + "loss": 0.0914, + "step": 64300 + }, + { + "epoch": 4.17, + "learning_rate": 4.998954794878494e-05, + "loss": 0.0951, + "step": 64310 + }, + { + "epoch": 4.17, + "learning_rate": 4.99503527567285e-05, + "loss": 0.091, + "step": 64320 + }, + { + "epoch": 4.18, + "learning_rate": 4.991115756467206e-05, + "loss": 0.1467, + "step": 64330 + }, + { + "epoch": 4.18, + "learning_rate": 4.9871962372615623e-05, + "loss": 0.1167, + "step": 64340 + }, + { + "epoch": 4.18, + "learning_rate": 4.983276718055918e-05, + "loss": 0.1056, + "step": 64350 + }, + { + "epoch": 4.18, + "learning_rate": 4.979357198850274e-05, + "loss": 0.087, + "step": 64360 + }, + { + "epoch": 4.18, + "learning_rate": 4.97543767964463e-05, + "loss": 0.1222, + "step": 64370 + }, + { + "epoch": 4.18, + "learning_rate": 4.971518160438986e-05, + "loss": 0.1439, + "step": 64380 + }, + { + "epoch": 4.18, + "learning_rate": 4.967598641233342e-05, + "loss": 0.1601, + "step": 64390 + }, + { + "epoch": 4.18, + "learning_rate": 4.963679122027698e-05, + "loss": 0.1065, + "step": 64400 + }, + { + "epoch": 4.18, + "learning_rate": 4.959759602822054e-05, + "loss": 0.0986, + "step": 64410 + }, + { + "epoch": 4.18, + "learning_rate": 4.955840083616409e-05, + "loss": 0.1164, + "step": 64420 + }, + { + "epoch": 4.18, + "learning_rate": 4.951920564410765e-05, + "loss": 0.1441, + "step": 64430 + }, + { + "epoch": 4.18, + "learning_rate": 4.9480010452051206e-05, + "loss": 0.1319, + "step": 64440 + }, + { + "epoch": 4.18, + "learning_rate": 4.944081525999477e-05, + "loss": 0.0934, + "step": 64450 + }, + { + "epoch": 4.18, + "learning_rate": 4.940162006793833e-05, + "loss": 0.0837, + "step": 64460 + }, + { + "epoch": 4.18, + "learning_rate": 4.9362424875881886e-05, + "loss": 0.1211, + "step": 64470 + }, + { + "epoch": 4.18, + "learning_rate": 4.932322968382545e-05, + "loss": 0.1086, + "step": 64480 + }, + { + "epoch": 4.19, + "learning_rate": 4.9284034491769004e-05, + "loss": 0.1276, + "step": 64490 + }, + { + "epoch": 4.19, + "learning_rate": 4.9244839299712566e-05, + "loss": 0.0986, + "step": 64500 + }, + { + "epoch": 4.19, + "learning_rate": 4.920564410765613e-05, + "loss": 0.1212, + "step": 64510 + }, + { + "epoch": 4.19, + "learning_rate": 4.9166448915599684e-05, + "loss": 0.1243, + "step": 64520 + }, + { + "epoch": 4.19, + "learning_rate": 4.9127253723543246e-05, + "loss": 0.1942, + "step": 64530 + }, + { + "epoch": 4.19, + "learning_rate": 4.90880585314868e-05, + "loss": 0.1134, + "step": 64540 + }, + { + "epoch": 4.19, + "learning_rate": 4.904886333943036e-05, + "loss": 0.1009, + "step": 64550 + }, + { + "epoch": 4.19, + "learning_rate": 4.900966814737391e-05, + "loss": 0.1045, + "step": 64560 + }, + { + "epoch": 4.19, + "learning_rate": 4.8970472955317475e-05, + "loss": 0.1213, + "step": 64570 + }, + { + "epoch": 4.19, + "learning_rate": 4.893127776326104e-05, + "loss": 0.1353, + "step": 64580 + }, + { + "epoch": 4.19, + "learning_rate": 4.889208257120459e-05, + "loss": 0.1391, + "step": 64590 + }, + { + "epoch": 4.19, + "learning_rate": 4.8852887379148155e-05, + "loss": 0.0967, + "step": 64600 + }, + { + "epoch": 4.19, + "learning_rate": 4.881369218709171e-05, + "loss": 0.1045, + "step": 64610 + }, + { + "epoch": 4.19, + "learning_rate": 4.877449699503527e-05, + "loss": 0.1287, + "step": 64620 + }, + { + "epoch": 4.19, + "learning_rate": 4.8735301802978835e-05, + "loss": 0.1421, + "step": 64630 + }, + { + "epoch": 4.2, + "learning_rate": 4.869610661092239e-05, + "loss": 0.1439, + "step": 64640 + }, + { + "epoch": 4.2, + "learning_rate": 4.865691141886595e-05, + "loss": 0.0977, + "step": 64650 + }, + { + "epoch": 4.2, + "learning_rate": 4.861771622680951e-05, + "loss": 0.1005, + "step": 64660 + }, + { + "epoch": 4.2, + "learning_rate": 4.8578521034753064e-05, + "loss": 0.1224, + "step": 64670 + }, + { + "epoch": 4.2, + "learning_rate": 4.853932584269662e-05, + "loss": 0.167, + "step": 64680 + }, + { + "epoch": 4.2, + "learning_rate": 4.850013065064018e-05, + "loss": 0.1139, + "step": 64690 + }, + { + "epoch": 4.2, + "learning_rate": 4.8460935458583744e-05, + "loss": 0.0984, + "step": 64700 + }, + { + "epoch": 4.2, + "learning_rate": 4.84217402665273e-05, + "loss": 0.0988, + "step": 64710 + }, + { + "epoch": 4.2, + "learning_rate": 4.838254507447086e-05, + "loss": 0.1072, + "step": 64720 + }, + { + "epoch": 4.2, + "learning_rate": 4.834334988241442e-05, + "loss": 0.1449, + "step": 64730 + }, + { + "epoch": 4.2, + "learning_rate": 4.830415469035798e-05, + "loss": 0.1132, + "step": 64740 + }, + { + "epoch": 4.2, + "learning_rate": 4.826495949830154e-05, + "loss": 0.0907, + "step": 64750 + }, + { + "epoch": 4.2, + "learning_rate": 4.82257643062451e-05, + "loss": 0.1128, + "step": 64760 + }, + { + "epoch": 4.2, + "learning_rate": 4.818656911418866e-05, + "loss": 0.1092, + "step": 64770 + }, + { + "epoch": 4.2, + "learning_rate": 4.8147373922132215e-05, + "loss": 0.1345, + "step": 64780 + }, + { + "epoch": 4.2, + "learning_rate": 4.810817873007577e-05, + "loss": 0.1212, + "step": 64790 + }, + { + "epoch": 4.21, + "learning_rate": 4.8068983538019326e-05, + "loss": 0.0807, + "step": 64800 + }, + { + "epoch": 4.21, + "learning_rate": 4.802978834596289e-05, + "loss": 0.1165, + "step": 64810 + }, + { + "epoch": 4.21, + "learning_rate": 4.799059315390645e-05, + "loss": 0.1169, + "step": 64820 + }, + { + "epoch": 4.21, + "learning_rate": 4.7951397961850006e-05, + "loss": 0.1679, + "step": 64830 + }, + { + "epoch": 4.21, + "learning_rate": 4.791220276979357e-05, + "loss": 0.1342, + "step": 64840 + }, + { + "epoch": 4.21, + "learning_rate": 4.7873007577737124e-05, + "loss": 0.0875, + "step": 64850 + }, + { + "epoch": 4.21, + "learning_rate": 4.7833812385680686e-05, + "loss": 0.0923, + "step": 64860 + }, + { + "epoch": 4.21, + "learning_rate": 4.779461719362425e-05, + "loss": 0.1135, + "step": 64870 + }, + { + "epoch": 4.21, + "learning_rate": 4.7755422001567804e-05, + "loss": 0.1192, + "step": 64880 + }, + { + "epoch": 4.21, + "learning_rate": 4.7716226809511366e-05, + "loss": 0.1154, + "step": 64890 + }, + { + "epoch": 4.21, + "learning_rate": 4.767703161745492e-05, + "loss": 0.1195, + "step": 64900 + }, + { + "epoch": 4.21, + "learning_rate": 4.763783642539848e-05, + "loss": 0.1137, + "step": 64910 + }, + { + "epoch": 4.21, + "learning_rate": 4.759864123334203e-05, + "loss": 0.1043, + "step": 64920 + }, + { + "epoch": 4.21, + "learning_rate": 4.7559446041285595e-05, + "loss": 0.1434, + "step": 64930 + }, + { + "epoch": 4.21, + "learning_rate": 4.752025084922916e-05, + "loss": 0.1058, + "step": 64940 + }, + { + "epoch": 4.22, + "learning_rate": 4.748105565717271e-05, + "loss": 0.0841, + "step": 64950 + }, + { + "epoch": 4.22, + "learning_rate": 4.7441860465116275e-05, + "loss": 0.109, + "step": 64960 + }, + { + "epoch": 4.22, + "learning_rate": 4.740266527305983e-05, + "loss": 0.0947, + "step": 64970 + }, + { + "epoch": 4.22, + "learning_rate": 4.736347008100339e-05, + "loss": 0.1248, + "step": 64980 + }, + { + "epoch": 4.22, + "learning_rate": 4.7324274888946955e-05, + "loss": 0.1311, + "step": 64990 + }, + { + "epoch": 4.22, + "learning_rate": 4.728507969689051e-05, + "loss": 0.0903, + "step": 65000 + }, + { + "epoch": 4.22, + "eval_loss": 0.18288911879062653, + "eval_runtime": 1005.6361, + "eval_samples_per_second": 15.501, + "eval_steps_per_second": 1.938, + "eval_wer": 0.17565493027817658, + "step": 65000 + }, + { + "epoch": 4.22, + "learning_rate": 4.724588450483407e-05, + "loss": 0.1046, + "step": 65010 + }, + { + "epoch": 4.22, + "learning_rate": 4.720668931277763e-05, + "loss": 0.1163, + "step": 65020 + }, + { + "epoch": 4.22, + "learning_rate": 4.716749412072119e-05, + "loss": 0.1301, + "step": 65030 + }, + { + "epoch": 4.22, + "learning_rate": 4.712829892866474e-05, + "loss": 0.1199, + "step": 65040 + }, + { + "epoch": 4.22, + "learning_rate": 4.70891037366083e-05, + "loss": 0.0919, + "step": 65050 + }, + { + "epoch": 4.22, + "learning_rate": 4.7049908544551864e-05, + "loss": 0.1005, + "step": 65060 + }, + { + "epoch": 4.22, + "learning_rate": 4.701071335249542e-05, + "loss": 0.0957, + "step": 65070 + }, + { + "epoch": 4.22, + "learning_rate": 4.697151816043898e-05, + "loss": 0.1431, + "step": 65080 + }, + { + "epoch": 4.22, + "learning_rate": 4.693232296838254e-05, + "loss": 0.1278, + "step": 65090 + }, + { + "epoch": 4.23, + "learning_rate": 4.68931277763261e-05, + "loss": 0.0812, + "step": 65100 + }, + { + "epoch": 4.23, + "learning_rate": 4.685393258426966e-05, + "loss": 0.0955, + "step": 65110 + }, + { + "epoch": 4.23, + "learning_rate": 4.681473739221322e-05, + "loss": 0.1108, + "step": 65120 + }, + { + "epoch": 4.23, + "learning_rate": 4.677554220015678e-05, + "loss": 0.1495, + "step": 65130 + }, + { + "epoch": 4.23, + "learning_rate": 4.6736347008100336e-05, + "loss": 0.1254, + "step": 65140 + }, + { + "epoch": 4.23, + "learning_rate": 4.66971518160439e-05, + "loss": 0.1058, + "step": 65150 + }, + { + "epoch": 4.23, + "learning_rate": 4.665795662398745e-05, + "loss": 0.1006, + "step": 65160 + }, + { + "epoch": 4.23, + "learning_rate": 4.661876143193101e-05, + "loss": 0.1239, + "step": 65170 + }, + { + "epoch": 4.23, + "learning_rate": 4.657956623987457e-05, + "loss": 0.1544, + "step": 65180 + }, + { + "epoch": 4.23, + "learning_rate": 4.654037104781813e-05, + "loss": 0.1179, + "step": 65190 + }, + { + "epoch": 4.23, + "learning_rate": 4.650117585576169e-05, + "loss": 0.0998, + "step": 65200 + }, + { + "epoch": 4.23, + "learning_rate": 4.6461980663705245e-05, + "loss": 0.0941, + "step": 65210 + }, + { + "epoch": 4.23, + "learning_rate": 4.642278547164881e-05, + "loss": 0.1168, + "step": 65220 + }, + { + "epoch": 4.23, + "learning_rate": 4.638359027959237e-05, + "loss": 0.165, + "step": 65230 + }, + { + "epoch": 4.23, + "learning_rate": 4.6344395087535925e-05, + "loss": 0.1301, + "step": 65240 + }, + { + "epoch": 4.23, + "learning_rate": 4.630519989547949e-05, + "loss": 0.1093, + "step": 65250 + }, + { + "epoch": 4.24, + "learning_rate": 4.626600470342304e-05, + "loss": 0.1024, + "step": 65260 + }, + { + "epoch": 4.24, + "learning_rate": 4.6226809511366605e-05, + "loss": 0.1032, + "step": 65270 + }, + { + "epoch": 4.24, + "learning_rate": 4.6187614319310154e-05, + "loss": 0.126, + "step": 65280 + }, + { + "epoch": 4.24, + "learning_rate": 4.6148419127253716e-05, + "loss": 0.1254, + "step": 65290 + }, + { + "epoch": 4.24, + "learning_rate": 4.610922393519728e-05, + "loss": 0.0902, + "step": 65300 + }, + { + "epoch": 4.24, + "learning_rate": 4.6070028743140834e-05, + "loss": 0.1079, + "step": 65310 + }, + { + "epoch": 4.24, + "learning_rate": 4.6030833551084396e-05, + "loss": 0.1139, + "step": 65320 + }, + { + "epoch": 4.24, + "learning_rate": 4.599163835902795e-05, + "loss": 0.1553, + "step": 65330 + }, + { + "epoch": 4.24, + "learning_rate": 4.5952443166971514e-05, + "loss": 0.1153, + "step": 65340 + }, + { + "epoch": 4.24, + "learning_rate": 4.5913247974915076e-05, + "loss": 0.0939, + "step": 65350 + }, + { + "epoch": 4.24, + "learning_rate": 4.587405278285863e-05, + "loss": 0.0956, + "step": 65360 + }, + { + "epoch": 4.24, + "learning_rate": 4.5834857590802194e-05, + "loss": 0.113, + "step": 65370 + }, + { + "epoch": 4.24, + "learning_rate": 4.5795662398745756e-05, + "loss": 0.1449, + "step": 65380 + }, + { + "epoch": 4.24, + "learning_rate": 4.575646720668931e-05, + "loss": 0.1417, + "step": 65390 + }, + { + "epoch": 4.24, + "learning_rate": 4.571727201463286e-05, + "loss": 0.0766, + "step": 65400 + }, + { + "epoch": 4.25, + "learning_rate": 4.567807682257642e-05, + "loss": 0.1154, + "step": 65410 + }, + { + "epoch": 4.25, + "learning_rate": 4.5638881630519985e-05, + "loss": 0.1315, + "step": 65420 + }, + { + "epoch": 4.25, + "learning_rate": 4.559968643846354e-05, + "loss": 0.1393, + "step": 65430 + }, + { + "epoch": 4.25, + "learning_rate": 4.55604912464071e-05, + "loss": 0.1133, + "step": 65440 + }, + { + "epoch": 4.25, + "learning_rate": 4.552129605435066e-05, + "loss": 0.1245, + "step": 65450 + }, + { + "epoch": 4.25, + "learning_rate": 4.548210086229422e-05, + "loss": 0.0893, + "step": 65460 + }, + { + "epoch": 4.25, + "learning_rate": 4.544290567023778e-05, + "loss": 0.1036, + "step": 65470 + }, + { + "epoch": 4.25, + "learning_rate": 4.540371047818134e-05, + "loss": 0.1588, + "step": 65480 + }, + { + "epoch": 4.25, + "learning_rate": 4.53645152861249e-05, + "loss": 0.111, + "step": 65490 + }, + { + "epoch": 4.25, + "learning_rate": 4.532532009406846e-05, + "loss": 0.0891, + "step": 65500 + }, + { + "epoch": 4.25, + "learning_rate": 4.528612490201202e-05, + "loss": 0.1011, + "step": 65510 + }, + { + "epoch": 4.25, + "learning_rate": 4.524692970995558e-05, + "loss": 0.1541, + "step": 65520 + }, + { + "epoch": 4.25, + "learning_rate": 4.520773451789913e-05, + "loss": 0.1452, + "step": 65530 + }, + { + "epoch": 4.25, + "learning_rate": 4.516853932584269e-05, + "loss": 0.1182, + "step": 65540 + }, + { + "epoch": 4.25, + "learning_rate": 4.512934413378625e-05, + "loss": 0.0994, + "step": 65550 + }, + { + "epoch": 4.25, + "learning_rate": 4.509014894172981e-05, + "loss": 0.0946, + "step": 65560 + }, + { + "epoch": 4.26, + "learning_rate": 4.505095374967337e-05, + "loss": 0.1113, + "step": 65570 + }, + { + "epoch": 4.26, + "learning_rate": 4.501175855761693e-05, + "loss": 0.1462, + "step": 65580 + }, + { + "epoch": 4.26, + "learning_rate": 4.497256336556049e-05, + "loss": 0.1148, + "step": 65590 + }, + { + "epoch": 4.26, + "learning_rate": 4.4933368173504045e-05, + "loss": 0.0955, + "step": 65600 + }, + { + "epoch": 4.26, + "learning_rate": 4.489417298144761e-05, + "loss": 0.0895, + "step": 65610 + }, + { + "epoch": 4.26, + "learning_rate": 4.485497778939117e-05, + "loss": 0.1025, + "step": 65620 + }, + { + "epoch": 4.26, + "learning_rate": 4.4815782597334725e-05, + "loss": 0.1116, + "step": 65630 + }, + { + "epoch": 4.26, + "learning_rate": 4.477658740527829e-05, + "loss": 0.1108, + "step": 65640 + }, + { + "epoch": 4.26, + "learning_rate": 4.4737392213221836e-05, + "loss": 0.0979, + "step": 65650 + }, + { + "epoch": 4.26, + "learning_rate": 4.46981970211654e-05, + "loss": 0.1128, + "step": 65660 + }, + { + "epoch": 4.26, + "learning_rate": 4.4659001829108954e-05, + "loss": 0.0867, + "step": 65670 + }, + { + "epoch": 4.26, + "learning_rate": 4.4619806637052516e-05, + "loss": 0.1212, + "step": 65680 + }, + { + "epoch": 4.26, + "learning_rate": 4.458061144499608e-05, + "loss": 0.1362, + "step": 65690 + }, + { + "epoch": 4.26, + "learning_rate": 4.4541416252939634e-05, + "loss": 0.1109, + "step": 65700 + }, + { + "epoch": 4.26, + "learning_rate": 4.4502221060883196e-05, + "loss": 0.1065, + "step": 65710 + }, + { + "epoch": 4.27, + "learning_rate": 4.446302586882675e-05, + "loss": 0.1209, + "step": 65720 + }, + { + "epoch": 4.27, + "learning_rate": 4.4423830676770314e-05, + "loss": 0.1532, + "step": 65730 + }, + { + "epoch": 4.27, + "learning_rate": 4.4384635484713877e-05, + "loss": 0.1427, + "step": 65740 + }, + { + "epoch": 4.27, + "learning_rate": 4.434544029265743e-05, + "loss": 0.0782, + "step": 65750 + }, + { + "epoch": 4.27, + "learning_rate": 4.4306245100600994e-05, + "loss": 0.0811, + "step": 65760 + }, + { + "epoch": 4.27, + "learning_rate": 4.426704990854454e-05, + "loss": 0.1126, + "step": 65770 + }, + { + "epoch": 4.27, + "learning_rate": 4.4227854716488105e-05, + "loss": 0.114, + "step": 65780 + }, + { + "epoch": 4.27, + "learning_rate": 4.418865952443166e-05, + "loss": 0.1068, + "step": 65790 + }, + { + "epoch": 4.27, + "learning_rate": 4.414946433237522e-05, + "loss": 0.0986, + "step": 65800 + }, + { + "epoch": 4.27, + "learning_rate": 4.4110269140318786e-05, + "loss": 0.0927, + "step": 65810 + }, + { + "epoch": 4.27, + "learning_rate": 4.407107394826234e-05, + "loss": 0.1267, + "step": 65820 + }, + { + "epoch": 4.27, + "learning_rate": 4.40318787562059e-05, + "loss": 0.1412, + "step": 65830 + }, + { + "epoch": 4.27, + "learning_rate": 4.399268356414946e-05, + "loss": 0.1255, + "step": 65840 + }, + { + "epoch": 4.27, + "learning_rate": 4.395348837209302e-05, + "loss": 0.146, + "step": 65850 + }, + { + "epoch": 4.27, + "learning_rate": 4.3914293180036583e-05, + "loss": 0.1077, + "step": 65860 + }, + { + "epoch": 4.28, + "learning_rate": 4.387509798798014e-05, + "loss": 0.0963, + "step": 65870 + }, + { + "epoch": 4.28, + "learning_rate": 4.38359027959237e-05, + "loss": 0.1141, + "step": 65880 + }, + { + "epoch": 4.28, + "learning_rate": 4.379670760386725e-05, + "loss": 0.1322, + "step": 65890 + }, + { + "epoch": 4.28, + "learning_rate": 4.375751241181081e-05, + "loss": 0.0923, + "step": 65900 + }, + { + "epoch": 4.28, + "learning_rate": 4.371831721975437e-05, + "loss": 0.0908, + "step": 65910 + }, + { + "epoch": 4.28, + "learning_rate": 4.367912202769793e-05, + "loss": 0.1037, + "step": 65920 + }, + { + "epoch": 4.28, + "learning_rate": 4.363992683564149e-05, + "loss": 0.1358, + "step": 65930 + }, + { + "epoch": 4.28, + "learning_rate": 4.360073164358505e-05, + "loss": 0.1093, + "step": 65940 + }, + { + "epoch": 4.28, + "learning_rate": 4.356153645152861e-05, + "loss": 0.0909, + "step": 65950 + }, + { + "epoch": 4.28, + "learning_rate": 4.3522341259472166e-05, + "loss": 0.0898, + "step": 65960 + }, + { + "epoch": 4.28, + "learning_rate": 4.348314606741573e-05, + "loss": 0.0943, + "step": 65970 + }, + { + "epoch": 4.28, + "learning_rate": 4.344395087535929e-05, + "loss": 0.1091, + "step": 65980 + }, + { + "epoch": 4.28, + "learning_rate": 4.3404755683302846e-05, + "loss": 0.0915, + "step": 65990 + }, + { + "epoch": 4.28, + "learning_rate": 4.336556049124641e-05, + "loss": 0.0987, + "step": 66000 + }, + { + "epoch": 4.28, + "eval_loss": 0.17928317189216614, + "eval_runtime": 1006.7744, + "eval_samples_per_second": 15.483, + "eval_steps_per_second": 1.936, + "eval_wer": 0.171873889496951, + "step": 66000 + }, + { + "epoch": 4.28, + "learning_rate": 4.3326365299189964e-05, + "loss": 0.103, + "step": 66010 + }, + { + "epoch": 4.28, + "learning_rate": 4.328717010713352e-05, + "loss": 0.1352, + "step": 66020 + }, + { + "epoch": 4.29, + "learning_rate": 4.3247974915077075e-05, + "loss": 0.1262, + "step": 66030 + }, + { + "epoch": 4.29, + "learning_rate": 4.320877972302064e-05, + "loss": 0.1109, + "step": 66040 + }, + { + "epoch": 4.29, + "learning_rate": 4.31695845309642e-05, + "loss": 0.0988, + "step": 66050 + }, + { + "epoch": 4.29, + "learning_rate": 4.3130389338907755e-05, + "loss": 0.1021, + "step": 66060 + }, + { + "epoch": 4.29, + "learning_rate": 4.309119414685132e-05, + "loss": 0.1211, + "step": 66070 + }, + { + "epoch": 4.29, + "learning_rate": 4.305199895479487e-05, + "loss": 0.1324, + "step": 66080 + }, + { + "epoch": 4.29, + "learning_rate": 4.3012803762738435e-05, + "loss": 0.1244, + "step": 66090 + }, + { + "epoch": 4.29, + "learning_rate": 4.2973608570682e-05, + "loss": 0.1139, + "step": 66100 + }, + { + "epoch": 4.29, + "learning_rate": 4.293441337862555e-05, + "loss": 0.1007, + "step": 66110 + }, + { + "epoch": 4.29, + "learning_rate": 4.2895218186569115e-05, + "loss": 0.0914, + "step": 66120 + }, + { + "epoch": 4.29, + "learning_rate": 4.285602299451267e-05, + "loss": 0.1238, + "step": 66130 + }, + { + "epoch": 4.29, + "learning_rate": 4.2816827802456226e-05, + "loss": 0.1638, + "step": 66140 + }, + { + "epoch": 4.29, + "learning_rate": 4.277763261039978e-05, + "loss": 0.0943, + "step": 66150 + }, + { + "epoch": 4.29, + "learning_rate": 4.2738437418343344e-05, + "loss": 0.091, + "step": 66160 + }, + { + "epoch": 4.29, + "learning_rate": 4.2699242226286906e-05, + "loss": 0.1162, + "step": 66170 + }, + { + "epoch": 4.3, + "learning_rate": 4.266004703423046e-05, + "loss": 0.1449, + "step": 66180 + }, + { + "epoch": 4.3, + "learning_rate": 4.2620851842174024e-05, + "loss": 0.1253, + "step": 66190 + }, + { + "epoch": 4.3, + "learning_rate": 4.258165665011758e-05, + "loss": 0.1102, + "step": 66200 + }, + { + "epoch": 4.3, + "learning_rate": 4.254246145806114e-05, + "loss": 0.0977, + "step": 66210 + }, + { + "epoch": 4.3, + "learning_rate": 4.2503266266004704e-05, + "loss": 0.1218, + "step": 66220 + }, + { + "epoch": 4.3, + "learning_rate": 4.246407107394826e-05, + "loss": 0.117, + "step": 66230 + }, + { + "epoch": 4.3, + "learning_rate": 4.242487588189182e-05, + "loss": 0.1003, + "step": 66240 + }, + { + "epoch": 4.3, + "learning_rate": 4.238568068983538e-05, + "loss": 0.0818, + "step": 66250 + }, + { + "epoch": 4.3, + "learning_rate": 4.234648549777893e-05, + "loss": 0.1285, + "step": 66260 + }, + { + "epoch": 4.3, + "learning_rate": 4.231120982492814e-05, + "loss": 0.1009, + "step": 66270 + }, + { + "epoch": 4.3, + "learning_rate": 4.22720146328717e-05, + "loss": 0.1487, + "step": 66280 + }, + { + "epoch": 4.3, + "learning_rate": 4.223281944081526e-05, + "loss": 0.0927, + "step": 66290 + }, + { + "epoch": 4.3, + "learning_rate": 4.219362424875882e-05, + "loss": 0.0926, + "step": 66300 + }, + { + "epoch": 4.3, + "learning_rate": 4.2154429056702374e-05, + "loss": 0.0956, + "step": 66310 + }, + { + "epoch": 4.3, + "learning_rate": 4.211523386464594e-05, + "loss": 0.109, + "step": 66320 + }, + { + "epoch": 4.3, + "learning_rate": 4.20760386725895e-05, + "loss": 0.1371, + "step": 66330 + }, + { + "epoch": 4.31, + "learning_rate": 4.203684348053305e-05, + "loss": 0.1257, + "step": 66340 + }, + { + "epoch": 4.31, + "learning_rate": 4.199764828847661e-05, + "loss": 0.1042, + "step": 66350 + }, + { + "epoch": 4.31, + "learning_rate": 4.1958453096420166e-05, + "loss": 0.1209, + "step": 66360 + }, + { + "epoch": 4.31, + "learning_rate": 4.191925790436373e-05, + "loss": 0.1187, + "step": 66370 + }, + { + "epoch": 4.31, + "learning_rate": 4.188006271230728e-05, + "loss": 0.1224, + "step": 66380 + }, + { + "epoch": 4.31, + "learning_rate": 4.1840867520250846e-05, + "loss": 0.1325, + "step": 66390 + }, + { + "epoch": 4.31, + "learning_rate": 4.180167232819441e-05, + "loss": 0.0922, + "step": 66400 + }, + { + "epoch": 4.31, + "learning_rate": 4.1762477136137963e-05, + "loss": 0.1007, + "step": 66410 + }, + { + "epoch": 4.31, + "learning_rate": 4.1723281944081526e-05, + "loss": 0.1022, + "step": 66420 + }, + { + "epoch": 4.31, + "learning_rate": 4.168408675202508e-05, + "loss": 0.124, + "step": 66430 + }, + { + "epoch": 4.31, + "learning_rate": 4.1644891559968644e-05, + "loss": 0.161, + "step": 66440 + }, + { + "epoch": 4.31, + "learning_rate": 4.1605696367912206e-05, + "loss": 0.0977, + "step": 66450 + }, + { + "epoch": 4.31, + "learning_rate": 4.1566501175855755e-05, + "loss": 0.0923, + "step": 66460 + }, + { + "epoch": 4.31, + "learning_rate": 4.152730598379932e-05, + "loss": 0.1097, + "step": 66470 + }, + { + "epoch": 4.31, + "learning_rate": 4.148811079174287e-05, + "loss": 0.1379, + "step": 66480 + }, + { + "epoch": 4.32, + "learning_rate": 4.1448915599686435e-05, + "loss": 0.1339, + "step": 66490 + }, + { + "epoch": 4.32, + "learning_rate": 4.140972040762999e-05, + "loss": 0.0942, + "step": 66500 + }, + { + "epoch": 4.32, + "learning_rate": 4.137052521557355e-05, + "loss": 0.1085, + "step": 66510 + }, + { + "epoch": 4.32, + "learning_rate": 4.1331330023517115e-05, + "loss": 0.1041, + "step": 66520 + }, + { + "epoch": 4.32, + "learning_rate": 4.129213483146067e-05, + "loss": 0.1566, + "step": 66530 + }, + { + "epoch": 4.32, + "learning_rate": 4.125293963940423e-05, + "loss": 0.1326, + "step": 66540 + }, + { + "epoch": 4.32, + "learning_rate": 4.121374444734779e-05, + "loss": 0.0951, + "step": 66550 + }, + { + "epoch": 4.32, + "learning_rate": 4.117454925529135e-05, + "loss": 0.094, + "step": 66560 + }, + { + "epoch": 4.32, + "learning_rate": 4.113535406323491e-05, + "loss": 0.0965, + "step": 66570 + }, + { + "epoch": 4.32, + "learning_rate": 4.109615887117846e-05, + "loss": 0.1553, + "step": 66580 + }, + { + "epoch": 4.32, + "learning_rate": 4.1056963679122024e-05, + "loss": 0.1306, + "step": 66590 + }, + { + "epoch": 4.32, + "learning_rate": 4.101776848706558e-05, + "loss": 0.0946, + "step": 66600 + }, + { + "epoch": 4.32, + "learning_rate": 4.097857329500914e-05, + "loss": 0.1088, + "step": 66610 + }, + { + "epoch": 4.32, + "learning_rate": 4.09393781029527e-05, + "loss": 0.1167, + "step": 66620 + }, + { + "epoch": 4.32, + "learning_rate": 4.090018291089626e-05, + "loss": 0.1514, + "step": 66630 + }, + { + "epoch": 4.33, + "learning_rate": 4.086098771883982e-05, + "loss": 0.1196, + "step": 66640 + }, + { + "epoch": 4.33, + "learning_rate": 4.082179252678338e-05, + "loss": 0.091, + "step": 66650 + }, + { + "epoch": 4.33, + "learning_rate": 4.078259733472694e-05, + "loss": 0.105, + "step": 66660 + }, + { + "epoch": 4.33, + "learning_rate": 4.0743402142670495e-05, + "loss": 0.1096, + "step": 66670 + }, + { + "epoch": 4.33, + "learning_rate": 4.070420695061406e-05, + "loss": 0.1229, + "step": 66680 + }, + { + "epoch": 4.33, + "learning_rate": 4.066501175855762e-05, + "loss": 0.1305, + "step": 66690 + }, + { + "epoch": 4.33, + "learning_rate": 4.0625816566501175e-05, + "loss": 0.0885, + "step": 66700 + }, + { + "epoch": 4.33, + "learning_rate": 4.058662137444473e-05, + "loss": 0.1108, + "step": 66710 + }, + { + "epoch": 4.33, + "learning_rate": 4.0547426182388286e-05, + "loss": 0.1183, + "step": 66720 + }, + { + "epoch": 4.33, + "learning_rate": 4.050823099033185e-05, + "loss": 0.1446, + "step": 66730 + }, + { + "epoch": 4.33, + "learning_rate": 4.0469035798275404e-05, + "loss": 0.1106, + "step": 66740 + }, + { + "epoch": 4.33, + "learning_rate": 4.0429840606218966e-05, + "loss": 0.0928, + "step": 66750 + }, + { + "epoch": 4.33, + "learning_rate": 4.039064541416253e-05, + "loss": 0.0971, + "step": 66760 + }, + { + "epoch": 4.33, + "learning_rate": 4.0351450222106084e-05, + "loss": 0.1075, + "step": 66770 + }, + { + "epoch": 4.33, + "learning_rate": 4.0312255030049646e-05, + "loss": 0.1396, + "step": 66780 + }, + { + "epoch": 4.33, + "learning_rate": 4.02730598379932e-05, + "loss": 0.1317, + "step": 66790 + }, + { + "epoch": 4.34, + "learning_rate": 4.0233864645936764e-05, + "loss": 0.089, + "step": 66800 + }, + { + "epoch": 4.34, + "learning_rate": 4.0194669453880326e-05, + "loss": 0.09, + "step": 66810 + }, + { + "epoch": 4.34, + "learning_rate": 4.015547426182388e-05, + "loss": 0.094, + "step": 66820 + }, + { + "epoch": 4.34, + "learning_rate": 4.011627906976744e-05, + "loss": 0.1353, + "step": 66830 + }, + { + "epoch": 4.34, + "learning_rate": 4.007708387771099e-05, + "loss": 0.1129, + "step": 66840 + }, + { + "epoch": 4.34, + "learning_rate": 4.0037888685654555e-05, + "loss": 0.1031, + "step": 66850 + }, + { + "epoch": 4.34, + "learning_rate": 3.999869349359811e-05, + "loss": 0.0861, + "step": 66860 + }, + { + "epoch": 4.34, + "learning_rate": 3.995949830154167e-05, + "loss": 0.1095, + "step": 66870 + }, + { + "epoch": 4.34, + "learning_rate": 3.9920303109485235e-05, + "loss": 0.1509, + "step": 66880 + }, + { + "epoch": 4.34, + "learning_rate": 3.988110791742879e-05, + "loss": 0.1141, + "step": 66890 + }, + { + "epoch": 4.34, + "learning_rate": 3.984191272537235e-05, + "loss": 0.0796, + "step": 66900 + }, + { + "epoch": 4.34, + "learning_rate": 3.980271753331591e-05, + "loss": 0.1118, + "step": 66910 + }, + { + "epoch": 4.34, + "learning_rate": 3.976352234125947e-05, + "loss": 0.0974, + "step": 66920 + }, + { + "epoch": 4.34, + "learning_rate": 3.972432714920303e-05, + "loss": 0.1279, + "step": 66930 + }, + { + "epoch": 4.34, + "learning_rate": 3.968513195714659e-05, + "loss": 0.1056, + "step": 66940 + }, + { + "epoch": 4.35, + "learning_rate": 3.9645936765090144e-05, + "loss": 0.099, + "step": 66950 + }, + { + "epoch": 4.35, + "learning_rate": 3.96067415730337e-05, + "loss": 0.1059, + "step": 66960 + }, + { + "epoch": 4.35, + "learning_rate": 3.956754638097726e-05, + "loss": 0.1098, + "step": 66970 + }, + { + "epoch": 4.35, + "learning_rate": 3.952835118892082e-05, + "loss": 0.113, + "step": 66980 + }, + { + "epoch": 4.35, + "learning_rate": 3.948915599686438e-05, + "loss": 0.1095, + "step": 66990 + }, + { + "epoch": 4.35, + "learning_rate": 3.944996080480794e-05, + "loss": 0.0923, + "step": 67000 + }, + { + "epoch": 4.35, + "eval_loss": 0.1796606183052063, + "eval_runtime": 1002.4557, + "eval_samples_per_second": 15.55, + "eval_steps_per_second": 1.944, + "eval_wer": 0.1699904763258518, + "step": 67000 + }, + { + "epoch": 4.35, + "learning_rate": 3.94107656127515e-05, + "loss": 0.1079, + "step": 67010 + }, + { + "epoch": 4.35, + "learning_rate": 3.937157042069506e-05, + "loss": 0.1201, + "step": 67020 + }, + { + "epoch": 4.35, + "learning_rate": 3.9332375228638615e-05, + "loss": 0.1446, + "step": 67030 + }, + { + "epoch": 4.35, + "learning_rate": 3.929318003658218e-05, + "loss": 0.1414, + "step": 67040 + }, + { + "epoch": 4.35, + "learning_rate": 3.925398484452574e-05, + "loss": 0.0962, + "step": 67050 + }, + { + "epoch": 4.35, + "learning_rate": 3.9214789652469296e-05, + "loss": 0.0928, + "step": 67060 + }, + { + "epoch": 4.35, + "learning_rate": 3.917559446041285e-05, + "loss": 0.1198, + "step": 67070 + }, + { + "epoch": 4.35, + "learning_rate": 3.9136399268356407e-05, + "loss": 0.1401, + "step": 67080 + }, + { + "epoch": 4.35, + "learning_rate": 3.909720407629997e-05, + "loss": 0.1264, + "step": 67090 + }, + { + "epoch": 4.35, + "learning_rate": 3.9058008884243524e-05, + "loss": 0.1024, + "step": 67100 + }, + { + "epoch": 4.36, + "learning_rate": 3.901881369218709e-05, + "loss": 0.1099, + "step": 67110 + }, + { + "epoch": 4.36, + "learning_rate": 3.897961850013065e-05, + "loss": 0.1436, + "step": 67120 + }, + { + "epoch": 4.36, + "learning_rate": 3.8940423308074204e-05, + "loss": 0.1349, + "step": 67130 + }, + { + "epoch": 4.36, + "learning_rate": 3.890122811601777e-05, + "loss": 0.1123, + "step": 67140 + }, + { + "epoch": 4.36, + "learning_rate": 3.886203292396132e-05, + "loss": 0.0789, + "step": 67150 + }, + { + "epoch": 4.36, + "learning_rate": 3.8822837731904885e-05, + "loss": 0.1104, + "step": 67160 + }, + { + "epoch": 4.36, + "learning_rate": 3.878364253984845e-05, + "loss": 0.0883, + "step": 67170 + }, + { + "epoch": 4.36, + "learning_rate": 3.8744447347792e-05, + "loss": 0.1362, + "step": 67180 + }, + { + "epoch": 4.36, + "learning_rate": 3.8705252155735565e-05, + "loss": 0.1342, + "step": 67190 + }, + { + "epoch": 4.36, + "learning_rate": 3.8666056963679113e-05, + "loss": 0.0941, + "step": 67200 + }, + { + "epoch": 4.36, + "learning_rate": 3.8626861771622676e-05, + "loss": 0.1141, + "step": 67210 + }, + { + "epoch": 4.36, + "learning_rate": 3.858766657956623e-05, + "loss": 0.0861, + "step": 67220 + }, + { + "epoch": 4.36, + "learning_rate": 3.8548471387509794e-05, + "loss": 0.1468, + "step": 67230 + }, + { + "epoch": 4.36, + "learning_rate": 3.8509276195453356e-05, + "loss": 0.0989, + "step": 67240 + }, + { + "epoch": 4.36, + "learning_rate": 3.847008100339691e-05, + "loss": 0.1087, + "step": 67250 + }, + { + "epoch": 4.37, + "learning_rate": 3.8430885811340474e-05, + "loss": 0.0975, + "step": 67260 + }, + { + "epoch": 4.37, + "learning_rate": 3.839169061928403e-05, + "loss": 0.1175, + "step": 67270 + }, + { + "epoch": 4.37, + "learning_rate": 3.835249542722759e-05, + "loss": 0.129, + "step": 67280 + }, + { + "epoch": 4.37, + "learning_rate": 3.8313300235171154e-05, + "loss": 0.1254, + "step": 67290 + }, + { + "epoch": 4.37, + "learning_rate": 3.827410504311471e-05, + "loss": 0.1066, + "step": 67300 + }, + { + "epoch": 4.37, + "learning_rate": 3.823490985105827e-05, + "loss": 0.0868, + "step": 67310 + }, + { + "epoch": 4.37, + "learning_rate": 3.819571465900182e-05, + "loss": 0.1146, + "step": 67320 + }, + { + "epoch": 4.37, + "learning_rate": 3.815651946694538e-05, + "loss": 0.1273, + "step": 67330 + }, + { + "epoch": 4.37, + "learning_rate": 3.811732427488894e-05, + "loss": 0.1314, + "step": 67340 + }, + { + "epoch": 4.37, + "learning_rate": 3.80781290828325e-05, + "loss": 0.0989, + "step": 67350 + }, + { + "epoch": 4.37, + "learning_rate": 3.8042853409981706e-05, + "loss": 0.0978, + "step": 67360 + }, + { + "epoch": 4.37, + "learning_rate": 3.800365821792527e-05, + "loss": 0.1032, + "step": 67370 + }, + { + "epoch": 4.37, + "learning_rate": 3.7964463025868824e-05, + "loss": 0.1103, + "step": 67380 + }, + { + "epoch": 4.37, + "learning_rate": 3.792526783381238e-05, + "loss": 0.1228, + "step": 67390 + }, + { + "epoch": 4.37, + "learning_rate": 3.7886072641755935e-05, + "loss": 0.1063, + "step": 67400 + }, + { + "epoch": 4.38, + "learning_rate": 3.78468774496995e-05, + "loss": 0.1011, + "step": 67410 + }, + { + "epoch": 4.38, + "learning_rate": 3.780768225764306e-05, + "loss": 0.1168, + "step": 67420 + }, + { + "epoch": 4.38, + "learning_rate": 3.7768487065586615e-05, + "loss": 0.1497, + "step": 67430 + }, + { + "epoch": 4.38, + "learning_rate": 3.772929187353018e-05, + "loss": 0.1082, + "step": 67440 + }, + { + "epoch": 4.38, + "learning_rate": 3.769009668147373e-05, + "loss": 0.0985, + "step": 67450 + }, + { + "epoch": 4.38, + "learning_rate": 3.7650901489417295e-05, + "loss": 0.0899, + "step": 67460 + }, + { + "epoch": 4.38, + "learning_rate": 3.761170629736086e-05, + "loss": 0.1257, + "step": 67470 + }, + { + "epoch": 4.38, + "learning_rate": 3.757251110530441e-05, + "loss": 0.1138, + "step": 67480 + }, + { + "epoch": 4.38, + "learning_rate": 3.7533315913247976e-05, + "loss": 0.1177, + "step": 67490 + }, + { + "epoch": 4.38, + "learning_rate": 3.749412072119153e-05, + "loss": 0.0825, + "step": 67500 + }, + { + "epoch": 4.38, + "learning_rate": 3.7454925529135087e-05, + "loss": 0.1049, + "step": 67510 + }, + { + "epoch": 4.38, + "learning_rate": 3.741573033707865e-05, + "loss": 0.117, + "step": 67520 + }, + { + "epoch": 4.38, + "learning_rate": 3.737653514502221e-05, + "loss": 0.1081, + "step": 67530 + }, + { + "epoch": 4.38, + "learning_rate": 3.733733995296577e-05, + "loss": 0.1477, + "step": 67540 + }, + { + "epoch": 4.38, + "learning_rate": 3.729814476090932e-05, + "loss": 0.084, + "step": 67550 + }, + { + "epoch": 4.38, + "learning_rate": 3.7258949568852884e-05, + "loss": 0.095, + "step": 67560 + }, + { + "epoch": 4.39, + "learning_rate": 3.721975437679644e-05, + "loss": 0.1023, + "step": 67570 + }, + { + "epoch": 4.39, + "learning_rate": 3.718055918474e-05, + "loss": 0.1497, + "step": 67580 + }, + { + "epoch": 4.39, + "learning_rate": 3.7141363992683565e-05, + "loss": 0.1019, + "step": 67590 + }, + { + "epoch": 4.39, + "learning_rate": 3.710216880062712e-05, + "loss": 0.0841, + "step": 67600 + }, + { + "epoch": 4.39, + "learning_rate": 3.7062973608570676e-05, + "loss": 0.1061, + "step": 67610 + }, + { + "epoch": 4.39, + "learning_rate": 3.702377841651424e-05, + "loss": 0.1108, + "step": 67620 + }, + { + "epoch": 4.39, + "learning_rate": 3.6984583224457793e-05, + "loss": 0.1143, + "step": 67630 + }, + { + "epoch": 4.39, + "learning_rate": 3.6945388032401356e-05, + "loss": 0.1278, + "step": 67640 + }, + { + "epoch": 4.39, + "learning_rate": 3.690619284034492e-05, + "loss": 0.0904, + "step": 67650 + }, + { + "epoch": 4.39, + "learning_rate": 3.6866997648288474e-05, + "loss": 0.1033, + "step": 67660 + }, + { + "epoch": 4.39, + "learning_rate": 3.682780245623203e-05, + "loss": 0.1038, + "step": 67670 + }, + { + "epoch": 4.39, + "learning_rate": 3.678860726417559e-05, + "loss": 0.1209, + "step": 67680 + }, + { + "epoch": 4.39, + "learning_rate": 3.674941207211915e-05, + "loss": 0.1093, + "step": 67690 + }, + { + "epoch": 4.39, + "learning_rate": 3.671021688006271e-05, + "loss": 0.1017, + "step": 67700 + }, + { + "epoch": 4.39, + "learning_rate": 3.667102168800627e-05, + "loss": 0.1203, + "step": 67710 + }, + { + "epoch": 4.4, + "learning_rate": 3.663182649594983e-05, + "loss": 0.1048, + "step": 67720 + }, + { + "epoch": 4.4, + "learning_rate": 3.659263130389338e-05, + "loss": 0.1575, + "step": 67730 + }, + { + "epoch": 4.4, + "learning_rate": 3.6553436111836945e-05, + "loss": 0.1309, + "step": 67740 + }, + { + "epoch": 4.4, + "learning_rate": 3.65142409197805e-05, + "loss": 0.0924, + "step": 67750 + }, + { + "epoch": 4.4, + "learning_rate": 3.647504572772406e-05, + "loss": 0.1082, + "step": 67760 + }, + { + "epoch": 4.4, + "learning_rate": 3.6435850535667625e-05, + "loss": 0.1121, + "step": 67770 + }, + { + "epoch": 4.4, + "learning_rate": 3.639665534361118e-05, + "loss": 0.1161, + "step": 67780 + }, + { + "epoch": 4.4, + "learning_rate": 3.6357460151554736e-05, + "loss": 0.1111, + "step": 67790 + }, + { + "epoch": 4.4, + "learning_rate": 3.63182649594983e-05, + "loss": 0.0962, + "step": 67800 + }, + { + "epoch": 4.4, + "learning_rate": 3.6279069767441854e-05, + "loss": 0.1195, + "step": 67810 + }, + { + "epoch": 4.4, + "learning_rate": 3.6239874575385416e-05, + "loss": 0.0931, + "step": 67820 + }, + { + "epoch": 4.4, + "learning_rate": 3.620067938332898e-05, + "loss": 0.1226, + "step": 67830 + }, + { + "epoch": 4.4, + "learning_rate": 3.6161484191272534e-05, + "loss": 0.1136, + "step": 67840 + }, + { + "epoch": 4.4, + "learning_rate": 3.612228899921609e-05, + "loss": 0.0788, + "step": 67850 + }, + { + "epoch": 4.4, + "learning_rate": 3.608309380715965e-05, + "loss": 0.1065, + "step": 67860 + }, + { + "epoch": 4.4, + "learning_rate": 3.604389861510321e-05, + "loss": 0.0943, + "step": 67870 + }, + { + "epoch": 4.41, + "learning_rate": 3.600470342304677e-05, + "loss": 0.1304, + "step": 67880 + }, + { + "epoch": 4.41, + "learning_rate": 3.596550823099033e-05, + "loss": 0.1223, + "step": 67890 + }, + { + "epoch": 4.41, + "learning_rate": 3.592631303893389e-05, + "loss": 0.0855, + "step": 67900 + }, + { + "epoch": 4.41, + "learning_rate": 3.588711784687744e-05, + "loss": 0.0859, + "step": 67910 + }, + { + "epoch": 4.41, + "learning_rate": 3.5847922654821005e-05, + "loss": 0.1003, + "step": 67920 + }, + { + "epoch": 4.41, + "learning_rate": 3.580872746276456e-05, + "loss": 0.1272, + "step": 67930 + }, + { + "epoch": 4.41, + "learning_rate": 3.576953227070812e-05, + "loss": 0.129, + "step": 67940 + }, + { + "epoch": 4.41, + "learning_rate": 3.5730337078651685e-05, + "loss": 0.0946, + "step": 67950 + }, + { + "epoch": 4.41, + "learning_rate": 3.569114188659524e-05, + "loss": 0.1083, + "step": 67960 + }, + { + "epoch": 4.41, + "learning_rate": 3.56519466945388e-05, + "loss": 0.107, + "step": 67970 + }, + { + "epoch": 4.41, + "learning_rate": 3.561275150248236e-05, + "loss": 0.1269, + "step": 67980 + }, + { + "epoch": 4.41, + "learning_rate": 3.5573556310425914e-05, + "loss": 0.1066, + "step": 67990 + }, + { + "epoch": 4.41, + "learning_rate": 3.5534361118369476e-05, + "loss": 0.0865, + "step": 68000 + }, + { + "epoch": 4.41, + "eval_loss": 0.17573896050453186, + "eval_runtime": 1009.3778, + "eval_samples_per_second": 15.443, + "eval_steps_per_second": 1.931, + "eval_wer": 0.16913760998422198, + "step": 68000 + }, + { + "epoch": 4.41, + "learning_rate": 3.549516592631304e-05, + "loss": 0.1156, + "step": 68010 + }, + { + "epoch": 4.41, + "learning_rate": 3.5455970734256594e-05, + "loss": 0.1253, + "step": 68020 + }, + { + "epoch": 4.42, + "learning_rate": 3.5416775542200156e-05, + "loss": 0.1215, + "step": 68030 + }, + { + "epoch": 4.42, + "learning_rate": 3.537758035014371e-05, + "loss": 0.1081, + "step": 68040 + }, + { + "epoch": 4.42, + "learning_rate": 3.533838515808727e-05, + "loss": 0.1261, + "step": 68050 + }, + { + "epoch": 4.42, + "learning_rate": 3.529918996603083e-05, + "loss": 0.0863, + "step": 68060 + }, + { + "epoch": 4.42, + "learning_rate": 3.525999477397439e-05, + "loss": 0.0985, + "step": 68070 + }, + { + "epoch": 4.42, + "learning_rate": 3.522079958191795e-05, + "loss": 0.1233, + "step": 68080 + }, + { + "epoch": 4.42, + "learning_rate": 3.518160438986151e-05, + "loss": 0.0983, + "step": 68090 + }, + { + "epoch": 4.42, + "learning_rate": 3.5142409197805065e-05, + "loss": 0.0936, + "step": 68100 + }, + { + "epoch": 4.42, + "learning_rate": 3.510321400574862e-05, + "loss": 0.1017, + "step": 68110 + }, + { + "epoch": 4.42, + "learning_rate": 3.506401881369218e-05, + "loss": 0.1034, + "step": 68120 + }, + { + "epoch": 4.42, + "learning_rate": 3.5024823621635745e-05, + "loss": 0.1249, + "step": 68130 + }, + { + "epoch": 4.42, + "learning_rate": 3.49856284295793e-05, + "loss": 0.1049, + "step": 68140 + }, + { + "epoch": 4.42, + "learning_rate": 3.494643323752286e-05, + "loss": 0.0959, + "step": 68150 + }, + { + "epoch": 4.42, + "learning_rate": 3.490723804546642e-05, + "loss": 0.1033, + "step": 68160 + }, + { + "epoch": 4.42, + "learning_rate": 3.486804285340998e-05, + "loss": 0.114, + "step": 68170 + }, + { + "epoch": 4.42, + "learning_rate": 3.4828847661353536e-05, + "loss": 0.1154, + "step": 68180 + }, + { + "epoch": 4.43, + "learning_rate": 3.47896524692971e-05, + "loss": 0.1217, + "step": 68190 + }, + { + "epoch": 4.43, + "learning_rate": 3.4750457277240654e-05, + "loss": 0.0798, + "step": 68200 + }, + { + "epoch": 4.43, + "learning_rate": 3.4711262085184217e-05, + "loss": 0.1004, + "step": 68210 + }, + { + "epoch": 4.43, + "learning_rate": 3.467206689312777e-05, + "loss": 0.1232, + "step": 68220 + }, + { + "epoch": 4.43, + "learning_rate": 3.4632871701071334e-05, + "loss": 0.1571, + "step": 68230 + }, + { + "epoch": 4.43, + "learning_rate": 3.459367650901489e-05, + "loss": 0.1216, + "step": 68240 + }, + { + "epoch": 4.43, + "learning_rate": 3.455448131695845e-05, + "loss": 0.0755, + "step": 68250 + }, + { + "epoch": 4.43, + "learning_rate": 3.451528612490201e-05, + "loss": 0.0958, + "step": 68260 + }, + { + "epoch": 4.43, + "learning_rate": 3.447609093284557e-05, + "loss": 0.0831, + "step": 68270 + }, + { + "epoch": 4.43, + "learning_rate": 3.4436895740789125e-05, + "loss": 0.1328, + "step": 68280 + }, + { + "epoch": 4.43, + "learning_rate": 3.439770054873269e-05, + "loss": 0.1213, + "step": 68290 + }, + { + "epoch": 4.43, + "learning_rate": 3.435850535667624e-05, + "loss": 0.0837, + "step": 68300 + }, + { + "epoch": 4.43, + "learning_rate": 3.4319310164619806e-05, + "loss": 0.0996, + "step": 68310 + }, + { + "epoch": 4.43, + "learning_rate": 3.428011497256336e-05, + "loss": 0.1124, + "step": 68320 + }, + { + "epoch": 4.43, + "learning_rate": 3.4240919780506923e-05, + "loss": 0.1227, + "step": 68330 + }, + { + "epoch": 4.44, + "learning_rate": 3.420172458845048e-05, + "loss": 0.1183, + "step": 68340 + }, + { + "epoch": 4.44, + "learning_rate": 3.416252939639404e-05, + "loss": 0.1067, + "step": 68350 + }, + { + "epoch": 4.44, + "learning_rate": 3.41233342043376e-05, + "loss": 0.0887, + "step": 68360 + }, + { + "epoch": 4.44, + "learning_rate": 3.408413901228116e-05, + "loss": 0.1116, + "step": 68370 + }, + { + "epoch": 4.44, + "learning_rate": 3.4044943820224715e-05, + "loss": 0.1231, + "step": 68380 + }, + { + "epoch": 4.44, + "learning_rate": 3.400574862816828e-05, + "loss": 0.119, + "step": 68390 + }, + { + "epoch": 4.44, + "learning_rate": 3.396655343611184e-05, + "loss": 0.0891, + "step": 68400 + }, + { + "epoch": 4.44, + "learning_rate": 3.3927358244055395e-05, + "loss": 0.1117, + "step": 68410 + }, + { + "epoch": 4.44, + "learning_rate": 3.388816305199895e-05, + "loss": 0.1316, + "step": 68420 + }, + { + "epoch": 4.44, + "learning_rate": 3.384896785994251e-05, + "loss": 0.1016, + "step": 68430 + }, + { + "epoch": 4.44, + "learning_rate": 3.380977266788607e-05, + "loss": 0.114, + "step": 68440 + }, + { + "epoch": 4.44, + "learning_rate": 3.377057747582963e-05, + "loss": 0.0965, + "step": 68450 + }, + { + "epoch": 4.44, + "learning_rate": 3.373138228377319e-05, + "loss": 0.1126, + "step": 68460 + }, + { + "epoch": 4.44, + "learning_rate": 3.369218709171675e-05, + "loss": 0.0951, + "step": 68470 + }, + { + "epoch": 4.44, + "learning_rate": 3.3652991899660304e-05, + "loss": 0.157, + "step": 68480 + }, + { + "epoch": 4.45, + "learning_rate": 3.3613796707603866e-05, + "loss": 0.1012, + "step": 68490 + }, + { + "epoch": 4.45, + "learning_rate": 3.357460151554742e-05, + "loss": 0.0934, + "step": 68500 + }, + { + "epoch": 4.45, + "learning_rate": 3.3535406323490984e-05, + "loss": 0.0914, + "step": 68510 + }, + { + "epoch": 4.45, + "learning_rate": 3.3496211131434546e-05, + "loss": 0.106, + "step": 68520 + }, + { + "epoch": 4.45, + "learning_rate": 3.34570159393781e-05, + "loss": 0.1046, + "step": 68530 + }, + { + "epoch": 4.45, + "learning_rate": 3.341782074732166e-05, + "loss": 0.0932, + "step": 68540 + }, + { + "epoch": 4.45, + "learning_rate": 3.337862555526522e-05, + "loss": 0.1095, + "step": 68550 + }, + { + "epoch": 4.45, + "learning_rate": 3.3339430363208775e-05, + "loss": 0.115, + "step": 68560 + }, + { + "epoch": 4.45, + "learning_rate": 3.330023517115234e-05, + "loss": 0.1215, + "step": 68570 + }, + { + "epoch": 4.45, + "learning_rate": 3.32610399790959e-05, + "loss": 0.1311, + "step": 68580 + }, + { + "epoch": 4.45, + "learning_rate": 3.3221844787039455e-05, + "loss": 0.1151, + "step": 68590 + }, + { + "epoch": 4.45, + "learning_rate": 3.318264959498301e-05, + "loss": 0.0893, + "step": 68600 + }, + { + "epoch": 4.45, + "learning_rate": 3.314345440292657e-05, + "loss": 0.1034, + "step": 68610 + }, + { + "epoch": 4.45, + "learning_rate": 3.310425921087013e-05, + "loss": 0.1062, + "step": 68620 + }, + { + "epoch": 4.45, + "learning_rate": 3.306506401881369e-05, + "loss": 0.1754, + "step": 68630 + }, + { + "epoch": 4.45, + "learning_rate": 3.302586882675725e-05, + "loss": 0.1039, + "step": 68640 + }, + { + "epoch": 4.46, + "learning_rate": 3.298667363470081e-05, + "loss": 0.0728, + "step": 68650 + }, + { + "epoch": 4.46, + "learning_rate": 3.2947478442644364e-05, + "loss": 0.1101, + "step": 68660 + }, + { + "epoch": 4.46, + "learning_rate": 3.2908283250587926e-05, + "loss": 0.1082, + "step": 68670 + }, + { + "epoch": 4.46, + "learning_rate": 3.286908805853148e-05, + "loss": 0.1071, + "step": 68680 + }, + { + "epoch": 4.46, + "learning_rate": 3.2829892866475044e-05, + "loss": 0.1218, + "step": 68690 + }, + { + "epoch": 4.46, + "learning_rate": 3.2790697674418606e-05, + "loss": 0.0964, + "step": 68700 + }, + { + "epoch": 4.46, + "learning_rate": 3.275150248236216e-05, + "loss": 0.1037, + "step": 68710 + }, + { + "epoch": 4.46, + "learning_rate": 3.271230729030572e-05, + "loss": 0.1248, + "step": 68720 + }, + { + "epoch": 4.46, + "learning_rate": 3.267311209824928e-05, + "loss": 0.1386, + "step": 68730 + }, + { + "epoch": 4.46, + "learning_rate": 3.2633916906192835e-05, + "loss": 0.1339, + "step": 68740 + }, + { + "epoch": 4.46, + "learning_rate": 3.25947217141364e-05, + "loss": 0.0928, + "step": 68750 + }, + { + "epoch": 4.46, + "learning_rate": 3.255552652207996e-05, + "loss": 0.0924, + "step": 68760 + }, + { + "epoch": 4.46, + "learning_rate": 3.2516331330023515e-05, + "loss": 0.1159, + "step": 68770 + }, + { + "epoch": 4.46, + "learning_rate": 3.247713613796707e-05, + "loss": 0.129, + "step": 68780 + }, + { + "epoch": 4.46, + "learning_rate": 3.243794094591063e-05, + "loss": 0.1138, + "step": 68790 + }, + { + "epoch": 4.47, + "learning_rate": 3.239874575385419e-05, + "loss": 0.0868, + "step": 68800 + }, + { + "epoch": 4.47, + "learning_rate": 3.235955056179775e-05, + "loss": 0.1032, + "step": 68810 + }, + { + "epoch": 4.47, + "learning_rate": 3.232035536974131e-05, + "loss": 0.102, + "step": 68820 + }, + { + "epoch": 4.47, + "learning_rate": 3.228116017768487e-05, + "loss": 0.1282, + "step": 68830 + }, + { + "epoch": 4.47, + "learning_rate": 3.2241964985628424e-05, + "loss": 0.1156, + "step": 68840 + }, + { + "epoch": 4.47, + "learning_rate": 3.2202769793571986e-05, + "loss": 0.1059, + "step": 68850 + }, + { + "epoch": 4.47, + "learning_rate": 3.216357460151554e-05, + "loss": 0.1034, + "step": 68860 + }, + { + "epoch": 4.47, + "learning_rate": 3.2124379409459104e-05, + "loss": 0.1274, + "step": 68870 + }, + { + "epoch": 4.47, + "learning_rate": 3.2085184217402666e-05, + "loss": 0.0992, + "step": 68880 + }, + { + "epoch": 4.47, + "learning_rate": 3.204598902534622e-05, + "loss": 0.1126, + "step": 68890 + }, + { + "epoch": 4.47, + "learning_rate": 3.200679383328978e-05, + "loss": 0.0854, + "step": 68900 + }, + { + "epoch": 4.47, + "learning_rate": 3.196759864123334e-05, + "loss": 0.0946, + "step": 68910 + }, + { + "epoch": 4.47, + "learning_rate": 3.1928403449176895e-05, + "loss": 0.1143, + "step": 68920 + }, + { + "epoch": 4.47, + "learning_rate": 3.188920825712046e-05, + "loss": 0.1407, + "step": 68930 + }, + { + "epoch": 4.47, + "learning_rate": 3.185001306506402e-05, + "loss": 0.1226, + "step": 68940 + }, + { + "epoch": 4.47, + "learning_rate": 3.1810817873007575e-05, + "loss": 0.0916, + "step": 68950 + }, + { + "epoch": 4.48, + "learning_rate": 3.177162268095113e-05, + "loss": 0.1014, + "step": 68960 + }, + { + "epoch": 4.48, + "learning_rate": 3.173242748889469e-05, + "loss": 0.1157, + "step": 68970 + }, + { + "epoch": 4.48, + "learning_rate": 3.169323229683825e-05, + "loss": 0.1033, + "step": 68980 + }, + { + "epoch": 4.48, + "learning_rate": 3.165403710478181e-05, + "loss": 0.1152, + "step": 68990 + }, + { + "epoch": 4.48, + "learning_rate": 3.161484191272537e-05, + "loss": 0.0871, + "step": 69000 + }, + { + "epoch": 4.48, + "eval_loss": 0.1751844882965088, + "eval_runtime": 1006.4507, + "eval_samples_per_second": 15.488, + "eval_steps_per_second": 1.937, + "eval_wer": 0.16581853847137923, + "step": 69000 + }, + { + "epoch": 4.48, + "learning_rate": 3.157564672066893e-05, + "loss": 0.1045, + "step": 69010 + }, + { + "epoch": 4.48, + "learning_rate": 3.1536451528612484e-05, + "loss": 0.0939, + "step": 69020 + }, + { + "epoch": 4.48, + "learning_rate": 3.1497256336556047e-05, + "loss": 0.1206, + "step": 69030 + }, + { + "epoch": 4.48, + "learning_rate": 3.14580611444996e-05, + "loss": 0.1066, + "step": 69040 + }, + { + "epoch": 4.48, + "learning_rate": 3.1418865952443164e-05, + "loss": 0.0832, + "step": 69050 + }, + { + "epoch": 4.48, + "learning_rate": 3.137967076038673e-05, + "loss": 0.1036, + "step": 69060 + }, + { + "epoch": 4.48, + "learning_rate": 3.134047556833028e-05, + "loss": 0.0934, + "step": 69070 + }, + { + "epoch": 4.48, + "learning_rate": 3.130128037627384e-05, + "loss": 0.0956, + "step": 69080 + }, + { + "epoch": 4.48, + "learning_rate": 3.12620851842174e-05, + "loss": 0.1507, + "step": 69090 + }, + { + "epoch": 4.48, + "learning_rate": 3.1222889992160956e-05, + "loss": 0.0784, + "step": 69100 + }, + { + "epoch": 4.49, + "learning_rate": 3.118369480010452e-05, + "loss": 0.1029, + "step": 69110 + }, + { + "epoch": 4.49, + "learning_rate": 3.114449960804808e-05, + "loss": 0.1141, + "step": 69120 + }, + { + "epoch": 4.49, + "learning_rate": 3.1105304415991636e-05, + "loss": 0.1234, + "step": 69130 + }, + { + "epoch": 4.49, + "learning_rate": 3.106610922393519e-05, + "loss": 0.1171, + "step": 69140 + }, + { + "epoch": 4.49, + "learning_rate": 3.1026914031878753e-05, + "loss": 0.1102, + "step": 69150 + }, + { + "epoch": 4.49, + "learning_rate": 3.098771883982231e-05, + "loss": 0.1243, + "step": 69160 + }, + { + "epoch": 4.49, + "learning_rate": 3.094852364776587e-05, + "loss": 0.0874, + "step": 69170 + }, + { + "epoch": 4.49, + "learning_rate": 3.0909328455709434e-05, + "loss": 0.1394, + "step": 69180 + }, + { + "epoch": 4.49, + "learning_rate": 3.087013326365299e-05, + "loss": 0.1129, + "step": 69190 + }, + { + "epoch": 4.49, + "learning_rate": 3.0830938071596545e-05, + "loss": 0.0865, + "step": 69200 + }, + { + "epoch": 4.49, + "learning_rate": 3.079174287954011e-05, + "loss": 0.0991, + "step": 69210 + }, + { + "epoch": 4.49, + "learning_rate": 3.075254768748366e-05, + "loss": 0.1152, + "step": 69220 + }, + { + "epoch": 4.49, + "learning_rate": 3.0713352495427225e-05, + "loss": 0.1126, + "step": 69230 + }, + { + "epoch": 4.49, + "learning_rate": 3.067415730337079e-05, + "loss": 0.1063, + "step": 69240 + }, + { + "epoch": 4.49, + "learning_rate": 3.063496211131434e-05, + "loss": 0.0917, + "step": 69250 + }, + { + "epoch": 4.5, + "learning_rate": 3.05957669192579e-05, + "loss": 0.1064, + "step": 69260 + }, + { + "epoch": 4.5, + "learning_rate": 3.055657172720146e-05, + "loss": 0.0955, + "step": 69270 + }, + { + "epoch": 4.5, + "learning_rate": 3.051737653514502e-05, + "loss": 0.1254, + "step": 69280 + }, + { + "epoch": 4.5, + "learning_rate": 3.0478181343088578e-05, + "loss": 0.1191, + "step": 69290 + }, + { + "epoch": 4.5, + "learning_rate": 3.0438986151032137e-05, + "loss": 0.0767, + "step": 69300 + }, + { + "epoch": 4.5, + "learning_rate": 3.03997909589757e-05, + "loss": 0.0959, + "step": 69310 + }, + { + "epoch": 4.5, + "learning_rate": 3.0360595766919258e-05, + "loss": 0.1002, + "step": 69320 + }, + { + "epoch": 4.5, + "learning_rate": 3.0321400574862814e-05, + "loss": 0.1123, + "step": 69330 + }, + { + "epoch": 4.5, + "learning_rate": 3.0282205382806373e-05, + "loss": 0.1254, + "step": 69340 + }, + { + "epoch": 4.5, + "learning_rate": 3.024301019074993e-05, + "loss": 0.1027, + "step": 69350 + }, + { + "epoch": 4.5, + "learning_rate": 3.020381499869349e-05, + "loss": 0.105, + "step": 69360 + }, + { + "epoch": 4.5, + "learning_rate": 3.0164619806637053e-05, + "loss": 0.1301, + "step": 69370 + }, + { + "epoch": 4.5, + "learning_rate": 3.012542461458061e-05, + "loss": 0.1474, + "step": 69380 + }, + { + "epoch": 4.5, + "learning_rate": 3.0086229422524167e-05, + "loss": 0.1531, + "step": 69390 + }, + { + "epoch": 4.5, + "learning_rate": 3.0047034230467726e-05, + "loss": 0.0852, + "step": 69400 + }, + { + "epoch": 4.5, + "learning_rate": 3.0007839038411285e-05, + "loss": 0.1137, + "step": 69410 + }, + { + "epoch": 4.51, + "learning_rate": 2.9968643846354844e-05, + "loss": 0.112, + "step": 69420 + }, + { + "epoch": 4.51, + "learning_rate": 2.9929448654298406e-05, + "loss": 0.1336, + "step": 69430 + }, + { + "epoch": 4.51, + "learning_rate": 2.9890253462241965e-05, + "loss": 0.1324, + "step": 69440 + }, + { + "epoch": 4.51, + "learning_rate": 2.985105827018552e-05, + "loss": 0.09, + "step": 69450 + }, + { + "epoch": 4.51, + "learning_rate": 2.981186307812908e-05, + "loss": 0.0944, + "step": 69460 + }, + { + "epoch": 4.51, + "learning_rate": 2.977266788607264e-05, + "loss": 0.0946, + "step": 69470 + }, + { + "epoch": 4.51, + "learning_rate": 2.9733472694016197e-05, + "loss": 0.1256, + "step": 69480 + }, + { + "epoch": 4.51, + "learning_rate": 2.969427750195976e-05, + "loss": 0.123, + "step": 69490 + }, + { + "epoch": 4.51, + "learning_rate": 2.965508230990332e-05, + "loss": 0.0765, + "step": 69500 + }, + { + "epoch": 4.51, + "learning_rate": 2.9615887117846874e-05, + "loss": 0.0922, + "step": 69510 + }, + { + "epoch": 4.51, + "learning_rate": 2.9576691925790433e-05, + "loss": 0.1312, + "step": 69520 + }, + { + "epoch": 4.51, + "learning_rate": 2.9537496733733992e-05, + "loss": 0.1219, + "step": 69530 + }, + { + "epoch": 4.51, + "learning_rate": 2.949830154167755e-05, + "loss": 0.0957, + "step": 69540 + }, + { + "epoch": 4.51, + "learning_rate": 2.9459106349621113e-05, + "loss": 0.0898, + "step": 69550 + }, + { + "epoch": 4.51, + "learning_rate": 2.9419911157564672e-05, + "loss": 0.0948, + "step": 69560 + }, + { + "epoch": 4.52, + "learning_rate": 2.9380715965508227e-05, + "loss": 0.0962, + "step": 69570 + }, + { + "epoch": 4.52, + "learning_rate": 2.9341520773451786e-05, + "loss": 0.1401, + "step": 69580 + }, + { + "epoch": 4.52, + "learning_rate": 2.9302325581395345e-05, + "loss": 0.1112, + "step": 69590 + }, + { + "epoch": 4.52, + "learning_rate": 2.9263130389338904e-05, + "loss": 0.0906, + "step": 69600 + }, + { + "epoch": 4.52, + "learning_rate": 2.9223935197282466e-05, + "loss": 0.1093, + "step": 69610 + }, + { + "epoch": 4.52, + "learning_rate": 2.9184740005226025e-05, + "loss": 0.1006, + "step": 69620 + }, + { + "epoch": 4.52, + "learning_rate": 2.914554481316958e-05, + "loss": 0.1188, + "step": 69630 + }, + { + "epoch": 4.52, + "learning_rate": 2.910634962111314e-05, + "loss": 0.1112, + "step": 69640 + }, + { + "epoch": 4.52, + "learning_rate": 2.90671544290567e-05, + "loss": 0.0866, + "step": 69650 + }, + { + "epoch": 4.52, + "learning_rate": 2.9027959237000257e-05, + "loss": 0.1029, + "step": 69660 + }, + { + "epoch": 4.52, + "learning_rate": 2.898876404494382e-05, + "loss": 0.1001, + "step": 69670 + }, + { + "epoch": 4.52, + "learning_rate": 2.894956885288738e-05, + "loss": 0.1289, + "step": 69680 + }, + { + "epoch": 4.52, + "learning_rate": 2.8910373660830934e-05, + "loss": 0.1002, + "step": 69690 + }, + { + "epoch": 4.52, + "learning_rate": 2.8871178468774493e-05, + "loss": 0.094, + "step": 69700 + }, + { + "epoch": 4.52, + "learning_rate": 2.8831983276718052e-05, + "loss": 0.0981, + "step": 69710 + }, + { + "epoch": 4.52, + "learning_rate": 2.879278808466161e-05, + "loss": 0.1055, + "step": 69720 + }, + { + "epoch": 4.53, + "learning_rate": 2.8753592892605173e-05, + "loss": 0.1337, + "step": 69730 + }, + { + "epoch": 4.53, + "learning_rate": 2.8714397700548732e-05, + "loss": 0.1038, + "step": 69740 + }, + { + "epoch": 4.53, + "learning_rate": 2.8675202508492288e-05, + "loss": 0.0964, + "step": 69750 + }, + { + "epoch": 4.53, + "learning_rate": 2.8636007316435847e-05, + "loss": 0.1131, + "step": 69760 + }, + { + "epoch": 4.53, + "learning_rate": 2.8596812124379405e-05, + "loss": 0.1147, + "step": 69770 + }, + { + "epoch": 4.53, + "learning_rate": 2.8557616932322964e-05, + "loss": 0.1343, + "step": 69780 + }, + { + "epoch": 4.53, + "learning_rate": 2.8518421740266527e-05, + "loss": 0.1293, + "step": 69790 + }, + { + "epoch": 4.53, + "learning_rate": 2.8479226548210086e-05, + "loss": 0.0808, + "step": 69800 + }, + { + "epoch": 4.53, + "learning_rate": 2.8440031356153644e-05, + "loss": 0.0816, + "step": 69810 + }, + { + "epoch": 4.53, + "learning_rate": 2.84008361640972e-05, + "loss": 0.0953, + "step": 69820 + }, + { + "epoch": 4.53, + "learning_rate": 2.836164097204076e-05, + "loss": 0.146, + "step": 69830 + }, + { + "epoch": 4.53, + "learning_rate": 2.8322445779984318e-05, + "loss": 0.1302, + "step": 69840 + }, + { + "epoch": 4.53, + "learning_rate": 2.828325058792788e-05, + "loss": 0.1003, + "step": 69850 + }, + { + "epoch": 4.53, + "learning_rate": 2.824405539587144e-05, + "loss": 0.0995, + "step": 69860 + }, + { + "epoch": 4.53, + "learning_rate": 2.8204860203814998e-05, + "loss": 0.1042, + "step": 69870 + }, + { + "epoch": 4.54, + "learning_rate": 2.8165665011758553e-05, + "loss": 0.1439, + "step": 69880 + }, + { + "epoch": 4.54, + "learning_rate": 2.8126469819702112e-05, + "loss": 0.1024, + "step": 69890 + }, + { + "epoch": 4.54, + "learning_rate": 2.808727462764567e-05, + "loss": 0.0918, + "step": 69900 + }, + { + "epoch": 4.54, + "learning_rate": 2.8048079435589233e-05, + "loss": 0.0967, + "step": 69910 + }, + { + "epoch": 4.54, + "learning_rate": 2.8008884243532792e-05, + "loss": 0.1352, + "step": 69920 + }, + { + "epoch": 4.54, + "learning_rate": 2.7973608570681995e-05, + "loss": 0.145, + "step": 69930 + }, + { + "epoch": 4.54, + "learning_rate": 2.7934413378625554e-05, + "loss": 0.1309, + "step": 69940 + }, + { + "epoch": 4.54, + "learning_rate": 2.789521818656911e-05, + "loss": 0.0907, + "step": 69950 + }, + { + "epoch": 4.54, + "learning_rate": 2.785602299451267e-05, + "loss": 0.0977, + "step": 69960 + }, + { + "epoch": 4.54, + "learning_rate": 2.781682780245623e-05, + "loss": 0.0915, + "step": 69970 + }, + { + "epoch": 4.54, + "learning_rate": 2.777763261039979e-05, + "loss": 0.1297, + "step": 69980 + }, + { + "epoch": 4.54, + "learning_rate": 2.773843741834335e-05, + "loss": 0.1227, + "step": 69990 + }, + { + "epoch": 4.54, + "learning_rate": 2.7699242226286907e-05, + "loss": 0.1161, + "step": 70000 + }, + { + "epoch": 4.54, + "eval_loss": 0.1720564067363739, + "eval_runtime": 1006.6441, + "eval_samples_per_second": 15.485, + "eval_steps_per_second": 1.936, + "eval_wer": 0.1680573126181575, + "step": 70000 + }, + { + "epoch": 4.54, + "learning_rate": 2.7660047034230463e-05, + "loss": 0.1002, + "step": 70010 + }, + { + "epoch": 4.54, + "learning_rate": 2.7620851842174022e-05, + "loss": 0.1098, + "step": 70020 + }, + { + "epoch": 4.55, + "learning_rate": 2.7581656650117584e-05, + "loss": 0.1598, + "step": 70030 + }, + { + "epoch": 4.55, + "learning_rate": 2.7542461458061143e-05, + "loss": 0.099, + "step": 70040 + }, + { + "epoch": 4.55, + "learning_rate": 2.7503266266004702e-05, + "loss": 0.1074, + "step": 70050 + }, + { + "epoch": 4.55, + "learning_rate": 2.746407107394826e-05, + "loss": 0.0909, + "step": 70060 + }, + { + "epoch": 4.55, + "learning_rate": 2.7424875881891816e-05, + "loss": 0.1102, + "step": 70070 + }, + { + "epoch": 4.55, + "learning_rate": 2.7385680689835375e-05, + "loss": 0.1432, + "step": 70080 + }, + { + "epoch": 4.55, + "learning_rate": 2.7346485497778937e-05, + "loss": 0.1103, + "step": 70090 + }, + { + "epoch": 4.55, + "learning_rate": 2.7307290305722496e-05, + "loss": 0.1007, + "step": 70100 + }, + { + "epoch": 4.55, + "learning_rate": 2.7268095113666055e-05, + "loss": 0.089, + "step": 70110 + }, + { + "epoch": 4.55, + "learning_rate": 2.7228899921609614e-05, + "loss": 0.0866, + "step": 70120 + }, + { + "epoch": 4.55, + "learning_rate": 2.7189704729553173e-05, + "loss": 0.1431, + "step": 70130 + }, + { + "epoch": 4.55, + "learning_rate": 2.715050953749673e-05, + "loss": 0.0956, + "step": 70140 + }, + { + "epoch": 4.55, + "learning_rate": 2.711131434544029e-05, + "loss": 0.1056, + "step": 70150 + }, + { + "epoch": 4.55, + "learning_rate": 2.707211915338385e-05, + "loss": 0.121, + "step": 70160 + }, + { + "epoch": 4.55, + "learning_rate": 2.703292396132741e-05, + "loss": 0.0971, + "step": 70170 + }, + { + "epoch": 4.55, + "learning_rate": 2.6993728769270968e-05, + "loss": 0.1144, + "step": 70180 + }, + { + "epoch": 4.56, + "learning_rate": 2.6954533577214527e-05, + "loss": 0.1389, + "step": 70190 + }, + { + "epoch": 4.56, + "learning_rate": 2.6915338385158082e-05, + "loss": 0.0796, + "step": 70200 + }, + { + "epoch": 4.56, + "learning_rate": 2.6876143193101644e-05, + "loss": 0.0899, + "step": 70210 + }, + { + "epoch": 4.56, + "learning_rate": 2.6836948001045203e-05, + "loss": 0.0968, + "step": 70220 + }, + { + "epoch": 4.56, + "learning_rate": 2.6797752808988762e-05, + "loss": 0.1278, + "step": 70230 + }, + { + "epoch": 4.56, + "learning_rate": 2.675855761693232e-05, + "loss": 0.1173, + "step": 70240 + }, + { + "epoch": 4.56, + "learning_rate": 2.671936242487588e-05, + "loss": 0.1061, + "step": 70250 + }, + { + "epoch": 4.56, + "learning_rate": 2.6680167232819435e-05, + "loss": 0.099, + "step": 70260 + }, + { + "epoch": 4.56, + "learning_rate": 2.6640972040762998e-05, + "loss": 0.0968, + "step": 70270 + }, + { + "epoch": 4.56, + "learning_rate": 2.6601776848706557e-05, + "loss": 0.1247, + "step": 70280 + }, + { + "epoch": 4.56, + "learning_rate": 2.6562581656650116e-05, + "loss": 0.1171, + "step": 70290 + }, + { + "epoch": 4.56, + "learning_rate": 2.6523386464593674e-05, + "loss": 0.0983, + "step": 70300 + }, + { + "epoch": 4.56, + "learning_rate": 2.6484191272537233e-05, + "loss": 0.0825, + "step": 70310 + }, + { + "epoch": 4.56, + "learning_rate": 2.644499608048079e-05, + "loss": 0.0896, + "step": 70320 + }, + { + "epoch": 4.56, + "learning_rate": 2.640580088842435e-05, + "loss": 0.1258, + "step": 70330 + }, + { + "epoch": 4.57, + "learning_rate": 2.636660569636791e-05, + "loss": 0.124, + "step": 70340 + }, + { + "epoch": 4.57, + "learning_rate": 2.632741050431147e-05, + "loss": 0.0946, + "step": 70350 + }, + { + "epoch": 4.57, + "learning_rate": 2.6288215312255028e-05, + "loss": 0.1211, + "step": 70360 + }, + { + "epoch": 4.57, + "learning_rate": 2.6249020120198587e-05, + "loss": 0.1046, + "step": 70370 + }, + { + "epoch": 4.57, + "learning_rate": 2.6209824928142142e-05, + "loss": 0.1552, + "step": 70380 + }, + { + "epoch": 4.57, + "learning_rate": 2.6170629736085705e-05, + "loss": 0.1191, + "step": 70390 + }, + { + "epoch": 4.57, + "learning_rate": 2.6131434544029263e-05, + "loss": 0.0923, + "step": 70400 + }, + { + "epoch": 4.57, + "learning_rate": 2.6092239351972822e-05, + "loss": 0.0908, + "step": 70410 + }, + { + "epoch": 4.57, + "learning_rate": 2.605304415991638e-05, + "loss": 0.0923, + "step": 70420 + }, + { + "epoch": 4.57, + "learning_rate": 2.601384896785994e-05, + "loss": 0.1263, + "step": 70430 + }, + { + "epoch": 4.57, + "learning_rate": 2.5974653775803496e-05, + "loss": 0.0936, + "step": 70440 + }, + { + "epoch": 4.57, + "learning_rate": 2.5935458583747058e-05, + "loss": 0.0916, + "step": 70450 + }, + { + "epoch": 4.57, + "learning_rate": 2.5896263391690617e-05, + "loss": 0.0986, + "step": 70460 + }, + { + "epoch": 4.57, + "learning_rate": 2.5857068199634176e-05, + "loss": 0.0881, + "step": 70470 + }, + { + "epoch": 4.57, + "learning_rate": 2.5817873007577735e-05, + "loss": 0.1228, + "step": 70480 + }, + { + "epoch": 4.57, + "learning_rate": 2.5778677815521297e-05, + "loss": 0.1009, + "step": 70490 + }, + { + "epoch": 4.58, + "learning_rate": 2.573948262346485e-05, + "loss": 0.0734, + "step": 70500 + }, + { + "epoch": 4.58, + "learning_rate": 2.570028743140841e-05, + "loss": 0.0919, + "step": 70510 + }, + { + "epoch": 4.58, + "learning_rate": 2.566109223935197e-05, + "loss": 0.1004, + "step": 70520 + }, + { + "epoch": 4.58, + "learning_rate": 2.562189704729553e-05, + "loss": 0.1528, + "step": 70530 + }, + { + "epoch": 4.58, + "learning_rate": 2.5582701855239088e-05, + "loss": 0.1023, + "step": 70540 + }, + { + "epoch": 4.58, + "learning_rate": 2.554350666318265e-05, + "loss": 0.1003, + "step": 70550 + }, + { + "epoch": 4.58, + "learning_rate": 2.5504311471126203e-05, + "loss": 0.0881, + "step": 70560 + }, + { + "epoch": 4.58, + "learning_rate": 2.5465116279069765e-05, + "loss": 0.0983, + "step": 70570 + }, + { + "epoch": 4.58, + "learning_rate": 2.5425921087013324e-05, + "loss": 0.1148, + "step": 70580 + }, + { + "epoch": 4.58, + "learning_rate": 2.5386725894956883e-05, + "loss": 0.1079, + "step": 70590 + }, + { + "epoch": 4.58, + "learning_rate": 2.534753070290044e-05, + "loss": 0.0885, + "step": 70600 + }, + { + "epoch": 4.58, + "learning_rate": 2.5308335510844004e-05, + "loss": 0.0897, + "step": 70610 + }, + { + "epoch": 4.58, + "learning_rate": 2.5269140318787563e-05, + "loss": 0.0917, + "step": 70620 + }, + { + "epoch": 4.58, + "learning_rate": 2.5229945126731118e-05, + "loss": 0.1344, + "step": 70630 + }, + { + "epoch": 4.58, + "learning_rate": 2.5190749934674677e-05, + "loss": 0.1519, + "step": 70640 + }, + { + "epoch": 4.59, + "learning_rate": 2.5151554742618236e-05, + "loss": 0.0885, + "step": 70650 + }, + { + "epoch": 4.59, + "learning_rate": 2.5112359550561795e-05, + "loss": 0.1014, + "step": 70660 + }, + { + "epoch": 4.59, + "learning_rate": 2.5073164358505357e-05, + "loss": 0.1074, + "step": 70670 + }, + { + "epoch": 4.59, + "learning_rate": 2.5033969166448916e-05, + "loss": 0.1389, + "step": 70680 + }, + { + "epoch": 4.59, + "learning_rate": 2.499477397439247e-05, + "loss": 0.103, + "step": 70690 + }, + { + "epoch": 4.59, + "learning_rate": 2.495557878233603e-05, + "loss": 0.0882, + "step": 70700 + }, + { + "epoch": 4.59, + "learning_rate": 2.491638359027959e-05, + "loss": 0.0976, + "step": 70710 + }, + { + "epoch": 4.59, + "learning_rate": 2.487718839822315e-05, + "loss": 0.1024, + "step": 70720 + }, + { + "epoch": 4.59, + "learning_rate": 2.483799320616671e-05, + "loss": 0.1332, + "step": 70730 + }, + { + "epoch": 4.59, + "learning_rate": 2.479879801411027e-05, + "loss": 0.159, + "step": 70740 + }, + { + "epoch": 4.59, + "learning_rate": 2.4759602822053825e-05, + "loss": 0.0631, + "step": 70750 + }, + { + "epoch": 4.59, + "learning_rate": 2.4720407629997384e-05, + "loss": 0.1023, + "step": 70760 + }, + { + "epoch": 4.59, + "learning_rate": 2.4681212437940943e-05, + "loss": 0.096, + "step": 70770 + }, + { + "epoch": 4.59, + "learning_rate": 2.4642017245884502e-05, + "loss": 0.1252, + "step": 70780 + }, + { + "epoch": 4.59, + "learning_rate": 2.4602822053828064e-05, + "loss": 0.109, + "step": 70790 + }, + { + "epoch": 4.6, + "learning_rate": 2.4563626861771623e-05, + "loss": 0.0877, + "step": 70800 + }, + { + "epoch": 4.6, + "learning_rate": 2.452443166971518e-05, + "loss": 0.091, + "step": 70810 + }, + { + "epoch": 4.6, + "learning_rate": 2.4485236477658737e-05, + "loss": 0.1154, + "step": 70820 + }, + { + "epoch": 4.6, + "learning_rate": 2.4446041285602296e-05, + "loss": 0.1039, + "step": 70830 + }, + { + "epoch": 4.6, + "learning_rate": 2.4406846093545855e-05, + "loss": 0.1304, + "step": 70840 + }, + { + "epoch": 4.6, + "learning_rate": 2.4367650901489417e-05, + "loss": 0.0958, + "step": 70850 + }, + { + "epoch": 4.6, + "learning_rate": 2.4328455709432976e-05, + "loss": 0.0979, + "step": 70860 + }, + { + "epoch": 4.6, + "learning_rate": 2.4289260517376532e-05, + "loss": 0.1106, + "step": 70870 + }, + { + "epoch": 4.6, + "learning_rate": 2.425006532532009e-05, + "loss": 0.1776, + "step": 70880 + }, + { + "epoch": 4.6, + "learning_rate": 2.421087013326365e-05, + "loss": 0.1499, + "step": 70890 + }, + { + "epoch": 4.6, + "learning_rate": 2.417167494120721e-05, + "loss": 0.0845, + "step": 70900 + }, + { + "epoch": 4.6, + "learning_rate": 2.413247974915077e-05, + "loss": 0.0935, + "step": 70910 + }, + { + "epoch": 4.6, + "learning_rate": 2.409328455709433e-05, + "loss": 0.1009, + "step": 70920 + }, + { + "epoch": 4.6, + "learning_rate": 2.4054089365037885e-05, + "loss": 0.146, + "step": 70930 + }, + { + "epoch": 4.6, + "learning_rate": 2.4014894172981444e-05, + "loss": 0.0854, + "step": 70940 + }, + { + "epoch": 4.6, + "learning_rate": 2.3975698980925003e-05, + "loss": 0.0958, + "step": 70950 + }, + { + "epoch": 4.61, + "learning_rate": 2.3936503788868562e-05, + "loss": 0.0926, + "step": 70960 + }, + { + "epoch": 4.61, + "learning_rate": 2.3897308596812124e-05, + "loss": 0.1131, + "step": 70970 + }, + { + "epoch": 4.61, + "learning_rate": 2.3858113404755683e-05, + "loss": 0.1185, + "step": 70980 + }, + { + "epoch": 4.61, + "learning_rate": 2.381891821269924e-05, + "loss": 0.121, + "step": 70990 + }, + { + "epoch": 4.61, + "learning_rate": 2.3779723020642798e-05, + "loss": 0.1062, + "step": 71000 + }, + { + "epoch": 4.61, + "eval_loss": 0.17095209658145905, + "eval_runtime": 1010.2468, + "eval_samples_per_second": 15.43, + "eval_steps_per_second": 1.929, + "eval_wer": 0.16382140978806273, + "step": 71000 + }, + { + "epoch": 4.61, + "learning_rate": 2.3740527828586357e-05, + "loss": 0.083, + "step": 71010 + }, + { + "epoch": 4.61, + "learning_rate": 2.3701332636529915e-05, + "loss": 0.1002, + "step": 71020 + }, + { + "epoch": 4.61, + "learning_rate": 2.3662137444473478e-05, + "loss": 0.1057, + "step": 71030 + }, + { + "epoch": 4.61, + "learning_rate": 2.3622942252417037e-05, + "loss": 0.1429, + "step": 71040 + }, + { + "epoch": 4.61, + "learning_rate": 2.3583747060360596e-05, + "loss": 0.0851, + "step": 71050 + }, + { + "epoch": 4.61, + "learning_rate": 2.354455186830415e-05, + "loss": 0.0818, + "step": 71060 + }, + { + "epoch": 4.61, + "learning_rate": 2.350535667624771e-05, + "loss": 0.0837, + "step": 71070 + }, + { + "epoch": 4.61, + "learning_rate": 2.346616148419127e-05, + "loss": 0.1299, + "step": 71080 + }, + { + "epoch": 4.61, + "learning_rate": 2.342696629213483e-05, + "loss": 0.1135, + "step": 71090 + }, + { + "epoch": 4.61, + "learning_rate": 2.338777110007839e-05, + "loss": 0.0863, + "step": 71100 + }, + { + "epoch": 4.62, + "learning_rate": 2.334857590802195e-05, + "loss": 0.0891, + "step": 71110 + }, + { + "epoch": 4.62, + "learning_rate": 2.3309380715965504e-05, + "loss": 0.1189, + "step": 71120 + }, + { + "epoch": 4.62, + "learning_rate": 2.3270185523909063e-05, + "loss": 0.1177, + "step": 71130 + }, + { + "epoch": 4.62, + "learning_rate": 2.3230990331852622e-05, + "loss": 0.1177, + "step": 71140 + }, + { + "epoch": 4.62, + "learning_rate": 2.3191795139796185e-05, + "loss": 0.0982, + "step": 71150 + }, + { + "epoch": 4.62, + "learning_rate": 2.3152599947739743e-05, + "loss": 0.096, + "step": 71160 + }, + { + "epoch": 4.62, + "learning_rate": 2.3113404755683302e-05, + "loss": 0.1129, + "step": 71170 + }, + { + "epoch": 4.62, + "learning_rate": 2.3074209563626858e-05, + "loss": 0.1296, + "step": 71180 + }, + { + "epoch": 4.62, + "learning_rate": 2.3035014371570417e-05, + "loss": 0.1067, + "step": 71190 + }, + { + "epoch": 4.62, + "learning_rate": 2.2995819179513976e-05, + "loss": 0.0932, + "step": 71200 + }, + { + "epoch": 4.62, + "learning_rate": 2.2956623987457538e-05, + "loss": 0.1095, + "step": 71210 + }, + { + "epoch": 4.62, + "learning_rate": 2.2917428795401097e-05, + "loss": 0.0983, + "step": 71220 + }, + { + "epoch": 4.62, + "learning_rate": 2.2878233603344656e-05, + "loss": 0.1406, + "step": 71230 + }, + { + "epoch": 4.62, + "learning_rate": 2.283903841128821e-05, + "loss": 0.1106, + "step": 71240 + }, + { + "epoch": 4.62, + "learning_rate": 2.279984321923177e-05, + "loss": 0.1038, + "step": 71250 + }, + { + "epoch": 4.62, + "learning_rate": 2.276064802717533e-05, + "loss": 0.0917, + "step": 71260 + }, + { + "epoch": 4.63, + "learning_rate": 2.272145283511889e-05, + "loss": 0.12, + "step": 71270 + }, + { + "epoch": 4.63, + "learning_rate": 2.268225764306245e-05, + "loss": 0.1005, + "step": 71280 + }, + { + "epoch": 4.63, + "learning_rate": 2.264306245100601e-05, + "loss": 0.1096, + "step": 71290 + }, + { + "epoch": 4.63, + "learning_rate": 2.2603867258949565e-05, + "loss": 0.08, + "step": 71300 + }, + { + "epoch": 4.63, + "learning_rate": 2.2564672066893124e-05, + "loss": 0.1051, + "step": 71310 + }, + { + "epoch": 4.63, + "learning_rate": 2.2525476874836686e-05, + "loss": 0.0994, + "step": 71320 + }, + { + "epoch": 4.63, + "learning_rate": 2.2486281682780245e-05, + "loss": 0.115, + "step": 71330 + }, + { + "epoch": 4.63, + "learning_rate": 2.2447086490723804e-05, + "loss": 0.1098, + "step": 71340 + }, + { + "epoch": 4.63, + "learning_rate": 2.2407891298667363e-05, + "loss": 0.0918, + "step": 71350 + }, + { + "epoch": 4.63, + "learning_rate": 2.2368696106610918e-05, + "loss": 0.1003, + "step": 71360 + }, + { + "epoch": 4.63, + "learning_rate": 2.2329500914554477e-05, + "loss": 0.097, + "step": 71370 + }, + { + "epoch": 4.63, + "learning_rate": 2.229030572249804e-05, + "loss": 0.1245, + "step": 71380 + }, + { + "epoch": 4.63, + "learning_rate": 2.2251110530441598e-05, + "loss": 0.1318, + "step": 71390 + }, + { + "epoch": 4.63, + "learning_rate": 2.2211915338385157e-05, + "loss": 0.0829, + "step": 71400 + }, + { + "epoch": 4.63, + "learning_rate": 2.2172720146328716e-05, + "loss": 0.1066, + "step": 71410 + }, + { + "epoch": 4.64, + "learning_rate": 2.213352495427227e-05, + "loss": 0.112, + "step": 71420 + }, + { + "epoch": 4.64, + "learning_rate": 2.209432976221583e-05, + "loss": 0.1007, + "step": 71430 + }, + { + "epoch": 4.64, + "learning_rate": 2.2055134570159393e-05, + "loss": 0.1014, + "step": 71440 + }, + { + "epoch": 4.64, + "learning_rate": 2.201593937810295e-05, + "loss": 0.0876, + "step": 71450 + }, + { + "epoch": 4.64, + "learning_rate": 2.197674418604651e-05, + "loss": 0.1072, + "step": 71460 + }, + { + "epoch": 4.64, + "learning_rate": 2.193754899399007e-05, + "loss": 0.0955, + "step": 71470 + }, + { + "epoch": 4.64, + "learning_rate": 2.1898353801933625e-05, + "loss": 0.1308, + "step": 71480 + }, + { + "epoch": 4.64, + "learning_rate": 2.1859158609877184e-05, + "loss": 0.1038, + "step": 71490 + }, + { + "epoch": 4.64, + "learning_rate": 2.1819963417820746e-05, + "loss": 0.0708, + "step": 71500 + }, + { + "epoch": 4.64, + "learning_rate": 2.1780768225764305e-05, + "loss": 0.1017, + "step": 71510 + }, + { + "epoch": 4.64, + "learning_rate": 2.1741573033707864e-05, + "loss": 0.0966, + "step": 71520 + }, + { + "epoch": 4.64, + "learning_rate": 2.1702377841651423e-05, + "loss": 0.1532, + "step": 71530 + }, + { + "epoch": 4.64, + "learning_rate": 2.1663182649594982e-05, + "loss": 0.1301, + "step": 71540 + }, + { + "epoch": 4.64, + "learning_rate": 2.1623987457538537e-05, + "loss": 0.1076, + "step": 71550 + }, + { + "epoch": 4.64, + "learning_rate": 2.15847922654821e-05, + "loss": 0.0996, + "step": 71560 + }, + { + "epoch": 4.64, + "learning_rate": 2.154559707342566e-05, + "loss": 0.105, + "step": 71570 + }, + { + "epoch": 4.65, + "learning_rate": 2.1506401881369217e-05, + "loss": 0.1149, + "step": 71580 + }, + { + "epoch": 4.65, + "learning_rate": 2.1467206689312776e-05, + "loss": 0.1262, + "step": 71590 + }, + { + "epoch": 4.65, + "learning_rate": 2.1428011497256335e-05, + "loss": 0.099, + "step": 71600 + }, + { + "epoch": 4.65, + "learning_rate": 2.138881630519989e-05, + "loss": 0.1074, + "step": 71610 + }, + { + "epoch": 4.65, + "learning_rate": 2.1349621113143453e-05, + "loss": 0.086, + "step": 71620 + }, + { + "epoch": 4.65, + "learning_rate": 2.1310425921087012e-05, + "loss": 0.1254, + "step": 71630 + }, + { + "epoch": 4.65, + "learning_rate": 2.127123072903057e-05, + "loss": 0.1155, + "step": 71640 + }, + { + "epoch": 4.65, + "learning_rate": 2.123203553697413e-05, + "loss": 0.0804, + "step": 71650 + }, + { + "epoch": 4.65, + "learning_rate": 2.119284034491769e-05, + "loss": 0.1119, + "step": 71660 + }, + { + "epoch": 4.65, + "learning_rate": 2.1153645152861244e-05, + "loss": 0.1153, + "step": 71670 + }, + { + "epoch": 4.65, + "learning_rate": 2.1114449960804806e-05, + "loss": 0.1179, + "step": 71680 + }, + { + "epoch": 4.65, + "learning_rate": 2.1075254768748365e-05, + "loss": 0.0976, + "step": 71690 + }, + { + "epoch": 4.65, + "learning_rate": 2.1036059576691924e-05, + "loss": 0.0893, + "step": 71700 + }, + { + "epoch": 4.65, + "learning_rate": 2.0996864384635483e-05, + "loss": 0.0987, + "step": 71710 + }, + { + "epoch": 4.65, + "learning_rate": 2.0957669192579042e-05, + "loss": 0.1116, + "step": 71720 + }, + { + "epoch": 4.66, + "learning_rate": 2.0918474000522598e-05, + "loss": 0.1008, + "step": 71730 + }, + { + "epoch": 4.66, + "learning_rate": 2.087927880846616e-05, + "loss": 0.1134, + "step": 71740 + }, + { + "epoch": 4.66, + "learning_rate": 2.084008361640972e-05, + "loss": 0.0775, + "step": 71750 + }, + { + "epoch": 4.66, + "learning_rate": 2.0800888424353278e-05, + "loss": 0.102, + "step": 71760 + }, + { + "epoch": 4.66, + "learning_rate": 2.0761693232296837e-05, + "loss": 0.0977, + "step": 71770 + }, + { + "epoch": 4.66, + "learning_rate": 2.0722498040240395e-05, + "loss": 0.1106, + "step": 71780 + }, + { + "epoch": 4.66, + "learning_rate": 2.068330284818395e-05, + "loss": 0.1099, + "step": 71790 + }, + { + "epoch": 4.66, + "learning_rate": 2.0644107656127513e-05, + "loss": 0.0876, + "step": 71800 + }, + { + "epoch": 4.66, + "learning_rate": 2.0604912464071072e-05, + "loss": 0.0884, + "step": 71810 + }, + { + "epoch": 4.66, + "learning_rate": 2.056571727201463e-05, + "loss": 0.0863, + "step": 71820 + }, + { + "epoch": 4.66, + "learning_rate": 2.052652207995819e-05, + "loss": 0.1184, + "step": 71830 + }, + { + "epoch": 4.66, + "learning_rate": 2.048732688790175e-05, + "loss": 0.1011, + "step": 71840 + }, + { + "epoch": 4.66, + "learning_rate": 2.0448131695845304e-05, + "loss": 0.0934, + "step": 71850 + }, + { + "epoch": 4.66, + "learning_rate": 2.0408936503788867e-05, + "loss": 0.0967, + "step": 71860 + }, + { + "epoch": 4.66, + "learning_rate": 2.0369741311732426e-05, + "loss": 0.1004, + "step": 71870 + }, + { + "epoch": 4.67, + "learning_rate": 2.0330546119675984e-05, + "loss": 0.1339, + "step": 71880 + }, + { + "epoch": 4.67, + "learning_rate": 2.0291350927619543e-05, + "loss": 0.1198, + "step": 71890 + }, + { + "epoch": 4.67, + "learning_rate": 2.0252155735563102e-05, + "loss": 0.1026, + "step": 71900 + }, + { + "epoch": 4.67, + "learning_rate": 2.0212960543506658e-05, + "loss": 0.0889, + "step": 71910 + }, + { + "epoch": 4.67, + "learning_rate": 2.017376535145022e-05, + "loss": 0.1017, + "step": 71920 + }, + { + "epoch": 4.67, + "learning_rate": 2.013457015939378e-05, + "loss": 0.1242, + "step": 71930 + }, + { + "epoch": 4.67, + "learning_rate": 2.0095374967337338e-05, + "loss": 0.1234, + "step": 71940 + }, + { + "epoch": 4.67, + "learning_rate": 2.0056179775280897e-05, + "loss": 0.1105, + "step": 71950 + }, + { + "epoch": 4.67, + "learning_rate": 2.001698458322446e-05, + "loss": 0.085, + "step": 71960 + }, + { + "epoch": 4.67, + "learning_rate": 1.9977789391168018e-05, + "loss": 0.1267, + "step": 71970 + }, + { + "epoch": 4.67, + "learning_rate": 1.9938594199111574e-05, + "loss": 0.1292, + "step": 71980 + }, + { + "epoch": 4.67, + "learning_rate": 1.9899399007055132e-05, + "loss": 0.0967, + "step": 71990 + }, + { + "epoch": 4.67, + "learning_rate": 1.986020381499869e-05, + "loss": 0.0766, + "step": 72000 + }, + { + "epoch": 4.67, + "eval_loss": 0.1713821291923523, + "eval_runtime": 1024.5294, + "eval_samples_per_second": 15.215, + "eval_steps_per_second": 1.902, + "eval_wer": 0.16301118676351437, + "step": 72000 + }, + { + "epoch": 4.67, + "learning_rate": 1.982100862294225e-05, + "loss": 0.0867, + "step": 72010 + }, + { + "epoch": 4.67, + "learning_rate": 1.9781813430885813e-05, + "loss": 0.0897, + "step": 72020 + }, + { + "epoch": 4.67, + "learning_rate": 1.974261823882937e-05, + "loss": 0.0901, + "step": 72030 + }, + { + "epoch": 4.68, + "learning_rate": 1.9703423046772927e-05, + "loss": 0.126, + "step": 72040 + }, + { + "epoch": 4.68, + "learning_rate": 1.9664227854716486e-05, + "loss": 0.0922, + "step": 72050 + }, + { + "epoch": 4.68, + "learning_rate": 1.9625032662660045e-05, + "loss": 0.0872, + "step": 72060 + }, + { + "epoch": 4.68, + "learning_rate": 1.9585837470603604e-05, + "loss": 0.1321, + "step": 72070 + }, + { + "epoch": 4.68, + "learning_rate": 1.9546642278547166e-05, + "loss": 0.1324, + "step": 72080 + }, + { + "epoch": 4.68, + "learning_rate": 1.9507447086490725e-05, + "loss": 0.1392, + "step": 72090 + }, + { + "epoch": 4.68, + "learning_rate": 1.946825189443428e-05, + "loss": 0.0788, + "step": 72100 + }, + { + "epoch": 4.68, + "learning_rate": 1.942905670237784e-05, + "loss": 0.1185, + "step": 72110 + }, + { + "epoch": 4.68, + "learning_rate": 1.9389861510321398e-05, + "loss": 0.0843, + "step": 72120 + }, + { + "epoch": 4.68, + "learning_rate": 1.9350666318264957e-05, + "loss": 0.1142, + "step": 72130 + }, + { + "epoch": 4.68, + "learning_rate": 1.931147112620852e-05, + "loss": 0.1132, + "step": 72140 + }, + { + "epoch": 4.68, + "learning_rate": 1.9272275934152078e-05, + "loss": 0.0963, + "step": 72150 + }, + { + "epoch": 4.68, + "learning_rate": 1.9233080742095634e-05, + "loss": 0.0971, + "step": 72160 + }, + { + "epoch": 4.68, + "learning_rate": 1.9193885550039193e-05, + "loss": 0.0937, + "step": 72170 + }, + { + "epoch": 4.68, + "learning_rate": 1.915469035798275e-05, + "loss": 0.1071, + "step": 72180 + }, + { + "epoch": 4.69, + "learning_rate": 1.911549516592631e-05, + "loss": 0.126, + "step": 72190 + }, + { + "epoch": 4.69, + "learning_rate": 1.9076299973869873e-05, + "loss": 0.0739, + "step": 72200 + }, + { + "epoch": 4.69, + "learning_rate": 1.903710478181343e-05, + "loss": 0.1119, + "step": 72210 + }, + { + "epoch": 4.69, + "learning_rate": 1.8997909589756987e-05, + "loss": 0.1011, + "step": 72220 + }, + { + "epoch": 4.69, + "learning_rate": 1.8958714397700546e-05, + "loss": 0.1145, + "step": 72230 + }, + { + "epoch": 4.69, + "learning_rate": 1.8919519205644105e-05, + "loss": 0.1364, + "step": 72240 + }, + { + "epoch": 4.69, + "learning_rate": 1.8880324013587664e-05, + "loss": 0.0815, + "step": 72250 + }, + { + "epoch": 4.69, + "learning_rate": 1.8841128821531226e-05, + "loss": 0.1184, + "step": 72260 + }, + { + "epoch": 4.69, + "learning_rate": 1.8801933629474785e-05, + "loss": 0.1121, + "step": 72270 + }, + { + "epoch": 4.69, + "learning_rate": 1.876273843741834e-05, + "loss": 0.1428, + "step": 72280 + }, + { + "epoch": 4.69, + "learning_rate": 1.8723543245361903e-05, + "loss": 0.0894, + "step": 72290 + }, + { + "epoch": 4.69, + "learning_rate": 1.868434805330546e-05, + "loss": 0.0859, + "step": 72300 + }, + { + "epoch": 4.69, + "learning_rate": 1.8645152861249017e-05, + "loss": 0.0819, + "step": 72310 + }, + { + "epoch": 4.69, + "learning_rate": 1.860595766919258e-05, + "loss": 0.1187, + "step": 72320 + }, + { + "epoch": 4.69, + "learning_rate": 1.8566762477136135e-05, + "loss": 0.1433, + "step": 72330 + }, + { + "epoch": 4.69, + "learning_rate": 1.8527567285079694e-05, + "loss": 0.1259, + "step": 72340 + }, + { + "epoch": 4.7, + "learning_rate": 1.8488372093023256e-05, + "loss": 0.0775, + "step": 72350 + }, + { + "epoch": 4.7, + "learning_rate": 1.8449176900966812e-05, + "loss": 0.0978, + "step": 72360 + }, + { + "epoch": 4.7, + "learning_rate": 1.840998170891037e-05, + "loss": 0.0909, + "step": 72370 + }, + { + "epoch": 4.7, + "learning_rate": 1.8370786516853933e-05, + "loss": 0.1495, + "step": 72380 + }, + { + "epoch": 4.7, + "learning_rate": 1.833159132479749e-05, + "loss": 0.1116, + "step": 72390 + }, + { + "epoch": 4.7, + "learning_rate": 1.8292396132741047e-05, + "loss": 0.0787, + "step": 72400 + }, + { + "epoch": 4.7, + "learning_rate": 1.825320094068461e-05, + "loss": 0.0944, + "step": 72410 + }, + { + "epoch": 4.7, + "learning_rate": 1.8214005748628165e-05, + "loss": 0.1194, + "step": 72420 + }, + { + "epoch": 4.7, + "learning_rate": 1.8174810556571724e-05, + "loss": 0.1044, + "step": 72430 + }, + { + "epoch": 4.7, + "learning_rate": 1.8135615364515286e-05, + "loss": 0.1168, + "step": 72440 + }, + { + "epoch": 4.7, + "learning_rate": 1.8096420172458842e-05, + "loss": 0.0877, + "step": 72450 + }, + { + "epoch": 4.7, + "learning_rate": 1.80572249804024e-05, + "loss": 0.0756, + "step": 72460 + }, + { + "epoch": 4.7, + "learning_rate": 1.8018029788345963e-05, + "loss": 0.0999, + "step": 72470 + }, + { + "epoch": 4.7, + "learning_rate": 1.797883459628952e-05, + "loss": 0.0985, + "step": 72480 + }, + { + "epoch": 4.7, + "learning_rate": 1.7939639404233078e-05, + "loss": 0.1173, + "step": 72490 + }, + { + "epoch": 4.71, + "learning_rate": 1.790044421217664e-05, + "loss": 0.1098, + "step": 72500 + }, + { + "epoch": 4.71, + "learning_rate": 1.7861249020120195e-05, + "loss": 0.0832, + "step": 72510 + }, + { + "epoch": 4.71, + "learning_rate": 1.7822053828063754e-05, + "loss": 0.098, + "step": 72520 + }, + { + "epoch": 4.71, + "learning_rate": 1.7782858636007317e-05, + "loss": 0.1257, + "step": 72530 + }, + { + "epoch": 4.71, + "learning_rate": 1.7743663443950872e-05, + "loss": 0.091, + "step": 72540 + }, + { + "epoch": 4.71, + "learning_rate": 1.770446825189443e-05, + "loss": 0.0733, + "step": 72550 + }, + { + "epoch": 4.71, + "learning_rate": 1.7665273059837993e-05, + "loss": 0.099, + "step": 72560 + }, + { + "epoch": 4.71, + "learning_rate": 1.762607786778155e-05, + "loss": 0.0856, + "step": 72570 + }, + { + "epoch": 4.71, + "learning_rate": 1.7586882675725108e-05, + "loss": 0.131, + "step": 72580 + }, + { + "epoch": 4.71, + "learning_rate": 1.754768748366867e-05, + "loss": 0.1244, + "step": 72590 + }, + { + "epoch": 4.71, + "learning_rate": 1.7508492291612226e-05, + "loss": 0.0728, + "step": 72600 + }, + { + "epoch": 4.71, + "learning_rate": 1.7469297099555784e-05, + "loss": 0.0849, + "step": 72610 + }, + { + "epoch": 4.71, + "learning_rate": 1.7430101907499347e-05, + "loss": 0.125, + "step": 72620 + }, + { + "epoch": 4.71, + "learning_rate": 1.7390906715442902e-05, + "loss": 0.1218, + "step": 72630 + }, + { + "epoch": 4.71, + "learning_rate": 1.735171152338646e-05, + "loss": 0.1078, + "step": 72640 + }, + { + "epoch": 4.72, + "learning_rate": 1.7312516331330023e-05, + "loss": 0.0947, + "step": 72650 + }, + { + "epoch": 4.72, + "learning_rate": 1.7273321139273582e-05, + "loss": 0.1141, + "step": 72660 + }, + { + "epoch": 4.72, + "learning_rate": 1.7234125947217138e-05, + "loss": 0.1071, + "step": 72670 + }, + { + "epoch": 4.72, + "learning_rate": 1.71949307551607e-05, + "loss": 0.1643, + "step": 72680 + }, + { + "epoch": 4.72, + "learning_rate": 1.715573556310426e-05, + "loss": 0.1017, + "step": 72690 + }, + { + "epoch": 4.72, + "learning_rate": 1.7116540371047815e-05, + "loss": 0.0937, + "step": 72700 + }, + { + "epoch": 4.72, + "learning_rate": 1.7077345178991377e-05, + "loss": 0.1244, + "step": 72710 + }, + { + "epoch": 4.72, + "learning_rate": 1.7038149986934936e-05, + "loss": 0.0834, + "step": 72720 + }, + { + "epoch": 4.72, + "learning_rate": 1.699895479487849e-05, + "loss": 0.1057, + "step": 72730 + }, + { + "epoch": 4.72, + "learning_rate": 1.6959759602822054e-05, + "loss": 0.1231, + "step": 72740 + }, + { + "epoch": 4.72, + "learning_rate": 1.6920564410765612e-05, + "loss": 0.0815, + "step": 72750 + }, + { + "epoch": 4.72, + "learning_rate": 1.688136921870917e-05, + "loss": 0.0927, + "step": 72760 + }, + { + "epoch": 4.72, + "learning_rate": 1.684217402665273e-05, + "loss": 0.1062, + "step": 72770 + }, + { + "epoch": 4.72, + "learning_rate": 1.680297883459629e-05, + "loss": 0.1065, + "step": 72780 + }, + { + "epoch": 4.72, + "learning_rate": 1.6763783642539848e-05, + "loss": 0.0915, + "step": 72790 + }, + { + "epoch": 4.72, + "learning_rate": 1.6724588450483407e-05, + "loss": 0.0884, + "step": 72800 + }, + { + "epoch": 4.73, + "learning_rate": 1.6685393258426966e-05, + "loss": 0.0918, + "step": 72810 + }, + { + "epoch": 4.73, + "learning_rate": 1.6646198066370525e-05, + "loss": 0.1115, + "step": 72820 + }, + { + "epoch": 4.73, + "learning_rate": 1.6607002874314084e-05, + "loss": 0.1326, + "step": 72830 + }, + { + "epoch": 4.73, + "learning_rate": 1.6567807682257643e-05, + "loss": 0.1018, + "step": 72840 + }, + { + "epoch": 4.73, + "learning_rate": 1.65286124902012e-05, + "loss": 0.0775, + "step": 72850 + }, + { + "epoch": 4.73, + "learning_rate": 1.648941729814476e-05, + "loss": 0.1096, + "step": 72860 + }, + { + "epoch": 4.73, + "learning_rate": 1.645022210608832e-05, + "loss": 0.1172, + "step": 72870 + }, + { + "epoch": 4.73, + "learning_rate": 1.6411026914031878e-05, + "loss": 0.1247, + "step": 72880 + }, + { + "epoch": 4.73, + "learning_rate": 1.6371831721975437e-05, + "loss": 0.1415, + "step": 72890 + }, + { + "epoch": 4.73, + "learning_rate": 1.6332636529918996e-05, + "loss": 0.0878, + "step": 72900 + }, + { + "epoch": 4.73, + "learning_rate": 1.6293441337862555e-05, + "loss": 0.0941, + "step": 72910 + }, + { + "epoch": 4.73, + "learning_rate": 1.6254246145806114e-05, + "loss": 0.1208, + "step": 72920 + }, + { + "epoch": 4.73, + "learning_rate": 1.6215050953749673e-05, + "loss": 0.0985, + "step": 72930 + }, + { + "epoch": 4.73, + "learning_rate": 1.617585576169323e-05, + "loss": 0.1305, + "step": 72940 + }, + { + "epoch": 4.73, + "learning_rate": 1.613666056963679e-05, + "loss": 0.0909, + "step": 72950 + }, + { + "epoch": 4.74, + "learning_rate": 1.609746537758035e-05, + "loss": 0.0932, + "step": 72960 + }, + { + "epoch": 4.74, + "learning_rate": 1.6058270185523908e-05, + "loss": 0.1033, + "step": 72970 + }, + { + "epoch": 4.74, + "learning_rate": 1.6019074993467467e-05, + "loss": 0.0933, + "step": 72980 + }, + { + "epoch": 4.74, + "learning_rate": 1.5979879801411026e-05, + "loss": 0.1096, + "step": 72990 + }, + { + "epoch": 4.74, + "learning_rate": 1.5940684609354585e-05, + "loss": 0.0857, + "step": 73000 + }, + { + "epoch": 4.74, + "eval_loss": 0.1680546998977661, + "eval_runtime": 1022.7446, + "eval_samples_per_second": 15.241, + "eval_steps_per_second": 1.906, + "eval_wer": 0.16137652627539054, + "step": 73000 + }, + { + "epoch": 4.74, + "learning_rate": 1.5901489417298144e-05, + "loss": 0.0858, + "step": 73010 + }, + { + "epoch": 4.74, + "learning_rate": 1.5862294225241703e-05, + "loss": 0.1094, + "step": 73020 + }, + { + "epoch": 4.74, + "learning_rate": 1.5823099033185262e-05, + "loss": 0.1201, + "step": 73030 + }, + { + "epoch": 4.74, + "learning_rate": 1.578390384112882e-05, + "loss": 0.1049, + "step": 73040 + }, + { + "epoch": 4.74, + "learning_rate": 1.574470864907238e-05, + "loss": 0.0871, + "step": 73050 + }, + { + "epoch": 4.74, + "learning_rate": 1.570551345701594e-05, + "loss": 0.1093, + "step": 73060 + }, + { + "epoch": 4.74, + "learning_rate": 1.5666318264959497e-05, + "loss": 0.1052, + "step": 73070 + }, + { + "epoch": 4.74, + "learning_rate": 1.5627123072903056e-05, + "loss": 0.1112, + "step": 73080 + }, + { + "epoch": 4.74, + "learning_rate": 1.5587927880846615e-05, + "loss": 0.0954, + "step": 73090 + }, + { + "epoch": 4.74, + "learning_rate": 1.5548732688790174e-05, + "loss": 0.0682, + "step": 73100 + }, + { + "epoch": 4.74, + "learning_rate": 1.5509537496733733e-05, + "loss": 0.1195, + "step": 73110 + }, + { + "epoch": 4.75, + "learning_rate": 1.5470342304677292e-05, + "loss": 0.1075, + "step": 73120 + }, + { + "epoch": 4.75, + "learning_rate": 1.543114711262085e-05, + "loss": 0.124, + "step": 73130 + }, + { + "epoch": 4.75, + "learning_rate": 1.539195192056441e-05, + "loss": 0.1193, + "step": 73140 + }, + { + "epoch": 4.75, + "learning_rate": 1.535275672850797e-05, + "loss": 0.0687, + "step": 73150 + }, + { + "epoch": 4.75, + "learning_rate": 1.5313561536451527e-05, + "loss": 0.0837, + "step": 73160 + }, + { + "epoch": 4.75, + "learning_rate": 1.5274366344395086e-05, + "loss": 0.0905, + "step": 73170 + }, + { + "epoch": 4.75, + "learning_rate": 1.5235171152338645e-05, + "loss": 0.112, + "step": 73180 + }, + { + "epoch": 4.75, + "learning_rate": 1.5195975960282204e-05, + "loss": 0.1054, + "step": 73190 + }, + { + "epoch": 4.75, + "learning_rate": 1.5156780768225763e-05, + "loss": 0.0979, + "step": 73200 + }, + { + "epoch": 4.75, + "learning_rate": 1.5117585576169322e-05, + "loss": 0.0904, + "step": 73210 + }, + { + "epoch": 4.75, + "learning_rate": 1.5078390384112881e-05, + "loss": 0.0888, + "step": 73220 + }, + { + "epoch": 4.75, + "learning_rate": 1.503919519205644e-05, + "loss": 0.1359, + "step": 73230 + }, + { + "epoch": 4.75, + "learning_rate": 1.4999999999999999e-05, + "loss": 0.1021, + "step": 73240 + }, + { + "epoch": 4.75, + "learning_rate": 1.4960804807943558e-05, + "loss": 0.0915, + "step": 73250 + }, + { + "epoch": 4.75, + "learning_rate": 1.4921609615887116e-05, + "loss": 0.0994, + "step": 73260 + }, + { + "epoch": 4.76, + "learning_rate": 1.4882414423830677e-05, + "loss": 0.0959, + "step": 73270 + }, + { + "epoch": 4.76, + "learning_rate": 1.4843219231774234e-05, + "loss": 0.148, + "step": 73280 + }, + { + "epoch": 4.76, + "learning_rate": 1.4804024039717793e-05, + "loss": 0.103, + "step": 73290 + }, + { + "epoch": 4.76, + "learning_rate": 1.4764828847661354e-05, + "loss": 0.0744, + "step": 73300 + }, + { + "epoch": 4.76, + "learning_rate": 1.4725633655604911e-05, + "loss": 0.109, + "step": 73310 + }, + { + "epoch": 4.76, + "learning_rate": 1.468643846354847e-05, + "loss": 0.109, + "step": 73320 + }, + { + "epoch": 4.76, + "learning_rate": 1.464724327149203e-05, + "loss": 0.1565, + "step": 73330 + }, + { + "epoch": 4.76, + "learning_rate": 1.4608048079435588e-05, + "loss": 0.0847, + "step": 73340 + }, + { + "epoch": 4.76, + "learning_rate": 1.4568852887379147e-05, + "loss": 0.0851, + "step": 73350 + }, + { + "epoch": 4.76, + "learning_rate": 1.4529657695322707e-05, + "loss": 0.1131, + "step": 73360 + }, + { + "epoch": 4.76, + "learning_rate": 1.4490462503266264e-05, + "loss": 0.106, + "step": 73370 + }, + { + "epoch": 4.76, + "learning_rate": 1.4451267311209823e-05, + "loss": 0.1394, + "step": 73380 + }, + { + "epoch": 4.76, + "learning_rate": 1.4412072119153384e-05, + "loss": 0.0907, + "step": 73390 + }, + { + "epoch": 4.76, + "learning_rate": 1.4372876927096941e-05, + "loss": 0.0922, + "step": 73400 + }, + { + "epoch": 4.76, + "learning_rate": 1.43336817350405e-05, + "loss": 0.1064, + "step": 73410 + }, + { + "epoch": 4.77, + "learning_rate": 1.429448654298406e-05, + "loss": 0.0984, + "step": 73420 + }, + { + "epoch": 4.77, + "learning_rate": 1.4255291350927618e-05, + "loss": 0.1556, + "step": 73430 + }, + { + "epoch": 4.77, + "learning_rate": 1.4216096158871177e-05, + "loss": 0.1097, + "step": 73440 + }, + { + "epoch": 4.77, + "learning_rate": 1.4176900966814737e-05, + "loss": 0.0938, + "step": 73450 + }, + { + "epoch": 4.77, + "learning_rate": 1.4137705774758295e-05, + "loss": 0.0844, + "step": 73460 + }, + { + "epoch": 4.77, + "learning_rate": 1.4098510582701853e-05, + "loss": 0.0959, + "step": 73470 + }, + { + "epoch": 4.77, + "learning_rate": 1.4059315390645414e-05, + "loss": 0.1478, + "step": 73480 + }, + { + "epoch": 4.77, + "learning_rate": 1.4020120198588971e-05, + "loss": 0.0965, + "step": 73490 + }, + { + "epoch": 4.77, + "learning_rate": 1.398092500653253e-05, + "loss": 0.0735, + "step": 73500 + }, + { + "epoch": 4.77, + "learning_rate": 1.394172981447609e-05, + "loss": 0.077, + "step": 73510 + }, + { + "epoch": 4.77, + "learning_rate": 1.3902534622419648e-05, + "loss": 0.1008, + "step": 73520 + }, + { + "epoch": 4.77, + "learning_rate": 1.3863339430363207e-05, + "loss": 0.1105, + "step": 73530 + }, + { + "epoch": 4.77, + "learning_rate": 1.3824144238306767e-05, + "loss": 0.1044, + "step": 73540 + }, + { + "epoch": 4.77, + "learning_rate": 1.3784949046250325e-05, + "loss": 0.0921, + "step": 73550 + }, + { + "epoch": 4.77, + "learning_rate": 1.3745753854193884e-05, + "loss": 0.1022, + "step": 73560 + }, + { + "epoch": 4.77, + "learning_rate": 1.3706558662137444e-05, + "loss": 0.0783, + "step": 73570 + }, + { + "epoch": 4.78, + "learning_rate": 1.3667363470081003e-05, + "loss": 0.1304, + "step": 73580 + }, + { + "epoch": 4.78, + "learning_rate": 1.362816827802456e-05, + "loss": 0.1176, + "step": 73590 + }, + { + "epoch": 4.78, + "learning_rate": 1.3588973085968121e-05, + "loss": 0.0855, + "step": 73600 + }, + { + "epoch": 4.78, + "learning_rate": 1.354977789391168e-05, + "loss": 0.0939, + "step": 73610 + }, + { + "epoch": 4.78, + "learning_rate": 1.3510582701855237e-05, + "loss": 0.1186, + "step": 73620 + }, + { + "epoch": 4.78, + "learning_rate": 1.3471387509798798e-05, + "loss": 0.1131, + "step": 73630 + }, + { + "epoch": 4.78, + "learning_rate": 1.3432192317742356e-05, + "loss": 0.1103, + "step": 73640 + }, + { + "epoch": 4.78, + "learning_rate": 1.3392997125685914e-05, + "loss": 0.0865, + "step": 73650 + }, + { + "epoch": 4.78, + "learning_rate": 1.3353801933629474e-05, + "loss": 0.0883, + "step": 73660 + }, + { + "epoch": 4.78, + "learning_rate": 1.3314606741573033e-05, + "loss": 0.08, + "step": 73670 + }, + { + "epoch": 4.78, + "learning_rate": 1.327541154951659e-05, + "loss": 0.1271, + "step": 73680 + }, + { + "epoch": 4.78, + "learning_rate": 1.3236216357460151e-05, + "loss": 0.1038, + "step": 73690 + }, + { + "epoch": 4.78, + "learning_rate": 1.319702116540371e-05, + "loss": 0.0856, + "step": 73700 + }, + { + "epoch": 4.78, + "learning_rate": 1.3157825973347267e-05, + "loss": 0.0887, + "step": 73710 + }, + { + "epoch": 4.78, + "learning_rate": 1.3118630781290828e-05, + "loss": 0.1095, + "step": 73720 + }, + { + "epoch": 4.79, + "learning_rate": 1.3079435589234387e-05, + "loss": 0.1256, + "step": 73730 + }, + { + "epoch": 4.79, + "learning_rate": 1.3040240397177944e-05, + "loss": 0.1064, + "step": 73740 + }, + { + "epoch": 4.79, + "learning_rate": 1.3001045205121504e-05, + "loss": 0.0929, + "step": 73750 + }, + { + "epoch": 4.79, + "learning_rate": 1.2961850013065063e-05, + "loss": 0.1223, + "step": 73760 + }, + { + "epoch": 4.79, + "learning_rate": 1.292265482100862e-05, + "loss": 0.104, + "step": 73770 + }, + { + "epoch": 4.79, + "learning_rate": 1.2883459628952181e-05, + "loss": 0.1193, + "step": 73780 + }, + { + "epoch": 4.79, + "learning_rate": 1.284426443689574e-05, + "loss": 0.1221, + "step": 73790 + }, + { + "epoch": 4.79, + "learning_rate": 1.2805069244839297e-05, + "loss": 0.093, + "step": 73800 + }, + { + "epoch": 4.79, + "learning_rate": 1.2765874052782858e-05, + "loss": 0.0939, + "step": 73810 + }, + { + "epoch": 4.79, + "learning_rate": 1.2726678860726417e-05, + "loss": 0.0795, + "step": 73820 + }, + { + "epoch": 4.79, + "learning_rate": 1.2687483668669974e-05, + "loss": 0.1212, + "step": 73830 + }, + { + "epoch": 4.79, + "learning_rate": 1.2648288476613535e-05, + "loss": 0.1048, + "step": 73840 + }, + { + "epoch": 4.79, + "learning_rate": 1.2609093284557093e-05, + "loss": 0.1017, + "step": 73850 + }, + { + "epoch": 4.79, + "learning_rate": 1.256989809250065e-05, + "loss": 0.1049, + "step": 73860 + }, + { + "epoch": 4.79, + "learning_rate": 1.2530702900444211e-05, + "loss": 0.1062, + "step": 73870 + }, + { + "epoch": 4.79, + "learning_rate": 1.249150770838777e-05, + "loss": 0.1145, + "step": 73880 + }, + { + "epoch": 4.8, + "learning_rate": 1.2452312516331327e-05, + "loss": 0.1165, + "step": 73890 + }, + { + "epoch": 4.8, + "learning_rate": 1.2413117324274888e-05, + "loss": 0.0962, + "step": 73900 + }, + { + "epoch": 4.8, + "learning_rate": 1.2373922132218447e-05, + "loss": 0.0826, + "step": 73910 + }, + { + "epoch": 4.8, + "learning_rate": 1.2334726940162004e-05, + "loss": 0.1058, + "step": 73920 + }, + { + "epoch": 4.8, + "learning_rate": 1.2295531748105565e-05, + "loss": 0.0976, + "step": 73930 + }, + { + "epoch": 4.8, + "learning_rate": 1.2256336556049124e-05, + "loss": 0.0967, + "step": 73940 + }, + { + "epoch": 4.8, + "learning_rate": 1.221714136399268e-05, + "loss": 0.0775, + "step": 73950 + }, + { + "epoch": 4.8, + "learning_rate": 1.2177946171936241e-05, + "loss": 0.1075, + "step": 73960 + }, + { + "epoch": 4.8, + "learning_rate": 1.21387509798798e-05, + "loss": 0.1016, + "step": 73970 + }, + { + "epoch": 4.8, + "learning_rate": 1.2099555787823358e-05, + "loss": 0.1423, + "step": 73980 + }, + { + "epoch": 4.8, + "learning_rate": 1.2060360595766918e-05, + "loss": 0.1303, + "step": 73990 + }, + { + "epoch": 4.8, + "learning_rate": 1.2021165403710477e-05, + "loss": 0.0994, + "step": 74000 + }, + { + "epoch": 4.8, + "eval_loss": 0.16801908612251282, + "eval_runtime": 978.4984, + "eval_samples_per_second": 15.931, + "eval_steps_per_second": 1.992, + "eval_wer": 0.15955707807991357, + "step": 74000 + }, + { + "epoch": 4.8, + "learning_rate": 1.1981970211654034e-05, + "loss": 0.0956, + "step": 74010 + }, + { + "epoch": 4.8, + "learning_rate": 1.1942775019597595e-05, + "loss": 0.1271, + "step": 74020 + }, + { + "epoch": 4.8, + "learning_rate": 1.1903579827541154e-05, + "loss": 0.1167, + "step": 74030 + }, + { + "epoch": 4.81, + "learning_rate": 1.1864384635484714e-05, + "loss": 0.0999, + "step": 74040 + }, + { + "epoch": 4.81, + "learning_rate": 1.1825189443428272e-05, + "loss": 0.0772, + "step": 74050 + }, + { + "epoch": 4.81, + "learning_rate": 1.178599425137183e-05, + "loss": 0.0907, + "step": 74060 + }, + { + "epoch": 4.81, + "learning_rate": 1.1746799059315391e-05, + "loss": 0.1255, + "step": 74070 + }, + { + "epoch": 4.81, + "learning_rate": 1.1707603867258948e-05, + "loss": 0.0876, + "step": 74080 + }, + { + "epoch": 4.81, + "learning_rate": 1.1668408675202507e-05, + "loss": 0.1121, + "step": 74090 + }, + { + "epoch": 4.81, + "learning_rate": 1.1629213483146068e-05, + "loss": 0.1063, + "step": 74100 + }, + { + "epoch": 4.81, + "learning_rate": 1.1590018291089625e-05, + "loss": 0.1147, + "step": 74110 + }, + { + "epoch": 4.81, + "learning_rate": 1.1550823099033184e-05, + "loss": 0.1136, + "step": 74120 + }, + { + "epoch": 4.81, + "learning_rate": 1.1511627906976744e-05, + "loss": 0.1082, + "step": 74130 + }, + { + "epoch": 4.81, + "learning_rate": 1.1472432714920302e-05, + "loss": 0.1023, + "step": 74140 + }, + { + "epoch": 4.81, + "learning_rate": 1.143323752286386e-05, + "loss": 0.0914, + "step": 74150 + }, + { + "epoch": 4.81, + "learning_rate": 1.1394042330807421e-05, + "loss": 0.0859, + "step": 74160 + }, + { + "epoch": 4.81, + "learning_rate": 1.1358766657956624e-05, + "loss": 0.1122, + "step": 74170 + }, + { + "epoch": 4.81, + "learning_rate": 1.1319571465900181e-05, + "loss": 0.1478, + "step": 74180 + }, + { + "epoch": 4.82, + "learning_rate": 1.128037627384374e-05, + "loss": 0.119, + "step": 74190 + }, + { + "epoch": 4.82, + "learning_rate": 1.12411810817873e-05, + "loss": 0.0879, + "step": 74200 + }, + { + "epoch": 4.82, + "learning_rate": 1.1201985889730858e-05, + "loss": 0.0931, + "step": 74210 + }, + { + "epoch": 4.82, + "learning_rate": 1.1162790697674417e-05, + "loss": 0.107, + "step": 74220 + }, + { + "epoch": 4.82, + "learning_rate": 1.1123595505617977e-05, + "loss": 0.1561, + "step": 74230 + }, + { + "epoch": 4.82, + "learning_rate": 1.1084400313561534e-05, + "loss": 0.1426, + "step": 74240 + }, + { + "epoch": 4.82, + "learning_rate": 1.1045205121505095e-05, + "loss": 0.0926, + "step": 74250 + }, + { + "epoch": 4.82, + "learning_rate": 1.1006009929448654e-05, + "loss": 0.0897, + "step": 74260 + }, + { + "epoch": 4.82, + "learning_rate": 1.0966814737392211e-05, + "loss": 0.1239, + "step": 74270 + }, + { + "epoch": 4.82, + "learning_rate": 1.0927619545335772e-05, + "loss": 0.1182, + "step": 74280 + }, + { + "epoch": 4.82, + "learning_rate": 1.088842435327933e-05, + "loss": 0.1004, + "step": 74290 + }, + { + "epoch": 4.82, + "learning_rate": 1.0849229161222888e-05, + "loss": 0.0833, + "step": 74300 + }, + { + "epoch": 4.82, + "learning_rate": 1.0810033969166448e-05, + "loss": 0.092, + "step": 74310 + }, + { + "epoch": 4.82, + "learning_rate": 1.0770838777110007e-05, + "loss": 0.1039, + "step": 74320 + }, + { + "epoch": 4.82, + "learning_rate": 1.0731643585053565e-05, + "loss": 0.1226, + "step": 74330 + }, + { + "epoch": 4.82, + "learning_rate": 1.0692448392997125e-05, + "loss": 0.1327, + "step": 74340 + }, + { + "epoch": 4.83, + "learning_rate": 1.0653253200940684e-05, + "loss": 0.0888, + "step": 74350 + }, + { + "epoch": 4.83, + "learning_rate": 1.0614058008884243e-05, + "loss": 0.0884, + "step": 74360 + }, + { + "epoch": 4.83, + "learning_rate": 1.0574862816827802e-05, + "loss": 0.1086, + "step": 74370 + }, + { + "epoch": 4.83, + "learning_rate": 1.053566762477136e-05, + "loss": 0.1106, + "step": 74380 + }, + { + "epoch": 4.83, + "learning_rate": 1.049647243271492e-05, + "loss": 0.1167, + "step": 74390 + }, + { + "epoch": 4.83, + "learning_rate": 1.0457277240658479e-05, + "loss": 0.0865, + "step": 74400 + }, + { + "epoch": 4.83, + "learning_rate": 1.0418082048602037e-05, + "loss": 0.094, + "step": 74410 + }, + { + "epoch": 4.83, + "learning_rate": 1.0378886856545596e-05, + "loss": 0.089, + "step": 74420 + }, + { + "epoch": 4.83, + "learning_rate": 1.0339691664489155e-05, + "loss": 0.0954, + "step": 74430 + }, + { + "epoch": 4.83, + "learning_rate": 1.0300496472432714e-05, + "loss": 0.1156, + "step": 74440 + }, + { + "epoch": 4.83, + "learning_rate": 1.0261301280376273e-05, + "loss": 0.0888, + "step": 74450 + }, + { + "epoch": 4.83, + "learning_rate": 1.0222106088319832e-05, + "loss": 0.0866, + "step": 74460 + }, + { + "epoch": 4.83, + "learning_rate": 1.0182910896263391e-05, + "loss": 0.0976, + "step": 74470 + }, + { + "epoch": 4.83, + "learning_rate": 1.014371570420695e-05, + "loss": 0.1156, + "step": 74480 + }, + { + "epoch": 4.83, + "learning_rate": 1.0104520512150509e-05, + "loss": 0.0995, + "step": 74490 + }, + { + "epoch": 4.84, + "learning_rate": 1.0065325320094068e-05, + "loss": 0.0805, + "step": 74500 + }, + { + "epoch": 4.84, + "learning_rate": 1.0026130128037627e-05, + "loss": 0.1041, + "step": 74510 + }, + { + "epoch": 4.84, + "learning_rate": 9.986934935981185e-06, + "loss": 0.1011, + "step": 74520 + }, + { + "epoch": 4.84, + "learning_rate": 9.947739743924744e-06, + "loss": 0.1147, + "step": 74530 + }, + { + "epoch": 4.84, + "learning_rate": 9.908544551868303e-06, + "loss": 0.1059, + "step": 74540 + }, + { + "epoch": 4.84, + "learning_rate": 9.869349359811862e-06, + "loss": 0.0915, + "step": 74550 + }, + { + "epoch": 4.84, + "learning_rate": 9.830154167755421e-06, + "loss": 0.0897, + "step": 74560 + }, + { + "epoch": 4.84, + "learning_rate": 9.79095897569898e-06, + "loss": 0.115, + "step": 74570 + }, + { + "epoch": 4.84, + "learning_rate": 9.751763783642539e-06, + "loss": 0.1613, + "step": 74580 + }, + { + "epoch": 4.84, + "learning_rate": 9.712568591586098e-06, + "loss": 0.1012, + "step": 74590 + }, + { + "epoch": 4.84, + "learning_rate": 9.673373399529658e-06, + "loss": 0.096, + "step": 74600 + }, + { + "epoch": 4.84, + "learning_rate": 9.634178207473216e-06, + "loss": 0.0997, + "step": 74610 + }, + { + "epoch": 4.84, + "learning_rate": 9.594983015416774e-06, + "loss": 0.0844, + "step": 74620 + }, + { + "epoch": 4.84, + "learning_rate": 9.555787823360335e-06, + "loss": 0.116, + "step": 74630 + }, + { + "epoch": 4.84, + "learning_rate": 9.516592631303892e-06, + "loss": 0.097, + "step": 74640 + }, + { + "epoch": 4.84, + "learning_rate": 9.477397439247451e-06, + "loss": 0.0809, + "step": 74650 + }, + { + "epoch": 4.85, + "learning_rate": 9.438202247191012e-06, + "loss": 0.0765, + "step": 74660 + }, + { + "epoch": 4.85, + "learning_rate": 9.399007055134569e-06, + "loss": 0.0964, + "step": 74670 + }, + { + "epoch": 4.85, + "learning_rate": 9.359811863078128e-06, + "loss": 0.1344, + "step": 74680 + }, + { + "epoch": 4.85, + "learning_rate": 9.320616671021687e-06, + "loss": 0.1147, + "step": 74690 + }, + { + "epoch": 4.85, + "learning_rate": 9.281421478965246e-06, + "loss": 0.085, + "step": 74700 + }, + { + "epoch": 4.85, + "learning_rate": 9.242226286908805e-06, + "loss": 0.0797, + "step": 74710 + }, + { + "epoch": 4.85, + "learning_rate": 9.203031094852365e-06, + "loss": 0.1052, + "step": 74720 + }, + { + "epoch": 4.85, + "learning_rate": 9.163835902795922e-06, + "loss": 0.1391, + "step": 74730 + }, + { + "epoch": 4.85, + "learning_rate": 9.124640710739481e-06, + "loss": 0.1072, + "step": 74740 + }, + { + "epoch": 4.85, + "learning_rate": 9.085445518683042e-06, + "loss": 0.1107, + "step": 74750 + }, + { + "epoch": 4.85, + "learning_rate": 9.046250326626599e-06, + "loss": 0.0911, + "step": 74760 + }, + { + "epoch": 4.85, + "learning_rate": 9.007055134570158e-06, + "loss": 0.0979, + "step": 74770 + }, + { + "epoch": 4.85, + "learning_rate": 8.967859942513719e-06, + "loss": 0.115, + "step": 74780 + }, + { + "epoch": 4.85, + "learning_rate": 8.928664750457276e-06, + "loss": 0.1089, + "step": 74790 + }, + { + "epoch": 4.85, + "learning_rate": 8.889469558400835e-06, + "loss": 0.0807, + "step": 74800 + }, + { + "epoch": 4.86, + "learning_rate": 8.850274366344395e-06, + "loss": 0.0971, + "step": 74810 + }, + { + "epoch": 4.86, + "learning_rate": 8.811079174287953e-06, + "loss": 0.1193, + "step": 74820 + }, + { + "epoch": 4.86, + "learning_rate": 8.771883982231511e-06, + "loss": 0.1235, + "step": 74830 + }, + { + "epoch": 4.86, + "learning_rate": 8.732688790175072e-06, + "loss": 0.0886, + "step": 74840 + }, + { + "epoch": 4.86, + "learning_rate": 8.69349359811863e-06, + "loss": 0.0963, + "step": 74850 + }, + { + "epoch": 4.86, + "learning_rate": 8.654298406062188e-06, + "loss": 0.0962, + "step": 74860 + }, + { + "epoch": 4.86, + "learning_rate": 8.615103214005749e-06, + "loss": 0.0911, + "step": 74870 + }, + { + "epoch": 4.86, + "learning_rate": 8.575908021949306e-06, + "loss": 0.1238, + "step": 74880 + }, + { + "epoch": 4.86, + "learning_rate": 8.536712829892865e-06, + "loss": 0.1194, + "step": 74890 + }, + { + "epoch": 4.86, + "learning_rate": 8.497517637836425e-06, + "loss": 0.0754, + "step": 74900 + }, + { + "epoch": 4.86, + "learning_rate": 8.458322445779983e-06, + "loss": 0.1092, + "step": 74910 + }, + { + "epoch": 4.86, + "learning_rate": 8.419127253723542e-06, + "loss": 0.0934, + "step": 74920 + }, + { + "epoch": 4.86, + "learning_rate": 8.379932061667102e-06, + "loss": 0.1392, + "step": 74930 + }, + { + "epoch": 4.86, + "learning_rate": 8.34073686961066e-06, + "loss": 0.104, + "step": 74940 + }, + { + "epoch": 4.86, + "learning_rate": 8.30154167755422e-06, + "loss": 0.0949, + "step": 74950 + }, + { + "epoch": 4.87, + "learning_rate": 8.262346485497779e-06, + "loss": 0.0927, + "step": 74960 + }, + { + "epoch": 4.87, + "learning_rate": 8.223151293441336e-06, + "loss": 0.1211, + "step": 74970 + }, + { + "epoch": 4.87, + "learning_rate": 8.183956101384897e-06, + "loss": 0.1144, + "step": 74980 + }, + { + "epoch": 4.87, + "learning_rate": 8.144760909328456e-06, + "loss": 0.0922, + "step": 74990 + }, + { + "epoch": 4.87, + "learning_rate": 8.105565717272014e-06, + "loss": 0.0955, + "step": 75000 + }, + { + "epoch": 4.87, + "eval_loss": 0.16729450225830078, + "eval_runtime": 981.9936, + "eval_samples_per_second": 15.874, + "eval_steps_per_second": 1.985, + "eval_wer": 0.1590027149578542, + "step": 75000 + }, + { + "epoch": 4.87, + "learning_rate": 8.066370525215573e-06, + "loss": 0.1054, + "step": 75010 + }, + { + "epoch": 4.87, + "learning_rate": 8.027175333159132e-06, + "loss": 0.0833, + "step": 75020 + }, + { + "epoch": 4.87, + "learning_rate": 7.987980141102691e-06, + "loss": 0.0987, + "step": 75030 + }, + { + "epoch": 4.87, + "learning_rate": 7.94878494904625e-06, + "loss": 0.119, + "step": 75040 + }, + { + "epoch": 4.87, + "learning_rate": 7.909589756989809e-06, + "loss": 0.0776, + "step": 75050 + }, + { + "epoch": 4.87, + "learning_rate": 7.870394564933368e-06, + "loss": 0.0834, + "step": 75060 + }, + { + "epoch": 4.87, + "learning_rate": 7.831199372876927e-06, + "loss": 0.0951, + "step": 75070 + }, + { + "epoch": 4.87, + "learning_rate": 7.792004180820486e-06, + "loss": 0.1102, + "step": 75080 + }, + { + "epoch": 4.87, + "learning_rate": 7.752808988764045e-06, + "loss": 0.0918, + "step": 75090 + }, + { + "epoch": 4.87, + "learning_rate": 7.713613796707603e-06, + "loss": 0.08, + "step": 75100 + }, + { + "epoch": 4.87, + "learning_rate": 7.674418604651162e-06, + "loss": 0.0994, + "step": 75110 + }, + { + "epoch": 4.88, + "learning_rate": 7.635223412594721e-06, + "loss": 0.1159, + "step": 75120 + }, + { + "epoch": 4.88, + "learning_rate": 7.59602822053828e-06, + "loss": 0.1203, + "step": 75130 + }, + { + "epoch": 4.88, + "learning_rate": 7.556833028481839e-06, + "loss": 0.0998, + "step": 75140 + }, + { + "epoch": 4.88, + "learning_rate": 7.517637836425397e-06, + "loss": 0.0815, + "step": 75150 + }, + { + "epoch": 4.88, + "learning_rate": 7.478442644368957e-06, + "loss": 0.0954, + "step": 75160 + }, + { + "epoch": 4.88, + "learning_rate": 7.439247452312516e-06, + "loss": 0.0799, + "step": 75170 + }, + { + "epoch": 4.88, + "learning_rate": 7.400052260256075e-06, + "loss": 0.0994, + "step": 75180 + }, + { + "epoch": 4.88, + "learning_rate": 7.360857068199634e-06, + "loss": 0.1105, + "step": 75190 + }, + { + "epoch": 4.88, + "learning_rate": 7.3216618761431925e-06, + "loss": 0.0861, + "step": 75200 + }, + { + "epoch": 4.88, + "learning_rate": 7.2824666840867514e-06, + "loss": 0.0837, + "step": 75210 + }, + { + "epoch": 4.88, + "learning_rate": 7.24327149203031e-06, + "loss": 0.1172, + "step": 75220 + }, + { + "epoch": 4.88, + "learning_rate": 7.204076299973869e-06, + "loss": 0.1206, + "step": 75230 + }, + { + "epoch": 4.88, + "learning_rate": 7.164881107917429e-06, + "loss": 0.1185, + "step": 75240 + }, + { + "epoch": 4.88, + "learning_rate": 7.125685915860987e-06, + "loss": 0.0734, + "step": 75250 + }, + { + "epoch": 4.88, + "learning_rate": 7.086490723804546e-06, + "loss": 0.0817, + "step": 75260 + }, + { + "epoch": 4.89, + "learning_rate": 7.047295531748106e-06, + "loss": 0.0986, + "step": 75270 + }, + { + "epoch": 4.89, + "learning_rate": 7.008100339691664e-06, + "loss": 0.1354, + "step": 75280 + }, + { + "epoch": 4.89, + "learning_rate": 6.968905147635223e-06, + "loss": 0.1314, + "step": 75290 + }, + { + "epoch": 4.89, + "learning_rate": 6.929709955578782e-06, + "loss": 0.0781, + "step": 75300 + }, + { + "epoch": 4.89, + "learning_rate": 6.8905147635223405e-06, + "loss": 0.0927, + "step": 75310 + }, + { + "epoch": 4.89, + "learning_rate": 6.851319571465899e-06, + "loss": 0.1133, + "step": 75320 + }, + { + "epoch": 4.89, + "learning_rate": 6.812124379409459e-06, + "loss": 0.141, + "step": 75330 + }, + { + "epoch": 4.89, + "learning_rate": 6.772929187353017e-06, + "loss": 0.1336, + "step": 75340 + }, + { + "epoch": 4.89, + "learning_rate": 6.733733995296576e-06, + "loss": 0.0832, + "step": 75350 + }, + { + "epoch": 4.89, + "learning_rate": 6.694538803240136e-06, + "loss": 0.0997, + "step": 75360 + }, + { + "epoch": 4.89, + "learning_rate": 6.655343611183694e-06, + "loss": 0.1151, + "step": 75370 + }, + { + "epoch": 4.89, + "learning_rate": 6.616148419127253e-06, + "loss": 0.1103, + "step": 75380 + }, + { + "epoch": 4.89, + "learning_rate": 6.5769532270708125e-06, + "loss": 0.1034, + "step": 75390 + }, + { + "epoch": 4.89, + "learning_rate": 6.537758035014371e-06, + "loss": 0.1032, + "step": 75400 + }, + { + "epoch": 4.89, + "learning_rate": 6.49856284295793e-06, + "loss": 0.0824, + "step": 75410 + }, + { + "epoch": 4.89, + "learning_rate": 6.459367650901489e-06, + "loss": 0.0969, + "step": 75420 + }, + { + "epoch": 4.9, + "learning_rate": 6.420172458845047e-06, + "loss": 0.1235, + "step": 75430 + }, + { + "epoch": 4.9, + "learning_rate": 6.380977266788607e-06, + "loss": 0.1109, + "step": 75440 + }, + { + "epoch": 4.9, + "learning_rate": 6.341782074732166e-06, + "loss": 0.0976, + "step": 75450 + }, + { + "epoch": 4.9, + "learning_rate": 6.302586882675724e-06, + "loss": 0.0969, + "step": 75460 + }, + { + "epoch": 4.9, + "learning_rate": 6.263391690619284e-06, + "loss": 0.111, + "step": 75470 + }, + { + "epoch": 4.9, + "learning_rate": 6.224196498562843e-06, + "loss": 0.1397, + "step": 75480 + }, + { + "epoch": 4.9, + "learning_rate": 6.185001306506401e-06, + "loss": 0.0984, + "step": 75490 + }, + { + "epoch": 4.9, + "learning_rate": 6.1458061144499605e-06, + "loss": 0.0835, + "step": 75500 + }, + { + "epoch": 4.9, + "learning_rate": 6.106610922393519e-06, + "loss": 0.0992, + "step": 75510 + }, + { + "epoch": 4.9, + "learning_rate": 6.067415730337077e-06, + "loss": 0.1134, + "step": 75520 + }, + { + "epoch": 4.9, + "learning_rate": 6.028220538280637e-06, + "loss": 0.1151, + "step": 75530 + }, + { + "epoch": 4.9, + "learning_rate": 5.989025346224196e-06, + "loss": 0.1087, + "step": 75540 + }, + { + "epoch": 4.9, + "learning_rate": 5.949830154167754e-06, + "loss": 0.0769, + "step": 75550 + }, + { + "epoch": 4.9, + "learning_rate": 5.910634962111314e-06, + "loss": 0.0893, + "step": 75560 + }, + { + "epoch": 4.9, + "learning_rate": 5.871439770054873e-06, + "loss": 0.1104, + "step": 75570 + }, + { + "epoch": 4.91, + "learning_rate": 5.832244577998431e-06, + "loss": 0.1353, + "step": 75580 + }, + { + "epoch": 4.91, + "learning_rate": 5.793049385941991e-06, + "loss": 0.1319, + "step": 75590 + }, + { + "epoch": 4.91, + "learning_rate": 5.7538541938855495e-06, + "loss": 0.1003, + "step": 75600 + }, + { + "epoch": 4.91, + "learning_rate": 5.7146590018291075e-06, + "loss": 0.0831, + "step": 75610 + }, + { + "epoch": 4.91, + "learning_rate": 5.675463809772667e-06, + "loss": 0.1013, + "step": 75620 + }, + { + "epoch": 4.91, + "learning_rate": 5.636268617716226e-06, + "loss": 0.1212, + "step": 75630 + }, + { + "epoch": 4.91, + "learning_rate": 5.597073425659786e-06, + "loss": 0.106, + "step": 75640 + }, + { + "epoch": 4.91, + "learning_rate": 5.557878233603344e-06, + "loss": 0.0951, + "step": 75650 + }, + { + "epoch": 4.91, + "learning_rate": 5.518683041546903e-06, + "loss": 0.0981, + "step": 75660 + }, + { + "epoch": 4.91, + "learning_rate": 5.479487849490463e-06, + "loss": 0.0868, + "step": 75670 + }, + { + "epoch": 4.91, + "learning_rate": 5.440292657434021e-06, + "loss": 0.1064, + "step": 75680 + }, + { + "epoch": 4.91, + "learning_rate": 5.40109746537758e-06, + "loss": 0.1193, + "step": 75690 + }, + { + "epoch": 4.91, + "learning_rate": 5.361902273321139e-06, + "loss": 0.0864, + "step": 75700 + }, + { + "epoch": 4.91, + "learning_rate": 5.322707081264697e-06, + "loss": 0.1021, + "step": 75710 + }, + { + "epoch": 4.91, + "learning_rate": 5.283511889208256e-06, + "loss": 0.118, + "step": 75720 + }, + { + "epoch": 4.91, + "learning_rate": 5.244316697151816e-06, + "loss": 0.1108, + "step": 75730 + }, + { + "epoch": 4.92, + "learning_rate": 5.205121505095374e-06, + "loss": 0.0944, + "step": 75740 + }, + { + "epoch": 4.92, + "learning_rate": 5.165926313038933e-06, + "loss": 0.077, + "step": 75750 + }, + { + "epoch": 4.92, + "learning_rate": 5.126731120982493e-06, + "loss": 0.0948, + "step": 75760 + }, + { + "epoch": 4.92, + "learning_rate": 5.087535928926051e-06, + "loss": 0.1039, + "step": 75770 + }, + { + "epoch": 4.92, + "learning_rate": 5.04834073686961e-06, + "loss": 0.1321, + "step": 75780 + }, + { + "epoch": 4.92, + "learning_rate": 5.0091455448131695e-06, + "loss": 0.1002, + "step": 75790 + }, + { + "epoch": 4.92, + "learning_rate": 4.9699503527567275e-06, + "loss": 0.0901, + "step": 75800 + }, + { + "epoch": 4.92, + "learning_rate": 4.9307551607002864e-06, + "loss": 0.0944, + "step": 75810 + }, + { + "epoch": 4.92, + "learning_rate": 4.891559968643846e-06, + "loss": 0.125, + "step": 75820 + }, + { + "epoch": 4.92, + "learning_rate": 4.852364776587405e-06, + "loss": 0.1252, + "step": 75830 + }, + { + "epoch": 4.92, + "learning_rate": 4.813169584530963e-06, + "loss": 0.109, + "step": 75840 + }, + { + "epoch": 4.92, + "learning_rate": 4.773974392474523e-06, + "loss": 0.1035, + "step": 75850 + }, + { + "epoch": 4.92, + "learning_rate": 4.734779200418082e-06, + "loss": 0.0837, + "step": 75860 + }, + { + "epoch": 4.92, + "learning_rate": 4.695584008361641e-06, + "loss": 0.0944, + "step": 75870 + }, + { + "epoch": 4.92, + "learning_rate": 4.6563888163052e-06, + "loss": 0.113, + "step": 75880 + }, + { + "epoch": 4.93, + "learning_rate": 4.6171936242487585e-06, + "loss": 0.117, + "step": 75890 + }, + { + "epoch": 4.93, + "learning_rate": 4.577998432192317e-06, + "loss": 0.0846, + "step": 75900 + }, + { + "epoch": 4.93, + "learning_rate": 4.538803240135876e-06, + "loss": 0.0859, + "step": 75910 + }, + { + "epoch": 4.93, + "learning_rate": 4.499608048079435e-06, + "loss": 0.109, + "step": 75920 + }, + { + "epoch": 4.93, + "learning_rate": 4.460412856022994e-06, + "loss": 0.1426, + "step": 75930 + }, + { + "epoch": 4.93, + "learning_rate": 4.421217663966553e-06, + "loss": 0.0966, + "step": 75940 + }, + { + "epoch": 4.93, + "learning_rate": 4.382022471910112e-06, + "loss": 0.0986, + "step": 75950 + }, + { + "epoch": 4.93, + "learning_rate": 4.342827279853671e-06, + "loss": 0.0905, + "step": 75960 + }, + { + "epoch": 4.93, + "learning_rate": 4.30363208779723e-06, + "loss": 0.0696, + "step": 75970 + }, + { + "epoch": 4.93, + "learning_rate": 4.264436895740789e-06, + "loss": 0.1258, + "step": 75980 + }, + { + "epoch": 4.93, + "learning_rate": 4.2252417036843475e-06, + "loss": 0.1109, + "step": 75990 + }, + { + "epoch": 4.93, + "learning_rate": 4.1860465116279065e-06, + "loss": 0.093, + "step": 76000 + }, + { + "epoch": 4.93, + "eval_loss": 0.16745918989181519, + "eval_runtime": 1570.8748, + "eval_samples_per_second": 9.923, + "eval_steps_per_second": 1.241, + "eval_wer": 0.1598271524214297, + "step": 76000 + }, + { + "epoch": 4.93, + "learning_rate": 4.146851319571465e-06, + "loss": 0.1012, + "step": 76010 + }, + { + "epoch": 4.93, + "learning_rate": 4.107656127515025e-06, + "loss": 0.1016, + "step": 76020 + }, + { + "epoch": 4.93, + "learning_rate": 4.068460935458583e-06, + "loss": 0.1043, + "step": 76030 + }, + { + "epoch": 4.94, + "learning_rate": 4.029265743402142e-06, + "loss": 0.1079, + "step": 76040 + }, + { + "epoch": 4.94, + "learning_rate": 3.990070551345702e-06, + "loss": 0.0759, + "step": 76050 + }, + { + "epoch": 4.94, + "learning_rate": 3.95087535928926e-06, + "loss": 0.077, + "step": 76060 + }, + { + "epoch": 4.94, + "learning_rate": 3.911680167232819e-06, + "loss": 0.1152, + "step": 76070 + }, + { + "epoch": 4.94, + "learning_rate": 3.8724849751763785e-06, + "loss": 0.1551, + "step": 76080 + }, + { + "epoch": 4.94, + "learning_rate": 3.833289783119937e-06, + "loss": 0.1042, + "step": 76090 + }, + { + "epoch": 4.94, + "learning_rate": 3.794094591063496e-06, + "loss": 0.0773, + "step": 76100 + }, + { + "epoch": 4.94, + "learning_rate": 3.754899399007055e-06, + "loss": 0.078, + "step": 76110 + }, + { + "epoch": 4.94, + "learning_rate": 3.7157042069506133e-06, + "loss": 0.1262, + "step": 76120 + }, + { + "epoch": 4.94, + "learning_rate": 3.6765090148941726e-06, + "loss": 0.1213, + "step": 76130 + }, + { + "epoch": 4.94, + "learning_rate": 3.6373138228377315e-06, + "loss": 0.1027, + "step": 76140 + }, + { + "epoch": 4.94, + "learning_rate": 3.598118630781291e-06, + "loss": 0.0881, + "step": 76150 + }, + { + "epoch": 4.94, + "learning_rate": 3.5589234387248493e-06, + "loss": 0.0892, + "step": 76160 + }, + { + "epoch": 4.94, + "learning_rate": 3.5197282466684082e-06, + "loss": 0.0741, + "step": 76170 + }, + { + "epoch": 4.94, + "learning_rate": 3.4805330546119676e-06, + "loss": 0.1396, + "step": 76180 + }, + { + "epoch": 4.94, + "learning_rate": 3.441337862555526e-06, + "loss": 0.0873, + "step": 76190 + }, + { + "epoch": 4.95, + "learning_rate": 3.402142670499085e-06, + "loss": 0.095, + "step": 76200 + }, + { + "epoch": 4.95, + "learning_rate": 3.3629474784426443e-06, + "loss": 0.0957, + "step": 76210 + }, + { + "epoch": 4.95, + "learning_rate": 3.3237522863862027e-06, + "loss": 0.1147, + "step": 76220 + }, + { + "epoch": 4.95, + "learning_rate": 3.284557094329762e-06, + "loss": 0.1672, + "step": 76230 + }, + { + "epoch": 4.95, + "learning_rate": 3.245361902273321e-06, + "loss": 0.1234, + "step": 76240 + }, + { + "epoch": 4.95, + "learning_rate": 3.20616671021688e-06, + "loss": 0.0834, + "step": 76250 + }, + { + "epoch": 4.95, + "learning_rate": 3.1669715181604388e-06, + "loss": 0.1013, + "step": 76260 + }, + { + "epoch": 4.95, + "learning_rate": 3.1277763261039977e-06, + "loss": 0.1107, + "step": 76270 + }, + { + "epoch": 4.95, + "learning_rate": 3.0885811340475566e-06, + "loss": 0.144, + "step": 76280 + }, + { + "epoch": 4.95, + "learning_rate": 3.0493859419911155e-06, + "loss": 0.0972, + "step": 76290 + }, + { + "epoch": 4.95, + "learning_rate": 3.0101907499346744e-06, + "loss": 0.0849, + "step": 76300 + }, + { + "epoch": 4.95, + "learning_rate": 2.9709955578782337e-06, + "loss": 0.0847, + "step": 76310 + }, + { + "epoch": 4.95, + "learning_rate": 2.931800365821792e-06, + "loss": 0.1085, + "step": 76320 + }, + { + "epoch": 4.95, + "learning_rate": 2.892605173765351e-06, + "loss": 0.1081, + "step": 76330 + }, + { + "epoch": 4.95, + "learning_rate": 2.8534099817089104e-06, + "loss": 0.1515, + "step": 76340 + }, + { + "epoch": 4.96, + "learning_rate": 2.818134308858113e-06, + "loss": 0.0908, + "step": 76350 + }, + { + "epoch": 4.96, + "learning_rate": 2.778939116801672e-06, + "loss": 0.0716, + "step": 76360 + }, + { + "epoch": 4.96, + "learning_rate": 2.7397439247452313e-06, + "loss": 0.1108, + "step": 76370 + }, + { + "epoch": 4.96, + "learning_rate": 2.70054873268879e-06, + "loss": 0.1225, + "step": 76380 + }, + { + "epoch": 4.96, + "learning_rate": 2.6613535406323487e-06, + "loss": 0.1221, + "step": 76390 + }, + { + "epoch": 4.96, + "learning_rate": 2.622158348575908e-06, + "loss": 0.0905, + "step": 76400 + }, + { + "epoch": 4.96, + "learning_rate": 2.5829631565194665e-06, + "loss": 0.0899, + "step": 76410 + }, + { + "epoch": 4.96, + "learning_rate": 2.5437679644630254e-06, + "loss": 0.119, + "step": 76420 + }, + { + "epoch": 4.96, + "learning_rate": 2.5045727724065847e-06, + "loss": 0.1192, + "step": 76430 + }, + { + "epoch": 4.96, + "learning_rate": 2.4653775803501432e-06, + "loss": 0.1264, + "step": 76440 + }, + { + "epoch": 4.96, + "learning_rate": 2.4301019074993467e-06, + "loss": 0.1046, + "step": 76450 + }, + { + "epoch": 4.96, + "learning_rate": 2.3909067154429056e-06, + "loss": 0.0879, + "step": 76460 + }, + { + "epoch": 4.96, + "learning_rate": 2.351711523386464e-06, + "loss": 0.1008, + "step": 76470 + }, + { + "epoch": 4.96, + "learning_rate": 2.3125163313300235e-06, + "loss": 0.1603, + "step": 76480 + }, + { + "epoch": 4.96, + "learning_rate": 2.2733211392735824e-06, + "loss": 0.1006, + "step": 76490 + }, + { + "epoch": 4.96, + "learning_rate": 2.2341259472171413e-06, + "loss": 0.0744, + "step": 76500 + }, + { + "epoch": 4.97, + "learning_rate": 2.1949307551607e-06, + "loss": 0.0905, + "step": 76510 + }, + { + "epoch": 4.97, + "learning_rate": 2.155735563104259e-06, + "loss": 0.0977, + "step": 76520 + }, + { + "epoch": 4.97, + "learning_rate": 2.116540371047818e-06, + "loss": 0.1385, + "step": 76530 + }, + { + "epoch": 4.97, + "learning_rate": 2.077345178991377e-06, + "loss": 0.1313, + "step": 76540 + }, + { + "epoch": 4.97, + "learning_rate": 2.0381499869349358e-06, + "loss": 0.0925, + "step": 76550 + }, + { + "epoch": 4.97, + "learning_rate": 1.9989547948784947e-06, + "loss": 0.0984, + "step": 76560 + }, + { + "epoch": 4.97, + "learning_rate": 1.9597596028220536e-06, + "loss": 0.1112, + "step": 76570 + }, + { + "epoch": 4.97, + "learning_rate": 1.9205644107656125e-06, + "loss": 0.1102, + "step": 76580 + }, + { + "epoch": 4.97, + "learning_rate": 1.8813692187091714e-06, + "loss": 0.123, + "step": 76590 + }, + { + "epoch": 4.97, + "learning_rate": 1.8421740266527305e-06, + "loss": 0.0966, + "step": 76600 + }, + { + "epoch": 4.97, + "learning_rate": 1.8029788345962894e-06, + "loss": 0.0986, + "step": 76610 + }, + { + "epoch": 4.97, + "learning_rate": 1.763783642539848e-06, + "loss": 0.1064, + "step": 76620 + }, + { + "epoch": 4.97, + "learning_rate": 1.7245884504834072e-06, + "loss": 0.1138, + "step": 76630 + }, + { + "epoch": 4.97, + "learning_rate": 1.6853932584269661e-06, + "loss": 0.1093, + "step": 76640 + }, + { + "epoch": 4.97, + "learning_rate": 1.6461980663705252e-06, + "loss": 0.0956, + "step": 76650 + }, + { + "epoch": 4.98, + "learning_rate": 1.607002874314084e-06, + "loss": 0.0808, + "step": 76660 + }, + { + "epoch": 4.98, + "learning_rate": 1.5678076822576428e-06, + "loss": 0.0977, + "step": 76670 + }, + { + "epoch": 4.98, + "learning_rate": 1.528612490201202e-06, + "loss": 0.1098, + "step": 76680 + }, + { + "epoch": 4.98, + "learning_rate": 1.4894172981447608e-06, + "loss": 0.118, + "step": 76690 + }, + { + "epoch": 4.98, + "learning_rate": 1.4502221060883197e-06, + "loss": 0.0791, + "step": 76700 + }, + { + "epoch": 4.98, + "learning_rate": 1.4110269140318786e-06, + "loss": 0.0872, + "step": 76710 + }, + { + "epoch": 4.98, + "learning_rate": 1.3718317219754375e-06, + "loss": 0.0909, + "step": 76720 + }, + { + "epoch": 4.98, + "learning_rate": 1.3326365299189967e-06, + "loss": 0.1113, + "step": 76730 + }, + { + "epoch": 4.98, + "learning_rate": 1.2934413378625554e-06, + "loss": 0.0853, + "step": 76740 + }, + { + "epoch": 4.98, + "learning_rate": 1.2542461458061143e-06, + "loss": 0.076, + "step": 76750 + }, + { + "epoch": 4.98, + "learning_rate": 1.2150509537496734e-06, + "loss": 0.0858, + "step": 76760 + }, + { + "epoch": 4.98, + "learning_rate": 1.175855761693232e-06, + "loss": 0.1171, + "step": 76770 + }, + { + "epoch": 4.98, + "learning_rate": 1.1366605696367912e-06, + "loss": 0.1199, + "step": 76780 + }, + { + "epoch": 4.98, + "learning_rate": 1.09746537758035e-06, + "loss": 0.0962, + "step": 76790 + }, + { + "epoch": 4.98, + "learning_rate": 1.058270185523909e-06, + "loss": 0.0802, + "step": 76800 + }, + { + "epoch": 4.99, + "learning_rate": 1.0190749934674679e-06, + "loss": 0.0998, + "step": 76810 + }, + { + "epoch": 4.99, + "learning_rate": 9.798798014110268e-07, + "loss": 0.0959, + "step": 76820 + }, + { + "epoch": 4.99, + "learning_rate": 9.406846093545857e-07, + "loss": 0.138, + "step": 76830 + }, + { + "epoch": 4.99, + "learning_rate": 9.014894172981447e-07, + "loss": 0.0913, + "step": 76840 + }, + { + "epoch": 4.99, + "learning_rate": 8.622942252417036e-07, + "loss": 0.0646, + "step": 76850 + }, + { + "epoch": 4.99, + "learning_rate": 8.230990331852626e-07, + "loss": 0.0772, + "step": 76860 + }, + { + "epoch": 4.99, + "learning_rate": 7.839038411288214e-07, + "loss": 0.0868, + "step": 76870 + }, + { + "epoch": 4.99, + "learning_rate": 7.447086490723804e-07, + "loss": 0.1398, + "step": 76880 + }, + { + "epoch": 4.99, + "learning_rate": 7.055134570159393e-07, + "loss": 0.1086, + "step": 76890 + }, + { + "epoch": 4.99, + "learning_rate": 6.663182649594983e-07, + "loss": 0.1092, + "step": 76900 + }, + { + "epoch": 4.99, + "learning_rate": 6.271230729030571e-07, + "loss": 0.0807, + "step": 76910 + }, + { + "epoch": 4.99, + "learning_rate": 5.87927880846616e-07, + "loss": 0.0885, + "step": 76920 + }, + { + "epoch": 4.99, + "learning_rate": 5.48732688790175e-07, + "loss": 0.1281, + "step": 76930 + }, + { + "epoch": 4.99, + "learning_rate": 5.095374967337339e-07, + "loss": 0.0872, + "step": 76940 + }, + { + "epoch": 4.99, + "learning_rate": 4.7034230467729285e-07, + "loss": 0.0792, + "step": 76950 + }, + { + "epoch": 4.99, + "learning_rate": 4.311471126208518e-07, + "loss": 0.1116, + "step": 76960 + }, + { + "epoch": 5.0, + "learning_rate": 3.919519205644107e-07, + "loss": 0.1004, + "step": 76970 + }, + { + "epoch": 5.0, + "learning_rate": 3.5275672850796966e-07, + "loss": 0.1303, + "step": 76980 + }, + { + "epoch": 5.0, + "learning_rate": 3.1356153645152856e-07, + "loss": 0.1369, + "step": 76990 + }, + { + "epoch": 5.0, + "learning_rate": 2.743663443950875e-07, + "loss": 0.0796, + "step": 77000 + }, + { + "epoch": 5.0, + "eval_loss": 0.1670677214860916, + "eval_runtime": 1632.0161, + "eval_samples_per_second": 9.551, + "eval_steps_per_second": 1.194, + "eval_wer": 0.15941493368964194, + "step": 77000 + } + ], + "max_steps": 77040, + "num_train_epochs": 5, + "total_flos": 2.1833194219936534e+20, + "trial_name": null, + "trial_params": null +}