whisper-small-en / trainer_state.json
lorenzoncina's picture
End of training
1f6eb94
{
"best_metric": 13.058509783761204,
"best_model_checkpoint": "./checkpoint-10000",
"epoch": 1.0,
"global_step": 10000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 4.4e-07,
"loss": 1.8084,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.400000000000001e-07,
"loss": 1.6156,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.44e-06,
"loss": 1.1067,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.94e-06,
"loss": 0.6228,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.45,
"step": 125
},
{
"epoch": 0.01,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.365,
"step": 150
},
{
"epoch": 0.02,
"learning_rate": 3.44e-06,
"loss": 0.3328,
"step": 175
},
{
"epoch": 0.02,
"learning_rate": 3.94e-06,
"loss": 0.3055,
"step": 200
},
{
"epoch": 0.02,
"learning_rate": 4.440000000000001e-06,
"loss": 0.3007,
"step": 225
},
{
"epoch": 0.03,
"learning_rate": 4.94e-06,
"loss": 0.2964,
"step": 250
},
{
"epoch": 0.03,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.2899,
"step": 275
},
{
"epoch": 0.03,
"learning_rate": 5.94e-06,
"loss": 0.2651,
"step": 300
},
{
"epoch": 0.03,
"learning_rate": 6.440000000000001e-06,
"loss": 0.2554,
"step": 325
},
{
"epoch": 0.04,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.2672,
"step": 350
},
{
"epoch": 0.04,
"learning_rate": 7.440000000000001e-06,
"loss": 0.2646,
"step": 375
},
{
"epoch": 0.04,
"learning_rate": 7.94e-06,
"loss": 0.2632,
"step": 400
},
{
"epoch": 0.04,
"learning_rate": 8.44e-06,
"loss": 0.2372,
"step": 425
},
{
"epoch": 0.04,
"learning_rate": 8.94e-06,
"loss": 0.2129,
"step": 450
},
{
"epoch": 0.05,
"learning_rate": 9.440000000000001e-06,
"loss": 0.2041,
"step": 475
},
{
"epoch": 0.05,
"learning_rate": 9.940000000000001e-06,
"loss": 0.2196,
"step": 500
},
{
"epoch": 0.05,
"learning_rate": 9.976842105263158e-06,
"loss": 0.2256,
"step": 525
},
{
"epoch": 0.06,
"learning_rate": 9.950526315789475e-06,
"loss": 0.2331,
"step": 550
},
{
"epoch": 0.06,
"learning_rate": 9.92421052631579e-06,
"loss": 0.2268,
"step": 575
},
{
"epoch": 0.06,
"learning_rate": 9.897894736842107e-06,
"loss": 0.2169,
"step": 600
},
{
"epoch": 0.06,
"learning_rate": 9.871578947368422e-06,
"loss": 0.234,
"step": 625
},
{
"epoch": 0.07,
"learning_rate": 9.845263157894738e-06,
"loss": 0.3794,
"step": 650
},
{
"epoch": 0.07,
"learning_rate": 9.818947368421053e-06,
"loss": 0.3477,
"step": 675
},
{
"epoch": 0.07,
"learning_rate": 9.79263157894737e-06,
"loss": 0.389,
"step": 700
},
{
"epoch": 0.07,
"learning_rate": 9.766315789473685e-06,
"loss": 0.3686,
"step": 725
},
{
"epoch": 0.07,
"learning_rate": 9.74e-06,
"loss": 0.3992,
"step": 750
},
{
"epoch": 0.08,
"learning_rate": 9.713684210526317e-06,
"loss": 0.3385,
"step": 775
},
{
"epoch": 0.08,
"learning_rate": 9.687368421052632e-06,
"loss": 0.2639,
"step": 800
},
{
"epoch": 0.08,
"learning_rate": 9.661052631578948e-06,
"loss": 0.2185,
"step": 825
},
{
"epoch": 0.09,
"learning_rate": 9.634736842105265e-06,
"loss": 0.2135,
"step": 850
},
{
"epoch": 0.09,
"learning_rate": 9.60842105263158e-06,
"loss": 0.2173,
"step": 875
},
{
"epoch": 0.09,
"learning_rate": 9.582105263157897e-06,
"loss": 0.205,
"step": 900
},
{
"epoch": 0.09,
"learning_rate": 9.555789473684211e-06,
"loss": 0.1955,
"step": 925
},
{
"epoch": 0.1,
"learning_rate": 9.529473684210528e-06,
"loss": 0.1661,
"step": 950
},
{
"epoch": 0.1,
"learning_rate": 9.503157894736843e-06,
"loss": 0.1495,
"step": 975
},
{
"epoch": 0.1,
"learning_rate": 9.476842105263158e-06,
"loss": 0.1537,
"step": 1000
},
{
"epoch": 0.1,
"eval_loss": 0.44048482179641724,
"eval_runtime": 1796.3495,
"eval_samples_per_second": 9.104,
"eval_steps_per_second": 0.569,
"eval_wer": 17.927577675673938,
"step": 1000
},
{
"epoch": 0.1,
"learning_rate": 9.450526315789475e-06,
"loss": 0.1677,
"step": 1025
},
{
"epoch": 0.1,
"learning_rate": 9.42421052631579e-06,
"loss": 0.1606,
"step": 1050
},
{
"epoch": 0.11,
"learning_rate": 9.397894736842106e-06,
"loss": 0.1391,
"step": 1075
},
{
"epoch": 0.11,
"learning_rate": 9.371578947368421e-06,
"loss": 0.1077,
"step": 1100
},
{
"epoch": 0.11,
"learning_rate": 9.345263157894738e-06,
"loss": 0.0953,
"step": 1125
},
{
"epoch": 0.12,
"learning_rate": 9.318947368421053e-06,
"loss": 0.1508,
"step": 1150
},
{
"epoch": 0.12,
"learning_rate": 9.292631578947368e-06,
"loss": 0.1422,
"step": 1175
},
{
"epoch": 0.12,
"learning_rate": 9.266315789473685e-06,
"loss": 0.1471,
"step": 1200
},
{
"epoch": 0.12,
"learning_rate": 9.240000000000001e-06,
"loss": 0.1589,
"step": 1225
},
{
"epoch": 0.12,
"learning_rate": 9.213684210526316e-06,
"loss": 0.1548,
"step": 1250
},
{
"epoch": 0.13,
"learning_rate": 9.187368421052633e-06,
"loss": 0.2132,
"step": 1275
},
{
"epoch": 0.13,
"learning_rate": 9.161052631578948e-06,
"loss": 0.2751,
"step": 1300
},
{
"epoch": 0.13,
"learning_rate": 9.134736842105265e-06,
"loss": 0.3139,
"step": 1325
},
{
"epoch": 0.14,
"learning_rate": 9.10842105263158e-06,
"loss": 0.3281,
"step": 1350
},
{
"epoch": 0.14,
"learning_rate": 9.082105263157896e-06,
"loss": 0.2389,
"step": 1375
},
{
"epoch": 0.14,
"learning_rate": 9.055789473684211e-06,
"loss": 0.2151,
"step": 1400
},
{
"epoch": 0.14,
"learning_rate": 9.029473684210526e-06,
"loss": 0.2243,
"step": 1425
},
{
"epoch": 0.14,
"learning_rate": 9.003157894736843e-06,
"loss": 0.3187,
"step": 1450
},
{
"epoch": 0.15,
"learning_rate": 8.976842105263158e-06,
"loss": 0.4891,
"step": 1475
},
{
"epoch": 0.15,
"learning_rate": 8.950526315789474e-06,
"loss": 0.2386,
"step": 1500
},
{
"epoch": 0.15,
"learning_rate": 8.92421052631579e-06,
"loss": 0.1652,
"step": 1525
},
{
"epoch": 0.15,
"learning_rate": 8.897894736842106e-06,
"loss": 0.2075,
"step": 1550
},
{
"epoch": 0.16,
"learning_rate": 8.871578947368423e-06,
"loss": 0.2272,
"step": 1575
},
{
"epoch": 0.16,
"learning_rate": 8.845263157894738e-06,
"loss": 0.2644,
"step": 1600
},
{
"epoch": 0.16,
"learning_rate": 8.818947368421054e-06,
"loss": 0.427,
"step": 1625
},
{
"epoch": 0.17,
"learning_rate": 8.79263157894737e-06,
"loss": 0.2881,
"step": 1650
},
{
"epoch": 0.17,
"learning_rate": 8.766315789473684e-06,
"loss": 0.2437,
"step": 1675
},
{
"epoch": 0.17,
"learning_rate": 8.740000000000001e-06,
"loss": 0.3028,
"step": 1700
},
{
"epoch": 0.17,
"learning_rate": 8.713684210526316e-06,
"loss": 0.2817,
"step": 1725
},
{
"epoch": 0.17,
"learning_rate": 8.687368421052633e-06,
"loss": 0.2488,
"step": 1750
},
{
"epoch": 0.18,
"learning_rate": 8.661052631578948e-06,
"loss": 0.2944,
"step": 1775
},
{
"epoch": 0.18,
"learning_rate": 8.634736842105264e-06,
"loss": 0.267,
"step": 1800
},
{
"epoch": 0.18,
"learning_rate": 8.60842105263158e-06,
"loss": 0.2292,
"step": 1825
},
{
"epoch": 0.18,
"learning_rate": 8.582105263157894e-06,
"loss": 0.2118,
"step": 1850
},
{
"epoch": 0.19,
"learning_rate": 8.555789473684213e-06,
"loss": 0.2364,
"step": 1875
},
{
"epoch": 0.19,
"learning_rate": 8.529473684210528e-06,
"loss": 0.2114,
"step": 1900
},
{
"epoch": 0.19,
"learning_rate": 8.503157894736843e-06,
"loss": 0.1788,
"step": 1925
},
{
"epoch": 0.2,
"learning_rate": 8.47684210526316e-06,
"loss": 0.2134,
"step": 1950
},
{
"epoch": 0.2,
"learning_rate": 8.450526315789474e-06,
"loss": 0.2168,
"step": 1975
},
{
"epoch": 0.2,
"learning_rate": 8.42421052631579e-06,
"loss": 0.2378,
"step": 2000
},
{
"epoch": 0.2,
"eval_loss": 0.4008975923061371,
"eval_runtime": 1762.3911,
"eval_samples_per_second": 9.279,
"eval_steps_per_second": 0.58,
"eval_wer": 15.988818558622242,
"step": 2000
},
{
"epoch": 0.2,
"learning_rate": 8.397894736842106e-06,
"loss": 0.2252,
"step": 2025
},
{
"epoch": 0.2,
"learning_rate": 8.371578947368422e-06,
"loss": 0.2883,
"step": 2050
},
{
"epoch": 0.21,
"learning_rate": 8.345263157894737e-06,
"loss": 0.3074,
"step": 2075
},
{
"epoch": 0.21,
"learning_rate": 8.318947368421052e-06,
"loss": 0.239,
"step": 2100
},
{
"epoch": 0.21,
"learning_rate": 8.292631578947369e-06,
"loss": 0.3065,
"step": 2125
},
{
"epoch": 0.21,
"learning_rate": 8.266315789473684e-06,
"loss": 0.1591,
"step": 2150
},
{
"epoch": 0.22,
"learning_rate": 8.24e-06,
"loss": 0.1774,
"step": 2175
},
{
"epoch": 0.22,
"learning_rate": 8.213684210526316e-06,
"loss": 0.2075,
"step": 2200
},
{
"epoch": 0.22,
"learning_rate": 8.187368421052632e-06,
"loss": 0.2126,
"step": 2225
},
{
"epoch": 0.23,
"learning_rate": 8.161052631578949e-06,
"loss": 0.1824,
"step": 2250
},
{
"epoch": 0.23,
"learning_rate": 8.134736842105264e-06,
"loss": 0.2051,
"step": 2275
},
{
"epoch": 0.23,
"learning_rate": 8.10842105263158e-06,
"loss": 0.2013,
"step": 2300
},
{
"epoch": 0.23,
"learning_rate": 8.082105263157896e-06,
"loss": 0.181,
"step": 2325
},
{
"epoch": 0.23,
"learning_rate": 8.05578947368421e-06,
"loss": 0.2462,
"step": 2350
},
{
"epoch": 0.24,
"learning_rate": 8.029473684210527e-06,
"loss": 0.2595,
"step": 2375
},
{
"epoch": 0.24,
"learning_rate": 8.003157894736842e-06,
"loss": 0.2019,
"step": 2400
},
{
"epoch": 0.24,
"learning_rate": 7.976842105263159e-06,
"loss": 0.1712,
"step": 2425
},
{
"epoch": 0.24,
"learning_rate": 7.950526315789474e-06,
"loss": 0.1848,
"step": 2450
},
{
"epoch": 0.25,
"learning_rate": 7.92421052631579e-06,
"loss": 0.1812,
"step": 2475
},
{
"epoch": 0.25,
"learning_rate": 7.897894736842106e-06,
"loss": 0.1555,
"step": 2500
},
{
"epoch": 0.25,
"learning_rate": 7.87157894736842e-06,
"loss": 0.1747,
"step": 2525
},
{
"epoch": 0.26,
"learning_rate": 7.845263157894739e-06,
"loss": 0.1798,
"step": 2550
},
{
"epoch": 0.26,
"learning_rate": 7.818947368421054e-06,
"loss": 0.2106,
"step": 2575
},
{
"epoch": 0.26,
"learning_rate": 7.792631578947369e-06,
"loss": 0.2513,
"step": 2600
},
{
"epoch": 0.26,
"learning_rate": 7.766315789473685e-06,
"loss": 0.2869,
"step": 2625
},
{
"epoch": 0.27,
"learning_rate": 7.74e-06,
"loss": 0.3266,
"step": 2650
},
{
"epoch": 0.27,
"learning_rate": 7.713684210526317e-06,
"loss": 0.2484,
"step": 2675
},
{
"epoch": 0.27,
"learning_rate": 7.687368421052632e-06,
"loss": 0.2938,
"step": 2700
},
{
"epoch": 0.27,
"learning_rate": 7.661052631578949e-06,
"loss": 0.2439,
"step": 2725
},
{
"epoch": 0.28,
"learning_rate": 7.634736842105264e-06,
"loss": 0.2393,
"step": 2750
},
{
"epoch": 0.28,
"learning_rate": 7.6084210526315795e-06,
"loss": 0.1989,
"step": 2775
},
{
"epoch": 0.28,
"learning_rate": 7.582105263157895e-06,
"loss": 0.1805,
"step": 2800
},
{
"epoch": 0.28,
"learning_rate": 7.555789473684211e-06,
"loss": 0.1616,
"step": 2825
},
{
"epoch": 0.28,
"learning_rate": 7.529473684210526e-06,
"loss": 0.1994,
"step": 2850
},
{
"epoch": 0.29,
"learning_rate": 7.503157894736842e-06,
"loss": 0.1989,
"step": 2875
},
{
"epoch": 0.29,
"learning_rate": 7.476842105263159e-06,
"loss": 0.1764,
"step": 2900
},
{
"epoch": 0.29,
"learning_rate": 7.4505263157894744e-06,
"loss": 0.1845,
"step": 2925
},
{
"epoch": 0.29,
"learning_rate": 7.42421052631579e-06,
"loss": 0.2311,
"step": 2950
},
{
"epoch": 0.3,
"learning_rate": 7.397894736842106e-06,
"loss": 0.2376,
"step": 2975
},
{
"epoch": 0.3,
"learning_rate": 7.371578947368422e-06,
"loss": 0.1709,
"step": 3000
},
{
"epoch": 0.3,
"eval_loss": 0.385152131319046,
"eval_runtime": 1741.3169,
"eval_samples_per_second": 9.392,
"eval_steps_per_second": 0.587,
"eval_wer": 15.495292870224592,
"step": 3000
},
{
"epoch": 0.3,
"learning_rate": 7.345263157894738e-06,
"loss": 0.2468,
"step": 3025
},
{
"epoch": 0.3,
"learning_rate": 7.3189473684210535e-06,
"loss": 0.3145,
"step": 3050
},
{
"epoch": 0.31,
"learning_rate": 7.292631578947369e-06,
"loss": 0.3232,
"step": 3075
},
{
"epoch": 0.31,
"learning_rate": 7.266315789473684e-06,
"loss": 0.2161,
"step": 3100
},
{
"epoch": 0.31,
"learning_rate": 7.24e-06,
"loss": 0.4712,
"step": 3125
},
{
"epoch": 0.32,
"learning_rate": 7.213684210526316e-06,
"loss": 0.8535,
"step": 3150
},
{
"epoch": 0.32,
"learning_rate": 7.187368421052632e-06,
"loss": 0.4093,
"step": 3175
},
{
"epoch": 0.32,
"learning_rate": 7.161052631578948e-06,
"loss": 0.2704,
"step": 3200
},
{
"epoch": 0.32,
"learning_rate": 7.134736842105264e-06,
"loss": 0.2258,
"step": 3225
},
{
"epoch": 0.33,
"learning_rate": 7.10842105263158e-06,
"loss": 0.236,
"step": 3250
},
{
"epoch": 0.33,
"learning_rate": 7.082105263157896e-06,
"loss": 0.2873,
"step": 3275
},
{
"epoch": 0.33,
"learning_rate": 7.055789473684212e-06,
"loss": 0.325,
"step": 3300
},
{
"epoch": 0.33,
"learning_rate": 7.0294736842105275e-06,
"loss": 0.2195,
"step": 3325
},
{
"epoch": 0.34,
"learning_rate": 7.0031578947368425e-06,
"loss": 0.2137,
"step": 3350
},
{
"epoch": 0.34,
"learning_rate": 6.976842105263158e-06,
"loss": 0.2159,
"step": 3375
},
{
"epoch": 0.34,
"learning_rate": 6.950526315789474e-06,
"loss": 0.2478,
"step": 3400
},
{
"epoch": 0.34,
"learning_rate": 6.92421052631579e-06,
"loss": 0.2416,
"step": 3425
},
{
"epoch": 0.34,
"learning_rate": 6.897894736842106e-06,
"loss": 0.218,
"step": 3450
},
{
"epoch": 0.35,
"learning_rate": 6.871578947368422e-06,
"loss": 0.1614,
"step": 3475
},
{
"epoch": 0.35,
"learning_rate": 6.845263157894737e-06,
"loss": 0.1286,
"step": 3500
},
{
"epoch": 0.35,
"learning_rate": 6.818947368421052e-06,
"loss": 0.2231,
"step": 3525
},
{
"epoch": 0.35,
"learning_rate": 6.792631578947368e-06,
"loss": 0.2648,
"step": 3550
},
{
"epoch": 0.36,
"learning_rate": 6.766315789473685e-06,
"loss": 0.2893,
"step": 3575
},
{
"epoch": 0.36,
"learning_rate": 6.740000000000001e-06,
"loss": 0.3837,
"step": 3600
},
{
"epoch": 0.36,
"learning_rate": 6.7136842105263165e-06,
"loss": 0.3255,
"step": 3625
},
{
"epoch": 0.36,
"learning_rate": 6.687368421052632e-06,
"loss": 0.2209,
"step": 3650
},
{
"epoch": 0.37,
"learning_rate": 6.661052631578948e-06,
"loss": 0.2279,
"step": 3675
},
{
"epoch": 0.37,
"learning_rate": 6.634736842105264e-06,
"loss": 0.2711,
"step": 3700
},
{
"epoch": 0.37,
"learning_rate": 6.60842105263158e-06,
"loss": 0.3661,
"step": 3725
},
{
"epoch": 0.38,
"learning_rate": 6.582105263157896e-06,
"loss": 0.4029,
"step": 3750
},
{
"epoch": 0.38,
"learning_rate": 6.5557894736842106e-06,
"loss": 0.2963,
"step": 3775
},
{
"epoch": 0.38,
"learning_rate": 6.529473684210526e-06,
"loss": 0.2796,
"step": 3800
},
{
"epoch": 0.38,
"learning_rate": 6.503157894736842e-06,
"loss": 0.2188,
"step": 3825
},
{
"epoch": 0.39,
"learning_rate": 6.476842105263158e-06,
"loss": 0.2262,
"step": 3850
},
{
"epoch": 0.39,
"learning_rate": 6.450526315789474e-06,
"loss": 0.2033,
"step": 3875
},
{
"epoch": 0.39,
"learning_rate": 6.42421052631579e-06,
"loss": 0.2373,
"step": 3900
},
{
"epoch": 0.39,
"learning_rate": 6.397894736842106e-06,
"loss": 0.2032,
"step": 3925
},
{
"epoch": 0.4,
"learning_rate": 6.371578947368422e-06,
"loss": 0.2088,
"step": 3950
},
{
"epoch": 0.4,
"learning_rate": 6.345263157894738e-06,
"loss": 0.2395,
"step": 3975
},
{
"epoch": 0.4,
"learning_rate": 6.318947368421054e-06,
"loss": 0.2792,
"step": 4000
},
{
"epoch": 0.4,
"eval_loss": 0.3699382245540619,
"eval_runtime": 1779.3947,
"eval_samples_per_second": 9.191,
"eval_steps_per_second": 0.575,
"eval_wer": 14.875815313433796,
"step": 4000
},
{
"epoch": 0.4,
"learning_rate": 6.292631578947369e-06,
"loss": 0.2441,
"step": 4025
},
{
"epoch": 0.41,
"learning_rate": 6.266315789473685e-06,
"loss": 0.2678,
"step": 4050
},
{
"epoch": 0.41,
"learning_rate": 6.24e-06,
"loss": 0.2758,
"step": 4075
},
{
"epoch": 0.41,
"learning_rate": 6.213684210526316e-06,
"loss": 0.2525,
"step": 4100
},
{
"epoch": 0.41,
"learning_rate": 6.187368421052632e-06,
"loss": 0.2469,
"step": 4125
},
{
"epoch": 0.41,
"learning_rate": 6.161052631578948e-06,
"loss": 0.2266,
"step": 4150
},
{
"epoch": 0.42,
"learning_rate": 6.134736842105264e-06,
"loss": 0.2407,
"step": 4175
},
{
"epoch": 0.42,
"learning_rate": 6.108421052631579e-06,
"loss": 0.2647,
"step": 4200
},
{
"epoch": 0.42,
"learning_rate": 6.0821052631578945e-06,
"loss": 0.226,
"step": 4225
},
{
"epoch": 0.42,
"learning_rate": 6.055789473684212e-06,
"loss": 0.2448,
"step": 4250
},
{
"epoch": 0.43,
"learning_rate": 6.029473684210527e-06,
"loss": 0.2175,
"step": 4275
},
{
"epoch": 0.43,
"learning_rate": 6.003157894736843e-06,
"loss": 0.2314,
"step": 4300
},
{
"epoch": 0.43,
"learning_rate": 5.976842105263159e-06,
"loss": 0.2548,
"step": 4325
},
{
"epoch": 0.43,
"learning_rate": 5.950526315789474e-06,
"loss": 0.2263,
"step": 4350
},
{
"epoch": 0.44,
"learning_rate": 5.92421052631579e-06,
"loss": 0.1998,
"step": 4375
},
{
"epoch": 0.44,
"learning_rate": 5.897894736842106e-06,
"loss": 0.2076,
"step": 4400
},
{
"epoch": 0.44,
"learning_rate": 5.871578947368422e-06,
"loss": 0.2066,
"step": 4425
},
{
"epoch": 0.45,
"learning_rate": 5.845263157894737e-06,
"loss": 0.2331,
"step": 4450
},
{
"epoch": 0.45,
"learning_rate": 5.818947368421053e-06,
"loss": 0.2692,
"step": 4475
},
{
"epoch": 0.45,
"learning_rate": 5.7926315789473685e-06,
"loss": 0.2328,
"step": 4500
},
{
"epoch": 0.45,
"learning_rate": 5.766315789473684e-06,
"loss": 0.1955,
"step": 4525
},
{
"epoch": 0.46,
"learning_rate": 5.74e-06,
"loss": 0.2384,
"step": 4550
},
{
"epoch": 0.46,
"learning_rate": 5.713684210526316e-06,
"loss": 0.2523,
"step": 4575
},
{
"epoch": 0.46,
"learning_rate": 5.687368421052633e-06,
"loss": 0.2033,
"step": 4600
},
{
"epoch": 0.46,
"learning_rate": 5.661052631578948e-06,
"loss": 0.2099,
"step": 4625
},
{
"epoch": 0.47,
"learning_rate": 5.634736842105264e-06,
"loss": 0.1867,
"step": 4650
},
{
"epoch": 0.47,
"learning_rate": 5.60842105263158e-06,
"loss": 0.2224,
"step": 4675
},
{
"epoch": 0.47,
"learning_rate": 5.582105263157895e-06,
"loss": 0.2485,
"step": 4700
},
{
"epoch": 0.47,
"learning_rate": 5.555789473684211e-06,
"loss": 0.2282,
"step": 4725
},
{
"epoch": 0.47,
"learning_rate": 5.529473684210527e-06,
"loss": 0.1791,
"step": 4750
},
{
"epoch": 0.48,
"learning_rate": 5.5031578947368425e-06,
"loss": 0.2362,
"step": 4775
},
{
"epoch": 0.48,
"learning_rate": 5.476842105263158e-06,
"loss": 0.1998,
"step": 4800
},
{
"epoch": 0.48,
"learning_rate": 5.450526315789474e-06,
"loss": 0.2081,
"step": 4825
},
{
"epoch": 0.48,
"learning_rate": 5.42421052631579e-06,
"loss": 0.2335,
"step": 4850
},
{
"epoch": 0.49,
"learning_rate": 5.397894736842105e-06,
"loss": 0.2228,
"step": 4875
},
{
"epoch": 0.49,
"learning_rate": 5.371578947368421e-06,
"loss": 0.1993,
"step": 4900
},
{
"epoch": 0.49,
"learning_rate": 5.345263157894738e-06,
"loss": 0.1847,
"step": 4925
},
{
"epoch": 0.49,
"learning_rate": 5.318947368421053e-06,
"loss": 0.1873,
"step": 4950
},
{
"epoch": 0.5,
"learning_rate": 5.292631578947369e-06,
"loss": 0.2249,
"step": 4975
},
{
"epoch": 0.5,
"learning_rate": 5.266315789473685e-06,
"loss": 0.2172,
"step": 5000
},
{
"epoch": 0.5,
"eval_loss": 0.3576693534851074,
"eval_runtime": 1769.8813,
"eval_samples_per_second": 9.24,
"eval_steps_per_second": 0.578,
"eval_wer": 14.265976930244515,
"step": 5000
},
{
"epoch": 0.5,
"learning_rate": 5.240000000000001e-06,
"loss": 0.1577,
"step": 5025
},
{
"epoch": 0.51,
"learning_rate": 5.2136842105263165e-06,
"loss": 0.1416,
"step": 5050
},
{
"epoch": 0.51,
"learning_rate": 5.187368421052632e-06,
"loss": 0.1306,
"step": 5075
},
{
"epoch": 0.51,
"learning_rate": 5.161052631578948e-06,
"loss": 0.1892,
"step": 5100
},
{
"epoch": 0.51,
"learning_rate": 5.134736842105263e-06,
"loss": 0.7114,
"step": 5125
},
{
"epoch": 0.52,
"learning_rate": 5.108421052631579e-06,
"loss": 0.8928,
"step": 5150
},
{
"epoch": 0.52,
"learning_rate": 5.082105263157895e-06,
"loss": 0.8808,
"step": 5175
},
{
"epoch": 0.52,
"learning_rate": 5.0557894736842105e-06,
"loss": 0.832,
"step": 5200
},
{
"epoch": 0.52,
"learning_rate": 5.029473684210526e-06,
"loss": 0.8431,
"step": 5225
},
{
"epoch": 0.53,
"learning_rate": 5.003157894736842e-06,
"loss": 0.8088,
"step": 5250
},
{
"epoch": 0.53,
"learning_rate": 4.976842105263158e-06,
"loss": 0.7433,
"step": 5275
},
{
"epoch": 0.53,
"learning_rate": 4.950526315789474e-06,
"loss": 0.7206,
"step": 5300
},
{
"epoch": 0.53,
"learning_rate": 4.92421052631579e-06,
"loss": 0.4545,
"step": 5325
},
{
"epoch": 0.54,
"learning_rate": 4.897894736842106e-06,
"loss": 0.2518,
"step": 5350
},
{
"epoch": 0.54,
"learning_rate": 4.871578947368421e-06,
"loss": 0.1921,
"step": 5375
},
{
"epoch": 0.54,
"learning_rate": 4.845263157894737e-06,
"loss": 0.154,
"step": 5400
},
{
"epoch": 0.54,
"learning_rate": 4.818947368421053e-06,
"loss": 0.1739,
"step": 5425
},
{
"epoch": 0.55,
"learning_rate": 4.792631578947369e-06,
"loss": 0.1653,
"step": 5450
},
{
"epoch": 0.55,
"learning_rate": 4.7663157894736845e-06,
"loss": 0.1808,
"step": 5475
},
{
"epoch": 0.55,
"learning_rate": 4.74e-06,
"loss": 0.177,
"step": 5500
},
{
"epoch": 0.55,
"learning_rate": 4.713684210526316e-06,
"loss": 0.1802,
"step": 5525
},
{
"epoch": 0.56,
"learning_rate": 4.687368421052632e-06,
"loss": 0.1805,
"step": 5550
},
{
"epoch": 0.56,
"learning_rate": 4.661052631578948e-06,
"loss": 0.1894,
"step": 5575
},
{
"epoch": 0.56,
"learning_rate": 4.634736842105264e-06,
"loss": 0.3482,
"step": 5600
},
{
"epoch": 0.56,
"learning_rate": 4.6084210526315794e-06,
"loss": 0.3409,
"step": 5625
},
{
"epoch": 0.56,
"learning_rate": 4.582105263157895e-06,
"loss": 0.3127,
"step": 5650
},
{
"epoch": 0.57,
"learning_rate": 4.55578947368421e-06,
"loss": 0.2032,
"step": 5675
},
{
"epoch": 0.57,
"learning_rate": 4.529473684210527e-06,
"loss": 0.1763,
"step": 5700
},
{
"epoch": 0.57,
"learning_rate": 4.503157894736843e-06,
"loss": 0.1707,
"step": 5725
},
{
"epoch": 0.57,
"learning_rate": 4.4768421052631585e-06,
"loss": 0.1734,
"step": 5750
},
{
"epoch": 0.58,
"learning_rate": 4.450526315789474e-06,
"loss": 0.1568,
"step": 5775
},
{
"epoch": 0.58,
"learning_rate": 4.424210526315789e-06,
"loss": 0.1352,
"step": 5800
},
{
"epoch": 0.58,
"learning_rate": 4.397894736842105e-06,
"loss": 0.1528,
"step": 5825
},
{
"epoch": 0.58,
"learning_rate": 4.371578947368421e-06,
"loss": 0.1609,
"step": 5850
},
{
"epoch": 0.59,
"learning_rate": 4.345263157894738e-06,
"loss": 0.1393,
"step": 5875
},
{
"epoch": 0.59,
"learning_rate": 4.3189473684210535e-06,
"loss": 0.1472,
"step": 5900
},
{
"epoch": 0.59,
"learning_rate": 4.2926315789473684e-06,
"loss": 0.1276,
"step": 5925
},
{
"epoch": 0.59,
"learning_rate": 4.266315789473684e-06,
"loss": 0.3361,
"step": 5950
},
{
"epoch": 0.6,
"learning_rate": 4.24e-06,
"loss": 0.4034,
"step": 5975
},
{
"epoch": 0.6,
"learning_rate": 4.213684210526316e-06,
"loss": 0.3616,
"step": 6000
},
{
"epoch": 0.6,
"eval_loss": 0.40417543053627014,
"eval_runtime": 1804.0486,
"eval_samples_per_second": 9.065,
"eval_steps_per_second": 0.567,
"eval_wer": 18.184622305047714,
"step": 6000
},
{
"epoch": 0.6,
"learning_rate": 4.1873684210526325e-06,
"loss": 0.3684,
"step": 6025
},
{
"epoch": 0.6,
"learning_rate": 4.1610526315789475e-06,
"loss": 0.3671,
"step": 6050
},
{
"epoch": 0.61,
"learning_rate": 4.134736842105263e-06,
"loss": 0.3644,
"step": 6075
},
{
"epoch": 0.61,
"learning_rate": 4.108421052631579e-06,
"loss": 0.354,
"step": 6100
},
{
"epoch": 0.61,
"learning_rate": 4.082105263157895e-06,
"loss": 0.3545,
"step": 6125
},
{
"epoch": 0.61,
"learning_rate": 4.055789473684211e-06,
"loss": 0.3182,
"step": 6150
},
{
"epoch": 0.62,
"learning_rate": 4.029473684210527e-06,
"loss": 0.328,
"step": 6175
},
{
"epoch": 0.62,
"learning_rate": 4.0031578947368424e-06,
"loss": 0.2427,
"step": 6200
},
{
"epoch": 0.62,
"learning_rate": 3.976842105263158e-06,
"loss": 0.1891,
"step": 6225
},
{
"epoch": 0.62,
"learning_rate": 3.950526315789474e-06,
"loss": 0.1886,
"step": 6250
},
{
"epoch": 0.63,
"learning_rate": 3.9252631578947366e-06,
"loss": 0.2127,
"step": 6275
},
{
"epoch": 0.63,
"learning_rate": 3.898947368421052e-06,
"loss": 0.2511,
"step": 6300
},
{
"epoch": 0.63,
"learning_rate": 3.872631578947369e-06,
"loss": 0.249,
"step": 6325
},
{
"epoch": 0.64,
"learning_rate": 3.846315789473685e-06,
"loss": 0.2476,
"step": 6350
},
{
"epoch": 0.64,
"learning_rate": 3.820000000000001e-06,
"loss": 0.2364,
"step": 6375
},
{
"epoch": 0.64,
"learning_rate": 3.793684210526316e-06,
"loss": 0.2561,
"step": 6400
},
{
"epoch": 0.64,
"learning_rate": 3.767368421052632e-06,
"loss": 0.2471,
"step": 6425
},
{
"epoch": 0.65,
"learning_rate": 3.7410526315789473e-06,
"loss": 0.2425,
"step": 6450
},
{
"epoch": 0.65,
"learning_rate": 3.714736842105263e-06,
"loss": 0.2539,
"step": 6475
},
{
"epoch": 0.65,
"learning_rate": 3.6884210526315794e-06,
"loss": 0.2504,
"step": 6500
},
{
"epoch": 0.65,
"learning_rate": 3.662105263157895e-06,
"loss": 0.2389,
"step": 6525
},
{
"epoch": 0.66,
"learning_rate": 3.635789473684211e-06,
"loss": 0.2513,
"step": 6550
},
{
"epoch": 0.66,
"learning_rate": 3.6094736842105264e-06,
"loss": 0.2612,
"step": 6575
},
{
"epoch": 0.66,
"learning_rate": 3.5831578947368422e-06,
"loss": 0.2542,
"step": 6600
},
{
"epoch": 0.66,
"learning_rate": 3.556842105263158e-06,
"loss": 0.2578,
"step": 6625
},
{
"epoch": 0.67,
"learning_rate": 3.5305263157894743e-06,
"loss": 0.2421,
"step": 6650
},
{
"epoch": 0.67,
"learning_rate": 3.50421052631579e-06,
"loss": 0.2496,
"step": 6675
},
{
"epoch": 0.67,
"learning_rate": 3.4778947368421055e-06,
"loss": 0.2372,
"step": 6700
},
{
"epoch": 0.67,
"learning_rate": 3.4515789473684213e-06,
"loss": 0.2263,
"step": 6725
},
{
"epoch": 0.68,
"learning_rate": 3.425263157894737e-06,
"loss": 0.2551,
"step": 6750
},
{
"epoch": 0.68,
"learning_rate": 3.398947368421053e-06,
"loss": 0.2403,
"step": 6775
},
{
"epoch": 0.68,
"learning_rate": 3.3726315789473683e-06,
"loss": 0.2478,
"step": 6800
},
{
"epoch": 0.68,
"learning_rate": 3.3463157894736846e-06,
"loss": 0.2371,
"step": 6825
},
{
"epoch": 0.69,
"learning_rate": 3.3200000000000004e-06,
"loss": 0.2572,
"step": 6850
},
{
"epoch": 0.69,
"learning_rate": 3.2936842105263162e-06,
"loss": 0.2288,
"step": 6875
},
{
"epoch": 0.69,
"learning_rate": 3.267368421052632e-06,
"loss": 0.2459,
"step": 6900
},
{
"epoch": 0.69,
"learning_rate": 3.2410526315789474e-06,
"loss": 0.2458,
"step": 6925
},
{
"epoch": 0.69,
"learning_rate": 3.2147368421052633e-06,
"loss": 0.2511,
"step": 6950
},
{
"epoch": 0.7,
"learning_rate": 3.188421052631579e-06,
"loss": 0.2594,
"step": 6975
},
{
"epoch": 0.7,
"learning_rate": 3.1621052631578953e-06,
"loss": 0.2456,
"step": 7000
},
{
"epoch": 0.7,
"eval_loss": 0.3374858796596527,
"eval_runtime": 1789.664,
"eval_samples_per_second": 9.138,
"eval_steps_per_second": 0.572,
"eval_wer": 13.309128297400635,
"step": 7000
},
{
"epoch": 0.7,
"learning_rate": 3.135789473684211e-06,
"loss": 0.2697,
"step": 7025
},
{
"epoch": 0.7,
"learning_rate": 3.1094736842105265e-06,
"loss": 0.2904,
"step": 7050
},
{
"epoch": 0.71,
"learning_rate": 3.0831578947368423e-06,
"loss": 0.2812,
"step": 7075
},
{
"epoch": 0.71,
"learning_rate": 3.056842105263158e-06,
"loss": 0.2849,
"step": 7100
},
{
"epoch": 0.71,
"learning_rate": 3.0305263157894736e-06,
"loss": 0.2844,
"step": 7125
},
{
"epoch": 0.71,
"learning_rate": 3.0042105263157894e-06,
"loss": 0.3018,
"step": 7150
},
{
"epoch": 0.72,
"learning_rate": 2.9778947368421056e-06,
"loss": 0.2803,
"step": 7175
},
{
"epoch": 0.72,
"learning_rate": 2.9515789473684214e-06,
"loss": 0.2986,
"step": 7200
},
{
"epoch": 0.72,
"learning_rate": 2.9252631578947373e-06,
"loss": 0.2981,
"step": 7225
},
{
"epoch": 0.72,
"learning_rate": 2.8989473684210526e-06,
"loss": 0.2913,
"step": 7250
},
{
"epoch": 0.73,
"learning_rate": 2.8726315789473685e-06,
"loss": 0.2883,
"step": 7275
},
{
"epoch": 0.73,
"learning_rate": 2.8463157894736843e-06,
"loss": 0.2856,
"step": 7300
},
{
"epoch": 0.73,
"learning_rate": 2.82e-06,
"loss": 0.2629,
"step": 7325
},
{
"epoch": 0.73,
"learning_rate": 2.7936842105263163e-06,
"loss": 0.2842,
"step": 7350
},
{
"epoch": 0.74,
"learning_rate": 2.7673684210526317e-06,
"loss": 0.2716,
"step": 7375
},
{
"epoch": 0.74,
"learning_rate": 2.7410526315789476e-06,
"loss": 0.3039,
"step": 7400
},
{
"epoch": 0.74,
"learning_rate": 2.7147368421052634e-06,
"loss": 0.2908,
"step": 7425
},
{
"epoch": 0.74,
"learning_rate": 2.688421052631579e-06,
"loss": 0.2968,
"step": 7450
},
{
"epoch": 0.75,
"learning_rate": 2.6621052631578946e-06,
"loss": 0.2587,
"step": 7475
},
{
"epoch": 0.75,
"learning_rate": 2.635789473684211e-06,
"loss": 0.3114,
"step": 7500
},
{
"epoch": 0.75,
"learning_rate": 2.6094736842105267e-06,
"loss": 0.2176,
"step": 7525
},
{
"epoch": 0.76,
"learning_rate": 2.5831578947368425e-06,
"loss": 0.3007,
"step": 7550
},
{
"epoch": 0.76,
"learning_rate": 2.5568421052631583e-06,
"loss": 0.2782,
"step": 7575
},
{
"epoch": 0.76,
"learning_rate": 2.5305263157894737e-06,
"loss": 0.1537,
"step": 7600
},
{
"epoch": 0.76,
"learning_rate": 2.5042105263157895e-06,
"loss": 0.2005,
"step": 7625
},
{
"epoch": 0.77,
"learning_rate": 2.4778947368421053e-06,
"loss": 0.1959,
"step": 7650
},
{
"epoch": 0.77,
"learning_rate": 2.451578947368421e-06,
"loss": 0.198,
"step": 7675
},
{
"epoch": 0.77,
"learning_rate": 2.425263157894737e-06,
"loss": 0.2075,
"step": 7700
},
{
"epoch": 0.77,
"learning_rate": 2.3989473684210528e-06,
"loss": 0.1737,
"step": 7725
},
{
"epoch": 0.78,
"learning_rate": 2.3726315789473686e-06,
"loss": 0.1238,
"step": 7750
},
{
"epoch": 0.78,
"learning_rate": 2.3463157894736844e-06,
"loss": 0.1584,
"step": 7775
},
{
"epoch": 0.78,
"learning_rate": 2.3200000000000002e-06,
"loss": 0.24,
"step": 7800
},
{
"epoch": 0.78,
"learning_rate": 2.293684210526316e-06,
"loss": 0.4397,
"step": 7825
},
{
"epoch": 0.79,
"learning_rate": 2.267368421052632e-06,
"loss": 0.3095,
"step": 7850
},
{
"epoch": 0.79,
"learning_rate": 2.2410526315789473e-06,
"loss": 0.2157,
"step": 7875
},
{
"epoch": 0.79,
"learning_rate": 2.2147368421052635e-06,
"loss": 0.238,
"step": 7900
},
{
"epoch": 0.79,
"learning_rate": 2.188421052631579e-06,
"loss": 0.2518,
"step": 7925
},
{
"epoch": 0.8,
"learning_rate": 2.1621052631578947e-06,
"loss": 0.2654,
"step": 7950
},
{
"epoch": 0.8,
"learning_rate": 2.135789473684211e-06,
"loss": 0.2588,
"step": 7975
},
{
"epoch": 0.8,
"learning_rate": 2.1094736842105264e-06,
"loss": 0.2505,
"step": 8000
},
{
"epoch": 0.8,
"eval_loss": 0.3394748270511627,
"eval_runtime": 1776.4137,
"eval_samples_per_second": 9.206,
"eval_steps_per_second": 0.576,
"eval_wer": 13.622722745236644,
"step": 8000
},
{
"epoch": 0.8,
"learning_rate": 2.083157894736842e-06,
"loss": 0.2013,
"step": 8025
},
{
"epoch": 0.81,
"learning_rate": 2.056842105263158e-06,
"loss": 0.2375,
"step": 8050
},
{
"epoch": 0.81,
"learning_rate": 2.030526315789474e-06,
"loss": 0.2546,
"step": 8075
},
{
"epoch": 0.81,
"learning_rate": 2.0042105263157896e-06,
"loss": 0.3597,
"step": 8100
},
{
"epoch": 0.81,
"learning_rate": 1.9778947368421055e-06,
"loss": 0.3629,
"step": 8125
},
{
"epoch": 0.81,
"learning_rate": 1.9515789473684213e-06,
"loss": 0.2185,
"step": 8150
},
{
"epoch": 0.82,
"learning_rate": 1.925263157894737e-06,
"loss": 0.2011,
"step": 8175
},
{
"epoch": 0.82,
"learning_rate": 1.8989473684210527e-06,
"loss": 0.1981,
"step": 8200
},
{
"epoch": 0.82,
"learning_rate": 1.8726315789473687e-06,
"loss": 0.1846,
"step": 8225
},
{
"epoch": 0.82,
"learning_rate": 1.8463157894736843e-06,
"loss": 0.2052,
"step": 8250
},
{
"epoch": 0.83,
"learning_rate": 1.8200000000000002e-06,
"loss": 0.2016,
"step": 8275
},
{
"epoch": 0.83,
"learning_rate": 1.7936842105263158e-06,
"loss": 0.1761,
"step": 8300
},
{
"epoch": 0.83,
"learning_rate": 1.7673684210526318e-06,
"loss": 0.1632,
"step": 8325
},
{
"epoch": 0.83,
"learning_rate": 1.742105263157895e-06,
"loss": 0.182,
"step": 8350
},
{
"epoch": 0.84,
"learning_rate": 1.7157894736842107e-06,
"loss": 0.182,
"step": 8375
},
{
"epoch": 0.84,
"learning_rate": 1.6894736842105264e-06,
"loss": 0.172,
"step": 8400
},
{
"epoch": 0.84,
"learning_rate": 1.6631578947368424e-06,
"loss": 0.1919,
"step": 8425
},
{
"epoch": 0.84,
"learning_rate": 1.636842105263158e-06,
"loss": 0.169,
"step": 8450
},
{
"epoch": 0.85,
"learning_rate": 1.6105263157894738e-06,
"loss": 0.2048,
"step": 8475
},
{
"epoch": 0.85,
"learning_rate": 1.5842105263157894e-06,
"loss": 0.2797,
"step": 8500
},
{
"epoch": 0.85,
"learning_rate": 1.5578947368421054e-06,
"loss": 0.2226,
"step": 8525
},
{
"epoch": 0.85,
"learning_rate": 1.5315789473684213e-06,
"loss": 0.1446,
"step": 8550
},
{
"epoch": 0.86,
"learning_rate": 1.5052631578947369e-06,
"loss": 0.1187,
"step": 8575
},
{
"epoch": 0.86,
"learning_rate": 1.478947368421053e-06,
"loss": 0.1332,
"step": 8600
},
{
"epoch": 0.86,
"learning_rate": 1.4526315789473685e-06,
"loss": 0.1805,
"step": 8625
},
{
"epoch": 0.86,
"learning_rate": 1.4263157894736843e-06,
"loss": 0.2092,
"step": 8650
},
{
"epoch": 0.87,
"learning_rate": 1.4000000000000001e-06,
"loss": 0.1891,
"step": 8675
},
{
"epoch": 0.87,
"learning_rate": 1.373684210526316e-06,
"loss": 0.1875,
"step": 8700
},
{
"epoch": 0.87,
"learning_rate": 1.3473684210526316e-06,
"loss": 0.2381,
"step": 8725
},
{
"epoch": 0.88,
"learning_rate": 1.3210526315789474e-06,
"loss": 0.2345,
"step": 8750
},
{
"epoch": 0.88,
"learning_rate": 1.2947368421052634e-06,
"loss": 0.2702,
"step": 8775
},
{
"epoch": 0.88,
"learning_rate": 1.268421052631579e-06,
"loss": 0.2574,
"step": 8800
},
{
"epoch": 0.88,
"learning_rate": 1.2421052631578948e-06,
"loss": 0.2513,
"step": 8825
},
{
"epoch": 0.89,
"learning_rate": 1.2157894736842107e-06,
"loss": 0.2413,
"step": 8850
},
{
"epoch": 0.89,
"learning_rate": 1.1894736842105265e-06,
"loss": 0.2531,
"step": 8875
},
{
"epoch": 0.89,
"learning_rate": 1.163157894736842e-06,
"loss": 0.2375,
"step": 8900
},
{
"epoch": 0.89,
"learning_rate": 1.136842105263158e-06,
"loss": 0.242,
"step": 8925
},
{
"epoch": 0.9,
"learning_rate": 1.110526315789474e-06,
"loss": 0.247,
"step": 8950
},
{
"epoch": 0.9,
"learning_rate": 1.0842105263157895e-06,
"loss": 0.2394,
"step": 8975
},
{
"epoch": 0.9,
"learning_rate": 1.0578947368421054e-06,
"loss": 0.2563,
"step": 9000
},
{
"epoch": 0.9,
"eval_loss": 0.33053791522979736,
"eval_runtime": 1767.455,
"eval_samples_per_second": 9.253,
"eval_steps_per_second": 0.579,
"eval_wer": 13.140764065160814,
"step": 9000
},
{
"epoch": 0.9,
"learning_rate": 1.0315789473684212e-06,
"loss": 0.2483,
"step": 9025
},
{
"epoch": 0.91,
"learning_rate": 1.005263157894737e-06,
"loss": 0.212,
"step": 9050
},
{
"epoch": 0.91,
"learning_rate": 9.789473684210526e-07,
"loss": 0.2196,
"step": 9075
},
{
"epoch": 0.91,
"learning_rate": 9.526315789473685e-07,
"loss": 0.2415,
"step": 9100
},
{
"epoch": 0.91,
"learning_rate": 9.263157894736844e-07,
"loss": 0.2211,
"step": 9125
},
{
"epoch": 0.92,
"learning_rate": 9.000000000000001e-07,
"loss": 0.2199,
"step": 9150
},
{
"epoch": 0.92,
"learning_rate": 8.736842105263159e-07,
"loss": 0.2434,
"step": 9175
},
{
"epoch": 0.92,
"learning_rate": 8.473684210526316e-07,
"loss": 0.2282,
"step": 9200
},
{
"epoch": 0.92,
"learning_rate": 8.210526315789474e-07,
"loss": 0.2262,
"step": 9225
},
{
"epoch": 0.93,
"learning_rate": 7.947368421052632e-07,
"loss": 0.2446,
"step": 9250
},
{
"epoch": 0.93,
"learning_rate": 7.684210526315789e-07,
"loss": 0.2329,
"step": 9275
},
{
"epoch": 0.93,
"learning_rate": 7.421052631578949e-07,
"loss": 0.2032,
"step": 9300
},
{
"epoch": 0.93,
"learning_rate": 7.157894736842106e-07,
"loss": 0.2149,
"step": 9325
},
{
"epoch": 0.94,
"learning_rate": 6.894736842105264e-07,
"loss": 0.2536,
"step": 9350
},
{
"epoch": 0.94,
"learning_rate": 6.631578947368422e-07,
"loss": 0.2662,
"step": 9375
},
{
"epoch": 0.94,
"learning_rate": 6.368421052631579e-07,
"loss": 0.2825,
"step": 9400
},
{
"epoch": 0.94,
"learning_rate": 6.105263157894738e-07,
"loss": 0.2655,
"step": 9425
},
{
"epoch": 0.94,
"learning_rate": 5.842105263157896e-07,
"loss": 0.2658,
"step": 9450
},
{
"epoch": 0.95,
"learning_rate": 5.578947368421053e-07,
"loss": 0.2761,
"step": 9475
},
{
"epoch": 0.95,
"learning_rate": 5.315789473684211e-07,
"loss": 0.2579,
"step": 9500
},
{
"epoch": 0.95,
"learning_rate": 5.052631578947369e-07,
"loss": 0.2751,
"step": 9525
},
{
"epoch": 0.95,
"learning_rate": 4.789473684210526e-07,
"loss": 0.2818,
"step": 9550
},
{
"epoch": 0.96,
"learning_rate": 4.526315789473685e-07,
"loss": 0.2633,
"step": 9575
},
{
"epoch": 0.96,
"learning_rate": 4.2631578947368427e-07,
"loss": 0.2897,
"step": 9600
},
{
"epoch": 0.96,
"learning_rate": 4.0000000000000003e-07,
"loss": 0.2602,
"step": 9625
},
{
"epoch": 0.96,
"learning_rate": 3.736842105263158e-07,
"loss": 0.2623,
"step": 9650
},
{
"epoch": 0.97,
"learning_rate": 3.4736842105263157e-07,
"loss": 0.2637,
"step": 9675
},
{
"epoch": 0.97,
"learning_rate": 3.210526315789474e-07,
"loss": 0.2586,
"step": 9700
},
{
"epoch": 0.97,
"learning_rate": 2.9473684210526315e-07,
"loss": 0.2367,
"step": 9725
},
{
"epoch": 0.97,
"learning_rate": 2.6842105263157897e-07,
"loss": 0.2334,
"step": 9750
},
{
"epoch": 0.98,
"learning_rate": 2.4210526315789473e-07,
"loss": 0.2291,
"step": 9775
},
{
"epoch": 0.98,
"learning_rate": 2.1578947368421053e-07,
"loss": 0.235,
"step": 9800
},
{
"epoch": 0.98,
"learning_rate": 1.8947368421052634e-07,
"loss": 0.2429,
"step": 9825
},
{
"epoch": 0.98,
"learning_rate": 1.631578947368421e-07,
"loss": 0.2358,
"step": 9850
},
{
"epoch": 0.99,
"learning_rate": 1.368421052631579e-07,
"loss": 0.2155,
"step": 9875
},
{
"epoch": 0.99,
"learning_rate": 1.1052631578947368e-07,
"loss": 0.224,
"step": 9900
},
{
"epoch": 0.99,
"learning_rate": 8.421052631578947e-08,
"loss": 0.2384,
"step": 9925
},
{
"epoch": 0.99,
"learning_rate": 5.789473684210527e-08,
"loss": 0.2163,
"step": 9950
},
{
"epoch": 1.0,
"learning_rate": 3.157894736842106e-08,
"loss": 0.2093,
"step": 9975
},
{
"epoch": 1.0,
"learning_rate": 5.263157894736842e-09,
"loss": 0.2395,
"step": 10000
},
{
"epoch": 1.0,
"eval_loss": 0.3268873691558838,
"eval_runtime": 1771.5472,
"eval_samples_per_second": 9.231,
"eval_steps_per_second": 0.577,
"eval_wer": 13.058509783761204,
"step": 10000
},
{
"epoch": 1.0,
"step": 10000,
"total_flos": 9.23473281024e+19,
"train_loss": 0.26370683794021604,
"train_runtime": 54817.416,
"train_samples_per_second": 5.838,
"train_steps_per_second": 0.182
}
],
"max_steps": 10000,
"num_train_epochs": 9223372036854775807,
"total_flos": 9.23473281024e+19,
"trial_name": null,
"trial_params": null
}