whisper-medium-ar / trainer_state.json
lorenzoncina's picture
End of training
cdca41b
{
"best_metric": 47.53066666666667,
"best_model_checkpoint": "./checkpoint-7000",
"epoch": 7.0865,
"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": 0.8205,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.400000000000001e-07,
"loss": 0.5841,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.44e-06,
"loss": 0.376,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.94e-06,
"loss": 0.3351,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.3202,
"step": 125
},
{
"epoch": 0.01,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.2705,
"step": 150
},
{
"epoch": 0.02,
"learning_rate": 3.44e-06,
"loss": 0.2686,
"step": 175
},
{
"epoch": 0.02,
"learning_rate": 3.94e-06,
"loss": 0.2507,
"step": 200
},
{
"epoch": 0.02,
"learning_rate": 4.440000000000001e-06,
"loss": 0.2769,
"step": 225
},
{
"epoch": 0.03,
"learning_rate": 4.94e-06,
"loss": 0.2783,
"step": 250
},
{
"epoch": 0.03,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.2515,
"step": 275
},
{
"epoch": 0.03,
"learning_rate": 5.94e-06,
"loss": 0.2237,
"step": 300
},
{
"epoch": 0.03,
"learning_rate": 6.440000000000001e-06,
"loss": 0.2962,
"step": 325
},
{
"epoch": 0.04,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.2655,
"step": 350
},
{
"epoch": 0.04,
"learning_rate": 7.440000000000001e-06,
"loss": 0.2719,
"step": 375
},
{
"epoch": 0.04,
"learning_rate": 7.94e-06,
"loss": 0.2923,
"step": 400
},
{
"epoch": 0.04,
"learning_rate": 8.44e-06,
"loss": 0.299,
"step": 425
},
{
"epoch": 0.04,
"learning_rate": 8.94e-06,
"loss": 0.3011,
"step": 450
},
{
"epoch": 0.05,
"learning_rate": 9.440000000000001e-06,
"loss": 0.2809,
"step": 475
},
{
"epoch": 0.05,
"learning_rate": 9.940000000000001e-06,
"loss": 0.2609,
"step": 500
},
{
"epoch": 0.05,
"learning_rate": 9.976842105263158e-06,
"loss": 0.2181,
"step": 525
},
{
"epoch": 0.06,
"learning_rate": 9.950526315789475e-06,
"loss": 0.2394,
"step": 550
},
{
"epoch": 0.06,
"learning_rate": 9.92421052631579e-06,
"loss": 0.2259,
"step": 575
},
{
"epoch": 0.06,
"learning_rate": 9.897894736842107e-06,
"loss": 0.1863,
"step": 600
},
{
"epoch": 0.06,
"learning_rate": 9.871578947368422e-06,
"loss": 0.1973,
"step": 625
},
{
"epoch": 0.07,
"learning_rate": 9.845263157894738e-06,
"loss": 0.211,
"step": 650
},
{
"epoch": 0.07,
"learning_rate": 9.818947368421053e-06,
"loss": 0.2216,
"step": 675
},
{
"epoch": 0.07,
"learning_rate": 9.79263157894737e-06,
"loss": 0.2231,
"step": 700
},
{
"epoch": 0.07,
"learning_rate": 9.766315789473685e-06,
"loss": 0.2117,
"step": 725
},
{
"epoch": 0.07,
"learning_rate": 9.74e-06,
"loss": 0.2187,
"step": 750
},
{
"epoch": 0.08,
"learning_rate": 9.713684210526317e-06,
"loss": 0.2017,
"step": 775
},
{
"epoch": 0.08,
"learning_rate": 9.687368421052632e-06,
"loss": 0.2117,
"step": 800
},
{
"epoch": 0.08,
"learning_rate": 9.661052631578948e-06,
"loss": 0.2357,
"step": 825
},
{
"epoch": 0.09,
"learning_rate": 9.634736842105265e-06,
"loss": 0.2567,
"step": 850
},
{
"epoch": 0.09,
"learning_rate": 9.60842105263158e-06,
"loss": 0.2742,
"step": 875
},
{
"epoch": 0.09,
"learning_rate": 9.582105263157897e-06,
"loss": 0.2451,
"step": 900
},
{
"epoch": 0.09,
"learning_rate": 9.555789473684211e-06,
"loss": 0.2464,
"step": 925
},
{
"epoch": 0.1,
"learning_rate": 9.529473684210528e-06,
"loss": 0.2208,
"step": 950
},
{
"epoch": 0.1,
"learning_rate": 9.503157894736843e-06,
"loss": 0.1912,
"step": 975
},
{
"epoch": 0.1,
"learning_rate": 9.476842105263158e-06,
"loss": 0.2215,
"step": 1000
},
{
"epoch": 0.1,
"eval_loss": 0.33608928322792053,
"eval_runtime": 2652.355,
"eval_samples_per_second": 3.936,
"eval_steps_per_second": 0.984,
"eval_wer": 49.93066666666667,
"step": 1000
},
{
"epoch": 0.1,
"learning_rate": 9.450526315789475e-06,
"loss": 0.2505,
"step": 1025
},
{
"epoch": 0.1,
"learning_rate": 9.42421052631579e-06,
"loss": 0.249,
"step": 1050
},
{
"epoch": 0.11,
"learning_rate": 9.397894736842106e-06,
"loss": 0.1987,
"step": 1075
},
{
"epoch": 0.11,
"learning_rate": 9.371578947368421e-06,
"loss": 0.215,
"step": 1100
},
{
"epoch": 0.11,
"learning_rate": 9.345263157894738e-06,
"loss": 0.1871,
"step": 1125
},
{
"epoch": 0.12,
"learning_rate": 9.318947368421053e-06,
"loss": 0.1725,
"step": 1150
},
{
"epoch": 0.12,
"learning_rate": 9.292631578947368e-06,
"loss": 0.1661,
"step": 1175
},
{
"epoch": 0.12,
"learning_rate": 9.266315789473685e-06,
"loss": 0.1691,
"step": 1200
},
{
"epoch": 0.12,
"learning_rate": 9.240000000000001e-06,
"loss": 0.1676,
"step": 1225
},
{
"epoch": 0.12,
"learning_rate": 9.213684210526316e-06,
"loss": 0.2093,
"step": 1250
},
{
"epoch": 0.13,
"learning_rate": 9.187368421052633e-06,
"loss": 0.2155,
"step": 1275
},
{
"epoch": 0.13,
"learning_rate": 9.161052631578948e-06,
"loss": 0.2282,
"step": 1300
},
{
"epoch": 1.0,
"learning_rate": 9.134736842105265e-06,
"loss": 0.1892,
"step": 1325
},
{
"epoch": 1.0,
"learning_rate": 9.10842105263158e-06,
"loss": 0.1552,
"step": 1350
},
{
"epoch": 1.01,
"learning_rate": 9.082105263157896e-06,
"loss": 0.1691,
"step": 1375
},
{
"epoch": 1.01,
"learning_rate": 9.055789473684211e-06,
"loss": 0.1747,
"step": 1400
},
{
"epoch": 1.01,
"learning_rate": 9.029473684210526e-06,
"loss": 0.2059,
"step": 1425
},
{
"epoch": 1.01,
"learning_rate": 9.003157894736843e-06,
"loss": 0.1624,
"step": 1450
},
{
"epoch": 1.02,
"learning_rate": 8.976842105263158e-06,
"loss": 0.1633,
"step": 1475
},
{
"epoch": 1.02,
"learning_rate": 8.950526315789474e-06,
"loss": 0.1652,
"step": 1500
},
{
"epoch": 1.02,
"learning_rate": 8.92421052631579e-06,
"loss": 0.1728,
"step": 1525
},
{
"epoch": 1.02,
"learning_rate": 8.897894736842106e-06,
"loss": 0.1525,
"step": 1550
},
{
"epoch": 1.03,
"learning_rate": 8.871578947368423e-06,
"loss": 0.1525,
"step": 1575
},
{
"epoch": 1.03,
"learning_rate": 8.845263157894738e-06,
"loss": 0.13,
"step": 1600
},
{
"epoch": 1.03,
"learning_rate": 8.818947368421054e-06,
"loss": 0.1488,
"step": 1625
},
{
"epoch": 1.03,
"learning_rate": 8.79263157894737e-06,
"loss": 0.1318,
"step": 1650
},
{
"epoch": 1.04,
"learning_rate": 8.766315789473684e-06,
"loss": 0.1525,
"step": 1675
},
{
"epoch": 1.04,
"learning_rate": 8.740000000000001e-06,
"loss": 0.152,
"step": 1700
},
{
"epoch": 1.04,
"learning_rate": 8.713684210526316e-06,
"loss": 0.148,
"step": 1725
},
{
"epoch": 1.04,
"learning_rate": 8.687368421052633e-06,
"loss": 0.1477,
"step": 1750
},
{
"epoch": 1.05,
"learning_rate": 8.661052631578948e-06,
"loss": 0.1219,
"step": 1775
},
{
"epoch": 1.05,
"learning_rate": 8.634736842105264e-06,
"loss": 0.1234,
"step": 1800
},
{
"epoch": 1.05,
"learning_rate": 8.60842105263158e-06,
"loss": 0.1198,
"step": 1825
},
{
"epoch": 1.05,
"learning_rate": 8.582105263157894e-06,
"loss": 0.1014,
"step": 1850
},
{
"epoch": 1.06,
"learning_rate": 8.555789473684213e-06,
"loss": 0.11,
"step": 1875
},
{
"epoch": 1.06,
"learning_rate": 8.529473684210528e-06,
"loss": 0.0894,
"step": 1900
},
{
"epoch": 1.06,
"learning_rate": 8.503157894736843e-06,
"loss": 0.104,
"step": 1925
},
{
"epoch": 1.06,
"learning_rate": 8.47684210526316e-06,
"loss": 0.0956,
"step": 1950
},
{
"epoch": 1.07,
"learning_rate": 8.450526315789474e-06,
"loss": 0.1233,
"step": 1975
},
{
"epoch": 1.07,
"learning_rate": 8.425263157894737e-06,
"loss": 0.1134,
"step": 2000
},
{
"epoch": 1.07,
"eval_loss": 0.3289839029312134,
"eval_runtime": 2679.1374,
"eval_samples_per_second": 3.897,
"eval_steps_per_second": 0.974,
"eval_wer": 56.76,
"step": 2000
},
{
"epoch": 1.07,
"learning_rate": 8.398947368421053e-06,
"loss": 0.0952,
"step": 2025
},
{
"epoch": 1.07,
"learning_rate": 8.372631578947368e-06,
"loss": 0.0909,
"step": 2050
},
{
"epoch": 1.08,
"learning_rate": 8.346315789473685e-06,
"loss": 0.0929,
"step": 2075
},
{
"epoch": 1.08,
"learning_rate": 8.32e-06,
"loss": 0.1026,
"step": 2100
},
{
"epoch": 1.08,
"learning_rate": 8.293684210526317e-06,
"loss": 0.0998,
"step": 2125
},
{
"epoch": 1.08,
"learning_rate": 8.267368421052632e-06,
"loss": 0.1381,
"step": 2150
},
{
"epoch": 1.09,
"learning_rate": 8.241052631578948e-06,
"loss": 0.1208,
"step": 2175
},
{
"epoch": 1.09,
"learning_rate": 8.214736842105265e-06,
"loss": 0.1197,
"step": 2200
},
{
"epoch": 1.09,
"learning_rate": 8.18842105263158e-06,
"loss": 0.1163,
"step": 2225
},
{
"epoch": 1.09,
"learning_rate": 8.162105263157895e-06,
"loss": 0.1117,
"step": 2250
},
{
"epoch": 1.1,
"learning_rate": 8.135789473684212e-06,
"loss": 0.0971,
"step": 2275
},
{
"epoch": 1.1,
"learning_rate": 8.109473684210527e-06,
"loss": 0.0996,
"step": 2300
},
{
"epoch": 1.1,
"learning_rate": 8.083157894736843e-06,
"loss": 0.1186,
"step": 2325
},
{
"epoch": 1.1,
"learning_rate": 8.056842105263158e-06,
"loss": 0.1379,
"step": 2350
},
{
"epoch": 1.11,
"learning_rate": 8.030526315789475e-06,
"loss": 0.118,
"step": 2375
},
{
"epoch": 1.11,
"learning_rate": 8.00421052631579e-06,
"loss": 0.1023,
"step": 2400
},
{
"epoch": 1.11,
"learning_rate": 7.977894736842105e-06,
"loss": 0.0809,
"step": 2425
},
{
"epoch": 1.11,
"learning_rate": 7.951578947368421e-06,
"loss": 0.0902,
"step": 2450
},
{
"epoch": 1.12,
"learning_rate": 7.925263157894736e-06,
"loss": 0.0836,
"step": 2475
},
{
"epoch": 1.12,
"learning_rate": 7.898947368421053e-06,
"loss": 0.0899,
"step": 2500
},
{
"epoch": 1.12,
"learning_rate": 7.87263157894737e-06,
"loss": 0.0789,
"step": 2525
},
{
"epoch": 1.12,
"learning_rate": 7.846315789473685e-06,
"loss": 0.0982,
"step": 2550
},
{
"epoch": 1.13,
"learning_rate": 7.820000000000001e-06,
"loss": 0.1085,
"step": 2575
},
{
"epoch": 1.13,
"learning_rate": 7.793684210526316e-06,
"loss": 0.1058,
"step": 2600
},
{
"epoch": 2.0,
"learning_rate": 7.767368421052633e-06,
"loss": 0.0932,
"step": 2625
},
{
"epoch": 2.0,
"learning_rate": 7.741052631578948e-06,
"loss": 0.0796,
"step": 2650
},
{
"epoch": 2.01,
"learning_rate": 7.714736842105263e-06,
"loss": 0.0895,
"step": 2675
},
{
"epoch": 2.01,
"learning_rate": 7.68842105263158e-06,
"loss": 0.0862,
"step": 2700
},
{
"epoch": 2.01,
"learning_rate": 7.662105263157895e-06,
"loss": 0.0945,
"step": 2725
},
{
"epoch": 2.01,
"learning_rate": 7.635789473684211e-06,
"loss": 0.0812,
"step": 2750
},
{
"epoch": 2.02,
"learning_rate": 7.609473684210526e-06,
"loss": 0.0798,
"step": 2775
},
{
"epoch": 2.02,
"learning_rate": 7.583157894736842e-06,
"loss": 0.0922,
"step": 2800
},
{
"epoch": 2.02,
"learning_rate": 7.556842105263158e-06,
"loss": 0.0878,
"step": 2825
},
{
"epoch": 2.02,
"learning_rate": 7.5305263157894745e-06,
"loss": 0.0864,
"step": 2850
},
{
"epoch": 2.03,
"learning_rate": 7.50421052631579e-06,
"loss": 0.0804,
"step": 2875
},
{
"epoch": 2.03,
"learning_rate": 7.477894736842106e-06,
"loss": 0.08,
"step": 2900
},
{
"epoch": 2.03,
"learning_rate": 7.451578947368422e-06,
"loss": 0.0641,
"step": 2925
},
{
"epoch": 2.03,
"learning_rate": 7.425263157894738e-06,
"loss": 0.0784,
"step": 2950
},
{
"epoch": 2.04,
"learning_rate": 7.398947368421054e-06,
"loss": 0.076,
"step": 2975
},
{
"epoch": 2.04,
"learning_rate": 7.3726315789473694e-06,
"loss": 0.0765,
"step": 3000
},
{
"epoch": 2.04,
"eval_loss": 0.3399508595466614,
"eval_runtime": 2652.4052,
"eval_samples_per_second": 3.936,
"eval_steps_per_second": 0.984,
"eval_wer": 54.394666666666666,
"step": 3000
},
{
"epoch": 2.04,
"learning_rate": 7.346315789473684e-06,
"loss": 0.0745,
"step": 3025
},
{
"epoch": 2.04,
"learning_rate": 7.32e-06,
"loss": 0.0754,
"step": 3050
},
{
"epoch": 2.05,
"learning_rate": 7.293684210526316e-06,
"loss": 0.0746,
"step": 3075
},
{
"epoch": 2.05,
"learning_rate": 7.267368421052632e-06,
"loss": 0.0663,
"step": 3100
},
{
"epoch": 2.05,
"learning_rate": 7.241052631578948e-06,
"loss": 0.0612,
"step": 3125
},
{
"epoch": 2.05,
"learning_rate": 7.2147368421052635e-06,
"loss": 0.06,
"step": 3150
},
{
"epoch": 2.06,
"learning_rate": 7.18842105263158e-06,
"loss": 0.0489,
"step": 3175
},
{
"epoch": 2.06,
"learning_rate": 7.162105263157896e-06,
"loss": 0.0553,
"step": 3200
},
{
"epoch": 2.06,
"learning_rate": 7.135789473684212e-06,
"loss": 0.0563,
"step": 3225
},
{
"epoch": 2.06,
"learning_rate": 7.109473684210528e-06,
"loss": 0.0565,
"step": 3250
},
{
"epoch": 2.07,
"learning_rate": 7.083157894736843e-06,
"loss": 0.0626,
"step": 3275
},
{
"epoch": 2.07,
"learning_rate": 7.056842105263158e-06,
"loss": 0.0494,
"step": 3300
},
{
"epoch": 2.07,
"learning_rate": 7.030526315789474e-06,
"loss": 0.0476,
"step": 3325
},
{
"epoch": 2.07,
"learning_rate": 7.00421052631579e-06,
"loss": 0.0506,
"step": 3350
},
{
"epoch": 2.08,
"learning_rate": 6.977894736842106e-06,
"loss": 0.0538,
"step": 3375
},
{
"epoch": 2.08,
"learning_rate": 6.951578947368422e-06,
"loss": 0.0443,
"step": 3400
},
{
"epoch": 2.08,
"learning_rate": 6.9252631578947375e-06,
"loss": 0.0528,
"step": 3425
},
{
"epoch": 2.08,
"learning_rate": 6.8989473684210525e-06,
"loss": 0.0664,
"step": 3450
},
{
"epoch": 2.09,
"learning_rate": 6.872631578947368e-06,
"loss": 0.0728,
"step": 3475
},
{
"epoch": 2.09,
"learning_rate": 6.846315789473684e-06,
"loss": 0.0546,
"step": 3500
},
{
"epoch": 2.09,
"learning_rate": 6.820000000000001e-06,
"loss": 0.0624,
"step": 3525
},
{
"epoch": 2.09,
"learning_rate": 6.793684210526317e-06,
"loss": 0.0608,
"step": 3550
},
{
"epoch": 2.1,
"learning_rate": 6.767368421052632e-06,
"loss": 0.0554,
"step": 3575
},
{
"epoch": 2.1,
"learning_rate": 6.741052631578948e-06,
"loss": 0.0502,
"step": 3600
},
{
"epoch": 2.1,
"learning_rate": 6.714736842105264e-06,
"loss": 0.0676,
"step": 3625
},
{
"epoch": 2.1,
"learning_rate": 6.68842105263158e-06,
"loss": 0.0646,
"step": 3650
},
{
"epoch": 2.11,
"learning_rate": 6.662105263157896e-06,
"loss": 0.0629,
"step": 3675
},
{
"epoch": 2.11,
"learning_rate": 6.635789473684211e-06,
"loss": 0.0529,
"step": 3700
},
{
"epoch": 2.11,
"learning_rate": 6.6094736842105265e-06,
"loss": 0.0557,
"step": 3725
},
{
"epoch": 2.11,
"learning_rate": 6.583157894736842e-06,
"loss": 0.0474,
"step": 3750
},
{
"epoch": 2.12,
"learning_rate": 6.556842105263158e-06,
"loss": 0.0454,
"step": 3775
},
{
"epoch": 2.12,
"learning_rate": 6.530526315789474e-06,
"loss": 0.0463,
"step": 3800
},
{
"epoch": 2.12,
"learning_rate": 6.50421052631579e-06,
"loss": 0.0468,
"step": 3825
},
{
"epoch": 2.12,
"learning_rate": 6.477894736842106e-06,
"loss": 0.0415,
"step": 3850
},
{
"epoch": 2.13,
"learning_rate": 6.451578947368422e-06,
"loss": 0.0546,
"step": 3875
},
{
"epoch": 2.13,
"learning_rate": 6.425263157894738e-06,
"loss": 0.0622,
"step": 3900
},
{
"epoch": 3.0,
"learning_rate": 6.398947368421054e-06,
"loss": 0.0492,
"step": 3925
},
{
"epoch": 3.0,
"learning_rate": 6.372631578947369e-06,
"loss": 0.0442,
"step": 3950
},
{
"epoch": 3.01,
"learning_rate": 6.346315789473685e-06,
"loss": 0.042,
"step": 3975
},
{
"epoch": 3.01,
"learning_rate": 6.3200000000000005e-06,
"loss": 0.0417,
"step": 4000
},
{
"epoch": 3.01,
"eval_loss": 0.359872967004776,
"eval_runtime": 2619.1281,
"eval_samples_per_second": 3.986,
"eval_steps_per_second": 0.997,
"eval_wer": 52.532000000000004,
"step": 4000
},
{
"epoch": 3.01,
"learning_rate": 6.293684210526316e-06,
"loss": 0.0426,
"step": 4025
},
{
"epoch": 3.01,
"learning_rate": 6.267368421052632e-06,
"loss": 0.0386,
"step": 4050
},
{
"epoch": 3.02,
"learning_rate": 6.241052631578948e-06,
"loss": 0.0446,
"step": 4075
},
{
"epoch": 3.02,
"learning_rate": 6.214736842105264e-06,
"loss": 0.0439,
"step": 4100
},
{
"epoch": 3.02,
"learning_rate": 6.188421052631579e-06,
"loss": 0.0424,
"step": 4125
},
{
"epoch": 3.02,
"learning_rate": 6.1621052631578945e-06,
"loss": 0.0443,
"step": 4150
},
{
"epoch": 3.03,
"learning_rate": 6.13578947368421e-06,
"loss": 0.0415,
"step": 4175
},
{
"epoch": 3.03,
"learning_rate": 6.109473684210527e-06,
"loss": 0.0485,
"step": 4200
},
{
"epoch": 3.03,
"learning_rate": 6.083157894736843e-06,
"loss": 0.0394,
"step": 4225
},
{
"epoch": 3.03,
"learning_rate": 6.056842105263159e-06,
"loss": 0.0407,
"step": 4250
},
{
"epoch": 3.04,
"learning_rate": 6.0305263157894745e-06,
"loss": 0.0403,
"step": 4275
},
{
"epoch": 3.04,
"learning_rate": 6.00421052631579e-06,
"loss": 0.045,
"step": 4300
},
{
"epoch": 3.04,
"learning_rate": 5.977894736842106e-06,
"loss": 0.0361,
"step": 4325
},
{
"epoch": 3.04,
"learning_rate": 5.951578947368422e-06,
"loss": 0.0402,
"step": 4350
},
{
"epoch": 3.05,
"learning_rate": 5.925263157894737e-06,
"loss": 0.0404,
"step": 4375
},
{
"epoch": 3.05,
"learning_rate": 5.898947368421053e-06,
"loss": 0.0314,
"step": 4400
},
{
"epoch": 3.05,
"learning_rate": 5.8726315789473685e-06,
"loss": 0.0344,
"step": 4425
},
{
"epoch": 3.05,
"learning_rate": 5.846315789473684e-06,
"loss": 0.0312,
"step": 4450
},
{
"epoch": 3.06,
"learning_rate": 5.82e-06,
"loss": 0.0231,
"step": 4475
},
{
"epoch": 3.06,
"learning_rate": 5.793684210526316e-06,
"loss": 0.0317,
"step": 4500
},
{
"epoch": 3.06,
"learning_rate": 5.767368421052632e-06,
"loss": 0.033,
"step": 4525
},
{
"epoch": 3.06,
"learning_rate": 5.7410526315789485e-06,
"loss": 0.03,
"step": 4550
},
{
"epoch": 3.07,
"learning_rate": 5.714736842105264e-06,
"loss": 0.0329,
"step": 4575
},
{
"epoch": 3.07,
"learning_rate": 5.68842105263158e-06,
"loss": 0.0294,
"step": 4600
},
{
"epoch": 3.07,
"learning_rate": 5.662105263157895e-06,
"loss": 0.0225,
"step": 4625
},
{
"epoch": 3.07,
"learning_rate": 5.635789473684211e-06,
"loss": 0.0238,
"step": 4650
},
{
"epoch": 3.08,
"learning_rate": 5.609473684210527e-06,
"loss": 0.0261,
"step": 4675
},
{
"epoch": 3.08,
"learning_rate": 5.5831578947368425e-06,
"loss": 0.0232,
"step": 4700
},
{
"epoch": 3.08,
"learning_rate": 5.556842105263158e-06,
"loss": 0.0289,
"step": 4725
},
{
"epoch": 3.08,
"learning_rate": 5.530526315789474e-06,
"loss": 0.032,
"step": 4750
},
{
"epoch": 3.09,
"learning_rate": 5.50421052631579e-06,
"loss": 0.0353,
"step": 4775
},
{
"epoch": 3.09,
"learning_rate": 5.477894736842105e-06,
"loss": 0.0347,
"step": 4800
},
{
"epoch": 3.09,
"learning_rate": 5.451578947368421e-06,
"loss": 0.0286,
"step": 4825
},
{
"epoch": 3.09,
"learning_rate": 5.425263157894737e-06,
"loss": 0.0272,
"step": 4850
},
{
"epoch": 3.1,
"learning_rate": 5.398947368421053e-06,
"loss": 0.0248,
"step": 4875
},
{
"epoch": 3.1,
"learning_rate": 5.372631578947369e-06,
"loss": 0.0252,
"step": 4900
},
{
"epoch": 3.1,
"learning_rate": 5.346315789473685e-06,
"loss": 0.032,
"step": 4925
},
{
"epoch": 3.1,
"learning_rate": 5.320000000000001e-06,
"loss": 0.0345,
"step": 4950
},
{
"epoch": 3.11,
"learning_rate": 5.2936842105263166e-06,
"loss": 0.0315,
"step": 4975
},
{
"epoch": 3.11,
"learning_rate": 5.267368421052632e-06,
"loss": 0.0364,
"step": 5000
},
{
"epoch": 3.11,
"eval_loss": 0.3740359842777252,
"eval_runtime": 2652.088,
"eval_samples_per_second": 3.937,
"eval_steps_per_second": 0.984,
"eval_wer": 55.565333333333335,
"step": 5000
},
{
"epoch": 3.11,
"learning_rate": 5.241052631578948e-06,
"loss": 0.0273,
"step": 5025
},
{
"epoch": 3.11,
"learning_rate": 5.214736842105263e-06,
"loss": 0.0206,
"step": 5050
},
{
"epoch": 3.12,
"learning_rate": 5.188421052631579e-06,
"loss": 0.021,
"step": 5075
},
{
"epoch": 3.12,
"learning_rate": 5.162105263157895e-06,
"loss": 0.0211,
"step": 5100
},
{
"epoch": 3.12,
"learning_rate": 5.135789473684211e-06,
"loss": 0.0234,
"step": 5125
},
{
"epoch": 3.12,
"learning_rate": 5.1094736842105264e-06,
"loss": 0.0272,
"step": 5150
},
{
"epoch": 3.13,
"learning_rate": 5.083157894736842e-06,
"loss": 0.026,
"step": 5175
},
{
"epoch": 3.13,
"learning_rate": 5.056842105263158e-06,
"loss": 0.0255,
"step": 5200
},
{
"epoch": 4.0,
"learning_rate": 5.030526315789475e-06,
"loss": 0.0283,
"step": 5225
},
{
"epoch": 4.0,
"learning_rate": 5.0042105263157906e-06,
"loss": 0.0228,
"step": 5250
},
{
"epoch": 4.01,
"learning_rate": 4.9778947368421055e-06,
"loss": 0.0249,
"step": 5275
},
{
"epoch": 4.01,
"learning_rate": 4.951578947368421e-06,
"loss": 0.0217,
"step": 5300
},
{
"epoch": 4.01,
"learning_rate": 4.925263157894737e-06,
"loss": 0.0246,
"step": 5325
},
{
"epoch": 4.01,
"learning_rate": 4.898947368421053e-06,
"loss": 0.0207,
"step": 5350
},
{
"epoch": 4.02,
"learning_rate": 4.872631578947369e-06,
"loss": 0.019,
"step": 5375
},
{
"epoch": 4.02,
"learning_rate": 4.846315789473685e-06,
"loss": 0.0184,
"step": 5400
},
{
"epoch": 4.02,
"learning_rate": 4.8200000000000004e-06,
"loss": 0.0215,
"step": 5425
},
{
"epoch": 4.02,
"learning_rate": 4.793684210526316e-06,
"loss": 0.0215,
"step": 5450
},
{
"epoch": 4.03,
"learning_rate": 4.767368421052632e-06,
"loss": 0.02,
"step": 5475
},
{
"epoch": 4.03,
"learning_rate": 4.741052631578948e-06,
"loss": 0.0196,
"step": 5500
},
{
"epoch": 4.03,
"learning_rate": 4.714736842105264e-06,
"loss": 0.0183,
"step": 5525
},
{
"epoch": 4.03,
"learning_rate": 4.6884210526315795e-06,
"loss": 0.021,
"step": 5550
},
{
"epoch": 4.04,
"learning_rate": 4.662105263157895e-06,
"loss": 0.0181,
"step": 5575
},
{
"epoch": 4.04,
"learning_rate": 4.63578947368421e-06,
"loss": 0.0224,
"step": 5600
},
{
"epoch": 4.04,
"learning_rate": 4.609473684210526e-06,
"loss": 0.0174,
"step": 5625
},
{
"epoch": 4.04,
"learning_rate": 4.583157894736843e-06,
"loss": 0.0198,
"step": 5650
},
{
"epoch": 4.05,
"learning_rate": 4.556842105263159e-06,
"loss": 0.0157,
"step": 5675
},
{
"epoch": 4.05,
"learning_rate": 4.5305263157894744e-06,
"loss": 0.0156,
"step": 5700
},
{
"epoch": 4.05,
"learning_rate": 4.504210526315789e-06,
"loss": 0.0189,
"step": 5725
},
{
"epoch": 4.05,
"learning_rate": 4.477894736842105e-06,
"loss": 0.0139,
"step": 5750
},
{
"epoch": 4.06,
"learning_rate": 4.451578947368421e-06,
"loss": 0.0139,
"step": 5775
},
{
"epoch": 4.06,
"learning_rate": 4.425263157894737e-06,
"loss": 0.0132,
"step": 5800
},
{
"epoch": 4.06,
"learning_rate": 4.3989473684210535e-06,
"loss": 0.0153,
"step": 5825
},
{
"epoch": 4.06,
"learning_rate": 4.3726315789473685e-06,
"loss": 0.0158,
"step": 5850
},
{
"epoch": 4.07,
"learning_rate": 4.346315789473684e-06,
"loss": 0.0146,
"step": 5875
},
{
"epoch": 4.07,
"learning_rate": 4.32e-06,
"loss": 0.0138,
"step": 5900
},
{
"epoch": 4.07,
"learning_rate": 4.293684210526316e-06,
"loss": 0.0114,
"step": 5925
},
{
"epoch": 4.07,
"learning_rate": 4.267368421052632e-06,
"loss": 0.0109,
"step": 5950
},
{
"epoch": 4.08,
"learning_rate": 4.241052631578948e-06,
"loss": 0.0088,
"step": 5975
},
{
"epoch": 4.08,
"learning_rate": 4.214736842105263e-06,
"loss": 0.0094,
"step": 6000
},
{
"epoch": 4.08,
"eval_loss": 0.4152411222457886,
"eval_runtime": 2659.2972,
"eval_samples_per_second": 3.926,
"eval_steps_per_second": 0.981,
"eval_wer": 56.43066666666666,
"step": 6000
},
{
"epoch": 4.08,
"learning_rate": 4.188421052631579e-06,
"loss": 0.0102,
"step": 6025
},
{
"epoch": 4.08,
"learning_rate": 4.162105263157895e-06,
"loss": 0.0116,
"step": 6050
},
{
"epoch": 4.09,
"learning_rate": 4.135789473684211e-06,
"loss": 0.0164,
"step": 6075
},
{
"epoch": 4.09,
"learning_rate": 4.109473684210527e-06,
"loss": 0.0162,
"step": 6100
},
{
"epoch": 4.09,
"learning_rate": 4.0831578947368425e-06,
"loss": 0.0159,
"step": 6125
},
{
"epoch": 4.09,
"learning_rate": 4.056842105263158e-06,
"loss": 0.0103,
"step": 6150
},
{
"epoch": 4.1,
"learning_rate": 4.030526315789474e-06,
"loss": 0.0135,
"step": 6175
},
{
"epoch": 4.1,
"learning_rate": 4.00421052631579e-06,
"loss": 0.0092,
"step": 6200
},
{
"epoch": 4.1,
"learning_rate": 3.977894736842106e-06,
"loss": 0.0145,
"step": 6225
},
{
"epoch": 4.1,
"learning_rate": 3.951578947368422e-06,
"loss": 0.0157,
"step": 6250
},
{
"epoch": 4.11,
"learning_rate": 3.9252631578947366e-06,
"loss": 0.0137,
"step": 6275
},
{
"epoch": 4.11,
"learning_rate": 3.898947368421052e-06,
"loss": 0.0165,
"step": 6300
},
{
"epoch": 4.11,
"learning_rate": 3.872631578947369e-06,
"loss": 0.0111,
"step": 6325
},
{
"epoch": 4.11,
"learning_rate": 3.846315789473685e-06,
"loss": 0.0094,
"step": 6350
},
{
"epoch": 4.12,
"learning_rate": 3.820000000000001e-06,
"loss": 0.0101,
"step": 6375
},
{
"epoch": 4.12,
"learning_rate": 3.793684210526316e-06,
"loss": 0.0094,
"step": 6400
},
{
"epoch": 4.12,
"learning_rate": 3.767368421052632e-06,
"loss": 0.0111,
"step": 6425
},
{
"epoch": 4.12,
"learning_rate": 3.7410526315789473e-06,
"loss": 0.0126,
"step": 6450
},
{
"epoch": 4.13,
"learning_rate": 3.714736842105263e-06,
"loss": 0.0113,
"step": 6475
},
{
"epoch": 4.13,
"learning_rate": 3.6884210526315794e-06,
"loss": 0.0129,
"step": 6500
},
{
"epoch": 4.13,
"learning_rate": 3.662105263157895e-06,
"loss": 0.0115,
"step": 6525
},
{
"epoch": 5.0,
"learning_rate": 3.635789473684211e-06,
"loss": 0.0105,
"step": 6550
},
{
"epoch": 5.0,
"learning_rate": 3.6094736842105264e-06,
"loss": 0.0098,
"step": 6575
},
{
"epoch": 5.01,
"learning_rate": 3.5831578947368422e-06,
"loss": 0.0123,
"step": 6600
},
{
"epoch": 5.01,
"learning_rate": 3.556842105263158e-06,
"loss": 0.0126,
"step": 6625
},
{
"epoch": 5.01,
"learning_rate": 3.5305263157894743e-06,
"loss": 0.0092,
"step": 6650
},
{
"epoch": 5.01,
"learning_rate": 3.50421052631579e-06,
"loss": 0.0096,
"step": 6675
},
{
"epoch": 5.02,
"learning_rate": 3.4778947368421055e-06,
"loss": 0.0081,
"step": 6700
},
{
"epoch": 5.02,
"learning_rate": 3.4515789473684213e-06,
"loss": 0.0058,
"step": 6725
},
{
"epoch": 5.02,
"learning_rate": 3.425263157894737e-06,
"loss": 0.0078,
"step": 6750
},
{
"epoch": 5.03,
"learning_rate": 3.398947368421053e-06,
"loss": 0.0091,
"step": 6775
},
{
"epoch": 5.03,
"learning_rate": 3.3726315789473683e-06,
"loss": 0.0112,
"step": 6800
},
{
"epoch": 5.03,
"learning_rate": 3.3463157894736846e-06,
"loss": 0.0095,
"step": 6825
},
{
"epoch": 5.03,
"learning_rate": 3.3200000000000004e-06,
"loss": 0.0083,
"step": 6850
},
{
"epoch": 5.04,
"learning_rate": 3.2936842105263162e-06,
"loss": 0.0073,
"step": 6875
},
{
"epoch": 5.04,
"learning_rate": 3.267368421052632e-06,
"loss": 0.0098,
"step": 6900
},
{
"epoch": 5.04,
"learning_rate": 3.2410526315789474e-06,
"loss": 0.0087,
"step": 6925
},
{
"epoch": 5.04,
"learning_rate": 3.2147368421052633e-06,
"loss": 0.0071,
"step": 6950
},
{
"epoch": 5.04,
"learning_rate": 3.188421052631579e-06,
"loss": 0.008,
"step": 6975
},
{
"epoch": 5.05,
"learning_rate": 3.1621052631578953e-06,
"loss": 0.0077,
"step": 7000
},
{
"epoch": 5.05,
"eval_loss": 0.42180904746055603,
"eval_runtime": 2581.8065,
"eval_samples_per_second": 4.044,
"eval_steps_per_second": 1.011,
"eval_wer": 47.53066666666667,
"step": 7000
},
{
"epoch": 5.05,
"learning_rate": 3.135789473684211e-06,
"loss": 0.0063,
"step": 7025
},
{
"epoch": 5.05,
"learning_rate": 3.1094736842105265e-06,
"loss": 0.0096,
"step": 7050
},
{
"epoch": 5.05,
"learning_rate": 3.0831578947368423e-06,
"loss": 0.0054,
"step": 7075
},
{
"epoch": 5.06,
"learning_rate": 3.056842105263158e-06,
"loss": 0.0049,
"step": 7100
},
{
"epoch": 5.06,
"learning_rate": 3.0305263157894736e-06,
"loss": 0.007,
"step": 7125
},
{
"epoch": 5.06,
"learning_rate": 3.0042105263157894e-06,
"loss": 0.0096,
"step": 7150
},
{
"epoch": 5.07,
"learning_rate": 2.9778947368421056e-06,
"loss": 0.0085,
"step": 7175
},
{
"epoch": 5.07,
"learning_rate": 2.9515789473684214e-06,
"loss": 0.0078,
"step": 7200
},
{
"epoch": 5.07,
"learning_rate": 2.9252631578947373e-06,
"loss": 0.004,
"step": 7225
},
{
"epoch": 5.07,
"learning_rate": 2.8989473684210526e-06,
"loss": 0.005,
"step": 7250
},
{
"epoch": 5.08,
"learning_rate": 2.8726315789473685e-06,
"loss": 0.0043,
"step": 7275
},
{
"epoch": 5.08,
"learning_rate": 2.8463157894736843e-06,
"loss": 0.0032,
"step": 7300
},
{
"epoch": 5.08,
"learning_rate": 2.82e-06,
"loss": 0.0057,
"step": 7325
},
{
"epoch": 5.08,
"learning_rate": 2.7936842105263163e-06,
"loss": 0.0059,
"step": 7350
},
{
"epoch": 5.08,
"learning_rate": 2.7673684210526317e-06,
"loss": 0.0055,
"step": 7375
},
{
"epoch": 5.09,
"learning_rate": 2.7410526315789476e-06,
"loss": 0.0058,
"step": 7400
},
{
"epoch": 5.09,
"learning_rate": 2.7147368421052634e-06,
"loss": 0.0067,
"step": 7425
},
{
"epoch": 5.09,
"learning_rate": 2.688421052631579e-06,
"loss": 0.0047,
"step": 7450
},
{
"epoch": 5.09,
"learning_rate": 2.6621052631578946e-06,
"loss": 0.0045,
"step": 7475
},
{
"epoch": 5.1,
"learning_rate": 2.635789473684211e-06,
"loss": 0.0061,
"step": 7500
},
{
"epoch": 5.1,
"learning_rate": 2.6094736842105267e-06,
"loss": 0.0052,
"step": 7525
},
{
"epoch": 5.1,
"learning_rate": 2.5831578947368425e-06,
"loss": 0.0088,
"step": 7550
},
{
"epoch": 5.11,
"learning_rate": 2.5568421052631583e-06,
"loss": 0.0086,
"step": 7575
},
{
"epoch": 5.11,
"learning_rate": 2.5305263157894737e-06,
"loss": 0.0061,
"step": 7600
},
{
"epoch": 5.11,
"learning_rate": 2.5042105263157895e-06,
"loss": 0.0037,
"step": 7625
},
{
"epoch": 5.11,
"learning_rate": 2.4778947368421053e-06,
"loss": 0.0035,
"step": 7650
},
{
"epoch": 5.12,
"learning_rate": 2.451578947368421e-06,
"loss": 0.0054,
"step": 7675
},
{
"epoch": 5.12,
"learning_rate": 2.425263157894737e-06,
"loss": 0.0037,
"step": 7700
},
{
"epoch": 5.12,
"learning_rate": 2.3989473684210528e-06,
"loss": 0.0047,
"step": 7725
},
{
"epoch": 5.12,
"learning_rate": 2.3726315789473686e-06,
"loss": 0.0039,
"step": 7750
},
{
"epoch": 5.12,
"learning_rate": 2.3463157894736844e-06,
"loss": 0.0044,
"step": 7775
},
{
"epoch": 5.13,
"learning_rate": 2.3200000000000002e-06,
"loss": 0.0059,
"step": 7800
},
{
"epoch": 5.13,
"learning_rate": 2.293684210526316e-06,
"loss": 0.0036,
"step": 7825
},
{
"epoch": 6.0,
"learning_rate": 2.267368421052632e-06,
"loss": 0.0063,
"step": 7850
},
{
"epoch": 6.0,
"learning_rate": 2.2410526315789473e-06,
"loss": 0.0052,
"step": 7875
},
{
"epoch": 6.01,
"learning_rate": 2.2147368421052635e-06,
"loss": 0.0041,
"step": 7900
},
{
"epoch": 6.01,
"learning_rate": 2.188421052631579e-06,
"loss": 0.0035,
"step": 7925
},
{
"epoch": 6.01,
"learning_rate": 2.1621052631578947e-06,
"loss": 0.0036,
"step": 7950
},
{
"epoch": 6.01,
"learning_rate": 2.135789473684211e-06,
"loss": 0.0032,
"step": 7975
},
{
"epoch": 6.02,
"learning_rate": 2.1094736842105264e-06,
"loss": 0.0018,
"step": 8000
},
{
"epoch": 6.02,
"eval_loss": 0.4556337594985962,
"eval_runtime": 2619.7381,
"eval_samples_per_second": 3.985,
"eval_steps_per_second": 0.996,
"eval_wer": 50.04933333333334,
"step": 8000
},
{
"epoch": 6.02,
"learning_rate": 2.083157894736842e-06,
"loss": 0.0019,
"step": 8025
},
{
"epoch": 6.02,
"learning_rate": 2.056842105263158e-06,
"loss": 0.005,
"step": 8050
},
{
"epoch": 6.02,
"learning_rate": 2.030526315789474e-06,
"loss": 0.0031,
"step": 8075
},
{
"epoch": 6.03,
"learning_rate": 2.0042105263157896e-06,
"loss": 0.0035,
"step": 8100
},
{
"epoch": 6.03,
"learning_rate": 1.9778947368421055e-06,
"loss": 0.0028,
"step": 8125
},
{
"epoch": 6.03,
"learning_rate": 1.9515789473684213e-06,
"loss": 0.0018,
"step": 8150
},
{
"epoch": 6.03,
"learning_rate": 1.925263157894737e-06,
"loss": 0.0034,
"step": 8175
},
{
"epoch": 6.04,
"learning_rate": 1.8989473684210527e-06,
"loss": 0.0043,
"step": 8200
},
{
"epoch": 6.04,
"learning_rate": 1.8726315789473687e-06,
"loss": 0.0035,
"step": 8225
},
{
"epoch": 6.04,
"learning_rate": 1.8463157894736843e-06,
"loss": 0.0029,
"step": 8250
},
{
"epoch": 6.04,
"learning_rate": 1.8200000000000002e-06,
"loss": 0.0032,
"step": 8275
},
{
"epoch": 6.05,
"learning_rate": 1.7936842105263158e-06,
"loss": 0.0033,
"step": 8300
},
{
"epoch": 6.05,
"learning_rate": 1.7673684210526318e-06,
"loss": 0.0029,
"step": 8325
},
{
"epoch": 6.05,
"learning_rate": 1.7410526315789474e-06,
"loss": 0.002,
"step": 8350
},
{
"epoch": 6.05,
"learning_rate": 1.7147368421052632e-06,
"loss": 0.0015,
"step": 8375
},
{
"epoch": 6.06,
"learning_rate": 1.6884210526315792e-06,
"loss": 0.0018,
"step": 8400
},
{
"epoch": 6.06,
"learning_rate": 1.6621052631578948e-06,
"loss": 0.0024,
"step": 8425
},
{
"epoch": 6.06,
"learning_rate": 1.6357894736842107e-06,
"loss": 0.0025,
"step": 8450
},
{
"epoch": 6.06,
"learning_rate": 1.6094736842105265e-06,
"loss": 0.0024,
"step": 8475
},
{
"epoch": 6.07,
"learning_rate": 1.5831578947368423e-06,
"loss": 0.0023,
"step": 8500
},
{
"epoch": 6.07,
"learning_rate": 1.556842105263158e-06,
"loss": 0.002,
"step": 8525
},
{
"epoch": 6.07,
"learning_rate": 1.5305263157894737e-06,
"loss": 0.0014,
"step": 8550
},
{
"epoch": 6.07,
"learning_rate": 1.5042105263157898e-06,
"loss": 0.0013,
"step": 8575
},
{
"epoch": 6.08,
"learning_rate": 1.4778947368421054e-06,
"loss": 0.0023,
"step": 8600
},
{
"epoch": 6.08,
"learning_rate": 1.4515789473684212e-06,
"loss": 0.0018,
"step": 8625
},
{
"epoch": 6.08,
"learning_rate": 1.425263157894737e-06,
"loss": 0.0029,
"step": 8650
},
{
"epoch": 6.08,
"learning_rate": 1.3989473684210528e-06,
"loss": 0.0019,
"step": 8675
},
{
"epoch": 6.09,
"learning_rate": 1.3726315789473684e-06,
"loss": 0.0019,
"step": 8700
},
{
"epoch": 6.09,
"learning_rate": 1.3463157894736842e-06,
"loss": 0.004,
"step": 8725
},
{
"epoch": 6.09,
"learning_rate": 1.32e-06,
"loss": 0.0013,
"step": 8750
},
{
"epoch": 6.09,
"learning_rate": 1.2936842105263159e-06,
"loss": 0.0013,
"step": 8775
},
{
"epoch": 6.1,
"learning_rate": 1.2673684210526315e-06,
"loss": 0.0012,
"step": 8800
},
{
"epoch": 6.1,
"learning_rate": 1.2410526315789475e-06,
"loss": 0.0013,
"step": 8825
},
{
"epoch": 6.1,
"learning_rate": 1.2147368421052633e-06,
"loss": 0.0026,
"step": 8850
},
{
"epoch": 6.1,
"learning_rate": 1.188421052631579e-06,
"loss": 0.004,
"step": 8875
},
{
"epoch": 6.11,
"learning_rate": 1.1621052631578948e-06,
"loss": 0.0019,
"step": 8900
},
{
"epoch": 6.11,
"learning_rate": 1.1357894736842106e-06,
"loss": 0.0018,
"step": 8925
},
{
"epoch": 6.11,
"learning_rate": 1.1094736842105264e-06,
"loss": 0.0016,
"step": 8950
},
{
"epoch": 6.11,
"learning_rate": 1.0831578947368422e-06,
"loss": 0.0026,
"step": 8975
},
{
"epoch": 6.12,
"learning_rate": 1.0568421052631578e-06,
"loss": 0.0012,
"step": 9000
},
{
"epoch": 6.12,
"eval_loss": 0.47603604197502136,
"eval_runtime": 2657.6698,
"eval_samples_per_second": 3.928,
"eval_steps_per_second": 0.982,
"eval_wer": 54.81466666666667,
"step": 9000
},
{
"epoch": 6.12,
"learning_rate": 1.0305263157894739e-06,
"loss": 0.0016,
"step": 9025
},
{
"epoch": 6.12,
"learning_rate": 1.0042105263157897e-06,
"loss": 0.0011,
"step": 9050
},
{
"epoch": 6.12,
"learning_rate": 9.778947368421053e-07,
"loss": 0.0016,
"step": 9075
},
{
"epoch": 6.13,
"learning_rate": 9.515789473684212e-07,
"loss": 0.0012,
"step": 9100
},
{
"epoch": 6.13,
"learning_rate": 9.252631578947368e-07,
"loss": 0.0012,
"step": 9125
},
{
"epoch": 7.0,
"learning_rate": 8.989473684210527e-07,
"loss": 0.0013,
"step": 9150
},
{
"epoch": 7.0,
"learning_rate": 8.726315789473686e-07,
"loss": 0.0016,
"step": 9175
},
{
"epoch": 7.01,
"learning_rate": 8.463157894736843e-07,
"loss": 0.0014,
"step": 9200
},
{
"epoch": 7.01,
"learning_rate": 8.200000000000001e-07,
"loss": 0.0033,
"step": 9225
},
{
"epoch": 7.01,
"learning_rate": 7.936842105263158e-07,
"loss": 0.0009,
"step": 9250
},
{
"epoch": 7.01,
"learning_rate": 7.673684210526316e-07,
"loss": 0.0008,
"step": 9275
},
{
"epoch": 7.02,
"learning_rate": 7.410526315789475e-07,
"loss": 0.0009,
"step": 9300
},
{
"epoch": 7.02,
"learning_rate": 7.147368421052632e-07,
"loss": 0.0009,
"step": 9325
},
{
"epoch": 7.02,
"learning_rate": 6.884210526315791e-07,
"loss": 0.0009,
"step": 9350
},
{
"epoch": 7.02,
"learning_rate": 6.621052631578948e-07,
"loss": 0.001,
"step": 9375
},
{
"epoch": 7.03,
"learning_rate": 6.357894736842106e-07,
"loss": 0.0009,
"step": 9400
},
{
"epoch": 7.03,
"learning_rate": 6.094736842105263e-07,
"loss": 0.0018,
"step": 9425
},
{
"epoch": 7.03,
"learning_rate": 5.831578947368421e-07,
"loss": 0.0009,
"step": 9450
},
{
"epoch": 7.03,
"learning_rate": 5.56842105263158e-07,
"loss": 0.0007,
"step": 9475
},
{
"epoch": 7.04,
"learning_rate": 5.305263157894737e-07,
"loss": 0.0018,
"step": 9500
},
{
"epoch": 7.04,
"learning_rate": 5.042105263157895e-07,
"loss": 0.0015,
"step": 9525
},
{
"epoch": 7.04,
"learning_rate": 4.778947368421053e-07,
"loss": 0.0017,
"step": 9550
},
{
"epoch": 7.04,
"learning_rate": 4.5157894736842107e-07,
"loss": 0.0011,
"step": 9575
},
{
"epoch": 7.05,
"learning_rate": 4.2526315789473684e-07,
"loss": 0.0008,
"step": 9600
},
{
"epoch": 7.05,
"learning_rate": 3.9894736842105266e-07,
"loss": 0.0008,
"step": 9625
},
{
"epoch": 7.05,
"learning_rate": 3.726315789473685e-07,
"loss": 0.001,
"step": 9650
},
{
"epoch": 7.05,
"learning_rate": 3.4631578947368424e-07,
"loss": 0.0008,
"step": 9675
},
{
"epoch": 7.06,
"learning_rate": 3.2e-07,
"loss": 0.001,
"step": 9700
},
{
"epoch": 7.06,
"learning_rate": 2.936842105263158e-07,
"loss": 0.0009,
"step": 9725
},
{
"epoch": 7.06,
"learning_rate": 2.6736842105263164e-07,
"loss": 0.0011,
"step": 9750
},
{
"epoch": 7.06,
"learning_rate": 2.410526315789474e-07,
"loss": 0.0009,
"step": 9775
},
{
"epoch": 7.07,
"learning_rate": 2.1473684210526317e-07,
"loss": 0.0007,
"step": 9800
},
{
"epoch": 7.07,
"learning_rate": 1.8842105263157897e-07,
"loss": 0.0005,
"step": 9825
},
{
"epoch": 7.07,
"learning_rate": 1.6210526315789476e-07,
"loss": 0.0004,
"step": 9850
},
{
"epoch": 7.07,
"learning_rate": 1.3578947368421055e-07,
"loss": 0.0016,
"step": 9875
},
{
"epoch": 7.08,
"learning_rate": 1.0947368421052632e-07,
"loss": 0.0004,
"step": 9900
},
{
"epoch": 7.08,
"learning_rate": 8.315789473684211e-08,
"loss": 0.0007,
"step": 9925
},
{
"epoch": 7.08,
"learning_rate": 5.68421052631579e-08,
"loss": 0.0015,
"step": 9950
},
{
"epoch": 7.08,
"learning_rate": 3.0526315789473686e-08,
"loss": 0.0012,
"step": 9975
},
{
"epoch": 7.09,
"learning_rate": 4.210526315789474e-09,
"loss": 0.0009,
"step": 10000
},
{
"epoch": 7.09,
"eval_loss": 0.47105905413627625,
"eval_runtime": 2626.7715,
"eval_samples_per_second": 3.974,
"eval_steps_per_second": 0.994,
"eval_wer": 48.75333333333333,
"step": 10000
},
{
"epoch": 7.09,
"step": 10000,
"total_flos": 1.6326820826578944e+20,
"train_loss": 0.06593986138729378,
"train_runtime": 64997.5895,
"train_samples_per_second": 2.462,
"train_steps_per_second": 0.154
}
],
"max_steps": 10000,
"num_train_epochs": 9223372036854775807,
"total_flos": 1.6326820826578944e+20,
"trial_name": null,
"trial_params": null
}