whisper-medium-gl / trainer_state.json
zuazo's picture
End of training
a320bc6
raw
history blame contribute delete
No virus
51.5 kB
{
"best_metric": 7.122654525386314,
"best_model_checkpoint": "./checkpoint-9000",
"epoch": 49.02,
"eval_steps": 1000,
"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": 1.4558,
"step": 25
},
{
"epoch": 0.01,
"learning_rate": 9.400000000000001e-07,
"loss": 1.0392,
"step": 50
},
{
"epoch": 0.01,
"learning_rate": 1.44e-06,
"loss": 0.4931,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.94e-06,
"loss": 0.368,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.294,
"step": 125
},
{
"epoch": 0.01,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.2645,
"step": 150
},
{
"epoch": 0.02,
"learning_rate": 3.44e-06,
"loss": 0.2372,
"step": 175
},
{
"epoch": 0.02,
"learning_rate": 3.94e-06,
"loss": 0.2033,
"step": 200
},
{
"epoch": 1.0,
"learning_rate": 4.440000000000001e-06,
"loss": 0.196,
"step": 225
},
{
"epoch": 1.0,
"learning_rate": 4.94e-06,
"loss": 0.2127,
"step": 250
},
{
"epoch": 1.01,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.181,
"step": 275
},
{
"epoch": 1.01,
"learning_rate": 5.94e-06,
"loss": 0.1592,
"step": 300
},
{
"epoch": 1.01,
"learning_rate": 6.440000000000001e-06,
"loss": 0.1337,
"step": 325
},
{
"epoch": 1.01,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.1235,
"step": 350
},
{
"epoch": 1.02,
"learning_rate": 7.440000000000001e-06,
"loss": 0.112,
"step": 375
},
{
"epoch": 1.02,
"learning_rate": 7.94e-06,
"loss": 0.0995,
"step": 400
},
{
"epoch": 2.0,
"learning_rate": 8.44e-06,
"loss": 0.0879,
"step": 425
},
{
"epoch": 2.0,
"learning_rate": 8.94e-06,
"loss": 0.1117,
"step": 450
},
{
"epoch": 2.01,
"learning_rate": 9.440000000000001e-06,
"loss": 0.0881,
"step": 475
},
{
"epoch": 2.01,
"learning_rate": 9.940000000000001e-06,
"loss": 0.0767,
"step": 500
},
{
"epoch": 2.01,
"learning_rate": 9.976842105263158e-06,
"loss": 0.0656,
"step": 525
},
{
"epoch": 2.02,
"learning_rate": 9.950526315789475e-06,
"loss": 0.0592,
"step": 550
},
{
"epoch": 2.02,
"learning_rate": 9.92421052631579e-06,
"loss": 0.0534,
"step": 575
},
{
"epoch": 2.02,
"learning_rate": 9.897894736842107e-06,
"loss": 0.0479,
"step": 600
},
{
"epoch": 3.0,
"learning_rate": 9.871578947368422e-06,
"loss": 0.0419,
"step": 625
},
{
"epoch": 3.0,
"learning_rate": 9.845263157894738e-06,
"loss": 0.0542,
"step": 650
},
{
"epoch": 3.01,
"learning_rate": 9.818947368421053e-06,
"loss": 0.0418,
"step": 675
},
{
"epoch": 3.01,
"learning_rate": 9.79263157894737e-06,
"loss": 0.0376,
"step": 700
},
{
"epoch": 3.01,
"learning_rate": 9.766315789473685e-06,
"loss": 0.0339,
"step": 725
},
{
"epoch": 3.02,
"learning_rate": 9.74e-06,
"loss": 0.0298,
"step": 750
},
{
"epoch": 3.02,
"learning_rate": 9.713684210526317e-06,
"loss": 0.0265,
"step": 775
},
{
"epoch": 3.02,
"learning_rate": 9.687368421052632e-06,
"loss": 0.0233,
"step": 800
},
{
"epoch": 4.0,
"learning_rate": 9.661052631578948e-06,
"loss": 0.0221,
"step": 825
},
{
"epoch": 4.0,
"learning_rate": 9.634736842105265e-06,
"loss": 0.0293,
"step": 850
},
{
"epoch": 4.01,
"learning_rate": 9.60842105263158e-06,
"loss": 0.0229,
"step": 875
},
{
"epoch": 4.01,
"learning_rate": 9.582105263157897e-06,
"loss": 0.0218,
"step": 900
},
{
"epoch": 4.01,
"learning_rate": 9.555789473684211e-06,
"loss": 0.0171,
"step": 925
},
{
"epoch": 4.01,
"learning_rate": 9.529473684210528e-06,
"loss": 0.0157,
"step": 950
},
{
"epoch": 4.02,
"learning_rate": 9.503157894736843e-06,
"loss": 0.014,
"step": 975
},
{
"epoch": 4.02,
"learning_rate": 9.476842105263158e-06,
"loss": 0.0124,
"step": 1000
},
{
"epoch": 4.02,
"eval_loss": 0.21937283873558044,
"eval_runtime": 1489.5407,
"eval_samples_per_second": 4.395,
"eval_steps_per_second": 0.138,
"eval_wer": 7.538286423841059,
"step": 1000
},
{
"epoch": 5.0,
"learning_rate": 9.450526315789475e-06,
"loss": 0.0129,
"step": 1025
},
{
"epoch": 5.0,
"learning_rate": 9.42421052631579e-06,
"loss": 0.0163,
"step": 1050
},
{
"epoch": 5.01,
"learning_rate": 9.397894736842106e-06,
"loss": 0.0123,
"step": 1075
},
{
"epoch": 5.01,
"learning_rate": 9.371578947368421e-06,
"loss": 0.0147,
"step": 1100
},
{
"epoch": 5.01,
"learning_rate": 9.345263157894738e-06,
"loss": 0.0103,
"step": 1125
},
{
"epoch": 5.01,
"learning_rate": 9.318947368421053e-06,
"loss": 0.0088,
"step": 1150
},
{
"epoch": 5.02,
"learning_rate": 9.292631578947368e-06,
"loss": 0.0081,
"step": 1175
},
{
"epoch": 5.02,
"learning_rate": 9.266315789473685e-06,
"loss": 0.0063,
"step": 1200
},
{
"epoch": 6.0,
"learning_rate": 9.240000000000001e-06,
"loss": 0.0078,
"step": 1225
},
{
"epoch": 6.0,
"learning_rate": 9.213684210526316e-06,
"loss": 0.0103,
"step": 1250
},
{
"epoch": 6.01,
"learning_rate": 9.187368421052633e-06,
"loss": 0.0081,
"step": 1275
},
{
"epoch": 6.01,
"learning_rate": 9.161052631578948e-06,
"loss": 0.0073,
"step": 1300
},
{
"epoch": 6.01,
"learning_rate": 9.134736842105265e-06,
"loss": 0.0081,
"step": 1325
},
{
"epoch": 6.01,
"learning_rate": 9.10842105263158e-06,
"loss": 0.0074,
"step": 1350
},
{
"epoch": 6.02,
"learning_rate": 9.082105263157896e-06,
"loss": 0.007,
"step": 1375
},
{
"epoch": 6.02,
"learning_rate": 9.055789473684211e-06,
"loss": 0.0055,
"step": 1400
},
{
"epoch": 7.0,
"learning_rate": 9.029473684210526e-06,
"loss": 0.0062,
"step": 1425
},
{
"epoch": 7.0,
"learning_rate": 9.003157894736843e-06,
"loss": 0.0063,
"step": 1450
},
{
"epoch": 7.01,
"learning_rate": 8.976842105263158e-06,
"loss": 0.0084,
"step": 1475
},
{
"epoch": 7.01,
"learning_rate": 8.950526315789474e-06,
"loss": 0.0052,
"step": 1500
},
{
"epoch": 7.01,
"learning_rate": 8.92421052631579e-06,
"loss": 0.0056,
"step": 1525
},
{
"epoch": 7.01,
"learning_rate": 8.897894736842106e-06,
"loss": 0.0038,
"step": 1550
},
{
"epoch": 7.02,
"learning_rate": 8.871578947368423e-06,
"loss": 0.005,
"step": 1575
},
{
"epoch": 7.02,
"learning_rate": 8.845263157894738e-06,
"loss": 0.0048,
"step": 1600
},
{
"epoch": 8.0,
"learning_rate": 8.818947368421054e-06,
"loss": 0.0035,
"step": 1625
},
{
"epoch": 8.01,
"learning_rate": 8.79263157894737e-06,
"loss": 0.0056,
"step": 1650
},
{
"epoch": 8.01,
"learning_rate": 8.766315789473684e-06,
"loss": 0.0045,
"step": 1675
},
{
"epoch": 8.01,
"learning_rate": 8.740000000000001e-06,
"loss": 0.004,
"step": 1700
},
{
"epoch": 8.01,
"learning_rate": 8.713684210526316e-06,
"loss": 0.0041,
"step": 1725
},
{
"epoch": 8.02,
"learning_rate": 8.687368421052633e-06,
"loss": 0.0038,
"step": 1750
},
{
"epoch": 8.02,
"learning_rate": 8.661052631578948e-06,
"loss": 0.0041,
"step": 1775
},
{
"epoch": 8.02,
"learning_rate": 8.634736842105264e-06,
"loss": 0.003,
"step": 1800
},
{
"epoch": 9.0,
"learning_rate": 8.60842105263158e-06,
"loss": 0.0031,
"step": 1825
},
{
"epoch": 9.01,
"learning_rate": 8.582105263157894e-06,
"loss": 0.0037,
"step": 1850
},
{
"epoch": 9.01,
"learning_rate": 8.555789473684213e-06,
"loss": 0.0035,
"step": 1875
},
{
"epoch": 9.01,
"learning_rate": 8.529473684210528e-06,
"loss": 0.0027,
"step": 1900
},
{
"epoch": 9.01,
"learning_rate": 8.503157894736843e-06,
"loss": 0.0026,
"step": 1925
},
{
"epoch": 9.02,
"learning_rate": 8.47684210526316e-06,
"loss": 0.0032,
"step": 1950
},
{
"epoch": 9.02,
"learning_rate": 8.450526315789474e-06,
"loss": 0.0032,
"step": 1975
},
{
"epoch": 9.02,
"learning_rate": 8.42421052631579e-06,
"loss": 0.0027,
"step": 2000
},
{
"epoch": 9.02,
"eval_loss": 0.2399628460407257,
"eval_runtime": 1439.8654,
"eval_samples_per_second": 4.546,
"eval_steps_per_second": 0.142,
"eval_wer": 7.338231236203091,
"step": 2000
},
{
"epoch": 10.0,
"learning_rate": 8.397894736842106e-06,
"loss": 0.0048,
"step": 2025
},
{
"epoch": 10.01,
"learning_rate": 8.371578947368422e-06,
"loss": 0.0037,
"step": 2050
},
{
"epoch": 10.01,
"learning_rate": 8.345263157894737e-06,
"loss": 0.0038,
"step": 2075
},
{
"epoch": 10.01,
"learning_rate": 8.318947368421052e-06,
"loss": 0.0033,
"step": 2100
},
{
"epoch": 10.01,
"learning_rate": 8.292631578947369e-06,
"loss": 0.0025,
"step": 2125
},
{
"epoch": 10.02,
"learning_rate": 8.266315789473684e-06,
"loss": 0.0032,
"step": 2150
},
{
"epoch": 10.02,
"learning_rate": 8.24e-06,
"loss": 0.0024,
"step": 2175
},
{
"epoch": 10.02,
"learning_rate": 8.213684210526316e-06,
"loss": 0.0032,
"step": 2200
},
{
"epoch": 11.0,
"learning_rate": 8.187368421052632e-06,
"loss": 0.0032,
"step": 2225
},
{
"epoch": 11.01,
"learning_rate": 8.161052631578949e-06,
"loss": 0.0029,
"step": 2250
},
{
"epoch": 11.01,
"learning_rate": 8.134736842105264e-06,
"loss": 0.003,
"step": 2275
},
{
"epoch": 11.01,
"learning_rate": 8.10842105263158e-06,
"loss": 0.0023,
"step": 2300
},
{
"epoch": 11.01,
"learning_rate": 8.082105263157896e-06,
"loss": 0.0022,
"step": 2325
},
{
"epoch": 11.02,
"learning_rate": 8.05578947368421e-06,
"loss": 0.0029,
"step": 2350
},
{
"epoch": 11.02,
"learning_rate": 8.029473684210527e-06,
"loss": 0.0016,
"step": 2375
},
{
"epoch": 11.02,
"learning_rate": 8.003157894736842e-06,
"loss": 0.0033,
"step": 2400
},
{
"epoch": 12.0,
"learning_rate": 7.976842105263159e-06,
"loss": 0.0019,
"step": 2425
},
{
"epoch": 12.01,
"learning_rate": 7.950526315789474e-06,
"loss": 0.0022,
"step": 2450
},
{
"epoch": 12.01,
"learning_rate": 7.92421052631579e-06,
"loss": 0.0022,
"step": 2475
},
{
"epoch": 12.01,
"learning_rate": 7.897894736842106e-06,
"loss": 0.0018,
"step": 2500
},
{
"epoch": 12.01,
"learning_rate": 7.87157894736842e-06,
"loss": 0.0015,
"step": 2525
},
{
"epoch": 12.02,
"learning_rate": 7.845263157894739e-06,
"loss": 0.002,
"step": 2550
},
{
"epoch": 12.02,
"learning_rate": 7.818947368421054e-06,
"loss": 0.0038,
"step": 2575
},
{
"epoch": 12.02,
"learning_rate": 7.792631578947369e-06,
"loss": 0.0017,
"step": 2600
},
{
"epoch": 13.0,
"learning_rate": 7.766315789473685e-06,
"loss": 0.0023,
"step": 2625
},
{
"epoch": 13.01,
"learning_rate": 7.74e-06,
"loss": 0.003,
"step": 2650
},
{
"epoch": 13.01,
"learning_rate": 7.713684210526317e-06,
"loss": 0.0021,
"step": 2675
},
{
"epoch": 13.01,
"learning_rate": 7.687368421052632e-06,
"loss": 0.0021,
"step": 2700
},
{
"epoch": 13.01,
"learning_rate": 7.661052631578949e-06,
"loss": 0.0031,
"step": 2725
},
{
"epoch": 13.02,
"learning_rate": 7.634736842105264e-06,
"loss": 0.0022,
"step": 2750
},
{
"epoch": 13.02,
"learning_rate": 7.6084210526315795e-06,
"loss": 0.002,
"step": 2775
},
{
"epoch": 13.02,
"learning_rate": 7.582105263157895e-06,
"loss": 0.0014,
"step": 2800
},
{
"epoch": 14.0,
"learning_rate": 7.555789473684211e-06,
"loss": 0.0018,
"step": 2825
},
{
"epoch": 14.01,
"learning_rate": 7.529473684210526e-06,
"loss": 0.0019,
"step": 2850
},
{
"epoch": 14.01,
"learning_rate": 7.503157894736842e-06,
"loss": 0.0022,
"step": 2875
},
{
"epoch": 14.01,
"learning_rate": 7.476842105263159e-06,
"loss": 0.0018,
"step": 2900
},
{
"epoch": 14.01,
"learning_rate": 7.4505263157894744e-06,
"loss": 0.0024,
"step": 2925
},
{
"epoch": 14.02,
"learning_rate": 7.42421052631579e-06,
"loss": 0.0024,
"step": 2950
},
{
"epoch": 14.02,
"learning_rate": 7.397894736842106e-06,
"loss": 0.002,
"step": 2975
},
{
"epoch": 14.02,
"learning_rate": 7.371578947368422e-06,
"loss": 0.0019,
"step": 3000
},
{
"epoch": 14.02,
"eval_loss": 0.242582768201828,
"eval_runtime": 1448.2754,
"eval_samples_per_second": 4.52,
"eval_steps_per_second": 0.142,
"eval_wer": 7.405491169977925,
"step": 3000
},
{
"epoch": 15.0,
"learning_rate": 7.345263157894738e-06,
"loss": 0.0016,
"step": 3025
},
{
"epoch": 15.01,
"learning_rate": 7.3189473684210535e-06,
"loss": 0.0028,
"step": 3050
},
{
"epoch": 15.01,
"learning_rate": 7.292631578947369e-06,
"loss": 0.0023,
"step": 3075
},
{
"epoch": 15.01,
"learning_rate": 7.266315789473684e-06,
"loss": 0.0018,
"step": 3100
},
{
"epoch": 15.01,
"learning_rate": 7.24e-06,
"loss": 0.002,
"step": 3125
},
{
"epoch": 15.02,
"learning_rate": 7.213684210526316e-06,
"loss": 0.0026,
"step": 3150
},
{
"epoch": 15.02,
"learning_rate": 7.187368421052632e-06,
"loss": 0.0013,
"step": 3175
},
{
"epoch": 15.02,
"learning_rate": 7.161052631578948e-06,
"loss": 0.0012,
"step": 3200
},
{
"epoch": 16.0,
"learning_rate": 7.134736842105264e-06,
"loss": 0.0015,
"step": 3225
},
{
"epoch": 16.0,
"learning_rate": 7.10842105263158e-06,
"loss": 0.0015,
"step": 3250
},
{
"epoch": 16.01,
"learning_rate": 7.082105263157896e-06,
"loss": 0.0008,
"step": 3275
},
{
"epoch": 16.01,
"learning_rate": 7.055789473684212e-06,
"loss": 0.0017,
"step": 3300
},
{
"epoch": 16.01,
"learning_rate": 7.0294736842105275e-06,
"loss": 0.002,
"step": 3325
},
{
"epoch": 16.02,
"learning_rate": 7.0031578947368425e-06,
"loss": 0.001,
"step": 3350
},
{
"epoch": 16.02,
"learning_rate": 6.976842105263158e-06,
"loss": 0.0011,
"step": 3375
},
{
"epoch": 16.02,
"learning_rate": 6.950526315789474e-06,
"loss": 0.0009,
"step": 3400
},
{
"epoch": 17.0,
"learning_rate": 6.92421052631579e-06,
"loss": 0.0012,
"step": 3425
},
{
"epoch": 17.0,
"learning_rate": 6.897894736842106e-06,
"loss": 0.0016,
"step": 3450
},
{
"epoch": 17.01,
"learning_rate": 6.871578947368422e-06,
"loss": 0.0016,
"step": 3475
},
{
"epoch": 17.01,
"learning_rate": 6.845263157894737e-06,
"loss": 0.0017,
"step": 3500
},
{
"epoch": 17.01,
"learning_rate": 6.818947368421052e-06,
"loss": 0.0015,
"step": 3525
},
{
"epoch": 17.02,
"learning_rate": 6.792631578947368e-06,
"loss": 0.001,
"step": 3550
},
{
"epoch": 17.02,
"learning_rate": 6.766315789473685e-06,
"loss": 0.0012,
"step": 3575
},
{
"epoch": 17.02,
"learning_rate": 6.740000000000001e-06,
"loss": 0.0016,
"step": 3600
},
{
"epoch": 18.0,
"learning_rate": 6.7136842105263165e-06,
"loss": 0.0019,
"step": 3625
},
{
"epoch": 18.0,
"learning_rate": 6.687368421052632e-06,
"loss": 0.0015,
"step": 3650
},
{
"epoch": 18.01,
"learning_rate": 6.661052631578948e-06,
"loss": 0.0019,
"step": 3675
},
{
"epoch": 18.01,
"learning_rate": 6.634736842105264e-06,
"loss": 0.0011,
"step": 3700
},
{
"epoch": 18.01,
"learning_rate": 6.60842105263158e-06,
"loss": 0.0007,
"step": 3725
},
{
"epoch": 18.02,
"learning_rate": 6.582105263157896e-06,
"loss": 0.001,
"step": 3750
},
{
"epoch": 18.02,
"learning_rate": 6.5557894736842106e-06,
"loss": 0.0013,
"step": 3775
},
{
"epoch": 18.02,
"learning_rate": 6.529473684210526e-06,
"loss": 0.0014,
"step": 3800
},
{
"epoch": 19.0,
"learning_rate": 6.503157894736842e-06,
"loss": 0.0026,
"step": 3825
},
{
"epoch": 19.0,
"learning_rate": 6.476842105263158e-06,
"loss": 0.001,
"step": 3850
},
{
"epoch": 19.01,
"learning_rate": 6.450526315789474e-06,
"loss": 0.0011,
"step": 3875
},
{
"epoch": 19.01,
"learning_rate": 6.42421052631579e-06,
"loss": 0.0011,
"step": 3900
},
{
"epoch": 19.01,
"learning_rate": 6.397894736842106e-06,
"loss": 0.0007,
"step": 3925
},
{
"epoch": 19.02,
"learning_rate": 6.371578947368422e-06,
"loss": 0.0006,
"step": 3950
},
{
"epoch": 19.02,
"learning_rate": 6.345263157894738e-06,
"loss": 0.0006,
"step": 3975
},
{
"epoch": 19.02,
"learning_rate": 6.318947368421054e-06,
"loss": 0.0011,
"step": 4000
},
{
"epoch": 19.02,
"eval_loss": 0.2689194083213806,
"eval_runtime": 1460.1887,
"eval_samples_per_second": 4.483,
"eval_steps_per_second": 0.14,
"eval_wer": 7.352028145695365,
"step": 4000
},
{
"epoch": 20.0,
"learning_rate": 6.292631578947369e-06,
"loss": 0.0007,
"step": 4025
},
{
"epoch": 20.0,
"learning_rate": 6.266315789473685e-06,
"loss": 0.0021,
"step": 4050
},
{
"epoch": 20.01,
"learning_rate": 6.24e-06,
"loss": 0.0009,
"step": 4075
},
{
"epoch": 20.01,
"learning_rate": 6.213684210526316e-06,
"loss": 0.0013,
"step": 4100
},
{
"epoch": 20.01,
"learning_rate": 6.187368421052632e-06,
"loss": 0.0012,
"step": 4125
},
{
"epoch": 20.02,
"learning_rate": 6.161052631578948e-06,
"loss": 0.0007,
"step": 4150
},
{
"epoch": 20.02,
"learning_rate": 6.134736842105264e-06,
"loss": 0.0012,
"step": 4175
},
{
"epoch": 20.02,
"learning_rate": 6.108421052631579e-06,
"loss": 0.0008,
"step": 4200
},
{
"epoch": 21.0,
"learning_rate": 6.0821052631578945e-06,
"loss": 0.0008,
"step": 4225
},
{
"epoch": 21.0,
"learning_rate": 6.055789473684212e-06,
"loss": 0.0015,
"step": 4250
},
{
"epoch": 21.01,
"learning_rate": 6.029473684210527e-06,
"loss": 0.0013,
"step": 4275
},
{
"epoch": 21.01,
"learning_rate": 6.003157894736843e-06,
"loss": 0.0011,
"step": 4300
},
{
"epoch": 21.01,
"learning_rate": 5.976842105263159e-06,
"loss": 0.0006,
"step": 4325
},
{
"epoch": 21.02,
"learning_rate": 5.950526315789474e-06,
"loss": 0.0015,
"step": 4350
},
{
"epoch": 21.02,
"learning_rate": 5.92421052631579e-06,
"loss": 0.0006,
"step": 4375
},
{
"epoch": 21.02,
"learning_rate": 5.897894736842106e-06,
"loss": 0.0009,
"step": 4400
},
{
"epoch": 22.0,
"learning_rate": 5.871578947368422e-06,
"loss": 0.0008,
"step": 4425
},
{
"epoch": 22.0,
"learning_rate": 5.845263157894737e-06,
"loss": 0.0012,
"step": 4450
},
{
"epoch": 22.01,
"learning_rate": 5.818947368421053e-06,
"loss": 0.0007,
"step": 4475
},
{
"epoch": 22.01,
"learning_rate": 5.7926315789473685e-06,
"loss": 0.0008,
"step": 4500
},
{
"epoch": 22.01,
"learning_rate": 5.766315789473684e-06,
"loss": 0.0008,
"step": 4525
},
{
"epoch": 22.02,
"learning_rate": 5.74e-06,
"loss": 0.0006,
"step": 4550
},
{
"epoch": 22.02,
"learning_rate": 5.713684210526316e-06,
"loss": 0.0007,
"step": 4575
},
{
"epoch": 22.02,
"learning_rate": 5.687368421052633e-06,
"loss": 0.001,
"step": 4600
},
{
"epoch": 23.0,
"learning_rate": 5.661052631578948e-06,
"loss": 0.0008,
"step": 4625
},
{
"epoch": 23.0,
"learning_rate": 5.634736842105264e-06,
"loss": 0.0011,
"step": 4650
},
{
"epoch": 23.01,
"learning_rate": 5.60842105263158e-06,
"loss": 0.0013,
"step": 4675
},
{
"epoch": 23.01,
"learning_rate": 5.582105263157895e-06,
"loss": 0.0014,
"step": 4700
},
{
"epoch": 23.01,
"learning_rate": 5.555789473684211e-06,
"loss": 0.0009,
"step": 4725
},
{
"epoch": 23.02,
"learning_rate": 5.529473684210527e-06,
"loss": 0.0012,
"step": 4750
},
{
"epoch": 23.02,
"learning_rate": 5.5031578947368425e-06,
"loss": 0.0009,
"step": 4775
},
{
"epoch": 23.02,
"learning_rate": 5.476842105263158e-06,
"loss": 0.0006,
"step": 4800
},
{
"epoch": 24.0,
"learning_rate": 5.450526315789474e-06,
"loss": 0.0008,
"step": 4825
},
{
"epoch": 24.0,
"learning_rate": 5.42421052631579e-06,
"loss": 0.0015,
"step": 4850
},
{
"epoch": 24.01,
"learning_rate": 5.397894736842105e-06,
"loss": 0.0008,
"step": 4875
},
{
"epoch": 24.01,
"learning_rate": 5.371578947368421e-06,
"loss": 0.0003,
"step": 4900
},
{
"epoch": 24.01,
"learning_rate": 5.345263157894738e-06,
"loss": 0.0005,
"step": 4925
},
{
"epoch": 24.02,
"learning_rate": 5.318947368421053e-06,
"loss": 0.0005,
"step": 4950
},
{
"epoch": 24.02,
"learning_rate": 5.292631578947369e-06,
"loss": 0.0015,
"step": 4975
},
{
"epoch": 24.02,
"learning_rate": 5.266315789473685e-06,
"loss": 0.0014,
"step": 5000
},
{
"epoch": 24.02,
"eval_loss": 0.2848970890045166,
"eval_runtime": 1475.3105,
"eval_samples_per_second": 4.437,
"eval_steps_per_second": 0.139,
"eval_wer": 7.531387969094923,
"step": 5000
},
{
"epoch": 25.0,
"learning_rate": 5.240000000000001e-06,
"loss": 0.0008,
"step": 5025
},
{
"epoch": 25.0,
"learning_rate": 5.2136842105263165e-06,
"loss": 0.0008,
"step": 5050
},
{
"epoch": 25.01,
"learning_rate": 5.187368421052632e-06,
"loss": 0.0019,
"step": 5075
},
{
"epoch": 25.01,
"learning_rate": 5.161052631578948e-06,
"loss": 0.0009,
"step": 5100
},
{
"epoch": 25.01,
"learning_rate": 5.134736842105263e-06,
"loss": 0.0015,
"step": 5125
},
{
"epoch": 25.02,
"learning_rate": 5.108421052631579e-06,
"loss": 0.0006,
"step": 5150
},
{
"epoch": 25.02,
"learning_rate": 5.082105263157895e-06,
"loss": 0.001,
"step": 5175
},
{
"epoch": 25.02,
"learning_rate": 5.0557894736842105e-06,
"loss": 0.0006,
"step": 5200
},
{
"epoch": 26.0,
"learning_rate": 5.029473684210526e-06,
"loss": 0.0007,
"step": 5225
},
{
"epoch": 26.0,
"learning_rate": 5.003157894736842e-06,
"loss": 0.0012,
"step": 5250
},
{
"epoch": 26.01,
"learning_rate": 4.976842105263158e-06,
"loss": 0.0013,
"step": 5275
},
{
"epoch": 26.01,
"learning_rate": 4.950526315789474e-06,
"loss": 0.0009,
"step": 5300
},
{
"epoch": 26.01,
"learning_rate": 4.92421052631579e-06,
"loss": 0.0005,
"step": 5325
},
{
"epoch": 26.02,
"learning_rate": 4.897894736842106e-06,
"loss": 0.0006,
"step": 5350
},
{
"epoch": 26.02,
"learning_rate": 4.871578947368421e-06,
"loss": 0.0007,
"step": 5375
},
{
"epoch": 26.02,
"learning_rate": 4.845263157894737e-06,
"loss": 0.0007,
"step": 5400
},
{
"epoch": 27.0,
"learning_rate": 4.818947368421053e-06,
"loss": 0.0005,
"step": 5425
},
{
"epoch": 27.0,
"learning_rate": 4.792631578947369e-06,
"loss": 0.0012,
"step": 5450
},
{
"epoch": 27.01,
"learning_rate": 4.7663157894736845e-06,
"loss": 0.0008,
"step": 5475
},
{
"epoch": 27.01,
"learning_rate": 4.74e-06,
"loss": 0.0007,
"step": 5500
},
{
"epoch": 27.01,
"learning_rate": 4.713684210526316e-06,
"loss": 0.0009,
"step": 5525
},
{
"epoch": 27.02,
"learning_rate": 4.687368421052632e-06,
"loss": 0.0009,
"step": 5550
},
{
"epoch": 27.02,
"learning_rate": 4.661052631578948e-06,
"loss": 0.0004,
"step": 5575
},
{
"epoch": 27.02,
"learning_rate": 4.634736842105264e-06,
"loss": 0.001,
"step": 5600
},
{
"epoch": 28.0,
"learning_rate": 4.6084210526315794e-06,
"loss": 0.0004,
"step": 5625
},
{
"epoch": 28.0,
"learning_rate": 4.582105263157895e-06,
"loss": 0.0004,
"step": 5650
},
{
"epoch": 28.01,
"learning_rate": 4.55578947368421e-06,
"loss": 0.0003,
"step": 5675
},
{
"epoch": 28.01,
"learning_rate": 4.529473684210527e-06,
"loss": 0.0004,
"step": 5700
},
{
"epoch": 28.01,
"learning_rate": 4.503157894736843e-06,
"loss": 0.0006,
"step": 5725
},
{
"epoch": 28.02,
"learning_rate": 4.4768421052631585e-06,
"loss": 0.0004,
"step": 5750
},
{
"epoch": 28.02,
"learning_rate": 4.450526315789474e-06,
"loss": 0.0003,
"step": 5775
},
{
"epoch": 28.02,
"learning_rate": 4.424210526315789e-06,
"loss": 0.0003,
"step": 5800
},
{
"epoch": 29.0,
"learning_rate": 4.397894736842105e-06,
"loss": 0.0002,
"step": 5825
},
{
"epoch": 29.0,
"learning_rate": 4.371578947368421e-06,
"loss": 0.0002,
"step": 5850
},
{
"epoch": 29.01,
"learning_rate": 4.345263157894738e-06,
"loss": 0.0002,
"step": 5875
},
{
"epoch": 29.01,
"learning_rate": 4.3189473684210535e-06,
"loss": 0.0002,
"step": 5900
},
{
"epoch": 29.01,
"learning_rate": 4.2926315789473684e-06,
"loss": 0.0002,
"step": 5925
},
{
"epoch": 29.02,
"learning_rate": 4.266315789473684e-06,
"loss": 0.0002,
"step": 5950
},
{
"epoch": 29.02,
"learning_rate": 4.24e-06,
"loss": 0.0003,
"step": 5975
},
{
"epoch": 29.02,
"learning_rate": 4.213684210526316e-06,
"loss": 0.0004,
"step": 6000
},
{
"epoch": 29.02,
"eval_loss": 0.2931840717792511,
"eval_runtime": 1500.3562,
"eval_samples_per_second": 4.363,
"eval_steps_per_second": 0.137,
"eval_wer": 7.258899006622517,
"step": 6000
},
{
"epoch": 30.0,
"learning_rate": 4.1873684210526325e-06,
"loss": 0.0002,
"step": 6025
},
{
"epoch": 30.0,
"learning_rate": 4.1610526315789475e-06,
"loss": 0.0002,
"step": 6050
},
{
"epoch": 30.01,
"learning_rate": 4.134736842105263e-06,
"loss": 0.0002,
"step": 6075
},
{
"epoch": 30.01,
"learning_rate": 4.108421052631579e-06,
"loss": 0.0002,
"step": 6100
},
{
"epoch": 30.01,
"learning_rate": 4.082105263157895e-06,
"loss": 0.0002,
"step": 6125
},
{
"epoch": 30.02,
"learning_rate": 4.055789473684211e-06,
"loss": 0.0002,
"step": 6150
},
{
"epoch": 30.02,
"learning_rate": 4.029473684210527e-06,
"loss": 0.0002,
"step": 6175
},
{
"epoch": 30.02,
"learning_rate": 4.0031578947368424e-06,
"loss": 0.0002,
"step": 6200
},
{
"epoch": 31.0,
"learning_rate": 3.976842105263158e-06,
"loss": 0.0002,
"step": 6225
},
{
"epoch": 31.0,
"learning_rate": 3.950526315789474e-06,
"loss": 0.0002,
"step": 6250
},
{
"epoch": 31.01,
"learning_rate": 3.92421052631579e-06,
"loss": 0.0002,
"step": 6275
},
{
"epoch": 31.01,
"learning_rate": 3.897894736842106e-06,
"loss": 0.0001,
"step": 6300
},
{
"epoch": 31.01,
"learning_rate": 3.8715789473684215e-06,
"loss": 0.0001,
"step": 6325
},
{
"epoch": 31.02,
"learning_rate": 3.845263157894737e-06,
"loss": 0.0001,
"step": 6350
},
{
"epoch": 31.02,
"learning_rate": 3.818947368421053e-06,
"loss": 0.0001,
"step": 6375
},
{
"epoch": 31.02,
"learning_rate": 3.792631578947369e-06,
"loss": 0.0001,
"step": 6400
},
{
"epoch": 32.0,
"learning_rate": 3.766315789473685e-06,
"loss": 0.0001,
"step": 6425
},
{
"epoch": 32.01,
"learning_rate": 3.74e-06,
"loss": 0.0003,
"step": 6450
},
{
"epoch": 32.01,
"learning_rate": 3.713684210526316e-06,
"loss": 0.0001,
"step": 6475
},
{
"epoch": 32.01,
"learning_rate": 3.687368421052632e-06,
"loss": 0.0001,
"step": 6500
},
{
"epoch": 32.01,
"learning_rate": 3.6610526315789472e-06,
"loss": 0.0001,
"step": 6525
},
{
"epoch": 32.02,
"learning_rate": 3.6347368421052635e-06,
"loss": 0.0001,
"step": 6550
},
{
"epoch": 32.02,
"learning_rate": 3.6084210526315793e-06,
"loss": 0.0001,
"step": 6575
},
{
"epoch": 32.02,
"learning_rate": 3.582105263157895e-06,
"loss": 0.0001,
"step": 6600
},
{
"epoch": 33.0,
"learning_rate": 3.555789473684211e-06,
"loss": 0.0001,
"step": 6625
},
{
"epoch": 33.01,
"learning_rate": 3.5294736842105263e-06,
"loss": 0.0002,
"step": 6650
},
{
"epoch": 33.01,
"learning_rate": 3.503157894736842e-06,
"loss": 0.0001,
"step": 6675
},
{
"epoch": 33.01,
"learning_rate": 3.476842105263158e-06,
"loss": 0.0001,
"step": 6700
},
{
"epoch": 33.01,
"learning_rate": 3.450526315789474e-06,
"loss": 0.0001,
"step": 6725
},
{
"epoch": 33.02,
"learning_rate": 3.42421052631579e-06,
"loss": 0.0001,
"step": 6750
},
{
"epoch": 33.02,
"learning_rate": 3.3978947368421054e-06,
"loss": 0.0001,
"step": 6775
},
{
"epoch": 33.02,
"learning_rate": 3.3715789473684212e-06,
"loss": 0.0001,
"step": 6800
},
{
"epoch": 34.0,
"learning_rate": 3.345263157894737e-06,
"loss": 0.0001,
"step": 6825
},
{
"epoch": 34.01,
"learning_rate": 3.318947368421053e-06,
"loss": 0.0001,
"step": 6850
},
{
"epoch": 34.01,
"learning_rate": 3.292631578947369e-06,
"loss": 0.0001,
"step": 6875
},
{
"epoch": 34.01,
"learning_rate": 3.2663157894736845e-06,
"loss": 0.0001,
"step": 6900
},
{
"epoch": 34.01,
"learning_rate": 3.2400000000000003e-06,
"loss": 0.0001,
"step": 6925
},
{
"epoch": 34.02,
"learning_rate": 3.213684210526316e-06,
"loss": 0.0001,
"step": 6950
},
{
"epoch": 34.02,
"learning_rate": 3.187368421052632e-06,
"loss": 0.0001,
"step": 6975
},
{
"epoch": 34.02,
"learning_rate": 3.1610526315789474e-06,
"loss": 0.0001,
"step": 7000
},
{
"epoch": 34.02,
"eval_loss": 0.30692604184150696,
"eval_runtime": 1437.4075,
"eval_samples_per_second": 4.554,
"eval_steps_per_second": 0.143,
"eval_wer": 7.148523730684326,
"step": 7000
},
{
"epoch": 35.0,
"learning_rate": 3.134736842105263e-06,
"loss": 0.0001,
"step": 7025
},
{
"epoch": 35.01,
"learning_rate": 3.1084210526315794e-06,
"loss": 0.0001,
"step": 7050
},
{
"epoch": 35.01,
"learning_rate": 3.0821052631578952e-06,
"loss": 0.0001,
"step": 7075
},
{
"epoch": 35.01,
"learning_rate": 3.055789473684211e-06,
"loss": 0.0001,
"step": 7100
},
{
"epoch": 35.01,
"learning_rate": 3.0294736842105264e-06,
"loss": 0.0001,
"step": 7125
},
{
"epoch": 35.02,
"learning_rate": 3.0031578947368423e-06,
"loss": 0.0001,
"step": 7150
},
{
"epoch": 35.02,
"learning_rate": 2.976842105263158e-06,
"loss": 0.0001,
"step": 7175
},
{
"epoch": 35.02,
"learning_rate": 2.9505263157894735e-06,
"loss": 0.0001,
"step": 7200
},
{
"epoch": 36.0,
"learning_rate": 2.92421052631579e-06,
"loss": 0.0001,
"step": 7225
},
{
"epoch": 36.01,
"learning_rate": 2.8978947368421055e-06,
"loss": 0.0001,
"step": 7250
},
{
"epoch": 36.01,
"learning_rate": 2.8715789473684214e-06,
"loss": 0.0001,
"step": 7275
},
{
"epoch": 36.01,
"learning_rate": 2.845263157894737e-06,
"loss": 0.0001,
"step": 7300
},
{
"epoch": 36.01,
"learning_rate": 2.8189473684210526e-06,
"loss": 0.0001,
"step": 7325
},
{
"epoch": 36.02,
"learning_rate": 2.7926315789473684e-06,
"loss": 0.0001,
"step": 7350
},
{
"epoch": 36.02,
"learning_rate": 2.766315789473684e-06,
"loss": 0.0001,
"step": 7375
},
{
"epoch": 36.02,
"learning_rate": 2.7400000000000004e-06,
"loss": 0.0001,
"step": 7400
},
{
"epoch": 37.0,
"learning_rate": 2.7136842105263163e-06,
"loss": 0.0001,
"step": 7425
},
{
"epoch": 37.01,
"learning_rate": 2.6873684210526317e-06,
"loss": 0.0001,
"step": 7450
},
{
"epoch": 37.01,
"learning_rate": 2.6610526315789475e-06,
"loss": 0.0001,
"step": 7475
},
{
"epoch": 37.01,
"learning_rate": 2.6347368421052633e-06,
"loss": 0.0001,
"step": 7500
},
{
"epoch": 37.01,
"learning_rate": 2.608421052631579e-06,
"loss": 0.0001,
"step": 7525
},
{
"epoch": 37.02,
"learning_rate": 2.5821052631578954e-06,
"loss": 0.0001,
"step": 7550
},
{
"epoch": 37.02,
"learning_rate": 2.5557894736842108e-06,
"loss": 0.0001,
"step": 7575
},
{
"epoch": 37.02,
"learning_rate": 2.5294736842105266e-06,
"loss": 0.0001,
"step": 7600
},
{
"epoch": 38.0,
"learning_rate": 2.5031578947368424e-06,
"loss": 0.0001,
"step": 7625
},
{
"epoch": 38.01,
"learning_rate": 2.476842105263158e-06,
"loss": 0.0001,
"step": 7650
},
{
"epoch": 38.01,
"learning_rate": 2.4505263157894736e-06,
"loss": 0.0001,
"step": 7675
},
{
"epoch": 38.01,
"learning_rate": 2.42421052631579e-06,
"loss": 0.0001,
"step": 7700
},
{
"epoch": 38.01,
"learning_rate": 2.3978947368421052e-06,
"loss": 0.0001,
"step": 7725
},
{
"epoch": 38.02,
"learning_rate": 2.371578947368421e-06,
"loss": 0.0001,
"step": 7750
},
{
"epoch": 38.02,
"learning_rate": 2.3452631578947373e-06,
"loss": 0.0001,
"step": 7775
},
{
"epoch": 38.02,
"learning_rate": 2.3189473684210527e-06,
"loss": 0.0001,
"step": 7800
},
{
"epoch": 39.0,
"learning_rate": 2.2926315789473685e-06,
"loss": 0.0001,
"step": 7825
},
{
"epoch": 39.01,
"learning_rate": 2.2663157894736843e-06,
"loss": 0.0001,
"step": 7850
},
{
"epoch": 39.01,
"learning_rate": 2.24e-06,
"loss": 0.0001,
"step": 7875
},
{
"epoch": 39.01,
"learning_rate": 2.213684210526316e-06,
"loss": 0.0001,
"step": 7900
},
{
"epoch": 39.01,
"learning_rate": 2.187368421052632e-06,
"loss": 0.0001,
"step": 7925
},
{
"epoch": 39.02,
"learning_rate": 2.1610526315789476e-06,
"loss": 0.0001,
"step": 7950
},
{
"epoch": 39.02,
"learning_rate": 2.1347368421052634e-06,
"loss": 0.0001,
"step": 7975
},
{
"epoch": 39.02,
"learning_rate": 2.1084210526315792e-06,
"loss": 0.0001,
"step": 8000
},
{
"epoch": 39.02,
"eval_loss": 0.314346045255661,
"eval_runtime": 1406.8839,
"eval_samples_per_second": 4.653,
"eval_steps_per_second": 0.146,
"eval_wer": 7.148523730684326,
"step": 8000
},
{
"epoch": 40.0,
"learning_rate": 2.082105263157895e-06,
"loss": 0.0001,
"step": 8025
},
{
"epoch": 40.01,
"learning_rate": 2.055789473684211e-06,
"loss": 0.0001,
"step": 8050
},
{
"epoch": 40.01,
"learning_rate": 2.0294736842105263e-06,
"loss": 0.0001,
"step": 8075
},
{
"epoch": 40.01,
"learning_rate": 2.003157894736842e-06,
"loss": 0.0001,
"step": 8100
},
{
"epoch": 40.01,
"learning_rate": 1.976842105263158e-06,
"loss": 0.0001,
"step": 8125
},
{
"epoch": 40.02,
"learning_rate": 1.9505263157894737e-06,
"loss": 0.0001,
"step": 8150
},
{
"epoch": 40.02,
"learning_rate": 1.9242105263157896e-06,
"loss": 0.0001,
"step": 8175
},
{
"epoch": 40.02,
"learning_rate": 1.8978947368421056e-06,
"loss": 0.0001,
"step": 8200
},
{
"epoch": 41.0,
"learning_rate": 1.8715789473684212e-06,
"loss": 0.0001,
"step": 8225
},
{
"epoch": 41.01,
"learning_rate": 1.845263157894737e-06,
"loss": 0.0001,
"step": 8250
},
{
"epoch": 41.01,
"learning_rate": 1.8189473684210528e-06,
"loss": 0.0001,
"step": 8275
},
{
"epoch": 41.01,
"learning_rate": 1.7926315789473686e-06,
"loss": 0.0001,
"step": 8300
},
{
"epoch": 41.01,
"learning_rate": 1.7663157894736843e-06,
"loss": 0.0001,
"step": 8325
},
{
"epoch": 41.02,
"learning_rate": 1.74e-06,
"loss": 0.0001,
"step": 8350
},
{
"epoch": 41.02,
"learning_rate": 1.713684210526316e-06,
"loss": 0.0001,
"step": 8375
},
{
"epoch": 41.02,
"learning_rate": 1.6873684210526317e-06,
"loss": 0.0001,
"step": 8400
},
{
"epoch": 42.0,
"learning_rate": 1.6610526315789473e-06,
"loss": 0.0001,
"step": 8425
},
{
"epoch": 42.01,
"learning_rate": 1.6347368421052633e-06,
"loss": 0.0001,
"step": 8450
},
{
"epoch": 42.01,
"learning_rate": 1.6084210526315792e-06,
"loss": 0.0001,
"step": 8475
},
{
"epoch": 42.01,
"learning_rate": 1.5821052631578948e-06,
"loss": 0.0001,
"step": 8500
},
{
"epoch": 42.01,
"learning_rate": 1.5557894736842106e-06,
"loss": 0.0001,
"step": 8525
},
{
"epoch": 42.02,
"learning_rate": 1.5294736842105264e-06,
"loss": 0.0001,
"step": 8550
},
{
"epoch": 42.02,
"learning_rate": 1.5031578947368422e-06,
"loss": 0.0001,
"step": 8575
},
{
"epoch": 42.02,
"learning_rate": 1.4768421052631578e-06,
"loss": 0.0001,
"step": 8600
},
{
"epoch": 43.0,
"learning_rate": 1.4505263157894739e-06,
"loss": 0.0001,
"step": 8625
},
{
"epoch": 43.01,
"learning_rate": 1.4242105263157897e-06,
"loss": 0.0001,
"step": 8650
},
{
"epoch": 43.01,
"learning_rate": 1.3978947368421053e-06,
"loss": 0.0001,
"step": 8675
},
{
"epoch": 43.01,
"learning_rate": 1.3715789473684213e-06,
"loss": 0.0001,
"step": 8700
},
{
"epoch": 43.01,
"learning_rate": 1.345263157894737e-06,
"loss": 0.0001,
"step": 8725
},
{
"epoch": 43.02,
"learning_rate": 1.3189473684210527e-06,
"loss": 0.0001,
"step": 8750
},
{
"epoch": 43.02,
"learning_rate": 1.2926315789473683e-06,
"loss": 0.0001,
"step": 8775
},
{
"epoch": 43.02,
"learning_rate": 1.2663157894736844e-06,
"loss": 0.0001,
"step": 8800
},
{
"epoch": 44.0,
"learning_rate": 1.2400000000000002e-06,
"loss": 0.0001,
"step": 8825
},
{
"epoch": 44.01,
"learning_rate": 1.213684210526316e-06,
"loss": 0.0001,
"step": 8850
},
{
"epoch": 44.01,
"learning_rate": 1.1873684210526316e-06,
"loss": 0.0001,
"step": 8875
},
{
"epoch": 44.01,
"learning_rate": 1.1610526315789474e-06,
"loss": 0.0001,
"step": 8900
},
{
"epoch": 44.01,
"learning_rate": 1.1347368421052633e-06,
"loss": 0.0001,
"step": 8925
},
{
"epoch": 44.02,
"learning_rate": 1.108421052631579e-06,
"loss": 0.0001,
"step": 8950
},
{
"epoch": 44.02,
"learning_rate": 1.082105263157895e-06,
"loss": 0.0001,
"step": 8975
},
{
"epoch": 44.02,
"learning_rate": 1.0557894736842105e-06,
"loss": 0.0001,
"step": 9000
},
{
"epoch": 44.02,
"eval_loss": 0.3196016252040863,
"eval_runtime": 1404.9195,
"eval_samples_per_second": 4.659,
"eval_steps_per_second": 0.146,
"eval_wer": 7.122654525386314,
"step": 9000
},
{
"epoch": 45.0,
"learning_rate": 1.0294736842105265e-06,
"loss": 0.0001,
"step": 9025
},
{
"epoch": 45.01,
"learning_rate": 1.0031578947368421e-06,
"loss": 0.0001,
"step": 9050
},
{
"epoch": 45.01,
"learning_rate": 9.76842105263158e-07,
"loss": 0.0001,
"step": 9075
},
{
"epoch": 45.01,
"learning_rate": 9.505263157894738e-07,
"loss": 0.0001,
"step": 9100
},
{
"epoch": 45.01,
"learning_rate": 9.242105263157895e-07,
"loss": 0.0001,
"step": 9125
},
{
"epoch": 45.02,
"learning_rate": 8.978947368421054e-07,
"loss": 0.0001,
"step": 9150
},
{
"epoch": 45.02,
"learning_rate": 8.715789473684211e-07,
"loss": 0.0001,
"step": 9175
},
{
"epoch": 45.02,
"learning_rate": 8.452631578947369e-07,
"loss": 0.0001,
"step": 9200
},
{
"epoch": 46.0,
"learning_rate": 8.189473684210527e-07,
"loss": 0.0001,
"step": 9225
},
{
"epoch": 46.01,
"learning_rate": 7.926315789473685e-07,
"loss": 0.0001,
"step": 9250
},
{
"epoch": 46.01,
"learning_rate": 7.663157894736843e-07,
"loss": 0.0001,
"step": 9275
},
{
"epoch": 46.01,
"learning_rate": 7.4e-07,
"loss": 0.0001,
"step": 9300
},
{
"epoch": 46.01,
"learning_rate": 7.136842105263158e-07,
"loss": 0.0001,
"step": 9325
},
{
"epoch": 46.02,
"learning_rate": 6.873684210526315e-07,
"loss": 0.0001,
"step": 9350
},
{
"epoch": 46.02,
"learning_rate": 6.610526315789475e-07,
"loss": 0.0001,
"step": 9375
},
{
"epoch": 46.02,
"learning_rate": 6.347368421052633e-07,
"loss": 0.0001,
"step": 9400
},
{
"epoch": 47.0,
"learning_rate": 6.08421052631579e-07,
"loss": 0.0001,
"step": 9425
},
{
"epoch": 47.01,
"learning_rate": 5.821052631578948e-07,
"loss": 0.0001,
"step": 9450
},
{
"epoch": 47.01,
"learning_rate": 5.557894736842106e-07,
"loss": 0.0001,
"step": 9475
},
{
"epoch": 47.01,
"learning_rate": 5.294736842105263e-07,
"loss": 0.0001,
"step": 9500
},
{
"epoch": 47.01,
"learning_rate": 5.031578947368422e-07,
"loss": 0.0001,
"step": 9525
},
{
"epoch": 47.02,
"learning_rate": 4.768421052631579e-07,
"loss": 0.0001,
"step": 9550
},
{
"epoch": 47.02,
"learning_rate": 4.505263157894737e-07,
"loss": 0.0001,
"step": 9575
},
{
"epoch": 47.02,
"learning_rate": 4.242105263157895e-07,
"loss": 0.0001,
"step": 9600
},
{
"epoch": 48.0,
"learning_rate": 3.9789473684210533e-07,
"loss": 0.0001,
"step": 9625
},
{
"epoch": 48.01,
"learning_rate": 3.715789473684211e-07,
"loss": 0.0001,
"step": 9650
},
{
"epoch": 48.01,
"learning_rate": 3.4526315789473686e-07,
"loss": 0.0001,
"step": 9675
},
{
"epoch": 48.01,
"learning_rate": 3.1894736842105263e-07,
"loss": 0.0001,
"step": 9700
},
{
"epoch": 48.01,
"learning_rate": 2.9263157894736845e-07,
"loss": 0.0001,
"step": 9725
},
{
"epoch": 48.02,
"learning_rate": 2.663157894736842e-07,
"loss": 0.0001,
"step": 9750
},
{
"epoch": 48.02,
"learning_rate": 2.4000000000000003e-07,
"loss": 0.0001,
"step": 9775
},
{
"epoch": 48.02,
"learning_rate": 2.136842105263158e-07,
"loss": 0.0001,
"step": 9800
},
{
"epoch": 49.0,
"learning_rate": 1.8736842105263162e-07,
"loss": 0.0001,
"step": 9825
},
{
"epoch": 49.01,
"learning_rate": 1.6105263157894738e-07,
"loss": 0.0001,
"step": 9850
},
{
"epoch": 49.01,
"learning_rate": 1.3473684210526317e-07,
"loss": 0.0001,
"step": 9875
},
{
"epoch": 49.01,
"learning_rate": 1.0842105263157895e-07,
"loss": 0.0001,
"step": 9900
},
{
"epoch": 49.01,
"learning_rate": 8.210526315789474e-08,
"loss": 0.0001,
"step": 9925
},
{
"epoch": 49.02,
"learning_rate": 5.578947368421053e-08,
"loss": 0.0001,
"step": 9950
},
{
"epoch": 49.02,
"learning_rate": 2.9473684210526318e-08,
"loss": 0.0001,
"step": 9975
},
{
"epoch": 49.02,
"learning_rate": 3.1578947368421054e-09,
"loss": 0.0001,
"step": 10000
},
{
"epoch": 49.02,
"eval_loss": 0.3218191862106323,
"eval_runtime": 1405.9186,
"eval_samples_per_second": 4.656,
"eval_steps_per_second": 0.146,
"eval_wer": 7.124379139072848,
"step": 10000
},
{
"epoch": 49.02,
"step": 10000,
"total_flos": 6.52880959340544e+20,
"train_loss": 0.017749687022855506,
"train_runtime": 135026.7905,
"train_samples_per_second": 4.74,
"train_steps_per_second": 0.074
}
],
"logging_steps": 25,
"max_steps": 10000,
"num_train_epochs": 9223372036854775807,
"save_steps": 1000,
"total_flos": 6.52880959340544e+20,
"trial_name": null,
"trial_params": null
}