whisper-large-v2-english-2k-steps / trainer_state.json
Elsayed Issa
first commit
ada48e6
raw
history blame
10.4 kB
{
"best_metric": 0.1143532423424451,
"best_model_checkpoint": "/storage/elsayedissa/whisper-large-v2-english/checkpoint-2000",
"epoch": 0.10540184453227931,
"global_step": 2000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 4.6000000000000004e-07,
"loss": 1.195,
"step": 25
},
{
"epoch": 0.0,
"learning_rate": 9.600000000000001e-07,
"loss": 0.9807,
"step": 50
},
{
"epoch": 0.0,
"learning_rate": 1.46e-06,
"loss": 0.4964,
"step": 75
},
{
"epoch": 0.01,
"learning_rate": 1.94e-06,
"loss": 0.322,
"step": 100
},
{
"epoch": 0.01,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.3187,
"step": 125
},
{
"epoch": 0.01,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.1916,
"step": 150
},
{
"epoch": 0.01,
"learning_rate": 3.44e-06,
"loss": 0.2396,
"step": 175
},
{
"epoch": 0.01,
"learning_rate": 3.94e-06,
"loss": 0.1802,
"step": 200
},
{
"epoch": 0.01,
"learning_rate": 4.440000000000001e-06,
"loss": 0.2344,
"step": 225
},
{
"epoch": 0.01,
"learning_rate": 4.94e-06,
"loss": 0.1607,
"step": 250
},
{
"epoch": 0.01,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.2193,
"step": 275
},
{
"epoch": 0.02,
"learning_rate": 5.94e-06,
"loss": 0.1683,
"step": 300
},
{
"epoch": 0.02,
"learning_rate": 6.42e-06,
"loss": 0.2214,
"step": 325
},
{
"epoch": 0.02,
"learning_rate": 6.92e-06,
"loss": 0.1616,
"step": 350
},
{
"epoch": 0.02,
"learning_rate": 7.420000000000001e-06,
"loss": 0.2247,
"step": 375
},
{
"epoch": 0.02,
"learning_rate": 7.92e-06,
"loss": 0.1811,
"step": 400
},
{
"epoch": 0.02,
"learning_rate": 8.42e-06,
"loss": 0.2259,
"step": 425
},
{
"epoch": 0.02,
"learning_rate": 8.920000000000001e-06,
"loss": 0.1733,
"step": 450
},
{
"epoch": 0.03,
"learning_rate": 9.42e-06,
"loss": 0.2147,
"step": 475
},
{
"epoch": 0.03,
"learning_rate": 9.920000000000002e-06,
"loss": 0.1708,
"step": 500
},
{
"epoch": 0.03,
"learning_rate": 9.953333333333333e-06,
"loss": 0.22,
"step": 525
},
{
"epoch": 0.03,
"learning_rate": 9.89777777777778e-06,
"loss": 0.1731,
"step": 550
},
{
"epoch": 0.03,
"learning_rate": 9.842222222222223e-06,
"loss": 0.2183,
"step": 575
},
{
"epoch": 0.03,
"learning_rate": 9.786666666666667e-06,
"loss": 0.1726,
"step": 600
},
{
"epoch": 0.03,
"learning_rate": 9.731111111111113e-06,
"loss": 0.2217,
"step": 625
},
{
"epoch": 0.03,
"learning_rate": 9.675555555555555e-06,
"loss": 0.1643,
"step": 650
},
{
"epoch": 0.04,
"learning_rate": 9.620000000000001e-06,
"loss": 0.2271,
"step": 675
},
{
"epoch": 0.04,
"learning_rate": 9.564444444444445e-06,
"loss": 0.1602,
"step": 700
},
{
"epoch": 0.04,
"learning_rate": 9.508888888888889e-06,
"loss": 0.2287,
"step": 725
},
{
"epoch": 0.04,
"learning_rate": 9.453333333333335e-06,
"loss": 0.1701,
"step": 750
},
{
"epoch": 0.04,
"learning_rate": 9.397777777777779e-06,
"loss": 0.2158,
"step": 775
},
{
"epoch": 0.04,
"learning_rate": 9.342222222222223e-06,
"loss": 0.1703,
"step": 800
},
{
"epoch": 0.04,
"learning_rate": 9.286666666666667e-06,
"loss": 0.211,
"step": 825
},
{
"epoch": 0.04,
"learning_rate": 9.231111111111111e-06,
"loss": 0.1581,
"step": 850
},
{
"epoch": 0.05,
"learning_rate": 9.175555555555557e-06,
"loss": 0.215,
"step": 875
},
{
"epoch": 0.05,
"learning_rate": 9.12e-06,
"loss": 0.1645,
"step": 900
},
{
"epoch": 0.05,
"learning_rate": 9.066666666666667e-06,
"loss": 0.2283,
"step": 925
},
{
"epoch": 0.05,
"learning_rate": 9.011111111111111e-06,
"loss": 0.1499,
"step": 950
},
{
"epoch": 0.05,
"learning_rate": 8.955555555555555e-06,
"loss": 0.2204,
"step": 975
},
{
"epoch": 0.05,
"learning_rate": 8.900000000000001e-06,
"loss": 0.1763,
"step": 1000
},
{
"epoch": 0.05,
"eval_loss": 0.23617663979530334,
"eval_runtime": 5930.7453,
"eval_samples_per_second": 2.757,
"eval_steps_per_second": 0.172,
"eval_wer": 0.12045520789770495,
"step": 1000
},
{
"epoch": 0.05,
"learning_rate": 8.844444444444445e-06,
"loss": 0.2088,
"step": 1025
},
{
"epoch": 0.06,
"learning_rate": 8.788888888888891e-06,
"loss": 0.1511,
"step": 1050
},
{
"epoch": 0.06,
"learning_rate": 8.733333333333333e-06,
"loss": 0.2093,
"step": 1075
},
{
"epoch": 0.06,
"learning_rate": 8.677777777777779e-06,
"loss": 0.1617,
"step": 1100
},
{
"epoch": 0.06,
"learning_rate": 8.622222222222223e-06,
"loss": 0.2099,
"step": 1125
},
{
"epoch": 0.06,
"learning_rate": 8.566666666666667e-06,
"loss": 0.1734,
"step": 1150
},
{
"epoch": 0.06,
"learning_rate": 8.511111111111113e-06,
"loss": 0.2148,
"step": 1175
},
{
"epoch": 0.06,
"learning_rate": 8.455555555555555e-06,
"loss": 0.1564,
"step": 1200
},
{
"epoch": 0.06,
"learning_rate": 8.400000000000001e-06,
"loss": 0.205,
"step": 1225
},
{
"epoch": 0.07,
"learning_rate": 8.344444444444445e-06,
"loss": 0.1704,
"step": 1250
},
{
"epoch": 0.07,
"learning_rate": 8.288888888888889e-06,
"loss": 0.1994,
"step": 1275
},
{
"epoch": 0.07,
"learning_rate": 8.233333333333335e-06,
"loss": 0.1779,
"step": 1300
},
{
"epoch": 0.07,
"learning_rate": 8.177777777777779e-06,
"loss": 0.2,
"step": 1325
},
{
"epoch": 0.07,
"learning_rate": 8.122222222222223e-06,
"loss": 0.1587,
"step": 1350
},
{
"epoch": 0.07,
"learning_rate": 8.066666666666667e-06,
"loss": 0.204,
"step": 1375
},
{
"epoch": 0.07,
"learning_rate": 8.013333333333333e-06,
"loss": 0.1599,
"step": 1400
},
{
"epoch": 0.08,
"learning_rate": 7.957777777777779e-06,
"loss": 0.2189,
"step": 1425
},
{
"epoch": 0.08,
"learning_rate": 7.902222222222223e-06,
"loss": 0.1586,
"step": 1450
},
{
"epoch": 0.08,
"learning_rate": 7.846666666666667e-06,
"loss": 0.1885,
"step": 1475
},
{
"epoch": 0.08,
"learning_rate": 7.791111111111111e-06,
"loss": 0.1591,
"step": 1500
},
{
"epoch": 0.08,
"learning_rate": 7.735555555555557e-06,
"loss": 0.2103,
"step": 1525
},
{
"epoch": 0.08,
"learning_rate": 7.680000000000001e-06,
"loss": 0.145,
"step": 1550
},
{
"epoch": 0.08,
"learning_rate": 7.624444444444445e-06,
"loss": 0.1984,
"step": 1575
},
{
"epoch": 0.08,
"learning_rate": 7.56888888888889e-06,
"loss": 0.1451,
"step": 1600
},
{
"epoch": 0.09,
"learning_rate": 7.513333333333334e-06,
"loss": 0.1965,
"step": 1625
},
{
"epoch": 0.09,
"learning_rate": 7.457777777777778e-06,
"loss": 0.1703,
"step": 1650
},
{
"epoch": 0.09,
"learning_rate": 7.402222222222223e-06,
"loss": 0.1989,
"step": 1675
},
{
"epoch": 0.09,
"learning_rate": 7.346666666666668e-06,
"loss": 0.1509,
"step": 1700
},
{
"epoch": 0.09,
"learning_rate": 7.291111111111112e-06,
"loss": 0.1976,
"step": 1725
},
{
"epoch": 0.09,
"learning_rate": 7.235555555555556e-06,
"loss": 0.1523,
"step": 1750
},
{
"epoch": 0.09,
"learning_rate": 7.180000000000001e-06,
"loss": 0.1923,
"step": 1775
},
{
"epoch": 0.09,
"learning_rate": 7.124444444444445e-06,
"loss": 0.143,
"step": 1800
},
{
"epoch": 0.1,
"learning_rate": 7.06888888888889e-06,
"loss": 0.1968,
"step": 1825
},
{
"epoch": 0.1,
"learning_rate": 7.0133333333333345e-06,
"loss": 0.1518,
"step": 1850
},
{
"epoch": 0.1,
"learning_rate": 6.9577777777777785e-06,
"loss": 0.192,
"step": 1875
},
{
"epoch": 0.1,
"learning_rate": 6.902222222222223e-06,
"loss": 0.1547,
"step": 1900
},
{
"epoch": 0.1,
"learning_rate": 6.846666666666667e-06,
"loss": 0.1955,
"step": 1925
},
{
"epoch": 0.1,
"learning_rate": 6.7911111111111115e-06,
"loss": 0.1493,
"step": 1950
},
{
"epoch": 0.1,
"learning_rate": 6.735555555555556e-06,
"loss": 0.2019,
"step": 1975
},
{
"epoch": 0.11,
"learning_rate": 6.680000000000001e-06,
"loss": 0.1604,
"step": 2000
},
{
"epoch": 0.11,
"eval_loss": 0.22141852974891663,
"eval_runtime": 9591.9258,
"eval_samples_per_second": 1.705,
"eval_steps_per_second": 0.107,
"eval_wer": 0.1143532423424451,
"step": 2000
}
],
"max_steps": 5000,
"num_train_epochs": 1,
"total_flos": 2.1231710208e+20,
"trial_name": null,
"trial_params": null
}