whisper-ckpt3000 / trainer_state.json
lekiet258's picture
Upload 6 files
18669cb
raw
history blame
No virus
15.3 kB
{
"best_metric": 41.8351610692255,
"best_model_checkpoint": "./whisper-small-vn/checkpoint-3000",
"epoch": 60.0,
"global_step": 3000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.5,
"learning_rate": 4.4e-07,
"loss": 3.1082,
"step": 25
},
{
"epoch": 1.0,
"learning_rate": 9.400000000000001e-07,
"loss": 2.4714,
"step": 50
},
{
"epoch": 1.5,
"learning_rate": 1.44e-06,
"loss": 1.9624,
"step": 75
},
{
"epoch": 2.0,
"learning_rate": 1.94e-06,
"loss": 1.6754,
"step": 100
},
{
"epoch": 2.5,
"learning_rate": 2.4400000000000004e-06,
"loss": 1.3215,
"step": 125
},
{
"epoch": 3.0,
"learning_rate": 2.9400000000000002e-06,
"loss": 1.1727,
"step": 150
},
{
"epoch": 3.5,
"learning_rate": 3.44e-06,
"loss": 1.0125,
"step": 175
},
{
"epoch": 4.0,
"learning_rate": 3.94e-06,
"loss": 1.0082,
"step": 200
},
{
"epoch": 4.5,
"learning_rate": 4.440000000000001e-06,
"loss": 0.8672,
"step": 225
},
{
"epoch": 5.0,
"learning_rate": 4.94e-06,
"loss": 0.78,
"step": 250
},
{
"epoch": 5.5,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.6437,
"step": 275
},
{
"epoch": 6.0,
"learning_rate": 5.94e-06,
"loss": 0.6069,
"step": 300
},
{
"epoch": 6.5,
"learning_rate": 6.440000000000001e-06,
"loss": 0.4016,
"step": 325
},
{
"epoch": 7.0,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.4129,
"step": 350
},
{
"epoch": 7.5,
"learning_rate": 7.440000000000001e-06,
"loss": 0.2727,
"step": 375
},
{
"epoch": 8.0,
"learning_rate": 7.94e-06,
"loss": 0.2735,
"step": 400
},
{
"epoch": 8.5,
"learning_rate": 8.44e-06,
"loss": 0.1796,
"step": 425
},
{
"epoch": 9.0,
"learning_rate": 8.94e-06,
"loss": 0.1699,
"step": 450
},
{
"epoch": 9.5,
"learning_rate": 9.440000000000001e-06,
"loss": 0.1024,
"step": 475
},
{
"epoch": 10.0,
"learning_rate": 9.940000000000001e-06,
"loss": 0.1121,
"step": 500
},
{
"epoch": 10.5,
"learning_rate": 9.937142857142858e-06,
"loss": 0.0645,
"step": 525
},
{
"epoch": 11.0,
"learning_rate": 9.865714285714285e-06,
"loss": 0.0727,
"step": 550
},
{
"epoch": 11.5,
"learning_rate": 9.794285714285714e-06,
"loss": 0.043,
"step": 575
},
{
"epoch": 12.0,
"learning_rate": 9.722857142857143e-06,
"loss": 0.0461,
"step": 600
},
{
"epoch": 12.5,
"learning_rate": 9.651428571428572e-06,
"loss": 0.0289,
"step": 625
},
{
"epoch": 13.0,
"learning_rate": 9.58e-06,
"loss": 0.0329,
"step": 650
},
{
"epoch": 13.5,
"learning_rate": 9.508571428571429e-06,
"loss": 0.0223,
"step": 675
},
{
"epoch": 14.0,
"learning_rate": 9.437142857142858e-06,
"loss": 0.0214,
"step": 700
},
{
"epoch": 14.5,
"learning_rate": 9.365714285714287e-06,
"loss": 0.0153,
"step": 725
},
{
"epoch": 15.0,
"learning_rate": 9.294285714285714e-06,
"loss": 0.018,
"step": 750
},
{
"epoch": 15.5,
"learning_rate": 9.222857142857143e-06,
"loss": 0.0143,
"step": 775
},
{
"epoch": 16.0,
"learning_rate": 9.151428571428572e-06,
"loss": 0.0144,
"step": 800
},
{
"epoch": 16.5,
"learning_rate": 9.080000000000001e-06,
"loss": 0.0124,
"step": 825
},
{
"epoch": 17.0,
"learning_rate": 9.00857142857143e-06,
"loss": 0.0122,
"step": 850
},
{
"epoch": 17.5,
"learning_rate": 8.937142857142857e-06,
"loss": 0.0095,
"step": 875
},
{
"epoch": 18.0,
"learning_rate": 8.865714285714287e-06,
"loss": 0.0085,
"step": 900
},
{
"epoch": 18.5,
"learning_rate": 8.794285714285716e-06,
"loss": 0.0072,
"step": 925
},
{
"epoch": 19.0,
"learning_rate": 8.722857142857145e-06,
"loss": 0.0075,
"step": 950
},
{
"epoch": 19.5,
"learning_rate": 8.651428571428572e-06,
"loss": 0.0066,
"step": 975
},
{
"epoch": 20.0,
"learning_rate": 8.580000000000001e-06,
"loss": 0.0069,
"step": 1000
},
{
"epoch": 20.0,
"eval_loss": 1.1462156772613525,
"eval_runtime": 178.292,
"eval_samples_per_second": 1.481,
"eval_steps_per_second": 0.185,
"eval_wer": 43.137422892392046,
"step": 1000
},
{
"epoch": 20.5,
"learning_rate": 8.50857142857143e-06,
"loss": 0.0061,
"step": 1025
},
{
"epoch": 21.0,
"learning_rate": 8.437142857142859e-06,
"loss": 0.0083,
"step": 1050
},
{
"epoch": 21.5,
"learning_rate": 8.365714285714286e-06,
"loss": 0.0054,
"step": 1075
},
{
"epoch": 22.0,
"learning_rate": 8.294285714285715e-06,
"loss": 0.0063,
"step": 1100
},
{
"epoch": 22.5,
"learning_rate": 8.222857142857144e-06,
"loss": 0.0055,
"step": 1125
},
{
"epoch": 23.0,
"learning_rate": 8.151428571428572e-06,
"loss": 0.0061,
"step": 1150
},
{
"epoch": 23.5,
"learning_rate": 8.08e-06,
"loss": 0.0037,
"step": 1175
},
{
"epoch": 24.0,
"learning_rate": 8.00857142857143e-06,
"loss": 0.0039,
"step": 1200
},
{
"epoch": 24.5,
"learning_rate": 7.937142857142857e-06,
"loss": 0.0032,
"step": 1225
},
{
"epoch": 25.0,
"learning_rate": 7.865714285714286e-06,
"loss": 0.0032,
"step": 1250
},
{
"epoch": 25.5,
"learning_rate": 7.794285714285715e-06,
"loss": 0.0031,
"step": 1275
},
{
"epoch": 26.0,
"learning_rate": 7.722857142857142e-06,
"loss": 0.0027,
"step": 1300
},
{
"epoch": 26.5,
"learning_rate": 7.651428571428571e-06,
"loss": 0.0027,
"step": 1325
},
{
"epoch": 27.0,
"learning_rate": 7.58e-06,
"loss": 0.0027,
"step": 1350
},
{
"epoch": 27.5,
"learning_rate": 7.508571428571429e-06,
"loss": 0.0018,
"step": 1375
},
{
"epoch": 28.0,
"learning_rate": 7.4371428571428575e-06,
"loss": 0.0023,
"step": 1400
},
{
"epoch": 28.5,
"learning_rate": 7.365714285714286e-06,
"loss": 0.0024,
"step": 1425
},
{
"epoch": 29.0,
"learning_rate": 7.294285714285715e-06,
"loss": 0.0034,
"step": 1450
},
{
"epoch": 29.5,
"learning_rate": 7.222857142857144e-06,
"loss": 0.0025,
"step": 1475
},
{
"epoch": 30.0,
"learning_rate": 7.151428571428573e-06,
"loss": 0.0026,
"step": 1500
},
{
"epoch": 30.5,
"learning_rate": 7.08e-06,
"loss": 0.002,
"step": 1525
},
{
"epoch": 31.0,
"learning_rate": 7.008571428571429e-06,
"loss": 0.002,
"step": 1550
},
{
"epoch": 31.5,
"learning_rate": 6.937142857142858e-06,
"loss": 0.0012,
"step": 1575
},
{
"epoch": 32.0,
"learning_rate": 6.865714285714287e-06,
"loss": 0.0018,
"step": 1600
},
{
"epoch": 32.5,
"learning_rate": 6.794285714285714e-06,
"loss": 0.0013,
"step": 1625
},
{
"epoch": 33.0,
"learning_rate": 6.722857142857143e-06,
"loss": 0.0014,
"step": 1650
},
{
"epoch": 33.5,
"learning_rate": 6.651428571428572e-06,
"loss": 0.0011,
"step": 1675
},
{
"epoch": 34.0,
"learning_rate": 6.5800000000000005e-06,
"loss": 0.0032,
"step": 1700
},
{
"epoch": 34.5,
"learning_rate": 6.5085714285714295e-06,
"loss": 0.0019,
"step": 1725
},
{
"epoch": 35.0,
"learning_rate": 6.437142857142858e-06,
"loss": 0.0026,
"step": 1750
},
{
"epoch": 35.5,
"learning_rate": 6.365714285714286e-06,
"loss": 0.0017,
"step": 1775
},
{
"epoch": 36.0,
"learning_rate": 6.294285714285715e-06,
"loss": 0.0023,
"step": 1800
},
{
"epoch": 36.5,
"learning_rate": 6.222857142857144e-06,
"loss": 0.0018,
"step": 1825
},
{
"epoch": 37.0,
"learning_rate": 6.151428571428571e-06,
"loss": 0.0023,
"step": 1850
},
{
"epoch": 37.5,
"learning_rate": 6.08e-06,
"loss": 0.0027,
"step": 1875
},
{
"epoch": 38.0,
"learning_rate": 6.008571428571429e-06,
"loss": 0.0018,
"step": 1900
},
{
"epoch": 38.5,
"learning_rate": 5.937142857142858e-06,
"loss": 0.0011,
"step": 1925
},
{
"epoch": 39.0,
"learning_rate": 5.865714285714286e-06,
"loss": 0.0014,
"step": 1950
},
{
"epoch": 39.5,
"learning_rate": 5.794285714285715e-06,
"loss": 0.0017,
"step": 1975
},
{
"epoch": 40.0,
"learning_rate": 5.722857142857144e-06,
"loss": 0.0011,
"step": 2000
},
{
"epoch": 40.0,
"eval_loss": 1.257658839225769,
"eval_runtime": 176.0293,
"eval_samples_per_second": 1.5,
"eval_steps_per_second": 0.187,
"eval_wer": 42.2978067169294,
"step": 2000
},
{
"epoch": 40.5,
"learning_rate": 5.651428571428572e-06,
"loss": 0.0008,
"step": 2025
},
{
"epoch": 41.0,
"learning_rate": 5.580000000000001e-06,
"loss": 0.0011,
"step": 2050
},
{
"epoch": 41.5,
"learning_rate": 5.508571428571429e-06,
"loss": 0.0007,
"step": 2075
},
{
"epoch": 42.0,
"learning_rate": 5.437142857142857e-06,
"loss": 0.0009,
"step": 2100
},
{
"epoch": 42.5,
"learning_rate": 5.365714285714286e-06,
"loss": 0.0007,
"step": 2125
},
{
"epoch": 43.0,
"learning_rate": 5.294285714285715e-06,
"loss": 0.0006,
"step": 2150
},
{
"epoch": 43.5,
"learning_rate": 5.2228571428571425e-06,
"loss": 0.0006,
"step": 2175
},
{
"epoch": 44.0,
"learning_rate": 5.1514285714285715e-06,
"loss": 0.0009,
"step": 2200
},
{
"epoch": 44.5,
"learning_rate": 5.0800000000000005e-06,
"loss": 0.0006,
"step": 2225
},
{
"epoch": 45.0,
"learning_rate": 5.0085714285714295e-06,
"loss": 0.0006,
"step": 2250
},
{
"epoch": 45.5,
"learning_rate": 4.937142857142858e-06,
"loss": 0.0006,
"step": 2275
},
{
"epoch": 46.0,
"learning_rate": 4.865714285714287e-06,
"loss": 0.0005,
"step": 2300
},
{
"epoch": 46.5,
"learning_rate": 4.794285714285715e-06,
"loss": 0.0005,
"step": 2325
},
{
"epoch": 47.0,
"learning_rate": 4.722857142857144e-06,
"loss": 0.0005,
"step": 2350
},
{
"epoch": 47.5,
"learning_rate": 4.651428571428572e-06,
"loss": 0.0005,
"step": 2375
},
{
"epoch": 48.0,
"learning_rate": 4.58e-06,
"loss": 0.0005,
"step": 2400
},
{
"epoch": 48.5,
"learning_rate": 4.508571428571429e-06,
"loss": 0.0004,
"step": 2425
},
{
"epoch": 49.0,
"learning_rate": 4.437142857142857e-06,
"loss": 0.0005,
"step": 2450
},
{
"epoch": 49.5,
"learning_rate": 4.3657142857142855e-06,
"loss": 0.0004,
"step": 2475
},
{
"epoch": 50.0,
"learning_rate": 4.2942857142857146e-06,
"loss": 0.0005,
"step": 2500
},
{
"epoch": 50.5,
"learning_rate": 4.222857142857143e-06,
"loss": 0.0004,
"step": 2525
},
{
"epoch": 51.0,
"learning_rate": 4.151428571428572e-06,
"loss": 0.0005,
"step": 2550
},
{
"epoch": 51.5,
"learning_rate": 4.08e-06,
"loss": 0.0008,
"step": 2575
},
{
"epoch": 52.0,
"learning_rate": 4.008571428571429e-06,
"loss": 0.0004,
"step": 2600
},
{
"epoch": 52.5,
"learning_rate": 3.937142857142858e-06,
"loss": 0.0004,
"step": 2625
},
{
"epoch": 53.0,
"learning_rate": 3.865714285714286e-06,
"loss": 0.0007,
"step": 2650
},
{
"epoch": 53.5,
"learning_rate": 3.7942857142857147e-06,
"loss": 0.0004,
"step": 2675
},
{
"epoch": 54.0,
"learning_rate": 3.722857142857143e-06,
"loss": 0.0004,
"step": 2700
},
{
"epoch": 54.5,
"learning_rate": 3.651428571428572e-06,
"loss": 0.0005,
"step": 2725
},
{
"epoch": 55.0,
"learning_rate": 3.58e-06,
"loss": 0.0004,
"step": 2750
},
{
"epoch": 55.5,
"learning_rate": 3.508571428571429e-06,
"loss": 0.0005,
"step": 2775
},
{
"epoch": 56.0,
"learning_rate": 3.437142857142857e-06,
"loss": 0.0004,
"step": 2800
},
{
"epoch": 56.5,
"learning_rate": 3.3657142857142862e-06,
"loss": 0.0006,
"step": 2825
},
{
"epoch": 57.0,
"learning_rate": 3.2942857142857144e-06,
"loss": 0.0003,
"step": 2850
},
{
"epoch": 57.5,
"learning_rate": 3.222857142857143e-06,
"loss": 0.0003,
"step": 2875
},
{
"epoch": 58.0,
"learning_rate": 3.151428571428572e-06,
"loss": 0.0004,
"step": 2900
},
{
"epoch": 58.5,
"learning_rate": 3.08e-06,
"loss": 0.0004,
"step": 2925
},
{
"epoch": 59.0,
"learning_rate": 3.008571428571429e-06,
"loss": 0.0003,
"step": 2950
},
{
"epoch": 59.5,
"learning_rate": 2.9371428571428573e-06,
"loss": 0.0003,
"step": 2975
},
{
"epoch": 60.0,
"learning_rate": 2.865714285714286e-06,
"loss": 0.0003,
"step": 3000
},
{
"epoch": 60.0,
"eval_loss": 1.3213410377502441,
"eval_runtime": 174.6225,
"eval_samples_per_second": 1.512,
"eval_steps_per_second": 0.189,
"eval_wer": 41.8351610692255,
"step": 3000
}
],
"max_steps": 4000,
"num_train_epochs": 80,
"total_flos": 1.37308933472256e+19,
"trial_name": null,
"trial_params": null
}