whisper-hu-small-augmented / trainer_state.json
Moreno La Quatra
End of training
e08e747
raw
history blame
15.2 kB
{
"best_metric": 31.12362881707679,
"best_model_checkpoint": "whisper-hu-small-augmented/checkpoint-5000",
"epoch": 74.6268656716418,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.75,
"learning_rate": 4.800000000000001e-06,
"loss": 2.6486,
"step": 50
},
{
"epoch": 1.49,
"learning_rate": 9.800000000000001e-06,
"loss": 1.1943,
"step": 100
},
{
"epoch": 2.24,
"learning_rate": 1.48e-05,
"loss": 0.7269,
"step": 150
},
{
"epoch": 2.99,
"learning_rate": 1.9800000000000004e-05,
"loss": 0.5174,
"step": 200
},
{
"epoch": 3.73,
"learning_rate": 2.48e-05,
"loss": 0.3459,
"step": 250
},
{
"epoch": 4.48,
"learning_rate": 2.98e-05,
"loss": 0.252,
"step": 300
},
{
"epoch": 5.22,
"learning_rate": 3.48e-05,
"loss": 0.1918,
"step": 350
},
{
"epoch": 5.97,
"learning_rate": 3.9800000000000005e-05,
"loss": 0.1365,
"step": 400
},
{
"epoch": 6.72,
"learning_rate": 4.4800000000000005e-05,
"loss": 0.0893,
"step": 450
},
{
"epoch": 7.46,
"learning_rate": 4.9800000000000004e-05,
"loss": 0.0763,
"step": 500
},
{
"epoch": 7.46,
"eval_loss": 0.5267707109451294,
"eval_runtime": 1348.4551,
"eval_samples_per_second": 3.434,
"eval_steps_per_second": 0.054,
"eval_wer": 40.50993181144381,
"step": 500
},
{
"epoch": 8.21,
"learning_rate": 4.9466666666666665e-05,
"loss": 0.0673,
"step": 550
},
{
"epoch": 8.96,
"learning_rate": 4.8911111111111116e-05,
"loss": 0.0532,
"step": 600
},
{
"epoch": 9.7,
"learning_rate": 4.835555555555556e-05,
"loss": 0.0396,
"step": 650
},
{
"epoch": 10.45,
"learning_rate": 4.78e-05,
"loss": 0.0349,
"step": 700
},
{
"epoch": 11.19,
"learning_rate": 4.724444444444445e-05,
"loss": 0.031,
"step": 750
},
{
"epoch": 11.94,
"learning_rate": 4.668888888888889e-05,
"loss": 0.0262,
"step": 800
},
{
"epoch": 12.69,
"learning_rate": 4.6133333333333334e-05,
"loss": 0.0222,
"step": 850
},
{
"epoch": 13.43,
"learning_rate": 4.557777777777778e-05,
"loss": 0.0203,
"step": 900
},
{
"epoch": 14.18,
"learning_rate": 4.502222222222223e-05,
"loss": 0.0183,
"step": 950
},
{
"epoch": 14.93,
"learning_rate": 4.4466666666666666e-05,
"loss": 0.0147,
"step": 1000
},
{
"epoch": 14.93,
"eval_loss": 0.5232693552970886,
"eval_runtime": 1472.4775,
"eval_samples_per_second": 3.145,
"eval_steps_per_second": 0.05,
"eval_wer": 36.142899495997625,
"step": 1000
},
{
"epoch": 15.67,
"learning_rate": 4.3911111111111116e-05,
"loss": 0.013,
"step": 1050
},
{
"epoch": 16.42,
"learning_rate": 4.335555555555556e-05,
"loss": 0.0126,
"step": 1100
},
{
"epoch": 17.16,
"learning_rate": 4.2800000000000004e-05,
"loss": 0.0117,
"step": 1150
},
{
"epoch": 17.91,
"learning_rate": 4.224444444444445e-05,
"loss": 0.0104,
"step": 1200
},
{
"epoch": 18.66,
"learning_rate": 4.168888888888889e-05,
"loss": 0.0062,
"step": 1250
},
{
"epoch": 19.4,
"learning_rate": 4.1133333333333335e-05,
"loss": 0.0058,
"step": 1300
},
{
"epoch": 20.15,
"learning_rate": 4.057777777777778e-05,
"loss": 0.0056,
"step": 1350
},
{
"epoch": 20.9,
"learning_rate": 4.002222222222222e-05,
"loss": 0.0059,
"step": 1400
},
{
"epoch": 21.64,
"learning_rate": 3.9466666666666666e-05,
"loss": 0.0058,
"step": 1450
},
{
"epoch": 22.39,
"learning_rate": 3.8911111111111117e-05,
"loss": 0.0064,
"step": 1500
},
{
"epoch": 22.39,
"eval_loss": 0.546664834022522,
"eval_runtime": 1325.7539,
"eval_samples_per_second": 3.493,
"eval_steps_per_second": 0.055,
"eval_wer": 35.09338867477023,
"step": 1500
},
{
"epoch": 23.13,
"learning_rate": 3.8355555555555553e-05,
"loss": 0.0063,
"step": 1550
},
{
"epoch": 23.88,
"learning_rate": 3.7800000000000004e-05,
"loss": 0.0052,
"step": 1600
},
{
"epoch": 24.63,
"learning_rate": 3.724444444444445e-05,
"loss": 0.004,
"step": 1650
},
{
"epoch": 25.37,
"learning_rate": 3.668888888888889e-05,
"loss": 0.0042,
"step": 1700
},
{
"epoch": 26.12,
"learning_rate": 3.6133333333333335e-05,
"loss": 0.0031,
"step": 1750
},
{
"epoch": 26.87,
"learning_rate": 3.557777777777778e-05,
"loss": 0.0028,
"step": 1800
},
{
"epoch": 27.61,
"learning_rate": 3.502222222222222e-05,
"loss": 0.003,
"step": 1850
},
{
"epoch": 28.36,
"learning_rate": 3.4466666666666666e-05,
"loss": 0.0038,
"step": 1900
},
{
"epoch": 29.1,
"learning_rate": 3.391111111111111e-05,
"loss": 0.0037,
"step": 1950
},
{
"epoch": 29.85,
"learning_rate": 3.3355555555555554e-05,
"loss": 0.0045,
"step": 2000
},
{
"epoch": 29.85,
"eval_loss": 0.5434173941612244,
"eval_runtime": 1311.8792,
"eval_samples_per_second": 3.53,
"eval_steps_per_second": 0.056,
"eval_wer": 34.2929143195968,
"step": 2000
},
{
"epoch": 30.6,
"learning_rate": 3.2800000000000004e-05,
"loss": 0.0038,
"step": 2050
},
{
"epoch": 31.34,
"learning_rate": 3.224444444444444e-05,
"loss": 0.0038,
"step": 2100
},
{
"epoch": 32.09,
"learning_rate": 3.168888888888889e-05,
"loss": 0.0033,
"step": 2150
},
{
"epoch": 32.84,
"learning_rate": 3.1133333333333336e-05,
"loss": 0.003,
"step": 2200
},
{
"epoch": 33.58,
"learning_rate": 3.057777777777778e-05,
"loss": 0.0027,
"step": 2250
},
{
"epoch": 34.33,
"learning_rate": 3.0022222222222223e-05,
"loss": 0.0029,
"step": 2300
},
{
"epoch": 35.07,
"learning_rate": 2.946666666666667e-05,
"loss": 0.0028,
"step": 2350
},
{
"epoch": 35.82,
"learning_rate": 2.891111111111111e-05,
"loss": 0.002,
"step": 2400
},
{
"epoch": 36.57,
"learning_rate": 2.8355555555555558e-05,
"loss": 0.002,
"step": 2450
},
{
"epoch": 37.31,
"learning_rate": 2.7800000000000005e-05,
"loss": 0.0019,
"step": 2500
},
{
"epoch": 37.31,
"eval_loss": 0.5347548723220825,
"eval_runtime": 1268.6864,
"eval_samples_per_second": 3.65,
"eval_steps_per_second": 0.058,
"eval_wer": 32.78683664393715,
"step": 2500
},
{
"epoch": 38.06,
"learning_rate": 2.7244444444444445e-05,
"loss": 0.0021,
"step": 2550
},
{
"epoch": 38.81,
"learning_rate": 2.6688888888888892e-05,
"loss": 0.002,
"step": 2600
},
{
"epoch": 39.55,
"learning_rate": 2.6133333333333333e-05,
"loss": 0.0014,
"step": 2650
},
{
"epoch": 40.3,
"learning_rate": 2.557777777777778e-05,
"loss": 0.0019,
"step": 2700
},
{
"epoch": 41.04,
"learning_rate": 2.5022222222222224e-05,
"loss": 0.0012,
"step": 2750
},
{
"epoch": 41.79,
"learning_rate": 2.4466666666666667e-05,
"loss": 0.0014,
"step": 2800
},
{
"epoch": 42.54,
"learning_rate": 2.391111111111111e-05,
"loss": 0.001,
"step": 2850
},
{
"epoch": 43.28,
"learning_rate": 2.3355555555555555e-05,
"loss": 0.0013,
"step": 2900
},
{
"epoch": 44.03,
"learning_rate": 2.2800000000000002e-05,
"loss": 0.0015,
"step": 2950
},
{
"epoch": 44.78,
"learning_rate": 2.2244444444444446e-05,
"loss": 0.0008,
"step": 3000
},
{
"epoch": 44.78,
"eval_loss": 0.5314146876335144,
"eval_runtime": 1200.2965,
"eval_samples_per_second": 3.858,
"eval_steps_per_second": 0.061,
"eval_wer": 32.0604802846131,
"step": 3000
},
{
"epoch": 45.52,
"learning_rate": 2.168888888888889e-05,
"loss": 0.0007,
"step": 3050
},
{
"epoch": 46.27,
"learning_rate": 2.1133333333333337e-05,
"loss": 0.001,
"step": 3100
},
{
"epoch": 47.01,
"learning_rate": 2.057777777777778e-05,
"loss": 0.0009,
"step": 3150
},
{
"epoch": 47.76,
"learning_rate": 2.0022222222222224e-05,
"loss": 0.0006,
"step": 3200
},
{
"epoch": 48.51,
"learning_rate": 1.9466666666666668e-05,
"loss": 0.0006,
"step": 3250
},
{
"epoch": 49.25,
"learning_rate": 1.891111111111111e-05,
"loss": 0.0007,
"step": 3300
},
{
"epoch": 50.0,
"learning_rate": 1.8355555555555555e-05,
"loss": 0.0006,
"step": 3350
},
{
"epoch": 50.75,
"learning_rate": 1.78e-05,
"loss": 0.0008,
"step": 3400
},
{
"epoch": 51.49,
"learning_rate": 1.7244444444444446e-05,
"loss": 0.0008,
"step": 3450
},
{
"epoch": 52.24,
"learning_rate": 1.668888888888889e-05,
"loss": 0.0008,
"step": 3500
},
{
"epoch": 52.24,
"eval_loss": 0.5438169836997986,
"eval_runtime": 1352.7551,
"eval_samples_per_second": 3.423,
"eval_steps_per_second": 0.054,
"eval_wer": 32.69196560924993,
"step": 3500
},
{
"epoch": 52.99,
"learning_rate": 1.6133333333333334e-05,
"loss": 0.001,
"step": 3550
},
{
"epoch": 53.73,
"learning_rate": 1.5577777777777777e-05,
"loss": 0.0007,
"step": 3600
},
{
"epoch": 54.48,
"learning_rate": 1.5022222222222224e-05,
"loss": 0.0008,
"step": 3650
},
{
"epoch": 55.22,
"learning_rate": 1.4466666666666667e-05,
"loss": 0.0008,
"step": 3700
},
{
"epoch": 55.97,
"learning_rate": 1.391111111111111e-05,
"loss": 0.0007,
"step": 3750
},
{
"epoch": 56.72,
"learning_rate": 1.3355555555555557e-05,
"loss": 0.0007,
"step": 3800
},
{
"epoch": 57.46,
"learning_rate": 1.2800000000000001e-05,
"loss": 0.0004,
"step": 3850
},
{
"epoch": 58.21,
"learning_rate": 1.2244444444444445e-05,
"loss": 0.0006,
"step": 3900
},
{
"epoch": 58.96,
"learning_rate": 1.168888888888889e-05,
"loss": 0.0006,
"step": 3950
},
{
"epoch": 59.7,
"learning_rate": 1.1133333333333334e-05,
"loss": 0.0005,
"step": 4000
},
{
"epoch": 59.7,
"eval_loss": 0.5428301095962524,
"eval_runtime": 1301.4476,
"eval_samples_per_second": 3.558,
"eval_steps_per_second": 0.056,
"eval_wer": 32.093092202786835,
"step": 4000
},
{
"epoch": 60.45,
"learning_rate": 1.0577777777777778e-05,
"loss": 0.0004,
"step": 4050
},
{
"epoch": 61.19,
"learning_rate": 1.0022222222222223e-05,
"loss": 0.0006,
"step": 4100
},
{
"epoch": 61.94,
"learning_rate": 9.466666666666667e-06,
"loss": 0.0005,
"step": 4150
},
{
"epoch": 62.69,
"learning_rate": 8.911111111111112e-06,
"loss": 0.0004,
"step": 4200
},
{
"epoch": 63.43,
"learning_rate": 8.355555555555556e-06,
"loss": 0.0004,
"step": 4250
},
{
"epoch": 64.18,
"learning_rate": 7.8e-06,
"loss": 0.0005,
"step": 4300
},
{
"epoch": 64.93,
"learning_rate": 7.244444444444445e-06,
"loss": 0.0005,
"step": 4350
},
{
"epoch": 65.67,
"learning_rate": 6.688888888888889e-06,
"loss": 0.0003,
"step": 4400
},
{
"epoch": 66.42,
"learning_rate": 6.133333333333334e-06,
"loss": 0.0006,
"step": 4450
},
{
"epoch": 67.16,
"learning_rate": 5.577777777777778e-06,
"loss": 0.0003,
"step": 4500
},
{
"epoch": 67.16,
"eval_loss": 0.5328006744384766,
"eval_runtime": 1200.6149,
"eval_samples_per_second": 3.857,
"eval_steps_per_second": 0.061,
"eval_wer": 31.25111176993774,
"step": 4500
},
{
"epoch": 67.91,
"learning_rate": 5.022222222222223e-06,
"loss": 0.0004,
"step": 4550
},
{
"epoch": 68.66,
"learning_rate": 4.4666666666666665e-06,
"loss": 0.0003,
"step": 4600
},
{
"epoch": 69.4,
"learning_rate": 3.911111111111111e-06,
"loss": 0.0003,
"step": 4650
},
{
"epoch": 70.15,
"learning_rate": 3.3555555555555557e-06,
"loss": 0.0003,
"step": 4700
},
{
"epoch": 70.9,
"learning_rate": 2.8000000000000003e-06,
"loss": 0.0003,
"step": 4750
},
{
"epoch": 71.64,
"learning_rate": 2.2444444444444445e-06,
"loss": 0.0002,
"step": 4800
},
{
"epoch": 72.39,
"learning_rate": 1.6888888888888888e-06,
"loss": 0.0003,
"step": 4850
},
{
"epoch": 73.13,
"learning_rate": 1.1333333333333334e-06,
"loss": 0.0002,
"step": 4900
},
{
"epoch": 73.88,
"learning_rate": 5.777777777777778e-07,
"loss": 0.0002,
"step": 4950
},
{
"epoch": 74.63,
"learning_rate": 2.2222222222222224e-08,
"loss": 0.0004,
"step": 5000
},
{
"epoch": 74.63,
"eval_loss": 0.529201865196228,
"eval_runtime": 1279.8356,
"eval_samples_per_second": 3.618,
"eval_steps_per_second": 0.057,
"eval_wer": 31.12362881707679,
"step": 5000
},
{
"epoch": 74.63,
"step": 5000,
"total_flos": 1.8388315405910016e+20,
"train_loss": 0.06695196392126382,
"train_runtime": 71734.9338,
"train_samples_per_second": 8.922,
"train_steps_per_second": 0.07
}
],
"max_steps": 5000,
"num_train_epochs": 75,
"total_flos": 1.8388315405910016e+20,
"trial_name": null,
"trial_params": null
}