whisper-small-hi / checkpoint-3000 /trainer_state.json
saswatapcs17's picture
Upload with huggingface_hub
c8f2bd5
raw
history blame contribute delete
No virus
15.3 kB
{
"best_metric": 84.61538461538461,
"best_model_checkpoint": "./whisper-small-hi/checkpoint-2000",
"epoch": 3000.0,
"global_step": 3000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 25.0,
"learning_rate": 4.6000000000000004e-07,
"loss": 2.7845,
"step": 25
},
{
"epoch": 50.0,
"learning_rate": 9.600000000000001e-07,
"loss": 1.8784,
"step": 50
},
{
"epoch": 75.0,
"learning_rate": 1.46e-06,
"loss": 0.7949,
"step": 75
},
{
"epoch": 100.0,
"learning_rate": 1.9600000000000003e-06,
"loss": 0.2619,
"step": 100
},
{
"epoch": 125.0,
"learning_rate": 2.46e-06,
"loss": 0.1719,
"step": 125
},
{
"epoch": 150.0,
"learning_rate": 2.96e-06,
"loss": 0.1447,
"step": 150
},
{
"epoch": 175.0,
"learning_rate": 3.46e-06,
"loss": 0.1234,
"step": 175
},
{
"epoch": 200.0,
"learning_rate": 3.96e-06,
"loss": 0.0997,
"step": 200
},
{
"epoch": 225.0,
"learning_rate": 4.4600000000000005e-06,
"loss": 0.0658,
"step": 225
},
{
"epoch": 250.0,
"learning_rate": 4.960000000000001e-06,
"loss": 0.0054,
"step": 250
},
{
"epoch": 275.0,
"learning_rate": 5.460000000000001e-06,
"loss": 0.0006,
"step": 275
},
{
"epoch": 300.0,
"learning_rate": 5.9600000000000005e-06,
"loss": 0.0003,
"step": 300
},
{
"epoch": 325.0,
"learning_rate": 6.460000000000001e-06,
"loss": 0.0002,
"step": 325
},
{
"epoch": 350.0,
"learning_rate": 6.96e-06,
"loss": 0.0002,
"step": 350
},
{
"epoch": 375.0,
"learning_rate": 7.4600000000000006e-06,
"loss": 0.0002,
"step": 375
},
{
"epoch": 400.0,
"learning_rate": 7.960000000000002e-06,
"loss": 0.0001,
"step": 400
},
{
"epoch": 425.0,
"learning_rate": 8.46e-06,
"loss": 0.0001,
"step": 425
},
{
"epoch": 450.0,
"learning_rate": 8.96e-06,
"loss": 0.0001,
"step": 450
},
{
"epoch": 475.0,
"learning_rate": 9.460000000000001e-06,
"loss": 0.0001,
"step": 475
},
{
"epoch": 500.0,
"learning_rate": 9.960000000000001e-06,
"loss": 0.0001,
"step": 500
},
{
"epoch": 525.0,
"learning_rate": 9.934285714285715e-06,
"loss": 0.0001,
"step": 525
},
{
"epoch": 550.0,
"learning_rate": 9.862857142857144e-06,
"loss": 0.0001,
"step": 550
},
{
"epoch": 575.0,
"learning_rate": 9.791428571428571e-06,
"loss": 0.0001,
"step": 575
},
{
"epoch": 600.0,
"learning_rate": 9.72e-06,
"loss": 0.0001,
"step": 600
},
{
"epoch": 625.0,
"learning_rate": 9.648571428571429e-06,
"loss": 0.0001,
"step": 625
},
{
"epoch": 650.0,
"learning_rate": 9.577142857142858e-06,
"loss": 0.0001,
"step": 650
},
{
"epoch": 675.0,
"learning_rate": 9.505714285714287e-06,
"loss": 0.0,
"step": 675
},
{
"epoch": 700.0,
"learning_rate": 9.434285714285714e-06,
"loss": 0.0,
"step": 700
},
{
"epoch": 725.0,
"learning_rate": 9.362857142857143e-06,
"loss": 0.0,
"step": 725
},
{
"epoch": 750.0,
"learning_rate": 9.291428571428572e-06,
"loss": 0.0,
"step": 750
},
{
"epoch": 775.0,
"learning_rate": 9.220000000000002e-06,
"loss": 0.0,
"step": 775
},
{
"epoch": 800.0,
"learning_rate": 9.148571428571429e-06,
"loss": 0.0,
"step": 800
},
{
"epoch": 825.0,
"learning_rate": 9.077142857142858e-06,
"loss": 0.0,
"step": 825
},
{
"epoch": 850.0,
"learning_rate": 9.005714285714287e-06,
"loss": 0.0,
"step": 850
},
{
"epoch": 875.0,
"learning_rate": 8.934285714285716e-06,
"loss": 0.0,
"step": 875
},
{
"epoch": 900.0,
"learning_rate": 8.862857142857143e-06,
"loss": 0.0,
"step": 900
},
{
"epoch": 925.0,
"learning_rate": 8.791428571428572e-06,
"loss": 0.0,
"step": 925
},
{
"epoch": 950.0,
"learning_rate": 8.720000000000001e-06,
"loss": 0.0,
"step": 950
},
{
"epoch": 975.0,
"learning_rate": 8.64857142857143e-06,
"loss": 0.0,
"step": 975
},
{
"epoch": 1000.0,
"learning_rate": 8.577142857142858e-06,
"loss": 0.0,
"step": 1000
},
{
"epoch": 1000.0,
"eval_loss": 1.5850915908813477,
"eval_runtime": 3.2322,
"eval_samples_per_second": 0.619,
"eval_steps_per_second": 0.309,
"eval_wer": 92.3076923076923,
"step": 1000
},
{
"epoch": 1025.0,
"learning_rate": 8.505714285714287e-06,
"loss": 0.0,
"step": 1025
},
{
"epoch": 1050.0,
"learning_rate": 8.434285714285716e-06,
"loss": 0.0,
"step": 1050
},
{
"epoch": 1075.0,
"learning_rate": 8.362857142857143e-06,
"loss": 0.0,
"step": 1075
},
{
"epoch": 1100.0,
"learning_rate": 8.291428571428572e-06,
"loss": 0.0,
"step": 1100
},
{
"epoch": 1125.0,
"learning_rate": 8.220000000000001e-06,
"loss": 0.0,
"step": 1125
},
{
"epoch": 1150.0,
"learning_rate": 8.148571428571428e-06,
"loss": 0.0,
"step": 1150
},
{
"epoch": 1175.0,
"learning_rate": 8.077142857142857e-06,
"loss": 0.0,
"step": 1175
},
{
"epoch": 1200.0,
"learning_rate": 8.005714285714286e-06,
"loss": 0.0,
"step": 1200
},
{
"epoch": 1225.0,
"learning_rate": 7.934285714285715e-06,
"loss": 0.0,
"step": 1225
},
{
"epoch": 1250.0,
"learning_rate": 7.862857142857143e-06,
"loss": 0.0,
"step": 1250
},
{
"epoch": 1275.0,
"learning_rate": 7.791428571428572e-06,
"loss": 0.0,
"step": 1275
},
{
"epoch": 1300.0,
"learning_rate": 7.72e-06,
"loss": 0.0,
"step": 1300
},
{
"epoch": 1325.0,
"learning_rate": 7.64857142857143e-06,
"loss": 0.0,
"step": 1325
},
{
"epoch": 1350.0,
"learning_rate": 7.577142857142857e-06,
"loss": 0.0,
"step": 1350
},
{
"epoch": 1375.0,
"learning_rate": 7.505714285714286e-06,
"loss": 0.0,
"step": 1375
},
{
"epoch": 1400.0,
"learning_rate": 7.434285714285715e-06,
"loss": 0.0,
"step": 1400
},
{
"epoch": 1425.0,
"learning_rate": 7.362857142857144e-06,
"loss": 0.0,
"step": 1425
},
{
"epoch": 1450.0,
"learning_rate": 7.291428571428571e-06,
"loss": 0.0,
"step": 1450
},
{
"epoch": 1475.0,
"learning_rate": 7.22e-06,
"loss": 0.0,
"step": 1475
},
{
"epoch": 1500.0,
"learning_rate": 7.148571428571429e-06,
"loss": 0.0,
"step": 1500
},
{
"epoch": 1525.0,
"learning_rate": 7.077142857142858e-06,
"loss": 0.0,
"step": 1525
},
{
"epoch": 1550.0,
"learning_rate": 7.0057142857142865e-06,
"loss": 0.0,
"step": 1550
},
{
"epoch": 1575.0,
"learning_rate": 6.934285714285715e-06,
"loss": 0.0,
"step": 1575
},
{
"epoch": 1600.0,
"learning_rate": 6.862857142857144e-06,
"loss": 0.0,
"step": 1600
},
{
"epoch": 1625.0,
"learning_rate": 6.791428571428572e-06,
"loss": 0.0,
"step": 1625
},
{
"epoch": 1650.0,
"learning_rate": 6.720000000000001e-06,
"loss": 0.0,
"step": 1650
},
{
"epoch": 1675.0,
"learning_rate": 6.648571428571429e-06,
"loss": 0.0,
"step": 1675
},
{
"epoch": 1700.0,
"learning_rate": 6.577142857142857e-06,
"loss": 0.0,
"step": 1700
},
{
"epoch": 1725.0,
"learning_rate": 6.505714285714286e-06,
"loss": 0.0,
"step": 1725
},
{
"epoch": 1750.0,
"learning_rate": 6.434285714285715e-06,
"loss": 0.0,
"step": 1750
},
{
"epoch": 1775.0,
"learning_rate": 6.3628571428571426e-06,
"loss": 0.0,
"step": 1775
},
{
"epoch": 1800.0,
"learning_rate": 6.2914285714285716e-06,
"loss": 0.0,
"step": 1800
},
{
"epoch": 1825.0,
"learning_rate": 6.220000000000001e-06,
"loss": 0.0,
"step": 1825
},
{
"epoch": 1850.0,
"learning_rate": 6.14857142857143e-06,
"loss": 0.0,
"step": 1850
},
{
"epoch": 1875.0,
"learning_rate": 6.077142857142858e-06,
"loss": 0.0,
"step": 1875
},
{
"epoch": 1900.0,
"learning_rate": 6.005714285714286e-06,
"loss": 0.0,
"step": 1900
},
{
"epoch": 1925.0,
"learning_rate": 5.934285714285715e-06,
"loss": 0.0,
"step": 1925
},
{
"epoch": 1950.0,
"learning_rate": 5.862857142857143e-06,
"loss": 0.0,
"step": 1950
},
{
"epoch": 1975.0,
"learning_rate": 5.791428571428572e-06,
"loss": 0.0,
"step": 1975
},
{
"epoch": 2000.0,
"learning_rate": 5.72e-06,
"loss": 0.0,
"step": 2000
},
{
"epoch": 2000.0,
"eval_loss": 2.00115704536438,
"eval_runtime": 3.6044,
"eval_samples_per_second": 0.555,
"eval_steps_per_second": 0.277,
"eval_wer": 84.61538461538461,
"step": 2000
},
{
"epoch": 2025.0,
"learning_rate": 5.6485714285714285e-06,
"loss": 0.0,
"step": 2025
},
{
"epoch": 2050.0,
"learning_rate": 5.5771428571428575e-06,
"loss": 0.0,
"step": 2050
},
{
"epoch": 2075.0,
"learning_rate": 5.5057142857142865e-06,
"loss": 0.0,
"step": 2075
},
{
"epoch": 2100.0,
"learning_rate": 5.4342857142857155e-06,
"loss": 0.0,
"step": 2100
},
{
"epoch": 2125.0,
"learning_rate": 5.362857142857143e-06,
"loss": 0.0,
"step": 2125
},
{
"epoch": 2150.0,
"learning_rate": 5.291428571428572e-06,
"loss": 0.0,
"step": 2150
},
{
"epoch": 2175.0,
"learning_rate": 5.220000000000001e-06,
"loss": 0.0,
"step": 2175
},
{
"epoch": 2200.0,
"learning_rate": 5.14857142857143e-06,
"loss": 0.0,
"step": 2200
},
{
"epoch": 2225.0,
"learning_rate": 5.077142857142857e-06,
"loss": 0.0,
"step": 2225
},
{
"epoch": 2250.0,
"learning_rate": 5.005714285714286e-06,
"loss": 0.0,
"step": 2250
},
{
"epoch": 2275.0,
"learning_rate": 4.934285714285715e-06,
"loss": 0.0,
"step": 2275
},
{
"epoch": 2300.0,
"learning_rate": 4.862857142857143e-06,
"loss": 0.0,
"step": 2300
},
{
"epoch": 2325.0,
"learning_rate": 4.7914285714285715e-06,
"loss": 0.0,
"step": 2325
},
{
"epoch": 2350.0,
"learning_rate": 4.7200000000000005e-06,
"loss": 0.0,
"step": 2350
},
{
"epoch": 2375.0,
"learning_rate": 4.648571428571429e-06,
"loss": 0.0,
"step": 2375
},
{
"epoch": 2400.0,
"learning_rate": 4.577142857142858e-06,
"loss": 0.0,
"step": 2400
},
{
"epoch": 2425.0,
"learning_rate": 4.505714285714286e-06,
"loss": 0.0,
"step": 2425
},
{
"epoch": 2450.0,
"learning_rate": 4.434285714285715e-06,
"loss": 0.0,
"step": 2450
},
{
"epoch": 2475.0,
"learning_rate": 4.362857142857143e-06,
"loss": 0.0,
"step": 2475
},
{
"epoch": 2500.0,
"learning_rate": 4.291428571428572e-06,
"loss": 0.0,
"step": 2500
},
{
"epoch": 2525.0,
"learning_rate": 4.22e-06,
"loss": 0.0,
"step": 2525
},
{
"epoch": 2550.0,
"learning_rate": 4.148571428571429e-06,
"loss": 0.0,
"step": 2550
},
{
"epoch": 2575.0,
"learning_rate": 4.0771428571428574e-06,
"loss": 0.0,
"step": 2575
},
{
"epoch": 2600.0,
"learning_rate": 4.0057142857142864e-06,
"loss": 0.0,
"step": 2600
},
{
"epoch": 2625.0,
"learning_rate": 3.934285714285715e-06,
"loss": 0.0,
"step": 2625
},
{
"epoch": 2650.0,
"learning_rate": 3.862857142857143e-06,
"loss": 0.0,
"step": 2650
},
{
"epoch": 2675.0,
"learning_rate": 3.7914285714285722e-06,
"loss": 0.0,
"step": 2675
},
{
"epoch": 2700.0,
"learning_rate": 3.7200000000000004e-06,
"loss": 0.0,
"step": 2700
},
{
"epoch": 2725.0,
"learning_rate": 3.648571428571429e-06,
"loss": 0.0,
"step": 2725
},
{
"epoch": 2750.0,
"learning_rate": 3.5771428571428576e-06,
"loss": 0.0,
"step": 2750
},
{
"epoch": 2775.0,
"learning_rate": 3.505714285714286e-06,
"loss": 0.0,
"step": 2775
},
{
"epoch": 2800.0,
"learning_rate": 3.4342857142857143e-06,
"loss": 0.0,
"step": 2800
},
{
"epoch": 2825.0,
"learning_rate": 3.3628571428571433e-06,
"loss": 0.0,
"step": 2825
},
{
"epoch": 2850.0,
"learning_rate": 3.2914285714285715e-06,
"loss": 0.0,
"step": 2850
},
{
"epoch": 2875.0,
"learning_rate": 3.2200000000000005e-06,
"loss": 0.0,
"step": 2875
},
{
"epoch": 2900.0,
"learning_rate": 3.1485714285714287e-06,
"loss": 0.0,
"step": 2900
},
{
"epoch": 2925.0,
"learning_rate": 3.0771428571428573e-06,
"loss": 0.0,
"step": 2925
},
{
"epoch": 2950.0,
"learning_rate": 3.005714285714286e-06,
"loss": 0.0,
"step": 2950
},
{
"epoch": 2975.0,
"learning_rate": 2.9342857142857144e-06,
"loss": 0.0,
"step": 2975
},
{
"epoch": 3000.0,
"learning_rate": 2.8628571428571435e-06,
"loss": 0.0,
"step": 3000
},
{
"epoch": 3000.0,
"eval_loss": 2.211437225341797,
"eval_runtime": 1.8782,
"eval_samples_per_second": 1.065,
"eval_steps_per_second": 0.532,
"eval_wer": 100.0,
"step": 3000
}
],
"max_steps": 4000,
"num_train_epochs": 4000,
"total_flos": 8.6575620096e+18,
"trial_name": null,
"trial_params": null
}