whisper-small-korean / trainer_state.json
steja's picture
korean whisper small ft
891da01
{
"best_metric": 13.012854375770383,
"best_model_checkpoint": "./whisper-small-korean/checkpoint-1000",
"epoch": 71.41379310344827,
"global_step": 1000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.76,
"learning_rate": 6.611096473669596e-06,
"loss": 0.2939,
"step": 25
},
{
"epoch": 3.55,
"learning_rate": 8.360489289678585e-06,
"loss": 0.0651,
"step": 50
},
{
"epoch": 5.34,
"learning_rate": 9.316614300602277e-06,
"loss": 0.0166,
"step": 75
},
{
"epoch": 7.14,
"learning_rate": 9.978175972987748e-06,
"loss": 0.0063,
"step": 100
},
{
"epoch": 8.9,
"learning_rate": 9.733333333333334e-06,
"loss": 0.0044,
"step": 125
},
{
"epoch": 10.69,
"learning_rate": 9.444444444444445e-06,
"loss": 0.0034,
"step": 150
},
{
"epoch": 12.48,
"learning_rate": 9.155555555555557e-06,
"loss": 0.0059,
"step": 175
},
{
"epoch": 14.28,
"learning_rate": 8.866666666666668e-06,
"loss": 0.004,
"step": 200
},
{
"epoch": 16.07,
"learning_rate": 8.577777777777778e-06,
"loss": 0.0028,
"step": 225
},
{
"epoch": 17.83,
"learning_rate": 8.3e-06,
"loss": 0.0014,
"step": 250
},
{
"epoch": 19.62,
"learning_rate": 8.011111111111113e-06,
"loss": 0.0011,
"step": 275
},
{
"epoch": 21.41,
"learning_rate": 7.722222222222223e-06,
"loss": 0.0009,
"step": 300
},
{
"epoch": 23.21,
"learning_rate": 7.433333333333334e-06,
"loss": 0.0009,
"step": 325
},
{
"epoch": 24.97,
"learning_rate": 7.1444444444444446e-06,
"loss": 0.0008,
"step": 350
},
{
"epoch": 26.76,
"learning_rate": 6.855555555555556e-06,
"loss": 0.0007,
"step": 375
},
{
"epoch": 28.55,
"learning_rate": 6.566666666666667e-06,
"loss": 0.0007,
"step": 400
},
{
"epoch": 30.34,
"learning_rate": 6.277777777777778e-06,
"loss": 0.0006,
"step": 425
},
{
"epoch": 32.14,
"learning_rate": 5.98888888888889e-06,
"loss": 0.0006,
"step": 450
},
{
"epoch": 33.9,
"learning_rate": 5.711111111111112e-06,
"loss": 0.0005,
"step": 475
},
{
"epoch": 35.69,
"learning_rate": 5.422222222222223e-06,
"loss": 0.0005,
"step": 500
},
{
"epoch": 35.69,
"eval_loss": 0.31884765625,
"eval_runtime": 89.0963,
"eval_samples_per_second": 4.287,
"eval_steps_per_second": 0.067,
"eval_wer": 13.030463109702412,
"step": 500
},
{
"epoch": 37.48,
"learning_rate": 5.133333333333334e-06,
"loss": 0.0005,
"step": 525
},
{
"epoch": 39.28,
"learning_rate": 4.8444444444444446e-06,
"loss": 0.0005,
"step": 550
},
{
"epoch": 41.07,
"learning_rate": 4.555555555555556e-06,
"loss": 0.0005,
"step": 575
},
{
"epoch": 42.83,
"learning_rate": 4.266666666666668e-06,
"loss": 0.0004,
"step": 600
},
{
"epoch": 44.62,
"learning_rate": 3.977777777777778e-06,
"loss": 0.0004,
"step": 625
},
{
"epoch": 46.41,
"learning_rate": 3.688888888888889e-06,
"loss": 0.0004,
"step": 650
},
{
"epoch": 48.21,
"learning_rate": 3.4000000000000005e-06,
"loss": 0.0004,
"step": 675
},
{
"epoch": 49.97,
"learning_rate": 3.1222222222222228e-06,
"loss": 0.0004,
"step": 700
},
{
"epoch": 51.76,
"learning_rate": 2.8333333333333335e-06,
"loss": 0.0004,
"step": 725
},
{
"epoch": 53.55,
"learning_rate": 2.5444444444444446e-06,
"loss": 0.0004,
"step": 750
},
{
"epoch": 55.34,
"learning_rate": 2.2555555555555557e-06,
"loss": 0.0004,
"step": 775
},
{
"epoch": 57.14,
"learning_rate": 1.9666666666666668e-06,
"loss": 0.0004,
"step": 800
},
{
"epoch": 58.9,
"learning_rate": 1.6777777777777779e-06,
"loss": 0.0004,
"step": 825
},
{
"epoch": 60.69,
"learning_rate": 1.3888888888888892e-06,
"loss": 0.0004,
"step": 850
},
{
"epoch": 62.48,
"learning_rate": 1.1e-06,
"loss": 0.0004,
"step": 875
},
{
"epoch": 64.28,
"learning_rate": 8.111111111111112e-07,
"loss": 0.0004,
"step": 900
},
{
"epoch": 66.07,
"learning_rate": 5.222222222222223e-07,
"loss": 0.0003,
"step": 925
},
{
"epoch": 67.83,
"learning_rate": 2.444444444444445e-07,
"loss": 0.0003,
"step": 950
},
{
"epoch": 69.62,
"learning_rate": 0.0,
"loss": 0.0003,
"step": 975
},
{
"epoch": 71.41,
"learning_rate": 0.0,
"loss": 0.0003,
"step": 1000
},
{
"epoch": 71.41,
"eval_loss": 0.33154296875,
"eval_runtime": 89.0871,
"eval_samples_per_second": 4.288,
"eval_steps_per_second": 0.067,
"eval_wer": 13.012854375770383,
"step": 1000
},
{
"epoch": 71.41,
"step": 1000,
"total_flos": 6.854445413862841e+19,
"train_loss": 0.010475540935993195,
"train_runtime": 5731.9768,
"train_samples_per_second": 5.583,
"train_steps_per_second": 0.174
}
],
"max_steps": 1000,
"num_train_epochs": 72,
"total_flos": 6.854445413862841e+19,
"trial_name": null,
"trial_params": null
}