whisper-large-v3-ca / trainer_state.json
zuazo's picture
End of training
e0526ab verified
{
"best_metric": 5.919405544441269,
"best_model_checkpoint": "./checkpoint-18000",
"epoch": 39.0625,
"eval_steps": 1000,
"global_step": 20000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.05,
"learning_rate": 5.000000000000001e-07,
"loss": 0.7211,
"step": 25
},
{
"epoch": 0.1,
"learning_rate": 1.0000000000000002e-06,
"loss": 0.3652,
"step": 50
},
{
"epoch": 0.15,
"learning_rate": 1.5e-06,
"loss": 0.1882,
"step": 75
},
{
"epoch": 0.2,
"learning_rate": 2.0000000000000003e-06,
"loss": 0.1748,
"step": 100
},
{
"epoch": 0.24,
"learning_rate": 2.5e-06,
"loss": 0.1762,
"step": 125
},
{
"epoch": 0.29,
"learning_rate": 3e-06,
"loss": 0.1548,
"step": 150
},
{
"epoch": 0.34,
"learning_rate": 3.5e-06,
"loss": 0.1513,
"step": 175
},
{
"epoch": 0.39,
"learning_rate": 4.000000000000001e-06,
"loss": 0.1558,
"step": 200
},
{
"epoch": 0.44,
"learning_rate": 4.5e-06,
"loss": 0.1383,
"step": 225
},
{
"epoch": 0.49,
"learning_rate": 5e-06,
"loss": 0.1368,
"step": 250
},
{
"epoch": 0.54,
"learning_rate": 5.500000000000001e-06,
"loss": 0.1452,
"step": 275
},
{
"epoch": 0.59,
"learning_rate": 6e-06,
"loss": 0.1499,
"step": 300
},
{
"epoch": 0.63,
"learning_rate": 6.5000000000000004e-06,
"loss": 0.135,
"step": 325
},
{
"epoch": 0.68,
"learning_rate": 7e-06,
"loss": 0.1395,
"step": 350
},
{
"epoch": 0.73,
"learning_rate": 7.500000000000001e-06,
"loss": 0.1371,
"step": 375
},
{
"epoch": 0.78,
"learning_rate": 8.000000000000001e-06,
"loss": 0.1437,
"step": 400
},
{
"epoch": 0.83,
"learning_rate": 8.5e-06,
"loss": 0.1421,
"step": 425
},
{
"epoch": 0.88,
"learning_rate": 9e-06,
"loss": 0.1363,
"step": 450
},
{
"epoch": 0.93,
"learning_rate": 9.5e-06,
"loss": 0.1439,
"step": 475
},
{
"epoch": 0.98,
"learning_rate": 1e-05,
"loss": 0.1335,
"step": 500
},
{
"epoch": 1.03,
"learning_rate": 9.987179487179488e-06,
"loss": 0.1194,
"step": 525
},
{
"epoch": 1.07,
"learning_rate": 9.974358974358974e-06,
"loss": 0.0915,
"step": 550
},
{
"epoch": 1.12,
"learning_rate": 9.961538461538463e-06,
"loss": 0.0911,
"step": 575
},
{
"epoch": 1.17,
"learning_rate": 9.94871794871795e-06,
"loss": 0.0905,
"step": 600
},
{
"epoch": 1.22,
"learning_rate": 9.935897435897437e-06,
"loss": 0.0909,
"step": 625
},
{
"epoch": 1.27,
"learning_rate": 9.923076923076923e-06,
"loss": 0.0907,
"step": 650
},
{
"epoch": 1.32,
"learning_rate": 9.910256410256412e-06,
"loss": 0.0966,
"step": 675
},
{
"epoch": 1.37,
"learning_rate": 9.897435897435899e-06,
"loss": 0.0894,
"step": 700
},
{
"epoch": 1.42,
"learning_rate": 9.884615384615386e-06,
"loss": 0.095,
"step": 725
},
{
"epoch": 1.46,
"learning_rate": 9.871794871794872e-06,
"loss": 0.0926,
"step": 750
},
{
"epoch": 1.51,
"learning_rate": 9.858974358974361e-06,
"loss": 0.096,
"step": 775
},
{
"epoch": 1.56,
"learning_rate": 9.846153846153848e-06,
"loss": 0.0969,
"step": 800
},
{
"epoch": 1.61,
"learning_rate": 9.833333333333333e-06,
"loss": 0.0895,
"step": 825
},
{
"epoch": 1.66,
"learning_rate": 9.820512820512821e-06,
"loss": 0.0903,
"step": 850
},
{
"epoch": 1.71,
"learning_rate": 9.807692307692308e-06,
"loss": 0.0909,
"step": 875
},
{
"epoch": 1.76,
"learning_rate": 9.794871794871795e-06,
"loss": 0.0923,
"step": 900
},
{
"epoch": 1.81,
"learning_rate": 9.782051282051282e-06,
"loss": 0.0987,
"step": 925
},
{
"epoch": 1.86,
"learning_rate": 9.76923076923077e-06,
"loss": 0.0933,
"step": 950
},
{
"epoch": 1.9,
"learning_rate": 9.756410256410257e-06,
"loss": 0.0941,
"step": 975
},
{
"epoch": 1.95,
"learning_rate": 9.743589743589744e-06,
"loss": 0.0988,
"step": 1000
},
{
"epoch": 1.95,
"eval_loss": 0.1487419754266739,
"eval_runtime": 3542.1773,
"eval_samples_per_second": 4.624,
"eval_steps_per_second": 0.289,
"eval_wer": 6.561874821377536,
"step": 1000
},
{
"epoch": 2.0,
"learning_rate": 9.730769230769231e-06,
"loss": 0.0942,
"step": 1025
},
{
"epoch": 2.05,
"learning_rate": 9.71794871794872e-06,
"loss": 0.0409,
"step": 1050
},
{
"epoch": 2.1,
"learning_rate": 9.705128205128206e-06,
"loss": 0.0408,
"step": 1075
},
{
"epoch": 2.15,
"learning_rate": 9.692307692307693e-06,
"loss": 0.0415,
"step": 1100
},
{
"epoch": 2.2,
"learning_rate": 9.67948717948718e-06,
"loss": 0.0483,
"step": 1125
},
{
"epoch": 2.25,
"learning_rate": 9.666666666666667e-06,
"loss": 0.0427,
"step": 1150
},
{
"epoch": 2.29,
"learning_rate": 9.653846153846155e-06,
"loss": 0.0509,
"step": 1175
},
{
"epoch": 2.34,
"learning_rate": 9.641025641025642e-06,
"loss": 0.0472,
"step": 1200
},
{
"epoch": 2.39,
"learning_rate": 9.628205128205129e-06,
"loss": 0.0455,
"step": 1225
},
{
"epoch": 2.44,
"learning_rate": 9.615384615384616e-06,
"loss": 0.044,
"step": 1250
},
{
"epoch": 2.49,
"learning_rate": 9.602564102564104e-06,
"loss": 0.046,
"step": 1275
},
{
"epoch": 2.54,
"learning_rate": 9.589743589743591e-06,
"loss": 0.043,
"step": 1300
},
{
"epoch": 2.59,
"learning_rate": 9.576923076923078e-06,
"loss": 0.0472,
"step": 1325
},
{
"epoch": 2.64,
"learning_rate": 9.564102564102565e-06,
"loss": 0.0447,
"step": 1350
},
{
"epoch": 2.69,
"learning_rate": 9.551282051282053e-06,
"loss": 0.0506,
"step": 1375
},
{
"epoch": 2.73,
"learning_rate": 9.53846153846154e-06,
"loss": 0.0473,
"step": 1400
},
{
"epoch": 2.78,
"learning_rate": 9.525641025641025e-06,
"loss": 0.0464,
"step": 1425
},
{
"epoch": 2.83,
"learning_rate": 9.512820512820514e-06,
"loss": 0.0439,
"step": 1450
},
{
"epoch": 2.88,
"learning_rate": 9.5e-06,
"loss": 0.0514,
"step": 1475
},
{
"epoch": 2.93,
"learning_rate": 9.487179487179487e-06,
"loss": 0.0479,
"step": 1500
},
{
"epoch": 2.98,
"learning_rate": 9.474358974358974e-06,
"loss": 0.0449,
"step": 1525
},
{
"epoch": 3.03,
"learning_rate": 9.461538461538463e-06,
"loss": 0.0339,
"step": 1550
},
{
"epoch": 3.08,
"learning_rate": 9.44871794871795e-06,
"loss": 0.0232,
"step": 1575
},
{
"epoch": 3.12,
"learning_rate": 9.435897435897436e-06,
"loss": 0.0213,
"step": 1600
},
{
"epoch": 3.17,
"learning_rate": 9.423076923076923e-06,
"loss": 0.0219,
"step": 1625
},
{
"epoch": 3.22,
"learning_rate": 9.410256410256412e-06,
"loss": 0.0251,
"step": 1650
},
{
"epoch": 3.27,
"learning_rate": 9.397435897435899e-06,
"loss": 0.0222,
"step": 1675
},
{
"epoch": 3.32,
"learning_rate": 9.384615384615385e-06,
"loss": 0.0232,
"step": 1700
},
{
"epoch": 3.37,
"learning_rate": 9.371794871794872e-06,
"loss": 0.0242,
"step": 1725
},
{
"epoch": 3.42,
"learning_rate": 9.358974358974359e-06,
"loss": 0.0229,
"step": 1750
},
{
"epoch": 3.47,
"learning_rate": 9.346153846153847e-06,
"loss": 0.0232,
"step": 1775
},
{
"epoch": 3.52,
"learning_rate": 9.333333333333334e-06,
"loss": 0.0254,
"step": 1800
},
{
"epoch": 3.56,
"learning_rate": 9.320512820512821e-06,
"loss": 0.0254,
"step": 1825
},
{
"epoch": 3.61,
"learning_rate": 9.307692307692308e-06,
"loss": 0.0231,
"step": 1850
},
{
"epoch": 3.66,
"learning_rate": 9.294871794871796e-06,
"loss": 0.0244,
"step": 1875
},
{
"epoch": 3.71,
"learning_rate": 9.282051282051283e-06,
"loss": 0.0231,
"step": 1900
},
{
"epoch": 3.76,
"learning_rate": 9.26923076923077e-06,
"loss": 0.0235,
"step": 1925
},
{
"epoch": 3.81,
"learning_rate": 9.256410256410257e-06,
"loss": 0.0234,
"step": 1950
},
{
"epoch": 3.86,
"learning_rate": 9.243589743589745e-06,
"loss": 0.0254,
"step": 1975
},
{
"epoch": 3.91,
"learning_rate": 9.230769230769232e-06,
"loss": 0.025,
"step": 2000
},
{
"epoch": 3.91,
"eval_loss": 0.1675654947757721,
"eval_runtime": 3504.3083,
"eval_samples_per_second": 4.674,
"eval_steps_per_second": 0.292,
"eval_wer": 6.315518719634181,
"step": 2000
},
{
"epoch": 3.96,
"learning_rate": 9.217948717948717e-06,
"loss": 0.0296,
"step": 2025
},
{
"epoch": 4.0,
"learning_rate": 9.205128205128206e-06,
"loss": 0.0223,
"step": 2050
},
{
"epoch": 4.05,
"learning_rate": 9.192307692307693e-06,
"loss": 0.0128,
"step": 2075
},
{
"epoch": 4.1,
"learning_rate": 9.17948717948718e-06,
"loss": 0.0126,
"step": 2100
},
{
"epoch": 4.15,
"learning_rate": 9.166666666666666e-06,
"loss": 0.0134,
"step": 2125
},
{
"epoch": 4.2,
"learning_rate": 9.153846153846155e-06,
"loss": 0.0137,
"step": 2150
},
{
"epoch": 4.25,
"learning_rate": 9.141025641025642e-06,
"loss": 0.0142,
"step": 2175
},
{
"epoch": 4.3,
"learning_rate": 9.128205128205129e-06,
"loss": 0.0112,
"step": 2200
},
{
"epoch": 4.35,
"learning_rate": 9.115384615384615e-06,
"loss": 0.0123,
"step": 2225
},
{
"epoch": 4.39,
"learning_rate": 9.102564102564104e-06,
"loss": 0.013,
"step": 2250
},
{
"epoch": 4.44,
"learning_rate": 9.08974358974359e-06,
"loss": 0.0131,
"step": 2275
},
{
"epoch": 4.49,
"learning_rate": 9.076923076923078e-06,
"loss": 0.0132,
"step": 2300
},
{
"epoch": 4.54,
"learning_rate": 9.064102564102564e-06,
"loss": 0.0137,
"step": 2325
},
{
"epoch": 4.59,
"learning_rate": 9.051282051282051e-06,
"loss": 0.0151,
"step": 2350
},
{
"epoch": 4.64,
"learning_rate": 9.03846153846154e-06,
"loss": 0.0149,
"step": 2375
},
{
"epoch": 4.69,
"learning_rate": 9.025641025641027e-06,
"loss": 0.0152,
"step": 2400
},
{
"epoch": 4.74,
"learning_rate": 9.012820512820513e-06,
"loss": 0.0137,
"step": 2425
},
{
"epoch": 4.79,
"learning_rate": 9e-06,
"loss": 0.014,
"step": 2450
},
{
"epoch": 4.83,
"learning_rate": 8.987179487179489e-06,
"loss": 0.0138,
"step": 2475
},
{
"epoch": 4.88,
"learning_rate": 8.974358974358976e-06,
"loss": 0.0146,
"step": 2500
},
{
"epoch": 4.93,
"learning_rate": 8.961538461538462e-06,
"loss": 0.015,
"step": 2525
},
{
"epoch": 4.98,
"learning_rate": 8.94871794871795e-06,
"loss": 0.0159,
"step": 2550
},
{
"epoch": 5.03,
"learning_rate": 8.935897435897438e-06,
"loss": 0.011,
"step": 2575
},
{
"epoch": 5.08,
"learning_rate": 8.923076923076925e-06,
"loss": 0.0093,
"step": 2600
},
{
"epoch": 5.13,
"learning_rate": 8.910256410256411e-06,
"loss": 0.0087,
"step": 2625
},
{
"epoch": 5.18,
"learning_rate": 8.897435897435898e-06,
"loss": 0.0079,
"step": 2650
},
{
"epoch": 5.22,
"learning_rate": 8.884615384615385e-06,
"loss": 0.0084,
"step": 2675
},
{
"epoch": 5.27,
"learning_rate": 8.871794871794872e-06,
"loss": 0.0069,
"step": 2700
},
{
"epoch": 5.32,
"learning_rate": 8.858974358974359e-06,
"loss": 0.008,
"step": 2725
},
{
"epoch": 5.37,
"learning_rate": 8.846153846153847e-06,
"loss": 0.0081,
"step": 2750
},
{
"epoch": 5.42,
"learning_rate": 8.833333333333334e-06,
"loss": 0.0088,
"step": 2775
},
{
"epoch": 5.47,
"learning_rate": 8.820512820512821e-06,
"loss": 0.0096,
"step": 2800
},
{
"epoch": 5.52,
"learning_rate": 8.807692307692308e-06,
"loss": 0.01,
"step": 2825
},
{
"epoch": 5.57,
"learning_rate": 8.794871794871796e-06,
"loss": 0.0087,
"step": 2850
},
{
"epoch": 5.62,
"learning_rate": 8.782051282051283e-06,
"loss": 0.0087,
"step": 2875
},
{
"epoch": 5.66,
"learning_rate": 8.76923076923077e-06,
"loss": 0.0086,
"step": 2900
},
{
"epoch": 5.71,
"learning_rate": 8.756410256410257e-06,
"loss": 0.0081,
"step": 2925
},
{
"epoch": 5.76,
"learning_rate": 8.743589743589743e-06,
"loss": 0.0083,
"step": 2950
},
{
"epoch": 5.81,
"learning_rate": 8.730769230769232e-06,
"loss": 0.0087,
"step": 2975
},
{
"epoch": 5.86,
"learning_rate": 8.717948717948719e-06,
"loss": 0.0105,
"step": 3000
},
{
"epoch": 5.86,
"eval_loss": 0.18706324696540833,
"eval_runtime": 3503.6718,
"eval_samples_per_second": 4.675,
"eval_steps_per_second": 0.292,
"eval_wer": 6.40354386967705,
"step": 3000
},
{
"epoch": 5.91,
"learning_rate": 8.705128205128206e-06,
"loss": 0.0086,
"step": 3025
},
{
"epoch": 5.96,
"learning_rate": 8.692307692307692e-06,
"loss": 0.0102,
"step": 3050
},
{
"epoch": 6.01,
"learning_rate": 8.679487179487181e-06,
"loss": 0.0092,
"step": 3075
},
{
"epoch": 6.05,
"learning_rate": 8.666666666666668e-06,
"loss": 0.0065,
"step": 3100
},
{
"epoch": 6.1,
"learning_rate": 8.653846153846155e-06,
"loss": 0.0055,
"step": 3125
},
{
"epoch": 6.15,
"learning_rate": 8.641025641025641e-06,
"loss": 0.0071,
"step": 3150
},
{
"epoch": 6.2,
"learning_rate": 8.62820512820513e-06,
"loss": 0.007,
"step": 3175
},
{
"epoch": 6.25,
"learning_rate": 8.615384615384617e-06,
"loss": 0.0069,
"step": 3200
},
{
"epoch": 6.3,
"learning_rate": 8.602564102564104e-06,
"loss": 0.006,
"step": 3225
},
{
"epoch": 6.35,
"learning_rate": 8.58974358974359e-06,
"loss": 0.0071,
"step": 3250
},
{
"epoch": 6.4,
"learning_rate": 8.576923076923077e-06,
"loss": 0.0068,
"step": 3275
},
{
"epoch": 6.45,
"learning_rate": 8.564102564102564e-06,
"loss": 0.0077,
"step": 3300
},
{
"epoch": 6.49,
"learning_rate": 8.551282051282051e-06,
"loss": 0.0067,
"step": 3325
},
{
"epoch": 6.54,
"learning_rate": 8.53846153846154e-06,
"loss": 0.0067,
"step": 3350
},
{
"epoch": 6.59,
"learning_rate": 8.525641025641026e-06,
"loss": 0.0075,
"step": 3375
},
{
"epoch": 6.64,
"learning_rate": 8.512820512820513e-06,
"loss": 0.0068,
"step": 3400
},
{
"epoch": 6.69,
"learning_rate": 8.5e-06,
"loss": 0.0078,
"step": 3425
},
{
"epoch": 6.74,
"learning_rate": 8.487179487179488e-06,
"loss": 0.007,
"step": 3450
},
{
"epoch": 6.79,
"learning_rate": 8.474358974358975e-06,
"loss": 0.0082,
"step": 3475
},
{
"epoch": 6.84,
"learning_rate": 8.461538461538462e-06,
"loss": 0.008,
"step": 3500
},
{
"epoch": 6.88,
"learning_rate": 8.448717948717949e-06,
"loss": 0.0078,
"step": 3525
},
{
"epoch": 6.93,
"learning_rate": 8.435897435897436e-06,
"loss": 0.0069,
"step": 3550
},
{
"epoch": 6.98,
"learning_rate": 8.423076923076924e-06,
"loss": 0.0079,
"step": 3575
},
{
"epoch": 7.03,
"learning_rate": 8.410256410256411e-06,
"loss": 0.0055,
"step": 3600
},
{
"epoch": 7.08,
"learning_rate": 8.397435897435898e-06,
"loss": 0.0058,
"step": 3625
},
{
"epoch": 7.13,
"learning_rate": 8.384615384615385e-06,
"loss": 0.005,
"step": 3650
},
{
"epoch": 7.18,
"learning_rate": 8.371794871794873e-06,
"loss": 0.0045,
"step": 3675
},
{
"epoch": 7.23,
"learning_rate": 8.35897435897436e-06,
"loss": 0.0063,
"step": 3700
},
{
"epoch": 7.28,
"learning_rate": 8.346153846153847e-06,
"loss": 0.005,
"step": 3725
},
{
"epoch": 7.32,
"learning_rate": 8.333333333333334e-06,
"loss": 0.0046,
"step": 3750
},
{
"epoch": 7.37,
"learning_rate": 8.320512820512822e-06,
"loss": 0.0047,
"step": 3775
},
{
"epoch": 7.42,
"learning_rate": 8.307692307692309e-06,
"loss": 0.005,
"step": 3800
},
{
"epoch": 7.47,
"learning_rate": 8.294871794871796e-06,
"loss": 0.0044,
"step": 3825
},
{
"epoch": 7.52,
"learning_rate": 8.282051282051283e-06,
"loss": 0.0054,
"step": 3850
},
{
"epoch": 7.57,
"learning_rate": 8.26923076923077e-06,
"loss": 0.0049,
"step": 3875
},
{
"epoch": 7.62,
"learning_rate": 8.256410256410256e-06,
"loss": 0.005,
"step": 3900
},
{
"epoch": 7.67,
"learning_rate": 8.243589743589743e-06,
"loss": 0.0044,
"step": 3925
},
{
"epoch": 7.71,
"learning_rate": 8.230769230769232e-06,
"loss": 0.0062,
"step": 3950
},
{
"epoch": 7.76,
"learning_rate": 8.217948717948719e-06,
"loss": 0.0055,
"step": 3975
},
{
"epoch": 7.81,
"learning_rate": 8.205128205128205e-06,
"loss": 0.0047,
"step": 4000
},
{
"epoch": 7.81,
"eval_loss": 0.1972731351852417,
"eval_runtime": 3494.1518,
"eval_samples_per_second": 4.688,
"eval_steps_per_second": 0.293,
"eval_wer": 6.486996284652758,
"step": 4000
},
{
"epoch": 7.86,
"learning_rate": 8.192307692307692e-06,
"loss": 0.0049,
"step": 4025
},
{
"epoch": 7.91,
"learning_rate": 8.17948717948718e-06,
"loss": 0.006,
"step": 4050
},
{
"epoch": 7.96,
"learning_rate": 8.166666666666668e-06,
"loss": 0.0061,
"step": 4075
},
{
"epoch": 8.01,
"learning_rate": 8.153846153846154e-06,
"loss": 0.0057,
"step": 4100
},
{
"epoch": 8.06,
"learning_rate": 8.141025641025641e-06,
"loss": 0.0043,
"step": 4125
},
{
"epoch": 8.11,
"learning_rate": 8.12820512820513e-06,
"loss": 0.0044,
"step": 4150
},
{
"epoch": 8.15,
"learning_rate": 8.115384615384617e-06,
"loss": 0.0033,
"step": 4175
},
{
"epoch": 8.2,
"learning_rate": 8.102564102564103e-06,
"loss": 0.0033,
"step": 4200
},
{
"epoch": 8.25,
"learning_rate": 8.08974358974359e-06,
"loss": 0.0043,
"step": 4225
},
{
"epoch": 8.3,
"learning_rate": 8.076923076923077e-06,
"loss": 0.0036,
"step": 4250
},
{
"epoch": 8.35,
"learning_rate": 8.064102564102566e-06,
"loss": 0.0036,
"step": 4275
},
{
"epoch": 8.4,
"learning_rate": 8.051282051282052e-06,
"loss": 0.0042,
"step": 4300
},
{
"epoch": 8.45,
"learning_rate": 8.03846153846154e-06,
"loss": 0.0043,
"step": 4325
},
{
"epoch": 8.5,
"learning_rate": 8.025641025641026e-06,
"loss": 0.003,
"step": 4350
},
{
"epoch": 8.54,
"learning_rate": 8.012820512820515e-06,
"loss": 0.0034,
"step": 4375
},
{
"epoch": 8.59,
"learning_rate": 8.000000000000001e-06,
"loss": 0.004,
"step": 4400
},
{
"epoch": 8.64,
"learning_rate": 7.987179487179488e-06,
"loss": 0.004,
"step": 4425
},
{
"epoch": 8.69,
"learning_rate": 7.974358974358975e-06,
"loss": 0.0036,
"step": 4450
},
{
"epoch": 8.74,
"learning_rate": 7.961538461538462e-06,
"loss": 0.0042,
"step": 4475
},
{
"epoch": 8.79,
"learning_rate": 7.948717948717949e-06,
"loss": 0.0048,
"step": 4500
},
{
"epoch": 8.84,
"learning_rate": 7.935897435897435e-06,
"loss": 0.0055,
"step": 4525
},
{
"epoch": 8.89,
"learning_rate": 7.923076923076924e-06,
"loss": 0.0059,
"step": 4550
},
{
"epoch": 8.94,
"learning_rate": 7.91025641025641e-06,
"loss": 0.0063,
"step": 4575
},
{
"epoch": 8.98,
"learning_rate": 7.897435897435898e-06,
"loss": 0.0043,
"step": 4600
},
{
"epoch": 9.03,
"learning_rate": 7.884615384615384e-06,
"loss": 0.0041,
"step": 4625
},
{
"epoch": 9.08,
"learning_rate": 7.871794871794873e-06,
"loss": 0.0041,
"step": 4650
},
{
"epoch": 9.13,
"learning_rate": 7.85897435897436e-06,
"loss": 0.0031,
"step": 4675
},
{
"epoch": 9.18,
"learning_rate": 7.846153846153847e-06,
"loss": 0.0033,
"step": 4700
},
{
"epoch": 9.23,
"learning_rate": 7.833333333333333e-06,
"loss": 0.0036,
"step": 4725
},
{
"epoch": 9.28,
"learning_rate": 7.820512820512822e-06,
"loss": 0.0045,
"step": 4750
},
{
"epoch": 9.33,
"learning_rate": 7.807692307692309e-06,
"loss": 0.0034,
"step": 4775
},
{
"epoch": 9.38,
"learning_rate": 7.794871794871796e-06,
"loss": 0.0047,
"step": 4800
},
{
"epoch": 9.42,
"learning_rate": 7.782051282051282e-06,
"loss": 0.0036,
"step": 4825
},
{
"epoch": 9.47,
"learning_rate": 7.76923076923077e-06,
"loss": 0.0028,
"step": 4850
},
{
"epoch": 9.52,
"learning_rate": 7.756410256410258e-06,
"loss": 0.0033,
"step": 4875
},
{
"epoch": 9.57,
"learning_rate": 7.743589743589745e-06,
"loss": 0.0045,
"step": 4900
},
{
"epoch": 9.62,
"learning_rate": 7.730769230769231e-06,
"loss": 0.0035,
"step": 4925
},
{
"epoch": 9.67,
"learning_rate": 7.717948717948718e-06,
"loss": 0.0049,
"step": 4950
},
{
"epoch": 9.72,
"learning_rate": 7.705128205128207e-06,
"loss": 0.0054,
"step": 4975
},
{
"epoch": 9.77,
"learning_rate": 7.692307692307694e-06,
"loss": 0.0061,
"step": 5000
},
{
"epoch": 9.77,
"eval_loss": 0.20856818556785583,
"eval_runtime": 3499.3081,
"eval_samples_per_second": 4.681,
"eval_steps_per_second": 0.293,
"eval_wer": 6.4835667333523865,
"step": 5000
},
{
"epoch": 9.81,
"learning_rate": 7.67948717948718e-06,
"loss": 0.0054,
"step": 5025
},
{
"epoch": 9.86,
"learning_rate": 7.666666666666667e-06,
"loss": 0.0052,
"step": 5050
},
{
"epoch": 9.91,
"learning_rate": 7.653846153846154e-06,
"loss": 0.0047,
"step": 5075
},
{
"epoch": 9.96,
"learning_rate": 7.641025641025641e-06,
"loss": 0.0039,
"step": 5100
},
{
"epoch": 10.01,
"learning_rate": 7.6282051282051286e-06,
"loss": 0.0041,
"step": 5125
},
{
"epoch": 10.06,
"learning_rate": 7.615384615384615e-06,
"loss": 0.0031,
"step": 5150
},
{
"epoch": 10.11,
"learning_rate": 7.602564102564103e-06,
"loss": 0.004,
"step": 5175
},
{
"epoch": 10.16,
"learning_rate": 7.58974358974359e-06,
"loss": 0.004,
"step": 5200
},
{
"epoch": 10.21,
"learning_rate": 7.5769230769230775e-06,
"loss": 0.0025,
"step": 5225
},
{
"epoch": 10.25,
"learning_rate": 7.564102564102564e-06,
"loss": 0.0022,
"step": 5250
},
{
"epoch": 10.3,
"learning_rate": 7.551282051282052e-06,
"loss": 0.0027,
"step": 5275
},
{
"epoch": 10.35,
"learning_rate": 7.538461538461539e-06,
"loss": 0.0028,
"step": 5300
},
{
"epoch": 10.4,
"learning_rate": 7.5256410256410265e-06,
"loss": 0.003,
"step": 5325
},
{
"epoch": 10.45,
"learning_rate": 7.512820512820513e-06,
"loss": 0.0036,
"step": 5350
},
{
"epoch": 10.5,
"learning_rate": 7.500000000000001e-06,
"loss": 0.0029,
"step": 5375
},
{
"epoch": 10.55,
"learning_rate": 7.487179487179488e-06,
"loss": 0.0019,
"step": 5400
},
{
"epoch": 10.6,
"learning_rate": 7.474358974358975e-06,
"loss": 0.002,
"step": 5425
},
{
"epoch": 10.64,
"learning_rate": 7.461538461538462e-06,
"loss": 0.0025,
"step": 5450
},
{
"epoch": 10.69,
"learning_rate": 7.448717948717949e-06,
"loss": 0.0028,
"step": 5475
},
{
"epoch": 10.74,
"learning_rate": 7.435897435897437e-06,
"loss": 0.0037,
"step": 5500
},
{
"epoch": 10.79,
"learning_rate": 7.423076923076924e-06,
"loss": 0.0029,
"step": 5525
},
{
"epoch": 10.84,
"learning_rate": 7.410256410256411e-06,
"loss": 0.0035,
"step": 5550
},
{
"epoch": 10.89,
"learning_rate": 7.397435897435898e-06,
"loss": 0.0034,
"step": 5575
},
{
"epoch": 10.94,
"learning_rate": 7.384615384615386e-06,
"loss": 0.0033,
"step": 5600
},
{
"epoch": 10.99,
"learning_rate": 7.371794871794873e-06,
"loss": 0.004,
"step": 5625
},
{
"epoch": 11.04,
"learning_rate": 7.35897435897436e-06,
"loss": 0.0027,
"step": 5650
},
{
"epoch": 11.08,
"learning_rate": 7.346153846153847e-06,
"loss": 0.0028,
"step": 5675
},
{
"epoch": 11.13,
"learning_rate": 7.333333333333333e-06,
"loss": 0.003,
"step": 5700
},
{
"epoch": 11.18,
"learning_rate": 7.320512820512821e-06,
"loss": 0.0022,
"step": 5725
},
{
"epoch": 11.23,
"learning_rate": 7.307692307692308e-06,
"loss": 0.0029,
"step": 5750
},
{
"epoch": 11.28,
"learning_rate": 7.294871794871795e-06,
"loss": 0.0028,
"step": 5775
},
{
"epoch": 11.33,
"learning_rate": 7.282051282051282e-06,
"loss": 0.0026,
"step": 5800
},
{
"epoch": 11.38,
"learning_rate": 7.26923076923077e-06,
"loss": 0.0021,
"step": 5825
},
{
"epoch": 11.43,
"learning_rate": 7.256410256410257e-06,
"loss": 0.002,
"step": 5850
},
{
"epoch": 11.47,
"learning_rate": 7.243589743589744e-06,
"loss": 0.0023,
"step": 5875
},
{
"epoch": 11.52,
"learning_rate": 7.230769230769231e-06,
"loss": 0.0026,
"step": 5900
},
{
"epoch": 11.57,
"learning_rate": 7.217948717948719e-06,
"loss": 0.0029,
"step": 5925
},
{
"epoch": 11.62,
"learning_rate": 7.205128205128206e-06,
"loss": 0.0031,
"step": 5950
},
{
"epoch": 11.67,
"learning_rate": 7.192307692307693e-06,
"loss": 0.0036,
"step": 5975
},
{
"epoch": 11.72,
"learning_rate": 7.17948717948718e-06,
"loss": 0.0034,
"step": 6000
},
{
"epoch": 11.72,
"eval_loss": 0.21721301972866058,
"eval_runtime": 3505.9605,
"eval_samples_per_second": 4.672,
"eval_steps_per_second": 0.292,
"eval_wer": 6.6441840525864535,
"step": 6000
},
{
"epoch": 11.77,
"learning_rate": 7.166666666666667e-06,
"loss": 0.0045,
"step": 6025
},
{
"epoch": 11.82,
"learning_rate": 7.153846153846155e-06,
"loss": 0.004,
"step": 6050
},
{
"epoch": 11.87,
"learning_rate": 7.1410256410256414e-06,
"loss": 0.0037,
"step": 6075
},
{
"epoch": 11.91,
"learning_rate": 7.128205128205129e-06,
"loss": 0.0032,
"step": 6100
},
{
"epoch": 11.96,
"learning_rate": 7.115384615384616e-06,
"loss": 0.0041,
"step": 6125
},
{
"epoch": 12.01,
"learning_rate": 7.102564102564104e-06,
"loss": 0.0036,
"step": 6150
},
{
"epoch": 12.06,
"learning_rate": 7.0897435897435904e-06,
"loss": 0.0032,
"step": 6175
},
{
"epoch": 12.11,
"learning_rate": 7.076923076923078e-06,
"loss": 0.0024,
"step": 6200
},
{
"epoch": 12.16,
"learning_rate": 7.064102564102565e-06,
"loss": 0.0024,
"step": 6225
},
{
"epoch": 12.21,
"learning_rate": 7.051282051282053e-06,
"loss": 0.003,
"step": 6250
},
{
"epoch": 12.26,
"learning_rate": 7.038461538461539e-06,
"loss": 0.0033,
"step": 6275
},
{
"epoch": 12.3,
"learning_rate": 7.025641025641025e-06,
"loss": 0.0018,
"step": 6300
},
{
"epoch": 12.35,
"learning_rate": 7.012820512820513e-06,
"loss": 0.0023,
"step": 6325
},
{
"epoch": 12.4,
"learning_rate": 7e-06,
"loss": 0.002,
"step": 6350
},
{
"epoch": 12.45,
"learning_rate": 6.9871794871794876e-06,
"loss": 0.0025,
"step": 6375
},
{
"epoch": 12.5,
"learning_rate": 6.974358974358974e-06,
"loss": 0.002,
"step": 6400
},
{
"epoch": 12.55,
"learning_rate": 6.961538461538462e-06,
"loss": 0.0023,
"step": 6425
},
{
"epoch": 12.6,
"learning_rate": 6.948717948717949e-06,
"loss": 0.0029,
"step": 6450
},
{
"epoch": 12.65,
"learning_rate": 6.9358974358974366e-06,
"loss": 0.0031,
"step": 6475
},
{
"epoch": 12.7,
"learning_rate": 6.923076923076923e-06,
"loss": 0.0035,
"step": 6500
},
{
"epoch": 12.74,
"learning_rate": 6.910256410256411e-06,
"loss": 0.0036,
"step": 6525
},
{
"epoch": 12.79,
"learning_rate": 6.897435897435898e-06,
"loss": 0.0036,
"step": 6550
},
{
"epoch": 12.84,
"learning_rate": 6.8846153846153855e-06,
"loss": 0.0034,
"step": 6575
},
{
"epoch": 12.89,
"learning_rate": 6.871794871794872e-06,
"loss": 0.0037,
"step": 6600
},
{
"epoch": 12.94,
"learning_rate": 6.858974358974359e-06,
"loss": 0.0029,
"step": 6625
},
{
"epoch": 12.99,
"learning_rate": 6.846153846153847e-06,
"loss": 0.003,
"step": 6650
},
{
"epoch": 13.04,
"learning_rate": 6.833333333333334e-06,
"loss": 0.0032,
"step": 6675
},
{
"epoch": 13.09,
"learning_rate": 6.820512820512821e-06,
"loss": 0.0024,
"step": 6700
},
{
"epoch": 13.13,
"learning_rate": 6.807692307692308e-06,
"loss": 0.0026,
"step": 6725
},
{
"epoch": 13.18,
"learning_rate": 6.794871794871796e-06,
"loss": 0.0019,
"step": 6750
},
{
"epoch": 13.23,
"learning_rate": 6.782051282051283e-06,
"loss": 0.0014,
"step": 6775
},
{
"epoch": 13.28,
"learning_rate": 6.76923076923077e-06,
"loss": 0.0029,
"step": 6800
},
{
"epoch": 13.33,
"learning_rate": 6.756410256410257e-06,
"loss": 0.002,
"step": 6825
},
{
"epoch": 13.38,
"learning_rate": 6.743589743589745e-06,
"loss": 0.0029,
"step": 6850
},
{
"epoch": 13.43,
"learning_rate": 6.730769230769232e-06,
"loss": 0.0029,
"step": 6875
},
{
"epoch": 13.48,
"learning_rate": 6.717948717948718e-06,
"loss": 0.0025,
"step": 6900
},
{
"epoch": 13.53,
"learning_rate": 6.705641025641027e-06,
"loss": 0.0037,
"step": 6925
},
{
"epoch": 13.57,
"learning_rate": 6.692820512820514e-06,
"loss": 0.003,
"step": 6950
},
{
"epoch": 13.62,
"learning_rate": 6.680000000000001e-06,
"loss": 0.0032,
"step": 6975
},
{
"epoch": 13.67,
"learning_rate": 6.667179487179488e-06,
"loss": 0.0036,
"step": 7000
},
{
"epoch": 13.67,
"eval_loss": 0.22047244012355804,
"eval_runtime": 3497.6451,
"eval_samples_per_second": 4.683,
"eval_steps_per_second": 0.293,
"eval_wer": 6.404115461560446,
"step": 7000
},
{
"epoch": 13.72,
"learning_rate": 6.654358974358976e-06,
"loss": 0.0032,
"step": 7025
},
{
"epoch": 13.77,
"learning_rate": 6.641538461538463e-06,
"loss": 0.0033,
"step": 7050
},
{
"epoch": 13.82,
"learning_rate": 6.6287179487179486e-06,
"loss": 0.0023,
"step": 7075
},
{
"epoch": 13.87,
"learning_rate": 6.615897435897436e-06,
"loss": 0.0024,
"step": 7100
},
{
"epoch": 13.92,
"learning_rate": 6.603076923076923e-06,
"loss": 0.0034,
"step": 7125
},
{
"epoch": 13.96,
"learning_rate": 6.590256410256411e-06,
"loss": 0.0033,
"step": 7150
},
{
"epoch": 14.01,
"learning_rate": 6.5774358974358976e-06,
"loss": 0.0025,
"step": 7175
},
{
"epoch": 14.06,
"learning_rate": 6.564615384615385e-06,
"loss": 0.0016,
"step": 7200
},
{
"epoch": 14.11,
"learning_rate": 6.551794871794872e-06,
"loss": 0.0019,
"step": 7225
},
{
"epoch": 14.16,
"learning_rate": 6.538974358974359e-06,
"loss": 0.0018,
"step": 7250
},
{
"epoch": 14.21,
"learning_rate": 6.5261538461538465e-06,
"loss": 0.0016,
"step": 7275
},
{
"epoch": 14.26,
"learning_rate": 6.513333333333333e-06,
"loss": 0.0011,
"step": 7300
},
{
"epoch": 14.31,
"learning_rate": 6.500512820512821e-06,
"loss": 0.001,
"step": 7325
},
{
"epoch": 14.36,
"learning_rate": 6.487692307692308e-06,
"loss": 0.0015,
"step": 7350
},
{
"epoch": 14.4,
"learning_rate": 6.4748717948717955e-06,
"loss": 0.0016,
"step": 7375
},
{
"epoch": 14.45,
"learning_rate": 6.462051282051282e-06,
"loss": 0.0016,
"step": 7400
},
{
"epoch": 14.5,
"learning_rate": 6.44923076923077e-06,
"loss": 0.0016,
"step": 7425
},
{
"epoch": 14.55,
"learning_rate": 6.436410256410257e-06,
"loss": 0.0022,
"step": 7450
},
{
"epoch": 14.6,
"learning_rate": 6.4235897435897445e-06,
"loss": 0.0016,
"step": 7475
},
{
"epoch": 14.65,
"learning_rate": 6.410769230769231e-06,
"loss": 0.0016,
"step": 7500
},
{
"epoch": 14.7,
"learning_rate": 6.397948717948719e-06,
"loss": 0.0013,
"step": 7525
},
{
"epoch": 14.75,
"learning_rate": 6.385128205128206e-06,
"loss": 0.0012,
"step": 7550
},
{
"epoch": 14.79,
"learning_rate": 6.3723076923076935e-06,
"loss": 0.0014,
"step": 7575
},
{
"epoch": 14.84,
"learning_rate": 6.35948717948718e-06,
"loss": 0.0014,
"step": 7600
},
{
"epoch": 14.89,
"learning_rate": 6.346666666666668e-06,
"loss": 0.0024,
"step": 7625
},
{
"epoch": 14.94,
"learning_rate": 6.333846153846155e-06,
"loss": 0.0026,
"step": 7650
},
{
"epoch": 14.99,
"learning_rate": 6.3210256410256425e-06,
"loss": 0.003,
"step": 7675
},
{
"epoch": 15.04,
"learning_rate": 6.3082051282051285e-06,
"loss": 0.0025,
"step": 7700
},
{
"epoch": 15.09,
"learning_rate": 6.295384615384615e-06,
"loss": 0.0028,
"step": 7725
},
{
"epoch": 15.14,
"learning_rate": 6.282564102564103e-06,
"loss": 0.0018,
"step": 7750
},
{
"epoch": 15.19,
"learning_rate": 6.26974358974359e-06,
"loss": 0.0023,
"step": 7775
},
{
"epoch": 15.23,
"learning_rate": 6.2569230769230775e-06,
"loss": 0.0017,
"step": 7800
},
{
"epoch": 15.28,
"learning_rate": 6.244102564102564e-06,
"loss": 0.0026,
"step": 7825
},
{
"epoch": 15.33,
"learning_rate": 6.231282051282051e-06,
"loss": 0.0028,
"step": 7850
},
{
"epoch": 15.38,
"learning_rate": 6.218461538461539e-06,
"loss": 0.002,
"step": 7875
},
{
"epoch": 15.43,
"learning_rate": 6.205641025641026e-06,
"loss": 0.0022,
"step": 7900
},
{
"epoch": 15.48,
"learning_rate": 6.192820512820513e-06,
"loss": 0.002,
"step": 7925
},
{
"epoch": 15.53,
"learning_rate": 6.18e-06,
"loss": 0.0017,
"step": 7950
},
{
"epoch": 15.58,
"learning_rate": 6.167179487179488e-06,
"loss": 0.0024,
"step": 7975
},
{
"epoch": 15.62,
"learning_rate": 6.154358974358975e-06,
"loss": 0.002,
"step": 8000
},
{
"epoch": 15.62,
"eval_loss": 0.22138619422912598,
"eval_runtime": 3512.6371,
"eval_samples_per_second": 4.663,
"eval_steps_per_second": 0.292,
"eval_wer": 6.43498142326379,
"step": 8000
},
{
"epoch": 15.67,
"learning_rate": 6.141538461538462e-06,
"loss": 0.0015,
"step": 8025
},
{
"epoch": 15.72,
"learning_rate": 6.128717948717949e-06,
"loss": 0.0017,
"step": 8050
},
{
"epoch": 15.77,
"learning_rate": 6.115897435897437e-06,
"loss": 0.0028,
"step": 8075
},
{
"epoch": 15.82,
"learning_rate": 6.103076923076924e-06,
"loss": 0.0022,
"step": 8100
},
{
"epoch": 15.87,
"learning_rate": 6.090256410256411e-06,
"loss": 0.0028,
"step": 8125
},
{
"epoch": 15.92,
"learning_rate": 6.077435897435898e-06,
"loss": 0.0032,
"step": 8150
},
{
"epoch": 15.97,
"learning_rate": 6.064615384615386e-06,
"loss": 0.0024,
"step": 8175
},
{
"epoch": 16.02,
"learning_rate": 6.051794871794873e-06,
"loss": 0.002,
"step": 8200
},
{
"epoch": 16.06,
"learning_rate": 6.03897435897436e-06,
"loss": 0.0013,
"step": 8225
},
{
"epoch": 16.11,
"learning_rate": 6.026153846153847e-06,
"loss": 0.0016,
"step": 8250
},
{
"epoch": 16.16,
"learning_rate": 6.013333333333335e-06,
"loss": 0.0021,
"step": 8275
},
{
"epoch": 16.21,
"learning_rate": 6.000512820512821e-06,
"loss": 0.0017,
"step": 8300
},
{
"epoch": 16.26,
"learning_rate": 5.9876923076923076e-06,
"loss": 0.0021,
"step": 8325
},
{
"epoch": 16.31,
"learning_rate": 5.974871794871795e-06,
"loss": 0.0018,
"step": 8350
},
{
"epoch": 16.36,
"learning_rate": 5.962051282051282e-06,
"loss": 0.0017,
"step": 8375
},
{
"epoch": 16.41,
"learning_rate": 5.94923076923077e-06,
"loss": 0.0023,
"step": 8400
},
{
"epoch": 16.46,
"learning_rate": 5.9364102564102566e-06,
"loss": 0.0014,
"step": 8425
},
{
"epoch": 16.5,
"learning_rate": 5.923589743589744e-06,
"loss": 0.0022,
"step": 8450
},
{
"epoch": 16.55,
"learning_rate": 5.910769230769231e-06,
"loss": 0.0018,
"step": 8475
},
{
"epoch": 16.6,
"learning_rate": 5.897948717948718e-06,
"loss": 0.0019,
"step": 8500
},
{
"epoch": 16.65,
"learning_rate": 5.8851282051282056e-06,
"loss": 0.0035,
"step": 8525
},
{
"epoch": 16.7,
"learning_rate": 5.872307692307692e-06,
"loss": 0.0029,
"step": 8550
},
{
"epoch": 16.75,
"learning_rate": 5.85948717948718e-06,
"loss": 0.002,
"step": 8575
},
{
"epoch": 16.8,
"learning_rate": 5.846666666666667e-06,
"loss": 0.0026,
"step": 8600
},
{
"epoch": 16.85,
"learning_rate": 5.8338461538461545e-06,
"loss": 0.0024,
"step": 8625
},
{
"epoch": 16.89,
"learning_rate": 5.821025641025641e-06,
"loss": 0.0029,
"step": 8650
},
{
"epoch": 16.94,
"learning_rate": 5.808205128205129e-06,
"loss": 0.0026,
"step": 8675
},
{
"epoch": 16.99,
"learning_rate": 5.795384615384616e-06,
"loss": 0.0035,
"step": 8700
},
{
"epoch": 17.04,
"learning_rate": 5.7825641025641035e-06,
"loss": 0.0017,
"step": 8725
},
{
"epoch": 17.09,
"learning_rate": 5.76974358974359e-06,
"loss": 0.002,
"step": 8750
},
{
"epoch": 17.14,
"learning_rate": 5.756923076923078e-06,
"loss": 0.0019,
"step": 8775
},
{
"epoch": 17.19,
"learning_rate": 5.744102564102565e-06,
"loss": 0.0016,
"step": 8800
},
{
"epoch": 17.24,
"learning_rate": 5.7312820512820525e-06,
"loss": 0.001,
"step": 8825
},
{
"epoch": 17.29,
"learning_rate": 5.718461538461539e-06,
"loss": 0.0012,
"step": 8850
},
{
"epoch": 17.33,
"learning_rate": 5.705641025641027e-06,
"loss": 0.0018,
"step": 8875
},
{
"epoch": 17.38,
"learning_rate": 5.692820512820513e-06,
"loss": 0.0019,
"step": 8900
},
{
"epoch": 17.43,
"learning_rate": 5.68e-06,
"loss": 0.001,
"step": 8925
},
{
"epoch": 17.48,
"learning_rate": 5.6671794871794875e-06,
"loss": 0.0013,
"step": 8950
},
{
"epoch": 17.53,
"learning_rate": 5.654358974358974e-06,
"loss": 0.0017,
"step": 8975
},
{
"epoch": 17.58,
"learning_rate": 5.641538461538462e-06,
"loss": 0.0011,
"step": 9000
},
{
"epoch": 17.58,
"eval_loss": 0.23390185832977295,
"eval_runtime": 3504.1726,
"eval_samples_per_second": 4.674,
"eval_steps_per_second": 0.292,
"eval_wer": 6.194341240354387,
"step": 9000
},
{
"epoch": 17.63,
"learning_rate": 5.628717948717949e-06,
"loss": 0.0009,
"step": 9025
},
{
"epoch": 17.68,
"learning_rate": 5.6158974358974365e-06,
"loss": 0.0019,
"step": 9050
},
{
"epoch": 17.72,
"learning_rate": 5.603589743589744e-06,
"loss": 0.0023,
"step": 9075
},
{
"epoch": 17.77,
"learning_rate": 5.590769230769231e-06,
"loss": 0.0021,
"step": 9100
},
{
"epoch": 17.82,
"learning_rate": 5.5779487179487176e-06,
"loss": 0.0023,
"step": 9125
},
{
"epoch": 17.87,
"learning_rate": 5.565128205128205e-06,
"loss": 0.0019,
"step": 9150
},
{
"epoch": 17.92,
"learning_rate": 5.552307692307692e-06,
"loss": 0.0019,
"step": 9175
},
{
"epoch": 17.97,
"learning_rate": 5.53948717948718e-06,
"loss": 0.0014,
"step": 9200
},
{
"epoch": 18.02,
"learning_rate": 5.5266666666666666e-06,
"loss": 0.0018,
"step": 9225
},
{
"epoch": 18.07,
"learning_rate": 5.513846153846154e-06,
"loss": 0.0019,
"step": 9250
},
{
"epoch": 18.12,
"learning_rate": 5.501025641025641e-06,
"loss": 0.0009,
"step": 9275
},
{
"epoch": 18.16,
"learning_rate": 5.488205128205129e-06,
"loss": 0.001,
"step": 9300
},
{
"epoch": 18.21,
"learning_rate": 5.4753846153846155e-06,
"loss": 0.0006,
"step": 9325
},
{
"epoch": 18.26,
"learning_rate": 5.462564102564103e-06,
"loss": 0.0008,
"step": 9350
},
{
"epoch": 18.31,
"learning_rate": 5.44974358974359e-06,
"loss": 0.0008,
"step": 9375
},
{
"epoch": 18.36,
"learning_rate": 5.436923076923078e-06,
"loss": 0.0018,
"step": 9400
},
{
"epoch": 18.41,
"learning_rate": 5.4241025641025645e-06,
"loss": 0.0017,
"step": 9425
},
{
"epoch": 18.46,
"learning_rate": 5.411282051282052e-06,
"loss": 0.0012,
"step": 9450
},
{
"epoch": 18.51,
"learning_rate": 5.398461538461539e-06,
"loss": 0.001,
"step": 9475
},
{
"epoch": 18.55,
"learning_rate": 5.385641025641027e-06,
"loss": 0.0016,
"step": 9500
},
{
"epoch": 18.6,
"learning_rate": 5.3728205128205135e-06,
"loss": 0.0016,
"step": 9525
},
{
"epoch": 18.65,
"learning_rate": 5.36e-06,
"loss": 0.0012,
"step": 9550
},
{
"epoch": 18.7,
"learning_rate": 5.347179487179488e-06,
"loss": 0.0011,
"step": 9575
},
{
"epoch": 18.75,
"learning_rate": 5.334358974358975e-06,
"loss": 0.0026,
"step": 9600
},
{
"epoch": 18.8,
"learning_rate": 5.3215384615384625e-06,
"loss": 0.0015,
"step": 9625
},
{
"epoch": 18.85,
"learning_rate": 5.308717948717949e-06,
"loss": 0.0017,
"step": 9650
},
{
"epoch": 18.9,
"learning_rate": 5.295897435897437e-06,
"loss": 0.0013,
"step": 9675
},
{
"epoch": 18.95,
"learning_rate": 5.283076923076923e-06,
"loss": 0.0013,
"step": 9700
},
{
"epoch": 18.99,
"learning_rate": 5.27025641025641e-06,
"loss": 0.0015,
"step": 9725
},
{
"epoch": 19.04,
"learning_rate": 5.2574358974358975e-06,
"loss": 0.0007,
"step": 9750
},
{
"epoch": 19.09,
"learning_rate": 5.244615384615384e-06,
"loss": 0.0007,
"step": 9775
},
{
"epoch": 19.14,
"learning_rate": 5.231794871794872e-06,
"loss": 0.0006,
"step": 9800
},
{
"epoch": 19.19,
"learning_rate": 5.218974358974359e-06,
"loss": 0.0008,
"step": 9825
},
{
"epoch": 19.24,
"learning_rate": 5.2061538461538465e-06,
"loss": 0.0006,
"step": 9850
},
{
"epoch": 19.29,
"learning_rate": 5.193333333333333e-06,
"loss": 0.0011,
"step": 9875
},
{
"epoch": 19.34,
"learning_rate": 5.180512820512821e-06,
"loss": 0.0008,
"step": 9900
},
{
"epoch": 19.38,
"learning_rate": 5.167692307692308e-06,
"loss": 0.0006,
"step": 9925
},
{
"epoch": 19.43,
"learning_rate": 5.1548717948717955e-06,
"loss": 0.0013,
"step": 9950
},
{
"epoch": 19.48,
"learning_rate": 5.142051282051282e-06,
"loss": 0.0009,
"step": 9975
},
{
"epoch": 19.53,
"learning_rate": 5.12923076923077e-06,
"loss": 0.0009,
"step": 10000
},
{
"epoch": 19.53,
"eval_loss": 0.23882965743541718,
"eval_runtime": 3519.6363,
"eval_samples_per_second": 4.654,
"eval_steps_per_second": 0.291,
"eval_wer": 6.2920834524149765,
"step": 10000
},
{
"epoch": 19.58,
"learning_rate": 5.116410256410257e-06,
"loss": 0.0014,
"step": 10025
},
{
"epoch": 19.63,
"learning_rate": 5.104102564102564e-06,
"loss": 0.0014,
"step": 10050
},
{
"epoch": 19.68,
"learning_rate": 5.091282051282052e-06,
"loss": 0.0007,
"step": 10075
},
{
"epoch": 19.73,
"learning_rate": 5.078461538461539e-06,
"loss": 0.0011,
"step": 10100
},
{
"epoch": 19.78,
"learning_rate": 5.065641025641026e-06,
"loss": 0.001,
"step": 10125
},
{
"epoch": 19.82,
"learning_rate": 5.052820512820513e-06,
"loss": 0.001,
"step": 10150
},
{
"epoch": 19.87,
"learning_rate": 5.04e-06,
"loss": 0.0008,
"step": 10175
},
{
"epoch": 19.92,
"learning_rate": 5.027179487179488e-06,
"loss": 0.0013,
"step": 10200
},
{
"epoch": 19.97,
"learning_rate": 5.0143589743589745e-06,
"loss": 0.0009,
"step": 10225
},
{
"epoch": 20.02,
"learning_rate": 5.001538461538462e-06,
"loss": 0.001,
"step": 10250
},
{
"epoch": 20.07,
"learning_rate": 4.988717948717949e-06,
"loss": 0.0017,
"step": 10275
},
{
"epoch": 20.12,
"learning_rate": 4.975897435897437e-06,
"loss": 0.0012,
"step": 10300
},
{
"epoch": 20.17,
"learning_rate": 4.9630769230769235e-06,
"loss": 0.0005,
"step": 10325
},
{
"epoch": 20.21,
"learning_rate": 4.95025641025641e-06,
"loss": 0.0011,
"step": 10350
},
{
"epoch": 20.26,
"learning_rate": 4.937435897435898e-06,
"loss": 0.0006,
"step": 10375
},
{
"epoch": 20.31,
"learning_rate": 4.924615384615385e-06,
"loss": 0.0017,
"step": 10400
},
{
"epoch": 20.36,
"learning_rate": 4.9117948717948725e-06,
"loss": 0.0008,
"step": 10425
},
{
"epoch": 20.41,
"learning_rate": 4.898974358974359e-06,
"loss": 0.0021,
"step": 10450
},
{
"epoch": 20.46,
"learning_rate": 4.886153846153847e-06,
"loss": 0.0019,
"step": 10475
},
{
"epoch": 20.51,
"learning_rate": 4.873333333333334e-06,
"loss": 0.0017,
"step": 10500
},
{
"epoch": 20.56,
"learning_rate": 4.860512820512821e-06,
"loss": 0.0024,
"step": 10525
},
{
"epoch": 20.61,
"learning_rate": 4.847692307692308e-06,
"loss": 0.0007,
"step": 10550
},
{
"epoch": 20.65,
"learning_rate": 4.834871794871795e-06,
"loss": 0.0008,
"step": 10575
},
{
"epoch": 20.7,
"learning_rate": 4.822051282051283e-06,
"loss": 0.001,
"step": 10600
},
{
"epoch": 20.75,
"learning_rate": 4.80923076923077e-06,
"loss": 0.0006,
"step": 10625
},
{
"epoch": 20.8,
"learning_rate": 4.7964102564102565e-06,
"loss": 0.0005,
"step": 10650
},
{
"epoch": 20.85,
"learning_rate": 4.783589743589744e-06,
"loss": 0.0009,
"step": 10675
},
{
"epoch": 20.9,
"learning_rate": 4.770769230769231e-06,
"loss": 0.0009,
"step": 10700
},
{
"epoch": 20.95,
"learning_rate": 4.757948717948719e-06,
"loss": 0.0007,
"step": 10725
},
{
"epoch": 21.0,
"learning_rate": 4.7451282051282055e-06,
"loss": 0.0008,
"step": 10750
},
{
"epoch": 21.04,
"learning_rate": 4.732307692307693e-06,
"loss": 0.0009,
"step": 10775
},
{
"epoch": 21.09,
"learning_rate": 4.71948717948718e-06,
"loss": 0.001,
"step": 10800
},
{
"epoch": 21.14,
"learning_rate": 4.706666666666667e-06,
"loss": 0.0004,
"step": 10825
},
{
"epoch": 21.19,
"learning_rate": 4.6938461538461544e-06,
"loss": 0.0014,
"step": 10850
},
{
"epoch": 21.24,
"learning_rate": 4.681025641025641e-06,
"loss": 0.0012,
"step": 10875
},
{
"epoch": 21.29,
"learning_rate": 4.668205128205129e-06,
"loss": 0.0009,
"step": 10900
},
{
"epoch": 21.34,
"learning_rate": 4.655384615384616e-06,
"loss": 0.0012,
"step": 10925
},
{
"epoch": 21.39,
"learning_rate": 4.642564102564103e-06,
"loss": 0.0011,
"step": 10950
},
{
"epoch": 21.44,
"learning_rate": 4.62974358974359e-06,
"loss": 0.0014,
"step": 10975
},
{
"epoch": 21.48,
"learning_rate": 4.616923076923077e-06,
"loss": 0.0011,
"step": 11000
},
{
"epoch": 21.48,
"eval_loss": 0.23270219564437866,
"eval_runtime": 3504.5983,
"eval_samples_per_second": 4.674,
"eval_steps_per_second": 0.292,
"eval_wer": 6.251500428693912,
"step": 11000
},
{
"epoch": 21.53,
"learning_rate": 4.604102564102565e-06,
"loss": 0.0015,
"step": 11025
},
{
"epoch": 21.58,
"learning_rate": 4.591282051282052e-06,
"loss": 0.0008,
"step": 11050
},
{
"epoch": 21.63,
"learning_rate": 4.578461538461539e-06,
"loss": 0.0007,
"step": 11075
},
{
"epoch": 21.68,
"learning_rate": 4.565641025641026e-06,
"loss": 0.0013,
"step": 11100
},
{
"epoch": 21.73,
"learning_rate": 4.552820512820513e-06,
"loss": 0.0017,
"step": 11125
},
{
"epoch": 21.78,
"learning_rate": 4.540000000000001e-06,
"loss": 0.0006,
"step": 11150
},
{
"epoch": 21.83,
"learning_rate": 4.527179487179487e-06,
"loss": 0.0016,
"step": 11175
},
{
"epoch": 21.88,
"learning_rate": 4.514358974358975e-06,
"loss": 0.001,
"step": 11200
},
{
"epoch": 21.92,
"learning_rate": 4.501538461538462e-06,
"loss": 0.0017,
"step": 11225
},
{
"epoch": 21.97,
"learning_rate": 4.488717948717949e-06,
"loss": 0.0015,
"step": 11250
},
{
"epoch": 22.02,
"learning_rate": 4.475897435897436e-06,
"loss": 0.0008,
"step": 11275
},
{
"epoch": 22.07,
"learning_rate": 4.463076923076923e-06,
"loss": 0.0012,
"step": 11300
},
{
"epoch": 22.12,
"learning_rate": 4.450256410256411e-06,
"loss": 0.0005,
"step": 11325
},
{
"epoch": 22.17,
"learning_rate": 4.437435897435898e-06,
"loss": 0.001,
"step": 11350
},
{
"epoch": 22.22,
"learning_rate": 4.424615384615385e-06,
"loss": 0.0006,
"step": 11375
},
{
"epoch": 22.27,
"learning_rate": 4.411794871794872e-06,
"loss": 0.0004,
"step": 11400
},
{
"epoch": 22.31,
"learning_rate": 4.398974358974359e-06,
"loss": 0.0007,
"step": 11425
},
{
"epoch": 22.36,
"learning_rate": 4.386153846153847e-06,
"loss": 0.0012,
"step": 11450
},
{
"epoch": 22.41,
"learning_rate": 4.3733333333333335e-06,
"loss": 0.0009,
"step": 11475
},
{
"epoch": 22.46,
"learning_rate": 4.360512820512821e-06,
"loss": 0.0006,
"step": 11500
},
{
"epoch": 22.51,
"learning_rate": 4.347692307692308e-06,
"loss": 0.0004,
"step": 11525
},
{
"epoch": 22.56,
"learning_rate": 4.334871794871795e-06,
"loss": 0.0003,
"step": 11550
},
{
"epoch": 22.61,
"learning_rate": 4.3220512820512825e-06,
"loss": 0.0004,
"step": 11575
},
{
"epoch": 22.66,
"learning_rate": 4.309230769230769e-06,
"loss": 0.0004,
"step": 11600
},
{
"epoch": 22.71,
"learning_rate": 4.296410256410257e-06,
"loss": 0.0007,
"step": 11625
},
{
"epoch": 22.75,
"learning_rate": 4.283589743589744e-06,
"loss": 0.0011,
"step": 11650
},
{
"epoch": 22.8,
"learning_rate": 4.2707692307692315e-06,
"loss": 0.0009,
"step": 11675
},
{
"epoch": 22.85,
"learning_rate": 4.257948717948718e-06,
"loss": 0.0005,
"step": 11700
},
{
"epoch": 22.9,
"learning_rate": 4.245128205128205e-06,
"loss": 0.0009,
"step": 11725
},
{
"epoch": 22.95,
"learning_rate": 4.232307692307693e-06,
"loss": 0.001,
"step": 11750
},
{
"epoch": 23.0,
"learning_rate": 4.21948717948718e-06,
"loss": 0.0005,
"step": 11775
},
{
"epoch": 23.05,
"learning_rate": 4.206666666666667e-06,
"loss": 0.0007,
"step": 11800
},
{
"epoch": 23.1,
"learning_rate": 4.193846153846154e-06,
"loss": 0.0012,
"step": 11825
},
{
"epoch": 23.14,
"learning_rate": 4.181025641025641e-06,
"loss": 0.0011,
"step": 11850
},
{
"epoch": 23.19,
"learning_rate": 4.168205128205129e-06,
"loss": 0.0009,
"step": 11875
},
{
"epoch": 23.24,
"learning_rate": 4.1553846153846155e-06,
"loss": 0.0009,
"step": 11900
},
{
"epoch": 23.29,
"learning_rate": 4.142564102564103e-06,
"loss": 0.0008,
"step": 11925
},
{
"epoch": 23.34,
"learning_rate": 4.12974358974359e-06,
"loss": 0.0004,
"step": 11950
},
{
"epoch": 23.39,
"learning_rate": 4.116923076923078e-06,
"loss": 0.0006,
"step": 11975
},
{
"epoch": 23.44,
"learning_rate": 4.1041025641025645e-06,
"loss": 0.0003,
"step": 12000
},
{
"epoch": 23.44,
"eval_loss": 0.24715840816497803,
"eval_runtime": 3506.3217,
"eval_samples_per_second": 4.672,
"eval_steps_per_second": 0.292,
"eval_wer": 6.205201486138897,
"step": 12000
},
{
"epoch": 23.49,
"learning_rate": 4.091282051282051e-06,
"loss": 0.0007,
"step": 12025
},
{
"epoch": 23.54,
"learning_rate": 4.078461538461539e-06,
"loss": 0.001,
"step": 12050
},
{
"epoch": 23.58,
"learning_rate": 4.065641025641026e-06,
"loss": 0.0011,
"step": 12075
},
{
"epoch": 23.63,
"learning_rate": 4.0528205128205135e-06,
"loss": 0.0008,
"step": 12100
},
{
"epoch": 23.68,
"learning_rate": 4.04e-06,
"loss": 0.0009,
"step": 12125
},
{
"epoch": 23.73,
"learning_rate": 4.027179487179487e-06,
"loss": 0.0007,
"step": 12150
},
{
"epoch": 23.78,
"learning_rate": 4.014358974358975e-06,
"loss": 0.0006,
"step": 12175
},
{
"epoch": 23.83,
"learning_rate": 4.001538461538462e-06,
"loss": 0.0004,
"step": 12200
},
{
"epoch": 23.88,
"learning_rate": 3.988717948717949e-06,
"loss": 0.0008,
"step": 12225
},
{
"epoch": 23.93,
"learning_rate": 3.975897435897436e-06,
"loss": 0.0004,
"step": 12250
},
{
"epoch": 23.97,
"learning_rate": 3.963076923076924e-06,
"loss": 0.0003,
"step": 12275
},
{
"epoch": 24.02,
"learning_rate": 3.950256410256411e-06,
"loss": 0.0012,
"step": 12300
},
{
"epoch": 24.07,
"learning_rate": 3.937435897435897e-06,
"loss": 0.0002,
"step": 12325
},
{
"epoch": 24.12,
"learning_rate": 3.924615384615385e-06,
"loss": 0.0004,
"step": 12350
},
{
"epoch": 24.17,
"learning_rate": 3.911794871794872e-06,
"loss": 0.0003,
"step": 12375
},
{
"epoch": 24.22,
"learning_rate": 3.89897435897436e-06,
"loss": 0.0005,
"step": 12400
},
{
"epoch": 24.27,
"learning_rate": 3.886153846153846e-06,
"loss": 0.0011,
"step": 12425
},
{
"epoch": 24.32,
"learning_rate": 3.873333333333333e-06,
"loss": 0.0007,
"step": 12450
},
{
"epoch": 24.37,
"learning_rate": 3.860512820512821e-06,
"loss": 0.0006,
"step": 12475
},
{
"epoch": 24.41,
"learning_rate": 3.847692307692308e-06,
"loss": 0.0004,
"step": 12500
},
{
"epoch": 24.46,
"learning_rate": 3.834871794871795e-06,
"loss": 0.0003,
"step": 12525
},
{
"epoch": 24.51,
"learning_rate": 3.822051282051282e-06,
"loss": 0.0003,
"step": 12550
},
{
"epoch": 24.56,
"learning_rate": 3.8092307692307695e-06,
"loss": 0.0006,
"step": 12575
},
{
"epoch": 24.61,
"learning_rate": 3.7964102564102567e-06,
"loss": 0.0002,
"step": 12600
},
{
"epoch": 24.66,
"learning_rate": 3.783589743589744e-06,
"loss": 0.0004,
"step": 12625
},
{
"epoch": 24.71,
"learning_rate": 3.7707692307692312e-06,
"loss": 0.0003,
"step": 12650
},
{
"epoch": 24.76,
"learning_rate": 3.7579487179487185e-06,
"loss": 0.0003,
"step": 12675
},
{
"epoch": 24.8,
"learning_rate": 3.7451282051282057e-06,
"loss": 0.0007,
"step": 12700
},
{
"epoch": 24.85,
"learning_rate": 3.732307692307693e-06,
"loss": 0.001,
"step": 12725
},
{
"epoch": 24.9,
"learning_rate": 3.7194871794871794e-06,
"loss": 0.0005,
"step": 12750
},
{
"epoch": 24.95,
"learning_rate": 3.7066666666666666e-06,
"loss": 0.0006,
"step": 12775
},
{
"epoch": 25.0,
"learning_rate": 3.693846153846154e-06,
"loss": 0.001,
"step": 12800
},
{
"epoch": 25.05,
"learning_rate": 3.681025641025641e-06,
"loss": 0.0009,
"step": 12825
},
{
"epoch": 25.1,
"learning_rate": 3.6682051282051284e-06,
"loss": 0.001,
"step": 12850
},
{
"epoch": 25.15,
"learning_rate": 3.6553846153846156e-06,
"loss": 0.0005,
"step": 12875
},
{
"epoch": 25.2,
"learning_rate": 3.642564102564103e-06,
"loss": 0.0005,
"step": 12900
},
{
"epoch": 25.24,
"learning_rate": 3.62974358974359e-06,
"loss": 0.0007,
"step": 12925
},
{
"epoch": 25.29,
"learning_rate": 3.6169230769230773e-06,
"loss": 0.0006,
"step": 12950
},
{
"epoch": 25.34,
"learning_rate": 3.6041025641025646e-06,
"loss": 0.0008,
"step": 12975
},
{
"epoch": 25.39,
"learning_rate": 3.591282051282052e-06,
"loss": 0.0012,
"step": 13000
},
{
"epoch": 25.39,
"eval_loss": 0.23819763958454132,
"eval_runtime": 3505.409,
"eval_samples_per_second": 4.673,
"eval_steps_per_second": 0.292,
"eval_wer": 6.289225492997999,
"step": 13000
},
{
"epoch": 25.44,
"learning_rate": 3.578461538461539e-06,
"loss": 0.0008,
"step": 13025
},
{
"epoch": 25.49,
"learning_rate": 3.5656410256410263e-06,
"loss": 0.0004,
"step": 13050
},
{
"epoch": 25.54,
"learning_rate": 3.5528205128205127e-06,
"loss": 0.0007,
"step": 13075
},
{
"epoch": 25.59,
"learning_rate": 3.54e-06,
"loss": 0.0005,
"step": 13100
},
{
"epoch": 25.63,
"learning_rate": 3.5271794871794872e-06,
"loss": 0.0003,
"step": 13125
},
{
"epoch": 25.68,
"learning_rate": 3.5143589743589745e-06,
"loss": 0.0007,
"step": 13150
},
{
"epoch": 25.73,
"learning_rate": 3.5015384615384617e-06,
"loss": 0.0003,
"step": 13175
},
{
"epoch": 25.78,
"learning_rate": 3.488717948717949e-06,
"loss": 0.0005,
"step": 13200
},
{
"epoch": 25.83,
"learning_rate": 3.4758974358974362e-06,
"loss": 0.0006,
"step": 13225
},
{
"epoch": 25.88,
"learning_rate": 3.4630769230769235e-06,
"loss": 0.0007,
"step": 13250
},
{
"epoch": 25.93,
"learning_rate": 3.4502564102564107e-06,
"loss": 0.0004,
"step": 13275
},
{
"epoch": 25.98,
"learning_rate": 3.437435897435898e-06,
"loss": 0.0006,
"step": 13300
},
{
"epoch": 26.03,
"learning_rate": 3.4246153846153852e-06,
"loss": 0.0006,
"step": 13325
},
{
"epoch": 26.07,
"learning_rate": 3.4117948717948725e-06,
"loss": 0.0003,
"step": 13350
},
{
"epoch": 26.12,
"learning_rate": 3.398974358974359e-06,
"loss": 0.0004,
"step": 13375
},
{
"epoch": 26.17,
"learning_rate": 3.386153846153846e-06,
"loss": 0.0004,
"step": 13400
},
{
"epoch": 26.22,
"learning_rate": 3.3733333333333334e-06,
"loss": 0.0005,
"step": 13425
},
{
"epoch": 26.27,
"learning_rate": 3.3605128205128206e-06,
"loss": 0.0004,
"step": 13450
},
{
"epoch": 26.32,
"learning_rate": 3.347692307692308e-06,
"loss": 0.0002,
"step": 13475
},
{
"epoch": 26.37,
"learning_rate": 3.334871794871795e-06,
"loss": 0.0003,
"step": 13500
},
{
"epoch": 26.42,
"learning_rate": 3.3220512820512824e-06,
"loss": 0.0002,
"step": 13525
},
{
"epoch": 26.46,
"learning_rate": 3.3092307692307696e-06,
"loss": 0.0005,
"step": 13550
},
{
"epoch": 26.51,
"learning_rate": 3.296410256410257e-06,
"loss": 0.0007,
"step": 13575
},
{
"epoch": 26.56,
"learning_rate": 3.283589743589744e-06,
"loss": 0.0003,
"step": 13600
},
{
"epoch": 26.61,
"learning_rate": 3.2707692307692313e-06,
"loss": 0.0004,
"step": 13625
},
{
"epoch": 26.66,
"learning_rate": 3.2579487179487186e-06,
"loss": 0.0002,
"step": 13650
},
{
"epoch": 26.71,
"learning_rate": 3.245128205128205e-06,
"loss": 0.0008,
"step": 13675
},
{
"epoch": 26.76,
"learning_rate": 3.2323076923076922e-06,
"loss": 0.0004,
"step": 13700
},
{
"epoch": 26.81,
"learning_rate": 3.2194871794871795e-06,
"loss": 0.0003,
"step": 13725
},
{
"epoch": 26.86,
"learning_rate": 3.2066666666666667e-06,
"loss": 0.0004,
"step": 13750
},
{
"epoch": 26.9,
"learning_rate": 3.193846153846154e-06,
"loss": 0.0002,
"step": 13775
},
{
"epoch": 26.95,
"learning_rate": 3.1810256410256412e-06,
"loss": 0.0003,
"step": 13800
},
{
"epoch": 27.0,
"learning_rate": 3.1682051282051285e-06,
"loss": 0.0005,
"step": 13825
},
{
"epoch": 27.05,
"learning_rate": 3.1553846153846157e-06,
"loss": 0.0002,
"step": 13850
},
{
"epoch": 27.1,
"learning_rate": 3.142564102564103e-06,
"loss": 0.0002,
"step": 13875
},
{
"epoch": 27.15,
"learning_rate": 3.1297435897435902e-06,
"loss": 0.0002,
"step": 13900
},
{
"epoch": 27.2,
"learning_rate": 3.1169230769230775e-06,
"loss": 0.0002,
"step": 13925
},
{
"epoch": 27.25,
"learning_rate": 3.1041025641025647e-06,
"loss": 0.0001,
"step": 13950
},
{
"epoch": 27.29,
"learning_rate": 3.091282051282051e-06,
"loss": 0.0001,
"step": 13975
},
{
"epoch": 27.34,
"learning_rate": 3.0784615384615384e-06,
"loss": 0.0001,
"step": 14000
},
{
"epoch": 27.34,
"eval_loss": 0.25495532155036926,
"eval_runtime": 3512.4249,
"eval_samples_per_second": 4.663,
"eval_steps_per_second": 0.292,
"eval_wer": 5.994855673049443,
"step": 14000
},
{
"epoch": 27.39,
"learning_rate": 3.0656410256410256e-06,
"loss": 0.0001,
"step": 14025
},
{
"epoch": 27.44,
"learning_rate": 3.052820512820513e-06,
"loss": 0.0003,
"step": 14050
},
{
"epoch": 27.49,
"learning_rate": 3.04e-06,
"loss": 0.0001,
"step": 14075
},
{
"epoch": 27.54,
"learning_rate": 3.0271794871794874e-06,
"loss": 0.0005,
"step": 14100
},
{
"epoch": 27.59,
"learning_rate": 3.0143589743589746e-06,
"loss": 0.0002,
"step": 14125
},
{
"epoch": 27.64,
"learning_rate": 3.001538461538462e-06,
"loss": 0.0002,
"step": 14150
},
{
"epoch": 27.69,
"learning_rate": 2.988717948717949e-06,
"loss": 0.0002,
"step": 14175
},
{
"epoch": 27.73,
"learning_rate": 2.9758974358974364e-06,
"loss": 0.0002,
"step": 14200
},
{
"epoch": 27.78,
"learning_rate": 2.9630769230769236e-06,
"loss": 0.0002,
"step": 14225
},
{
"epoch": 27.83,
"learning_rate": 2.950769230769231e-06,
"loss": 0.0003,
"step": 14250
},
{
"epoch": 27.88,
"learning_rate": 2.9379487179487183e-06,
"loss": 0.0001,
"step": 14275
},
{
"epoch": 27.93,
"learning_rate": 2.9251282051282055e-06,
"loss": 0.0001,
"step": 14300
},
{
"epoch": 27.98,
"learning_rate": 2.9123076923076928e-06,
"loss": 0.0001,
"step": 14325
},
{
"epoch": 28.03,
"learning_rate": 2.89948717948718e-06,
"loss": 0.0002,
"step": 14350
},
{
"epoch": 28.08,
"learning_rate": 2.8866666666666673e-06,
"loss": 0.0001,
"step": 14375
},
{
"epoch": 28.12,
"learning_rate": 2.873846153846154e-06,
"loss": 0.0002,
"step": 14400
},
{
"epoch": 28.17,
"learning_rate": 2.8610256410256413e-06,
"loss": 0.0002,
"step": 14425
},
{
"epoch": 28.22,
"learning_rate": 2.848205128205128e-06,
"loss": 0.0005,
"step": 14450
},
{
"epoch": 28.27,
"learning_rate": 2.8353846153846154e-06,
"loss": 0.0001,
"step": 14475
},
{
"epoch": 28.32,
"learning_rate": 2.8225641025641027e-06,
"loss": 0.0001,
"step": 14500
},
{
"epoch": 28.37,
"learning_rate": 2.80974358974359e-06,
"loss": 0.0003,
"step": 14525
},
{
"epoch": 28.42,
"learning_rate": 2.796923076923077e-06,
"loss": 0.0004,
"step": 14550
},
{
"epoch": 28.47,
"learning_rate": 2.7841025641025644e-06,
"loss": 0.0006,
"step": 14575
},
{
"epoch": 28.52,
"learning_rate": 2.7712820512820516e-06,
"loss": 0.0001,
"step": 14600
},
{
"epoch": 28.56,
"learning_rate": 2.758461538461539e-06,
"loss": 0.0003,
"step": 14625
},
{
"epoch": 28.61,
"learning_rate": 2.745641025641026e-06,
"loss": 0.0002,
"step": 14650
},
{
"epoch": 28.66,
"learning_rate": 2.7328205128205134e-06,
"loss": 0.0002,
"step": 14675
},
{
"epoch": 28.71,
"learning_rate": 2.7200000000000002e-06,
"loss": 0.0001,
"step": 14700
},
{
"epoch": 28.76,
"learning_rate": 2.7071794871794875e-06,
"loss": 0.0003,
"step": 14725
},
{
"epoch": 28.81,
"learning_rate": 2.6943589743589743e-06,
"loss": 0.0001,
"step": 14750
},
{
"epoch": 28.86,
"learning_rate": 2.6815384615384615e-06,
"loss": 0.0004,
"step": 14775
},
{
"epoch": 28.91,
"learning_rate": 2.6687179487179488e-06,
"loss": 0.0006,
"step": 14800
},
{
"epoch": 28.96,
"learning_rate": 2.655897435897436e-06,
"loss": 0.0005,
"step": 14825
},
{
"epoch": 29.0,
"learning_rate": 2.6430769230769233e-06,
"loss": 0.0002,
"step": 14850
},
{
"epoch": 29.05,
"learning_rate": 2.6302564102564105e-06,
"loss": 0.0006,
"step": 14875
},
{
"epoch": 29.1,
"learning_rate": 2.6174358974358978e-06,
"loss": 0.0004,
"step": 14900
},
{
"epoch": 29.15,
"learning_rate": 2.604615384615385e-06,
"loss": 0.0005,
"step": 14925
},
{
"epoch": 29.2,
"learning_rate": 2.5917948717948723e-06,
"loss": 0.0002,
"step": 14950
},
{
"epoch": 29.25,
"learning_rate": 2.5789743589743595e-06,
"loss": 0.0005,
"step": 14975
},
{
"epoch": 29.3,
"learning_rate": 2.5661538461538463e-06,
"loss": 0.0006,
"step": 15000
},
{
"epoch": 29.3,
"eval_loss": 0.25740066170692444,
"eval_runtime": 3497.5649,
"eval_samples_per_second": 4.683,
"eval_steps_per_second": 0.293,
"eval_wer": 6.360674478422407,
"step": 15000
},
{
"epoch": 29.35,
"learning_rate": 2.5533333333333336e-06,
"loss": 0.0005,
"step": 15025
},
{
"epoch": 29.39,
"learning_rate": 2.5405128205128204e-06,
"loss": 0.0003,
"step": 15050
},
{
"epoch": 29.44,
"learning_rate": 2.5276923076923077e-06,
"loss": 0.0002,
"step": 15075
},
{
"epoch": 29.49,
"learning_rate": 2.514871794871795e-06,
"loss": 0.0004,
"step": 15100
},
{
"epoch": 29.54,
"learning_rate": 2.502051282051282e-06,
"loss": 0.0006,
"step": 15125
},
{
"epoch": 29.59,
"learning_rate": 2.4892307692307694e-06,
"loss": 0.0004,
"step": 15150
},
{
"epoch": 29.64,
"learning_rate": 2.4764102564102567e-06,
"loss": 0.0008,
"step": 15175
},
{
"epoch": 29.69,
"learning_rate": 2.463589743589744e-06,
"loss": 0.0006,
"step": 15200
},
{
"epoch": 29.74,
"learning_rate": 2.450769230769231e-06,
"loss": 0.0003,
"step": 15225
},
{
"epoch": 29.79,
"learning_rate": 2.4379487179487184e-06,
"loss": 0.001,
"step": 15250
},
{
"epoch": 29.83,
"learning_rate": 2.4251282051282056e-06,
"loss": 0.0002,
"step": 15275
},
{
"epoch": 29.88,
"learning_rate": 2.4123076923076925e-06,
"loss": 0.0007,
"step": 15300
},
{
"epoch": 29.93,
"learning_rate": 2.3994871794871797e-06,
"loss": 0.0002,
"step": 15325
},
{
"epoch": 29.98,
"learning_rate": 2.386666666666667e-06,
"loss": 0.0003,
"step": 15350
},
{
"epoch": 30.03,
"learning_rate": 2.3738461538461542e-06,
"loss": 0.0005,
"step": 15375
},
{
"epoch": 30.08,
"learning_rate": 2.3610256410256415e-06,
"loss": 0.0002,
"step": 15400
},
{
"epoch": 30.13,
"learning_rate": 2.3482051282051287e-06,
"loss": 0.0005,
"step": 15425
},
{
"epoch": 30.18,
"learning_rate": 2.3353846153846155e-06,
"loss": 0.0003,
"step": 15450
},
{
"epoch": 30.22,
"learning_rate": 2.3225641025641028e-06,
"loss": 0.0003,
"step": 15475
},
{
"epoch": 30.27,
"learning_rate": 2.30974358974359e-06,
"loss": 0.0002,
"step": 15500
},
{
"epoch": 30.32,
"learning_rate": 2.2969230769230773e-06,
"loss": 0.0002,
"step": 15525
},
{
"epoch": 30.37,
"learning_rate": 2.2841025641025645e-06,
"loss": 0.0001,
"step": 15550
},
{
"epoch": 30.42,
"learning_rate": 2.2712820512820518e-06,
"loss": 0.0002,
"step": 15575
},
{
"epoch": 30.47,
"learning_rate": 2.2584615384615386e-06,
"loss": 0.0001,
"step": 15600
},
{
"epoch": 30.52,
"learning_rate": 2.245641025641026e-06,
"loss": 0.0005,
"step": 15625
},
{
"epoch": 30.57,
"learning_rate": 2.232820512820513e-06,
"loss": 0.0007,
"step": 15650
},
{
"epoch": 30.62,
"learning_rate": 2.2200000000000003e-06,
"loss": 0.0004,
"step": 15675
},
{
"epoch": 30.66,
"learning_rate": 2.2071794871794876e-06,
"loss": 0.0003,
"step": 15700
},
{
"epoch": 30.71,
"learning_rate": 2.194358974358975e-06,
"loss": 0.0002,
"step": 15725
},
{
"epoch": 30.76,
"learning_rate": 2.1815384615384617e-06,
"loss": 0.0002,
"step": 15750
},
{
"epoch": 30.81,
"learning_rate": 2.168717948717949e-06,
"loss": 0.0007,
"step": 15775
},
{
"epoch": 30.86,
"learning_rate": 2.155897435897436e-06,
"loss": 0.0002,
"step": 15800
},
{
"epoch": 30.91,
"learning_rate": 2.1430769230769234e-06,
"loss": 0.0001,
"step": 15825
},
{
"epoch": 30.96,
"learning_rate": 2.1302564102564107e-06,
"loss": 0.0001,
"step": 15850
},
{
"epoch": 31.01,
"learning_rate": 2.117435897435898e-06,
"loss": 0.0001,
"step": 15875
},
{
"epoch": 31.05,
"learning_rate": 2.1046153846153847e-06,
"loss": 0.0001,
"step": 15900
},
{
"epoch": 31.1,
"learning_rate": 2.091794871794872e-06,
"loss": 0.0001,
"step": 15925
},
{
"epoch": 31.15,
"learning_rate": 2.0789743589743592e-06,
"loss": 0.0001,
"step": 15950
},
{
"epoch": 31.2,
"learning_rate": 2.0661538461538465e-06,
"loss": 0.0001,
"step": 15975
},
{
"epoch": 31.25,
"learning_rate": 2.0533333333333337e-06,
"loss": 0.0001,
"step": 16000
},
{
"epoch": 31.25,
"eval_loss": 0.2583903968334198,
"eval_runtime": 3498.2451,
"eval_samples_per_second": 4.682,
"eval_steps_per_second": 0.293,
"eval_wer": 6.014289797084881,
"step": 16000
},
{
"epoch": 31.3,
"learning_rate": 2.040512820512821e-06,
"loss": 0.0001,
"step": 16025
},
{
"epoch": 31.35,
"learning_rate": 2.027692307692308e-06,
"loss": 0.0001,
"step": 16050
},
{
"epoch": 31.4,
"learning_rate": 2.014871794871795e-06,
"loss": 0.0001,
"step": 16075
},
{
"epoch": 31.45,
"learning_rate": 2.0020512820512823e-06,
"loss": 0.0001,
"step": 16100
},
{
"epoch": 31.49,
"learning_rate": 1.9892307692307695e-06,
"loss": 0.0001,
"step": 16125
},
{
"epoch": 31.54,
"learning_rate": 1.9764102564102568e-06,
"loss": 0.0001,
"step": 16150
},
{
"epoch": 31.59,
"learning_rate": 1.963589743589744e-06,
"loss": 0.0001,
"step": 16175
},
{
"epoch": 31.64,
"learning_rate": 1.950769230769231e-06,
"loss": 0.0001,
"step": 16200
},
{
"epoch": 31.69,
"learning_rate": 1.937948717948718e-06,
"loss": 0.0001,
"step": 16225
},
{
"epoch": 31.74,
"learning_rate": 1.9251282051282054e-06,
"loss": 0.0001,
"step": 16250
},
{
"epoch": 31.79,
"learning_rate": 1.9123076923076926e-06,
"loss": 0.0001,
"step": 16275
},
{
"epoch": 31.84,
"learning_rate": 1.8994871794871796e-06,
"loss": 0.0001,
"step": 16300
},
{
"epoch": 31.88,
"learning_rate": 1.8866666666666669e-06,
"loss": 0.0001,
"step": 16325
},
{
"epoch": 31.93,
"learning_rate": 1.873846153846154e-06,
"loss": 0.0001,
"step": 16350
},
{
"epoch": 31.98,
"learning_rate": 1.8610256410256412e-06,
"loss": 0.0001,
"step": 16375
},
{
"epoch": 32.03,
"learning_rate": 1.8482051282051284e-06,
"loss": 0.0001,
"step": 16400
},
{
"epoch": 32.08,
"learning_rate": 1.8353846153846155e-06,
"loss": 0.0001,
"step": 16425
},
{
"epoch": 32.13,
"learning_rate": 1.8225641025641027e-06,
"loss": 0.0001,
"step": 16450
},
{
"epoch": 32.18,
"learning_rate": 1.80974358974359e-06,
"loss": 0.0001,
"step": 16475
},
{
"epoch": 32.23,
"learning_rate": 1.796923076923077e-06,
"loss": 0.0001,
"step": 16500
},
{
"epoch": 32.28,
"learning_rate": 1.7841025641025642e-06,
"loss": 0.0001,
"step": 16525
},
{
"epoch": 32.32,
"learning_rate": 1.7712820512820515e-06,
"loss": 0.0001,
"step": 16550
},
{
"epoch": 32.37,
"learning_rate": 1.7584615384615385e-06,
"loss": 0.0001,
"step": 16575
},
{
"epoch": 32.42,
"learning_rate": 1.7456410256410258e-06,
"loss": 0.0001,
"step": 16600
},
{
"epoch": 32.47,
"learning_rate": 1.732820512820513e-06,
"loss": 0.0001,
"step": 16625
},
{
"epoch": 32.52,
"learning_rate": 1.72e-06,
"loss": 0.0001,
"step": 16650
},
{
"epoch": 32.57,
"learning_rate": 1.7071794871794873e-06,
"loss": 0.0001,
"step": 16675
},
{
"epoch": 32.62,
"learning_rate": 1.6943589743589745e-06,
"loss": 0.0001,
"step": 16700
},
{
"epoch": 32.67,
"learning_rate": 1.6815384615384616e-06,
"loss": 0.0001,
"step": 16725
},
{
"epoch": 32.71,
"learning_rate": 1.6687179487179488e-06,
"loss": 0.0001,
"step": 16750
},
{
"epoch": 32.76,
"learning_rate": 1.655897435897436e-06,
"loss": 0.0001,
"step": 16775
},
{
"epoch": 32.81,
"learning_rate": 1.6430769230769233e-06,
"loss": 0.0001,
"step": 16800
},
{
"epoch": 32.86,
"learning_rate": 1.6302564102564104e-06,
"loss": 0.0001,
"step": 16825
},
{
"epoch": 32.91,
"learning_rate": 1.6174358974358976e-06,
"loss": 0.0001,
"step": 16850
},
{
"epoch": 32.96,
"learning_rate": 1.6046153846153846e-06,
"loss": 0.0001,
"step": 16875
},
{
"epoch": 33.01,
"learning_rate": 1.5917948717948719e-06,
"loss": 0.0001,
"step": 16900
},
{
"epoch": 33.06,
"learning_rate": 1.5789743589743591e-06,
"loss": 0.0001,
"step": 16925
},
{
"epoch": 33.11,
"learning_rate": 1.5661538461538464e-06,
"loss": 0.0001,
"step": 16950
},
{
"epoch": 33.15,
"learning_rate": 1.5533333333333334e-06,
"loss": 0.0001,
"step": 16975
},
{
"epoch": 33.2,
"learning_rate": 1.5405128205128207e-06,
"loss": 0.0001,
"step": 17000
},
{
"epoch": 33.2,
"eval_loss": 0.26860281825065613,
"eval_runtime": 3497.2481,
"eval_samples_per_second": 4.684,
"eval_steps_per_second": 0.293,
"eval_wer": 5.948556730494427,
"step": 17000
},
{
"epoch": 33.25,
"learning_rate": 1.5276923076923077e-06,
"loss": 0.0001,
"step": 17025
},
{
"epoch": 33.3,
"learning_rate": 1.514871794871795e-06,
"loss": 0.0001,
"step": 17050
},
{
"epoch": 33.35,
"learning_rate": 1.5020512820512822e-06,
"loss": 0.0001,
"step": 17075
},
{
"epoch": 33.4,
"learning_rate": 1.4892307692307695e-06,
"loss": 0.0001,
"step": 17100
},
{
"epoch": 33.45,
"learning_rate": 1.4764102564102565e-06,
"loss": 0.0001,
"step": 17125
},
{
"epoch": 33.5,
"learning_rate": 1.4635897435897437e-06,
"loss": 0.0001,
"step": 17150
},
{
"epoch": 33.54,
"learning_rate": 1.4507692307692308e-06,
"loss": 0.0001,
"step": 17175
},
{
"epoch": 33.59,
"learning_rate": 1.437948717948718e-06,
"loss": 0.0001,
"step": 17200
},
{
"epoch": 33.64,
"learning_rate": 1.4251282051282053e-06,
"loss": 0.0001,
"step": 17225
},
{
"epoch": 33.69,
"learning_rate": 1.4123076923076925e-06,
"loss": 0.0001,
"step": 17250
},
{
"epoch": 33.74,
"learning_rate": 1.3994871794871796e-06,
"loss": 0.0001,
"step": 17275
},
{
"epoch": 33.79,
"learning_rate": 1.3866666666666668e-06,
"loss": 0.0001,
"step": 17300
},
{
"epoch": 33.84,
"learning_rate": 1.3738461538461538e-06,
"loss": 0.0001,
"step": 17325
},
{
"epoch": 33.89,
"learning_rate": 1.361025641025641e-06,
"loss": 0.0001,
"step": 17350
},
{
"epoch": 33.94,
"learning_rate": 1.3482051282051283e-06,
"loss": 0.0001,
"step": 17375
},
{
"epoch": 33.98,
"learning_rate": 1.3353846153846156e-06,
"loss": 0.0001,
"step": 17400
},
{
"epoch": 34.03,
"learning_rate": 1.3225641025641026e-06,
"loss": 0.0001,
"step": 17425
},
{
"epoch": 34.08,
"learning_rate": 1.3097435897435899e-06,
"loss": 0.0001,
"step": 17450
},
{
"epoch": 34.13,
"learning_rate": 1.296923076923077e-06,
"loss": 0.0001,
"step": 17475
},
{
"epoch": 34.18,
"learning_rate": 1.2841025641025641e-06,
"loss": 0.0001,
"step": 17500
},
{
"epoch": 34.23,
"learning_rate": 1.2712820512820514e-06,
"loss": 0.0001,
"step": 17525
},
{
"epoch": 34.28,
"learning_rate": 1.2584615384615386e-06,
"loss": 0.0001,
"step": 17550
},
{
"epoch": 34.33,
"learning_rate": 1.2456410256410257e-06,
"loss": 0.0001,
"step": 17575
},
{
"epoch": 34.38,
"learning_rate": 1.232820512820513e-06,
"loss": 0.0001,
"step": 17600
},
{
"epoch": 34.42,
"learning_rate": 1.2200000000000002e-06,
"loss": 0.0001,
"step": 17625
},
{
"epoch": 34.47,
"learning_rate": 1.2071794871794872e-06,
"loss": 0.0001,
"step": 17650
},
{
"epoch": 34.52,
"learning_rate": 1.1943589743589745e-06,
"loss": 0.0001,
"step": 17675
},
{
"epoch": 34.57,
"learning_rate": 1.1815384615384617e-06,
"loss": 0.0001,
"step": 17700
},
{
"epoch": 34.62,
"learning_rate": 1.1687179487179487e-06,
"loss": 0.0001,
"step": 17725
},
{
"epoch": 34.67,
"learning_rate": 1.155897435897436e-06,
"loss": 0.0001,
"step": 17750
},
{
"epoch": 34.72,
"learning_rate": 1.1430769230769232e-06,
"loss": 0.0001,
"step": 17775
},
{
"epoch": 34.77,
"learning_rate": 1.1302564102564103e-06,
"loss": 0.0001,
"step": 17800
},
{
"epoch": 34.81,
"learning_rate": 1.1174358974358975e-06,
"loss": 0.0001,
"step": 17825
},
{
"epoch": 34.86,
"learning_rate": 1.1046153846153848e-06,
"loss": 0.0001,
"step": 17850
},
{
"epoch": 34.91,
"learning_rate": 1.0917948717948718e-06,
"loss": 0.0001,
"step": 17875
},
{
"epoch": 34.96,
"learning_rate": 1.078974358974359e-06,
"loss": 0.0001,
"step": 17900
},
{
"epoch": 35.01,
"learning_rate": 1.0661538461538463e-06,
"loss": 0.0001,
"step": 17925
},
{
"epoch": 35.06,
"learning_rate": 1.0533333333333333e-06,
"loss": 0.0001,
"step": 17950
},
{
"epoch": 35.11,
"learning_rate": 1.0405128205128206e-06,
"loss": 0.0,
"step": 17975
},
{
"epoch": 35.16,
"learning_rate": 1.0276923076923078e-06,
"loss": 0.0,
"step": 18000
},
{
"epoch": 35.16,
"eval_loss": 0.2735806107521057,
"eval_runtime": 3545.141,
"eval_samples_per_second": 4.62,
"eval_steps_per_second": 0.289,
"eval_wer": 5.919405544441269,
"step": 18000
},
{
"epoch": 35.21,
"learning_rate": 1.0148717948717949e-06,
"loss": 0.0,
"step": 18025
},
{
"epoch": 35.25,
"learning_rate": 1.0020512820512821e-06,
"loss": 0.0,
"step": 18050
},
{
"epoch": 35.3,
"learning_rate": 9.892307692307694e-07,
"loss": 0.0001,
"step": 18075
},
{
"epoch": 35.35,
"learning_rate": 9.764102564102564e-07,
"loss": 0.0001,
"step": 18100
},
{
"epoch": 35.4,
"learning_rate": 9.635897435897437e-07,
"loss": 0.0001,
"step": 18125
},
{
"epoch": 35.45,
"learning_rate": 9.507692307692308e-07,
"loss": 0.0001,
"step": 18150
},
{
"epoch": 35.5,
"learning_rate": 9.37948717948718e-07,
"loss": 0.0,
"step": 18175
},
{
"epoch": 35.55,
"learning_rate": 9.251282051282053e-07,
"loss": 0.0001,
"step": 18200
},
{
"epoch": 35.6,
"learning_rate": 9.123076923076923e-07,
"loss": 0.0,
"step": 18225
},
{
"epoch": 35.64,
"learning_rate": 8.994871794871796e-07,
"loss": 0.0,
"step": 18250
},
{
"epoch": 35.69,
"learning_rate": 8.866666666666668e-07,
"loss": 0.0,
"step": 18275
},
{
"epoch": 35.74,
"learning_rate": 8.738461538461539e-07,
"loss": 0.0,
"step": 18300
},
{
"epoch": 35.79,
"learning_rate": 8.610256410256411e-07,
"loss": 0.0,
"step": 18325
},
{
"epoch": 35.84,
"learning_rate": 8.482051282051284e-07,
"loss": 0.0,
"step": 18350
},
{
"epoch": 35.89,
"learning_rate": 8.353846153846154e-07,
"loss": 0.0001,
"step": 18375
},
{
"epoch": 35.94,
"learning_rate": 8.225641025641026e-07,
"loss": 0.0,
"step": 18400
},
{
"epoch": 35.99,
"learning_rate": 8.097435897435899e-07,
"loss": 0.0,
"step": 18425
},
{
"epoch": 36.04,
"learning_rate": 7.969230769230769e-07,
"loss": 0.0,
"step": 18450
},
{
"epoch": 36.08,
"learning_rate": 7.841025641025642e-07,
"loss": 0.0,
"step": 18475
},
{
"epoch": 36.13,
"learning_rate": 7.712820512820514e-07,
"loss": 0.0,
"step": 18500
},
{
"epoch": 36.18,
"learning_rate": 7.584615384615385e-07,
"loss": 0.0,
"step": 18525
},
{
"epoch": 36.23,
"learning_rate": 7.456410256410257e-07,
"loss": 0.0,
"step": 18550
},
{
"epoch": 36.28,
"learning_rate": 7.32820512820513e-07,
"loss": 0.0,
"step": 18575
},
{
"epoch": 36.33,
"learning_rate": 7.2e-07,
"loss": 0.0,
"step": 18600
},
{
"epoch": 36.38,
"learning_rate": 7.071794871794872e-07,
"loss": 0.0,
"step": 18625
},
{
"epoch": 36.43,
"learning_rate": 6.943589743589745e-07,
"loss": 0.0,
"step": 18650
},
{
"epoch": 36.47,
"learning_rate": 6.815384615384615e-07,
"loss": 0.0,
"step": 18675
},
{
"epoch": 36.52,
"learning_rate": 6.687179487179488e-07,
"loss": 0.0,
"step": 18700
},
{
"epoch": 36.57,
"learning_rate": 6.55897435897436e-07,
"loss": 0.0,
"step": 18725
},
{
"epoch": 36.62,
"learning_rate": 6.43076923076923e-07,
"loss": 0.0,
"step": 18750
},
{
"epoch": 36.67,
"learning_rate": 6.302564102564103e-07,
"loss": 0.0,
"step": 18775
},
{
"epoch": 36.72,
"learning_rate": 6.174358974358975e-07,
"loss": 0.0,
"step": 18800
},
{
"epoch": 36.77,
"learning_rate": 6.046153846153847e-07,
"loss": 0.0,
"step": 18825
},
{
"epoch": 36.82,
"learning_rate": 5.917948717948718e-07,
"loss": 0.0,
"step": 18850
},
{
"epoch": 36.87,
"learning_rate": 5.789743589743591e-07,
"loss": 0.0,
"step": 18875
},
{
"epoch": 36.91,
"learning_rate": 5.661538461538462e-07,
"loss": 0.0,
"step": 18900
},
{
"epoch": 36.96,
"learning_rate": 5.533333333333334e-07,
"loss": 0.0,
"step": 18925
},
{
"epoch": 37.01,
"learning_rate": 5.405128205128206e-07,
"loss": 0.0,
"step": 18950
},
{
"epoch": 37.06,
"learning_rate": 5.276923076923078e-07,
"loss": 0.0,
"step": 18975
},
{
"epoch": 37.11,
"learning_rate": 5.148717948717949e-07,
"loss": 0.0,
"step": 19000
},
{
"epoch": 37.11,
"eval_loss": 0.2767850160598755,
"eval_runtime": 3545.9564,
"eval_samples_per_second": 4.619,
"eval_steps_per_second": 0.289,
"eval_wer": 5.964561303229495,
"step": 19000
},
{
"epoch": 37.16,
"learning_rate": 5.020512820512821e-07,
"loss": 0.0,
"step": 19025
},
{
"epoch": 37.21,
"learning_rate": 4.892307692307693e-07,
"loss": 0.0,
"step": 19050
},
{
"epoch": 37.26,
"learning_rate": 4.7641025641025643e-07,
"loss": 0.0,
"step": 19075
},
{
"epoch": 37.3,
"learning_rate": 4.635897435897437e-07,
"loss": 0.0,
"step": 19100
},
{
"epoch": 37.35,
"learning_rate": 4.507692307692308e-07,
"loss": 0.0,
"step": 19125
},
{
"epoch": 37.4,
"learning_rate": 4.3794871794871796e-07,
"loss": 0.0,
"step": 19150
},
{
"epoch": 37.45,
"learning_rate": 4.251282051282052e-07,
"loss": 0.0,
"step": 19175
},
{
"epoch": 37.5,
"learning_rate": 4.1230769230769235e-07,
"loss": 0.0,
"step": 19200
},
{
"epoch": 37.55,
"learning_rate": 3.994871794871795e-07,
"loss": 0.0,
"step": 19225
},
{
"epoch": 37.6,
"learning_rate": 3.8666666666666674e-07,
"loss": 0.0,
"step": 19250
},
{
"epoch": 37.65,
"learning_rate": 3.738461538461539e-07,
"loss": 0.0,
"step": 19275
},
{
"epoch": 37.7,
"learning_rate": 3.61025641025641e-07,
"loss": 0.0,
"step": 19300
},
{
"epoch": 37.74,
"learning_rate": 3.4820512820512827e-07,
"loss": 0.0,
"step": 19325
},
{
"epoch": 37.79,
"learning_rate": 3.353846153846154e-07,
"loss": 0.0,
"step": 19350
},
{
"epoch": 37.84,
"learning_rate": 3.2256410256410256e-07,
"loss": 0.0,
"step": 19375
},
{
"epoch": 37.89,
"learning_rate": 3.097435897435898e-07,
"loss": 0.0,
"step": 19400
},
{
"epoch": 37.94,
"learning_rate": 2.9692307692307695e-07,
"loss": 0.0,
"step": 19425
},
{
"epoch": 37.99,
"learning_rate": 2.8410256410256414e-07,
"loss": 0.0,
"step": 19450
},
{
"epoch": 38.04,
"learning_rate": 2.7128205128205134e-07,
"loss": 0.0,
"step": 19475
},
{
"epoch": 38.09,
"learning_rate": 2.584615384615385e-07,
"loss": 0.0,
"step": 19500
},
{
"epoch": 38.13,
"learning_rate": 2.456410256410257e-07,
"loss": 0.0,
"step": 19525
},
{
"epoch": 38.18,
"learning_rate": 2.3282051282051284e-07,
"loss": 0.0,
"step": 19550
},
{
"epoch": 38.23,
"learning_rate": 2.2e-07,
"loss": 0.0,
"step": 19575
},
{
"epoch": 38.28,
"learning_rate": 2.071794871794872e-07,
"loss": 0.0,
"step": 19600
},
{
"epoch": 38.33,
"learning_rate": 1.9435897435897437e-07,
"loss": 0.0,
"step": 19625
},
{
"epoch": 38.38,
"learning_rate": 1.8153846153846154e-07,
"loss": 0.0,
"step": 19650
},
{
"epoch": 38.43,
"learning_rate": 1.6871794871794874e-07,
"loss": 0.0,
"step": 19675
},
{
"epoch": 38.48,
"learning_rate": 1.558974358974359e-07,
"loss": 0.0,
"step": 19700
},
{
"epoch": 38.53,
"learning_rate": 1.430769230769231e-07,
"loss": 0.0,
"step": 19725
},
{
"epoch": 38.57,
"learning_rate": 1.3025641025641027e-07,
"loss": 0.0,
"step": 19750
},
{
"epoch": 38.62,
"learning_rate": 1.1743589743589745e-07,
"loss": 0.0,
"step": 19775
},
{
"epoch": 38.67,
"learning_rate": 1.0461538461538462e-07,
"loss": 0.0,
"step": 19800
},
{
"epoch": 38.72,
"learning_rate": 9.179487179487179e-08,
"loss": 0.0,
"step": 19825
},
{
"epoch": 38.77,
"learning_rate": 7.897435897435898e-08,
"loss": 0.0,
"step": 19850
},
{
"epoch": 38.82,
"learning_rate": 6.615384615384615e-08,
"loss": 0.0,
"step": 19875
},
{
"epoch": 38.87,
"learning_rate": 5.3333333333333334e-08,
"loss": 0.0,
"step": 19900
},
{
"epoch": 38.92,
"learning_rate": 4.0512820512820516e-08,
"loss": 0.0,
"step": 19925
},
{
"epoch": 38.96,
"learning_rate": 2.769230769230769e-08,
"loss": 0.0,
"step": 19950
},
{
"epoch": 39.01,
"learning_rate": 1.4871794871794873e-08,
"loss": 0.0,
"step": 19975
},
{
"epoch": 39.06,
"learning_rate": 2.051282051282051e-09,
"loss": 0.0,
"step": 20000
},
{
"epoch": 39.06,
"eval_loss": 0.2782536447048187,
"eval_runtime": 3558.7248,
"eval_samples_per_second": 4.603,
"eval_steps_per_second": 0.288,
"eval_wer": 5.971420405830237,
"step": 20000
},
{
"epoch": 39.06,
"step": 20000,
"total_flos": 4.3477378129251533e+21,
"train_loss": 0.010816604709852254,
"train_runtime": 181349.8483,
"train_samples_per_second": 7.058,
"train_steps_per_second": 0.11
}
],
"logging_steps": 25,
"max_steps": 20000,
"num_input_tokens_seen": 0,
"num_train_epochs": 40,
"save_steps": 1000,
"total_flos": 4.3477378129251533e+21,
"train_batch_size": 32,
"trial_name": null,
"trial_params": null
}