whisper-small-id / trainer_state.json
cahya's picture
finetuned with more dataset
fe26a48
{
"best_metric": 6.059208706077654,
"best_model_checkpoint": "./whisper-small-id/checkpoint-7000",
"epoch": 6.618133686300463,
"global_step": 10000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.02,
"learning_rate": 4.800000000000001e-07,
"loss": 0.6222,
"step": 25
},
{
"epoch": 0.03,
"learning_rate": 9.800000000000001e-07,
"loss": 0.5723,
"step": 50
},
{
"epoch": 0.05,
"learning_rate": 1.48e-06,
"loss": 0.4941,
"step": 75
},
{
"epoch": 0.07,
"learning_rate": 1.98e-06,
"loss": 0.4433,
"step": 100
},
{
"epoch": 0.08,
"learning_rate": 2.4800000000000004e-06,
"loss": 0.3808,
"step": 125
},
{
"epoch": 0.1,
"learning_rate": 2.9800000000000003e-06,
"loss": 0.3698,
"step": 150
},
{
"epoch": 0.12,
"learning_rate": 3.48e-06,
"loss": 0.336,
"step": 175
},
{
"epoch": 0.13,
"learning_rate": 3.980000000000001e-06,
"loss": 0.3213,
"step": 200
},
{
"epoch": 0.15,
"learning_rate": 4.48e-06,
"loss": 0.2962,
"step": 225
},
{
"epoch": 0.17,
"learning_rate": 4.980000000000001e-06,
"loss": 0.266,
"step": 250
},
{
"epoch": 0.18,
"learning_rate": 5.480000000000001e-06,
"loss": 0.2579,
"step": 275
},
{
"epoch": 0.2,
"learning_rate": 5.98e-06,
"loss": 0.2627,
"step": 300
},
{
"epoch": 0.22,
"learning_rate": 6.480000000000001e-06,
"loss": 0.2557,
"step": 325
},
{
"epoch": 0.23,
"learning_rate": 6.98e-06,
"loss": 0.2568,
"step": 350
},
{
"epoch": 0.25,
"learning_rate": 7.48e-06,
"loss": 0.2668,
"step": 375
},
{
"epoch": 0.26,
"learning_rate": 7.980000000000002e-06,
"loss": 0.2397,
"step": 400
},
{
"epoch": 0.28,
"learning_rate": 8.48e-06,
"loss": 0.2181,
"step": 425
},
{
"epoch": 0.3,
"learning_rate": 8.98e-06,
"loss": 0.2231,
"step": 450
},
{
"epoch": 0.31,
"learning_rate": 9.48e-06,
"loss": 0.2178,
"step": 475
},
{
"epoch": 0.33,
"learning_rate": 9.980000000000001e-06,
"loss": 0.2073,
"step": 500
},
{
"epoch": 0.35,
"learning_rate": 9.974736842105263e-06,
"loss": 0.1962,
"step": 525
},
{
"epoch": 0.36,
"learning_rate": 9.94842105263158e-06,
"loss": 0.203,
"step": 550
},
{
"epoch": 0.38,
"learning_rate": 9.922105263157895e-06,
"loss": 0.185,
"step": 575
},
{
"epoch": 0.4,
"learning_rate": 9.895789473684212e-06,
"loss": 0.1874,
"step": 600
},
{
"epoch": 0.41,
"learning_rate": 9.869473684210528e-06,
"loss": 0.1829,
"step": 625
},
{
"epoch": 0.43,
"learning_rate": 9.843157894736843e-06,
"loss": 0.175,
"step": 650
},
{
"epoch": 0.45,
"learning_rate": 9.816842105263158e-06,
"loss": 0.1751,
"step": 675
},
{
"epoch": 0.46,
"learning_rate": 9.790526315789475e-06,
"loss": 0.1796,
"step": 700
},
{
"epoch": 0.48,
"learning_rate": 9.76421052631579e-06,
"loss": 0.1754,
"step": 725
},
{
"epoch": 0.5,
"learning_rate": 9.737894736842107e-06,
"loss": 0.1715,
"step": 750
},
{
"epoch": 0.51,
"learning_rate": 9.711578947368422e-06,
"loss": 0.1639,
"step": 775
},
{
"epoch": 0.53,
"learning_rate": 9.685263157894738e-06,
"loss": 0.1867,
"step": 800
},
{
"epoch": 0.55,
"learning_rate": 9.658947368421053e-06,
"loss": 0.1678,
"step": 825
},
{
"epoch": 0.56,
"learning_rate": 9.63263157894737e-06,
"loss": 0.1758,
"step": 850
},
{
"epoch": 0.58,
"learning_rate": 9.606315789473685e-06,
"loss": 0.1748,
"step": 875
},
{
"epoch": 0.6,
"learning_rate": 9.58e-06,
"loss": 0.165,
"step": 900
},
{
"epoch": 0.61,
"learning_rate": 9.553684210526316e-06,
"loss": 0.1713,
"step": 925
},
{
"epoch": 0.63,
"learning_rate": 9.527368421052631e-06,
"loss": 0.1523,
"step": 950
},
{
"epoch": 0.65,
"learning_rate": 9.501052631578948e-06,
"loss": 0.151,
"step": 975
},
{
"epoch": 0.66,
"learning_rate": 9.474736842105265e-06,
"loss": 0.173,
"step": 1000
},
{
"epoch": 0.66,
"eval_loss": 0.165363147854805,
"eval_runtime": 974.6629,
"eval_samples_per_second": 3.712,
"eval_steps_per_second": 0.233,
"eval_wer": 9.87734021949645,
"step": 1000
},
{
"epoch": 0.68,
"learning_rate": 9.44842105263158e-06,
"loss": 0.1652,
"step": 1025
},
{
"epoch": 0.69,
"learning_rate": 9.422105263157896e-06,
"loss": 0.1556,
"step": 1050
},
{
"epoch": 0.71,
"learning_rate": 9.395789473684211e-06,
"loss": 0.1588,
"step": 1075
},
{
"epoch": 0.73,
"learning_rate": 9.369473684210528e-06,
"loss": 0.1525,
"step": 1100
},
{
"epoch": 0.74,
"learning_rate": 9.343157894736843e-06,
"loss": 0.1345,
"step": 1125
},
{
"epoch": 0.76,
"learning_rate": 9.316842105263158e-06,
"loss": 0.1747,
"step": 1150
},
{
"epoch": 0.78,
"learning_rate": 9.290526315789475e-06,
"loss": 0.166,
"step": 1175
},
{
"epoch": 0.79,
"learning_rate": 9.26421052631579e-06,
"loss": 0.1695,
"step": 1200
},
{
"epoch": 0.81,
"learning_rate": 9.237894736842106e-06,
"loss": 0.1369,
"step": 1225
},
{
"epoch": 0.83,
"learning_rate": 9.211578947368421e-06,
"loss": 0.1507,
"step": 1250
},
{
"epoch": 0.84,
"learning_rate": 9.185263157894738e-06,
"loss": 0.1495,
"step": 1275
},
{
"epoch": 0.86,
"learning_rate": 9.158947368421055e-06,
"loss": 0.1618,
"step": 1300
},
{
"epoch": 0.88,
"learning_rate": 9.13263157894737e-06,
"loss": 0.1722,
"step": 1325
},
{
"epoch": 0.89,
"learning_rate": 9.106315789473684e-06,
"loss": 0.1539,
"step": 1350
},
{
"epoch": 0.91,
"learning_rate": 9.080000000000001e-06,
"loss": 0.1276,
"step": 1375
},
{
"epoch": 0.93,
"learning_rate": 9.053684210526316e-06,
"loss": 0.146,
"step": 1400
},
{
"epoch": 0.94,
"learning_rate": 9.027368421052633e-06,
"loss": 0.1508,
"step": 1425
},
{
"epoch": 0.96,
"learning_rate": 9.001052631578948e-06,
"loss": 0.137,
"step": 1450
},
{
"epoch": 0.98,
"learning_rate": 8.974736842105264e-06,
"loss": 0.1345,
"step": 1475
},
{
"epoch": 0.99,
"learning_rate": 8.94842105263158e-06,
"loss": 0.1279,
"step": 1500
},
{
"epoch": 1.01,
"learning_rate": 8.922105263157896e-06,
"loss": 0.0942,
"step": 1525
},
{
"epoch": 1.03,
"learning_rate": 8.895789473684211e-06,
"loss": 0.0657,
"step": 1550
},
{
"epoch": 1.04,
"learning_rate": 8.869473684210526e-06,
"loss": 0.0814,
"step": 1575
},
{
"epoch": 1.06,
"learning_rate": 8.843157894736843e-06,
"loss": 0.071,
"step": 1600
},
{
"epoch": 1.08,
"learning_rate": 8.816842105263158e-06,
"loss": 0.0655,
"step": 1625
},
{
"epoch": 1.09,
"learning_rate": 8.790526315789474e-06,
"loss": 0.0756,
"step": 1650
},
{
"epoch": 1.11,
"learning_rate": 8.764210526315791e-06,
"loss": 0.0685,
"step": 1675
},
{
"epoch": 1.13,
"learning_rate": 8.737894736842106e-06,
"loss": 0.0676,
"step": 1700
},
{
"epoch": 1.14,
"learning_rate": 8.711578947368423e-06,
"loss": 0.0818,
"step": 1725
},
{
"epoch": 1.16,
"learning_rate": 8.685263157894738e-06,
"loss": 0.0632,
"step": 1750
},
{
"epoch": 1.17,
"learning_rate": 8.658947368421054e-06,
"loss": 0.0537,
"step": 1775
},
{
"epoch": 1.19,
"learning_rate": 8.63263157894737e-06,
"loss": 0.0692,
"step": 1800
},
{
"epoch": 1.21,
"learning_rate": 8.606315789473684e-06,
"loss": 0.0763,
"step": 1825
},
{
"epoch": 1.22,
"learning_rate": 8.580000000000001e-06,
"loss": 0.0729,
"step": 1850
},
{
"epoch": 1.24,
"learning_rate": 8.553684210526316e-06,
"loss": 0.0731,
"step": 1875
},
{
"epoch": 1.26,
"learning_rate": 8.527368421052632e-06,
"loss": 0.0591,
"step": 1900
},
{
"epoch": 1.27,
"learning_rate": 8.501052631578947e-06,
"loss": 0.0717,
"step": 1925
},
{
"epoch": 1.29,
"learning_rate": 8.474736842105264e-06,
"loss": 0.0715,
"step": 1950
},
{
"epoch": 1.31,
"learning_rate": 8.44842105263158e-06,
"loss": 0.0728,
"step": 1975
},
{
"epoch": 1.32,
"learning_rate": 8.422105263157896e-06,
"loss": 0.0771,
"step": 2000
},
{
"epoch": 1.32,
"eval_loss": 0.12896065413951874,
"eval_runtime": 942.4009,
"eval_samples_per_second": 3.839,
"eval_steps_per_second": 0.241,
"eval_wer": 7.751544775431153,
"step": 2000
},
{
"epoch": 1.34,
"learning_rate": 8.395789473684212e-06,
"loss": 0.0771,
"step": 2025
},
{
"epoch": 1.36,
"learning_rate": 8.369473684210527e-06,
"loss": 0.0653,
"step": 2050
},
{
"epoch": 1.37,
"learning_rate": 8.343157894736842e-06,
"loss": 0.0704,
"step": 2075
},
{
"epoch": 1.39,
"learning_rate": 8.316842105263159e-06,
"loss": 0.0691,
"step": 2100
},
{
"epoch": 1.41,
"learning_rate": 8.290526315789474e-06,
"loss": 0.0879,
"step": 2125
},
{
"epoch": 1.42,
"learning_rate": 8.26421052631579e-06,
"loss": 0.0655,
"step": 2150
},
{
"epoch": 1.44,
"learning_rate": 8.237894736842106e-06,
"loss": 0.0793,
"step": 2175
},
{
"epoch": 1.46,
"learning_rate": 8.211578947368422e-06,
"loss": 0.0689,
"step": 2200
},
{
"epoch": 1.47,
"learning_rate": 8.185263157894737e-06,
"loss": 0.0604,
"step": 2225
},
{
"epoch": 1.49,
"learning_rate": 8.158947368421052e-06,
"loss": 0.0749,
"step": 2250
},
{
"epoch": 1.51,
"learning_rate": 8.132631578947369e-06,
"loss": 0.0712,
"step": 2275
},
{
"epoch": 1.52,
"learning_rate": 8.106315789473684e-06,
"loss": 0.0719,
"step": 2300
},
{
"epoch": 1.54,
"learning_rate": 8.08e-06,
"loss": 0.0752,
"step": 2325
},
{
"epoch": 1.56,
"learning_rate": 8.053684210526317e-06,
"loss": 0.0815,
"step": 2350
},
{
"epoch": 1.57,
"learning_rate": 8.027368421052632e-06,
"loss": 0.0749,
"step": 2375
},
{
"epoch": 1.59,
"learning_rate": 8.001052631578949e-06,
"loss": 0.0612,
"step": 2400
},
{
"epoch": 1.6,
"learning_rate": 7.974736842105264e-06,
"loss": 0.0774,
"step": 2425
},
{
"epoch": 1.62,
"learning_rate": 7.94842105263158e-06,
"loss": 0.0652,
"step": 2450
},
{
"epoch": 1.64,
"learning_rate": 7.922105263157895e-06,
"loss": 0.0757,
"step": 2475
},
{
"epoch": 1.65,
"learning_rate": 7.89578947368421e-06,
"loss": 0.0686,
"step": 2500
},
{
"epoch": 1.67,
"learning_rate": 7.869473684210527e-06,
"loss": 0.0626,
"step": 2525
},
{
"epoch": 1.69,
"learning_rate": 7.843157894736842e-06,
"loss": 0.0659,
"step": 2550
},
{
"epoch": 1.7,
"learning_rate": 7.816842105263159e-06,
"loss": 0.0773,
"step": 2575
},
{
"epoch": 1.72,
"learning_rate": 7.790526315789474e-06,
"loss": 0.0705,
"step": 2600
},
{
"epoch": 1.74,
"learning_rate": 7.76421052631579e-06,
"loss": 0.0672,
"step": 2625
},
{
"epoch": 1.75,
"learning_rate": 7.737894736842105e-06,
"loss": 0.0626,
"step": 2650
},
{
"epoch": 1.77,
"learning_rate": 7.711578947368422e-06,
"loss": 0.0753,
"step": 2675
},
{
"epoch": 1.79,
"learning_rate": 7.685263157894739e-06,
"loss": 0.0735,
"step": 2700
},
{
"epoch": 1.8,
"learning_rate": 7.658947368421054e-06,
"loss": 0.0588,
"step": 2725
},
{
"epoch": 1.82,
"learning_rate": 7.632631578947369e-06,
"loss": 0.0633,
"step": 2750
},
{
"epoch": 1.84,
"learning_rate": 7.606315789473685e-06,
"loss": 0.0689,
"step": 2775
},
{
"epoch": 1.85,
"learning_rate": 7.58e-06,
"loss": 0.0591,
"step": 2800
},
{
"epoch": 1.87,
"learning_rate": 7.553684210526316e-06,
"loss": 0.0653,
"step": 2825
},
{
"epoch": 1.89,
"learning_rate": 7.527368421052632e-06,
"loss": 0.0658,
"step": 2850
},
{
"epoch": 1.9,
"learning_rate": 7.501052631578948e-06,
"loss": 0.0739,
"step": 2875
},
{
"epoch": 1.92,
"learning_rate": 7.4747368421052635e-06,
"loss": 0.0569,
"step": 2900
},
{
"epoch": 1.94,
"learning_rate": 7.448421052631579e-06,
"loss": 0.0565,
"step": 2925
},
{
"epoch": 1.95,
"learning_rate": 7.422105263157895e-06,
"loss": 0.0754,
"step": 2950
},
{
"epoch": 1.97,
"learning_rate": 7.395789473684211e-06,
"loss": 0.0551,
"step": 2975
},
{
"epoch": 1.99,
"learning_rate": 7.369473684210528e-06,
"loss": 0.0569,
"step": 3000
},
{
"epoch": 1.99,
"eval_loss": 0.1056058332324028,
"eval_runtime": 968.944,
"eval_samples_per_second": 3.734,
"eval_steps_per_second": 0.234,
"eval_wer": 7.147468412800885,
"step": 3000
},
{
"epoch": 2.0,
"learning_rate": 7.3431578947368435e-06,
"loss": 0.0685,
"step": 3025
},
{
"epoch": 2.02,
"learning_rate": 7.3168421052631585e-06,
"loss": 0.0293,
"step": 3050
},
{
"epoch": 2.04,
"learning_rate": 7.290526315789474e-06,
"loss": 0.0275,
"step": 3075
},
{
"epoch": 2.05,
"learning_rate": 7.26421052631579e-06,
"loss": 0.0251,
"step": 3100
},
{
"epoch": 2.07,
"learning_rate": 7.237894736842106e-06,
"loss": 0.0218,
"step": 3125
},
{
"epoch": 2.08,
"learning_rate": 7.211578947368422e-06,
"loss": 0.0284,
"step": 3150
},
{
"epoch": 2.1,
"learning_rate": 7.1852631578947375e-06,
"loss": 0.0289,
"step": 3175
},
{
"epoch": 2.12,
"learning_rate": 7.158947368421053e-06,
"loss": 0.0236,
"step": 3200
},
{
"epoch": 2.13,
"learning_rate": 7.132631578947369e-06,
"loss": 0.0348,
"step": 3225
},
{
"epoch": 2.15,
"learning_rate": 7.106315789473684e-06,
"loss": 0.028,
"step": 3250
},
{
"epoch": 2.17,
"learning_rate": 7.08e-06,
"loss": 0.0285,
"step": 3275
},
{
"epoch": 2.18,
"learning_rate": 7.053684210526316e-06,
"loss": 0.0221,
"step": 3300
},
{
"epoch": 2.2,
"learning_rate": 7.027368421052632e-06,
"loss": 0.0243,
"step": 3325
},
{
"epoch": 2.22,
"learning_rate": 7.001052631578948e-06,
"loss": 0.0251,
"step": 3350
},
{
"epoch": 2.23,
"learning_rate": 6.974736842105264e-06,
"loss": 0.0261,
"step": 3375
},
{
"epoch": 2.25,
"learning_rate": 6.94842105263158e-06,
"loss": 0.0192,
"step": 3400
},
{
"epoch": 2.27,
"learning_rate": 6.922105263157896e-06,
"loss": 0.0274,
"step": 3425
},
{
"epoch": 2.28,
"learning_rate": 6.8957894736842116e-06,
"loss": 0.0289,
"step": 3450
},
{
"epoch": 2.3,
"learning_rate": 6.869473684210527e-06,
"loss": 0.0226,
"step": 3475
},
{
"epoch": 2.32,
"learning_rate": 6.843157894736842e-06,
"loss": 0.0221,
"step": 3500
},
{
"epoch": 2.33,
"learning_rate": 6.816842105263158e-06,
"loss": 0.0301,
"step": 3525
},
{
"epoch": 2.35,
"learning_rate": 6.790526315789474e-06,
"loss": 0.0251,
"step": 3550
},
{
"epoch": 2.37,
"learning_rate": 6.76421052631579e-06,
"loss": 0.0244,
"step": 3575
},
{
"epoch": 2.38,
"learning_rate": 6.737894736842106e-06,
"loss": 0.0251,
"step": 3600
},
{
"epoch": 2.4,
"learning_rate": 6.7115789473684214e-06,
"loss": 0.0323,
"step": 3625
},
{
"epoch": 2.42,
"learning_rate": 6.685263157894737e-06,
"loss": 0.0254,
"step": 3650
},
{
"epoch": 2.43,
"learning_rate": 6.658947368421054e-06,
"loss": 0.0339,
"step": 3675
},
{
"epoch": 2.45,
"learning_rate": 6.63263157894737e-06,
"loss": 0.0239,
"step": 3700
},
{
"epoch": 2.47,
"learning_rate": 6.606315789473685e-06,
"loss": 0.0279,
"step": 3725
},
{
"epoch": 2.48,
"learning_rate": 6.5800000000000005e-06,
"loss": 0.0358,
"step": 3750
},
{
"epoch": 2.5,
"learning_rate": 6.553684210526316e-06,
"loss": 0.0254,
"step": 3775
},
{
"epoch": 2.51,
"learning_rate": 6.527368421052632e-06,
"loss": 0.0218,
"step": 3800
},
{
"epoch": 2.53,
"learning_rate": 6.501052631578948e-06,
"loss": 0.0239,
"step": 3825
},
{
"epoch": 2.55,
"learning_rate": 6.474736842105264e-06,
"loss": 0.0287,
"step": 3850
},
{
"epoch": 2.56,
"learning_rate": 6.44842105263158e-06,
"loss": 0.0278,
"step": 3875
},
{
"epoch": 2.58,
"learning_rate": 6.4221052631578954e-06,
"loss": 0.0191,
"step": 3900
},
{
"epoch": 2.6,
"learning_rate": 6.39578947368421e-06,
"loss": 0.0222,
"step": 3925
},
{
"epoch": 2.61,
"learning_rate": 6.369473684210526e-06,
"loss": 0.0205,
"step": 3950
},
{
"epoch": 2.63,
"learning_rate": 6.343157894736842e-06,
"loss": 0.0222,
"step": 3975
},
{
"epoch": 2.65,
"learning_rate": 6.316842105263158e-06,
"loss": 0.0274,
"step": 4000
},
{
"epoch": 2.65,
"eval_loss": 0.10442622750997543,
"eval_runtime": 937.0642,
"eval_samples_per_second": 3.861,
"eval_steps_per_second": 0.242,
"eval_wer": 6.626394909157981,
"step": 4000
},
{
"epoch": 2.66,
"learning_rate": 6.2905263157894745e-06,
"loss": 0.0231,
"step": 4025
},
{
"epoch": 2.68,
"learning_rate": 6.26421052631579e-06,
"loss": 0.0206,
"step": 4050
},
{
"epoch": 2.7,
"learning_rate": 6.237894736842106e-06,
"loss": 0.0266,
"step": 4075
},
{
"epoch": 2.71,
"learning_rate": 6.211578947368422e-06,
"loss": 0.0262,
"step": 4100
},
{
"epoch": 2.73,
"learning_rate": 6.185263157894738e-06,
"loss": 0.0229,
"step": 4125
},
{
"epoch": 2.75,
"learning_rate": 6.158947368421054e-06,
"loss": 0.0272,
"step": 4150
},
{
"epoch": 2.76,
"learning_rate": 6.132631578947369e-06,
"loss": 0.0258,
"step": 4175
},
{
"epoch": 2.78,
"learning_rate": 6.106315789473684e-06,
"loss": 0.0218,
"step": 4200
},
{
"epoch": 2.8,
"learning_rate": 6.08e-06,
"loss": 0.0251,
"step": 4225
},
{
"epoch": 2.81,
"learning_rate": 6.053684210526316e-06,
"loss": 0.0242,
"step": 4250
},
{
"epoch": 2.83,
"learning_rate": 6.027368421052632e-06,
"loss": 0.0267,
"step": 4275
},
{
"epoch": 2.85,
"learning_rate": 6.001052631578948e-06,
"loss": 0.0294,
"step": 4300
},
{
"epoch": 2.86,
"learning_rate": 5.9747368421052635e-06,
"loss": 0.0187,
"step": 4325
},
{
"epoch": 2.88,
"learning_rate": 5.94842105263158e-06,
"loss": 0.0208,
"step": 4350
},
{
"epoch": 2.9,
"learning_rate": 5.922105263157896e-06,
"loss": 0.0229,
"step": 4375
},
{
"epoch": 2.91,
"learning_rate": 5.895789473684212e-06,
"loss": 0.0242,
"step": 4400
},
{
"epoch": 2.93,
"learning_rate": 5.869473684210527e-06,
"loss": 0.025,
"step": 4425
},
{
"epoch": 2.95,
"learning_rate": 5.843157894736843e-06,
"loss": 0.0251,
"step": 4450
},
{
"epoch": 2.96,
"learning_rate": 5.816842105263158e-06,
"loss": 0.0321,
"step": 4475
},
{
"epoch": 2.98,
"learning_rate": 5.790526315789474e-06,
"loss": 0.0268,
"step": 4500
},
{
"epoch": 2.99,
"learning_rate": 5.76421052631579e-06,
"loss": 0.0266,
"step": 4525
},
{
"epoch": 3.01,
"learning_rate": 5.737894736842106e-06,
"loss": 0.012,
"step": 4550
},
{
"epoch": 3.03,
"learning_rate": 5.711578947368422e-06,
"loss": 0.0103,
"step": 4575
},
{
"epoch": 3.04,
"learning_rate": 5.685263157894737e-06,
"loss": 0.0093,
"step": 4600
},
{
"epoch": 3.06,
"learning_rate": 5.6589473684210525e-06,
"loss": 0.0096,
"step": 4625
},
{
"epoch": 3.08,
"learning_rate": 5.632631578947368e-06,
"loss": 0.0111,
"step": 4650
},
{
"epoch": 3.09,
"learning_rate": 5.606315789473684e-06,
"loss": 0.0106,
"step": 4675
},
{
"epoch": 3.11,
"learning_rate": 5.580000000000001e-06,
"loss": 0.0117,
"step": 4700
},
{
"epoch": 3.13,
"learning_rate": 5.553684210526317e-06,
"loss": 0.0104,
"step": 4725
},
{
"epoch": 3.14,
"learning_rate": 5.527368421052632e-06,
"loss": 0.0106,
"step": 4750
},
{
"epoch": 3.16,
"learning_rate": 5.501052631578948e-06,
"loss": 0.0131,
"step": 4775
},
{
"epoch": 3.18,
"learning_rate": 5.474736842105264e-06,
"loss": 0.0085,
"step": 4800
},
{
"epoch": 3.19,
"learning_rate": 5.44842105263158e-06,
"loss": 0.0118,
"step": 4825
},
{
"epoch": 3.21,
"learning_rate": 5.422105263157895e-06,
"loss": 0.011,
"step": 4850
},
{
"epoch": 3.23,
"learning_rate": 5.395789473684211e-06,
"loss": 0.0083,
"step": 4875
},
{
"epoch": 3.24,
"learning_rate": 5.3694736842105265e-06,
"loss": 0.0104,
"step": 4900
},
{
"epoch": 3.26,
"learning_rate": 5.343157894736842e-06,
"loss": 0.007,
"step": 4925
},
{
"epoch": 3.28,
"learning_rate": 5.316842105263158e-06,
"loss": 0.0081,
"step": 4950
},
{
"epoch": 3.29,
"learning_rate": 5.290526315789474e-06,
"loss": 0.0104,
"step": 4975
},
{
"epoch": 3.31,
"learning_rate": 5.26421052631579e-06,
"loss": 0.0072,
"step": 5000
},
{
"epoch": 3.31,
"eval_loss": 0.10227037966251373,
"eval_runtime": 938.4381,
"eval_samples_per_second": 3.855,
"eval_steps_per_second": 0.242,
"eval_wer": 6.354329982477175,
"step": 5000
},
{
"epoch": 3.33,
"learning_rate": 5.237894736842105e-06,
"loss": 0.0093,
"step": 5025
},
{
"epoch": 3.34,
"learning_rate": 5.211578947368422e-06,
"loss": 0.0082,
"step": 5050
},
{
"epoch": 3.36,
"learning_rate": 5.185263157894738e-06,
"loss": 0.0124,
"step": 5075
},
{
"epoch": 3.38,
"learning_rate": 5.158947368421053e-06,
"loss": 0.0147,
"step": 5100
},
{
"epoch": 3.39,
"learning_rate": 5.132631578947369e-06,
"loss": 0.0136,
"step": 5125
},
{
"epoch": 3.41,
"learning_rate": 5.106315789473685e-06,
"loss": 0.0088,
"step": 5150
},
{
"epoch": 3.42,
"learning_rate": 5.0800000000000005e-06,
"loss": 0.0104,
"step": 5175
},
{
"epoch": 3.44,
"learning_rate": 5.053684210526316e-06,
"loss": 0.01,
"step": 5200
},
{
"epoch": 3.46,
"learning_rate": 5.027368421052632e-06,
"loss": 0.0095,
"step": 5225
},
{
"epoch": 3.47,
"learning_rate": 5.001052631578948e-06,
"loss": 0.0105,
"step": 5250
},
{
"epoch": 3.49,
"learning_rate": 4.974736842105263e-06,
"loss": 0.0109,
"step": 5275
},
{
"epoch": 3.51,
"learning_rate": 4.94842105263158e-06,
"loss": 0.0074,
"step": 5300
},
{
"epoch": 3.52,
"learning_rate": 4.922105263157895e-06,
"loss": 0.0069,
"step": 5325
},
{
"epoch": 3.54,
"learning_rate": 4.895789473684211e-06,
"loss": 0.0111,
"step": 5350
},
{
"epoch": 3.56,
"learning_rate": 4.869473684210527e-06,
"loss": 0.0069,
"step": 5375
},
{
"epoch": 3.57,
"learning_rate": 4.843157894736842e-06,
"loss": 0.0119,
"step": 5400
},
{
"epoch": 3.59,
"learning_rate": 4.816842105263158e-06,
"loss": 0.0157,
"step": 5425
},
{
"epoch": 3.61,
"learning_rate": 4.790526315789474e-06,
"loss": 0.0099,
"step": 5450
},
{
"epoch": 3.62,
"learning_rate": 4.76421052631579e-06,
"loss": 0.0093,
"step": 5475
},
{
"epoch": 3.64,
"learning_rate": 4.737894736842106e-06,
"loss": 0.0148,
"step": 5500
},
{
"epoch": 3.66,
"learning_rate": 4.711578947368421e-06,
"loss": 0.0092,
"step": 5525
},
{
"epoch": 3.67,
"learning_rate": 4.685263157894737e-06,
"loss": 0.0063,
"step": 5550
},
{
"epoch": 3.69,
"learning_rate": 4.658947368421053e-06,
"loss": 0.0116,
"step": 5575
},
{
"epoch": 3.71,
"learning_rate": 4.6326315789473686e-06,
"loss": 0.0103,
"step": 5600
},
{
"epoch": 3.72,
"learning_rate": 4.606315789473685e-06,
"loss": 0.0153,
"step": 5625
},
{
"epoch": 3.74,
"learning_rate": 4.58e-06,
"loss": 0.0098,
"step": 5650
},
{
"epoch": 3.76,
"learning_rate": 4.553684210526316e-06,
"loss": 0.0101,
"step": 5675
},
{
"epoch": 3.77,
"learning_rate": 4.527368421052632e-06,
"loss": 0.0103,
"step": 5700
},
{
"epoch": 3.79,
"learning_rate": 4.501052631578948e-06,
"loss": 0.0097,
"step": 5725
},
{
"epoch": 3.81,
"learning_rate": 4.4747368421052635e-06,
"loss": 0.011,
"step": 5750
},
{
"epoch": 3.82,
"learning_rate": 4.448421052631579e-06,
"loss": 0.007,
"step": 5775
},
{
"epoch": 3.84,
"learning_rate": 4.422105263157895e-06,
"loss": 0.0055,
"step": 5800
},
{
"epoch": 3.86,
"learning_rate": 4.395789473684211e-06,
"loss": 0.0091,
"step": 5825
},
{
"epoch": 3.87,
"learning_rate": 4.369473684210527e-06,
"loss": 0.0156,
"step": 5850
},
{
"epoch": 3.89,
"learning_rate": 4.3431578947368426e-06,
"loss": 0.0088,
"step": 5875
},
{
"epoch": 3.9,
"learning_rate": 4.316842105263158e-06,
"loss": 0.0105,
"step": 5900
},
{
"epoch": 3.92,
"learning_rate": 4.290526315789474e-06,
"loss": 0.0101,
"step": 5925
},
{
"epoch": 3.94,
"learning_rate": 4.264210526315789e-06,
"loss": 0.0082,
"step": 5950
},
{
"epoch": 3.95,
"learning_rate": 4.237894736842106e-06,
"loss": 0.0086,
"step": 5975
},
{
"epoch": 3.97,
"learning_rate": 4.211578947368422e-06,
"loss": 0.009,
"step": 6000
},
{
"epoch": 3.97,
"eval_loss": 0.10000595450401306,
"eval_runtime": 937.2515,
"eval_samples_per_second": 3.86,
"eval_steps_per_second": 0.242,
"eval_wer": 6.335884902702205,
"step": 6000
},
{
"epoch": 3.99,
"learning_rate": 4.1852631578947375e-06,
"loss": 0.0086,
"step": 6025
},
{
"epoch": 4.0,
"learning_rate": 4.16e-06,
"loss": 0.0079,
"step": 6050
},
{
"epoch": 4.02,
"learning_rate": 4.133684210526316e-06,
"loss": 0.0051,
"step": 6075
},
{
"epoch": 4.04,
"learning_rate": 4.1073684210526325e-06,
"loss": 0.0041,
"step": 6100
},
{
"epoch": 4.05,
"learning_rate": 4.0810526315789474e-06,
"loss": 0.003,
"step": 6125
},
{
"epoch": 4.07,
"learning_rate": 4.054736842105263e-06,
"loss": 0.0045,
"step": 6150
},
{
"epoch": 4.09,
"learning_rate": 4.028421052631579e-06,
"loss": 0.0029,
"step": 6175
},
{
"epoch": 4.1,
"learning_rate": 4.002105263157895e-06,
"loss": 0.0024,
"step": 6200
},
{
"epoch": 4.12,
"learning_rate": 3.975789473684211e-06,
"loss": 0.0066,
"step": 6225
},
{
"epoch": 4.14,
"learning_rate": 3.9494736842105265e-06,
"loss": 0.0032,
"step": 6250
},
{
"epoch": 4.15,
"learning_rate": 3.923157894736842e-06,
"loss": 0.004,
"step": 6275
},
{
"epoch": 4.17,
"learning_rate": 3.896842105263158e-06,
"loss": 0.0047,
"step": 6300
},
{
"epoch": 4.19,
"learning_rate": 3.870526315789474e-06,
"loss": 0.0039,
"step": 6325
},
{
"epoch": 4.2,
"learning_rate": 3.84421052631579e-06,
"loss": 0.004,
"step": 6350
},
{
"epoch": 4.22,
"learning_rate": 3.817894736842106e-06,
"loss": 0.0029,
"step": 6375
},
{
"epoch": 4.24,
"learning_rate": 3.791578947368421e-06,
"loss": 0.0056,
"step": 6400
},
{
"epoch": 4.25,
"learning_rate": 3.7652631578947373e-06,
"loss": 0.0062,
"step": 6425
},
{
"epoch": 4.27,
"learning_rate": 3.738947368421053e-06,
"loss": 0.0071,
"step": 6450
},
{
"epoch": 4.29,
"learning_rate": 3.712631578947369e-06,
"loss": 0.0058,
"step": 6475
},
{
"epoch": 4.3,
"learning_rate": 3.6863157894736847e-06,
"loss": 0.0057,
"step": 6500
},
{
"epoch": 4.32,
"learning_rate": 3.66e-06,
"loss": 0.0055,
"step": 6525
},
{
"epoch": 4.33,
"learning_rate": 3.633684210526316e-06,
"loss": 0.0036,
"step": 6550
},
{
"epoch": 4.35,
"learning_rate": 3.6073684210526317e-06,
"loss": 0.0037,
"step": 6575
},
{
"epoch": 4.37,
"learning_rate": 3.581052631578948e-06,
"loss": 0.0029,
"step": 6600
},
{
"epoch": 4.38,
"learning_rate": 3.554736842105264e-06,
"loss": 0.0027,
"step": 6625
},
{
"epoch": 4.4,
"learning_rate": 3.528421052631579e-06,
"loss": 0.0041,
"step": 6650
},
{
"epoch": 4.42,
"learning_rate": 3.502105263157895e-06,
"loss": 0.0029,
"step": 6675
},
{
"epoch": 4.43,
"learning_rate": 3.475789473684211e-06,
"loss": 0.0024,
"step": 6700
},
{
"epoch": 4.45,
"learning_rate": 3.4494736842105262e-06,
"loss": 0.0029,
"step": 6725
},
{
"epoch": 4.47,
"learning_rate": 3.423157894736842e-06,
"loss": 0.0039,
"step": 6750
},
{
"epoch": 4.48,
"learning_rate": 3.3968421052631583e-06,
"loss": 0.0061,
"step": 6775
},
{
"epoch": 4.5,
"learning_rate": 3.370526315789474e-06,
"loss": 0.0049,
"step": 6800
},
{
"epoch": 4.52,
"learning_rate": 3.34421052631579e-06,
"loss": 0.0054,
"step": 6825
},
{
"epoch": 4.53,
"learning_rate": 3.3178947368421053e-06,
"loss": 0.0028,
"step": 6850
},
{
"epoch": 4.55,
"learning_rate": 3.291578947368421e-06,
"loss": 0.0043,
"step": 6875
},
{
"epoch": 4.57,
"learning_rate": 3.265263157894737e-06,
"loss": 0.0036,
"step": 6900
},
{
"epoch": 4.58,
"learning_rate": 3.238947368421053e-06,
"loss": 0.003,
"step": 6925
},
{
"epoch": 4.6,
"learning_rate": 3.212631578947369e-06,
"loss": 0.0025,
"step": 6950
},
{
"epoch": 4.62,
"learning_rate": 3.1863157894736844e-06,
"loss": 0.0026,
"step": 6975
},
{
"epoch": 4.63,
"learning_rate": 3.1600000000000002e-06,
"loss": 0.0033,
"step": 7000
},
{
"epoch": 4.63,
"eval_loss": 0.10218288749456406,
"eval_runtime": 937.592,
"eval_samples_per_second": 3.859,
"eval_steps_per_second": 0.242,
"eval_wer": 6.059208706077654,
"step": 7000
},
{
"epoch": 4.65,
"learning_rate": 3.133684210526316e-06,
"loss": 0.0023,
"step": 7025
},
{
"epoch": 4.67,
"learning_rate": 3.107368421052632e-06,
"loss": 0.0035,
"step": 7050
},
{
"epoch": 4.68,
"learning_rate": 3.0810526315789473e-06,
"loss": 0.0052,
"step": 7075
},
{
"epoch": 4.7,
"learning_rate": 3.0547368421052635e-06,
"loss": 0.004,
"step": 7100
},
{
"epoch": 4.72,
"learning_rate": 3.0284210526315793e-06,
"loss": 0.0024,
"step": 7125
},
{
"epoch": 4.73,
"learning_rate": 3.002105263157895e-06,
"loss": 0.0064,
"step": 7150
},
{
"epoch": 4.75,
"learning_rate": 2.975789473684211e-06,
"loss": 0.0033,
"step": 7175
},
{
"epoch": 4.77,
"learning_rate": 2.9494736842105264e-06,
"loss": 0.0043,
"step": 7200
},
{
"epoch": 4.78,
"learning_rate": 2.923157894736842e-06,
"loss": 0.0045,
"step": 7225
},
{
"epoch": 4.8,
"learning_rate": 2.896842105263158e-06,
"loss": 0.0034,
"step": 7250
},
{
"epoch": 4.81,
"learning_rate": 2.8705263157894742e-06,
"loss": 0.0056,
"step": 7275
},
{
"epoch": 4.83,
"learning_rate": 2.84421052631579e-06,
"loss": 0.0062,
"step": 7300
},
{
"epoch": 4.85,
"learning_rate": 2.8178947368421055e-06,
"loss": 0.0029,
"step": 7325
},
{
"epoch": 4.86,
"learning_rate": 2.7915789473684213e-06,
"loss": 0.0033,
"step": 7350
},
{
"epoch": 4.88,
"learning_rate": 2.765263157894737e-06,
"loss": 0.0039,
"step": 7375
},
{
"epoch": 4.9,
"learning_rate": 2.7389473684210525e-06,
"loss": 0.0039,
"step": 7400
},
{
"epoch": 4.91,
"learning_rate": 2.7126315789473683e-06,
"loss": 0.0069,
"step": 7425
},
{
"epoch": 4.93,
"learning_rate": 2.6863157894736845e-06,
"loss": 0.0033,
"step": 7450
},
{
"epoch": 4.95,
"learning_rate": 2.6600000000000004e-06,
"loss": 0.0031,
"step": 7475
},
{
"epoch": 4.96,
"learning_rate": 2.633684210526316e-06,
"loss": 0.003,
"step": 7500
},
{
"epoch": 4.98,
"learning_rate": 2.6073684210526316e-06,
"loss": 0.0023,
"step": 7525
},
{
"epoch": 5.0,
"learning_rate": 2.5810526315789474e-06,
"loss": 0.0045,
"step": 7550
},
{
"epoch": 5.01,
"learning_rate": 2.5547368421052632e-06,
"loss": 0.0017,
"step": 7575
},
{
"epoch": 5.03,
"learning_rate": 2.528421052631579e-06,
"loss": 0.0026,
"step": 7600
},
{
"epoch": 5.05,
"learning_rate": 2.5021052631578953e-06,
"loss": 0.0027,
"step": 7625
},
{
"epoch": 5.06,
"learning_rate": 2.4757894736842107e-06,
"loss": 0.0021,
"step": 7650
},
{
"epoch": 5.08,
"learning_rate": 2.4494736842105265e-06,
"loss": 0.0014,
"step": 7675
},
{
"epoch": 5.1,
"learning_rate": 2.4231578947368423e-06,
"loss": 0.0017,
"step": 7700
},
{
"epoch": 5.11,
"learning_rate": 2.396842105263158e-06,
"loss": 0.0023,
"step": 7725
},
{
"epoch": 5.13,
"learning_rate": 2.370526315789474e-06,
"loss": 0.002,
"step": 7750
},
{
"epoch": 5.15,
"learning_rate": 2.3442105263157898e-06,
"loss": 0.0019,
"step": 7775
},
{
"epoch": 5.16,
"learning_rate": 2.317894736842105e-06,
"loss": 0.0034,
"step": 7800
},
{
"epoch": 5.18,
"learning_rate": 2.2915789473684214e-06,
"loss": 0.0014,
"step": 7825
},
{
"epoch": 5.2,
"learning_rate": 2.2652631578947372e-06,
"loss": 0.0034,
"step": 7850
},
{
"epoch": 5.21,
"learning_rate": 2.2389473684210526e-06,
"loss": 0.0018,
"step": 7875
},
{
"epoch": 5.23,
"learning_rate": 2.2126315789473684e-06,
"loss": 0.0016,
"step": 7900
},
{
"epoch": 5.24,
"learning_rate": 2.1863157894736843e-06,
"loss": 0.0017,
"step": 7925
},
{
"epoch": 5.26,
"learning_rate": 2.16e-06,
"loss": 0.0011,
"step": 7950
},
{
"epoch": 5.28,
"learning_rate": 2.133684210526316e-06,
"loss": 0.0015,
"step": 7975
},
{
"epoch": 5.29,
"learning_rate": 2.1073684210526317e-06,
"loss": 0.002,
"step": 8000
},
{
"epoch": 5.29,
"eval_loss": 0.10514529049396515,
"eval_runtime": 939.1907,
"eval_samples_per_second": 3.852,
"eval_steps_per_second": 0.242,
"eval_wer": 6.156045374896246,
"step": 8000
},
{
"epoch": 5.31,
"learning_rate": 2.0810526315789475e-06,
"loss": 0.0014,
"step": 8025
},
{
"epoch": 5.33,
"learning_rate": 2.0547368421052633e-06,
"loss": 0.0024,
"step": 8050
},
{
"epoch": 5.34,
"learning_rate": 2.028421052631579e-06,
"loss": 0.0011,
"step": 8075
},
{
"epoch": 5.36,
"learning_rate": 2.002105263157895e-06,
"loss": 0.0015,
"step": 8100
},
{
"epoch": 5.38,
"learning_rate": 1.975789473684211e-06,
"loss": 0.0027,
"step": 8125
},
{
"epoch": 5.39,
"learning_rate": 1.949473684210526e-06,
"loss": 0.001,
"step": 8150
},
{
"epoch": 5.41,
"learning_rate": 1.9231578947368424e-06,
"loss": 0.0022,
"step": 8175
},
{
"epoch": 5.43,
"learning_rate": 1.896842105263158e-06,
"loss": 0.0018,
"step": 8200
},
{
"epoch": 5.44,
"learning_rate": 1.8705263157894737e-06,
"loss": 0.0024,
"step": 8225
},
{
"epoch": 5.46,
"learning_rate": 1.8442105263157897e-06,
"loss": 0.0021,
"step": 8250
},
{
"epoch": 5.48,
"learning_rate": 1.8178947368421055e-06,
"loss": 0.0021,
"step": 8275
},
{
"epoch": 5.49,
"learning_rate": 1.7915789473684211e-06,
"loss": 0.0014,
"step": 8300
},
{
"epoch": 5.51,
"learning_rate": 1.7652631578947371e-06,
"loss": 0.0018,
"step": 8325
},
{
"epoch": 5.53,
"learning_rate": 1.7389473684210527e-06,
"loss": 0.0016,
"step": 8350
},
{
"epoch": 5.54,
"learning_rate": 1.7126315789473686e-06,
"loss": 0.001,
"step": 8375
},
{
"epoch": 5.56,
"learning_rate": 1.6863157894736842e-06,
"loss": 0.0039,
"step": 8400
},
{
"epoch": 5.58,
"learning_rate": 1.6610526315789473e-06,
"loss": 0.0022,
"step": 8425
},
{
"epoch": 5.59,
"learning_rate": 1.6347368421052633e-06,
"loss": 0.0019,
"step": 8450
},
{
"epoch": 5.61,
"learning_rate": 1.6084210526315792e-06,
"loss": 0.0019,
"step": 8475
},
{
"epoch": 5.63,
"learning_rate": 1.5821052631578948e-06,
"loss": 0.0016,
"step": 8500
},
{
"epoch": 5.64,
"learning_rate": 1.5557894736842106e-06,
"loss": 0.0033,
"step": 8525
},
{
"epoch": 5.66,
"learning_rate": 1.5294736842105264e-06,
"loss": 0.0013,
"step": 8550
},
{
"epoch": 5.68,
"learning_rate": 1.5031578947368422e-06,
"loss": 0.0016,
"step": 8575
},
{
"epoch": 5.69,
"learning_rate": 1.4768421052631578e-06,
"loss": 0.0025,
"step": 8600
},
{
"epoch": 5.71,
"learning_rate": 1.4505263157894739e-06,
"loss": 0.0014,
"step": 8625
},
{
"epoch": 5.72,
"learning_rate": 1.4242105263157897e-06,
"loss": 0.0012,
"step": 8650
},
{
"epoch": 5.74,
"learning_rate": 1.3978947368421053e-06,
"loss": 0.0026,
"step": 8675
},
{
"epoch": 5.76,
"learning_rate": 1.3715789473684213e-06,
"loss": 0.0013,
"step": 8700
},
{
"epoch": 5.77,
"learning_rate": 1.345263157894737e-06,
"loss": 0.0024,
"step": 8725
},
{
"epoch": 5.79,
"learning_rate": 1.3189473684210527e-06,
"loss": 0.0017,
"step": 8750
},
{
"epoch": 5.81,
"learning_rate": 1.2926315789473683e-06,
"loss": 0.0023,
"step": 8775
},
{
"epoch": 5.82,
"learning_rate": 1.2663157894736844e-06,
"loss": 0.0014,
"step": 8800
},
{
"epoch": 5.84,
"learning_rate": 1.2400000000000002e-06,
"loss": 0.0018,
"step": 8825
},
{
"epoch": 5.86,
"learning_rate": 1.213684210526316e-06,
"loss": 0.0022,
"step": 8850
},
{
"epoch": 5.87,
"learning_rate": 1.1873684210526316e-06,
"loss": 0.0022,
"step": 8875
},
{
"epoch": 5.89,
"learning_rate": 1.1610526315789474e-06,
"loss": 0.0013,
"step": 8900
},
{
"epoch": 5.91,
"learning_rate": 1.1347368421052633e-06,
"loss": 0.0025,
"step": 8925
},
{
"epoch": 5.92,
"learning_rate": 1.108421052631579e-06,
"loss": 0.002,
"step": 8950
},
{
"epoch": 5.94,
"learning_rate": 1.082105263157895e-06,
"loss": 0.001,
"step": 8975
},
{
"epoch": 5.96,
"learning_rate": 1.0557894736842105e-06,
"loss": 0.0028,
"step": 9000
},
{
"epoch": 5.96,
"eval_loss": 0.10519474744796753,
"eval_runtime": 938.0428,
"eval_samples_per_second": 3.857,
"eval_steps_per_second": 0.242,
"eval_wer": 6.100710135571337,
"step": 9000
},
{
"epoch": 5.97,
"learning_rate": 1.0294736842105265e-06,
"loss": 0.0012,
"step": 9025
},
{
"epoch": 5.99,
"learning_rate": 1.0031578947368421e-06,
"loss": 0.0022,
"step": 9050
},
{
"epoch": 6.01,
"learning_rate": 9.76842105263158e-07,
"loss": 0.001,
"step": 9075
},
{
"epoch": 6.02,
"learning_rate": 9.505263157894738e-07,
"loss": 0.0013,
"step": 9100
},
{
"epoch": 6.04,
"learning_rate": 9.242105263157895e-07,
"loss": 0.001,
"step": 9125
},
{
"epoch": 6.06,
"learning_rate": 8.978947368421054e-07,
"loss": 0.0021,
"step": 9150
},
{
"epoch": 6.07,
"learning_rate": 8.715789473684211e-07,
"loss": 0.0011,
"step": 9175
},
{
"epoch": 6.09,
"learning_rate": 8.452631578947369e-07,
"loss": 0.0011,
"step": 9200
},
{
"epoch": 6.11,
"learning_rate": 8.189473684210527e-07,
"loss": 0.0009,
"step": 9225
},
{
"epoch": 6.12,
"learning_rate": 7.926315789473685e-07,
"loss": 0.0008,
"step": 9250
},
{
"epoch": 6.14,
"learning_rate": 7.663157894736843e-07,
"loss": 0.0015,
"step": 9275
},
{
"epoch": 6.15,
"learning_rate": 7.4e-07,
"loss": 0.0013,
"step": 9300
},
{
"epoch": 6.17,
"learning_rate": 7.136842105263158e-07,
"loss": 0.0008,
"step": 9325
},
{
"epoch": 6.19,
"learning_rate": 6.873684210526315e-07,
"loss": 0.0012,
"step": 9350
},
{
"epoch": 6.2,
"learning_rate": 6.610526315789475e-07,
"loss": 0.0011,
"step": 9375
},
{
"epoch": 6.22,
"learning_rate": 6.347368421052633e-07,
"loss": 0.0009,
"step": 9400
},
{
"epoch": 6.24,
"learning_rate": 6.08421052631579e-07,
"loss": 0.0009,
"step": 9425
},
{
"epoch": 6.25,
"learning_rate": 5.821052631578948e-07,
"loss": 0.0017,
"step": 9450
},
{
"epoch": 6.27,
"learning_rate": 5.557894736842106e-07,
"loss": 0.0013,
"step": 9475
},
{
"epoch": 6.29,
"learning_rate": 5.294736842105263e-07,
"loss": 0.0011,
"step": 9500
},
{
"epoch": 6.3,
"learning_rate": 5.031578947368422e-07,
"loss": 0.0014,
"step": 9525
},
{
"epoch": 6.32,
"learning_rate": 4.768421052631579e-07,
"loss": 0.0014,
"step": 9550
},
{
"epoch": 6.34,
"learning_rate": 4.505263157894737e-07,
"loss": 0.0011,
"step": 9575
},
{
"epoch": 6.35,
"learning_rate": 4.242105263157895e-07,
"loss": 0.0012,
"step": 9600
},
{
"epoch": 6.37,
"learning_rate": 3.9789473684210533e-07,
"loss": 0.0013,
"step": 9625
},
{
"epoch": 6.39,
"learning_rate": 3.715789473684211e-07,
"loss": 0.0013,
"step": 9650
},
{
"epoch": 6.4,
"learning_rate": 3.4526315789473686e-07,
"loss": 0.0015,
"step": 9675
},
{
"epoch": 6.42,
"learning_rate": 3.1894736842105263e-07,
"loss": 0.0015,
"step": 9700
},
{
"epoch": 6.44,
"learning_rate": 2.9263157894736845e-07,
"loss": 0.001,
"step": 9725
},
{
"epoch": 6.45,
"learning_rate": 2.663157894736842e-07,
"loss": 0.001,
"step": 9750
},
{
"epoch": 6.47,
"learning_rate": 2.4000000000000003e-07,
"loss": 0.0011,
"step": 9775
},
{
"epoch": 6.49,
"learning_rate": 2.136842105263158e-07,
"loss": 0.0018,
"step": 9800
},
{
"epoch": 6.5,
"learning_rate": 1.8736842105263162e-07,
"loss": 0.0008,
"step": 9825
},
{
"epoch": 6.52,
"learning_rate": 1.6105263157894738e-07,
"loss": 0.0013,
"step": 9850
},
{
"epoch": 6.54,
"learning_rate": 1.3473684210526317e-07,
"loss": 0.0013,
"step": 9875
},
{
"epoch": 6.55,
"learning_rate": 1.0842105263157895e-07,
"loss": 0.0009,
"step": 9900
},
{
"epoch": 6.57,
"learning_rate": 8.210526315789474e-08,
"loss": 0.0011,
"step": 9925
},
{
"epoch": 6.59,
"learning_rate": 5.578947368421053e-08,
"loss": 0.0012,
"step": 9950
},
{
"epoch": 6.6,
"learning_rate": 2.9473684210526318e-08,
"loss": 0.001,
"step": 9975
},
{
"epoch": 6.62,
"learning_rate": 3.1578947368421054e-09,
"loss": 0.0013,
"step": 10000
},
{
"epoch": 6.62,
"eval_loss": 0.10633327811956406,
"eval_runtime": 938.2307,
"eval_samples_per_second": 3.856,
"eval_steps_per_second": 0.242,
"eval_wer": 6.137600295121277,
"step": 10000
},
{
"epoch": 6.62,
"step": 10000,
"total_flos": 9.231962390396928e+19,
"train_loss": 0.04964755083955824,
"train_runtime": 49853.2971,
"train_samples_per_second": 6.419,
"train_steps_per_second": 0.201
}
],
"max_steps": 10000,
"num_train_epochs": 7,
"total_flos": 9.231962390396928e+19,
"trial_name": null,
"trial_params": null
}