whisper-small-marathi / trainer_state.json
steja's picture
marathi whisper small ft
0cbb265
raw
history blame
No virus
25.7 kB
{
"best_metric": 18.578158046529513,
"best_model_checkpoint": "./whisper-small-marathi/checkpoint-3000",
"epoch": 94.33962264150944,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.47,
"learning_rate": 4.6000000000000004e-07,
"loss": 1.1274,
"step": 25
},
{
"epoch": 0.94,
"learning_rate": 9.600000000000001e-07,
"loss": 0.9372,
"step": 50
},
{
"epoch": 1.42,
"learning_rate": 1.46e-06,
"loss": 0.6953,
"step": 75
},
{
"epoch": 1.89,
"learning_rate": 1.9600000000000003e-06,
"loss": 0.5522,
"step": 100
},
{
"epoch": 2.36,
"learning_rate": 2.46e-06,
"loss": 0.4549,
"step": 125
},
{
"epoch": 2.83,
"learning_rate": 2.96e-06,
"loss": 0.4127,
"step": 150
},
{
"epoch": 3.3,
"learning_rate": 3.46e-06,
"loss": 0.3451,
"step": 175
},
{
"epoch": 3.77,
"learning_rate": 3.96e-06,
"loss": 0.315,
"step": 200
},
{
"epoch": 4.25,
"learning_rate": 4.4600000000000005e-06,
"loss": 0.2816,
"step": 225
},
{
"epoch": 4.72,
"learning_rate": 4.960000000000001e-06,
"loss": 0.248,
"step": 250
},
{
"epoch": 5.19,
"learning_rate": 5.460000000000001e-06,
"loss": 0.2281,
"step": 275
},
{
"epoch": 5.66,
"learning_rate": 5.9600000000000005e-06,
"loss": 0.1972,
"step": 300
},
{
"epoch": 6.13,
"learning_rate": 6.460000000000001e-06,
"loss": 0.1764,
"step": 325
},
{
"epoch": 6.6,
"learning_rate": 6.96e-06,
"loss": 0.145,
"step": 350
},
{
"epoch": 7.08,
"learning_rate": 7.4600000000000006e-06,
"loss": 0.1403,
"step": 375
},
{
"epoch": 7.55,
"learning_rate": 7.960000000000002e-06,
"loss": 0.1011,
"step": 400
},
{
"epoch": 8.02,
"learning_rate": 8.46e-06,
"loss": 0.099,
"step": 425
},
{
"epoch": 8.49,
"learning_rate": 8.96e-06,
"loss": 0.0678,
"step": 450
},
{
"epoch": 8.96,
"learning_rate": 9.460000000000001e-06,
"loss": 0.0672,
"step": 475
},
{
"epoch": 9.43,
"learning_rate": 9.960000000000001e-06,
"loss": 0.0436,
"step": 500
},
{
"epoch": 9.91,
"learning_rate": 9.94888888888889e-06,
"loss": 0.044,
"step": 525
},
{
"epoch": 10.38,
"learning_rate": 9.893333333333334e-06,
"loss": 0.0311,
"step": 550
},
{
"epoch": 10.85,
"learning_rate": 9.837777777777778e-06,
"loss": 0.0255,
"step": 575
},
{
"epoch": 11.32,
"learning_rate": 9.782222222222222e-06,
"loss": 0.0198,
"step": 600
},
{
"epoch": 11.79,
"learning_rate": 9.726666666666668e-06,
"loss": 0.0169,
"step": 625
},
{
"epoch": 12.26,
"learning_rate": 9.671111111111112e-06,
"loss": 0.0136,
"step": 650
},
{
"epoch": 12.74,
"learning_rate": 9.615555555555558e-06,
"loss": 0.0104,
"step": 675
},
{
"epoch": 13.21,
"learning_rate": 9.56e-06,
"loss": 0.0096,
"step": 700
},
{
"epoch": 13.68,
"learning_rate": 9.504444444444446e-06,
"loss": 0.0096,
"step": 725
},
{
"epoch": 14.15,
"learning_rate": 9.44888888888889e-06,
"loss": 0.0079,
"step": 750
},
{
"epoch": 14.62,
"learning_rate": 9.393333333333334e-06,
"loss": 0.0065,
"step": 775
},
{
"epoch": 15.09,
"learning_rate": 9.33777777777778e-06,
"loss": 0.0066,
"step": 800
},
{
"epoch": 15.57,
"learning_rate": 9.282222222222222e-06,
"loss": 0.005,
"step": 825
},
{
"epoch": 16.04,
"learning_rate": 9.226666666666668e-06,
"loss": 0.0052,
"step": 850
},
{
"epoch": 16.51,
"learning_rate": 9.171111111111112e-06,
"loss": 0.0037,
"step": 875
},
{
"epoch": 16.98,
"learning_rate": 9.115555555555556e-06,
"loss": 0.0043,
"step": 900
},
{
"epoch": 17.45,
"learning_rate": 9.060000000000001e-06,
"loss": 0.0033,
"step": 925
},
{
"epoch": 17.92,
"learning_rate": 9.004444444444445e-06,
"loss": 0.004,
"step": 950
},
{
"epoch": 18.4,
"learning_rate": 8.94888888888889e-06,
"loss": 0.0039,
"step": 975
},
{
"epoch": 18.87,
"learning_rate": 8.893333333333333e-06,
"loss": 0.0027,
"step": 1000
},
{
"epoch": 18.87,
"eval_loss": 0.5089536309242249,
"eval_runtime": 506.9498,
"eval_samples_per_second": 3.582,
"eval_steps_per_second": 0.057,
"eval_wer": 20.032205345125938,
"step": 1000
},
{
"epoch": 19.34,
"learning_rate": 8.83777777777778e-06,
"loss": 0.003,
"step": 1025
},
{
"epoch": 19.81,
"learning_rate": 8.782222222222223e-06,
"loss": 0.003,
"step": 1050
},
{
"epoch": 20.28,
"learning_rate": 8.726666666666667e-06,
"loss": 0.0029,
"step": 1075
},
{
"epoch": 20.75,
"learning_rate": 8.671111111111113e-06,
"loss": 0.0026,
"step": 1100
},
{
"epoch": 21.23,
"learning_rate": 8.615555555555555e-06,
"loss": 0.0028,
"step": 1125
},
{
"epoch": 21.7,
"learning_rate": 8.560000000000001e-06,
"loss": 0.002,
"step": 1150
},
{
"epoch": 22.17,
"learning_rate": 8.504444444444445e-06,
"loss": 0.0019,
"step": 1175
},
{
"epoch": 22.64,
"learning_rate": 8.448888888888889e-06,
"loss": 0.0023,
"step": 1200
},
{
"epoch": 23.11,
"learning_rate": 8.393333333333335e-06,
"loss": 0.0025,
"step": 1225
},
{
"epoch": 23.58,
"learning_rate": 8.337777777777777e-06,
"loss": 0.0017,
"step": 1250
},
{
"epoch": 24.06,
"learning_rate": 8.282222222222223e-06,
"loss": 0.0021,
"step": 1275
},
{
"epoch": 24.53,
"learning_rate": 8.226666666666667e-06,
"loss": 0.0022,
"step": 1300
},
{
"epoch": 25.0,
"learning_rate": 8.171111111111113e-06,
"loss": 0.0015,
"step": 1325
},
{
"epoch": 25.47,
"learning_rate": 8.115555555555557e-06,
"loss": 0.0013,
"step": 1350
},
{
"epoch": 25.94,
"learning_rate": 8.06e-06,
"loss": 0.0012,
"step": 1375
},
{
"epoch": 26.42,
"learning_rate": 8.004444444444445e-06,
"loss": 0.0012,
"step": 1400
},
{
"epoch": 26.89,
"learning_rate": 7.948888888888889e-06,
"loss": 0.0012,
"step": 1425
},
{
"epoch": 27.36,
"learning_rate": 7.893333333333335e-06,
"loss": 0.0011,
"step": 1450
},
{
"epoch": 27.83,
"learning_rate": 7.837777777777779e-06,
"loss": 0.0008,
"step": 1475
},
{
"epoch": 28.3,
"learning_rate": 7.782222222222223e-06,
"loss": 0.0004,
"step": 1500
},
{
"epoch": 28.77,
"learning_rate": 7.726666666666667e-06,
"loss": 0.0005,
"step": 1525
},
{
"epoch": 29.25,
"learning_rate": 7.67111111111111e-06,
"loss": 0.0012,
"step": 1550
},
{
"epoch": 29.72,
"learning_rate": 7.6155555555555564e-06,
"loss": 0.0007,
"step": 1575
},
{
"epoch": 30.19,
"learning_rate": 7.5600000000000005e-06,
"loss": 0.0013,
"step": 1600
},
{
"epoch": 30.66,
"learning_rate": 7.504444444444445e-06,
"loss": 0.0018,
"step": 1625
},
{
"epoch": 31.13,
"learning_rate": 7.44888888888889e-06,
"loss": 0.0018,
"step": 1650
},
{
"epoch": 31.6,
"learning_rate": 7.393333333333333e-06,
"loss": 0.0015,
"step": 1675
},
{
"epoch": 32.08,
"learning_rate": 7.337777777777778e-06,
"loss": 0.0019,
"step": 1700
},
{
"epoch": 32.55,
"learning_rate": 7.282222222222222e-06,
"loss": 0.0012,
"step": 1725
},
{
"epoch": 33.02,
"learning_rate": 7.226666666666667e-06,
"loss": 0.0016,
"step": 1750
},
{
"epoch": 33.49,
"learning_rate": 7.171111111111112e-06,
"loss": 0.0012,
"step": 1775
},
{
"epoch": 33.96,
"learning_rate": 7.115555555555557e-06,
"loss": 0.0011,
"step": 1800
},
{
"epoch": 34.43,
"learning_rate": 7.06e-06,
"loss": 0.0009,
"step": 1825
},
{
"epoch": 34.91,
"learning_rate": 7.004444444444445e-06,
"loss": 0.0009,
"step": 1850
},
{
"epoch": 35.38,
"learning_rate": 6.948888888888889e-06,
"loss": 0.0007,
"step": 1875
},
{
"epoch": 35.85,
"learning_rate": 6.893333333333334e-06,
"loss": 0.0011,
"step": 1900
},
{
"epoch": 36.32,
"learning_rate": 6.837777777777779e-06,
"loss": 0.0009,
"step": 1925
},
{
"epoch": 36.79,
"learning_rate": 6.782222222222222e-06,
"loss": 0.0009,
"step": 1950
},
{
"epoch": 37.26,
"learning_rate": 6.726666666666667e-06,
"loss": 0.0007,
"step": 1975
},
{
"epoch": 37.74,
"learning_rate": 6.671111111111112e-06,
"loss": 0.0007,
"step": 2000
},
{
"epoch": 37.74,
"eval_loss": 0.5729004740715027,
"eval_runtime": 475.4466,
"eval_samples_per_second": 3.82,
"eval_steps_per_second": 0.061,
"eval_wer": 19.00115362430302,
"step": 2000
},
{
"epoch": 38.21,
"learning_rate": 6.615555555555556e-06,
"loss": 0.0004,
"step": 2025
},
{
"epoch": 38.68,
"learning_rate": 6.560000000000001e-06,
"loss": 0.0004,
"step": 2050
},
{
"epoch": 39.15,
"learning_rate": 6.504444444444446e-06,
"loss": 0.0008,
"step": 2075
},
{
"epoch": 39.62,
"learning_rate": 6.448888888888889e-06,
"loss": 0.0004,
"step": 2100
},
{
"epoch": 40.09,
"learning_rate": 6.393333333333334e-06,
"loss": 0.0004,
"step": 2125
},
{
"epoch": 40.57,
"learning_rate": 6.34e-06,
"loss": 0.0005,
"step": 2150
},
{
"epoch": 41.04,
"learning_rate": 6.284444444444445e-06,
"loss": 0.0006,
"step": 2175
},
{
"epoch": 41.51,
"learning_rate": 6.22888888888889e-06,
"loss": 0.0008,
"step": 2200
},
{
"epoch": 41.98,
"learning_rate": 6.173333333333333e-06,
"loss": 0.0007,
"step": 2225
},
{
"epoch": 42.45,
"learning_rate": 6.117777777777778e-06,
"loss": 0.0005,
"step": 2250
},
{
"epoch": 42.92,
"learning_rate": 6.062222222222223e-06,
"loss": 0.0006,
"step": 2275
},
{
"epoch": 43.4,
"learning_rate": 6.006666666666667e-06,
"loss": 0.0005,
"step": 2300
},
{
"epoch": 43.87,
"learning_rate": 5.951111111111112e-06,
"loss": 0.0005,
"step": 2325
},
{
"epoch": 44.34,
"learning_rate": 5.895555555555557e-06,
"loss": 0.0003,
"step": 2350
},
{
"epoch": 44.81,
"learning_rate": 5.84e-06,
"loss": 0.0004,
"step": 2375
},
{
"epoch": 45.28,
"learning_rate": 5.784444444444445e-06,
"loss": 0.0004,
"step": 2400
},
{
"epoch": 45.75,
"learning_rate": 5.72888888888889e-06,
"loss": 0.0002,
"step": 2425
},
{
"epoch": 46.23,
"learning_rate": 5.673333333333334e-06,
"loss": 0.0003,
"step": 2450
},
{
"epoch": 46.7,
"learning_rate": 5.617777777777779e-06,
"loss": 0.0002,
"step": 2475
},
{
"epoch": 47.17,
"learning_rate": 5.562222222222222e-06,
"loss": 0.0001,
"step": 2500
},
{
"epoch": 47.64,
"learning_rate": 5.506666666666667e-06,
"loss": 0.0001,
"step": 2525
},
{
"epoch": 48.11,
"learning_rate": 5.451111111111112e-06,
"loss": 0.0001,
"step": 2550
},
{
"epoch": 48.58,
"learning_rate": 5.3955555555555565e-06,
"loss": 0.0001,
"step": 2575
},
{
"epoch": 49.06,
"learning_rate": 5.3400000000000005e-06,
"loss": 0.0001,
"step": 2600
},
{
"epoch": 49.53,
"learning_rate": 5.2844444444444454e-06,
"loss": 0.0001,
"step": 2625
},
{
"epoch": 50.0,
"learning_rate": 5.228888888888889e-06,
"loss": 0.0001,
"step": 2650
},
{
"epoch": 50.47,
"learning_rate": 5.1733333333333335e-06,
"loss": 0.0001,
"step": 2675
},
{
"epoch": 50.94,
"learning_rate": 5.117777777777778e-06,
"loss": 0.0001,
"step": 2700
},
{
"epoch": 51.42,
"learning_rate": 5.062222222222222e-06,
"loss": 0.0001,
"step": 2725
},
{
"epoch": 51.89,
"learning_rate": 5.006666666666667e-06,
"loss": 0.0001,
"step": 2750
},
{
"epoch": 52.36,
"learning_rate": 4.951111111111111e-06,
"loss": 0.0001,
"step": 2775
},
{
"epoch": 52.83,
"learning_rate": 4.895555555555556e-06,
"loss": 0.0001,
"step": 2800
},
{
"epoch": 53.3,
"learning_rate": 4.84e-06,
"loss": 0.0001,
"step": 2825
},
{
"epoch": 53.77,
"learning_rate": 4.784444444444445e-06,
"loss": 0.0001,
"step": 2850
},
{
"epoch": 54.25,
"learning_rate": 4.728888888888889e-06,
"loss": 0.0001,
"step": 2875
},
{
"epoch": 54.72,
"learning_rate": 4.673333333333333e-06,
"loss": 0.0001,
"step": 2900
},
{
"epoch": 55.19,
"learning_rate": 4.617777777777778e-06,
"loss": 0.0001,
"step": 2925
},
{
"epoch": 55.66,
"learning_rate": 4.562222222222222e-06,
"loss": 0.0001,
"step": 2950
},
{
"epoch": 56.13,
"learning_rate": 4.506666666666667e-06,
"loss": 0.0,
"step": 2975
},
{
"epoch": 56.6,
"learning_rate": 4.451111111111112e-06,
"loss": 0.0,
"step": 3000
},
{
"epoch": 56.6,
"eval_loss": 0.6230812072753906,
"eval_runtime": 475.4387,
"eval_samples_per_second": 3.82,
"eval_steps_per_second": 0.061,
"eval_wer": 18.578158046529513,
"step": 3000
},
{
"epoch": 57.08,
"learning_rate": 4.395555555555556e-06,
"loss": 0.0,
"step": 3025
},
{
"epoch": 57.55,
"learning_rate": 4.34e-06,
"loss": 0.0,
"step": 3050
},
{
"epoch": 58.02,
"learning_rate": 4.284444444444445e-06,
"loss": 0.0,
"step": 3075
},
{
"epoch": 58.49,
"learning_rate": 4.228888888888889e-06,
"loss": 0.0,
"step": 3100
},
{
"epoch": 58.96,
"learning_rate": 4.173333333333334e-06,
"loss": 0.0,
"step": 3125
},
{
"epoch": 59.43,
"learning_rate": 4.117777777777779e-06,
"loss": 0.0,
"step": 3150
},
{
"epoch": 59.91,
"learning_rate": 4.062222222222223e-06,
"loss": 0.0,
"step": 3175
},
{
"epoch": 60.38,
"learning_rate": 4.006666666666667e-06,
"loss": 0.0,
"step": 3200
},
{
"epoch": 60.85,
"learning_rate": 3.951111111111112e-06,
"loss": 0.0,
"step": 3225
},
{
"epoch": 61.32,
"learning_rate": 3.895555555555556e-06,
"loss": 0.0,
"step": 3250
},
{
"epoch": 61.79,
"learning_rate": 3.8400000000000005e-06,
"loss": 0.0,
"step": 3275
},
{
"epoch": 62.26,
"learning_rate": 3.784444444444445e-06,
"loss": 0.0,
"step": 3300
},
{
"epoch": 62.74,
"learning_rate": 3.728888888888889e-06,
"loss": 0.0,
"step": 3325
},
{
"epoch": 63.21,
"learning_rate": 3.673333333333334e-06,
"loss": 0.0,
"step": 3350
},
{
"epoch": 63.68,
"learning_rate": 3.617777777777778e-06,
"loss": 0.0,
"step": 3375
},
{
"epoch": 64.15,
"learning_rate": 3.5622222222222224e-06,
"loss": 0.0,
"step": 3400
},
{
"epoch": 64.62,
"learning_rate": 3.5066666666666673e-06,
"loss": 0.0,
"step": 3425
},
{
"epoch": 65.09,
"learning_rate": 3.4511111111111113e-06,
"loss": 0.0,
"step": 3450
},
{
"epoch": 65.57,
"learning_rate": 3.3955555555555558e-06,
"loss": 0.0,
"step": 3475
},
{
"epoch": 66.04,
"learning_rate": 3.3400000000000006e-06,
"loss": 0.0,
"step": 3500
},
{
"epoch": 66.51,
"learning_rate": 3.2844444444444447e-06,
"loss": 0.0,
"step": 3525
},
{
"epoch": 66.98,
"learning_rate": 3.228888888888889e-06,
"loss": 0.0,
"step": 3550
},
{
"epoch": 67.45,
"learning_rate": 3.173333333333334e-06,
"loss": 0.0,
"step": 3575
},
{
"epoch": 67.92,
"learning_rate": 3.117777777777778e-06,
"loss": 0.0,
"step": 3600
},
{
"epoch": 68.4,
"learning_rate": 3.0622222222222225e-06,
"loss": 0.0,
"step": 3625
},
{
"epoch": 68.87,
"learning_rate": 3.0066666666666674e-06,
"loss": 0.0,
"step": 3650
},
{
"epoch": 69.34,
"learning_rate": 2.9511111111111114e-06,
"loss": 0.0,
"step": 3675
},
{
"epoch": 69.81,
"learning_rate": 2.895555555555556e-06,
"loss": 0.0,
"step": 3700
},
{
"epoch": 70.28,
"learning_rate": 2.84e-06,
"loss": 0.0,
"step": 3725
},
{
"epoch": 70.75,
"learning_rate": 2.784444444444445e-06,
"loss": 0.0,
"step": 3750
},
{
"epoch": 71.23,
"learning_rate": 2.7288888888888893e-06,
"loss": 0.0,
"step": 3775
},
{
"epoch": 71.7,
"learning_rate": 2.6733333333333333e-06,
"loss": 0.0,
"step": 3800
},
{
"epoch": 72.17,
"learning_rate": 2.617777777777778e-06,
"loss": 0.0,
"step": 3825
},
{
"epoch": 72.64,
"learning_rate": 2.5622222222222226e-06,
"loss": 0.0,
"step": 3850
},
{
"epoch": 73.11,
"learning_rate": 2.5066666666666667e-06,
"loss": 0.0,
"step": 3875
},
{
"epoch": 73.58,
"learning_rate": 2.451111111111111e-06,
"loss": 0.0,
"step": 3900
},
{
"epoch": 74.06,
"learning_rate": 2.3955555555555556e-06,
"loss": 0.0,
"step": 3925
},
{
"epoch": 74.53,
"learning_rate": 2.3400000000000005e-06,
"loss": 0.0,
"step": 3950
},
{
"epoch": 75.0,
"learning_rate": 2.2844444444444445e-06,
"loss": 0.0,
"step": 3975
},
{
"epoch": 75.47,
"learning_rate": 2.228888888888889e-06,
"loss": 0.0,
"step": 4000
},
{
"epoch": 75.47,
"eval_loss": 0.6435311436653137,
"eval_runtime": 476.3235,
"eval_samples_per_second": 3.813,
"eval_steps_per_second": 0.061,
"eval_wer": 18.623822341857334,
"step": 4000
},
{
"epoch": 75.94,
"learning_rate": 2.1733333333333334e-06,
"loss": 0.0,
"step": 4025
},
{
"epoch": 76.42,
"learning_rate": 2.117777777777778e-06,
"loss": 0.0,
"step": 4050
},
{
"epoch": 76.89,
"learning_rate": 2.0622222222222223e-06,
"loss": 0.0,
"step": 4075
},
{
"epoch": 77.36,
"learning_rate": 2.006666666666667e-06,
"loss": 0.0,
"step": 4100
},
{
"epoch": 77.83,
"learning_rate": 1.9511111111111113e-06,
"loss": 0.0,
"step": 4125
},
{
"epoch": 78.3,
"learning_rate": 1.8955555555555557e-06,
"loss": 0.0,
"step": 4150
},
{
"epoch": 78.77,
"learning_rate": 1.8400000000000002e-06,
"loss": 0.0,
"step": 4175
},
{
"epoch": 79.25,
"learning_rate": 1.7844444444444444e-06,
"loss": 0.0,
"step": 4200
},
{
"epoch": 79.72,
"learning_rate": 1.728888888888889e-06,
"loss": 0.0,
"step": 4225
},
{
"epoch": 80.19,
"learning_rate": 1.6733333333333335e-06,
"loss": 0.0,
"step": 4250
},
{
"epoch": 80.66,
"learning_rate": 1.6177777777777778e-06,
"loss": 0.0,
"step": 4275
},
{
"epoch": 81.13,
"learning_rate": 1.5622222222222225e-06,
"loss": 0.0,
"step": 4300
},
{
"epoch": 81.6,
"learning_rate": 1.506666666666667e-06,
"loss": 0.0,
"step": 4325
},
{
"epoch": 82.08,
"learning_rate": 1.4511111111111112e-06,
"loss": 0.0,
"step": 4350
},
{
"epoch": 82.55,
"learning_rate": 1.3955555555555556e-06,
"loss": 0.0,
"step": 4375
},
{
"epoch": 83.02,
"learning_rate": 1.34e-06,
"loss": 0.0,
"step": 4400
},
{
"epoch": 83.49,
"learning_rate": 1.2844444444444445e-06,
"loss": 0.0,
"step": 4425
},
{
"epoch": 83.96,
"learning_rate": 1.228888888888889e-06,
"loss": 0.0,
"step": 4450
},
{
"epoch": 84.43,
"learning_rate": 1.1733333333333335e-06,
"loss": 0.0,
"step": 4475
},
{
"epoch": 84.91,
"learning_rate": 1.117777777777778e-06,
"loss": 0.0,
"step": 4500
},
{
"epoch": 85.38,
"learning_rate": 1.0622222222222222e-06,
"loss": 0.0,
"step": 4525
},
{
"epoch": 85.85,
"learning_rate": 1.0066666666666668e-06,
"loss": 0.0,
"step": 4550
},
{
"epoch": 86.32,
"learning_rate": 9.511111111111111e-07,
"loss": 0.0,
"step": 4575
},
{
"epoch": 86.79,
"learning_rate": 8.955555555555557e-07,
"loss": 0.0,
"step": 4600
},
{
"epoch": 87.26,
"learning_rate": 8.400000000000001e-07,
"loss": 0.0,
"step": 4625
},
{
"epoch": 87.74,
"learning_rate": 7.844444444444445e-07,
"loss": 0.0,
"step": 4650
},
{
"epoch": 88.21,
"learning_rate": 7.28888888888889e-07,
"loss": 0.0,
"step": 4675
},
{
"epoch": 88.68,
"learning_rate": 6.733333333333334e-07,
"loss": 0.0,
"step": 4700
},
{
"epoch": 89.15,
"learning_rate": 6.177777777777778e-07,
"loss": 0.0,
"step": 4725
},
{
"epoch": 89.62,
"learning_rate": 5.622222222222223e-07,
"loss": 0.0,
"step": 4750
},
{
"epoch": 90.09,
"learning_rate": 5.066666666666667e-07,
"loss": 0.0,
"step": 4775
},
{
"epoch": 90.57,
"learning_rate": 4.511111111111111e-07,
"loss": 0.0,
"step": 4800
},
{
"epoch": 91.04,
"learning_rate": 3.9555555555555557e-07,
"loss": 0.0,
"step": 4825
},
{
"epoch": 91.51,
"learning_rate": 3.4000000000000003e-07,
"loss": 0.0,
"step": 4850
},
{
"epoch": 91.98,
"learning_rate": 2.844444444444445e-07,
"loss": 0.0,
"step": 4875
},
{
"epoch": 92.45,
"learning_rate": 2.2888888888888892e-07,
"loss": 0.0,
"step": 4900
},
{
"epoch": 92.92,
"learning_rate": 1.7333333333333335e-07,
"loss": 0.0,
"step": 4925
},
{
"epoch": 93.4,
"learning_rate": 1.1777777777777778e-07,
"loss": 0.0,
"step": 4950
},
{
"epoch": 93.87,
"learning_rate": 6.222222222222223e-08,
"loss": 0.0,
"step": 4975
},
{
"epoch": 94.34,
"learning_rate": 6.666666666666667e-09,
"loss": 0.0,
"step": 5000
},
{
"epoch": 94.34,
"eval_loss": 0.6519695520401001,
"eval_runtime": 477.6808,
"eval_samples_per_second": 3.802,
"eval_steps_per_second": 0.061,
"eval_wer": 18.604595270140358,
"step": 5000
},
{
"epoch": 94.34,
"step": 5000,
"total_flos": 9.1587773092709e+19,
"train_loss": 0.03470887508669403,
"train_runtime": 10504.9163,
"train_samples_per_second": 30.462,
"train_steps_per_second": 0.476
}
],
"max_steps": 5000,
"num_train_epochs": 95,
"total_flos": 9.1587773092709e+19,
"trial_name": null,
"trial_params": null
}