whisper-small-oriya-v1.3 / trainer_state.json
chandan3007's picture
Upload 20 files
469ac65
raw
history blame contribute delete
No virus
25.8 kB
{
"best_metric": 29.055007052186177,
"best_model_checkpoint": "./checkpoint-2000",
"epoch": 250.0,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.25,
"learning_rate": 4.0000000000000003e-07,
"loss": 4.7471,
"step": 25
},
{
"epoch": 2.5,
"learning_rate": 9.000000000000001e-07,
"loss": 2.5917,
"step": 50
},
{
"epoch": 3.75,
"learning_rate": 1.4000000000000001e-06,
"loss": 1.7827,
"step": 75
},
{
"epoch": 5.0,
"learning_rate": 1.9000000000000002e-06,
"loss": 1.4544,
"step": 100
},
{
"epoch": 6.25,
"learning_rate": 2.4000000000000003e-06,
"loss": 1.2362,
"step": 125
},
{
"epoch": 7.5,
"learning_rate": 2.9e-06,
"loss": 1.0861,
"step": 150
},
{
"epoch": 8.75,
"learning_rate": 3.4000000000000005e-06,
"loss": 1.0142,
"step": 175
},
{
"epoch": 10.0,
"learning_rate": 3.900000000000001e-06,
"loss": 0.9151,
"step": 200
},
{
"epoch": 11.25,
"learning_rate": 4.4e-06,
"loss": 0.6885,
"step": 225
},
{
"epoch": 12.5,
"learning_rate": 4.9000000000000005e-06,
"loss": 0.4378,
"step": 250
},
{
"epoch": 13.75,
"learning_rate": 5.400000000000001e-06,
"loss": 0.2751,
"step": 275
},
{
"epoch": 15.0,
"learning_rate": 5.9e-06,
"loss": 0.185,
"step": 300
},
{
"epoch": 16.25,
"learning_rate": 6.4000000000000006e-06,
"loss": 0.1255,
"step": 325
},
{
"epoch": 17.5,
"learning_rate": 6.9e-06,
"loss": 0.0927,
"step": 350
},
{
"epoch": 18.75,
"learning_rate": 7.4e-06,
"loss": 0.0697,
"step": 375
},
{
"epoch": 20.0,
"learning_rate": 7.9e-06,
"loss": 0.0539,
"step": 400
},
{
"epoch": 21.25,
"learning_rate": 8.400000000000001e-06,
"loss": 0.0351,
"step": 425
},
{
"epoch": 22.5,
"learning_rate": 8.900000000000001e-06,
"loss": 0.0254,
"step": 450
},
{
"epoch": 23.75,
"learning_rate": 9.4e-06,
"loss": 0.0228,
"step": 475
},
{
"epoch": 25.0,
"learning_rate": 9.9e-06,
"loss": 0.0163,
"step": 500
},
{
"epoch": 26.25,
"learning_rate": 9.955555555555556e-06,
"loss": 0.013,
"step": 525
},
{
"epoch": 27.5,
"learning_rate": 9.9e-06,
"loss": 0.0108,
"step": 550
},
{
"epoch": 28.75,
"learning_rate": 9.844444444444446e-06,
"loss": 0.0085,
"step": 575
},
{
"epoch": 30.0,
"learning_rate": 9.78888888888889e-06,
"loss": 0.0085,
"step": 600
},
{
"epoch": 31.25,
"learning_rate": 9.733333333333334e-06,
"loss": 0.0078,
"step": 625
},
{
"epoch": 32.5,
"learning_rate": 9.677777777777778e-06,
"loss": 0.0057,
"step": 650
},
{
"epoch": 33.75,
"learning_rate": 9.622222222222222e-06,
"loss": 0.0053,
"step": 675
},
{
"epoch": 35.0,
"learning_rate": 9.566666666666668e-06,
"loss": 0.0059,
"step": 700
},
{
"epoch": 36.25,
"learning_rate": 9.511111111111112e-06,
"loss": 0.0044,
"step": 725
},
{
"epoch": 37.5,
"learning_rate": 9.455555555555557e-06,
"loss": 0.0043,
"step": 750
},
{
"epoch": 38.75,
"learning_rate": 9.4e-06,
"loss": 0.0029,
"step": 775
},
{
"epoch": 40.0,
"learning_rate": 9.344444444444446e-06,
"loss": 0.0038,
"step": 800
},
{
"epoch": 41.25,
"learning_rate": 9.28888888888889e-06,
"loss": 0.0038,
"step": 825
},
{
"epoch": 42.5,
"learning_rate": 9.233333333333334e-06,
"loss": 0.0033,
"step": 850
},
{
"epoch": 43.75,
"learning_rate": 9.17777777777778e-06,
"loss": 0.0032,
"step": 875
},
{
"epoch": 45.0,
"learning_rate": 9.122222222222223e-06,
"loss": 0.0025,
"step": 900
},
{
"epoch": 46.25,
"learning_rate": 9.066666666666667e-06,
"loss": 0.0019,
"step": 925
},
{
"epoch": 47.5,
"learning_rate": 9.011111111111111e-06,
"loss": 0.0022,
"step": 950
},
{
"epoch": 48.75,
"learning_rate": 8.955555555555555e-06,
"loss": 0.0021,
"step": 975
},
{
"epoch": 50.0,
"learning_rate": 8.900000000000001e-06,
"loss": 0.0025,
"step": 1000
},
{
"epoch": 50.0,
"eval_loss": 0.46390804648399353,
"eval_runtime": 38.9032,
"eval_samples_per_second": 5.655,
"eval_steps_per_second": 0.18,
"eval_wer": 35.34555712270804,
"step": 1000
},
{
"epoch": 51.25,
"learning_rate": 8.844444444444445e-06,
"loss": 0.0024,
"step": 1025
},
{
"epoch": 52.5,
"learning_rate": 8.788888888888891e-06,
"loss": 0.0034,
"step": 1050
},
{
"epoch": 53.75,
"learning_rate": 8.733333333333333e-06,
"loss": 0.002,
"step": 1075
},
{
"epoch": 55.0,
"learning_rate": 8.677777777777779e-06,
"loss": 0.0018,
"step": 1100
},
{
"epoch": 56.25,
"learning_rate": 8.622222222222223e-06,
"loss": 0.0018,
"step": 1125
},
{
"epoch": 57.5,
"learning_rate": 8.566666666666667e-06,
"loss": 0.0018,
"step": 1150
},
{
"epoch": 58.75,
"learning_rate": 8.511111111111113e-06,
"loss": 0.0023,
"step": 1175
},
{
"epoch": 60.0,
"learning_rate": 8.455555555555555e-06,
"loss": 0.0027,
"step": 1200
},
{
"epoch": 61.25,
"learning_rate": 8.400000000000001e-06,
"loss": 0.0022,
"step": 1225
},
{
"epoch": 62.5,
"learning_rate": 8.344444444444445e-06,
"loss": 0.0027,
"step": 1250
},
{
"epoch": 63.75,
"learning_rate": 8.288888888888889e-06,
"loss": 0.0006,
"step": 1275
},
{
"epoch": 65.0,
"learning_rate": 8.233333333333335e-06,
"loss": 0.0005,
"step": 1300
},
{
"epoch": 66.25,
"learning_rate": 8.177777777777779e-06,
"loss": 0.002,
"step": 1325
},
{
"epoch": 67.5,
"learning_rate": 8.122222222222223e-06,
"loss": 0.0023,
"step": 1350
},
{
"epoch": 68.75,
"learning_rate": 8.066666666666667e-06,
"loss": 0.0016,
"step": 1375
},
{
"epoch": 70.0,
"learning_rate": 8.011111111111113e-06,
"loss": 0.0008,
"step": 1400
},
{
"epoch": 71.25,
"learning_rate": 7.955555555555557e-06,
"loss": 0.0014,
"step": 1425
},
{
"epoch": 72.5,
"learning_rate": 7.9e-06,
"loss": 0.0016,
"step": 1450
},
{
"epoch": 73.75,
"learning_rate": 7.844444444444446e-06,
"loss": 0.0011,
"step": 1475
},
{
"epoch": 75.0,
"learning_rate": 7.788888888888889e-06,
"loss": 0.0017,
"step": 1500
},
{
"epoch": 76.25,
"learning_rate": 7.733333333333334e-06,
"loss": 0.0013,
"step": 1525
},
{
"epoch": 77.5,
"learning_rate": 7.677777777777778e-06,
"loss": 0.0012,
"step": 1550
},
{
"epoch": 78.75,
"learning_rate": 7.622222222222223e-06,
"loss": 0.0012,
"step": 1575
},
{
"epoch": 80.0,
"learning_rate": 7.566666666666667e-06,
"loss": 0.0014,
"step": 1600
},
{
"epoch": 81.25,
"learning_rate": 7.511111111111111e-06,
"loss": 0.0019,
"step": 1625
},
{
"epoch": 82.5,
"learning_rate": 7.455555555555556e-06,
"loss": 0.0013,
"step": 1650
},
{
"epoch": 83.75,
"learning_rate": 7.4e-06,
"loss": 0.0012,
"step": 1675
},
{
"epoch": 85.0,
"learning_rate": 7.344444444444445e-06,
"loss": 0.0008,
"step": 1700
},
{
"epoch": 86.25,
"learning_rate": 7.28888888888889e-06,
"loss": 0.0015,
"step": 1725
},
{
"epoch": 87.5,
"learning_rate": 7.233333333333334e-06,
"loss": 0.0012,
"step": 1750
},
{
"epoch": 88.75,
"learning_rate": 7.177777777777778e-06,
"loss": 0.0009,
"step": 1775
},
{
"epoch": 90.0,
"learning_rate": 7.122222222222222e-06,
"loss": 0.001,
"step": 1800
},
{
"epoch": 91.25,
"learning_rate": 7.066666666666667e-06,
"loss": 0.0012,
"step": 1825
},
{
"epoch": 92.5,
"learning_rate": 7.011111111111112e-06,
"loss": 0.0008,
"step": 1850
},
{
"epoch": 93.75,
"learning_rate": 6.955555555555557e-06,
"loss": 0.0011,
"step": 1875
},
{
"epoch": 95.0,
"learning_rate": 6.9e-06,
"loss": 0.0013,
"step": 1900
},
{
"epoch": 96.25,
"learning_rate": 6.844444444444445e-06,
"loss": 0.0006,
"step": 1925
},
{
"epoch": 97.5,
"learning_rate": 6.788888888888889e-06,
"loss": 0.0005,
"step": 1950
},
{
"epoch": 98.75,
"learning_rate": 6.733333333333334e-06,
"loss": 0.0002,
"step": 1975
},
{
"epoch": 100.0,
"learning_rate": 6.677777777777779e-06,
"loss": 0.0002,
"step": 2000
},
{
"epoch": 100.0,
"eval_loss": 0.5570347309112549,
"eval_runtime": 38.7413,
"eval_samples_per_second": 5.679,
"eval_steps_per_second": 0.181,
"eval_wer": 29.055007052186177,
"step": 2000
},
{
"epoch": 101.25,
"learning_rate": 6.6222222222222236e-06,
"loss": 0.001,
"step": 2025
},
{
"epoch": 102.5,
"learning_rate": 6.566666666666667e-06,
"loss": 0.001,
"step": 2050
},
{
"epoch": 103.75,
"learning_rate": 6.511111111111112e-06,
"loss": 0.001,
"step": 2075
},
{
"epoch": 105.0,
"learning_rate": 6.455555555555556e-06,
"loss": 0.0006,
"step": 2100
},
{
"epoch": 106.25,
"learning_rate": 6.4000000000000006e-06,
"loss": 0.0005,
"step": 2125
},
{
"epoch": 107.5,
"learning_rate": 6.3444444444444454e-06,
"loss": 0.0012,
"step": 2150
},
{
"epoch": 108.75,
"learning_rate": 6.28888888888889e-06,
"loss": 0.0014,
"step": 2175
},
{
"epoch": 110.0,
"learning_rate": 6.2333333333333335e-06,
"loss": 0.0008,
"step": 2200
},
{
"epoch": 111.25,
"learning_rate": 6.177777777777778e-06,
"loss": 0.0008,
"step": 2225
},
{
"epoch": 112.5,
"learning_rate": 6.1222222222222224e-06,
"loss": 0.0008,
"step": 2250
},
{
"epoch": 113.75,
"learning_rate": 6.066666666666667e-06,
"loss": 0.0003,
"step": 2275
},
{
"epoch": 115.0,
"learning_rate": 6.011111111111112e-06,
"loss": 0.0004,
"step": 2300
},
{
"epoch": 116.25,
"learning_rate": 5.955555555555555e-06,
"loss": 0.0005,
"step": 2325
},
{
"epoch": 117.5,
"learning_rate": 5.9e-06,
"loss": 0.0003,
"step": 2350
},
{
"epoch": 118.75,
"learning_rate": 5.844444444444445e-06,
"loss": 0.0003,
"step": 2375
},
{
"epoch": 120.0,
"learning_rate": 5.788888888888889e-06,
"loss": 0.0011,
"step": 2400
},
{
"epoch": 121.25,
"learning_rate": 5.733333333333334e-06,
"loss": 0.0004,
"step": 2425
},
{
"epoch": 122.5,
"learning_rate": 5.677777777777779e-06,
"loss": 0.0001,
"step": 2450
},
{
"epoch": 123.75,
"learning_rate": 5.622222222222222e-06,
"loss": 0.0003,
"step": 2475
},
{
"epoch": 125.0,
"learning_rate": 5.566666666666667e-06,
"loss": 0.0007,
"step": 2500
},
{
"epoch": 126.25,
"learning_rate": 5.511111111111112e-06,
"loss": 0.0004,
"step": 2525
},
{
"epoch": 127.5,
"learning_rate": 5.455555555555556e-06,
"loss": 0.0008,
"step": 2550
},
{
"epoch": 128.75,
"learning_rate": 5.400000000000001e-06,
"loss": 0.0008,
"step": 2575
},
{
"epoch": 130.0,
"learning_rate": 5.344444444444446e-06,
"loss": 0.0011,
"step": 2600
},
{
"epoch": 131.25,
"learning_rate": 5.288888888888889e-06,
"loss": 0.0004,
"step": 2625
},
{
"epoch": 132.5,
"learning_rate": 5.233333333333334e-06,
"loss": 0.0002,
"step": 2650
},
{
"epoch": 133.75,
"learning_rate": 5.177777777777779e-06,
"loss": 0.0004,
"step": 2675
},
{
"epoch": 135.0,
"learning_rate": 5.122222222222223e-06,
"loss": 0.0001,
"step": 2700
},
{
"epoch": 136.25,
"learning_rate": 5.0666666666666676e-06,
"loss": 0.0001,
"step": 2725
},
{
"epoch": 137.5,
"learning_rate": 5.011111111111111e-06,
"loss": 0.0002,
"step": 2750
},
{
"epoch": 138.75,
"learning_rate": 4.9555555555555565e-06,
"loss": 0.0004,
"step": 2775
},
{
"epoch": 140.0,
"learning_rate": 4.9000000000000005e-06,
"loss": 0.0023,
"step": 2800
},
{
"epoch": 141.25,
"learning_rate": 4.8444444444444446e-06,
"loss": 0.0003,
"step": 2825
},
{
"epoch": 142.5,
"learning_rate": 4.7888888888888894e-06,
"loss": 0.0002,
"step": 2850
},
{
"epoch": 143.75,
"learning_rate": 4.7333333333333335e-06,
"loss": 0.0001,
"step": 2875
},
{
"epoch": 145.0,
"learning_rate": 4.677777777777778e-06,
"loss": 0.0002,
"step": 2900
},
{
"epoch": 146.25,
"learning_rate": 4.622222222222222e-06,
"loss": 0.0002,
"step": 2925
},
{
"epoch": 147.5,
"learning_rate": 4.566666666666667e-06,
"loss": 0.0001,
"step": 2950
},
{
"epoch": 148.75,
"learning_rate": 4.511111111111111e-06,
"loss": 0.0001,
"step": 2975
},
{
"epoch": 150.0,
"learning_rate": 4.455555555555555e-06,
"loss": 0.0004,
"step": 3000
},
{
"epoch": 150.0,
"eval_loss": 0.620841383934021,
"eval_runtime": 38.8643,
"eval_samples_per_second": 5.661,
"eval_steps_per_second": 0.18,
"eval_wer": 30.634696755994355,
"step": 3000
},
{
"epoch": 151.25,
"learning_rate": 4.4e-06,
"loss": 0.0001,
"step": 3025
},
{
"epoch": 152.5,
"learning_rate": 4.344444444444445e-06,
"loss": 0.0,
"step": 3050
},
{
"epoch": 153.75,
"learning_rate": 4.288888888888889e-06,
"loss": 0.0,
"step": 3075
},
{
"epoch": 155.0,
"learning_rate": 4.233333333333334e-06,
"loss": 0.0,
"step": 3100
},
{
"epoch": 156.25,
"learning_rate": 4.177777777777778e-06,
"loss": 0.0,
"step": 3125
},
{
"epoch": 157.5,
"learning_rate": 4.122222222222222e-06,
"loss": 0.0,
"step": 3150
},
{
"epoch": 158.75,
"learning_rate": 4.066666666666667e-06,
"loss": 0.0,
"step": 3175
},
{
"epoch": 160.0,
"learning_rate": 4.011111111111111e-06,
"loss": 0.0,
"step": 3200
},
{
"epoch": 161.25,
"learning_rate": 3.955555555555556e-06,
"loss": 0.0,
"step": 3225
},
{
"epoch": 162.5,
"learning_rate": 3.900000000000001e-06,
"loss": 0.0,
"step": 3250
},
{
"epoch": 163.75,
"learning_rate": 3.844444444444445e-06,
"loss": 0.0,
"step": 3275
},
{
"epoch": 165.0,
"learning_rate": 3.7888888888888893e-06,
"loss": 0.0,
"step": 3300
},
{
"epoch": 166.25,
"learning_rate": 3.7333333333333337e-06,
"loss": 0.0,
"step": 3325
},
{
"epoch": 167.5,
"learning_rate": 3.6777777777777778e-06,
"loss": 0.0,
"step": 3350
},
{
"epoch": 168.75,
"learning_rate": 3.6222222222222226e-06,
"loss": 0.0,
"step": 3375
},
{
"epoch": 170.0,
"learning_rate": 3.566666666666667e-06,
"loss": 0.0,
"step": 3400
},
{
"epoch": 171.25,
"learning_rate": 3.511111111111111e-06,
"loss": 0.0,
"step": 3425
},
{
"epoch": 172.5,
"learning_rate": 3.455555555555556e-06,
"loss": 0.0,
"step": 3450
},
{
"epoch": 173.75,
"learning_rate": 3.4000000000000005e-06,
"loss": 0.0,
"step": 3475
},
{
"epoch": 175.0,
"learning_rate": 3.3444444444444445e-06,
"loss": 0.0,
"step": 3500
},
{
"epoch": 176.25,
"learning_rate": 3.2888888888888894e-06,
"loss": 0.0,
"step": 3525
},
{
"epoch": 177.5,
"learning_rate": 3.2333333333333334e-06,
"loss": 0.0,
"step": 3550
},
{
"epoch": 178.75,
"learning_rate": 3.177777777777778e-06,
"loss": 0.0,
"step": 3575
},
{
"epoch": 180.0,
"learning_rate": 3.1222222222222228e-06,
"loss": 0.0,
"step": 3600
},
{
"epoch": 181.25,
"learning_rate": 3.066666666666667e-06,
"loss": 0.0,
"step": 3625
},
{
"epoch": 182.5,
"learning_rate": 3.0111111111111113e-06,
"loss": 0.0,
"step": 3650
},
{
"epoch": 183.75,
"learning_rate": 2.955555555555556e-06,
"loss": 0.0,
"step": 3675
},
{
"epoch": 185.0,
"learning_rate": 2.9e-06,
"loss": 0.0,
"step": 3700
},
{
"epoch": 186.25,
"learning_rate": 2.8444444444444446e-06,
"loss": 0.0,
"step": 3725
},
{
"epoch": 187.5,
"learning_rate": 2.788888888888889e-06,
"loss": 0.0,
"step": 3750
},
{
"epoch": 188.75,
"learning_rate": 2.7333333333333336e-06,
"loss": 0.0,
"step": 3775
},
{
"epoch": 190.0,
"learning_rate": 2.677777777777778e-06,
"loss": 0.0,
"step": 3800
},
{
"epoch": 191.25,
"learning_rate": 2.6222222222222225e-06,
"loss": 0.0,
"step": 3825
},
{
"epoch": 192.5,
"learning_rate": 2.566666666666667e-06,
"loss": 0.0,
"step": 3850
},
{
"epoch": 193.75,
"learning_rate": 2.5111111111111114e-06,
"loss": 0.0,
"step": 3875
},
{
"epoch": 195.0,
"learning_rate": 2.455555555555556e-06,
"loss": 0.0,
"step": 3900
},
{
"epoch": 196.25,
"learning_rate": 2.4000000000000003e-06,
"loss": 0.0,
"step": 3925
},
{
"epoch": 197.5,
"learning_rate": 2.3444444444444448e-06,
"loss": 0.0,
"step": 3950
},
{
"epoch": 198.75,
"learning_rate": 2.2888888888888892e-06,
"loss": 0.0,
"step": 3975
},
{
"epoch": 200.0,
"learning_rate": 2.2333333333333333e-06,
"loss": 0.0,
"step": 4000
},
{
"epoch": 200.0,
"eval_loss": 0.8745277523994446,
"eval_runtime": 39.3221,
"eval_samples_per_second": 5.595,
"eval_steps_per_second": 0.178,
"eval_wer": 33.540197461212976,
"step": 4000
},
{
"epoch": 201.25,
"learning_rate": 2.1777777777777777e-06,
"loss": 0.0,
"step": 4025
},
{
"epoch": 202.5,
"learning_rate": 2.1222222222222226e-06,
"loss": 0.0,
"step": 4050
},
{
"epoch": 203.75,
"learning_rate": 2.0666666666666666e-06,
"loss": 0.0,
"step": 4075
},
{
"epoch": 205.0,
"learning_rate": 2.011111111111111e-06,
"loss": 0.0,
"step": 4100
},
{
"epoch": 206.25,
"learning_rate": 1.955555555555556e-06,
"loss": 0.0,
"step": 4125
},
{
"epoch": 207.5,
"learning_rate": 1.9000000000000002e-06,
"loss": 0.0,
"step": 4150
},
{
"epoch": 208.75,
"learning_rate": 1.8444444444444445e-06,
"loss": 0.0,
"step": 4175
},
{
"epoch": 210.0,
"learning_rate": 1.788888888888889e-06,
"loss": 0.0,
"step": 4200
},
{
"epoch": 211.25,
"learning_rate": 1.7333333333333336e-06,
"loss": 0.0,
"step": 4225
},
{
"epoch": 212.5,
"learning_rate": 1.6777777777777779e-06,
"loss": 0.0,
"step": 4250
},
{
"epoch": 213.75,
"learning_rate": 1.6222222222222223e-06,
"loss": 0.0,
"step": 4275
},
{
"epoch": 215.0,
"learning_rate": 1.566666666666667e-06,
"loss": 0.0,
"step": 4300
},
{
"epoch": 216.25,
"learning_rate": 1.5111111111111112e-06,
"loss": 0.0,
"step": 4325
},
{
"epoch": 217.5,
"learning_rate": 1.4555555555555557e-06,
"loss": 0.0,
"step": 4350
},
{
"epoch": 218.75,
"learning_rate": 1.4000000000000001e-06,
"loss": 0.0,
"step": 4375
},
{
"epoch": 220.0,
"learning_rate": 1.3444444444444446e-06,
"loss": 0.0,
"step": 4400
},
{
"epoch": 221.25,
"learning_rate": 1.288888888888889e-06,
"loss": 0.0,
"step": 4425
},
{
"epoch": 222.5,
"learning_rate": 1.2333333333333335e-06,
"loss": 0.0,
"step": 4450
},
{
"epoch": 223.75,
"learning_rate": 1.1777777777777778e-06,
"loss": 0.0,
"step": 4475
},
{
"epoch": 225.0,
"learning_rate": 1.1222222222222222e-06,
"loss": 0.0,
"step": 4500
},
{
"epoch": 226.25,
"learning_rate": 1.066666666666667e-06,
"loss": 0.0,
"step": 4525
},
{
"epoch": 227.5,
"learning_rate": 1.0111111111111111e-06,
"loss": 0.0,
"step": 4550
},
{
"epoch": 228.75,
"learning_rate": 9.555555555555556e-07,
"loss": 0.0,
"step": 4575
},
{
"epoch": 230.0,
"learning_rate": 9.000000000000001e-07,
"loss": 0.0,
"step": 4600
},
{
"epoch": 231.25,
"learning_rate": 8.444444444444445e-07,
"loss": 0.0,
"step": 4625
},
{
"epoch": 232.5,
"learning_rate": 7.888888888888889e-07,
"loss": 0.0,
"step": 4650
},
{
"epoch": 233.75,
"learning_rate": 7.333333333333334e-07,
"loss": 0.0,
"step": 4675
},
{
"epoch": 235.0,
"learning_rate": 6.777777777777779e-07,
"loss": 0.0,
"step": 4700
},
{
"epoch": 236.25,
"learning_rate": 6.222222222222223e-07,
"loss": 0.0,
"step": 4725
},
{
"epoch": 237.5,
"learning_rate": 5.666666666666667e-07,
"loss": 0.0,
"step": 4750
},
{
"epoch": 238.75,
"learning_rate": 5.111111111111112e-07,
"loss": 0.0,
"step": 4775
},
{
"epoch": 240.0,
"learning_rate": 4.5555555555555563e-07,
"loss": 0.0,
"step": 4800
},
{
"epoch": 241.25,
"learning_rate": 4.0000000000000003e-07,
"loss": 0.0,
"step": 4825
},
{
"epoch": 242.5,
"learning_rate": 3.444444444444445e-07,
"loss": 0.0,
"step": 4850
},
{
"epoch": 243.75,
"learning_rate": 2.888888888888889e-07,
"loss": 0.0,
"step": 4875
},
{
"epoch": 245.0,
"learning_rate": 2.3333333333333336e-07,
"loss": 0.0,
"step": 4900
},
{
"epoch": 246.25,
"learning_rate": 1.777777777777778e-07,
"loss": 0.0,
"step": 4925
},
{
"epoch": 247.5,
"learning_rate": 1.2222222222222225e-07,
"loss": 0.0,
"step": 4950
},
{
"epoch": 248.75,
"learning_rate": 6.666666666666668e-08,
"loss": 0.0,
"step": 4975
},
{
"epoch": 250.0,
"learning_rate": 1.1111111111111112e-08,
"loss": 0.0,
"step": 5000
},
{
"epoch": 250.0,
"eval_loss": 0.9034796953201294,
"eval_runtime": 39.275,
"eval_samples_per_second": 5.602,
"eval_steps_per_second": 0.178,
"eval_wer": 33.004231311706626,
"step": 5000
},
{
"epoch": 250.0,
"step": 5000,
"total_flos": 4.458644434944e+19,
"train_loss": 0.08519221190060271,
"train_runtime": 18645.4161,
"train_samples_per_second": 8.581,
"train_steps_per_second": 0.268
}
],
"max_steps": 5000,
"num_train_epochs": 250,
"total_flos": 4.458644434944e+19,
"trial_name": null,
"trial_params": null
}