whisper-medium-az / trainer_state.json
vumichien's picture
End of training
2231191
{
"best_metric": 47.337278106508876,
"best_model_checkpoint": "./checkpoint-1000",
"epoch": 2499.0004,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 12.0,
"learning_rate": 4.800000000000001e-07,
"loss": 1.1584,
"step": 25
},
{
"epoch": 24.0,
"learning_rate": 9.800000000000001e-07,
"loss": 0.5673,
"step": 50
},
{
"epoch": 37.0,
"learning_rate": 1.48e-06,
"loss": 0.0905,
"step": 75
},
{
"epoch": 49.0,
"learning_rate": 1.98e-06,
"loss": 0.0068,
"step": 100
},
{
"epoch": 62.0,
"learning_rate": 2.4800000000000004e-06,
"loss": 0.0023,
"step": 125
},
{
"epoch": 74.0,
"learning_rate": 2.9800000000000003e-06,
"loss": 0.0014,
"step": 150
},
{
"epoch": 87.0,
"learning_rate": 3.48e-06,
"loss": 0.0009,
"step": 175
},
{
"epoch": 99.0,
"learning_rate": 3.980000000000001e-06,
"loss": 0.0007,
"step": 200
},
{
"epoch": 112.0,
"learning_rate": 4.48e-06,
"loss": 0.0005,
"step": 225
},
{
"epoch": 124.0,
"learning_rate": 4.980000000000001e-06,
"loss": 0.0004,
"step": 250
},
{
"epoch": 137.0,
"learning_rate": 5.480000000000001e-06,
"loss": 0.0003,
"step": 275
},
{
"epoch": 149.0,
"learning_rate": 5.98e-06,
"loss": 0.0003,
"step": 300
},
{
"epoch": 162.0,
"learning_rate": 6.480000000000001e-06,
"loss": 0.0002,
"step": 325
},
{
"epoch": 174.0,
"learning_rate": 6.98e-06,
"loss": 0.0002,
"step": 350
},
{
"epoch": 187.0,
"learning_rate": 7.48e-06,
"loss": 0.0002,
"step": 375
},
{
"epoch": 199.0,
"learning_rate": 7.980000000000002e-06,
"loss": 0.0001,
"step": 400
},
{
"epoch": 212.0,
"learning_rate": 8.48e-06,
"loss": 0.0001,
"step": 425
},
{
"epoch": 224.0,
"learning_rate": 8.98e-06,
"loss": 0.0001,
"step": 450
},
{
"epoch": 237.0,
"learning_rate": 9.48e-06,
"loss": 0.0001,
"step": 475
},
{
"epoch": 249.0,
"learning_rate": 9.980000000000001e-06,
"loss": 0.0001,
"step": 500
},
{
"epoch": 262.0,
"learning_rate": 9.946666666666667e-06,
"loss": 0.0001,
"step": 525
},
{
"epoch": 274.0,
"learning_rate": 9.891111111111113e-06,
"loss": 0.0001,
"step": 550
},
{
"epoch": 287.0,
"learning_rate": 9.835555555555556e-06,
"loss": 0.0001,
"step": 575
},
{
"epoch": 299.0,
"learning_rate": 9.780000000000001e-06,
"loss": 0.0001,
"step": 600
},
{
"epoch": 312.0,
"learning_rate": 9.724444444444445e-06,
"loss": 0.0001,
"step": 625
},
{
"epoch": 324.0,
"learning_rate": 9.66888888888889e-06,
"loss": 0.0,
"step": 650
},
{
"epoch": 337.0,
"learning_rate": 9.613333333333335e-06,
"loss": 0.0,
"step": 675
},
{
"epoch": 349.0,
"learning_rate": 9.557777777777777e-06,
"loss": 0.0,
"step": 700
},
{
"epoch": 362.0,
"learning_rate": 9.502222222222223e-06,
"loss": 0.0,
"step": 725
},
{
"epoch": 374.0,
"learning_rate": 9.446666666666667e-06,
"loss": 0.0,
"step": 750
},
{
"epoch": 387.0,
"learning_rate": 9.391111111111111e-06,
"loss": 0.0,
"step": 775
},
{
"epoch": 399.0,
"learning_rate": 9.335555555555557e-06,
"loss": 0.0,
"step": 800
},
{
"epoch": 412.0,
"learning_rate": 9.280000000000001e-06,
"loss": 0.0,
"step": 825
},
{
"epoch": 424.0,
"learning_rate": 9.224444444444445e-06,
"loss": 0.0,
"step": 850
},
{
"epoch": 437.0,
"learning_rate": 9.168888888888889e-06,
"loss": 0.0,
"step": 875
},
{
"epoch": 449.0,
"learning_rate": 9.113333333333335e-06,
"loss": 0.0,
"step": 900
},
{
"epoch": 462.0,
"learning_rate": 9.057777777777779e-06,
"loss": 0.0,
"step": 925
},
{
"epoch": 474.0,
"learning_rate": 9.002222222222223e-06,
"loss": 0.0,
"step": 950
},
{
"epoch": 487.0,
"learning_rate": 8.946666666666669e-06,
"loss": 0.0,
"step": 975
},
{
"epoch": 499.0,
"learning_rate": 8.891111111111111e-06,
"loss": 0.0,
"step": 1000
},
{
"epoch": 499.0,
"eval_loss": 0.7816162109375,
"eval_runtime": 25.5855,
"eval_samples_per_second": 0.86,
"eval_steps_per_second": 0.078,
"eval_wer": 47.337278106508876,
"step": 1000
},
{
"epoch": 512.0,
"learning_rate": 8.835555555555557e-06,
"loss": 0.0,
"step": 1025
},
{
"epoch": 524.0,
"learning_rate": 8.78e-06,
"loss": 0.0,
"step": 1050
},
{
"epoch": 537.0,
"learning_rate": 8.724444444444445e-06,
"loss": 0.0,
"step": 1075
},
{
"epoch": 549.0,
"learning_rate": 8.66888888888889e-06,
"loss": 0.0,
"step": 1100
},
{
"epoch": 562.0,
"learning_rate": 8.613333333333333e-06,
"loss": 0.0,
"step": 1125
},
{
"epoch": 574.0,
"learning_rate": 8.557777777777778e-06,
"loss": 0.0,
"step": 1150
},
{
"epoch": 587.0,
"learning_rate": 8.502222222222223e-06,
"loss": 0.0,
"step": 1175
},
{
"epoch": 599.0,
"learning_rate": 8.446666666666668e-06,
"loss": 0.0,
"step": 1200
},
{
"epoch": 612.0,
"learning_rate": 8.391111111111112e-06,
"loss": 0.0,
"step": 1225
},
{
"epoch": 624.0,
"learning_rate": 8.335555555555556e-06,
"loss": 0.0,
"step": 1250
},
{
"epoch": 637.0,
"learning_rate": 8.28e-06,
"loss": 0.0,
"step": 1275
},
{
"epoch": 649.0,
"learning_rate": 8.224444444444444e-06,
"loss": 0.0,
"step": 1300
},
{
"epoch": 662.0,
"learning_rate": 8.16888888888889e-06,
"loss": 0.0,
"step": 1325
},
{
"epoch": 674.0,
"learning_rate": 8.113333333333334e-06,
"loss": 0.0,
"step": 1350
},
{
"epoch": 687.0,
"learning_rate": 8.057777777777778e-06,
"loss": 0.0,
"step": 1375
},
{
"epoch": 699.0,
"learning_rate": 8.002222222222222e-06,
"loss": 0.0,
"step": 1400
},
{
"epoch": 712.0,
"learning_rate": 7.946666666666666e-06,
"loss": 0.0,
"step": 1425
},
{
"epoch": 724.0,
"learning_rate": 7.891111111111112e-06,
"loss": 0.0,
"step": 1450
},
{
"epoch": 737.0,
"learning_rate": 7.835555555555556e-06,
"loss": 0.0,
"step": 1475
},
{
"epoch": 749.0,
"learning_rate": 7.78e-06,
"loss": 0.0,
"step": 1500
},
{
"epoch": 762.0,
"learning_rate": 7.724444444444446e-06,
"loss": 0.0,
"step": 1525
},
{
"epoch": 774.0,
"learning_rate": 7.66888888888889e-06,
"loss": 0.0,
"step": 1550
},
{
"epoch": 787.0,
"learning_rate": 7.613333333333334e-06,
"loss": 0.0,
"step": 1575
},
{
"epoch": 799.0,
"learning_rate": 7.557777777777779e-06,
"loss": 0.0,
"step": 1600
},
{
"epoch": 812.0,
"learning_rate": 7.502222222222223e-06,
"loss": 0.0,
"step": 1625
},
{
"epoch": 824.0,
"learning_rate": 7.446666666666668e-06,
"loss": 0.0,
"step": 1650
},
{
"epoch": 837.0,
"learning_rate": 7.3911111111111125e-06,
"loss": 0.0,
"step": 1675
},
{
"epoch": 849.0,
"learning_rate": 7.335555555555556e-06,
"loss": 0.0,
"step": 1700
},
{
"epoch": 862.0,
"learning_rate": 7.280000000000001e-06,
"loss": 0.0,
"step": 1725
},
{
"epoch": 874.0,
"learning_rate": 7.224444444444445e-06,
"loss": 0.0,
"step": 1750
},
{
"epoch": 887.0,
"learning_rate": 7.1688888888888895e-06,
"loss": 0.0,
"step": 1775
},
{
"epoch": 899.0,
"learning_rate": 7.113333333333334e-06,
"loss": 0.0,
"step": 1800
},
{
"epoch": 912.0,
"learning_rate": 7.057777777777778e-06,
"loss": 0.0,
"step": 1825
},
{
"epoch": 924.0,
"learning_rate": 7.0022222222222225e-06,
"loss": 0.0,
"step": 1850
},
{
"epoch": 937.0,
"learning_rate": 6.946666666666667e-06,
"loss": 0.0,
"step": 1875
},
{
"epoch": 949.0,
"learning_rate": 6.891111111111111e-06,
"loss": 0.0,
"step": 1900
},
{
"epoch": 962.0,
"learning_rate": 6.835555555555556e-06,
"loss": 0.0,
"step": 1925
},
{
"epoch": 974.0,
"learning_rate": 6.780000000000001e-06,
"loss": 0.0,
"step": 1950
},
{
"epoch": 987.0,
"learning_rate": 6.724444444444444e-06,
"loss": 0.0,
"step": 1975
},
{
"epoch": 999.0,
"learning_rate": 6.668888888888889e-06,
"loss": 0.0,
"step": 2000
},
{
"epoch": 999.0,
"eval_loss": 0.9050376415252686,
"eval_runtime": 23.012,
"eval_samples_per_second": 0.956,
"eval_steps_per_second": 0.087,
"eval_wer": 47.337278106508876,
"step": 2000
},
{
"epoch": 1012.0,
"learning_rate": 6.613333333333334e-06,
"loss": 0.0,
"step": 2025
},
{
"epoch": 1024.0,
"learning_rate": 6.557777777777778e-06,
"loss": 0.0,
"step": 2050
},
{
"epoch": 1037.0,
"learning_rate": 6.502222222222223e-06,
"loss": 0.0,
"step": 2075
},
{
"epoch": 1049.0,
"learning_rate": 6.446666666666668e-06,
"loss": 0.0,
"step": 2100
},
{
"epoch": 1062.0,
"learning_rate": 6.391111111111111e-06,
"loss": 0.0,
"step": 2125
},
{
"epoch": 1074.0,
"learning_rate": 6.335555555555556e-06,
"loss": 0.0,
"step": 2150
},
{
"epoch": 1087.0,
"learning_rate": 6.280000000000001e-06,
"loss": 0.0,
"step": 2175
},
{
"epoch": 1099.0,
"learning_rate": 6.224444444444445e-06,
"loss": 0.0,
"step": 2200
},
{
"epoch": 1112.0,
"learning_rate": 6.16888888888889e-06,
"loss": 0.0,
"step": 2225
},
{
"epoch": 1124.0,
"learning_rate": 6.113333333333333e-06,
"loss": 0.0,
"step": 2250
},
{
"epoch": 1137.0,
"learning_rate": 6.057777777777778e-06,
"loss": 0.0,
"step": 2275
},
{
"epoch": 1149.0,
"learning_rate": 6.002222222222223e-06,
"loss": 0.0,
"step": 2300
},
{
"epoch": 1162.0,
"learning_rate": 5.946666666666668e-06,
"loss": 0.0,
"step": 2325
},
{
"epoch": 1174.0,
"learning_rate": 5.891111111111112e-06,
"loss": 0.0,
"step": 2350
},
{
"epoch": 1187.0,
"learning_rate": 5.8355555555555565e-06,
"loss": 0.0,
"step": 2375
},
{
"epoch": 1199.0,
"learning_rate": 5.78e-06,
"loss": 0.0,
"step": 2400
},
{
"epoch": 1212.0,
"learning_rate": 5.724444444444445e-06,
"loss": 0.0,
"step": 2425
},
{
"epoch": 1224.0,
"learning_rate": 5.6688888888888895e-06,
"loss": 0.0,
"step": 2450
},
{
"epoch": 1237.0,
"learning_rate": 5.613333333333334e-06,
"loss": 0.0,
"step": 2475
},
{
"epoch": 1249.0,
"learning_rate": 5.557777777777778e-06,
"loss": 0.0,
"step": 2500
},
{
"epoch": 1262.0,
"learning_rate": 5.5022222222222224e-06,
"loss": 0.0,
"step": 2525
},
{
"epoch": 1274.0,
"learning_rate": 5.4466666666666665e-06,
"loss": 0.0,
"step": 2550
},
{
"epoch": 1287.0,
"learning_rate": 5.391111111111111e-06,
"loss": 0.0,
"step": 2575
},
{
"epoch": 1299.0,
"learning_rate": 5.335555555555556e-06,
"loss": 0.0,
"step": 2600
},
{
"epoch": 1312.0,
"learning_rate": 5.28e-06,
"loss": 0.0,
"step": 2625
},
{
"epoch": 1324.0,
"learning_rate": 5.224444444444445e-06,
"loss": 0.0,
"step": 2650
},
{
"epoch": 1337.0,
"learning_rate": 5.168888888888889e-06,
"loss": 0.0,
"step": 2675
},
{
"epoch": 1349.0,
"learning_rate": 5.113333333333333e-06,
"loss": 0.0,
"step": 2700
},
{
"epoch": 1362.0,
"learning_rate": 5.057777777777778e-06,
"loss": 0.0,
"step": 2725
},
{
"epoch": 1374.0,
"learning_rate": 5.002222222222223e-06,
"loss": 0.0,
"step": 2750
},
{
"epoch": 1387.0,
"learning_rate": 4.946666666666667e-06,
"loss": 0.0,
"step": 2775
},
{
"epoch": 1399.0,
"learning_rate": 4.891111111111111e-06,
"loss": 0.0,
"step": 2800
},
{
"epoch": 1412.0,
"learning_rate": 4.835555555555556e-06,
"loss": 0.0,
"step": 2825
},
{
"epoch": 1424.0,
"learning_rate": 4.78e-06,
"loss": 0.0,
"step": 2850
},
{
"epoch": 1437.0,
"learning_rate": 4.724444444444445e-06,
"loss": 0.0,
"step": 2875
},
{
"epoch": 1449.0,
"learning_rate": 4.66888888888889e-06,
"loss": 0.0,
"step": 2900
},
{
"epoch": 1462.0,
"learning_rate": 4.613333333333334e-06,
"loss": 0.0,
"step": 2925
},
{
"epoch": 1474.0,
"learning_rate": 4.557777777777778e-06,
"loss": 0.0,
"step": 2950
},
{
"epoch": 1487.0,
"learning_rate": 4.502222222222223e-06,
"loss": 0.0,
"step": 2975
},
{
"epoch": 1499.0,
"learning_rate": 4.446666666666667e-06,
"loss": 0.0,
"step": 3000
},
{
"epoch": 1499.0,
"eval_loss": 0.9688135385513306,
"eval_runtime": 23.0541,
"eval_samples_per_second": 0.954,
"eval_steps_per_second": 0.087,
"eval_wer": 47.928994082840234,
"step": 3000
},
{
"epoch": 1512.0,
"learning_rate": 4.391111111111112e-06,
"loss": 0.0,
"step": 3025
},
{
"epoch": 1524.0,
"learning_rate": 4.3355555555555565e-06,
"loss": 0.0,
"step": 3050
},
{
"epoch": 1537.0,
"learning_rate": 4.2800000000000005e-06,
"loss": 0.0,
"step": 3075
},
{
"epoch": 1549.0,
"learning_rate": 4.2244444444444446e-06,
"loss": 0.0,
"step": 3100
},
{
"epoch": 1562.0,
"learning_rate": 4.168888888888889e-06,
"loss": 0.0,
"step": 3125
},
{
"epoch": 1574.0,
"learning_rate": 4.1133333333333335e-06,
"loss": 0.0,
"step": 3150
},
{
"epoch": 1587.0,
"learning_rate": 4.057777777777778e-06,
"loss": 0.0,
"step": 3175
},
{
"epoch": 1599.0,
"learning_rate": 4.002222222222222e-06,
"loss": 0.0,
"step": 3200
},
{
"epoch": 1612.0,
"learning_rate": 3.946666666666667e-06,
"loss": 0.0,
"step": 3225
},
{
"epoch": 1624.0,
"learning_rate": 3.891111111111111e-06,
"loss": 0.0,
"step": 3250
},
{
"epoch": 1637.0,
"learning_rate": 3.835555555555555e-06,
"loss": 0.0,
"step": 3275
},
{
"epoch": 1649.0,
"learning_rate": 3.7800000000000002e-06,
"loss": 0.0,
"step": 3300
},
{
"epoch": 1662.0,
"learning_rate": 3.724444444444445e-06,
"loss": 0.0,
"step": 3325
},
{
"epoch": 1674.0,
"learning_rate": 3.668888888888889e-06,
"loss": 0.0,
"step": 3350
},
{
"epoch": 1687.0,
"learning_rate": 3.6133333333333336e-06,
"loss": 0.0,
"step": 3375
},
{
"epoch": 1699.0,
"learning_rate": 3.5577777777777785e-06,
"loss": 0.0,
"step": 3400
},
{
"epoch": 1712.0,
"learning_rate": 3.5022222222222225e-06,
"loss": 0.0,
"step": 3425
},
{
"epoch": 1724.0,
"learning_rate": 3.446666666666667e-06,
"loss": 0.0,
"step": 3450
},
{
"epoch": 1737.0,
"learning_rate": 3.391111111111111e-06,
"loss": 0.0,
"step": 3475
},
{
"epoch": 1749.0,
"learning_rate": 3.335555555555556e-06,
"loss": 0.0,
"step": 3500
},
{
"epoch": 1762.0,
"learning_rate": 3.2800000000000004e-06,
"loss": 0.0,
"step": 3525
},
{
"epoch": 1774.0,
"learning_rate": 3.2244444444444444e-06,
"loss": 0.0,
"step": 3550
},
{
"epoch": 1787.0,
"learning_rate": 3.1688888888888893e-06,
"loss": 0.0,
"step": 3575
},
{
"epoch": 1799.0,
"learning_rate": 3.1133333333333337e-06,
"loss": 0.0,
"step": 3600
},
{
"epoch": 1812.0,
"learning_rate": 3.0577777777777778e-06,
"loss": 0.0,
"step": 3625
},
{
"epoch": 1824.0,
"learning_rate": 3.0022222222222227e-06,
"loss": 0.0,
"step": 3650
},
{
"epoch": 1837.0,
"learning_rate": 2.946666666666667e-06,
"loss": 0.0,
"step": 3675
},
{
"epoch": 1849.0,
"learning_rate": 2.891111111111111e-06,
"loss": 0.0,
"step": 3700
},
{
"epoch": 1862.0,
"learning_rate": 2.835555555555556e-06,
"loss": 0.0,
"step": 3725
},
{
"epoch": 1874.0,
"learning_rate": 2.7800000000000005e-06,
"loss": 0.0,
"step": 3750
},
{
"epoch": 1887.0,
"learning_rate": 2.7244444444444445e-06,
"loss": 0.0,
"step": 3775
},
{
"epoch": 1899.0,
"learning_rate": 2.6688888888888894e-06,
"loss": 0.0,
"step": 3800
},
{
"epoch": 1912.0,
"learning_rate": 2.6133333333333334e-06,
"loss": 0.0,
"step": 3825
},
{
"epoch": 1924.0,
"learning_rate": 2.557777777777778e-06,
"loss": 0.0,
"step": 3850
},
{
"epoch": 1937.0,
"learning_rate": 2.5022222222222224e-06,
"loss": 0.0,
"step": 3875
},
{
"epoch": 1949.0,
"learning_rate": 2.446666666666667e-06,
"loss": 0.0,
"step": 3900
},
{
"epoch": 1962.0,
"learning_rate": 2.3911111111111113e-06,
"loss": 0.0,
"step": 3925
},
{
"epoch": 1974.0,
"learning_rate": 2.3355555555555557e-06,
"loss": 0.0,
"step": 3950
},
{
"epoch": 1987.0,
"learning_rate": 2.28e-06,
"loss": 0.0,
"step": 3975
},
{
"epoch": 1999.0,
"learning_rate": 2.2244444444444447e-06,
"loss": 0.0,
"step": 4000
},
{
"epoch": 1999.0,
"eval_loss": 1.0181673765182495,
"eval_runtime": 22.8111,
"eval_samples_per_second": 0.964,
"eval_steps_per_second": 0.088,
"eval_wer": 47.928994082840234,
"step": 4000
},
{
"epoch": 2012.0,
"learning_rate": 2.168888888888889e-06,
"loss": 0.0,
"step": 4025
},
{
"epoch": 2024.0,
"learning_rate": 2.1133333333333336e-06,
"loss": 0.0,
"step": 4050
},
{
"epoch": 2037.0,
"learning_rate": 2.057777777777778e-06,
"loss": 0.0,
"step": 4075
},
{
"epoch": 2049.0,
"learning_rate": 2.0022222222222225e-06,
"loss": 0.0,
"step": 4100
},
{
"epoch": 2062.0,
"learning_rate": 1.9466666666666665e-06,
"loss": 0.0,
"step": 4125
},
{
"epoch": 2074.0,
"learning_rate": 1.8911111111111114e-06,
"loss": 0.0,
"step": 4150
},
{
"epoch": 2087.0,
"learning_rate": 1.8355555555555557e-06,
"loss": 0.0,
"step": 4175
},
{
"epoch": 2099.0,
"learning_rate": 1.7800000000000001e-06,
"loss": 0.0,
"step": 4200
},
{
"epoch": 2112.0,
"learning_rate": 1.7244444444444448e-06,
"loss": 0.0,
"step": 4225
},
{
"epoch": 2124.0,
"learning_rate": 1.668888888888889e-06,
"loss": 0.0,
"step": 4250
},
{
"epoch": 2137.0,
"learning_rate": 1.6133333333333335e-06,
"loss": 0.0,
"step": 4275
},
{
"epoch": 2149.0,
"learning_rate": 1.5577777777777777e-06,
"loss": 0.0,
"step": 4300
},
{
"epoch": 2162.0,
"learning_rate": 1.5022222222222224e-06,
"loss": 0.0,
"step": 4325
},
{
"epoch": 2174.0,
"learning_rate": 1.4466666666666669e-06,
"loss": 0.0,
"step": 4350
},
{
"epoch": 2187.0,
"learning_rate": 1.3911111111111111e-06,
"loss": 0.0,
"step": 4375
},
{
"epoch": 2199.0,
"learning_rate": 1.3355555555555558e-06,
"loss": 0.0,
"step": 4400
},
{
"epoch": 2212.0,
"learning_rate": 1.28e-06,
"loss": 0.0,
"step": 4425
},
{
"epoch": 2224.0,
"learning_rate": 1.2244444444444445e-06,
"loss": 0.0,
"step": 4450
},
{
"epoch": 2237.0,
"learning_rate": 1.168888888888889e-06,
"loss": 0.0,
"step": 4475
},
{
"epoch": 2249.0,
"learning_rate": 1.1133333333333334e-06,
"loss": 0.0,
"step": 4500
},
{
"epoch": 2262.0,
"learning_rate": 1.0577777777777779e-06,
"loss": 0.0,
"step": 4525
},
{
"epoch": 2274.0,
"learning_rate": 1.0022222222222223e-06,
"loss": 0.0,
"step": 4550
},
{
"epoch": 2287.0,
"learning_rate": 9.466666666666667e-07,
"loss": 0.0,
"step": 4575
},
{
"epoch": 2299.0,
"learning_rate": 8.911111111111112e-07,
"loss": 0.0,
"step": 4600
},
{
"epoch": 2312.0,
"learning_rate": 8.355555555555556e-07,
"loss": 0.0,
"step": 4625
},
{
"epoch": 2324.0,
"learning_rate": 7.8e-07,
"loss": 0.0,
"step": 4650
},
{
"epoch": 2337.0,
"learning_rate": 7.244444444444446e-07,
"loss": 0.0,
"step": 4675
},
{
"epoch": 2349.0,
"learning_rate": 6.68888888888889e-07,
"loss": 0.0,
"step": 4700
},
{
"epoch": 2362.0,
"learning_rate": 6.133333333333333e-07,
"loss": 0.0,
"step": 4725
},
{
"epoch": 2374.0,
"learning_rate": 5.577777777777779e-07,
"loss": 0.0,
"step": 4750
},
{
"epoch": 2387.0,
"learning_rate": 5.022222222222222e-07,
"loss": 0.0,
"step": 4775
},
{
"epoch": 2399.0,
"learning_rate": 4.466666666666667e-07,
"loss": 0.0,
"step": 4800
},
{
"epoch": 2412.0,
"learning_rate": 3.9111111111111115e-07,
"loss": 0.0,
"step": 4825
},
{
"epoch": 2424.0,
"learning_rate": 3.3555555555555556e-07,
"loss": 0.0,
"step": 4850
},
{
"epoch": 2437.0,
"learning_rate": 2.8e-07,
"loss": 0.0,
"step": 4875
},
{
"epoch": 2449.0,
"learning_rate": 2.2444444444444445e-07,
"loss": 0.0,
"step": 4900
},
{
"epoch": 2462.0,
"learning_rate": 1.6888888888888888e-07,
"loss": 0.0,
"step": 4925
},
{
"epoch": 2474.0,
"learning_rate": 1.1333333333333336e-07,
"loss": 0.0,
"step": 4950
},
{
"epoch": 2487.0,
"learning_rate": 5.777777777777778e-08,
"loss": 0.0,
"step": 4975
},
{
"epoch": 2499.0,
"learning_rate": 2.2222222222222225e-09,
"loss": 0.0,
"step": 5000
},
{
"epoch": 2499.0,
"eval_loss": 1.0293190479278564,
"eval_runtime": 23.932,
"eval_samples_per_second": 0.919,
"eval_steps_per_second": 0.084,
"eval_wer": 47.928994082840234,
"step": 5000
},
{
"epoch": 2499.0,
"step": 5000,
"total_flos": 1.02060490752e+20,
"train_loss": 0.009165090815824806,
"train_runtime": 20280.5808,
"train_samples_per_second": 7.889,
"train_steps_per_second": 0.247
}
],
"max_steps": 5000,
"num_train_epochs": 9223372036854775807,
"total_flos": 1.02060490752e+20,
"trial_name": null,
"trial_params": null
}