whisper-large-v2-az / trainer_state.json
geninhu's picture
End of training
97707a8
{
"best_metric": 38.46153846153847,
"best_model_checkpoint": "./checkpoint-2000",
"epoch": 4999.0002,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 24.0,
"learning_rate": 4.4e-07,
"loss": 1.004,
"step": 25
},
{
"epoch": 49.0,
"learning_rate": 9.400000000000001e-07,
"loss": 0.3701,
"step": 50
},
{
"epoch": 74.0,
"learning_rate": 1.44e-06,
"loss": 0.0331,
"step": 75
},
{
"epoch": 99.0,
"learning_rate": 1.94e-06,
"loss": 0.0055,
"step": 100
},
{
"epoch": 124.0,
"learning_rate": 2.4400000000000004e-06,
"loss": 0.0026,
"step": 125
},
{
"epoch": 149.0,
"learning_rate": 2.9400000000000002e-06,
"loss": 0.0016,
"step": 150
},
{
"epoch": 174.0,
"learning_rate": 3.44e-06,
"loss": 0.0011,
"step": 175
},
{
"epoch": 199.0,
"learning_rate": 3.94e-06,
"loss": 0.0008,
"step": 200
},
{
"epoch": 224.0,
"learning_rate": 4.440000000000001e-06,
"loss": 0.0006,
"step": 225
},
{
"epoch": 249.0,
"learning_rate": 4.94e-06,
"loss": 0.0004,
"step": 250
},
{
"epoch": 274.0,
"learning_rate": 5.4400000000000004e-06,
"loss": 0.0003,
"step": 275
},
{
"epoch": 299.0,
"learning_rate": 5.94e-06,
"loss": 0.0003,
"step": 300
},
{
"epoch": 324.0,
"learning_rate": 6.440000000000001e-06,
"loss": 0.0002,
"step": 325
},
{
"epoch": 349.0,
"learning_rate": 6.9400000000000005e-06,
"loss": 0.0002,
"step": 350
},
{
"epoch": 374.0,
"learning_rate": 7.440000000000001e-06,
"loss": 0.0002,
"step": 375
},
{
"epoch": 399.0,
"learning_rate": 7.94e-06,
"loss": 0.0001,
"step": 400
},
{
"epoch": 424.0,
"learning_rate": 8.44e-06,
"loss": 0.0001,
"step": 425
},
{
"epoch": 449.0,
"learning_rate": 8.94e-06,
"loss": 0.0001,
"step": 450
},
{
"epoch": 474.0,
"learning_rate": 9.440000000000001e-06,
"loss": 0.0001,
"step": 475
},
{
"epoch": 499.0,
"learning_rate": 9.940000000000001e-06,
"loss": 0.0001,
"step": 500
},
{
"epoch": 524.0,
"learning_rate": 9.951111111111111e-06,
"loss": 0.0001,
"step": 525
},
{
"epoch": 549.0,
"learning_rate": 9.895555555555557e-06,
"loss": 0.0001,
"step": 550
},
{
"epoch": 574.0,
"learning_rate": 9.84e-06,
"loss": 0.0001,
"step": 575
},
{
"epoch": 599.0,
"learning_rate": 9.784444444444445e-06,
"loss": 0.0,
"step": 600
},
{
"epoch": 624.0,
"learning_rate": 9.72888888888889e-06,
"loss": 0.0,
"step": 625
},
{
"epoch": 649.0,
"learning_rate": 9.673333333333334e-06,
"loss": 0.0,
"step": 650
},
{
"epoch": 674.0,
"learning_rate": 9.617777777777778e-06,
"loss": 0.0,
"step": 675
},
{
"epoch": 699.0,
"learning_rate": 9.562222222222223e-06,
"loss": 0.0,
"step": 700
},
{
"epoch": 724.0,
"learning_rate": 9.506666666666667e-06,
"loss": 0.0,
"step": 725
},
{
"epoch": 749.0,
"learning_rate": 9.451111111111112e-06,
"loss": 0.0,
"step": 750
},
{
"epoch": 774.0,
"learning_rate": 9.395555555555556e-06,
"loss": 0.0,
"step": 775
},
{
"epoch": 799.0,
"learning_rate": 9.340000000000002e-06,
"loss": 0.0,
"step": 800
},
{
"epoch": 824.0,
"learning_rate": 9.284444444444444e-06,
"loss": 0.0,
"step": 825
},
{
"epoch": 849.0,
"learning_rate": 9.22888888888889e-06,
"loss": 0.0,
"step": 850
},
{
"epoch": 874.0,
"learning_rate": 9.173333333333334e-06,
"loss": 0.0,
"step": 875
},
{
"epoch": 899.0,
"learning_rate": 9.117777777777778e-06,
"loss": 0.0,
"step": 900
},
{
"epoch": 924.0,
"learning_rate": 9.062222222222224e-06,
"loss": 0.0,
"step": 925
},
{
"epoch": 949.0,
"learning_rate": 9.006666666666666e-06,
"loss": 0.0,
"step": 950
},
{
"epoch": 974.0,
"learning_rate": 8.951111111111112e-06,
"loss": 0.0,
"step": 975
},
{
"epoch": 999.0,
"learning_rate": 8.895555555555556e-06,
"loss": 0.0,
"step": 1000
},
{
"epoch": 999.0,
"eval_loss": 0.8372936844825745,
"eval_runtime": 38.9838,
"eval_samples_per_second": 0.564,
"eval_steps_per_second": 0.077,
"eval_wer": 39.64497041420118,
"step": 1000
},
{
"epoch": 1024.0,
"learning_rate": 8.84e-06,
"loss": 0.0,
"step": 1025
},
{
"epoch": 1049.0,
"learning_rate": 8.784444444444446e-06,
"loss": 0.0,
"step": 1050
},
{
"epoch": 1074.0,
"learning_rate": 8.72888888888889e-06,
"loss": 0.0,
"step": 1075
},
{
"epoch": 1099.0,
"learning_rate": 8.673333333333334e-06,
"loss": 0.0,
"step": 1100
},
{
"epoch": 1124.0,
"learning_rate": 8.617777777777778e-06,
"loss": 0.0,
"step": 1125
},
{
"epoch": 1149.0,
"learning_rate": 8.562222222222224e-06,
"loss": 0.0,
"step": 1150
},
{
"epoch": 1174.0,
"learning_rate": 8.506666666666668e-06,
"loss": 0.0,
"step": 1175
},
{
"epoch": 1199.0,
"learning_rate": 8.451111111111112e-06,
"loss": 0.0,
"step": 1200
},
{
"epoch": 1224.0,
"learning_rate": 8.395555555555557e-06,
"loss": 0.0,
"step": 1225
},
{
"epoch": 1249.0,
"learning_rate": 8.34e-06,
"loss": 0.0,
"step": 1250
},
{
"epoch": 1274.0,
"learning_rate": 8.284444444444446e-06,
"loss": 0.0,
"step": 1275
},
{
"epoch": 1299.0,
"learning_rate": 8.22888888888889e-06,
"loss": 0.0,
"step": 1300
},
{
"epoch": 1324.0,
"learning_rate": 8.173333333333334e-06,
"loss": 0.0,
"step": 1325
},
{
"epoch": 1349.0,
"learning_rate": 8.11777777777778e-06,
"loss": 0.0,
"step": 1350
},
{
"epoch": 1374.0,
"learning_rate": 8.062222222222222e-06,
"loss": 0.0,
"step": 1375
},
{
"epoch": 1399.0,
"learning_rate": 8.006666666666667e-06,
"loss": 0.0,
"step": 1400
},
{
"epoch": 1424.0,
"learning_rate": 7.951111111111111e-06,
"loss": 0.0,
"step": 1425
},
{
"epoch": 1449.0,
"learning_rate": 7.895555555555557e-06,
"loss": 0.0,
"step": 1450
},
{
"epoch": 1474.0,
"learning_rate": 7.840000000000001e-06,
"loss": 0.0,
"step": 1475
},
{
"epoch": 1499.0,
"learning_rate": 7.784444444444445e-06,
"loss": 0.0,
"step": 1500
},
{
"epoch": 1524.0,
"learning_rate": 7.72888888888889e-06,
"loss": 0.0,
"step": 1525
},
{
"epoch": 1549.0,
"learning_rate": 7.673333333333333e-06,
"loss": 0.0,
"step": 1550
},
{
"epoch": 1574.0,
"learning_rate": 7.617777777777778e-06,
"loss": 0.0,
"step": 1575
},
{
"epoch": 1599.0,
"learning_rate": 7.562222222222223e-06,
"loss": 0.0,
"step": 1600
},
{
"epoch": 1624.0,
"learning_rate": 7.506666666666668e-06,
"loss": 0.0,
"step": 1625
},
{
"epoch": 1649.0,
"learning_rate": 7.451111111111111e-06,
"loss": 0.0,
"step": 1650
},
{
"epoch": 1674.0,
"learning_rate": 7.395555555555556e-06,
"loss": 0.0,
"step": 1675
},
{
"epoch": 1699.0,
"learning_rate": 7.340000000000001e-06,
"loss": 0.0,
"step": 1700
},
{
"epoch": 1724.0,
"learning_rate": 7.284444444444445e-06,
"loss": 0.0,
"step": 1725
},
{
"epoch": 1749.0,
"learning_rate": 7.22888888888889e-06,
"loss": 0.0,
"step": 1750
},
{
"epoch": 1774.0,
"learning_rate": 7.173333333333335e-06,
"loss": 0.0,
"step": 1775
},
{
"epoch": 1799.0,
"learning_rate": 7.117777777777778e-06,
"loss": 0.0,
"step": 1800
},
{
"epoch": 1824.0,
"learning_rate": 7.062222222222223e-06,
"loss": 0.0,
"step": 1825
},
{
"epoch": 1849.0,
"learning_rate": 7.006666666666667e-06,
"loss": 0.0,
"step": 1850
},
{
"epoch": 1874.0,
"learning_rate": 6.951111111111112e-06,
"loss": 0.0,
"step": 1875
},
{
"epoch": 1899.0,
"learning_rate": 6.8955555555555565e-06,
"loss": 0.0,
"step": 1900
},
{
"epoch": 1924.0,
"learning_rate": 6.8400000000000014e-06,
"loss": 0.0,
"step": 1925
},
{
"epoch": 1949.0,
"learning_rate": 6.784444444444445e-06,
"loss": 0.0,
"step": 1950
},
{
"epoch": 1974.0,
"learning_rate": 6.7288888888888895e-06,
"loss": 0.0,
"step": 1975
},
{
"epoch": 1999.0,
"learning_rate": 6.6733333333333335e-06,
"loss": 0.0,
"step": 2000
},
{
"epoch": 1999.0,
"eval_loss": 0.9434587359428406,
"eval_runtime": 40.9588,
"eval_samples_per_second": 0.537,
"eval_steps_per_second": 0.073,
"eval_wer": 38.46153846153847,
"step": 2000
},
{
"epoch": 2024.0,
"learning_rate": 6.617777777777778e-06,
"loss": 0.0,
"step": 2025
},
{
"epoch": 2049.0,
"learning_rate": 6.562222222222223e-06,
"loss": 0.0,
"step": 2050
},
{
"epoch": 2074.0,
"learning_rate": 6.5066666666666665e-06,
"loss": 0.0,
"step": 2075
},
{
"epoch": 2099.0,
"learning_rate": 6.451111111111111e-06,
"loss": 0.0,
"step": 2100
},
{
"epoch": 2124.0,
"learning_rate": 6.395555555555556e-06,
"loss": 0.0,
"step": 2125
},
{
"epoch": 2149.0,
"learning_rate": 6.34e-06,
"loss": 0.0,
"step": 2150
},
{
"epoch": 2174.0,
"learning_rate": 6.284444444444445e-06,
"loss": 0.0,
"step": 2175
},
{
"epoch": 2199.0,
"learning_rate": 6.22888888888889e-06,
"loss": 0.0,
"step": 2200
},
{
"epoch": 2224.0,
"learning_rate": 6.173333333333333e-06,
"loss": 0.0,
"step": 2225
},
{
"epoch": 2249.0,
"learning_rate": 6.117777777777778e-06,
"loss": 0.0,
"step": 2250
},
{
"epoch": 2274.0,
"learning_rate": 6.062222222222223e-06,
"loss": 0.0,
"step": 2275
},
{
"epoch": 2299.0,
"learning_rate": 6.006666666666667e-06,
"loss": 0.0,
"step": 2300
},
{
"epoch": 2324.0,
"learning_rate": 5.951111111111112e-06,
"loss": 0.0,
"step": 2325
},
{
"epoch": 2349.0,
"learning_rate": 5.895555555555557e-06,
"loss": 0.0,
"step": 2350
},
{
"epoch": 2374.0,
"learning_rate": 5.84e-06,
"loss": 0.0,
"step": 2375
},
{
"epoch": 2399.0,
"learning_rate": 5.784444444444445e-06,
"loss": 0.0,
"step": 2400
},
{
"epoch": 2424.0,
"learning_rate": 5.72888888888889e-06,
"loss": 0.0,
"step": 2425
},
{
"epoch": 2449.0,
"learning_rate": 5.673333333333334e-06,
"loss": 0.0,
"step": 2450
},
{
"epoch": 2474.0,
"learning_rate": 5.617777777777779e-06,
"loss": 0.0,
"step": 2475
},
{
"epoch": 2499.0,
"learning_rate": 5.562222222222222e-06,
"loss": 0.0,
"step": 2500
},
{
"epoch": 2524.0,
"learning_rate": 5.506666666666667e-06,
"loss": 0.0,
"step": 2525
},
{
"epoch": 2549.0,
"learning_rate": 5.451111111111112e-06,
"loss": 0.0,
"step": 2550
},
{
"epoch": 2574.0,
"learning_rate": 5.3955555555555565e-06,
"loss": 0.0,
"step": 2575
},
{
"epoch": 2599.0,
"learning_rate": 5.3400000000000005e-06,
"loss": 0.0,
"step": 2600
},
{
"epoch": 2624.0,
"learning_rate": 5.2844444444444454e-06,
"loss": 0.0,
"step": 2625
},
{
"epoch": 2649.0,
"learning_rate": 5.228888888888889e-06,
"loss": 0.0,
"step": 2650
},
{
"epoch": 2674.0,
"learning_rate": 5.1733333333333335e-06,
"loss": 0.0,
"step": 2675
},
{
"epoch": 2699.0,
"learning_rate": 5.117777777777778e-06,
"loss": 0.0,
"step": 2700
},
{
"epoch": 2724.0,
"learning_rate": 5.062222222222222e-06,
"loss": 0.0,
"step": 2725
},
{
"epoch": 2749.0,
"learning_rate": 5.006666666666667e-06,
"loss": 0.0,
"step": 2750
},
{
"epoch": 2774.0,
"learning_rate": 4.951111111111111e-06,
"loss": 0.0,
"step": 2775
},
{
"epoch": 2799.0,
"learning_rate": 4.895555555555556e-06,
"loss": 0.0,
"step": 2800
},
{
"epoch": 2824.0,
"learning_rate": 4.84e-06,
"loss": 0.0,
"step": 2825
},
{
"epoch": 2849.0,
"learning_rate": 4.784444444444445e-06,
"loss": 0.0,
"step": 2850
},
{
"epoch": 2874.0,
"learning_rate": 4.728888888888889e-06,
"loss": 0.0,
"step": 2875
},
{
"epoch": 2899.0,
"learning_rate": 4.673333333333333e-06,
"loss": 0.0,
"step": 2900
},
{
"epoch": 2924.0,
"learning_rate": 4.617777777777778e-06,
"loss": 0.0,
"step": 2925
},
{
"epoch": 2949.0,
"learning_rate": 4.562222222222222e-06,
"loss": 0.0,
"step": 2950
},
{
"epoch": 2974.0,
"learning_rate": 4.506666666666667e-06,
"loss": 0.0,
"step": 2975
},
{
"epoch": 2999.0,
"learning_rate": 4.451111111111112e-06,
"loss": 0.0,
"step": 3000
},
{
"epoch": 2999.0,
"eval_loss": 1.0010170936584473,
"eval_runtime": 39.2141,
"eval_samples_per_second": 0.561,
"eval_steps_per_second": 0.077,
"eval_wer": 43.19526627218935,
"step": 3000
},
{
"epoch": 3024.0,
"learning_rate": 4.395555555555556e-06,
"loss": 0.0,
"step": 3025
},
{
"epoch": 3049.0,
"learning_rate": 4.34e-06,
"loss": 0.0,
"step": 3050
},
{
"epoch": 3074.0,
"learning_rate": 4.284444444444445e-06,
"loss": 0.0,
"step": 3075
},
{
"epoch": 3099.0,
"learning_rate": 4.228888888888889e-06,
"loss": 0.0,
"step": 3100
},
{
"epoch": 3124.0,
"learning_rate": 4.173333333333334e-06,
"loss": 0.0,
"step": 3125
},
{
"epoch": 3149.0,
"learning_rate": 4.117777777777779e-06,
"loss": 0.0,
"step": 3150
},
{
"epoch": 3174.0,
"learning_rate": 4.062222222222223e-06,
"loss": 0.0,
"step": 3175
},
{
"epoch": 3199.0,
"learning_rate": 4.006666666666667e-06,
"loss": 0.0,
"step": 3200
},
{
"epoch": 3224.0,
"learning_rate": 3.951111111111112e-06,
"loss": 0.0,
"step": 3225
},
{
"epoch": 3249.0,
"learning_rate": 3.895555555555556e-06,
"loss": 0.0,
"step": 3250
},
{
"epoch": 3274.0,
"learning_rate": 3.8400000000000005e-06,
"loss": 0.0,
"step": 3275
},
{
"epoch": 3299.0,
"learning_rate": 3.784444444444445e-06,
"loss": 0.0,
"step": 3300
},
{
"epoch": 3324.0,
"learning_rate": 3.728888888888889e-06,
"loss": 0.0,
"step": 3325
},
{
"epoch": 3349.0,
"learning_rate": 3.673333333333334e-06,
"loss": 0.0,
"step": 3350
},
{
"epoch": 3374.0,
"learning_rate": 3.617777777777778e-06,
"loss": 0.0,
"step": 3375
},
{
"epoch": 3399.0,
"learning_rate": 3.5622222222222224e-06,
"loss": 0.0,
"step": 3400
},
{
"epoch": 3424.0,
"learning_rate": 3.5066666666666673e-06,
"loss": 0.0,
"step": 3425
},
{
"epoch": 3449.0,
"learning_rate": 3.4511111111111113e-06,
"loss": 0.0,
"step": 3450
},
{
"epoch": 3474.0,
"learning_rate": 3.3955555555555558e-06,
"loss": 0.0,
"step": 3475
},
{
"epoch": 3499.0,
"learning_rate": 3.3400000000000006e-06,
"loss": 0.0,
"step": 3500
},
{
"epoch": 3524.0,
"learning_rate": 3.2844444444444447e-06,
"loss": 0.0,
"step": 3525
},
{
"epoch": 3549.0,
"learning_rate": 3.228888888888889e-06,
"loss": 0.0,
"step": 3550
},
{
"epoch": 3574.0,
"learning_rate": 3.173333333333334e-06,
"loss": 0.0,
"step": 3575
},
{
"epoch": 3599.0,
"learning_rate": 3.117777777777778e-06,
"loss": 0.0,
"step": 3600
},
{
"epoch": 3624.0,
"learning_rate": 3.0622222222222225e-06,
"loss": 0.0,
"step": 3625
},
{
"epoch": 3649.0,
"learning_rate": 3.0066666666666674e-06,
"loss": 0.0,
"step": 3650
},
{
"epoch": 3674.0,
"learning_rate": 2.9511111111111114e-06,
"loss": 0.0,
"step": 3675
},
{
"epoch": 3699.0,
"learning_rate": 2.895555555555556e-06,
"loss": 0.0,
"step": 3700
},
{
"epoch": 3724.0,
"learning_rate": 2.84e-06,
"loss": 0.0,
"step": 3725
},
{
"epoch": 3749.0,
"learning_rate": 2.784444444444445e-06,
"loss": 0.0,
"step": 3750
},
{
"epoch": 3774.0,
"learning_rate": 2.7288888888888893e-06,
"loss": 0.0,
"step": 3775
},
{
"epoch": 3799.0,
"learning_rate": 2.6733333333333333e-06,
"loss": 0.0,
"step": 3800
},
{
"epoch": 3824.0,
"learning_rate": 2.617777777777778e-06,
"loss": 0.0,
"step": 3825
},
{
"epoch": 3849.0,
"learning_rate": 2.5622222222222226e-06,
"loss": 0.0,
"step": 3850
},
{
"epoch": 3874.0,
"learning_rate": 2.5066666666666667e-06,
"loss": 0.0,
"step": 3875
},
{
"epoch": 3899.0,
"learning_rate": 2.451111111111111e-06,
"loss": 0.0,
"step": 3900
},
{
"epoch": 3924.0,
"learning_rate": 2.3955555555555556e-06,
"loss": 0.0,
"step": 3925
},
{
"epoch": 3949.0,
"learning_rate": 2.3400000000000005e-06,
"loss": 0.0,
"step": 3950
},
{
"epoch": 3974.0,
"learning_rate": 2.2844444444444445e-06,
"loss": 0.0,
"step": 3975
},
{
"epoch": 3999.0,
"learning_rate": 2.228888888888889e-06,
"loss": 0.0,
"step": 4000
},
{
"epoch": 3999.0,
"eval_loss": 1.038047432899475,
"eval_runtime": 38.9006,
"eval_samples_per_second": 0.566,
"eval_steps_per_second": 0.077,
"eval_wer": 44.37869822485207,
"step": 4000
},
{
"epoch": 4024.0,
"learning_rate": 2.1733333333333334e-06,
"loss": 0.0,
"step": 4025
},
{
"epoch": 4049.0,
"learning_rate": 2.117777777777778e-06,
"loss": 0.0,
"step": 4050
},
{
"epoch": 4074.0,
"learning_rate": 2.0622222222222223e-06,
"loss": 0.0,
"step": 4075
},
{
"epoch": 4099.0,
"learning_rate": 2.006666666666667e-06,
"loss": 0.0,
"step": 4100
},
{
"epoch": 4124.0,
"learning_rate": 1.9511111111111113e-06,
"loss": 0.0,
"step": 4125
},
{
"epoch": 4149.0,
"learning_rate": 1.8955555555555557e-06,
"loss": 0.0,
"step": 4150
},
{
"epoch": 4174.0,
"learning_rate": 1.8400000000000002e-06,
"loss": 0.0,
"step": 4175
},
{
"epoch": 4199.0,
"learning_rate": 1.7844444444444444e-06,
"loss": 0.0,
"step": 4200
},
{
"epoch": 4224.0,
"learning_rate": 1.728888888888889e-06,
"loss": 0.0,
"step": 4225
},
{
"epoch": 4249.0,
"learning_rate": 1.6733333333333335e-06,
"loss": 0.0,
"step": 4250
},
{
"epoch": 4274.0,
"learning_rate": 1.6177777777777778e-06,
"loss": 0.0,
"step": 4275
},
{
"epoch": 4299.0,
"learning_rate": 1.5622222222222225e-06,
"loss": 0.0,
"step": 4300
},
{
"epoch": 4324.0,
"learning_rate": 1.506666666666667e-06,
"loss": 0.0,
"step": 4325
},
{
"epoch": 4349.0,
"learning_rate": 1.4511111111111112e-06,
"loss": 0.0,
"step": 4350
},
{
"epoch": 4374.0,
"learning_rate": 1.3955555555555556e-06,
"loss": 0.0,
"step": 4375
},
{
"epoch": 4399.0,
"learning_rate": 1.34e-06,
"loss": 0.0,
"step": 4400
},
{
"epoch": 4424.0,
"learning_rate": 1.2844444444444445e-06,
"loss": 0.0,
"step": 4425
},
{
"epoch": 4449.0,
"learning_rate": 1.228888888888889e-06,
"loss": 0.0,
"step": 4450
},
{
"epoch": 4474.0,
"learning_rate": 1.1733333333333335e-06,
"loss": 0.0,
"step": 4475
},
{
"epoch": 4499.0,
"learning_rate": 1.117777777777778e-06,
"loss": 0.0,
"step": 4500
},
{
"epoch": 4524.0,
"learning_rate": 1.0622222222222222e-06,
"loss": 0.0,
"step": 4525
},
{
"epoch": 4549.0,
"learning_rate": 1.0066666666666668e-06,
"loss": 0.0,
"step": 4550
},
{
"epoch": 4574.0,
"learning_rate": 9.511111111111111e-07,
"loss": 0.0,
"step": 4575
},
{
"epoch": 4599.0,
"learning_rate": 8.955555555555557e-07,
"loss": 0.0,
"step": 4600
},
{
"epoch": 4624.0,
"learning_rate": 8.400000000000001e-07,
"loss": 0.0,
"step": 4625
},
{
"epoch": 4649.0,
"learning_rate": 7.844444444444445e-07,
"loss": 0.0,
"step": 4650
},
{
"epoch": 4674.0,
"learning_rate": 7.28888888888889e-07,
"loss": 0.0,
"step": 4675
},
{
"epoch": 4699.0,
"learning_rate": 6.733333333333334e-07,
"loss": 0.0,
"step": 4700
},
{
"epoch": 4724.0,
"learning_rate": 6.177777777777778e-07,
"loss": 0.0,
"step": 4725
},
{
"epoch": 4749.0,
"learning_rate": 5.622222222222223e-07,
"loss": 0.0,
"step": 4750
},
{
"epoch": 4774.0,
"learning_rate": 5.066666666666667e-07,
"loss": 0.0,
"step": 4775
},
{
"epoch": 4799.0,
"learning_rate": 4.511111111111111e-07,
"loss": 0.0,
"step": 4800
},
{
"epoch": 4824.0,
"learning_rate": 3.9555555555555557e-07,
"loss": 0.0,
"step": 4825
},
{
"epoch": 4849.0,
"learning_rate": 3.4000000000000003e-07,
"loss": 0.0,
"step": 4850
},
{
"epoch": 4874.0,
"learning_rate": 2.844444444444445e-07,
"loss": 0.0,
"step": 4875
},
{
"epoch": 4899.0,
"learning_rate": 2.2888888888888892e-07,
"loss": 0.0,
"step": 4900
},
{
"epoch": 4924.0,
"learning_rate": 1.7333333333333335e-07,
"loss": 0.0,
"step": 4925
},
{
"epoch": 4949.0,
"learning_rate": 1.1777777777777778e-07,
"loss": 0.0,
"step": 4950
},
{
"epoch": 4974.0,
"learning_rate": 6.222222222222223e-08,
"loss": 0.0,
"step": 4975
},
{
"epoch": 4999.0,
"learning_rate": 6.666666666666667e-09,
"loss": 0.0,
"step": 5000
},
{
"epoch": 4999.0,
"eval_loss": 1.0529390573501587,
"eval_runtime": 38.0353,
"eval_samples_per_second": 0.578,
"eval_steps_per_second": 0.079,
"eval_wer": 43.786982248520715,
"step": 5000
},
{
"epoch": 4999.0,
"step": 5000,
"total_flos": 4.246172187918336e+20,
"train_loss": 0.007115483383736864,
"train_runtime": 57931.0141,
"train_samples_per_second": 2.762,
"train_steps_per_second": 0.086
}
],
"max_steps": 5000,
"num_train_epochs": 9223372036854775807,
"total_flos": 4.246172187918336e+20,
"trial_name": null,
"trial_params": null
}