whisper-large-v2-amet / trainer_state.json
drmeeseeks's picture
End of training
8b48c2c
raw
history blame contribute delete
No virus
25.7 kB
{
"best_metric": 102.80748663101605,
"best_model_checkpoint": "./whisper-large-v2-amet/checkpoint-3000",
"epoch": 5000.0,
"global_step": 5000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 25.0,
"learning_rate": 4.0000000000000003e-07,
"loss": 2.7655,
"step": 25
},
{
"epoch": 50.0,
"learning_rate": 9.000000000000001e-07,
"loss": 1.7532,
"step": 50
},
{
"epoch": 75.0,
"learning_rate": 1.4000000000000001e-06,
"loss": 0.3058,
"step": 75
},
{
"epoch": 100.0,
"learning_rate": 1.9000000000000002e-06,
"loss": 0.0006,
"step": 100
},
{
"epoch": 125.0,
"learning_rate": 2.4000000000000003e-06,
"loss": 0.0001,
"step": 125
},
{
"epoch": 150.0,
"learning_rate": 2.9e-06,
"loss": 0.0001,
"step": 150
},
{
"epoch": 175.0,
"learning_rate": 3.4000000000000005e-06,
"loss": 0.0001,
"step": 175
},
{
"epoch": 200.0,
"learning_rate": 3.900000000000001e-06,
"loss": 0.0001,
"step": 200
},
{
"epoch": 225.0,
"learning_rate": 4.4e-06,
"loss": 0.0001,
"step": 225
},
{
"epoch": 250.0,
"learning_rate": 4.9000000000000005e-06,
"loss": 0.0001,
"step": 250
},
{
"epoch": 275.0,
"learning_rate": 5.400000000000001e-06,
"loss": 0.0001,
"step": 275
},
{
"epoch": 300.0,
"learning_rate": 5.9e-06,
"loss": 0.0001,
"step": 300
},
{
"epoch": 325.0,
"learning_rate": 6.4000000000000006e-06,
"loss": 0.0001,
"step": 325
},
{
"epoch": 350.0,
"learning_rate": 6.9e-06,
"loss": 0.0001,
"step": 350
},
{
"epoch": 375.0,
"learning_rate": 7.4e-06,
"loss": 0.0001,
"step": 375
},
{
"epoch": 400.0,
"learning_rate": 7.9e-06,
"loss": 0.0001,
"step": 400
},
{
"epoch": 425.0,
"learning_rate": 8.400000000000001e-06,
"loss": 0.0001,
"step": 425
},
{
"epoch": 450.0,
"learning_rate": 8.900000000000001e-06,
"loss": 0.0001,
"step": 450
},
{
"epoch": 475.0,
"learning_rate": 9.4e-06,
"loss": 0.0001,
"step": 475
},
{
"epoch": 500.0,
"learning_rate": 9.9e-06,
"loss": 0.0001,
"step": 500
},
{
"epoch": 525.0,
"learning_rate": 9.955555555555556e-06,
"loss": 0.0001,
"step": 525
},
{
"epoch": 550.0,
"learning_rate": 9.9e-06,
"loss": 0.0,
"step": 550
},
{
"epoch": 575.0,
"learning_rate": 9.844444444444446e-06,
"loss": 0.0,
"step": 575
},
{
"epoch": 600.0,
"learning_rate": 9.78888888888889e-06,
"loss": 0.0,
"step": 600
},
{
"epoch": 625.0,
"learning_rate": 9.733333333333334e-06,
"loss": 0.0001,
"step": 625
},
{
"epoch": 650.0,
"learning_rate": 9.677777777777778e-06,
"loss": 0.0001,
"step": 650
},
{
"epoch": 675.0,
"learning_rate": 9.622222222222222e-06,
"loss": 0.0001,
"step": 675
},
{
"epoch": 700.0,
"learning_rate": 9.566666666666668e-06,
"loss": 0.0001,
"step": 700
},
{
"epoch": 725.0,
"learning_rate": 9.511111111111112e-06,
"loss": 0.0,
"step": 725
},
{
"epoch": 750.0,
"learning_rate": 9.455555555555557e-06,
"loss": 0.0,
"step": 750
},
{
"epoch": 775.0,
"learning_rate": 9.4e-06,
"loss": 0.0,
"step": 775
},
{
"epoch": 800.0,
"learning_rate": 9.344444444444446e-06,
"loss": 0.0,
"step": 800
},
{
"epoch": 825.0,
"learning_rate": 9.28888888888889e-06,
"loss": 0.0,
"step": 825
},
{
"epoch": 850.0,
"learning_rate": 9.233333333333334e-06,
"loss": 0.0,
"step": 850
},
{
"epoch": 875.0,
"learning_rate": 9.17777777777778e-06,
"loss": 0.0,
"step": 875
},
{
"epoch": 900.0,
"learning_rate": 9.122222222222223e-06,
"loss": 0.0,
"step": 900
},
{
"epoch": 925.0,
"learning_rate": 9.066666666666667e-06,
"loss": 0.0,
"step": 925
},
{
"epoch": 950.0,
"learning_rate": 9.011111111111111e-06,
"loss": 0.0,
"step": 950
},
{
"epoch": 975.0,
"learning_rate": 8.955555555555555e-06,
"loss": 0.0,
"step": 975
},
{
"epoch": 1000.0,
"learning_rate": 8.900000000000001e-06,
"loss": 0.0,
"step": 1000
},
{
"epoch": 1000.0,
"eval_loss": 8.382182121276855,
"eval_runtime": 384.7534,
"eval_samples_per_second": 0.13,
"eval_steps_per_second": 0.003,
"eval_wer": 156.01604278074865,
"step": 1000
},
{
"epoch": 1025.0,
"learning_rate": 8.844444444444445e-06,
"loss": 0.0,
"step": 1025
},
{
"epoch": 1050.0,
"learning_rate": 8.788888888888891e-06,
"loss": 0.0,
"step": 1050
},
{
"epoch": 1075.0,
"learning_rate": 8.733333333333333e-06,
"loss": 0.0,
"step": 1075
},
{
"epoch": 1100.0,
"learning_rate": 8.677777777777779e-06,
"loss": 0.0,
"step": 1100
},
{
"epoch": 1125.0,
"learning_rate": 8.622222222222223e-06,
"loss": 0.0,
"step": 1125
},
{
"epoch": 1150.0,
"learning_rate": 8.566666666666667e-06,
"loss": 0.0,
"step": 1150
},
{
"epoch": 1175.0,
"learning_rate": 8.511111111111113e-06,
"loss": 0.0,
"step": 1175
},
{
"epoch": 1200.0,
"learning_rate": 8.455555555555555e-06,
"loss": 0.0,
"step": 1200
},
{
"epoch": 1225.0,
"learning_rate": 8.400000000000001e-06,
"loss": 0.0,
"step": 1225
},
{
"epoch": 1250.0,
"learning_rate": 8.344444444444445e-06,
"loss": 0.0,
"step": 1250
},
{
"epoch": 1275.0,
"learning_rate": 8.288888888888889e-06,
"loss": 0.0,
"step": 1275
},
{
"epoch": 1300.0,
"learning_rate": 8.233333333333335e-06,
"loss": 0.0,
"step": 1300
},
{
"epoch": 1325.0,
"learning_rate": 8.177777777777779e-06,
"loss": 0.0,
"step": 1325
},
{
"epoch": 1350.0,
"learning_rate": 8.122222222222223e-06,
"loss": 0.0,
"step": 1350
},
{
"epoch": 1375.0,
"learning_rate": 8.066666666666667e-06,
"loss": 0.0,
"step": 1375
},
{
"epoch": 1400.0,
"learning_rate": 8.011111111111113e-06,
"loss": 0.0,
"step": 1400
},
{
"epoch": 1425.0,
"learning_rate": 7.955555555555557e-06,
"loss": 0.0001,
"step": 1425
},
{
"epoch": 1450.0,
"learning_rate": 7.9e-06,
"loss": 0.0,
"step": 1450
},
{
"epoch": 1475.0,
"learning_rate": 7.844444444444446e-06,
"loss": 0.0,
"step": 1475
},
{
"epoch": 1500.0,
"learning_rate": 7.788888888888889e-06,
"loss": 0.0,
"step": 1500
},
{
"epoch": 1525.0,
"learning_rate": 7.733333333333334e-06,
"loss": 0.0001,
"step": 1525
},
{
"epoch": 1550.0,
"learning_rate": 7.677777777777778e-06,
"loss": 0.0,
"step": 1550
},
{
"epoch": 1575.0,
"learning_rate": 7.622222222222223e-06,
"loss": 0.0,
"step": 1575
},
{
"epoch": 1600.0,
"learning_rate": 7.566666666666667e-06,
"loss": 0.0,
"step": 1600
},
{
"epoch": 1625.0,
"learning_rate": 7.511111111111111e-06,
"loss": 0.0,
"step": 1625
},
{
"epoch": 1650.0,
"learning_rate": 7.455555555555556e-06,
"loss": 0.0,
"step": 1650
},
{
"epoch": 1675.0,
"learning_rate": 7.4e-06,
"loss": 0.0,
"step": 1675
},
{
"epoch": 1700.0,
"learning_rate": 7.344444444444445e-06,
"loss": 0.0,
"step": 1700
},
{
"epoch": 1725.0,
"learning_rate": 7.28888888888889e-06,
"loss": 0.0,
"step": 1725
},
{
"epoch": 1750.0,
"learning_rate": 7.233333333333334e-06,
"loss": 0.0,
"step": 1750
},
{
"epoch": 1775.0,
"learning_rate": 7.177777777777778e-06,
"loss": 0.0,
"step": 1775
},
{
"epoch": 1800.0,
"learning_rate": 7.122222222222222e-06,
"loss": 0.0,
"step": 1800
},
{
"epoch": 1825.0,
"learning_rate": 7.066666666666667e-06,
"loss": 0.0,
"step": 1825
},
{
"epoch": 1850.0,
"learning_rate": 7.011111111111112e-06,
"loss": 0.0,
"step": 1850
},
{
"epoch": 1875.0,
"learning_rate": 6.955555555555557e-06,
"loss": 0.0,
"step": 1875
},
{
"epoch": 1900.0,
"learning_rate": 6.9e-06,
"loss": 0.0,
"step": 1900
},
{
"epoch": 1925.0,
"learning_rate": 6.844444444444445e-06,
"loss": 0.0,
"step": 1925
},
{
"epoch": 1950.0,
"learning_rate": 6.788888888888889e-06,
"loss": 0.0,
"step": 1950
},
{
"epoch": 1975.0,
"learning_rate": 6.733333333333334e-06,
"loss": 0.0,
"step": 1975
},
{
"epoch": 2000.0,
"learning_rate": 6.677777777777779e-06,
"loss": 0.0,
"step": 2000
},
{
"epoch": 2000.0,
"eval_loss": 9.796119689941406,
"eval_runtime": 384.86,
"eval_samples_per_second": 0.13,
"eval_steps_per_second": 0.003,
"eval_wer": 110.42780748663101,
"step": 2000
},
{
"epoch": 2025.0,
"learning_rate": 6.6222222222222236e-06,
"loss": 0.0,
"step": 2025
},
{
"epoch": 2050.0,
"learning_rate": 6.566666666666667e-06,
"loss": 0.0,
"step": 2050
},
{
"epoch": 2075.0,
"learning_rate": 6.511111111111112e-06,
"loss": 0.0,
"step": 2075
},
{
"epoch": 2100.0,
"learning_rate": 6.455555555555556e-06,
"loss": 0.0,
"step": 2100
},
{
"epoch": 2125.0,
"learning_rate": 6.4000000000000006e-06,
"loss": 0.0,
"step": 2125
},
{
"epoch": 2150.0,
"learning_rate": 6.3444444444444454e-06,
"loss": 0.0,
"step": 2150
},
{
"epoch": 2175.0,
"learning_rate": 6.28888888888889e-06,
"loss": 0.0,
"step": 2175
},
{
"epoch": 2200.0,
"learning_rate": 6.2333333333333335e-06,
"loss": 0.0,
"step": 2200
},
{
"epoch": 2225.0,
"learning_rate": 6.177777777777778e-06,
"loss": 0.0,
"step": 2225
},
{
"epoch": 2250.0,
"learning_rate": 6.1222222222222224e-06,
"loss": 0.0,
"step": 2250
},
{
"epoch": 2275.0,
"learning_rate": 6.066666666666667e-06,
"loss": 0.0,
"step": 2275
},
{
"epoch": 2300.0,
"learning_rate": 6.011111111111112e-06,
"loss": 0.0,
"step": 2300
},
{
"epoch": 2325.0,
"learning_rate": 5.955555555555555e-06,
"loss": 0.0,
"step": 2325
},
{
"epoch": 2350.0,
"learning_rate": 5.9e-06,
"loss": 0.0,
"step": 2350
},
{
"epoch": 2375.0,
"learning_rate": 5.844444444444445e-06,
"loss": 0.0,
"step": 2375
},
{
"epoch": 2400.0,
"learning_rate": 5.788888888888889e-06,
"loss": 0.0,
"step": 2400
},
{
"epoch": 2425.0,
"learning_rate": 5.733333333333334e-06,
"loss": 0.0,
"step": 2425
},
{
"epoch": 2450.0,
"learning_rate": 5.677777777777779e-06,
"loss": 0.0,
"step": 2450
},
{
"epoch": 2475.0,
"learning_rate": 5.622222222222222e-06,
"loss": 0.0,
"step": 2475
},
{
"epoch": 2500.0,
"learning_rate": 5.566666666666667e-06,
"loss": 0.0,
"step": 2500
},
{
"epoch": 2525.0,
"learning_rate": 5.511111111111112e-06,
"loss": 0.0,
"step": 2525
},
{
"epoch": 2550.0,
"learning_rate": 5.455555555555556e-06,
"loss": 0.0,
"step": 2550
},
{
"epoch": 2575.0,
"learning_rate": 5.400000000000001e-06,
"loss": 0.0,
"step": 2575
},
{
"epoch": 2600.0,
"learning_rate": 5.344444444444446e-06,
"loss": 0.0,
"step": 2600
},
{
"epoch": 2625.0,
"learning_rate": 5.288888888888889e-06,
"loss": 0.0,
"step": 2625
},
{
"epoch": 2650.0,
"learning_rate": 5.233333333333334e-06,
"loss": 0.0,
"step": 2650
},
{
"epoch": 2675.0,
"learning_rate": 5.177777777777779e-06,
"loss": 0.0,
"step": 2675
},
{
"epoch": 2700.0,
"learning_rate": 5.122222222222223e-06,
"loss": 0.0,
"step": 2700
},
{
"epoch": 2725.0,
"learning_rate": 5.0666666666666676e-06,
"loss": 0.0,
"step": 2725
},
{
"epoch": 2750.0,
"learning_rate": 5.011111111111111e-06,
"loss": 0.0,
"step": 2750
},
{
"epoch": 2775.0,
"learning_rate": 4.9555555555555565e-06,
"loss": 0.0,
"step": 2775
},
{
"epoch": 2800.0,
"learning_rate": 4.9000000000000005e-06,
"loss": 0.0,
"step": 2800
},
{
"epoch": 2825.0,
"learning_rate": 4.8444444444444446e-06,
"loss": 0.0,
"step": 2825
},
{
"epoch": 2850.0,
"learning_rate": 4.7888888888888894e-06,
"loss": 0.0,
"step": 2850
},
{
"epoch": 2875.0,
"learning_rate": 4.7333333333333335e-06,
"loss": 0.0,
"step": 2875
},
{
"epoch": 2900.0,
"learning_rate": 4.677777777777778e-06,
"loss": 0.0,
"step": 2900
},
{
"epoch": 2925.0,
"learning_rate": 4.622222222222222e-06,
"loss": 0.0,
"step": 2925
},
{
"epoch": 2950.0,
"learning_rate": 4.566666666666667e-06,
"loss": 0.0,
"step": 2950
},
{
"epoch": 2975.0,
"learning_rate": 4.511111111111111e-06,
"loss": 0.0,
"step": 2975
},
{
"epoch": 3000.0,
"learning_rate": 4.455555555555555e-06,
"loss": 0.0,
"step": 3000
},
{
"epoch": 3000.0,
"eval_loss": 12.001350402832031,
"eval_runtime": 384.6328,
"eval_samples_per_second": 0.13,
"eval_steps_per_second": 0.003,
"eval_wer": 102.80748663101605,
"step": 3000
},
{
"epoch": 3025.0,
"learning_rate": 4.4e-06,
"loss": 0.0,
"step": 3025
},
{
"epoch": 3050.0,
"learning_rate": 4.344444444444445e-06,
"loss": 0.0,
"step": 3050
},
{
"epoch": 3075.0,
"learning_rate": 4.288888888888889e-06,
"loss": 0.0,
"step": 3075
},
{
"epoch": 3100.0,
"learning_rate": 4.233333333333334e-06,
"loss": 0.0,
"step": 3100
},
{
"epoch": 3125.0,
"learning_rate": 4.177777777777778e-06,
"loss": 0.0,
"step": 3125
},
{
"epoch": 3150.0,
"learning_rate": 4.122222222222222e-06,
"loss": 0.0,
"step": 3150
},
{
"epoch": 3175.0,
"learning_rate": 4.066666666666667e-06,
"loss": 0.0,
"step": 3175
},
{
"epoch": 3200.0,
"learning_rate": 4.011111111111111e-06,
"loss": 0.0,
"step": 3200
},
{
"epoch": 3225.0,
"learning_rate": 3.955555555555556e-06,
"loss": 0.0,
"step": 3225
},
{
"epoch": 3250.0,
"learning_rate": 3.900000000000001e-06,
"loss": 0.0,
"step": 3250
},
{
"epoch": 3275.0,
"learning_rate": 3.844444444444445e-06,
"loss": 0.0,
"step": 3275
},
{
"epoch": 3300.0,
"learning_rate": 3.7888888888888893e-06,
"loss": 0.0,
"step": 3300
},
{
"epoch": 3325.0,
"learning_rate": 3.7333333333333337e-06,
"loss": 0.0,
"step": 3325
},
{
"epoch": 3350.0,
"learning_rate": 3.6777777777777778e-06,
"loss": 0.0,
"step": 3350
},
{
"epoch": 3375.0,
"learning_rate": 3.6222222222222226e-06,
"loss": 0.0,
"step": 3375
},
{
"epoch": 3400.0,
"learning_rate": 3.566666666666667e-06,
"loss": 0.0,
"step": 3400
},
{
"epoch": 3425.0,
"learning_rate": 3.511111111111111e-06,
"loss": 0.0,
"step": 3425
},
{
"epoch": 3450.0,
"learning_rate": 3.455555555555556e-06,
"loss": 0.0,
"step": 3450
},
{
"epoch": 3475.0,
"learning_rate": 3.4000000000000005e-06,
"loss": 0.0,
"step": 3475
},
{
"epoch": 3500.0,
"learning_rate": 3.3444444444444445e-06,
"loss": 0.0,
"step": 3500
},
{
"epoch": 3525.0,
"learning_rate": 3.2888888888888894e-06,
"loss": 0.0,
"step": 3525
},
{
"epoch": 3550.0,
"learning_rate": 3.2333333333333334e-06,
"loss": 0.0,
"step": 3550
},
{
"epoch": 3575.0,
"learning_rate": 3.177777777777778e-06,
"loss": 0.0,
"step": 3575
},
{
"epoch": 3600.0,
"learning_rate": 3.1222222222222228e-06,
"loss": 0.0,
"step": 3600
},
{
"epoch": 3625.0,
"learning_rate": 3.066666666666667e-06,
"loss": 0.0,
"step": 3625
},
{
"epoch": 3650.0,
"learning_rate": 3.0111111111111113e-06,
"loss": 0.0,
"step": 3650
},
{
"epoch": 3675.0,
"learning_rate": 2.955555555555556e-06,
"loss": 0.0,
"step": 3675
},
{
"epoch": 3700.0,
"learning_rate": 2.9e-06,
"loss": 0.0,
"step": 3700
},
{
"epoch": 3725.0,
"learning_rate": 2.8444444444444446e-06,
"loss": 0.0,
"step": 3725
},
{
"epoch": 3750.0,
"learning_rate": 2.788888888888889e-06,
"loss": 0.0,
"step": 3750
},
{
"epoch": 3775.0,
"learning_rate": 2.7333333333333336e-06,
"loss": 0.0,
"step": 3775
},
{
"epoch": 3800.0,
"learning_rate": 2.677777777777778e-06,
"loss": 0.0,
"step": 3800
},
{
"epoch": 3825.0,
"learning_rate": 2.6222222222222225e-06,
"loss": 0.0,
"step": 3825
},
{
"epoch": 3850.0,
"learning_rate": 2.566666666666667e-06,
"loss": 0.0,
"step": 3850
},
{
"epoch": 3875.0,
"learning_rate": 2.5111111111111114e-06,
"loss": 0.0,
"step": 3875
},
{
"epoch": 3900.0,
"learning_rate": 2.455555555555556e-06,
"loss": 0.0,
"step": 3900
},
{
"epoch": 3925.0,
"learning_rate": 2.4000000000000003e-06,
"loss": 0.0,
"step": 3925
},
{
"epoch": 3950.0,
"learning_rate": 2.3444444444444448e-06,
"loss": 0.0,
"step": 3950
},
{
"epoch": 3975.0,
"learning_rate": 2.2888888888888892e-06,
"loss": 0.0,
"step": 3975
},
{
"epoch": 4000.0,
"learning_rate": 2.2333333333333333e-06,
"loss": 0.0,
"step": 4000
},
{
"epoch": 4000.0,
"eval_loss": 12.263284683227539,
"eval_runtime": 384.6578,
"eval_samples_per_second": 0.13,
"eval_steps_per_second": 0.003,
"eval_wer": 103.34224598930481,
"step": 4000
},
{
"epoch": 4025.0,
"learning_rate": 2.1777777777777777e-06,
"loss": 0.0,
"step": 4025
},
{
"epoch": 4050.0,
"learning_rate": 2.1222222222222226e-06,
"loss": 0.0,
"step": 4050
},
{
"epoch": 4075.0,
"learning_rate": 2.0666666666666666e-06,
"loss": 0.0,
"step": 4075
},
{
"epoch": 4100.0,
"learning_rate": 2.011111111111111e-06,
"loss": 0.0,
"step": 4100
},
{
"epoch": 4125.0,
"learning_rate": 1.955555555555556e-06,
"loss": 0.0,
"step": 4125
},
{
"epoch": 4150.0,
"learning_rate": 1.9000000000000002e-06,
"loss": 0.0,
"step": 4150
},
{
"epoch": 4175.0,
"learning_rate": 1.8444444444444445e-06,
"loss": 0.0,
"step": 4175
},
{
"epoch": 4200.0,
"learning_rate": 1.788888888888889e-06,
"loss": 0.0,
"step": 4200
},
{
"epoch": 4225.0,
"learning_rate": 1.7333333333333336e-06,
"loss": 0.0,
"step": 4225
},
{
"epoch": 4250.0,
"learning_rate": 1.6777777777777779e-06,
"loss": 0.0,
"step": 4250
},
{
"epoch": 4275.0,
"learning_rate": 1.6222222222222223e-06,
"loss": 0.0,
"step": 4275
},
{
"epoch": 4300.0,
"learning_rate": 1.566666666666667e-06,
"loss": 0.0,
"step": 4300
},
{
"epoch": 4325.0,
"learning_rate": 1.5111111111111112e-06,
"loss": 0.0,
"step": 4325
},
{
"epoch": 4350.0,
"learning_rate": 1.4555555555555557e-06,
"loss": 0.0,
"step": 4350
},
{
"epoch": 4375.0,
"learning_rate": 1.4000000000000001e-06,
"loss": 0.0,
"step": 4375
},
{
"epoch": 4400.0,
"learning_rate": 1.3444444444444446e-06,
"loss": 0.0,
"step": 4400
},
{
"epoch": 4425.0,
"learning_rate": 1.288888888888889e-06,
"loss": 0.0,
"step": 4425
},
{
"epoch": 4450.0,
"learning_rate": 1.2333333333333335e-06,
"loss": 0.0,
"step": 4450
},
{
"epoch": 4475.0,
"learning_rate": 1.1777777777777778e-06,
"loss": 0.0,
"step": 4475
},
{
"epoch": 4500.0,
"learning_rate": 1.1222222222222222e-06,
"loss": 0.0,
"step": 4500
},
{
"epoch": 4525.0,
"learning_rate": 1.066666666666667e-06,
"loss": 0.0,
"step": 4525
},
{
"epoch": 4550.0,
"learning_rate": 1.0111111111111111e-06,
"loss": 0.0,
"step": 4550
},
{
"epoch": 4575.0,
"learning_rate": 9.555555555555556e-07,
"loss": 0.0,
"step": 4575
},
{
"epoch": 4600.0,
"learning_rate": 9.000000000000001e-07,
"loss": 0.0,
"step": 4600
},
{
"epoch": 4625.0,
"learning_rate": 8.444444444444445e-07,
"loss": 0.0,
"step": 4625
},
{
"epoch": 4650.0,
"learning_rate": 7.888888888888889e-07,
"loss": 0.0,
"step": 4650
},
{
"epoch": 4675.0,
"learning_rate": 7.333333333333334e-07,
"loss": 0.0,
"step": 4675
},
{
"epoch": 4700.0,
"learning_rate": 6.777777777777779e-07,
"loss": 0.0,
"step": 4700
},
{
"epoch": 4725.0,
"learning_rate": 6.222222222222223e-07,
"loss": 0.0,
"step": 4725
},
{
"epoch": 4750.0,
"learning_rate": 5.666666666666667e-07,
"loss": 0.0,
"step": 4750
},
{
"epoch": 4775.0,
"learning_rate": 5.111111111111112e-07,
"loss": 0.0,
"step": 4775
},
{
"epoch": 4800.0,
"learning_rate": 4.5555555555555563e-07,
"loss": 0.0,
"step": 4800
},
{
"epoch": 4825.0,
"learning_rate": 4.0000000000000003e-07,
"loss": 0.0,
"step": 4825
},
{
"epoch": 4850.0,
"learning_rate": 3.444444444444445e-07,
"loss": 0.0,
"step": 4850
},
{
"epoch": 4875.0,
"learning_rate": 2.888888888888889e-07,
"loss": 0.0,
"step": 4875
},
{
"epoch": 4900.0,
"learning_rate": 2.3333333333333336e-07,
"loss": 0.0,
"step": 4900
},
{
"epoch": 4925.0,
"learning_rate": 1.777777777777778e-07,
"loss": 0.0,
"step": 4925
},
{
"epoch": 4950.0,
"learning_rate": 1.2222222222222225e-07,
"loss": 0.0,
"step": 4950
},
{
"epoch": 4975.0,
"learning_rate": 6.666666666666668e-08,
"loss": 0.0,
"step": 4975
},
{
"epoch": 5000.0,
"learning_rate": 1.1111111111111112e-08,
"loss": 0.0,
"step": 5000
},
{
"epoch": 5000.0,
"eval_loss": 12.240767478942871,
"eval_runtime": 384.5003,
"eval_samples_per_second": 0.13,
"eval_steps_per_second": 0.003,
"eval_wer": 102.94117647058823,
"step": 5000
},
{
"epoch": 5000.0,
"step": 5000,
"total_flos": 1.0615855104e+19,
"train_loss": 0.024147475125084566,
"train_runtime": 5167.5516,
"train_samples_per_second": 123.85,
"train_steps_per_second": 0.968
}
],
"max_steps": 5000,
"num_train_epochs": 5000,
"total_flos": 1.0615855104e+19,
"trial_name": null,
"trial_params": null
}