learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.5553	2.4752	500	0.4724	0.9855
0.3331	4.9505	1000	0.2739	0.9504
0.2707	7.4257	1500	0.2362	0.9635
0.2411	9.9010	2000	0.2191	0.9122
0.2246	12.3762	2500	0.2068	0.8596
0.2088	14.8515	3000	0.1971	0.8365
0.1951	17.3267	3500	0.1918	0.8261
0.187	19.8020	4000	0.1864	0.8227
0.1769	22.2772	4500	0.1851	0.8210
0.1728	24.7525	5000	0.1811	0.8250
0.1682	27.2277	5500	0.1730	0.8135
0.1662	29.7030	6000	0.1708	0.7941

nh0znoisung/hubert-large-timit-upsample-4-50

hubert-large-timit-upsample-4-50