learning_rate: 0.0001
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 32
optimizer: Use OptimizerNames.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: 1000
training_steps: 3500
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.16	2.7056	500	0.2363	0.1710	0.0594
0.1467	5.4071	1000	0.2485	0.1882	0.0678
0.1392	8.1085	1500	0.2382	0.1650	0.0568
0.11	10.8141	2000	0.2240	0.1623	0.0568
0.1006	13.5156	2500	0.1995	0.1510	0.0519
0.0788	16.2171	3000	0.2202	0.1454	0.0507
0.0678	18.9227	3500	0.1992	0.1411	0.0495

AndersenC4/wav2vec2-medical

wav2vec2-medical