learning_rate: 0.0001
train_batch_size: 48
eval_batch_size: 32
seed: 42
optimizer: Use adamw_torch_fused 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: 8
num_epochs: 3.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
1.3855	0.1786	5	1.3324	44.2692
1.4496	0.3571	10	1.2169	42.1878
1.1894	0.5357	15	1.0753	39.6771
0.8815	0.7143	20	0.9507	39.3317
0.8743	0.8929	25	0.8451	32.7049
0.7159	1.0714	30	0.7612	29.3821
0.6921	1.25	35	0.7015	22.5033
0.7305	1.4286	40	0.6655	21.8779
0.7068	1.6071	45	0.6364	21.4859
0.573	1.7857	50	0.6118	20.7299
0.7105	1.9643	55	0.5943	18.7605
0.687	2.1429	60	0.5809	22.5126
0.6132	2.3214	65	0.5696	20.7952
0.6703	2.5	70	0.5614	22.3446
0.6001	2.6786	75	0.5558	19.2738
0.5669	2.8571	80	0.5525	19.2365

greenw0lf/whisper-nonnat-5h-meta-seed333

whisper-nonnat-5h-meta-seed333