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
1.1292	0.1786	5	1.3299	44.6892
1.0845	0.3571	10	1.2101	43.3638
0.939	0.5357	15	1.0710	42.3838
0.9406	0.7143	20	0.9471	40.2277
0.7859	0.8929	25	0.8397	33.8809
0.6389	1.0714	30	0.7542	30.5768
0.6165	1.25	35	0.6979	22.2326
0.5693	1.4286	40	0.6582	22.0646
0.6091	1.6071	45	0.6262	20.6646
0.5544	1.7857	50	0.6023	22.3913
0.5429	1.9643	55	0.5842	20.2632
0.5064	2.1429	60	0.5708	20.1232
0.5022	2.3214	65	0.5619	18.4432
0.5135	2.5	70	0.5562	19.6005
0.4744	2.6786	75	0.5523	19.2832
0.4698	2.8571	80	0.5500	20.1699

greenw0lf/whisper-nonnat-5h-mmr-xlsr

whisper-nonnat-5h-mmr-xlsr