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: 73
num_epochs: 3.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.4319	0.2980	73	0.4746	23.6455
0.3107	0.5959	146	0.3979	19.2539
0.5531	0.8939	219	0.3770	18.7775
0.3599	1.1918	292	0.3672	17.9924
0.3292	1.4898	365	0.3600	17.9622
0.2894	1.7878	438	0.3585	17.3080
0.1794	2.0857	511	0.3540	18.3413
0.2068	2.3837	584	0.3548	17.5328
0.3404	2.6816	657	0.3520	17.7240
0.2577	2.9796	730	0.3515	17.6167

greenw0lf/whisper-ecapa-caps-10h

whisper-ecapa-caps-10h