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: 95
num_epochs: 3.0
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
0.4803	0.2969	95	0.4288	21.3675
0.3662	0.5938	190	0.3694	17.5160
0.3262	0.8906	285	0.3527	16.2378
0.3975	1.1875	380	0.3456	15.8721
0.3515	1.4844	475	0.3379	16.3586
0.2394	1.7812	570	0.3321	16.5163
0.3971	2.0781	665	0.3288	15.9761
0.3413	2.375	760	0.3275	14.9730
0.3109	2.6719	855	0.3258	15.2984
0.2904	2.9688	950	0.3245	14.9529

greenw0lf/whisper-ecapa-mmr-20h

whisper-ecapa-mmr-20h