learning_rate: 5e-05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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: 500
num_epochs: 1
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Cer	Wer
0.5975	0.0548	1000	0.3457	0.9317	0.3544
0.5729	0.1096	2000	0.3527	0.9454	0.3633
0.519	0.1644	3000	0.3491	0.9362	0.3597
0.5355	0.2192	4000	0.3435	0.9349	0.3451
0.5285	0.2740	5000	0.3419	0.9564	0.3680
0.5043	0.3288	6000	0.3389	0.9385	0.3531
0.4749	0.3837	7000	0.3353	0.9450	0.3491
0.4515	0.4385	8000	0.3315	0.9321	0.3510
0.4544	0.4933	9000	0.3316	0.9275	0.3533
0.4697	0.5481	10000	0.3257	0.9417	0.3418
0.4492	0.6029	11000	0.3182	0.9280	0.3422
0.4188	0.6577	12000	0.3208	0.9289	0.3422
0.4187	0.7125	13000	0.3199	0.9289	0.3455
0.4034	0.7673	14000	0.3193	0.9193	0.3432
0.3974	0.8221	15000	0.3125	0.9225	0.3443
0.4285	0.8769	16000	0.3088	0.9225	0.3423
0.3937	0.9317	17000	0.3079	0.9206	0.3469
0.3852	0.9865	18000	0.3065	0.9216	0.3467

ganga4364/whisper-small-v5

whisper-small-v5