learning_rate: 1e-05
train_batch_size: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 4
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	Wer
0.4934	0.0933	1000	0.5257	40.4572
0.3632	0.1866	2000	0.4365	34.0404
0.3164	0.2799	3000	0.3980	30.8892
0.2853	0.3732	4000	0.3731	29.0710
0.2674	0.4664	5000	0.3553	27.9257
0.2557	0.5597	6000	0.3414	26.5430
0.2451	0.6530	7000	0.3340	25.9229
0.2375	0.7463	8000	0.3242	25.4436
0.2334	0.8396	9000	0.3204	25.0639
0.2268	0.9329	10000	0.3168	24.9003

mondhs/l3-whisper-small-l2c_v4

l3-whisper-small-l2c_v4