learning_rate: 1e-05
train_batch_size: 32
eval_batch_size: 16
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: 100
num_epochs: 15
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
3.5743	1.0	18	3.0139	187.2881
2.6299	2.0	36	2.1201	99.0315
1.7123	3.0	54	1.6249	116.5860
1.1523	4.0	72	1.3463	97.5787
0.7429	5.0	90	1.1763	72.2760
0.4828	6.0	108	1.1024	68.4019
0.2375	7.0	126	1.0793	64.7700
0.1053	8.0	144	1.0770	63.5593
0.0498	9.0	162	1.0751	102.4213
0.0264	10.0	180	1.0856	74.9395
0.0152	11.0	198	1.0842	84.0194
0.0101	12.0	216	1.0852	83.4140
0.0076	13.0	234	1.0987	61.8644
0.0068	14.0	252	1.1034	62.1065
0.0064	15.0	270	1.1039	84.8668

zionia/whisper-small-isizulu

Whisper Small IsiZulu