learning_rate: 1e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Use 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
training_steps: 10000
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
2.1826	0.1123	500	1.2597	13.2054
1.1149	0.2247	1000	1.1731	12.1589
1.0703	0.3370	1500	1.1441	11.9967
1.0787	0.4493	2000	1.1242	11.7956
1.0588	0.5617	2500	1.1149	11.5296
1.0346	0.6740	3000	1.1049	11.4707
1.0206	0.7863	3500	1.0982	11.3632
1.024	0.8987	4000	1.0873	11.3176
1.0147	1.0110	4500	1.0833	11.1196
0.9595	1.1233	5000	1.0844	11.6402
0.9653	1.2357	5500	1.0825	11.2186
0.9606	1.3480	6000	1.0773	11.1749

SatwikDutta/2026-02-25_09-56-36

2026-02-25_09-56-36