whisper-small-lug-grain

This model is a fine-tuned version of sulaimank/whisper-small-lug-4src on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1864
• Wer: 0.0936
• Cer: 0.0387

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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: 500
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
0.3627	0.6549	1000	0.0612	0.1947	0.1571
0.1263	1.3098	2000	0.1511	0.1339	0.0630
0.1213	1.9646	3000	0.1367	0.1280	0.0557
0.0273	2.6195	4000	0.1531	0.1243	0.0545
0.0066	3.2744	5000	0.1693	0.1167	0.0563
0.0087	3.9293	6000	0.1531	0.1092	0.0516
0.0052	4.5842	7000	0.1610	0.1087	0.0499
0.0032	5.2390	8000	0.1614	0.1065	0.0491
0.0047	5.8939	9000	0.1674	0.1022	0.0426
0.0019	6.5488	10000	0.1729	0.1001	0.0428
0.0024	7.2037	11000	0.1773	0.1006	0.0426
0.0030	7.8585	12000	0.1916	0.1011	0.0417
0.0012	8.5134	13000	0.1759	0.0958	0.0416
0.0018	9.1683	14000	0.1797	0.0995	0.0438
0.0024	9.8232	15000	0.1803	0.0936	0.0400
0.0011	10.4781	16000	0.1869	0.1060	0.0477
0.0043	11.1329	17000	0.1871	0.1017	0.0427
0.0008	11.7878	18000	0.1820	0.0995	0.0435
0.0006	12.4427	19000	0.1864	0.0936	0.0387

Framework versions

• Transformers 5.3.0
• Pytorch 2.10.0+cu130
• Datasets 4.0.0
• Tokenizers 0.22.2