jssaluja/fb-mms-1b-cleaned-jssaluja_rajinder_singh-epochs-12-test-datasets-10-20250812_232950

This model is a fine-tuned version of facebook/mms-1b-all on the jssaluja/rajinder_singh dataset. It achieves the following results on the evaluation set:

• Loss: 0.3189
• Wer: 0.3560
• Wil: 0.5533
• Mer: 0.3498
• Cer: 0.0917

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• 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
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Mer	Wer	Wil
0.387	1.0	7548	0.1258	0.4401	0.4452	0.4738	0.6682
0.2054	2.0	15096	0.1060	0.3444	0.3981	0.4064	0.6145
0.1811	3.0	22644	0.3718	0.4129	0.6234	0.4041	0.1102
0.1666	4.0	30192	0.3334	0.4028	0.6108	0.3993	0.1021
0.1548	5.0	37740	0.3263	0.3748	0.5769	0.3675	0.0968
0.1452	6.0	45288	0.3267	0.3879	0.5943	0.3837	0.0986
0.1382	7.0	52836	0.3242	0.3774	0.5809	0.3725	0.0960
0.1314	8.0	60384	0.3273	0.3661	0.5656	0.3580	0.0949
0.1257	9.0	67932	0.3203	0.3690	0.5698	0.3620	0.0945
0.1195	10.0	75480	0.3153	0.3584	0.5576	0.3528	0.0921
0.1141	11.0	83028	0.3187	0.3613	0.5596	0.3554	0.0920
0.1098	12.0	90576	0.3189	0.3560	0.5534	0.3498	0.0917

Framework versions

• Transformers 4.55.0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.4