jssaluja/fb-mms-1b-cer-jssaluja_rajinder_singh-epochs-10-test-datasets-10-20251026_094103

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: 0.4526
• Wer: 0.4014
• Wil: 0.6121
• Mer: 0.3942
• Cer: 0.1163
• Ncer: 0.0694
• Ccer: 0.1248

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: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Ccer	Cer	Validation Loss	Mer	Ncer	Wer	Wil
0.7855	1.0	7548	0.1771	0.1714	0.7379	0.5481	0.1145	0.5560	0.7709
0.3187	2.0	15096	0.1575	0.1473	0.6830	0.4840	0.0947	0.4984	0.7138
0.2592	3.0	22644	0.1579	0.1453	0.6231	0.4508	0.0998	0.4656	0.6786
0.2319	4.0	30192	0.1752	0.1618	0.7575	0.4672	0.1157	0.4766	0.6934
0.2093	5.0	37740	0.1500	0.1388	0.6409	0.4389	0.0912	0.4468	0.6647
0.1912	6.0	45288	0.1329	0.1228	0.5108	0.4153	0.0754	0.4234	0.6380
0.1777	7.0	52836	0.1378	0.1272	0.5269	0.4168	0.0817	0.4234	0.6390
0.1614	8.0	60384	0.1321	0.1230	0.5047	0.4092	0.0762	0.4163	0.6301
0.1463	9.0	67932	0.1267	0.1178	0.4550	0.4033	0.0696	0.4094	0.6235
0.1347	10.0	75480	0.4526	0.4014	0.6121	0.3942	0.1163	0.0694	0.1248

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.2