mms-1b-nyagen-balanced-model

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

• Loss: 0.1803
• Wer: 0.2549

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• 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: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.181	0.2762	100	0.6055	0.5247
0.5071	0.5525	200	0.2452	0.3594
0.3791	0.8287	300	0.2159	0.3232
0.3464	1.1050	400	0.2059	0.3047
0.3326	1.3812	500	0.1919	0.2943
0.322	1.6575	600	0.1868	0.2861
0.3025	1.9337	700	0.1850	0.2902
0.2939	2.2099	800	0.1777	0.2698
0.2971	2.4862	900	0.1716	0.2675
0.2787	2.7624	1000	0.1750	0.2716
0.32	3.0387	1100	0.1725	0.2734
0.2738	3.3149	1200	0.1803	0.2544

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0