mms-1b-all-bemgen-male-52

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

• Loss: 0.1838
• Wer: 0.3999

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: 8
• eval_batch_size: 4
• seed: 52
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
• lr_scheduler_warmup_steps: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.0656	0.5236	100	0.5987	0.6894
0.6923	1.0471	200	0.2096	0.4492
0.5391	1.5707	300	0.1898	0.4076
0.5182	2.0942	400	0.1838	0.4002
0.4984	2.6178	500	0.1802	0.4009
0.4791	3.1414	600	0.1793	0.3906
0.4669	3.6649	700	0.1784	0.4056
0.4646	4.1885	800	0.1833	0.4025
0.4526	4.7120	900	0.1755	0.3975
0.4582	5.2356	1000	0.1766	0.3976
0.4434	5.7592	1100	0.1773	0.3925
0.4292	6.2827	1200	0.1757	0.3883

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0