mms-300m-ewe-innocent

This model is a fine-tuned version of facebook/mms-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4485
• Wer: 0.3152
• Cer: 0.0983

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• 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
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
21.0099	1.0617	500	2.4328	1.0	0.8118
3.7724	2.1233	1000	0.4691	0.3473	0.1071
3.2197	3.1850	1500	0.4333	0.3302	0.1019
2.8604	4.2467	2000	0.4280	0.3242	0.0996
2.6069	5.3083	2500	0.4244	0.3147	0.0980
3.9345	6.3700	3000	0.4149	0.3164	0.0986
2.4257	7.4317	3500	0.4234	0.3155	0.0982
2.3694	8.4934	4000	0.4515	0.3147	0.0987
2.0703	9.5550	4500	0.4485	0.3152	0.0983

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2