mms-300m-ful-victor

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: 0.4118
• Wer: 0.4305
• Cer: 0.1115

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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
5.9195	0.8439	500	2.9152	1.0	1.0
1.7186	1.6869	1000	0.7183	0.6875	0.1960
1.0733	2.5300	1500	0.4555	0.4797	0.1263
0.8422	3.3730	2000	0.4021	0.4286	0.1121
0.7292	4.2160	2500	0.3830	0.4079	0.1051
0.6939	5.0591	3000	0.3765	0.4006	0.1030
0.6966	5.9030	3500	0.3658	0.4002	0.1026
0.6921	6.7460	4000	0.3789	0.4035	0.1032
0.7452	7.5890	4500	0.3977	0.4203	0.1085
0.7932	8.4321	5000	0.4118	0.4305	0.1115

Framework versions

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