shaobai880824/breeze-asr-25-local-hokkien_v1

breeze-asr-25-local-hokkian_v1

This model is a fine-tuned version of MediaTek-Research/Breeze-ASR-25 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6065
• Semantic Similarity: 0.8492

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: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• 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: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Semantic Similarity
1.3345	0.1282	500	1.2993	0.7334
1.1385	0.2564	1000	1.1114	0.7649
1.024	0.3846	1500	1.0116	0.7790
0.922	0.5128	2000	0.9015	0.8011
0.8232	0.6410	2500	0.8240	0.8102
0.7668	0.7691	3000	0.7536	0.8223
0.691	0.8973	3500	0.6940	0.8340
0.5624	1.0254	4000	0.6480	0.8410
0.4132	1.1536	4500	0.6199	0.8470
0.4085	1.2818	5000	0.6065	0.8492

Framework versions

• Transformers 4.53.3
• Pytorch 2.7.1+cu126
• Datasets 4.0.0
• Tokenizers 0.21.2