whisper-tiny-bn-custom-spm

This model is a fine-tuned version of Noobbbbb/whisper-tiny-bn-custom-spm on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4220
• Wer: 72.73
• Cer: 46.62

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: 3e-06
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• 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: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 13
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.2248	1.4535	1000	1.6454	76.44	49.73
2.1715	2.9070	2000	1.5724	75.21	49.2
2.0676	4.3605	3000	1.5242	74.82	48.7
2.0005	5.8140	4000	1.4818	73.67	47.12
1.9536	7.2674	5000	1.4661	73.07	46.64
1.8984	8.7209	6000	1.4387	72.99	46.82
1.9172	10.1744	7000	1.4269	72.79	46.59
1.8687	11.6279	8000	1.4229	72.67	46.51
1.9018	13.0	8944	1.4220	72.73	46.62

Framework versions

• Transformers 5.12.0
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2