bezaisingh/whisper-small-bn-cv17

Whisper Small Bengali CV17

This model is a fine-tuned version of openai/whisper-small on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1083
• Wer: 31.8770

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 6000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2261	0.2618	500	0.2448	64.0053
0.1627	0.5236	1000	0.1681	48.7944
0.135	0.7853	1500	0.1401	43.1451
0.0831	1.0471	2000	0.1229	38.6268
0.0828	1.3089	2500	0.1179	37.0647
0.0778	1.5707	3000	0.1114	35.6588
0.0786	1.8325	3500	0.1043	34.2412
0.0485	2.0942	4000	0.1047	32.9504
0.047	2.3560	4500	0.1056	32.4125
0.0518	2.6178	5000	0.1021	32.5969
0.051	2.8796	5500	0.1005	32.0766
0.0267	3.1414	6000	0.1083	31.8770

Framework versions

• Transformers 4.43.3
• Pytorch 2.4.0
• Datasets 2.20.0
• Tokenizers 0.19.1