bezaisingh/whisper-tiny-bn-cv17

Whisper Tiny Bengali

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

• Loss: 0.1903
• Wer: 57.0863

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
1.1049	0.2618	500	1.0603	235.8867
0.3583	0.5236	1000	0.3693	96.8899
0.2893	0.7853	1500	0.3039	84.1197
0.2304	1.0471	2000	0.2714	79.2737
0.2191	1.3089	2500	0.2520	83.5195
0.213	1.5707	3000	0.2347	71.8450
0.1972	1.8325	3500	0.2225	69.6217
0.1659	2.0942	4000	0.2145	61.4613
0.165	2.3560	4500	0.2059	61.1559
0.1584	2.6178	5000	0.1979	61.7949
0.1556	2.8796	5500	0.1923	57.9308
0.1263	3.1414	6000	0.1903	57.0863

Framework versions

• Transformers 4.45.2
• Pytorch 2.2.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1