binbin123/whisper-small-ug

Whisper Small ug - Sanchit Gandhi

This model is a fine-tuned version of openai/whisper-small on the Common Voice 11 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3966
• Wer Ortho: 39.5518
• Wer: 34.2595

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: 2e-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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0842	3.9216	1000	0.2856	44.0534	37.9371
0.0173	7.8431	2000	0.3364	40.8646	34.9089
0.0087	11.7647	3000	0.3656	39.9169	34.3824
0.0065	15.6863	4000	0.3966	39.5518	34.2595
0.0068	19.6078	5000	0.3997	39.5790	34.3210

Framework versions

• Transformers 4.45.2
• Pytorch 1.12.0+cu113
• Datasets 3.0.2
• Tokenizers 0.20.1