Whisper Small Hi - Vyapar V2
This model is a fine-tuned version of openai/whisper-small on the Vyapar Calling Data 1200 hours dataset. It achieves the following results on the evaluation set:
- Loss: 0.9627
- Wer: 48.4083
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use 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: 4000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 1.8904 | 0.9251 | 1000 | 1.0187 | 56.1507 |
| 1.7278 | 1.8501 | 2000 | 0.9433 | 53.3196 |
| 1.1629 | 2.7752 | 3000 | 0.9364 | 48.6768 |
| 0.8499 | 3.7003 | 4000 | 0.9627 | 48.4083 |
Framework versions
- Transformers 4.47.1
- Pytorch 2.5.1+cu118
- Datasets 3.2.0
- Tokenizers 0.21.0