Whisper small Hi - CKS 2411 graminVoice
This model is a fine-tuned version of CKSINGH/whisper-small-hi-firefox on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.3893
- Wer: 16.1932
- Cer: 12.5576
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: 1.75e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 3000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.0115 | 17.24 | 1000 | 0.3435 | 16.9823 | 13.1975 |
| 0.0004 | 34.48 | 2000 | 0.3805 | 16.4186 | 12.5888 |
| 0.0001 | 51.72 | 3000 | 0.3893 | 16.1932 | 12.5576 |
Framework versions
- Transformers 4.36.0.dev0
- Pytorch 2.1.0+cu118
- Datasets 2.15.0
- Tokenizers 0.15.0