whisper-small-santali-sanlish-1934
This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6691
- Wer: 34.5088
- Cer: 7.6606
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 153
- num_epochs: 25
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 3.4164 | 1.0 | 103 | 0.6553 | 44.3325 | 10.3516 |
| 0.9734 | 2.0 | 206 | 0.5723 | 38.7909 | 8.9410 |
| 0.4124 | 3.0 | 309 | 0.5563 | 38.0353 | 8.3984 |
| 0.1842 | 4.0 | 412 | 0.5620 | 35.8942 | 7.8993 |
| 0.1116 | 5.0 | 515 | 0.5717 | 33.8791 | 7.8993 |
| 0.0794 | 6.0 | 618 | 0.5897 | 34.1310 | 7.8125 |
| 0.0470 | 7.0 | 721 | 0.6306 | 34.8866 | 7.8776 |
| 0.0363 | 8.0 | 824 | 0.6719 | 35.1385 | 7.8993 |
| 0.0266 | 9.0 | 927 | 0.6565 | 35.5164 | 7.9210 |
| 0.0132 | 10.0 | 1030 | 0.6691 | 34.5088 | 7.6606 |
Framework versions
- Transformers 5.15.0
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2