whisper-small-combined-lin-sna-lug-olusola
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: 0.5546
- Wer: 0.3443
- Cer: 0.1257
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: 0.0001
- train_batch_size: 16
- eval_batch_size: 64
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- 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: 500
- num_epochs: 30
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.2661 | 0.5020 | 500 | 0.6305 | 0.4220 | 0.1551 |
| 1.0202 | 1.0040 | 1000 | 0.5507 | 0.4619 | 0.1931 |
| 0.8280 | 1.5060 | 1500 | 0.5235 | 0.3643 | 0.1377 |
| 0.6772 | 2.0080 | 2000 | 0.5089 | 0.3957 | 0.1495 |
| 0.6185 | 2.5100 | 2500 | 0.5049 | 0.3441 | 0.1230 |
| 0.4801 | 3.0120 | 3000 | 0.5298 | 0.3308 | 0.1152 |
| 0.4279 | 3.5141 | 3500 | 0.5320 | 0.3468 | 0.1252 |
| 0.3562 | 4.0161 | 4000 | 0.5546 | 0.3443 | 0.1257 |
Framework versions
- Transformers 5.12.1
- Pytorch 2.11.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2