whisper-small-lingala-qlora
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: 4.7410
- Wer: 0.9973
- Cer: 0.9793
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: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 4
- optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT 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: 50
- training_steps: 500
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 22.1056 | 0.4306 | 100 | 5.4382 | 1.0155 | 0.9853 |
| 19.9251 | 0.8611 | 200 | 5.0677 | 0.9898 | 0.9502 |
| 19.3998 | 1.2885 | 300 | 4.8921 | 0.9763 | 0.9285 |
| 18.6250 | 1.7191 | 400 | 4.7937 | 0.9905 | 0.9595 |
| 18.0820 | 2.1464 | 500 | 4.7410 | 0.9973 | 0.9793 |
Framework versions
- PEFT 0.19.1
- Transformers 5.13.0
- Pytorch 2.11.0+cu128
- Datasets 5.0.0
- Tokenizers 0.22.2