Whisper Base GA-EN Speech Translation
This model is a fine-tuned version of openai/whisper-base on an unknown dataset. The best model based on ChrF (this version) is at checkpoint 1000, epoch 3.72, and it achieves the following results on the evaluation set:
- Loss: 2.2482
- Bleu: 20.8
- Chrf: 35.56
- Wer: 84.0162
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: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 0.03
- training_steps: 1000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Bleu | Chrf | Wer |
|---|---|---|---|---|---|---|
| 1.5709 | 0.37 | 100 | 2.1099 | 5.49 | 22.56 | 144.5745 |
| 0.9426 | 0.74 | 200 | 2.0613 | 10.65 | 26.37 | 130.0315 |
| 0.3912 | 1.12 | 300 | 2.1207 | 13.43 | 29.77 | 103.9172 |
| 0.3943 | 1.49 | 400 | 2.1177 | 16.64 | 32.27 | 97.3435 |
| 0.3605 | 1.86 | 500 | 2.1689 | 18.41 | 32.69 | 87.1679 |
| 0.1164 | 2.23 | 600 | 2.1506 | 20.49 | 33.74 | 82.3953 |
| 0.1371 | 2.6 | 700 | 2.1397 | 19.86 | 34.97 | 84.9167 |
| 0.1263 | 2.97 | 800 | 2.1849 | 21.11 | 34.92 | 81.3147 |
| 0.049 | 3.35 | 900 | 2.2424 | 21.24 | 35.22 | 83.6110 |
| 0.0462 | 3.72 | 1000 | 2.2482 | 20.8 | 35.56 | 84.0162 |
Framework versions
- Transformers 4.39.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2