metadata
license: apache-2.0
tags:
- generated_from_trainer
base_model: openai/whisper-large-v2
metrics:
- wer
model-index:
- name: BA_Model_V3
results: []
BA_Model_V3
This model is a fine-tuned version of openai/whisper-large-v2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.5502
- Wer: 24.6079
- Cer: 13.4358
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: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 15
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 0.3329 | 1.0 | 278 | 0.3232 | 21.6707 | 12.0794 |
| 0.2364 | 2.0 | 556 | 0.3756 | 23.9367 | 13.5422 |
| 0.1902 | 3.0 | 834 | 0.4029 | 38.9055 | 23.2284 |
| 0.1324 | 4.0 | 1112 | 0.4090 | 24.9568 | 13.9016 |
| 0.0797 | 5.0 | 1390 | 0.4468 | 28.4789 | 15.9201 |
| 0.0505 | 6.0 | 1668 | 0.4519 | 26.8873 | 14.9458 |
| 0.0317 | 7.0 | 1946 | 0.4599 | 32.2435 | 18.9045 |
| 0.0198 | 8.0 | 2224 | 0.4814 | 25.7675 | 14.1053 |
| 0.0124 | 9.0 | 2502 | 0.4716 | 26.4620 | 15.1289 |
| 0.007 | 10.0 | 2780 | 0.5003 | 24.9236 | 13.6666 |
| 0.0024 | 11.0 | 3058 | 0.5149 | 25.2459 | 14.1990 |
| 0.0018 | 12.0 | 3336 | 0.5262 | 24.2856 | 13.5232 |
| 0.0006 | 13.0 | 3614 | 0.5360 | 24.0896 | 13.4676 |
| 0.0006 | 14.0 | 3892 | 0.5473 | 24.2059 | 13.1781 |
| 0.0005 | 15.0 | 4170 | 0.5502 | 24.6079 | 13.4358 |
Framework versions
- Transformers 4.38.1
- Pytorch 2.2.1+cu121
- Datasets 2.17.2.dev0
- Tokenizers 0.15.2