exp_008b_base_cv24_vanilla_no_aug
This model is a fine-tuned version of openai/whisper-base on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.3538
- Wer: 28.6338
- Wer Ortho: 31.5190
- Cer: 9.4386
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: 64
- eval_batch_size: 32
- seed: 42
- 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: 2176
- num_epochs: 15
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Wer Ortho | Cer |
|---|---|---|---|---|---|---|
| 0.5162 | 0.3444 | 500 | 0.5790 | 51.4221 | 54.3859 | 16.3021 |
| 0.3936 | 0.6887 | 1000 | 0.4636 | 40.6877 | 43.8625 | 13.3724 |
| 0.3192 | 1.0331 | 1500 | 0.4237 | 37.1166 | 40.2916 | 12.2298 |
| 0.3004 | 1.3774 | 2000 | 0.4044 | 35.4258 | 38.4338 | 11.7855 |
| 0.2904 | 1.7218 | 2500 | 0.3844 | 34.7824 | 37.6633 | 12.3830 |
| 0.2073 | 2.0661 | 3000 | 0.3724 | 32.2113 | 35.0304 | 10.6682 |
| 0.2124 | 2.4105 | 3500 | 0.3539 | 30.7883 | 33.7801 | 10.2432 |
| 0.2079 | 2.7548 | 4000 | 0.3436 | 30.6876 | 33.5884 | 10.0643 |
| 0.1405 | 3.0992 | 4500 | 0.3538 | 29.7492 | 32.7413 | 9.9562 |
| 0.1498 | 3.4435 | 5000 | 0.3421 | 29.6484 | 32.5631 | 9.7255 |
| 0.1566 | 3.7879 | 5500 | 0.3374 | 29.0830 | 32.0248 | 9.5126 |
| 0.0981 | 4.1322 | 6000 | 0.3621 | 29.0041 | 32.0248 | 9.7219 |
| 0.1059 | 4.4766 | 6500 | 0.3602 | 28.9169 | 31.8070 | 9.4974 |
| 0.1109 | 4.8209 | 7000 | 0.3538 | 28.6338 | 31.5190 | 9.4386 |
| 0.0685 | 5.1653 | 7500 | 0.3843 | 28.9487 | 31.9060 | 9.4235 |
| 0.0752 | 5.5096 | 8000 | 0.3902 | 28.7436 | 31.7404 | 9.4488 |
| 0.0777 | 5.8540 | 8500 | 0.3862 | 28.7363 | 31.6513 | 9.5293 |
| 0.0435 | 6.1983 | 9000 | 0.4288 | 29.4597 | 32.4632 | 9.8163 |
| 0.0473 | 6.5427 | 9500 | 0.4399 | 29.0994 | 32.0879 | 9.5760 |
Framework versions
- Transformers 5.2.0
- Pytorch 2.10.0+cu128
- Datasets 3.6.0
- Tokenizers 0.22.2