Whisper Base N - Fine-Tuned
This model is a fine-tuned version of openai/whisper-base on the N Demo dataset. It achieves the following results on the evaluation set:
- Loss: 0.8412
- Wer: 60.6884
- Cer: 23.9254
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: 1e-05
- train_batch_size: 16
- eval_batch_size: 8
- 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: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 2.2957 | 1.0 | 27 | 1.8628 | 214.4928 | 171.5234 |
| 1.5812 | 2.0 | 54 | 1.2720 | 129.3478 | 84.3552 |
| 1.1056 | 3.0 | 81 | 0.9757 | 92.9348 | 40.4551 |
| 0.8301 | 4.0 | 108 | 0.8126 | 82.0652 | 34.3236 |
| 0.647 | 5.0 | 135 | 0.7076 | 80.6159 | 32.7434 |
| 0.5085 | 6.0 | 162 | 0.6479 | 75.5435 | 29.9621 |
| 0.3937 | 7.0 | 189 | 0.6231 | 74.0942 | 29.8673 |
| 0.3025 | 8.0 | 216 | 0.5965 | 72.2826 | 27.5601 |
| 0.2285 | 9.0 | 243 | 0.5880 | 69.2029 | 26.4223 |
| 0.1701 | 10.0 | 270 | 0.6018 | 68.8406 | 26.5171 |
| 0.1261 | 11.0 | 297 | 0.6133 | 66.6667 | 26.0746 |
| 0.0921 | 12.0 | 324 | 0.6170 | 62.8623 | 25.3793 |
| 0.0649 | 13.0 | 351 | 0.6466 | 62.1377 | 25.3477 |
| 0.0452 | 14.0 | 378 | 0.6619 | 63.5870 | 25.6321 |
| 0.0332 | 15.0 | 405 | 0.6742 | 62.1377 | 24.2099 |
| 0.0228 | 16.0 | 432 | 0.6976 | 62.3188 | 24.4311 |
| 0.0184 | 17.0 | 459 | 0.7067 | 63.2246 | 24.1466 |
| 0.014 | 18.0 | 486 | 0.7276 | 59.9638 | 24.9684 |
| 0.0107 | 19.0 | 513 | 0.7442 | 61.5942 | 23.6726 |
| 0.0088 | 20.0 | 540 | 0.7347 | 61.9565 | 23.9886 |
| 0.0072 | 21.0 | 567 | 0.7440 | 60.8696 | 24.5575 |
| 0.0052 | 22.0 | 594 | 0.7598 | 59.7826 | 23.4513 |
| 0.0036 | 23.0 | 621 | 0.7754 | 60.3261 | 24.0834 |
| 0.0034 | 24.0 | 648 | 0.7911 | 59.7826 | 24.2731 |
| 0.0026 | 25.0 | 675 | 0.7969 | 60.1449 | 23.7358 |
| 0.0019 | 26.0 | 702 | 0.8012 | 59.4203 | 23.7042 |
| 0.0016 | 27.0 | 729 | 0.8079 | 58.8768 | 23.7674 |
| 0.0014 | 28.0 | 756 | 0.8144 | 59.9638 | 23.9254 |
| 0.0013 | 29.0 | 783 | 0.8201 | 59.9638 | 23.7674 |
| 0.0012 | 30.0 | 810 | 0.8249 | 59.0580 | 23.6726 |
| 0.0011 | 31.0 | 837 | 0.8293 | 60.5072 | 24.1783 |
| 0.0011 | 32.0 | 864 | 0.8325 | 60.5072 | 24.1150 |
| 0.001 | 33.0 | 891 | 0.8358 | 59.7826 | 23.9254 |
| 0.001 | 34.0 | 918 | 0.8388 | 60.1449 | 24.0834 |
| 0.0009 | 35.0 | 945 | 0.8412 | 60.6884 | 23.9254 |
Framework versions
- Transformers 4.57.6
- Pytorch 2.9.0+cu126
- Datasets 4.5.0
- Tokenizers 0.22.2