Whisper Small N - Final Augmented
This model is a fine-tuned version of openai/whisper-small on the N Demo Final (Augmented) dataset. It achieves the following results on the evaluation set:
- Loss: 0.5052
- Wer: 48.7617
- Cer: 16.8914
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: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 16
- 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_steps: 0.1
- num_epochs: 20
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 2.3469 | 0.4986 | 88 | 0.9121 | 91.0842 | 36.4790 |
| 1.4826 | 0.9972 | 176 | 0.6352 | 79.1414 | 29.2724 |
| 1.2327 | 1.4929 | 264 | 0.5404 | 77.6555 | 29.0543 |
| 1.1068 | 1.9915 | 352 | 0.4834 | 67.3638 | 23.6420 |
| 0.9248 | 2.4873 | 440 | 0.4396 | 66.4832 | 24.7026 |
| 0.8094 | 2.9858 | 528 | 0.4209 | 60.2091 | 20.4203 |
| 0.6628 | 3.4816 | 616 | 0.4094 | 57.8976 | 19.9048 |
| 0.6277 | 3.9802 | 704 | 0.3889 | 55.6412 | 19.2803 |
| 0.4893 | 4.4759 | 792 | 0.3984 | 55.9163 | 18.8442 |
| 0.4862 | 4.9745 | 880 | 0.3801 | 53.4948 | 18.9036 |
| 0.3425 | 5.4703 | 968 | 0.4037 | 55.8063 | 19.2010 |
| 0.3797 | 5.9688 | 1056 | 0.3896 | 51.1282 | 18.9334 |
| 0.2309 | 6.4646 | 1144 | 0.4003 | 52.3390 | 18.4179 |
| 0.2416 | 6.9632 | 1232 | 0.4048 | 57.7876 | 20.8763 |
| 0.1828 | 7.4589 | 1320 | 0.4177 | 51.4034 | 18.1701 |
| 0.1982 | 7.9575 | 1408 | 0.4211 | 51.2383 | 17.9718 |
| 0.1416 | 8.4533 | 1496 | 0.4379 | 51.4034 | 17.8628 |
| 0.1390 | 8.9518 | 1584 | 0.4391 | 50.4128 | 17.4861 |
| 0.1169 | 9.4476 | 1672 | 0.4437 | 49.7523 | 17.4762 |
| 0.1106 | 9.9462 | 1760 | 0.4507 | 50.7980 | 17.6745 |
| 0.0935 | 10.4419 | 1848 | 0.4551 | 49.9725 | 17.8628 |
| 0.0882 | 10.9405 | 1936 | 0.4590 | 49.3671 | 17.6348 |
| 0.0660 | 11.4363 | 2024 | 0.4662 | 49.7523 | 17.3473 |
| 0.0803 | 11.9348 | 2112 | 0.4685 | 49.7523 | 17.5952 |
| 0.0719 | 12.4306 | 2200 | 0.4869 | 50.0275 | 17.4960 |
| 0.0612 | 12.9292 | 2288 | 0.4838 | 48.7067 | 17.1491 |
| 0.0456 | 13.4249 | 2376 | 0.4830 | 48.2113 | 17.1392 |
| 0.0407 | 13.9235 | 2464 | 0.4867 | 49.1469 | 17.1491 |
| 0.0399 | 14.4193 | 2552 | 0.4914 | 49.9174 | 17.7042 |
| 0.0439 | 14.9178 | 2640 | 0.4974 | 48.9268 | 16.9112 |
| 0.0432 | 15.4136 | 2728 | 0.4989 | 49.4772 | 17.3969 |
| 0.0292 | 15.9122 | 2816 | 0.5004 | 51.1282 | 18.3089 |
| 0.0316 | 16.4079 | 2904 | 0.5029 | 48.9268 | 17.1590 |
| 0.0332 | 16.9065 | 2992 | 0.5020 | 48.6516 | 16.8715 |
| 0.0203 | 17.4023 | 3080 | 0.5011 | 48.5416 | 17.0500 |
| 0.0355 | 17.9008 | 3168 | 0.5060 | 48.6516 | 17.0599 |
| 0.0289 | 18.3966 | 3256 | 0.5048 | 48.5966 | 17.0599 |
| 0.0313 | 18.8952 | 3344 | 0.5051 | 48.5416 | 16.8616 |
| 0.0319 | 19.3909 | 3432 | 0.5052 | 48.7617 | 16.8914 |
Framework versions
- Transformers 5.1.0
- Pytorch 2.9.0+cu126
- Datasets 4.5.0
- Tokenizers 0.22.2