|
--- |
|
library_name: transformers |
|
license: apache-2.0 |
|
base_model: openai/whisper-small |
|
tags: |
|
- generated_from_trainer |
|
metrics: |
|
- wer |
|
model-index: |
|
- name: whisper-small-CV-Fleurs-lg-313hrs-v1 |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
# whisper-small-CV-Fleurs-lg-313hrs-v1 |
|
|
|
This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Cer: 0.0803 |
|
- Loss: 0.7644 |
|
- Wer: 0.2848 |
|
|
|
## 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: 4 |
|
- eval_batch_size: 2 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 8 |
|
- optimizer: Use adamw_hf with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_ratio: 0.1 |
|
- num_epochs: 30 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Cer | Validation Loss | Wer | |
|
|:-------------:|:-----:|:-------:|:------:|:---------------:|:------:| |
|
| 0.9959 | 1.0 | 30487 | 0.4780 | 0.5805 | 1.0507 | |
|
| 0.2917 | 2.0 | 60974 | 0.8109 | 0.4413 | 1.4684 | |
|
| 0.1945 | 3.0 | 91461 | 0.1449 | 0.4016 | 0.4604 | |
|
| 0.139 | 4.0 | 121948 | 0.0933 | 0.3980 | 0.3740 | |
|
| 0.0986 | 5.0 | 152435 | 0.0910 | 0.4101 | 0.3542 | |
|
| 0.0698 | 6.0 | 182922 | 0.0935 | 0.4245 | 0.3447 | |
|
| 0.0515 | 7.0 | 213409 | 0.0824 | 0.4571 | 0.3246 | |
|
| 0.0412 | 8.0 | 243896 | 0.0843 | 0.4769 | 0.3185 | |
|
| 0.0362 | 9.0 | 274383 | 0.0812 | 0.4909 | 0.3106 | |
|
| 0.0339 | 10.0 | 304870 | 0.0819 | 0.5001 | 0.3119 | |
|
| 0.0301 | 11.0 | 335357 | 0.0848 | 0.5248 | 0.3142 | |
|
| 0.0243 | 12.0 | 365844 | 0.0843 | 0.5246 | 0.3045 | |
|
| 0.0199 | 13.0 | 396331 | 0.0801 | 0.5518 | 0.3004 | |
|
| 0.0167 | 14.0 | 426818 | 0.0857 | 0.5877 | 0.3085 | |
|
| 0.0143 | 15.0 | 457305 | 0.0806 | 0.5835 | 0.3024 | |
|
| 0.0124 | 16.0 | 487792 | 0.0819 | 0.5982 | 0.2995 | |
|
| 0.011 | 17.0 | 518279 | 0.0845 | 0.5933 | 0.3022 | |
|
| 0.0099 | 18.0 | 548766 | 0.0866 | 0.6195 | 0.2996 | |
|
| 0.0088 | 19.0 | 579253 | 0.0825 | 0.6577 | 0.2966 | |
|
| 0.0079 | 20.0 | 609740 | 0.0843 | 0.6416 | 0.2991 | |
|
| 0.0073 | 21.0 | 640227 | 0.0810 | 0.6536 | 0.2938 | |
|
| 0.0065 | 22.0 | 670714 | 0.0829 | 0.6708 | 0.2990 | |
|
| 0.006 | 23.0 | 701201 | 0.0867 | 0.6726 | 0.2978 | |
|
| 0.0056 | 24.0 | 731688 | 0.0819 | 0.6944 | 0.2921 | |
|
| 0.0053 | 25.0 | 762175 | 0.0824 | 0.6845 | 0.2942 | |
|
| 0.0049 | 26.0 | 792662 | 0.0840 | 0.6856 | 0.2926 | |
|
| 0.0046 | 27.0 | 823149 | 0.0829 | 0.6926 | 0.2914 | |
|
| 0.0042 | 28.0 | 853636 | 0.0832 | 0.7022 | 0.2866 | |
|
| 0.004 | 29.0 | 884123 | 0.0800 | 0.7230 | 0.2898 | |
|
| 0.0037 | 30.0 | 914610 | 0.0824 | 0.7287 | 0.2925 | |
|
| 0.0034 | 31.0 | 945097 | 0.0801 | 0.7363 | 0.2860 | |
|
| 0.0033 | 32.0 | 975584 | 0.0803 | 0.7497 | 0.2866 | |
|
| 0.0032 | 33.0 | 1006071 | 0.0814 | 0.7478 | 0.2827 | |
|
| 0.0029 | 34.0 | 1036558 | 0.0791 | 0.7292 | 0.2845 | |
|
| 0.0028 | 35.0 | 1067045 | 0.0829 | 0.7657 | 0.2891 | |
|
| 0.0026 | 36.0 | 1097532 | 0.0803 | 0.7644 | 0.2848 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.47.0 |
|
- Pytorch 2.1.0+cu118 |
|
- Datasets 3.2.0 |
|
- Tokenizers 0.21.0 |
|
|
