openai/whisper-small
This model is a fine-tuned version of openai/whisper-small on the pphuc25/VietMed-split-8-2 dataset. It achieves the following results on the evaluation set:
- Loss: 0.9214
- Wer: 21.2777
- Cer: 17.6019
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: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 20
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
0.624 | 1.0 | 569 | 0.6172 | 28.3507 | 24.2846 |
0.4335 | 2.0 | 1138 | 0.5851 | 24.6604 | 20.0193 |
0.2821 | 3.0 | 1707 | 0.6146 | 24.1333 | 19.6217 |
0.1837 | 4.0 | 2276 | 0.6578 | 24.6458 | 20.3639 |
0.0938 | 5.0 | 2845 | 0.7273 | 24.2724 | 19.7762 |
0.0787 | 6.0 | 3414 | 0.7621 | 23.7232 | 19.1921 |
0.0536 | 7.0 | 3983 | 0.8131 | 23.9649 | 19.5506 |
0.0424 | 8.0 | 4552 | 0.8392 | 24.1552 | 19.7242 |
0.0205 | 9.0 | 5121 | 0.8504 | 22.8409 | 18.6722 |
0.018 | 10.0 | 5690 | 0.8700 | 23.0240 | 18.7208 |
0.016 | 11.0 | 6259 | 0.8755 | 22.8666 | 18.6357 |
0.0152 | 12.0 | 6828 | 0.8713 | 22.5773 | 18.3892 |
0.0062 | 13.0 | 7397 | 0.8858 | 21.9623 | 18.0203 |
0.0051 | 14.0 | 7966 | 0.9012 | 21.9001 | 18.0021 |
0.0024 | 15.0 | 8535 | 0.8900 | 21.7133 | 17.8484 |
0.001 | 16.0 | 9104 | 0.9155 | 21.9586 | 18.0594 |
0.0006 | 17.0 | 9673 | 0.9122 | 21.5047 | 17.7451 |
0.0005 | 18.0 | 10242 | 0.9147 | 21.2997 | 17.5880 |
0.0011 | 19.0 | 10811 | 0.9194 | 21.3143 | 17.6132 |
0.0001 | 20.0 | 11380 | 0.9214 | 21.2777 | 17.6019 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 7
This model does not have enough activity to be deployed to Inference API (serverless) yet.
Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.