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: 0.9411
- Wer: 24.8801
- Cer: 21.1598
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 |
---|---|---|---|---|---|
0.5524 | 1.0 | 569 | 0.5578 | 28.6033 | 24.0945 |
0.3686 | 2.0 | 1138 | 0.5504 | 30.1336 | 26.4337 |
0.2033 | 3.0 | 1707 | 0.6189 | 28.4239 | 24.3193 |
0.0994 | 4.0 | 2276 | 0.7076 | 26.9339 | 22.6996 |
0.0404 | 5.0 | 2845 | 0.7609 | 26.8827 | 22.0729 |
0.0317 | 6.0 | 3414 | 0.7954 | 27.0108 | 23.2898 |
0.0254 | 7.0 | 3983 | 0.8337 | 26.8534 | 22.7960 |
0.0118 | 8.0 | 4552 | 0.8676 | 28.1091 | 23.5103 |
0.016 | 9.0 | 5121 | 0.8757 | 25.6928 | 21.6242 |
0.0072 | 10.0 | 5690 | 0.8893 | 25.2169 | 21.3377 |
0.0032 | 11.0 | 6259 | 0.9026 | 24.7593 | 20.8569 |
0.0047 | 12.0 | 6828 | 0.9188 | 26.1798 | 22.3055 |
0.002 | 13.0 | 7397 | 0.9011 | 25.7844 | 22.0373 |
0.0005 | 14.0 | 7966 | 0.9134 | 25.6745 | 21.8073 |
0.0004 | 15.0 | 8535 | 0.9110 | 25.4000 | 21.5877 |
0.0001 | 16.0 | 9104 | 0.9231 | 24.6971 | 20.8421 |
0.0001 | 17.0 | 9673 | 0.9301 | 25.1766 | 21.4002 |
0.0001 | 18.0 | 10242 | 0.9363 | 24.8984 | 21.1641 |
0.0001 | 19.0 | 10811 | 0.9399 | 24.8398 | 21.1155 |
0.0 | 20.0 | 11380 | 0.9411 | 24.8801 | 21.1598 |
Framework versions
- Transformers 4.41.1
- Pytorch 2.3.0
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 20
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.