|
|
--- |
|
|
pipeline_tag: audio-classification |
|
|
datasets: |
|
|
- common_language |
|
|
--- |
|
|
# Language Classification |
|
|
|
|
|
A model trained for language classification. Thanks to @sanchit-gandhi for [this code](https://huggingface.co/sanchit-gandhi/whisper-base-ft-common-language-id) which was used to train the model. |
|
|
|
|
|
This model was trained for 15 epochs. |
|
|
|
|
|
## Evaluation |
|
|
|
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 1.1229 |
|
|
- Accuracy: 0.7401 |
|
|
|
|
|
## Hyperparameters |
|
|
|
|
|
The following hyperparameters were used during training: |
|
|
- learning_rate: 1e-05 |
|
|
- train_batch_size: 64 |
|
|
- eval_batch_size: 16 |
|
|
- seed: 0 |
|
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
|
- lr_scheduler_type: linear |
|
|
- lr_scheduler_warmup_ratio: 0.1 |
|
|
- num_epochs: 15.0 |
|
|
- mixed_precision_training: Native AMP |
|
|
|
|
|
## Training Results |
|
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Accuracy | |
|
|
|:-------------:|:-----:|:----:|:---------------:|:--------:| |
|
|
| 3.1638 | 1.0 | 347 | 3.0152 | 0.4200 | |
|
|
| 2.0788 | 2.0 | 694 | 1.9700 | 0.5504 | |
|
|
| 1.4236 | 3.0 | 1041 | 1.5048 | 0.6374 | |
|
|
| 1.0305 | 4.0 | 1388 | 1.2979 | 0.6685 | |
|
|
| 0.7651 | 5.0 | 1735 | 1.1692 | 0.7023 | |
|
|
| 0.5782 | 6.0 | 2082 | 1.0896 | 0.7227 | |
|
|
| 0.4483 | 7.0 | 2429 | 1.0605 | 0.7198 | |
|
|
| 0.3253 | 8.0 | 2776 | 1.0255 | 0.7376 | |
|
|
| 0.2589 | 9.0 | 3123 | 1.0478 | 0.7354 | |
|
|
| 0.1825 | 10.0 | 3470 | 1.0677 | 0.7318 | |
|
|
| 0.1489 | 11.0 | 3817 | 1.0946 | 0.7373 | |
|
|
| 0.1274 | 12.0 | 4164 | 1.1180 | 0.7376 | |
|
|
| 0.1074 | 13.0 | 4511 | 1.1229 | 0.7401 | |
|
|
| 0.0979 | 14.0 | 4858 | 1.1523 | 0.7383 | |
|
|
| 0.0914 | 15.0 | 5205 | 1.1498 | 0.7401 | |
