wav2vec-best-CREMA-sentiment-analysis-best3
This model is a fine-tuned version of facebook/wav2vec2-base on Supreeta03/CREMA-audioData. It achieves the following results on the evaluation set:
- top2 Accuracy: 0.7824
- Loss: 1.1563
- Accuracy: 0.5601
Model description
Fine tuned from [facebook/wav2vec2-base] for performing sentiment analysis on audio data.
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 20
Training results
| Training Loss | Epoch | Step | Top2 Accuracy | Validation Loss | Accuracy |
|---|---|---|---|---|---|
| 1.7555 | 0.99 | 37 | 0.5281 | 1.7048 | 0.2905 |
| 1.5612 | 1.99 | 74 | 0.5819 | 1.5406 | 0.3493 |
| 1.4333 | 2.98 | 111 | 0.6373 | 1.4668 | 0.3778 |
| 1.3933 | 4.0 | 149 | 0.6809 | 1.3798 | 0.4450 |
| 1.3418 | 4.99 | 186 | 0.7045 | 1.3120 | 0.4719 |
| 1.2238 | 5.99 | 223 | 0.7263 | 1.2718 | 0.4979 |
| 1.1896 | 6.98 | 260 | 0.7313 | 1.2430 | 0.5113 |
| 1.1501 | 8.0 | 298 | 0.7296 | 1.2631 | 0.5088 |
| 1.1052 | 8.99 | 335 | 0.7506 | 1.2462 | 0.5097 |
| 1.068 | 9.99 | 372 | 0.7641 | 1.1822 | 0.5399 |
| 1.0594 | 10.98 | 409 | 0.7590 | 1.1700 | 0.5575 |
| 0.9519 | 12.0 | 447 | 0.7733 | 1.1465 | 0.5516 |
| 0.9513 | 12.99 | 484 | 0.7918 | 1.1428 | 0.5676 |
| 0.9324 | 13.99 | 521 | 0.7666 | 1.1721 | 0.5634 |
| 0.9173 | 14.98 | 558 | 0.7825 | 1.1494 | 0.5584 |
| 0.8781 | 16.0 | 596 | 0.7918 | 1.1468 | 0.5718 |
| 0.8627 | 16.99 | 633 | 0.7775 | 1.1554 | 0.5575 |
| 0.83 | 17.99 | 670 | 0.7817 | 1.1438 | 0.5718 |
| 0.8305 | 18.98 | 707 | 0.7935 | 1.1323 | 0.5760 |
| 0.8314 | 19.87 | 740 | 0.7851 | 1.1341 | 0.5726 |
Framework versions
- Transformers 4.38.2
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.15.2
- Downloads last month
- 42