wav2vec2-xlsr-korean-speech-emotion-recognition3
This model is a fine-tuned version of jungjongho/wav2vec2-large-xlsr-korean-demo-colab_epoch15 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.0600
- Accuracy: 0.9876
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: 2e-05
- train_batch_size: 2
- eval_batch_size: 2
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 4
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 1
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.6472 | 0.08 | 1500 | 0.3769 | 0.8705 |
| 0.3873 | 0.15 | 3000 | 0.3814 | 0.9127 |
| 0.3002 | 0.23 | 4500 | 0.2617 | 0.9429 |
| 0.2399 | 0.3 | 6000 | 0.1336 | 0.9693 |
| 0.2181 | 0.38 | 7500 | 0.1360 | 0.9728 |
| 0.1992 | 0.46 | 9000 | 0.1239 | 0.9717 |
| 0.1556 | 0.53 | 10500 | 0.1053 | 0.9781 |
| 0.1412 | 0.61 | 12000 | 0.0915 | 0.9810 |
| 0.1396 | 0.69 | 13500 | 0.0777 | 0.9826 |
| 0.1159 | 0.76 | 15000 | 0.0801 | 0.9831 |
| 0.1156 | 0.84 | 16500 | 0.0667 | 0.9867 |
| 0.1149 | 0.91 | 18000 | 0.0670 | 0.9860 |
| 0.0929 | 0.99 | 19500 | 0.0600 | 0.9876 |
Framework versions
- Transformers 4.22.0.dev0
- Pytorch 1.12.1+cu113
- Datasets 2.4.1.dev0
- Tokenizers 0.12.1