distiled_flip_model_emotion_alpha_0.8_epoch5_v1
This model is a fine-tuned version of ArafatBHossain/distill_bert_fine_tuned_emotion_dataset on the emotion dataset. It achieves the following results on the evaluation set:
- Loss: 0.1476
- Accuracy: 0.942
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: 5e-05
- 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
- num_epochs: 5
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.1966 | 1.0 | 2000 | 0.2675 | 0.9315 |
| 0.154 | 2.0 | 4000 | 0.2265 | 0.9355 |
| 0.1214 | 3.0 | 6000 | 0.1805 | 0.9375 |
| 0.078 | 4.0 | 8000 | 0.1401 | 0.9385 |
| 0.0652 | 5.0 | 10000 | 0.1476 | 0.942 |
Framework versions
- Transformers 4.23.1
- Pytorch 1.11.0
- Datasets 2.6.1
- Tokenizers 0.12.1