<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
distilbert-base-uncased__sst2__train-32-8
This model is a fine-tuned version of distilbert-base-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.6880
- Accuracy: 0.5014
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: 4
- eval_batch_size: 4
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.712 | 1.0 | 13 | 0.6936 | 0.5385 |
| 0.665 | 2.0 | 26 | 0.6960 | 0.3846 |
| 0.6112 | 3.0 | 39 | 0.7138 | 0.3846 |
| 0.4521 | 4.0 | 52 | 0.8243 | 0.4615 |
| 0.2627 | 5.0 | 65 | 0.7723 | 0.6154 |
| 0.0928 | 6.0 | 78 | 1.2666 | 0.5385 |
| 0.0312 | 7.0 | 91 | 1.2306 | 0.6154 |
| 0.0132 | 8.0 | 104 | 1.3385 | 0.6154 |
| 0.0082 | 9.0 | 117 | 1.4584 | 0.6154 |
| 0.0063 | 10.0 | 130 | 1.5429 | 0.6154 |
| 0.0049 | 11.0 | 143 | 1.5913 | 0.6154 |
Framework versions
- Transformers 4.15.0
- Pytorch 1.10.2+cu102
- Datasets 1.18.2
- Tokenizers 0.10.3
