| ## TextAttack Model Card | |
| This `roberta-base` model was fine-tuned for sequence classification using TextAttack | |
| and the glue dataset loaded using the `nlp` library. The model was fine-tuned | |
| for 5 epochs with a batch size of 8, a learning | |
| rate of 2e-05, and a maximum sequence length of 128. | |
| Since this was a regression task, the model was trained with a mean squared error loss function. | |
| The best score the model achieved on this task was 0.9108696741479216, as measured by the | |
| eval set pearson correlation, found after 4 epochs. | |
| For more information, check out [TextAttack on Github](https://github.com/QData/TextAttack). | |