|
--- |
|
license: mit |
|
datasets: |
|
- vishnun/CodevsNL |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
library_name: transformers |
|
pipeline_tag: text-classification |
|
tags: |
|
- code |
|
- nli |
|
--- |
|
|
|
## PreFace |
|
Code vs Natural language classification using bert-small from prajwall, below are the metrics achieved |
|
|
|
## Training Metrics |
|
|
|
| Epoch | Training Loss | Validation Loss | Accuracy | |
|
|-------|---------------|-----------------|----------| |
|
| 1 | 0.022500 | 0.012705 | 0.997203 | |
|
| 2 | 0.008700 | 0.013107 | 0.996880 | |
|
| 3 | 0.002700 | 0.014081 | 0.997633 | |
|
| 4 | 0.001800 | 0.010666 | 0.997526 | |
|
| 5 | 0.000900 | 0.010800 | 0.998063 | |
|
|
|
## More |
|
|
|
- Github repo for installable python package: https://github.com/Vishnunkumar |
|
- Space on the extraction of code blocks from screenshots: https://huggingface.co/spaces/vishnun/SnapCode |
|
|
