| | --- |
| | license: apache-2.0 |
| | base_model: |
| | - microsoft/deberta-v3-base |
| | pipeline_tag: text-classification |
| | library_name: transformers |
| | tags: |
| | - deberta-v2 |
| | - secrets |
| | - secret-detection |
| | - secure-codeing |
| | - Github-pipeline |
| | - deberta |
| | - cybersecurity |
| | - code-analysis |
| | --- |
| | |
| | # DeBERTa Secret Detection Model |
| |
|
| | This model detects secrets in git diff lines. |
| |
|
| | ## Model Details |
| |
|
| | - Base Model: microsoft/deberta-base |
| | - Task: Binary sequence classification |
| | - Labels: |
| | - LABEL_0: Normal |
| | - LABEL_1: Secret |
| |
|
| | ## Training |
| |
|
| | - Loss: Weighted cross-entropy |
| | - Metric for best model: F1 |
| | - BF16 training |
| | - Gradient checkpointing enabled |
| |
|
| | ## Usage |
| |
|
| | ```python |
| | from transformers import pipeline |
| | |
| | classifier = pipeline( |
| | "text-classification", |
| | model="hypn05/secret-detector-deberta-base" |
| | ) |
| | |
| | classifier("+ password='secret123'") |
| | ``` |
| |
|
| | ### Intended Use |
| |
|
| | Production secret detection in git diffs. |
| |
|
