---
license: gpl-3.0
datasets:
- OSS-forge/PyResBugs
base_model:
- Salesforce/codet5p-770m
pipeline_tag: text-generation
library_name: transformers
---

This model is a CodeT5+ 770M sequence-to-sequence model fine-tuned on the PyResBugs dataset, a curated collection of 5,007 residual Python bugs paired with their fixed versions and multi-level natural-language descriptions.
This model is designed for bug generation: given a clean Python function withing a natural-language description, it produces an implementation intentionally containing a realistic software bug.

This enables controlled fault injection, robustness testing of APR systems, fuzzing pipelines, and research on the security posture and reliability of code LLMs.