metadata

license: mit
dataset_info:
  features:
    - name: code
      dtype: string
    - name: label
      dtype: string
  splits:
    - name: train
      num_bytes: 3851220
      num_examples: 2493
  download_size: 1555926
  dataset_size: 3851220
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

Python Linting Dataset

Python Static Analysis Dataset created by running pylint on stack-v2 and manually labelling other issues not found by pylint. Data Augmentation is used on the manually labelled dataset using python's ast module to parse the code and modify name nodes.

Scripts to download more data or modify this dataset further:

Github Repository

use augmentation.py to augment your manually labelled data.
run stackv2.py to download stackv2 python files with linenos < 200, then run genlabels.py to generate the lables from pylint.
append augmented dataset to this to get the combined dataset for training.
use train.py to fine-tune using unsloth.

Model Trained on this dataset

To use this Linter in CLI:

run ollama create llm4lint7b -f
pip install llm4lint To use it as a vscode extension: https://github.com/ahmedhus22/llm4lint-vsc-extension