Skip to content

CHANGELOG.md

Latest commit

 

History

History
47 lines (42 loc) · 2.43 KB

CHANGELOG.md

File metadata and controls

47 lines (42 loc) · 2.43 KB
Raw

Changelog

All notable changes to this project are documented in this file. This project adheres to Semantic Versioning.

[2.0.0] - 2026-06-25

Added

  • Installable Python library graphcodebert_interpretability with a clean, documented public API (compare, token_alignment, saliency, project, similarity_matrix, embed, baselines, datasets, set_seed, ...) that works on any code, not just the bundled examples.
  • gcbi command-line interface: compare, saliency, project, heatmap, report, and reproduce.
  • Model-agnostic model loading (load_model) with caching - supports GraphCodeBERT, CodeBERT, UniXcoder and other compatible checkpoints.
  • Automated benchmark reports: generate_report emits a Markdown table, a booktabs LaTeX table and a 300-DPI comparison figure across GraphCodeBERT, AST and TF-IDF lenses.
  • Dataset loaders: bundled sorting corpus plus load_snippets_from_dir and load_jsonl for user-provided code.
  • pytest test suite (40+ fast tests, no model download required) and GitHub Actions CI (lint + tests on Python 3.9-3.12, plus a build job).
  • pyproject.toml packaging with optional extras ([model], [umap], [full], [dev]) and a .gitignore.

Changed

  • The nine analysis scripts (comparison.py, heatmap.py, pca.py, tsne.py, pumap.py, saliency_maps.py, ablation.py, ast-s.py, tf-s.py) are now thin wrappers over the library, eliminating the per-script duplication of the sorting-algorithm corpus and the model-loading boilerplate.
  • README rewritten around the library and CLI, with installation, API quickstart, capability matrix and SEO-friendly headings.
  • Stronger, centralised determinism via set_seed (Python, NumPy, PyTorch).

[1.1.0] - 2026-05-23

Added

  • Comprehensive README overhaul with expanded methodology, usage, results, and citation guidance.
  • requirements.txt for reproducible environment setup.
  • CONTRIBUTING.md with contribution and coding-style guidance.
  • docs/REPRODUCIBILITY.md with end-to-end reproduction steps, runtimes, and determinism notes.
  • examples/demo_similarity.ipynb notebook with an Open in Colab badge.
  • GitHub issue templates for bug reports and feature requests.

[1.0.0] - 2024-09-01

Initial release

  • Initial publication of GraphCodeBERT interpretability scripts and generated visualizations.
  • MIT licensing and citation metadata.