chore: Release 0.5.0

.github/workflows/release.yml

@@ -26,15 +26,12 @@ jobs:
       - name: Install dependencies
         run: |
           python3 -m pip install --upgrade pip
-          python3 -m pip install --upgrade build setuptools setuptools_scm
+          python3 -m pip install --upgrade build setuptools
 
       - name: Setup | Force correct release branch on workflow sha
         run: |
           git checkout -B ${{ github.ref_name }} ${{ github.sha }}
 
-      - name: Verify dynamic version
-        run: python -m setuptools_scm
-
       - name: Use Python Semantic Release to prepare release
         id: release
         uses: python-semantic-release/python-semantic-release@v8.3.0
@@ -66,10 +63,10 @@ jobs:
           git checkout -b version-bump-${{ steps.release.outputs.tag }}
 
           # Add the changes made by semantic-release (version update and changelog)
-          git add CHANGELOG.md
+          git add CHANGELOG.md pyproject.toml
 
           # Commit the changes (version bump and changelog update)
-          git commit -m "chore: Update changelog after version bump ${{ steps.release.outputs.tag }}"
+          git commit -m "chore: bump version to ${{ steps.release.outputs.tag }} and update changelog"
 
           # Push changes to the new branch
           git push origin version-bump-${{ steps.release.outputs.tag }}

.gitignore

@@ -93,3 +93,4 @@ scripts/plot_results_LOCAL.py
 scripts/**/plot_*_LOCAL.sh
 scripts/PLOT_ALL_LOCAL.sh
 slurm_LOCAL/*
+.pypirc

CHANGELOG.md

@@ -1 +1,38 @@
+# Changelog
 
+## 0.5.0 (2026-05-11)
+
+This release reworks `quanda` around ready-to-use benchmark assets and
+extends it beyond image classification to text classification and causal
+language modeling.
+
+### Features
+
+- **Explainers**
+  - New wrappers: Kronfluence, `dattri` explainers
+  - Support for text classification and causal language modeling tasks
+- **Benchmarks**
+  - Fully refactored interface
+  - Full MNIST benchmark suite
+  - Full CIFAR-10 / ResNet-9 benchmark suite
+  - BERT/QNLI benchmarks (mislabeling, mixed datasets, LDS, SGD variants)
+  - AwA2 / ResNet-50 benchmark configurations
+  - Fact-tracing benchmarks for GPT-2 smal/ T-REx
+- **Metrics**
+  - Recall-at-s Class/Subclass Detection metrics
+- **Tooling**
+  - Metadata class for handling transformed dataset metadata
+  - YAML-based benchmark configuration
+
+### Fixes & refactors
+
+- Numerous fixes across benchmark training, dataset handling
+  (splits, HF cache reuse, shuffling), checkpoint loading, and device
+  placement
+- Consistent benchmark flags (e.g., `use_prediction`) aligned with originating papers
+- Reworked explainer dataset/device handling
+
+### Docs
+
+- Updated README, docs, and tutorials
+- Quickstart and integration tests are now used as doc snippets

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "quanda"
-dynamic = ["version"]
+version = "0.5.0"
 authors = [
   { name="Dilyara Bareeva", email="dilyabareeva@gmail.com" },
     {name = "Galip Ümit Yolcu", email = "galip.uemit.yolcu@hhi.fraunhofer.de" },
@@ -46,13 +46,9 @@ Homepage = "https://github.com/dilyabareeva/quanda"
 Issues = "https://github.com/dilyabareeva/quanda/issues"
 
 [build-system]
-requires = ["setuptools>=42", "setuptools-scm[toml]>=6.0"]
+requires = ["setuptools>=42"]
 build-backend = "setuptools.build_meta"
 
-[tool.setuptools_scm]
-version_scheme = "post-release"
-local_scheme = "node-and-date"
-
 [tool.isort]
 profile = "black"
 line_length = 79