Add Python Testing Infrastructure

Summary

This PR establishes a comprehensive testing infrastructure for the Vision Transformer (ViT) explainability project using Poetry as the package manager and pytest as the testing framework.

Changes Made

Package Management

• Poetry Configuration: Created pyproject.toml with Poetry setup as the primary package manager
• Dependencies: Migrated project dependencies (torch, torchvision, Pillow, numpy, opencv-python) to Poetry
• Development Dependencies: Added pytest, pytest-cov, and pytest-mock as dev dependencies

Testing Framework Setup

• pytest Configuration:

  • Configured test discovery patterns
  • Set up coverage reporting with 80% threshold
  • Added custom markers: unit, integration, and slow
  • Configured both HTML and XML coverage report generation

• Directory Structure:

  tests/
  ├── __init__.py
  ├── conftest.py          # Shared fixtures
  ├── unit/                # Unit tests
  │   └── __init__.py
  ├── integration/         # Integration tests
  │   └── __init__.py
  └── test_infrastructure_validation.py  # Setup validation
  

Test Fixtures

Created comprehensive fixtures in conftest.py:

• temp_dir: Temporary directory management
• sample_image: Test image generation
• sample_tensor: PyTorch tensor creation
• mock_model: Mock Vision Transformer model
• attention_rollout_config & grad_rollout_config: Test configurations
• mock_attention_weights & mock_gradients: Mock data for testing
• output_dir: Test output directory management
• reset_torch_hub_dir: Prevents model downloads during tests

Additional Configuration

• Updated .gitignore: Added Python testing artifacts, virtual environments, and IDE files
• Poetry Scripts: Configured both poetry run test and poetry run tests commands

How to Use

1. Install dependencies:

   poetry install

2. Run all tests:

   poetry run test
   # or
   poetry run tests

3. Run specific test types:

   # Only unit tests
   poetry run pytest -m unit
   
   # Skip slow tests
   poetry run pytest -m "not slow"
   
   # Only integration tests
   poetry run pytest -m integration

4. View coverage report:

   # After running tests, open the HTML report
   open htmlcov/index.html

Notes

• The testing infrastructure is ready for immediate use
• Coverage is configured with an 80% threshold
• The validation test suite verifies the setup is working correctly
• OpenCV (cv2) may require additional system dependencies on some platforms
• Poetry.lock file is created and should be committed to ensure reproducible builds

Next Steps

Developers can now start writing unit and integration tests for the existing modules:

• vit_explain.py
• vit_grad_rollout.py
• vit_rollout.py