tests/
├── conftest.py                 # Pytest configuration and fixtures
├── pytest.ini                  # Pytest settings (in parent directory)
├── smoke_test.py              # Basic import checks (no model loading)
└── integration_real/          # Real model integration tests
    └── test_da3_small_real.py # Tests with actual DA3 models on bridge image

Test Types

1. Smoke Tests

File: smoke_test.py Purpose: Verify basic imports and module structure Runtime: < 5 seconds Requirements: No model downloads needed

Run with:

python tests/smoke_test.py

2. Integration Tests (Real Models)

Directory: integration_real/ Purpose: Test actual model loading and inference on the bridge image Runtime: ~2-5 minutes (first run downloads ~300MB model) Requirements: Downloads DA3 Small model from HuggingFace

Run with:

# Run all integration tests
pytest tests/integration_real/ -v -m real_model

# Run specific test
pytest tests/integration_real/test_da3_small_real.py::test_da3_small_inference_bridge -v -s

GitHub Actions Workflows

Smoke Tests (Fast)

Linux: .github/workflows/test-unit.yml
Windows: .github/workflows/test-windows.yml
macOS: .github/workflows/test-macos.yml

Runs on: Python 3.10, 3.11, 3.12 Duration: ~1 minute per platform

Integration Tests (Slow)

Linux: .github/workflows/test-linux-integration.yml
Windows: .github/workflows/test-windows-integration.yml
macOS: .github/workflows/test-macos-integration.yml

Runs on: Python 3.10 Duration: ~3-5 minutes per platform (cached models make reruns faster)

Artifacts: Each integration test uploads depth map outputs as artifacts for visual inspection.

Test Image

All integration tests use the bridge image from assets/bridge.jpeg:

Black and white architectural bridge photo
Resized to 768x768 for testing
Good depth variation (close bridge structure, distant background)

Output Artifacts

Integration tests save outputs to tests/test_outputs/:

bridge_input.png - Resized input image
bridge_da3_small_depth.png - Generated depth map

These are uploaded as GitHub Actions artifacts for inspection.

Running Tests Locally

Install test dependencies:

pip install -r .github/requirements-ci.txt
pip install -r requirements-dev.txt
pip install pytest pillow

Run smoke tests:

python tests/smoke_test.py

Run integration tests:

# Run all real model tests
pytest tests/integration_real/ -v -m real_model -s

# Run with GPU (if available)
pytest tests/integration_real/ -v -m real_model --use-gpu

Markers

@pytest.mark.unit - Fast unit tests (not implemented yet)
@pytest.mark.real_model - Tests that download and use real models

Model Caching

Models are cached in ~/.cache/huggingface/ to speed up subsequent test runs.