Skip to content

README.md

Latest commit

 

History

History
92 lines (68 loc) · 2.45 KB

README.md

File metadata and controls

92 lines (68 loc) · 2.45 KB
Raw

mpBAX Tests

Comprehensive test suite for the mpBAX framework.

Test Structure

Tests are organized by component for better maintainability:

  • test_data_handler.py - DataHandler: add/get data, loop tracking, save/load, multi-output
  • test_evaluator.py - Evaluator: oracle evaluation, shape validation, eval counting
  • test_model.py - Models: DummyModel, custom models, train/predict, save/load
  • test_algorithm.py - Algorithms: RandomSampling, GreedySampling, proposals
  • test_checkpoint.py - CheckpointManager: save/load/rollback, oracle naming
  • test_engine.py - Engine integration: instance/string configs, multi-oracle, finetune/retrain modes

Running Tests

With installation (recommended):

First, install mpBAX in development mode with test dependencies:

cd /path/to/mpBAX
pip install -e ".[dev]"

Then run tests:

# Run all tests
pytest tests/ -v

# Run specific test file
pytest tests/test_engine.py -v

# Or run individual files directly
python tests/test_data_handler.py
python tests/test_evaluator.py

Without installation:

Set PYTHONPATH and run tests:

export PYTHONPATH=/path/to/mpBAX:$PYTHONPATH

# With pytest
pytest tests/ -v

# Or directly
python tests/test_engine.py

Test Coverage

Tests cover:

  • ✓ Core components (DataHandler, Evaluator, Model, Algorithm, CheckpointManager)
  • ✓ Engine end-to-end integration
  • ✓ Instance-based config (Python API)
  • ✓ String-based config (YAML compatibility)
  • ✓ Multi-oracle optimization
  • ✓ Multi-output oracles
  • ✓ Finetune vs retrain modes
  • ✓ Loop tracking and metadata
  • ✓ Checkpoint/resume/rollback functionality
  • ✓ Shape validation and error handling

Adding New Tests

When adding functionality to mpBAX:

  1. Add test to appropriate test file (or create new file if needed)
  2. Follow existing test patterns: clear names, assertions, print statements
  3. Test both success and error cases
  4. Run full test suite to ensure no regressions

Example:

def test_my_feature():
    """Test description."""
    print("Testing my feature...")

    # Test code here
    assert something == expected

    print("  ✓ My feature tests passed")

Test Principles

  • Simple and direct: Each test function tests one thing
  • Instance-based: Use direct instances, not string imports (easier to test)
  • Isolated: Each test is independent, uses temp directories
  • Clear output: Print statements show progress and what passed