README.md

Test Suite

This directory contains the current pytest-based test suite for GloViTa.

Scope

The tests primarily cover:

• dataset classes
• augmentation policy construction
• dataset factory / dataloader wiring

The test suite is useful for:

• validating refactors
• checking that config/factory integration works
• smoke-testing new dataset additions

Current Test Files

tests/
├── conftest.py
├── test_all_datasets.py
├── test_augmentations.py
├── test_datamodules.py
├── test_generic_image_dataset.py
└── test_datasets.py

test_datamodules.py covers the dataset factory and dataloader path.

Running Tests

Run everything:

pytest

Verbose output:

pytest -v

Run one file:

pytest tests/test_datamodules.py

Markers

Current markers used in the suite include:

• unit
• integration
• requires_data
• slow
• augmentation
• datamodule

Examples:

pytest -m unit
pytest -m "integration and requires_data"
pytest -m "not slow"
pytest -m augmentation

Dataset-Dependent Tests

Some tests require real dataset files to be present locally. Those are marked with:

• integration
• requires_data

If the expected dataset path is missing, those tests usually skip rather than fail.

Common Usage Patterns

Check Dataset Factory Wiring

pytest tests/test_datamodules.py -v

Check Augmentation Construction

pytest tests/test_augmentations.py -v

Run Only Fast Tests

pytest -m "unit and not slow"

When Adding A New Dataset

When you add a new dataset, check at least:

1. the dataset config class exists
2. the dataset is registered in the dataset factory
3. one dataloader batch can be constructed

In practice, that usually means updating:

• tests/conftest.py
• and re-running:
  • tests/test_datasets.py
  • tests/test_datamodules.py

When Adding A New Augmentation Policy

For a new augmentation policy, re-run:

pytest tests/test_augmentations.py -v

This is the quickest sanity check that the policy module still builds valid train/test transforms.