ml-validate

What problem does it solve?

ML projects often fail at runtime due to broken Hydra configuration, invalid imports for target classes, missing project structure, or basic code-quality/dependency problems that only surface after you start training.

Core Features & Use Cases

• Project structure validation: Confirms required directories and key files like src/train.py and configs/config.yaml exist.
• Configuration validation: Checks YAML syntax and verifies Hydra config composition and required top-level fields (model, data, trainer).
• Training readiness checks: Runs ruff for code quality, imports required/optional dependencies, checks CUDA availability, and exercises model and DataModule instantiation plus a fast dev run.
• Use case: Before running expensive experiments, validate that your Hydra target paths import correctly and that the model/data/trainer components can be instantiated and execute a minimal train loop.

Quick Start

Run the full validator by executing the validation script in your project: python scripts/validate_project.py