ml-code-standards

What problem does it solve?

Enforces reproducibility, consistent configuration, and opinionated coding standards so machine learning experiments remain traceable, debuggable, and repeatable from day one rather than becoming ad hoc and unreplicable later.

Core Features & Use Cases

• Shape documentation: Requires inline tensor shape comments and concise math-oriented variable naming to make debugging and reviews faster.
• Hydra-first configuration: Mandates Hydra YAML configs and override patterns for consistent experiment parameterization and composition.
• Reproducible run infrastructure: Defines per-run directories with frozen config snapshots, logs, PIDs, and checkpointing plus guidance for MLflow integration and background training scripts.
• Use case: During code review or onboarding, apply these standards to immediately surface missing configs, absent shape annotations, or missing run-management scripts and produce a prioritized remediation plan.

Quick Start

Ask the assistant to audit the repository for reproducibility gaps including missing Hydra configs, absent shape comments, no per-run snapshots, and missing run-management scripts and return a prioritized list of fixes.