sue-interface-check

What problem does it solve?

This skill solves the problem of teachers and operators needing to validate that student-submitted ML experiment codebases are portable, safe, and ready for large-scale runs across HPC and cloud sandboxes without exposing private infrastructure details or wasting expensive compute on broken code.

Core Features & Use Cases

• 4 Hard Gate Enforcement: Automatically checks for the 4 mandatory scale-up readiness controls: real dry-run profiling, inode-saving output mode, first-run smoke test, and committed running README under scale_up_scripts/.
• 22 Prioritized Audit Rules: Evaluates codebases against a ranked set of rules covering W&B logging, CSV result/progress ledgers, Hydra configuration, PyTorch DDP support, and queryable progress monitoring.
• Use Case: A machine learning educator can run this audit on student experiment submissions before allocating GPU resources on LUMI, RunPod, or other sandboxes, catching missing interfaces early to avoid failed runs and wasted cloud costs.

Quick Start

Use the sue-interface-check skill to audit a student-provided ML experiment codebase for portable scale-up readiness before submitting teacher-led training or generation runs to a sandbox backend.