training-chain

What problem does it solve?

The Training Chain Visualizer renders the full ancestry of SLURM training jobs, turning scattered logs and squeue outputs into a single, easy-to-read chain from the root smoke test to the current running job, with step numbers, W&B slugs, exit states, and transition reasons.

Core Features & Use Cases

• Visualizes the end-to-end training chain for long HPC experiments to simplify debugging and auditability.
• Tracks statuses, transitions, and checkpoint events across multiple jobs to reveal bottlenecks and restarts.
• Use Case: verify that a resumed chain correctly connects from the initial root job to the latest checkpoint, including any auto-resume transitions.

Quick Start

Run the training-chain skill with a target SLURM job ID to render the full execution ancestry from root to current.