Standardizes experiment runs with configurable parameters and structured output formats.
npx skills add https://github.com/gyy0592/claude-config --skill experiment-run
Please help me install this Agent Skill. Skill: experiment-run Source: https://github.com/gyy0592/claude-config/tree/main/skills/experiment-run Command: npx skills add https://github.com/gyy0592/claude-config --skill experiment-run
Ensures every computational experiment is launched, recorded, and validated so results are reproducible, discoverable, and auditable months later. It prevents lost metrics, hidden hyperparameters, and irreproducible outputs by enforcing config-driven runs, structured output directories, and immediate recording.
Prepare a config, negotiate what metrics and per-item fields to record, confirm the pre-flight summary, then create the exp/ directory, snapshot the code, and generate and submit the job.
High-intent search queries and answers about installing and using this skill.
To make experiments reproducible on a Slurm cluster, enforce config-driven parameters, generate Slurm job scripts, snapshot code, and record structured metrics in scalars.csv and events.jsonl formats for consistent auditing.
The best way to record training metrics for reproducibility is enforcing config-first runs and capturing outputs in structured exp/ directories with exact scalars.csv and events.jsonl schemas to prevent lost metrics and hidden hyperparameters.
To validate experiment outputs after data processing, run a post-run validation step that checks the structured exp/ directory and produces a run_checks.json file, ensuring all recorded metrics and artifacts match the initial config.
Yes, config-driven experiment tracking works for local data processing workflows by generating local job scripts, performing pre-flight confirmation, and enforcing structured output directories to preserve checkpoints and metrics.
Pre-flight confirmation and code snapshots are needed before submitting a training job to capture the exact runtime state and config parameters, preventing irreproducible outputs and ensuring the run_manifest accurately reflects the executed code.
Yes, this experiment recording approach supports per-sample logging during repeated evaluation experiments by negotiating intermediate and artifact fields interactively, capturing them incrementally with flushes into structured events.jsonl files.
maggitDiagnose environment setup, dependencies, and configuration issues in a project.
ovachieverVisualize ML training metrics and debugging signals via interactive dashboards.
yo-stevenCompile TensorRT-LLM on a SLURM cluster via enroot container batch jobs.
kaminocorpLog metrics, artifacts, and hyperparameters for ML experiment tracking.
