perform-sweep
Design, configure, launch, and analyze ablation sweeps for GRPO training. Use for hypothesis testing, hyperparameter experiments, and systematic comparisons.
Full skill instructions
End-to-end workflow for running ablation experiments on the Diplomacy GRPO training pipeline.
| Phase | Action | Command |
|---|---|---|
| Configure | Create sweep.yaml | See YAML Reference |
| Validate | Dry run | python scripts/launch_sweep.py <path> --dry-run |
| Info | Show config | python scripts/launch_sweep.py <path> --info |
| Launch | Start sweep | python scripts/launch_sweep.py <path> |
| Status | Check progress | python scripts/launch_sweep.py <path> --status |
| List | List all sweeps | python scripts/launch_sweep.py --list |
| Analyze | Compare results | Use experiment-analysis skill |
experiments/experiment-tracker.mdhypothesis fieldCreate experiments/sweeps/<name>/sweep.yaml:
metadata:
name: "my-ablation"
description: "Testing hypothesis X"
hypothesis: "Longer horizons should improve strategic play"
experiment_tag_prefix: "my-ablation"
defaults:
total_steps: 100
runs:
A:
name: "control"
description: "Baseline configuration"
config:
experiment_tag: "${metadata.experiment_tag_prefix}-A"
B:
name: "treatment"
description: "With longer horizon"
config:
rollout_horizon_years: 8
experiment_tag: "${metadata.experiment_tag_prefix}-B"
See YAML Reference for full schema.
# Show sweep info
python scripts/launch_sweep.py experiments/sweeps/<name>/ --info
# Dry run (validates config, shows what would run)
python scripts/launch_sweep.py experiments/sweeps/<name>/ --dry-run
# Launch (fire-and-forget - runs in cloud)
python scripts/launch_sweep.py experiments/sweeps/<name>/
# Check status anytime
python scripts/launch_sweep.py experiments/sweeps/<name>/ --status
# List all sweeps
python scripts/launch_sweep.py --list
After sweep completes, use the experiment-analysis skill:
# Full analysis for each run
uv run python .claude/skills/experiment-analysis/analyze_elo.py <run-name>
# Compare in WandB
# Filter by experiment_tag_prefix (e.g., "my-ablation")
See existing sweeps in experiments/sweeps/:
longer-horizon-inverted-weight-ablation/ - 2x2 ablation on horizon and scoringexperiment-analysis skill for post-sweep metrics analysisexperiment_tag for filteringresults.md
Comprehensive Spring Framework and Spring Boot best practices including dependency injection patterns, bean lifecycle and scopes, REST API development, Spring Data JPA, service layer design, Spring Security, testing strategies, caching, AOP, async processing, error handling, and common anti-patte...
Dispatch to the right problem-solving technique based on how you're stuck
TypeScript厳格モードによる型安全性設計を専門とするスキル。 📖 参照書籍: - 『Effective TypeScript』（Dan Vanderkam）: 型設計 📚 リソース参照: - `resources/Level1_basics.md`: レベル1の基礎ガイド - `resources/Level2_intermediate.md`: レベル2の実務ガイド - `resources/Level3_advanced.md`: レベル3の応用ガイド - `resources/Level4_expert.md`: レベル4の専門ガイド - `resources/di...
Deploys applications to cloud Kubernetes (AKS/GKE/DOKS) with CI/CD pipelines. Use when deploying to production, setting up GitHub Actions, troubleshooting deployments. Covers build-time vs runtime vars, architecture matching, and battle-tested debugging.
19-agent team structure, decision trees for agent selection, Haiku vs Sonnet model selection, and agent collaboration principles. Use when deciding which sub-agent to invoke, understanding team responsibilities, or learning multi-agent orchestration.
Converts architecture descriptions, module specs, or workflow docs into Mermaid diagrams. Use when visualizing brick module relationships, workflows (DDD, investigation), or system architecture. Supports: flowcharts, sequence diagrams, class diagrams, state machines, entity relationship diagrams,...
INVOKE FIRST before any code work. Validates git workflow (branch, issue, worklog) and checks approach. Use at START of every task and END before completing. Prevents skipped steps.
This skill should be used when working with browser-rendered artifacts (i.e. Bun, React/React Native) to proactively validate that development work on localhost is being built appropriately and to support debugging browser-rendered content. Use this skill after making frontend changes to verify t...
Lead Développeur - Coordination technique opérationnelle, code review, mentoring et livraison. Pair de web-dev-process au niveau OPÉRATIONS.
Monitor Docker workers, RQ queue health, and auto-scale workers (max 4). Use when: checking job progress, monitoring queue depth, scaling workers up/down, diagnosing slow processing, or waiting for jobs to complete. Referenced by data-quality skill during reprocessing.
Proactive quality guidance system that monitors artifact creation and provides real-time feedback on documentation quality
This skill should be used when the user wants to bump the version number in the workspace. It updates versions across all pyproject.toml files (root, tde, and tda packages) and the CHANGELOG.md to keep them synchronized.
Generate, edit, and beat-sync AI video with leading models in one workspace.
The world's fastest calendar for remote work
Transform Your Design with AI Designer by ImgCreator.ai
Revolutionizing Video Production with AI-Powered Creativity
Extend an image past the frame and let AI fill the new aspect ratio.
Discover your celebrity doppelgänger with StarByFace!
ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.
Opus.ai: Revolutionize Your Web Experience