redai-infra
- 16 skills
- 0 followers
- 7 hours ago last updated
- ▌ Relax Dev Debug · redai-infraDevelop and debug the Relax reinforcement learning project. Use this skill whenever modifying code in the relax/ directory, or running remote training jobs on a Ray cluster for validation. Also use it when the user mentions training, debugging training runs, submitting Ray jobs, or fixing training errors.
- ▌ Debug Hang · redai-infra bundle自动排查 Ray 调度的分布式训练任务 hang 问题。使用当训练任务无响应、资源利用率异常、任务长时间无进度时。自动收集集群状态、任务调用栈、Actor 状态，分析阻塞链条并定位根因。
- ▌ Doc Writer · redai-infra bundleWrite and maintain bilingual (English + Chinese) documentation for the Relax project. Use when user asks to create, update, or translate documentation pages. Ensures format correctness (VitePress, sidebar config, bilingual parity) and content correctness (matches actual codebase, no fabricated features).
- ▌ Git Commit · redai-infra bundleCreates git commits following Conventional Commits format with type/scope/subject and detailed markdown body. Use when user wants to commit changes, create commit, save work, or stage and commit. Enforces project-specific conventions from CLAUDE.md. Each change type gets its own markdown heading (# emoji + type), with detailed item lists under each.
- ▌ Opd Tuning · redai-infra bundleTune Relax On-Policy Distillation (OPD/MOPD) runs — teacher SGLang engine knobs, logits-memory limits, student/teacher GPU split, and multimodal caches. Use when a user asks to speed up or debug an OPD run, mentions "OPD", "MOPD", "on-policy distillation", "蒸馏", a teacher engine OOM, teacher timeouts, idle student/teacher GPUs, or any `--teacher-sglang-*` / `--opd-*` flag. Produces a diagnosis report with cited flags, measured evidence, and concrete fixes; applies the edits only when asked.
- ▌ Code Review · redai-infra bundleExpert code review of current git changes with a senior engineer lens. Detects SOLID violations, security risks, Python anti-patterns, and ML/distributed training issues. Tailored for the Relax reinforcement learning framework.
- ▌ Perf Doctor · redai-infra bundleDiagnose Relax training launch scripts for misconfigured flags that hurt performance (time/MFU) or waste GPU memory (cards needed). Use when user asks to review/audit/check a training script, mentions "perf doctor", suspects a config is slow or OOM-prone, or wants a sanity check before launching. Produces a two-section markdown report (Performance + Memory) with cited flags, severity, and concrete fixes.
- ▌ Sync Github · redai-infra bundleUse when syncing Relax code between internal GitLab and external GitHub, especially gitlab/dev, gitlab/main, github/main, internal CR/MR handoff, linear main history, sensitive-content checks, GitHub Actions CI validation, or guarded GitHub pushes.
- ▌ Verl To Relax · redai-infra bundleMigrate RL training recipes from verl to Relax framework. Use when user wants to port reward functions, tool environments, training scripts, or any recipe code from the verl (volcengine/verl) codebase to Relax. Handles reward, rollout, tool/env, dataset, and launch script conversion. Supports both colocate (default) and fully async deployment modes.
- ▌ Sglang Upgrade · redai-infra bundleUpgrade the sglang version used by Relax's training Docker image. Use when bumping/upgrading sglang (changing the base image tag), rebasing docker/patch/latest/sglang.patch onto a new sglang release, or porting Relax's sglang customizations forward to a new version. Covers the version-determination mechanism, using THUDM/slime as the patch-rebase reference, classifying private patches into upstreamed-vs-still-needed, and validating the rebased patch against the exact target source.
- ▌ Agentic Rollout · redai-infra bundleAssess an agent, bridge it into Relax, check launch readiness, and demonstrate resident Agentic rollout with evidence. Use when evaluating an agent app before integration, connecting an external agent with --use-agentic-rollout, auditing model-visible context linearity or SessionForest export, checking remote or large-scale resident concurrency, or preparing an Agentic experiment.
- ▌ Creating Skills · redai-infra bundleGuide for creating Claude Code skills following Anthropic's official best practices. Use when user wants to create a new skill, build a skill, write SKILL.md, update an existing skill, or needs skill creation guidelines. Provides structure, frontmatter fields, naming conventions, and new features like dynamic context injection and subagent execution.
- ▌ Ssh Ray Cluster · redai-infra3-step debug loop for remote Ray cluster — submit task via SSH, check logs locally, analyze errors and fix code, repeat until resolved.
- ▌ Model Integration · redai-infra bundleGuide for integrating a new model into the Relax training pipeline. Use when adding a new model architecture, writing Megatron-to-HF weight converters, implementing custom TP all-gather/chunk logic, debugging weight sync issues, or adapting models for colocate or fully-async mode. Covers Megatron backend (bridge and raw modes) and FSDP backend.
- ▌ Redaccel To Relax · redai-infra bundleMigrate RL training algorithms from RedAccel to Relax framework. Use when user wants to port reward functions, agent environments, training scripts, or any algorithm code from the RedAccel (redaccelrl) codebase to Relax. Handles reward, environment, rollout, and launch script conversion.
- ▌ Nemo Gym Recipe Integration · redai-infraIntegrate a new NVIDIA NeMo Gym environment into Relax as a three-step recipe. Use when adding or debugging a recipe under examples/nemo_gym_agentic/recipes; covers data preparation, a local private Gym service, direct Ray training launch, verifier validation, callback networking, lifecycle cleanup, and failure triage.