[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78T4C0ZPY6M20XMNJ1WH18X":3,"skill-same:01M78T4C0ZPY6M20XMNJ1WH18X":71},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":46},"01M78T4C0ZPY6M20XMNJ1WH18X","rl-and-rewards",null,"Guides agents configuring Axolotl GRPO, EBFT, custom reward functions, vLLM training servers, NeMo Gym, Hatchery, and async online RL checks.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: rl-and-rewards\ndescription: \"Guides agents configuring Axolotl GRPO, EBFT, custom reward\n  functions, vLLM training servers, NeMo Gym, Hatchery, and async online RL\n  checks.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# RL and Rewards\n\nUse this sub-skill when the task involves Axolotl online RL or reward-driven training: GRPO, GDPO-style reward aggregation, custom reward functions, `axolotl vllm-serve` for training, EBFT feature-matching rewards, NeMo Gym environment rewards, Hatchery-style remote RL flows, or async rollout debugging.\n\n## Route By Task\n\n- For GRPO configs, vLLM serving, async prefetch, replay, re-roll, and importance sampling, read [references\u002Fgrpo.md](references\u002Fgrpo.md).\n- For EBFT structured or strided feature-matching runs, read [references\u002Febft.md](references\u002Febft.md).\n- For reward function signatures, deterministic checks, multiple rewards, rollout functions, NeMo Gym rewards, and Hatchery reward hooks, read [references\u002Freward-functions.md](references\u002Freward-functions.md).\n- For symptom-driven fixes around rewards, vLLM, async rollouts, EBFT mode selection, and resource constraints, read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n- Before training, validate a local reward function with [scripts\u002Fvalidate_reward_function.py](scripts\u002Fvalidate_reward_function.py): `python scripts\u002Fvalidate_reward_function.py rewards.py:accuracy_reward`.\n\n## Quick Workflow\n\n1. Classify the method first: `rl: grpo` for programmatic verifiers or environment rewards, `rl: ebft` for feature matching against ground-truth completions, and route DPO\u002FKTO\u002FORPO\u002FSimPO preference data to `preference-tuning`.\n2. For GRPO, define the YAML, reward module, and vLLM plan together; `trl.reward_funcs`, `trl.num_generations`, `trl.max_completion_length`, `trl.use_vllm`, and `vllm.host`\u002F`vllm.port` must agree.\n3. For EBFT, choose `ebft.mode` before copying settings: structured mode uses prompt plus `ground_truth` and usually vLLM, while strided mode uses tokenized document anchors and does not require vLLM.\n4. Run local deterministic reward checks before `axolotl train`; then use `axolotl preprocess config.yaml --debug` only for data\u002Ftokenization visibility and `axolotl train config.yaml` only after the user accepts runtime cost.\n5. For server-backed RL, start or verify the generation\u002Fenvironment services before training; `axolotl vllm-serve config.yaml` must serve the same base model as the training config.\n\n## Boundaries\n\n- This sub-skill owns GRPO\u002FGDPO online RL, reward functions, vLLM serving relationships, async rollout features, EBFT structured\u002Fstrided setup, NeMo Gym reward routing, Hatchery reward hooks, and safe local reward validation.\n- Route static DPO, KTO, ORPO, SimPO, reward-model, and process-reward-model preference workflows to `preference-tuning`.\n- Route generic dataset format mapping, `chat_template`, schema discovery, and YAML-only validation to `data-and-configs`.\n- Route cluster launch, DeepSpeed\u002FFSDP tuning, GPU placement, and throughput engineering to `distributed-and-performance`.\n- Route model architecture quirks, tokenizer templates, LoRA\u002FQLoRA target modules, and adapter loading to `model-loading-and-adapters`.\n- Route CLI mechanics, fetch\u002Fdocs commands, inference, and merge-lora operations to `cli-and-operations`.\n\n## Evidence Notes\n\nThis guidance is distilled from Axolotl GRPO, EBFT, vLLM serving, RLHF, schema, integration, example, and unit-test evidence. It does not claim live training, vLLM, GPU, model-loading, or environment-service verification.\n","# RL and Rewards\n\nUse this sub-skill when the task involves Axolotl online RL or reward-driven training: GRPO, GDPO-style reward aggregation, custom reward functions, `axolotl vllm-serve` for training, EBFT feature-matching rewards, NeMo Gym environment rewards, Hatchery-style remote RL flows, or async rollout debugging.\n\n## Route By Task\n\n- For GRPO configs, vLLM serving, async prefetch, replay, re-roll, and importance sampling, read [references\u002Fgrpo.md](references\u002Fgrpo.md).\n- For EBFT structured or strided feature-matching runs, read [references\u002Febft.md](references\u002Febft.md).\n- For reward function signatures, deterministic checks, multiple rewards, rollout functions, NeMo Gym rewards, and Hatchery reward hooks, read [references\u002Freward-functions.md](references\u002Freward-functions.md).\n- For symptom-driven fixes around rewards, vLLM, async rollouts, EBFT mode selection, and resource constraints, read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md).\n- Before training, validate a local reward function with [scripts\u002Fvalidate_reward_function.py](scripts\u002Fvalidate_reward_function.py): `python scripts\u002Fvalidate_reward_function.py rewards.py:accuracy_reward`.\n\n## Quick Workflow\n\n1. Classify the method first: `rl: grpo` for programmatic verifiers or environment rewards, `rl: ebft` for feature matching against ground-truth completions, and route DPO\u002FKTO\u002FORPO\u002FSimPO preference data to `preference-tuning`.\n2. For GRPO, define the YAML, reward module, and vLLM plan together; `trl.reward_funcs`, `trl.num_generations`, `trl.max_completion_length`, `trl.use_vllm`, and `vllm.host`\u002F`vllm.port` must agree.\n3. For EBFT, choose `ebft.mode` before copying settings: structured mode uses prompt plus `ground_truth` and usually vLLM, while strided mode uses tokenized document anchors and does not require vLLM.\n4. Run local deterministic reward checks before `axolotl train`; then use `axolotl preprocess config.yaml --debug` only for data\u002Ftokenization visibility and `axolotl train config.yaml` only after the user accepts runtime cost.\n5. For server-backed RL, start or verify the generation\u002Fenvironment services before training; `axolotl vllm-serve config.yaml` must serve the same base model as the training config.\n\n## Boundaries\n\n- This sub-skill owns GRPO\u002FGDPO online RL, reward functions, vLLM serving relationships, async rollout features, EBFT structured\u002Fstrided setup, NeMo Gym reward routing, Hatchery reward hooks, and safe local reward validation.\n- Route static DPO, KTO, ORPO, SimPO, reward-model, and process-reward-model preference workflows to `preference-tuning`.\n- Route generic dataset format mapping, `chat_template`, schema discovery, and YAML-only validation to `data-and-configs`.\n- Route cluster launch, DeepSpeed\u002FFSDP tuning, GPU placement, and throughput engineering to `distributed-and-performance`.\n- Route model architecture quirks, tokenizer templates, LoRA\u002FQLoRA target modules, and adapter loading to `model-loading-and-adapters`.\n- Route CLI mechanics, fetch\u002Fdocs commands, inference, and merge-lora operations to `cli-and-operations`.\n\n## Evidence Notes\n\nThis guidance is distilled from Axolotl GRPO, EBFT, vLLM serving, RLHF, schema, integration, example, and unit-test evidence. It does not claim live training, vLLM, GPU, model-loading, or environment-service verification.\n",{"name":5,"license":23,"metadata":24,"description":7,"disable-model-invocation":26},"Apache 2.0",{"disco-role":25},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":26},[36,39,42],{"tool":37,"command":38},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-and-rewards",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-and-rewards",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-and-rewards","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002FSKILL.md",[47,53,57,61,65],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Febft.md","REFERENCE","参考文档",5160,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002Freferences\u002Febft.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fgrpo.md",6399,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002Freferences\u002Fgrpo.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Freward-functions.md",5252,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002Freferences\u002Freward-functions.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Ftroubleshooting.md",6523,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002Freferences\u002Ftroubleshooting.md",{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fvalidate_reward_function.py","SCRIPT","脚本",9589,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faxolotl\u002Fsub-skills\u002Frl-and-rewards\u002Fscripts\u002Fvalidate_reward_function.py",[72,79,86,93,100,107],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M799NYSH1R3RVAHSM2D88Y84","passnet-triton-opt","Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block\u002Fgrid\u002Fwarp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"publishedAt":18},"01M799NS5APEC0N1FZ7XZVW6BN","passnet-solve","END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-solve\u002FSKILL.md",[],[],{"skillNo":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"publishedAt":18},"01M799NNY9BKSM21V5R7QVTBJX","passnet-skill","PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under .\u002Fpass_dir\u002F, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-skill\u002FSKILL.md",[],[],{"skillNo":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"publishedAt":18},"01M799NHH6NP7Z1RK77KGXAN6S","passnet-pattern-fusion","Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()\u002Freplacement_args()\u002Freplacement_func() files, when a pass \"failed to match\", or when deciding how to fuse multiple kernels\u002Fops.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-pattern-fusion\u002FSKILL.md",[],[],{"skillNo":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"publishedAt":18},"01M799ND2J4WNC2PGAZ0A06NDF","passnet-orchestrate","Round\u002Fbatch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep\u002Frevert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-orchestrate\u002FSKILL.md",[],[],{"skillNo":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"publishedAt":18},"01M799N8SBTMY02ZDG2AZF5ZWE","passnet-feedback","Fast, deterministic feedback for PassNet work: pre-flight pattern\u002Fmatch verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-feedback\u002FSKILL.md",[],[]]