[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78SSQ2G3PQ4B8F2PC5WABYM":3,"skill-same:01M78SSQ2G3PQ4B8F2PC5WABYM":66},{"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":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":32,"install":33,"resources":45},"01M78SSQ2G3PQ4B8F2PC5WABYM","training-and-evaluation",null,"Train, evaluate, and export HumanoidVerse policies with PPO, PPODeltaA, Hydra overrides, simulator selection, motion tracking, locomotion, and delta-action workflows.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"MIT","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training-and-evaluation\ndescription: \"Train, evaluate, and export HumanoidVerse policies with PPO,\n  PPODeltaA, Hydra overrides, simulator selection, motion tracking, locomotion,\n  and delta-action workflows.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# training-and-evaluation\n\nUse this sub-skill when a user asks for HumanoidVerse policy training, checkpoint evaluation, ONNX export, simulator\u002Fbackend choice, Hydra override selection, phase-based motion tracking, locomotion, open-loop delta-action training, or closed-loop delta-action finetuning.\n\nRun ASAP training\u002Fevaluation entry points from the repository checkout root. The bundled helper scripts are skill-root-relative and accept `--repo-root \u003Casap-checkout>`. The repo entry points are scripts, not package modules: `python humanoidverse\u002Ftrain_agent.py ...` and `python humanoidverse\u002Feval_agent.py ...`. Do not use `python -m humanoidverse.train_agent`, because the source imports `utils.config_utils` relative to the `humanoidverse\u002F` script directory.\n\n## Do not use this sub-skill for\n\n- SMPL\u002FAMASS shape fitting, motion retargeting, robot XML fitting, or MuJoCo retargeting visualization. Use [`..\u002Fmotion-retargeting\u002FSKILL.md`](..\u002Fmotion-retargeting\u002FSKILL.md).\n- Sim2sim or sim2real runtime control, ROS2\u002FUnitree bridges, joystick control, and live robot deployment. Use [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md).\n- Generic PPO theory or unrelated simulator stacks that are not launched through this repository's HumanoidVerse Hydra configs.\n\n## Required reading order\n\n1. Read the root router: [`..\u002F..\u002FSKILL.md`](..\u002F..\u002FSKILL.md).\n2. Read root install\u002Fbackend guidance before choosing a simulator: [`..\u002F..\u002Freferences\u002Finstall-and-backends.md`](..\u002F..\u002Freferences\u002Finstall-and-backends.md).\n3. Read this sub-skill's command recipes: [`references\u002Fworkflows.md`](references\u002Fworkflows.md).\n4. Read CLI\u002Fconfig\u002Fclass behavior before changing Hydra groups or checkpoint\u002Fexport assumptions: [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md).\n5. If anything fails, read this sub-skill's predictable failures first: [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md), then root troubleshooting [`..\u002F..\u002Freferences\u002Ftroubleshooting.md`](..\u002F..\u002Freferences\u002Ftroubleshooting.md).\n\n## Safe preflight checks\n\nThese checks do not start simulator training:\n\n```bash\n# Confirm that Hydra can list config groups.\npython humanoidverse\u002Ftrain_agent.py --help\npython humanoidverse\u002Feval_agent.py --help\n\n# Build a command without executing it.\npython sub-skills\u002Ftraining-and-evaluation\u002Fscripts\u002Fbuild_training_command.py \\\n  --repo-root \u003Casap-checkout> \\\n  --workflow motion-tracking \\\n  --motion-file humanoidverse\u002Fdata\u002Fmotions\u002Fg1_29dof_anneal_23dof\u002FTairanTestbed\u002Fsingles\u002F0-TairanTestbed_TairanTestbed_CR7_video_CR7_level1_filter_amass.pkl\n\n# Turn a training command into a Hydra composition-only smoke check.\npython sub-skills\u002Ftraining-and-evaluation\u002Fscripts\u002Fbuild_training_command.py \\\n  --repo-root \u003Casap-checkout> \\\n  --workflow motion-tracking \\\n  --motion-file humanoidverse\u002Fdata\u002Fmotions\u002Fg1_29dof_anneal_23dof\u002FTairanTestbed\u002Fsingles\u002F0-TairanTestbed_TairanTestbed_CR7_video_CR7_level1_filter_amass.pkl \\\n  --cfg-job\n```\n\nRun the root doctor script before runtime commands to get a single import\u002Fhelp report:\n\n```bash\npython scripts\u002Fasap_doctor.py --repo-root \u003Casap-checkout> --section core\n```\n\n## Route by prompt\n\n| User asks for | Use | Primary recipe |\n| --- | --- | --- |\n| \"train motion tracking\", \"imitate this motion\", \"CR7 policy\" | `train_agent.py` with `+exp=motion_tracking` | [`references\u002Fworkflows.md#motion-tracking-training`](references\u002Fworkflows.md#motion-tracking-training) |\n| \"train locomotion\" or \"test IsaacGym install with locomotion\" | `train_agent.py` with `+exp=locomotion` | [`references\u002Fworkflows.md#locomotion-training-and-smoke-test`](references\u002Fworkflows.md#locomotion-training-and-smoke-test) |\n| \"train delta action model\" | `train_agent.py` with `+exp=train_delta_a_open_loop` | [`references\u002Fworkflows.md#open-loop-delta-action-training`](references\u002Fworkflows.md#open-loop-delta-action-training) |\n| \"finetune with delta A\" or \"closed-loop delta action\" | `train_agent.py` with `+exp=train_delta_a_closed_loop` and `algo.config.policy_checkpoint=...` | [`references\u002Fworkflows.md#closed-loop-delta-action-finetuning`](references\u002Fworkflows.md#closed-loop-delta-action-finetuning) |\n| \"evaluate checkpoint\", \"visualize policy\" | `eval_agent.py +checkpoint=...` | [`references\u002Fworkflows.md#checkpoint-evaluation`](references\u002Fworkflows.md#checkpoint-evaluation) |\n| \"export ONNX\" | `eval_agent.py +checkpoint=...`; export happens before the eval loop | [`references\u002Fworkflows.md#onnx-export`](references\u002Fworkflows.md#onnx-export) |\n| \"which Hydra overrides?\" | Choose config groups and scalar overrides from the workflow table | [`references\u002Fapi-reference.md#hydra-config-groups`](references\u002Fapi-reference.md#hydra-config-groups) |\n\n## Backend selection rules\n\n- Default to `+simulator=isaacgym` for README-aligned training recipes when IsaacGym Preview 4 is installed and CUDA is available.\n- Use `+simulator=isaacsim` only in an IsaacLab\u002FIsaacSim environment with `omni.isaac.lab`; the entry points launch an IsaacSim app before importing Torch.\n- Use `+simulator=genesis` only after installing `genesis-world` in the same environment.\n- `+simulator=mujoco` is present as a Hydra config, but deployment\u002Fcontrol recipes live in [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md); do not infer that every training recipe has been validated on MuJoCo.\n- The verified inspection environment had CUDA-capable Torch, Hydra, `humanoidverse`, `asap` distribution metadata, `onnx`, and `onnxruntime`, but did not have IsaacGym, Genesis, or IsaacLab installed. Treat actual simulator launch as backend-gated.\n\n## Common outputs\n\n- Training experiment directory: `logs\u002F\u003Cproject_name>\u002F\u003Ctimestamp>-\u003Cexperiment_name>-\u003Clog_task_name>-\u003Crobot_type>\u002F`.\n- Training checkpoints: `model_\u003Citeration>.pt` in the experiment directory; `PPO.save()` writes actor, critic, optimizer states, iteration, and infos.\n- Saved training config: `config.yaml` in the experiment directory; Hydra's own run directory is `\u003Cexperiment_dir>\u002F.hydra\u002F` and `train.log` is written there.\n- Training renderings: `\u003Cexperiment_dir>\u002Frenderings_training\u002F` when the simulator records them.\n- Evaluation logs: `logs_eval\u002F\u003Ceval_name>\u002F\u003Ceval_timestamp>\u002Fconfig.yaml` and `eval.log` under the Hydra eval run directory.\n- Evaluation renderings: `\u003Ccheckpoint_dir>\u002Frenderings\u002Fckpt_\u003CN>\u002F`, where `\u003CN>` comes from `model_\u003CN>.pt`.\n- ONNX export: `\u003Ccheckpoint_dir>\u002Fexported\u002Fmodel_\u003CN>.onnx`; `eval_agent.py` exports before it enters the infinite evaluation loop.\n- Evaluation motion recording with `+opt=record`: `\u003Ccheckpoint_dir>\u002Fmotions\u002F\u003Csave_note>_\u003Ceval_timestamp>.pkl` or `\u003Ceval_timestamp>_\u003Cdump_motion_name>.pkl`; those dumps include an `action` key useful for delta-action datasets.\n\n## Cross-links\n\n- Root router: [`..\u002F..\u002FSKILL.md`](..\u002F..\u002FSKILL.md).\n- Root install and backend notes: [`..\u002F..\u002Freferences\u002Finstall-and-backends.md`](..\u002F..\u002Freferences\u002Finstall-and-backends.md).\n- Root troubleshooting: [`..\u002F..\u002Freferences\u002Ftroubleshooting.md`](..\u002F..\u002Freferences\u002Ftroubleshooting.md).\n- Motion retargeting: [`..\u002Fmotion-retargeting\u002FSKILL.md`](..\u002Fmotion-retargeting\u002FSKILL.md) for producing motion `.pkl` files consumed by motion-tracking training.\n- Sim2real deployment: [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md) for using exported ONNX policies in MuJoCo\u002FROS2\u002FUnitree runtime flows.\n","# training-and-evaluation\n\nUse this sub-skill when a user asks for HumanoidVerse policy training, checkpoint evaluation, ONNX export, simulator\u002Fbackend choice, Hydra override selection, phase-based motion tracking, locomotion, open-loop delta-action training, or closed-loop delta-action finetuning.\n\nRun ASAP training\u002Fevaluation entry points from the repository checkout root. The bundled helper scripts are skill-root-relative and accept `--repo-root \u003Casap-checkout>`. The repo entry points are scripts, not package modules: `python humanoidverse\u002Ftrain_agent.py ...` and `python humanoidverse\u002Feval_agent.py ...`. Do not use `python -m humanoidverse.train_agent`, because the source imports `utils.config_utils` relative to the `humanoidverse\u002F` script directory.\n\n## Do not use this sub-skill for\n\n- SMPL\u002FAMASS shape fitting, motion retargeting, robot XML fitting, or MuJoCo retargeting visualization. Use [`..\u002Fmotion-retargeting\u002FSKILL.md`](..\u002Fmotion-retargeting\u002FSKILL.md).\n- Sim2sim or sim2real runtime control, ROS2\u002FUnitree bridges, joystick control, and live robot deployment. Use [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md).\n- Generic PPO theory or unrelated simulator stacks that are not launched through this repository's HumanoidVerse Hydra configs.\n\n## Required reading order\n\n1. Read the root router: [`..\u002F..\u002FSKILL.md`](..\u002F..\u002FSKILL.md).\n2. Read root install\u002Fbackend guidance before choosing a simulator: [`..\u002F..\u002Freferences\u002Finstall-and-backends.md`](..\u002F..\u002Freferences\u002Finstall-and-backends.md).\n3. Read this sub-skill's command recipes: [`references\u002Fworkflows.md`](references\u002Fworkflows.md).\n4. Read CLI\u002Fconfig\u002Fclass behavior before changing Hydra groups or checkpoint\u002Fexport assumptions: [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md).\n5. If anything fails, read this sub-skill's predictable failures first: [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md), then root troubleshooting [`..\u002F..\u002Freferences\u002Ftroubleshooting.md`](..\u002F..\u002Freferences\u002Ftroubleshooting.md).\n\n## Safe preflight checks\n\nThese checks do not start simulator training:\n\n```bash\n# Confirm that Hydra can list config groups.\npython humanoidverse\u002Ftrain_agent.py --help\npython humanoidverse\u002Feval_agent.py --help\n\n# Build a command without executing it.\npython sub-skills\u002Ftraining-and-evaluation\u002Fscripts\u002Fbuild_training_command.py \\\n  --repo-root \u003Casap-checkout> \\\n  --workflow motion-tracking \\\n  --motion-file humanoidverse\u002Fdata\u002Fmotions\u002Fg1_29dof_anneal_23dof\u002FTairanTestbed\u002Fsingles\u002F0-TairanTestbed_TairanTestbed_CR7_video_CR7_level1_filter_amass.pkl\n\n# Turn a training command into a Hydra composition-only smoke check.\npython sub-skills\u002Ftraining-and-evaluation\u002Fscripts\u002Fbuild_training_command.py \\\n  --repo-root \u003Casap-checkout> \\\n  --workflow motion-tracking \\\n  --motion-file humanoidverse\u002Fdata\u002Fmotions\u002Fg1_29dof_anneal_23dof\u002FTairanTestbed\u002Fsingles\u002F0-TairanTestbed_TairanTestbed_CR7_video_CR7_level1_filter_amass.pkl \\\n  --cfg-job\n```\n\nRun the root doctor script before runtime commands to get a single import\u002Fhelp report:\n\n```bash\npython scripts\u002Fasap_doctor.py --repo-root \u003Casap-checkout> --section core\n```\n\n## Route by prompt\n\n| User asks for | Use | Primary recipe |\n| --- | --- | --- |\n| \"train motion tracking\", \"imitate this motion\", \"CR7 policy\" | `train_agent.py` with `+exp=motion_tracking` | [`references\u002Fworkflows.md#motion-tracking-training`](references\u002Fworkflows.md#motion-tracking-training) |\n| \"train locomotion\" or \"test IsaacGym install with locomotion\" | `train_agent.py` with `+exp=locomotion` | [`references\u002Fworkflows.md#locomotion-training-and-smoke-test`](references\u002Fworkflows.md#locomotion-training-and-smoke-test) |\n| \"train delta action model\" | `train_agent.py` with `+exp=train_delta_a_open_loop` | [`references\u002Fworkflows.md#open-loop-delta-action-training`](references\u002Fworkflows.md#open-loop-delta-action-training) |\n| \"finetune with delta A\" or \"closed-loop delta action\" | `train_agent.py` with `+exp=train_delta_a_closed_loop` and `algo.config.policy_checkpoint=...` | [`references\u002Fworkflows.md#closed-loop-delta-action-finetuning`](references\u002Fworkflows.md#closed-loop-delta-action-finetuning) |\n| \"evaluate checkpoint\", \"visualize policy\" | `eval_agent.py +checkpoint=...` | [`references\u002Fworkflows.md#checkpoint-evaluation`](references\u002Fworkflows.md#checkpoint-evaluation) |\n| \"export ONNX\" | `eval_agent.py +checkpoint=...`; export happens before the eval loop | [`references\u002Fworkflows.md#onnx-export`](references\u002Fworkflows.md#onnx-export) |\n| \"which Hydra overrides?\" | Choose config groups and scalar overrides from the workflow table | [`references\u002Fapi-reference.md#hydra-config-groups`](references\u002Fapi-reference.md#hydra-config-groups) |\n\n## Backend selection rules\n\n- Default to `+simulator=isaacgym` for README-aligned training recipes when IsaacGym Preview 4 is installed and CUDA is available.\n- Use `+simulator=isaacsim` only in an IsaacLab\u002FIsaacSim environment with `omni.isaac.lab`; the entry points launch an IsaacSim app before importing Torch.\n- Use `+simulator=genesis` only after installing `genesis-world` in the same environment.\n- `+simulator=mujoco` is present as a Hydra config, but deployment\u002Fcontrol recipes live in [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md); do not infer that every training recipe has been validated on MuJoCo.\n- The verified inspection environment had CUDA-capable Torch, Hydra, `humanoidverse`, `asap` distribution metadata, `onnx`, and `onnxruntime`, but did not have IsaacGym, Genesis, or IsaacLab installed. Treat actual simulator launch as backend-gated.\n\n## Common outputs\n\n- Training experiment directory: `logs\u002F\u003Cproject_name>\u002F\u003Ctimestamp>-\u003Cexperiment_name>-\u003Clog_task_name>-\u003Crobot_type>\u002F`.\n- Training checkpoints: `model_\u003Citeration>.pt` in the experiment directory; `PPO.save()` writes actor, critic, optimizer states, iteration, and infos.\n- Saved training config: `config.yaml` in the experiment directory; Hydra's own run directory is `\u003Cexperiment_dir>\u002F.hydra\u002F` and `train.log` is written there.\n- Training renderings: `\u003Cexperiment_dir>\u002Frenderings_training\u002F` when the simulator records them.\n- Evaluation logs: `logs_eval\u002F\u003Ceval_name>\u002F\u003Ceval_timestamp>\u002Fconfig.yaml` and `eval.log` under the Hydra eval run directory.\n- Evaluation renderings: `\u003Ccheckpoint_dir>\u002Frenderings\u002Fckpt_\u003CN>\u002F`, where `\u003CN>` comes from `model_\u003CN>.pt`.\n- ONNX export: `\u003Ccheckpoint_dir>\u002Fexported\u002Fmodel_\u003CN>.onnx`; `eval_agent.py` exports before it enters the infinite evaluation loop.\n- Evaluation motion recording with `+opt=record`: `\u003Ccheckpoint_dir>\u002Fmotions\u002F\u003Csave_note>_\u003Ceval_timestamp>.pkl` or `\u003Ceval_timestamp>_\u003Cdump_motion_name>.pkl`; those dumps include an `action` key useful for delta-action datasets.\n\n## Cross-links\n\n- Root router: [`..\u002F..\u002FSKILL.md`](..\u002F..\u002FSKILL.md).\n- Root install and backend notes: [`..\u002F..\u002Freferences\u002Finstall-and-backends.md`](..\u002F..\u002Freferences\u002Finstall-and-backends.md).\n- Root troubleshooting: [`..\u002F..\u002Freferences\u002Ftroubleshooting.md`](..\u002F..\u002Freferences\u002Ftroubleshooting.md).\n- Motion retargeting: [`..\u002Fmotion-retargeting\u002FSKILL.md`](..\u002Fmotion-retargeting\u002FSKILL.md) for producing motion `.pkl` files consumed by motion-tracking training.\n- Sim2real deployment: [`..\u002Fsim2real-deployment\u002FSKILL.md`](..\u002Fsim2real-deployment\u002FSKILL.md) for using exported ONNX policies in MuJoCo\u002FROS2\u002FUnitree runtime flows.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":25},[35,38,41],{"tool":36,"command":37},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill training-and-evaluation",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training-and-evaluation",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training-and-evaluation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002FSKILL.md",[46,52,56,60],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fapi-reference.md","REFERENCE","参考文档",15039,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002Freferences\u002Fapi-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Ftroubleshooting.md",7335,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002Freferences\u002Ftroubleshooting.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fworkflows.md",16408,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002Freferences\u002Fworkflows.md",{"path":61,"resourceType":62,"resourceTypeDesc":63,"sizeBytes":64,"sourceHtmlUrl":65},"scripts\u002Fbuild_training_command.py","SCRIPT","脚本",17842,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fasap\u002Fsub-skills\u002Ftraining-and-evaluation\u002Fscripts\u002Fbuild_training_command.py",[67,75,82,89,96,103],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"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.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"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":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"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":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"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":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"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":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":71,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"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",[],[]]