[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M790WXGZHSF54286TPHCEH61":3,"skill-same:01M790WXGZHSF54286TPHCEH61":67},{"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},"01M790WXGZHSF54286TPHCEH61","training",null,"Configure, validate, launch, and resume Motus three-stage training on a single node or SLURM cluster without confusing fine-tuning, checkpoint resume, or backbone initialization.","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\u002Fmotus\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training\ndescription: \"Configure, validate, launch, and resume Motus three-stage training\n  on a single node or SLURM cluster without confusing fine-tuning, checkpoint\n  resume, or backbone initialization.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Motus training\n\nUse this skill when a user needs to prepare a Motus training configuration, check\nit safely, construct a `torchrun`\u002FDeepSpeed command, or reason about resuming a\nrun. It covers the three-stage intent, not dataset conversion or inference.\n\n- Read [configuration.md](references\u002Fconfiguration.md) for the config contract,\n  derived values, and checkpoint-mode rules.\n- Read [workflows.md](references\u002Fworkflows.md) for single-node, SLURM, and\n  checkpoint export procedures. Launchers in the repository are reference-only:\n  they are long-running and may submit external cluster jobs.\n- Read [troubleshooting.md](references\u002Ftroubleshooting.md) when validation,\n  imports, CUDA, flash-attn, NCCL, logging, data, or checkpoint loading fails.\n- For dataset directory\u002Flayout preparation, use the sibling\n  [data-preparation skill](..\u002Fdata-preparation\u002FSKILL.md). For inference, use\n  [model-inference](..\u002Fmodel-inference\u002FSKILL.md).\n\n## Operating procedure\n\n1. **Identify intent before editing YAML.** Select `pretrain`, `finetune`, or\n   resume. Record whether this is Stage 1 VGM training, Stage 2 Motus\n   pretraining with latent actions, or Stage 3 target-robot SFT with actions.\n   A checkpoint resume is a fourth *execution mode*, not a new training stage.\n2. **Choose an example config.** Use `configs\u002Flatent_action.yaml` for the\n   explicit `training_mode: pretrain` latent-action pipeline; use the\n   embodiment config matching the dataset (`robotwin`, `ac_one`,\n   `aloha_agilex_2`, or `lerobot`) for fine-tuning. Replace all example paths\n   with paths valid on the target machine.\n3. **Run safe checks before any launch.** Parse YAML, check the required\n   sections and enum-like fields, calculate `action_chunk_size`, verify local\n   dataset\u002Fcheckpoint\u002FVAE\u002Fconfig paths where available, and check GPU count,\n   CUDA, bfloat16, DeepSpeed, and flash-attn. A parser\u002Fimport check is allowed\n   on CPU; model construction, VAE access, and training require CUDA and the\n   large WAN\u002FVLM checkpoints.\n4. **Make checkpoint intent mutually clear.** Never set a resume checkpoint and\n   a fine-tune checkpoint accidentally. Resume loads the complete Accelerator\n   state and continues its step number; fine-tune partially initializes a\n   `finetune` model and skips action input\u002Fdecoder weights; scratch loads WAN\n   and VLM backbones from `model.*.checkpoint_path`.\n5. **Review the rendered command.** Check node count, GPUs per node, master\n   address\u002Fport, config path, DeepSpeed JSON, logging mode, output\u002Fcheckpoint\n   directory, and the absence of accidental shell comments after `\\\\`.\n   Prefer a short, bounded smoke configuration before an expensive run.\n6. **Launch only with explicit approval.** `torchrun`, `sbatch`, `srun`, WAN\u002FVLM\n   downloads, and full training are external, long-running, or expensive\n   operations. This skill explains and constructs them but does not silently\n   execute them. Use the repository launchers only as templates; do not copy\n   their placeholder paths into a public skill or run them as validation.\n7. **Verify expected signals.** Rank 0 should report the selected dataset and\n   training mode, checkpoint directory, logging backends, model creation,\n   dataloader creation, and step\u002Floss lines. A saved checkpoint directory\n   should contain Accelerator\u002FDeepSpeed state and the filtered `config.json`.\n   Use [scripts\u002Fexport_config_json.py](scripts\u002Fexport_config_json.py) to create\n   that JSON deliberately for an existing checkpoint directory.\n\n## Hard boundary\n\nThis skill does not prescribe dataset layouts, conversion, or inference CLI\nflags. It does not promise CPU training: Motus uses CUDA autocast and a very\nlarge WAN\u002FVLM stack, and the documented training requirement is more than 80\nGB VRAM (typically eight 80-GB GPUs for the supplied examples). Keep missing\nmodel checkpoints, datasets, optional packages, cluster topology, and\nunverified hardware as explicit blockers rather than guessing.\n","# Motus training\n\nUse this skill when a user needs to prepare a Motus training configuration, check\nit safely, construct a `torchrun`\u002FDeepSpeed command, or reason about resuming a\nrun. It covers the three-stage intent, not dataset conversion or inference.\n\n- Read [configuration.md](references\u002Fconfiguration.md) for the config contract,\n  derived values, and checkpoint-mode rules.\n- Read [workflows.md](references\u002Fworkflows.md) for single-node, SLURM, and\n  checkpoint export procedures. Launchers in the repository are reference-only:\n  they are long-running and may submit external cluster jobs.\n- Read [troubleshooting.md](references\u002Ftroubleshooting.md) when validation,\n  imports, CUDA, flash-attn, NCCL, logging, data, or checkpoint loading fails.\n- For dataset directory\u002Flayout preparation, use the sibling\n  [data-preparation skill](..\u002Fdata-preparation\u002FSKILL.md). For inference, use\n  [model-inference](..\u002Fmodel-inference\u002FSKILL.md).\n\n## Operating procedure\n\n1. **Identify intent before editing YAML.** Select `pretrain`, `finetune`, or\n   resume. Record whether this is Stage 1 VGM training, Stage 2 Motus\n   pretraining with latent actions, or Stage 3 target-robot SFT with actions.\n   A checkpoint resume is a fourth *execution mode*, not a new training stage.\n2. **Choose an example config.** Use `configs\u002Flatent_action.yaml` for the\n   explicit `training_mode: pretrain` latent-action pipeline; use the\n   embodiment config matching the dataset (`robotwin`, `ac_one`,\n   `aloha_agilex_2`, or `lerobot`) for fine-tuning. Replace all example paths\n   with paths valid on the target machine.\n3. **Run safe checks before any launch.** Parse YAML, check the required\n   sections and enum-like fields, calculate `action_chunk_size`, verify local\n   dataset\u002Fcheckpoint\u002FVAE\u002Fconfig paths where available, and check GPU count,\n   CUDA, bfloat16, DeepSpeed, and flash-attn. A parser\u002Fimport check is allowed\n   on CPU; model construction, VAE access, and training require CUDA and the\n   large WAN\u002FVLM checkpoints.\n4. **Make checkpoint intent mutually clear.** Never set a resume checkpoint and\n   a fine-tune checkpoint accidentally. Resume loads the complete Accelerator\n   state and continues its step number; fine-tune partially initializes a\n   `finetune` model and skips action input\u002Fdecoder weights; scratch loads WAN\n   and VLM backbones from `model.*.checkpoint_path`.\n5. **Review the rendered command.** Check node count, GPUs per node, master\n   address\u002Fport, config path, DeepSpeed JSON, logging mode, output\u002Fcheckpoint\n   directory, and the absence of accidental shell comments after `\\\\`.\n   Prefer a short, bounded smoke configuration before an expensive run.\n6. **Launch only with explicit approval.** `torchrun`, `sbatch`, `srun`, WAN\u002FVLM\n   downloads, and full training are external, long-running, or expensive\n   operations. This skill explains and constructs them but does not silently\n   execute them. Use the repository launchers only as templates; do not copy\n   their placeholder paths into a public skill or run them as validation.\n7. **Verify expected signals.** Rank 0 should report the selected dataset and\n   training mode, checkpoint directory, logging backends, model creation,\n   dataloader creation, and step\u002Floss lines. A saved checkpoint directory\n   should contain Accelerator\u002FDeepSpeed state and the filtered `config.json`.\n   Use [scripts\u002Fexport_config_json.py](scripts\u002Fexport_config_json.py) to create\n   that JSON deliberately for an existing checkpoint directory.\n\n## Hard boundary\n\nThis skill does not prescribe dataset layouts, conversion, or inference CLI\nflags. It does not promise CPU training: Motus uses CUDA autocast and a very\nlarge WAN\u002FVLM stack, and the documented training requirement is more than 80\nGB VRAM (typically eight 80-GB GPUs for the supplied examples). Keep missing\nmodel checkpoints, datasets, optional packages, cluster topology, and\nunverified hardware as explicit blockers rather than guessing.\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\u002Fmotus\u002Fsub-skills\u002Ftraining",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 training",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmotus\u002Fsub-skills\u002Ftraining\u002FSKILL.md",[47,53,57,61],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fconfiguration.md","REFERENCE","参考文档",11825,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmotus\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fconfiguration.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Ftroubleshooting.md",12002,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmotus\u002Fsub-skills\u002Ftraining\u002Freferences\u002Ftroubleshooting.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fworkflows.md",11311,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmotus\u002Fsub-skills\u002Ftraining\u002Freferences\u002Fworkflows.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Fexport_config_json.py","SCRIPT","脚本",3752,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmotus\u002Fsub-skills\u002Ftraining\u002Fscripts\u002Fexport_config_json.py",[68,75,82,89,96,103],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"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.","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":14,"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":14,"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":14,"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":14,"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":14,"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",[],[]]