[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M794KA37KX4B489AJKE798CP":3,"skill-same:01M794KA37KX4B489AJKE798CP":75},{"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},"01M794KA37KX4B489AJKE798CP","training-data-configs",null,"Plan Sana training data schemas, config overrides, launch commands, and safe validation for image, video, LoRA, Sprint, WM, Streaming, and Sol-RL workflows.","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\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training-data-configs\ndescription: \"Plan Sana training data schemas, config overrides, launch\n  commands, and safe validation for image, video, LoRA, Sprint, WM, Streaming,\n  and Sol-RL workflows.\"\nmetadata:\n  disco-role: operating\ndisable-model-invocation: true\nlicense: Apache 2.0\n---\n\n# Sana training data and config planner\n\nUse this sub-skill when a user asks how to prepare data, choose a training recipe, plan config overrides, or draft a safe launch command for Sana training or post-training. It covers image-text training, multi-scale WIDS\u002FWebDataset data, DDP\u002FFSDP image training, Sprint sCM\u002FLADD, DreamBooth LoRA, SANA-Video, LongSANA, SANA-WM Stage-1 and distillation, SANA-Streaming V2V, Sol-RL, and the boundary to Cosmos-RL.\n\nDo not use this sub-skill for inference-only requests. Route image inference to the image-generation sub-skill, video\u002Fworld\u002Fstreaming inference to the video-world-streaming sub-skill, and metrics, conversion, or deployment requests to the evaluation-conversion-deployment sub-skill.\n\n## Fast operating path\n\n1. Classify the request into one training family:\n   - image: SanaImgDataset image-text pairs or SanaWebDatasetMS WIDS\u002FWebDataset shards.\n   - sprint: one-step Sana-Sprint sCM plus LADD training.\n   - lora: DreamBooth LoRA over a diffusers Sana base model.\n   - video: SANA-Video 480p WanVAE or 720p LTX2 VAE training.\n   - longsana: LongSANA ODE, self-forcing, or long-video stage.\n   - wm-stage1: SANA-WM bidirectional or chunk-causal Stage-1 teacher.\n   - wm-distill: SANA-WM ODE, T43 self-forcing, or T121 self-forcing\u002FDMD distillation.\n   - streaming-v2v: SANA-Streaming bidirectional short V2V or long V2V fine-tuning.\n   - sol-rl: in-repo reinforcement-learning post-training launchers.\n   - cosmos-rl: external Cosmos-RL SFT\u002FRL integration boundary.\n2. Identify the data schema and verify it before planning training. Use the bundled validator when the user gives a local data path:\n\n```bash\npython skills\u002Fdisco\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fvalidate_dataset_layout.py \\\n  --path data\u002Fmy_sana_data \\\n  --mode auto \\\n  --max-samples 20\n```\n\n3. Generate a safe command plan instead of launching training. Use the bundled planner for a first draft:\n\n```bash\npython skills\u002Fdisco\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fplan_training_command.py \\\n  --family image \\\n  --data-dir data\u002Fmy_sana_data \\\n  --data-type SanaImgDataset \\\n  --work-dir output\u002Fmy_sana_run \\\n  --gpus 2 \\\n  --batch-size 8 \\\n  --num-workers 4\n```\n\n4. Explain what the plan can and cannot verify locally. Dataset shape and manifest checks are safe; actual model loading, HF downloads, CUDA kernels, FSDP, CP, VAE encoding, reward services, and quality convergence are not verified by these helpers.\n5. When a command depends on external assets, call out the asset explicitly: HF model or dataset repo id, local checkpoint, reward checkpoint, VAE cache, WIDS metadata, wandb login, or distributed rendezvous settings.\n\n## Bundled references\n\n- `references\u002Fdata-and-configs.md`: dataset layouts, WIDS\u002FWebDataset conversion planning, config override syntax, and preflight checks.\n- `references\u002Ftraining-workflows.md`: command templates for image, FSDP, Sprint, video, LongSANA, SANA-WM, SANA-Streaming, checkpoints, and native test candidates.\n- `references\u002Flora-and-post-training.md`: DreamBooth LoRA, Sol-RL, Cosmos-RL boundaries, rewards, and logging.\n- `references\u002Ftroubleshooting.md`: data, cache, VAE\u002Ftext-feature, torchrun, FSDP, OOM, HF, wandb, resume, and license failure modes.\n\n## Source evidence labels\n\nThis sub-skill was distilled from Sana documentation, config classes, dataset loaders, train launchers, and training smoke tests. Evidence labels include `docs\u002Fsana.md`, `docs\u002Fsana_sprint.md`, `docs\u002Fsana_lora_dreambooth.md`, `docs\u002Fsana_video.md`, `docs\u002Flongsana.md`, `docs\u002Fsana_wm.md`, `docs\u002Fsana_streaming.md`, `docs\u002Fsol_rl.md`, `docs\u002Fsana_cosmos_rl.md`, `diffusion\u002Futils\u002Fconfig.py`, `diffusion\u002Fdata\u002F`, `train_scripts\u002F`, `train_video_scripts\u002F`, `configs\u002F`, and `tests\u002Fbash\u002Ftraining\u002F`.\n","# Sana training data and config planner\n\nUse this sub-skill when a user asks how to prepare data, choose a training recipe, plan config overrides, or draft a safe launch command for Sana training or post-training. It covers image-text training, multi-scale WIDS\u002FWebDataset data, DDP\u002FFSDP image training, Sprint sCM\u002FLADD, DreamBooth LoRA, SANA-Video, LongSANA, SANA-WM Stage-1 and distillation, SANA-Streaming V2V, Sol-RL, and the boundary to Cosmos-RL.\n\nDo not use this sub-skill for inference-only requests. Route image inference to the image-generation sub-skill, video\u002Fworld\u002Fstreaming inference to the video-world-streaming sub-skill, and metrics, conversion, or deployment requests to the evaluation-conversion-deployment sub-skill.\n\n## Fast operating path\n\n1. Classify the request into one training family:\n   - image: SanaImgDataset image-text pairs or SanaWebDatasetMS WIDS\u002FWebDataset shards.\n   - sprint: one-step Sana-Sprint sCM plus LADD training.\n   - lora: DreamBooth LoRA over a diffusers Sana base model.\n   - video: SANA-Video 480p WanVAE or 720p LTX2 VAE training.\n   - longsana: LongSANA ODE, self-forcing, or long-video stage.\n   - wm-stage1: SANA-WM bidirectional or chunk-causal Stage-1 teacher.\n   - wm-distill: SANA-WM ODE, T43 self-forcing, or T121 self-forcing\u002FDMD distillation.\n   - streaming-v2v: SANA-Streaming bidirectional short V2V or long V2V fine-tuning.\n   - sol-rl: in-repo reinforcement-learning post-training launchers.\n   - cosmos-rl: external Cosmos-RL SFT\u002FRL integration boundary.\n2. Identify the data schema and verify it before planning training. Use the bundled validator when the user gives a local data path:\n\n```bash\npython skills\u002Fdisco\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fvalidate_dataset_layout.py \\\n  --path data\u002Fmy_sana_data \\\n  --mode auto \\\n  --max-samples 20\n```\n\n3. Generate a safe command plan instead of launching training. Use the bundled planner for a first draft:\n\n```bash\npython skills\u002Fdisco\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fplan_training_command.py \\\n  --family image \\\n  --data-dir data\u002Fmy_sana_data \\\n  --data-type SanaImgDataset \\\n  --work-dir output\u002Fmy_sana_run \\\n  --gpus 2 \\\n  --batch-size 8 \\\n  --num-workers 4\n```\n\n4. Explain what the plan can and cannot verify locally. Dataset shape and manifest checks are safe; actual model loading, HF downloads, CUDA kernels, FSDP, CP, VAE encoding, reward services, and quality convergence are not verified by these helpers.\n5. When a command depends on external assets, call out the asset explicitly: HF model or dataset repo id, local checkpoint, reward checkpoint, VAE cache, WIDS metadata, wandb login, or distributed rendezvous settings.\n\n## Bundled references\n\n- `references\u002Fdata-and-configs.md`: dataset layouts, WIDS\u002FWebDataset conversion planning, config override syntax, and preflight checks.\n- `references\u002Ftraining-workflows.md`: command templates for image, FSDP, Sprint, video, LongSANA, SANA-WM, SANA-Streaming, checkpoints, and native test candidates.\n- `references\u002Flora-and-post-training.md`: DreamBooth LoRA, Sol-RL, Cosmos-RL boundaries, rewards, and logging.\n- `references\u002Ftroubleshooting.md`: data, cache, VAE\u002Ftext-feature, torchrun, FSDP, OOM, HF, wandb, resume, and license failure modes.\n\n## Source evidence labels\n\nThis sub-skill was distilled from Sana documentation, config classes, dataset loaders, train launchers, and training smoke tests. Evidence labels include `docs\u002Fsana.md`, `docs\u002Fsana_sprint.md`, `docs\u002Fsana_lora_dreambooth.md`, `docs\u002Fsana_video.md`, `docs\u002Flongsana.md`, `docs\u002Fsana_wm.md`, `docs\u002Fsana_streaming.md`, `docs\u002Fsol_rl.md`, `docs\u002Fsana_cosmos_rl.md`, `diffusion\u002Futils\u002Fconfig.py`, `diffusion\u002Fdata\u002F`, `train_scripts\u002F`, `train_video_scripts\u002F`, `configs\u002F`, and `tests\u002Fbash\u002Ftraining\u002F`.\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\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs",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-data-configs",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training-data-configs",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training-data-configs","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002FSKILL.md",[47,53,57,61,65,71],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fdata-and-configs.md","REFERENCE","参考文档",12119,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Freferences\u002Fdata-and-configs.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Flora-and-post-training.md",9139,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Freferences\u002Flora-and-post-training.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Ftraining-workflows.md",12779,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Freferences\u002Ftraining-workflows.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Ftroubleshooting.md",9641,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Freferences\u002Ftroubleshooting.md",{"path":66,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fplan_training_command.py","SCRIPT","脚本",21497,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fplan_training_command.py",{"path":72,"resourceType":67,"resourceTypeDesc":68,"sizeBytes":73,"sourceHtmlUrl":74},"scripts\u002Fvalidate_dataset_layout.py",19395,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsana\u002Fsub-skills\u002Ftraining-data-configs\u002Fscripts\u002Fvalidate_dataset_layout.py",[76,83,90,97,104,111],{"skillNo":77,"name":78,"nameZh":6,"description":79,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"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":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"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":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"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":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":103,"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":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"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":112,"name":113,"nameZh":6,"description":114,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":115,"categories":116,"keywords":117,"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",[],[]]