[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M753JSVEPN7AW94MXN0EE3DQ":3,"skill-same:01M753JSVEPN7AW94MXN0EE3DQ":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},"01M753JSVEPN7AW94MXN0EE3DQ","training-loop-integration",null,"Migrate raw PyTorch training and evaluation loops to Hugging Face Accelerate using Accelerator, prepare(), backward(), gradient accumulation, dataloader behavior, gather\u002Freduce, mixed precision, DDP kwargs, local SGD, communication hooks, and basic distributed logging.","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\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: training-loop-integration\ndescription: \"Migrate raw PyTorch training and evaluation loops to Hugging Face\n  Accelerate using Accelerator, prepare(), backward(), gradient accumulation,\n  dataloader behavior, gather\u002Freduce, mixed precision, DDP kwargs, local SGD,\n  communication hooks, and basic distributed logging.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Training Loop Integration\n\nUse this sub-skill when adapting an existing PyTorch model, optimizer, dataloader, scheduler, or evaluation loop to Accelerate while keeping the training code mostly framework-neutral.\n\n## Route first\n\n- For `accelerate config`, `accelerate launch`, launch flags, config files, or process-count setup, use `..\u002Fconfiguration-and-cli\u002F`.\n- For DeepSpeed, FSDP, TPU, Megatron-LM, FP8 backend recipes, or backend-specific plugin tuning, use `..\u002Fdistributed-training-backends\u002F`.\n- For `save_state`, `load_state`, model checkpoint saving, trackers, profiler, or experiment dashboards, use `..\u002Fcheckpointing-and-tracking\u002F`.\n- For large-model loading, device maps, offload, or inference dispatch, use the root skill routing to the big-model sub-skill.\n\n## Start here\n\n1. Read `references\u002Fworkflows.md` for migration recipes and loop patterns.\n2. Read `references\u002Fapi-reference.md` for constructor options, wrappers, dataloader semantics, gather\u002Freduce, kwargs handlers, logging, and scheduler behavior.\n3. Read `references\u002Ftroubleshooting.md` when a distributed loop hangs, metrics are wrong, gradients do not sync, or device placement behaves unexpectedly.\n4. Run `scripts\u002Faccelerator_loop_smoke.py` to verify a minimal CPU training loop in the current environment.\n\n## Core migration checklist\n\n- Create one `Accelerator` early, before distributed-sensitive objects need its state.\n- Remove hard-coded `.cuda()` and direct `.to(\"cuda\")`; prefer automatic device placement through `accelerator.prepare()`.\n- Pass related PyTorch objects to `accelerator.prepare(...)` and unpack them in the same order.\n- Replace `loss.backward()` with `accelerator.backward(loss)`.\n- For accumulation, create `Accelerator(gradient_accumulation_steps=N)` or a `GradientAccumulationPlugin`, then wrap each minibatch body in `with accelerator.accumulate(model):`.\n- Use `accelerator.gather_for_metrics(...)` for evaluation metrics and `accelerator.pad_across_processes(...)` before gathering variable-size tensors.\n- Use `accelerator.print(...)` or `accelerate.logging.get_logger(...)` instead of unsynchronized process-wide `print()` spam.\n\n## Smoke check\n\n```bash\npython skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Fscripts\u002Faccelerator_loop_smoke.py\npython skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Fscripts\u002Faccelerator_loop_smoke.py --help\n```\n\nThe smoke script is intentionally CPU-safe and tiny. It validates importability, `Accelerator.prepare`, `Accelerator.accumulate`, `Accelerator.backward`, scheduler wrapping, and metric gathering without requiring a distributed launcher.\n","# Training Loop Integration\n\nUse this sub-skill when adapting an existing PyTorch model, optimizer, dataloader, scheduler, or evaluation loop to Accelerate while keeping the training code mostly framework-neutral.\n\n## Route first\n\n- For `accelerate config`, `accelerate launch`, launch flags, config files, or process-count setup, use `..\u002Fconfiguration-and-cli\u002F`.\n- For DeepSpeed, FSDP, TPU, Megatron-LM, FP8 backend recipes, or backend-specific plugin tuning, use `..\u002Fdistributed-training-backends\u002F`.\n- For `save_state`, `load_state`, model checkpoint saving, trackers, profiler, or experiment dashboards, use `..\u002Fcheckpointing-and-tracking\u002F`.\n- For large-model loading, device maps, offload, or inference dispatch, use the root skill routing to the big-model sub-skill.\n\n## Start here\n\n1. Read `references\u002Fworkflows.md` for migration recipes and loop patterns.\n2. Read `references\u002Fapi-reference.md` for constructor options, wrappers, dataloader semantics, gather\u002Freduce, kwargs handlers, logging, and scheduler behavior.\n3. Read `references\u002Ftroubleshooting.md` when a distributed loop hangs, metrics are wrong, gradients do not sync, or device placement behaves unexpectedly.\n4. Run `scripts\u002Faccelerator_loop_smoke.py` to verify a minimal CPU training loop in the current environment.\n\n## Core migration checklist\n\n- Create one `Accelerator` early, before distributed-sensitive objects need its state.\n- Remove hard-coded `.cuda()` and direct `.to(\"cuda\")`; prefer automatic device placement through `accelerator.prepare()`.\n- Pass related PyTorch objects to `accelerator.prepare(...)` and unpack them in the same order.\n- Replace `loss.backward()` with `accelerator.backward(loss)`.\n- For accumulation, create `Accelerator(gradient_accumulation_steps=N)` or a `GradientAccumulationPlugin`, then wrap each minibatch body in `with accelerator.accumulate(model):`.\n- Use `accelerator.gather_for_metrics(...)` for evaluation metrics and `accelerator.pad_across_processes(...)` before gathering variable-size tensors.\n- Use `accelerator.print(...)` or `accelerate.logging.get_logger(...)` instead of unsynchronized process-wide `print()` spam.\n\n## Smoke check\n\n```bash\npython skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Fscripts\u002Faccelerator_loop_smoke.py\npython skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Fscripts\u002Faccelerator_loop_smoke.py --help\n```\n\nThe smoke script is intentionally CPU-safe and tiny. It validates importability, `Accelerator.prepare`, `Accelerator.accumulate`, `Accelerator.backward`, scheduler wrapping, and metric gathering without requiring a distributed launcher.\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\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration",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-loop-integration",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training-loop-integration",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training-loop-integration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002FSKILL.md",[47,53,57,61],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fapi-reference.md","REFERENCE","参考文档",11839,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Ftroubleshooting.md",8632,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Freferences\u002Ftroubleshooting.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fworkflows.md",10931,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Freferences\u002Fworkflows.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Faccelerator_loop_smoke.py","SCRIPT","脚本",3820,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Faccelerate\u002Fsub-skills\u002Ftraining-loop-integration\u002Fscripts\u002Faccelerator_loop_smoke.py",[68,76,83,90,97,104],{"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":75},"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",[],[],"2026-11-15 14:19:17",{"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":75},"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":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":75},"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":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":75},"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":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":75},"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":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":75},"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",[],[]]