[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78ZJY1526R37BC414QCSJ67":3,"skill-same:01M78ZJY1526R37BC414QCSJ67":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},"01M78ZJY1526R37BC414QCSJ67","nlu-glue-adaptation",null,"Construct and troubleshoot LoRA fine-tuning and evaluation workflows for RoBERTa or DeBERTa-v2 GLUE-style sequence classification.","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\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: nlu-glue-adaptation\ndescription: \"Construct and troubleshoot LoRA fine-tuning and evaluation\n  workflows for RoBERTa or DeBERTa-v2 GLUE-style sequence classification.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# NLU GLUE adaptation\n\nUse this sub-skill for the repository's LoRA-aware sequence-classification\nworkflow: model configuration, query\u002Fvalue projection replacement, GLUE task\ncommands, adapter checkpoint transfer, and training-resource checks.\n\n## Route here\n\n- Build a LoRA GLUE command for RoBERTa or DeBERTa-v2.\n- Explain `--apply_lora`, `--lora_r`, `--lora_alpha`, and `--lora_path`.\n- Port the repository's query\u002Fvalue adaptation into another Transformers\n  version or a custom encoder.\n- Diagnose missing checkpoints, label-head mismatches, or CUDA\u002Fdistributed\n  launch failures.\n\n## Start fast\n\n1. Install the base model's Transformers and dataset dependencies in the\n   target environment, then verify that the runner's `--help` works before\n   downloading a model.\n2. Choose the model family and task. The archived recipes use LoRA on query and\n   value projections and leave key projections unchanged.\n3. Set `--apply_lora`, a positive `--lora_r`, and `--lora_alpha`. Use\n   `--lora_path` only when transferring an adapter trained on a compatible base\n   architecture\u002Ftask.\n4. Start with one GPU or a dry command. Add `--fp16` and distributed launch\n   only after the model loads and a tiny batch succeeds.\n5. Save output and logs in a new directory; do not overwrite a non-empty output\n   directory unless resuming intentionally.\n\nGenerate a safe command without opening the archived shell launchers:\n\n```bash\npython scripts\u002Fbuild_glue_lora_command.py \\\n  --model roberta-base --task mnli --script run_glue.py --num-gpus 1\n```\n\n## Model-specific defaults\n\n- RoBERTa recipes generally use `r=8`, `alpha=16`, sequence length 512, and a\n  higher learning rate than ordinary full fine-tuning.\n- DeBERTa-v2 XXL recipes use `r=16`, `alpha=32`, sequence length 256, fp16, and\n  distributed launch. Treat those as resource-heavy starting points, not\n  universal hyperparameters.\n- For MRPC, RTE, and STS-B transfer workflows, the archived recipes expect a\n  LoRA-adapted MNLI checkpoint. Ensure the adapter path and base model match.\n\n## Reroute\n\n- Layer constructor, trainable parameters, and state-dict details: use\n  `..\u002Fcore-lora-api\u002FSKILL.md`.\n- GPT-2 data-to-text training and decoding: use\n  `..\u002Fnlg-gpt2-adaptation\u002FSKILL.md`.\n\n## References\n\n- Read [GLUE workflows](references\u002Fglue-lora-workflows.md) for flags,\n  launcher patterns, task\u002Fdata layouts, and checkpoint transfer.\n- Read [model integration](references\u002Fmodel-integration.md) for the exact\n  query\u002Fvalue insertion behavior and porting checklist.\n- Read [troubleshooting](references\u002Ftroubleshooting.md) before changing rank,\n  alpha, batch size, or model code to fix a runtime failure.\n\n## Common request patterns\n\nUse this sub-skill when the user asks to:\n\n- adapt an existing RoBERTa\u002FDeBERTa classification model with LoRA;\n- translate an archived launcher into a safer single-device smoke command;\n- attach an MNLI-trained LoRA checkpoint to MRPC, RTE, or STS-B;\n- understand why a `run_glue.py` call needs `--apply_lora`, `--lora_r`, and\n  `--lora_alpha` together; or\n- trace a missing-key error back to a changed module path, task head, or model\n  family.\n\n## Exit checklist\n\nBefore returning, confirm the answer states:\n\n1. which projections receive LoRA and which stay unchanged;\n2. the command arguments that must match between training and transfer;\n3. whether the suggested run assumes CUDA, distributed launch, or a CPU smoke;\n4. where the adapter checkpoint should be loaded; and\n5. what failure signals imply a bad base model, bad head, or stale launcher.\n","# NLU GLUE adaptation\n\nUse this sub-skill for the repository's LoRA-aware sequence-classification\nworkflow: model configuration, query\u002Fvalue projection replacement, GLUE task\ncommands, adapter checkpoint transfer, and training-resource checks.\n\n## Route here\n\n- Build a LoRA GLUE command for RoBERTa or DeBERTa-v2.\n- Explain `--apply_lora`, `--lora_r`, `--lora_alpha`, and `--lora_path`.\n- Port the repository's query\u002Fvalue adaptation into another Transformers\n  version or a custom encoder.\n- Diagnose missing checkpoints, label-head mismatches, or CUDA\u002Fdistributed\n  launch failures.\n\n## Start fast\n\n1. Install the base model's Transformers and dataset dependencies in the\n   target environment, then verify that the runner's `--help` works before\n   downloading a model.\n2. Choose the model family and task. The archived recipes use LoRA on query and\n   value projections and leave key projections unchanged.\n3. Set `--apply_lora`, a positive `--lora_r`, and `--lora_alpha`. Use\n   `--lora_path` only when transferring an adapter trained on a compatible base\n   architecture\u002Ftask.\n4. Start with one GPU or a dry command. Add `--fp16` and distributed launch\n   only after the model loads and a tiny batch succeeds.\n5. Save output and logs in a new directory; do not overwrite a non-empty output\n   directory unless resuming intentionally.\n\nGenerate a safe command without opening the archived shell launchers:\n\n```bash\npython scripts\u002Fbuild_glue_lora_command.py \\\n  --model roberta-base --task mnli --script run_glue.py --num-gpus 1\n```\n\n## Model-specific defaults\n\n- RoBERTa recipes generally use `r=8`, `alpha=16`, sequence length 512, and a\n  higher learning rate than ordinary full fine-tuning.\n- DeBERTa-v2 XXL recipes use `r=16`, `alpha=32`, sequence length 256, fp16, and\n  distributed launch. Treat those as resource-heavy starting points, not\n  universal hyperparameters.\n- For MRPC, RTE, and STS-B transfer workflows, the archived recipes expect a\n  LoRA-adapted MNLI checkpoint. Ensure the adapter path and base model match.\n\n## Reroute\n\n- Layer constructor, trainable parameters, and state-dict details: use\n  `..\u002Fcore-lora-api\u002FSKILL.md`.\n- GPT-2 data-to-text training and decoding: use\n  `..\u002Fnlg-gpt2-adaptation\u002FSKILL.md`.\n\n## References\n\n- Read [GLUE workflows](references\u002Fglue-lora-workflows.md) for flags,\n  launcher patterns, task\u002Fdata layouts, and checkpoint transfer.\n- Read [model integration](references\u002Fmodel-integration.md) for the exact\n  query\u002Fvalue insertion behavior and porting checklist.\n- Read [troubleshooting](references\u002Ftroubleshooting.md) before changing rank,\n  alpha, batch size, or model code to fix a runtime failure.\n\n## Common request patterns\n\nUse this sub-skill when the user asks to:\n\n- adapt an existing RoBERTa\u002FDeBERTa classification model with LoRA;\n- translate an archived launcher into a safer single-device smoke command;\n- attach an MNLI-trained LoRA checkpoint to MRPC, RTE, or STS-B;\n- understand why a `run_glue.py` call needs `--apply_lora`, `--lora_r`, and\n  `--lora_alpha` together; or\n- trace a missing-key error back to a changed module path, task head, or model\n  family.\n\n## Exit checklist\n\nBefore returning, confirm the answer states:\n\n1. which projections receive LoRA and which stay unchanged;\n2. the command arguments that must match between training and transfer;\n3. whether the suggested run assumes CUDA, distributed launch, or a CPU smoke;\n4. where the adapter checkpoint should be loaded; and\n5. what failure signals imply a bad base model, bad head, or stale launcher.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation",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 nlu-glue-adaptation",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill nlu-glue-adaptation",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill nlu-glue-adaptation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002FSKILL.md",[46,52,56,60],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fglue-lora-workflows.md","REFERENCE","参考文档",2890,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002Freferences\u002Fglue-lora-workflows.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fmodel-integration.md",2158,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002Freferences\u002Fmodel-integration.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Ftroubleshooting.md",2476,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002Freferences\u002Ftroubleshooting.md",{"path":61,"resourceType":62,"resourceTypeDesc":63,"sizeBytes":64,"sourceHtmlUrl":65},"scripts\u002Fbuild_glue_lora_command.py","SCRIPT","脚本",3541,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Flora\u002Fsub-skills\u002Fnlu-glue-adaptation\u002Fscripts\u002Fbuild_glue_lora_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",[],[]]