[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7F9V61DWSF2F8T56TQV24D4":3,"skill-same:01M7F9V61DWSF2F8T56TQV24D4":63},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":28,"commitSha":29,"skillDirectory":30,"forksCount":31,"repositoryHtmlUrl":32,"repositoryHomepageUrl":6,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":34,"updatedAt":35,"install":36,"resources":48},"01M7F9V61DWSF2F8T56TQV24D4","training-evaluation",null,"Run and troubleshoot HRM CUDA training, checkpoint evaluation, W&B logging, and ARC prediction post-processing.","lza6\u002FArmor-piercing-system-prompts","lza6","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F128880206?v=4","01M7EB3K7WPX5T28EJ2NVHE6EG","Chinese-tingfeng","01M7EB3K97E1KQ16QWKJMFNGX8",11,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation\u002FSKILL.md",[],[],"2026-11-16 10:05:34",false,"---\nname: training-evaluation\ndescription: \"Run and troubleshoot HRM CUDA training, checkpoint evaluation, W&B\n  logging, and ARC prediction post-processing.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# HRM Training and Evaluation\n\nUse this sub-skill when the task is to launch HRM experiments, adapt Hydra\ntraining overrides, evaluate checkpoints, interpret saved prediction shards, or\ntroubleshoot CUDA\u002FW&B\u002Fcheckpoint failures.\n\n## When to use\n\n- The user wants to run `pretrain.py`, `torchrun ... pretrain.py`, or\n  `evaluate.py checkpoint=\u003CCHECKPOINT_PATH>`.\n- The task mentions `global_batch_size`, `epochs`, `eval_interval`, W&B,\n  checkpoint paths, `all_config.yaml`, `*_all_preds.\u003Crank>`, `eval\u002Fexact_accuracy`,\n  or ARC top-K aggregation.\n- A run fails due to CUDA, FlashAttention, `adam_atan2_backend`, Hydra\n  overrides, distributed launch, dataset batch sizing, W&B login\u002Foffline mode,\n  or checkpoint\u002Fconfig mismatch.\n\n## Route map\n\n1. Read [references\u002Fworkflows.md](references\u002Fworkflows.md) for verified train,\n   evaluate, checkpoint, and ARC post-processing commands.\n2. Read [references\u002Fconfiguration.md](references\u002Fconfiguration.md) for Hydra\n   defaults and important overrides.\n3. Read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md) for CUDA,\n   W&B, checkpoint, distributed, and evaluation failures.\n4. Run [scripts\u002Fcheck_training_env.py](scripts\u002Fcheck_training_env.py) before a\n   long run to validate imports and CUDA dependency readiness without training.\n5. Use [scripts\u002Farc_postprocess.py](scripts\u002Farc_postprocess.py) after\n   `evaluate.py` saves ARC prediction shards.\n\n## Baseline commands\n\nSingle-GPU Sudoku quick demo after dataset creation:\n\n```bash\nOMP_NUM_THREADS=8 python pretrain.py \\\n  data_path=data\u002Fsudoku-extreme-1k-aug-1000 \\\n  epochs=20000 eval_interval=2000 global_batch_size=384 \\\n  lr=7e-5 puzzle_emb_lr=7e-5 \\\n  weight_decay=1.0 puzzle_emb_weight_decay=1.0\n```\n\nEight-GPU ARC-1 default training:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 pretrain.py\n```\n\nARC-2:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 pretrain.py data_path=data\u002Farc-2-aug-1000\n```\n\nEvaluate an ARC checkpoint:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 evaluate.py checkpoint=\u003CCHECKPOINT_PATH>\n```\n\n## Bounded readiness check\n\n```bash\npython \u003Cskill>\u002Fsub-skills\u002Ftraining-evaluation\u002Fscripts\u002Fcheck_training_env.py \\\n  --repo-root \u002Fpath\u002Fto\u002FHRM --require-cuda\n```\n\nThis imports `pretrain.py` and `evaluate.py`, checks PyTorch CUDA, and verifies\nFlashAttention\u002Fadam-atan2 backend imports. It does not start training, download\ndata, contact W&B, or require checkpoints.\n\n## Boundaries\n\n- Use `data-preparation` for dataset creation and validation before training.\n- Use `model-architecture` for model internals, losses, identifiers, and\n  architecture-level CUDA import issues.\n- This sub-skill treats real HRM training\u002Fevaluation as CUDA-required. CPU-only\n  runs are not a verified substitute.\n- Do not run full training or checkpoint evaluation as a smoke test unless the\n  user explicitly accepts runtime, GPU, data, storage, and W&B implications.\n","# HRM Training and Evaluation\n\nUse this sub-skill when the task is to launch HRM experiments, adapt Hydra\ntraining overrides, evaluate checkpoints, interpret saved prediction shards, or\ntroubleshoot CUDA\u002FW&B\u002Fcheckpoint failures.\n\n## When to use\n\n- The user wants to run `pretrain.py`, `torchrun ... pretrain.py`, or\n  `evaluate.py checkpoint=\u003CCHECKPOINT_PATH>`.\n- The task mentions `global_batch_size`, `epochs`, `eval_interval`, W&B,\n  checkpoint paths, `all_config.yaml`, `*_all_preds.\u003Crank>`, `eval\u002Fexact_accuracy`,\n  or ARC top-K aggregation.\n- A run fails due to CUDA, FlashAttention, `adam_atan2_backend`, Hydra\n  overrides, distributed launch, dataset batch sizing, W&B login\u002Foffline mode,\n  or checkpoint\u002Fconfig mismatch.\n\n## Route map\n\n1. Read [references\u002Fworkflows.md](references\u002Fworkflows.md) for verified train,\n   evaluate, checkpoint, and ARC post-processing commands.\n2. Read [references\u002Fconfiguration.md](references\u002Fconfiguration.md) for Hydra\n   defaults and important overrides.\n3. Read [references\u002Ftroubleshooting.md](references\u002Ftroubleshooting.md) for CUDA,\n   W&B, checkpoint, distributed, and evaluation failures.\n4. Run [scripts\u002Fcheck_training_env.py](scripts\u002Fcheck_training_env.py) before a\n   long run to validate imports and CUDA dependency readiness without training.\n5. Use [scripts\u002Farc_postprocess.py](scripts\u002Farc_postprocess.py) after\n   `evaluate.py` saves ARC prediction shards.\n\n## Baseline commands\n\nSingle-GPU Sudoku quick demo after dataset creation:\n\n```bash\nOMP_NUM_THREADS=8 python pretrain.py \\\n  data_path=data\u002Fsudoku-extreme-1k-aug-1000 \\\n  epochs=20000 eval_interval=2000 global_batch_size=384 \\\n  lr=7e-5 puzzle_emb_lr=7e-5 \\\n  weight_decay=1.0 puzzle_emb_weight_decay=1.0\n```\n\nEight-GPU ARC-1 default training:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 pretrain.py\n```\n\nARC-2:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 pretrain.py data_path=data\u002Farc-2-aug-1000\n```\n\nEvaluate an ARC checkpoint:\n\n```bash\nOMP_NUM_THREADS=8 torchrun --nproc-per-node 8 evaluate.py checkpoint=\u003CCHECKPOINT_PATH>\n```\n\n## Bounded readiness check\n\n```bash\npython \u003Cskill>\u002Fsub-skills\u002Ftraining-evaluation\u002Fscripts\u002Fcheck_training_env.py \\\n  --repo-root \u002Fpath\u002Fto\u002FHRM --require-cuda\n```\n\nThis imports `pretrain.py` and `evaluate.py`, checks PyTorch CUDA, and verifies\nFlashAttention\u002Fadam-atan2 backend imports. It does not start training, download\ndata, contact W&B, or require checkpoints.\n\n## Boundaries\n\n- Use `data-preparation` for dataset creation and validation before training.\n- Use `model-architecture` for model internals, losses, identifiers, and\n  architecture-level CUDA import issues.\n- This sub-skill treats real HRM training\u002Fevaluation as CUDA-required. CPU-only\n  runs are not a verified substitute.\n- Do not run full training or checkpoint evaluation as a smoke test unless the\n  user explicitly accepts runtime, GPU, data, storage, and W&B implications.\n",{"name":5,"license":24,"metadata":25,"description":7,"disable-model-invocation":27},"Apache 2.0",{"disco-role":26},"operating",true,"main","d69096e239f2a89aabc2cd078c47664a9d3b1e11","06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation",3,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts","https:\u002F\u002Fgithub.com\u002Flza6","个人vx：Tf00798\r\n个人✈：@TFGY999","2026-10-08 14:29:22",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":27},[38,41,44],{"tool":39,"command":40},"npx","npx skills add lza6\u002FArmor-piercing-system-prompts --skill training-evaluation",{"tool":42,"command":43},"bunx","bunx skills add lza6\u002FArmor-piercing-system-prompts --skill training-evaluation",{"tool":45,"command":46},"pnpm","pnpm dlx skills add lza6\u002FArmor-piercing-system-prompts --skill training-evaluation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation\u002FSKILL.md",[49,55,59],{"path":50,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":53,"sourceHtmlUrl":54},"references\u002Fconfiguration.md","REFERENCE","参考文档",2951,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation\u002Freferences\u002Fconfiguration.md",{"path":56,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":57,"sourceHtmlUrl":58},"references\u002Ftroubleshooting.md",3523,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation\u002Freferences\u002Ftroubleshooting.md",{"path":60,"resourceType":51,"resourceTypeDesc":52,"sizeBytes":61,"sourceHtmlUrl":62},"references\u002Fworkflows.md",3224,"https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F06-前端与设计\u002Fhrm\u002Fsub-skills\u002Ftraining-evaluation\u002Freferences\u002Fworkflows.md",[64,72,77,84,91,98],{"skillNo":65,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":19},"01M7FFG485KY83TD6HSC4YPCMC","obsidian-clipper-template-creator","Guide for creating templates for the Obsidian Web Clipper. Use when you want to create a new clipping template, understand available variables, or format clipped content.","MIT","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator\u002FSKILL.md",[],[],{"skillNo":73,"name":66,"nameZh":6,"description":67,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M7FFG0T6AZ0STEJPSB3CX84K","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobsidian-clipper-template-creator-antigravity-awesome-skills-main\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M7FFFY0QEM14SKWPDYGMXM2P","python-api","Programmatic Object-Detection-Metrics API guidance for BoundingBox, BoundingBoxes, Evaluator, enums, coordinate conversion, AP\u002FmAP semantics, and source-style import diagnostics.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Fpython-api\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M7FFFVZQZ2GJ0C0TGP7JHD20","file-evaluation","Folder-based PASCAL VOC AP\u002FmAP evaluation from ground-truth and detection text files.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002Fsub-skills\u002Ffile-evaluation\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M7FFFSFNYSE77SJ6PE1CK2DA","object-detection-metrics","Guides legacy Object-Detection-Metrics PASCAL VOC AP\u002FmAP evaluation from detection text folders and direct Python API classes.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fobject-detection-metrics\u002FSKILL.md",[],[],{"skillNo":99,"name":100,"nameZh":6,"description":101,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":68,"licenseFileUrl":6,"sourceHtmlUrl":102,"categories":103,"keywords":104,"publishedAt":19},"01M7FFFP0GCK1RVXBCPQ8APBMT","nx-workspace-patterns","Configure and optimize Nx monorepo workspaces. Use when setting up Nx, configuring project boundaries, optimizing build caching, or implementing affected commands.","https:\u002F\u002Fgithub.com\u002Flza6\u002FArmor-piercing-system-prompts\u002Fblob\u002Fmain\u002F07-后端与架构\u002Fnx-workspace-patterns\u002FSKILL.md",[],[]]