[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M796N6CVS8SE1V4BDF02E6ZP":3,"skill-same:01M796N6CVS8SE1V4BDF02E6ZP":74},{"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},"01M796N6CVS8SE1V4BDF02E6ZP","model-training",null,"Guides TransFuser imitation-model training, dataset and configuration validation, backbone and PointPillars selection, single-GPU or torchrun DDP launch, loss interpretation, and checkpoint recovery.","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\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: model-training\ndescription: \"Guides TransFuser imitation-model training, dataset and\n  configuration validation, backbone and PointPillars selection, single-GPU or\n  torchrun DDP launch, loss interpretation, and checkpoint recovery.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# TransFuser Model Training\n\nUse this sub-skill when a task involves training or resuming a TransFuser-family model, choosing a fusion backbone, validating training data, preparing a single-GPU or DDP command, interpreting losses and logs, or diagnosing a checkpoint.\n\n## Non-negotiable runtime boundary\n\n- Training requires an NVIDIA CUDA runtime. The repository training path unconditionally selects a CUDA device; there is **no CPU training substitute**.\n- Preserve the legacy package family unless deliberately porting and revalidating it: Python 3.7, PyTorch 1.12.1+cu113, `mmcv-full` 1.6.0, `mmdet` 2.25.0, `mmsegmentation` 0.25.0, `mmcls` 0.25.0, `torch-scatter` 2.1.0, and `timm` 0.6.7. This combination passed imports and `pip check` with CUDA visible on an NVIDIA A100 during skill construction.\n- Do not treat a successful CPU import, `pip check`, or CLI parser run as proof that training works. Require `torch.cuda.is_available()`, a CUDA allocation, and matching compiled extensions.\n- Model construction can ask `timm` for pretrained image weights. Prepare an approved local cache or explicitly authorize network access before constructing a fresh model in a network-restricted environment.\n\n## Route the task\n\n1. **Prepare a new or resumed run:** read [training-workflow.md](references\u002Ftraining-workflow.md). It contains every `train.py` flag\u002Fdefault, single-GPU and `torchrun` recipes, logging, losses, caching, debug behavior, and resume semantics.\n2. **Validate data or diagnose a shape\u002Fkey failure:** read [data-format.md](references\u002Fdata-format.md), then run [validate_training_setup.py](scripts\u002Fvalidate_training_setup.py). The helper only inspects paths, metadata, environment variables, and optional imports; it never trains or downloads.\n3. **Choose a backbone, architecture, config override, or PointPillars:** read [api-reference.md](references\u002Fapi-reference.md). It covers `GlobalConfig`, `CARLA_Data`, `LidarCenterNet`, all four fusion backbones, tensor contracts, and the PointPillars path.\n4. **Resume, compare, or diagnose weights:** first read the checkpoint section in [training-workflow.md](references\u002Ftraining-workflow.md), then run [inspect_checkpoint.py](scripts\u002Finspect_checkpoint.py). Its default mode does not unpickle the checkpoint; only use `--unsafe-load` for a trusted file.\n5. **Recover from a failure:** read [troubleshooting.md](references\u002Ftroubleshooting.md) for missing paths, DDP rank variables, invalid backbones, checkpoint incompatibility, cache\u002Fdebug failures, and missing OpenMMLab or `torch-scatter` components.\n\n## Safe first pass\n\nFrom any directory, validate the intended run before launching it:\n\n```bash\npython \u003Cthis-sub-skill>\u002Fscripts\u002Fvalidate_training_setup.py \\\n  --repo-root \u003Ctransfuser-checkout> \\\n  --dataset-root \u003Cdataset-root> \\\n  --setting 02_05_withheld \\\n  --backbone transFuser \\\n  --parallel-training 0\n```\n\nAdd `--check-runtime` only in the environment intended for training. It imports the compatibility stack and repository training modules and requires working CUDA, but it does not instantiate a model or start training.\n\n## Boundaries and handoffs\n\n- Dataset acquisition, CARLA route\u002Fscenario generation, and 210-GB collection are owned by [data-and-routes](..\u002Fdata-and-routes\u002FSKILL.md).\n- Loading a trained model into `HybridAgent`, handling `args.txt` at inference time, and CARLA control are owned by [sensor-agent](..\u002Fsensor-agent\u002FSKILL.md).\n- Longest6 execution and result parsing are owned by [carla-evaluation](..\u002Fcarla-evaluation\u002FSKILL.md).\n- Full training, model-weight downloads, CARLA simulation, Docker builds, and cloud submission are intentionally not performed by bundled helpers.\n\n## Completion criteria\n\nBefore launching training, require all of the following:\n\n- the dataset split selects at least one train route and, for `02_05_withheld`, at least one withheld validation route;\n- required modality files and future labels exist for eligible samples;\n- the exact backbone string and architecture choices are compatible;\n- CUDA and compiled dependency probes pass;\n- single-GPU runs set `--parallel_training 0`, while DDP runs are launched by `torchrun` with `RANK`, `LOCAL_RANK`, and `WORLD_SIZE` populated;\n- resume runs have both `model_\u003Cepoch>.pth` and the derived `optimizer_\u003Cepoch>.pth`, a deliberate `--start_epoch`, and matching architecture\u002Fconfig provenance;\n- the output resolves to `\u003Clogdir>\u002F\u003Cid>` and has sufficient storage for checkpoints, TensorBoard events, optional visualizations, and any disk cache.\n","# TransFuser Model Training\n\nUse this sub-skill when a task involves training or resuming a TransFuser-family model, choosing a fusion backbone, validating training data, preparing a single-GPU or DDP command, interpreting losses and logs, or diagnosing a checkpoint.\n\n## Non-negotiable runtime boundary\n\n- Training requires an NVIDIA CUDA runtime. The repository training path unconditionally selects a CUDA device; there is **no CPU training substitute**.\n- Preserve the legacy package family unless deliberately porting and revalidating it: Python 3.7, PyTorch 1.12.1+cu113, `mmcv-full` 1.6.0, `mmdet` 2.25.0, `mmsegmentation` 0.25.0, `mmcls` 0.25.0, `torch-scatter` 2.1.0, and `timm` 0.6.7. This combination passed imports and `pip check` with CUDA visible on an NVIDIA A100 during skill construction.\n- Do not treat a successful CPU import, `pip check`, or CLI parser run as proof that training works. Require `torch.cuda.is_available()`, a CUDA allocation, and matching compiled extensions.\n- Model construction can ask `timm` for pretrained image weights. Prepare an approved local cache or explicitly authorize network access before constructing a fresh model in a network-restricted environment.\n\n## Route the task\n\n1. **Prepare a new or resumed run:** read [training-workflow.md](references\u002Ftraining-workflow.md). It contains every `train.py` flag\u002Fdefault, single-GPU and `torchrun` recipes, logging, losses, caching, debug behavior, and resume semantics.\n2. **Validate data or diagnose a shape\u002Fkey failure:** read [data-format.md](references\u002Fdata-format.md), then run [validate_training_setup.py](scripts\u002Fvalidate_training_setup.py). The helper only inspects paths, metadata, environment variables, and optional imports; it never trains or downloads.\n3. **Choose a backbone, architecture, config override, or PointPillars:** read [api-reference.md](references\u002Fapi-reference.md). It covers `GlobalConfig`, `CARLA_Data`, `LidarCenterNet`, all four fusion backbones, tensor contracts, and the PointPillars path.\n4. **Resume, compare, or diagnose weights:** first read the checkpoint section in [training-workflow.md](references\u002Ftraining-workflow.md), then run [inspect_checkpoint.py](scripts\u002Finspect_checkpoint.py). Its default mode does not unpickle the checkpoint; only use `--unsafe-load` for a trusted file.\n5. **Recover from a failure:** read [troubleshooting.md](references\u002Ftroubleshooting.md) for missing paths, DDP rank variables, invalid backbones, checkpoint incompatibility, cache\u002Fdebug failures, and missing OpenMMLab or `torch-scatter` components.\n\n## Safe first pass\n\nFrom any directory, validate the intended run before launching it:\n\n```bash\npython \u003Cthis-sub-skill>\u002Fscripts\u002Fvalidate_training_setup.py \\\n  --repo-root \u003Ctransfuser-checkout> \\\n  --dataset-root \u003Cdataset-root> \\\n  --setting 02_05_withheld \\\n  --backbone transFuser \\\n  --parallel-training 0\n```\n\nAdd `--check-runtime` only in the environment intended for training. It imports the compatibility stack and repository training modules and requires working CUDA, but it does not instantiate a model or start training.\n\n## Boundaries and handoffs\n\n- Dataset acquisition, CARLA route\u002Fscenario generation, and 210-GB collection are owned by [data-and-routes](..\u002Fdata-and-routes\u002FSKILL.md).\n- Loading a trained model into `HybridAgent`, handling `args.txt` at inference time, and CARLA control are owned by [sensor-agent](..\u002Fsensor-agent\u002FSKILL.md).\n- Longest6 execution and result parsing are owned by [carla-evaluation](..\u002Fcarla-evaluation\u002FSKILL.md).\n- Full training, model-weight downloads, CARLA simulation, Docker builds, and cloud submission are intentionally not performed by bundled helpers.\n\n## Completion criteria\n\nBefore launching training, require all of the following:\n\n- the dataset split selects at least one train route and, for `02_05_withheld`, at least one withheld validation route;\n- required modality files and future labels exist for eligible samples;\n- the exact backbone string and architecture choices are compatible;\n- CUDA and compiled dependency probes pass;\n- single-GPU runs set `--parallel_training 0`, while DDP runs are launched by `torchrun` with `RANK`, `LOCAL_RANK`, and `WORLD_SIZE` populated;\n- resume runs have both `model_\u003Cepoch>.pth` and the derived `optimizer_\u003Cepoch>.pth`, a deliberate `--start_epoch`, and matching architecture\u002Fconfig provenance;\n- the output resolves to `\u003Clogdir>\u002F\u003Cid>` and has sufficient storage for checkpoints, TensorBoard events, optional visualizations, and any disk cache.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training",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 model-training",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill model-training",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill model-training","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002FSKILL.md",[46,52,56,60,64,70],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fapi-reference.md","REFERENCE","参考文档",10982,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Freferences\u002Fapi-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fdata-format.md",9413,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Freferences\u002Fdata-format.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Ftraining-workflow.md",15574,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Freferences\u002Ftraining-workflow.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Ftroubleshooting.md",13239,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Freferences\u002Ftroubleshooting.md",{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},"scripts\u002Finspect_checkpoint.py","SCRIPT","脚本",7421,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Fscripts\u002Finspect_checkpoint.py",{"path":71,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":72,"sourceHtmlUrl":73},"scripts\u002Fvalidate_training_setup.py",17711,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Ftransfuser\u002Fsub-skills\u002Fmodel-training\u002Fscripts\u002Fvalidate_training_setup.py",[75,83,90,97,104,111],{"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":79,"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.","Apache-2.0","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":79,"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":79,"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":79,"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":79,"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":79,"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",[],[]]