[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M794TJC2EYC2W9X3NJR51KVW":3,"skill-same:01M794TJC2EYC2W9X3NJR51KVW":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},"01M794TJC2EYC2W9X3NJR51KVW","training-and-inference",null,"Route legacy SECOND and PointPillars model construction, training, evaluation, checkpoint restoration, and inference with explicit configuration, device, and backend guards.","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\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: training-and-inference\ndescription: \"Route legacy SECOND and PointPillars model construction, training,\n  evaluation, checkpoint restoration, and inference with explicit configuration,\n  device, and backend guards.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# Training and inference\n\nUse this sub-skill when a task asks to train or evaluate SECOND\u002FPointPillars,\nbuild a `VoxelNet`, restore a `.tckpt`, run inference, select a model config,\nor reason about multi-GPU, fp16, NMS, or spconv errors.\n\n## Safety and compatibility boundary\n\n- This repository is deprecated. Its public README recommends OpenPCDet or\n  MMDetection3D for new work; prefer a maintained implementation unless the\n  historical checkpoint or experiment specifically requires this code.\n- The detector route is **legacy-backend guarded and unverified**. Source code\n  depends on old `spconv` APIs (`VoxelGeneratorV2` and NMS symbols), legacy\n  Numba CUDA behavior, and an old PyTorch integration. A modern spconv 2.x\n  import is not proof of compatibility.\n- Run the bundled non-invasive probe before any detector import:\n  `python \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py`.\n  Add `--require-detector` only when deciding whether to attempt a detector\n  route. Without it, the probe is diagnostic and returns success even when the\n  legacy symbols are absent.\n- CUDA availability, a successful CUDA tensor smoke test, or a CPU import does\n  not prove that sparse model construction, NMS, training, or inference works.\n  Do not claim detector execution was verified from those observations.\n- This package has no setup metadata. Use an explicit, isolated environment\n  only when a separately supplied compatible checkout is intentionally in\n  scope. Do not assume a package install or silently mutate an existing\n  environment. This skill does not bundle the historical detector entry point.\n\n## Route the request\n\n1. **Config\u002Fdata contract:** choose a config and validate dataset paths,\n   generated info\u002Fdatabase files, class order, voxel range, and point feature\n   count. Route dataset generation and layout preparation to\n   [data-preparation](..\u002Fdata-preparation\u002FSKILL.md).\n2. **Model\u002Ftraining:** map `network_class_name`, VFE, middle extractor, and RPN\n   names through the registries; use the Fire commands and output semantics in\n   [workflows](references\u002Fworkflows.md).\n3. **Evaluation\u002Fgeometry:** keep box encoding, NMS, score thresholds, and\n   evaluator implementation questions with\n   [geometry-and-evaluation](..\u002Fgeometry-and-evaluation\u002FSKILL.md).\n4. **Viewer\u002Fservice:** route web API, browser, and server operations to\n   [visualization-and-serving](..\u002Fvisualization-and-serving\u002FSKILL.md); do not\n   start the legacy viewer as a training smoke test.\n\n## Reference map\n\n- [workflows](references\u002Fworkflows.md): Fire commands, guarded train\u002Fresume,\n  multi-GPU scaling, evaluation outputs, inference, pretrained loading, and\n  NuScenes tuning notes.\n- [configuration](references\u002Fconfiguration.md): config families, protobuf\n  fields, model registries, class order, optimizers, and schedules.\n- [API reference](references\u002Fapi-reference.md): public signatures, model\n  methods, builder contracts, freeze filters, and torchplus checkpoints.\n- [compatibility](references\u002Fcompatibility.md): exact backend gate, known\n  modern-spconv gaps, and evidence needed to widen executable scope.\n- [troubleshooting](references\u002Ftroubleshooting.md): install\u002Fimport, optional\n  dependencies, data\u002Fconfig, CLI\u002FAPI, checkpoint, multi-GPU, fp16, and\n  workflow-specific failures.\n\n## Guarded command patterns\n\nFirst run the bundled non-invasive probe:\n\n```bash\npython \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py\npython \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py --require-detector\n```\n\nThe second command is the detector gate. This skill does not bundle or invoke\nthat historical writer\u002Frunner, so no source-checkout command is presented as a\nrunnable recipe. If the probe reports missing legacy symbols, stop before any\nhistorical `train`, `evaluate`, or `TorchInferenceContext` attempt; capture the\nexact missing names and use [compatibility](references\u002Fcompatibility.md) for\nrecovery. If a user separately supplies a compatible checkout, the historical\nargument shapes are documented in [workflows](references\u002Fworkflows.md).\n\nThese are guarded historical routes, not verified recipes. `train` refuses an\nexisting `model_dir` unless `--resume=True`; `--create_folder=True` creates a\nnew timestamped folder when the requested prefix already exists. Keep model\noutputs outside the source tree and back up checkpoints before experiments.\n\n## Core decisions\n\n- `batch_size` and `num_workers` in the training input config are **per GPU**.\n  The implementation multiplies both by the visible GPU count for\n  `--multi_gpu=True`; do not multiply the config values yourself.\n- The README says to divide `train_config.steps` and `steps_per_eval` by the\n  number of GPUs when scaling a single-GPU schedule to multi-GPU. Verify this\n  arithmetic before launching; an undivided schedule changes total optimizer\n  updates and evaluation cadence.\n- fp16 requires the historical Apex path and a compatible sparse backend;\n  `enable_mixed_precision: true` is not equivalent to modern PyTorch AMP.\n  Honor the source assertion that `max_number_of_voxels * batch_size \u003C 65535`.\n- `measure_time=True` enables CUDA synchronization\u002Ftimers in model paths and\n  can distort throughput. Use it only for a deliberate timing run.\n- `evaluate` writes `result.pkl` below `\u003Cresult-path>\u002Fstep_\u003Cglobal_step>` and\n  invokes the dataset evaluator. The historical signature does not accept the\n  README-era `--pickle_result` keyword; do not pass it to a separately supplied\n  runner unless its help output proves otherwise. Route official label\n  conversion to geometry\u002Fevaluation after checking the exact dataset API.\n- Checkpoints are named by model\u002Foptimizer plus step and indexed in\n  `checkpoints.json`. Restore the latest compatible state, or pass an explicit\n  `.tckpt`; never mix a checkpoint with an incompatible config.\n\nFor exact signatures, config fields, model registry names, checkpoint utilities,\nand failure recovery, read the linked references before giving a detailed\ncommand or API answer.\n","# Training and inference\n\nUse this sub-skill when a task asks to train or evaluate SECOND\u002FPointPillars,\nbuild a `VoxelNet`, restore a `.tckpt`, run inference, select a model config,\nor reason about multi-GPU, fp16, NMS, or spconv errors.\n\n## Safety and compatibility boundary\n\n- This repository is deprecated. Its public README recommends OpenPCDet or\n  MMDetection3D for new work; prefer a maintained implementation unless the\n  historical checkpoint or experiment specifically requires this code.\n- The detector route is **legacy-backend guarded and unverified**. Source code\n  depends on old `spconv` APIs (`VoxelGeneratorV2` and NMS symbols), legacy\n  Numba CUDA behavior, and an old PyTorch integration. A modern spconv 2.x\n  import is not proof of compatibility.\n- Run the bundled non-invasive probe before any detector import:\n  `python \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py`.\n  Add `--require-detector` only when deciding whether to attempt a detector\n  route. Without it, the probe is diagnostic and returns success even when the\n  legacy symbols are absent.\n- CUDA availability, a successful CUDA tensor smoke test, or a CPU import does\n  not prove that sparse model construction, NMS, training, or inference works.\n  Do not claim detector execution was verified from those observations.\n- This package has no setup metadata. Use an explicit, isolated environment\n  only when a separately supplied compatible checkout is intentionally in\n  scope. Do not assume a package install or silently mutate an existing\n  environment. This skill does not bundle the historical detector entry point.\n\n## Route the request\n\n1. **Config\u002Fdata contract:** choose a config and validate dataset paths,\n   generated info\u002Fdatabase files, class order, voxel range, and point feature\n   count. Route dataset generation and layout preparation to\n   [data-preparation](..\u002Fdata-preparation\u002FSKILL.md).\n2. **Model\u002Ftraining:** map `network_class_name`, VFE, middle extractor, and RPN\n   names through the registries; use the Fire commands and output semantics in\n   [workflows](references\u002Fworkflows.md).\n3. **Evaluation\u002Fgeometry:** keep box encoding, NMS, score thresholds, and\n   evaluator implementation questions with\n   [geometry-and-evaluation](..\u002Fgeometry-and-evaluation\u002FSKILL.md).\n4. **Viewer\u002Fservice:** route web API, browser, and server operations to\n   [visualization-and-serving](..\u002Fvisualization-and-serving\u002FSKILL.md); do not\n   start the legacy viewer as a training smoke test.\n\n## Reference map\n\n- [workflows](references\u002Fworkflows.md): Fire commands, guarded train\u002Fresume,\n  multi-GPU scaling, evaluation outputs, inference, pretrained loading, and\n  NuScenes tuning notes.\n- [configuration](references\u002Fconfiguration.md): config families, protobuf\n  fields, model registries, class order, optimizers, and schedules.\n- [API reference](references\u002Fapi-reference.md): public signatures, model\n  methods, builder contracts, freeze filters, and torchplus checkpoints.\n- [compatibility](references\u002Fcompatibility.md): exact backend gate, known\n  modern-spconv gaps, and evidence needed to widen executable scope.\n- [troubleshooting](references\u002Ftroubleshooting.md): install\u002Fimport, optional\n  dependencies, data\u002Fconfig, CLI\u002FAPI, checkpoint, multi-GPU, fp16, and\n  workflow-specific failures.\n\n## Guarded command patterns\n\nFirst run the bundled non-invasive probe:\n\n```bash\npython \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py\npython \u003Ctraining-skill-root>\u002Fscripts\u002Fcheck_legacy_backend.py --require-detector\n```\n\nThe second command is the detector gate. This skill does not bundle or invoke\nthat historical writer\u002Frunner, so no source-checkout command is presented as a\nrunnable recipe. If the probe reports missing legacy symbols, stop before any\nhistorical `train`, `evaluate`, or `TorchInferenceContext` attempt; capture the\nexact missing names and use [compatibility](references\u002Fcompatibility.md) for\nrecovery. If a user separately supplies a compatible checkout, the historical\nargument shapes are documented in [workflows](references\u002Fworkflows.md).\n\nThese are guarded historical routes, not verified recipes. `train` refuses an\nexisting `model_dir` unless `--resume=True`; `--create_folder=True` creates a\nnew timestamped folder when the requested prefix already exists. Keep model\noutputs outside the source tree and back up checkpoints before experiments.\n\n## Core decisions\n\n- `batch_size` and `num_workers` in the training input config are **per GPU**.\n  The implementation multiplies both by the visible GPU count for\n  `--multi_gpu=True`; do not multiply the config values yourself.\n- The README says to divide `train_config.steps` and `steps_per_eval` by the\n  number of GPUs when scaling a single-GPU schedule to multi-GPU. Verify this\n  arithmetic before launching; an undivided schedule changes total optimizer\n  updates and evaluation cadence.\n- fp16 requires the historical Apex path and a compatible sparse backend;\n  `enable_mixed_precision: true` is not equivalent to modern PyTorch AMP.\n  Honor the source assertion that `max_number_of_voxels * batch_size \u003C 65535`.\n- `measure_time=True` enables CUDA synchronization\u002Ftimers in model paths and\n  can distort throughput. Use it only for a deliberate timing run.\n- `evaluate` writes `result.pkl` below `\u003Cresult-path>\u002Fstep_\u003Cglobal_step>` and\n  invokes the dataset evaluator. The historical signature does not accept the\n  README-era `--pickle_result` keyword; do not pass it to a separately supplied\n  runner unless its help output proves otherwise. Route official label\n  conversion to geometry\u002Fevaluation after checking the exact dataset API.\n- Checkpoints are named by model\u002Foptimizer plus step and indexed in\n  `checkpoints.json`. Restore the latest compatible state, or pass an explicit\n  `.tckpt`; never mix a checkpoint with an incompatible config.\n\nFor exact signatures, config fields, model registry names, checkpoint utilities,\nand failure recovery, read the linked references before giving a detailed\ncommand or API answer.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference",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 training-and-inference",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill training-and-inference",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill training-and-inference","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002FSKILL.md",[46,52,56,60,64,68],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fapi-reference.md","REFERENCE","参考文档",7710,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Freferences\u002Fapi-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fcompatibility.md",6453,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Freferences\u002Fcompatibility.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fconfiguration.md",7718,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Freferences\u002Fconfiguration.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Ftroubleshooting.md",10814,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Freferences\u002Ftroubleshooting.md",{"path":65,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":66,"sourceHtmlUrl":67},"references\u002Fworkflows.md",10071,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Freferences\u002Fworkflows.md",{"path":69,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":72,"sourceHtmlUrl":73},"scripts\u002Fcheck_legacy_backend.py","SCRIPT","脚本",6077,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsecond-pytorch\u002Fsub-skills\u002Ftraining-and-inference\u002Fscripts\u002Fcheck_legacy_backend.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",[],[]]