[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78ZXGD7T32T2GHDH16E9QJ4":3,"skill-same:01M78ZXGD7T32T2GHDH16E9QJ4":70},{"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},"01M78ZXGD7T32T2GHDH16E9QJ4","generation",null,"Route safe MatterGen crystal generation from pretrained or local checkpoints, including unconditional, property-conditional, multi-property, CSP composition-targeted sampling, sampling overrides, trajectories, and the CrystalGenerator API.","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\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: generation\ndescription: \"Route safe MatterGen crystal generation from pretrained or local\n  checkpoints, including unconditional, property-conditional, multi-property,\n  CSP composition-targeted sampling, sampling overrides, trajectories, and the\n  CrystalGenerator API.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# MatterGen generation\n\nUse this sub-skill when the task is to generate inorganic crystal candidates with\nMatterGen 1.0.3. It owns checkpoint selection, conditioning, CSP compositions,\nsampling configuration, output artifacts, and the public `CrystalGenerator` API.\nIt does **not** own training, fine-tuning, relaxation\u002Fevaluation, dataset\ncreation, or paper-result interpretation; route those tasks to the appropriate\nMatterGen skill instead.\n\n## Route before running\n\n1. Decide whether the checkpoint is a named pretrained model or a local\n   checkpoint directory. `pretrained_name` and `model_path` are mutually\n   exclusive, and at least one is required.\n2. Decide whether the request is unconditional, property-conditioned, or CSP.\n   CSP means fixed element counts and requires a CSP-trained checkpoint plus the\n   `csp` sampling config; a normal base checkpoint cannot be turned into CSP by\n   adding a composition argument.\n3. Check that every requested property is a condition the selected checkpoint\n   was trained to accept. For the catalog and known condition sets, read\n   [model-overview.md](references\u002Fmodel-overview.md).\n4. Bound `batch_size * num_batches`, GPU memory, trajectory storage, and output\n   location before loading a model. Use the safe validator first:\n   [scripts\u002Fgenerate_materials.py](scripts\u002Fgenerate_materials.py).\n5. Read [workflows.md](references\u002Fworkflows.md) for the matching invocation and\n   [troubleshooting.md](references\u002Ftroubleshooting.md) before retrying failures.\n\n## Supported routes\n\n- **Unconditional:** use `mattergen_base` or `mp_20_base` with no property map\n  and no target composition. `diffusion_guidance_factor=0.0` is unconditional\n  behavior.\n- **Single property:** use a compatible fine-tuned model and pass a mapping such\n  as `{'dft_mag_density': 0.15}`. The CLI is Fire-based; quote the whole map\n  and do not put whitespace around its key\u002Fvalue separator.\n- **Multiple properties:** pass one mapping, for example\n  `{'energy_above_hull': 0.05, 'chemical_system': 'Li-O'}`, to a checkpoint\n  trained jointly for those conditions. A property name being globally known\n  does not prove that every checkpoint supports it.\n- **CSP:** pass a list of element-count mappings, such as\n  `[{'Na': 1, 'Cl': 1}]`, and `--sampling-config-name=csp`. The CSP config omits\n  atomic-number predictor\u002Fcorrector parts because atom types are fixed by the\n  requested composition. Use a CSP-trained local model; the public pretrained\n  catalog is primarily the non-CSP catalog.\n- **Python API:** construct `MatterGenCheckpointInfo`, then `CrystalGenerator`,\n  and call `generate`. Use the exact public signatures in\n  [api-reference.md](references\u002Fapi-reference.md); do not reproduce private\n  model-loading internals in a caller script.\n\n## Safe execution procedure\n\n1. Run the bundled helper with `--help`, then without `--run`, to parse and\n   validate the request. It never downloads a checkpoint or starts sampling by\n   default. Fix errors before using `--run`.\n2. For a Hub model, verify network\u002Fcache permission and choose a small smoke\n   batch first. `MatterGenCheckpointInfo.from_hf_hub(...)` downloads missing\n   files only when explicitly running. For a local model, verify its directory\n   contains `config.yaml` and hydrated checkpoint files; a Git-LFS pointer is\n   not a usable checkpoint.\n3. Set `--record-trajectories=False` (native CLI spelling) when trajectory ZIPs\n   are not needed. Keep the first run small; increase batch size only after a\n   successful memory smoke test. MatterGen selects CUDA when available, then\n   MPS, then CPU; generation is substantially slower and may be impractical on\n   CPU.\n4. Run into a new or deliberately chosen output directory. Never assume a\n   partially written output is a complete sample set.\n5. Inspect the returned\u002Fgenerated structure count and the expected files:\n   `generated_crystals_cif.zip`, `generated_crystals.extxyz`, and, when enabled,\n   `generated_trajectories.zip`. Preserve the resolved checkpoint\u002Fconfig and\n   sampling overrides with the experiment record outside this skill tree.\n\n## Conditioning and guidance rules\n\n`properties_to_condition_on` is a `dict` whose keys must be model-supported\nproperty source IDs. `diffusion_guidance_factor` is the classifier-free guidance\nweight: 0 selects the unconditional score, 1 selects the conditional score,\nand larger nonnegative values increasingly favor the requested condition while\nusually reducing diversity\u002Frealism. Guidance cannot add a missing property\nhead, repair malformed values, or make an incompatible checkpoint conditional.\n\n`sampling_config_overrides` are Hydra override strings. The installed default\nsampling configuration uses 1,000 diffusion steps and atom-number denoising;\nthe installed CSP configuration uses composition conditioning and omits\natomic-number denoising. Use overrides only for known config keys and record\nthem exactly. Avoid changing sampler fields until a baseline completes.\n\n## Failure recovery\n\n- Missing assets: stop and acquire\u002Fhydrate the selected checkpoint (Hub or LFS);\n  do not confuse CUDA readiness with asset readiness.\n- Import\u002Fbackend errors: run the helper's validation without `--run`, then use\n  the environment's public `python -c` import\u002Fhelp probes. Check optional ASE,\n  PyTorch Geometric CUDA wheels, `pymatgen`, Hydra, and MatterSim dependencies;\n  do not silently switch to a different package version.\n- Config\u002Fcondition errors: reduce to an unconditional one-batch run, then add\n  one condition, then multiple conditions. For CSP, switch both the checkpoint\n  and config rather than only changing the CLI flag.\n- Out-of-memory or stalled jobs: lower `batch_size`, lower `num_batches`, turn\n  off trajectories, and retry in a fresh output directory. Do not infer model\n  quality from an interrupted run.\n- Output errors: ensure the output directory is writable and has space; verify\n  ZIP members and frame count before downstream evaluation.\n\n## Bundled operating material\n\n- [API signatures and contracts](references\u002Fapi-reference.md)\n- [CLI, API, CSP, and override workflows](references\u002Fworkflows.md)\n- [Checkpoint catalog and model limits](references\u002Fmodel-overview.md)\n- [Failure diagnosis and recovery matrix](references\u002Ftroubleshooting.md)\n- [Safe parser\u002Fvalidator and explicit-run helper](scripts\u002Fgenerate_materials.py)\n","# MatterGen generation\n\nUse this sub-skill when the task is to generate inorganic crystal candidates with\nMatterGen 1.0.3. It owns checkpoint selection, conditioning, CSP compositions,\nsampling configuration, output artifacts, and the public `CrystalGenerator` API.\nIt does **not** own training, fine-tuning, relaxation\u002Fevaluation, dataset\ncreation, or paper-result interpretation; route those tasks to the appropriate\nMatterGen skill instead.\n\n## Route before running\n\n1. Decide whether the checkpoint is a named pretrained model or a local\n   checkpoint directory. `pretrained_name` and `model_path` are mutually\n   exclusive, and at least one is required.\n2. Decide whether the request is unconditional, property-conditioned, or CSP.\n   CSP means fixed element counts and requires a CSP-trained checkpoint plus the\n   `csp` sampling config; a normal base checkpoint cannot be turned into CSP by\n   adding a composition argument.\n3. Check that every requested property is a condition the selected checkpoint\n   was trained to accept. For the catalog and known condition sets, read\n   [model-overview.md](references\u002Fmodel-overview.md).\n4. Bound `batch_size * num_batches`, GPU memory, trajectory storage, and output\n   location before loading a model. Use the safe validator first:\n   [scripts\u002Fgenerate_materials.py](scripts\u002Fgenerate_materials.py).\n5. Read [workflows.md](references\u002Fworkflows.md) for the matching invocation and\n   [troubleshooting.md](references\u002Ftroubleshooting.md) before retrying failures.\n\n## Supported routes\n\n- **Unconditional:** use `mattergen_base` or `mp_20_base` with no property map\n  and no target composition. `diffusion_guidance_factor=0.0` is unconditional\n  behavior.\n- **Single property:** use a compatible fine-tuned model and pass a mapping such\n  as `{'dft_mag_density': 0.15}`. The CLI is Fire-based; quote the whole map\n  and do not put whitespace around its key\u002Fvalue separator.\n- **Multiple properties:** pass one mapping, for example\n  `{'energy_above_hull': 0.05, 'chemical_system': 'Li-O'}`, to a checkpoint\n  trained jointly for those conditions. A property name being globally known\n  does not prove that every checkpoint supports it.\n- **CSP:** pass a list of element-count mappings, such as\n  `[{'Na': 1, 'Cl': 1}]`, and `--sampling-config-name=csp`. The CSP config omits\n  atomic-number predictor\u002Fcorrector parts because atom types are fixed by the\n  requested composition. Use a CSP-trained local model; the public pretrained\n  catalog is primarily the non-CSP catalog.\n- **Python API:** construct `MatterGenCheckpointInfo`, then `CrystalGenerator`,\n  and call `generate`. Use the exact public signatures in\n  [api-reference.md](references\u002Fapi-reference.md); do not reproduce private\n  model-loading internals in a caller script.\n\n## Safe execution procedure\n\n1. Run the bundled helper with `--help`, then without `--run`, to parse and\n   validate the request. It never downloads a checkpoint or starts sampling by\n   default. Fix errors before using `--run`.\n2. For a Hub model, verify network\u002Fcache permission and choose a small smoke\n   batch first. `MatterGenCheckpointInfo.from_hf_hub(...)` downloads missing\n   files only when explicitly running. For a local model, verify its directory\n   contains `config.yaml` and hydrated checkpoint files; a Git-LFS pointer is\n   not a usable checkpoint.\n3. Set `--record-trajectories=False` (native CLI spelling) when trajectory ZIPs\n   are not needed. Keep the first run small; increase batch size only after a\n   successful memory smoke test. MatterGen selects CUDA when available, then\n   MPS, then CPU; generation is substantially slower and may be impractical on\n   CPU.\n4. Run into a new or deliberately chosen output directory. Never assume a\n   partially written output is a complete sample set.\n5. Inspect the returned\u002Fgenerated structure count and the expected files:\n   `generated_crystals_cif.zip`, `generated_crystals.extxyz`, and, when enabled,\n   `generated_trajectories.zip`. Preserve the resolved checkpoint\u002Fconfig and\n   sampling overrides with the experiment record outside this skill tree.\n\n## Conditioning and guidance rules\n\n`properties_to_condition_on` is a `dict` whose keys must be model-supported\nproperty source IDs. `diffusion_guidance_factor` is the classifier-free guidance\nweight: 0 selects the unconditional score, 1 selects the conditional score,\nand larger nonnegative values increasingly favor the requested condition while\nusually reducing diversity\u002Frealism. Guidance cannot add a missing property\nhead, repair malformed values, or make an incompatible checkpoint conditional.\n\n`sampling_config_overrides` are Hydra override strings. The installed default\nsampling configuration uses 1,000 diffusion steps and atom-number denoising;\nthe installed CSP configuration uses composition conditioning and omits\natomic-number denoising. Use overrides only for known config keys and record\nthem exactly. Avoid changing sampler fields until a baseline completes.\n\n## Failure recovery\n\n- Missing assets: stop and acquire\u002Fhydrate the selected checkpoint (Hub or LFS);\n  do not confuse CUDA readiness with asset readiness.\n- Import\u002Fbackend errors: run the helper's validation without `--run`, then use\n  the environment's public `python -c` import\u002Fhelp probes. Check optional ASE,\n  PyTorch Geometric CUDA wheels, `pymatgen`, Hydra, and MatterSim dependencies;\n  do not silently switch to a different package version.\n- Config\u002Fcondition errors: reduce to an unconditional one-batch run, then add\n  one condition, then multiple conditions. For CSP, switch both the checkpoint\n  and config rather than only changing the CLI flag.\n- Out-of-memory or stalled jobs: lower `batch_size`, lower `num_batches`, turn\n  off trajectories, and retry in a fresh output directory. Do not infer model\n  quality from an interrupted run.\n- Output errors: ensure the output directory is writable and has space; verify\n  ZIP members and frame count before downstream evaluation.\n\n## Bundled operating material\n\n- [API signatures and contracts](references\u002Fapi-reference.md)\n- [CLI, API, CSP, and override workflows](references\u002Fworkflows.md)\n- [Checkpoint catalog and model limits](references\u002Fmodel-overview.md)\n- [Failure diagnosis and recovery matrix](references\u002Ftroubleshooting.md)\n- [Safe parser\u002Fvalidator and explicit-run helper](scripts\u002Fgenerate_materials.py)\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration",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 generation",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill generation",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill generation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002FSKILL.md",[46,52,56,60,64],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fapi-reference.md","REFERENCE","参考文档",5634,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002Freferences\u002Fapi-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fmodel-overview.md",4822,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002Freferences\u002Fmodel-overview.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Ftroubleshooting.md",5265,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002Freferences\u002Ftroubleshooting.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Fworkflows.md",5992,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002Freferences\u002Fworkflows.md",{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},"scripts\u002Fgenerate_materials.py","SCRIPT","脚本",15400,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fgeneration\u002Fscripts\u002Fgenerate_materials.py",[71,79,86,93,100,107],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"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":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"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":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"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":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"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":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"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":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"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",[],[]]