[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M792QTRX7FQWE0V4NF82ZXRC":3,"skill-same:01M792QTRX7FQWE0V4NF82ZXRC":71},{"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":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":46},"01M792QTRX7FQWE0V4NF82ZXRC","toolkit",null,"Build, encode, decode, sample, inspect, optimize, and reload PhySO symbolic programs without running SR orchestration.","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\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: toolkit\ndescription: \"Build, encode, decode, sample, inspect, optimize, and reload PhySO\n  symbolic programs without running SR orchestration.\"\nread_when: \"Use when a task concerns physo.toolkit, expression prefix encodings,\n  token libraries, free constants, program trees, display, monitoring, or\n  Pareto\u002Fresult loading.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# PhySO toolkit\n\n## Role\n\nUse this sub-skill for the low-level symbolic-expression layer of PhySO. It\nturns a token library and prefix expressions into executable `Program` or\n`VectPrograms` objects, and provides safe routes for sampling, inspecting,\noptimizing, displaying, and reloading those objects. The verified baseline is\nPhySO 1.2.0 in a CPU-only Python environment. Import-time warnings about\nmissing system LaTeX are expected and do not invalidate plain-text workflows.\n\n## Boundaries\n\n### Included\n\n- `physo.toolkit.get_library`, `get_expressions`, `get_prior`, and\n  `sample_random_expressions`.\n- `Library.encode`\u002F`decode`, prefix-token validation, one-hot encodings, and\n  length\u002Fstructural priors.\n- `Dataset`, `Token` and its public typed subclasses, `Library`, `Program`,\n  `VectPrograms`, `Cursor`, `FreeConstantsTable`, and differentiable\n  `candidate_wrapper` usage at the expression layer.\n- CPU-safe execution, autograd inspection, free-constant optimization, class versus\n  realization-specific constants, tree relationships, and plain or optional\n  display representations.\n- `RunLogger`, Pareto-front inspection, `physo.read_pareto_csv`,\n  `physo.read_pareto_pkl`, and `physo.load_expr`.\n\n### Excluded\n\n- `physo.SR` and `physo.ClassSR` orchestration, run presets, stopping rules,\n  and dataset\u002Ftask selection. Route those questions to\n  [the SR sibling](..\u002Fsr\u002FSKILL.md) or [the Class SR sibling](..\u002Fclass-sr\u002FSKILL.md).\n- Feynman\u002FClass benchmark problem definitions, selectors, and equivalence\n  workflows. Route them to [the benchmarks sibling](..\u002Fbenchmarks\u002FSKILL.md).\n- Maintainer benchmark runners, cluster jobfile generation, long training,\n  plotting-heavy analysis, CUDA-specific claims, and undocumented private\n  helpers.\n\n## Verified entry points\n\n| Need | Start here | Result |\n|---|---|---|\n| Define variables, constants, operations, units, and free constants | `physo.toolkit.get_library(...)` | A `physo.physym.library.Library` |\n| Encode readable prefix names | `library.encode(expressions_str, one_hot=False)` | One integer-token list per expression; `one_hot=True` returns one-hot matrices |\n| Decode integer prefixes | `library.decode(expressions_idx, n_realizations=1)` | A `VectPrograms` batch with padded invalid placeholders |\n| Make an empty program batch | `physo.toolkit.get_expressions(batch_size, max_length, library, ...)` | A `VectPrograms` ready for `append` or `set_programs` |\n| Build priors for an existing batch | `physo.toolkit.get_prior(priors_config, max_length, expressions, library)` | A prior collection callable |\n| Generate random valid candidates | `physo.toolkit.sample_random_expressions(...)` | A sampled `VectPrograms` batch |\n| Inspect one candidate | `batch.get_prog(i)` | A `Program` with execution, constants, text, SymPy, and save methods |\n| Inspect a batch\u002Ftree | `batch.status()`, `batch.tokens`, `Cursor`, family methods | Names, arities, units, parent\u002Fchild\u002Fsibling\u002Fancestor relations |\n| Fit constants | `program.optimize_constants(...)`, `batch.batch_optimize_constants(...)` | Updated constant tables and a loss history for the single-program route |\n| Reload results | `physo.read_pareto_pkl`, `physo.read_pareto_csv`, `physo.load_expr` | Programs or SymPy expressions, subject to the CSV\u002FClass SR caveat |\n\nSee [API reference](references\u002Fapi-reference.md) for source-verified\nsignatures, [workflows](references\u002Fworkflows.md) for compact recipes, and the\n[PhySO root skill](..\u002F..\u002FSKILL.md) for install\u002Fbackend routing.\n\n## Operating workflow\n\n1. **Fix the data contract first.** For each realization, use `X` with shape\n   `(n_dim, n_samples)` and `y` with shape `(n_samples,)`, both floating-point\n   tensors or safely convertible arrays. A `Dataset` receives lists\n   `multi_X` and `multi_y`, even for one realization (`[X]`, `[y]`). Keep all\n   `X`, `y`, and weights on the same device.\n2. **Build the smallest library.** Set `X_names`, `y_name`, aligned fixed and\n   free-constant arrays, and an explicit `op_names` list. Use protected\n   operations unless you deliberately need unprotected numerical behavior.\n   With units, every terminal units vector must be aligned and consistent.\n3. **Encode only choosable names.** Write expressions in prefix order, e.g.\n   `['mul', 'c', 'x']`. Check `library.choosable_names`; placeholders such as\n   the superparent, dummy, and invalid token are not encodable user tokens.\n4. **Decode and validate.** Pass a batch-shaped list of Python `int` indices to\n   `Library.decode`. Inspect `batch.status()`, `batch.n_lengths`,\n   `batch.is_complete`, and `batch.get_infix_str(i)` before executing. A prefix\n   expression must be a complete tree: `len(tokens) - sum(token.arity) == 1`.\n5. **Sample under bounded priors.** Use a small `batch_size` and `max_length`\n   for exploration. If `priors_config` is supplied, include a\n   `HardLengthPrior`; its `max_length` controls the sampler and takes\n   precedence over the top-level value. Seed NumPy and PyTorch when a\n   reproducible fixture matters.\n6. **Execute and inspect.** Get a `Program` with `get_prog`, call it with a\n   tensor, or use `execute(..., i_realization=..., n_samples_per_dataset=...)`\n   for realization-specific constants. Prefer `get_infix_str`,\n   `get_infix_pretty`, and `get_infix_sympy` for non-graph output.\n7. **Optimize constants only after shapes are stable.** Use differentiable\n   PyTorch operations and a float target. If using a `candidate_wrapper`, it\n   must be callable as `(func, X) -> wrapped_output` and remain differentiable.\n   `Program.optimize_constants` uses\n   MSE plus LBFGS by default; inspect `history`, `free_consts.class_values`,\n   `free_consts.spe_values`, `is_opti`, and `opti_steps`.\n8. **Save and reload deliberately.** Use the run visualizer\u002Flogger paths to\n   identify the generated Pareto CSV\u002FPKL. Prefer PKL for programs and Class SR\n   expressions with spe constants. Use CSV loading only for non-spe Pareto\n   fronts and pass a SymPy symbol dictionary when assumptions matter.\n9. **Fall back when optional display tools are absent.** Missing `latex`,\n   `pygraphviz`, `dot2tex`, `pdflatex`, `pdf2image`, or Pillow affects images\n   and tree rendering, not prefix encoding or plain text. Do not turn a display\n   dependency into a core-workflow failure.\n\n## Input\u002Foutput contracts\n\n- **Library inputs:** `X_names` length must match the first dimension of data;\n  units rows must align with their names and use at most seven dimensions.\n  `fixed_consts_units`, `free_consts_units`, and spe-constant units align with\n  their corresponding value\u002Fname arrays. For the toolkit convenience API,\n  `free_consts_init_val` and `spe_free_consts_init_val` are array-like values\n  aligned with names; low-level tokenization may use name-to-value dicts.\n- **Prefix inputs:** outer sequence is a batch; each inner sequence is a list\n  or NumPy array of strings for `encode`, or Python integers for `decode`.\n  Variable expression lengths are accepted by `decode` and padded internally.\n  One-hot output is for model input and is not a valid direct input to\n  `decode`.\n- **Program inputs:** `X` is `(n_dim, n_samples)`, target and weights are\n  `(n_samples,)`. For flattened multiple realizations, concatenate datasets in\n  realization order and pass `n_samples_per_dataset` with the matching counts.\n- **Constant outputs:** `class_values` has shape\n  `(batch_size, n_class_free_const)`; `spe_values` has shape\n  `(batch_size, n_spe_free_const, n_realizations)`. A scalar spe initializer is\n  broadcast; a vector must have exactly `n_realizations` entries.\n- **Monitoring outputs:** `RunLogger.get_pareto_front()` returns\n  `(complexities, programs, rewards, rmse)`. `read_pareto_pkl` returns a list\n  of `Program` objects; `read_pareto_csv` returns SymPy expressions, or a\n  `(expressions, DataFrame)` tuple when `return_df=True`.\n\n## Verification and native evidence\n\nThe related native candidates are:\n\n- `physo.toolkit.tests.codec_UnitTest`\n- `physo.toolkit.tests.random_sampler_UnitTest`\n- `physo.physym.tests.program_UnitTest`\n- `physo.physym.tests.vect_programs_UnitTest`\n- `physo.physym.tests.library_UnitTest`\n- `physo.physym.tests.token_UnitTest`\n- `physo.physym.tests.tokenize_UnitTest`\n- `physo.physym.tests.dataset_UnitTest`\n- `physo.physym.tests.free_const_UnitTest`\n- `physo.physym.tests.program_display_UnitTest`\n- `physo.learn.tests.monitoring_UnitTest`\n\nRun the bundled CPU-safe checks before using a generated result:\n\n```bash\npython skills\u002Fdisco\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fcheck_toolkit_roundtrip.py\npython skills\u002Fdisco\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fsmoke_toolkit.py\n```\n\nThese scripts do not access the network, plot, or launch a long search. They\nexercise encoding\u002Fdecoding, an expected unknown-token failure, bounded random\nsampling, `Dataset` shape handling, and a small LBFGS constant fit.\n\n## Troubleshooting\n\nUse [troubleshooting](references\u002Ftroubleshooting.md) when an error is not\nimmediately explained. In particular, check exact library names and token\narity before changing code; check spe initializer and flattened-data shapes\nbefore changing an optimizer; and use text\u002FSymPy output before installing\noptional renderers. If the request asks for an SR\u002FClass SR run or a benchmark\nproblem loader, hand it to the sibling skill instead of extending this route.\n","# PhySO toolkit\n\n## Role\n\nUse this sub-skill for the low-level symbolic-expression layer of PhySO. It\nturns a token library and prefix expressions into executable `Program` or\n`VectPrograms` objects, and provides safe routes for sampling, inspecting,\noptimizing, displaying, and reloading those objects. The verified baseline is\nPhySO 1.2.0 in a CPU-only Python environment. Import-time warnings about\nmissing system LaTeX are expected and do not invalidate plain-text workflows.\n\n## Boundaries\n\n### Included\n\n- `physo.toolkit.get_library`, `get_expressions`, `get_prior`, and\n  `sample_random_expressions`.\n- `Library.encode`\u002F`decode`, prefix-token validation, one-hot encodings, and\n  length\u002Fstructural priors.\n- `Dataset`, `Token` and its public typed subclasses, `Library`, `Program`,\n  `VectPrograms`, `Cursor`, `FreeConstantsTable`, and differentiable\n  `candidate_wrapper` usage at the expression layer.\n- CPU-safe execution, autograd inspection, free-constant optimization, class versus\n  realization-specific constants, tree relationships, and plain or optional\n  display representations.\n- `RunLogger`, Pareto-front inspection, `physo.read_pareto_csv`,\n  `physo.read_pareto_pkl`, and `physo.load_expr`.\n\n### Excluded\n\n- `physo.SR` and `physo.ClassSR` orchestration, run presets, stopping rules,\n  and dataset\u002Ftask selection. Route those questions to\n  [the SR sibling](..\u002Fsr\u002FSKILL.md) or [the Class SR sibling](..\u002Fclass-sr\u002FSKILL.md).\n- Feynman\u002FClass benchmark problem definitions, selectors, and equivalence\n  workflows. Route them to [the benchmarks sibling](..\u002Fbenchmarks\u002FSKILL.md).\n- Maintainer benchmark runners, cluster jobfile generation, long training,\n  plotting-heavy analysis, CUDA-specific claims, and undocumented private\n  helpers.\n\n## Verified entry points\n\n| Need | Start here | Result |\n|---|---|---|\n| Define variables, constants, operations, units, and free constants | `physo.toolkit.get_library(...)` | A `physo.physym.library.Library` |\n| Encode readable prefix names | `library.encode(expressions_str, one_hot=False)` | One integer-token list per expression; `one_hot=True` returns one-hot matrices |\n| Decode integer prefixes | `library.decode(expressions_idx, n_realizations=1)` | A `VectPrograms` batch with padded invalid placeholders |\n| Make an empty program batch | `physo.toolkit.get_expressions(batch_size, max_length, library, ...)` | A `VectPrograms` ready for `append` or `set_programs` |\n| Build priors for an existing batch | `physo.toolkit.get_prior(priors_config, max_length, expressions, library)` | A prior collection callable |\n| Generate random valid candidates | `physo.toolkit.sample_random_expressions(...)` | A sampled `VectPrograms` batch |\n| Inspect one candidate | `batch.get_prog(i)` | A `Program` with execution, constants, text, SymPy, and save methods |\n| Inspect a batch\u002Ftree | `batch.status()`, `batch.tokens`, `Cursor`, family methods | Names, arities, units, parent\u002Fchild\u002Fsibling\u002Fancestor relations |\n| Fit constants | `program.optimize_constants(...)`, `batch.batch_optimize_constants(...)` | Updated constant tables and a loss history for the single-program route |\n| Reload results | `physo.read_pareto_pkl`, `physo.read_pareto_csv`, `physo.load_expr` | Programs or SymPy expressions, subject to the CSV\u002FClass SR caveat |\n\nSee [API reference](references\u002Fapi-reference.md) for source-verified\nsignatures, [workflows](references\u002Fworkflows.md) for compact recipes, and the\n[PhySO root skill](..\u002F..\u002FSKILL.md) for install\u002Fbackend routing.\n\n## Operating workflow\n\n1. **Fix the data contract first.** For each realization, use `X` with shape\n   `(n_dim, n_samples)` and `y` with shape `(n_samples,)`, both floating-point\n   tensors or safely convertible arrays. A `Dataset` receives lists\n   `multi_X` and `multi_y`, even for one realization (`[X]`, `[y]`). Keep all\n   `X`, `y`, and weights on the same device.\n2. **Build the smallest library.** Set `X_names`, `y_name`, aligned fixed and\n   free-constant arrays, and an explicit `op_names` list. Use protected\n   operations unless you deliberately need unprotected numerical behavior.\n   With units, every terminal units vector must be aligned and consistent.\n3. **Encode only choosable names.** Write expressions in prefix order, e.g.\n   `['mul', 'c', 'x']`. Check `library.choosable_names`; placeholders such as\n   the superparent, dummy, and invalid token are not encodable user tokens.\n4. **Decode and validate.** Pass a batch-shaped list of Python `int` indices to\n   `Library.decode`. Inspect `batch.status()`, `batch.n_lengths`,\n   `batch.is_complete`, and `batch.get_infix_str(i)` before executing. A prefix\n   expression must be a complete tree: `len(tokens) - sum(token.arity) == 1`.\n5. **Sample under bounded priors.** Use a small `batch_size` and `max_length`\n   for exploration. If `priors_config` is supplied, include a\n   `HardLengthPrior`; its `max_length` controls the sampler and takes\n   precedence over the top-level value. Seed NumPy and PyTorch when a\n   reproducible fixture matters.\n6. **Execute and inspect.** Get a `Program` with `get_prog`, call it with a\n   tensor, or use `execute(..., i_realization=..., n_samples_per_dataset=...)`\n   for realization-specific constants. Prefer `get_infix_str`,\n   `get_infix_pretty`, and `get_infix_sympy` for non-graph output.\n7. **Optimize constants only after shapes are stable.** Use differentiable\n   PyTorch operations and a float target. If using a `candidate_wrapper`, it\n   must be callable as `(func, X) -> wrapped_output` and remain differentiable.\n   `Program.optimize_constants` uses\n   MSE plus LBFGS by default; inspect `history`, `free_consts.class_values`,\n   `free_consts.spe_values`, `is_opti`, and `opti_steps`.\n8. **Save and reload deliberately.** Use the run visualizer\u002Flogger paths to\n   identify the generated Pareto CSV\u002FPKL. Prefer PKL for programs and Class SR\n   expressions with spe constants. Use CSV loading only for non-spe Pareto\n   fronts and pass a SymPy symbol dictionary when assumptions matter.\n9. **Fall back when optional display tools are absent.** Missing `latex`,\n   `pygraphviz`, `dot2tex`, `pdflatex`, `pdf2image`, or Pillow affects images\n   and tree rendering, not prefix encoding or plain text. Do not turn a display\n   dependency into a core-workflow failure.\n\n## Input\u002Foutput contracts\n\n- **Library inputs:** `X_names` length must match the first dimension of data;\n  units rows must align with their names and use at most seven dimensions.\n  `fixed_consts_units`, `free_consts_units`, and spe-constant units align with\n  their corresponding value\u002Fname arrays. For the toolkit convenience API,\n  `free_consts_init_val` and `spe_free_consts_init_val` are array-like values\n  aligned with names; low-level tokenization may use name-to-value dicts.\n- **Prefix inputs:** outer sequence is a batch; each inner sequence is a list\n  or NumPy array of strings for `encode`, or Python integers for `decode`.\n  Variable expression lengths are accepted by `decode` and padded internally.\n  One-hot output is for model input and is not a valid direct input to\n  `decode`.\n- **Program inputs:** `X` is `(n_dim, n_samples)`, target and weights are\n  `(n_samples,)`. For flattened multiple realizations, concatenate datasets in\n  realization order and pass `n_samples_per_dataset` with the matching counts.\n- **Constant outputs:** `class_values` has shape\n  `(batch_size, n_class_free_const)`; `spe_values` has shape\n  `(batch_size, n_spe_free_const, n_realizations)`. A scalar spe initializer is\n  broadcast; a vector must have exactly `n_realizations` entries.\n- **Monitoring outputs:** `RunLogger.get_pareto_front()` returns\n  `(complexities, programs, rewards, rmse)`. `read_pareto_pkl` returns a list\n  of `Program` objects; `read_pareto_csv` returns SymPy expressions, or a\n  `(expressions, DataFrame)` tuple when `return_df=True`.\n\n## Verification and native evidence\n\nThe related native candidates are:\n\n- `physo.toolkit.tests.codec_UnitTest`\n- `physo.toolkit.tests.random_sampler_UnitTest`\n- `physo.physym.tests.program_UnitTest`\n- `physo.physym.tests.vect_programs_UnitTest`\n- `physo.physym.tests.library_UnitTest`\n- `physo.physym.tests.token_UnitTest`\n- `physo.physym.tests.tokenize_UnitTest`\n- `physo.physym.tests.dataset_UnitTest`\n- `physo.physym.tests.free_const_UnitTest`\n- `physo.physym.tests.program_display_UnitTest`\n- `physo.learn.tests.monitoring_UnitTest`\n\nRun the bundled CPU-safe checks before using a generated result:\n\n```bash\npython skills\u002Fdisco\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fcheck_toolkit_roundtrip.py\npython skills\u002Fdisco\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fsmoke_toolkit.py\n```\n\nThese scripts do not access the network, plot, or launch a long search. They\nexercise encoding\u002Fdecoding, an expected unknown-token failure, bounded random\nsampling, `Dataset` shape handling, and a small LBFGS constant fit.\n\n## Troubleshooting\n\nUse [troubleshooting](references\u002Ftroubleshooting.md) when an error is not\nimmediately explained. In particular, check exact library names and token\narity before changing code; check spe initializer and flattened-data shapes\nbefore changing an optimizer; and use text\u002FSymPy output before installing\noptional renderers. If the request asks for an SR\u002FClass SR run or a benchmark\nproblem loader, hand it to the sibling skill instead of extending this route.\n",{"name":5,"license":14,"metadata":23,"read_when":25,"description":7,"disable-model-invocation":26},{"disco-role":24},"operating","Use when a task concerns physo.toolkit, expression prefix encodings, token libraries, free constants, program trees, display, monitoring, or Pareto\u002Fresult loading.",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":26},[36,39,42],{"tool":37,"command":38},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill toolkit",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill toolkit",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill toolkit","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002FSKILL.md",[47,53,57,61,67],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fapi-reference.md","REFERENCE","参考文档",11830,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Ftroubleshooting.md",8071,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Freferences\u002Ftroubleshooting.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fworkflows.md",10500,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Freferences\u002Fworkflows.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Fcheck_toolkit_roundtrip.py","SCRIPT","脚本",2351,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fcheck_toolkit_roundtrip.py",{"path":68,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":69,"sourceHtmlUrl":70},"scripts\u002Fsmoke_toolkit.py",2790,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fphyso\u002Fsub-skills\u002Ftoolkit\u002Fscripts\u002Fsmoke_toolkit.py",[72,80,87,94,101,108],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"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":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"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":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"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":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"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":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"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":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":76,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"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",[],[]]