[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-qiskit-to-cudaq":3,"mdc--a8m9fv-key":38,"related-repo-nvidia-qiskit-to-cudaq":568,"related-org-nvidia-qiskit-to-cudaq":577},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":33,"sourceUrl":36,"mdContent":37},"qiskit-to-cudaq","port Qiskit circuits to CUDA-Q kernels","Use when porting Qiskit Python circuits to CUDA-Q kernels while preserving algorithms and validation fidelity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,14,17],{"name":9,"slug":8,"type":13},"tag",{"name":15,"slug":16,"type":13},"Python","python",{"name":18,"slug":19,"type":13},"Quantum Computing","quantum-computing",1089,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fcuda-quantum","2026-07-14T05:31:30.970327","Apache-2.0",420,[26,27,16,28,29,19,30,31,32],"cpp","hacktoberfest","quantum","quantum-algorithms","quantum-machine-learning","quantum-programming-language","unitaryhack",{"repoUrl":21,"stars":20,"forks":24,"topics":34,"description":35},[26,27,16,28,29,19,30,31,32],"C++ and Python support for the CUDA Quantum programming model for heterogeneous quantum-classical workflows","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fcuda-quantum\u002Ftree\u002FHEAD\u002Fskills\u002Fqiskit-to-cudaq","---\nname: \"qiskit-to-cudaq\"\ntitle: \"Qiskit to CUDA-Q\"\ndescription: \"Use when porting Qiskit Python circuits to CUDA-Q kernels while preserving algorithms and validation fidelity.\"\nversion: \"1.0.0\"\nauthor: \"CUDA-Q Team \u003Ccuda-quantum@nvidia.com>\"\ntags: [cuda-quantum, quantum-computing, qiskit, porting, migration, kernels, nvidia]\ntools: [Read, Glob, Grep]\nlicense: \"Apache-2.0\"\ncompatibility: \"Python 3.10+\"\nmetadata:\n    author: \"CUDA-Q Team \u003Ccuda-quantum@nvidia.com>\"\n    short-description: \"Port Qiskit circuits to CUDA-Q\"\n    tags:\n        - cuda-quantum\n        - quantum-computing\n        - qiskit\n        - porting\n        - migration\n        - nvidia\n    languages:\n        - python\n    domain: \"quantum\"\n---\n\n# Qiskit to CUDA-Q\n\n## Purpose\n\nUse this skill to port Qiskit Python code, or code with Qiskit-style circuit\nconstruction, to CUDA-Q Python kernels. The goal is a framework-free CUDA-Q port\nthat preserves the source quantum algorithm, matches source behavior at small\ntest sizes, and documents any unavoidable CUDA-Q limitations.\n\n## Prerequisites\n\n- Python 3.10+.\n- CUDA-Q installed in the target environment. Check the runtime with:\n  `python -c \"import cudaq; print(getattr(cudaq, '__version__', 'unknown'))\"`.\n- Access to the source implementation and a way to run or inspect its expected\n  behavior.\n- For validation against Qiskit, Qiskit\u002FAer must be installed in the validation\n  environment. The final CUDA-Q port itself must not require Qiskit.\n- When using CUDA-Q documentation or repository MCP connectors, verify the\n  connector is available before relying on it; otherwise use local docs or the\n  source tree.\n- When debugging and the installed CUDA-Q version differs from the latest\n  documentation, review relevant documentation or source changes before\n  treating a behavior difference as a porting bug.\n\n## Workflow\n\n1. Read the source circuit construction and identify the exact algorithm,\n   qubit\u002Fregister layout, measurement behavior, and any framework helpers.\n2. Preserve the high-level quantum algorithm. Do not replace mid-circuit\n   measurement, QPE structure, oracle definitions, or decomposition strategy\n   without explicit user permission.\n3. Select the CUDA-Q execution pattern:\n   - Use `cudaq.sample` for final-measurement sampling.\n   - Use `cudaq.run` when mid-circuit measurement values must be returned or\n     used per shot.\n   - Use runtime-argument kernels instead of generated per-size kernels unless\n     CUDA-Q requires a fixed-length return shape.\n4. Translate gates and subcircuits. For detailed gate mappings, ordering rules,\n   precision guidance, and helper-extraction patterns, read\n   [references\u002Fporting-reference.md](references\u002Fporting-reference.md).\n5. Remove runtime source-framework dependencies from the CUDA-Q port. Extract\n   pure helpers into framework-free modules.\n6. Validate with small deterministic inputs before scaling. Compare raw count\n   keys and distributions, not just aggregate fidelity.\n7. Re-run any previously failing configurations after every fix.\n\n## Core Rules\n\n- Keep the source algorithm intact unless the user approves a change.\n- Do not introduce fixed qubit caps, fixed control arities, or source-framework\n  imports unless they are genuinely unavoidable and documented.\n- Prefer native CUDA-Q gates (`r1.ctrl`, `x.ctrl`, `swap.ctrl`, etc.) over\n  transpiling through Qiskit.\n- Keep bit-order conversion at the port boundary: allocation order,\n  measurement return list, or final count-key formatting.\n- Match floating-point precision when comparing CUDA-Q and Qiskit results if\n  fidelity differences matter.\n- Accept source flags that become no-ops in CUDA-Q when doing so preserves\n  source-compatible behavior.\n\n## When to Read the Reference\n\nRead [references\u002Fporting-reference.md](references\u002Fporting-reference.md) when\nyou need any of the following:\n\n- Qiskit-to-CUDA-Q gate translation table.\n- Bit-ordering and count-key conventions.\n- CUDA-Q fp32 vs Qiskit fp64 precision implications.\n- Pure-Python helper extraction and import-blocker validation.\n- Recursive-constructor emitters or gate-recorder patterns.\n- Detailed port validation checklist and external CUDA-Q references.\n\n## Limitations\n\n- Guidance targets CUDA-Q 0.14\u002F0.15 decorator-mode Python APIs. Re-check\n  behavior against the installed CUDA-Q version for version-sensitive features.\n- Some CUDA-Q kernel-language constructs are constrained compared with normal\n  Python; use the companion `cudaq-guide` skill for core CUDA-Q authoring\n  constraints and shared kernel patterns.\n- CUDA-Q and Qiskit differ in default precision and count-key display order.\n  Apparent fidelity or bitstring mismatches may be convention differences.\n- Hardware-target behavior, available backends, and target options depend on\n  the local CUDA-Q installation.\n- This skill does not guarantee equivalent performance; it focuses on\n  correctness-preserving ports.\n\n## Troubleshooting\n\nUse this format when diagnosing failures:\n\n- **Error:** `ModuleNotFoundError: qiskit` from a CUDA-Q path.\n  **Cause:** The port still imports the source framework.\n  **Solution:** Move pure helpers into a framework-free module and verify with\n  the import-blocker pattern in the reference.\n\n- **Error:** Fidelity looks plausible but raw keys are reversed.\n  **Cause:** Qiskit and CUDA-Q count-key ordering differ.\n  **Solution:** Fix allocation, return-list order, or formatting at the port\n  boundary. Do not alter the algorithm.\n\n- **Error:** Deep-circuit fidelity differs between frameworks.\n  **Cause:** CUDA-Q and Qiskit may be using different floating-point precision.\n  **Solution:** Match precision before comparing, then rerun the smallest\n  failing deterministic case.\n\n- **Error:** A multi-controlled operation works for small controls but fails or\n  silently changes behavior at higher arity.\n  **Cause:** The port used a fixed-arity dispatcher.\n  **Solution:** Use CUDA-Q control-list patterns for arbitrary arity.\n\n- **Error:** MCP documentation or repository lookup fails.\n  **Cause:** Connector unavailable, stale, or transiently failing.\n  **Solution:** Verify the connector\u002Fresource list, retry transient failures\n  once, then fall back to local docs\u002Fsource or official CUDA-Q docs. Do not\n  change the port based on unverified MCP results.\n\n- **Error:** CUDA-Q behavior conflicts with documentation while debugging.\n  **Cause:** The installed CUDA-Q version may differ from the latest\n  documentation.\n  **Solution:** Check `cudaq.__version__`, then review relevant documentation or\n  source changes between the installed version and latest before changing the\n  port.\n\n## References\n\n- [Detailed porting reference](references\u002Fporting-reference.md)\n- Companion skill: `cudaq-guide` (`\u002Fcudaq-guide author`) for CUDA-Q authoring\n  patterns, kernel-language constraints, execution APIs, and debugging workflow.\n",{"data":39,"body":58},{"name":4,"title":40,"description":6,"version":41,"author":42,"tags":43,"tools":49,"license":23,"compatibility":53,"metadata":54},"Qiskit to CUDA-Q","1.0.0","CUDA-Q Team \u003Ccuda-quantum@nvidia.com>",[44,19,45,46,47,48,8],"cuda-quantum","qiskit","porting","migration","kernels",[50,51,52],"Read","Glob","Grep","Python 3.10+",{"author":42,"short-description":55,"tags":56,"languages":57,"domain":28},"Port Qiskit circuits to CUDA-Q",[44,19,45,46,47,8],[16],{"type":59,"children":60},"root",[61,69,76,82,88,132,138,217,223,279,285,296,329,335,371,377,382,531,537],{"type":62,"tag":63,"props":64,"children":66},"element","h1",{"id":65},"qiskit-to-cuda-q",[67],{"type":68,"value":40},"text",{"type":62,"tag":70,"props":71,"children":73},"h2",{"id":72},"purpose",[74],{"type":68,"value":75},"Purpose",{"type":62,"tag":77,"props":78,"children":79},"p",{},[80],{"type":68,"value":81},"Use this skill to port Qiskit Python code, or code with Qiskit-style circuit\nconstruction, to CUDA-Q Python kernels. The goal is a framework-free CUDA-Q port\nthat preserves the source quantum algorithm, matches source behavior at small\ntest sizes, and documents any unavoidable CUDA-Q limitations.",{"type":62,"tag":70,"props":83,"children":85},{"id":84},"prerequisites",[86],{"type":68,"value":87},"Prerequisites",{"type":62,"tag":89,"props":90,"children":91},"ul",{},[92,98,112,117,122,127],{"type":62,"tag":93,"props":94,"children":95},"li",{},[96],{"type":68,"value":97},"Python 3.10+.",{"type":62,"tag":93,"props":99,"children":100},{},[101,103,110],{"type":68,"value":102},"CUDA-Q installed in the target environment. Check the runtime with:\n",{"type":62,"tag":104,"props":105,"children":107},"code",{"className":106},[],[108],{"type":68,"value":109},"python -c \"import cudaq; print(getattr(cudaq, '__version__', 'unknown'))\"",{"type":68,"value":111},".",{"type":62,"tag":93,"props":113,"children":114},{},[115],{"type":68,"value":116},"Access to the source implementation and a way to run or inspect its expected\nbehavior.",{"type":62,"tag":93,"props":118,"children":119},{},[120],{"type":68,"value":121},"For validation against Qiskit, Qiskit\u002FAer must be installed in the validation\nenvironment. The final CUDA-Q port itself must not require Qiskit.",{"type":62,"tag":93,"props":123,"children":124},{},[125],{"type":68,"value":126},"When using CUDA-Q documentation or repository MCP connectors, verify the\nconnector is available before relying on it; otherwise use local docs or the\nsource tree.",{"type":62,"tag":93,"props":128,"children":129},{},[130],{"type":68,"value":131},"When debugging and the installed CUDA-Q version differs from the latest\ndocumentation, review relevant documentation or source changes before\ntreating a behavior difference as a porting bug.",{"type":62,"tag":70,"props":133,"children":135},{"id":134},"workflow",[136],{"type":68,"value":137},"Workflow",{"type":62,"tag":139,"props":140,"children":141},"ol",{},[142,147,152,190,202,207,212],{"type":62,"tag":93,"props":143,"children":144},{},[145],{"type":68,"value":146},"Read the source circuit construction and identify the exact algorithm,\nqubit\u002Fregister layout, measurement behavior, and any framework helpers.",{"type":62,"tag":93,"props":148,"children":149},{},[150],{"type":68,"value":151},"Preserve the high-level quantum algorithm. Do not replace mid-circuit\nmeasurement, QPE structure, oracle definitions, or decomposition strategy\nwithout explicit user permission.",{"type":62,"tag":93,"props":153,"children":154},{},[155,157],{"type":68,"value":156},"Select the CUDA-Q execution pattern:\n",{"type":62,"tag":89,"props":158,"children":159},{},[160,173,185],{"type":62,"tag":93,"props":161,"children":162},{},[163,165,171],{"type":68,"value":164},"Use ",{"type":62,"tag":104,"props":166,"children":168},{"className":167},[],[169],{"type":68,"value":170},"cudaq.sample",{"type":68,"value":172}," for final-measurement sampling.",{"type":62,"tag":93,"props":174,"children":175},{},[176,177,183],{"type":68,"value":164},{"type":62,"tag":104,"props":178,"children":180},{"className":179},[],[181],{"type":68,"value":182},"cudaq.run",{"type":68,"value":184}," when mid-circuit measurement values must be returned or\nused per shot.",{"type":62,"tag":93,"props":186,"children":187},{},[188],{"type":68,"value":189},"Use runtime-argument kernels instead of generated per-size kernels unless\nCUDA-Q requires a fixed-length return shape.",{"type":62,"tag":93,"props":191,"children":192},{},[193,195,201],{"type":68,"value":194},"Translate gates and subcircuits. For detailed gate mappings, ordering rules,\nprecision guidance, and helper-extraction patterns, read\n",{"type":62,"tag":196,"props":197,"children":199},"a",{"href":198},"references\u002Fporting-reference.md",[200],{"type":68,"value":198},{"type":68,"value":111},{"type":62,"tag":93,"props":203,"children":204},{},[205],{"type":68,"value":206},"Remove runtime source-framework dependencies from the CUDA-Q port. Extract\npure helpers into framework-free modules.",{"type":62,"tag":93,"props":208,"children":209},{},[210],{"type":68,"value":211},"Validate with small deterministic inputs before scaling. Compare raw count\nkeys and distributions, not just aggregate fidelity.",{"type":62,"tag":93,"props":213,"children":214},{},[215],{"type":68,"value":216},"Re-run any previously failing configurations after every fix.",{"type":62,"tag":70,"props":218,"children":220},{"id":219},"core-rules",[221],{"type":68,"value":222},"Core Rules",{"type":62,"tag":89,"props":224,"children":225},{},[226,231,236,264,269,274],{"type":62,"tag":93,"props":227,"children":228},{},[229],{"type":68,"value":230},"Keep the source algorithm intact unless the user approves a change.",{"type":62,"tag":93,"props":232,"children":233},{},[234],{"type":68,"value":235},"Do not introduce fixed qubit caps, fixed control arities, or source-framework\nimports unless they are genuinely unavoidable and documented.",{"type":62,"tag":93,"props":237,"children":238},{},[239,241,247,249,255,256,262],{"type":68,"value":240},"Prefer native CUDA-Q gates (",{"type":62,"tag":104,"props":242,"children":244},{"className":243},[],[245],{"type":68,"value":246},"r1.ctrl",{"type":68,"value":248},", ",{"type":62,"tag":104,"props":250,"children":252},{"className":251},[],[253],{"type":68,"value":254},"x.ctrl",{"type":68,"value":248},{"type":62,"tag":104,"props":257,"children":259},{"className":258},[],[260],{"type":68,"value":261},"swap.ctrl",{"type":68,"value":263},", etc.) over\ntranspiling through Qiskit.",{"type":62,"tag":93,"props":265,"children":266},{},[267],{"type":68,"value":268},"Keep bit-order conversion at the port boundary: allocation order,\nmeasurement return list, or final count-key formatting.",{"type":62,"tag":93,"props":270,"children":271},{},[272],{"type":68,"value":273},"Match floating-point precision when comparing CUDA-Q and Qiskit results if\nfidelity differences matter.",{"type":62,"tag":93,"props":275,"children":276},{},[277],{"type":68,"value":278},"Accept source flags that become no-ops in CUDA-Q when doing so preserves\nsource-compatible behavior.",{"type":62,"tag":70,"props":280,"children":282},{"id":281},"when-to-read-the-reference",[283],{"type":68,"value":284},"When to Read the Reference",{"type":62,"tag":77,"props":286,"children":287},{},[288,290,294],{"type":68,"value":289},"Read ",{"type":62,"tag":196,"props":291,"children":292},{"href":198},[293],{"type":68,"value":198},{"type":68,"value":295}," when\nyou need any of the following:",{"type":62,"tag":89,"props":297,"children":298},{},[299,304,309,314,319,324],{"type":62,"tag":93,"props":300,"children":301},{},[302],{"type":68,"value":303},"Qiskit-to-CUDA-Q gate translation table.",{"type":62,"tag":93,"props":305,"children":306},{},[307],{"type":68,"value":308},"Bit-ordering and count-key conventions.",{"type":62,"tag":93,"props":310,"children":311},{},[312],{"type":68,"value":313},"CUDA-Q fp32 vs Qiskit fp64 precision implications.",{"type":62,"tag":93,"props":315,"children":316},{},[317],{"type":68,"value":318},"Pure-Python helper extraction and import-blocker validation.",{"type":62,"tag":93,"props":320,"children":321},{},[322],{"type":68,"value":323},"Recursive-constructor emitters or gate-recorder patterns.",{"type":62,"tag":93,"props":325,"children":326},{},[327],{"type":68,"value":328},"Detailed port validation checklist and external CUDA-Q references.",{"type":62,"tag":70,"props":330,"children":332},{"id":331},"limitations",[333],{"type":68,"value":334},"Limitations",{"type":62,"tag":89,"props":336,"children":337},{},[338,343,356,361,366],{"type":62,"tag":93,"props":339,"children":340},{},[341],{"type":68,"value":342},"Guidance targets CUDA-Q 0.14\u002F0.15 decorator-mode Python APIs. Re-check\nbehavior against the installed CUDA-Q version for version-sensitive features.",{"type":62,"tag":93,"props":344,"children":345},{},[346,348,354],{"type":68,"value":347},"Some CUDA-Q kernel-language constructs are constrained compared with normal\nPython; use the companion ",{"type":62,"tag":104,"props":349,"children":351},{"className":350},[],[352],{"type":68,"value":353},"cudaq-guide",{"type":68,"value":355}," skill for core CUDA-Q authoring\nconstraints and shared kernel patterns.",{"type":62,"tag":93,"props":357,"children":358},{},[359],{"type":68,"value":360},"CUDA-Q and Qiskit differ in default precision and count-key display order.\nApparent fidelity or bitstring mismatches may be convention differences.",{"type":62,"tag":93,"props":362,"children":363},{},[364],{"type":68,"value":365},"Hardware-target behavior, available backends, and target options depend on\nthe local CUDA-Q installation.",{"type":62,"tag":93,"props":367,"children":368},{},[369],{"type":68,"value":370},"This skill does not guarantee equivalent performance; it focuses on\ncorrectness-preserving ports.",{"type":62,"tag":70,"props":372,"children":374},{"id":373},"troubleshooting",[375],{"type":68,"value":376},"Troubleshooting",{"type":62,"tag":77,"props":378,"children":379},{},[380],{"type":68,"value":381},"Use this format when diagnosing failures:",{"type":62,"tag":89,"props":383,"children":384},{},[385,418,439,460,481,502],{"type":62,"tag":93,"props":386,"children":387},{},[388,394,396,402,404,409,411,416],{"type":62,"tag":389,"props":390,"children":391},"strong",{},[392],{"type":68,"value":393},"Error:",{"type":68,"value":395}," ",{"type":62,"tag":104,"props":397,"children":399},{"className":398},[],[400],{"type":68,"value":401},"ModuleNotFoundError: qiskit",{"type":68,"value":403}," from a CUDA-Q path.\n",{"type":62,"tag":389,"props":405,"children":406},{},[407],{"type":68,"value":408},"Cause:",{"type":68,"value":410}," The port still imports the source framework.\n",{"type":62,"tag":389,"props":412,"children":413},{},[414],{"type":68,"value":415},"Solution:",{"type":68,"value":417}," Move pure helpers into a framework-free module and verify with\nthe import-blocker pattern in the reference.",{"type":62,"tag":93,"props":419,"children":420},{},[421,425,427,431,433,437],{"type":62,"tag":389,"props":422,"children":423},{},[424],{"type":68,"value":393},{"type":68,"value":426}," Fidelity looks plausible but raw keys are reversed.\n",{"type":62,"tag":389,"props":428,"children":429},{},[430],{"type":68,"value":408},{"type":68,"value":432}," Qiskit and CUDA-Q count-key ordering differ.\n",{"type":62,"tag":389,"props":434,"children":435},{},[436],{"type":68,"value":415},{"type":68,"value":438}," Fix allocation, return-list order, or formatting at the port\nboundary. Do not alter the algorithm.",{"type":62,"tag":93,"props":440,"children":441},{},[442,446,448,452,454,458],{"type":62,"tag":389,"props":443,"children":444},{},[445],{"type":68,"value":393},{"type":68,"value":447}," Deep-circuit fidelity differs between frameworks.\n",{"type":62,"tag":389,"props":449,"children":450},{},[451],{"type":68,"value":408},{"type":68,"value":453}," CUDA-Q and Qiskit may be using different floating-point precision.\n",{"type":62,"tag":389,"props":455,"children":456},{},[457],{"type":68,"value":415},{"type":68,"value":459}," Match precision before comparing, then rerun the smallest\nfailing deterministic case.",{"type":62,"tag":93,"props":461,"children":462},{},[463,467,469,473,475,479],{"type":62,"tag":389,"props":464,"children":465},{},[466],{"type":68,"value":393},{"type":68,"value":468}," A multi-controlled operation works for small controls but fails or\nsilently changes behavior at higher arity.\n",{"type":62,"tag":389,"props":470,"children":471},{},[472],{"type":68,"value":408},{"type":68,"value":474}," The port used a fixed-arity dispatcher.\n",{"type":62,"tag":389,"props":476,"children":477},{},[478],{"type":68,"value":415},{"type":68,"value":480}," Use CUDA-Q control-list patterns for arbitrary arity.",{"type":62,"tag":93,"props":482,"children":483},{},[484,488,490,494,496,500],{"type":62,"tag":389,"props":485,"children":486},{},[487],{"type":68,"value":393},{"type":68,"value":489}," MCP documentation or repository lookup fails.\n",{"type":62,"tag":389,"props":491,"children":492},{},[493],{"type":68,"value":408},{"type":68,"value":495}," Connector unavailable, stale, or transiently failing.\n",{"type":62,"tag":389,"props":497,"children":498},{},[499],{"type":68,"value":415},{"type":68,"value":501}," Verify the connector\u002Fresource list, retry transient failures\nonce, then fall back to local docs\u002Fsource or official CUDA-Q docs. Do not\nchange the port based on unverified MCP results.",{"type":62,"tag":93,"props":503,"children":504},{},[505,509,511,515,517,521,523,529],{"type":62,"tag":389,"props":506,"children":507},{},[508],{"type":68,"value":393},{"type":68,"value":510}," CUDA-Q behavior conflicts with documentation while debugging.\n",{"type":62,"tag":389,"props":512,"children":513},{},[514],{"type":68,"value":408},{"type":68,"value":516}," The installed CUDA-Q version may differ from the latest\ndocumentation.\n",{"type":62,"tag":389,"props":518,"children":519},{},[520],{"type":68,"value":415},{"type":68,"value":522}," Check ",{"type":62,"tag":104,"props":524,"children":526},{"className":525},[],[527],{"type":68,"value":528},"cudaq.__version__",{"type":68,"value":530},", then review relevant documentation or\nsource changes between the installed version and latest before changing the\nport.",{"type":62,"tag":70,"props":532,"children":534},{"id":533},"references",[535],{"type":68,"value":536},"References",{"type":62,"tag":89,"props":538,"children":539},{},[540,548],{"type":62,"tag":93,"props":541,"children":542},{},[543],{"type":62,"tag":196,"props":544,"children":545},{"href":198},[546],{"type":68,"value":547},"Detailed porting reference",{"type":62,"tag":93,"props":549,"children":550},{},[551,553,558,560,566],{"type":68,"value":552},"Companion skill: ",{"type":62,"tag":104,"props":554,"children":556},{"className":555},[],[557],{"type":68,"value":353},{"type":68,"value":559}," (",{"type":62,"tag":104,"props":561,"children":563},{"className":562},[],[564],{"type":68,"value":565},"\u002Fcudaq-guide author",{"type":68,"value":567},") for CUDA-Q authoring\npatterns, kernel-language constraints, execution APIs, and debugging workflow.",{"items":569,"total":576},[570],{"slug":4,"name":4,"fn":5,"description":6,"org":571,"tags":572,"stars":20,"repoUrl":21,"updatedAt":22},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[573,574,575],{"name":9,"slug":8,"type":13},{"name":15,"slug":16,"type":13},{"name":18,"slug":19,"type":13},1,{"items":578,"total":734},[579,597,613,624,636,650,663,676,689,700,714,723],{"slug":580,"name":580,"fn":581,"description":582,"org":583,"tags":584,"stars":594,"repoUrl":595,"updatedAt":596},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[585,588,591],{"name":586,"slug":587,"type":13},"Documentation","documentation",{"name":589,"slug":590,"type":13},"MCP","mcp",{"name":592,"slug":593,"type":13},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":598,"name":598,"fn":599,"description":600,"org":601,"tags":602,"stars":610,"repoUrl":611,"updatedAt":612},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[603,606,609],{"name":604,"slug":605,"type":13},"Containers","containers",{"name":607,"slug":608,"type":13},"Deployment","deployment",{"name":15,"slug":16,"type":13},17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":614,"name":614,"fn":615,"description":616,"org":617,"tags":618,"stars":610,"repoUrl":611,"updatedAt":623},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[619,622],{"name":620,"slug":621,"type":13},"CI\u002FCD","ci-cd",{"name":607,"slug":608,"type":13},"2026-07-14T05:25:59.97109",{"slug":625,"name":625,"fn":626,"description":627,"org":628,"tags":629,"stars":610,"repoUrl":611,"updatedAt":635},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[630,631,632],{"name":620,"slug":621,"type":13},{"name":607,"slug":608,"type":13},{"name":633,"slug":634,"type":13},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":637,"name":637,"fn":638,"description":639,"org":640,"tags":641,"stars":610,"repoUrl":611,"updatedAt":649},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[642,645,646],{"name":643,"slug":644,"type":13},"Debugging","debugging",{"name":633,"slug":634,"type":13},{"name":647,"slug":648,"type":13},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":651,"name":651,"fn":652,"description":653,"org":654,"tags":655,"stars":610,"repoUrl":611,"updatedAt":662},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[656,659],{"name":657,"slug":658,"type":13},"Best Practices","best-practices",{"name":660,"slug":661,"type":13},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":664,"name":664,"fn":665,"description":666,"org":667,"tags":668,"stars":610,"repoUrl":611,"updatedAt":675},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[669,672,674],{"name":670,"slug":671,"type":13},"Machine Learning","machine-learning",{"name":673,"slug":47,"type":13},"Migration",{"name":9,"slug":8,"type":13},"2026-07-17T06:07:11.777011",{"slug":677,"name":677,"fn":678,"description":679,"org":680,"tags":681,"stars":610,"repoUrl":611,"updatedAt":688},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[682,685],{"name":683,"slug":684,"type":13},"QA","qa",{"name":686,"slug":687,"type":13},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":690,"name":690,"fn":691,"description":692,"org":693,"tags":694,"stars":610,"repoUrl":611,"updatedAt":699},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[695,696],{"name":607,"slug":608,"type":13},{"name":697,"slug":698,"type":13},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":701,"name":701,"fn":702,"description":703,"org":704,"tags":705,"stars":610,"repoUrl":611,"updatedAt":713},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[706,709,710],{"name":707,"slug":708,"type":13},"Code Review","code-review",{"name":633,"slug":634,"type":13},{"name":711,"slug":712,"type":13},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":715,"name":715,"fn":716,"description":717,"org":718,"tags":719,"stars":610,"repoUrl":611,"updatedAt":722},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[720,721],{"name":683,"slug":684,"type":13},{"name":686,"slug":687,"type":13},"2026-07-14T05:25:54.928983",{"slug":724,"name":724,"fn":725,"description":726,"org":727,"tags":728,"stars":610,"repoUrl":611,"updatedAt":733},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[729,732],{"name":730,"slug":731,"type":13},"Automation","automation",{"name":620,"slug":621,"type":13},"2026-07-30T05:29:03.275638",496]