[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-mcore-split-pr":3,"mdc--973xqc-key":36,"related-repo-nvidia-mcore-split-pr":483,"related-org-nvidia-mcore-split-pr":577},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":22,"repoUrl":23,"updatedAt":24,"license":25,"forks":26,"topics":27,"repo":31,"sourceUrl":34,"mdContent":35},"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},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19],{"name":13,"slug":14,"type":15},"GitHub","github","tag",{"name":17,"slug":18,"type":15},"Code Review","code-review",{"name":20,"slug":21,"type":15},"Pull Requests","pull-requests",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-14T05:26:01.226578","Apache-2.0",4230,[28,29,30],"large-language-models","model-para","transformers",{"repoUrl":23,"stars":22,"forks":26,"topics":32,"description":33},[28,29,30],"Ongoing research training transformer models at scale","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM\u002Ftree\u002FHEAD\u002Fskills\u002Fmcore-split-pr","---\nname: mcore-split-pr\ndescription: Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.\nlicense: Apache-2.0\nwhen_to_use: User asks to split a PR, reduce reviewer groups, or break up a large PR; 'too many CODEOWNERS', 'split this PR', 'break up PR', 'reduce reviewers needed'.\nuser_invocable: true\nargument: \"PR URL or number\"\nmetadata:\n  author: Philip Petrakian \u003Cppetrakian@nvidia.com>\n---\n\n# Split PR by CODEOWNERS Groups\n\nSplit a large pull request into multiple smaller PRs, where each PR touches\nthe fewest possible CODEOWNERS reviewer groups. The goal is to reduce review\nburden: a PR that only touches `megatron\u002Fcore\u002F` needs only the core reviewers,\nwhile a PR that also touches `examples\u002F`, `tools\u002F`, and `megatron\u002Ftraining\u002F`\npulls in many additional groups.\n\n## Answer-First Constraints\n\nFor split-planning questions, lead with these constraints before the full\nworkflow:\n\n- Minimize CODEOWNERS reviewer groups per PR, but each resulting PR must still\n  be independently mergeable and reviewable.\n- Tests travel with the production code they validate; do not split tests into a\n  separate PR just to reduce reviewer groups.\n- If PR B depends on symbols renamed in PR A, call out the dependency and put\n  backward-compatible aliases, re-exports, or shims in PR A when needed.\n- When creating dependent PRs, set the dependent PR's GitHub base\u002Fdiffbase to\n  `pull-request\u002F\u003Cbase PR number>`, not the base PR author's branch.\n- Before merging a base PR, retarget each dependent PR back to `main` and\n  refresh it against `main`; otherwise GitHub may automatically close the\n  dependent PR, losing approvals and review discussion.\n- Wait for user approval before execution.\n- Execution creates draft PRs from the right base, applies file-scoped diffs\n  with `git diff upstream\u002Fmain..\u003Csource-branch> -- \u003Cpaths> | git apply`, pushes\n  to the user's fork, and never pushes directly to upstream.\n\n## Workflow\n\n### 1. Analyze the PR\n\n1. Fetch the PR details: `gh pr view \u003Cnumber> --repo NVIDIA\u002FMegatron-LM --json title,body,headRefName,author` and `gh pr diff \u003Cnumber> --repo NVIDIA\u002FMegatron-LM --stat`. Also determine the current GitHub user with `gh api user --jq .login`.\n2. Parse `.github\u002FCODEOWNERS` to build a mapping from file path patterns to owner groups.\n3. For each changed file in the PR, determine which CODEOWNERS groups would be required to review it.\n4. Build a summary table grouped by CODEOWNERS group, showing which files pull in which groups.\n5. Count the total number of distinct reviewer groups the PR currently requires.\n\n### 2. Propose a split that minimizes reviewer groups per PR\n\nThe primary optimization goal: **minimize the number of CODEOWNERS reviewer groups required for each resulting PR**.\n\nStrategy:\n1. Cluster files by their CODEOWNERS groups. Files owned by the same set of groups naturally belong together.\n2. Identify the largest cluster — this becomes the first (and usually largest) PR.\n3. Remaining files form one or more additional PRs, each ideally requiring only one or two reviewer groups.\n4. If a split creates a dependency (e.g., PR B uses symbols renamed in PR A), the dependent PR must be merged after the first. Note this explicitly.\n5. Each PR must be independently mergeable to main — no broken imports, no missing symbols. Backward-compatible aliases and re-export stubs in the first PR can make this possible.\n\nPresent the proposed split as a table:\n- PR name\u002Fdescription\n- Files included\n- CODEOWNERS groups required\n- Dependencies on other PRs (if any)\n\nWait for user approval before proceeding.\n\n### 3. Execute the split (after user approval)\n\nFor each new PR:\n1. Create a new branch from the appropriate local base (`main`, or a dependency PR's branch).\n2. Extract the relevant changes: `git diff upstream\u002Fmain..\u003Csource-branch> -- \u003Cfile paths> | git apply`.\n3. Stage, commit with a clear message, and push to the user's fork.\n4. Create the PR as a **draft** (per repo contributing guidelines). For dependent PRs, set the GitHub base\u002Fdiffbase to `pull-request\u002F\u003Cbase PR number>`.\n5. If the original PR needs to be narrowed in scope, confirm with the user before force-pushing.\n6. Report all PR URLs when done.\n\n## Important guidelines\n\n- Always create PRs as **drafts** and push to the user's fork, never directly to upstream.\n- Backward-compatible changes (aliases, re-exports, deprecation shims) should go in the first PR so subsequent PRs can depend on them.\n- Dependent PRs should target `pull-request\u002F\u003Cbase PR number>` while stacked, then be retargeted and refreshed to `main` before the base PR is merged.\n- Test files should go with the production code they test, not in a separate PR.\n- Prefer a single clean commit per split PR over replaying the original commit history.\n- If a file is hard to categorize (e.g., it touches two groups), ask the user which PR it should go in.\n- If the current GitHub user is not the author of the original PR, each new PR's description must explicitly credit the original author (e.g., \"Original changes by @\u003Cauthor> in #\u003Cnumber>\").\n",{"data":37,"body":43},{"name":4,"description":6,"license":25,"when_to_use":38,"user_invocable":39,"argument":40,"metadata":41},"User asks to split a PR, reduce reviewer groups, or break up a large PR; 'too many CODEOWNERS', 'split this PR', 'break up PR', 'reduce reviewers needed'.",true,"PR URL or number",{"author":42},"Philip Petrakian \u003Cppetrakian@nvidia.com>",{"type":44,"children":45},"root",[46,55,94,101,106,177,183,190,251,257,269,274,302,307,330,335,341,346,406,412],{"type":47,"tag":48,"props":49,"children":51},"element","h1",{"id":50},"split-pr-by-codeowners-groups",[52],{"type":53,"value":54},"text","Split PR by CODEOWNERS Groups",{"type":47,"tag":56,"props":57,"children":58},"p",{},[59,61,68,70,76,78,84,86,92],{"type":53,"value":60},"Split a large pull request into multiple smaller PRs, where each PR touches\nthe fewest possible CODEOWNERS reviewer groups. The goal is to reduce review\nburden: a PR that only touches ",{"type":47,"tag":62,"props":63,"children":65},"code",{"className":64},[],[66],{"type":53,"value":67},"megatron\u002Fcore\u002F",{"type":53,"value":69}," needs only the core reviewers,\nwhile a PR that also touches ",{"type":47,"tag":62,"props":71,"children":73},{"className":72},[],[74],{"type":53,"value":75},"examples\u002F",{"type":53,"value":77},", ",{"type":47,"tag":62,"props":79,"children":81},{"className":80},[],[82],{"type":53,"value":83},"tools\u002F",{"type":53,"value":85},", and ",{"type":47,"tag":62,"props":87,"children":89},{"className":88},[],[90],{"type":53,"value":91},"megatron\u002Ftraining\u002F",{"type":53,"value":93},"\npulls in many additional groups.",{"type":47,"tag":95,"props":96,"children":98},"h2",{"id":97},"answer-first-constraints",[99],{"type":53,"value":100},"Answer-First Constraints",{"type":47,"tag":56,"props":102,"children":103},{},[104],{"type":53,"value":105},"For split-planning questions, lead with these constraints before the full\nworkflow:",{"type":47,"tag":107,"props":108,"children":109},"ul",{},[110,116,121,126,139,159,164],{"type":47,"tag":111,"props":112,"children":113},"li",{},[114],{"type":53,"value":115},"Minimize CODEOWNERS reviewer groups per PR, but each resulting PR must still\nbe independently mergeable and reviewable.",{"type":47,"tag":111,"props":117,"children":118},{},[119],{"type":53,"value":120},"Tests travel with the production code they validate; do not split tests into a\nseparate PR just to reduce reviewer groups.",{"type":47,"tag":111,"props":122,"children":123},{},[124],{"type":53,"value":125},"If PR B depends on symbols renamed in PR A, call out the dependency and put\nbackward-compatible aliases, re-exports, or shims in PR A when needed.",{"type":47,"tag":111,"props":127,"children":128},{},[129,131,137],{"type":53,"value":130},"When creating dependent PRs, set the dependent PR's GitHub base\u002Fdiffbase to\n",{"type":47,"tag":62,"props":132,"children":134},{"className":133},[],[135],{"type":53,"value":136},"pull-request\u002F\u003Cbase PR number>",{"type":53,"value":138},", not the base PR author's branch.",{"type":47,"tag":111,"props":140,"children":141},{},[142,144,150,152,157],{"type":53,"value":143},"Before merging a base PR, retarget each dependent PR back to ",{"type":47,"tag":62,"props":145,"children":147},{"className":146},[],[148],{"type":53,"value":149},"main",{"type":53,"value":151}," and\nrefresh it against ",{"type":47,"tag":62,"props":153,"children":155},{"className":154},[],[156],{"type":53,"value":149},{"type":53,"value":158},"; otherwise GitHub may automatically close the\ndependent PR, losing approvals and review discussion.",{"type":47,"tag":111,"props":160,"children":161},{},[162],{"type":53,"value":163},"Wait for user approval before execution.",{"type":47,"tag":111,"props":165,"children":166},{},[167,169,175],{"type":53,"value":168},"Execution creates draft PRs from the right base, applies file-scoped diffs\nwith ",{"type":47,"tag":62,"props":170,"children":172},{"className":171},[],[173],{"type":53,"value":174},"git diff upstream\u002Fmain..\u003Csource-branch> -- \u003Cpaths> | git apply",{"type":53,"value":176},", pushes\nto the user's fork, and never pushes directly to upstream.",{"type":47,"tag":95,"props":178,"children":180},{"id":179},"workflow",[181],{"type":53,"value":182},"Workflow",{"type":47,"tag":184,"props":185,"children":187},"h3",{"id":186},"_1-analyze-the-pr",[188],{"type":53,"value":189},"1. Analyze the PR",{"type":47,"tag":191,"props":192,"children":193},"ol",{},[194,223,236,241,246],{"type":47,"tag":111,"props":195,"children":196},{},[197,199,205,207,213,215,221],{"type":53,"value":198},"Fetch the PR details: ",{"type":47,"tag":62,"props":200,"children":202},{"className":201},[],[203],{"type":53,"value":204},"gh pr view \u003Cnumber> --repo NVIDIA\u002FMegatron-LM --json title,body,headRefName,author",{"type":53,"value":206}," and ",{"type":47,"tag":62,"props":208,"children":210},{"className":209},[],[211],{"type":53,"value":212},"gh pr diff \u003Cnumber> --repo NVIDIA\u002FMegatron-LM --stat",{"type":53,"value":214},". Also determine the current GitHub user with ",{"type":47,"tag":62,"props":216,"children":218},{"className":217},[],[219],{"type":53,"value":220},"gh api user --jq .login",{"type":53,"value":222},".",{"type":47,"tag":111,"props":224,"children":225},{},[226,228,234],{"type":53,"value":227},"Parse ",{"type":47,"tag":62,"props":229,"children":231},{"className":230},[],[232],{"type":53,"value":233},".github\u002FCODEOWNERS",{"type":53,"value":235}," to build a mapping from file path patterns to owner groups.",{"type":47,"tag":111,"props":237,"children":238},{},[239],{"type":53,"value":240},"For each changed file in the PR, determine which CODEOWNERS groups would be required to review it.",{"type":47,"tag":111,"props":242,"children":243},{},[244],{"type":53,"value":245},"Build a summary table grouped by CODEOWNERS group, showing which files pull in which groups.",{"type":47,"tag":111,"props":247,"children":248},{},[249],{"type":53,"value":250},"Count the total number of distinct reviewer groups the PR currently requires.",{"type":47,"tag":184,"props":252,"children":254},{"id":253},"_2-propose-a-split-that-minimizes-reviewer-groups-per-pr",[255],{"type":53,"value":256},"2. Propose a split that minimizes reviewer groups per PR",{"type":47,"tag":56,"props":258,"children":259},{},[260,262,268],{"type":53,"value":261},"The primary optimization goal: ",{"type":47,"tag":263,"props":264,"children":265},"strong",{},[266],{"type":53,"value":267},"minimize the number of CODEOWNERS reviewer groups required for each resulting PR",{"type":53,"value":222},{"type":47,"tag":56,"props":270,"children":271},{},[272],{"type":53,"value":273},"Strategy:",{"type":47,"tag":191,"props":275,"children":276},{},[277,282,287,292,297],{"type":47,"tag":111,"props":278,"children":279},{},[280],{"type":53,"value":281},"Cluster files by their CODEOWNERS groups. Files owned by the same set of groups naturally belong together.",{"type":47,"tag":111,"props":283,"children":284},{},[285],{"type":53,"value":286},"Identify the largest cluster — this becomes the first (and usually largest) PR.",{"type":47,"tag":111,"props":288,"children":289},{},[290],{"type":53,"value":291},"Remaining files form one or more additional PRs, each ideally requiring only one or two reviewer groups.",{"type":47,"tag":111,"props":293,"children":294},{},[295],{"type":53,"value":296},"If a split creates a dependency (e.g., PR B uses symbols renamed in PR A), the dependent PR must be merged after the first. Note this explicitly.",{"type":47,"tag":111,"props":298,"children":299},{},[300],{"type":53,"value":301},"Each PR must be independently mergeable to main — no broken imports, no missing symbols. Backward-compatible aliases and re-export stubs in the first PR can make this possible.",{"type":47,"tag":56,"props":303,"children":304},{},[305],{"type":53,"value":306},"Present the proposed split as a table:",{"type":47,"tag":107,"props":308,"children":309},{},[310,315,320,325],{"type":47,"tag":111,"props":311,"children":312},{},[313],{"type":53,"value":314},"PR name\u002Fdescription",{"type":47,"tag":111,"props":316,"children":317},{},[318],{"type":53,"value":319},"Files included",{"type":47,"tag":111,"props":321,"children":322},{},[323],{"type":53,"value":324},"CODEOWNERS groups required",{"type":47,"tag":111,"props":326,"children":327},{},[328],{"type":53,"value":329},"Dependencies on other PRs (if any)",{"type":47,"tag":56,"props":331,"children":332},{},[333],{"type":53,"value":334},"Wait for user approval before proceeding.",{"type":47,"tag":184,"props":336,"children":338},{"id":337},"_3-execute-the-split-after-user-approval",[339],{"type":53,"value":340},"3. Execute the split (after user approval)",{"type":47,"tag":56,"props":342,"children":343},{},[344],{"type":53,"value":345},"For each new PR:",{"type":47,"tag":191,"props":347,"children":348},{},[349,361,373,378,396,401],{"type":47,"tag":111,"props":350,"children":351},{},[352,354,359],{"type":53,"value":353},"Create a new branch from the appropriate local base (",{"type":47,"tag":62,"props":355,"children":357},{"className":356},[],[358],{"type":53,"value":149},{"type":53,"value":360},", or a dependency PR's branch).",{"type":47,"tag":111,"props":362,"children":363},{},[364,366,372],{"type":53,"value":365},"Extract the relevant changes: ",{"type":47,"tag":62,"props":367,"children":369},{"className":368},[],[370],{"type":53,"value":371},"git diff upstream\u002Fmain..\u003Csource-branch> -- \u003Cfile paths> | git apply",{"type":53,"value":222},{"type":47,"tag":111,"props":374,"children":375},{},[376],{"type":53,"value":377},"Stage, commit with a clear message, and push to the user's fork.",{"type":47,"tag":111,"props":379,"children":380},{},[381,383,388,390,395],{"type":53,"value":382},"Create the PR as a ",{"type":47,"tag":263,"props":384,"children":385},{},[386],{"type":53,"value":387},"draft",{"type":53,"value":389}," (per repo contributing guidelines). For dependent PRs, set the GitHub base\u002Fdiffbase to ",{"type":47,"tag":62,"props":391,"children":393},{"className":392},[],[394],{"type":53,"value":136},{"type":53,"value":222},{"type":47,"tag":111,"props":397,"children":398},{},[399],{"type":53,"value":400},"If the original PR needs to be narrowed in scope, confirm with the user before force-pushing.",{"type":47,"tag":111,"props":402,"children":403},{},[404],{"type":53,"value":405},"Report all PR URLs when done.",{"type":47,"tag":95,"props":407,"children":409},{"id":408},"important-guidelines",[410],{"type":53,"value":411},"Important guidelines",{"type":47,"tag":107,"props":413,"children":414},{},[415,427,432,451,456,461,466],{"type":47,"tag":111,"props":416,"children":417},{},[418,420,425],{"type":53,"value":419},"Always create PRs as ",{"type":47,"tag":263,"props":421,"children":422},{},[423],{"type":53,"value":424},"drafts",{"type":53,"value":426}," and push to the user's fork, never directly to upstream.",{"type":47,"tag":111,"props":428,"children":429},{},[430],{"type":53,"value":431},"Backward-compatible changes (aliases, re-exports, deprecation shims) should go in the first PR so subsequent PRs can depend on them.",{"type":47,"tag":111,"props":433,"children":434},{},[435,437,442,444,449],{"type":53,"value":436},"Dependent PRs should target ",{"type":47,"tag":62,"props":438,"children":440},{"className":439},[],[441],{"type":53,"value":136},{"type":53,"value":443}," while stacked, then be retargeted and refreshed to ",{"type":47,"tag":62,"props":445,"children":447},{"className":446},[],[448],{"type":53,"value":149},{"type":53,"value":450}," before the base PR is merged.",{"type":47,"tag":111,"props":452,"children":453},{},[454],{"type":53,"value":455},"Test files should go with the production code they test, not in a separate PR.",{"type":47,"tag":111,"props":457,"children":458},{},[459],{"type":53,"value":460},"Prefer a single clean commit per split PR over replaying the original commit history.",{"type":47,"tag":111,"props":462,"children":463},{},[464],{"type":53,"value":465},"If a file is hard to categorize (e.g., it touches two groups), ask the user which PR it should go in.",{"type":47,"tag":111,"props":467,"children":468},{},[469,471],{"type":53,"value":470},"If the current GitHub user is not the author of the original PR, each new PR's description must explicitly credit the original author (e.g., \"Original changes by @",{"type":47,"tag":472,"props":473,"children":474},"author",{},[475,477],{"type":53,"value":476}," in #",{"type":47,"tag":478,"props":479,"children":480},"number",{},[481],{"type":53,"value":482},"\").",{"items":484,"total":576},[485,501,512,522,536,549,563],{"slug":486,"name":486,"fn":487,"description":488,"org":489,"tags":490,"stars":22,"repoUrl":23,"updatedAt":500},"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},[491,494,497],{"name":492,"slug":493,"type":15},"Containers","containers",{"name":495,"slug":496,"type":15},"Deployment","deployment",{"name":498,"slug":499,"type":15},"Python","python","2026-07-27T06:06:11.249662",{"slug":502,"name":502,"fn":503,"description":504,"org":505,"tags":506,"stars":22,"repoUrl":23,"updatedAt":511},"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},[507,510],{"name":508,"slug":509,"type":15},"CI\u002FCD","ci-cd",{"name":495,"slug":496,"type":15},"2026-07-14T05:25:59.97109",{"slug":513,"name":513,"fn":514,"description":515,"org":516,"tags":517,"stars":22,"repoUrl":23,"updatedAt":521},"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},[518,519,520],{"name":508,"slug":509,"type":15},{"name":495,"slug":496,"type":15},{"name":13,"slug":14,"type":15},"2026-07-27T06:06:12.278222",{"slug":523,"name":523,"fn":524,"description":525,"org":526,"tags":527,"stars":22,"repoUrl":23,"updatedAt":535},"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},[528,531,532],{"name":529,"slug":530,"type":15},"Debugging","debugging",{"name":13,"slug":14,"type":15},{"name":533,"slug":534,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":537,"name":537,"fn":538,"description":539,"org":540,"tags":541,"stars":22,"repoUrl":23,"updatedAt":548},"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},[542,545],{"name":543,"slug":544,"type":15},"Best Practices","best-practices",{"name":546,"slug":547,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":550,"name":550,"fn":551,"description":552,"org":553,"tags":554,"stars":22,"repoUrl":23,"updatedAt":562},"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},[555,558,561],{"name":556,"slug":557,"type":15},"Machine Learning","machine-learning",{"name":559,"slug":560,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":564,"name":564,"fn":565,"description":566,"org":567,"tags":568,"stars":22,"repoUrl":23,"updatedAt":575},"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},[569,572],{"name":570,"slug":571,"type":15},"QA","qa",{"name":573,"slug":574,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",13,{"items":578,"total":673},[579,597,603,608,614,620,625,631,636,647,653,662],{"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":15},"Documentation","documentation",{"name":589,"slug":590,"type":15},"MCP","mcp",{"name":592,"slug":593,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":486,"name":486,"fn":487,"description":488,"org":598,"tags":599,"stars":22,"repoUrl":23,"updatedAt":500},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[600,601,602],{"name":492,"slug":493,"type":15},{"name":495,"slug":496,"type":15},{"name":498,"slug":499,"type":15},{"slug":502,"name":502,"fn":503,"description":504,"org":604,"tags":605,"stars":22,"repoUrl":23,"updatedAt":511},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[606,607],{"name":508,"slug":509,"type":15},{"name":495,"slug":496,"type":15},{"slug":513,"name":513,"fn":514,"description":515,"org":609,"tags":610,"stars":22,"repoUrl":23,"updatedAt":521},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[611,612,613],{"name":508,"slug":509,"type":15},{"name":495,"slug":496,"type":15},{"name":13,"slug":14,"type":15},{"slug":523,"name":523,"fn":524,"description":525,"org":615,"tags":616,"stars":22,"repoUrl":23,"updatedAt":535},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[617,618,619],{"name":529,"slug":530,"type":15},{"name":13,"slug":14,"type":15},{"name":533,"slug":534,"type":15},{"slug":537,"name":537,"fn":538,"description":539,"org":621,"tags":622,"stars":22,"repoUrl":23,"updatedAt":548},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[623,624],{"name":543,"slug":544,"type":15},{"name":546,"slug":547,"type":15},{"slug":550,"name":550,"fn":551,"description":552,"org":626,"tags":627,"stars":22,"repoUrl":23,"updatedAt":562},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[628,629,630],{"name":556,"slug":557,"type":15},{"name":559,"slug":560,"type":15},{"name":9,"slug":8,"type":15},{"slug":564,"name":564,"fn":565,"description":566,"org":632,"tags":633,"stars":22,"repoUrl":23,"updatedAt":575},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[634,635],{"name":570,"slug":571,"type":15},{"name":573,"slug":574,"type":15},{"slug":637,"name":637,"fn":638,"description":639,"org":640,"tags":641,"stars":22,"repoUrl":23,"updatedAt":646},"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},[642,643],{"name":495,"slug":496,"type":15},{"name":644,"slug":645,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":4,"name":4,"fn":5,"description":6,"org":648,"tags":649,"stars":22,"repoUrl":23,"updatedAt":24},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[650,651,652],{"name":17,"slug":18,"type":15},{"name":13,"slug":14,"type":15},{"name":20,"slug":21,"type":15},{"slug":654,"name":654,"fn":655,"description":656,"org":657,"tags":658,"stars":22,"repoUrl":23,"updatedAt":661},"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},[659,660],{"name":570,"slug":571,"type":15},{"name":573,"slug":574,"type":15},"2026-07-14T05:25:54.928983",{"slug":663,"name":663,"fn":664,"description":665,"org":666,"tags":667,"stars":22,"repoUrl":23,"updatedAt":672},"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},[668,671],{"name":669,"slug":670,"type":15},"Automation","automation",{"name":508,"slug":509,"type":15},"2026-07-30T05:29:03.275638",496]