[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-compileiq-bootstrap":3,"mdc-vlfhd8-key":31,"related-repo-nvidia-compileiq-bootstrap":984,"related-org-nvidia-compileiq-bootstrap":1068},{"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":26,"sourceUrl":29,"mdContent":30},"compileiq-bootstrap","bootstrap CompileIQ project environments","Use when starting a fresh CompileIQ project, hitting a socket timeout, or before running any other compileiq-* skill. Verifies CUDA 13.3+, ptxas, GPU access, that `from compileiq.ciq import Search` and friends resolve, and that `PtxasSearchSpace().retrieve()` returns a real path. Documents the env vars that control timeouts, caching, and search-space mirroring. Triggers on \"set up compileiq\", \"compileiq doesn't work\", \"socket timeout\", \"where do search spaces come from\", \"air-gapped compileiq\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17],{"name":13,"slug":14,"type":15},"Onboarding","onboarding","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Engineering","engineering",107,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FCompileIQ","2026-07-14T05:32:11.472149","Apache-2.0",8,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"An Optimizer for Nvidia Compilers.","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FCompileIQ\u002Ftree\u002FHEAD\u002Fagent-skills\u002Fcompileiq-bootstrap","---\nname: compileiq-bootstrap\ndescription: >\n  Use when starting a fresh CompileIQ project, hitting a socket timeout, or before\n  running any other compileiq-* skill. Verifies CUDA 13.3+, ptxas, GPU access,\n  that `from compileiq.ciq import Search` and friends resolve, and that\n  `PtxasSearchSpace().retrieve()` returns a real path. Documents the env vars\n  that control timeouts, caching, and search-space mirroring. Triggers on\n  \"set up compileiq\", \"compileiq doesn't work\", \"socket timeout\", \"where do\n  search spaces come from\", \"air-gapped compileiq\".\nwhen_to_use: |\n  - First time a user opens a CompileIQ project.\n  - Search() hangs on its first evaluation (almost always a non-BLAS cause now).\n  - Before invoking any other compileiq-* skill in this session.\n  Don't use when: the env is already validated in this session.\nlicense: Apache-2.0\nmetadata:\n  version: \"1.0.0\"\n  author: NVIDIA CompileIQ\n  domain: compiler-optimization\nallowed-tools: Bash Read\npaths: [\"**\u002F*.py\", \"**\u002F*.cu\", \"**\u002F*.acf\", \"**\u002F*.yaml\", \"**\u002F*.yml\"]\n---\n\n# compileiq-bootstrap\n\nValidate that a host can run CompileIQ. No installation of system libraries\n(that's distro-specific and out of scope). No state files. Just a checklist\nthat either passes or prints the precise next fix.\n\n## When\n\n- First-time setup on a new host or container.\n- `Search().start()` hangs on the first evaluation.\n- After upgrading CUDA, the `compileiq` wheel, or the Python venv.\n\n## Steps\n\n### 1. CUDA toolkit 13.3+\n\nCompileIQ's `--apply-controls` mechanism requires CUDA 13.3 or later\n(see `examples\u002Fcompilers\u002Fnvbench_example\u002Foptimize_reduction.py:302`).\n\n```bash\nnvcc --version | grep -E \"release (1[3-9]|[2-9][0-9])\\.[3-9]\"\nptxas --version | grep -E \"V(1[3-9]|[2-9][0-9])\\.[3-9]\"\n```\n\nIf either command exits non-zero, install or upgrade the CUDA toolkit and put\n`\u002Fusr\u002Flocal\u002Fcuda\u002Fbin` on `PATH` and `\u002Fusr\u002Flocal\u002Fcuda\u002Flib64` on\n`LD_LIBRARY_PATH`.\n\n### 2. GPU is visible\n\n```bash\nnvidia-smi --query-gpu=name,compute_cap --format=csv\n```\n\nCompileIQ targets compute capability 9.0 (Hopper \u002F H100) and 10.0 (Blackwell \u002F\nB200) most aggressively, but works on any GPU PTXAS supports for the chosen\narch.\n\n### 3. CompileIQ imports resolve\n\nOne shot that covers everything callers will need:\n\n```bash\npython -c \"\nfrom compileiq.ciq import Search\nfrom compileiq.types import INVALID_SCORE, BASELINE_CONFIG, WorkerTypes, ProblemType, SearchConfiguration\nfrom compileiq.search_spaces.compilers import PtxasSearchSpace, NvccSearchSpace, LocalSearchSpaceBin\nfrom compileiq.utils.helpers import save_compiler_config, load_compiler_config\nfrom compileiq.worker import MultiProcessWorker, IsoMultiProcessWorker, RayWorker, AsyncWorker\nprint('imports OK')\n\"\n```\n\nIf this fails, run `pip install compileiq` (or `pip install -e .` from a\nsource checkout) and re-run.\n\n### 4. Search-space resolution round-trip\n\nThis catches network or air-gapped issues that otherwise surface as\nsocket-timeout hangs deep inside the first evaluation:\n\n```bash\npython -c \"\nfrom compileiq.search_spaces.compilers import PtxasSearchSpace\np = PtxasSearchSpace().retrieve()\nassert p.exists() and p.stat().st_size > 0, p\nprint(f'resolved: {p}')\n\"\n```\n\nThe first run downloads from GitHub releases and caches under\n`~\u002F.cache\u002Fcompileiq\u002F\u003Ctag>\u002F`. Subsequent runs hit the cache.\n\n### 5. Env vars to know\n\n| Variable | Default | What it controls |\n|---|---|---|\n| `CIQ_SOCKET_TIMEOUT` | `20` | Seconds to wait on IPC with the core. Raise to 60-120 for large search spaces; raise much higher if you regularly see a hang on the first evaluation. |\n| `CIQ_KEEP_CACHE` | unset (`0`) | Set to `1`\u002F`true`\u002F`yes` to keep `.cache` files after a run for post-mortem replay. |\n| `CIQ_PROCESS_MODE` | `forkserver` | `forkserver`, `fork`, or `spawn`. Switch to `spawn` if `forkserver` fails on a constrained host. `IsoMultiProcessWorker` defaults to `fork` regardless. |\n| `CIQ_SEARCH_SPACES_DIR` | unset | Path to a local mirror containing `manifest.json` plus the referenced `.bin` files. Set this on air-gapped hosts to skip network. |\n| `CIQ_SEARCH_SPACES_REPO` | `NVIDIA\u002FCompileIQ` | Override the GitHub repo that release-backed search-space resolution queries. Useful for staging or forks. |\n| `CIQ_SS_TAG_PREFIX` | `search-spaces-` | Tag prefix the resolver uses when `tag=\"latest\"`. Rarely needs changing. |\n\n## Self-test\n\n```bash\nbash scripts\u002Fcheck_env.sh\n```\n\nExits 0 if every step passes. Exits with a non-zero count of failures and\nprints, for each failure, the precise command the user should run next.\n\n## Gotchas\n\n- **Socket timeout on first eval is *not* BLAS.** Current shipped binaries do\n  not require BLAS\u002FLAPACK. If a hang occurs:\n  1. Raise `CIQ_SOCKET_TIMEOUT=120` and retry.\n  2. If still hangs, mirror search spaces locally and set\n     `CIQ_SEARCH_SPACES_DIR=\u002Fpath\u002Fto\u002Fmirror`.\n  3. If still hangs, switch process mode: `CIQ_PROCESS_MODE=spawn`.\n- **Don't install system libraries in this skill.** Distro-specific package\n  installs require sudo and break in containers and on shared clusters. If\n  `nvcc`, `ptxas`, or a Python interpreter is missing, surface the error\n  and let the user decide how to install.\n\n## Next\n\n- For attention workloads: `compileiq-search-space` (variant=\"att\" by default).\n- Before paying for a full search: try `compileiq-booster-pack`.\n- Writing an objective function: `compileiq-author-objective`.\n",{"data":32,"body":45},{"name":4,"description":6,"when_to_use":33,"license":23,"metadata":34,"allowed-tools":38,"paths":39},"- First time a user opens a CompileIQ project.\n- Search() hangs on its first evaluation (almost always a non-BLAS cause now).\n- Before invoking any other compileiq-* skill in this session.\nDon't use when: the env is already validated in this session.\n",{"version":35,"author":36,"domain":37},"1.0.0","NVIDIA CompileIQ","compiler-optimization","Bash Read",[40,41,42,43,44],"**\u002F*.py","**\u002F*.cu","**\u002F*.acf","**\u002F*.yaml","**\u002F*.yml",{"type":46,"children":47},"root",[48,55,61,68,103,109,116,137,235,272,278,303,308,314,319,404,425,431,436,497,510,516,810,816,835,840,846,932,938,978],{"type":49,"tag":50,"props":51,"children":52},"element","h1",{"id":4},[53],{"type":54,"value":4},"text",{"type":49,"tag":56,"props":57,"children":58},"p",{},[59],{"type":54,"value":60},"Validate that a host can run CompileIQ. No installation of system libraries\n(that's distro-specific and out of scope). No state files. Just a checklist\nthat either passes or prints the precise next fix.",{"type":49,"tag":62,"props":63,"children":65},"h2",{"id":64},"when",[66],{"type":54,"value":67},"When",{"type":49,"tag":69,"props":70,"children":71},"ul",{},[72,78,90],{"type":49,"tag":73,"props":74,"children":75},"li",{},[76],{"type":54,"value":77},"First-time setup on a new host or container.",{"type":49,"tag":73,"props":79,"children":80},{},[81,88],{"type":49,"tag":82,"props":83,"children":85},"code",{"className":84},[],[86],{"type":54,"value":87},"Search().start()",{"type":54,"value":89}," hangs on the first evaluation.",{"type":49,"tag":73,"props":91,"children":92},{},[93,95,101],{"type":54,"value":94},"After upgrading CUDA, the ",{"type":49,"tag":82,"props":96,"children":98},{"className":97},[],[99],{"type":54,"value":100},"compileiq",{"type":54,"value":102}," wheel, or the Python venv.",{"type":49,"tag":62,"props":104,"children":106},{"id":105},"steps",[107],{"type":54,"value":108},"Steps",{"type":49,"tag":110,"props":111,"children":113},"h3",{"id":112},"_1-cuda-toolkit-133",[114],{"type":54,"value":115},"1. CUDA toolkit 13.3+",{"type":49,"tag":56,"props":117,"children":118},{},[119,121,127,129,135],{"type":54,"value":120},"CompileIQ's ",{"type":49,"tag":82,"props":122,"children":124},{"className":123},[],[125],{"type":54,"value":126},"--apply-controls",{"type":54,"value":128}," mechanism requires CUDA 13.3 or later\n(see ",{"type":49,"tag":82,"props":130,"children":132},{"className":131},[],[133],{"type":54,"value":134},"examples\u002Fcompilers\u002Fnvbench_example\u002Foptimize_reduction.py:302",{"type":54,"value":136},").",{"type":49,"tag":138,"props":139,"children":144},"pre",{"className":140,"code":141,"language":142,"meta":143,"style":143},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","nvcc --version | grep -E \"release (1[3-9]|[2-9][0-9])\\.[3-9]\"\nptxas --version | grep -E \"V(1[3-9]|[2-9][0-9])\\.[3-9]\"\n","bash","",[145],{"type":49,"tag":82,"props":146,"children":147},{"__ignoreMap":143},[148,197],{"type":49,"tag":149,"props":150,"children":153},"span",{"class":151,"line":152},"line",1,[154,160,166,172,177,182,187,192],{"type":49,"tag":149,"props":155,"children":157},{"style":156},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[158],{"type":54,"value":159},"nvcc",{"type":49,"tag":149,"props":161,"children":163},{"style":162},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[164],{"type":54,"value":165}," --version",{"type":49,"tag":149,"props":167,"children":169},{"style":168},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[170],{"type":54,"value":171}," |",{"type":49,"tag":149,"props":173,"children":174},{"style":156},[175],{"type":54,"value":176}," grep",{"type":49,"tag":149,"props":178,"children":179},{"style":162},[180],{"type":54,"value":181}," -E",{"type":49,"tag":149,"props":183,"children":184},{"style":168},[185],{"type":54,"value":186}," \"",{"type":49,"tag":149,"props":188,"children":189},{"style":162},[190],{"type":54,"value":191},"release (1[3-9]|[2-9][0-9])\\.[3-9]",{"type":49,"tag":149,"props":193,"children":194},{"style":168},[195],{"type":54,"value":196},"\"\n",{"type":49,"tag":149,"props":198,"children":200},{"class":151,"line":199},2,[201,206,210,214,218,222,226,231],{"type":49,"tag":149,"props":202,"children":203},{"style":156},[204],{"type":54,"value":205},"ptxas",{"type":49,"tag":149,"props":207,"children":208},{"style":162},[209],{"type":54,"value":165},{"type":49,"tag":149,"props":211,"children":212},{"style":168},[213],{"type":54,"value":171},{"type":49,"tag":149,"props":215,"children":216},{"style":156},[217],{"type":54,"value":176},{"type":49,"tag":149,"props":219,"children":220},{"style":162},[221],{"type":54,"value":181},{"type":49,"tag":149,"props":223,"children":224},{"style":168},[225],{"type":54,"value":186},{"type":49,"tag":149,"props":227,"children":228},{"style":162},[229],{"type":54,"value":230},"V(1[3-9]|[2-9][0-9])\\.[3-9]",{"type":49,"tag":149,"props":232,"children":233},{"style":168},[234],{"type":54,"value":196},{"type":49,"tag":56,"props":236,"children":237},{},[238,240,246,248,254,256,262,264,270],{"type":54,"value":239},"If either command exits non-zero, install or upgrade the CUDA toolkit and put\n",{"type":49,"tag":82,"props":241,"children":243},{"className":242},[],[244],{"type":54,"value":245},"\u002Fusr\u002Flocal\u002Fcuda\u002Fbin",{"type":54,"value":247}," on ",{"type":49,"tag":82,"props":249,"children":251},{"className":250},[],[252],{"type":54,"value":253},"PATH",{"type":54,"value":255}," and ",{"type":49,"tag":82,"props":257,"children":259},{"className":258},[],[260],{"type":54,"value":261},"\u002Fusr\u002Flocal\u002Fcuda\u002Flib64",{"type":54,"value":263}," on\n",{"type":49,"tag":82,"props":265,"children":267},{"className":266},[],[268],{"type":54,"value":269},"LD_LIBRARY_PATH",{"type":54,"value":271},".",{"type":49,"tag":110,"props":273,"children":275},{"id":274},"_2-gpu-is-visible",[276],{"type":54,"value":277},"2. GPU is visible",{"type":49,"tag":138,"props":279,"children":281},{"className":140,"code":280,"language":142,"meta":143,"style":143},"nvidia-smi --query-gpu=name,compute_cap --format=csv\n",[282],{"type":49,"tag":82,"props":283,"children":284},{"__ignoreMap":143},[285],{"type":49,"tag":149,"props":286,"children":287},{"class":151,"line":152},[288,293,298],{"type":49,"tag":149,"props":289,"children":290},{"style":156},[291],{"type":54,"value":292},"nvidia-smi",{"type":49,"tag":149,"props":294,"children":295},{"style":162},[296],{"type":54,"value":297}," --query-gpu=name,compute_cap",{"type":49,"tag":149,"props":299,"children":300},{"style":162},[301],{"type":54,"value":302}," --format=csv\n",{"type":49,"tag":56,"props":304,"children":305},{},[306],{"type":54,"value":307},"CompileIQ targets compute capability 9.0 (Hopper \u002F H100) and 10.0 (Blackwell \u002F\nB200) most aggressively, but works on any GPU PTXAS supports for the chosen\narch.",{"type":49,"tag":110,"props":309,"children":311},{"id":310},"_3-compileiq-imports-resolve",[312],{"type":54,"value":313},"3. CompileIQ imports resolve",{"type":49,"tag":56,"props":315,"children":316},{},[317],{"type":54,"value":318},"One shot that covers everything callers will need:",{"type":49,"tag":138,"props":320,"children":322},{"className":140,"code":321,"language":142,"meta":143,"style":143},"python -c \"\nfrom compileiq.ciq import Search\nfrom compileiq.types import INVALID_SCORE, BASELINE_CONFIG, WorkerTypes, ProblemType, SearchConfiguration\nfrom compileiq.search_spaces.compilers import PtxasSearchSpace, NvccSearchSpace, LocalSearchSpaceBin\nfrom compileiq.utils.helpers import save_compiler_config, load_compiler_config\nfrom compileiq.worker import MultiProcessWorker, IsoMultiProcessWorker, RayWorker, AsyncWorker\nprint('imports OK')\n\"\n",[323],{"type":49,"tag":82,"props":324,"children":325},{"__ignoreMap":143},[326,344,352,361,370,379,388,397],{"type":49,"tag":149,"props":327,"children":328},{"class":151,"line":152},[329,334,339],{"type":49,"tag":149,"props":330,"children":331},{"style":156},[332],{"type":54,"value":333},"python",{"type":49,"tag":149,"props":335,"children":336},{"style":162},[337],{"type":54,"value":338}," -c",{"type":49,"tag":149,"props":340,"children":341},{"style":168},[342],{"type":54,"value":343}," \"\n",{"type":49,"tag":149,"props":345,"children":346},{"class":151,"line":199},[347],{"type":49,"tag":149,"props":348,"children":349},{"style":162},[350],{"type":54,"value":351},"from compileiq.ciq import Search\n",{"type":49,"tag":149,"props":353,"children":355},{"class":151,"line":354},3,[356],{"type":49,"tag":149,"props":357,"children":358},{"style":162},[359],{"type":54,"value":360},"from compileiq.types import INVALID_SCORE, BASELINE_CONFIG, WorkerTypes, ProblemType, SearchConfiguration\n",{"type":49,"tag":149,"props":362,"children":364},{"class":151,"line":363},4,[365],{"type":49,"tag":149,"props":366,"children":367},{"style":162},[368],{"type":54,"value":369},"from compileiq.search_spaces.compilers import PtxasSearchSpace, NvccSearchSpace, LocalSearchSpaceBin\n",{"type":49,"tag":149,"props":371,"children":373},{"class":151,"line":372},5,[374],{"type":49,"tag":149,"props":375,"children":376},{"style":162},[377],{"type":54,"value":378},"from compileiq.utils.helpers import save_compiler_config, load_compiler_config\n",{"type":49,"tag":149,"props":380,"children":382},{"class":151,"line":381},6,[383],{"type":49,"tag":149,"props":384,"children":385},{"style":162},[386],{"type":54,"value":387},"from compileiq.worker import MultiProcessWorker, IsoMultiProcessWorker, RayWorker, AsyncWorker\n",{"type":49,"tag":149,"props":389,"children":391},{"class":151,"line":390},7,[392],{"type":49,"tag":149,"props":393,"children":394},{"style":162},[395],{"type":54,"value":396},"print('imports OK')\n",{"type":49,"tag":149,"props":398,"children":399},{"class":151,"line":24},[400],{"type":49,"tag":149,"props":401,"children":402},{"style":168},[403],{"type":54,"value":196},{"type":49,"tag":56,"props":405,"children":406},{},[407,409,415,417,423],{"type":54,"value":408},"If this fails, run ",{"type":49,"tag":82,"props":410,"children":412},{"className":411},[],[413],{"type":54,"value":414},"pip install compileiq",{"type":54,"value":416}," (or ",{"type":49,"tag":82,"props":418,"children":420},{"className":419},[],[421],{"type":54,"value":422},"pip install -e .",{"type":54,"value":424}," from a\nsource checkout) and re-run.",{"type":49,"tag":110,"props":426,"children":428},{"id":427},"_4-search-space-resolution-round-trip",[429],{"type":54,"value":430},"4. Search-space resolution round-trip",{"type":49,"tag":56,"props":432,"children":433},{},[434],{"type":54,"value":435},"This catches network or air-gapped issues that otherwise surface as\nsocket-timeout hangs deep inside the first evaluation:",{"type":49,"tag":138,"props":437,"children":439},{"className":140,"code":438,"language":142,"meta":143,"style":143},"python -c \"\nfrom compileiq.search_spaces.compilers import PtxasSearchSpace\np = PtxasSearchSpace().retrieve()\nassert p.exists() and p.stat().st_size > 0, p\nprint(f'resolved: {p}')\n\"\n",[440],{"type":49,"tag":82,"props":441,"children":442},{"__ignoreMap":143},[443,458,466,474,482,490],{"type":49,"tag":149,"props":444,"children":445},{"class":151,"line":152},[446,450,454],{"type":49,"tag":149,"props":447,"children":448},{"style":156},[449],{"type":54,"value":333},{"type":49,"tag":149,"props":451,"children":452},{"style":162},[453],{"type":54,"value":338},{"type":49,"tag":149,"props":455,"children":456},{"style":168},[457],{"type":54,"value":343},{"type":49,"tag":149,"props":459,"children":460},{"class":151,"line":199},[461],{"type":49,"tag":149,"props":462,"children":463},{"style":162},[464],{"type":54,"value":465},"from compileiq.search_spaces.compilers import PtxasSearchSpace\n",{"type":49,"tag":149,"props":467,"children":468},{"class":151,"line":354},[469],{"type":49,"tag":149,"props":470,"children":471},{"style":162},[472],{"type":54,"value":473},"p = PtxasSearchSpace().retrieve()\n",{"type":49,"tag":149,"props":475,"children":476},{"class":151,"line":363},[477],{"type":49,"tag":149,"props":478,"children":479},{"style":162},[480],{"type":54,"value":481},"assert p.exists() and p.stat().st_size > 0, p\n",{"type":49,"tag":149,"props":483,"children":484},{"class":151,"line":372},[485],{"type":49,"tag":149,"props":486,"children":487},{"style":162},[488],{"type":54,"value":489},"print(f'resolved: {p}')\n",{"type":49,"tag":149,"props":491,"children":492},{"class":151,"line":381},[493],{"type":49,"tag":149,"props":494,"children":495},{"style":168},[496],{"type":54,"value":196},{"type":49,"tag":56,"props":498,"children":499},{},[500,502,508],{"type":54,"value":501},"The first run downloads from GitHub releases and caches under\n",{"type":49,"tag":82,"props":503,"children":505},{"className":504},[],[506],{"type":54,"value":507},"~\u002F.cache\u002Fcompileiq\u002F\u003Ctag>\u002F",{"type":54,"value":509},". Subsequent runs hit the cache.",{"type":49,"tag":110,"props":511,"children":513},{"id":512},"_5-env-vars-to-know",[514],{"type":54,"value":515},"5. Env vars to know",{"type":49,"tag":517,"props":518,"children":519},"table",{},[520,544],{"type":49,"tag":521,"props":522,"children":523},"thead",{},[524],{"type":49,"tag":525,"props":526,"children":527},"tr",{},[528,534,539],{"type":49,"tag":529,"props":530,"children":531},"th",{},[532],{"type":54,"value":533},"Variable",{"type":49,"tag":529,"props":535,"children":536},{},[537],{"type":54,"value":538},"Default",{"type":49,"tag":529,"props":540,"children":541},{},[542],{"type":54,"value":543},"What it controls",{"type":49,"tag":545,"props":546,"children":547},"tbody",{},[548,575,636,712,750,776],{"type":49,"tag":525,"props":549,"children":550},{},[551,561,570],{"type":49,"tag":552,"props":553,"children":554},"td",{},[555],{"type":49,"tag":82,"props":556,"children":558},{"className":557},[],[559],{"type":54,"value":560},"CIQ_SOCKET_TIMEOUT",{"type":49,"tag":552,"props":562,"children":563},{},[564],{"type":49,"tag":82,"props":565,"children":567},{"className":566},[],[568],{"type":54,"value":569},"20",{"type":49,"tag":552,"props":571,"children":572},{},[573],{"type":54,"value":574},"Seconds to wait on IPC with the core. Raise to 60-120 for large search spaces; raise much higher if you regularly see a hang on the first evaluation.",{"type":49,"tag":525,"props":576,"children":577},{},[578,587,600],{"type":49,"tag":552,"props":579,"children":580},{},[581],{"type":49,"tag":82,"props":582,"children":584},{"className":583},[],[585],{"type":54,"value":586},"CIQ_KEEP_CACHE",{"type":49,"tag":552,"props":588,"children":589},{},[590,592,598],{"type":54,"value":591},"unset (",{"type":49,"tag":82,"props":593,"children":595},{"className":594},[],[596],{"type":54,"value":597},"0",{"type":54,"value":599},")",{"type":49,"tag":552,"props":601,"children":602},{},[603,605,611,613,619,620,626,628,634],{"type":54,"value":604},"Set to ",{"type":49,"tag":82,"props":606,"children":608},{"className":607},[],[609],{"type":54,"value":610},"1",{"type":54,"value":612},"\u002F",{"type":49,"tag":82,"props":614,"children":616},{"className":615},[],[617],{"type":54,"value":618},"true",{"type":54,"value":612},{"type":49,"tag":82,"props":621,"children":623},{"className":622},[],[624],{"type":54,"value":625},"yes",{"type":54,"value":627}," to keep ",{"type":49,"tag":82,"props":629,"children":631},{"className":630},[],[632],{"type":54,"value":633},".cache",{"type":54,"value":635}," files after a run for post-mortem replay.",{"type":49,"tag":525,"props":637,"children":638},{},[639,648,657],{"type":49,"tag":552,"props":640,"children":641},{},[642],{"type":49,"tag":82,"props":643,"children":645},{"className":644},[],[646],{"type":54,"value":647},"CIQ_PROCESS_MODE",{"type":49,"tag":552,"props":649,"children":650},{},[651],{"type":49,"tag":82,"props":652,"children":654},{"className":653},[],[655],{"type":54,"value":656},"forkserver",{"type":49,"tag":552,"props":658,"children":659},{},[660,665,667,673,675,681,683,688,690,695,697,703,705,710],{"type":49,"tag":82,"props":661,"children":663},{"className":662},[],[664],{"type":54,"value":656},{"type":54,"value":666},", ",{"type":49,"tag":82,"props":668,"children":670},{"className":669},[],[671],{"type":54,"value":672},"fork",{"type":54,"value":674},", or ",{"type":49,"tag":82,"props":676,"children":678},{"className":677},[],[679],{"type":54,"value":680},"spawn",{"type":54,"value":682},". Switch to ",{"type":49,"tag":82,"props":684,"children":686},{"className":685},[],[687],{"type":54,"value":680},{"type":54,"value":689}," if ",{"type":49,"tag":82,"props":691,"children":693},{"className":692},[],[694],{"type":54,"value":656},{"type":54,"value":696}," fails on a constrained host. ",{"type":49,"tag":82,"props":698,"children":700},{"className":699},[],[701],{"type":54,"value":702},"IsoMultiProcessWorker",{"type":54,"value":704}," defaults to ",{"type":49,"tag":82,"props":706,"children":708},{"className":707},[],[709],{"type":54,"value":672},{"type":54,"value":711}," regardless.",{"type":49,"tag":525,"props":713,"children":714},{},[715,724,729],{"type":49,"tag":552,"props":716,"children":717},{},[718],{"type":49,"tag":82,"props":719,"children":721},{"className":720},[],[722],{"type":54,"value":723},"CIQ_SEARCH_SPACES_DIR",{"type":49,"tag":552,"props":725,"children":726},{},[727],{"type":54,"value":728},"unset",{"type":49,"tag":552,"props":730,"children":731},{},[732,734,740,742,748],{"type":54,"value":733},"Path to a local mirror containing ",{"type":49,"tag":82,"props":735,"children":737},{"className":736},[],[738],{"type":54,"value":739},"manifest.json",{"type":54,"value":741}," plus the referenced ",{"type":49,"tag":82,"props":743,"children":745},{"className":744},[],[746],{"type":54,"value":747},".bin",{"type":54,"value":749}," files. Set this on air-gapped hosts to skip network.",{"type":49,"tag":525,"props":751,"children":752},{},[753,762,771],{"type":49,"tag":552,"props":754,"children":755},{},[756],{"type":49,"tag":82,"props":757,"children":759},{"className":758},[],[760],{"type":54,"value":761},"CIQ_SEARCH_SPACES_REPO",{"type":49,"tag":552,"props":763,"children":764},{},[765],{"type":49,"tag":82,"props":766,"children":768},{"className":767},[],[769],{"type":54,"value":770},"NVIDIA\u002FCompileIQ",{"type":49,"tag":552,"props":772,"children":773},{},[774],{"type":54,"value":775},"Override the GitHub repo that release-backed search-space resolution queries. Useful for staging or forks.",{"type":49,"tag":525,"props":777,"children":778},{},[779,788,797],{"type":49,"tag":552,"props":780,"children":781},{},[782],{"type":49,"tag":82,"props":783,"children":785},{"className":784},[],[786],{"type":54,"value":787},"CIQ_SS_TAG_PREFIX",{"type":49,"tag":552,"props":789,"children":790},{},[791],{"type":49,"tag":82,"props":792,"children":794},{"className":793},[],[795],{"type":54,"value":796},"search-spaces-",{"type":49,"tag":552,"props":798,"children":799},{},[800,802,808],{"type":54,"value":801},"Tag prefix the resolver uses when ",{"type":49,"tag":82,"props":803,"children":805},{"className":804},[],[806],{"type":54,"value":807},"tag=\"latest\"",{"type":54,"value":809},". Rarely needs changing.",{"type":49,"tag":62,"props":811,"children":813},{"id":812},"self-test",[814],{"type":54,"value":815},"Self-test",{"type":49,"tag":138,"props":817,"children":819},{"className":140,"code":818,"language":142,"meta":143,"style":143},"bash scripts\u002Fcheck_env.sh\n",[820],{"type":49,"tag":82,"props":821,"children":822},{"__ignoreMap":143},[823],{"type":49,"tag":149,"props":824,"children":825},{"class":151,"line":152},[826,830],{"type":49,"tag":149,"props":827,"children":828},{"style":156},[829],{"type":54,"value":142},{"type":49,"tag":149,"props":831,"children":832},{"style":162},[833],{"type":54,"value":834}," scripts\u002Fcheck_env.sh\n",{"type":49,"tag":56,"props":836,"children":837},{},[838],{"type":54,"value":839},"Exits 0 if every step passes. Exits with a non-zero count of failures and\nprints, for each failure, the precise command the user should run next.",{"type":49,"tag":62,"props":841,"children":843},{"id":842},"gotchas",[844],{"type":54,"value":845},"Gotchas",{"type":49,"tag":69,"props":847,"children":848},{},[849,909],{"type":49,"tag":73,"props":850,"children":851},{},[852,866,868],{"type":49,"tag":853,"props":854,"children":855},"strong",{},[856,858,864],{"type":54,"value":857},"Socket timeout on first eval is ",{"type":49,"tag":859,"props":860,"children":861},"em",{},[862],{"type":54,"value":863},"not",{"type":54,"value":865}," BLAS.",{"type":54,"value":867}," Current shipped binaries do\nnot require BLAS\u002FLAPACK. If a hang occurs:\n",{"type":49,"tag":869,"props":870,"children":871},"ol",{},[872,885,897],{"type":49,"tag":73,"props":873,"children":874},{},[875,877,883],{"type":54,"value":876},"Raise ",{"type":49,"tag":82,"props":878,"children":880},{"className":879},[],[881],{"type":54,"value":882},"CIQ_SOCKET_TIMEOUT=120",{"type":54,"value":884}," and retry.",{"type":49,"tag":73,"props":886,"children":887},{},[888,890,896],{"type":54,"value":889},"If still hangs, mirror search spaces locally and set\n",{"type":49,"tag":82,"props":891,"children":893},{"className":892},[],[894],{"type":54,"value":895},"CIQ_SEARCH_SPACES_DIR=\u002Fpath\u002Fto\u002Fmirror",{"type":54,"value":271},{"type":49,"tag":73,"props":898,"children":899},{},[900,902,908],{"type":54,"value":901},"If still hangs, switch process mode: ",{"type":49,"tag":82,"props":903,"children":905},{"className":904},[],[906],{"type":54,"value":907},"CIQ_PROCESS_MODE=spawn",{"type":54,"value":271},{"type":49,"tag":73,"props":910,"children":911},{},[912,917,919,924,925,930],{"type":49,"tag":853,"props":913,"children":914},{},[915],{"type":54,"value":916},"Don't install system libraries in this skill.",{"type":54,"value":918}," Distro-specific package\ninstalls require sudo and break in containers and on shared clusters. If\n",{"type":49,"tag":82,"props":920,"children":922},{"className":921},[],[923],{"type":54,"value":159},{"type":54,"value":666},{"type":49,"tag":82,"props":926,"children":928},{"className":927},[],[929],{"type":54,"value":205},{"type":54,"value":931},", or a Python interpreter is missing, surface the error\nand let the user decide how to install.",{"type":49,"tag":62,"props":933,"children":935},{"id":934},"next",[936],{"type":54,"value":937},"Next",{"type":49,"tag":69,"props":939,"children":940},{},[941,954,966],{"type":49,"tag":73,"props":942,"children":943},{},[944,946,952],{"type":54,"value":945},"For attention workloads: ",{"type":49,"tag":82,"props":947,"children":949},{"className":948},[],[950],{"type":54,"value":951},"compileiq-search-space",{"type":54,"value":953}," (variant=\"att\" by default).",{"type":49,"tag":73,"props":955,"children":956},{},[957,959,965],{"type":54,"value":958},"Before paying for a full search: try ",{"type":49,"tag":82,"props":960,"children":962},{"className":961},[],[963],{"type":54,"value":964},"compileiq-booster-pack",{"type":54,"value":271},{"type":49,"tag":73,"props":967,"children":968},{},[969,971,977],{"type":54,"value":970},"Writing an objective function: ",{"type":49,"tag":82,"props":972,"children":974},{"className":973},[],[975],{"type":54,"value":976},"compileiq-author-objective",{"type":54,"value":271},{"type":49,"tag":979,"props":980,"children":981},"style",{},[982],{"type":54,"value":983},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":985,"total":390},[986,997,1009,1015,1030,1042,1053],{"slug":976,"name":976,"fn":987,"description":988,"org":989,"tags":990,"stars":20,"repoUrl":21,"updatedAt":996},"author CompileIQ objective functions","Use when writing the objective_function= passed to Search(). Covers the two legal signatures (compiler-only str vs mixed list), the baseline-knockout branch, per-eval cache busting, framework-specific --apply-controls injection (raw PTXAS, NVCC, Triton, Helion, cuTeDSL\u002FFA4, FlashInfer), correctness-before-timing, INVALID_SCORE handling, and the Debug-pack O0\u002FO3 ACF-injection canary that must pass before launching a search. Triggers on \"objective function\", \"apply-controls\", \"INVALID_SCORE\", \"save_compiler_config\", \"baseline knockout\", \"BASELINE_CONFIG\", \"every config returns the same score\", \"TypeError fromhex\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[991,992,993],{"name":18,"slug":19,"type":15},{"name":9,"slug":8,"type":15},{"name":994,"slug":995,"type":15},"Performance","performance","2026-07-14T05:32:10.176779",{"slug":964,"name":964,"fn":998,"description":999,"org":1000,"tags":1001,"stars":20,"repoUrl":21,"updatedAt":1008},"apply CompileIQ booster packs to compilers","Use BEFORE running a full CompileIQ search. Walks through downloading a Booster Pack from NVIDIA\u002FCompileIQ GitHub Releases, applying ACF candidates one at a time to the user's compiler (raw PTXAS, NVCC, Triton, Helion, FlashInfer), and keeping only candidates that compile, pass correctness, and beat the no-ACF baseline. Includes the mandatory Debug-pack O0\u002FO3 ACF-injection canary that proves the ACF is reaching PTXAS. Triggers on \"booster pack\", \"ACF\", \"apply-controls\", \"speed up without searching\", \"helion fp8\", \"flashinfer batch decode\", \"debug pack\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1002,1003,1004,1007],{"name":18,"slug":19,"type":15},{"name":9,"slug":8,"type":15},{"name":1005,"slug":1006,"type":15},"Optimization","optimization",{"name":994,"slug":995,"type":15},"2026-07-14T05:32:12.791444",{"slug":4,"name":4,"fn":5,"description":6,"org":1010,"tags":1011,"stars":20,"repoUrl":21,"updatedAt":22},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1012,1013,1014],{"name":18,"slug":19,"type":15},{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},{"slug":1016,"name":1016,"fn":1017,"description":1018,"org":1019,"tags":1020,"stars":20,"repoUrl":21,"updatedAt":1029},"compileiq-debug","debug CompileIQ search and evaluation failures","Use when something is wrong: Search() hangs, all evaluations return INVALID_SCORE, scores aren't improving, every config returns the same number, ptxas errors fill the log, CV% is too high, or a winning ACF candidate needs NCU profiling to explain. Symptom-indexed table on top. Triggers on \"compileiq hang\", \"socket timeout\", \"INVALID_SCORE\", \"not converging\", \"every score is the same\", \"TypeError fromhex\", \"ncu profile\", \"register spill\", \"ptxas error\", \"not in expected format\", \"high cv\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1021,1024,1027,1028],{"name":1022,"slug":1023,"type":15},"Debugging","debugging",{"name":1025,"slug":1026,"type":15},"Evals","evals",{"name":9,"slug":8,"type":15},{"name":994,"slug":995,"type":15},"2026-07-14T05:32:05.08078",{"slug":1031,"name":1031,"fn":1032,"description":1033,"org":1034,"tags":1035,"stars":20,"repoUrl":21,"updatedAt":1041},"compileiq-run-search","execute CompileIQ search workflows","Use when composing the Search(...) call and calling .start(). Covers the four worker classes (MultiProcessWorker \u002F IsoMultiProcessWorker \u002F RayWorker \u002F AsyncWorker) and when to pick each, SearchConfiguration sizing rules, dump_results checkpointing, tracker_config choice (Disabled \u002F Loguru \u002F MLflow), num_workers\u002Ftask_timeout semantics, and GPU clock locking for stable measurements. Triggers on \"Search()\", \"tuner.start()\", \"pool_size\", \"num_workers\", \"task_timeout\", \"IsoMultiProcessWorker\", \"RayWorker\", \"dump_results\", \"MLflow\", \"GPU clocks\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1036,1039,1040],{"name":1037,"slug":1038,"type":15},"Automation","automation",{"name":9,"slug":8,"type":15},{"name":994,"slug":995,"type":15},"2026-07-14T05:32:08.913442",{"slug":951,"name":951,"fn":1043,"description":1044,"org":1045,"tags":1046,"stars":20,"repoUrl":21,"updatedAt":1052},"configure CompileIQ search spaces","Use when picking the search_space= argument for Search(). Covers the three provider classes (PtxasSearchSpace, NvccSearchSpace, LocalSearchSpaceBin), how to pin a version\u002Fvariant\u002Ftag, the attention-focused 'att' variant for attention kernels (FlashAttention, GQA, MHA, MLA, FlashInfer Batch Decode), air-gapped mirroring via CIQ_SEARCH_SPACES_DIR, and custom user-defined search spaces built from compileiq.search_spaces.base primitives. Triggers on \"search space\", \"PtxasSearchSpace\", \"NvccSearchSpace\", \"air-gapped compileiq\", \"offline compileiq\", \"CIQ_SEARCH_SPACES_DIR\", \"attention variant\", \"ptxas att\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1047,1050,1051],{"name":1048,"slug":1049,"type":15},"Configuration","configuration",{"name":9,"slug":8,"type":15},{"name":994,"slug":995,"type":15},"2026-07-14T05:32:07.605639",{"slug":1054,"name":1054,"fn":1055,"description":1056,"org":1057,"tags":1058,"stars":20,"repoUrl":21,"updatedAt":1067},"compileiq-validate-result","validate CompileIQ search results","Use AFTER a Search has completed and BEFORE claiming any speedup or shipping an ACF. Loads the dump_results CSV, extracts top-K candidates (single-objective) or the Pareto front (multi-objective), re-measures each against the no-ACF baseline with 100+ trials on fresh caches, runs Welch's t-test plus Cohen's d, rejects three classic false-positive patterns (lucky-min \u002F higher-variance \u002F multiple-comparisons-of-N), and saves the validated winner as best.acf. Triggers on \"validate result\", \"extract best config\", \"Welch's t-test\", \"is my speedup real\", \"save best ACF\", \"pareto front\", \"claim speedup\", \"ship config\".\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1059,1062,1063,1064],{"name":1060,"slug":1061,"type":15},"Data Analysis","data-analysis",{"name":9,"slug":8,"type":15},{"name":994,"slug":995,"type":15},{"name":1065,"slug":1066,"type":15},"QA","qa","2026-07-14T05:32:06.343444",{"items":1069,"total":1221},[1070,1088,1105,1116,1128,1140,1153,1167,1178,1189,1203,1212],{"slug":1071,"name":1071,"fn":1072,"description":1073,"org":1074,"tags":1075,"stars":1085,"repoUrl":1086,"updatedAt":1087},"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},[1076,1079,1082],{"name":1077,"slug":1078,"type":15},"Documentation","documentation",{"name":1080,"slug":1081,"type":15},"MCP","mcp",{"name":1083,"slug":1084,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1089,"name":1089,"fn":1090,"description":1091,"org":1092,"tags":1093,"stars":1102,"repoUrl":1103,"updatedAt":1104},"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},[1094,1097,1100],{"name":1095,"slug":1096,"type":15},"Containers","containers",{"name":1098,"slug":1099,"type":15},"Deployment","deployment",{"name":1101,"slug":333,"type":15},"Python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1106,"name":1106,"fn":1107,"description":1108,"org":1109,"tags":1110,"stars":1102,"repoUrl":1103,"updatedAt":1115},"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},[1111,1114],{"name":1112,"slug":1113,"type":15},"CI\u002FCD","ci-cd",{"name":1098,"slug":1099,"type":15},"2026-07-14T05:25:59.97109",{"slug":1117,"name":1117,"fn":1118,"description":1119,"org":1120,"tags":1121,"stars":1102,"repoUrl":1103,"updatedAt":1127},"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},[1122,1123,1124],{"name":1112,"slug":1113,"type":15},{"name":1098,"slug":1099,"type":15},{"name":1125,"slug":1126,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1129,"name":1129,"fn":1130,"description":1131,"org":1132,"tags":1133,"stars":1102,"repoUrl":1103,"updatedAt":1139},"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},[1134,1135,1136],{"name":1022,"slug":1023,"type":15},{"name":1125,"slug":1126,"type":15},{"name":1137,"slug":1138,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1141,"name":1141,"fn":1142,"description":1143,"org":1144,"tags":1145,"stars":1102,"repoUrl":1103,"updatedAt":1152},"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},[1146,1149],{"name":1147,"slug":1148,"type":15},"Best Practices","best-practices",{"name":1150,"slug":1151,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1154,"name":1154,"fn":1155,"description":1156,"org":1157,"tags":1158,"stars":1102,"repoUrl":1103,"updatedAt":1166},"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},[1159,1162,1165],{"name":1160,"slug":1161,"type":15},"Machine Learning","machine-learning",{"name":1163,"slug":1164,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1168,"name":1168,"fn":1169,"description":1170,"org":1171,"tags":1172,"stars":1102,"repoUrl":1103,"updatedAt":1177},"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},[1173,1174],{"name":1065,"slug":1066,"type":15},{"name":1175,"slug":1176,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":1179,"name":1179,"fn":1180,"description":1181,"org":1182,"tags":1183,"stars":1102,"repoUrl":1103,"updatedAt":1188},"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},[1184,1185],{"name":1098,"slug":1099,"type":15},{"name":1186,"slug":1187,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1190,"name":1190,"fn":1191,"description":1192,"org":1193,"tags":1194,"stars":1102,"repoUrl":1103,"updatedAt":1202},"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},[1195,1198,1199],{"name":1196,"slug":1197,"type":15},"Code Review","code-review",{"name":1125,"slug":1126,"type":15},{"name":1200,"slug":1201,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1204,"name":1204,"fn":1205,"description":1206,"org":1207,"tags":1208,"stars":1102,"repoUrl":1103,"updatedAt":1211},"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},[1209,1210],{"name":1065,"slug":1066,"type":15},{"name":1175,"slug":1176,"type":15},"2026-07-14T05:25:54.928983",{"slug":1213,"name":1213,"fn":1214,"description":1215,"org":1216,"tags":1217,"stars":1102,"repoUrl":1103,"updatedAt":1220},"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},[1218,1219],{"name":1037,"slug":1038,"type":15},{"name":1112,"slug":1113,"type":15},"2026-07-30T05:29:03.275638",496]