[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-mcore-onboard-gb200-1node-tests":3,"mdc--f704hv-key":33,"related-repo-nvidia-mcore-onboard-gb200-1node-tests":1595,"related-org-nvidia-mcore-onboard-gb200-1node-tests":1682},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":19,"repoUrl":20,"updatedAt":21,"license":22,"forks":23,"topics":24,"repo":28,"sourceUrl":31,"mdContent":32},"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},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16],{"name":13,"slug":14,"type":15},"QA","qa","tag",{"name":17,"slug":18,"type":15},"Testing","testing",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-14T05:25:53.673039","Apache-2.0",4230,[25,26,27],"large-language-models","model-para","transformers",{"repoUrl":20,"stars":19,"forks":23,"topics":29,"description":30},[25,26,27],"Ongoing research training transformer models at scale","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM\u002Ftree\u002FHEAD\u002Fskills\u002Fmcore-onboard-gb200-1node-tests","---\nname: mcore-onboard-gb200-1node-tests\ndescription: Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.\nlicense: Apache-2.0\nwhen_to_use: Adding GB200 github-mr tests; creating single-node variants of existing tests; expanding CI coverage for GB200; 'add GB200 MR tests', 'onboard GB200 1-node', 'create single-node variant'.\nuser_invocable: true\nargument: \"[model-yaml]  # optional: gpt, moe, or both (default: both)\"\nmetadata:\n  author: Oliver Koenig \u003Cokoenig@nvidia.com>\n---\n\n# Onboard GB200 1-Node GitHub MR Tests\n\nCreate 1-node (`mr-github`) variants of existing 2-node (`mr`-scoped) GB200 functional tests.\nEach GB200 node has **4 GPUs**. A 2-node test uses 8 GPUs total; the 1-node variant uses 4.\n\n---\n\n## Background\n\nGB200 functional tests live in `tests\u002Ftest_utils\u002Frecipes\u002Fgb200\u002F`:\n\n| Recipe file | Notes |\n|-------------|-------|\n| `gpt.yaml` | GPT dense tests, `nodes: 2, gpus: 4` (8 total) |\n| `moe.yaml` | MoE tests, `nodes: 2, gpus: 4` (8 total) |\n| `moe-1node.yaml` | Existing 1-node MoE tests, `nodes: 1, gpus: 4` (4 total) |\n| `gpt-1node.yaml` | 1-node GPT tests (create if not present) |\n\nModel configs live at:\n`tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}\u002Fmodel_config.yaml`\n\n1-node test cases use the `_1node` suffix:\n`tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\u002Fmodel_config.yaml`\n\n---\n\n## Workflow\n\n### Step 1 — Find candidate tests\n\nScan the `products:` block in `gpt.yaml` and `moe.yaml` for entries with `scope: [mr, ...]` or `scope: [mr-slim, ...]`. These are the 2-node tests that need 1-node `mr-github` counterparts.\n\nIgnore tests already covered in `*-1node.yaml` files, and ignore `nightly`, `weekly`, `mr-broken` scopes.\n\n### Step 2 — Read each model config\n\nFor each candidate, read its `model_config.yaml` and extract the key parallelism arguments:\n\n```\n--tensor-model-parallel-size   (TP)\n--pipeline-model-parallel-size (PP)\n--expert-model-parallel-size   (EP)\n--expert-tensor-parallel-size  (ETP)\n--context-parallel-size        (CP)\n--global-batch-size\n--micro-batch-size\n```\n\n### Step 3 — Classify: trivial copy vs. needs adaptation\n\nThe world size formula is: `world_size = TP × PP × DP` where `DP ≥ EP`.\n\nGoing from 8 GPUs → 4 GPUs:\n\n| Condition | Action |\n|-----------|--------|\n| `TP × PP ≤ 4` | **Trivial copy.** Config unchanged; DP is halved automatically. |\n| `TP × PP = 8` (e.g. tp4 pp2) | **Reduce PP.** Set `PP = PP \u002F 2` (e.g. pp2→1). Verify `TP × PP_new ≤ 4`. |\n| `EP > 4` (e.g. ep8 with tp1 pp1) | **Reduce EP.** Set `EP = 4`. Experts stay at `num-experts` (each EP rank holds more experts). |\n| `EP > 4` **and** `TP × PP > 4` | Reduce both PP and EP as above. |\n| ETP test (ep × etp ≤ TP × DP) | Check `EP × ETP ≤ TP × DP_new` after PP reduction. Usually satisfied when pp→1. |\n\n**Do not change GBS** — let gradient accumulation absorb the reduced DP.\n\n### Step 4 — Create `_1node` model config directories\n\n```bash\n# Trivial copy\nmkdir -p tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\ncp tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}\u002Fmodel_config.yaml \\\n   tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\u002Fmodel_config.yaml\n\n# Then apply any parallelism changes (EP or PP) with Edit tool\n```\n\n### Step 5 — Create or update recipe files\n\n**For GPT tests** — create `tests\u002Ftest_utils\u002Frecipes\u002Fgb200\u002Fgpt-1node.yaml` (if absent) by cloning `gpt.yaml`'s spec block with `nodes: 1`. Use this template for the spec:\n\n```yaml\ntype: basic\nformat_version: 1\nmaintainers: [mcore]\nloggers: [stdout]\nspec:\n  name: \"{test_case}_{environment}_{platforms}\"\n  model: gpt          # or moe\n  build: mcore-pyt-{environment}\n  nodes: 1\n  gpus: 4\n  n_repeat: 5\n  platforms: dgx_gb200\n  script_setup: |    # copy verbatim from gpt.yaml \u002F moe.yaml\n    ...\n  script: |-         # copy verbatim from gpt.yaml \u002F moe.yaml\n    ...\n```\n\n**For MoE tests** — append entries to the existing `moe-1node.yaml`.\n\n### Step 6 — Add products entries\n\nScope convention:\n- **1–2 most representative tests** per recipe: `scope: [mr-github, mr-github-slim]`\n- **All other tests**: `scope: [mr-github]`\n\n```yaml\nproducts:\n  - test_case: [\u003Ctest_case>_1node]\n    products:\n      - environment: [dev]\n        scope: [mr-github, mr-github-slim]   # or [mr-github]\n        platforms: [dgx_gb200]\n```\n\n---\n\n## Quick parallelism reference\n\n| Original (8 GPUs) | 1-node config (4 GPUs) | Notes |\n|-------------------|----------------------|-------|\n| tp1 pp1 ep1 → dp8 | tp1 pp1 ep1 → dp4 | trivial |\n| tp2 pp1 ep1 → dp4 | tp2 pp1 ep1 → dp2 | trivial |\n| tp1 pp2 ep1 → dp4 | tp1 pp2 ep1 → dp2 | trivial |\n| tp4 pp1 ep1 → dp2 | tp4 pp1 ep1 → dp1 | trivial |\n| tp1 pp4 ep1 → dp2 | tp1 pp4 ep1 → dp1 | trivial |\n| tp1 pp1 ep8 → dp8 | tp1 pp1 ep4 → dp4 | ep 8→4 |\n| tp4 pp2 ep2 etp2 → dp1 | tp4 pp1 ep2 etp2 → dp1 | pp 2→1 |\n\n---\n\n## Checklist\n\n- [ ] Identified all `mr`-scoped tests in `gpt.yaml` and `moe.yaml` not yet in `*-1node.yaml`\n- [ ] Read model config for each candidate\n- [ ] Classified trivial vs. adaptation needed\n- [ ] Created `_1node\u002Fmodel_config.yaml` for each test\n- [ ] Applied EP or PP reductions where needed\n- [ ] Created\u002Fupdated recipe YAML with `nodes: 1, gpus: 4`\n- [ ] Assigned `mr-github` scope (+ `mr-github-slim` for 1–2 representative tests per recipe)\n- [ ] Verified no `mr-github-slim` overload (slim suite should stay small)\n",{"data":34,"body":40},{"name":4,"description":6,"license":22,"when_to_use":35,"user_invocable":36,"argument":37,"metadata":38},"Adding GB200 github-mr tests; creating single-node variants of existing tests; expanding CI coverage for GB200; 'add GB200 MR tests', 'onboard GB200 1-node', 'create single-node variant'.",true,"[model-yaml]  # optional: gpt, moe, or both (default: both)",{"author":39},"Oliver Koenig \u003Cokoenig@nvidia.com>",{"type":41,"children":42},"root",[43,52,83,87,94,107,225,236,255,258,264,271,321,357,363,376,388,394,415,420,593,603,616,705,711,744,1063,1079,1085,1090,1127,1283,1286,1292,1440,1443,1449,1589],{"type":44,"tag":45,"props":46,"children":48},"element","h1",{"id":47},"onboard-gb200-1-node-github-mr-tests",[49],{"type":50,"value":51},"text","Onboard GB200 1-Node GitHub MR Tests",{"type":44,"tag":53,"props":54,"children":55},"p",{},[56,58,65,67,73,75,81],{"type":50,"value":57},"Create 1-node (",{"type":44,"tag":59,"props":60,"children":62},"code",{"className":61},[],[63],{"type":50,"value":64},"mr-github",{"type":50,"value":66},") variants of existing 2-node (",{"type":44,"tag":59,"props":68,"children":70},{"className":69},[],[71],{"type":50,"value":72},"mr",{"type":50,"value":74},"-scoped) GB200 functional tests.\nEach GB200 node has ",{"type":44,"tag":76,"props":77,"children":78},"strong",{},[79],{"type":50,"value":80},"4 GPUs",{"type":50,"value":82},". A 2-node test uses 8 GPUs total; the 1-node variant uses 4.",{"type":44,"tag":84,"props":85,"children":86},"hr",{},[],{"type":44,"tag":88,"props":89,"children":91},"h2",{"id":90},"background",[92],{"type":50,"value":93},"Background",{"type":44,"tag":53,"props":95,"children":96},{},[97,99,105],{"type":50,"value":98},"GB200 functional tests live in ",{"type":44,"tag":59,"props":100,"children":102},{"className":101},[],[103],{"type":50,"value":104},"tests\u002Ftest_utils\u002Frecipes\u002Fgb200\u002F",{"type":50,"value":106},":",{"type":44,"tag":108,"props":109,"children":110},"table",{},[111,130],{"type":44,"tag":112,"props":113,"children":114},"thead",{},[115],{"type":44,"tag":116,"props":117,"children":118},"tr",{},[119,125],{"type":44,"tag":120,"props":121,"children":122},"th",{},[123],{"type":50,"value":124},"Recipe file",{"type":44,"tag":120,"props":126,"children":127},{},[128],{"type":50,"value":129},"Notes",{"type":44,"tag":131,"props":132,"children":133},"tbody",{},[134,160,183,208],{"type":44,"tag":116,"props":135,"children":136},{},[137,147],{"type":44,"tag":138,"props":139,"children":140},"td",{},[141],{"type":44,"tag":59,"props":142,"children":144},{"className":143},[],[145],{"type":50,"value":146},"gpt.yaml",{"type":44,"tag":138,"props":148,"children":149},{},[150,152,158],{"type":50,"value":151},"GPT dense tests, ",{"type":44,"tag":59,"props":153,"children":155},{"className":154},[],[156],{"type":50,"value":157},"nodes: 2, gpus: 4",{"type":50,"value":159}," (8 total)",{"type":44,"tag":116,"props":161,"children":162},{},[163,172],{"type":44,"tag":138,"props":164,"children":165},{},[166],{"type":44,"tag":59,"props":167,"children":169},{"className":168},[],[170],{"type":50,"value":171},"moe.yaml",{"type":44,"tag":138,"props":173,"children":174},{},[175,177,182],{"type":50,"value":176},"MoE tests, ",{"type":44,"tag":59,"props":178,"children":180},{"className":179},[],[181],{"type":50,"value":157},{"type":50,"value":159},{"type":44,"tag":116,"props":184,"children":185},{},[186,195],{"type":44,"tag":138,"props":187,"children":188},{},[189],{"type":44,"tag":59,"props":190,"children":192},{"className":191},[],[193],{"type":50,"value":194},"moe-1node.yaml",{"type":44,"tag":138,"props":196,"children":197},{},[198,200,206],{"type":50,"value":199},"Existing 1-node MoE tests, ",{"type":44,"tag":59,"props":201,"children":203},{"className":202},[],[204],{"type":50,"value":205},"nodes: 1, gpus: 4",{"type":50,"value":207}," (4 total)",{"type":44,"tag":116,"props":209,"children":210},{},[211,220],{"type":44,"tag":138,"props":212,"children":213},{},[214],{"type":44,"tag":59,"props":215,"children":217},{"className":216},[],[218],{"type":50,"value":219},"gpt-1node.yaml",{"type":44,"tag":138,"props":221,"children":222},{},[223],{"type":50,"value":224},"1-node GPT tests (create if not present)",{"type":44,"tag":53,"props":226,"children":227},{},[228,230],{"type":50,"value":229},"Model configs live at:\n",{"type":44,"tag":59,"props":231,"children":233},{"className":232},[],[234],{"type":50,"value":235},"tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}\u002Fmodel_config.yaml",{"type":44,"tag":53,"props":237,"children":238},{},[239,241,247,249],{"type":50,"value":240},"1-node test cases use the ",{"type":44,"tag":59,"props":242,"children":244},{"className":243},[],[245],{"type":50,"value":246},"_1node",{"type":50,"value":248}," suffix:\n",{"type":44,"tag":59,"props":250,"children":252},{"className":251},[],[253],{"type":50,"value":254},"tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\u002Fmodel_config.yaml",{"type":44,"tag":84,"props":256,"children":257},{},[],{"type":44,"tag":88,"props":259,"children":261},{"id":260},"workflow",[262],{"type":50,"value":263},"Workflow",{"type":44,"tag":265,"props":266,"children":268},"h3",{"id":267},"step-1-find-candidate-tests",[269],{"type":50,"value":270},"Step 1 — Find candidate tests",{"type":44,"tag":53,"props":272,"children":273},{},[274,276,282,284,289,291,296,298,304,306,312,314,319],{"type":50,"value":275},"Scan the ",{"type":44,"tag":59,"props":277,"children":279},{"className":278},[],[280],{"type":50,"value":281},"products:",{"type":50,"value":283}," block in ",{"type":44,"tag":59,"props":285,"children":287},{"className":286},[],[288],{"type":50,"value":146},{"type":50,"value":290}," and ",{"type":44,"tag":59,"props":292,"children":294},{"className":293},[],[295],{"type":50,"value":171},{"type":50,"value":297}," for entries with ",{"type":44,"tag":59,"props":299,"children":301},{"className":300},[],[302],{"type":50,"value":303},"scope: [mr, ...]",{"type":50,"value":305}," or ",{"type":44,"tag":59,"props":307,"children":309},{"className":308},[],[310],{"type":50,"value":311},"scope: [mr-slim, ...]",{"type":50,"value":313},". These are the 2-node tests that need 1-node ",{"type":44,"tag":59,"props":315,"children":317},{"className":316},[],[318],{"type":50,"value":64},{"type":50,"value":320}," counterparts.",{"type":44,"tag":53,"props":322,"children":323},{},[324,326,332,334,340,342,348,349,355],{"type":50,"value":325},"Ignore tests already covered in ",{"type":44,"tag":59,"props":327,"children":329},{"className":328},[],[330],{"type":50,"value":331},"*-1node.yaml",{"type":50,"value":333}," files, and ignore ",{"type":44,"tag":59,"props":335,"children":337},{"className":336},[],[338],{"type":50,"value":339},"nightly",{"type":50,"value":341},", ",{"type":44,"tag":59,"props":343,"children":345},{"className":344},[],[346],{"type":50,"value":347},"weekly",{"type":50,"value":341},{"type":44,"tag":59,"props":350,"children":352},{"className":351},[],[353],{"type":50,"value":354},"mr-broken",{"type":50,"value":356}," scopes.",{"type":44,"tag":265,"props":358,"children":360},{"id":359},"step-2-read-each-model-config",[361],{"type":50,"value":362},"Step 2 — Read each model config",{"type":44,"tag":53,"props":364,"children":365},{},[366,368,374],{"type":50,"value":367},"For each candidate, read its ",{"type":44,"tag":59,"props":369,"children":371},{"className":370},[],[372],{"type":50,"value":373},"model_config.yaml",{"type":50,"value":375}," and extract the key parallelism arguments:",{"type":44,"tag":377,"props":378,"children":382},"pre",{"className":379,"code":381,"language":50},[380],"language-text","--tensor-model-parallel-size   (TP)\n--pipeline-model-parallel-size (PP)\n--expert-model-parallel-size   (EP)\n--expert-tensor-parallel-size  (ETP)\n--context-parallel-size        (CP)\n--global-batch-size\n--micro-batch-size\n",[383],{"type":44,"tag":59,"props":384,"children":386},{"__ignoreMap":385},"",[387],{"type":50,"value":381},{"type":44,"tag":265,"props":389,"children":391},{"id":390},"step-3-classify-trivial-copy-vs-needs-adaptation",[392],{"type":50,"value":393},"Step 3 — Classify: trivial copy vs. needs adaptation",{"type":44,"tag":53,"props":395,"children":396},{},[397,399,405,407,413],{"type":50,"value":398},"The world size formula is: ",{"type":44,"tag":59,"props":400,"children":402},{"className":401},[],[403],{"type":50,"value":404},"world_size = TP × PP × DP",{"type":50,"value":406}," where ",{"type":44,"tag":59,"props":408,"children":410},{"className":409},[],[411],{"type":50,"value":412},"DP ≥ EP",{"type":50,"value":414},".",{"type":44,"tag":53,"props":416,"children":417},{},[418],{"type":50,"value":419},"Going from 8 GPUs → 4 GPUs:",{"type":44,"tag":108,"props":421,"children":422},{},[423,439],{"type":44,"tag":112,"props":424,"children":425},{},[426],{"type":44,"tag":116,"props":427,"children":428},{},[429,434],{"type":44,"tag":120,"props":430,"children":431},{},[432],{"type":50,"value":433},"Condition",{"type":44,"tag":120,"props":435,"children":436},{},[437],{"type":50,"value":438},"Action",{"type":44,"tag":131,"props":440,"children":441},{},[442,464,503,542,572],{"type":44,"tag":116,"props":443,"children":444},{},[445,454],{"type":44,"tag":138,"props":446,"children":447},{},[448],{"type":44,"tag":59,"props":449,"children":451},{"className":450},[],[452],{"type":50,"value":453},"TP × PP ≤ 4",{"type":44,"tag":138,"props":455,"children":456},{},[457,462],{"type":44,"tag":76,"props":458,"children":459},{},[460],{"type":50,"value":461},"Trivial copy.",{"type":50,"value":463}," Config unchanged; DP is halved automatically.",{"type":44,"tag":116,"props":465,"children":466},{},[467,478],{"type":44,"tag":138,"props":468,"children":469},{},[470,476],{"type":44,"tag":59,"props":471,"children":473},{"className":472},[],[474],{"type":50,"value":475},"TP × PP = 8",{"type":50,"value":477}," (e.g. tp4 pp2)",{"type":44,"tag":138,"props":479,"children":480},{},[481,486,488,494,496,502],{"type":44,"tag":76,"props":482,"children":483},{},[484],{"type":50,"value":485},"Reduce PP.",{"type":50,"value":487}," Set ",{"type":44,"tag":59,"props":489,"children":491},{"className":490},[],[492],{"type":50,"value":493},"PP = PP \u002F 2",{"type":50,"value":495}," (e.g. pp2→1). Verify ",{"type":44,"tag":59,"props":497,"children":499},{"className":498},[],[500],{"type":50,"value":501},"TP × PP_new ≤ 4",{"type":50,"value":414},{"type":44,"tag":116,"props":504,"children":505},{},[506,517],{"type":44,"tag":138,"props":507,"children":508},{},[509,515],{"type":44,"tag":59,"props":510,"children":512},{"className":511},[],[513],{"type":50,"value":514},"EP > 4",{"type":50,"value":516}," (e.g. ep8 with tp1 pp1)",{"type":44,"tag":138,"props":518,"children":519},{},[520,525,526,532,534,540],{"type":44,"tag":76,"props":521,"children":522},{},[523],{"type":50,"value":524},"Reduce EP.",{"type":50,"value":487},{"type":44,"tag":59,"props":527,"children":529},{"className":528},[],[530],{"type":50,"value":531},"EP = 4",{"type":50,"value":533},". Experts stay at ",{"type":44,"tag":59,"props":535,"children":537},{"className":536},[],[538],{"type":50,"value":539},"num-experts",{"type":50,"value":541}," (each EP rank holds more experts).",{"type":44,"tag":116,"props":543,"children":544},{},[545,567],{"type":44,"tag":138,"props":546,"children":547},{},[548,553,555,560,561],{"type":44,"tag":59,"props":549,"children":551},{"className":550},[],[552],{"type":50,"value":514},{"type":50,"value":554}," ",{"type":44,"tag":76,"props":556,"children":557},{},[558],{"type":50,"value":559},"and",{"type":50,"value":554},{"type":44,"tag":59,"props":562,"children":564},{"className":563},[],[565],{"type":50,"value":566},"TP × PP > 4",{"type":44,"tag":138,"props":568,"children":569},{},[570],{"type":50,"value":571},"Reduce both PP and EP as above.",{"type":44,"tag":116,"props":573,"children":574},{},[575,580],{"type":44,"tag":138,"props":576,"children":577},{},[578],{"type":50,"value":579},"ETP test (ep × etp ≤ TP × DP)",{"type":44,"tag":138,"props":581,"children":582},{},[583,585,591],{"type":50,"value":584},"Check ",{"type":44,"tag":59,"props":586,"children":588},{"className":587},[],[589],{"type":50,"value":590},"EP × ETP ≤ TP × DP_new",{"type":50,"value":592}," after PP reduction. Usually satisfied when pp→1.",{"type":44,"tag":53,"props":594,"children":595},{},[596,601],{"type":44,"tag":76,"props":597,"children":598},{},[599],{"type":50,"value":600},"Do not change GBS",{"type":50,"value":602}," — let gradient accumulation absorb the reduced DP.",{"type":44,"tag":265,"props":604,"children":606},{"id":605},"step-4-create-_1node-model-config-directories",[607,609,614],{"type":50,"value":608},"Step 4 — Create ",{"type":44,"tag":59,"props":610,"children":612},{"className":611},[],[613],{"type":50,"value":246},{"type":50,"value":615}," model config directories",{"type":44,"tag":377,"props":617,"children":621},{"className":618,"code":619,"language":620,"meta":385,"style":385},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","# Trivial copy\nmkdir -p tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\ncp tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}\u002Fmodel_config.yaml \\\n   tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\u002Fmodel_config.yaml\n\n# Then apply any parallelism changes (EP or PP) with Edit tool\n","bash",[622],{"type":44,"tag":59,"props":623,"children":624},{"__ignoreMap":385},[625,637,658,678,687,696],{"type":44,"tag":626,"props":627,"children":630},"span",{"class":628,"line":629},"line",1,[631],{"type":44,"tag":626,"props":632,"children":634},{"style":633},"--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic",[635],{"type":50,"value":636},"# Trivial copy\n",{"type":44,"tag":626,"props":638,"children":640},{"class":628,"line":639},2,[641,647,653],{"type":44,"tag":626,"props":642,"children":644},{"style":643},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[645],{"type":50,"value":646},"mkdir",{"type":44,"tag":626,"props":648,"children":650},{"style":649},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[651],{"type":50,"value":652}," -p",{"type":44,"tag":626,"props":654,"children":655},{"style":649},[656],{"type":50,"value":657}," tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\n",{"type":44,"tag":626,"props":659,"children":661},{"class":628,"line":660},3,[662,667,672],{"type":44,"tag":626,"props":663,"children":664},{"style":643},[665],{"type":50,"value":666},"cp",{"type":44,"tag":626,"props":668,"children":669},{"style":649},[670],{"type":50,"value":671}," tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}\u002Fmodel_config.yaml",{"type":44,"tag":626,"props":673,"children":675},{"style":674},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[676],{"type":50,"value":677}," \\\n",{"type":44,"tag":626,"props":679,"children":681},{"class":628,"line":680},4,[682],{"type":44,"tag":626,"props":683,"children":684},{"style":649},[685],{"type":50,"value":686},"   tests\u002Ffunctional_tests\u002Ftest_cases\u002F{model}\u002F{test_case}_1node\u002Fmodel_config.yaml\n",{"type":44,"tag":626,"props":688,"children":690},{"class":628,"line":689},5,[691],{"type":44,"tag":626,"props":692,"children":693},{"emptyLinePlaceholder":36},[694],{"type":50,"value":695},"\n",{"type":44,"tag":626,"props":697,"children":699},{"class":628,"line":698},6,[700],{"type":44,"tag":626,"props":701,"children":702},{"style":633},[703],{"type":50,"value":704},"# Then apply any parallelism changes (EP or PP) with Edit tool\n",{"type":44,"tag":265,"props":706,"children":708},{"id":707},"step-5-create-or-update-recipe-files",[709],{"type":50,"value":710},"Step 5 — Create or update recipe files",{"type":44,"tag":53,"props":712,"children":713},{},[714,719,721,727,729,734,736,742],{"type":44,"tag":76,"props":715,"children":716},{},[717],{"type":50,"value":718},"For GPT tests",{"type":50,"value":720}," — create ",{"type":44,"tag":59,"props":722,"children":724},{"className":723},[],[725],{"type":50,"value":726},"tests\u002Ftest_utils\u002Frecipes\u002Fgb200\u002Fgpt-1node.yaml",{"type":50,"value":728}," (if absent) by cloning ",{"type":44,"tag":59,"props":730,"children":732},{"className":731},[],[733],{"type":50,"value":146},{"type":50,"value":735},"'s spec block with ",{"type":44,"tag":59,"props":737,"children":739},{"className":738},[],[740],{"type":50,"value":741},"nodes: 1",{"type":50,"value":743},". Use this template for the spec:",{"type":44,"tag":377,"props":745,"children":749},{"className":746,"code":747,"language":748,"meta":385,"style":385},"language-yaml shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","type: basic\nformat_version: 1\nmaintainers: [mcore]\nloggers: [stdout]\nspec:\n  name: \"{test_case}_{environment}_{platforms}\"\n  model: gpt          # or moe\n  build: mcore-pyt-{environment}\n  nodes: 1\n  gpus: 4\n  n_repeat: 5\n  platforms: dgx_gb200\n  script_setup: |    # copy verbatim from gpt.yaml \u002F moe.yaml\n    ...\n  script: |-         # copy verbatim from gpt.yaml \u002F moe.yaml\n    ...\n","yaml",[750],{"type":44,"tag":59,"props":751,"children":752},{"__ignoreMap":385},[753,772,790,817,842,855,882,905,923,940,958,976,994,1018,1027,1055],{"type":44,"tag":626,"props":754,"children":755},{"class":628,"line":629},[756,762,767],{"type":44,"tag":626,"props":757,"children":759},{"style":758},"--shiki-light:#E53935;--shiki-default:#F07178;--shiki-dark:#F07178",[760],{"type":50,"value":761},"type",{"type":44,"tag":626,"props":763,"children":765},{"style":764},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[766],{"type":50,"value":106},{"type":44,"tag":626,"props":768,"children":769},{"style":649},[770],{"type":50,"value":771}," basic\n",{"type":44,"tag":626,"props":773,"children":774},{"class":628,"line":639},[775,780,784],{"type":44,"tag":626,"props":776,"children":777},{"style":758},[778],{"type":50,"value":779},"format_version",{"type":44,"tag":626,"props":781,"children":782},{"style":764},[783],{"type":50,"value":106},{"type":44,"tag":626,"props":785,"children":787},{"style":786},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[788],{"type":50,"value":789}," 1\n",{"type":44,"tag":626,"props":791,"children":792},{"class":628,"line":660},[793,798,802,807,812],{"type":44,"tag":626,"props":794,"children":795},{"style":758},[796],{"type":50,"value":797},"maintainers",{"type":44,"tag":626,"props":799,"children":800},{"style":764},[801],{"type":50,"value":106},{"type":44,"tag":626,"props":803,"children":804},{"style":764},[805],{"type":50,"value":806}," [",{"type":44,"tag":626,"props":808,"children":809},{"style":649},[810],{"type":50,"value":811},"mcore",{"type":44,"tag":626,"props":813,"children":814},{"style":764},[815],{"type":50,"value":816},"]\n",{"type":44,"tag":626,"props":818,"children":819},{"class":628,"line":680},[820,825,829,833,838],{"type":44,"tag":626,"props":821,"children":822},{"style":758},[823],{"type":50,"value":824},"loggers",{"type":44,"tag":626,"props":826,"children":827},{"style":764},[828],{"type":50,"value":106},{"type":44,"tag":626,"props":830,"children":831},{"style":764},[832],{"type":50,"value":806},{"type":44,"tag":626,"props":834,"children":835},{"style":649},[836],{"type":50,"value":837},"stdout",{"type":44,"tag":626,"props":839,"children":840},{"style":764},[841],{"type":50,"value":816},{"type":44,"tag":626,"props":843,"children":844},{"class":628,"line":689},[845,850],{"type":44,"tag":626,"props":846,"children":847},{"style":758},[848],{"type":50,"value":849},"spec",{"type":44,"tag":626,"props":851,"children":852},{"style":764},[853],{"type":50,"value":854},":\n",{"type":44,"tag":626,"props":856,"children":857},{"class":628,"line":698},[858,863,867,872,877],{"type":44,"tag":626,"props":859,"children":860},{"style":758},[861],{"type":50,"value":862},"  name",{"type":44,"tag":626,"props":864,"children":865},{"style":764},[866],{"type":50,"value":106},{"type":44,"tag":626,"props":868,"children":869},{"style":764},[870],{"type":50,"value":871}," \"",{"type":44,"tag":626,"props":873,"children":874},{"style":649},[875],{"type":50,"value":876},"{test_case}_{environment}_{platforms}",{"type":44,"tag":626,"props":878,"children":879},{"style":764},[880],{"type":50,"value":881},"\"\n",{"type":44,"tag":626,"props":883,"children":885},{"class":628,"line":884},7,[886,891,895,900],{"type":44,"tag":626,"props":887,"children":888},{"style":758},[889],{"type":50,"value":890},"  model",{"type":44,"tag":626,"props":892,"children":893},{"style":764},[894],{"type":50,"value":106},{"type":44,"tag":626,"props":896,"children":897},{"style":649},[898],{"type":50,"value":899}," gpt",{"type":44,"tag":626,"props":901,"children":902},{"style":633},[903],{"type":50,"value":904},"          # or moe\n",{"type":44,"tag":626,"props":906,"children":908},{"class":628,"line":907},8,[909,914,918],{"type":44,"tag":626,"props":910,"children":911},{"style":758},[912],{"type":50,"value":913},"  build",{"type":44,"tag":626,"props":915,"children":916},{"style":764},[917],{"type":50,"value":106},{"type":44,"tag":626,"props":919,"children":920},{"style":649},[921],{"type":50,"value":922}," mcore-pyt-{environment}\n",{"type":44,"tag":626,"props":924,"children":926},{"class":628,"line":925},9,[927,932,936],{"type":44,"tag":626,"props":928,"children":929},{"style":758},[930],{"type":50,"value":931},"  nodes",{"type":44,"tag":626,"props":933,"children":934},{"style":764},[935],{"type":50,"value":106},{"type":44,"tag":626,"props":937,"children":938},{"style":786},[939],{"type":50,"value":789},{"type":44,"tag":626,"props":941,"children":943},{"class":628,"line":942},10,[944,949,953],{"type":44,"tag":626,"props":945,"children":946},{"style":758},[947],{"type":50,"value":948},"  gpus",{"type":44,"tag":626,"props":950,"children":951},{"style":764},[952],{"type":50,"value":106},{"type":44,"tag":626,"props":954,"children":955},{"style":786},[956],{"type":50,"value":957}," 4\n",{"type":44,"tag":626,"props":959,"children":961},{"class":628,"line":960},11,[962,967,971],{"type":44,"tag":626,"props":963,"children":964},{"style":758},[965],{"type":50,"value":966},"  n_repeat",{"type":44,"tag":626,"props":968,"children":969},{"style":764},[970],{"type":50,"value":106},{"type":44,"tag":626,"props":972,"children":973},{"style":786},[974],{"type":50,"value":975}," 5\n",{"type":44,"tag":626,"props":977,"children":979},{"class":628,"line":978},12,[980,985,989],{"type":44,"tag":626,"props":981,"children":982},{"style":758},[983],{"type":50,"value":984},"  platforms",{"type":44,"tag":626,"props":986,"children":987},{"style":764},[988],{"type":50,"value":106},{"type":44,"tag":626,"props":990,"children":991},{"style":649},[992],{"type":50,"value":993}," dgx_gb200\n",{"type":44,"tag":626,"props":995,"children":997},{"class":628,"line":996},13,[998,1003,1007,1013],{"type":44,"tag":626,"props":999,"children":1000},{"style":758},[1001],{"type":50,"value":1002},"  script_setup",{"type":44,"tag":626,"props":1004,"children":1005},{"style":764},[1006],{"type":50,"value":106},{"type":44,"tag":626,"props":1008,"children":1010},{"style":1009},"--shiki-light:#39ADB5;--shiki-light-font-style:italic;--shiki-default:#89DDFF;--shiki-default-font-style:italic;--shiki-dark:#89DDFF;--shiki-dark-font-style:italic",[1011],{"type":50,"value":1012}," |",{"type":44,"tag":626,"props":1014,"children":1015},{"style":633},[1016],{"type":50,"value":1017},"    # copy verbatim from gpt.yaml \u002F moe.yaml\n",{"type":44,"tag":626,"props":1019,"children":1021},{"class":628,"line":1020},14,[1022],{"type":44,"tag":626,"props":1023,"children":1024},{"style":649},[1025],{"type":50,"value":1026},"    ...\n",{"type":44,"tag":626,"props":1028,"children":1030},{"class":628,"line":1029},15,[1031,1036,1040,1044,1050],{"type":44,"tag":626,"props":1032,"children":1033},{"style":758},[1034],{"type":50,"value":1035},"  script",{"type":44,"tag":626,"props":1037,"children":1038},{"style":764},[1039],{"type":50,"value":106},{"type":44,"tag":626,"props":1041,"children":1042},{"style":1009},[1043],{"type":50,"value":1012},{"type":44,"tag":626,"props":1045,"children":1047},{"style":1046},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[1048],{"type":50,"value":1049},"-",{"type":44,"tag":626,"props":1051,"children":1052},{"style":633},[1053],{"type":50,"value":1054},"         # copy verbatim from gpt.yaml \u002F moe.yaml\n",{"type":44,"tag":626,"props":1056,"children":1058},{"class":628,"line":1057},16,[1059],{"type":44,"tag":626,"props":1060,"children":1061},{"style":649},[1062],{"type":50,"value":1026},{"type":44,"tag":53,"props":1064,"children":1065},{},[1066,1071,1073,1078],{"type":44,"tag":76,"props":1067,"children":1068},{},[1069],{"type":50,"value":1070},"For MoE tests",{"type":50,"value":1072}," — append entries to the existing ",{"type":44,"tag":59,"props":1074,"children":1076},{"className":1075},[],[1077],{"type":50,"value":194},{"type":50,"value":414},{"type":44,"tag":265,"props":1080,"children":1082},{"id":1081},"step-6-add-products-entries",[1083],{"type":50,"value":1084},"Step 6 — Add products entries",{"type":44,"tag":53,"props":1086,"children":1087},{},[1088],{"type":50,"value":1089},"Scope convention:",{"type":44,"tag":1091,"props":1092,"children":1093},"ul",{},[1094,1111],{"type":44,"tag":1095,"props":1096,"children":1097},"li",{},[1098,1103,1105],{"type":44,"tag":76,"props":1099,"children":1100},{},[1101],{"type":50,"value":1102},"1–2 most representative tests",{"type":50,"value":1104}," per recipe: ",{"type":44,"tag":59,"props":1106,"children":1108},{"className":1107},[],[1109],{"type":50,"value":1110},"scope: [mr-github, mr-github-slim]",{"type":44,"tag":1095,"props":1112,"children":1113},{},[1114,1119,1121],{"type":44,"tag":76,"props":1115,"children":1116},{},[1117],{"type":50,"value":1118},"All other tests",{"type":50,"value":1120},": ",{"type":44,"tag":59,"props":1122,"children":1124},{"className":1123},[],[1125],{"type":50,"value":1126},"scope: [mr-github]",{"type":44,"tag":377,"props":1128,"children":1130},{"className":746,"code":1129,"language":748,"meta":385,"style":385},"products:\n  - test_case: [\u003Ctest_case>_1node]\n    products:\n      - environment: [dev]\n        scope: [mr-github, mr-github-slim]   # or [mr-github]\n        platforms: [dgx_gb200]\n",[1131],{"type":44,"tag":59,"props":1132,"children":1133},{"__ignoreMap":385},[1134,1146,1176,1188,1218,1258],{"type":44,"tag":626,"props":1135,"children":1136},{"class":628,"line":629},[1137,1142],{"type":44,"tag":626,"props":1138,"children":1139},{"style":758},[1140],{"type":50,"value":1141},"products",{"type":44,"tag":626,"props":1143,"children":1144},{"style":764},[1145],{"type":50,"value":854},{"type":44,"tag":626,"props":1147,"children":1148},{"class":628,"line":639},[1149,1154,1159,1163,1167,1172],{"type":44,"tag":626,"props":1150,"children":1151},{"style":764},[1152],{"type":50,"value":1153},"  -",{"type":44,"tag":626,"props":1155,"children":1156},{"style":758},[1157],{"type":50,"value":1158}," test_case",{"type":44,"tag":626,"props":1160,"children":1161},{"style":764},[1162],{"type":50,"value":106},{"type":44,"tag":626,"props":1164,"children":1165},{"style":764},[1166],{"type":50,"value":806},{"type":44,"tag":626,"props":1168,"children":1169},{"style":649},[1170],{"type":50,"value":1171},"\u003Ctest_case>_1node",{"type":44,"tag":626,"props":1173,"children":1174},{"style":764},[1175],{"type":50,"value":816},{"type":44,"tag":626,"props":1177,"children":1178},{"class":628,"line":660},[1179,1184],{"type":44,"tag":626,"props":1180,"children":1181},{"style":758},[1182],{"type":50,"value":1183},"    products",{"type":44,"tag":626,"props":1185,"children":1186},{"style":764},[1187],{"type":50,"value":854},{"type":44,"tag":626,"props":1189,"children":1190},{"class":628,"line":680},[1191,1196,1201,1205,1209,1214],{"type":44,"tag":626,"props":1192,"children":1193},{"style":764},[1194],{"type":50,"value":1195},"      -",{"type":44,"tag":626,"props":1197,"children":1198},{"style":758},[1199],{"type":50,"value":1200}," environment",{"type":44,"tag":626,"props":1202,"children":1203},{"style":764},[1204],{"type":50,"value":106},{"type":44,"tag":626,"props":1206,"children":1207},{"style":764},[1208],{"type":50,"value":806},{"type":44,"tag":626,"props":1210,"children":1211},{"style":649},[1212],{"type":50,"value":1213},"dev",{"type":44,"tag":626,"props":1215,"children":1216},{"style":764},[1217],{"type":50,"value":816},{"type":44,"tag":626,"props":1219,"children":1220},{"class":628,"line":689},[1221,1226,1230,1234,1238,1243,1248,1253],{"type":44,"tag":626,"props":1222,"children":1223},{"style":758},[1224],{"type":50,"value":1225},"        scope",{"type":44,"tag":626,"props":1227,"children":1228},{"style":764},[1229],{"type":50,"value":106},{"type":44,"tag":626,"props":1231,"children":1232},{"style":764},[1233],{"type":50,"value":806},{"type":44,"tag":626,"props":1235,"children":1236},{"style":649},[1237],{"type":50,"value":64},{"type":44,"tag":626,"props":1239,"children":1240},{"style":764},[1241],{"type":50,"value":1242},",",{"type":44,"tag":626,"props":1244,"children":1245},{"style":649},[1246],{"type":50,"value":1247}," mr-github-slim",{"type":44,"tag":626,"props":1249,"children":1250},{"style":764},[1251],{"type":50,"value":1252},"]",{"type":44,"tag":626,"props":1254,"children":1255},{"style":633},[1256],{"type":50,"value":1257},"   # or [mr-github]\n",{"type":44,"tag":626,"props":1259,"children":1260},{"class":628,"line":698},[1261,1266,1270,1274,1279],{"type":44,"tag":626,"props":1262,"children":1263},{"style":758},[1264],{"type":50,"value":1265},"        platforms",{"type":44,"tag":626,"props":1267,"children":1268},{"style":764},[1269],{"type":50,"value":106},{"type":44,"tag":626,"props":1271,"children":1272},{"style":764},[1273],{"type":50,"value":806},{"type":44,"tag":626,"props":1275,"children":1276},{"style":649},[1277],{"type":50,"value":1278},"dgx_gb200",{"type":44,"tag":626,"props":1280,"children":1281},{"style":764},[1282],{"type":50,"value":816},{"type":44,"tag":84,"props":1284,"children":1285},{},[],{"type":44,"tag":88,"props":1287,"children":1289},{"id":1288},"quick-parallelism-reference",[1290],{"type":50,"value":1291},"Quick parallelism reference",{"type":44,"tag":108,"props":1293,"children":1294},{},[1295,1315],{"type":44,"tag":112,"props":1296,"children":1297},{},[1298],{"type":44,"tag":116,"props":1299,"children":1300},{},[1301,1306,1311],{"type":44,"tag":120,"props":1302,"children":1303},{},[1304],{"type":50,"value":1305},"Original (8 GPUs)",{"type":44,"tag":120,"props":1307,"children":1308},{},[1309],{"type":50,"value":1310},"1-node config (4 GPUs)",{"type":44,"tag":120,"props":1312,"children":1313},{},[1314],{"type":50,"value":129},{"type":44,"tag":131,"props":1316,"children":1317},{},[1318,1336,1353,1370,1387,1404,1422],{"type":44,"tag":116,"props":1319,"children":1320},{},[1321,1326,1331],{"type":44,"tag":138,"props":1322,"children":1323},{},[1324],{"type":50,"value":1325},"tp1 pp1 ep1 → dp8",{"type":44,"tag":138,"props":1327,"children":1328},{},[1329],{"type":50,"value":1330},"tp1 pp1 ep1 → dp4",{"type":44,"tag":138,"props":1332,"children":1333},{},[1334],{"type":50,"value":1335},"trivial",{"type":44,"tag":116,"props":1337,"children":1338},{},[1339,1344,1349],{"type":44,"tag":138,"props":1340,"children":1341},{},[1342],{"type":50,"value":1343},"tp2 pp1 ep1 → dp4",{"type":44,"tag":138,"props":1345,"children":1346},{},[1347],{"type":50,"value":1348},"tp2 pp1 ep1 → dp2",{"type":44,"tag":138,"props":1350,"children":1351},{},[1352],{"type":50,"value":1335},{"type":44,"tag":116,"props":1354,"children":1355},{},[1356,1361,1366],{"type":44,"tag":138,"props":1357,"children":1358},{},[1359],{"type":50,"value":1360},"tp1 pp2 ep1 → dp4",{"type":44,"tag":138,"props":1362,"children":1363},{},[1364],{"type":50,"value":1365},"tp1 pp2 ep1 → dp2",{"type":44,"tag":138,"props":1367,"children":1368},{},[1369],{"type":50,"value":1335},{"type":44,"tag":116,"props":1371,"children":1372},{},[1373,1378,1383],{"type":44,"tag":138,"props":1374,"children":1375},{},[1376],{"type":50,"value":1377},"tp4 pp1 ep1 → dp2",{"type":44,"tag":138,"props":1379,"children":1380},{},[1381],{"type":50,"value":1382},"tp4 pp1 ep1 → dp1",{"type":44,"tag":138,"props":1384,"children":1385},{},[1386],{"type":50,"value":1335},{"type":44,"tag":116,"props":1388,"children":1389},{},[1390,1395,1400],{"type":44,"tag":138,"props":1391,"children":1392},{},[1393],{"type":50,"value":1394},"tp1 pp4 ep1 → dp2",{"type":44,"tag":138,"props":1396,"children":1397},{},[1398],{"type":50,"value":1399},"tp1 pp4 ep1 → dp1",{"type":44,"tag":138,"props":1401,"children":1402},{},[1403],{"type":50,"value":1335},{"type":44,"tag":116,"props":1405,"children":1406},{},[1407,1412,1417],{"type":44,"tag":138,"props":1408,"children":1409},{},[1410],{"type":50,"value":1411},"tp1 pp1 ep8 → dp8",{"type":44,"tag":138,"props":1413,"children":1414},{},[1415],{"type":50,"value":1416},"tp1 pp1 ep4 → dp4",{"type":44,"tag":138,"props":1418,"children":1419},{},[1420],{"type":50,"value":1421},"ep 8→4",{"type":44,"tag":116,"props":1423,"children":1424},{},[1425,1430,1435],{"type":44,"tag":138,"props":1426,"children":1427},{},[1428],{"type":50,"value":1429},"tp4 pp2 ep2 etp2 → dp1",{"type":44,"tag":138,"props":1431,"children":1432},{},[1433],{"type":50,"value":1434},"tp4 pp1 ep2 etp2 → dp1",{"type":44,"tag":138,"props":1436,"children":1437},{},[1438],{"type":50,"value":1439},"pp 2→1",{"type":44,"tag":84,"props":1441,"children":1442},{},[],{"type":44,"tag":88,"props":1444,"children":1446},{"id":1445},"checklist",[1447],{"type":50,"value":1448},"Checklist",{"type":44,"tag":1091,"props":1450,"children":1453},{"className":1451},[1452],"contains-task-list",[1454,1491,1500,1509,1526,1535,1549,1573],{"type":44,"tag":1095,"props":1455,"children":1458},{"className":1456},[1457],"task-list-item",[1459,1464,1466,1471,1473,1478,1479,1484,1486],{"type":44,"tag":1460,"props":1461,"children":1463},"input",{"disabled":36,"type":1462},"checkbox",[],{"type":50,"value":1465}," Identified all ",{"type":44,"tag":59,"props":1467,"children":1469},{"className":1468},[],[1470],{"type":50,"value":72},{"type":50,"value":1472},"-scoped tests in ",{"type":44,"tag":59,"props":1474,"children":1476},{"className":1475},[],[1477],{"type":50,"value":146},{"type":50,"value":290},{"type":44,"tag":59,"props":1480,"children":1482},{"className":1481},[],[1483],{"type":50,"value":171},{"type":50,"value":1485}," not yet in ",{"type":44,"tag":59,"props":1487,"children":1489},{"className":1488},[],[1490],{"type":50,"value":331},{"type":44,"tag":1095,"props":1492,"children":1494},{"className":1493},[1457],[1495,1498],{"type":44,"tag":1460,"props":1496,"children":1497},{"disabled":36,"type":1462},[],{"type":50,"value":1499}," Read model config for each candidate",{"type":44,"tag":1095,"props":1501,"children":1503},{"className":1502},[1457],[1504,1507],{"type":44,"tag":1460,"props":1505,"children":1506},{"disabled":36,"type":1462},[],{"type":50,"value":1508}," Classified trivial vs. adaptation needed",{"type":44,"tag":1095,"props":1510,"children":1512},{"className":1511},[1457],[1513,1516,1518,1524],{"type":44,"tag":1460,"props":1514,"children":1515},{"disabled":36,"type":1462},[],{"type":50,"value":1517}," Created ",{"type":44,"tag":59,"props":1519,"children":1521},{"className":1520},[],[1522],{"type":50,"value":1523},"_1node\u002Fmodel_config.yaml",{"type":50,"value":1525}," for each test",{"type":44,"tag":1095,"props":1527,"children":1529},{"className":1528},[1457],[1530,1533],{"type":44,"tag":1460,"props":1531,"children":1532},{"disabled":36,"type":1462},[],{"type":50,"value":1534}," Applied EP or PP reductions where needed",{"type":44,"tag":1095,"props":1536,"children":1538},{"className":1537},[1457],[1539,1542,1544],{"type":44,"tag":1460,"props":1540,"children":1541},{"disabled":36,"type":1462},[],{"type":50,"value":1543}," Created\u002Fupdated recipe YAML with ",{"type":44,"tag":59,"props":1545,"children":1547},{"className":1546},[],[1548],{"type":50,"value":205},{"type":44,"tag":1095,"props":1550,"children":1552},{"className":1551},[1457],[1553,1556,1558,1563,1565,1571],{"type":44,"tag":1460,"props":1554,"children":1555},{"disabled":36,"type":1462},[],{"type":50,"value":1557}," Assigned ",{"type":44,"tag":59,"props":1559,"children":1561},{"className":1560},[],[1562],{"type":50,"value":64},{"type":50,"value":1564}," scope (+ ",{"type":44,"tag":59,"props":1566,"children":1568},{"className":1567},[],[1569],{"type":50,"value":1570},"mr-github-slim",{"type":50,"value":1572}," for 1–2 representative tests per recipe)",{"type":44,"tag":1095,"props":1574,"children":1576},{"className":1575},[1457],[1577,1580,1582,1587],{"type":44,"tag":1460,"props":1578,"children":1579},{"disabled":36,"type":1462},[],{"type":50,"value":1581}," Verified no ",{"type":44,"tag":59,"props":1583,"children":1585},{"className":1584},[],[1586],{"type":50,"value":1570},{"type":50,"value":1588}," overload (slim suite should stay small)",{"type":44,"tag":1590,"props":1591,"children":1592},"style",{},[1593],{"type":50,"value":1594},"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":1596,"total":996},[1597,1613,1624,1636,1650,1663,1677],{"slug":1598,"name":1598,"fn":1599,"description":1600,"org":1601,"tags":1602,"stars":19,"repoUrl":20,"updatedAt":1612},"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},[1603,1606,1609],{"name":1604,"slug":1605,"type":15},"Containers","containers",{"name":1607,"slug":1608,"type":15},"Deployment","deployment",{"name":1610,"slug":1611,"type":15},"Python","python","2026-07-27T06:06:11.249662",{"slug":1614,"name":1614,"fn":1615,"description":1616,"org":1617,"tags":1618,"stars":19,"repoUrl":20,"updatedAt":1623},"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},[1619,1622],{"name":1620,"slug":1621,"type":15},"CI\u002FCD","ci-cd",{"name":1607,"slug":1608,"type":15},"2026-07-14T05:25:59.97109",{"slug":1625,"name":1625,"fn":1626,"description":1627,"org":1628,"tags":1629,"stars":19,"repoUrl":20,"updatedAt":1635},"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},[1630,1631,1632],{"name":1620,"slug":1621,"type":15},{"name":1607,"slug":1608,"type":15},{"name":1633,"slug":1634,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1637,"name":1637,"fn":1638,"description":1639,"org":1640,"tags":1641,"stars":19,"repoUrl":20,"updatedAt":1649},"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},[1642,1645,1646],{"name":1643,"slug":1644,"type":15},"Debugging","debugging",{"name":1633,"slug":1634,"type":15},{"name":1647,"slug":1648,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1651,"name":1651,"fn":1652,"description":1653,"org":1654,"tags":1655,"stars":19,"repoUrl":20,"updatedAt":1662},"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},[1656,1659],{"name":1657,"slug":1658,"type":15},"Best Practices","best-practices",{"name":1660,"slug":1661,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1664,"name":1664,"fn":1665,"description":1666,"org":1667,"tags":1668,"stars":19,"repoUrl":20,"updatedAt":1676},"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},[1669,1672,1675],{"name":1670,"slug":1671,"type":15},"Machine Learning","machine-learning",{"name":1673,"slug":1674,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":4,"name":4,"fn":5,"description":6,"org":1678,"tags":1679,"stars":19,"repoUrl":20,"updatedAt":21},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1680,1681],{"name":13,"slug":14,"type":15},{"name":17,"slug":18,"type":15},{"items":1683,"total":1786},[1684,1702,1708,1713,1719,1725,1730,1736,1741,1752,1766,1775],{"slug":1685,"name":1685,"fn":1686,"description":1687,"org":1688,"tags":1689,"stars":1699,"repoUrl":1700,"updatedAt":1701},"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},[1690,1693,1696],{"name":1691,"slug":1692,"type":15},"Documentation","documentation",{"name":1694,"slug":1695,"type":15},"MCP","mcp",{"name":1697,"slug":1698,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1598,"name":1598,"fn":1599,"description":1600,"org":1703,"tags":1704,"stars":19,"repoUrl":20,"updatedAt":1612},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1705,1706,1707],{"name":1604,"slug":1605,"type":15},{"name":1607,"slug":1608,"type":15},{"name":1610,"slug":1611,"type":15},{"slug":1614,"name":1614,"fn":1615,"description":1616,"org":1709,"tags":1710,"stars":19,"repoUrl":20,"updatedAt":1623},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1711,1712],{"name":1620,"slug":1621,"type":15},{"name":1607,"slug":1608,"type":15},{"slug":1625,"name":1625,"fn":1626,"description":1627,"org":1714,"tags":1715,"stars":19,"repoUrl":20,"updatedAt":1635},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1716,1717,1718],{"name":1620,"slug":1621,"type":15},{"name":1607,"slug":1608,"type":15},{"name":1633,"slug":1634,"type":15},{"slug":1637,"name":1637,"fn":1638,"description":1639,"org":1720,"tags":1721,"stars":19,"repoUrl":20,"updatedAt":1649},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1722,1723,1724],{"name":1643,"slug":1644,"type":15},{"name":1633,"slug":1634,"type":15},{"name":1647,"slug":1648,"type":15},{"slug":1651,"name":1651,"fn":1652,"description":1653,"org":1726,"tags":1727,"stars":19,"repoUrl":20,"updatedAt":1662},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1728,1729],{"name":1657,"slug":1658,"type":15},{"name":1660,"slug":1661,"type":15},{"slug":1664,"name":1664,"fn":1665,"description":1666,"org":1731,"tags":1732,"stars":19,"repoUrl":20,"updatedAt":1676},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1733,1734,1735],{"name":1670,"slug":1671,"type":15},{"name":1673,"slug":1674,"type":15},{"name":9,"slug":8,"type":15},{"slug":4,"name":4,"fn":5,"description":6,"org":1737,"tags":1738,"stars":19,"repoUrl":20,"updatedAt":21},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1739,1740],{"name":13,"slug":14,"type":15},{"name":17,"slug":18,"type":15},{"slug":1742,"name":1742,"fn":1743,"description":1744,"org":1745,"tags":1746,"stars":19,"repoUrl":20,"updatedAt":1751},"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},[1747,1748],{"name":1607,"slug":1608,"type":15},{"name":1749,"slug":1750,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1753,"name":1753,"fn":1754,"description":1755,"org":1756,"tags":1757,"stars":19,"repoUrl":20,"updatedAt":1765},"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},[1758,1761,1762],{"name":1759,"slug":1760,"type":15},"Code Review","code-review",{"name":1633,"slug":1634,"type":15},{"name":1763,"slug":1764,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1767,"name":1767,"fn":1768,"description":1769,"org":1770,"tags":1771,"stars":19,"repoUrl":20,"updatedAt":1774},"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},[1772,1773],{"name":13,"slug":14,"type":15},{"name":17,"slug":18,"type":15},"2026-07-14T05:25:54.928983",{"slug":1776,"name":1776,"fn":1777,"description":1778,"org":1779,"tags":1780,"stars":19,"repoUrl":20,"updatedAt":1785},"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},[1781,1784],{"name":1782,"slug":1783,"type":15},"Automation","automation",{"name":1620,"slug":1621,"type":15},"2026-07-30T05:29:03.275638",496]