[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-mig-configure":3,"mdc--hbwym9-key":34,"related-org-nvidia-mig-configure":1561,"related-repo-nvidia-mig-configure":1720},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"mig-configure","configure NVIDIA MIG GPU partitions","Configure NVIDIA MIG (Multi-Instance GPU) partitions on the DGX Station GB300, including enabling MIG mode, choosing a profile layout, creating instances, and retrieving MIG UUIDs. Use when the user asks to partition the GB300, set up MIG, run multiple models in isolation on one GPU, or reconfigure existing MIG instances.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Performance","performance","tag",{"name":17,"slug":18,"type":15},"Deployment","deployment",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"AI Infrastructure","ai-infrastructure",1144,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fdgx-spark-playbooks","2026-07-14T05:31:02.816078",null,249,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"Collection of step-by-step playbooks for setting up AI\u002FML workloads on NVIDIA DGX Spark devices with Blackwell architecture.","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fdgx-spark-playbooks\u002Ftree\u002FHEAD\u002Fnvidia\u002Fstation-ai-skills\u002Fassets\u002Fskills\u002Fmig-configure","---\nname: mig-configure\ndescription: Configure NVIDIA MIG (Multi-Instance GPU) partitions on the DGX Station GB300, including enabling MIG mode, choosing a profile layout, creating instances, and retrieving MIG UUIDs. Use when the user asks to partition the GB300, set up MIG, run multiple models in isolation on one GPU, or reconfigure existing MIG instances.\nmetadata:\n  publisher: nvidia\n  hardware: DGX Station GB300\n---\n\n# MIG Configuration on DGX Station\n\nConfigure MIG (Multi-Instance GPU) partitions on the DGX Station GB300.\n\n## Steps\n\n1. **Find the GB300 GPU index.** Run:\n   ```bash\n   nvidia-smi --query-gpu=index,name --format=csv,noheader\n   ```\n\n2. **Check current MIG state:**\n   ```bash\n   nvidia-smi -i \u003CGB300_INDEX> -q | grep -i \"MIG Mode\"\n   ```\n\n3. **If MIG is already enabled, show current instances:**\n   ```bash\n   nvidia-smi mig -lgi -i \u003CGB300_INDEX>\n   nvidia-smi mig -lci -i \u003CGB300_INDEX>\n   ```\n   If the user wants to reconfigure, destroy existing instances first (step 6).\n\n4. **If MIG is not enabled, enable it.** All GPU processes must be stopped first:\n   ```bash\n   # Check for running GPU processes\n   sudo fuser -v \u002Fdev\u002Fnvidia*\n\n   # Enable MIG\n   sudo nvidia-smi -i \u003CGB300_INDEX> -mig 1\n\n   # Verify\n   nvidia-smi -i \u003CGB300_INDEX> -q | grep -i \"MIG Mode\"\n   ```\n\n5. **Show available profiles and help the user choose a layout:**\n   ```bash\n   nvidia-smi mig -lgip -i \u003CGB300_INDEX>\n   ```\n\n   Common GB300 MIG profiles:\n\n   | ID | Profile name (driver-dependent) | Approx. memory | Use case |\n   |----|----------------------------------|----------------|----------|\n   | 19 | `1g.35gb` (59x) · `1g.31gb` (61x) | ~30 GB | Small models (7-8B), dev\u002Ftest |\n   | 20 | `1g.35gb+me` · `1g.31gb+me` | ~30 GB | Same + media extensions |\n   | 15 | `1g.70gb` | ~68 GB | Slightly larger inference |\n   | 14 | `2g.70gb` | ~68 GB | Medium models (14-30B) |\n   | 9  | `3g.139gb` (59x) · `3g.126gb` (61x) | ~137 GB | Large models (70B quantized) |\n   | 5  | `4g.139gb` · `4g.126gb` | ~137 GB | Large models, more compute |\n   | 0  | `7g.278gb` (59x) · `7g.251gb` (61x) | ~276 GB | Full GPU as single instance |\n\n   > **Profile names depend on your driver version; the profile IDs do not.** Always read the exact\n   > names and sizes on your box with `nvidia-smi mig -lgip -i \u003CGB300_INDEX>`, and create instances by\n   > **ID**. (Driver 59x reports the `…35gb\u002F139gb\u002F278gb` names; 61x reports `…31gb\u002F126gb\u002F251gb` for the\n   > same IDs.)\n\n   Suggest layouts based on the user's workload (use the stable IDs). Examples:\n   - **Two models (70B + smaller):** one `3g` + two `1g.70gb` → IDs `9,15,15`\n   - **Many small models:** three `1g` → IDs `19,19,19`\n   - **One large model with isolation:** the full `7g` → ID `0`\n\n   > **MIG layouts are constrained by fixed memory-slice placement, not just total memory** — never\n   > sum nominal GB and assume any combination fits. A `3g + 2g + 2g` layout (`9,14,14`) is **not**\n   > realizable, for example, because the second `2g` has no legal placement after a `3g`. And\n   > `nvidia-smi mig -lgip` `Free\u002FTotal` tracks **compute** (GPC) slices, so it overstates the number\n   > of instances you can actually create (QA observed only 3 creatable `1g` instances on a 61x\n   > Station even though `Free\u002FTotal` reported 7). Always validate a specific layout with\n   > `nvidia-smi mig -lgipp` before relying on it.\n\n   Ask the user what models they want to run before suggesting a layout.\n\n6. **Create (or recreate) instances:**\n\n   If reconfiguring, destroy existing instances first:\n   ```bash\n   sudo nvidia-smi mig -dci -i \u003CGB300_INDEX>\n   sudo nvidia-smi mig -dgi -i \u003CGB300_INDEX>\n   ```\n\n   Then create the new layout:\n   ```bash\n   sudo nvidia-smi mig -cgi \u003CPROFILE_IDS> -C -i \u003CGB300_INDEX>\n   ```\n\n7. **Get the MIG device UUIDs:**\n   ```bash\n   nvidia-smi -L\n   ```\n   Note the `MIG-\u003Cuuid>` entries — these are used to target specific MIG instances.\n\n8. **Show the user how to use MIG devices:**\n   ```bash\n   # Bare metal\n   export CUDA_VISIBLE_DEVICES=MIG-\u003Cuuid>\n\n   # Docker\n   docker run --gpus '\"device=MIG-\u003Cuuid>\"' ...\n   ```\n\n9. **Report the final layout** to the user with UUIDs and suggested docker commands for each instance.\n\n## Disabling MIG\n\nIf the user wants to return to full-GPU mode:\n\n```bash\n# Stop all workloads using MIG instances first\nsudo nvidia-smi mig -dci -i \u003CGB300_INDEX>\nsudo nvidia-smi mig -dgi -i \u003CGB300_INDEX>\nsudo nvidia-smi -i \u003CGB300_INDEX> -mig 0\n```\n\n> **Do not run `nvidia-fabricmanager` on DGX Station.** It has a single GB300 over NVLink-C2C (no\n> NVSwitch fabric), so Fabric Manager is not installed and `systemctl start nvidia-fabricmanager`\n> fails with \"Unit not found.\" NVLink-C2C re-initializes automatically after MIG is disabled. If MIG\n> mode is stuck in a \"pending\" state, reset the GPU instead: `sudo nvidia-smi -i \u003CGB300_INDEX> --gpu-reset`.\n",{"data":35,"body":38},{"name":4,"description":6,"metadata":36},{"publisher":8,"hardware":37},"DGX Station GB300",{"type":39,"children":40},"root",[41,50,56,63,1374,1380,1385,1518,1555],{"type":42,"tag":43,"props":44,"children":46},"element","h1",{"id":45},"mig-configuration-on-dgx-station",[47],{"type":48,"value":49},"text","MIG Configuration on DGX Station",{"type":42,"tag":51,"props":52,"children":53},"p",{},[54],{"type":48,"value":55},"Configure MIG (Multi-Instance GPU) partitions on the DGX Station GB300.",{"type":42,"tag":57,"props":58,"children":60},"h2",{"id":59},"steps",[61],{"type":48,"value":62},"Steps",{"type":42,"tag":64,"props":65,"children":66},"ol",{},[67,114,197,293,483,1035,1210,1250,1364],{"type":42,"tag":68,"props":69,"children":70},"li",{},[71,77,79],{"type":42,"tag":72,"props":73,"children":74},"strong",{},[75],{"type":48,"value":76},"Find the GB300 GPU index.",{"type":48,"value":78}," Run:",{"type":42,"tag":80,"props":81,"children":86},"pre",{"className":82,"code":83,"language":84,"meta":85,"style":85},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","nvidia-smi --query-gpu=index,name --format=csv,noheader\n","bash","",[87],{"type":42,"tag":88,"props":89,"children":90},"code",{"__ignoreMap":85},[91],{"type":42,"tag":92,"props":93,"children":96},"span",{"class":94,"line":95},"line",1,[97,103,109],{"type":42,"tag":92,"props":98,"children":100},{"style":99},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[101],{"type":48,"value":102},"nvidia-smi",{"type":42,"tag":92,"props":104,"children":106},{"style":105},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[107],{"type":48,"value":108}," --query-gpu=index,name",{"type":42,"tag":92,"props":110,"children":111},{"style":105},[112],{"type":48,"value":113}," --format=csv,noheader\n",{"type":42,"tag":68,"props":115,"children":116},{},[117,122],{"type":42,"tag":72,"props":118,"children":119},{},[120],{"type":48,"value":121},"Check current MIG state:",{"type":42,"tag":80,"props":123,"children":125},{"className":82,"code":124,"language":84,"meta":85,"style":85},"nvidia-smi -i \u003CGB300_INDEX> -q | grep -i \"MIG Mode\"\n",[126],{"type":42,"tag":88,"props":127,"children":128},{"__ignoreMap":85},[129],{"type":42,"tag":92,"props":130,"children":131},{"class":94,"line":95},[132,136,141,147,152,158,163,168,173,178,182,187,192],{"type":42,"tag":92,"props":133,"children":134},{"style":99},[135],{"type":48,"value":102},{"type":42,"tag":92,"props":137,"children":138},{"style":105},[139],{"type":48,"value":140}," -i",{"type":42,"tag":92,"props":142,"children":144},{"style":143},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[145],{"type":48,"value":146}," \u003C",{"type":42,"tag":92,"props":148,"children":149},{"style":105},[150],{"type":48,"value":151},"GB300_INDE",{"type":42,"tag":92,"props":153,"children":155},{"style":154},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[156],{"type":48,"value":157},"X",{"type":42,"tag":92,"props":159,"children":160},{"style":143},[161],{"type":48,"value":162},">",{"type":42,"tag":92,"props":164,"children":165},{"style":105},[166],{"type":48,"value":167}," -q",{"type":42,"tag":92,"props":169,"children":170},{"style":143},[171],{"type":48,"value":172}," |",{"type":42,"tag":92,"props":174,"children":175},{"style":99},[176],{"type":48,"value":177}," grep",{"type":42,"tag":92,"props":179,"children":180},{"style":105},[181],{"type":48,"value":140},{"type":42,"tag":92,"props":183,"children":184},{"style":143},[185],{"type":48,"value":186}," \"",{"type":42,"tag":92,"props":188,"children":189},{"style":105},[190],{"type":48,"value":191},"MIG Mode",{"type":42,"tag":92,"props":193,"children":194},{"style":143},[195],{"type":48,"value":196},"\"\n",{"type":42,"tag":68,"props":198,"children":199},{},[200,205,287,291],{"type":42,"tag":72,"props":201,"children":202},{},[203],{"type":48,"value":204},"If MIG is already enabled, show current instances:",{"type":42,"tag":80,"props":206,"children":208},{"className":82,"code":207,"language":84,"meta":85,"style":85},"nvidia-smi mig -lgi -i \u003CGB300_INDEX>\nnvidia-smi mig -lci -i \u003CGB300_INDEX>\n",[209],{"type":42,"tag":88,"props":210,"children":211},{"__ignoreMap":85},[212,250],{"type":42,"tag":92,"props":213,"children":214},{"class":94,"line":95},[215,219,224,229,233,237,241,245],{"type":42,"tag":92,"props":216,"children":217},{"style":99},[218],{"type":48,"value":102},{"type":42,"tag":92,"props":220,"children":221},{"style":105},[222],{"type":48,"value":223}," mig",{"type":42,"tag":92,"props":225,"children":226},{"style":105},[227],{"type":48,"value":228}," -lgi",{"type":42,"tag":92,"props":230,"children":231},{"style":105},[232],{"type":48,"value":140},{"type":42,"tag":92,"props":234,"children":235},{"style":143},[236],{"type":48,"value":146},{"type":42,"tag":92,"props":238,"children":239},{"style":105},[240],{"type":48,"value":151},{"type":42,"tag":92,"props":242,"children":243},{"style":154},[244],{"type":48,"value":157},{"type":42,"tag":92,"props":246,"children":247},{"style":143},[248],{"type":48,"value":249},">\n",{"type":42,"tag":92,"props":251,"children":253},{"class":94,"line":252},2,[254,258,262,267,271,275,279,283],{"type":42,"tag":92,"props":255,"children":256},{"style":99},[257],{"type":48,"value":102},{"type":42,"tag":92,"props":259,"children":260},{"style":105},[261],{"type":48,"value":223},{"type":42,"tag":92,"props":263,"children":264},{"style":105},[265],{"type":48,"value":266}," -lci",{"type":42,"tag":92,"props":268,"children":269},{"style":105},[270],{"type":48,"value":140},{"type":42,"tag":92,"props":272,"children":273},{"style":143},[274],{"type":48,"value":146},{"type":42,"tag":92,"props":276,"children":277},{"style":105},[278],{"type":48,"value":151},{"type":42,"tag":92,"props":280,"children":281},{"style":154},[282],{"type":48,"value":157},{"type":42,"tag":92,"props":284,"children":285},{"style":143},[286],{"type":48,"value":249},{"type":42,"tag":288,"props":289,"children":290},"br",{},[],{"type":48,"value":292},"If the user wants to reconfigure, destroy existing instances first (step 6).",{"type":42,"tag":68,"props":294,"children":295},{},[296,301,303],{"type":42,"tag":72,"props":297,"children":298},{},[299],{"type":48,"value":300},"If MIG is not enabled, enable it.",{"type":48,"value":302}," All GPU processes must be stopped first:",{"type":42,"tag":80,"props":304,"children":306},{"className":82,"code":305,"language":84,"meta":85,"style":85},"# Check for running GPU processes\nsudo fuser -v \u002Fdev\u002Fnvidia*\n\n# Enable MIG\nsudo nvidia-smi -i \u003CGB300_INDEX> -mig 1\n\n# Verify\nnvidia-smi -i \u003CGB300_INDEX> -q | grep -i \"MIG Mode\"\n",[307],{"type":42,"tag":88,"props":308,"children":309},{"__ignoreMap":85},[310,319,347,357,366,410,418,427],{"type":42,"tag":92,"props":311,"children":312},{"class":94,"line":95},[313],{"type":42,"tag":92,"props":314,"children":316},{"style":315},"--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic",[317],{"type":48,"value":318},"# Check for running GPU processes\n",{"type":42,"tag":92,"props":320,"children":321},{"class":94,"line":252},[322,327,332,337,342],{"type":42,"tag":92,"props":323,"children":324},{"style":99},[325],{"type":48,"value":326},"sudo",{"type":42,"tag":92,"props":328,"children":329},{"style":105},[330],{"type":48,"value":331}," fuser",{"type":42,"tag":92,"props":333,"children":334},{"style":105},[335],{"type":48,"value":336}," -v",{"type":42,"tag":92,"props":338,"children":339},{"style":105},[340],{"type":48,"value":341}," \u002Fdev\u002Fnvidia",{"type":42,"tag":92,"props":343,"children":344},{"style":154},[345],{"type":48,"value":346},"*\n",{"type":42,"tag":92,"props":348,"children":350},{"class":94,"line":349},3,[351],{"type":42,"tag":92,"props":352,"children":354},{"emptyLinePlaceholder":353},true,[355],{"type":48,"value":356},"\n",{"type":42,"tag":92,"props":358,"children":360},{"class":94,"line":359},4,[361],{"type":42,"tag":92,"props":362,"children":363},{"style":315},[364],{"type":48,"value":365},"# Enable MIG\n",{"type":42,"tag":92,"props":367,"children":369},{"class":94,"line":368},5,[370,374,379,383,387,391,395,399,404],{"type":42,"tag":92,"props":371,"children":372},{"style":99},[373],{"type":48,"value":326},{"type":42,"tag":92,"props":375,"children":376},{"style":105},[377],{"type":48,"value":378}," nvidia-smi",{"type":42,"tag":92,"props":380,"children":381},{"style":105},[382],{"type":48,"value":140},{"type":42,"tag":92,"props":384,"children":385},{"style":143},[386],{"type":48,"value":146},{"type":42,"tag":92,"props":388,"children":389},{"style":105},[390],{"type":48,"value":151},{"type":42,"tag":92,"props":392,"children":393},{"style":154},[394],{"type":48,"value":157},{"type":42,"tag":92,"props":396,"children":397},{"style":143},[398],{"type":48,"value":162},{"type":42,"tag":92,"props":400,"children":401},{"style":105},[402],{"type":48,"value":403}," -mig",{"type":42,"tag":92,"props":405,"children":407},{"style":406},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[408],{"type":48,"value":409}," 1\n",{"type":42,"tag":92,"props":411,"children":413},{"class":94,"line":412},6,[414],{"type":42,"tag":92,"props":415,"children":416},{"emptyLinePlaceholder":353},[417],{"type":48,"value":356},{"type":42,"tag":92,"props":419,"children":421},{"class":94,"line":420},7,[422],{"type":42,"tag":92,"props":423,"children":424},{"style":315},[425],{"type":48,"value":426},"# Verify\n",{"type":42,"tag":92,"props":428,"children":430},{"class":94,"line":429},8,[431,435,439,443,447,451,455,459,463,467,471,475,479],{"type":42,"tag":92,"props":432,"children":433},{"style":99},[434],{"type":48,"value":102},{"type":42,"tag":92,"props":436,"children":437},{"style":105},[438],{"type":48,"value":140},{"type":42,"tag":92,"props":440,"children":441},{"style":143},[442],{"type":48,"value":146},{"type":42,"tag":92,"props":444,"children":445},{"style":105},[446],{"type":48,"value":151},{"type":42,"tag":92,"props":448,"children":449},{"style":154},[450],{"type":48,"value":157},{"type":42,"tag":92,"props":452,"children":453},{"style":143},[454],{"type":48,"value":162},{"type":42,"tag":92,"props":456,"children":457},{"style":105},[458],{"type":48,"value":167},{"type":42,"tag":92,"props":460,"children":461},{"style":143},[462],{"type":48,"value":172},{"type":42,"tag":92,"props":464,"children":465},{"style":99},[466],{"type":48,"value":177},{"type":42,"tag":92,"props":468,"children":469},{"style":105},[470],{"type":48,"value":140},{"type":42,"tag":92,"props":472,"children":473},{"style":143},[474],{"type":48,"value":186},{"type":42,"tag":92,"props":476,"children":477},{"style":105},[478],{"type":48,"value":191},{"type":42,"tag":92,"props":480,"children":481},{"style":143},[482],{"type":48,"value":196},{"type":42,"tag":68,"props":484,"children":485},{},[486,491,534,537,539,804,848,851,853,934,1030,1033],{"type":42,"tag":72,"props":487,"children":488},{},[489],{"type":48,"value":490},"Show available profiles and help the user choose a layout:",{"type":42,"tag":80,"props":492,"children":494},{"className":82,"code":493,"language":84,"meta":85,"style":85},"nvidia-smi mig -lgip -i \u003CGB300_INDEX>\n",[495],{"type":42,"tag":88,"props":496,"children":497},{"__ignoreMap":85},[498],{"type":42,"tag":92,"props":499,"children":500},{"class":94,"line":95},[501,505,509,514,518,522,526,530],{"type":42,"tag":92,"props":502,"children":503},{"style":99},[504],{"type":48,"value":102},{"type":42,"tag":92,"props":506,"children":507},{"style":105},[508],{"type":48,"value":223},{"type":42,"tag":92,"props":510,"children":511},{"style":105},[512],{"type":48,"value":513}," -lgip",{"type":42,"tag":92,"props":515,"children":516},{"style":105},[517],{"type":48,"value":140},{"type":42,"tag":92,"props":519,"children":520},{"style":143},[521],{"type":48,"value":146},{"type":42,"tag":92,"props":523,"children":524},{"style":105},[525],{"type":48,"value":151},{"type":42,"tag":92,"props":527,"children":528},{"style":154},[529],{"type":48,"value":157},{"type":42,"tag":92,"props":531,"children":532},{"style":143},[533],{"type":48,"value":249},{"type":42,"tag":288,"props":535,"children":536},{},[],{"type":48,"value":538},"Common GB300 MIG profiles:",{"type":42,"tag":540,"props":541,"children":542},"table",{},[543,572],{"type":42,"tag":544,"props":545,"children":546},"thead",{},[547],{"type":42,"tag":548,"props":549,"children":550},"tr",{},[551,557,562,567],{"type":42,"tag":552,"props":553,"children":554},"th",{},[555],{"type":48,"value":556},"ID",{"type":42,"tag":552,"props":558,"children":559},{},[560],{"type":48,"value":561},"Profile name (driver-dependent)",{"type":42,"tag":552,"props":563,"children":564},{},[565],{"type":48,"value":566},"Approx. memory",{"type":42,"tag":552,"props":568,"children":569},{},[570],{"type":48,"value":571},"Use case",{"type":42,"tag":573,"props":574,"children":575},"tbody",{},[576,614,648,675,701,736,769],{"type":42,"tag":548,"props":577,"children":578},{},[579,585,604,609],{"type":42,"tag":580,"props":581,"children":582},"td",{},[583],{"type":48,"value":584},"19",{"type":42,"tag":580,"props":586,"children":587},{},[588,594,596,602],{"type":42,"tag":88,"props":589,"children":591},{"className":590},[],[592],{"type":48,"value":593},"1g.35gb",{"type":48,"value":595}," (59x) · ",{"type":42,"tag":88,"props":597,"children":599},{"className":598},[],[600],{"type":48,"value":601},"1g.31gb",{"type":48,"value":603}," (61x)",{"type":42,"tag":580,"props":605,"children":606},{},[607],{"type":48,"value":608},"~30 GB",{"type":42,"tag":580,"props":610,"children":611},{},[612],{"type":48,"value":613},"Small models (7-8B), dev\u002Ftest",{"type":42,"tag":548,"props":615,"children":616},{},[617,622,639,643],{"type":42,"tag":580,"props":618,"children":619},{},[620],{"type":48,"value":621},"20",{"type":42,"tag":580,"props":623,"children":624},{},[625,631,633],{"type":42,"tag":88,"props":626,"children":628},{"className":627},[],[629],{"type":48,"value":630},"1g.35gb+me",{"type":48,"value":632}," · ",{"type":42,"tag":88,"props":634,"children":636},{"className":635},[],[637],{"type":48,"value":638},"1g.31gb+me",{"type":42,"tag":580,"props":640,"children":641},{},[642],{"type":48,"value":608},{"type":42,"tag":580,"props":644,"children":645},{},[646],{"type":48,"value":647},"Same + media extensions",{"type":42,"tag":548,"props":649,"children":650},{},[651,656,665,670],{"type":42,"tag":580,"props":652,"children":653},{},[654],{"type":48,"value":655},"15",{"type":42,"tag":580,"props":657,"children":658},{},[659],{"type":42,"tag":88,"props":660,"children":662},{"className":661},[],[663],{"type":48,"value":664},"1g.70gb",{"type":42,"tag":580,"props":666,"children":667},{},[668],{"type":48,"value":669},"~68 GB",{"type":42,"tag":580,"props":671,"children":672},{},[673],{"type":48,"value":674},"Slightly larger inference",{"type":42,"tag":548,"props":676,"children":677},{},[678,683,692,696],{"type":42,"tag":580,"props":679,"children":680},{},[681],{"type":48,"value":682},"14",{"type":42,"tag":580,"props":684,"children":685},{},[686],{"type":42,"tag":88,"props":687,"children":689},{"className":688},[],[690],{"type":48,"value":691},"2g.70gb",{"type":42,"tag":580,"props":693,"children":694},{},[695],{"type":48,"value":669},{"type":42,"tag":580,"props":697,"children":698},{},[699],{"type":48,"value":700},"Medium models (14-30B)",{"type":42,"tag":548,"props":702,"children":703},{},[704,709,726,731],{"type":42,"tag":580,"props":705,"children":706},{},[707],{"type":48,"value":708},"9",{"type":42,"tag":580,"props":710,"children":711},{},[712,718,719,725],{"type":42,"tag":88,"props":713,"children":715},{"className":714},[],[716],{"type":48,"value":717},"3g.139gb",{"type":48,"value":595},{"type":42,"tag":88,"props":720,"children":722},{"className":721},[],[723],{"type":48,"value":724},"3g.126gb",{"type":48,"value":603},{"type":42,"tag":580,"props":727,"children":728},{},[729],{"type":48,"value":730},"~137 GB",{"type":42,"tag":580,"props":732,"children":733},{},[734],{"type":48,"value":735},"Large models (70B quantized)",{"type":42,"tag":548,"props":737,"children":738},{},[739,744,760,764],{"type":42,"tag":580,"props":740,"children":741},{},[742],{"type":48,"value":743},"5",{"type":42,"tag":580,"props":745,"children":746},{},[747,753,754],{"type":42,"tag":88,"props":748,"children":750},{"className":749},[],[751],{"type":48,"value":752},"4g.139gb",{"type":48,"value":632},{"type":42,"tag":88,"props":755,"children":757},{"className":756},[],[758],{"type":48,"value":759},"4g.126gb",{"type":42,"tag":580,"props":761,"children":762},{},[763],{"type":48,"value":730},{"type":42,"tag":580,"props":765,"children":766},{},[767],{"type":48,"value":768},"Large models, more compute",{"type":42,"tag":548,"props":770,"children":771},{},[772,777,794,799],{"type":42,"tag":580,"props":773,"children":774},{},[775],{"type":48,"value":776},"0",{"type":42,"tag":580,"props":778,"children":779},{},[780,786,787,793],{"type":42,"tag":88,"props":781,"children":783},{"className":782},[],[784],{"type":48,"value":785},"7g.278gb",{"type":48,"value":595},{"type":42,"tag":88,"props":788,"children":790},{"className":789},[],[791],{"type":48,"value":792},"7g.251gb",{"type":48,"value":603},{"type":42,"tag":580,"props":795,"children":796},{},[797],{"type":48,"value":798},"~276 GB",{"type":42,"tag":580,"props":800,"children":801},{},[802],{"type":48,"value":803},"Full GPU as single instance",{"type":42,"tag":805,"props":806,"children":807},"blockquote",{},[808],{"type":42,"tag":51,"props":809,"children":810},{},[811,816,818,824,826,830,832,838,840,846],{"type":42,"tag":72,"props":812,"children":813},{},[814],{"type":48,"value":815},"Profile names depend on your driver version; the profile IDs do not.",{"type":48,"value":817}," Always read the exact\nnames and sizes on your box with ",{"type":42,"tag":88,"props":819,"children":821},{"className":820},[],[822],{"type":48,"value":823},"nvidia-smi mig -lgip -i \u003CGB300_INDEX>",{"type":48,"value":825},", and create instances by\n",{"type":42,"tag":72,"props":827,"children":828},{},[829],{"type":48,"value":556},{"type":48,"value":831},". (Driver 59x reports the ",{"type":42,"tag":88,"props":833,"children":835},{"className":834},[],[836],{"type":48,"value":837},"…35gb\u002F139gb\u002F278gb",{"type":48,"value":839}," names; 61x reports ",{"type":42,"tag":88,"props":841,"children":843},{"className":842},[],[844],{"type":48,"value":845},"…31gb\u002F126gb\u002F251gb",{"type":48,"value":847}," for the\nsame IDs.)",{"type":42,"tag":288,"props":849,"children":850},{},[],{"type":48,"value":852},"Suggest layouts based on the user's workload (use the stable IDs). Examples:",{"type":42,"tag":854,"props":855,"children":856},"ul",{},[857,888,911],{"type":42,"tag":68,"props":858,"children":859},{},[860,865,867,873,875,880,882],{"type":42,"tag":72,"props":861,"children":862},{},[863],{"type":48,"value":864},"Two models (70B + smaller):",{"type":48,"value":866}," one ",{"type":42,"tag":88,"props":868,"children":870},{"className":869},[],[871],{"type":48,"value":872},"3g",{"type":48,"value":874}," + two ",{"type":42,"tag":88,"props":876,"children":878},{"className":877},[],[879],{"type":48,"value":664},{"type":48,"value":881}," → IDs ",{"type":42,"tag":88,"props":883,"children":885},{"className":884},[],[886],{"type":48,"value":887},"9,15,15",{"type":42,"tag":68,"props":889,"children":890},{},[891,896,898,904,905],{"type":42,"tag":72,"props":892,"children":893},{},[894],{"type":48,"value":895},"Many small models:",{"type":48,"value":897}," three ",{"type":42,"tag":88,"props":899,"children":901},{"className":900},[],[902],{"type":48,"value":903},"1g",{"type":48,"value":881},{"type":42,"tag":88,"props":906,"children":908},{"className":907},[],[909],{"type":48,"value":910},"19,19,19",{"type":42,"tag":68,"props":912,"children":913},{},[914,919,921,927,929],{"type":42,"tag":72,"props":915,"children":916},{},[917],{"type":48,"value":918},"One large model with isolation:",{"type":48,"value":920}," the full ",{"type":42,"tag":88,"props":922,"children":924},{"className":923},[],[925],{"type":48,"value":926},"7g",{"type":48,"value":928}," → ID ",{"type":42,"tag":88,"props":930,"children":932},{"className":931},[],[933],{"type":48,"value":776},{"type":42,"tag":805,"props":935,"children":936},{},[937],{"type":42,"tag":51,"props":938,"children":939},{},[940,945,947,953,955,961,963,968,970,976,978,983,985,991,993,999,1001,1006,1008,1013,1015,1020,1022,1028],{"type":42,"tag":72,"props":941,"children":942},{},[943],{"type":48,"value":944},"MIG layouts are constrained by fixed memory-slice placement, not just total memory",{"type":48,"value":946}," — never\nsum nominal GB and assume any combination fits. A ",{"type":42,"tag":88,"props":948,"children":950},{"className":949},[],[951],{"type":48,"value":952},"3g + 2g + 2g",{"type":48,"value":954}," layout (",{"type":42,"tag":88,"props":956,"children":958},{"className":957},[],[959],{"type":48,"value":960},"9,14,14",{"type":48,"value":962},") is ",{"type":42,"tag":72,"props":964,"children":965},{},[966],{"type":48,"value":967},"not",{"type":48,"value":969},"\nrealizable, for example, because the second ",{"type":42,"tag":88,"props":971,"children":973},{"className":972},[],[974],{"type":48,"value":975},"2g",{"type":48,"value":977}," has no legal placement after a ",{"type":42,"tag":88,"props":979,"children":981},{"className":980},[],[982],{"type":48,"value":872},{"type":48,"value":984},". And\n",{"type":42,"tag":88,"props":986,"children":988},{"className":987},[],[989],{"type":48,"value":990},"nvidia-smi mig -lgip",{"type":48,"value":992}," ",{"type":42,"tag":88,"props":994,"children":996},{"className":995},[],[997],{"type":48,"value":998},"Free\u002FTotal",{"type":48,"value":1000}," tracks ",{"type":42,"tag":72,"props":1002,"children":1003},{},[1004],{"type":48,"value":1005},"compute",{"type":48,"value":1007}," (GPC) slices, so it overstates the number\nof instances you can actually create (QA observed only 3 creatable ",{"type":42,"tag":88,"props":1009,"children":1011},{"className":1010},[],[1012],{"type":48,"value":903},{"type":48,"value":1014}," instances on a 61x\nStation even though ",{"type":42,"tag":88,"props":1016,"children":1018},{"className":1017},[],[1019],{"type":48,"value":998},{"type":48,"value":1021}," reported 7). Always validate a specific layout with\n",{"type":42,"tag":88,"props":1023,"children":1025},{"className":1024},[],[1026],{"type":48,"value":1027},"nvidia-smi mig -lgipp",{"type":48,"value":1029}," before relying on it.",{"type":42,"tag":288,"props":1031,"children":1032},{},[],{"type":48,"value":1034},"Ask the user what models they want to run before suggesting a layout.",{"type":42,"tag":68,"props":1036,"children":1037},{},[1038,1043,1046,1048,1135,1138,1140],{"type":42,"tag":72,"props":1039,"children":1040},{},[1041],{"type":48,"value":1042},"Create (or recreate) instances:",{"type":42,"tag":288,"props":1044,"children":1045},{},[],{"type":48,"value":1047},"If reconfiguring, destroy existing instances first:",{"type":42,"tag":80,"props":1049,"children":1051},{"className":82,"code":1050,"language":84,"meta":85,"style":85},"sudo nvidia-smi mig -dci -i \u003CGB300_INDEX>\nsudo nvidia-smi mig -dgi -i \u003CGB300_INDEX>\n",[1052],{"type":42,"tag":88,"props":1053,"children":1054},{"__ignoreMap":85},[1055,1095],{"type":42,"tag":92,"props":1056,"children":1057},{"class":94,"line":95},[1058,1062,1066,1070,1075,1079,1083,1087,1091],{"type":42,"tag":92,"props":1059,"children":1060},{"style":99},[1061],{"type":48,"value":326},{"type":42,"tag":92,"props":1063,"children":1064},{"style":105},[1065],{"type":48,"value":378},{"type":42,"tag":92,"props":1067,"children":1068},{"style":105},[1069],{"type":48,"value":223},{"type":42,"tag":92,"props":1071,"children":1072},{"style":105},[1073],{"type":48,"value":1074}," -dci",{"type":42,"tag":92,"props":1076,"children":1077},{"style":105},[1078],{"type":48,"value":140},{"type":42,"tag":92,"props":1080,"children":1081},{"style":143},[1082],{"type":48,"value":146},{"type":42,"tag":92,"props":1084,"children":1085},{"style":105},[1086],{"type":48,"value":151},{"type":42,"tag":92,"props":1088,"children":1089},{"style":154},[1090],{"type":48,"value":157},{"type":42,"tag":92,"props":1092,"children":1093},{"style":143},[1094],{"type":48,"value":249},{"type":42,"tag":92,"props":1096,"children":1097},{"class":94,"line":252},[1098,1102,1106,1110,1115,1119,1123,1127,1131],{"type":42,"tag":92,"props":1099,"children":1100},{"style":99},[1101],{"type":48,"value":326},{"type":42,"tag":92,"props":1103,"children":1104},{"style":105},[1105],{"type":48,"value":378},{"type":42,"tag":92,"props":1107,"children":1108},{"style":105},[1109],{"type":48,"value":223},{"type":42,"tag":92,"props":1111,"children":1112},{"style":105},[1113],{"type":48,"value":1114}," -dgi",{"type":42,"tag":92,"props":1116,"children":1117},{"style":105},[1118],{"type":48,"value":140},{"type":42,"tag":92,"props":1120,"children":1121},{"style":143},[1122],{"type":48,"value":146},{"type":42,"tag":92,"props":1124,"children":1125},{"style":105},[1126],{"type":48,"value":151},{"type":42,"tag":92,"props":1128,"children":1129},{"style":154},[1130],{"type":48,"value":157},{"type":42,"tag":92,"props":1132,"children":1133},{"style":143},[1134],{"type":48,"value":249},{"type":42,"tag":288,"props":1136,"children":1137},{},[],{"type":48,"value":1139},"Then create the new layout:",{"type":42,"tag":80,"props":1141,"children":1143},{"className":82,"code":1142,"language":84,"meta":85,"style":85},"sudo nvidia-smi mig -cgi \u003CPROFILE_IDS> -C -i \u003CGB300_INDEX>\n",[1144],{"type":42,"tag":88,"props":1145,"children":1146},{"__ignoreMap":85},[1147],{"type":42,"tag":92,"props":1148,"children":1149},{"class":94,"line":95},[1150,1154,1158,1162,1167,1171,1176,1181,1185,1190,1194,1198,1202,1206],{"type":42,"tag":92,"props":1151,"children":1152},{"style":99},[1153],{"type":48,"value":326},{"type":42,"tag":92,"props":1155,"children":1156},{"style":105},[1157],{"type":48,"value":378},{"type":42,"tag":92,"props":1159,"children":1160},{"style":105},[1161],{"type":48,"value":223},{"type":42,"tag":92,"props":1163,"children":1164},{"style":105},[1165],{"type":48,"value":1166}," -cgi",{"type":42,"tag":92,"props":1168,"children":1169},{"style":143},[1170],{"type":48,"value":146},{"type":42,"tag":92,"props":1172,"children":1173},{"style":105},[1174],{"type":48,"value":1175},"PROFILE_ID",{"type":42,"tag":92,"props":1177,"children":1178},{"style":154},[1179],{"type":48,"value":1180},"S",{"type":42,"tag":92,"props":1182,"children":1183},{"style":143},[1184],{"type":48,"value":162},{"type":42,"tag":92,"props":1186,"children":1187},{"style":105},[1188],{"type":48,"value":1189}," -C",{"type":42,"tag":92,"props":1191,"children":1192},{"style":105},[1193],{"type":48,"value":140},{"type":42,"tag":92,"props":1195,"children":1196},{"style":143},[1197],{"type":48,"value":146},{"type":42,"tag":92,"props":1199,"children":1200},{"style":105},[1201],{"type":48,"value":151},{"type":42,"tag":92,"props":1203,"children":1204},{"style":154},[1205],{"type":48,"value":157},{"type":42,"tag":92,"props":1207,"children":1208},{"style":143},[1209],{"type":48,"value":249},{"type":42,"tag":68,"props":1211,"children":1212},{},[1213,1218,1237,1240,1242,1248],{"type":42,"tag":72,"props":1214,"children":1215},{},[1216],{"type":48,"value":1217},"Get the MIG device UUIDs:",{"type":42,"tag":80,"props":1219,"children":1221},{"className":82,"code":1220,"language":84,"meta":85,"style":85},"nvidia-smi -L\n",[1222],{"type":42,"tag":88,"props":1223,"children":1224},{"__ignoreMap":85},[1225],{"type":42,"tag":92,"props":1226,"children":1227},{"class":94,"line":95},[1228,1232],{"type":42,"tag":92,"props":1229,"children":1230},{"style":99},[1231],{"type":48,"value":102},{"type":42,"tag":92,"props":1233,"children":1234},{"style":105},[1235],{"type":48,"value":1236}," -L\n",{"type":42,"tag":288,"props":1238,"children":1239},{},[],{"type":48,"value":1241},"Note the ",{"type":42,"tag":88,"props":1243,"children":1245},{"className":1244},[],[1246],{"type":48,"value":1247},"MIG-\u003Cuuid>",{"type":48,"value":1249}," entries — these are used to target specific MIG instances.",{"type":42,"tag":68,"props":1251,"children":1252},{},[1253,1258],{"type":42,"tag":72,"props":1254,"children":1255},{},[1256],{"type":48,"value":1257},"Show the user how to use MIG devices:",{"type":42,"tag":80,"props":1259,"children":1261},{"className":82,"code":1260,"language":84,"meta":85,"style":85},"# Bare metal\nexport CUDA_VISIBLE_DEVICES=MIG-\u003Cuuid>\n\n# Docker\ndocker run --gpus '\"device=MIG-\u003Cuuid>\"' ...\n",[1262],{"type":42,"tag":88,"props":1263,"children":1264},{"__ignoreMap":85},[1265,1273,1311,1318,1326],{"type":42,"tag":92,"props":1266,"children":1267},{"class":94,"line":95},[1268],{"type":42,"tag":92,"props":1269,"children":1270},{"style":315},[1271],{"type":48,"value":1272},"# Bare metal\n",{"type":42,"tag":92,"props":1274,"children":1275},{"class":94,"line":252},[1276,1282,1287,1292,1297,1302,1307],{"type":42,"tag":92,"props":1277,"children":1279},{"style":1278},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[1280],{"type":48,"value":1281},"export",{"type":42,"tag":92,"props":1283,"children":1284},{"style":154},[1285],{"type":48,"value":1286}," CUDA_VISIBLE_DEVICES",{"type":42,"tag":92,"props":1288,"children":1289},{"style":143},[1290],{"type":48,"value":1291},"=",{"type":42,"tag":92,"props":1293,"children":1294},{"style":154},[1295],{"type":48,"value":1296},"MIG-",{"type":42,"tag":92,"props":1298,"children":1299},{"style":143},[1300],{"type":48,"value":1301},"\u003C",{"type":42,"tag":92,"props":1303,"children":1304},{"style":154},[1305],{"type":48,"value":1306},"uuid",{"type":42,"tag":92,"props":1308,"children":1309},{"style":143},[1310],{"type":48,"value":249},{"type":42,"tag":92,"props":1312,"children":1313},{"class":94,"line":349},[1314],{"type":42,"tag":92,"props":1315,"children":1316},{"emptyLinePlaceholder":353},[1317],{"type":48,"value":356},{"type":42,"tag":92,"props":1319,"children":1320},{"class":94,"line":359},[1321],{"type":42,"tag":92,"props":1322,"children":1323},{"style":315},[1324],{"type":48,"value":1325},"# Docker\n",{"type":42,"tag":92,"props":1327,"children":1328},{"class":94,"line":368},[1329,1334,1339,1344,1349,1354,1359],{"type":42,"tag":92,"props":1330,"children":1331},{"style":99},[1332],{"type":48,"value":1333},"docker",{"type":42,"tag":92,"props":1335,"children":1336},{"style":105},[1337],{"type":48,"value":1338}," run",{"type":42,"tag":92,"props":1340,"children":1341},{"style":105},[1342],{"type":48,"value":1343}," --gpus",{"type":42,"tag":92,"props":1345,"children":1346},{"style":143},[1347],{"type":48,"value":1348}," '",{"type":42,"tag":92,"props":1350,"children":1351},{"style":105},[1352],{"type":48,"value":1353},"\"device=MIG-\u003Cuuid>\"",{"type":42,"tag":92,"props":1355,"children":1356},{"style":143},[1357],{"type":48,"value":1358},"'",{"type":42,"tag":92,"props":1360,"children":1361},{"style":105},[1362],{"type":48,"value":1363}," ...\n",{"type":42,"tag":68,"props":1365,"children":1366},{},[1367,1372],{"type":42,"tag":72,"props":1368,"children":1369},{},[1370],{"type":48,"value":1371},"Report the final layout",{"type":48,"value":1373}," to the user with UUIDs and suggested docker commands for each instance.",{"type":42,"tag":57,"props":1375,"children":1377},{"id":1376},"disabling-mig",[1378],{"type":48,"value":1379},"Disabling MIG",{"type":42,"tag":51,"props":1381,"children":1382},{},[1383],{"type":48,"value":1384},"If the user wants to return to full-GPU mode:",{"type":42,"tag":80,"props":1386,"children":1388},{"className":82,"code":1387,"language":84,"meta":85,"style":85},"# Stop all workloads using MIG instances first\nsudo nvidia-smi mig -dci -i \u003CGB300_INDEX>\nsudo nvidia-smi mig -dgi -i \u003CGB300_INDEX>\nsudo nvidia-smi -i \u003CGB300_INDEX> -mig 0\n",[1389],{"type":42,"tag":88,"props":1390,"children":1391},{"__ignoreMap":85},[1392,1400,1439,1478],{"type":42,"tag":92,"props":1393,"children":1394},{"class":94,"line":95},[1395],{"type":42,"tag":92,"props":1396,"children":1397},{"style":315},[1398],{"type":48,"value":1399},"# Stop all workloads using MIG instances first\n",{"type":42,"tag":92,"props":1401,"children":1402},{"class":94,"line":252},[1403,1407,1411,1415,1419,1423,1427,1431,1435],{"type":42,"tag":92,"props":1404,"children":1405},{"style":99},[1406],{"type":48,"value":326},{"type":42,"tag":92,"props":1408,"children":1409},{"style":105},[1410],{"type":48,"value":378},{"type":42,"tag":92,"props":1412,"children":1413},{"style":105},[1414],{"type":48,"value":223},{"type":42,"tag":92,"props":1416,"children":1417},{"style":105},[1418],{"type":48,"value":1074},{"type":42,"tag":92,"props":1420,"children":1421},{"style":105},[1422],{"type":48,"value":140},{"type":42,"tag":92,"props":1424,"children":1425},{"style":143},[1426],{"type":48,"value":146},{"type":42,"tag":92,"props":1428,"children":1429},{"style":105},[1430],{"type":48,"value":151},{"type":42,"tag":92,"props":1432,"children":1433},{"style":154},[1434],{"type":48,"value":157},{"type":42,"tag":92,"props":1436,"children":1437},{"style":143},[1438],{"type":48,"value":249},{"type":42,"tag":92,"props":1440,"children":1441},{"class":94,"line":349},[1442,1446,1450,1454,1458,1462,1466,1470,1474],{"type":42,"tag":92,"props":1443,"children":1444},{"style":99},[1445],{"type":48,"value":326},{"type":42,"tag":92,"props":1447,"children":1448},{"style":105},[1449],{"type":48,"value":378},{"type":42,"tag":92,"props":1451,"children":1452},{"style":105},[1453],{"type":48,"value":223},{"type":42,"tag":92,"props":1455,"children":1456},{"style":105},[1457],{"type":48,"value":1114},{"type":42,"tag":92,"props":1459,"children":1460},{"style":105},[1461],{"type":48,"value":140},{"type":42,"tag":92,"props":1463,"children":1464},{"style":143},[1465],{"type":48,"value":146},{"type":42,"tag":92,"props":1467,"children":1468},{"style":105},[1469],{"type":48,"value":151},{"type":42,"tag":92,"props":1471,"children":1472},{"style":154},[1473],{"type":48,"value":157},{"type":42,"tag":92,"props":1475,"children":1476},{"style":143},[1477],{"type":48,"value":249},{"type":42,"tag":92,"props":1479,"children":1480},{"class":94,"line":359},[1481,1485,1489,1493,1497,1501,1505,1509,1513],{"type":42,"tag":92,"props":1482,"children":1483},{"style":99},[1484],{"type":48,"value":326},{"type":42,"tag":92,"props":1486,"children":1487},{"style":105},[1488],{"type":48,"value":378},{"type":42,"tag":92,"props":1490,"children":1491},{"style":105},[1492],{"type":48,"value":140},{"type":42,"tag":92,"props":1494,"children":1495},{"style":143},[1496],{"type":48,"value":146},{"type":42,"tag":92,"props":1498,"children":1499},{"style":105},[1500],{"type":48,"value":151},{"type":42,"tag":92,"props":1502,"children":1503},{"style":154},[1504],{"type":48,"value":157},{"type":42,"tag":92,"props":1506,"children":1507},{"style":143},[1508],{"type":48,"value":162},{"type":42,"tag":92,"props":1510,"children":1511},{"style":105},[1512],{"type":48,"value":403},{"type":42,"tag":92,"props":1514,"children":1515},{"style":406},[1516],{"type":48,"value":1517}," 0\n",{"type":42,"tag":805,"props":1519,"children":1520},{},[1521],{"type":42,"tag":51,"props":1522,"children":1523},{},[1524,1537,1539,1545,1547,1553],{"type":42,"tag":72,"props":1525,"children":1526},{},[1527,1529,1535],{"type":48,"value":1528},"Do not run ",{"type":42,"tag":88,"props":1530,"children":1532},{"className":1531},[],[1533],{"type":48,"value":1534},"nvidia-fabricmanager",{"type":48,"value":1536}," on DGX Station.",{"type":48,"value":1538}," It has a single GB300 over NVLink-C2C (no\nNVSwitch fabric), so Fabric Manager is not installed and ",{"type":42,"tag":88,"props":1540,"children":1542},{"className":1541},[],[1543],{"type":48,"value":1544},"systemctl start nvidia-fabricmanager",{"type":48,"value":1546},"\nfails with \"Unit not found.\" NVLink-C2C re-initializes automatically after MIG is disabled. If MIG\nmode is stuck in a \"pending\" state, reset the GPU instead: ",{"type":42,"tag":88,"props":1548,"children":1550},{"className":1549},[],[1551],{"type":48,"value":1552},"sudo nvidia-smi -i \u003CGB300_INDEX> --gpu-reset",{"type":48,"value":1554},".",{"type":42,"tag":1556,"props":1557,"children":1558},"style",{},[1559],{"type":48,"value":1560},"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":1562,"total":1719},[1563,1581,1597,1608,1620,1634,1647,1661,1674,1685,1699,1708],{"slug":1564,"name":1564,"fn":1565,"description":1566,"org":1567,"tags":1568,"stars":1578,"repoUrl":1579,"updatedAt":1580},"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},[1569,1572,1575],{"name":1570,"slug":1571,"type":15},"Documentation","documentation",{"name":1573,"slug":1574,"type":15},"MCP","mcp",{"name":1576,"slug":1577,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1582,"name":1582,"fn":1583,"description":1584,"org":1585,"tags":1586,"stars":1594,"repoUrl":1595,"updatedAt":1596},"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},[1587,1590,1591],{"name":1588,"slug":1589,"type":15},"Containers","containers",{"name":17,"slug":18,"type":15},{"name":1592,"slug":1593,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1598,"name":1598,"fn":1599,"description":1600,"org":1601,"tags":1602,"stars":1594,"repoUrl":1595,"updatedAt":1607},"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},[1603,1606],{"name":1604,"slug":1605,"type":15},"CI\u002FCD","ci-cd",{"name":17,"slug":18,"type":15},"2026-07-14T05:25:59.97109",{"slug":1609,"name":1609,"fn":1610,"description":1611,"org":1612,"tags":1613,"stars":1594,"repoUrl":1595,"updatedAt":1619},"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},[1614,1615,1616],{"name":1604,"slug":1605,"type":15},{"name":17,"slug":18,"type":15},{"name":1617,"slug":1618,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1621,"name":1621,"fn":1622,"description":1623,"org":1624,"tags":1625,"stars":1594,"repoUrl":1595,"updatedAt":1633},"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},[1626,1629,1630],{"name":1627,"slug":1628,"type":15},"Debugging","debugging",{"name":1617,"slug":1618,"type":15},{"name":1631,"slug":1632,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1635,"name":1635,"fn":1636,"description":1637,"org":1638,"tags":1639,"stars":1594,"repoUrl":1595,"updatedAt":1646},"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},[1640,1643],{"name":1641,"slug":1642,"type":15},"Best Practices","best-practices",{"name":1644,"slug":1645,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1648,"name":1648,"fn":1649,"description":1650,"org":1651,"tags":1652,"stars":1594,"repoUrl":1595,"updatedAt":1660},"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},[1653,1656,1659],{"name":1654,"slug":1655,"type":15},"Machine Learning","machine-learning",{"name":1657,"slug":1658,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1662,"name":1662,"fn":1663,"description":1664,"org":1665,"tags":1666,"stars":1594,"repoUrl":1595,"updatedAt":1673},"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},[1667,1670],{"name":1668,"slug":1669,"type":15},"QA","qa",{"name":1671,"slug":1672,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":1675,"name":1675,"fn":1676,"description":1677,"org":1678,"tags":1679,"stars":1594,"repoUrl":1595,"updatedAt":1684},"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},[1680,1681],{"name":17,"slug":18,"type":15},{"name":1682,"slug":1683,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1686,"name":1686,"fn":1687,"description":1688,"org":1689,"tags":1690,"stars":1594,"repoUrl":1595,"updatedAt":1698},"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},[1691,1694,1695],{"name":1692,"slug":1693,"type":15},"Code Review","code-review",{"name":1617,"slug":1618,"type":15},{"name":1696,"slug":1697,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1700,"name":1700,"fn":1701,"description":1702,"org":1703,"tags":1704,"stars":1594,"repoUrl":1595,"updatedAt":1707},"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},[1705,1706],{"name":1668,"slug":1669,"type":15},{"name":1671,"slug":1672,"type":15},"2026-07-14T05:25:54.928983",{"slug":1709,"name":1709,"fn":1710,"description":1711,"org":1712,"tags":1713,"stars":1594,"repoUrl":1595,"updatedAt":1718},"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},[1714,1717],{"name":1715,"slug":1716,"type":15},"Automation","automation",{"name":1604,"slug":1605,"type":15},"2026-07-30T05:29:03.275638",496,{"items":1721,"total":1820},[1722,1740,1753,1768,1783,1794,1807],{"slug":1723,"name":1723,"fn":1724,"description":1725,"org":1726,"tags":1727,"stars":23,"repoUrl":24,"updatedAt":1739},"analysis-methods","write Python analysis code for FHIR data","Teaches the analyst agent how to write correct, robust Python analysis code for FHIR clinical data using pandas, matplotlib, and scipy.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1728,1731,1734,1737,1738],{"name":1729,"slug":1730,"type":15},"Data Analysis","data-analysis",{"name":1732,"slug":1733,"type":15},"FHIR","fhir",{"name":1735,"slug":1736,"type":15},"Healthcare","healthcare",{"name":9,"slug":8,"type":15},{"name":1592,"slug":1593,"type":15},"2026-07-14T05:35:59.037962",{"slug":1741,"name":1741,"fn":1742,"description":1743,"org":1744,"tags":1745,"stars":23,"repoUrl":24,"updatedAt":1752},"case-summary","summarize clinical patient cases from FHIR","Prepare a complete clinical case summary for a patient from FHIR endpoints. Use when asked to summarize a patient, compile a case, or prepare for tumor board.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1746,1747,1748,1749],{"name":1732,"slug":1733,"type":15},{"name":1735,"slug":1736,"type":15},{"name":9,"slug":8,"type":15},{"name":1750,"slug":1751,"type":15},"Summarization","summarization","2026-07-14T05:35:52.790528",{"slug":1754,"name":1754,"fn":1755,"description":1756,"org":1757,"tags":1758,"stars":23,"repoUrl":24,"updatedAt":1767},"clinical-delegation","delegate clinical tasks to specialist agents","How to delegate clinical tasks to specialist agents. Always use sub-agent runtime with explicit agentId — never ACP. Never call FHIR via web_fetch.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1759,1762,1763,1766],{"name":1760,"slug":1761,"type":15},"Agents","agents",{"name":1735,"slug":1736,"type":15},{"name":1764,"slug":1765,"type":15},"Multi-Agent","multi-agent",{"name":9,"slug":8,"type":15},"2026-07-14T05:35:55.294972",{"slug":1769,"name":1769,"fn":1770,"description":1771,"org":1772,"tags":1773,"stars":23,"repoUrl":24,"updatedAt":1782},"clinical-knowledge","provide clinical reference and regulatory context","Teaches agents clinical reference ranges, condition codes, quality measure definitions, drug classifications, and regulatory context so they can flag abnormal values and identify care gaps.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1774,1777,1778,1779],{"name":1775,"slug":1776,"type":15},"Clinical Trials","clinical-trials",{"name":1735,"slug":1736,"type":15},{"name":9,"slug":8,"type":15},{"name":1780,"slug":1781,"type":15},"Regulatory Compliance","regulatory-compliance","2026-07-14T05:35:56.550833",{"slug":1784,"name":1784,"fn":1785,"description":1786,"org":1787,"tags":1788,"stars":23,"repoUrl":24,"updatedAt":1793},"cohort-compare","analyze patient cohorts from FHIR endpoints","Analyze a cohort of patients from FHIR endpoints to find care gaps and patterns. Use when asked to compare patients, find quality gaps, or analyze a population.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1789,1790,1791,1792],{"name":1729,"slug":1730,"type":15},{"name":1732,"slug":1733,"type":15},{"name":1735,"slug":1736,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:35:51.543095",{"slug":1795,"name":1795,"fn":1796,"description":1797,"org":1798,"tags":1799,"stars":23,"repoUrl":24,"updatedAt":1806},"dgx-diagnose","diagnose NVIDIA DGX Station hardware issues","Diagnose common DGX Station GB300 issues — CUDA crashes, wrong-GPU targeting, vLLM\u002FSGLang container bugs, MIG state problems, NVLink\u002FFabric Manager errors, X\u002FVulkan failures, HuggingFace auth, and port conflicts. Use when the user reports a GPU error, inference server crash, MIG problem, or any unexplained DGX Station failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1800,1801,1802,1803],{"name":21,"slug":22,"type":15},{"name":1627,"slug":1628,"type":15},{"name":9,"slug":8,"type":15},{"name":1804,"slug":1805,"type":15},"Observability","observability","2026-07-14T05:31:04.085598",{"slug":1808,"name":1808,"fn":1809,"description":1810,"org":1811,"tags":1812,"stars":23,"repoUrl":24,"updatedAt":1819},"fhir-basics","query and parse FHIR R4 API resources","Teaches agents how FHIR R4 APIs work, what resources are available, how to query them with search parameters, and how to correctly parse all response formats including component Observations.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1813,1816,1817,1818],{"name":1814,"slug":1815,"type":15},"API Development","api-development",{"name":1732,"slug":1733,"type":15},{"name":1735,"slug":1736,"type":15},{"name":9,"slug":8,"type":15},"2026-07-14T05:35:57.797226",11]