[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-jetson-memory-audit":3,"mdc--cfc5do-key":34,"related-repo-nvidia-jetson-memory-audit":1896,"related-org-nvidia-jetson-memory-audit":2000},{"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},"jetson-memory-audit","audit Jetson DRAM and memory usage","Measure Jetson DRAM\u002FNvMap usage and verify before\u002Fafter memory reclamation with live audit data.",{"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},"Monitoring","monitoring",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Debugging","debugging",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:28:35.749212","Apache-2.0",281,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fjetson-memory-audit","---\nname: jetson-memory-audit\ndescription: Measure Jetson DRAM\u002FNvMap usage and verify before\u002Fafter memory reclamation with live audit data.\nversion: 0.0.1\nlicense: \"Apache-2.0\"\nmetadata:\n  author: \"Jetson Team\"\n  tags: [jetson, memory, audit]\n  languages: [bash]\n  data-classification: public\n---\n\n# Jetson Memory Audit\n\nRead-only memory-focused snapshot for a Jetson, plus the `drop_caches` verify-loop helper that confirms freed memory actually shows up as free instead of cached.\n\n## Purpose\n\nMeasure current Jetson memory consumers, capture before\u002Fafter baselines, and verify whether user-approved changes actually reclaimed DRAM. Use live device data rather than estimates from container size, model size, or generic process memory.\n\n## CRITICAL: Memory appears stuck after stopping vLLM \u002F sglang (JetPack below 7.2 \u002F L4T below r39.0)\n\n**This is the most common memory confusion on Jetson releases before JetPack 7.2 or before L4T r39.0.**\n\nAfter you stop a vLLM, sglang, or Ollama server (or any CUDA workload), the memory shown as free by `free -h` or `tegrastats` may not recover — even though the process is gone. `nvidia-smi` may also show misleadingly low free GPU memory.\n\n**Root cause:** The Thor RM (resource manager) holds freed sysmem pages in its own pool after a CUDA context exits. On Unified Memory Architecture (UMA) devices like Jetson, `cudaMemGetInfo` reads RM pool state and reports far less free memory than is actually available to a new process.\n\n**Workaround (for JetPack below 7.2 or L4T below r39.0):**\n```bash\nsudo sync && sudo sysctl -w vm.drop_caches=3\n```\nRun this on the **host**, not inside a container. The important operation is `sudo sysctl -w vm.drop_caches=3`; keep `sudo sync` immediately before it so dirty data is flushed before reclaimable page\u002Fdentry\u002Finode caches are dropped. After running it, `free -h` and `tegrastats` will reflect the true available memory.\n\nFor affected releases, recommend this command when a user says:\n- \"Memory didn't free after I stopped vLLM\u002Fsglang\"\n- \"Why does tegrastats still show high usage after my container exited?\"\n- \"OOM even though nothing is running\"\n- \"Memory was fine yesterday but now it's full\"\n\nOn JetPack below 7.2 or L4T below r39.0, `drop_caches` is the reliable workaround when memory appears stuck after a CUDA workload exits; on newer releases, use it only if the same symptom is observed and the user approves.\n\n## When to use\n\n- \"How much memory is in use on this Jetson? What's holding it?\"\n- \"I disabled the GUI \u002F stopped vLLM \u002F quit my container — did memory actually free?\"\n- \"Why does `free -h` still show low free memory after I stopped my workload?\"\n- As the **baseline** before applying `jetson-headless-mode` or other memory-related changes, and again **after** to compute the actual delta.\n\n## Prerequisites\n\n- Run on the Jetson host, or in a sandbox\u002Fcontainer with host-visible `\u002Fproc`, `\u002Fetc\u002Fnv_tegra_release`, `tegrastats`, and process data.\n- NvMap debugfs reads may require root. If unavailable, report that GPU memory attribution is limited rather than guessing.\n- `drop_caches.sh` requires root or passwordless `sudo -n`; run it only after the user explicitly authorizes cache dropping.\n\n## Available Scripts\n\n| Script | Purpose | Arguments |\n|--------|---------|-----------|\n| `scripts\u002Faudit.sh` | Emits a JSON snapshot from `jetson-diagnostic\u002Fscripts\u002Fsnapshot.sh` for memory audit workflows. | No arguments. |\n| `scripts\u002Fdrop_caches.sh` | Flushes reclaimable page\u002Fdentry\u002Finode caches and prints before\u002Fafter memory deltas. | `--mode 1\\|2\\|3`, `--quiet`. |\n\nIf your agent runtime supports `run_script`, use it to run `scripts\u002Faudit.sh` or `scripts\u002Fdrop_caches.sh` and summarize the returned output. Otherwise run the scripts with `bash` from the repository root.\n\n## Instructions\n\nFor \"how much memory is in use right now?\" questions, run `scripts\u002Faudit.sh` and report only values from the JSON snapshot.\n\n## Reporting guidance\n\nDo not only print or mention the path to a helper. Invoke the helper and then summarize the returned data.\n\n- For \"how much memory is in use\" prompts, run `scripts\u002Faudit.sh` and quote `mem_total_gb`, `memory_kb.available`, and the leading `procrank_top` process or `nvmap.top_clients` consumer.\n- For GUI\u002Fdesktop memory prompts, run `scripts\u002Faudit.sh` and report `default_systemd_target` plus any display manager in `candidate_services` (`gdm3`, `gdm`, `lightdm`, `sddm`, or `display-manager`). Do not disable anything; hand off to `jetson-headless-mode` for a plan.\n- For prompts that explicitly authorize cache dropping after a stopped workload, run `scripts\u002Fdrop_caches.sh` (equivalent to `sudo sync && sudo sysctl -w vm.drop_caches=3` by default) and report its before\u002Fafter free, available, and cached deltas. If root is unavailable, explain that it must be run on the host with sudo.\n\nIf your agent runtime does not execute helper scripts relative to this skill directory, resolve script paths with the AgentSkills `{baseDir}` placeholder:\n\n```bash\n{baseDir}\u002Fscripts\u002Faudit.sh\n{baseDir}\u002Fscripts\u002Fdrop_caches.sh\n```\n\nDo not call `jetson-memory-audit` as a tool name unless the runtime explicitly registers skills as callable tools; Agent Skills are normally instructions plus files, not direct tool functions.\n\nSandbox note for agents: seeing this skill file does not guarantee access to Jetson host memory data. If `\u002Fproc\u002Fdevice-tree\u002Fmodel`, `\u002Fetc\u002Fnv_tegra_release`, `tegrastats`, `\u002Fsys\u002Fkernel\u002Fdebug\u002Fnvmap`, or host process data are missing inside a NemoClaw\u002FOpenClaw sandbox, say the sandbox lacks Jetson host visibility and ask the user to run on the Jetson host or relaunch with a host-visible sandbox profile. Do not fabricate memory totals, available memory, PSS, NvMap, or reclamation deltas.\n\nFor \"how much memory did this change free?\" questions, use a before\u002Fafter delta. Do not estimate freed memory from container size, image size, RSS, or a single post-change snapshot.\n\n1. Before the change, run `scripts\u002Faudit.sh` and save the JSON baseline.\n2. Make the user-approved change (stop the container, switch mode, apply a tuning recommendation, etc.).\n3. On JetPack below 7.2 \u002F L4T below r39.0, or when the same stuck-memory symptom is observed on a newer release, flush reclaimable page cache on the **host** (not inside a container) so freed pages show up as free instead of cached:\n   ```bash\n   sudo sync && sudo sysctl -w vm.drop_caches=3\n   ```\n4. Re-run `scripts\u002Faudit.sh` and compare `memory_kb.available` before vs after — that delta is the real reclamation.\n\nIf the user already made the change and no baseline exists, say that the exact freed amount cannot be recovered from the current snapshot alone. Capture a new baseline now so the next change can be measured.\n\nUse live audit data as the source of truth. Memory totals, available memory, NvMap totals, PSS values, display-manager state, and savings deltas must come from `scripts\u002Faudit.sh`, `free -h`, or `tegrastats` on the actual device. If a number is not present in those outputs, do not guess it.\n\n## Output contract for `audit.sh`\n\n```json\n{\n  \"sku\": \"orin-nano\",\n  \"variant\": \"orin-nano-8gb\",\n  \"mem_total_gb\": 8,\n  \"l4t_version\": \"36.4.0\",\n  \"product_model\": \"nvidia jetson orin nano developer kit\",\n  \"memory_kb\": { \"total\": 8123456, \"available\": 4123456, \"free\": 1023456, \"cached\": 1234567, \"swap_total\": 0, \"swap_free\": 0 },\n  \"default_systemd_target\": \"graphical.target\",\n  \"candidate_services\": { \"gdm3\": { \"active\": \"active\", \"enabled\": \"enabled\" } },\n  \"tegrastats_sample\": \"RAM 4011\u002F8138MB (lfb 8x4MB) ...\",\n  \"nvmap\": { \"readable\": false, \"total_kb\": 0, \"top_clients\": [] },\n  \"procrank_top\": [ { \"pid\": 4321, \"pss_kb\": 4000000, \"cmd\": \"vllm\" } ]\n}\n```\n\n## Limitations\n\n- Exact freed-memory deltas require a before snapshot, the user-approved change, cache flush when appropriate, and an after snapshot.\n- NvMap attribution depends on host-visible debugfs access; if it is unavailable, report limited GPU memory attribution instead of guessing.\n- Sandbox\u002Fcontainer runs may not see host `\u002Fproc`, `tegrastats`, systemd, or NvMap data unless the runtime exposes them.\n\n## Error handling\n\n- If `scripts\u002Faudit.sh` cannot access host Jetson data, report the missing visibility and ask to rerun on the Jetson host or in a host-visible sandbox.\n- If `scripts\u002Fdrop_caches.sh` lacks root or passwordless `sudo -n`, report that cache dropping must be run on the host with sudo approval.\n- If no before snapshot exists, say the exact reclaimed amount cannot be recovered from the current state alone and capture a new baseline for the next change.\n\n## Safety\n\nRead-only. `drop_caches` is non-destructive (kernel only releases pages it could reclaim under pressure anyway; `sync` runs first to preserve dirty data).\n\n## Hand off to\n\n- `jetson-headless-mode` — biggest single user-space win on systems still booting `graphical.target`.\n- `jetson-inference-mem-tune` — when a model server is the top NvMap \u002F PSS consumer.\n- If runtime changes cannot hit the target, report that further reclamation is outside this skill's scope rather than suggesting unsafe boot-time edits.\n",{"data":35,"body":46},{"name":4,"description":6,"version":36,"license":26,"metadata":37},"0.0.1",{"author":38,"tags":39,"languages":43,"data-classification":45},"Jetson Team",[40,41,42],"jetson","memory","audit",[44],"bash","public",{"type":47,"children":48},"root",[49,57,72,79,84,90,99,128,146,154,214,256,261,286,298,304,356,362,416,422,519,552,558,570,576,581,719,732,757,769,801,806,898,903,927,939,1742,1748,1779,1785,1823,1829,1849,1855,1890],{"type":50,"tag":51,"props":52,"children":53},"element","h1",{"id":4},[54],{"type":55,"value":56},"text","Jetson Memory Audit",{"type":50,"tag":58,"props":59,"children":60},"p",{},[61,63,70],{"type":55,"value":62},"Read-only memory-focused snapshot for a Jetson, plus the ",{"type":50,"tag":64,"props":65,"children":67},"code",{"className":66},[],[68],{"type":55,"value":69},"drop_caches",{"type":55,"value":71}," verify-loop helper that confirms freed memory actually shows up as free instead of cached.",{"type":50,"tag":73,"props":74,"children":76},"h2",{"id":75},"purpose",[77],{"type":55,"value":78},"Purpose",{"type":50,"tag":58,"props":80,"children":81},{},[82],{"type":55,"value":83},"Measure current Jetson memory consumers, capture before\u002Fafter baselines, and verify whether user-approved changes actually reclaimed DRAM. Use live device data rather than estimates from container size, model size, or generic process memory.",{"type":50,"tag":73,"props":85,"children":87},{"id":86},"critical-memory-appears-stuck-after-stopping-vllm-sglang-jetpack-below-72-l4t-below-r390",[88],{"type":55,"value":89},"CRITICAL: Memory appears stuck after stopping vLLM \u002F sglang (JetPack below 7.2 \u002F L4T below r39.0)",{"type":50,"tag":58,"props":91,"children":92},{},[93],{"type":50,"tag":94,"props":95,"children":96},"strong",{},[97],{"type":55,"value":98},"This is the most common memory confusion on Jetson releases before JetPack 7.2 or before L4T r39.0.",{"type":50,"tag":58,"props":100,"children":101},{},[102,104,110,112,118,120,126],{"type":55,"value":103},"After you stop a vLLM, sglang, or Ollama server (or any CUDA workload), the memory shown as free by ",{"type":50,"tag":64,"props":105,"children":107},{"className":106},[],[108],{"type":55,"value":109},"free -h",{"type":55,"value":111}," or ",{"type":50,"tag":64,"props":113,"children":115},{"className":114},[],[116],{"type":55,"value":117},"tegrastats",{"type":55,"value":119}," may not recover — even though the process is gone. ",{"type":50,"tag":64,"props":121,"children":123},{"className":122},[],[124],{"type":55,"value":125},"nvidia-smi",{"type":55,"value":127}," may also show misleadingly low free GPU memory.",{"type":50,"tag":58,"props":129,"children":130},{},[131,136,138,144],{"type":50,"tag":94,"props":132,"children":133},{},[134],{"type":55,"value":135},"Root cause:",{"type":55,"value":137}," The Thor RM (resource manager) holds freed sysmem pages in its own pool after a CUDA context exits. On Unified Memory Architecture (UMA) devices like Jetson, ",{"type":50,"tag":64,"props":139,"children":141},{"className":140},[],[142],{"type":55,"value":143},"cudaMemGetInfo",{"type":55,"value":145}," reads RM pool state and reports far less free memory than is actually available to a new process.",{"type":50,"tag":58,"props":147,"children":148},{},[149],{"type":50,"tag":94,"props":150,"children":151},{},[152],{"type":55,"value":153},"Workaround (for JetPack below 7.2 or L4T below r39.0):",{"type":50,"tag":155,"props":156,"children":160},"pre",{"className":157,"code":158,"language":44,"meta":159,"style":159},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","sudo sync && sudo sysctl -w vm.drop_caches=3\n","",[161],{"type":50,"tag":64,"props":162,"children":163},{"__ignoreMap":159},[164],{"type":50,"tag":165,"props":166,"children":169},"span",{"class":167,"line":168},"line",1,[170,176,182,188,193,198,203,208],{"type":50,"tag":165,"props":171,"children":173},{"style":172},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[174],{"type":55,"value":175},"sudo",{"type":50,"tag":165,"props":177,"children":179},{"style":178},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[180],{"type":55,"value":181}," sync",{"type":50,"tag":165,"props":183,"children":185},{"style":184},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[186],{"type":55,"value":187}," &&",{"type":50,"tag":165,"props":189,"children":190},{"style":172},[191],{"type":55,"value":192}," sudo",{"type":50,"tag":165,"props":194,"children":195},{"style":178},[196],{"type":55,"value":197}," sysctl",{"type":50,"tag":165,"props":199,"children":200},{"style":178},[201],{"type":55,"value":202}," -w",{"type":50,"tag":165,"props":204,"children":205},{"style":178},[206],{"type":55,"value":207}," vm.drop_caches=",{"type":50,"tag":165,"props":209,"children":211},{"style":210},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[212],{"type":55,"value":213},"3\n",{"type":50,"tag":58,"props":215,"children":216},{},[217,219,224,226,232,234,240,242,247,249,254],{"type":55,"value":218},"Run this on the ",{"type":50,"tag":94,"props":220,"children":221},{},[222],{"type":55,"value":223},"host",{"type":55,"value":225},", not inside a container. The important operation is ",{"type":50,"tag":64,"props":227,"children":229},{"className":228},[],[230],{"type":55,"value":231},"sudo sysctl -w vm.drop_caches=3",{"type":55,"value":233},"; keep ",{"type":50,"tag":64,"props":235,"children":237},{"className":236},[],[238],{"type":55,"value":239},"sudo sync",{"type":55,"value":241}," immediately before it so dirty data is flushed before reclaimable page\u002Fdentry\u002Finode caches are dropped. After running it, ",{"type":50,"tag":64,"props":243,"children":245},{"className":244},[],[246],{"type":55,"value":109},{"type":55,"value":248}," and ",{"type":50,"tag":64,"props":250,"children":252},{"className":251},[],[253],{"type":55,"value":117},{"type":55,"value":255}," will reflect the true available memory.",{"type":50,"tag":58,"props":257,"children":258},{},[259],{"type":55,"value":260},"For affected releases, recommend this command when a user says:",{"type":50,"tag":262,"props":263,"children":264},"ul",{},[265,271,276,281],{"type":50,"tag":266,"props":267,"children":268},"li",{},[269],{"type":55,"value":270},"\"Memory didn't free after I stopped vLLM\u002Fsglang\"",{"type":50,"tag":266,"props":272,"children":273},{},[274],{"type":55,"value":275},"\"Why does tegrastats still show high usage after my container exited?\"",{"type":50,"tag":266,"props":277,"children":278},{},[279],{"type":55,"value":280},"\"OOM even though nothing is running\"",{"type":50,"tag":266,"props":282,"children":283},{},[284],{"type":55,"value":285},"\"Memory was fine yesterday but now it's full\"",{"type":50,"tag":58,"props":287,"children":288},{},[289,291,296],{"type":55,"value":290},"On JetPack below 7.2 or L4T below r39.0, ",{"type":50,"tag":64,"props":292,"children":294},{"className":293},[],[295],{"type":55,"value":69},{"type":55,"value":297}," is the reliable workaround when memory appears stuck after a CUDA workload exits; on newer releases, use it only if the same symptom is observed and the user approves.",{"type":50,"tag":73,"props":299,"children":301},{"id":300},"when-to-use",[302],{"type":55,"value":303},"When to use",{"type":50,"tag":262,"props":305,"children":306},{},[307,312,317,329],{"type":50,"tag":266,"props":308,"children":309},{},[310],{"type":55,"value":311},"\"How much memory is in use on this Jetson? What's holding it?\"",{"type":50,"tag":266,"props":313,"children":314},{},[315],{"type":55,"value":316},"\"I disabled the GUI \u002F stopped vLLM \u002F quit my container — did memory actually free?\"",{"type":50,"tag":266,"props":318,"children":319},{},[320,322,327],{"type":55,"value":321},"\"Why does ",{"type":50,"tag":64,"props":323,"children":325},{"className":324},[],[326],{"type":55,"value":109},{"type":55,"value":328}," still show low free memory after I stopped my workload?\"",{"type":50,"tag":266,"props":330,"children":331},{},[332,334,339,341,347,349,354],{"type":55,"value":333},"As the ",{"type":50,"tag":94,"props":335,"children":336},{},[337],{"type":55,"value":338},"baseline",{"type":55,"value":340}," before applying ",{"type":50,"tag":64,"props":342,"children":344},{"className":343},[],[345],{"type":55,"value":346},"jetson-headless-mode",{"type":55,"value":348}," or other memory-related changes, and again ",{"type":50,"tag":94,"props":350,"children":351},{},[352],{"type":55,"value":353},"after",{"type":55,"value":355}," to compute the actual delta.",{"type":50,"tag":73,"props":357,"children":359},{"id":358},"prerequisites",[360],{"type":55,"value":361},"Prerequisites",{"type":50,"tag":262,"props":363,"children":364},{},[365,392,397],{"type":50,"tag":266,"props":366,"children":367},{},[368,370,376,378,384,385,390],{"type":55,"value":369},"Run on the Jetson host, or in a sandbox\u002Fcontainer with host-visible ",{"type":50,"tag":64,"props":371,"children":373},{"className":372},[],[374],{"type":55,"value":375},"\u002Fproc",{"type":55,"value":377},", ",{"type":50,"tag":64,"props":379,"children":381},{"className":380},[],[382],{"type":55,"value":383},"\u002Fetc\u002Fnv_tegra_release",{"type":55,"value":377},{"type":50,"tag":64,"props":386,"children":388},{"className":387},[],[389],{"type":55,"value":117},{"type":55,"value":391},", and process data.",{"type":50,"tag":266,"props":393,"children":394},{},[395],{"type":55,"value":396},"NvMap debugfs reads may require root. If unavailable, report that GPU memory attribution is limited rather than guessing.",{"type":50,"tag":266,"props":398,"children":399},{},[400,406,408,414],{"type":50,"tag":64,"props":401,"children":403},{"className":402},[],[404],{"type":55,"value":405},"drop_caches.sh",{"type":55,"value":407}," requires root or passwordless ",{"type":50,"tag":64,"props":409,"children":411},{"className":410},[],[412],{"type":55,"value":413},"sudo -n",{"type":55,"value":415},"; run it only after the user explicitly authorizes cache dropping.",{"type":50,"tag":73,"props":417,"children":419},{"id":418},"available-scripts",[420],{"type":55,"value":421},"Available Scripts",{"type":50,"tag":423,"props":424,"children":425},"table",{},[426,449],{"type":50,"tag":427,"props":428,"children":429},"thead",{},[430],{"type":50,"tag":431,"props":432,"children":433},"tr",{},[434,440,444],{"type":50,"tag":435,"props":436,"children":437},"th",{},[438],{"type":55,"value":439},"Script",{"type":50,"tag":435,"props":441,"children":442},{},[443],{"type":55,"value":78},{"type":50,"tag":435,"props":445,"children":446},{},[447],{"type":55,"value":448},"Arguments",{"type":50,"tag":450,"props":451,"children":452},"tbody",{},[453,484],{"type":50,"tag":431,"props":454,"children":455},{},[456,466,479],{"type":50,"tag":457,"props":458,"children":459},"td",{},[460],{"type":50,"tag":64,"props":461,"children":463},{"className":462},[],[464],{"type":55,"value":465},"scripts\u002Faudit.sh",{"type":50,"tag":457,"props":467,"children":468},{},[469,471,477],{"type":55,"value":470},"Emits a JSON snapshot from ",{"type":50,"tag":64,"props":472,"children":474},{"className":473},[],[475],{"type":55,"value":476},"jetson-diagnostic\u002Fscripts\u002Fsnapshot.sh",{"type":55,"value":478}," for memory audit workflows.",{"type":50,"tag":457,"props":480,"children":481},{},[482],{"type":55,"value":483},"No arguments.",{"type":50,"tag":431,"props":485,"children":486},{},[487,496,501],{"type":50,"tag":457,"props":488,"children":489},{},[490],{"type":50,"tag":64,"props":491,"children":493},{"className":492},[],[494],{"type":55,"value":495},"scripts\u002Fdrop_caches.sh",{"type":50,"tag":457,"props":497,"children":498},{},[499],{"type":55,"value":500},"Flushes reclaimable page\u002Fdentry\u002Finode caches and prints before\u002Fafter memory deltas.",{"type":50,"tag":457,"props":502,"children":503},{},[504,510,511,517],{"type":50,"tag":64,"props":505,"children":507},{"className":506},[],[508],{"type":55,"value":509},"--mode 1|2|3",{"type":55,"value":377},{"type":50,"tag":64,"props":512,"children":514},{"className":513},[],[515],{"type":55,"value":516},"--quiet",{"type":55,"value":518},".",{"type":50,"tag":58,"props":520,"children":521},{},[522,524,530,532,537,538,543,545,550],{"type":55,"value":523},"If your agent runtime supports ",{"type":50,"tag":64,"props":525,"children":527},{"className":526},[],[528],{"type":55,"value":529},"run_script",{"type":55,"value":531},", use it to run ",{"type":50,"tag":64,"props":533,"children":535},{"className":534},[],[536],{"type":55,"value":465},{"type":55,"value":111},{"type":50,"tag":64,"props":539,"children":541},{"className":540},[],[542],{"type":55,"value":495},{"type":55,"value":544}," and summarize the returned output. Otherwise run the scripts with ",{"type":50,"tag":64,"props":546,"children":548},{"className":547},[],[549],{"type":55,"value":44},{"type":55,"value":551}," from the repository root.",{"type":50,"tag":73,"props":553,"children":555},{"id":554},"instructions",[556],{"type":55,"value":557},"Instructions",{"type":50,"tag":58,"props":559,"children":560},{},[561,563,568],{"type":55,"value":562},"For \"how much memory is in use right now?\" questions, run ",{"type":50,"tag":64,"props":564,"children":566},{"className":565},[],[567],{"type":55,"value":465},{"type":55,"value":569}," and report only values from the JSON snapshot.",{"type":50,"tag":73,"props":571,"children":573},{"id":572},"reporting-guidance",[574],{"type":55,"value":575},"Reporting guidance",{"type":50,"tag":58,"props":577,"children":578},{},[579],{"type":55,"value":580},"Do not only print or mention the path to a helper. Invoke the helper and then summarize the returned data.",{"type":50,"tag":262,"props":582,"children":583},{},[584,627,699],{"type":50,"tag":266,"props":585,"children":586},{},[587,589,594,596,602,603,609,611,617,619,625],{"type":55,"value":588},"For \"how much memory is in use\" prompts, run ",{"type":50,"tag":64,"props":590,"children":592},{"className":591},[],[593],{"type":55,"value":465},{"type":55,"value":595}," and quote ",{"type":50,"tag":64,"props":597,"children":599},{"className":598},[],[600],{"type":55,"value":601},"mem_total_gb",{"type":55,"value":377},{"type":50,"tag":64,"props":604,"children":606},{"className":605},[],[607],{"type":55,"value":608},"memory_kb.available",{"type":55,"value":610},", and the leading ",{"type":50,"tag":64,"props":612,"children":614},{"className":613},[],[615],{"type":55,"value":616},"procrank_top",{"type":55,"value":618}," process or ",{"type":50,"tag":64,"props":620,"children":622},{"className":621},[],[623],{"type":55,"value":624},"nvmap.top_clients",{"type":55,"value":626}," consumer.",{"type":50,"tag":266,"props":628,"children":629},{},[630,632,637,639,645,647,653,655,661,662,668,669,675,676,682,684,690,692,697],{"type":55,"value":631},"For GUI\u002Fdesktop memory prompts, run ",{"type":50,"tag":64,"props":633,"children":635},{"className":634},[],[636],{"type":55,"value":465},{"type":55,"value":638}," and report ",{"type":50,"tag":64,"props":640,"children":642},{"className":641},[],[643],{"type":55,"value":644},"default_systemd_target",{"type":55,"value":646}," plus any display manager in ",{"type":50,"tag":64,"props":648,"children":650},{"className":649},[],[651],{"type":55,"value":652},"candidate_services",{"type":55,"value":654}," (",{"type":50,"tag":64,"props":656,"children":658},{"className":657},[],[659],{"type":55,"value":660},"gdm3",{"type":55,"value":377},{"type":50,"tag":64,"props":663,"children":665},{"className":664},[],[666],{"type":55,"value":667},"gdm",{"type":55,"value":377},{"type":50,"tag":64,"props":670,"children":672},{"className":671},[],[673],{"type":55,"value":674},"lightdm",{"type":55,"value":377},{"type":50,"tag":64,"props":677,"children":679},{"className":678},[],[680],{"type":55,"value":681},"sddm",{"type":55,"value":683},", or ",{"type":50,"tag":64,"props":685,"children":687},{"className":686},[],[688],{"type":55,"value":689},"display-manager",{"type":55,"value":691},"). Do not disable anything; hand off to ",{"type":50,"tag":64,"props":693,"children":695},{"className":694},[],[696],{"type":55,"value":346},{"type":55,"value":698}," for a plan.",{"type":50,"tag":266,"props":700,"children":701},{},[702,704,709,711,717],{"type":55,"value":703},"For prompts that explicitly authorize cache dropping after a stopped workload, run ",{"type":50,"tag":64,"props":705,"children":707},{"className":706},[],[708],{"type":55,"value":495},{"type":55,"value":710}," (equivalent to ",{"type":50,"tag":64,"props":712,"children":714},{"className":713},[],[715],{"type":55,"value":716},"sudo sync && sudo sysctl -w vm.drop_caches=3",{"type":55,"value":718}," by default) and report its before\u002Fafter free, available, and cached deltas. If root is unavailable, explain that it must be run on the host with sudo.",{"type":50,"tag":58,"props":720,"children":721},{},[722,724,730],{"type":55,"value":723},"If your agent runtime does not execute helper scripts relative to this skill directory, resolve script paths with the AgentSkills ",{"type":50,"tag":64,"props":725,"children":727},{"className":726},[],[728],{"type":55,"value":729},"{baseDir}",{"type":55,"value":731}," placeholder:",{"type":50,"tag":155,"props":733,"children":735},{"className":157,"code":734,"language":44,"meta":159,"style":159},"{baseDir}\u002Fscripts\u002Faudit.sh\n{baseDir}\u002Fscripts\u002Fdrop_caches.sh\n",[736],{"type":50,"tag":64,"props":737,"children":738},{"__ignoreMap":159},[739,748],{"type":50,"tag":165,"props":740,"children":741},{"class":167,"line":168},[742],{"type":50,"tag":165,"props":743,"children":745},{"style":744},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[746],{"type":55,"value":747},"{baseDir}\u002Fscripts\u002Faudit.sh\n",{"type":50,"tag":165,"props":749,"children":751},{"class":167,"line":750},2,[752],{"type":50,"tag":165,"props":753,"children":754},{"style":744},[755],{"type":55,"value":756},"{baseDir}\u002Fscripts\u002Fdrop_caches.sh\n",{"type":50,"tag":58,"props":758,"children":759},{},[760,762,767],{"type":55,"value":761},"Do not call ",{"type":50,"tag":64,"props":763,"children":765},{"className":764},[],[766],{"type":55,"value":4},{"type":55,"value":768}," as a tool name unless the runtime explicitly registers skills as callable tools; Agent Skills are normally instructions plus files, not direct tool functions.",{"type":50,"tag":58,"props":770,"children":771},{},[772,774,780,781,786,787,792,793,799],{"type":55,"value":773},"Sandbox note for agents: seeing this skill file does not guarantee access to Jetson host memory data. If ",{"type":50,"tag":64,"props":775,"children":777},{"className":776},[],[778],{"type":55,"value":779},"\u002Fproc\u002Fdevice-tree\u002Fmodel",{"type":55,"value":377},{"type":50,"tag":64,"props":782,"children":784},{"className":783},[],[785],{"type":55,"value":383},{"type":55,"value":377},{"type":50,"tag":64,"props":788,"children":790},{"className":789},[],[791],{"type":55,"value":117},{"type":55,"value":377},{"type":50,"tag":64,"props":794,"children":796},{"className":795},[],[797],{"type":55,"value":798},"\u002Fsys\u002Fkernel\u002Fdebug\u002Fnvmap",{"type":55,"value":800},", or host process data are missing inside a NemoClaw\u002FOpenClaw sandbox, say the sandbox lacks Jetson host visibility and ask the user to run on the Jetson host or relaunch with a host-visible sandbox profile. Do not fabricate memory totals, available memory, PSS, NvMap, or reclamation deltas.",{"type":50,"tag":58,"props":802,"children":803},{},[804],{"type":55,"value":805},"For \"how much memory did this change free?\" questions, use a before\u002Fafter delta. Do not estimate freed memory from container size, image size, RSS, or a single post-change snapshot.",{"type":50,"tag":807,"props":808,"children":809},"ol",{},[810,822,827,879],{"type":50,"tag":266,"props":811,"children":812},{},[813,815,820],{"type":55,"value":814},"Before the change, run ",{"type":50,"tag":64,"props":816,"children":818},{"className":817},[],[819],{"type":55,"value":465},{"type":55,"value":821}," and save the JSON baseline.",{"type":50,"tag":266,"props":823,"children":824},{},[825],{"type":55,"value":826},"Make the user-approved change (stop the container, switch mode, apply a tuning recommendation, etc.).",{"type":50,"tag":266,"props":828,"children":829},{},[830,832,836,838],{"type":55,"value":831},"On JetPack below 7.2 \u002F L4T below r39.0, or when the same stuck-memory symptom is observed on a newer release, flush reclaimable page cache on the ",{"type":50,"tag":94,"props":833,"children":834},{},[835],{"type":55,"value":223},{"type":55,"value":837}," (not inside a container) so freed pages show up as free instead of cached:\n",{"type":50,"tag":155,"props":839,"children":840},{"className":157,"code":158,"language":44,"meta":159,"style":159},[841],{"type":50,"tag":64,"props":842,"children":843},{"__ignoreMap":159},[844],{"type":50,"tag":165,"props":845,"children":846},{"class":167,"line":168},[847,851,855,859,863,867,871,875],{"type":50,"tag":165,"props":848,"children":849},{"style":172},[850],{"type":55,"value":175},{"type":50,"tag":165,"props":852,"children":853},{"style":178},[854],{"type":55,"value":181},{"type":50,"tag":165,"props":856,"children":857},{"style":184},[858],{"type":55,"value":187},{"type":50,"tag":165,"props":860,"children":861},{"style":172},[862],{"type":55,"value":192},{"type":50,"tag":165,"props":864,"children":865},{"style":178},[866],{"type":55,"value":197},{"type":50,"tag":165,"props":868,"children":869},{"style":178},[870],{"type":55,"value":202},{"type":50,"tag":165,"props":872,"children":873},{"style":178},[874],{"type":55,"value":207},{"type":50,"tag":165,"props":876,"children":877},{"style":210},[878],{"type":55,"value":213},{"type":50,"tag":266,"props":880,"children":881},{},[882,884,889,891,896],{"type":55,"value":883},"Re-run ",{"type":50,"tag":64,"props":885,"children":887},{"className":886},[],[888],{"type":55,"value":465},{"type":55,"value":890}," and compare ",{"type":50,"tag":64,"props":892,"children":894},{"className":893},[],[895],{"type":55,"value":608},{"type":55,"value":897}," before vs after — that delta is the real reclamation.",{"type":50,"tag":58,"props":899,"children":900},{},[901],{"type":55,"value":902},"If the user already made the change and no baseline exists, say that the exact freed amount cannot be recovered from the current snapshot alone. Capture a new baseline now so the next change can be measured.",{"type":50,"tag":58,"props":904,"children":905},{},[906,908,913,914,919,920,925],{"type":55,"value":907},"Use live audit data as the source of truth. Memory totals, available memory, NvMap totals, PSS values, display-manager state, and savings deltas must come from ",{"type":50,"tag":64,"props":909,"children":911},{"className":910},[],[912],{"type":55,"value":465},{"type":55,"value":377},{"type":50,"tag":64,"props":915,"children":917},{"className":916},[],[918],{"type":55,"value":109},{"type":55,"value":683},{"type":50,"tag":64,"props":921,"children":923},{"className":922},[],[924],{"type":55,"value":117},{"type":55,"value":926}," on the actual device. If a number is not present in those outputs, do not guess it.",{"type":50,"tag":73,"props":928,"children":930},{"id":929},"output-contract-for-auditsh",[931,933],{"type":55,"value":932},"Output contract for ",{"type":50,"tag":64,"props":934,"children":936},{"className":935},[],[937],{"type":55,"value":938},"audit.sh",{"type":50,"tag":155,"props":940,"children":944},{"className":941,"code":942,"language":943,"meta":159,"style":159},"language-json shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","{\n  \"sku\": \"orin-nano\",\n  \"variant\": \"orin-nano-8gb\",\n  \"mem_total_gb\": 8,\n  \"l4t_version\": \"36.4.0\",\n  \"product_model\": \"nvidia jetson orin nano developer kit\",\n  \"memory_kb\": { \"total\": 8123456, \"available\": 4123456, \"free\": 1023456, \"cached\": 1234567, \"swap_total\": 0, \"swap_free\": 0 },\n  \"default_systemd_target\": \"graphical.target\",\n  \"candidate_services\": { \"gdm3\": { \"active\": \"active\", \"enabled\": \"enabled\" } },\n  \"tegrastats_sample\": \"RAM 4011\u002F8138MB (lfb 8x4MB) ...\",\n  \"nvmap\": { \"readable\": false, \"total_kb\": 0, \"top_clients\": [] },\n  \"procrank_top\": [ { \"pid\": 4321, \"pss_kb\": 4000000, \"cmd\": \"vllm\" } ]\n}\n","json",[945],{"type":50,"tag":64,"props":946,"children":947},{"__ignoreMap":159},[948,956,999,1037,1066,1104,1142,1325,1362,1477,1515,1613,1733],{"type":50,"tag":165,"props":949,"children":950},{"class":167,"line":168},[951],{"type":50,"tag":165,"props":952,"children":953},{"style":184},[954],{"type":55,"value":955},"{\n",{"type":50,"tag":165,"props":957,"children":958},{"class":167,"line":750},[959,964,970,975,980,985,990,994],{"type":50,"tag":165,"props":960,"children":961},{"style":184},[962],{"type":55,"value":963},"  \"",{"type":50,"tag":165,"props":965,"children":967},{"style":966},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[968],{"type":55,"value":969},"sku",{"type":50,"tag":165,"props":971,"children":972},{"style":184},[973],{"type":55,"value":974},"\"",{"type":50,"tag":165,"props":976,"children":977},{"style":184},[978],{"type":55,"value":979},":",{"type":50,"tag":165,"props":981,"children":982},{"style":184},[983],{"type":55,"value":984}," \"",{"type":50,"tag":165,"props":986,"children":987},{"style":178},[988],{"type":55,"value":989},"orin-nano",{"type":50,"tag":165,"props":991,"children":992},{"style":184},[993],{"type":55,"value":974},{"type":50,"tag":165,"props":995,"children":996},{"style":184},[997],{"type":55,"value":998},",\n",{"type":50,"tag":165,"props":1000,"children":1002},{"class":167,"line":1001},3,[1003,1007,1012,1016,1020,1024,1029,1033],{"type":50,"tag":165,"props":1004,"children":1005},{"style":184},[1006],{"type":55,"value":963},{"type":50,"tag":165,"props":1008,"children":1009},{"style":966},[1010],{"type":55,"value":1011},"variant",{"type":50,"tag":165,"props":1013,"children":1014},{"style":184},[1015],{"type":55,"value":974},{"type":50,"tag":165,"props":1017,"children":1018},{"style":184},[1019],{"type":55,"value":979},{"type":50,"tag":165,"props":1021,"children":1022},{"style":184},[1023],{"type":55,"value":984},{"type":50,"tag":165,"props":1025,"children":1026},{"style":178},[1027],{"type":55,"value":1028},"orin-nano-8gb",{"type":50,"tag":165,"props":1030,"children":1031},{"style":184},[1032],{"type":55,"value":974},{"type":50,"tag":165,"props":1034,"children":1035},{"style":184},[1036],{"type":55,"value":998},{"type":50,"tag":165,"props":1038,"children":1040},{"class":167,"line":1039},4,[1041,1045,1049,1053,1057,1062],{"type":50,"tag":165,"props":1042,"children":1043},{"style":184},[1044],{"type":55,"value":963},{"type":50,"tag":165,"props":1046,"children":1047},{"style":966},[1048],{"type":55,"value":601},{"type":50,"tag":165,"props":1050,"children":1051},{"style":184},[1052],{"type":55,"value":974},{"type":50,"tag":165,"props":1054,"children":1055},{"style":184},[1056],{"type":55,"value":979},{"type":50,"tag":165,"props":1058,"children":1059},{"style":210},[1060],{"type":55,"value":1061}," 8",{"type":50,"tag":165,"props":1063,"children":1064},{"style":184},[1065],{"type":55,"value":998},{"type":50,"tag":165,"props":1067,"children":1069},{"class":167,"line":1068},5,[1070,1074,1079,1083,1087,1091,1096,1100],{"type":50,"tag":165,"props":1071,"children":1072},{"style":184},[1073],{"type":55,"value":963},{"type":50,"tag":165,"props":1075,"children":1076},{"style":966},[1077],{"type":55,"value":1078},"l4t_version",{"type":50,"tag":165,"props":1080,"children":1081},{"style":184},[1082],{"type":55,"value":974},{"type":50,"tag":165,"props":1084,"children":1085},{"style":184},[1086],{"type":55,"value":979},{"type":50,"tag":165,"props":1088,"children":1089},{"style":184},[1090],{"type":55,"value":984},{"type":50,"tag":165,"props":1092,"children":1093},{"style":178},[1094],{"type":55,"value":1095},"36.4.0",{"type":50,"tag":165,"props":1097,"children":1098},{"style":184},[1099],{"type":55,"value":974},{"type":50,"tag":165,"props":1101,"children":1102},{"style":184},[1103],{"type":55,"value":998},{"type":50,"tag":165,"props":1105,"children":1107},{"class":167,"line":1106},6,[1108,1112,1117,1121,1125,1129,1134,1138],{"type":50,"tag":165,"props":1109,"children":1110},{"style":184},[1111],{"type":55,"value":963},{"type":50,"tag":165,"props":1113,"children":1114},{"style":966},[1115],{"type":55,"value":1116},"product_model",{"type":50,"tag":165,"props":1118,"children":1119},{"style":184},[1120],{"type":55,"value":974},{"type":50,"tag":165,"props":1122,"children":1123},{"style":184},[1124],{"type":55,"value":979},{"type":50,"tag":165,"props":1126,"children":1127},{"style":184},[1128],{"type":55,"value":984},{"type":50,"tag":165,"props":1130,"children":1131},{"style":178},[1132],{"type":55,"value":1133},"nvidia jetson orin nano developer kit",{"type":50,"tag":165,"props":1135,"children":1136},{"style":184},[1137],{"type":55,"value":974},{"type":50,"tag":165,"props":1139,"children":1140},{"style":184},[1141],{"type":55,"value":998},{"type":50,"tag":165,"props":1143,"children":1145},{"class":167,"line":1144},7,[1146,1150,1155,1159,1163,1168,1172,1177,1181,1185,1190,1195,1199,1204,1208,1212,1217,1221,1225,1230,1234,1238,1243,1247,1251,1256,1260,1264,1269,1273,1277,1282,1286,1290,1295,1299,1303,1308,1312,1316,1320],{"type":50,"tag":165,"props":1147,"children":1148},{"style":184},[1149],{"type":55,"value":963},{"type":50,"tag":165,"props":1151,"children":1152},{"style":966},[1153],{"type":55,"value":1154},"memory_kb",{"type":50,"tag":165,"props":1156,"children":1157},{"style":184},[1158],{"type":55,"value":974},{"type":50,"tag":165,"props":1160,"children":1161},{"style":184},[1162],{"type":55,"value":979},{"type":50,"tag":165,"props":1164,"children":1165},{"style":184},[1166],{"type":55,"value":1167}," {",{"type":50,"tag":165,"props":1169,"children":1170},{"style":184},[1171],{"type":55,"value":984},{"type":50,"tag":165,"props":1173,"children":1174},{"style":172},[1175],{"type":55,"value":1176},"total",{"type":50,"tag":165,"props":1178,"children":1179},{"style":184},[1180],{"type":55,"value":974},{"type":50,"tag":165,"props":1182,"children":1183},{"style":184},[1184],{"type":55,"value":979},{"type":50,"tag":165,"props":1186,"children":1187},{"style":210},[1188],{"type":55,"value":1189}," 8123456",{"type":50,"tag":165,"props":1191,"children":1192},{"style":184},[1193],{"type":55,"value":1194},",",{"type":50,"tag":165,"props":1196,"children":1197},{"style":184},[1198],{"type":55,"value":984},{"type":50,"tag":165,"props":1200,"children":1201},{"style":172},[1202],{"type":55,"value":1203},"available",{"type":50,"tag":165,"props":1205,"children":1206},{"style":184},[1207],{"type":55,"value":974},{"type":50,"tag":165,"props":1209,"children":1210},{"style":184},[1211],{"type":55,"value":979},{"type":50,"tag":165,"props":1213,"children":1214},{"style":210},[1215],{"type":55,"value":1216}," 4123456",{"type":50,"tag":165,"props":1218,"children":1219},{"style":184},[1220],{"type":55,"value":1194},{"type":50,"tag":165,"props":1222,"children":1223},{"style":184},[1224],{"type":55,"value":984},{"type":50,"tag":165,"props":1226,"children":1227},{"style":172},[1228],{"type":55,"value":1229},"free",{"type":50,"tag":165,"props":1231,"children":1232},{"style":184},[1233],{"type":55,"value":974},{"type":50,"tag":165,"props":1235,"children":1236},{"style":184},[1237],{"type":55,"value":979},{"type":50,"tag":165,"props":1239,"children":1240},{"style":210},[1241],{"type":55,"value":1242}," 1023456",{"type":50,"tag":165,"props":1244,"children":1245},{"style":184},[1246],{"type":55,"value":1194},{"type":50,"tag":165,"props":1248,"children":1249},{"style":184},[1250],{"type":55,"value":984},{"type":50,"tag":165,"props":1252,"children":1253},{"style":172},[1254],{"type":55,"value":1255},"cached",{"type":50,"tag":165,"props":1257,"children":1258},{"style":184},[1259],{"type":55,"value":974},{"type":50,"tag":165,"props":1261,"children":1262},{"style":184},[1263],{"type":55,"value":979},{"type":50,"tag":165,"props":1265,"children":1266},{"style":210},[1267],{"type":55,"value":1268}," 1234567",{"type":50,"tag":165,"props":1270,"children":1271},{"style":184},[1272],{"type":55,"value":1194},{"type":50,"tag":165,"props":1274,"children":1275},{"style":184},[1276],{"type":55,"value":984},{"type":50,"tag":165,"props":1278,"children":1279},{"style":172},[1280],{"type":55,"value":1281},"swap_total",{"type":50,"tag":165,"props":1283,"children":1284},{"style":184},[1285],{"type":55,"value":974},{"type":50,"tag":165,"props":1287,"children":1288},{"style":184},[1289],{"type":55,"value":979},{"type":50,"tag":165,"props":1291,"children":1292},{"style":210},[1293],{"type":55,"value":1294}," 0",{"type":50,"tag":165,"props":1296,"children":1297},{"style":184},[1298],{"type":55,"value":1194},{"type":50,"tag":165,"props":1300,"children":1301},{"style":184},[1302],{"type":55,"value":984},{"type":50,"tag":165,"props":1304,"children":1305},{"style":172},[1306],{"type":55,"value":1307},"swap_free",{"type":50,"tag":165,"props":1309,"children":1310},{"style":184},[1311],{"type":55,"value":974},{"type":50,"tag":165,"props":1313,"children":1314},{"style":184},[1315],{"type":55,"value":979},{"type":50,"tag":165,"props":1317,"children":1318},{"style":210},[1319],{"type":55,"value":1294},{"type":50,"tag":165,"props":1321,"children":1322},{"style":184},[1323],{"type":55,"value":1324}," },\n",{"type":50,"tag":165,"props":1326,"children":1328},{"class":167,"line":1327},8,[1329,1333,1337,1341,1345,1349,1354,1358],{"type":50,"tag":165,"props":1330,"children":1331},{"style":184},[1332],{"type":55,"value":963},{"type":50,"tag":165,"props":1334,"children":1335},{"style":966},[1336],{"type":55,"value":644},{"type":50,"tag":165,"props":1338,"children":1339},{"style":184},[1340],{"type":55,"value":974},{"type":50,"tag":165,"props":1342,"children":1343},{"style":184},[1344],{"type":55,"value":979},{"type":50,"tag":165,"props":1346,"children":1347},{"style":184},[1348],{"type":55,"value":984},{"type":50,"tag":165,"props":1350,"children":1351},{"style":178},[1352],{"type":55,"value":1353},"graphical.target",{"type":50,"tag":165,"props":1355,"children":1356},{"style":184},[1357],{"type":55,"value":974},{"type":50,"tag":165,"props":1359,"children":1360},{"style":184},[1361],{"type":55,"value":998},{"type":50,"tag":165,"props":1363,"children":1365},{"class":167,"line":1364},9,[1366,1370,1374,1378,1382,1386,1390,1394,1398,1402,1406,1410,1415,1419,1423,1427,1431,1435,1439,1443,1448,1452,1456,1460,1464,1468,1473],{"type":50,"tag":165,"props":1367,"children":1368},{"style":184},[1369],{"type":55,"value":963},{"type":50,"tag":165,"props":1371,"children":1372},{"style":966},[1373],{"type":55,"value":652},{"type":50,"tag":165,"props":1375,"children":1376},{"style":184},[1377],{"type":55,"value":974},{"type":50,"tag":165,"props":1379,"children":1380},{"style":184},[1381],{"type":55,"value":979},{"type":50,"tag":165,"props":1383,"children":1384},{"style":184},[1385],{"type":55,"value":1167},{"type":50,"tag":165,"props":1387,"children":1388},{"style":184},[1389],{"type":55,"value":984},{"type":50,"tag":165,"props":1391,"children":1392},{"style":172},[1393],{"type":55,"value":660},{"type":50,"tag":165,"props":1395,"children":1396},{"style":184},[1397],{"type":55,"value":974},{"type":50,"tag":165,"props":1399,"children":1400},{"style":184},[1401],{"type":55,"value":979},{"type":50,"tag":165,"props":1403,"children":1404},{"style":184},[1405],{"type":55,"value":1167},{"type":50,"tag":165,"props":1407,"children":1408},{"style":184},[1409],{"type":55,"value":984},{"type":50,"tag":165,"props":1411,"children":1412},{"style":210},[1413],{"type":55,"value":1414},"active",{"type":50,"tag":165,"props":1416,"children":1417},{"style":184},[1418],{"type":55,"value":974},{"type":50,"tag":165,"props":1420,"children":1421},{"style":184},[1422],{"type":55,"value":979},{"type":50,"tag":165,"props":1424,"children":1425},{"style":184},[1426],{"type":55,"value":984},{"type":50,"tag":165,"props":1428,"children":1429},{"style":178},[1430],{"type":55,"value":1414},{"type":50,"tag":165,"props":1432,"children":1433},{"style":184},[1434],{"type":55,"value":974},{"type":50,"tag":165,"props":1436,"children":1437},{"style":184},[1438],{"type":55,"value":1194},{"type":50,"tag":165,"props":1440,"children":1441},{"style":184},[1442],{"type":55,"value":984},{"type":50,"tag":165,"props":1444,"children":1445},{"style":210},[1446],{"type":55,"value":1447},"enabled",{"type":50,"tag":165,"props":1449,"children":1450},{"style":184},[1451],{"type":55,"value":974},{"type":50,"tag":165,"props":1453,"children":1454},{"style":184},[1455],{"type":55,"value":979},{"type":50,"tag":165,"props":1457,"children":1458},{"style":184},[1459],{"type":55,"value":984},{"type":50,"tag":165,"props":1461,"children":1462},{"style":178},[1463],{"type":55,"value":1447},{"type":50,"tag":165,"props":1465,"children":1466},{"style":184},[1467],{"type":55,"value":974},{"type":50,"tag":165,"props":1469,"children":1470},{"style":184},[1471],{"type":55,"value":1472}," }",{"type":50,"tag":165,"props":1474,"children":1475},{"style":184},[1476],{"type":55,"value":1324},{"type":50,"tag":165,"props":1478,"children":1480},{"class":167,"line":1479},10,[1481,1485,1490,1494,1498,1502,1507,1511],{"type":50,"tag":165,"props":1482,"children":1483},{"style":184},[1484],{"type":55,"value":963},{"type":50,"tag":165,"props":1486,"children":1487},{"style":966},[1488],{"type":55,"value":1489},"tegrastats_sample",{"type":50,"tag":165,"props":1491,"children":1492},{"style":184},[1493],{"type":55,"value":974},{"type":50,"tag":165,"props":1495,"children":1496},{"style":184},[1497],{"type":55,"value":979},{"type":50,"tag":165,"props":1499,"children":1500},{"style":184},[1501],{"type":55,"value":984},{"type":50,"tag":165,"props":1503,"children":1504},{"style":178},[1505],{"type":55,"value":1506},"RAM 4011\u002F8138MB (lfb 8x4MB) ...",{"type":50,"tag":165,"props":1508,"children":1509},{"style":184},[1510],{"type":55,"value":974},{"type":50,"tag":165,"props":1512,"children":1513},{"style":184},[1514],{"type":55,"value":998},{"type":50,"tag":165,"props":1516,"children":1518},{"class":167,"line":1517},11,[1519,1523,1528,1532,1536,1540,1544,1549,1553,1557,1562,1566,1571,1575,1579,1583,1587,1591,1596,1600,1604,1609],{"type":50,"tag":165,"props":1520,"children":1521},{"style":184},[1522],{"type":55,"value":963},{"type":50,"tag":165,"props":1524,"children":1525},{"style":966},[1526],{"type":55,"value":1527},"nvmap",{"type":50,"tag":165,"props":1529,"children":1530},{"style":184},[1531],{"type":55,"value":974},{"type":50,"tag":165,"props":1533,"children":1534},{"style":184},[1535],{"type":55,"value":979},{"type":50,"tag":165,"props":1537,"children":1538},{"style":184},[1539],{"type":55,"value":1167},{"type":50,"tag":165,"props":1541,"children":1542},{"style":184},[1543],{"type":55,"value":984},{"type":50,"tag":165,"props":1545,"children":1546},{"style":172},[1547],{"type":55,"value":1548},"readable",{"type":50,"tag":165,"props":1550,"children":1551},{"style":184},[1552],{"type":55,"value":974},{"type":50,"tag":165,"props":1554,"children":1555},{"style":184},[1556],{"type":55,"value":979},{"type":50,"tag":165,"props":1558,"children":1559},{"style":184},[1560],{"type":55,"value":1561}," false,",{"type":50,"tag":165,"props":1563,"children":1564},{"style":184},[1565],{"type":55,"value":984},{"type":50,"tag":165,"props":1567,"children":1568},{"style":172},[1569],{"type":55,"value":1570},"total_kb",{"type":50,"tag":165,"props":1572,"children":1573},{"style":184},[1574],{"type":55,"value":974},{"type":50,"tag":165,"props":1576,"children":1577},{"style":184},[1578],{"type":55,"value":979},{"type":50,"tag":165,"props":1580,"children":1581},{"style":210},[1582],{"type":55,"value":1294},{"type":50,"tag":165,"props":1584,"children":1585},{"style":184},[1586],{"type":55,"value":1194},{"type":50,"tag":165,"props":1588,"children":1589},{"style":184},[1590],{"type":55,"value":984},{"type":50,"tag":165,"props":1592,"children":1593},{"style":172},[1594],{"type":55,"value":1595},"top_clients",{"type":50,"tag":165,"props":1597,"children":1598},{"style":184},[1599],{"type":55,"value":974},{"type":50,"tag":165,"props":1601,"children":1602},{"style":184},[1603],{"type":55,"value":979},{"type":50,"tag":165,"props":1605,"children":1606},{"style":184},[1607],{"type":55,"value":1608}," []",{"type":50,"tag":165,"props":1610,"children":1611},{"style":184},[1612],{"type":55,"value":1324},{"type":50,"tag":165,"props":1614,"children":1616},{"class":167,"line":1615},12,[1617,1621,1625,1629,1633,1638,1642,1646,1651,1655,1659,1664,1668,1672,1677,1681,1685,1690,1694,1698,1703,1707,1711,1715,1720,1724,1728],{"type":50,"tag":165,"props":1618,"children":1619},{"style":184},[1620],{"type":55,"value":963},{"type":50,"tag":165,"props":1622,"children":1623},{"style":966},[1624],{"type":55,"value":616},{"type":50,"tag":165,"props":1626,"children":1627},{"style":184},[1628],{"type":55,"value":974},{"type":50,"tag":165,"props":1630,"children":1631},{"style":184},[1632],{"type":55,"value":979},{"type":50,"tag":165,"props":1634,"children":1635},{"style":184},[1636],{"type":55,"value":1637}," [",{"type":50,"tag":165,"props":1639,"children":1640},{"style":184},[1641],{"type":55,"value":1167},{"type":50,"tag":165,"props":1643,"children":1644},{"style":184},[1645],{"type":55,"value":984},{"type":50,"tag":165,"props":1647,"children":1648},{"style":172},[1649],{"type":55,"value":1650},"pid",{"type":50,"tag":165,"props":1652,"children":1653},{"style":184},[1654],{"type":55,"value":974},{"type":50,"tag":165,"props":1656,"children":1657},{"style":184},[1658],{"type":55,"value":979},{"type":50,"tag":165,"props":1660,"children":1661},{"style":210},[1662],{"type":55,"value":1663}," 4321",{"type":50,"tag":165,"props":1665,"children":1666},{"style":184},[1667],{"type":55,"value":1194},{"type":50,"tag":165,"props":1669,"children":1670},{"style":184},[1671],{"type":55,"value":984},{"type":50,"tag":165,"props":1673,"children":1674},{"style":172},[1675],{"type":55,"value":1676},"pss_kb",{"type":50,"tag":165,"props":1678,"children":1679},{"style":184},[1680],{"type":55,"value":974},{"type":50,"tag":165,"props":1682,"children":1683},{"style":184},[1684],{"type":55,"value":979},{"type":50,"tag":165,"props":1686,"children":1687},{"style":210},[1688],{"type":55,"value":1689}," 4000000",{"type":50,"tag":165,"props":1691,"children":1692},{"style":184},[1693],{"type":55,"value":1194},{"type":50,"tag":165,"props":1695,"children":1696},{"style":184},[1697],{"type":55,"value":984},{"type":50,"tag":165,"props":1699,"children":1700},{"style":172},[1701],{"type":55,"value":1702},"cmd",{"type":50,"tag":165,"props":1704,"children":1705},{"style":184},[1706],{"type":55,"value":974},{"type":50,"tag":165,"props":1708,"children":1709},{"style":184},[1710],{"type":55,"value":979},{"type":50,"tag":165,"props":1712,"children":1713},{"style":184},[1714],{"type":55,"value":984},{"type":50,"tag":165,"props":1716,"children":1717},{"style":178},[1718],{"type":55,"value":1719},"vllm",{"type":50,"tag":165,"props":1721,"children":1722},{"style":184},[1723],{"type":55,"value":974},{"type":50,"tag":165,"props":1725,"children":1726},{"style":184},[1727],{"type":55,"value":1472},{"type":50,"tag":165,"props":1729,"children":1730},{"style":184},[1731],{"type":55,"value":1732}," ]\n",{"type":50,"tag":165,"props":1734,"children":1736},{"class":167,"line":1735},13,[1737],{"type":50,"tag":165,"props":1738,"children":1739},{"style":184},[1740],{"type":55,"value":1741},"}\n",{"type":50,"tag":73,"props":1743,"children":1745},{"id":1744},"limitations",[1746],{"type":55,"value":1747},"Limitations",{"type":50,"tag":262,"props":1749,"children":1750},{},[1751,1756,1761],{"type":50,"tag":266,"props":1752,"children":1753},{},[1754],{"type":55,"value":1755},"Exact freed-memory deltas require a before snapshot, the user-approved change, cache flush when appropriate, and an after snapshot.",{"type":50,"tag":266,"props":1757,"children":1758},{},[1759],{"type":55,"value":1760},"NvMap attribution depends on host-visible debugfs access; if it is unavailable, report limited GPU memory attribution instead of guessing.",{"type":50,"tag":266,"props":1762,"children":1763},{},[1764,1766,1771,1772,1777],{"type":55,"value":1765},"Sandbox\u002Fcontainer runs may not see host ",{"type":50,"tag":64,"props":1767,"children":1769},{"className":1768},[],[1770],{"type":55,"value":375},{"type":55,"value":377},{"type":50,"tag":64,"props":1773,"children":1775},{"className":1774},[],[1776],{"type":55,"value":117},{"type":55,"value":1778},", systemd, or NvMap data unless the runtime exposes them.",{"type":50,"tag":73,"props":1780,"children":1782},{"id":1781},"error-handling",[1783],{"type":55,"value":1784},"Error handling",{"type":50,"tag":262,"props":1786,"children":1787},{},[1788,1800,1818],{"type":50,"tag":266,"props":1789,"children":1790},{},[1791,1793,1798],{"type":55,"value":1792},"If ",{"type":50,"tag":64,"props":1794,"children":1796},{"className":1795},[],[1797],{"type":55,"value":465},{"type":55,"value":1799}," cannot access host Jetson data, report the missing visibility and ask to rerun on the Jetson host or in a host-visible sandbox.",{"type":50,"tag":266,"props":1801,"children":1802},{},[1803,1804,1809,1811,1816],{"type":55,"value":1792},{"type":50,"tag":64,"props":1805,"children":1807},{"className":1806},[],[1808],{"type":55,"value":495},{"type":55,"value":1810}," lacks root or passwordless ",{"type":50,"tag":64,"props":1812,"children":1814},{"className":1813},[],[1815],{"type":55,"value":413},{"type":55,"value":1817},", report that cache dropping must be run on the host with sudo approval.",{"type":50,"tag":266,"props":1819,"children":1820},{},[1821],{"type":55,"value":1822},"If no before snapshot exists, say the exact reclaimed amount cannot be recovered from the current state alone and capture a new baseline for the next change.",{"type":50,"tag":73,"props":1824,"children":1826},{"id":1825},"safety",[1827],{"type":55,"value":1828},"Safety",{"type":50,"tag":58,"props":1830,"children":1831},{},[1832,1834,1839,1841,1847],{"type":55,"value":1833},"Read-only. ",{"type":50,"tag":64,"props":1835,"children":1837},{"className":1836},[],[1838],{"type":55,"value":69},{"type":55,"value":1840}," is non-destructive (kernel only releases pages it could reclaim under pressure anyway; ",{"type":50,"tag":64,"props":1842,"children":1844},{"className":1843},[],[1845],{"type":55,"value":1846},"sync",{"type":55,"value":1848}," runs first to preserve dirty data).",{"type":50,"tag":73,"props":1850,"children":1852},{"id":1851},"hand-off-to",[1853],{"type":55,"value":1854},"Hand off to",{"type":50,"tag":262,"props":1856,"children":1857},{},[1858,1874,1885],{"type":50,"tag":266,"props":1859,"children":1860},{},[1861,1866,1868,1873],{"type":50,"tag":64,"props":1862,"children":1864},{"className":1863},[],[1865],{"type":55,"value":346},{"type":55,"value":1867}," — biggest single user-space win on systems still booting ",{"type":50,"tag":64,"props":1869,"children":1871},{"className":1870},[],[1872],{"type":55,"value":1353},{"type":55,"value":518},{"type":50,"tag":266,"props":1875,"children":1876},{},[1877,1883],{"type":50,"tag":64,"props":1878,"children":1880},{"className":1879},[],[1881],{"type":55,"value":1882},"jetson-inference-mem-tune",{"type":55,"value":1884}," — when a model server is the top NvMap \u002F PSS consumer.",{"type":50,"tag":266,"props":1886,"children":1887},{},[1888],{"type":55,"value":1889},"If runtime changes cannot hit the target, report that further reclamation is outside this skill's scope rather than suggesting unsafe boot-time edits.",{"type":50,"tag":1891,"props":1892,"children":1893},"style",{},[1894],{"type":55,"value":1895},"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":1897,"total":1999},[1898,1913,1927,1941,1953,1970,1985],{"slug":1899,"name":1899,"fn":1900,"description":1901,"org":1902,"tags":1903,"stars":23,"repoUrl":24,"updatedAt":1912},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1904,1907,1910,1911],{"name":1905,"slug":1906,"type":15},"Data Analysis","data-analysis",{"name":1908,"slug":1909,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},"2026-07-14T05:28:43.176466",{"slug":1914,"name":1914,"fn":1915,"description":1916,"org":1917,"tags":1918,"stars":23,"repoUrl":24,"updatedAt":1926},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1919,1922,1925],{"name":1920,"slug":1921,"type":15},"Deployment","deployment",{"name":1923,"slug":1924,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1928,"name":1928,"fn":1929,"description":1930,"org":1931,"tags":1932,"stars":23,"repoUrl":24,"updatedAt":1940},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1933,1936,1937],{"name":1934,"slug":1935,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1938,"slug":1939,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1942,"name":1942,"fn":1943,"description":1944,"org":1945,"tags":1946,"stars":23,"repoUrl":24,"updatedAt":1952},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1947,1948,1949],{"name":1905,"slug":1906,"type":15},{"name":9,"slug":8,"type":15},{"name":1950,"slug":1951,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1954,"name":1954,"fn":1955,"description":1956,"org":1957,"tags":1958,"stars":23,"repoUrl":24,"updatedAt":1969},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1959,1962,1965,1966],{"name":1960,"slug":1961,"type":15},"Automation","automation",{"name":1963,"slug":1964,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1967,"slug":1968,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1971,"name":1971,"fn":1972,"description":1973,"org":1974,"tags":1975,"stars":23,"repoUrl":24,"updatedAt":1984},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1976,1977,1980,1981],{"name":1920,"slug":1921,"type":15},{"name":1978,"slug":1979,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1982,"slug":1983,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1986,"name":1986,"fn":1987,"description":1988,"org":1989,"tags":1990,"stars":23,"repoUrl":24,"updatedAt":1998},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1991,1992,1995],{"name":9,"slug":8,"type":15},{"name":1993,"slug":1994,"type":15},"Quantum Computing","quantum-computing",{"name":1996,"slug":1997,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2001,"total":2150},[2002,2020,2036,2047,2059,2071,2084,2098,2109,2118,2132,2141],{"slug":2003,"name":2003,"fn":2004,"description":2005,"org":2006,"tags":2007,"stars":2017,"repoUrl":2018,"updatedAt":2019},"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},[2008,2011,2014],{"name":2009,"slug":2010,"type":15},"Documentation","documentation",{"name":2012,"slug":2013,"type":15},"MCP","mcp",{"name":2015,"slug":2016,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2021,"name":2021,"fn":2022,"description":2023,"org":2024,"tags":2025,"stars":2033,"repoUrl":2034,"updatedAt":2035},"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},[2026,2029,2030],{"name":2027,"slug":2028,"type":15},"Containers","containers",{"name":1920,"slug":1921,"type":15},{"name":2031,"slug":2032,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2037,"name":2037,"fn":2038,"description":2039,"org":2040,"tags":2041,"stars":2033,"repoUrl":2034,"updatedAt":2046},"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},[2042,2045],{"name":2043,"slug":2044,"type":15},"CI\u002FCD","ci-cd",{"name":1920,"slug":1921,"type":15},"2026-07-14T05:25:59.97109",{"slug":2048,"name":2048,"fn":2049,"description":2050,"org":2051,"tags":2052,"stars":2033,"repoUrl":2034,"updatedAt":2058},"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},[2053,2054,2055],{"name":2043,"slug":2044,"type":15},{"name":1920,"slug":1921,"type":15},{"name":2056,"slug":2057,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2060,"name":2060,"fn":2061,"description":2062,"org":2063,"tags":2064,"stars":2033,"repoUrl":2034,"updatedAt":2070},"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},[2065,2066,2067],{"name":21,"slug":22,"type":15},{"name":2056,"slug":2057,"type":15},{"name":2068,"slug":2069,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2072,"name":2072,"fn":2073,"description":2074,"org":2075,"tags":2076,"stars":2033,"repoUrl":2034,"updatedAt":2083},"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},[2077,2080],{"name":2078,"slug":2079,"type":15},"Best Practices","best-practices",{"name":2081,"slug":2082,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2085,"name":2085,"fn":2086,"description":2087,"org":2088,"tags":2089,"stars":2033,"repoUrl":2034,"updatedAt":2097},"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},[2090,2093,2096],{"name":2091,"slug":2092,"type":15},"Machine Learning","machine-learning",{"name":2094,"slug":2095,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2099,"name":2099,"fn":2100,"description":2101,"org":2102,"tags":2103,"stars":2033,"repoUrl":2034,"updatedAt":2108},"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},[2104,2107],{"name":2105,"slug":2106,"type":15},"QA","qa",{"name":1950,"slug":1951,"type":15},"2026-07-14T05:25:53.673039",{"slug":2110,"name":2110,"fn":2111,"description":2112,"org":2113,"tags":2114,"stars":2033,"repoUrl":2034,"updatedAt":2117},"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},[2115,2116],{"name":1920,"slug":1921,"type":15},{"name":1923,"slug":1924,"type":15},"2026-07-14T05:25:49.362534",{"slug":2119,"name":2119,"fn":2120,"description":2121,"org":2122,"tags":2123,"stars":2033,"repoUrl":2034,"updatedAt":2131},"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},[2124,2127,2128],{"name":2125,"slug":2126,"type":15},"Code Review","code-review",{"name":2056,"slug":2057,"type":15},{"name":2129,"slug":2130,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2133,"name":2133,"fn":2134,"description":2135,"org":2136,"tags":2137,"stars":2033,"repoUrl":2034,"updatedAt":2140},"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},[2138,2139],{"name":2105,"slug":2106,"type":15},{"name":1950,"slug":1951,"type":15},"2026-07-14T05:25:54.928983",{"slug":2142,"name":2142,"fn":2143,"description":2144,"org":2145,"tags":2146,"stars":2033,"repoUrl":2034,"updatedAt":2149},"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},[2147,2148],{"name":1960,"slug":1961,"type":15},{"name":2043,"slug":2044,"type":15},"2026-07-30T05:29:03.275638",496]