[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-holoscan-setup":3,"mdc--nr117y-key":31,"related-repo-nvidia-holoscan-setup":2055,"related-org-nvidia-holoscan-setup":2158},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":26,"sourceUrl":29,"mdContent":30},"holoscan-setup","install Holoscan SDK on host systems","Guides Holoscan SDK installation: inspects the host, assesses platform compatibility, recommends an install method, and delegates to the matching install skill.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,17],{"name":13,"slug":14,"type":15},"Deployment","deployment","tag",{"name":9,"slug":8,"type":15},{"name":18,"slug":19,"type":15},"Engineering","engineering",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:28:00.614445","Apache-2.0",281,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fholoscan-setup","---\nname: holoscan-setup\nversion: \"1.0.0\"\ndescription: \"Guides Holoscan SDK installation: inspects the host, assesses platform compatibility, recommends an install method, and delegates to the matching install skill.\"\nlicense: Apache-2.0\nmetadata:\n  author: \"Holoscan Team \u003Choloscan-team@nvidia.com>\"\n  github-url: \"https:\u002F\u002Fgithub.com\u002Fnvidia-holoscan\u002Fholoscan-sdk\"\n  tags:\n    - holoscan\n    - installation\n    - nvidia\n    - sdk\n    - setup\n---\n\n# Holoscan SDK Setup\n\n## Purpose\n\nDetermines the correct Holoscan SDK installation method for the current host by inspecting hardware, OS, CUDA driver, and existing tooling, then delegates to a method-specific install skill. Covers NGC container, Debian\u002Fapt, pip wheel, Conda, and source builds across Ubuntu, RHEL, IGX Orin, Jetson, and DGX Spark \u002F Grace-Hopper platforms.\n\n## Prerequisites\n\n- Linux host (Ubuntu 22.04\u002F24.04, RHEL 9.x, IGX Orin, Jetson, or DGX Spark \u002F Grace-Hopper)\n- NVIDIA GPU with a working driver (`nvidia-smi` returns a CUDA Version)\n- Network access to `docs.nvidia.com` and NGC\n- One of: Docker + NVIDIA Container Toolkit, `apt`, Python 3.10–3.13 with `pip`, Conda, or a build toolchain — depending on chosen method\n\n## Available Scripts\n\n| Script | Purpose | Arguments |\n|--------|---------|-----------|\n| `scripts\u002Fcheck_conda.sh` | Detects Conda installs even when not on PATH (searches `~\u002Fminiconda3`, `~\u002Fminiforge3`, `~\u002Fanaconda3`, `~\u002Fmambaforge`, `\u002Fopt\u002Fconda`, and shell rc files); reports envs and which have `holoscan` importable. | none |\n| `scripts\u002Fcheck_ngc_image.sh` | Checks whether the NGC Holoscan container image for a given CUDA tag suffix is pulled or available. | `\u003Ccuda-tag-suffix>` — one of `cuda13`, `cuda12-dgpu`, `cuda12-igpu` |\n\nInvoke scripts with `run_script(\"scripts\u002Fcheck_conda.sh\")` and `run_script(\"scripts\u002Fcheck_ngc_image.sh\", \"cuda13\")`. Trust the script output over bare commands such as `which conda` or `docker images`.\n\n## Instructions\n\nBe conversational and step-by-step — do not front-load all the information. Complete each step and report back before moving on.\n\n### Workflow rules (must follow)\n\n1. End Step 5 with a **bolded one-line recommendation** that names the method (e.g. `**Recommendation:** NGC Container — bundles all deps, fastest path to a working install.`).\n2. For a first-time user on a supported x86_64 host with Docker available, that recommendation **must** be **NGC Container**.\n3. After the recommendation, **stop and ask** which method to use. Do not paste `docker pull`, `docker run`, `apt install`, `pip install`, or other install commands in that turn — those belong to the delegated install skill in Step 6.\n4. If the container path is in play, verify Docker + GPU passthrough **yourself** in Step 4 (run the command shown there). Do not ask the user to run `nvidia-smi` or `docker --version` for you.\n\n### Step 1: Read the Docs First\n\nFetch `https:\u002F\u002Fdocs.nvidia.com\u002Fholoscan\u002Fsdk-user-guide\u002F` then `sdk_installation.html` to get the current release's supported platforms, package names, and install requirements. Do not rely on hardcoded assumptions.\n\n### Step 2: Inspect the Machine\n\nRun in parallel:\n\n```bash\nuname -a && (lsb_release -a 2>\u002Fdev\u002Fnull || cat \u002Fetc\u002Fos-release)\nuname -m\nnvidia-smi 2>&1 | head -10\nnproc && free -h | head -2\n```\n\n**Key:** Read the \"CUDA Version\" field from `nvidia-smi` (top-right of the table header) — this is the *maximum* CUDA version the driver supports, and drives `cuda12` vs `cuda13` package selection.\n\n### Step 3: Assess Compatibility\n\n| Platform | Methods Available |\n|----------|-------------------|\n| Ubuntu 22.04\u002F24.04, x86_64 | Container, Debian\u002Fapt, pip wheel, Conda, Source |\n| RHEL 9.x, x86_64 | Container only |\n| IGX Orin (ARM64) | Container, Debian\u002Fapt, Source |\n| Jetson AGX Orin \u002F Orin Nano | Container, Debian\u002Fapt (iGPU) |\n| Jetson AGX Thor | Container, Debian\u002Fapt |\n| DGX Spark \u002F Grace-Hopper | Container (check docs for OS requirements) |\n| Other Linux, x86_64 | Container may work; pip wheel if glibc ≥ 2.35 |\n\n### Step 4: Check Tools and Present Options\n\nRun in parallel:\n\n```bash\ndocker --version 2>&1 | head -1; python3 --version 2>&1; pip3 --version 2>&1\ndpkg -l | grep holoscan || true\npip3 show holoscan 2>\u002Fdev\u002Fnull | grep -E \"^(Name|Version)\" || true\n~\u002Fholoscan\u002Fvenv\u002Fbin\u002Fpip show holoscan 2>\u002Fdev\u002Fnull | grep -E \"^(Name|Version)\" | sed 's\u002F^\u002Fvenv: \u002F' || true\n```\n\nThen verify GPU passthrough yourself — do **not** ask the user to run this:\n\n```bash\ndocker run --rm --gpus all ubuntu:22.04 nvidia-smi 2>&1 | tail -5 || true\n```\n\nInterpret the result for the Status column in Step 5:\n- `docker` missing → container row Status `✗ — Docker not installed`.\n- Docker present but `could not select device driver \"nvidia\"` → `✗ — NVIDIA Container Toolkit missing`.\n- `nvidia-smi` output appears → `✓`.\n\nThen invoke the detection scripts via `run_script`:\n\n- `run_script(\"scripts\u002Fcheck_conda.sh\")` — see Available Scripts above for why this is preferred over `conda --version`.\n- `run_script(\"scripts\u002Fcheck_ngc_image.sh\", \"\u003Ccuda-tag-suffix>\")` — replace `\u003Ccuda-tag-suffix>` with the tag determined from Step 2 (e.g. `cuda13`, `cuda12-dgpu`, `cuda12-igpu`).\n\nIf Holoscan is already installed, note the version and ask whether to upgrade or verify the existing install.\n\n**CUDA variant rule** (canonical reference — apply this in all steps below):\n\n| nvidia-smi CUDA Version | Native packages | Container tag |\n|------------------------|-----------------|---------------|\n| 13.x+ | `holoscan-cu13` \u002F `holoscan-cuda-13` | `cuda13` |\n| 12.x, Blackwell GPU | `holoscan-cu12` \u002F `holoscan-cuda-12` | `cuda13` (Forward Compat) or `cuda12-dgpu` |\n| 12.x, Ampere\u002FAda dGPU | `holoscan-cu12` \u002F `holoscan-cuda-12` | `cuda12-dgpu` |\n| ARM64 iGPU (Jetson, IGX) | `holoscan` | `cuda12-igpu` |\n\nNative installs treat the driver CUDA version as a hard ceiling. Containers support Forward Compatibility (banner saying \"CUDA Forward Compatibility mode ENABLED\" is expected, not an error).\n\n### Step 5: Present Options and Recommend\n\nAlways present **all methods** in the table — never omit a row. Use the Status column to indicate availability on the host (unavailable methods show ✗ with a short reason). Use this table format:\n\n| Method | Best for | Status |\n|--------|----------|--------|\n| **NGC Container** | All deps bundled (CUDA, TensorRT, LibTorch, ONNX Runtime, Vulkan); C++ + Python. Needs Docker + NVIDIA Container Toolkit. | ✓\u002F✗ based on docker presence |\n| **Debian\u002Fapt** | Native Ubuntu; C++ only | ✓\u002F✗ if package is installed |\n| **pip wheel** | Python-only projects; needs CUDA Toolkit on PATH; Python 3.10–3.13. | ✓\u002F✗ if wheel is installed in virtual env at ~\u002Fholoscan\u002Fvenv |\n| **Conda** | CUDA 13 only; good if already in a conda environment. | ✓\u002F✗ based on `check_conda.sh` output (not just `which conda`) |\n| **Source** | Modifying SDK internals, custom CMake flags, debug symbols, unsupported platform, or unreleased branch. | ✓\u002F✗ if already cloned at ~\u002Fholoscan\u002Fholoscan-sdk |\n\nAfter the table, end the turn with this exact two-line shape:\n\n> **Recommendation:** `\u003Cmethod>` — `\u003Cone-line why>`\n>\n> **Which method would you like to use?** (container \u002F apt \u002F wheel \u002F conda \u002F source)\n\nIf the user is new to Holoscan and the host is a supported x86_64 platform with Docker available, recommend **NGC Container**. For RHEL 9 or other container-only hosts, recommend container. For Python-only projects on a Docker-less host, recommend pip wheel.\n\nDo **not** include `docker pull`, `docker run`, `apt install`, or `pip install` commands in this turn — those live in the install skill invoked in Step 6. Keep this response short to avoid being truncated mid-table.\n\n### Step 6: Delegate to the Install Skill\n\nOnce a method is picked, invoke the corresponding skill — do not repeat the install steps inline:\n\n| Method | Skill to invoke |\n|--------|-----------------|\n| NGC Container | `\u002Fholoscan-install-container` |\n| Debian\u002Fapt | `\u002Fholoscan-install-debian` |\n| pip wheel | `\u002Fholoscan-install-wheel` |\n| Conda | `\u002Fholoscan-install-conda` |\n| Source | `\u002Fholoscan-install-source` |\n\nPass the CUDA variant (cu12\u002Fcu13\u002Figpu) and any other relevant facts from Steps 2–4 as context when invoking the skill.\n\nThe install skill owns the full command set — including the recommended container flags (`--gpus all`, `--ipc=host`, `--ulimit memlock=-1`, `--ulimit stack=67108864`, inner `ulimit -s 32768`) and verification examples. Do not restate them from `holoscan-setup`; delegate and let the install skill produce them.\n\n### Step 7: Summary\n\nIf installation was successful and tests were run, print a table summary of test results.\n\n## Limitations\n\n- RHEL 9.x supports the NGC container method only — native packages are not published.\n- Conda packages are CUDA 13 only; CUDA 12 hosts must use container, apt, pip wheel, or source.\n- Debian\u002Fapt installs C++ only since Holoscan v3.0.0; Python support requires an additional pip wheel install.\n- pip wheel requires glibc ≥ 2.35 and Python 3.10–3.13.\n- Native installs cannot exceed the driver's reported CUDA Version; only containers can use CUDA Forward Compatibility.\n- DGX Spark \u002F Grace-Hopper OS requirements change between releases — always re-check `sdk_installation.html`.\n\n## Troubleshooting\n\n- **`conda --version` says \"command not found\" but Conda is installed** — common in zsh setups with lazy-loaded conda or when only `.bashrc` ran `conda init`. Use `run_script(\"scripts\u002Fcheck_conda.sh\")`; it searches install dirs and rc files.\n- **`nvidia-smi` shows a lower CUDA Version than expected** — that field is the driver's max supported CUDA, not the installed toolkit. Upgrade the driver before installing a newer-CUDA package.\n- **Debian install succeeds but `import holoscan` fails in Python** — apt installs C++ only since v3.0.0. Follow up with `\u002Fholoscan-install-wheel`.\n- **`pip install holoscan` fails with glibc errors** — host glibc is \u003C 2.35. Use container or apt instead.\n- **`check_ngc_image.sh` reports image missing** — confirm NGC login (`docker login nvcr.io`) and that the tag suffix matches the CUDA variant rule in Step 4.\n",{"data":32,"body":42},{"name":4,"version":33,"description":6,"license":23,"metadata":34},"1.0.0",{"author":35,"github-url":36,"tags":37},"Holoscan Team \u003Choloscan-team@nvidia.com>","https:\u002F\u002Fgithub.com\u002Fnvidia-holoscan\u002Fholoscan-sdk",[38,39,8,40,41],"holoscan","installation","sdk","setup",{"type":43,"children":44},"root",[45,54,61,67,73,131,137,283,320,326,331,338,448,454,475,481,486,642,682,688,801,807,811,1060,1072,1142,1147,1204,1217,1273,1278,1288,1441,1446,1452,1464,1610,1615,1653,1664,1701,1707,1712,1813,1818,1867,1873,1878,1884,1923,1929,2049],{"type":46,"tag":47,"props":48,"children":50},"element","h1",{"id":49},"holoscan-sdk-setup",[51],{"type":52,"value":53},"text","Holoscan SDK Setup",{"type":46,"tag":55,"props":56,"children":58},"h2",{"id":57},"purpose",[59],{"type":52,"value":60},"Purpose",{"type":46,"tag":62,"props":63,"children":64},"p",{},[65],{"type":52,"value":66},"Determines the correct Holoscan SDK installation method for the current host by inspecting hardware, OS, CUDA driver, and existing tooling, then delegates to a method-specific install skill. Covers NGC container, Debian\u002Fapt, pip wheel, Conda, and source builds across Ubuntu, RHEL, IGX Orin, Jetson, and DGX Spark \u002F Grace-Hopper platforms.",{"type":46,"tag":55,"props":68,"children":70},{"id":69},"prerequisites",[71],{"type":52,"value":72},"Prerequisites",{"type":46,"tag":74,"props":75,"children":76},"ul",{},[77,83,97,110],{"type":46,"tag":78,"props":79,"children":80},"li",{},[81],{"type":52,"value":82},"Linux host (Ubuntu 22.04\u002F24.04, RHEL 9.x, IGX Orin, Jetson, or DGX Spark \u002F Grace-Hopper)",{"type":46,"tag":78,"props":84,"children":85},{},[86,88,95],{"type":52,"value":87},"NVIDIA GPU with a working driver (",{"type":46,"tag":89,"props":90,"children":92},"code",{"className":91},[],[93],{"type":52,"value":94},"nvidia-smi",{"type":52,"value":96}," returns a CUDA Version)",{"type":46,"tag":78,"props":98,"children":99},{},[100,102,108],{"type":52,"value":101},"Network access to ",{"type":46,"tag":89,"props":103,"children":105},{"className":104},[],[106],{"type":52,"value":107},"docs.nvidia.com",{"type":52,"value":109}," and NGC",{"type":46,"tag":78,"props":111,"children":112},{},[113,115,121,123,129],{"type":52,"value":114},"One of: Docker + NVIDIA Container Toolkit, ",{"type":46,"tag":89,"props":116,"children":118},{"className":117},[],[119],{"type":52,"value":120},"apt",{"type":52,"value":122},", Python 3.10–3.13 with ",{"type":46,"tag":89,"props":124,"children":126},{"className":125},[],[127],{"type":52,"value":128},"pip",{"type":52,"value":130},", Conda, or a build toolchain — depending on chosen method",{"type":46,"tag":55,"props":132,"children":134},{"id":133},"available-scripts",[135],{"type":52,"value":136},"Available Scripts",{"type":46,"tag":138,"props":139,"children":140},"table",{},[141,164],{"type":46,"tag":142,"props":143,"children":144},"thead",{},[145],{"type":46,"tag":146,"props":147,"children":148},"tr",{},[149,155,159],{"type":46,"tag":150,"props":151,"children":152},"th",{},[153],{"type":52,"value":154},"Script",{"type":46,"tag":150,"props":156,"children":157},{},[158],{"type":52,"value":60},{"type":46,"tag":150,"props":160,"children":161},{},[162],{"type":52,"value":163},"Arguments",{"type":46,"tag":165,"props":166,"children":167},"tbody",{},[168,235],{"type":46,"tag":146,"props":169,"children":170},{},[171,181,230],{"type":46,"tag":172,"props":173,"children":174},"td",{},[175],{"type":46,"tag":89,"props":176,"children":178},{"className":177},[],[179],{"type":52,"value":180},"scripts\u002Fcheck_conda.sh",{"type":46,"tag":172,"props":182,"children":183},{},[184,186,192,194,200,201,207,208,214,215,221,223,228],{"type":52,"value":185},"Detects Conda installs even when not on PATH (searches ",{"type":46,"tag":89,"props":187,"children":189},{"className":188},[],[190],{"type":52,"value":191},"~\u002Fminiconda3",{"type":52,"value":193},", ",{"type":46,"tag":89,"props":195,"children":197},{"className":196},[],[198],{"type":52,"value":199},"~\u002Fminiforge3",{"type":52,"value":193},{"type":46,"tag":89,"props":202,"children":204},{"className":203},[],[205],{"type":52,"value":206},"~\u002Fanaconda3",{"type":52,"value":193},{"type":46,"tag":89,"props":209,"children":211},{"className":210},[],[212],{"type":52,"value":213},"~\u002Fmambaforge",{"type":52,"value":193},{"type":46,"tag":89,"props":216,"children":218},{"className":217},[],[219],{"type":52,"value":220},"\u002Fopt\u002Fconda",{"type":52,"value":222},", and shell rc files); reports envs and which have ",{"type":46,"tag":89,"props":224,"children":226},{"className":225},[],[227],{"type":52,"value":38},{"type":52,"value":229}," importable.",{"type":46,"tag":172,"props":231,"children":232},{},[233],{"type":52,"value":234},"none",{"type":46,"tag":146,"props":236,"children":237},{},[238,247,252],{"type":46,"tag":172,"props":239,"children":240},{},[241],{"type":46,"tag":89,"props":242,"children":244},{"className":243},[],[245],{"type":52,"value":246},"scripts\u002Fcheck_ngc_image.sh",{"type":46,"tag":172,"props":248,"children":249},{},[250],{"type":52,"value":251},"Checks whether the NGC Holoscan container image for a given CUDA tag suffix is pulled or available.",{"type":46,"tag":172,"props":253,"children":254},{},[255,261,263,269,270,276,277],{"type":46,"tag":89,"props":256,"children":258},{"className":257},[],[259],{"type":52,"value":260},"\u003Ccuda-tag-suffix>",{"type":52,"value":262}," — one of ",{"type":46,"tag":89,"props":264,"children":266},{"className":265},[],[267],{"type":52,"value":268},"cuda13",{"type":52,"value":193},{"type":46,"tag":89,"props":271,"children":273},{"className":272},[],[274],{"type":52,"value":275},"cuda12-dgpu",{"type":52,"value":193},{"type":46,"tag":89,"props":278,"children":280},{"className":279},[],[281],{"type":52,"value":282},"cuda12-igpu",{"type":46,"tag":62,"props":284,"children":285},{},[286,288,294,296,302,304,310,312,318],{"type":52,"value":287},"Invoke scripts with ",{"type":46,"tag":89,"props":289,"children":291},{"className":290},[],[292],{"type":52,"value":293},"run_script(\"scripts\u002Fcheck_conda.sh\")",{"type":52,"value":295}," and ",{"type":46,"tag":89,"props":297,"children":299},{"className":298},[],[300],{"type":52,"value":301},"run_script(\"scripts\u002Fcheck_ngc_image.sh\", \"cuda13\")",{"type":52,"value":303},". Trust the script output over bare commands such as ",{"type":46,"tag":89,"props":305,"children":307},{"className":306},[],[308],{"type":52,"value":309},"which conda",{"type":52,"value":311}," or ",{"type":46,"tag":89,"props":313,"children":315},{"className":314},[],[316],{"type":52,"value":317},"docker images",{"type":52,"value":319},".",{"type":46,"tag":55,"props":321,"children":323},{"id":322},"instructions",[324],{"type":52,"value":325},"Instructions",{"type":46,"tag":62,"props":327,"children":328},{},[329],{"type":52,"value":330},"Be conversational and step-by-step — do not front-load all the information. Complete each step and report back before moving on.",{"type":46,"tag":332,"props":333,"children":335},"h3",{"id":334},"workflow-rules-must-follow",[336],{"type":52,"value":337},"Workflow rules (must follow)",{"type":46,"tag":339,"props":340,"children":341},"ol",{},[342,363,381,422],{"type":46,"tag":78,"props":343,"children":344},{},[345,347,353,355,361],{"type":52,"value":346},"End Step 5 with a ",{"type":46,"tag":348,"props":349,"children":350},"strong",{},[351],{"type":52,"value":352},"bolded one-line recommendation",{"type":52,"value":354}," that names the method (e.g. ",{"type":46,"tag":89,"props":356,"children":358},{"className":357},[],[359],{"type":52,"value":360},"**Recommendation:** NGC Container — bundles all deps, fastest path to a working install.",{"type":52,"value":362},").",{"type":46,"tag":78,"props":364,"children":365},{},[366,368,373,375,380],{"type":52,"value":367},"For a first-time user on a supported x86_64 host with Docker available, that recommendation ",{"type":46,"tag":348,"props":369,"children":370},{},[371],{"type":52,"value":372},"must",{"type":52,"value":374}," be ",{"type":46,"tag":348,"props":376,"children":377},{},[378],{"type":52,"value":379},"NGC Container",{"type":52,"value":319},{"type":46,"tag":78,"props":382,"children":383},{},[384,386,391,393,399,400,406,407,413,414,420],{"type":52,"value":385},"After the recommendation, ",{"type":46,"tag":348,"props":387,"children":388},{},[389],{"type":52,"value":390},"stop and ask",{"type":52,"value":392}," which method to use. Do not paste ",{"type":46,"tag":89,"props":394,"children":396},{"className":395},[],[397],{"type":52,"value":398},"docker pull",{"type":52,"value":193},{"type":46,"tag":89,"props":401,"children":403},{"className":402},[],[404],{"type":52,"value":405},"docker run",{"type":52,"value":193},{"type":46,"tag":89,"props":408,"children":410},{"className":409},[],[411],{"type":52,"value":412},"apt install",{"type":52,"value":193},{"type":46,"tag":89,"props":415,"children":417},{"className":416},[],[418],{"type":52,"value":419},"pip install",{"type":52,"value":421},", or other install commands in that turn — those belong to the delegated install skill in Step 6.",{"type":46,"tag":78,"props":423,"children":424},{},[425,427,432,434,439,440,446],{"type":52,"value":426},"If the container path is in play, verify Docker + GPU passthrough ",{"type":46,"tag":348,"props":428,"children":429},{},[430],{"type":52,"value":431},"yourself",{"type":52,"value":433}," in Step 4 (run the command shown there). Do not ask the user to run ",{"type":46,"tag":89,"props":435,"children":437},{"className":436},[],[438],{"type":52,"value":94},{"type":52,"value":311},{"type":46,"tag":89,"props":441,"children":443},{"className":442},[],[444],{"type":52,"value":445},"docker --version",{"type":52,"value":447}," for you.",{"type":46,"tag":332,"props":449,"children":451},{"id":450},"step-1-read-the-docs-first",[452],{"type":52,"value":453},"Step 1: Read the Docs First",{"type":46,"tag":62,"props":455,"children":456},{},[457,459,465,467,473],{"type":52,"value":458},"Fetch ",{"type":46,"tag":89,"props":460,"children":462},{"className":461},[],[463],{"type":52,"value":464},"https:\u002F\u002Fdocs.nvidia.com\u002Fholoscan\u002Fsdk-user-guide\u002F",{"type":52,"value":466}," then ",{"type":46,"tag":89,"props":468,"children":470},{"className":469},[],[471],{"type":52,"value":472},"sdk_installation.html",{"type":52,"value":474}," to get the current release's supported platforms, package names, and install requirements. Do not rely on hardcoded assumptions.",{"type":46,"tag":332,"props":476,"children":478},{"id":477},"step-2-inspect-the-machine",[479],{"type":52,"value":480},"Step 2: Inspect the Machine",{"type":46,"tag":62,"props":482,"children":483},{},[484],{"type":52,"value":485},"Run in parallel:",{"type":46,"tag":487,"props":488,"children":493},"pre",{"className":489,"code":490,"language":491,"meta":492,"style":492},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","uname -a && (lsb_release -a 2>\u002Fdev\u002Fnull || cat \u002Fetc\u002Fos-release)\nuname -m\nnvidia-smi 2>&1 | head -10\nnproc && free -h | head -2\n","bash","",[494],{"type":46,"tag":89,"props":495,"children":496},{"__ignoreMap":492},[497,565,578,606],{"type":46,"tag":498,"props":499,"children":502},"span",{"class":500,"line":501},"line",1,[503,509,515,521,526,531,535,540,545,550,555,560],{"type":46,"tag":498,"props":504,"children":506},{"style":505},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[507],{"type":52,"value":508},"uname",{"type":46,"tag":498,"props":510,"children":512},{"style":511},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[513],{"type":52,"value":514}," -a",{"type":46,"tag":498,"props":516,"children":518},{"style":517},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[519],{"type":52,"value":520}," &&",{"type":46,"tag":498,"props":522,"children":523},{"style":517},[524],{"type":52,"value":525}," (",{"type":46,"tag":498,"props":527,"children":528},{"style":505},[529],{"type":52,"value":530},"lsb_release",{"type":46,"tag":498,"props":532,"children":533},{"style":511},[534],{"type":52,"value":514},{"type":46,"tag":498,"props":536,"children":537},{"style":517},[538],{"type":52,"value":539}," 2>",{"type":46,"tag":498,"props":541,"children":542},{"style":511},[543],{"type":52,"value":544},"\u002Fdev\u002Fnull",{"type":46,"tag":498,"props":546,"children":547},{"style":517},[548],{"type":52,"value":549}," ||",{"type":46,"tag":498,"props":551,"children":552},{"style":505},[553],{"type":52,"value":554}," cat",{"type":46,"tag":498,"props":556,"children":557},{"style":511},[558],{"type":52,"value":559}," \u002Fetc\u002Fos-release",{"type":46,"tag":498,"props":561,"children":562},{"style":517},[563],{"type":52,"value":564},")\n",{"type":46,"tag":498,"props":566,"children":568},{"class":500,"line":567},2,[569,573],{"type":46,"tag":498,"props":570,"children":571},{"style":505},[572],{"type":52,"value":508},{"type":46,"tag":498,"props":574,"children":575},{"style":511},[576],{"type":52,"value":577}," -m\n",{"type":46,"tag":498,"props":579,"children":581},{"class":500,"line":580},3,[582,586,591,596,601],{"type":46,"tag":498,"props":583,"children":584},{"style":505},[585],{"type":52,"value":94},{"type":46,"tag":498,"props":587,"children":588},{"style":517},[589],{"type":52,"value":590}," 2>&1",{"type":46,"tag":498,"props":592,"children":593},{"style":517},[594],{"type":52,"value":595}," |",{"type":46,"tag":498,"props":597,"children":598},{"style":505},[599],{"type":52,"value":600}," head",{"type":46,"tag":498,"props":602,"children":603},{"style":511},[604],{"type":52,"value":605}," -10\n",{"type":46,"tag":498,"props":607,"children":609},{"class":500,"line":608},4,[610,615,619,624,629,633,637],{"type":46,"tag":498,"props":611,"children":612},{"style":505},[613],{"type":52,"value":614},"nproc",{"type":46,"tag":498,"props":616,"children":617},{"style":517},[618],{"type":52,"value":520},{"type":46,"tag":498,"props":620,"children":621},{"style":505},[622],{"type":52,"value":623}," free",{"type":46,"tag":498,"props":625,"children":626},{"style":511},[627],{"type":52,"value":628}," -h",{"type":46,"tag":498,"props":630,"children":631},{"style":517},[632],{"type":52,"value":595},{"type":46,"tag":498,"props":634,"children":635},{"style":505},[636],{"type":52,"value":600},{"type":46,"tag":498,"props":638,"children":639},{"style":511},[640],{"type":52,"value":641}," -2\n",{"type":46,"tag":62,"props":643,"children":644},{},[645,650,652,657,659,665,667,673,675,680],{"type":46,"tag":348,"props":646,"children":647},{},[648],{"type":52,"value":649},"Key:",{"type":52,"value":651}," Read the \"CUDA Version\" field from ",{"type":46,"tag":89,"props":653,"children":655},{"className":654},[],[656],{"type":52,"value":94},{"type":52,"value":658}," (top-right of the table header) — this is the ",{"type":46,"tag":660,"props":661,"children":662},"em",{},[663],{"type":52,"value":664},"maximum",{"type":52,"value":666}," CUDA version the driver supports, and drives ",{"type":46,"tag":89,"props":668,"children":670},{"className":669},[],[671],{"type":52,"value":672},"cuda12",{"type":52,"value":674}," vs ",{"type":46,"tag":89,"props":676,"children":678},{"className":677},[],[679],{"type":52,"value":268},{"type":52,"value":681}," package selection.",{"type":46,"tag":332,"props":683,"children":685},{"id":684},"step-3-assess-compatibility",[686],{"type":52,"value":687},"Step 3: Assess Compatibility",{"type":46,"tag":138,"props":689,"children":690},{},[691,707],{"type":46,"tag":142,"props":692,"children":693},{},[694],{"type":46,"tag":146,"props":695,"children":696},{},[697,702],{"type":46,"tag":150,"props":698,"children":699},{},[700],{"type":52,"value":701},"Platform",{"type":46,"tag":150,"props":703,"children":704},{},[705],{"type":52,"value":706},"Methods Available",{"type":46,"tag":165,"props":708,"children":709},{},[710,723,736,749,762,775,788],{"type":46,"tag":146,"props":711,"children":712},{},[713,718],{"type":46,"tag":172,"props":714,"children":715},{},[716],{"type":52,"value":717},"Ubuntu 22.04\u002F24.04, x86_64",{"type":46,"tag":172,"props":719,"children":720},{},[721],{"type":52,"value":722},"Container, Debian\u002Fapt, pip wheel, Conda, Source",{"type":46,"tag":146,"props":724,"children":725},{},[726,731],{"type":46,"tag":172,"props":727,"children":728},{},[729],{"type":52,"value":730},"RHEL 9.x, x86_64",{"type":46,"tag":172,"props":732,"children":733},{},[734],{"type":52,"value":735},"Container only",{"type":46,"tag":146,"props":737,"children":738},{},[739,744],{"type":46,"tag":172,"props":740,"children":741},{},[742],{"type":52,"value":743},"IGX Orin (ARM64)",{"type":46,"tag":172,"props":745,"children":746},{},[747],{"type":52,"value":748},"Container, Debian\u002Fapt, Source",{"type":46,"tag":146,"props":750,"children":751},{},[752,757],{"type":46,"tag":172,"props":753,"children":754},{},[755],{"type":52,"value":756},"Jetson AGX Orin \u002F Orin Nano",{"type":46,"tag":172,"props":758,"children":759},{},[760],{"type":52,"value":761},"Container, Debian\u002Fapt (iGPU)",{"type":46,"tag":146,"props":763,"children":764},{},[765,770],{"type":46,"tag":172,"props":766,"children":767},{},[768],{"type":52,"value":769},"Jetson AGX Thor",{"type":46,"tag":172,"props":771,"children":772},{},[773],{"type":52,"value":774},"Container, Debian\u002Fapt",{"type":46,"tag":146,"props":776,"children":777},{},[778,783],{"type":46,"tag":172,"props":779,"children":780},{},[781],{"type":52,"value":782},"DGX Spark \u002F Grace-Hopper",{"type":46,"tag":172,"props":784,"children":785},{},[786],{"type":52,"value":787},"Container (check docs for OS requirements)",{"type":46,"tag":146,"props":789,"children":790},{},[791,796],{"type":46,"tag":172,"props":792,"children":793},{},[794],{"type":52,"value":795},"Other Linux, x86_64",{"type":46,"tag":172,"props":797,"children":798},{},[799],{"type":52,"value":800},"Container may work; pip wheel if glibc ≥ 2.35",{"type":46,"tag":332,"props":802,"children":804},{"id":803},"step-4-check-tools-and-present-options",[805],{"type":52,"value":806},"Step 4: Check Tools and Present Options",{"type":46,"tag":62,"props":808,"children":809},{},[810],{"type":52,"value":485},{"type":46,"tag":487,"props":812,"children":814},{"className":489,"code":813,"language":491,"meta":492,"style":492},"docker --version 2>&1 | head -1; python3 --version 2>&1; pip3 --version 2>&1\ndpkg -l | grep holoscan || true\npip3 show holoscan 2>\u002Fdev\u002Fnull | grep -E \"^(Name|Version)\" || true\n~\u002Fholoscan\u002Fvenv\u002Fbin\u002Fpip show holoscan 2>\u002Fdev\u002Fnull | grep -E \"^(Name|Version)\" | sed 's\u002F^\u002Fvenv: \u002F' || true\n",[815],{"type":46,"tag":89,"props":816,"children":817},{"__ignoreMap":492},[818,881,918,979],{"type":46,"tag":498,"props":819,"children":820},{"class":500,"line":501},[821,826,831,835,839,843,848,853,858,862,867,872,876],{"type":46,"tag":498,"props":822,"children":823},{"style":505},[824],{"type":52,"value":825},"docker",{"type":46,"tag":498,"props":827,"children":828},{"style":511},[829],{"type":52,"value":830}," --version",{"type":46,"tag":498,"props":832,"children":833},{"style":517},[834],{"type":52,"value":590},{"type":46,"tag":498,"props":836,"children":837},{"style":517},[838],{"type":52,"value":595},{"type":46,"tag":498,"props":840,"children":841},{"style":505},[842],{"type":52,"value":600},{"type":46,"tag":498,"props":844,"children":845},{"style":511},[846],{"type":52,"value":847}," -1",{"type":46,"tag":498,"props":849,"children":850},{"style":517},[851],{"type":52,"value":852},";",{"type":46,"tag":498,"props":854,"children":855},{"style":505},[856],{"type":52,"value":857}," python3",{"type":46,"tag":498,"props":859,"children":860},{"style":511},[861],{"type":52,"value":830},{"type":46,"tag":498,"props":863,"children":864},{"style":517},[865],{"type":52,"value":866}," 2>&1;",{"type":46,"tag":498,"props":868,"children":869},{"style":505},[870],{"type":52,"value":871}," pip3",{"type":46,"tag":498,"props":873,"children":874},{"style":511},[875],{"type":52,"value":830},{"type":46,"tag":498,"props":877,"children":878},{"style":517},[879],{"type":52,"value":880}," 2>&1\n",{"type":46,"tag":498,"props":882,"children":883},{"class":500,"line":567},[884,889,894,898,903,908,912],{"type":46,"tag":498,"props":885,"children":886},{"style":505},[887],{"type":52,"value":888},"dpkg",{"type":46,"tag":498,"props":890,"children":891},{"style":511},[892],{"type":52,"value":893}," -l",{"type":46,"tag":498,"props":895,"children":896},{"style":517},[897],{"type":52,"value":595},{"type":46,"tag":498,"props":899,"children":900},{"style":505},[901],{"type":52,"value":902}," grep",{"type":46,"tag":498,"props":904,"children":905},{"style":511},[906],{"type":52,"value":907}," holoscan",{"type":46,"tag":498,"props":909,"children":910},{"style":517},[911],{"type":52,"value":549},{"type":46,"tag":498,"props":913,"children":915},{"style":914},"--shiki-light:#6182B8;--shiki-default:#82AAFF;--shiki-dark:#82AAFF",[916],{"type":52,"value":917}," true\n",{"type":46,"tag":498,"props":919,"children":920},{"class":500,"line":580},[921,926,931,935,939,943,947,951,956,961,966,971,975],{"type":46,"tag":498,"props":922,"children":923},{"style":505},[924],{"type":52,"value":925},"pip3",{"type":46,"tag":498,"props":927,"children":928},{"style":511},[929],{"type":52,"value":930}," show",{"type":46,"tag":498,"props":932,"children":933},{"style":511},[934],{"type":52,"value":907},{"type":46,"tag":498,"props":936,"children":937},{"style":517},[938],{"type":52,"value":539},{"type":46,"tag":498,"props":940,"children":941},{"style":511},[942],{"type":52,"value":544},{"type":46,"tag":498,"props":944,"children":945},{"style":517},[946],{"type":52,"value":595},{"type":46,"tag":498,"props":948,"children":949},{"style":505},[950],{"type":52,"value":902},{"type":46,"tag":498,"props":952,"children":953},{"style":511},[954],{"type":52,"value":955}," -E",{"type":46,"tag":498,"props":957,"children":958},{"style":517},[959],{"type":52,"value":960}," \"",{"type":46,"tag":498,"props":962,"children":963},{"style":511},[964],{"type":52,"value":965},"^(Name|Version)",{"type":46,"tag":498,"props":967,"children":968},{"style":517},[969],{"type":52,"value":970},"\"",{"type":46,"tag":498,"props":972,"children":973},{"style":517},[974],{"type":52,"value":549},{"type":46,"tag":498,"props":976,"children":977},{"style":914},[978],{"type":52,"value":917},{"type":46,"tag":498,"props":980,"children":981},{"class":500,"line":608},[982,987,993,998,1003,1008,1012,1016,1020,1024,1028,1032,1037,1042,1047,1052,1056],{"type":46,"tag":498,"props":983,"children":984},{"style":517},[985],{"type":52,"value":986},"~",{"type":46,"tag":498,"props":988,"children":990},{"style":989},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[991],{"type":52,"value":992},"\u002Fholoscan\u002Fvenv\u002Fbin\u002Fpip show holoscan ",{"type":46,"tag":498,"props":994,"children":995},{"style":517},[996],{"type":52,"value":997},"2>",{"type":46,"tag":498,"props":999,"children":1000},{"style":989},[1001],{"type":52,"value":1002},"\u002Fdev\u002Fnull ",{"type":46,"tag":498,"props":1004,"children":1005},{"style":517},[1006],{"type":52,"value":1007},"|",{"type":46,"tag":498,"props":1009,"children":1010},{"style":505},[1011],{"type":52,"value":902},{"type":46,"tag":498,"props":1013,"children":1014},{"style":511},[1015],{"type":52,"value":955},{"type":46,"tag":498,"props":1017,"children":1018},{"style":517},[1019],{"type":52,"value":960},{"type":46,"tag":498,"props":1021,"children":1022},{"style":511},[1023],{"type":52,"value":965},{"type":46,"tag":498,"props":1025,"children":1026},{"style":517},[1027],{"type":52,"value":970},{"type":46,"tag":498,"props":1029,"children":1030},{"style":517},[1031],{"type":52,"value":595},{"type":46,"tag":498,"props":1033,"children":1034},{"style":505},[1035],{"type":52,"value":1036}," sed",{"type":46,"tag":498,"props":1038,"children":1039},{"style":517},[1040],{"type":52,"value":1041}," '",{"type":46,"tag":498,"props":1043,"children":1044},{"style":511},[1045],{"type":52,"value":1046},"s\u002F^\u002Fvenv: \u002F",{"type":46,"tag":498,"props":1048,"children":1049},{"style":517},[1050],{"type":52,"value":1051},"'",{"type":46,"tag":498,"props":1053,"children":1054},{"style":517},[1055],{"type":52,"value":549},{"type":46,"tag":498,"props":1057,"children":1058},{"style":914},[1059],{"type":52,"value":917},{"type":46,"tag":62,"props":1061,"children":1062},{},[1063,1065,1070],{"type":52,"value":1064},"Then verify GPU passthrough yourself — do ",{"type":46,"tag":348,"props":1066,"children":1067},{},[1068],{"type":52,"value":1069},"not",{"type":52,"value":1071}," ask the user to run this:",{"type":46,"tag":487,"props":1073,"children":1075},{"className":489,"code":1074,"language":491,"meta":492,"style":492},"docker run --rm --gpus all ubuntu:22.04 nvidia-smi 2>&1 | tail -5 || true\n",[1076],{"type":46,"tag":89,"props":1077,"children":1078},{"__ignoreMap":492},[1079],{"type":46,"tag":498,"props":1080,"children":1081},{"class":500,"line":501},[1082,1086,1091,1096,1101,1106,1111,1116,1120,1124,1129,1134,1138],{"type":46,"tag":498,"props":1083,"children":1084},{"style":505},[1085],{"type":52,"value":825},{"type":46,"tag":498,"props":1087,"children":1088},{"style":511},[1089],{"type":52,"value":1090}," run",{"type":46,"tag":498,"props":1092,"children":1093},{"style":511},[1094],{"type":52,"value":1095}," --rm",{"type":46,"tag":498,"props":1097,"children":1098},{"style":511},[1099],{"type":52,"value":1100}," --gpus",{"type":46,"tag":498,"props":1102,"children":1103},{"style":511},[1104],{"type":52,"value":1105}," all",{"type":46,"tag":498,"props":1107,"children":1108},{"style":511},[1109],{"type":52,"value":1110}," ubuntu:22.04",{"type":46,"tag":498,"props":1112,"children":1113},{"style":511},[1114],{"type":52,"value":1115}," nvidia-smi",{"type":46,"tag":498,"props":1117,"children":1118},{"style":517},[1119],{"type":52,"value":590},{"type":46,"tag":498,"props":1121,"children":1122},{"style":517},[1123],{"type":52,"value":595},{"type":46,"tag":498,"props":1125,"children":1126},{"style":505},[1127],{"type":52,"value":1128}," tail",{"type":46,"tag":498,"props":1130,"children":1131},{"style":511},[1132],{"type":52,"value":1133}," -5",{"type":46,"tag":498,"props":1135,"children":1136},{"style":517},[1137],{"type":52,"value":549},{"type":46,"tag":498,"props":1139,"children":1140},{"style":914},[1141],{"type":52,"value":917},{"type":46,"tag":62,"props":1143,"children":1144},{},[1145],{"type":52,"value":1146},"Interpret the result for the Status column in Step 5:",{"type":46,"tag":74,"props":1148,"children":1149},{},[1150,1167,1187],{"type":46,"tag":78,"props":1151,"children":1152},{},[1153,1158,1160,1166],{"type":46,"tag":89,"props":1154,"children":1156},{"className":1155},[],[1157],{"type":52,"value":825},{"type":52,"value":1159}," missing → container row Status ",{"type":46,"tag":89,"props":1161,"children":1163},{"className":1162},[],[1164],{"type":52,"value":1165},"✗ — Docker not installed",{"type":52,"value":319},{"type":46,"tag":78,"props":1168,"children":1169},{},[1170,1172,1178,1180,1186],{"type":52,"value":1171},"Docker present but ",{"type":46,"tag":89,"props":1173,"children":1175},{"className":1174},[],[1176],{"type":52,"value":1177},"could not select device driver \"nvidia\"",{"type":52,"value":1179}," → ",{"type":46,"tag":89,"props":1181,"children":1183},{"className":1182},[],[1184],{"type":52,"value":1185},"✗ — NVIDIA Container Toolkit missing",{"type":52,"value":319},{"type":46,"tag":78,"props":1188,"children":1189},{},[1190,1195,1197,1203],{"type":46,"tag":89,"props":1191,"children":1193},{"className":1192},[],[1194],{"type":52,"value":94},{"type":52,"value":1196}," output appears → ",{"type":46,"tag":89,"props":1198,"children":1200},{"className":1199},[],[1201],{"type":52,"value":1202},"✓",{"type":52,"value":319},{"type":46,"tag":62,"props":1205,"children":1206},{},[1207,1209,1215],{"type":52,"value":1208},"Then invoke the detection scripts via ",{"type":46,"tag":89,"props":1210,"children":1212},{"className":1211},[],[1213],{"type":52,"value":1214},"run_script",{"type":52,"value":1216},":",{"type":46,"tag":74,"props":1218,"children":1219},{},[1220,1237],{"type":46,"tag":78,"props":1221,"children":1222},{},[1223,1228,1230,1236],{"type":46,"tag":89,"props":1224,"children":1226},{"className":1225},[],[1227],{"type":52,"value":293},{"type":52,"value":1229}," — see Available Scripts above for why this is preferred over ",{"type":46,"tag":89,"props":1231,"children":1233},{"className":1232},[],[1234],{"type":52,"value":1235},"conda --version",{"type":52,"value":319},{"type":46,"tag":78,"props":1238,"children":1239},{},[1240,1246,1248,1253,1255,1260,1261,1266,1267,1272],{"type":46,"tag":89,"props":1241,"children":1243},{"className":1242},[],[1244],{"type":52,"value":1245},"run_script(\"scripts\u002Fcheck_ngc_image.sh\", \"\u003Ccuda-tag-suffix>\")",{"type":52,"value":1247}," — replace ",{"type":46,"tag":89,"props":1249,"children":1251},{"className":1250},[],[1252],{"type":52,"value":260},{"type":52,"value":1254}," with the tag determined from Step 2 (e.g. ",{"type":46,"tag":89,"props":1256,"children":1258},{"className":1257},[],[1259],{"type":52,"value":268},{"type":52,"value":193},{"type":46,"tag":89,"props":1262,"children":1264},{"className":1263},[],[1265],{"type":52,"value":275},{"type":52,"value":193},{"type":46,"tag":89,"props":1268,"children":1270},{"className":1269},[],[1271],{"type":52,"value":282},{"type":52,"value":362},{"type":46,"tag":62,"props":1274,"children":1275},{},[1276],{"type":52,"value":1277},"If Holoscan is already installed, note the version and ask whether to upgrade or verify the existing install.",{"type":46,"tag":62,"props":1279,"children":1280},{},[1281,1286],{"type":46,"tag":348,"props":1282,"children":1283},{},[1284],{"type":52,"value":1285},"CUDA variant rule",{"type":52,"value":1287}," (canonical reference — apply this in all steps below):",{"type":46,"tag":138,"props":1289,"children":1290},{},[1291,1312],{"type":46,"tag":142,"props":1292,"children":1293},{},[1294],{"type":46,"tag":146,"props":1295,"children":1296},{},[1297,1302,1307],{"type":46,"tag":150,"props":1298,"children":1299},{},[1300],{"type":52,"value":1301},"nvidia-smi CUDA Version",{"type":46,"tag":150,"props":1303,"children":1304},{},[1305],{"type":52,"value":1306},"Native packages",{"type":46,"tag":150,"props":1308,"children":1309},{},[1310],{"type":52,"value":1311},"Container tag",{"type":46,"tag":165,"props":1313,"children":1314},{},[1315,1348,1387,1417],{"type":46,"tag":146,"props":1316,"children":1317},{},[1318,1323,1340],{"type":46,"tag":172,"props":1319,"children":1320},{},[1321],{"type":52,"value":1322},"13.x+",{"type":46,"tag":172,"props":1324,"children":1325},{},[1326,1332,1334],{"type":46,"tag":89,"props":1327,"children":1329},{"className":1328},[],[1330],{"type":52,"value":1331},"holoscan-cu13",{"type":52,"value":1333}," \u002F ",{"type":46,"tag":89,"props":1335,"children":1337},{"className":1336},[],[1338],{"type":52,"value":1339},"holoscan-cuda-13",{"type":46,"tag":172,"props":1341,"children":1342},{},[1343],{"type":46,"tag":89,"props":1344,"children":1346},{"className":1345},[],[1347],{"type":52,"value":268},{"type":46,"tag":146,"props":1349,"children":1350},{},[1351,1356,1372],{"type":46,"tag":172,"props":1352,"children":1353},{},[1354],{"type":52,"value":1355},"12.x, Blackwell GPU",{"type":46,"tag":172,"props":1357,"children":1358},{},[1359,1365,1366],{"type":46,"tag":89,"props":1360,"children":1362},{"className":1361},[],[1363],{"type":52,"value":1364},"holoscan-cu12",{"type":52,"value":1333},{"type":46,"tag":89,"props":1367,"children":1369},{"className":1368},[],[1370],{"type":52,"value":1371},"holoscan-cuda-12",{"type":46,"tag":172,"props":1373,"children":1374},{},[1375,1380,1382],{"type":46,"tag":89,"props":1376,"children":1378},{"className":1377},[],[1379],{"type":52,"value":268},{"type":52,"value":1381}," (Forward Compat) or ",{"type":46,"tag":89,"props":1383,"children":1385},{"className":1384},[],[1386],{"type":52,"value":275},{"type":46,"tag":146,"props":1388,"children":1389},{},[1390,1395,1409],{"type":46,"tag":172,"props":1391,"children":1392},{},[1393],{"type":52,"value":1394},"12.x, Ampere\u002FAda dGPU",{"type":46,"tag":172,"props":1396,"children":1397},{},[1398,1403,1404],{"type":46,"tag":89,"props":1399,"children":1401},{"className":1400},[],[1402],{"type":52,"value":1364},{"type":52,"value":1333},{"type":46,"tag":89,"props":1405,"children":1407},{"className":1406},[],[1408],{"type":52,"value":1371},{"type":46,"tag":172,"props":1410,"children":1411},{},[1412],{"type":46,"tag":89,"props":1413,"children":1415},{"className":1414},[],[1416],{"type":52,"value":275},{"type":46,"tag":146,"props":1418,"children":1419},{},[1420,1425,1433],{"type":46,"tag":172,"props":1421,"children":1422},{},[1423],{"type":52,"value":1424},"ARM64 iGPU (Jetson, IGX)",{"type":46,"tag":172,"props":1426,"children":1427},{},[1428],{"type":46,"tag":89,"props":1429,"children":1431},{"className":1430},[],[1432],{"type":52,"value":38},{"type":46,"tag":172,"props":1434,"children":1435},{},[1436],{"type":46,"tag":89,"props":1437,"children":1439},{"className":1438},[],[1440],{"type":52,"value":282},{"type":46,"tag":62,"props":1442,"children":1443},{},[1444],{"type":52,"value":1445},"Native installs treat the driver CUDA version as a hard ceiling. Containers support Forward Compatibility (banner saying \"CUDA Forward Compatibility mode ENABLED\" is expected, not an error).",{"type":46,"tag":332,"props":1447,"children":1449},{"id":1448},"step-5-present-options-and-recommend",[1450],{"type":52,"value":1451},"Step 5: Present Options and Recommend",{"type":46,"tag":62,"props":1453,"children":1454},{},[1455,1457,1462],{"type":52,"value":1456},"Always present ",{"type":46,"tag":348,"props":1458,"children":1459},{},[1460],{"type":52,"value":1461},"all methods",{"type":52,"value":1463}," in the table — never omit a row. Use the Status column to indicate availability on the host (unavailable methods show ✗ with a short reason). Use this table format:",{"type":46,"tag":138,"props":1465,"children":1466},{},[1467,1488],{"type":46,"tag":142,"props":1468,"children":1469},{},[1470],{"type":46,"tag":146,"props":1471,"children":1472},{},[1473,1478,1483],{"type":46,"tag":150,"props":1474,"children":1475},{},[1476],{"type":52,"value":1477},"Method",{"type":46,"tag":150,"props":1479,"children":1480},{},[1481],{"type":52,"value":1482},"Best for",{"type":46,"tag":150,"props":1484,"children":1485},{},[1486],{"type":52,"value":1487},"Status",{"type":46,"tag":165,"props":1489,"children":1490},{},[1491,1511,1532,1553,1589],{"type":46,"tag":146,"props":1492,"children":1493},{},[1494,1501,1506],{"type":46,"tag":172,"props":1495,"children":1496},{},[1497],{"type":46,"tag":348,"props":1498,"children":1499},{},[1500],{"type":52,"value":379},{"type":46,"tag":172,"props":1502,"children":1503},{},[1504],{"type":52,"value":1505},"All deps bundled (CUDA, TensorRT, LibTorch, ONNX Runtime, Vulkan); C++ + Python. Needs Docker + NVIDIA Container Toolkit.",{"type":46,"tag":172,"props":1507,"children":1508},{},[1509],{"type":52,"value":1510},"✓\u002F✗ based on docker presence",{"type":46,"tag":146,"props":1512,"children":1513},{},[1514,1522,1527],{"type":46,"tag":172,"props":1515,"children":1516},{},[1517],{"type":46,"tag":348,"props":1518,"children":1519},{},[1520],{"type":52,"value":1521},"Debian\u002Fapt",{"type":46,"tag":172,"props":1523,"children":1524},{},[1525],{"type":52,"value":1526},"Native Ubuntu; C++ only",{"type":46,"tag":172,"props":1528,"children":1529},{},[1530],{"type":52,"value":1531},"✓\u002F✗ if package is installed",{"type":46,"tag":146,"props":1533,"children":1534},{},[1535,1543,1548],{"type":46,"tag":172,"props":1536,"children":1537},{},[1538],{"type":46,"tag":348,"props":1539,"children":1540},{},[1541],{"type":52,"value":1542},"pip wheel",{"type":46,"tag":172,"props":1544,"children":1545},{},[1546],{"type":52,"value":1547},"Python-only projects; needs CUDA Toolkit on PATH; Python 3.10–3.13.",{"type":46,"tag":172,"props":1549,"children":1550},{},[1551],{"type":52,"value":1552},"✓\u002F✗ if wheel is installed in virtual env at ~\u002Fholoscan\u002Fvenv",{"type":46,"tag":146,"props":1554,"children":1555},{},[1556,1564,1569],{"type":46,"tag":172,"props":1557,"children":1558},{},[1559],{"type":46,"tag":348,"props":1560,"children":1561},{},[1562],{"type":52,"value":1563},"Conda",{"type":46,"tag":172,"props":1565,"children":1566},{},[1567],{"type":52,"value":1568},"CUDA 13 only; good if already in a conda environment.",{"type":46,"tag":172,"props":1570,"children":1571},{},[1572,1574,1580,1582,1587],{"type":52,"value":1573},"✓\u002F✗ based on ",{"type":46,"tag":89,"props":1575,"children":1577},{"className":1576},[],[1578],{"type":52,"value":1579},"check_conda.sh",{"type":52,"value":1581}," output (not just ",{"type":46,"tag":89,"props":1583,"children":1585},{"className":1584},[],[1586],{"type":52,"value":309},{"type":52,"value":1588},")",{"type":46,"tag":146,"props":1590,"children":1591},{},[1592,1600,1605],{"type":46,"tag":172,"props":1593,"children":1594},{},[1595],{"type":46,"tag":348,"props":1596,"children":1597},{},[1598],{"type":52,"value":1599},"Source",{"type":46,"tag":172,"props":1601,"children":1602},{},[1603],{"type":52,"value":1604},"Modifying SDK internals, custom CMake flags, debug symbols, unsupported platform, or unreleased branch.",{"type":46,"tag":172,"props":1606,"children":1607},{},[1608],{"type":52,"value":1609},"✓\u002F✗ if already cloned at ~\u002Fholoscan\u002Fholoscan-sdk",{"type":46,"tag":62,"props":1611,"children":1612},{},[1613],{"type":52,"value":1614},"After the table, end the turn with this exact two-line shape:",{"type":46,"tag":1616,"props":1617,"children":1618},"blockquote",{},[1619,1643],{"type":46,"tag":62,"props":1620,"children":1621},{},[1622,1627,1629,1635,1637],{"type":46,"tag":348,"props":1623,"children":1624},{},[1625],{"type":52,"value":1626},"Recommendation:",{"type":52,"value":1628}," ",{"type":46,"tag":89,"props":1630,"children":1632},{"className":1631},[],[1633],{"type":52,"value":1634},"\u003Cmethod>",{"type":52,"value":1636}," — ",{"type":46,"tag":89,"props":1638,"children":1640},{"className":1639},[],[1641],{"type":52,"value":1642},"\u003Cone-line why>",{"type":46,"tag":62,"props":1644,"children":1645},{},[1646,1651],{"type":46,"tag":348,"props":1647,"children":1648},{},[1649],{"type":52,"value":1650},"Which method would you like to use?",{"type":52,"value":1652}," (container \u002F apt \u002F wheel \u002F conda \u002F source)",{"type":46,"tag":62,"props":1654,"children":1655},{},[1656,1658,1662],{"type":52,"value":1657},"If the user is new to Holoscan and the host is a supported x86_64 platform with Docker available, recommend ",{"type":46,"tag":348,"props":1659,"children":1660},{},[1661],{"type":52,"value":379},{"type":52,"value":1663},". For RHEL 9 or other container-only hosts, recommend container. For Python-only projects on a Docker-less host, recommend pip wheel.",{"type":46,"tag":62,"props":1665,"children":1666},{},[1667,1669,1673,1675,1680,1681,1686,1687,1692,1694,1699],{"type":52,"value":1668},"Do ",{"type":46,"tag":348,"props":1670,"children":1671},{},[1672],{"type":52,"value":1069},{"type":52,"value":1674}," include ",{"type":46,"tag":89,"props":1676,"children":1678},{"className":1677},[],[1679],{"type":52,"value":398},{"type":52,"value":193},{"type":46,"tag":89,"props":1682,"children":1684},{"className":1683},[],[1685],{"type":52,"value":405},{"type":52,"value":193},{"type":46,"tag":89,"props":1688,"children":1690},{"className":1689},[],[1691],{"type":52,"value":412},{"type":52,"value":1693},", or ",{"type":46,"tag":89,"props":1695,"children":1697},{"className":1696},[],[1698],{"type":52,"value":419},{"type":52,"value":1700}," commands in this turn — those live in the install skill invoked in Step 6. Keep this response short to avoid being truncated mid-table.",{"type":46,"tag":332,"props":1702,"children":1704},{"id":1703},"step-6-delegate-to-the-install-skill",[1705],{"type":52,"value":1706},"Step 6: Delegate to the Install Skill",{"type":46,"tag":62,"props":1708,"children":1709},{},[1710],{"type":52,"value":1711},"Once a method is picked, invoke the corresponding skill — do not repeat the install steps inline:",{"type":46,"tag":138,"props":1713,"children":1714},{},[1715,1730],{"type":46,"tag":142,"props":1716,"children":1717},{},[1718],{"type":46,"tag":146,"props":1719,"children":1720},{},[1721,1725],{"type":46,"tag":150,"props":1722,"children":1723},{},[1724],{"type":52,"value":1477},{"type":46,"tag":150,"props":1726,"children":1727},{},[1728],{"type":52,"value":1729},"Skill to invoke",{"type":46,"tag":165,"props":1731,"children":1732},{},[1733,1749,1765,1781,1797],{"type":46,"tag":146,"props":1734,"children":1735},{},[1736,1740],{"type":46,"tag":172,"props":1737,"children":1738},{},[1739],{"type":52,"value":379},{"type":46,"tag":172,"props":1741,"children":1742},{},[1743],{"type":46,"tag":89,"props":1744,"children":1746},{"className":1745},[],[1747],{"type":52,"value":1748},"\u002Fholoscan-install-container",{"type":46,"tag":146,"props":1750,"children":1751},{},[1752,1756],{"type":46,"tag":172,"props":1753,"children":1754},{},[1755],{"type":52,"value":1521},{"type":46,"tag":172,"props":1757,"children":1758},{},[1759],{"type":46,"tag":89,"props":1760,"children":1762},{"className":1761},[],[1763],{"type":52,"value":1764},"\u002Fholoscan-install-debian",{"type":46,"tag":146,"props":1766,"children":1767},{},[1768,1772],{"type":46,"tag":172,"props":1769,"children":1770},{},[1771],{"type":52,"value":1542},{"type":46,"tag":172,"props":1773,"children":1774},{},[1775],{"type":46,"tag":89,"props":1776,"children":1778},{"className":1777},[],[1779],{"type":52,"value":1780},"\u002Fholoscan-install-wheel",{"type":46,"tag":146,"props":1782,"children":1783},{},[1784,1788],{"type":46,"tag":172,"props":1785,"children":1786},{},[1787],{"type":52,"value":1563},{"type":46,"tag":172,"props":1789,"children":1790},{},[1791],{"type":46,"tag":89,"props":1792,"children":1794},{"className":1793},[],[1795],{"type":52,"value":1796},"\u002Fholoscan-install-conda",{"type":46,"tag":146,"props":1798,"children":1799},{},[1800,1804],{"type":46,"tag":172,"props":1801,"children":1802},{},[1803],{"type":52,"value":1599},{"type":46,"tag":172,"props":1805,"children":1806},{},[1807],{"type":46,"tag":89,"props":1808,"children":1810},{"className":1809},[],[1811],{"type":52,"value":1812},"\u002Fholoscan-install-source",{"type":46,"tag":62,"props":1814,"children":1815},{},[1816],{"type":52,"value":1817},"Pass the CUDA variant (cu12\u002Fcu13\u002Figpu) and any other relevant facts from Steps 2–4 as context when invoking the skill.",{"type":46,"tag":62,"props":1819,"children":1820},{},[1821,1823,1829,1830,1836,1837,1843,1844,1850,1852,1858,1860,1865],{"type":52,"value":1822},"The install skill owns the full command set — including the recommended container flags (",{"type":46,"tag":89,"props":1824,"children":1826},{"className":1825},[],[1827],{"type":52,"value":1828},"--gpus all",{"type":52,"value":193},{"type":46,"tag":89,"props":1831,"children":1833},{"className":1832},[],[1834],{"type":52,"value":1835},"--ipc=host",{"type":52,"value":193},{"type":46,"tag":89,"props":1838,"children":1840},{"className":1839},[],[1841],{"type":52,"value":1842},"--ulimit memlock=-1",{"type":52,"value":193},{"type":46,"tag":89,"props":1845,"children":1847},{"className":1846},[],[1848],{"type":52,"value":1849},"--ulimit stack=67108864",{"type":52,"value":1851},", inner ",{"type":46,"tag":89,"props":1853,"children":1855},{"className":1854},[],[1856],{"type":52,"value":1857},"ulimit -s 32768",{"type":52,"value":1859},") and verification examples. Do not restate them from ",{"type":46,"tag":89,"props":1861,"children":1863},{"className":1862},[],[1864],{"type":52,"value":4},{"type":52,"value":1866},"; delegate and let the install skill produce them.",{"type":46,"tag":332,"props":1868,"children":1870},{"id":1869},"step-7-summary",[1871],{"type":52,"value":1872},"Step 7: Summary",{"type":46,"tag":62,"props":1874,"children":1875},{},[1876],{"type":52,"value":1877},"If installation was successful and tests were run, print a table summary of test results.",{"type":46,"tag":55,"props":1879,"children":1881},{"id":1880},"limitations",[1882],{"type":52,"value":1883},"Limitations",{"type":46,"tag":74,"props":1885,"children":1886},{},[1887,1892,1897,1902,1907,1912],{"type":46,"tag":78,"props":1888,"children":1889},{},[1890],{"type":52,"value":1891},"RHEL 9.x supports the NGC container method only — native packages are not published.",{"type":46,"tag":78,"props":1893,"children":1894},{},[1895],{"type":52,"value":1896},"Conda packages are CUDA 13 only; CUDA 12 hosts must use container, apt, pip wheel, or source.",{"type":46,"tag":78,"props":1898,"children":1899},{},[1900],{"type":52,"value":1901},"Debian\u002Fapt installs C++ only since Holoscan v3.0.0; Python support requires an additional pip wheel install.",{"type":46,"tag":78,"props":1903,"children":1904},{},[1905],{"type":52,"value":1906},"pip wheel requires glibc ≥ 2.35 and Python 3.10–3.13.",{"type":46,"tag":78,"props":1908,"children":1909},{},[1910],{"type":52,"value":1911},"Native installs cannot exceed the driver's reported CUDA Version; only containers can use CUDA Forward Compatibility.",{"type":46,"tag":78,"props":1913,"children":1914},{},[1915,1917,1922],{"type":52,"value":1916},"DGX Spark \u002F Grace-Hopper OS requirements change between releases — always re-check ",{"type":46,"tag":89,"props":1918,"children":1920},{"className":1919},[],[1921],{"type":52,"value":472},{"type":52,"value":319},{"type":46,"tag":55,"props":1924,"children":1926},{"id":1925},"troubleshooting",[1927],{"type":52,"value":1928},"Troubleshooting",{"type":46,"tag":74,"props":1930,"children":1931},{},[1932,1970,1985,2009,2025],{"type":46,"tag":78,"props":1933,"children":1934},{},[1935,1945,1947,1953,1955,1961,1963,1968],{"type":46,"tag":348,"props":1936,"children":1937},{},[1938,1943],{"type":46,"tag":89,"props":1939,"children":1941},{"className":1940},[],[1942],{"type":52,"value":1235},{"type":52,"value":1944}," says \"command not found\" but Conda is installed",{"type":52,"value":1946}," — common in zsh setups with lazy-loaded conda or when only ",{"type":46,"tag":89,"props":1948,"children":1950},{"className":1949},[],[1951],{"type":52,"value":1952},".bashrc",{"type":52,"value":1954}," ran ",{"type":46,"tag":89,"props":1956,"children":1958},{"className":1957},[],[1959],{"type":52,"value":1960},"conda init",{"type":52,"value":1962},". Use ",{"type":46,"tag":89,"props":1964,"children":1966},{"className":1965},[],[1967],{"type":52,"value":293},{"type":52,"value":1969},"; it searches install dirs and rc files.",{"type":46,"tag":78,"props":1971,"children":1972},{},[1973,1983],{"type":46,"tag":348,"props":1974,"children":1975},{},[1976,1981],{"type":46,"tag":89,"props":1977,"children":1979},{"className":1978},[],[1980],{"type":52,"value":94},{"type":52,"value":1982}," shows a lower CUDA Version than expected",{"type":52,"value":1984}," — that field is the driver's max supported CUDA, not the installed toolkit. Upgrade the driver before installing a newer-CUDA package.",{"type":46,"tag":78,"props":1986,"children":1987},{},[1988,2001,2003,2008],{"type":46,"tag":348,"props":1989,"children":1990},{},[1991,1993,1999],{"type":52,"value":1992},"Debian install succeeds but ",{"type":46,"tag":89,"props":1994,"children":1996},{"className":1995},[],[1997],{"type":52,"value":1998},"import holoscan",{"type":52,"value":2000}," fails in Python",{"type":52,"value":2002}," — apt installs C++ only since v3.0.0. Follow up with ",{"type":46,"tag":89,"props":2004,"children":2006},{"className":2005},[],[2007],{"type":52,"value":1780},{"type":52,"value":319},{"type":46,"tag":78,"props":2010,"children":2011},{},[2012,2023],{"type":46,"tag":348,"props":2013,"children":2014},{},[2015,2021],{"type":46,"tag":89,"props":2016,"children":2018},{"className":2017},[],[2019],{"type":52,"value":2020},"pip install holoscan",{"type":52,"value":2022}," fails with glibc errors",{"type":52,"value":2024}," — host glibc is \u003C 2.35. Use container or apt instead.",{"type":46,"tag":78,"props":2026,"children":2027},{},[2028,2039,2041,2047],{"type":46,"tag":348,"props":2029,"children":2030},{},[2031,2037],{"type":46,"tag":89,"props":2032,"children":2034},{"className":2033},[],[2035],{"type":52,"value":2036},"check_ngc_image.sh",{"type":52,"value":2038}," reports image missing",{"type":52,"value":2040}," — confirm NGC login (",{"type":46,"tag":89,"props":2042,"children":2044},{"className":2043},[],[2045],{"type":52,"value":2046},"docker login nvcr.io",{"type":52,"value":2048},") and that the tag suffix matches the CUDA variant rule in Step 4.",{"type":46,"tag":2050,"props":2051,"children":2052},"style",{},[2053],{"type":52,"value":2054},"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":2056,"total":2157},[2057,2074,2086,2100,2112,2129,2143],{"slug":2058,"name":2058,"fn":2059,"description":2060,"org":2061,"tags":2062,"stars":20,"repoUrl":21,"updatedAt":2073},"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},[2063,2066,2069,2070],{"name":2064,"slug":2065,"type":15},"Data Analysis","data-analysis",{"name":2067,"slug":2068,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2071,"slug":2072,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":2075,"name":2075,"fn":2076,"description":2077,"org":2078,"tags":2079,"stars":20,"repoUrl":21,"updatedAt":2085},"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},[2080,2081,2084],{"name":13,"slug":14,"type":15},{"name":2082,"slug":2083,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":2087,"name":2087,"fn":2088,"description":2089,"org":2090,"tags":2091,"stars":20,"repoUrl":21,"updatedAt":2099},"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},[2092,2095,2096],{"name":2093,"slug":2094,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":2097,"slug":2098,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":2101,"name":2101,"fn":2102,"description":2103,"org":2104,"tags":2105,"stars":20,"repoUrl":21,"updatedAt":2111},"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},[2106,2107,2108],{"name":2064,"slug":2065,"type":15},{"name":9,"slug":8,"type":15},{"name":2109,"slug":2110,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":2113,"name":2113,"fn":2114,"description":2115,"org":2116,"tags":2117,"stars":20,"repoUrl":21,"updatedAt":2128},"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},[2118,2121,2124,2125],{"name":2119,"slug":2120,"type":15},"Automation","automation",{"name":2122,"slug":2123,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":2126,"slug":2127,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":2130,"name":2130,"fn":2131,"description":2132,"org":2133,"tags":2134,"stars":20,"repoUrl":21,"updatedAt":2142},"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},[2135,2136,2138,2139],{"name":13,"slug":14,"type":15},{"name":2137,"slug":825,"type":15},"Docker",{"name":9,"slug":8,"type":15},{"name":2140,"slug":2141,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":2144,"name":2144,"fn":2145,"description":2146,"org":2147,"tags":2148,"stars":20,"repoUrl":21,"updatedAt":2156},"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},[2149,2150,2153],{"name":9,"slug":8,"type":15},{"name":2151,"slug":2152,"type":15},"Quantum Computing","quantum-computing",{"name":2154,"slug":2155,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":2159,"total":2310},[2160,2178,2194,2205,2217,2231,2244,2258,2269,2278,2292,2301],{"slug":2161,"name":2161,"fn":2162,"description":2163,"org":2164,"tags":2165,"stars":2175,"repoUrl":2176,"updatedAt":2177},"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},[2166,2169,2172],{"name":2167,"slug":2168,"type":15},"Documentation","documentation",{"name":2170,"slug":2171,"type":15},"MCP","mcp",{"name":2173,"slug":2174,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":2179,"name":2179,"fn":2180,"description":2181,"org":2182,"tags":2183,"stars":2191,"repoUrl":2192,"updatedAt":2193},"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},[2184,2187,2188],{"name":2185,"slug":2186,"type":15},"Containers","containers",{"name":13,"slug":14,"type":15},{"name":2189,"slug":2190,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":2195,"name":2195,"fn":2196,"description":2197,"org":2198,"tags":2199,"stars":2191,"repoUrl":2192,"updatedAt":2204},"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},[2200,2203],{"name":2201,"slug":2202,"type":15},"CI\u002FCD","ci-cd",{"name":13,"slug":14,"type":15},"2026-07-14T05:25:59.97109",{"slug":2206,"name":2206,"fn":2207,"description":2208,"org":2209,"tags":2210,"stars":2191,"repoUrl":2192,"updatedAt":2216},"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},[2211,2212,2213],{"name":2201,"slug":2202,"type":15},{"name":13,"slug":14,"type":15},{"name":2214,"slug":2215,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":2218,"name":2218,"fn":2219,"description":2220,"org":2221,"tags":2222,"stars":2191,"repoUrl":2192,"updatedAt":2230},"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},[2223,2226,2227],{"name":2224,"slug":2225,"type":15},"Debugging","debugging",{"name":2214,"slug":2215,"type":15},{"name":2228,"slug":2229,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":2232,"name":2232,"fn":2233,"description":2234,"org":2235,"tags":2236,"stars":2191,"repoUrl":2192,"updatedAt":2243},"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},[2237,2240],{"name":2238,"slug":2239,"type":15},"Best Practices","best-practices",{"name":2241,"slug":2242,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":2245,"name":2245,"fn":2246,"description":2247,"org":2248,"tags":2249,"stars":2191,"repoUrl":2192,"updatedAt":2257},"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},[2250,2253,2256],{"name":2251,"slug":2252,"type":15},"Machine Learning","machine-learning",{"name":2254,"slug":2255,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":2259,"name":2259,"fn":2260,"description":2261,"org":2262,"tags":2263,"stars":2191,"repoUrl":2192,"updatedAt":2268},"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},[2264,2267],{"name":2265,"slug":2266,"type":15},"QA","qa",{"name":2109,"slug":2110,"type":15},"2026-07-14T05:25:53.673039",{"slug":2270,"name":2270,"fn":2271,"description":2272,"org":2273,"tags":2274,"stars":2191,"repoUrl":2192,"updatedAt":2277},"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},[2275,2276],{"name":13,"slug":14,"type":15},{"name":2082,"slug":2083,"type":15},"2026-07-14T05:25:49.362534",{"slug":2279,"name":2279,"fn":2280,"description":2281,"org":2282,"tags":2283,"stars":2191,"repoUrl":2192,"updatedAt":2291},"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},[2284,2287,2288],{"name":2285,"slug":2286,"type":15},"Code Review","code-review",{"name":2214,"slug":2215,"type":15},{"name":2289,"slug":2290,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2293,"name":2293,"fn":2294,"description":2295,"org":2296,"tags":2297,"stars":2191,"repoUrl":2192,"updatedAt":2300},"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},[2298,2299],{"name":2265,"slug":2266,"type":15},{"name":2109,"slug":2110,"type":15},"2026-07-14T05:25:54.928983",{"slug":2302,"name":2302,"fn":2303,"description":2304,"org":2305,"tags":2306,"stars":2191,"repoUrl":2192,"updatedAt":2309},"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},[2307,2308],{"name":2119,"slug":2120,"type":15},{"name":2201,"slug":2202,"type":15},"2026-07-30T05:29:03.275638",496]