[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-holoscan-install-container":3,"mdc--ehdrh5-key":31,"related-repo-nvidia-holoscan-install-container":1787,"related-org-nvidia-holoscan-install-container":1890},{"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-install-container","install Holoscan SDK via Docker container","Install Holoscan SDK via the NGC Docker container. Use for container-based installs; not for native apt\u002Fpip\u002FConda installs.",{"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],{"name":13,"slug":14,"type":15},"Containers","containers","tag",{"name":17,"slug":18,"type":15},"Deployment","deployment",{"name":9,"slug":8,"type":15},2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:30:27.924792","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-install-container","---\nname: holoscan-install-container\nversion: \"1.0.0\"\ndescription: \"Install Holoscan SDK via the NGC Docker container. Use for container-based installs; not for native apt\u002Fpip\u002FConda installs.\"\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    - install\n    - container\n    - docker\n    - ngc\n---\n\n# Holoscan NGC Container Installation\n\n## Purpose\n\nPull and verify the official Holoscan SDK container from NGC (`nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan`), selecting the right CUDA\u002Farch tag for the host GPU and validating with the bundled Python and C++ examples.\n\n## Prerequisites\n\n- Linux host with an NVIDIA GPU and a working driver (`nvidia-smi`).\n- Docker installed and the user in the `docker` group (or `sudo`).\n- NVIDIA Container Toolkit installed (`docker run --gpus all` works).\n- ~10–20 GB free disk for the image pull.\n- Network access to `nvcr.io` and `docs.nvidia.com`.\n\n## Limitations\n\n- Container images cover only the tag matrix below — no Conda\u002Fpip env inside.\n- GUI examples require X11 forwarding; this skill runs Holoviz headless to avoid that.\n- Tag suffix must match the host GPU\u002Fdriver (cuda13 \u002F cuda12-dgpu \u002F cuda12-igpu) — wrong suffix → CUDA init failures.\n\n## Instructions\n\n- Container repo: `nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan`.\n- The doc page at https:\u002F\u002Fdocs.nvidia.com\u002Fholoscan\u002Fsdk-user-guide\u002Fsdk_installation.html is canonical — fetch it if anything below disagrees.\n- Work through the steps below in order: pick the tag, verify GPU passthrough and pull, verify with the six examples, then hand off the launch command.\n\n## Step 1: Pick the tag\n\nTag = `\u003Cversion>-\u003Csuffix>`, e.g. `v4.1.0-cuda13`. Get the current SDK version from the doc page above; pick the suffix from `nvidia-smi` (the \"CUDA Version\" field, top-right of the table header):\n\n| `nvidia-smi` CUDA Version | Suffix |\n|---|---|\n| 13.x+ | `cuda13` |\n| 12.x, Ampere\u002FAda dGPU | `cuda12-dgpu` |\n| 12.x, ARM64 iGPU (nvgpu) | `cuda12-igpu` |\n\nThe \"CUDA Forward Compatibility mode ENABLED\" banner is expected — not an error — when the container ships a newer CUDA minor version than the host driver supports. The forward-compat shim lets the container's CUDA runtime work against the older host driver within the same major version.\n\n## Step 2: Verify GPU passthrough, then pull\n\n```bash\ndocker run --rm --gpus all ubuntu:22.04 nvidia-smi 2>&1 | tail -5\n```\n\nIf Docker is missing → install from https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F. If GPU passthrough fails → install the NVIDIA Container Toolkit per https:\u002F\u002Fdocs.nvidia.com\u002Fdatacenter\u002Fcloud-native\u002Fcontainer-toolkit\u002Flatest\u002Finstall-guide.html, then retry.\n\nPull (~10–20 GB — warn the user before starting):\n\n```bash\ndocker pull nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\n```\n\n## Step 3: Verify with six examples\n\nTests cover: bare Python binding (1a), bare C++ runtime (1b, 2a), Python + Holoviz\u002FVulkan (2b, 3a), and C++ + Holoviz\u002FVulkan (3b). Holoviz examples always run headless (inject `headless: true` into the YAML) — this works whether or not a display is attached and avoids GUI failure modes over SSH.\n\n```bash\nIMG=nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\nRUN=(docker run --rm --runtime=nvidia --gpus all --cap-add CAP_SYS_PTRACE --ipc=host --ulimit memlock=-1 --ulimit stack=67108864)\n\n# 1a. hello_world (Python) — expect \"Hello World!\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && python3 \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fpython\u002Fhello_world.py\"\n\n# 1b. hello_world (C++) — expect \"Hello World!\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fcpp\u002Fhello_world\"\n\n# 2a. tensor_interop (C++) — expect tensors doubling each pass, \"Graph execution finished.\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fcpp\u002Ftensor_interop\"\n\n# 2b. tensor_interop (Python, 10 frames) — Holoviz, headless. The YAML has no\n#     headless field by default, so inject one under `holoviz:`. Expect\n#     \"message received (count: 10)\".\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^holoviz:\u002Fholoviz:\\n  headless: true\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\u002Ftensor_interop.yaml > \u002Ftmp\u002Fti.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\n  python3 tensor_interop.py --config \u002Ftmp\u002Fti.yaml\n\"\n\n# 3a. video_replayer (Python, 10 frames) — Holoviz, headless. Inject `headless: true`\n#     under `holoviz:` (above `width: 854`). Same sed works for the C++ YAML in 3b —\n#     both files share the same `holoviz:` section shape.\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^  width: 854\u002F  headless: true\\n  width: 854\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\n  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata python3 video_replayer.py --config \u002Ftmp\u002Fvr.yaml\n\"\n\n# 3b. video_replayer (C++, 10 frames) — same headless injection as 3a. The C++\n#     YAML hard-codes `directory: \"..\u002Fdata\u002Fracerx\"`, but HOLOSCAN_INPUT_PATH\n#     overrides it, so we don't need to patch that field.\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^  width: 854\u002F  headless: true\\n  width: 854\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr_cpp.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\n  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata .\u002Fvideo_replayer --config \u002Ftmp\u002Fvr_cpp.yaml\n\"\n```\n\n## Step 4: Launch command\n\n- Read https:\u002F\u002Fcatalog.ngc.nvidia.com\u002Forgs\u002Fnvidia\u002Fteams\u002Fclara-holoscan\u002Fcontainers\u002Fholoscan.\n- Explain the docker flags below to the user.\n- Refer the user to that link for additional flags (e.g., how to mount V4L2 video devices).\n\n```bash\ndocker run -it --rm \\\n  --runtime=nvidia --gpus all --cap-add CAP_SYS_PTRACE \\\n  --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 \\\n  nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\n# Examples: \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002F\n# Mount files: -v \u002Fhost\u002Fpath:\u002Fcontainer\u002Fpath\n# GUI examples: add -v \u002Ftmp\u002F.X11-unix:\u002Ftmp\u002F.X11-unix -e DISPLAY=$DISPLAY\n```\n\nNext:\n- Explore: `ls \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002F`\n- Walk through one: `\u002Fholoscan-explain-example`\n\n## Troubleshooting\n\n- **`docker: Error response from daemon: could not select device driver \"nvidia\"`.** NVIDIA Container Toolkit is missing or not configured. Install per the link in Step 2 and restart Docker.\n- **CUDA init failure inside the container.** Tag suffix doesn't match the host. Re-check `nvidia-smi` CUDA Version and the table in Step 1.\n- **Segmentation fault when launching an example.** `ulimit -s 32768` wasn't applied inside the container. Use the `bash -c \"ulimit -s 32768 && ...\"` pattern shown in Step 3.\n- **Holoviz example hangs \u002F no window over SSH.** YAML wasn't patched to `headless: true`. Use the `sed` injection shown in Step 3.\n- **`video_replayer` can't find data.** Set `HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata` — overrides the YAML's hard-coded path.\n",{"data":32,"body":43},{"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",[38,39,40,41,42],"holoscan","install","container","docker","ngc",{"type":44,"children":45},"root",[46,55,62,77,83,159,165,183,189,223,229,257,341,346,352,426,447,452,497,503,516,1458,1464,1489,1634,1639,1664,1670,1781],{"type":47,"tag":48,"props":49,"children":51},"element","h1",{"id":50},"holoscan-ngc-container-installation",[52],{"type":53,"value":54},"text","Holoscan NGC Container Installation",{"type":47,"tag":56,"props":57,"children":59},"h2",{"id":58},"purpose",[60],{"type":53,"value":61},"Purpose",{"type":47,"tag":63,"props":64,"children":65},"p",{},[66,68,75],{"type":53,"value":67},"Pull and verify the official Holoscan SDK container from NGC (",{"type":47,"tag":69,"props":70,"children":72},"code",{"className":71},[],[73],{"type":53,"value":74},"nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan",{"type":53,"value":76},"), selecting the right CUDA\u002Farch tag for the host GPU and validating with the bundled Python and C++ examples.",{"type":47,"tag":56,"props":78,"children":80},{"id":79},"prerequisites",[81],{"type":53,"value":82},"Prerequisites",{"type":47,"tag":84,"props":85,"children":86},"ul",{},[87,101,120,133,138],{"type":47,"tag":88,"props":89,"children":90},"li",{},[91,93,99],{"type":53,"value":92},"Linux host with an NVIDIA GPU and a working driver (",{"type":47,"tag":69,"props":94,"children":96},{"className":95},[],[97],{"type":53,"value":98},"nvidia-smi",{"type":53,"value":100},").",{"type":47,"tag":88,"props":102,"children":103},{},[104,106,111,113,119],{"type":53,"value":105},"Docker installed and the user in the ",{"type":47,"tag":69,"props":107,"children":109},{"className":108},[],[110],{"type":53,"value":41},{"type":53,"value":112}," group (or ",{"type":47,"tag":69,"props":114,"children":116},{"className":115},[],[117],{"type":53,"value":118},"sudo",{"type":53,"value":100},{"type":47,"tag":88,"props":121,"children":122},{},[123,125,131],{"type":53,"value":124},"NVIDIA Container Toolkit installed (",{"type":47,"tag":69,"props":126,"children":128},{"className":127},[],[129],{"type":53,"value":130},"docker run --gpus all",{"type":53,"value":132}," works).",{"type":47,"tag":88,"props":134,"children":135},{},[136],{"type":53,"value":137},"~10–20 GB free disk for the image pull.",{"type":47,"tag":88,"props":139,"children":140},{},[141,143,149,151,157],{"type":53,"value":142},"Network access to ",{"type":47,"tag":69,"props":144,"children":146},{"className":145},[],[147],{"type":53,"value":148},"nvcr.io",{"type":53,"value":150}," and ",{"type":47,"tag":69,"props":152,"children":154},{"className":153},[],[155],{"type":53,"value":156},"docs.nvidia.com",{"type":53,"value":158},".",{"type":47,"tag":56,"props":160,"children":162},{"id":161},"limitations",[163],{"type":53,"value":164},"Limitations",{"type":47,"tag":84,"props":166,"children":167},{},[168,173,178],{"type":47,"tag":88,"props":169,"children":170},{},[171],{"type":53,"value":172},"Container images cover only the tag matrix below — no Conda\u002Fpip env inside.",{"type":47,"tag":88,"props":174,"children":175},{},[176],{"type":53,"value":177},"GUI examples require X11 forwarding; this skill runs Holoviz headless to avoid that.",{"type":47,"tag":88,"props":179,"children":180},{},[181],{"type":53,"value":182},"Tag suffix must match the host GPU\u002Fdriver (cuda13 \u002F cuda12-dgpu \u002F cuda12-igpu) — wrong suffix → CUDA init failures.",{"type":47,"tag":56,"props":184,"children":186},{"id":185},"instructions",[187],{"type":53,"value":188},"Instructions",{"type":47,"tag":84,"props":190,"children":191},{},[192,203,218],{"type":47,"tag":88,"props":193,"children":194},{},[195,197,202],{"type":53,"value":196},"Container repo: ",{"type":47,"tag":69,"props":198,"children":200},{"className":199},[],[201],{"type":53,"value":74},{"type":53,"value":158},{"type":47,"tag":88,"props":204,"children":205},{},[206,208,216],{"type":53,"value":207},"The doc page at ",{"type":47,"tag":209,"props":210,"children":214},"a",{"href":211,"rel":212},"https:\u002F\u002Fdocs.nvidia.com\u002Fholoscan\u002Fsdk-user-guide\u002Fsdk_installation.html",[213],"nofollow",[215],{"type":53,"value":211},{"type":53,"value":217}," is canonical — fetch it if anything below disagrees.",{"type":47,"tag":88,"props":219,"children":220},{},[221],{"type":53,"value":222},"Work through the steps below in order: pick the tag, verify GPU passthrough and pull, verify with the six examples, then hand off the launch command.",{"type":47,"tag":56,"props":224,"children":226},{"id":225},"step-1-pick-the-tag",[227],{"type":53,"value":228},"Step 1: Pick the tag",{"type":47,"tag":63,"props":230,"children":231},{},[232,234,240,242,248,250,255],{"type":53,"value":233},"Tag = ",{"type":47,"tag":69,"props":235,"children":237},{"className":236},[],[238],{"type":53,"value":239},"\u003Cversion>-\u003Csuffix>",{"type":53,"value":241},", e.g. ",{"type":47,"tag":69,"props":243,"children":245},{"className":244},[],[246],{"type":53,"value":247},"v4.1.0-cuda13",{"type":53,"value":249},". Get the current SDK version from the doc page above; pick the suffix from ",{"type":47,"tag":69,"props":251,"children":253},{"className":252},[],[254],{"type":53,"value":98},{"type":53,"value":256}," (the \"CUDA Version\" field, top-right of the table header):",{"type":47,"tag":258,"props":259,"children":260},"table",{},[261,285],{"type":47,"tag":262,"props":263,"children":264},"thead",{},[265],{"type":47,"tag":266,"props":267,"children":268},"tr",{},[269,280],{"type":47,"tag":270,"props":271,"children":272},"th",{},[273,278],{"type":47,"tag":69,"props":274,"children":276},{"className":275},[],[277],{"type":53,"value":98},{"type":53,"value":279}," CUDA Version",{"type":47,"tag":270,"props":281,"children":282},{},[283],{"type":53,"value":284},"Suffix",{"type":47,"tag":286,"props":287,"children":288},"tbody",{},[289,307,324],{"type":47,"tag":266,"props":290,"children":291},{},[292,298],{"type":47,"tag":293,"props":294,"children":295},"td",{},[296],{"type":53,"value":297},"13.x+",{"type":47,"tag":293,"props":299,"children":300},{},[301],{"type":47,"tag":69,"props":302,"children":304},{"className":303},[],[305],{"type":53,"value":306},"cuda13",{"type":47,"tag":266,"props":308,"children":309},{},[310,315],{"type":47,"tag":293,"props":311,"children":312},{},[313],{"type":53,"value":314},"12.x, Ampere\u002FAda dGPU",{"type":47,"tag":293,"props":316,"children":317},{},[318],{"type":47,"tag":69,"props":319,"children":321},{"className":320},[],[322],{"type":53,"value":323},"cuda12-dgpu",{"type":47,"tag":266,"props":325,"children":326},{},[327,332],{"type":47,"tag":293,"props":328,"children":329},{},[330],{"type":53,"value":331},"12.x, ARM64 iGPU (nvgpu)",{"type":47,"tag":293,"props":333,"children":334},{},[335],{"type":47,"tag":69,"props":336,"children":338},{"className":337},[],[339],{"type":53,"value":340},"cuda12-igpu",{"type":47,"tag":63,"props":342,"children":343},{},[344],{"type":53,"value":345},"The \"CUDA Forward Compatibility mode ENABLED\" banner is expected — not an error — when the container ships a newer CUDA minor version than the host driver supports. The forward-compat shim lets the container's CUDA runtime work against the older host driver within the same major version.",{"type":47,"tag":56,"props":347,"children":349},{"id":348},"step-2-verify-gpu-passthrough-then-pull",[350],{"type":53,"value":351},"Step 2: Verify GPU passthrough, then pull",{"type":47,"tag":353,"props":354,"children":359},"pre",{"className":355,"code":356,"language":357,"meta":358,"style":358},"language-bash shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","docker run --rm --gpus all ubuntu:22.04 nvidia-smi 2>&1 | tail -5\n","bash","",[360],{"type":47,"tag":69,"props":361,"children":362},{"__ignoreMap":358},[363],{"type":47,"tag":364,"props":365,"children":368},"span",{"class":366,"line":367},"line",1,[369,374,380,385,390,395,400,405,411,416,421],{"type":47,"tag":364,"props":370,"children":372},{"style":371},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[373],{"type":53,"value":41},{"type":47,"tag":364,"props":375,"children":377},{"style":376},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[378],{"type":53,"value":379}," run",{"type":47,"tag":364,"props":381,"children":382},{"style":376},[383],{"type":53,"value":384}," --rm",{"type":47,"tag":364,"props":386,"children":387},{"style":376},[388],{"type":53,"value":389}," --gpus",{"type":47,"tag":364,"props":391,"children":392},{"style":376},[393],{"type":53,"value":394}," all",{"type":47,"tag":364,"props":396,"children":397},{"style":376},[398],{"type":53,"value":399}," ubuntu:22.04",{"type":47,"tag":364,"props":401,"children":402},{"style":376},[403],{"type":53,"value":404}," nvidia-smi",{"type":47,"tag":364,"props":406,"children":408},{"style":407},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[409],{"type":53,"value":410}," 2>&1",{"type":47,"tag":364,"props":412,"children":413},{"style":407},[414],{"type":53,"value":415}," |",{"type":47,"tag":364,"props":417,"children":418},{"style":371},[419],{"type":53,"value":420}," tail",{"type":47,"tag":364,"props":422,"children":423},{"style":376},[424],{"type":53,"value":425}," -5\n",{"type":47,"tag":63,"props":427,"children":428},{},[429,431,437,439,445],{"type":53,"value":430},"If Docker is missing → install from ",{"type":47,"tag":209,"props":432,"children":435},{"href":433,"rel":434},"https:\u002F\u002Fdocs.docker.com\u002Fengine\u002Finstall\u002F",[213],[436],{"type":53,"value":433},{"type":53,"value":438},". If GPU passthrough fails → install the NVIDIA Container Toolkit per ",{"type":47,"tag":209,"props":440,"children":443},{"href":441,"rel":442},"https:\u002F\u002Fdocs.nvidia.com\u002Fdatacenter\u002Fcloud-native\u002Fcontainer-toolkit\u002Flatest\u002Finstall-guide.html",[213],[444],{"type":53,"value":441},{"type":53,"value":446},", then retry.",{"type":47,"tag":63,"props":448,"children":449},{},[450],{"type":53,"value":451},"Pull (~10–20 GB — warn the user before starting):",{"type":47,"tag":353,"props":453,"children":455},{"className":355,"code":454,"language":357,"meta":358,"style":358},"docker pull nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\n",[456],{"type":47,"tag":69,"props":457,"children":458},{"__ignoreMap":358},[459],{"type":47,"tag":364,"props":460,"children":461},{"class":366,"line":367},[462,466,471,476,481,486,492],{"type":47,"tag":364,"props":463,"children":464},{"style":371},[465],{"type":53,"value":41},{"type":47,"tag":364,"props":467,"children":468},{"style":376},[469],{"type":53,"value":470}," pull",{"type":47,"tag":364,"props":472,"children":473},{"style":376},[474],{"type":53,"value":475}," nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:",{"type":47,"tag":364,"props":477,"children":478},{"style":407},[479],{"type":53,"value":480},"\u003C",{"type":47,"tag":364,"props":482,"children":483},{"style":376},[484],{"type":53,"value":485},"TA",{"type":47,"tag":364,"props":487,"children":489},{"style":488},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[490],{"type":53,"value":491},"G",{"type":47,"tag":364,"props":493,"children":494},{"style":407},[495],{"type":53,"value":496},">\n",{"type":47,"tag":56,"props":498,"children":500},{"id":499},"step-3-verify-with-six-examples",[501],{"type":53,"value":502},"Step 3: Verify with six examples",{"type":47,"tag":63,"props":504,"children":505},{},[506,508,514],{"type":53,"value":507},"Tests cover: bare Python binding (1a), bare C++ runtime (1b, 2a), Python + Holoviz\u002FVulkan (2b, 3a), and C++ + Holoviz\u002FVulkan (3b). Holoviz examples always run headless (inject ",{"type":47,"tag":69,"props":509,"children":511},{"className":510},[],[512],{"type":53,"value":513},"headless: true",{"type":53,"value":515}," into the YAML) — this works whether or not a display is attached and avoids GUI failure modes over SSH.",{"type":47,"tag":353,"props":517,"children":519},{"className":355,"code":518,"language":357,"meta":358,"style":358},"IMG=nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\nRUN=(docker run --rm --runtime=nvidia --gpus all --cap-add CAP_SYS_PTRACE --ipc=host --ulimit memlock=-1 --ulimit stack=67108864)\n\n# 1a. hello_world (Python) — expect \"Hello World!\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && python3 \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fpython\u002Fhello_world.py\"\n\n# 1b. hello_world (C++) — expect \"Hello World!\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fcpp\u002Fhello_world\"\n\n# 2a. tensor_interop (C++) — expect tensors doubling each pass, \"Graph execution finished.\"\n\"${RUN[@]}\" \"$IMG\" bash -c \\\n  \"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fcpp\u002Ftensor_interop\"\n\n# 2b. tensor_interop (Python, 10 frames) — Holoviz, headless. The YAML has no\n#     headless field by default, so inject one under `holoviz:`. Expect\n#     \"message received (count: 10)\".\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^holoviz:\u002Fholoviz:\\n  headless: true\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\u002Ftensor_interop.yaml > \u002Ftmp\u002Fti.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\n  python3 tensor_interop.py --config \u002Ftmp\u002Fti.yaml\n\"\n\n# 3a. video_replayer (Python, 10 frames) — Holoviz, headless. Inject `headless: true`\n#     under `holoviz:` (above `width: 854`). Same sed works for the C++ YAML in 3b —\n#     both files share the same `holoviz:` section shape.\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^  width: 854\u002F  headless: true\\n  width: 854\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\n  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata python3 video_replayer.py --config \u002Ftmp\u002Fvr.yaml\n\"\n\n# 3b. video_replayer (C++, 10 frames) — same headless injection as 3a. The C++\n#     YAML hard-codes `directory: \"..\u002Fdata\u002Fracerx\"`, but HOLOSCAN_INPUT_PATH\n#     overrides it, so we don't need to patch that field.\n\"${RUN[@]}\" \"$IMG\" bash -c \"\n  ulimit -s 32768\n  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' \\\n      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' \\\n      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' \\\n      -e 's\u002F^  width: 854\u002F  headless: true\\n  width: 854\u002F' \\\n      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr_cpp.yaml\n  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\n  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata .\u002Fvideo_replayer --config \u002Ftmp\u002Fvr_cpp.yaml\n\"\n",[520],{"type":47,"tag":69,"props":521,"children":522},{"__ignoreMap":358},[523,554,669,679,689,745,764,772,781,829,846,854,863,911,928,936,945,954,963,1012,1021,1035,1048,1061,1074,1083,1092,1101,1109,1117,1126,1135,1144,1192,1200,1212,1224,1236,1249,1258,1267,1276,1284,1292,1301,1310,1319,1367,1375,1387,1399,1411,1423,1432,1441,1450],{"type":47,"tag":364,"props":524,"children":525},{"class":366,"line":367},[526,531,536,541,545,550],{"type":47,"tag":364,"props":527,"children":528},{"style":488},[529],{"type":53,"value":530},"IMG",{"type":47,"tag":364,"props":532,"children":533},{"style":407},[534],{"type":53,"value":535},"=",{"type":47,"tag":364,"props":537,"children":538},{"style":376},[539],{"type":53,"value":540},"nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:",{"type":47,"tag":364,"props":542,"children":543},{"style":407},[544],{"type":53,"value":480},{"type":47,"tag":364,"props":546,"children":547},{"style":376},[548],{"type":53,"value":549},"TAG",{"type":47,"tag":364,"props":551,"children":552},{"style":407},[553],{"type":53,"value":496},{"type":47,"tag":364,"props":555,"children":557},{"class":366,"line":556},2,[558,563,568,572,576,580,586,590,594,598,602,607,612,617,621,626,631,636,640,646,650,655,659,664],{"type":47,"tag":364,"props":559,"children":560},{"style":488},[561],{"type":53,"value":562},"RUN",{"type":47,"tag":364,"props":564,"children":565},{"style":407},[566],{"type":53,"value":567},"=(",{"type":47,"tag":364,"props":569,"children":570},{"style":376},[571],{"type":53,"value":41},{"type":47,"tag":364,"props":573,"children":574},{"style":376},[575],{"type":53,"value":379},{"type":47,"tag":364,"props":577,"children":578},{"style":376},[579],{"type":53,"value":384},{"type":47,"tag":364,"props":581,"children":583},{"style":582},"--shiki-light:#9C3EDA;--shiki-default:#C792EA;--shiki-dark:#C792EA",[584],{"type":53,"value":585}," --runtime",{"type":47,"tag":364,"props":587,"children":588},{"style":407},[589],{"type":53,"value":535},{"type":47,"tag":364,"props":591,"children":592},{"style":376},[593],{"type":53,"value":8},{"type":47,"tag":364,"props":595,"children":596},{"style":376},[597],{"type":53,"value":389},{"type":47,"tag":364,"props":599,"children":600},{"style":376},[601],{"type":53,"value":394},{"type":47,"tag":364,"props":603,"children":604},{"style":376},[605],{"type":53,"value":606}," --cap-add",{"type":47,"tag":364,"props":608,"children":609},{"style":376},[610],{"type":53,"value":611}," CAP_SYS_PTRACE",{"type":47,"tag":364,"props":613,"children":614},{"style":582},[615],{"type":53,"value":616}," --ipc",{"type":47,"tag":364,"props":618,"children":619},{"style":407},[620],{"type":53,"value":535},{"type":47,"tag":364,"props":622,"children":623},{"style":376},[624],{"type":53,"value":625},"host",{"type":47,"tag":364,"props":627,"children":628},{"style":376},[629],{"type":53,"value":630}," --ulimit",{"type":47,"tag":364,"props":632,"children":633},{"style":582},[634],{"type":53,"value":635}," memlock",{"type":47,"tag":364,"props":637,"children":638},{"style":407},[639],{"type":53,"value":535},{"type":47,"tag":364,"props":641,"children":643},{"style":642},"--shiki-light:#F76D47;--shiki-default:#F78C6C;--shiki-dark:#F78C6C",[644],{"type":53,"value":645},"-1",{"type":47,"tag":364,"props":647,"children":648},{"style":376},[649],{"type":53,"value":630},{"type":47,"tag":364,"props":651,"children":652},{"style":582},[653],{"type":53,"value":654}," stack",{"type":47,"tag":364,"props":656,"children":657},{"style":407},[658],{"type":53,"value":535},{"type":47,"tag":364,"props":660,"children":661},{"style":642},[662],{"type":53,"value":663},"67108864",{"type":47,"tag":364,"props":665,"children":666},{"style":407},[667],{"type":53,"value":668},")\n",{"type":47,"tag":364,"props":670,"children":672},{"class":366,"line":671},3,[673],{"type":47,"tag":364,"props":674,"children":676},{"emptyLinePlaceholder":675},true,[677],{"type":53,"value":678},"\n",{"type":47,"tag":364,"props":680,"children":682},{"class":366,"line":681},4,[683],{"type":47,"tag":364,"props":684,"children":686},{"style":685},"--shiki-light:#90A4AE;--shiki-light-font-style:italic;--shiki-default:#546E7A;--shiki-default-font-style:italic;--shiki-dark:#676E95;--shiki-dark-font-style:italic",[687],{"type":53,"value":688},"# 1a. hello_world (Python) — expect \"Hello World!\"\n",{"type":47,"tag":364,"props":690,"children":692},{"class":366,"line":691},5,[693,698,703,707,712,716,721,726,730,735,740],{"type":47,"tag":364,"props":694,"children":695},{"style":371},[696],{"type":53,"value":697},"\"",{"type":47,"tag":364,"props":699,"children":700},{"style":407},[701],{"type":53,"value":702},"${",{"type":47,"tag":364,"props":704,"children":705},{"style":488},[706],{"type":53,"value":562},{"type":47,"tag":364,"props":708,"children":709},{"style":407},[710],{"type":53,"value":711},"[@]}",{"type":47,"tag":364,"props":713,"children":714},{"style":371},[715],{"type":53,"value":697},{"type":47,"tag":364,"props":717,"children":718},{"style":407},[719],{"type":53,"value":720}," \"",{"type":47,"tag":364,"props":722,"children":723},{"style":488},[724],{"type":53,"value":725},"$IMG",{"type":47,"tag":364,"props":727,"children":728},{"style":407},[729],{"type":53,"value":697},{"type":47,"tag":364,"props":731,"children":732},{"style":376},[733],{"type":53,"value":734}," bash",{"type":47,"tag":364,"props":736,"children":737},{"style":376},[738],{"type":53,"value":739}," -c",{"type":47,"tag":364,"props":741,"children":742},{"style":488},[743],{"type":53,"value":744}," \\\n",{"type":47,"tag":364,"props":746,"children":748},{"class":366,"line":747},6,[749,754,759],{"type":47,"tag":364,"props":750,"children":751},{"style":407},[752],{"type":53,"value":753},"  \"",{"type":47,"tag":364,"props":755,"children":756},{"style":376},[757],{"type":53,"value":758},"ulimit -s 32768 && python3 \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fpython\u002Fhello_world.py",{"type":47,"tag":364,"props":760,"children":761},{"style":407},[762],{"type":53,"value":763},"\"\n",{"type":47,"tag":364,"props":765,"children":767},{"class":366,"line":766},7,[768],{"type":47,"tag":364,"props":769,"children":770},{"emptyLinePlaceholder":675},[771],{"type":53,"value":678},{"type":47,"tag":364,"props":773,"children":775},{"class":366,"line":774},8,[776],{"type":47,"tag":364,"props":777,"children":778},{"style":685},[779],{"type":53,"value":780},"# 1b. hello_world (C++) — expect \"Hello World!\"\n",{"type":47,"tag":364,"props":782,"children":784},{"class":366,"line":783},9,[785,789,793,797,801,805,809,813,817,821,825],{"type":47,"tag":364,"props":786,"children":787},{"style":371},[788],{"type":53,"value":697},{"type":47,"tag":364,"props":790,"children":791},{"style":407},[792],{"type":53,"value":702},{"type":47,"tag":364,"props":794,"children":795},{"style":488},[796],{"type":53,"value":562},{"type":47,"tag":364,"props":798,"children":799},{"style":407},[800],{"type":53,"value":711},{"type":47,"tag":364,"props":802,"children":803},{"style":371},[804],{"type":53,"value":697},{"type":47,"tag":364,"props":806,"children":807},{"style":407},[808],{"type":53,"value":720},{"type":47,"tag":364,"props":810,"children":811},{"style":488},[812],{"type":53,"value":725},{"type":47,"tag":364,"props":814,"children":815},{"style":407},[816],{"type":53,"value":697},{"type":47,"tag":364,"props":818,"children":819},{"style":376},[820],{"type":53,"value":734},{"type":47,"tag":364,"props":822,"children":823},{"style":376},[824],{"type":53,"value":739},{"type":47,"tag":364,"props":826,"children":827},{"style":488},[828],{"type":53,"value":744},{"type":47,"tag":364,"props":830,"children":832},{"class":366,"line":831},10,[833,837,842],{"type":47,"tag":364,"props":834,"children":835},{"style":407},[836],{"type":53,"value":753},{"type":47,"tag":364,"props":838,"children":839},{"style":376},[840],{"type":53,"value":841},"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fhello_world\u002Fcpp\u002Fhello_world",{"type":47,"tag":364,"props":843,"children":844},{"style":407},[845],{"type":53,"value":763},{"type":47,"tag":364,"props":847,"children":849},{"class":366,"line":848},11,[850],{"type":47,"tag":364,"props":851,"children":852},{"emptyLinePlaceholder":675},[853],{"type":53,"value":678},{"type":47,"tag":364,"props":855,"children":857},{"class":366,"line":856},12,[858],{"type":47,"tag":364,"props":859,"children":860},{"style":685},[861],{"type":53,"value":862},"# 2a. tensor_interop (C++) — expect tensors doubling each pass, \"Graph execution finished.\"\n",{"type":47,"tag":364,"props":864,"children":866},{"class":366,"line":865},13,[867,871,875,879,883,887,891,895,899,903,907],{"type":47,"tag":364,"props":868,"children":869},{"style":371},[870],{"type":53,"value":697},{"type":47,"tag":364,"props":872,"children":873},{"style":407},[874],{"type":53,"value":702},{"type":47,"tag":364,"props":876,"children":877},{"style":488},[878],{"type":53,"value":562},{"type":47,"tag":364,"props":880,"children":881},{"style":407},[882],{"type":53,"value":711},{"type":47,"tag":364,"props":884,"children":885},{"style":371},[886],{"type":53,"value":697},{"type":47,"tag":364,"props":888,"children":889},{"style":407},[890],{"type":53,"value":720},{"type":47,"tag":364,"props":892,"children":893},{"style":488},[894],{"type":53,"value":725},{"type":47,"tag":364,"props":896,"children":897},{"style":407},[898],{"type":53,"value":697},{"type":47,"tag":364,"props":900,"children":901},{"style":376},[902],{"type":53,"value":734},{"type":47,"tag":364,"props":904,"children":905},{"style":376},[906],{"type":53,"value":739},{"type":47,"tag":364,"props":908,"children":909},{"style":488},[910],{"type":53,"value":744},{"type":47,"tag":364,"props":912,"children":914},{"class":366,"line":913},14,[915,919,924],{"type":47,"tag":364,"props":916,"children":917},{"style":407},[918],{"type":53,"value":753},{"type":47,"tag":364,"props":920,"children":921},{"style":376},[922],{"type":53,"value":923},"ulimit -s 32768 && \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fcpp\u002Ftensor_interop",{"type":47,"tag":364,"props":925,"children":926},{"style":407},[927],{"type":53,"value":763},{"type":47,"tag":364,"props":929,"children":931},{"class":366,"line":930},15,[932],{"type":47,"tag":364,"props":933,"children":934},{"emptyLinePlaceholder":675},[935],{"type":53,"value":678},{"type":47,"tag":364,"props":937,"children":939},{"class":366,"line":938},16,[940],{"type":47,"tag":364,"props":941,"children":942},{"style":685},[943],{"type":53,"value":944},"# 2b. tensor_interop (Python, 10 frames) — Holoviz, headless. The YAML has no\n",{"type":47,"tag":364,"props":946,"children":948},{"class":366,"line":947},17,[949],{"type":47,"tag":364,"props":950,"children":951},{"style":685},[952],{"type":53,"value":953},"#     headless field by default, so inject one under `holoviz:`. Expect\n",{"type":47,"tag":364,"props":955,"children":957},{"class":366,"line":956},18,[958],{"type":47,"tag":364,"props":959,"children":960},{"style":685},[961],{"type":53,"value":962},"#     \"message received (count: 10)\".\n",{"type":47,"tag":364,"props":964,"children":966},{"class":366,"line":965},19,[967,971,975,979,983,987,991,995,999,1003,1007],{"type":47,"tag":364,"props":968,"children":969},{"style":371},[970],{"type":53,"value":697},{"type":47,"tag":364,"props":972,"children":973},{"style":407},[974],{"type":53,"value":702},{"type":47,"tag":364,"props":976,"children":977},{"style":488},[978],{"type":53,"value":562},{"type":47,"tag":364,"props":980,"children":981},{"style":407},[982],{"type":53,"value":711},{"type":47,"tag":364,"props":984,"children":985},{"style":371},[986],{"type":53,"value":697},{"type":47,"tag":364,"props":988,"children":989},{"style":407},[990],{"type":53,"value":720},{"type":47,"tag":364,"props":992,"children":993},{"style":488},[994],{"type":53,"value":725},{"type":47,"tag":364,"props":996,"children":997},{"style":407},[998],{"type":53,"value":697},{"type":47,"tag":364,"props":1000,"children":1001},{"style":376},[1002],{"type":53,"value":734},{"type":47,"tag":364,"props":1004,"children":1005},{"style":376},[1006],{"type":53,"value":739},{"type":47,"tag":364,"props":1008,"children":1009},{"style":407},[1010],{"type":53,"value":1011}," \"\n",{"type":47,"tag":364,"props":1013,"children":1015},{"class":366,"line":1014},20,[1016],{"type":47,"tag":364,"props":1017,"children":1018},{"style":376},[1019],{"type":53,"value":1020},"  ulimit -s 32768\n",{"type":47,"tag":364,"props":1022,"children":1024},{"class":366,"line":1023},21,[1025,1030],{"type":47,"tag":364,"props":1026,"children":1027},{"style":376},[1028],{"type":53,"value":1029},"  sed -e 's\u002Fcount: 0\u002Fcount: 10\u002F' ",{"type":47,"tag":364,"props":1031,"children":1032},{"style":488},[1033],{"type":53,"value":1034},"\\\n",{"type":47,"tag":364,"props":1036,"children":1038},{"class":366,"line":1037},22,[1039,1044],{"type":47,"tag":364,"props":1040,"children":1041},{"style":376},[1042],{"type":53,"value":1043},"      -e 's\u002Frepeat: true\u002Frepeat: false\u002F' ",{"type":47,"tag":364,"props":1045,"children":1046},{"style":488},[1047],{"type":53,"value":1034},{"type":47,"tag":364,"props":1049,"children":1051},{"class":366,"line":1050},23,[1052,1057],{"type":47,"tag":364,"props":1053,"children":1054},{"style":376},[1055],{"type":53,"value":1056},"      -e 's\u002Frealtime: true\u002Frealtime: false\u002F' ",{"type":47,"tag":364,"props":1058,"children":1059},{"style":488},[1060],{"type":53,"value":1034},{"type":47,"tag":364,"props":1062,"children":1064},{"class":366,"line":1063},24,[1065,1070],{"type":47,"tag":364,"props":1066,"children":1067},{"style":376},[1068],{"type":53,"value":1069},"      -e 's\u002F^holoviz:\u002Fholoviz:\\n  headless: true\u002F' ",{"type":47,"tag":364,"props":1071,"children":1072},{"style":488},[1073],{"type":53,"value":1034},{"type":47,"tag":364,"props":1075,"children":1077},{"class":366,"line":1076},25,[1078],{"type":47,"tag":364,"props":1079,"children":1080},{"style":376},[1081],{"type":53,"value":1082},"      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\u002Ftensor_interop.yaml > \u002Ftmp\u002Fti.yaml\n",{"type":47,"tag":364,"props":1084,"children":1086},{"class":366,"line":1085},26,[1087],{"type":47,"tag":364,"props":1088,"children":1089},{"style":376},[1090],{"type":53,"value":1091},"  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Ftensor_interop\u002Fpython\n",{"type":47,"tag":364,"props":1093,"children":1095},{"class":366,"line":1094},27,[1096],{"type":47,"tag":364,"props":1097,"children":1098},{"style":376},[1099],{"type":53,"value":1100},"  python3 tensor_interop.py --config \u002Ftmp\u002Fti.yaml\n",{"type":47,"tag":364,"props":1102,"children":1104},{"class":366,"line":1103},28,[1105],{"type":47,"tag":364,"props":1106,"children":1107},{"style":407},[1108],{"type":53,"value":763},{"type":47,"tag":364,"props":1110,"children":1112},{"class":366,"line":1111},29,[1113],{"type":47,"tag":364,"props":1114,"children":1115},{"emptyLinePlaceholder":675},[1116],{"type":53,"value":678},{"type":47,"tag":364,"props":1118,"children":1120},{"class":366,"line":1119},30,[1121],{"type":47,"tag":364,"props":1122,"children":1123},{"style":685},[1124],{"type":53,"value":1125},"# 3a. video_replayer (Python, 10 frames) — Holoviz, headless. Inject `headless: true`\n",{"type":47,"tag":364,"props":1127,"children":1129},{"class":366,"line":1128},31,[1130],{"type":47,"tag":364,"props":1131,"children":1132},{"style":685},[1133],{"type":53,"value":1134},"#     under `holoviz:` (above `width: 854`). Same sed works for the C++ YAML in 3b —\n",{"type":47,"tag":364,"props":1136,"children":1138},{"class":366,"line":1137},32,[1139],{"type":47,"tag":364,"props":1140,"children":1141},{"style":685},[1142],{"type":53,"value":1143},"#     both files share the same `holoviz:` section shape.\n",{"type":47,"tag":364,"props":1145,"children":1147},{"class":366,"line":1146},33,[1148,1152,1156,1160,1164,1168,1172,1176,1180,1184,1188],{"type":47,"tag":364,"props":1149,"children":1150},{"style":371},[1151],{"type":53,"value":697},{"type":47,"tag":364,"props":1153,"children":1154},{"style":407},[1155],{"type":53,"value":702},{"type":47,"tag":364,"props":1157,"children":1158},{"style":488},[1159],{"type":53,"value":562},{"type":47,"tag":364,"props":1161,"children":1162},{"style":407},[1163],{"type":53,"value":711},{"type":47,"tag":364,"props":1165,"children":1166},{"style":371},[1167],{"type":53,"value":697},{"type":47,"tag":364,"props":1169,"children":1170},{"style":407},[1171],{"type":53,"value":720},{"type":47,"tag":364,"props":1173,"children":1174},{"style":488},[1175],{"type":53,"value":725},{"type":47,"tag":364,"props":1177,"children":1178},{"style":407},[1179],{"type":53,"value":697},{"type":47,"tag":364,"props":1181,"children":1182},{"style":376},[1183],{"type":53,"value":734},{"type":47,"tag":364,"props":1185,"children":1186},{"style":376},[1187],{"type":53,"value":739},{"type":47,"tag":364,"props":1189,"children":1190},{"style":407},[1191],{"type":53,"value":1011},{"type":47,"tag":364,"props":1193,"children":1195},{"class":366,"line":1194},34,[1196],{"type":47,"tag":364,"props":1197,"children":1198},{"style":376},[1199],{"type":53,"value":1020},{"type":47,"tag":364,"props":1201,"children":1203},{"class":366,"line":1202},35,[1204,1208],{"type":47,"tag":364,"props":1205,"children":1206},{"style":376},[1207],{"type":53,"value":1029},{"type":47,"tag":364,"props":1209,"children":1210},{"style":488},[1211],{"type":53,"value":1034},{"type":47,"tag":364,"props":1213,"children":1215},{"class":366,"line":1214},36,[1216,1220],{"type":47,"tag":364,"props":1217,"children":1218},{"style":376},[1219],{"type":53,"value":1043},{"type":47,"tag":364,"props":1221,"children":1222},{"style":488},[1223],{"type":53,"value":1034},{"type":47,"tag":364,"props":1225,"children":1227},{"class":366,"line":1226},37,[1228,1232],{"type":47,"tag":364,"props":1229,"children":1230},{"style":376},[1231],{"type":53,"value":1056},{"type":47,"tag":364,"props":1233,"children":1234},{"style":488},[1235],{"type":53,"value":1034},{"type":47,"tag":364,"props":1237,"children":1239},{"class":366,"line":1238},38,[1240,1245],{"type":47,"tag":364,"props":1241,"children":1242},{"style":376},[1243],{"type":53,"value":1244},"      -e 's\u002F^  width: 854\u002F  headless: true\\n  width: 854\u002F' ",{"type":47,"tag":364,"props":1246,"children":1247},{"style":488},[1248],{"type":53,"value":1034},{"type":47,"tag":364,"props":1250,"children":1252},{"class":366,"line":1251},39,[1253],{"type":47,"tag":364,"props":1254,"children":1255},{"style":376},[1256],{"type":53,"value":1257},"      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr.yaml\n",{"type":47,"tag":364,"props":1259,"children":1261},{"class":366,"line":1260},40,[1262],{"type":47,"tag":364,"props":1263,"children":1264},{"style":376},[1265],{"type":53,"value":1266},"  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fpython\n",{"type":47,"tag":364,"props":1268,"children":1270},{"class":366,"line":1269},41,[1271],{"type":47,"tag":364,"props":1272,"children":1273},{"style":376},[1274],{"type":53,"value":1275},"  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata python3 video_replayer.py --config \u002Ftmp\u002Fvr.yaml\n",{"type":47,"tag":364,"props":1277,"children":1279},{"class":366,"line":1278},42,[1280],{"type":47,"tag":364,"props":1281,"children":1282},{"style":407},[1283],{"type":53,"value":763},{"type":47,"tag":364,"props":1285,"children":1287},{"class":366,"line":1286},43,[1288],{"type":47,"tag":364,"props":1289,"children":1290},{"emptyLinePlaceholder":675},[1291],{"type":53,"value":678},{"type":47,"tag":364,"props":1293,"children":1295},{"class":366,"line":1294},44,[1296],{"type":47,"tag":364,"props":1297,"children":1298},{"style":685},[1299],{"type":53,"value":1300},"# 3b. video_replayer (C++, 10 frames) — same headless injection as 3a. The C++\n",{"type":47,"tag":364,"props":1302,"children":1304},{"class":366,"line":1303},45,[1305],{"type":47,"tag":364,"props":1306,"children":1307},{"style":685},[1308],{"type":53,"value":1309},"#     YAML hard-codes `directory: \"..\u002Fdata\u002Fracerx\"`, but HOLOSCAN_INPUT_PATH\n",{"type":47,"tag":364,"props":1311,"children":1313},{"class":366,"line":1312},46,[1314],{"type":47,"tag":364,"props":1315,"children":1316},{"style":685},[1317],{"type":53,"value":1318},"#     overrides it, so we don't need to patch that field.\n",{"type":47,"tag":364,"props":1320,"children":1322},{"class":366,"line":1321},47,[1323,1327,1331,1335,1339,1343,1347,1351,1355,1359,1363],{"type":47,"tag":364,"props":1324,"children":1325},{"style":371},[1326],{"type":53,"value":697},{"type":47,"tag":364,"props":1328,"children":1329},{"style":407},[1330],{"type":53,"value":702},{"type":47,"tag":364,"props":1332,"children":1333},{"style":488},[1334],{"type":53,"value":562},{"type":47,"tag":364,"props":1336,"children":1337},{"style":407},[1338],{"type":53,"value":711},{"type":47,"tag":364,"props":1340,"children":1341},{"style":371},[1342],{"type":53,"value":697},{"type":47,"tag":364,"props":1344,"children":1345},{"style":407},[1346],{"type":53,"value":720},{"type":47,"tag":364,"props":1348,"children":1349},{"style":488},[1350],{"type":53,"value":725},{"type":47,"tag":364,"props":1352,"children":1353},{"style":407},[1354],{"type":53,"value":697},{"type":47,"tag":364,"props":1356,"children":1357},{"style":376},[1358],{"type":53,"value":734},{"type":47,"tag":364,"props":1360,"children":1361},{"style":376},[1362],{"type":53,"value":739},{"type":47,"tag":364,"props":1364,"children":1365},{"style":407},[1366],{"type":53,"value":1011},{"type":47,"tag":364,"props":1368,"children":1370},{"class":366,"line":1369},48,[1371],{"type":47,"tag":364,"props":1372,"children":1373},{"style":376},[1374],{"type":53,"value":1020},{"type":47,"tag":364,"props":1376,"children":1378},{"class":366,"line":1377},49,[1379,1383],{"type":47,"tag":364,"props":1380,"children":1381},{"style":376},[1382],{"type":53,"value":1029},{"type":47,"tag":364,"props":1384,"children":1385},{"style":488},[1386],{"type":53,"value":1034},{"type":47,"tag":364,"props":1388,"children":1390},{"class":366,"line":1389},50,[1391,1395],{"type":47,"tag":364,"props":1392,"children":1393},{"style":376},[1394],{"type":53,"value":1043},{"type":47,"tag":364,"props":1396,"children":1397},{"style":488},[1398],{"type":53,"value":1034},{"type":47,"tag":364,"props":1400,"children":1402},{"class":366,"line":1401},51,[1403,1407],{"type":47,"tag":364,"props":1404,"children":1405},{"style":376},[1406],{"type":53,"value":1056},{"type":47,"tag":364,"props":1408,"children":1409},{"style":488},[1410],{"type":53,"value":1034},{"type":47,"tag":364,"props":1412,"children":1414},{"class":366,"line":1413},52,[1415,1419],{"type":47,"tag":364,"props":1416,"children":1417},{"style":376},[1418],{"type":53,"value":1244},{"type":47,"tag":364,"props":1420,"children":1421},{"style":488},[1422],{"type":53,"value":1034},{"type":47,"tag":364,"props":1424,"children":1426},{"class":366,"line":1425},53,[1427],{"type":47,"tag":364,"props":1428,"children":1429},{"style":376},[1430],{"type":53,"value":1431},"      \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\u002Fvideo_replayer.yaml > \u002Ftmp\u002Fvr_cpp.yaml\n",{"type":47,"tag":364,"props":1433,"children":1435},{"class":366,"line":1434},54,[1436],{"type":47,"tag":364,"props":1437,"children":1438},{"style":376},[1439],{"type":53,"value":1440},"  cd \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002Fvideo_replayer\u002Fcpp\n",{"type":47,"tag":364,"props":1442,"children":1444},{"class":366,"line":1443},55,[1445],{"type":47,"tag":364,"props":1446,"children":1447},{"style":376},[1448],{"type":53,"value":1449},"  HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata .\u002Fvideo_replayer --config \u002Ftmp\u002Fvr_cpp.yaml\n",{"type":47,"tag":364,"props":1451,"children":1453},{"class":366,"line":1452},56,[1454],{"type":47,"tag":364,"props":1455,"children":1456},{"style":407},[1457],{"type":53,"value":763},{"type":47,"tag":56,"props":1459,"children":1461},{"id":1460},"step-4-launch-command",[1462],{"type":53,"value":1463},"Step 4: Launch command",{"type":47,"tag":84,"props":1465,"children":1466},{},[1467,1479,1484],{"type":47,"tag":88,"props":1468,"children":1469},{},[1470,1472,1478],{"type":53,"value":1471},"Read ",{"type":47,"tag":209,"props":1473,"children":1476},{"href":1474,"rel":1475},"https:\u002F\u002Fcatalog.ngc.nvidia.com\u002Forgs\u002Fnvidia\u002Fteams\u002Fclara-holoscan\u002Fcontainers\u002Fholoscan",[213],[1477],{"type":53,"value":1474},{"type":53,"value":158},{"type":47,"tag":88,"props":1480,"children":1481},{},[1482],{"type":53,"value":1483},"Explain the docker flags below to the user.",{"type":47,"tag":88,"props":1485,"children":1486},{},[1487],{"type":53,"value":1488},"Refer the user to that link for additional flags (e.g., how to mount V4L2 video devices).",{"type":47,"tag":353,"props":1490,"children":1492},{"className":355,"code":1491,"language":357,"meta":358,"style":358},"docker run -it --rm \\\n  --runtime=nvidia --gpus all --cap-add CAP_SYS_PTRACE \\\n  --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 \\\n  nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:\u003CTAG>\n# Examples: \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002F\n# Mount files: -v \u002Fhost\u002Fpath:\u002Fcontainer\u002Fpath\n# GUI examples: add -v \u002Ftmp\u002F.X11-unix:\u002Ftmp\u002F.X11-unix -e DISPLAY=$DISPLAY\n",[1493],{"type":47,"tag":69,"props":1494,"children":1495},{"__ignoreMap":358},[1496,1520,1548,1586,1610,1618,1626],{"type":47,"tag":364,"props":1497,"children":1498},{"class":366,"line":367},[1499,1503,1507,1512,1516],{"type":47,"tag":364,"props":1500,"children":1501},{"style":371},[1502],{"type":53,"value":41},{"type":47,"tag":364,"props":1504,"children":1505},{"style":376},[1506],{"type":53,"value":379},{"type":47,"tag":364,"props":1508,"children":1509},{"style":376},[1510],{"type":53,"value":1511}," -it",{"type":47,"tag":364,"props":1513,"children":1514},{"style":376},[1515],{"type":53,"value":384},{"type":47,"tag":364,"props":1517,"children":1518},{"style":488},[1519],{"type":53,"value":744},{"type":47,"tag":364,"props":1521,"children":1522},{"class":366,"line":556},[1523,1528,1532,1536,1540,1544],{"type":47,"tag":364,"props":1524,"children":1525},{"style":376},[1526],{"type":53,"value":1527},"  --runtime=nvidia",{"type":47,"tag":364,"props":1529,"children":1530},{"style":376},[1531],{"type":53,"value":389},{"type":47,"tag":364,"props":1533,"children":1534},{"style":376},[1535],{"type":53,"value":394},{"type":47,"tag":364,"props":1537,"children":1538},{"style":376},[1539],{"type":53,"value":606},{"type":47,"tag":364,"props":1541,"children":1542},{"style":376},[1543],{"type":53,"value":611},{"type":47,"tag":364,"props":1545,"children":1546},{"style":488},[1547],{"type":53,"value":744},{"type":47,"tag":364,"props":1549,"children":1550},{"class":366,"line":671},[1551,1556,1560,1565,1569,1573,1578,1582],{"type":47,"tag":364,"props":1552,"children":1553},{"style":376},[1554],{"type":53,"value":1555},"  --ipc=host",{"type":47,"tag":364,"props":1557,"children":1558},{"style":376},[1559],{"type":53,"value":630},{"type":47,"tag":364,"props":1561,"children":1562},{"style":376},[1563],{"type":53,"value":1564}," memlock=",{"type":47,"tag":364,"props":1566,"children":1567},{"style":642},[1568],{"type":53,"value":645},{"type":47,"tag":364,"props":1570,"children":1571},{"style":376},[1572],{"type":53,"value":630},{"type":47,"tag":364,"props":1574,"children":1575},{"style":376},[1576],{"type":53,"value":1577}," stack=",{"type":47,"tag":364,"props":1579,"children":1580},{"style":642},[1581],{"type":53,"value":663},{"type":47,"tag":364,"props":1583,"children":1584},{"style":488},[1585],{"type":53,"value":744},{"type":47,"tag":364,"props":1587,"children":1588},{"class":366,"line":681},[1589,1594,1598,1602,1606],{"type":47,"tag":364,"props":1590,"children":1591},{"style":376},[1592],{"type":53,"value":1593},"  nvcr.io\u002Fnvidia\u002Fclara-holoscan\u002Fholoscan:",{"type":47,"tag":364,"props":1595,"children":1596},{"style":407},[1597],{"type":53,"value":480},{"type":47,"tag":364,"props":1599,"children":1600},{"style":376},[1601],{"type":53,"value":485},{"type":47,"tag":364,"props":1603,"children":1604},{"style":488},[1605],{"type":53,"value":491},{"type":47,"tag":364,"props":1607,"children":1608},{"style":407},[1609],{"type":53,"value":496},{"type":47,"tag":364,"props":1611,"children":1612},{"class":366,"line":691},[1613],{"type":47,"tag":364,"props":1614,"children":1615},{"style":685},[1616],{"type":53,"value":1617},"# Examples: \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002F\n",{"type":47,"tag":364,"props":1619,"children":1620},{"class":366,"line":747},[1621],{"type":47,"tag":364,"props":1622,"children":1623},{"style":685},[1624],{"type":53,"value":1625},"# Mount files: -v \u002Fhost\u002Fpath:\u002Fcontainer\u002Fpath\n",{"type":47,"tag":364,"props":1627,"children":1628},{"class":366,"line":766},[1629],{"type":47,"tag":364,"props":1630,"children":1631},{"style":685},[1632],{"type":53,"value":1633},"# GUI examples: add -v \u002Ftmp\u002F.X11-unix:\u002Ftmp\u002F.X11-unix -e DISPLAY=$DISPLAY\n",{"type":47,"tag":63,"props":1635,"children":1636},{},[1637],{"type":53,"value":1638},"Next:",{"type":47,"tag":84,"props":1640,"children":1641},{},[1642,1653],{"type":47,"tag":88,"props":1643,"children":1644},{},[1645,1647],{"type":53,"value":1646},"Explore: ",{"type":47,"tag":69,"props":1648,"children":1650},{"className":1649},[],[1651],{"type":53,"value":1652},"ls \u002Fopt\u002Fnvidia\u002Fholoscan\u002Fexamples\u002F",{"type":47,"tag":88,"props":1654,"children":1655},{},[1656,1658],{"type":53,"value":1657},"Walk through one: ",{"type":47,"tag":69,"props":1659,"children":1661},{"className":1660},[],[1662],{"type":53,"value":1663},"\u002Fholoscan-explain-example",{"type":47,"tag":56,"props":1665,"children":1667},{"id":1666},"troubleshooting",[1668],{"type":53,"value":1669},"Troubleshooting",{"type":47,"tag":84,"props":1671,"children":1672},{},[1673,1689,1706,1732,1757],{"type":47,"tag":88,"props":1674,"children":1675},{},[1676,1687],{"type":47,"tag":1677,"props":1678,"children":1679},"strong",{},[1680,1686],{"type":47,"tag":69,"props":1681,"children":1683},{"className":1682},[],[1684],{"type":53,"value":1685},"docker: Error response from daemon: could not select device driver \"nvidia\"",{"type":53,"value":158},{"type":53,"value":1688}," NVIDIA Container Toolkit is missing or not configured. Install per the link in Step 2 and restart Docker.",{"type":47,"tag":88,"props":1690,"children":1691},{},[1692,1697,1699,1704],{"type":47,"tag":1677,"props":1693,"children":1694},{},[1695],{"type":53,"value":1696},"CUDA init failure inside the container.",{"type":53,"value":1698}," Tag suffix doesn't match the host. Re-check ",{"type":47,"tag":69,"props":1700,"children":1702},{"className":1701},[],[1703],{"type":53,"value":98},{"type":53,"value":1705}," CUDA Version and the table in Step 1.",{"type":47,"tag":88,"props":1707,"children":1708},{},[1709,1714,1716,1722,1724,1730],{"type":47,"tag":1677,"props":1710,"children":1711},{},[1712],{"type":53,"value":1713},"Segmentation fault when launching an example.",{"type":53,"value":1715}," ",{"type":47,"tag":69,"props":1717,"children":1719},{"className":1718},[],[1720],{"type":53,"value":1721},"ulimit -s 32768",{"type":53,"value":1723}," wasn't applied inside the container. Use the ",{"type":47,"tag":69,"props":1725,"children":1727},{"className":1726},[],[1728],{"type":53,"value":1729},"bash -c \"ulimit -s 32768 && ...\"",{"type":53,"value":1731}," pattern shown in Step 3.",{"type":47,"tag":88,"props":1733,"children":1734},{},[1735,1740,1742,1747,1749,1755],{"type":47,"tag":1677,"props":1736,"children":1737},{},[1738],{"type":53,"value":1739},"Holoviz example hangs \u002F no window over SSH.",{"type":53,"value":1741}," YAML wasn't patched to ",{"type":47,"tag":69,"props":1743,"children":1745},{"className":1744},[],[1746],{"type":53,"value":513},{"type":53,"value":1748},". Use the ",{"type":47,"tag":69,"props":1750,"children":1752},{"className":1751},[],[1753],{"type":53,"value":1754},"sed",{"type":53,"value":1756}," injection shown in Step 3.",{"type":47,"tag":88,"props":1758,"children":1759},{},[1760,1771,1773,1779],{"type":47,"tag":1677,"props":1761,"children":1762},{},[1763,1769],{"type":47,"tag":69,"props":1764,"children":1766},{"className":1765},[],[1767],{"type":53,"value":1768},"video_replayer",{"type":53,"value":1770}," can't find data.",{"type":53,"value":1772}," Set ",{"type":47,"tag":69,"props":1774,"children":1776},{"className":1775},[],[1777],{"type":53,"value":1778},"HOLOSCAN_INPUT_PATH=\u002Fopt\u002Fnvidia\u002Fholoscan\u002Fdata",{"type":53,"value":1780}," — overrides the YAML's hard-coded path.",{"type":47,"tag":1782,"props":1783,"children":1784},"style",{},[1785],{"type":53,"value":1786},"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":1788,"total":1889},[1789,1806,1818,1832,1844,1861,1875],{"slug":1790,"name":1790,"fn":1791,"description":1792,"org":1793,"tags":1794,"stars":20,"repoUrl":21,"updatedAt":1805},"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},[1795,1798,1801,1802],{"name":1796,"slug":1797,"type":15},"Data Analysis","data-analysis",{"name":1799,"slug":1800,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1803,"slug":1804,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1807,"name":1807,"fn":1808,"description":1809,"org":1810,"tags":1811,"stars":20,"repoUrl":21,"updatedAt":1817},"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},[1812,1813,1816],{"name":17,"slug":18,"type":15},{"name":1814,"slug":1815,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1819,"name":1819,"fn":1820,"description":1821,"org":1822,"tags":1823,"stars":20,"repoUrl":21,"updatedAt":1831},"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},[1824,1827,1828],{"name":1825,"slug":1826,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1829,"slug":1830,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1833,"name":1833,"fn":1834,"description":1835,"org":1836,"tags":1837,"stars":20,"repoUrl":21,"updatedAt":1843},"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},[1838,1839,1840],{"name":1796,"slug":1797,"type":15},{"name":9,"slug":8,"type":15},{"name":1841,"slug":1842,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1845,"name":1845,"fn":1846,"description":1847,"org":1848,"tags":1849,"stars":20,"repoUrl":21,"updatedAt":1860},"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},[1850,1853,1856,1857],{"name":1851,"slug":1852,"type":15},"Automation","automation",{"name":1854,"slug":1855,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1858,"slug":1859,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1862,"name":1862,"fn":1863,"description":1864,"org":1865,"tags":1866,"stars":20,"repoUrl":21,"updatedAt":1874},"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},[1867,1868,1870,1871],{"name":17,"slug":18,"type":15},{"name":1869,"slug":41,"type":15},"Docker",{"name":9,"slug":8,"type":15},{"name":1872,"slug":1873,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1876,"name":1876,"fn":1877,"description":1878,"org":1879,"tags":1880,"stars":20,"repoUrl":21,"updatedAt":1888},"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},[1881,1882,1885],{"name":9,"slug":8,"type":15},{"name":1883,"slug":1884,"type":15},"Quantum Computing","quantum-computing",{"name":1886,"slug":1887,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":1891,"total":2040},[1892,1910,1924,1935,1947,1961,1974,1988,1999,2008,2022,2031],{"slug":1893,"name":1893,"fn":1894,"description":1895,"org":1896,"tags":1897,"stars":1907,"repoUrl":1908,"updatedAt":1909},"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},[1898,1901,1904],{"name":1899,"slug":1900,"type":15},"Documentation","documentation",{"name":1902,"slug":1903,"type":15},"MCP","mcp",{"name":1905,"slug":1906,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1911,"name":1911,"fn":1912,"description":1913,"org":1914,"tags":1915,"stars":1921,"repoUrl":1922,"updatedAt":1923},"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},[1916,1917,1918],{"name":13,"slug":14,"type":15},{"name":17,"slug":18,"type":15},{"name":1919,"slug":1920,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1925,"name":1925,"fn":1926,"description":1927,"org":1928,"tags":1929,"stars":1921,"repoUrl":1922,"updatedAt":1934},"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},[1930,1933],{"name":1931,"slug":1932,"type":15},"CI\u002FCD","ci-cd",{"name":17,"slug":18,"type":15},"2026-07-14T05:25:59.97109",{"slug":1936,"name":1936,"fn":1937,"description":1938,"org":1939,"tags":1940,"stars":1921,"repoUrl":1922,"updatedAt":1946},"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},[1941,1942,1943],{"name":1931,"slug":1932,"type":15},{"name":17,"slug":18,"type":15},{"name":1944,"slug":1945,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1948,"name":1948,"fn":1949,"description":1950,"org":1951,"tags":1952,"stars":1921,"repoUrl":1922,"updatedAt":1960},"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},[1953,1956,1957],{"name":1954,"slug":1955,"type":15},"Debugging","debugging",{"name":1944,"slug":1945,"type":15},{"name":1958,"slug":1959,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1962,"name":1962,"fn":1963,"description":1964,"org":1965,"tags":1966,"stars":1921,"repoUrl":1922,"updatedAt":1973},"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},[1967,1970],{"name":1968,"slug":1969,"type":15},"Best Practices","best-practices",{"name":1971,"slug":1972,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1975,"name":1975,"fn":1976,"description":1977,"org":1978,"tags":1979,"stars":1921,"repoUrl":1922,"updatedAt":1987},"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},[1980,1983,1986],{"name":1981,"slug":1982,"type":15},"Machine Learning","machine-learning",{"name":1984,"slug":1985,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1989,"name":1989,"fn":1990,"description":1991,"org":1992,"tags":1993,"stars":1921,"repoUrl":1922,"updatedAt":1998},"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},[1994,1997],{"name":1995,"slug":1996,"type":15},"QA","qa",{"name":1841,"slug":1842,"type":15},"2026-07-14T05:25:53.673039",{"slug":2000,"name":2000,"fn":2001,"description":2002,"org":2003,"tags":2004,"stars":1921,"repoUrl":1922,"updatedAt":2007},"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},[2005,2006],{"name":17,"slug":18,"type":15},{"name":1814,"slug":1815,"type":15},"2026-07-14T05:25:49.362534",{"slug":2009,"name":2009,"fn":2010,"description":2011,"org":2012,"tags":2013,"stars":1921,"repoUrl":1922,"updatedAt":2021},"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},[2014,2017,2018],{"name":2015,"slug":2016,"type":15},"Code Review","code-review",{"name":1944,"slug":1945,"type":15},{"name":2019,"slug":2020,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":2023,"name":2023,"fn":2024,"description":2025,"org":2026,"tags":2027,"stars":1921,"repoUrl":1922,"updatedAt":2030},"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},[2028,2029],{"name":1995,"slug":1996,"type":15},{"name":1841,"slug":1842,"type":15},"2026-07-14T05:25:54.928983",{"slug":2032,"name":2032,"fn":2033,"description":2034,"org":2035,"tags":2036,"stars":1921,"repoUrl":1922,"updatedAt":2039},"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},[2037,2038],{"name":1851,"slug":1852,"type":15},{"name":1931,"slug":1932,"type":15},"2026-07-30T05:29:03.275638",496]