[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-nre":3,"mdc--wh977a-key":34,"related-repo-nvidia-nre":2916,"related-org-nvidia-nre":2988},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":29,"sourceUrl":32,"mdContent":33},"nre","train 3D Gaussian Splat models with NRE","Use to drive NVIDIA Omniverse NuRec \u002F Neural Reconstruction Engine (NRE) via the public NGC containers nvcr.io\u002Fnvidia\u002Fnre\u002Fnre and nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools (NGC_API_KEY required) — train 3DGUT Gaussian reconstructions from NCore clips, generate aux data, render frames or LiDAR sweeps (local or warm `serve-grpc`), export PLY\u002Fdepth\u002Fmesh\u002FUSDZ, edit actors, and evaluate metrics. Do NOT use for per-object asset capture (use `asset-harvester`) or sensor-to-NCore conversion (use `ncore`).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20],{"name":13,"slug":14,"type":15},"Machine Learning","machine-learning","tag",{"name":17,"slug":18,"type":15},"3D","3d",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Simulation","simulation",16,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills","2026-07-14T05:32:34.548655","CC-BY-4.0 AND Apache-2.0",4,[],{"repoUrl":24,"stars":23,"forks":27,"topics":30,"description":31},[],"Agent skills for Neural Reconstruction Engine","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\u002Ftree\u002FHEAD\u002Fskills\u002Fnre","---\nname: nre\ndescription: >-\n Use to drive NVIDIA Omniverse NuRec \u002F Neural Reconstruction\n Engine (NRE) via the public NGC containers nvcr.io\u002Fnvidia\u002Fnre\u002Fnre\n and nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools (NGC_API_KEY required) — train\n 3DGUT Gaussian reconstructions from NCore clips, generate aux\n data, render frames or LiDAR sweeps (local or warm `serve-grpc`),\n export PLY\u002Fdepth\u002Fmesh\u002FUSDZ, edit actors, and evaluate metrics. Do\n NOT use for per-object asset capture (use `asset-harvester`) or\n sensor-to-NCore conversion (use `ncore`).\nversion: \"0.2.2\"\ntools:\n - Shell\n - Read\n - Write\nlicense: CC-BY-4.0 AND Apache-2.0\ncompatibility: >-\n Linux x86_64, 1+ NVIDIA GPU (Ampere A100\u002FA10\u002FA40\u002FRTX A6000, Ada\n L20\u002FL40\u002FL40S, Hopper H100\u002FH20, or Blackwell RTX Pro 6000D) with\n CUDA 12.8 and >= 24 GB VRAM (48+ GB recommended); driver R570+\n recommended (R580+ on Blackwell, R535+ minimum for Fixer-only).\n Docker >= 23.0.1 + NVIDIA Container Toolkit >= 1.13.5, NGC\n account with NGC_API_KEY exported.\ndependencies:\n - bash\n - docker\n - python3\nmetadata:\n author: NVIDIA NRS \u003Cnurec-skills@nvidia.com>\n tags:\n - nurec\n - autonomous-vehicles\n - neural-reconstruction\n - rendering\n - container\n product_page: https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fomniverse\u002Fnurec\u002F\n ngc_container: nvcr.io\u002Fnvidia\u002Fnre\u002Fnre:latest\n ngc_tools_container: nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools:latest\n ngc_fixer_model: https:\u002F\u002Fcatalog.ngc.nvidia.com\u002Forgs\u002Fnvidia\u002Fteams\u002Fnre\u002Fmodels\u002Fnurec-fixer\n hf_datasets: https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fnvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec\n hf_fixer_model: https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FDifix3D\n release_date: \"2026-04-30\"\n time-estimate: \"2h\"\n---\n\n# NRE — NVIDIA Omniverse NuRec (Neural Reconstruction Engine)\n\n## Purpose\n\nDrive the public NVIDIA Omniverse NuRec \u002F Neural Reconstruction\nEngine containers (`nvcr.io\u002Fnvidia\u002Fnre\u002Fnre`,\n`nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools`) to train a 3DGUT\u002F3DGRT Gaussian\nreconstruction from an NCore V4 camera+LiDAR clip, render novel\nviews (locally or via gRPC), generate aux data, export\nPLY\u002Fdepth\u002Fmesh\u002Fego-mask\u002Ftracks, package Asset Harvester output\ninto a USDZ, and evaluate rendering metrics.\n\nThis skill carries the host-side toolkit around the NRE CLI: NGC\ncredential resolution, cached-image notes, local render recipes,\nMP4 encoding, warm `serve-grpc` boot\u002Fteardown scripts, a thin\nPython gRPC client for repeated RGB renders, bundled rig JSONs,\npre-baked custom-rig trajectories, and bash \u002F Hydra \u002F OSMO\nworkflow templates.\n\n## When to Use \u002F When NOT to Use\n\n**Use this skill when** the user has an NCore V4 clip (or a USDZ +\nNRE artifact pair) on a Linux x86_64 host with an NVIDIA GPU and\nan NGC API key, and wants to train, render, generate aux data,\nexport artifacts, insert\u002Fremove actors, run the gRPC server, or\nevaluate metrics. Concrete triggers:\n\n- Train a multi-camera + LiDAR AV clip into a renderable USDZ\n scene with 3DGUT (or 3DGRT ray-traced) Gaussians.\n- Generate NuRec auxiliary data (seg, depth, ego mask, DINOv2,\n LiDAR-seg visibility) using `nre-tools`.\n- Render frames locally (no server) along the training rig or a\n custom rig + offsets.\n- Render novel views via the sensorsim gRPC API (CARLA, Isaac\n Sim, AlpaSim, custom simulator), optionally with Difix\n artifact-removal.\n- Render LiDAR sweeps via `render-grpc --lidar`.\n- Export PLY \u002F ego masks \u002F depth \u002F Poisson mesh \u002F ground mesh \u002F\n point clouds \u002F cuboid tracks \u002F NCore tracks \u002F custom rig\n trajectories.\n- Insert \u002F remove \u002F replace 3D actors with `export-external-assets`\n + `render-grpc --edit-assets`.\n- Render the gated HF dataset\n `nvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec`.\n- Upgrade an old USDZ once (`upgrade-artifact`).\n- Inspect \u002F evaluate (`export-parsed-config`, `gaussian-statistics`,\n `eval-rendering-metrics`, `compute-metrics`,\n `eval-ground-mesh`).\n- Browse a USDZ or PLY in the in-container viewer.\n\n**Do NOT use this skill when:**\n\n- The user still needs to convert raw sensor data into NCore V4\n (use the `ncore` skill first; NRE consumes NCore-formatted\n shards).\n- The user wants per-object 3D asset extraction from sparse views\n (use `asset-harvester`; NRE only *consumes* AH outputs via\n `export-external-assets`).\n- The user only needs to clean up already-rendered frames using\n the standalone Cosmos-based Fixer (use `nurec-fixer`). NRE's\n inline `--enable-difix` flag is still on this skill's surface,\n but the standalone harmonizer pipeline is owned by\n `nurec-fixer`.\n\n## Inputs\n\n- **dataset_dir** — host directory holding the NCore shards\n (`\u003CNAME>.zarr.itar`, `\u003CNAME>.json`, and any pre-generated\n `\u003CNAME>.aux.*.zarr` auxiliary shards). Required.\n- **dataset_name** — basename of the NCore dataset (the part\n before `.zarr.itar`). Required.\n- **output_dir** — host directory NRE will fill with checkpoints,\n parsed config, metrics, videos, and USDZ artifacts. Required.\n- **camera_ids \u002F lidar_ids** — sensor IDs from the NCore JSON to\n include. Default: all sensors per recipe.\n- **config_name** — Hydra config path resolved inside the\n container. Pick by source dataset:\n - **Waymo Open Dataset** →\n `configs\u002Fapps\u002FAV\u002FWaymo\u002F3dgut_dynamic.yaml` (and its\n `_mcmc` \u002F `_road_semantic` \u002F `_static` siblings). These are\n Waymo-only — they bake in the Waymo sensor rig and conventions.\n - **NVIDIA Physical AI Autonomous Vehicles (PAI)** →\n `\u002Fapps\u002Fprod\u002FHyperion-8.1\u002Fcar2sim_6cam.yaml` (the Hyperion-8.1\n car2sim 6-camera recipe used by the Maglev PAI pipeline).\n Typically referenced via the small overlay shipped at\n [`references\u002Fconfigs\u002Fpai.yaml`](references\u002Fconfigs\u002Fpai.yaml),\n which extends `car2sim_6cam.yaml` with PAI's\n `lidar_top_360fov` ID, six-camera validation set, and lidar\n `intensity` supervision; mount it as\n `{nre_config_dir}\u002Fexternal_overrides.yaml` and pass\n `--config-name=external_overrides`.\n - **PandaSet \u002F NVIDIA AV (NV) \u002F Tesla \u002F Alpasim** → see the\n matching `configs\u002Fapps\u002FAV\u002F{PandaSet,NV,Tesla}\u002F…` or\n `configs\u002Fapps\u002FAlpasim\u002F…` recipes in\n `references\u002Fconfiguration.md`.\n - Do **not** use the Waymo recipes for PAI clips — the\n sensor rig, validation cameras, and lidar IDs differ.\n- **mode** — `train`, `val`, or `trainval`. Default: `trainval`.\n- **NGC_API_KEY** — required env var. Generate at\n \u003Chttps:\u002F\u002Forg.ngc.nvidia.com\u002Fsetup\u002Fapi-keys>.\n\n## Instructions\n\n1. **Validate prerequisites.** Have the agent execute\n `scripts\u002Fvalidate_setup.py` via its standard script runner\n (`run_script(\"scripts\u002Fvalidate_setup.py\")`, or\n `python scripts\u002Fvalidate_setup.py [--strict]`). It checks\n Docker, NVIDIA Container Toolkit, GPU\u002Fdriver, and\n `NGC_API_KEY`. Resolve any FAIL line before pulling the image.\n2. **Authenticate Docker to NGC + pull the public containers.**\n See [`references\u002Finstall.md`](references\u002Finstall.md). Pull both\n `nvcr.io\u002Fnvidia\u002Fnre\u002Fnre:latest` and\n `nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools:latest`.\n3. **Confirm input layout.** The dataset directory must contain\n `\u003CNAME>.zarr.itar`, `\u003CNAME>.json`, and any `\u003CNAME>.aux.*.zarr`\n shards. If the NCore data is fresh, generate auxiliary data\n first — see `references\u002Faux-data.md`.\n4. **Train \u002F validate the reconstruction.** Run the train recipe\n in [`references\u002Fcookbook.md`](references\u002Fcookbook.md) with the\n chosen Hydra `--config-name`, `mode`, `dataset.path`, and\n `out_dir`. For multi-GPU append `trainer.world_size=\u003CN>\n trainer.num_nodes=\u003CM>` (see Workflow D). Set\n `checkpoint.artifact.enabled=true` if you intend to render or\n serve the result.\n5. **Export downstream artifacts.** Use export sub-commands\n (`export-gaussian-plys`, `export-mesh`, `export-ground-mesh`,\n `export-ego-mask`, `export-depth`, `export-sequence-tracks`,\n `export-ncore-tracks`, …) — full surface in\n `references\u002Fcli-reference.md`.\n6. **Render novel views — pick the backend.**\n - **Local CLI** — `nre render --artifact-path \u003Cusdz>` writes\n frames on disk along the training trajectory, with optional\n rig offsets or `--custom-rig-trajectory`. No gRPC server.\n See `references\u002Flocal-render.md`.\n - **Warm RGB service** — boot `serve-grpc` once with\n `scripts\u002Fsession_warm_server.sh`, extract protobuf stubs,\n and use\n `references\u002FNRE_RenderClient\u002Fscripts\u002Fthin_client.py` for\n repeated single-camera or `batch_render_rgb` calls.\n - **Remote CLI \u002F simulator integration** — `serve-grpc` +\n `render-grpc` (or your own client via `nre.grpc.protos`).\n Required for LiDAR rendering, simulator loops, Difix, or\n `--edit-assets`. See `references\u002Fgrpc-api.md` and\n `references\u002Fphysical-ai-render.md`.\n7. **Edit actors (optional).** Run `export-external-assets` to\n repackage Asset-Harvester output into a new USDZ, then pass\n the produced `edit-assets.json` to `render-grpc --edit-assets`\n (with `serve-grpc --enable-editing-actors`). See\n `references\u002Fasset-editing.md`.\n8. **Validate the result.** Confirm\n `\u003Coutput_dir>\u002F\u003CRUN-ID>\u002Fusd-out\u002Flast.usdz` opens, `metrics.yaml`\n reports a reasonable `test\u002Fpsnr`, and the generated MP4s\n render. For more thorough metrics use Workflow I (eval) in\n [`references\u002Fworkflows.md`](references\u002Fworkflows.md). Tear\n down any gRPC server (`Ctrl-C` or `docker rm -f`).\n\nFor any NRE task expected to run 5 minutes or longer (training,\nOSMO jobs, multi-clip renders), follow\n`references\u002Flong-running-tasks.md`: delegate to a subagent \u002F\nbackground job and report compact status at least every 5\nminutes.\n\n## Examples\n\n### Example 1 — End-to-end NCore → USDZ → render\n\nWalk Workflow A in\n[`references\u002Fworkflows.md`](references\u002Fworkflows.md): validate\nhost → generate aux data → train (cookbook recipe) → export →\nlocal render or `serve-grpc`. Concrete commands live in the\nreferenced files; this index does not duplicate them.\n\n### Example 2 — Skip training, render the gated Physical AI dataset\n\nWalk Workflow B: download\n`nvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec` from HuggingFace,\nthen jump to `serve-grpc` + a Python client. Coordinate-frame\nconversion code is in `references\u002Fphysical-ai-render.md`.\n\n### Example 3 — Insert Asset-Harvester actors into a USDZ\n\nWalk Workflow C: run `asset-harvester`, then\n`export-external-assets`, edit `edit-assets.json`, and call\n`serve-grpc --enable-editing-actors` + `render-grpc\n--edit-assets`. Schema lives in `references\u002Fasset-editing.md`.\n\n### Example 4 — Warm-server thin-client for repeated RGB renders\n\nWalk the warm-server quick start at the bottom of\n[`references\u002Fworkflows.md`](references\u002Fworkflows.md). Boot\n`scripts\u002Fsession_warm_server.sh`, render with\n`thin_client.py`, tear down with `scripts\u002Fsession_teardown.sh`.\n\n## Backend Selection\n\nPick the smallest backend that exposes the requested feature:\n\n- **Local Docker, single command.** Use `nre render`, `render-grpc`,\n or an export sub-command directly. Simplest for one-off renders,\n LiDAR sweeps, actor edits, rolling shutter, in-container video\n export, or exact `--replicate-training-views` behavior. See\n `references\u002Flocal-render.md`, `references\u002Fnre-image-notes.md`,\n and `references\u002Fmp4-encoding.md`.\n- **Local Docker, warm `serve-grpc` + thin host client.** Use for\n render-heavy RGB sessions where repeated Docker\u002FPython\u002FCUDA\n cold-start dominates latency, or where multiple cameras should\n be rendered through one `batch_render_rgb` RPC. See\n `references\u002FNRE_RenderClient\u002FREADME.md` and\n `scripts\u002Fsession_warm_server.sh` \u002F `scripts\u002Fsession_teardown.sh`.\n- **OSMO \u002F cluster workflows.** Use the templates under\n `references\u002Fexample-workflows\u002Fosmo\u002F` for multi-clip fan-out,\n isolation from the local machine, or training jobs that should\n not run on the user's workstation. Follow\n `references\u002Fngc-and-registry.md` for registry credentials and\n `references\u002Flong-running-tasks.md` for polling discipline.\n\n## Output Format\n\nStructured deliverables placed under `${output_dir}\u002F${RUN_ID}\u002F`\nby the NRE container (no JSON state file required from the\nagent):\n\n- `config\u002Fparsed.yaml` — Hydra-resolved training config.\n- `checkpoints\u002Flast.ckpt` (plus periodic snapshots).\n- `val\u002Fmetrics.yaml` — per-frame PSNR \u002F SSIM \u002F LPIPS under\n `test\u002F*`.\n- `val\u002F*.mp4`, `val\u002F\u003Cframe>\u002F*.png` — depth, opacity, segmentation,\n RGB visualisations.\n- `usd-out\u002Flast.usdz` — USDZ containing the trained reconstruction,\n `data_info.json`, `rig_trajectories.json`,\n `sequence_tracks.json`, `parsed_config.yaml`, `checkpoint.ckpt`,\n optional `mesh.ply`, and `map.xodr`. Render with `nre render`,\n `serve-grpc` + `render-grpc`, the in-container `viewer`, or hand\n to a downstream simulator (CARLA, AlpaSim, Isaac Sim).\n- `*.ply` \u002F `ego_mask\u002F*` \u002F `depth\u002F*` \u002F `sequence_tracks.json` \u002F\n `ncore_tracks.json` \u002F `mesh.ply` \u002F `ground_mesh.ply` — produced\n by the matching export sub-command.\n\n## Scripts\n\n| Script | Purpose | Usage |\n|--------|---------|-------|\n| `scripts\u002Fvalidate_setup.py` | Verify Docker, NVIDIA Container Toolkit, GPU\u002Fdriver R570+ (R535+ minimum), NGC login, and `NGC_API_KEY` env var. No network calls. | `run_script(\"scripts\u002Fvalidate_setup.py\")` or `python scripts\u002Fvalidate_setup.py [--strict]` |\n| `scripts\u002Fsession_warm_server.sh` | Idempotently boot a session-scoped `nre serve-grpc` container for the thin Python client. Discovers a cached 26.04+ renderer image, mounts the USDZ root, waits for readiness. | `NRE_GRPC_USDZ_HOST_DIR=\u002Fpath\u002Fto\u002Fusdz\u002Froot bash scripts\u002Fsession_warm_server.sh` |\n| `scripts\u002Fsession_teardown.sh` | Stop and remove the warm `serve-grpc` container and clear its state file without racing the next boot. | `bash scripts\u002Fsession_teardown.sh` |\n\n## References\n\nRead these on demand; keep `SKILL.md` as the routing layer.\n\n- [`references\u002Finstall.md`](references\u002Finstall.md) — `docker\n login nvcr.io`, image pull, full prerequisite matrix, and safe\n secret-handling for `NGC_API_KEY` \u002F `HF_TOKEN`.\n- [`references\u002Fcookbook.md`](references\u002Fcookbook.md) — most-used\n `docker run` invocations: train + validate, re-validate with\n shift, local render at quarter or native res, `serve-grpc` boot,\n LiDAR sweep, in-container `--help`.\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) — workflows\n A – I end-to-end, plus the warm-server thin-client quick start.\n- [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md)\n — extended error matrix (`OOM`, `wandb` blocking,\n `--artifact-glob` mismatches, deprecated flags, gRPC LiDAR size,\n etc.).\n- [`references\u002Fteardown.md`](references\u002Fteardown.md) — disk\n cleanup, post-teardown verification, ownership-recovery.\n- `references\u002Fcli-reference.md` — full sub-command surface of the\n NRE container (training, validation, `render`, `serve-grpc`,\n `render-grpc`, `render-novel-trajectory`, every `export-*`,\n `upgrade-config` \u002F `upgrade-artifact`, `gaussian-statistics`,\n `eval-rendering-metrics`, `compute-metrics`, `viewer`,\n `ply_viewer`, `profile-dataloader`, `run-script`, the\n `nre-tools` aux-data + AH entry points).\n- `references\u002Fconfiguration.md` — Hydra recipe map for Waymo \u002F NV\n \u002F PandaSet \u002F Tesla \u002F Alpasim, plus override matrix.\n- `references\u002Faux-data.md` — `nre-tools` auxiliary-data CLI.\n- `references\u002Flocal-render.md` — host-side `docker run … render`\n recipes for rig offsets and `export-custom-rig-trajectory`.\n- `references\u002FNRE_RenderClient\u002FREADME.md` — warm-server thin\n Python gRPC client.\n- `references\u002Fgrpc-api.md` — sensorsim gRPC server flags + Python\n client cookbook.\n- `references\u002Fnre-image-notes.md` — cached-image discovery, 26.04+\n vs 26.03 vs pre-26.03 flags.\n- `references\u002Fngc-and-registry.md` — NGC API key resolution.\n- `references\u002Fmp4-encoding.md` — host-side ffmpeg recipe.\n- `references\u002Fasset-editing.md` — `export-external-assets` +\n `edit-assets.json` schema.\n- `references\u002Fphysical-ai-render.md` — recipe for rendering the\n HuggingFace NuRec dataset.\n- `references\u002Fexample-workflows\u002F` — bash, Hydra, and OSMO\n templates.\n- `references\u002Frig-json\u002F` — bundled `rig.json` and\n `augmented_rig.json`.\n- `references\u002Fcustom-rig-trajectories\u002F` — pre-baked\n `export-custom-rig-trajectory` outputs.\n- `references\u002Flong-running-tasks.md` — background-job + 5-minute\n status reporting convention.\n- `references\u002Fnurec-skill-catalog.md` — routing table for sibling\n NuRec-stack skills.\n- Public product page: \u003Chttps:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fomniverse\u002Fnurec\u002F>\n- HF dataset: \u003Chttps:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fnvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec>\n- HF Fixer model: \u003Chttps:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FDifix3D>\n- NGC Fixer model card: \u003Chttps:\u002F\u002Fcatalog.ngc.nvidia.com\u002Forgs\u002Fnvidia\u002Fteams\u002Fnre\u002Fmodels\u002Fnurec-fixer>\n\n## Prerequisites\n\nLinux x86_64 + NVIDIA GPU + Docker 23+ + NVIDIA Container Toolkit\n1.13+ + `NGC_API_KEY`. Full matrix (driver minimums per arch,\nshm-size, file ownership, GPU-tier guidance) lives in\n[`references\u002Finstall.md`](references\u002Finstall.md). Always verify\nvia `scripts\u002Fvalidate_setup.py` before pulling the image.\n\n## Limitations\n\n- **Linux x86_64 only.** aarch64 (e.g. Jetson) is not supported.\n- **Internal source not redistributable.** Use only the public\n NGC containers and the public NuRec docs.\n- **Multi-GPU defaults are conservative.** Set\n `trainer.world_size` \u002F `trainer.num_nodes` explicitly to scale\n out; SLURM is auto-detected when both are `0`. Quality plateaus\n past ~6 GPUs; per release notes, multi-GPU +\n `dataset.aux_data=false` is a known crash combination.\n- **`--config-name` paths differ between train and val\u002Fexport.**\n Training uses container-bundled recipes; validation and exports\n re-pass the `parsed.yaml` written under\n `\u003Coutput_dir>\u002F\u003CRUN-ID>\u002Fconfig\u002F`.\n- **Validation may prompt for `wandb`.** Choose option 3 to skip\n in non-interactive runs, or pass `logger=tensorboard` \u002F\n `logger=dummy`.\n- **Render gRPC is data-format-pinned.** Older releases warn \u002F\n reject artifacts that pre-date them; check release notes when\n mixing client \u002F server versions.\n- **Asset-Harvester input only.** `export-external-assets`\n requires AH outputs; raw `.ply` files won't carry the\n per-asset cuboid metadata.\n- **Difix variants are pluggable.** The container ships both the\n Cosmos Difix variant (default since 25.09 —\n `difix=cosmos_difix`) and the legacy Stable-Diffusion variant\n (`difix=sd_difix`). The newer Cosmos-Predict-based Fixer\n variants live in the `nurec-fixer` skill.\n- **`render` ↔ `render-grpc` overlap.** `render` runs\n in-container without a server; `render-grpc` requires an active\n `serve-grpc`. Use `render` for batch novel-view jobs and\n `render-grpc` when you need actor editing, LiDAR rendering, or\n a long-lived service.\n\n## Troubleshooting (top 4)\n\n| Error | Cause | Fix |\n|-------|-------|-----|\n| `Unable to find image 'nvcr.io\u002Fnvidia\u002Fnre\u002Fnre:latest'` | Docker not authenticated to NGC. | `docker login nvcr.io` with `Username: $oauthtoken`. |\n| `OOM Killed` \u002F `CUDA out of memory` during training | Default recipe needs >= 48 GB VRAM. | Reduce `dataset.camera_ids`, lower `trainer.max_epochs`, switch to `trainer.precision=16-mixed`, or use a 48 GB+ GPU. |\n| `serve-grpc` fails to find the USDZ | `--artifact-glob` must end in `.usdz` and be quoted. | Use e.g. `--artifact-glob \u002Fworkdir\u002Foutput\u002F\u003CRUN-ID>\u002Fusd-out\u002Flast.usdz`. |\n| Edits silently ignored from `render-grpc --edit-assets` | Server started without `--enable-editing-actors`. | Restart `serve-grpc` with that flag. |\n\nFull matrix in\n[`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md).\n\n## Teardown\n\nFull inventory, ownership-recovery, and post-teardown verification\ncommands live in [`references\u002Fteardown.md`](references\u002Fteardown.md).\nHeadline: stop `serve-grpc` containers, `docker image rm`\nnre\u002Fnre-tools, `rm -rf ${HOME}\u002F.cache\u002Fnre` and your\n`\u003Coutput_dir>\u002F\u003CRUN-ID>\u002F`. Do not revoke `NGC_API_KEY` unless you\nsuspect it has been leaked.\n",{"data":35,"body":62},{"name":4,"description":6,"version":36,"tools":37,"license":26,"compatibility":41,"dependencies":42,"metadata":46},"0.2.2",[38,39,40],"Shell","Read","Write","Linux x86_64, 1+ NVIDIA GPU (Ampere A100\u002FA10\u002FA40\u002FRTX A6000, Ada L20\u002FL40\u002FL40S, Hopper H100\u002FH20, or Blackwell RTX Pro 6000D) with CUDA 12.8 and >= 24 GB VRAM (48+ GB recommended); driver R570+ recommended (R580+ on Blackwell, R535+ minimum for Fixer-only). Docker >= 23.0.1 + NVIDIA Container Toolkit >= 1.13.5, NGC account with NGC_API_KEY exported.",[43,44,45],"bash","docker","python3",{"author":47,"tags":48,"product_page":54,"ngc_container":55,"ngc_tools_container":56,"ngc_fixer_model":57,"hf_datasets":58,"hf_fixer_model":59,"release_date":60,"time-estimate":61},"NVIDIA NRS \u003Cnurec-skills@nvidia.com>",[49,50,51,52,53],"nurec","autonomous-vehicles","neural-reconstruction","rendering","container","https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fomniverse\u002Fnurec\u002F","nvcr.io\u002Fnvidia\u002Fnre\u002Fnre:latest","nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools:latest","https:\u002F\u002Fcatalog.ngc.nvidia.com\u002Forgs\u002Fnvidia\u002Fteams\u002Fnre\u002Fmodels\u002Fnurec-fixer","https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fnvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec","https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FDifix3D","2026-04-30","2h",{"type":63,"children":64},"root",[65,74,81,104,117,123,134,279,287,357,363,661,667,1161,1174,1180,1187,1209,1215,1240,1246,1292,1298,1335,1341,1346,1478,1484,1497,1700,1706,1840,1846,1859,2352,2358,2387,2393,2641,2647,2843,2857,2863],{"type":66,"tag":67,"props":68,"children":70},"element","h1",{"id":69},"nre-nvidia-omniverse-nurec-neural-reconstruction-engine",[71],{"type":72,"value":73},"text","NRE — NVIDIA Omniverse NuRec (Neural Reconstruction Engine)",{"type":66,"tag":75,"props":76,"children":78},"h2",{"id":77},"purpose",[79],{"type":72,"value":80},"Purpose",{"type":66,"tag":82,"props":83,"children":84},"p",{},[85,87,94,96,102],{"type":72,"value":86},"Drive the public NVIDIA Omniverse NuRec \u002F Neural Reconstruction\nEngine containers (",{"type":66,"tag":88,"props":89,"children":91},"code",{"className":90},[],[92],{"type":72,"value":93},"nvcr.io\u002Fnvidia\u002Fnre\u002Fnre",{"type":72,"value":95},",\n",{"type":66,"tag":88,"props":97,"children":99},{"className":98},[],[100],{"type":72,"value":101},"nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools",{"type":72,"value":103},") to train a 3DGUT\u002F3DGRT Gaussian\nreconstruction from an NCore V4 camera+LiDAR clip, render novel\nviews (locally or via gRPC), generate aux data, export\nPLY\u002Fdepth\u002Fmesh\u002Fego-mask\u002Ftracks, package Asset Harvester output\ninto a USDZ, and evaluate rendering metrics.",{"type":66,"tag":82,"props":105,"children":106},{},[107,109,115],{"type":72,"value":108},"This skill carries the host-side toolkit around the NRE CLI: NGC\ncredential resolution, cached-image notes, local render recipes,\nMP4 encoding, warm ",{"type":66,"tag":88,"props":110,"children":112},{"className":111},[],[113],{"type":72,"value":114},"serve-grpc",{"type":72,"value":116}," boot\u002Fteardown scripts, a thin\nPython gRPC client for repeated RGB renders, bundled rig JSONs,\npre-baked custom-rig trajectories, and bash \u002F Hydra \u002F OSMO\nworkflow templates.",{"type":66,"tag":75,"props":118,"children":120},{"id":119},"when-to-use-when-not-to-use",[121],{"type":72,"value":122},"When to Use \u002F When NOT to Use",{"type":66,"tag":82,"props":124,"children":125},{},[126,132],{"type":66,"tag":127,"props":128,"children":129},"strong",{},[130],{"type":72,"value":131},"Use this skill when",{"type":72,"value":133}," the user has an NCore V4 clip (or a USDZ +\nNRE artifact pair) on a Linux x86_64 host with an NVIDIA GPU and\nan NGC API key, and wants to train, render, generate aux data,\nexport artifacts, insert\u002Fremove actors, run the gRPC server, or\nevaluate metrics. Concrete triggers:",{"type":66,"tag":135,"props":136,"children":137},"ul",{},[138,144,157,162,167,179,184,208,220,233,274],{"type":66,"tag":139,"props":140,"children":141},"li",{},[142],{"type":72,"value":143},"Train a multi-camera + LiDAR AV clip into a renderable USDZ\nscene with 3DGUT (or 3DGRT ray-traced) Gaussians.",{"type":66,"tag":139,"props":145,"children":146},{},[147,149,155],{"type":72,"value":148},"Generate NuRec auxiliary data (seg, depth, ego mask, DINOv2,\nLiDAR-seg visibility) using ",{"type":66,"tag":88,"props":150,"children":152},{"className":151},[],[153],{"type":72,"value":154},"nre-tools",{"type":72,"value":156},".",{"type":66,"tag":139,"props":158,"children":159},{},[160],{"type":72,"value":161},"Render frames locally (no server) along the training rig or a\ncustom rig + offsets.",{"type":66,"tag":139,"props":163,"children":164},{},[165],{"type":72,"value":166},"Render novel views via the sensorsim gRPC API (CARLA, Isaac\nSim, AlpaSim, custom simulator), optionally with Difix\nartifact-removal.",{"type":66,"tag":139,"props":168,"children":169},{},[170,172,178],{"type":72,"value":171},"Render LiDAR sweeps via ",{"type":66,"tag":88,"props":173,"children":175},{"className":174},[],[176],{"type":72,"value":177},"render-grpc --lidar",{"type":72,"value":156},{"type":66,"tag":139,"props":180,"children":181},{},[182],{"type":72,"value":183},"Export PLY \u002F ego masks \u002F depth \u002F Poisson mesh \u002F ground mesh \u002F\npoint clouds \u002F cuboid tracks \u002F NCore tracks \u002F custom rig\ntrajectories.",{"type":66,"tag":139,"props":185,"children":186},{},[187,189,195],{"type":72,"value":188},"Insert \u002F remove \u002F replace 3D actors with ",{"type":66,"tag":88,"props":190,"children":192},{"className":191},[],[193],{"type":72,"value":194},"export-external-assets",{"type":66,"tag":135,"props":196,"children":197},{},[198],{"type":66,"tag":139,"props":199,"children":200},{},[201,207],{"type":66,"tag":88,"props":202,"children":204},{"className":203},[],[205],{"type":72,"value":206},"render-grpc --edit-assets",{"type":72,"value":156},{"type":66,"tag":139,"props":209,"children":210},{},[211,213,219],{"type":72,"value":212},"Render the gated HF dataset\n",{"type":66,"tag":88,"props":214,"children":216},{"className":215},[],[217],{"type":72,"value":218},"nvidia\u002FPhysicalAI-Autonomous-Vehicles-NuRec",{"type":72,"value":156},{"type":66,"tag":139,"props":221,"children":222},{},[223,225,231],{"type":72,"value":224},"Upgrade an old USDZ once (",{"type":66,"tag":88,"props":226,"children":228},{"className":227},[],[229],{"type":72,"value":230},"upgrade-artifact",{"type":72,"value":232},").",{"type":66,"tag":139,"props":234,"children":235},{},[236,238,244,246,252,253,259,260,266,267,273],{"type":72,"value":237},"Inspect \u002F evaluate (",{"type":66,"tag":88,"props":239,"children":241},{"className":240},[],[242],{"type":72,"value":243},"export-parsed-config",{"type":72,"value":245},", ",{"type":66,"tag":88,"props":247,"children":249},{"className":248},[],[250],{"type":72,"value":251},"gaussian-statistics",{"type":72,"value":95},{"type":66,"tag":88,"props":254,"children":256},{"className":255},[],[257],{"type":72,"value":258},"eval-rendering-metrics",{"type":72,"value":245},{"type":66,"tag":88,"props":261,"children":263},{"className":262},[],[264],{"type":72,"value":265},"compute-metrics",{"type":72,"value":95},{"type":66,"tag":88,"props":268,"children":270},{"className":269},[],[271],{"type":72,"value":272},"eval-ground-mesh",{"type":72,"value":232},{"type":66,"tag":139,"props":275,"children":276},{},[277],{"type":72,"value":278},"Browse a USDZ or PLY in the in-container viewer.",{"type":66,"tag":82,"props":280,"children":281},{},[282],{"type":66,"tag":127,"props":283,"children":284},{},[285],{"type":72,"value":286},"Do NOT use this skill when:",{"type":66,"tag":135,"props":288,"children":289},{},[290,303,330],{"type":66,"tag":139,"props":291,"children":292},{},[293,295,301],{"type":72,"value":294},"The user still needs to convert raw sensor data into NCore V4\n(use the ",{"type":66,"tag":88,"props":296,"children":298},{"className":297},[],[299],{"type":72,"value":300},"ncore",{"type":72,"value":302}," skill first; NRE consumes NCore-formatted\nshards).",{"type":66,"tag":139,"props":304,"children":305},{},[306,308,314,316,322,324,329],{"type":72,"value":307},"The user wants per-object 3D asset extraction from sparse views\n(use ",{"type":66,"tag":88,"props":309,"children":311},{"className":310},[],[312],{"type":72,"value":313},"asset-harvester",{"type":72,"value":315},"; NRE only ",{"type":66,"tag":317,"props":318,"children":319},"em",{},[320],{"type":72,"value":321},"consumes",{"type":72,"value":323}," AH outputs via\n",{"type":66,"tag":88,"props":325,"children":327},{"className":326},[],[328],{"type":72,"value":194},{"type":72,"value":232},{"type":66,"tag":139,"props":331,"children":332},{},[333,335,341,343,349,351,356],{"type":72,"value":334},"The user only needs to clean up already-rendered frames using\nthe standalone Cosmos-based Fixer (use ",{"type":66,"tag":88,"props":336,"children":338},{"className":337},[],[339],{"type":72,"value":340},"nurec-fixer",{"type":72,"value":342},"). NRE's\ninline ",{"type":66,"tag":88,"props":344,"children":346},{"className":345},[],[347],{"type":72,"value":348},"--enable-difix",{"type":72,"value":350}," flag is still on this skill's surface,\nbut the standalone harmonizer pipeline is owned by\n",{"type":66,"tag":88,"props":352,"children":354},{"className":353},[],[355],{"type":72,"value":340},{"type":72,"value":156},{"type":66,"tag":75,"props":358,"children":360},{"id":359},"inputs",[361],{"type":72,"value":362},"Inputs",{"type":66,"tag":135,"props":364,"children":365},{},[366,399,417,427,437,604,643],{"type":66,"tag":139,"props":367,"children":368},{},[369,374,376,382,383,389,391,397],{"type":66,"tag":127,"props":370,"children":371},{},[372],{"type":72,"value":373},"dataset_dir",{"type":72,"value":375}," — host directory holding the NCore shards\n(",{"type":66,"tag":88,"props":377,"children":379},{"className":378},[],[380],{"type":72,"value":381},"\u003CNAME>.zarr.itar",{"type":72,"value":245},{"type":66,"tag":88,"props":384,"children":386},{"className":385},[],[387],{"type":72,"value":388},"\u003CNAME>.json",{"type":72,"value":390},", and any pre-generated\n",{"type":66,"tag":88,"props":392,"children":394},{"className":393},[],[395],{"type":72,"value":396},"\u003CNAME>.aux.*.zarr",{"type":72,"value":398}," auxiliary shards). Required.",{"type":66,"tag":139,"props":400,"children":401},{},[402,407,409,415],{"type":66,"tag":127,"props":403,"children":404},{},[405],{"type":72,"value":406},"dataset_name",{"type":72,"value":408}," — basename of the NCore dataset (the part\nbefore ",{"type":66,"tag":88,"props":410,"children":412},{"className":411},[],[413],{"type":72,"value":414},".zarr.itar",{"type":72,"value":416},"). Required.",{"type":66,"tag":139,"props":418,"children":419},{},[420,425],{"type":66,"tag":127,"props":421,"children":422},{},[423],{"type":72,"value":424},"output_dir",{"type":72,"value":426}," — host directory NRE will fill with checkpoints,\nparsed config, metrics, videos, and USDZ artifacts. Required.",{"type":66,"tag":139,"props":428,"children":429},{},[430,435],{"type":66,"tag":127,"props":431,"children":432},{},[433],{"type":72,"value":434},"camera_ids \u002F lidar_ids",{"type":72,"value":436}," — sensor IDs from the NCore JSON to\ninclude. Default: all sensors per recipe.",{"type":66,"tag":139,"props":438,"children":439},{},[440,445,447],{"type":66,"tag":127,"props":441,"children":442},{},[443],{"type":72,"value":444},"config_name",{"type":72,"value":446}," — Hydra config path resolved inside the\ncontainer. Pick by source dataset:\n",{"type":66,"tag":135,"props":448,"children":449},{},[450,491,559,592],{"type":66,"tag":139,"props":451,"children":452},{},[453,458,460,466,468,474,476,482,483,489],{"type":66,"tag":127,"props":454,"children":455},{},[456],{"type":72,"value":457},"Waymo Open Dataset",{"type":72,"value":459}," →\n",{"type":66,"tag":88,"props":461,"children":463},{"className":462},[],[464],{"type":72,"value":465},"configs\u002Fapps\u002FAV\u002FWaymo\u002F3dgut_dynamic.yaml",{"type":72,"value":467}," (and its\n",{"type":66,"tag":88,"props":469,"children":471},{"className":470},[],[472],{"type":72,"value":473},"_mcmc",{"type":72,"value":475}," \u002F ",{"type":66,"tag":88,"props":477,"children":479},{"className":478},[],[480],{"type":72,"value":481},"_road_semantic",{"type":72,"value":475},{"type":66,"tag":88,"props":484,"children":486},{"className":485},[],[487],{"type":72,"value":488},"_static",{"type":72,"value":490}," siblings). These are\nWaymo-only — they bake in the Waymo sensor rig and conventions.",{"type":66,"tag":139,"props":492,"children":493},{},[494,499,500,506,508,518,520,526,528,534,536,542,544,550,552,558],{"type":66,"tag":127,"props":495,"children":496},{},[497],{"type":72,"value":498},"NVIDIA Physical AI Autonomous Vehicles (PAI)",{"type":72,"value":459},{"type":66,"tag":88,"props":501,"children":503},{"className":502},[],[504],{"type":72,"value":505},"\u002Fapps\u002Fprod\u002FHyperion-8.1\u002Fcar2sim_6cam.yaml",{"type":72,"value":507}," (the Hyperion-8.1\ncar2sim 6-camera recipe used by the Maglev PAI pipeline).\nTypically referenced via the small overlay shipped at\n",{"type":66,"tag":509,"props":510,"children":512},"a",{"href":511},"references\u002Fconfigs\u002Fpai.yaml",[513],{"type":66,"tag":88,"props":514,"children":516},{"className":515},[],[517],{"type":72,"value":511},{"type":72,"value":519},",\nwhich extends ",{"type":66,"tag":88,"props":521,"children":523},{"className":522},[],[524],{"type":72,"value":525},"car2sim_6cam.yaml",{"type":72,"value":527}," with PAI's\n",{"type":66,"tag":88,"props":529,"children":531},{"className":530},[],[532],{"type":72,"value":533},"lidar_top_360fov",{"type":72,"value":535}," ID, six-camera validation set, and lidar\n",{"type":66,"tag":88,"props":537,"children":539},{"className":538},[],[540],{"type":72,"value":541},"intensity",{"type":72,"value":543}," supervision; mount it as\n",{"type":66,"tag":88,"props":545,"children":547},{"className":546},[],[548],{"type":72,"value":549},"{nre_config_dir}\u002Fexternal_overrides.yaml",{"type":72,"value":551}," and pass\n",{"type":66,"tag":88,"props":553,"children":555},{"className":554},[],[556],{"type":72,"value":557},"--config-name=external_overrides",{"type":72,"value":156},{"type":66,"tag":139,"props":560,"children":561},{},[562,567,569,575,577,583,585,591],{"type":66,"tag":127,"props":563,"children":564},{},[565],{"type":72,"value":566},"PandaSet \u002F NVIDIA AV (NV) \u002F Tesla \u002F Alpasim",{"type":72,"value":568}," → see the\nmatching ",{"type":66,"tag":88,"props":570,"children":572},{"className":571},[],[573],{"type":72,"value":574},"configs\u002Fapps\u002FAV\u002F{PandaSet,NV,Tesla}\u002F…",{"type":72,"value":576}," or\n",{"type":66,"tag":88,"props":578,"children":580},{"className":579},[],[581],{"type":72,"value":582},"configs\u002Fapps\u002FAlpasim\u002F…",{"type":72,"value":584}," recipes in\n",{"type":66,"tag":88,"props":586,"children":588},{"className":587},[],[589],{"type":72,"value":590},"references\u002Fconfiguration.md",{"type":72,"value":156},{"type":66,"tag":139,"props":593,"children":594},{},[595,597,602],{"type":72,"value":596},"Do ",{"type":66,"tag":127,"props":598,"children":599},{},[600],{"type":72,"value":601},"not",{"type":72,"value":603}," use the Waymo recipes for PAI clips — the\nsensor rig, validation cameras, and lidar IDs differ.",{"type":66,"tag":139,"props":605,"children":606},{},[607,612,614,620,621,627,629,635,637,642],{"type":66,"tag":127,"props":608,"children":609},{},[610],{"type":72,"value":611},"mode",{"type":72,"value":613}," — ",{"type":66,"tag":88,"props":615,"children":617},{"className":616},[],[618],{"type":72,"value":619},"train",{"type":72,"value":245},{"type":66,"tag":88,"props":622,"children":624},{"className":623},[],[625],{"type":72,"value":626},"val",{"type":72,"value":628},", or ",{"type":66,"tag":88,"props":630,"children":632},{"className":631},[],[633],{"type":72,"value":634},"trainval",{"type":72,"value":636},". Default: ",{"type":66,"tag":88,"props":638,"children":640},{"className":639},[],[641],{"type":72,"value":634},{"type":72,"value":156},{"type":66,"tag":139,"props":644,"children":645},{},[646,651,653,660],{"type":66,"tag":127,"props":647,"children":648},{},[649],{"type":72,"value":650},"NGC_API_KEY",{"type":72,"value":652}," — required env var. Generate at\n",{"type":66,"tag":509,"props":654,"children":658},{"href":655,"rel":656},"https:\u002F\u002Forg.ngc.nvidia.com\u002Fsetup\u002Fapi-keys",[657],"nofollow",[659],{"type":72,"value":655},{"type":72,"value":156},{"type":66,"tag":75,"props":662,"children":664},{"id":663},"instructions",[665],{"type":72,"value":666},"Instructions",{"type":66,"tag":668,"props":669,"children":670},"ol",{},[671,712,746,783,849,916,1054,1101],{"type":66,"tag":139,"props":672,"children":673},{},[674,679,681,687,689,695,697,703,705,710],{"type":66,"tag":127,"props":675,"children":676},{},[677],{"type":72,"value":678},"Validate prerequisites.",{"type":72,"value":680}," Have the agent execute\n",{"type":66,"tag":88,"props":682,"children":684},{"className":683},[],[685],{"type":72,"value":686},"scripts\u002Fvalidate_setup.py",{"type":72,"value":688}," via its standard script runner\n(",{"type":66,"tag":88,"props":690,"children":692},{"className":691},[],[693],{"type":72,"value":694},"run_script(\"scripts\u002Fvalidate_setup.py\")",{"type":72,"value":696},", or\n",{"type":66,"tag":88,"props":698,"children":700},{"className":699},[],[701],{"type":72,"value":702},"python scripts\u002Fvalidate_setup.py [--strict]",{"type":72,"value":704},"). It checks\nDocker, NVIDIA Container Toolkit, GPU\u002Fdriver, and\n",{"type":66,"tag":88,"props":706,"children":708},{"className":707},[],[709],{"type":72,"value":650},{"type":72,"value":711},". Resolve any FAIL line before pulling the image.",{"type":66,"tag":139,"props":713,"children":714},{},[715,720,722,731,733,738,740,745],{"type":66,"tag":127,"props":716,"children":717},{},[718],{"type":72,"value":719},"Authenticate Docker to NGC + pull the public containers.",{"type":72,"value":721},"\nSee ",{"type":66,"tag":509,"props":723,"children":725},{"href":724},"references\u002Finstall.md",[726],{"type":66,"tag":88,"props":727,"children":729},{"className":728},[],[730],{"type":72,"value":724},{"type":72,"value":732},". Pull both\n",{"type":66,"tag":88,"props":734,"children":736},{"className":735},[],[737],{"type":72,"value":55},{"type":72,"value":739}," and\n",{"type":66,"tag":88,"props":741,"children":743},{"className":742},[],[744],{"type":72,"value":56},{"type":72,"value":156},{"type":66,"tag":139,"props":747,"children":748},{},[749,754,756,761,762,767,769,774,776,782],{"type":66,"tag":127,"props":750,"children":751},{},[752],{"type":72,"value":753},"Confirm input layout.",{"type":72,"value":755}," The dataset directory must contain\n",{"type":66,"tag":88,"props":757,"children":759},{"className":758},[],[760],{"type":72,"value":381},{"type":72,"value":245},{"type":66,"tag":88,"props":763,"children":765},{"className":764},[],[766],{"type":72,"value":388},{"type":72,"value":768},", and any ",{"type":66,"tag":88,"props":770,"children":772},{"className":771},[],[773],{"type":72,"value":396},{"type":72,"value":775},"\nshards. If the NCore data is fresh, generate auxiliary data\nfirst — see ",{"type":66,"tag":88,"props":777,"children":779},{"className":778},[],[780],{"type":72,"value":781},"references\u002Faux-data.md",{"type":72,"value":156},{"type":66,"tag":139,"props":784,"children":785},{},[786,791,793,802,804,810,811,816,817,823,825,831,833,839,841,847],{"type":66,"tag":127,"props":787,"children":788},{},[789],{"type":72,"value":790},"Train \u002F validate the reconstruction.",{"type":72,"value":792}," Run the train recipe\nin ",{"type":66,"tag":509,"props":794,"children":796},{"href":795},"references\u002Fcookbook.md",[797],{"type":66,"tag":88,"props":798,"children":800},{"className":799},[],[801],{"type":72,"value":795},{"type":72,"value":803}," with the\nchosen Hydra ",{"type":66,"tag":88,"props":805,"children":807},{"className":806},[],[808],{"type":72,"value":809},"--config-name",{"type":72,"value":245},{"type":66,"tag":88,"props":812,"children":814},{"className":813},[],[815],{"type":72,"value":611},{"type":72,"value":245},{"type":66,"tag":88,"props":818,"children":820},{"className":819},[],[821],{"type":72,"value":822},"dataset.path",{"type":72,"value":824},", and\n",{"type":66,"tag":88,"props":826,"children":828},{"className":827},[],[829],{"type":72,"value":830},"out_dir",{"type":72,"value":832},". For multi-GPU append ",{"type":66,"tag":88,"props":834,"children":836},{"className":835},[],[837],{"type":72,"value":838},"trainer.world_size=\u003CN> trainer.num_nodes=\u003CM>",{"type":72,"value":840}," (see Workflow D). Set\n",{"type":66,"tag":88,"props":842,"children":844},{"className":843},[],[845],{"type":72,"value":846},"checkpoint.artifact.enabled=true",{"type":72,"value":848}," if you intend to render or\nserve the result.",{"type":66,"tag":139,"props":850,"children":851},{},[852,857,859,865,866,872,873,879,880,886,887,893,894,900,901,907,909,915],{"type":66,"tag":127,"props":853,"children":854},{},[855],{"type":72,"value":856},"Export downstream artifacts.",{"type":72,"value":858}," Use export sub-commands\n(",{"type":66,"tag":88,"props":860,"children":862},{"className":861},[],[863],{"type":72,"value":864},"export-gaussian-plys",{"type":72,"value":245},{"type":66,"tag":88,"props":867,"children":869},{"className":868},[],[870],{"type":72,"value":871},"export-mesh",{"type":72,"value":245},{"type":66,"tag":88,"props":874,"children":876},{"className":875},[],[877],{"type":72,"value":878},"export-ground-mesh",{"type":72,"value":95},{"type":66,"tag":88,"props":881,"children":883},{"className":882},[],[884],{"type":72,"value":885},"export-ego-mask",{"type":72,"value":245},{"type":66,"tag":88,"props":888,"children":890},{"className":889},[],[891],{"type":72,"value":892},"export-depth",{"type":72,"value":245},{"type":66,"tag":88,"props":895,"children":897},{"className":896},[],[898],{"type":72,"value":899},"export-sequence-tracks",{"type":72,"value":95},{"type":66,"tag":88,"props":902,"children":904},{"className":903},[],[905],{"type":72,"value":906},"export-ncore-tracks",{"type":72,"value":908},", …) — full surface in\n",{"type":66,"tag":88,"props":910,"children":912},{"className":911},[],[913],{"type":72,"value":914},"references\u002Fcli-reference.md",{"type":72,"value":156},{"type":66,"tag":139,"props":917,"children":918},{},[919,924],{"type":66,"tag":127,"props":920,"children":921},{},[922],{"type":72,"value":923},"Render novel views — pick the backend.",{"type":66,"tag":135,"props":925,"children":926},{},[927,959,1000],{"type":66,"tag":139,"props":928,"children":929},{},[930,935,936,942,944,950,952,958],{"type":66,"tag":127,"props":931,"children":932},{},[933],{"type":72,"value":934},"Local CLI",{"type":72,"value":613},{"type":66,"tag":88,"props":937,"children":939},{"className":938},[],[940],{"type":72,"value":941},"nre render --artifact-path \u003Cusdz>",{"type":72,"value":943}," writes\nframes on disk along the training trajectory, with optional\nrig offsets or ",{"type":66,"tag":88,"props":945,"children":947},{"className":946},[],[948],{"type":72,"value":949},"--custom-rig-trajectory",{"type":72,"value":951},". No gRPC server.\nSee ",{"type":66,"tag":88,"props":953,"children":955},{"className":954},[],[956],{"type":72,"value":957},"references\u002Flocal-render.md",{"type":72,"value":156},{"type":66,"tag":139,"props":960,"children":961},{},[962,967,969,974,976,982,984,990,992,998],{"type":66,"tag":127,"props":963,"children":964},{},[965],{"type":72,"value":966},"Warm RGB service",{"type":72,"value":968}," — boot ",{"type":66,"tag":88,"props":970,"children":972},{"className":971},[],[973],{"type":72,"value":114},{"type":72,"value":975}," once with\n",{"type":66,"tag":88,"props":977,"children":979},{"className":978},[],[980],{"type":72,"value":981},"scripts\u002Fsession_warm_server.sh",{"type":72,"value":983},", extract protobuf stubs,\nand use\n",{"type":66,"tag":88,"props":985,"children":987},{"className":986},[],[988],{"type":72,"value":989},"references\u002FNRE_RenderClient\u002Fscripts\u002Fthin_client.py",{"type":72,"value":991}," for\nrepeated single-camera or ",{"type":66,"tag":88,"props":993,"children":995},{"className":994},[],[996],{"type":72,"value":997},"batch_render_rgb",{"type":72,"value":999}," calls.",{"type":66,"tag":139,"props":1001,"children":1002},{},[1003,1008,1009,1014,1016,1022,1024,1030,1032,1038,1040,1046,1047,1053],{"type":66,"tag":127,"props":1004,"children":1005},{},[1006],{"type":72,"value":1007},"Remote CLI \u002F simulator integration",{"type":72,"value":613},{"type":66,"tag":88,"props":1010,"children":1012},{"className":1011},[],[1013],{"type":72,"value":114},{"type":72,"value":1015}," +\n",{"type":66,"tag":88,"props":1017,"children":1019},{"className":1018},[],[1020],{"type":72,"value":1021},"render-grpc",{"type":72,"value":1023}," (or your own client via ",{"type":66,"tag":88,"props":1025,"children":1027},{"className":1026},[],[1028],{"type":72,"value":1029},"nre.grpc.protos",{"type":72,"value":1031},").\nRequired for LiDAR rendering, simulator loops, Difix, or\n",{"type":66,"tag":88,"props":1033,"children":1035},{"className":1034},[],[1036],{"type":72,"value":1037},"--edit-assets",{"type":72,"value":1039},". See ",{"type":66,"tag":88,"props":1041,"children":1043},{"className":1042},[],[1044],{"type":72,"value":1045},"references\u002Fgrpc-api.md",{"type":72,"value":739},{"type":66,"tag":88,"props":1048,"children":1050},{"className":1049},[],[1051],{"type":72,"value":1052},"references\u002Fphysical-ai-render.md",{"type":72,"value":156},{"type":66,"tag":139,"props":1055,"children":1056},{},[1057,1062,1064,1069,1071,1077,1079,1084,1086,1092,1094,1100],{"type":66,"tag":127,"props":1058,"children":1059},{},[1060],{"type":72,"value":1061},"Edit actors (optional).",{"type":72,"value":1063}," Run ",{"type":66,"tag":88,"props":1065,"children":1067},{"className":1066},[],[1068],{"type":72,"value":194},{"type":72,"value":1070}," to\nrepackage Asset-Harvester output into a new USDZ, then pass\nthe produced ",{"type":66,"tag":88,"props":1072,"children":1074},{"className":1073},[],[1075],{"type":72,"value":1076},"edit-assets.json",{"type":72,"value":1078}," to ",{"type":66,"tag":88,"props":1080,"children":1082},{"className":1081},[],[1083],{"type":72,"value":206},{"type":72,"value":1085},"\n(with ",{"type":66,"tag":88,"props":1087,"children":1089},{"className":1088},[],[1090],{"type":72,"value":1091},"serve-grpc --enable-editing-actors",{"type":72,"value":1093},"). See\n",{"type":66,"tag":88,"props":1095,"children":1097},{"className":1096},[],[1098],{"type":72,"value":1099},"references\u002Fasset-editing.md",{"type":72,"value":156},{"type":66,"tag":139,"props":1102,"children":1103},{},[1104,1109,1111,1117,1119,1125,1127,1133,1135,1144,1146,1152,1154,1160],{"type":66,"tag":127,"props":1105,"children":1106},{},[1107],{"type":72,"value":1108},"Validate the result.",{"type":72,"value":1110}," Confirm\n",{"type":66,"tag":88,"props":1112,"children":1114},{"className":1113},[],[1115],{"type":72,"value":1116},"\u003Coutput_dir>\u002F\u003CRUN-ID>\u002Fusd-out\u002Flast.usdz",{"type":72,"value":1118}," opens, ",{"type":66,"tag":88,"props":1120,"children":1122},{"className":1121},[],[1123],{"type":72,"value":1124},"metrics.yaml",{"type":72,"value":1126},"\nreports a reasonable ",{"type":66,"tag":88,"props":1128,"children":1130},{"className":1129},[],[1131],{"type":72,"value":1132},"test\u002Fpsnr",{"type":72,"value":1134},", and the generated MP4s\nrender. For more thorough metrics use Workflow I (eval) in\n",{"type":66,"tag":509,"props":1136,"children":1138},{"href":1137},"references\u002Fworkflows.md",[1139],{"type":66,"tag":88,"props":1140,"children":1142},{"className":1141},[],[1143],{"type":72,"value":1137},{"type":72,"value":1145},". Tear\ndown any gRPC server (",{"type":66,"tag":88,"props":1147,"children":1149},{"className":1148},[],[1150],{"type":72,"value":1151},"Ctrl-C",{"type":72,"value":1153}," or ",{"type":66,"tag":88,"props":1155,"children":1157},{"className":1156},[],[1158],{"type":72,"value":1159},"docker rm -f",{"type":72,"value":232},{"type":66,"tag":82,"props":1162,"children":1163},{},[1164,1166,1172],{"type":72,"value":1165},"For any NRE task expected to run 5 minutes or longer (training,\nOSMO jobs, multi-clip renders), follow\n",{"type":66,"tag":88,"props":1167,"children":1169},{"className":1168},[],[1170],{"type":72,"value":1171},"references\u002Flong-running-tasks.md",{"type":72,"value":1173},": delegate to a subagent \u002F\nbackground job and report compact status at least every 5\nminutes.",{"type":66,"tag":75,"props":1175,"children":1177},{"id":1176},"examples",[1178],{"type":72,"value":1179},"Examples",{"type":66,"tag":1181,"props":1182,"children":1184},"h3",{"id":1183},"example-1-end-to-end-ncore-usdz-render",[1185],{"type":72,"value":1186},"Example 1 — End-to-end NCore → USDZ → render",{"type":66,"tag":82,"props":1188,"children":1189},{},[1190,1192,1200,1202,1207],{"type":72,"value":1191},"Walk Workflow A in\n",{"type":66,"tag":509,"props":1193,"children":1194},{"href":1137},[1195],{"type":66,"tag":88,"props":1196,"children":1198},{"className":1197},[],[1199],{"type":72,"value":1137},{"type":72,"value":1201},": validate\nhost → generate aux data → train (cookbook recipe) → export →\nlocal render or ",{"type":66,"tag":88,"props":1203,"children":1205},{"className":1204},[],[1206],{"type":72,"value":114},{"type":72,"value":1208},". Concrete commands live in the\nreferenced files; this index does not duplicate them.",{"type":66,"tag":1181,"props":1210,"children":1212},{"id":1211},"example-2-skip-training-render-the-gated-physical-ai-dataset",[1213],{"type":72,"value":1214},"Example 2 — Skip training, render the gated Physical AI dataset",{"type":66,"tag":82,"props":1216,"children":1217},{},[1218,1220,1225,1227,1232,1234,1239],{"type":72,"value":1219},"Walk Workflow B: download\n",{"type":66,"tag":88,"props":1221,"children":1223},{"className":1222},[],[1224],{"type":72,"value":218},{"type":72,"value":1226}," from HuggingFace,\nthen jump to ",{"type":66,"tag":88,"props":1228,"children":1230},{"className":1229},[],[1231],{"type":72,"value":114},{"type":72,"value":1233}," + a Python client. Coordinate-frame\nconversion code is in ",{"type":66,"tag":88,"props":1235,"children":1237},{"className":1236},[],[1238],{"type":72,"value":1052},{"type":72,"value":156},{"type":66,"tag":1181,"props":1241,"children":1243},{"id":1242},"example-3-insert-asset-harvester-actors-into-a-usdz",[1244],{"type":72,"value":1245},"Example 3 — Insert Asset-Harvester actors into a USDZ",{"type":66,"tag":82,"props":1247,"children":1248},{},[1249,1251,1256,1258,1263,1265,1270,1272,1277,1279,1284,1286,1291],{"type":72,"value":1250},"Walk Workflow C: run ",{"type":66,"tag":88,"props":1252,"children":1254},{"className":1253},[],[1255],{"type":72,"value":313},{"type":72,"value":1257},", then\n",{"type":66,"tag":88,"props":1259,"children":1261},{"className":1260},[],[1262],{"type":72,"value":194},{"type":72,"value":1264},", edit ",{"type":66,"tag":88,"props":1266,"children":1268},{"className":1267},[],[1269],{"type":72,"value":1076},{"type":72,"value":1271},", and call\n",{"type":66,"tag":88,"props":1273,"children":1275},{"className":1274},[],[1276],{"type":72,"value":1091},{"type":72,"value":1278}," + ",{"type":66,"tag":88,"props":1280,"children":1282},{"className":1281},[],[1283],{"type":72,"value":206},{"type":72,"value":1285},". Schema lives in ",{"type":66,"tag":88,"props":1287,"children":1289},{"className":1288},[],[1290],{"type":72,"value":1099},{"type":72,"value":156},{"type":66,"tag":1181,"props":1293,"children":1295},{"id":1294},"example-4-warm-server-thin-client-for-repeated-rgb-renders",[1296],{"type":72,"value":1297},"Example 4 — Warm-server thin-client for repeated RGB renders",{"type":66,"tag":82,"props":1299,"children":1300},{},[1301,1303,1311,1313,1318,1320,1326,1328,1334],{"type":72,"value":1302},"Walk the warm-server quick start at the bottom of\n",{"type":66,"tag":509,"props":1304,"children":1305},{"href":1137},[1306],{"type":66,"tag":88,"props":1307,"children":1309},{"className":1308},[],[1310],{"type":72,"value":1137},{"type":72,"value":1312},". Boot\n",{"type":66,"tag":88,"props":1314,"children":1316},{"className":1315},[],[1317],{"type":72,"value":981},{"type":72,"value":1319},", render with\n",{"type":66,"tag":88,"props":1321,"children":1323},{"className":1322},[],[1324],{"type":72,"value":1325},"thin_client.py",{"type":72,"value":1327},", tear down with ",{"type":66,"tag":88,"props":1329,"children":1331},{"className":1330},[],[1332],{"type":72,"value":1333},"scripts\u002Fsession_teardown.sh",{"type":72,"value":156},{"type":66,"tag":75,"props":1336,"children":1338},{"id":1337},"backend-selection",[1339],{"type":72,"value":1340},"Backend Selection",{"type":66,"tag":82,"props":1342,"children":1343},{},[1344],{"type":72,"value":1345},"Pick the smallest backend that exposes the requested feature:",{"type":66,"tag":135,"props":1347,"children":1348},{},[1349,1402,1445],{"type":66,"tag":139,"props":1350,"children":1351},{},[1352,1357,1359,1365,1366,1371,1373,1379,1381,1386,1387,1393,1395,1401],{"type":66,"tag":127,"props":1353,"children":1354},{},[1355],{"type":72,"value":1356},"Local Docker, single command.",{"type":72,"value":1358}," Use ",{"type":66,"tag":88,"props":1360,"children":1362},{"className":1361},[],[1363],{"type":72,"value":1364},"nre render",{"type":72,"value":245},{"type":66,"tag":88,"props":1367,"children":1369},{"className":1368},[],[1370],{"type":72,"value":1021},{"type":72,"value":1372},",\nor an export sub-command directly. Simplest for one-off renders,\nLiDAR sweeps, actor edits, rolling shutter, in-container video\nexport, or exact ",{"type":66,"tag":88,"props":1374,"children":1376},{"className":1375},[],[1377],{"type":72,"value":1378},"--replicate-training-views",{"type":72,"value":1380}," behavior. See\n",{"type":66,"tag":88,"props":1382,"children":1384},{"className":1383},[],[1385],{"type":72,"value":957},{"type":72,"value":245},{"type":66,"tag":88,"props":1388,"children":1390},{"className":1389},[],[1391],{"type":72,"value":1392},"references\u002Fnre-image-notes.md",{"type":72,"value":1394},",\nand ",{"type":66,"tag":88,"props":1396,"children":1398},{"className":1397},[],[1399],{"type":72,"value":1400},"references\u002Fmp4-encoding.md",{"type":72,"value":156},{"type":66,"tag":139,"props":1403,"children":1404},{},[1405,1417,1419,1424,1426,1432,1433,1438,1439,1444],{"type":66,"tag":127,"props":1406,"children":1407},{},[1408,1410,1415],{"type":72,"value":1409},"Local Docker, warm ",{"type":66,"tag":88,"props":1411,"children":1413},{"className":1412},[],[1414],{"type":72,"value":114},{"type":72,"value":1416}," + thin host client.",{"type":72,"value":1418}," Use for\nrender-heavy RGB sessions where repeated Docker\u002FPython\u002FCUDA\ncold-start dominates latency, or where multiple cameras should\nbe rendered through one ",{"type":66,"tag":88,"props":1420,"children":1422},{"className":1421},[],[1423],{"type":72,"value":997},{"type":72,"value":1425}," RPC. See\n",{"type":66,"tag":88,"props":1427,"children":1429},{"className":1428},[],[1430],{"type":72,"value":1431},"references\u002FNRE_RenderClient\u002FREADME.md",{"type":72,"value":739},{"type":66,"tag":88,"props":1434,"children":1436},{"className":1435},[],[1437],{"type":72,"value":981},{"type":72,"value":475},{"type":66,"tag":88,"props":1440,"children":1442},{"className":1441},[],[1443],{"type":72,"value":1333},{"type":72,"value":156},{"type":66,"tag":139,"props":1446,"children":1447},{},[1448,1453,1455,1461,1463,1469,1471,1476],{"type":66,"tag":127,"props":1449,"children":1450},{},[1451],{"type":72,"value":1452},"OSMO \u002F cluster workflows.",{"type":72,"value":1454}," Use the templates under\n",{"type":66,"tag":88,"props":1456,"children":1458},{"className":1457},[],[1459],{"type":72,"value":1460},"references\u002Fexample-workflows\u002Fosmo\u002F",{"type":72,"value":1462}," for multi-clip fan-out,\nisolation from the local machine, or training jobs that should\nnot run on the user's workstation. Follow\n",{"type":66,"tag":88,"props":1464,"children":1466},{"className":1465},[],[1467],{"type":72,"value":1468},"references\u002Fngc-and-registry.md",{"type":72,"value":1470}," for registry credentials and\n",{"type":66,"tag":88,"props":1472,"children":1474},{"className":1473},[],[1475],{"type":72,"value":1171},{"type":72,"value":1477}," for polling discipline.",{"type":66,"tag":75,"props":1479,"children":1481},{"id":1480},"output-format",[1482],{"type":72,"value":1483},"Output Format",{"type":66,"tag":82,"props":1485,"children":1486},{},[1487,1489,1495],{"type":72,"value":1488},"Structured deliverables placed under ",{"type":66,"tag":88,"props":1490,"children":1492},{"className":1491},[],[1493],{"type":72,"value":1494},"${output_dir}\u002F${RUN_ID}\u002F",{"type":72,"value":1496},"\nby the NRE container (no JSON state file required from the\nagent):",{"type":66,"tag":135,"props":1498,"children":1499},{},[1500,1511,1522,1540,1558,1648],{"type":66,"tag":139,"props":1501,"children":1502},{},[1503,1509],{"type":66,"tag":88,"props":1504,"children":1506},{"className":1505},[],[1507],{"type":72,"value":1508},"config\u002Fparsed.yaml",{"type":72,"value":1510}," — Hydra-resolved training config.",{"type":66,"tag":139,"props":1512,"children":1513},{},[1514,1520],{"type":66,"tag":88,"props":1515,"children":1517},{"className":1516},[],[1518],{"type":72,"value":1519},"checkpoints\u002Flast.ckpt",{"type":72,"value":1521}," (plus periodic snapshots).",{"type":66,"tag":139,"props":1523,"children":1524},{},[1525,1531,1533,1539],{"type":66,"tag":88,"props":1526,"children":1528},{"className":1527},[],[1529],{"type":72,"value":1530},"val\u002Fmetrics.yaml",{"type":72,"value":1532}," — per-frame PSNR \u002F SSIM \u002F LPIPS under\n",{"type":66,"tag":88,"props":1534,"children":1536},{"className":1535},[],[1537],{"type":72,"value":1538},"test\u002F*",{"type":72,"value":156},{"type":66,"tag":139,"props":1541,"children":1542},{},[1543,1549,1550,1556],{"type":66,"tag":88,"props":1544,"children":1546},{"className":1545},[],[1547],{"type":72,"value":1548},"val\u002F*.mp4",{"type":72,"value":245},{"type":66,"tag":88,"props":1551,"children":1553},{"className":1552},[],[1554],{"type":72,"value":1555},"val\u002F\u003Cframe>\u002F*.png",{"type":72,"value":1557}," — depth, opacity, segmentation,\nRGB visualisations.",{"type":66,"tag":139,"props":1559,"children":1560},{},[1561,1567,1569,1575,1576,1582,1583,1589,1590,1596,1597,1603,1605,1611,1613,1619,1621,1626,1627,1632,1633,1638,1640,1646],{"type":66,"tag":88,"props":1562,"children":1564},{"className":1563},[],[1565],{"type":72,"value":1566},"usd-out\u002Flast.usdz",{"type":72,"value":1568}," — USDZ containing the trained reconstruction,\n",{"type":66,"tag":88,"props":1570,"children":1572},{"className":1571},[],[1573],{"type":72,"value":1574},"data_info.json",{"type":72,"value":245},{"type":66,"tag":88,"props":1577,"children":1579},{"className":1578},[],[1580],{"type":72,"value":1581},"rig_trajectories.json",{"type":72,"value":95},{"type":66,"tag":88,"props":1584,"children":1586},{"className":1585},[],[1587],{"type":72,"value":1588},"sequence_tracks.json",{"type":72,"value":245},{"type":66,"tag":88,"props":1591,"children":1593},{"className":1592},[],[1594],{"type":72,"value":1595},"parsed_config.yaml",{"type":72,"value":245},{"type":66,"tag":88,"props":1598,"children":1600},{"className":1599},[],[1601],{"type":72,"value":1602},"checkpoint.ckpt",{"type":72,"value":1604},",\noptional ",{"type":66,"tag":88,"props":1606,"children":1608},{"className":1607},[],[1609],{"type":72,"value":1610},"mesh.ply",{"type":72,"value":1612},", and ",{"type":66,"tag":88,"props":1614,"children":1616},{"className":1615},[],[1617],{"type":72,"value":1618},"map.xodr",{"type":72,"value":1620},". Render with ",{"type":66,"tag":88,"props":1622,"children":1624},{"className":1623},[],[1625],{"type":72,"value":1364},{"type":72,"value":95},{"type":66,"tag":88,"props":1628,"children":1630},{"className":1629},[],[1631],{"type":72,"value":114},{"type":72,"value":1278},{"type":66,"tag":88,"props":1634,"children":1636},{"className":1635},[],[1637],{"type":72,"value":1021},{"type":72,"value":1639},", the in-container ",{"type":66,"tag":88,"props":1641,"children":1643},{"className":1642},[],[1644],{"type":72,"value":1645},"viewer",{"type":72,"value":1647},", or hand\nto a downstream simulator (CARLA, AlpaSim, Isaac Sim).",{"type":66,"tag":139,"props":1649,"children":1650},{},[1651,1657,1658,1664,1665,1671,1672,1677,1679,1685,1686,1691,1692,1698],{"type":66,"tag":88,"props":1652,"children":1654},{"className":1653},[],[1655],{"type":72,"value":1656},"*.ply",{"type":72,"value":475},{"type":66,"tag":88,"props":1659,"children":1661},{"className":1660},[],[1662],{"type":72,"value":1663},"ego_mask\u002F*",{"type":72,"value":475},{"type":66,"tag":88,"props":1666,"children":1668},{"className":1667},[],[1669],{"type":72,"value":1670},"depth\u002F*",{"type":72,"value":475},{"type":66,"tag":88,"props":1673,"children":1675},{"className":1674},[],[1676],{"type":72,"value":1588},{"type":72,"value":1678}," \u002F\n",{"type":66,"tag":88,"props":1680,"children":1682},{"className":1681},[],[1683],{"type":72,"value":1684},"ncore_tracks.json",{"type":72,"value":475},{"type":66,"tag":88,"props":1687,"children":1689},{"className":1688},[],[1690],{"type":72,"value":1610},{"type":72,"value":475},{"type":66,"tag":88,"props":1693,"children":1695},{"className":1694},[],[1696],{"type":72,"value":1697},"ground_mesh.ply",{"type":72,"value":1699}," — produced\nby the matching export sub-command.",{"type":66,"tag":75,"props":1701,"children":1703},{"id":1702},"scripts",[1704],{"type":72,"value":1705},"Scripts",{"type":66,"tag":1707,"props":1708,"children":1709},"table",{},[1710,1733],{"type":66,"tag":1711,"props":1712,"children":1713},"thead",{},[1714],{"type":66,"tag":1715,"props":1716,"children":1717},"tr",{},[1718,1724,1728],{"type":66,"tag":1719,"props":1720,"children":1721},"th",{},[1722],{"type":72,"value":1723},"Script",{"type":66,"tag":1719,"props":1725,"children":1726},{},[1727],{"type":72,"value":80},{"type":66,"tag":1719,"props":1729,"children":1730},{},[1731],{"type":72,"value":1732},"Usage",{"type":66,"tag":1734,"props":1735,"children":1736},"tbody",{},[1737,1775,1808],{"type":66,"tag":1715,"props":1738,"children":1739},{},[1740,1749,1761],{"type":66,"tag":1741,"props":1742,"children":1743},"td",{},[1744],{"type":66,"tag":88,"props":1745,"children":1747},{"className":1746},[],[1748],{"type":72,"value":686},{"type":66,"tag":1741,"props":1750,"children":1751},{},[1752,1754,1759],{"type":72,"value":1753},"Verify Docker, NVIDIA Container Toolkit, GPU\u002Fdriver R570+ (R535+ minimum), NGC login, and ",{"type":66,"tag":88,"props":1755,"children":1757},{"className":1756},[],[1758],{"type":72,"value":650},{"type":72,"value":1760}," env var. No network calls.",{"type":66,"tag":1741,"props":1762,"children":1763},{},[1764,1769,1770],{"type":66,"tag":88,"props":1765,"children":1767},{"className":1766},[],[1768],{"type":72,"value":694},{"type":72,"value":1153},{"type":66,"tag":88,"props":1771,"children":1773},{"className":1772},[],[1774],{"type":72,"value":702},{"type":66,"tag":1715,"props":1776,"children":1777},{},[1778,1786,1799],{"type":66,"tag":1741,"props":1779,"children":1780},{},[1781],{"type":66,"tag":88,"props":1782,"children":1784},{"className":1783},[],[1785],{"type":72,"value":981},{"type":66,"tag":1741,"props":1787,"children":1788},{},[1789,1791,1797],{"type":72,"value":1790},"Idempotently boot a session-scoped ",{"type":66,"tag":88,"props":1792,"children":1794},{"className":1793},[],[1795],{"type":72,"value":1796},"nre serve-grpc",{"type":72,"value":1798}," container for the thin Python client. Discovers a cached 26.04+ renderer image, mounts the USDZ root, waits for readiness.",{"type":66,"tag":1741,"props":1800,"children":1801},{},[1802],{"type":66,"tag":88,"props":1803,"children":1805},{"className":1804},[],[1806],{"type":72,"value":1807},"NRE_GRPC_USDZ_HOST_DIR=\u002Fpath\u002Fto\u002Fusdz\u002Froot bash scripts\u002Fsession_warm_server.sh",{"type":66,"tag":1715,"props":1809,"children":1810},{},[1811,1819,1831],{"type":66,"tag":1741,"props":1812,"children":1813},{},[1814],{"type":66,"tag":88,"props":1815,"children":1817},{"className":1816},[],[1818],{"type":72,"value":1333},{"type":66,"tag":1741,"props":1820,"children":1821},{},[1822,1824,1829],{"type":72,"value":1823},"Stop and remove the warm ",{"type":66,"tag":88,"props":1825,"children":1827},{"className":1826},[],[1828],{"type":72,"value":114},{"type":72,"value":1830}," container and clear its state file without racing the next boot.",{"type":66,"tag":1741,"props":1832,"children":1833},{},[1834],{"type":66,"tag":88,"props":1835,"children":1837},{"className":1836},[],[1838],{"type":72,"value":1839},"bash scripts\u002Fsession_teardown.sh",{"type":66,"tag":75,"props":1841,"children":1843},{"id":1842},"references",[1844],{"type":72,"value":1845},"References",{"type":66,"tag":82,"props":1847,"children":1848},{},[1849,1851,1857],{"type":72,"value":1850},"Read these on demand; keep ",{"type":66,"tag":88,"props":1852,"children":1854},{"className":1853},[],[1855],{"type":72,"value":1856},"SKILL.md",{"type":72,"value":1858}," as the routing layer.",{"type":66,"tag":135,"props":1860,"children":1861},{},[1862,1895,1930,1943,1980,1994,2104,2114,2130,2155,2165,2175,2185,2195,2205,2227,2237,2248,2273,2291,2301,2312,2322,2332,2342],{"type":66,"tag":139,"props":1863,"children":1864},{},[1865,1873,1874,1880,1882,1887,1888,1894],{"type":66,"tag":509,"props":1866,"children":1867},{"href":724},[1868],{"type":66,"tag":88,"props":1869,"children":1871},{"className":1870},[],[1872],{"type":72,"value":724},{"type":72,"value":613},{"type":66,"tag":88,"props":1875,"children":1877},{"className":1876},[],[1878],{"type":72,"value":1879},"docker login nvcr.io",{"type":72,"value":1881},", image pull, full prerequisite matrix, and safe\nsecret-handling for ",{"type":66,"tag":88,"props":1883,"children":1885},{"className":1884},[],[1886],{"type":72,"value":650},{"type":72,"value":475},{"type":66,"tag":88,"props":1889,"children":1891},{"className":1890},[],[1892],{"type":72,"value":1893},"HF_TOKEN",{"type":72,"value":156},{"type":66,"tag":139,"props":1896,"children":1897},{},[1898,1906,1908,1914,1916,1921,1923,1929],{"type":66,"tag":509,"props":1899,"children":1900},{"href":795},[1901],{"type":66,"tag":88,"props":1902,"children":1904},{"className":1903},[],[1905],{"type":72,"value":795},{"type":72,"value":1907}," — most-used\n",{"type":66,"tag":88,"props":1909,"children":1911},{"className":1910},[],[1912],{"type":72,"value":1913},"docker run",{"type":72,"value":1915}," invocations: train + validate, re-validate with\nshift, local render at quarter or native res, ",{"type":66,"tag":88,"props":1917,"children":1919},{"className":1918},[],[1920],{"type":72,"value":114},{"type":72,"value":1922}," boot,\nLiDAR sweep, in-container ",{"type":66,"tag":88,"props":1924,"children":1926},{"className":1925},[],[1927],{"type":72,"value":1928},"--help",{"type":72,"value":156},{"type":66,"tag":139,"props":1931,"children":1932},{},[1933,1941],{"type":66,"tag":509,"props":1934,"children":1935},{"href":1137},[1936],{"type":66,"tag":88,"props":1937,"children":1939},{"className":1938},[],[1940],{"type":72,"value":1137},{"type":72,"value":1942}," — workflows\nA – I end-to-end, plus the warm-server thin-client quick start.",{"type":66,"tag":139,"props":1944,"children":1945},{},[1946,1955,1957,1963,1964,1970,1972,1978],{"type":66,"tag":509,"props":1947,"children":1949},{"href":1948},"references\u002Ftroubleshooting.md",[1950],{"type":66,"tag":88,"props":1951,"children":1953},{"className":1952},[],[1954],{"type":72,"value":1948},{"type":72,"value":1956},"\n— extended error matrix (",{"type":66,"tag":88,"props":1958,"children":1960},{"className":1959},[],[1961],{"type":72,"value":1962},"OOM",{"type":72,"value":245},{"type":66,"tag":88,"props":1965,"children":1967},{"className":1966},[],[1968],{"type":72,"value":1969},"wandb",{"type":72,"value":1971}," blocking,\n",{"type":66,"tag":88,"props":1973,"children":1975},{"className":1974},[],[1976],{"type":72,"value":1977},"--artifact-glob",{"type":72,"value":1979}," mismatches, deprecated flags, gRPC LiDAR size,\netc.).",{"type":66,"tag":139,"props":1981,"children":1982},{},[1983,1992],{"type":66,"tag":509,"props":1984,"children":1986},{"href":1985},"references\u002Fteardown.md",[1987],{"type":66,"tag":88,"props":1988,"children":1990},{"className":1989},[],[1991],{"type":72,"value":1985},{"type":72,"value":1993}," — disk\ncleanup, post-teardown verification, ownership-recovery.",{"type":66,"tag":139,"props":1995,"children":1996},{},[1997,2002,2004,2010,2011,2016,2017,2022,2023,2029,2031,2037,2038,2044,2045,2050,2051,2056,2057,2062,2063,2068,2069,2074,2075,2081,2082,2088,2089,2095,2097,2102],{"type":66,"tag":88,"props":1998,"children":2000},{"className":1999},[],[2001],{"type":72,"value":914},{"type":72,"value":2003}," — full sub-command surface of the\nNRE container (training, validation, ",{"type":66,"tag":88,"props":2005,"children":2007},{"className":2006},[],[2008],{"type":72,"value":2009},"render",{"type":72,"value":245},{"type":66,"tag":88,"props":2012,"children":2014},{"className":2013},[],[2015],{"type":72,"value":114},{"type":72,"value":95},{"type":66,"tag":88,"props":2018,"children":2020},{"className":2019},[],[2021],{"type":72,"value":1021},{"type":72,"value":245},{"type":66,"tag":88,"props":2024,"children":2026},{"className":2025},[],[2027],{"type":72,"value":2028},"render-novel-trajectory",{"type":72,"value":2030},", every ",{"type":66,"tag":88,"props":2032,"children":2034},{"className":2033},[],[2035],{"type":72,"value":2036},"export-*",{"type":72,"value":95},{"type":66,"tag":88,"props":2039,"children":2041},{"className":2040},[],[2042],{"type":72,"value":2043},"upgrade-config",{"type":72,"value":475},{"type":66,"tag":88,"props":2046,"children":2048},{"className":2047},[],[2049],{"type":72,"value":230},{"type":72,"value":245},{"type":66,"tag":88,"props":2052,"children":2054},{"className":2053},[],[2055],{"type":72,"value":251},{"type":72,"value":95},{"type":66,"tag":88,"props":2058,"children":2060},{"className":2059},[],[2061],{"type":72,"value":258},{"type":72,"value":245},{"type":66,"tag":88,"props":2064,"children":2066},{"className":2065},[],[2067],{"type":72,"value":265},{"type":72,"value":245},{"type":66,"tag":88,"props":2070,"children":2072},{"className":2071},[],[2073],{"type":72,"value":1645},{"type":72,"value":95},{"type":66,"tag":88,"props":2076,"children":2078},{"className":2077},[],[2079],{"type":72,"value":2080},"ply_viewer",{"type":72,"value":245},{"type":66,"tag":88,"props":2083,"children":2085},{"className":2084},[],[2086],{"type":72,"value":2087},"profile-dataloader",{"type":72,"value":245},{"type":66,"tag":88,"props":2090,"children":2092},{"className":2091},[],[2093],{"type":72,"value":2094},"run-script",{"type":72,"value":2096},", the\n",{"type":66,"tag":88,"props":2098,"children":2100},{"className":2099},[],[2101],{"type":72,"value":154},{"type":72,"value":2103}," aux-data + AH entry points).",{"type":66,"tag":139,"props":2105,"children":2106},{},[2107,2112],{"type":66,"tag":88,"props":2108,"children":2110},{"className":2109},[],[2111],{"type":72,"value":590},{"type":72,"value":2113}," — Hydra recipe map for Waymo \u002F NV\n\u002F PandaSet \u002F Tesla \u002F Alpasim, plus override matrix.",{"type":66,"tag":139,"props":2115,"children":2116},{},[2117,2122,2123,2128],{"type":66,"tag":88,"props":2118,"children":2120},{"className":2119},[],[2121],{"type":72,"value":781},{"type":72,"value":613},{"type":66,"tag":88,"props":2124,"children":2126},{"className":2125},[],[2127],{"type":72,"value":154},{"type":72,"value":2129}," auxiliary-data CLI.",{"type":66,"tag":139,"props":2131,"children":2132},{},[2133,2138,2140,2146,2148,2154],{"type":66,"tag":88,"props":2134,"children":2136},{"className":2135},[],[2137],{"type":72,"value":957},{"type":72,"value":2139}," — host-side ",{"type":66,"tag":88,"props":2141,"children":2143},{"className":2142},[],[2144],{"type":72,"value":2145},"docker run … render",{"type":72,"value":2147},"\nrecipes for rig offsets and ",{"type":66,"tag":88,"props":2149,"children":2151},{"className":2150},[],[2152],{"type":72,"value":2153},"export-custom-rig-trajectory",{"type":72,"value":156},{"type":66,"tag":139,"props":2156,"children":2157},{},[2158,2163],{"type":66,"tag":88,"props":2159,"children":2161},{"className":2160},[],[2162],{"type":72,"value":1431},{"type":72,"value":2164}," — warm-server thin\nPython gRPC client.",{"type":66,"tag":139,"props":2166,"children":2167},{},[2168,2173],{"type":66,"tag":88,"props":2169,"children":2171},{"className":2170},[],[2172],{"type":72,"value":1045},{"type":72,"value":2174}," — sensorsim gRPC server flags + Python\nclient cookbook.",{"type":66,"tag":139,"props":2176,"children":2177},{},[2178,2183],{"type":66,"tag":88,"props":2179,"children":2181},{"className":2180},[],[2182],{"type":72,"value":1392},{"type":72,"value":2184}," — cached-image discovery, 26.04+\nvs 26.03 vs pre-26.03 flags.",{"type":66,"tag":139,"props":2186,"children":2187},{},[2188,2193],{"type":66,"tag":88,"props":2189,"children":2191},{"className":2190},[],[2192],{"type":72,"value":1468},{"type":72,"value":2194}," — NGC API key resolution.",{"type":66,"tag":139,"props":2196,"children":2197},{},[2198,2203],{"type":66,"tag":88,"props":2199,"children":2201},{"className":2200},[],[2202],{"type":72,"value":1400},{"type":72,"value":2204}," — host-side ffmpeg recipe.",{"type":66,"tag":139,"props":2206,"children":2207},{},[2208,2213,2214,2219,2220,2225],{"type":66,"tag":88,"props":2209,"children":2211},{"className":2210},[],[2212],{"type":72,"value":1099},{"type":72,"value":613},{"type":66,"tag":88,"props":2215,"children":2217},{"className":2216},[],[2218],{"type":72,"value":194},{"type":72,"value":1015},{"type":66,"tag":88,"props":2221,"children":2223},{"className":2222},[],[2224],{"type":72,"value":1076},{"type":72,"value":2226}," schema.",{"type":66,"tag":139,"props":2228,"children":2229},{},[2230,2235],{"type":66,"tag":88,"props":2231,"children":2233},{"className":2232},[],[2234],{"type":72,"value":1052},{"type":72,"value":2236}," — recipe for rendering the\nHuggingFace NuRec dataset.",{"type":66,"tag":139,"props":2238,"children":2239},{},[2240,2246],{"type":66,"tag":88,"props":2241,"children":2243},{"className":2242},[],[2244],{"type":72,"value":2245},"references\u002Fexample-workflows\u002F",{"type":72,"value":2247}," — bash, Hydra, and OSMO\ntemplates.",{"type":66,"tag":139,"props":2249,"children":2250},{},[2251,2257,2259,2265,2266,2272],{"type":66,"tag":88,"props":2252,"children":2254},{"className":2253},[],[2255],{"type":72,"value":2256},"references\u002Frig-json\u002F",{"type":72,"value":2258}," — bundled ",{"type":66,"tag":88,"props":2260,"children":2262},{"className":2261},[],[2263],{"type":72,"value":2264},"rig.json",{"type":72,"value":739},{"type":66,"tag":88,"props":2267,"children":2269},{"className":2268},[],[2270],{"type":72,"value":2271},"augmented_rig.json",{"type":72,"value":156},{"type":66,"tag":139,"props":2274,"children":2275},{},[2276,2282,2284,2289],{"type":66,"tag":88,"props":2277,"children":2279},{"className":2278},[],[2280],{"type":72,"value":2281},"references\u002Fcustom-rig-trajectories\u002F",{"type":72,"value":2283}," — pre-baked\n",{"type":66,"tag":88,"props":2285,"children":2287},{"className":2286},[],[2288],{"type":72,"value":2153},{"type":72,"value":2290}," outputs.",{"type":66,"tag":139,"props":2292,"children":2293},{},[2294,2299],{"type":66,"tag":88,"props":2295,"children":2297},{"className":2296},[],[2298],{"type":72,"value":1171},{"type":72,"value":2300}," — background-job + 5-minute\nstatus reporting convention.",{"type":66,"tag":139,"props":2302,"children":2303},{},[2304,2310],{"type":66,"tag":88,"props":2305,"children":2307},{"className":2306},[],[2308],{"type":72,"value":2309},"references\u002Fnurec-skill-catalog.md",{"type":72,"value":2311}," — routing table for sibling\nNuRec-stack skills.",{"type":66,"tag":139,"props":2313,"children":2314},{},[2315,2317],{"type":72,"value":2316},"Public product page: ",{"type":66,"tag":509,"props":2318,"children":2320},{"href":54,"rel":2319},[657],[2321],{"type":72,"value":54},{"type":66,"tag":139,"props":2323,"children":2324},{},[2325,2327],{"type":72,"value":2326},"HF dataset: ",{"type":66,"tag":509,"props":2328,"children":2330},{"href":58,"rel":2329},[657],[2331],{"type":72,"value":58},{"type":66,"tag":139,"props":2333,"children":2334},{},[2335,2337],{"type":72,"value":2336},"HF Fixer model: ",{"type":66,"tag":509,"props":2338,"children":2340},{"href":59,"rel":2339},[657],[2341],{"type":72,"value":59},{"type":66,"tag":139,"props":2343,"children":2344},{},[2345,2347],{"type":72,"value":2346},"NGC Fixer model card: ",{"type":66,"tag":509,"props":2348,"children":2350},{"href":57,"rel":2349},[657],[2351],{"type":72,"value":57},{"type":66,"tag":75,"props":2353,"children":2355},{"id":2354},"prerequisites",[2356],{"type":72,"value":2357},"Prerequisites",{"type":66,"tag":82,"props":2359,"children":2360},{},[2361,2363,2368,2370,2378,2380,2385],{"type":72,"value":2362},"Linux x86_64 + NVIDIA GPU + Docker 23+ + NVIDIA Container Toolkit\n1.13+ + ",{"type":66,"tag":88,"props":2364,"children":2366},{"className":2365},[],[2367],{"type":72,"value":650},{"type":72,"value":2369},". Full matrix (driver minimums per arch,\nshm-size, file ownership, GPU-tier guidance) lives in\n",{"type":66,"tag":509,"props":2371,"children":2372},{"href":724},[2373],{"type":66,"tag":88,"props":2374,"children":2376},{"className":2375},[],[2377],{"type":72,"value":724},{"type":72,"value":2379},". Always verify\nvia ",{"type":66,"tag":88,"props":2381,"children":2383},{"className":2382},[],[2384],{"type":72,"value":686},{"type":72,"value":2386}," before pulling the image.",{"type":66,"tag":75,"props":2388,"children":2390},{"id":2389},"limitations",[2391],{"type":72,"value":2392},"Limitations",{"type":66,"tag":135,"props":2394,"children":2395},{},[2396,2406,2416,2457,2487,2517,2527,2552,2585],{"type":66,"tag":139,"props":2397,"children":2398},{},[2399,2404],{"type":66,"tag":127,"props":2400,"children":2401},{},[2402],{"type":72,"value":2403},"Linux x86_64 only.",{"type":72,"value":2405}," aarch64 (e.g. Jetson) is not supported.",{"type":66,"tag":139,"props":2407,"children":2408},{},[2409,2414],{"type":66,"tag":127,"props":2410,"children":2411},{},[2412],{"type":72,"value":2413},"Internal source not redistributable.",{"type":72,"value":2415}," Use only the public\nNGC containers and the public NuRec docs.",{"type":66,"tag":139,"props":2417,"children":2418},{},[2419,2424,2426,2432,2433,2439,2441,2447,2449,2455],{"type":66,"tag":127,"props":2420,"children":2421},{},[2422],{"type":72,"value":2423},"Multi-GPU defaults are conservative.",{"type":72,"value":2425}," Set\n",{"type":66,"tag":88,"props":2427,"children":2429},{"className":2428},[],[2430],{"type":72,"value":2431},"trainer.world_size",{"type":72,"value":475},{"type":66,"tag":88,"props":2434,"children":2436},{"className":2435},[],[2437],{"type":72,"value":2438},"trainer.num_nodes",{"type":72,"value":2440}," explicitly to scale\nout; SLURM is auto-detected when both are ",{"type":66,"tag":88,"props":2442,"children":2444},{"className":2443},[],[2445],{"type":72,"value":2446},"0",{"type":72,"value":2448},". Quality plateaus\npast ~6 GPUs; per release notes, multi-GPU +\n",{"type":66,"tag":88,"props":2450,"children":2452},{"className":2451},[],[2453],{"type":72,"value":2454},"dataset.aux_data=false",{"type":72,"value":2456}," is a known crash combination.",{"type":66,"tag":139,"props":2458,"children":2459},{},[2460,2470,2472,2478,2480,2486],{"type":66,"tag":127,"props":2461,"children":2462},{},[2463,2468],{"type":66,"tag":88,"props":2464,"children":2466},{"className":2465},[],[2467],{"type":72,"value":809},{"type":72,"value":2469}," paths differ between train and val\u002Fexport.",{"type":72,"value":2471},"\nTraining uses container-bundled recipes; validation and exports\nre-pass the ",{"type":66,"tag":88,"props":2473,"children":2475},{"className":2474},[],[2476],{"type":72,"value":2477},"parsed.yaml",{"type":72,"value":2479}," written under\n",{"type":66,"tag":88,"props":2481,"children":2483},{"className":2482},[],[2484],{"type":72,"value":2485},"\u003Coutput_dir>\u002F\u003CRUN-ID>\u002Fconfig\u002F",{"type":72,"value":156},{"type":66,"tag":139,"props":2488,"children":2489},{},[2490,2501,2503,2509,2510,2516],{"type":66,"tag":127,"props":2491,"children":2492},{},[2493,2495,2500],{"type":72,"value":2494},"Validation may prompt for ",{"type":66,"tag":88,"props":2496,"children":2498},{"className":2497},[],[2499],{"type":72,"value":1969},{"type":72,"value":156},{"type":72,"value":2502}," Choose option 3 to skip\nin non-interactive runs, or pass ",{"type":66,"tag":88,"props":2504,"children":2506},{"className":2505},[],[2507],{"type":72,"value":2508},"logger=tensorboard",{"type":72,"value":1678},{"type":66,"tag":88,"props":2511,"children":2513},{"className":2512},[],[2514],{"type":72,"value":2515},"logger=dummy",{"type":72,"value":156},{"type":66,"tag":139,"props":2518,"children":2519},{},[2520,2525],{"type":66,"tag":127,"props":2521,"children":2522},{},[2523],{"type":72,"value":2524},"Render gRPC is data-format-pinned.",{"type":72,"value":2526}," Older releases warn \u002F\nreject artifacts that pre-date them; check release notes when\nmixing client \u002F server versions.",{"type":66,"tag":139,"props":2528,"children":2529},{},[2530,2535,2537,2542,2544,2550],{"type":66,"tag":127,"props":2531,"children":2532},{},[2533],{"type":72,"value":2534},"Asset-Harvester input only.",{"type":72,"value":2536}," ",{"type":66,"tag":88,"props":2538,"children":2540},{"className":2539},[],[2541],{"type":72,"value":194},{"type":72,"value":2543},"\nrequires AH outputs; raw ",{"type":66,"tag":88,"props":2545,"children":2547},{"className":2546},[],[2548],{"type":72,"value":2549},".ply",{"type":72,"value":2551}," files won't carry the\nper-asset cuboid metadata.",{"type":66,"tag":139,"props":2553,"children":2554},{},[2555,2560,2562,2568,2570,2576,2578,2583],{"type":66,"tag":127,"props":2556,"children":2557},{},[2558],{"type":72,"value":2559},"Difix variants are pluggable.",{"type":72,"value":2561}," The container ships both the\nCosmos Difix variant (default since 25.09 —\n",{"type":66,"tag":88,"props":2563,"children":2565},{"className":2564},[],[2566],{"type":72,"value":2567},"difix=cosmos_difix",{"type":72,"value":2569},") and the legacy Stable-Diffusion variant\n(",{"type":66,"tag":88,"props":2571,"children":2573},{"className":2572},[],[2574],{"type":72,"value":2575},"difix=sd_difix",{"type":72,"value":2577},"). The newer Cosmos-Predict-based Fixer\nvariants live in the ",{"type":66,"tag":88,"props":2579,"children":2581},{"className":2580},[],[2582],{"type":72,"value":340},{"type":72,"value":2584}," skill.",{"type":66,"tag":139,"props":2586,"children":2587},{},[2588,2605,2606,2611,2613,2618,2620,2625,2627,2632,2634,2639],{"type":66,"tag":127,"props":2589,"children":2590},{},[2591,2596,2598,2603],{"type":66,"tag":88,"props":2592,"children":2594},{"className":2593},[],[2595],{"type":72,"value":2009},{"type":72,"value":2597}," ↔ ",{"type":66,"tag":88,"props":2599,"children":2601},{"className":2600},[],[2602],{"type":72,"value":1021},{"type":72,"value":2604}," overlap.",{"type":72,"value":2536},{"type":66,"tag":88,"props":2607,"children":2609},{"className":2608},[],[2610],{"type":72,"value":2009},{"type":72,"value":2612}," runs\nin-container without a server; ",{"type":66,"tag":88,"props":2614,"children":2616},{"className":2615},[],[2617],{"type":72,"value":1021},{"type":72,"value":2619}," requires an active\n",{"type":66,"tag":88,"props":2621,"children":2623},{"className":2622},[],[2624],{"type":72,"value":114},{"type":72,"value":2626},". Use ",{"type":66,"tag":88,"props":2628,"children":2630},{"className":2629},[],[2631],{"type":72,"value":2009},{"type":72,"value":2633}," for batch novel-view jobs and\n",{"type":66,"tag":88,"props":2635,"children":2637},{"className":2636},[],[2638],{"type":72,"value":1021},{"type":72,"value":2640}," when you need actor editing, LiDAR rendering, or\na long-lived service.",{"type":66,"tag":75,"props":2642,"children":2644},{"id":2643},"troubleshooting-top-4",[2645],{"type":72,"value":2646},"Troubleshooting (top 4)",{"type":66,"tag":1707,"props":2648,"children":2649},{},[2650,2671],{"type":66,"tag":1711,"props":2651,"children":2652},{},[2653],{"type":66,"tag":1715,"props":2654,"children":2655},{},[2656,2661,2666],{"type":66,"tag":1719,"props":2657,"children":2658},{},[2659],{"type":72,"value":2660},"Error",{"type":66,"tag":1719,"props":2662,"children":2663},{},[2664],{"type":72,"value":2665},"Cause",{"type":66,"tag":1719,"props":2667,"children":2668},{},[2669],{"type":72,"value":2670},"Fix",{"type":66,"tag":1734,"props":2672,"children":2673},{},[2674,2708,2763,2806],{"type":66,"tag":1715,"props":2675,"children":2676},{},[2677,2686,2691],{"type":66,"tag":1741,"props":2678,"children":2679},{},[2680],{"type":66,"tag":88,"props":2681,"children":2683},{"className":2682},[],[2684],{"type":72,"value":2685},"Unable to find image 'nvcr.io\u002Fnvidia\u002Fnre\u002Fnre:latest'",{"type":66,"tag":1741,"props":2687,"children":2688},{},[2689],{"type":72,"value":2690},"Docker not authenticated to NGC.",{"type":66,"tag":1741,"props":2692,"children":2693},{},[2694,2699,2701,2707],{"type":66,"tag":88,"props":2695,"children":2697},{"className":2696},[],[2698],{"type":72,"value":1879},{"type":72,"value":2700}," with ",{"type":66,"tag":88,"props":2702,"children":2704},{"className":2703},[],[2705],{"type":72,"value":2706},"Username: $oauthtoken",{"type":72,"value":156},{"type":66,"tag":1715,"props":2709,"children":2710},{},[2711,2729,2734],{"type":66,"tag":1741,"props":2712,"children":2713},{},[2714,2720,2721,2727],{"type":66,"tag":88,"props":2715,"children":2717},{"className":2716},[],[2718],{"type":72,"value":2719},"OOM Killed",{"type":72,"value":475},{"type":66,"tag":88,"props":2722,"children":2724},{"className":2723},[],[2725],{"type":72,"value":2726},"CUDA out of memory",{"type":72,"value":2728}," during training",{"type":66,"tag":1741,"props":2730,"children":2731},{},[2732],{"type":72,"value":2733},"Default recipe needs >= 48 GB VRAM.",{"type":66,"tag":1741,"props":2735,"children":2736},{},[2737,2739,2745,2747,2753,2755,2761],{"type":72,"value":2738},"Reduce ",{"type":66,"tag":88,"props":2740,"children":2742},{"className":2741},[],[2743],{"type":72,"value":2744},"dataset.camera_ids",{"type":72,"value":2746},", lower ",{"type":66,"tag":88,"props":2748,"children":2750},{"className":2749},[],[2751],{"type":72,"value":2752},"trainer.max_epochs",{"type":72,"value":2754},", switch to ",{"type":66,"tag":88,"props":2756,"children":2758},{"className":2757},[],[2759],{"type":72,"value":2760},"trainer.precision=16-mixed",{"type":72,"value":2762},", or use a 48 GB+ GPU.",{"type":66,"tag":1715,"props":2764,"children":2765},{},[2766,2776,2794],{"type":66,"tag":1741,"props":2767,"children":2768},{},[2769,2774],{"type":66,"tag":88,"props":2770,"children":2772},{"className":2771},[],[2773],{"type":72,"value":114},{"type":72,"value":2775}," fails to find the USDZ",{"type":66,"tag":1741,"props":2777,"children":2778},{},[2779,2784,2786,2792],{"type":66,"tag":88,"props":2780,"children":2782},{"className":2781},[],[2783],{"type":72,"value":1977},{"type":72,"value":2785}," must end in ",{"type":66,"tag":88,"props":2787,"children":2789},{"className":2788},[],[2790],{"type":72,"value":2791},".usdz",{"type":72,"value":2793}," and be quoted.",{"type":66,"tag":1741,"props":2795,"children":2796},{},[2797,2799,2805],{"type":72,"value":2798},"Use e.g. ",{"type":66,"tag":88,"props":2800,"children":2802},{"className":2801},[],[2803],{"type":72,"value":2804},"--artifact-glob \u002Fworkdir\u002Foutput\u002F\u003CRUN-ID>\u002Fusd-out\u002Flast.usdz",{"type":72,"value":156},{"type":66,"tag":1715,"props":2807,"children":2808},{},[2809,2819,2831],{"type":66,"tag":1741,"props":2810,"children":2811},{},[2812,2814],{"type":72,"value":2813},"Edits silently ignored from ",{"type":66,"tag":88,"props":2815,"children":2817},{"className":2816},[],[2818],{"type":72,"value":206},{"type":66,"tag":1741,"props":2820,"children":2821},{},[2822,2824,2830],{"type":72,"value":2823},"Server started without ",{"type":66,"tag":88,"props":2825,"children":2827},{"className":2826},[],[2828],{"type":72,"value":2829},"--enable-editing-actors",{"type":72,"value":156},{"type":66,"tag":1741,"props":2832,"children":2833},{},[2834,2836,2841],{"type":72,"value":2835},"Restart ",{"type":66,"tag":88,"props":2837,"children":2839},{"className":2838},[],[2840],{"type":72,"value":114},{"type":72,"value":2842}," with that flag.",{"type":66,"tag":82,"props":2844,"children":2845},{},[2846,2848,2856],{"type":72,"value":2847},"Full matrix in\n",{"type":66,"tag":509,"props":2849,"children":2850},{"href":1948},[2851],{"type":66,"tag":88,"props":2852,"children":2854},{"className":2853},[],[2855],{"type":72,"value":1948},{"type":72,"value":156},{"type":66,"tag":75,"props":2858,"children":2860},{"id":2859},"teardown",[2861],{"type":72,"value":2862},"Teardown",{"type":66,"tag":82,"props":2864,"children":2865},{},[2866,2868,2876,2878,2883,2885,2891,2893,2899,2901,2907,2909,2914],{"type":72,"value":2867},"Full inventory, ownership-recovery, and post-teardown verification\ncommands live in ",{"type":66,"tag":509,"props":2869,"children":2870},{"href":1985},[2871],{"type":66,"tag":88,"props":2872,"children":2874},{"className":2873},[],[2875],{"type":72,"value":1985},{"type":72,"value":2877},".\nHeadline: stop ",{"type":66,"tag":88,"props":2879,"children":2881},{"className":2880},[],[2882],{"type":72,"value":114},{"type":72,"value":2884}," containers, ",{"type":66,"tag":88,"props":2886,"children":2888},{"className":2887},[],[2889],{"type":72,"value":2890},"docker image rm",{"type":72,"value":2892},"\nnre\u002Fnre-tools, ",{"type":66,"tag":88,"props":2894,"children":2896},{"className":2895},[],[2897],{"type":72,"value":2898},"rm -rf ${HOME}\u002F.cache\u002Fnre",{"type":72,"value":2900}," and your\n",{"type":66,"tag":88,"props":2902,"children":2904},{"className":2903},[],[2905],{"type":72,"value":2906},"\u003Coutput_dir>\u002F\u003CRUN-ID>\u002F",{"type":72,"value":2908},". Do not revoke ",{"type":66,"tag":88,"props":2910,"children":2912},{"className":2911},[],[2913],{"type":72,"value":650},{"type":72,"value":2915}," unless you\nsuspect it has been leaked.",{"items":2917,"total":2987},[2918,2930,2944,2951,2961,2971],{"slug":313,"name":313,"fn":2919,"description":2920,"org":2921,"tags":2922,"stars":23,"repoUrl":24,"updatedAt":2929},"extract 3D Gaussian assets with Asset Harvester","Use to install and run NVIDIA Asset Harvester (Apache-2.0) to extract per-object 3D Gaussian Splat assets (`gaussians.ply`) from AV NCore V4 clips or masked single images via SparseViewDiT + TokenGS, optionally producing `metadata.yaml` for NuRec object insertion. Do NOT use for full-scene reconstruction (use `nre`) or for inputs without per-object masks.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2923,2924,2927,2928],{"name":17,"slug":18,"type":15},{"name":2925,"slug":2926,"type":15},"Computer Vision","computer-vision",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"2026-07-14T05:32:37.216976",{"slug":300,"name":300,"fn":2931,"description":2932,"org":2933,"tags":2934,"stars":23,"repoUrl":24,"updatedAt":2943},"convert sensor datasets to NVIDIA NCore format","Use when converting any sensor dataset into NVIDIA NCore V4 format (and feeding it to NuRec or a robotics-to-sim \"r2s\" pipeline). Covers ingesting raw cameras, LiDARs, radars, IMUs, depth or stereo into V4 sequences; authoring a new converter from the template; adapting PAI \u002F Waymo \u002F PandaSet \u002F NuScenes to V4; handling non-AV rigs (mono+depth, mono+lidar, stereo, multi-stereo, RGB-D, COLMAP \u002F SfM, ROS2 bag); and diagnosing a broken converter against `validate.py`. Do NOT use to train reconstructions (use `nre`) or to extract per-object 3D assets (use `asset-harvester`). Trigger keywords: ncore, ncore v4, convert, ingest, zarr, itar, nurec, waymo, pandaset, nuscenes, pai, hyperion, colmap, scannetpp, stereo, multi-stereo, mono+depth, mono+lidar, kitti, sfm, camera, lidar, radar, imu, cuboid, poses, intrinsics, ego mask, ros2, rosbag, mcap, realsense, zed, rgb-d, r2s, robotics, sam2.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2935,2938,2939,2942],{"name":2936,"slug":2937,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":2940,"slug":2941,"type":15},"Robotics","robotics",{"name":21,"slug":22,"type":15},"2026-07-14T05:32:35.853952",{"slug":4,"name":4,"fn":5,"description":6,"org":2945,"tags":2946,"stars":23,"repoUrl":24,"updatedAt":25},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2947,2948,2949,2950],{"name":17,"slug":18,"type":15},{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},{"slug":340,"name":340,"fn":2952,"description":2953,"org":2954,"tags":2955,"stars":23,"repoUrl":24,"updatedAt":2960},"harmonize and fine-tune 3D reconstruction frames","Use to run NVIDIA DiffusionHarmonizer (public successor to the older Fixer recipes) to enhance, harmonize, evaluate, or fine-tune novel-view frames from NRE \u002F NuRec \u002F 3DGS \u002F NeRF reconstructions. Do NOT use for training the 3D reconstruction itself (use `nre`) or for sensor-to-NCore conversion (use `ncore`).",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2956,2957,2958,2959],{"name":17,"slug":18,"type":15},{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"2026-07-14T05:32:38.465728",{"slug":2962,"name":2962,"fn":2963,"description":2964,"org":2965,"tags":2966,"stars":23,"repoUrl":24,"updatedAt":2970},"nurec-index","route NVIDIA NuRec and asset tasks","Router for NVIDIA NuRec \u002F NRE \u002F 3DGUT \u002F USDZ \u002F NCore V4 \u002F asset harvest \u002F frame cleanup tasks — picks the right sibling (nre, ncore, asset-harvester, nurec-fixer, physical-ai-datasets). Use when the sub-skill is unclear or a multi-stage pipeline is needed; do NOT use for non-NuRec tasks or to run any pipeline itself.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2967,2968,2969],{"name":17,"slug":18,"type":15},{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"2026-07-14T05:32:31.902229",{"slug":2972,"name":2972,"fn":2973,"description":2974,"org":2975,"tags":2976,"stars":23,"repoUrl":24,"updatedAt":2986},"physical-ai-datasets","find and download NVIDIA Physical AI datasets","Use when the user wants to find, download, or pick a NVIDIA Physical AI dataset on Hugging Face for autonomous-vehicle, robotics, spatial intelligence, manipulation, or neural-reconstruction workflows. Catalog of every dataset under huggingface.co\u002Fnvidia with the `PhysicalAI-` prefix, organised by domain (AV, Robotics-Manipulation, Robotics-GR00T, Robotics-mindmap, Robotics-NuRec, Spatial Intelligence, Grasping, Healthcare, Sim-Ready, Material properties), with per-dataset size, format, gating, license, and the downstream sibling skill (`ncore`, `nre`, `asset-harvester`, `nurec-fixer`) or upstream tool (Isaac Sim, CARLA, Isaac-GR00T, Cosmos-*) that consumes it. Do NOT use as a runtime — it routes you elsewhere. Trigger keywords: nvidia physical ai dataset, PhysicalAI- dataset, hf nvidia dataset, NCore dataset, NuRec dataset, GR00T dataset, GraspGen, SimReady, Cosmos-Drive-Dreams, Lyra SDG, Open-H-Embodiment, huggingface-cli download, physical_ai_av, dataset gated, RDS-HQ.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[2977,2980,2983,2984,2985],{"name":2978,"slug":2979,"type":15},"Datasets","datasets",{"name":2981,"slug":2982,"type":15},"Hugging Face","hugging-face",{"name":13,"slug":14,"type":15},{"name":9,"slug":8,"type":15},{"name":2940,"slug":2941,"type":15},"2026-07-14T05:32:33.182445",6,{"items":2989,"total":3146},[2990,3008,3026,3037,3049,3063,3076,3088,3101,3112,3126,3135],{"slug":2991,"name":2991,"fn":2992,"description":2993,"org":2994,"tags":2995,"stars":3005,"repoUrl":3006,"updatedAt":3007},"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},[2996,2999,3002],{"name":2997,"slug":2998,"type":15},"Documentation","documentation",{"name":3000,"slug":3001,"type":15},"MCP","mcp",{"name":3003,"slug":3004,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":3009,"name":3009,"fn":3010,"description":3011,"org":3012,"tags":3013,"stars":3023,"repoUrl":3024,"updatedAt":3025},"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},[3014,3017,3020],{"name":3015,"slug":3016,"type":15},"Containers","containers",{"name":3018,"slug":3019,"type":15},"Deployment","deployment",{"name":3021,"slug":3022,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":3027,"name":3027,"fn":3028,"description":3029,"org":3030,"tags":3031,"stars":3023,"repoUrl":3024,"updatedAt":3036},"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},[3032,3035],{"name":3033,"slug":3034,"type":15},"CI\u002FCD","ci-cd",{"name":3018,"slug":3019,"type":15},"2026-07-14T05:25:59.97109",{"slug":3038,"name":3038,"fn":3039,"description":3040,"org":3041,"tags":3042,"stars":3023,"repoUrl":3024,"updatedAt":3048},"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},[3043,3044,3045],{"name":3033,"slug":3034,"type":15},{"name":3018,"slug":3019,"type":15},{"name":3046,"slug":3047,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":3050,"name":3050,"fn":3051,"description":3052,"org":3053,"tags":3054,"stars":3023,"repoUrl":3024,"updatedAt":3062},"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},[3055,3058,3059],{"name":3056,"slug":3057,"type":15},"Debugging","debugging",{"name":3046,"slug":3047,"type":15},{"name":3060,"slug":3061,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":3064,"name":3064,"fn":3065,"description":3066,"org":3067,"tags":3068,"stars":3023,"repoUrl":3024,"updatedAt":3075},"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},[3069,3072],{"name":3070,"slug":3071,"type":15},"Best Practices","best-practices",{"name":3073,"slug":3074,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":3077,"name":3077,"fn":3078,"description":3079,"org":3080,"tags":3081,"stars":3023,"repoUrl":3024,"updatedAt":3087},"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},[3082,3083,3086],{"name":13,"slug":14,"type":15},{"name":3084,"slug":3085,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":3089,"name":3089,"fn":3090,"description":3091,"org":3092,"tags":3093,"stars":3023,"repoUrl":3024,"updatedAt":3100},"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},[3094,3097],{"name":3095,"slug":3096,"type":15},"QA","qa",{"name":3098,"slug":3099,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":3102,"name":3102,"fn":3103,"description":3104,"org":3105,"tags":3106,"stars":3023,"repoUrl":3024,"updatedAt":3111},"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},[3107,3108],{"name":3018,"slug":3019,"type":15},{"name":3109,"slug":3110,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":3113,"name":3113,"fn":3114,"description":3115,"org":3116,"tags":3117,"stars":3023,"repoUrl":3024,"updatedAt":3125},"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},[3118,3121,3122],{"name":3119,"slug":3120,"type":15},"Code Review","code-review",{"name":3046,"slug":3047,"type":15},{"name":3123,"slug":3124,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":3127,"name":3127,"fn":3128,"description":3129,"org":3130,"tags":3131,"stars":3023,"repoUrl":3024,"updatedAt":3134},"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},[3132,3133],{"name":3095,"slug":3096,"type":15},{"name":3098,"slug":3099,"type":15},"2026-07-14T05:25:54.928983",{"slug":3136,"name":3136,"fn":3137,"description":3138,"org":3139,"tags":3140,"stars":3023,"repoUrl":3024,"updatedAt":3145},"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},[3141,3144],{"name":3142,"slug":3143,"type":15},"Automation","automation",{"name":3033,"slug":3034,"type":15},"2026-07-30T05:29:03.275638",496]