[maxkra15]

Description

Sharpa startup repeated texture decoding, built large nested Python payloads, and imported visual-only Newton geometry for native USD renderers. This keeps preparation with its existing owner and makes the default Newton GL scenes smaller.

• Load the detailed RJ45 switch front from one prepared USD; preserve all 1,151,285 triangles, authored normals and three material groups. --backdrop datacenter retains one full GL rack.
• Keep coffee/teapot mesh and particle payloads in NumPy arrays. Reuse existing mesh triangulation, avoid redundant config/normal copies, and scope shared texture decoding and optional downsampling to model upload.
• Default GL coffee/teapot to their studios, with full backdrops available through --backdrop. Kit/OVRTX retain full native scenes; skip their unused visual-only Newton imports.
• Restore teapot Kit/OVRTX camera preparation, hide the coffee studio floor before import, and match the GL cable's authored green color.
• Preserve RJ45 failure exit status before Kit closes. Returning failure inside the launcher context previously let Kit exit 0; unsuccessful cycles now raise SystemExit(1) inside that context. Environment cleanup and success criteria are unchanged.

The lean follow-up removes 22 net lines across six files. The separately discovered exit-status fix adds four net lines including its changelog. No framework, source module or tests added. Existing launch metadata was aligned and the user-approved obsolete GL-only rejection check was removed. Current main is merged, including exporter #8769; no companion overlay is required.

Validation

• Reviewed every changed block against current source and recent [6Biv] Separate shared visualizer settings from backend options #8704, [Fix] Cut ~4 s from the first articulation spawn by deferring the TorchScript converter import #8762 and [PR9B] Simplify teleoperation and recording script setup #8774 patterns. Full formatting/lint/changelog checks passed.
• Existing standalone CPU suite: 39 passed, 396 optional native matrix rows skipped. Focused CPU checks preserve exact teapot geometry/order/dtypes, transformed polygons/instances, RJ45 material batches, texture-cache cleanup and legacy config validation, and reset lifecycle behavior.
• Native HORDE L40: all nine demo/visualizer combinations plus the optional GL kitchen produced usable frames after three demo/control steps and 20 render-only frames. Sampled body/joint/particle states were finite; every sampled fluid particle moved, and coffee publication buffers matched the solver. Initial Kit RJ45 captured correctly but exposed the exit-status bug. All three RJ45 backends were rerun after that fix; the expected bounded-run status is 1, not full-cycle success. Screenshots were inspected.

Earlier paired startup measurements remain scoped to their recorded revisions: prepared-front GL first RGB changed from 42.7–43.8 s to 29.2–33.7 s on cached assets (9b6d304/83611cf, before/after/after/before). The compact-array CPU preparation scope changed from 3.75–3.77 s to 0.029–0.038 s (83611cf/dcf896d, fixed 135,314-position cloud, excluding sampling/native import/GPU/rendering). Shared triangulation removes duplication; it did not improve isolated CPU timing. These results do not guarantee startup below 30 seconds.

Matched coffee backdrop runs at 9b6d304 found modest differences: studio/kitchen first-RGB medians were GL 47.5/50.9 s, Kit 48.6/48.4 s, and OVRTX 46.9/50.4 s. With only two runs per scene, this does not establish a precise overhead; the kitchen was not the dominant startup cost.

Asset dependency and remaining limits

The prepared 30,464,855-byte front currently uses the development server requested for this change:

omniverse://isaac-dev.ov.nvidia.com/Isaac/IsaacLab/Environments/RizonSharpaRJ45/Environments/GB300/switch_front_c0ddcdf9d4ef.usdc

SHA-256: c0ddcdf9d4ef0845c6e30dc841ffb53d1edc3ebba4a64ab023cef8633c1b8c09. It is self-contained; provenance and bake source are beside it. Nucleus access is required until public promotion. HORDE uses a byte-identical mirror via the existing visual-root override; the actual URL was independently loaded and verified on the authenticated client.

RJ45 full-cycle reliability remains 9/10 at dcf896d, using fixed geometry and seeds 17–26. Seed 23 lost its grasp during final insertion (8.011 mm slip against the 8 mm guard); two successful instrumented runs had brief capture-flag drops. No controller, physics or success-threshold fix is claimed. Full coffee/teapot sequences were not run. The native RJ45 datacenter still emits an OmniPBR material-resolution warning; captured geometry/materials were usable.

[maxkra15]

run-ci

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge; no actionable defects were established.

Summary

This PR reduces presentation work when the Sharpa demos start.

• Coffee starts with a studio in Newton GL and keeps the kitchen optional.
• The teapot demo runs with Kit, Newton GL, OVRTX, or no visualizer.
• RJ45 Newton GL shows the switch front and ports by default.
• Teapot setup assigns its large lists after building the scene config.
• Newton GL reuses file textures and can limit their upload size.

No actionable defects were established. maxkra15 explicitly documents that no new tests were added.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Choose visualizers and backdrop] --> B[Launcher resolves viewers and cameras]
  B --> C{Newton GL selected?}
  C -->|Yes| D[Import Newton visual meshes]
  C -->|No| E[Skip visual-only Newton meshes]
  D --> F[Build scene]
  E --> F
  F --> G[GL uploads model textures]
  F --> H[Kit and OVRTX read USD scene]
  G --> I[Reuse decoded textures during upload]

Loading

_{Reviews (1) · Last reviewed commit: "Reduce Sharpa demo scene startup costs" · Reviewed by Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

Needs attention: Style consistency (4). Other reviewed categories look good.

• PR description: Add the omitted cable-color fix and reproducible base-versus-head coffee/RJ45 startup comparisons with revisions, command, environment count, hardware, backend, warmup, and metric.
• Design and architecture: Architecture looks good.
• Scope and over-engineering: No over-engineering issues.
• API: API changes look good.
• Compatibility and deprecation: No breaking changes identified.
• Implementation: Implementation looks good.
• Style consistency: Added backdrop documentation is inaccurate and duplicated; correct the robot-only room claim and consolidate the RJ45 default description.
• Test quality: Test quality looks good.

Minor fixes needed. Posted 4 inline findings.

Automated review; human maintainers own approval decisions.

[isaaclab-review-bot]

🟡 Warning · Style Consistency — Backdrop help overstates room contents

cfg.room is only added when --robot rizon_sharpa is set, so --backdrop room loads just the kitchen island by default. State that the enclosed room applies only to the robot variant.

[isaaclab-review-bot]

🟡 Warning · Style Consistency — Duplicated backdrop prose in README

Lines 21-24 restate the Newton GL and Kit/RTX backdrop defaults already given above, and break the sentence mid-clause after "In GL,". Keep one statement of the defaults and reflow it.

[isaaclab-review-bot]

🟡 Warning · Style Consistency — New methods precede class properties

set_model and log_mesh sit between init and the camera_speed property, breaking the prescribed order of dunders, properties, then public instance methods. Move both below the property block.

[isaaclab-review-bot]

🟡 Warning · Style Consistency — getattr default hides backend-only field

The guide forbids getattr defaults for known attributes; only NewtonGLBackendCfg declares texture_max_size, so this shared initializer hides the capability difference. Read it in the GL visualizer or declare it on both backends.

[maxkra15]

run-ci