Errors with hydra scene-graph instancing

I am using Orbit with the GridCloner while enabling the “hydra scene-graph instancing” and “flatcache” using the following:

if sim.get_physics_context().use_gpu_pipeline:
set_carb_setting(sim._settings,“/persistent/omnihydra/useSceneGraphInstancing”, True)

I get two sets of warnings while running the simulation from a standalone python script:

[Warning] [omni.hydra.scene_delegate.plugin] Calling getBypassRenderSkelMeshProcessing for prim /World/envs/env_0/table/leg_03.proto_leg_id0 that has not been populated

[Warning] [omni.usd] Coding Error: in _GetProtoPrim at line 1866 of /buildAgent/work/ca6c508eae419cf8/USD/pxr/usdImaging/usdImaging/instanceAdapter.cpp – Failed verification: ’ r ’ – instancer = /World/envs/env_0/kona_module/bolts/bolt_M5X16_tight, cachePath = /World/envs/env_0/kona_module/bolts/bolt_M5X16_tight

[Warning] [omni.usd] Coding Error: in ProcessPropertyChange at line 1313 of /buildAgent/work/ca6c508eae419cf8/USD/pxr/usdImaging/usdImaging/instanceAdapter.cpp – Failed verification: ’ proto.adapter ’ – /World/envs/env_0/kona_module/bolts/bolt_M5X16_tight

during scene loading


[Warning] [omni.hydra.scene_delegate.plugin] InstanceAdapter - cannot find cache item for proto /World/envs/env_0/kona_module/bolts/bolt_M5X16_tight.proto_bolt_M5X10_tight_id0

during the execution.

Interestingly, the second set of warnings depends on the cloned environment number; some numbers do not create the warning. When these warnings are there, I see undesired effects, such that I cannot set the pose of the prims, etc.

What do these warnings mean? How can I prevent them (and their effects)?

Note: Isaac Sim 2022.2.1 docker container on Ubuntu 20.04. Nvidia driver Version: 525.60.11. CUDA Version: 12.0.

Thanks in advance for the support.

Any news on this?

Hi @ozhanozen

We have asked the rendering to help answer this issue since it is more related to how that pipeline works.

As a quick thing, can you clear your cache and see if that helps?

./_isaac_sim/ --clear-data --clear-cache

Also can you please share your error log files?

@ozhanozen I have been asked to share a repro. Can you make one for us to share internally please?

Thanks for your reply @mmittal.

We have noticed these warnings only occur for a specific asset and when we access the sub-elements of the asset using the API (e.g., using XformPrim.get_forld_pose to access the bolt_M5X16_tight element of kona_module). This asset is directly imported from Onshape using your Onshape importer. Moreover, the aforementioned undesired effects seem to be related to another problem.

What do you mean by repro, a USD file? If so, how can we share one with you?