Tyler Lum

I encountered the same issue! This fix should work, but I think a cleaner solution would be to avoid making a change in `train.py` ("feels" more like a hack), but...

NOTE: edited solution above to include `device_name`. This fixes the problem for `python train.py task=Ant headless=True sim_device=cpu rl_device=cpu test=True`. This doesn't fix the other issue though. I believe this is...

Found related forum post: https://forums.developer.nvidia.com/t/error-in-creating-camera-sensor-handle-while-headless/239808/2

I see. I believe that should work. Are you suggesting that we don't add the `use_link_poses` argument, and just always pass in `link_poses` to all `plan_*`? And then the user...

Another observation: If you forgot to put NEW_SCENE.obj into the right folder, the `reproduce_update_world_bug.py` example does not even raise an error, so it clearly is not reading the obj file,...

Clear diffs between base_case.py and reproduce_update_world_bug.py (only uses NEW_SCENE.obj vs. using base_scene then updating to NEW_SCENE.obj) ``` diff base_case.py reproduce_update_world_bug.py 30c30 < "NEW_scene": { --- > "base_scene": { 46,48c46 <...

The absolute path vs. relative path makes sense. "cuRobo treats meshes as assets so that you can use the same mesh across different environments. This would mean that that cuRobo...

Another undesired symptom of this bug. When I was first debugging this, I used `motion_gen.world_coll_checker.world_model.save_world_as_mesh("world.obj")` to visualize the scene and check if the scene was being updated correctly. It _**appeared**_...

This is very helpful thank you!

If there's any way to check if the update world is given an object with the same name but different file path, then giving a big warning, that'd be very...