Docker and PhysX Visual Debugger (PVD)

Hi,

I followed the step to build and run Isaac-Gym into a docker, according to Install in a Docker container but when I run an example (even in an LXDE environment) I got the next error related to the connection with PhysX Visual Debugger (PVD).

root@d084a87cc264:/workspace/isaacgym/python/examples# python joint_monkey.py
Importing module 'gym_36' (/workspace/isaacgym/python/isaacgym/_bindings/linux-x86_64/gym_36.so)
Setting GYM_USD_PLUG_INFO_PATH to /workspace/isaacgym/python/isaacgym/_bindings/linux-x86_64/usd/plugInfo.json
Not connected to PVD
Segmentation fault (core dumped)

Any idea to solve it?
Thanks in advance.

Hi @toni.sm

The issue here is that the docker container does not support rendering - that’s the actual cause of the crash.

Per the documentation:

Note that rendering is not supported when running from a docker container, so only RL training running in headless mode will work.

You should be able to run the rl training examples in the container, so long as you use the --headless flag.

The docker setup here allows you to train on clusters you may have access to like that, and then you can take the trained policy and run it on your local machine, with rendering on. This won’t work for any policy that requires camera rendering during training of course.

Take care,
-Gav

@gstate Hi Gav,

Is there a way to install PVD in linux ? I only found PVD pack for windows in Nvidia downloading center. Thanks~

Hi @qiurh097,

I’m afraid that at the moment the PVD debugger is only available in Windows. You can save a PVD capture made using the Gym CPU pipeline on Linux and open it using PVD in Windows however.

Given that this thread started out with a question about docker and rendering, note that with Isaac Gym Preview 3 we include the appropriate components in the sample Dockerfile to support rendering from the container as well.

Take care,
-Gav