Can I use camera obervation with 1 GPU

[huang1997214]

When I make the environment.

env = nl_guide_lift_env(
robots="Sawyer", # try with other robots like "Sawyer" and "Jaco"
has_renderer=False,
has_offscreen_renderer=True,
use_camera_obs=True,
ignore_done=True,
render_visual_mesh=True,
control_freq=20,
)

The following error occurs
File "mujoco_py/mjrendercontext.pyx", line 45, in mujoco_py.cymj.MjRenderContext.init
File "mujoco_py/mjrendercontext.pyx", line 113, in mujoco_py.cymj.MjRenderContext._setup_opengl_context
File "mujoco_py/opengl_context.pyx", line 130, in mujoco_py.cymj.OffscreenOpenGLContext.init
RuntimeError: Failed to initialize OpenGL

This seems to result in 1 GPU. I wondering if there is some way that I can use 1GPU to get the image observation.
I want to use image observation to train a DQN.
Thanks!

[hermanjakobsen]

This is a known issue with mujoco-py and rendering. I have posted a similar issue some time ago :)

[huang1997214]

Sorry, I am not familiar with mujoco-py.

[RayYoh]

This is a known issue with mujoco-py and rendering. I have posted a similar issue some time ago :)

So how to solve this problem?

[peterdavidfagan]

@RayYoh it is a bit of a hack but you can use the following from my experience: openai/mujoco-py#268 (comment)

Kudos to @HesNobi for this.

[RayYoh]

@RayYoh it is a bit of a hack but you can use the following from my experience: openai/mujoco-py#268 (comment)

Kudos to @HesNobi for this.

Thanks a lot. Hope it works.