Closed PrettyMeng closed 4 years ago
Yes, you can train without a window.
Calling env.step()
does not open a window.
Yes, you can train without a window.
Calling
env.step()
does not open a window.
Thanks for your quick reply! I would like to know does this error occur because of trying to open a window?
pyglet.canvas.xlib.NoSuchDisplayException: Cannot connect to "None"
which is also mentioned in #13 .
That has to do with trying to connect to a display for rendering. Have you tried using xvfb-run
as outlined here?
That has to do with trying to connect to a display for rendering. Have you tried using
xvfb-run
as outlined here?
Thanks again. Actually the cluster I use cannot install xvfb for some reason. I wonder whether it is possible to do training without displaying or rendering.
You should contact the cluster administrators and try to get them to install xvfb. It's necessary for many environments that use OpenGL for rendering, unfortunately.
I'm trying to get the people behind the pyglet package to remove this dependency, by implementing EGL support, but that hasn't been moving.
You should contact the cluster administrators and try to get them to install xvfb. It's necessary for many environments that use OpenGL for rendering, unfortunately.
I'm trying to get the people behind the pyglet package to remove this dependency, by implementing EGL support, but that hasn't been moving.
Thank you so much! I will try.
Hi there, is it possible to train an RL agent without opening the simulation window? I'm trying to do training on my server but it can be problematic to open a window there.
Any help or response will be highly appreciated. Thanks!