Offscreen render does not work (GLEW initalization error: Missing GL version)

[caprest]

Describe the bug
Rendering with off screen fails to initialize.

To Reproduce

import mujoco_py
model = """
<mujoco model="cartpole">
  <worldbody>
    <geom name="floor" pos="0 0 -3.0" size="40 40 40" type="plane"/>
  </worldbody>
</mujoco>
"""


model = mujoco_py.load_model_from_xml(model)
sim = mujoco_py.MjSim(model)
mujoco_py.MjRenderContextOffscreen(sim, device_id=0)

Error Messages
ERROR: GLEW initalization error: Missing GL version

It seems like the error occurs at the first line of this function (mjr_makeContext(self._model_ptr, &self._con, mjFONTSCALE_150))

mujoco-py/mujoco_py/mjrendercontext.pyx

Lines 83 to 93 in 1aeb02e

	def _set_mujoco_buffers(self):
	mjr_makeContext(self._model_ptr, &self._con, mjFONTSCALE_150)
	if self.offscreen:
	mjr_setBuffer(mjFB_OFFSCREEN, &self._con);
	if self._con.currentBuffer != mjFB_OFFSCREEN:
	raise RuntimeError('Offscreen rendering not supported')
	else:
	mjr_setBuffer(mjFB_WINDOW, &self._con);
	if self._con.currentBuffer != mjFB_WINDOW:
	raise RuntimeError('Window rendering not supported')
	self.con = WrapMjrContext(&self._con)

(found by print debug)

Furthermore when I add "None" argument at mysterious position
mujoco_py.MjRenderContextOffscreen(sim, None,device_id=0)
images are correctly rendered, though black winodow appears.

Desktop (please complete the following information):

• OS: [Ubuntu18.04 ]
• Python Version [3.7.3]
• Mujoco Version [2.0.0]
• mujoco-py version [2.0.2.2]

Environment

• output of: echo $LD_LIBRARY_PATH /home/user/.mujoco/mujoco200/bin

Additional context
OpenGL version: 4.6.0
Nvidia Driver: 418.67

[katadh]

I am having the same problem.

[gitanshu]

Have you already looked at #268 ?

[caprest]

At least, I already tried the solution for LD_PRELOAD and seems like to have correct output

>>> echo LD_PRELOAD
/usr/lib/x86_64-linux-gnu/libGLEW.so

[ZiwenZhuang]

It seems like if you have it render in 'window' mode for the first time, there will be no error. But the following offscreen render will be all zeros (black) matrix. (So, I still didn't fix it.)

[ZiwenZhuang]

After gained some insight from my several tests, I found that letting the system render on screen or make a 'window' mode render context will get rid of the problem. But that would lead to a serious buffer problem and get only all zero numpy arrays if you keep using that context viewer, which might due to some OpenGL buffer problems.
Also, I've noticed that PyOpenGL supports OpenGL only from v1.1 to 4.4 (referring to here). But you reported your OpenGL version is 4.6 :)

Solution

• Theoretically, you can downgrade your OpenGL version via reinstall a lower version of Nvidia driver. But that would lead to a series of re-installation of Cuda, or more. So I didn't try that.

• Let the entire program initialize an on-screen render context ('window' mode) and abandon it. Then, make your off-screen render context as normal, which will ensure you get the image you expected. Although this will leave a black window on your screen and system keep reporting 'mujoco-py not responding'.

[dingchenghu]

I face exactly the same issue.
Any other idea on how to solve that? Thanks

[RajK853]

This solved my issue in Ubuntu 18.04.3 LTS:
sudo mkdir -p /usr/lib/nvidia-000
Then add this in ~/.bashrc file:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia-000

[Rowing0914]

Issue's been solved!!

See #408 (comment)

I was having the same problem.

I tried below yet, none of them had worked.

• https://github.com/deepmind/dm_control#rendering
• GLEW initalization error: Missing GL version #268

Spec

• OS: Ubuntu 18.04.3 LTS
• GPU: GeForce RTX 2080 with Max-Q Design
• Driver: 430
• mujoco-py==2.0.2.7

import gym
from gym.wrappers import Monitor

env = gym.make('HalfCheetah-v2')
env = Monitor(env, './video', force=True)
env.reset()
while True:
    obs, r, done, info = env.step([0, 0, 0, 0, 0, 0])
    if done: break

[RajK853]

Recently while installing mujoco-py in a new conda environment to use tensorflow-gpu 1.13.1, I faced the same problem again. My problem was resolved by installing tensorflow-gpu and glew using conda instead of pip. This will also install the supported cuda version in the conda and it might have solved my issue.

[Rowing0914]

This solved my issue in Ubuntu 18.04.3 LTS:
sudo mkdir -p /use/lib/nvidia-000
Then add this in ~/.bashrc file:
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia-000

@RajK853
Thank you for your sharing! In my case, which is described above, your advice worked well and solved my issue.

So, after I created the empty dir named nvidia-OOO(OOO == 430 in my case), at the first execution of import mujoco_py it started compiling the library again though, everything seems to settle now!

Edit: I am running scripts on PyCharm, which requires some tricky bit in terms of environment variables....

[robotlearning123]

Hi, I have the folder /usr/lib/nvidia-410 but it's not empty. How to deal with it?

$ ll /usr/lib/nvidia-410
total 196224
drwxr-xr-x   6 root root     4096 Jul 30 11:34 ./
drwxr-xr-x 188 root root    20480 Jan 13 23:30 ../
-rw-r--r--   1 root root        0 Nov 15  2018 alt_ld.so.conf
drwxr-xr-x   2 root root     4096 Jul 30 11:34 bin/
-rw-r--r--   1 root root       42 Nov 15  2018 ld.so.conf
lrwxrwxrwx   1 root root       23 Nov 15  2018 libEGL_nvidia.so.0 -> libEGL_nvidia.so.410.78
-rw-r--r--   1 root root  1031552 Nov 11  2018 libEGL_nvidia.so.410.78
lrwxrwxrwx   1 root root       16 Nov 15  2018 libEGL.so -> libEGL.so.410.78
lrwxrwxrwx   1 root root       16 Nov 15  2018 libEGL.so.1 -> libEGL.so.410.78
-rw-r--r--   1 root root    73328 Nov 11  2018 libEGL.so.1.1.0
-rw-r--r--   1 root root    23392 Nov 11  2018 libEGL.so.410.78
-rw-r--r--   1 root root   732408 Nov 11  2018 libGLdispatch.so.0
lrwxrwxrwx   1 root root       29 Nov 15  2018 libGLESv1_CM_nvidia.so.1 -> libGLESv1_CM_nvidia.so.410.78
-rw-r--r--   1 root root    60200 Nov 11  2018 libGLESv1_CM_nvidia.so.410.78
lrwxrwxrwx   1 root root       17 Nov 15  2018 libGLESv1_CM.so -> libGLESv1_CM.so.1
lrwxrwxrwx   1 root root       21 Jul 30 11:34 libGLESv1_CM.so.1 -> libGLESv1_CM.so.1.2.0
-rw-r--r--   1 root root    43696 Nov 11  2018 libGLESv1_CM.so.1.2.0
lrwxrwxrwx   1 root root       26 Nov 15  2018 libGLESv2_nvidia.so.2 -> libGLESv2_nvidia.so.410.78
-rw-r--r--   1 root root   111400 Nov 11  2018 libGLESv2_nvidia.so.410.78
lrwxrwxrwx   1 root root       14 Nov 15  2018 libGLESv2.so -> libGLESv2.so.2
lrwxrwxrwx   1 root root       18 Jul 30 11:34 libGLESv2.so.2 -> libGLESv2.so.2.1.0
-rw-r--r--   1 root root    83280 Nov 11  2018 libGLESv2.so.2.1.0
lrwxrwxrwx   1 root root       10 Nov 15  2018 libGL.so -> libGL.so.1
lrwxrwxrwx   1 root root       15 Nov 15  2018 libGL.so.1 -> libGL.so.410.78
-rw-r--r--   1 root root   685848 Nov 11  2018 libGL.so.1.7.0
-rw-r--r--   1 root root  1270576 Nov 11  2018 libGL.so.410.78
lrwxrwxrwx   1 root root       23 Nov 15  2018 libGLX_indirect.so.0 -> libGLX_nvidia.so.410.78
lrwxrwxrwx   1 root root       23 Nov 15  2018 libGLX_nvidia.so.0 -> libGLX_nvidia.so.410.78
-rw-r--r--   1 root root  1270576 Nov 11  2018 libGLX_nvidia.so.410.78
lrwxrwxrwx   1 root root       11 Nov 15  2018 libGLX.so -> libGLX.so.0
-rw-r--r--   1 root root    65168 Nov 11  2018 libGLX.so.0
lrwxrwxrwx   1 root root       15 Nov 15  2018 libnvcuvid.so -> libnvcuvid.so.1
lrwxrwxrwx   1 root root       20 Nov 15  2018 libnvcuvid.so.1 -> libnvcuvid.so.410.78
-rw-r--r--   1 root root  2807440 Nov 11  2018 libnvcuvid.so.410.78
-rw-r--r--   1 root root   370984 Nov 11  2018 libnvidia-cbl.so.410.78
lrwxrwxrwx   1 root root       18 Nov 15  2018 libnvidia-cfg.so -> libnvidia-cfg.so.1
lrwxrwxrwx   1 root root       23 Nov 15  2018 libnvidia-cfg.so.1 -> libnvidia-cfg.so.410.78
-rw-r--r--   1 root root   179528 Nov 11  2018 libnvidia-cfg.so.410.78
lrwxrwxrwx   1 root root       23 Nov 15  2018 libnvidia-compiler.so -> libnvidia-compiler.so.1
lrwxrwxrwx   1 root root       28 Nov 15  2018 libnvidia-compiler.so.1 -> libnvidia-compiler.so.410.78
-rw-r--r--   1 root root 47842480 Nov 11  2018 libnvidia-compiler.so.410.78
-rw-r--r--   1 root root 25215360 Nov 11  2018 libnvidia-eglcore.so.410.78
lrwxrwxrwx   1 root root       30 Nov 15  2018 libnvidia-egl-wayland.so.1 -> libnvidia-egl-wayland.so.1.1.0
-rw-r--r--   1 root root    32904 Nov 11  2018 libnvidia-egl-wayland.so.1.1.0
lrwxrwxrwx   1 root root       21 Nov 15  2018 libnvidia-encode.so -> libnvidia-encode.so.1
lrwxrwxrwx   1 root root       26 Nov 15  2018 libnvidia-encode.so.1 -> libnvidia-encode.so.410.78
-rw-r--r--   1 root root   168184 Nov 11  2018 libnvidia-encode.so.410.78
-rw-r--r--   1 root root   292840 Nov 11  2018 libnvidia-fatbinaryloader.so.410.78
lrwxrwxrwx   1 root root       18 Nov 15  2018 libnvidia-fbc.so -> libnvidia-fbc.so.1
lrwxrwxrwx   1 root root       23 Nov 15  2018 libnvidia-fbc.so.1 -> libnvidia-fbc.so.410.78
-rw-r--r--   1 root root   119048 Nov 11  2018 libnvidia-fbc.so.410.78
-rw-r--r--   1 root root 27018280 Nov 11  2018 libnvidia-glcore.so.410.78
-rw-r--r--   1 root root   580856 Nov 11  2018 libnvidia-glsi.so.410.78
-rw-r--r--   1 root root 13792536 Nov 11  2018 libnvidia-glvkspirv.so.410.78
lrwxrwxrwx   1 root root       18 Nov 15  2018 libnvidia-ifr.so -> libnvidia-ifr.so.1
lrwxrwxrwx   1 root root       23 Nov 15  2018 libnvidia-ifr.so.1 -> libnvidia-ifr.so.410.78
-rw-r--r--   1 root root   206888 Nov 11  2018 libnvidia-ifr.so.410.78
lrwxrwxrwx   1 root root       17 Nov 15  2018 libnvidia-ml.so -> libnvidia-ml.so.1
lrwxrwxrwx   1 root root       22 Nov 15  2018 libnvidia-ml.so.1 -> libnvidia-ml.so.410.78
-rw-r--r--   1 root root  1528152 Nov 11  2018 libnvidia-ml.so.410.78
lrwxrwxrwx   1 root root       34 Jul 30 11:34 libnvidia-ptxjitcompiler.so.1 -> libnvidia-ptxjitcompiler.so.410.78
-rw-r--r--   1 root root 12129192 Nov 11  2018 libnvidia-ptxjitcompiler.so.410.78
-rw-r--r--   1 root root 27151728 Nov 11  2018 libnvidia-rtcore.so.410.78
-rw-r--r--   1 root root    13080 Nov 11  2018 libnvidia-tls.so.410.78
lrwxrwxrwx   1 root root       23 Jul 30 11:34 libnvidia-wfb.so.1 -> libnvidia-wfb.so.410.78
-rw-r--r--   1 root root   295416 Jun  4  2018 libnvidia-wfb.so.410.78
lrwxrwxrwx   1 root root       20 Jul 30 11:34 libnvoptix.so.1 -> libnvoptix.so.410.78
-rw-r--r--   1 root root 35398672 Nov 11  2018 libnvoptix.so.410.78
lrwxrwxrwx   1 root root       14 Nov 15  2018 libOpenGL.so -> libOpenGL.so.0
-rw-r--r--   1 root root   215824 Nov 11  2018 libOpenGL.so.0
drwxr-xr-x   2 root root     4096 Jul 30 11:34 tls/
drwxr-xr-x   2 root root     4096 Jul 30 11:34 vdpau/
drwxr-xr-x   2 root root     4096 Jul 30 11:34 xorg/

And

echo $LD_LIBRARY_PATH
/home/cong/ros_ws/dual_ws/devel/lib:/opt/ros/kinetic/lib:/opt/ros/kinetic/lib/x86_64-linux-gnu:/usr/local/cuda-10.0/lib64:/home/cong/.mujoco/mujoco200/bin:/usr/lib/nvidia-410

When I run the code

import gym
from gym.wrappers import Monitor

env = gym.make('HalfCheetah-v2')
env = Monitor(env, './video', force=True)
env.reset()
while True:
    obs, r, done, info = env.step([0, 0, 0, 0, 0, 0])
    if done: break

I also get the error:

ERROR: GLEW initalization error: Missing GL version

Press Enter to exit ...

[seyeeet]

@wangcongrobot did you find the solution?