search

Sentdex / Carla-RL

Reinforcement Learning codebase for self-driving car in Carla
MIT License
364 stars 98 forks source link

Error: BadShmSeg #6

Closed tom-doerr closed 4 years ago

tom-doerr commented 4 years ago

Even after over 5000 Episodes the Training doesn't start. The following error messages keep popping up:

X Error: BadShmSeg (invalid shared segment parameter) 128
  Extension:    130 (MIT-SHM)
  Minor opcode: 3 (X_ShmPutImage)
  Resource id:  0x2c0000e

I'm running the training inside a docker container using the following command:

docker run -it --gpus all -v /tmp/.X11-unix:/tmp/.X11-unix \
 -v ~/.Xauthority:/home/root/.Xauthority \
--link carla_server \
-u 1000 \
--env DISPLAY=:0 sentdex_carla_rl_image bash -c '
echo started & \
socat TCP-LISTEN:2000,fork TCP:carla_server:2000 & \
socat TCP-LISTEN:2001,fork TCP:carla_server:2001 & \
sleep 1 && \
#./carla_095/PythonAPI/examples/manual_control.py && \
python3 train.py
'
tom-doerr commented 4 years ago

I fixed it by setting the following environment variable: 

QT_X11_NO_MITSHM=1