This is a docker for running Rope using headless VNC environments. Useful for running Rope on cloud GPU services like Runpod.io or Vast.ai.
The Docker image is installed with the following components:
- Rope Pearl-00
- JupyterLab (default http port
8080
) - VNC-Server (default VNC port
5901
) - noVNC 1.5.0 - HTML5 VNC client (default http port
6901
) - CUDA Toolkit 11.8
- Pytorch 2.0.1+cu118
- Desktop environment Xfce4
- fiebrowser (default http port
8585
) - Mozilla Firefox
-
Run command with mapping to local port
5901
(vnc protocol),6901
(vnc web access),8080
(JupyterLab), and8585
(filebrowser):docker run -d --gpus all -p 5901:5901 -p 6901:6901 -p 8080:8080 -p 8585:8585 -e VNC_PASSWORDLESS=true -e VNC_RESOLUTION=1024x768 asyafiqe/rope_vnc:latest
For more options, please check ConSol's docker-headless-vnc-container github.
-
Build an image from scratch:
docker build -t asyafiqe/rope_vnc .
-
Vast.ai template: Put
asyafiqe/rope_vnc:latest
in image path/tag. Docker options:-p 5901:5901 -p 6901:6901 -p 8080:8080 -p 8585:8585 -e VNC_PASSWORDLESS=true -e VNC_RESOLUTION=1024x768
In Launch Mode select 'Run interactive shell server, SSH'. Check 'Use direct SSH connection'. On-start Script:
env | grep _ >> /etc/environment; echo 'starting up' /dockerstartup/vnc_startup.sh sleep infinity
-
Runpod.io template:
- Container image:
asyafiqe/rope_vnc:latest
. - Docker command:
-p 5901:5901 -p 6901:6901 -p 8080:8080 -p 8585:8585
- Container image: