Open LunarEngineer opened 3 years ago
@aKutluozen @hagopi1611 I thought I'd add you to this conversation.
If anyone does not have the capability to run on an 18.04 system, the cschranz image can be removed from the dockerfile in order to allow running from the command line. This image is what's enabling GPU support, but is not needed immediately during development.
If you guys are planning on building the image and running from the image and there is a desire, then I can spend some time and get a replacement base docker image with jupyter that doesn't support GPU (likely a bunch available) working.
Building and running the image likely is not necessary if you're simply managing the pip install of the repo yourself. The odds of dependencies falling out of scope during the time frame is minimal.
Build Dependencies
The docker container has an operating system and cuda version dependency that comes from the cschranz Jupyter image I used as a base.
The operating system is a hard dependency at Ubuntu 18.04, while the cuda version is a softer dependency ranging from the available images.
In order for this to build correctly you would need to either match the dependencies specified in the environment, or any of those available here.
@cizumigawa3 I thought I would tag you in this issue, since you mentioned you were having difficulties getting the docker container up and running.