Closed practical-dreamer closed 3 weeks ago
Update: I'm pretty sure this issue is not specific to docker - I adjusted INSTALL_EXTENSIONS=FALSE
in the dockerfile and it finally worked. I'm now getting Error response from daemon: could not select device driver "nvidia" with capabilities: [[gpu]]
probably because I need the NVIDIA container toolkit but at least the build completed... My theory is the package dependency issue is a problem with one of the extensions and would be observed if I called INSTALL_EXTENSIONS=FALSE ./start_linux.sh
directly...
Unsure what extension is the issue
I am closing this issue as my original objective of building a docker image can be accomplished by modifying INSTALL_EXTENSIONS=TRUE
to INSTALL_EXTENSIONS=FALSE
in the dockerfile.
However, I may end up reopening or reinvestigating why pip is getting stuck when trying to download dependencies for the extensions... Not sure which extension is causing problems but this probably is causing problems for others and should be looked into...
Also just FYI if anyone gets to this page and is trying to figure how to get docker with cuda support to work I made a short gist hoping to help anyone along the path - accessible here: https://gist.github.com/practical-dreamer/79c26931d99bc6a7f28271b3612907a9
Describe the bug
Objective:
Having trouble building docker image from 'Alternative: Docker' section of README.md. - hoping some chad out there will help me figure why it's hanging.
Issue:
After half an hour docker compose is still running (see my logs) - doesn't seem normal when
pip install -r requirements
takes under 2 minutes in a fresh conda env.Observation:
According to stdout from
docker compose up --build
it seems I'm downloading MULTIPLE versions of packages which does not occur when runningpip install -r requirements
from a conda env.Here's a small sample (see the logs section for the complete output)
Is there an existing issue for this?
Reproduction
curl -fsSL https://get.docker.com -o get-docker.sh sudo sh get-docker.sh sudo usermod -aG docker $USER
ln -s docker/{nvidia/Dockerfile,nvidia/docker-compose.yml,.dockerignore} . cp docker/.env.example .env mkdir -p logs cache docker compose up --build | tee docker_output.log
Screenshot
Logs
System Info