Conversation
|
I've added a dockerfile for both non-cuda and cuda. They both work on my intel macbook. Both images have been pushed to dockerhub. Would be nice to have someone test the cuda image on an actual nvidia machine, to see if GPU mode works. I did both after all because of the massive size difference. Compressed, the non-cuda is 2.33 GB and the cuda image is 7.97 GB. I have no idea why there is that large of a discrepancy. Yes they use different bases (but still on Ubuntu 22.04), and cuda-compiled versions of various packages, but I wouldn't have expected THAT much of a difference.
|
|
Thanks man, nice work! I finally also traced the issue I was seeing on my side, it is a "Beth didn't upgrade her computer in a long time and there's a now-patched bug in Rosetta" :shame: :shame: :shame: |
- update pixi version to latest - match ubuntu versions in build and production (jammy, ie 22.04) - install InstaSeg as editable
|
Is it intentional to have the rather large |
|
Ohh sorry, forgot to insert the folder name in gitignore! - Will fix that in next commit |
- keep pixi around and use it as shell for downtream containers
gnodar01
left a comment
gnodar01
left a comment
There was a problem hiding this comment.
I think at this point we can get rid of Dockerfile_conda, Dockerfile_mamba, and Dockerfile_working_mamba
| # Attempt to pull the image from DockerHub | ||
| try: | ||
| print(f'Trying to pull the Docker image: {docker_image_name}') | ||
| session.run('docker', 'pull', "--platform", "linux/amd64", "bilayer/instanseg:1.0.2") |
There was a problem hiding this comment.
Shouldn't be hardcoding "bilayer/instanseg" here.
|
Closing for now, as it's seeming more likely we want to probably only cherry pick very small parts of this branch out for real, and other parts (like the various dockerfiles) will only be for documentation. Please don't delete the branch though until that work is done. |
3 participants
No description provided.