Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Upgrade CUDA on machine images to 11+ #174

Closed
0x2b3bfa0 opened this issue Aug 5, 2021 · 4 comments
Closed

Upgrade CUDA on machine images to 11+ #174

0x2b3bfa0 opened this issue Aug 5, 2021 · 4 comments
Assignees
@0x2b3bfa0
Labels
discussion Waiting for team decision invalid This doesn't seem right machine-image

Comments

@0x2b3bfa0
Copy link
Member

0x2b3bfa0 commented Aug 5, 2021 •
edited
Loading

Follow-up of iterative/cml#654

Machine images are still using CUDA 10, but we also offer support for CUDA 11 on container images. It would be awesome if we could upgrade it on our machine images without breaking backwards compatibility.

Related to #146
@0x2b3bfa0 0x2b3bfa0 added machine-image discussion Waiting for team decision labels Aug 5, 2021
@0x2b3bfa0 0x2b3bfa0 self-assigned this Aug 5, 2021
@0x2b3bfa0 0x2b3bfa0 mentioned this issue Aug 5, 2021
Closed
@DavidGOrtega
Copy link
Contributor

Ideally users would have to use images in their workflow. And never use the bare metal instance, anyway we could have AMIs with CUDA upgrade

@0x2b3bfa0
Copy link
Member Author

Please refer to iterative/cml#654 for context: the CUDA version on the bare metal instance affects the version of the libraries exposed to the container through the NVIDIA runtime.

@casperdcl
Copy link
Contributor

Odd. I thought the entire point is the bare metal only needs NVIDIA runtime drivers but not CUDA libraries. The container contains CUDA libraries which require the bare metal to have compatible drivers.

@0x2b3bfa0 0x2b3bfa0 mentioned this issue Feb 7, 2022
Merged
@0x2b3bfa0 0x2b3bfa0 mentioned this issue Feb 23, 2022
Closed
@0x2b3bfa0
Copy link
Member Author

libcusolver.so.11 is, in fact, present in out container images; I was probably wrong with iterative/cml#654

@0x2b3bfa0 0x2b3bfa0 added the invalid This doesn't seem right label Apr 24, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@0x2b3bfa0 0x2b3bfa0

Labels
discussion Waiting for team decision invalid This doesn't seem right machine-image
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
3 participants
@DavidGOrtega @casperdcl @0x2b3bfa0