Skip to content

Conversation

@xuzhao9
Copy link
Contributor

@xuzhao9 xuzhao9 commented Dec 10, 2025

Do not install CUDA compile toolchain in the docker to speedup the docker build.

Now the docker build finishes in < 20min, and it should reduce the chance of connection timeout when downloading the docker image.

Test plan:

https://github.com/meta-pytorch/tritonbench/actions/runs/20100057335

@xuzhao9 xuzhao9 temporarily deployed to docker-s3-upload December 10, 2025 04:01 — with GitHub Actions Inactive
@meta-cla meta-cla bot added the cla signed label Dec 10, 2025
@xuzhao9 xuzhao9 temporarily deployed to docker-s3-upload December 10, 2025 19:03 — with GitHub Actions Inactive
@xuzhao9 xuzhao9 temporarily deployed to docker-s3-upload December 10, 2025 19:03 — with GitHub Actions Inactive
@xuzhao9 xuzhao9 changed the title [docker] fix the nv docker build [docker] speedup the nv docker build Dec 10, 2025
@xuzhao9 xuzhao9 requested a review from FindHao December 10, 2025 19:29
Copy link
Member

@FindHao FindHao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. one question, are we using the docker's system cuda now?

@xuzhao9
Copy link
Contributor Author

xuzhao9 commented Dec 10, 2025

@FindHao Previously we were using it to build fbgemm wheels (and for bisection, if required). We are not doing this anymore.

@xuzhao9 xuzhao9 merged commit cdc2808 into main Dec 10, 2025
7 of 9 checks passed
@xuzhao9 xuzhao9 deleted the xz9/fix-docker branch December 10, 2025 20:02
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants