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.

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

Enable tokenizers in shortfin packages on Linux x86_64. #688

Open
wants to merge 1 commit into
base: main
Choose a base branch
from

Conversation

ScottTodd
Copy link
Member

Fixes #679.

Multi-arch support is hacky at the moment, but we are only building for x86_64 today and not aarch64. I think the ideal path forward there is to publish our own multi-platform images that include all of our deps (including rust and ccache), so I started on that with ScottTodd/nod-ai-base-docker-images@bdc5cc5. Another option is to continue using the upstream manylinux dockerfiles and install more deps on-demand like yum install -y rust cargo).

Tested package build times and sizes:

Description Logs Duration Package size Cache hit rate
Baseline rc build logs here 5m15s 2.6mb
New rc build logs here 8m40s 11.8mb
Baseline dev build 1 logs here 2m49s 1.2mb 0%
Baseline dev build 2 logs here 2m9s 1.2mb 45%
New dev build 1 logs here 4m1s 5.8mb 58%
New dev build 2 logs here 3m52s 5.8mb 67%

Dev package builds will be used for #584, so keeping those times low is a priority. Could disable tokenizers there, at least as long as using the native tokenizers library is optional

Observations:

  • 30s docker pull before, 45s docker pull after
    • Could try to prune more dependencies
  • 365 build actions before, 412 build actions after
    • Still lower cache hit rates than expected

@ScottTodd ScottTodd requested a review from marbre December 12, 2024 18:13
Copy link
Collaborator

@marbre marbre left a comment

Choose a reason for hiding this comment

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

Let's wait for a signal from @stellaraccident before the release to decide if this should be included in the stable releases.

@ScottTodd
Copy link
Member Author

Let's wait for a signal from @stellaraccident before the release to decide if this should be included in the stable releases.

SGTM!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Enable SHORTFIN_ENABLE_TOKENIZERS in Linux package builds
2 participants