ARM docker build with 24.12 pytorch fw image

[trvachov]

Description

Updates ARM Dockerfile to work with 24.12 pytorch FW image

Type of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Refactor
• Documentation update
• Other (please describe):

CI Pipeline Configuration

ARM build is not covered by pre-merge CI, is covered nightly. Having said that, this change might not work with CI at all because the blossom runners have a kernel/CUDA mismatch on their Grace systems.... more info TBD

Usage

docker run --gpus all --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 -it build-image-name:tag /bin/bash

Pre-submit Checklist

• I have tested these changes locally
• [N/A] I have updated the documentation accordingly
• [N/A] I have added/updated tests as needed
• All existing tests pass successfully

[pstjohn]

Are there tests for this anywhere ATM? Do we need to be testing ARM platforms as part of our pre-merge testing generally? That would definitely need to be something we discuss with the azure runner folks if so

[trvachov]

todo remove

[trvachov]

Fixed.

[trvachov]

@pstjohn I broke up this run command into two -- my understanding is i'm doing this in an earlier stage of the docker build therefore this doesn't have negative image size consequences -- am I correct?

[pstjohn]

I think all these layers end up in our release image, groan. I was thinking I would work a bit on trying to clean that up. But I think if you want to split these, that's fine; but you should move the

rm -rf llvm-project

line up to this block (if you can)?

[trvachov]

Cant move the rm -rf llvm-project up, because that's where the static libraries are against which triton compiles. I will just merge them back. I split these commands for debugging purposes.

[gagank1]

+1 to merging them back. I put it together because the LLVM build cache is huge and must be deleted in the same RUN command to actually reduce size

[trvachov]

Yep -- makes total sense -- thanks for the feedback.
Merged the two commands again. @gagank1 can you re-review?

[trvachov]

/build-ci

[pstjohn]

Sure LGTM if the tests pass 😅 , but tbh I'm not 100% clear on what's going on here with the ffmpeg stuff

[trvachov]

/build-ci

[gagank1]

for my own knowledge - what is this for?

[trvachov]

Decord doesn't build againt the latest ffmpeg. Rather than downgrading ffmpeg (which caused issues for me), someone provided a patch file to make decord compatible with the newest ffmpeg. The link in the comment has a description: dmlc/decord#186 (comment)

[gagank1]

interesting, i wonder why ffmpeg started causing issues now

[trvachov]

/build-ci

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 87.73%. Comparing base (43d2ca3) to head (335758f).

✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #581   +/-   ##
=======================================
  Coverage   87.73%   87.73%           
=======================================
  Files          89       89           
  Lines        5758     5758           
=======================================
  Hits         5052     5052           
  Misses        706      706           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.