Skip to content

Test different input sequence lengths for Llama #1070

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

Merged
merged 6 commits into from
Mar 6, 2025

Conversation

pmarkovicTT
Copy link
Contributor

@pmarkovicTT pmarkovicTT commented Jan 20, 2025

Add test to make sure Llama compiles and run fwd pass with different input sequence lengths as we will have inputs of various lengths during training.

Close #1071

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests823 ran490 passed333 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests665 ran434 passed231 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests665 ran437 passed228 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests823 ran492 passed331 skipped0 failed
TestResult
No test annotations available

@pmarkovicTT pmarkovicTT force-pushed the pmarkovic/test-input-seq-lengths-llama branch from 5b45490 to 710afb4 Compare February 4, 2025 16:47
Copy link

github-actions bot commented Feb 4, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests510 ran451 passed59 skipped0 failed
TestResult
No test annotations available

1 similar comment
Copy link

github-actions bot commented Feb 4, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests510 ran451 passed59 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Feb 4, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests568 ran489 passed79 skipped0 failed
TestResult
No test annotations available

1 similar comment
Copy link

github-actions bot commented Feb 4, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests568 ran489 passed79 skipped0 failed
TestResult
No test annotations available

@nvukobratTT
Copy link
Collaborator

@pmarkovicTT is this one ready for review and potential merge? If not, can we move it to draft?

@pmarkovicTT pmarkovicTT force-pushed the pmarkovic/test-input-seq-lengths-llama branch from 710afb4 to f081375 Compare February 25, 2025 13:01
Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests670 ran536 passed134 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests611 ran482 passed129 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ☑️Skipped ⚠️Failed ❌️
TT-Forge-FE Tests611 ran481 passed129 skipped1 failed
TestResult
TT-Forge-FE Tests
pytest
test_dla.test_dla_pytorch[dla34]❌ failure

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests670 ran536 passed134 skipped0 failed
TestResult
No test annotations available

@pmarkovicTT
Copy link
Contributor Author

@nvukobratTT PR is ready for review/merge.

Summary of previous conversations:

  • We selected input sequence sizes that matter to us (128, 512, 2048)
  • Added test for only 1 layer as we talked in one of previous syncs to run test faster
  • Embedding inputs have to be int and it's explicitly casted to int32 due to following issue Invalid Runtime inputs to embedding #952
  • When it comes to training/testing in lower precision, that's something we will do and incorporate in future training tests we add

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests616 ran484 passed132 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests675 ran538 passed137 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests616 ran484 passed132 skipped0 failed
TestResult
No test annotations available

Copy link

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests675 ran538 passed137 skipped0 failed
TestResult
No test annotations available

Copy link
Collaborator

@nvukobratTT nvukobratTT left a comment

Choose a reason for hiding this comment

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

Not to block anymore, left are minor details, so I'm approving this PR :))

Regarding CI, I would suggest adding 1B variant just to make sure that it's passing. 3B we can skip.

And make sure to check do you need input IDs DF (non-blocker) :))

@pmarkovicTT pmarkovicTT force-pushed the pmarkovic/test-input-seq-lengths-llama branch from 2ab4f57 to 463414e Compare March 6, 2025 11:05
@codecov-commenter
Copy link

codecov-commenter commented Mar 6, 2025

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 43.40%. Comparing base (9213144) to head (97c3e69).
Report is 2 commits behind head on main.

Additional details and impacted files
@@           Coverage Diff           @@
##             main    #1070   +/-   ##
=======================================
  Coverage   43.40%   43.40%           
=======================================
  Files          48       48           
  Lines        7860     7860           
=======================================
  Hits         3412     3412           
  Misses       4448     4448           

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

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests625 ran489 passed136 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests684 ran541 passed143 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests625 ran489 passed136 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests684 ran541 passed143 skipped0 failed
TestResult
No test annotations available

Verified

This commit was created on GitHub.com and signed with GitHub’s verified signature.
Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests634 ran492 passed142 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests693 ran556 passed137 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests634 ran492 passed142 skipped0 failed
TestResult
No test annotations available

Copy link

github-actions bot commented Mar 6, 2025

TestsPassed ✅Skipped ⚠️Failed
TT-Forge-FE Tests693 ran556 passed137 skipped0 failed
TestResult
No test annotations available

@pmarkovicTT pmarkovicTT merged commit 70f40ea into main Mar 6, 2025
11 checks passed
@pmarkovicTT pmarkovicTT deleted the pmarkovic/test-input-seq-lengths-llama branch March 6, 2025 16:52
pdeviTT pushed a commit that referenced this pull request Mar 25, 2025
Add test to make sure Llama compiles and run fwd pass with different
input sequence lengths as we will have inputs of various lengths during
training.

Close #1071
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.

Different input sequence length
4 participants