Add THD support for RoPE#1284
Open
niyunsheng wants to merge 2 commits into
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Add THD packed-input support for RoPE. This introduces a THD RoPE autograd wrapper and public functional/transformers APIs for query/key tensors shaped
(total_tokens, num_heads, head_dim)withcos/sinshaped(total_tokens, head_dim).Details
The THD path reuses the existing Triton RoPE kernel with a packed-token layout wrapper, keeping the existing batch-major RoPE API unchanged. Documentation and RoPE unit tests were updated to cover the new public THD API.
This is an initial RoPE-only step toward THD / packed-sequence training support requested in #1281. It does not close the issue, since broader packed-layout support across kernels remains future work.
Testing Done
make testto ensure correctnessmake checkstyleto ensure code stylemake test-convergenceto ensure convergenceAdditional targeted testing:
python -m pytest test/transformers/test_rope.py