Resolve issue with async scheduling when decode and prompt tokens are mixed #642
+22
−8
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
… mixed Signed-off-by: Tianmu Li <tianmu.li@intel.com>
tianmu-li
requested review from
adobrzyn,
afierka-intel,
iboiko-habana,
kamil-kaczor,
ksmusz,
kzawora-intel,
mgawarkiewicz-intel,
michalkuligowski,
vivekgoe and
xuechendi
as code owners
Contributor
There was a problem hiding this comment.
Pull request overview
This PR resolves an issue with async scheduling when decode and prompt tokens are mixed in a batch. The fix ensures that tokens from the previous batch can be correctly copied to their target positions when decode tokens are not strictly positioned before prompt tokens.
Key Changes:
- Modified
_prepare_input_idsto optionally return index tensor for reordered batches - Updated
create_unified_batchto accept and use decode indices for correct token placement
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| vllm_gaudi/v1/worker/hpu_model_runner.py | Added return_index parameter to _prepare_input_ids and logic to return index tensor when batch is reordered; updated batch preparation to pass decode index |
| vllm_gaudi/extension/unified_batch.py | Added decode_index parameter and conditional logic to copy tokens using indices instead of assuming sequential placement |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Signed-off-by: Tianmu Li <tianmu.li@intel.com>
✅ CI PassedAll checks passed successfully against the following vllm commit: |
adobrzyn
approved these changes
Dec 3, 2025
tianmu-li
added a commit
to tianmu-li/vllm-gaudi
that referenced
this pull request
Dec 3, 2025
… mixed (vllm-project#642) When decode tokens are not strictly before prompt tokens, tokens from the previous batch cannot be copied using :num_decodes when using async scheduling. --------- Signed-off-by: Tianmu Li <tianmu.li@intel.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
When decode tokens are not strictly before prompt tokens, tokens from the previous batch cannot be copied using :num_decodes when using async scheduling.