Make _prepare_sample non blocking and pin memory of CPU input buffers #2207
Conversation
This file contains 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
|
Looks good - unfortunately we cannot use pinned memory if we are in wsl, can we add a check for that? |
Yard1
reviewed
Dec 19, 2023
Just added. Please check |
Yard1
reviewed
Dec 19, 2023
|
@hanzhi713 A quick question: do you happen to evaluate the performance impact of this change? Just wondering, because the input preparation part only takes 1~3% of the overall running time in our profiling results. |
About 1% for 70B tp4 bs=64. Just a minor optimization. Merge at your discretion 😃 |
Co-authored-by: Antoni Baum <antoni.baum@protonmail.com>
WoosukKwon
reviewed
Dec 20, 2023
WoosukKwon
reviewed
Dec 20, 2023
WoosukKwon
reviewed
Dec 20, 2023
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
WoosukKwon
approved these changes
Dec 20, 2023
hongxiayang
pushed a commit
to hongxiayang/vllm
that referenced
this pull request
Feb 13, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Hide
_prepare_samplelatency with model execution since it looks like it doesn't depend on model forward.We can use a copy stream for h2ds in
_prepare_sample, but we probably don't really need to because these h2ds are very short.cc @Yard1