[Fix] Fix major performance bug in certain cases #1563
Merged
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
Ying1123
force-pushed
the
ying-perf-fix
branch
2 times, most recently
from
1559c1e to
e6113bd
Compare
Ying1123
changed the title
Ying1123
force-pushed
the
ying-perf-fix
branch
8 times, most recently
from
eba3ac8 to
f58a5cf
Compare
Ying1123
force-pushed
the
ying-perf-fix
branch
from
f58a5cf to
5dc697c
Compare
Ying1123
force-pushed
the
ying-perf-fix
branch
from
5dc697c to
a465494
Compare
Ying1123
force-pushed
the
ying-perf-fix
branch
from
a465494 to
88c6701
Compare
hnyls2002
approved these changes
Oct 4, 2024
Ying1123
force-pushed
the
ying-perf-fix
branch
from
e6376fd to
3834869
Compare
hnyls2002
force-pushed
the
ying-perf-fix
branch
from
4dd9f38 to
7ec9132
Compare
timethink
pushed a commit
to timethink/sglang
that referenced
this pull request
Mar 9, 2025
Co-authored-by: hnyls2002 <[email protected]>
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Outdated suggestions cannot be applied.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.
The issue could happen in non-streaming scenarios with small batch sizes. It could break continuous batching because of the wrong manipulation of
do_not_get_new_batch.This PR fixes the issue.