Support page size > 1 #4356
Merged
Support page size > 1 #4356
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
merrymercy
requested review from
Ying1123,
zhyncs,
hnyls2002,
ispobock,
HaiShaw and
ByronHsu
as code owners
merrymercy
force-pushed
the
lianmin/page-size
branch
from
7b15727 to
3b48346
Compare
yangw1234
reviewed
Mar 13, 2025
merrymercy
force-pushed
the
lianmin/page-size
branch
from
fdabcd0 to
bb81342
Compare
67 tasks
This was referenced Mar 13, 2025
hebiao064
pushed a commit
to hebiao064/sglang
that referenced
this pull request
Mar 13, 2025
6 tasks
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Outdated suggestions cannot be applied.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.
PagedTokenToKVPoolAllocator. It has almost the same interface as the oldTokenToKVPoolAllocator, but it guarantees that the returned kv cache locations of one request are always page aligned. It is compatible with prefix cache.batch_is_fulltoScheduleBatch)Page size = 1, 4, 8, 16 shows no perf diff, because the page table management is all overlapped with GPU computation.
This PR will enable the integration of more attention kernels (e.g., FlashMLA and CuDNN attention)
Todos for the next PRs: