feat: support uniform cross-layer KV cache layout for CPU offloading#361
Open
lianghao208 wants to merge 1 commit into
Open
feat: support uniform cross-layer KV cache layout for CPU offloading#361lianghao208 wants to merge 1 commit into
lianghao208 wants to merge 1 commit into
Conversation
lianghao208
force-pushed
the
support_uniform_kvcache
branch
from
4d05151 to
c314a74
Compare
lianghao208
force-pushed
the
support_uniform_kvcache
branch
from
c314a74 to
2733377
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
#269 introduces a approach for supporting CPU offloading in kvcached, it forces vLLM to fall back on per-layer KV cache management rather than a uniform, all-layer strategy, which results in implicit extra overhead.
This PR proposes a new approach that allows CPU offloading services to utilize the kvcached VMM API to allocate cross-layer contiguous memory layout. This effectively preserves system elasticity without sacrificing performance.