Enable Xeon optimizations like Tensor Parallel and AMX from vLLM 0.10.0 #2106
+77
−1
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
Dependency Review✅ No vulnerabilities or license issues found.Scanned FilesNone |
This was referenced Jul 1, 2025
louie-tsai
force-pushed
the
vllm-optimize
branch
2 times, most recently
from
cccd778 to
7072d2c
Compare
chensuyue
reviewed
Jul 1, 2025
chensuyue
reviewed
Jul 1, 2025
louie-tsai
force-pushed
the
vllm-optimize
branch
3 times, most recently
from
3740497 to
2ee36fc
Compare
This was
linked to
issues
Jul 9, 2025
|
This PR is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 7 days. |
louie-tsai
force-pushed
the
vllm-optimize
branch
3 times, most recently
from
8e0c6d5 to
b19733c
Compare
louie-tsai
force-pushed
the
vllm-optimize
branch
from
b19733c to
9169670
Compare
louie-tsai
changed the title
louie-tsai
force-pushed
the
vllm-optimize
branch
3 times, most recently
from
b5d3362 to
f484e04
Compare
|
all docker instances up and healthy on local env. original compose.yaml also works |
louie-tsai
force-pushed
the
vllm-optimize
branch
from
f484e04 to
6d82af0
Compare
louie-tsai
force-pushed
the
vllm-optimize
branch
2 times, most recently
from
42b33da to
bb1c060
Compare
louie-tsai
requested review from
ZePan110,
ftian1,
lkk12014402,
minmin-intel and
rbrugaro
as code owners
louie-tsai
force-pushed
the
vllm-optimize
branch
2 times, most recently
from
b90bd59 to
83668ea
Compare
letonghan
approved these changes
Aug 15, 2025
louie-tsai
force-pushed
the
vllm-optimize
branch
3 times, most recently
from
9b82ce3 to
368c98a
Compare
ZePan110
approved these changes
Aug 15, 2025
louie-tsai
force-pushed
the
vllm-optimize
branch
6 times, most recently
from
e87a6d9 to
fab7efa
Compare
|
@chensuyue 
|
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
Signed-off-by: Tsai, Louie <louie.tsai@intel.com> Signed-off-by: Tsai, Louie <louie.tsai@intel.com>
louie-tsai
force-pushed
the
vllm-optimize
branch
from
fab7efa to
39f7187
Compare
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.
Description
Added an additional compose.perf.yaml file, when users want to have more vLLM optimization, they just apply with one more yaml file during docker compose
docker compose -f compose.yaml -f compose.perf.yaml upit includes most of the Xeon optimizations from public vLLM 0.9.2 which plan to be release this week.
Assume that we use a system with 2 NUMA nodes and AMX support.
Issues
[#2045 ]
[#2044 ]
Type of change
List the type of change like below. Please delete options that are not relevant.
Dependencies
vLLM 0.9.2 release
https://github.com/vllm-project/vllm/releases/tag/v0.9.2
https://gallery.ecr.aws/q9t5s3a7/vllm-cpu-release-repo
Tests
Throughput speedup is ~1.3X. TTFT and TPOT are reduced to 75% comparing to original OPEA vLLM
perf_comparison.html