Skip to content

Conversation

panpan0000
Copy link
Contributor

@panpan0000 panpan0000 commented May 15, 2025

FIX #NA (link existing issues this PR will resolve)

As KVCache-Sharing/Offloading become more essential in production, but the current default image vllm/vllm-openai:v0.8.* doesn't include the kv-connector dependencies, for example , when run vllm image with --kv-transfer-config '{"kv_connector":"LMCacheConnector","kv_role":"kv_both","kv_rank":0,"kv_parallel_size":2}'

[engine.py:448]   File "/usr/local/lib/python3.12/dist-packages/vllm/distributed/kv_transfer/kv_connector/lmcache_connector.py", line 37, in __init__
[engine.py:448]     from lmcache.experimental.cache_engine import LMCacheEngineBuilder 
[engine.py:448] ModuleNotFoundError: No module named 'lmcache'

So introducing a new requirements.txt for kv-connector(put lmcache for now, mooncake & nixl later)

And I enable kv-connector into dockerfile in CI scripts, I'm not sure if it's proper or maybe we should do it in vllm-production-stack project? Please correct me.

since we don't have a up to date CI image for vllm+lmcache, the image lmcache/vllm-openai:latest seems still using old vllm(v0.7.0) and image not build from a trackable github CI. @KuntaiDu , any suggestion?

Copy link

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

@mergify mergify bot added the ci/build label May 15, 2025
@panpan0000 panpan0000 force-pushed the kv-cache-in-dockerfile branch from 4f33c86 to 5a4aeab Compare May 15, 2025 07:55
@panpan0000
Copy link
Contributor Author

panpan0000 commented Jun 3, 2025

@YuhanLiu11 What's your suggestion ?

@YuhanLiu11
Copy link

@panpan0000 this is a great idea! Also, now LMCache has bdist package and it can be easily installed without extra compilation.
cc @ApostaC

@KuntaiDu
Copy link
Collaborator

KuntaiDu commented Jun 3, 2025

Yeah agree --- and we can contribute some perf CIs for CPU offloading and KV cache sharing. @youkaichao @simon-mo @DarkLight1337 what's your thoughts on this?

@panpan0000
Copy link
Contributor Author

So how to push thing forward?

@KuntaiDu
Copy link
Collaborator

So how to push thing forward?

I am pinging people for feedbacks. Could you please fix the merge conflict first so that we can start the full CI?

@panpan0000 panpan0000 force-pushed the kv-cache-in-dockerfile branch from 5a4aeab to 3bbaded Compare June 30, 2025 01:58
@panpan0000
Copy link
Contributor Author

So how to push thing forward?

I am pinging people for feedbacks. Could you please fix the merge conflict first so that we can start the full CI?

Thanks for reminder. done.

Copy link
Collaborator

@KuntaiDu KuntaiDu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

@KuntaiDu KuntaiDu added the ready ONLY add when PR is ready to merge/full CI is needed label Jun 30, 2025
@KuntaiDu
Copy link
Collaborator

I am running some test now to make sure compatibility issues and other things, will get back to you soon.

Copy link
Collaborator

@simon-mo simon-mo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit
s/INSTALL_KV_CONNECTOR/INSTALL_KV_CONNECTORS

@panpan0000 panpan0000 force-pushed the kv-cache-in-dockerfile branch from 3bbaded to 76d6761 Compare July 2, 2025 12:42
@panpan0000
Copy link
Contributor Author

nit s/INSTALL_KV_CONNECTOR/INSTALL_KV_CONNECTORS

done :-) Thanks @simon-mo

@KuntaiDu
Copy link
Collaborator

KuntaiDu commented Jul 2, 2025

I tested this PR and it is functional.

@KuntaiDu KuntaiDu enabled auto-merge (squash) July 2, 2025 18:24
@KuntaiDu
Copy link
Collaborator

KuntaiDu commented Jul 2, 2025

@panpan0000 I enabled automerge, could you please fix the linting errors? Many thanks for your contribution!

Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
auto-merge was automatically disabled July 3, 2025 03:16

Head branch was pushed to by a user without write access

@panpan0000 panpan0000 force-pushed the kv-cache-in-dockerfile branch from 76d6761 to 5cee0d5 Compare July 3, 2025 03:16
@vllm-bot vllm-bot merged commit 5561681 into vllm-project:main Jul 4, 2025
103 of 105 checks passed
sfeng33 pushed a commit to sfeng33/vllm that referenced this pull request Jul 6, 2025
huydhn pushed a commit to huydhn/vllm that referenced this pull request Jul 8, 2025
Chen-zexi pushed a commit to Chen-zexi/vllm that referenced this pull request Jul 13, 2025
patrickvonplaten pushed a commit to patrickvonplaten/vllm that referenced this pull request Jul 15, 2025
vllm-project#18193)

Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
LyrisZhong pushed a commit to LyrisZhong/vllm that referenced this pull request Jul 23, 2025
avigny pushed a commit to avigny/vllm that referenced this pull request Jul 31, 2025
vllm-project#18193)

Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: avigny <47987522+avigny@users.noreply.github.com>
Pradyun92 pushed a commit to Pradyun92/vllm that referenced this pull request Aug 6, 2025
npanpaliya pushed a commit to odh-on-pz/vllm-upstream that referenced this pull request Aug 6, 2025
jinzhen-lin pushed a commit to jinzhen-lin/vllm that referenced this pull request Aug 9, 2025
vllm-project#18193)

Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>
googlercolin pushed a commit to googlercolin/vllm that referenced this pull request Aug 29, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
ci/build ready ONLY add when PR is ready to merge/full CI is needed
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants