Skip to content

[Docs][EC Connector] CPU EC Connector usage Docs - #54522

Merged
ywang96 merged 4 commits into
vllm-project:mainfrom
omerpaz95:worktree-doc_cpu_ec_connector
Sep 9, 2026
Merged

ywang96 merged 4 commits into
vllm-project:mainfrom
omerpaz95:worktree-doc_cpu_ec_connector

Conversation

@omerpaz95

Copy link
Copy Markdown
Contributor

Purpose

Introducing a basic usage manual to the CPU EC Connector.
This includes feature flags, environment variables, basic setup, NIXL-based EC sharing setup, protocol description.

Signed-off-by: omerpaz95 <omerpaz95@gmail.com>

@claude claude Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

@mergify

mergify Bot commented Aug 31, 2026

Copy link
Copy Markdown
Contributor

Documentation preview: https://vllm--54522.org.readthedocs.build/en/54522/

@mergify mergify Bot added the documentation Improvements or additions to documentation label Aug 31, 2026
Signed-off-by: omerpaz95 <omerpaz95@gmail.com>
Comment thread docs/features/ec_cpu_connector.md Outdated
}'
```

`ec_role="ec_producer"` alone already sets `vllm_config.is_encoder_only=True` (skips the language model, sampler, and pooler); add `--mm-encoder-only` only if you need encoder-only execution independent of `ec_transfer_config`.

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

This is now vllm_config.is_mm_encoder_only on main.

Comment thread docs/features/ec_cpu_connector.md Outdated
| `ec_connector` | `str \| None` | `None` | Connector class name. Use `"ECCPUConnector"`. |
| `ec_role` | `"ec_producer" \| "ec_consumer" \| "ec_both" \| None` | `None` | Required whenever `ec_connector` is set. `ec_producer` offloads GPU→CPU only, `ec_consumer` reloads CPU→GPU only, `ec_both` does both in the same process. |
| `ec_connector_extra_config` | `dict[str, Any]` | `{}` | Connector-specific settings, including `ec_enable_nixl` — see [`ec_connector_extra_config` Reference](#ec_connector_extra_config-reference). |
| `engine_id` | `str \| None` | random UUID4 | Names the shared mmap file: `/dev/shm/vllm_ec_{engine_id}_dp{dp_rank}.mmap`. Also seeds the NIXL agent name when `ec_enable_nixl=True`. |

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

The mmap filename now uses vllm_config.instance_id, not ec_transfer_config.engine_id. The latter is still used for the NIXL agent name.

@gty111 gty111 left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

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

LGTM

@gty111 gty111 added the ready ONLY add when PR is ready to merge/full CI is needed label Sep 9, 2026
@gty111

gty111 commented Sep 9, 2026

Copy link
Copy Markdown
Collaborator

/ci run

@github-actions

github-actions Bot commented Sep 9, 2026

Copy link
Copy Markdown

✅ @omerpaz95, CI is now available for this PR.

  • /ci run starts upstream CI; /amd-ci run starts AMD CI only.
  • /ci retry retries failed jobs in the CI build for the current PR head. If the current head has no CI build, it starts a new CI build for the current head containing only jobs that failed in the latest earlier CI build for this PR.
  • /amd-ci retry retries failed jobs in AMD CI for the current PR head. Use /amd-ci run when the current head has no AMD CI build.
  • /ci cancel cancels scheduled or running CI builds for this PR branch; /amd-ci cancel does the same for AMD CI only.

@github-actions

github-actions Bot commented Sep 9, 2026

Copy link
Copy Markdown

✅ Triggered Buildkite CI #87798 for commit 783f258f102b.

@ywang96
ywang96 merged commit a97dacb into vllm-project:main Sep 9, 2026
16 checks passed
ItsRoy69 pushed a commit to ItsRoy69/vllm that referenced this pull request Sep 10, 2026
Signed-off-by: omerpaz95 <omerpaz95@gmail.com>
Signed-off-by: Jyotirmoy Roy <jyotirmoyroy649@gmail.com>
michael-gpu pushed a commit to practical-inference/vllm-ampere-extended that referenced this pull request Oct 20, 2026
Signed-off-by: omerpaz95 <omerpaz95@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

build-docs documentation Improvements or additions to documentation 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.

3 participants