Skip to content

Conversation

@wuxibin89
Copy link
Collaborator

What does this PR do?

Following #3456, support vllm/sglang DP+EP in server mode.

Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

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

Code Review

This pull request adds support for Data Parallelism (DP) and Expert Parallelism (EP) in vLLM/SGLang server mode. The changes are extensive, touching configuration, testing, and core distributed logic. My review has identified several critical issues, including a syntax error in a shell script, a broken f-string in an assertion, and incorrect usage of uvicorn that would prevent the server from starting. I've also pointed out some areas for improvement in test portability and configuration validation. Please address these critical issues to ensure the new functionality works as expected.

@wuxibin89 wuxibin89 force-pushed the wuxibin/rollout_dp_ep branch from 884392a to d11e124 Compare September 24, 2025 13:41
@wuxibin89 wuxibin89 marked this pull request as ready for review September 25, 2025 08:21
@wuxibin89 wuxibin89 requested a review from ISEEKYAN as a code owner September 25, 2025 08:21
@wuxibin89 wuxibin89 changed the title [WIP][2/N][rollout] feat: support vllm/sglang DP+EP in server mode [2/N][rollout] feat: support vllm/sglang DP+EP in server mode Sep 25, 2025
@wuxibin89
Copy link
Collaborator Author

For vllm, sleep level=2 does not work with expert parallel, fixed in vllm-project/vllm#25458. As a workaround, when EP enabled, sleep(level=1) for now.

@wuxibin89
Copy link
Collaborator Author

For sglang, there's same issue which is already fixed in sgl-project/sglang#8676, we should upgrade sglang to 0.5.2.

vermouth1992 pushed a commit that referenced this pull request Sep 26, 2025
@wuxibin89 wuxibin89 force-pushed the wuxibin/rollout_dp_ep branch from d11e124 to 14dd442 Compare September 26, 2025 03:13
@vermouth1992 vermouth1992 merged commit 84d5619 into volcengine:main Sep 26, 2025
65 of 66 checks passed
from vllm.outputs import RequestOutput
from vllm.utils import FlexibleArgumentParser
from vllm.usage.usage_lib import UsageContext
from vllm.utils import FlexibleArgumentParser, get_tcp_uri
Copy link
Contributor

Choose a reason for hiding this comment

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

get_tcp_uri is introduced in vllm v0.9.0, the installation script needs to be updated.

masoudhashemi pushed a commit to masoudhashemi/verl that referenced this pull request Oct 19, 2025
masoudhashemi pushed a commit to masoudhashemi/verl that referenced this pull request Oct 19, 2025
…gine#3530)

### What does this PR do?

Following volcengine#3456, support
vllm/sglang DP+EP in server mode.
techkang pushed a commit to techkang/verl that referenced this pull request Oct 31, 2025
techkang pushed a commit to techkang/verl that referenced this pull request Oct 31, 2025
…gine#3530)

### What does this PR do?

Following volcengine#3456, support
vllm/sglang DP+EP in server mode.
mtian8 pushed a commit to mtian8/verl that referenced this pull request Nov 1, 2025
mtian8 pushed a commit to mtian8/verl that referenced this pull request Nov 1, 2025
…gine#3530)

### What does this PR do?

Following volcengine#3456, support
vllm/sglang DP+EP in server mode.
wangboxiong320 pushed a commit to wangboxiong320/verl that referenced this pull request Nov 1, 2025
wangboxiong320 pushed a commit to wangboxiong320/verl that referenced this pull request Nov 1, 2025
…gine#3530)

### What does this PR do?

Following volcengine#3456, support
vllm/sglang DP+EP in server mode.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants