Repository navigation
chore(deps): bump sglang from 0.5.9/0.5.10rc0 to 0.5.10 - #1064
Conversation
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughBump SGLang references to stable v0.5.10 across CI/workflow, Docker comment, packaging, and install scripts; remove gRPC repinning in local install; change multimodal embedding handling to use Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 3✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
github-actions
Bot
added
dependencies
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9d52216a4c
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request updates the SGLang dependency to the stable version 0.5.10 across the Docker configuration, pyproject.toml, and CI installation scripts. It also removes the temporary workaround for pinning gRPC packages, as the new SGLang release is now available. I have no feedback to provide as the changes correctly reflect the transition to the stable release.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/release-sglang-docker.yml:
- Line 63: The build matrix for base_image defined in the workflow variable
base_image contains a duplicate "lmsysorg/sglang:v0.5.10" entry causing
identical builds to run twice; update the matrix used in the base_image
expression to remove or replace the duplicate (e.g., change the second
"lmsysorg/sglang:v0.5.10" to a different version like "lmsysorg/sglang:v0.5.9"
or remove it entirely) so the base_image array used by the matrix is unique and
no redundant CI job runs.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 44243f20-e8d8-41e6-9062-749b81081276
📒 Files selected for processing (5)
.github/workflows/release-sglang-docker.ymldocker/engine.Dockerfilegrpc_servicer/pyproject.tomlscripts/ci_install_sglang.shscripts/installation/install-sglang.sh
💤 Files with no reviewable changes (1)
- scripts/installation/install-sglang.sh
SGLang 0.5.10 stable is now released, so we can drop the pre-release pin and related workarounds: - release-sglang-docker.yml: update default base image and matrix from v0.5.9 to v0.5.10 - grpc_servicer/pyproject.toml: change sglang dep from >=0.5.10rc0 to >=0.5.10 - scripts/ci_install_sglang.sh: remove TODO comment and --prerelease flag, pin to stable 0.5.10 - scripts/installation/install-sglang.sh: remove grpcio 1.78.0 pin workaround (no longer needed with 0.5.10) - docker/engine.Dockerfile: update comment example to v0.5.10 Signed-off-by: Simo Lin <simolin@users.noreply.github.com> Signed-off-by: Simo Lin <linsimo.mark@gmail.com>
9d52216 to
aaa08a9
Compare
v0.5.10 is the first version we fully support — drop v0.5.7 and the duplicate v0.5.10 entry from the push-triggered matrix. Signed-off-by: Simo Lin <simolin@users.noreply.github.com> Signed-off-by: Simo Lin <linsimo.mark@gmail.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 78e2ccf8cc
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@scripts/ci_install_sglang.sh`:
- Line 44: The pip install line currently forces prerelease acceptance: update
the command string uv pip install --prerelease=allow "sglang[all]==0.5.10" by
removing the --prerelease=allow flag (so it becomes uv pip install
"sglang[all]==0.5.10") unless you intentionally need prerelease allowance for
transitive dependencies; change the invocation accordingly where this exact
command appears.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: da9a9324-d3f0-4dd8-a989-bdcfd09052a8
📒 Files selected for processing (5)
.github/workflows/release-sglang-docker.ymldocker/engine.Dockerfilegrpc_servicer/pyproject.tomlscripts/ci_install_sglang.shscripts/installation/install-sglang.sh
💤 Files with no reviewable changes (1)
- scripts/installation/install-sglang.sh
SGLang 0.5.10 changed _get_multimodal_inputs to call
MultimodalInputs.from_processor_output() which accesses obj.mm_items
as an attribute. Passing a plain dict {"mm_items": []} causes
AttributeError: 'dict' object has no attribute 'mm_items'.
Since embedding requests without multimodal content don't need
image_inputs, pass None instead. The scheduler's handle_embedding_request
already guards with `if recv_req.image_inputs is not None` and skips
the multimodal path entirely.
Fixes both the health check warmup path (line 354) and the embed
request conversion path (line 830).
Signed-off-by: Simo Lin <simolin@users.noreply.github.com>
Signed-off-by: Simo Lin <linsimo.mark@gmail.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 958d756f33
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Signed-off-by: Chang Su <chang.s.su@oracle.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: fd71251c1f
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
…nputs sglang 0.5.10's unwrap_shm_features() accesses obj.mm_inputs.mm_items via attribute access, but _parse_mm_inputs was returning a plain dict which doesn't support dot notation. This crashed the scheduler for multimodal models (e.g. Qwen3-VL) with: AttributeError: 'dict' object has no attribute 'mm_items' Signed-off-by: Chang Su <chang.s.su@oracle.com>
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@grpc_servicer/smg_grpc_servicer/sglang/servicer.py`:
- Around line 775-776: The docstring for _parse_mm_inputs is stale: update its
return description to state it returns a MultimodalInputs instance (not a dict)
and briefly describe the structure/semantics of the returned MultimodalInputs
object; keep the existing description of the mm_proto parameter and ensure the
one-line summary matches the new return type so callers of _parse_mm_inputs and
tools generating docs won't be misled.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: ASSERTIVE
Plan: Pro
Run ID: 0a73d0ff-a135-4972-9651-11eb024136f1
📒 Files selected for processing (1)
grpc_servicer/smg_grpc_servicer/sglang/servicer.py
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6b1f1e5d44
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Signed-off-by: Chang Su <chang.s.su@oracle.com>
The go-bindings-e2e job was not collecting worker logs on failure, making it impossible to diagnose why sglang workers crash. Add E2E_LOG_DIR and artifact upload steps matching e2e-gpu-job.yml. Signed-off-by: Chang Su <chang.s.su@oracle.com>
sglang 0.5.10 explicitly disables multimodal for Llama4ForConditionalGeneration by default (mm_disabled_models list in model_config.py). Without --enable-multimodal, the vision encoder is not loaded, causing the scheduler to crash with "no embedding method found for Modality.IMAGE" when processing image inputs. Signed-off-by: Chang Su <chang.s.su@oracle.com>
Signed-off-by: Chang Su <chang.s.su@oracle.com>
This reverts commit ae7db73. Signed-off-by: Chang Su <chang.s.su@oracle.com>
Proto3 defaults unset numeric fields to 0, causing repetition_penalty to be sent as 0.0 instead of the correct default 1.0. sglang 0.5.10 added BatchedRepetitionPenalizer which divides logits by this value, causing division by zero and a CUDA crash. Signed-off-by: key4ng <keyang@lightseek.org> Signed-off-by: key4ng <rukeyang@gmail.com>
|
Update the TRT-LLM Docker release workflow: - Default base image: 1.3.0rc8 → 1.3.0rc11 - Release matrix: [rc8, 1.2.0, 1.1.0] → [rc11, rc10, rc9] - Description examples updated to rc11 Follows the same pattern as the recent sglang (#1064) and vllm (#1066) engine version bumps. Signed-off-by: Simo Lin <linsimo.mark@gmail.com>
Update the TRT-LLM Docker release workflow: - Default base image: 1.3.0rc8 → 1.3.0rc10 - Release matrix: [rc8, 1.2.0, 1.1.0] → [rc10, rc9, rc8] - Description examples updated to rc10 Note: 1.3.0rc11 does not exist on nvcr.io yet. Follows the same pattern as the recent sglang (#1064) and vllm (#1066) engine version bumps. Signed-off-by: Simo Lin <linsimo.mark@gmail.com>
3 participants
Summary
SGLang 0.5.10 stable is now released. This bumps all references from 0.5.9/0.5.10rc0 to 0.5.10 and removes related workarounds.
Changes
.github/workflows/release-sglang-docker.yml: update default base image and build matrix from v0.5.9 to v0.5.10grpc_servicer/pyproject.toml:sglang>=0.5.10rc0→sglang>=0.5.10scripts/ci_install_sglang.sh: remove TODO comment and--prereleaseflag, pin to stable 0.5.10scripts/installation/install-sglang.sh: remove grpcio 1.78.0 pin workaround (no longer needed)docker/engine.Dockerfile: update comment example to v0.5.10Not updated
1.3.3-sglang-v0.5.9— no1.3.3-sglang-v0.5.10image exists yet. Will update when the next SMG release builds against 0.5.10.getting-started/index.md) — same reason, image tags reference 1.3.3.Test Plan
Summary by CodeRabbit