Update NVIDIA DeepSeek sglang Docker image from v0.5.5 to v0.5.6

[Copilot]

Updates all NVIDIA DeepSeek sglang configurations to use the latest Docker image version.

Changes

• Updated Docker image reference from lmsysorg/sglang:v0.5.5-cu129-amd64 to lmsysorg/sglang:v0.5.6-cu129-amd64 in .github/configs/nvidia-master.yaml
• Affected configurations:
  • dsr1-fp4-b200-sglang
  • dsr1-fp8-b200-sglang
  • dsr1-fp8-h200-sglang

Original prompt

---

This section details on the original issue you should resolve

<issue_title>update all nvidia DeepSeek sglang from lmsysorg/sglang:v0.5.5-cu129-amd64 to lmsysorg/sglang:v0.5.6-cu129-amd64 in .github/configs/nvidia-master.yaml</issue_title>
<issue_description>https://hub.docker.com/layers/lmsysorg/sglang/v0.5.6-cu129-arm64/images/sha256-5d7c5dccf0e384f88a2f2cdfb807644e6d079bdd040942c96b49b8c1a1acb6a7</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes InferenceMAX/InferenceMAX#275

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

---

Note

Updates NVIDIA DeepSeek SGLang configs to latest Docker image.

• In .github/configs/nvidia-master.yaml, change image from lmsysorg/sglang:v0.5.5-cu129-amd64 to lmsysorg/sglang:v0.5.6-cu129-amd64 for dsr1-fp4-b200-sglang, dsr1-fp8-b200-sglang, and dsr1-fp8-h200-sglang
• Append corresponding entry to perf-changelog.yaml documenting the image bump and PR link

^{Written by Cursor Bugbot for commit e55faff. This will update automatically on new commits. Configure here.}

[cquil11]

@kedarpotdar-nv cc @ankursingh-nv results here https://github.com/InferenceMAX/InferenceMAX/actions/runs/19933312619
is this good with yall?

[cquil11]

lgtm

[kedarpotdar-nv]

We're reviewing - please hold on merging.

[cquil11]

@kedarpotdar-nv @ankursingh-nv

Reminder:

PR 267 has been merged. With this. sweeps will no longer run nightly, rather they will run only when necessary as indicated by the perf-changelog.yaml file at the root of the repo. Going forward, when developers make changes to configs that have performance impact, they must note that change in perf-changelog.yaml and give a brief description of the changes. Once their PR is ready for review, they can add the sweep-enabled label to trigger a test sweep on their local branch. Once everything looks good, they can merge to main and an official sweep will be run for the specified configs.

So for this PR, you will add something like the following entry to the bottom of perf-changelog.yaml:

- config-keys:
    - dsr1-fp4-b200-sglang
    - dsr1-fp8-b200-sglang
    - dsr1-fp8-h200-sglang
  description: |
    - Update NVIDIA DeepSeek sglang Docker image from v0.5.5 to v0.5.6
    PR: https://github.com/InferenceMAX/InferenceMAX/pull/276

Then add the sweep-enabled tag to the PR after marking it ready for review to run a test sweep. After the test sweep is done, please link the run in your PR description.

[cquil11]

please see comment about perf changelog

[cquil11]

@ankursingh-nv where are we on this? I made changes to perf changelog and kicked off test run
https://github.com/InferenceMAX/InferenceMAX/actions/runs/20286929996

[ankursingh-nv]

Thanks @cquil11, a few jobs failing. Need to look into it before we merge.

[cursor]

Cleanup command accidentally removed from workflow

The line sudo find . -not -user $USER -exec rm -rf {} + was removed from the cleanup section, but this change is not mentioned in the PR description which only discusses Docker image version updates. The identical cleanup command still exists in benchmark-multinode-tmpl.yml, suggesting this removal is accidental. This command removes files not owned by the current user before checkout, and its absence could cause permission-related issues during git checkout or subsequent workflow runs.

 

[cquil11]

lgtm