Skip to content
This repository was archived by the owner on Apr 20, 2026. It is now read-only.

Fix sglang recipes for sglang v0.5.5#96

Merged
ishandhanani merged 1 commit intoishandhanani:mainfrom
nvcastet:fix_recipes
Jan 23, 2026
Merged

Fix sglang recipes for sglang v0.5.5#96
ishandhanani merged 1 commit intoishandhanani:mainfrom
nvcastet:fix_recipes

Conversation

@nvcastet
Copy link
Copy Markdown
Contributor

@nvcastet nvcastet commented Jan 23, 2026

sglang v0.5.5 uses SGLANG_CUTEDSL_MOE_NVFP4_DISPATCH instead of SGLANG_MOE_NVFP4_DISPATCH, setting both to support older sglang versions.

Summary by CodeRabbit

  • Chores
    • Updated GB200-FP4 recipe configurations to improve compatibility with older sglang versions by adding environment variable support across multiple deployment profiles.

✏️ Tip: You can customize this high-level summary in your review settings.

@coderabbitai
Copy link
Copy Markdown
Contributor

coderabbitai bot commented Jan 23, 2026

Caution

Review failed

The pull request is closed.

📝 Walkthrough

Walkthrough

Add environment variable SGLANG_CUTEDSL_MOE_NVFP4_DISPATCH with value "1" across GB200-FP4 recipe YAML configurations for backward compatibility with older sglang versions, including minor whitespace normalization in two files.

Changes

Cohort / File(s) Summary
1k1k variant recipes
recipes/gb200-fp4/1k1k/max-tpt-mtp.yaml, recipes/gb200-fp4/1k1k/max-tpt.yaml, recipes/gb200-fp4/1k1k/mid-curve.yaml
Add SGLANG_CUTEDSL_MOE_NVFP4_DISPATCH: "1" to prefill and/or decode environment blocks for older sglang version compatibility
1k8k variant recipes
recipes/gb200-fp4/1k8k/max-tpt.yaml, recipes/gb200-fp4/1k8k/mid-curve.yaml
Add SGLANG_CUTEDSL_MOE_NVFP4_DISPATCH: "1" to decode_environment sections
8k1k variant recipes
recipes/gb200-fp4/8k1k/max-tpt.yaml, recipes/gb200-fp4/8k1k/mid-curve.yaml
Add SGLANG_CUTEDSL_MOE_NVFP4_DISPATCH: "1" to decode_environment; normalize trailing whitespace across configuration lines

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~5 minutes

Possibly related PRs

Suggested reviewers

  • ishandhanani

Poem

🐰 A backward glance through config files,
One variable whispered across the aisles,
Older sglang versions now at rest,
With NVFP4 dispatch—put to the test!


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands and usage tips.

@ishandhanani ishandhanani merged commit 0dbeb25 into ishandhanani:main Jan 23, 2026
1 check was pending
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants