Skip to content

[Klaud Cold] Update minimaxm2.5-fp8-b300-vllm vLLM image to v0.21.0 #1454

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
functionstackx merged 3 commits into from
May 18, 2026
+7 −1
Merged

[Klaud Cold] Update minimaxm2.5-fp8-b300-vllm vLLM image to v0.21.0 #1454

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
e4a190b
Update minimaxm2.5-fp8-b300-vllm vLLM image to v0.21.0
functionstackx May 17, 2026
a1d561e
chore: fill pr-link for #1454
functionstackx May 17, 2026
3c9f5d8
Merge branch 'main' into update-minimaxm2.5-fp8-b300-vllm-v0.21.0
functionstackx May 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/configs/nvidia-master.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4303,7 +4303,7 @@ minimaxm2.5-fp8-b200-vllm-agentic:
# MiniMax-M2.5 FP8 B200 vLLM recipe as-is until B300-specific tuning is available.

minimaxm2.5-fp8-b300-vllm:
image: vllm/vllm-openai:v0.19.0-cu130
image: vllm/vllm-openai:v0.21.0

@claude claude Bot May 17, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

🟡 The divergence comment block at .github/configs/nvidia-master.yaml:4329-4332 becomes stale after this bump: it still claims the b300-vllm entry is 'left identical to origin/main' (no longer true) and records the divergence as 'v0.19.0-cu130 -> v0.19.1' when the actual divergence is now 'v0.21.0 -> v0.19.1' (with b300-vllm-agentic at line 4334 still pinned to v0.19.1). Either update the comment to reflect the new divergence, or reconsider whether bumping b300-vllm while leaving b300-vllm-agentic pinned is intended.

Extended reasoning...

What the bug is\n\nThis PR bumps minimaxm2.5-fp8-b300-vllm at .github/configs/nvidia-master.yaml:4306 from vllm/vllm-openai:v0.19.0-cu130 to vllm/vllm-openai:v0.21.0, but does not touch the adjacent divergence comment block at lines 4329-4332 that documents this exact image. The b300-vllm-agentic sibling at line 4334 remains pinned to vllm/vllm-openai:v0.19.1.\n\n### Why both claims in the comment are now wrong\n\nThe comment currently reads:\n\nyaml\n# Diverged from minimaxm2.5-fp8-b300-vllm (agentic-coding sibling). Reasons below;\n# the original minimaxm2.5-fp8-b300-vllm entry is left identical to origin/main so\n# its fixed-seq-len sweep is unaffected.\n# - image: 'vllm/vllm-openai:v0.19.0-cu130' -> 'vllm/vllm-openai:v0.19.1'\n\n\n1. Claim 1 — 'left identical to origin/main': This PR explicitly modifies the b300-vllm entry, so it is no longer identical to origin/main. The rationale for forking the -agentic sibling (preserving an unchanged fixed-seq-len sweep history on the upstream image) is no longer being honored.\n2. Claim 2 — divergence delta: The recorded delta 'v0.19.0-cu130' -> 'v0.19.1' is now inaccurate on both sides: the upstream b300-vllm is now v0.21.0 (different version, -cu130 suffix dropped), so the actual divergence becomes v0.21.0 -> v0.19.1.\n\n### Step-by-step proof\n\n1. git blame shows the comment was introduced by commit e92a9bf9 (Cameron Quilici, 'Migrate agentic-coding benchmarks to aiperf v0.2'). At that time, b300-vllm was on v0.19.0-cu130 and the -agentic sibling was intentionally forked to v0.19.1 to keep the original sweep on the upstream image.\n2. This PR changes line 4306 image: vllm/vllm-openai:v0.19.0-cu130 → image: vllm/vllm-openai:v0.21.0.\n3. Line 4334 (b300-vllm-agentic image) is untouched by this PR — it still reads vllm/vllm-openai:v0.19.1.\n4. Lines 4329-4332 are also untouched — they still describe the prior divergence delta and assert b300-vllm is identical to origin/main.\n5. Therefore, after the merge, the comment misrepresents both the divergence delta (says v0.19.0-cu130 -> v0.19.1, actually v0.21.0 -> v0.19.1) and the design invariant (says b300-vllm is unchanged from origin/main, but it just changed).\n\n### Impact and fix\n\nThe comment is pure YAML #-prefixed text — there is no runtime impact and the sweep configuration itself is consistent. However, the comment actively misleads future readers and reviewers about why the -agentic sibling exists as a separate entry. The fix is to either (a) update the comment block to record the new v0.21.0 -> v0.19.1 divergence and drop the 'identical to origin/main' wording, or (b) also bump b300-vllm-agentic to v0.21.0 in this PR so the divergence is closed and the comment can be removed entirely. The b200 sibling block at lines 4278-4282 is unaffected.

model: MiniMaxAI/MiniMax-M2.5
model-prefix: minimaxm2.5
runner: b300
Expand Down
6 changes: 6 additions & 0 deletions perf-changelog.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2629,3 +2629,9 @@
description:
- "Update vLLM ROCm image from v0.18.0 to v0.21.0"
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1404

- config-keys:
- minimaxm2.5-fp8-b300-vllm
description:
- "Update vLLM image from v0.19.0-cu130 (26d old) to v0.21.0"
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1454