Update minimax h100 & h200 vLLM image to v0.18.0

[functionstackx]

Update vllm/vllm-openai image tag from v0.16.0 to v0.18.0 for minimaxm2.5-fp8-h100-vllm and minimaxm2.5-fp8-h200-vllm configs.

Closes #955

Generated with Claude Code

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://cookbook.sglang.io/
• https://docs.vllm.ai/projects/recipes/en/latest/

If it is not, please create a PR first before we can merge your PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sgl-cookbook

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://cookbook.sglang.io/
• https://docs.vllm.ai/projects/recipes/en/latest/

If it is not, please create a PR first before we can merge your PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sgl-cookbook

[functionstackx]

@claude remove disable log requests

vllm: error: unrecognized arguments: --disable-log-requests

https://github.com/SemiAnalysisAI/InferenceX/actions/runs/23635447185/job/68843638787?pr=958

[Klaud-Cold]

Claude finished @functionstackx's task in 2m 0s —— View job

---

Removed --disable-log-requests from vLLM scripts

• Identify where --disable-log-requests is used
• Remove the flag from relevant vLLM benchmark scripts
• Commit and push changes

Removed --disable-log-requests from both minimaxm2.5_fp8_h100.sh and minimaxm2.5_fp8_h200.sh. This flag is no longer recognized in vLLM v0.18.0.
| Branch

[claude]

LGTM — straightforward version bump from v0.16.0 to v0.18.0; the inline comment flags the placeholder in the changelog entry.

Extended reasoning...

Overview

Two-file change: two single-line image tag bumps in .github/configs/nvidia-master.yaml (h100 and h200 minimax configs: v0.16.0 → v0.18.0) and a new changelog entry in perf-changelog.yaml with a placeholder PR link.

Security risks

None. This is a container image version update with no auth, crypto, or permission changes.

Level of scrutiny

Low. Version bumps following the established pattern in this repo (matching recent similar PRs like #950 for kimik2.5-int4-mi355x and #936 for kimik2.5-fp4-mi355x) require minimal scrutiny. The only issue is a nit already captured as an inline comment.

Other factors

The placeholder pull/XXX in perf-changelog.yaml is a pre-existing pattern in this file (several other entries also use it), so while not ideal, it is consistent with existing practice. No prior reviews from me on this PR.

[claude]

🟡 The new changelog entry for minimaxm2.5-fp8-h100-vllm and minimaxm2.5-fp8-h200-vllm uses the placeholder pull/XXX instead of the actual PR number. The pr-link on line 7 should be updated to https://github.com/SemiAnalysisAI/InferenceX/pull/958.

Extended reasoning...

Bug: Placeholder PR link not replaced with actual PR number

The new changelog entry added at the top of perf-changelog.yaml (lines 1-7) contains a placeholder pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX instead of the actual PR number.

How it manifests: The diff clearly shows the new entry was added with the XXX placeholder:

- config-keys:
    - minimaxm2.5-fp8-h100-vllm
    - minimaxm2.5-fp8-h200-vllm
  description:
    - "Update vLLM image from v0.16.0 to v0.18.0 for minimax h100 and h200 configs"
  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX

Why existing code doesn't prevent it: There is no automated validation or CI check that enforces that pr-link values in perf-changelog.yaml must reference real PR numbers rather than the XXX placeholder. The file already contains several other pre-existing XXX placeholders (for entries like dsr1-fp8-h200-sglang, glm5-fp8-mi355x-sglang, minimaxm2.5-fp8-h200-vllm, qwen3.5-bf16-mi325x-sglang, and qwen3.5-fp8-mi325x-sglang), so no pattern-match check exists to flag this.

Why this is different from the other XXX entries: Unlike the pre-existing XXX entries where the PR number may have been unknown at the time of submission, this entry was added as part of PR #958 itself. The PR number was already known and available — it is even referenced in the PR description (Closes #955 and this is PR #958). This is a straightforward oversight where the author forgot to replace the placeholder before submitting.

Impact: The pr-link field is changelog metadata that helps trace which PR introduced each configuration change. A broken link reduces traceability and makes it harder to find the associated discussion, review comments, and rationale for the change. While this does not affect any functionality, it degrades the quality of the changelog as a reference document.

Step-by-step proof:

1. This PR is Update minimax h100 & h200 vLLM image to v0.18.0 #958 (as confirmed by the PR metadata)
2. The PR adds a new entry at the top of perf-changelog.yaml
3. That entry has pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/XXX
4. Visiting the link results in a 404 (or a random PR if XXX was meant as a literal string)
5. The correct link should be https://github.com/SemiAnalysisAI/InferenceX/pull/958

Fix: Replace pull/XXX with pull/958 on line 7 of perf-changelog.yaml.