[codex] Update MiniMax M3 B200 FlashInfer image

[cquil11]

Summary

• update minimaxm3-fp8-b200-vllm to vllm/vllm-openai:minimax-m3-0618
• enable FlashInfer TRT-LLM attention with FP8 indexer KV and KV cache
• use the image's default FlashInfer TRT-LLM MoE backend instead of explicitly forcing Marlin

Validation

• benchmark script passes bash -n
• MiniMax M3 B200-DGXC vLLM sweep generation succeeds

---

Note

Medium Risk
Changes inference stack (attention/MoE/KV dtypes) and mutates installed vLLM source at job start; benchmark behavior and numerics may shift versus the prior Marlin image.

Overview
Aligns MiniMax M3 B200 single-node benchmarking with the newer minimax-m3-0618-x86_64-cu130 vLLM image (replacing vllm/vllm-openai:minimax-m3 in nvidia-master.yaml).

The B200 launch script now matches the B300 recipe: a startup Python patch on sparse_attention_msa.py adds .contiguous() on sliced prefill top-k indices so MSA CSR construction works under TP with CUDA graphs; TP-only runs drop --moe-backend marlin in favor of the image default FlashInfer TRT-LLM MoE; vllm serve gains FlashInfer TRT-LLM attention, FP8 indexer KV, and FP8 KV cache. perf-changelog.yaml documents the config change.

^{Reviewed by Cursor Bugbot for commit e57575d. Bugbot is set up for automated code reviews on this repo. Configure here.}

[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://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node 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/sglang/tree/main/docs_new

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27792967822
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27792967822

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27796120896
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27796120896

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 10af6e2. Configure here.}

[cursor]

Changelog not appended at end

High Severity

The new minimaxm3-fp8-b200-vllm block is inserted immediately before the existing trailing minimaxm3-fp8-mi300x-vllm entry instead of after it. Repository changelog validation requires append-only edits with unchanged historical bytes, so this diff fails check-changelog and can block sweep setup.

 

^{Reviewed by Cursor Bugbot for commit 10af6e2. Configure here.}

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=27805371252
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=27805371252

[Oseltamivir]

lgtm

[Oseltamivir]

/reuse-sweep-run 27805371252

[github-actions]

see unofficial run visualizer at https://inferencex.semianalysis.com/inference?unofficialRun=28000080406
see unofficial run visualizer at https://inferencex.semianalysis.com/evaluation?unofficialRun=28000080406