Revert "[Klaud Cold] MI300X MiniMax-M3 nightly image and FP8 KV cache"

[cquil11]

Reverts #1837.

What changed

• restore the MI300X MiniMax-M3 STP image to vllm/vllm-openai-rocm:minimax-m3
• restore the default BF16 KV cache
• restore the prior MI300X Slurm exclusion list
• remove the [Klaud Cold] MI300X MiniMax-M3 nightly image and FP8 KV cache #1837 performance changelog entry while preserving later entries

Why

The #1837 merge commit contained [skip-sweep], so its post-merge main workflow skipped the benchmark, collection, and ingestion jobs. This revert allows the change to be reopened and merged through a clean official workflow.

Validation

• bash -n benchmarks/single_node/fixed_seq_len/minimaxm3_fp8_mi300x.sh
• generated targeted MI300X MiniMax-M3 vLLM sweep configuration successfully
• git diff --check

---

Note

Low Risk
Benchmark and launcher config only; restores known-good serving settings and widens the MI300X node pool with no auth or data-path changes.

Overview
Reverts #1837 so the MI300X MiniMax-M3 MXFP8 vLLM recipe matches the prior day-zero setup and can land again through a normal post-merge workflow (the original merge used [skip-sweep]).

minimaxm3-fp8-mi300x-vllm again pins vllm/vllm-openai-rocm:minimax-m3 instead of the ROCm nightly, and config comments call out default BF16 KV cache on gfx942 because the checkpoint lacks calibrated ROCm FP8 attention scales.

minimaxm3_fp8_mi300x.sh drops --kv-cache-dtype fp8 from vllm serve for the same accuracy reason (vLLM’s 1.0 fallback corrupts outputs).

launch_mi300x-amds.sh stops excluding chi-mi300x-121 from Slurm allocation. perf-changelog.yaml removes the #1837 entry for this config while leaving later changelog rows intact.

^{Reviewed by Cursor Bugbot for commit 57d8a51. 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.

---

感谢你的贡献！对于 vLLM 与 SGLang，请确保你的 recipe 与官方 vLLM recipes 和/或 SGLang cookbook 保持一致

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

如果不一致，请先创建一个 PR，之后我们才能将你的单节点 PR 合并到 master 分支。让我们确保文档保持一流水准，使整个 ML 社区都能从你的辛勤工作中受益！谢谢

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。如果选择重新运行失败的任务，PR 作者有责任确保其最终通过。参见 GitHub 关于重新运行失败任务的文档：https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

一般而言，PR 作者应先向相应公司的 CODEOWNERS 请求审阅并获得 PR 批准，然后再请求核心维护者审阅。

如需更多帮助，PR 作者可通过 Slack 联系核心维护者。