From e7b92e1c89825630be2d95075b01474091bcad95 Mon Sep 17 00:00:00 2001 From: "claude[bot]" <41898282+claude[bot]@users.noreply.github.com> Date: Mon, 21 Sep 2026 06:58:08 +0000 Subject: [PATCH 1/2] perf(h200): update MiniMax-M3 FP8 vLLM AgentX image to nightly-3df4ae15 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Update minimaxm3-fp8-h200-vllm-agentic-mtp from vllm/vllm-openai:nightly-d9105ea8001e0a6d77a96327d17515bb5791fb36 (2026-09-07 nightly, tag no longer published on Docker Hub) to vllm/vllm-openai:nightly-3df4ae153eb385e27b52f26c81f8edb9e20b9984 (2026-09-21 nightly), pinned by its manifest-list digest. The recipe script, search space, evals and golden acceptance length are unchanged. 将 minimaxm3-fp8-h200-vllm-agentic-mtp 的 vLLM 镜像从 vllm/vllm-openai:nightly-d9105ea8001e0a6d77a96327d17515bb5791fb36 （2026-09-07 nightly，该 tag 已不再在 Docker Hub 发布）更新为 vllm/vllm-openai:nightly-3df4ae153eb385e27b52f26c81f8edb9e20b9984 （2026-09-21 nightly），并按 manifest-list digest 固定。配方脚本、 搜索空间、评测与 golden 接受长度均保持不变。 Co-Authored-By: Claude Fable 5.1 --- configs/nvidia-master.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/configs/nvidia-master.yaml b/configs/nvidia-master.yaml index c5d5edac68..00473162a0 100644 --- a/configs/nvidia-master.yaml +++ b/configs/nvidia-master.yaml @@ -5162,7 +5162,7 @@ minimaxm3-fp8-h100-vllm-agentic-mtp: - { tp: 8, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: mooncake, version: "0.3.11.post1" }, conc-list: [6, 8] } minimaxm3-fp8-h200-vllm-agentic-mtp: - image: vllm/vllm-openai:nightly-d9105ea8001e0a6d77a96327d17515bb5791fb36 + image: vllm/vllm-openai:nightly-3df4ae153eb385e27b52f26c81f8edb9e20b9984@sha256:1e1f56a164a3dfdf87a57168983602421a295b0b3570d27c5b49c54b97b341b4 model: MiniMaxAI/MiniMax-M3-MXFP8 model-prefix: minimaxm3 runner: cluster:h200-dgxc From d98915cf1841eb6e7ea679739b2a0f4f85dccbfc Mon Sep 17 00:00:00 2001 From: "claude[bot]" <41898282+claude[bot]@users.noreply.github.com> Date: Mon, 21 Sep 2026 09:05:15 +0000 Subject: [PATCH 2/2] chore(changelog): record MiniMax-M3 FP8 H200 vLLM image update MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Append the perf-changelog entry for minimaxm3-fp8-h200-vllm-agentic-mtp after the updated-image smoke run (run 35570778795) passed c1/c12 throughput and both minimax_m3_full evals. 为 minimaxm3-fp8-h200-vllm-agentic-mtp 追加 perf-changelog 条目； 更新镜像的 smoke 运行（run 35570778795）已通过 c1/c12 吞吐与两项 minimax_m3_full 评测。 Co-Authored-By: Claude Fable 5.1 --- perf-changelog.yaml | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/perf-changelog.yaml b/perf-changelog.yaml index c5b4309f5f..f32b100f13 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -8455,3 +8455,9 @@ description: - "Update B200 vLLM AgentX to DSpark6 and a new image with TP8 and DEP8 configurations." pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3274 + +- config-keys: + - minimaxm3-fp8-h200-vllm-agentic-mtp + description: + - "Update vLLM image from nightly-d9105ea8 to nightly-3df4ae15 (2026-09-21 upstream nightly, pinned by digest)." + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/3327