From 97045041f5c0dbe62b2d1baef9b56e05b84a5958 Mon Sep 17 00:00:00 2001 From: "claude[bot]" Date: Sat, 5 Sep 2026 21:02:53 +0000 Subject: [PATCH 1/2] chore(amd): update dsr1-fp8-mi325x-sglang-mtp image to v0.5.19-rocm700-mi30x MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Bump the dsr1-fp8-mi325x-sglang-mtp master family image from lmsysorg/sglang:v0.5.12-rocm700-mi30x to lmsysorg/sglang:v0.5.19-rocm700-mi30x. No other fields or families change. 将 dsr1-fp8-mi325x-sglang-mtp 家族的镜像从 lmsysorg/sglang:v0.5.12-rocm700-mi30x 更新为 lmsysorg/sglang:v0.5.19-rocm700-mi30x。不改动其他字段或家族。 Co-Authored-By: Claude Fable 5.1 --- configs/amd-master.yaml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/configs/amd-master.yaml b/configs/amd-master.yaml index 7fa80d605..6e27d3cdb 100644 --- a/configs/amd-master.yaml +++ b/configs/amd-master.yaml @@ -1286,7 +1286,7 @@ dsv4-fp4-mi355x-atom-mtp: # works only because it's non-MTP (no draft KV layer). MTP stays tp8-only. dsr1-fp8-mi325x-sglang-mtp: - image: lmsysorg/sglang:v0.5.12-rocm700-mi30x + image: lmsysorg/sglang:v0.5.19-rocm700-mi30x model: deepseek-ai/DeepSeek-R1-0528 model-prefix: dsr1 runner: mi325x From 1fceee6e753e4cf7959fee2e52bd46fcc276feec Mon Sep 17 00:00:00 2001 From: "claude[bot]" Date: Sat, 5 Sep 2026 21:36:20 +0000 Subject: [PATCH 2/2] chore(changelog): add dsr1-fp8-mi325x-sglang-mtp v0.5.19 image entry MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Append the perf-changelog.yaml entry for the dsr1-fp8-mi325x-sglang-mtp image update to lmsysorg/sglang:v0.5.19-rocm700-mi30x (PR #2844). 为 dsr1-fp8-mi325x-sglang-mtp 更新至 lmsysorg/sglang:v0.5.19-rocm700-mi30x 的镜像变更追加 perf-changelog.yaml 条目（PR #2844）。 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 912728b2e..98cd5bb79 100644 --- a/perf-changelog.yaml +++ b/perf-changelog.yaml @@ -6902,3 +6902,9 @@ - "Lower mem-fraction-static from 0.89 to 0.86 on every arm: TP4, TP8 and TP8 with DP attention. swa-full-tokens-ratio becomes per-arm: 0.10 on the tensor-parallel arms as before, 0.15 under DP attention." - "Drop concurrency 2 and 10 from the TP4 arm, leaving [1, 4, 8], and drop concurrency 16 from the TP8 hicache arm, leaving [32, 48]. Concurrency 16 remains on the TP8 no-offload arm." pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2800 + +- config-keys: + - dsr1-fp8-mi325x-sglang-mtp + description: + - "Update SGLang image from v0.5.12-rocm700-mi30x to v0.5.19-rocm700-mi30x" + pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/2844