From 5169262fa692e3b09666e6e91a689ef8c225d69b Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sun, 17 May 2026 19:45:57 -0400
Subject: [PATCH 1/3] [Klaud Cold] Update dsv4-fp8-h200-sglang (+mtp) SGLang
 image to v0.5.12-cu130

---
 .github/configs/nvidia-master.yaml | 4 ++--
 perf-changelog.yaml                | 7 +++++++
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/.github/configs/nvidia-master.yaml b/.github/configs/nvidia-master.yaml
index 37dd5af3f2..4537658730 100644
--- a/.github/configs/nvidia-master.yaml
+++ b/.github/configs/nvidia-master.yaml
@@ -2914,7 +2914,7 @@ dsv4-fp8-h200-vllm-agentic:
 # --speculative-config '{"method":"mtp","num_speculative_tokens":2}'.
 
 dsv4-fp8-h200-sglang:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:7f19c6dc092e47a10fac2e41f47eab78970280d06648b8e50d312a82f0ae722f
+  image: lmsysorg/sglang:v0.5.12-cu130
   model: deepseek-ai/DeepSeek-V4-Pro
   model-prefix: dsv4
   runner: h200-dgxc
@@ -2938,7 +2938,7 @@ dsv4-fp8-h200-sglang:
 # runner pool, search space) and adds EAGLE speculative decoding via
 # --speculative-algorithm EAGLE with the (3,1,4) chain matching dsv4-fp4-b300-sglang-mtp.
 dsv4-fp8-h200-sglang-mtp:
-  image: lmsysorg/sglang:deepseek-v4-hopper@sha256:7f19c6dc092e47a10fac2e41f47eab78970280d06648b8e50d312a82f0ae722f
+  image: lmsysorg/sglang:v0.5.12-cu130
   model: deepseek-ai/DeepSeek-V4-Pro
   model-prefix: dsv4
   runner: h200-dgxc
diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index 13a695453a..d4b3a63880 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -2629,3 +2629,10 @@
   description:
     - "Update vLLM ROCm image from v0.18.0 to v0.21.0"
   pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1404
+
+- config-keys:
+    - dsv4-fp8-h200-sglang
+    - dsv4-fp8-h200-sglang-mtp
+  description:
+    - "Update SGLang image from SHA-pinned deepseek-v4-hopper custom build (15/14d old) to v0.5.12-cu130"
+  pr-link: PLACEHOLDER

From 2044102d8234bd5ac8001a2659fe2a37263d5e54 Mon Sep 17 00:00:00 2001
From: functionstackx <47992694+functionstackx@users.noreply.github.com>
Date: Sun, 17 May 2026 19:46:00 -0400
Subject: [PATCH 2/3] chore: fill pr-link for #1460

---
 perf-changelog.yaml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/perf-changelog.yaml b/perf-changelog.yaml
index d4b3a63880..8b5511d23b 100644
--- a/perf-changelog.yaml
+++ b/perf-changelog.yaml
@@ -2635,4 +2635,4 @@
     - dsv4-fp8-h200-sglang-mtp
   description:
     - "Update SGLang image from SHA-pinned deepseek-v4-hopper custom build (15/14d old) to v0.5.12-cu130"
-  pr-link: PLACEHOLDER
+  pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/1460

From 6d2b2bef0838a8be08081485740cd57f31466817 Mon Sep 17 00:00:00 2001
From: Oseltamivir <58582368+Oseltamivir@users.noreply.github.com>
Date: Thu, 21 May 2026 17:00:25 -0700
Subject: [PATCH 3/3] fix: raise dsv4 h200 sglang memory fraction

---
 benchmarks/single_node/dsv4_fp8_h200_sglang.sh     | 2 +-
 benchmarks/single_node/dsv4_fp8_h200_sglang_mtp.sh | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/benchmarks/single_node/dsv4_fp8_h200_sglang.sh b/benchmarks/single_node/dsv4_fp8_h200_sglang.sh
index bf5c6f7b21..a669528790 100644
--- a/benchmarks/single_node/dsv4_fp8_h200_sglang.sh
+++ b/benchmarks/single_node/dsv4_fp8_h200_sglang.sh
@@ -43,7 +43,7 @@ PYTHONNOUSERSITE=1 sglang serve \
     --chunked-prefill-size 4096 \
     --disable-flashinfer-autotune \
     --disable-radix-cache \
-    --mem-fraction-static 0.88 \
+    --mem-fraction-static 0.92 \
     --max-running-requests "$(( CONC * 3 / 2 > 8 ? CONC * 3 / 2 : 8 ))" \
     $EVAL_CONTEXT_ARGS >> $SERVER_LOG 2>&1 &
 
diff --git a/benchmarks/single_node/dsv4_fp8_h200_sglang_mtp.sh b/benchmarks/single_node/dsv4_fp8_h200_sglang_mtp.sh
index bcba415435..d50c184d7d 100644
--- a/benchmarks/single_node/dsv4_fp8_h200_sglang_mtp.sh
+++ b/benchmarks/single_node/dsv4_fp8_h200_sglang_mtp.sh
@@ -43,7 +43,7 @@ PYTHONNOUSERSITE=1 sglang serve \
     --chunked-prefill-size 4096 \
     --disable-flashinfer-autotune \
     --disable-radix-cache \
-    --mem-fraction-static 0.88 \
+    --mem-fraction-static 0.92 \
     --max-running-requests "$(( CONC * 3 / 2 > 8 ? CONC * 3 / 2 : 8 ))" \
     --speculative-algorithm EAGLE \
     --speculative-num-steps 3 \