From d9a2e3b5e56ed73f841dab283663707088c987c8 Mon Sep 17 00:00:00 2001
From: shishi <shishi@meta-stone.com>
Date: Mon, 3 Aug 2026 17:22:04 +0800
Subject: [PATCH] =?UTF-8?q?fix(910c/glm52):=20=E9=95=9C=E5=83=8F=E4=B8=8E?=
 =?UTF-8?q?=20client=20=E4=BF=AE=E6=AD=A3=20-=20=E9=83=A8=E7=BD=B2?=
 =?UTF-8?q?=E5=9C=A8=20910c.2=20=E7=9A=84=20GLM5.2-tuned=20=E9=95=9C?=
 =?UTF-8?q?=E5=83=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- profile DOCKER_IMAGE 改回 local/vllm-ascend:0.23-a3-20260718-sglang
  (910c.1 的 glm5.2-a3-openeuler 缺 expert_map_manager 模块无法启动)
- config.env DOCKER_CLIENT_IMAGE 指向 910c.2 本地 tuned 镜像(带 bench_serving)
- 实测: 910c.2 TP8/DP2 smoke 40/40, TTFT 1062ms, TPOT 51ms
---
 deploy/profiles/910c/glm52_910c_vllm_tp_dp_matrix.env    | 2 +-
 experiments/910c/glm52_910c_vllm_tp_dp_matrix/config.env | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/deploy/profiles/910c/glm52_910c_vllm_tp_dp_matrix.env b/deploy/profiles/910c/glm52_910c_vllm_tp_dp_matrix.env
index 7582abf..99f2195 100644
--- a/deploy/profiles/910c/glm52_910c_vllm_tp_dp_matrix.env
+++ b/deploy/profiles/910c/glm52_910c_vllm_tp_dp_matrix.env
@@ -9,7 +9,7 @@ EXPERIMENT=glm52_910c_vllm_tp_dp_matrix
 MODEL_NAME=GLM-5.2
 ENGINE=vllm
 RUNTIME=docker
-DOCKER_IMAGE=quay.io/ascend/vllm-ascend:glm5.2-a3-openeuler
+DOCKER_IMAGE=local/vllm-ascend:0.23-a3-20260718-sglang
 CONTAINER_NAME=vllm-ascend-glm52-910c_tp${TP}_dp${DP}
 MODEL_PATH=/mnt/models/GLM-5.2-w4a8c8
 SERVED_MODEL_NAME=glm-5.2
diff --git a/experiments/910c/glm52_910c_vllm_tp_dp_matrix/config.env b/experiments/910c/glm52_910c_vllm_tp_dp_matrix/config.env
index 87005c0..251a38b 100644
--- a/experiments/910c/glm52_910c_vllm_tp_dp_matrix/config.env
+++ b/experiments/910c/glm52_910c_vllm_tp_dp_matrix/config.env
@@ -33,7 +33,7 @@ DOCKER_IMAGE="${DOCKER_IMAGE:-local/vllm-ascend:0.23-a3-20260718-sglang}"
 # (local/vllm-ascend:0.23-a3-*-sglang, see envs/ASCEND_910C_ENV_SETUP.md §8.1)
 # carry sglang 0.5.2 bench_serving, so the client runs as a standalone container
 # (TORCH_DEVICE_BACKEND_AUTOLOAD=0 injected by the runner, no NPU needed).
-DOCKER_CLIENT_IMAGE="${DOCKER_CLIENT_IMAGE:-local/vllm-ascend:0.23-a3-dsv4-sglang}"
+DOCKER_CLIENT_IMAGE="${DOCKER_CLIENT_IMAGE:-local/vllm-ascend:0.23-a3-20260718-sglang}"
 USE_DOCKER_CLIENT="${USE_DOCKER_CLIENT:-1}"
 
 # Device selection. ASCEND_VISIBLE_DEVICES selects NPU cards 0..7; the Ascend