.

Dockerfile

@@ -15,7 +15,7 @@ ENV USE_CUDA=1 \
 ARG MAX_JOBS=90                       
 ENV DEBIAN_FRONTEND=noninteractive \
     PYTHONUNBUFFERED=1 LANG=C.UTF-8 LC_ALL=C.UTF-8 \
-    TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6;9.0"
+    TORCH_CUDA_ARCH_LIST="7.5;8.0;8.6;8.9;9.0"
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
       python3 python3-dev python3-pip python3-distutils git cmake ninja-build \
@@ -84,7 +84,7 @@ RUN python3 -m pip install ".[srt,openai]" --no-build-isolation && \
 
 
 # ── 🔄 下载 sgl-kernel（与 sglang 同步）───────────────────────────────────────
-RUN pip download --only-binary=:all: --no-deps sgl-kernel==0.1.9 -d /tmp/sgl_kernel_wheels
+RUN pip download --only-binary=:all: --no-deps sgl-kernel==0.3.7 -d /tmp/sgl_kernel_wheels
 
 # ── 收集所有 wheel 到 /wheels ──────────────────────────────────────────────
 RUN mkdir -p /wheels && \
@@ -128,9 +128,6 @@ COPY --from=builder-extras /usr/local/cuda/lib64/libcupti.so /usr/lib/x86_64-lin
 # 👇建议在后面补上
 RUN ldconfig
 
-# ---- 拷贝预调优的 MoE Triton kernel config ----------------------------
-COPY moe_kernels /usr/local/lib/python3.10/dist-packages/sglang/srt/layers/moe/fused_moe_triton/configs
-
 
 COPY --from=builder-extras /wheels /tmp/wheels
 
@@ -158,6 +155,9 @@ ENV PROMETHEUS_MULTIPROC_DIR=/tmp/prometheus
 # ✅ 确保目录存在
 RUN mkdir -p /tmp/prometheus
 
+# ---- 拷贝预调优的 MoE Triton kernel config ----------------------------
+COPY moe_kernels /usr/local/lib/python3.10/dist-packages/sglang/srt/layers/moe/fused_moe_triton/configs
+
 # ✅ 添加 Tini（推荐）
 ENV TINI_VERSION=v0.19.0
 ADD https://github.com/krallin/tini/releases/download/${TINI_VERSION}/tini /tini