feat: add cu128 identifier for sgl-kernel (#5287)

.github/workflows/release-whl-kernel-cu128.yml

@@ -14,7 +14,7 @@ on:
 jobs:
   build-wheels:
     if: github.repository == 'sgl-project/sglang'
-    runs-on: ubuntu-latest
+    runs-on: sgl-kernel-build-node
     strategy:
       matrix:
         python-version: ['3.9']

sgl-kernel/build.sh

@@ -4,7 +4,6 @@ set -ex
 PYTHON_VERSION=$1
 CUDA_VERSION=$2
 PYTHON_ROOT_PATH=/opt/python/cp${PYTHON_VERSION//.}-cp${PYTHON_VERSION//.}
-ENABLE_SM90A=$(( ${CUDA_VERSION%.*} >= 12 ? ON : OFF ))
 
 if [ ${CUDA_VERSION} = "12.8" ]; then
    DOCKER_IMAGE="pytorch/manylinux2_28-builder:cuda${CUDA_VERSION}"

sgl-kernel/rename_wheels.sh

@@ -5,13 +5,17 @@ WHEEL_DIR="dist"
 
 wheel_files=($WHEEL_DIR/*.whl)
 for wheel in "${wheel_files[@]}"; do
-    new_wheel="${wheel/linux/manylinux2014}"
+    intermediate_wheel="${wheel/linux/manylinux2014}"
+
+    if ls /usr/local/ | grep -q "12.8"; then
+        new_wheel="${intermediate_wheel/-cp39/+cu128-cp39}"
+    else
+        new_wheel="$intermediate_wheel"
+    fi
 
     if [[ "$wheel" != "$new_wheel" ]]; then
         echo "Renaming $wheel to $new_wheel"
         mv -- "$wheel" "$new_wheel"
-
     fi
 done
-
 echo "Wheel renaming completed."