chore: upgrade cutlass to v4.0 (#460)

Author: guocuimi

.github/workflows/build.yml

@@ -11,7 +11,7 @@ on:
       - "**/CMakeLists.txt"
       - ".github/workflows/build.yml"
       - "vcpkg.json"
-    branches: 
+    branches:
       - main
 
   pull_request:
@@ -40,9 +40,9 @@ jobs:
       matrix:
         # Customize the CMake build type here (Release, Debug, RelWithDebInfo, etc.)
         build_type: [Release]
-    
+
     runs-on: [self-hosted, linux, x64, build]
-    
+
     env:
         BUILD_TYPE: ${{ matrix.build_type }}
         # Tells vcpkg where binary packages are stored.

.github/workflows/package_test.yml

@@ -39,7 +39,7 @@ jobs:
       fail-fast: false
       matrix:
         python: ["3.12"]
-        cuda: ["12.6"]
+        cuda: ["12.8"]
         torch: ["2.7.0"]
     runs-on: [self-hosted, linux, build]
     env:

third_party/cutlass

@@ -30,15 +30,15 @@ function usage() {
   echo "Examples:"
   echo "  ${progname} cmake -G Ninja -S . -B build"
   echo "  ${progname} cmake --build build --target all"
-  echo "  ${progname} -ni -d vectorchai/scalellm_devel:cuda12.1 -o '--gpus=all' ctest"
+  echo "  ${progname} -ni -d vectorchai/scalellm_devel:cuda12.8 -o '--gpus=all' ctest"
   echo ""
   exit 1
 }
 
 (( $# < 1 )) && usage
 
 # Default image
-IMAGE="vectorchai/scalellm_devel:cuda12.6"
+IMAGE="vectorchai/scalellm_devel:cuda12.8"
 RUN_OPTS=()
 
 INTERACTIVE=1