default native cuda architecture when possible

CMakeLists.txt

@@ -19,9 +19,11 @@ set(GGML_CCACHE ON)
 set(GGML_BACKEND_DL ON)
 set(GGML_BACKEND_SHARED ON)
 set(GGML_SCHED_MAX_COPIES 4)
+
+set(GGML_LLAMAFILE ON)
 set(GGML_CPU_ALL_VARIANTS ON)
 set(GGML_CUDA_PEER_MAX_BATCH_SIZE 128)
-set(GGML_LLAMAFILE ON)
+set(GGML_CUDA_GRAPHS ON)
 
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
@@ -37,6 +39,10 @@ set_property(TARGET ggml PROPERTY EXCLUDE_FROM_ALL TRUE)
 
 check_language(CUDA)
 if(CMAKE_CUDA_COMPILER)
+    if(CMAKE_VERSION VERSION_GREATER_EQUAL "3.24" AND NOT CMAKE_CUDA_ARCHITECTURES)
+        set(CMAKE_CUDA_ARCHITECTURES "native")
+    endif()
+
     add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/ml/backend/ggml/ggml/src/ggml-cuda)
 endif()