diff --git a/CMakeLists.txt b/CMakeLists.txt
index 4ee6210..00147e6 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -52,17 +52,7 @@ endif()
 set(TRITON_TENSORRT_BACKEND_LIBNAME triton_tensorrt)
 set(TRITON_TENSORRT_BACKEND_INSTALLDIR ${CMAKE_INSTALL_PREFIX}/backends/tensorrt)
 
-set(RHEL_NVINFER_PATH "")
-if(LINUX)
-  file(STRINGS "/etc/os-release" DISTRO_ID_LIKE REGEX "ID_LIKE")
-  if(${DISTRO_ID_LIKE} MATCHES "rhel|centos")
-    if(${CMAKE_SYSTEM_PROCESSOR} STREQUAL "aarch64")
-      set(RHEL_NVINFER_PATH "/usr/local/cuda/targets/sbsa-linux/lib")
-    else()
-      set(RHEL_NVINFER_PATH "/usr/local/cuda/targets/x86_64-linux/lib")
-    endif()
-  endif(${DISTRO_ID_LIKE} MATCHES "rhel|centos")
-endif(LINUX)
+set(CMAKE_PREFIX_PATH "/usr/local/cuda/targets/sbsa-linux/lib;/usr/local/cuda/targets/x86_64-linux/lib")
 
 #
 # Dependencies
@@ -259,8 +249,8 @@ ENDFOREACH(p)
 
 # NOTE: TRT 10 for Windows added the version suffix to the library names. See the release notes:
 # https://docs.nvidia.com/deeplearning/tensorrt/release-notes/index.html#tensorrt-10
-find_library(NVINFER_LIBRARY NAMES nvinfer nvinfer_10 PATHS ${RHEL_NVINFER_PATH})
-find_library(NVINFER_PLUGIN_LIBRARY NAMES nvinfer_plugin nvinfer_plugin_10 PATHS ${RHEL_NVINFER_PATH})
+find_library(NVINFER_LIBRARY NAMES nvinfer nvinfer_10)
+find_library(NVINFER_PLUGIN_LIBRARY NAMES nvinfer_plugin nvinfer_plugin_10)
 target_link_libraries(
   triton-tensorrt-backend
   PRIVATE