Using cuda11.8 container for ORT build

Adding 'libcuff-11-8' installation (#5241)

Include cuda 11.8 libraries in system path (#5243)

Mchornyi r23.01 libcufft (#5244)

Adding 'libcuff-11-8' installation

Removing invalid package installation (#5246)

Fix cufft installation (#5247)

build.py

@@ -632,15 +632,15 @@ def onnxruntime_cmake_args(images, library_paths):
         ]
     else:
         if target_platform() == 'windows':
-            if 'base' in images:
+            if 'ort-base' in images:
                 cargs.append(
                     cmake_backend_arg('onnxruntime', 'TRITON_BUILD_CONTAINER',
-                                      None, images['base']))
+                                      None, images['ort-base']))
         else:
-            if 'base' in images:
+            if 'ort-base' in images:
                 cargs.append(
                     cmake_backend_arg('onnxruntime', 'TRITON_BUILD_CONTAINER',
-                                      None, images['base']))
+                                      None, images['ort-base']))
             else:
                 cargs.append(
                     cmake_backend_arg('onnxruntime',
@@ -1026,6 +1026,7 @@ def dockerfile_prepare_container_linux(argmap, backends, enable_gpu,
     if 'onnxruntime' in backends:
         df += '''
 ENV LD_LIBRARY_PATH /opt/tritonserver/backends/onnxruntime:${LD_LIBRARY_PATH}
+ENV LD_LIBRARY_PATH /usr/local/cuda-11.8/lib64:${LD_LIBRARY_PATH}
 '''
 
     backend_dependencies = ""
@@ -1081,6 +1082,10 @@ def dockerfile_prepare_container_linux(argmap, backends, enable_gpu,
 
     if enable_gpu:
         df += install_dcgm_libraries(argmap['DCGM_VERSION'], target_machine)
+        # This is temporary solution to support 23.01
+        df += '''
+RUN apt-get update && apt-get install -y libcufft-11-8
+'''
         df += '''
 # Extra defensive wiring for CUDA Compat lib
 RUN ln -sf ${_CUDA_COMPAT_PATH}/lib.real ${_CUDA_COMPAT_PATH}/lib \
@@ -2102,9 +2107,12 @@ def enable_all():
         fail_if(
             len(parts) != 2,
             '--image must specify <image-name>,<full-image-registry>')
+        # REMOVEME: ONNXRUNTIME 1.13.1 build is failing with cuda 12. Hence, using
+        # cuda 11.8 container as build image for ORT.
         fail_if(
             parts[0] not in [
-                'base', 'gpu-base', 'pytorch', 'tensorflow1', 'tensorflow2'
+                'base', 'gpu-base', 'pytorch', 'tensorflow1', 'tensorflow2',
+                'ort-base'
             ], 'unsupported value for --image')
         log('image "{}": "{}"'.format(parts[0], parts[1]))
         images[parts[0]] = parts[1]