diff --git a/cmake/onnxruntime_unittests.cmake b/cmake/onnxruntime_unittests.cmake
index ec396fbcc6117..d5c3af748e528 100644
--- a/cmake/onnxruntime_unittests.cmake
+++ b/cmake/onnxruntime_unittests.cmake
@@ -79,6 +79,9 @@ function(AddTest)
     if (onnxruntime_USE_NCCL)
       target_include_directories(${_UT_TARGET} PRIVATE ${NCCL_INCLUDE_DIRS})
     endif()
+    if(onnxruntime_CUDA_MINIMAL)
+      target_compile_definitions(${_UT_TARGET} PRIVATE -DUSE_CUDA_MINIMAL)
+    endif()
   endif()
   if (onnxruntime_USE_TENSORRT)
     # used for instantiating placeholder TRT builder to mitigate TRT library load/unload overhead
diff --git a/onnxruntime/test/contrib_ops/bias_dropout_op_test.cc b/onnxruntime/test/contrib_ops/bias_dropout_op_test.cc
index f700e31003012..027d4b3fff1b0 100644
--- a/onnxruntime/test/contrib_ops/bias_dropout_op_test.cc
+++ b/onnxruntime/test/contrib_ops/bias_dropout_op_test.cc
@@ -2,7 +2,7 @@
 // Licensed under the MIT License.
 
 // BiasDropout kernel is only implemented for CUDA/ROCM
-#if defined(USE_CUDA) || defined(USE_ROCM)
+#if (defined(USE_CUDA) && !defined(USE_CUDA_MINIMAL)) || defined(USE_ROCM)
 
 #ifdef _MSC_VER
 #pragma warning(disable : 4389)
diff --git a/onnxruntime/test/framework/allocation_planner_test.cc b/onnxruntime/test/framework/allocation_planner_test.cc
index 4e9e80b180e9c..43d3782be3280 100644
--- a/onnxruntime/test/framework/allocation_planner_test.cc
+++ b/onnxruntime/test/framework/allocation_planner_test.cc
@@ -2078,6 +2078,7 @@ TEST(AllocationPlannerTest, ReusedInputCrossDifferentStreams) {
   ASSERT_EQ(plan->allocation_plan[14].alloc_kind, AllocKind::kReuse) << "The input of reshape and gather will reuse the output of shape";
 
   int gather_count = 0;
+  ASSERT_GT(plan->execution_plan.size(), 1) << "Number of execution plans should be greater than 1";
   for (size_t i = 0; i < plan->execution_plan[1]->steps_.size(); i++) {
     if (strstr(typeid(*(plan->execution_plan[1]->steps_[i])).name(), "LaunchKernelStep")) {
       const Node* node = sess.GetSessionState().GetGraphViewer().GetNode(plan->execution_plan[1]->steps_[i]->GetNodeIndex());