diff --git a/onnxruntime/core/providers/cuda/math/topk_impl.cu b/onnxruntime/core/providers/cuda/math/topk_impl.cu
index 9f81f428dd..f24a0f9ffb 100644
--- a/onnxruntime/core/providers/cuda/math/topk_impl.cu
+++ b/onnxruntime/core/providers/cuda/math/topk_impl.cu
@@ -56,14 +56,15 @@ struct NumericLimits<MLFloat16> {
 
 template <typename T>
 __global__ void BitonicTopK(const T* X, T* V, int64_t* I, const TArray<int64_t> elem_nums, size_t size, int32_t axis, int64_t K, int64_t aligned_K, int64_t largest, int64_t sorted, int64_t dimension, int64_t aligned_dimension, T type_min, T type_max) {
-  auto tid = threadIdx.x;
-  auto bid = blockIdx.x;
+  int64_t tid = threadIdx.x;
+  int64_t bid = blockIdx.x;
+  int64_t bdim = blockDim.x;
   extern __shared__ char shared_mem[];
   auto S = (KV<T>*)(shared_mem);
   auto mid_dim = axis == size - 1 ? 1 : elem_nums[axis + 1];
   auto left_dim = bid / mid_dim * elem_nums[axis];
   auto right_dim = axis == size - 1 ? 0 : bid % elem_nums[axis + 1];
-  for (auto i = tid; i < aligned_dimension; i += blockDim.x) {
+  for (auto i = tid; i < aligned_dimension; i += bdim) {
     S[i].key = i < dimension ? X[FROM(i)] : TRIVIAL;
     S[i].val = i;
   }
diff --git a/onnxruntime/core/providers/rocm/rocm_execution_provider.cc b/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
index 0d60177c3b..5ba380f0e4 100644
--- a/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
+++ b/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
@@ -1551,7 +1551,7 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, int32_t, NonZero)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, int64_t, NonZero)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, float, NonZero)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 1, 9, TopK)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 1, 9, TopK)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 8, 8, Scan)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 10, Scan)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 1, 10, Loop)>,
@@ -1581,7 +1581,7 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
     // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, float, ThresholdedRelu)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, double, ThresholdedRelu)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, MLFloat16, ThresholdedRelu)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, 10, TopK)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, 10, TopK)>,
     // BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 1, 10, If)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, int8_t, QuantizeLinear)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 10, uint8_t, QuantizeLinear)>,
@@ -1659,7 +1659,7 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, LogSoftmax)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, 12, Split)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, 12, Squeeze)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, TopK)>,
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, TopK)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, SequenceAt)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, SequenceConstruct)>,
     BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 11, SequenceEmpty)>,
diff --git a/tools/ci_build/amd_hipify.py b/tools/ci_build/amd_hipify.py
index 99de6b213c..bd88a243ed 100644
--- a/tools/ci_build/amd_hipify.py
+++ b/tools/ci_build/amd_hipify.py
@@ -90,10 +90,6 @@ provider_excluded_files = [
                 'math/matmul_integer.h',
                 'math/softmax_impl.cu',
                 'math/softmax.cc',
-                'math/topk.cc',
-                'math/topk.h',
-                'math/topk_impl.cu',
-                'math/topk_impl.h',
                 'nn/batch_norm.cc',
                 'nn/batch_norm.h',
                 'nn/conv.cc',
@@ -244,6 +240,8 @@ def hipify(src_file_path, dst_file_path):
     s = s.replace('std::log', 'logf')
     s = s.replace('#include <cub/device/device_radix_sort.cuh>',
                   '#include <hipcub/hipcub.hpp>\n#include <hipcub/backend/rocprim/device/device_radix_sort.hpp>')
+    s = s.replace('#include "cub/device/device_radix_sort.cuh"',
+                  '#include <hipcub/hipcub.hpp>\n#include <hipcub/backend/rocprim/device/device_radix_sort.hpp>')
     s = s.replace('#include <cub/device/device_reduce.cuh>',
                   '#include <hipcub/backend/rocprim/device/device_reduce.hpp>')
     s = s.replace('#include <cub/device/device_run_length_encode.cuh>',
@@ -254,6 +252,14 @@ def hipify(src_file_path, dst_file_path):
                   '#include <hipcub/backend/rocprim/iterator/counting_input_iterator.hpp>')
     s = s.replace('#include <cub/iterator/discard_output_iterator.cuh>',
                   '#include <hipcub/backend/rocprim/iterator/discard_output_iterator.hpp>')
+    s = s.replace('#include <cub/util_allocator.cuh>',
+                  '#include <hipcub/util_allocator.hpp>')
+    s = s.replace('#include "cub/util_allocator.cuh"',
+                  '#include <hipcub/util_allocator.hpp>')
+    s = s.replace('#include <cub/util_type.cuh>',
+                  '#include <hipcub/backend/rocprim/util_type.hpp>')
+    s = s.replace('#include "cub/util_type.cuh"',
+                  '#include <hipcub/backend/rocprim/util_type.hpp>')
     s = s.replace('typedef half MappedType', 'typedef __half MappedType')
 
     # CUBLAS -> HIPBLAS
@@ -301,6 +307,9 @@ def hipify(src_file_path, dst_file_path):
     s = s.replace('ROCM_VERSION', 'CUDA_VERSION')  # semantically different meanings, cannot hipify
     s = s.replace('__ROCM_ARCH__', '__CUDA_ARCH__')  # semantically different meanings, cannot hipify
 
+    # Deletions
+    s = s.replace('#include "device_atomic_functions.h"', '')  # HIP atomics in main hip header already
+
     do_write = True
     if os.path.exists(dst_file_path):
         with open(dst_file_path, 'r', encoding='utf-8') as fout_old: