diff --git a/.clang-tidy b/.clang-tidy
index 7e2313b94c06..45b4374aa09c 100644
--- a/.clang-tidy
+++ b/.clang-tidy
@@ -42,7 +42,6 @@ misc-*,
 -misc-non-private-member-variables-in-classes,
 -misc-confusable-identifiers,
 modernize-*,
--modernize-concat-nested-namespaces,
 -modernize-macro-to-enum,
 -modernize-return-braced-init-list,
 -modernize-use-auto,
diff --git a/c10/core/WrapDimMinimal.cpp b/c10/core/WrapDimMinimal.cpp
index 2375dc3ac5cf..c95753432580 100644
--- a/c10/core/WrapDimMinimal.cpp
+++ b/c10/core/WrapDimMinimal.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/WrapDimMinimal.h>
 
-namespace c10 {
-namespace detail {
+namespace c10::detail {
 
 template <typename T>
 T maybe_wrap_dim_slow(T dim, T dim_post_expr, bool wrap_scalar) {
@@ -40,5 +39,4 @@ maybe_wrap_dim_slow(int64_t dim, int64_t dim_post_expr, bool wrap_scalar);
 template C10_API SymInt
 maybe_wrap_dim_slow(SymInt dim, SymInt dim_post_expr, bool wrap_scalar);
 
-} // namespace detail
-} // namespace c10
+} // namespace c10::detail
diff --git a/c10/core/impl/DeviceGuardImplInterface.cpp b/c10/core/impl/DeviceGuardImplInterface.cpp
index cc54c54c9708..581f32f1e130 100644
--- a/c10/core/impl/DeviceGuardImplInterface.cpp
+++ b/c10/core/impl/DeviceGuardImplInterface.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/impl/DeviceGuardImplInterface.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // NOLINTNEXTLINE(modernize-avoid-c-arrays,cppcoreguidelines-avoid-c-arrays)
 std::atomic<const DeviceGuardImplInterface*>
@@ -14,5 +13,4 @@ DeviceGuardImplRegistrar::DeviceGuardImplRegistrar(
   device_guard_impl_registry[static_cast<size_t>(type)].store(impl);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/FakeGuardImpl.h b/c10/core/impl/FakeGuardImpl.h
index c9b472156edd..c8bfe91619ed 100644
--- a/c10/core/impl/FakeGuardImpl.h
+++ b/c10/core/impl/FakeGuardImpl.h
@@ -4,8 +4,7 @@
 
 #include <array>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // FakeGuardImpl is hardcoded to have eight devices.  Not for
 // any good reason, just to simplify code.
@@ -100,5 +99,4 @@ template <DeviceType T>
 thread_local std::array<StreamId, kFakeGuardImplMaxDevices>
     FakeGuardImpl<T>::current_streams_ = {0, 0, 0, 0, 0, 0, 0, 0};
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/GPUTrace.cpp b/c10/core/impl/GPUTrace.cpp
index 403f1e65ce56..ac1f08d7e091 100644
--- a/c10/core/impl/GPUTrace.cpp
+++ b/c10/core/impl/GPUTrace.cpp
@@ -1,8 +1,7 @@
 #include <c10/core/impl/GPUTrace.h>
 #include <c10/util/CallOnce.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 std::atomic<const PyInterpreter*> GPUTrace::gpuTraceState{nullptr};
 
@@ -16,5 +15,4 @@ void GPUTrace::set_trace(const PyInterpreter* trace) {
   });
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/GPUTrace.h b/c10/core/impl/GPUTrace.h
index 377af88be034..3acb875b54a3 100644
--- a/c10/core/impl/GPUTrace.h
+++ b/c10/core/impl/GPUTrace.h
@@ -2,8 +2,7 @@
 
 #include <c10/core/impl/PyInterpreter.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct C10_API GPUTrace {
   // On the x86 architecture the atomic operations are lock-less.
@@ -26,5 +25,4 @@ struct C10_API GPUTrace {
   }
 };
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/HermeticPyObjectTLS.cpp b/c10/core/impl/HermeticPyObjectTLS.cpp
index a7eb89430be8..ae1f6fd8ee6b 100644
--- a/c10/core/impl/HermeticPyObjectTLS.cpp
+++ b/c10/core/impl/HermeticPyObjectTLS.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/impl/HermeticPyObjectTLS.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local std::atomic<bool> hermeticPyObjectState{false};
 
@@ -19,5 +18,4 @@ void HermeticPyObjectTLS::init_state() {
   haveState_ = true;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/LocalDispatchKeySet.cpp b/c10/core/impl/LocalDispatchKeySet.cpp
index 0b6cf6fa66f8..0b3d87fce241 100644
--- a/c10/core/impl/LocalDispatchKeySet.cpp
+++ b/c10/core/impl/LocalDispatchKeySet.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/impl/LocalDispatchKeySet.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 // NB: POD, must be zero initialized!
 // Note [TLS Initialization]
@@ -115,5 +114,4 @@ bool tls_is_dispatch_keyset_excluded(DispatchKeySet ks) {
 bool tls_is_dispatch_keyset_included(DispatchKeySet ks) {
   return raw_local_dispatch_key_set.included().isSupersetOf(ks);
 }
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/PyInterpreter.cpp b/c10/core/impl/PyInterpreter.cpp
index f555ee5c3458..dc0ec52760c5 100644
--- a/c10/core/impl/PyInterpreter.cpp
+++ b/c10/core/impl/PyInterpreter.cpp
@@ -2,8 +2,7 @@
 #include <c10/core/TensorImpl.h>
 #include <c10/core/impl/PyInterpreter.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct NoopPyInterpreterVTable final : public PyInterpreterVTable {
   std::string name() const override {
@@ -125,5 +124,4 @@ void PyInterpreter::disarm() noexcept {
   vtable_ = &noop_vtable;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/PyObjectSlot.cpp b/c10/core/impl/PyObjectSlot.cpp
index ac6d5915afb6..400903bc7a65 100644
--- a/c10/core/impl/PyObjectSlot.cpp
+++ b/c10/core/impl/PyObjectSlot.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/impl/PyObjectSlot.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 PyObjectSlot::PyObjectSlot() : pyobj_interpreter_(nullptr), pyobj_(nullptr) {}
 
@@ -71,5 +70,4 @@ void PyObjectSlot::set_owns_pyobj(bool b) {
       reinterpret_cast<uintptr_t>(_unchecked_untagged_pyobj()) | b);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/PythonDispatcherTLS.cpp b/c10/core/impl/PythonDispatcherTLS.cpp
index 5ca57a9da466..a53e293eaa32 100644
--- a/c10/core/impl/PythonDispatcherTLS.cpp
+++ b/c10/core/impl/PythonDispatcherTLS.cpp
@@ -2,8 +2,7 @@
 #include <c10/core/impl/LocalDispatchKeySet.h>
 #include <c10/core/impl/PythonDispatcherTLS.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local PyInterpreter* pythonDispatcherState;
 
@@ -27,5 +26,4 @@ void PythonDispatcherTLS::reset_state() {
       DispatchKey::PythonDispatcher, false);
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/PythonDispatcherTLS.h b/c10/core/impl/PythonDispatcherTLS.h
index 8998935279ee..9016c3e11e15 100644
--- a/c10/core/impl/PythonDispatcherTLS.h
+++ b/c10/core/impl/PythonDispatcherTLS.h
@@ -3,8 +3,7 @@
 #include <c10/core/impl/PyInterpreter.h>
 #include <c10/macros/Export.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 struct C10_API PythonDispatcherTLS {
   static void set_state(PyInterpreter* state);
@@ -22,5 +21,4 @@ struct C10_API DisablePythonDispatcher {
   PyInterpreter* old_;
 };
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/SizesAndStrides.cpp b/c10/core/impl/SizesAndStrides.cpp
index ca859ae0681d..6c63109ea5af 100644
--- a/c10/core/impl/SizesAndStrides.cpp
+++ b/c10/core/impl/SizesAndStrides.cpp
@@ -1,7 +1,6 @@
 #include <c10/core/impl/SizesAndStrides.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 void SizesAndStrides::resizeSlowPath(
     const size_t newSize,
@@ -77,5 +76,4 @@ void SizesAndStrides::resizeSlowPath(
   size_ = newSize;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/TorchDispatchModeTLS.cpp b/c10/core/impl/TorchDispatchModeTLS.cpp
index 1ee5aa85cc23..37c75003e200 100644
--- a/c10/core/impl/TorchDispatchModeTLS.cpp
+++ b/c10/core/impl/TorchDispatchModeTLS.cpp
@@ -6,8 +6,7 @@
 
 #include <utility>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 thread_local TorchDispatchModeTLS torchDispatchModeState;
 
@@ -192,5 +191,4 @@ std::string to_string(TorchDispatchModeKey mode_key) {
   }
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/core/impl/TorchDispatchModeTLS.h b/c10/core/impl/TorchDispatchModeTLS.h
index 55d0d92b230f..50a92459e61e 100644
--- a/c10/core/impl/TorchDispatchModeTLS.h
+++ b/c10/core/impl/TorchDispatchModeTLS.h
@@ -3,8 +3,7 @@
 #include <c10/core/SafePyObject.h>
 #include <c10/macros/Export.h>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 enum class TorchDispatchModeKey : int8_t {
   FAKE,
@@ -61,5 +60,4 @@ C10_API bool dispatch_mode_enabled();
 
 C10_API std::string to_string(TorchDispatchModeKey mode_key);
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/cuda/CUDAAllocatorConfig.cpp b/c10/cuda/CUDAAllocatorConfig.cpp
index 5a709f981b0e..eb7d8d49754f 100644
--- a/c10/cuda/CUDAAllocatorConfig.cpp
+++ b/c10/cuda/CUDAAllocatorConfig.cpp
@@ -6,9 +6,7 @@
 #include <c10/cuda/driver_api.h>
 #endif
 
-namespace c10 {
-namespace cuda {
-namespace CUDACachingAllocator {
+namespace c10::cuda::CUDACachingAllocator {
 
 constexpr size_t kRoundUpPowerOfTwoIntervals = 16;
 
@@ -354,6 +352,4 @@ void setAllocatorSettings(const std::string& env) {
   CUDACachingAllocator::CUDAAllocatorConfig::instance().parseArgs(env.c_str());
 }
 
-} // namespace CUDACachingAllocator
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::CUDACachingAllocator
diff --git a/c10/cuda/CUDACachingAllocator.cpp b/c10/cuda/CUDACachingAllocator.cpp
index ebe1fd2b72e8..f44703c5a9a0 100644
--- a/c10/cuda/CUDACachingAllocator.cpp
+++ b/c10/cuda/CUDACachingAllocator.cpp
@@ -40,8 +40,7 @@ namespace c10 {
 
 C10_DEFINE_REGISTRY(FreeCudaMemoryCallbacksRegistry, FreeMemoryCallback);
 
-namespace cuda {
-namespace CUDACachingAllocator {
+namespace cuda::CUDACachingAllocator {
 
 // Included here as this is externally used in CUDAAllocatorConfig
 const size_t kLargeBuffer =
@@ -3347,6 +3346,6 @@ struct BackendStaticInitializer {
 std::atomic<CUDAAllocator*> allocator;
 BackendStaticInitializer backend_static_initializer;
 
-} // namespace CUDACachingAllocator
-} // namespace cuda
+} // namespace cuda::CUDACachingAllocator
+
 } // namespace c10
diff --git a/c10/cuda/CUDADeviceAssertionHost.cpp b/c10/cuda/CUDADeviceAssertionHost.cpp
index 1a3b2a68c69c..b954dcf26ebe 100644
--- a/c10/cuda/CUDADeviceAssertionHost.cpp
+++ b/c10/cuda/CUDADeviceAssertionHost.cpp
@@ -26,8 +26,7 @@
         false);                                                     \
   } while (0)
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 namespace {
 
@@ -343,5 +342,4 @@ bool CUDAKernelLaunchRegistry::has_failed() const {
   return false;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
diff --git a/c10/cuda/CUDAException.cpp b/c10/cuda/CUDAException.cpp
index 24f3d928af69..2d725747c969 100644
--- a/c10/cuda/CUDAException.cpp
+++ b/c10/cuda/CUDAException.cpp
@@ -6,8 +6,7 @@
 
 #include <string>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 void c10_cuda_check_implementation(
     const int32_t err,
@@ -44,5 +43,4 @@ void c10_cuda_check_implementation(
   TORCH_CHECK(false, check_message);
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
diff --git a/c10/cuda/CUDAMallocAsyncAllocator.cpp b/c10/cuda/CUDAMallocAsyncAllocator.cpp
index a4503e383039..b80c91d00819 100644
--- a/c10/cuda/CUDAMallocAsyncAllocator.cpp
+++ b/c10/cuda/CUDAMallocAsyncAllocator.cpp
@@ -9,10 +9,7 @@
 #include <unordered_set>
 #include <vector>
 
-namespace c10 {
-namespace cuda {
-namespace CUDACachingAllocator {
-namespace CudaMallocAsync {
+namespace c10::cuda::CUDACachingAllocator::CudaMallocAsync {
 
 #if CUDA_VERSION >= 11040
 // CUDA device allocator that uses cudaMallocAsync to implement
@@ -898,7 +895,4 @@ CUDAAllocator* allocator() {
 
 #endif
 
-} // namespace CudaMallocAsync
-} // namespace CUDACachingAllocator
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::CUDACachingAllocator::CudaMallocAsync
diff --git a/c10/cuda/CUDAMiscFunctions.cpp b/c10/cuda/CUDAMiscFunctions.cpp
index 37225d2a9a0b..212139ddfacd 100644
--- a/c10/cuda/CUDAMiscFunctions.cpp
+++ b/c10/cuda/CUDAMiscFunctions.cpp
@@ -1,8 +1,7 @@
 #include <c10/cuda/CUDAMiscFunctions.h>
 #include <stdlib.h>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 const char* get_cuda_check_suffix() noexcept {
   static char* device_blocking_flag = getenv("CUDA_LAUNCH_BLOCKING");
@@ -21,5 +20,4 @@ std::mutex* getFreeMutex() {
   return &cuda_free_mutex;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
diff --git a/c10/cuda/CUDAStream.cpp b/c10/cuda/CUDAStream.cpp
index 08916cfee4a6..88150334d678 100644
--- a/c10/cuda/CUDAStream.cpp
+++ b/c10/cuda/CUDAStream.cpp
@@ -11,8 +11,7 @@
 #include <mutex>
 #include <vector>
 
-namespace c10 {
-namespace cuda {
+namespace c10::cuda {
 
 namespace {
 
@@ -339,5 +338,4 @@ std::ostream& operator<<(std::ostream& stream, const CUDAStream& s) {
   return stream << s.unwrap();
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
diff --git a/c10/cuda/driver_api.cpp b/c10/cuda/driver_api.cpp
index 56243e663c79..5fc4364e9f02 100644
--- a/c10/cuda/driver_api.cpp
+++ b/c10/cuda/driver_api.cpp
@@ -4,8 +4,8 @@
 #include <c10/util/Exception.h>
 #include <dlfcn.h>
 #include <iostream>
-namespace c10 {
-namespace cuda {
+
+namespace c10::cuda {
 
 namespace {
 
@@ -42,7 +42,6 @@ C10_EXPORT DriverAPI* DriverAPI::get() {
   return &singleton;
 }
 
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda
 
 #endif
diff --git a/c10/cuda/impl/CUDAGuardImpl.cpp b/c10/cuda/impl/CUDAGuardImpl.cpp
index 584430f7a220..35161209a15f 100644
--- a/c10/cuda/impl/CUDAGuardImpl.cpp
+++ b/c10/cuda/impl/CUDAGuardImpl.cpp
@@ -1,11 +1,7 @@
 #include <c10/cuda/impl/CUDAGuardImpl.h>
 
-namespace c10 {
-namespace cuda {
-namespace impl {
+namespace c10::cuda::impl {
 
 C10_REGISTER_GUARD_IMPL(CUDA, CUDAGuardImpl);
 
-} // namespace impl
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::impl
diff --git a/c10/cuda/impl/CUDAGuardImpl.h b/c10/cuda/impl/CUDAGuardImpl.h
index f98f6a98a3fa..2be8cb1b1921 100644
--- a/c10/cuda/impl/CUDAGuardImpl.h
+++ b/c10/cuda/impl/CUDAGuardImpl.h
@@ -18,9 +18,7 @@
 #include <cuda_runtime_api.h>
 #include <cstdint>
 
-namespace c10 {
-namespace cuda {
-namespace impl {
+namespace c10::cuda::impl {
 
 struct CUDAGuardImpl final : public c10::impl::DeviceGuardImplInterface {
   static constexpr DeviceType static_type = DeviceType::CUDA;
@@ -211,6 +209,4 @@ struct CUDAGuardImpl final : public c10::impl::DeviceGuardImplInterface {
   }
 };
 
-} // namespace impl
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::impl
diff --git a/c10/cuda/impl/CUDATest.cpp b/c10/cuda/impl/CUDATest.cpp
index f2d4ae976512..287479dab945 100644
--- a/c10/cuda/impl/CUDATest.cpp
+++ b/c10/cuda/impl/CUDATest.cpp
@@ -5,9 +5,7 @@
 
 #include <cuda_runtime.h>
 
-namespace c10 {
-namespace cuda {
-namespace impl {
+namespace c10::cuda::impl {
 
 bool has_cuda_gpu() {
   int count = 0;
@@ -29,6 +27,4 @@ int c10_cuda_private_test() {
   return 2;
 }
 
-} // namespace impl
-} // namespace cuda
-} // namespace c10
+} // namespace c10::cuda::impl
diff --git a/c10/cuda/impl/CUDATest.h b/c10/cuda/impl/CUDATest.h
index 593905d15672..c4316e8a6b73 100644
--- a/c10/cuda/impl/CUDATest.h
+++ b/c10/cuda/impl/CUDATest.h
@@ -2,12 +2,8 @@
 
 #include <c10/cuda/CUDAMacros.h>
 
-namespace c10 {
-namespace cuda {
-namespace impl {
+namespace c10::cuda::impl {
 
 C10_CUDA_API int c10_cuda_test();
 
 }
-} // namespace cuda
-} // namespace c10
diff --git a/c10/util/DeadlockDetection.cpp b/c10/util/DeadlockDetection.cpp
index bb95939fc53f..320fa7873c6f 100644
--- a/c10/util/DeadlockDetection.cpp
+++ b/c10/util/DeadlockDetection.cpp
@@ -2,8 +2,7 @@
 
 #include <cstdlib>
 
-namespace c10 {
-namespace impl {
+namespace c10::impl {
 
 namespace {
 PythonGILHooks* python_gil_hooks = nullptr;
@@ -28,5 +27,4 @@ void SetPythonGILHooks(PythonGILHooks* hooks) {
   python_gil_hooks = hooks;
 }
 
-} // namespace impl
-} // namespace c10
+} // namespace c10::impl
diff --git a/c10/util/Logging.cpp b/c10/util/Logging.cpp
index 862bc8d19a0d..f4eef35b147e 100644
--- a/c10/util/Logging.cpp
+++ b/c10/util/Logging.cpp
@@ -420,8 +420,7 @@ MessageLogger::~MessageLogger() {
 
 #endif // !C10_USE_GLOG
 
-namespace c10 {
-namespace detail {
+namespace c10::detail {
 namespace {
 
 void setLogLevelFlagFromEnv() {
@@ -467,5 +466,4 @@ void setLogLevelFlagFromEnv() {
 }
 
 } // namespace
-} // namespace detail
-} // namespace c10
+} // namespace c10::detail
diff --git a/c10/util/UniqueVoidPtr.cpp b/c10/util/UniqueVoidPtr.cpp
index d4e9e32810ff..3554bcf8ee62 100644
--- a/c10/util/UniqueVoidPtr.cpp
+++ b/c10/util/UniqueVoidPtr.cpp
@@ -1,9 +1,7 @@
 #include <c10/util/UniqueVoidPtr.h>
 
-namespace c10 {
-namespace detail {
+namespace c10::detail {
 
 void deleteNothing(void*) {}
 
-} // namespace detail
-} // namespace c10
+} // namespace c10::detail
diff --git a/torch/csrc/itt.cpp b/torch/csrc/itt.cpp
index 0e0ad78b4bac..868235d90ad7 100644
--- a/torch/csrc/itt.cpp
+++ b/torch/csrc/itt.cpp
@@ -1,8 +1,7 @@
 #include <torch/csrc/itt_wrapper.h>
 #include <torch/csrc/utils/pybind.h>
 
-namespace torch {
-namespace profiler {
+namespace torch::profiler {
 void initIttBindings(PyObject* module) {
   auto m = py::handle(module).cast<py::module>();
 
@@ -12,5 +11,4 @@ void initIttBindings(PyObject* module) {
   itt.def("rangePop", itt_range_pop);
   itt.def("mark", itt_mark);
 }
-} // namespace profiler
-} // namespace torch
+} // namespace torch::profiler
diff --git a/torch/csrc/itt_wrapper.cpp b/torch/csrc/itt_wrapper.cpp
index 432ef4313559..2719e11e1952 100644
--- a/torch/csrc/itt_wrapper.cpp
+++ b/torch/csrc/itt_wrapper.cpp
@@ -2,8 +2,7 @@
 #include <torch/csrc/itt_wrapper.h>
 #include <torch/csrc/profiler/stubs/base.h>
 
-namespace torch {
-namespace profiler {
+namespace torch::profiler {
 __itt_domain* _itt_domain = __itt_domain_create("PyTorch");
 
 bool itt_is_available() {
@@ -24,5 +23,4 @@ void itt_mark(const char* msg) {
   __itt_task_begin(_itt_domain, __itt_null, __itt_null, hsMsg);
   __itt_task_end(_itt_domain);
 }
-} // namespace profiler
-} // namespace torch
+} // namespace torch::profiler
diff --git a/torch/csrc/itt_wrapper.h b/torch/csrc/itt_wrapper.h
index a24fcb1a6991..502af374ff3d 100644
--- a/torch/csrc/itt_wrapper.h
+++ b/torch/csrc/itt_wrapper.h
@@ -2,13 +2,11 @@
 #define PROFILER_ITT_H
 #include <c10/macros/Export.h>
 
-namespace torch {
-namespace profiler {
+namespace torch::profiler {
 TORCH_API bool itt_is_available();
 TORCH_API void itt_range_push(const char* msg);
 TORCH_API void itt_range_pop();
 TORCH_API void itt_mark(const char* msg);
-} // namespace profiler
-} // namespace torch
+} // namespace torch::profiler
 
 #endif // PROFILER_ITT_H