diff --git a/.clang-format b/.clang-format
index 31be219b186..0b94540e7a2 100644
--- a/.clang-format
+++ b/.clang-format
@@ -106,10 +106,11 @@ StatementMacros:
   - C10_DEFINE_int32
   - C10_DEFINE_int64
   - C10_DEFINE_string
+  - DEFINE_BINARY
   - PyObject_HEAD
   - PyObject_VAR_HEAD
   - PyException_HEAD
-  - DEFINE_BINARY
+  - TORCH_DECLARE_bool
 
 TabWidth:        8
 UseTab:          Never
diff --git a/aten/src/ATen/native/DispatchStub.h b/aten/src/ATen/native/DispatchStub.h
index 54835cefbaf..fc8a5f1962d 100644
--- a/aten/src/ATen/native/DispatchStub.h
+++ b/aten/src/ATen/native/DispatchStub.h
@@ -20,7 +20,7 @@
 //
 // In native/MyKernel.h:
 //   using fn_type = void(*)(const Tensor& x);
-//   DECLARE_DISPATCH(fn_type, stub);
+//   DECLARE_DISPATCH(fn_type, stub)
 //
 // In native/MyKernel.cpp
 //   DEFINE_DISPATCH(stub);
diff --git a/aten/src/ATen/native/cpu/ReducedPrecisionFloatGemvFastPathKernel.cpp b/aten/src/ATen/native/cpu/ReducedPrecisionFloatGemvFastPathKernel.cpp
index e266dcf202a..6eced4b7a4f 100644
--- a/aten/src/ATen/native/cpu/ReducedPrecisionFloatGemvFastPathKernel.cpp
+++ b/aten/src/ATen/native/cpu/ReducedPrecisionFloatGemvFastPathKernel.cpp
@@ -476,8 +476,8 @@ void bf16_gemv_trans(
 #if !defined(C10_MOBILE)
 REGISTER_DISPATCH(fp16_dot_with_fp32_arith_stub, &fp16_dot_with_fp32_arith)
 REGISTER_DISPATCH(fp16_gemv_trans_stub, &fp16_gemv_trans)
-REGISTER_DISPATCH(bf16_dot_with_fp32_arith_stub, &bf16_dot_with_fp32_arith);
-REGISTER_DISPATCH(bf16_gemv_trans_stub, &bf16_gemv_trans);
+REGISTER_DISPATCH(bf16_dot_with_fp32_arith_stub, &bf16_dot_with_fp32_arith)
+REGISTER_DISPATCH(bf16_gemv_trans_stub, &bf16_gemv_trans)
 #endif //!defined(C10_MOBILE)
 
 } // namespace at::native
diff --git a/cmake/public/utils.cmake b/cmake/public/utils.cmake
index c796fab1e9a..a1b2f0b5ec3 100644
--- a/cmake/public/utils.cmake
+++ b/cmake/public/utils.cmake
@@ -387,9 +387,8 @@ function(torch_compile_options libname)
       list(APPEND private_compile_options -Wunused-but-set-variable)
     endif()
     if("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang")
-      list(APPEND private_compile_options -Wunused-private-field)
-    endif()
-    if(NOT "${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang")
+      list(APPEND private_compile_options -Wunused-private-field -Wextra-semi -Wno-error=extra-semi)
+    else()
       list(APPEND private_compile_options
         # Considered to be flaky.  See the discussion at
         # https://github.com/pytorch/pytorch/pull/9608
diff --git a/torch/csrc/DataLoader.cpp b/torch/csrc/DataLoader.cpp
index 50054f529a2..7303ef5f680 100644
--- a/torch/csrc/DataLoader.cpp
+++ b/torch/csrc/DataLoader.cpp
@@ -70,15 +70,15 @@ SIGNAL_HANDLER(
     SIGBUS,
     handler_SIGBUS,
     "ERROR: Unexpected bus error encountered in worker. "
-    "This might be caused by insufficient shared memory (shm).\n");
+    "This might be caused by insufficient shared memory (shm).\n")
 SIGNAL_HANDLER(
     SIGSEGV,
     handler_SIGSEGV,
-    "ERROR: Unexpected segmentation fault encountered in worker.\n");
+    "ERROR: Unexpected segmentation fault encountered in worker.\n")
 SIGNAL_HANDLER(
     SIGFPE,
     handler_SIGFPE,
-    "ERROR: Unexpected floating-point exception encountered in worker.\n");
+    "ERROR: Unexpected floating-point exception encountered in worker.\n")
 
 // When an error happened in DataLoader methods and Python starts to exit, the
 // error trace will keep the loader alive, and Python may kill the children
diff --git a/torch/csrc/Exceptions.h b/torch/csrc/Exceptions.h
index 8c1f37a10c2..7a927c3f03f 100644
--- a/torch/csrc/Exceptions.h
+++ b/torch/csrc/Exceptions.h
@@ -339,7 +339,7 @@ struct noop_gil_scoped_release {
   // user-defined constructor (i.e. not defaulted) to avoid
   // unused-variable warnings at usage sites of this class
   // NOLINTNEXTLINE(modernize-use-equals-default)
-  noop_gil_scoped_release(){};
+  noop_gil_scoped_release() {}
 };
 
 template <bool release_gil>
diff --git a/torch/csrc/PyInterpreter.cpp b/torch/csrc/PyInterpreter.cpp
index 57410c5cadf..6d285759e28 100644
--- a/torch/csrc/PyInterpreter.cpp
+++ b/torch/csrc/PyInterpreter.cpp
@@ -273,14 +273,14 @@ void ConcretePyInterpreterVTable::decref(PyObject* pyobj, bool has_pyobj_slot)
     }
   }
   Py_DECREF(pyobj);
-};
+}
 
 void ConcretePyInterpreterVTable::incref(PyObject* pyobj) const {
   if (!Py_IsInitialized())
     return;
   pybind11::gil_scoped_acquire gil;
   Py_INCREF(pyobj);
-};
+}
 
 bool isPythonTensor(const at::Tensor& tensor) {
   return tensor.unsafeGetTensorImpl()->key_set().has(c10::DispatchKey::Python);
diff --git a/torch/csrc/Storage.h b/torch/csrc/Storage.h
index 55deb18892b..fc63d14ab93 100644
--- a/torch/csrc/Storage.h
+++ b/torch/csrc/Storage.h
@@ -10,7 +10,7 @@
 #define THPStorageStr "torch.UntypedStorage"
 
 struct THPStorage {
-  PyObject_HEAD;
+  PyObject_HEAD
   c10::MaybeOwned<c10::Storage> cdata;
   bool is_hermetic;
 };
diff --git a/torch/csrc/autograd/FunctionsManual.cpp b/torch/csrc/autograd/FunctionsManual.cpp
index 2af649f468c..86771b9b30a 100644
--- a/torch/csrc/autograd/FunctionsManual.cpp
+++ b/torch/csrc/autograd/FunctionsManual.cpp
@@ -5249,7 +5249,7 @@ static Tensor apply_simple_transformation(
       return condition_with_I ? K - transformation : -transformation;
     }
   }
-};
+}
 
 std::tuple<Tensor, Tensor> householder_product_backward(
     const Tensor& grad,
diff --git a/torch/csrc/autograd/profiler_python.cpp b/torch/csrc/autograd/profiler_python.cpp
index d3f3b1bd214..4b9232eff76 100644
--- a/torch/csrc/autograd/profiler_python.cpp
+++ b/torch/csrc/autograd/profiler_python.cpp
@@ -77,7 +77,7 @@ PyCodeObject* getCode<CallType::PyModuleCall>() {
     return (PyCodeObject*)res;
   }();
   return module_call_code;
-};
+}
 
 template <>
 PyCodeObject* getCode<CallType::PyOptimizerCall>() {
@@ -92,7 +92,7 @@ PyCodeObject* getCode<CallType::PyOptimizerCall>() {
     return (PyCodeObject*)res;
   }();
   return optimizer_step_code;
-};
+}
 
 } // namespace
 } // namespace torch::profiler::impl
@@ -548,7 +548,7 @@ struct TraceKeyCacheState {
 // `PyEval_SetProfile`.
 struct ThreadLocalResults;
 struct TraceContext {
-  PyObject_HEAD;
+  PyObject_HEAD
   ThreadLocalResults* thread_local_results_;
 };
 
@@ -795,7 +795,7 @@ PythonTracer::PythonTracer(torch::profiler::impl::RecordQueue* queue)
     //   cannot be round tripped via `sys.settrace(sys.gettrace())`
     PyEval_SetProfile(PythonTracer::pyProfileFn, (PyObject*)ctx);
   }
-};
+}
 
 void PythonTracer::stop() {
   gil_and_restore_thread gil;
diff --git a/torch/csrc/autograd/python_variable.h b/torch/csrc/autograd/python_variable.h
index 32cc5c930ca..82939211eb5 100644
--- a/torch/csrc/autograd/python_variable.h
+++ b/torch/csrc/autograd/python_variable.h
@@ -15,7 +15,7 @@ namespace py = pybind11;
 
 // Python object that backs torch.autograd.Variable
 struct THPVariable {
-  PyObject_HEAD;
+  PyObject_HEAD
   // Payload
   c10::MaybeOwned<at::Tensor> cdata;
   // Hooks to be run on backwards pass (corresponds to Python attr
diff --git a/torch/csrc/distributed/autograd/rpc_messages/cleanup_autograd_context_req.cpp b/torch/csrc/distributed/autograd/rpc_messages/cleanup_autograd_context_req.cpp
index 54db496ea01..7134715d6a5 100644
--- a/torch/csrc/distributed/autograd/rpc_messages/cleanup_autograd_context_req.cpp
+++ b/torch/csrc/distributed/autograd/rpc_messages/cleanup_autograd_context_req.cpp
@@ -5,7 +5,7 @@
 namespace torch::distributed::autograd {
 
 CleanupAutogradContextReq::CleanupAutogradContextReq(int64_t context_id)
-    : context_id_(context_id){};
+    : context_id_(context_id) {}
 
 int64_t CleanupAutogradContextReq::getContextId() {
   return context_id_;
diff --git a/torch/csrc/distributed/c10d/DMAConnectivity.cpp b/torch/csrc/distributed/c10d/DMAConnectivity.cpp
index 50c34f62426..3e5efa19049 100644
--- a/torch/csrc/distributed/c10d/DMAConnectivity.cpp
+++ b/torch/csrc/distributed/c10d/DMAConnectivity.cpp
@@ -65,7 +65,7 @@ class DetectorMap {
       cached_;
 };
 
-}; // namespace
+} // namespace
 
 namespace c10d {
 
diff --git a/torch/csrc/distributed/c10d/init.cpp b/torch/csrc/distributed/c10d/init.cpp
index 58bc7d14bdb..5e7cb1aed7e 100644
--- a/torch/csrc/distributed/c10d/init.cpp
+++ b/torch/csrc/distributed/c10d/init.cpp
@@ -139,7 +139,7 @@ class IntrusivePtrNoGilDestructor {
 
 } // anonymous namespace
 
-PYBIND11_DECLARE_HOLDER_TYPE(T, IntrusivePtrNoGilDestructor<T>, true);
+PYBIND11_DECLARE_HOLDER_TYPE(T, IntrusivePtrNoGilDestructor<T>, true)
 
 namespace torch::distributed::c10d {
 
diff --git a/torch/csrc/distributed/c10d/reducer.cpp b/torch/csrc/distributed/c10d/reducer.cpp
index bf21bab37ce..e31431ef271 100644
--- a/torch/csrc/distributed/c10d/reducer.cpp
+++ b/torch/csrc/distributed/c10d/reducer.cpp
@@ -43,7 +43,7 @@ C10_DEFINE_TYPED_REGISTRY( // NOLINT
     c10::DeviceType,
     Timer,
     std::unique_ptr,
-    c10::Device);
+    c10::Device)
 
 namespace {
 
@@ -67,7 +67,7 @@ class CpuTimer : public Timer {
   }
 };
 
-C10_REGISTER_TYPED_CLASS(TimerRegistry, c10::kCPU, CpuTimer);
+C10_REGISTER_TYPED_CLASS(TimerRegistry, c10::kCPU, CpuTimer)
 
 std::vector<at::Tensor> extractTensors(const c10::IValue& result) {
   if (result.isPyObject()) {
diff --git a/torch/csrc/distributed/rpc/metrics/registry.cpp b/torch/csrc/distributed/rpc/metrics/registry.cpp
index b787390fda5..c70a5f1a711 100644
--- a/torch/csrc/distributed/rpc/metrics/registry.cpp
+++ b/torch/csrc/distributed/rpc/metrics/registry.cpp
@@ -3,5 +3,5 @@
 namespace torch::distributed::rpc {
 C10_DEFINE_REGISTRY(
     RpcMetricsHandlerRegistry,
-    torch::distributed::rpc::RpcMetricsHandler);
+    torch::distributed::rpc::RpcMetricsHandler)
 } // namespace torch::distributed::rpc
diff --git a/torch/csrc/dynamo/guards.cpp b/torch/csrc/dynamo/guards.cpp
index f538085b083..c39ac420ced 100644
--- a/torch/csrc/dynamo/guards.cpp
+++ b/torch/csrc/dynamo/guards.cpp
@@ -229,7 +229,7 @@ namespace {
 typedef std::vector<TensorCheck> ChecksList;
 
 typedef struct {
-  PyObject_HEAD;
+  PyObject_HEAD
   ChecksList* checks;
 } TensorGuards;
 
@@ -510,7 +510,7 @@ static PyTypeObject TensorGuardsType = { PyVarObject_HEAD_INIT(nullptr, 0)
 // merged.
 // NOLINTNEXTLINE(cppcoreguidelines-pro-type-member-init)
 struct GlobalStateGuard {
-  PyObject_HEAD;
+  PyObject_HEAD
 
   inline void init() {
     auto& ctx = at::globalContext();
diff --git a/torch/csrc/dynamo/init.cpp b/torch/csrc/dynamo/init.cpp
index 16a3f1e2c97..ce8f5db1c0a 100644
--- a/torch/csrc/dynamo/init.cpp
+++ b/torch/csrc/dynamo/init.cpp
@@ -15,7 +15,7 @@
 static struct PyModuleDef _module =
     {PyModuleDef_HEAD_INIT, "torch._C._dynamo", "", -1, nullptr};
 
-PYBIND11_MAKE_OPAQUE(std::vector<uint8_t>);
+PYBIND11_MAKE_OPAQUE(std::vector<uint8_t>)
 
 namespace torch::dynamo {
 
diff --git a/torch/csrc/jit/api/function_impl.cpp b/torch/csrc/jit/api/function_impl.cpp
index b6aa3703a52..820ecef66a8 100644
--- a/torch/csrc/jit/api/function_impl.cpp
+++ b/torch/csrc/jit/api/function_impl.cpp
@@ -13,6 +13,7 @@
 #include <torch/csrc/jit/passes/autocast.h>
 #endif
 
+// clang-format off
 C10_DEFINE_bool(
     torch_jit_do_not_store_optimized_graph,
     false,
diff --git a/torch/csrc/jit/passes/dtype_analysis.cpp b/torch/csrc/jit/passes/dtype_analysis.cpp
index 3aee83379bd..9cbe6a93623 100644
--- a/torch/csrc/jit/passes/dtype_analysis.cpp
+++ b/torch/csrc/jit/passes/dtype_analysis.cpp
@@ -61,7 +61,7 @@ std::unique_ptr<Stack> MTensorArgumentCreator(Node* n) {
     }
   }
   return stack;
-};
+}
 
 bool MTensorNodeArgValid(Value* value) {
   auto tensor_type = value->type()->cast<TensorType>();
diff --git a/torch/csrc/jit/passes/onnx/naming.cpp b/torch/csrc/jit/passes/onnx/naming.cpp
index 432c5efd1e5..692d60a2d3d 100644
--- a/torch/csrc/jit/passes/onnx/naming.cpp
+++ b/torch/csrc/jit/passes/onnx/naming.cpp
@@ -79,7 +79,7 @@ namespace {
 
 class NodeNameGenerator {
  public:
-  NodeNameGenerator(std::shared_ptr<Graph> g) : graph_(std::move(g)){};
+  NodeNameGenerator(std::shared_ptr<Graph> g) : graph_(std::move(g)) {}
   virtual ~NodeNameGenerator() = 0;
   void PopulateNodeNames();
 
@@ -105,7 +105,7 @@ NodeNameGenerator::~NodeNameGenerator() = default;
 class ScopedNodeNameGenerator : public NodeNameGenerator {
  public:
   ScopedNodeNameGenerator(std::shared_ptr<Graph> g)
-      : NodeNameGenerator(std::move(g)){};
+      : NodeNameGenerator(std::move(g)) {}
 
  protected:
   void CreateNodeName(Node* n) override;
diff --git a/torch/csrc/jit/passes/symbolic_shape_cache.cpp b/torch/csrc/jit/passes/symbolic_shape_cache.cpp
index 6a265a943d5..0cca03d6f74 100644
--- a/torch/csrc/jit/passes/symbolic_shape_cache.cpp
+++ b/torch/csrc/jit/passes/symbolic_shape_cache.cpp
@@ -205,5 +205,5 @@ bool operator==(
     const CanonicalizedSymbolicShape& a,
     const CanonicalizedSymbolicShape& b) {
   return a.values_ == b.values_;
-};
+}
 } // namespace torch::jit
diff --git a/torch/csrc/jit/passes/symbolic_shape_runtime_fusion.cpp b/torch/csrc/jit/passes/symbolic_shape_runtime_fusion.cpp
index b82e5594511..723f3c9cf75 100644
--- a/torch/csrc/jit/passes/symbolic_shape_runtime_fusion.cpp
+++ b/torch/csrc/jit/passes/symbolic_shape_runtime_fusion.cpp
@@ -154,7 +154,7 @@ static std::vector<StrideInput> summarizeInputStrides(const TensorType& tt) {
         summarizeStrideDim(sizes, strides, dim, stride_inputs, 0));
   }
   return stride_inputs;
-};
+}
 
 // Todo: incorporate in codegen
 static StrideInput summarizeOutputStrides(const TensorType& tt) {
diff --git a/torch/csrc/jit/passes/tensorexpr_fuser.cpp b/torch/csrc/jit/passes/tensorexpr_fuser.cpp
index 9162bef4258..01ca666eb66 100644
--- a/torch/csrc/jit/passes/tensorexpr_fuser.cpp
+++ b/torch/csrc/jit/passes/tensorexpr_fuser.cpp
@@ -26,6 +26,7 @@
 
 #include <utility>
 
+// clang-format off
 C10_DEFINE_bool(
     torch_jit_disable_cat,
     false,
diff --git a/torch/csrc/jit/python/pybind.h b/torch/csrc/jit/python/pybind.h
index eb9b59d08d8..5bab3878f3b 100644
--- a/torch/csrc/jit/python/pybind.h
+++ b/torch/csrc/jit/python/pybind.h
@@ -65,7 +65,7 @@ class unwrapping_shared_ptr {
 
 } // namespace torch::jit
 
-PYBIND11_DECLARE_HOLDER_TYPE(T, torch::jit::unwrapping_shared_ptr<T>, true);
+PYBIND11_DECLARE_HOLDER_TYPE(T, torch::jit::unwrapping_shared_ptr<T>, true)
 
 namespace pybind11::detail {
 
diff --git a/torch/csrc/jit/python/python_dict.h b/torch/csrc/jit/python/python_dict.h
index c8433a7df6c..5e8fdbfe9a0 100644
--- a/torch/csrc/jit/python/python_dict.h
+++ b/torch/csrc/jit/python/python_dict.h
@@ -98,12 +98,12 @@ class ScriptDict final {
   // not exist.
   at::IValue getItem(const at::IValue& key) {
     return dict_.at(key);
-  };
+  }
 
   // Set the value for the given key.
   void setItem(const at::IValue& key, const at::IValue& value) {
     dict_.insert_or_assign(key, value);
-  };
+  }
 
   // Check whether the dictionary contains the given key.
   bool contains(const at::IValue& key) {
diff --git a/torch/csrc/jit/python/python_list.h b/torch/csrc/jit/python/python_list.h
index 783e429946f..83955a9f3d5 100644
--- a/torch/csrc/jit/python/python_list.h
+++ b/torch/csrc/jit/python/python_list.h
@@ -92,7 +92,7 @@ class ScriptList final {
   at::IValue getItem(diff_type idx) {
     idx = wrap_index(idx);
     return list_.get(idx);
-  };
+  }
 
   // Set the value corresponding to the given index.
   void setItem(diff_type idx, const at::IValue& value) {
diff --git a/torch/csrc/jit/python/python_sugared_value.h b/torch/csrc/jit/python/python_sugared_value.h
index 314b00bcf38..15cc2445fd5 100644
--- a/torch/csrc/jit/python/python_sugared_value.h
+++ b/torch/csrc/jit/python/python_sugared_value.h
@@ -127,7 +127,7 @@ struct VISIBILITY_HIDDEN ConstantParameterList : public SugaredValue {
 
 struct VISIBILITY_HIDDEN ModuleDictMethod : public SugaredValue {
   explicit ModuleDictMethod(SugaredValuePtr iterable, std::string name)
-      : iterable_(std::move(iterable)), name_(std::move(name)){};
+      : iterable_(std::move(iterable)), name_(std::move(name)) {}
 
   std::string kind() const override {
     return name_;
@@ -286,7 +286,7 @@ struct VISIBILITY_HIDDEN SugaredDict : public SugaredValue {
 
   SugaredValuePtr iter(const SourceRange& loc, GraphFunction& m) override {
     return keys_;
-  };
+  }
 
   std::shared_ptr<ModuleValue> self_;
   std::shared_ptr<SugaredTupleValue> keys_;
diff --git a/torch/csrc/jit/runtime/argument_spec.cpp b/torch/csrc/jit/runtime/argument_spec.cpp
index f8c99f40290..48e45ab7a3a 100644
--- a/torch/csrc/jit/runtime/argument_spec.cpp
+++ b/torch/csrc/jit/runtime/argument_spec.cpp
@@ -66,7 +66,7 @@ void ArgumentSpecCreator::scan(
   } else {
     instructions_.emplace_back(SKIP);
   }
-};
+}
 
 // this is a coarse-grained guarantee that the slots of a class will not be
 // modified by the function. It works fine for things that used be read-only
diff --git a/torch/csrc/jit/runtime/graph_executor.cpp b/torch/csrc/jit/runtime/graph_executor.cpp
index 7b948911df7..8f89584fb7c 100644
--- a/torch/csrc/jit/runtime/graph_executor.cpp
+++ b/torch/csrc/jit/runtime/graph_executor.cpp
@@ -53,6 +53,7 @@
 #include <utility>
 #include <vector>
 
+// clang-format off
 C10_DEFINE_bool(
     torch_jit_execution_plan_reuse_code_graph,
     false,
diff --git a/torch/csrc/jit/runtime/interpreter.cpp b/torch/csrc/jit/runtime/interpreter.cpp
index b3545a3666e..d42c4c69b6f 100644
--- a/torch/csrc/jit/runtime/interpreter.cpp
+++ b/torch/csrc/jit/runtime/interpreter.cpp
@@ -46,6 +46,7 @@ using torch::distributed::autograd::DistAutogradContainer;
 #include <utility>
 #include <vector>
 
+// clang-format off
 C10_DEFINE_bool(
     torch_jit_enable_rethrow_caught_exception,
     false,
diff --git a/torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp b/torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp
index af9deaf476c..98acf24dd1d 100644
--- a/torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp
+++ b/torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp
@@ -38,6 +38,7 @@
 #include <mutex>
 #include <optional>
 
+// clang-format off
 C10_DEFINE_bool(
     torch_jit_enable_new_executor,
     true,
diff --git a/torch/csrc/jit/runtime/static/impl.cpp b/torch/csrc/jit/runtime/static/impl.cpp
index a503d6112ac..8eef32b9c95 100644
--- a/torch/csrc/jit/runtime/static/impl.cpp
+++ b/torch/csrc/jit/runtime/static/impl.cpp
@@ -47,6 +47,7 @@
 #endif
 
 // used in test only
+// clang-format off
 C10_DEFINE_bool(
     static_runtime_disable_debug_memory_overlap_check,
     false,
diff --git a/torch/csrc/jit/runtime/static/ops.cpp b/torch/csrc/jit/runtime/static/ops.cpp
index a3e01e307f5..de41be15889 100644
--- a/torch/csrc/jit/runtime/static/ops.cpp
+++ b/torch/csrc/jit/runtime/static/ops.cpp
@@ -40,6 +40,7 @@
 
 #include <ATen/CompositeExplicitAutogradFunctions.h>
 
+// clang-format off
 C10_DEFINE_bool(
     static_runtime_enable_fast_math,
     true,
diff --git a/torch/csrc/jit/runtime/static/passes.cpp b/torch/csrc/jit/runtime/static/passes.cpp
index 8214c374ddf..0632970e1ca 100644
--- a/torch/csrc/jit/runtime/static/passes.cpp
+++ b/torch/csrc/jit/runtime/static/passes.cpp
@@ -9,6 +9,7 @@
 #include <torch/csrc/jit/runtime/graph_iterator.h>
 #include <torch/csrc/jit/runtime/static/ops.h>
 
+// clang-format off
 C10_DEFINE_bool(
     enable_clip_ranges_gather_fusions,
     true,
diff --git a/torch/csrc/jit/testing/file_check.cpp b/torch/csrc/jit/testing/file_check.cpp
index ba36ab2bb9c..c17a84ac191 100644
--- a/torch/csrc/jit/testing/file_check.cpp
+++ b/torch/csrc/jit/testing/file_check.cpp
@@ -85,7 +85,7 @@ std::ostream& operator<<(std::ostream& out, const Check& c) {
   }
   out << ": " << c.search_str_;
   return out;
-};
+}
 
 namespace {
 
diff --git a/torch/csrc/lazy/core/config.cpp b/torch/csrc/lazy/core/config.cpp
index ce6ce27ca17..00bf247cabd 100644
--- a/torch/csrc/lazy/core/config.cpp
+++ b/torch/csrc/lazy/core/config.cpp
@@ -1,7 +1,7 @@
 #include <c10/util/env.h>
 #include <torch/csrc/lazy/core/config.h>
 
-C10_DEFINE_bool(torch_lazy_ir_debug, false, "Enable lazy tensor IR debugging");
+C10_DEFINE_bool(torch_lazy_ir_debug, false, "Enable lazy tensor IR debugging")
 
 C10_DEFINE_bool(
     torch_lazy_param_aliasing,
diff --git a/torch/csrc/lazy/core/ir.cpp b/torch/csrc/lazy/core/ir.cpp
index 6bb2c0ba0e9..8dac3e563cb 100644
--- a/torch/csrc/lazy/core/ir.cpp
+++ b/torch/csrc/lazy/core/ir.cpp
@@ -7,6 +7,7 @@
 
 // Enables caching on for dynamic shapes (aka disable hash on shapes)
 // NOLINTNEXTLINE(misc-use-internal-linkage)
+// clang-format off
 C10_DEFINE_bool(
     ltc_enable_dynamic_shapes,
     false,
diff --git a/torch/csrc/lazy/core/shape.cpp b/torch/csrc/lazy/core/shape.cpp
index 359fb7a9665..f59684d4ffc 100644
--- a/torch/csrc/lazy/core/shape.cpp
+++ b/torch/csrc/lazy/core/shape.cpp
@@ -9,7 +9,7 @@
 C10_DEFINE_bool(
     ltc_enable_symbolic_shapes,
     false,
-    "Enables calculation of if dims are symbolic");
+    "Enables calculation of if dims are symbolic")
 
 namespace torch::lazy {
 
diff --git a/torch/csrc/lazy/ts_backend/config.cpp b/torch/csrc/lazy/ts_backend/config.cpp
index ee2ce424262..ec098d4dc6e 100644
--- a/torch/csrc/lazy/ts_backend/config.cpp
+++ b/torch/csrc/lazy/ts_backend/config.cpp
@@ -5,7 +5,7 @@
 C10_DEFINE_bool(
     torch_lazy_ts_tensor_update_sync,
     true,
-    "Use synchronous copy inside _copy_from op");
+    "Use synchronous copy inside _copy_from op")
 
 // TODO(whc) we need to hook up these flags in a more useful way
 // possibly also keep LTC_TS_CUDA env working?
@@ -13,4 +13,4 @@ C10_DEFINE_bool(
 C10_DEFINE_bool(
     torch_lazy_ts_cuda,
     false,
-    "Use cuda device for torchscript backend (instead of CPU)");
+    "Use cuda device for torchscript backend (instead of CPU)")
diff --git a/torch/csrc/profiler/python/init.h b/torch/csrc/profiler/python/init.h
index 28fae14988a..a14cc213e8f 100644
--- a/torch/csrc/profiler/python/init.h
+++ b/torch/csrc/profiler/python/init.h
@@ -12,12 +12,12 @@ using torch::profiler::impl::TensorID;
   template <>                         \
   struct type_caster<T> : public strong_pointer_type_caster<T> {};
 
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::StorageImplData);
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::AllocationID);
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::TensorImplAddress);
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyModuleSelf);
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyModuleCls);
-STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyOptimizerSelf);
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::StorageImplData)
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::AllocationID)
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::TensorImplAddress)
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyModuleSelf)
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyModuleCls)
+STRONG_POINTER_TYPE_CASTER(torch::profiler::impl::PyOptimizerSelf)
 #undef STRONG_POINTER_TYPE_CASTER
 
 template <>
diff --git a/torch/csrc/utils/invalid_arguments.cpp b/torch/csrc/utils/invalid_arguments.cpp
index d26f8c2ee1d..c2825f7d945 100644
--- a/torch/csrc/utils/invalid_arguments.cpp
+++ b/torch/csrc/utils/invalid_arguments.cpp
@@ -27,7 +27,7 @@ struct Type {
 };
 
 struct SimpleType : public Type {
-  SimpleType(std::string& name) : name(name){};
+  SimpleType(std::string& name) : name(name) {}
 
   bool is_matching(PyObject* object) override {
     return py_typename(object) == name;
@@ -38,7 +38,7 @@ struct SimpleType : public Type {
 
 struct MultiType : public Type {
   MultiType(std::initializer_list<std::string> accepted_types)
-      : types(accepted_types){};
+      : types(accepted_types) {}
 
   bool is_matching(PyObject* object) override {
     auto it = std::find(types.begin(), types.end(), py_typename(object));
@@ -49,7 +49,7 @@ struct MultiType : public Type {
 };
 
 struct NullableType : public Type {
-  NullableType(std::unique_ptr<Type> type) : type(std::move(type)){};
+  NullableType(std::unique_ptr<Type> type) : type(std::move(type)) {}
 
   bool is_matching(PyObject* object) override {
     return object == Py_None || type->is_matching(object);
@@ -60,7 +60,7 @@ struct NullableType : public Type {
 
 struct TupleType : public Type {
   TupleType(std::vector<std::unique_ptr<Type>> types)
-      : types(std::move(types)){};
+      : types(std::move(types)) {}
 
   bool is_matching(PyObject* object) override {
     if (!PyTuple_Check(object))
@@ -79,7 +79,7 @@ struct TupleType : public Type {
 };
 
 struct SequenceType : public Type {
-  SequenceType(std::unique_ptr<Type> type) : type(std::move(type)){};
+  SequenceType(std::unique_ptr<Type> type) : type(std::move(type)) {}
 
   bool is_matching(PyObject* object) override {
     if (!PySequence_Check(object))
@@ -99,7 +99,7 @@ struct SequenceType : public Type {
 
 struct Argument {
   Argument(std::string name, std::unique_ptr<Type> type)
-      : name(std::move(name)), type(std::move(type)){};
+      : name(std::move(name)), type(std::move(type)) {}
 
   std::string name;
   std::unique_ptr<Type> type;
@@ -109,9 +109,9 @@ struct Option {
   Option(std::vector<Argument> arguments, bool is_variadic, bool has_out)
       : arguments(std::move(arguments)),
         is_variadic(is_variadic),
-        has_out(has_out){};
+        has_out(has_out) {}
   Option(bool is_variadic, bool has_out)
-      : arguments(), is_variadic(is_variadic), has_out(has_out){};
+      : arguments(), is_variadic(is_variadic), has_out(has_out) {}
   Option(const Option&) = delete;
   Option(Option&& other) noexcept = default;
   Option& operator=(const Option&) = delete;
diff --git a/torch/csrc/utils/object_ptr.h b/torch/csrc/utils/object_ptr.h
index 70a887860ff..983a7a2ae07 100644
--- a/torch/csrc/utils/object_ptr.h
+++ b/torch/csrc/utils/object_ptr.h
@@ -7,15 +7,15 @@
 template <class T>
 class TORCH_PYTHON_API THPPointer {
  public:
-  THPPointer() : ptr(nullptr){};
-  explicit THPPointer(T* ptr) noexcept : ptr(ptr){};
+  THPPointer() : ptr(nullptr) {}
+  explicit THPPointer(T* ptr) noexcept : ptr(ptr) {}
   THPPointer(THPPointer&& p) noexcept : ptr(std::exchange(p.ptr, nullptr)) {}
   THPPointer(const THPPointer& p) = delete;
   THPPointer& operator=(const THPPointer&) = delete;
 
   ~THPPointer() {
     free();
-  };
+  }
   T* get() {
     return ptr;
   }
diff --git a/torch/csrc/utils/pybind.h b/torch/csrc/utils/pybind.h
index 3c0f32864a0..a22a08cc222 100644
--- a/torch/csrc/utils/pybind.h
+++ b/torch/csrc/utils/pybind.h
@@ -24,10 +24,10 @@ namespace py = pybind11;
 // This makes intrusive_ptr to be available as a custom pybind11 holder type,
 // see
 // https://pybind11.readthedocs.io/en/stable/advanced/smart_ptrs.html#custom-smart-pointers
-PYBIND11_DECLARE_HOLDER_TYPE(T, c10::intrusive_ptr<T>, true);
+PYBIND11_DECLARE_HOLDER_TYPE(T, c10::intrusive_ptr<T>, true)
 
-PYBIND11_DECLARE_HOLDER_TYPE(T, c10::SingletonOrSharedTypePtr<T>);
-PYBIND11_DECLARE_HOLDER_TYPE(T, c10::SingletonTypePtr<T>, true);
+PYBIND11_DECLARE_HOLDER_TYPE(T, c10::SingletonOrSharedTypePtr<T>)
+PYBIND11_DECLARE_HOLDER_TYPE(T, c10::SingletonTypePtr<T>, true)
 
 namespace pybind11::detail {
 
diff --git a/torch/csrc/utils/python_symnode.h b/torch/csrc/utils/python_symnode.h
index 5a1f43d1bc8..43ef85ad8fc 100644
--- a/torch/csrc/utils/python_symnode.h
+++ b/torch/csrc/utils/python_symnode.h
@@ -35,7 +35,7 @@ class PythonSymNodeImpl : public c10::SymNodeImpl {
   PythonSymNodeImpl(py::object pyobj) : c10::SymNodeImpl() {
     pyobj_ = std::make_shared<c10::SafePyObject>(
         pyobj.release().ptr(), getPyInterpreter());
-  };
+  }
 
   c10::SymNode wrap_int(int64_t num) override {
     py::gil_scoped_acquire acquire;