Fix usage of forwarding references (#161094)

I found a number of places that seem to want forwarding
references but the type signature does not reflect that

Pull Request resolved: https://github.com/pytorch/pytorch/pull/161094
Approved by: https://github.com/malfet

aten/src/ATen/native/quantized/cpu/qlinear_prepack.cpp

@@ -374,7 +374,7 @@ static inline std::vector<c10::IValue> makeStack(Inputs&&... inputs) {
 template <class... Args>
 static inline std::vector<c10::IValue> callOpByHandle(
     const c10::OperatorHandle& op,
-    Args... args) {
+    Args&&... args) {
   auto stack = makeStack(std::forward<Args>(args)...);
   c10::Dispatcher::singleton().callBoxed(op, &stack);
   return stack;
@@ -384,7 +384,7 @@ template <class... Args>
 static inline std::vector<c10::IValue> callOpByName(
     const char* func_name,
     const char* overload_name,
-    Args... args) {
+    Args&&... args) {
   const std::optional<c10::OperatorHandle> op_handle =
       c10::Dispatcher::singleton().findSchema({func_name, overload_name});
   assert(op_handle.has_value());

functorch/csrc/dim/minpybind.h

@@ -232,7 +232,7 @@ struct base {
         return obj<T>::steal(self);
     }
     template<typename ... Args>
-    static obj<T> create(Args ... args) {
+    static obj<T> create(Args&&... args) {
         auto self = alloc();
         self->init(std::forward<Args>(args)...);
         return self;

torch/csrc/autograd/profiler_python.cpp

@@ -657,7 +657,7 @@ struct ThreadLocalResults {
   }
 
   template <CallType C, EventType E, typename Ephemeral, typename... Args>
-  TraceKey intern(Ephemeral ephemeral, Args... args) {
+  TraceKey intern(Ephemeral ephemeral, Args&&... args) {
     static_assert(
         Config<C>::event_type == E,
         "ThreadLocalResults.intern called from the wrong typed context.");

torch/csrc/jit/tensorexpr/eval.h

@@ -267,7 +267,7 @@ class ExprEval {
   }
 
   template <typename T, typename... Ts>
-  T value(Ts... ts) {
+  T value(Ts&&... ts) {
     call(std::forward<Ts>(ts)...);
     return ret_value_.as<T>();
   }

torch/nativert/common/FileUtil.cpp

@@ -81,7 +81,7 @@ int filterCloseReturn(int r) {
 
 // Wrap call to f(args) in loop to retry on EINTR
 template <class F, class... Args>
-ssize_t wrapNoInt(F f, Args... args) {
+ssize_t wrapNoInt(F f, Args&&... args) {
   ssize_t r = -1;
   do {
     r = f(std::forward<Args>(args)...);