Adding preprocessor checks for torch version during torch cpp extensi…

…ons compilation (#8989)

orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/aten_op_executor/aten_op_executor.cc

@@ -180,7 +180,24 @@ std::vector<DLManagedTensor*> ExecuteATenOperator(const char* op_name, const cha
     torch::jit::push(stack, arguments[i]);
   }
 
+#ifndef TORCH_VERSION_PREEQ
+#define TORCH_VERSION_PREEQ(x, y) \
+  ((TORCH_VERSION_MAJOR == (x) && TORCH_VERSION_MINOR >= (y)) || \
+     (TORCH_VERSION_MAJOR > (x)))
+#endif
+
+// pull request https://github.com/pytorch/pytorch/pull/63414 introduced
+// a backwards incompatibility by changing the API. To make ORTModule
+// work with both torch versions >=1.10 as well as < 1.10, we need
+// preprocessor checks
+#if TORCH_VERSION_PREEQ(1, 10)
+  // torch version is >= 1.10
+  aten_op.op->getOperation()(stack);
+#else
+  // torch version is < 1.10
   aten_op.op->getOperation()(&stack);
+#endif
+
   std::vector<DLManagedTensor*> result;
   for (const auto& ret : torch::jit::pop(stack, aten_op.return_size)) {
     const auto& tensor = ret.toTensor();