fix the bug of input format inconsistency

mindspore/ccsrc/runtime/graph_scheduler/actor/kernel_actor.cc

@@ -48,7 +48,8 @@ void KernelActor::Init() {
     const auto &input_device_tensor = AnfAlgo::GetPrevNodeMutableOutputAddr(kernel_, i, false);
     MS_EXCEPTION_IF_NULL(input_device_tensor);
     (void)real_input_data_infos_.emplace_back(
-      std::make_pair(input_device_tensor->format(), input_device_tensor->host_shape()));
+      std::make_shared<InputDataInfo>(input_device_tensor->format(), input_device_tensor->host_shape(),
+                                      input_device_tensor->GetSize(), input_device_tensor->type_id()));
   }
 
   // Init the device tensors and kernel launch info.
@@ -311,8 +312,9 @@ void KernelActor::CopyInputDeviceTensor(const OpData<DeviceTensor> *input_data,
     SET_OPCONTEXT_FAIL_RET_WITH_ERROR_BY_STRATEGY(strategy_, *context, "The input index is of range.");
   }
   auto &real_input_info = real_input_data_infos_[input_data->index_];
+  MS_EXCEPTION_IF_NULL(real_input_info);
   if ((input_data->data_->DeviceType() == device_contexts_[0]->GetDeviceAddressType()) &&
-      (input_data->data_->format() == real_input_info.first)) {
+      (input_data->data_->format() == real_input_info->format_)) {
     return;
   }
 
@@ -320,14 +322,15 @@ void KernelActor::CopyInputDeviceTensor(const OpData<DeviceTensor> *input_data,
     SET_OPCONTEXT_FAIL_RET_WITH_ERROR_BY_STRATEGY(strategy_, *context, "The input index is of range.");
   }
   if (copy_input_device_tensors_[input_data->index_] == nullptr) {
-    copy_input_device_tensors_[input_data->index_] =
-      device_contexts_[0]->CreateDeviceAddress(nullptr, input_data->data_->GetSize(), real_input_info.first,
-                                               input_data->data_->type_id(), real_input_info.second);
+    copy_input_device_tensors_[input_data->index_] = device_contexts_[0]->CreateDeviceAddress(
+      nullptr, real_input_info->size_, real_input_info->format_, real_input_info->type_id_, real_input_info->shape_);
   }
   auto &new_device_tensor = copy_input_device_tensors_[input_data->index_];
   MS_EXCEPTION_IF_NULL(new_device_tensor);
   // Dynamic shape need update size.
-  new_device_tensor->SetSize(input_data->data_->GetSize());
+  if (AnfUtils::IsShapeDynamic(real_input_info->shape_)) {
+    new_device_tensor->SetSize(input_data->data_->GetSize());
+  }
   // Update the input device tensor.
   input_device_tensors_[input_data->index_] = new_device_tensor.get();
 

mindspore/ccsrc/runtime/graph_scheduler/actor/kernel_actor.h

@@ -39,6 +39,15 @@ using mindspore::kernel::Address;
 using mindspore::kernel::KernelLaunchInfo;
 using mindspore::tensor::TensorPtr;
 
+struct InputDataInfo {
+  InputDataInfo(const std::string &format, const ShapeVector &shape, size_t size, TypeId type_id)
+      : format_(format), shape_(shape), size_(size), type_id_(type_id) {}
+  std::string format_;
+  ShapeVector shape_;
+  size_t size_;
+  TypeId type_id_;
+};
+
 // The kernel actor is used to receive the device tensors and control info to luanch kernel.
 // The processing flow is RunOpData/RunOpControl -> CheckRunningCondition -> SendMemoryAllocReq
 // -> OnMemoryAllocFinish -> LaunchKernel -> SendMemoryFreeReq -> SendOutput.
@@ -129,8 +138,8 @@ class KernelActor : public DebugAwareActor {
   // The received input device type and format may be different from the formal parameter in the control flow scenarios,
   // so it needs to be copied from the input data to real data that kernel launch needs.
   std::vector<DeviceTensorPtr> copy_input_device_tensors_;
-  // Real data info <format, host_shape> that kernel launch needs, used to check the consistency of received input data.
-  std::vector<std::pair<std::string, ShapeVector>> real_input_data_infos_;
+  // Real data info that kernel launch needs, used to check the consistency of received input data.
+  std::vector<std::shared_ptr<InputDataInfo>> real_input_data_infos_;
 
   // The device tensors for memory alloc and free.
   // output + workspace

mindspore/core/utils/anf_utils.cc

@@ -142,6 +142,10 @@ bool AnfUtils::IsShapeDynamic(const std::vector<size_t> &shape) {
   return std::any_of(shape.begin(), shape.end(), [](int64_t s) { return s < 0; });
 }
 
+bool AnfUtils::IsShapeDynamic(const std::vector<int64_t> &shape) {
+  return std::any_of(shape.begin(), shape.end(), [](int64_t s) { return s < 0; });
+}
+
 bool AnfUtils::IsNodeOutputDynamicShape(const CNodePtr &node) {
   MS_EXCEPTION_IF_NULL(node);
   auto base_shape = node->Shape();

mindspore/core/utils/anf_utils.h

@@ -65,6 +65,7 @@ class MS_CORE_API AnfUtils {
   static bool IsDimUnknown(const abstract::ShapePtr &shape);
   static bool IsShapeDynamic(const abstract::ShapePtr &shape);
   static bool IsShapeDynamic(const std::vector<size_t> &shape);
+  static bool IsShapeDynamic(const std::vector<int64_t> &shape);
   static bool IsNodeOutputDynamicShape(const CNodePtr &node);
   static bool IsDimUnknown(const AnfNodePtr &node);
   // check whether the anf node is a real kernel that can run on device,parameter and constant is real kernel too