fix resnext_50 segment bug

mindspore/lite/src/common/graph_util.cc

@@ -86,15 +86,8 @@ std::vector<size_t> GetLinkedPostNodeIdx(const lite::Model *model, const size_t
   return post_node_idxes;
 }
 
+// only support op_type from current schema
 bool IsPackedOp(int op_type) {
-#ifdef ENABLE_V0
-  static std::vector<int> v0_packed_ops = {
-    schema::v0::PrimitiveType_Conv2D, schema::v0::PrimitiveType_DeConv2D, schema::v0::PrimitiveType_DepthwiseConv2D,
-    schema::v0::PrimitiveType_DeDepthwiseConv2D, schema::v0::PrimitiveType_MatMul};
-  if (VersionManager::GetInstance()->CheckV0Schema()) {
-    return IsContain(v0_packed_ops, op_type);
-  }
-#endif
   static std::vector<int> packed_ops = {schema::PrimitiveType_Conv2DFusion, schema::PrimitiveType_Conv2dTransposeFusion,
                                         schema::PrimitiveType_MatMul};
   return IsContain(packed_ops, op_type);

mindspore/lite/src/common/graph_util.h

@@ -31,13 +31,14 @@ namespace mindspore {
 namespace lite {
 using NODE_ID = std::string;
 
+// only support op_type from current schema
+bool IsPackedOp(int op_type);
+
 std::vector<size_t> GetGraphInputNodes(const lite::Model *model);
 
 std::vector<size_t> GetGraphOutputNodes(const lite::Model *model);
 
 std::vector<size_t> GetLinkedPostNodeIdx(const lite::Model *model, size_t tensor_idx);
-
-bool IsPackedOp(int op_type);
 }  // namespace lite
 }  // namespace mindspore
 

mindspore/lite/src/scheduler.cc

@@ -281,24 +281,26 @@ int CopyConstTensorData(const std::vector<Tensor *> &tensors, int op_type) {
     return RET_OK;
   }
   for (auto *tensor : tensors) {
-    // only cast const tensor
-    // tensorlist not support fp16 now
-    if (!tensor->IsConst() || tensor->data_type() == kObjectTypeTensorType) {
+    // only copy non-copied const tensor
+    if (!tensor->IsConst() || tensor->own_data()) {
       continue;
     }
-    if (tensor->own_data()) {
-      continue;
+    if (tensor->data_type() == kObjectTypeTensorType) {
+      // tensorlist's data is nullptr since ConvertTensors
+      // we never set or malloc data of tensorlist but malloc tensors in tensorlist
+      MS_ASSERT(tensor->data_c() == nullptr);
+    } else {
+      auto copy_tensor = Tensor::CopyTensor(*tensor, true);
+      if (copy_tensor == nullptr) {
+        MS_LOG(ERROR) << "Copy tensor failed";
+        return RET_ERROR;
+      }
+      tensor->FreeData();
+      tensor->set_data(copy_tensor->data_c());
+      tensor->set_own_data(true);
+      copy_tensor->set_data(nullptr);
+      delete (copy_tensor);
     }
-    auto copy_tensor = Tensor::CopyTensor(*tensor, true);
-    if (copy_tensor == nullptr) {
-      MS_LOG(ERROR) << "Copy tensor failed";
-      return RET_ERROR;
-    }
-    tensor->FreeData();
-    tensor->set_data(copy_tensor->data_c());
-    tensor->set_own_data(true);
-    copy_tensor->set_data(nullptr);
-    delete (copy_tensor);
   }
   return RET_OK;
 }

mindspore/lite/test/models_compatibility.cfg

@@ -1,4 +1,3 @@
 inception_v4_ms_r1.0
 mobilenet_v2_1.0_224_ms_r1.0
-# Add input shapes to fix Segmentation fault at Samsung phone
-resnext50_ms_r1.0;1,224,224,3
+resnext50_ms_r1.0