!6475 [MSLITE] Fix bug of resize parser.

Merge pull request !6475 from wangshaocong/lite_convert

mindspore/lite/src/ops/resize.cc

@@ -97,11 +97,29 @@ int Resize::UnPackToFlatBuilder(const schema::Primitive *primitive, flatbuffers:
 namespace {
 constexpr int kInputRank = 4;
 }  // namespace
+template <typename T>
+void CalShape(const T *data, const std::vector<Tensor *> &inputs, std::vector<int> *out_shape, int shape_size) {
+  int input_count = inputs[0]->ElementsNum();
+  int index = 0;
+  int size = 1;
+  for (int i = 0; i < shape_size; i++) {
+    if (static_cast<int>(data[i]) == -1) {
+      index = i;
+    } else {
+      size *= data[i];
+    }
+    out_shape->push_back(data[i]);
+  }
+  if (static_cast<int>(data[index]) == -1) {
+    (*out_shape)[index] = input_count / size;
+  }
+}
+
 int Resize::InferShape(std::vector<lite::Tensor *> inputs_, std::vector<lite::Tensor *> outputs_) {
   MS_ASSERT(this->primitive_ != nullptr);
   auto input = inputs_.front();
   if (input == nullptr) {
-    return 1;
+    return RET_ERROR;
   }
   if (input->shape().size() != kInputRank) {
     MS_LOG(ERROR) << "Size of input shape is wrong.";
@@ -110,20 +128,58 @@ int Resize::InferShape(std::vector<lite::Tensor *> inputs_, std::vector<lite::Te
 
   auto output = outputs_.front();
   if (output == nullptr) {
-    return 1;
+    return RET_NULL_PTR;
   }
   output->set_data_type(input->data_type());
   output->SetFormat(input->GetFormat());
   if (!GetInferFlag()) {
     return RET_OK;
   }
-  auto new_height = GetNewHeight();
-  auto new_width = GetNewWidth();
 
   std::vector<int> output_shape;
   output_shape.push_back(input->Batch());
-  output_shape.push_back(new_height);
-  output_shape.push_back(new_width);
+  if (inputs_.size() == kDoubleNum) {
+    auto shape_tensor = inputs_.at(1);
+    if (shape_tensor->data_c() == nullptr) {
+      MS_LOG(INFO) << "Do infer shape in runtime.";
+      return RET_INFER_INVALID;
+    }
+    size_t shape_size = shape_tensor->ElementsNum();
+    switch (shape_tensor->data_type()) {
+      case kNumberTypeInt8: {
+        auto data = reinterpret_cast<int8_t *>(shape_tensor->MutableData());
+        CalShape<int8_t>(data, inputs_, &output_shape, shape_size);
+      } break;
+      case kNumberTypeInt32: {
+        auto data = reinterpret_cast<int32_t *>(shape_tensor->MutableData());
+        CalShape<int32_t>(data, inputs_, &output_shape, shape_size);
+      } break;
+      case kNumberTypeInt64: {
+        auto data = reinterpret_cast<int64_t *>(shape_tensor->MutableData());
+        CalShape<int64_t>(data, inputs_, &output_shape, shape_size);
+      } break;
+      case kNumberTypeFloat: {
+        auto data = reinterpret_cast<float *>(shape_tensor->MutableData());
+        CalShape<float>(data, inputs_, &output_shape, shape_size);
+      } break;
+      case kNumberTypeUInt32: {
+        auto data = reinterpret_cast<uint32_t *>(shape_tensor->MutableData());
+        CalShape<uint32_t>(data, inputs_, &output_shape, shape_size);
+      } break;
+      default: {
+        MS_LOG(ERROR) << "Reshape weight tensor has unsupported dataType: " << shape_tensor->data_type();
+        return RET_INFER_ERR;
+      }
+    }
+  } else if (inputs_.size() == kSingleNum) {
+    auto new_height = GetNewHeight();
+    auto new_width = GetNewWidth();
+    output_shape.push_back(new_height);
+    output_shape.push_back(new_width);
+  } else {
+    MS_LOG(ERROR) << "inputs tensor size invalid.";
+    return RET_INFER_ERR;
+  }
   output_shape.push_back(input->Channel());
   output->set_shape(output_shape);
 

mindspore/lite/tools/converter/parser/tflite/tflite_resize_parser.cc

@@ -85,16 +85,22 @@ STATUS TfliteResizeParser::Parse(const std::unique_ptr<tflite::OperatorT> &tflit
     return RET_NULL_PTR;
   }
   auto buffData = reinterpret_cast<int32_t *>(buff->data.data());
-  auto height = buffData[0];
-  auto width = buffData[1];
-  attr->newWidth = width;
-  attr->newHeight = height;
+  if (buffData != nullptr) {
+    auto height = buffData[0];
+    auto width = buffData[1];
+    attr->newWidth = width;
+    attr->newHeight = height;
+  }
 
   op->primitive->value.type = schema::PrimitiveType_Resize;
   op->primitive->value.value = attr.release();
 
   AddOpInput(op, tensors_id, tensors_format, tensors_id_map, tflite_op->inputs[0], tensors_id->size(),
              tflite_tensors.size(), schema::Format::Format_NHWC);
+  if (buffData == nullptr) {
+    AddOpInput(op, tensors_id, tensors_format, tensors_id_map, tflite_op->inputs[1], tensors_id->size(),
+               tflite_tensors.size(), schema::Format::Format_NHWC);
+  }
   AddOpOutput(op, tensors_id, tensors_format, tensors_id_map, tflite_op->outputs[0], tensors_id->size(),
               tflite_tensors.size(), schema::Format::Format_NHWC);
   return RET_OK;