!24 Change strategy for structure output

Merge pull request !24 from 步学/structure-output

mindspore/ccsrc/pipeline/pipeline.cc

@@ -725,23 +725,15 @@ py::object ExecutorPy::Run(const py::tuple& args, const py::object& phase) {
   return BaseRefToPyData(value);
 }
 
-py::object StructureOutput(const AbstractBasePtr& output, const py::tuple& data, size_t* count) {
+py::object ExtractGeneralCnodeRet(const AbstractBasePtr& cnode_data, const py::tuple& data, size_t* count) {
-  MS_EXCEPTION_IF_NULL(output);
+  MS_EXCEPTION_IF_NULL(cnode_data);
+  if (*count >= data.size()) {
+    MS_LOG(EXCEPTION) << "The number of elements in the outputs : " << data.size()
+                      << " less than the number of elements required. ";
+  }
 
-  if (!output->isa<AbstractTuple>()) {
+  if (cnode_data->isa<AbstractTensor>()) {
-    ValuePtr value = output->BuildValue();
+    BaseShapePtr shape = cnode_data->BuildShape();
-    if (value != kAnyValue) {
-      return ValuePtrToPyData(value);
-    }
-    if (!output->isa<AbstractTensor>()) {
-      MS_LOG(EXCEPTION) << "Output can only be tensor except for constants, but got "
-                        << output->BuildValue()->ToString() << ".";
-    }
-    if (*count >= data.size()) {
-      MS_LOG(EXCEPTION) << "The number of elements in the outputs : " << data.size()
-                        << " less than the number of elements required. ";
-    }
-    auto shape = output->BuildShape();
     auto shape_act = shape->cast<abstract::ShapePtr>()->shape();
     Tensor tensor_exp = py::cast<Tensor>(data[*count]);
     if (shape_act != tensor_exp.shape()) {
@@ -751,16 +743,58 @@ py::object StructureOutput(const AbstractBasePtr& output, const py::tuple& data,
     return data[(*count)++];
   }
 
-  auto tuple_output = output->cast<AbstractTuplePtr>();
+  if (!cnode_data->isa<AbstractTuple>()) {
-  AbstractBasePtrList elements = tuple_output->elements();
+    MS_LOG(EXCEPTION) << "The output of operator in the final anf graph could "
-  size_t size = elements.size();
+                      << "only be a tensor or a tuple of tensor, but got " << cnode_data->BuildValue()->ToString()
+                      << ".";
+  }
+  auto data_tp = cnode_data->cast<AbstractTuplePtr>();
+  auto elements = data_tp->elements();
+  size_t size = data_tp->size();
   py::tuple tp = py::tuple(size);
   for (size_t i = 0; i < size; i++) {
-    tp[i] = StructureOutput(elements[i], data, count);
+    tp[i] = ExtractGeneralCnodeRet(elements[i], data, count);
   }
   return std::move(tp);
 }
 
+py::object StructureOutput(const AnfNodePtr& output_node, const py::tuple& data, size_t* count) {
+  MS_EXCEPTION_IF_NULL(output_node);
+
+  if (output_node->isa<ValueNode>()) {
+    return ValuePtrToPyData(GetValueNode(output_node));
+  }
+
+  if (*count >= data.size()) {
+    MS_LOG(EXCEPTION) << "The number of elements in the outputs : " << data.size()
+                      << " less than the number of elements required. ";
+  }
+  if (output_node->isa<Parameter>()) {
+    return data[(*count)++];
+  }
+
+  auto output_c = output_node->cast<CNodePtr>();
+  if (output_c == nullptr) {
+    MS_LOG(EXCEPTION) << "The final anf graph could only have constant, parameter, and operator, but got "
+                      << output_node->ToString();
+  }
+
+  if (output_c->IsApply(prim::kPrimMakeTuple)) {
+    auto input_list = output_c->inputs();
+    size_t size = input_list.size();
+    py::tuple tp = py::tuple(size - 1);
+    for (size_t i = 1; i < size; i++) {
+      tp[i - 1] = StructureOutput(input_list[i], data, count);
+    }
+    return std::move(tp);
+  }
+  if (output_c->IsApply(prim::kPrimDepend)) {
+    return StructureOutput(output_c->input(1), data, count);
+  }
+
+  return ExtractGeneralCnodeRet(output_c->abstract(), data, count);
+}
+
 std::shared_ptr<py::object> DoExecGraph(const FuncGraphPtr& graph, const std::vector<MeTensorPtr>& inputs,
                                         const std::string& phase) {
   std::vector<GeTensorPtr> ge_tensors = TransformUtil::ConvertInputTensors(inputs, kOpFormat_NCHW);
@@ -806,11 +840,10 @@ std::shared_ptr<py::object> DoExecGraph(const FuncGraphPtr& graph, const std::ve
   std::shared_ptr<py::object> ret = nullptr;
 
 #ifdef ENABLE_GE
-  AnfNodePtr root = graph->get_return();
+  AnfNodePtr output_node = graph->get_return()->input(1);
-  MS_EXCEPTION_IF_NULL(root);
+  MS_EXCEPTION_IF_NULL(output_node);
-  AbstractBasePtr output = root->abstract();
   size_t count = 0;
-  py::object oj = StructureOutput(output, outputs, &count);
+  py::object oj = StructureOutput(output_node, outputs, &count);
   ret = std::make_shared<py::object>(oj);
 #else
   if (outputs.size() == 1) {

tests/ut/python/nn/test_structure_output.py

@@ -236,7 +236,7 @@ def test_soft():
         def __init__(self):
             super(SoftmaxCrossEntropyWithLogitsNet, self).__init__()
             self.soft = P.SoftmaxCrossEntropyWithLogits()
-            self.value = (Tensor(np.zeros((2,)).astype(np.float32)), Tensor(np.ones((2,)).astype(np.float32)))
+            self.value = (Tensor(np.zeros((2, 2)).astype(np.float32)), Tensor(np.ones((2, 2)).astype(np.float32)))
 
         def construct(self, x, y, z):
             xx = x + y
@@ -246,8 +246,30 @@ def test_soft():
             ret = (ret, self.value)
             return ret
 
-    input1 = Tensor(np.zeros((2,)).astype(np.float32))
+    input1 = Tensor(np.zeros((2, 2)).astype(np.float32))
-    input2 = Tensor(np.ones((2,)).astype(np.float32))
+    input2 = Tensor(np.ones((2, 2)).astype(np.float32))
-    input3 = Tensor((np.ones((2,)) + np.ones((2,))).astype(np.float32))
+    input3 = Tensor((np.ones((2, 2)) + np.ones((2, 2))).astype(np.float32))
     net = SoftmaxCrossEntropyWithLogitsNet()
-    print(net(input1, input2, input3))
+    net(input1, input2, input3)
+
+
+def test_const_depend():
+    class ConstDepend(Cell):
+        def __init__(self):
+            super(ConstDepend, self).__init__()
+            self.value = (Tensor(np.zeros((2, 3)).astype(np.float32)), Tensor(np.ones((2, 3)).astype(np.float32)))
+            self.soft = P.SoftmaxCrossEntropyWithLogits()
+            self.depend = depend
+
+        def construct(self, x, y, z):
+            ret = x + y
+            ret = ret * z
+            ret = self.depend(self.value, ret)
+            ret = (ret, self.soft(x, y))
+            return ret
+
+    input1 = Tensor(np.zeros((2, 2)).astype(np.float32))
+    input2 = Tensor(np.ones((2, 2)).astype(np.float32))
+    input3 = Tensor((np.ones((2, 2)) + np.ones((2, 2))).astype(np.float32))
+    net = ConstDepend()
+    net(input1, input2, input3)