MindSpore报TypeError: Cannot joined the return values

1 报错描述

MindSpore报TypeError: Cannot joined the return values of different branches, perhaps you need to make them equal.

1.1 系统环境
*Hardware Environment(Ascend/GPU/CPU): Ascend/GPU/CPU

Software Environment:

– MindSpore version (source or binary): 1.6.0

– Python version (e.g., Python 3.7.5): 3.7.6

– OS platform and distribution (e.g., Linux Ubuntu 16.04): Ubuntu 4.15.0-74-generic

– GCC/Compiler version (if compiled from source):

1.2 基本信息

1.2.1 脚本

在控制流场景中，对同一个变量在不同分支赋不同类型的数据。

import mindspore as ms
import mindspore.nn as nn
from mindspore import ms_function, context

context.set_context(mode=context.GRAPH_MODE)

def test_join():
  class JoinNet(nn.Cell):
    def construct(self, x):
      out = [1, 2, 3]
      if x > 0:
        out = x
      return out
  x = ms.Tensor([1])
  net = JoinNet()
  res = net(x)
  print("res:", res)

1.2.2 报错

这里报错信息如下：

TypeError: mindspre/ccsrc/pipeline/jit/static_analysis/static_analysis.cc:780 ProcessEvalResults] Cannot join the return values of different branches, perhaps you need to make them equal.
Type Join Failed: Abstract type AbstractTensor cannot join with AbstractList.
In file ../test.py(11)
   if  x > 0:
   ^

原因分析:

我们看报错信息，在TypeError中，写到Cannot join the return values of different branches，意思是不支持两个不同分支的join操作，out在true分支被赋值为Tensor类型，在false分支被赋值为list类型，条件x > y为变量，因此在return out语句中无法确定out的数据类型，会导致图模式下的编译出现异常。

2 解决方法

基于上面已知的原因，很容易做出如下修改：

import mindspore as ms
import mindspore.nn as nn
from mindspore import ms_function, context

context.set_context(mode=context.GRAPH_MODE)

def test_join():
  class JoinNet(nn.Cell):
    def construct(self, x):
      out = Tensor([1, 2, 3])
      if x > 0:
        out = x
      return out
  x = ms.Tensor([1])
  net = JoinNet()
  res = net(x)
  print("res:", res)

此时执行成功，输出如下：

res: [1]

3 总结

定位报错问题的步骤：

1、找到报错的用户代码行：if x > 0:
2、分析控制流的两个分支的out数据类型，将其调整成相同类型。关注变量定义、初始化的正确性。

4 参考文档

https://www.mindspore.cn/docs/programming_guide/zh-CN/r1.6/control_flow.html