1 报错描述
1.1 系统环境
Hardware Environment(Ascend/GPU/CPU): Ascend
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
训练脚本是通过构建BinaryCrossEntropy单算子网络，计算两个变量二进制交叉熵的例子。脚本如下：
import mindspore.nn as nn
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
self.binary_cross_entropy = ops.BinaryCrossEntropy()
self.weight = None
def construct(self, logits, labels):
result = self.binary_cross_entropy(logits, labels)
return result
logits = Tensor(np.random.uniform(0, 1, (4, 3, 388, 388)).astype(np.float32))
labels = Tensor(np.random.randint(0, 2, (4, 2, 388, 388)).astype(np.float32))
net = Net()
out = net(logits, labels)
print(‘out:’,out)
1.2.2 报错
这里报错信息如下：
Traceback (most recent call last):
File "demo.py", line 14, in <module>
out = net(logits, labels)
File "../lib/python3.7/site-packages/mindspore/nn/cell.py", line 569, in __call__
out = self.compile_and_run(*args)
File "../lib/python3.7/site-packages/mindspore/nn/cell.py", line 899, in compile_and_run
self.compile(*inputs)
File "../lib/python3.7/site-packages/mindspore/nn/cell.py", line 884, in compile
_cell_graph_executor.compile(self, *inputs, phase=self.phase, auto_parallel_mode=self._auto_parallel_mode)
File "../lib/python3.7/site-packages/mindspore/common/api.py", line 784, in compile
result = self._graph_executor.compile(obj, args_list, phase, self._use_vm_mode())
ValueError: mindspore/core/utils/check_convert_utils.h:216 Check] For primitive[BinaryCrossEntropy], the x shape: [4,3,388,388,] must be equal to [4,2,388,388,]
The function call stack (See file 'demo/rank_0/om/analyze_fail.dat' for more details):
# 0 In file demo.py(08)
result = self.binary_cross_entropy(logits, labels, self.weight)
原因分析
我们看报错信息，在ValueError中，写到For primitive[BinaryCrossEntropy], the x shape: [4,3,388,388,] must be equal to [4,2,388,388,]，意思是传的labels的shape是 [4, 2, 388, 388]，但是你传进去的logits的shape却是[4, 3, 388, 388]，而所有的损失函数都会要求logits和labels的shape必须一样，在官网中对BinaryCrossEntropy输入logits和labels的shape做了限制说明。
检查代码发现，12行代码labels的维度为确实不等于logits，此时需要检查传入值是否有问题，使logits和labels的shape相等。
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.nn as nn
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
self.binary_cross_entropy = ops.BinaryCrossEntropy()
self.weight = None
def construct(self, logits, labels):
result = self.binary_cross_entropy(logits, labels)
return result
logits = Tensor(np.random.uniform(0, 1, (4, 3, 388, 388)).astype(np.float32))
labels = Tensor(np.random.randint(0, 2, (4, 3, 388, 388)).astype(np.float32))
net = Net()
out = net(logits, labels)
print(‘out:’,out)
此时执行成功，输出如下：
out: 1.0000657
3 总结
定位报错问题的步骤：
1、找到报错的用户代码行： result = self.binary_cross_entropy(logits, labels, self.weight) ;
2、 根据日志报错信息中的关键字，缩小分析问题的范围： the x shape: [4,3,388,388,] must be equal to [4,2,388,388,] ;
3、需要重点关注变量定义、初始化的正确性。