MINet 报错：(原因已找到)The size of tensor a (2) must match the size of tensor b (4) at non-singleton

遇到问题如下：

Traceback (most recent call last):                                              
  File "main.py", line 29, in
    solver.test()
  File "/home/nk/zjc/PycharmProjects/3ClassicAlgorithm/MINet/MINet(Relu)/MINet (weight)(FPNres) (selfmade-channel)(x2)/MINet-master/code/utils/solver.py", line 236, in test
    results = self.__test_process(save_pre=self.save_pre)
  File "/home/nk/zjc/PycharmProjects/3ClassicAlgorithm/MINet/MINet(Relu)/MINet (weight)(FPNres) (selfmade-channel)(x2)/MINet-master/code/utils/solver.py", line 274, in __test_process
    outputs = self.net(in_imgs)
  File "/home/nk/anaconda3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/nk/zjc/PycharmProjects/3ClassicAlgorithm/MINet/MINet(Relu)/MINet (weight)(FPNres) (selfmade-channel)(x2)/MINet-master/code/network/MINet.py", line 171, in forward
    in_data_1, in_data_2, in_data_4, in_data_8, in_data_16, w0a, w1a, w2a, w3a, w4a, w0b, w1b, w2b, w3b, w4b   # todo W
  File "/home/nk/anaconda3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/nk/zjc/PycharmProjects/3ClassicAlgorithm/MINet/MINet(Relu)/MINet (weight)(FPNres) (selfmade-channel)(x2)/MINet-master/code/module/MyModule.py", line 468, in forward
    out_xs.append(self.conv1(in_data_1, in_data_2, in_data_4, w1a, w1b))  # todo
  File "/home/nk/anaconda3/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/nk/zjc/PycharmProjects/3ClassicAlgorithm/MINet/MINet(Relu)/MINet (weight)(FPNres) (selfmade-channel)(x2)/MINet-master/code/module/MyModule.py", line 299, in forward
    out = self.relu(self.bnm_3(self.m2m_3(m)) * wa + self.identity(in_m) * wb)  # todo
RuntimeError: The size of tensor a (2) must match the size of tensor b (4) at non-singleton dimension 0

 

原因如下：

运行Pytorch时出现了这样一个错误，原因在于计算二分类交叉熵损失函数时是在每个batch中进行的，而总的图片数量并不能被所设置的batch_size整除，造成最后一个batch的图片数量与batch_size不相等。

因为我设置的值是4个一组，但是实际情况并不是能被4整除

https://blog.csdn.net/S20144144/article/details/100015058