如何修复 UnboundLocalError：赋值前引用的局部变量“o1”

Question

我正在制作一个二元分类器并遇到了这个问题。

代码：

import torch.nn as nn
import torch.nn.functional as F
import numpy as np

class BinCNN(nn.Module):
    def __init__(self, n, i, j, i1, j1, d, fc1, fc2):
        super().__init__()
        """
        n - source frame number of channels
        i - 1-st kernel_size height
        j - 1-st kernel_size width
        i1 - 2-nd kernel_size height
        j1 - 2-nd kernel_size width
        d - dropout rate
        fc1 - 1-st Linear layer size
        fc2 - 2-nd Linear layer size
        """
        self.conv1 = nn.Conv2d(n, o1, kernel_size=[i, j])
        self.conv2 = nn.Conv2d(o1, o2, kernel_size=[i1, j1])
        o1 = self._get_conv1_out(n)
        o2 = self._get_conv2_out(o1)
        self.drop = nn.Dropout(d)
        self.fcl1 = nn.Linear(fc1, fc2)
        self.fcl2 = nn.Linear(fc2, 2)

    def _get_conv1_out(self, shape):
        o = self.conv1(torch.zeros(1, *shape))
        return int(np.prod(o.size()))

    def _get_conv2_out(self, shape):
        o = self.conv2(torch.zeros(1, *shape))
        return int(np.prod(o.size()))

当我写

sd = BinCNN(3, 2, 2, 1, 1, 0.3, 300, 150)

print(sd)

这给了我给定的错误：

Traceback (most recent call last):
  File "/home/name/Документы/work/science/train/train.py", line 46, in <module>
    sd = BinCNN(3, 2, 2, 1, 1, 0.3, 300, 150)
  File "/home/name/Документы/work/science/train/train.py", line 21, in __init__
    self.conv1 = nn.Conv2d(n, o1, kernel_size=[i, j])
UnboundLocalError: local variable 'o1' referenced before assignment

我试图通过在self.conv1之前添加model来修复它，但它没有帮助

我用谷歌搜索了这个但一无所获。 请帮帮我

（我删除def forward (self, x)因为 stackoverflow 发誓“有很多代码”）

Answer 1

在分别定义conv1和_get_conv2_out之前，您正在调用_get_conv1_out和conv2 。 它们还没有被定义：

    o1 = self._get_conv1_out(n)
    o2 = self._get_conv2_out(o1)
    self.conv1 = nn.Conv2d(n, o1, kernel_size=[i, j])
    self.conv2 = nn.Conv2d(o1, o2, kernel_size=[i1, j1])