解决错误AttributeError: module ‘torch.nn.parameter‘ has no attribute ‘UninitializedParameter‘的方法

在创建pyG中model的过程中

出现报错：

Traceback (most recent call last):
  File "C:\Users\Wang\PycharmProjects\pythonProject\唐老师\mmmmmmmm.py", line 31, in <module>
    model = GCN()
  File "C:\Users\Wang\PycharmProjects\pythonProject\唐老师\mmmmmmmm.py", line 12, in __init__
    self.conv1 = GCNConv(dataset.num_features, 4)
  File "F:\anaconda\envs\pyG\lib\site-packages\torch_geometric\nn\conv\gcn_conv.py", line 140, in __init__
    weight_initializer='glorot')
  File "F:\anaconda\envs\pyG\lib\site-packages\torch_geometric\nn\dense\linear.py", line 65, in __init__
    self.reset_parameters()
  File "F:\anaconda\envs\pyG\lib\site-packages\torch_geometric\nn\dense\linear.py", line 78, in reset_parameters
    if isinstance(self.weight, nn.parameter.UninitializedParameter):
AttributeError: module 'torch.nn.parameter' has no attribute 'UninitializedParameter'

原因是torch的版本不同，导致其中没有UninitializedParameter这个属性

可以更换torch版本或修改代码去掉这个变量，但是更换版本要重装，这里学习了网上的经验修改代码解决了问题

不会用到 ‘torch.nn.parameter’ 的’UninitializedParameter’属性了，因此修改后就不会报错了

1.按照上面报错的提示打开文件：

File "F:\anaconda\envs\pyG\lib\site-packages\torch_geometric\nn\dense\linear.py"

2.修改该linear文件

修改前：

from typing import Optional

import copy
import math

import torch
from torch import nn
from torch import Tensor
import torch.nn.functional as F
from torch.nn.parameter import Parameter

from torch_geometric.nn import inits


class Linear(torch.nn.Module):
    r"""Applies a linear tranformation to the incoming data

    .. math::
        \mathbf{x}^{\prime} = \mathbf{x} \mathbf{W}^{\top} + \mathbf{b}

    similar to :class:`torch.nn.Linear`.
    It supports lazy initialization and customizable weight and bias
    initialization.

    Args:
        in_channels (int): Size of each input sample. Will be initialized
            lazily in case it is given as :obj:`-1`.
        out_channels (int): Size of each output sample.
        bias (bool, optional): If set to :obj:`False`, the layer will not learn
            an additive bias. (default: :obj:`True`)
        weight_initializer (str, optional): The initializer for the weight
            matrix (:obj:`"glorot"`, :obj:`"uniform"`, :obj:`"kaiming_uniform"`
            or :obj:`None`).
            If set to :obj:`None`, will match default weight initialization of
            :class:`torch.nn.Linear`. (default: :obj:`None`)
        bias_initializer (str, optional): The initializer for the bias vector
            (:obj:`"zeros"` or :obj:`None`).
            If set to :obj:`None`, will match default bias initialization of
            :class:`torch.nn.Linear`. (default: :obj:`None`)
    """
    def __init__(self, in_channels: int, out_channels: int, bias: bool = True,
                 weight_initializer: Optional[str] = None,
                 bias_initializer: Optional[str] = None):
        super().__init__()
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.weight_initializer = weight_initializer
        self.bias_initializer = bias_initializer

        if in_channels > 0:
            self.weight = Parameter(torch.Tensor(out_channels, in_channels))
        else:
            self.weight = nn.parameter.UninitializedParameter()
            self._hook = self.register_forward_pre_hook(
                self.initialize_parameters)

        if bias:
            self.bias = Parameter(torch.Tensor(out_channels))