如何在pytorch中将CNN模型转换为宽卷积？

import torch
import torch.nn as nn

class CNNModel(nn.Module):
    def __init__(self):
        super(CNNModel, self).__init__()
        # 定义你的CNN模型结构
        self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size)
        self.conv2 = nn.Conv2d(in_channels, out_channels, kernel_size)
        # ...

    def forward(self, x):
        # 定义CNN模型的前向传播过程
        x = self.conv1(x)
        x = self.conv2(x)
        # ...
        return x

class WideConvModel(nn.Module):
    def __init__(self, cnn_model):
        super(WideConvModel, self).__init__()
        self.cnn_model = cnn_model
        # 在这里定义宽卷积的层

    def forward(self, x):
        # 在这里实现宽卷积的前向传播过程
        x = self.cnn_model(x)
        # ...
        return x

cnn_model = CNNModel()
cnn_model.load_state_dict(torch.load('cnn_model_weights.pth'))

wide_conv_model = WideConvModel(cnn_model)