Resnet网络回归的简单实现与经典错误解析

注：参考《动手学深度学习》一书

错误分析参考网页：

1.https://blog.csdn.net/qq_35235032/article/details/107384079

2.https://betheme.net/news/txtlist_i43568v.html?action=onClick

原始运行代码请参考以上两个链接，为避免侵权

本解决方案包括大部分报错，直接获取以上链接中代码，进行调试即可

俩链接的代码都一样

经典主要错误与解决方式

AttributeError: 'NoneType' object has no attribute 'parameters'

原因是博客中给的代码未加入resnet层
解决方法：把ResNet()函数内的代码改成如下（将模型返回+加入resnet_block）,增加返回值，可以解决此问题

def ResNet():
    model = torch.nn.Sequential(
        torch.nn.Conv2d(1, 4, kernel_size=2, padding=1, stride=1), 
        torch.nn.BatchNorm2d(4),  
        torch.nn.ReLU(),
        torch.nn.MaxPool2d(kernel_size=2)
    )
 
    model.add_module("global_avg_pool", GlobalAvgPool2d())
    model.add_module("fc", torch.nn.Sequential(FlattenLayer(), torch.nn.Linear(32, 2)))
    return model
# 此处上面一行，加入了返回值 ！！！！！

 AttributeError: type object 'datetime.time' has no attribute 'time'

导入包 import time 

AttributeError: '_MultiProcessingDataLoaderIter' object has no attribute 'next'

dataloader是pytorch加载数据里常用方法，原作者在这里简化了包名，以下是dataloader原始使用方式：

    train_data = TensorDataset(train_x_tensor, train_y_tensor)
    test_data = TensorDataset(train_x_tensor, train_y_tensor)

    tr_loader = torch.utils.data.DataLoader(train_data, batch_size, True)
    test_loader = torch.utils.data.DataLoader(train_data, batch_size, True)

 解决方案：

第一，使用原始方式，所以的dataloader加载代码更换如下格式：

# 源代码：

test_batch = Data.DataLoader(torch_testset, batch_size=batch_size, shuffle=False, num_workers=works_num)

# 更改为：

test_batch = torch.utils.data.DataLoader(torch_testset, batch_size=batch_size, shuffle=False, num_workers=works_num)

或者，导入简化的包，代码如下：

import sys
from torch.utils.data import Dataset
import torch.utils.data as Data
import torch.nn.functional as F

RuntimeError: mat1 and mat2 shapes cannot be multiplied (100x4 and 32x2)

解决方式：原因是博客中给的代码未加入resnet层
两个报错的解决方法：把ResNet()函数内的代码改成如下（将模型返回+加入resnet_block）

def ResNet():
    model = torch.nn.Sequential(
        torch.nn.Conv2d(1, 4, kernel_size=2, padding=1, stride=1), 
        torch.nn.BatchNorm2d(4),  
        torch.nn.ReLU(),
        torch.nn.MaxPool2d(kernel_size=2),
        resnet_block(4, 32, 2) # 这里加入resnet ！！！！！！
    )
 
    model.add_module("global_avg_pool", GlobalAvgPool2d())
    model.add_module("fc", torch.nn.Sequential(FlattenLayer(), torch.nn.Linear(32, 2)))
    return model

model = model.to(device)
File "E:\software\Anaconda\lib\site-packages\torch\nn\modules\module.py", line 908, in to
device, dtype, non_blocking, convert_to_format = torch._C._nn._parse_to(*args, **kwargs)
TypeError: to() received an invalid combination of arguments - got (type), but expected one of:
* (torch.device device, torch.dtype dtype, bool non_blocking, bool copy, *, torch.memory_format memory_format)
* (torch.dtype dtype, bool non_blocking, bool copy, *, torch.memory_format memory_format)
* (Tensor tensor, bool non_blocking, bool copy, *, torch.memory_format memory_format)

解决方式：加上这段代码

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

在此处

# 3、训练验证模型
# 3.1 训练模型
def train(model, train_batch, test_batch, batch_size, optimizer, device, num_epochs):
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    model = model.to(device)
    print("run in ", device)

运行结果：

2023.4.24 

本文转载

Resnet网络回归的简单实现_resnet回归_飘满红楼的博客-CSDN博客

只是解决了作者代码的问题和在互联网其他地方搜集的解决方法，仅供参考，如有侵权可以联系删除