PyTorch实现CNN遇RuntimeError：多进程启动异常求助

我在使用PyTorch实现CNN时，添加「训练与验证数据集」相关代码后遇到如下RuntimeError，此前所有步骤均正常运行。运行环境为Windows 11系统、PyCharm编辑器、Python 3.8版本。

错误信息 #

RuntimeError: 
        An attempt has been made to start a new process before the
        current process has finished its bootstrapping phase.

        This probably means that you are not using fork to start your
        child processes and you have forgotten to use the proper idiom
        in the main module:

            if __name__ == '__main__':
                freeze_support()
                ...

        The "freeze_support()" line can be omitted if the program
        is not going to be frozen to produce an executable.

我的代码 #

# Import whatever libraries/modules you need

import numpy as np
import tensorflow as tf
from torchvision.datasets import ImageFolder
from torchvision.transforms import ToTensor
import matplotlib
import matplotlib.pyplot as plt

import os
import torch
import torchvision
import tarfile
from torchvision.datasets.utils import download_url
from torch.utils.data import random_split
from torch.utils.data.dataloader import DataLoader
from torchvision.utils import make_grid

# Dowload the dataset
dataset_url = "https://s3.amazonaws.com/fast-ai-imageclas/cifar10.tgz"
download_url(dataset_url, '.')

# # Extract from archive to data/cifar10, containing training set and test set
# with tarfile.open('./cifar10.tgz', 'r:gz') as tar:
#     tar.extractall(path='./data')

# Verify using os.listdir
data_dir = './data/cifar10'

print(os.listdir(data_dir))
classes = os.listdir(data_dir + "/train")
print(classes)

# Load Data as PyTorch Tensors
dataset = ImageFolder(data_dir+'/train', transform=ToTensor())

#Display image using matplotlib but change the tensor dimensions to (32, 32, 3)
matplotlib.rcParams['figure.facecolor'] = '#ffffff'

def show_example(img, label):
    print('Label: ', dataset.classes[label], "("+str(label)+")")
    plt.imshow(img.permute(1, 2, 0))
    plt.show()

show_example(*dataset[0])
show_example(*dataset[1099])

# Training
random_seed = 42
torch.manual_seed(random_seed);
val_size = 5000
train_size = len(dataset) - val_size


train_ds, val_ds = random_split(dataset, [train_size, val_size])
len(train_ds), len(val_ds)

batch_size=128
train_dl = DataLoader(train_ds, batch_size, shuffle=True, num_workers=4, pin_memory=True)
val_dl = DataLoader(val_ds, batch_size*2, num_workers=4, pin_memory=True)

# Look at batches of images from the dataset using the make grid method
# Gets a random batch each time since the sampler shuffles the indices before creating batches

def show_batch(dl):
    for images, labels in dl:
        fig, ax = plt.subplots(figsize=(12, 6))
        ax.set_xticks([]); ax.set_yticks([])
        ax.imshow(make_grid(images, nrow=16).permute(1, 2, 0))
        break
show_batch(train_dl)

解决方案 #

Windows系统下Python多进程采用spawn而非Linux的fork机制，子进程会重新执行整个脚本的全局代码，因此必须将触发多进程的逻辑（如DataLoader初始化、数据加载执行）放在if __name__ == '__main__':代码块内，避免子进程重复初始化导致错误。

修改后的代码示例：

# Import whatever libraries/modules you need

import numpy as np
import tensorflow as tf
from torchvision.datasets import ImageFolder
from torchvision.transforms import ToTensor
import matplotlib
import matplotlib.pyplot as plt

import os
import torch
import torchvision
import tarfile
from torchvision.datasets.utils import download_url
from torch.utils.data import random_split
from torch.utils.data.dataloader import DataLoader
from torchvision.utils import make_grid

# Display image using matplotlib but change the tensor dimensions to (32, 32, 3)
matplotlib.rcParams['figure.facecolor'] = '#ffffff'

def show_example(img, label, classes):
    print('Label: ', classes[label], "("+str(label)+")")
    plt.imshow(img.permute(1, 2, 0))
    plt.show()

def show_batch(dl):
    for images, labels in dl:
        fig, ax = plt.subplots(figsize=(12, 6))
        ax.set_xticks([]); ax.set_yticks([])
        ax.imshow(make_grid(images, nrow=16).permute(1, 2, 0))
        break

if __name__ == '__main__':
    # Dowload the dataset
    dataset_url = "https://s3.amazonaws.com/fast-ai-imageclas/cifar10.tgz"
    download_url(dataset_url, '.')

    # # Extract from archive to data/cifar10, containing training set and test set
    # with tarfile.open('./cifar10.tgz', 'r:gz') as tar:
    #     tar.extractall(path='./data')

    # Verify using os.listdir
    data_dir = './data/cifar10'

    print(os.listdir(data_dir))
    classes = os.listdir(data_dir + "/train")
    print(classes)

    # Load Data as PyTorch Tensors
    dataset = ImageFolder(data_dir+'/train', transform=ToTensor())

    show_example(*dataset[0], classes)
    show_example(*dataset[1099], classes)

    # Training
    random_seed = 42
    torch.manual_seed(random_seed)
    val_size = 5000
    train_size = len(dataset) - val_size

    train_ds, val_ds = random_split(dataset, [train_size, val_size])
    print(len(train_ds), len(val_ds))

    batch_size=128
    train_dl = DataLoader(train_ds, batch_size, shuffle=True, num_workers=4, pin_memory=True)
    val_dl = DataLoader(val_ds, batch_size*2, num_workers=4, pin_memory=True)

    show_batch(train_dl)

说明：

• 将所有执行逻辑（包括数据下载、加载、DataLoader初始化、图像展示）放入if __name__ == '__main__':块内
• 调整show_example函数，将classes作为参数传入，避免子进程中无法访问全局变量的问题
• 保留函数定义、库导入在全局区域，不影响子进程调用

内容的提问来源于stack exchange，提问作者Derz