通过随机种子使两个dataset打乱的顺序相同

import numpy as np
import torch
from torch.utils.data import Dataset, DataLoader

x = np.arange(11)
y = np.arange(11)
x2 = np.arange(0, 110, 10)
y2 = np.arange(0, 110, 10)

class SequenceDataset(Dataset):
    def __init__(self, input, target):
        self.input = torch.tensor(input)
        self.target = torch.tensor(target)

    def __len__(self):
        return len(self.target)

    def __getitem__(self, item):
        return self.input[item], self.target[item]

sequence_dataset = SequenceDataset(x, y)
loader = DataLoader(sequence_dataset, batch_size= 5, shuffle=True)

loader2 = DataLoader(SequenceDataset(x2, y2), batch_size= 5, shuffle=True)


for epoch in range(3):
    torch.manual_seed(0)
    for (x, y) in loader:
        print(x, y)
    torch.manual_seed(0)
    for (x2, y2) in loader2:
        print(x2, y2)

在每一次使用dataloader前，两个dataloader设置相同的随机种子即可。

输出结果：

 

 设置随机种子控制shuffle顺序的原理参考：https://www.zhihu.com/question/437269017