python-pytorch关于next和iter使用时没有效果的笔记

问题

自己写来一个自定义数据集的类，使用dataloader去加载，然后使用next和iter去访问，每次访问到的数据都是一致的

datas,lables=next(iter(train_dataloader))
print(lables)
datas1,lables1=next(iter(train_dataloader))
print(lables1)

输出

tensor([9, 0, 0, 3, 0, 2, 7, 2, 5, 5, 0, 9, 5, 5, 7, 9, 1, 0, 6, 4, 3, 1, 4, 8,
        4, 3, 0, 2, 4, 4, 5, 3, 6, 6, 0, 8, 5, 2, 1, 6, 6, 7, 9, 5, 9, 2, 7, 3,
        0, 3, 3, 3, 7, 2, 2, 6, 6, 8, 3, 3, 5, 0, 5, 5])
tensor([9, 0, 0, 3, 0, 2, 7, 2, 5, 5, 0, 9, 5, 5, 7, 9, 1, 0, 6, 4, 3, 1, 4, 8,
        4, 3, 0, 2, 4, 4, 5, 3, 6, 6, 0, 8, 5, 2, 1, 6, 6, 7, 9, 5, 9, 2, 7, 3,
        0, 3, 3, 3, 7, 2, 2, 6, 6, 8, 3, 3, 5, 0, 5, 5])

原因

每次都根据数据集生成了一个迭代器，所以执行结果是一样的

解决

改成同一迭代器，再next就还可以了，问题比较低级

cc2=iter(train_dataloader)
datas,lables=next(cc2)
print(lables)
datas1,lables1=next(cc2)
print(lables1)

输出

tensor([9, 0, 0, 3, 0, 2, 7, 2, 5, 5, 0, 9, 5, 5, 7, 9, 1, 0, 6, 4, 3, 1, 4, 8,
        4, 3, 0, 2, 4, 4, 5, 3, 6, 6, 0, 8, 5, 2, 1, 6, 6, 7, 9, 5, 9, 2, 7, 3,
        0, 3, 3, 3, 7, 2, 2, 6, 6, 8, 3, 3, 5, 0, 5, 5])
tensor([0, 2, 0, 0, 4, 1, 3, 1, 6, 3, 1, 4, 4, 6, 1, 9, 1, 3, 5, 7, 9, 7, 1, 7,
        9, 9, 9, 3, 2, 9, 3, 6, 4, 1, 1, 8, 8, 0, 1, 1, 6, 8, 1, 9, 7, 8, 8, 9,
        6, 6, 3, 1, 5, 4, 6, 7, 5, 5, 9, 2, 2, 2, 7, 6])
        ```