pytorch实现批训练

Posted 2021-03-10 loyolh

代码：

#进行批训练
import torch
import torch.utils.data as Data

BATCH_SIZE = 5  #每批5个数据

if __name__ == ‘__main__‘:
    x = torch.linspace(1, 10, 10)  #x是从1到10共10个数据
    y = torch.linspace(10, 1, 10)  #y是从10到1共10个数据

    #torch_dataset = Data.TensorDataset(data_tensor = x, target_tensor=y)会报错
    torch_dataset = Data.TensorDataset(x,y)
    loader = Data.DataLoader(      #使我们的训练变成一小批一小批的
        dataset = torch_dataset,   #将所有数据放入dataset中
        batch_size= BATCH_SIZE,
        shuffle=True,              #true训练的时候随机打乱数据,false不打乱
        num_workers=2,             #每次训练用两个线程或进程进行提取
    )   

    for epoch in range(3):
        for step, (batch_x, batch_y) in enumerate(loader):  #利用enumerate可以同时获得索引(step)和值
            print(‘Epoch:‘, epoch, ‘| Step:‘, step, ‘| batch_x:‘, 
            batch_x.numpy(), ‘| batch_y:‘, batch_y.numpy())

过程中遇到了问题，问题及解决办法都在https://blog.csdn.net/thunderf/article/details/94733747