pytorch 之 batch_train

Posted 2021-03-24 dhname

 1 import torch
 2 import torch.utils.data as Data
 3 
 4 torch.manual_seed(1)    # reproducible
 5 
 6 BATCH_SIZE = 5
 7 # BATCH_SIZE = 8
 8 
 9 x = torch.linspace(1, 10, 10)       # this is x data (torch tensor)
10 y = torch.linspace(10, 1, 10)       # this is y data (torch tensor)
11 
12 torch_dataset = Data.TensorDataset(x, y)
13 loader = Data.DataLoader(
14     dataset=torch_dataset,      # torch TensorDataset format
15     batch_size=BATCH_SIZE,      # mini batch size
16     shuffle=True,               # random shuffle for training
17     num_workers=2,              # subprocesses for loading data
18 )
19 
20 
21 def show_batch():
22     for epoch in range(3):   # train entire dataset 3 times
23         for step, (batch_x, batch_y) in enumerate(loader):  # for each training step
24             # train your data...
25             print(‘Epoch: ‘, epoch, ‘| Step: ‘, step, ‘| batch x: ‘,
26                   batch_x.numpy(), ‘| batch y: ‘, batch_y.numpy())
27 
28 
29 if __name__ == ‘__main__‘:
30     show_batch()