如何在 pytorch 的 DataLoader 中查看数据

Question

I see something like the following in the examples on Github. How can I see the type of this data (shape and the other properties)?

train_data = MyDataset(int(1e3), length=50)
train_iterator = DataLoader(train_data, batch_size=1000, shuffle=True)

Answer 1

You can inspect the data with following statements:

data = train_iterator.dataset.data 
shape = train_iterator.dataset.data.shape  
datatype = train_iterator.dataset.data.dtype

You can iterate the data and feed to a network as:

for nth_batch, (batch,_) in enumerate(train_iterator):
    feedable = Variable(batch)
    #here goes neural nets part

As Ivan stated in comments Variable is deprecated (although it still works fine) and Tensor itself now supports autograd, so the batch can be used in neural net.

for nth_batch, (batch,_) in enumerate(train_iterator):
    #feedforward the batch