pytorch-custom dataset

import torch

import torch.utils.data as data

1.

 class CustomDataset(data.Dataset):

__init__(self, transform)

__len__(self)

__getitem__(self, index): 

    transform(data)

loader = data.DataLoader(dataset=torch_dataset, batch_size= bz, shuffle=True/False, num_workers=3)

for epoch in range(5):

    for step, (batch_x, batch_y) in enumerate(loader):

          pass

transform:

transform = transforms.Compose([transforms.Scale(40), transforms.RandomHorizontalFlip(), transforms.RandomCrop(32), transforms.ToTensor()])

transforms.Normalize((0.1307,), (0.3081,))、

    mean=[0.485,0.456,0.406],std=[0.229,0.224,0.225]

----------------------------------------------------------reference---------------------

1. https://github.com/pytorch/vision/blob/master/torchvision/datasets/mnist.py

2. http://blog.csdn.net/u012436149/article/details/69061711