- 导入一些必要的库
import torch
import torch.utils.dada as Data
import torch.nn as nn
import torchvision
2.准备数据,这次需要MNIST手写数字数据集,利用torchvision来获得数据集。,torchvisino除了MNIST,还有ciffar10等等许多数据集。有了torchvison模块,很方便对数据进行下载,处理.
train_data=
data_loader=
3.定义网络
class Net():
def __init__(self,n_input):
super(Net,self).__init__()
self.conv1=nn.Sequecial(
)
def forward():
4.训练
loss_func=
optimizer=torch.optim.ENTROPYloss()
for epoch in EPOCH:
for step,(b_x,b_y) in enumerate (data_loader):
output =net(b_x)
loss=loss_func(output,b_y)
optimizer.zero_grad()
loss.backward()
optimizer.step()