加载数据集
import torch
from torch import nn
from torch.nn import functional as F
from torch import optim
import torchvision
from matplotlib import pyplot as plt
from utils import plot_image,plot_curve,one_hot
batch_size=512
train_loader=torch.utils.data.DataLoader(
torchvision.datasets.MNIST('mnist data',train=True,download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
(0.1307,),(0.3081,)
)
])),
batch_size=batch_size,shuffle=True
)
test_loader=torch.utils.data.DataLoader(
torchvision.datasets.MNIST('mnist data',train=False,download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
(0.1307,),(0.3081,)
)
])),
batch_size=batch_size,shuffle=True
)
x,y=next(iter(train_loader))
print(x.shape,y.shape,x.min(),x.max())
plot_image(x,y,'image sample')结果
torch.Size([512, 1, 28, 28]) torch.Size([512]) tensor(-0.4242) tensor(2.8215)
训练
# 创建网络：三层非线性层的嵌套
class Net(nn.Module):#继承nn.Module
def __init__(self):
super(Net, self).__init__()
# xw+b 线性层
self.fc1=nn.Linear(28*28,256)#输入的是28*28维特征（固定）的x->输出256（经验得出）维特征的y 此y就是一个中间过度的输出、
self.fc2=nn.Linear(256, 64)
self.fc3=nn.Linear(64,10)#10分类需要是个输出节点 所以一定是10 这是个维度就表示每一个样本对于0-9不同的概率
def forward(self, x):
#x:[b,1,28,28] b表示b章图片
x=F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
# 最后一层加不加激活函数取决与经验值 这个就不加
x=(self.fc3(x))
return x
# 训练 求导跟更新
# 网络对象
net=Net()
# 优化器
optimizer = optim.SGD(net.parameters(), lr=0.01,momentum=0.9)
train_loss=[]
# 对数据集迭代三次
for epoch in range(3):
# 对整个数据集迭代一次
for batch_idx,(x,y) in enumerate(train_loader):
# print(x.shape, y.shape)
# x:[b=512,1,28,28],y:[512]
# 全连接层xw+b层 此网络只能接受[b,feature]这样类型，而实际x:[b=512,1,28,28]，所以需要打平
# 打平[b,1,28,28]=>[b,784]
x=x.view(x.size(0),28*28)
out=net(x)
# print(out.shape)=torch.Size([512, 10])
y_onehot=one_hot(y)
loss=F.mse_loss(out, y_onehot)
train_loss.append(loss.item())
optimizer.zero_grad()
loss.backward()
# w=w-lr*grad
optimizer.step()
if batch_idx % 10 == 0:
print(epoch,batch_idx,loss.item())
plot_curve(train_loss)
# 得到了[w1,b1,w2,b2,w3,b3]
0 0 0.11067488044500351
0 10 0.0943082645535469
0 20 0.08495017886161804
0 30 0.07744021713733673
0 40 0.0714396983385086
0 50 0.0674654096364975
0 60 0.06324262917041779
0 70 0.059055496007204056
0 80 0.05903492122888565
0 90 0.0553865060210228
0 100 0.051882218569517136
0 110 0.04975864663720131
1 0 0.05097641423344612
1 10 0.045732591301202774
1 20 0.04423248767852783
1 30 0.04524523392319679
1 40 0.04314319044351578
1 50 0.04154777526855469
1 60 0.041426219046115875
1 70 0.038350772112607956
1 80 0.03864293918013573
1 90 0.039954546838998795
1 100 0.03705741837620735
1 110 0.03812757134437561
2 0 0.0377984382212162
2 10 0.03647816926240921
2 20 0.03585110604763031
2 30 0.03720710426568985
2 40 0.03573155403137207
2 50 0.035967521369457245
2 60 0.03404301032423973
2 70 0.03343154117465019
2 80 0.03460792452096939
2 90 0.03182953968644142
2 100 0.032444968819618225
2 110 0.029968297109007835
测试 求准确率
#准确度测试
total_correct=0
for x,y in test_loader:
x=x.view(x.size(0), 28 * 28)
out = net(x)
# out:[b,10]
# 求第一个维度上面 （不是第0个 就是10所在的维度）最大值的索引
# pred:[b]
pred=out.argmax(dim=1)
correct = pred.eq(y).sum().float()
total_correct += correct.item()
total_num = len(test_loader.dataset)
acc=total_correct / total_num
print(acc)结果0.8792
import torch
from torch import nn
from torch.nn import functional as F
from torch import optim
import torchvision
from matplotlib import pyplot as plt
from utils import plot_image,plot_curve,one_hot
# 加载数据
batch_size=512
train_loader=torch.utils.data.DataLoader(
torchvision.datasets.MNIST('mnist data',train=True,download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
(0.1307,),(0.3081,)
)
])),
batch_size=batch_size,shuffle=True
)
test_loader=torch.utils.data.DataLoader(
torchvision.datasets.MNIST('mnist data',train=False,download=True,
transform=torchvision.transforms.Compose([
torchvision.transforms.ToTensor(),
torchvision.transforms.Normalize(
(0.1307,),(0.3081,)
)
])),
batch_size=batch_size,shuffle=True
)
# x,y=next(iter(train_loader))
# print(x.shape,y.shape,x.min(),x.max())
# torch.Size([512, 1, 28, 28]) torch.Size([512]) tensor(-0.4242) tensor(2.8215)
# plot_image(x,y,'image sample')
# 创建网络：三层非线性层的嵌套
class Net(nn.Module):#继承nn.Module
def __init__(self):
super(Net, self).__init__()
# xw+b 线性层
self.fc1=nn.Linear(28*28,256)#输入的是28*28维特征（固定）的x->输出256（经验得出）维特征的y 此y就是一个中间过度的输出、
self.fc2=nn.Linear(256, 64)
self.fc3=nn.Linear(64,10)#10分类需要是个输出节点 所以一定是10 这是个维度就表示每一个样本对于0-9不同的概率
def forward(self, x):
#x:[b,1,28,28] b表示b章图片
x=F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
# 最后一层加不加激活函数取决与经验值 这个就不加
x=(self.fc3(x))
return x
# 训练 求导跟更新
# 网络对象
net=Net()
# 优化器
optimizer = optim.SGD(net.parameters(), lr=0.01,momentum=0.9)
train_loss=[]
# 对数据集迭代三次
for epoch in range(3):
# 对整个数据集迭代一次
for batch_idx,(x,y) in enumerate(train_loader):
# print(x.shape, y.shape)
# x:[b=512,1,28,28],y:[512]
# 全连接层xw+b层 此网络只能接受[b,feature]这样类型，而实际x:[b=512,1,28,28]，所以需要打平
# 打平[b,1,28,28]=>[b,784]
x=x.view(x.size(0),28*28)
out=net(x)
# print(out.shape)=torch.Size([512, 10])
y_onehot=one_hot(y)
loss=F.mse_loss(out, y_onehot)
train_loss.append(loss.item())
optimizer.zero_grad()
loss.backward()
# w=w-lr*grad
optimizer.step()
# if batch_idx % 10 == 0:
# print(epoch,batch_idx,loss.item())
# plot_curve(train_loss)
# 得到了[w1,b1,w2,b2,w3,b3]
#准确度测试
total_correct=0
for x,y in test_loader:
x=x.view(x.size(0), 28 * 28)
out = net(x)
# out:[b,10]
# 求第一个维度上面 （不是第0个 就是10所在的维度）最大值的索引
# pred:[b]
pred=out.argmax(dim=1)
correct = pred.eq(y).sum().float()
total_correct += correct.item()
total_num = len(test_loader.dataset)
acc=total_correct / total_num
print(acc)
4409
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
