Pytorch github代码链接:https://github.com/L1aoXingyu/pytorch-beginner
ch3:多层全连接神经网络
3.1 热身:Pytorch基础
3.1.1 Tensor(张量)
- 不同数据类型的tensor:
有32位浮点型torch.FloatTensor #Tensor默认数据类型
64位浮点型torch.DoubleTensor
16位整型torch.ShortTensor
32位整型torch.IntTensor
64位整型torch.LongTensor
tensor 应用实例:
import torch
a=torch.Tensor([[1,2],[2,3],[3,4]]);
print('a is:{}'.format(a))
print('a size is :{}'.format(a.size()))
#改变tensor的类型
b=torch.LongTensor([[1,2],[2,3],[3,4]])
print('b is:{}'.format(b))
#建立一个全零的tensor
c=torch.zeros((3,2))
print('c is:{}'.format(c))
#取一个正态分布作为随机初始值
d=torch.randn((3,2))
print('d is:{}'.format(d))
运行结果:
a is:
1 2
2 3
3 4
[torch.FloatTensor of size 3x2]
a size is :torch.Size([3, 2])
b is:
1 2
2 3
3 4
[torch.LongTensor of size 3x2]
c is:
0 0
0 0
0 0
[torch.FloatTensor of size 3x2]
d is:
-0.2037 -0.6000
-0.4322 0.2700
0.6836 0.3145
[torch.FloatTensor of size 3x2]
- 可以像numpy一样通过索引的方式取得其中的元素,同时改变它的值:
比如:
a[0,1]=100
print('a is:{}'.format(a))
运行结果为:
a is:
1 100
2 3
3 4
[torch.FloatTensor of size 3x2]
- 除此之外,还可以在Tensor和numpy.ndarray 之间相互转换。
通过b.numpy()可以将tensor b转换为numpy
通过torch.from_numpy(e)可以讲numpy e转换为tensor;
numpy_b=b.numpy()
print('b convert to numpy is:{}'.format(numpy_b))
运行结果为:
b convert to numpy is:[[1 2]
[2 3]
[3 4]]
e=np.array([[2,3],[4,5]])
print('e is:{}'.format(e))
torch_e=torch.from_numpy(e)
print('torch_e is:{}'.format(torch_e))
运行结果为:
e is:[[2 3]
[4 5]]
torch_e is:[[2 3]
[4 5]]
如果需要改变tensor的数据类型,只需要在转换后的tensor后面加上所需的数据类型即可。
f_torche=torch_e.float()
print('f_torche is:{}'.format(f_torche))
结果为:
f_torche is:
2 3
4 5
[torch.FloatTensor of size 2x2]
如果需要将Tensor放到GPU上,只需要a.cuda()就能将tensor a放到GPU上了。
if torch.cuda.is_available(): #判断是否支持GPU
a_cuda=a.cuda()
print(a_cuda)
3.1.2:Variable(变量)
Variable(变量)提供了自动求导的功能。
将一个tensor a 变成Variable(变量),只需要Variable(a)就可以了。
Variable有三个比较重要的属性:data, grad 和 grad_fn.
通过data可以取出Variable里面的tensor数值,grad_fn表示的是得到这个Variable的操作,比如通过加减还是乘除来得到的, 最后grad就是这个Variable的反向传播梯度。
1)标量求导
注意:
1.构建变量时,要注意参数requires_grad=True, 这个参数表示是否对这个变量求梯度,默认的是false——不求梯度。
2.y.backward()
,这一行的代码就是所谓的自动求导,这个函数其实等价于y.backward(torch.FloatTensor([1]))
, 只不过对于标量求导里面的参数可以不写。
自动求导不需要明确地写出那个函数对哪个函数求导,直接通过这行代码就可以对所有的需要梯度的变量进行求导,得到他们的梯度,然后通过x.grad就可以得到x 的梯度。
import torch
from torch.autograd import Variable
#创建变量
x=Variable(torch.Tensor([1]), requires_grad=True)
w=Variable(torch.Tensor([2]), requires_grad=True)
b=Variable(torch.Tensor([3]), requires_grad=True)
#建立计算图
y=w*x+b #y=2*x+3
#计算梯度
y.backward()
#打印梯度
print(x.grad) #x.grad=2
print(w.grad) #w.grad=1
print(b.grad) #b.grad=1
结果为:
Variable containing:
2
[torch.FloatTensor of size 1]
Variable containing:
1
[torch.FloatTensor of size 1]
Variable containing:
1
[torch.FloatTensor of size 1]
2)矩阵求导
x=torch.randn(3)
print('x is :{}'.format(x))
x=Variable(x,requires_grad=True)
y=x*2
print(y)
y.backward(torch.FloatTensor([1,0.1,0.01]))
print(x.grad)
相当于给出了一个三维向量去做运算,这时候得到的结果
就是一个向量。对这个向量求导就不能写成y.backward()
, 这样程序会报错。
这个时候需要传入参数生命,比如y.backward(torch.FloatTensor([1,1,1]))
,这样得到的结果就是他们每个分量的梯度;或者可以传入y.backward(torch.FloatTensor([1,0.1,0.01]))
,这样得到的梯度就是他们原本的梯度分别乘上1, 0.1, 0.01.
Dataset(数据集)
3.2.3 多维线性回归
3.2.4 一维线性回归
文章链接:10分钟快速入门 PyTorch (1) - 线性回归
训练数据
散点图用matplotlib画出来如图:
(matplotlib用法链接:python学习之matplotlib绘制散点图实例)
#import torch
import numpy as np
import matplotlib.pyplot as plt
#from torch.autograd import Variable
x_train=np.array([[3.3],[4.4],[5.5],[6.71],[6.93],[4.168],[9.779],
[6.182],[7.59],[2.167],[7.042],[10.791],[5.313],[7.997],[3.1]],dtype=np.float32)
y_train=np.array([[1.7],[2.76],[2.09],[3.19],[1.694],[1.573],[3.366],[2.596],
[2.53],[1.22],[2.827],[3.465],[1.65],[2.904],[1.3]],dtype=np.float)
#plt.plot(x_train,y_train)
plt.scatter(x_train, y_train, s=50)
# 设置图表标题并给坐标轴加上标签
plt.title('Numbers', fontsize=24)
plt.xlabel('x_Value', fontsize=14)
plt.ylabel('y_Value', fontsize=14)
# 设置刻度标记的大小
plt.tick_params(axis='both', which='major', labelsize=14)
# 设置每个坐标轴的取值范围
#函数axis()要求提供四个值:x、y坐标轴的最小值和最大值。[xmin,xmax,ymin,ymax]
plt.axis([0, 15, 0, 4])
plt.show()
我们想要做的事情就是找一条直线去逼近这些点,希望这条直线离这些点的距离之和最小。
1)先将numpy.array 转换成Tensor, 因为Pytorch里面的处理单元是Tensor.
x_train = torch.from_numpy(x_train)
y_train = torch.from_numpy(y_train)
2)接着需要建立模型:
class LinearRegression(nn.Module):
def __init__(self):
super(LinearRegression, self).__init__()
self.linear = nn.Linear(1, 1) # input and output is 1 dimension
def forward(self, x):
out = self.linear(x)
return out
if torch.cuda.is_available():
model = LinearRegression().cuda()
else:
model = LinearRegression()
3)然后需要定义loss和optimizer,就是误差和优化函数
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr=1e-4)
4)接着就可以训练模型了:
# 开始训练
num_epochs = 1000
for epoch in range(num_epochs):
inputs = Variable(x_train)
target = Variable(y_train)
# forward
out = model(inputs)
loss = criterion(out, target)
# backward
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (epoch+1) % 20 == 0:
print('Epoch[{}/{}], loss: {:.6f}'
.format(epoch+1, num_epochs, loss.data[0]))
5)做完训练之后可以预测一下结果:
model.eval()
predict = model(Variable(x_train))
predict = predict.data.numpy()
plt.plot(x_train.numpy(), y_train.numpy(), 'ro', label='Original data')
plt.plot(x_train.numpy(), predict, label='Fitting Line')
总的代码实现为:
'''
__author__ = 'SherlockLiao'
'''
#/home/np/BoostTree.py
import torch
from torch import nn, optim
from torch.autograd import Variable
import numpy as np
import matplotlib.pyplot as plt
x_train = np.array([[3.3], [4.4], [5.5], [6.71], [6.93], [4.168],
[9.779], [6.182], [7.59], [2.167], [7.042],
[10.791], [5.313], [7.997], [3.1]], dtype=np.float32)
y_train = np.array([[1.7], [2.76], [2.09], [3.19], [1.694], [1.573],
[3.366], [2.596], [2.53], [1.221], [2.827],
[3.465], [1.65], [2.904], [1.3]], dtype=np.float32)
x_train = torch.from_numpy(x_train)
y_train = torch.from_numpy(y_train)
# Linear Regression Model
class LinearRegression(nn.Module):
def __init__(self):
super(LinearRegression, self).__init__()
self.linear = nn.Linear(1, 1) # input and output is 1 dimension
def forward(self, x):
out = self.linear(x)
return out
if torch.cuda.is_available():
model = LinearRegression().cuda()
else:
model = LinearRegression()
# 定义loss和优化函数
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr=1e-4)
# 开始训练
num_epochs = 1000
for epoch in range(num_epochs):
inputs = Variable(x_train)
target = Variable(y_train)
# forward
out = model(inputs)
loss = criterion(out, target)
# backward
optimizer.zero_grad()
loss.backward()
optimizer.step()
if (epoch+1) % 20 == 0:
print('Epoch[{}/{}], loss: {:.6f}'
.format(epoch+1, num_epochs, loss.data[0]))
model.eval()
predict = model(Variable(x_train))
predict = predict.data.numpy()
plt.plot(x_train.numpy(), y_train.numpy(), 'ro', label='Original data')
plt.plot(x_train.numpy(), predict, label='Fitting Line')
# 显示图例
plt.legend()
plt.show()
# 保存模型
torch.save(model.state_dict(), './linear.pth')
3.2.5 多项式回归
链接:Pytorch 系列教程之一 使用Pytorch拟合多项式(多项式回归)
实现代码:
import torch
from torch.autograd import Variable
from torch import nn
from torch import optim
import matplotlib.pyplot as plt
import numpy as np
def make_features(x):
x = x.unsqueeze(1)
return torch.cat([x ** i for i in range(1,4)] , 1)
def f(x):
return x.mm(w_target)+b_target[0]
def get_batch(batch_size=32):
random = torch.randn(batch_size)
x = make_features(random)
'''Compute the actual results'''
y = f(x)
if torch.cuda.is_available():
return Variable(x).cuda(), Variable(y).cuda()
else:
return Variable(x), Variable(y)
class poly_model(nn.Module):
def __init__(self):
super(poly_model, self).__init__()
self.poly = nn.Linear(3,1)
def forward(self, x):
out = self.poly(x)
return out
w_target = torch.FloatTensor([0.5,3,2.4]).unsqueeze(1)
b_target = torch.FloatTensor([0.9])
if torch.cuda.is_available():
model = poly_model().cuda()
else:
model = poly_model()
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(), lr = 1e-3)
epoch = 0
while True:
batch_x,batch_y = get_batch()
output = model(batch_x)
loss = criterion(output,batch_y)
print_loss = loss.data[0]
optimizer.zero_grad()
loss.backward()
optimizer.step()
epoch+=1
if print_loss < 1e-3:
break
'''Generate some random numbers to see the results'''
x_test = np.linspace(-5,5,50).astype(np.float32)
y_test = 0.9+0.5*x_test+np.square(x_test)*3+np.power(x_test,3)*2.4
model.eval()
predict = model(make_features(Variable(torch.from_numpy(x_test))))
predict = predict.cpu()
predict = predict.data.numpy()
plt.figure()
plt.plot(x_test,y_test,'-r',label = 'Original Data')
plt.scatter(x_test, predict)
plt.legend()
plt.show()
逻辑回归
__author__ = 'SherlockLiao'
import torch
from torch import nn, optim
#import torch.nn.functional as F
from torch.autograd import Variable
from torch.utils.data import DataLoader
from torchvision import transforms
from torchvision import datasets
import time
# 定义超参数
batch_size = 32
learning_rate = 1e-3
num_epoches = 100
# 下载训练集 MNIST 手写数字训练集
train_dataset = datasets.MNIST(
root='./data', train=True, transform=transforms.ToTensor(), download=True)
test_dataset = datasets.MNIST(
root='./data', train=False, transform=transforms.ToTensor())
train_loader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False)
# 定义 Logistic Regression 模型
class Logstic_Regression(nn.Module):
def __init__(self, in_dim, n_class):
super(Logstic_Regression, self).__init__()
self.logstic = nn.Linear(in_dim, n_class)
def forward(self, x):
out = self.logstic(x)
return out
model = Logstic_Regression(28 * 28, 10) # 图片大小是28x28
use_gpu = torch.cuda.is_available() # 判断是否有GPU加速
if use_gpu:
model = model.cuda()
# 定义loss和optimizer
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=learning_rate)
# 开始训练
for epoch in f(num_epoches):
print('*' * 10)
print('epoch {}'.format(epoch + 1))
since = time.time()
running_loss = 0.0
running_acc = 0.0
for i, data in enumerate(train_loader, 1):
img, label = data
img = img.view(img.size(0), -1) # 将图片展开成 28x28
if use_gpu:
img = Variable(img).cuda()
label = Variable(label).cuda()
else:
img = Variable(img)
label = Variable(label)
# 向前传播
out = model(img)
loss = criterion(out, label)
running_loss += loss.data[0] * label.size(0)
_, pred = torch.max(out, 1)
num_correct = (pred == label).sum()
running_acc += num_correct.data[0]
# 向后传播
optimizer.zero_grad()
loss.backward()
optimizer.step()
if i % 300 == 0:
print('[{}/{}] Loss: {:.6f}, Acc: {:.6f}'.format(
epoch + 1, num_epoches, running_loss / (batch_size * i),
running_acc / (batch_size * i)))
print('Finish {} epoch, Loss: {:.6f}, Acc: {:.6f}'.format(
epoch + 1, running_loss / (len(train_dataset)), running_acc / (len(
train_dataset))))
model.eval()
eval_loss = 0.
eval_acc = 0.
for data in test_loader:
img, label = data
img = img.view(img.size(0), -1)
if use_gpu:
img = Variable(img, volatile=True).cuda()
label = Variable(label, volatile=True).cuda()
else:
img = Variable(img, volatile=True)
label = Variable(label, volatile=True)
out = model(img)
loss = criterion(out, label)
eval_loss += loss.data[0] * label.size(0)
_, pred = torch.max(out, 1)
num_correct = (pred == label).sum()
eval_acc += num_correct.data[0]
print('Test Loss: {:.6f}, Acc: {:.6f}'.format(eval_loss / (len(
test_dataset)), eval_acc / (len(test_dataset))))
print('Time:{:.1f} s'.format(time.time() - since))
print()
# 保存模型
torch.save(model.state_dict(), './logstic.pth')