1、pytorch入门

介绍

Pytorch是一款facebook发布的深度学生框架，由其易用性，友好性，深受广大用户青睐

安装

张量Tensor

创建张量

张量的属性

tensor常见数据类型

常见方法

CUDA Tensor

常用数学运算

自动求导

前向计算

梯度计算

案例：Pytorch完成线性回归

import torch
import numpy as np
import matplotlib.pyplot as plt

x = torch.rand([50])
y = 3*x+0.8

w = torch.rand(1,requires_grad=True)
b = torch.rand(1,requires_grad=True)

def loss_fn(y,y_predict):
    loss = (y_predict - y).pow(2).mean()
    for i in [w,b]:
        if i.grad is not None:
            i.grad.data.zero_()
    loss.backward()
    return loss.data

def optimize(lr):
    w.data -= lr * w.grad.data
    b.data -= lr * b.grad.data

for i in range(3000):
    y_predict = x*w+b
    loss = loss_fn(y,y_predict)
    if i%200 == 0:
        print(i,loss)
    optimize(0.01)
predict = x*w+b
plt.scatter(x.data.numpy(),y.data.numpy(),c='r')
plt.plot(x.data.numpy(),predict.data.numpy())
plt.show()

Pytorch基础模型组件

nn.Module

nn.Sequential

优化器

损失函数

案例：线性回归

import torch
import torch.optim as optim
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt

x = torch.rand([50,1])
y = 3*x+0.8

class Lr(nn.Module):
    def __init__(self):
        super(Lr,self).__init__()
        self.linear = nn.Linear(1,1)
        
    def forward(self,x):
        out = self.linear(x)
        return out

model = Lr()
criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(),lr=1e-3)

for i in range(30000):
    out = model(x)
    loss = criterion(y,out)
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
    if (i+1)%1000 == 0:
        print("Epch{}/{},Loss{:.6f}".format(i,5000,loss.data))
model.eval()
predict = model(x)
output = predict.data.numpy()
plt.scatter(x.data.numpy(),y.data.numpy(),c='r')
plt.plot(x.data.numpy(),output)
plt.show()

GPU运行代码

案例修改

import torch
import torch.optim as optim
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

x = torch.rand([50,1])
y = 3*x+0.8

class Lr(nn.Module):
    def __init__(self):
        super(Lr,self).__init__()
        self.linear = nn.Linear(1,1)
        
    def forward(self,x):
        out = self.linear(x)
        return out

x = x.to(device)
y = y.to(device)

model = Lr().to(device)

criterion = nn.MSELoss()
optimizer = optim.SGD(model.parameters(),lr=1e-3)

for i in range(30000):
    out = model(x)
    loss = criterion(y,out)
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()
    if (i+1)%1000 == 0:
        print("Epch{}/{},Loss{:.6f}".format(i,5000,loss.data))
model.eval()
predict = model(x)
output = predict.cpu().detach().numpy()

plt.scatter(x.cpu().data.numpy(),y.cpu().data.numpy(),c='r')

plt.plot(x.cpu().data.numpy(),output)

plt.show()

常见优化算法

避免过拟合/加速训练