目录
初识pytorch,本文基于pytorch构建最基本的神经网络，实现线性回归模型。
(1)构造一组输入数据X和其对应的标签y
x_values = [i for i in range(11)]
x_train = np.array(x_values, dtype=np.float32)#np.array格式
x_train = x_train.reshape(-1, 1)#把数据转换成矩阵，防止出错
y_values = [2*i + 1 for i in x_values]#定义回归方程y=2x+1
y_train = np.array(y_values, dtype=np.float32)#np.array格式
y_train = y_train.reshape(-1, 1)
(2)构建模型
class LinearRegressionModel(nn.Module):#定义类，import torch.nn as nn，nn.Module：只用写用哪个层
def __init__(self, input_dim, output_dim):#构造函数，写用到哪些层
super(LinearRegressionModel, self).__init__()
self.linear = nn.Linear(input_dim, output_dim) #nn.Linear()全连接层，传入输入维度，输出维度
def forward(self, x):#前向传播函数，用到的层如何使用的
out = self.linear(x)#在全连接层中输入x得到结果
return out
(3)指定好参数和损失函数
input_dim = 1
output_dim = 1
model = LinearRegressionModel(input_dim, output_dim)#定义模型
epochs = 1000#一共迭代1000次
learning_rate = 0.01#学习率
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)#优化器，SGD，传入需要优化的参数和学习率
criterion = nn.MSELoss()#损失函数，分类任务常用交叉熵损失，回归任务常用MSE均方差损失函数
(4)训练模型
for epoch in range(epochs):#遍历1000次
epoch += 1
# np.array格式不能直接进行训练，转换成tensor格式
inputs = torch.from_numpy(x_train)
labels = torch.from_numpy(y_train)
# 梯度要清零每一次迭代
optimizer.zero_grad() #如果不清空梯度会累加起来
# 前向传播
outputs = model(inputs)
# 计算损失
loss = criterion(outputs, labels)
# 返向传播
loss.backward()
# 更新权重参数
optimizer.step()
if epoch % 50 == 0:
print('epoch {}, loss {}'.format(epoch, loss.item()))
结果：
(5)测试模型预测结果
predicted = model(torch.from_numpy(x_train).requires_grad_()).data.numpy()#进行一次前向传播
结果：
(6)模型的保存与读取
torch.save(model.state_dict(), 'model.pkl')#model.state_dict()：模型的权重参数
model.load_state_dict(torch.load('model.pkl'))#读取
(7)使用GPU进行训练
只需要把数据和模型传入到cuda里面就可以了，与CPU训练代码有两点不同，已加注释说明
import torch
import torch.nn as nn
import numpy as np
class LinearRegressionModel(nn.Module):
def __init__(self, input_dim, output_dim):
super(LinearRegressionModel, self).__init__()
self.linear = nn.Linear(input_dim, output_dim)
def forward(self, x):
out = self.linear(x)
return out
input_dim = 1
output_dim = 1
model = LinearRegressionModel(input_dim, output_dim)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")#指定设备cuda,如果cuda配置好的用cuda否则用CPU
model.to(device)#把模型放入cuda
criterion = nn.MSELoss()
learning_rate = 0.01
optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate)
epochs = 1000
for epoch in range(epochs):
epoch += 1
inputs = torch.from_numpy(x_train).to(device)#把输入x传入cuda
labels = torch.from_numpy(y_train).to(device)#把输入y传入cuda
optimizer.zero_grad()
outputs = model(inputs)
loss = criterion(outputs, labels)
loss.backward()
optimizer.step()
if epoch % 50 == 0:
print('epoch {}, loss {}'.format(epoch, loss.item()))