import torch
import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self, **kwargs):
        super(Model, self).__init__(**kwargs)
    //可以在这你定义你的链接层
        pass
    def reset_parameters(self):
    //在这里初始化权重
        pass
    def forward(self, inputs):
    //在这里定义前向传播
        pass

推理

model = Model()

model(inputs) # 或model.forward(inputs)

优化器和损失函数定义

optimize=torch.optim.SGD(model.parameters(),lr=0.1)
loss_fn=nn.MSELoss()

训练

for epoch in range(10):
    //清空梯度
    optimize.zero_grad()
    //前向传播
    pred=model(features)
    //损失值
    loss=loss_fn(pred,labels)
    //后向传播
    loss.backward()
    //优化器更新参数
    optimize.step()
    //打印损失
    print(loss)

实例

产生数据集

%matplotlib inline
import random
import torch

def synthetic_data(w, b, num_examples):  #@save
    """生成y=Xw+b+噪声"""
    X = torch.normal(0, 1, (num_examples, len(w)))
    y = torch.matmul(X, w) + b
    y += torch.normal(0, 0.01, y.shape)
    return X, y.reshape((-1, 1))

true_w = torch.tensor([2, -3.4])
true_b = 4.2
features, labels = synthetic_data(true_w, true_b, 1000)

训练网络

import torch
import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self, **kwargs):
        super(Model, self).__init__(**kwargs)
        self.dense=nn.Linear(2,1)
    def forward(self, inputs):
        outputs=self.dense(inputs)
        return outputs

model=Model()

optimize=torch.optim.SGD(model.parameters(),lr=0.1)
loss_fn=nn.MSELoss()

for epoch in range(10):
    optimize.zero_grad()
    pred=model(features)
    loss=loss_fn(pred,labels)
    loss.backward()
    optimize.step()
    print(loss)