pytorch编写代码基本步骤(4步)

https://mp.weixin.qq.com/s/yyyCKd9W86H5MtpkdEmXPw
1 Prepare dataset
2 Design model using Class

• inherit from nn.Module 计算y^hat

3 Construct loss and optimizer

• using PyTorch API 计算loss和optimizer

4 Training cycle

• forward, backward, update 前馈，反馈，更新

1 Prepare dataset

图像数据

点云数据

2 Design model using Class

• inherit from nn.Module 计算y^hat

  model

  class xxxModel(torch.nn.Module): class类里面至少实现两个函数（backword内部实现，不用写）

• def init(self，传递参数) # 初始化对象的时候默认调用的函数（构造函数）

• def forward(slf, x) # 前馈的时候所要执行的计算（backword反向自动根据计算图实现，不用写）

"""
our model class should be inherit from nn.Module, which is base class for all neural network modules.
member methods __init__() and forward() have to be implemented
class nn.linear contain two member Tensors: weight and bias
class nn.Linear has implemented the magic method __call__(),which enable the instance of the class can
be called just like a function.Normally the forward() will be called 
"""
class LinearModel(torch.nn.Module):  # 继承torch.nn.Module， torch.nn.Module是父类  
    def __init__(self):  # 构造函数(初始化)
        super(LinearModel, self).__init__() # 调用父类的init     super(LinearModel, self)  == torch.nn.Module
        # (1,1)是指输入x和输出y的特征维度，这里数据集中的x和y的特征都是1维的
        # 该线性层需要学习的参数是w和b  获取w/b的方式分别是~linear.weight/linear.bias
        self.linear = torch.nn.Linear(1, 1) # 构造对象，执行wx+b的操作，Linear也是继承自Model，并说明输入输出的维数，第三个参数默认为true，表示用到b
    def forward(self, x):
        y_pred = self.linear(x) #可调用对象（对象括号里面加参数x），计算y=wx+b 
        return y_pred
model = LinearModel() # 自动构建计算图，实现反向（自带的，不用写）

3 Construct loss and optimizer 损失&优化器

• using PyTorch API 计算loss和optimizer

4 Training cycle 训练epoch

• forward, backward, update 前馈，反馈，更新

    for epoch in range(epochs): # 100============================================
      acc_loss = 0.0
      num_samples = 0
      start_tic = time.time()
      for x, y in train_loader:      #  Iteration=样本数/batch_size 遍历数据(每次train_loader不一样？)=
        optimizer.zero_grad() # 0 set grad to zero
        out = model(x) # 1 模型输入,求解y
        loss = softXEnt(out, y) # 2 计算损失
        loss.backward() # 3 loss backward
        optimizer.step() # 4 update network's param

5 Test 测试