柳杨

实际中，经常会出现需要使用之前训练好的模型，去掉后面若干层，然后在模型最后增加一个针对本次任务的 head 的情况。 上述操作，尤其是**去掉后面若干层**，似乎在 pytorch 中比较麻烦。因为pytorch 好像不能取其前向反馈(forward)中的部分操作。 假设定义一个网络（卷积参数瞎写的）： ``` class Test(nn.Module): def __init__(self): super(Test, self).__init__() self.a = nn.Conv2d(2, 3, 3) self.b = nn.Conv2d(3, 4, 3) def forward(self, x): # x...

### Please confirm if feature request does NOT exists already ? - [X] I confirm there is no existing issue for this ### Describe the usecase for the feature I...

### Please confirm if feature request does NOT exists already ? - [X] I confirm there is no existing issue for this ### Describe the usecase for the feature When...

I cannot find the related content in doc or example. and how to continune code interpret chat? Do i have to upload the file again?