pytorch中对模型结构进行的常用操作，如冻结参数添加改变层等

torch 笔记

一、模型操作

添加层

b = torchvision.models.alexnet().features
d = torchvision.models.alexnet().avgpool
for i in d.named_modules():
    b.add_module('12',i[1])#12为位置索引

改变层结构

b[1] = Linear(2048, 10)

冻结卷积层参数

for i in range(10):#具体层数要看模型结构
  if isintance(b[i], Conv2d):
    b[i].weight.requires_grad = False
    b[i].bias.requires_grad = False
    #b[i].weight.requires_grad_(False)

summary 计算网络参数个数

from torchsummary import summary
print(summary(conv, (16, 64, 64), batch_size=1))
#网络，输入，batch

本文地址：https://blog.csdn.net/FLY__1997/article/details/107658240