ResNet选择冻结任意层不进行参数更新

mymodel=models.resnet50(pretrained=True)
#print(mymodel)

for name, value in mymodel.named_parameters():
    #print(name)
    #print(value.requires_grad)
    #print(value)
    if name == 'conv1.weight':
        #confient
        value.requires_grad= False
    print(name)
    print(value.requires_grad)
    #print(value)
    #print(value)

这里只让conv1.weight这一层冻结了，咋一层在新的参数训练中不参与梯度下降并不再更新本层参数（个人理解），