resnet = resnet50(pretrained=False)
resnet.load_state_dict(torch.load(args.predir))
 
res_conv31 = Bottleneck_dilated(1024, 256,dilated_rate = 2)
print("---------------------",res_conv31)
print("---------------------",resnet.layer3[1])
 
res_conv31.load_state_dict(resnet.layer3[1].state_dict())

model_dict = model.state_dict()
# print(model_dict)
pretrained_dict = torch.load("/yzc/reid_testpcb/se_resnet50-ce0d4300.pth")
keys = []
for k, v in pretrained_dict.items():
       keys.append(k)
i = 0
for k, v in model_dict.items():
    if v.size() == pretrained_dict[keys[i]].size():
         model_dict[k] = pretrained_dict[keys[i]]
         #print(model_dict[k])
         i = i + 1
model.load_state_dict(model_dict)

from collections import OrderedDict
pretrained_dict = torch.load('premodel')
 
new_state_dict = OrderedDict()
 
# for k, v in mgn_state_dict.items():
#     name = k[7:]  # remove `module.`
#     new_state_dict[name] = v
# self.model = self.model.load_state_dict(new_state_dict)
 
for k, v in pretrained_dict.items():
    name = "model.module."+k   # remove `module.`
    # print(name)
    new_state_dict[name] = v
self.model.load_state_dict(new_state_dict)

#注意：需要重新訓練的層的名字要和之前的不同。
model=resnet()#自己構建的模型，以resnet為例
model_dict = model.state_dict()
pretrained_dict = torch.load('xxx.pkl')
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}
model_dict.update(pretrained_dict)
model.load_state_dict(model_dict)

#k是可訓練參數(shù)的名字，v是包含可訓練參數(shù)的一個實體
#可以先print（k）,找到自己想進行調整的層，并將該層的名字加入到if語句中：
for k,v in model.named_parameters():
    if k!='xxx.weight' and k!='xxx.bias' :
        v.requires_grad=False#固定參數(shù)

#將要訓練的參數(shù)放入優(yōu)化器
optimizer2=torch.optim.Adam(params=[model.xxx.weight，model.xxx.bias],lr=learning_rate,betas=(0.9,0.999),weight_decay=1e-5)

for k,v in model.named_parameters():
   if k!='xxx.weight' and k!='xxx.bias' :
   print(v.requires_grad)#理想狀態(tài)下，所有值都是False