pytorch加载预训练模型或者自己的模型，并修改最后的分类输出层strict=False

参数是一个字典，所以可以按照字典的格式
去除键值对

        if args.pretrained:
            print("=> using pre-trained model '{}'".format(args.arch))
            model = models.__dict__[args.arch](pretrained=True)
            num_fc_ftr = model.fc.in_features ##重写分类输出层
            model.fc = torch.nn.Linear(num_fc_ftr, num_classes)
            ###############加载自己训练的模型
            print ("加载res34自己训练的模型")
            pretrained_dict = torch.load("day_label_smooth_resnet34_model_best.pth.tar")["state_dict"]
            
            for k,v in model.state_dict().items():
                print (k)
                print (v.shape)
            pretrained_dict.pop('fc.weight') # 加载的参数直接删除全连接层的参数，打印名字看看
            pretrained_dict.pop('fc.bias')
            model.load_state_dict(pretrained_dict, strict=False)

修改 模型名字，按照名字匹配

     model = DPN(num_init_features=64, k_R=96, G=32, k_sec=(3,4,20,3), inc_sec=(16,32,24,128), num_classes=1,decoder=args.decoder)
     http = {'url': 'http://data.lip6.fr/cadene/pretrainedmodels/dpn92_extra-b040e4a9b.pth'}
     pretrained_dict=model_zoo.load_url(http['url'])
     model_dict = model.state_dict()
     pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}#filter out unnecessary keys 
     model_dict.update(pretrained_dict)
     model.load_state_dict(model_dict)
     model = torch.nn.DataParallel(model).cuda()