YangXuhui

LoadModel.py ``` if self.use_dcl: if config.cls_2: self.classifier_swap = nn.Linear(2048, 2, bias=False) if config.cls_2xmul: self.classifier_swap = nn.Linear(2048, 2*self.num_classes, bias=False) self.Convmask = nn.Conv2d(2048, 1, 1, stride=1, padding=0, bias=True) self.avgpool2 = nn.AvgPool2d(2, stride=2)...

请问问题解决了吗，我也遇到这个问题，stanford cars的测试效果上不去，不管是否使用该数据集提供的bbox标签

I just jumped to the function definition of dist.init() in tools/train.py and changed master_host: ``` # master_host = 'tcp://' + os.environ['MASTER_HOST'] master_host = None ``` , then it is ok...

> I just jumped to the function definition of dist.init() and changed master_host: > > ``` > # master_host = 'tcp://' + os.environ['MASTER_HOST'] > master_host = None > ``` >...

> I found that the camera extrinsics generated by CVT are different from those generated by Fiery, I'm confused, too. Do you know why now? Maybe it is the inverse...

Have you checked the MPI installation?

> > I just jumped to the function definition of dist.init() in tools/train.py and changed master_host: > > ``` > > # master_host = 'tcp://' + os.environ['MASTER_HOST'] > > master_host...

> I also have this problem. Have you solved it? > > I've trained 20K steps. May be it is not enough? Not solved. Have you tried the new code?