Wang Liu
Results
3
issues of
Wang Liu
您好！非常感谢您的工作！我有一个关于基准测试的问题，请问一下下面的结果是在验证集上计算得到的吗？还是说需要将测试集提交到测试服务器？ 
Nice work! In the UDA setting, the labels of the target domain cannot be used to choose hyper-parameters. Why use the target val datasets to evaluate the trained model？
line 142 in test.py: "pred = pred.cpu().numpy().squeeze().argmax(0)" should change to "pred = pred.squeeze().argmax(0).cpu().numpy()".