heha102 comments

Results 5 comments of


                                            heha102

important code error: losses["loss"].detach() in train_util.py line 246

I met the same situation. Do you know how to solve it now?

AttributeError: ConsistentTeacher: 'NoneType' object has no attribute 'num_scores'

https://github.com/Adamdad/ConsistentTeacher/issues/15 this issue solved the problem

make install

The same problem.

assert len(indices) == len(self)

参考https://github.com/microsoft/SoftTeacher/issues/109#issuecomment-976197003 修改semi_sampler.py的 211行，将 sum(self.epoch_length) * self.samples_per_gpu，修改为sum(self.epoch_length)。这样是可以的。

set randomness= dict(seed=138011772,deterministic=True) in config file, RuntimeError in anchor_head.py

I tried what @ttt0666 suggested, but I'm get a new error. I referred to this https://www.neusncp.com/user/blog?id=766 ， added `os.environ['CUBLAS_WORKSPACE_CONFIG']=':4096:2'` at the top of tools/train.py, which successfully addressed the problem.