Problem with using model se_resnext50

When I use code `pretrainedmodels.se_resnext50_32x4d(pretrained='imagenet')` to download model se_resnext50, error comes up :

> File "/home/xx/anaconda3/envs/torch37/lib/python3.7/site-packages/torch/serialization.py", line 764, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
_pickle.UnpicklingError: invalid load key, '<'.

But when I tried other models such as se_resnext101,resnet ,there's no error. What goes wrong?

Start by reproducing pretrainedmodels.se_resnext50_32x4d(pretrained='imagenet') and inspect the failure at torch/serialization.py during model loading. Compare its download or load path with se_resnext101 and resnet; done means se_resnext50 downloads and loads successfully without the reported UnpicklingError.

Written by the indexing model from the issue text.