There is a known bug in the earlier versions of pytorch including the version used in this project. As of now, trying to train the model yourself may result in the following error:
Traceback (most recent call last):
File "train.py", line 11, in <module>
from torch.utils.tensorboard import SummaryWriter
File "/Users/smmousavi/.pyenv/versions/miniconda3-4.7.10/lib/python3.7/site-packages/torch/utils/tensorboard/__init__.py", line 4, in <module>
LooseVersion = distutils.version.LooseVersion
AttributeError: module 'distutils' has no attribute 'version'
To use version 1.10.0 of pytorch required by this project, downgrading setuptools to version 59.5.0 will solve this problem. The version of setuptools in environment.yamlfile is below this version but if someone wants to install the dependencies on their own, they need to manually downgrade the setuptools. For users not on linux and without CUDA the environment can not be resolved and it is needed to manually install the dependencies.
I added an explanation how to fix this bug in the README.md file.
There is a known bug in the earlier versions of
pytorch
including the version used in this project. As of now, trying to train the model yourself may result in the following error:This bug has been fixed in newer versions: https://github.com/pytorch/pytorch/pull/69904
To use version
1.10.0
ofpytorch
required by this project, downgradingsetuptools
to version59.5.0
will solve this problem. The version ofsetuptools
inenvironment.yaml
file is below this version but if someone wants to install the dependencies on their own, they need to manually downgrade thesetuptools
. For users not on linux and withoutCUDA
the environment can not be resolved and it is needed to manually install the dependencies.I added an explanation how to fix this bug in the
README.md
file.