Closed songquanpeng closed 4 years ago
Looks like it helps if you set num_workders=0
when you call the DataLoader function. So I think it's a problem caused by multiprocessing.
def __iter__(self):
if self.num_workers == 0:
return _SingleProcessDataLoaderIter(self)
else:
return _MultiProcessingDataLoaderIter(self)
May I ask which operating system do you use in this experiment?
Our experiments are running on Ubuntu 16.04.5 LTS (GNU/Linux 4.4.0-87-generic x86_64). Based on your description, I am more inclined that this is an operating system related issue rather than package issue. And the multiprocessing seems not a package but a file in pytorch. The requirement file is here. Hope it can help you.
Thanks for your help. Looks like multiprocessing is a standard package of Python. And my env name is pytorch which may confuse you.
Hello, when I try to run your code with the following argument:
--dataset miniimagenet --mode train
, a strange error occurred:It looks like something wrong with the data load part. So I want to make sure all the packages such as
multiprocessing
have the proper version to help locate the error.Would you please offer a requirements.txt exported by
pip freeze > requirements.txt
? Thanks in advance.