Can't pickle local object 'MultiThreadedAugmenter._start.<locals>.producer about batchgenerators HOT 8 CLOSED

At some point we had another user having issues with anaconda. Sometimes anaconda installs funky combinations of python package version. One thing you could try is to use a regular python virtualenv and install everything with pip.

virtualenv -p python3 my_venv_name
source my_venv_name/bin/activate
pip3 install numpy==1.14.5 # and whatever else you need
cd my_batchgenerators_folder
pip3 install --upgrade .

from batchgenerators.

What I was trying to say is that sometimes conda instally funky versions of other python packages we depend on. If anything is installed by conda then you should try installing everything by pip instead

from batchgenerators.

My guess still is that conda botched up some libraries.
It's most likely not your dataset that causes the problem.

AttributeError: Can't pickle local object 'MultiThreadedAugmenter._start..producer'

That is an error arising from our code, not your dataset.

producer was so far defined within the MultiThreadedAugmenter. That worked well for all of us in the department. I just pushed a new commit where I moved it out of MultiThreadedAugmenter. Please pull, reinstall (use either pip install -e or pip install --upgrade to make sure changes are applied) and try again.

Best,
Fabian

from batchgenerators.

Hi,
I just pulled the most recent master and tried some of my scripts without errors. I cannot reproduce your problem. Please provide a minimalistic example that crashes, otherwise I am afraid I can't help :-/
Best,
Fabian

from batchgenerators.

Thanks for your suggestion. Since I installed the package via pip in conda (forgot to mention that) it should actually be the same.

I'll see if I can break it down to a minimal example, since I have lot's of internal dependencies there.

from batchgenerators.

Did that resolve your problem?

from batchgenerators.

Unfortunately not.
However, since this isn't directly related to this repo, I closed this issue.

There's probably a small issue that prevents my dataset from being pickled. Because of that, I'll just rewrite my dataset from scratch and try to fix it that way.

from batchgenerators.

That fixed it. Thanks a lot!

from batchgenerators.