search

pesser / edflow

Reduce boilerplate code for your ML projects. TensorFlow and PyTorch.
https://edflow.readthedocs.io/
MIT License
24 stars 13 forks source link

eval mode does not produce any outputs after completed training #232

Closed jhaux closed 4 years ago

jhaux commented 4 years ago

Describe the bug Calling edflow -p <project> -b <project>/configs/<train_config> Produces no outputs without throwing an error as the condition Iterator.get_global_step() >= config['num_steps'] is met.

To Reproduce Run an experiment as described above

Expected behavior During evaluation mode, this condition should be never met as we are assuming that there will only be one epoch and we will iterate over the entire data of that epoch.

Desktop (please complete the following information): Experiments we run on the IWR machines with current pesser/master code.

pesser commented 4 years ago

should already be fixed on dev branch. Can you confirm? Do you think we should fix it seperately on master or just try to merge dev as soon as possible?

jhaux commented 4 years ago

I cannot yet confirm, but in my opinion we should try to merge into master as soon as possible. The reason being, that the tests would make sense again.

pesser commented 4 years ago

definitely fixed in #250