Resuming experiment results in incorrect test logs

MichiganCOG / ViP

Video Platform for Action Recognition and Object Detection in Pytorch

MIT License

219 stars 37 forks source link

Resuming experiment results in incorrect test logs #20

Closed lemmersj closed 5 years ago

lemmersj commented 5 years ago

I've been seeing this phenomenon in my experiments, when restarting from already loaded weights. Could be related to #19? Note that logging is fine for training.

Attached are the relevant config files (in a tarball).

jump

configs.tar.gz

ehofesmann commented 5 years ago

The issue here was that you switched load_type from train to train_val. It was originally validating on your training set and then when you loaded it started validating on your validation set. Since the sizes are different for len(val_loader) it plotted at a different point. We are going to update this so that all plots use the same x axis #24