search

kr-colab / diploSHIC

feature-based deep learning for the identification of selective sweeps
MIT License
49 stars 14 forks source link

Weights not written for trained model #23

Closed nspope closed 2 years ago

nspope commented 4 years ago

Hi, thanks for developing this intuitive and useful tool.

With python 3.69, tensorflow 1.14.0, keras 2.3.1: weights from the trained model are not saved. This is fixed by changing monitor = 'val_acc' to monitor = 'val_accuracy' at:

https://github.com/kern-lab/diploSHIC/blob/7d55417af438e321c374dadc2f7e58f56107443e/diploSHIC.py#L250

For early stopping to work, the same change needs to made at:

https://github.com/kern-lab/diploSHIC/blob/7d55417af438e321c374dadc2f7e58f56107443e/diploSHIC.py#L243

Perhaps due to a recent change in keras or documentation thereof? (See https://github.com/tensorflow/tensorflow/issues/33163)

Nate

andrewkern commented 4 years ago

sorry just seeing this issue? is this still affecting things?

andrewkern commented 2 years ago

closing this issue as it seems stale. feel free to reopen