The train script is now TF2 compatible but should still run with >=1.14. Unfortunately the implementation of PeriodicConv2D throws a warning for <2.x which can be ignored, though. The old implementation is still in train_nn.py.
Also, I removed the exact version numbers again from environment.yml because this seems to have caused some problems (see #12). But I added the exact version numbers I used as a comment in the environment file.
The train script is now TF2 compatible but should still run with >=1.14. Unfortunately the implementation of
PeriodicConv2Dthrows a warning for <2.x which can be ignored, though. The old implementation is still intrain_nn.py.Also, I removed the exact version numbers again from
environment.ymlbecause this seems to have caused some problems (see #12). But I added the exact version numbers I used as a comment in the environment file.