2. Potential Pitfalls

This document covers things we have learned during development that might cause problems or hard to find bugs.

2.1. Privileged training argument in the call() method cannot be passed via kwargs

For layers that have different behavior at training and inference time (e.g., dropout or batch normalization layers), a boolean training argument can be exposed, see this section of the Keras documentation. If we want to pass this manually, we have to do so explicitly and not as part of a set of keyword arguments via **kwargs.

@Lars: Maybe you can add more details on what is going on behind the scenes.