layer names

[jackd]

Style request: can layer names conform to tensorflow conventions? Specifically:

• no appended Layer or KerasLayer to tf.keras.layers.Layer implementations - it should be obvious from the module (e.g. tf.nn.layer) and convention.
• no naming of functions finishing in layer if they aren't returning / using a tf.keras.layers.Layer.

Happy to put in a PR with these changes (non-breaking with deprecated warnings on existing functions) if it's a desired direction.

I'd also love the layer submodule renamed layers, but that's probably a bit much to ask at this point (though if that's desired by others I can do this as well).

[julienvalentin]

Hi @jackd !

Makes sense! Let's maybe start with one of these layers to see what the suggested change implies?

Best.

[taiya]

+1 to both

I think via some init.py magic you could get layer → layers (agreed it is better) change to be non-breaking. TF does a fair bit of init magic in this sense too.