Closed milani closed 6 years ago
keras floatx is configuration dependent, I think. But this one should always use float32.
Why should it always use float32?
@jhung0 Sorry I replied in the email and it appeared not in the correct place. I answered your question in the right thread.
The two commits try to fix #159 and also part of #158.
As described in the commit descriptions, I am not sure if the typecast was there on purpose and it made problems. Removing it does not seem to break the tests at least on my system.