I'm trying to migrate code from Tensorflow 1.13 to Tensorflow 2.0 and I'm having issues with the session declaration step.
For now, I replaced tf.keras.backend.set_session by tf.compat.v1.keras.backend.set_session.
config = tf.compat.v1.ConfigProto()
config.gpu_options.allow_growth = True
if self._gpu_fraction:
config.gpu_options.per_process_gpu_memory_fraction = self._gpu_fraction
sess = tf.compat.v1.Session(config=config)
tf.compat.v1.keras.backend.set_session(sess)
The code is running but I get a warning:
The name tf.keras.backend.set_session is deprecated. Please use tf.compat.v1.keras.backend.set_session instead.
I understood that sessions evolved a lot with Tensorflow 2 but I'm not sure if I can/should get rid of this set_session.
I'm also not sure about the warning: is tf.compat.v1.keras.backend.set_session calling the old tf.keras.backend.set_session?
Thank you for any guidance you can provide!