Hi everyone,
I'm using IBM Watson Machine Learning to train TensorFlow models that include custom layers. When I attempt to export these models for deployment, the process fails with errors related to serialization of the custom components. I need to understand why Watson ML cannot handle these layers during export and whether there is a supported way to include custom TensorFlow layers in a model so that it can be successfully exported and deployed. I'm specifically looking for technical details about the serialization process or API limitations.
Regards from Berlin
Antonio B. Flowers - Germany
CTO Münsterheim
------------------------------
Antonio Bernal
------------------------------