This message was deleted.

I can only see 2 possibilities here: 1. Copy the encoder pickle/model file to your current machine and save as

custom_objects

. 2. Create an API endpoint to perform this encoding and call this API endpoint during inference.