Closed
Description
Description
Currently ONNX api and TensorFlow api containers are made up of different images with different python files.
pkg/workloads/tf_api/
pkg/workloads/onnx_serve/
pkg/workloads/predictor_serve/
Motivation
Refactoring them will reduce code duplicity.
Also revisit Python error wrapping
Additional Context
- ONNX API pod is just the flask container where is TF API pod consists of the flask container and a TensorFlow serving container
- The metadata injected into request handlers is different