RFC: TFX Tuner Component

[1025KB]

This RFC will be open for comment until Friday, May 8, 2020.

TFX Tuner Component

Status	Proposed
RFC #	236
Author(s)	Jiayi Zhao ([email protected]), Amy Wu ([email protected])
Sponsor	Zhitao Li ([email protected]), Tom O'Malley ([email protected]), Matthieu Monsch ([email protected]), Makoto Uchida ([email protected]), Goutham Bhat ([email protected])
Updated	2020-04-20

Objective

Goal

• A new Tuner component in TFX for automated hyper-parameter tuning, which is based on abstractions from KerasTuner library, in order to reuse abstractions and algorithms from latter.

Non Goal

• Natively support multi-worker tuning by the system. As TFX doesn't have ability to manage multi-worker clusters, running multiple trials in parallel (parallel tuning) and running each trial in distributed env (distributed training) are not supported natively. Parallel tuning may instead be realized by a particular implementation of TFX Tuner (custom Executor), e.g., in Google Cloud environment.
• Implementation of custom tuner for KerasTuner library is out of scope of this design discussion, e.g., a built-in EstimatorTuner support. However, user project can still implement a tuner that inherits from kerastuner.BaseTuner and provide it to the proposed TFX Tuner component.

[ematejska]

The review happened today. There was no extra feedback and the Design is accepted as is.