README.md

Introduction

This set of examples uses Ray in a variety of Metaflow workflows to train or tune many models in parallel, evaluate them, and serve with Ray Serve and Fast API.

Training

Single Node

python train.py --environment=conda run

Multinode

python train_parallel.py --environment=conda run

Tuning

Single Node

python tune.py --environment=conda run

Multinode

python tune_parallel.py --environment=conda run

Scoring

python score.py --environment=conda

Serving

serve run server:batch_preds