yamlflow

Yet Another ML flow

We follow convention over configuration (also known as coding by convention) software design paradigm.

Here are some of the features the yamlflow provides.

1. Build and publish your ML solution as a RESTful Web Service with yaml.

   • You don't need to write web realated code, or dockerfiles.

   • You don't need to benchmark which python web server or framework is best in terms of performance.

   • WE do it for you. All the best, packed in.

Metrics for inference

• Throughput (How many requests can server process in some interval of time)

• Latency (How long does it take to get a prediction for a single request)

Python REST API

async
    - web-server(ASGI): Uvicorn, Hypercorn, Daphne
    - web-(micro)framework: Starlete, ...
    - API-framework: FastAPI, ...

sync
    - web-server(WSGI): gunicorn, uWSGI, Gevent, Twisted Web
    - web-(micro)framework: Flask
    - API-framework: None

Project structure

.
├── model
│   ├── ...
│   ├── pipeline.py
│   └── requirements.txt
├── service
│   ├── objects
│   ├── config.py
│   └── predictor.py
├── train
│   ├── ...
│   ├── train.py
│   └── requirements.txt
├── README.md
└── yamlflow.yaml

example yamlflow.yaml

kind: Service
meta:
  registry: your.docker.registry
  user: dockerusername
project:
  name: ml-project
  version: 0.1.0
backend:
  runtime: torch
  device: cpu
port: 8002

Installation guide

pip install yamlflow

User guide

yamlflow init
yamlflow build
yamlflow run

Developer guide

pyenv install 3.8.6
poetry env use ~/.pyenv/versions/3.8.6/bin/python
poetry shell
poetry install