sktmls 2020.12.16rc42

MLS SDK

mls-model-registry (sktmls)

Contents

• Description
• How to use
• Development
  • Requirements for development
  • Local model registry
  • Python environment
  • Documents generation
• Version

Description

A Python package for MLS model registry.

This python package includes

• Customized prediction pipelines inheriting MLSModel
• Model uploader to AWS S3 for meta management and online prediction

Installation

Installation is automatically done by training containers in YE. If you want to install manually for local machines,

# develop
pip install --index-url https://test.pypi.org/simple/ --no-deps sktmls

# production
pip install sktmls

How to use

• MLS Docs: https://ab.sktmls.com/docs/model-registry
• sktmls Docs: https://sktaiflow.github.io/mls-sdk/sktmls

Development

Requirements for development

• Python 3.6
• requirements.txt
• requirements-dev.txt

Local model registry

To enable all model related features in local environment, you need to create a directory models in your home directory.

$ cd ~/
$ mkdir models

Python environment

First you need to do the followings

$ python -V # Check if the version is 3.6.
$ python -m venv env # Create a virtualenv.
$ . env/bin/activate # Activate the env.
$ pip install "numpy>=1.19.4,<1.20" # Install numpy to avoid a requirement error.
$ pip install -r requirements.txt # Install required packages.
$ pip install -r requirements-dev.txt # Install required dev packages.

Documents generation

Before a commit, generate documents if any docstring has been changed

rm -rf docs
pdoc --html --config show_source_code=False -f -o ./docs sktmls

Version

sktmls package version is automatically genereated followd by a production release on format YY.MM.DD
We use Calendar Versioning. For version available, see the tags on this repository.