mlspeclib 0.0.3

MLSpec helper library to making using metadata in ML workflows easier

ML Spec Library

The purpose of MLSpec Lib is to provide a library that can easily read and write schema-tized metadata related to the steps in an ML Pipeline.

The original set of schema was drawn from here - - as a set of illustrative data for initial specs.

If you had a sample pipeline that did the following:

• Read data
• Transformed the data in some way and saved it to a new directory
• Trained a model based on the data
• Wrote the results of the training to a new directory
• Converted the model to a training format
• Created a serving endpoint for the model

At each step, you would want to use first class python objects to drive the workflow, and, ideally, you would want to read/write metadata about what you were doing to have a permenant record.

This is what MLSpecLib is designed to do:

• Read the metadata in
• Provide a first class Python object for use
• Allow validation of the Python object according to a schema
• Write the object to disk in a readable YAML format

To see it in action, go to the Sample Notebook.

*** NOTE: The Sample Notebook is illustrative only! It does not actually execute any actual training and the schema are just for examples. ***

• Getting Started *

• Fork this repo to a directory.
• It is highly recommended you have a virtual environment set up - please see how to do that here: https://uoa-eresearch.github.io/eresearch-cookbook/recipe/2014/11/26/python-virtual-env/

# Go into the directory and install all packages
pip3 install -r requirements.txt

# Change into the notebook directory and install jupyter
pip3 install jupyter

# Start a notebook server
jupyter notebook

Now go to the URL that jupyter gives you, and run through the Sample Notebook. It should provide examples of how to read and write full objects from disk using native feeling Python.

TODO:

• Support better typing / flexibility in lists (including allowed)
• Real world examples of usage - more notebooks

---

Library structure and tools are derived from the work of Kenneth Reitz:

• Learn more <http://www.kennethreitz.org/essays/repository-structure-and-python>_.
• If you want to learn more about setup.py files, check out this repository <https://github.com/kennethreitz/setup.py>_.