/grubhub_test

Primary LanguageJupyter Notebook

Model Train and Save Test

Author: Brenton Mallen

This is a package to build a ML model and save it to disk.

Installation

pip install mallen_grubhub_test

or install manually by cloning the repo and running

python setup.py install

To Run

The CLI requires an integer argument called --region_id.

Example run call: train --region_id 1 (This will train a classifier on region 1)

For CLI help type python src/train.py -h in the package root directory

Run Tests

To run tests, run the following command in the package root directory: make test

Development

Either a Conda environment using conda env update in the root directory or by using virtualenv by first initializing the environment with virtualenv grubhub-venv, activating with grubhub-venv/bin/activate and installing the dependencies using pip install -r requirements.txt

Updating pypi package Upload to pypi can be done using the make package-test command. It will ask for credentials for a test pypi account.

Similarly, production can be updated using the make package-prod command.

For reference, see the tutorial

TODOs

Outside of the time constrain, there are a number of improvements:

  1. Continuous integration and deployment with something like Jenkins
  2. Implement mypy for type checking to reduce potential errors
  3. Implement linting for quality checking
  4. Explore using tox for platform testing