A software engineering framework to jumpstart your Machine Learning projects
Install lolpop by cd'ing to the this directory and executing:
poetry install
pip install -e .Note: getting the poetry install to work has been difficult, due to some of conflicts among packages (I think primarly dbt. If you're having issues getting it to work, feel free to try removing things that are troublesome and/or moving it into a requirements.txt file and installing via pip)
Try out the classification example in examples/classification/mlflow. To run it, simply execute
python3 examples/classification/mlflow/mlrunner.py Note: You'll want to update examples/classification/mlflow/local_dev.yaml with your own configuration. In particular, update the mlflow configuration. You also need to download the test.csv and train.csv files
from Kaggle and update your conf to point to their locations.
You'll probably hit a few errors in trying to run the first time. Let me know what they are.
If you're interested in contributing, there is a basic CLI tool that can boostrap a new runner/pipeline/component for you. To use, simply execute:
python3 cli/cli.py create component/pipeline/runner <component_type> <componenet_class>Component type should be something like metadata_tracker, metrics_tracker, model_trainer, etc, and component class is the name of the class. This will create a new component type in the specified directory in lolpop.
This will boostrap a cookiecutter project in the provided --component-dir which you can then edit as desired.