ml-pipelines
Applications for managing machine learning pipelines and human workflows around them, published under BSD-2 license.
At the moment of writing this repository includes only the MLP Catalog - a web aplication for exploring projects and datasets stored within the metadata library.
To understand context of the project, check out STUDY.md.
Explore Datasets
Find relevant data by searching across all datasets.
View Projects
Organize elements of ML Pipelines into projects.
View Datasets
View dataset properties and relations.
Specs
Application functionality is being covered with event-driven specs. This captures business logic and UX flows in non-fragile way.
Getting started
Application is build and tested with Python 3.7.
Prerequisites:
- Python 3.7 with dev libraries:
apt install python3.7 python3.7-dev
-
graphviz:
apt install graphviz
- virtualenv
To get started, go to the mlp
folder and:
- set up a virtualenv in
mlp
folder and activate it; -
pip install -r requirements.txt
- install all the dependencies; -
python manage.py specs
- to run tests; -
python manage.py demo && python manage.py runserver
to fill up DB with demo data and launch the web UI (available at localhost:8000)