A drag-and-drop visual designer for data analysis and machine learning, using popular Scikit-learn and other related packages. The goal of this project is to create an environment for non-programmers to easily create baseline models, focus on data rathar than coding and debugging.
- Intergrated with well-known Scikit-learn framework, and supports related libraries
- Drag-and-drop editor, top-to-bottom straight-forward pipeline
- Can be use from beginner works to business-critical ML models
- Platform independent web interface, supports cloud deployment
- Export model in runtime mode for on-line prediction
- Support plugins for more sophisticated machine learning algorithm
Use Dockerfile to build and run docker image
#build docker image
docker build -t libretto/editor .
#start docker container
docker run -p 6789:6789 -d --name libretto-editor libretto/editorNOTE: First run takes time because required packages are being installed in first run, to check installation progress, simply attach to docker container with:
docker attach libretto-editorConnect to libretto editor with any browser to http://localhost:6789
-
Python 3
Libretto is developed with Python 3.9, other versions may work
-
clone/download this repo
-
python main.pyMain script will automatically create a virtual environment and install required packages at first run.
Examples are provided, launch Libretto, browse to example folder, drag-and-drop the json file to the center panel (or right-click -> load as receipe)
Libretto aims at One-liner is a bare-pass, coding is fail, if user has to code for the machine learning model to function then it should be built as a receipe block which can be drag-and-drop in the editor instead. While Libretto does not aim at mobile ML development, one should be able to create a ML model on their smartphone with their finger tips.
- Keras / Pytorch intergration
- Non-tabular data support
- Pre-built macro blocks for one-click machine learning model
- One click model deployment
See README.md in libretto/plugin directory