Skip to content

ALFIE AutoML Engine

An AutoML engine written for the ALFIE project with the following features - AutoML for tabular - AutoML for vision (this is a WIP) - Website accessibility checker - Image to website tool - Part of a host of tools made for better and more informed generation of AI models

Quickstart

Installation

  • First, follow the steps here, installation
  • Then, follow this to install AutoDW

Generate sample data (optional)

The repo includes mk_sample_data.sh which downloads small datasets under sample_data/:

1
2
# From project root
bash mk_sample_data.sh
What it fetches: - sample_data/knot_theory/{train.csv,test.csv} - sample_data/m4_hourly_subset/{train.csv,test.csv}

If wget is missing on macOS: brew install wget.

Configuration

You can set environment variables via the .env file in the project root.

  • Copy the .env.template to .env and fill in whatever is missing
  • Change the ports if needed
  • Uploads are saved under uploaded_data/.
  • AutoML artifacts (from training) are written alongside the uploaded session folder in automl_data_path/.

Running the AutoML

  • Once you have everything setup, and the data uploaded to AutoDW, you can use the AUotML!
    • Note that if you only want to use AutoMLPlus, you don't need to use AutoDW or have it running
  • Just follow this to run the service of your choosing

System components

System components

Loading the trained model for inference

  • Follow the instructions here
  • Once the AutoML tool is done, it will point you to a folder with