README.md

Flow toolkit

---

A simple development toolkit for streamlining ML and DL workflows.

Getting started

Installation

The base compiler can be installed via pip:

pip install flow-toolkit

Next, install a plugin for your intended output. For example:

flow -i https://github.com/StealthyPanda/pytorch-backend.git

Any github repository can be used as a valid flow plugin, as long as it contains a plugin.py in its root directory, and contains a main function.

Quick Start

A simple flow for a dense neural network would be:

// example.fl

flow linear(x) [weights, biases] {
    return
        (weights @ x) + biases;
}

flow NeuralNetwork (x) {
    let linear l1;
    let linear l2;

    y = l1(x);
    y = l2(y);

    return y;
}

build NeuralNetwork simple {
    x => 784;
    output => 10;
    l1 => {
        output => 16;
    };
}

Build the flow to a pytorch model with:

flow -f example.fl -o example