Neurula, a website for graphically designing neural networks

A drag-n-drop based GUI for designing neural networks.

The blocks represent the layers of the network, or operations on tensors.

Once the network is designed you can download auto-generated code that implements that network in pytorch or tensorflow.

Code we wrote

The code we wrote can be found in front/src/ as well as front/build/tf_code_generator.py front/build/pytorch_code_generator front/build/index.html

the rest is auto generated by svelte

Index.html

The webpage can be found in front/build/index.html

Hosted website

can be found at https://neurula.onrender.com/

HOW TO RUN LOCALLY

This directory always refers to the directory this README is contained in (the root directory of the repository) These steps work in an unix or mac enviroment

Initial set up

In this directory:

If you're not on a Mac: Install npm 
$ sudo apt install npm

    if some fail to install, do
    $ sudo apt-get update
    then run again


Install requirements
$ pip install fastapi
$ pip install uvicorn

Now to set up svelte

enter front/
$ cd front/

Install requirements
$ npm install

Build svelte app
$ npm run build

To run on local host

In this directory:
$ uvicorn api:app --reload

Once you're done
Ctrl-C to kill process

To see a change youve made

Rebuild svelte app
$ cd front
$ npm run build

You may have to clear your browser cache
    For chrome
        click three dots on right
        select more tools
        select Clear Browsing Data
        click clear data