/DLMML

Auto-DL helps you make Deep Learning models without writing a single line of code and giving as little input as possible.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

made-with-python Contributions welcome GitHub issues GitHub closed issues GitHub pull requests GitHub closed pull requests

DLMML

The intermediary representation of the generated model. This ideally would be done using our FrontEnd.

How to run

There are essentially three ways to run this code

  1. The code is an API, which is called by the Generator frontend. For instructions on how to run the React and Django Servers (GUI) please head on to User Guidlelines of the Generator Repository

  2. Run the dev-server and call the API with a POST request (no-GUI). However, tools like postman can be used.

  3. Don't want to run the server? Prefer jupyter notebooks instead? Head on to our notebooks directory

Steps for method 2:

  1.  # clone the repo
 git clone https://github.com/Auto-DL/DLMML.git

  2. Activate your environment (not compulsory but highly recommended)

  3. Create a MongoDB (local or atlas) database and put the configuration details in a ".env" file (without quotes)

  4. Place data in the ./data directory

  5.  # change dir
 cd dev_server
 # run the flask app
 python app.py

  6. Make the post request.

Note: For an example post request and to know how the data is expected in the ./data directory please head on to the User Guide

What if I want to go for method 3:

  • Follow steps 1 and 2 of method 2
  • Run the jupyter-notebook command
  • Using the GUI, navigate and run the notebooks
  • This can be a good starting point

Where to go next?

To know more about the project and initiative, please visit our website

Curious to know about our front-end or backend development? Here, Have a look :)

Note

Contributing

Please take a look at our contributing guidelines if you're interested in helping!

Features to add

  • Check if generated code is correct (current thought is to call model.compile and return errors if any)

  • Add predict functionality to the generated model

  • Add different model evaluation parameters

  • Test for backward compatibility of libraries versions

  • Benchmarking parameters

  • Add model generation code for pytorch

  • Visualization and data preprocessing steps to be added