Giter VIP home page Giter VIP logo

mt-cgcnn's Introduction

Bayesian Graph Convolutional Neural Networks for Crystallography

This repository contains work on Bayesian GCNs for crystallography. The goal of the project is to infer material properties with uncertainty measurements from material properties. We build upon CGCNN and MT-CGCNN.

Table of Contents

Prerequisites

pip install -r requirements.txt

Usage

Download CIF files from the Materials Project

To download CIF files, you will need to first [generate an API KEY]. Then run:

python utils.py --api_key API_KEY --mp_ids csvs/mpids_full.csv --output data/materials_project

Training

To see training of an instance of MT-CGCNN, you can just run the following:

python init.py

This will run a demo using datasource: data/sample/ and some pre-defined set of parameters. But, you can change and play with the parameters of the model using the init.py file. To run multiple iterations of the same experiment (one motivation can be to get average error), you can run the following code:

python init.py --idx 1

To reproduce results stated in the paper, you might need to tune the parameters in the hyperparameter space mentioned in the paper. Also, average MAE is reported for 5 runs of the experiment.

After training, you will get multiple files in results folder present within the datasource (For eg., for demo case results will be saved in data/sample/results/0/). The most important ones are:

  • model_best.pth.tar: stores the MT-CGCNN model with the best validation accuracy.
  • checkpoint.pth.tar: stores the MT-CGCNN model at the last epoch.
  • test_results.csv: stores the ID, target values, and predicted values for each crystal in test set.
  • logfile.log: A complete log of the experiment (useful for DEBUGGING purposes)

The other files are useful to understand how well-trained the model is and can be referred for DEBUGGING purposes. Briefly, the files contain information about the training & validation losses, training & validation errors and also some useful plots.

License

MT-CGCNN is released under the MIT License.

mt-cgcnn's People

Contributors

soumyasanyal avatar

Watchers

 avatar  avatar  avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.