/SurVAE-replication

A replication of "SurVAE Flows: Surjections to Bridge the Gap between VAEs and Flows", implemented as the final project for the "Generative Neural Networks" class of the University of Heidelberg, 2024.

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

SurVAE

A replication of "SurVAE Flows: Surjections to Bridge the Gap between VAEs and Flows", implemented as the final project for the "Generative Neural Networks" class of the University of Heidelberg, 2024.

[Our Report] [Original Paper]

Overview image

Structure

The repository contains the following:

├── assets     # relevant PDFs
├── report     # source code for the report
├── notebooks  # Jupyter notebooks
├── saves      # save states for MNIST
└── survae     # source code

Running the code

  1. install the requirements (pip install -r requirements.txt)
  2. open and run any notebook in notebooks/

Tests

To run all of the tests, use the following command:

python3 -m survae.test