This repository is a collection of generative models for scenes in PyTorch. The following models have been implemented-
- Variational Autoencoders (VAEs)
vae.py
- Generative Adversarial Networks (GANs)
gan.py
- Energy-based Model (EBM)
ebm.py
- Variational Energy-based Model (VAEBM)
vaebm.py
- Energy-based Autoencoder (EAE)
ebmvae_1.py
- Score-based Model/Noise Contrastive Score Network (NCSN)
ncsn.py
Download the manipulation scenes dataset from this link and place it in the data
directory.
Train a model by running its respective file. For instance, train the VAEBM using the following-
python train_vaebm.py
If you find this repository helpful then please cite the following-
@misc{ebmrl,
author = {Suri, Karush},
title = {{Generative Models in Torch}},
url = {https://github.com/karush17/ebm-rl},
year = {2021}
}