/torch-generative-models

Primary LanguagePython

Generative Models in Torch

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

Usage

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

Citation

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}
}