generative_minimal

a repository for minimal implementations of common generative models

each model class contains a train.py script to train the associated model

Models:

Generative Adversarial Networks (GANs)
- Generative Adversarial Network (GAN)
- Deep Convolutional Generative Adversarial Network (DCGAN)
- Wasserstein Generative Adversarial Network (WGAN)
Variational Auto Encoder (VAEs)
- Variation Auto Encoder (VAE)
- Conditional Variational Auto Encoder (CVAE)
Normalizing Flows
- Normalizing Flow
Diffusion Models
- Denoising Diffusion Probabalistic Model (DDPM)
Transformers
- Encoder-only
- Decoder-only
- Encoder-Decoder

feel free to submit an issue or a pull request with additional models following the same format used here

Setup

conda create -n generative -y python=3.11 && conda activate generative
git clone [email protected]:KyleM73/generative_minimal.git
cd generative_minimal
pip install -e .
cd generative_minimal
python get_datasets.py
python models/<model>/train.py

Datasets

MNIST