# igul222/improved_wgan_training

Code for reproducing experiments in "Improved Training of Wasserstein GANs"

Repository: https://github.com/igul222/improved_wgan_training
Canonical: https://ross.abutalabs.com/products/improved_wgan_training
Language: Python
License: MIT
License Family: permissive
Last push: 2018-06-06T13:53:11+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3441, "days_push": 3010, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2411, forks 661 (observed 2026-08-28T04:06:44.441816+00:00)

## What it is
Reference implementation of the 'Improved Training of Wasserstein GANs' (WGAN-GP) paper, providing TensorFlow code to reproduce its experiments. It includes training scripts for toy datasets, MNIST, CIFAR-10, 64x64 images, and character-level language modeling.

## Use cases
- reproduce WGAN-GP paper experiments
- train a Wasserstein GAN with gradient penalty on MNIST
- compare GAN loss functions on toy datasets like 8 Gaussians and Swiss Roll
- learn how to implement gradient penalty in TensorFlow
- train a GAN on CIFAR-10 or 64x64 images
- experiment with GANs for character-level language modeling

## When to choose
- you want to reproduce or verify the WGAN-GP paper's results
- you need a clear reference implementation of the gradient penalty technique
- you are studying GAN training stability with simple, readable TensorFlow code

## When to avoid
- you need a production-ready or actively maintained GAN framework
- you use PyTorch or modern TensorFlow 2.x APIs
- you need flexible configuration, checkpointing, or multi-GPU support out of the box

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning, artificial-intelligence
- platform: python
- tags: wgan, gan, wasserstein, gradient-penalty, tensorflow, research-code, generative-models, linux, gpu

## Member repositories
- igul222/improved_wgan_training (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:44.441816+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:33:58.333471+00:00, confidence not recorded.
  - readme: https://github.com/igul222/improved_wgan_training (fetched 2026-08-28T04:06:44.441816+00:00, sha 27c104a4708e)
- Data as of 2026-08-30T08:39:29.467469+00:00.