hojonathanho/diffusion
Official TensorFlow implementation of the DDPM paper for training and evaluating diffusion probabilistic models on image generation.

Velocity · 7d
+2.4
★ / day
Trend
→steady
star history
This repository contains the reference implementation of the 2020 NeurIPS paper introducing Denoising Diffusion Probabilistic Models. It provides training and evaluation scripts for a generative model that learns to reverse a gradual noising process to produce images. The model was trained on Google Cloud TPU v3-8 hardware using TensorFlow 1.15.