arpitbansal297/Cold-Diffusion-Models

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1136 stars · 83 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official PyTorch implementation of Cold Diffusion, a research paper showing that diffusion-style generative models can invert arbitrary image degradations (blur, masking, downsampling, desaturation) rather than only Gaussian noise. It includes training and sampling scripts plus pretrained models for MNIST, CIFAR-10, CelebA, and AFHQ.

Use cases

• train a diffusion model that uses blur instead of noise
• generate images with generalized diffusion models
• reproduce the Cold Diffusion paper experiments
• experiment with non-Gaussian degradation processes in diffusion
• compare hot vs cold diffusion sampling algorithms
• generate CelebA or AFHQ faces with pretrained models

When to choose

• you want to research or extend generalized diffusion models with arbitrary degradations
• you need a PyTorch codebase for the Cold Diffusion paper
• you want to test whether deterministic degradations suffice for generation

When to avoid

• you need a production-ready, licensed generative image library
• you want standard DDPM denoising with active maintenance and community support
• you need a stable API or long-term support

Facets

library · maturity maintenance

machine-learning deep-learning image-processing machine-learning computer-vision image-processing deep-learning python diffusion-models generative-models pytorch research-code cold-diffusion image-generation

Member repositories

For agents

markdown · JSON · MCP: product_card(name="arpitbansal297/Cold-Diffusion-Models")