OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Autoregressive Diffusion Pytorch

Autoregressive Diffusion Pytorch is a Python/PyTorch implementation of an autoregressive diffusion architecture for image generation without vector quantization. It provides models for tensor sequences and images, with documented training-loss, sampling, image-dataset and trainer interfaces.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Autoregressive Diffusion Pytorch — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when a local or self-hosted PyTorch implementation is suitable for research or model-building work and you need documented sequence-level or image-level diffusion components.

Who it is for

  • PyTorch researchers studying autoregressive diffusion
  • Developers prototyping image-generation training pipelines
  • Users experimenting with custom image data and sequence representations

Documented uses

  • Train an image-generation model with autoregressive diffusion.
  • Prototype comparisons between autoregressive diffusion and other image-generation approaches.
  • Build a PyTorch training pipeline around the provided model modules.
  • Experiment with image datasets represented as sequences of patches or tokens.
Before you choose

Requirements & limits

The repository is public, active and MIT-licensed. It documents installation as a Python package with pip. No account or pricing requirement is provided.

  • The implementation is Python/PyTorch rather than a hosted image-generation application.
  • The image workflow uses image_size and patch_size to configure image inputs.
  • The repository includes both direct model usage and an ImageDataset/ImageTrainer workflow.
  • A flow-matching variant is documented through ImageAutoregressiveFlow and AutoregressiveFlow.
  • The available material does not establish a complete end-to-end pipeline for custom datasets.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up an image-training workflow

Install the package and configure the documented image dataset, model and trainer components.

Before you start

  • Python environment with PyTorch
  • The autoregressive-diffusion-pytorch package
  • An image directory
  • An image size

Steps

  1. Install the package with pip install autoregressive-diffusion-pytorch.

  2. Create an ImageDataset from an image directory and specify image_size.

  3. Create an ImageAutoregressiveDiffusion model with model settings, image_size and patch_size.

  4. Create an ImageTrainer with the model and dataset, then invoke the trainer as shown in the repository example.

Expected outcome

A configured image-training workflow using ImageDataset, ImageAutoregressiveDiffusion and ImageTrainer.

Guide limitations

  • The supplied material documents the setup and trainer invocation but does not establish training results or custom-dataset compatibility beyond the shown dataset interface.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
4 October 2026
Original website title
lucidrains/autoregressive-diffusion-pytorch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction