OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Parti Pytorch

Parti Pytorch is a Python and PyTorch implementation of Parti, a pure-attention text-to-image neural network, with code for training a ViT VQ-GAN VAE and examples connecting it to Parti for model training and prompt-based image generation.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Parti Pytorch — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use the repository when you want code and examples for experimenting with a locally trained Parti-style text-to-image model, rather than a hosted prompt-to-image application.

Who it is for

  • Developers building text-to-image research workflows in PyTorch.
  • Researchers exploring Parti-style transformer models and VAE training.

Documented uses

  • Train a ViT VQ-GAN VAE on an image folder.
  • Connect a trained VAE to a Parti model and calculate training loss from text-image batches.
  • Generate PIL images from text prompts after training.
  • Pre-encode text with T5 and pass token embeddings and masks into the model workflow.
Before you choose

Requirements & limits

The repository is public on GitHub and reports an MIT license. Its README gives pip install instructions; the examples use CUDA-backed tensors.

  • The documented example trains the VAE before connecting it to Parti.
  • The sample workflow uses CUDA and provides no hardware compatibility guarantee.
  • The README lists Parti training code, inference caching, and super-resolution training code as TODO items.

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

Train the ViT VQ-GAN VAE

Set up and start VAE training on an image folder using the documented Python example.

Before you start

  • Install the parti-pytorch package with pip.
  • Provide an image folder for training.
  • Use the documented CUDA-backed model configuration.

Steps

  1. Import VitVQGanVAE and VQGanVAETrainer from parti_pytorch.

  2. Create a VitVQGanVAE and move it to CUDA.

  3. Initialize VQGanVAETrainer with the model and image folder, then call trainer.train().

Expected outcome

The trainer runs the VAE training procedure configured in the example.

Guide limitations

  • This covers VAE training only; it does not train Parti or demonstrate generating images.
  • The example uses CUDA, and the repository lists Parti training code as unfinished.
Documentation 1 ↗
Evidence and transparency

Sources & review

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