VAR logo

VAR

FoundationVision/VAR

Official implementation of Visual Autoregressive Modeling for scalable image generation

GraphCanon updated 4d · GitHub synced 4d · 33 views this month

8.7k stars571 forksLast push 9mo Jupyter Notebook MIT

Decision brief

VAR is an ultra-simple user-friendly state-of-the-art codebase for autoregressive image generation

Good fit when

  • When you prefer a straightforward implementation with minimal configuration effort
  • If fast attention computation acceleration from flash-attn or xformers is beneficial

Avoid when

  • Avoid if your project requires complex customization beyond basic VAR parameters
  • Not ideal when you do not have access to or willingness to prepare the ImageNet dataset in a specific structure

Observed Jul 15, 2026 · Source: enrich:decision_facts

Verify the decision

Maintenance and security

Full trust report
Maintenance
Slowing (279d since push)
As of 4d
Provenance
Not a fork · Organization account
As of 4d
Security (OSV)
No criticals
As of 1mo

Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.

Install

git clone https://github.com/FoundationVision/VAR

How it fits your stack(4)

Typed graph edges - alternatives, integrations, successors, and dependencies. Ranked by relationship type, not raw GitHub stars.

Relationship graph

Optional deeper exploration of typed edges and category neighbours.

Similar tools

Same-category neighbours not already linked as typed edges.

Evidence and technical details

Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.

Overview

Ultra-simple and user-friendly state-of-the-art codebase for autoregressive image generation

Capability facts

Languages
jupyter notebook

Source: github.language · Aug 17, 2026

Categories

Tags

README

Installation

  1. Install torch>=2.0.0.

  2. Install other pip packages via pip3 install -r requirements.txt.

  3. Prepare the ImageNet dataset

    assume the ImageNet is in `/path/to/imagenet`. It should be like this:
    /path/to/imagenet/:
        train/:
            n01440764: 
                many_images.JPEG ...
            n01443537:
                many_images.JPEG ...
        val/:
            n01440764:
                ILSVRC2012_val_00000293.JPEG ...
            n01443537:
                ILSVRC2012_val_00000236.JPEG ...
    

    NOTE: The arg --data_path=/path/to/imagenet should be passed to the training script.

  4. (Optional) install and compile flash-attn and xformers for faster attention computation. Our code will automatically use them if installed. See models/basic_var.py#L15-L30.


License

This project is licensed under the MIT License - see the LICENSE file for details.

For agents

This page has a .md twin and JSON over the API.

Was this helpful?

Anonymous feedback helps us improve pages and translations.