# xdit-project/xDiT

xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism

Repository: https://github.com/xdit-project/xDiT
Canonical: https://ross.abutalabs.com/products/xdit
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T13:33:13+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 56, longevity 64
- inputs: {"age_days": 897, "days_push": 7, "days_rel": 294, "gap_med": 16, "n_releases_24m": 14}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2699, forks 341 (observed 2026-08-28T04:07:11.289658+00:00)

## What it is
xDiT is a scalable inference engine for Diffusion Transformers (DiTs) that enables parallel deployment across multiple GPUs and machines. It provides parallel methods like PipeFusion, Unified Sequence Parallel, CFG Parallel, and parallel VAE, plus single-GPU accelerations via compilation and caching.

## Use cases
- run diffusion transformer inference in parallel across multiple gpus
- speed up flux or other dit image generation for real-time serving
- generate images and videos faster with sequence parallelism
- deploy diffusion models on multi-machine clusters
- accelerate comfyui diffusion model inference
- reduce latency of attention-heavy diffusion model inference

## When to choose
- you need real-time or low-latency diffusion transformer inference
- you are serving DiT-based image or video generation models at scale
- a single GPU cannot meet your throughput or latency requirements
- you use ComfyUI and want parallel acceleration

## When to avoid
- you only run small diffusion models on a single GPU with no latency pressure
- you need training or fine-tuning rather than inference
- your models are not diffusion transformer based

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, gpu-computing, image-processing, video-processing, machine-learning
- domain: deep-learning, gpu-computing, image-processing, artificial-intelligence
- platform: python
- tags: diffusion-transformers, parallel-inference, sequence-parallelism, pipefusion, multi-gpu, image-generation, video-generation, comfyui, video, gpu, linux, docker

## Member repositories
- xdit-project/xDiT (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:11.289658+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:16:08.242355+00:00, confidence not recorded.
  - readme: https://github.com/xdit-project/xDiT (fetched 2026-08-28T04:07:11.289658+00:00, sha 231cb1e66307)
- Data as of 2026-08-30T08:39:29.467469+00:00.