flash-attention-3 开源项目推荐（13 个）

xlite-dev /Awesome-LLM-Inference

📚A curated list of Awesome LLM/VLM Inference Papers with Codes: Flash-Attention, Paged-Attention, WINT8/4, Parallelism, etc.🎉

shreyansh26 /FlashAttention-PyTorch

Implementation of FlashAttention (FA1-FA4) in PyTorch for educational and algorithmic clarity

harleyszhang /rapid_llm

RapidLLM is an LLM inference framework with continuous batching, tensor and data parallelism, quantized weights, and with pluggable Triton/CUDA kernels.

PRITHIVSAKTHIUR /Qwen-Image-Edit-2511-LoRAs-Fast-Lazy-Load

Qwen-Image-Edit-2511型号的演示，配备懒惰加载的LoRA适配器，用于高级单图和多图编辑。支持7+专业LoRAs，包括照片转动画、多角度摄像控制、姿势转换（任意姿势）、放大、风格转移、光线迁移和漫画色调。具备快速推理（默认4步）。

windreamer /flash-attention3-wheels

Pre-built wheels that erase Flash Attention 3 installation headaches.

PRITHIVSAKTHIUR /FireRed-Image-Edit-1.0-Fast

FireRed-Image-Edit-1.0-Fast is a high-performance, AI-driven image editing application that utilizes advanced diffusers and the QIE+ Pipeline for precise, prompt-based image modifications.

PRITHIVSAKTHIUR /Qwen-Image-Edit-2509-LoRAs-Fast

Qwen-Image-Edit-2509-LoRAs-Fast is a high-performance, user-friendly web application built with Gradio that leverages the advanced Qwen/Qwen-Image-Edit-2509 model from Hugging Face for seamless image editing tasks.

PRITHIVSAKTHIUR /Qwen-Image-Edit-3D-Lighting-Control

Demonstration for the Qwen/Qwen-Image-Edit-2509 model with 3D lighting control using the Multi-Angle-Lighting LoRA adapter. Allows precise control of light source direction via interactive 3D viewport or sliders for azimuth and elevation..

PRITHIVSAKTHIUR /Qwen-Image-Edit-Object-Manipulator

Qwen-Image-Edit-Object-Manipulator is an advanced, high-performance image manipulation workspace built on top of the Qwen/Qwen-Image-Edit-2511 base pipeline and enhanced by a custom optimized transformer.

PRITHIVSAKTHIUR /LTX-2-LoRAs-Camera-Control-Dolly

Demonstration for the Lightricks LTX-2 Distilled model, enhanced with specialized LoRA adapters for cinematic camera movements (dolly left/right/in/out, jib up/down, static). Generates animated videos from text prompts or input images, with optional prompt enhancement using Gemma-3-12b.

gietema /attention

Toy Flash Attention implementation in torch

aw920h /flash-attn-3-wheels

Collection of Flash attention 3 precompiled wheels for direct install on Hopper series NVIDIA GPU (H100, H200 sm_90a)

PRITHIVSAKTHIUR /Qwen-Image-Edit-2509-LoRAs-Fast-Fusion-Lazy-Load

Qwen-Image-Edit-2509-LoRAs-Fast-Fusion-Lazy-Load is an advanced multi-image generative studio and reference-based composition platform. Built upon the Qwen/Qwen-Image-Edit-2509 base architecture and an optimized transformer backend.