
SpargeAttn
SpargeAttention: A training-free sparse attention that can accelerate any model inference.
Stars: 719

SpargeAttn is an official implementation designed for accelerating any model inference by providing accurate sparse attention. It offers a significant speedup in model performance while maintaining quality. The tool is based on SageAttention and SageAttention2, providing options for different levels of optimization. Users can easily install the package and utilize the available APIs for their specific needs. SpargeAttn is particularly useful for tasks requiring efficient attention mechanisms in deep learning models.
README:
The official implementation of SpargeAttn, a universal training-free sparse attention accelerating language, image, and video models.
- Sparse SageAttention1 API and Sparse SageAttention2 API can compute attention with any block sparse pattern very fast.
- SpargeAttn based on SageAttention2++ will be released around June 25.
- [2025-05-11]: Add a very simple usage without tuning or calibration:
o = spas_sage2_attn_meansim_cuda(q, k, v)
. - [2025-05-02]: 🎉SpargeAttn and SageAttention2 are accepted by ICML 2025!
- [2025-01-24]: 🎉SageAttention is accepted by ICLR 2025!
-
python>=3.9
,torch>=2.3.0
-
CUDA
:-
>=12.8
for Blackwell,>=12.4
for fp8 support on Ada,>=12.3
for fp8 support on Hopper,>=12.0
for Ampere
-
pip install ninja # for parallel compilation
python setup.py install # or pip install -e .
-
spas_sage2_attn_meansim_cuda
: SpargeAttn based on SageAttention2. -
spas_sage_attn_meansim_cuda
: SpargeAttn based on SageAttention.
from spas_sage_attn import block_sparse_sage2_attn_cuda
block_sparse_sage2_attn_cuda(q, k, v, mask_id=None, scale=None, pvthreshd=20, attention_sink=False, tensor_layout="HND", return_sparsity=False):
In this API, we support computing $S=QK^T$ in any block sparse pattern per attention head. And we compute $PV$ multiplication with further acceleration. Specifically, the attention mask per head, mask_id
, is of shape (batch_size, num_qo_heads, qo_seq_len // BLOCK_M, kv_seq_len // BLOCK_N)
. Currently, the supported block size is aligned to that of SpargeAttention, which is (BLOCK_M = 128, BLOCK_N = 64). The lower pvthreshd
, the more sparsity for PV
Matmul and faster attention.
from spas_sage_attn import spas_sage2_attn_meansim_cuda
attn_output = spas_sage2_attn_meansim_cuda(q, k, v, simthreshd1=0.6, cdfthreshd=0.98, is_causal=False)
You can tune simthreshd1
and cdfthreshd
to balance between attention accuracy (higher values) and sparsity (lower values). However, for optimal accuracy and sparsity performance, we recommend running a tuning process before inference, as described below.
Tuning:
# sequential tuning
python evaluate/cogvideo_example.py --use_spas_sage_attn --model_out_path evaluate/models_dict/CogVideoX-2b_0.06_0.07.pt --tune
# parallel tuning, this will use all gpu available on the machine
python evaluate/cogvideo_example.py --use_spas_sage_attn --model_out_path evaluate/models_dict/CogVideoX-2b_0.06_0.07.pt --tune --parallel_tune
Inference:
# `--compile` is optional and will slow the first time inference.
python evaluate/cogvideo_example.py --use_spas_sage_attn --model_out_path evaluate/models_dict/CogVideoX-2b_0.06_0.07.pt --compile
Note: We provide pre-tuned hyper-parameters
CogVideoX-2b_0.06_0.07.pt
that allow running the inference script directly. However, for better performance in both speed and quality, we recommend re-tuning because the provided hyper-parameters are tuned with SpargeAttn based on SageAttention, whereas the default API is based on SageAttention2 now.
Note:
--compile
is optional and will further accelerate video generation but bring an overhead for the first video generation.
The tuning and inference usage is similar to CogVideoX.
Here’s a list of the tuned models so far, go to hugginface to see all tuned ckpt. Our approach is universal, and we warmly welcome contributions! Feel free to submit a pull request to support more models. 🚀
model name | example script | tuned ckpt |
---|---|---|
CogVideoX-2b | evaluate/cogvideo_example.py | link |
want2v-1.3B | evaluate/wan_example.py | link |
Flux | evaluate/flux_example.py | TBD |
Note: All experiments in the above Table and our paper used SpargeAttn based on SageAttention. An updated implementation based on SageAttention2, is available now. It further offers a 30% speedup.
![]() The quality of video generation on Mochi. |
![]() End-to-end performance of NIAH. |
If you use this code or find our work valuable, please cite:
@inproceedings{zhang2025spargeattn,
title={Spargeattn: Accurate sparse attention accelerating any model inference},
author={Zhang, Jintao and Xiang, Chendong and Huang, Haofeng and Wei, Jia and Xi, Haocheng and Zhu, Jun and Chen, Jianfei},
booktitle={International Conference on Machine Learning (ICML)},
year={2025}
}
@inproceedings{zhang2025sageattention,
title={SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Acceleration},
author={Zhang, Jintao and Wei, Jia and Zhang, Pengle and Zhu, Jun and Chen, Jianfei},
booktitle={International Conference on Learning Representations (ICLR)},
year={2025}
}
@inproceedings{zhang2024sageattention2,
title={Sageattention2: Efficient attention with thorough outlier smoothing and per-thread int4 quantization},
author={Zhang, Jintao and Huang, Haofeng and Zhang, Pengle and Wei, Jia and Zhu, Jun and Chen, Jianfei},
booktitle={International Conference on Machine Learning (ICML)},
year={2025}
}
For Tasks:
Click tags to check more tools for each tasksFor Jobs:
Alternative AI tools for SpargeAttn
Similar Open Source Tools

SpargeAttn
SpargeAttn is an official implementation designed for accelerating any model inference by providing accurate sparse attention. It offers a significant speedup in model performance while maintaining quality. The tool is based on SageAttention and SageAttention2, providing options for different levels of optimization. Users can easily install the package and utilize the available APIs for their specific needs. SpargeAttn is particularly useful for tasks requiring efficient attention mechanisms in deep learning models.

SageAttention
SageAttention is an official implementation of an accurate 8-bit attention mechanism for plug-and-play inference acceleration. It is optimized for RTX4090 and RTX3090 GPUs, providing performance improvements for specific GPU architectures. The tool offers a technique called 'smooth_k' to ensure accuracy in processing FP16/BF16 data. Users can easily replace 'scaled_dot_product_attention' with SageAttention for faster video processing.

Vitron
Vitron is a unified pixel-level vision LLM designed for comprehensive understanding, generating, segmenting, and editing static images and dynamic videos. It addresses challenges in existing vision LLMs such as superficial instance-level understanding, lack of unified support for images and videos, and insufficient coverage across various vision tasks. The tool requires Python >= 3.8, Pytorch == 2.1.0, and CUDA Version >= 11.8 for installation. Users can deploy Gradio demo locally and fine-tune their models for specific tasks.

paxml
Pax is a framework to configure and run machine learning experiments on top of Jax.

KVCache-Factory
KVCache-Factory is a unified framework for KV Cache compression of diverse models. It supports multi-GPUs inference with big LLMs and various attention implementations. The tool enables KV cache compression without Flash Attention v2, multi-GPU inference, and specific models like Mistral. It also provides functions for KV cache budget allocation and batch inference. The visualization tools help in understanding the attention patterns of models.

zml
ZML is a high-performance AI inference stack built for production, using Zig language, MLIR, and Bazel. It allows users to create exciting AI projects, run pre-packaged models like MNIST, TinyLlama, OpenLLama, and Meta Llama, and compile models for accelerator runtimes. Users can also run tests, explore examples, and contribute to the project. ZML is licensed under the Apache 2.0 license.

Pixel-Reasoner
Pixel Reasoner is a framework that introduces reasoning in the pixel-space for Vision-Language Models (VLMs), enabling them to directly inspect, interrogate, and infer from visual evidences. This enhances reasoning fidelity for visual tasks by equipping VLMs with visual reasoning operations like zoom-in and select-frame. The framework addresses challenges like model's imbalanced competence and reluctance to adopt pixel-space operations through a two-phase training approach involving instruction tuning and curiosity-driven reinforcement learning. With these visual operations, VLMs can interact with complex visual inputs such as images or videos to gather necessary information, leading to improved performance across visual reasoning benchmarks.

litserve
LitServe is a high-throughput serving engine for deploying AI models at scale. It generates an API endpoint for a model, handles batching, streaming, autoscaling across CPU/GPUs, and more. Built for enterprise scale, it supports every framework like PyTorch, JAX, Tensorflow, and more. LitServe is designed to let users focus on model performance, not the serving boilerplate. It is like PyTorch Lightning for model serving but with broader framework support and scalability.

can-ai-code
Can AI Code is a self-evaluating interview tool for AI coding models. It includes interview questions written by humans and tests taken by AI, inference scripts for common API providers and CUDA-enabled quantization runtimes, a Docker-based sandbox environment for validating untrusted Python and NodeJS code, and the ability to evaluate the impact of prompting techniques and sampling parameters on large language model (LLM) coding performance. Users can also assess LLM coding performance degradation due to quantization. The tool provides test suites for evaluating LLM coding performance, a webapp for exploring results, and comparison scripts for evaluations. It supports multiple interviewers for API and CUDA runtimes, with detailed instructions on running the tool in different environments. The repository structure includes folders for interviews, prompts, parameters, evaluation scripts, comparison scripts, and more.

mem-kk-logic
This repository provides a PyTorch implementation of the paper 'On Memorization of Large Language Models in Logical Reasoning'. The work investigates memorization of Large Language Models (LLMs) in reasoning tasks, proposing a memorization metric and a logical reasoning benchmark based on Knights and Knaves puzzles. It shows that LLMs heavily rely on memorization to solve training puzzles but also improve generalization performance through fine-tuning. The repository includes code, data, and tools for evaluation, fine-tuning, probing model internals, and sample classification.

gateway
Adaline Gateway is a fully local production-grade Super SDK that offers a unified interface for calling over 200+ LLMs. It is production-ready, supports batching, retries, caching, callbacks, and OpenTelemetry. Users can create custom plugins and providers for seamless integration with their infrastructure.

lloco
LLoCO is a technique that learns documents offline through context compression and in-domain parameter-efficient finetuning using LoRA, which enables LLMs to handle long context efficiently.

mLoRA
mLoRA (Multi-LoRA Fine-Tune) is an open-source framework for efficient fine-tuning of multiple Large Language Models (LLMs) using LoRA and its variants. It allows concurrent fine-tuning of multiple LoRA adapters with a shared base model, efficient pipeline parallelism algorithm, support for various LoRA variant algorithms, and reinforcement learning preference alignment algorithms. mLoRA helps save computational and memory resources when training multiple adapters simultaneously, achieving high performance on consumer hardware.

Fira
Fira is a memory-efficient training framework for Large Language Models (LLMs) that enables full-rank training under low-rank constraint. It introduces a method for training with full-rank gradients of full-rank weights, achieved with just two lines of equations. The framework includes pre-training and fine-tuning functionalities, packaged as a Python library for easy use. Fira utilizes Adam optimizer by default and provides options for weight decay. It supports pre-training LLaMA models on the C4 dataset and fine-tuning LLaMA-7B models on commonsense reasoning tasks.

LLMBox
LLMBox is a comprehensive library designed for implementing Large Language Models (LLMs) with a focus on a unified training pipeline and comprehensive model evaluation. It serves as a one-stop solution for training and utilizing LLMs, offering flexibility and efficiency in both training and utilization stages. The library supports diverse training strategies, comprehensive datasets, tokenizer vocabulary merging, data construction strategies, parameter efficient fine-tuning, and efficient training methods. For utilization, LLMBox provides comprehensive evaluation on various datasets, in-context learning strategies, chain-of-thought evaluation, evaluation methods, prefix caching for faster inference, support for specific LLM models like vLLM and Flash Attention, and quantization options. The tool is suitable for researchers and developers working with LLMs for natural language processing tasks.

DeepPavlov
DeepPavlov is an open-source conversational AI library built on PyTorch. It is designed for the development of production-ready chatbots and complex conversational systems, as well as for research in the area of NLP and dialog systems. The library offers a wide range of models for tasks such as Named Entity Recognition, Intent/Sentence Classification, Question Answering, Sentence Similarity/Ranking, Syntactic Parsing, and more. DeepPavlov also provides embeddings like BERT, ELMo, and FastText for various languages, along with AutoML capabilities and integrations with REST API, Socket API, and Amazon AWS.
For similar tasks

AutoGPTQ
AutoGPTQ is an easy-to-use LLM quantization package with user-friendly APIs, based on GPTQ algorithm (weight-only quantization). It provides a simple and efficient way to quantize large language models (LLMs) to reduce their size and computational cost while maintaining their performance. AutoGPTQ supports a wide range of LLM models, including GPT-2, GPT-J, OPT, and BLOOM. It also supports various evaluation tasks, such as language modeling, sequence classification, and text summarization. With AutoGPTQ, users can easily quantize their LLM models and deploy them on resource-constrained devices, such as mobile phones and embedded systems.

Qwen-TensorRT-LLM
Qwen-TensorRT-LLM is a project developed for the NVIDIA TensorRT Hackathon 2023, focusing on accelerating inference for the Qwen-7B-Chat model using TRT-LLM. The project offers various functionalities such as FP16/BF16 support, INT8 and INT4 quantization options, Tensor Parallel for multi-GPU parallelism, web demo setup with gradio, Triton API deployment for maximum throughput/concurrency, fastapi integration for openai requests, CLI interaction, and langchain support. It supports models like qwen2, qwen, and qwen-vl for both base and chat models. The project also provides tutorials on Bilibili and blogs for adapting Qwen models in NVIDIA TensorRT-LLM, along with hardware requirements and quick start guides for different model types and quantization methods.

stable-diffusion.cpp
The stable-diffusion.cpp repository provides an implementation for inferring stable diffusion in pure C/C++. It offers features such as support for different versions of stable diffusion, lightweight and dependency-free implementation, various quantization support, memory-efficient CPU inference, GPU acceleration, and more. Users can download the built executable program or build it manually. The repository also includes instructions for downloading weights, building from scratch, using different acceleration methods, running the tool, converting weights, and utilizing various features like Flash Attention, ESRGAN upscaling, PhotoMaker support, and more. Additionally, it mentions future TODOs and provides information on memory requirements, bindings, UIs, contributors, and references.

LMOps
LMOps is a research initiative focusing on fundamental research and technology for building AI products with foundation models, particularly enabling AI capabilities with Large Language Models (LLMs) and Generative AI models. The project explores various aspects such as prompt optimization, longer context handling, LLM alignment, acceleration of LLMs, LLM customization, and understanding in-context learning. It also includes tools like Promptist for automatic prompt optimization, Structured Prompting for efficient long-sequence prompts consumption, and X-Prompt for extensible prompts beyond natural language. Additionally, LLMA accelerators are developed to speed up LLM inference by referencing and copying text spans from documents. The project aims to advance technologies that facilitate prompting language models and enhance the performance of LLMs in various scenarios.

Awesome-Efficient-LLM
Awesome-Efficient-LLM is a curated list focusing on efficient large language models. It includes topics such as knowledge distillation, network pruning, quantization, inference acceleration, efficient MOE, efficient architecture of LLM, KV cache compression, text compression, low-rank decomposition, hardware/system, tuning, and survey. The repository provides a collection of papers and projects related to improving the efficiency of large language models through various techniques like sparsity, quantization, and compression.

TensorRT-Model-Optimizer
The NVIDIA TensorRT Model Optimizer is a library designed to quantize and compress deep learning models for optimized inference on GPUs. It offers state-of-the-art model optimization techniques including quantization and sparsity to reduce inference costs for generative AI models. Users can easily stack different optimization techniques to produce quantized checkpoints from torch or ONNX models. The quantized checkpoints are ready for deployment in inference frameworks like TensorRT-LLM or TensorRT, with planned integrations for NVIDIA NeMo and Megatron-LM. The tool also supports 8-bit quantization with Stable Diffusion for enterprise users on NVIDIA NIM. Model Optimizer is available for free on NVIDIA PyPI, and this repository serves as a platform for sharing examples, GPU-optimized recipes, and collecting community feedback.

lightning-bolts
Bolts package provides a variety of components to extend PyTorch Lightning, such as callbacks & datasets, for applied research and production. Users can accelerate Lightning training with the Torch ORT Callback to optimize ONNX graph for faster training & inference. Additionally, users can introduce sparsity with the SparseMLCallback to accelerate inference by leveraging the DeepSparse engine. Specific research implementations are encouraged, with contributions that help train SSL models and integrate with Lightning Flash for state-of-the-art models in applied research.

ms-swift
ms-swift is an official framework provided by the ModelScope community for fine-tuning and deploying large language models and multi-modal large models. It supports training, inference, evaluation, quantization, and deployment of over 400 large models and 100+ multi-modal large models. The framework includes various training technologies and accelerates inference, evaluation, and deployment modules. It offers a Gradio-based Web-UI interface and best practices for easy application of large models. ms-swift supports a wide range of model types, dataset types, hardware support, lightweight training methods, distributed training techniques, quantization training, RLHF training, multi-modal training, interface training, plugin and extension support, inference acceleration engines, model evaluation, and model quantization.
For similar jobs

sweep
Sweep is an AI junior developer that turns bugs and feature requests into code changes. It automatically handles developer experience improvements like adding type hints and improving test coverage.

teams-ai
The Teams AI Library is a software development kit (SDK) that helps developers create bots that can interact with Teams and Microsoft 365 applications. It is built on top of the Bot Framework SDK and simplifies the process of developing bots that interact with Teams' artificial intelligence capabilities. The SDK is available for JavaScript/TypeScript, .NET, and Python.

ai-guide
This guide is dedicated to Large Language Models (LLMs) that you can run on your home computer. It assumes your PC is a lower-end, non-gaming setup.

classifai
Supercharge WordPress Content Workflows and Engagement with Artificial Intelligence. Tap into leading cloud-based services like OpenAI, Microsoft Azure AI, Google Gemini and IBM Watson to augment your WordPress-powered websites. Publish content faster while improving SEO performance and increasing audience engagement. ClassifAI integrates Artificial Intelligence and Machine Learning technologies to lighten your workload and eliminate tedious tasks, giving you more time to create original content that matters.

chatbot-ui
Chatbot UI is an open-source AI chat app that allows users to create and deploy their own AI chatbots. It is easy to use and can be customized to fit any need. Chatbot UI is perfect for businesses, developers, and anyone who wants to create a chatbot.

BricksLLM
BricksLLM is a cloud native AI gateway written in Go. Currently, it provides native support for OpenAI, Anthropic, Azure OpenAI and vLLM. BricksLLM aims to provide enterprise level infrastructure that can power any LLM production use cases. Here are some use cases for BricksLLM: * Set LLM usage limits for users on different pricing tiers * Track LLM usage on a per user and per organization basis * Block or redact requests containing PIIs * Improve LLM reliability with failovers, retries and caching * Distribute API keys with rate limits and cost limits for internal development/production use cases * Distribute API keys with rate limits and cost limits for students

uAgents
uAgents is a Python library developed by Fetch.ai that allows for the creation of autonomous AI agents. These agents can perform various tasks on a schedule or take action on various events. uAgents are easy to create and manage, and they are connected to a fast-growing network of other uAgents. They are also secure, with cryptographically secured messages and wallets.

griptape
Griptape is a modular Python framework for building AI-powered applications that securely connect to your enterprise data and APIs. It offers developers the ability to maintain control and flexibility at every step. Griptape's core components include Structures (Agents, Pipelines, and Workflows), Tasks, Tools, Memory (Conversation Memory, Task Memory, and Meta Memory), Drivers (Prompt and Embedding Drivers, Vector Store Drivers, Image Generation Drivers, Image Query Drivers, SQL Drivers, Web Scraper Drivers, and Conversation Memory Drivers), Engines (Query Engines, Extraction Engines, Summary Engines, Image Generation Engines, and Image Query Engines), and additional components (Rulesets, Loaders, Artifacts, Chunkers, and Tokenizers). Griptape enables developers to create AI-powered applications with ease and efficiency.