A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training
☆650Feb 27, 2026Updated last week
Alternatives and similar repositories for MagiAttention
Users that are interested in MagiAttention are comparing it to the libraries listed below
Sorting:
- MAGI-1: Autoregressive Video Generation at Scale☆3,643Jun 17, 2025Updated 8 months ago
- Distributed Compiler based on Triton for Parallel Systems☆1,371Feb 13, 2026Updated 3 weeks ago
- A fast communication-overlapping library for tensor/expert parallelism on GPUs.☆1,264Aug 28, 2025Updated 6 months ago
- ☆65Apr 26, 2025Updated 10 months ago
- xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism☆2,549Feb 26, 2026Updated last week
- Ring attention implementation with flash attention☆987Sep 10, 2025Updated 5 months ago
- Quantized Attention on GPU☆44Nov 22, 2024Updated last year
- 🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"☆969Feb 5, 2026Updated last month
- [ICCV 2025] Official repo for "GigaTok: Scaling Visual Tokenizers to 3 Billion Parameters for Autoregressive Image Generation"☆198Jan 7, 2026Updated last month
- USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference☆644Jan 15, 2026Updated last month
- 🚀 Efficient implementations of state-of-the-art linear attention models☆4,474Updated this week
- ☆52May 19, 2025Updated 9 months ago
- [ICML 2025] XAttention: Block Sparse Attention with Antidiagonal Scoring☆269Jul 6, 2025Updated 8 months ago
- Helpful tools and examples for working with flex-attention☆1,140Feb 8, 2026Updated 3 weeks ago
- A unified inference and post-training framework for accelerated video generation.☆3,111Updated this week
- DeeperGEMM: crazy optimized version☆74May 5, 2025Updated 10 months ago
- [ICML2025, NeurIPS2025 Spotlight] Sparse VideoGen 1 & 2: Accelerating Video Diffusion Transformers with Sparse Attention☆629Feb 3, 2026Updated last month
- Tile primitives for speedy kernels☆3,202Feb 24, 2026Updated last week
- [ICML2025] SpargeAttention: A training-free sparse attention that accelerates any model inference.☆952Feb 25, 2026Updated last week
- VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo☆1,676Updated this week
- DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling☆21Feb 9, 2026Updated 3 weeks ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆147May 10, 2025Updated 9 months ago
- Accelerate LLM preference tuning via prefix sharing with a single line of code☆51Jul 4, 2025Updated 8 months ago
- Perplexity GPU Kernels☆567Nov 7, 2025Updated 3 months ago
- A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on H…☆3,176Updated this week
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆239Jun 15, 2025Updated 8 months ago
- [ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-t…☆3,192Jan 17, 2026Updated last month
- FlashInfer: Kernel Library for LLM Serving☆5,057Updated this week
- flex-block-attn: an efficient block sparse attention computation library☆124Dec 26, 2025Updated 2 months ago
- ☆118May 19, 2025Updated 9 months ago
- ☆262Jul 11, 2024Updated last year
- Muon is Scalable for LLM Training☆1,440Aug 3, 2025Updated 7 months ago
- https://wavespeed.ai/ Context parallel attention that accelerates DiT model inference with dynamic caching☆424Jul 5, 2025Updated 8 months ago
- Dynamic Memory Management for Serving LLMs without PagedAttention☆464May 30, 2025Updated 9 months ago
- [ICLR 2025] Autoregressive Video Generation without Vector Quantization☆629Oct 29, 2025Updated 4 months ago
- ☆38Aug 7, 2025Updated 6 months ago
- ByteCheckpoint: An Unified Checkpointing Library for LFMs☆270Feb 2, 2026Updated last month
- Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels☆5,284Updated this week
- FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.☆1,025Sep 4, 2024Updated last year