A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training
☆952Oct 13, 2026Updated this week
Alternatives and similar repositories for MagiAttention
Users that are interested in MagiAttention are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.
Sorting:
- A plug-and-play compiler that delivers free-lunch optimizations for both inference and training.☆334Oct 1, 2026Updated last week
- Accelerating MoE with IO and Tile-aware Optimizations☆779Aug 29, 2026Updated last month
- Distributed Compiler and Optimized Parallel Kernels☆1,566Sep 18, 2026Updated 3 weeks ago
- MAGI-1: Autoregressive Video Generation at Scale☆3,795Jun 17, 2026Updated 3 months ago
- VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo☆2,236Updated this week
- Deploy to Railway using AI coding agents - Free Credits Offer • AdUse Claude Code, Codex, OpenCode, and more. Autonomous software development now has the infrastructure to match with Railway.
- A fast communication-overlapping library for tensor/expert parallelism on GPUs.☆1,371Aug 28, 2025Updated last year
- ☆255Nov 19, 2025Updated 10 months ago
- [ICML2025, NeurIPS2025 Spotlight] Sparse VideoGen 1 & 2: Accelerating Video Diffusion Transformers with Sparse Attention☆709Jul 4, 2026Updated 3 months ago
- A unified inference and post-training framework for accelerated video generation.☆4,596Updated this week
- A Quirky Assortment of CuTe Kernels☆1,173Sep 17, 2026Updated 3 weeks ago
- [ICML2025] SpargeAttention: A training-free sparse attention that accelerates any model inference.☆1,260Feb 25, 2026Updated 7 months ago
- Ring attention implementation with flash attention☆1,065Sep 10, 2025Updated last year
- 🚀 Efficient implementations for emerging model architectures☆5,855Updated this week
- USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference☆696Sep 16, 2026Updated 3 weeks ago
- Managed Kubernetes at scale on DigitalOcean • AdDigitalOcean Kubernetes includes the control plane, bandwidth allowance, container registry, automatic updates, and more for free.
- A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point (FP8 and FP4) precision on H…☆3,576Updated this week
- xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism☆2,737Updated this week
- Helpful tools and examples for working with flex-attention☆1,256Updated this week
- 🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"☆1,023Feb 5, 2026Updated 8 months ago
- FlashInfer: Kernel Library for LLM Serving☆6,594Updated this week
- [ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-t…☆3,968Jan 17, 2026Updated 8 months ago
- ☆37Aug 7, 2025Updated last year
- Tile primitives for speedy kernels☆3,749Sep 12, 2026Updated last month
- [NeurIPS 2025] Radial Attention: O(nlogn) Sparse Attention with Energy Decay for Long Video Generation☆607Nov 11, 2025Updated 11 months ago
- Open source password manager - Proton Pass • AdSecurely store, share, and autofill your credentials with Proton Pass, the end-to-end encrypted password manager trusted by millions.
- flex-block-attn: an efficient block sparse attention computation library☆132Dec 26, 2025Updated 9 months ago
- An efficient implementation of the NSA (Native Sparse Attention) kernel☆135Jun 24, 2025Updated last year
- A sparse attention kernel supporting mix sparse patterns☆560Jan 18, 2026Updated 8 months ago
- DeepSeek-V3.2-Exp DSA Warmup Lightning Indexer training operator based on tilelang☆52Nov 19, 2025Updated 10 months ago
- CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.☆555Updated this week
- Perplexity GPU Kernels☆608Nov 7, 2025Updated 11 months ago
- high-performance linear attention kernel library built on TileLang☆724Sep 30, 2026Updated last week
- NVSHMEM‑Tutorial: Build a DeepEP‑like GPU Buffer☆206Feb 11, 2026Updated 8 months ago
- Kernel Library for Large Headdim Attention (64~1024, BF16/FP8/FP4), 1.5x~15x↑ vs PyTorch SDPA.☆340Updated this week
- End-to-end encrypted email - Proton Mail • AdSpecial offer: 40% Off Yearly / 80% Off First Month. All Proton services are open source and independently audited for security.
- Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels☆8,594Updated this week
- A Survey of Efficient Attention Methods: Hardware-efficient, Sparse, Compact, and Linear Attention☆312Dec 1, 2025Updated 10 months ago
- Allow torch tensor memory to be released and resumed later☆279Sep 29, 2026Updated 2 weeks ago
- Efficient triton implementation of Native Sparse Attention.☆283May 23, 2025Updated last year
- Kernels, of the mega variety :)☆847May 26, 2026Updated 4 months ago
- A PyTorch-native inference engine with cache, parallelism, quantization and cpu offload for DiTs.☆1,291Updated this week
- Distributed MoE in a Single Kernel [NeurIPS '25]☆310May 5, 2026Updated 5 months ago