SandAI-org / MagiAttention
A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training
☆351Updated this week
Alternatives and similar repositories for MagiAttention
Users that are interested in MagiAttention are comparing it to the libraries listed below
Sorting:
- ☆165Updated 4 months ago
- [ICML2025] Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity☆249Updated 2 weeks ago
- SpargeAttention: A training-free sparse attention that can accelerate any model inference.☆547Updated this week
- USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference☆492Updated 3 weeks ago
- VeOmni: Scaling any Modality Model Training to any Accelerators with PyTorch native Training Framework☆326Updated this week
- A sparse attention kernel supporting mix sparse patterns☆205Updated 3 months ago
- A parallelism VAE avoids OOM for high resolution image generation☆61Updated 3 months ago
- 📖A curated list of Awesome Diffusion Inference Papers with codes: Sampling, Caching, Multi-GPUs, etc. 🎉🎉☆232Updated last week
- Adaptive Caching for Faster Video Generation with Diffusion Transformers☆147Updated 6 months ago
- Model Compression Toolbox for Large Language Models and Diffusion Models☆463Updated last month
- XAttention: Block Sparse Attention with Antidiagonal Scoring☆147Updated this week
- [CVPR 2024 Highlight] DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models☆680Updated 5 months ago
- [ICLR 2025] COAT: Compressing Optimizer States and Activation for Memory-Efficient FP8 Training☆190Updated 3 weeks ago
- Ring attention implementation with flash attention☆764Updated last month
- 📚 Collection of awesome generation acceleration resources.☆235Updated 3 weeks ago
- Efficient triton implementation of Native Sparse Attention.☆147Updated this week
- An open-source implementation of Regional Adaptive Sampling (RAS), a novel diffusion model sampling strategy that introduces regional var…☆126Updated 2 months ago
- Patch convolution to avoid large GPU memory usage of Conv2D☆87Updated 3 months ago
- https://wavespeed.ai/ Context parallel attention that accelerates DiT model inference with dynamic caching☆271Updated last week
- 🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"☆665Updated last month
- [ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads☆459Updated 3 months ago
- 🔥 A minimal training framework for scaling FLA models☆128Updated last week
- [ICLR'25] ViDiT-Q: Efficient and Accurate Quantization of Diffusion Transformers for Image and Video Generation☆92Updated last month
- ByteCheckpoint: An Unified Checkpointing Library for LFMs☆209Updated last month
- An official implementation of Flow-GRPO: Training Flow Matching Models via Online RL☆526Updated this week
- 🎬 3.7× faster video generation E2E 🖼️ 1.6× faster image generation E2E ⚡ ColumnSparseAttn 9.3× vs FlashAttn‑3 💨 ColumnSparseGEMM 2.5× …☆58Updated last week
- Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs☆165Updated this week
- Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models☆595Updated last month
- Scaling Diffusion Transformers with Mixture of Experts☆321Updated 8 months ago
- [NeurIPS 2024] AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoising☆202Updated 2 months ago