thu-ml / SageAttention
Quantized Attention that achieves speedups of 2.1x and 2.7x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.
☆370Updated this week
Related projects ⓘ
Alternatives and complementary repositories for SageAttention
- xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism☆677Updated this week
- Model Compression Toolbox for Large Language Models and Diffusion Models☆188Updated this week
- QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving☆438Updated this week
- 📖A small curated list of Awesome SD/DiT/ViT/Diffusion Inference with Distributed/Caching/Sampling: DistriFusion, PipeFusion, AsyncDiff, …☆89Updated 2 months ago
- Code for Neurips24 paper: QuaRot, an end-to-end 4-bit inference of large language models.☆282Updated 3 months ago
- [ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference☆196Updated last week
- [NeurIPS'23] H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models.☆388Updated 3 months ago
- [MLSys'24] Atom: Low-bit Quantization for Efficient and Accurate LLM Serving☆277Updated 4 months ago
- [CVPR 2024 Highlight] DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models☆588Updated last week
- Ring attention implementation with flash attention☆578Updated this week
- USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference☆353Updated last week
- [EMNLP 2024 Industry Track] This is the official PyTorch implementation of "LLMC: Benchmarking Large Language Model Quantization with a V…☆315Updated this week
- DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads☆348Updated last week
- ☆97Updated last month
- EfficientQAT: Efficient Quantization-Aware Training for Large Language Models☆222Updated last month
- flash attention tutorial written in python, triton, cuda, cutlass☆195Updated 4 months ago
- ☆285Updated 7 months ago
- Dynamic Memory Management for Serving LLMs without PagedAttention☆228Updated this week
- A collection of memory efficient attention operators implemented in the Triton language.☆217Updated 5 months ago
- BitBLAS is a library to support mixed-precision matrix multiplications, especially for quantized LLM deployment.☆410Updated this week
- [ICML 2024] KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache☆241Updated last month
- SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models☆246Updated this week
- FlagGems is an operator library for large language models implemented in Triton Language.☆335Updated this week
- A fast communication-overlapping library for tensor parallelism on GPUs.☆219Updated 2 weeks ago
- Mirage: Automatically Generating Fast GPU Kernels without Programming in Triton/CUDA☆601Updated last week
- An easy-to-use package for implementing SmoothQuant for LLMs☆82Updated 5 months ago
- An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).☆196Updated 2 weeks ago
- ☆183Updated 6 months ago
- Analyze the inference of Large Language Models (LLMs). Analyze aspects like computation, storage, transmission, and hardware roofline mod…☆310Updated 2 months ago
- [NeurIPS 2024] KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization☆303Updated 2 months ago