vllm-project / flash-attentionLinks
Fast and memory-efficient exact attention
☆95Updated this week
Alternatives and similar repositories for flash-attention
Users that are interested in flash-attention are comparing it to the libraries listed below
Sorting:
- ☆95Updated 6 months ago
- Compare different hardware platforms via the Roofline Model for LLM inference tasks.☆115Updated last year
- ☆98Updated last year
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆118Updated 4 months ago
- A lightweight design for computation-communication overlap.☆177Updated 2 weeks ago
- A prefill & decode disaggregated LLM serving framework with shared GPU memory and fine-grained compute isolation.☆111Updated 4 months ago
- An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).☆265Updated 2 months ago
- NVSHMEM‑Tutorial: Build a DeepEP‑like GPU Buffer☆130Updated 2 weeks ago
- ☆147Updated 6 months ago
- ☆88Updated last week
- AI Accelerator Benchmark focuses on evaluating AI Accelerators from a practical production perspective, including the ease of use and ver…☆266Updated last month
- Dynamic Memory Management for Serving LLMs without PagedAttention☆421Updated 4 months ago
- Pipeline Parallelism Emulation and Visualization☆67Updated 3 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆230Updated this week
- PyTorch bindings for CUTLASS grouped GEMM.☆121Updated 4 months ago
- 🤖FFPA: Extend FlashAttention-2 with Split-D, ~O(1) SRAM complexity for large headdim, 1.8x~3x↑🎉 vs SDPA EA.☆219Updated last month
- DeeperGEMM: crazy optimized version☆71Updated 4 months ago
- ☆90Updated 10 months ago
- High performance Transformer implementation in C++.☆134Updated 8 months ago
- ☆298Updated last week
- ☆98Updated 4 months ago
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆186Updated 8 months ago
- Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity☆221Updated 2 years ago
- A collection of memory efficient attention operators implemented in the Triton language.☆279Updated last year
- [DAC'25] Official implement of "HybriMoE: Hybrid CPU-GPU Scheduling and Cache Management for Efficient MoE Inference"☆72Updated 3 months ago
- Train speculative decoding models effortlessly and port them smoothly to SGLang serving.☆412Updated this week
- Standalone Flash Attention v2 kernel without libtorch dependency☆110Updated last year
- A low-latency & high-throughput serving engine for LLMs☆422Updated 4 months ago
- ☆64Updated 5 months ago
- Allow torch tensor memory to be released and resumed later☆142Updated last week