xlite-dev / ffpa-attnLinks
🤖FFPA: Extend FlashAttention-2 with Split-D, ~O(1) SRAM complexity for large headdim, 1.8x~3x↑🎉 vs SDPA EA.
☆211Updated 2 weeks ago
Alternatives and similar repositories for ffpa-attn
Users that are interested in ffpa-attn are comparing it to the libraries listed below
Sorting:
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆100Updated 3 months ago
- Implement Flash Attention using Cute.☆94Updated 8 months ago
- ☆97Updated 11 months ago
- ☆92Updated 5 months ago
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆183Updated 6 months ago
- Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.☆40Updated 5 months ago
- A collection of memory efficient attention operators implemented in the Triton language.☆278Updated last year
- ☆100Updated 3 months ago
- Examples of CUDA implementations by Cutlass CuTe☆219Updated last month
- ☆146Updated 5 months ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆111Updated 11 months ago
- A lightweight design for computation-communication overlap.☆157Updated last week
- flash attention tutorial written in python, triton, cuda, cutlass☆404Updated 3 months ago
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆74Updated last year
- ☆61Updated 4 months ago
- A Easy-to-understand TensorOp Matmul Tutorial☆372Updated 11 months ago
- ☆53Updated last month
- ☆91Updated last week
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆207Updated this week
- An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).☆261Updated last month
- High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.☆113Updated last year
- ☆229Updated last year
- QQQ is an innovative and hardware-optimized W4A8 quantization solution for LLMs.☆138Updated this week
- DeeperGEMM: crazy optimized version☆71Updated 3 months ago
- ☆138Updated last year
- ☆150Updated 7 months ago
- Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.☆40Updated 2 months ago
- ☆171Updated 2 years ago
- [DAC'25] Official implement of "HybriMoE: Hybrid CPU-GPU Scheduling and Cache Management for Efficient MoE Inference"☆66Updated 2 months ago
- Benchmark code for the "Online normalizer calculation for softmax" paper☆98Updated 7 years ago