cutlass

📚200+ Tensor/CUDA Cores Kernels, ⚡️flash-attn-mma, ⚡️hgemm with WMMA, MMA and CuTe (98%~100% TFLOPS of cuBLAS/FA2 🎉🎉).

A fast communication-overlapping library for tensor/expert parallelism on GPUs.

#大语言模型#🔥🔥🔥 A collection of some awesome public CUDA, cuBLAS, cuDNN, CUTLASS, TensorRT, TensorRT-LLM, Triton, TVM, MLIR and High Performance Computing (HPC) projects.

Examples of CUDA implementations by Cutlass CuTe

CUTLASS and CuTe Examples

#大语言模型#Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.

#计算机科学#GEMM and Winograd based convolutions using CUTLASS

study of cutlass

#大语言模型#Multiple GEMM operators are constructed with cutlass to support LLM inference.

Kernel Library Wheel for SGLang

A cutlass cute implementation of headdim-64 flashattentionv2 TensorRT plugin for LightGlue. Run on Jetson Orin NX 8GB with TensorRT 8.5.2.

pytorch implements block sparse