MoonshotAI / FlashKDA
FlashKDA: high-performance Kimi Delta Attention kernels
See what the GitHub community is most excited about today.
FlashKDA: high-performance Kimi Delta Attention kernels
Instant neural graphics primitives: lightning fast NeRF and more
Mirage Persistent Kernel: Compiling LLMs into a MegaKernel
DeepEP: an efficient expert-parallel communication library
NCCL Tests
RAPIDS Accelerator JNI For Apache Spark
Causal depthwise conv1d in CUDA, with a PyTorch interface
Tile primitives for speedy kernels
This package contains the original 2012 AlexNet code.
GPU accelerated decision optimization
DeepGEMM: clean and efficient BLAS kernel library on GPU
[ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-to-end metrics across language, image, and video models.
LLM training in simple, raw C/CUDA
CUDA Kernel Benchmarking Library
how to optimize some algorithm in cuda.
[ARCHIVED] Cooperative primitives for CUDA C++. See https://github.com/NVIDIA/cccl
cuVS - a library for vector search and clustering on the GPU