Pinned Loading

1. Mini-Attention Mini-Attention Public

   FP16 Flash Attention 2 from scratch in CUDA C++ acheving 96% of CuDnn performance on SM120 (RTX 5090)

   Cuda 5

2. 100-days-cuda 100-days-cuda Public

   This repository documents my 100-day journey of learning and writing CUDA kernels.

   Jupyter Notebook 28 1

3. xDiT xDiT Public

   Forked from xdit-project/xDiT

   xDiT: A Scalable Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelism

   Python

4. HazyResearch/ThunderKittens HazyResearch/ThunderKittens Public

   Tile primitives for speedy kernels

   Cuda 3.3k 273

5. SageAttention SageAttention Public

   Forked from thu-ml/SageAttention

   [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.

   Cuda