A massively parallel, optimal functional runtime in Rust

DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling

[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.

Sample codes for my CUDA programming book

[ICML2025] SpargeAttention: A training-free sparse attention that accelerates any model inference.

Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.