Mirage Persistent Kernel: Compiling LLMs into a MegaKernel
[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.
how to optimize some algorithm in cuda.
RTP-LLM: Alibaba's high-performance LLM inference engine for diverse applications.
RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing high performance applications.