Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.

Kernel Library Wheel for SGLang

a powerful, large-scale, multimodal model for Text-to-Image generation.

⚡ Optimize attention mechanisms with FlashMLA, a library of advanced sparse and dense kernels for DeepSeek models, improving performance and efficiency.

🚀 Accelerate attention mechanisms with FlashMLA, featuring optimized kernels for DeepSeek models, enhancing performance through sparse and dense attention.