Jinye Du, Quan Yuan, Zuyao Zhang, Yanzhi Yi, Jiahui Hu, Wangyi Chen, Yiyang Zhu, Qishui Zheng, Wenxiang Zou, Xiangyu Chang, Zuohe Zheng, Zichun Ye, Chao Liu, Shanni Li, Renwei Zhang, Yiping Deng, Xinwei Hu, Xuefeng Jin, Jie Zhao
The AKG kernel agent is a multi-agent system that automates the generation and tuning of computation kernels for various hardware platforms, achieving significant speedups in AI workloads.
As AI models become more complex, the need for efficient computation increases, particularly with the diverse hardware and frequent updates in the field. Manually optimizing these computation kernels for different platforms is becoming a bottleneck. The AKG kernel agent addresses this by using AI to automate the creation and tuning of these kernels, supporting multiple programming languages to ensure compatibility with various hardware. This system has shown to improve performance significantly compared to traditional methods.
