Zhengning Liu
Non-Ten Technology CTO,Head of Fitten Code AI Programming Assistant
Zhengning Liu received his Ph.D. degree from Tsinghua University, where he was engaged in the research and development of large model systems, deep learning frameworks, 3D computer vision and computer graphics. He has published articles in ACM TOG, IEEE TVCG, CVMJ, ECCV and other journals. Dr. Liu Zhengning is the technical director of Beijing Non-Ten Technology, launched Fitten Code AI programming assistant, and is the core developer of Jittor, a domestic deep learning framework, and is now dedicated to AI services and applications, as well as deep learning framework open source.
Topic
AI-assisted development based on the Mapping Framework
Code generation understanding based on large language models is forming a new software development paradigm. Based on the domestic deep learning framework Jittor, Non-Ten breaks through the difficulties of high training resource requirement, high inference delay and high deployment cost of large models. Thanks to Jittor's meta-arithmetic fusion technology, high-efficiency distributed computation, dynamic swap mechanism, etc., Jittor significantly improves the speed of training and inference, reduces the cost of deploying and applying the large models of code, and based on this, it provides a swift and accurate AI Code assistant service, compared with international similar services to shorten the delay time by about 70%, accuracy increased by about 20%.