AI工程師、機器學習工程師、深度學習工程師、資料科學家、Machine Learning Engineer、Deep Learning Engineer、Data Scientist
parameter training volume to 2% , decreasing memory usage by 70%, and increasing training speed by 25%. Architecture Optimization: Used Direct Preference Optimization (DPO) to optimize Reinforcement Learning, improving model efficiency and decision quality. [ G itHub ] Graduate Research Assistant OctOct 2023 Institute of Information Science, Academia Sinica, Taiwan Natural Language and Knowledge Processing Lab (NLP Lab) Publication: Published in ACM CIKMSequential Text-based Knowledge Update with Self-Supervised Learning for Generative Language Models . [ Paper | GitHub ] Performance has improved by 18.8% compared to traditional methods in the past and surpassed SOTA LLMs by 39.6%, e
國立政治大學(National Chengchi University)・
資訊科學系