all AI news
Transformers Get Stable: An End-to-End Signal Propagation Theory for Language Models
March 15, 2024, 4:42 a.m. | Akhil Kedia, Mohd Abbas Zaidi, Sushil Khyalia, Jungho Jung, Harshith Goka, Haejun Lee
cs.LG updates on arXiv.org arxiv.org
Abstract: In spite of their huge success, transformer models remain difficult to scale in depth. In this work, we develop a unified signal propagation theory and provide formulae that govern the moments of the forward and backward signal through the transformer model. Our framework can be used to understand and mitigate vanishing/exploding gradients, rank collapse, and instability associated with high attention scores. We also propose DeepScaleLM, an initialization and scaling scheme that conserves unit output/gradient moments …
abstract arxiv cs.ai cs.cl cs.cv cs.lg framework language language models moments propagation scale signal success theory through transformer transformer model transformer models transformers type work
More from arxiv.org / cs.LG updates on arXiv.org
Jobs in AI, ML, Big Data
Data Architect
@ University of Texas at Austin | Austin, TX
Data ETL Engineer
@ University of Texas at Austin | Austin, TX
Lead GNSS Data Scientist
@ Lurra Systems | Melbourne
Senior Machine Learning Engineer (MLOps)
@ Promaton | Remote, Europe
Senior Data Engineer
@ Quantexa | Sydney, New South Wales, Australia
Staff Analytics Engineer
@ Warner Bros. Discovery | NY New York 230 Park Avenue South