Novel Adaptive Quantization Methodology for 8-bit Floating-Point DNN Training - Department of Electrical Engineering and Information Technology at RPTU

Division of Microelectronic Systems Design (EMS)

Novel Adaptive Quantization Methodology for 8-bit Floating-Point DNN Training
M. H. Sadi, C. Sudarshan, N. Wehn. Design Automation for Embedded Systems, Springer Journal, Volume 28, pages 91–110, (2024)
Link *)

*) This paper is supported by European Union’s Horizon Europe research and innovation program (HORIZON-CL4- 2021-HUMAN-01) under grant agreement No 101070408, project SustainML.