BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20240626T180034Z
LOCATION:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20240625T180000
DTEND;TZID=America/Los_Angeles:20240625T190000
UID:dac_DAC 2024_sess236_RESEARCH1587@linklings.com
SUMMARY:Additive Partial Sum Quantization
DESCRIPTION:Work-in-Progress Poster\n\nPingcheng Dong, Yonghao Tan, Dong Z
 hang, Yongkun Wu, Xijie Huang, and Shi-Yang Liu (Hong Kong University of S
 cience and Technology (HKUST)); Yu Liu, Xuejiao Liu, Peng Luo, and Luhong 
 Liang (AI Chip Center for Emerging Smart System (ACCESS)); Fengwei An (Sou
 thern University of Science and Technology); and Kwang-Ting Cheng (Hong Ko
 ng University of Science and Technology (HKUST))\n\nDNN accelerators, sign
 ificantly advanced by model compression and specialized dataflow technique
 s, have marked considerable progress. However, the frequent access of high
 -precision partial sums (PSUMs) leads to excessive memory demands in archi
 tectures utilizing weight/input stationary dataflows. Traditional compress
 ion strategies have typically overlooked PSUM quantization, a gap recently
  explored in compute-in-memory research. Moreover, these approaches are ma
 inly toward reducing the Analog-to-Digital Converter (ADC) overhead, negle
 cting the critical issue of intensive memory access. This study introduces
  a novel Additive Partial Sum Quantization (APSQ) method, seamlessly integ
 rating PSUM accumulation into the quantization framework. We further propo
 se a grouping strategy that combines APSQ with PSQ enhanced by a floating-
 point regularization technique to boost accuracy. The experiments indicate
  that APSQ can efficiently compress PSUMs to INT-8, incurring a negligible
  degradation of accuracy for the Segformer-B0 and EfficientViT-B0 on the c
 hallenging Cityscapes dataset. This leads to a notable reduction in energy
  costs by 30~45%.\n\nTopic: AI, Autonomous Systems, Cloud, Design, EDA, Em
 bedded Systems, IP, Security
END:VEVENT
END:VCALENDAR