QUIDAM: A Framework for <u>Qu</u> ant <u>i</u> zation-aware <u>D</u> NN <u>A</u> ccelerator and <u>M</u> odel Co-Exploration

نویسندگان

چکیده

As the machine learning and systems communities strive to achieve higher energy efficiency through custom deep neural network (DNN) accelerators, varied precision or quantization levels, model compression techniques, there is a need for design space exploration frameworks that incorporate quantization-aware processing elements into accelerator while having accurate fast power, performance, area models. In this work, we present QUIDAM , highly parameterized DNN co-exploration framework. Our framework can facilitate future research on of accelerators various choices such as bit precision, element type, scratchpad sizes elements, global buffer size, number total configurations. results show different precisions types lead significant differences in terms performance per energy. Specifically, our identifies wide range points where varies more than 5× 35×, respectively. With proposed framework, lightweight par accuracy up 5.7× improvement when compared best 16-bit integer quantization–based implementation. Finally, due pre-characterized models, speed process by three four orders magnitude it removes expensive synthesis characterization each design.

برای دانلود باید عضویت طلایی داشته باشید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

a framework for identifying and prioritizing factors affecting customers’ online shopping behavior in iran

the purpose of this study is identifying effective factors which make customers shop online in iran and investigating the importance of discovered factors in online customers’ decision. in the identifying phase, to discover the factors affecting online shopping behavior of customers in iran, the derived reference model summarizing antecedents of online shopping proposed by change et al. was us...

15 صفحه اول

passivity in waiting for godot and endgame: a psychoanalytic reading

this study intends to investigate samuel beckett’s waiting for godot and endgame under the lacanian psychoanalysis. it begins by explaining the most important concepts of lacanian psychoanalysis. the beckettian characters are studied regarding their state of unconscious, and not the state of consciousness as is common in most beckett studies. according to lacan, language plays the sole role in ...

A framework for reliability-aware design exploration on MPSoC based systems

Applying system-level fault-tolerant techniques such as active redundancy is a promising way to enhance the system reliability for safety-related applications. Embedded system design using active redundancy is a challenging task that involves solving two major problems, namely finding the optimal redundancy configuration and mapping/scheduling of the application (including the redundant compone...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

ژورنال

عنوان ژورنال: ACM Transactions in Embedded Computing Systems

سال: 2023

ISSN: ['1539-9087', '1558-3465']

DOI: https://doi.org/10.1145/3555807