D-NAT: Data-Driven Non-Ideality Aware Training Framework for Fabricated Computing-In-Memory Macros
- 29 April 2022
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. 12 (2) , 381-392
Abstract
To enable energy-efficient computation for deep neural networks (DNNs) at edge, computing-in-memory (CIM) is proposed to reduce the energy costs during intense off-chip memory access. However, CIM is prone to multiply-accumulate (MAC) errors due to non-idealities of memory crossbars and peripheral circuits, which severely degrade the accuracy of DNNs. In this work, we propose a Data-Driven Non-ideality Aware Training (D-NAT) framework to compensate for the accuracy degradation. The proposed D-NAT framework has the following contributions: 1) We measured a fabricated SRAM-based CIM macro to obtain a data-driven MAC error model (D-MAC-EM). Based on the derived D-MAC-EM, we analyze the impact of the non-idealities on DNN’s accuracy. 2) To make DNNs robust to the non-idealities of CIM macros, we incorporate the measured D-MAC-EM into DNN’s training procedure. Moreover, we propose a statistical training mechanism to better estimate the gradients of the discrete D-MAC-EM. 3) We investigate trade-offs between quantization range and quantization errors. To mitigate the quantization errors in activations, we introduce extended PACT (E-PACT) that adaptively learns the upper and lower bounds of input activations for each layer. Simulation results show that our proposed D-NAT improves the accuracy of ResNet20, VGG8, ResNet34, and VGG16 by 78.98%, 71.8%, 72.04%, and 57.85%, respectively, which reaches the ideal upper bound of the quantized model. Lastly, the D-NAT framework is validated on an FPGA platform with the fabricated SRAM-based CIM macro chip. Based on the measurement results, D-NAT successfully recovers the accuracy under non-idealities of a real SRAM-based CIM macro.Keywords
Funding Information
- Ministry of Science and Technology of Taiwan (MOST 110-2218-E-002-034-MBK, MOST 111-2218-E-002-018-MBK)
- MediaTek Inc. (MTKC-2021-0167)
This publication has 23 references indexed in Scilit:
- CxDNNACM Transactions on Embedded Computing Systems, 2019
- In-Memory Computing: Advances and ProspectsIEEE Solid-State Circuits Magazine, 2019
- Resistive Memory‐Based In‐Memory Computing: From Device and Large‐Scale Integration System PerspectivesAdvanced Intelligent Systems, 2019
- Impact of Non-Ideal Characteristics of Resistive Synaptic Devices on Implementing Convolutional Neural NetworksIEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2019
- Stochastic Data-driven Hardware Resilience to Efficiently Train Inference Models for Stochastic Hardware Implementations2019
- Design of Reliable DNN Accelerator with Un-reliable ReRAM2019
- Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference2018
- In-Memory Computation of a Machine-Learning Classifier in a Standard 6T SRAM ArrayIEEE Journal of Solid-State Circuits, 2017
- Analysis of Nonideal Behaviors Based on INL/DNL Plots for SAR ADCsIEEE Transactions on Instrumentation and Measurement, 2016
- A 10-bit 50-MS/s SAR ADC With a Monotonic Capacitor Switching ProcedureIEEE Journal of Solid-State Circuits, 2010