skip to main content


Title: AccPar: Tensor Partitioning for Heterogeneous Deep Learning Accelerators
Deep neural network (DNN) accelerators as an example of domain-specific architecture have demonstrated great success in DNN inference. However, the architecture acceleration for equally important DNN training has not yet been fully studied. With data forward, error backward and gradient calculation, DNN training is a more complicated process with higher computation and communication intensity. Because the recent research demonstrates a diminishing specialization return, namely, “accelerator wall”, we believe that a promising approach is to explore coarse-grained parallelism among multiple performance-bounded accelerators to support DNN training. Distributing computations on multiple heterogeneous accelerators to achieve high throughput and balanced execution, however, remaining challenging. We present ACCPAR, a principled and systematic method of determining the tensor partition among heterogeneous accelerator arrays. Compared to prior empirical or unsystematic methods, ACCPAR considers the complete tensor partition space and can reveal previously unknown new parallelism configurations. ACCPAR optimizes the performance based on a cost model that takes into account both computation and communication costs of a heterogeneous execution environment. Hence, our method can avoid the drawbacks of existing approaches that use communication as a proxy of the performance. The enhanced flexibility of tensor partitioning in ACCPAR allows the flexible ratio of computations to be distributed among accelerators with different performances. The proposed search algorithm is also applicable to the emerging multi-path patterns in modern DNNs such as ResNet. We simulate ACCPAR on a heterogeneous accelerator array composed of both TPU-v2 and TPU-v3 accelerators for the training of large-scale DNN models such as Alexnet, Vgg series and Resnet series. The average performance improvements of the state-of-the-art “one weird trick” (OWT) and HYPAR, and ACCPAR, normalized to the baseline data parallelism scheme where each accelerator replicates the model and processes different input data in parallel, are 2.98×, 3.78×, and 6.30×, respectively.  more » « less
Award ID(s):
1910299 1717885
PAR ID:
10179973
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
2020 IEEE International Symposium on High Performance Computer Architecture (HPCA)
Page Range / eLocation ID:
342 to 355
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. In pursuit of higher inference accuracy, deep neural network (DNN) models have significantly increased in complexity and size. To overcome the consequent computational challenges, scalable chiplet-based accelerators have been proposed. However, data communication using metallic-based interconnects in these chiplet-based DNN accelerators is becoming a primary obstacle to performance, energy efficiency, and scalability. The photonic interconnects can provide adequate data communication support due to some superior properties like low latency, high bandwidth and energy efficiency, and ease of broadcast communication. In this paper, we propose SPACX: a Silicon Photonics-based Chiplet ACcelerator for DNN inference applications. Specifically, SPACX includes a photonic network design that enables seamless single-chiplet and cross-chiplet broadcast communications, and a tailored dataflow that promotes data broadcast and maximizes parallelism. Furthermore, we explore the broadcast granularities of the photonic network and implications on system performance and energy efficiency. A flexible bandwidth allocation scheme is also proposed to dynamically adjust communication bandwidths for different types of data. Simulation results using several DNN models show that SPACX can achieve 78% and 75% reduction in execution time and energy, respectively, as compared to other state-of-the-art chiplet-based DNN accelerators. 
    more » « less
  2. DNN training is extremely time-consuming, necessitating efficient multi-accelerator parallelization. Current approaches to parallelizing training primarily use intra-batch parallelization, where a single iteration of training is split over the available workers, but suffer from diminishing returns at higher worker counts. We present PipeDream, a system that adds inter-batch pipelining to intra-batch parallelism to further improve parallel training throughput, helping to better overlap computation with communication and reduce the amount of communication when possible. Unlike traditional pipelining, DNN training is bi-directional, where a forward pass through the computation graph is followed by a backward pass that uses state and intermediate data computed during the forward pass. Naïve pipelining can thus result in mismatches in state versions used in the forward and backward passes, or excessive pipeline flushes and lower hardware efficiency. To address these challenges, PipeDream versions model parameters for numerically correct gradient computations, and schedules forward and backward passes of different minibatches concurrently on different workers with minimal pipeline stalls. PipeDream also automatically partitions DNN layers among workers to balance work and minimize communication. Extensive experimentation with a range of DNN tasks, models, and hardware configurations shows that PipeDream trains models to high accuracy up to 5.3X faster than commonly used intra-batch parallelism techniques. 
    more » « less
  3. The AI boom is bringing a plethora of domain-specific architectures for Neural Network computations. Google’s Tensor Processing Unit (TPU), a Deep Neural Network (DNN) accelerator, has replaced the CPUs/GPUs in its data centers, claiming more than 15X rate of inference. However, the unprecedented growth in DNN workloads with the widespread use of AI services projects an increasing energy consumption of TPU based data centers. In this work, we parametrize the extreme hardware underutilization in TPU systolic array and propose UPTPU: an intelligent, dataflow adaptive power-gating paradigm to provide a staggering 3.5X - 6.5X energy efficiency to TPU for different input batch sizes. 
    more » « less
  4. In Deep Neural Network (DNN) applications, energy consumption and performance cost of moving data between memory hierarchy and computational units are significantly higher than that of the computation itself. Process-in-memory (PIM) architecture such as Hybrid Memory Cube (HMC), becomes an excellent candidate to improve the data locality for efficient DNN execution. However, it’s still hard to efficiently deploy large-scale matrix computation in DNN on HMC because of its coarse grained packet protocol. In this work, we propose NeuralHMC, the first HMC-based accelerator tailored for efficient DNN execution. Experimental results show that NeuralHMC reduces the data movement by 1.4x to 2.5x (depending on the DNN data reuse strategy) compared to Von Neumann architecture. Furthermore, compared to state-of-the-art PIM-based DNN accelerator, NeuralHMC can promisingly improve the system performance by 4.1x and reduces energy by 1.5x, on average. 
    more » « less
  5. Architecture reverse engineering has become an emerging attack against deep neural network (DNN) implemen- tations. Several prior works have utilized side-channel leakage to recover the model architecture while the an DNN is executing on a hardware acceleration platform. In this work, we target an open- source deep-learning accelerator, Versatile Tensor Accelerator (VTA), and utilize electromagnetic (EM) side-channel leakage to comprehensively learn the association between DNN architecture configurations and EM emanations. We also consider the holistic system – including the low-level tensor program code of the VTA accelerator on a Xilinx FPGA, and explore the effect of such low- level configurations on the EM leakage. Our study demonstrates that both the optimization and configuration of tensor programs will affect the EM side-channel leakage. Gaining knowledge of the association between low-level tensor program and the EM emanations, we propose NNReArch, a lightweight tensor program scheduling framework against side- channel-based DNN model architecture reverse engineering. Specifically, NNReArch targets reshaping the EM traces of different DNN operators, through scheduling the tensor program execution of the DNN model so as to confuse the adversary. NNReArch is a comprehensive protection framework supporting two modes, a balanced mode that strikes a balance between the DNN model confidentiality and execution performance, and a secure mode where the most secure setting is chosen. We imple- ment and evaluate the proposed framework on the open-source VTA with state-of-the-art DNN architectures. The experimental results demonstrate that NNReArch can efficiently enhance the model architecture security with a small performance overhead. In addition, the proposed obfuscation technique makes reverse engineering of the DNN architecture significantly harder. 
    more » « less