Recently, Vision Transformer (ViT) has continuously established new milestones in the computer vision field, while the high computation and memory cost makes its propagation in industrial production difficult. Pruning, a traditional model compression paradigm for hardware efficiency, has been widely applied in various DNN structures. Nevertheless, it stays ambiguous on how to perform exclusive pruning on the ViT structure. Considering three key points: the structural characteristics, the internal data pattern of ViTs, and the related edge device deployment, we leverage the input token sparsity and propose a computation-aware soft pruning framework, which can be set up on vanilla Transformers of both flatten and CNN-type structures, such as Pooling-based ViT (PiT). More concretely, we design a dynamic attention-based multi-head token selector, which is a lightweight module for adaptive instance-wise token selection. We further introduce a soft pruning technique, which integrates the less informative tokens generated by the selector module into a package token that will participate in subsequent calculations rather than being completely discarded. Our framework is bound to the trade-off between accuracy and computation constraints of specific edge devices through our proposed computation-aware training strategy. Experimental results show that our framework significantly reduces the computation cost of ViTs while maintaining comparable performance on image classification. Moreover, our framework can guarantee the identified model to meet resource specifications of mobile devices and FPGA, and even achieve the real-time execution of DeiT-T on mobile platforms. For example, our method reduces the latency of DeiT-T to 26 ms (26%$\sim $41% superior to existing works) on the mobile device with 0.25%$\sim $4% higher top-1 accuracy on ImageNet. Our code will be released soon.
This work targets the commonly used FPGA (field-programmable gate array) devices as the hardware platform for DNN edge computing. We focus on DNN quantization as the main model compression technique. The novelty of this work is: We use a quantization method that supports multiple precisions along the intra-layer dimension, while the existing quantization methods apply multi-precision quantization along the inter-layer dimension. The intra-layer multi-precision method can uniform the hardware configurations for different layers to reduce computation overhead and at the same time preserve the model accuracy as the inter-layer approach. Our proposed ILMPQ DNN quantization framework achieves 70.73 Top1 accuracy in ResNet-18 on the ImageNet dataset. We also validate the proposed MSP framework on two FPGA devices i.e., Xilinx XC7Z020 and XC7Z045. We achieve 3.65x speedup in end-to-end inference time on the ImageNet, compared with the fixed-point quantization method.
This work proposes a novel Deep Neural Network (DNN) quantization framework, namely RMSMP, with a Row-wise Mixed-Scheme and Multi-Precision approach. Specifically, this is the first effort to assign mixed quantization schemes and multiple precisions within layers -- among rows of the DNN weight matrix, for simplified operations in hardware inference, while preserving accuracy. Furthermore, this paper makes a different observation from the prior work that the quantization error does not necessarily exhibit the layer-wise sensitivity, and actually can be mitigated as long as a certain portion of the weights in every layer are in higher precisions. This observation enables layer-wise uniformality in the hardware implementation towards guaranteed inference acceleration, while still enjoying row-wise flexibility of mixed schemes and multiple precisions to boost accuracy. The candidates of schemes and precisions are derived practically and effectively with a highly hardware-informative strategy to reduce the problem search space. With the offline determined ratio of different quantization schemes and precisions for all the layers, the RMSMP quantization algorithm uses the Hessian and variance-based method to effectively assign schemes and precisions for each row. The proposed RMSMP is tested for the image classification and natural language processing (BERT) applications and achieves the best accuracy performance among state-of-the-arts under the same equivalent precisions. The RMSMP is implemented on FPGA devices, achieving 3.65x speedup in the end-to-end inference time for ResNet-18 on ImageNet, compared with the 4-bit Fixed-point baseline.
Deep Neural Networks (DNNs) have achieved extraordinary performance in various application domains. To support diverse DNN models, efficient implementations of DNN inference on edge-computing platforms, e.g., ASICs, FPGAs, and embedded systems, are extensively investigated. Due to the huge model size and computation amount, model compression is a critical step to deploy DNN models on edge devices. This paper focuses on weight quantization, a hardware-friendly model compression approach that is complementary to weight pruning. Unlike existing methods that use the same quantization scheme for all weights, we propose the first solution that applies different quantization schemes for different rows of the weight matrix. It is motivated by (1) the distribution of the weights in the different rows are not the same; and (2) the potential of achieving better utilization of heterogeneous FPGA hardware resources. To achieve that, we first propose a hardware-friendly quantization scheme named sum-of-power-of-2 (SP2) suitable for Gaussian-like weight distribution, in which the multiplication arithmetic can be replaced with logic shifter and adder, thereby enabling highly efficient implementations with the FPGA LUT resources. In contrast, the existing fixed-point quantization is suitable for Uniform-like weight distribution and can be implemented efficiently by DSP. Then to fully explore the resources, we propose an FPGA-centric mixed scheme quantization (MSQ) with an ensemble of the proposed SP2 and the fixed-point schemes. Combining the two schemes can maintain, or even increase accuracy due to better matching with weight distributions.
With the tremendous success of deep learning, there exists imminent need to deploy deep learning models onto edge devices. To tackle the limited computing and storage resources in edge devices, model compression techniques have been widely used to trim deep neural network (DNN) models for on-device inference execution. This paper targets the commonly used FPGA (field programmable gate array) devices as the hardware platforms for DNN edge computing. We focus on the DNN quantization as the main model compression technique, since DNN quantization has been of great importance for the implementations of DNN models on the hardware platforms. The novelty of this work comes in twofold: (i) We propose a mixed-scheme DNN quantization method that incorporates both the linear and non-linear number systems for quantization, with the aim to boost the utilization of the heterogeneous computing resources, i.e., LUTs (look up tables) and DSPs (digital signal processors) on an FPGA. Note that all the existing (single-scheme) quantization methods can only utilize one type of resources (either LUTs or DSPs for the MAC (multiply-accumulate) operations in deep learning computations. (ii) We use a quantization method that supports multiple precisions along the intra-layer dimension, while the existing quantization methods apply multi-precision quantization along the inter-layer dimension. The intra-layer multi-precision method can uniform the hardware configurations for different layers to reduce computation overhead and at the same time preserve the model accuracy as the inter-layer approach.
Mobile devices are becoming an important carrier for deep learning tasks, as they are being equipped with powerful, high-end mobile CPUs and GPUs. However, it is still a challenging task to execute 3D Convolutional Neural Networks (CNNs) targeting for real-time performance, besides high inference accuracy. The reason is more complex model structure and higher model dimensionality overwhelm the available computation/storage resources on mobile devices. A natural way may be turning to deep learning weight pruning techniques. However, the direct generalization of existing 2D CNN weight pruning methods to 3D CNNs is not ideal for fully exploiting mobile parallelism while achieving high inference accuracy. This paper proposes RT3D, a model compression and mobile acceleration framework for 3D CNNs, seamlessly integrating neural network weight pruning and compiler code generation techniques. We propose and investigate two structured sparsity schemes i.e., the vanilla structured sparsity and kernel group structured (KGS) sparsity that are mobile acceleration friendly. The vanilla sparsity removes whole kernel groups, while KGS sparsity is a more fine-grained structured sparsity that enjoys higher flexibility while exploiting full on-device parallelism. We propose a reweighted regularization pruning algorithm to achieve the proposed sparsity schemes. The inference time speedup due to sparsity is approaching the pruning rate of the whole model FLOPs (floating point operations). RT3D demonstrates up to 29.1$\times$ speedup in end-to-end inference time comparing with current mobile frameworks supporting 3D CNNs, with moderate 1%-1.5% accuracy loss. The end-to-end inference time for 16 video frames could be within 150 ms, when executing representative C3D and R(2+1)D models on a cellphone. For the first time, real-time execution of 3D CNNs is achieved on off-the-shelf mobiles.
Graph Neural Networks (GNNs) have made significant advances on several fundamental inference tasks. As a result, there is a surge of interest in using these models for making potentially important decisions in high-regret applications. However, despite GNNs' impressive performance, it has been observed that carefully crafted perturbations on graph structures (or nodes attributes) lead them to make wrong predictions. Presence of these adversarial examples raises serious security concerns. Most of the existing robust GNN design/training methods are only applicable to white-box settings where model parameters are known and gradient based methods can be used by performing convex relaxation of the discrete graph domain. More importantly, these methods are not efficient and scalable which make them infeasible in time sensitive tasks and massive graph datasets. To overcome these limitations, we propose a general framework which leverages the greedy search algorithms and zeroth-order methods to obtain robust GNNs in a generic and an efficient manner. On several applications, we show that the proposed techniques are significantly less computationally expensive and, in some cases, more robust than the state-of-the-art methods making them suitable to large-scale problems which were out of the reach of traditional robust training methods.
Structured weight pruning is a representative model compression technique of DNNs for hardware efficiency and inference accelerations. Previous works in this area leave great space for improvement since sparse structures with combinations of different structured pruning schemes are not exploited fully and efficiently. To mitigate the limitations, we propose SS-Auto, a single-shot, automatic structured pruning framework that can achieve row pruning and column pruning simultaneously. We adopt soft constraint-based formulation to alleviate the strong non-convexity of l0-norm constraints used in state-of-the-art ADMM-based methods for faster convergence and fewer hyperparameters. Instead of solving the problem directly, a Primal-Proximal solution is proposed to avoid the pitfall of penalizing all weights equally, thereby enhancing the accuracy. Extensive experiments on CIFAR-10 and CIFAR-100 datasets demonstrate that the proposed framework can achieve ultra-high pruning rates while maintaining accuracy. Furthermore, significant inference speedup has been observed from the proposed framework through actual measurements on the smartphone.
It is known that deep neural networks (DNNs) are vulnerable to adversarial attacks. The so-called physical adversarial examples deceive DNN-based decision makers by attaching adversarial patches to real objects. However, most of the existing works on physical adversarial attacks focus on static objects such as glass frames, stop signs and images attached to cardboard. In this work, we propose Adversarial T-shirts, a robust physical adversarial example for evading person detectors even if it could undergo non-rigid deformation due to a moving person's pose changes. To the best of our knowledge, this is the first work that models the effect of deformation for designing physical adversarial examples with respect to non-rigid objects such as T-shirts. We show that the proposed method achieves 74% and 57% attack success rates in digital and physical worlds respectively against YOLOv2. In contrast, the state-of-the-art physical attack method to fool a person detector only achieves 18% attack success rate. Furthermore, by leveraging min-max optimization, we extend our method to the ensemble attack setting against two object detectors YOLO-v2 and Faster R-CNN simultaneously.
It is known that deep neural networks (DNNs) could be vulnerable to adversarial attacks. The so-called physical adversarial examples deceive DNN-based decision makers by attaching adversarial patches to real objects. However, most of the existing works on physical adversarial attacks focus on static objects such as glass frame, stop sign and image attached to a cardboard. In this work, we proposed adversarial T-shirt, a robust physical adversarial example for evading person detectors even if it suffers from deformation due toa moving person's pose change. To the best of our knowledge, the effect of deformation is first modeled for designing physical adversarial examples with respect to non-rigid objects such as T-shirts. We show that the proposed method achieves 79% and 63% attack success rates in digital and physical worlds respectively against YOLOv2. In contrast, the state-of-the-art physical attack method to fool a person detector only achieves 27% attack success rate. Furthermore, by leveraging min-max optimization, we extend our method to the ensemble attack setting against object detectors YOLOv2 and Faster R-CNN simultaneously.