No Arabic abstract
Quantization is spearheading the increase in performance and efficiency of neural network computing systems making headway into commodity hardware. We present SWIS - Shared Weight bIt Sparsity, a quantization framework for efficient neural network inference acceleration delivering improved performance and storage compression through an offline weight decomposition and scheduling algorithm. SWIS can achieve up to 54.3% (19.8%) point accuracy improvement compared to weight truncation when quantizing MobileNet-v2 to 4 (2) bits post-training (with retraining) showing the strength of leveraging shared bit-sparsity in weights. SWIS accelerator gives up to 6x speedup and 1.9x energy improvement overstate of the art bit-serial architectures.
Mixed-precision quantization can potentially achieve the optimal tradeoff between performance and compression rate of deep neural networks, and thus, have been widely investigated. However, it lacks a systematic method to determine the exact quantization scheme. Previous methods either examine only a small manually-designed search space or utilize a cumbersome neural architecture search to explore the vast search space. These approaches cannot lead to an optimal quantization scheme efficiently. This work proposes bit-level sparsity quantization (BSQ) to tackle the mixed-precision quantization from a new angle of inducing bit-level sparsity. We consider each bit of quantized weights as an independent trainable variable and introduce a differentiable bit-sparsity regularizer. BSQ can induce all-zero bits across a group of weight elements and realize the dynamic precision reduction, leading to a mixed-precision quantization scheme of the original model. Our method enables the exploration of the full mixed-precision space with a single gradient-based optimization process, with only one hyperparameter to tradeoff the performance and compression. BSQ achieves both higher accuracy and higher bit reduction on various model architectures on the CIFAR-10 and ImageNet datasets comparing to previous methods.
Emerging resistive random-access memory (ReRAM) has recently been intensively investigated to accelerate the processing of deep neural networks (DNNs). Due to the in-situ computation capability, analog ReRAM crossbars yield significant throughput improvement and energy reduction compared to traditional digital methods. However, the power hungry analog-to-digital converters (ADCs) prevent the practical deployment of ReRAM-based DNN accelerators on end devices with limited chip area and power budget. We observe that due to the limited bit-density of ReRAM cells, DNN weights are bit sliced and correspondingly stored on multiple ReRAM bitlines. The accumulated current on bitlines resulted by weights directly dictates the overhead of ADCs. As such, bitwise weight sparsity rather than the sparsity of the full weight, is desirable for efficient ReRAM deployment. In this work, we propose bit-slice L1, the first algorithm to induce bit-slice sparsity during the training of dynamic fixed-point DNNs. Experiment results show that our approach achieves 2x sparsity improvement compared to previous algorithms. The resulting sparsity allows the ADC resolution to be reduced to 1-bit of the most significant bit-slice and down to 3-bit for the others bits, which significantly speeds up processing and reduces power and area overhead.
Compressing Deep Neural Network (DNN) models to alleviate the storage and computation requirements is essential for practical applications, especially for resource limited devices. Although capable of reducing a reasonable amount of model parameters, previous unstructured or structured weight pruning methods can hardly truly accelerate inference, either due to the poor hardware compatibility of the unstructured sparsity or due to the low sparse rate of the structurally pruned network. Aiming at reducing both storage and computation, as well as preserving the original task performance, we propose a generalized weight unification framework at a hardware compatible micro-structured level to achieve high amount of compression and acceleration. Weight coefficients of a selected micro-structured block are unified to reduce the storage and computation of the block without changing the neuron connections, which turns to a micro-structured pruning special case when all unified coefficients are set to zero, where neuron connections (hence storage and computation) are completely removed. In addition, we developed an effective training framework based on the alternating direction method of multipliers (ADMM), which converts our complex constrained optimization into separately solvable subproblems. Through iteratively optimizing the subproblems, the desired micro-structure can be ensured with high compression ratio and low performance degradation. We extensively evaluated our method using a variety of benchmark models and datasets for different applications. Experimental results demonstrate state-of-the-art performance.
Existing neural network verifiers compute a proof that each input is handled correctly under a given perturbation by propagating a convex set of reachable values at each layer. This process is repeated independently for each input (e.g., image) and perturbation (e.g., rotation), leading to an expensive overall proof effort when handling an entire dataset. In this work we introduce a new method for reducing this verification cost based on the key insight that convex sets obtained at intermediate layers can overlap across different inputs and perturbations. Leveraging this insight, we introduce the general concept of shared certificates, enabling proof effort reuse across multiple inputs and driving down overall verification costs. We validate our insight via an extensive experimental evaluation and demonstrate the effectiveness of shared certificates on a range of datasets and attack specifications including geometric, patch and $ell_infty$ input perturbations.
Accelerating the inference speed of CNNs is critical to their deployment in real-world applications. Among all the pruning approaches, those implementing a sparsity learning framework have shown to be effective as they learn and prune the models in an end-to-end data-driven manner. However, these works impose the same sparsity regularization on all filters indiscriminately, which can hardly result in an optimal structure-sparse network. In this paper, we propose a Saliency-Adaptive Sparsity Learning (SASL) approach for further optimization. A novel and effective estimation of each filter, i.e., saliency, is designed, which is measured from two aspects: the importance for the prediction performance and the consumed computational resources. During sparsity learning, the regularization strength is adjusted according to the saliency, so our optimized format can better preserve the prediction performance while zeroing out more computation-heavy filters. The calculation for saliency introduces minimum overhead to the training process, which means our SASL is very efficient. During the pruning phase, in order to optimize the proposed data-dependent criterion, a hard sample mining strategy is utilized, which shows higher effectiveness and efficiency. Extensive experiments demonstrate the superior performance of our method. Notably, on ILSVRC-2012 dataset, our approach can reduce 49.7% FLOPs of ResNet-50 with very negligible 0.39% top-1 and 0.05% top-5 accuracy degradation.