Pre-trained language models (PLMs) are engineered to be robust in contextual understanding and exhibit outstanding performance in various natural language processing tasks. However, their considerable size incurs significant computational and storage costs. Modern pruning strategies employ retraining-free one-shot techniques to compress PLMs; however, these approaches often lead to an indispensable reduction in performance. In this paper, we propose SDS, a Sparse-Dense-Sparse pruning framework to enhance the performance of the pruned PLMs from a weight distribution optimization perspective. We outline the pruning process in three steps. Initially, we prune less critical connections in the model using conventional one-shot pruning methods. Next, we reconstruct a dense model featuring a pruning-friendly weight distribution by reactivating pruned connections with sparse regularization. Finally, we perform a second pruning round, yielding a superior pruned model compared to the initial pruning. Experiments demonstrate that SDS outperforms the state-of-the-art pruning techniques SparseGPT and Wanda under an identical sparsity configuration. For instance, SDS reduces perplexity by 5.16 on Raw-Wikitext2 and improves average accuracy by 3.86% across multiple zero-shot benchmarks for LLaMA-3-8B compared to Wanda with 2:4 sparsity.

One-shot post-training pruning enhances the deployment of billion-scale large language models (LLMs), with the pruning metric playing a pivotal role in determining which weights to remove. However, existing metrics underperform due to their reliance on a simple symbolic combination of weights and activations, overlooking imbalanced weight magnitudes and the disproportionate influence of activation outliers. To overcome these limitations, we introduce \textbf{BaWA}, a novel pruning metric that systematically \underline{Ba}lances \underline{W}eight and \underline{A}ctivation distributions for more effective pruning. BaWA introduces two key innovations: \textbf{magnitude normalization}, which mitigates weight imbalance across channels for fairer pruning decisions, and \textbf{outlier regularization}, which reduces the impact of activation outliers, ensuring more appropriate channel prioritization. To further enhance its effectiveness, BaWA incorporates an efficient and automatic framework for optimizing normalization and regularization hyperparameters. Extensive experiments validate BaWA as a state-of-the-art (SOTA) pruning metric. For instance, applying BaWA to induce 2:4 sparsity in Mistral-7B reduces perplexity in language comprehension by 2.49 and improves average downstream task accuracy by 3.08\%, outperforming the previous SOTA method Wanda.

Efficient deep learning models, especially optimized for edge devices, benefit from low inference latency to efficient energy consumption. Two classical techniques for efficient model inference are lightweight neural architecture search (NAS), which automatically designs compact network models, and quantization, which reduces the bit-precision of neural network models. As a consequence, joint design for both neural architecture and quantization precision settings is becoming increasingly popular. There are three main aspects that affect the performance of the joint optimization between neural architecture and quantization: quantization precision selection (QPS), quantization aware training (QAT), and neural architecture searching (NAS). However, existing works focus on at most twofold of these aspects, and result in secondary performance. To this end, we proposed a novel automatic optimization framework, DAQUDAQU is an ancient liquor fermentation process., that allows jointly searching for Pareto-optimal neural architecture and quantization precision combination among more than 1047 quantized subnet models. To overcome the instability of the conventional automatic optimization framework, DAQU incorporates a warm-up strategy to reduce the accuracy gap among different neural architectures, and a precision-transfer training approach to maintain flexibility among different quantization precision settings. Our experiments show that the quantized lightweight neural networks generated by DAQU consistently outperform state-of-the-art NAS and quantization joint optimization methods.

Bit-serial architectures (BSAs) are becoming increasingly popular in low-power neural network processor (NNP) designs for edge scenarios. However, the performance and energy efficiency of state-of-the-art BSA NNPs heavily depends on both the proportion and distribution of ineffectual weight bits in neural networks (NNs). To boost the performance of typical BSA accelerators, we present Bit-Pruner, a software approach to learn BSA-favored NNs without resorting to hardware modifications. Bit-Pruner not only progressively prunes but also restructures the nonzero bits in weights so that the number of nonzero bits in the model can be reduced and the corresponding computing can be load-balanced to suit the target BSA accelerators. On top of Bit-Pruner, we further propose a Pareto frontier optimization algorithm to adjust the bit-pruning rate across network layers and fulfill diverse NN processing requirements in terms of performance and accuracy for various edge scenarios. However, an aggressive Bit-Pruner can lead to nontrivial accuracy loss, especially for lightweight NNs and complex tasks. To this end, the alternating direction method of multipliers (ADMMs) is adapted to the retraining phase in Bit-Pruner to smooth the abrupt disturbance due to bit-pruning and enhance the resulting model accuracy. According to the experiments, Bit-Pruner increases the bit-sparsity up to 94.4% with negligible accuracy degradation and achieves an optimized tradeoff between NN accuracy and energy efficiency even under very-aggressive performance constraints. When pruned models are deployed onto typical BSA accelerators, the average performance is$2.1\times $and$1.6\times $higher than the baseline networks without pruning and those with classical weight pruning, respectively.

Multi-task learning systems are commonly adopted in many real-world AI applications such as intelligent robots and self-driving vehicles. Instead of improving single-network performance, this work proposes a specialized Multi-Task Deep Learning Accelerator architecture, MT-DLA, to improve the performance of concurrent networks by exploiting the shared feature and parameters across these models. It is shown in our evaluation with realistic multi-task workloads, MT-DLA dramatically eliminates the memory and computation overhead caused by the shared parameters, activations and computation result. In the experiments with real-world multi-task learning workloads, MT-DLA brings about 1.4x-7.0x energy efficiency boost when compared to the baseline neural network accelerator without multi-task support.

Resistive RAM (ReRAM) is a promising device to realize the Computing in Memory (CiM) architecture, suitable for power-constrained IoT systems. Because of low leakage, the dot-production operations in ReRAM crossbars dominate the chip power, especially when implementing low-precision neural networks. This work investigates the correlation between the cell resistance state and the crossbar operation power, and proposes a State-Aware ReRAM Accelerator (SARA) architecture for energy-efficient neural networks. With the proposed state-aware network training and mapping strategy, crossbars in the ReRAM accelerator can perform in a lower-power state. The evaluation shows that our design reduces 47% energy over the baseline without compromising the network accuracy.

Convolutional neural networks (CNNs) have become the state-of-the-art technique in many classification tasks in IoT system. However, the low-power and area-constraint edge devices are unable to afford the expensive cost of CNNs. Resistive random access memory (RRAM) is attractive for establishing the CNN accelerator at the edge end due to the features of scalability, low-power and in-situ dot-product. However, mapping a random network architecture onto a general-purpose RRAM accelerator suffers a severe issue of resource underutilization. The neural network quantization offers an opportunity to rescue the degraded resource utilization. Selecting the bit-width for the vast parameters is impractically completed by human labor. This paper proposes an AutoML-based array-aware quantization and mapping framework that generates the fine-grained mixed-precision neural networks to optimize resource utilization in RRAM. In this framework, we design a two-stage learning and array-aware grouping strategy to quickly explore the huge searching space. The experimental results show that the proposed framework achieves 18.2%~36.1% improvement in resource utilization and 0.9%~3.3% increase in model accuracy over prior coarse-grained quantization methods.

Bit-serial architectures (BSAs) are becoming increasingly popular in low power neural network processor (NNP) design. However, the performance and efficiency of state-of-the-art BSA NNPs are heavily depending on the distribution of ineffectual weight-bits of the running neural network. To boost the efficiency of third-party BSA accelerators, this work presents Bit-Pruner, a software approach to learn BSA-favored neural networks without resorting to hardware modifications. The techniques proposed in this work not only progressively prune but also structure the non-zero bits in weights, so that the number of zero-bits in the model can be increased and also load-balanced to suit the architecture of the target BSA accelerators. According to our experiments on a set of representative neural networks, Bit-Pruner increases the bit-sparsity up to 94.4% with negligible accuracy degradation. When the bit-pruned models are deployed onto typical BSA accelerators, the average performance is 2.1X and 1.5X higher than the baselines running non-pruned and weight-pruned networks, respectively.