The rapidly advancing field of Augmented and Virtual Reality (AR/VR) demands real-time, photorealistic rendering on resource-constrained platforms. 3D Gaussian Splatting, delivering state-of-the-art (SOTA) performance in rendering efficiency and quality, has emerged as a promising solution across a broad spectrum of AR/VR applications. However, despite its effectiveness on high-end GPUs, it struggles on edge systems like the Jetson Orin NX Edge GPU, achieving only 7-17 FPS—well below the over 60 FPS standard required for truly immersive AR/VR experiences. Addressing this challenge, we perform a comprehensive analysis of Gaussian-based AR/VR applications and identify the Gaussian Blending Stage, which intensively calculates each Gaussian’s contribution at every pixel, as the primary bottleneck. In response, we propose a Gaussian Blending Unit (GBU), an edge GPU plug-in module for real-time rendering in AR/VR applications. Notably, our GBU can be seamlessly integrated into conventional edge GPUs and collaboratively supports a wide range of AR/VR applications. Specifically, GBU incorporates an intra-row sequential shading (IRSS) dataflow that shades each row of pixels sequentially from left to right, utilizing a two-step coordinate transformation. This transformation enables (1) the sharing of intermediate values between adjacent pixels, reducing pixel-wise computation costs by up to $5.5 \times$, and (2) the early identification and skipping of Gaussians that minimally contribute to the pixels, reducing per-pixel computation by up to $\mathbf{9 3 \%}$. When directly deployed on a GPU, the proposed dataflow achieved a non-trivial $1.72 \times$ speedup on real-world static scenes, though still falls short of real-time rendering performance. Recognizing the limited compute utilization in the GPU-based implementation, GBU enhances rendering speed with a dedicated rendering engine that balances the workload across rows by aggregating computations from multiple Gaussians. Additionally, GBU integrates a Gaussian Reuse Cache, reducing off-chip memory accesses by 44.9% and resulting in a $1.14 \times$ speedup in rendering. Experiments across representative AR/VR applications demonstrate that our GBU provides a unified solution for on-device real-time rendering while maintaining SOTA rendering quality.

Diffusion models have shown remarkable success across generative tasks, yet their high computational demands challenge deployment on resource-limited platforms. This paper investigates a critical question for compute-optimal diffusion model deployment: Under a post-training setting without fine-tuning, is it more effective to reduce the number of denoising steps or to use a cheaper per-step inference? Intuitively, reducing the number of denoising steps increases the variability of the distributions across steps, making the model more sensitive to compression. In contrast, keeping more denoising steps makes the differences smaller, preserving redundancy, and making post-training compression more feasible. To systematically examine this, we propose PostDiff, a training-free framework for accelerating pre-trained diffusion models by reducing redundancy at both the input level and module level in a post-training manner. At the input level, we propose a mixed-resolution denoising scheme based on the insight that reducing generation resolution in early denoising steps can enhance low-frequency components and improve final generation fidelity. At the module level, we employ a hybrid module caching strategy to reuse computations across denoising steps. Extensive experiments and ablation studies demonstrate that (1) PostDiff can significantly improve the fidelity-efficiency trade-off of state-of-the-art diffusion models, and (2) to boost efficiency while maintaining decent generation fidelity, reducing per-step inference cost is often more effective than reducing the number of denoising steps. Our code is available at https://github.com/GATECH-EIC/PostDiff.

We propose Hymba, a family of small language models featuring a hybrid-head parallel architecture that integrates attention mechanisms and state space models (SSMs) within the same layer, offering parallel and complementary processing of the same inputs. In this hybrid-head module, attention heads provide high-resolution recall, while SSM heads facilitate efficient context summarization. Additionally, we introduce learnable meta tokens, which are prepended to prompts to store critical meta information, guiding subsequent tokens and alleviating the “forced-to-attend” burden associated with attention mechanisms. Thanks to the global context summarized by SSMs, the attention heads in our model can be further optimized through cross-layer key-value (KV) sharing and a mix of global and local attention, resulting in a compact cache size without compromising accuracy. Notably, Hymba achieves state-of-the-art performance among small LMs: Our Hymba-1.5B-Base model surpasses all sub-2B public models and even outperforms Llama-3.2-3B, achieving 1.32\% higher average accuracy, an 11.67$\times$ reduction in cache size, and 3.49$\times$ higher throughput.

State space models (SSMs) have emerged as an efficient alternative to Transformer models for language modeling, offering linear computational complexity and constant memory usage as context length increases. However, despite their efficiency in handling long contexts, recent studies have shown that SSMs, such as Mamba models, generally underperform compared to Transformers in long-context understanding tasks. To address this significant shortfall and achieve both efficient and accurate long-context understanding, we propose LongMamba, a training-free technique that significantly enhances the long-context capabilities of Mamba models. LongMamba builds on our discovery that the hidden channels in Mamba can be categorized into local and global channels based on their receptive field lengths, with global channels primarily responsible for long-context capability. These global channels can become the key bottleneck as the input context lengthens. Specifically, when input lengths largely exceed the training sequence length, global channels exhibit limitations in adaptively extend their receptive fields, leading to Mamba’s poor long-context performance. The key idea of LongMamba is to mitigate the hidden state memory decay in these global channels by preventing the accumulation of unimportant tokens in their memory. This is achieved by first identifying critical tokens in the global channels and then applying token filtering to accumulate only those critical tokens. Through extensive benchmarking across synthetic and real-world long-context scenarios, LongMamba sets a new standard for Mamba’s long-context performance, significantly extending its operational range without requiring additional training. Our code is available at https://github.com/GATECH-EIC/LongMamba.

Recent advancements in Large Language Models (LLMs) have spurred interest in numerous applications requiring robust long-range capabilities, essential for processing extensive input contexts and continuously generating extended outputs. As sequence lengths increase, the number of Key-Value (KV) pairs in LLMs escalates, creating a significant efficiency bottleneck. In this paper, we propose a new KV cache optimization paradigm called LaCache, a training-free method for efficient and accurate generative inference of LLMs. LaCache enables LLMs to simultaneously address both of the critical challenges in long-range modeling: robust long-range capabilities and continuous generation without running out-of-memory (OOM). Specifically, LaCache integrates two key innovations: (1) a ladder-shaped KV cache pattern that stores KV pairs not only sequentially (left-to-right within each layer) but also across layers (from shallow to deep), providing an extended span for capturing long-range dependencies under a fixed storage budget, thereby boosting long-range capabilities; and (2) an iterative compaction mechanism that progressively compresses older caches, freeing up space for new tokens within a fixed cache size. This token distance-based dynamic compression enables more effective continuous generation under constrained cache budgets. Experiments across various tasks, benchmarks, and LLM models consistently validate LaCache’s effectiveness in enhancing LLMs’ long-range capabilities. Our code is available at https://github.com/GATECH-EIC/LaCache.

Recent breakthroughs in AI reasoning, enabled by test-time compute (TTC) on compact large language models (LLMs), offer great potential for edge devices to effectively execute complex reasoning tasks.However, the intricate inference pipelines associated with TTC pose new efficiency bottlenecks, limiting achievable latency and hindering widespread adoption.Through an in-depth analysis, we identify three key barriers: (1) variable parallelism, characterized by inference-dependent dynamic control flows and varying batch sizes, complicating workload scheduling; (2) branch dependencies, hindering efficient pipelining across sequential reasoning steps; and (3) branch pruning, causing memory fragmentation and irregular data access patterns.Motivated by the memory-bound nature of LLMs and Processing-in-Memory (PIM)'s capability to reduce data movement, we propose ORCHES, a novel GPU-PIM collaborative system specifically designed to address these barriers.ORCHES integrates three key innovations: (1) adaptive workload assignment, dynamically balancing workloads between GPU and PIM units to maximize parallelism despite unpredictable branching;(2) branch-aware pipelining, leveraging speculative execution to substantially reduce inter-step pipeline stalls; and (3) fragmentationaware memory structuring, enhancing data locality and access efficiency through coordinated caching and optimized memory layout reorganization.Experimental results demonstrate that ORCHES * Sixu Li and Yuzhou Chen contributed equally to this work.

Pre-training datasets are typically collected from web content and lack inherent domain divisions. For instance, widely used datasets like Common Crawl do not include explicit domain labels, while manually curating labeled datasets such as The Pile is labor-intensive. Consequently, identifying an optimal pre-training data mixture remains a challenging problem, despite its significant benefits for pre-training performance. To address these challenges, we propose CLustering-based Iterative Data Mixture Bootstrapping (Nemotron-CLIMB), an automated framework that discovers, evaluates, and refines data mixtures in a pre-training setting. Specifically, Nemotron-CLIMB embeds and clusters large-scale datasets in a semantic space and then iteratively searches for optimal mixtures using a smaller proxy model and a predictor. This strategy enables effective domain adaptation without relying solely on curated data. When continuously trained on 400B tokens with this mixture, our 1B model exceeds the state-of-the-art Llama-3.2-1B by 2.0%. Moreover, we observe that optimizing for a specific domain (e.g., Social Sciences) yields a 5% improvement over random sampling. Finally, we introduce Nemotron-ClimbLab, a filtered 1.2-trillion-token corpus with 20 clusters as a research playground, and Nemotron-ClimbMix, a compact yet powerful 400-billion-token dataset designed for efficient pre-training that delivers superior performance under an equal token budget. We analyze the final data mixture, elucidating the characteristics of an optimal data mixture.

Efficient deployment of small language models (SLMs) is essential for numerous real-world applications with stringent latency constraints.While previous work on SLM design has primarily focused on reducing the number of parameters to achieve parameter-optimal SLMs, parameter efficiency does not necessarily translate into proportional real-device speed-ups. This work aims to identify the key determinants of SLMs' real-device latency and offer generalizable principles and methodologies for SLM design and training when real-device latency is the primary consideration. Specifically, we identify two central architectural factors: depth–width ratios and operator choices. The former is crucial for small-batch-size latency, while the latter affects both latency and large-batch-size throughput. In light of this, we first study latency-optimal depth–width ratios, with the key finding that although deep–thin models generally achieve better accuracy under the same parameter budget, they may not lie on the accuracy–latency trade-off frontier. Next, we explore emerging efficient attention alternatives to evaluate their potential as candidate building operators. Using the identified promising operators, we construct an evolutionary search framework to automatically discover latency-optimal combinations of these operators within hybrid SLMs, thereby advancing the accuracy–latency frontier. In addition to architectural improvements, we further enhance SLM training using a weight normalization technique that enables more effective weight updates and improves final convergence. This technique can serve as a generalizable component for future SLMs. Combining these methods, we introduce a new family of hybrid SLMs, called Nemotron-Flash, which significantly advances the accuracy–efficiency frontier of state-of-the-art SLMs, e.g., achieving over +5.5\% average accuracy, 1.3$\times$/1.9$\times$ lower latency, and 18.7$\times$/45.6$\times$ higher throughput compared to Qwen3-1.7B/0.6B, respectively.

Real-time Codec Avatars, which employ deep generative models for 3-D reconstruction of human features, are crucial for immersive telepresence in augmented reality and virtual reality (AR/VR) environments. However, deploying these avatars in real-time on AR/VR headsets is challenging due to the inability of existing devices to achieve satisfying performance within stringent hardware resource constraints. To address these challenges, we introduce Re-CATA, an innovative full-stack and flexible Codec Avatar accelerator design framework. Re-CATA is designed to deliver real-time throughput (greater than 120 FPS) for the complete Codec Avatar processing pipeline within an edge-level power budget of 5 W under FPGA prototyping. Our approach begins by abstracting the operation mapping and scheduling challenges inherent in Codec Avatars, which require both centralized and distributed processing to handle dynamically changing workloads. We propose a novel hardware resource and workload partitioning scheme optimized for these fluctuating demands. To complement this, we introduce an agile runtime scheduling system for efficient workload reallocation among computing units as needed, recognizing the limitations of static partitioning in rapidly evolving workload scenarios. Furthermore, our micro-architecture design incorporates unified computing modules and efficient hardware peripherals, enabling seamless workload balancing across the Codec Avatar processing pipeline. We evaluate the Re-CATA accelerators via on-board FPGA prototyping, comparing them to various baselines, including commercial AR/VR system-on-chips and academic accelerators. This evaluation demonstrates a maximum speedup of up to$5.95\times $under similar settings.

Attention is a fundamental component behind the remarkable achievements of large language models (LLMs). However, our current understanding of the attention mechanism, especially regarding how attention distributions are established, remains limited. Inspired by recent studies that explore the presence of attention sink in the initial token, which receives disproportionately large attention scores despite their lack of semantic importance, this work delves deeper into this phenomenon. We aim to provide a more profound understanding of the existence of attention sinks within LLMs and to uncover ways to enhance the achievable accuracy of LLMs by directly optimizing the attention distributions, without the need for weight finetuning. Specifically, this work begins with comprehensive visualizations of the attention distributions in LLMs during inference across various inputs and tasks. Based on these visualizations, to the best of our knowledge, we are the first to discover that (1) attention sinks occur not only at the start of sequences but also within later tokens of the input, and (2) not all attention sinks have a positive impact on the achievable accuracy of LLMs. Building upon our findings, we propose a training-free Attention Calibration Technique (ACT) that automatically optimizes the attention distributions on the fly during inference in an input-adaptive manner. Extensive experiments validate that ACT consistently enhances the accuracy of various LLMs across different applications. Specifically, ACT achieves an average improvement of up to $7.30%$ in accuracy across different datasets when applied to Llama-30B.

The remarkable advancements in artificial intel-ligence (AI), primarily driven by deep neural networks, are facing challenges surrounding unsustainable computational tra-jectories, limited robustness, and a lack of explainability. To develop next-generation cognitive AI systems, neuro-symbolic AI emerges as a promising paradigm, fusing neural and symbolic approaches to enhance interpretability, robustness, and trustwor-thiness, while facilitating learning from much less data. Recent neuro-symbolic systems have demonstrated great potential in collaborative human-AI scenarios with reasoning and cognitive capabilities. In this paper, we aim to understand the workload characteristics and potential architectures for neuro-symbolic AI. We first systematically categorize neuro-symbolic AI algorithms, and then experimentally evaluate and analyze them in terms of runtime, memory, computational operators, sparsity, and system characteristics on CPUs, GPUs, and edge SoCs. Our studies reveal that neuro-symbolic models suffer from inefficiencies on off-the-shelf hardware, due to the memory-bound nature of vector-symbolic and logical operations, complex flow control, data dependencies, sparsity variations, and limited scalability. Based on profiling insights, we suggest cross-layer optimization solutions to improve the performance, efficiency, and scalability of neuro-symbolic computing. Finally, we discuss the challenges and potential future directions of neuro-symbolic AI from both system and architectural perspectives.

Motivated by the transformative capabilities of large language models (LLMs) across various natural language tasks, there has been a growing demand to deploy these models effectively across diverse real-world applications and platforms. However, the challenge of efficiently deploying LLMs has become increasingly pronounced due to the varying application-specific performance requirements and the rapid evolution of computational platforms, which feature diverse resource constraints and deployment flows. These varying requirements necessitate LLMs that can adapt their structures (depth and width) for optimal efficiency across different platforms and application specifications. To address this critical gap, we propose AmoebaLLM, a novel framework designed to enable the instant derivation of LLM subnets of arbitrary shapes, which achieve the accuracy-efficiency frontier and can be extracted immediately after a one-time fine-tuning. In this way, AmoebaLLM significantly facilitates rapid deployment tailored to various platforms and applications. Specifically, AmoebaLLM integrates three innovative components: (1) a knowledge-preserving subnet selection strategy that features a dynamic-programming approach for depth shrinking and an importance-driven method for width shrinking; (2) a shape-aware mixture of LoRAs to mitigate gradient conflicts among subnets during fine-tuning; and (3) an in-place distillation scheme with loss-magnitude balancing as the fine-tuning objective. Extensive experiments validate that AmoebaLLM not only sets new standards in LLM adaptability but also successfully delivers subnets that achieve state-of-the-art trade-offs between accuracy and efficiency.

Although the neural radiance field (NeRF) exhibits high-fidelity visualization on the rendering task, it still suffers from rendering defects, especially in complex scenes. In this paper, we delve into the reason for the unsatisfactory performance and conjecture that it comes from interference in the training process. Due to occlusions in complex scenes, a 3D point may be invisible to some rays. On such a point, training with those rays that do not contain valid information about the point might interfere with the NeRF training. Based on the above intuition, we decouple the training process of NeRF in the ray dimension softly and propose a Ray-decoupled Training Framework for neural rendering (Rad-NeRF). Specifically, we construct an ensemble of sub-NeRFs and train a soft gate module to assign the gating scores to these sub-NeRFs based on specific rays. The gate module is jointly optimized with the sub-NeRF ensemble to learn the preference of sub-NeRFs for different rays automatically. Furthermore, we introduce depth-based mutual learning to enhance the rendering consistency among multiple sub-NeRFs and mitigate the depth ambiguity. Experiments on five datasets demonstrate that Rad-NeRF can enhance the rendering performance across a wide range of scene types compared with existing single-NeRF and multi-NeRF methods. With only 0.2% extra parameters, Rad-NeRF improves rendering performance by up to 1.5dB. Code is available at https://github.com/thu-nics/Rad-NeRF.

Real-time and robust photorealistic avatars for telepresence in AR/VR have been highly desired for enabling im-mersive photorealistic telepresence. However, there still exists one key bottleneck: the considerable computational expense needed to accurately infer facial expressions captured from headset-mounted cameras with a quality level that can match the realism of the avatar's human appearance. To this end, we propose a framework called Auto-CARD, which for the first time enables realtime and robust driving of Codec Avatars when exclusively using merely on-device computing resources. This is achieved by minimizing two sources of redundancy. First, we develop a dedicated neural architecture search technique called AVE-NAS for avatar encoding in AR/VR, which explicitly boosts both the searched architectures' robustness in the presence of extreme facial ex-pressions and hardware friendliness on fast evolving AR/VR headsets. Second, we leverage the temporal redundancy in consecutively captured images during continuous rendering and develop a mechanism dubbed LATEX to skip the computation of redundant frames. Specifically, we first identify an opportunity from the linearity of the latent space derived by the avatar decoder and then propose to perform adaptive latent extrapolation for redundant frames. For evaluation, we demonstrate the efficacy of our Auto-CARD framework in realtime Codec Avatar driving settings, where we achieve a$5.05\times$speedup on Meta Quest 2 while maintaining a compa-rable or even better animation quality than state-of-the-art avatar encoder designs.

Despite the growing demand for tuningfoundation vision transformers (FViTs) on downstream tasks, fully unleashing FViTs' potential under data-limited scenarios (e.g., few-shot tuning) remains a challenge due to FViTs' data-hungry nature. Common data augmentation techniques fall short in this context due to the limited features contained in the few-shot tuning data. To tackle this challenge, we first identify an opportunity for FViTs in few-shot tuning: pretrained FViTs themselves have already learned highly representative features from large-scale pretraining data, which are fully preserved during widely used parameter-efficient tuning. We thus hypothesize that leveraging those learned features to augment the tuning data can boost the effectiveness of few-shot FViT tuning. To this end, we propose a framework called Hint-based Data Augmentation (Hint-Aug), which aims to boost FViT in few-shot tuning byaugmenting the over-fitted parts of tuning samples with the learned features of pretrained FViTs. Specifically, Hint-Aug integrates two key enablers: (1) an Attentive Over-fitting Detector (AOD) to detect over-confident patches offoundation ViTs for potentially alleviating their over-fitting on the few-shot tuning data and (2) a Confusion-based Feature Infusion (CFI) module to infuse easy-to-confuse features from the pretrained FViTs with the over-confident patches detected by the above AOD in order to enhance the feature diversity during tuning. Extensive experiments and ablation studies on five datasets and three parameter-efficient tuning techniques consistently validate Hint-Aug's effectiveness: 0.04% ~ 32.91 % higher accuracy over the state-of-the-art (SOTA) data augmentation method under various low-shot settings. For example, on the Pet dataset, Hint-Aug achieves a 2.22% higher accuracy with 50% less training data over SOTA data augmentation methods.

Transfer learning leverages feature representations of deep neural networks (DNNs) pretrained on source tasks with rich data to empower effective finetuning on downstream tasks. However, the pre-trained models are often prohibitively large for delivering generalizable representations, which limits their deployment on edge devices with constrained resources. To close this gap, we propose a new transfer learning pipeline, which leverages our finding that robust tickets can transfer better, i.e., subnetworks drawn with properly induced adversarial robustness can win better transferability over vanilla lottery ticket subnetworks. Extensive experiments and ablation studies validate that our proposed transfer learning pipeline can achieve enhanced accuracy-sparsity trade-offs across both diverse downstream tasks and sparsity patterns, further enriching the lottery ticket hypothesis.

Tiny deep learning has attracted increasing attention driven by the substantial demand for deploying deep learning on numerous intelligent Internet-of-Things devices. However, it is still challenging to unleash tiny deep learning’s full potential on both large-scale datasets and downstream tasks due to the under-fitting issues caused by the limited model capacity of tiny neural networks (TNNs). To this end, we propose a framework called NetBooster to empower tiny deep learning by augmenting the architectures of TNNs via an expansion-then-contraction strategy. Extensive experiments show that NetBooster consistently outperforms state-of-the-art tiny deep learning solutions.

The remarkable capabilities and intricate nature of Artificial Intelligence (AI) have dramatically escalated the imperative for specialized AI accelerators. Nonetheless, designing these accelerators for various AI workloads remains both labor- and time-intensive. While existing design exploration and automation tools can partially alleviate the need for extensive human involvement, they still demand substantial hardware expertise, posing a barrier to non-experts and stifling AI accelerator development. Motivated by the astonishing potential of large language models (LLMs) for generating high-quality content in response to human language instructions, we embark on this work to examine the possibility of harnessing LLMs to automate AI accelerator design. Through this endeavor, we develop GPT4AIGChip, a framework intended to democratize AI accelerator design by leveraging human natural languages instead of domain-specific languages. Specifically, we first perform an in-depth investigation into LLMs' limitations and capabilities for AI accelerator design, thus aiding our understanding of our current position and garnering insights into LLM-powered automated AI accelerator design. Furthermore, drawing inspiration from the above insights, we develop a framework called GPT4AIGChip, which features an automated demo-augmented prompt-generation pipeline utilizing in-context learning to guide LLMs towards creating high-quality AI accelerator design. To our knowledge, this work is the first to demonstrate an effective pipeline for LLM-powered automated AI accelerator generation. Accordingly, we anticipate that our insights and framework can serve as a catalyst for innovations in next-generation LLM-powered design automation tools.

Generalizable Neural Radiance Fields (GNeRF) are one of the most promising real-world solutions for novel view synthesis, thanks to their cross-scene generalization capability and thus the possibility of instant rendering on new scenes. While adversarial robustness is essential for real-world applications, little study has been devoted to understanding its implication on GNeRF. We hypothesize that because GNeRF is implemented by conditioning on the source views from new scenes, which are often acquired from the Internet or third-party providers, there are potential new security concerns regarding its real-world applications. Meanwhile, existing understanding and solutions for neural networks’ adversarial robustness may not be applicable to GNeRF, due to its 3D nature and uniquely diverse operations. To this end, we present NeRFool, which to the best of our knowledge is the first work that sets out to understand the adversarial robustness of GNeRF. Specifically, NeRFool unveils the vulnerability patterns and important insights regarding GNeRF’s adversarial robustness. Built upon the above insights gained from NeRFool, we further develop NeRFool$^+$, which integrates two techniques capable of effectively attacking GNeRF across a wide range of target views, and provide guidelines for defending against our proposed attacks. We believe that our NeRFool/NeRFool$^+$ lays the initial foundation for future innovations in developing robust real-world GNeRF solutions. Our codes are available at: https://github.com/GATECH-EIC/NeRFool.

Despite the impressive performance recently achieved by automatic speech recognition (ASR), we observe two primary challenges that hinder its broader applications: (1) The difficulty of introducing scalability into the model to support more languages with limited training, inference, and storage overhead; (2) The low-resource adaptation ability that enables effective low-resource adaptation while avoiding over fitting and catastrophic forgetting issues. Inspired by recent findings, we hypothesize that we can address the above challenges with modules widely shared across languages. To this end, we propose an ASR framework, dubbed Master-ASR, that, for the first time, simultaneously achieves strong multilingual scalability and low-resource adaptation ability thanks to its modularize-then-assemble strategy. Specifically, Master-ASR learns a small set of generalizable sub-modules and adaptively assembles them for different languages to reduce the multilingual overhead and enable effective knowledge transfer for low-resource adaptation. Extensive experiments and visualizations demonstrate that Master-ASR can effectively discover language similarity and improve multilingual and low-resource ASR performance over state-of-the-art (SOTA) methods, e.g., under multilingual-ASR, our framework achieves a 0.13∼2.41 lower character error rate (CER) with 30% smaller inference overhead over SOTA solutions on multilingual ASR and a comparable CER with nearly 100 times fewer trainable parameters over SOTA solutions on low-resource tuning, respectively.

Novel view synthesis is an essential functionality for enabling immersive experiences in various Augmented- and Virtual-Reality (AR/VR) applications, for which Neural Radiance Field (NeRF) has emerged as the state-of-the-art (SOTA) technique. In particular, generalizable NeRFs have gained increasing popularity thanks to their cross-scene generalization capability, which enables NeRFs to be instantly serviceable for new scenes without per-scene training. Despite their promise, generalizable NeRFs aggravate the prohibitive complexity of NeRFs due to their required extra memory accesses needed to acquire scene features, causing NeRFs' ray marching process to be memory-bounded. To tackle this dilemma, existing sparsity-exploitation techniques for NeRFs fall short, because they require knowledge of the sparsity distribution of the target 3D scene which is unknown when generalizing NeRFs to a new scene.

The record-breaking performance of deep neural networks (DNNs) comes with heavy parameter budgets, which leads to external dynamic random access memory (DRAM) for storage. The prohibitive energy of DRAM accesses makes it nontrivial for DNN deployment on resource-constrained devices, calling for minimizing the movements of weights and data in order to improve the energy efficiency. Driven by this critical bottleneck, we present SmartDeal, a hardware-friendly algorithm framework to trade higher-cost memory storage/access for lower-cost computation, in order to aggressively boost the storage and energy efficiency, for both DNN inference and training. The core technique of SmartDeal is a novel DNN weight matrix decomposition framework with respective structural constraints on each matrix factor, carefully crafted to unleash the hardware-aware efficiency potential. Specifically, we decompose each weight tensor as the product of a small basis matrix and a large structurally sparse coefficient matrix whose nonzero elements are readily quantized to the power-of-2. The resulting sparse and readily quantized DNNs enjoy greatly reduced energy consumption in data movement as well as weight storage, while incurring minimal overhead to recover the original weights thanks to the required sparse bit-operations and cost-favorable computations. Beyond inference, we take another leap to embrace energy-efficient training, by introducing several customized techniques to address the unique roadblocks arising in training while preserving the SmartDeal structures. We also design a dedicated hardware accelerator to fully utilize the new weight structure to improve the real energy efficiency and latency performance. We conduct experiments on both vision and language tasks, with nine models, four datasets, and three settings (inference-only, adaptation, and fine-tuning). Our extensive results show that 1) being applied to inference, SmartDeal achieves up to 2.44× improvement in energy efficiency as evaluated using real hardware implementations and 2) being applied to training, SmartDeal can lead to 10.56× and 4.48× reduction in the storage and the training energy cost, respectively, with usually negligible accuracy loss, compared to state-of-the-art training baselines. Our source codes are available at: https://github.com/VITA-Group/SmartDeal.

Graph Convolutional Networks (GCNs) have emerged as the state-of-the-art deep learning model for representation learning on graphs. However, it remains notoriously challenging to train and inference GCNs over large graph datasets, limiting their application to large real-world graphs and hindering the exploration of deeper and more sophisticated GCN graphs. This is because as the graph size grows, the sheer number of node features and the large adjacency matrix can easily explode the required memory and data movements. To tackle the aforementioned challenges, we explore the possibility of drawing lottery tickets when sparsifying GCN graphs, i.e., subgraphs that largely shrink the adjacency matrix yet are capable of achieving accuracy comparable to or even better than their full graphs. Specifically, we for the first time discover the existence of graph early-bird (GEB) tickets that emerge at the very early stage when sparsifying GCN graphs, and propose a simple yet effective detector to automatically identify the emergence of such GEB tickets. Furthermore, we advocate graph-model co-optimization and develop a generic efficient GCN early-bird training framework dubbed GEBT that can significantly boost the efficiency of GCN training by (1) drawing joint early-bird tickets between the GCN graphs and models and (2) enabling simultaneously sparsification of both the GCN graphs and models. Experiments on various GCN models and datasets consistently validate our GEB finding and the effectiveness of our GEBT, e.g., our GEBT achieves up to 80.2% ~ 85.6% and 84.6% ~ 87.5% savings of GCN training and inference costs while offering a comparable or even better accuracy as compared to state-of-the-art methods. Our source code and supplementary appendix are available at https://github.com/RICE-EIC/Early-Bird-GCN.

Vision transformers have recently demonstrated great success in various computer vision tasks, motivating a tremendously increased interest in their deployment into many real-world IoT applications. However, powerful ViTs are often too computationally expensive to be fitted onto real-world resource-constrained platforms, due to (1) their quadratically increased complexity with the number of input tokens and (2) their overparameterized self-attention heads and model depth. In parallel, different images are of varied complexity and their different regions can contain various levels of visual information, e.g., a sky background is not as informative as a foreground object in object classification tasks, indicating that treating those regions equally in terms of model complexity is unnecessary while such opportunities for trimming down ViTs' complexity have not been fully exploited. To this end, we propose a Multi-grained Input-Adaptive Vision Transformer framework dubbed MIA-Former that can input-adaptively adjust the structure of ViTs at three coarse-to-fine-grained granularities (i.e., model depth and the number of model heads/tokens). In particular, our MIA-Former adopts a low-cost network trained with a hybrid supervised and reinforcement learning method to skip the unnecessary layers, heads, and tokens in an input adaptive manner, reducing the overall computational cost. Furthermore, an interesting side effect of our MIA-Former is that its resulting ViTs are naturally equipped with improved robustness against adversarial attacks over their static counterparts, because MIA-Former's multi-grained dynamic control improves the model diversity similar to the effect of ensemble and thus increases the difficulty of adversarial attacks against all its sub-models. Extensive experiments and ablation studies validate that the proposed MIA-Former framework can (1) effectively allocate adaptive computation budgets to the difficulty of input images, achieving state-of-the-art (SOTA) accuracy-efficiency trade-offs, e.g., up to 16.5\% computation savings with the same or even a higher accuracy compared with the SOTA dynamic transformer models, and (2) boost ViTs' robustness accuracy under various adversarial attacks over their vanilla counterparts by 2.4\% and 3.0\%, respectively. Our code is available at https://github.com/RICE-EIC/MIA-Former.

Contrastive learning learns visual representations by enforcing feature consistency under different augmented views. In this work, we explore contrastive learning from a new perspective. Interestingly, we find that quantization, when properly engineered, can enhance the effectiveness of contrastive learning. To this end, we propose a novel contrastive learning framework, dubbed Contrastive Quant, to encourage feature consistency under both differently augmented inputs via various data transformations and differently augmented weights/activations via various quantization levels. Extensive experiments, built on top of two state-of-the-art contrastive learning methods SimCLR and BYOL, show that Contrastive Quant consistently improves the learned visual representation.

Efficient deep neural network (DNN) models equipped with compact operators (e.g., depthwise convolutions) have shown great potential in reducing DNNs’ theoretical complexity (e.g., the total number of weights/operations) while maintaining a decent model accuracy. However, existing efficient DNNs are still limited in fulfilling their promise in boosting real-hardware efficiency, due to their commonly adopted compact operators’ low hardware utilization. In this work, we open up a new compression paradigm for developing real-hardware efficient DNNs, leading to boosted hardware efficiency while maintaining model accuracy. Interestingly, we observe that while some DNN layers’ activation functions help DNNs’ training optimization and achievable accuracy, they can be properly removed after training without compromising the model accuracy. Inspired by this observation, we propose a framework dubbed DepthShrinker, which develops hardware-friendly compact networks via shrinking the basic building blocks of existing efficient DNNs that feature irregular computation patterns into dense ones with much improved hardware utilization and thus real-hardware efficiency. Excitingly, our DepthShrinker framework delivers hardware-friendly compact networks that outperform both state-of-the-art efficient DNNs and compression techniques, e.g., a 3.06% higher accuracy and 1.53x throughput on Tesla V100 over SOTA channel-wise pruning method MetaPruning. Our codes are available at: https://github.com/facebookresearch/DepthShrinker.

Neural networks (NNs) with intensive multiplications (e.g., convolutions and transformers) are powerful yet power hungry, impeding their more extensive deployment into resource-constrained edge devices. As such, multiplication-free networks, which follow a common practice in energy-efficient hardware implementation to parameterize NNs with more efficient operators (e.g., bitwise shifts and additions), have gained growing attention. However, multiplication-free networks in general under-perform their vanilla counterparts in terms of the achieved accuracy. To this end, this work advocates hybrid NNs that consist of both powerful yet costly multiplications and efficient yet less powerful operators for marrying the best of both worlds, and proposes ShiftAddNAS, which can automatically search for more accurate and more efficient NNs. Our ShiftAddNAS highlights two enablers. Specifically, it integrates (1) the first hybrid search space that incorporates both multiplication-based and multiplication-free operators for facilitating the development of both accurate and efficient hybrid NNs; and (2) a novel weight sharing strategy that enables effective weight sharing among different operators that follow heterogeneous distributions (e.g., Gaussian for convolutions vs. Laplacian for add operators) and simultaneously leads to a largely reduced supernet size and much better searched networks. Extensive experiments and ablation studies on various models, datasets, and tasks consistently validate the effectiveness of ShiftAddNAS, e.g., achieving up to a +7.7% higher accuracy or a +4.9 better BLEU score as compared to state-of-the-art expert-designed and neural architecture searched NNs, while leading to up to 93% or 69% energy and latency savings, respectively. Codes and pretrained models are available at https://github.com/RICE-EIC/ShiftAddNAS.

Eye tracking has become an essential human-machine interaction modality for providing immersive experience in numerous virtual and augmented reality (VR/AR) applications desiring high throughput (e.g., 240 FPS), small-form, and enhanced visual privacy. However, existing eye tracking systems are still limited by their: (1) large form-factor largely due to the adopted bulky lens-based cameras; (2) high communication cost required between the camera and backend processor; and (3) potentially concerned low visual privacy, thus prohibiting their more extensive applications. To this end, we propose, develop, and validate a lensless FlatCambased eye tracking algorithm and accelerator co-design framework dubbed EyeCoD to enable eye tracking systems with a much reduced form-factor and boosted system efficiency without sacrificing the tracking accuracy, paving the way for next-generation eye tracking solutions. On the system level, we advocate the use of lensless FlatCams instead of lens-based cameras to facilitate the small form-factor need in mobile eye tracking systems, which also leaves rooms for a dedicated sensing-processor co-design to reduce the required camera-processor communication latency. On the algorithm level, EyeCoD integrates a predict-then-focus pipeline that first predicts the region-of-interest (ROI) via segmentation and then only focuses on the ROI parts to estimate gaze directions, greatly reducing redundant computations and data movements. On the hardware level, we further develop a dedicated accelerator that (1) integrates a novel workload orchestration between the aforementioned segmentation and gaze estimation models, (2) leverages intra-channel reuse opportunities for depth-wise layers, (3) utilizes input feature-wise partition to save activation memory size, and (4) develops a sequential-write-parallel-read input buffer to alleviate the bandwidth requirement for the activation global buffer. On-silicon measurement and extensive experiments validate that our EyeCoD consistently reduces both the communication and computation costs, leading to an overall system speedup of 10.95×, 3.21×, and 12.85× over general computing platforms including CPUs and GPUs, and a prior-art eye tracking processor called CIS-GEP, respectively, while maintaining the tracking accuracy. Codes are available at https://github.com/RICE-EIC/EyeCoD.

Self-supervised learning (SSL) for rich speech representations has achieved empirical success in low-resource Automatic Speech Recognition (ASR) and other speech processing tasks, which can mitigate the necessity of a large amount of transcribed speech and thus has driven a growing demand for on-device ASR and other speech processing. However, advanced speech SSL models have become increasingly large, which contradicts the limited on-device resources. This gap could be more severe in multilingual/multitask scenarios requiring simultaneously recognizing multiple languages or executing multiple speech processing tasks. Additionally, strongly overparameterized speech SSL models tend to suffer from overfitting when being finetuned on low-resource speech corpus. This work aims to enhance the practical usage of speech SSL models towards a win-win in both enhanced efficiency and alleviated overfitting via our proposed S$^3$-Router framework, which for the first time discovers that simply discarding no more than 10% of model weights via only finetuning model connections of speech SSL models can achieve better accuracy over standard weight finetuning on downstream speech processing tasks. More importantly, S$^3$-Router can serve as an all-in-one technique to enable (1) a new finetuning scheme, (2) an efficient multilingual/multitask solution, (3) a state-of-the-art pruning technique, and (4) a new tool to quantitatively analyze the learned speech representation. We believe S$^3$-Router has provided a new perspective for practical deployment of speech SSL models. Our codes are available at: https://github.com/GATECH-EIC/S3-Router.

There exists a gap in terms of the signals provided by pacemakers (i.e., intracardiac electrogram (EGM)) and the signals doctors use (i.e., 12-lead electrocardiogram (ECG)) to diagnose abnormal rhythms. Therefore, the former, even if remotely transmitted, are not sufficient for doctors to provide a precise diagnosis, let alone make a timely intervention. To close this gap and make a heuristic step towards real-time critical intervention in instant response to irregular and infrequent ventricular rhythms, we propose a new framework dubbed RT-RCG to automatically search for (1) efficient Deep Neural Network (DNN) structures and then (2) corresponding accelerators, to enable R eal- T ime and high-quality R econstruction of E C G signals from E G M signals. Specifically, RT-RCG proposes a new DNN search space tailored for ECG reconstruction from EGM signals and incorporates a differentiable acceleration search (DAS) engine to efficiently navigate over the large and discrete accelerator design space to generate optimized accelerators. Extensive experiments and ablation studies under various settings consistently validate the effectiveness of our RT-RCG. To the best of our knowledge, RT-RCG is the first to leverage neural architecture search (NAS) to simultaneously tackle both reconstruction efficacy and efficiency.

Semantic segmentation for scene understanding is nowadays widely demanded, raising significant challenges for the algorithm efficiency, especially its applications on resource-limited platforms. Current segmentation models are trained and evaluated on massive high-resolution scene images (“data-level”) and suffer from the expensive computation arising from the required multi-scale aggregation (“network level”). In both folds, the computational and energy costs in training and inference are notable due to the often desired large input resolutions and heavy computational burden of segmentation models. To this end, we propose DANCE, general automated DA ta- N etwork C o-optimization for E fficient segmentation model training and inference . Distinct from existing efficient segmentation approaches that focus merely on light-weight network design, DANCE distinguishes itself as an automated simultaneous data-network co-optimization via both input data manipulation and network architecture slimming. Specifically, DANCE integrates automated data slimming which adaptively downsamples/drops input images and controls their corresponding contribution to the training loss guided by the images’ spatial complexity. Such a downsampling operation, in addition to slimming down the cost associated with the input size directly, also shrinks the dynamic range of input object and context scales, therefore motivating us to also adaptively slim the network to match the downsampled data. Extensive experiments and ablating studies (on four SOTA segmentation models with three popular segmentation datasets under two training settings) demonstrate that DANCE can achieve “all-win” towards efficient segmentation (reduced training cost, less expensive inference, and better mean Intersection-over-Union (mIoU)). Specifically, DANCE can reduce ↓25%–↓77% energy consumption in training, ↓31%–↓56% in inference, while boosting the mIoU by ↓0.71%–↑ 13.34%.

The promise of Deep Neural Network (DNN) powered Internet of Thing (IoT) devices has motivated a tremendous demand for automated solutions to enable fast development and deployment of efficient (1) DNNs equipped with instantaneous accuracy-efficiency trade-off capability to accommodate the time-varying resources at IoT devices and (2) dataflows to optimize DNNs’ execution efficiency on different devices. Therefore, we propose InstantNet to automatically generate and deploy instantaneously switchable-precision networks which operates at variable bit-widths. Extensive experiments show that the proposed InstantNet consistently outperforms state-of-the-art designs. Our codes are available at: https://github.com/RICE-EIC/InstantNet.

Driven by the explosive interest in applying deep reinforcement learning (DRL) agents to numerous real-time control and decision-making applications, there has been a growing demand to deploy DRL agents to empower daily-life intelligent devices, while the prohibitive complexity of DRL stands at odds with limited on-device resources. In this work, we propose an Automated Agent Accelerator Co-Search (A3C-S) framework, which to our best knowledge is the first to automatically co-search the optimally matched DRL agents and accelerators that maximize both test scores and hardware efficiency. Extensive experiments consistently validate the superiority of our A3C-S over state-of-the-art techniques.

The recent breakthroughs and prohibitive complexities of Deep Neural Networks (DNNs) have excited extensive interest in domain specific DNN accelerators, among which optical DNN accelerators are particularly promising thanks to their unprecedented potential of achieving superior performance-per-watt. However, the development of optical DNN accelerators is much slower than that of electrical DNN accelerators. One key challenge is that while many techniques have been developed to facilitate the development of electrical DNN accelerators, techniques that support or expedite optical DNN accelerator design remain much less explored, limiting both the achievable performance and the innovation development of optical DNN accelerators. To this end, we develop the first-of-its-kind framework dubbed O-HAS, which for the first time demonstrates automated Optical Hardware Accelerator Search for boosting both the acceleration efficiency and development speed of optical DNN accelerators. Specifically, our O-HAS consists of two integrated enablers: (1) an O-Cost Predictor, which can accurately yet efficiently predict an optical accelerator's energy and latency based on the DNN model parameters and the optical accelerator design; and (2) an O-Search Engine, which can automatically explore the large design space of optical DNN accelerators and identify the optimal accelerators (i.e., the micro-architectures and algorithm-to-accelerator mapping methods) in order to maximize the target acceleration efficiency. Extensive experiments and ablation studies consistently validate the effectiveness of both our O-Cost Predictor and O-Search Engine as well as the excellent efficiency of O-HAS generated optical accelerators.

Graph Neural Networks (GNNs) have emerged as the state-of-the-art (SOTA) method for graph-based learning tasks. However, it still remains prohibitively challenging to inference GNNs over large graph datasets, limiting their application to large-scale real-world tasks. While end-to-end jointly optimizing GNNs and their accelerators is promising in boosting GNNs' inference efficiency and expediting the design process, it is still underexplored due to the vast and distinct design spaces of GNNs and their accelerators. In this work, we propose G-CoS, a GNN and accelerator co-search framework that can automatically search for matched GNN structures and accelerators to maximize both task accuracy and acceleration efficiency. Specifically, G-CoS integrates two major enabling components: (1) a generic GNN accelerator search space which is applicable to various GNN structures and (2) a one-shot GNN and accelerator co-search algorithm that enables simultaneous and efficient search for optimal GNN structures and their matched accelerators. To the best of our knowledge, G-CoS is the first co-search framework for GNNs and their accelerators. Extensive experiments and ablation studies show that the GNNs and accelerators generated by G-CoS consistently outperform SOTA GNNs and GNN accelerators in terms of both task accuracy and hardware efficiency, while only requiring a few hours for the end-to-end generation of the best matched GNNs and their accelerators.

There has been a booming demand for integrating Convolutional Neural Networks (CNNs) powered functionalities into Internet-of-Thing (IoT) devices to enable ubiquitous intelligent "IoT cameras". However, more extensive applications of such IoT systems are still limited by two challenges. First, some applications, especially medicine-and wearable-related ones, impose stringent requirements on the camera form factor. Second, powerful CNNs often require considerable storage and energy cost, whereas IoT devices often suffer from limited resources. PhlatCam, with its form factor potentially reduced by orders of magnitude, has emerged as a promising solution to the first aforementioned challenge, while the second one remains a bottleneck. Existing compression techniques, which can potentially tackle the second challenge, are far from realizing the full potential in storage and energy reduction, because they mostly focus on the CNN algorithm itself. To this end, this work proposes SACoD, a Sensor Algorithm Co-Design framework to develop more efficient CNN-powered PhlatCam. In particular, the mask coded in the Phlat-Cam sensor and the backend CNN model are jointly optimized in terms of both model parameters and architectures via differential neural architecture search. Extensive experiments including both simulation and physical measurement on manufactured masks show that the proposed SACoD framework achieves aggressive model compression and energy savings while maintaining or even boosting the task accuracy, when benchmarking over two state-of-the-art (SOTA) designs with six datasets across four different vision tasks including classification, segmentation, image translation, and face recognition. Our codes are available at: https://github.com/RICE-EIC/SACoD.

Quantization is promising in enabling powerful yet complex deep neural networks (DNNs) to be deployed into resource constrained platforms. However, quantized DNNs are vulnerable to adversarial attacks unless being equipped with sophisticated techniques, leading to a dilemma of struggling between DNNs’ efficiency and robustness. In this work, we demonstrate a new perspective regarding quantization’s role in DNNs’ robustness, advocating that quantization can be leveraged to largely boost DNNs’ robustness, and propose a framework dubbed Double-Win Quant that can boost the robustness of quantized DNNs over their full precision counterparts by a large margin. Specifically, we for the first time identify that when an adversarially trained model is quantized to different precisions in a post-training manner, the associated adversarial attacks transfer poorly between different precisions. Leveraging this intriguing observation, we further develop Double-Win Quant integrating random precision inference and training to further reduce and utilize the poor adversarial transferability, enabling an aggressive “win-win" in terms of DNNs’ robustness and efficiency. Extensive experiments and ablation studies consistently validate Double-Win Quant’s effectiveness and advantages over state-of-the-art (SOTA) adversarial training methods across various attacks/models/datasets. Our codes are available at: https://github.com/RICE-EIC/Double-Win-Quant.

While maximizing deep neural networks’ (DNNs’) acceleration efficiency requires a joint search/design of three different yet highly coupled aspects, including the networks, bitwidths, and accelerators, the challenges associated with such a joint search have not yet been fully understood and addressed. The key challenges include (1) the dilemma of whether to explode the memory consumption due to the huge joint space or achieve sub-optimal designs, (2) the discrete nature of the accelerator design space that is coupled yet different from that of the networks and bitwidths, and (3) the chicken and egg problem associated with network-accelerator co-search, i.e., co-search requires operation-wise hardware cost, which is lacking during search as the optimal accelerator depending on the whole network is still unknown during search. To tackle these daunting challenges towards optimal and fast development of DNN accelerators, we propose a framework dubbed Auto-NBA to enable jointly searching for the Networks, Bitwidths, and Accelerators, by efficiently localizing the optimal design within the huge joint design space for each target dataset and acceleration specification. Our Auto-NBA integrates a heterogeneous sampling strategy to achieve unbiased search with constant memory consumption, and a novel joint-search pipeline equipped with a generic differentiable accelerator search engine. Extensive experiments and ablation studies validate that both Auto-NBA generated networks and accelerators consistently outperform state-of-the-art designs (including co-search/exploration techniques, hardware-aware NAS methods, and DNN accelerators), in terms of search time, task accuracy, and accelerator efficiency. Our codes are available at: https://github.com/RICE-EIC/Auto-NBA.

We present DIAN, a Differentiable Accelerator-Network Co-Search framework for automatically searching for matched networks and accelerators to maximize both the accuracy and efficiency. Specifically, DIAN integrates two enablers: (1) a generic design space for DNN accelerators that is applicable to both FPGA- and ASIC-based DNN accelerators; and (2) a joint DNN network and accelerator co-search algorithm that enables the simultaneous search for optimal DNN structures and their accelerators. Experiments and ablation studies based on FPGA measurements and ASIC synthesis show that the matched networks and accelerators generated by DIAN consistently outperform state-of-the-art (SOTA) DNNs and DNN accelerators (e.g., 3.04× better FPS with a 5.46% higher accuracy on ImageNet), while requiring notably reduced search time (up to $1234.3\times)$ over SOTA co-exploration methods, when evaluated over ten SOTA baselines on three datasets.

The recent breakthroughs of deep neural networks (DNNs) and the advent of billions of Internet of Things (IoT) devices have excited an explosive demand for intelligent IoT devices equipped with domain-specific DNN accelerators. However, the deployment of DNN accelerator enabled intelligent functionality into real-world IoT devices still remains particularly challenging. First, powerful DNNs often come at prohibitive complexities, whereas IoT devices often suffer from stringent resource constraints. Second, while DNNs are vulnerable to adversarial attacks especially on IoT devices exposed to complex real-world environments, many IoT applications require strict security. Existing DNN accelerators mostly tackle only one of the two aforementioned challenges (i.e., efficiency or adversarial robustness) while neglecting or even sacrificing the other. To this end, we propose a 2-in-1 Accelerator, an integrated algorithm-accelerator co-design framework aiming at winning both the adversarial robustness and efficiency of DNN accelerators. Specifically, we first propose a Random Precision Switch (RPS) algorithm that can effectively defend DNNs against adversarial attacks by enabling random DNN quantization as an in-situ model switch during training and inference. Furthermore, we propose a new precision-scalable accelerator featuring (1) a new precision-scalable MAC unit architecture which spatially tiles the temporal MAC units to boost both the achievable efficiency and flexibility and (2) a systematically optimized dataflow that is searched by our generic accelerator optimizer. Extensive experiments and ablation studies validate that our 2-in-1 Accelerator can not only aggressively boost both the adversarial robustness and efficiency of DNN accelerators under various attacks, but also naturally support instantaneous robustness-efficiency trade-offs adapting to varied resources without the necessity of DNN retraining. We believe our 2-in-1 Accelerator has opened up an exciting perspective for robust and efficient accelerator design.

Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks, i.e., an imperceptible perturbation to the input can mislead DNNs trained on clean images into making erroneous predictions. To tackle this, adversarial training is currently the most effective defense method, by augmenting the training set with adversarial samples generated on the fly. \textbf{Interestingly, we discover for the first time that there exist subnetworks with inborn robustness, matching or surpassing the robust accuracy of the adversarially trained networks with comparable model sizes, within randomly initialized networks without any model training}, indicating that adversarial training on model weights is not indispensable towards adversarial robustness. We name such subnetworks Robust Scratch Tickets (RSTs), which are also by nature efficient. Distinct from the popular lottery ticket hypothesis, neither the original dense networks nor the identified RSTs need to be trained. To validate and understand this fascinating finding, we further conduct extensive experiments to study the existence and properties of RSTs under different models, datasets, sparsity patterns, and attacks, drawing insights regarding the relationship between DNNs’ robustness and their initialization/overparameterization. Furthermore, we identify the poor adversarial transferability between RSTs of different sparsity ratios drawn from the same randomly initialized dense network, and propose a Random RST Switch (R2S) technique, which randomly switches between different RSTs, as a novel defense method built on top of RSTs. We believe our findings about RSTs have opened up a new perspective to study model robustness and extend the lottery ticket hypothesis.

While increasingly deep networks are still in general desired for achieving state-of-the-art performance, for many specific inputs a simpler network might already suffice. Existing works exploited this observation by learning to skip convolutional layers in an input-dependent manner. However, we argue their binary decision scheme, i.e., either fully executing or completely bypassing one layer for a specific input, can be enhanced by introducing finer-grained, “softer” decisions. We therefore propose a Dynamic Fractional Skipping (DFS) framework. The core idea of DFS is to hypothesize layer-wise quantization (to different bitwidths) as intermediate “soft” choices to be made between fully utilizing and skipping a layer. For each input, DFS dynamically assigns a bitwidth to both weights and activations of each layer, where fully executing and skipping could be viewed as two “extremes” (i.e., full bitwidth and zero bitwidth). In this way, DFS can “fractionally” exploit a layer's expressive power during input-adaptive inference, enabling finer-grained accuracy-computational cost trade-offs. It presents a unified view to link input-adaptive layer skipping and input-adaptive hybrid quantization. Extensive experimental results demonstrate the superior tradeoff between computational cost and model expressive power (accuracy) achieved by DFS. More visualizations also indicate a smooth and consistent transition in the DFS behaviors, especially the learned choices between layer skipping and different quantizations when the total computational budgets vary, validating our hypothesis that layer quantization could be viewed as intermediate variants of layer skipping. Our source code and supplementary material are available at https://github.com/Torment123/DFS.

The compression of Generative Adversarial Networks (GANs) has lately drawn attention, due to the increasing demand for deploying GANs into mobile devices for numerous applications such as image translation, enhancement and editing. However, compared to the substantial efforts to compressing other deep models, the research on compressing GANs (usually the generators) remains at its infancy stage. Existing GAN compression algorithms are limited to handling specific GAN architectures and losses. Inspired by the recent success of AutoML in deep compression, we introduce AutoML to GAN compression and develop an AutoGAN-Distiller (AGD) framework. Starting with a specifically designed efficient search space, AGD performs an end-to-end discovery for new efficient generators, given the target computational resource constraints. The search is guided by the original GAN model via knowledge distillation, therefore fulfilling the compression. AGD is fully automatic, standalone (i.e., needing no trained discriminators), and generically applicable to various GAN models. We evaluate AGD in two representative GAN tasks: image translation and super resolution. Without bells and whistles, AGD yields remarkably lightweight yet more competitive compressed models, that largely outperform existing alternatives. Our codes and pretrained models are available at: https://github.com/TAMU-VITA/AGD.

We present SmartExchange, an algorithm-hardware co-design framework to trade higher-cost memory storage/access for lower-cost computation, for energy-efficient inference of deep neural networks (DNNs). We develop a novel algorithm to enforce a specially favorable DNN weight structure, where each layerwise weight matrix can be stored as the product of a small basis matrix and a large sparse coefficient matrix whose non-zero elements are all power-of-2. To our best knowledge, this algorithm is the first formulation that integrates three mainstream model compression ideas: sparsification or pruning, decomposition, and quantization, into one unified framework. The resulting sparse and readily-quantized DNN thus enjoys greatly reduced energy consumption in data movement as well as weight storage. On top of that, we further design a dedicated accelerator to fully utilize the SmartExchange-enforced weights to improve both energy efficiency and latency performance. Extensive experiments show that 1) on the algorithm level, SmartExchange outperforms stateof-the-art compression techniques, including merely sparsification or pruning, decomposition, and quantization, in various ablation studies based on nine models and four datasets; and 2) on the hardware level, SmartExchange can boost the energy efficiency by up to 6.7× and reduce the latency by up to 19.2× over four state-of-the-art DNN accelerators, when benchmarked on seven DNN models (including four standard DNNs, two compact DNN models, and one segmentation model) and three datasets.

Recent breakthroughs in deep neural networks (DNNs) have fueled a tremendous demand for intelligent edge devices featuring on-site learning, while the practical realization of such systems remains a challenge due to the limited resources available at the edge and the required massive training costs for state-of-the-art (SOTA) DNNs. As reducing precision is one of the most effective knobs for boosting training time/energy efficiency, there has been a growing interest in low-precision DNN training. In this paper, we explore from an orthogonal direction: how to fractionally squeeze out more training cost savings from the most redundant bit level, progressively along the training trajectory and dynamically per input. Specifically, we propose FracTrain that integrates (i) progressive fractional quantization which gradually increases the precision of activations, weights, and gradients that will not reach the precision of SOTA static quantized DNN training until the final training stage, and (ii) dynamic fractional quantization which assigns precisions to both the activations and gradients of each layer in an input-adaptive manner, for only "fractionally" updating layer parameters. Extensive simulations and ablation studies (six models, four datasets, and three training settings including standard, adaptation, and fine-tuning) validate the effectiveness of FracTrain in reducing computational cost and hardware-quantified energy/latency of DNN training while achieving a comparable or better (-0.12%~+1.87%) accuracy. For example, when training ResNet-74 on CIFAR-10, FracTrain achieves 77.6% and 53.5% computational cost and training latency savings, respectively, compared with the best SOTA baseline, while achieving a comparable (-0.07%) accuracy. Our codes are available at: https://github.com/RICE-EIC/FracTrain.

We propose a multi-modal method for the One-Minute Empathy Prediction competition. First, we use bottleneck residual and fully-connected network to encode facial images and speeches of the listener. Second, we propose to use the current time stage as a temporal feature and encoded it into the proposed multi-modal network. Third, we select a subset training data based on its performance of empathy prediction on the validation data. Experimental results on the testing set show that the proposed method outperforms the baseline methods significantly according to the CCC metric (0.14 vs 0.06).