arXiv subjects

Junyi Wu

Publications and source records attributed to Junyi Wu.

44 records · Page 3Linked to original sources

arXiv2024

PTQ4DiT: Post-training Quantization for Diffusion Transformers

The recent introduction of Diffusion Transformers (DiTs) has demonstrated exceptional capabilities in image generation by using a different backbone architecture, departing from traditional U-Nets and embracing the scalable nature of transformers. Despite their advanced capabilities, the wide deployment of DiTs, particularly for real-time applications, is currently hampered by considerable computational demands at the inference stage. Post-training Quantization (PTQ) has emerged as a fast and data-efficient solution that can significantly reduce computation and memory footprint by using low-bit weights and activations. However, its applicability to DiTs has not yet been explored and faces non-trivial difficulties due to the unique design of DiTs. In this paper, we propose PTQ4DiT, a specifically designed PTQ method for DiTs. We discover two primary quantization challenges inherent in DiTs, notably the presence of salient channels with extreme magnitudes and the temporal variability in distributions of salient activation over multiple timesteps. To tackle these challenges, we propose Channel-wise Salience Balancing (CSB) and Spearmen's $ρ$-guided Salience Calibration (SSC). CSB leverages the complementarity property of channel magnitudes to redistribute the extremes, alleviating quantization errors for both activations and weights. SSC extends this approach by dynamically adjusting the balanced salience to capture the temporal variations in activation. Additionally, to eliminate extra computational costs caused by PTQ4DiT during inference, we design an offline re-parameterization strategy for DiTs. Experiments demonstrate that our PTQ4DiT successfully quantizes DiTs to 8-bit precision (W8A8) while preserving comparable generation ability and further enables effective quantization to 4-bit weight precision (W4A8) for the first time.

cs.CV↗

arXiv2024

Dataset Quantization with Active Learning based Adaptive Sampling

Deep learning has made remarkable progress recently, largely due to the availability of large, well-labeled datasets. However, the training on such datasets elevates costs and computational demands. To address this, various techniques like coreset selection, dataset distillation, and dataset quantization have been explored in the literature. Unlike traditional techniques that depend on uniform sample distributions across different classes, our research demonstrates that maintaining performance is feasible even with uneven distributions. We find that for certain classes, the variation in sample quantity has a minimal impact on performance. Inspired by this observation, an intuitive idea is to reduce the number of samples for stable classes and increase the number of samples for sensitive classes to achieve a better performance with the same sampling ratio. Then the question arises: how can we adaptively select samples from a dataset to achieve optimal performance? In this paper, we propose a novel active learning based adaptive sampling strategy, Dataset Quantization with Active Learning based Adaptive Sampling (DQAS), to optimize the sample selection. In addition, we introduce a novel pipeline for dataset quantization, utilizing feature space from the final stage of dataset quantization to generate more precise dataset bins. Our comprehensive evaluations on the multiple datasets show that our approach outperforms the state-of-the-art dataset compression methods.

cs.CV↗

arXiv2024

Visual Grounding with Attention-Driven Constraint Balancing

Unlike Object Detection, Visual Grounding task necessitates the detection of an object described by complex free-form language. To simultaneously model such complex semantic and visual representations, recent state-of-the-art studies adopt transformer-based models to fuse features from both modalities, further introducing various modules that modulate visual features to align with the language expressions and eliminate the irrelevant redundant information. However, their loss function, still adopting common Object Detection losses, solely governs the bounding box regression output, failing to fully optimize for the above objectives. To tackle this problem, in this paper, we first analyze the attention mechanisms of transformer-based models. Building upon this, we further propose a novel framework named Attention-Driven Constraint Balancing (AttBalance) to optimize the behavior of visual features within language-relevant regions. Extensive experimental results show that our method brings impressive improvements. Specifically, we achieve constant improvements over five different models evaluated on four different benchmarks. Moreover, we attain a new state-of-the-art performance by integrating our method into QRNet.

cs.CV↗

arXiv2024

"I see it as a wellspring for my positive and upward journey in life.": Understanding Current Practices of Assistive Technology's Customized Modification in China

Due to the significant differences in physical conditions and living environments of people with disabilities, standardized assistive technologies (ATs) often fail to meet their needs. Modified AT, especially DIY (Do It Yourself) ATs, are a popular solution in many high-income countries, but there is a lack of documentation for low- and middle-income areas, especially in China, where the culture of philanthropy is undeveloped. To understand the current situation in this paper, we conducted semi-structured interviews with 10 individuals with disabilities using modified ATs and 10 individuals involved in providing these, including family members, standard assistive device manufacturers, and individuals employed for their modification skills, etc. Based on the results of the thematic analysis, we have summarized the general process of modified ATs for people with disabilities in China and the benefits these devices bring. We found that modified ATs not only make the lives of people with disabilities more comfortable and convenient but also bring them confidence, reduce social pressure, and even help them achieve self-realization. Additionally, we summarized the challenges they encountered before, during, and after the modification, including awareness gaps, family resistance, a lack of a business model, and so on. Specifically, we conducted a special case study about the typical business models and challenges currently faced by AT modification organizations in China. Our research provides important design foundations and research insights for the future of universal and personalized production of AT.

cs.HC↗

arXiv2024

HASS: Hardware-Aware Sparsity Search for Dataflow DNN Accelerator

Deep Neural Networks (DNNs) excel in learning hierarchical representations from raw data, such as images, audio, and text. To compute these DNN models with high performance and energy efficiency, these models are usually deployed onto customized hardware accelerators. Among various accelerator designs, dataflow architecture has shown promising performance due to its layer-pipelined structure and its scalability in data parallelism. Exploiting weights and activations sparsity can further enhance memory storage and computation efficiency. However, existing approaches focus on exploiting sparsity in non-dataflow accelerators, which cannot be applied onto dataflow accelerators because of the large hardware design space introduced. As such, this could miss opportunities to find an optimal combination of sparsity features and hardware designs. In this paper, we propose a novel approach to exploit unstructured weights and activations sparsity for dataflow accelerators, using software and hardware co-optimization. We propose a Hardware-Aware Sparsity Search (HASS) to systematically determine an efficient sparsity solution for dataflow accelerators. Over a set of models, we achieve an efficiency improvement ranging from 1.3$\times$ to 4.2$\times$ compared to existing sparse designs, which are either non-dataflow or non-hardware-aware. Particularly, the throughput of MobileNetV3 can be optimized to 4895 images per second. HASS is open-source: \url{https://github.com/Yu-Zhewen/HASS}

cs.AR↗

arXiv2024

Token Transformation Matters: Towards Faithful Post-hoc Explanation for Vision Transformer

While Transformers have rapidly gained popularity in various computer vision applications, post-hoc explanations of their internal mechanisms remain largely unexplored. Vision Transformers extract visual information by representing image regions as transformed tokens and integrating them via attention weights. However, existing post-hoc explanation methods merely consider these attention weights, neglecting crucial information from the transformed tokens, which fails to accurately illustrate the rationales behind the models' predictions. To incorporate the influence of token transformation into interpretation, we propose TokenTM, a novel post-hoc explanation method that utilizes our introduced measurement of token transformation effects. Specifically, we quantify token transformation effects by measuring changes in token lengths and correlations in their directions pre- and post-transformation. Moreover, we develop initialization and aggregation rules to integrate both attention weights and token transformation effects across all layers, capturing holistic token contributions throughout the model. Experimental results on segmentation and perturbation tests demonstrate the superiority of our proposed TokenTM compared to state-of-the-art Vision Transformer explanation methods.

cs.CV↗

arXiv2017

Magnetic order induces symmetry breaking in the single crystalline orthorhombic CuMnAs semimetal

Recently, orthorhombic CuMnAs has been proposed to be a magnetic material where topological fermions exist around the Fermi level. Here we report the magnetic structure of the orthorhombic Cu0.95MnAs and Cu0.98Mn0.96As single crystals. While Cu0.95MnAs is a commensurate antiferromagnet (C-AFM) below 360 K with a propagation vector of k = 0, Cu0.98Mn0.96As undergoes a second-order paramagnetic to incommensurate antiferromagnetic (IC-AFM) phase transition at 320 K with k = (0.1,0,0), followed by a second-order IC-AFM to C-AFM phase transition at 230 K. In the C-AFM state, the Mn spins order parallel to the b-axis but antiparallel to their nearest-neighbors with the easy axis along the b axis. This magnetic order breaks Ry gliding and S2z rotational symmetries, the two crucial for symmetry analysis, resulting in finite band gaps at the crossing point and the disappearance of the massless topological fermions. However, the spin-polarized surface states and signature induced by non-trivial topology still can be observed in this system, which makes orthorhombic CuMnAs promising in antiferromagnetic spintronics.

cond-mat.mtrl-sci↗

arXiv2014

Effect of low-intensity pulsed ultrasound on biocompatibility and cellular uptake of chitosan-TPP nanoparticles

Using low molecular weight chitosan nanoparticles (CNPs) prepared by an ionic gelation method, we report the effect of low-intensity pulsed ultrasound (US) on cell viability and nanoparticle uptake in cultured murine pre-osteoblasts. Particle size and zeta potential are measured using Dynamic Light Scattering (DLS), and cell viability is evaluated using the MTS assay. Results show that 30 min delivery of CNPs at 0.5 mg/mL is able to prevent loss of cell viability due to either serum starvation or subsequent exposure to US (1 W/cm2 or 2 W/cm2, up to 1 min). Additionally, flow cytometry data suggest that there is a close association between cellular membrane integrity and the presence of CNPs when US at 2 W/cm2 is administered.

q-bio.BM↗