We introduce PropNEAT, a fast backpropagation implementation of NEAT that uses a bidirectional mapping of the genome graph to a layer-based architecture that preserves the NEAT genomes whilst enabling efficient GPU backpropagation. We test PropNEAT on 58 binary classification datasets from the Penn Machine Learning Benchmarks database, comparing the performance against logistic regression, dense neural networks and random forests, as well as a densely retrained variant of the final PropNEAT model. PropNEAT had the second best overall performance, behind Random Forest, though the difference between the models was not statistically significant apart from between Random Forest in comparison with logistic regression and the PropNEAT retrain models. PropNEAT was substantially faster than a naive backpropagation method, and both were substantially faster and had better performance than the original NEAT implementation. We demonstrate that the per-epoch training time for PropNEAT scales linearly with network depth, and is efficient on GPU implementations for backpropagation. This implementation could be extended to support reinforcement learning or convolutional networks, and is able to find sparser and smaller networks with potential for applications in low-power contexts.
Semi-supervised video object segmentation (VOS) has been largely driven by space-time memory (STM) networks, which store past frame features in a spatiotemporal memory to segment the current frame via softmax attention. However, STM networks face memory limitations due to the quadratic complexity of softmax matching, restricting their applicability as video length and resolution increase. To address this, we propose LiVOS, a lightweight memory network that employs linear matching via linear attention, reformulating memory matching into a recurrent process that reduces the quadratic attention matrix to a constant-size, spatiotemporal-agnostic 2D state. To enhance selectivity, we introduce gated linear matching, where a data-dependent gate matrix is multiplied with the state matrix to control what information to retain or discard. Experiments on diverse benchmarks demonstrated the effectiveness of our method. It achieved 64.8 J&F on MOSE and 85.1 J&F on DAVIS, surpassing all non-STM methods and narrowing the gap with STM-based approaches. For longer and higher-resolution videos, it matched STM-based methods with 53% less GPU memory and supports 4096p inference on a 32G consumer-grade GPU--a previously cost-prohibitive capability--opening the door for long and high-resolution video foundation models.
Multimodal foundation models (MFMs) have revolutionized sequential recommender systems through advanced representation learning. While Parameter-efficient Fine-tuning (PEFT) is commonly used to adapt these models, studies often prioritize parameter efficiency, neglecting GPU memory and training speed. To address this, we introduced the IISAN framework, significantly enhancing efficiency. However, IISAN was limited to symmetrical MFMs and identical text and image encoders, preventing the use of state-of-the-art Large Language Models. To overcome this, we developed IISAN-Versa, a versatile plug-and-play architecture compatible with both symmetrical and asymmetrical MFMs. IISAN-Versa employs a Decoupled PEFT structure and utilizes both intra- and inter-modal adaptation. It effectively handles asymmetry through a simple yet effective combination of group layer-dropping and dimension transformation alignment. Our research demonstrates that IISAN-Versa effectively adapts large text encoders, and we further identify a scaling effect where larger encoders generally perform better. IISAN-Versa also demonstrates strong versatility in our defined multimodal scenarios, which include raw titles and captions generated from images and videos. Additionally, IISAN-Versa achieved state-of-the-art performance on the Microlens public benchmark. We will release our code and datasets to support future research.
We introduce the CRONOS algorithm for convex optimization of two-layer neural networks. CRONOS is the first algorithm capable of scaling to high-dimensional datasets such as ImageNet, which are ubiquitous in modern deep learning. This significantly improves upon prior work, which has been restricted to downsampled versions of MNIST and CIFAR-10. Taking CRONOS as a primitive, we then develop a new algorithm called CRONOS-AM, which combines CRONOS with alternating minimization, to obtain an algorithm capable of training multi-layer networks with arbitrary architectures. Our theoretical analysis proves that CRONOS converges to the global minimum of the convex reformulation under mild assumptions. In addition, we validate the efficacy of CRONOS and CRONOS-AM through extensive large-scale numerical experiments with GPU acceleration in JAX. Our results show that CRONOS-AM can obtain comparable or better validation accuracy than predominant tuned deep learning optimizers on vision and language tasks with benchmark datasets such as ImageNet and IMDb. To the best of our knowledge, CRONOS is the first algorithm which utilizes the convex reformulation to enhance performance on large-scale learning tasks.
Model selection in Gaussian processes scales prohibitively with the size of the training dataset, both in time and memory. While many approximations exist, all incur inevitable approximation error. Recent work accounts for this error in the form of computational uncertainty, which enables -- at the cost of quadratic complexity -- an explicit tradeoff between computation and precision. Here we extend this development to model selection, which requires significant enhancements to the existing approach, including linear-time scaling in the size of the dataset. We propose a novel training loss for hyperparameter optimization and demonstrate empirically that the resulting method can outperform SGPR, CGGP and SVGP, state-of-the-art methods for GP model selection, on medium to large-scale datasets. Our experiments show that model selection for computation-aware GPs trained on 1.8 million data points can be done within a few hours on a single GPU. As a result of this work, Gaussian processes can be trained on large-scale datasets without significantly compromising their ability to quantify uncertainty -- a fundamental prerequisite for optimal decision-making.
Multi-Object Tracking (MOT) poses significant challenges in computer vision. Despite its wide application in robotics, autonomous driving, and smart manufacturing, there is limited literature addressing the specific challenges of running MOT on embedded devices. State-of-the-art MOT trackers designed for high-end GPUs often experience low processing rates (<11fps) when deployed on embedded devices. Existing MOT frameworks for embedded devices proposed strategies such as fusing the detector model with the feature embedding model to reduce inference latency or combining different trackers to improve tracking accuracy, but tend to compromise one for the other. This paper introduces HopTrack, a real-time multi-object tracking system tailored for embedded devices. Our system employs a novel discretized static and dynamic matching approach along with an innovative content-aware dynamic sampling technique to enhance tracking accuracy while meeting the real-time requirement. Compared with the best high-end GPU modified baseline Byte (Embed) and the best existing baseline on embedded devices MobileNet-JDE, HopTrack achieves a processing speed of up to 39.29 fps on NVIDIA AGX Xavier with a multi-object tracking accuracy (MOTA) of up to 63.12% on the MOT16 benchmark, outperforming both counterparts by 2.15% and 4.82%, respectively. Additionally, the accuracy improvement is coupled with the reduction in energy consumption (20.8%), power (5%), and memory usage (8%), which are crucial resources on embedded devices. HopTrack is also detector agnostic allowing the flexibility of plug-and-play.
Clinical diagnosis of breast malignancy (BM) is a challenging problem in the recent era. In particular, Deep learning (DL) models have continued to offer important solutions for early BM diagnosis but their performance experiences overfitting due to the limited volume of breast ultrasound (BUS) image data. Further, large BUS datasets are difficult to manage due to privacy and legal concerns. Hence, image augmentation is a necessary and challenging step to improve the performance of the DL models. However, the current DL-based augmentation models are inadequate and operate as a black box resulting lack of information and justifications about their suitability and efficacy. Additionally, pre and post-augmentation need high-performance computational resources and time to produce the augmented image and evaluate the model performance. Thus, this study aims to develop a novel efficient augmentation approach for BUS images with advanced neural style transfer (NST) and Explainable AI (XAI) harnessing GPU-based parallel infrastructure. We scale and distribute the training of the augmentation model across 8 GPUs using the Horovod framework on a DGX cluster, achieving a 5.09 speedup while maintaining the model's accuracy. The proposed model is evaluated on 800 (348 benign and 452 malignant) BUS images and its performance is analyzed with other progressive techniques, using different quantitative analyses. The result indicates that the proposed approach can successfully augment the BUS images with 92.47% accuracy.
We present a novel algorithm for real-time planar semantic mapping tailored for humanoid robots navigating complex terrains such as staircases. Our method is adaptable to any odometry input and leverages GPU-accelerated processes for planar extraction, enabling the rapid generation of globally consistent semantic maps. We utilize an anisotropic diffusion filter on depth images to effectively minimize noise from gradient jumps while preserving essential edge details, enhancing normal vector images' accuracy and smoothness. Both the anisotropic diffusion and the RANSAC-based plane extraction processes are optimized for parallel processing on GPUs, significantly enhancing computational efficiency. Our approach achieves real-time performance, processing single frames at rates exceeding $30~Hz$, which facilitates detailed plane extraction and map management swiftly and efficiently. Extensive testing underscores the algorithm's capabilities in real-time scenarios and demonstrates its practical application in humanoid robot gait planning, significantly improving its ability to navigate dynamic environments.
Recent advancements in text-to-image diffusion models have enabled the personalization of these models to generate custom images from textual prompts. This paper presents an efficient LoRA-based personalization approach for on-device subject-driven generation, where pre-trained diffusion models are fine-tuned with user-specific data on resource-constrained devices. Our method, termed Hollowed Net, enhances memory efficiency during fine-tuning by modifying the architecture of a diffusion U-Net to temporarily remove a fraction of its deep layers, creating a hollowed structure. This approach directly addresses on-device memory constraints and substantially reduces GPU memory requirements for training, in contrast to previous methods that primarily focus on minimizing training steps and reducing the number of parameters to update. Additionally, the personalized Hollowed Net can be transferred back into the original U-Net, enabling inference without additional memory overhead. Quantitative and qualitative analyses demonstrate that our approach not only reduces training memory to levels as low as those required for inference but also maintains or improves personalization performance compared to existing methods.
The rapid progress in artificial intelligence-generated content (AIGC), especially with diffusion models, has significantly advanced development of high-quality video generation. However, current video diffusion models exhibit demanding computational requirements and high peak memory usage, especially for generating longer and higher-resolution videos. These limitations greatly hinder the practical application of video diffusion models on standard hardware platforms. To tackle this issue, we present a novel, training-free framework named Streamlined Inference, which leverages the temporal and spatial properties of video diffusion models. Our approach integrates three core components: Feature Slicer, Operator Grouping, and Step Rehash. Specifically, Feature Slicer effectively partitions input features into sub-features and Operator Grouping processes each sub-feature with a group of consecutive operators, resulting in significant memory reduction without sacrificing the quality or speed. Step Rehash further exploits the similarity between adjacent steps in diffusion, and accelerates inference through skipping unnecessary steps. Extensive experiments demonstrate that our approach significantly reduces peak memory and computational overhead, making it feasible to generate high-quality videos on a single consumer GPU (e.g., reducing peak memory of AnimateDiff from 42GB to 11GB, featuring faster inference on 2080Ti).
Online LLM inference powers many exciting applications such as intelligent chatbots and autonomous agents. Modern LLM inference engines widely rely on request batching to improve inference throughput, aiming to make it cost-efficient when running on expensive GPU accelerators. However, the limited GPU memory has largely limited the batch size achieved in practice, leaving significant GPU compute resources wasted. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU, effectively increasing the GPU batch size and thus inference throughput. To this end, NEO proposes asymmetric GPU-CPU pipelining and load-aware scheduling to balance GPU and CPU loads and fully utilize their compute and memory resources. We evaluate NEO on a wide range of workloads (i.e., code generation, text summarization), GPUs (i.e., T4, A10G, H100), and LLM models (i.e., 7B, 8B, 70B). NEO achieves up to 7.5$\times$, 26%, and 14% higher throughput compared to GPU-only approach on T4, A10G, and H100 GPUs, respectively, while maintaining the same latency; with more powerful CPUs, NEO achieves up to 79.3% throughput gain on A10G GPU.
We present context parallelism for long-context large language model inference, which achieves near-linear scaling for long-context prefill latency with up to 128 H100 GPUs across 16 nodes. Particularly, our method achieves 1M context prefill with Llama3 405B model in 77s (93% parallelization efficiency, 63% FLOPS utilization) and 128K context prefill in 3.8s. We develop two lossless exact ring attention variants: pass-KV and pass-Q to cover a wide range of use cases with the state-of-the-art performance: full prefill, persistent KV prefill and decode. Benchmarks on H100 GPU hosts inter-connected with RDMA and TCP both show similar scalability for long-context prefill, demonstrating that our method scales well using common commercial data center with medium-to-low inter-host bandwidth.
Retrieval-augmented generation (RAG) combines knowledge from domain-specific sources into large language models to ground answer generation. Current RAG systems lack customizable visibility on the context documents and the model's attentiveness towards such documents. We propose RAGViz, a RAG diagnosis tool that visualizes the attentiveness of the generated tokens in retrieved documents. With a built-in user interface, retrieval index, and Large Language Model (LLM) backbone, RAGViz provides two main functionalities: (1) token and document-level attention visualization, and (2) generation comparison upon context document addition and removal. As an open-source toolkit, RAGViz can be easily hosted with a custom embedding model and HuggingFace-supported LLM backbone. Using a hybrid ANN (Approximate Nearest Neighbor) index, memory-efficient LLM inference tool, and custom context snippet method, RAGViz operates efficiently with a median query time of about 5 seconds on a moderate GPU node. Our code is available at https://github.com/cxcscmu/RAGViz. A demo video of RAGViz can be found at https://youtu.be/cTAbuTu6ur4.
Diffusion models are pivotal for generating high-quality images and videos. Inspired by the success of OpenAI's Sora, the backbone of diffusion models is evolving from U-Net to Transformer, known as Diffusion Transformers (DiTs). However, generating high-quality content necessitates longer sequence lengths, exponentially increasing the computation required for the attention mechanism, and escalating DiTs inference latency. Parallel inference is essential for real-time DiTs deployments, but relying on a single parallel method is impractical due to poor scalability at large scales. This paper introduces xDiT, a comprehensive parallel inference engine for DiTs. After thoroughly investigating existing DiTs parallel approaches, xDiT chooses Sequence Parallel (SP) and PipeFusion, a novel Patch-level Pipeline Parallel method, as intra-image parallel strategies, alongside CFG parallel for inter-image parallelism. xDiT can flexibly combine these parallel approaches in a hybrid manner, offering a robust and scalable solution. Experimental results on two 8xL40 GPUs (PCIe) nodes interconnected by Ethernet and an 8xA100 (NVLink) node showcase xDiT's exceptional scalability across five state-of-the-art DiTs. Notably, we are the first to demonstrate DiTs scalability on Ethernet-connected GPU clusters. xDiT is available at https://github.com/xdit-project/xDiT.
One of the major bottlenecks for efficient deployment of neural network based recommendation systems is the memory footprint of their embedding tables. Although many neural network based recommendation systems could benefit from the faster on-chip memory access and increased computational power of hardware accelerators, the large embedding tables in these models often cannot fit on the constrained memory of accelerators. Despite the pervasiveness of these models, prior methods in memory optimization and parallelism fail to address the memory and communication costs of large embedding tables on accelerators. As a result, the majority of models are trained on CPUs, while current implementations of accelerators are hindered by issues such as bottlenecks in inter-device communication and main memory lookups. In this paper, we propose a theoretical framework that analyses the communication costs of arbitrary distributed systems that use lookup tables. We use this framework to propose algorithms that maximize throughput subject to memory, computation, and communication constraints. Furthermore, we demonstrate that our method achieves strong theoretical performance across dataset distributions and memory constraints, applicable to a wide range of use cases from mobile federated learning to warehouse-scale computation. We implement our framework and algorithms in PyTorch and achieve up to 6x increases in training throughput on GPU systems over baselines, on the Criteo Terabytes dataset.
MLLMs have demonstrated remarkable comprehension and reasoning capabilities with complex language and visual data. These advances have spurred the vision of establishing a generalist robotic MLLM proficient in understanding complex human instructions and accomplishing various embodied tasks. However, developing MLLMs for real-world robots is challenging due to the typically limited computation and memory capacities available on robotic platforms. In contrast, the inference of MLLMs involves storing billions of parameters and performing tremendous computation, imposing significant hardware demands. In our paper, we propose a Dynamic Early-Exit Framework for Robotic Vision-Language-Action Model (DeeR-VLA, or simply DeeR) that automatically adjusts the size of the activated MLLM based on each situation at hand. The approach leverages a multi-exit architecture in MLLMs, which allows the model to terminate processing once a proper size of the model has been activated for a specific situation, thus avoiding further redundant computation. Additionally, we develop novel algorithms that establish early-termination criteria for DeeR, conditioned on predefined demands such as average computational cost (i.e., power consumption), as well as peak computational consumption (i.e., latency) and GPU memory usage. These enhancements ensure that DeeR operates efficiently under varying resource constraints while maintaining competitive performance. On the CALVIN robot manipulation benchmark, DeeR demonstrates significant reductions in computational costs of LLM by 5.2-6.5x and GPU memory of LLM by 2-6x without compromising performance. Code and checkpoints are available at https://github.com/yueyang130/DeeR-VLA.
Despite the popularity of large language model (LLM) quantization for inference acceleration, significant uncertainty remains regarding the accuracy-performance trade-offs associated with various quantization formats. We present a comprehensive empirical study of quantized accuracy, evaluating popular quantization formats (FP8, INT8, INT4) across academic benchmarks and real-world tasks, on the entire Llama-3.1 model family. Additionally, our study examines the difference in text generated by quantized models versus their uncompressed counterparts. Beyond benchmarks, we also present a couple of quantization improvements which allowed us to obtain state-of-the-art accuracy recovery results. Our investigation, encompassing over 500,000 individual evaluations, yields several key findings: (1) FP8 weight and activation quantization (W8A8-FP) is lossless across all model scales, (2) INT8 weight and activation quantization (W8A8-INT), when properly tuned, incurs surprisingly low 1-3% accuracy degradation, and (3) INT4 weight-only quantization (W4A16-INT) is competitive with 8-bit integer weight and activation quantization. To address the question of the "best" format for a given deployment environment, we conduct inference performance analysis using the popular open-source vLLM framework on various GPU architectures. We find that W4A16 offers the best cost-efficiency for synchronous deployments, and for asynchronous deployment on mid-tier GPUs. At the same time, W8A8 formats excel in asynchronous "continuous batching" deployment of mid- and large-size models on high-end GPUs. Our results provide a set of practical guidelines for deploying quantized LLMs across scales and performance requirements.
Accurate channel modeling in real-time faces remarkable challenge due to the complexities of traditional methods such as ray tracing and field measurements. AI-based techniques have emerged to address these limitations, offering rapid, precise predictions of channel properties through ground truth data. This paper introduces an innovative approach to real-time, high-fidelity propagation modeling through advanced deep learning. Our model integrates 3D geographical data and rough propagation estimates to generate precise path gain predictions. By positioning the transmitter centrally, we simplify the model and enhance its computational efficiency, making it amenable to larger scenarios. Our approach achieves a normalized Root Mean Squared Error of less than 0.035 dB over a 37,210 square meter area, processing in just 46 ms on a GPU and 183 ms on a CPU. This performance significantly surpasses traditional high-fidelity ray tracing methods, which require approximately three orders of magnitude more time. Additionally, the model's adaptability to real-world data highlights its potential to revolutionize wireless network design and optimization, through enabling real-time creation of adaptive digital twins of real-world wireless scenarios in dynamic environments.
Scaling has been critical in improving model performance and generalization in machine learning. It involves how a model's performance changes with increases in model size or input data, as well as how efficiently computational resources are utilized to support this growth. Despite successes in other areas, the study of scaling in Neural Network Interatomic Potentials (NNIPs) remains limited. NNIPs act as surrogate models for ab initio quantum mechanical calculations. The dominant paradigm here is to incorporate many physical domain constraints into the model, such as rotational equivariance. We contend that these complex constraints inhibit the scaling ability of NNIPs, and are likely to lead to performance plateaus in the long run. In this work, we take an alternative approach and start by systematically studying NNIP scaling strategies. Our findings indicate that scaling the model through attention mechanisms is efficient and improves model expressivity. These insights motivate us to develop an NNIP architecture designed for scalability: the Efficiently Scaled Attention Interatomic Potential (EScAIP). EScAIP leverages a multi-head self-attention formulation within graph neural networks, applying attention at the neighbor-level representations. Implemented with highly-optimized attention GPU kernels, EScAIP achieves substantial gains in efficiency--at least 10x faster inference, 5x less memory usage--compared to existing NNIPs. EScAIP also achieves state-of-the-art performance on a wide range of datasets including catalysts (OC20 and OC22), molecules (SPICE), and materials (MPTrj). We emphasize that our approach should be thought of as a philosophy rather than a specific model, representing a proof-of-concept for developing general-purpose NNIPs that achieve better expressivity through scaling, and continue to scale efficiently with increased computational resources and training data.
Positron emission tomography (PET) is a critical tool for diagnosing tumors and neurological disorders but poses radiation risks to patients, particularly to sensitive populations. While reducing injected radiation dose mitigates this risk, it often compromises image quality. To reconstruct full-dose-quality images from low-dose scans, we propose a Cycle-constrained Adversarial Denoising Convolutional Network (Cycle-DCN). This model integrates a noise predictor, two discriminators, and a consistency network, and is optimized using a combination of supervised loss, adversarial loss, cycle consistency loss, identity loss, and neighboring Structural Similarity Index (SSIM) loss. Experiments were conducted on a large dataset consisting of raw PET brain data from 1,224 patients, acquired using a Siemens Biograph Vision PET/CT scanner. Each patient underwent a 120-seconds brain scan. To simulate low-dose PET conditions, images were reconstructed from shortened scan durations of 30, 12, and 5 seconds, corresponding to 1/4, 1/10, and 1/24 of the full-dose acquisition, respectively, using a custom-developed GPU-based image reconstruction software. The results show that Cycle-DCN significantly improves average Peak Signal-to-Noise Ratio (PSNR), SSIM, and Normalized Root Mean Square Error (NRMSE) across three dose levels, with improvements of up to 56%, 35%, and 71%, respectively. Additionally, it achieves contrast-to-noise ratio (CNR) and Edge Preservation Index (EPI) values that closely align with full-dose images, effectively preserving image details, tumor shape, and contrast, while resolving issues with blurred edges. The results of reader studies indicated that the images restored by Cycle-DCN consistently received the highest ratings from nuclear medicine physicians, highlighting their strong clinical relevance.