Collaborative filtering (CF) has been proven to be one of the most effective techniques for recommendation. Among all CF approaches, SimpleX is the state-of-the-art method that adopts a novel loss function and a proper number of negative samples. However, there is no work that optimizes SimpleX on multi-core CPUs, leading to limited performance. To this end, we perform an in-depth profiling and analysis of existing SimpleX implementations and identify their performance bottlenecks including (1) irregular memory accesses, (2) unnecessary memory copies, and (3) redundant computations. To address these issues, we propose an efficient CF training system (called HEAT) that fully enables the multi-level caching and multi-threading capabilities of modern CPUs. Specifically, the optimization of HEAT is threefold: (1) It tiles the embedding matrix to increase data locality and reduce cache misses (thus reduces read latency); (2) It optimizes stochastic gradient descent (SGD) with sampling by parallelizing vector products instead of matrix-matrix multiplications, in particular the similarity computation therein, to avoid memory copies for matrix data preparation; and (3) It aggressively reuses intermediate results from the forward phase in the backward phase to alleviate redundant computation. Evaluation on five widely used datasets with both x86- and ARM-architecture processors shows that HEAT achieves up to 45.2X speedup over existing CPU solution and 4.5X speedup and 7.9X cost reduction in Cloud over existing GPU solution with NVIDIA V100 GPU.
Dynamic Graph Neural Networks (DGNNs) are becoming increasingly popular due to their effectiveness in analyzing and predicting the evolution of complex interconnected graph-based systems. However, hardware deployment of DGNNs still remains a challenge. First, DGNNs do not fully utilize hardware resources because temporal data dependencies cause low hardware parallelism. Additionally, there is currently a lack of generic DGNN hardware accelerator frameworks, and existing GNN accelerator frameworks have limited ability to handle dynamic graphs with changing topologies and node features. To address the aforementioned challenges, in this paper, we propose DGNN-Booster, which is a novel Field-Programmable Gate Array (FPGA) accelerator framework for real-time DGNN inference using High-Level Synthesis (HLS). It includes two different FPGA accelerator designs with different dataflows that can support the most widely used DGNNs. We showcase the effectiveness of our designs by implementing and evaluating two representative DGNN models on ZCU102 board and measuring the end-to-end performance. The experiment results demonstrate that DGNN-Booster can achieve a speedup of up to 5.6x compared to the CPU baseline (6226R), 8.4x compared to the GPU baseline (A6000) and 2.1x compared to the FPGA baseline without applying optimizations proposed in this paper. Moreover, DGNN-Booster can achieve over 100x and over 1000x runtime energy efficiency than the CPU and GPU baseline respectively. Our implementation code and on-board measurements are publicly available at https://github.com/sharc-lab/DGNN-Booster.
N-body simulations are the most powerful method to study the non-linear evolution of large-scale structure. However, they require large amounts of computational resources, making unfeasible their direct adoption in scenarios that require broad explorations of parameter spaces. In this work, we show that it is possible to perform fast dark matter density field emulations with competitive accuracy using simple machine-learning approaches. We build an emulator based on dimensionality reduction and machine learning regression combining simple Principal Component Analysis and supervised learning methods. For the estimations with a single free parameter, we train on the dark matter density parameter, $\Omega_m$, while for emulations with two free parameters, we train on a range of $\Omega_m$ and redshift. The method first adopts a projection of a grid of simulations on a given basis; then, a machine learning regression is trained on this projected grid. Finally, new density cubes for different cosmological parameters can be estimated without relying directly on new N-body simulations by predicting and de-projecting the basis coefficients. We show that the proposed emulator can generate density cubes at non-linear cosmological scales with density distributions within a few percent compared to the corresponding N-body simulations. The method enables gains of three orders of magnitude in CPU run times compared to performing a full N-body simulation while reproducing the power spectrum and bispectrum within $\sim 1\%$ and $\sim 3\%$, respectively, for the single free parameter emulation and $\sim 5\%$ and $\sim 15\%$ for two free parameters. This can significantly accelerate the generation of density cubes for a wide variety of cosmological models, opening the doors to previously unfeasible applications, such as parameter and model inferences at full survey scales as the ESA/NASA Euclid mission.
Recent works on personalized text-to-image generation usually learn to bind a special token with specific subjects or styles of a few given images by tuning its embedding through gradient descent. It is natural to question whether we can optimize the textual inversions by only accessing the process of model inference. As only requiring the forward computation to determine the textual inversion retains the benefits of less GPU memory, simple deployment, and secure access for scalable models. In this paper, we introduce a \emph{gradient-free} framework to optimize the continuous textual inversion in an iterative evolutionary strategy. Specifically, we first initialize an appropriate token embedding for textual inversion with the consideration of visual and text vocabulary information. Then, we decompose the optimization of evolutionary strategy into dimension reduction of searching space and non-convex gradient-free optimization in subspace, which significantly accelerates the optimization process with negligible performance loss. Experiments in several applications demonstrate that the performance of text-to-image model equipped with our proposed gradient-free method is comparable to that of gradient-based counterparts with variant GPU/CPU platforms, flexible employment, as well as computational efficiency.
Given the ubiquity of non-separable optimization problems in real worlds, in this paper we analyze and extend the large-scale version of the well-known cooperative coevolution (CC), a divide-and-conquer black-box optimization framework, on non-separable functions. First, we reveal empirical reasons of when decomposition-based methods are preferred or not in practice on some non-separable large-scale problems, which have not been clearly pointed out in many previous CC papers. Then, we formalize CC to a continuous-game model via simplification, but without losing its essential property. Different from previous evolutionary game theory for CC, our new model provides a much simpler but useful viewpoint to analyze its convergence, since only the pure Nash equilibrium concept is needed and more general fitness landscapes can be explicitly considered. Based on convergence analyses, we propose a hierarchical decomposition strategy for better generalization, as for any decomposition, there is a risk of getting trapped into a suboptimal Nash equilibrium. Finally, we use powerful distributed computing to accelerate it under the recent multi-level learning framework, which combines the fine-tuning ability from decomposition with the invariance property of CMA-ES. Experiments on a set of high-dimensional test functions validate both its search performance and scalability (w.r.t. CPU cores) on a clustering computing platform with 400 CPU cores.
In the past few years, more and more AI applications have been applied to edge devices. However, models trained by data scientists with machine learning frameworks, such as PyTorch or TensorFlow, can not be seamlessly executed on edge. In this paper, we develop an end-to-end code generator parsing a pre-trained model to C source libraries for the backend using MicroTVM, a machine learning compiler framework extension addressing inference on bare metal devices. An analysis shows that specific compute-intensive operators can be easily offloaded to the dedicated accelerator with a Universal Modular Accelerator (UMA) interface, while others are processed in the CPU cores. By using the automatically generated ahead-of-time C runtime, we conduct a hand gesture recognition experiment on an ARM Cortex M4F core.
GPU-based HPC clusters are attracting more scientific application developers due to their extensive parallelism and energy efficiency. In order to achieve portability among a variety of multi/many core architectures, a popular choice for an application developer is to utilize directive-based parallel programming models, such as OpenMP. However, even with OpenMP, the developer must choose from among many strategies for exploiting a GPU or a CPU. Recently, Machine Learning (ML) approaches have brought significant advances in the optimizations of HPC applications. To this end, several ways have been proposed to represent application characteristics for ML models. However, the available techniques fail to capture features that are crucial for exposing parallelism. In this paper, we introduce a new graph-based program representation for parallel applications that extends the Abstract Syntax Tree to represent control and data flow information. The originality of this work lies in the addition of new edges exploiting the implicit ordering and parent-child relationships in ASTs, as well as the introduction of edge weights to account for loop and condition information. We evaluate our proposed representation by training a Graph Neural Network (GNN) to predict the runtime of an OpenMP code region across CPUs and GPUs. Various transformations utilizing collapse and data transfer between the CPU and GPU are used to construct the dataset. The predicted runtime of the model is used to determine which transformation provides the best performance. Results show that our approach is indeed effective and has normalized RMSE as low as 0.004 to at most 0.01 in its runtime predictions.
Interactive image segmentation enables annotators to efficiently perform pixel-level annotation for segmentation tasks. However, the existing interactive segmentation pipeline suffers from inefficient computations of interactive models because of the following two issues. First, annotators' later click is based on models' feedback of annotators' former click. This serial interaction is unable to utilize model's parallelism capabilities. Second, in each interaction step, the model handles the invariant image along with the sparse variable clicks, resulting in a process that's highly repetitive and redundant. For efficient computations, we propose a method named InterFormer that follows a new pipeline to address these issues. InterFormer extracts and preprocesses the computationally time-consuming part i.e. image processing from the existing process. Specifically, InterFormer employs a large vision transformer (ViT) on high-performance devices to preprocess images in parallel, and then uses a lightweight module called interactive multi-head self attention (I-MSA) for interactive segmentation. Furthermore, the I-MSA module's deployment on low-power devices extends the practical application of interactive segmentation. The I-MSA module utilizes the preprocessed features to efficiently response to the annotator inputs in real-time. The experiments on several datasets demonstrate the effectiveness of InterFormer, which outperforms previous interactive segmentation models in terms of computational efficiency and segmentation quality, achieve real-time high-quality interactive segmentation on CPU-only devices. The code is available at https://github.com/YouHuang67/InterFormer.
This paper addresses real-time dense 3D reconstruction for a resource-constrained Autonomous Underwater Vehicle (AUV). Underwater vision-guided operations are among the most challenging as they combine 3D motion in the presence of external forces, limited visibility, and absence of global positioning. Obstacle avoidance and effective path planning require online dense reconstructions of the environment. Autonomous operation is central to environmental monitoring, marine archaeology, resource utilization, and underwater cave exploration. To address this problem, we propose to use SVIn2, a robust VIO method, together with a real-time 3D reconstruction pipeline. We provide extensive evaluation on four challenging underwater datasets. Our pipeline produces comparable reconstruction with that of COLMAP, the state-of-the-art offline 3D reconstruction method, at high frame rates on a single CPU.
Images captured through smartphone cameras often suffer from degradation, blur being one of the major ones, posing a challenge in processing these images for downstream tasks. In this paper we propose low-compute lightweight patch-wise features for image quality assessment. Using our method we can discriminate between blur vs sharp image degradation. To this end, we train a decision-tree based XGBoost model on various intuitive image features like gray level variance, first and second order gradients, texture features like local binary patterns. Experiments conducted on an open dataset show that the proposed low compute method results in 90.1% mean accuracy on the validation set, which is comparable to the accuracy of a compute-intensive VGG16 network with 94% mean accuracy fine-tuned to this task. To demonstrate the generalizability of our proposed features and model we test the model on BHBID dataset and an internal dataset where we attain accuracy of 98% and 91%, respectively. The proposed method is 10x faster than the VGG16 based model on CPU and scales linearly to the input image size making it suitable to be implemented on low compute edge devices.
Processing-in-memory (PIM) promises to alleviate the data movement bottleneck in modern computing systems. However, current real-world PIM systems have the inherent disadvantage that their hardware is more constrained than in conventional processors (CPU, GPU), due to the difficulty and cost of building processing elements near or inside the memory. As a result, general-purpose PIM architectures support fairly limited instruction sets and struggle to execute complex operations such as transcendental functions and other hard-to-calculate operations (e.g., square root). These operations are particularly important for some modern workloads, e.g., activation functions in machine learning applications. In order to provide support for transcendental (and other hard-to-calculate) functions in general-purpose PIM systems, we present \emph{TransPimLib}, a library that provides CORDIC-based and LUT-based methods for trigonometric functions, hyperbolic functions, exponentiation, logarithm, square root, etc. We develop an implementation of TransPimLib for the UPMEM PIM architecture and perform a thorough evaluation of TransPimLib's methods in terms of performance and accuracy, using microbenchmarks and three full workloads (Blackscholes, Sigmoid, Softmax). We open-source all our code and datasets at~\url{https://github.com/CMU-SAFARI/transpimlib}.
Deep learning recommendation models (DLRM) rely on large embedding tables to manage categorical sparse features. Expanding such embedding tables can significantly enhance model performance, but at the cost of increased GPU/CPU/memory usage. Meanwhile, tech companies have built extensive cloud-based services to accelerate training DLRM models at scale. In this paper, we conduct a deep investigation of the DLRM training platforms at AntGroup and reveal two critical challenges: low resource utilization due to suboptimal configurations by users and the tendency to encounter abnormalities due to an unstable cloud environment. To overcome them, we introduce DLRover-RM, an elastic training framework for DLRMs designed to increase resource utilization and handle the instability of a cloud environment. DLRover-RM develops a resource-performance model by considering the unique characteristics of DLRMs and a three-stage heuristic strategy to automatically allocate and dynamically adjust resources for DLRM training jobs for higher resource utilization. Further, DLRover-RM develops multiple mechanisms to ensure efficient and reliable execution of DLRM training jobs. Our extensive evaluation shows that DLRover-RM reduces job completion times by 31%, increases the job completion rate by 6%, enhances CPU usage by 15%, and improves memory utilization by 20%, compared to state-of-the-art resource scheduling frameworks. DLRover-RM has been widely deployed at AntGroup and processes thousands of DLRM training jobs on a daily basis. DLRover-RM is open-sourced and has been adopted by 10+ companies.
Efficient large-scale neural network training and inference on commodity CPU hardware is of immense practical significance in democratizing deep learning (DL) capabilities. Presently, the process of training massive models consisting of hundreds of millions to billions of parameters requires the extensive use of specialized hardware accelerators, such as GPUs, which are only accessible to a limited number of institutions with considerable financial resources. Moreover, there is often an alarming carbon footprint associated with training and deploying these models. In this paper, we take a step towards addressing these challenges by introducing BOLT, a sparse deep learning library for training large-scale search and recommendation models on standard CPU hardware. BOLT provides a flexible, high-level API for constructing models that will be familiar to users of existing popular DL frameworks. By automatically tuning specialized hyperparameters, BOLT also abstracts away the algorithmic details of sparse network training. We evaluate BOLT on a number of information retrieval tasks including product recommendations, text classification, graph neural networks, and personalization. We find that our proposed system achieves competitive performance with state-of-the-art techniques at a fraction of the cost and energy consumption and an order-of-magnitude faster inference time. BOLT has also been successfully deployed by multiple businesses to address critical problems, and we highlight one customer case study in the field of e-commerce.
Video instance segmentation on mobile devices is an important yet very challenging edge AI problem. It mainly suffers from (1) heavy computation and memory costs for frame-by-frame pixel-level instance perception and (2) complicated heuristics for tracking objects. To address those issues, we present MobileInst, a lightweight and mobile-friendly framework for video instance segmentation on mobile devices. Firstly, MobileInst adopts a mobile vision transformer to extract multi-level semantic features and presents an efficient query-based dual-transformer instance decoder for mask kernels and a semantic-enhanced mask decoder to generate instance segmentation per frame. Secondly, MobileInst exploits simple yet effective kernel reuse and kernel association to track objects for video instance segmentation. Further, we propose temporal query passing to enhance the tracking ability for kernels. We conduct experiments on COCO and YouTube-VIS datasets to demonstrate the superiority of MobileInst and evaluate the inference latency on one single CPU core of Snapdragon 778G Mobile Platform, without other methods of acceleration. On the COCO dataset, MobileInst achieves 31.2 mask AP and 433 ms on the mobile CPU, which reduces the latency by 50% compared to the previous SOTA. For video instance segmentation, MobileInst achieves 35.0 AP on YouTube-VIS 2019 and 30.1 AP on YouTube-VIS 2021. Code will be available to facilitate real-world applications and future research.
There are plenty of graph neural network (GNN) accelerators being proposed. However, they highly rely on users' hardware expertise and are usually optimized for one specific GNN model, making them challenging for practical use. Therefore, in this work, we propose GNNBuilder, the first automated, generic, end-to-end GNN accelerator generation framework. It features four advantages: (1) GNNBuilder can automatically generate GNN accelerators for a wide range of GNN models arbitrarily defined by users; (2) GNNBuilder takes standard PyTorch programming interface, introducing zero overhead for algorithm developers; (3) GNNBuilder supports end-to-end code generation, simulation, accelerator optimization, and hardware deployment, realizing a push-button fashion for GNN accelerator design; (4) GNNBuilder is equipped with accurate performance models of its generated accelerator, enabling fast and flexible design space exploration (DSE). In the experiments, first, we show that our accelerator performance model has errors within $36\%$ for latency prediction and $18\%$ for BRAM count prediction. Second, we show that our generated accelerators can outperform CPU by $6.33\times$ and GPU by $6.87\times$. This framework is open-source, and the code is available at https://github.com/sharc-lab/gnn-builder.
Graph embedding maps graph nodes to low-dimensional vectors, and is widely adopted in machine learning tasks. The increasing availability of billion-edge graphs underscores the importance of learning efficient and effective embeddings on large graphs, such as link prediction on Twitter with over one billion edges. Most existing graph embedding methods fall short of reaching high data scalability. In this paper, we present a general-purpose, distributed, information-centric random walk-based graph embedding framework, DistGER, which can scale to embed billion-edge graphs. DistGER incrementally computes information-centric random walks. It further leverages a multi-proximity-aware, streaming, parallel graph partitioning strategy, simultaneously achieving high local partition quality and excellent workload balancing across machines. DistGER also improves the distributed Skip-Gram learning model to generate node embeddings by optimizing the access locality, CPU throughput, and synchronization efficiency. Experiments on real-world graphs demonstrate that compared to state-of-the-art distributed graph embedding frameworks, including KnightKing, DistDGL, and Pytorch-BigGraph, DistGER exhibits 2.33x-129x acceleration, 45% reduction in cross-machines communication, and > 10% effectiveness improvement in downstream tasks.
Functional regression analysis is an established tool for many contemporary scientific applications. Regression problems involving large and complex data sets are ubiquitous, and feature selection is crucial for avoiding overfitting and achieving accurate predictions. We propose a new, flexible and ultra-efficient approach to perform feature selection in a sparse high dimensional function-on-function regression problem, and we show how to extend it to the scalar-on-function framework. Our method, called FAStEN, combines functional data, optimization, and machine learning techniques to perform feature selection and parameter estimation simultaneously. We exploit the properties of Functional Principal Components and the sparsity inherent to the Dual Augmented Lagrangian problem to significantly reduce computational cost, and we introduce an adaptive scheme to improve selection accuracy. In addition, we derive asymptotic oracle properties, which guarantee estimation and selection consistency for the proposed FAStEN estimator. Through an extensive simulation study, we benchmark our approach to the best existing competitors and demonstrate a massive gain in terms of CPU time and selection performance, without sacrificing the quality of the coefficients' estimation. The theoretical derivations and the simulation study provide a strong motivation for our approach. Finally, we present an application to brain fMRI data from the AOMIC PIOP1 study. Complete FAStEN code is provided at https://github.com/IBM/funGCN.
Positioning problem can be solved with several approaches in 5G. This paper introduces RT (Ray Tracing) technique for small-cell Outdoor positioning and tracking for systems with distributed architecture with multipath processing gain. Proposed approach exploits high-resolution angular spectrum estimation and known radio propagation environment. It solves positioning problem robustly even in NLOS cases via CoMP (Cooperative Multi Point) reception and joint processing at cloud CPU. The aim of this paper is to disclose proposed algorithm for NR and compare it to GNSS measurements.
Many ML applications and products train on medium amounts of input data but get bottlenecked in real-time inference. When implementing ML systems, conventional wisdom favors segregating ML code into services queried by product code via Remote Procedure Call (RPC) APIs. This approach clarifies the overall software architecture and simplifies product code by abstracting away ML internals. However, the separation adds network latency and entails additional CPU overhead. Hence, we simplify inference algorithms and embed them into the product code to reduce network communication. For public datasets and a high-performance real-time platform that deals with tabular data, we show that over half of the inputs are often amenable to such optimization, while the remainder can be handled by the original model. By applying our optimization with AutoML to both training and inference, we reduce inference latency by 1.3x, CPU resources by 30%, and network communication between application front-end and ML back-end by about 50% for a commercial end-to-end ML platform that serves millions of real-time decisions per second.
Large pretrained language models have achieved state-of-the-art results on a variety of downstream tasks. Knowledge Distillation (KD) into a smaller student model addresses their inefficiency, allowing for deployment in resource-constrained environments. However, KD can be ineffective when the student is manually selected from a set of existing options, since it can be a sub-optimal choice within the space of all possible student architectures. We develop multilingual KD-NAS, the use of Neural Architecture Search (NAS) guided by KD to find the optimal student architecture for task agnostic distillation from a multilingual teacher. In each episode of the search process, a NAS controller predicts a reward based on the distillation loss and latency of inference. The top candidate architectures are then distilled from the teacher on a small proxy set. Finally the architecture(s) with the highest reward is selected, and distilled on the full training corpus. KD-NAS can automatically trade off efficiency and effectiveness, and recommends architectures suitable to various latency budgets. Using our multi-layer hidden state distillation process, our KD-NAS student model achieves a 7x speedup on CPU inference (2x on GPU) compared to a XLM-Roberta Base Teacher, while maintaining 90% performance, and has been deployed in 3 software offerings requiring large throughput, low latency and deployment on CPU.