PyTorchKR Papers

최신 논문

총 2,153편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2208.14754공식 코드 · Papers with Code

LexMAE: Lexicon-Bottlenecked Pretraining for Large-Scale Retrieval

In large-scale retrieval, the lexicon-weighting paradigm, learning weighted sparse representations in vocabulary space, has shown promising results with high quality and low latency. Despite it deeply exploiting the lexicon-representing capability of pre-trained language models, a crucial gap remains between language modeling and lexicon-weighting retrieval -- the former preferring certain or low-entropy words whereas the latter favoring pivot or high-entropy words -- becoming the main barrier to lexicon-weighting performance for large-scale retrieval. To bridge this gap, we propose a brand-new pre-training framework, lexicon-bottlenecked masked autoencoder (LexMAE), to learn importance-aware lexicon representations. Essentially, we present a lexicon-bottlenecked module between a normal language modeling encoder and a weakened decoder, where a continuous bag-of-words bottleneck is constructed to learn a lexicon-importance distribution in an unsupervised fashion. The pre-trained LexMAE is readily transferred to the lexicon-weighting retrieval via fine-tuning. On the ad-hoc retrieval benchmark, MS-Marco, it achieves 42.6% MRR@10 with 45.8 QPS for the passage dataset and 44.4% MRR@100 with 134.8 QPS for the document dataset, by a CPU machine. And LexMAE shows state-of-the-art zero-shot transfer capability on BEIR benchmark with 12 datasets.

arXiv:2208.14124

Towards making the most of NLP-based device mapping optimization for OpenCL kernels

Nowadays, we are living in an era of extreme device heterogeneity. Despite the high variety of conventional CPU architectures, accelerator devices, such as GPUs and FPGAs, also appear in the foreground exploding the pool of available solutions to execute applications. However, choosing the appropriate device per application needs is an extremely challenging task due to the abstract relationship between hardware and software. Automatic optimization algorithms that are accurate are required to cope with the complexity and variety of current hardware and software. Optimal execution has always relied on time-consuming trial and error approaches. Machine learning (ML) and Natural Language Processing (NLP) has flourished over the last decade with research focusing on deep architectures. In this context, the use of natural language processing techniques to source code in order to conduct autotuning tasks is an emerging field of study. In this paper, we extend the work of Cummins et al., namely Deeptune, that tackles the problem of optimal device selection (CPU or GPU) for accelerated OpenCL kernels. We identify three major limitations of Deeptune and, based on these, we propose four different DNN models that provide enhanced contextual information of source codes. Experimental results show that our proposed methodology surpasses that of Cummins et al. work, providing up to 4\% improvement in prediction accuracy.

arXiv:2208.14059

Distributed CPU Scheduling Subject to Nonlinear Constraints

This paper considers a network of collaborating agents for local resource allocation subject to nonlinear model constraints. In many applications, it is required (or desirable) that the solution be anytime feasible in terms of satisfying the sum-preserving global constraint. Motivated by this, sufficient conditions on the nonlinear mapping for anytime feasibility (or non-asymptotic feasibility) are addressed in this paper. For the two proposed distributed solutions, one converges over directed weight-balanced networks and the other one over undirected networks. In particular, we elaborate on uniform quantization and discuss the notion of {\epsilon}-accurate solution, which gives an estimate of how close we can get to the exact optimizer subject to different quantization levels. This work, further, handles general (possibly non-quadratic) strictly convex objective functions with application to CPU allocation among a cloud of data centers via distributed solutions. The results can be used as a coordination mechanism to optimally balance the tasks and CPU resources among a group of networked servers while addressing quantization or limited server capacity. Index Terms: multi-agent systems, sum-preserving resource allocation, distributed optimization, anytime feasibility

arXiv:2208.13958

Joint Optimization of Resource Allocation, Phase Shift and UAV Trajectory for Energy-Efficient RIS-Assisted UAV-Enabled MEC Systems

The unmanned aerial vehicle (UAV) enabled mobile edge computing (MEC) has been deemed a promising paradigm to provide ubiquitous communication and computing services for the Internet of Things (IoT). Besides, by intelligently reflecting the received signals, the reconfigurable intelligent surface (RIS) can significantly improve the propagation environment and further enhance the service quality of the UAV-enabled MEC. Motivated by this vision, in this paper, we consider both the amount of completed task bits and the energy consumption to maximize the energy efficiency of the RIS-assisted UAV-enabled MEC systems, where the bit allocation, transmit power, phase shift, and UAV trajectory are jointly optimized by an iterative algorithm with a double-loop structure based on the Dinkelbach's method and block coordinate decent (BCD) technique. Simulation results demonstrate that: 1) with the deployment of RIS, our proposed algorithm can achieve higher energy efficiency than baseline schemes while satisfying the task tolerance latency; 2) the energy efficiency first increases and then decreases with the increase of the mission period and the total amount of task-input bits of IoT devices; 3) when the CPU cycles required for computing 1-bit of task-input data becomes larger, more task bits will be offloaded to the UAV while the energy efficiency will be decreased.

arXiv:2208.14390공식 코드 · Papers with Code

k-MS: A novel clustering algorithm based on morphological reconstruction

This work proposes a clusterization algorithm called k-Morphological Sets (k-MS), based on morphological reconstruction and heuristics. k-MS is faster than the CPU-parallel k-Means in worst case scenarios and produces enhanced visualizations of the dataset as well as very distinct clusterizations. It is also faster than similar clusterization methods that are sensitive to density and shapes such as Mitosis and TRICLUST. In addition, k-MS is deterministic and has an intrinsic sense of maximal clusters that can be created for a given input sample and input parameters, differing from k-Means and other clusterization algorithms. In other words, given a constant k, a structuring element and a dataset, k-MS produces k or less clusters without using random/ pseudo-random functions. Finally, the proposed algorithm also provides a straightforward means for removing noise from images or datasets in general.

arXiv:2208.13285

Computing with Hypervectors for Efficient Speaker Identification

We introduce a method to identify speakers by computing with high-dimensional random vectors. Its strengths are simplicity and speed. With only 1.02k active parameters and a 128-minute pass through the training data we achieve Top-1 and Top-5 scores of 31% and 52% on the VoxCeleb1 dataset of 1,251 speakers. This is in contrast to CNN models requiring several million parameters and orders of magnitude higher computational complexity for only a 2$\times$ gain in discriminative power as measured in mutual information. An additional 92 seconds of training with Generalized Learning Vector Quantization (GLVQ) raises the scores to 48% and 67%. A trained classifier classifies 1 second of speech in 5.7 ms. All processing was done on standard CPU-based machines.

arXiv:2208.12187공식 코드 · Papers with Code

JAXFit: Trust Region Method for Nonlinear Least-Squares Curve Fitting on the GPU

We implement a trust region method on the GPU for nonlinear least squares curve fitting problems using a new deep learning Python library called JAX. Our open source package, JAXFit, works for both unconstrained and constrained curve fitting problems and allows the fit functions to be defined in Python alone -- without any specialized knowledge of either the GPU or CUDA programming. Since JAXFit runs on the GPU, it is much faster than CPU based libraries and even other GPU based libraries, despite being very easy to use. Additionally, due to JAX's deep learning foundations, the Jacobian in JAXFit's trust region algorithm is calculated with automatic differentiation, rather than than using derivative approximations or requiring the user to define the fit function's partial derivatives.

arXiv:2208.11666

Efficient Heterogeneous Video Segmentation at the Edge

We introduce an efficient video segmentation system for resource-limited edge devices leveraging heterogeneous compute. Specifically, we design network models by searching across multiple dimensions of specifications for the neural architectures and operations on top of already light-weight backbones, targeting commercially available edge inference engines. We further analyze and optimize the heterogeneous data flows in our systems across the CPU, the GPU and the NPU. Our approach has empirically factored well into our real-time AR system, enabling remarkably higher accuracy with quadrupled effective resolutions, yet at much shorter end-to-end latency, much higher frame rate, and even lower power consumption on edge platforms.

arXiv:2208.11527

Fast and Precise Binary Instance Segmentation of 2D Objects for Automotive Applications

In this paper, we focus on improving binary 2D instance segmentation to assist humans in labeling ground truth datasets with polygons. Humans labeler just have to draw boxes around objects, and polygons are generated automatically. To be useful, our system has to run on CPUs in real-time. The most usual approach for binary instance segmentation involves encoder-decoder networks. This report evaluates state-of-the-art encoder-decoder networks and proposes a method for improving instance segmentation quality using these networks. Alongside network architecture improvements, our proposed method relies upon providing extra information to the network input, so-called extreme points, i.e. the outermost points on the object silhouette. The user can label them instead of a bounding box almost as quickly. The bounding box can be deduced from the extreme points as well. This method produces better IoU compared to other state-of-the-art encoder-decoder networks and also runs fast enough when it is deployed on a CPU.

arXiv:2208.09151공식 코드 · Papers with Code

Ginex: SSD-enabled Billion-scale Graph Neural Network Training on a Single Machine via Provably Optimal In-memory Caching

Recently, Graph Neural Networks (GNNs) have been receiving a spotlight as a powerful tool that can effectively serve various inference tasks on graph structured data. As the size of real-world graphs continues to scale, the GNN training system faces a scalability challenge. Distributed training is a popular approach to address this challenge by scaling out CPU nodes. However, not much attention has been paid to disk-based GNN training, which can scale up the single-node system in a more cost-effective manner by leveraging high-performance storage devices like NVMe SSDs. We observe that the data movement between the main memory and the disk is the primary bottleneck in the SSD-based training system, and that the conventional GNN training pipeline is sub-optimal without taking this overhead into account. Thus, we propose Ginex, the first SSD-based GNN training system that can process billion-scale graph datasets on a single machine. Inspired by the inspector-executor execution model in compiler optimization, Ginex restructures the GNN training pipeline by separating sample and gather stages. This separation enables Ginex to realize a provably optimal replacement algorithm, known as Belady's algorithm, for caching feature vectors in memory, which account for the dominant portion of I/O accesses. According to our evaluation with four billion-scale graph datasets, Ginex achieves 2.11x higher training throughput on average (up to 2.67x at maximum) than the SSD-extended PyTorch Geometric.

arXiv:2208.08711공식 코드 · Papers with Code

L3: Accelerator-Friendly Lossless Image Format for High-Resolution, High-Throughput DNN Training

The training process of deep neural networks (DNNs) is usually pipelined with stages for data preparation on CPUs followed by gradient computation on accelerators like GPUs. In an ideal pipeline, the end-to-end training throughput is eventually limited by the throughput of the accelerator, not by that of data preparation. In the past, the DNN training pipeline achieved a near-optimal throughput by utilizing datasets encoded with a lightweight, lossy image format like JPEG. However, as high-resolution, losslessly-encoded datasets become more popular for applications requiring high accuracy, a performance problem arises in the data preparation stage due to low-throughput image decoding on the CPU. Thus, we propose L3, a custom lightweight, lossless image format for high-resolution, high-throughput DNN training. The decoding process of L3 is effectively parallelized on the accelerator, thus minimizing CPU intervention for data preparation during DNN training. L3 achieves a 9.29x higher data preparation throughput than PNG, the most popular lossless image format, for the Cityscapes dataset on NVIDIA A100 GPU, which leads to 1.71x higher end-to-end training throughput. Compared to JPEG and WebP, two popular lossy image formats, L3 provides up to 1.77x and 2.87x higher end-to-end training throughput for ImageNet, respectively, at equivalent metric performance.

arXiv:2208.08706공식 코드 · Papers with Code

Musika! Fast Infinite Waveform Music Generation

Fast and user-controllable music generation could enable novel ways of composing or performing music. However, state-of-the-art music generation systems require large amounts of data and computational resources for training, and are slow at inference. This makes them impractical for real-time interactive use. In this work, we introduce Musika, a music generation system that can be trained on hundreds of hours of music using a single consumer GPU, and that allows for much faster than real-time generation of music of arbitrary length on a consumer CPU. We achieve this by first learning a compact invertible representation of spectrogram magnitudes and phases with adversarial autoencoders, then training a Generative Adversarial Network (GAN) on this representation for a particular music domain. A latent coordinate system enables generating arbitrarily long sequences of excerpts in parallel, while a global context vector allows the music to remain stylistically coherent through time. We perform quantitative evaluations to assess the quality of the generated samples and showcase options for user control in piano and techno music generation. We release the source code and pretrained autoencoder weights at github.com/marcoppasini/musika, such that a GAN can be trained on a new music domain with a single GPU in a matter of hours.

arXiv:2208.07657

Uconv-Conformer: High Reduction of Input Sequence Length for End-to-End Speech Recognition

Optimization of modern ASR architectures is among the highest priority tasks since it saves many computational resources for model training and inference. The work proposes a new Uconv-Conformer architecture based on the standard Conformer model. It consistently reduces the input sequence length by 16 times, which results in speeding up the work of the intermediate layers. To solve the convergence issue connected with such a significant reduction of the time dimension, we use upsampling blocks like in the U-Net architecture to ensure the correct CTC loss calculation and stabilize network training. The Uconv-Conformer architecture appears to be not only faster in terms of training and inference speed but also shows better WER compared to the baseline Conformer. Our best Uconv-Conformer model shows 47.8% and 23.5% inference acceleration on the CPU and GPU, respectively. Relative WER reduction is 7.3% and 9.2% on LibriSpeech test_clean and test_other respectively.

arXiv:2208.06555공식 코드 · Papers with Code

BenchPress: A Deep Active Benchmark Generator

We develop BenchPress, the first ML benchmark generator for compilers that is steerable within feature space representations of source code. BenchPress synthesizes compiling functions by adding new code in any part of an empty or existing sequence by jointly observing its left and right context, achieving excellent compilation rate. BenchPress steers benchmark generation towards desired target features that has been impossible for state of the art synthesizers (or indeed humans) to reach. It performs better in targeting the features of Rodinia benchmarks in 3 different feature spaces compared with (a) CLgen - a state of the art ML synthesizer, (b) CLSmith fuzzer, (c) SRCIROR mutator or even (d) human-written code from GitHub. BenchPress is the first generator to search the feature space with active learning in order to generate benchmarks that will improve a downstream task. We show how using BenchPress, Grewe's et al. CPU vs GPU heuristic model can obtain a higher speedup when trained on BenchPress's benchmarks compared to other techniques. BenchPress is a powerful code generator: Its generated samples compile at a rate of 86%, compared to CLgen's 2.33%. Starting from an empty fixed input, BenchPress produces 10x more unique, compiling OpenCL benchmarks than CLgen, which are significantly larger and more feature diverse.

arXiv:2208.06150공식 코드 · Papers with Code

Pre-training Tasks for User Intent Detection and Embedding Retrieval in E-commerce Search

BERT-style models pre-trained on the general corpus (e.g., Wikipedia) and fine-tuned on specific task corpus, have recently emerged as breakthrough techniques in many NLP tasks: question answering, text classification, sequence labeling and so on. However, this technique may not always work, especially for two scenarios: a corpus that contains very different text from the general corpus Wikipedia, or a task that learns embedding spacial distribution for a specific purpose (e.g., approximate nearest neighbor search). In this paper, to tackle the above two scenarios that we have encountered in an industrial e-commerce search system, we propose customized and novel pre-training tasks for two critical modules: user intent detection and semantic embedding retrieval. The customized pre-trained models after fine-tuning, being less than 10% of BERT-base's size in order to be feasible for cost-efficient CPU serving, significantly improve the other baseline models: 1) no pre-training model and 2) fine-tuned model from the official pre-trained BERT using general corpus, on both offline datasets and online system. We have open sourced our datasets for the sake of reproducibility and future works.

arXiv:2208.05215

User Association and Hybrid Beamforming Designs for Cooperative mmWave MIMO Systems

Hybrid analog and digital beamforming has emerged as a key enabling technology for millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) communication systems since it can balance the trade-off between system performance and hardware efficiency. Owing to the strong ability of central control, cooperative networks show great potential to enhance the spectral efficiency of mmWave communications. In this paper, we consider cooperative mmWave MIMO systems and propose user association and hybrid beamforming design algorithms for three typical hybrid beamforming architectures. The central processing unit (CPU) of the cooperative networks first matches the service pairs of base stations (BSs) and users. Then, an iterative hybrid beamforming design algorithm is proposed to maximize the weighted achievable sum-rate performance of the mmWave MIMO system with fully connected hybrid beamforming architecture. Moreover, a heuristic analog beamforming design algorithm is introduced for the fixed subarray hybrid beamforming architecture. In an effort to further exploit multiple-antenna diversities, we also consider the dynamic subarray architecture and propose a novel antenna design algorithm for the analog beamforming design. Simulation results illustrate that the proposed hybrid beamforming algorithms achieve a significant performance improvement than other existing approaches and the dynamic subarray architecture has great advantages of improving the energy efficiency (EE) performance.

arXiv:2208.05210

Partially Distributed Beamforming Design for RIS-Aided Cell-Free Networks

Cell-free networks are regarded as a promising technology to meet higher rate requirements for beyond fifth-generation (5G) communications. Most works on cell-free networks focus on either fully centralized beamforming to maximally enhance system performance, or fully distributed beamforming to avoid extensive channel state information (CSI) exchange among access points (APs). In order to achieve both network capacity improvement and CSI exchange reduction, we propose a partially distributed beamforming design algorithm for reconfigurable intelligent surface (RIS)-aided cell-free networks. We aim at maximizing the weighted sum-rate of all users by designing active and passive beamforming subject to transmit power constraints of APs and unit-modulus constraints of RIS elements. The weighted sum-rate maximization problem is first transformed into an equivalent weighted sum-mean-square-error (sum-MSE) minimization problem, and then alternating optimization (AO) approach is adopted to iteratively design active and passive beamformer. Specifically, active beamforming vectors are obtained by local APs and passive beamforming vector is optimized by central processing unit (CPU). Numerical results not only illustrate the proposed partially distributed algorithm achieves the remarkable performance improvement compared with conventional local beamforming methods, but also further show the considerable potential of deploying RIS in cell-free networks.

arXiv:2208.05321공식 코드 · Papers with Code

A Frequency-aware Software Cache for Large Recommendation System Embeddings

Deep learning recommendation models (DLRMs) have been widely applied in Internet companies. The embedding tables of DLRMs are too large to fit on GPU memory entirely. We propose a GPU-based software cache approaches to dynamically manage the embedding table in the CPU and GPU memory space by leveraging the id's frequency statistics of the target dataset. Our proposed software cache is efficient in training entire DLRMs on GPU in a synchronized update manner. It is also scaled to multiple GPUs in combination with the widely used hybrid parallel training approaches. Evaluating our prototype system shows that we can keep only 1.5% of the embedding parameters in the GPU to obtain a decent end-to-end training speed.

arXiv:2208.04238

POSET-RL: Phase ordering for Optimizing Size and Execution Time using Reinforcement Learning

The ever increasing memory requirements of several applications has led to increased demands which might not be met by embedded devices. Constraining the usage of memory in such cases is of paramount importance. It is important that such code size improvements should not have a negative impact on the runtime. Improving the execution time while optimizing for code size is a non-trivial but a significant task. The ordering of standard optimization sequences in modern compilers is fixed, and are heuristically created by the compiler domain experts based on their expertise. However, this ordering is sub-optimal, and does not generalize well across all the cases. We present a reinforcement learning based solution to the phase ordering problem, where the ordering improves both the execution time and code size. We propose two different approaches to model the sequences: one by manual ordering, and other based on a graph called Oz Dependence Graph (ODG). Our approach uses minimal data as training set, and is integrated with LLVM. We show results on x86 and AArch64 architectures on the benchmarks from SPEC-CPU 2006, SPEC-CPU 2017 and MiBench. We observe that the proposed model based on ODG outperforms the current Oz sequence both in terms of size and execution time by 6.19% and 11.99% in SPEC 2017 benchmarks, on an average.

arXiv:2208.03646

A Length Adaptive Algorithm-Hardware Co-design of Transformer on FPGA Through Sparse Attention and Dynamic Pipelining

Transformers are considered one of the most important deep learning models since 2018, in part because it establishes state-of-the-art (SOTA) records and could potentially replace existing Deep Neural Networks (DNNs). Despite the remarkable triumphs, the prolonged turnaround time of Transformer models is a widely recognized roadblock. The variety of sequence lengths imposes additional computing overhead where inputs need to be zero-padded to the maximum sentence length in the batch to accommodate the parallel computing platforms. This paper targets the field-programmable gate array (FPGA) and proposes a coherent sequence length adaptive algorithm-hardware co-design for Transformer acceleration. Particularly, we develop a hardware-friendly sparse attention operator and a length-aware hardware resource scheduling algorithm. The proposed sparse attention operator brings the complexity of attention-based models down to linear complexity and alleviates the off-chip memory traffic. The proposed length-aware resource hardware scheduling algorithm dynamically allocates the hardware resources to fill up the pipeline slots and eliminates bubbles for NLP tasks. Experiments show that our design has very small accuracy loss and has 80.2 $\times$ and 2.6 $\times$ speedup compared to CPU and GPU implementation, and 4 $\times$ higher energy efficiency than state-of-the-art GPU accelerator optimized via CUBLAS GEMM.