PyTorchKR Papers

최신 논문

총 5,357편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2404.08002

ApproxDARTS: Differentiable Neural Architecture Search with Approximate Multipliers

Integrating the principles of approximate computing into the design of hardware-aware deep neural networks (DNN) has led to DNNs implementations showing good output quality and highly optimized hardware parameters such as low latency or inference energy. In this work, we present ApproxDARTS, a neural architecture search (NAS) method enabling the popular differentiable neural architecture search method called DARTS to exploit approximate multipliers and thus reduce the power consumption of generated neural networks. We showed on the CIFAR-10 data set that the ApproxDARTS is able to perform a complete architecture search within less than $10$ GPU hours and produce competitive convolutional neural networks (CNN) containing approximate multipliers in convolutional layers. For example, ApproxDARTS created a CNN showing an energy consumption reduction of (a) $53.84\%$ in the arithmetic operations of the inference phase compared to the CNN utilizing the native $32$-bit floating-point multipliers and (b) $5.97\%$ compared to the CNN utilizing the exact $8$-bit fixed-point multipliers, in both cases with a negligible accuracy drop. Moreover, the ApproxDARTS is $2.3\times$ faster than a similar but evolutionary algorithm-based method called EvoApproxNAS.

arXiv:2404.08399

Mitigating Challenges of the Space Environment for Onboard Artificial Intelligence: Design Overview of the Imaging Payload on SpIRIT

Artificial intelligence (AI) and autonomous edge computing in space are emerging areas of interest to augment capabilities of nanosatellites, where modern sensors generate orders of magnitude more data than can typically be transmitted to mission control. Here, we present the hardware and software design of an onboard AI subsystem hosted on SpIRIT. The system is optimised for on-board computer vision experiments based on visible light and long wave infrared cameras. This paper highlights the key design choices made to maximise the robustness of the system in harsh space conditions, and their motivation relative to key mission requirements, such as limited compute resources, resilience to cosmic radiation, extreme temperature variations, distribution shifts, and very low transmission bandwidths. The payload, called Loris, consists of six visible light cameras, three infrared cameras, a camera control board and a Graphics Processing Unit (GPU) system-on-module. Loris enables the execution of AI models with on-orbit fine-tuning as well as a next-generation image compression algorithm, including progressive coding. This innovative approach not only enhances the data processing capabilities of nanosatellites but also lays the groundwork for broader applications to remote sensing from space.

arXiv:2404.08080

Variance-reduced Zeroth-Order Methods for Fine-Tuning Language Models

Fine-tuning language models (LMs) has demonstrated success in a wide array of downstream tasks. However, as LMs are scaled up, the memory requirements for backpropagation become prohibitively high. Zeroth-order (ZO) optimization methods can leverage memory-efficient forward passes to estimate gradients. More recently, MeZO, an adaptation of ZO-SGD, has been shown to consistently outperform zero-shot and in-context learning when combined with suitable task prompts. In this work, we couple ZO methods with variance reduction techniques to enhance stability and convergence for inference-based LM fine-tuning. We introduce Memory-Efficient Zeroth-Order Stochastic Variance-Reduced Gradient (MeZO-SVRG) and demonstrate its efficacy across multiple LM fine-tuning tasks, eliminating the reliance on task-specific prompts. Evaluated across a range of both masked and autoregressive LMs on benchmark GLUE tasks, MeZO-SVRG outperforms MeZO with up to 20% increase in test accuracies in both full- and partial-parameter fine-tuning settings. MeZO-SVRG benefits from reduced computation time as it often surpasses MeZO's peak test accuracy with a $2\times$ reduction in GPU-hours. MeZO-SVRG significantly reduces the required memory footprint compared to first-order SGD, i.e. by $2\times$ for autoregressive models. Our experiments highlight that MeZO-SVRG's memory savings progressively improve compared to SGD with larger batch sizes.

arXiv:2404.07413공식 코드 · Papers with Code

JetMoE: Reaching Llama2 Performance with 0.1M Dollars

Large Language Models (LLMs) have achieved remarkable results, but their increasing resource demand has become a major obstacle to the development of powerful and accessible super-human intelligence. This report introduces JetMoE-8B, a new LLM trained with less than $0.1 million, using 1.25T tokens from carefully mixed open-source corpora and 30,000 H100 GPU hours. Despite its low cost, the JetMoE-8B demonstrates impressive performance, with JetMoE-8B outperforming the Llama2-7B model and JetMoE-8B-Chat surpassing the Llama2-13B-Chat model. These results suggest that LLM training can be much more cost-effective than generally thought. JetMoE-8B is based on an efficient Sparsely-gated Mixture-of-Experts (SMoE) architecture, composed of attention and feedforward experts. Both layers are sparsely activated, allowing JetMoE-8B to have 8B parameters while only activating 2B for each input token, reducing inference computation by about 70% compared to Llama2-7B. Moreover, JetMoE-8B is highly open and academia-friendly, using only public datasets and training code. All training parameters and data mixtures have been detailed in this report to facilitate future efforts in the development of open foundation models. This transparency aims to encourage collaboration and further advancements in the field of accessible and efficient LLMs. The model weights are publicly available at https://github.com/myshell-ai/JetMoE.

arXiv:2404.04270

Accelerating Recommender Model Training by Dynamically Skipping Stale Embeddings

Training recommendation models pose significant challenges regarding resource utilization and performance. Prior research has proposed an approach that categorizes embeddings into popular and non-popular classes to reduce the training time for recommendation models. We observe that, even among the popular embeddings, certain embeddings undergo rapid training and exhibit minimal subsequent variation, resulting in saturation. Consequently, updates to these embeddings lack any contribution to model quality. This paper presents Slipstream, a software framework that identifies stale embeddings on the fly and skips their updates to enhance performance. This capability enables Slipstream to achieve substantial speedup, optimize CPU-GPU bandwidth usage, and eliminate unnecessary memory access. SlipStream showcases training time reductions of 2x, 2.4x, 1.2x, and 1.175x across real-world datasets and configurations, compared to Baseline XDL, Intel-optimized DRLM, FAE, and Hotline, respectively.

arXiv:2404.07177공식 코드 · Papers with Code

Scaling Laws for Data Filtering -- Data Curation cannot be Compute Agnostic

Vision-language models (VLMs) are trained for thousands of GPU hours on carefully curated web datasets. In recent times, data curation has gained prominence with several works developing strategies to retain 'high-quality' subsets of 'raw' scraped data. For instance, the LAION public dataset retained only 10% of the total crawled data. However, these strategies are typically developed agnostic of the available compute for training. In this paper, we first demonstrate that making filtering decisions independent of training compute is often suboptimal: the limited high-quality data rapidly loses its utility when repeated, eventually requiring the inclusion of 'unseen' but 'lower-quality' data. To address this quality-quantity tradeoff ($\texttt{QQT}$), we introduce neural scaling laws that account for the non-homogeneous nature of web data, an angle ignored in existing literature. Our scaling laws (i) characterize the $\textit{differing}$ 'utility' of various quality subsets of web data; (ii) account for how utility diminishes for a data point at its 'nth' repetition; and (iii) formulate the mutual interaction of various data pools when combined, enabling the estimation of model performance on a combination of multiple data pools without ever jointly training on them. Our key message is that data curation $\textit{cannot}$ be agnostic of the total compute that a model will be trained for. Our scaling laws allow us to curate the best possible pool for achieving top performance on Datacomp at various compute budgets, carving out a pareto-frontier for data curation. Code is available at https://github.com/locuslab/scaling_laws_data_filtering.

arXiv:2404.07188

GCV-Turbo: End-to-end Acceleration of GNN-based Computer Vision Tasks on FPGA

Graph neural networks (GNNs) have recently empowered various novel computer vision (CV) tasks. In GNN-based CV tasks, a combination of CNN layers and GNN layers or only GNN layers are employed. This paper introduces GCV-Turbo, a domain-specific accelerator on FPGA for end-to-end acceleration of GNN-based CV tasks. GCV-Turbo consists of two key components: (1) a \emph{novel} hardware architecture optimized for the computation kernels in both CNNs and GNNs using the same set of computation resources. (2) a PyTorch-compatible compiler that takes a user-defined model as input, performs end-to-end optimization for the computation graph of a given GNN-based CV task, and produces optimized code for hardware execution. The hardware architecture and the compiler work synergistically to support a variety of GNN-based CV tasks. We implement GCV-Turbo on a state-of-the-art FPGA and evaluate its performance across six representative GNN-based CV tasks with diverse input data modalities (e.g., image, human skeleton, point cloud). Compared with state-of-the-art CPU (GPU) implementations, GCV-Turbo achieves an average latency reduction of $68.4\times$ ($4.1\times$) on these six GNN-based CV tasks. Moreover, GCV-Turbo supports the execution of the standalone CNNs or GNNs, achieving performance comparable to that of state-of-the-art CNN (GNN) accelerators for widely used CNN-only (GNN-only) models.

arXiv:2404.06744

YOLO based Ocean Eddy Localization with AWS SageMaker

Ocean eddies play a significant role both on the sea surface and beneath it, contributing to the sustainability of marine life dependent on oceanic behaviors. Therefore, it is crucial to investigate ocean eddies to monitor changes in the Earth, particularly in the oceans, and their impact on climate. This study aims to pinpoint ocean eddies using AWS cloud services, specifically SageMaker. The primary objective is to detect small-scale (<20km) ocean eddies from satellite remote images and assess the feasibility of utilizing SageMaker, which offers tools for deploying AI applications. Moreover, this research not only explores the deployment of cloud-based services for remote sensing of Earth data but also evaluates several YOLO (You Only Look Once) models using single and multi-GPU-based services in the cloud. Furthermore, this study underscores the potential of these services, their limitations, challenges related to deployment and resource management, and their user-riendliness for Earth science projects.

arXiv:2404.07164공식 코드 · Papers with Code

PIM-Opt: Demystifying Distributed Optimization Algorithms on a Real-World Processing-In-Memory System

Modern Machine Learning (ML) training on large-scale datasets is a very time-consuming workload. It relies on the optimization algorithm Stochastic Gradient Descent (SGD) due to its effectiveness, simplicity, and generalization performance. Processor-centric architectures (e.g., CPUs, GPUs) commonly used for modern ML training workloads based on SGD are bottlenecked by data movement between the processor and memory units due to the poor data locality in accessing large datasets. As a result, processor-centric architectures suffer from low performance and high energy consumption while executing ML training workloads. Processing-In-Memory (PIM) is a promising solution to alleviate the data movement bottleneck by placing the computation mechanisms inside or near memory. Our goal is to understand the capabilities of popular distributed SGD algorithms on real-world PIM systems to accelerate data-intensive ML training workloads. To this end, we 1) implement several representative centralized parallel SGD algorithms on the real-world UPMEM PIM system, 2) rigorously evaluate these algorithms for ML training on large-scale datasets in terms of performance, accuracy, and scalability, 3) compare to conventional CPU and GPU baselines, and 4) discuss implications for future PIM hardware and highlight the need for a shift to an algorithm-hardware codesign. Our results demonstrate three major findings: 1) The UPMEM PIM system can be a viable alternative to state-of-the-art CPUs and GPUs for many memory-bound ML training workloads, especially when operations and datatypes are natively supported by PIM hardware, 2) it is important to carefully choose the optimization algorithms that best fit PIM, and 3) the UPMEM PIM system does not scale approximately linearly with the number of nodes for many data-intensive ML training workloads. We open source all our code to facilitate future research.

arXiv:2404.06003공식 코드 · Papers with Code

FreeEval: A Modular Framework for Trustworthy and Efficient Evaluation of Large Language Models

The rapid development of large language model (LLM) evaluation methodologies and datasets has led to a profound challenge: integrating state-of-the-art evaluation techniques cost-effectively while ensuring reliability, reproducibility, and efficiency. Currently, there is a notable absence of a unified and adaptable framework that seamlessly integrates various evaluation approaches. Moreover, the reliability of evaluation findings is often questionable due to potential data contamination, with the evaluation efficiency commonly overlooked when facing the substantial costs associated with LLM inference. In response to these challenges, we introduce FreeEval, a modular and scalable framework crafted to enable trustworthy and efficient automatic evaluations of LLMs. Firstly, FreeEval's unified abstractions simplify the integration and improve the transparency of diverse evaluation methodologies, encompassing dynamic evaluation that demand sophisticated LLM interactions. Secondly, the framework integrates meta-evaluation techniques like human evaluation and data contamination detection, which, along with dynamic evaluation modules in the platform, enhance the fairness of the evaluation outcomes. Lastly, FreeEval is designed with a high-performance infrastructure, including distributed computation and caching strategies, enabling extensive evaluations across multi-node, multi-GPU clusters for open-source and proprietary LLMs.

arXiv:2404.06075

LIPT: Latency-aware Image Processing Transformer

Transformer is leading a trend in the field of image processing. Despite the great success that existing lightweight image processing transformers have achieved, they are tailored to FLOPs or parameters reduction, rather than practical inference acceleration. In this paper, we present a latency-aware image processing transformer, termed LIPT. We devise the low-latency proportion LIPT block that substitutes memory-intensive operators with the combination of self-attention and convolutions to achieve practical speedup. Specifically, we propose a novel non-volatile sparse masking self-attention (NVSM-SA) that utilizes a pre-computing sparse mask to capture contextual information from a larger window with no extra computation overload. Besides, a high-frequency reparameterization module (HRM) is proposed to make LIPT block reparameterization friendly, which improves the model's detail reconstruction capability. Extensive experiments on multiple image processing tasks (e.g., image super-resolution (SR), JPEG artifact reduction, and image denoising) demonstrate the superiority of LIPT on both latency and PSNR. LIPT achieves real-time GPU inference with state-of-the-art performance on multiple image SR benchmarks.

arXiv:2404.05911

LATUP-Net: A Lightweight 3D Attention U-Net with Parallel Convolutions for Brain Tumor Segmentation

Early-stage 3D brain tumor segmentation from magnetic resonance imaging (MRI) scans is crucial for prompt and effective treatment. However, this process faces the challenge of precise delineation due to the tumors' complex heterogeneity. Moreover, energy sustainability targets and resource limitations, especially in developing countries, require efficient and accessible medical imaging solutions. The proposed architecture, a Lightweight 3D ATtention U-Net with Parallel convolutions, LATUP-Net, addresses these issues. It is specifically designed to reduce computational requirements significantly while maintaining high segmentation performance. By incorporating parallel convolutions, it enhances feature representation by capturing multi-scale information. It further integrates an attention mechanism to refine segmentation through selective feature recalibration. LATUP-Net achieves promising segmentation performance: the average Dice scores for the whole tumor, tumor core, and enhancing tumor on the BraTS 2020 dataset are 88.41%, 83.82%, and 73.67%, and on the BraTS 2021 dataset, they are 90.29%, 89.54%, and 83.92%, respectively. Hausdorff distance metrics further indicate its improved ability to delineate tumor boundaries. With its significantly reduced computational demand using only 3.07M parameters, about 59 times fewer than other state-of-the-art models, and running on a single NVIDIA GeForce RTX3060 12GB GPU, LATUP-Net requires just 15.79 GFLOPs. This makes it a promising solution for real-world clinical applications, particularly in settings with limited resources. Investigations into the model's interpretability, utilizing gradient-weighted class activation mapping and confusion matrices, reveal that while attention mechanisms enhance the segmentation of small regions, their impact is nuanced. Achieving the most [...]. The code is available at https://qyber.black/ca/code-bca.

arXiv:2404.04895공식 코드 · Papers with Code

Tensorized Ant Colony Optimization for GPU Acceleration

Ant Colony Optimization (ACO) is renowned for its effectiveness in solving Traveling Salesman Problems, yet it faces computational challenges in CPU-based environments, particularly with large-scale instances. In response, we introduce a Tensorized Ant Colony Optimization (TensorACO) to utilize the advancements of GPU acceleration. As the core, TensorACO fully transforms ant system and ant path into tensor forms, a process we refer to as tensorization. For the tensorization of ant system, we propose a preprocessing method to reduce the computational overhead by calculating the probability transition matrix. In the tensorization of ant path, we propose an index mapping method to accelerate the update of pheromone matrix by replacing the mechanism of sequential path update with parallel matrix operations. Additionally, we introduce an Adaptive Independent Roulette (AdaIR) method to overcome the challenges of parallelizing ACO's selection mechanism on GPUs. Comprehensive experiments demonstrate the superior performance of TensorACO achieving up to 1921$\times$ speedup over standard ACO. Moreover, the AdaIR method further improves TensorACO's convergence speed by 80% and solution quality by 2%. Source codes are available at https://github.com/EMI-Group/tensoraco.

arXiv:2404.04815공식 코드 · Papers with Code

Allo: A Programming Model for Composable Accelerator Design

Special-purpose hardware accelerators are increasingly pivotal for sustaining performance improvements in emerging applications, especially as the benefits of technology scaling continue to diminish. However, designers currently lack effective tools and methodologies to construct complex, high-performance accelerator architectures in a productive manner. Existing high-level synthesis (HLS) tools often require intrusive source-level changes to attain satisfactory quality of results. Despite the introduction of several new accelerator design languages (ADLs) aiming to enhance or replace HLS, their advantages are more evident in relatively simple applications with a single kernel. Existing ADLs prove less effective for realistic hierarchical designs with multiple kernels, even if the design hierarchy is flattened. In this paper, we introduce Allo, a composable programming model for efficient spatial accelerator design. Allo decouples hardware customizations, including compute, memory, communication, and data type from algorithm specification, and encapsulates them as a set of customization primitives. Allo preserves the hierarchical structure of an input program by combining customizations from different functions in a bottom-up, type-safe manner. This approach facilitates holistic optimizations that span across function boundaries. We conduct comprehensive experiments on commonly-used HLS benchmarks and several realistic deep learning models. Our evaluation shows that Allo can outperform state-of-the-art HLS tools and ADLs on all test cases in the PolyBench. For the GPT2 model, the inference latency of the Allo generated accelerator is 1.7x faster than the NVIDIA A100 GPU with 5.4x higher energy efficiency, demonstrating the capability of Allo to handle large-scale designs.

arXiv:2404.05726공식 코드 · Papers with Code

MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding

With the success of large language models (LLMs), integrating the vision model into LLMs to build vision-language foundation models has gained much more interest recently. However, existing LLM-based large multimodal models (e.g., Video-LLaMA, VideoChat) can only take in a limited number of frames for short video understanding. In this study, we mainly focus on designing an efficient and effective model for long-term video understanding. Instead of trying to process more frames simultaneously like most existing work, we propose to process videos in an online manner and store past video information in a memory bank. This allows our model to reference historical video content for long-term analysis without exceeding LLMs' context length constraints or GPU memory limits. Our memory bank can be seamlessly integrated into current multimodal LLMs in an off-the-shelf manner. We conduct extensive experiments on various video understanding tasks, such as long-video understanding, video question answering, and video captioning, and our model can achieve state-of-the-art performances across multiple datasets. Code available at https://boheumd.github.io/MA-LMM/.

arXiv:2404.04871공식 코드 · Papers with Code

Data Stream Sampling with Fuzzy Task Boundaries and Noisy Labels

In the realm of continual learning, the presence of noisy labels within data streams represents a notable obstacle to model reliability and fairness. We focus on the data stream scenario outlined in pertinent literature, characterized by fuzzy task boundaries and noisy labels. To address this challenge, we introduce a novel and intuitive sampling method called Noisy Test Debiasing (NTD) to mitigate noisy labels in evolving data streams and establish a fair and robust continual learning algorithm. NTD is straightforward to implement, making it feasible across various scenarios. Our experiments benchmark four datasets, including two synthetic noise datasets (CIFAR10 and CIFAR100) and real-world noise datasets (mini-WebVision and Food-101N). The results validate the efficacy of NTD for online continual learning in scenarios with noisy labels in data streams. Compared to the previous leading approach, NTD achieves a training speedup enhancement over two times while maintaining or surpassing accuracy levels. Moreover, NTD utilizes less than one-fifth of the GPU memory resources compared to previous leading methods.

arXiv:2404.05567

Dense Training, Sparse Inference: Rethinking Training of Mixture-of-Experts Language Models

Mixture-of-Experts (MoE) language models can reduce computational costs by 2-4$\times$ compared to dense models without sacrificing performance, making them more efficient in computation-bounded scenarios. However, MoE models generally require 2-4$\times$ times more parameters to achieve comparable performance to a dense model, which incurs larger GPU memory requirements and makes MoE models less efficient in I/O-bounded scenarios like autoregressive generation. In this work, we propose a hybrid dense training and sparse inference framework for MoE models (DS-MoE) which achieves strong computation and parameter efficiency by employing dense computation across all experts during training and sparse computation during inference. Our experiments on training LLMs demonstrate that our DS-MoE models are more parameter-efficient than standard sparse MoEs and are on par with dense models in terms of total parameter size and performance while being computationally cheaper (activating 30-40% of the model's parameters). Performance tests using vLLM show that our DS-MoE-6B model runs up to $1.86\times$ faster than similar dense models like Mistral-7B, and between $1.50\times$ and $1.71\times$ faster than comparable MoEs, such as DeepSeekMoE-16B and Qwen1.5-MoE-A2.7B.

arXiv:2404.04118

GNNBENCH: Fair and Productive Benchmarking for Single-GPU GNN System

We hypothesize that the absence of a standardized benchmark has allowed several fundamental pitfalls in GNN System design and evaluation that the community has overlooked. In this work, we propose GNNBench, a plug-and-play benchmarking platform focused on system innovation. GNNBench presents a new protocol to exchange their captive tensor data, supports custom classes in System APIs, and allows automatic integration of the same system module to many deep learning frameworks, such as PyTorch and TensorFlow. To demonstrate the importance of such a benchmark framework, we integrated several GNN systems. Our results show that integration with GNNBench helped us identify several measurement issues that deserve attention from the community.

arXiv:2404.03019공식 코드 · Papers with Code

GeoT: Tensor Centric Library for Graph Neural Network via Efficient Segment Reduction on GPU

In recent years, Graph Neural Networks (GNNs) have ignited a surge of innovation, significantly enhancing the processing of geometric data structures such as graphs, point clouds, and meshes. As the domain continues to evolve, a series of frameworks and libraries are being developed to push GNN efficiency to new heights. While graph-centric libraries have achieved success in the past, the advent of efficient tensor compilers has highlighted the urgent need for tensor-centric libraries. Yet, efficient tensor-centric frameworks for GNNs remain scarce due to unique challenges and limitations encountered when implementing segment reduction in GNN contexts. We introduce GeoT, a cutting-edge tensor-centric library designed specifically for GNNs via efficient segment reduction. GeoT debuts innovative parallel algorithms that not only introduce new design principles but also expand the available design space. Importantly, GeoT is engineered for straightforward fusion within a computation graph, ensuring compatibility with contemporary tensor-centric machine learning frameworks and compilers. Setting a new performance benchmark, GeoT marks a considerable advancement by showcasing an average operator speedup of 1.80x and an end-to-end speedup of 1.68x.

arXiv:2404.03605공식 코드 · Papers with Code

Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization

We consider the problem of accurate quantization for language models, where both the weights and activations are uniformly quantized to 4 bits per parameter, the lowest bitwidth format natively supported by GPU hardware. In this context, the key challenge is activation quantization: it is known that language models contain outlier channels whose values on average are orders of magnitude higher than than other channels, which prevents accurate low-bitwidth quantization with known techniques. We systematically study this phenomena and find that these outlier channels emerge early in training, and that they occur more frequently in layers with residual streams. We then propose a simple strategy which regularizes a layer's inputs via quantization-aware training (QAT) and its outputs via activation kurtosis regularization. We show that regularizing both the inputs and outputs is crucial for preventing a model's "migrating" the difficulty in input quantization to the weights, which makes post-training quantization (PTQ) of weights more difficult. When combined with weight PTQ, we show that our approach can obtain a W4A4 model that performs competitively to the standard-precision W16A16 baseline.