PyTorchKR Papers

최신 논문

총 5,357편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2409.14023

FAMOUS: Flexible Accelerator for the Attention Mechanism of Transformer on UltraScale+ FPGAs

Transformer neural networks (TNNs) are being applied across a widening range of application domains, including natural language processing (NLP), machine translation, and computer vision (CV). Their popularity is largely attributed to the exceptional performance of their multi-head self-attention blocks when analyzing sequential data and extracting features. To date, there are limited hardware accelerators tailored for this mechanism, which is the first step before designing an accelerator for a complete model. This paper proposes \textit{FAMOUS}, a flexible hardware accelerator for dense multi-head attention (MHA) computation of TNNs on field-programmable gate arrays (FPGAs). It is optimized for high utilization of processing elements and on-chip memories to improve parallelism and reduce latency. An efficient tiling of large matrices has been employed to distribute memory and computing resources across different modules on various FPGA platforms. The design is evaluated on Xilinx Alveo U55C and U200 data center cards containing Ultrascale+ FPGAs. Experimental results are presented that show that it can attain a maximum throughput, number of parallel attention heads, embedding dimension and tile size of 328 (giga operations/second (GOPS)), 8, 768 and 64 respectively on the U55C. Furthermore, it is 3.28$\times$ and 2.6$\times$ faster than the Intel Xeon Gold 5220R CPU and NVIDIA V100 GPU respectively. It is also 1.3$\times$ faster than the fastest state-of-the-art FPGA-based accelerator.

arXiv:2409.13975

ProTEA: Programmable Transformer Encoder Acceleration on FPGA

Transformer neural networks (TNN) have been widely utilized on a diverse range of applications, including natural language processing (NLP), machine translation, and computer vision (CV). Their widespread adoption has been primarily driven by the exceptional performance of their multi-head self-attention block used to extract key features from sequential data. The multi-head self-attention block is followed by feedforward neural networks, which play a crucial role in introducing non-linearity to assist the model in learning complex patterns. Despite the popularity of TNNs, there has been limited numbers of hardware accelerators targeting these two critical blocks. Most prior works have concentrated on sparse architectures that are not flexible for popular TNN variants. This paper introduces \textit{ProTEA}, a runtime programmable accelerator tailored for the dense computations of most of state-of-the-art transformer encoders. \textit{ProTEA} is designed to reduce latency by maximizing parallelism. We introduce an efficient tiling of large matrices that can distribute memory and computing resources across different hardware components within the FPGA. We provide run time evaluations of \textit{ProTEA} on a Xilinx Alveo U55C high-performance data center accelerator card. Experimental results demonstrate that \textit{ProTEA} can host a wide range of popular transformer networks and achieve near optimal performance with a tile size of 64 in the multi-head self-attention block and 6 in the feedforward networks block when configured with 8 parallel attention heads, 12 layers, and an embedding dimension of 768 on the U55C. Comparative results are provided showing \textit{ProTEA} is 2.5$\times$ faster than an NVIDIA Titan XP GPU. Results also show that it achieves 1.3 -- 2.8$\times$ speed up compared with current state-of-the-art custom designed FPGA accelerators.

arXiv:2409.14918

A Realistic Simulation Framework for Analog/Digital Neuromorphic Architectures

Developing dedicated mixed-signal neuromorphic computing systems optimized for real-time sensory-processing in extreme edge-computing applications requires time-consuming design, fabrication, and deployment of full-custom neuromorphic processors. To ensure that initial prototyping efforts, exploring the properties of different network architectures and parameter settings, lead to realistic results, it is important to use simulation frameworks that match as best as possible the properties of the final hardware. This is particularly challenging for neuromorphic hardware platforms made using mixed-signal analog/digital circuits, due to the variability and noise sensitivity of their components. In this paper, we address this challenge by developing a software spiking neural network simulator explicitly designed to account for the properties of mixed-signal neuromorphic circuits, including device mismatch variability. The simulator, called ARCANA (A Realistic Simulation Framework for Analog/Digital Neuromorphic Architectures), is designed to reproduce the dynamics of mixed-signal synapse and neuron electronic circuits with autogradient differentiation for parameter optimization and GPU acceleration. We demonstrate the effectiveness of this approach by matching software simulation results with measurements made from an existing neuromorphic processor. We show how the results obtained provide a reliable estimate of the behavior of the spiking neural network trained in software, once deployed in hardware. This framework enables the development and innovation of new learning rules and processing architectures in neuromorphic embedded systems.

arXiv:2409.13997

Drift to Remember

Lifelong learning in artificial intelligence (AI) aims to mimic the biological brain's ability to continuously learn and retain knowledge, yet it faces challenges such as catastrophic forgetting. Recent neuroscience research suggests that neural activity in biological systems undergoes representational drift, where neural responses evolve over time, even with consistent inputs and tasks. We hypothesize that representational drift can alleviate catastrophic forgetting in AI during new task acquisition. To test this, we introduce DriftNet, a network designed to constantly explore various local minima in the loss landscape while dynamically retrieving relevant tasks. This approach ensures efficient integration of new information and preserves existing knowledge. Experimental studies in image classification and natural language processing demonstrate that DriftNet outperforms existing models in lifelong learning. Importantly, DriftNet is scalable in handling a sequence of tasks such as sentiment analysis and question answering using large language models (LLMs) with billions of parameters on a single Nvidia A100 GPU. DriftNet efficiently updates LLMs using only new data, avoiding the need for full dataset retraining. Tested on GPT-2 and RoBERTa, DriftNet is a robust, cost-effective solution for lifelong learning in LLMs. This study not only advances AI systems to emulate biological learning, but also provides insights into the adaptive mechanisms of biological neural systems, deepening our understanding of lifelong learning in nature.

arXiv:2409.14162

On Importance of Pruning and Distillation for Efficient Low Resource NLP

The rise of large transformer models has revolutionized Natural Language Processing, leading to significant advances in tasks like text classification. However, this progress demands substantial computational resources, escalating training duration, and expenses with larger model sizes. Efforts have been made to downsize and accelerate English models (e.g., Distilbert, MobileBert). Yet, research in this area is scarce for low-resource languages. In this study, we explore the case of the low-resource Indic language Marathi. Leveraging the marathi-topic-all-doc-v2 model as our baseline, we implement optimization techniques to reduce computation time and memory usage. Our focus is on enhancing the efficiency of Marathi transformer models while maintaining top-tier accuracy and reducing computational demands. Using the MahaNews document classification dataset and the marathi-topic-all-doc-v2 model from L3Cube, we apply Block Movement Pruning, Knowledge Distillation, and Mixed Precision methods individually and in combination to boost efficiency. We demonstrate the importance of strategic pruning levels in achieving desired efficiency gains. Furthermore, we analyze the balance between efficiency improvements and environmental impact, highlighting how optimized model architectures can contribute to a more sustainable computational ecosystem. Implementing these techniques on a single GPU system, we determine that the optimal configuration is 25\% pruning + knowledge distillation. This approach yielded a 2.56x speedup in computation time while maintaining baseline accuracy levels.

arXiv:2409.14939공식 코드 · Papers with Code

FastGL: A GPU-Efficient Framework for Accelerating Sampling-Based GNN Training at Large Scale

Graph Neural Networks (GNNs) have shown great superiority on non-Euclidean graph data, achieving ground-breaking performance on various graph-related tasks. As a practical solution to train GNN on large graphs with billions of nodes and edges, the sampling-based training is widely adopted by existing training frameworks. However, through an in-depth analysis, we observe that the efficiency of existing sampling-based training frameworks is still limited due to the key bottlenecks lying in all three phases of sampling-based training, i.e., subgraph sample, memory IO, and computation. To this end, we propose FastGL, a GPU-efficient Framework for accelerating sampling-based training of GNN at Large scale by simultaneously optimizing all above three phases, taking into account both GPU characteristics and graph structure. Specifically, by exploiting the inherent overlap within graph structures, FastGL develops the Match-Reorder strategy to reduce the data traffic, which accelerates the memory IO without incurring any GPU memory overhead. Additionally, FastGL leverages a Memory-Aware computation method, harnessing the GPU memory's hierarchical nature to mitigate irregular data access during computation. FastGL further incorporates the Fused-Map approach aimed at diminishing the synchronization overhead during sampling. Extensive experiments demonstrate that FastGL can achieve an average speedup of 11.8x, 2.2x and 1.5x over the state-of-the-art frameworks PyG, DGL, and GNNLab, respectively.Our code is available at https://github.com/a1bc2def6g/fastgl-ae.

arXiv:2409.12892공식 코드 · Papers with Code

3DGS-LM: Faster Gaussian-Splatting Optimization with Levenberg-Marquardt

We present 3DGS-LM, a new method that accelerates the reconstruction of 3D Gaussian Splatting (3DGS) by replacing its ADAM optimizer with a tailored Levenberg-Marquardt (LM). Existing methods reduce the optimization time by decreasing the number of Gaussians or by improving the implementation of the differentiable rasterizer. However, they still rely on the ADAM optimizer to fit Gaussian parameters of a scene in thousands of iterations, which can take up to an hour. To this end, we change the optimizer to LM that runs in conjunction with the 3DGS differentiable rasterizer. For efficient GPU parallization, we propose a caching data structure for intermediate gradients that allows us to efficiently calculate Jacobian-vector products in custom CUDA kernels. In every LM iteration, we calculate update directions from multiple image subsets using these kernels and combine them in a weighted mean. Overall, our method is 30% faster than the original 3DGS while obtaining the same reconstruction quality. Our optimization is also agnostic to other methods that acclerate 3DGS, thus enabling even faster speedups compared to vanilla 3DGS.

arXiv:2409.12878

Impact of ML Optimization Tactics on Greener Pre-Trained ML Models

Background: Given the fast-paced nature of today's technology, which has surpassed human performance in tasks like image classification, visual reasoning, and English understanding, assessing the impact of Machine Learning (ML) on energy consumption is crucial. Traditionally, ML projects have prioritized accuracy over energy, creating a gap in energy consumption during model inference. Aims: This study aims to (i) analyze image classification datasets and pre-trained models, (ii) improve inference efficiency by comparing optimized and non-optimized models, and (iii) assess the economic impact of the optimizations. Method: We conduct a controlled experiment to evaluate the impact of various PyTorch optimization techniques (dynamic quantization, torch.compile, local pruning, and global pruning) to 42 Hugging Face models for image classification. The metrics examined include GPU utilization, power and energy consumption, accuracy, time, computational complexity, and economic costs. The models are repeatedly evaluated to quantify the effects of these software engineering tactics. Results: Dynamic quantization demonstrates significant reductions in inference time and energy consumption, making it highly suitable for large-scale systems. Additionally, torch.compile balances accuracy and energy. In contrast, local pruning shows no positive impact on performance, and global pruning's longer optimization times significantly impact costs. Conclusions: This study highlights the role of software engineering tactics in achieving greener ML models, offering guidelines for practitioners to make informed decisions on optimization methods that align with sustainability goals.

arXiv:2409.12903공식 코드 · Papers with Code

Scaling Smart: Accelerating Large Language Model Pre-training with Small Model Initialization

The pre-training phase of language models often begins with randomly initialized parameters. With the current trends in scaling models, training their large number of parameters can be extremely slow and costly. In contrast, small language models are less expensive to train, but they often cannot achieve the accuracy of large models. In this paper, we explore an intriguing idea to connect these two different regimes: Can we develop a method to initialize large language models using smaller pre-trained models? Will such initialization bring any benefits in terms of training time and final accuracy? In this paper, we introduce HyperCloning, a method that can expand the parameters of a pre-trained language model to those of a larger model with increased hidden dimensions. Our method ensures that the larger model retains the functionality of the smaller model. As a result, the larger model already inherits the predictive power and accuracy of the smaller model before the training starts. We demonstrate that training such an initialized model results in significant savings in terms of GPU hours required for pre-training large language models.

arXiv:2409.12815

Graph Convolutional Neural Networks as Surrogate Models for Climate Simulation

Many climate processes are characterized using large systems of nonlinear differential equations; this, along with the immense amount of data required to parameterize complex interactions, means that Earth-System Model (ESM) simulations may take weeks to run on large clusters. Uncertainty quantification may require thousands of runs, making ESM simulations impractical for preliminary assessment. Alternatives may include simplifying the processes in the model, but recent efforts have focused on using machine learning to complement these models or even act as full surrogates. \textit{We leverage machine learning, specifically fully-connected neural networks (FCNNs) and graph convolutional neural networks (GCNNs), to enable rapid simulation and uncertainty quantification in order to inform more extensive ESM simulations.} Our surrogate simulated 80 years in approximately 310 seconds on a single A100 GPU, compared to weeks for the ESM model while having mean temperature errors below $0.1^{\circ}C$ and maximum errors below $2^{\circ}C$.

arXiv:2409.12490공식 코드 · Papers with Code

CritiPrefill: A Segment-wise Criticality-based Approach for Prefilling Acceleration in LLMs

Large language models have achieved notable success across various domains, yet efficient inference is still limited by the quadratic computation complexity of the attention mechanism. The inference consists of prefilling and decoding phases. Although several attempts have been made to accelerate decoding, the inefficiency of the prefilling phase, especially for long-context tasks, remains a challenge. In this paper, we observe a locality in query criticality during the prefilling phase of long-context processing: adjacent query tokens tend to focus on similar subsets of the past Key-Value (KV) cache. Based on this observation, we propose CritiPrefill, a criticality-based segment-wise prefilling method. This method partitions the input sequence's queries and KV cache into segments and blocks, utilizing a segment-wise algorithm to estimate the query criticality. By pruning non-critical computations between query segments and cache blocks in the self-attention mechanism, the prefilling process can be significantly accelerated. Extensive evaluations on multiple long-context datasets show up to 2.7x speedup on Llama3-8B and 3.0x speedup on Yi-9B for 128K context length on a single A100 GPU, with minimal quality degradation.

arXiv:2409.12264

User-friendly Foundation Model Adapters for Multivariate Time Series Classification

Foundation models, while highly effective, are often resource-intensive, requiring substantial inference time and memory. This paper addresses the challenge of making these models more accessible with limited computational resources by exploring dimensionality reduction techniques. Our goal is to enable users to run large pre-trained foundation models on standard GPUs without sacrificing performance. We investigate classical methods such as Principal Component Analysis alongside neural network-based adapters, aiming to reduce the dimensionality of multivariate time series data while preserving key features. Our experiments show up to a 10x speedup compared to the baseline model, without performance degradation, and enable up to 4.5x more datasets to fit on a single GPU, paving the way for more user-friendly and scalable foundation models.

arXiv:2409.11155

ISO: Overlap of Computation and Communication within Seqenence For LLM Inference

In the realm of Large Language Model (LLM) inference, the inherent structure of transformer models coupled with the multi-GPU tensor parallelism strategy leads to a sequential execution of computation and communication. This results in substantial underutilization of computing resources during the communication phase. To mitigate this inefficiency, various techniques have been developed to optimize the use of computational power throughout the communication process. These strategies primarily involve overlapping matrix computations and communications, as well as interleaving micro-batches across different requests. Nonetheless, these approaches either fall short of achieving ideal overlap or impose certain limitations on their application. To overcome these challenges, this paper introduces a novel strategy for computation-communication overlap that operates at the sequence level. This method not only enhances the degree of overlap but also minimizes the constraints on its applicability. Experimental evaluations conducted using 30b/70b models have demonstrated significant improvements in efficiency. Specifically, the proposed technique has been shown to reduce time consumption by approximately 35% on 4090 GPU and by roughly 15% on A800 GPU during the prefill stage of LLM inference.

arXiv:2409.11902

Less Memory Means smaller GPUs: Backpropagation with Compressed Activations

The ever-growing scale of deep neural networks (DNNs) has lead to an equally rapid growth in computational resource requirements. Many recent architectures, most prominently Large Language Models, have to be trained using supercomputers with thousands of accelerators, such as GPUs or TPUs. Next to the vast number of floating point operations the memory footprint of DNNs is also exploding. In contrast, GPU architectures are notoriously short on memory. Even comparatively small architectures like some EfficientNet variants cannot be trained on a single consumer-grade GPU at reasonable mini-batch sizes. During training, intermediate input activations have to be stored until backpropagation for gradient calculation. These make up the vast majority of the memory footprint. In this work we therefore consider compressing activation maps for the backward pass using pooling, which can reduce both the memory footprint and amount of data movement. The forward computation remains uncompressed. We empirically show convergence and study effects on feature detection at the example of the common vision architecture ResNet. With this approach we are able to reduce the peak memory consumption by 29% at the cost of a longer training schedule, while maintaining prediction accuracy compared to an uncompressed baseline.

arXiv:2409.11783공식 코드 · Papers with Code

Development and bilingual evaluation of Japanese medical large language model within reasonably low computational resources

The recent success of large language models (LLMs) and the scaling law has led to a widespread adoption of larger models. Particularly in the healthcare industry, there is an increasing demand for locally operated LLMs due to security concerns. However, the majority of high quality open-source LLMs have a size of 70B parameters, imposing significant financial burdens on users for GPU preparation and operation. To overcome these issues, we present a medical adaptation based on the recent 7B models, which enables the operation in low computational resources. We compare the performance on medical question-answering benchmarks in two languages (Japanese and English), demonstrating that its scores reach parity with or surpass those of currently existing medical LLMs that are ten times larger. We find that fine-tuning an English-centric base model on Japanese medical dataset improves the score in both language, supporting the effect of cross-lingual knowledge transfer. We hope that this study will alleviate financial challenges, serving as a stepping stone for clinical institutions to practically utilize LLMs locally. Our evaluation code is available at https://github.com/stardust-coder/japanese-lm-med-harness.

arXiv:2409.12190

Bundle Adjustment in the Eager Mode

Bundle adjustment (BA) is a critical technique in various robotic applications, such as simultaneous localization and mapping (SLAM), augmented reality (AR), and photogrammetry. BA optimizes parameters such as camera poses and 3D landmarks to align them with observations. With the growing importance of deep learning in perception systems, there is an increasing need to integrate BA with deep learning frameworks for enhanced reliability and performance. However, widely-used C++-based BA frameworks, such as GTSAM, g$^2$o, and Ceres, lack native integration with modern deep learning libraries like PyTorch. This limitation affects their flexibility, adaptability, ease of debugging, and overall implementation efficiency. To address this gap, we introduce an eager-mode BA framework seamlessly integrated with PyPose, providing PyTorch-compatible interfaces with high efficiency. Our approach includes GPU-accelerated, differentiable, and sparse operations designed for 2nd-order optimization, Lie group and Lie algebra operations, and linear solvers. Our eager-mode BA on GPU demonstrates substantial runtime efficiency, achieving an average speedup of 18.5$\times$, 22$\times$, and 23$\times$ compared to GTSAM, g$^2$o, and Ceres, respectively.

arXiv:2409.11513공식 코드 · Papers with Code

Mamba Fusion: Learning Actions Through Questioning

Video Language Models (VLMs) are crucial for generalizing across diverse tasks and using language cues to enhance learning. While transformer-based architectures have been the de facto in vision-language training, they face challenges like quadratic computational complexity, high GPU memory usage, and difficulty with long-term dependencies. To address these limitations, we introduce MambaVL, a novel model that leverages recent advancements in selective state space modality fusion to efficiently capture long-range dependencies and learn joint representations for vision and language data. MambaVL utilizes a shared state transition matrix across both modalities, allowing the model to capture information about actions from multiple perspectives within the scene. Furthermore, we propose a question-answering task that helps guide the model toward relevant cues. These questions provide critical information about actions, objects, and environmental context, leading to enhanced performance. As a result, MambaVL achieves state-of-the-art performance in action recognition on the Epic-Kitchens-100 dataset and outperforms baseline methods in action anticipation.

arXiv:2409.11440

MARCA: Mamba Accelerator with ReConfigurable Architecture

We propose a Mamba accelerator with reconfigurable architecture, MARCA.We propose three novel approaches in this paper. (1) Reduction alternative PE array architecture for both linear and element-wise operations. For linear operations, the reduction tree connected to PE arrays is enabled and executes the reduction operation. For element-wise operations, the reduction tree is disabled and the output bypasses. (2) Reusable nonlinear function unit based on the reconfigurable PE. We decompose the exponential function into element-wise operations and a shift operation by a fast biased exponential algorithm, and the activation function (SiLU) into a range detection and element-wise operations by a piecewise approximation algorithm. Thus, the reconfigurable PEs are reused to execute nonlinear functions with negligible accuracy loss.(3) Intra-operation and inter-operation buffer management strategy. We propose intra-operation buffer management strategy to maximize input data sharing for linear operations within operations, and inter-operation strategy for element-wise operations between operations. We conduct extensive experiments on Mamba model families with different sizes.MARCA achieves up to 463.22$\times$/11.66$\times$ speedup and up to 9761.42$\times$/242.52$\times$ energy efficiency compared to Intel Xeon 8358P CPU and NVIDIA Tesla A100 GPU implementations, respectively.

arXiv:2409.11786

Efficient Low-Resolution Face Recognition via Bridge Distillation

Face recognition in the wild is now advancing towards light-weight models, fast inference speed and resolution-adapted capability. In this paper, we propose a bridge distillation approach to turn a complex face model pretrained on private high-resolution faces into a light-weight one for low-resolution face recognition. In our approach, such a cross-dataset resolution-adapted knowledge transfer problem is solved via two-step distillation. In the first step, we conduct cross-dataset distillation to transfer the prior knowledge from private high-resolution faces to public high-resolution faces and generate compact and discriminative features. In the second step, the resolution-adapted distillation is conducted to further transfer the prior knowledge to synthetic low-resolution faces via multi-task learning. By learning low-resolution face representations and mimicking the adapted high-resolution knowledge, a light-weight student model can be constructed with high efficiency and promising accuracy in recognizing low-resolution faces. Experimental results show that the student model performs impressively in recognizing low-resolution faces with only 0.21M parameters and 0.057MB memory. Meanwhile, its speed reaches up to 14,705, ~934 and 763 faces per second on GPU, CPU and mobile phone, respectively.

arXiv:2409.11129공식 코드 · Papers with Code

Can Graph Reordering Speed Up Graph Neural Network Training? An Experimental Study

Graph neural networks (GNNs) are a type of neural network capable of learning on graph-structured data. However, training GNNs on large-scale graphs is challenging due to iterative aggregations of high-dimensional features from neighboring vertices within sparse graph structures combined with neural network operations. The sparsity of graphs frequently results in suboptimal memory access patterns and longer training time. Graph reordering is an optimization strategy aiming to improve the graph data layout. It has shown to be effective to speed up graph analytics workloads, but its effect on the performance of GNN training has not been investigated yet. The generalization of reordering to GNN performance is nontrivial, as multiple aspects must be considered: GNN hyper-parameters such as the number of layers, the number of hidden dimensions, and the feature size used in the GNN model, neural network operations, large intermediate vertex states, and GPU acceleration. In our work, we close this gap by performing an empirical evaluation of 12 reordering strategies in two state-of-the-art GNN systems, PyTorch Geometric and Deep Graph Library. Our results show that graph reordering is effective in reducing training time for CPU- and GPU-based training, respectively. Further, we find that GNN hyper-parameters influence the effectiveness of reordering, that reordering metrics play an important role in selecting a reordering strategy, that lightweight reordering performs better for GPU-based than for CPU-based training, and that invested reordering time can in many cases be amortized.