PyTorchKR Papers

최신 논문

총 5,357편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2404.11848공식 코드 · Papers with Code

Partial Large Kernel CNNs for Efficient Super-Resolution

Recently, in the super-resolution (SR) domain, transformers have outperformed CNNs with fewer FLOPs and fewer parameters since they can deal with long-range dependency and adaptively adjust weights based on instance. In this paper, we demonstrate that CNNs, although less focused on in the current SR domain, surpass Transformers in direct efficiency measures. By incorporating the advantages of Transformers into CNNs, we aim to achieve both computational efficiency and enhanced performance. However, using a large kernel in the SR domain, which mainly processes large images, incurs a large computational overhead. To overcome this, we propose novel approaches to employing the large kernel, which can reduce latency by 86\% compared to the naive large kernel, and leverage an Element-wise Attention module to imitate instance-dependent weights. As a result, we introduce Partial Large Kernel CNNs for Efficient Super-Resolution (PLKSR), which achieves state-of-the-art performance on four datasets at a scale of $\times$4, with reductions of 68.1\% in latency and 80.2\% in maximum GPU memory occupancy compared to SRFormer-light.

arXiv:2404.10779

Fine Tuning LLM for Enterprise: Practical Guidelines and Recommendations

There is a compelling necessity from enterprises for fine tuning LLMs (Large Language Models) o get them trained on proprietary domain knowledge. The challenge is to imbibe the LLMs with domain specific knowledge using the most optimial resource and cost and in the best possible time. Many enterprises rely on RAG (Retrieval Augmented Generation) which does not need LLMs to be ine-tuned but they are limited by the quality of vector databases and their retrieval capabilities rather than the intrinsic capabilities of the LLMs themselves. In our current work we focus on fine tuning LLaMA, an open source LLM using proprietary documents and code from an enterprise repository and use the fine tuned models to evaluate the quality of responses. As part of this work, we aim to guide beginners on how to start with fine tuning an LLM for documentation and code by making educated guesses on size of GPU required and options that are available for formatting the data. We also propose pre processing recipes for both documentation and code to prepare dataset in different formats. The proposed methods of data preparation for document datasets are forming paragraph chunks, forming question and answer pairs and forming keyword and paragraph chunk pairs. For code dataset we propose forming summary and function pairs. Further, we qualitatively evaluate the results of the models for domain specific queries. Finally, we also propose practical guidelines and recommendations for fine tuning LLMs.

arXiv:2404.10631

Parallel Implementations Assessment of a Spatial-Spectral Classifier for Hyperspectral Clinical Applications

Hyperspectral (HS) imaging presents itself as a non-contact, non-ionizing and non-invasive technique, proven to be suitable for medical diagnosis. However, the volume of information contained in these images makes difficult providing the surgeon with information about the boundaries in real-time. To that end, High-Performance-Computing (HPC) platforms become necessary. This paper presents a comparison between the performances provided by five different HPC platforms while processing a spatial-spectral approach to classify HS images, assessing their main benefits and drawbacks. To provide a complete study, two different medical applications, with two different requirements, have been analyzed. The first application consists of HS images taken from neurosurgical operations; the second one presents HS images taken from dermatological interventions. While the main constraint for neurosurgical applications is the processing time, in other environments, as the dermatological one, other requirements can be considered. In that sense, energy efficiency is becoming a major challenge, since this kind of applications are usually developed as hand-held devices, thus depending on the battery capacity. These requirements have been considered to choose the target platforms: on the one hand, three of the most powerful Graphic Processing Units (GPUs) available in the market; and, on the other hand, a low-power GPU and a manycore architecture, both specifically thought for being used in battery-dependent environments.

arXiv:2404.10934공식 코드 · Papers with Code

Shears: Unstructured Sparsity with Neural Low-rank Adapter Search

Recently, several approaches successfully demonstrated that weight-sharing Neural Architecture Search (NAS) can effectively explore a search space of elastic low-rank adapters (LoRA), allowing the parameter-efficient fine-tuning (PEFT) and compression of large language models. In this paper, we introduce a novel approach called Shears, demonstrating how the integration of cost-effective sparsity and a proposed Neural Low-rank adapter Search (NLS) algorithm can further improve the efficiency of PEFT approaches. Results demonstrate the benefits of Shears compared to other methods, reaching high sparsity levels while improving or with little drop in accuracy, utilizing a single GPU for a pair of hours.

arXiv:2404.10933공식 코드 · Papers with Code

LLMem: Estimating GPU Memory Usage for Fine-Tuning Pre-Trained LLMs

Fine-tuning pre-trained large language models (LLMs) with limited hardware presents challenges due to GPU memory constraints. Various distributed fine-tuning methods have been proposed to alleviate memory constraints on GPU. However, determining the most effective method for achieving rapid fine-tuning while preventing GPU out-of-memory issues in a given environment remains unclear. To address this challenge, we introduce LLMem, a solution that estimates the GPU memory consumption when applying distributed fine-tuning methods across multiple GPUs and identifies the optimal method. We conduct GPU memory usage estimation prior to fine-tuning, leveraging the fundamental structure of transformer-based decoder models and the memory usage distribution of each method. Experimental results show that LLMem accurately estimates peak GPU memory usage on a single GPU, with error rates of up to 1.6%. Additionally, it shows an average error rate of 3.0% when applying distributed fine-tuning methods to LLMs with more than a billion parameters on multi-GPU setups.

arXiv:2404.11118공식 코드 · Papers with Code

FastFace: Fast-converging Scheduler for Large-scale Face Recognition Training with One GPU

Computing power has evolved into a foundational and indispensable resource in the area of deep learning, particularly in tasks such as Face Recognition (FR) model training on large-scale datasets, where multiple GPUs are often a necessity. Recognizing this challenge, some FR methods have started exploring ways to compress the fully-connected layer in FR models. Unlike other approaches, our observations reveal that without prompt scheduling of the learning rate (LR) during FR model training, the loss curve tends to exhibit numerous stationary subsequences. To address this issue, we introduce a novel LR scheduler leveraging Exponential Moving Average (EMA) and Haar Convolutional Kernel (HCK) to eliminate stationary subsequences, resulting in a significant reduction in converging time. However, the proposed scheduler incurs a considerable computational overhead due to its time complexity. To overcome this limitation, we propose FastFace, a fast-converging scheduler with negligible time complexity, i.e. O(1) per iteration, during training. In practice, FastFace is able to accelerate FR model training to a quarter of its original time without sacrificing more than 1% accuracy, making large-scale FR training feasible even with just one single GPU in terms of both time and space complexity. Extensive experiments validate the efficiency and effectiveness of FastFace. The code is publicly available at: https://github.com/amoonfana/FastFace

arXiv:2404.10572

Label merge-and-split: A graph-colouring approach for memory-efficient brain parcellation

Whole brain parcellation requires inferring hundreds of segmentation labels in large image volumes and thus presents significant practical challenges for deep learning approaches. We introduce label merge-and-split, a method that first greatly reduces the effective number of labels required for learning-based whole brain parcellation and then recovers original labels. Using a greedy graph colouring algorithm, our method automatically groups and merges multiple spatially separate labels prior to model training and inference. The merged labels may be semantically unrelated. A deep learning model is trained to predict merged labels. At inference time, original labels are restored using atlas-based influence regions. In our experiments, the proposed approach reduces the number of labels by up to 68% while achieving segmentation accuracy comparable to the baseline method without label merging and splitting. Moreover, model training and inference times as well as GPU memory requirements were reduced significantly. The proposed method can be applied to all semantic segmentation tasks with a large number of spatially separate classes within an atlas-based prior.

arXiv:2404.10445

SparseDM: Toward Sparse Efficient Diffusion Models

Diffusion models represent a powerful family of generative models widely used for image and video generation. However, the time-consuming deployment, long inference time, and requirements on large memory hinder their applications on resource constrained devices. In this paper, we propose a method based on the improved Straight-Through Estimator to improve the deployment efficiency of diffusion models. Specifically, we add sparse masks to the Convolution and Linear layers in a pre-trained diffusion model, then transfer learn the sparse model during the fine-tuning stage and turn on the sparse masks during inference. Experimental results on a Transformer and UNet-based diffusion models demonstrate that our method reduces MACs by 50% while maintaining FID. Sparse models are accelerated by approximately 1.2x on the GPU. Under other MACs conditions, the FID is also lower than 1 compared to other methods.

arXiv:2404.10296공식 코드 · Papers with Code

Interpolating neural network: A novel unification of machine learning and interpolation theory

Artificial intelligence (AI) has revolutionized software development, shifting from task-specific codes (Software 1.0) to neural network-based approaches (Software 2.0). However, applying this transition in engineering software presents challenges, including low surrogate model accuracy, the curse of dimensionality in inverse design, and rising complexity in physical simulations. We introduce an interpolating neural network (INN), grounded in interpolation theory and tensor decomposition, to realize Engineering Software 2.0 by advancing data training, partial differential equation solving, and parameter calibration. INN offers orders of magnitude fewer trainable/solvable parameters for comparable model accuracy than traditional multi-layer perceptron (MLP) or physics-informed neural networks (PINN). Demonstrated in metal additive manufacturing, INN rapidly constructs an accurate surrogate model of Laser Powder Bed Fusion (L-PBF) heat transfer simulation, achieving sub-10-micrometer resolution for a 10 mm path in under 15 minutes on a single GPU. This makes a transformative step forward across all domains essential to engineering software.

arXiv:2404.10162

Optimal Kernel Tuning Parameter Prediction using Deep Sequence Models

GPU kernels have come to the forefront of comput- ing due to their utility in varied fields, from high-performance computing to machine learning. A typical GPU compute kernel is invoked millions, if not billions of times in a typical application, which makes their performance highly critical. Due to the unknown nature of the optimization surface, an exhaustive search is required to discover the global optimum, which is infeasible due to the possible exponential number of parameter combinations. In this work, we propose a methodology that uses deep sequence- to-sequence models to predict the optimal tuning parameters governing compute kernels. This work considers the prediction of kernel parameters as a sequence to the sequence translation problem, borrowing models from the Natural Language Process- ing (NLP) domain. Parameters describing the input, output and weight tensors are considered as the input language to the model that emits the corresponding kernel parameters. In essence, the model translates the problem parameter language to kernel parameter language. The core contributions of this work are: a) Proposing that a sequence to sequence model can accurately learn the performance dynamics of a GPU compute kernel b) A novel network architecture which predicts the kernel tuning parameters for GPU kernels, c) A constrained beam search which incorporates the physical limits of the GPU hardware as well as other expert knowledge reducing the search space. The proposed algorithm can achieve more than 90% accuracy on various convolutional kernels in MIOpen, the AMD machine learning primitives library. As a result, the proposed technique can reduce the development time and compute resources required to tune unseen input configurations, resulting in shorter development cycles, reduced development costs, and better user experience.

arXiv:2404.10730

Insight Gained from Migrating a Machine Learning Model to Intelligence Processing Units

The discoveries in this paper show that Intelligence Processing Units (IPUs) offer a viable accelerator alternative to GPUs for machine learning (ML) applications within the fields of materials science and battery research. We investigate the process of migrating a model from GPU to IPU and explore several optimization techniques, including pipelining and gradient accumulation, aimed at enhancing the performance of IPU-based models. Furthermore, we have effectively migrated a specialized model to the IPU platform. This model is employed for predicting effective conductivity, a parameter crucial in ion transport processes, which govern the performance of multiple charge and discharge cycles of batteries. The model utilizes a Convolutional Neural Network (CNN) architecture to perform prediction tasks for effective conductivity. The performance of this model on the IPU is found to be comparable to its execution on GPUs. We also analyze the utilization and performance of Graphcore's Bow IPU. Through benchmark tests, we observe significantly improved performance with the Bow IPU when compared to its predecessor, the Colossus IPU.

arXiv:2404.08698공식 코드 · Papers with Code

Lossless Acceleration of Large Language Model via Adaptive N-gram Parallel Decoding

While Large Language Models (LLMs) have shown remarkable abilities, they are hindered by significant resource consumption and considerable latency due to autoregressive processing. In this study, we introduce Adaptive N-gram Parallel Decoding (ANPD), an innovative and lossless approach that accelerates inference by allowing the simultaneous generation of multiple tokens. ANPD incorporates a two-stage approach: it begins with a rapid drafting phase that employs an N-gram module, which adapts based on the current interactive context, followed by a verification phase, during which the original LLM assesses and confirms the proposed tokens. Consequently, ANPD preserves the integrity of the LLM's original output while enhancing processing speed. We further leverage a multi-level architecture for the N-gram module to enhance the precision of the initial draft, consequently reducing inference latency. ANPD eliminates the need for retraining or extra GPU memory, making it an efficient and plug-and-play enhancement. In our experiments, models such as LLaMA and its fine-tuned variants have shown speed improvements up to 3.67x, validating the effectiveness of our proposed ANPD.

arXiv:2404.09967

Ctrl-Adapter: An Efficient and Versatile Framework for Adapting Diverse Controls to Any Diffusion Model

ControlNets are widely used for adding spatial control to text-to-image diffusion models with different conditions, such as depth maps, scribbles/sketches, and human poses. However, when it comes to controllable video generation, ControlNets cannot be directly integrated into new backbones due to feature space mismatches, and training ControlNets for new backbones can be a significant burden for many users. Furthermore, applying ControlNets independently to different frames cannot effectively maintain object temporal consistency. To address these challenges, we introduce Ctrl-Adapter, an efficient and versatile framework that adds diverse controls to any image/video diffusion model through the adaptation of pretrained ControlNets. Ctrl-Adapter offers strong and diverse capabilities, including image and video control, sparse-frame video control, fine-grained patch-level multi-condition control (via an MoE router), zero-shot adaptation to unseen conditions, and supports a variety of downstream tasks beyond spatial control, including video editing, video style transfer, and text-guided motion control. With six diverse U-Net/DiT-based image/video diffusion models (SDXL, PixArt-$\alpha$, I2VGen-XL, SVD, Latte, Hotshot-XL), Ctrl-Adapter matches the performance of pretrained ControlNets on COCO and achieves the state-of-the-art on DAVIS 2017 with significantly lower computation (< 10 GPU hours).

arXiv:2404.09857

Empowering Embodied Visual Tracking with Visual Foundation Models and Offline RL

Embodied visual tracking is to follow a target object in dynamic 3D environments using an agent's egocentric vision. This is a vital and challenging skill for embodied agents. However, existing methods suffer from inefficient training and poor generalization. In this paper, we propose a novel framework that combines visual foundation models(VFM) and offline reinforcement learning(offline RL) to empower embodied visual tracking. We use a pre-trained VFM, such as "Tracking Anything", to extract semantic segmentation masks with text prompts. We then train a recurrent policy network with offline RL, e.g., Conservative Q-Learning, to learn from the collected demonstrations without online interactions. To further improve the robustness and generalization of the policy network, we also introduce a mask re-targeting mechanism and a multi-level data collection strategy. In this way, we can train a robust policy within an hour on a consumer-level GPU, e.g., Nvidia RTX 3090. We evaluate our agent on several high-fidelity environments with challenging situations, such as distraction and occlusion. The results show that our agent outperforms state-of-the-art methods in terms of sample efficiency, robustness to distractors, and generalization to unseen scenarios and targets. We also demonstrate the transferability of the learned agent from virtual environments to a real-world robot.

arXiv:2404.08763공식 코드 · Papers with Code

CATS: Contextually-Aware Thresholding for Sparsity in Large Language Models

Large Language Models (LLMs) have dramatically advanced AI applications, yet their deployment remains challenging due to their immense inference costs. Recent studies ameliorate the computational costs of LLMs by increasing their activation sparsity but suffer from significant performance degradation on downstream tasks. In this work, we introduce a new framework for sparsifying the activations of base LLMs and reducing inference costs, dubbed Contextually Aware Thresholding for Sparsity (CATS). CATS is relatively simple, easy to implement, and highly effective. At the heart of our framework is a new non-linear activation function. We demonstrate that CATS can be applied to various base models, including Mistral-7B and Llama2-7B, and outperforms existing sparsification techniques in downstream task performance. More precisely, CATS-based models often achieve downstream task performance within 1-2% of their base models without any fine-tuning and even at activation sparsity levels of 50%. Furthermore, CATS-based models converge faster and display better task performance than competing techniques when fine-tuning is applied. Finally, we develop a custom GPU kernel for efficient implementation of CATS that translates the activation of sparsity of CATS to real wall-clock time speedups. Our custom kernel implementation of CATS results in a ~15% improvement in wall-clock inference latency of token generation on both Llama-7B and Mistral-7B.

arXiv:2404.09526공식 코드 · Papers with Code

LoongServe: Efficiently Serving Long-Context Large Language Models with Elastic Sequence Parallelism

The context window of large language models (LLMs) is rapidly increasing, leading to a huge variance in resource usage between different requests as well as between different phases of the same request. Restricted by static parallelism strategies, existing LLM serving systems cannot efficiently utilize the underlying resources to serve variable-length requests in different phases. To address this problem, we propose a new parallelism paradigm, elastic sequence parallelism (ESP), to elastically adapt to the variance between different requests and phases. Based on ESP, we design and build LoongServe, an LLM serving system that (1) improves computation efficiency by elastically adjusting the degree of parallelism in real-time, (2) improves communication efficiency by reducing key-value cache migration overhead and overlapping partial decoding communication with computation, and (3) improves GPU memory efficiency by reducing key-value cache fragmentation across instances. Our evaluation under diverse real-world datasets shows that LoongServe improves the maximum throughput by up to 3.85$\times$ compared to the chunked prefill and 5.81$\times$ compared to the prefill-decoding disaggregation.

arXiv:2404.08788

Detecting AI-Generated Images via CLIP

As AI-generated image (AIGI) methods become more powerful and accessible, it has become a critical task to determine if an image is real or AI-generated. Because AIGI lack the signatures of photographs and have their own unique patterns, new models are needed to determine if an image is AI-generated. In this paper, we investigate the ability of the Contrastive Language-Image Pre-training (CLIP) architecture, pre-trained on massive internet-scale data sets, to perform this differentiation. We fine-tune CLIP on real images and AIGI from several generative models, enabling CLIP to determine if an image is AI-generated and, if so, determine what generation method was used to create it. We show that the fine-tuned CLIP architecture is able to differentiate AIGI as well or better than models whose architecture is specifically designed to detect AIGI. Our method will significantly increase access to AIGI-detecting tools and reduce the negative effects of AIGI on society, as our CLIP fine-tuning procedures require no architecture changes from publicly available model repositories and consume significantly less GPU resources than other AIGI detection models.

arXiv:2404.09683공식 코드 · Papers with Code

Post-Training Network Compression for 3D Medical Image Segmentation: Reducing Computational Efforts via Tucker Decomposition

We address the computational barrier of deploying advanced deep learning segmentation models in clinical settings by studying the efficacy of network compression through tensor decomposition. We propose a post-training Tucker factorization that enables the decomposition of pre-existing models to reduce computational requirements without impeding segmentation accuracy. We applied Tucker decomposition to the convolutional kernels of the TotalSegmentator (TS) model, an nnU-Net model trained on a comprehensive dataset for automatic segmentation of 117 anatomical structures. Our approach reduced the floating-point operations (FLOPs) and memory required during inference, offering an adjustable trade-off between computational efficiency and segmentation quality. This study utilized the publicly available TS dataset, employing various downsampling factors to explore the relationship between model size, inference speed, and segmentation performance. The application of Tucker decomposition to the TS model substantially reduced the model parameters and FLOPs across various compression rates, with limited loss in segmentation accuracy. We removed up to 88% of the model's parameters with no significant performance changes in the majority of classes after fine-tuning. Practical benefits varied across different graphics processing unit (GPU) architectures, with more distinct speed-ups on less powerful hardware. Post-hoc network compression via Tucker decomposition presents a viable strategy for reducing the computational demand of medical image segmentation models without substantially sacrificing accuracy. This approach enables the broader adoption of advanced deep learning technologies in clinical practice, offering a way to navigate the constraints of hardware capabilities.

arXiv:2404.08811

Reducing the Barriers to Entry for Foundation Model Training

The world has recently witnessed an unprecedented acceleration in demands for Machine Learning and Artificial Intelligence applications. This spike in demand has imposed tremendous strain on the underlying technology stack in supply chain, GPU-accelerated hardware, software, datacenter power density, and energy consumption. If left on the current technological trajectory, future demands show insurmountable spending trends, further limiting market players, stifling innovation, and widening the technology gap. To address these challenges, we propose a fundamental change in the AI training infrastructure throughout the technology ecosystem. The changes require advancements in supercomputing and novel AI training approaches, from high-end software to low-level hardware, microprocessor, and chip design, while advancing the energy efficiency required by a sustainable infrastructure. This paper presents the analytical framework that quantitatively highlights the challenges and points to the opportunities to reduce the barriers to entry for training large language models.

arXiv:2404.09172

LoopAnimate: Loopable Salient Object Animation

Research on diffusion model-based video generation has advanced rapidly. However, limitations in object fidelity and generation length hinder its practical applications. Additionally, specific domains like animated wallpapers require seamless looping, where the first and last frames of the video match seamlessly. To address these challenges, this paper proposes LoopAnimate, a novel method for generating videos with consistent start and end frames. To enhance object fidelity, we introduce a framework that decouples multi-level image appearance and textual semantic information. Building upon an image-to-image diffusion model, our approach incorporates both pixel-level and feature-level information from the input image, injecting image appearance and textual semantic embeddings at different positions of the diffusion model. Existing UNet-based video generation models require to input the entire videos during training to encode temporal and positional information at once. However, due to limitations in GPU memory, the number of frames is typically restricted to 16. To address this, this paper proposes a three-stage training strategy with progressively increasing frame numbers and reducing fine-tuning modules. Additionally, we introduce the Temporal E nhanced Motion Module(TEMM) to extend the capacity for encoding temporal and positional information up to 36 frames. The proposed LoopAnimate, which for the first time extends the single-pass generation length of UNet-based video generation models to 35 frames while maintaining high-quality video generation. Experiments demonstrate that LoopAnimate achieves state-of-the-art performance in both objective metrics, such as fidelity and temporal consistency, and subjective evaluation results.