PyTorchKR Papers

최신 논문

총 5,576편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2305.07027공식 코드 · Papers with Code

EfficientViT: Memory Efficient Vision Transformer with Cascaded Group Attention

Vision transformers have shown great success due to their high model capabilities. However, their remarkable performance is accompanied by heavy computation costs, which makes them unsuitable for real-time applications. In this paper, we propose a family of high-speed vision transformers named EfficientViT. We find that the speed of existing transformer models is commonly bounded by memory inefficient operations, especially the tensor reshaping and element-wise functions in MHSA. Therefore, we design a new building block with a sandwich layout, i.e., using a single memory-bound MHSA between efficient FFN layers, which improves memory efficiency while enhancing channel communication. Moreover, we discover that the attention maps share high similarities across heads, leading to computational redundancy. To address this, we present a cascaded group attention module feeding attention heads with different splits of the full feature, which not only saves computation cost but also improves attention diversity. Comprehensive experiments demonstrate EfficientViT outperforms existing efficient models, striking a good trade-off between speed and accuracy. For instance, our EfficientViT-M5 surpasses MobileNetV3-Large by 1.9% in accuracy, while getting 40.4% and 45.2% higher throughput on Nvidia V100 GPU and Intel Xeon CPU, respectively. Compared to the recent efficient model MobileViT-XXS, EfficientViT-M2 achieves 1.8% superior accuracy, while running 5.8x/3.7x faster on the GPU/CPU, and 7.4x faster when converted to ONNX format. Code and models are available at https://github.com/microsoft/Cream/tree/main/EfficientViT.

arXiv:2305.06908공식 코드 · Papers with Code

CoMoSpeech: One-Step Speech and Singing Voice Synthesis via Consistency Model

Denoising diffusion probabilistic models (DDPMs) have shown promising performance for speech synthesis. However, a large number of iterative steps are required to achieve high sample quality, which restricts the inference speed. Maintaining sample quality while increasing sampling speed has become a challenging task. In this paper, we propose a "Co"nsistency "Mo"del-based "Speech" synthesis method, CoMoSpeech, which achieve speech synthesis through a single diffusion sampling step while achieving high audio quality. The consistency constraint is applied to distill a consistency model from a well-designed diffusion-based teacher model, which ultimately yields superior performances in the distilled CoMoSpeech. Our experiments show that by generating audio recordings by a single sampling step, the CoMoSpeech achieves an inference speed more than 150 times faster than real-time on a single NVIDIA A100 GPU, which is comparable to FastSpeech2, making diffusion-sampling based speech synthesis truly practical. Meanwhile, objective and subjective evaluations on text-to-speech and singing voice synthesis show that the proposed teacher models yield the best audio quality, and the one-step sampling based CoMoSpeech achieves the best inference speed with better or comparable audio quality to other conventional multi-step diffusion model baselines. Audio samples are available at https://comospeech.github.io/.

arXiv:2305.06404

LACoS-BLOOM: Low-rank Adaptation with Contrastive objective on 8 bits Siamese-BLOOM

Text embeddings are useful features for several NLP applications, such as sentence similarity, text clustering, and semantic search. In this paper, we present a Low-rank Adaptation with a Contrastive objective on top of 8-bit Siamese-BLOOM, a multilingual large language model optimized to produce semantically meaningful word embeddings. The innovation is threefold. First, we cast BLOOM weights to 8-bit values. Second, we fine-tune BLOOM with a scalable adapter (LoRA) and 8-bit Adam optimizer for sentence similarity classification. Third, we apply a Siamese architecture on BLOOM model with a contrastive objective to ease the multi-lingual labeled data scarcity. The experiment results show the quality of learned embeddings from LACoS-BLOOM is proportional to the number of model parameters and the amount of unlabeled training data. With the parameter efficient fine-tuning design, we are able to run BLOOM 7.1 billion parameters end-to-end on a single GPU machine with 32GB memory. Compared to previous solution Sentence-BERT, we achieve significant improvement on both English and multi-lingual STS tasks.

arXiv:2305.05920

Fast Distributed Inference Serving for Large Language Models

Large language models (LLMs) power a new generation of interactive AI applications exemplified by ChatGPT. The interactive nature of these applications demands low latency for LLM inference. Existing LLM serving systems use run-to-completion processing for inference jobs, which suffers from head-of-line blocking and long latency. We present FastServe, a distributed inference serving system for LLMs. FastServe exploits the autoregressive pattern of LLM inference to enable preemption at the granularity of each output token. FastServe uses preemptive scheduling to minimize latency with a novel skip-join Multi-Level Feedback Queue scheduler. Based on the new semi-information-agnostic setting of LLM inference, the scheduler leverages the input length information to assign an appropriate initial queue for each arrival job to join. The higher priority queues than the joined queue are skipped to reduce demotions. We design an efficient GPU memory management mechanism that proactively offloads and uploads intermediate state between GPU memory and host memory for LLM inference. We build a system prototype of FastServe and experimental results show that compared to the state-of-the-art solution vLLM, FastServe improves the throughput by up to 31.4x and 17.9x under the same average and tail latency requirements, respectively.

arXiv:2305.05780

Enhancing Gappy Speech Audio Signals with Generative Adversarial Networks

Gaps, dropouts and short clips of corrupted audio are a common problem and particularly annoying when they occur in speech. This paper uses machine learning to regenerate gaps of up to 320ms in an audio speech signal. Audio regeneration is translated into image regeneration by transforming audio into a Mel-spectrogram and using image in-painting to regenerate the gaps. The full Mel-spectrogram is then transferred back to audio using the Parallel-WaveGAN vocoder and integrated into the audio stream. Using a sample of 1300 spoken audio clips of between 1 and 10 seconds taken from the publicly-available LJSpeech dataset our results show regeneration of audio gaps in close to real time using GANs with a GPU equipped system. As expected, the smaller the gap in the audio, the better the quality of the filled gaps. On a gap of 240ms the average mean opinion score (MOS) for the best performing models was 3.737, on a scale of 1 (worst) to 5 (best) which is sufficient for a human to perceive as close to uninterrupted human speech.

arXiv:2305.05947

iEdit: Localised Text-guided Image Editing with Weak Supervision

Diffusion models (DMs) can generate realistic images with text guidance using large-scale datasets. However, they demonstrate limited controllability in the output space of the generated images. We propose a novel learning method for text-guided image editing, namely \texttt{iEdit}, that generates images conditioned on a source image and a textual edit prompt. As a fully-annotated dataset with target images does not exist, previous approaches perform subject-specific fine-tuning at test time or adopt contrastive learning without a target image, leading to issues on preserving the fidelity of the source image. We propose to automatically construct a dataset derived from LAION-5B, containing pseudo-target images with their descriptive edit prompts given input image-caption pairs. This dataset gives us the flexibility of introducing a weakly-supervised loss function to generate the pseudo-target image from the latent noise of the source image conditioned on the edit prompt. To encourage localised editing and preserve or modify spatial structures in the image, we propose a loss function that uses segmentation masks to guide the editing during training and optionally at inference. Our model is trained on the constructed dataset with 200K samples and constrained GPU resources. It shows favourable results against its counterparts in terms of image fidelity, CLIP alignment score and qualitatively for editing both generated and real images.

arXiv:2305.05490공식 코드 · Papers with Code

Real-time instance segmentation with polygons using an Intersection-over-Union loss

Predicting a binary mask for an object is more accurate but also more computationally expensive than a bounding box. Polygonal masks as developed in CenterPoly can be a good compromise. In this paper, we improve over CenterPoly by enhancing the classical regression L1 loss with a novel region-based loss and a novel order loss, as well as with a new training process for the vertices prediction head. Moreover, the previous methods that predict polygonal masks use different coordinate systems, but it is not clear if one is better than another, if we abstract the architecture requirement. We therefore investigate their impact on the prediction. We also use a new evaluation protocol with oracle predictions for the detection head, to further isolate the segmentation process and better compare the polygonal masks with binary masks. Our instance segmentation method is trained and tested with challenging datasets containing urban scenes, with a high density of road users. Experiments show, in particular, that using a combination of a regression loss and a region-based loss allows significant improvements on the Cityscapes and IDD test set compared to CenterPoly. Moreover the inference stage remains fast enough to reach real-time performance with an average of 0.045 s per frame for 2048$\times$1024 images on a single RTX 2070 GPU. The code is available $\href{https://github.com/KatiaJDL/CenterPoly-v2}{\text{here}}$.

arXiv:2305.04526

CrAFT: Compression-Aware Fine-Tuning for Efficient Visual Task Adaptation

Transfer learning has become a popular task adaptation method in the era of foundation models. However, many foundation models require large storage and computing resources, which makes off-the-shelf deployment impractical. Post-training compression techniques such as pruning and quantization can help lower deployment costs. Unfortunately, the resulting performance degradation limits the usability and benefits of such techniques. To close this performance gap, we propose CrAFT, a simple fine-tuning framework that enables effective post-training network compression. In CrAFT, users simply employ the default fine-tuning schedule along with sharpness minimization objective, simultaneously facilitating task adaptation and compression-friendliness. Contrary to the conventional sharpness minimization techniques, which are applied during pretraining, the CrAFT approach adds negligible training overhead as fine-tuning is done in under a couple of minutes or hours with a single GPU. The effectiveness of CrAFT, which is a general-purpose tool that can significantly boost one-shot pruning and post-training quantization, is demonstrated on both convolution-based and attention-based vision foundation models on a variety of target tasks. The code will be made publicly available.

arXiv:2305.04374

Spatiotemporally Consistent HDR Indoor Lighting Estimation

We propose a physically-motivated deep learning framework to solve a general version of the challenging indoor lighting estimation problem. Given a single LDR image with a depth map, our method predicts spatially consistent lighting at any given image position. Particularly, when the input is an LDR video sequence, our framework not only progressively refines the lighting prediction as it sees more regions, but also preserves temporal consistency by keeping the refinement smooth. Our framework reconstructs a spherical Gaussian lighting volume (SGLV) through a tailored 3D encoder-decoder, which enables spatially consistent lighting prediction through volume ray tracing, a hybrid blending network for detailed environment maps, an in-network Monte-Carlo rendering layer to enhance photorealism for virtual object insertion, and recurrent neural networks (RNN) to achieve temporally consistent lighting prediction with a video sequence as the input. For training, we significantly enhance the OpenRooms public dataset of photorealistic synthetic indoor scenes with around 360K HDR environment maps of much higher resolution and 38K video sequences, rendered with GPU-based path tracing. Experiments show that our framework achieves lighting prediction with higher quality compared to state-of-the-art single-image or video-based methods, leading to photorealistic AR applications such as object insertion.

arXiv:2305.03152공식 코드 · Papers with Code

Communication-Efficient Graph Neural Networks with Probabilistic Neighborhood Expansion Analysis and Caching

Training and inference with graph neural networks (GNNs) on massive graphs has been actively studied since the inception of GNNs, owing to the widespread use and success of GNNs in applications such as recommendation systems and financial forensics. This paper is concerned with minibatch training and inference with GNNs that employ node-wise sampling in distributed settings, where the necessary partitioning of vertex features across distributed storage causes feature communication to become a major bottleneck that hampers scalability. To significantly reduce the communication volume without compromising prediction accuracy, we propose a policy for caching data associated with frequently accessed vertices in remote partitions. The proposed policy is based on an analysis of vertex-wise inclusion probabilities (VIP) during multi-hop neighborhood sampling, which may expand the neighborhood far beyond the partition boundaries of the graph. VIP analysis not only enables the elimination of the communication bottleneck, but it also offers a means to organize in-memory data by prioritizing GPU storage for the most frequently accessed vertex features. We present SALIENT++, which extends the prior state-of-the-art SALIENT system to work with partitioned feature data and leverages the VIP-driven caching policy. SALIENT++ retains the local training efficiency and scalability of SALIENT by using a deep pipeline and drastically reducing communication volume while consuming only a fraction of the storage required by SALIENT. We provide experimental results with the Open Graph Benchmark data sets and demonstrate that training a 3-layer GraphSAGE model with SALIENT++ on 8 single-GPU machines is 7.1 faster than with SALIENT on 1 single-GPU machine, and 12.7 faster than with DistDGL on 8 single-GPU machines.

arXiv:2305.03726공식 코드 · Papers with Code

Otter: A Multi-Modal Model with In-Context Instruction Tuning

Large language models (LLMs) have demonstrated significant universal capabilities as few/zero-shot learners in various tasks due to their pre-training on vast amounts of text data, as exemplified by GPT-3, which boosted to InstrctGPT and ChatGPT, effectively following natural language instructions to accomplish real-world tasks. In this paper, we propose to introduce instruction tuning into multi-modal models, motivated by the Flamingo model's upstream interleaved format pretraining dataset. We adopt a similar approach to construct our MultI-Modal In-Context Instruction Tuning (MIMIC-IT) dataset. We then introduce Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following ability and in-context learning. We also optimize OpenFlamingo's implementation for researchers, democratizing the required training resources from 1$\times$ A100 GPU to 4$\times$ RTX-3090 GPUs, and integrate both OpenFlamingo and Otter into Huggingface Transformers for more researchers to incorporate the models into their customized training and inference pipelines.

arXiv:2305.03716공식 코드 · Papers with Code

3D Small Object Detection with Dynamic Spatial Pruning

In this paper, we propose an efficient feature pruning strategy for 3D small object detection. Conventional 3D object detection methods struggle on small objects due to the weak geometric information from a small number of points. Although increasing the spatial resolution of feature representations can improve the detection performance on small objects, the additional computational overhead is unaffordable. With in-depth study, we observe the growth of computation mainly comes from the upsampling operation in the decoder of 3D detector. Motivated by this, we present a multi-level 3D detector named DSPDet3D which benefits from high spatial resolution to achieves high accuracy on small object detection, while reducing redundant computation by only focusing on small object areas. Specifically, we theoretically derive a dynamic spatial pruning (DSP) strategy to prune the redundant spatial representation of 3D scene in a cascade manner according to the distribution of objects. Then we design DSP module following this strategy and construct DSPDet3D with this efficient module. On ScanNet and TO-SCENE dataset, our method achieves leading performance on small object detection. Moreover, DSPDet3D trained with only ScanNet rooms can generalize well to scenes in larger scale. It takes less than 2s to directly process a whole building consisting of more than 4500k points while detecting out almost all objects, ranging from cups to beds, on a single RTX 3090 GPU. Project page: https://xuxw98.github.io/DSPDet3D/.

arXiv:2305.03416

Evolution under Length Constraints for CNN Architecture design

In recent years, the CNN architectures designed by evolution algorithms have proven to be competitive with handcrafted architectures designed by experts. However, these algorithms need a lot of computational power, which is beyond the capabilities of most researchers and engineers. To overcome this problem, we propose an evolution architecture under length constraints. It consists of two algorithms: a search length strategy to find an optimal space and a search architecture strategy based on genetic algorithm to find the best individual in the optimal space. Our algorithms reduce drastically resource cost and also keep good performance. On the Cifar-10 dataset, our framework presents outstanding performance with an error rate of 5.12% and only 4.6 GPU a day to converge to the optimal individual -22 GPU a day less than the lowest cost automatic evolutionary algorithm in the peer competition.

arXiv:2305.03252

HeteroEdge: Addressing Asymmetry in Heterogeneous Collaborative Autonomous Systems

Gathering knowledge about surroundings and generating situational awareness for IoT devices is of utmost importance for systems developed for smart urban and uncontested environments. For example, a large-area surveillance system is typically equipped with multi-modal sensors such as cameras and LIDARs and is required to execute deep learning algorithms for action, face, behavior, and object recognition. However, these systems face power and memory constraints due to their ubiquitous nature, making it crucial to optimize data processing, deep learning algorithm input, and model inference communication. In this paper, we propose a self-adaptive optimization framework for a testbed comprising two Unmanned Ground Vehicles (UGVs) and two NVIDIA Jetson devices. This framework efficiently manages multiple tasks (storage, processing, computation, transmission, inference) on heterogeneous nodes concurrently. It involves compressing and masking input image frames, identifying similar frames, and profiling devices to obtain boundary conditions for optimization.. Finally, we propose and optimize a novel parameter split-ratio, which indicates the proportion of the data required to be offloaded to another device while considering the networking bandwidth, busy factor, memory (CPU, GPU, RAM), and power constraints of the devices in the testbed. Our evaluations captured while executing multiple tasks (e.g., PoseNet, SegNet, ImageNet, DetectNet, DepthNet) simultaneously, reveal that executing 70% (split-ratio=70%) of the data on the auxiliary node minimizes the offloading latency by approx. 33% (18.7 ms/image to 12.5 ms/image) and the total operation time by approx. 47% (69.32s to 36.43s) compared to the baseline configuration (executing on the primary node).

arXiv:2305.03173공식 코드 · Papers with Code

New Adversarial Image Detection Based on Sentiment Analysis

Deep Neural Networks (DNNs) are vulnerable to adversarial examples, while adversarial attack models, e.g., DeepFool, are on the rise and outrunning adversarial example detection techniques. This paper presents a new adversarial example detector that outperforms state-of-the-art detectors in identifying the latest adversarial attacks on image datasets. Specifically, we propose to use sentiment analysis for adversarial example detection, qualified by the progressively manifesting impact of an adversarial perturbation on the hidden-layer feature maps of a DNN under attack. Accordingly, we design a modularized embedding layer with the minimum learnable parameters to embed the hidden-layer feature maps into word vectors and assemble sentences ready for sentiment analysis. Extensive experiments demonstrate that the new detector consistently surpasses the state-of-the-art detection algorithms in detecting the latest attacks launched against ResNet and Inception neutral networks on the CIFAR-10, CIFAR-100 and SVHN datasets. The detector only has about 2 million parameters, and takes shorter than 4.6 milliseconds to detect an adversarial example generated by the latest attack models using a Tesla K80 GPU card.

arXiv:2305.03045공식 코드 · Papers with Code

OctFormer: Octree-based Transformers for 3D Point Clouds

We propose octree-based transformers, named OctFormer, for 3D point cloud learning. OctFormer can not only serve as a general and effective backbone for 3D point cloud segmentation and object detection but also have linear complexity and is scalable for large-scale point clouds. The key challenge in applying transformers to point clouds is reducing the quadratic, thus overwhelming, computation complexity of attentions. To combat this issue, several works divide point clouds into non-overlapping windows and constrain attentions in each local window. However, the point number in each window varies greatly, impeding the efficient execution on GPU. Observing that attentions are robust to the shapes of local windows, we propose a novel octree attention, which leverages sorted shuffled keys of octrees to partition point clouds into local windows containing a fixed number of points while permitting shapes of windows to change freely. And we also introduce dilated octree attention to expand the receptive field further. Our octree attention can be implemented in 10 lines of code with open-sourced libraries and runs 17 times faster than other point cloud attentions when the point number exceeds 200k. Built upon the octree attention, OctFormer can be easily scaled up and achieves state-of-the-art performances on a series of 3D segmentation and detection benchmarks, surpassing previous sparse-voxel-based CNNs and point cloud transformers in terms of both efficiency and effectiveness. Notably, on the challenging ScanNet200 dataset, OctFormer outperforms sparse-voxel-based CNNs by 7.3 in mIoU. Our code and trained models are available at https://wang-ps.github.io/octformer.

arXiv:2305.02510공식 코드 · Papers with Code

SuperNeuro: A Fast and Scalable Simulator for Neuromorphic Computing

In many neuromorphic workflows, simulators play a vital role for important tasks such as training spiking neural networks (SNNs), running neuroscience simulations, and designing, implementing and testing neuromorphic algorithms. Currently available simulators are catered to either neuroscience workflows (such as NEST and Brian2) or deep learning workflows (such as BindsNET). While the neuroscience-based simulators are slow and not very scalable, the deep learning-based simulators do not support certain functionalities such as synaptic delay that are typical of neuromorphic workloads. In this paper, we address this gap in the literature and present SuperNeuro, which is a fast and scalable simulator for neuromorphic computing, capable of both homogeneous and heterogeneous simulations as well as GPU acceleration. We also present preliminary results comparing SuperNeuro to widely used neuromorphic simulators such as NEST, Brian2 and BindsNET in terms of computation times. We demonstrate that SuperNeuro can be approximately 10--300 times faster than some of the other simulators for small sparse networks. On large sparse and large dense networks, SuperNeuro can be approximately 2.2 and 3.4 times faster than the other simulators respectively.

arXiv:2305.01883공식 코드 · Papers with Code

A Lightweight CNN-Transformer Model for Learning Traveling Salesman Problems

Several studies have attempted to solve traveling salesman problems (TSPs) using various deep learning techniques. Among them, Transformer-based models show state-of-the-art performance even for large-scale Traveling Salesman Problems (TSPs). However, they are based on fully-connected attention models and suffer from large computational complexity and GPU memory usage. Our work is the first CNN-Transformer model based on a CNN embedding layer and partial self-attention for TSP. Our CNN-Transformer model is able to better learn spatial features from input data using a CNN embedding layer compared with the standard Transformer-based models. It also removes considerable redundancy in fully-connected attention models using the proposed partial self-attention. Experimental results show that the proposed CNN embedding layer and partial self-attention are very effective in improving performance and computational complexity. The proposed model exhibits the best performance in real-world datasets and outperforms other existing state-of-the-art (SOTA) Transformer-based models in various aspects. Our code is publicly available at https://github.com/cm8908/CNN_Transformer3.

arXiv:2305.02299공식 코드 · Papers with Code

Dynamic Sparse Training with Structured Sparsity

Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically less computationally expensive, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity by imposing a constant fan-in constraint. Using our empirical analysis of existing DST methods at high sparsity, we additionally employ a neuron ablation method which enables SRigL to achieve state-of-the-art sparse-to-sparse structured DST performance on a variety of Neural Network (NN) architectures. Using a 90% sparse linear layer, we demonstrate a real-world acceleration of 3.4x/2.5x on CPU for online inference and 1.7x/13.0x on GPU for inference with a batch size of 256 when compared to equivalent dense/unstructured (CSR) sparse layers, respectively.

arXiv:2305.01255공식 코드 · Papers with Code

RT-K-Net: Revisiting K-Net for Real-Time Panoptic Segmentation

Panoptic segmentation is one of the most challenging scene parsing tasks, combining the tasks of semantic segmentation and instance segmentation. While much progress has been made, few works focus on the real-time application of panoptic segmentation methods. In this paper, we revisit the recently introduced K-Net architecture. We propose vital changes to the architecture, training, and inference procedure, which massively decrease latency and improve performance. Our resulting RT-K-Net sets a new state-of-the-art performance for real-time panoptic segmentation methods on the Cityscapes dataset and shows promising results on the challenging Mapillary Vistas dataset. On Cityscapes, RT-K-Net reaches 60.2 % PQ with an average inference time of 32 ms for full resolution 1024x2048 pixel images on a single Titan RTX GPU. On Mapillary Vistas, RT-K-Net reaches 33.2 % PQ with an average inference time of 69 ms. Source code is available at https://github.com/markusschoen/RT-K-Net.