اشترك بالحزمة الذهبية واحصل على وصول غير محدود شمرا أكاديميا
تسجيل مستخدم جديدQuantization of Generative Adversarial Networks for Efficient Inference: a Methodological Study
65
0
0.0
(
0
)
تأليف
Pavel Andreev
اسأل ChatGPT حول البحث
ﻻ يوجد ملخص باللغة العربية
Generative adversarial networks (GANs) have an enormous potential impact on digital content creation, e.g., photo-realistic digital avatars, semantic content editing, and quality enhancement of speech and images. However, the performance of modern GANs comes together with massive amounts of computations performed during the inference and high energy consumption. That complicates, or even makes impossible, their deployment on edge devices. The problem can be reduced with quantization -- a neural network compression technique that facilitates hardware-friendly inference by replacing floating-point computations with low-bit integer ones. While quantization is well established for discriminative models, the performance of modern quantization techniques in application to GANs remains unclear. GANs generate content of a more complex structure than discriminative models, and thus quantization of GANs is significantly more challenging. To tackle this problem, we perform an extensive experimental study of state-of-art quantization techniques on three diverse GAN architectures, namely StyleGAN, Self-Attention GAN, and CycleGAN. As a result, we discovered practical recipes that allowed us to successfully quantize these models for inference with 4/8-bit weights and 8-bit activations while preserving the quality of the original full-precision models.
قيم البحث
اقرأ أيضاً
Subsampling unconditional generative adversarial networks (GANs) to improve the overall image quality has been studied recently. However, these methods often require high training costs (e.g., storage space, parameter tuning) and may be inefficient o
r even inapplicable for subsampling conditional GANs, such as class-conditional GANs and continuous conditional GANs (CcGANs), when the condition has many distinct values. In this paper, we propose an efficient method called conditional density ratio estimation in feature space with conditional Softplus loss (cDRE-F-cSP). With cDRE-F-cSP, we estimate an images conditional density ratio based on a novel conditional Softplus (cSP) loss in the feature space learned by a specially designed ResNet-34 or sparse autoencoder. We then derive the error bound of a conditional density ratio model trained with the proposed cSP loss. Finally, we propose a rejection sampling scheme, termed cDRE-F-cSP+RS, which can subsample both class-conditional GANs and CcGANs efficiently. An extra filtering scheme is also developed for CcGANs to increase the label consistency. Experiments on CIFAR-10 and Tiny-ImageNet datasets show that cDRE-F-cSP+RS can substantially improve the Intra-FID and FID scores of BigGAN. Experiments on RC-49 and UTKFace datasets demonstrate that cDRE-F-cSP+RS also improves Intra-FID, Diversity, and Label Score of CcGANs. Moreover, to show the high efficiency of cDRE-F-cSP+RS, we compare it with the state-of-the-art unconditional subsampling method (i.e., DRE-F-SP+RS). With comparable or even better performance, cDRE-F-cSP+RS only requires about textbf{10}% and textbf{1.7}% of the training costs spent respectively on CIFAR-10 and UTKFace by DRE-F-SP+RS.
Recently, sampling methods have been successfully applied to enhance the sample quality of Generative Adversarial Networks (GANs). However, in practice, they typically have poor sample efficiency because of the independent proposal sampling from the
generator. In this work, we propose REP-GAN, a novel sampling method that allows general dependent proposals by REParameterizing the Markov chains into the latent space of the generator. Theoretically, we show that our reparameterized proposal admits a closed-form Metropolis-Hastings acceptance ratio. Empirically, extensive experiments on synthetic and real datasets demonstrate that our REP-GAN largely improves the sample efficiency and obtains better sample quality simultaneously.
The advent of generative adversarial networks (GAN) has enabled new capabilities in synthesis, interpolation, and data augmentation heretofore considered very challenging. However, one of the common assumptions in most GAN architectures is the assump
tion of simple parametric latent-space distributions. While easy to implement, a simple latent-space distribution can be problematic for uses such as interpolation. This is due to distributional mismatches when samples are interpolated in the latent space. We present a straightforward formalization of this problem; using basic results from probability theory and off-the-shelf-optimization tools, we develop ways to arrive at appropriate non-parametric priors. The obtained prior exhibits unusual qualitative properties in terms of its shape, and quantitative benefits in terms of lower divergence with its mid-point distribution. We demonstrate that our designed prior helps improve image generation along any Euclidean straight line during interpolation, both qualitatively and quantitatively, without any additional training or architectural modifications. The proposed formulation is quite flexible, paving the way to impose newer constraints on the latent-space statistics.
LHCb is one of the major experiments operating at the Large Hadron Collider at CERN. The richness of the physics program and the increasing precision of the measurements in LHCb lead to the need of ever larger simulated samples. This need will increa
se further when the upgraded LHCb detector will start collecting data in the LHC Run 3. Given the computing resources pledged for the production of Monte Carlo simulated events in the next years, the use of fast simulation techniques will be mandatory to cope with the expected dataset size. In LHCb generative models, which are nowadays widely used for computer vision and image processing are being investigated in order to accelerate the generation of showers in the calorimeter and high-level responses of Cherenkov detector. We demonstrate that this approach provides high-fidelity results along with a significant speed increase and discuss possible implication of these results. We also present an implementation of this algorithm into LHCb simulation software and validation tests.
Generative Adversarial Networks (GANs) have been employed with certain success for image translation tasks between optical and real-valued SAR intensity imagery. Applications include aiding interpretability of SAR scenes with their optical counterpar
ts by artificial patch generation and automatic SAR-optical scene matching. The synthesis of artificial complex-valued InSAR image stacks asks for, besides good perceptual quality, more stringent quality metrics like phase noise and phase coherence. This paper provides a signal processing model of generative CNN structures, describes effects influencing those quality metrics and presents a mapping scheme of complex-valued data to given CNN structures based on popular Deep Learning frameworks.
الأسئلة المقترحة
سجل دخول لتتمكن من نشر تعليقات
التعليقات
جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
