最近,稀疏培训已成为有希望的范式,可在边缘设备上有效地深入学习。当前的研究主要致力于通过进一步增加模型稀疏性来降低培训成本。但是,增加的稀疏性并不总是理想的,因为它不可避免地会在极高的稀疏度下引入严重的准确性降解。本文打算探索其他可能的方向,以有效,有效地降低稀疏培训成本,同时保持准确性。为此,我们研究了两种技术,即层冻结和数据筛分。首先,层冻结方法在密集的模型训练和微调方面取得了成功,但在稀疏训练域中从未采用过。然而,稀疏训练的独特特征可能会阻碍层冻结技术的结合。因此,我们分析了在稀疏培训中使用层冻结技术的可行性和潜力,并发现它有可能节省大量培训成本。其次,我们提出了一种用于数据集有效培训的数据筛分方法,该方法通过确保在整个培训过程中仅使用部分数据集来进一步降低培训成本。我们表明,这两种技术都可以很好地整合到稀疏训练算法中,以形成一个通用框架,我们将其配置为SPFDE。我们的广泛实验表明,SPFDE可以显着降低培训成本,同时从三个维度中保留准确性:重量稀疏性,层冻结和数据集筛分。
translated by 谷歌翻译
持续学习的现有工作(CL)的重点是减轻灾难性遗忘,即学习新任务时过去任务的模型绩效恶化。但是,CL系统的训练效率不足,这限制了CL系统在资源有限的方案下的现实应用。在这项工作中,我们提出了一个名为“稀疏持续学习”(SPARCL)的新颖框架,这是第一个利用稀疏性以使边缘设备上具有成本效益的持续学习的研究。 SPARCL通过三个方面的协同作用来实现训练加速度和准确性保护:体重稀疏性,数据效率和梯度稀疏性。具体而言,我们建议在整个CL过程中学习一个稀疏网络,动态数据删除(DDR),以删除信息较少的培训数据和动态梯度掩盖(DGM),以稀疏梯度更新。他们每个人不仅提高了效率,而且进一步减轻了灾难性的遗忘。 SPARCL始终提高现有最新CL方法(SOTA)CL方法的训练效率最多减少了训练失败,而且令人惊讶的是,SOTA的准确性最多最多提高了1.7%。 SPARCL还优于通过将SOTA稀疏训练方法适应CL设置的效率和准确性获得的竞争基线。我们还评估了SPARCL在真实手机上的有效性,进一步表明了我们方法的实际潜力。
translated by 谷歌翻译
深度神经网络(DNN)在解决许多真实问题方面都有效。较大的DNN模型通常表现出更好的质量(例如,精度,精度),但它们的过度计算会导致长期推理时间。模型稀疏可以降低计算和内存成本,同时保持模型质量。大多数现有的稀疏算法是单向移除的重量,而其他人则随机或贪婪地探索每层进行修剪的小权重子集。这些算法的局限性降低了可实现的稀疏性水平。此外,许多算法仍然需要预先训练的密集模型,因此遭受大的内存占地面积。在本文中,我们提出了一种新颖的预定生长和修剪(间隙)方法,而无需预先培训密集模型。它通过反复生长一个层次的层来解决以前的作品的缺点,然后在一些训练后修剪回到稀疏。实验表明,使用所提出的方法修剪模型匹配或击败高度优化的密集模型的质量,在各种任务中以80%的稀疏度,例如图像分类,客观检测,3D对象分段和翻译。它们还优于模型稀疏的其他最先进的(SOTA)方法。作为一个例子,通过间隙获得的90%不均匀的稀疏resnet-50模型在想象中实现了77.9%的前1个精度,提高了先前的SOTA结果1.5%。所有代码将公开发布。
translated by 谷歌翻译
网络的稀疏性主要是由于其降低网络复杂性的能力而受欢迎。广泛的研究挖掘了梯度驱动的稀疏性。通常,这些方法是在体重独立性前提下构建的,但是与重量受到相互影响的事实相反。因此,他们的性能仍有待改进。在本文中,我们建议通过解决这种独立悖论来进一步优化梯度驱动的稀疏性(OPTG)。我们的动机来自最近对超级策略训练的进步,该进步表明,稀疏子网可以通过简单地更新掩码值而无需修改任何权重的情况下将其位于随机初始化的网络中。我们证明,超级手机训练是积累重量梯度,并可以部分解决独立悖论。因此,OPTG将Supermask训练集成到梯度驱动的稀疏度中,并且设计了专门的掩模优化器来解决独立悖论。实验表明,OPTG可以很好地超越许多现有的最先进的竞争对手。我们的代码可在\ url {https://github.com/zyxxmu/optg}上找到。
translated by 谷歌翻译
通过强迫连续重量的最多n非零,最近的N:M网络稀疏性因其两个有吸引力的优势而受到越来越多的关注:1)高稀疏性的有希望的表现。 2)对NVIDIA A100 GPU的显着加速。最近的研究需要昂贵的训练阶段或重型梯度计算。在本文中,我们表明N:M学习可以自然地将其描述为一个组合问题,该问题可以在有限的集合中寻找最佳组合候选者。由这种特征激励,我们以有效的分裂方式解决了n:m的稀疏性。首先,我们将重量向量分为$ c _ {\ text {m}}}^{\ text {n}} $组合s子集的固定大小N。然后,我们通过分配每个组合来征服组合问题,一个可学习的分数是共同优化了其关联权重。我们证明,引入的评分机制可以很好地模拟组合子集之间的相对重要性。通过逐渐去除低得分的子集,可以在正常训练阶段有效地优化N:M细粒稀疏性。全面的实验表明,我们的学习最佳组合(LBC)的表现始终如一,始终如一地比现成的N:m稀疏方法更好。我们的代码在\ url {https://github.com/zyxxmu/lbc}上发布。
translated by 谷歌翻译
Many applications require sparse neural networks due to space or inference time restrictions. There is a large body of work on training dense networks to yield sparse networks for inference, but this limits the size of the largest trainable sparse model to that of the largest trainable dense model. In this paper we introduce a method to train sparse neural networks with a fixed parameter count and a fixed computational cost throughout training, without sacrificing accuracy relative to existing dense-tosparse training methods. Our method updates the topology of the sparse network during training by using parameter magnitudes and infrequent gradient calculations. We show that this approach requires fewer floating-point operations (FLOPs) to achieve a given level of accuracy compared to prior techniques. We demonstrate state-of-the-art sparse training results on a variety of networks and datasets, including ResNet-50, MobileNets on Imagenet-2012, and RNNs on WikiText-103. Finally, we provide some insights into why allowing the topology to change during the optimization can overcome local minima encountered when the topology remains static * .
translated by 谷歌翻译
彩票票证假设(LTH)表明,密集的模型包含高度稀疏的子网(即获奖门票),可以隔离培训以完全准确。尽管做出了许多激动人心的努力,但仍有一个“常识”很少受到挑战:通过迭代级修剪(IMP)发现了一张获胜的票,因此由此产生的修剪子网仅具有非结构化的稀疏性。这一差距限制了在实践中赢得门票的吸引力,因为高度不规则的稀疏模式在硬件上加速的挑战是挑战性的。同时,直接将结构化修剪替换为非结构化的修剪,以更严重地损害绩效,并且通常无法找到获胜的票。在本文中,我们证明了第一个积极的结果是,总体上可以有效地找到结构上稀疏的获胜票。核心思想是在每一轮(非结构化)IMP之后附加“后处理技术”,以实施结构稀疏的形成。具体而言,我们首先在某些被认为很重要的通道中“重新填充”修剪元素,然后“重新组”非零元素以创建灵活的群体结构模式。我们确定的渠道和团体结构子网都赢得了彩票,并以现有硬件很容易支持的大量推理加速。广泛的实验,在多个网络骨架的不同数据集上进行,一致验证了我们的建议,表明LTH的硬件加速障碍现在已被删除。具体而言,结构上的获胜票最多可获得{64.93%,64.84%,60.23%}的运行时间节省,以{36%〜80%,74%,58%}的稀疏性在{Cifar,cifar,tiny-imageNet,imageNet}上保持可比较的精度。代码在https://github.com/vita-group/structure-lth上。
translated by 谷歌翻译
重量修剪是一种有效的模型压缩技术,可以解决在移动设备上实现实时深神经网络(DNN)推断的挑战。然而,由于精度劣化,难以利用硬件加速度,以及某些类型的DNN层的限制,难以降低的应用方案具有有限的应用方案。在本文中,我们提出了一般的细粒度的结构化修剪方案和相应的编译器优化,适用于任何类型的DNN层,同时实现高精度和硬件推理性能。随着使用我们的编译器优化所支持的不同层的灵活性,我们进一步探讨了确定最佳修剪方案的新问题,了解各种修剪方案的不同加速度和精度性能。两个修剪方案映射方法,一个是基于搜索,另一个是基于规则的,建议自动推导出任何给定DNN的每层的最佳修剪规则和块大小。实验结果表明,我们的修剪方案映射方法,以及一般细粒化结构修剪方案,优于最先进的DNN优化框架,最高可达2.48 $ \ times $和1.73 $ \ times $ DNN推理加速在CiFar-10和Imagenet DataSet上没有准确性损失。
translated by 谷歌翻译
深度神经网络(DNN)的记录断裂性能具有沉重的参数化,导致外部动态随机存取存储器(DRAM)进行存储。 DRAM访问的禁用能量使得在资源受限的设备上部署DNN是不普遍的,呼叫最小化重量和数据移动以提高能量效率。我们呈现SmartDeal(SD),算法框架,以进行更高成本的存储器存储/访问的较低成本计算,以便在推理和培训中积极提高存储和能量效率。 SD的核心是一种具有结构约束的新型重量分解,精心制作以释放硬件效率潜力。具体地,我们将每个重量张量分解为小基矩阵的乘积以及大的结构稀疏系数矩阵,其非零被量化为-2的功率。由此产生的稀疏和量化的DNN致力于为数据移动和重量存储而大大降低的能量,因为由于稀疏的比特 - 操作和成本良好的计算,恢复原始权重的最小开销。除了推理之外,我们采取了另一次飞跃来拥抱节能培训,引入创新技术,以解决培训时出现的独特障碍,同时保留SD结构。我们还设计专用硬件加速器,充分利用SD结构来提高实际能源效率和延迟。我们在不同的设置中对多个任务,模型和数据集进行实验。结果表明:1)应用于推理,SD可实现高达2.44倍的能效,通过实际硬件实现评估; 2)应用于培训,储存能量降低10.56倍,减少了10.56倍和4.48倍,与最先进的训练基线相比,可忽略的准确性损失。我们的源代码在线提供。
translated by 谷歌翻译
新兴的边缘情报应用程序要求服务器重新训练和更新部署在远程边缘节点上的深神经网络,以利用新收集的数据示例。不幸的是,由于高度严格的通信资源,在实践中可能不可能连续向这些边缘节点发送全面更新的权重。在本文中,我们提出了重量的深层部分更新范式,该范式巧妙地选择了一小部分权重以在每个服务器到边缘通信中进行更新,同时与完整更新相比实现了相似的性能。我们的方法是通过分析上限的部分更新和完整更新之间的损失差异来建立的,并且只能更新权重,从而对上限产生最大的贡献。广泛的实验结果证明了我们部分更新方法的功效,该方法在更新少量的权重的同时,可以达到高推理精度。
translated by 谷歌翻译
深度神经网络(DNN)的计算要求增加导致获得稀疏,且准确的DNN模型的兴趣。最近的工作已经调查了稀疏训练的更加困难的情况,其中DNN重量尽可能稀少,以减少训练期间的计算成本。现有的稀疏训练方法通常是经验的,并且可以具有相对于致密基线的准确性较低。在本文中,我们介绍了一种称为交替压缩/解压缩(AC / DC)训练DNN的一般方法,证明了算法变体的收敛,并表明AC / DC在类似的计算预算中准确地表现出现有的稀疏训练方法;在高稀疏水平下,AC / DC甚至优于现有的现有方法,依赖于准确的预训练密集模型。 AC / DC的一个重要属性是它允许联合培训密集和稀疏的型号,在训练过程结束时产生精确的稀疏密集模型对。这在实践中是有用的,其中压缩变体可能是为了在资源受限的设置中进行部署而不重新执行整个训练流,并且还为我们提供了深入和压缩模型之间的精度差距的见解。代码可在:https://github.com/ist-daslab/acdc。
translated by 谷歌翻译
Neural network pruning has been a well-established compression technique to enable deep learning models on resource-constrained devices. The pruned model is usually specialized to meet specific hardware platforms and training tasks (defined as deployment scenarios). However, existing pruning approaches rely heavily on training data to trade off model size, efficiency, and accuracy, which becomes ineffective for federated learning (FL) over distributed and confidential datasets. Moreover, the memory- and compute-intensive pruning process of most existing approaches cannot be handled by most FL devices with resource limitations. In this paper, we develop FedTiny, a novel distributed pruning framework for FL, to obtain specialized tiny models for memory- and computing-constrained participating devices with confidential local data. To alleviate biased pruning due to unseen heterogeneous data over devices, FedTiny introduces an adaptive batch normalization (BN) selection module to adaptively obtain an initially pruned model to fit deployment scenarios. Besides, to further improve the initial pruning, FedTiny develops a lightweight progressive pruning module for local finer pruning under tight memory and computational budgets, where the pruning policy for each layer is gradually determined rather than evaluating the overall deep model structure. Extensive experimental results demonstrate the effectiveness of FedTiny, which outperforms state-of-the-art baseline approaches, especially when compressing deep models to extremely sparse tiny models.
translated by 谷歌翻译
Turning the weights to zero when training a neural network helps in reducing the computational complexity at inference. To progressively increase the sparsity ratio in the network without causing sharp weight discontinuities during training, our work combines soft-thresholding and straight-through gradient estimation to update the raw, i.e. non-thresholded, version of zeroed weights. Our method, named ST-3 for straight-through/soft-thresholding/sparse-training, obtains SoA results, both in terms of accuracy/sparsity and accuracy/FLOPS trade-offs, when progressively increasing the sparsity ratio in a single training cycle. In particular, despite its simplicity, ST-3 favorably compares to the most recent methods, adopting differentiable formulations or bio-inspired neuroregeneration principles. This suggests that the key ingredients for effective sparsification primarily lie in the ability to give the weights the freedom to evolve smoothly across the zero state while progressively increasing the sparsity ratio. Source code and weights available at https://github.com/vanderschuea/stthree
translated by 谷歌翻译
Network pruning is widely used for reducing the heavy inference cost of deep models in low-resource settings. A typical pruning algorithm is a three-stage pipeline, i.e., training (a large model), pruning and fine-tuning. During pruning, according to a certain criterion, redundant weights are pruned and important weights are kept to best preserve the accuracy. In this work, we make several surprising observations which contradict common beliefs. For all state-of-the-art structured pruning algorithms we examined, fine-tuning a pruned model only gives comparable or worse performance than training that model with randomly initialized weights. For pruning algorithms which assume a predefined target network architecture, one can get rid of the full pipeline and directly train the target network from scratch. Our observations are consistent for multiple network architectures, datasets, and tasks, which imply that: 1) training a large, over-parameterized model is often not necessary to obtain an efficient final model, 2) learned "important" weights of the large model are typically not useful for the small pruned model, 3) the pruned architecture itself, rather than a set of inherited "important" weights, is more crucial to the efficiency in the final model, which suggests that in some cases pruning can be useful as an architecture search paradigm. Our results suggest the need for more careful baseline evaluations in future research on structured pruning methods. We also compare with the "Lottery Ticket Hypothesis" (Frankle & Carbin, 2019), and find that with optimal learning rate, the "winning ticket" initialization as used in Frankle & Carbin (2019) does not bring improvement over random initialization. * Equal contribution. † Work done while visiting UC Berkeley.
translated by 谷歌翻译
This paper proposed a Soft Filter Pruning (SFP) method to accelerate the inference procedure of deep Convolutional Neural Networks (CNNs). Specifically, the proposed SFP enables the pruned filters to be updated when training the model after pruning. SFP has two advantages over previous works: (1) Larger model capacity. Updating previously pruned filters provides our approach with larger optimization space than fixing the filters to zero. Therefore, the network trained by our method has a larger model capacity to learn from the training data. (2) Less dependence on the pretrained model. Large capacity enables SFP to train from scratch and prune the model simultaneously. In contrast, previous filter pruning methods should be conducted on the basis of the pre-trained model to guarantee their performance. Empirically, SFP from scratch outperforms the previous filter pruning methods. Moreover, our approach has been demonstrated effective for many advanced CNN architectures. Notably, on ILSCRC-2012, SFP reduces more than 42% FLOPs on ResNet-101 with even 0.2% top-5 accuracy improvement, which has advanced the state-of-the-art. Code is publicly available on GitHub: https://github.com/he-y/softfilter-pruning
translated by 谷歌翻译
最近对稀疏神经网络的作品已经证明了独立从头开始训练稀疏子网,以匹配其相应密集网络的性能。然而,识别这种稀疏的子网(获奖票)涉及昂贵的迭代火车 - 培训 - 培训过程(例如,彩票票证假设)或过度扩展的训练时间(例如,动态稀疏训练)。在这项工作中,我们在稀疏神经网络训练和深度合并技术之间汲取了独特的联系,产生了一个名为FreeTickets的新型集合学习框架。 FreeTickets而不是从密集的网络开始,随机初始化稀疏的子网,然后在动态调整其稀疏掩码的同时列举子网,从而在整个训练过程中产生许多不同的稀疏子网。 FreeTickets被定义为这些稀疏子网的集合,在这种单次通过,稀疏稀疏训练中自由获得,其仅使用Vanilla密集培训所需的计算资源的一小部分。此外,尽管是模型的集合,但与单一密集模型相比,FreeTickets的参数和训练拖鞋更少:这种看似反向直观的结果是由于每个子网的高稀疏性。与标准致密基线相比,观察到惯性基因术,以预测准确性,不确定度估计,鲁棒性和效率相比表现出显着的全面改进。 FreeTickets在ImageNet上只使用后者所需的四分之一的培训拖鞋,可以轻松地表达Naive Deep EndleBe。我们的结果提供了对稀疏神经网络的强度的见解,并表明稀疏性的好处超出了通常预期的推理效率。
translated by 谷歌翻译
The mainstream approach for filter pruning is usually either to force a hard-coded importance estimation upon a computation-heavy pretrained model to select "important" filters, or to impose a hyperparameter-sensitive sparse constraint on the loss objective to regularize the network training. In this paper, we present a novel filter pruning method, dubbed dynamic-coded filter fusion (DCFF), to derive compact CNNs in a computation-economical and regularization-free manner for efficient image classification. Each filter in our DCFF is firstly given an inter-similarity distribution with a temperature parameter as a filter proxy, on top of which, a fresh Kullback-Leibler divergence based dynamic-coded criterion is proposed to evaluate the filter importance. In contrast to simply keeping high-score filters in other methods, we propose the concept of filter fusion, i.e., the weighted averages using the assigned proxies, as our preserved filters. We obtain a one-hot inter-similarity distribution as the temperature parameter approaches infinity. Thus, the relative importance of each filter can vary along with the training of the compact CNN, leading to dynamically changeable fused filters without both the dependency on the pretrained model and the introduction of sparse constraints. Extensive experiments on classification benchmarks demonstrate the superiority of our DCFF over the compared counterparts. For example, our DCFF derives a compact VGGNet-16 with only 72.77M FLOPs and 1.06M parameters while reaching top-1 accuracy of 93.47% on CIFAR-10. A compact ResNet-50 is obtained with 63.8% FLOPs and 58.6% parameter reductions, retaining 75.60% top-1 accuracy on ILSVRC-2012. Our code, narrower models and training logs are available at https://github.com/lmbxmu/DCFF.
translated by 谷歌翻译
边缘设备上有限且动态的资源激励我们部署优化的深神经网络,该网络可以调整其子网络以适应不同的资源约束。但是,现有作品通常通过在手工制作的采样空间中搜索不同的网络体系结构来构建子网络,这不仅可以导致低标准的性能,而且可能导致设备上的重新配置开销。在本文中,我们提出了一种新颖的培训算法,动态的实时稀疏子网(着装)。着装通过基于行的非结构化稀疏度从相同的骨干网络采样多个子网络,并与加权损失并联训练这些子网络。着装还利用包括参数重复使用和基于行的细粒抽样在内的策略,以进行有效的存储消耗和有效的机上适应。公共视觉数据集的广泛实验表明,与最先进的子网络相比,着装的准确性明显更高。
translated by 谷歌翻译
关于稀疏神经网络训练(稀疏训练)的最新研究表明,通过从头开始训练本质上稀疏的神经网络可以实现绩效和效率之间的令人信服的权衡。现有的稀疏训练方法通常努力在一次跑步中找到最佳的稀疏子网,而无需涉及任何昂贵的密集或预训练步骤。例如,作为最突出的方向之一,动态稀疏训练(DST)能够通过在训练过程中迭代发展稀疏拓扑来实现竞争性训练的竞争性能。在本文中,我们认为最好分配有限的资源来创建多个低损失的稀疏子网并将其超级置于更强的基因,而不是完全分配所有资源以找到单个子网络。为了实现这一目标,需要两个Desiderata:(1)在一个培训过程中有效生产许多低损失的子网,即所谓的廉价门票,仅限于用于密集培训的标准培训时间; (2)将这些廉价的门票有效地超级为一个更强的子网,而无需超越约束参数预算。为了证实我们的猜想,我们提出了一种新颖的稀疏训练方法,称为\ textbf {sup-tickets},可以在单个稀疏到较小的训练过程中同时满足上述两个desiderata。在CIFAR-10/100和Imagenet上的各种现代体系结构中,我们表明,SUP-Tickets与现有的稀疏训练方法无缝集成,并显示出一致的性能提高。
translated by 谷歌翻译
Over-parameterization of deep neural networks (DNNs) has shown high prediction accuracy for many applications. Although effective, the large number of parameters hinders its popularity on resource-limited devices and has an outsize environmental impact. Sparse training (using a fixed number of nonzero weights in each iteration) could significantly mitigate the training costs by reducing the model size. However, existing sparse training methods mainly use either random-based or greedy-based drop-and-grow strategies, resulting in local minimal and low accuracy. In this work, to assist explainable sparse training, we propose important weights Exploitation and coverage Exploration to characterize Dynamic Sparse Training (DST-EE), and provide quantitative analysis of these two metrics. We further design an acquisition function and provide the theoretical guarantees for the proposed method and clarify its convergence property. Experimental results show that sparse models (up to 98\% sparsity) obtained by our proposed method outperform the SOTA sparse training methods on a wide variety of deep learning tasks. On VGG-19 / CIFAR-100, ResNet-50 / CIFAR-10, ResNet-50 / CIFAR-100, our method has even higher accuracy than dense models. On ResNet-50 / ImageNet, the proposed method has up to 8.2\% accuracy improvement compared to SOTA sparse training methods.
translated by 谷歌翻译