转移学习是一种经典范式，通过该范式，在大型“上游”数据集上佩戴的模型适于在“下游”专业数据集中产生良好的结果。通常，据了解，“上游”数据集上的更准确的模型将提供更好的转移精度“下游”。在这项工作中，我们在想象的神经网络（CNNS）的背景下对这种现象进行了深入的调查，这些现象已经在想象的数据集上训练的情况下被修剪 - 这是通过缩小它们的连接来压缩。具体地，我们考虑使用通过应用几种最先进的修剪方法而获得的非结构化修剪模型的转移，包括基于幅度的，二阶，重新增长和正规化方法，在12个标准转移任务的上下文中。简而言之，我们的研究表明，即使在高稀稀物质，稀疏的型号也可以匹配或甚至优于致密模型的转移性能，并且在此操作时，可以导致显着的推论甚至培训加速度。与此同时，我们观察和分析不同修剪方法行为的显着差异。

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。

最近对深神经网络（DNN）效率的重点已导致了模型压缩方法的重要工作，其中重量修剪是最受欢迎的方法之一。同时，有快速增长的计算支持，以有效地执行通过修剪获得的非结构化模型。但是，大多数现有的修剪方法最小化仅剩余权重的数量，即模型的大小，而不是针对推理时间进行优化。我们通过引入SPDY来解决这一差距，SPDY是一种新的压缩方法，该方法会自动确定层次的稀疏性目标，可以在给定系统上实现所需的推理速度，同时最大程度地减少准确性损失。 SPDY由两种新技术组成：第一个是一种有效的动态编程算法，用于求解一组给定的层敏感性得分，以解决加速约束的层压缩问题；第二个是一个局部搜索程序，用于确定准确的层敏感性得分。跨流行视觉和语言模型的实验表明，SPDY可以保证相对于现有策略的恢复较高的准确性，无论是一次性和逐步修剪方案，并且与大多数现有的修剪方法兼容。我们还将方法扩展到了最近实施的修剪任务，几乎没有数据，在该数据中，我们在修剪GPU支持的2：4稀疏模式时实现了最著名的准确性恢复。

Transfer learning is a cornerstone of computer vision, yet little work has been done to evaluate the relationship between architecture and transfer. An implicit hypothesis in modern computer vision research is that models that perform better on ImageNet necessarily perform better on other vision tasks. However, this hypothesis has never been systematically tested. Here, we compare the performance of 16 classification networks on 12 image classification datasets. We find that, when networks are used as fixed feature extractors or fine-tuned, there is a strong correlation between ImageNet accuracy and transfer accuracy (r = 0.99 and 0.96, respectively). In the former setting, we find that this relationship is very sensitive to the way in which networks are trained on ImageNet; many common forms of regularization slightly improve ImageNet accuracy but yield penultimate layer features that are much worse for transfer learning. Additionally, we find that, on two small fine-grained image classification datasets, pretraining on ImageNet provides minimal benefits, indicating the learned features from Ima-geNet do not transfer well to fine-grained tasks. Together, our results show that ImageNet architectures generalize well across datasets, but ImageNet features are less general than previously suggested.

我们研究了基于SGD的深神经网络（DNN）的优化是否可以适应高度准确且易于压缩的模型。我们提出了一种新的压缩意识的最小化器，称为CRAM，它以原则性的方式修改了SGD训练迭代，以产生在压缩操作（例如减肥或量化）下局部损失行为稳定的模型。标准图像分类任务的实验结果表明，CRAM产生的密集模型比标准SGD型基准线更准确，但在重量修剪下令人惊讶的是稳定的：例如，对于Imagenet上的Resnet50，CRAM训练的模型可能会损失到。他们的重量的70％一次性只有微小的精度损失。

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.

有效地近似损失函数的局部曲率信息是用于深神经网络的优化和压缩的关键工具。然而，大多数现有方法近似二阶信息具有高计算或存储成本，这可以限制其实用性。在这项工作中，我们调查矩阵，用于估计逆象征的矢量产品（IHVPS）的矩阵线性时间方法，因为当Hessian可以近似为乘语 - 一个矩阵的总和时，如Hessian的经典近似由经验丰富的Fisher矩阵。我们提出了两个新的算法作为称为M-FAC的框架的一部分：第一个算法朝着网络压缩量身定制，如果Hessian给出了M $等级的总和，则可以计算Dimension $ D $的IHVP。 ，使用$ O（DM ^ 2）$预压制，$ O（DM）$代价计算IHVP，并查询逆Hessian的任何单个元素的费用$ O（m）$。第二算法针对优化设置，我们希望在反向Hessian之间计算产品，估计在优化步骤的滑动窗口和给定梯度方向上，根据预先说明的SGD所需的梯度方向。我们为计算IHVP和OHVP和O（DM + M ^ 3）$ of $ o（dm + m ^ 2）$提供算法，以便从滑动窗口添加或删除任何渐变。这两种算法产生最先进的结果，用于网络修剪和相对于现有二阶方法的计算开销的优化。在[9]和[17]可用实现。

Pruning refers to the elimination of trivial weights from neural networks. The sub-networks within an overparameterized model produced after pruning are often called Lottery tickets. This research aims to generate winning lottery tickets from a set of lottery tickets that can achieve similar accuracy to the original unpruned network. We introduce a novel winning ticket called Cyclic Overlapping Lottery Ticket (COLT) by data splitting and cyclic retraining of the pruned network from scratch. We apply a cyclic pruning algorithm that keeps only the overlapping weights of different pruned models trained on different data segments. Our results demonstrate that COLT can achieve similar accuracies (obtained by the unpruned model) while maintaining high sparsities. We show that the accuracy of COLT is on par with the winning tickets of Lottery Ticket Hypothesis (LTH) and, at times, is better. Moreover, COLTs can be generated using fewer iterations than tickets generated by the popular Iterative Magnitude Pruning (IMP) method. In addition, we also notice COLTs generated on large datasets can be transferred to small ones without compromising performance, demonstrating its generalizing capability. We conduct all our experiments on Cifar-10, Cifar-100 & TinyImageNet datasets and report superior performance than the state-of-the-art methods.

Self-supervised visual representation learning has seen huge progress recently, but no large scale evaluation has compared the many models now available. We evaluate the transfer performance of 13 top self-supervised models on 40 downstream tasks, including many-shot and few-shot recognition, object detection, and dense prediction. We compare their performance to a supervised baseline and show that on most tasks the best self-supervised models outperform supervision, confirming the recently observed trend in the literature. We find ImageNet Top-1 accuracy to be highly correlated with transfer to many-shot recognition, but increasingly less so for few-shot, object detection and dense prediction. No single self-supervised method dominates overall, suggesting that universal pre-training is still unsolved. Our analysis of features suggests that top self-supervised learners fail to preserve colour information as well as supervised alternatives, but tend to induce better classifier calibration, and less attentive overfitting than supervised learners.

我们考虑在具有挑战性的训练后环境中，深度神经网络（DNN）的模型压缩问题，在该设置中，我们将获得精确的训练模型，并且必须仅基于少量校准输入数据而无需任何重新培训即可压缩它。鉴于新兴软件和硬件支持通过加速修剪和/或量化压缩的模型，并且已经针对两种压缩方法独立提出了良好的表现解决方案，因此该问题已变得流行。在本文中，我们引入了一个新的压缩框架，该框架涵盖了统一环境中的重量修剪和量化，时间和空间效率高，并且在现有的后训练方法的实际性能上大大改善。在技​​术层面上，我们的方法基于[Lecun，Denker和Solla，1990年]在现代DNN的规模上的经典最佳脑外科医生（OBS）框架的第一个精确实现，我们进一步扩展到覆盖范围。重量量化。这是通过一系列可能具有独立利益的算法开发来实现的。从实际的角度来看，我们的实验结果表明，它可以在现有后训练方法的压缩 - 准确性权衡方面显着改善，并且甚至可以在训练后进行修剪和量化的准确共同应用。

网络修剪是一种广泛使用的技术，用于有效地压缩深神经网络，几乎没有在推理期间在性能下降低。迭代幅度修剪（IMP）是由几种迭代训练和修剪步骤组成的网络修剪的最熟悉的方法之一，其中在修剪后丢失了大量网络的性能，然后在随后的再培训阶段中恢复。虽然常用为基准参考，但经常认为a）通过不将稀疏纳入训练阶段来达到次优状态，b）其全球选择标准未能正确地确定最佳层面修剪速率和c）其迭代性质使它变得缓慢和不竞争。根据最近提出的再培训技术，我们通过严格和一致的实验来调查这些索赔，我们将Impr到培训期间的训练算法进行比较，评估其选择标准的建议修改，并研究实际需要的迭代次数和总培训时间。我们发现IMP与SLR进行再培训，可以优于最先进的修剪期间，没有或仅具有很少的计算开销，即全局幅度选择标准在很大程度上具有更复杂的方法，并且只有几个刷新时期在实践中需要达到大部分稀疏性与IMP的诽谤 - 与性能权衡。我们的目标既可以证明基本的进攻已经可以提供最先进的修剪结果，甚至优于更加复杂或大量参数化方法，也可以为未来的研究建立更加现实但易于可实现的基线。

Neural network pruning-the task of reducing the size of a network by removing parameters-has been the subject of a great deal of work in recent years. We provide a meta-analysis of the literature, including an overview of approaches to pruning and consistent findings in the literature. After aggregating results across 81 papers and pruning hundreds of models in controlled conditions, our clearest finding is that the community suffers from a lack of standardized benchmarks and metrics. This deficiency is substantial enough that it is hard to compare pruning techniques to one another or determine how much progress the field has made over the past three decades. To address this situation, we identify issues with current practices, suggest concrete remedies, and introduce ShrinkBench, an open-source framework to facilitate standardized evaluations of pruning methods. We use ShrinkBench to compare various pruning techniques and show that its comprehensive evaluation can prevent common pitfalls when comparing pruning methods.

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 * .

We revisit the performance of the classic gradual magnitude pruning (GMP) baseline for large language models, focusing on the classic BERT benchmark on various popular tasks. Despite existing evidence in the literature that GMP performs poorly, we show that a simple and general variant, which we call GMP*, can match and sometimes outperform more complex state-of-the-art methods. Our results provide a simple yet strong baseline for future work, highlight the importance of parameter tuning for baselines, and even improve the performance of the state-of-the-art second-order pruning method in this setting.

This paper presents a method for adding multiple tasks to a single deep neural network while avoiding catastrophic forgetting. Inspired by network pruning techniques, we exploit redundancies in large deep networks to free up parameters that can then be employed to learn new tasks. By performing iterative pruning and network re-training, we are able to sequentially "pack" multiple tasks into a single network while ensuring minimal drop in performance and minimal storage overhead. Unlike prior work that uses proxy losses to maintain accuracy on older tasks, we always optimize for the task at hand. We perform extensive experiments on a variety of network architectures and largescale datasets, and observe much better robustness against catastrophic forgetting than prior work. In particular, we are able to add three fine-grained classification tasks to a single ImageNet-trained VGG-16 network and achieve accuracies close to those of separately trained networks for each task. Code available at https://github.com/ arunmallya/packnet

视觉变压器（VIT）已被证明可以在广泛的视觉应用中获得高度竞争性的性能，例如图像分类，对象检测和语义图像分割。与卷积神经网络相比，通常发现视觉变压器的较弱的电感偏差会在较小的培训数据集上培训时，会增加对模型正则化或数据增强的依赖（简称为“ AUGREG”）。我们进行了一项系统的实证研究，以便更好地了解培训数据，AUGREG，模型大小和计算预算之间的相互作用。作为这项研究的一个结果，我们发现增加的计算和AUGREG的组合可以产生与在数量级上训练的模型相同的训练数据的模型：我们在公共Imagenet-21K数据集中培训各种尺寸的VIT模型在较大的JFT-300M数据集上匹配或超越其对手的培训。

野外的深度学习（DL）的成功采用需要模型：（1）紧凑，（2）准确，（3）强大的分布换档。不幸的是，同时满足这些要求的努力主要是不成功的。这提出了一个重要问题：无法创建紧凑，准确，强大的深神经网络（卡）基础？为了回答这个问题，我们对流行的模型压缩技术进行了大规模分析，该技术揭示了几种有趣模式。值得注意的是，与传统的修剪方法相比（例如，微调和逐渐修剪），我们发现“彩票式风格”方法令人惊讶地用于生产卡，包括二进制牌。具体而言，我们能够创建极其紧凑的卡，与其较大的对应物相比，具有类似的测试精度和匹配（或更好）的稳健性 - 仅通过修剪和（可选）量化。利用卡的紧凑性，我们开发了一种简单的域 - 自适应测试时间合并方法（卡片 - 甲板），它使用门控模块根据与测试样本的光谱相似性动态地选择相应的卡片。该拟议的方法建立了一个“赢得胜利”的卡片，即在CiFar-10-C精度（即96.8％标准和92.75％的鲁棒）和CiFar-100- C精度（80.6％标准和71.3％的稳健性），内存使用率比非压缩基线（Https://github.com/robustbench/robustbench提供的预制卡和卡片 - 甲板）。最后，我们为我们的理论支持提供了理论支持经验研究结果。

在实现最先进的性能和在实际应用中负担得起的大型模型之间，计算机视觉的差异越来越大。在本文中，我们解决了这个问题，并显着弥合了这两种模型之间的差距。在我们的实证研究中，我们不一定要提出一种新方法，而是要努力确定一个可靠的有效食谱，以使最先进的大型模型在实践中负担得起。我们证明，当正确执行时，知识蒸馏可以成为减少大型尺寸而不损害其性能的强大工具。特别是，我们发现存在某些隐式设计选择，这可能会严重影响蒸馏的有效性。我们的关键贡献是对这些设计选择的明确识别，这些选择以前在文献中尚未阐明。我们通过一项全面的实证研究备份了我们的发现，在广泛的视觉数据集上展示了令人信服的结果，尤其是获得了最先进的Imagenet Resnet-50模型，该模型可实现82.8％的Top-1准确性。 。

修剪是稀疏深神经网络的任务，最近受到了越来越多的关注。尽管最先进的修剪方法提取了高度稀疏的模型，但它们忽略了两个主要挑战：（1）寻找这些稀疏模型的过程通常非常昂贵； （2）非结构化的修剪在GPU记忆，训练时间或碳排放方面没有提供好处。我们提出了通过梯度流量保存（早期CROP）提出的早期压缩，该压缩在训练挑战（1）的培训（1）中有效提取最先进的稀疏模型，并且可以以结构化的方式应用来应对挑战（2）。这使我们能够在商品GPU上训练稀疏的网络，该商品GPU的密集版本太大，从而节省了成本并减少了硬件要求。我们从经验上表明，早期杂交的表现优于许多任务（包括分类，回归）和域（包括计算机视觉，自然语言处理和增强学习）的丰富基线。早期杂交导致准确性与密集训练相当，同时超过修剪基线。

Transfer of pre-trained representations improves sample efficiency and simplifies hyperparameter tuning when training deep neural networks for vision. We revisit the paradigm of pre-training on large supervised datasets and fine-tuning the model on a target task. We scale up pre-training, and propose a simple recipe that we call Big Transfer (BiT). By combining a few carefully selected components, and transferring using a simple heuristic, we achieve strong performance on over 20 datasets. BiT performs well across a surprisingly wide range of data regimes -from 1 example per class to 1 M total examples. BiT achieves 87.5% top-1 accuracy on ILSVRC-2012, 99.4% on CIFAR-10, and 76.3% on the 19 task Visual Task Adaptation Benchmark (VTAB). On small datasets, BiT attains 76.8% on ILSVRC-2012 with 10 examples per class, and 97.0% on CIFAR-10 with 10 examples per class. We conduct detailed analysis of the main components that lead to high transfer performance.