We present RecD (Recommendation Deduplication), a suite of end-to-end infrastructure optimizations across the Deep Learning Recommendation Model (DLRM) training pipeline. RecD addresses immense storage, preprocessing, and training overheads caused by feature duplication inherent in industry-scale DLRM training datasets. Feature duplication arises because DLRM datasets are generated from interactions. While each user session can generate multiple training samples, many features' values do not change across these samples. We demonstrate how RecD exploits this property, end-to-end, across a deployed training pipeline. RecD optimizes data generation pipelines to decrease dataset storage and preprocessing resource demands and to maximize duplication within a training batch. RecD introduces a new tensor format, InverseKeyedJaggedTensors (IKJTs), to deduplicate feature values in each batch. We show how DLRM model architectures can leverage IKJTs to drastically increase training throughput. RecD improves the training and preprocessing throughput and storage efficiency by up to 2.49x, 1.79x, and 3.71x, respectively, in an industry-scale DLRM training system.

学习在线推荐模型的关键挑战之一是时间域移动，这会导致培训与测试数据分布之间的不匹配以及域的概括错误。为了克服，我们建议学习一个未来的梯度生成器，该生成器可以预测培训未来数据分配的梯度信息，以便可以对建议模型进行培训，就像我们能够展望其部署的未来一样。与批处理更新相比，我们的理论表明，所提出的算法达到了较小的时间域概括误差，该误差通过梯度变异项在局部遗憾中衡量。我们通过与各种代表性基线进行比较来证明经验优势。

嵌入学习是深度建议模型中的重要技术，可以将分类特征映射到密集的矢量。但是，嵌入表通常需要大量参数，这些参数成为存储和效率瓶颈。已经采用了分布式培训解决方案将嵌入表分配到多个设备中。但是，如果不仔细分区，则嵌入表很容易导致失衡。这是名为“嵌入桌碎片”的分布式系统的重大设计挑战，即，我们应该如何对嵌入表进行分配以平衡跨设备的成本，这是一项非平凡的任务，因为1）很难有效，精确地衡量成本，和2）已知分区问题是NP-HARD。在这项工作中，我们在Meta中介绍了新颖的实践，即Autoshard，该实践使用神经成本模型直接预测多桌成本和利用深度强化学习以解决分区问题。开源的大规模合成数据集和Meta生产数据集的实验结果证明了Autoshard的优越性优于启发式方法。此外，Autoshard的学习政策可以转移到具有不同数量的表和不同表格比率的碎片任务中，而无需进行任何微调。此外，Autoshard可以在几秒钟内有效地将数百张桌子碎片。 Autoshard的有效性，可转移性和效率使其适合生产使用。我们的算法已在元生产环境中部署。可以在https://github.com/daochenzha/autoshard上获得原型

嵌入学习在其他域中发现了推荐系统和自然语言建模中的广泛应用。为了有效地学习质量嵌入，自适应学习率算法已经证明了SGD的卓越经验性能，主要是对其令牌依赖学习率的认可。然而，令牌依赖学习率效率的潜在机制仍然是缺乏缺陷的。我们表明，在嵌入学习问题中结合令牌的频率信息导致可提供的可提供有效的算法，并且证明普通的自适应算法在很大程度上隐含地利用频率信息。具体地，我们提出（基于计数器的）频率感知随机梯度下降，其为每个令牌应用频率相关的学习率，并且当令牌分布不平衡时，与SGD相比表现出可提供的速度。凭经验，我们显示所提出的算法能够改进或匹配基准推荐任务和大型工业推荐系统的自适应算法，关闭SGD和自适应算法之间的性能差距。我们的结果是第一个显示令牌依赖学习率，可否改善非凸嵌入学习问题的收敛。

Graph Neural Networks (GNNs) have become increasingly important in recent years due to their state-of-the-art performance on many important downstream applications. Existing GNNs have mostly focused on learning a single node representation, despite that a node often exhibits polysemous behavior in different contexts. In this work, we develop a persona-based graph neural network framework called PersonaSAGE that learns multiple persona-based embeddings for each node in the graph. Such disentangled representations are more interpretable and useful than a single embedding. Furthermore, PersonaSAGE learns the appropriate set of persona embeddings for each node in the graph, and every node can have a different number of assigned persona embeddings. The framework is flexible enough and the general design helps in the wide applicability of the learned embeddings to suit the domain. We utilize publicly available benchmark datasets to evaluate our approach and against a variety of baselines. The experiments demonstrate the effectiveness of PersonaSAGE for a variety of important tasks including link prediction where we achieve an average gain of 15% while remaining competitive for node classification. Finally, we also demonstrate the utility of PersonaSAGE with a case study for personalized recommendation of different entity types in a data management platform.

Reinforcement learning can enable robots to navigate to distant goals while optimizing user-specified reward functions, including preferences for following lanes, staying on paved paths, or avoiding freshly mowed grass. However, online learning from trial-and-error for real-world robots is logistically challenging, and methods that instead can utilize existing datasets of robotic navigation data could be significantly more scalable and enable broader generalization. In this paper, we present ReViND, the first offline RL system for robotic navigation that can leverage previously collected data to optimize user-specified reward functions in the real-world. We evaluate our system for off-road navigation without any additional data collection or fine-tuning, and show that it can navigate to distant goals using only offline training from this dataset, and exhibit behaviors that qualitatively differ based on the user-specified reward function.

We propose an approach for semantic imitation, which uses demonstrations from a source domain, e.g. human videos, to accelerate reinforcement learning (RL) in a different target domain, e.g. a robotic manipulator in a simulated kitchen. Instead of imitating low-level actions like joint velocities, our approach imitates the sequence of demonstrated semantic skills like "opening the microwave" or "turning on the stove". This allows us to transfer demonstrations across environments (e.g. real-world to simulated kitchen) and agent embodiments (e.g. bimanual human demonstration to robotic arm). We evaluate on three challenging cross-domain learning problems and match the performance of demonstration-accelerated RL approaches that require in-domain demonstrations. In a simulated kitchen environment, our approach learns long-horizon robot manipulation tasks, using less than 3 minutes of human video demonstrations from a real-world kitchen. This enables scaling robot learning via the reuse of demonstrations, e.g. collected as human videos, for learning in any number of target domains.

Navigation is one of the most heavily studied problems in robotics, and is conventionally approached as a geometric mapping and planning problem. However, real-world navigation presents a complex set of physical challenges that defies simple geometric abstractions. Machine learning offers a promising way to go beyond geometry and conventional planning, allowing for navigational systems that make decisions based on actual prior experience. Such systems can reason about traversability in ways that go beyond geometry, accounting for the physical outcomes of their actions and exploiting patterns in real-world environments. They can also improve as more data is collected, potentially providing a powerful network effect. In this article, we present a general toolkit for experiential learning of robotic navigation skills that unifies several recent approaches, describe the underlying design principles, summarize experimental results from several of our recent papers, and discuss open problems and directions for future work.

Iterative text revision improves text quality by fixing grammatical errors, rephrasing for better readability or contextual appropriateness, or reorganizing sentence structures throughout a document. Most recent research has focused on understanding and classifying different types of edits in the iterative revision process from human-written text instead of building accurate and robust systems for iterative text revision. In this work, we aim to build an end-to-end text revision system that can iteratively generate helpful edits by explicitly detecting editable spans (where-to-edit) with their corresponding edit intents and then instructing a revision model to revise the detected edit spans. Leveraging datasets from other related text editing NLP tasks, combined with the specification of editable spans, leads our system to more accurately model the process of iterative text refinement, as evidenced by empirical results and human evaluations. Our system significantly outperforms previous baselines on our text revision tasks and other standard text revision tasks, including grammatical error correction, text simplification, sentence fusion, and style transfer. Through extensive qualitative and quantitative analysis, we make vital connections between edit intentions and writing quality, and better computational modeling of iterative text revisions.

Semantic navigation is necessary to deploy mobile robots in uncontrolled environments like our homes, schools, and hospitals. Many learning-based approaches have been proposed in response to the lack of semantic understanding of the classical pipeline for spatial navigation, which builds a geometric map using depth sensors and plans to reach point goals. Broadly, end-to-end learning approaches reactively map sensor inputs to actions with deep neural networks, while modular learning approaches enrich the classical pipeline with learning-based semantic sensing and exploration. But learned visual navigation policies have predominantly been evaluated in simulation. How well do different classes of methods work on a robot? We present a large-scale empirical study of semantic visual navigation methods comparing representative methods from classical, modular, and end-to-end learning approaches across six homes with no prior experience, maps, or instrumentation. We find that modular learning works well in the real world, attaining a 90% success rate. In contrast, end-to-end learning does not, dropping from 77% simulation to 23% real-world success rate due to a large image domain gap between simulation and reality. For practitioners, we show that modular learning is a reliable approach to navigate to objects: modularity and abstraction in policy design enable Sim-to-Real transfer. For researchers, we identify two key issues that prevent today's simulators from being reliable evaluation benchmarks - (A) a large Sim-to-Real gap in images and (B) a disconnect between simulation and real-world error modes - and propose concrete steps forward.