diff --git a/chapter_recommender_system/system_architecture.md b/chapter_recommender_system/system_architecture.md index 0f69acd..c476fc8 100644 --- a/chapter_recommender_system/system_architecture.md +++ b/chapter_recommender_system/system_architecture.md @@ -11,7 +11,7 @@ Merlin是英伟达公司开发的一个开源推荐系统框架，帮助使用 3. 推理服务：类似地，推理服务器在接到一批用户的推荐请求后，从参数服务器拉去相应的嵌入项和稠密神经网络参数来响应用户的请求。推荐系统的推理服务对延迟十分敏感，例如脸书公司的DLRM :cite:`naumov2019deep`基准在MLPerf评测中的服务器延迟限定在30ms[^1]。因此如何在限定延迟（latency-bounded）的情况下尽可能提升吞吐（throughput）是推理服务面临的关键问题。在GPU推理场景下，常见的优化手段有：请求动态合批处理、核融合、低精度部署等 :cite:`10.1145/3437801.3441578,wang-etal-2021-lightseq`. Triton提供了请求调度的功能并且支持多种不同的机器学习框架作为后端。 -在工业界，为了提升系统在发生故障的情况下的可用性，以上介绍的各个组件在实际中部署中都应该具备基本的容灾和故障恢复能力。以推理服务为例，在线服务中的深度学习推荐模型通常都采用多副本分布式部署。同一个模型的多个副本通常会被部署在至少两个不同的地理区域内的多个数据中心中，如 :numref:`ch10-recommendation-systems`，以应对大面积停电或者网络中断而导致整个地区的所有副本都不可用。除了容错方面的考虑，部署多个副本还有其他几点优势。首先，将模型部署在靠近用户的云服务器上可以提升响应速度。其次，部署多份副本也可以拓展模型推理服务的吞吐率。 +在工业界，为了提升系统在发生故障的情况下的可用性，以上介绍的各个组件在实际中部署中都应该具备基本的容灾和故障恢复能力。以推理服务为例，在线服务中的深度学习推荐模型通常都采用多副本分布式部署。同一个模型的多个副本通常会被部署在至少两个不同的地理区域内的多个数据中心中，如图 :numref:`ch10-recommendation-systems`所示，以应对大面积停电或者网络中断而导致整个地区的所有副本都不可用。除了容错方面的考虑，部署多个副本还有其他几点优势。首先，将模型部署在靠近用户的云服务器上可以提升响应速度。其次，部署多份副本也可以拓展模型推理服务的吞吐率。 ![推荐系统的分布式架构](../img/ch10/ch10-recommendation-systems.png) :width:`800px` diff --git a/references/recommender.bib b/references/recommender.bib index b4c0286..3b5a3e8 100644 --- a/references/recommender.bib +++ b/references/recommender.bib @@ -1,31 +1,31 @@ @inproceedings{10.1145/2988450.2988454, author = {Cheng, Heng-Tze and Koc, Levent and Harmsen, Jeremiah and Shaked, Tal and Chandra, Tushar and Aradhye, Hrishi and Anderson, Glen and Corrado, Greg and Chai, Wei and Ispir, Mustafa and Anil, Rohan and Haque, Zakaria and Hong, Lichan and Jain, Vihan and Liu, Xiaobing and Shah, Hemal}, -title = {Wide & Deep Learning for Recommender Systems}, +title = {Wide \& Deep Learning for Recommender Systems}, year = {2016}, isbn = {9781450347952}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/2988450.2988454}, doi = {10.1145/2988450.2988454}, -abstract = {Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs. Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort. With less feature engineering, deep neural networks can generalize better to unseen feature combinations through low-dimensional dense embeddings learned for the sparse features. However, deep neural networks with embeddings can over-generalize and recommend less relevant items when the user-item interactions are sparse and high-rank. In this paper, we present Wide & Deep learning---jointly trained wide linear models and deep neural networks---to combine the benefits of memorization and generalization for recommender systems. We productionized and evaluated the system on Google Play, a commercial mobile app store with over one billion active users and over one million apps. Online experiment results show that Wide & Deep significantly increased app acquisitions compared with wide-only and deep-only models. We have also open-sourced our implementation in TensorFlow.}, +abstract = {Generalized linear models with nonlinear feature transformations are widely used for large-scale regression and classification problems with sparse inputs. Memorization of feature interactions through a wide set of cross-product feature transformations are effective and interpretable, while generalization requires more feature engineering effort. With less feature engineering, deep neural networks can generalize better to unseen feature combinations through low-dimensional dense embeddings learned for the sparse features. However, deep neural networks with embeddings can over-generalize and recommend less relevant items when the user-item interactions are sparse and high-rank. In this paper, we present Wide \& Deep learning---jointly trained wide linear models and deep neural networks---to combine the benefits of memorization and generalization for recommender systems. We productionized and evaluated the system on Google Play, a commercial mobile app store with over one billion active users and over one million apps. Online experiment results show that Wide \& Deep significantly increased app acquisitions compared with wide-only and deep-only models. We have also open-sourced our implementation in TensorFlow.}, booktitle = {Proceedings of the 1st Workshop on Deep Learning for Recommender Systems}, pages = {7-10}, numpages = {4}, -keywords = {Recommender Systems, Wide & Deep Learning}, +keywords = {Recommender Systems, Wide \& Deep Learning}, location = {Boston, MA, USA}, series = {DLRS 2016} } @inproceedings{10.1145/3124749.3124754, author = {Wang, Ruoxi and Fu, Bin and Fu, Gang and Wang, Mingliang}, -title = {Deep & Cross Network for Ad Click Predictions}, +title = {Deep \& Cross Network for Ad Click Predictions}, year = {2017}, isbn = {9781450351942}, publisher = {Association for Computing Machinery}, address = {New York, NY, USA}, url = {https://doi.org/10.1145/3124749.3124754}, doi = {10.1145/3124749.3124754}, -abstract = {Feature engineering has been the key to the success of many prediction models. However, the process is nontrivial and often requires manual feature engineering or exhaustive searching. DNNs are able to automatically learn feature interactions; however, they generate all the interactions implicitly, and are not necessarily efficient in learning all types of cross features. In this paper, we propose the Deep & Cross Network (DCN) which keeps the benefits of a DNN model, and beyond that, it introduces a novel cross network that is more efficient in learning certain bounded-degree feature interactions. In particular, DCN explicitly applies feature crossing at each layer, requires no manual feature engineering, and adds negligible extra complexity to the DNN model. Our experimental results have demonstrated its superiority over the state-of-art algorithms on the CTR prediction dataset and dense classification dataset, in terms of both model accuracy and memory usage.}, +abstract = {Feature engineering has been the key to the success of many prediction models. However, the process is nontrivial and often requires manual feature engineering or exhaustive searching. DNNs are able to automatically learn feature interactions; however, they generate all the interactions implicitly, and are not necessarily efficient in learning all types of cross features. In this paper, we propose the Deep \& Cross Network (DCN) which keeps the benefits of a DNN model, and beyond that, it introduces a novel cross network that is more efficient in learning certain bounded-degree feature interactions. In particular, DCN explicitly applies feature crossing at each layer, requires no manual feature engineering, and adds negligible extra complexity to the DNN model. Our experimental results have demonstrated its superiority over the state-of-art algorithms on the CTR prediction dataset and dense classification dataset, in terms of both model accuracy and memory usage.}, booktitle = {Proceedings of the ADKDD'17}, articleno = {12}, numpages = {7}, @@ -266,7 +266,7 @@ address = {New York, NY, USA}, url = {https://doi.org/10.1145/3394486.3403059}, doi = {10.1145/3394486.3403059}, abstract = {}, -booktitle = {Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining}, +booktitle = {Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining}, pages = {165–175}, numpages = {11}, keywords = {model compression, recommendation systems, embeddings},