%0 Conference Paper %1 Baylor:2017:TTP:3097983.3098021 %A Baylor, Denis %A Breck, Eric %A Cheng, Heng-Tze %A Fiedel, Noah %A Foo, Chuan Yu %A Haque, Zakaria %A Haykal, Salem %A Ispir, Mustafa %A Jain, Vihan %A Koc, Levent %A Koo, Chiu Yuen %A Lew, Lukasz %A Mewald, Clemens %A Modi, Akshay Naresh %A Polyzotis, Neoklis %A Ramesh, Sukriti %A Roy, Sudip %A Whang, Steven Euijong %A Wicke, Martin %A Wilkiewicz, Jarek %A Zhang, Xin %A Zinkevich, Martin %B Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining %C New York, NY, USA %D 2017 %I ACM %K machine_learning ml_operations ml_pipeline tensorflow %P 1387--1395 %R 10.1145/3097983.3098021 %T TFX: A TensorFlow-Based Production-Scale Machine Learning Platform %U http://doi.acm.org/10.1145/3097983.3098021 %X Creating and maintaining a platform for reliably producing and deploying machine learning models requires careful orchestration of many components---a learner for generating models based on training data, modules for analyzing and validating both data as well as models, and finally infrastructure for serving models in production. This becomes particularly challenging when data changes over time and fresh models need to be produced continuously. Unfortunately, such orchestration is often done ad hoc using glue code and custom scripts developed by individual teams for specific use cases, leading to duplicated effort and fragile systems with high technical debt. We present TensorFlow Extended (TFX), a TensorFlow-based general-purpose machine learning platform implemented at Google. By integrating the aforementioned components into one platform, we were able to standardize the components, simplify the platform configuration, and reduce the time to production from the order of months to weeks, while providing platform stability that minimizes disruptions. We present the case study of one deployment of TFX in the Google Play app store, where the machine learning models are refreshed continuously as new data arrive. Deploying TFX led to reduced custom code, faster experiment cycles, and a 2% increase in app installs resulting from improved data and model analysis. %@ 978-1-4503-4887-4

@inproceedings{Baylor:2017:TTP:3097983.3098021, abstract = {Creating and maintaining a platform for reliably producing and deploying machine learning models requires careful orchestration of many components---a learner for generating models based on training data, modules for analyzing and validating both data as well as models, and finally infrastructure for serving models in production. This becomes particularly challenging when data changes over time and fresh models need to be produced continuously. Unfortunately, such orchestration is often done ad hoc using glue code and custom scripts developed by individual teams for specific use cases, leading to duplicated effort and fragile systems with high technical debt. We present TensorFlow Extended (TFX), a TensorFlow-based general-purpose machine learning platform implemented at Google. By integrating the aforementioned components into one platform, we were able to standardize the components, simplify the platform configuration, and reduce the time to production from the order of months to weeks, while providing platform stability that minimizes disruptions. We present the case study of one deployment of TFX in the Google Play app store, where the machine learning models are refreshed continuously as new data arrive. Deploying TFX led to reduced custom code, faster experiment cycles, and a 2% increase in app installs resulting from improved data and model analysis.}, acmid = {3098021}, added-at = {2019-03-16T18:52:29.000+0100}, address = {New York, NY, USA}, author = {Baylor, Denis and Breck, Eric and Cheng, Heng-Tze and Fiedel, Noah and Foo, Chuan Yu and Haque, Zakaria and Haykal, Salem and Ispir, Mustafa and Jain, Vihan and Koc, Levent and Koo, Chiu Yuen and Lew, Lukasz and Mewald, Clemens and Modi, Akshay Naresh and Polyzotis, Neoklis and Ramesh, Sukriti and Roy, Sudip and Whang, Steven Euijong and Wicke, Martin and Wilkiewicz, Jarek and Zhang, Xin and Zinkevich, Martin}, biburl = {https://www.bibsonomy.org/bibtex/2576848f099cd4863d0b16279d19859e3/carand}, booktitle = {Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining}, description = {TFX}, doi = {10.1145/3097983.3098021}, interhash = {bfc3cac87f3aae1e7238fbcb1eb22349}, intrahash = {576848f099cd4863d0b16279d19859e3}, isbn = {978-1-4503-4887-4}, keywords = {machine_learning ml_operations ml_pipeline tensorflow}, location = {Halifax, NS, Canada}, numpages = {9}, pages = {1387--1395}, publisher = {ACM}, series = {KDD '17}, timestamp = {2019-03-16T18:52:29.000+0100}, title = {TFX: A TensorFlow-Based Production-Scale Machine Learning Platform}, url = {http://doi.acm.org/10.1145/3097983.3098021}, year = 2017 }