%0 Generic %1 shen2019lingvo %A Shen, Jonathan %A Nguyen, Patrick %A Wu, Yonghui %A Chen, Zhifeng %A Chen, Mia X. %A Jia, Ye %A Kannan, Anjuli %A Sainath, Tara %A Cao, Yuan %A Chiu, Chung-Cheng %A He, Yanzhang %A Chorowski, Jan %A Hinsu, Smit %A Laurenzo, Stella %A Qin, James %A Firat, Orhan %A Macherey, Wolfgang %A Gupta, Suyog %A Bapna, Ankur %A Zhang, Shuyuan %A Pang, Ruoming %A Weiss, Ron J. %A Prabhavalkar, Rohit %A Liang, Qiao %A Jacob, Benoit %A Liang, Bowen %A Lee, HyoukJoong %A Chelba, Ciprian %A Jean, Sébastien %A Li, Bo %A Johnson, Melvin %A Anil, Rohan %A Tibrewal, Rajat %A Liu, Xiaobing %A Eriguchi, Akiko %A Jaitly, Navdeep %A Ari, Naveen %A Cherry, Colin %A Haghani, Parisa %A Good, Otavio %A Cheng, Youlong %A Alvarez, Raziel %A Caswell, Isaac %A Hsu, Wei-Ning %A Yang, Zongheng %A Wang, Kuan-Chieh %A Gonina, Ekaterina %A Tomanek, Katrin %A Vanik, Ben %A Wu, Zelin %A Jones, Llion %A Schuster, Mike %A Huang, Yanping %A Chen, Dehao %A Irie, Kazuki %A Foster, George %A Richardson, John %A Macherey, Klaus %A Bruguier, Antoine %A Zen, Heiga %A Raffel, Colin %A Kumar, Shankar %A Rao, Kanishka %A Rybach, David %A Murray, Matthew %A Peddinti, Vijayaditya %A Krikun, Maxim %A Bacchiani, Michiel A. U. %A Jablin, Thomas B. %A Suderman, Rob %A Williams, Ian %A Lee, Benjamin %A Bhatia, Deepti %A Carlson, Justin %A Yavuz, Semih %A Zhang, Yu %A McGraw, Ian %A Galkin, Max %A Ge, Qi %A Pundak, Golan %A Whipkey, Chad %A Wang, Todd %A Alon, Uri %A Lepikhin, Dmitry %A Tian, Ye %A Sabour, Sara %A Chan, William %A Toshniwal, Shubham %A Liao, Baohua %A Nirschl, Michael %A Rondon, Pat %D 2019 %K speech %T Lingvo: a Modular and Scalable Framework for Sequence-to-Sequence Modeling %U http://arxiv.org/abs/1902.08295 %X Lingvo is a Tensorflow framework offering a complete solution for collaborative deep learning research, with a particular focus towards sequence-to-sequence models. Lingvo models are composed of modular building blocks that are flexible and easily extensible, and experiment configurations are centralized and highly customizable. Distributed training and quantized inference are supported directly within the framework, and it contains existing implementations of a large number of utilities, helper functions, and the newest research ideas. Lingvo has been used in collaboration by dozens of researchers in more than 20 papers over the last two years. This document outlines the underlying design of Lingvo and serves as an introduction to the various pieces of the framework, while also offering examples of advanced features that showcase the capabilities of the framework.

@misc{shen2019lingvo, abstract = {Lingvo is a Tensorflow framework offering a complete solution for collaborative deep learning research, with a particular focus towards sequence-to-sequence models. Lingvo models are composed of modular building blocks that are flexible and easily extensible, and experiment configurations are centralized and highly customizable. Distributed training and quantized inference are supported directly within the framework, and it contains existing implementations of a large number of utilities, helper functions, and the newest research ideas. Lingvo has been used in collaboration by dozens of researchers in more than 20 papers over the last two years. This document outlines the underlying design of Lingvo and serves as an introduction to the various pieces of the framework, while also offering examples of advanced features that showcase the capabilities of the framework.}, added-at = {2019-07-09T15:45:02.000+0200}, author = {Shen, Jonathan and Nguyen, Patrick and Wu, Yonghui and Chen, Zhifeng and Chen, Mia X. and Jia, Ye and Kannan, Anjuli and Sainath, Tara and Cao, Yuan and Chiu, Chung-Cheng and He, Yanzhang and Chorowski, Jan and Hinsu, Smit and Laurenzo, Stella and Qin, James and Firat, Orhan and Macherey, Wolfgang and Gupta, Suyog and Bapna, Ankur and Zhang, Shuyuan and Pang, Ruoming and Weiss, Ron J. and Prabhavalkar, Rohit and Liang, Qiao and Jacob, Benoit and Liang, Bowen and Lee, HyoukJoong and Chelba, Ciprian and Jean, Sébastien and Li, Bo and Johnson, Melvin and Anil, Rohan and Tibrewal, Rajat and Liu, Xiaobing and Eriguchi, Akiko and Jaitly, Navdeep and Ari, Naveen and Cherry, Colin and Haghani, Parisa and Good, Otavio and Cheng, Youlong and Alvarez, Raziel and Caswell, Isaac and Hsu, Wei-Ning and Yang, Zongheng and Wang, Kuan-Chieh and Gonina, Ekaterina and Tomanek, Katrin and Vanik, Ben and Wu, Zelin and Jones, Llion and Schuster, Mike and Huang, Yanping and Chen, Dehao and Irie, Kazuki and Foster, George and Richardson, John and Macherey, Klaus and Bruguier, Antoine and Zen, Heiga and Raffel, Colin and Kumar, Shankar and Rao, Kanishka and Rybach, David and Murray, Matthew and Peddinti, Vijayaditya and Krikun, Maxim and Bacchiani, Michiel A. U. and Jablin, Thomas B. and Suderman, Rob and Williams, Ian and Lee, Benjamin and Bhatia, Deepti and Carlson, Justin and Yavuz, Semih and Zhang, Yu and McGraw, Ian and Galkin, Max and Ge, Qi and Pundak, Golan and Whipkey, Chad and Wang, Todd and Alon, Uri and Lepikhin, Dmitry and Tian, Ye and Sabour, Sara and Chan, William and Toshniwal, Shubham and Liao, Baohua and Nirschl, Michael and Rondon, Pat}, biburl = {https://www.bibsonomy.org/bibtex/246149fb59e13ea0867aba2b86fb51986/topel}, interhash = {60b3c3b19b1a211ebca9607feea3db7d}, intrahash = {46149fb59e13ea0867aba2b86fb51986}, keywords = {speech}, note = {cite arxiv:1902.08295}, timestamp = {2019-07-09T15:45:02.000+0200}, title = {Lingvo: a Modular and Scalable Framework for Sequence-to-Sequence Modeling}, url = {http://arxiv.org/abs/1902.08295}, year = 2019 }