cff-version: 1.2.0
message: "If you use TensorFlow in your research, please cite it using these metadata. Software is available from tensorflow.org."
title: TensorFlow, Large-scale machine learning on heterogeneous systems
abstract: TensorFlow is a machine learning system that operates at large scale and in heterogeneous environments. TensorFlow uses dataflow graphs to represent computation, shared state, and the operations that mutate that state. It maps the nodes of a dataflow graph across many machines in a cluster, and within a machine across multiple computational devices, including multicore CPUs, general purpose GPUs, and custom-designed ASICs known as Tensor Processing Units (TPUs). This architecture gives flexibility to the application developer, whereas in previous “parameter server” designs the management of shared state is built into the system, TensorFlow enables developers to experiment with novel optimizations and training algorithms. TensorFlow supports a variety of applications, with a focus on training and inference on deep neural networks. Several Google services use TensorFlow in production, we have released it as an open-source project, and it has become widely used for machine learning research. In this paper, we describe the TensorFlow dataflow model and demonstrate the compelling performance that TensorFlow achieves for several real-world applications.
authors:
  - family-names: Abadi
    given-names: Martín
  - family-names: Agarwal
    given-names: Ashish
  - family-names: Barham
    given-names: Paul
  - family-names: Brevdo
    given-names: Eugene
  - family-names: Chen
    given-names: Zhifeng
  - family-names: Citro
    given-names: Craig
  - family-names: Corrado
    given-names: Greg S.
  - family-names: Davis
    given-names: Andy
  - family-names: Dean
    given-names: Jeffrey
  - family-names: Devin
    given-names: Matthieu
  - family-names: Ghemawat
    given-names: Sanjay
  - family-names: Goodfellow
    given-names: Ian
  - family-names: Harp
    given-names: Andrew
  - family-names: Irving
    given-names: Geoffrey
  - family-names: Isard
    given-names: Michael
  - family-names: Jozefowicz
    given-names: Rafal
  - family-names: Jia
    given-names: Yangqing
  - family-names: Kaiser
    given-names: Lukasz
  - family-names: Kudlur
    given-names: Manjunath
  - family-names: Levenberg
    given-names: Josh
  - family-names: Mané
    given-names: Dan
  - family-names: Schuster
    given-names: Mike
  - family-names: Monga
    given-names: Rajat
  - family-names: Moore
    given-names: Sherry
  - family-names: Murray
    given-names: Derek
  - family-names: Olah
    given-names: Chris
  - family-names: Shlens
    given-names: Jonathon
  - family-names: Steiner
    given-names: Benoit
  - family-names: Sutskever
    given-names: Ilya
  - family-names: Talwar
    given-names: Kunal
  - family-names: Tucker
    given-names: Paul
  - family-names: Vanhoucke
    given-names: Vincent
  - family-names: Vasudevan
    given-names: Vijay
  - family-names: Viégas
    given-names: Fernanda
  - family-names: Vinyals
    given-names: Oriol
  - family-names: Warden
    given-names: Pete
  - family-names: Wattenberg
    given-names: Martin
  - family-names: Wicke
    given-names: Martin
  - family-names: Yu
    given-names: Yuan
  - family-names: Zheng
    given-names: Xiaoqiang
identifiers:
  - type: doi
    value: 10.5281/zenodo.4724125
    description: The concept DOI for the collection containing all versions of the Citation File Format.
date-released: "2015-11-09"
license: "Apache-2.0"
doi: 10.5281/zenodo.4724125