Deep learning frameworks have often focused on either usability or speed, but
not both. PyTorch is a machine learning library that shows that these two goals
are in fact compatible: it provides an imperative and Pythonic programming
style that supports code as a model, makes debugging easy and is consistent
with other popular scientific computing libraries, while remaining efficient
and supporting hardware accelerators such as GPUs.
In this paper, we detail the principles that drove the implementation of
PyTorch and how they are reflected in its architecture. We emphasize that every
aspect of PyTorch is a regular Python program under the full control of its
user. We also explain how the careful and pragmatic implementation of the key
components of its runtime enables them to work together to achieve compelling
performance.
We demonstrate the efficiency of individual subsystems, as well as the
overall speed of PyTorch on several common benchmarks.
Description
[1912.01703] PyTorch: An Imperative Style, High-Performance Deep Learning Library
%0 Generic
%1 paszke2019pytorch
%A Paszke, Adam
%A Gross, Sam
%A Massa, Francisco
%A Lerer, Adam
%A Bradbury, James
%A Chanan, Gregory
%A Killeen, Trevor
%A Lin, Zeming
%A Gimelshein, Natalia
%A Antiga, Luca
%A Desmaison, Alban
%A Köpf, Andreas
%A Yang, Edward
%A DeVito, Zach
%A Raison, Martin
%A Tejani, Alykhan
%A Chilamkurthy, Sasank
%A Steiner, Benoit
%A Fang, Lu
%A Bai, Junjie
%A Chintala, Soumith
%D 2019
%K 2019 deep-learning facebook library pytorch
%T PyTorch: An Imperative Style, High-Performance Deep Learning Library
%U http://arxiv.org/abs/1912.01703
%X Deep learning frameworks have often focused on either usability or speed, but
not both. PyTorch is a machine learning library that shows that these two goals
are in fact compatible: it provides an imperative and Pythonic programming
style that supports code as a model, makes debugging easy and is consistent
with other popular scientific computing libraries, while remaining efficient
and supporting hardware accelerators such as GPUs.
In this paper, we detail the principles that drove the implementation of
PyTorch and how they are reflected in its architecture. We emphasize that every
aspect of PyTorch is a regular Python program under the full control of its
user. We also explain how the careful and pragmatic implementation of the key
components of its runtime enables them to work together to achieve compelling
performance.
We demonstrate the efficiency of individual subsystems, as well as the
overall speed of PyTorch on several common benchmarks.
@misc{paszke2019pytorch,
abstract = {Deep learning frameworks have often focused on either usability or speed, but
not both. PyTorch is a machine learning library that shows that these two goals
are in fact compatible: it provides an imperative and Pythonic programming
style that supports code as a model, makes debugging easy and is consistent
with other popular scientific computing libraries, while remaining efficient
and supporting hardware accelerators such as GPUs.
In this paper, we detail the principles that drove the implementation of
PyTorch and how they are reflected in its architecture. We emphasize that every
aspect of PyTorch is a regular Python program under the full control of its
user. We also explain how the careful and pragmatic implementation of the key
components of its runtime enables them to work together to achieve compelling
performance.
We demonstrate the efficiency of individual subsystems, as well as the
overall speed of PyTorch on several common benchmarks.},
added-at = {2020-01-10T22:19:45.000+0100},
author = {Paszke, Adam and Gross, Sam and Massa, Francisco and Lerer, Adam and Bradbury, James and Chanan, Gregory and Killeen, Trevor and Lin, Zeming and Gimelshein, Natalia and Antiga, Luca and Desmaison, Alban and Köpf, Andreas and Yang, Edward and DeVito, Zach and Raison, Martin and Tejani, Alykhan and Chilamkurthy, Sasank and Steiner, Benoit and Fang, Lu and Bai, Junjie and Chintala, Soumith},
biburl = {https://www.bibsonomy.org/bibtex/2018e9538aaad774ee16f704b858a78dd/analyst},
description = {[1912.01703] PyTorch: An Imperative Style, High-Performance Deep Learning Library},
interhash = {a3caca5456f48ad236ed06df5cf0ecca},
intrahash = {018e9538aaad774ee16f704b858a78dd},
keywords = {2019 deep-learning facebook library pytorch},
note = {cite arxiv:1912.01703Comment: 12 pages, 3 figures, NeurIPS 2019},
timestamp = {2021-04-21T16:03:01.000+0200},
title = {PyTorch: An Imperative Style, High-Performance Deep Learning Library},
url = {http://arxiv.org/abs/1912.01703},
year = 2019
}