Deep learning has been popularized by its recent successes on challenging
artificial intelligence problems. One of the reasons for its dominance is also
an ongoing challenge: the need for immense amounts of computational power.
Hardware architects have responded by proposing a wide array of promising
ideas, but to date, the majority of the work has focused on specific algorithms
in somewhat narrow application domains. While their specificity does not
diminish these approaches, there is a clear need for more flexible solutions.
We believe the first step is to examine the characteristics of cutting edge
models from across the deep learning community.
Consequently, we have assembled Fathom: a collection of eight archetypal deep
learning workloads for study. Each of these models comes from a seminal work in
the deep learning community, ranging from the familiar deep convolutional
neural network of Krizhevsky et al., to the more exotic memory networks from
Facebook's AI research group. Fathom has been released online, and this paper
focuses on understanding the fundamental performance characteristics of each
model. We use a set of application-level modeling tools built around the
TensorFlow deep learning framework in order to analyze the behavior of the
Fathom workloads. We present a breakdown of where time is spent, the
similarities between the performance profiles of our models, an analysis of
behavior in inference and training, and the effects of parallelism on scaling.
%0 Generic
%1 adolf2016fathom
%A Adolf, Robert
%A Rama, Saketh
%A Reagen, Brandon
%A Wei, Gu-Yeon
%A Brooks, David
%D 2016
%K benchmark speedup
%R 10.1109/IISWC.2016.7581275
%T Fathom: Reference Workloads for Modern Deep Learning Methods
%U http://arxiv.org/abs/1608.06581
%X Deep learning has been popularized by its recent successes on challenging
artificial intelligence problems. One of the reasons for its dominance is also
an ongoing challenge: the need for immense amounts of computational power.
Hardware architects have responded by proposing a wide array of promising
ideas, but to date, the majority of the work has focused on specific algorithms
in somewhat narrow application domains. While their specificity does not
diminish these approaches, there is a clear need for more flexible solutions.
We believe the first step is to examine the characteristics of cutting edge
models from across the deep learning community.
Consequently, we have assembled Fathom: a collection of eight archetypal deep
learning workloads for study. Each of these models comes from a seminal work in
the deep learning community, ranging from the familiar deep convolutional
neural network of Krizhevsky et al., to the more exotic memory networks from
Facebook's AI research group. Fathom has been released online, and this paper
focuses on understanding the fundamental performance characteristics of each
model. We use a set of application-level modeling tools built around the
TensorFlow deep learning framework in order to analyze the behavior of the
Fathom workloads. We present a breakdown of where time is spent, the
similarities between the performance profiles of our models, an analysis of
behavior in inference and training, and the effects of parallelism on scaling.
@misc{adolf2016fathom,
abstract = {Deep learning has been popularized by its recent successes on challenging
artificial intelligence problems. One of the reasons for its dominance is also
an ongoing challenge: the need for immense amounts of computational power.
Hardware architects have responded by proposing a wide array of promising
ideas, but to date, the majority of the work has focused on specific algorithms
in somewhat narrow application domains. While their specificity does not
diminish these approaches, there is a clear need for more flexible solutions.
We believe the first step is to examine the characteristics of cutting edge
models from across the deep learning community.
Consequently, we have assembled Fathom: a collection of eight archetypal deep
learning workloads for study. Each of these models comes from a seminal work in
the deep learning community, ranging from the familiar deep convolutional
neural network of Krizhevsky et al., to the more exotic memory networks from
Facebook's AI research group. Fathom has been released online, and this paper
focuses on understanding the fundamental performance characteristics of each
model. We use a set of application-level modeling tools built around the
TensorFlow deep learning framework in order to analyze the behavior of the
Fathom workloads. We present a breakdown of where time is spent, the
similarities between the performance profiles of our models, an analysis of
behavior in inference and training, and the effects of parallelism on scaling.},
added-at = {2018-02-10T15:26:52.000+0100},
author = {Adolf, Robert and Rama, Saketh and Reagen, Brandon and Wei, Gu-Yeon and Brooks, David},
biburl = {https://www.bibsonomy.org/bibtex/231667a0495568840fc1b561f57478e68/jk_itwm},
description = {1608.06581.pdf},
doi = {10.1109/IISWC.2016.7581275},
interhash = {8479d0d7bc07419c8ade0622e755366e},
intrahash = {31667a0495568840fc1b561f57478e68},
keywords = {benchmark speedup},
note = {cite arxiv:1608.06581Comment: Proceedings of the IEEE International Symposium on Workload Characterization, 2016},
timestamp = {2018-02-10T15:26:52.000+0100},
title = {Fathom: Reference Workloads for Modern Deep Learning Methods},
url = {http://arxiv.org/abs/1608.06581},
year = 2016
}