https://arxiv.org/abs/2101.03961

close this message

Donate to arXiv

Please join the Simons Foundation and our generous member
organizations in supporting arXiv during our giving campaign
September 23-27. 100% of your contribution will fund improvements and
new initiatives to benefit arXiv's global scientific community.

DONATE

[secure site, no need to create account]

Skip to main content
Cornell University
We gratefully acknowledge support from
the Simons Foundation and member institutions.
 
arXiv.org > cs > arXiv:2101.03961
[                    ]

Help | Advanced Search

[All fields        ]
Search
arXiv
Cornell University Logo
[                    ] GO
quick links

  * Login
  * Help Pages
  * About

Computer Science > Machine Learning

arXiv:2101.03961 (cs)
[Submitted on 11 Jan 2021]

Title:Switch Transformers: Scaling to Trillion Parameter Models with
Simple and Efficient Sparsity

Authors:William Fedus, Barret Zoph, Noam Shazeer
Download PDF

    Abstract: In deep learning, models typically reuse the same
    parameters for all inputs. Mixture of Experts (MoE) defies this
    and instead selects different parameters for each incoming
    example. The result is a sparsely-activated model -- with
    outrageous numbers of parameters -- but a constant computational
    cost. However, despite several notable successes of MoE,
    widespread adoption has been hindered by complexity,
    communication costs and training instability -- we address these
    with the Switch Transformer. We simplify the MoE routing
    algorithm and design intuitive improved models with reduced
    communication and computational costs. Our proposed training
    techniques help wrangle the instabilities and we show large
    sparse models may be trained, for the first time, with lower
    precision (bfloat16) formats. We design models based off T5-Base
    and T5-Large to obtain up to 7x increases in pre-training speed
    with the same computational resources. These improvements extend
    into multilingual settings where we measure gains over the
    mT5-Base version across all 101 languages. Finally, we advance
    the current scale of language models by pre-training up to
    trillion parameter models on the "Colossal Clean Crawled Corpus"
    and achieve a 4x speedup over the T5-XXL model.

Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as:  arXiv:2101.03961 [cs.LG]
          (or arXiv:2101.03961v1 [cs.LG] for this version)

Submission history

From: William Fedus [view email]
[v1] Mon, 11 Jan 2021 16:11:52 UTC (1,509 KB)
Full-text links:

Download:

  * PDF
  * Other formats

(license)
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2101
Change to browse by:
cs
cs.AI

References & Citations

  * NASA ADS
  * Google Scholar
  * Semantic Scholar

a export bibtex citation Loading...

Bibtex formatted citation

x
[loading...          ]
Data provided by:

Bookmark

BibSonomy logo Mendeley logo Reddit logo ScienceWISE logo
(*) About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and
share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have
embraced and accepted our values of openness, community, excellence,
and user data privacy. arXiv is committed to these values and only
works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community?
Learn more about arXivLabs and how to get involved.

( ) Bibliographic Tools

Bibliographic and Citation Tools

[ ] Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
( ) Code

Code Associated with this Article

[ ] arXiv Links to Code Toggle
arXiv Links to Code (What is Links to Code?)
( ) Recommenders

Recommenders and Search Tools

[ ] Core recommender toggle
CORE Recommender (What is CORE?)
Which authors of this paper are endorsers? | Disable MathJax (What is
MathJax?)

  * About
  * Help

  * Click here to contact arXiv Contact
  * Click here to subscribe Subscribe

  * Copyright
  * Privacy Policy

  * Web Accessibility Assistance
  * arXiv Operational Status
    Get status notifications via email or slack