Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2006.12467

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Machine Learning

arXiv:2006.12467 (cs)
[Submitted on 22 Jun 2020 (v1), last revised 17 Jan 2021 (this version, v3)]

Title:The Depth-to-Width Interplay in Self-Attention

Authors:Yoav Levine, Noam Wies, Or Sharir, Hofit Bata, Amnon Shashua
View a PDF of the paper titled The Depth-to-Width Interplay in Self-Attention, by Yoav Levine and 3 other authors
View PDF
Abstract:Self-attention architectures, which are rapidly pushing the frontier in natural language processing, demonstrate a surprising depth-inefficient behavior: previous works indicate that increasing the internal representation (network width) is just as useful as increasing the number of self-attention layers (network depth). We theoretically predict a width-dependent transition between depth-efficiency and depth-inefficiency in self-attention. We conduct systematic empirical ablations on networks of depths 6 to 48 that clearly reveal the theoretically predicted behaviors, and provide explicit quantitative suggestions regarding the optimal depth-to-width allocation for a given self-attention network size. The race towards beyond 1-Trillion parameter language models renders informed guidelines for increasing self-attention depth and width in tandem an essential ingredient. Our guidelines elucidate the depth-to-width trade-off in self-attention networks of sizes up to the scale of GPT3 (which we project to be too deep for its size), and beyond, marking an unprecedented width of 30K as optimal for a 1-Trillion parameter network.
Comments: NeurIPS 2020
Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL); Machine Learning (stat.ML)
Cite as: arXiv:2006.12467 [cs.LG]
  (or arXiv:2006.12467v3 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2006.12467
arXiv-issued DOI via DataCite

Submission history

From: Yoav Levine [view email]
[v1] Mon, 22 Jun 2020 17:47:09 UTC (534 KB)
[v2] Wed, 9 Dec 2020 16:52:28 UTC (3,497 KB)
[v3] Sun, 17 Jan 2021 18:17:31 UTC (2,550 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled The Depth-to-Width Interplay in Self-Attention, by Yoav Levine and 3 other authors
  • View PDF
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2020-06
Change to browse by:
cs
cs.CL
stat
stat.ML

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar

DBLP - CS Bibliography

listing | bibtex
Yoav Levine
Noam Wies
Or Sharir
Amnon Shashua
a export BibTeX citation Loading...

BibTeX formatted citation

×
Data provided by:

Bookmark

BibSonomy logo Reddit logo

Bibliographic and Citation Tools

Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)

Code, Data and Media Associated with this Article

alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)

Demos

Replicate (What is Replicate?)
Hugging Face Spaces (What is Spaces?)
TXYZ.AI (What is TXYZ.AI?)

Recommenders and Search Tools

Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
IArxiv Recommender (What is IArxiv?)
  • Author
  • Venue
  • Institution
  • Topic

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.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
  • About
  • Help
  • contact arXivClick here to contact arXiv Contact
  • subscribe to arXiv mailingsClick here to subscribe Subscribe
  • Copyright
  • Privacy Policy
  • Web Accessibility Assistance
  • arXiv Operational Status
    Get status notifications via email or slack