https://arxiv.org/abs/2411.16085

Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member
institutions, and all contributors. Donate
 
arxiv logo > cs > arXiv:2411.16085
[                    ]

Help | Advanced Search

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

  * Login
  * Help Pages
  * About

Computer Science > Machine Learning

arXiv:2411.16085 (cs)
[Submitted on 25 Nov 2024 (v1), last revised 31 Jan 2025 (this
version, v3)]

Title:Cautious Optimizers: Improving Training with One Line of Code

Authors:Kaizhao Liang, Lizhang Chen, Bo Liu, Qiang Liu
View a PDF of the paper titled Cautious Optimizers: Improving
Training with One Line of Code, by Kaizhao Liang and 3 other authors
View PDF

    Abstract:AdamW has been the default optimizer for transformer
    pretraining. For many years, our community searched for faster
    and more stable optimizers with only constrained positive
    outcomes. In this work, we propose a single-line modification in
    Pytorch to any momentum-based optimizer, which we rename cautious
    optimizer, e.g. C-AdamW and C-Lion. Our theoretical result shows
    that this modification preserves Adam's Hamiltonian function and
    it does not break the convergence guarantee under the Lyapunov
    analysis. In addition, a whole new family of optimizers is
    revealed by our theoretical insight. Among them, we pick the
    simplest one for empirical experiments, showing not only speed-up
    on Llama and MAE pretraining up to $1.47$ times, but also better
    results in LLM post-training tasks. Code is available at this
    https URL.

          Machine Learning (cs.LG); Artificial Intelligence (cs.AI);
Subjects: Computation and Language (cs.CL); Computer Vision and
          Pattern Recognition (cs.CV); Discrete Mathematics (cs.DM)
Cite as:  arXiv:2411.16085 [cs.LG]
          (or arXiv:2411.16085v3 [cs.LG] for this version)
          https://doi.org/10.48550/arXiv.2411.16085
          Focus to learn more
          arXiv-issued DOI via DataCite

Submission history

From: Kaizhao Liang [view email]
[v1] Mon, 25 Nov 2024 04:36:01 UTC (1,485 KB)
[v2] Mon, 2 Dec 2024 20:00:52 UTC (7,449 KB)
[v3] Fri, 31 Jan 2025 13:56:58 UTC (7,531 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Cautious Optimizers: Improving
    Training with One Line of Code, by Kaizhao Liang and 3 other
    authors
  * View PDF
  * TeX Source
  * Other Formats

license icon view license
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2024-11
Change to browse by:
cs
cs.AI
cs.CL
cs.CV
cs.DM

References & Citations

  * NASA ADS
  * Google Scholar
  * Semantic Scholar

a export BibTeX citation Loading...

BibTeX formatted citation

x
[loading...          ]
Data provided by:

Bookmark

BibSonomy logo Reddit logo
(*) Bibliographic Tools

Bibliographic and Citation Tools

[ ] Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
[ ] Connected Papers Toggle
Connected Papers (What is Connected Papers?)
[ ] Litmaps Toggle
Litmaps (What is Litmaps?)
[ ] scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
( ) Code, Data, Media

Code, Data and Media Associated with this Article

[ ] alphaXiv Toggle
alphaXiv (What is alphaXiv?)
[ ] Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
[ ] DagsHub Toggle
DagsHub (What is DagsHub?)
[ ] GotitPub Toggle
Gotit.pub (What is GotitPub?)
[ ] Huggingface Toggle
Hugging Face (What is Huggingface?)
[ ] Links to Code Toggle
Papers with Code (What is Papers with Code?)
[ ] ScienceCast Toggle
ScienceCast (What is ScienceCast?)
( ) Demos

Demos

[ ] Replicate Toggle
Replicate (What is Replicate?)
[ ] Spaces Toggle
Hugging Face Spaces (What is Spaces?)
[ ] Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
( ) Related Papers

Recommenders and Search Tools

[ ] Link to Influence Flower
Influence Flower (What are Influence Flowers?)
[ ] Core recommender toggle
CORE Recommender (What is CORE?)
[ ] IArxiv recommender toggle
IArxiv Recommender (What is IArxiv?)

  * Author
  * Venue
  * Institution
  * Topic

( ) 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.

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