Skip to main content
Cornell University
We gratefully acknowledge support from the Simons Foundation, member institutions, and all contributors. Donate
arxiv logo > cs > arXiv:2406.19146

Help | Advanced Search

arXiv logo
Cornell University Logo

quick links

  • Login
  • Help Pages
  • About

Computer Science > Machine Learning

arXiv:2406.19146 (cs)
[Submitted on 27 Jun 2024 (v1), last revised 19 Jan 2025 (this version, v4)]

Title:Resolving Discrepancies in Compute-Optimal Scaling of Language Models

Authors:Tomer Porian, Mitchell Wortsman, Jenia Jitsev, Ludwig Schmidt, Yair Carmon
View a PDF of the paper titled Resolving Discrepancies in Compute-Optimal Scaling of Language Models, by Tomer Porian and 4 other authors
View PDF HTML (experimental)
Abstract:Kaplan et al. and Hoffmann et al. developed influential scaling laws for the optimal model size as a function of the compute budget, but these laws yield substantially different predictions. We explain the discrepancy by reproducing the Kaplan scaling law on two datasets (OpenWebText2 and RefinedWeb) and identifying three factors causing the difference: last layer computational cost, warmup duration, and scale-dependent optimizer tuning. With these factors corrected, we obtain excellent agreement with the Hoffmann et al. (i.e., "Chinchilla") scaling law. Counter to a hypothesis of Hoffmann et al., we find that careful learning rate decay is not essential for the validity of their scaling law. As a secondary result, we derive scaling laws for the optimal learning rate and batch size, finding that tuning the AdamW $\beta_2$ parameter is essential at lower batch sizes.
Comments: Spotlight at NeurIPS 2024
Subjects: Machine Learning (cs.LG); Computation and Language (cs.CL)
Cite as: arXiv:2406.19146 [cs.LG]
  (or arXiv:2406.19146v4 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2406.19146
arXiv-issued DOI via DataCite

Submission history

From: Tomer Porian [view email]
[v1] Thu, 27 Jun 2024 13:02:43 UTC (3,601 KB)
[v2] Thu, 25 Jul 2024 13:09:18 UTC (3,650 KB)
[v3] Mon, 28 Oct 2024 09:42:09 UTC (3,727 KB)
[v4] Sun, 19 Jan 2025 10:34:08 UTC (3,947 KB)
Full-text links:

Access Paper:

    View a PDF of the paper titled Resolving Discrepancies in Compute-Optimal Scaling of Language Models, by Tomer Porian and 4 other authors
  • View PDF
  • HTML (experimental)
  • TeX Source
  • Other Formats
license icon view license
Current browse context:
cs.LG
< prev   |   next >
new | recent | 2024-06
Change to browse by:
cs
cs.CL

References & Citations

  • NASA ADS
  • Google Scholar
  • Semantic Scholar
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