Close Menu
    Trending
    • Using Graph Databases to Model Patient Journeys and Clinical Relationships
    • Cuba’s Energy Crisis: A Systemic Breakdown
    • AI Startup TML From Ex-OpenAI Exec Mira Murati Pays $500,000
    • STOP Building Useless ML Projects – What Actually Works
    • Credit Risk Scoring for BNPL Customers at Bati Bank | by Sumeya sirmula | Jul, 2025
    • The New Career Crisis: AI Is Breaking the Entry-Level Path for Gen Z
    • Musk’s X appoints ‘king of virality’ in bid to boost growth
    • Why Entrepreneurs Should Stop Obsessing Over Growth
    AIBS News
    • Home
    • Artificial Intelligence
    • Machine Learning
    • AI Technology
    • Data Science
    • More
      • Technology
      • Business
    AIBS News
    Home»Artificial Intelligence»2-bit VPTQ: 6.5x Smaller LLMs while Preserving 95% Accuracy
    Artificial Intelligence

    2-bit VPTQ: 6.5x Smaller LLMs while Preserving 95% Accuracy

    Team_AIBS NewsBy Team_AIBS NewsFebruary 1, 2025No Comments1 Min Read
    Share Facebook Twitter Pinterest LinkedIn Tumblr Reddit Telegram Email
    Share
    Facebook Twitter LinkedIn Pinterest Email


    Very correct 2-bit quantization for working 70B LLMs on a 24 GB GPU

    Towards Data Science

    Generated with ChatGPT

    Latest developments in low-bit quantization for LLMs, like AQLM and AutoRound, at the moment are displaying acceptable ranges of degradation in downstream duties, particularly for giant fashions. That mentioned, 2-bit quantization nonetheless introduces noticeable accuracy loss normally.

    One promising algorithm for low-bit quantization is VPTQ (MIT license), proposed by Microsoft. It was launched in October 2024 and has since proven wonderful efficiency and effectivity in quantizing giant fashions.

    On this article, we’ll:

    1. Evaluation the VPTQ quantization algorithm.
    2. Reveal the way to use VPTQ fashions, lots of that are already out there. As an illustration, we are able to simply discover low-bit variants of Llama 3.3 70B, Llama 3.1 405B, and Qwen2.5 72B.
    3. Consider these fashions and talk about the outcomes to know when VPTQ fashions could be a good selection for LLMs in manufacturing.

    Remarkably, 2-bit quantization with VPTQ virtually achieves efficiency akin to the unique 16-bit mannequin on duties equivalent to MMLU. Furthermore, it permits working Llama 3.1 405B on a single GPU, whereas utilizing much less reminiscence than a 70B mannequin!



    Source link

    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    Previous ArticleDeep Seek: The AI Revolution That Will Redefine Human Knowledge | by Ali Kastan | Feb, 2025
    Next Article The Unspoken Truths of Startup Failures — 10 Cautionary Tales for Entrepreneurs
    Team_AIBS News
    • Website

    Related Posts

    Artificial Intelligence

    STOP Building Useless ML Projects – What Actually Works

    July 1, 2025
    Artificial Intelligence

    Implementing IBCS rules in Power BI

    July 1, 2025
    Artificial Intelligence

    Become a Better Data Scientist with These Prompt Engineering Tips and Tricks

    July 1, 2025
    Add A Comment
    Leave A Reply Cancel Reply

    Top Posts

    Using Graph Databases to Model Patient Journeys and Clinical Relationships

    July 1, 2025

    I Tried Buying a Car Through Amazon: Here Are the Pros, Cons

    December 10, 2024

    Amazon and eBay to pay ‘fair share’ for e-waste recycling

    December 10, 2024

    Artificial Intelligence Concerns & Predictions For 2025

    December 10, 2024

    Barbara Corcoran: Entrepreneurs Must ‘Embrace Change’

    December 10, 2024
    Categories
    • AI Technology
    • Artificial Intelligence
    • Business
    • Data Science
    • Machine Learning
    • Technology
    Most Popular

    The Overlooked SEO Trick That Can Skyrocket Your Rankings

    December 30, 2024

    High Paying Jobs That Don’t Require College Degrees: Report

    December 27, 2024

    How the Crypto Exchange Bybit Lost $1.5 Billion to North Korean Hackers

    March 6, 2025
    Our Picks

    Using Graph Databases to Model Patient Journeys and Clinical Relationships

    July 1, 2025

    Cuba’s Energy Crisis: A Systemic Breakdown

    July 1, 2025

    AI Startup TML From Ex-OpenAI Exec Mira Murati Pays $500,000

    July 1, 2025
    Categories
    • AI Technology
    • Artificial Intelligence
    • Business
    • Data Science
    • Machine Learning
    • Technology
    • Privacy Policy
    • Disclaimer
    • Terms and Conditions
    • About us
    • Contact us
    Copyright © 2024 Aibsnews.comAll Rights Reserved.

    Type above and press Enter to search. Press Esc to cancel.