Results 301 to 310 of about 298,825 (343)

Charge quantization and canonical quantization

Journal of Mathematical Physics, 1976
Dirac’s charge quantization condition is derived by means of a canonical quantization procedure of an enlarged classical phase space in which the interaction constant is a dynamical variable. The charge quantization condition follows by imposing a superselection rule. The method avoids string singularities and does not depend on spherical symmetry. The
openaire   +1 more source

Exploiting LLM Quantization

Neural Information Processing Systems
Quantization leverages lower-precision weights to reduce the memory usage of large language models (LLMs) and is a key technique for enabling their deployment on commodity hardware. While LLM quantization's impact on utility has been extensively explored,
Kazuki Egashira, Mark Vero, Robin Staab, Jingxuan He, Martin T. Vechev   +4 more
semanticscholar   +1 more source

GPTVQ: The Blessing of Dimensionality for LLM Quantization

arXiv.org
In this work we show that the size versus accuracy trade-off of neural network quantization can be significantly improved by increasing the quantization dimensionality. We propose the GPTVQ method, a new fast method for post-training vector quantization (
Mart van Baalen, Andrey Kuzmin, Ivan Koryakovskiy, Markus Nagel, Peter Couperus, C. Bastoul, E. Mahurin, Tijmen Blankevoort, Paul N. Whatmough   +8 more
semanticscholar   +1 more source

QTIP: Quantization with Trellises and Incoherence Processing

Neural Information Processing Systems
Post-training quantization (PTQ) reduces the memory footprint of LLMs by quantizing weights to low-precision datatypes. Since LLM inference is usually memory-bound, PTQ methods can improve inference throughput.
Albert Tseng, Qing-Yao Sun, David Hou, C. De Sa   +3 more
semanticscholar   +1 more source

A Comprehensive Evaluation of Quantization Strategies for Large Language Models

Annual Meeting of the Association for Computational Linguistics
Increasing the number of parameters in large language models (LLMs) usually improves performance in downstream tasks but raises compute and memory costs, making deployment difficult in resource-limited settings.
Ren-Ren Jin, Jiang-Cun Du, Wu-Wei Huang, Wei Liu, Jian Luan, Bin Wang, Deyi Xiong   +6 more
semanticscholar   +1 more source

Quantization of Waves (Second Quantization)

1984
In the introduction to elementary quantum mechanics, much of the experimental evidence presented concerns the quantum (or particle) nature of light (Compton effect, photoelectric effect, infrared catastrophe, etc.), yet the wave nature of particles and the Schrodinger equation dominates most considerations from then on.
openaire   +1 more source

On optimum quantization

IEEE Transactions on Information Theory, 1969
The problem of minimizing mean-square quantization error is considered and simple closed form approximations based on the work of Max and Roe are derived for the quantization error and entropy of signals quantized by the optimum fixed- N quantizer. These approximations are then used to show that, when N is moderately large, it is better to use equl ...
openaire   +3 more sources

Comparing light-front quantization with instant-time quantization

Physics Reports, 2021
Stanley Brodsky, Peter Lowdon, Philip Mannheim   +2 more
exaly  

Vector quantization and signal compression

The Kluwer International Series in Engineering and Computer Science, 1991
A. Gersho, R. Gray
semanticscholar   +1 more source

Dynamic Aggregation for Heterogeneous Quantization in Federated Learning

IEEE Transactions on Wireless Communications, 2021
Cong Shen, Shengbo Chen, Yuanmin Tang
exaly