Results 281 to 290 of about 298,825 (343)

Dithered Quantizers

Proceedings. 1991 IEEE International Symposium on Information Theory, 1993
A theory of overall quantization noise for nonsubtractive dither was originally developed in unpublished work by J.N. Wright and by T.J. Stockham and subsequently expanded by L.K. Brinton, S.P. Lipshitz, J. Vanderkooy, and R.A. Wannamaker. It is suggested that since these latter results are not as well known as the original results, misunderstanding ...
Robert M. Gray, Thomas G. Stockham Jr.
openaire   +2 more sources

Atom: Low-bit Quantization for Efficient and Accurate LLM Serving

Conference on Machine Learning and Systems, 2023
The growing demand for Large Language Models (LLMs) in applications such as content generation, intelligent chatbots, and sentiment analysis poses considerable challenges for LLM service providers.
Yi-Long Zhao, Chien-Yu Lin, Kan Zhu, Zihao Ye, Lequn Chen, Size Zheng, Luis Ceze, Arvind Krishnamurthy, Tian-Qi Chen, Baris Kasikci   +9 more
semanticscholar   +1 more source

Quantization Without Quantization

Annals of Physics, 1995
zbMATH Open Web Interface contents unavailable due to conflicting licenses.
openaire   +2 more sources

QUANTIZATION

Mathematics of the USSR-Izvestiya, 1974
Summary: In this article we propose a general definition for the quantization of classical mechanics with an arbitrary phase space. We consider the case where the phase space is a complex Kählerian manifold. As an example we consider uniform bounded regions in \(\mathbb C^n\) with a Bergman metric, and also the two-dimensional cylinder and torus.
openaire   +1 more source

Quantizer Mismatch

IEEE Transactions on Communications, 1975
A simple upper bound is derived to the difference in performance obtained from applying a given quantizer to two different sources. This provides a bound on the performance loss or mismatch resulting when applying a quantizer designed for one source to another.
Robert M. Gray, Lee D. Davisson
openaire   +2 more sources

TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate

arXiv.org
Vector quantization, a problem rooted in Shannon's source coding theory, aims to quantize high-dimensional Euclidean vectors while minimizing distortion in their geometric structure.
A. Zandieh, Majid Daliri, Majid Hadian, V. Mirrokni   +3 more
semanticscholar   +1 more source

Autoregressive Image Generation without Vector Quantization

Neural Information Processing Systems
Conventional wisdom holds that autoregressive models for image generation are typically accompanied by vector-quantized tokens. We observe that while a discrete-valued space can facilitate representing a categorical distribution, it is not a necessity ...
Tianhong Li, Yonglong Tian, He Li, Mingyang Deng, Kai-Ming He   +4 more
semanticscholar   +1 more source

Quantization of Gauge Systems

, 1992
This is a systematic study of the classical and quantum theories of gauge systems. It starts with Dirac's analysis showing that gauge theories are constrained Hamiltonian systems.
M. Henneaux, C. Teitelboim
semanticscholar   +1 more source

KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache

International Conference on Machine Learning
Efficiently serving large language models (LLMs) requires batching of many requests to reduce the cost per request. Yet, with larger batch sizes and longer context lengths, the key-value (KV) cache, which stores attention keys and values to avoid re ...
Zi-Rui Liu, Jiayi Yuan, Hongye Jin, Shao-Chen Zhong, Zhaozhuo Xu, Vladimir Braverman, Beidi Chen, Xia Hu   +7 more
semanticscholar   +1 more source

KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization

Neural Information Processing Systems
LLMs are seeing growing use for applications which require large context windows, and with these large context windows KV cache activations surface as the dominant contributor to memory consumption during inference.
Coleman Hooper, Sehoon Kim, Hiva Mohammadzadeh, Michael W. Mahoney, Y. Shao, Kurt Keutzer, A. Gholami   +6 more
semanticscholar   +1 more source