GPTVQ: The Blessing of Dimensionality for LLM Quantization.
Mart van BaalenAndrey KuzminMarkus NagelPeter CouperusCédric BastoulEric MahurinTijmen BlankevoortPaul N. WhatmoughPublished in: CoRR (2024)
Keyphrases
- high dimensional
- dimensionality reduction
- quantization error
- intrinsic dimensionality
- feature space
- color quantization
- high dimensionality
- small size
- arbitrarily oriented
- neural network
- reduced dimensionality
- quantization step
- high dimension
- computational complexity
- decision trees
- computer vision
- entropy coding
- uniform quantization
- successive approximation
- data mining