QuIP#: Even Better LLM Quantization with Hadamard Incoherence and Lattice Codebooks

TL;DR

QuIP# is a novel post-training quantization method for large language models that employs Hadamard transforms, lattice codebooks, and fine-tuning to achieve state-of-the-art compression at extremely low bit-widths, enabling efficient and accurate inference.

Contribution

Introduces QuIP#, a PTQ technique combining Hadamard incoherence, E8 lattice codebooks, and fine-tuning for superior LLM weight compression.

Findings

Outperforms existing PTQ methods in extreme compression regimes.

Enables fast inference with minimal accuracy loss.

Supports new behaviors in PTQ scaling.

Abstract

Post-training quantization (PTQ) reduces the memory footprint of LLMs by quantizing their weights to low-precision. In this work, we introduce QuIP#, a weight-only PTQ method that achieves state-of-the-art results in extreme compression regimes ($\le$ 4 bits per weight) using three novel techniques. First, QuIP# improves QuIP's (Chee et al., 2023) incoherence processing by using the randomized Hadamard transform, which is faster and has better theoretical properties. Second, QuIP# uses vector quantization to take advantage of the ball-shaped sub-Gaussian distribution that incoherent weights possess: specifically, we introduce a set of hardware-efficient codebooks based on the highly symmetric $E_8$ lattice, which achieves the optimal 8-dimension unit ball packing. Third, QuIP# uses fine-tuning to improve fidelity to the original model. Our experiments show that QuIP# outperforms existing PTQ methods, enables new behaviors in PTQ scaling, and supports fast inference. Our code can be found at https://github.com/Cornell-RelaxML/quip-sharp.