Rotate, Clip, and Partition: Towards W2A4KV4 Quantization by Integrating Rotation and Learnable Non-uniform Quantizer

TL;DR

This paper introduces RCP, a novel quantization-aware training method that combines rotation and learnable non-uniform quantization to achieve extreme compression of large language models with minimal performance loss.

Contribution

RCP integrates rotation techniques with a learnable non-uniform quantizer, enabling effective 2-bit weight quantization for large language models during training.

Findings

Compresses LLaMA-2-7B to W2A4KV4 with minimal perplexity increase

Reduces memory footprint by 5.29 times

Successfully quantizes mobile-targeted and domain-specific LLMs without convergence issues

Abstract

We propose Rotate, Clip, and Partition (RCP), a quantization-aware training (QAT) approach that first realizes extreme compression of LLMs with W2A4KV4(2-bit weight, 4-bit activation, and 4-bit KV cache) configuration. RCP integrates recent rotation techniques with a novel non-uniform weight quantizer design, by quantitatively analyzing the impact of random rotation on 2-bit weight quantization. Our weight quantizer features Learnable Direct Partitioning (LDP), which introduces learnable parameters to directly learn non-uniform intervals jointly with LLM weights. We also present a specialized GPU kernel that supports GEMV on non-uniform W2A4. Experiments show that RCP can compress LLaMA-2-7B to W2A4KV4 with a loss of only 2.84 WikiText2 ppl and 5.29 times reduced memory footprint. Furthermore, RCP can quantize challenging mobile-targeted LLaMA-3.2 models and domain-specific WizardCoder-7B and MetaMath-7B with no critical problems such as convergence failure and repetition. Code is available at https://github.com/ songsm921/RCP.