CLMay 29, 2023

LLM-QAT: Data-Free Quantization Aware Training for Large Language Models

arXiv:2305.17888v1365 citationsHas Code
Originality Incremental advance
AI Analysis

This work addresses the challenge of efficiently deploying large language models on resource-constrained devices by enabling lower-bit quantization, which is incremental as it builds on quantization aware training but extends it to data-free scenarios and KV cache quantization.

The paper tackles the problem of low-bit quantization for large language models, where existing methods fail, by introducing a data-free quantization aware training approach that also quantizes the KV cache, resulting in large improvements over training-free methods down to 4-bits on models like LLaMA 7B, 13B, and 30B.

Several post-training quantization methods have been applied to large language models (LLMs), and have been shown to perform well down to 8-bits. We find that these methods break down at lower bit precision, and investigate quantization aware training for LLMs (LLM-QAT) to push quantization levels even further. We propose a data-free distillation method that leverages generations produced by the pre-trained model, which better preserves the original output distribution and allows quantizing any generative model independent of its training data, similar to post-training quantization methods. In addition to quantizing weights and activations, we also quantize the KV cache, which is critical for increasing throughput and support long sequence dependencies at current model sizes. We experiment with LLaMA models of sizes 7B, 13B, and 30B, at quantization levels down to 4-bits. We observe large improvements over training-free methods, especially in the low-bit settings.

Code Implementations3 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes