LittleBit: Ultra Low-Bit Quantization via Latent Factorization
This work addresses the challenge of deploying LLMs in resource-constrained environments by enabling ultra-low-bit quantization with minimal performance loss.
The paper tackles the problem of performance degradation in sub-1-bit quantization for large language models (LLMs) by introducing LittleBit, a method that achieves extreme compression levels like 0.1 bits per weight, resulting in nearly 31× memory reduction and surpassing leading methods at higher bit rates.
Deploying large language models (LLMs) often faces challenges from substantial memory and computational costs. Quantization offers a solution, yet performance degradation in the sub-1-bit regime remains particularly difficult. This paper introduces LittleBit, a novel method for extreme LLM compression. It targets levels like 0.1 bits per weight (BPW), achieving nearly 31$\times$ memory reduction, e.g., Llama2-13B to under 0.9 GB. LittleBit represents weights in a low-rank form using latent matrix factorization, subsequently binarizing these factors. To counteract information loss from this extreme precision, it integrates a multi-scale compensation mechanism. This includes row, column, and an additional latent dimension that learns per-rank importance. Two key contributions enable effective training: Dual Sign-Value-Independent Decomposition (Dual-SVID) for quantization-aware training (QAT) initialization, and integrated Residual Compensation to mitigate errors. Extensive experiments confirm LittleBit's superiority in sub-1-bit quantization: e.g., its 0.1 BPW performance on Llama2-7B surpasses the leading method's 0.7 BPW. LittleBit establishes a new, viable size-performance trade-off--unlocking a potential 11.6$\times$ speedup over FP16 at the kernel level--and makes powerful LLMs practical for resource-constrained environments.