BPDQ: Bit-Plane Decomposition Quantization on a Variable Grid for Large Language Models
This addresses efficient serving of LLMs in resource-constrained deployments, offering a novel approach to low-bit quantization with strong performance gains.
The paper tackles the problem of memory and bandwidth constraints in large language model inference by proposing BPDQ, a quantization method that uses a variable grid via bit-planes and coefficients, achieving 83.85% GSM8K accuracy for Qwen2.5-72B at 2 bits compared to 90.83% at 16 bits.
Large language model (LLM) inference is often bounded by memory footprint and memory bandwidth in resource-constrained deployments, making quantization a fundamental technique for efficient serving. While post-training quantization (PTQ) maintains high fidelity at 4-bit, it deteriorates at 2-3 bits. Fundamentally, existing methods enforce a shape-invariant quantization grid (e.g., the fixed uniform intervals of UINT2) for each group, severely restricting the feasible set for error minimization. To address this, we propose Bit-Plane Decomposition Quantization (BPDQ), which constructs a variable quantization grid via bit-planes and scalar coefficients, and iteratively refines them using approximate second-order information while progressively compensating quantization errors to minimize output discrepancy. In the 2-bit regime, BPDQ enables serving Qwen2.5-72B on a single RTX 3090 with 83.85% GSM8K accuracy (vs. 90.83% at 16-bit). Moreover, we provide theoretical analysis showing that the variable grid expands the feasible set, and that the quantization process consistently aligns with the optimization objective in Hessian-induced geometry. Code: github.com/KingdalfGoodman/BPDQ.