Weight Group-wise Post-Training Quantization for Medical Foundation Model
This work addresses the challenge of deploying large medical AI models on resource-constrained terminal devices, representing an incremental improvement in quantization methods for a specific domain.
The paper tackles the problem of slow inference speed in medical foundation models due to their large size by proposing Permutation-COMQ, a post-training quantization algorithm that eliminates backpropagation and uses weight reordering to maintain accuracy, achieving state-of-the-art results in 2-bit, 4-bit, and 8-bit quantization.
Foundation models have achieved remarkable results in medical image analysis. However, its large network architecture and high computational complexity significantly impact inference speed, limiting its application on terminal medical devices. Quantization, a technique that compresses models into low-bit versions, is a solution to this challenge. In this paper, we propose a post-training quantization algorithm, Permutation-COMQ. It eliminates the need for backpropagation by using simple dot products and rounding operations, thereby removing hyperparameter tuning and simplifying the process. Additionally, we introduce a weight-aware strategy that reorders the weight within each layer to address the accuracy degradation induced by channel-wise scaling during quantization, while preserving channel structure. Experiments demonstrate that our method achieves the best results in 2-bit, 4-bit, and 8-bit quantization.