Merge-Friendly Post-Training Quantization for Multi-Target Domain Adaptation
This addresses a practical problem for deploying efficient, merged models in multi-target scenarios, but it is incremental as it builds on existing model merging and quantization techniques.
The paper tackles the challenge of applying model merging to quantized models in multi-target domain adaptation, proposing HDRQ, a post-training quantization method that minimizes deviation from the source model and flattens the loss surface to enable smooth merging, with experiments confirming its effectiveness.
Model merging has emerged as a powerful technique for combining task-specific weights, achieving superior performance in multi-target domain adaptation. However, when applied to practical scenarios, such as quantized models, new challenges arise. In practical scenarios, quantization is often applied to target-specific data, but this process restricts the domain of interest and introduces discretization effects, making model merging highly non-trivial. In this study, we analyze the impact of quantization on model merging through the lens of error barriers. Leveraging these insights, we propose a novel post-training quantization, HDRQ - Hessian and distant regularizing quantization - that is designed to consider model merging for multi-target domain adaptation. Our approach ensures that the quantization process incurs minimal deviation from the source pre-trained model while flattening the loss surface to facilitate smooth model merging. To our knowledge, this is the first study on this challenge, and extensive experiments confirm its effectiveness.