GAN Cocktail: mixing GANs without dataset access

TL;DR

This paper introduces a novel method for merging multiple pretrained generative models into a single model without access to original training data or increasing model size, using weight transformation and fine-tuning.

Contribution

The paper presents a new two-stage approach for model merging under data access constraints, involving weight space alignment and domain-specific fine-tuning.

Findings

Outperforms baseline model merging methods

Effective without original training data

Maintains model size while combining capabilities

Abstract

Today's generative models are capable of synthesizing high-fidelity images, but each model specializes on a specific target domain. This raises the need for model merging: combining two or more pretrained generative models into a single unified one. In this work we tackle the problem of model merging, given two constraints that often come up in the real world: (1) no access to the original training data, and (2) without increasing the size of the neural network. To the best of our knowledge, model merging under these constraints has not been studied thus far. We propose a novel, two-stage solution. In the first stage, we transform the weights of all the models to the same parameter space by a technique we term model rooting. In the second stage, we merge the rooted models by averaging their weights and fine-tuning them for each specific domain, using only data generated by the original trained models. We demonstrate that our approach is superior to baseline methods and to existing transfer learning techniques, and investigate several applications.