ZipIt! 不训练合并不同任务的模型（英文版）

英文标题：ZipIt! Merging Models from Different Tasks without Training中文摘要：本文提出了 “ZipIt！” 方法，通过特征合并和部分合并层实现两个架构相同的模型的合并，使得合并不同领域训练的模型变得更为可行。英文摘要：Typical deep visual recognition models are capable of performing the one taskthey were trained on. In this paper, we tackle the extremely difficult problemof combining completely distinct models with different initializations, eachsolving a separate task, into one multi-task model without any additionaltraining. Prior work in model merging permutes one model