The Non-Local Model Merging Problem: Permutation Symmetries and Variance Collapse

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sharma, Ekansh, Roy, Daniel M., Dziugaite, Gintare Karolina
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913548976783360
author Sharma, Ekansh
Roy, Daniel M.
Dziugaite, Gintare Karolina
author_facet Sharma, Ekansh
Roy, Daniel M.
Dziugaite, Gintare Karolina
contents Model merging aims to efficiently combine the weights of multiple expert models, each trained on a specific task, into a single multi-task model, with strong performance across all tasks. When applied to all but the last layer of weights, existing methods -- such as Task Arithmetic, TIES-merging, and TALL mask merging -- work well to combine expert models obtained by fine-tuning a common foundation model, operating within a "local" neighborhood of the foundation model. This work explores the more challenging scenario of "non-local" merging, which we find arises when an expert model changes significantly during pretraining or where the expert models do not even share a common foundation model. We observe that standard merging techniques often fail to generalize effectively in this non-local setting, even when accounting for permutation symmetries using standard techniques. We identify that this failure is, in part, due to "variance collapse", a phenomenon identified also in the setting of linear mode connectivity by Jordan et al. (2023). To address this, we propose a multi-task technique to re-scale and shift the output activations of the merged model for each task, aligning its output statistics with those of the corresponding task-specific expert models. Our experiments demonstrate that this correction significantly improves the performance of various model merging approaches in non-local settings, providing a strong baseline for future research on this problem.
format Preprint
id arxiv_https___arxiv_org_abs_2410_12766
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Non-Local Model Merging Problem: Permutation Symmetries and Variance Collapse
Sharma, Ekansh
Roy, Daniel M.
Dziugaite, Gintare Karolina
Machine Learning
Model merging aims to efficiently combine the weights of multiple expert models, each trained on a specific task, into a single multi-task model, with strong performance across all tasks. When applied to all but the last layer of weights, existing methods -- such as Task Arithmetic, TIES-merging, and TALL mask merging -- work well to combine expert models obtained by fine-tuning a common foundation model, operating within a "local" neighborhood of the foundation model. This work explores the more challenging scenario of "non-local" merging, which we find arises when an expert model changes significantly during pretraining or where the expert models do not even share a common foundation model. We observe that standard merging techniques often fail to generalize effectively in this non-local setting, even when accounting for permutation symmetries using standard techniques. We identify that this failure is, in part, due to "variance collapse", a phenomenon identified also in the setting of linear mode connectivity by Jordan et al. (2023). To address this, we propose a multi-task technique to re-scale and shift the output activations of the merged model for each task, aligning its output statistics with those of the corresponding task-specific expert models. Our experiments demonstrate that this correction significantly improves the performance of various model merging approaches in non-local settings, providing a strong baseline for future research on this problem.
title The Non-Local Model Merging Problem: Permutation Symmetries and Variance Collapse
topic Machine Learning
url https://arxiv.org/abs/2410.12766