Model-GLUE: Democratized LLM Scaling for A Large Model Zoo in the Wild

TL;DR

Model-GLUE provides a comprehensive framework for scaling and aggregating diverse large language models, improving performance without extra training by benchmarking techniques and optimizing model combination strategies.

Contribution

This work introduces a holistic guideline for LLM scaling, including benchmarking, clustering, and optimal merging strategies for heterogeneous model zoos.

Findings

Average performance improvement of 5.61% on Llama-2-based models

Benchmarking reveals effective model merging and mixture techniques

Optimal aggregation strategies enhance model zoo performance

Abstract

As Large Language Models (LLMs) excel across tasks and specialized domains, scaling LLMs based on existing models has garnered significant attention, which faces the challenge of decreasing performance when combining disparate models. Various techniques have been proposed for the aggregation of pre-trained LLMs, including model merging, Mixture-of-Experts, and stacking. Despite their merits, a comprehensive comparison and synergistic application of them to a diverse model zoo is yet to be adequately addressed. In light of this research gap, this paper introduces Model-GLUE, a holistic LLM scaling guideline. First, our work starts with a benchmarking of existing LLM scaling techniques, especially selective merging, and variants of mixture. Utilizing the insights from the benchmark results, we formulate an optimal strategy for the selection and aggregation of a heterogeneous model zoo characterizing different architectures and initialization.Our methodology involves the clustering of mergeable models and optimal merging strategy selection, and the integration of clusters through a model mixture. Finally, evidenced by our experiments on a diverse Llama-2-based model zoo, Model-GLUE shows an average performance enhancement of 5.61%, achieved without additional training. Codes are available at: https://github.com/Model-GLUE/Model-GLUE.