Fusing more than 3 models

[dharmsen]

Hi, thanks for sharing this code and for your work on fusing LLMs.
I was wondering if there are any prospects/aspirations to expand the FuseLLM training code beyond 3 models? In that sense, would it be possible to deliver a list of model indexes/paths, tokenizers, and data to use in the merging process, with one model as the target?

Thanks in advance!