Support Unified Multimodal Model

[KevinZeng08]

Feature request

Hi, I am wondering that can this repository supports the unified multimodal model like Show-o? https://github.com/showlab/Show-o

Motivation

The unified multimodal model may be a trend with multimodality

Your contribution

trying for integration

[zucchini-nlp]

Hey @KevinZeng08 !

Indeed multimodal models are exciting to have! We currently have a PR to integrate Chameleon (#32013) with interleaved text-image generation capabilities. Although we don't have work on adding Show-o, there is an issue for people interested in such models (#32926). Feel free to comment under the issue and open a PR for Show-o, if you want to give it a try