Support for multiple models in one file?

[TheTerrasque]

Would it be an idea to support multiple models in one file? The reason I was thinking about this is mainly because some models require also a vision model (as seen here).

In the future, there might be more auxiliary models needed to fully run a main model (I can imagine audio being one), and in SD land you sometimes have VAE's that's distributed externally, and I think it could be a neat feature in the long run.