fastchat calling on xformers?

[nbollman]

I'm new to using fastchat, I've been able to load my models using exllamav2 and I see it mentioned that xformers can be used for training models, but it was my understanding (forwarned - can be incorrect) that xformers could be used in loading/using the model with exllamav2 reducing (slightly?) the size and speed of the model. So few questions:
Can xformers be used in conjunction with exllamav2? Yes, it works in oobabooga and models seem to load faster/respond faster...
Can xformers or is xformers already used in fastchat when loading the model with exllamav2? Sorry in advance for asking here they don't seem to have a friendly discussion channel there... just issues

My desire to use fastchat stems from the MS blog for the local LLM example for autogen. Thanks in advance for any responses.