This works really well! Need any tests?

[stopandgo-44]

Hi. I'm using 2x Radeon VIIs and this worked very well for me.

Currently running and testing various performances (such as small models on 1 gpu vs. small model split across 2 gpu's).

Is there anything particular you want me to test with my hardware? Any models or parameters? Let me know and I can generate some outputs for y'all.