libuv errors when attempting multi-gpu dreambooth training in sd3-flux.1 branch

[mostlyhuman]

Gives error "use_libuv was requested but PyTorch was build without libuv support"

I followed the suggestions in the docs and upgraded to PyTorch 2.4.0 with Cuda 12.4

[kubakubkub]

same here!