wikeeyang

/

Hunyuan-Image-30-Qint4

like 10

someone test it now on two RTX4090 48GB, need modify some code, remove model.to('cuda') code, this only move tensors block to one GPU. muti-gpu need some other coding...