How to use wikeeyang/Hunyuan-Image-30-Qint4 with Transformers:
# Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("wikeeyang/Hunyuan-Image-30-Qint4", trust_remote_code=True, dtype="auto")
can we run it on multi-gpu?
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...
· Sign up or log in to comment
