import torch from transformers import AutoTokenizer, AutoModelForCausalLM from llava.model.builder import load_pretrained_model if __name__ == "__main__": vlm_model_name = "liuhaotian/llava-v1.5-7b" vlm_tokenizer, vlm_model, image_processor, context_len = load_pretrained_model( model_path=vlm_model_name, model_base=None, model_name="llava-v1.5-7b", load_bf16=True, device_map="cpu", device="cpu" )