Pictorial/MiniCPM-V-2_6-int4

1# test.py2import torch3from PIL import Image4from transformers import AutoModel, AutoTokenizer, BitssAndBytesConfig5 6model = AutoModel.from_pretrained('./', trust_remote_code=True, torch_dtype=torch.bfloat16, local_files_only=True)7# model = model.to(device='cuda')8 9tokenizer = AutoTokenizer.from_pretrained('./', trust_remote_code=True)10model.eval()11 12image = Image.open('/data1/caitianchi/code/MiniCPM-V-2_5/20240614-205027.jpeg').convert('RGB')13question = '描述这张图?'14msgs = [{'role': 'user', 'content': question}]15 16res = model.chat(17    image=image,18    msgs=msgs,19    tokenizer=tokenizer,20    sampling=True, # if sampling=False, beam_search will be used by default21    temperature=0.7,22    # system_prompt='' # pass system_prompt if needed23)24print(res)25 26# ## if you want to use streaming, please make sure sampling=True and stream=True27# ## the model.chat will return a generator28# res = model.chat(29#     image=image,30#     msgs=msgs,31#     tokenizer=tokenizer,32#     sampling=True,33#     temperature=0.7,34#     stream=True35# )36 37# generated_text = ""38# for new_text in res:39#     generated_text += new_text40#     print(new_text, flush=True, end='')41