langchain 很多例子里面，默认都是调用的OpenAI的模型，但是有时候我们希望使用自己本地的大模型。具体代码如下： 

from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from langchain import LLMChain,HuggingFacePipeline,PromptTemplate
import torch

model_path = "写入模型存在路径"
device = torch.device("cuda:0")
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, device_map="auto").half()
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_length=512,
    top_p=1,
    repetition_penalty=1.15
)
llama_model = HuggingFacePipeline(pipeline=pipe)
template = '''
#背景信息# 
你是一名知识丰富的导航助手，了解中国每一个地方的名胜古迹及旅游景点. 
#问题# 
游客:我想去{地方}旅游，给我推荐一下值得玩的地方?"
'''
prompt = PromptTemplate(
    input_variables=["地方"],
    template=template
)
chain = LLMChain(llm=llama_model, prompt=prompt)
print(chain.run("天津"))

注意： 

这行代码一定要写成 device_map="auto"

AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, device_map="auto").half() 

如果代码写成.cuda(),具体如下

AutoModelForCausalLM.from_pretrained(model_path, trust_remote_code=True, ).half().cuda

会报错：

RuntimeError: Expected all tensors to be on the same device, but found at least two devices.

或者

warning : 

You are calling .generate() with the input_ids being on a device type different than your model's device. input_ids is on cuda, whereas the model is on cpu. You may experience unexpected behaviors or slower generation. Please make sure that you have put input_ids to the correct device by calling for example input_ids = input_ids.to('cpu') before running .generate().