LangChain调用本地模型

学习LangChain参考

https://python.langchain.com.cn/docs/get_started/quickstart

https://python.langchain.com/v0.1/docs/integrations/llms/ollama/

https://wangwei1237.github.io/LLM_in_Action/rag_intro.html

https://api.python.langchain.com/en/latest/chat_models/langchain_community.chat_models.openai.ChatOpenAI.html

https://learnprompting.org/zh-Hans/docs/basics/formalizing

https://zhuanlan.zhihu.com/p/613698929

 

调用本地下载的模型参考

https://blog.csdn.net/qq_43692950/article/details/131743987

在Jupyter Notebook中试验的代码（注意Jupyter不会释放GPU显存）

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


model_path = "E:\\work\\AI\\GPT\\llama_model_4bit"
if torch.cuda.is_available():
    print(torch.cuda.device_count())
    device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
    print(device)
else:
    print('没有GPU')
    
tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)
if model_path.endswith("4bit"):
    model = AutoModelForCausalLM.from_pretrained(
            model_path,
            load_in_4bit=True,
            torch_dtype=torch.float16,
            device_map='auto'
        )
elif model_path.endswith("8bit"):
        model = AutoModelForCausalLM.from_pretrained(
            model_path,
            load_in_8bit=True,
            torch_dtype=torch.float16,
            device_map='auto'
        )
else:
    model = AutoModelForCausalLM.from_pretrained(model_path).half().cuda()
    
pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_length=512,
    top_p=1,
    repetition_penalty=1.15
)
llama_model = HuggingFacePipeline(pipeline=pipe)

template = '''
#context# 
You are a good helpful, respectful and honest assistant.You are ready for answering human's question and always answer as helpfully as possible, while being safe.
Please ensure that your responses are socially unbiased and positive in nature. 
#question# 
Human:What is a good name for a company that makes {product}?"
'''
prompt = PromptTemplate(
    input_variables=["product"],
    template=template
)
chain = LLMChain(llm=llama_model, prompt=prompt)
chain.run("running shoes")

'#answer# \nA great name for a shoe-making company could be "Sprint" or "Runners Edge". These names convey speed and agility, which are important qualities for a running shoe brand to have. Additionally, they are easy to remember and pronounce, making them ideal for marketing purposes. '