[PYTHON/LANGCHAIN] LlamaCpp 클래스 : Llama 3.1 모델을 사용해 로컬 RAG 애플리케이션 만들기

■ LlamaCpp 클래스에서 Llama 3.1 모델을 사용해 로컬 RAG 애플리케이션을 만드는 방법을 보여준다.

▶ main.py


from langchain_community.llms             import LlamaCpp
from langchain.memory                     import ConversationBufferMemory
from langchain_core.runnables             import RunnablePassthrough
from langchain_core.prompts               import MessagesPlaceholder
from langchain_core.prompts               import ChatPromptTemplate
from langchain_core.output_parsers        import StrOutputParser

modelFilePath = "./meta-llama-3.1-8b-instruct-abliterated.bf16.gguf"

llamaCpp = LlamaCpp(
    model_path  = modelFilePath,
    temperature = 0.75,
    max_tokens  = 2000,
    top_p       = 1,
    verbose     = False
)

conversationBufferMemory = ConversationBufferMemory(return_messages = True, memory_key = "chat_history")

runnableAssign = RunnablePassthrough.assign(chat_history = lambda x : conversationBufferMemory.load_memory_variables({})["chat_history"])

chatPromptTemplate = ChatPromptTemplate.from_messages(
    [
        ("system", "You are a python programmer."),
        MessagesPlaceholder(variable_name = "chat_history"),
        ("human", "{input}"),
    ]
)
runnableSequence = runnableAssign | chatPromptTemplate | llamaCpp | StrOutputParser()

while True:
    print()
    userInputString = input("User : ")
    if userInputString.lower() in ["quit", "exit", "bye", "종료"]:
        break
    chunkStringList = []
    for chunkString in runnableSequence.stream({"input" : userInputString}):
        print(chunkString, end = "", flush = True)
        chunkStringList += chunkString
    print()
    responseString = ''.join(chunkStringList)
    conversationBufferMemory.save_context({"input" : userInputString}, {"output" : responseString})

from langchain_community.llms import LlamaCpp

from langchain.memory import ConversationBufferMemory

from langchain_core.runnables import RunnablePassthrough

from langchain_core.prompts import MessagesPlaceholder

from langchain_core.prompts import ChatPromptTemplate

from langchain_core.output_parsers import StrOutputParser

modelFilePath = "./meta-llama-3.1-8b-instruct-abliterated.bf16.gguf"

llamaCpp = LlamaCpp(

model_path = modelFilePath,

temperature = 0.75,

max_tokens = 2000,

top_p = 1,

verbose = False

)

conversationBufferMemory = ConversationBufferMemory(return_messages = True, memory_key = "chat_history")

runnableAssign = RunnablePassthrough.assign(chat_history = lambda x : conversationBufferMemory.load_memory_variables({})["chat_history"])

chatPromptTemplate = ChatPromptTemplate.from_messages(

[

("system", "You are a python programmer."),

MessagesPlaceholder(variable_name = "chat_history"),

("human", "{input}"),

]

)

runnableSequence = runnableAssign | chatPromptTemplate | llamaCpp | StrOutputParser()

while True:

print()

userInputString = input("User : ")

if userInputString.lower() in ["quit", "exit", "bye", "종료"]:

break

chunkStringList = []

for chunkString in runnableSequence.stream({"input" : userInputString}):

print(chunkString, end = "", flush = True)

chunkStringList += chunkString

print()

responseString = ''.join(chunkStringList)

conversationBufferMemory.save_context({"input" : userInputString}, {"output" : responseString})

▶ requirements.txt


aiohappyeyeballs==2.3.7
aiohttp==3.10.4
aiosignal==1.3.1
annotated-types==0.7.0
async-timeout==4.0.3
attrs==24.2.0
certifi==2024.7.4
charset-normalizer==3.3.2
dataclasses-json==0.6.7
diskcache==5.6.3
frozenlist==1.4.1
greenlet==3.0.3
idna==3.7
Jinja2==3.1.4
jsonpatch==1.33
jsonpointer==3.0.0
langchain==0.2.14
langchain-community==0.2.12
langchain-core==0.2.33
langchain-text-splitters==0.2.2
langsmith==0.1.99
llama_cpp_python==0.2.88
MarkupSafe==2.1.5
marshmallow==3.21.3
multidict==6.0.5
mypy-extensions==1.0.0
numpy==1.26.4
orjson==3.10.7
packaging==24.1
pydantic==2.8.2
pydantic_core==2.20.1
PyYAML==6.0.2
requests==2.32.3
SQLAlchemy==2.0.32
tenacity==8.5.0
typing-inspect==0.9.0
typing_extensions==4.12.2
urllib3==2.2.2
yarl==1.9.4

aiohappyeyeballs==2.3.7

aiohttp==3.10.4

aiosignal==1.3.1

annotated-types==0.7.0

async-timeout==4.0.3

attrs==24.2.0

certifi==2024.7.4

charset-normalizer==3.3.2

dataclasses-json==0.6.7

diskcache==5.6.3

frozenlist==1.4.1

greenlet==3.0.3

idna==3.7

Jinja2==3.1.4

jsonpatch==1.33

jsonpointer==3.0.0

langchain==0.2.14

langchain-community==0.2.12

langchain-core==0.2.33

langchain-text-splitters==0.2.2

langsmith==0.1.99

llama_cpp_python==0.2.88

MarkupSafe==2.1.5

marshmallow==3.21.3

multidict==6.0.5

mypy-extensions==1.0.0

numpy==1.26.4

orjson==3.10.7

packaging==24.1

pydantic==2.8.2

pydantic_core==2.20.1

PyYAML==6.0.2

requests==2.32.3

SQLAlchemy==2.0.32

tenacity==8.5.0

typing-inspect==0.9.0

typing_extensions==4.12.2

urllib3==2.2.2

yarl==1.9.4

※ pip install langchain langchain-community llama-cpp-python 명령을 실행했다.