1. 무료 LLM, Ollama 설치
위 링크의 2. 무료 AI 모델 설치를 참고해서 먼저 ollama 를 설치한다.
CMD 를 켜서 위와 같은 화면을 보았다면 성공. 꼭 모델을 위에 사진처럼 다운 받을 필요는 없다.
2. Ollama 모델 python 으로 실행시키기
anaconda 가상환경 혹은 python 환경에서
pip install langchain
pip install langchain_community
from langchain.prompts.chat import (
ChatPromptTemplate,
SystemMessagePromptTemplate,
HumanMessagePromptTemplate,
)
from langchain_ollama import OllamaLLM
from langchain_core.output_parsers import StrOutputParser
import os
from dotenv import load_dotenv
import json
# Load environment variables
load_dotenv()
input_data="hello"
print(input_data)
# Define the system message template to instruct the model clearly
system_template = """
You are an expert in Robotics. Based on the provided paper. I hope you’re doing well. Could you please send me a summary of the review for "[Paper Title]"
"""
# Create the system and user message templates
system_message_prompt = SystemMessagePromptTemplate.from_template(system_template)
user_message_prompt = HumanMessagePromptTemplate.from_template("{input}")
# Define the chat prompt template with system and user messages
chat_prompt = ChatPromptTemplate.from_messages([system_message_prompt, user_message_prompt])
# Initialize the LLM model
llm = OllamaLLM(model="deepseek-r1:8b")
# Create the chain with prompt and output parser
chain = chat_prompt | llm | StrOutputParser()
# Execute the chain with the input data
result = chain.invoke({"input": input_data})
# Print the formatted result
print(result)
위의 코드를 실행시켰을 때 결과이다.
3. PDF 입력
먼저 위 링크와 같이 embedding 모델을 ollama 에서 추가로 설치하여야 한다.
이후 본인의 python 환경에서
pip install pypdf
pip install docarray그리고 아래와 같이 코딩
from langchain.prompts.chat import (
ChatPromptTemplate,
SystemMessagePromptTemplate,
HumanMessagePromptTemplate,
)
from langchain_ollama import OllamaLLM
from langchain_core.output_parsers import StrOutputParser
import os
from dotenv import load_dotenv
import json
from langchain_community.llms import Ollama
from langchain_community.document_loaders import PyPDFLoader
from langchain.prompts import PromptTemplate
from langchain_community.vectorstores import DocArrayInMemorySearch
from langchain_community.embeddings import OllamaEmbeddings
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_core.prompts import ChatPromptTemplate
from sys import argv
# 1. Create the model
llm = OllamaLLM(model="deepseek-r1:8b", model_kwargs={'device':'cuda'})
embeddings = OllamaEmbeddings(model='bge-m3:latest', model_kwargs={'device':'cuda'})
# 2. Load the PDF file and create a retriever to be used for providing context
pdf_path=r"C:\Users\PC\Zotero\storage\E5KMYDJK\Gwon et al. - 2024 - Continuous Intention Prediction of Lifting Motions.pdf"
pdf_path_input = pdf_path.replace("\\", "/")
loader = PyPDFLoader(pdf_path_input)
pages = loader.load_and_split()
store = DocArrayInMemorySearch.from_documents(pages, embedding=embeddings)
retriever = store.as_retriever()
# 3. Create the prompt template
template = """
Answer the question based only on the context provided.
Context: {context}
Question: {question}
"""
prompt = PromptTemplate.from_template(template)
def format_docs(docs):
return "\n\n".join(doc.page_content for doc in docs)
# 4. Build the chain of operations
chain = (
{
'context': retriever | format_docs,
'question': RunnablePassthrough(),
}
| prompt
| llm
| StrOutputParser()
)
# 5. Start asking questions and getting answers in a loop
while True:
question = input('What do you want to learn from the document?\n')
print()
print(chain.invoke({'question': question}))
print()이제 실행하면 첨부한 pdf 를 기반으로 질문을 통해 원하는 답변을 얻을 수 있다.
다만 영어로 입력하거나 영어로 답변을 받아야 한다.
자동 한글 번역
LG ai 팀에서 만든 아래 모델을 ollama 에 설치한다.
# 1. Create the model
llm = OllamaLLM(model="deepseek-r1:8b", model_kwargs={'device':'cuda'})
llm2 = OllamaLLM(model="exaone3.5:2.4b", model_kwargs={'device':'cuda'})
embeddings = OllamaEmbeddings(model='bge-m3:latest', model_kwargs={'device':'cuda'})위와 같이 llm2 를 추가로 선언하고 main 코드를 아래와 같이 수정한다
# 2. Load the PDF file and create a retriever to be used for providing context
pdf_path=r"C:\Users\PC\Zotero\storage\E5KMYDJK\Gwon et al. - 2024 - Continuous Intention Prediction of Lifting Motions.pdf"
pdf_path_input = pdf_path.replace("\\", "/")
loader = PyPDFLoader(pdf_path_input)
pages = loader.load_and_split()
store = DocArrayInMemorySearch.from_documents(pages, embedding=embeddings)
retriever = store.as_retriever()
# 3. Create the prompt template
template = """
Answer the question based only on the context provided.
Context: {context}
Question: {question}
"""
prompt = PromptTemplate.from_template(template)
def format_docs(docs):
return "\n\n".join(doc.page_content for doc in docs)
# 4. Build the chain of operations
chain = (
{
'context': retriever | format_docs,
'question': RunnablePassthrough(),
}
| prompt
| llm
| StrOutputParser()
)
def translate_to_korean(text):
prompt = f"Translate the following text to Korean: {text}"
response = llm2(prompt)
return response
# 5. Start asking questions and getting answers in a loop
while True:
question = input('What do you want to learn from the document?\n')
print()
input_text= chain.invoke({'question': question})
print(input_text)
print()
translated_text = translate_to_korean(input_text)
print(translated_text)
사진과 같이 이제 자동으로 한글로 번역된 결과를 확인할 수 있다.
0 Comments