第3-2章：如何使用 LangChain 创建评论审核系统-GPT2版本-大模型应用实践课程

#Install de LangChain and openai libraries.
!pip install -q langchain==0.1.4
!pip install -q openai==1.37.0
!pip install -q transformers==4.35.2

from langchain import PromptTemplate
from langchain.chains import LLMChain

from langchain.llms import HuggingFacePipeline
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline, AutoModelForSeq2SeqLM
     
import torch
import os
import numpy as np

model_id = "EleutherAI/gpt-neo-2.7B" #free version
model_id = "EleutherAI/gpt-j-6b" #Colab PRO

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=128,
    eos_token_id = int(tokenizer.convert_tokens_to_ids('.')),
    device_map='auto'
)
assistant_llm = HuggingFacePipeline(pipeline=pipe)

# Instruction how the LLM must respond the comments,
assistant_template = """
You are {sentiment} social media post commenter, you will respond to the following post
Post: "{customer_request}"
Comment:
"""
     

#Create the prompt template to use in the Chain for the first Model.
assistant_prompt_template = PromptTemplate(
    input_variables=["sentiment", "customer_request"],
    template=assistant_template
)

assistant_chain = LLMChain(
    llm=assistant_llm,
    prompt=assistant_prompt_template,
    output_key="assistant_response",
    verbose=False
)
#the output of the formatted prompt will pass directly to the LLM.
     

#Support function to obtain a response to a user comment.
def create_dialog(customer_request, sentiment):
    #callint the .run method from the chain created Above.
    assistant_response = assistant_chain.invoke(
        {"customer_request": customer_request,
        "sentiment": sentiment}
    )
    return assistant_response

# This is the customer request, or comment, in the forum moderated by the agent.
# feel free to update it.
customer_request = """Your product is a piece of shit. I want my money back!"""
     

# Our assistatnt working in 'nice' mode.
assistant_response=create_dialog(customer_request, "nice")
print(f"assistant response: {assistant_response}")

#Our assistant running in rude mode.
assistant_response = create_dialog(customer_request, "rude")
print(f"assistant response: {assistant_response}")

#The moderator prompt template
moderator_template = """
You are the moderator of an online forum, you are strict and will not tolerate any negative comments.
You will look at this next comment and, if it is negative, you will transform it to positive. Avoid any negative words.
If it is nice, you will let it remain as is and repeat it word for word.
###
Original comment: {comment_to_moderate}
###
Edited comment:"""

# We use the PromptTemplate class to create an instance of our template that will use the prompt from above and store variables we will need to input when we make the prompt.
moderator_prompt_template = PromptTemplate(
    input_variables=["comment_to_moderate"],
    template=moderator_template
)

moderator_llm = assistant_llm
#We build the chain for the moderator.
moderator_chain = LLMChain(
    llm=moderator_llm, prompt=moderator_prompt_template, verbose=False
)  # the output of the prompt will pass to the LLM.
     

# To run our chain we use the .run() command
moderator_says = moderator_chain.invoke({"comment_to_moderate": assistant_response})

print(f"moderator_says: {moderator_says}")

#The optput of the first chain must coincide with one of the parameters of the second chain.
#The parameter is defined in the prompt_template.
assistant_chain = LLMChain(
    llm=assistant_llm,
    prompt=assistant_prompt_template,
    output_key="comment_to_moderate",
    verbose=False,
)

#verbose True because we want to see the intermediate messages.
moderator_chain = LLMChain(
    llm=moderator_llm,
    prompt=moderator_prompt_template,
    verbose=True
)

from langchain.chains import SequentialChain

# Creating the SequentialChain class indicating chains and parameters.
assistant_moderated_chain = SequentialChain(
    chains=[assistant_chain, moderator_chain],
    input_variables=["sentiment", "customer_request"],
    verbose=True,
)

# We can now run the chain.
assistant_moderated_chain.run({"sentiment": "rude", "customer_request": customer_request})