moslem-bot-be / app.py
Bofandra's picture
Update app.py
c7e3111 verified
raw
history blame
4.67 kB
import gradio as gr
from sentence_transformers import SentenceTransformer
from huggingface_hub import InferenceClient
import pandas as pd
import torch
import math
import httpcore
setattr(httpcore, 'SyncHTTPTransport', 'AsyncHTTPProxy')
"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
def get_detailed_instruct(task_description: str, query: str) -> str:
return f'Instruct: {task_description}\nQuery: {query}'
def respond(
message,
history: list[tuple[str, str]],
max_tokens = 2048,
temperature = 0.7,
top_p = 0.95,
):
#system role
messages = [{"role": "system", "content": "You are a moslem bot that always give answer based on quran and hadith!"}]
#make a moslem bot
messages.append({"role": "user", "content": "I want you to answer strictly based on quran and hadith"})
messages.append({"role": "assistant", "content": "I'd be happy to help! Please go ahead and provide the sentence you'd like me to analyze. Please specify whether you're referencing a particular verse or hadith (Prophetic tradition) from the Quran or Hadith, or if you're asking me to analyze a general statement."})
#adding references
df = pd.read_csv("moslem-bot-reference.csv")
for index, row in df.iterrows():
messages.append({"role": "user", "content": row['user']})
messages.append({"role": "assistant", "content": row['assistant']})
#adding more references
selected_references = torch.load('selected_references.sav', map_location=torch.device('cpu'))
encoded_questions = torch.load('encoded_questions.sav', map_location=torch.device('cpu'))
task = 'Given a web search query, retrieve relevant passages that answer the query'
queries = [
get_detailed_instruct(task, message)
]
model = SentenceTransformer('intfloat/multilingual-e5-large-instruct')
query_embeddings = model.encode(queries, convert_to_tensor=True, normalize_embeddings=True)
scores = (query_embeddings @ encoded_questions.T) * 100
selected_references['similarity'] = scores.tolist()[0]
sorted_references = selected_references.sort_values(by='similarity', ascending=False)
sorted_references = sorted_references.head(3)
sorted_references = selected_references.sort_values(by='similarity', ascending=True)
from googletrans import Translator
translator = Translator()
for index, row in sorted_references.iterrows():
print(index)
print(row['user'])
user = translator.translate(row['user'])
print(user)
print(row['assistant'])
assistant = translator.translate(row['assistant'])
print(assistant)
messages.append({"role": "user", "content":user })
messages.append({"role": "assistant", "content": assistant})
#history from chat session
for val in history:
if val[0]:
messages.append({"role": "user", "content": val[0]})
if val[1]:
messages.append({"role": "assistant", "content": val[1]})
#latest user question
messages.append({"role": "user", "content": message})
print(messages)
response = ""
for message in client.chat_completion(
messages,
max_tokens=max_tokens,
stream=True,
temperature=temperature,
top_p=top_p,
):
token = message.choices[0].delta.content
response += token
yield response
"""
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
"""
demo = gr.ChatInterface(
respond,
additional_inputs=[
gr.Slider(minimum=1, maximum=2048, value=2048, step=1, label="Max new tokens"),
gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
gr.Slider(
minimum=0.1,
maximum=1.0,
value=0.95,
step=0.05,
label="Top-p (nucleus sampling)",
),
],
examples=[
["Why is men created?"],
["How is life after death?"],
["Please tell me about superstition!"],
["How moses defeat pharaoh?"],
["Please tell me about inheritance law in Islam!"],
["A woman not wear hijab"],
["Worshipping God beside Allah"],
["Blindly obey a person"],
["Make profit from lending money to a friend"],
],
)
if __name__ == "__main__":
demo.launch()