利用Streamlit 和 Hugging Face 创建免费AI故事机

mkdir AI-Story
cd AI-Story
python3 -m venv venv

source venv/bin/activate  #for mac
venv\Scripts\activate     #for windows users

pip install --upgrade pip
pip install transformers     # 用于与LLM的互动
pip install huggingface_hub  # hugging face library
pip install langchain 
pip install streamlit==1.24.0
pip install python-dotenv

from dotenv import load_dotenv
import os
load_dotenv()
HUGGING_FACE_API_TOKEN=os.environ.get('HUGGING_FACE_API_TOKEN')

HUGGING_FACE_API_TOKEN=" "

# AI 推理库
from huggingface_hub import InferenceClient
from langchain import HuggingFaceHub
import requests
# 配置文件
from config import (HUGGING_FACE_API_TOKEN)
import os
import datetime

python3 app.py

hfApiToken = HUGGING_FACE_API_TOKEN 
# 仅 HuggingFace Hub 推论
model_Image2Text = "Salesforce/blip-image-captioning-base"

def imageToText(url):
    from huggingface_hub import InferenceClient
    client = InferenceClient(token=hfApiToken)
    model_Image2Text = "Salesforce/blip-image-captioning-base"
    # 来自huggingface.co/tasks
    text = client.image_to_text(url,model=model_Image2Text)
    print(text)
    return text

# AI 推理库
from huggingface_hub import InferenceClient
from langchain import HuggingFaceHub
import requests
# 配置文件
from config import (HUGGING_FACE_API_TOKEN)
import os
import datetime
hfApiToken = HUGGING_FACE_API_TOKEN
# 仅 HuggingFace Hub 推论
model_Image2Text = "Salesforce/blip-image-captioning-base"
def imageToText(url):
    from huggingface_hub import InferenceClient
    client = InferenceClient(token=hfApiToken)
    # 来自huggingface.co/tasks
    text = client.image_to_text(url,model=model_Image2Text)
    print(text)
    return text

basetext = imageToText("./images/a-boy.jpg")

python3 app.py

a boy sitting on top of a pile of books

Once upon a time, there was a boy named Jack who loved to read. He would spend hours buried in a pile of books, lost in the worlds that only books could create.
One day, Jack was sitting on top of the pile, lost in a book about dragons and knights. Suddenly, he felt a tug on his book. He looked up to see a small creature trying to get his attention.
"What are you doing?" Jack asked.
The creature looked up at him with big, round eyes. "I'm trying to get your attention," it said. "I'm a bookworm too. We're both in here, aren't we?"
Jack smiled. "I suppose we are," he said. "What are you reading?"
The creature looked down at the book in its hand. "This one," it said. "It's about a brave knight who saves a princess from a dragon."
Jack nodded. "I've been reading a lot of those lately," he said.

# Langchain 到 HuggingFace 的推论
def LC_TextGeneration(model, basetext):
    from langchain import PromptTemplate, LLMChain
    os.environ["HUGGINGFACEHUB_API_TOKEN"] = hfApiToken
    llm = HuggingFaceHub(repo_id=model , model_kwargs={"temperature":0.45,"min_length":30, "max_length":250})
    print(f"Running repo: {model}")    
    print("Preparing template")
    template = """<human>: write a very short story about {basetext}.
    The story must be a one paragraph.
    <bot>: """
    prompt = PromptTemplate(template=template, input_variables=["basetext"])
    llm_chain = LLMChain(prompt=prompt, llm=llm)
    start = datetime.datetime.now() 
    print("Running chain...")
    story = llm_chain.run(basetext)
    stop = datetime.datetime.now()  
    elapsed = stop - start
    print(f"Executed in {elapsed}")
    print(story)
    return story

template = """"""<human>: write a very short story about {basetext}.
    The story must be a one paragraph.
    <bot>: """"""

basetext = imageToText("./images/a-boy.jpg")
model_TextGeneration="togethercomputer/RedPajama-INCITE-Chat-3B-v1"
mystory = LC_TextGeneration(model_TextGeneration, basetext)
print("="*50)
finalstory = mystory.split('\n\n')[0]
print(finalstory)

Once upon a time, there was a boy named Jack who loved to read. He would spend hours buried in a pile of books, lost in the worlds that only books could create.
One day, Jack was sitting on top of the pile, lost in a book about dragons and knights. Suddenly, he felt a tug on his book. He looked up to see a small creature trying to get his attention.
"What are you doing?" Jack asked.
The creature looked up at him with big, round eyes. "I'm trying to get your attention," it said. "I'm a bookworm too. We're both in here, aren't we?"
Jack smiled. "I suppose we are," he said. "What are you reading?"
The creature looked down at the book in its hand. "This one," it said. "It's about a brave knight who saves a princess from a dragon."
Jack nodded. "I've been reading a lot of those lately," he said.

# 使用 Streamlit 进行 HuggingFace 推理的 Python 应用程序
# AI 推理库
from huggingface_hub import InferenceClient
from langchain import HuggingFaceHub
import requests
# 内部使用
import os
import datetime
import uuid
# STREAMLIT
import streamlit as st
# 配置文件
from config import (HUGGING_FACE_API_TOKEN)
hfApiToken = HUGGING_FACE_API_TOKEN
# 只有HuggingFace Hub 推理
model_TextGeneration="togethercomputer/RedPajama-INCITE-Chat-3B-v1"
model_Image2Text = "Salesforce/blip-image-captioning-base"
model_Text2Speech="espnet/kan-bayashi_ljspeech_vits"
def imageToText(url):
    from huggingface_hub import InferenceClient
    client = InferenceClient(token=hfApiToken)
    model_Image2Text = "Salesforce/blip-image-captioning-base"
    text = client.image_to_text(url,
                                model=model_Image2Text)
    print(text)
    return text
def  text2speech(text):
  import requests
  API_URL = "https://api-inference.huggingface.co/models/espnet/kan-bayashi_ljspeech_vits"
  headers = {"Authorization": f"Bearer {hfApiToken}"}
  payloads = {
      "inputs" : "".join(text.split('\n\n'))
  }
  response = requests.post(API_URL, headers=headers, json=payloads)
  with open('audiostory.flac', 'wb') as file:
    file.write(response.content)
# Langchain 到 Hugging Face 的推理
def LC_TextGeneration(model, basetext):
    from langchain import PromptTemplate, LLMChain
    os.environ["HUGGINGFACEHUB_API_TOKEN"] = hfApiToken
    llm = HuggingFaceHub(repo_id=model , model_kwargs={"temperature":0.45,"min_length":30, "max_length":250})
    print(f"Running repo: {model}")    
    print("Preparing template")
    template = """<human>: write a very short story about {basetext}.
    The story must be a one paragraph.
    <bot>: """
    prompt = PromptTemplate(template=template, input_variables=["basetext"])
    llm_chain = LLMChain(prompt=prompt, llm=llm)
    start = datetime.datetime.now()  
    print("Running chain...")
    story = llm_chain.run(basetext)
    stop = datetime.datetime.now() 
    elapsed = stop - start
    print(f"Executed in {elapsed}")
    print(story)
    return story
def generate_uuid():
    return uuid.uuid4().hex
def main():
  st.set_page_config(page_title="Your Photo Story Creatror App", page_icon='📱')
  st.header("Turn your Photos into Amazing Audio Stories")
  st.image('./assets/banner.png', use_column_width=True)
  st.markdown("1. Select a photo from your pc\n 2. AI detect the photo description\n3. AI write a story about the photo\n4. AI generate an audio file of the story")
  image_file = st.file_uploader("Choose an image...", type=['png', 'jpg'] )
  if image_file is not None:
    print(image_file)
    bytes_data = image_file.getvalue()
    save_name = generate_uuid()
    upload_path = f"./upload/{save_name}.jpg"
    with open(upload_path, "wb") as file:
      file.write(bytes_data)
    st.image(image_file, caption="Uploaded Image...",
             use_column_width=True)
    st.warning("Generating Photo description",  icon="🤖")
    basetext = imageToText(upload_path)
    with st.expander("Photo Description"):
      st.write(basetext)    
    st.warning("Generating Photo Story",  icon="🤖")
    mystory = LC_TextGeneration(model_TextGeneration, basetext)
    finalstory = mystory
    print("="*50)
    with st.expander("Photo Story"):
      st.write(finalstory)    
    st.warning("Generating Audio Story",  icon="🤖")
    text2speech(finalstory)
    st.audio('audiostory.flac')
    st.success("Audio Story completed!")
if __name__ == '__main__':
   main()

streamlit run main.py