Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,102 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import gradio as gr
|
| 3 |
+
from langchain.agents import initialize_agent,AgentType
|
| 4 |
+
from langchain.chat_models import AzureChatOpenAI
|
| 5 |
+
from langchain.chains.conversation.memory import ConversationBufferWindowMemory
|
| 6 |
+
import torch
|
| 7 |
+
from transformers import BlipProcessor,BlipForConditionalGeneration
|
| 8 |
+
import requests
|
| 9 |
+
from PIL import Image
|
| 10 |
+
from langchain.tools import BaseTool
|
| 11 |
+
from langchain.chains import LLMChain
|
| 12 |
+
from langchain import PromptTemplate, FewShotPromptTemplate
|
| 13 |
+
|
| 14 |
+
OPENAI_API_KEY=os.getenv("OPENAI_API_KEY")
|
| 15 |
+
OPENAI_API_BASE=os.getenv("OPENAI_API_BASE")
|
| 16 |
+
DEP_NAME=os.getenv("deployment_name")
|
| 17 |
+
llm=AzureChatOpenAI(deployment_name=DEP_NAME,openai_api_base=OPENAI_API_BASE,openai_api_key=OPENAI_API_KEY,openai_api_version="2023-03-15-preview",model_name="gpt-3.5-turbo")
|
| 18 |
+
|
| 19 |
+
image_to_text_model="Salesforce/blip-image-captioning-large"
|
| 20 |
+
device= 'cuda' if torch.cuda.is_available() else 'cpu'
|
| 21 |
+
|
| 22 |
+
processor=BlipProcessor.from_pretrained(image_to_text_model)
|
| 23 |
+
|
| 24 |
+
model=BlipForConditionalGeneration.from_pretrained(image_to_text_model).to(device)
|
| 25 |
+
|
| 26 |
+
def descImage(image_url):
|
| 27 |
+
image_obj=Image.open(image_url).convert('RGB')
|
| 28 |
+
inputs=processor(image_obj,return_tensors='pt').to(device)
|
| 29 |
+
outputs=model.generate(**inputs)
|
| 30 |
+
return processor.decode(outputs[0],skip_special_tokens=True)
|
| 31 |
+
|
| 32 |
+
def toChinese(en:str):
|
| 33 |
+
pp="翻译下面语句到中文\n{en}"
|
| 34 |
+
prompt = PromptTemplate(
|
| 35 |
+
input_variables=["en"],
|
| 36 |
+
template=pp
|
| 37 |
+
)
|
| 38 |
+
llchain=LLMChain(llm=llm,prompt=prompt)
|
| 39 |
+
return llchain.run(en)
|
| 40 |
+
|
| 41 |
+
class DescTool(BaseTool):
|
| 42 |
+
name="Describe Image Tool"
|
| 43 |
+
description="use this tool to describe an image"
|
| 44 |
+
|
| 45 |
+
def _run(self,url:str):
|
| 46 |
+
description=descImage(url)
|
| 47 |
+
return description
|
| 48 |
+
def _arun(
|
| 49 |
+
self,query:str):
|
| 50 |
+
raise NotImplementedError('未实现')
|
| 51 |
+
|
| 52 |
+
tools=[DescTool()]
|
| 53 |
+
memory=ConversationBufferWindowMemory(
|
| 54 |
+
memory_key='chat_history',
|
| 55 |
+
k=5,
|
| 56 |
+
return_messages=True
|
| 57 |
+
)
|
| 58 |
+
|
| 59 |
+
agent=initialize_agent(
|
| 60 |
+
agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION,
|
| 61 |
+
tools=tools,
|
| 62 |
+
llm=llm,
|
| 63 |
+
verbose=False,
|
| 64 |
+
max_iterations=3,
|
| 65 |
+
early_stopping_method='generate',
|
| 66 |
+
memory=memory
|
| 67 |
+
)
|
| 68 |
+
def reset_user_input():
|
| 69 |
+
return gr.update(value='')
|
| 70 |
+
def reset_state():
|
| 71 |
+
return [], []
|
| 72 |
+
|
| 73 |
+
def predict(file,input, chatbot,history):
|
| 74 |
+
input1=f""+input+"\n"+file
|
| 75 |
+
out=agent(input1)
|
| 76 |
+
anws=toChinese(out['output'])
|
| 77 |
+
chatbot.append(input)
|
| 78 |
+
chatbot[-1] = (input, anws)
|
| 79 |
+
yield chatbot, history
|
| 80 |
+
return
|
| 81 |
+
|
| 82 |
+
with gr.Blocks(css=".chat-blocks{height:calc(100vh - 332px);} .mychat{flex:1} .mychat .block{min-height:100%} .mychat .block .wrap{max-height: calc(100vh - 330px);} .myinput{flex:initial !important;min-height:180px}") as demo:
|
| 83 |
+
title = '图像识别'
|
| 84 |
+
demo.title=title
|
| 85 |
+
with gr.Column(elem_classes="chat-blocks"):
|
| 86 |
+
with gr.Row(elem_classes="mychat"):
|
| 87 |
+
file = gr.Image(type="filepath")
|
| 88 |
+
chatbot = gr.Chatbot(label="图像识别", show_label=False)
|
| 89 |
+
with gr.Column(elem_classes="myinput"):
|
| 90 |
+
user_input = gr.Textbox(show_label=False, placeholder="请输入...", lines=1).style(
|
| 91 |
+
container=False)
|
| 92 |
+
submitBtn = gr.Button("提交", variant="primary", elem_classes="btn1")
|
| 93 |
+
emptyBtn = gr.Button("清除历史").style(container=False)
|
| 94 |
+
|
| 95 |
+
history = gr.State([])
|
| 96 |
+
submitBtn.click(predict, [file,user_input, chatbot,history], [chatbot, history],
|
| 97 |
+
show_progress=True)
|
| 98 |
+
submitBtn.click(reset_user_input, [], [user_input])
|
| 99 |
+
emptyBtn.click(reset_state, outputs=[chatbot, history], show_progress=True)
|
| 100 |
+
|
| 101 |
+
|
| 102 |
+
demo.queue(api_open=False,concurrency_count=20).launch()
|