修改app.py

app.py

@@ -4,7 +4,6 @@ from langchain.agents import load_tools
 from langchain.agents import initialize_agent
 from langchain.agents import AgentType
 from langchain.llms import OpenAI
-from langchain.chat_models import AzureChatOpenAI
 from langchain.chat_models import ChatOpenAI
 from langchain.chains.conversation.memory import ConversationBufferWindowMemory
 from transformers import BlipProcessor,BlipForConditionalGeneration
@@ -13,15 +12,20 @@ import requests
 from PIL import Image
 from langchain.tools import BaseTool
 import gradio as gr
+from translate import Translator
+from langchain import PromptTemplate, FewShotPromptTemplate, LLMChain
 
-OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+#OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+OPENAI_API_KEY = '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'
 # OPENAI_API_BASE = os.getenv("OPENAI_AI_BASE")
 # DEP_NAME = os.getenv("deployment name")
 
 llm = ChatOpenAI(openai_api_key=OPENAI_API_KEY, temperature=0, model_name='gpt-3.5-turbo')
 # 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")
 
-image_to_text_model = "Salesforce/blip-image-captioning-large"
+# image_to_text_model = "Salesforce/blip-image-captioning-large"
+image_to_text_model = "F:\\code\\Anaconda\\blip-image-captioning-large"
+
 device = 'cuda' if torch.cuda.is_available() else 'cpu'
 
 processor = BlipProcessor.from_pretrained(image_to_text_model)
@@ -34,7 +38,7 @@ def describeImage(image_url):
   outputs = model.generate(**inputs)
   return processor.decode(outputs[0], skip_special_tokens=True)
 
-img_url = 'https://img1.baidu.com/it/u=1919509102,1927615551&fm=253&fmt=auto&app=120&f=JPEG?w=889&h=500'
+#img_url = 'https://img1.baidu.com/it/u=1919509102,1927615551&fm=253&fmt=auto&app=120&f=JPEG?w=889&h=500'
 # description = describeImage(img_url)
 # print(description)
 
@@ -68,21 +72,39 @@ agent = initialize_agent(
     )
 )
 
-def descImage(req , image_url) :
-    output = agent(f"{req}:\n{image_url}")
+# 翻译
+def to_chinese(title):
+    pp = "翻译下面语句到中文\n{en}"
+    prompt = PromptTemplate(
+        input_variables=["en"],
+        template=pp
+    )
+    llchain = LLMChain(llm=llm, prompt=prompt)
+    return llchain.run(title)
+
+
+
+def descImage(input_text , image_url) :
+    output = agent(f"{input_text}:\n{image_url}")
     print( output )
     desc = output['output']
-    return desc
+    # 对描述文字进行转换
+    desc_ch = to_chinese(desc)
+    return desc_ch
+
+# ret = descImage("描述该图片","https://lmg.jj20.com/up/allimg/4k/s/02/2109250006343S5-0-lp.jpg")
+# print(ret)
 
-desc = descImage('描述该图片' , img_url)
-# print(desc)
+with gr.Blocks() as demo:
+    with gr.Column():
+        file = gr.Image(type='filepath')
+        user_input = gr.Textbox(show_label=False,placeholder="请输入问题",lines=1)
+    with gr.Column():
+        submitBtn = gr.Button("提交",variant="primary")
 
-output = agent(f"将该段翻译为中文：{desc}")
-print(output['output'])
+    with gr.Column():
+        output = gr.TextArea(show_label=False,placeholder="输出结果",lines=5)
 
-def greet(name):
-    return "Hello " + name + "!!"
+    submitBtn.click(descImage,[user_input,file],output,show_progress=True)
 
-if __name__ == "__main__":
-    iface = gr.Interface(fn=descImage,inputs="text",outputs="text")
-    iface.launch()
+demo.launch()