Update app.py

app.py

@@ -1,107 +1,97 @@
 # -*- coding: utf-8 -*-
 """
-鸟类知识科普系统（Qwen3优化版） by [你的名字]
-ISOM5240 Group Project
+鸟类知识智能科普系统
 """
-import transformers
-import gradio as gr
-from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer
+
+import streamlit as st
 from PIL import Image
+import tempfile
+from transformers import pipeline, AutoConfig
 import torch
 
+# ========== 模型配置 ==========
+MODEL_CONFIG = {
+    "image_to_text": {
+        "model": "chriamue/bird-species-classifier",
+        "config": {"use_fast": True}  # 强制启用快速处理器
+    },
+    "text_generation": {
+        "model": "Qwen/Qwen-7B-Chat",
+        "config": AutoConfig.from_pretrained("Qwen/Qwen-7B-Chat", revision="main")
+    },
+    "text_to_speech": {
+        "model": "facebook/mms-tts-eng",
+        "config": {"speaker_id": 6}  # 儿童音色
+    }
+}
 
-
-# 初始化模型（兼容Qwen3）
-def init_models():
-    # 鸟类分类模型（保持不变）
-    classifier = pipeline(
-        "image-classification",
-        model="chriamue/bird-species-classifier",
-        device=0 if torch.cuda.is_available() else -1
-    )
-    
-    # 更新为Qwen3模型（官方支持版本）
-    text_generator = pipeline(
-        "text-generation",
-        model="Qwen/Qwen-7B-Chat",  # 使用官方维护版本
-        device_map="auto",
-        torch_dtype=torch.bfloat16,
-        trust_remote_code=True,     # 必须开启
-        model_kwargs={
-            "revision": "main",
-            "force_download": True  # 替换弃用参数
-        }
-    )
-    
-    # 语音合成模型（保持不变）
-    tts = pipeline(
-        "text-to-speech",
-        model="facebook/mms-tts-eng",
-        device=0 if torch.cuda.is_available() else -1
-    )
-    
-    return classifier, text_generator, tts
-
-# 生成儿童友好的鸟类描述
-def generate_child_friendly_text(bird_name):
-    PROMPT = f"""以6-12岁儿童能理解的方式描述{bird_name}：
-1. 用比喻手法（如：羽毛像彩虹糖纸）
-2. 包含一个趣味冷知识（例如：每天吃相当于自身体重30%的食物）
-3. 语句长度不超过15个英文单词
-4. 避免使用专业术语"""
-    
-    response = text_generator(
-        PROMPT,
-        max_new_tokens=150,
-        temperature=0.7,
-        do_sample=True
-    )
-    
-    return response[0]['generated_text'].split('\n')[2]
-
-# 主处理流程
-def process_image(image):
+# ========== 模型初始化 ==========
+@st.cache_resource
+def init_pipelines():
+    """缓存模型加载结果避免重复初始化"""
     try:
-        classification = classifier(image)
-        bird_name = classification[0]['label']
-        description = generate_child_friendly_text(bird_name)
-        speech = tts(description, forward_params={"speaker_id": 6})
+        img_pipeline = pipeline(
+            "image-classification", 
+            model=MODEL_CONFIG["image_to_text"]["model"],
+            **MODEL_CONFIG["image_to_text"]["config"]
+        )
+        
+        text_pipeline = pipeline(
+            "text-generation",
+            model=MODEL_CONFIG["text_generation"]["model"],
+            config=MODEL_CONFIG["text_generation"]["config"],
+            torch_dtype=torch.bfloat16,
+            device_map="auto"
+        )
         
-        return {
-            "bird_name": bird_name,
-            "description": description,
-            "audio": speech["audio"]
-        }
+        tts_pipeline = pipeline(
+            "text-to-speech",
+            model=MODEL_CONFIG["text_to_speech"]["model"],
+            **MODEL_CONFIG["text_to_speech"]["config"]
+        )
+        
+        return img_pipeline, text_pipeline, tts_pipeline
+    
     except Exception as e:
-        return f"处理错误: {str(e)}"
+        st.error(f"模型加载失败: {str(e)}")
+        st.stop()
 
-# 初始化模型
-classifier, text_generator, tts = init_models()
+# ========== 核心功能 ==========
+def generate_description(_pipe, bird_name):
+    """生成儿童友好型描述"""
+    prompt = f"用6-12岁儿童能理解的语言描述{bird_name}，使用比喻和趣味知识："
+    return _pipe(prompt, max_new_tokens=120)[0]['generated_text'].split("：")[-1]
 
-# 创建Gradio界面
-with gr.Blocks(theme=gr.themes.Soft(), css=".gradio-container {max-width: 800px}") as demo:
-    gr.Markdown("# 🐦 鸟类知识小课堂（Qwen3版）")
-    
-    with gr.Row():
-        image_input = gr.Image(type="pil", label="上传鸟类图片", height=300)
-        audio_output = gr.Audio(label="语音讲解", autoplay=True)
-    
-    with gr.Column():
-        name_output = gr.Textbox(label="识别到的鸟类")
-        text_output = gr.Textbox(label="趣味知识", lines=4)
+# ========== 界面设计 ==========
+st.set_page_config(page_title="鸟类知识百科", page_icon="🐦")
+st.title("🐦 智能鸟类科普系统")
+st.markdown("上传鸟类图片，获取趣味知识讲解")
+
+# 主流程
+def main():
+    img_pipe, text_pipe, tts_pipe = init_pipelines()
     
-    examples = gr.Examples(
-        examples=["eagle.jpg", "penguin.jpg", "peacock.jpg"],
-        inputs=image_input,
-        label="示例图片"
-    )
+    uploaded_file = st.file_uploader("选择图片文件", type=["jpg", "png", "jpeg"])
     
-    image_input.change(
-        process_image,
-        inputs=image_input,
-        outputs=[name_output, text_output, audio_output]
-    )
+    if uploaded_file:
+        with tempfile.NamedTemporaryFile(suffix=".jpg") as tmp_file:
+            # 保存临时文件
+            tmp_file.write(uploaded_file.getvalue())
+            
+            with st.spinner("识别中..."):
+                # 识别鸟类
+                result = img_pipe(Image.open(tmp_file.name))
+                bird_name = result[0]['label']
+                st.success(f"识别结果：{bird_name}")
+                
+                # 生成描述
+                desc = generate_description(text_pipe, bird_name)
+                st.subheader("趣味知识")
+                st.write(desc)
+                
+                # 语音合成
+                audio = tts_pipe(desc[:1000])  # 限制文本长度
+                st.audio(audio["audio"], sample_rate=audio["sampling_rate"])
 
-# 部署配置
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)
+    main()