Update app.py

app.py

@@ -3,48 +3,109 @@ from transformers import Qwen2VLForConditionalGeneration, AutoTokenizer, AutoPro
 from qwen_vl_utils import process_vision_info
 import gradio as gr
 from PIL import Image
+from huggingface_hub import login
+import os
+
+# ========== 使用你的 secret 名称 fmv 登录 ==========
+token = os.getenv("fmv")  # 读取名为 fmv 的 secret
+if token:
+    login(token=token)
+    print("成功使用 token 登录!")
+else:
+    print("警告:未找到 token")
+# ==========================================
 
 # Hugging Face 模型仓库路径
-model_path = "hiko1999/Qwen2-Wildfire-2B"  # 替换为你的模型路径
+model_path = "hiko1999/Qwen2-Wildfire-2B"
 
-# 加载 Hugging Face 上的模型和 processor
+# 加载模型和 processor
+print(f"正在加载模型: {model_path}")
 tokenizer = AutoTokenizer.from_pretrained(model_path)
-model = Qwen2VLForConditionalGeneration.from_pretrained(model_path, torch_dtype=torch.bfloat16)  # 移除 device_map 参数以避免自动分配到 GPU
+model = Qwen2VLForConditionalGeneration.from_pretrained(
+    model_path, 
+    torch_dtype=torch.bfloat16,
+    device_map="cpu"
+)
 processor = AutoProcessor.from_pretrained(model_path)
+print("模型加载完成!")
 
 # 定义预测函数
 def predict(image):
-    # 将上传的图片处理为模型需要的格式
-    messages = [{"role": "user",
-                 "content": [{"type": "image", "image": image}, {"type": "text", "text": "Describe this image."}]}]
-
-    # 处理图片输入
-    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
-    image_inputs, video_inputs = process_vision_info(messages)
-    inputs = processor(text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt")
-
-    # 将数据转移到 CPU
-    inputs = inputs.to("cpu")  # 使用 CPU 而不是 CUDA
-
-    # 生成模型输出
-    generated_ids = model.generate(**inputs, max_new_tokens=128)
-    generated_ids_trimmed = [out_ids[len(in_ids):] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)]
-    output_text = processor.batch_decode(generated_ids_trimmed, skip_special_tokens=True,
-                                         clean_up_tokenization_spaces=False)
-
-    return output_text[0]  # 返回生成的文本
+    """处理图片并生成描述"""
+    if image is None:
+        return "错误:未上传图片"
+    
+    try:
+        # 构建消息
+        messages = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "image", "image": image},
+                    {"type": "text", "text": "请描述这张图片中的火灾情况。"}
+                ]
+            }
+        ]
+        
+        # 处理输入
+        text = processor.apply_chat_template(
+            messages, 
+            tokenize=False, 
+            add_generation_prompt=True
+        )
+        image_inputs, video_inputs = process_vision_info(messages)
+        inputs = processor(
+            text=[text], 
+            images=image_inputs, 
+            videos=video_inputs, 
+            padding=True, 
+            return_tensors="pt"
+        )
+        
+        # 确保在 CPU 上运行
+        inputs = inputs.to("cpu")
+        
+        # 生成输出
+        generated_ids = model.generate(
+            **inputs, 
+            max_new_tokens=256,
+            do_sample=True,
+            temperature=0.7
+        )
+        
+        # 解码输出
+        generated_ids_trimmed = [
+            out_ids[len(in_ids):] 
+            for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
+        ]
+        output_text = processor.batch_decode(
+            generated_ids_trimmed, 
+            skip_special_tokens=True,
+            clean_up_tokenization_spaces=False
+        )
+        
+        return output_text[0]
+        
+    except Exception as e:
+        return f"预测失败: {str(e)}"
 
-# Gradio界面
+# Gradio 界面函数
 def gradio_interface(image):
+    """Gradio 界面的主函数"""
     result = predict(image)
-    return f"预测结果：{result}"
+    return result
 
-# 创建Gradio接口
-interface = gr.Interface(fn=gradio_interface,
-                         inputs=gr.Image(type="pil"),  # 输入的图像
-                         outputs="text",  # 输出结果
-                         title="火灾场景多模态模型预测",
-                         description="上传图片进行火灾预测。")
+# 创建 Gradio 界面
+interface = gr.Interface(
+    fn=gradio_interface,
+    inputs=gr.Image(type="pil", label="上传火灾图片"),
+    outputs=gr.Textbox(label="AI 分析结果", lines=10),
+    title="🔥 火灾场景智能分析系统",
+    description="上传火灾相关图片,AI 将自动分析并描述火灾情况。",
+    theme=gr.themes.Soft(),
+    allow_flagging="never"
+)
 
 # 启动接口
-interface.launch()
+if __name__ == "__main__":
+    interface.launch()