update sth. at 2024-11-15 18:37:03

app.py

@@ -1,12 +1,10 @@
 import gradio as gr
 from huggingface_hub import InferenceClient
 from PIL import Image
-import requests
-import os
 
 # ===================== 核心逻辑模块 =====================
 
-# 初始化所需的模型客户端
+# 初始化模型客户端
 try:
     # 文本聊天模型
     client_text = InferenceClient("meta-llama/Llama-3.2-11B-Vision-Instruct")
@@ -52,14 +50,19 @@ def visual_qa(image, question):
     调用视觉文档问答模型回答图像问题。
     """
     try:
-        # 如果输入是路径或URL，则直接加载
-        if isinstance(image, str):
-            response = client_vqa.visual_question_answering(image=image, question=question)
+        # 检查输入是否为URL路径或本地路径
+        if isinstance(image, str) and image.startswith("http"):
+            # 如果是网络路径，直接使用
+            image_path = image
         else:
-            # 将本地图像保存为临时文件以供模型使用
-            temp_image_path = f"/tmp/{os.path.basename(image.filename)}"
-            image.save(temp_image_path)
-            response = client_vqa.visual_question_answering(image=temp_image_path, question=question)
+            # 如果是本地图像，获取Gradio的文件路径
+            image_path = image.name  # Gradio会传递上传文件的路径
+
+        # 调用视觉问答API
+        response = client_vqa.visual_question_answering(
+            image=image_path,
+            question=question
+        )
         return response["answer"]
     except Exception as e:
         print(f"Visual QA failed: {e}")