Update app.py

app.py

@@ -1,38 +1,39 @@
 import gradio as gr
-from transformers import AutoProcessor, AutoModelForCausalLM
+from transformers import AutoTokenizer, AutoModelForCausalLM, AutoImageProcessor
 import torch
+from PIL import Image
 import os
 from huggingface_hub import login
 
-# ✅ 设置你的 Hugging Face Token
-HF_TOKEN = os.environ.get("HF_TOKEN")# ← 这里替换为你的 token
+# ✅ 登入 Token（注意，不要寫死 token，請用 Secrets）
+HF_TOKEN = os.environ.get("HF_TOKEN")
 login(token=HF_TOKEN)
 
-# ✅ 指定模型
+# ✅ 模型與處理器
 MODEL_ID = "Qwen/Qwen-VL-Chat"
 
-# 加载模型
-processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True, token=HF_TOKEN)
-model = AutoModelForCausalLM.from_pretrained(MODEL_ID, trust_remote_code=True, device_map="auto", token=HF_TOKEN)
-model.eval()
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True, token=HF_TOKEN)
+image_processor = AutoImageProcessor.from_pretrained(MODEL_ID, trust_remote_code=True, token=HF_TOKEN)
+model = AutoModelForCausalLM.from_pretrained(MODEL_ID, trust_remote_code=True, token=HF_TOKEN).eval()
 
-# 推理函数
+# ✅ 推理函數
 def ask(image, prompt):
-    inputs = processor.from_list_format([
-        {"image": image},
-        {"text": prompt}
-    ])
-    inputs = processor(inputs, return_tensors="pt").to(model.device)
-    outputs = model.generate(**inputs, max_new_tokens=512)
-    response = processor.batch_decode(outputs, skip_special_tokens=True)[0]
+    image_tensor = image_processor(image, return_tensors="pt")["pixel_values"].to(model.device)
+    text_input = tokenizer(prompt, return_tensors="pt").to(model.device)
+    inputs = {
+        "input_ids": text_input["input_ids"],
+        "pixel_values": image_tensor
+    }
+    output = model.generate(**inputs, max_new_tokens=512)
+    response = tokenizer.decode(output[0], skip_special_tokens=True)
     return response
 
-# Gradio 页面
+# ✅ Gradio UI
 demo = gr.Interface(
     fn=ask,
-    inputs=[gr.Image(type="pil"), gr.Textbox(label="请输入问题")],
+    inputs=[gr.Image(type="pil"), gr.Textbox(label="請輸入問題")],
     outputs="text",
-    title="Qwen1.5-VL-Chat 在线体验"
+    title="🧠 Qwen-VL 圖文問答 Demo"
 )
 
 if __name__ == "__main__":