Update app.py

app.py

@@ -16,21 +16,17 @@ def respond(message, history):
     if model is None:
         model = AutoModelForCausalLM.from_pretrained(
             MODEL_ID,
-            dtype=torch.float16, # 修复弃用警告：torch_dtype -> dtype
+            dtype=torch.float16, 
             trust_remote_code=True,
         ).to("cuda")
 
-    # 更加鲁棒的 history 处理逻辑
+    # 针对旧版 Gradio 结构处理：history 是 [[q1, a1], [q2, a2]]
     messages = [{"role": "system", "content": "You are Spartacus, a helpful assistant."}]
     
-    for entry in history:
-        # 自动兼容 字典格式 {'role': '...', 'content': '...'}
-        if isinstance(entry, dict):
-            messages.append(entry)
-        # 自动兼容 元组/列表格式 [user, assistant]
-        elif isinstance(entry, (list, tuple)) and len(entry) == 2:
-            messages.append({"role": "user", "content": entry[0]})
-            messages.append({"role": "assistant", "content": entry[1]})
+    if history:
+        for user_msg, assistant_msg in history:
+            messages.append({"role": "user", "content": user_msg})
+            messages.append({"role": "assistant", "content": assistant_msg})
     
     messages.append({"role": "user", "content": message})
 
@@ -43,7 +39,7 @@ def respond(message, history):
     generate_kwargs = dict(
         input_ids=input_ids,
         streamer=streamer,
-        max_new_tokens=1024,
+        max_new_tokens=2048,
         temperature=0.5,
         top_p=0.9,
         do_sample=True,
@@ -57,10 +53,9 @@ def respond(message, history):
         response += token
         yield response
 
-# 强制指定 type="tuples" 来适配上面的 history 处理，避免 unpacked 错误
+# 彻底删掉 type 参数，只保留最基础的配置
 demo = gr.ChatInterface(
     fn=respond,
-    type="tuples", # 显式声明，确保 entry 是 [user, bot] 结构
     title="Spartacus Chat",
     description="Chat with NoesisLab/Spartacus-1B-Instruct",
 )