Update app.py

app.py

@@ -115,14 +115,14 @@ def gpu_predict(model_path, device_map, torch_dtype,
         # 兼容原逻辑：提取 think> 之后的内容
         try:
             assert output_res.count("think>") == 2
-            new_prompt = output_res.split("think>")[-1]
-            if new_prompt.startswith("\n"):
+            new_prompt_ret = output_res.split("think>")[-1]
+            if new_prompt_ret.startswith("\n"):
                 new_prompt = new_prompt[1:]
             new_prompt = replace_single_quotes(new_prompt)
         except Exception:
             # 如果格式不符合预期，则直接回退为原始输入
-            new_prompt = org_prompt_cot
-        return new_prompt, ""
+            new_prompt_ret = org_prompt_cot
+        return new_prompt_ret, ""
     except Exception as e:
         # 失败则返回原始提示词和错误信息
         return org_prompt_cot, f"推理失败：{e}"
@@ -198,7 +198,7 @@ with gr.Blocks(title="Prompt Enhancer_V2") as demo:
             )
             with gr.Row():
                 temperature = gr.Slider(0, 1, value=0.1, step=0.05, label="Temperature")
-                max_new_tokens = gr.Slider(16, 4096, value=2048, step=16, label="Max New Tokens")
+                max_new_tokens = gr.Slider(16, 16384, value=4096, step=16, label="Max New Tokens")
                 device = gr.Dropdown(choices=["cuda", "cpu"], value="cuda", label="推理device")
 
     state = gr.State(value=None)