Update app.py

app.py

@@ -36,7 +36,6 @@ model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 )
 processor = AutoProcessor.from_pretrained(MODEL_DIR)
 
-# --------- Chat Inference Function ---------
 def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_tokens: int = 1024):
     # —— 原有多模态输入构造 —— #
     messages = [
@@ -63,7 +62,7 @@ def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_
     # 1. 构造 streamer，用 processor.tokenizer（AutoProcessor 内部自带 tokenizer）
     streamer = TextIteratorStreamer(
         processor.tokenizer,
-        timeout=10.0,
+        timeout=100.0,
         skip_prompt=True,
         skip_special_tokens=True
     )
@@ -75,9 +74,8 @@ def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_
         top_k=1024,
         max_new_tokens=max_new_tokens,
         temperature=temperature,
-        top_p=0.1,
-        eos_token_id=terminators,  # 你的结束符 ID 列表
-    )
+        top_p=0.1
+        )
     # 如果需要零温度贪心，则关闭采样
     if gen_kwargs["temperature"] == 0:
         gen_kwargs["do_sample"] = False
@@ -92,6 +90,7 @@ def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_
         # 每次拿到新片段就拼接并输出
         yield "".join(buffer)
         
+        
 # --------- 3D Mesh Coloring Function ---------
 def apply_gradient_color(mesh_text: str) -> str:
     """