Update app.py

app.py

@@ -52,7 +52,7 @@ def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_
     )
     print(text)
     image_inputs, video_inputs = process_vision_info(messages)
-    inputs = processor(
+    input_ids = processor(
         text=[text],
         images=image_inputs,
         videos=video_inputs,
@@ -61,19 +61,34 @@ def chat_qwen_vl(message: str, history: list, temperature: float = 0.1, max_new_
     ).to(model.device)
 
     # 2. 把 streamer 和生成参数一起传给 model.generate
-    gen_kwargs = dict(
-        **inputs,           # 包含 input_ids, pixel_values, attention_mask 等
-        top_k=1024,
+    streamer = TextIteratorStreamer(
+        tokenizer,
+        timeout=100.0,
+        skip_prompt=True,
+        skip_special_tokens=True
+    )
+    #print(input_ids)
+    generate_kwargs = dict(
+        input_ids= input_ids["input_ids"],
+        streamer=streamer,
         max_new_tokens=max_new_tokens,
+        do_sample=True,
         temperature=temperature,
+        eos_token_id=terminators,
+        top_k=1024,
         top_p=0.1
-        )
-    generated_ids = model.generate(**gen_kwargs)
-    generated_ids_trimmed = [
-        out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)]
-    output_text = processor.batch_decode(
-        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
-    yield output_text
+    )
+
+    # 4. 后台线程启动生成
+    Thread(target=model.generate, kwargs=generate_kwargs).start()
+
+    # 5. 主线程读取 streamer 并 yield
+    buffer = []
+    for chunk in streamer:
+        print(chunk)
+        buffer.append(chunk)
+        # 每次新到一个片段，就拼接并返回给前端
+        yield "".join(buffer)
         
         
 # --------- 3D Mesh Coloring Function ---------