更新

app.py

@@ -4,7 +4,7 @@ from os import getenv
 from threading import Thread
 from transformers import Qwen2VLForConditionalGeneration, Qwen2VLProcessor, TextIteratorStreamer, AutoProcessor, BatchFeature
 from qwen_vl_utils import process_vision_info
-from gradio import ChatInterface, Textbox, Slider
+from gradio import ChatInterface, Slider
 
 model_path = "Pectics/Softie-VL-7B-250123"
 
@@ -47,18 +47,22 @@ def infer(
         yield response
 
 def respond(
-    message: str,
+    message: str | list[object],
     history: list[object],
     max_tokens: int,
     temperature: float,
     top_p: float,
 ):
-    if len(history) == 0 or history[0]['role'] != 'system':
+    print('message', message)
+    print('history', history)
+    if isinstance(message, str) and (len(history) == 0 or history[0]['role'] != 'system'):
         history.insert(0, {"role": "system", "content": """You are Softie, or 小软 in Chinese.
 You are an intelligent assistant developed by the School of Software at Hefei University of Technology.
 You like to chat with people and help them solve problems."""})
-    history.append({"role": "user", "content": message})
-    text_inputs = processor.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
+        history.append({"role": "user", "content": message})
+        text_inputs = processor.apply_chat_template(history, tokenize=False, add_generation_prompt=True)
+    else:
+        text_inputs = processor.apply_chat_template(message, tokenize=False, add_generation_prompt=True)
     image_inputs, video_inputs = process_vision_info(history)
     for response in infer((text_inputs, image_inputs, video_inputs), max_tokens, temperature, top_p):
         yield response