Update app.py

app.py

@@ -57,19 +57,15 @@ def process_audio_input(audio):
         print(f"Error processing audio: {e}")
         return None
 
-@spaces.GPU(duration=180)  # 使用ZeroGPU，3分钟超时
-def process_audio_text(audio):
-    """主要处理函数"""
+@spaces.GPU(duration=120)
+def process_audio_text(text, audio):
     global _MODEL_ON_CUDA, inference_model
     
-    # 初始化模型（如果还没初始化）
     if inference_model is None:
         init_model()
     
-    # 首次使用GPU时移动模型
     if not _MODEL_ON_CUDA:
         try:
-            # 将模型移动到GPU
             if hasattr(inference_model, 'model'):
                 inference_model.model = inference_model.model.to("cuda")
             if hasattr(inference_model, 'unit_translator'):
@@ -81,11 +77,8 @@ def process_audio_text(audio):
         except Exception as e:
             print(f"Error moving model to GPU: {e}")
     
-    # 处理音频输入
     audio_path = process_audio_input(audio)
     
-    text = ""
-    
     tmp = [{
         "conversations": [
             {
@@ -111,26 +104,25 @@ def process_audio_text(audio):
     except Exception as e:
         yield None, f"Error: {str(e)}"
     finally:
-        # 清理临时文件
         if audio_path and audio_path != audio and os.path.exists(audio_path):
             try:
                 os.unlink(audio_path)
             except:
                 pass
 
-# 初始化模型（在CPU上）
 init_model()
 
 if __name__ == "__main__":
     examples = [
-        ["./show_case/1.wav"],
-        ["./show_case/2.wav"],
+        ["Provide a short answer.", "./show_case/1.wav"],
+        ["", "./show_case/2.wav"],
     ]  
 
     iface = gr.Interface(
         fn=process_audio_text,
         inputs=[
-            gr.Audio(type="filepath", label="Upload Audio")
+            gr.Textbox(label="Input Text", value=examples[0][0]),
+            gr.Audio(type="filepath", label="Upload Audio", value=examples[0][1])
         ],
         outputs=[
             gr.Audio(label="Streamed Audio", streaming=True, autoplay=True),