feat: update duration

app.py

@@ -114,7 +114,7 @@ def get_gpu_duration(audio_input, text_input, target_text, task_type, task_info)
 
     if common_tts_engine is None:
         # First call - need time for model loading (up to 5 minutes)
-        return 300  # Maximum allowed duration for model initialization
+        return 120  # Maximum allowed duration for model initialization
     else:
         # Subsequent calls - only inference time needed
         return 120  # Standard inference duration

model_loader.py

@@ -104,7 +104,6 @@ class UnifiedModelLoader:
         modelscope_patterns = []
         return any(pattern in model_path for pattern in modelscope_patterns)
 
-    @spaces.GPU
     def load_transformers_model(
         self,
         model_path: str,
@@ -189,7 +188,6 @@ class UnifiedModelLoader:
             self.logger.error(f"Failed to load model from {source}: {e}")
             raise
 
-    @spaces.GPU
     def load_funasr_model(
         self,
         repo_path: str,