Update app.py

app.py

@@ -21,12 +21,16 @@ snapshot_download(repo_id="jzq11111/mooncast", local_dir='./resources/')
 from inference import Model
 import base64
 
-model = Model()
-model.generate_config.max_new_tokens = 50 * 50 # no more than 20s per turn
+# model = Model()
+# model.generate_config.max_new_tokens = 50 * 50 # no more than 20s per turn
+model = None
 
 @spaces.GPU(duration=120)
 def process_json_and_generate_audio(prompt_audio_role0_file, prompt_text_role0, prompt_audio_role1_file, prompt_text_role1, json_dialogue_input_str):
     try:
+        if model is None:
+            model = Model()
+            model.generate_config.max_new_tokens = 50 * 50 # no more than 20s per turn
         print(json_dialogue_input_str, type(json_dialogue_input_str))
         print(prompt_audio_role0_file, prompt_text_role0, prompt_audio_role1_file, prompt_text_role1)
         # json_data = json.loads(json_dialogue_input_str)