fix global vars

app.py

@@ -234,6 +234,9 @@ config = Config()
 logging.getLogger("numba").setLevel(logging.WARNING)
 
 hubert_model = None
+voicecraft_model = None
+voicecraft_config = None
+phn2num = None
 
 def load_hubert():
     global hubert_model
@@ -256,9 +259,7 @@ def load_hubert():
     # hubert_model = models[0]
 
 def load_voicecraft():
-    global voicecraft_model
-    global phn2num
-    global voicecraft_config
+    global voicecraft_model, phn2num, voicecraft_config
 
     from lib.voicecraft.models import voicecraft
     voicecraft_name = "giga330M.pth"
@@ -1493,6 +1494,8 @@ def transcribe_btn_click(audio_choice):
 
 def run(seed, stop_repetition, sample_batch_size, left_margin, right_margin, codec_audio_sr, codec_sr, top_k, top_p,
         temperature, kvcache, cutoff_value, target_transcript, silence_tokens, transcribed_text):
+    global voicecraft_model, voicecraft_config, phn2num
+    
     os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
     os.environ["CUDA_VISIBLE_DEVICES"] = "0"
     # take a look at demo/temp/mfa_alignment, decide which part of the audio to use as prompt