Update app.py

app.py

@@ -15,6 +15,13 @@ MODELS = {
 }
 from ttts.gpt.voice_tokenizer import VoiceBpeTokenizer
 import torch.nn.functional as F
+cond_audio = 'ttts/3.wav'
+audio,sr = torchaudio.load(cond_audio)
+if audio.shape[0]>1:
+    audio = audio[0].unsqueeze(0)
+audio = torchaudio.transforms.Resample(sr, 24000)(audio)
+cond_mel = MelSpectrogramFeatures()(audio).to(device)
+print(cond_mel.shape)
 auto_conditioning = cond_mel
 settings = {'temperature': .8, 'length_penalty': 1.0, 'repetition_penalty': 2.0,
                     'top_p': .8,
@@ -44,13 +51,6 @@ def speak(text):
     print(text_tokens)
     gpt = load_model('gpt',MODELS['gpt.pth'], './ttts/gpt/config.json',device)
     gpt.post_init_gpt2_config(use_deepspeed=False, kv_cache=False, half=False)
-    cond_audio = 'ttts/3.wav'
-    audio,sr = torchaudio.load(cond_audio)
-    if audio.shape[0]>1:
-        audio = audio[0].unsqueeze(0)
-    audio = torchaudio.transforms.Resample(sr, 24000)(audio)
-    cond_mel = MelSpectrogramFeatures()(audio).to(device)
-    print(cond_mel.shape)
     codes = gpt.inference_speech(auto_conditioning, text_tokens,
                                 do_sample=True,
                                 top_p=top_p,