Update app.py

app.py

@@ -11,12 +11,12 @@ DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 model = OratorTTS.from_pretrained(DEVICE)
 
 def generate(text, audio_prompt_path, exaggeration, pace, temperature, seed_num):
-
-    wav = model.generate(
-        text,
-        audio_prompt_path=audio_prompt_path,
-        emotion_adv=exaggeration,
-    )
+    with torch.inference_mode():
+        wav = model.generate(
+            text,
+            audio_prompt_path=audio_prompt_path,
+            emotion_adv=exaggeration,
+        )
     return model.sr, wav.squeeze(0).numpy()