minor change

app.py

@@ -4,7 +4,7 @@ import os
 import sys
 import soundfile as sf
 import numpy as np
-import torch.cuda
+import torch
 
 repo_url = "https://huggingface.co/dangtr0408/StyleTTS2-lite-vi"
 repo_dir = "StyleTTS2-lite-vi"
@@ -20,7 +20,7 @@ device = 'cuda' if torch.cuda.is_available() else 'cpu'
 
 config_path = os.path.join(repo_dir, "Models", "config.yml")
 models_path = os.path.join(repo_dir, "Models", "model.pth")
-model = StyleTTS2(config_path, models_path).to(device)
+model = StyleTTS2(config_path, models_path).eval().to(device)
 
 # Core inference function
 def process_inputs(text_prompt, reference_audio_paths,
@@ -35,7 +35,8 @@ def process_inputs(text_prompt, reference_audio_paths,
             "speed": 1.1
     }
 
-    r = model.generate(text_prompt, speakers, avg_style, stabilize, denoise, n_merge, "[id_1]")
+    with torch.no_grad():
+        r = model.generate(text_prompt, speakers, avg_style, stabilize, denoise, n_merge, "[id_1]")
 
     r = r / np.abs(r).max()
     sf.write("output.wav", r, samplerate=24000)