Update app.py

app.py

@@ -14,7 +14,7 @@ processor = AutoProcessor.from_pretrained("suno/bark-small")
 #model = BetterTransformer.transform(model, keep_original_model=False)
 
 # Enable CPU offload
-model.enable_cpu_offload()
+#model.enable_cpu_offload()
 
 import numpy as np
 from scipy.io import wavfile
@@ -23,6 +23,7 @@ from scipy.io import wavfile
 def infer(text_prompt):
     text_prompt = text_prompt
     inputs = processor(text_prompt).to(device)
+    
     with torch.inference_mode():
         speech_output = model.generate(**inputs, do_sample = True, fine_temperature = 0.4, coarse_temperature = 0.8)
         audio_out = speech_output[0].cpu().numpy()