Added basic optimization according to model card.

app.py

@@ -1,16 +1,17 @@
 import gradio as gr
-from transformers import AutoProcessor, AutoModel
+from transformers import AutoProcessor, BarkModel
 import scipy
 
 processor = AutoProcessor.from_pretrained("suno/bark-small")
-model = AutoModel.from_pretrained("suno/bark-small")
+model = BarkModel.from_pretrained("suno/bark-small")
+model = model.to_bettertransformer()
 
 def greet(text):
     inputs = processor(
         text=[text],
         return_tensors="pt",
     )
-    speech_values = model.generate(**inputs, do_sample=True)
+    speech_values = model.generate(**inputs, do_sample=True, use_flash_attention_2=True)
     scipy.io.wavfile.write("tmp.wav", rate=24000, data=speech_values.cpu().numpy().squeeze())
     return open("tmp.wav", "rb").read()