Update app.py

app.py

@@ -10,7 +10,7 @@ if torch.cuda.is_available():
 
 @spaces.GPU
 def predict(prompt, temperature, max_length):
-    return pipe(prompt, temperature=temperature, max_length=max_length)[0]["generated_text"]
+    return pipe(prompt, temperature=temperature, max_length=max_length, top_p=0.95, top_k=50, do_sample=True)[0]["generated_text"]
 
 demo = gr.Interface(
     fn=predict,