Update app.py

app.py

@@ -29,7 +29,7 @@ def generate(model, input_ids, past_key_values, max_new_tokens):
     output_ids = input_ids.clone()
     next_token = input_ids
     with torch.no_grad():
-        for _ in range(50):
+        for _ in range(128):
             out = model(
                 input_ids=next_token,
                 past_key_values=past_key_values,