Update modeling_qwen2_flash.py

modeling_qwen2_flash.py

@@ -1027,6 +1027,7 @@ class Qwen2Model_Flash(Qwen2PreTrainedModel):
                 (batch_size, seq_length),
                 inputs_embeds,
                 past_key_values_length,
+                sliding_window=self.config.sliding_window,
             )
         else:
             # 4d mask is passed through the layers