diff --git a/vllm/model_executor/models/qwen2.py b/vllm/model_executor/models/qwen2.py
index 46e56c4272a69..9ccf0d4fd8ca0 100644
--- a/vllm/model_executor/models/qwen2.py
+++ b/vllm/model_executor/models/qwen2.py
@@ -1,6 +1,7 @@
 # coding=utf-8
 # Adapted from
 # https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/qwen2/modeling_qwen2.py
+# Copyright 2024 Intel Corporation
 # Copyright 2024 The Qwen team.
 # Copyright 2023 The vLLM team.
 # Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
@@ -448,3 +449,6 @@ def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
                            if self.config.tie_word_embeddings else None),
         )
         loader.load_weights(weights)
+
+        if current_platform.is_hpu():
+            torch.hpu.synchronize()