diff --git a/python/llm/src/bigdl/llm/transformers/convert.py b/python/llm/src/bigdl/llm/transformers/convert.py
index 7c1aaec8..882683d0 100644
--- a/python/llm/src/bigdl/llm/transformers/convert.py
+++ b/python/llm/src/bigdl/llm/transformers/convert.py
@@ -546,7 +546,8 @@ def ggml_convert_low_bit(model, qtype, optimize_model=True,
     _enable_ipex = os.getenv("BIGDL_OPT_IPEX")
     _enable_ipex = (_enable_ipex is not None) and (_enable_ipex.lower() == "true")
     _enable_ipex = _enable_ipex and (qtype == ggml_tensor_qtype["bf16"])
-    logger.info(f"BIGDL_OPT_IPEX: {_enable_ipex}")
+    if (device == "cpu") and (qtype == ggml_tensor_qtype["bf16"]):
+        logger.info(f"BIGDL_OPT_IPEX: {_enable_ipex}")
     if _enable_ipex:
         model = _optimize_ipex(model)
         return model