diff --git a/paddleformers/mergekit/merge_model.py b/paddleformers/mergekit/merge_model.py
index ade00dc80dc..1ec9b05a03b 100644
--- a/paddleformers/mergekit/merge_model.py
+++ b/paddleformers/mergekit/merge_model.py
@@ -590,7 +590,7 @@ def shard_lora_merge(self, base_index, shard_file, lora_config, file_type_list,
                     tensor = paddle.Tensor.__call__(tensor, zero_copy=True)
                     lora_A_tensor = paddle.Tensor.__call__(lora_A_tensor, zero_copy=True)
                     lora_B_tensor = paddle.Tensor.__call__(lora_B_tensor, zero_copy=True)
-                    if self.is_cpu and is_bf16 or self.merge_config.save_to_hf:
+                    if self.is_cpu and is_bf16:
                         tensor = tensor.astype("float32")
                         lora_A_tensor = lora_A_tensor.astype("float32")
                         lora_B_tensor = lora_B_tensor.astype("float32")
diff --git a/paddleformers/transformers/model_utils.py b/paddleformers/transformers/model_utils.py
index f7fce5b0d94..a7ef574cd88 100644
--- a/paddleformers/transformers/model_utils.py
+++ b/paddleformers/transformers/model_utils.py
@@ -598,7 +598,7 @@ def load_state_dict(
 def prepare_safe_save_state_dict(state_dict, save_to_hf=False):
     for k in list(state_dict.keys()):
         if isinstance(state_dict[k], paddle.Tensor):
-            if save_to_hf:
+            if state_dict[k].dtype == paddle.bfloat16:
                 state_dict[k] = state_dict.pop(k).astype("float32").cpu().numpy().astype(ml_dtypes.bfloat16)
             else:
                 state_dict[k] = state_dict.pop(k).cpu().numpy()