diff --git a/fastdeploy/model_executor/layers/quantization/kv_cache.py b/fastdeploy/model_executor/layers/quantization/kv_cache.py
index cd461fde799..2ccc06f0e45 100644
--- a/fastdeploy/model_executor/layers/quantization/kv_cache.py
+++ b/fastdeploy/model_executor/layers/quantization/kv_cache.py
@@ -263,10 +263,11 @@ def process_weights_after_loading(self, layer: nn.Layer):
         """
         use for loader v1
         """
-        if layer.cache_k_scale._is_initialized():
-            layer.cache_k_out_scale.set_value(1 / layer.cache_k_scale)
-        if layer.cache_v_scale._is_initialized():
-            layer.cache_v_out_scale.set_value(1 / layer.cache_v_scale)
+        if "block_wise" not in layer.cache_quant_type_str:
+            if layer.cache_k_scale._is_initialized():
+                layer.cache_k_out_scale.set_value(1 / layer.cache_k_scale)
+            if layer.cache_v_scale._is_initialized():
+                layer.cache_v_out_scale.set_value(1 / layer.cache_v_scale)
 
     def apply(self, layer):
         """