diff --git a/colossalai/legacy/engine/_base_engine.py b/colossalai/legacy/engine/_base_engine.py
index a4e90b218489..9a1a2dc325a3 100644
--- a/colossalai/legacy/engine/_base_engine.py
+++ b/colossalai/legacy/engine/_base_engine.py
@@ -157,7 +157,7 @@ def step(self):
         """Execute parameter update
         """
         self._all_reduce_gradients()
-        self.optimizer.clip_grad_norm(self.model, self._clip_grad_norm)
+        self.optimizer.clip_grad_by_norm(self._clip_grad_norm)
         return self.optimizer.step()
 
     def backward(self, loss: Tensor):
diff --git a/colossalai/legacy/engine/gradient_accumulation/_gradient_accumulation.py b/colossalai/legacy/engine/gradient_accumulation/_gradient_accumulation.py
index 1f91286a6656..c2270dc53a50 100644
--- a/colossalai/legacy/engine/gradient_accumulation/_gradient_accumulation.py
+++ b/colossalai/legacy/engine/gradient_accumulation/_gradient_accumulation.py
@@ -74,7 +74,7 @@ def clip_grad_norm(self, model: nn.Module, max_norm: float) -> None:
         if self.accumulate_step < self.accumulate_size:
             pass
         else:
-            self.optim.clip_grad_norm(model, max_norm)
+            self.optim.clip_grad_by_norm(max_norm)
 
     def backward(self, loss: Tensor) -> None:
         """Execute backward pass.