diff --git a/python/paddle/fluid/optimizer.py b/python/paddle/fluid/optimizer.py
index e1051db52b41a90418bcfb2f14ae1db73853d438..6a97ab0cbd87da9395f1b0a458f1264a66e32d89 100755
--- a/python/paddle/fluid/optimizer.py
+++ b/python/paddle/fluid/optimizer.py
@@ -2809,7 +2809,7 @@ class AdamaxOptimizer(Optimizer):
             with param.block.program._optimized_guard(
                 [param, grad]
             ), name_scope('adamx'):
-                beta1_pow_acc = self._get_accumulator(
+                beta1_pow_acc = self._get_accumulator_master(
                     self._beta1_pow_acc_str, param
                 )
                 if in_dygraph_mode():
diff --git a/python/paddle/fluid/tests/unittests/test_adamax_op.py b/python/paddle/fluid/tests/unittests/test_adamax_op.py
index 7d4c2849bdc8f4f7d71a4ceadbc8dc998fcc2f93..8acad2b4bfb4896d35ba06dc767a815a6789c03b 100644
--- a/python/paddle/fluid/tests/unittests/test_adamax_op.py
+++ b/python/paddle/fluid/tests/unittests/test_adamax_op.py
@@ -31,6 +31,7 @@ def adamx_wrapper(
     beta1=0.78,
     beta2=0.899,
     epsilon=1e-5,
+    find_master=False,
 ):
     return paddle._C_ops.adamax_(
         param,
@@ -43,6 +44,7 @@ def adamx_wrapper(
         beta1,
         beta2,
         epsilon,
+        find_master,
     )