diff --git a/paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cc b/paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cc
similarity index 92%
rename from paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cc
rename to paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cc
index fa5f996f5c1506a68f1cf8464887c6b313f98660..c737ba361e0f2573d46def53d1b566774a4bd90f 100644
--- a/paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cc
+++ b/paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cc
@@ -10,7 +10,7 @@ distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. */
-#include "paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.h"
+#include "paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.h"
 #include "paddle/fluid/framework/generator.h"
 #include "paddle/fluid/framework/op_registry.h"
 namespace paddle {
@@ -82,7 +82,7 @@ class SoftmaxMaskFuseUpperTriangleGradOpMaker
 
  protected:
   void Apply(GradOpPtr<T> op) const override {
-    op->SetType("softmax_mask_fuse_upper_triangle_grad");
+    op->SetType("fused_softmax_mask_upper_triangle_grad");
     op->SetInput("Softmax", this->Output("Out"));
     op->SetInput(framework::GradVarName("Out"), this->OutputGrad("Out"));
     op->SetOutput(framework::GradVarName("X"), this->InputGrad("X"));
@@ -94,13 +94,13 @@ class SoftmaxMaskFuseUpperTriangleGradOpMaker
 
 namespace ops = paddle::operators;
 REGISTER_OPERATOR(
-    softmax_mask_fuse_upper_triangle, ops::SoftmaxMaskFuseUpperTriangleOp,
+    fused_softmax_mask_upper_triangle, ops::SoftmaxMaskFuseUpperTriangleOp,
     ops::SoftmaxMaskFuseUpperTriangleOpMaker,
     ops::SoftmaxMaskFuseUpperTriangleGradOpMaker<paddle::framework::OpDesc>,
     ops::SoftmaxMaskFuseUpperTriangleGradOpMaker<paddle::imperative::OpBase>);
-REGISTER_OPERATOR(softmax_mask_fuse_upper_triangle_grad,
+REGISTER_OPERATOR(fused_softmax_mask_upper_triangle_grad,
                   ops::SoftmaxMaskFuseUpperTriangleOpGrad);
-REGISTER_OP_CPU_KERNEL(softmax_mask_fuse_upper_triangle,
+REGISTER_OP_CPU_KERNEL(fused_softmax_mask_upper_triangle,
                        ops::SoftmaxMaskFuseUpperTriangleCPUKernel<
                            paddle::platform::CPUDeviceContext, float>,
                        ops::SoftmaxMaskFuseUpperTriangleCPUKernel<
diff --git a/paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cu b/paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cu
similarity index 99%
rename from paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cu
rename to paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cu
index 9a1b4332e8b9f39d4ad405cce664208a8b34b4cf..3bebbee1fb7ccb0e465d84a542f214cb59ed54c6 100644
--- a/paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.cu
+++ b/paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.cu
@@ -31,7 +31,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/generator.h"
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/memory/memcpy.h"
-#include "paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.h"
+#include "paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.h"
 #include "paddle/fluid/platform/float16.h"
 
 namespace paddle {
@@ -534,12 +534,12 @@ class SoftmaxMaskFuseUpperTriangleGradKernel : public framework::OpKernel<T> {
 namespace ops = paddle::operators;
 namespace plat = paddle::platform;
 REGISTER_OP_CUDA_KERNEL(
-    softmax_mask_fuse_upper_triangle,
+    fused_softmax_mask_upper_triangle,
     ops::SoftmaxMaskFuseUpperTriangleKernel<plat::CUDADeviceContext,
                                             plat::float16>,
     ops::SoftmaxMaskFuseUpperTriangleKernel<plat::CUDADeviceContext, float>);
 REGISTER_OP_CUDA_KERNEL(
-    softmax_mask_fuse_upper_triangle_grad,
+    fused_softmax_mask_upper_triangle_grad,
     ops::SoftmaxMaskFuseUpperTriangleGradKernel<plat::CUDADeviceContext,
                                                 plat::float16>,
     ops::SoftmaxMaskFuseUpperTriangleGradKernel<plat::CUDADeviceContext,
diff --git a/paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.h b/paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.h
similarity index 100%
rename from paddle/fluid/operators/softmax_mask_fuse_upper_triangle_op.h
rename to paddle/fluid/operators/fused_softmax_mask_upper_triangle_op.h
diff --git a/python/paddle/fluid/tests/unittests/test_softmax_mask_fuse_upper_triangle_op.py b/python/paddle/fluid/tests/unittests/test_softmax_mask_fuse_upper_triangle_op.py
index a5f59c6d1f26167bb77b5314f28fcb823177e099..8b6d37882ba1a28264ee0e068a0528fc49735973 100644
--- a/python/paddle/fluid/tests/unittests/test_softmax_mask_fuse_upper_triangle_op.py
+++ b/python/paddle/fluid/tests/unittests/test_softmax_mask_fuse_upper_triangle_op.py
@@ -42,7 +42,7 @@ def _get_softmax_upper(x, fp16=True):
                  "core is not compiled with CUDA")
 class TestSoftmaxMaskFuseOp(OpTest):
     def setUp(self):
-        self.op_type = "softmax_mask_fuse_upper_triangle"
+        self.op_type = "fused_softmax_mask_upper_triangle"
         x = np.random.random((1, 1, 32, 32)).astype("float16")
         self.inputs = {'X': x}
         rst = _get_softmax_upper(x)
@@ -59,7 +59,7 @@ class TestSoftmaxMaskFuseOp(OpTest):
                  "core is not compiled with CUDA")
 class TestSoftmaxMaskFuseOp1(OpTest):
     def setUp(self):
-        self.op_type = "softmax_mask_fuse_upper_triangle"
+        self.op_type = "fused_softmax_mask_upper_triangle"
         x = np.random.random((1, 1, 32, 32))
         self.inputs = {'X': x}
         rst = _get_softmax_upper(x)
diff --git a/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py b/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py
index b81ad4ecdc82a528f7e4d69df94dfb81d7890d4d..636d0f5f9dd3b637e63963ee328aeb4924c44a9c 100644
--- a/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py
+++ b/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py
@@ -28,15 +28,15 @@ def softmax_mask_fuse_upper_triangle(x):
     :return: the result of softmax mask fuse (upper triangle)
     """
     if in_dygraph_mode():
-        out = core.ops.softmax_mask_fuse_upper_triangle(x)
+        out = core.ops.fused_softmax_mask_upper_triangle(x)
         return out
 
-    helper = LayerHelper('softmax_mask_fuse_upper_triangle', **locals())
+    helper = LayerHelper('fused_softmax_mask_upper_triangle', **locals())
 
     out = helper.create_variable_for_type_inference(dtype=x.dtype)
 
     helper.append_op(
-        type='softmax_mask_fuse_upper_triangle',
+        type='fused_softmax_mask_upper_triangle',
         inputs={'X': [x]},
         outputs={'Out': [out]})
     return out