diff --git a/paddle/fluid/operators/argsort_op.cc b/paddle/fluid/operators/argsort_op.cc
index 999e2510a8db5859b248cf6582106411a02bb55f..1995b7ba048bb71b6c1fe357967e7317230825e8 100644
--- a/paddle/fluid/operators/argsort_op.cc
+++ b/paddle/fluid/operators/argsort_op.cc
@@ -93,4 +93,6 @@ REGISTER_OPERATOR(
     paddle::framework::EmptyGradOpMaker<paddle::imperative::OpBase>);
 REGISTER_OP_CPU_KERNEL(argsort,
                        ops::ArgsortKernel<paddle::platform::CPUPlace, float>,
-                       ops::ArgsortKernel<paddle::platform::CPUPlace, double>);
+                       ops::ArgsortKernel<paddle::platform::CPUPlace, double>,
+                       ops::ArgsortKernel<paddle::platform::CPUPlace, int>,
+                       ops::ArgsortKernel<paddle::platform::CPUPlace, int64_t>);
diff --git a/paddle/fluid/operators/argsort_op.cu b/paddle/fluid/operators/argsort_op.cu
index 8eb56e70b59ca486601862b501eca0b5731d2b31..0ea7e3dcb14867002b48e66c322160f8a2c49ba7 100644
--- a/paddle/fluid/operators/argsort_op.cu
+++ b/paddle/fluid/operators/argsort_op.cu
@@ -240,4 +240,6 @@ class ArgsortOpCUDAKernel : public framework::OpKernel<T> {
 REGISTER_OP_CUDA_KERNEL(
     argsort, paddle::operators::ArgsortOpCUDAKernel<float>,
     paddle::operators::ArgsortOpCUDAKernel<double>,
+    paddle::operators::ArgsortOpCUDAKernel<int>,
+    paddle::operators::ArgsortOpCUDAKernel<int64_t>,
     paddle::operators::ArgsortOpCUDAKernel<paddle::platform::float16>);
diff --git a/paddle/fluid/operators/cumsum_op.cc b/paddle/fluid/operators/cumsum_op.cc
index ab81a2873bf12455baa6ba222ff9aa50db0f9008..89cd5c697743e61fa945680cf84d81aea74ecfa9 100644
--- a/paddle/fluid/operators/cumsum_op.cc
+++ b/paddle/fluid/operators/cumsum_op.cc
@@ -81,4 +81,5 @@ REGISTER_OPERATOR(cumsum, ops::CumOp, ops::CumsumOpMaker,
                   ops::CumsumGradMaker<paddle::imperative::OpBase>);
 REGISTER_OP_CPU_KERNEL(cumsum, ops::CumKernel<CPU, ops::CumsumFunctor<float>>,
                        ops::CumKernel<CPU, ops::CumsumFunctor<double>>,
-                       ops::CumKernel<CPU, ops::CumsumFunctor<int>>);
+                       ops::CumKernel<CPU, ops::CumsumFunctor<int>>,
+                       ops::CumKernel<CPU, ops::CumsumFunctor<int64_t>>);
diff --git a/paddle/fluid/operators/cumsum_op.cu b/paddle/fluid/operators/cumsum_op.cu
index eb5fd99ccb844b1f1717b818e7807a384d6515eb..a1388f20dc5203867134b952d82fea7f1c87337f 100644
--- a/paddle/fluid/operators/cumsum_op.cu
+++ b/paddle/fluid/operators/cumsum_op.cu
@@ -19,4 +19,5 @@ using CUDA = paddle::platform::CUDADeviceContext;
 
 REGISTER_OP_CUDA_KERNEL(cumsum, ops::CumKernel<CUDA, ops::CumsumFunctor<float>>,
                         ops::CumKernel<CUDA, ops::CumsumFunctor<double>>,
-                        ops::CumKernel<CUDA, ops::CumsumFunctor<int>>);
+                        ops::CumKernel<CUDA, ops::CumsumFunctor<int>>,
+                        ops::CumKernel<CUDA, ops::CumsumFunctor<int64_t>>);
diff --git a/paddle/fluid/operators/expand_as_op.cc b/paddle/fluid/operators/expand_as_op.cc
index e72a16cb5f0b8d7109140f2861e9c7901a525ab5..da20ed3bbc2322bf3f124c488f628ffa7f18ac51 100644
--- a/paddle/fluid/operators/expand_as_op.cc
+++ b/paddle/fluid/operators/expand_as_op.cc
@@ -130,8 +130,11 @@ REGISTER_OP_CPU_KERNEL(
     expand_as, ops::ExpandAsKernel<paddle::platform::CPUDeviceContext, float>,
     ops::ExpandAsKernel<paddle::platform::CPUDeviceContext, double>,
     ops::ExpandAsKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::ExpandAsKernel<paddle::platform::CPUDeviceContext, int64_t>,
     ops::ExpandAsKernel<paddle::platform::CPUDeviceContext, bool>);
 REGISTER_OP_CPU_KERNEL(
     expand_as_grad,
+    ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, int64_t>,
     ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::ExpandAsGradKernel<paddle::platform::CPUDeviceContext, double>);
diff --git a/paddle/fluid/operators/expand_as_op.cu b/paddle/fluid/operators/expand_as_op.cu
index d7c894d24816298d4bd9b61ac0dfbb7f40679f1c..dbb1fcf3ab32619b73363bf064cb1cc286efde6e 100755
--- a/paddle/fluid/operators/expand_as_op.cu
+++ b/paddle/fluid/operators/expand_as_op.cu
@@ -15,8 +15,11 @@ REGISTER_OP_CUDA_KERNEL(
     expand_as, ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, float>,
     ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, double>,
     ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, int64_t>,
     ops::ExpandAsKernel<paddle::platform::CUDADeviceContext, bool>);
 REGISTER_OP_CUDA_KERNEL(
     expand_as_grad,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, int64_t>,
     ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, float>,
     ops::ExpandAsGradKernel<paddle::platform::CUDADeviceContext, double>);
diff --git a/paddle/fluid/operators/isfinite_op.cc b/paddle/fluid/operators/isfinite_op.cc
index f8c584163fbef040a96d696e5da07095854f1cc3..8936a8c9a2a7da1af8df3f199091192ab18e65ce 100644
--- a/paddle/fluid/operators/isfinite_op.cc
+++ b/paddle/fluid/operators/isfinite_op.cc
@@ -104,6 +104,8 @@ namespace ops = paddle::operators;
   REGISTER_OP_CPU_KERNEL(                                                   \
       op_type, ops::OverflowKernel<paddle::platform::CPUDeviceContext, int, \
                                    ops::functor>,                           \
+      ops::OverflowKernel<paddle::platform::CPUDeviceContext, int64_t,      \
+                          ops::functor>,                                    \
       ops::OverflowKernel<paddle::platform::CPUDeviceContext, float,        \
                           ops::functor>,                                    \
       ops::OverflowKernel<paddle::platform::CPUDeviceContext, double,       \
diff --git a/paddle/fluid/operators/pad2d_op.cc b/paddle/fluid/operators/pad2d_op.cc
index 71919667ba1bbaf3456b074c261dd714078681bf..9dfdcfcf13969ca030466c1ae66a28817e6947d1 100644
--- a/paddle/fluid/operators/pad2d_op.cc
+++ b/paddle/fluid/operators/pad2d_op.cc
@@ -661,5 +661,8 @@ REGISTER_OPERATOR(pad2d, ops::Pad2dOp, ops::Pad2dOpMaker,
                   ops::Pad2dOpGradMaker<paddle::imperative::OpBase>);
 REGISTER_OPERATOR(pad2d_grad, ops::Pad2dOpGrad,
                   ops::Pad2dOpGradNoNeedBufferVarsInference);
-REGISTER_OP_CPU_KERNEL(pad2d, ops::Pad2dCPUKernel<float>);
-REGISTER_OP_CPU_KERNEL(pad2d_grad, ops::Pad2dGradCPUKernel<float>);
+REGISTER_OP_CPU_KERNEL(pad2d, ops::Pad2dCPUKernel<float>,
+                       ops::Pad2dCPUKernel<double>, ops::Pad2dCPUKernel<int>,
+                       ops::Pad2dCPUKernel<int64_t>);
+REGISTER_OP_CPU_KERNEL(pad2d_grad, ops::Pad2dGradCPUKernel<float>,
+                       ops::Pad2dGradCPUKernel<double>);
diff --git a/paddle/fluid/operators/pad2d_op.cu b/paddle/fluid/operators/pad2d_op.cu
index 72eca08b06b144335424a669241b5754beda758d..4b6e3e9521d675964e61061d17f49d274bfe1606 100644
--- a/paddle/fluid/operators/pad2d_op.cu
+++ b/paddle/fluid/operators/pad2d_op.cu
@@ -459,5 +459,8 @@ class Pad2dGradCUDAKernel : public framework::OpKernel<T> {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(pad2d, ops::Pad2dCUDAKernel<float>);
-REGISTER_OP_CUDA_KERNEL(pad2d_grad, ops::Pad2dGradCUDAKernel<float>);
+REGISTER_OP_CUDA_KERNEL(pad2d, ops::Pad2dCUDAKernel<float>,
+                        ops::Pad2dCUDAKernel<double>, ops::Pad2dCUDAKernel<int>,
+                        ops::Pad2dCUDAKernel<int64_t>);
+REGISTER_OP_CUDA_KERNEL(pad2d_grad, ops::Pad2dGradCUDAKernel<float>,
+                        ops::Pad2dGradCUDAKernel<double>);
diff --git a/paddle/fluid/operators/pad_constant_like_op.cc b/paddle/fluid/operators/pad_constant_like_op.cc
index d408275d8e5ca9ee97b3ca4dc99f7154ffa6cc5a..4fb4f7a022f6bd8ff0843b62d735a3535ae77d81 100644
--- a/paddle/fluid/operators/pad_constant_like_op.cc
+++ b/paddle/fluid/operators/pad_constant_like_op.cc
@@ -224,8 +224,13 @@ REGISTER_OPERATOR(pad_constant_like_grad, ops::PadConstantLikeOpGrad);
 REGISTER_OP_CPU_KERNEL(
     pad_constant_like,
     ops::PadConstantLikeKernel<paddle::platform::CPUDeviceContext, float>,
-    ops::PadConstantLikeKernel<paddle::platform::CPUDeviceContext, double>);
+    ops::PadConstantLikeKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::PadConstantLikeKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::PadConstantLikeKernel<paddle::platform::CPUDeviceContext, int64_t>);
 REGISTER_OP_CPU_KERNEL(
     pad_constant_like_grad,
     ops::PadConstantLikeGradKernel<paddle::platform::CPUDeviceContext, float>,
-    ops::PadConstantLikeGradKernel<paddle::platform::CPUDeviceContext, double>);
+    ops::PadConstantLikeGradKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::PadConstantLikeGradKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::PadConstantLikeGradKernel<paddle::platform::CPUDeviceContext,
+                                   int64_t>);
diff --git a/paddle/fluid/operators/pad_constant_like_op.cu b/paddle/fluid/operators/pad_constant_like_op.cu
index 9e62a6dc9d34a96c59a08d0e5fd6cdd9f0d6d51d..76faf30ed92000d7093eb73bf6499a43f6ab5b57 100644
--- a/paddle/fluid/operators/pad_constant_like_op.cu
+++ b/paddle/fluid/operators/pad_constant_like_op.cu
@@ -17,9 +17,14 @@ namespace ops = paddle::operators;
 REGISTER_OP_CUDA_KERNEL(
     pad_constant_like,
     ops::PadConstantLikeKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::PadConstantLikeKernel<paddle::platform::CUDADeviceContext, double>);
+    ops::PadConstantLikeKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::PadConstantLikeKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::PadConstantLikeKernel<paddle::platform::CUDADeviceContext, int64_t>);
 REGISTER_OP_CUDA_KERNEL(
     pad_constant_like_grad,
+    ops::PadConstantLikeGradKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::PadConstantLikeGradKernel<paddle::platform::CUDADeviceContext,
+                                   int64_t>,
     ops::PadConstantLikeGradKernel<paddle::platform::CUDADeviceContext, float>,
     ops::PadConstantLikeGradKernel<paddle::platform::CUDADeviceContext,
                                    double>);
diff --git a/paddle/fluid/operators/pad_op.cc b/paddle/fluid/operators/pad_op.cc
index b98f201e8a1699577d8948f3754ddc2c00c05841..a604c28cb006f8d7e5696ed8937fd1212a79173b 100644
--- a/paddle/fluid/operators/pad_op.cc
+++ b/paddle/fluid/operators/pad_op.cc
@@ -147,6 +147,10 @@ REGISTER_OPERATOR(pad, ops::PadOp, ops::PadOpMaker,
                   ops::PadOpGradMaker<paddle::imperative::OpBase>);
 REGISTER_OPERATOR(pad_grad, ops::PadOpGrad);
 REGISTER_OP_CPU_KERNEL(
-    pad, ops::PadKernel<paddle::platform::CPUDeviceContext, float>);
+    pad, ops::PadKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::PadKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::PadKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::PadKernel<paddle::platform::CPUDeviceContext, int64_t>);
 REGISTER_OP_CPU_KERNEL(
-    pad_grad, ops::PadGradKernel<paddle::platform::CPUDeviceContext, float>);
+    pad_grad, ops::PadGradKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::PadGradKernel<paddle::platform::CPUDeviceContext, double>);
diff --git a/paddle/fluid/operators/pad_op.cu b/paddle/fluid/operators/pad_op.cu
index e07c99b12ba818d126ffbb68f7a4928987792e08..391e305352e55188fb0c502b8efe03af597d48ca 100644
--- a/paddle/fluid/operators/pad_op.cu
+++ b/paddle/fluid/operators/pad_op.cu
@@ -18,6 +18,8 @@ namespace plat = paddle::platform;
 REGISTER_OP_CUDA_KERNEL(
     pad, ops::PadKernel<paddle::platform::CUDADeviceContext, double>,
     ops::PadKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::PadKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::PadKernel<paddle::platform::CUDADeviceContext, int64_t>,
     ops::PadKernel<paddle::platform::CUDADeviceContext, plat::float16>);
 REGISTER_OP_CUDA_KERNEL(
     pad_grad, ops::PadGradKernel<paddle::platform::CUDADeviceContext, double>,
diff --git a/paddle/fluid/operators/save_combine_op.cc b/paddle/fluid/operators/save_combine_op.cc
index 953e2655d13328b986a67398dca54f8a5e3aedcf..68144e603bddf3457bde373b69f99b67625534d0 100644
--- a/paddle/fluid/operators/save_combine_op.cc
+++ b/paddle/fluid/operators/save_combine_op.cc
@@ -86,4 +86,5 @@ REGISTER_OP_CPU_KERNEL(
     save_combine,
     ops::SaveCombineOpKernel<paddle::platform::CPUDeviceContext, float>,
     ops::SaveCombineOpKernel<paddle::platform::CPUDeviceContext, double>,
-    ops::SaveCombineOpKernel<paddle::platform::CPUDeviceContext, int>);
+    ops::SaveCombineOpKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::SaveCombineOpKernel<paddle::platform::CPUDeviceContext, int64_t>);
diff --git a/paddle/fluid/operators/save_combine_op.cu b/paddle/fluid/operators/save_combine_op.cu
index 78607823a0368d216310bbbb390fd7face002839..71476fd802bdd9e617e69d422b1e931178f05322 100644
--- a/paddle/fluid/operators/save_combine_op.cu
+++ b/paddle/fluid/operators/save_combine_op.cu
@@ -20,4 +20,5 @@ REGISTER_OP_CUDA_KERNEL(
     save_combine,
     ops::SaveCombineOpKernel<paddle::platform::CUDADeviceContext, float>,
     ops::SaveCombineOpKernel<paddle::platform::CUDADeviceContext, double>,
-    ops::SaveCombineOpKernel<paddle::platform::CUDADeviceContext, int>);
+    ops::SaveCombineOpKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::SaveCombineOpKernel<paddle::platform::CUDADeviceContext, int64_t>);
diff --git a/paddle/fluid/operators/scatter_op.cc b/paddle/fluid/operators/scatter_op.cc
index 6ba5ba7bb277c21384c12b60f9f573e4afc31513..12dadc4eb9288e334dfc90b83f4ccd5658abf565 100644
--- a/paddle/fluid/operators/scatter_op.cc
+++ b/paddle/fluid/operators/scatter_op.cc
@@ -146,5 +146,10 @@ REGISTER_OPERATOR(scatter, ops::ScatterOp, ops::ScatterOpMaker,
 REGISTER_OPERATOR(scatter_grad, ops::ScatterGradOp,
                   ops::ScatterGradNoNeedBufferVarsInference,
                   ops::ScatterGradInplaceInferer);
-REGISTER_OP_CPU_KERNEL(scatter, ops::ScatterOpKernel<float>);
-REGISTER_OP_CPU_KERNEL(scatter_grad, ops::ScatterGradientOpKernel<float>);
+REGISTER_OP_CPU_KERNEL(scatter, ops::ScatterOpKernel<float>,
+                       ops::ScatterOpKernel<double>, ops::ScatterOpKernel<int>,
+                       ops::ScatterOpKernel<int64_t>);
+REGISTER_OP_CPU_KERNEL(scatter_grad, ops::ScatterGradientOpKernel<float>,
+                       ops::ScatterGradientOpKernel<double>,
+                       ops::ScatterGradientOpKernel<int>,
+                       ops::ScatterGradientOpKernel<int64_t>);
diff --git a/paddle/fluid/operators/scatter_op.cu b/paddle/fluid/operators/scatter_op.cu
index f3e0faa164c2f95bcb2402a7ba92847cfb67167c..938616e86c71f2787aa847d68d73b9e1f85de71d 100644
--- a/paddle/fluid/operators/scatter_op.cu
+++ b/paddle/fluid/operators/scatter_op.cu
@@ -94,5 +94,9 @@ class ScatterGradOpCUDAKernel : public framework::OpKernel<T> {
 }  // namespace paddle
 
 namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(scatter, ops::ScatterOpCUDAKernel<float>);
-REGISTER_OP_CUDA_KERNEL(scatter_grad, ops::ScatterGradOpCUDAKernel<float>);
+REGISTER_OP_CUDA_KERNEL(scatter, ops::ScatterOpCUDAKernel<float>,
+                        ops::ScatterOpCUDAKernel<double>,
+                        ops::ScatterOpCUDAKernel<int>,
+                        ops::ScatterOpCUDAKernel<int64_t>);
+REGISTER_OP_CUDA_KERNEL(scatter_grad, ops::ScatterGradOpCUDAKernel<float>,
+                        ops::ScatterGradOpCUDAKernel<double>);
diff --git a/paddle/fluid/operators/sequence_ops/sequence_concat_op.cc b/paddle/fluid/operators/sequence_ops/sequence_concat_op.cc
index ac513dee29ef8632b74f64d4b3e75ba2a9bc7a85..d8214ac2ca62a870e4cccd37eedd75913d11a701 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_concat_op.cc
+++ b/paddle/fluid/operators/sequence_ops/sequence_concat_op.cc
@@ -126,7 +126,7 @@ REGISTER_OPERATOR(sequence_concat, op::SequenceConcatOp, op::SeqConcatOpMaker,
 template <typename T>
 using Kernel = op::SeqConcatKernel<paddle::platform::CPUDeviceContext, T>;
 REGISTER_OP_CPU_KERNEL(sequence_concat, Kernel<float>, Kernel<double>,
-                       Kernel<int64_t>);
+                       Kernel<int>, Kernel<int64_t>);
 
 REGISTER_OPERATOR(sequence_concat_grad, op::SeqConcatGradOp,
                   op::SeqConcatGradNoNeedBufferVarsInference);
@@ -134,4 +134,5 @@ template <typename T>
 using GradKernel =
     op::SeqConcatGradKernel<paddle::platform::CPUDeviceContext, T>;
 REGISTER_OP_CPU_KERNEL(sequence_concat_grad, GradKernel<float>,
-                       GradKernel<double>, GradKernel<int64_t>);
+                       GradKernel<double>, GradKernel<int>,
+                       GradKernel<int64_t>);
diff --git a/paddle/fluid/operators/sequence_ops/sequence_concat_op.cu.cc b/paddle/fluid/operators/sequence_ops/sequence_concat_op.cu.cc
index 7b8043bc4538b486bb73e005769e1585e5c4817e..43fd84a711ff7fbf999a66bfdfe320b4daceb1a7 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_concat_op.cu.cc
+++ b/paddle/fluid/operators/sequence_ops/sequence_concat_op.cu.cc
@@ -17,10 +17,12 @@
 template <typename T>
 using Kernel =
     paddle::operators::SeqConcatKernel<paddle::platform::CUDADeviceContext, T>;
-REGISTER_OP_CUDA_KERNEL(sequence_concat, Kernel<float>, Kernel<double>);
+REGISTER_OP_CUDA_KERNEL(sequence_concat, Kernel<float>, Kernel<double>,
+                        Kernel<int>, Kernel<int64_t>);
 template <typename T>
 using GradKernel =
     paddle::operators::SeqConcatGradKernel<paddle::platform::CUDADeviceContext,
                                            T>;
 REGISTER_OP_CUDA_KERNEL(sequence_concat_grad, GradKernel<float>,
-                        GradKernel<double>);
+                        GradKernel<double>, GradKernel<int>,
+                        GradKernel<int64_t>);
diff --git a/paddle/fluid/operators/sequence_ops/sequence_slice_op.cc b/paddle/fluid/operators/sequence_ops/sequence_slice_op.cc
index 9e31a81c4353ea8622c258fc5cc4d89eabee4f28..655afe7e53f19c9bbdc3e8f6bf4560cc31e3a7cd 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_slice_op.cc
+++ b/paddle/fluid/operators/sequence_ops/sequence_slice_op.cc
@@ -150,7 +150,14 @@ REGISTER_OPERATOR(sequence_slice_grad, ops::SequenceSliceGradOp,
                   ops::SequenceSliceGradNoNeedBufferVarsInference);
 REGISTER_OP_CPU_KERNEL(
     sequence_slice,
-    ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, float>);
+    ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::SequenceSliceOpKernel<paddle::platform::CPUDeviceContext, int64_t>);
 REGISTER_OP_CPU_KERNEL(
     sequence_slice_grad,
-    ops::SequenceSliceGradOpKernel<paddle::platform::CPUDeviceContext, float>);
+    ops::SequenceSliceGradOpKernel<paddle::platform::CPUDeviceContext, float>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CPUDeviceContext, int>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CPUDeviceContext,
+                                   int64_t>);
diff --git a/paddle/fluid/operators/sequence_ops/sequence_slice_op.cu b/paddle/fluid/operators/sequence_ops/sequence_slice_op.cu
index 1e4a1b8323dbaacdf3f74c33e7aa4484d9be2478..a4b0ea2e5b2f4972a3004e427bcae96e87a5a9e9 100644
--- a/paddle/fluid/operators/sequence_ops/sequence_slice_op.cu
+++ b/paddle/fluid/operators/sequence_ops/sequence_slice_op.cu
@@ -17,7 +17,14 @@ limitations under the License. */
 namespace ops = paddle::operators;
 REGISTER_OP_CUDA_KERNEL(
     sequence_slice,
-    ops::SequenceSliceOpKernel<paddle::platform::CUDADeviceContext, float>);
+    ops::SequenceSliceOpKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::SequenceSliceOpKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::SequenceSliceOpKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::SequenceSliceOpKernel<paddle::platform::CUDADeviceContext, int64_t>);
 REGISTER_OP_CUDA_KERNEL(
     sequence_slice_grad,
-    ops::SequenceSliceGradOpKernel<paddle::platform::CUDADeviceContext, float>);
+    ops::SequenceSliceGradOpKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CUDADeviceContext, int>,
+    ops::SequenceSliceGradOpKernel<paddle::platform::CUDADeviceContext,
+                                   int64_t>);
diff --git a/paddle/fluid/operators/space_to_depth_op.cc b/paddle/fluid/operators/space_to_depth_op.cc
index 133be773ac76e376b3bcb57e0b3a9061091cc16b..1042d2acd3928a2a1287917cb50c53f0f9654bde 100644
--- a/paddle/fluid/operators/space_to_depth_op.cc
+++ b/paddle/fluid/operators/space_to_depth_op.cc
@@ -187,9 +187,11 @@ REGISTER_OP_CPU_KERNEL(
     space_to_depth,
     ops::SpaceToDepthKernel<paddle::platform::CPUDeviceContext, float>,
     ops::SpaceToDepthKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::SpaceToDepthKernel<paddle::platform::CPUDeviceContext, int>,
     ops::SpaceToDepthKernel<paddle::platform::CPUDeviceContext, int64_t>);
 REGISTER_OP_CPU_KERNEL(
     space_to_depth_grad,
     ops::SpaceToDepthGradKernel<paddle::platform::CPUDeviceContext, float>,
     ops::SpaceToDepthGradKernel<paddle::platform::CPUDeviceContext, double>,
+    ops::SpaceToDepthGradKernel<paddle::platform::CPUDeviceContext, int>,
     ops::SpaceToDepthGradKernel<paddle::platform::CPUDeviceContext, int64_t>);
diff --git a/paddle/fluid/operators/space_to_depth_op.cu b/paddle/fluid/operators/space_to_depth_op.cu
index 38d0a662733222386b8ecd68d064f3d1abe56c3b..5a9f7c288d12c571df02828f9ad5a07563f127b4 100644
--- a/paddle/fluid/operators/space_to_depth_op.cu
+++ b/paddle/fluid/operators/space_to_depth_op.cu
@@ -21,10 +21,12 @@ REGISTER_OP_CUDA_KERNEL(
     space_to_depth,
     ops::SpaceToDepthKernel<paddle::platform::CUDADeviceContext, float>,
     ops::SpaceToDepthKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::SpaceToDepthKernel<paddle::platform::CUDADeviceContext, int>,
     ops::SpaceToDepthKernel<paddle::platform::CUDADeviceContext, int64_t>);
 
 REGISTER_OP_CUDA_KERNEL(
     space_to_depth_grad,
     ops::SpaceToDepthGradKernel<paddle::platform::CUDADeviceContext, float>,
     ops::SpaceToDepthGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::SpaceToDepthGradKernel<paddle::platform::CUDADeviceContext, int>,
     ops::SpaceToDepthGradKernel<paddle::platform::CUDADeviceContext, int64_t>);
diff --git a/paddle/fluid/operators/top_k_op.cc b/paddle/fluid/operators/top_k_op.cc
index 7b9960c8b47e73c7f7a8c29a789f02eadee06eed..c18ec5d41884f7cec09236c5b16b010616cce4b3 100644
--- a/paddle/fluid/operators/top_k_op.cc
+++ b/paddle/fluid/operators/top_k_op.cc
@@ -99,4 +99,6 @@ REGISTER_OPERATOR(
     paddle::framework::EmptyGradOpMaker<paddle::imperative::OpBase>);
 REGISTER_OP_CPU_KERNEL(top_k,
                        ops::TopkKernel<paddle::platform::CPUPlace, float>,
-                       ops::TopkKernel<paddle::platform::CPUPlace, double>);
+                       ops::TopkKernel<paddle::platform::CPUPlace, double>,
+                       ops::TopkKernel<paddle::platform::CPUPlace, int>,
+                       ops::TopkKernel<paddle::platform::CPUPlace, int64_t>);
diff --git a/paddle/fluid/operators/top_k_op.cu b/paddle/fluid/operators/top_k_op.cu
index 0375611dfb5b9c129e9cfa26808ec40185eac47d..c56716f8ed37965474e980d0f1338317369b18d1 100644
--- a/paddle/fluid/operators/top_k_op.cu
+++ b/paddle/fluid/operators/top_k_op.cu
@@ -549,4 +549,6 @@ class TopkOpCUDAKernel : public framework::OpKernel<T> {
 REGISTER_OP_CUDA_KERNEL(
     top_k, paddle::operators::TopkOpCUDAKernel<float>,
     paddle::operators::TopkOpCUDAKernel<double>,
+    paddle::operators::TopkOpCUDAKernel<int>,
+    paddle::operators::TopkOpCUDAKernel<int64_t>,
     paddle::operators::TopkOpCUDAKernel<paddle::platform::float16>);