diff --git a/paddle/fluid/framework/op_desc.cc b/paddle/fluid/framework/op_desc.cc
index a02466c04e9138615cf07068e04dc98033d53fd9..f31fefcfade89886acfc475ddd99e8310f677090 100644
--- a/paddle/fluid/framework/op_desc.cc
+++ b/paddle/fluid/framework/op_desc.cc
@@ -777,10 +777,17 @@ void OpDesc::CheckAttrs() {
   checker->Check(&attrs_);
 }
 
-void OpDesc::InferShape(const BlockDesc &block) const {
+void OpDesc::InferShape(const BlockDesc &block) {
   try {
     VLOG(3) << "CompileTime infer shape on " << Type();
-    auto &infer_shape = OpInfoMap::Instance().Get(this->Type()).infer_shape_;
+    auto &op_info = OpInfoMap::Instance().Get(this->Type());
+    auto *checker = op_info.Checker();
+    if (checker != nullptr) {
+      // set dafault value here
+      VLOG(10) << "begin to check attribute of " << Type();
+      checker->Check(&attrs_);
+    }
+    auto &infer_shape = op_info.infer_shape_;
     PADDLE_ENFORCE_EQ(
         static_cast<bool>(infer_shape), true,
         platform::errors::NotFound(
diff --git a/paddle/fluid/framework/op_desc.h b/paddle/fluid/framework/op_desc.h
index 82e15d40bee78d94e123105b6fa0095c75c930b1..0afe6796dad7a54f2f39c32e7afad6910baf8ea5 100644
--- a/paddle/fluid/framework/op_desc.h
+++ b/paddle/fluid/framework/op_desc.h
@@ -142,7 +142,7 @@ class OpDesc {
 
   void CheckAttrs();
 
-  void InferShape(const BlockDesc &block) const;
+  void InferShape(const BlockDesc &block);
 
   void InferVarType(BlockDesc *block) const;
 
diff --git a/paddle/fluid/operators/dropout_op.cc b/paddle/fluid/operators/dropout_op.cc
index 3d9950902acfe80a3cfef6c9efa2c6370e685c32..8d033ea3194b9bec0b5dbe1f267e3299b0cf367b 100644
--- a/paddle/fluid/operators/dropout_op.cc
+++ b/paddle/fluid/operators/dropout_op.cc
@@ -16,7 +16,7 @@ limitations under the License. */
 #include <string>
 #include "paddle/fluid/framework/infershape_utils.h"
 #include "paddle/fluid/framework/op_registry.h"
-#include "paddle/phi/infermeta/unary.h"
+#include "paddle/phi/infermeta/binary.h"
 
 namespace paddle {
 namespace operators {
diff --git a/paddle/phi/infermeta/binary.cc b/paddle/phi/infermeta/binary.cc
index ab13df081aa289cc3fbf43acaedc2a196d8d311f..60db5d342b8b371653317a1cb0736f56f57af293 100644
--- a/paddle/phi/infermeta/binary.cc
+++ b/paddle/phi/infermeta/binary.cc
@@ -776,6 +776,26 @@ void DistInferMeta(const MetaTensor& x,
   out->set_dtype(x.dtype());
 }
 
+void DropoutInferMeta(const MetaTensor& x,
+                      paddle::optional<const MetaTensor&> seed_tensor,
+                      float p,
+                      bool is_test,
+                      const std::string& mode,
+                      int seed,
+                      bool fix_seed,
+                      MetaTensor* out,
+                      MetaTensor* mask) {
+  auto x_dims = x.dims();
+  out->set_dims(x_dims);
+  out->share_lod(x);
+  out->set_dtype(x.dtype());
+
+  if (mask != nullptr) {
+    mask->set_dims(x_dims);
+    mask->set_dtype(DataType::UINT8);
+  }
+}
+
 void DotInferMeta(const MetaTensor& x, const MetaTensor& y, MetaTensor* out) {
   auto x_dims = x.dims();
   auto x_rank = static_cast<size_t>(x_dims.size());
diff --git a/paddle/phi/infermeta/binary.h b/paddle/phi/infermeta/binary.h
index 3fcbf69c35e25b7d4d9dd24711cbcca58e632f83..296c05756f29138e2c2209e389faf9a705e55e98 100644
--- a/paddle/phi/infermeta/binary.h
+++ b/paddle/phi/infermeta/binary.h
@@ -124,6 +124,16 @@ void DistInferMeta(const MetaTensor& x,
 
 void DotInferMeta(const MetaTensor& x, const MetaTensor& y, MetaTensor* out);
 
+void DropoutInferMeta(const MetaTensor& x,
+                      paddle::optional<const MetaTensor&> seed_tensor,
+                      float p,
+                      bool is_test,
+                      const std::string& mode,
+                      int seed,
+                      bool fix_seed,
+                      MetaTensor* out,
+                      MetaTensor* mask);
+
 void ElementwiseInferMeta(const MetaTensor& x,
                           const MetaTensor& y,
                           MetaTensor* out);
diff --git a/paddle/phi/infermeta/unary.cc b/paddle/phi/infermeta/unary.cc
index 36c192cbf27488343a0cd9d2870fe15c753804bd..e0ea637074c2027402317c49a210abd3325f83f5 100644
--- a/paddle/phi/infermeta/unary.cc
+++ b/paddle/phi/infermeta/unary.cc
@@ -24,6 +24,7 @@ limitations under the License. */
 #include "paddle/phi/core/infermeta_utils.h"
 #include "paddle/phi/kernels/funcs/parse_qr_mode.h"
 #include "paddle/phi/kernels/funcs/pooling.h"
+#include "paddle/phi/kernels/funcs/slice_utils.h"
 #include "paddle/phi/kernels/funcs/strided_slice.h"
 #include "paddle/phi/kernels/funcs/unfold_functor.h"
 #include "paddle/phi/kernels/funcs/unsqueeze.h"
@@ -360,17 +361,6 @@ void DiagonalInferMeta(const MetaTensor& input,
   out->set_dims(phi::make_ddim(out_dims));
 }
 
-void DropoutInferMeta(const MetaTensor& x, MetaTensor* out, MetaTensor* mask) {
-  auto x_dims = x.dims();
-  out->set_dims(x_dims);
-  out->share_lod(x);
-  out->set_dtype(x.dtype());
-
-  if (mask != nullptr) {
-    mask->set_dims(x_dims);
-  }
-}
-
 void EighInferMeta(const MetaTensor& x,
                    const std::string& uplo,
                    MetaTensor* out_w,
@@ -1738,6 +1728,51 @@ void SizeInferMeta(const MetaTensor& input, MetaTensor* out) {
   out->set_dims({1});
 }
 
+void SliceRawInferMeta(const MetaTensor& input,
+                       const std::vector<int64_t>& axes,
+                       const IntArray& starts_arr,
+                       const IntArray& ends_arr,
+                       const std::vector<int64_t>& infer_flags_t,
+                       const std::vector<int64_t>& decrease_axis,
+                       MetaTensor* out,
+                       MetaConfig config) {
+  auto in_dims = input.dims();
+  PADDLE_ENFORCE_LT(
+      in_dims.size(),
+      7,
+      phi::errors::InvalidArgument("The rank of input should be less than 7."));
+  DDim out_dims(in_dims);
+
+  std::vector<int64_t> infer_flags = infer_flags_t;
+  if (infer_flags.empty()) {
+    // Initialize infer_flags with 1.
+    // To be compatible with other op tests in which infer_flags is not set.
+    infer_flags = std::vector<int64_t>(axes.size(), 1);
+  }
+
+  // 2.1 Check attrs.
+  std::vector<int64_t> starts = starts_arr.GetData();
+  std::vector<int64_t> ends = ends_arr.GetData();
+
+  phi::funcs::CheckAndUpdateSliceAttrs<int64_t>(
+      in_dims, axes, &starts, &ends, nullptr, &infer_flags);
+
+  auto slice_dims = phi::funcs::GetSliceDims<int64_t>(
+      in_dims, axes, starts, ends, nullptr, &infer_flags);
+  if (config.is_runtime) {
+    out_dims = phi::funcs::GetDecreasedDims<int64_t>(
+        slice_dims, decrease_axis, &infer_flags);
+  } else {
+    out_dims = phi::funcs::GetDecreasedDims<int64_t>(
+        slice_dims, decrease_axis, nullptr);
+  }
+
+  out->set_dims(out_dims);
+  if (axes.size() > 0 && axes[0] != 0) {
+    out->share_lod(input);
+  }
+}
+
 void SoftmaxInferMeta(const MetaTensor& x, int axis, MetaTensor* out) {
   auto dim_x = x.dims();
   auto rank_x = dim_x.size();
diff --git a/paddle/phi/infermeta/unary.h b/paddle/phi/infermeta/unary.h
index bda9c83fce1f2867b431f40c93749f057f124739..5106c6f4487336741fda8855caab8e6628a0c2e9 100644
--- a/paddle/phi/infermeta/unary.h
+++ b/paddle/phi/infermeta/unary.h
@@ -80,8 +80,6 @@ void DiagInferMeta(const MetaTensor& x,
 void DiagonalInferMeta(
     const MetaTensor& input, int offset, int axis1, int axis2, MetaTensor* out);
 
-void DropoutInferMeta(const MetaTensor& x, MetaTensor* out, MetaTensor* mask);
-
 void EighInferMeta(const MetaTensor& x,
                    const std::string& uplo,
                    MetaTensor* out_w,
@@ -271,6 +269,15 @@ void ShardIndexInferMeta(const MetaTensor& in,
 
 void SizeInferMeta(const MetaTensor& input, MetaTensor* out);
 
+void SliceRawInferMeta(const MetaTensor& input,
+                       const std::vector<int64_t>& axes,
+                       const IntArray& starts,
+                       const IntArray& ends,
+                       const std::vector<int64_t>& infer_flags,
+                       const std::vector<int64_t>& decrease_axis,
+                       MetaTensor* out,
+                       MetaConfig config = MetaConfig());
+
 void SoftmaxInferMeta(const MetaTensor& x, int axis, MetaTensor* out);
 
 void SplitInferMeta(const MetaTensor& x_meta,
diff --git a/python/paddle/fluid/backward.py b/python/paddle/fluid/backward.py
index 0988f6709552b91e5fe2eebfc379ef4e3e15c4ac..ba7692b442f82c2ae6fefa159854e3b88907ceb8 100755
--- a/python/paddle/fluid/backward.py
+++ b/python/paddle/fluid/backward.py
@@ -1337,6 +1337,7 @@ def _append_backward_vars_(block, start_op_idx, grad_to_var, grad_info_map):
                 continue
             grad_info_map[grad_to_var[grad_var_name]] = (grad_var_name, block)
         # infer_shape and infer_type
+        op_desc.check_attrs()
         op_desc.infer_var_type(block.desc)
         op_desc.infer_shape(block.desc)
 
diff --git a/python/paddle/fluid/layers/nn.py b/python/paddle/fluid/layers/nn.py
index d7ec3276d8b7905cb60d004fe30fb8b5ed655f01..9f971faed34356f20a16847047aa5e82ce46a8ba 100755
--- a/python/paddle/fluid/layers/nn.py
+++ b/python/paddle/fluid/layers/nn.py
@@ -5141,7 +5141,6 @@ def l2_normalize(x, axis, epsilon=1e-12, name=None):
         #  [-0.33972208 -0.43014923  0.31772556  0.76617881 -0.10761525]]
 
     """
-
     if len(x.shape) == 1:
         axis = 0
     if _non_static_mode():
@@ -11199,18 +11198,15 @@ def slice(input, axes, starts, ends):
         infer_flags = list(1 for i in range(len(axes)))
 
         tmp_tensor_type = core.eager.Tensor
-
         if isinstance(starts, (list, tuple)):
             starts = [
                 item.numpy().item(0)
                 if isinstance(item, tmp_tensor_type) else item
                 for item in starts
             ]
-            attrs += ('starts', starts)
         elif isinstance(starts, tmp_tensor_type):
-            starts_tensor = starts
-            starts.stop_gradient = True
-            infer_flags = list(-1 for i in range(len(axes)))
+            tensor_t = starts.numpy()
+            starts = [ele for ele in tensor_t]
 
         if isinstance(ends, (list, tuple)):
             ends = [
@@ -11219,12 +11215,11 @@ def slice(input, axes, starts, ends):
             ]
             attrs += ('ends', ends)
         elif isinstance(ends, tmp_tensor_type):
-            ends_tensor = ends
-            ends_tensor.stop_gradient = True
-            infer_flags = list(-1 for i in range(len(axes)))
+            tensor_t = ends.numpy()
+            ends = [ele for ele in tensor_t]
 
-        return _C_ops.slice(input, starts_tensor, ends_tensor, None, None,
-                            'axes', axes, 'infer_flags', infer_flags, *attrs)
+        return _C_ops.final_state_slice(input, axes, starts, ends, infer_flags,
+                                        [])
     else:
         if _in_legacy_dygraph():
             attrs = ()
diff --git a/python/paddle/fluid/tests/unittests/test_dropout_op.py b/python/paddle/fluid/tests/unittests/test_dropout_op.py
index 09712005d412539f2e7ca3637e1a584f63508c9f..d8a4eb8f45f7d3220f4d5b52927dfcc8d2e0f8c7 100644
--- a/python/paddle/fluid/tests/unittests/test_dropout_op.py
+++ b/python/paddle/fluid/tests/unittests/test_dropout_op.py
@@ -22,8 +22,11 @@ import paddle
 import paddle.static as static
 import paddle.fluid as fluid
 from paddle.fluid import Program, program_guard
+from paddle.fluid.framework import _test_eager_guard
 import os
 
+from paddle import _C_ops
+
 
 class TestDropoutOp(OpTest):
     def setUp(self):
@@ -960,6 +963,19 @@ class TestDropoutBackward(unittest.TestCase):
                     np.array_equal(input.gradient(
                     ), self.cal_grad_downscale_in_infer(mask.numpy())))
 
+    def test_backward_downscale_in_infer_eager(self):
+        for place in self.places:
+            with fluid.dygraph.guard(place):
+                with _test_eager_guard():
+                    input = paddle.uniform([40, 40], dtype="float32")
+                    input.stop_gradient = False
+                    out, mask = _C_ops.final_state_dropout(
+                        input, None, 0.5, False, "downgrade_in_infer", 0, False)
+                    out.backward()
+                    self.assertTrue(
+                        np.array_equal(input.gradient(
+                        ), self.cal_grad_downscale_in_infer(mask.numpy())))
+
     def test_backward_upscale_train(self):
         for place in self.places:
             with fluid.dygraph.guard(place):
@@ -976,6 +992,21 @@ class TestDropoutBackward(unittest.TestCase):
                     np.allclose(input.gradient(
                     ), self.cal_grad_upscale_train(mask.numpy(), prob)))
 
+    def test_backward_upscale_train_eager(self):
+        for place in self.places:
+            with fluid.dygraph.guard(place):
+                with _test_eager_guard():
+                    prob = 0.5
+                    input = paddle.uniform([40, 40], dtype="float32")
+                    input.stop_gradient = False
+                    out, mask = _C_ops.final_state_dropout(
+                        input, None, 0.5, False, "upscale_in_train", 0, False)
+                    out.backward()
+
+                    self.assertTrue(
+                        np.allclose(input.gradient(
+                        ), self.cal_grad_upscale_train(mask.numpy(), prob)))
+
     def test_backward_upscale_train_2(self):
         for place in self.places:
             with fluid.dygraph.guard(place):
diff --git a/python/paddle/fluid/tests/unittests/test_slice_op.py b/python/paddle/fluid/tests/unittests/test_slice_op.py
index 71869b96aedf055a6ed8d6222ed060c50df091ba..a565bba304184384c6e760f56baa3bb069fef5ac 100644
--- a/python/paddle/fluid/tests/unittests/test_slice_op.py
+++ b/python/paddle/fluid/tests/unittests/test_slice_op.py
@@ -21,6 +21,7 @@ from op_test import OpTest, convert_float_to_uint16
 import paddle.fluid as fluid
 import paddle.fluid.layers as layers
 import paddle
+from paddle.fluid.framework import _test_eager_guard
 
 paddle.enable_static()
 
@@ -599,6 +600,31 @@ class TestSliceApiWithTensor(unittest.TestCase):
             self.assertTrue(np.array_equal(y_paddle.numpy(), y_np))
 
 
+class TestSliceApiEager(unittest.TestCase):
+    def test_slice_api(self):
+        with paddle.fluid.dygraph.guard():
+            with _test_eager_guard():
+                a = paddle.rand(shape=[4, 5, 6], dtype='float32')
+                a.stop_gradient = False
+                axes = [0, 1, 2]
+                starts = [-3, 0, 2]
+                ends = [3, 2, 4]
+                a_1 = paddle.slice(a, axes=axes, starts=starts, ends=ends)
+
+                a_2 = paddle.slice(
+                    a,
+                    axes=axes,
+                    starts=paddle.to_tensor(starts),
+                    ends=paddle.to_tensor(ends))
+
+                a_1.backward()
+                grad_truth = paddle.zeros_like(a)
+                grad_truth[-3:3, 0:2, 2:4] = 1
+                self.assertTrue(np.array_equal(grad_truth, a.gradient()))
+
+                self.assertTrue(np.allclose(a_1.numpy(), a[-3:3, 0:2, 2:4]))
+
+
 class TestSliceApiWithLoDTensorArray(unittest.TestCase):
     def setUp(self):
         self.shape = (3, 4)
diff --git a/python/paddle/nn/functional/common.py b/python/paddle/nn/functional/common.py
index 131d31aa02405a667e8a72817539f1eaf0136498..74df8f6ed5c344e9951e5193b2c17db893adcbb1 100644
--- a/python/paddle/nn/functional/common.py
+++ b/python/paddle/nn/functional/common.py
@@ -28,7 +28,7 @@ from ...tensor import clip
 from ...tensor import sum
 from ...tensor import sqrt
 from ...fluid.data_feeder import check_variable_and_dtype, check_dtype
-from ...fluid.framework import _varbase_creator, _in_legacy_dygraph, in_dygraph_mode
+from ...fluid.framework import _varbase_creator, _in_legacy_dygraph, in_dygraph_mode, _non_static_mode
 
 from ...fluid import dygraph_utils
 from ...fluid import layers
@@ -895,9 +895,15 @@ def dropout(x,
         seed = None
         mode = 'downgrade_in_infer' if mode == 'downscale_in_infer' else mode  #semantic transfer
 
-        if in_dynamic_mode():
+        if _non_static_mode():
             if default_main_program().random_seed != 0:
                 seed = default_main_program().random_seed
+
+            if in_dygraph_mode():
+                out, mask = _C_ops.final_state_dropout( x, None, p, not training, mode, \
+                    seed if seed is not None else 0, seed is not None)
+
+                return out
             out, mask = _C_ops.dropout(
                 x, 'dropout_prob', p, 'is_test', not training, 'fix_seed',
                 seed is not None, 'seed', seed
diff --git a/python/paddle/utils/code_gen/api.yaml b/python/paddle/utils/code_gen/api.yaml
index 4f05f107bc2fc55877f811fddd67e5eea7413561..2b0c562dbf9bdebffd84d25e898f31e194d95679 100644
--- a/python/paddle/utils/code_gen/api.yaml
+++ b/python/paddle/utils/code_gen/api.yaml
@@ -463,6 +463,16 @@
   kernel :
     func : dot
 
+- api : dropout
+  args : (Tensor x, Tensor seed_tensor, float p, bool is_test, str mode, int seed, bool fix_seed)
+  output : Tensor(out), Tensor(mask)
+  infer_meta :
+    func : DropoutInferMeta
+  kernel :
+    func : dropout
+  optional : seed_tensor
+  backward : dropout_grad
+
 # eigh
 - api : eigh
   args : (Tensor x, str uplo)
@@ -1504,6 +1514,15 @@
   kernel :
     func : size
 
+- api : slice
+  args : (Tensor input, int64_t[] axes, IntArray starts, IntArray ends, int64_t[] infer_flags, int64_t[] decrease_axis)
+  output : Tensor
+  infer_meta :
+    func : SliceRawInferMeta
+  kernel :
+    func : slice
+  backward : slice_grad
+
 # soft_shrink
 - api : soft_shrink
   args : (Tensor x, float lambda)
diff --git a/python/paddle/utils/code_gen/backward.yaml b/python/paddle/utils/code_gen/backward.yaml
index db1fe6cdf5220f0ad38694ad478a8058e8268219..cbcfc02ea099207da3ed39442401e1c54ed570d9 100644
--- a/python/paddle/utils/code_gen/backward.yaml
+++ b/python/paddle/utils/code_gen/backward.yaml
@@ -301,6 +301,17 @@
   kernel :
     func : divide_grad
 
+- backward_api : dropout_grad
+  forward : dropout (Tensor x, Tensor seed_tensor, float p, bool is_test, str mode, int seed, bool fix_seed) -> Tensor(out), Tensor(mask)
+  args : (Tensor mask, Tensor out_grad, float p, bool is_test, str mode)
+  output : Tensor(x_grad)
+  infer_meta :
+    func : UnchangedInferMeta
+    param : [out_grad]
+  kernel :
+    func : dropout_grad
+  optional : seed_tensor
+
 - backward_api : eigh_grad
   forward : eigh (Tensor x, str uplo) -> Tensor(out_w), Tensor(out_v)
   args : (Tensor out_w, Tensor out_v, Tensor out_w_grad, Tensor out_v_grad)
@@ -1054,6 +1065,16 @@
   kernel :
     func : sinh_grad
 
+- backward_api : slice_grad
+  forward : slice (Tensor input, int64_t[] axes, IntArray starts, IntArray ends, int64_t[] infer_flags, int64_t[] decrease_axis) -> Tensor(out)
+  args : (Tensor input, Tensor out_grad, int64_t[] axes, IntArray starts, IntArray ends, int64_t[] infer_flags, int64_t[] decrease_axis)
+  output : Tensor(input_grad)
+  infer_meta :
+    func : UnchangedInferMeta
+    param : [input]
+  kernel :
+    func : slice_grad
+
 - backward_api : soft_shrink_grad
   forward : soft_shrink (Tensor x, float lambda) -> Tensor(out)
   args : (Tensor x, Tensor out_grad, float lambda)
diff --git a/tools/infrt/skipped_phi_api.json b/tools/infrt/skipped_phi_api.json
index eef57a2d6b7bc7e83f0634a9e8a29343d0cab27a..74cb6fb0e535659024898ea905c20430db21dd5b 100644
--- a/tools/infrt/skipped_phi_api.json
+++ b/tools/infrt/skipped_phi_api.json
@@ -1,4 +1,4 @@
 {
-"phi_apis":["conj", "nll_loss", "flatten"],
+"phi_apis":["conj", "nll_loss", "dropout", "flatten"],
 "phi_kernels":["equal_all"]
 }