From 3794d171f36d0cfc8db518bce4bf3718a5bda31e Mon Sep 17 00:00:00 2001
From: Meteor Liu <liudongxue01@baidu.com>
Date: Mon, 22 May 2023 20:56:38 +0800
Subject: [PATCH] [dygraph]unify _non_static_mode() in_dygraph_mode() and
 in_dynamic_mode() (#53856)

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* [dygraph]unify _non_static_mode() in_dygraph_mode() and in_dynamic_mode()

* fixed cyclic reference that caused patial import

* fixed bad change

* fix bad import

* fix bad import

* fix bad import

* fix ut failed caused by change in_dynamic_mode

* fix ut failed caused by change in_dynamic_mode

* fixed usage of in_dynamic_mode() or in_dygraph_mode()

* revert python3 to python in .pre-commit-config.yaml

* fix merge conflicts
---
 python/paddle/amp/grad_scaler.py              |   5 +-
 python/paddle/autograd/autograd.py            |   2 +-
 python/paddle/common_ops_import.py            |   1 -
 .../distributed/auto_parallel/engine.py       |   8 +-
 .../auto_parallel/process_group.py            |   4 +-
 python/paddle/distributed/collective.py       |   8 +-
 .../distributed/communication/all_gather.py   |   2 +-
 .../communication/batch_isend_irecv.py        |   2 +-
 .../distributed/communication/broadcast.py    |   2 +-
 .../distributed/communication/gather.py       |   2 +-
 .../paddle/distributed/communication/group.py |   8 +-
 .../distributed/communication/reduce.py       |   2 +-
 .../distributed/communication/scatter.py      |   2 +-
 .../communication/stream/all_gather.py        |   2 +-
 .../communication/stream/all_reduce.py        |   2 +-
 .../communication/stream/all_to_all.py        |   4 +-
 .../communication/stream/broadcast.py         |   2 +-
 .../communication/stream/gather.py            |   2 +-
 .../distributed/communication/stream/recv.py  |   2 +-
 .../communication/stream/reduce.py            |   2 +-
 .../communication/stream/reduce_scatter.py    |   4 +-
 .../communication/stream/scatter.py           |   2 +-
 .../distributed/communication/stream/send.py  |   2 +-
 .../distributed/fleet/base/role_maker.py      |   2 +-
 python/paddle/distributed/fleet/fleet.py      |   9 +-
 .../distributed/fleet/layers/mpu/mp_ops.py    |  18 +-
 .../distributed/fleet/layers/mpu/random.py    |   5 +-
 .../fleet/meta_optimizers/dgc_optimizer.py    |   4 +-
 .../parallel_layers/pp_layers.py              |   2 +-
 .../fleet/meta_parallel/pipeline_parallel.py  |   2 +-
 .../pp_utils/p2p_communication.py             |   6 +-
 .../sharding/group_sharded_stage3.py          |   2 +-
 python/paddle/distributed/fleet/optimizer.py  |   4 +-
 .../distributed/fleet/recompute/recompute.py  |   4 +-
 .../fleet/recompute/recompute_hybrid.py       |   2 +-
 .../fleet/utils/hybrid_parallel_inference.py  |   5 +-
 .../fleet/utils/hybrid_parallel_util.py       |  10 +-
 .../fleet/utils/mix_precision_utils.py        |   4 +-
 python/paddle/distributed/models/moe/utils.py |  12 +-
 python/paddle/distributed/parallel.py         |  16 +-
 python/paddle/distributed/utils/moe_utils.py  |   6 +-
 python/paddle/distribution/bernoulli.py       |  18 +-
 python/paddle/distribution/categorical.py     |   8 +-
 python/paddle/distribution/dirichlet.py       |   4 +-
 python/paddle/distribution/distribution.py    |   4 +-
 .../paddle/distribution/exponential_family.py |   4 +-
 python/paddle/distribution/kl.py              |   4 +-
 python/paddle/distribution/normal.py          |   8 +-
 python/paddle/distribution/uniform.py         |  10 +-
 python/paddle/fft.py                          |  14 +-
 python/paddle/fluid/contrib/optimizer.py      |   2 +-
 python/paddle/fluid/data_feeder.py            |   8 +-
 python/paddle/fluid/dygraph/base.py           |   2 +-
 .../fluid/dygraph/tensor_patch_methods.py     |  25 +-
 python/paddle/fluid/framework.py              |  49 +---
 .../incubate/checkpoint/auto_checkpoint.py    |   4 +-
 python/paddle/fluid/layer_helper.py           |   4 +-
 python/paddle/fluid/layer_helper_base.py      |   6 +-
 python/paddle/fluid/lazy_init.py              |   2 +-
 python/paddle/fluid/reader.py                 |   6 +-
 .../fleet/hybrid_parallel_pp_recompute.py     |   8 +-
 .../fleet/parallel_dygraph_se_resnext.py      |   2 +-
 .../tests/unittests/test_context_manager.py   |   4 +-
 .../tests/unittests/test_dropout_nd_op.py     |   4 +-
 .../unittests/test_imperative_decorator.py    |   2 +-
 .../unittests/test_imperative_double_grad.py  |   2 +-
 .../test_imperative_ocr_attention_model.py    |   2 +-
 .../tests/unittests/test_imperative_resnet.py |   2 +-
 .../test_imperative_resnet_sorted_gradient.py |   2 +-
 .../unittests/test_imperative_se_resnext.py   |   2 +-
 ...perative_star_gan_with_gradient_penalty.py |   6 +-
 .../test_imperative_tensor_clear_gradient.py  |   2 +-
 .../unittests/test_imperative_triple_grad.py  |   2 +-
 .../fluid/tests/unittests/test_logical_op.py  |   6 +-
 .../tests/unittests/test_multiclass_nms_op.py |  30 +--
 .../unittests/test_squared_l2_norm_op.py      |   4 +-
 python/paddle/fluid/unique_name.py            |   4 +-
 python/paddle/fluid/variable_index.py         |  11 +-
 python/paddle/framework/__init__.py           |   6 +-
 python/paddle/framework/io.py                 |  12 +-
 python/paddle/geometric/math.py               |  10 +-
 .../geometric/message_passing/send_recv.py    |   9 +-
 python/paddle/geometric/reindex.py            |   7 +-
 python/paddle/geometric/sampling/neighbors.py |   6 +-
 python/paddle/hapi/model.py                   |  25 +-
 python/paddle/incubate/autograd/functional.py |  10 +-
 .../distributed/models/moe/moe_layer.py       |   4 +-
 .../incubate/distributed/models/moe/utils.py  |   4 +-
 python/paddle/incubate/layers/nn.py           |   6 +-
 .../nn/functional/fused_dropout_add.py        |   5 +-
 .../nn/functional/fused_gate_attention.py     |   4 +-
 .../nn/functional/fused_matmul_bias.py        |   4 +-
 .../nn/functional/fused_transformer.py        |  11 +-
 .../incubate/nn/layer/fused_dropout_nd.py     |   4 +-
 .../incubate/nn/layer/fused_transformer.py    |   5 +-
 python/paddle/incubate/nn/loss.py             |   4 +-
 .../incubate/nn/memory_efficient_attention.py |   4 +-
 .../incubate/operators/graph_khop_sampler.py  |   4 +-
 .../incubate/operators/graph_reindex.py       |   4 +-
 .../operators/graph_sample_neighbors.py       |   4 +-
 .../incubate/operators/graph_send_recv.py     |   5 +-
 .../incubate/operators/softmax_mask_fuse.py   |   4 +-
 .../softmax_mask_fuse_upper_triangle.py       |   4 +-
 .../optimizer/distributed_fused_lamb.py       |   4 +-
 .../paddle/incubate/optimizer/modelaverage.py |  38 +--
 python/paddle/incubate/tensor/manipulation.py |   4 +-
 python/paddle/incubate/tensor/math.py         |  10 +-
 python/paddle/incubate/xpu/resnet_block.py    |   2 +-
 .../paddle/io/dataloader/dataloader_iter.py   |   8 +-
 python/paddle/io/dataloader/dataset.py        |   2 +-
 python/paddle/io/reader.py                    |   7 +-
 python/paddle/jit/api.py                      |   7 +-
 .../jit/dy2static/program_translator.py       |   5 +-
 python/paddle/jit/translated_layer.py         |   5 +-
 python/paddle/metric/metrics.py               |   5 +-
 python/paddle/nn/clip.py                      |  19 +-
 python/paddle/nn/decode.py                    |   4 +-
 python/paddle/nn/functional/activation.py     |  65 +++---
 python/paddle/nn/functional/common.py         | 217 +++++++-----------
 python/paddle/nn/functional/conv.py           |  18 +-
 python/paddle/nn/functional/distance.py       |   4 +-
 python/paddle/nn/functional/extension.py      |  13 +-
 python/paddle/nn/functional/input.py          |   6 +-
 python/paddle/nn/functional/loss.py           |  54 ++---
 python/paddle/nn/layer/norm.py                |  42 +---
 python/paddle/nn/layer/rnn.py                 |   3 +-
 python/paddle/nn/quant/format.py              |   6 +-
 python/paddle/nn/utils/weight_norm_hook.py    |   4 +-
 python/paddle/optimizer/adadelta.py           |   4 +-
 python/paddle/optimizer/adamw.py              |   2 +-
 python/paddle/optimizer/momentum.py           |   8 +-
 python/paddle/optimizer/optimizer.py          |  18 +-
 .../paddle/quantization/quanters/abs_max.py   |   2 +-
 python/paddle/regularizer.py                  |  42 ++--
 python/paddle/signal.py                       |  21 +-
 python/paddle/sparse/unary.py                 |   4 +-
 python/paddle/static/nn/common.py             |  10 +-
 python/paddle/static/nn/control_flow.py       |   6 +-
 python/paddle/static/nn/metric.py             |   4 +-
 python/paddle/static/nn/sequence_lod.py       |  28 +--
 python/paddle/tensor/array.py                 |  10 +-
 python/paddle/tensor/creation.py              |  50 ++--
 .../paddle/tensor/layer_function_generator.py |   6 +-
 python/paddle/tensor/linalg.py                |  76 +++---
 python/paddle/tensor/logic.py                 |  44 ++--
 python/paddle/tensor/manipulation.py          |  86 +++----
 python/paddle/tensor/math.py                  | 154 ++++++-------
 python/paddle/tensor/ops.py                   |  47 ++--
 python/paddle/tensor/random.py                |  25 +-
 python/paddle/tensor/search.py                |  28 +--
 python/paddle/tensor/stat.py                  |  14 +-
 python/paddle/text/viterbi_decode.py          |  12 +-
 .../utils/cpp_extension/extension_utils.py    |   4 +-
 python/paddle/utils/dlpack.py                 |   6 +-
 python/paddle/utils/inplace_utils.py          |   4 +-
 python/paddle/utils/layers_utils.py           |   6 +-
 test/amp/amp_base_models.py                   |   8 +-
 test/auto_parallel/test_to_static.py          |   6 +-
 test/dygraph_to_static/test_convert_call.py   |   4 +-
 test/dygraph_to_static/test_lac.py            |   8 +-
 .../inference/test_trt_multiclass_nms3_op.py  |   4 +-
 test/legacy_test/test_dlpack.py               |   2 +-
 .../test_composite_layer_norm.py              |   4 +-
 test/tokenizer/test_faster_tokenizer_op.py    |   5 +-
 164 files changed, 869 insertions(+), 1072 deletions(-)

diff --git a/python/paddle/amp/grad_scaler.py b/python/paddle/amp/grad_scaler.py
index 2cade3482e9..e7774a35c1c 100644
--- a/python/paddle/amp/grad_scaler.py
+++ b/python/paddle/amp/grad_scaler.py
@@ -19,10 +19,11 @@ from enum import Enum
 import numpy as np
 
 from paddle import _C_ops, _legacy_C_ops
-from paddle.fluid import core, in_dygraph_mode
+from paddle.fluid import core
 from paddle.fluid.data_feeder import check_type
 from paddle.fluid.dygraph import to_variable
 from paddle.fluid.framework import _dygraph_tracer, dygraph_only
+from paddle.framework import in_dynamic_mode
 
 from .auto_cast import amp_global_state
 
@@ -307,7 +308,7 @@ class AmpScaler:
                         else:
                             param_grads_fp32.append(param._grad_ivar())
         else:
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 # It is very time-consuming to call c++ functions in a loop on the python side.
                 # We put this part of the code on the c++ side to improve the speed in eager mode.
                 (
diff --git a/python/paddle/autograd/autograd.py b/python/paddle/autograd/autograd.py
index e61a7ff0092..96f8ca1bbfa 100644
--- a/python/paddle/autograd/autograd.py
+++ b/python/paddle/autograd/autograd.py
@@ -689,7 +689,7 @@ def _grad_for_jacobian(ys, xs, v=None):
             Tensor is the sum of gradients of outputs with respect to the i-th
             inputs.
     """
-    if paddle.fluid._non_static_mode():
+    if paddle.in_dynamic_mode():
         # paddle.grad returns a list though the inputs is a signle Tensor. The
         # follow code snippet fixes the problem by return the first element of
         # xs_grad when the xs is a signle Tensor.
diff --git a/python/paddle/common_ops_import.py b/python/paddle/common_ops_import.py
index 62a55884f02..042934d66f7 100644
--- a/python/paddle/common_ops_import.py
+++ b/python/paddle/common_ops_import.py
@@ -25,7 +25,6 @@ from paddle.fluid.framework import (  # noqa: F401
     Variable,
     _create_tensor,
     _dygraph_tracer,
-    _non_static_mode,
     convert_np_dtype_to_dtype_,
     default_main_program,
     device_guard,
diff --git a/python/paddle/distributed/auto_parallel/engine.py b/python/paddle/distributed/auto_parallel/engine.py
index cb735fe11b8..728a60e18a4 100644
--- a/python/paddle/distributed/auto_parallel/engine.py
+++ b/python/paddle/distributed/auto_parallel/engine.py
@@ -28,7 +28,7 @@ from paddle.distributed import fleet
 from paddle.fluid.executor import _to_name_str
 from paddle.framework import IrGraph
 from paddle.framework import _current_expected_place as _get_device
-from paddle.framework import core, in_dygraph_mode
+from paddle.framework import core, in_dynamic_mode
 from paddle.metric import Metric
 from paddle.static import InputSpec, Operator, Variable, global_scope
 
@@ -312,7 +312,7 @@ class Engine:
         return inputs_spec, labels_spec
 
     def _prepare_data_tensor(self, inputs_spec, labels_spec, inputs, labels):
-        if in_dygraph_mode() or self._dygraph_mode:
+        if in_dynamic_mode() or self._dygraph_mode:
             raise ValueError("Only support static graph mode.")
 
         if inputs_spec:
@@ -561,7 +561,7 @@ class Engine:
         self._has_prepared[mode] = True
 
     def _build(self, mode):
-        if in_dygraph_mode() or self._dygraph_mode:
+        if in_dynamic_mode() or self._dygraph_mode:
             paddle.disable_static()
             self._dygraph_mode = True
             self._logger.info("Building model with 'to_static' method.")
@@ -1789,7 +1789,7 @@ class Engine:
             self._build(mode)
             self._plan(mode)
         else:
-            if in_dygraph_mode() or self._dygraph_mode:
+            if in_dynamic_mode() or self._dygraph_mode:
                 raise ValueError(
                     "Please call `prepare()` or `fit()` or  `evaluate()` or  `predict()` before calling `cost()`."
                 )
diff --git a/python/paddle/distributed/auto_parallel/process_group.py b/python/paddle/distributed/auto_parallel/process_group.py
index b5652353311..b5669b850b6 100644
--- a/python/paddle/distributed/auto_parallel/process_group.py
+++ b/python/paddle/distributed/auto_parallel/process_group.py
@@ -16,7 +16,7 @@ from collections import OrderedDict
 
 import paddle
 from paddle import _legacy_C_ops
-from paddle.framework import core, in_dygraph_mode
+from paddle.framework import core, in_dynamic_mode
 from paddle.tensor import fill_constant
 
 from ..collective import _get_global_env, _new_ring_id
@@ -177,7 +177,7 @@ class ProcessGroup:
                 )
             tmp = (
                 paddle.to_tensor([1], dtype="int32")
-                if in_dygraph_mode()
+                if in_dynamic_mode()
                 else fill_constant([0], dtype="int32", value="1")
             )
             # use legacy ops
diff --git a/python/paddle/distributed/collective.py b/python/paddle/distributed/collective.py
index 210dff45de2..2011fe19811 100644
--- a/python/paddle/distributed/collective.py
+++ b/python/paddle/distributed/collective.py
@@ -19,7 +19,7 @@ import paddle
 
 # (TODO: GhostScreaming) It will be removed later.
 from paddle.fluid import core
-from paddle.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 from .communication.group import Group, _add_new_group, is_initialized
 from .fleet.layers.mpu.mp_ops import _c_concat  # noqa: F401
@@ -128,7 +128,7 @@ def _set_group_map_backend(group, backend):
 
 def _new_ring_id():
     # NOTE(liyurui): For compatible reason, auto parallel and eager mode relay on previous syntax.
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         global _start_ring_id
         _start_ring_id += 1
         return _start_ring_id + max(_get_global_env().nrings, 9)
@@ -198,7 +198,7 @@ def new_group(ranks=None, backend=None, timeout=_default_timeout):
     """
     global _custom_gid
     global _group_map
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         global _default_group_name
         gid = _custom_gid if _custom_gid else _new_ring_id()
         group_name = _default_group_name + str(gid)
@@ -292,7 +292,7 @@ def new_group(ranks=None, backend=None, timeout=_default_timeout):
     # hang caused by cross-creation of new_group
     tmp = (
         paddle.to_tensor([1], dtype="int32")
-        if in_dygraph_mode()
+        if in_dynamic_mode()
         else paddle.full([0], 1, dtype="int32")
     )
     paddle.distributed.all_reduce(tmp, sync_op=True)
diff --git a/python/paddle/distributed/communication/all_gather.py b/python/paddle/distributed/communication/all_gather.py
index 0e479841af5..47844cebdf0 100644
--- a/python/paddle/distributed/communication/all_gather.py
+++ b/python/paddle/distributed/communication/all_gather.py
@@ -102,7 +102,7 @@ def all_gather_object(object_list, obj, group=None):
             # [{'foo': [1, 2, 3]}, {'bar': [4, 5, 6]}] (2 GPUs)
     """
     assert (
-        framework.in_dygraph_mode()
+        framework.in_dynamic_mode()
     ), "all_gather_object doesn't support static graph mode."
 
     tensor, len_of_tensor = convert_object_to_tensor(obj)
diff --git a/python/paddle/distributed/communication/batch_isend_irecv.py b/python/paddle/distributed/communication/batch_isend_irecv.py
index abdeae5bff5..17d374fe99f 100644
--- a/python/paddle/distributed/communication/batch_isend_irecv.py
+++ b/python/paddle/distributed/communication/batch_isend_irecv.py
@@ -159,7 +159,7 @@ def batch_isend_irecv(p2p_op_list):
     if _warn_cur_rank_not_in_group(group):
         return
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         backend = group.backend
         tasks = []
diff --git a/python/paddle/distributed/communication/broadcast.py b/python/paddle/distributed/communication/broadcast.py
index dd15ee529f1..dc2b6919427 100644
--- a/python/paddle/distributed/communication/broadcast.py
+++ b/python/paddle/distributed/communication/broadcast.py
@@ -102,7 +102,7 @@ def broadcast_object_list(object_list, src, group=None):
             # [{"bar": [4, 5, 6]}] (2 GPUs)
     """
     assert (
-        framework.in_dygraph_mode()
+        framework.in_dynamic_mode()
     ), "broadcast_object_list doesn't support static graph mode."
 
     rank = dist.get_rank()
diff --git a/python/paddle/distributed/communication/gather.py b/python/paddle/distributed/communication/gather.py
index 089f73d1d0c..34d44a7e2b4 100644
--- a/python/paddle/distributed/communication/gather.py
+++ b/python/paddle/distributed/communication/gather.py
@@ -55,6 +55,6 @@ def gather(tensor, gather_list=None, dst=0, group=None, sync_op=True):
             # [] (2 GPUs, out for rank 1)
     """
     assert (
-        framework.in_dygraph_mode()
+        framework.in_dynamic_mode()
     ), "gather doesn't support static graph mode yet."
     return stream.gather(tensor, gather_list, dst, group, sync_op)
diff --git a/python/paddle/distributed/communication/group.py b/python/paddle/distributed/communication/group.py
index 5fff4440877..e722d6fed1d 100644
--- a/python/paddle/distributed/communication/group.py
+++ b/python/paddle/distributed/communication/group.py
@@ -231,7 +231,7 @@ def get_group(id=0):
 
 
 def _sync_calc_stream(tensor):
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         return paddle._legacy_C_ops.c_sync_calc_stream(tensor, tensor)
     else:
         op_type = 'c_sync_calc_stream'
@@ -244,7 +244,7 @@ def _sync_calc_stream(tensor):
 
 
 def _sync_comm_stream(tensor, ring_id=0):
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         return paddle._legacy_C_ops.c_sync_comm_stream(
             [tensor], [tensor], 'ring_id', ring_id
         )
@@ -318,7 +318,7 @@ def barrier(group=None):
     if group is not None and not group.is_member():
         return
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         place = framework._current_expected_place()
         if isinstance(place, framework.CPUPlace):
@@ -332,7 +332,7 @@ def barrier(group=None):
     ring_id = 0 if group is None else group.id
 
     barrier_tensor = paddle.full([1], 1, dtype="int32")
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         return paddle._legacy_C_ops.barrier(
             barrier_tensor, barrier_tensor, 'ring_id', ring_id
         )
diff --git a/python/paddle/distributed/communication/reduce.py b/python/paddle/distributed/communication/reduce.py
index 4ee2142856f..a44c3144889 100644
--- a/python/paddle/distributed/communication/reduce.py
+++ b/python/paddle/distributed/communication/reduce.py
@@ -56,7 +56,7 @@ class ReduceOp:
 
 
 def _get_reduce_op(reduce_op, func_name):
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         if reduce_op == ReduceOp.SUM:
             return framework.core.ReduceOp.SUM
         elif reduce_op == ReduceOp.MAX:
diff --git a/python/paddle/distributed/communication/scatter.py b/python/paddle/distributed/communication/scatter.py
index 2826779d557..f3ae2e358e0 100644
--- a/python/paddle/distributed/communication/scatter.py
+++ b/python/paddle/distributed/communication/scatter.py
@@ -108,7 +108,7 @@ def scatter_object_list(
             # [{'bar': [4, 5, 6]}] (2 GPUs, out for rank 1)
     """
     assert (
-        framework.in_dygraph_mode()
+        framework.in_dynamic_mode()
     ), "scatter_object_list doesn't support static graph mode."
 
     rank = dist.get_rank()
diff --git a/python/paddle/distributed/communication/stream/all_gather.py b/python/paddle/distributed/communication/stream/all_gather.py
index 9f4b19fd5ce..83cbe7ac1fc 100644
--- a/python/paddle/distributed/communication/stream/all_gather.py
+++ b/python/paddle/distributed/communication/stream/all_gather.py
@@ -171,7 +171,7 @@ def all_gather(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         if paddle.is_tensor(tensor_or_tensor_list):
             return _all_gather_into_tensor_in_dygraph(
                 tensor_or_tensor_list, tensor, group, sync_op, use_calc_stream
diff --git a/python/paddle/distributed/communication/stream/all_reduce.py b/python/paddle/distributed/communication/stream/all_reduce.py
index 6b38bffc0bf..8ddeebd43cd 100644
--- a/python/paddle/distributed/communication/stream/all_reduce.py
+++ b/python/paddle/distributed/communication/stream/all_reduce.py
@@ -116,7 +116,7 @@ def all_reduce(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         return _all_reduce_in_dygraph(
             tensor, op, group, sync_op, use_calc_stream
diff --git a/python/paddle/distributed/communication/stream/all_to_all.py b/python/paddle/distributed/communication/stream/all_to_all.py
index df9c72c1da4..2425d1c06c8 100644
--- a/python/paddle/distributed/communication/stream/all_to_all.py
+++ b/python/paddle/distributed/communication/stream/all_to_all.py
@@ -185,7 +185,7 @@ def alltoall(
     if in_tensor_or_tensor_list is None:
         raise RuntimeError("The input should be specified.")
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         out_is_tensor = paddle.is_tensor(out_tensor_or_tensor_list)
         in_is_tensor = paddle.is_tensor(in_tensor_or_tensor_list)
@@ -335,7 +335,7 @@ def alltoall_single(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         return _alltoall_single_in_dygraph(
             out_tensor,
diff --git a/python/paddle/distributed/communication/stream/broadcast.py b/python/paddle/distributed/communication/stream/broadcast.py
index 91fab7c7d5c..2b3634f1a61 100644
--- a/python/paddle/distributed/communication/stream/broadcast.py
+++ b/python/paddle/distributed/communication/stream/broadcast.py
@@ -117,7 +117,7 @@ def broadcast(tensor, src, group=None, sync_op=True, use_calc_stream=False):
             "use_calc_stream can only be True in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         src_rank_in_group = _get_or_throw_group_rank(src, group)
 
diff --git a/python/paddle/distributed/communication/stream/gather.py b/python/paddle/distributed/communication/stream/gather.py
index df3db07eb59..67ec7bf248e 100644
--- a/python/paddle/distributed/communication/stream/gather.py
+++ b/python/paddle/distributed/communication/stream/gather.py
@@ -99,7 +99,7 @@ def gather(
     """
 
     assert (
-        framework.in_dygraph_mode()
+        framework.in_dynamic_mode()
     ), "gather doesn't support static graph mode yet."
 
     if _warn_cur_rank_not_in_group(group):
diff --git a/python/paddle/distributed/communication/stream/recv.py b/python/paddle/distributed/communication/stream/recv.py
index 8fbbfbf0988..d35b005544b 100644
--- a/python/paddle/distributed/communication/stream/recv.py
+++ b/python/paddle/distributed/communication/stream/recv.py
@@ -105,7 +105,7 @@ def recv(tensor, src=0, group=None, sync_op=True, use_calc_stream=False):
             "use_calc_stream can only be True in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         src_rank_in_group = _get_or_throw_group_rank(src, group)
 
diff --git a/python/paddle/distributed/communication/stream/reduce.py b/python/paddle/distributed/communication/stream/reduce.py
index a952e06a910..ba97943103c 100644
--- a/python/paddle/distributed/communication/stream/reduce.py
+++ b/python/paddle/distributed/communication/stream/reduce.py
@@ -131,7 +131,7 @@ def reduce(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         dst_rank_in_group = _get_or_throw_group_rank(dst, group)
         return _reduce_in_dygraph(
diff --git a/python/paddle/distributed/communication/stream/reduce_scatter.py b/python/paddle/distributed/communication/stream/reduce_scatter.py
index 5812ec5d8de..69d1b03aeb2 100644
--- a/python/paddle/distributed/communication/stream/reduce_scatter.py
+++ b/python/paddle/distributed/communication/stream/reduce_scatter.py
@@ -158,7 +158,7 @@ def reduce_scatter(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         if paddle.is_tensor(tensor_or_tensor_list):
             return _reduce_scatter_tensor_in_dygraph(
@@ -243,7 +243,7 @@ def _reduce_scatter_base(
             "use_calc_stream can only be true in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         return _reduce_scatter_tensor_in_dygraph(
             out_tensor,
diff --git a/python/paddle/distributed/communication/stream/scatter.py b/python/paddle/distributed/communication/stream/scatter.py
index 13f9c3ecf64..d6ab2a9305a 100644
--- a/python/paddle/distributed/communication/stream/scatter.py
+++ b/python/paddle/distributed/communication/stream/scatter.py
@@ -201,7 +201,7 @@ def scatter(
             )
         tensor_or_tensor_list = []
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         src_rank_in_group = _get_or_throw_group_rank(src, group)
         if paddle.is_tensor(tensor_or_tensor_list):
diff --git a/python/paddle/distributed/communication/stream/send.py b/python/paddle/distributed/communication/stream/send.py
index 0de989042b9..43c6d6302ca 100644
--- a/python/paddle/distributed/communication/stream/send.py
+++ b/python/paddle/distributed/communication/stream/send.py
@@ -104,7 +104,7 @@ def send(tensor, dst=0, group=None, sync_op=True, use_calc_stream=False):
             "use_calc_stream can only be True in sync op behavior."
         )
 
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = _get_global_group() if group is None else group
         dst_rank_in_group = _get_or_throw_group_rank(dst, group)
 
diff --git a/python/paddle/distributed/fleet/base/role_maker.py b/python/paddle/distributed/fleet/base/role_maker.py
index 0bf7db4a2c3..a490e09c9dd 100755
--- a/python/paddle/distributed/fleet/base/role_maker.py
+++ b/python/paddle/distributed/fleet/base/role_maker.py
@@ -1176,7 +1176,7 @@ class PaddleCloudRoleMaker(RoleMakerBase):
             else:
                 self._collective_env()
             self._role_is_generated = True
-            if not paddle.framework.in_dynamic_mode():
+            if not paddle.in_dynamic_mode():
                 self._gloo_init()
 
 
diff --git a/python/paddle/distributed/fleet/fleet.py b/python/paddle/distributed/fleet/fleet.py
index e9ee02b8e51..39948ab28e6 100755
--- a/python/paddle/distributed/fleet/fleet.py
+++ b/python/paddle/distributed/fleet/fleet.py
@@ -17,9 +17,8 @@ import os
 
 import paddle
 from paddle.fluid import compiler
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.wrapped_decorator import wrap_decorator
-from paddle.framework import _global_flags
+from paddle.framework import _global_flags, in_dynamic_mode
 from paddle.framework.ir import apply_build_strategy
 
 from .base import topology as tp
@@ -281,7 +280,7 @@ class Fleet:
                         "CUDA_VISIBLE_DEVICES shoule be set only 1 card if you use `python` to launch fleet program."
                     )
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if self.worker_num() == 1:
                 # if worker_num is 1, should construct default topology & hcg
                 self._topology = tp.CommunicateTopology()
@@ -1270,7 +1269,7 @@ class Fleet:
             )
         else:
             if (
-                in_dygraph_mode()
+                in_dynamic_mode()
                 or self._role_maker._is_non_distributed()
                 or self._is_collective
             ):
@@ -1286,7 +1285,7 @@ class Fleet:
         context["user_defined_strategy"] = copy.deepcopy(
             self._user_defined_strategy
         )
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             # imitate target optimizer retrieval
             target_opt = self.user_defined_optimizer
             self._context = context
diff --git a/python/paddle/distributed/fleet/layers/mpu/mp_ops.py b/python/paddle/distributed/fleet/layers/mpu/mp_ops.py
index b435795aded..322281d3c9f 100644
--- a/python/paddle/distributed/fleet/layers/mpu/mp_ops.py
+++ b/python/paddle/distributed/fleet/layers/mpu/mp_ops.py
@@ -16,7 +16,7 @@ import paddle
 from paddle import _legacy_C_ops
 from paddle.distributed import collective
 from paddle.fluid.data_feeder import check_dtype, check_variable_and_dtype
-from paddle.framework import LayerHelper, _create_tensor, in_dygraph_mode
+from paddle.framework import LayerHelper, _create_tensor, in_dynamic_mode
 from paddle.nn import Layer
 from paddle.nn.utils import dygraph_utils
 
@@ -39,7 +39,7 @@ def _c_identity(tensor, group=None):
         return
     ring_id = 0 if group is None else group.id
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         from paddle.autograd import PyLayer
 
         class c_identity_eager(PyLayer):
@@ -108,7 +108,7 @@ def _c_concat(tensor, group=None):
     rank = group.rank
     nranks = group.nranks
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.c_concat(
             tensor,
             'ring_id',
@@ -174,7 +174,7 @@ def _c_split(tensor, group=None):
         else group.nranks
     )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.c_split(
             tensor,
             'use_calc_stream',
@@ -226,7 +226,7 @@ def _mp_allreduce(
     if group is not None and not group.is_member():
         return
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         group = collective._get_default_group() if group is None else group
         assert op == ReduceOp.SUM, f"Unknown parameter: {op}."
 
@@ -308,7 +308,7 @@ def _c_lookup_table(table, index, start_index=0, name=None):
     Returns:
         Tensor.
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.c_embedding(
             table, index, "start_index", start_index
         )
@@ -401,7 +401,7 @@ def _c_softmax_with_cross_entropy(
     if input_dims - 1 == label_dims:
         label = paddle.unsqueeze(label, axis=-1)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         softmax, loss = _legacy_C_ops.c_softmax_with_cross_entropy(
             logits,
             label,
@@ -445,7 +445,7 @@ def _linear(x, weight, bias=None, name=None):
     """
     Fuction Linear
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         pre_bias = _create_tensor(dtype=x.dtype)
         _legacy_C_ops.matmul(
             x,
@@ -810,7 +810,7 @@ def split(
             supported_operations
         )
     )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         raise ValueError(
             "paddle.distributed.split cannot be used in dynamic "
             "graph mode, plese use ParallelEmbedding, ParallelRowLinear, "
diff --git a/python/paddle/distributed/fleet/layers/mpu/random.py b/python/paddle/distributed/fleet/layers/mpu/random.py
index 1447fccb66e..aabb8dc3fa1 100644
--- a/python/paddle/distributed/fleet/layers/mpu/random.py
+++ b/python/paddle/distributed/fleet/layers/mpu/random.py
@@ -21,8 +21,7 @@ from paddle import _legacy_C_ops
 from paddle.common_ops_import import Variable
 from paddle.fluid import core
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import in_dygraph_mode
-from paddle.framework import LayerHelper
+from paddle.framework import LayerHelper, in_dynamic_mode
 
 __all__ = []
 
@@ -218,7 +217,7 @@ def dropout(
     )  # semantic transfer
 
     # dygraph using tracker, doesn't need determinate seed
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out, mask = _legacy_C_ops.dropout(
             x,
             'dropout_prob',
diff --git a/python/paddle/distributed/fleet/meta_optimizers/dgc_optimizer.py b/python/paddle/distributed/fleet/meta_optimizers/dgc_optimizer.py
index 27c1a5b651b..f1967b5e1de 100644
--- a/python/paddle/distributed/fleet/meta_optimizers/dgc_optimizer.py
+++ b/python/paddle/distributed/fleet/meta_optimizers/dgc_optimizer.py
@@ -22,7 +22,7 @@ import paddle
 from paddle.common_ops_import import LayerHelper
 from paddle.fluid.dygraph import base as imperative_base
 from paddle.fluid.optimizer import Momentum, Optimizer
-from paddle.framework import core, in_dygraph_mode
+from paddle.framework import core, in_dynamic_mode
 from paddle.nn.clip import ClipGradByNorm, append_gradient_clip_ops
 from paddle.regularizer import L1Decay, L2Decay
 from paddle.static import create_global_var
@@ -46,7 +46,7 @@ class DGCMomentumOptimizer(Optimizer):
         grad_clip=None,
         name=None,
     ):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             raise Exception("In dygraph, don't support DGCMomentumOptimizer.")
 
         assert (
diff --git a/python/paddle/distributed/fleet/meta_parallel/parallel_layers/pp_layers.py b/python/paddle/distributed/fleet/meta_parallel/parallel_layers/pp_layers.py
index bc403da76ba..c18d6311618 100755
--- a/python/paddle/distributed/fleet/meta_parallel/parallel_layers/pp_layers.py
+++ b/python/paddle/distributed/fleet/meta_parallel/parallel_layers/pp_layers.py
@@ -533,7 +533,7 @@ class PipelineLayer(nn.Layer):
         for key, comm in self.shared_comm.items():
             param = getattr(self.shared_layers[key], comm['weight_attr'])
             # need use trace_op to allreduce weight
-            if framework.in_dygraph_mode():
+            if framework.in_dynamic_mode():
                 with paddle.framework.no_grad():
                     paddle.distributed.all_reduce(
                         param.grad
diff --git a/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py b/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py
index 5ff01d35ec5..84a32379866 100755
--- a/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py
+++ b/python/paddle/distributed/fleet/meta_parallel/pipeline_parallel.py
@@ -555,7 +555,7 @@ class PipelineParallelWithInterleave(PipelineParallel):
         super().__init__(layers=layers, hcg=hcg, strategy=strategy)
         assert layers.get_num_virtual_stages() > 1
         assert (
-            framework.in_dygraph_mode()
+            framework.in_dynamic_mode()
         ), "virtual pipeline stage with interleave only support eager dygraph mode"
         # setup for interleave scheduler
         self.num_model_chunks = layers.get_num_virtual_stages()
diff --git a/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py b/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py
index 1da86c83640..4983de3f2ab 100644
--- a/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py
+++ b/python/paddle/distributed/fleet/meta_parallel/pp_utils/p2p_communication.py
@@ -220,7 +220,7 @@ def _partial_send_op(
     tensor, group, use_calc_stream, ring_id, dst, nranks, rank_id
 ):
     dst_rank_in_group = dst if group is None else group.get_group_rank(dst)
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         group = (
             paddle.distributed.collective._get_default_group()
             if group is None
@@ -291,7 +291,7 @@ def recv_partial(
     else:
         if use_calc_stream:
             recv_op = paddle.distributed.recv
-        elif framework.in_dygraph_mode():
+        elif framework.in_dynamic_mode():
             recv_op = paddle.distributed.irecv
         return recv_op(tensor.detach(), src=src_rank, group=group)
 
@@ -656,7 +656,7 @@ def _p2p_helper(
                     tasks.append(task)
         _xpu_comm_group_end()
     if not sync_recv:
-        if framework.in_dygraph_mode():
+        if framework.in_dynamic_mode():
             # wait irecv tasks in eager dygraph mode with new comm library
             for task in tasks:
                 assert task is not None
diff --git a/python/paddle/distributed/fleet/meta_parallel/sharding/group_sharded_stage3.py b/python/paddle/distributed/fleet/meta_parallel/sharding/group_sharded_stage3.py
index bdf0ec29cd7..b1c47de593b 100644
--- a/python/paddle/distributed/fleet/meta_parallel/sharding/group_sharded_stage3.py
+++ b/python/paddle/distributed/fleet/meta_parallel/sharding/group_sharded_stage3.py
@@ -38,7 +38,7 @@ def _all_gather(tensor, buffer_size, group):
     """
 
     assert group is not None
-    if framework.in_dygraph_mode():
+    if framework.in_dynamic_mode():
         out = paddle.zeros([buffer_size], dtype=tensor.dtype)
         task = group.process_group.all_gather(tensor, out)
         return out, task
diff --git a/python/paddle/distributed/fleet/optimizer.py b/python/paddle/distributed/fleet/optimizer.py
index 5abe7c47e9b..6fc414d0a65 100755
--- a/python/paddle/distributed/fleet/optimizer.py
+++ b/python/paddle/distributed/fleet/optimizer.py
@@ -15,7 +15,7 @@
 import copy
 
 from paddle.distributed import fleet
-from paddle.fluid.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 from .meta_optimizers import HeterParallelOptimizer, HybridParallelOptimizer
 from .utils.log_util import logger
@@ -81,7 +81,7 @@ def _dygraph_distributed_optimizer(optimizer, strategy=None):
 
 
 def distributed_optimizer(*args, **kwargs):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _dygraph_distributed_optimizer(*args, **kwargs)
     else:
         return fleet.fleet.distributed_optimizer(*args, **kwargs)
diff --git a/python/paddle/distributed/fleet/recompute/recompute.py b/python/paddle/distributed/fleet/recompute/recompute.py
index 8bbbe8e4e7e..ec0690a10bc 100755
--- a/python/paddle/distributed/fleet/recompute/recompute.py
+++ b/python/paddle/distributed/fleet/recompute/recompute.py
@@ -21,7 +21,7 @@ from paddle.autograd import PyLayer
 from paddle.distributed.fleet.meta_parallel.parallel_layers.random import (
     get_rng_state_tracker,
 )
-from paddle.framework import core, in_dygraph_mode
+from paddle.framework import core, in_dynamic_mode
 
 from ..utils.log_util import logger
 
@@ -198,7 +198,7 @@ class RecomputeFunction(PyLayer):
                     forward_outputs_with_grad, backward_inputs_with_grad
                 )
 
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 grads = tuple(
                     inp._grad_ivar()
                     for inp in detached_inputs
diff --git a/python/paddle/distributed/fleet/recompute/recompute_hybrid.py b/python/paddle/distributed/fleet/recompute/recompute_hybrid.py
index a5689020eb0..c83cf549217 100644
--- a/python/paddle/distributed/fleet/recompute/recompute_hybrid.py
+++ b/python/paddle/distributed/fleet/recompute/recompute_hybrid.py
@@ -161,7 +161,7 @@ class _HPRecomputeFunction(PyLayer):
                 #  If not marked non_differentiable, all output tensors' attr `stop gradient`
                 #  will be reset to `False` in c++ backend.
                 #  See https://github.com/PaddlePaddle/Paddle/blob/9d62efb0e6e5373823039d9eda96cd5905426c0a/paddle/fluid/pybind/eager_py_layer.cc#L388
-                if framework.in_dygraph_mode() and state:
+                if framework.in_dynamic_mode() and state:
                     ctx.mark_non_differentiable(arg)
             else:
                 ctx.inputs.append(arg)
diff --git a/python/paddle/distributed/fleet/utils/hybrid_parallel_inference.py b/python/paddle/distributed/fleet/utils/hybrid_parallel_inference.py
index 68b5581d3be..cdec2cfd63f 100644
--- a/python/paddle/distributed/fleet/utils/hybrid_parallel_inference.py
+++ b/python/paddle/distributed/fleet/utils/hybrid_parallel_inference.py
@@ -20,8 +20,7 @@ from paddle.distributed import fleet
 
 # (TODO: GhostScreaming) It will be removed later.
 from paddle.fluid import core
-from paddle.fluid.framework import in_dygraph_mode
-from paddle.framework import Block, Program
+from paddle.framework import Block, Program, in_dynamic_mode
 
 
 class HybridParallelInferenceHelper:
@@ -205,7 +204,7 @@ class HybridParallelInferenceHelper:
             self._device = "gpu"
         assert self._device, "Only gpu are supported."
 
-        assert not in_dygraph_mode(), "Only static graph mode is supported."
+        assert not in_dynamic_mode(), "Only static graph mode is supported."
 
         op_maker = core.op_proto_and_checker_maker
         self._op_role = op_maker.OpRole
diff --git a/python/paddle/distributed/fleet/utils/hybrid_parallel_util.py b/python/paddle/distributed/fleet/utils/hybrid_parallel_util.py
index cd0455c8069..e07fee4fb56 100644
--- a/python/paddle/distributed/fleet/utils/hybrid_parallel_util.py
+++ b/python/paddle/distributed/fleet/utils/hybrid_parallel_util.py
@@ -17,7 +17,7 @@ from paddle import framework
 from paddle.distributed.parallel import (
     _split_tensors,
     build_groups,
-    in_dygraph_mode,
+    in_dynamic_mode,
     sync_params_buffers,
 )
 
@@ -131,7 +131,7 @@ def _broadcast_data_help(data, shape, dtype, hcg):
     )
 
     if mp_rank != 0:
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             data._clear_data()
             input_data._share_buffer_to(data)
         else:
@@ -174,7 +174,7 @@ def broadcast_input_data(hcg, *inputs, **kwargs):
     for v in inputs:
         if isinstance(v, core.eager.Tensor):
             with framework.no_grad():
-                if in_dygraph_mode() and not eval(f"v.place.is_{dev}_place")():
+                if in_dynamic_mode() and not eval(f"v.place.is_{dev}_place")():
                     v_gpu = v._copy_to(place, True)
                     v._clear_data()
                     v_gpu._share_buffer_to(v)
@@ -185,7 +185,7 @@ def broadcast_input_data(hcg, *inputs, **kwargs):
     for k, v in kwargs.items():
         if isinstance(v, core.eager.Tensor):
             with framework.no_grad():
-                if in_dygraph_mode() and not eval(f"v.place.is_{dev}_place")():
+                if in_dynamic_mode() and not eval(f"v.place.is_{dev}_place")():
                     v_gpu = v._copy_to(place, True)
                     v._clear_data()
                     v_gpu._share_buffer_to(v)
@@ -217,7 +217,7 @@ def fused_allreduce_gradients_with_group(
 ):
     apply_func = (
         _apply_collective_grads_eager
-        if in_dygraph_mode()
+        if in_dynamic_mode()
         else _apply_collective_grads
     )
     with framework.no_grad():
diff --git a/python/paddle/distributed/fleet/utils/mix_precision_utils.py b/python/paddle/distributed/fleet/utils/mix_precision_utils.py
index 9e172d840bd..ef8357cc309 100644
--- a/python/paddle/distributed/fleet/utils/mix_precision_utils.py
+++ b/python/paddle/distributed/fleet/utils/mix_precision_utils.py
@@ -120,7 +120,7 @@ class MixPrecisionOptimizer:
                 if param.stop_gradient:
                     continue
                 grad_var = param.main_grad
-                if framework.in_dygraph_mode():
+                if paddle.in_dynamic_mode():
                     if (
                         hasattr(grad_var, "is_selected_rows")
                         and grad_var.is_selected_rows()
@@ -151,7 +151,7 @@ class MixPrecisionOptimizer:
                     if param.stop_gradient:
                         continue
                     grad_var = param.main_grad
-                    if framework.in_dygraph_mode():
+                    if paddle.in_dynamic_mode():
                         if (
                             hasattr(grad_var, "is_selected_rows")
                             and grad_var.is_selected_rows()
diff --git a/python/paddle/distributed/models/moe/utils.py b/python/paddle/distributed/models/moe/utils.py
index 8a9d199ceea..69d43862bdb 100644
--- a/python/paddle/distributed/models/moe/utils.py
+++ b/python/paddle/distributed/models/moe/utils.py
@@ -14,7 +14,7 @@
 
 from paddle import _legacy_C_ops
 from paddle.common_ops_import import check_variable_and_dtype
-from paddle.framework import LayerHelper, in_dygraph_mode
+from paddle.framework import LayerHelper, in_dynamic_mode
 
 
 def _number_count(numbers, upper_range):
@@ -39,7 +39,7 @@ def _number_count(numbers, upper_range):
             number_count = paddle.distributed.utils.number_count(numbers, upper_range)
             print(number_count) # the result: [2, 0, 2, 0, 0, 0]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.number_count(numbers, 'upper_range', upper_range)
     else:
         op_type = 'number_count'
@@ -86,7 +86,7 @@ def _assign_pos(x, cum_count):
             pos = paddle.distributed.utils.assign_pos(x=numbers, cum_count=num_cum)
             print(pos) # the result: (2, 0, 3, 1)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.assign_pos(x, cum_count, cum_count[-1])
     else:
         op_type = 'assign_pos'
@@ -121,7 +121,7 @@ def _random_routing(topk_idx, topk_value, prob, topk=2):
         prob: random prob, shape=(topk_idx.shape[0],)
     """
     if topk == 2:
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _legacy_C_ops.random_routing(prob, topk_value, topk_idx)
         else:
             raise RuntimeError("Not supporting static graph mode now")
@@ -150,7 +150,7 @@ def _limit_by_capacity(expert_count, capacity, n_worker):
             out = paddle.distributed.utils.limit_by_capacity(expert_count, capacity, n_work)
             print(out) # the result: [1, 2, 2, 4, 3, 3]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.limit_by_capacity(
             expert_count, capacity, 'n_worker', n_worker
         )
@@ -195,7 +195,7 @@ def _prune_gate_by_capacity(gate_idx, expert_count, n_expert, n_worker):
             # Tensor(shape=[8], dtype=int32, place=CUDAPlace(0), stop_gradient=True,
               [1, 3, 3, 3, -1, 2, 1, 1])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.prune_gate_by_capacity(
             gate_idx, expert_count, "n_expert", n_expert, "n_worker", n_worker
         )
diff --git a/python/paddle/distributed/parallel.py b/python/paddle/distributed/parallel.py
index 1737d0731a9..f8888b9f71c 100644
--- a/python/paddle/distributed/parallel.py
+++ b/python/paddle/distributed/parallel.py
@@ -47,7 +47,7 @@ from paddle.distributed.fleet.launch_utils import check_backend
 # (TODO: GhostScreaming) It will be removed later.
 from paddle.framework import _set_expected_place
 from paddle.framework import base as imperative_base
-from paddle.framework import core, in_dygraph_mode
+from paddle.framework import core, in_dynamic_mode
 from paddle.nn.layer import layers
 from paddle.utils import deprecated
 
@@ -101,7 +101,7 @@ def _reshape_inplace(x, shape):
 
 @framework.dygraph_only
 def _split_tensors(coalesced_grads_and_grad_vars):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         for (
             coalesced_grad,
             origin_grad_vars,
@@ -356,7 +356,7 @@ class DataParallel(layers.Layer):
         super().__init__(layers.full_name() + "_data_parallel")
 
         assert (
-            in_dygraph_mode()
+            in_dynamic_mode()
         ), "It's not supported to construct DataParallel in static graph mode."
 
         self._layers = layers
@@ -381,7 +381,7 @@ class DataParallel(layers.Layer):
                 "constructing the DataParallel."
             )
 
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 self.group = (
                     paddle.distributed.collective._get_default_group()
                     if self.group is None
@@ -456,7 +456,7 @@ class DataParallel(layers.Layer):
             check_layer_sparse(sublayer) for sublayer, _ in layers_param
         ]
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             self.group_indices = core.eager_assign_group_by_size(
                 trainable_parameters,
                 is_sparse_gradient,
@@ -1041,7 +1041,7 @@ def init_parallel_env():
 
     group = None
 
-    if backend in _valid_backend_list and in_dygraph_mode():
+    if backend in _valid_backend_list and in_dynamic_mode():
         if _default_group_name in _get_group_map_by_name():
             return _get_group_map_by_name()[_default_group_name]
         _set_default_backend(backend)
@@ -1212,7 +1212,7 @@ def get_rank(group=None):
             print("The rank is %d" % dist.get_rank())
             # The rank is 0
     """
-    if in_dygraph_mode() and group:
+    if in_dynamic_mode() and group:
         return group.rank
 
     assert group is None, "Only support group argument in eager mode."
@@ -1244,7 +1244,7 @@ def get_world_size(group=None):
             print("The world_size is %d" % dist.get_world_size())
             # The world_size is 1
     """
-    if in_dygraph_mode() and group:
+    if in_dynamic_mode() and group:
         return group.world_size
 
     assert group is None, "Only support group argument in eager mode."
diff --git a/python/paddle/distributed/utils/moe_utils.py b/python/paddle/distributed/utils/moe_utils.py
index fa003c1a00d..ff0d04b0742 100644
--- a/python/paddle/distributed/utils/moe_utils.py
+++ b/python/paddle/distributed/utils/moe_utils.py
@@ -14,7 +14,7 @@
 
 from paddle import _legacy_C_ops
 from paddle.common_ops_import import check_variable_and_dtype
-from paddle.framework import LayerHelper, in_dygraph_mode
+from paddle.framework import LayerHelper, in_dynamic_mode
 
 
 def global_scatter(
@@ -102,7 +102,7 @@ def global_scatter(
         return
 
     ring_id = 0 if group is None else group.id
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.global_scatter(
             x,
             local_count,
@@ -219,7 +219,7 @@ def global_gather(
         return
 
     ring_id = 0 if group is None else group.id
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.global_gather(
             x,
             local_count,
diff --git a/python/paddle/distribution/bernoulli.py b/python/paddle/distribution/bernoulli.py
index e961ee9d018..0cecc6024f3 100644
--- a/python/paddle/distribution/bernoulli.py
+++ b/python/paddle/distribution/bernoulli.py
@@ -18,8 +18,8 @@ import numpy as np
 import paddle
 from paddle.distribution import exponential_family
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layers import tensor
+from paddle.framework import in_dynamic_mode
 from paddle.nn.functional import (
     binary_cross_entropy_with_logits,
     sigmoid,
@@ -93,7 +93,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
 
     def __init__(self, probs, name=None):
         self.name = name or 'Bernoulli'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 probs,
                 'probs',
@@ -110,7 +110,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
             self.dtype = paddle.get_default_dtype()
 
         # Check probs range [0, 1].
-        if _non_static_mode():
+        if in_dynamic_mode():
             """Not use `paddle.any` in static mode, which always be `True`."""
             if (
                 paddle.any(self.probs < 0)
@@ -176,7 +176,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 # [100, 2, 2]
         """
         name = self.name + '_sample'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 shape,
                 'shape',
@@ -255,7 +255,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 #        288.66418457)
         """
         name = self.name + '_rsample'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 shape,
                 'shape',
@@ -317,7 +317,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 #        [1.])
         """
         name = self.name + '_cdf'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(value, 'value', tensor.Variable, name)
 
         value = self._check_values_dtype_in_probs(self.probs, value)
@@ -355,7 +355,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 #        [-1.20397282])
         """
         name = self.name + '_log_prob'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(value, 'value', tensor.Variable, name)
 
         value = self._check_values_dtype_in_probs(self.probs, value)
@@ -394,7 +394,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 #        [0.29999998])
         """
         name = self.name + '_prob'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(value, 'value', tensor.Variable, name)
 
         return self.log_prob(value).exp(name=name)
@@ -459,7 +459,7 @@ class Bernoulli(exponential_family.ExponentialFamily):
                 #        0.33891910)
         """
         name = self.name + '_kl_divergence'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(other, 'other', Bernoulli, name)
 
         a_logits = self.logits
diff --git a/python/paddle/distribution/categorical.py b/python/paddle/distribution/categorical.py
index 7a02942b89c..6e3f7c44697 100644
--- a/python/paddle/distribution/categorical.py
+++ b/python/paddle/distribution/categorical.py
@@ -17,8 +17,8 @@ import numpy as np
 import paddle
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layers import tensor
+from paddle.framework import in_dynamic_mode
 from paddle.tensor import multinomial
 
 
@@ -90,7 +90,7 @@ class Categorical(distribution.Distribution):
             logits(list|tuple|numpy.ndarray|Tensor): The logits input of categorical distribution. The data type is float32 or float64.
             name(str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
         """
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 logits,
                 'logits',
@@ -146,7 +146,7 @@ class Categorical(distribution.Distribution):
 
         """
         name = self.name + '_sample'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(shape, 'shape', (list), 'sample')
 
         num_samples = np.prod(np.array(shape))
@@ -208,7 +208,7 @@ class Categorical(distribution.Distribution):
 
         """
         name = self.name + '_kl_divergence'
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(other, 'other', Categorical, 'kl_divergence')
 
         logits = self.logits - paddle.max(self.logits, axis=-1, keepdim=True)
diff --git a/python/paddle/distribution/dirichlet.py b/python/paddle/distribution/dirichlet.py
index ca55285540d..5639ef8e3c3 100644
--- a/python/paddle/distribution/dirichlet.py
+++ b/python/paddle/distribution/dirichlet.py
@@ -15,8 +15,8 @@
 import paddle
 from paddle.distribution import exponential_family
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 
 class Dirichlet(exponential_family.ExponentialFamily):
@@ -159,7 +159,7 @@ class Dirichlet(exponential_family.ExponentialFamily):
 
 def _dirichlet(concentration, name=None):
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return paddle._C_ops.dirichlet(concentration)
     else:
         op_type = 'dirichlet'
diff --git a/python/paddle/distribution/distribution.py b/python/paddle/distribution/distribution.py
index ae532f45f0a..af63c201810 100644
--- a/python/paddle/distribution/distribution.py
+++ b/python/paddle/distribution/distribution.py
@@ -26,8 +26,8 @@ import numpy as np
 import paddle
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype, convert_dtype
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layers import tensor
+from paddle.framework import in_dynamic_mode
 
 
 class Distribution:
@@ -221,7 +221,7 @@ class Distribution:
         Returns:
             value (Tensor): Change value's dtype if value's dtype is different from param.
         """
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if value.dtype != param.dtype and convert_dtype(value.dtype) in [
                 'float32',
                 'float64',
diff --git a/python/paddle/distribution/exponential_family.py b/python/paddle/distribution/exponential_family.py
index 0939e73f726..fbe7e71ebb1 100644
--- a/python/paddle/distribution/exponential_family.py
+++ b/python/paddle/distribution/exponential_family.py
@@ -14,7 +14,7 @@
 
 import paddle
 from paddle.distribution import distribution
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 
 class ExponentialFamily(distribution.Distribution):
@@ -61,7 +61,7 @@ class ExponentialFamily(distribution.Distribution):
 
         log_norm = self._log_normalizer(*natural_parameters)
 
-        if _non_static_mode():
+        if in_dynamic_mode():
             grads = paddle.grad(
                 log_norm.sum(), natural_parameters, create_graph=True
             )
diff --git a/python/paddle/distribution/kl.py b/python/paddle/distribution/kl.py
index 40be41e06c5..1b6e1a4b8d9 100644
--- a/python/paddle/distribution/kl.py
+++ b/python/paddle/distribution/kl.py
@@ -27,7 +27,7 @@ from paddle.distribution.laplace import Laplace
 from paddle.distribution.lognormal import LogNormal
 from paddle.distribution.normal import Normal
 from paddle.distribution.uniform import Uniform
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 __all__ = ["register_kl", "kl_divergence"]
 
@@ -229,7 +229,7 @@ def _kl_expfamily_expfamily(p, q):
     p_log_norm = p._log_normalizer(*p_natural_params)
 
     try:
-        if _non_static_mode():
+        if in_dynamic_mode():
             p_grads = paddle.grad(
                 p_log_norm, p_natural_params, create_graph=True
             )
diff --git a/python/paddle/distribution/normal.py b/python/paddle/distribution/normal.py
index 31a3750a048..7ba987819a3 100644
--- a/python/paddle/distribution/normal.py
+++ b/python/paddle/distribution/normal.py
@@ -20,8 +20,8 @@ import numpy as np
 import paddle
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layers import tensor
+from paddle.framework import in_dynamic_mode
 from paddle.tensor import random
 
 
@@ -87,7 +87,7 @@ class Normal(distribution.Distribution):
     """
 
     def __init__(self, loc, scale, name=None):
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 loc,
                 'loc',
@@ -166,7 +166,7 @@ class Normal(distribution.Distribution):
         if not isinstance(shape, Iterable):
             raise TypeError('sample shape must be Iterable object.')
 
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(seed, 'seed', (int), 'sample')
 
         shape = list(shape)
@@ -321,7 +321,7 @@ class Normal(distribution.Distribution):
             Tensor, kl-divergence between two normal distributions.The data type is float32.
 
         """
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(other, 'other', Normal, 'kl_divergence')
 
         name = self.name + '_kl_divergence'
diff --git a/python/paddle/distribution/uniform.py b/python/paddle/distribution/uniform.py
index 8547c6f9900..1d2cb885b56 100644
--- a/python/paddle/distribution/uniform.py
+++ b/python/paddle/distribution/uniform.py
@@ -18,8 +18,8 @@ import paddle
 from paddle import _C_ops
 from paddle.distribution import distribution
 from paddle.fluid.data_feeder import check_type, convert_dtype
-from paddle.fluid.framework import _non_static_mode, in_dygraph_mode
 from paddle.fluid.layers import tensor
+from paddle.framework import in_dynamic_mode
 from paddle.tensor import random
 
 
@@ -92,7 +92,7 @@ class Uniform(distribution.Distribution):
     """
 
     def __init__(self, low, high, name=None):
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(
                 low,
                 'low',
@@ -152,7 +152,7 @@ class Uniform(distribution.Distribution):
             Tensor, A tensor with prepended dimensions shape. The data type is float32.
 
         """
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             check_type(shape, 'shape', (list), 'sample')
             check_type(seed, 'seed', (int), 'sample')
 
@@ -205,7 +205,7 @@ class Uniform(distribution.Distribution):
 
         """
         value = self._check_values_dtype_in_probs(self.low, value)
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             # ensure value in [low, high]
             lb_bool = self.low < value
             ub_bool = value < self.high
@@ -234,7 +234,7 @@ class Uniform(distribution.Distribution):
 
         """
         value = self._check_values_dtype_in_probs(self.low, value)
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             lb_bool = self.low < value
             ub_bool = value < self.high
             lb = _C_ops.cast(lb_bool, value.dtype)
diff --git a/python/paddle/fft.py b/python/paddle/fft.py
index 438c65ae2f0..3e3b88cc3aa 100644
--- a/python/paddle/fft.py
+++ b/python/paddle/fft.py
@@ -20,8 +20,8 @@ import paddle
 
 from . import _C_ops
 from .fluid.data_feeder import check_variable_and_dtype
-from .fluid.framework import in_dygraph_mode
 from .fluid.layer_helper import LayerHelper
+from .framework import in_dynamic_mode
 from .tensor.attribute import is_floating_point, is_integer
 from .tensor.creation import _complex_to_real_dtype, _real_to_complex_dtype
 
@@ -1437,7 +1437,7 @@ def fft_c2c(x, n, axis, norm, forward, name):
         s = [n]
         x = _resize_fft_input(x, s, axes)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.fft_c2c(x, axes, norm, forward)
     else:
         op_type = 'fft_c2c'
@@ -1468,7 +1468,7 @@ def fft_r2c(x, n, axis, norm, forward, onesided, name):
         _check_fft_n(n)
         s = [n]
         x = _resize_fft_input(x, s, axes)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.fft_r2c(x, axes, norm, forward, onesided)
     else:
         op_type = 'fft_r2c'
@@ -1511,7 +1511,7 @@ def fft_c2r(x, n, axis, norm, forward, name):
         s = [n // 2 + 1]
         x = _resize_fft_input(x, s, axes)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if n is not None:
             out = _C_ops.fft_c2r(x, axes, norm, forward, n)
         else:
@@ -1570,7 +1570,7 @@ def fftn_c2c(x, s, axes, norm, forward, name):
     if s is not None:
         x = _resize_fft_input(x, s, axes)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.fft_c2c(x, axes, norm, forward)
     else:
         op_type = 'fft_c2c'
@@ -1620,7 +1620,7 @@ def fftn_r2c(x, s, axes, norm, forward, onesided, name):
     if s is not None:
         x = _resize_fft_input(x, s, axes)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.fft_r2c(x, axes, norm, forward, onesided)
     else:
         op_type = 'fft_r2c'
@@ -1684,7 +1684,7 @@ def fftn_c2r(x, s, axes, norm, forward, name):
         fft_input_shape[-1] = fft_input_shape[-1] // 2 + 1
         x = _resize_fft_input(x, fft_input_shape, axes)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if s is not None:
             out = _C_ops.fft_c2r(x, axes, norm, forward, s[-1])
         else:
diff --git a/python/paddle/fluid/contrib/optimizer.py b/python/paddle/fluid/contrib/optimizer.py
index 57f6e8114db..e97b9d8c49c 100644
--- a/python/paddle/fluid/contrib/optimizer.py
+++ b/python/paddle/fluid/contrib/optimizer.py
@@ -229,7 +229,7 @@ class Momentum(Optimizer):
             else None
         )
 
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             _, _, _ = _legacy_C_ops.momentum(
                 param_and_grad[0],
                 param_and_grad[1],
diff --git a/python/paddle/fluid/data_feeder.py b/python/paddle/fluid/data_feeder.py
index ef711fb4fa1..687504e3670 100644
--- a/python/paddle/fluid/data_feeder.py
+++ b/python/paddle/fluid/data_feeder.py
@@ -22,8 +22,8 @@ import struct
 from .framework import (
     Variable,
     default_main_program,
+    in_dygraph_mode,
     _current_expected_place,
-    _non_static_mode,
 )
 from .framework import _cpu_num, _cuda_ids
 
@@ -140,7 +140,7 @@ def check_type(input, input_name, expected_type, op_name, extra_message=''):
     # in dynamic graph mode.
     # 2. Performance considerations. Because these checks are executed at
     # each step in dynamic graph mode, it will bring a heavy performance burden.
-    if _non_static_mode():
+    if in_dygraph_mode():
         return
 
     # NOTE: `in_declarative_mode` is used to determined whether this op is called under
@@ -171,7 +171,7 @@ def check_dtype(
     input_dtype, input_name, expected_dtype, op_name, extra_message=''
 ):
     # See NOTE [ Why skip dynamic graph check ]
-    if _non_static_mode():
+    if in_dygraph_mode():
         return
     if convert_dtype(input_dtype) in ['float16']:
         warnings.warn(
@@ -208,7 +208,7 @@ def check_shape(
     expected_tensor_dtype=('int32', 'int64'),
 ):
     # See NOTE [ Why skip dynamic graph check ]
-    if _non_static_mode():
+    if in_dygraph_mode():
         return
     check_type(shape, 'shape', expected_shape_type, op_name)
     if expected_element_type is not None and not isinstance(shape, Variable):
diff --git a/python/paddle/fluid/dygraph/base.py b/python/paddle/fluid/dygraph/base.py
index 3430d78d6d7..538f5a0492f 100644
--- a/python/paddle/fluid/dygraph/base.py
+++ b/python/paddle/fluid/dygraph/base.py
@@ -106,7 +106,7 @@ def program_desc_tracing_guard(enable):
 @signature_safe_contextmanager
 def param_guard(parameters):
     # Note: parameters is a reference of self._parameters or self._buffers
-    if in_declarative_mode() and not framework.in_dygraph_mode() and parameters:
+    if in_declarative_mode() and not paddle.in_dynamic_mode() and parameters:
         origin_parameters = parameters.copy()
         for name, var_base in parameters.items():
             if isinstance(var_base, list):
diff --git a/python/paddle/fluid/dygraph/tensor_patch_methods.py b/python/paddle/fluid/dygraph/tensor_patch_methods.py
index ad51d8ff31a..58f99b9e98a 100644
--- a/python/paddle/fluid/dygraph/tensor_patch_methods.py
+++ b/python/paddle/fluid/dygraph/tensor_patch_methods.py
@@ -270,7 +270,7 @@ def monkey_patch_tensor():
                 # 4: [5000.]
 
         """
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             if in_profiler_mode():
                 record_event = profiler.RecordEvent(
                     "Gradient Backward", profiler.TracerEventType.Backward
@@ -978,21 +978,20 @@ def monkey_patch_tensor():
         ("values", values),
         ("to_dense", to_dense),
         ("to_sparse_coo", to_sparse_coo),
+        ("_set_grad_ivar", _set_grad_ivar),
+        ("value", value),
+        ("cpu", cpu),
+        ("cuda", cuda),
+        ("pin_memory", pin_memory),
+        ("_slice", _slice),
+        ("_numel", _numel),
+        ("_uva", _uva),
+        ("_clear_data", _clear_data),
+        ("__hash__", __hash__),
+        ("_use_gpudnn", _use_gpudnn),
     ):
         setattr(core.eager.Tensor, method_name, method)
 
-    setattr(core.eager.Tensor, "_set_grad_ivar", _set_grad_ivar)
-    setattr(core.eager.Tensor, "value", value)
-    setattr(core.eager.Tensor, "cpu", cpu)
-    setattr(core.eager.Tensor, "cuda", cuda)
-    setattr(core.eager.Tensor, "pin_memory", pin_memory)
-    setattr(core.eager.Tensor, "_slice", _slice)
-    setattr(core.eager.Tensor, "_numel", _numel)
-    setattr(core.eager.Tensor, "_uva", _uva)
-    setattr(core.eager.Tensor, "_clear_data", _clear_data)
-    setattr(core.eager.Tensor, "__hash__", __hash__)
-    setattr(core.eager.Tensor, "_use_gpudnn", _use_gpudnn)
-
     global _already_patch_repr
     if not _already_patch_repr:
         # NOTE(zhiqiu): pybind11 will set a default __str__ method of enum class.
diff --git a/python/paddle/fluid/framework.py b/python/paddle/fluid/framework.py
index ad7140c14b6..a3d589fc616 100644
--- a/python/paddle/fluid/framework.py
+++ b/python/paddle/fluid/framework.py
@@ -52,7 +52,6 @@ __all__ = [
     'cpu_places',
     'xpu_places',
     'cuda_pinned_places',
-    '_non_static_mode',
     'in_dygraph_mode',
     'is_compiled_with_cinn',
     'is_compiled_with_cuda',
@@ -156,28 +155,6 @@ extra_op_attrs = {
     "unique": ["is_sorted"],
 }
 
-# Some explanation of our execution system 2022.03
-# For now we have 3 kinds of execution system, since we refactored dygraph mode to
-# build a fast execution system for dynamic mode. But we can't just remove all legacy
-# code once we present the new system for some historical reason. That's why we have
-# these flags.
-#
-# 1. _non_static_mode():
-# _non_static_mode means  we are now running in legacy dygraph mode or dygraph mode.
-# 2. dygraph_mode():
-# This flags inidicates we are now running in dygraph mode which called eager mode before.
-# 3. _in_legacy_dygraph():
-# This flags has been deprecated
-#
-# They have a relation ship as below:
-# Since _in_legacy_graph is deprecated, so dygraph_mode is _non_static_mode
-#
-# Why we have to make different of _in_legacy_dygraph and dygraph_mode?
-# In some performance issue, we find that python if statement cause server performance problem
-# and we need our new dygraph mode becomes as fast as it could be. That's why we make these flags
-# to make sure in most case, we find new dygraph mode first with only one if statement.
-
-
 # FIXME(dev): We haven't fully verified eager mode on XPU et.al but
 # only GPU/CPU. Remove this after we improve this feature.
 _is_first_import_ = True
@@ -213,10 +190,6 @@ def in_dygraph_mode():
     return global_var._dygraph_tracer_ is not None
 
 
-def _non_static_mode():
-    return global_var._dygraph_tracer_ is not None
-
-
 global_ipu_index = -1
 global_ipu_stage = -1
 ipu_index_attr_name = 'ipu_index'
@@ -459,7 +432,7 @@ def require_version(min_version, max_version=None):
 
 def _dygraph_not_support_(func):
     def __impl__(*args, **kwargs):
-        assert not _non_static_mode(), (
+        assert not in_dygraph_mode(), (
             "We don't support %s in dynamic graph mode" % func.__name__
         )
         return func(*args, **kwargs)
@@ -469,7 +442,7 @@ def _dygraph_not_support_(func):
 
 def _dygraph_only_(func):
     def __impl__(*args, **kwargs):
-        assert _non_static_mode(), (
+        assert in_dygraph_mode(), (
             "We only support '%s()' in dynamic graph mode, please call 'paddle.disable_static()' to enter dynamic graph mode."
             % func.__name__
         )
@@ -482,7 +455,7 @@ def _non_static_only_(func):
     def __impl__(*args, **kwargs):
         from .dygraph.base import in_declarative_mode
 
-        assert _non_static_mode() or in_declarative_mode(), (
+        assert in_dygraph_mode() or in_declarative_mode(), (
             "We only support '%s()' in dynamic graph mode, please call 'paddle.disable_static()' to enter dynamic graph mode."
             % func.__name__
         )
@@ -493,7 +466,7 @@ def _non_static_only_(func):
 
 def _static_only_(func):
     def __impl__(*args, **kwargs):
-        assert not _non_static_mode(), (
+        assert not in_dygraph_mode(), (
             "In PaddlePaddle 2.x, we turn on dynamic graph mode by default, and '%s()' is only supported in static graph mode. So if you want to use this api, please call 'paddle.enable_static()' before this api to enter static graph mode."
             % func.__name__
         )
@@ -971,7 +944,7 @@ def name_scope(prefix=None):
     """
     # TODO(panyx0718): Only [0-9a-z].
     # in dygraph we don't need namescope since it will cause mem leak
-    if _non_static_mode():
+    if in_dygraph_mode():
         yield
     else:
         assert prefix, "namescope prefix can not be empty."
@@ -2738,7 +2711,7 @@ class Operator:
         except ValueError:
             pass
 
-        if _non_static_mode():
+        if in_dygraph_mode():
             if type is None:
                 raise ValueError(
                     "`type` to initialized an Operator can not be None."
@@ -2924,7 +2897,7 @@ class Operator:
                         else:
                             out_arg_names.append(arg.name)
                         # TODO(minqiyang): could we remove variable's op in static graph mode?
-                        if not _non_static_mode():
+                        if not in_dygraph_mode():
                             if isinstance(arg, str):
                                 block.var(arg).op = self
                             else:
@@ -3799,7 +3772,7 @@ class Block:
         )
 
     def create_var(self, *args, **kwargs):
-        if _non_static_mode():
+        if in_dygraph_mode():
             var = _create_tensor(*args, **kwargs)
         else:
             var = Variable(block=self, *args, **kwargs)
@@ -3956,7 +3929,7 @@ class Block:
             Operator: the append Operator.
         """
         op_type = kwargs.get("type", None)
-        if _non_static_mode():
+        if in_dygraph_mode():
             attrs = kwargs.get("attrs", {})
             inplace_map = kwargs.get("inplace_map", None)
             warnings.warn(
@@ -4093,7 +4066,7 @@ class Block:
         return self.ops[start:end]
 
     def _prepend_op(self, *args, **kwargs):
-        if _non_static_mode():
+        if in_dygraph_mode():
             type = kwargs.get("type", None)
             attrs = kwargs.get("attrs", {})
             op = Operator(
@@ -7469,7 +7442,7 @@ def _cuda_graph_guard(cuda_graph_attr=None):
                                    cuda_graph_capture_mode;memory_pool_id;cuda_graph_id
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "cuda_graph_guard only works under static graph mode"
     assert (
         core.is_compiled_with_cuda()
diff --git a/python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py b/python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py
index 33c95f03b1f..a840b702140 100644
--- a/python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py
+++ b/python/paddle/fluid/incubate/checkpoint/auto_checkpoint.py
@@ -24,7 +24,7 @@ from contextlib import contextmanager
 
 from paddle.fluid import unique_name, compiler
 from .checkpoint_saver import SerializableBase, CheckpointSaver, PaddleModel
-from paddle.fluid.framework import _non_static_mode, Program
+from paddle.fluid.framework import in_dygraph_mode, Program
 
 g_train_epoch_range = None
 g_checker = None
@@ -138,7 +138,7 @@ class AutoCheckpointChecker:
         return self._save_checkpoint_inter
 
     def valid(self):
-        if _non_static_mode():
+        if in_dygraph_mode():
             return False
 
         return (
diff --git a/python/paddle/fluid/layer_helper.py b/python/paddle/fluid/layer_helper.py
index 0342017822c..96722b231ac 100644
--- a/python/paddle/fluid/layer_helper.py
+++ b/python/paddle/fluid/layer_helper.py
@@ -17,7 +17,7 @@ import paddle
 from .framework import (
     Parameter,
     dtype_is_floating,
-    _non_static_mode,
+    in_dygraph_mode,
     OpProtoHolder,
     _global_flags,
 )
@@ -159,7 +159,7 @@ class LayerHelper(LayerHelperBase):
         if use_mkldnn:
             act['use_mkldnn'] = use_mkldnn
         act_type = act.pop('type')
-        if _non_static_mode():
+        if in_dygraph_mode():
             res = _append_activation_in_dygraph(
                 input_var, act_type, use_cudnn, use_mkldnn
             )
diff --git a/python/paddle/fluid/layer_helper_base.py b/python/paddle/fluid/layer_helper_base.py
index 4a9b71c0cdf..0579bc98563 100644
--- a/python/paddle/fluid/layer_helper_base.py
+++ b/python/paddle/fluid/layer_helper_base.py
@@ -20,7 +20,7 @@ from .framework import (
     Variable,
     default_main_program,
     default_startup_program,
-    _non_static_mode,
+    in_dygraph_mode,
     _current_expected_place,
 )
 from . import unique_name
@@ -409,7 +409,7 @@ class LayerHelperBase:
             param = self._create_weight_normalize(attr, shape, dtype)
             WeightNormParamAttr.params_with_weight_norm.append(param)
             return param
-        if _non_static_mode():
+        if in_dygraph_mode():
             # In dygraph mode, we want the returned parameter to be
             # initialized so that it can be used imperatively.
             # check parameter name
@@ -527,7 +527,7 @@ class LayerHelperBase:
             initializer: initializer to use
         """
         assert isinstance(var, Variable)
-        if _non_static_mode():
+        if in_dygraph_mode():
             initializer(var, self.main_program.global_block())
         else:
             self.startup_program.global_block().create_var(
diff --git a/python/paddle/fluid/lazy_init.py b/python/paddle/fluid/lazy_init.py
index 1851056f2c2..3a577eb9084 100644
--- a/python/paddle/fluid/lazy_init.py
+++ b/python/paddle/fluid/lazy_init.py
@@ -37,7 +37,7 @@ class LazyInitHelper:
         if self._state:
             return
         assert (
-            framework._non_static_mode()
+            framework.in_dygraph_mode()
         ), "LazyInit.enable() is only available in dygraph mode."
         self._state = True
 
diff --git a/python/paddle/fluid/reader.py b/python/paddle/fluid/reader.py
index 6dfd487ef49..c08fdedba3b 100644
--- a/python/paddle/fluid/reader.py
+++ b/python/paddle/fluid/reader.py
@@ -26,7 +26,7 @@ from .framework import (
     program_guard,
     default_main_program,
     default_startup_program,
-    _non_static_mode,
+    in_dygraph_mode,
     cpu_places,
     _current_expected_place,
 )
@@ -417,7 +417,7 @@ class DataLoader:
                             epoch_id, batch_id, np.mean(loss.numpy())))
 
         """
-        if _non_static_mode():
+        if in_dygraph_mode():
             return DygraphGeneratorLoader(
                 feed_list,
                 capacity,
@@ -1605,7 +1605,7 @@ class DatasetLoader(DataLoaderBase):
             dataset, paddle.distributed.fleet.dataset.DatasetBase
         ), "dataset must be type of DatasetBase"
         assert (
-            not _non_static_mode()
+            not in_dygraph_mode()
         ), "DatasetLoader is not supported in dygraph mode yet"
         if isinstance(places, (list, tuple)):
             places = _get_paddle_place_list(places)
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_pp_recompute.py b/python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_pp_recompute.py
index 36fc010fd8e..793f2effd09 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_pp_recompute.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/hybrid_parallel_pp_recompute.py
@@ -87,7 +87,7 @@ class TransformerNet(Layer):
 
 class EmbeddingPipe(EmbeddingNet):
     def forward(self, tensors):
-        if framework.in_dygraph_mode():
+        if framework.in_dynamic_mode():
             stable, x = tensors
             return stable, super().forward(x)
         else:
@@ -96,7 +96,7 @@ class EmbeddingPipe(EmbeddingNet):
 
 class TransformerNetPipe(TransformerNet):
     def forward(self, tensors):
-        if framework.in_dygraph_mode():
+        if framework.in_dynamic_mode():
             stable, x = tensors
             output = super().forward(x)
             return stable, output
@@ -109,7 +109,7 @@ class CriterionPipe(Layer):
         super().__init__()
 
     def forward(self, out, label):
-        if framework.in_dygraph_mode():
+        if framework.in_dynamic_mode():
             out = out[-1]
         loss = out.mean()
         return loss
@@ -179,7 +179,7 @@ class TestDistPPTraning(unittest.TestCase):
             x_data = np.random.randint(0, vocab_size, size=[batch_size, length])
             x = paddle.to_tensor(x_data)
             x.stop_gradient = True
-            input_ = (x, x) if framework.in_dygraph_mode() else x
+            input_ = (x, x) if framework.in_dynamic_mode() else x
             loss = model.train_batch([input_, x], optimizer, scheduler)
             # TODO(shenliang03) add utest for loss
             print("loss: ", loss)
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_se_resnext.py b/python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_se_resnext.py
index 0aae9d7cc63..2e6026e99bc 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_se_resnext.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/parallel_dygraph_se_resnext.py
@@ -55,7 +55,7 @@ def optimizer_setting(params, parameter_list=None):
     bd = [step * e for e in ls["epochs"]]
     lr = params["lr"]
     num_epochs = params["num_epochs"]
-    if fluid._non_static_mode():
+    if fluid.in_dygraph_mode():
         optimizer = fluid.optimizer.Momentum(
             learning_rate=fluid.layers.cosine_decay(
                 learning_rate=lr, step_each_epoch=step, epochs=num_epochs
diff --git a/python/paddle/fluid/tests/unittests/test_context_manager.py b/python/paddle/fluid/tests/unittests/test_context_manager.py
index b11b37050e1..43ddca5fb83 100644
--- a/python/paddle/fluid/tests/unittests/test_context_manager.py
+++ b/python/paddle/fluid/tests/unittests/test_context_manager.py
@@ -29,8 +29,8 @@ class TestContextManagerRaiseException(unittest.TestCase):
 
     def test_func2(self):
         # After test_func1 executed, if fluid.dygraph.guard() in test_func1 safely exited,
-        # fluid._non_static_mode() should be false.
-        self.assertEqual(fluid._non_static_mode(), False)
+        # fluid.in_dygraph_mode() should be false.
+        self.assertEqual(fluid.in_dygraph_mode(), False)
 
 
 if __name__ == '__main__':
diff --git a/python/paddle/fluid/tests/unittests/test_dropout_nd_op.py b/python/paddle/fluid/tests/unittests/test_dropout_nd_op.py
index e2d5ae8685c..070e0f102a8 100644
--- a/python/paddle/fluid/tests/unittests/test_dropout_nd_op.py
+++ b/python/paddle/fluid/tests/unittests/test_dropout_nd_op.py
@@ -21,7 +21,7 @@ import paddle
 from paddle import _legacy_C_ops, fluid
 from paddle.fluid import core
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode
+from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
 from paddle.static import default_main_program
 
@@ -34,7 +34,7 @@ def dropout_nd(
     mode = (
         'downgrade_in_infer' if mode == 'downscale_in_infer' else mode
     )  # semantic transfer
-    if _non_static_mode():
+    if in_dygraph_mode():
         if default_main_program().random_seed != 0:
             seed = default_main_program().random_seed
 
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_decorator.py b/python/paddle/fluid/tests/unittests/test_imperative_decorator.py
index e7c78189eea..d02819ff293 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_decorator.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_decorator.py
@@ -27,7 +27,7 @@ class TestTracerMode(unittest.TestCase):
         self.init_mode = True
 
     def get_tracer_mode(self):
-        assert fluid._non_static_mode(), "Dygraph mode must be enabled"
+        assert framework.in_dygraph_mode(), "Dygraph mode must be enabled"
 
     @fluid.dygraph.no_grad
     def no_grad_func(self, a):
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_double_grad.py b/python/paddle/fluid/tests/unittests/test_imperative_double_grad.py
index 30e8119ce16..069a99b5587 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_double_grad.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_double_grad.py
@@ -26,7 +26,7 @@ from paddle.vision.models import resnet50, resnet101
 
 def _dygraph_guard_(func):
     def __impl__(*args, **kwargs):
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             return func(*args, **kwargs)
         else:
             with fluid.dygraph.guard():
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_ocr_attention_model.py b/python/paddle/fluid/tests/unittests/test_imperative_ocr_attention_model.py
index 7adda0bd9f3..7afbf1ba5f9 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_ocr_attention_model.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_ocr_attention_model.py
@@ -206,7 +206,7 @@ class EncoderNet(paddle.nn.Layer):
             initializer=paddle.nn.initializer.Normal(0.0, 0.02),
             learning_rate=2.0,
         )
-        if fluid.framework._non_static_mode():
+        if fluid.framework.in_dygraph_mode():
             h_0 = np.zeros(
                 (Config.batch_size, rnn_hidden_size), dtype="float32"
             )
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_resnet.py b/python/paddle/fluid/tests/unittests/test_imperative_resnet.py
index af046305ff7..d130f97686a 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_resnet.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_resnet.py
@@ -58,7 +58,7 @@ def optimizer_setting(params, parameter_list=None):
         base_lr = params["lr"]
         lr = []
         lr = [base_lr * (0.1**i) for i in range(len(bd) + 1)]
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             optimizer = fluid.optimizer.SGD(
                 learning_rate=0.01, parameter_list=parameter_list
             )
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_resnet_sorted_gradient.py b/python/paddle/fluid/tests/unittests/test_imperative_resnet_sorted_gradient.py
index bc1dbd49af3..c537da047b8 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_resnet_sorted_gradient.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_resnet_sorted_gradient.py
@@ -54,7 +54,7 @@ def optimizer_setting(params, parameter_list=None):
         base_lr = params["lr"]
         lr = []
         lr = [base_lr * (0.1**i) for i in range(len(bd) + 1)]
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             optimizer = fluid.optimizer.SGD(
                 learning_rate=0.01, parameter_list=parameter_list
             )
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_se_resnext.py b/python/paddle/fluid/tests/unittests/test_imperative_se_resnext.py
index 1f98ac13155..e727c16ccf3 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_se_resnext.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_se_resnext.py
@@ -54,7 +54,7 @@ def optimizer_setting(params, parameter_list=None):
         # bd = [step * e for e in ls["epochs"]]
         # base_lr = params["lr"]
         # lr = [base_lr * (0.1**i) for i in range(len(bd) + 1)]
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             optimizer = fluid.optimizer.SGD(
                 learning_rate=0.01, parameter_list=parameter_list
             )
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_star_gan_with_gradient_penalty.py b/python/paddle/fluid/tests/unittests/test_imperative_star_gan_with_gradient_penalty.py
index 693dd5c7b88..0eda5bcc349 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_star_gan_with_gradient_penalty.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_star_gan_with_gradient_penalty.py
@@ -114,7 +114,7 @@ class InstanceNorm(paddle.nn.Layer):
         self.bias = self.create_parameter(shape=[num_channels], is_bias=True)
 
     def forward(self, input):
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             out, _, _ = _legacy_C_ops.instance_norm(
                 input, self.scale, self.bias, 'epsilon', self.epsilon
             )
@@ -387,7 +387,7 @@ def loss_cls(cls, label, cfg):
 
 
 def calc_gradients(outputs, inputs, no_grad_set):
-    if fluid._non_static_mode():
+    if fluid.in_dygraph_mode():
         return fluid.dygraph.grad(
             outputs=outputs,
             inputs=inputs,
@@ -481,7 +481,7 @@ def build_optimizer(layer, cfg, loss=None):
     learning_rate = 1e-3
     beta1 = 0.5
     beta2 = 0.999
-    if fluid._non_static_mode():
+    if fluid.in_dygraph_mode():
         return fluid.optimizer.Adam(
             learning_rate=learning_rate,
             beta1=beta1,
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_tensor_clear_gradient.py b/python/paddle/fluid/tests/unittests/test_imperative_tensor_clear_gradient.py
index e068b595da4..11864a59027 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_tensor_clear_gradient.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_tensor_clear_gradient.py
@@ -24,7 +24,7 @@ from paddle.fluid.wrapped_decorator import wrap_decorator
 
 def _dygraph_guard_(func):
     def __impl__(*args, **kwargs):
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             return func(*args, **kwargs)
         else:
             with fluid.dygraph.guard():
diff --git a/python/paddle/fluid/tests/unittests/test_imperative_triple_grad.py b/python/paddle/fluid/tests/unittests/test_imperative_triple_grad.py
index 70df83b98e1..72e2689da77 100644
--- a/python/paddle/fluid/tests/unittests/test_imperative_triple_grad.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_triple_grad.py
@@ -24,7 +24,7 @@ from paddle.fluid.wrapped_decorator import wrap_decorator
 
 def _dygraph_guard_(func):
     def __impl__(*args, **kwargs):
-        if fluid._non_static_mode():
+        if fluid.in_dygraph_mode():
             return func(*args, **kwargs)
         else:
             with fluid.dygraph.guard():
diff --git a/python/paddle/fluid/tests/unittests/test_logical_op.py b/python/paddle/fluid/tests/unittests/test_logical_op.py
index 14da477f18d..80a42d41451 100755
--- a/python/paddle/fluid/tests/unittests/test_logical_op.py
+++ b/python/paddle/fluid/tests/unittests/test_logical_op.py
@@ -18,7 +18,7 @@ import numpy as np
 from eager_op_test import convert_float_to_uint16
 
 import paddle
-from paddle.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 from paddle.static import Executor, Program, program_guard
 
 SUPPORTED_DTYPES = [
@@ -182,11 +182,11 @@ def test_type_error(unit_test, use_gpu, type_str_map):
         if binary_op:
             if type_str_map['x'] != type_str_map['y']:
                 unit_test.assertRaises(error_type, op, x=x, y=y)
-            if not _non_static_mode():
+            if not in_dynamic_mode():
                 error_type = TypeError
                 unit_test.assertRaises(error_type, op, x=x, y=y, out=1)
         else:
-            if not _non_static_mode():
+            if not in_dynamic_mode():
                 error_type = TypeError
                 unit_test.assertRaises(error_type, op, x=x, out=1)
 
diff --git a/python/paddle/fluid/tests/unittests/test_multiclass_nms_op.py b/python/paddle/fluid/tests/unittests/test_multiclass_nms_op.py
index 423e7039dd9..29ce4be023f 100644
--- a/python/paddle/fluid/tests/unittests/test_multiclass_nms_op.py
+++ b/python/paddle/fluid/tests/unittests/test_multiclass_nms_op.py
@@ -19,8 +19,8 @@ import numpy as np
 from eager_op_test import OpTest
 
 import paddle
-from paddle import _C_ops, _legacy_C_ops
-from paddle.fluid import _non_static_mode, core, in_dygraph_mode
+from paddle import _C_ops
+from paddle.fluid import core
 from paddle.fluid.layer_helper import LayerHelper
 
 
@@ -42,7 +42,7 @@ def multiclass_nms3(
 
     helper = LayerHelper('multiclass_nms3', **locals())
 
-    if in_dygraph_mode():
+    if paddle.in_dynamic_mode():
         attrs = (
             score_threshold,
             nms_top_k,
@@ -58,30 +58,6 @@ def multiclass_nms3(
         if not return_index:
             index = None
         return output, index, nms_rois_num
-    elif _non_static_mode():
-        attrs = (
-            'background_label',
-            background_label,
-            'score_threshold',
-            score_threshold,
-            'nms_top_k',
-            nms_top_k,
-            'nms_threshold',
-            nms_threshold,
-            'keep_top_k',
-            keep_top_k,
-            'nms_eta',
-            nms_eta,
-            'normalized',
-            normalized,
-        )
-        output, index, nms_rois_num = _legacy_C_ops.multiclass_nms3(
-            bboxes, scores, rois_num, *attrs
-        )
-        if not return_index:
-            index = None
-        return output, index, nms_rois_num
-
     else:
         output = helper.create_variable_for_type_inference(dtype=bboxes.dtype)
         index = helper.create_variable_for_type_inference(dtype='int32')
diff --git a/python/paddle/fluid/tests/unittests/test_squared_l2_norm_op.py b/python/paddle/fluid/tests/unittests/test_squared_l2_norm_op.py
index 6ad9f721d66..17586d94f3b 100755
--- a/python/paddle/fluid/tests/unittests/test_squared_l2_norm_op.py
+++ b/python/paddle/fluid/tests/unittests/test_squared_l2_norm_op.py
@@ -20,11 +20,11 @@ from numpy import linalg as LA
 
 import paddle
 from paddle import _C_ops, _legacy_C_ops
-from paddle.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 
 def test_squared_l2_norm(x):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.squared_l2_norm(x)
     else:
         return _legacy_C_ops.squared_l2_norm(x)
diff --git a/python/paddle/fluid/unique_name.py b/python/paddle/fluid/unique_name.py
index c881196ef56..863883a220c 100644
--- a/python/paddle/fluid/unique_name.py
+++ b/python/paddle/fluid/unique_name.py
@@ -120,9 +120,9 @@ def generate(key):
 # NOTE(zhiqiu): use c++ unique_name_generator in dygraph mode,
 # in order to keep name consistency.
 def generate_with_ignorable_key(key):
-    from .framework import _non_static_mode, _dygraph_tracer
+    from .framework import in_dygraph_mode, _dygraph_tracer
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         return _dygraph_tracer()._generate_unique_name()
 
     return generator(key)
diff --git a/python/paddle/fluid/variable_index.py b/python/paddle/fluid/variable_index.py
index fbd55ba83aa..dae0deb135c 100644
--- a/python/paddle/fluid/variable_index.py
+++ b/python/paddle/fluid/variable_index.py
@@ -436,8 +436,7 @@ def _getitem_impl_(var, item):
                 start = 0 if step > 0 else MAX_INTEGER
             if end is None:
                 if (
-                    paddle.fluid.framework._non_static_mode()
-                    or not is_tensor_array
+                    paddle.in_dynamic_mode() or not is_tensor_array
                 ) and var.shape[dim] != -1:
                     end = var.shape[dim] if step > 0 else -1
                 else:
@@ -550,7 +549,7 @@ def _getitem_impl_(var, item):
     out = var
     if len(axes) > 0:
         op_type = "strided_slice" if use_strided_slice else "slice"
-        if paddle.fluid.framework.in_dygraph_mode() and op_type == "slice":
+        if paddle.in_dynamic_mode() and op_type == "slice":
             if "StartsTensorList" in inputs.keys():
                 st = inputs['StartsTensorList']
             else:
@@ -620,11 +619,11 @@ def _setitem_for_tensor_array(var, item, value):
     If item is case (1), we perform paddle.tensor.array_write,
     in other cases, we raise a NotImplementedError.
     """
-    from ..framework import LayerHelper, core, _non_static_mode
+    from ..framework import LayerHelper, core
     from .framework import Variable
 
     assert (
-        not _non_static_mode()
+        not paddle.in_dynamic_mode()
     ), "setitem for tensor_array must be called in static graph mode."
     if isinstance(item, (Variable, int)):
         from paddle.jit.dy2static.variable_trans_func import (
@@ -808,7 +807,7 @@ def _setitem_impl_(var, item, value):
             )
         )
 
-    if paddle.fluid.framework._non_static_mode():
+    if paddle.in_dynamic_mode():
         var._bump_inplace_version()
 
     cur_block = default_main_program().current_block()
diff --git a/python/paddle/framework/__init__.py b/python/paddle/framework/__init__.py
index 0ae39627d9a..16f12fd41ec 100755
--- a/python/paddle/framework/__init__.py
+++ b/python/paddle/framework/__init__.py
@@ -55,10 +55,7 @@ from ..fluid.framework import set_flags  # noqa: F401
 from ..fluid.framework import Parameter
 from ..fluid.dygraph.base import enable_dygraph as disable_static  # noqa: F401
 from ..fluid.dygraph.base import disable_dygraph as enable_static  # noqa: F401
-from ..fluid.framework import _non_static_mode as in_dynamic_mode  # noqa: F401
-from ..fluid.framework import (  # noqa: F401
-    _non_static_mode,  # temporary used for hackson
-)
+from ..fluid.framework import in_dygraph_mode as in_dynamic_mode  # noqa: F401
 from ..fluid.framework import (
     _current_expected_place,
     _get_paddle_place,
@@ -74,7 +71,6 @@ from ..fluid.framework import _dygraph_tracer  # noqa: F401
 from ..fluid.framework import generate_control_dev_var_name  # noqa: F401
 
 from ..fluid.layer_helper import LayerHelper  # noqa: F401
-from ..fluid.framework import in_dygraph_mode  # noqa: F401
 from ..fluid.framework import _global_flags  # noqa: F401
 from ..fluid.framework import _apply_pass  # noqa: F401
 from ..fluid.framework import switch_main_program
diff --git a/python/paddle/framework/io.py b/python/paddle/framework/io.py
index e5d011453f2..055b4aa5fa4 100644
--- a/python/paddle/framework/io.py
+++ b/python/paddle/framework/io.py
@@ -34,7 +34,7 @@ from paddle.fluid.framework import (
     _create_tensor,
     _current_expected_place,
     _dygraph_tracer,
-    _non_static_mode,
+    in_dygraph_mode,
 )
 
 from .io_utils import (
@@ -438,7 +438,7 @@ def _to_LodTensor(ndarray):
 def _tuple_to_tensor(obj, return_numpy):
     if return_numpy:
         return obj[1]
-    if _non_static_mode():
+    if in_dygraph_mode():
         t = paddle.to_tensor(obj[1])
         # This function does modify the name of return value.
         # Loading the same variable multiple times may cause the same name.
@@ -451,7 +451,7 @@ def _tuple_to_tensor(obj, return_numpy):
 def _ndarray_to_tensor(obj, return_numpy):
     if return_numpy:
         return obj
-    if _non_static_mode():
+    if in_dygraph_mode():
         return paddle.to_tensor(obj)
     else:
         return _to_LodTensor(obj)
@@ -508,7 +508,7 @@ def _parse_load_result(obj, return_numpy):
         return obj
 
     if _contain_x(obj, is_layer):
-        if not _non_static_mode():
+        if not in_dygraph_mode():
             raise ValueError(
                 "Layer can only be loaded in dynamic graph mode, but now in static graph mode."
             )
@@ -819,7 +819,7 @@ def save(obj, path, protocol=4, **configs):
                 f.write(obj.desc.serialize_to_string())
 
         elif _is_state_dict(obj):
-            if _non_static_mode():
+            if in_dygraph_mode():
                 _legacy_save(obj, path, protocol)
             else:
                 _legacy_static_save(obj, path, protocol)
@@ -1110,7 +1110,7 @@ def load(path, **configs):
                     if config.return_numpy:
                         return np.array(tensor)
                     else:
-                        if _non_static_mode():
+                        if in_dygraph_mode():
                             return _lod_tensor2varbase(tensor)
                         return tensor
                 except:
diff --git a/python/paddle/geometric/math.py b/python/paddle/geometric/math.py
index 9e07e1c77a7..9db0303a2f6 100644
--- a/python/paddle/geometric/math.py
+++ b/python/paddle/geometric/math.py
@@ -14,8 +14,8 @@
 
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 __all__ = []
 
@@ -50,7 +50,7 @@ def segment_sum(data, segment_ids, name=None):
             #Outputs: [[4., 4., 4.], [4., 5., 6.]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "SUM")
     else:
         check_variable_and_dtype(
@@ -107,7 +107,7 @@ def segment_mean(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "MEAN")
     else:
 
@@ -164,7 +164,7 @@ def segment_min(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "MIN")
     else:
         check_variable_and_dtype(
@@ -220,7 +220,7 @@ def segment_max(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "MAX")
     else:
         check_variable_and_dtype(
diff --git a/python/paddle/geometric/message_passing/send_recv.py b/python/paddle/geometric/message_passing/send_recv.py
index 5f3ca7afe3f..e37868a229d 100644
--- a/python/paddle/geometric/message_passing/send_recv.py
+++ b/python/paddle/geometric/message_passing/send_recv.py
@@ -20,8 +20,9 @@ from paddle.fluid.data_feeder import (
     check_type,
     check_variable_and_dtype,
 )
-from paddle.fluid.framework import Variable, in_dygraph_mode
+from paddle.fluid.framework import Variable
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 from .utils import (
     convert_out_size_to_list,
@@ -118,7 +119,7 @@ def send_u_recv(
 
     # TODO(daisiming): Should we add judgement for out_size: max(dst_index) + 1.
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out_size = convert_out_size_to_list(out_size)
         return _C_ops.send_u_recv(
             x, src_index, dst_index, reduce_op.upper(), out_size
@@ -295,7 +296,7 @@ def send_ue_recv(
 
     # TODO(daisiming): Should we add judgement for out_size: max(dst_index) + 1.
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out_size = convert_out_size_to_list(out_size)
         return _C_ops.send_ue_recv(
             x,
@@ -451,7 +452,7 @@ def send_uv(x, y, src_index, dst_index, message_op="add", name=None):
         message_op = 'mul'
         y = 1.0 / (y + 1e-12)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.send_uv(x, y, src_index, dst_index, message_op.upper())
     else:
 
diff --git a/python/paddle/geometric/reindex.py b/python/paddle/geometric/reindex.py
index 0456750d4a5..c65fc36a417 100644
--- a/python/paddle/geometric/reindex.py
+++ b/python/paddle/geometric/reindex.py
@@ -15,8 +15,9 @@
 import paddle
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import Variable, _non_static_mode
+from paddle.fluid.framework import Variable
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 __all__ = []
 
@@ -86,7 +87,7 @@ def reindex_graph(
         True if value_buffer is not None and index_buffer is not None else False
     )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         reindex_src, reindex_dst, out_nodes = _C_ops.reindex_graph(
             x,
             neighbors,
@@ -205,7 +206,7 @@ def reindex_heter_graph(
         True if value_buffer is not None and index_buffer is not None else False
     )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         neighbors = paddle.concat(neighbors, axis=0)
         count = paddle.concat(count, axis=0)
         reindex_src, reindex_dst, out_nodes = _C_ops.reindex_graph(
diff --git a/python/paddle/geometric/sampling/neighbors.py b/python/paddle/geometric/sampling/neighbors.py
index c8d907c078b..1bd7c001d48 100644
--- a/python/paddle/geometric/sampling/neighbors.py
+++ b/python/paddle/geometric/sampling/neighbors.py
@@ -14,8 +14,8 @@
 
 from paddle import _C_ops, _legacy_C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode, in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 __all__ = []
 
@@ -100,7 +100,7 @@ def sample_neighbors(
 
     use_perm_buffer = True if perm_buffer is not None else False
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         (
             out_neighbors,
             out_count,
@@ -251,7 +251,7 @@ def weighted_sample_neighbors(
                 "`eids` should not be None if `return_eids` is True."
             )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         (
             out_neighbors,
             out_count,
diff --git a/python/paddle/hapi/model.py b/python/paddle/hapi/model.py
index 4169902401f..a20ee3169ad 100644
--- a/python/paddle/hapi/model.py
+++ b/python/paddle/hapi/model.py
@@ -33,7 +33,8 @@ from paddle.fluid.dygraph.base import to_variable
 from paddle.fluid.executor import global_scope
 from paddle.fluid.framework import Variable
 from paddle.fluid.framework import _current_expected_place as _get_device
-from paddle.fluid.framework import _get_paddle_place, _non_static_mode
+from paddle.fluid.framework import _get_paddle_place
+from paddle.framework import in_dynamic_mode
 from paddle.framework.io_utils import is_belong_to_optimizer
 from paddle.io import DataLoader, Dataset, DistributedBatchSampler
 from paddle.jit.translated_layer import INFER_MODEL_SUFFIX, INFER_PARAMS_SUFFIX
@@ -256,7 +257,7 @@ def prepare_distributed_context(place=None):
             exe = fluid.Executor(place)
             exe.run(communicator_prog)
 
-        if fluid._non_static_mode():
+        if in_dynamic_mode():
             fluid.disable_dygraph()
             _init_context()
             fluid.enable_dygraph(place)
@@ -1170,7 +1171,7 @@ class Model:
         self._test_dataloader = None
         self.stop_training = False
 
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             if not isinstance(inputs, (list, tuple, dict, Input)):
                 raise TypeError(
                     "'inputs' must be list or tuple or dict, and couldn't be None."
@@ -1182,7 +1183,7 @@ class Model:
         self._labels = self._verify_spec(labels)
 
         # init backend
-        if fluid._non_static_mode():
+        if in_dynamic_mode():
             self._adapter = DynamicGraphAdapter(self)
         else:
             self._adapter = StaticGraphAdapter(self)
@@ -1238,7 +1239,7 @@ class Model:
 
         """
         loss = self._adapter.train_batch(inputs, labels, update)
-        if fluid._non_static_mode() and self._input_info is None:
+        if in_dynamic_mode() and self._input_info is None:
             self._update_inputs()
         return loss
 
@@ -1292,7 +1293,7 @@ class Model:
 
         """
         loss = self._adapter.eval_batch(inputs, labels)
-        if fluid._non_static_mode() and self._input_info is None:
+        if in_dynamic_mode() and self._input_info is None:
             self._update_inputs()
         return loss
 
@@ -1341,7 +1342,7 @@ class Model:
 
         """
         loss = self._adapter.predict_batch(inputs)
-        if fluid._non_static_mode() and self._input_info is None:
+        if in_dynamic_mode() and self._input_info is None:
             self._update_inputs()
         return loss
 
@@ -1527,7 +1528,7 @@ class Model:
         )
 
         # TODO: support save/load scaler state in static graph
-        if _non_static_mode():
+        if in_dynamic_mode():
             scaler_state = None
             if hasattr(self, '_scaler') and self._scaler is not None:
                 if os.path.exists(path + '.pdscaler'):
@@ -1644,7 +1645,7 @@ class Model:
                 )
 
             if 'use_fp16_guard' in amp_config_key_set:
-                if _non_static_mode():
+                if in_dynamic_mode():
                     raise ValueError(
                         "'use_fp16_guard' is supported in static graph mode only."
                     )
@@ -1702,7 +1703,7 @@ class Model:
                 paddle.distributed.ParallelEnv().nranks > 1
                 and not _parallel_context_initialized
             ):
-                if fluid._non_static_mode():
+                if in_dynamic_mode():
                     main_prog_seed = fluid.default_main_program().random_seed
                     startup_prog_seed = (
                         fluid.default_startup_program().random_seed
@@ -2228,7 +2229,7 @@ class Model:
             None
         """
 
-        if fluid._non_static_mode():
+        if in_dynamic_mode():
             with fluid.framework._dygraph_guard(None):
                 layer = self.network
                 if self._input_info is None:  # No provided or inferred
@@ -2428,7 +2429,7 @@ class Model:
                 if (
                     shapes is not None
                     and dtypes is not None
-                    and fluid._non_static_mode()
+                    and in_dynamic_mode()
                 ):
                     out_specs = [
                         Input(name=n, dtype=dtypes[i], shape=shapes[i])
diff --git a/python/paddle/incubate/autograd/functional.py b/python/paddle/incubate/autograd/functional.py
index 218c125a92e..8f6f012ac39 100644
--- a/python/paddle/incubate/autograd/functional.py
+++ b/python/paddle/incubate/autograd/functional.py
@@ -69,7 +69,7 @@ def vjp(func, xs, v=None):
 
     # ``_seprate`` breaks the dependencies between ``xs`` and other
     # variables. See more ``_seprate`` .
-    if paddle.fluid._non_static_mode() or not utils.prim_enabled():
+    if framework.in_dygraph_mode() or not utils.prim_enabled():
         xs, v = _separate(xs), _separate(v)
     ys = func(*xs) if isinstance(xs, typing.Sequence) else func(xs)
     _check_v_shape(v, ys)
@@ -130,12 +130,12 @@ def jvp(func, xs, v=None):
     _check_inputs(func, xs, v)
     # ``_seprate`` breaks the dependencies between ``xs`` and other
     # variables. See more ``_seprate`` .
-    if paddle.fluid._non_static_mode() or not utils.prim_enabled():
+    if framework.in_dygraph_mode() or not utils.prim_enabled():
         xs, v = _separate(xs), _separate(v)
     ys = func(*xs) if isinstance(xs, typing.Sequence) else func(xs)
     _check_v_shape(v, xs)
 
-    if not paddle.fluid._non_static_mode() and utils.prim_enabled():
+    if not framework.in_dygraph_mode() and utils.prim_enabled():
         return ys, primapi.forward_grad(ys, xs, v)
     else:
         return ys, _double_backward_trick(ys, xs, v)
@@ -352,7 +352,7 @@ class _Jacobian:
     def __init__(self, func, xs):
         # Skip separating in prim mode temporarily, as detach and clone are not
         # primitive operators.
-        if not paddle.fluid._non_static_mode() and utils.prim_enabled():
+        if not framework.in_dygraph_mode() and utils.prim_enabled():
             self._xs = xs
         else:
             self._xs = _separate(xs)
@@ -580,7 +580,7 @@ def _grad(ys, xs, v=None):
             Tensor is the sum of gradients of outputs with respect to the i-th
             inputs.
     """
-    if paddle.fluid._non_static_mode():
+    if framework.in_dygraph_mode():
         # paddle.grad returns a list though the inputs is a signle Tensor. The
         # follow code snippet fixes the problem by return the first element of
         # xs_grad when the xs is a signle Tensor.
diff --git a/python/paddle/incubate/distributed/models/moe/moe_layer.py b/python/paddle/incubate/distributed/models/moe/moe_layer.py
index 9fd8081306c..1923db43b78 100644
--- a/python/paddle/incubate/distributed/models/moe/moe_layer.py
+++ b/python/paddle/incubate/distributed/models/moe/moe_layer.py
@@ -26,7 +26,7 @@ from paddle import nn
 from paddle.autograd import PyLayer
 from paddle.distributed.utils.moe_utils import global_gather, global_scatter
 from paddle.distributed.utils.nccl_utils import check_nccl_version_for_p2p
-from paddle.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 from paddle.incubate.distributed.fleet import recompute_hybrid
 
 from .gate import BaseGate, GShardGate, NaiveGate, SwitchGate
@@ -63,7 +63,7 @@ def _all_gather(tensor, group=None, use_calc_stream=True):
     if group is not None and not group.is_member():
         return
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         group = (
             paddle.distributed.collective._get_default_group()
             if group is None
diff --git a/python/paddle/incubate/distributed/models/moe/utils.py b/python/paddle/incubate/distributed/models/moe/utils.py
index e7c0fa9af54..f385baa1c2f 100644
--- a/python/paddle/incubate/distributed/models/moe/utils.py
+++ b/python/paddle/incubate/distributed/models/moe/utils.py
@@ -26,14 +26,14 @@ from paddle.distributed.models.moe.utils import (
     _number_count,
     _prune_gate_by_capacity,
 )
-from paddle.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 
 def _alltoall(in_tensor_list, group=None, use_calc_stream=True):
     if group is not None and not group.is_member():
         return
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         group = (
             paddle.distributed.collective._get_default_group()
             if group is None
diff --git a/python/paddle/incubate/layers/nn.py b/python/paddle/incubate/layers/nn.py
index 113b35b6e09..5ceea828109 100644
--- a/python/paddle/incubate/layers/nn.py
+++ b/python/paddle/incubate/layers/nn.py
@@ -1174,7 +1174,7 @@ def bilateral_slice(x, guide, grid, has_offset, name=None):
             output = paddle.incubate.layers.bilateral_slice(x, guide, grid, has_offset=True)
 
     """
-    if paddle.fluid._non_static_mode():
+    if paddle.in_dynamic_mode():
         attrs = ('has_offset', has_offset)
         return _legacy_C_ops.bilateral_slice(x, grid, guide, *attrs)
 
@@ -1252,7 +1252,7 @@ def correlation(
 
     """
 
-    if paddle.fluid._non_static_mode():
+    if paddle.in_dynamic_mode():
         attrs = (
             "pad_size",
             pad_size,
@@ -1501,7 +1501,7 @@ def fused_bn_add_act(
 def pow2_decay_with_linear_warmup(
     warmup_steps, total_steps, base_lr, end_lr, dtype='float32', name=None
 ):
-    if paddle.fluid._non_static_mode():
+    if paddle.in_dynamic_mode():
         raise NotImplementedError(
             "pow2_decay_with_linear_warmup does not support dygraph mode yet."
         )
diff --git a/python/paddle/incubate/nn/functional/fused_dropout_add.py b/python/paddle/incubate/nn/functional/fused_dropout_add.py
index 5f1b94cd896..79f5adfcc33 100644
--- a/python/paddle/incubate/nn/functional/fused_dropout_add.py
+++ b/python/paddle/incubate/nn/functional/fused_dropout_add.py
@@ -16,8 +16,7 @@
 from paddle import _C_ops
 from paddle.common_ops_import import default_main_program
 from paddle.fluid import core
-from paddle.fluid.framework import in_dygraph_mode
-from paddle.framework import LayerHelper
+from paddle.framework import LayerHelper, in_dynamic_mode
 
 
 def fused_dropout_add(
@@ -73,7 +72,7 @@ def fused_dropout_add(
             "mode argument should be 'downscale_in_infer' or 'upscale_in_train'"
         )
     seed = None
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if default_main_program().random_seed != 0:
             seed = default_main_program().random_seed
         out, seed_offset = _C_ops.fused_dropout_add(
diff --git a/python/paddle/incubate/nn/functional/fused_gate_attention.py b/python/paddle/incubate/nn/functional/fused_gate_attention.py
index 78f4abac823..5bc2211c33c 100644
--- a/python/paddle/incubate/nn/functional/fused_gate_attention.py
+++ b/python/paddle/incubate/nn/functional/fused_gate_attention.py
@@ -13,7 +13,7 @@
 # limitations under the License.
 
 from paddle import _legacy_C_ops
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 
 def fused_gate_attention(
@@ -142,7 +142,7 @@ def fused_gate_attention(
             # [2, 4, 2, 4]
 
     """
-    if _non_static_mode():
+    if in_dynamic_mode():
         _, _, _, _, _, _, _, _, out = _legacy_C_ops.fused_gate_attention(
             query,
             key,
diff --git a/python/paddle/incubate/nn/functional/fused_matmul_bias.py b/python/paddle/incubate/nn/functional/fused_matmul_bias.py
index c80a437b390..72fbdea4535 100644
--- a/python/paddle/incubate/nn/functional/fused_matmul_bias.py
+++ b/python/paddle/incubate/nn/functional/fused_matmul_bias.py
@@ -13,8 +13,8 @@
 # limitations under the License.
 
 from paddle import _legacy_C_ops
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.tensor.linalg import matmul
 
 
@@ -53,7 +53,7 @@ def fused_matmul_bias(
     """
     if bias is None:
         return matmul(x, y, transpose_x, transpose_y, name)
-    if _non_static_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.fused_gemm_epilogue(
             x, y, bias, 'trans_x', transpose_x, 'trans_y', transpose_y
         )
diff --git a/python/paddle/incubate/nn/functional/fused_transformer.py b/python/paddle/incubate/nn/functional/fused_transformer.py
index 59a15750c7d..469aea26cc6 100644
--- a/python/paddle/incubate/nn/functional/fused_transformer.py
+++ b/python/paddle/incubate/nn/functional/fused_transformer.py
@@ -15,8 +15,9 @@
 from paddle import _legacy_C_ops
 from paddle.fluid import core
 from paddle.fluid.data_feeder import check_dtype, check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode, default_main_program
+from paddle.fluid.framework import default_main_program
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 __all__ = []
 
@@ -132,7 +133,7 @@ def fused_feedforward(
         'downgrade_in_infer' if mode == 'downscale_in_infer' else mode
     )  # semantic transfer
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if default_main_program().random_seed != 0:
             seed = default_main_program().random_seed
         out, _, _, _, _, _, _, _, _, _, _ = _legacy_C_ops.fused_feedforward(
@@ -363,7 +364,7 @@ def fused_bias_dropout_residual_layer_norm(
             x.shape[len(x.shape) - 1] == ln_bias.shape[0]
         ), "The dim of ln_bias must equal to the last dim of x."
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if default_main_program().random_seed != 0:
             seed = default_main_program().random_seed
         (
@@ -620,7 +621,7 @@ def fused_multi_head_attention(
             f"The rank of the x should be 3, but received {x.ndim}."
         )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if default_main_program().random_seed != 0:
             seed = default_main_program().random_seed
         # pre_ln_mean, pre_ln_variance, pre_ln_out, qkv_out, qkv_bias_out, transpose_out, qk_out,
@@ -1046,7 +1047,7 @@ def fused_multi_transformer(
         'downgrade_in_infer' if mode == 'downscale_in_infer' else mode
     )  # semantic transfer
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         cache_kv_out, final_out = _legacy_C_ops.fused_multi_transformer(
             x,
             ln_scales,
diff --git a/python/paddle/incubate/nn/layer/fused_dropout_nd.py b/python/paddle/incubate/nn/layer/fused_dropout_nd.py
index b8aecf0c397..156880f7328 100644
--- a/python/paddle/incubate/nn/layer/fused_dropout_nd.py
+++ b/python/paddle/incubate/nn/layer/fused_dropout_nd.py
@@ -14,7 +14,7 @@
 
 import paddle
 from paddle import _legacy_C_ops
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 
 class FusedDropout(paddle.nn.Layer):
@@ -104,7 +104,7 @@ class FusedDropout(paddle.nn.Layer):
         if self.p == 0:
             return input
 
-        if self.axis is not None and _non_static_mode():
+        if self.axis is not None and in_dynamic_mode():
             seed = None
             if paddle.static.default_main_program().random_seed != 0:
                 seed = paddle.static.default_main_program().random_seed
diff --git a/python/paddle/incubate/nn/layer/fused_transformer.py b/python/paddle/incubate/nn/layer/fused_transformer.py
index 31ea1e8d663..c385d0284a1 100644
--- a/python/paddle/incubate/nn/layer/fused_transformer.py
+++ b/python/paddle/incubate/nn/layer/fused_transformer.py
@@ -17,7 +17,8 @@ import paddle
 from paddle.fluid import core
 from paddle.fluid.core import VarDesc
 from paddle.fluid.dygraph import no_grad
-from paddle.fluid.framework import _non_static_mode, convert_np_dtype_to_dtype_
+from paddle.fluid.framework import convert_np_dtype_to_dtype_
+from paddle.framework import in_dynamic_mode
 from paddle.incubate.nn import functional as incubate_f
 from paddle.nn import Layer
 from paddle.nn.initializer import Constant
@@ -34,7 +35,7 @@ def _set_var_distributed(var):
 
     var.is_distributed = True
 
-    if not _non_static_mode():
+    if not in_dynamic_mode():
         # NOTE: use current_block and find_var_recursive to support while_loop
         startup_block = paddle.static.default_startup_program().current_block()
         main_block = paddle.static.default_main_program().current_block()
diff --git a/python/paddle/incubate/nn/loss.py b/python/paddle/incubate/nn/loss.py
index 9d5d2618391..09d41e3f82d 100644
--- a/python/paddle/incubate/nn/loss.py
+++ b/python/paddle/incubate/nn/loss.py
@@ -15,7 +15,7 @@
 from paddle import _legacy_C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
 from paddle.fluid.layer_helper import LayerHelper
-from paddle.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 
 def identity_loss(x, reduction="none"):
@@ -59,7 +59,7 @@ def identity_loss(x, reduction="none"):
         if reduction is None:
             raise Exception("Unsupported reduction type.")
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.identity_loss(x, "reduction", reduction)
 
     check_variable_and_dtype(x, 'x', ['float32', 'float64'], "identity_loss")
diff --git a/python/paddle/incubate/nn/memory_efficient_attention.py b/python/paddle/incubate/nn/memory_efficient_attention.py
index 049f00a70cf..565809cf7df 100644
--- a/python/paddle/incubate/nn/memory_efficient_attention.py
+++ b/python/paddle/incubate/nn/memory_efficient_attention.py
@@ -21,8 +21,8 @@
 
 import paddle
 from paddle import _C_ops
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 from .attn_bias import (
     BlockDiagonalCausalMask,
@@ -99,7 +99,7 @@ def memory_efficient_attention(
     bias = _get_tensor_bias(attn_bias)
     is_test = not training
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         output, logsumexp, seed_and_offset = _C_ops.memory_efficient_attention(
             query,
             key,
diff --git a/python/paddle/incubate/operators/graph_khop_sampler.py b/python/paddle/incubate/operators/graph_khop_sampler.py
index 7bea5dbe762..06b079d7f1c 100644
--- a/python/paddle/incubate/operators/graph_khop_sampler.py
+++ b/python/paddle/incubate/operators/graph_khop_sampler.py
@@ -14,8 +14,8 @@
 
 from paddle import _legacy_C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 
 def graph_khop_sampler(
@@ -84,7 +84,7 @@ def graph_khop_sampler(
 
     """
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if return_eids:
             if sorted_eids is None:
                 raise ValueError(
diff --git a/python/paddle/incubate/operators/graph_reindex.py b/python/paddle/incubate/operators/graph_reindex.py
index 2e2a0a1642f..2594ed7ce05 100644
--- a/python/paddle/incubate/operators/graph_reindex.py
+++ b/python/paddle/incubate/operators/graph_reindex.py
@@ -14,8 +14,8 @@
 
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.utils import deprecated
 
 
@@ -116,7 +116,7 @@ def graph_reindex(
                 "be None if `flag_buffer_hashtable` is True."
             )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         reindex_src, reindex_dst, out_nodes = _C_ops.reindex_graph(
             x,
             neighbors,
diff --git a/python/paddle/incubate/operators/graph_sample_neighbors.py b/python/paddle/incubate/operators/graph_sample_neighbors.py
index b670674c565..169acca5fdc 100644
--- a/python/paddle/incubate/operators/graph_sample_neighbors.py
+++ b/python/paddle/incubate/operators/graph_sample_neighbors.py
@@ -14,8 +14,8 @@
 
 from paddle import _legacy_C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.utils import deprecated
 
 
@@ -109,7 +109,7 @@ def graph_sample_neighbors(
                 "is True."
             )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         (
             out_neighbors,
             out_count,
diff --git a/python/paddle/incubate/operators/graph_send_recv.py b/python/paddle/incubate/operators/graph_send_recv.py
index 701f7a7fa15..7a874f19249 100644
--- a/python/paddle/incubate/operators/graph_send_recv.py
+++ b/python/paddle/incubate/operators/graph_send_recv.py
@@ -22,8 +22,9 @@ from paddle.fluid.data_feeder import (
     check_variable_and_dtype,
     convert_dtype,
 )
-from paddle.fluid.framework import Variable, in_dygraph_mode
+from paddle.fluid.framework import Variable
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.utils import deprecated
 
 
@@ -124,7 +125,7 @@ def graph_send_recv(
 
     # TODO(daisiming): Should we add judgement for out_size: max(dst_index) + 1.
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out_size = convert_out_size_to_list(out_size)
         return _C_ops.send_u_recv(
             x, src_index, dst_index, pool_type.upper(), out_size
diff --git a/python/paddle/incubate/operators/softmax_mask_fuse.py b/python/paddle/incubate/operators/softmax_mask_fuse.py
index 399f8e9bd98..178cfd9a046 100644
--- a/python/paddle/incubate/operators/softmax_mask_fuse.py
+++ b/python/paddle/incubate/operators/softmax_mask_fuse.py
@@ -13,8 +13,8 @@
 # limitations under the License.
 
 from paddle import _legacy_C_ops
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 
 def softmax_mask_fuse(x, mask, name=None):
@@ -55,7 +55,7 @@ def softmax_mask_fuse(x, mask, name=None):
             rst = incubate.softmax_mask_fuse(x, mask)
             # [[[[0.02404429, 0.04658398, 0.02746007, ..., 0.01489375, 0.02397441, 0.02851614] ... ]]]
     """
-    if _non_static_mode():
+    if in_dynamic_mode():
         out = _legacy_C_ops.fused_softmax_mask(x, mask)
         return out
     helper = LayerHelper('fused_softmax_mask', **locals())
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 ffe8d8ac5ad..dd8e229a1e9 100644
--- a/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py
+++ b/python/paddle/incubate/operators/softmax_mask_fuse_upper_triangle.py
@@ -13,8 +13,8 @@
 # limitations under the License.
 
 from paddle import _legacy_C_ops
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 
 def softmax_mask_fuse_upper_triangle(x):
@@ -55,7 +55,7 @@ def softmax_mask_fuse_upper_triangle(x):
             #    [0.32674268, 0.28156221, 0.39169508, ..., 0., 0., 0.]
             #     ... ]]]
     """
-    if _non_static_mode():
+    if in_dynamic_mode():
         out = _legacy_C_ops.fused_softmax_mask_upper_triangle(x)
         return out
 
diff --git a/python/paddle/incubate/optimizer/distributed_fused_lamb.py b/python/paddle/incubate/optimizer/distributed_fused_lamb.py
index a9d9e941d25..1efc75cb7d6 100644
--- a/python/paddle/incubate/optimizer/distributed_fused_lamb.py
+++ b/python/paddle/incubate/optimizer/distributed_fused_lamb.py
@@ -15,7 +15,7 @@
 import os
 
 import paddle
-from paddle.fluid import core, framework, unique_name
+from paddle.fluid import core, unique_name
 from paddle.fluid.executor import global_scope
 from paddle.fluid.framework import Variable, name_scope
 from paddle.fluid.layer_helper import LayerHelper
@@ -129,7 +129,7 @@ class DistributedFusedLamb(Optimizer):
         name=None,
     ):
         assert (
-            not framework._non_static_mode()
+            not paddle.in_dynamic_mode()
         ), "DistributedFusedLamb does not support dygraph mode"
         super().__init__(learning_rate=learning_rate, grad_clip=None, name=name)
 
diff --git a/python/paddle/incubate/optimizer/modelaverage.py b/python/paddle/incubate/optimizer/modelaverage.py
index 338b3351021..d0c73e306e7 100644
--- a/python/paddle/incubate/optimizer/modelaverage.py
+++ b/python/paddle/incubate/optimizer/modelaverage.py
@@ -13,12 +13,13 @@
 # limitations under the License.
 
 import paddle
-from paddle import _C_ops, _legacy_C_ops
+from paddle import _C_ops
 from paddle.fluid import framework
 from paddle.fluid.dygraph import base as imperative_base
-from paddle.fluid.framework import Program, in_dygraph_mode
+from paddle.fluid.framework import Program
 from paddle.fluid.layer_helper import LayerHelper
 from paddle.fluid.wrapped_decorator import signature_safe_contextmanager
+from paddle.framework import in_dynamic_mode
 from paddle.optimizer import Optimizer
 
 __all__ = []
@@ -183,7 +184,7 @@ class ModelAverage(Optimizer):
         self.max_average_window = max_average_window
         self.type = "average_accumulates"
 
-        if not framework._non_static_mode():
+        if not in_dynamic_mode():
             global_block = framework.default_main_program().global_block()
             all_parameters = (
                 parameters if parameters else global_block.all_parameters()
@@ -235,7 +236,7 @@ class ModelAverage(Optimizer):
         )
         num_updates = self._get_accumulator('num_updates', param_and_grad[0])
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             _, _, _, _, _, _ = _C_ops.average_accumulates_(
                 param_and_grad[0],
                 sum_1,
@@ -249,29 +250,6 @@ class ModelAverage(Optimizer):
                 self.min_average_window,
             )
             return None
-        elif framework._non_static_mode():
-            _, _, _, _, _, _ = _legacy_C_ops.average_accumulates(
-                param_and_grad[0],
-                sum_1,
-                sum_2,
-                sum_3,
-                num_accumulates,
-                old_num_accumulates,
-                num_updates,
-                sum_1,
-                sum_2,
-                sum_3,
-                num_accumulates,
-                old_num_accumulates,
-                num_updates,
-                'average_window',
-                self.average_window,
-                'min_average_window',
-                self.min_average_window,
-                'max_average_window',
-                self.max_average_window,
-            )
-            return None
 
         block = framework.default_main_program().global_block()
         attrs = {
@@ -358,7 +336,7 @@ class ModelAverage(Optimizer):
                 modelaverage.clear_grad()
 
         """
-        if framework._non_static_mode():
+        if in_dynamic_mode():
             self.step()
 
     @framework.dygraph_only
@@ -443,7 +421,7 @@ class ModelAverage(Optimizer):
                 for param in linear.parameters():
                     print(param)
         """
-        if framework._non_static_mode():
+        if in_dynamic_mode():
             for param in self._parameter_list:
                 num_accumulates = self._get_accumulator(
                     'num_accumulates', param
@@ -522,7 +500,7 @@ class ModelAverage(Optimizer):
                 for param in linear.parameters():
                     print(param)
         """
-        if framework._non_static_mode():
+        if in_dynamic_mode():
             for param in self._parameter_list:
                 param_restore = self._get_accumulator('restore', param)
                 paddle.assign(param_restore, param)
diff --git a/python/paddle/incubate/tensor/manipulation.py b/python/paddle/incubate/tensor/manipulation.py
index 4e39dfa0c79..4e1380a0bd7 100644
--- a/python/paddle/incubate/tensor/manipulation.py
+++ b/python/paddle/incubate/tensor/manipulation.py
@@ -14,8 +14,8 @@
 
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 __all__ = []
 
@@ -47,7 +47,7 @@ def _npu_identity(x, format=-1):
             y = paddle.incubate._npu_identity(x, 3) # ACL_FORMAT_NC1HWC0 = 3
             # y.shape = [1, 1, 1, 1, 16]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.npu_identity(x, format)
     else:
         check_variable_and_dtype(
diff --git a/python/paddle/incubate/tensor/math.py b/python/paddle/incubate/tensor/math.py
index e7f8246ba90..04ed1da3e19 100644
--- a/python/paddle/incubate/tensor/math.py
+++ b/python/paddle/incubate/tensor/math.py
@@ -14,8 +14,8 @@
 
 from paddle import _C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.utils import deprecated
 
 __all__ = []
@@ -63,7 +63,7 @@ def segment_sum(data, segment_ids, name=None):
             #Outputs: [[4., 4., 4.], [4., 5., 6.]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "SUM")
     else:
         check_variable_and_dtype(
@@ -129,7 +129,7 @@ def segment_mean(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "MEAN")
 
     check_variable_and_dtype(
@@ -194,7 +194,7 @@ def segment_min(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.segment_pool(data, segment_ids, "MIN")
 
     check_variable_and_dtype(
@@ -259,7 +259,7 @@ def segment_max(data, segment_ids, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.segment_pool(data, segment_ids, "MAX")
         return out
 
diff --git a/python/paddle/incubate/xpu/resnet_block.py b/python/paddle/incubate/xpu/resnet_block.py
index d5c70e08605..1694679babd 100644
--- a/python/paddle/incubate/xpu/resnet_block.py
+++ b/python/paddle/incubate/xpu/resnet_block.py
@@ -61,7 +61,7 @@ def resnet_basic_block(
     find_conv_max=True,
 ):
 
-    if fluid.framework._non_static_mode():
+    if fluid.framework.in_dygraph_mode():
         attrs = (
             'stride1',
             stride1,
diff --git a/python/paddle/io/dataloader/dataloader_iter.py b/python/paddle/io/dataloader/dataloader_iter.py
index 43b749c869d..b212dd42f1d 100644
--- a/python/paddle/io/dataloader/dataloader_iter.py
+++ b/python/paddle/io/dataloader/dataloader_iter.py
@@ -29,7 +29,7 @@ from paddle.fluid.framework import _current_expected_place, _set_expected_place
 from paddle.profiler.timer import benchmark
 from paddle.profiler.utils import in_profiler_mode
 
-from ...framework import core, in_dygraph_mode
+from ...framework import core, in_dynamic_mode
 from ..multiprocess_utils import (
     MP_STATUS_CHECK_INTERVAL,
     CleanupFuncRegistrar,
@@ -286,7 +286,7 @@ class _DataLoaderIterSingleProcess(_DataLoaderIterBase):
         try:
             benchmark().check_if_need_record(self)
             benchmark().before_reader()
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 data = core.eager.read_next_tensor_list(
                     self._reader.read_next_list()[0]
                 )
@@ -535,7 +535,7 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
         # in order not to restart the thread, we just clear
         # the blocking_queue cachees instead of recreating one
         while self._blocking_queue.size() >= len(self._places):
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 data = core.eager.read_next_tensor_list(
                     self._reader.read_next_list()[0]
                 )
@@ -820,7 +820,7 @@ class _DataLoaderIterMultiProcess(_DataLoaderIterBase):
                     self._thread_done_event.set()
                     self._blocking_queue.close()
 
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 data = core.eager.read_next_tensor_list(
                     self._reader.read_next_list()[0]
                 )
diff --git a/python/paddle/io/dataloader/dataset.py b/python/paddle/io/dataloader/dataset.py
index e8bb6bbd364..4442531f7d1 100755
--- a/python/paddle/io/dataloader/dataset.py
+++ b/python/paddle/io/dataloader/dataset.py
@@ -268,7 +268,7 @@ class TensorDataset(Dataset):
     """
 
     def __init__(self, tensors):
-        if not framework._non_static_mode():
+        if not framework.in_dynamic_mode():
             raise RuntimeError(
                 "TensorDataset con only be used in imperative mode"
             )
diff --git a/python/paddle/io/reader.py b/python/paddle/io/reader.py
index 6698caa435f..861d1253dcf 100644
--- a/python/paddle/io/reader.py
+++ b/python/paddle/io/reader.py
@@ -27,9 +27,8 @@ from ..fluid.framework import (
     _current_expected_place,
     _get_paddle_place,
     _get_paddle_place_list,
-    _non_static_mode,
 )
-from ..framework import core
+from ..framework import core, in_dynamic_mode
 from .dataloader import BatchSampler, IterableDataset, Subset
 from .dataloader.batch_sampler import _InfiniteIterableSampler
 from .dataloader.dataloader_iter import (
@@ -413,7 +412,7 @@ class DataLoader:
 
         self.dataset = dataset
 
-        if not return_list and not _non_static_mode():
+        if not return_list and not in_dynamic_mode():
             assert (
                 feed_list is not None
             ), "feed_list should be set when return_list=False"
@@ -494,7 +493,7 @@ class DataLoader:
         self.auto_collate_batch = self.batch_sampler is not None
 
         self.pin_memory = False
-        if _non_static_mode():
+        if in_dynamic_mode():
             self.pin_memory = (
                 True if use_pinned_memory() is None else use_pinned_memory()
             )
diff --git a/python/paddle/jit/api.py b/python/paddle/jit/api.py
index 14c5a4cc0fc..827bb9806a1 100644
--- a/python/paddle/jit/api.py
+++ b/python/paddle/jit/api.py
@@ -69,9 +69,10 @@ from paddle.fluid.framework import (
     _dygraph_guard,
     _dygraph_tracer,
 )
-from paddle.fluid.framework import dygraph_only, _non_static_mode
+from paddle.fluid.framework import dygraph_only
 from paddle.fluid.wrapped_decorator import wrap_decorator
 from paddle.fluid.io import save_inference_model
+from paddle.framework import in_dynamic_mode
 
 
 def create_program_from_desc(program_desc):
@@ -154,7 +155,7 @@ def _dygraph_to_static_func_(dygraph_func):
     # TODO: remove this decorator after we finalize training API
     def __impl__(*args, **kwargs):
         program_translator = ProgramTranslator()
-        if _non_static_mode() or not program_translator.enable_to_static:
+        if in_dynamic_mode() or not program_translator.enable_to_static:
             logging_utils.warn(
                 "The decorator 'dygraph_to_static_func' doesn't work in "
                 "dygraph mode or set 'paddle.jit.enable_to_static' to False. "
@@ -1711,7 +1712,7 @@ class TracedLayer:
         ), "Inputs should be a list or tuple of variables"
         assert len(inputs) == len(self._feed_names)
         feed_dict = {}
-        if _non_static_mode():
+        if in_dynamic_mode():
             for x, name in zip(inputs, self._feed_names):
                 feed_dict[name] = x.value().get_tensor()
         else:
diff --git a/python/paddle/jit/dy2static/program_translator.py b/python/paddle/jit/dy2static/program_translator.py
index a8be1abb2a1..38801a36606 100644
--- a/python/paddle/jit/dy2static/program_translator.py
+++ b/python/paddle/jit/dy2static/program_translator.py
@@ -19,13 +19,14 @@ import threading
 import warnings
 import weakref
 
-from paddle.fluid import _non_static_mode, core, framework
+from paddle.fluid import core, framework
 from paddle.fluid.data_feeder import check_type
 from paddle.fluid.dygraph.base import (
     _switch_declarative_mode_guard_,
     param_guard,
     switch_to_static_graph,
 )
+from paddle.framework import in_dynamic_mode
 from paddle.nn.layer import layers
 from paddle.utils import flatten, gast
 
@@ -457,7 +458,7 @@ class StaticFunction:
             )
             return self._call_dygraph_function(*args, **kwargs)
 
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             raise RuntimeError(
                 "Failed to run the callable object {} decorated by '@paddle.jit.to_static', "
                 "because it is NOT in dynamic mode. Please disable the static graph mode to enter dynamic mode with the "
diff --git a/python/paddle/jit/translated_layer.py b/python/paddle/jit/translated_layer.py
index 06a89ce3640..80706eac762 100644
--- a/python/paddle/jit/translated_layer.py
+++ b/python/paddle/jit/translated_layer.py
@@ -22,7 +22,8 @@ from paddle import _legacy_C_ops
 from paddle.fluid import backward, core, framework, unique_name
 from paddle.fluid.data_feeder import check_type
 from paddle.fluid.dygraph.base import switch_to_static_graph
-from paddle.fluid.framework import OpProtoHolder, _non_static_mode
+from paddle.fluid.framework import OpProtoHolder
+from paddle.framework import in_dynamic_mode
 from paddle.jit.dy2static.partial_program import (
     LazyInitialized,
     add_build_strategy_for,
@@ -1486,7 +1487,7 @@ class TranslatedLayer(layers.Layer):
             program_holder = self._program_holder_dict[__i_m_p_l__.__name__]
             # When using jit.save, it runs in static graph mode.
             # Run in dynamic graph mode when the model is inferring.
-            if _non_static_mode():
+            if in_dynamic_mode():
                 return _run_dygraph(self, input, program_holder)
             else:
                 # NOTE(weixin): [ why not use 'program_holder.infer_program' directly? ]
diff --git a/python/paddle/metric/metrics.py b/python/paddle/metric/metrics.py
index bf3f164c338..826e781c61d 100644
--- a/python/paddle/metric/metrics.py
+++ b/python/paddle/metric/metrics.py
@@ -20,8 +20,9 @@ import paddle
 from paddle import _legacy_C_ops
 
 from ..fluid.data_feeder import check_variable_and_dtype
-from ..fluid.framework import _create_tensor, _non_static_mode
+from ..fluid.framework import _create_tensor
 from ..fluid.layer_helper import LayerHelper
+from ..framework import in_dynamic_mode
 
 __all__ = []
 
@@ -802,7 +803,7 @@ def accuracy(input, label, k=1, correct=None, total=None, name=None):
     """
     if label.dtype == paddle.int32:
         label = paddle.cast(label, paddle.int64)
-    if _non_static_mode():
+    if in_dynamic_mode():
         if correct is None:
             correct = _create_tensor(dtype="int32")
         if total is None:
diff --git a/python/paddle/nn/clip.py b/python/paddle/nn/clip.py
index 1952ea0514b..f8c674c8a0b 100644
--- a/python/paddle/nn/clip.py
+++ b/python/paddle/nn/clip.py
@@ -18,11 +18,11 @@ from sqlite3 import NotSupportedError
 
 import paddle
 import paddle.autograd as imperative_base
-from paddle import _C_ops, _legacy_C_ops
+from paddle import _C_ops
 from paddle.common_ops_import import Variable, check_type, default_main_program
 from paddle.fluid import core, framework, unique_name
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.framework import LayerHelper, _non_static_mode, in_dygraph_mode
+from paddle.framework import LayerHelper, in_dynamic_mode
 from paddle.tensor.layer_function_generator import templatedoc
 
 __all__ = []
@@ -57,10 +57,8 @@ def clip_by_norm(x, max_norm, name=None):
             # [[0.5, 0.5], [0.5, 0.5]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.clip_by_norm(x, max_norm)
-    if _non_static_mode():
-        return _legacy_C_ops.clip_by_norm(x, 'max_norm', max_norm)
 
     helper = LayerHelper("clip_by_norm", **locals())
     check_variable_and_dtype(
@@ -109,12 +107,9 @@ def merge_selected_rows(x, name=None):
                 type=fluid.core.VarDesc.VarType.SELECTED_ROWS)
             y = nn.merge_selected_rows(var)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.merge_selected_rows(x)
 
-    if _non_static_mode():
-        return _legacy_C_ops.merge_selected_rows(x)
-
     helper = LayerHelper("merge_selected_rows", **locals())
     out = helper.create_variable_for_type_inference(dtype=x.dtype)
     helper.append_op(
@@ -216,7 +211,7 @@ def _squared_l2_norm(x):
         sum_square = paddle.sum(square)
         return sum_square
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.squared_l2_norm(x)
 
     op_type = 'squared_l2_norm'
@@ -335,7 +330,7 @@ class ClipGradBase:
         raise NotImplementedError
 
     def __call__(self, params_grads):
-        if _non_static_mode():
+        if in_dynamic_mode():
             return self._dygraph_clip(params_grads)
         else:
             for p, g in params_grads:
@@ -655,7 +650,7 @@ class ClipGradByGlobalNorm(ClipGradBase):
                 continue
             merge_grad = g
 
-            if in_dygraph_mode() and g.is_selected_rows():
+            if in_dynamic_mode() and g.is_selected_rows():
                 merge_grad = merge_selected_rows(g)
                 merge_grad = merge_grad._get_tensor_from_selected_rows()
 
diff --git a/python/paddle/nn/decode.py b/python/paddle/nn/decode.py
index 44598d3917a..030a2a04881 100644
--- a/python/paddle/nn/decode.py
+++ b/python/paddle/nn/decode.py
@@ -20,7 +20,7 @@ import numpy as np
 
 import paddle
 from paddle.common_ops_import import default_main_program
-from paddle.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 from ..fluid.data_feeder import convert_dtype
 
@@ -1071,7 +1071,7 @@ def dynamic_decode(
                                     inits=decoder_cell.get_initial_states(encoder_output),
                                     max_step_num=10)
     """
-    if _non_static_mode():
+    if in_dynamic_mode():
         return _dynamic_decode_imperative(
             decoder,
             inits,
diff --git a/python/paddle/nn/functional/activation.py b/python/paddle/nn/functional/activation.py
index 5bb9d2b1d03..9220d14ab80 100644
--- a/python/paddle/nn/functional/activation.py
+++ b/python/paddle/nn/functional/activation.py
@@ -18,7 +18,7 @@ from paddle.framework import core
 from paddle.utils.inplace_utils import inplace_apis_in_dygraph_only
 
 from ...fluid.data_feeder import check_dtype, check_variable_and_dtype
-from ...fluid.framework import convert_np_dtype_to_dtype_, in_dygraph_mode
+from ...fluid.framework import convert_np_dtype_to_dtype_
 from ...fluid.layer_helper import LayerHelper
 from ...tensor.manipulation import chunk
 from ...tensor.math import tanh  # noqa: F401
@@ -58,7 +58,7 @@ def celu(x, alpha=1.0, name=None):
     """
     if alpha == 0:
         raise ZeroDivisionError("alpha cannot be 0 for celu")
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.celu(x, alpha)
     else:
         check_variable_and_dtype(
@@ -109,7 +109,7 @@ def elu(x, alpha=1.0, name=None):
             #  [ 1.          15.6      ]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.elu(x, alpha)
 
     else:
@@ -134,7 +134,7 @@ def elu_(x, alpha=1.0, name=None):
     Please refer to :ref:`api_nn_cn_elu`.
     """
     assert alpha >= 0.0, "elu_ only support alpha >= 0, please use elu instead."
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.elu_(x, alpha)
     return _legacy_C_ops.elu_(x, 'alpha', alpha)
 
@@ -180,7 +180,7 @@ def gelu(x, approximate=False, name=None):
             #  [ 0.84119201,  1.39957154]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.gelu(x, approximate)
     else:
         check_variable_and_dtype(
@@ -230,7 +230,7 @@ def hardshrink(x, threshold=0.5, name=None):
             out = F.hardshrink(x) # [-1., 0., 2.5]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.hardshrink(x, threshold)
     else:
         check_variable_and_dtype(
@@ -281,7 +281,7 @@ def hardtanh(x, min=-1.0, max=1.0, name=None):
             out = F.hardtanh(x) # [-1., 0.3, 1.]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.hardtanh(x, min, max)
     else:
         check_variable_and_dtype(
@@ -335,7 +335,7 @@ def hardsigmoid(x, slope=0.1666667, offset=0.5, name=None):
             out = F.hardsigmoid(x) # [0., 1., 0.666667]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.hardsigmoid(x, slope, offset)
     else:
         check_variable_and_dtype(
@@ -386,7 +386,7 @@ def hardswish(x, name=None):
             x = paddle.to_tensor([-4., 5., 1.])
             out = F.hardswish(x) # [0., 5., 0.666667]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.hardswish(x)
     else:
         check_variable_and_dtype(
@@ -435,7 +435,7 @@ def leaky_relu(x, negative_slope=0.01, name=None):
             # [-0.02, 0., 1.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.leaky_relu(x, negative_slope)
     else:
         check_variable_and_dtype(
@@ -534,7 +534,7 @@ def prelu(x, weight, data_format="NCHW", name=None):
             ), "The weight size should be equal to x input channel in prelu() when weight shape is not [1]."
         mode = 'channel'
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.prelu(x, weight, data_format, mode)
     else:
         check_variable_and_dtype(
@@ -657,7 +657,7 @@ def rrelu(x, lower=1.0 / 8.0, upper=1.0 / 3.0, training=True, name=None):
 
     is_test = not training
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.rrelu(x, lower, upper, is_test)
     else:
         check_variable_and_dtype(
@@ -705,7 +705,7 @@ def relu(x, name=None):
             # [0., 0., 1.]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.relu(x)
     else:
         check_variable_and_dtype(
@@ -751,7 +751,7 @@ def log_sigmoid(x, name=None):
             out = F.log_sigmoid(x) # [-0.313262 -0.126928 -0.0485874 -0.0181499]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logsigmoid(x)
     else:
         check_variable_and_dtype(
@@ -819,7 +819,7 @@ def maxout(x, groups, axis=1, name=None):
             #    [0.95313174 0.6228939  0.7129065  0.7087491 ]
             #    [0.7142536  0.88725346 0.61093384 0.38833922]]]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.maxout(x, groups, axis)
     else:
         check_variable_and_dtype(
@@ -871,10 +871,8 @@ def relu6(x, name=None):
             # [0, 0.3, 6]
     """
     threshold = 6.0
-    if in_dygraph_mode():
-        return _C_ops.relu6(x)
     if in_dynamic_mode():
-        return _legacy_C_ops.relu6(x, 'threshold', threshold)
+        return _C_ops.relu6(x)
 
     check_variable_and_dtype(
         x, 'x', ['float16', 'uint16', 'float32', 'float64'], 'relu6'
@@ -939,7 +937,7 @@ def selu(
             f"The alpha must be no less than zero. Received: {alpha}."
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.selu(x, scale, alpha)
     else:
         check_variable_and_dtype(
@@ -983,7 +981,7 @@ def silu(x, name=None):
             out = F.silu(x) # [ 0.731059, 1.761594, 2.857722, 3.928055 ]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.silu(x)
     else:
         check_variable_and_dtype(
@@ -1110,7 +1108,7 @@ def softmax(x, axis=-1, dtype=None, name=None):
 
     if (dtype is not None) and (not isinstance(dtype, core.VarDesc.VarType)):
         dtype = convert_np_dtype_to_dtype_(dtype)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         outs_cast = x if dtype is None else _C_ops.cast(x, dtype)
         return _C_ops.softmax(outs_cast, axis)
     else:
@@ -1197,7 +1195,7 @@ def softplus(x, beta=1, threshold=20, name=None):
             out = F.softplus(x) # [0.513015, 0.598139, 0.744397, 0.854355]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.softplus(x, beta, threshold)
     else:
         check_variable_and_dtype(
@@ -1256,7 +1254,7 @@ def softshrink(x, threshold=0.5, name=None):
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.softshrink(x, threshold)
     else:
         check_variable_and_dtype(
@@ -1300,10 +1298,8 @@ def softsign(x, name=None):
             # Tensor(shape=[4], dtype=float32, place=Place(gpu:0), stop_gradient=True,
             #        [-0.28571430, -0.16666666,  0.09090909,  0.23076925])
     """
-    if in_dygraph_mode():
-        return _C_ops.softsign(x)
     if in_dynamic_mode():
-        return _legacy_C_ops.softsign(x)
+        return _C_ops.softsign(x)
 
     check_variable_and_dtype(
         x, 'x', ['float16', 'uint16', 'float32', 'float64'], 'softsign'
@@ -1341,7 +1337,7 @@ def swish(x, name=None):
             # Tensor(shape=[3], dtype=float32, place=Place(gpu:0), stop_gradient=True,
             #        [-0.23840584,  0.        ,  0.73105854])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.swish(x)
     else:
         check_variable_and_dtype(
@@ -1387,7 +1383,7 @@ def mish(x, name=None):
             x = paddle.to_tensor([-5., 0., 5.])
             out = F.mish(x) # [-0.03357624, 0., 4.99955208]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.mish(x, 20)
     else:
         check_variable_and_dtype(
@@ -1426,7 +1422,7 @@ def tanhshrink(x, name=None):
             # Tensor(shape=[4], dtype=float32, place=Place(gpu:0), stop_gradient=True,
             #        [-0.02005106, -0.00262468,  0.00033200,  0.00868741])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.tanh_shrink(x)
     else:
         check_variable_and_dtype(
@@ -1476,7 +1472,7 @@ def thresholded_relu(x, threshold=1.0, name=None):
             #        [2., 0., 0.])
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.thresholded_relu(x, threshold)
     else:
         check_variable_and_dtype(
@@ -1554,7 +1550,7 @@ def log_softmax(x, axis=-1, dtype=None, name=None):
     if (dtype is not None) and (not isinstance(dtype, core.VarDesc.VarType)):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if dtype is not None:
             x = _C_ops.cast(x, dtype)
         return _C_ops.log_softmax(x, axis)
@@ -1709,13 +1705,8 @@ def gumbel_softmax(x, temperature=1.0, hard=False, axis=-1, name=None):
             # [0.00000000, 0.00000000, 0.00000000, 0.00001258, 0.99998736, 0.00000000]]
 
     """
-    if in_dygraph_mode():
-        return _C_ops.gumbel_softmax(x, temperature, hard, axis)
-
     if in_dynamic_mode():
-        return _legacy_C_ops.gumbel_softmax(
-            x, 'temperature', temperature, 'hard', hard, 'axis', axis
-        )
+        return _C_ops.gumbel_softmax(x, temperature, hard, axis)
 
     helper = LayerHelper("gumbel_softmax", **locals())
     check_variable_and_dtype(
diff --git a/python/paddle/nn/functional/common.py b/python/paddle/nn/functional/common.py
index 01c323d6596..d788cc46d0a 100644
--- a/python/paddle/nn/functional/common.py
+++ b/python/paddle/nn/functional/common.py
@@ -15,7 +15,7 @@
 import numpy
 
 import paddle
-from paddle import _C_ops, _legacy_C_ops
+from paddle import _C_ops
 from paddle.common_ops_import import Variable, default_main_program
 from paddle.fluid.layer_helper import LayerHelper
 from paddle.framework import core, in_dynamic_mode
@@ -26,7 +26,6 @@ from ...fluid.data_feeder import (
     check_type,
     check_variable_and_dtype,
 )
-from ...fluid.framework import in_dygraph_mode
 from ...tensor import clip, concat, sqrt, sum
 from ...tensor.creation import zeros
 
@@ -147,7 +146,7 @@ def unfold(x, kernel_sizes, strides=1, paddings=0, dilations=1, name=None):
             "of 2 or 4 integers"
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.unfold(x, kernel_sizes, strides, paddings, dilations)
 
     out = helper.create_variable_for_type_inference(dtype=x.dtype)
@@ -605,95 +604,80 @@ def interpolate(
         dy_attr = tuple(attr_list)
 
         if resample_type == "linear":
-            if in_dygraph_mode():
-                out = _C_ops.linear_interp(
-                    x,
-                    inputs['OutSize'] if 'OutSize' in inputs else None,
-                    inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
-                    inputs['Scale'] if 'Scale' in inputs else None,
-                    attrs['data_layout'],
-                    attrs['out_d'],
-                    attrs['out_h'],
-                    attrs['out_w'],
-                    attrs['scale'] if 'scale' in attrs else [],
-                    attrs['interp_method'],
-                    attrs['align_corners'],
-                    attrs['align_mode'],
-                )
-            else:
-                out = _legacy_C_ops.linear_interp_v2(x, *dy_attr)
+            out = _C_ops.linear_interp(
+                x,
+                inputs['OutSize'] if 'OutSize' in inputs else None,
+                inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
+                inputs['Scale'] if 'Scale' in inputs else None,
+                attrs['data_layout'],
+                attrs['out_d'],
+                attrs['out_h'],
+                attrs['out_w'],
+                attrs['scale'] if 'scale' in attrs else [],
+                attrs['interp_method'],
+                attrs['align_corners'],
+                attrs['align_mode'],
+            )
         elif resample_type == "bilinear":
-            if in_dygraph_mode():
-                out = _C_ops.bilinear_interp(
-                    x,
-                    inputs['OutSize'] if 'OutSize' in inputs else None,
-                    inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
-                    inputs['Scale'] if 'Scale' in inputs else None,
-                    attrs['data_layout'],
-                    attrs['out_d'],
-                    attrs['out_h'],
-                    attrs['out_w'],
-                    attrs['scale'] if 'scale' in attrs else [],
-                    attrs['interp_method'],
-                    attrs['align_corners'],
-                    attrs['align_mode'],
-                )
-            else:
-                out = _legacy_C_ops.bilinear_interp_v2(x, *dy_attr)
+            out = _C_ops.bilinear_interp(
+                x,
+                inputs['OutSize'] if 'OutSize' in inputs else None,
+                inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
+                inputs['Scale'] if 'Scale' in inputs else None,
+                attrs['data_layout'],
+                attrs['out_d'],
+                attrs['out_h'],
+                attrs['out_w'],
+                attrs['scale'] if 'scale' in attrs else [],
+                attrs['interp_method'],
+                attrs['align_corners'],
+                attrs['align_mode'],
+            )
         elif resample_type == "trilinear":
-            if in_dygraph_mode():
-                out = _C_ops.trilinear_interp(
-                    x,
-                    inputs['OutSize'] if 'OutSize' in inputs else None,
-                    inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
-                    inputs['Scale'] if 'Scale' in inputs else None,
-                    attrs['data_layout'],
-                    attrs['out_d'],
-                    attrs['out_h'],
-                    attrs['out_w'],
-                    attrs['scale'] if 'scale' in attrs else [],
-                    attrs['interp_method'],
-                    attrs['align_corners'],
-                    attrs['align_mode'],
-                )
-            else:
-                out = _legacy_C_ops.trilinear_interp_v2(x, *dy_attr)
+            out = _C_ops.trilinear_interp(
+                x,
+                inputs['OutSize'] if 'OutSize' in inputs else None,
+                inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
+                inputs['Scale'] if 'Scale' in inputs else None,
+                attrs['data_layout'],
+                attrs['out_d'],
+                attrs['out_h'],
+                attrs['out_w'],
+                attrs['scale'] if 'scale' in attrs else [],
+                attrs['interp_method'],
+                attrs['align_corners'],
+                attrs['align_mode'],
+            )
         elif resample_type == "nearest":
-            if in_dygraph_mode():
-                out = _C_ops.nearest_interp(
-                    x,
-                    inputs['OutSize'] if 'OutSize' in inputs else None,
-                    inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
-                    inputs['Scale'] if 'Scale' in inputs else None,
-                    attrs['data_layout'],
-                    attrs['out_d'],
-                    attrs['out_h'],
-                    attrs['out_w'],
-                    attrs['scale'] if 'scale' in attrs else [],
-                    attrs['interp_method'],
-                    attrs['align_corners'],
-                    attrs['align_mode'],
-                )
-            else:
-                out = _legacy_C_ops.nearest_interp_v2(x, *dy_attr)
+            out = _C_ops.nearest_interp(
+                x,
+                inputs['OutSize'] if 'OutSize' in inputs else None,
+                inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
+                inputs['Scale'] if 'Scale' in inputs else None,
+                attrs['data_layout'],
+                attrs['out_d'],
+                attrs['out_h'],
+                attrs['out_w'],
+                attrs['scale'] if 'scale' in attrs else [],
+                attrs['interp_method'],
+                attrs['align_corners'],
+                attrs['align_mode'],
+            )
         elif resample_type == "bicubic":
-            if in_dygraph_mode():
-                out = _C_ops.bicubic_interp(
-                    x,
-                    inputs['OutSize'] if 'OutSize' in inputs else None,
-                    inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
-                    inputs['Scale'] if 'Scale' in inputs else None,
-                    attrs['data_layout'],
-                    attrs['out_d'],
-                    attrs['out_h'],
-                    attrs['out_w'],
-                    attrs['scale'] if 'scale' in attrs else [],
-                    attrs['interp_method'],
-                    attrs['align_corners'],
-                    attrs['align_mode'],
-                )
-            else:
-                out = _legacy_C_ops.bicubic_interp_v2(x, *dy_attr)
+            out = _C_ops.bicubic_interp(
+                x,
+                inputs['OutSize'] if 'OutSize' in inputs else None,
+                inputs['SizeTensor'] if 'SizeTensor' in inputs else None,
+                inputs['Scale'] if 'Scale' in inputs else None,
+                attrs['data_layout'],
+                attrs['out_d'],
+                attrs['out_h'],
+                attrs['out_w'],
+                attrs['scale'] if 'scale' in attrs else [],
+                attrs['interp_method'],
+                attrs['align_corners'],
+                attrs['align_mode'],
+            )
         return out
 
     dtype = helper.input_dtype(input_param_name='x')
@@ -941,7 +925,7 @@ def bilinear(x1, x2, weight, bias=None, name=None):
             # [5, 1000]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.bilinear(x1, x2, weight, bias)
     else:
         check_variable_and_dtype(x1, 'x1', ['float32', 'float64'], 'bilinear')
@@ -1132,7 +1116,7 @@ def dropout(
             'downgrade_in_infer' if mode == 'downscale_in_infer' else mode
         )  # semantic transfer
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if default_main_program().random_seed != 0:
                 seed = default_main_program().random_seed
 
@@ -1583,7 +1567,7 @@ def pad(x, pad, mode='constant', value=0.0, data_format="NCHW", name=None):
         paddings = pad
         pad_value = value
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             out = _C_ops.pad(x, paddings, float(pad_value))
             return out
 
@@ -1681,7 +1665,7 @@ def pad(x, pad, mode='constant', value=0.0, data_format="NCHW", name=None):
                 unsqueezed_dim = [1]
                 x = unsqueeze(x, axis=unsqueezed_dim)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(pad, Variable):
             pad = pad.tolist()
         out = _C_ops.pad3d(x, pad, mode, value, data_format)
@@ -1853,7 +1837,7 @@ def linear(x, weight, bias=None, name=None):
           #     [0.9440598  0.9440598  0.9440598  0.9440598 ]
           #     [2.1077576  2.1077576  2.1077576  2.1077576 ]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         # TODO(jiabin): using addmm for fast forward route
         return _C_ops.linear(x, weight, bias)
     else:
@@ -1951,14 +1935,9 @@ def label_smooth(label, prior_dist=None, epsilon=0.1, name=None):
     if epsilon > 1.0 or epsilon < 0.0:
         raise ValueError("The value of epsilon must be between 0 and 1.")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.label_smooth(label, prior_dist, float(epsilon))
 
-    elif paddle.in_dynamic_mode():
-        return _legacy_C_ops.label_smooth(
-            label, prior_dist, 'epsilon', float(epsilon)
-        )
-
     check_variable_and_dtype(
         label,
         'label',
@@ -2141,7 +2120,7 @@ def class_center_sample(label, num_classes, num_samples, group=None):
     if (seed is None or seed == 0) and default_main_program().random_seed != 0:
         seed = default_main_program().random_seed
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.class_center_sample(
             label,
             num_classes,
@@ -2152,28 +2131,6 @@ def class_center_sample(label, num_classes, num_samples, group=None):
             seed is not None,
             seed if seed is not None else 0,
         )
-    elif paddle.in_dynamic_mode():
-        (
-            remapped_label,
-            sampled_class_center,
-        ) = _legacy_C_ops.class_center_sample(
-            label,
-            'num_classes',
-            num_classes,
-            'num_samples',
-            num_samples,
-            'ring_id',
-            ring_id,
-            'nranks',
-            nranks,
-            'rank',
-            rank,
-            'fix_seed',
-            seed is not None,
-            'seed',
-            seed if seed is not None else 0,
-        )
-        return remapped_label, sampled_class_center
 
     check_variable_and_dtype(
         label, 'label', ['int64', 'int32'], 'class_center_sample'
@@ -2321,24 +2278,10 @@ def fold(
             "of 2 or 4 integers"
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.fold(
             x, output_sizes, kernel_sizes, strides, paddings, dilations
         )
-    elif in_dynamic_mode():
-        out = _legacy_C_ops.fold(
-            x,
-            "output_sizes",
-            output_sizes,
-            "kernel_sizes",
-            kernel_sizes,
-            "strides",
-            strides,
-            "paddings",
-            paddings,
-            "dilations",
-            dilations,
-        )
     else:
         out = helper.create_variable_for_type_inference(dtype=x.dtype)
         helper.append_op(
diff --git a/python/paddle/nn/functional/conv.py b/python/paddle/nn/functional/conv.py
index e44847b0a9b..671ed1765cf 100644
--- a/python/paddle/nn/functional/conv.py
+++ b/python/paddle/nn/functional/conv.py
@@ -18,7 +18,7 @@ from paddle.device import (
     is_compiled_with_cuda,
     is_compiled_with_rocm,
 )
-from paddle.fluid.framework import _global_flags, in_dygraph_mode
+from paddle.fluid.framework import _global_flags
 from paddle.tensor.manipulation import reshape
 from paddle.tensor.math import _add_with_axis
 
@@ -129,7 +129,7 @@ def _conv_nd(
 ):
 
     # Due to the poor performance of NHWC, we transpose the input to NCHW.
-    if in_dygraph_mode() and op_type == "conv2d":
+    if in_dynamic_mode() and op_type == "conv2d":
         pre_bias = _C_ops.conv2d(
             x,
             weight,
@@ -158,7 +158,7 @@ def _conv_nd(
         else:
             return pre_bias
 
-    if in_dygraph_mode() and op_type == "depthwise_conv2d":
+    if in_dynamic_mode() and op_type == "depthwise_conv2d":
         pre_bias = _C_ops.depthwise_conv2d(
             x,
             weight,
@@ -177,7 +177,7 @@ def _conv_nd(
         else:
             return pre_bias
 
-    if in_dygraph_mode() and op_type == "conv3d":
+    if in_dynamic_mode() and op_type == "conv3d":
         pre_bias = _C_ops.conv3d(
             x,
             weight,
@@ -467,7 +467,7 @@ def conv1d(
     squeeze_aixs = -3 if channel_last else -2
     x = unsqueeze(x, axis=[squeeze_aixs])
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if l_type == 'conv2d':
             out = _C_ops.conv2d(
                 x,
@@ -707,7 +707,7 @@ def conv2d(
         else:
             use_cudnn = False
     else:
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             pre_bias = _C_ops.conv2d(
                 x,
                 weight,
@@ -1012,7 +1012,7 @@ def conv1d_transpose(
     x = unsqueeze(x, axis=[squeeze_axis])
     weight = unsqueeze(weight, axis=[-1])
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = getattr(_C_ops, op_type)(
             x,
             weight,
@@ -1293,7 +1293,7 @@ def conv2d_transpose(
         op_type = 'depthwise_conv2d_transpose'
         use_cudnn = False
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         op = (
             _C_ops.conv2d_transpose
             if op_type == 'conv2d_transpose'
@@ -1776,7 +1776,7 @@ def conv3d_transpose(
     op_type = 'conv3d_transpose'
     data_format_ = "NHWC" if channel_last else "NCHW"
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         pre_bias = _C_ops.conv3d_transpose(
             x,
             weight,
diff --git a/python/paddle/nn/functional/distance.py b/python/paddle/nn/functional/distance.py
index 067a49816dd..f1155852b00 100644
--- a/python/paddle/nn/functional/distance.py
+++ b/python/paddle/nn/functional/distance.py
@@ -14,7 +14,7 @@
 
 import paddle
 from paddle import _C_ops
-from paddle.fluid.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 from ...fluid.data_feeder import check_type, check_variable_and_dtype
 from ...fluid.layer_helper import LayerHelper
@@ -68,7 +68,7 @@ def pairwise_distance(x, y, p=2.0, epsilon=1e-6, keepdim=False, name=None):
     #              [4.99999860, 4.99999860])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         sub = _C_ops.subtract(x, y)
         # p_norm op has not used epsilon, so change it to the following.
         if epsilon != 0.0:
diff --git a/python/paddle/nn/functional/extension.py b/python/paddle/nn/functional/extension.py
index b9d226374ae..a5c1e2ef42e 100644
--- a/python/paddle/nn/functional/extension.py
+++ b/python/paddle/nn/functional/extension.py
@@ -24,7 +24,6 @@ from ...fluid.data_feeder import (
     check_type,
     check_variable_and_dtype,
 )
-from ...fluid.framework import in_dygraph_mode
 from ...fluid.layer_helper import LayerHelper
 from ...framework import convert_np_dtype_to_dtype_, core
 from ...tensor.creation import assign
@@ -102,12 +101,8 @@ def diag_embed(input, offset=0, dim1=-2, dim2=-1):
     if not isinstance(input, Variable):
         input = assign(input)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.diag_embed(input, offset, dim1, dim2)
-    elif in_dynamic_mode():
-        return _legacy_C_ops.diag_embed(
-            input, "offset", offset, "dim1", dim1, "dim2", dim2
-        )
 
     inputs = {'Input': [input]}
     attrs = {'offset': offset, 'dim1': dim1, 'dim2': dim2}
@@ -218,7 +213,7 @@ def sequence_mask(x, maxlen=None, dtype='int64', name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if not isinstance(dtype, core.VarDesc.VarType):
             dtype = convert_np_dtype_to_dtype_(dtype)
         if maxlen is not None:
@@ -318,7 +313,7 @@ def gather_tree(ids, parents):
     if ids.ndim != parents.ndim:
         raise ValueError("The ids's shape must be the same as parents' shape. ")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.gather_tree(ids, parents)
     else:
         helper = LayerHelper('gather_tree', **locals())
@@ -404,7 +399,7 @@ def temporal_shift(x, seg_num, shift_ratio=0.25, name=None, data_format="NCHW"):
             "Attr(data_format) should be 'NCHW' or 'NHWC'. "
             "Received Attr(data_format): {}.".format(data_format)
         )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.temporal_shift(x, seg_num, shift_ratio, data_format)
     else:
         helper = LayerHelper("temporal_shift", **locals())
diff --git a/python/paddle/nn/functional/input.py b/python/paddle/nn/functional/input.py
index eccaffcb729..793daddd684 100644
--- a/python/paddle/nn/functional/input.py
+++ b/python/paddle/nn/functional/input.py
@@ -16,8 +16,8 @@ from paddle import _C_ops
 
 from ...common_ops_import import Variable
 from ...fluid.data_feeder import check_variable_and_dtype
-from ...fluid.framework import in_dygraph_mode
 from ...fluid.layer_helper import LayerHelper
+from ...framework import in_dynamic_mode
 
 __all__ = []
 
@@ -85,7 +85,7 @@ def one_hot(x, num_classes, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.one_hot(x, num_classes)
     else:
         check_variable_and_dtype(x, 'input', ['int32', 'int64'], 'one_hot_v2')
@@ -201,7 +201,7 @@ def embedding(x, weight, padding_idx=None, sparse=False, name=None):
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.embedding(x, weight, padding_idx, sparse)
     else:
         helper = LayerHelper('embedding', **locals())
diff --git a/python/paddle/nn/functional/loss.py b/python/paddle/nn/functional/loss.py
index 7d71178d4a3..403fb299cbc 100644
--- a/python/paddle/nn/functional/loss.py
+++ b/python/paddle/nn/functional/loss.py
@@ -23,7 +23,7 @@ from paddle.utils import deprecated
 
 from ...common_ops_import import Variable
 from ...fluid.data_feeder import check_variable_and_dtype
-from ...fluid.framework import _current_expected_place, in_dygraph_mode
+from ...fluid.framework import _current_expected_place
 from ...fluid.layer_helper import LayerHelper
 from ...tensor.manipulation import reshape
 
@@ -143,7 +143,7 @@ def log_loss(input, label, epsilon=1e-4, name=None):
           prob = paddle.randn((10,1))
           cost = F.log_loss(input=prob, label=label)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.log_loss(input, label, epsilon)
 
     helper = LayerHelper('log_loss', **locals())
@@ -268,7 +268,7 @@ def fluid_softmax_with_cross_entropy(
         )
     if input_dims - 1 == label_dims:
         label = paddle.unsqueeze(label, axis=axis)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         softmax, loss = _C_ops.cross_entropy_with_softmax(
             logits,
             label,
@@ -418,7 +418,7 @@ def square_error_cost(input, label):
             # [0.01, 0.01]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         minus_out = _C_ops.subtract(input, label)
         square_out = _C_ops.square(minus_out)
         return square_out
@@ -545,7 +545,7 @@ def edit_distance(
         )
         label = erased_label
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.edit_distance(
             input, label, input_length, label_length, normalized
         )
@@ -644,7 +644,7 @@ def binary_cross_entropy(
             % reduction
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.bce_loss(input, label)
         if weight is not None:
             out = _C_ops.multiply(out, weight, 'axis', -1)
@@ -784,7 +784,7 @@ def binary_cross_entropy_with_logits(
             % reduction
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         one = _C_ops.full(
             [1],
             float(1.0),
@@ -963,7 +963,7 @@ def hsigmoid_loss(
     if num_classes < 2:
         raise ValueError(f'Expected num_classes >= 2 (got {num_classes})')
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out, _, _ = _C_ops.hsigmoid_loss(
             input,
             label,
@@ -1080,7 +1080,7 @@ def smooth_l1_loss(input, label, reduction='mean', delta=1.0, name=None):
             # 0.068004
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.huber_loss(input, label, delta)
     else:
         check_variable_and_dtype(
@@ -1172,7 +1172,7 @@ def margin_ranking_loss(
             "The value of 'reduction' in MarginRankingLoss should be 'sum', 'mean' or 'none', but "
             "received %s, which is not allowed." % reduction
         )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.subtract(other, input)
         out = _C_ops.multiply(out, label)
         if margin != 0.0:
@@ -1303,7 +1303,7 @@ def l1_loss(input, label, reduction='mean', name=None):
             "received %s, which is not allowed." % reduction
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         unreduced = _C_ops.abs(_C_ops.subtract(input, label))
 
         if reduction == 'mean':
@@ -1416,7 +1416,7 @@ def nll_loss(
 
     n = input_shape[0]
     c = input_shape[1]
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if input_dims != 2 and input_dims != 4:
             input = _C_ops.reshape(input, [n, c, 1, -1])
             label = _C_ops.reshape(label, [n, 1, -1])
@@ -1646,7 +1646,7 @@ def kl_div(input, label, reduction='mean', name=None):
     ):
         label = paddle.cast(label, 'float64')
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.kldiv_loss(input, label, 'none')
         if reduction == 'mean':
             out = paddle.mean(out)
@@ -1853,7 +1853,7 @@ def ctc_loss(
         input_length=None,
         label_length=None,
     ):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if input_length is None or label_length is None:
                 raise ValueError(
                     "input_length and label_length must not be None in dygraph mode!"
@@ -1974,7 +1974,7 @@ def rnnt_loss(
     def warprnnt(
         input, label, input_length, label_length, blank=0, fastemit_lambda=0.001
     ):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             loss_out = _C_ops.warprnnt(
                 input,
                 label,
@@ -2268,7 +2268,7 @@ def margin_cross_entropy(
     if input_dims - 1 == label_dims:
         label = paddle.unsqueeze(label, axis=-1)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         softmax, loss = _C_ops.margin_cross_entropy(
             logits,
             label,
@@ -2699,7 +2699,7 @@ def cross_entropy(
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if not soft_label:
             valid_label = (
                 paddle.cast(label != ignore_index, dtype=label.dtype) * label
@@ -3058,7 +3058,7 @@ def sigmoid_focal_loss(
                 )
             )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         place = _current_expected_place()
         one = _C_ops.full(logit.shape, float(1.0), logit.dtype, place)
 
@@ -3203,7 +3203,7 @@ def multi_label_soft_margin_loss(
             "but received {}!={}".format(input.shape, label.shape)
         )
 
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             input,
             'input',
@@ -3223,7 +3223,7 @@ def multi_label_soft_margin_loss(
     )
 
     if weight is not None:
-        if not in_dygraph_mode():
+        if not in_dynamic_mode():
             check_variable_and_dtype(
                 weight,
                 'weight',
@@ -3322,7 +3322,7 @@ def hinge_embedding_loss(input, label, margin=1.0, reduction='mean', name=None):
             "but received {}.".format(reduction)
         )
 
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             input, 'input', ['float32', 'float64'], 'hinge_embedding_loss'
         )
@@ -3547,7 +3547,7 @@ def triplet_margin_with_distance_loss(
         raise ValueError(
             "The margin between positive samples and negative samples should be greater than 0."
         )
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             input,
             'input',
@@ -3696,7 +3696,7 @@ def triplet_margin_loss(
         raise ValueError(
             "The margin between positive samples and negative samples should be greater than 0."
         )
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             input, 'input', ['float32', 'float64'], 'triplet_margin_loss'
         )
@@ -3806,7 +3806,7 @@ def multi_margin_loss(
             "but received {}.".format(reduction)
         )
 
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             input, 'input', ['float32', 'float64'], 'multi_margin_loss'
         )
@@ -3823,7 +3823,7 @@ def multi_margin_loss(
     label = label.reshape((-1, 1))
     index_sample = paddle.index_sample(input, label)
     if weight is not None:
-        if not in_dygraph_mode():
+        if not in_dynamic_mode():
             check_variable_and_dtype(
                 weight, 'weight', ['float32', 'float64'], 'multi_margin_loss'
             )
@@ -3927,7 +3927,7 @@ def soft_margin_loss(input, label, reduction='mean', name=None):
             % reduction
         )
 
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         fluid.data_feeder.check_variable_and_dtype(
             input, 'input', ['float32', 'float64'], 'soft_margin_loss'
         )
@@ -4071,7 +4071,7 @@ def gaussian_nll_loss(
         'gaussian_nll_loss',
     )
     # Entries of variance must be non-negative
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         condition = paddle.all(variance > 0)
         Assert(condition, [variance], 6)
     else:
diff --git a/python/paddle/nn/layer/norm.py b/python/paddle/nn/layer/norm.py
index 3e6cb3d387c..0320cf2a6bf 100644
--- a/python/paddle/nn/layer/norm.py
+++ b/python/paddle/nn/layer/norm.py
@@ -32,9 +32,8 @@ import warnings
 
 import numpy as np
 
-from paddle import _C_ops, _legacy_C_ops, in_dynamic_mode
+from paddle import _C_ops, in_dynamic_mode
 from paddle.device import get_all_custom_device_type
-from paddle.fluid.framework import in_dygraph_mode
 
 from ...fluid import dygraph_utils
 from ...fluid.data_feeder import check_variable_and_dtype
@@ -476,7 +475,7 @@ class GroupNorm(Layer):
             )
 
     def forward(self, input):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _C_ops.group_norm(
                 input,
                 self.weight,
@@ -1015,7 +1014,7 @@ class BatchNorm(Layer):
         self._trainable_statistics = trainable_statistics
 
     def forward(self, input):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             batch_norm_out, t1, t2, t3, t4, _ = _C_ops.batch_norm(
                 input,
                 self._mean,
@@ -1543,7 +1542,7 @@ class SyncBatchNorm(_BatchNormBase):
 
         # train mode: use mini-batch stats, eval mode: use global stats
         # use_global_stats only support False in sync_batch_norm
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             sync_batch_norm_out, _, _, _, _, _ = _C_ops.sync_batch_norm_(
                 x,
                 self._mean,
@@ -1559,37 +1558,6 @@ class SyncBatchNorm(_BatchNormBase):
             )
             return sync_batch_norm_out
 
-        elif in_dynamic_mode():
-            attrs = (
-                "momentum",
-                self._momentum,
-                "epsilon",
-                self._epsilon,
-                "is_test",
-                not self.training,
-                "data_layout",
-                self._data_format,
-                "use_mkldnn",
-                False,
-                "fuse_with_relu",
-                False,
-                "use_global_stats",
-                False,
-                'trainable_statistics',
-                False,
-            )
-            sync_batch_norm_out, _, _, _, _, _ = _legacy_C_ops.sync_batch_norm(
-                x,
-                self.weight,
-                self.bias,
-                self._mean,
-                self._variance,
-                mean_out,
-                variance_out,
-                *attrs,
-            )
-            return sync_batch_norm_out
-
         check_variable_and_dtype(
             x,
             'input',
@@ -1885,7 +1853,7 @@ class SpectralNorm(Layer):
 
     def forward(self, x):
         weight = x
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _C_ops.spectral_norm(
                 weight,
                 self.weight_u,
diff --git a/python/paddle/nn/layer/rnn.py b/python/paddle/nn/layer/rnn.py
index 2a0c9157a7a..8990cccce36 100644
--- a/python/paddle/nn/layer/rnn.py
+++ b/python/paddle/nn/layer/rnn.py
@@ -24,7 +24,6 @@ from paddle.common_ops_import import Variable
 from paddle.fluid.data_feeder import check_type, check_variable_and_dtype
 from paddle.fluid.dygraph.base import NON_PERSISTABLE_VAR_NAME_SUFFIX
 from paddle.fluid.framework import (
-    _non_static_mode,
     default_startup_program,
     in_dygraph_mode,
     program_guard,
@@ -102,7 +101,7 @@ def rnn(
 
     """
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         return _rnn_dynamic_graph(
             cell,
             inputs,
diff --git a/python/paddle/nn/quant/format.py b/python/paddle/nn/quant/format.py
index edd72ad8ce3..245c65510ab 100644
--- a/python/paddle/nn/quant/format.py
+++ b/python/paddle/nn/quant/format.py
@@ -17,7 +17,7 @@ from typing import List, Tuple
 
 import paddle
 from paddle import _legacy_C_ops as _C_ops
-from paddle.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 from paddle.nn import Layer
 
 
@@ -57,7 +57,7 @@ class LinearQuanter(Layer):
         self._bit_length = bit_length
 
     def forward(self, input):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _C_ops.quantize_linear(
                 input,
                 self._scales,
@@ -108,7 +108,7 @@ class LinearDequanter(Layer):
         self._bit_length = bit_length
 
     def forward(self, input):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _C_ops.dequantize_linear(
                 input,
                 self._scales,
diff --git a/python/paddle/nn/utils/weight_norm_hook.py b/python/paddle/nn/utils/weight_norm_hook.py
index 36e62128c08..0c0851cdf31 100644
--- a/python/paddle/nn/utils/weight_norm_hook.py
+++ b/python/paddle/nn/utils/weight_norm_hook.py
@@ -16,7 +16,7 @@ from paddle import _C_ops
 
 from ...fluid.data_feeder import check_variable_and_dtype
 from ...fluid.layer_helper import LayerHelper
-from ...framework import in_dygraph_mode
+from ...framework import in_dynamic_mode
 
 __all__ = []
 
@@ -25,7 +25,7 @@ def l2_norm(x, axis, epsilon=1e-12, name=None):
     if len(x.shape) == 1:
         axis = 0
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out, norm = _C_ops.norm(x, 1 if axis is None else axis, epsilon, False)
         return paddle.squeeze(norm, axis=[axis])
 
diff --git a/python/paddle/optimizer/adadelta.py b/python/paddle/optimizer/adadelta.py
index c760c535da0..0edebb17625 100644
--- a/python/paddle/optimizer/adadelta.py
+++ b/python/paddle/optimizer/adadelta.py
@@ -18,7 +18,7 @@ from paddle import _C_ops
 
 from ..fluid import framework
 from ..fluid.dygraph import no_grad
-from ..framework import in_dygraph_mode
+from ..framework import in_dynamic_mode
 from .optimizer import Optimizer
 
 __all__ = []
@@ -190,7 +190,7 @@ class Adadelta(Optimizer):
             else None
         )
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             with no_grad():
                 _C_ops.adadelta_(
                     param_and_grad[0],
diff --git a/python/paddle/optimizer/adamw.py b/python/paddle/optimizer/adamw.py
index 3af9457142c..284beec9ba5 100644
--- a/python/paddle/optimizer/adamw.py
+++ b/python/paddle/optimizer/adamw.py
@@ -205,7 +205,7 @@ class AdamW(Optimizer):
             self._parameter_list = None
 
         self._name = name
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             if self._parameter_list is None:
                 raise AttributeError(
                     "parameters argument given to the Optimizer should not be None in dygraph mode."
diff --git a/python/paddle/optimizer/momentum.py b/python/paddle/optimizer/momentum.py
index f46cde0b099..e01466bc0f6 100644
--- a/python/paddle/optimizer/momentum.py
+++ b/python/paddle/optimizer/momentum.py
@@ -16,7 +16,7 @@ import warnings
 
 import paddle
 from paddle import _C_ops
-from paddle.fluid.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 from paddle.regularizer import L2Decay
 
 from ..fluid import core, framework
@@ -200,7 +200,7 @@ class Momentum(Optimizer):
 
     def _create_accumulators(self, block, parameters):
         '''
-        if framework._non_static_mode():
+        if framework.in_dynamic_mode():
             return
         '''
         assert isinstance(block, framework.Block)
@@ -275,7 +275,7 @@ class Momentum(Optimizer):
             else None
         )
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if isinstance(param_and_grad, dict):
                 self._update_regularization(param_and_grad['weight_decay'])
             return _C_ops.momentum_(
@@ -471,7 +471,7 @@ class Momentum(Optimizer):
                     else None
                 )
 
-                if in_dygraph_mode():
+                if in_dynamic_mode():
                     found_inf = self._get_auxiliary_var('found_inf')
                     if found_inf:
                         if isinstance(found_inf, core.eager.Tensor):
diff --git a/python/paddle/optimizer/optimizer.py b/python/paddle/optimizer/optimizer.py
index df20134e923..5792497a2a3 100644
--- a/python/paddle/optimizer/optimizer.py
+++ b/python/paddle/optimizer/optimizer.py
@@ -194,7 +194,7 @@ class Optimizer:
             self._parameter_list = None
 
         self._name = name
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             if self._parameter_list is None:
                 raise AttributeError(
                     "parameters argument given to the Optimizer should not be None in dygraph mode."
@@ -743,7 +743,7 @@ class Optimizer:
             name in self._accumulators
             and param.name in self._accumulators[name]
         ):
-            if framework._non_static_mode():
+            if framework.in_dygraph_mode():
                 return self._accumulators[name][param.name]
             raise Exception(
                 "Accumulator {} already exists for parameter {}".format(
@@ -790,7 +790,7 @@ class Optimizer:
                     ),
                 )
 
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             if len(self._accumulators_holder) > 0:
                 assert (
                     var_name in self._accumulators_holder
@@ -949,7 +949,7 @@ class Optimizer:
                         ],
                         param_group_idx,
                     )
-            if framework._non_static_mode():
+            if framework.in_dygraph_mode():
                 self._append_optimize_multi_tensor_op(
                     target_block,
                     parameters_and_grads,
@@ -980,7 +980,7 @@ class Optimizer:
                             param_group_idx=param_group_idx,
                         )
         else:
-            if not framework._non_static_mode():
+            if not framework.in_dygraph_mode():
                 params_grads_device_map = (
                     parameters_and_grads['params']
                     if isinstance(parameters_and_grads, dict)
@@ -1010,7 +1010,7 @@ class Optimizer:
                 with paddle.fluid.framework.dygraph_guard_if_declarative():
                     self._create_accumulators(target_block, params_acc_dict)
 
-            if framework._non_static_mode():
+            if framework.in_dygraph_mode():
                 found_inf = self._get_auxiliary_var('found_inf')
                 if found_inf:
                     if isinstance(found_inf, core.eager.Tensor):
@@ -1114,7 +1114,7 @@ class Optimizer:
                 adam.clear_grad()
         """
         act_no_grad_set = None
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             pass
         else:
             act_no_grad_set = self._get_no_grad_set(loss, no_grad_set)
@@ -1219,7 +1219,7 @@ class Optimizer:
         Returns:
             list: A list of operators appended to the current program.
         """
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             with program_guard(
                 framework.default_main_program(),
                 framework.default_startup_program(),
@@ -1320,7 +1320,7 @@ class Optimizer:
             Exception: Unknown regularization type
         """
         params_and_grads = []
-        if framework._non_static_mode():
+        if framework.in_dygraph_mode():
             for param, grad in parameters_and_grads:
                 new_grad = self._create_regularization_of_grad(
                     param, grad, regularization
diff --git a/python/paddle/quantization/quanters/abs_max.py b/python/paddle/quantization/quanters/abs_max.py
index ce1e0233e2f..23e50d4f775 100644
--- a/python/paddle/quantization/quanters/abs_max.py
+++ b/python/paddle/quantization/quanters/abs_max.py
@@ -212,7 +212,7 @@ class FakeQuanterWithAbsMaxObserverLayer(BaseQuanter):
         return quant_out
 
     def forward(self, input):
-        if paddle.framework.in_dynamic_mode():
+        if paddle.in_dynamic_mode():
             return self.dynamic_forward(input)
         else:
             return self.static_forward(input)
diff --git a/python/paddle/regularizer.py b/python/paddle/regularizer.py
index 8c245f99b8b..e10bde313fd 100644
--- a/python/paddle/regularizer.py
+++ b/python/paddle/regularizer.py
@@ -13,7 +13,7 @@
 # limitations under the License.
 
 
-from paddle import _C_ops, _legacy_C_ops
+from paddle import _C_ops
 from paddle.fluid import framework
 from paddle.fluid.framework import in_dygraph_mode
 
@@ -122,9 +122,9 @@ class L1Decay(WeightDecayRegularizer):
         assert isinstance(param, framework.Variable)
         assert isinstance(block, framework.Block)
 
-        if framework._non_static_mode():
-            sign = block.create_var(dtype=param.dtype, shape=param.shape)
-            decay = block.create_var(dtype=param.dtype, shape=param.shape)
+        if in_dygraph_mode():
+            sign = _C_ops.sign(param)
+            return _C_ops.scale(sign, self._coeff, 0.0, True)
         else:
             sign = block.create_var(
                 dtype=param.dtype, shape=param.shape, lod_level=param.lod_level
@@ -132,22 +132,19 @@ class L1Decay(WeightDecayRegularizer):
             decay = block.create_var(
                 dtype=param.dtype, shape=param.shape, lod_level=param.lod_level
             )
-        if in_dygraph_mode():
-            sign = _C_ops.sign(param)
-            return _C_ops.scale(sign, self._coeff, 0.0, True)
-
-        # Append sign op
-        block.append_op(type='sign', inputs={"X": param}, outputs={"Out": sign})
-
-        # Append scale op to the output of sign op
-        block.append_op(
-            type='scale',
-            inputs={"X": sign},
-            outputs={"Out": decay},
-            attrs={"scale": self._coeff},
-        )
+            # Append sign op
+            block.append_op(
+                type='sign', inputs={"X": param}, outputs={"Out": sign}
+            )
 
-        return decay
+            # Append scale op to the output of sign op
+            block.append_op(
+                type='scale',
+                inputs={"X": sign},
+                outputs={"Out": decay},
+                attrs={"scale": self._coeff},
+            )
+            return decay
 
     def __str__(self):
         return "L1Decay, coeff=%f" % self._coeff
@@ -231,11 +228,8 @@ class L2Decay(WeightDecayRegularizer):
         assert isinstance(param, framework.Variable)
         assert isinstance(block, framework.Block)
 
-        if framework._non_static_mode():
-            if framework.in_dygraph_mode():
-                return _C_ops.scale(param, self._coeff, 0.0, True)
-            else:
-                return _legacy_C_ops.scale(param, "scale", self._coeff)
+        if in_dygraph_mode():
+            return _C_ops.scale(param, self._coeff, 0.0, True)
         else:
             decay = block.create_var(
                 dtype=param.dtype, shape=param.shape, lod_level=param.lod_level
diff --git a/python/paddle/signal.py b/python/paddle/signal.py
index e1580b00075..f61df7a2b07 100644
--- a/python/paddle/signal.py
+++ b/python/paddle/signal.py
@@ -13,12 +13,11 @@
 # limitations under the License.
 
 import paddle
-from paddle import _C_ops, _legacy_C_ops
-from paddle.fluid.framework import in_dygraph_mode
+from paddle import _C_ops
+from paddle.framework import in_dynamic_mode
 
 from .fft import fft_c2c, fft_c2r, fft_r2c
 from .fluid.data_feeder import check_variable_and_dtype
-from .fluid.framework import _non_static_mode
 from .fluid.layer_helper import LayerHelper
 from .tensor.attribute import is_complex
 
@@ -118,14 +117,12 @@ def frame(x, frame_length, hop_length, axis=-1, name=None):
             f'Unexpected hop_length: {hop_length}. It should be an positive integer.'
         )
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if frame_length > x.shape[axis]:
             raise ValueError(
                 f'Attribute frame_length should be less equal than sequence length, '
                 f'but got ({frame_length}) > ({x.shape[axis]}).'
             )
-
-    if in_dygraph_mode():
         return _C_ops.frame(x, frame_length, hop_length, axis)
     else:
         op_type = 'frame'
@@ -211,12 +208,8 @@ def overlap_add(x, hop_length, axis=-1, name=None):
 
     op_type = 'overlap_add'
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.overlap_add(x, hop_length, axis)
-    elif paddle.in_dynamic_mode():
-        attrs = ('hop_length', hop_length, 'axis', axis)
-        op = getattr(_legacy_C_ops, op_type)
-        out = op(x, *attrs)
     else:
         check_variable_and_dtype(
             x,
@@ -328,7 +321,7 @@ def stft(
     if win_length is None:
         win_length = n_fft
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         assert (
             0 < n_fft <= x.shape[-1]
         ), f'n_fft should be in (0, seq_length({x.shape[-1]})], but got {n_fft}.'
@@ -522,7 +515,7 @@ def istft(
     n_frames = x.shape[-1]
     fft_size = x.shape[-2]
 
-    if _non_static_mode():
+    if in_dynamic_mode():
         if onesided:
             assert (
                 fft_size == n_fft // 2 + 1
@@ -610,7 +603,7 @@ def istft(
         window_envelop = window_envelop[start : start + length]
 
     # Check whether the Nonzero Overlap Add (NOLA) constraint is met.
-    if _non_static_mode() and window_envelop.abs().min().item() < 1e-11:
+    if in_dynamic_mode() and window_envelop.abs().min().item() < 1e-11:
         raise ValueError(
             'Abort istft because Nonzero Overlap Add (NOLA) condition failed. For more information about NOLA constraint please see `scipy.signal.check_NOLA`(https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.check_NOLA.html).'
         )
diff --git a/python/paddle/sparse/unary.py b/python/paddle/sparse/unary.py
index 45398022589..ef7daa5bbb6 100644
--- a/python/paddle/sparse/unary.py
+++ b/python/paddle/sparse/unary.py
@@ -22,7 +22,7 @@ from paddle.fluid.framework import (
     core,
     dygraph_only,
 )
-from paddle.framework import LayerHelper, in_dygraph_mode
+from paddle.framework import LayerHelper
 
 __all__ = []
 
@@ -198,7 +198,7 @@ def sum(x, axis=None, dtype=None, keepdim=False, name=None):
         dtype_flag = True
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sparse_sum(x, axis, dtype, keepdim)
     else:
         if axis is None:
diff --git a/python/paddle/static/nn/common.py b/python/paddle/static/nn/common.py
index cd0b96c04e8..6aeef4934e5 100644
--- a/python/paddle/static/nn/common.py
+++ b/python/paddle/static/nn/common.py
@@ -26,7 +26,7 @@ from paddle.common_ops_import import (
 )
 from paddle.fluid import core
 from paddle.fluid.data_feeder import check_dtype
-from paddle.fluid.framework import Variable, _non_static_mode, static_only
+from paddle.fluid.framework import Variable, in_dygraph_mode, static_only
 from paddle.fluid.layers.layer_function_generator import templatedoc
 from paddle.fluid.param_attr import ParamAttr
 from paddle.nn.initializer import Constant, Normal
@@ -1656,7 +1656,7 @@ def conv2d_transpose(
     if filter_size is None:
         if output_size is []:
             raise ValueError("output_size must be set when filter_size is None")
-        if not _non_static_mode():
+        if not in_dygraph_mode():
             if isinstance(output_size, Variable) or paddle.utils._contain_var(
                 output_size
             ):
@@ -2808,7 +2808,7 @@ def batch_norm(
     # variance and variance_out share the same memory
     variance_out = variance
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         inputs_has_MomemtumTensor = False
         attrs_has_momentum = False
         tmp_tensor_type = core.eager.Tensor
@@ -3478,7 +3478,7 @@ def spectral_norm(weight, dim=0, power_iters=1, eps=1e-12, name=None):
     )
     v.stop_gradient = True
 
-    if paddle.framework.in_dygraph_mode():
+    if in_dygraph_mode():
         return paddle._C_ops.spectral_norm(weight, u, v, dim, power_iters, eps)
 
     inputs = {'Weight': weight}
@@ -3583,7 +3583,7 @@ def layer_norm(
             print(output.shape)  # [8, 32, 32]
     """
     assert (
-        _non_static_mode() is not True
+        in_dygraph_mode() is not True
     ), "please use LayerNorm instead of layer_norm in dygraph mode!"
     helper = LayerHelper('layer_norm', **locals())
     check_variable_and_dtype(
diff --git a/python/paddle/static/nn/control_flow.py b/python/paddle/static/nn/control_flow.py
index 79cc848804b..093d63c96e7 100644
--- a/python/paddle/static/nn/control_flow.py
+++ b/python/paddle/static/nn/control_flow.py
@@ -18,10 +18,10 @@ from functools import partial, reduce
 import paddle
 from paddle.common_ops_import import (
     LayerHelper,
-    _non_static_mode,
     check_type,
     check_variable_and_dtype,
     convert_dtype,
+    in_dygraph_mode,
 )
 from paddle.fluid import core
 from paddle.fluid.framework import Operator, Program, Variable, static_only
@@ -469,7 +469,7 @@ def while_loop(cond, body, loop_vars, is_test=False, name=None):
             f"but given shape as {list(pre_cond.shape)}."
         )
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         now_cond = pre_cond.item()
         while now_cond:
             output_vars = body(*loop_vars)
@@ -969,7 +969,7 @@ def cond(pred, true_fn=None, false_fn=None, name=None, return_names=None):
             #           [ True  True  True]]
 
     """
-    if _non_static_mode():
+    if in_dygraph_mode():
         assert isinstance(pred, Variable), "The pred in cond must be Variable"
         assert pred.size == 1, "condition input's numel should be 1"
         pred = pred.item()
diff --git a/python/paddle/static/nn/metric.py b/python/paddle/static/nn/metric.py
index 3c1e0bf3852..e63a5236726 100644
--- a/python/paddle/static/nn/metric.py
+++ b/python/paddle/static/nn/metric.py
@@ -19,7 +19,7 @@ import numpy as np
 import paddle
 from paddle import _legacy_C_ops
 from paddle.fluid.data_feeder import check_variable_and_dtype
-from paddle.fluid.framework import Variable, _create_tensor, _non_static_mode
+from paddle.fluid.framework import Variable, _create_tensor, in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
 from paddle.nn.initializer import ConstantInitializer
 
@@ -72,7 +72,7 @@ def accuracy(input, label, k=1, correct=None, total=None):
             # [array(0.33333334, dtype=float32)]
 
     """
-    if _non_static_mode():
+    if in_dygraph_mode():
         if correct is None:
             correct = _create_tensor(dtype="int32")
         if total is None:
diff --git a/python/paddle/static/nn/sequence_lod.py b/python/paddle/static/nn/sequence_lod.py
index 882d4a05acd..f9b9ea355e2 100644
--- a/python/paddle/static/nn/sequence_lod.py
+++ b/python/paddle/static/nn/sequence_lod.py
@@ -15,7 +15,7 @@
 import paddle
 from paddle.fluid.core import VarDesc
 from paddle.fluid.data_feeder import check_type, check_variable_and_dtype
-from paddle.fluid.framework import Variable, _non_static_mode
+from paddle.fluid.framework import Variable, in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
 from paddle.fluid.layers.layer_function_generator import templatedoc
 
@@ -131,7 +131,7 @@ def sequence_conv(
     """
 
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     check_variable_and_dtype(
         input, 'input', ['float32', 'float64'], 'sequence_conv'
@@ -233,7 +233,7 @@ def sequence_softmax(input, use_cudnn=False, name=None):
              x_sequence_softmax_2 = paddle.static.nn.sequence_softmax(input=y)
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_softmax', **locals())
     check_variable_and_dtype(
@@ -338,7 +338,7 @@ def sequence_pool(input, pool_type, is_test=False, pad_value=0.0):
             first_x = paddle.static.nn.sequence_pool(input=x, pool_type='first')
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     check_variable_and_dtype(
         input, 'input', ['float32', 'float64'], 'sequence_pool'
@@ -416,7 +416,7 @@ def sequence_concat(input, name=None):
             out = paddle.static.nn.sequence_concat(input=[x, y])
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_concat', **locals())
 
@@ -617,7 +617,7 @@ def sequence_slice(input, offset, length, name=None):
                                                    length=length)
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper("sequence_slice", **locals())
 
@@ -771,7 +771,7 @@ def sequence_expand(x, y, ref_level=-1, name=None):
             #    data: [1 2 1 2 3 4 3 4]
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     check_variable_and_dtype(
         x, 'x', ['float32', 'float64', 'int32', 'int64'], 'sequence_expand'
@@ -893,7 +893,7 @@ def sequence_expand_as(x, y, name=None):
             #    data: [1 1 1 2 2 2 3 4]
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     check_variable_and_dtype(
         x, 'x', ['float32', 'float64', 'int32', 'int64'], 'sequence_expand_as'
@@ -996,7 +996,7 @@ def sequence_pad(x, pad_value, maxlen=None, name=None):
     """
 
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_pad', **locals())
     check_variable_and_dtype(
@@ -1085,7 +1085,7 @@ def sequence_unpad(x, length, name=None):
     """
 
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_unpad', **locals())
     check_variable_and_dtype(
@@ -1163,7 +1163,7 @@ def sequence_reshape(input, new_dim):
             x_reshaped = paddle.static.nn.sequence_reshape(input=x, new_dim=4)
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_reshape', **locals())
     check_variable_and_dtype(
@@ -1248,7 +1248,7 @@ def sequence_scatter(input, index, updates, name=None):
 
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper('sequence_scatter', **locals())
 
@@ -1330,7 +1330,7 @@ def sequence_enumerate(input, win_size, pad_value=0, name=None):
             out = paddle.static.nn.sequence_enumerate(input=x, win_size=3, pad_value=0)
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     check_variable_and_dtype(
         input, 'input', ['int32', 'int64'], 'sequence_enumerate'
@@ -1459,7 +1459,7 @@ def sequence_reverse(x, name=None):
             x_reversed = paddle.static.nn.sequence_reverse(x)
     """
     assert (
-        not _non_static_mode()
+        not in_dygraph_mode()
     ), "sequence layer is not supported in dygraph mode yet."
     helper = LayerHelper("sequence_reverse", **locals())
     check_variable_and_dtype(
diff --git a/python/paddle/tensor/array.py b/python/paddle/tensor/array.py
index eefdf27c1a1..c4503bb1aec 100644
--- a/python/paddle/tensor/array.py
+++ b/python/paddle/tensor/array.py
@@ -16,7 +16,7 @@
 
 from ..common_ops_import import Variable
 from ..fluid.data_feeder import check_type, check_variable_and_dtype
-from ..framework import LayerHelper, core, in_dygraph_mode
+from ..framework import LayerHelper, core, in_dynamic_mode
 
 __all__ = []
 
@@ -45,7 +45,7 @@ def array_length(array):
             arr_len = paddle.tensor.array_length(arr)
             print(arr_len)  # 1
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         assert isinstance(
             array, list
         ), "The 'array' in array_write must be a list in dygraph mode"
@@ -109,7 +109,7 @@ def array_read(array, i):
             item = paddle.tensor.array_read(arr, i)
             print(item)     # [[5., 5., 5.]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         assert isinstance(
             array, list
         ), "The 'array' in array_read must be list in dygraph mode"
@@ -169,7 +169,7 @@ def array_write(x, i, array=None):
             item = paddle.tensor.array_read(arr, i)
             print(item)     # [[5., 5., 5.]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         assert isinstance(
             x, Variable
         ), "The input data 'x' in array_write must be Variable in dygraph mode"
@@ -267,7 +267,7 @@ def create_array(dtype, initialized_list=None):
                 )
             )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return array
     else:
         helper = LayerHelper("array", **locals())
diff --git a/python/paddle/tensor/creation.py b/python/paddle/tensor/creation.py
index bf6c603a7ec..bc9790e5cb1 100644
--- a/python/paddle/tensor/creation.py
+++ b/python/paddle/tensor/creation.py
@@ -38,7 +38,7 @@ from ..framework import (
     _get_paddle_place,
     convert_np_dtype_to_dtype_,
     core,
-    in_dygraph_mode,
+    in_dynamic_mode,
 )
 
 __all__ = []
@@ -312,7 +312,7 @@ def linspace(start, stop, num, dtype=None, name=None):
     if not isinstance(num, Variable):
         with device_guard("cpu"):
             tensor_num = fill_constant([1], 'int32', num, force_cpu=True)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.linspace(
             tensor_start,
             tensor_stop,
@@ -445,7 +445,7 @@ def logspace(start, stop, num, base=10.0, dtype=None, name=None):
     if not isinstance(base, Variable):
         with device_guard("cpu"):
             tensor_base = fill_constant([1], dtype, base)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logspace(
             tensor_start,
             tensor_stop,
@@ -569,12 +569,12 @@ def _to_tensor_non_static(data, dtype=None, place=None, stop_gradient=True):
                     "\n\tFaild to convert input data to a regular ndarray :\n\t - Usually "
                     "this means the input data contains nested lists with different lengths. "
                 )
-        elif isinstance(data, paddle.Tensor) and not in_dygraph_mode():
+        elif isinstance(data, paddle.Tensor) and not in_dynamic_mode():
             data = data._copy_to(place, False)
             data = _handle_tensor_dtype(data, dtype)
             data.stop_gradient = stop_gradient
             return data
-        elif isinstance(data, core.eager.Tensor) and in_dygraph_mode():
+        elif isinstance(data, core.eager.Tensor) and in_dynamic_mode():
             data = data._copy_to(place, False)
             data = _handle_tensor_dtype(data, dtype)
             data.stop_gradient = stop_gradient
@@ -583,7 +583,7 @@ def _to_tensor_non_static(data, dtype=None, place=None, stop_gradient=True):
             # should't expose it to users, just for internal use.
             # convert core.Tensor/core.LoDTensor to Tensor first
             # Currenly, there is no copy when places are same
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 data = core.eager.Tensor(data)
             else:
                 data = paddle.Tensor(data)
@@ -777,7 +777,7 @@ def to_tensor(data, dtype=None, place=None, stop_gradient=True):
     if place is None:
         place = _current_expected_place()
 
-    if paddle.fluid.framework._non_static_mode():
+    if in_dynamic_mode():
         return _to_tensor_non_static(data, dtype, place, stop_gradient)
 
     # call assign for static graph
@@ -821,7 +821,7 @@ def full_like(x, fill_value, dtype=None, name=None):
     else:
         if not isinstance(dtype, core.VarDesc.VarType):
             dtype = convert_np_dtype_to_dtype_(dtype)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.full_like(x, fill_value, dtype, x.place)
     else:
         helper = LayerHelper("full_like", **locals())
@@ -868,7 +868,7 @@ def full_like(x, fill_value, dtype=None, name=None):
 
 
 def fill_constant(shape, dtype, value, force_cpu=False, out=None, name=None):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         place = _current_expected_place()
         if force_cpu:
             place = core.CPUPlace()
@@ -1154,7 +1154,7 @@ def eye(num_rows, num_columns=None, dtype=None, name=None):
     else:
         num_columns = num_rows
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.eye(
             num_rows, num_columns, dtype, _current_expected_place()
         )
@@ -1302,7 +1302,7 @@ def arange(start=0, end=None, step=1, dtype=None, name=None):
         start = 0
 
     out_shape = None
-    if not in_dygraph_mode() and (
+    if not in_dynamic_mode() and (
         not isinstance(start, Variable)
         and not isinstance(end, Variable)
         and not isinstance(step, Variable)
@@ -1330,7 +1330,7 @@ def arange(start=0, end=None, step=1, dtype=None, name=None):
     elif step.dtype != dtype:
         step = paddle.cast(step, dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.arange(start, end, step, dtype, _current_expected_place())
     else:
         check_dtype(
@@ -1445,7 +1445,7 @@ def tril(x, diagonal=0, name=None):
             #         [5 , 0 , 0 , 0 ],
             #         [9 , 10, 0 , 0 ]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.tril(x, diagonal)
     else:
         return _tril_triu_op(LayerHelper('tril', **locals()))
@@ -1507,7 +1507,7 @@ def triu(x, diagonal=0, name=None):
             #         [0 , 10, 11, 12]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.triu(x, diagonal)
     else:
         return _tril_triu_op(LayerHelper('triu', **locals()))
@@ -1548,7 +1548,7 @@ def meshgrid(*args, **kwargs):
 
     if len(args) == 1 and isinstance(args[0], (list, tuple)):
         args = args[0]
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.meshgrid(list(args))
     else:
         name = kwargs.get("name", None)
@@ -1664,7 +1664,7 @@ def diagflat(x, offset=0, name=None):
             #         [0, 0, 3, 0, 0],
             #         [0, 0, 0, 4, 0]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if len(x.shape) <= 1:
             return _C_ops.diag(x, offset, 0)
         else:
@@ -1787,7 +1787,7 @@ def diag(x, offset=0, padding_value=0, name=None):
             # Tensor(shape=[1], dtype=int64, place=Place(cpu), stop_gradient=True,
             #        [4])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.diag(x, offset, padding_value)
     else:
         check_type(x, 'x', (Variable), 'diag_v2')
@@ -1869,7 +1869,7 @@ def empty(shape, dtype=None, name=None):
 
     dtype = convert_dtype(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         shape = paddle.utils.convert_shape_to_list(shape)
         out = _C_ops.empty(
             shape, convert_np_dtype_to_dtype_(dtype), _current_expected_place()
@@ -1950,7 +1950,7 @@ def empty_like(x, dtype=None, name=None):
         dtype = x.dtype
     dtype = convert_dtype(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.empty(
             x.shape,
             convert_np_dtype_to_dtype_(dtype),
@@ -2056,11 +2056,11 @@ def assign(x, output=None):
         input = np.array(input)
     # NOTE(Aurelius84): Why we judge core.Tensor?
     # In case of @to_static, a Tensor can be as input of `assign`,
-    # but _non_static_mode()==False under @to_static, which means
+    # but in_dynamic_mode()==False under @to_static, which means
     # isinstance(Tensor, Variable) == False. It will cause return None
     # after this api.
     if isinstance(input, (Variable, core.eager.Tensor)):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if output is None:
                 output = _C_ops.assign(input)
             else:
@@ -2154,7 +2154,7 @@ def assign(x, output=None):
                 "The size of input is too big. Please consider "
                 "saving it to file and 'load_op' to load it"
             )
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if output is None:
                 output = zeros(list(input.shape), dtype)
             _C_ops.assign_value_(
@@ -2313,7 +2313,7 @@ def complex(real, imag, name=None):
             #        [[0j    , 1j    , 2j    ],
             #         [(1+0j), (1+1j), (1+2j)]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.complex(real, imag)
     else:
         check_variable_and_dtype(
@@ -2385,7 +2385,7 @@ def tril_indices(row, col, offset=0, dtype='int64'):
     if not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if col is None:
             col = row
         out = _C_ops.tril_indices(
@@ -2464,7 +2464,7 @@ def triu_indices(row, col=None, offset=0, dtype='int64'):
     if not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if col is None:
             col = row
         out = _C_ops.triu_indices(
diff --git a/python/paddle/tensor/layer_function_generator.py b/python/paddle/tensor/layer_function_generator.py
index a2ab51d286d..884cecac0f5 100644
--- a/python/paddle/tensor/layer_function_generator.py
+++ b/python/paddle/tensor/layer_function_generator.py
@@ -27,7 +27,7 @@ from ..framework import (
     OpProtoHolder,
     convert_np_dtype_to_dtype_,
     core,
-    in_dygraph_mode,
+    in_dynamic_mode,
 )
 
 __all__ = []
@@ -267,7 +267,7 @@ def generate_activation_fn(op_type):
     op_proto = OpProtoHolder.instance().get_op_proto(op_type)
 
     def func(x, name=None):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if hasattr(_C_ops, op_type):
                 op = getattr(_C_ops, op_type)
                 return op(x)
@@ -346,7 +346,7 @@ def generate_inplace_fn(inplace_op_type):
 
     def func(x, name=None):
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if hasattr(_C_ops, inplace_op_type):
                 op = getattr(_C_ops, inplace_op_type)
                 return op(x)
diff --git a/python/paddle/tensor/linalg.py b/python/paddle/tensor/linalg.py
index 3b3c7565ab3..db0805669a1 100644
--- a/python/paddle/tensor/linalg.py
+++ b/python/paddle/tensor/linalg.py
@@ -24,7 +24,7 @@ from ..fluid.data_feeder import (
     check_type,
     check_variable_and_dtype,
 )
-from ..framework import LayerHelper, in_dygraph_mode
+from ..framework import LayerHelper, in_dynamic_mode
 from .creation import full
 from .manipulation import cast
 
@@ -84,7 +84,7 @@ def transpose(x, perm, name=None):
             # [3L, 2L, 4L]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.transpose(x, perm)
     else:
         check_variable_and_dtype(
@@ -229,7 +229,7 @@ def matmul(x, y, transpose_x=False, transpose_y=False, name=None):
             # (10, 3, 5, 5)
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.matmul(x, y, transpose_x, transpose_y)
     else:
         attrs = {
@@ -367,7 +367,7 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
                 "The dim of frobenius norm op should be None or two elements list!"
             )
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if dim is None:
                 return _C_ops.frobenius_norm(input, [], keepdim, True)
             return _C_ops.frobenius_norm(input, dim, keepdim, False)
@@ -403,7 +403,7 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
           axis (int, optional): None for last dimension.
           keepdim (bool, optional): Whether keep the dimensions as the `input`, Default False.
         """
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if axis is None:
                 axis = -1
             return _C_ops.p_norm(input, porder, axis, 1e-12, keepdim, asvector)
@@ -442,7 +442,7 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
     def inf_norm(
         input, porder=None, axis=axis, keepdim=False, asvector=False, name=None
     ):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             out = _C_ops.abs(input)
             if porder == np.float64('inf'):
                 return _C_ops.max(out, axis, keepdim)
@@ -486,7 +486,7 @@ def norm(x, p='fro', axis=None, keepdim=False, name=None):
         NOTE:
             This function actually treats the matrix as flattened vector to calculate vector norm instead of matrix norm.
         """
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             abs_out = _C_ops.abs(input)
             pow_out = _C_ops.pow(abs_out, porder)
             sum_out = _C_ops.sum(pow_out, axis, None, keepdim)
@@ -701,7 +701,7 @@ def dist(x, y, p=2, name=None):
             out = paddle.dist(x, y, float("-inf"))
             print(out) # out = 0.
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.dist(x, y, p)
 
     check_variable_and_dtype(x, 'dtype', ['float32', 'float64'], 'dist')
@@ -825,7 +825,7 @@ def cond(x, p=None, name=None):
             Calculate the matrix norm of a square matrix or batches of square matrices,
             when porder is in (1, -1, inf, -inf)
         """
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             abs_out = _C_ops.abs(input)
             sum_out = _C_ops.sum(abs_out, axis, None, False)
 
@@ -888,7 +888,7 @@ def cond(x, p=None, name=None):
         NOTE:
             Calculate the frobenius norm of a square matrix or batches of square matrices.
         """
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             pow_out = _C_ops.pow(input, porder)
             sum_out_1 = _C_ops.sum(pow_out, axis, None, False)
             sum_out_2 = _C_ops.sum(sum_out_1, axis, None, False)
@@ -950,7 +950,7 @@ def cond(x, p=None, name=None):
         """
         u, s, vh = svd(input, full_matrices=False)
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             if porder == "nuc":
                 return _C_ops.sum(s, axis, None, False)
             max_out = _C_ops.max(s, axis, False)
@@ -1021,7 +1021,7 @@ def cond(x, p=None, name=None):
                 return out
 
     def empty_tensor(input, shape):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return input.reshape(shape)
         raise ValueError(
             "only support x is nonempty tensor in static graph mode"
@@ -1104,7 +1104,7 @@ def dot(x, y, name=None):
         print(z)  # [32, 64]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.dot(x, y)
     else:
         op_type = 'dot'
@@ -1306,7 +1306,7 @@ def t(input, name=None):
             "length of Input(input) is %s. Perhaps you can use paddle."
             "tensor.transpose() instead." % len(input.shape)
         )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if len(input.shape) <= 1:
             return input
         # 2-D tensor
@@ -1374,7 +1374,7 @@ def cross(x, y, axis=9, name=None):
             #  [0. 0. 0.]
             #  [0. 0. 0.]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         axis = K_DEFAULT_DIM if axis is None else axis
         return _C_ops.cross(x, y, axis)
     else:
@@ -1440,7 +1440,7 @@ def cholesky(x, upper=False, name=None):
             out = paddle.linalg.cholesky(x, upper=False)
             print(out)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.cholesky(x, upper)
     else:
         check_variable_and_dtype(x, 'dtype', ['float32', 'float64'], 'cholesky')
@@ -1495,7 +1495,7 @@ def matrix_rank(x, tol=None, hermitian=False, name=None):
             #      [1, 1, 1, 1]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(tol, Variable):
             if tol.dtype != x.dtype:
                 tol_tensor = cast(tol, x.dtype)
@@ -1580,7 +1580,7 @@ def bmm(x, y, name=None):
             #          [60., 60.]]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.bmm(x, y)
     else:
         x_shape = x.shape
@@ -1636,7 +1636,7 @@ def histogram(input, bins=100, min=0, max=0, name=None):
             result = paddle.histogram(inputs, bins=4, min=0, max=3)
             print(result) # [0, 2, 1, 0]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.histogram(input, bins, min, max)
     else:
         helper = LayerHelper('histogram', **locals())
@@ -1683,7 +1683,7 @@ def bincount(x, weights=None, minlength=0, name=None):
     if x.dtype not in [paddle.int32, paddle.int64]:
         raise TypeError("Elements in Input(x) should all be integers")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.bincount(x, weights, minlength)
     else:
         helper = LayerHelper('bincount', **locals())
@@ -1739,7 +1739,7 @@ def mv(x, vec, name=None):
             # Tensor(shape=[2], dtype=float64, place=Place(cpu), stop_gradient=True,
             #        [14., 10.])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.mv(x, vec)
     else:
 
@@ -1804,7 +1804,7 @@ def det(x, name=None):
 
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.det(x)
     else:
         check_dtype(x.dtype, 'Input', ['float16', 'float32', 'float64'], 'det')
@@ -1863,7 +1863,7 @@ def slogdet(x, name=None):
             # [-0.98610914, -0.43010661, -0.10872950]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.slogdet(x)
     else:
         check_dtype(x.dtype, 'Input', ['float32', 'float64'], 'slogdet')
@@ -1945,7 +1945,7 @@ def svd(x, full_matrices=False, name=None):
             #                  V * VH == I
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.svd(x, full_matrices)
     else:
         check_variable_and_dtype(x, 'dtype', ['float32', 'float64'], 'svd')
@@ -2019,7 +2019,7 @@ def matrix_power(x, n, name=None):
             #  [-7.66666667 ,  8.         , -1.83333333 ],
             #  [ 1.80555556 , -1.91666667 ,  0.44444444 ]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.matrix_power(x, n)
     else:
         check_variable_and_dtype(
@@ -2079,7 +2079,7 @@ def qr(x, mode="reduced", name=None):
 
             # one can verify : X = Q * R ;
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         q, r = _C_ops.qr(x, mode)
         if mode == "r":
             return r
@@ -2182,7 +2182,7 @@ def lu(x, pivot=True, get_infos=False, name=None):
             # one can verify : X = P @ L @ U ;
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         lu, p, info = _C_ops.lu(x, pivot)
     else:
         check_variable_and_dtype(x, 'dtype', ['float32', 'float64'], 'lu')
@@ -2277,7 +2277,7 @@ def lu_unpack(x, y, unpack_ludata=True, unpack_pivots=True, name=None):
             # one can verify : X = P @ L @ U ;
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         P, L, U = _C_ops.lu_unpack(x, y, unpack_ludata, unpack_pivots)
         return P, L, U
     else:
@@ -2348,7 +2348,7 @@ def eig(x, name=None):
             #         (-0.21026087843552282+0j)])
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.eig(x)
     else:
         check_variable_and_dtype(
@@ -2417,7 +2417,7 @@ def eigvals(x, name=None):
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.eigvals(x)
     else:
         check_variable_and_dtype(
@@ -2488,7 +2488,7 @@ def multi_dot(x, name=None):
         # [10, 7]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.multi_dot(x)
     else:
         check_type(x, 'x', (list, tuple), 'multi_dot')
@@ -2546,7 +2546,7 @@ def eigh(x, UPLO='L', name=None):
             #[ 0.3826834323650898j    , -0.9238795325112867j    ]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.eigh(x, UPLO)
     else:
 
@@ -2653,7 +2653,7 @@ def pinv(x, rcond=1e-15, hermitian=False, name=None):
             # one can verify : x * out * x = x ;
             # or              out * x * out = x ;
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if not hermitian:
             # combine svd and matmul op
             u, s, vt = _C_ops.svd(x, False)
@@ -2876,7 +2876,7 @@ def solve(x, y, name=None):
             print(out)
             # [2., 3.])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.solve(x, y)
     else:
         inputs = {"X": [x], "Y": [y]}
@@ -2945,7 +2945,7 @@ def triangular_solve(
             print(out)
             # [7, -2, -5]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.triangular_solve(x, y, upper, transpose, unitriangular)
     else:
         inputs = {"X": [x], "Y": [y]}
@@ -3004,7 +3004,7 @@ def cholesky_solve(x, y, upper=False, name=None):
             print(out)
             # [-2.5, -7, 9.5]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.cholesky_solve(x, y, upper)
     else:
         helper = LayerHelper("cholesky_solve", **locals())
@@ -3051,7 +3051,7 @@ def eigvalsh(x, UPLO='L', name=None):
             # Tensor(shape=[2], dtype=float32, place=Place(cpu), stop_gradient=True,
             #        [0.17157286, 5.82842731])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         values, _ = _C_ops.eigvalsh(x, UPLO, x.stop_gradient)
         return values
     else:
@@ -3202,7 +3202,7 @@ def lstsq(x, y, rcond=None, driver=None, name=None):
         elif x.dtype == paddle.float64:
             rcond = 1e-15 * max(x.shape[-2], x.shape[-1])
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         solution, residuals, rank, singular_values = _C_ops.lstsq(
             x, y, rcond, driver
         )
diff --git a/python/paddle/tensor/logic.py b/python/paddle/tensor/logic.py
index cf3c1f5284d..8848f80949d 100755
--- a/python/paddle/tensor/logic.py
+++ b/python/paddle/tensor/logic.py
@@ -25,13 +25,13 @@ Tensor = paddle.fluid.framework.core.eager.Tensor
 from paddle import _C_ops
 from paddle.tensor.creation import full
 
-from ..framework import LayerHelper, in_dygraph_mode
+from ..framework import LayerHelper, in_dynamic_mode
 
 __all__ = []
 
 
 def _logical_op(op_name, x, y, out=None, name=None, binary_op=True):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         op = getattr(_C_ops, op_name)
         if binary_op:
             return op(x, y)
@@ -131,7 +131,7 @@ def logical_and(x, y, out=None, name=None):
             res = paddle.logical_and(x, y)
             print(res) # [True False True False]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logical_and(x, y)
 
     return _logical_op(
@@ -176,7 +176,7 @@ def logical_or(x, y, out=None, name=None):
             #        [[True , True ],
             #         [True , False]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logical_or(x, y)
     return _logical_op(
         op_name="logical_or", x=x, y=y, name=name, out=out, binary_op=True
@@ -220,7 +220,7 @@ def logical_xor(x, y, out=None, name=None):
             #        [[False, True ],
             #         [True , False]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logical_xor(x, y)
 
     return _logical_op(
@@ -260,7 +260,7 @@ def logical_not(x, out=None, name=None):
             res = paddle.logical_not(x)
             print(res) # [False  True False  True]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logical_not(x)
     return _logical_op(
         op_name="logical_not", x=x, y=None, name=name, out=out, binary_op=False
@@ -292,7 +292,7 @@ def is_empty(x, name=None):
             #        False)
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.is_empty(x)
     else:
         check_variable_and_dtype(
@@ -338,7 +338,7 @@ def equal_all(x, y, name=None):
           result2 = paddle.equal_all(x, z)
           print(result2) # result2 = False
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.equal_all(x, y)
     else:
         helper = LayerHelper("equal_all", **locals())
@@ -400,7 +400,7 @@ def allclose(x, y, rtol=1e-05, atol=1e-08, equal_nan=False, name=None):
           # True
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.allclose(x, y, rtol, atol, equal_nan)
     else:
         check_variable_and_dtype(
@@ -464,7 +464,7 @@ def equal(x, y, name=None):
     if not isinstance(y, Variable):
         y = full(shape=[], dtype=x.dtype, fill_value=y)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.equal(x, y)
     else:
         check_variable_and_dtype(
@@ -533,7 +533,7 @@ def greater_equal(x, y, name=None):
             result1 = paddle.greater_equal(x, y)
             print(result1)  # result1 = [True False True]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.greater_equal(x, y)
     else:
         check_variable_and_dtype(
@@ -602,7 +602,7 @@ def greater_than(x, y, name=None):
             result1 = paddle.greater_than(x, y)
             print(result1)  # result1 = [False False True]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.greater_than(x, y)
     else:
         check_variable_and_dtype(
@@ -672,7 +672,7 @@ def less_equal(x, y, name=None):
             result1 = paddle.less_equal(x, y)
             print(result1)  # result1 = [True True False]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.less_equal(x, y)
     else:
         check_variable_and_dtype(
@@ -742,7 +742,7 @@ def less_than(x, y, name=None):
             result1 = paddle.less_than(x, y)
             print(result1)  # result1 = [False True False]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.less_than(x, y)
     else:
         check_variable_and_dtype(
@@ -812,7 +812,7 @@ def not_equal(x, y, name=None):
             result1 = paddle.not_equal(x, y)
             print(result1)  # result1 = [False True True]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.not_equal(x, y)
     else:
         check_variable_and_dtype(
@@ -880,14 +880,14 @@ def is_tensor(x):
             print(check)  #False
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return isinstance(x, (Tensor, paddle.fluid.core.eager.Tensor))
     else:
         return isinstance(x, Variable)
 
 
 def _bitwise_op(op_name, x, y, out=None, name=None, binary_op=True):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         op = getattr(_C_ops, op_name)
         if binary_op:
             return op(x, y)
@@ -959,7 +959,7 @@ def bitwise_and(x, y, out=None, name=None):
             res = paddle.bitwise_and(x, y)
             print(res)  # [0, 2, 1]
     """
-    if in_dygraph_mode() and out is None:
+    if in_dynamic_mode() and out is None:
         return _C_ops.bitwise_and(x, y)
     return _bitwise_op(
         op_name="bitwise_and", x=x, y=y, name=name, out=out, binary_op=True
@@ -996,7 +996,7 @@ def bitwise_or(x, y, out=None, name=None):
             res = paddle.bitwise_or(x, y)
             print(res)  # [-1, -1, -3]
     """
-    if in_dygraph_mode() and out is None:
+    if in_dynamic_mode() and out is None:
         return _C_ops.bitwise_or(x, y)
 
     return _bitwise_op(
@@ -1034,7 +1034,7 @@ def bitwise_xor(x, y, out=None, name=None):
             res = paddle.bitwise_xor(x, y)
             print(res) # [-1, -3, -4]
     """
-    if in_dygraph_mode() and out is None:
+    if in_dynamic_mode() and out is None:
         return _C_ops.bitwise_xor(x, y)
     return _bitwise_op(
         op_name="bitwise_xor", x=x, y=y, name=name, out=out, binary_op=True
@@ -1069,7 +1069,7 @@ def bitwise_not(x, out=None, name=None):
             res = paddle.bitwise_not(x)
             print(res) # [4, 0, -2]
     """
-    if in_dygraph_mode() and out is None:
+    if in_dynamic_mode() and out is None:
         return _C_ops.bitwise_not(x)
 
     return _bitwise_op(
@@ -1126,7 +1126,7 @@ def isclose(x, y, rtol=1e-05, atol=1e-08, equal_nan=False, name=None):
           # [True, True]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.isclose(x, y, rtol, atol, equal_nan)
     else:
         check_variable_and_dtype(
diff --git a/python/paddle/tensor/manipulation.py b/python/paddle/tensor/manipulation.py
index 1371db3ea93..3828b097f8d 100644
--- a/python/paddle/tensor/manipulation.py
+++ b/python/paddle/tensor/manipulation.py
@@ -33,7 +33,7 @@ from ..framework import (
     convert_np_dtype_to_dtype_,
     core,
     dygraph_only,
-    in_dygraph_mode,
+    in_dynamic_mode,
 )
 from .creation import _complex_to_real_dtype, _real_to_complex_dtype, zeros
 
@@ -120,7 +120,7 @@ def tensor_array_to_tensor(input, axis=1, use_stack=False, name=None):
             paddle.tensor.array.array_write(x1, i + 1, array)
             output, output_index = paddle.tensor.manipulation.tensor_array_to_tensor(input=array)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         assert isinstance(
             input, list
         ), "The 'input' in tensor_array_to_tensor must be list"
@@ -178,7 +178,7 @@ def cast(x, dtype):
             x = paddle.to_tensor([2, 3, 4], 'float64')
             y = paddle.cast(x, 'uint8')
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if not isinstance(dtype, core.VarDesc.VarType):
             dtype = convert_np_dtype_to_dtype_(dtype)
         return _C_ops.cast(x, dtype)
@@ -298,7 +298,7 @@ def slice(input, axes, starts, ends):
             sliced_2 = paddle.slice(input, axes=axes, starts=[minus_3, 0, 2], ends=ends)
             # sliced_2 is input[1:3, 0:2, 2:4].
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         attrs = ()
         starts_tensor = None
         ends_tensor = None
@@ -465,7 +465,7 @@ def transpose(x, perm, name=None):
             # [3L, 2L, 4L]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.transpose(x, perm)
     else:
         check_variable_and_dtype(
@@ -545,7 +545,7 @@ def unstack(x, axis=0, num=None):
         raise ValueError(
             '`axis` must be in the range [-{0}, {0})'.format(x.ndim)
         )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if num is None:
             num = x.shape[axis]
         if num == 0:
@@ -618,7 +618,7 @@ def shard_index(input, index_num, nshards, shard_id, ignore_value=-1):
             print(shard_label)
             # [[-1], [1]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.shard_index(
             input, index_num, nshards, shard_id, ignore_value
         )
@@ -751,7 +751,7 @@ def crop(x, shape=None, offsets=None, name=None):
     if shape is None:
         shape = x.shape
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.crop(x, shape, offsets)
 
     out = helper.create_variable_for_type_inference(x.dtype)
@@ -934,7 +934,7 @@ def fill_diagonal_(x, value, offset=0, wrap=False, name=None):
             x.fill_diagonal_(1.0)
             print(x.tolist())   #[[1.0, 2.0, 2.0], [2.0, 1.0, 2.0], [2.0, 2.0, 1.0], [2.0, 2.0, 2.0]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if len(x.shape) == 2:
             return _C_ops.fill_diagonal_(x, value, offset, wrap)
         return _C_ops.fill_diagonal_(x, value, offset, True)
@@ -1113,7 +1113,7 @@ def concat(x, axis=0, name=None):
             #  [14 15 16]]
     """
     input = x
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(axis, Variable):
             axis = axis.item(0)
         if not isinstance(input, Variable):
@@ -1226,7 +1226,7 @@ def broadcast_tensors(input, name=None):
     """
 
     num_inputs = len(input)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.broadcast_tensors(input)
     else:
         check_type(input, 'input', (list, tuple), 'broadcast_tensors')
@@ -1340,7 +1340,7 @@ def flip(x, axis, name=None):
     if isinstance(axis, int):
         axis = [axis]
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.flip(x, axis)
     else:
         helper = LayerHelper("flip", **locals())
@@ -1580,7 +1580,7 @@ def flatten(x, start_axis=0, stop_axis=-1, name=None):
         if start_axis > stop_axis:
             raise ValueError("The stop_axis should be larger than stat_axis")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.flatten(x, start_axis, stop_axis)
     else:
         check_variable_and_dtype(
@@ -1644,7 +1644,7 @@ def flatten_(x, start_axis=0, stop_axis=-1, name=None):
     if start_axis > stop_axis:
         raise ValueError("The stop_axis should be larger than stat_axis")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.flatten_(x, start_axis, stop_axis)
 
 
@@ -1709,7 +1709,7 @@ def roll(x, shifts, axis=None, name=None):
     else:
         axis = []
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.roll(x, shifts, axis)
     else:
         check_variable_and_dtype(
@@ -1838,7 +1838,7 @@ def stack(x, axis=0, name=None):
     """
     axis = 0 if axis is None else axis
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.stack(x, axis)
     else:
         if not isinstance(x, list) and not isinstance(x, tuple):
@@ -1952,7 +1952,7 @@ def split(x, num_or_sections, axis=0, name=None):
     """
     input = x
     dim = axis
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(dim, Variable):
             dim = dim.item(0)
         assert len(input.shape) + dim >= 0, "(rank(x) + axis) must >= 0"
@@ -2201,7 +2201,7 @@ def squeeze(x, axis=None, name=None):
 
     input = x
     axes = axis
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.squeeze(input, axes)
     else:
         helper = LayerHelper("squeeze", **locals())
@@ -2261,7 +2261,7 @@ def squeeze_(x, axis=None, name=None):
 
     input = x
     axes = axis
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.squeeze_(input, axes)
 
 
@@ -2342,7 +2342,7 @@ def unique_consecutive(
     else:
         axis = [axis]
     attr_dtype = convert_np_dtype_to_dtype_(dtype)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out, inverse, counts = _C_ops.unique_consecutive(
             x, return_inverse, return_counts, axis, attr_dtype
         )
@@ -2469,7 +2469,7 @@ def unique(
     else:
         axis = [axis]
     attr_dtype = convert_np_dtype_to_dtype_(dtype)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out, indices, inverse, counts = _C_ops.unique(
             x, return_index, return_inverse, return_counts, axis, attr_dtype
         )
@@ -2592,7 +2592,7 @@ def unsqueeze(x, axis, name=None):
     """
     input = x
     axes = axis
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(axes, int):
             axes = [axes]
         elif isinstance(axes, Variable):
@@ -2721,7 +2721,7 @@ def gather(x, index, axis=None, name=None):
     if axis is None:
         axis = 0
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.gather(x, index, axis)
     else:
         check_variable_and_dtype(
@@ -2805,7 +2805,7 @@ def unbind(input, axis=0):
             f'The axis must in range({-input.ndim}, {input.ndim}).'
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.unbind(input, axis)
     else:
         if isinstance(axis, np.generic):
@@ -2919,7 +2919,7 @@ def scatter(x, index, updates, overwrite=True, name=None):
             #  [2., 2.],
             #  [1., 1.]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.scatter(x, index, updates, overwrite)
     else:
         check_variable_and_dtype(
@@ -3018,7 +3018,7 @@ def scatter_nd_add(x, index, updates, name=None):
             print(output.shape)
             # [3, 5, 9, 10]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.scatter_nd_add(x, index, updates)
     else:
         if x.dtype != updates.dtype:
@@ -3156,7 +3156,7 @@ def tile(x, repeat_times, name=None):
             # Tensor(shape=[1, 6], dtype=int32, place=Place(gpu:0), stop_gradient=True,
             #        [[1, 2, 3, 1, 2, 3]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(repeat_times, core.eager.Tensor):
             assert (
                 repeat_times.ndim == 1
@@ -3269,7 +3269,7 @@ def expand_as(x, y, name=None):
             #        [[1, 2, 3],
             #         [1, 2, 3]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.expand_as(x, None, y.shape)
     else:
         check_variable_and_dtype(
@@ -3336,7 +3336,7 @@ def broadcast_to(x, shape, name=None):
             print(out)
             # [[1, 2, 3], [1, 2, 3]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.expand(x, shape)
     else:
         if isinstance(shape, Variable):
@@ -3438,7 +3438,7 @@ def expand(x, shape, name=None):
             print(out)
             # [[1, 2, 3], [1, 2, 3]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.expand(x, shape)
     else:
         if isinstance(shape, Variable):
@@ -3572,7 +3572,7 @@ def reshape(x, shape, name=None):
             # the value is [10.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(shape, (list, tuple)):
             new_shape = []
             for ele in shape:
@@ -3681,7 +3681,7 @@ def reshape_(x, shape, name=None):
     Inplace version of ``reshape`` API, the output Tensor will be inplaced with input ``x``.
     Please refer to :ref:`api_paddle_tensor_reshape`.
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         tmp_tensor_type = core.eager.Tensor
         if isinstance(shape, (list, tuple)):
             shape = [
@@ -3776,7 +3776,7 @@ def gather_nd(x, index, name=None):
             output = paddle.gather_nd(x, index) #[[3, 4]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.gather_nd(x, index)
     else:
         check_variable_and_dtype(
@@ -3894,7 +3894,7 @@ def strided_slice(x, axes, starts, ends, strides, name=None):
             sliced_2 = paddle.strided_slice(x, axes=axes, starts=[minus_3, 0, 2], ends=ends, strides=strides_2)
             # sliced_2 is x[:, 1:3:1, 0:2:1, 2:4:2].
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.strided_slice(x, axes, starts, ends, strides)
     else:
         helper = LayerHelper('strided_slice', **locals())
@@ -4143,7 +4143,7 @@ def tensordot(x, y, axes=2, name=None):
     check_type(axes, 'axes', (int, tuple, list, Variable), op_type)
 
     def _var_to_list(var):
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return tolist(var)
         raise TypeError(
             "The 'axes' with type 'Tensor' in "
@@ -4266,7 +4266,7 @@ def as_complex(x, name=None):
             #        [[1j      , (2+3j)  , (4+5j)  ],
             #         [(6+7j)  , (8+9j)  , (10+11j)]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.as_complex(x)
     else:
         check_variable_and_dtype(x, 'x', ['float32', 'float64'], 'as_complex')
@@ -4319,7 +4319,7 @@ def as_real(x, name=None):
             #          [8. , 9. ],
             #          [10., 11.]]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.as_real(x)
     else:
         check_variable_and_dtype(x, 'x', ['complex64', 'complex128'], 'as_real')
@@ -4374,7 +4374,7 @@ def repeat_interleave(x, repeats, axis=None, name=None):
         x = paddle.flatten(x)
         axis = 0
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(repeats, Variable):
             return _C_ops.repeat_interleave_with_tensor_index(x, repeats, axis)
         return _C_ops.repeat_interleave(x, repeats, axis)
@@ -4485,7 +4485,7 @@ def moveaxis(x, source, destination, name=None):
     for i in range(len(src_dims)):
         perm[dst_dims[i]] = src_dims[i]
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.transpose(x, perm)
         return out
     else:
@@ -4578,7 +4578,7 @@ def take_along_axis(arr, indices, axis):
     if not broadcast_shape:
         # if indices matrix have larger size than arr, arr should broadcast into indices shape.
         broadcast_shape = indices.shape
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         indices = paddle.broadcast_to(indices, broadcast_shape)
         broadcast_shape_list = list(broadcast_shape)
         broadcast_shape_list[axis] = list(arr.shape)[axis]
@@ -4655,7 +4655,7 @@ def put_along_axis(arr, indices, values, axis, reduce='assign'):
         )
     axis = non_negative_axis(arr, axis)
     broadcast_shape = infer_broadcast_shape(arr, indices, axis)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         values = (
             paddle.to_tensor(values)
             if not isinstance(values, paddle.Tensor)
@@ -4752,7 +4752,7 @@ def index_add(x, index, axis, value, name=None):
             #         [1., 1., 1.],
             #         [2., 2., 2.]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.index_add(x, index, value, axis)
 
     helper = LayerHelper("index_add", **locals())
@@ -4887,7 +4887,7 @@ def index_put(x, indices, value, accumulate=False, name=None):
             #         [0., 0., 1.],
             #         [0., 1., 0.]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.index_put(x, indices, value, accumulate)
 
     helper = LayerHelper("index_put", **locals())
diff --git a/python/paddle/tensor/math.py b/python/paddle/tensor/math.py
index d36e0ff3064..e5a4cbe9a8e 100644
--- a/python/paddle/tensor/math.py
+++ b/python/paddle/tensor/math.py
@@ -37,7 +37,7 @@ from ..framework import (
     _dygraph_tracer,
     convert_np_dtype_to_dtype_,
     core,
-    in_dygraph_mode,
+    in_dynamic_mode,
 )
 from .creation import _complex_to_real_dtype
 from .layer_function_generator import generate_layer_fn, templatedoc
@@ -155,7 +155,7 @@ def log(x, name=None):
             res = paddle.log(x)
             # [[0.693147, 1.09861, 1.38629], [1.94591, 2.07944, 2.19722]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.log(x)
     else:
         check_variable_and_dtype(
@@ -216,7 +216,7 @@ def scale(x, scale=1.0, bias=0.0, bias_after_scale=True, act=None, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if act is None:
             return _C_ops.scale(x, scale, float(bias), bias_after_scale)
         out = _C_ops.scale(x, scale, float(bias), bias_after_scale)
@@ -284,7 +284,7 @@ def stanh(x, scale_a=0.67, scale_b=1.7159, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.stanh(x, scale_a, scale_b)
     else:
         check_variable_and_dtype(
@@ -352,7 +352,7 @@ def multiplex(inputs, index, name=None):
             print(res) # Tensor([[5., 6.], [3., 4.]], dtype=float32)
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.multiplex(inputs, index)
     else:
         helper = LayerHelper('multiplex', **locals())
@@ -388,7 +388,7 @@ def scale_(x, scale=1.0, bias=0.0, bias_after_scale=True, act=None, name=None):
     Inplace version of ``scale`` API, the output Tensor will be inplaced with input ``x``.
     Please refer to :ref:`api_tensor_scale`.
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.scale_(x, scale, float(bias), bias_after_scale)
 
 
@@ -440,7 +440,7 @@ def pow(x, y, name=None):
 
     """
     # in dynamic graph mode
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(y, (int, float)):
             return _C_ops.pow(x, y)
         elif isinstance(y, (paddle.Tensor, Variable)):
@@ -612,7 +612,7 @@ def add(x, y, name=None):
             print(z)  # [3., 8., 6. ]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.add(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_add', **locals()))
@@ -750,7 +750,7 @@ def subtract(x, y, name=None):
             # Tensor(shape=[3], dtype=float64, place=Place(cpu), stop_gradient=True,
             #        [ 4.  ,  inf., -inf.])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.subtract(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_sub', **locals()))
@@ -806,7 +806,7 @@ def divide(x, y, name=None):
             print(z)  # [2., 0.6, 2.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.divide(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_div', **locals()))
@@ -849,7 +849,7 @@ def floor_divide(x, y, name=None):
             print(z)  # [2, 0, 2, 2]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.floor_divide(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_floordiv', **locals()))
@@ -888,7 +888,7 @@ def remainder(x, y, name=None):
             print(z)  # [0, 3, 2, 1]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.remainder(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_mod', **locals()))
@@ -951,7 +951,7 @@ def multiply(x, y, name=None):
             print(res) # [[[2, 4, 6], [2, 4, 6]]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.multiply(x, y)
     else:
         if x.dtype != y.dtype:
@@ -990,8 +990,8 @@ def _elementwise_op_with_axis_in_dygraph(
     x, y, axis=-1, name=None, op_type="Undifined"
 ):
     assert (
-        in_dygraph_mode()
-    ), "You can only call `_elementwise_op_with_axis_in_dygraph` function within in_dygraph_mode"
+        in_dynamic_mode()
+    ), "You can only call `_elementwise_op_with_axis_in_dygraph` function within in_dynamic_mode"
     assert op_type in ["add", "subtract", "multiply", "divide"], (
         "op_name input error! _elementwise_op_with_axis is an inner function to replace elementwise_add/sub/mul/div. Input op_name=%s, Expect op_name=[add|subtract|multiply|divide]\n"
         % op_type
@@ -1012,7 +1012,7 @@ def _elementwise_op_with_axis_in_dygraph(
 
 def _add_with_axis(x, y, axis=-1, name=None):
     # opt performance, only dynamic mode needs reshape
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _elementwise_op_with_axis_in_dygraph(x, y, axis, name, "add")
     else:
         op_type = 'elementwise_add'
@@ -1021,7 +1021,7 @@ def _add_with_axis(x, y, axis=-1, name=None):
 
 def _subtract_with_axis(x, y, axis=-1, name=None):
     # opt performance, only dynamic mode needs reshape
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _elementwise_op_with_axis_in_dygraph(
             x, y, axis, name, "subtract"
         )
@@ -1032,7 +1032,7 @@ def _subtract_with_axis(x, y, axis=-1, name=None):
 
 def _multiply_with_axis(x, y, axis=-1, name=None):
     # opt performance, only dynamic mode needs reshape
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _elementwise_op_with_axis_in_dygraph(
             x, y, axis, name, "multiply"
         )
@@ -1043,7 +1043,7 @@ def _multiply_with_axis(x, y, axis=-1, name=None):
 
 def _divide_with_axis(x, y, axis=-1, name=None):
     # opt performance, only dynamic mode needs reshape
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _elementwise_op_with_axis_in_dygraph(x, y, axis, name, "divide")
     else:
         op_type = 'elementwise_div'
@@ -1106,7 +1106,7 @@ def maximum(x, y, name=None):
             # Tensor(shape=[3], dtype=float32, place=Place(cpu), stop_gradient=True,
             #        [5.  , 3.  , inf.])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.maximum(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_max', **locals()))
@@ -1168,7 +1168,7 @@ def minimum(x, y, name=None):
             # Tensor(shape=[3], dtype=float64, place=Place(cpu), stop_gradient=True,
             #        [ 1.  , -inf.,  5.  ])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.minimum(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_min', **locals()))
@@ -1232,7 +1232,7 @@ def fmax(x, y, name=None):
             # Tensor(shape=[3], dtype=float32, place=Place(cpu), stop_gradient=True,
             #        [5.  , 3.  , inf.])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.fmax(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_fmax', **locals()))
@@ -1296,7 +1296,7 @@ def fmin(x, y, name=None):
             # Tensor(shape=[3], dtype=float64, place=Place(cpu), stop_gradient=True,
             #        [ 1.  , -inf.,  5.  ])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.fmin(x, y)
     else:
         return _elementwise_op(LayerHelper('elementwise_fmin', **locals()))
@@ -1367,7 +1367,7 @@ def sum(x, axis=None, dtype=None, keepdim=False, name=None):
         dtype_flag = True
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sum(x, axis, dtype, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis_with_tensor(axis, x)
@@ -1711,7 +1711,7 @@ def add_n(inputs, name=None):
             # [[8., 10., 12.],
             #  [14., 16., 18.]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(inputs, Variable):
             inputs = [inputs]
         return _C_ops.add_n(inputs)
@@ -1783,7 +1783,7 @@ def trunc(input, name=None):
             #         [[0., 0.],
             #         [0., 0.]]))
     '''
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.trunc(input)
     else:
         inputs = {"X": input}
@@ -1868,7 +1868,7 @@ def mm(input, mat2, name=None):
 
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.matmul(input, mat2, False, False)
     else:
 
@@ -2009,7 +2009,7 @@ def addmm(input, x, y, beta=1.0, alpha=1.0, name=None):
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.addmm(input, x, y, beta, alpha)
     else:
         inputs = {'Input': input, "X": x, "Y": y}
@@ -2082,7 +2082,7 @@ def renorm(x, p, axis, max_norm):
                 )
             )
         axis = axis + len(input_shape)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         out = _C_ops.renorm(x, p, axis, max_norm)
         return out
     else:
@@ -2137,7 +2137,7 @@ def inner(x, y, name=None):
         nx = x.reshape((-1, xshape[-1]))
         ny = y.reshape((-1, yshape[-1]))
 
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             return _C_ops.matmul(nx, ny.T, False, False).reshape(dstshape)
         else:
 
@@ -2204,7 +2204,7 @@ def outer(x, y, name=None):
     nx = x.reshape((-1, 1))
     ny = y.reshape((1, -1))
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.matmul(nx, ny, False, False)
     else:
 
@@ -2269,7 +2269,7 @@ def logsumexp(x, axis=None, keepdim=False, name=None):
     """
     reduce_all, axis = _get_reduce_axis(axis, x)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logsumexp(x, axis, keepdim, reduce_all)
     else:
         check_variable_and_dtype(
@@ -2312,7 +2312,7 @@ def inverse(x, name=None):
             print(inv) # [[0.5, 0], [0, 0.5]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.inverse(x)
     else:
 
@@ -2412,7 +2412,7 @@ def max(x, axis=None, keepdim=False, name=None):
             #[7., 8.], [[[0., 0.], [0., 0.]], [[0., 0.], [1., 1.]]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.max(x, axis, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis_with_tensor(axis, x)
@@ -2513,7 +2513,7 @@ def min(x, axis=None, keepdim=False, name=None):
             #[1., 2.], [[[1., 1.], [0., 0.]], [[0., 0.], [0., 0.]]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.min(x, axis, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis_with_tensor(axis, x)
@@ -2624,7 +2624,7 @@ def amax(x, axis=None, keepdim=False, name=None):
             print(result6, y.grad)
             #[0.9., 0.9], [[[0., 0.3333], [0.5, 0.3333]], [[0.5, 0.3333], [1., 1.]]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.amax(x, axis, keepdim)
 
     else:
@@ -2734,7 +2734,7 @@ def amin(x, axis=None, keepdim=False, name=None):
             print(result6, y.grad)
             #[0.1., 0.1], [[[0., 0.3333], [0.5, 0.3333]], [[0.5, 0.3333], [1., 1.]]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.amin(x, axis, keepdim)
 
     else:
@@ -2778,7 +2778,7 @@ def log1p(x, name=None):
             # [[0.], [0.6931472]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.log1p(x)
     else:
         check_variable_and_dtype(
@@ -2830,7 +2830,7 @@ def log2(x, name=None):
             res = paddle.log2(x_i)
             print(res) # [1.0]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.log2(x)
     else:
         check_variable_and_dtype(
@@ -2882,7 +2882,7 @@ def log10(x, name=None):
             res = paddle.log10(x_i)
             print(res) # [1.0]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.log10(x)
     else:
         check_variable_and_dtype(
@@ -2946,7 +2946,7 @@ def clip(x, min=None, max=None, name=None):
         min_ = float(np.finfo(np.float32).min)
         max_ = float(np.finfo(np.float32).max)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if isinstance(min, Variable):
             min = min.item(0)
         if isinstance(max, Variable):
@@ -3024,7 +3024,7 @@ def clip_(x, min=None, max=None, name=None):
     min = fmin if min is None else min
     max = fmax if max is None else max
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.clip_(x, min, max)
 
 
@@ -3101,7 +3101,7 @@ def trace(x, offset=0, axis1=0, axis2=1, name=None):
             "But received axis1 = %d, axis2 = %d\n" % (axis1, axis2)
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.trace(x, offset, axis1, axis2)
     else:
         __check_input(x, offset, axis1, axis2)
@@ -3183,7 +3183,7 @@ def diagonal(x, offset=0, axis1=0, axis2=1, name=None):
             #        [0.17020577, 0.27325270]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.diagonal(x, offset, axis1, axis2)
     else:
 
@@ -3284,7 +3284,7 @@ def kron(x, y, name=None):
             #         [12, 15, 18, 16, 20, 24],
             #         [21, 24, 27, 28, 32, 36]])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _legacy_C_ops.kron(x, y)
     else:
         helper = LayerHelper('kron', **locals())
@@ -3350,7 +3350,7 @@ def cumsum(x, axis=None, dtype=None, name=None):
     if dtype is not None and x.dtype != convert_np_dtype_to_dtype_(dtype):
         x = cast(x, dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if axis is None:
             axis = -1
         return _C_ops.cumsum(x, axis, flatten, False, False)
@@ -3426,7 +3426,7 @@ def logcumsumexp(x, axis=None, dtype=None, name=None):
     if dtype is not None and x.dtype != convert_np_dtype_to_dtype_(dtype):
         x = cast(x, dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if axis is None:
             axis = -1
         return _C_ops.logcumsumexp(x, axis, flatten, False, False)
@@ -3500,7 +3500,7 @@ def cumprod(x, dim=None, dtype=None, name=None):
     if dtype is not None and x.dtype != convert_np_dtype_to_dtype_(dtype):
         x = cast(x, dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.cumprod(x, dim)
     else:
         check_variable_and_dtype(
@@ -3552,7 +3552,7 @@ def isfinite(x, name=None):
             out = paddle.isfinite(x)
             print(out)  # [False  True  True False  True False False]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.isfinite(x)
     else:
         helper = LayerHelper("isfinite_v2", **locals())
@@ -3597,7 +3597,7 @@ def isinf(x, name=None):
             out = paddle.isinf(x)
             print(out)  # [ True False False  True False False False]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.isinf(x)
     else:
         helper = LayerHelper("isinf_v2", **locals())
@@ -3640,7 +3640,7 @@ def isnan(x, name=None):
             out = paddle.isnan(x)
             print(out)  # [False False False False False  True  True]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.isnan(x)
     else:
         helper = LayerHelper("isnan_v2", **locals())
@@ -3727,7 +3727,7 @@ def prod(x, axis=None, keepdim=False, dtype=None, name=None):
             x = cast(x, dtype)
 
     reduce_all, axis = _get_reduce_axis_with_tensor(axis, x)
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.prod(x, axis, keepdim, reduce_all)
     else:
         helper = LayerHelper('reduce_prod', **locals())
@@ -3769,7 +3769,7 @@ def sign(x, name=None):
           out = paddle.sign(x=x)
           print(out)  # [1.0, 0.0, -1.0, 1.0]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sign(x)
     else:
         check_variable_and_dtype(
@@ -3808,7 +3808,7 @@ def tanh(x, name=None):
             print(out)
             # [-0.37994896 -0.19737532  0.09966799  0.29131261]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.tanh(x)
     else:
         check_variable_and_dtype(
@@ -3853,7 +3853,7 @@ def increment(x, value=1.0, name=None):
             # [1.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.increment_(x, value)
     else:
         check_variable_and_dtype(
@@ -3918,7 +3918,7 @@ def all(x, axis=None, keepdim=False, name=None):
             print(out4)
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.all(x, axis, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis(axis, x)
@@ -3992,7 +3992,7 @@ def any(x, axis=None, keepdim=False, name=None):
             print(out4)
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.any(x, axis, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis(axis, x)
@@ -4078,7 +4078,7 @@ def conj(x, name=None):
           #        [(4-4j), (5-5j), (6-6j)]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.conj(x)
     else:
         check_variable_and_dtype(
@@ -4132,7 +4132,7 @@ def digamma(x, name=None):
             #        [ nan       ,  5.32286835]])
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.digamma(x)
     else:
         check_variable_and_dtype(
@@ -4169,7 +4169,7 @@ def lgamma(x, name=None):
             print(out)
             # [1.31452441, 1.76149750, 2.25271273, 1.09579802]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.lgamma(x)
     else:
         check_variable_and_dtype(
@@ -4251,7 +4251,7 @@ def atan2(x, y, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.atan2(x, y)
     else:
         check_variable_and_dtype(
@@ -4316,7 +4316,7 @@ def logit(x, eps=None, name=None):
     """
     if eps is None:
         eps = 0.0
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.logit(x, eps)
     else:
         check_variable_and_dtype(
@@ -4366,7 +4366,7 @@ def lerp(x, y, weight, name=None):
     if isinstance(weight, float):
         weight = paddle.full(shape=[], fill_value=weight, dtype=x.dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.lerp(x, y, weight)
     else:
         check_variable_and_dtype(
@@ -4432,7 +4432,7 @@ def erfinv(x, name=None):
             # out: [0, 0.4769, -inf]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.erfinv(x)
     else:
         check_variable_and_dtype(x, 'x', ['float32', 'float64'], 'erfinv')
@@ -4494,7 +4494,7 @@ def rad2deg(x, name=None):
             #        57.29578018)
     """
     rad2deg_scale = 180 / np.pi
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if convert_dtype(x.dtype) in ['int32', 'int64']:
             x = cast(x, dtype="float32")
         return _C_ops.scale(x, rad2deg_scale, 0.0, True)
@@ -4558,7 +4558,7 @@ def deg2rad(x, name=None):
             #        3.14159274)
     """
     deg2rad_scale = np.pi / 180.0
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if convert_dtype(x.dtype) in ['int32', 'int64']:
             x = cast(x, dtype="float32")
         return _C_ops.scale(x, deg2rad_scale, 0.0, True)
@@ -4661,7 +4661,7 @@ def gcd(x, y, name=None):
         )
         return (paddle.where(x < y, y, x), paddle.where(x < y, x, y))
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         while _gcd_cond_fn(x, y):
             x, y = _gcd_body_fn(x, y)
 
@@ -4798,7 +4798,7 @@ def diff(x, n=1, axis=-1, prepend=None, append=None, name=None):
     dtype = x.dtype
     axes = [axis]
     infer_flags = [1 for i in range(len(axes))]
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         has_pend = False
         input_list = []
         if prepend is not None and append is not None:
@@ -4950,7 +4950,7 @@ def angle(x, name=None):
             #         [-1.10714877, -0.78539819,  0.        ,  0.78539819]])
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.angle(x)
     else:
         check_variable_and_dtype(
@@ -5018,7 +5018,7 @@ def heaviside(x, y, name=None):
             #    [[0.        , 0.20000000, 1.        ],
             #     [0.        , 1.        , 0.30000001]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.heaviside(x, y)
     else:
         op_type = 'elementwise_heaviside'
@@ -5060,7 +5060,7 @@ def frac(x, name=None):
                 x.dtype
             )
         )
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         y = _C_ops.trunc(x)
         return _C_ops.subtract(x, y)
     else:
@@ -5200,7 +5200,7 @@ def take(x, index, mode='raise', name=None):
             )
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if not isinstance(index, (paddle.Tensor, Variable)):
             raise TypeError(
                 "The type of 'index' must be Tensor, but got {}".format(
@@ -5575,7 +5575,7 @@ def nextafter(x, y, name=None):
             #Tensor(shape=[2], dtype=float32, place=Place(cpu), stop_gradient=True,
             #       [1.00000012, 1.99999988])
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.nextafter(x, y)
     else:
         check_variable_and_dtype(x, 'x', ['float32', 'float64'], 'nextafter')
@@ -5614,7 +5614,7 @@ def i0(x, name=None):
             print(paddle.i0(x))
             # (Tensor(shape=[5], dtype=float32, place=Place(cpu), stop_gradient=True, [0.99999994 , 1.26606596 , 2.27958512 , 4.88079262 , 11.30192089]),
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.i0(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "i0")
@@ -5651,7 +5651,7 @@ def i0e(x, name=None):
             print(paddle.i0e(x))
             # (Tensor(shape=[5], dtype=float32, place=Place(cpu), stop_gradient=True, [1., 0.46575961, 0.30850832, 0.24300035, 0.20700192]),
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.i0e(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "i0e")
@@ -5682,7 +5682,7 @@ def i1(x, name=None):
             print(paddle.i1(x))
             # (Tensor(shape=[5], dtype=float32, place=Place(cpu), stop_gradient=True, [0., 0.5651591 , 1.59063685 , 3.95337022 , 9.75946515]),
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.i1(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "i1")
@@ -5716,7 +5716,7 @@ def i1e(x, name=None):
             print(paddle.i1e(x))
             # (Tensor(shape=[5], dtype=float32, place=Place(cpu), stop_gradient=True, [0., 0.20791042, 0.21526929, 0.24300035, 0.17875084]),
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.i1e(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "i1e")
diff --git a/python/paddle/tensor/ops.py b/python/paddle/tensor/ops.py
index 5fd9372c9cd..2218235d172 100644
--- a/python/paddle/tensor/ops.py
+++ b/python/paddle/tensor/ops.py
@@ -14,8 +14,7 @@
 
 from .. import _C_ops
 from ..fluid.data_feeder import check_variable_and_dtype
-from ..fluid.framework import in_dygraph_mode
-from ..framework import LayerHelper
+from ..framework import LayerHelper, in_dynamic_mode
 from .layer_function_generator import (
     add_sample_code,
     generate_activation_fn,
@@ -217,7 +216,7 @@ def acos(x, name=None):
             # [1.98231317 1.77215425 1.47062891 1.26610367]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.acos(x)
     else:
         check_variable_and_dtype(
@@ -254,7 +253,7 @@ def acosh(x, name=None):
             # [0.        , 1.76274729, 2.06343699, 2.29243159]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.acosh(x)
     else:
         check_variable_and_dtype(
@@ -291,7 +290,7 @@ def asin(x, name=None):
             # [-0.41151685 -0.20135792  0.10016742  0.30469265]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.asin(x)
     else:
         check_variable_and_dtype(
@@ -328,7 +327,7 @@ def asinh(x, name=None):
             # [-0.39003533, -0.19869010,  0.09983408,  0.29567307]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.asinh(x)
     else:
         check_variable_and_dtype(
@@ -365,7 +364,7 @@ def atan(x, name=None):
             # [-0.38050638 -0.19739556  0.09966865  0.29145679]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.atan(x)
     else:
         check_variable_and_dtype(
@@ -402,7 +401,7 @@ def atanh(x, name=None):
             # [-0.42364895, -0.20273256,  0.10033535,  0.30951962]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.atanh(x)
     else:
         check_variable_and_dtype(
@@ -440,7 +439,7 @@ def ceil(x, name=None):
             # [-0. -0.  1.  1.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.ceil(x)
     else:
         check_variable_and_dtype(
@@ -479,7 +478,7 @@ def cos(x, name=None):
             # [0.92106099 0.98006658 0.99500417 0.95533649]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.cos(x)
     else:
         check_variable_and_dtype(
@@ -518,7 +517,7 @@ def cosh(x, name=None):
             # [1.08107237 1.02006676 1.00500417 1.04533851]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.cosh(x)
     else:
         check_variable_and_dtype(
@@ -556,7 +555,7 @@ def exp(x, name=None):
             # [0.67032005 0.81873075 1.10517092 1.34985881]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.exp(x)
     else:
         check_variable_and_dtype(
@@ -607,7 +606,7 @@ def expm1(x, name=None):
             # [-0.32967997, -0.18126924,  0.10517092,  0.34985882]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.expm1(x)
     else:
         check_variable_and_dtype(
@@ -645,7 +644,7 @@ def floor(x, name=None):
             # [-1. -1.  0.  0.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.floor(x)
     else:
         check_variable_and_dtype(
@@ -683,7 +682,7 @@ def reciprocal(x, name=None):
             # [-2.5        -5.         10.          3.33333333]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.reciprocal(x)
     else:
         check_variable_and_dtype(
@@ -730,7 +729,7 @@ def round(x, name=None):
             # [-1. -0.  1.  2.]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.round(x)
     else:
         check_variable_and_dtype(
@@ -769,7 +768,7 @@ def rsqrt(x, name=None):
             # [3.16227766 2.23606798 1.82574186 1.58113883]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.rsqrt(x)
     else:
         check_variable_and_dtype(
@@ -807,7 +806,7 @@ def sigmoid(x, name=None):
             # [0.40131234 0.450166   0.52497919 0.57444252]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sigmoid(x)
     else:
         check_variable_and_dtype(
@@ -844,7 +843,7 @@ def sin(x, name=None):
             # [-0.38941834 -0.19866933  0.09983342  0.29552021]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sin(x)
     else:
         check_variable_and_dtype(
@@ -881,7 +880,7 @@ def sinh(x, name=None):
             # [-0.41075233 -0.201336    0.10016675  0.30452029]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sinh(x)
     else:
         check_variable_and_dtype(
@@ -917,7 +916,7 @@ def sqrt(x, name=None):
             print(out)
             # [0.31622777 0.4472136  0.54772256 0.63245553]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.sqrt(x)
     else:
         check_variable_and_dtype(
@@ -956,7 +955,7 @@ def square(x, name=None):
             print(out)
             # [0.16 0.04 0.01 0.09]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.square(x)
     else:
         check_variable_and_dtype(
@@ -1006,7 +1005,7 @@ def tan(x, name=None):
             # [-0.42279324, -0.20271005, 0.10033467, 0.30933627]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.tan(x)
     else:
         check_variable_and_dtype(
@@ -1022,7 +1021,7 @@ _erf_ = generate_layer_fn('erf')
 
 
 def erf(x, name=None):
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.erf(x)
 
     locals_var = locals().copy()
diff --git a/python/paddle/tensor/random.py b/python/paddle/tensor/random.py
index 109dbcdfd97..a8206ff95bf 100644
--- a/python/paddle/tensor/random.py
+++ b/python/paddle/tensor/random.py
@@ -17,7 +17,8 @@
 import paddle
 from paddle import _C_ops, _legacy_C_ops
 from paddle.common_ops_import import Variable
-from paddle.fluid.framework import _current_expected_place, in_dygraph_mode
+from paddle.fluid.framework import _current_expected_place
+from paddle.framework import in_dynamic_mode
 
 from ..fluid.data_feeder import (
     check_dtype,
@@ -73,7 +74,7 @@ def bernoulli(x, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.bernoulli(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "bernoulli")
@@ -119,7 +120,7 @@ def poisson(x, name=None):
             # [5., 1., 3.]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.poisson(x)
     else:
         check_variable_and_dtype(x, "x", ["float32", "float64"], "poisson")
@@ -184,7 +185,7 @@ def multinomial(x, num_samples=1, replacement=False, name=None):
         not core.is_compiled_with_rocm()
     ), "multinomial op is not supported on ROCM yet."
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.multinomial(x, num_samples, replacement)
     else:
         check_variable_and_dtype(
@@ -337,7 +338,7 @@ def gaussian(shape, mean=0.0, std=1.0, seed=0, dtype=None, name=None):
     if not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         shape = paddle.utils.convert_shape_to_list(shape)
         place = _current_expected_place()
         return _C_ops.gaussian(
@@ -525,7 +526,7 @@ def normal(mean=0.0, std=1.0, shape=None, name=None):
             # [1.00780561 3.78457445 5.81058198]  # random
 
     """
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_type(mean, 'mean', (int, float, Variable), 'normal')
         check_type(std, 'std', (int, float, Variable), 'normal')
         if isinstance(mean, Variable):
@@ -563,7 +564,7 @@ def normal(mean=0.0, std=1.0, shape=None, name=None):
         return gaussian(shape=shape, mean=mean, std=std, name=name)
 
     out = out * std + mean
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         out.stop_grediant = True
     return out
 
@@ -646,7 +647,7 @@ def uniform(shape, dtype=None, min=-1.0, max=1.0, seed=0, name=None):
     if not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         shape = paddle.utils.convert_shape_to_list(shape)
         return _C_ops.uniform(
             shape,
@@ -796,7 +797,7 @@ def randint(low=0, high=None, shape=[1], dtype=None, name=None):
     elif not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         shape = paddle.utils.convert_shape_to_list(shape)
         place = _current_expected_place()
         return _C_ops.randint(low, high, shape, dtype, place)
@@ -969,7 +970,7 @@ def randint_like(x, low=0, high=None, dtype=None, name=None):
             f"high = {high}"
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         shape = paddle.utils.convert_shape_to_list(shape)
         out = _legacy_C_ops.randint(
             'shape',
@@ -1047,7 +1048,7 @@ def randperm(n, dtype="int64", name=None):
     if not isinstance(dtype, core.VarDesc.VarType):
         dtype = convert_np_dtype_to_dtype_(dtype)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.randperm(n, dtype, _current_expected_place())
     else:
         if n < 1:
@@ -1151,7 +1152,7 @@ def exponential_(x, lam=1.0, name=None):
             #  [0.72520673, 0.45208144, 0.30234432]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.exponential_(x, lam)
     else:
         check_variable_and_dtype(
diff --git a/python/paddle/tensor/search.py b/python/paddle/tensor/search.py
index 9fc8e39a9ed..908f0836bf8 100755
--- a/python/paddle/tensor/search.py
+++ b/python/paddle/tensor/search.py
@@ -25,7 +25,7 @@ from ..framework import (
     LayerHelper,
     convert_np_dtype_to_dtype_,
     core,
-    in_dygraph_mode,
+    in_dynamic_mode,
 )
 
 # from ..fluid.layers import has_inf  #DEFINE_ALIAS
@@ -94,7 +94,7 @@ def argsort(x, axis=-1, descending=False, name=None):
             #  [1 1 0 2]
             #  [0 2 1 1]]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         _, ids = _C_ops.argsort(x, axis, descending)
         return ids
     else:
@@ -186,7 +186,7 @@ def argmax(x, axis=None, keepdim=False, dtype="int64", name=None):
         flatten = True
         axis = 0
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.argmax(x, axis, keepdim, flatten, var_dtype)
     else:
         helper = LayerHelper("argmax", **locals())
@@ -276,7 +276,7 @@ def argmin(x, axis=None, keepdim=False, dtype="int64", name=None):
         flatten = True
         axis = 0
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.argmin(x, axis, keepdim, flatten, var_dtype)
     else:
         helper = LayerHelper("argmin", **locals())
@@ -345,7 +345,7 @@ def index_select(x, index, axis=0, name=None):
             # [ 9. 10. 10.]]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.index_select(x, index, axis)
     else:
         helper = LayerHelper("index_select", **locals())
@@ -429,7 +429,7 @@ def nonzero(x, as_tuple=False):
     shape = x.shape
     rank = len(shape)
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         outs = _C_ops.nonzero(x)
     else:
         check_variable_and_dtype(
@@ -526,7 +526,7 @@ def sort(x, axis=-1, descending=False, name=None):
             #  [4. 7. 4. 6.]
             #  [5. 7. 7. 9.]]]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         outs, _ = _C_ops.argsort(x, axis, descending)
         return outs
     else:
@@ -577,7 +577,7 @@ def mode(x, axis=-1, keepdim=False, name=None):
            #    [2, 1]]))
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.mode(x, axis, keepdim)
     else:
         helper = LayerHelper("mode", **locals())
@@ -676,7 +676,7 @@ def where(condition, x=None, y=None, name=None):
         broadcast_condition = paddle.add(cast_cond, broadcast_zeros)
         broadcast_condition = paddle.cast(broadcast_condition, 'bool')
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.where(broadcast_condition, broadcast_x, broadcast_y)
     else:
         check_variable_and_dtype(condition, 'condition', ['bool'], 'where')
@@ -781,7 +781,7 @@ def index_sample(x, index):
             # [1200 1100]]
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.index_sample(x, index)
     else:
         helper = LayerHelper("index_sample", **locals())
@@ -836,7 +836,7 @@ def masked_select(x, mask, name=None):
             #[1.0 5.0 6.0 9.0]
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.masked_select(x, mask)
 
     else:
@@ -907,7 +907,7 @@ def topk(x, k, axis=None, largest=True, sorted=True, name=None):
 
     """
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if axis is None:
             axis = -1
         out, indices = _C_ops.topk(x, k, axis, largest, sorted)
@@ -1040,7 +1040,7 @@ def searchsorted(
             #         [1, 3, 4, 5]])
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.searchsorted(sorted_sequence, values, out_int32, right)
     else:
         check_variable_and_dtype(
@@ -1107,7 +1107,7 @@ def kthvalue(x, k, axis=None, keepdim=False, name=None):
             #  [[0, 2],
             #  [1, 2]]))
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         if axis is not None:
             return _C_ops.kthvalue(x, k, axis, keepdim)
         else:
diff --git a/python/paddle/tensor/stat.py b/python/paddle/tensor/stat.py
index c2332da9ea7..fa41f5a93b8 100644
--- a/python/paddle/tensor/stat.py
+++ b/python/paddle/tensor/stat.py
@@ -16,7 +16,7 @@
 
 import paddle
 from paddle import _C_ops
-from paddle.fluid.framework import in_dygraph_mode
+from paddle.framework import in_dynamic_mode
 
 from ..common_ops_import import Variable
 from ..fluid.data_feeder import check_type, check_variable_and_dtype
@@ -79,7 +79,7 @@ def mean(x, axis=None, keepdim=False, name=None):
             out4 = paddle.mean(x, axis=[0, 2])
             # [ 8.5 12.5 16.5]
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.mean(x, axis, keepdim)
     else:
         reduce_all, axis = _get_reduce_axis_with_tensor(axis, x)
@@ -144,7 +144,7 @@ def var(x, axis=None, unbiased=True, keepdim=False, name=None):
             out2 = paddle.var(x, axis=1)
             # [1.         4.33333333]
     """
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             x, 'x', ['float16', 'float32', 'float64'], 'var'
         )
@@ -212,7 +212,7 @@ def std(x, axis=None, unbiased=True, keepdim=False, name=None):
             # [1.       2.081666]
 
     """
-    if not in_dygraph_mode():
+    if not in_dynamic_mode():
         check_variable_and_dtype(
             x, 'x', ['float16', 'float32', 'float64'], 'std'
         )
@@ -242,7 +242,7 @@ def numel(x, name=None):
 
 
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.numel(x)
     else:
         if not isinstance(x, Variable):
@@ -330,7 +330,7 @@ def nanmedian(x, axis=None, keepdim=True, name=None):
     if len(axis) != len(set(axis)):
         raise ValueError("Axis has duplicated elements.")
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.nanmedian(x, axis, keepdim)
     else:
         check_variable_and_dtype(
@@ -549,7 +549,7 @@ def _compute_quantile(x, q, axis=None, keepdim=False, ignore_nan=False):
     for q_num in q:
         if q_num < 0 or q_num > 1:
             raise ValueError("q should be in range [0, 1]")
-        if in_dygraph_mode():
+        if in_dynamic_mode():
             q_num = paddle.to_tensor(q_num, dtype='float64')
         if ignore_nan:
             indices.append(q_num * (valid_counts - 1))
diff --git a/python/paddle/text/viterbi_decode.py b/python/paddle/text/viterbi_decode.py
index 0fbb2c20a6f..d0e8d120faa 100644
--- a/python/paddle/text/viterbi_decode.py
+++ b/python/paddle/text/viterbi_decode.py
@@ -12,10 +12,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from paddle import _C_ops, _legacy_C_ops
+from paddle import _C_ops
 
 from ..fluid.data_feeder import check_type, check_variable_and_dtype
-from ..fluid.framework import _non_static_mode, in_dygraph_mode
+from ..fluid.framework import in_dygraph_mode
 from ..fluid.layer_helper import LayerHelper
 from ..nn import Layer
 
@@ -62,14 +62,6 @@ def viterbi_decode(
             potentials, transition_params, lengths, include_bos_eos_tag
         )
 
-    if _non_static_mode():
-        return _legacy_C_ops.viterbi_decode(
-            potentials,
-            transition_params,
-            lengths,
-            'include_bos_eos_tag',
-            include_bos_eos_tag,
-        )
     check_variable_and_dtype(
         potentials, 'input', ['float32', 'float64'], 'viterbi_decode'
     )
diff --git a/python/paddle/utils/cpp_extension/extension_utils.py b/python/paddle/utils/cpp_extension/extension_utils.py
index f5fbcc35d46..1d6fa487c16 100644
--- a/python/paddle/utils/cpp_extension/extension_utils.py
+++ b/python/paddle/utils/cpp_extension/extension_utils.py
@@ -1142,13 +1142,13 @@ def _custom_api_content(op_name):
     API_TEMPLATE = textwrap.dedent(
         """
         import paddle.fluid.core as core
-        from paddle.fluid.framework import in_dygraph_mode
+        from paddle.framework import in_dynamic_mode
         from paddle.fluid.layer_helper import LayerHelper
 
         def {op_name}({params_list}):
             # The output variable's dtype use default value 'float32',
             # and the actual dtype of output variable will be inferred in runtime.
-            if in_dygraph_mode():
+            if in_dynamic_mode():
                 outs = core.eager._run_custom_op("{op_name}", {params_list})
                 {dynamic_content}
             else:
diff --git a/python/paddle/utils/dlpack.py b/python/paddle/utils/dlpack.py
index af4f9fe6c95..d9821698f08 100644
--- a/python/paddle/utils/dlpack.py
+++ b/python/paddle/utils/dlpack.py
@@ -16,7 +16,7 @@ import paddle
 
 from ..fluid.core import LoDTensor
 from ..fluid.data_feeder import check_type
-from ..fluid.framework import _non_static_mode
+from ..fluid.framework import in_dygraph_mode
 
 __all__ = [
     'to_dlpack',
@@ -48,7 +48,7 @@ def to_dlpack(x):
             # <capsule object "dltensor" at 0x7f6103c681b0>
     """
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         if not isinstance(x, (paddle.Tensor, paddle.fluid.core.eager.Tensor)):
             raise TypeError(
                 "The type of 'x' in to_dlpack must be paddle.Tensor,"
@@ -96,7 +96,7 @@ def from_dlpack(dlpack):
             " but received {}.".format(type(dlpack))
         )
 
-    if _non_static_mode():
+    if in_dygraph_mode():
         out = paddle.fluid.core.from_dlpack(dlpack)
         out = paddle.to_tensor(out)
         return out
diff --git a/python/paddle/utils/inplace_utils.py b/python/paddle/utils/inplace_utils.py
index b56f766f115..e02ddbeb758 100644
--- a/python/paddle/utils/inplace_utils.py
+++ b/python/paddle/utils/inplace_utils.py
@@ -15,8 +15,8 @@
 import warnings
 
 import paddle  # noqa: F401
-from paddle.fluid.framework import _non_static_mode
 from paddle.fluid.wrapped_decorator import wrap_decorator
+from paddle.framework import in_dynamic_mode
 
 
 # NOTE(pangyoki): The Inplace APIs with underline(`_`) is only valid for the method of calling `_C_ops`
@@ -24,7 +24,7 @@ from paddle.fluid.wrapped_decorator import wrap_decorator
 # of the original API will be called.
 def _inplace_apis_in_dygraph_only_(func):
     def __impl__(*args, **kwargs):
-        if not _non_static_mode():
+        if not in_dynamic_mode():
             origin_api_name = func.__name__[:-1]
             warnings.warn(
                 "In static graph mode, {}() is the same as {}() and does not perform inplace operation.".format(
diff --git a/python/paddle/utils/layers_utils.py b/python/paddle/utils/layers_utils.py
index d04c2bfeac6..fbb0781db03 100644
--- a/python/paddle/utils/layers_utils.py
+++ b/python/paddle/utils/layers_utils.py
@@ -21,7 +21,7 @@ from weakref import WeakKeyDictionary
 import paddle
 
 from ..fluid.data_feeder import check_dtype, convert_dtype
-from ..fluid.framework import Block, Variable, _non_static_mode
+from ..fluid.framework import Block, Variable, in_dygraph_mode
 
 
 def convert_to_list(value, n, name, dtype=int):
@@ -493,7 +493,7 @@ def try_set_static_shape_tensor(tensor, shape):
     # (-1, 2)
 
     """
-    if not _non_static_mode():
+    if not in_dygraph_mode():
         # static graph mode, and shape is not all inferred (contains -1)
         if -1 in tensor.shape:
             if isinstance(shape, Variable):
@@ -516,7 +516,7 @@ def try_get_constant_shape_from_tensor(shape_tensor):
     # (-1, 2)
 
     """
-    if not _non_static_mode():
+    if not in_dygraph_mode():
         try:
             if shape_tensor.op is not None:
                 generate_op = shape_tensor.op
diff --git a/test/amp/amp_base_models.py b/test/amp/amp_base_models.py
index 29e8234e913..6d08b0d1483 100644
--- a/test/amp/amp_base_models.py
+++ b/test/amp/amp_base_models.py
@@ -21,7 +21,7 @@ import numpy as np
 import paddle
 from paddle import nn
 from paddle.fluid import core
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 
 
 def copy_bits_from_float_to_uint16(f):
@@ -68,7 +68,7 @@ def _build_optimizer(
         grad_clip = paddle.nn.ClipGradByGlobalNorm(clip_norm=1.0)
     else:
         grad_clip = None
-    if _non_static_mode():
+    if in_dynamic_mode():
         assert model is not None
         parameters = model.parameters()
     else:
@@ -82,7 +82,7 @@ def _build_optimizer(
         epsilon=1e-4,
         weight_decay=0.01,
     )
-    if not _non_static_mode() and use_amp:
+    if not in_dynamic_mode() and use_amp:
         optimizer = paddle.static.amp.decorate(
             optimizer,
             amp_lists,
@@ -178,7 +178,7 @@ class SimpleConvNet(nn.Layer):
 def build_conv_model(
     use_amp, amp_dtype="float16", amp_level="O1", use_promote=False
 ):
-    if _non_static_mode():
+    if in_dynamic_mode():
         model = SimpleConvNet()
         optimizer = _build_optimizer(use_amp=False, model=model)
         if use_amp and amp_dtype == "float16":
diff --git a/test/auto_parallel/test_to_static.py b/test/auto_parallel/test_to_static.py
index 04d2591fb89..2057d509ad1 100644
--- a/test/auto_parallel/test_to_static.py
+++ b/test/auto_parallel/test_to_static.py
@@ -21,7 +21,7 @@ import paddle.nn.functional as F
 from paddle import LazyGuard, nn
 from paddle.distributed.auto_parallel.helper import ProgramHelper, ProxyLayer
 from paddle.distributed.fleet import auto
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 from paddle.io import Dataset
 from paddle.jit.dy2static.utils import is_paddle_func
 from paddle.nn import Sequential
@@ -144,7 +144,7 @@ class TestToStatic(unittest.TestCase):
         # inputs = InputSpec([batch_size, hidden_size], 'float32', 'x')
         # labels = InputSpec([batch_size], 'int64', 'label')
 
-        assert _non_static_mode()
+        assert in_dynamic_mode()
         engine = auto.Engine(
             model=mlp,
             loss=loss,
@@ -155,7 +155,7 @@ class TestToStatic(unittest.TestCase):
         engine.fit(dataset, batch_size=batch_size)
         engine.evaluate(dataset, batch_size=batch_size)
         engine.predict(dataset, batch_size=batch_size)
-        assert not _non_static_mode()
+        assert not in_dynamic_mode()
 
 
 class TestLazyInit(unittest.TestCase):
diff --git a/test/dygraph_to_static/test_convert_call.py b/test/dygraph_to_static/test_convert_call.py
index 4b325f6282c..11f947d1832 100644
--- a/test/dygraph_to_static/test_convert_call.py
+++ b/test/dygraph_to_static/test_convert_call.py
@@ -282,13 +282,13 @@ class TestConvertPaddleAPI(unittest.TestCase):
         func = paddle.nn.functional.relu
         func = paddle.jit.to_static(func)
         self.assertNotIn("_jst.IfElse", func.code)
-        self.assertIn("if in_dygraph_mode()", func.code)
+        self.assertIn("if in_dynamic_mode()", func.code)
 
     def test_class_api(self):
         bn = paddle.nn.SyncBatchNorm(2)
         paddle.jit.to_static(bn)
         self.assertNotIn("_jst.IfElse", bn.forward.code)
-        self.assertIn("if in_dygraph_mode()", bn.forward.code)
+        self.assertIn("if in_dynamic_mode()", bn.forward.code)
 
     def test_class_patch_api(self):
         paddle.nn.SyncBatchNorm.forward = forward
diff --git a/test/dygraph_to_static/test_lac.py b/test/dygraph_to_static/test_lac.py
index 933fc00d735..ead65802136 100644
--- a/test/dygraph_to_static/test_lac.py
+++ b/test/dygraph_to_static/test_lac.py
@@ -25,7 +25,7 @@ os.environ["CUDA_VISIBLE_DEVICES"] = "2"
 import paddle
 from paddle import _legacy_C_ops, fluid
 from paddle.fluid.dygraph import to_variable
-from paddle.fluid.framework import _non_static_mode
+from paddle.framework import in_dynamic_mode
 from paddle.jit.api import to_static
 from paddle.jit.translated_layer import INFER_MODEL_SUFFIX, INFER_PARAMS_SUFFIX
 
@@ -172,7 +172,7 @@ class LinearChainCRF(paddle.nn.Layer):
         self._transition = value
 
     def forward(self, input, label, length=None):
-        if _non_static_mode():
+        if in_dynamic_mode():
             _, _, _, log_likelihood = _legacy_C_ops.linear_chain_crf(
                 input, self._transition, label, length, "is_test", self._is_test
             )
@@ -236,7 +236,7 @@ class CRFDecoding(paddle.nn.Layer):
         self._transition = value
 
     def forward(self, input, label=None, length=None):
-        if _non_static_mode():
+        if in_dynamic_mode():
             return _legacy_C_ops.crf_decoding(
                 input, self._transition, label, length, "is_test", self._is_test
             )
@@ -272,7 +272,7 @@ class ChunkEval(paddle.nn.Layer):
         self.excluded_chunk_types = excluded_chunk_types
 
     def forward(self, input, label, seq_length=None):
-        if _non_static_mode():
+        if in_dynamic_mode():
             return _legacy_C_ops.chunk_eval(
                 input,
                 label,
diff --git a/test/ir/inference/test_trt_multiclass_nms3_op.py b/test/ir/inference/test_trt_multiclass_nms3_op.py
index 6ad37b18985..9b83863a87e 100644
--- a/test/ir/inference/test_trt_multiclass_nms3_op.py
+++ b/test/ir/inference/test_trt_multiclass_nms3_op.py
@@ -22,8 +22,8 @@ import paddle
 from paddle import fluid
 from paddle.fluid import core
 from paddle.fluid.core import AnalysisConfig, PassVersionChecker
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.static import nn
 
 
@@ -131,7 +131,7 @@ def multiclass_nms(
                                             normalized=False,
                                             return_index=True)
     """
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         attrs = (
             'background_label',
             background_label,
diff --git a/test/legacy_test/test_dlpack.py b/test/legacy_test/test_dlpack.py
index dbfd08316c5..d908b316c06 100644
--- a/test/legacy_test/test_dlpack.py
+++ b/test/legacy_test/test_dlpack.py
@@ -27,7 +27,7 @@ class TestDLPack(unittest.TestCase):
         tensor = paddle.to_tensor(np.array([1, 2, 3, 4]).astype('int'))
         dlpack = paddle.utils.dlpack.to_dlpack(tensor)
         out_from_dlpack = paddle.utils.dlpack.from_dlpack(dlpack)
-        if paddle.fluid.framework.in_dygraph_mode():
+        if paddle.in_dynamic_mode():
             self.assertTrue(
                 isinstance(out_from_dlpack, paddle.fluid.core.eager.Tensor)
             )
diff --git a/test/prim/composite_ops/test_composite_layer_norm.py b/test/prim/composite_ops/test_composite_layer_norm.py
index f56dbea1556..31893397abd 100644
--- a/test/prim/composite_ops/test_composite_layer_norm.py
+++ b/test/prim/composite_ops/test_composite_layer_norm.py
@@ -20,8 +20,8 @@ from utils import SUB_TOLERANCE
 import paddle
 from paddle import _C_ops
 from paddle.fluid import core, framework
-from paddle.fluid.framework import in_dygraph_mode
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 from paddle.incubate.autograd import primapi
 from paddle.nn import LayerNorm
 
@@ -56,7 +56,7 @@ def layer_norm_wrapper(
             + str(input_shape)
         )
 
-    if in_dygraph_mode():
+    if in_dynamic_mode():
         return _C_ops.layer_norm(x, weight, bias, epsilon, begin_norm_axis)
 
     else:
diff --git a/test/tokenizer/test_faster_tokenizer_op.py b/test/tokenizer/test_faster_tokenizer_op.py
index 37bb09a514a..a5abbef8ab6 100755
--- a/test/tokenizer/test_faster_tokenizer_op.py
+++ b/test/tokenizer/test_faster_tokenizer_op.py
@@ -21,8 +21,9 @@ from bert_tokenizer import BertTokenizer
 
 import paddle
 from paddle import _legacy_C_ops, nn
-from paddle.fluid.framework import _non_static_mode, core
+from paddle.fluid.framework import core
 from paddle.fluid.layer_helper import LayerHelper
+from paddle.framework import in_dynamic_mode
 
 
 def to_string_tensor(string_values, name):
@@ -77,7 +78,7 @@ class FasterTokenizer(nn.Layer):
         is_split_into_words=False,
         pad_to_max_seq_len=False,
     ):
-        if _non_static_mode():
+        if in_dynamic_mode():
             input_ids, seg_ids = _legacy_C_ops.faster_tokenizer(
                 self.vocab,
                 text,
-- 
GitLab