From 529e74e41a56dba4bf14885af9d317e53b75d30a Mon Sep 17 00:00:00 2001
From: Roc <30228238+sljlp@users.noreply.github.com>
Date: Thu, 1 Dec 2022 17:52:07 +0800
Subject: [PATCH] [Clean Fluid] replace accuracy and auc  and remove
 get_places, distributions(#48554)

* mv accuracy and auc

* rm distributions

* rm get_places

* replace metric
---
 .../distributed/fleet/metrics/metric.py       |   6 +-
 .../contrib/slim/tests/test_imperative_ptq.py |   4 +-
 .../contrib/slim/tests/test_imperative_qat.py |  10 +-
 .../slim/tests/test_imperative_qat_amp.py     |   8 +-
 .../slim/tests/test_imperative_qat_lsq.py     |   6 +-
 .../tests/test_image_classification_fp16.py   |   2 +-
 .../incubate/fleet/tests/fleet_deep_ctr.py    |   4 +-
 .../fluid/incubate/fleet/utils/fleet_util.py  |   6 +-
 python/paddle/fluid/layers/__init__.py        |   8 -
 python/paddle/fluid/layers/device.py          |  43 -
 python/paddle/fluid/layers/distributions.py   | 721 ----------------
 python/paddle/fluid/metrics.py                |   2 +-
 .../tests/book/notest_understand_sentiment.py |   4 +-
 .../tests/book/test_image_classification.py   |   2 +-
 .../fluid/tests/book/test_recognize_digits.py |   2 +-
 .../tests/unittests/check_nan_inf_base.py     |   2 +-
 .../fleet/dist_mnist_gradient_merge.py        |   2 +-
 .../collective/fleet/pipeline_mnist.py        |   2 +-
 .../fleet/pipeline_mnist_multi_device.py      |   2 +-
 .../fleet/pipeline_mnist_one_device.py        |   2 +-
 .../tests/unittests/dist_allreduce_op.py      |   2 +-
 .../paddle/fluid/tests/unittests/dist_ctr.py  |   4 +-
 .../fluid/tests/unittests/dist_fleet_ctr.py   |   4 +-
 .../dist_fleet_raw_program_optimizer.py       |   2 +-
 ...et_raw_program_optimizer_fuse_allreduce.py |   2 +-
 .../dist_fleet_sparse_embedding_ctr.py        |   4 +-
 .../fluid/tests/unittests/dist_mnist.py       |   2 +-
 .../tests/unittests/dist_mnist_batch_merge.py |   2 +-
 .../unittests/dist_mnist_fp16_allreduce.py    |   2 +-
 .../fluid/tests/unittests/dist_mnist_lars.py  |   2 +-
 .../fluid/tests/unittests/dist_se_resnext.py  |   4 +-
 .../unittests/dist_text_classification.py     |   2 +-
 .../distribution/test_distribution.py         | 162 ----
 .../dygraph_to_static/bert_dygraph_model.py   |   2 +-
 .../unittests/dygraph_to_static/test_mnist.py |   2 +-
 .../dygraph_to_static/test_mobile_net.py      |   4 +-
 .../dygraph_to_static/test_resnet.py          |   4 +-
 .../dygraph_to_static/test_resnet_amp.py      |   4 +-
 .../test_resnet_pure_fp16.py                  |   4 +-
 .../dygraph_to_static/test_se_resnet.py       |   4 +-
 .../dygraph_to_static/test_sentiment.py       |   8 +-
 .../unittests/dygraph_to_static/test_tsm.py   |   4 +-
 .../unittests/mlu/test_accuracy_op_mlu.py     |   6 +-
 .../fluid/tests/unittests/test_accuracy_op.py |   6 +-
 .../test_async_ssa_graph_executor_mnist.py    |   2 +-
 .../fluid/tests/unittests/test_auc_op.py      |   2 +-
 .../fluid/tests/unittests/test_desc_clone.py  |   2 +-
 .../tests/unittests/test_distributions.py     | 799 ------------------
 .../tests/unittests/test_get_places_op.py     |  48 --
 .../unittests/test_inference_model_io.py      |   4 +-
 .../fluid/tests/unittests/test_layers.py      |  14 +-
 .../fluid/tests/unittests/test_profiler.py    |   2 +-
 .../unittests/test_program_prune_backward.py  |   2 +-
 python/paddle/static/__init__.py              |   5 +-
 .../metric_op.py => static/nn/metric.py}      |  18 +-
 55 files changed, 93 insertions(+), 1885 deletions(-)
 delete mode 100644 python/paddle/fluid/layers/device.py
 delete mode 100644 python/paddle/fluid/layers/distributions.py
 delete mode 100644 python/paddle/fluid/tests/unittests/test_get_places_op.py
 rename python/paddle/{fluid/layers/metric_op.py => static/nn/metric.py} (97%)
 mode change 100755 => 100644

diff --git a/python/paddle/distributed/fleet/metrics/metric.py b/python/paddle/distributed/fleet/metrics/metric.py
index aaf1115af8..ba2acd5ee3 100644
--- a/python/paddle/distributed/fleet/metrics/metric.py
+++ b/python/paddle/distributed/fleet/metrics/metric.py
@@ -148,8 +148,8 @@ def auc(stat_pos, stat_neg, scope=None, util=None):
     distributed auc in fleet
 
     Args:
-        stat_pos(numpy.array|Variable|string): stat_pos in output of fluid.layers.auc
-        stat_neg(numpy.array|Variable|string): stat_neg in output of fluid.layers.auc
+        stat_pos(numpy.array|Variable|string): stat_pos in output of paddle.static.auc
+        stat_neg(numpy.array|Variable|string): stat_neg in output of paddle.static.auc
         scope(Scope): specific scope
 
     Returns:
@@ -163,7 +163,7 @@ def auc(stat_pos, stat_neg, scope=None, util=None):
           binary_predict = fluid.layers.concat(
               input=[fluid.layers.elementwise_sub(fluid.layers.ceil(similarity_norm), similarity_norm), similarity_norm], axis=1)
           self.auc, batch_auc, [batch_stat_pos, batch_stat_neg, stat_pos, stat_neg] =
-              fluid.layers.auc(input=binary_predict, label=label, curve='ROC', num_thresholds=4096)
+              paddle.static.auc(input=binary_predict, label=label, curve='ROC', num_thresholds=4096)
 
           # in train.py, after train or infer
           pos = np.array(scope.find_var(stat_pos.name).get_tensor())
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_ptq.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_ptq.py
index 759e74907e..df182f6c9c 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_ptq.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_ptq.py
@@ -150,8 +150,8 @@ class TestImperativePTQ(unittest.TestCase):
             label = paddle.to_tensor(y_data)
 
             out = model(img)
-            acc_top1 = fluid.layers.accuracy(input=out, label=label, k=1)
-            acc_top5 = fluid.layers.accuracy(input=out, label=label, k=5)
+            acc_top1 = paddle.static.accuracy(input=out, label=label, k=1)
+            acc_top5 = paddle.static.accuracy(input=out, label=label, k=5)
             eval_acc_top1_list.append(float(acc_top1.numpy()))
 
             if batch_id % 50 == 0:
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat.py
index aff07fb397..0e0fbd752b 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat.py
@@ -129,7 +129,7 @@ class TestImperativeQat(unittest.TestCase):
                     img = fluid.dygraph.to_variable(x_data)
                     label = fluid.dygraph.to_variable(y_data)
                     out = lenet(img)
-                    acc = fluid.layers.accuracy(out, label)
+                    acc = paddle.static.accuracy(out, label)
                     loss = fluid.layers.cross_entropy(out, label)
                     avg_loss = paddle.mean(loss)
                     avg_loss.backward()
@@ -160,10 +160,10 @@ class TestImperativeQat(unittest.TestCase):
                     label = fluid.dygraph.to_variable(y_data)
 
                     out = lenet(img)
-                    acc_top1 = fluid.layers.accuracy(
+                    acc_top1 = paddle.static.accuracy(
                         input=out, label=label, k=1
                     )
-                    acc_top5 = fluid.layers.accuracy(
+                    acc_top5 = paddle.static.accuracy(
                         input=out, label=label, k=5
                     )
 
@@ -200,7 +200,7 @@ class TestImperativeQat(unittest.TestCase):
             label = fluid.dygraph.to_variable(y_data)
             lenet.eval()
             fp32_out = lenet(test_img)
-            fp32_acc = fluid.layers.accuracy(fp32_out, label).numpy()
+            fp32_acc = paddle.static.accuracy(fp32_out, label).numpy()
 
         with tempfile.TemporaryDirectory(prefix="qat_save_path_") as tmpdir:
             # save inference quantized model
@@ -237,7 +237,7 @@ class TestImperativeQat(unittest.TestCase):
             )
             paddle.disable_static()
             quant_out = fluid.dygraph.to_variable(quant_out)
-            quant_acc = fluid.layers.accuracy(quant_out, label).numpy()
+            quant_acc = paddle.static.accuracy(quant_out, label).numpy()
             paddle.enable_static()
             delta_value = fp32_acc - quant_acc
             self.assertLessEqual(delta_value, self.diff_threshold)
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_amp.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_amp.py
index ee0edb445e..d01fc2e63c 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_amp.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_amp.py
@@ -118,7 +118,7 @@ class TestImperativeQatAmp(unittest.TestCase):
             if use_amp:
                 with paddle.amp.auto_cast():
                     out = model(img)
-                    acc = fluid.layers.accuracy(out, label)
+                    acc = paddle.static.accuracy(out, label)
                     loss = fluid.layers.cross_entropy(out, label)
                     avg_loss = paddle.mean(loss)
                 scaled_loss = scaler.scale(avg_loss)
@@ -128,7 +128,7 @@ class TestImperativeQatAmp(unittest.TestCase):
                 adam.clear_gradients()
             else:
                 out = model(img)
-                acc = fluid.layers.accuracy(out, label)
+                acc = paddle.static.accuracy(out, label)
                 loss = fluid.layers.cross_entropy(out, label)
                 avg_loss = paddle.mean(loss)
                 avg_loss.backward()
@@ -167,8 +167,8 @@ class TestImperativeQatAmp(unittest.TestCase):
 
             with paddle.amp.auto_cast(use_amp):
                 out = model(img)
-                acc_top1 = fluid.layers.accuracy(input=out, label=label, k=1)
-                acc_top5 = fluid.layers.accuracy(input=out, label=label, k=5)
+                acc_top1 = paddle.static.accuracy(input=out, label=label, k=1)
+                acc_top5 = paddle.static.accuracy(input=out, label=label, k=5)
 
             acc_top1_list.append(float(acc_top1.numpy()))
             if batch_id % 100 == 0:
diff --git a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
index 1b54a5b55b..2b06ee5bf0 100644
--- a/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
+++ b/python/paddle/fluid/contrib/slim/tests/test_imperative_qat_lsq.py
@@ -170,7 +170,7 @@ class TestImperativeQatLSQ(unittest.TestCase):
                 img = fluid.dygraph.to_variable(x_data)
                 label = fluid.dygraph.to_variable(y_data)
                 out = lenet(img)
-                acc = fluid.layers.accuracy(out, label)
+                acc = paddle.static.accuracy(out, label)
                 loss = fluid.layers.cross_entropy(out, label)
                 avg_loss = paddle.mean(loss)
 
@@ -202,10 +202,10 @@ class TestImperativeQatLSQ(unittest.TestCase):
                     label = fluid.dygraph.to_variable(y_data)
 
                     out = lenet(img)
-                    acc_top1 = fluid.layers.accuracy(
+                    acc_top1 = paddle.static.accuracy(
                         input=out, label=label, k=1
                     )
-                    acc_top5 = fluid.layers.accuracy(
+                    acc_top5 = paddle.static.accuracy(
                         input=out, label=label, k=5
                     )
 
diff --git a/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py b/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py
index 7edaeb2760..ab9ebfa719 100644
--- a/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py
+++ b/python/paddle/fluid/contrib/tests/test_image_classification_fp16.py
@@ -131,7 +131,7 @@ def train(net_type, use_cuda, save_dirname, is_local):
             logits, label, return_softmax=True
         )
         avg_cost = paddle.mean(cost)
-        acc = fluid.layers.accuracy(input=predict, label=label)
+        acc = paddle.static.accuracy(input=predict, label=label)
 
         # Test program
         test_program = train_program.clone(for_test=True)
diff --git a/python/paddle/fluid/incubate/fleet/tests/fleet_deep_ctr.py b/python/paddle/fluid/incubate/fleet/tests/fleet_deep_ctr.py
index d2c843ca4d..3c034d60f3 100644
--- a/python/paddle/fluid/incubate/fleet/tests/fleet_deep_ctr.py
+++ b/python/paddle/fluid/incubate/fleet/tests/fleet_deep_ctr.py
@@ -146,8 +146,8 @@ def model():
     merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
 
     predict = fluid.layers.fc(input=merge_layer, size=2, act='softmax')
-    acc = fluid.layers.accuracy(input=predict, label=label)
-    auc_var, batch_auc_var, auc_states = fluid.layers.auc(
+    acc = paddle.static.accuracy(input=predict, label=label)
+    auc_var, batch_auc_var, auc_states = paddle.static.auc(
         input=predict, label=label
     )
     cost = fluid.layers.cross_entropy(input=predict, label=label)
diff --git a/python/paddle/fluid/incubate/fleet/utils/fleet_util.py b/python/paddle/fluid/incubate/fleet/utils/fleet_util.py
index 8d5203f201..9678914b50 100644
--- a/python/paddle/fluid/incubate/fleet/utils/fleet_util.py
+++ b/python/paddle/fluid/incubate/fleet/utils/fleet_util.py
@@ -192,7 +192,7 @@ class FleetUtil:
                       fluid.layers.ceil(similarity_norm), similarity_norm),\
                   similarity_norm], axis=1)
               auc, batch_auc, [batch_stat_pos, batch_stat_neg, stat_pos, \
-                  stat_neg] = fluid.layers.auc(input=binary_predict,\
+                  stat_neg] = paddle.static.auc(input=binary_predict,\
                                                label=label, curve='ROC',\
                                                num_thresholds=4096)
 
@@ -1381,7 +1381,7 @@ class FleetUtil:
                       fluid.layers.ceil(similarity_norm), similarity_norm),\
                   similarity_norm], axis=1)
               auc, batch_auc, [batch_stat_pos, batch_stat_neg, stat_pos, \
-                  stat_neg] = fluid.layers.auc(input=binary_predict,\
+                  stat_neg] = paddle.static.auc(input=binary_predict,\
                                                label=label, curve='ROC',\
                                                num_thresholds=4096)
               local_sqrerr, local_abserr, local_prob, local_q, local_pos_ins,\
@@ -1581,7 +1581,7 @@ class FleetUtil:
                       fluid.layers.ceil(similarity_norm), similarity_norm),\
                   similarity_norm], axis=1)
               auc, batch_auc, [batch_stat_pos, batch_stat_neg, stat_pos, \
-                  stat_neg] = fluid.layers.auc(input=binary_predict,\
+                  stat_neg] = paddle.static.auc(input=binary_predict,\
                                                label=label, curve='ROC',\
                                                num_thresholds=4096)
               local_sqrerr, local_abserr, local_prob, local_q, local_pos_ins, \
diff --git a/python/paddle/fluid/layers/__init__.py b/python/paddle/fluid/layers/__init__.py
index 9ce0c0afeb..6a49ad0117 100644
--- a/python/paddle/fluid/layers/__init__.py
+++ b/python/paddle/fluid/layers/__init__.py
@@ -20,19 +20,14 @@ from . import tensor
 from .tensor import *
 from . import control_flow
 from .control_flow import *
-from . import device
-from .device import *
 from . import math_op_patch
 from .math_op_patch import *
 from . import loss
 from .loss import *
 from . import detection
 from .detection import *
-from . import metric_op
-from .metric_op import *
 from .learning_rate_scheduler import *
 from .collective import *
-from .distributions import *
 from .sequence_lod import *
 from . import rnn
 
@@ -41,11 +36,8 @@ __all__ += nn.__all__
 __all__ += io.__all__
 __all__ += tensor.__all__
 __all__ += control_flow.__all__
-__all__ += device.__all__
 __all__ += detection.__all__
-__all__ += metric_op.__all__
 __all__ += learning_rate_scheduler.__all__
-__all__ += distributions.__all__
 __all__ += sequence_lod.__all__
 __all__ += loss.__all__
 __all__ += rnn.__all__
diff --git a/python/paddle/fluid/layers/device.py b/python/paddle/fluid/layers/device.py
deleted file mode 100644
index ac352ef52b..0000000000
--- a/python/paddle/fluid/layers/device.py
+++ /dev/null
@@ -1,43 +0,0 @@
-#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-"""
-All util layers.
-"""
-
-from .layer_function_generator import autodoc
-from ..framework import unique_name
-from ..layer_helper import LayerHelper
-from paddle.utils import deprecated
-
-__all__ = []
-
-
-@deprecated(since='0.15.0', update_to="paddle.fluid.ParallelExecutor")
-@autodoc()
-def get_places(device_count=None, device_type=None):
-    helper = LayerHelper('get_places', **locals())
-    out_places = helper.create_variable(
-        name=unique_name.generate_with_ignorable_key(helper.name + ".out")
-    )
-    attrs = dict()
-    if device_count is not None:
-        attrs['device_count'] = int(device_count)
-    if device_type is not None:
-        attrs['device_type'] = str(device_type)
-
-    helper.append_op(
-        type='get_places', outputs={"Out": [out_places]}, attrs=attrs
-    )
-
-    return out_places
diff --git a/python/paddle/fluid/layers/distributions.py b/python/paddle/fluid/layers/distributions.py
deleted file mode 100644
index a54403013c..0000000000
--- a/python/paddle/fluid/layers/distributions.py
+++ /dev/null
@@ -1,721 +0,0 @@
-# Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-from . import control_flow
-from . import tensor
-from . import nn
-import math
-import numpy as np
-import warnings
-import paddle
-
-from ..data_feeder import (
-    convert_dtype,
-    check_variable_and_dtype,
-    check_type,
-    check_dtype,
-)
-
-__all__ = ['Uniform', 'Normal', 'Categorical', 'MultivariateNormalDiag']
-
-
-class Distribution:
-    """
-    Distribution is the abstract base class for probability distributions.
-    """
-
-    def sample(self):
-        """Sampling from the distribution."""
-        raise NotImplementedError
-
-    def entropy(self):
-        """The entropy of the distribution."""
-        raise NotImplementedError
-
-    def kl_divergence(self, other):
-        """The KL-divergence between self distributions and other."""
-        raise NotImplementedError
-
-    def log_prob(self, value):
-        """Log probability density/mass function."""
-        raise NotImplementedError
-
-    def _validate_args(self, *args):
-        """
-        Argument validation for distribution args
-        Args:
-            value (float, list, numpy.ndarray, Variable)
-        Raises
-            ValueError: if one argument is Variable, all arguments should be Variable
-        """
-        is_variable = False
-        is_number = False
-        for arg in args:
-            if isinstance(arg, tensor.Variable):
-                is_variable = True
-            else:
-                is_number = True
-
-        if is_variable and is_number:
-            raise ValueError(
-                'if one argument is Variable, all arguments should be Variable'
-            )
-
-        return is_variable
-
-    def _to_variable(self, *args):
-        """
-        Argument convert args to Variable
-
-        Args:
-            value (float, list, numpy.ndarray, Variable)
-        Returns:
-            Variable of args.
-        """
-        numpy_args = []
-        variable_args = []
-        tmp = 0.0
-
-        for arg in args:
-            valid_arg = False
-            for cls in [float, list, np.ndarray, tensor.Variable]:
-                if isinstance(arg, cls):
-                    valid_arg = True
-                    break
-            assert (
-                valid_arg
-            ), "type of input args must be float, list, numpy.ndarray or Variable."
-            if isinstance(arg, float):
-                arg = np.zeros(1) + arg
-            arg_np = np.array(arg)
-            arg_dtype = arg_np.dtype
-            if str(arg_dtype) not in ['float32']:
-                warnings.warn(
-                    "data type of argument only support float32, your argument will be convert to float32."
-                )
-                arg_np = arg_np.astype('float32')
-            tmp = tmp + arg_np
-            numpy_args.append(arg_np)
-
-        dtype = tmp.dtype
-        for arg in numpy_args:
-            arg_broadcasted, _ = np.broadcast_arrays(arg, tmp)
-            arg_variable = tensor.create_tensor(dtype=dtype)
-            tensor.assign(arg_broadcasted, arg_variable)
-            variable_args.append(arg_variable)
-
-        return tuple(variable_args)
-
-
-class Uniform(Distribution):
-    r"""Uniform distribution with `low` and `high` parameters.
-
-    Mathematical Details
-
-    The probability density function (pdf) is,
-
-    .. math::
-
-        pdf(x; a, b) = \\frac{1}{Z}, \ a <=x <b
-
-    .. math::
-
-        Z = b - a
-
-    In the above equation:
-
-    * :math:`low = a`,
-    * :math:`high = b`,
-    * :math:`Z`: is the normalizing constant.
-
-    The parameters `low` and `high` must be shaped in a way that supports
-    broadcasting (e.g., `high - low` is a valid operation).
-
-    Args:
-        low(float|list|numpy.ndarray|Variable): The lower boundary of uniform distribution.The data type is float32
-        high(float|list|numpy.ndarray|Variable): The higher boundary of uniform distribution.The data type is float32
-
-    Examples:
-        .. code-block:: python
-
-          import numpy as np
-          from paddle.fluid import layers
-          from paddle.fluid.layers import Uniform
-
-          # Without broadcasting, a single uniform distribution [3, 4]:
-          u1 = Uniform(low=3.0, high=4.0)
-          # 2 distributions [1, 3], [2, 4]
-          u2 = Uniform(low=[1.0, 2.0],
-                        high=[3.0, 4.0])
-          # 4 distributions
-          u3 = Uniform(low=[[1.0, 2.0],
-                    [3.0, 4.0]],
-               high=[[1.5, 2.5],
-                     [3.5, 4.5]])
-
-          # With broadcasting:
-          u4 = Uniform(low=3.0, high=[5.0, 6.0, 7.0])
-
-          # Complete example
-          value_npdata = np.array([0.8], dtype="float32")
-          value_tensor = layers.create_tensor(dtype="float32")
-          layers.assign(value_npdata, value_tensor)
-
-          uniform = Uniform([0.], [2.])
-
-          sample = uniform.sample([2])
-          # a random tensor created by uniform distribution with shape: [2, 1]
-          entropy = uniform.entropy()
-          # [0.6931472] with shape: [1]
-          lp = uniform.log_prob(value_tensor)
-          # [-0.6931472] with shape: [1]
-    """
-
-    def __init__(self, low, high):
-        check_type(
-            low, 'low', (float, np.ndarray, tensor.Variable, list), 'Uniform'
-        )
-        check_type(
-            high, 'high', (float, np.ndarray, tensor.Variable, list), 'Uniform'
-        )
-
-        self.all_arg_is_float = False
-        self.batch_size_unknown = False
-        if self._validate_args(low, high):
-            self.batch_size_unknown = True
-            self.low = low
-            self.high = high
-        else:
-            if isinstance(low, float) and isinstance(high, float):
-                self.all_arg_is_float = True
-            self.low, self.high = self._to_variable(low, high)
-
-    def sample(self, shape, seed=0):
-        """Generate samples of the specified shape.
-
-        Args:
-          shape (list): 1D `int32`. Shape of the generated samples.
-          seed (int): Python integer number.
-
-        Returns:
-          Variable: A tensor with prepended dimensions shape.The data type is float32.
-
-        """
-        check_type(shape, 'shape', (list), 'sample')
-        check_type(seed, 'seed', (int), 'sample')
-
-        batch_shape = list((self.low + self.high).shape)
-        if self.batch_size_unknown:
-            output_shape = shape + batch_shape
-            zero_tmp = tensor.fill_constant_batch_size_like(
-                self.low + self.high, batch_shape + shape, self.low.dtype, 0.0
-            )
-            uniform_random_tmp = (
-                paddle.tensor.random.uniform_random_batch_size_like(
-                    zero_tmp, zero_tmp.shape, min=0.0, max=1.0, seed=seed
-                )
-            )
-            output = (
-                uniform_random_tmp * (zero_tmp + self.high - self.low)
-                + self.low
-            )
-            return paddle.reshape(output, output_shape)
-        else:
-            output_shape = shape + batch_shape
-            output = (
-                nn.uniform_random(output_shape, seed=seed)
-                * (
-                    tensor.zeros(output_shape, dtype=self.low.dtype)
-                    + (self.high - self.low)
-                )
-                + self.low
-            )
-            if self.all_arg_is_float:
-                return paddle.reshape(output, shape)
-            else:
-                return output
-
-    def log_prob(self, value):
-        """Log probability density/mass function.
-
-        Args:
-          value (Variable): The input tensor.
-
-        Returns:
-          Variable: log probability.The data type is same with value.
-
-        """
-        check_variable_and_dtype(
-            value, 'value', ['float32', 'float64'], 'log_prob'
-        )
-
-        lb_bool = control_flow.less_than(self.low, value)
-        ub_bool = control_flow.less_than(value, self.high)
-        lb = tensor.cast(lb_bool, dtype=value.dtype)
-        ub = tensor.cast(ub_bool, dtype=value.dtype)
-        return paddle.log(lb * ub) - paddle.log(self.high - self.low)
-
-    def entropy(self):
-        """Shannon entropy in nats.
-
-        Returns:
-          Variable: Shannon entropy of uniform distribution.The data type is float32.
-
-        """
-        return paddle.log(self.high - self.low)
-
-
-class Normal(Distribution):
-    r"""The Normal distribution with location `loc` and `scale` parameters.
-
-    Mathematical details
-
-    The probability density function (pdf) is,
-
-    .. math::
-
-        pdf(x; \mu, \sigma) = \\frac{1}{Z}e^{\\frac {-0.5 (x - \mu)^2}  {\sigma^2} }
-
-    .. math::
-
-        Z = (2 \pi \sigma^2)^{0.5}
-
-    In the above equation:
-
-    * :math:`loc = \mu`: is the mean.
-    * :math:`scale = \sigma`: is the std.
-    * :math:`Z`: is the normalization constant.
-
-    Args:
-        loc(float|list|numpy.ndarray|Variable): The mean of normal distribution.The data type is float32.
-        scale(float|list|numpy.ndarray|Variable): The std of normal distribution.The data type is float32.
-
-    Examples:
-        .. code-block:: python
-
-          import numpy as np
-          from paddle.fluid import layers
-          from paddle.fluid.layers import Normal
-
-          # Define a single scalar Normal distribution.
-          dist = Normal(loc=0., scale=3.)
-          # Define a batch of two scalar valued Normals.
-          # The first has mean 1 and standard deviation 11, the second 2 and 22.
-          dist = Normal(loc=[1., 2.], scale=[11., 22.])
-          # Get 3 samples, returning a 3 x 2 tensor.
-          dist.sample([3])
-
-          # Define a batch of two scalar valued Normals.
-          # Both have mean 1, but different standard deviations.
-          dist = Normal(loc=1., scale=[11., 22.])
-
-          # Complete example
-          value_npdata = np.array([0.8], dtype="float32")
-          value_tensor = layers.create_tensor(dtype="float32")
-          layers.assign(value_npdata, value_tensor)
-
-          normal_a = Normal([0.], [1.])
-          normal_b = Normal([0.5], [2.])
-
-          sample = normal_a.sample([2])
-          # a random tensor created by normal distribution with shape: [2, 1]
-          entropy = normal_a.entropy()
-          # [1.4189385] with shape: [1]
-          lp = normal_a.log_prob(value_tensor)
-          # [-1.2389386] with shape: [1]
-          kl = normal_a.kl_divergence(normal_b)
-          # [0.34939718] with shape: [1]
-    """
-
-    def __init__(self, loc, scale):
-        check_type(
-            loc, 'loc', (float, np.ndarray, tensor.Variable, list), 'Normal'
-        )
-        check_type(
-            scale, 'scale', (float, np.ndarray, tensor.Variable, list), 'Normal'
-        )
-
-        self.batch_size_unknown = False
-        self.all_arg_is_float = False
-        if self._validate_args(loc, scale):
-            self.batch_size_unknown = True
-            self.loc = loc
-            self.scale = scale
-        else:
-            if isinstance(loc, float) and isinstance(scale, float):
-                self.all_arg_is_float = True
-            self.loc, self.scale = self._to_variable(loc, scale)
-
-    def sample(self, shape, seed=0):
-        """Generate samples of the specified shape.
-
-        Args:
-          shape (list): 1D `int32`. Shape of the generated samples.
-          seed (int): Python integer number.
-
-        Returns:
-          Variable: A tensor with prepended dimensions shape.The data type is float32.
-
-        """
-
-        check_type(shape, 'shape', (list), 'sample')
-        check_type(seed, 'seed', (int), 'sample')
-
-        batch_shape = list((self.loc + self.scale).shape)
-
-        if self.batch_size_unknown:
-            output_shape = shape + batch_shape
-            zero_tmp = tensor.fill_constant_batch_size_like(
-                self.loc + self.scale, batch_shape + shape, self.loc.dtype, 0.0
-            )
-            zero_tmp_shape = nn.shape(zero_tmp)
-            normal_random_tmp = nn.gaussian_random(
-                zero_tmp_shape, mean=0.0, std=1.0, seed=seed
-            )
-            output = normal_random_tmp * (zero_tmp + self.scale) + self.loc
-            return paddle.reshape(output, output_shape)
-        else:
-            output_shape = shape + batch_shape
-            output = (
-                nn.gaussian_random(output_shape, mean=0.0, std=1.0, seed=seed)
-                * (
-                    tensor.zeros(output_shape, dtype=self.loc.dtype)
-                    + self.scale
-                )
-                + self.loc
-            )
-            if self.all_arg_is_float:
-                return paddle.reshape(output, shape)
-            else:
-                return output
-
-    def entropy(self):
-        """Shannon entropy in nats.
-
-        Returns:
-          Variable: Shannon entropy of normal distribution.The data type is float32.
-
-        """
-        batch_shape = list((self.loc + self.scale).shape)
-        zero_tmp = tensor.fill_constant_batch_size_like(
-            self.loc + self.scale, batch_shape, self.loc.dtype, 0.0
-        )
-        return (
-            0.5
-            + 0.5 * math.log(2 * math.pi)
-            + paddle.log((self.scale + zero_tmp))
-        )
-
-    def log_prob(self, value):
-        """Log probability density/mass function.
-
-        Args:
-          value (Variable): The input tensor.
-
-        Returns:
-          Variable: log probability.The data type is same with value.
-
-        """
-        check_variable_and_dtype(
-            value, 'value', ['float32', 'float64'], 'log_prob'
-        )
-
-        var = self.scale * self.scale
-        log_scale = paddle.log(self.scale)
-        return (
-            -1.0 * ((value - self.loc) * (value - self.loc)) / (2.0 * var)
-            - log_scale
-            - math.log(math.sqrt(2.0 * math.pi))
-        )
-
-    def kl_divergence(self, other):
-        """The KL-divergence between two normal distributions.
-
-        Args:
-            other (Normal): instance of Normal.
-
-        Returns:
-            Variable: kl-divergence between two normal distributions.The data type is float32.
-
-        """
-
-        check_type(other, 'other', Normal, 'kl_divergence')
-
-        var_ratio = self.scale / other.scale
-        var_ratio = var_ratio * var_ratio
-        t1 = (self.loc - other.loc) / other.scale
-        t1 = t1 * t1
-        return 0.5 * (var_ratio + t1 - 1.0 - paddle.log(var_ratio))
-
-
-class Categorical(Distribution):
-    r"""
-    Categorical distribution is a discrete probability distribution that
-    describes the possible results of a random variable that can take on
-    one of K possible categories, with the probability of each category
-    separately specified.
-
-    The probability mass function (pmf) is:
-
-    .. math::
-
-        pmf(k; p_i) = \prod_{i=1}^{k} p_i^{[x=i]}
-
-    In the above equation:
-
-    * :math:`[x=i]` : it evaluates to 1 if :math:`x==i` , 0 otherwise.
-
-    Args:
-        logits(list|numpy.ndarray|Variable): The logits input of categorical distribution. The data type is float32.
-
-    Examples:
-        .. code-block:: python
-
-          import numpy as np
-          from paddle.fluid import layers
-          from paddle.fluid.layers import Categorical
-
-          a_logits_npdata = np.array([-0.602,-0.602], dtype="float32")
-          a_logits_tensor = layers.create_tensor(dtype="float32")
-          layers.assign(a_logits_npdata, a_logits_tensor)
-
-          b_logits_npdata = np.array([-0.102,-0.112], dtype="float32")
-          b_logits_tensor = layers.create_tensor(dtype="float32")
-          layers.assign(b_logits_npdata, b_logits_tensor)
-
-          a = Categorical(a_logits_tensor)
-          b = Categorical(b_logits_tensor)
-
-          a.entropy()
-          # [0.6931472] with shape: [1]
-
-          b.entropy()
-          # [0.6931347] with shape: [1]
-
-          a.kl_divergence(b)
-          # [1.2516975e-05] with shape: [1]
-
-    """
-
-    def __init__(self, logits):
-        """
-        Args:
-            logits(list|numpy.ndarray|Variable): The logits input of categorical distribution. The data type is float32.
-        """
-        check_type(
-            logits, 'logits', (np.ndarray, tensor.Variable, list), 'Categorical'
-        )
-
-        if self._validate_args(logits):
-            self.logits = logits
-        else:
-            self.logits = self._to_variable(logits)[0]
-
-    def kl_divergence(self, other):
-        """The KL-divergence between two Categorical distributions.
-
-        Args:
-            other (Categorical): instance of Categorical. The data type is float32.
-
-        Returns:
-            Variable: kl-divergence between two Categorical distributions.
-
-        """
-        check_type(other, 'other', Categorical, 'kl_divergence')
-
-        logits = self.logits - paddle.max(self.logits, axis=-1, keepdim=True)
-        other_logits = other.logits - paddle.max(
-            other.logits, axis=-1, keepdim=True
-        )
-        e_logits = paddle.exp(logits)
-        other_e_logits = paddle.exp(other_logits)
-        z = paddle.sum(e_logits, axis=-1, keepdim=True)
-        other_z = paddle.sum(other_e_logits, axis=-1, keepdim=True)
-        prob = e_logits / z
-        kl = paddle.sum(
-            prob
-            * (logits - paddle.log(z) - other_logits + paddle.log(other_z)),
-            axis=-1,
-            keepdim=True,
-        )
-
-        return kl
-
-    def entropy(self):
-        """Shannon entropy in nats.
-
-        Returns:
-          Variable: Shannon entropy of Categorical distribution. The data type is float32.
-
-        """
-        logits = self.logits - paddle.max(self.logits, axis=-1, keepdim=True)
-        e_logits = paddle.exp(logits)
-        z = paddle.sum(e_logits, axis=-1, keepdim=True)
-
-        prob = e_logits / z
-        entropy = -1.0 * paddle.sum(
-            prob * (logits - paddle.log(z)), axis=-1, keepdim=True
-        )
-
-        return entropy
-
-
-class MultivariateNormalDiag(Distribution):
-    r"""
-    A multivariate normal (also called Gaussian) distribution parameterized by a mean vector
-    and a covariance matrix.
-
-    The probability density function (pdf) is:
-
-    .. math::
-
-        pdf(x; loc, scale) = \\frac{e^{-\\frac{||y||^2}{2}}}{Z}
-
-    where:
-    .. math::
-
-        y = inv(scale) @ (x - loc)
-        Z = (2\\pi)^{0.5k} |det(scale)|
-
-
-    In the above equation:
-
-    * :math:`inv` : denotes to take the inverse of the matrix.
-    * :math:`@` : denotes matrix multiplication.
-    * :math:`det` : denotes to evaluate the determinant.
-
-    Args:
-        loc(list|numpy.ndarray|Variable): The mean of multivariateNormal distribution with shape :math:`[k]` .
-            The data type is float32.
-        scale(list|numpy.ndarray|Variable): The positive definite diagonal covariance matrix of multivariateNormal
-            distribution  with shape :math:`[k, k]` . All elements are 0 except diagonal elements. The data type is
-            float32.
-
-    Examples:
-        .. code-block:: python
-
-            import numpy as np
-            from paddle.fluid import layers
-            from paddle.fluid.layers import MultivariateNormalDiag
-
-            a_loc_npdata = np.array([0.3,0.5],dtype="float32")
-            a_loc_tensor = layers.create_tensor(dtype="float32")
-            layers.assign(a_loc_npdata, a_loc_tensor)
-
-
-            a_scale_npdata = np.array([[0.4,0],[0,0.5]],dtype="float32")
-            a_scale_tensor = layers.create_tensor(dtype="float32")
-            layers.assign(a_scale_npdata, a_scale_tensor)
-
-            b_loc_npdata = np.array([0.2,0.4],dtype="float32")
-            b_loc_tensor = layers.create_tensor(dtype="float32")
-            layers.assign(b_loc_npdata, b_loc_tensor)
-
-            b_scale_npdata = np.array([[0.3,0],[0,0.4]],dtype="float32")
-            b_scale_tensor = layers.create_tensor(dtype="float32")
-            layers.assign(b_scale_npdata, b_scale_tensor)
-
-            a = MultivariateNormalDiag(a_loc_tensor, a_scale_tensor)
-            b = MultivariateNormalDiag(b_loc_tensor, b_scale_tensor)
-
-            a.entropy()
-            # [2.033158] with shape: [1]
-            b.entropy()
-            # [1.7777451] with shape: [1]
-
-            a.kl_divergence(b)
-            # [0.06542051] with shape: [1]
-
-    """
-
-    def __init__(self, loc, scale):
-        check_type(
-            loc,
-            'loc',
-            (np.ndarray, tensor.Variable, list),
-            'MultivariateNormalDiag',
-        )
-        check_type(
-            scale,
-            'scale',
-            (np.ndarray, tensor.Variable, list),
-            'MultivariateNormalDiag',
-        )
-
-        if self._validate_args(loc, scale):
-            self.loc = loc
-            self.scale = scale
-        else:
-            self.loc, self.scale = self._to_variable(loc, scale)
-
-    def _det(self, value):
-
-        batch_shape = list(value.shape)
-        one_all = paddle.ones(shape=batch_shape, dtype=self.loc.dtype)
-        one_diag = tensor.diag(
-            paddle.ones(shape=[batch_shape[0]], dtype=self.loc.dtype)
-        )
-        det_diag = paddle.prod(value + one_all - one_diag)
-
-        return det_diag
-
-    def _inv(self, value):
-
-        batch_shape = list(value.shape)
-        one_all = paddle.ones(shape=batch_shape, dtype=self.loc.dtype)
-        one_diag = tensor.diag(
-            paddle.ones(shape=[batch_shape[0]], dtype=self.loc.dtype)
-        )
-        inv_diag = paddle.pow(value, (one_all - 2 * one_diag))
-
-        return inv_diag
-
-    def entropy(self):
-        """Shannon entropy in nats.
-
-        Returns:
-          Variable: Shannon entropy of Multivariate Normal distribution. The data type is float32.
-
-        """
-        entropy = 0.5 * (
-            self.scale.shape[0] * (1.0 + math.log(2 * math.pi))
-            + paddle.log(self._det(self.scale))
-        )
-
-        return entropy
-
-    def kl_divergence(self, other):
-        """The KL-divergence between two Multivariate Normal distributions.
-
-        Args:
-            other (MultivariateNormalDiag): instance of Multivariate Normal.
-
-        Returns:
-            Variable: kl-divergence between two Multivariate Normal distributions. The data type is float32.
-
-        """
-        check_type(other, 'other', MultivariateNormalDiag, 'kl_divergence')
-
-        tr_cov_matmul = paddle.sum(self._inv(other.scale) * self.scale)
-        loc_matmul_cov = nn.matmul(
-            (other.loc - self.loc), self._inv(other.scale)
-        )
-        tri_matmul = nn.matmul(loc_matmul_cov, (other.loc - self.loc))
-        k = list(self.scale.shape)[0]
-        ln_cov = paddle.log(self._det(other.scale)) - paddle.log(
-            self._det(self.scale)
-        )
-        kl = 0.5 * (tr_cov_matmul + tri_matmul - k + ln_cov)
-
-        return kl
diff --git a/python/paddle/fluid/metrics.py b/python/paddle/fluid/metrics.py
index c775bff5f3..8e79a2429d 100644
--- a/python/paddle/fluid/metrics.py
+++ b/python/paddle/fluid/metrics.py
@@ -714,7 +714,7 @@ class Auc(MetricBase):
     The auc metric is for binary classification.
     Refer to https://en.wikipedia.org/wiki/Receiver_operating_characteristic#Area_under_the_curve.
     Please notice that the auc metric is implemented with python, which may be a little bit slow.
-    If you concern the speed, please use the fluid.layers.auc instead.
+    If you concern the speed, please use the paddle.static.auc instead.
 
     The `auc` function creates four local variables, `true_positives`,
     `true_negatives`, `false_positives` and `false_negatives` that are used to
diff --git a/python/paddle/fluid/tests/book/notest_understand_sentiment.py b/python/paddle/fluid/tests/book/notest_understand_sentiment.py
index ab220911b6..3809e797f3 100644
--- a/python/paddle/fluid/tests/book/notest_understand_sentiment.py
+++ b/python/paddle/fluid/tests/book/notest_understand_sentiment.py
@@ -49,7 +49,7 @@ def convolution_net(
     )
     cost = fluid.layers.cross_entropy(input=prediction, label=label)
     avg_cost = paddle.mean(cost)
-    accuracy = fluid.layers.accuracy(input=prediction, label=label)
+    accuracy = paddle.static.accuracy(input=prediction, label=label)
     return avg_cost, accuracy, prediction
 
 
@@ -84,7 +84,7 @@ def stacked_lstm_net(
     )
     cost = fluid.layers.cross_entropy(input=prediction, label=label)
     avg_cost = paddle.mean(cost)
-    accuracy = fluid.layers.accuracy(input=prediction, label=label)
+    accuracy = paddle.static.accuracy(input=prediction, label=label)
     return avg_cost, accuracy, prediction
 
 
diff --git a/python/paddle/fluid/tests/book/test_image_classification.py b/python/paddle/fluid/tests/book/test_image_classification.py
index 0ac7adb994..beb562bee5 100644
--- a/python/paddle/fluid/tests/book/test_image_classification.py
+++ b/python/paddle/fluid/tests/book/test_image_classification.py
@@ -121,7 +121,7 @@ def train(net_type, use_cuda, save_dirname, is_local):
     predict = fluid.layers.fc(input=net, size=classdim, act='softmax')
     cost = fluid.layers.cross_entropy(input=predict, label=label)
     avg_cost = paddle.mean(cost)
-    acc = fluid.layers.accuracy(input=predict, label=label)
+    acc = paddle.static.accuracy(input=predict, label=label)
 
     # Test program
     test_program = fluid.default_main_program().clone(for_test=True)
diff --git a/python/paddle/fluid/tests/book/test_recognize_digits.py b/python/paddle/fluid/tests/book/test_recognize_digits.py
index 892782820b..a89cb1617a 100644
--- a/python/paddle/fluid/tests/book/test_recognize_digits.py
+++ b/python/paddle/fluid/tests/book/test_recognize_digits.py
@@ -32,7 +32,7 @@ def loss_net(hidden, label):
     prediction = fluid.layers.fc(input=hidden, size=10, act='softmax')
     loss = fluid.layers.cross_entropy(input=prediction, label=label)
     avg_loss = paddle.mean(loss)
-    acc = fluid.layers.accuracy(input=prediction, label=label)
+    acc = paddle.static.accuracy(input=prediction, label=label)
     return prediction, avg_loss, acc
 
 
diff --git a/python/paddle/fluid/tests/unittests/check_nan_inf_base.py b/python/paddle/fluid/tests/unittests/check_nan_inf_base.py
index d96ebe3e43..c6e9a36ddf 100644
--- a/python/paddle/fluid/tests/unittests/check_nan_inf_base.py
+++ b/python/paddle/fluid/tests/unittests/check_nan_inf_base.py
@@ -66,7 +66,7 @@ def net():
     cost, y_predict = fluid.layers.softmax_with_cross_entropy(
         hidden, y, return_softmax=True
     )
-    acc_top1 = fluid.layers.accuracy(input=y_predict, label=y, k=1)
+    acc_top1 = paddle.static.accuracy(input=y_predict, label=y, k=1)
     avg_cost = paddle.mean(cost)
 
     sgd_optimizer = fluid.optimizer.SGD(learning_rate=0.05)
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/dist_mnist_gradient_merge.py b/python/paddle/fluid/tests/unittests/collective/fleet/dist_mnist_gradient_merge.py
index 35c6763b47..1e67d72204 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/dist_mnist_gradient_merge.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/dist_mnist_gradient_merge.py
@@ -39,7 +39,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist.py b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist.py
index caffe9c4e0..59572a5e7d 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist.py
@@ -106,7 +106,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
         # Evaluator
         with fluid.device_guard("gpu:1"):
             batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-            batch_acc = fluid.layers.accuracy(
+            batch_acc = paddle.static.accuracy(
                 input=predict, label=label, total=batch_size_tensor
             )
 
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_multi_device.py b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_multi_device.py
index faf74d9bf4..e60b6bbbcd 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_multi_device.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_multi_device.py
@@ -106,7 +106,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
         # Evaluator
         with fluid.device_guard("gpu:1"):
             batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-            batch_acc = fluid.layers.accuracy(
+            batch_acc = paddle.static.accuracy(
                 input=predict, label=label, total=batch_size_tensor
             )
 
diff --git a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_one_device.py b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_one_device.py
index 8ff918d0c7..8ecea66aaa 100644
--- a/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_one_device.py
+++ b/python/paddle/fluid/tests/unittests/collective/fleet/pipeline_mnist_one_device.py
@@ -98,7 +98,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
         # Evaluator
         with fluid.device_guard("gpu:0"):
             batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-            batch_acc = fluid.layers.accuracy(
+            batch_acc = paddle.static.accuracy(
                 input=predict, label=label, total=batch_size_tensor
             )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_allreduce_op.py b/python/paddle/fluid/tests/unittests/dist_allreduce_op.py
index d5ec419d79..bd3a6d659a 100644
--- a/python/paddle/fluid/tests/unittests/dist_allreduce_op.py
+++ b/python/paddle/fluid/tests/unittests/dist_allreduce_op.py
@@ -82,7 +82,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_ctr.py b/python/paddle/fluid/tests/unittests/dist_ctr.py
index f0d25089f3..73e4f7fffc 100644
--- a/python/paddle/fluid/tests/unittests/dist_ctr.py
+++ b/python/paddle/fluid/tests/unittests/dist_ctr.py
@@ -99,8 +99,8 @@ class TestDistCTR2x2(TestDistRunnerBase):
         merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
 
         predict = fluid.layers.fc(input=merge_layer, size=2, act='softmax')
-        acc = fluid.layers.accuracy(input=predict, label=label)
-        auc_var, batch_auc_var, auc_states = fluid.layers.auc(
+        acc = paddle.static.accuracy(input=predict, label=label)
+        auc_var, batch_auc_var, auc_states = paddle.static.auc(
             input=predict, label=label
         )
         cost = fluid.layers.cross_entropy(input=predict, label=label)
diff --git a/python/paddle/fluid/tests/unittests/dist_fleet_ctr.py b/python/paddle/fluid/tests/unittests/dist_fleet_ctr.py
index 988e77fc60..3826bbae5c 100644
--- a/python/paddle/fluid/tests/unittests/dist_fleet_ctr.py
+++ b/python/paddle/fluid/tests/unittests/dist_fleet_ctr.py
@@ -148,9 +148,9 @@ class TestDistCTR2x2(FleetDistRunnerBase):
         merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
 
         predict = fluid.layers.fc(input=merge_layer, size=2, act='softmax')
-        acc = fluid.layers.accuracy(input=predict, label=label)
+        acc = paddle.static.accuracy(input=predict, label=label)
 
-        auc_var, batch_auc_var, auc_states = fluid.layers.auc(
+        auc_var, batch_auc_var, auc_states = paddle.static.auc(
             input=predict, label=label
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer.py b/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer.py
index d7f5a23370..90c1ea16a8 100644
--- a/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer.py
+++ b/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer.py
@@ -84,7 +84,7 @@ class TestFleetMetaOptimizerPrecision(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer_fuse_allreduce.py b/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer_fuse_allreduce.py
index ae352e686e..98d7ef1d15 100644
--- a/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer_fuse_allreduce.py
+++ b/python/paddle/fluid/tests/unittests/dist_fleet_raw_program_optimizer_fuse_allreduce.py
@@ -84,7 +84,7 @@ class TestFleetMetaOptimizerFuseAllReducePrecision(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_fleet_sparse_embedding_ctr.py b/python/paddle/fluid/tests/unittests/dist_fleet_sparse_embedding_ctr.py
index fd12a0ecfa..1868f81dd1 100644
--- a/python/paddle/fluid/tests/unittests/dist_fleet_sparse_embedding_ctr.py
+++ b/python/paddle/fluid/tests/unittests/dist_fleet_sparse_embedding_ctr.py
@@ -138,8 +138,8 @@ class TestDistCTR2x2(FleetDistRunnerBase):
         merge_layer = fluid.layers.concat(input=[dnn_out, lr_pool], axis=1)
         predict = fluid.layers.fc(input=merge_layer, size=2, act='softmax')
 
-        acc = fluid.layers.accuracy(input=predict, label=label)
-        auc_var, _, _ = fluid.layers.auc(input=predict, label=label)
+        acc = paddle.static.accuracy(input=predict, label=label)
+        auc_var, _, _ = paddle.static.auc(input=predict, label=label)
         cost = fluid.layers.cross_entropy(input=predict, label=label)
         avg_cost = paddle.mean(x=cost)
 
diff --git a/python/paddle/fluid/tests/unittests/dist_mnist.py b/python/paddle/fluid/tests/unittests/dist_mnist.py
index 4af3aba8b8..3cecc8b32c 100644
--- a/python/paddle/fluid/tests/unittests/dist_mnist.py
+++ b/python/paddle/fluid/tests/unittests/dist_mnist.py
@@ -83,7 +83,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_mnist_batch_merge.py b/python/paddle/fluid/tests/unittests/dist_mnist_batch_merge.py
index 993ec3ec6d..4cda9dd53a 100644
--- a/python/paddle/fluid/tests/unittests/dist_mnist_batch_merge.py
+++ b/python/paddle/fluid/tests/unittests/dist_mnist_batch_merge.py
@@ -48,7 +48,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_mnist_fp16_allreduce.py b/python/paddle/fluid/tests/unittests/dist_mnist_fp16_allreduce.py
index 626233b1f0..53819ca549 100644
--- a/python/paddle/fluid/tests/unittests/dist_mnist_fp16_allreduce.py
+++ b/python/paddle/fluid/tests/unittests/dist_mnist_fp16_allreduce.py
@@ -42,7 +42,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_mnist_lars.py b/python/paddle/fluid/tests/unittests/dist_mnist_lars.py
index 4466b82cb4..347692afdd 100644
--- a/python/paddle/fluid/tests/unittests/dist_mnist_lars.py
+++ b/python/paddle/fluid/tests/unittests/dist_mnist_lars.py
@@ -39,7 +39,7 @@ class TestDistMnist2x2(TestDistRunnerBase):
 
         # Evaluator
         batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-        batch_acc = fluid.layers.accuracy(
+        batch_acc = paddle.static.accuracy(
             input=predict, label=label, total=batch_size_tensor
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dist_se_resnext.py b/python/paddle/fluid/tests/unittests/dist_se_resnext.py
index 3461be8228..05b3f3b093 100644
--- a/python/paddle/fluid/tests/unittests/dist_se_resnext.py
+++ b/python/paddle/fluid/tests/unittests/dist_se_resnext.py
@@ -224,8 +224,8 @@ class DistSeResneXt2x2(TestDistRunnerBase):
         cost = fluid.layers.cross_entropy(input=out, label=label)
 
         avg_cost = paddle.mean(x=cost)
-        acc_top1 = fluid.layers.accuracy(input=out, label=label, k=1)
-        acc_top5 = fluid.layers.accuracy(input=out, label=label, k=5)
+        acc_top1 = paddle.static.accuracy(input=out, label=label, k=1)
+        acc_top5 = paddle.static.accuracy(input=out, label=label, k=5)
 
         # Evaluator
         test_program = fluid.default_main_program().clone(for_test=True)
diff --git a/python/paddle/fluid/tests/unittests/dist_text_classification.py b/python/paddle/fluid/tests/unittests/dist_text_classification.py
index 3312974657..73e153164b 100644
--- a/python/paddle/fluid/tests/unittests/dist_text_classification.py
+++ b/python/paddle/fluid/tests/unittests/dist_text_classification.py
@@ -134,7 +134,7 @@ class TestDistTextClassification2x2(TestDistRunnerBase):
         predict = conv_net(data, dict_dim)
         cost = fluid.layers.cross_entropy(input=predict, label=label)
         avg_cost = paddle.mean(x=cost)
-        acc = fluid.layers.accuracy(input=predict, label=label)
+        acc = paddle.static.accuracy(input=predict, label=label)
         inference_program = fluid.default_main_program().clone()
 
         # Optimization
diff --git a/python/paddle/fluid/tests/unittests/distribution/test_distribution.py b/python/paddle/fluid/tests/unittests/distribution/test_distribution.py
index e3b6276529..bbe564eef3 100644
--- a/python/paddle/fluid/tests/unittests/distribution/test_distribution.py
+++ b/python/paddle/fluid/tests/unittests/distribution/test_distribution.py
@@ -12,18 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import unittest
-
-import config
-import numpy as np
-import parameterize
-
-import paddle
-from paddle.distribution import Categorical, Normal, Uniform
-from paddle.fluid import layers
-
-paddle.enable_static()
-
 
 class DistributionNumpy:
     def sample(self):
@@ -40,153 +28,3 @@ class DistributionNumpy:
 
     def probs(self, value):
         raise NotImplementedError
-
-
-class DistributionTestName(unittest.TestCase):
-    def get_prefix(self, string):
-        return string.split('.')[0]
-
-    def test_normal_name(self):
-        name = 'test_normal'
-        normal1 = Normal(0.0, 1.0, name=name)
-        self.assertEqual(normal1.name, name)
-
-        normal2 = Normal(0.0, 1.0)
-        self.assertEqual(normal2.name, 'Normal')
-
-        paddle.enable_static()
-
-        sample = normal1.sample([2])
-        self.assertEqual(self.get_prefix(sample.name), name + '_sample')
-
-        entropy = normal1.entropy()
-        self.assertEqual(self.get_prefix(entropy.name), name + '_entropy')
-
-        value_npdata = np.array([0.8], dtype="float32")
-        value_tensor = layers.create_tensor(dtype="float32")
-        layers.assign(value_npdata, value_tensor)
-
-        lp = normal1.log_prob(value_tensor)
-        self.assertEqual(self.get_prefix(lp.name), name + '_log_prob')
-
-        p = normal1.probs(value_tensor)
-        self.assertEqual(self.get_prefix(p.name), name + '_probs')
-
-        kl = normal1.kl_divergence(normal2)
-        self.assertEqual(self.get_prefix(kl.name), name + '_kl_divergence')
-
-    def test_uniform_name(self):
-        name = 'test_uniform'
-        uniform1 = Uniform(0.0, 1.0, name=name)
-        self.assertEqual(uniform1.name, name)
-
-        uniform2 = Uniform(0.0, 1.0)
-        self.assertEqual(uniform2.name, 'Uniform')
-
-        paddle.enable_static()
-
-        sample = uniform1.sample([2])
-        self.assertEqual(self.get_prefix(sample.name), name + '_sample')
-
-        entropy = uniform1.entropy()
-        self.assertEqual(self.get_prefix(entropy.name), name + '_entropy')
-
-        value_npdata = np.array([0.8], dtype="float32")
-        value_tensor = layers.create_tensor(dtype="float32")
-        layers.assign(value_npdata, value_tensor)
-
-        lp = uniform1.log_prob(value_tensor)
-        self.assertEqual(self.get_prefix(lp.name), name + '_log_prob')
-
-        p = uniform1.probs(value_tensor)
-        self.assertEqual(self.get_prefix(p.name), name + '_probs')
-
-    def test_categorical_name(self):
-        name = 'test_categorical'
-        categorical1 = Categorical([0.4, 0.6], name=name)
-        self.assertEqual(categorical1.name, name)
-
-        categorical2 = Categorical([0.5, 0.5])
-        self.assertEqual(categorical2.name, 'Categorical')
-
-        paddle.enable_static()
-
-        sample = categorical1.sample([2])
-        self.assertEqual(self.get_prefix(sample.name), name + '_sample')
-
-        entropy = categorical1.entropy()
-        self.assertEqual(self.get_prefix(entropy.name), name + '_entropy')
-
-        kl = categorical1.kl_divergence(categorical2)
-        self.assertEqual(self.get_prefix(kl.name), name + '_kl_divergence')
-
-        value_npdata = np.array([0], dtype="int64")
-        value_tensor = layers.create_tensor(dtype="int64")
-        layers.assign(value_npdata, value_tensor)
-
-        p = categorical1.probs(value_tensor)
-        self.assertEqual(self.get_prefix(p.name), name + '_probs')
-
-        lp = categorical1.log_prob(value_tensor)
-        self.assertEqual(self.get_prefix(lp.name), name + '_log_prob')
-
-
-@parameterize.place(config.DEVICES)
-@parameterize.parameterize_cls(
-    (parameterize.TEST_CASE_NAME, 'batch_shape', 'event_shape'),
-    [
-        ('test-tuple', (10, 20), (10, 20)),
-        ('test-list', [100, 100], [100, 200, 300]),
-        ('test-null-eventshape', (100, 100), ()),
-    ],
-)
-class TestDistributionShape(unittest.TestCase):
-    def setUp(self):
-        paddle.disable_static()
-        self.dist = paddle.distribution.Distribution(
-            batch_shape=self.batch_shape, event_shape=self.event_shape
-        )
-
-    def tearDown(self):
-        paddle.enable_static()
-
-    def test_batch_shape(self):
-        self.assertTrue(isinstance(self.dist.batch_shape, tuple))
-        self.assertTrue(self.dist.batch_shape == tuple(self.batch_shape))
-
-    def test_event_shape(self):
-        self.assertTrue(isinstance(self.dist.event_shape, tuple))
-        self.assertTrue(self.dist.event_shape == tuple(self.event_shape))
-
-    def test_prob(self):
-        with self.assertRaises(NotImplementedError):
-            self.dist.prob(paddle.to_tensor(parameterize.xrand()))
-
-    def test_extend_shape(self):
-        shapes = [(34, 20), (56,), ()]
-        for shape in shapes:
-            self.assertTrue(
-                self.dist._extend_shape(shape),
-                shape + self.dist.batch_shape + self.dist.event_shape,
-            )
-
-
-class TestDistributionException(unittest.TestCase):
-    def setUp(self):
-        self._d = paddle.distribution.Distribution()
-
-    def test_mean(self):
-        with self.assertRaises(NotImplementedError):
-            self._d.mean
-
-    def test_variance(self):
-        with self.assertRaises(NotImplementedError):
-            self._d.variance
-
-    def test_rsample(self):
-        with self.assertRaises(NotImplementedError):
-            self._d.rsample(())
-
-
-if __name__ == '__main__':
-    unittest.main()
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/bert_dygraph_model.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/bert_dygraph_model.py
index 3e372a6d94..64d0a8dc73 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/bert_dygraph_model.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/bert_dygraph_model.py
@@ -424,7 +424,7 @@ class PretrainModelLayer(Layer):
             logits=next_sent_fc_out, label=labels, return_softmax=True
         )
 
-        next_sent_acc = fluid.layers.accuracy(
+        next_sent_acc = paddle.static.accuracy(
             input=next_sent_softmax, label=labels
         )
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mnist.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mnist.py
index 43947a505e..93f01d165f 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mnist.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mnist.py
@@ -110,7 +110,7 @@ class MNIST(fluid.dygraph.Layer):
     def forward(self, inputs, label=None):
         x = self.inference(inputs)
         if label is not None:
-            acc = fluid.layers.accuracy(input=x, label=label)
+            acc = paddle.static.accuracy(input=x, label=label)
             loss = fluid.layers.cross_entropy(x, label)
             avg_loss = paddle.mean(loss)
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mobile_net.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mobile_net.py
index 2c458b006d..4c5e306718 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mobile_net.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_mobile_net.py
@@ -540,8 +540,8 @@ def train_mobilenet(args, to_static):
                     input=softmax_out, label=label
                 )
                 avg_loss = paddle.mean(x=loss)
-                acc_top1 = fluid.layers.accuracy(input=out, label=label, k=1)
-                acc_top5 = fluid.layers.accuracy(input=out, label=label, k=5)
+                acc_top1 = paddle.static.accuracy(input=out, label=label, k=1)
+                acc_top5 = paddle.static.accuracy(input=out, label=label, k=5)
                 t_start_back = time.time()
 
                 loss_data.append(avg_loss.numpy())
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet.py
index 00d6f24da3..6323ba5ee7 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet.py
@@ -276,10 +276,10 @@ class ResNetHelper:
                     pred = resnet(img)
                     loss = fluid.layers.cross_entropy(input=pred, label=label)
                     avg_loss = paddle.mean(x=loss)
-                    acc_top1 = fluid.layers.accuracy(
+                    acc_top1 = paddle.static.accuracy(
                         input=pred, label=label, k=1
                     )
-                    acc_top5 = fluid.layers.accuracy(
+                    acc_top5 = paddle.static.accuracy(
                         input=pred, label=label, k=5
                     )
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_amp.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_amp.py
index ae7368ac98..4ce379a704 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_amp.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_amp.py
@@ -76,8 +76,8 @@ def train(to_static, build_strategy=None):
                     # If we remove it, the loss between dygraph and dy2stat is exactly same.
                     loss = fluid.layers.cross_entropy(input=pred, label=label)
                 avg_loss = paddle.mean(x=pred)
-                acc_top1 = fluid.layers.accuracy(input=pred, label=label, k=1)
-                acc_top5 = fluid.layers.accuracy(input=pred, label=label, k=5)
+                acc_top1 = paddle.static.accuracy(input=pred, label=label, k=1)
+                acc_top5 = paddle.static.accuracy(input=pred, label=label, k=5)
 
                 scaled = scaler.scale(avg_loss)
                 scaled.backward()
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_pure_fp16.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_pure_fp16.py
index 058be1b07a..21d081b009 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_pure_fp16.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_resnet_pure_fp16.py
@@ -77,8 +77,8 @@ def train(to_static, build_strategy=None):
                 pred = resnet(img)
                 loss = fluid.layers.cross_entropy(input=pred, label=label)
             avg_loss = paddle.mean(x=pred)
-            acc_top1 = fluid.layers.accuracy(input=pred, label=label, k=1)
-            acc_top5 = fluid.layers.accuracy(input=pred, label=label, k=5)
+            acc_top1 = paddle.static.accuracy(input=pred, label=label, k=1)
+            acc_top5 = paddle.static.accuracy(input=pred, label=label, k=5)
 
             scaled = scaler.scale(avg_loss)
             scaled.backward()
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_se_resnet.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_se_resnet.py
index 269f50a8dd..109fc99754 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_se_resnet.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_se_resnet.py
@@ -347,8 +347,8 @@ class SeResNeXt(fluid.dygraph.Layer):
         loss = fluid.layers.cross_entropy(input=softmax_out, label=label)
         avg_loss = paddle.mean(x=loss)
 
-        acc_top1 = fluid.layers.accuracy(input=softmax_out, label=label, k=1)
-        acc_top5 = fluid.layers.accuracy(input=softmax_out, label=label, k=5)
+        acc_top1 = paddle.static.accuracy(input=softmax_out, label=label, k=1)
+        acc_top5 = paddle.static.accuracy(input=softmax_out, label=label, k=5)
         return out, avg_loss, acc_top1, acc_top5
 
 
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_sentiment.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_sentiment.py
index ef9436864a..c644cdfe0c 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_sentiment.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_sentiment.py
@@ -109,7 +109,7 @@ class CNN(fluid.dygraph.Layer):
 
         cost = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_cost = paddle.mean(x=cost)
-        acc = fluid.layers.accuracy(input=prediction, label=label)
+        acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
 
 
@@ -152,7 +152,7 @@ class BOW(fluid.dygraph.Layer):
 
         cost = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_cost = paddle.mean(x=cost)
-        acc = fluid.layers.accuracy(input=prediction, label=label)
+        acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
 
 
@@ -198,7 +198,7 @@ class GRU(fluid.dygraph.Layer):
         prediction = paddle.nn.functional.softmax(prediction)
         cost = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_cost = paddle.mean(x=cost)
-        acc = fluid.layers.accuracy(input=prediction, label=label)
+        acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
 
 
@@ -257,7 +257,7 @@ class BiGRU(fluid.dygraph.Layer):
         # if label is not None:
         cost = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_cost = paddle.mean(x=cost)
-        acc = fluid.layers.accuracy(input=prediction, label=label)
+        acc = paddle.static.accuracy(input=prediction, label=label)
         return avg_cost, prediction, acc
         # else:
         #     return prediction
diff --git a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tsm.py b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tsm.py
index 9c52786842..7cd74f3025 100644
--- a/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tsm.py
+++ b/python/paddle/fluid/tests/unittests/dygraph_to_static/test_tsm.py
@@ -334,10 +334,10 @@ def train(args, fake_data_reader, to_static):
                     input=outputs, label=labels, ignore_index=-1
                 )
                 avg_loss = paddle.mean(loss)
-                acc_top1 = fluid.layers.accuracy(
+                acc_top1 = paddle.static.accuracy(
                     input=outputs, label=labels, k=1
                 )
-                acc_top5 = fluid.layers.accuracy(
+                acc_top5 = paddle.static.accuracy(
                     input=outputs, label=labels, k=5
                 )
 
diff --git a/python/paddle/fluid/tests/unittests/mlu/test_accuracy_op_mlu.py b/python/paddle/fluid/tests/unittests/mlu/test_accuracy_op_mlu.py
index e37be2f9f3..a32765b8d3 100755
--- a/python/paddle/fluid/tests/unittests/mlu/test_accuracy_op_mlu.py
+++ b/python/paddle/fluid/tests/unittests/mlu/test_accuracy_op_mlu.py
@@ -74,14 +74,14 @@ class TestAccuracyOpError(unittest.TestCase):
             label = fluid.layers.data(
                 name='label', shape=[-1, 1], dtype="int32"
             )
-            self.assertRaises(TypeError, fluid.layers.accuracy, x1, label)
+            self.assertRaises(TypeError, paddle.static.accuracy, x1, label)
             self.assertRaises(TypeError, paddle.metric.accuracy, x1, label)
             # The input dtype of accuracy_op must be float32 or float64.
             x2 = fluid.layers.data(name='x2', shape=[4], dtype="int32")
-            self.assertRaises(TypeError, fluid.layers.accuracy, x2, label)
+            self.assertRaises(TypeError, paddle.static.accuracy, x2, label)
             self.assertRaises(TypeError, paddle.metric.accuracy, x2, label)
             x3 = fluid.layers.data(name='input', shape=[-1, 2], dtype="float16")
-            fluid.layers.accuracy(input=x3, label=label)
+            paddle.static.accuracy(input=x3, label=label)
             paddle.metric.accuracy(input=x3, label=label)
 
 
diff --git a/python/paddle/fluid/tests/unittests/test_accuracy_op.py b/python/paddle/fluid/tests/unittests/test_accuracy_op.py
index fc7666bcba..446240e9fb 100755
--- a/python/paddle/fluid/tests/unittests/test_accuracy_op.py
+++ b/python/paddle/fluid/tests/unittests/test_accuracy_op.py
@@ -69,14 +69,14 @@ class TestAccuracyOpError(unittest.TestCase):
             label = fluid.layers.data(
                 name='label', shape=[-1, 1], dtype="int32"
             )
-            self.assertRaises(TypeError, fluid.layers.accuracy, x1, label)
+            self.assertRaises(TypeError, paddle.static.accuracy, x1, label)
             self.assertRaises(TypeError, paddle.metric.accuracy, x1, label)
             # The input dtype of accuracy_op must be float32 or float64.
             x2 = fluid.layers.data(name='x2', shape=[4], dtype="int32")
-            self.assertRaises(TypeError, fluid.layers.accuracy, x2, label)
+            self.assertRaises(TypeError, paddle.static.accuracy, x2, label)
             self.assertRaises(TypeError, paddle.metric.accuracy, x2, label)
             x3 = fluid.layers.data(name='input', shape=[-1, 2], dtype="float16")
-            fluid.layers.accuracy(input=x3, label=label)
+            paddle.static.accuracy(input=x3, label=label)
             paddle.metric.accuracy(input=x3, label=label)
 
 
diff --git a/python/paddle/fluid/tests/unittests/test_async_ssa_graph_executor_mnist.py b/python/paddle/fluid/tests/unittests/test_async_ssa_graph_executor_mnist.py
index 9caae0bea6..26eb0a628a 100644
--- a/python/paddle/fluid/tests/unittests/test_async_ssa_graph_executor_mnist.py
+++ b/python/paddle/fluid/tests/unittests/test_async_ssa_graph_executor_mnist.py
@@ -59,7 +59,7 @@ def convolutional_neural_network(use_py_reader):
         prediction = fluid.layers.fc(input=conv_pool_2, size=10, act='softmax')
         loss = fluid.layers.cross_entropy(input=prediction, label=label)
         avg_loss = paddle.mean(loss)
-        acc = fluid.layers.accuracy(input=prediction, label=label)
+        acc = paddle.static.accuracy(input=prediction, label=label)
         i = fluid.layers.zeros(shape=[1], dtype='int64')
         array = fluid.layers.array_write(x=prediction, i=i)
         fluid.layers.increment(i)
diff --git a/python/paddle/fluid/tests/unittests/test_auc_op.py b/python/paddle/fluid/tests/unittests/test_auc_op.py
index 4a693500bc..48fe682b14 100644
--- a/python/paddle/fluid/tests/unittests/test_auc_op.py
+++ b/python/paddle/fluid/tests/unittests/test_auc_op.py
@@ -158,7 +158,7 @@ class TestAucOpError(unittest.TestCase):
                     name="input2", shape=[-1, 2], dtype="float32"
                 )
                 label2 = fluid.data(name="label2", shape=[-1], dtype="float32")
-                result2 = fluid.layers.auc(input=data2, label=label2)
+                result2 = paddle.static.auc(input=data2, label=label2)
 
             self.assertRaises(TypeError, test_type2)
 
diff --git a/python/paddle/fluid/tests/unittests/test_desc_clone.py b/python/paddle/fluid/tests/unittests/test_desc_clone.py
index e2a90cb544..7d4f1f0975 100644
--- a/python/paddle/fluid/tests/unittests/test_desc_clone.py
+++ b/python/paddle/fluid/tests/unittests/test_desc_clone.py
@@ -78,7 +78,7 @@ def get_model(batch_size):
 
     # Evaluator
     batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
-    batch_acc = fluid.layers.accuracy(
+    batch_acc = paddle.static.accuracy(
         input=predict, label=label, total=batch_size_tensor
     )
 
diff --git a/python/paddle/fluid/tests/unittests/test_distributions.py b/python/paddle/fluid/tests/unittests/test_distributions.py
index df8fe980e8..4656c20811 100644
--- a/python/paddle/fluid/tests/unittests/test_distributions.py
+++ b/python/paddle/fluid/tests/unittests/test_distributions.py
@@ -12,20 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import math
-import unittest
-
-import numpy as np
-
-from paddle import fluid
-from paddle.fluid import layers
-from paddle.fluid.layers.distributions import (
-    Categorical,
-    MultivariateNormalDiag,
-    Normal,
-    Uniform,
-)
-
 
 class DistributionNumpy:
     """
@@ -47,788 +33,3 @@ class DistributionNumpy:
     def log_prob(self, value):
         """Log probability density/mass function."""
         raise NotImplementedError
-
-
-class UniformNumpy(DistributionNumpy):
-    def __init__(self, low, high):
-        self.low = np.array(low).astype('float32')
-        self.high = np.array(high).astype('float32')
-
-    def sample(self, shape):
-        shape = tuple(shape) + (self.low + self.high).shape
-        return self.low + (
-            np.random.uniform(size=shape) * (self.high - self.low)
-        )
-
-    def log_prob(self, value):
-        lb = np.less(self.low, value).astype('float32')
-        ub = np.less(value, self.high).astype('float32')
-        return np.log(lb * ub) - np.log(self.high - self.low)
-
-    def entropy(self):
-        return np.log(self.high - self.low)
-
-
-class NormalNumpy(DistributionNumpy):
-    def __init__(self, loc, scale):
-        self.loc = np.array(loc).astype('float32')
-        self.scale = np.array(scale).astype('float32')
-
-    def sample(self, shape):
-        shape = tuple(shape) + (self.loc + self.scale).shape
-        return self.loc + (np.random.randn(*shape) * self.scale)
-
-    def log_prob(self, value):
-        var = self.scale * self.scale
-        log_scale = np.log(self.scale)
-        return (
-            -((value - self.loc) * (value - self.loc)) / (2.0 * var)
-            - log_scale
-            - math.log(math.sqrt(2.0 * math.pi))
-        )
-
-    def entropy(self):
-        return (
-            0.5
-            + 0.5 * np.log(np.array(2.0 * math.pi).astype('float32'))
-            + np.log(self.scale)
-        )
-
-    def kl_divergence(self, other):
-        var_ratio = self.scale / other.scale
-        var_ratio = var_ratio * var_ratio
-        t1 = (self.loc - other.loc) / other.scale
-        t1 = t1 * t1
-        return 0.5 * (var_ratio + t1 - 1 - np.log(var_ratio))
-
-
-class CategoricalNumpy(DistributionNumpy):
-    def __init__(self, logits):
-        self.logits = np.array(logits).astype('float32')
-
-    def entropy(self):
-        logits = self.logits - np.max(self.logits, axis=-1, keepdims=True)
-        e_logits = np.exp(logits)
-        z = np.sum(e_logits, axis=-1, keepdims=True)
-        prob = e_logits / z
-        return -1.0 * np.sum(
-            prob * (logits - np.log(z)), axis=-1, keepdims=True
-        )
-
-    def kl_divergence(self, other):
-        logits = self.logits - np.max(self.logits, axis=-1, keepdims=True)
-        other_logits = other.logits - np.max(
-            other.logits, axis=-1, keepdims=True
-        )
-        e_logits = np.exp(logits)
-        other_e_logits = np.exp(other_logits)
-        z = np.sum(e_logits, axis=-1, keepdims=True)
-        other_z = np.sum(other_e_logits, axis=-1, keepdims=True)
-        prob = e_logits / z
-        return np.sum(
-            prob * (logits - np.log(z) - other_logits + np.log(other_z)),
-            axis=-1,
-            keepdims=True,
-        )
-
-
-class MultivariateNormalDiagNumpy(DistributionNumpy):
-    def __init__(self, loc, scale):
-        self.loc = np.array(loc).astype('float32')
-        self.scale = np.array(scale).astype('float32')
-
-    def _det(self, value):
-        batch_shape = list(value.shape)
-        one_all = np.ones(shape=batch_shape, dtype='float32')
-        one_diag = np.eye(batch_shape[0], dtype='float32')
-        det_diag = np.prod(value + one_all - one_diag)
-
-        return det_diag
-
-    def _inv(self, value):
-        batch_shape = list(value.shape)
-        one_all = np.ones(shape=batch_shape, dtype='float32')
-        one_diag = np.eye(batch_shape[0], dtype='float32')
-        inv_diag = np.power(value, (one_all - 2 * one_diag))
-
-        return inv_diag
-
-    def entropy(self):
-        return 0.5 * (
-            self.scale.shape[0]
-            * (1.0 + np.log(np.array(2 * math.pi).astype('float32')))
-            + np.log(self._det(self.scale))
-        )
-
-    def kl_divergence(self, other):
-        tr_cov_matmul = np.sum(self._inv(other.scale) * self.scale)
-        loc_matmul_cov = np.matmul(
-            (other.loc - self.loc), self._inv(other.scale)
-        )
-        tri_matmul = np.matmul(loc_matmul_cov, (other.loc - self.loc))
-        k = list(self.scale.shape)[0]
-        ln_cov = np.log(self._det(other.scale)) - np.log(self._det(self.scale))
-        kl = 0.5 * (tr_cov_matmul + tri_matmul - k + ln_cov)
-
-        return kl
-
-
-class DistributionTest(unittest.TestCase):
-    def setUp(self, use_gpu=False):
-        self.use_gpu = use_gpu
-        if not use_gpu:
-            place = fluid.CPUPlace()
-            self.gpu_id = -1
-        else:
-            place = fluid.CUDAPlace(0)
-            self.gpu_id = 0
-        self.executor = fluid.Executor(place)
-
-    def build_normal_program(
-        self,
-        test_program,
-        batch_size,
-        dims,
-        loc_float,
-        scale_float,
-        other_loc_float,
-        other_scale_float,
-        scale_np,
-        other_scale_np,
-        loc_np,
-        other_loc_np,
-        values_np,
-    ):
-        with fluid.program_guard(test_program):
-            loc = layers.data(name='loc', shape=[dims], dtype='float32')
-            scale = layers.data(name='scale', shape=[dims], dtype='float32')
-
-            other_loc = layers.data(
-                name='other_loc', shape=[dims], dtype='float32'
-            )
-            other_scale = layers.data(
-                name='other_scale', shape=[dims], dtype='float32'
-            )
-
-            values = layers.data(name='values', shape=[dims], dtype='float32')
-
-            normal_float = Normal(loc_float, scale_float)
-            other_normal_float = Normal(other_loc_float, other_scale_float)
-
-            normal_float_np_broadcast = Normal(loc_float, scale_np)
-            other_normal_float_np_broadcast = Normal(
-                other_loc_float, other_scale_np
-            )
-
-            normal_np = Normal(loc_np, scale_np)
-            other_normal_np = Normal(other_loc_np, other_scale_np)
-
-            normal_variable = Normal(loc, scale)
-            other_normal_variable = Normal(other_loc, other_scale)
-
-            sample_float = normal_float.sample([batch_size, dims])
-            sample_float_np_broadcast = normal_float_np_broadcast.sample(
-                [batch_size, dims]
-            )
-            sample_np = normal_np.sample([batch_size, dims])
-            sample_variable = normal_variable.sample([batch_size, dims])
-
-            entropy_float = normal_float.entropy()
-            entropy_float_np_broadcast = normal_float_np_broadcast.entropy()
-            entropy_np = normal_np.entropy()
-            entropy_variable = normal_variable.entropy()
-
-            lp_float_np_broadcast = normal_float_np_broadcast.log_prob(values)
-            lp_np = normal_np.log_prob(values)
-            lp_variable = normal_variable.log_prob(values)
-
-            kl_float = normal_float.kl_divergence(other_normal_float)
-            kl_float_np_broadcast = normal_float_np_broadcast.kl_divergence(
-                other_normal_float_np_broadcast
-            )
-            kl_np = normal_np.kl_divergence(other_normal_np)
-            kl_variable = normal_variable.kl_divergence(other_normal_variable)
-
-        fetch_list = [
-            sample_float,
-            sample_float_np_broadcast,
-            sample_np,
-            sample_variable,
-            entropy_float,
-            entropy_float_np_broadcast,
-            entropy_np,
-            entropy_variable,
-            lp_float_np_broadcast,
-            lp_np,
-            lp_variable,
-            kl_float,
-            kl_float_np_broadcast,
-            kl_np,
-            kl_variable,
-        ]
-        feed_vars = {
-            'loc': loc_np,
-            'scale': scale_np,
-            'other_loc': other_loc_np,
-            'other_scale': other_scale_np,
-            'values': values_np,
-        }
-        return feed_vars, fetch_list
-
-    def get_normal_random_input(self, batch_size, dims):
-        loc_np = np.random.randn(batch_size, dims).astype('float32')
-        other_loc_np = np.random.randn(batch_size, dims).astype('float32')
-
-        loc_float = (np.random.ranf() - 0.5) * 4
-        scale_float = (np.random.ranf() - 0.5) * 4
-        while scale_float < 0:
-            scale_float = (np.random.ranf() - 0.5) * 4
-
-        other_loc_float = (np.random.ranf() - 0.5) * 4
-        other_scale_float = (np.random.ranf() - 0.5) * 4
-        while other_scale_float < 0:
-            other_scale_float = (np.random.ranf() - 0.5) * 4
-
-        scale_np = np.random.randn(batch_size, dims).astype('float32')
-        other_scale_np = np.random.randn(batch_size, dims).astype('float32')
-        values_np = np.random.randn(batch_size, dims).astype('float32')
-
-        while not np.all(scale_np > 0):
-            scale_np = np.random.randn(batch_size, dims).astype('float32')
-        while not np.all(other_scale_np > 0):
-            other_scale_np = np.random.randn(batch_size, dims).astype('float32')
-        return (
-            loc_np,
-            other_loc_np,
-            loc_float,
-            scale_float,
-            other_loc_float,
-            other_scale_float,
-            scale_np,
-            other_scale_np,
-            values_np,
-        )
-
-    def test_normal_distribution(self, batch_size=2, dims=3, tolerance=1e-6):
-        test_program = fluid.Program()
-        (
-            loc_np,
-            other_loc_np,
-            loc_float,
-            scale_float,
-            other_loc_float,
-            other_scale_float,
-            scale_np,
-            other_scale_np,
-            values_np,
-        ) = self.get_normal_random_input(batch_size, dims)
-
-        feed_vars, fetch_list = self.build_normal_program(
-            test_program,
-            batch_size,
-            dims,
-            loc_float,
-            scale_float,
-            other_loc_float,
-            other_scale_float,
-            scale_np,
-            other_scale_np,
-            loc_np,
-            other_loc_np,
-            values_np,
-        )
-        self.executor.run(fluid.default_startup_program())
-
-        np_normal_float = NormalNumpy(loc_float, scale_float)
-        np_other_normal_float = NormalNumpy(other_loc_float, other_scale_float)
-        np_normal_float_np_broadcast = NormalNumpy(loc_float, scale_np)
-        np_other_normal_float_np_broadcast = NormalNumpy(
-            other_loc_float, other_scale_np
-        )
-        np_normal = NormalNumpy(loc_np, scale_np)
-        np_other_normal = NormalNumpy(other_loc_np, other_scale_np)
-
-        gt_sample_float = np_normal_float.sample([batch_size, dims])
-        gt_sample_float_np_broadcast = np_normal_float_np_broadcast.sample(
-            [batch_size, dims]
-        )
-        gt_sample_np = np_normal.sample([batch_size, dims])
-        gt_entropy_float = np_normal_float.entropy()
-        gt_entropy_float_np_broadcast = np_normal_float_np_broadcast.entropy()
-        gt_entropy = np_normal.entropy()
-        gt_lp_float_np_broadcast = np_normal_float_np_broadcast.log_prob(
-            values_np
-        )
-        gt_lp = np_normal.log_prob(values_np)
-        gt_kl_float = np_normal_float.kl_divergence(np_other_normal_float)
-        gt_kl_float_np_broadcast = np_normal_float_np_broadcast.kl_divergence(
-            np_other_normal_float_np_broadcast
-        )
-        gt_kl = np_normal.kl_divergence(np_other_normal)
-
-        [
-            output_sample_float,
-            output_sample_float_np_broadcast,
-            output_sample_np,
-            output_sample_variable,
-            output_entropy_float,
-            output_entropy_float_np_broadcast,
-            output_entropy_np,
-            output_entropy_variable,
-            output_lp_float_np_broadcast,
-            output_lp_np,
-            output_lp_variable,
-            output_kl_float,
-            output_kl_float_np_broadcast,
-            output_kl_np,
-            output_kl_variable,
-        ] = self.executor.run(
-            program=test_program, feed=feed_vars, fetch_list=fetch_list
-        )
-
-        np.testing.assert_allclose(
-            output_sample_float.shape,
-            gt_sample_float.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_float_np_broadcast.shape,
-            gt_sample_float_np_broadcast.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_np.shape,
-            gt_sample_np.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_variable.shape,
-            gt_sample_np.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_float,
-            gt_entropy_float,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_float_np_broadcast,
-            gt_entropy_float_np_broadcast,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_np, gt_entropy, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_entropy_variable, gt_entropy, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_lp_float_np_broadcast,
-            gt_lp_float_np_broadcast,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_lp_np, gt_lp, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_lp_variable, gt_lp, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_kl_float, gt_kl_float, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_kl_float_np_broadcast,
-            gt_kl_float_np_broadcast,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_kl_np, gt_kl, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_kl_variable, gt_kl, rtol=tolerance, atol=tolerance
-        )
-
-    def build_uniform_program(
-        self,
-        test_program,
-        batch_size,
-        dims,
-        low_float,
-        high_float,
-        high_np,
-        low_np,
-        values_np,
-    ):
-        with fluid.program_guard(test_program):
-            low = layers.data(name='low', shape=[dims], dtype='float32')
-            high = layers.data(name='high', shape=[dims], dtype='float32')
-
-            values = layers.data(name='values', shape=[dims], dtype='float32')
-
-            uniform_float = Uniform(low_float, high_float)
-            uniform_float_np_broadcast = Uniform(low_float, high_np)
-            uniform_np = Uniform(low_np, high_np)
-            uniform_variable = Uniform(low, high)
-
-            sample_float = uniform_float.sample([batch_size, dims])
-            sample_float_np_broadcast = uniform_float_np_broadcast.sample(
-                [batch_size, dims]
-            )
-            sample_np = uniform_np.sample([batch_size, dims])
-            sample_variable = uniform_variable.sample([batch_size, dims])
-
-            entropy_float = uniform_float.entropy()
-            entropy_float_np_broadcast = uniform_float_np_broadcast.entropy()
-            entropy_np = uniform_np.entropy()
-            entropy_variable = uniform_variable.entropy()
-
-            lp_float_np_broadcast = uniform_float_np_broadcast.log_prob(values)
-            lp_np = uniform_np.log_prob(values)
-            lp_variable = uniform_variable.log_prob(values)
-
-        fetch_list = [
-            sample_float,
-            sample_float_np_broadcast,
-            sample_np,
-            sample_variable,
-            entropy_float,
-            entropy_float_np_broadcast,
-            entropy_np,
-            entropy_variable,
-            lp_float_np_broadcast,
-            lp_np,
-            lp_variable,
-        ]
-        feed_vars = {'low': low_np, 'high': high_np, 'values': values_np}
-        return feed_vars, fetch_list
-
-    def test_uniform_distribution(self, batch_size=2, dims=3, tolerance=1e-6):
-        test_program = fluid.Program()
-
-        low_np = np.random.randn(batch_size, dims).astype('float32')
-        low_float = np.random.uniform(-2, 1)
-        high_float = np.random.uniform(1, 3)
-        high_np = np.random.uniform(-5.0, 5.0, (batch_size, dims)).astype(
-            'float32'
-        )
-        values_np = np.random.randn(batch_size, dims).astype('float32')
-
-        feed_vars, fetch_list = self.build_uniform_program(
-            test_program,
-            batch_size,
-            dims,
-            low_float,
-            high_float,
-            high_np,
-            low_np,
-            values_np,
-        )
-
-        self.executor.run(fluid.default_startup_program())
-
-        np_uniform_float = UniformNumpy(low_float, high_float)
-        np_uniform_float_np_broadcast = UniformNumpy(low_float, high_np)
-        np_uniform = UniformNumpy(low_np, high_np)
-
-        gt_sample_float = np_uniform_float.sample([batch_size, dims])
-        gt_sample_float_np_broadcast = np_uniform_float_np_broadcast.sample(
-            [batch_size, dims]
-        )
-        gt_sample_np = np_uniform.sample([batch_size, dims])
-        gt_entropy_float = np_uniform_float.entropy()
-        gt_entropy_float_np_broadcast = np_uniform_float_np_broadcast.entropy()
-        gt_entropy = np_uniform.entropy()
-        gt_lp_float_np_broadcast = np_uniform_float_np_broadcast.log_prob(
-            values_np
-        )
-        gt_lp = np_uniform.log_prob(values_np)
-
-        # result calculated by paddle
-        [
-            output_sample_float,
-            output_sample_float_np_broadcast,
-            output_sample_np,
-            output_sample_variable,
-            output_entropy_float,
-            output_entropy_float_np_broadcast,
-            output_entropy_np,
-            output_entropy_variable,
-            output_lp_float_np_broadcast,
-            output_lp_np,
-            output_lp_variable,
-        ] = self.executor.run(
-            program=test_program, feed=feed_vars, fetch_list=fetch_list
-        )
-
-        np.testing.assert_allclose(
-            output_sample_float.shape,
-            gt_sample_float.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_float_np_broadcast.shape,
-            gt_sample_float_np_broadcast.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_np.shape,
-            gt_sample_np.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_sample_variable.shape,
-            gt_sample_np.shape,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_float,
-            gt_entropy_float,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_float_np_broadcast,
-            gt_entropy_float_np_broadcast,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_entropy_np, gt_entropy, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_entropy_variable, gt_entropy, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_lp_float_np_broadcast,
-            gt_lp_float_np_broadcast,
-            rtol=tolerance,
-            atol=tolerance,
-        )
-        np.testing.assert_allclose(
-            output_lp_np, gt_lp, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_lp_variable, gt_lp, rtol=tolerance, atol=tolerance
-        )
-
-    def test_categorical_distribution(
-        self, batch_size=2, dims=3, tolerance=1e-6
-    ):
-        test_program = fluid.Program()
-
-        logits_np = np.random.randn(batch_size, dims).astype('float32')
-        other_logits_np = np.random.randn(batch_size, dims).astype('float32')
-
-        with fluid.program_guard(test_program):
-            logits = layers.data(name='logits', shape=[dims], dtype='float32')
-            other_logits = layers.data(
-                name='other_logits', shape=[dims], dtype='float32'
-            )
-
-            categorical_np = Categorical(logits_np)
-            other_categorical_np = Categorical(other_logits_np)
-
-            entropy_np = categorical_np.entropy()
-            kl_np = categorical_np.kl_divergence(other_categorical_np)
-
-        self.executor.run(fluid.default_main_program())
-
-        np_categorical = CategoricalNumpy(logits_np)
-        np_other_categorical = CategoricalNumpy(other_logits_np)
-        gt_entropy_np = np_categorical.entropy()
-        gt_kl_np = np_categorical.kl_divergence(np_other_categorical)
-
-        # result calculated by paddle
-        [output_entropy_np, output_kl_np] = self.executor.run(
-            program=test_program,
-            feed={'logits': logits_np},
-            fetch_list=[entropy_np, kl_np],
-        )
-        np.testing.assert_allclose(
-            output_entropy_np, gt_entropy_np, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_kl_np, gt_kl_np, rtol=tolerance, atol=tolerance
-        )
-
-    def test_multivariateNormalDiag_distribution(
-        self, batch_size=2, tolerance=1e-6
-    ):
-        test_program = fluid.Program()
-
-        loc_np = np.random.random(
-            batch_size,
-        ).astype('float32')
-        scale_np = np.diag(
-            np.random.random(
-                batch_size,
-            )
-        ).astype('float32')
-        other_loc_np = np.random.random(
-            batch_size,
-        ).astype('float32')
-        other_scale_np = np.diag(
-            np.random.random(
-                batch_size,
-            )
-        ).astype('float32')
-
-        with fluid.program_guard(test_program):
-            loc = layers.data(
-                name='loc',
-                shape=[
-                    batch_size,
-                ],
-                dtype='float32',
-                append_batch_size=False,
-            )
-            scale = layers.data(
-                name='scale',
-                shape=[batch_size, batch_size],
-                dtype='float32',
-                append_batch_size=False,
-            )
-            other_loc = layers.data(
-                name='other_loc',
-                shape=[
-                    batch_size,
-                ],
-                dtype='float32',
-                append_batch_size=False,
-            )
-            other_scale = layers.data(
-                name='other_scale',
-                shape=[batch_size, batch_size],
-                dtype='float32',
-                append_batch_size=False,
-            )
-
-            multivariate_np = MultivariateNormalDiag(loc, scale)
-            other_multivariate_np = MultivariateNormalDiag(
-                other_loc, other_scale
-            )
-
-            entropy_np = multivariate_np.entropy()
-            other_entropy_np = other_multivariate_np.entropy()
-            kl_np = multivariate_np.kl_divergence(other_multivariate_np)
-
-        self.executor.run(fluid.default_main_program())
-
-        np_multivariate = MultivariateNormalDiagNumpy(loc_np, scale_np)
-        np_other_multivariate = MultivariateNormalDiagNumpy(
-            other_loc_np, other_scale_np
-        )
-        gt_entropy_np = np_multivariate.entropy()
-        gt_kl_np = np_multivariate.kl_divergence(np_other_multivariate)
-
-        # result calculated by paddle
-        [output_entropy_np, output_kl_np] = self.executor.run(
-            program=test_program,
-            feed={
-                'loc': loc_np,
-                'scale': scale_np,
-                'other_loc': other_loc_np,
-                'other_scale': other_scale_np,
-            },
-            fetch_list=[entropy_np, kl_np],
-        )
-        np.testing.assert_allclose(
-            output_entropy_np, gt_entropy_np, rtol=tolerance, atol=tolerance
-        )
-        np.testing.assert_allclose(
-            output_kl_np, gt_kl_np, rtol=tolerance, atol=tolerance
-        )
-
-
-class DistributionTestError(unittest.TestCase):
-    def test_normal_error(self):
-        loc = int(1)
-        scale = int(1)
-
-        # type of loc and scale must be float, list, numpy.ndarray, Variable
-        self.assertRaises(TypeError, Normal, loc, 1.0)
-        self.assertRaises(TypeError, Normal, 1.0, scale)
-
-        normal = Normal(0.0, 1.0)
-
-        value = [1.0, 2.0]
-        # type of value must be variable
-        self.assertRaises(TypeError, normal.log_prob, value)
-
-        shape = 1.0
-        # type of shape must be list
-        self.assertRaises(TypeError, normal.sample, shape)
-
-        seed = 1.0
-        # type of seed must be int
-        self.assertRaises(TypeError, normal.sample, [2, 3], seed)
-
-        normal_other = Uniform(1.0, 2.0)
-        # type of other must be an instance of Normal
-        self.assertRaises(TypeError, normal.kl_divergence, normal_other)
-
-    def test_uniform_error(self):
-        low = int(1)
-        high = int(1)
-
-        # type of loc and scale must be float, list, numpy.ndarray, Variable
-        self.assertRaises(TypeError, Uniform, low, 1.0)
-        self.assertRaises(TypeError, Uniform, 1.0, high)
-
-        uniform = Uniform(0.0, 1.0)
-
-        value = [1.0, 2.0]
-        # type of value must be variable
-        self.assertRaises(TypeError, uniform.log_prob, value)
-
-        shape = 1.0
-        # type of shape must be list
-        self.assertRaises(TypeError, uniform.sample, shape)
-
-        seed = 1.0
-        # type of seed must be int
-        self.assertRaises(TypeError, uniform.sample, [2, 3], seed)
-
-    def test_categorical_error(self):
-        logit = 1.0
-
-        # type of loc and scale must be list, numpy.ndarray, Variable
-        self.assertRaises(TypeError, Categorical, logit)
-
-        categorical = Categorical([-0.602, -0.602])
-
-        categorical_other = Normal(1.0, 2.0)
-        # type of other must be an instance of Normal
-        self.assertRaises(
-            TypeError, categorical.kl_divergence, categorical_other
-        )
-
-    def test_multivariate_normal_diag_error(self):
-        loc = 1.0
-        scale = 1.0
-
-        # type of loc and scale must be list, numpy.ndarray, Variable
-        self.assertRaises(TypeError, MultivariateNormalDiag, loc, [1.0])
-        self.assertRaises(TypeError, MultivariateNormalDiag, [1.0], scale)
-
-        mnd = MultivariateNormalDiag([0.3, 0.5], [[0.4, 0], [0, 0.5]])
-
-        categorical_other = Normal(1.0, 2.0)
-        # type of other must be an instance of Normal
-        self.assertRaises(TypeError, mnd.kl_divergence, categorical_other)
-
-
-if __name__ == '__main__':
-    unittest.main()
diff --git a/python/paddle/fluid/tests/unittests/test_get_places_op.py b/python/paddle/fluid/tests/unittests/test_get_places_op.py
deleted file mode 100644
index 03f32c78b5..0000000000
--- a/python/paddle/fluid/tests/unittests/test_get_places_op.py
+++ /dev/null
@@ -1,48 +0,0 @@
-#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-import unittest
-
-from decorator_helper import prog_scope
-
-import paddle.fluid as fluid
-import paddle.fluid.core as core
-from paddle.fluid.layers.device import get_places
-
-
-class TestGetPlaces(unittest.TestCase):
-    @prog_scope()
-    def check_get_cpu_places(self):
-        places = get_places()
-        cpu = fluid.CPUPlace()
-        exe = fluid.Executor(cpu)
-        exe.run(fluid.default_main_program())
-        self.assertEqual(places.type, fluid.core.VarDesc.VarType.PLACE_LIST)
-
-    @prog_scope()
-    def check_get_gpu_places(self):
-        places = get_places(device_type='CUDA')
-        gpu = fluid.CUDAPlace(0)
-        exe = fluid.Executor(gpu)
-        exe.run(fluid.default_main_program())
-        self.assertEqual(places.type, fluid.core.VarDesc.VarType.PLACE_LIST)
-
-    def test_main(self):
-        if core.is_compiled_with_cuda():
-            self.check_get_gpu_places()
-        self.check_get_cpu_places()
-
-
-if __name__ == '__main__':
-    unittest.main()
diff --git a/python/paddle/fluid/tests/unittests/test_inference_model_io.py b/python/paddle/fluid/tests/unittests/test_inference_model_io.py
index b7aa6e7ba0..9cf82e16f7 100644
--- a/python/paddle/fluid/tests/unittests/test_inference_model_io.py
+++ b/python/paddle/fluid/tests/unittests/test_inference_model_io.py
@@ -174,8 +174,8 @@ class TestSaveInferenceModel(unittest.TestCase):
             x = layers.data(name='x', shape=[2], dtype='float32')
             y = layers.data(name='y', shape=[1], dtype='int32')
             predict = fluid.layers.fc(input=x, size=2, act='softmax')
-            acc = fluid.layers.accuracy(input=predict, label=y)
-            auc_var, batch_auc_var, auc_states = fluid.layers.auc(
+            acc = paddle.static.accuracy(input=predict, label=y)
+            auc_var, batch_auc_var, auc_states = paddle.static.auc(
                 input=predict, label=y
             )
             cost = fluid.layers.cross_entropy(input=predict, label=y)
diff --git a/python/paddle/fluid/tests/unittests/test_layers.py b/python/paddle/fluid/tests/unittests/test_layers.py
index 912382f49a..8807b77664 100644
--- a/python/paddle/fluid/tests/unittests/test_layers.py
+++ b/python/paddle/fluid/tests/unittests/test_layers.py
@@ -34,7 +34,6 @@ from paddle.fluid.framework import (
     program_guard,
 )
 from paddle.fluid.initializer import Constant
-from paddle.fluid.layers.device import get_places
 from paddle.fluid.param_attr import ParamAttr
 from paddle.tensor import random
 
@@ -2895,7 +2894,7 @@ class TestLayer(LayerTest):
             label = fluid.data(name="label", shape=[-1, 1], dtype="int")
             fc_out = fluid.layers.fc(input=data, size=10)
             predict = fluid.layers.softmax(input=fc_out)
-            result = fluid.layers.accuracy(input=predict, label=label, k=5)
+            result = paddle.static.accuracy(input=predict, label=label, k=5)
             place = fluid.CPUPlace()
             exe = fluid.Executor(place)
 
@@ -2911,7 +2910,9 @@ class TestLayer(LayerTest):
             label = base.to_variable(y)
             fc_out = fluid.layers.fc(data, size=10)
             predict = fluid.layers.softmax(fc_out)
-            dynamic_out = fluid.layers.accuracy(input=predict, label=label, k=5)
+            dynamic_out = paddle.static.accuracy(
+                input=predict, label=label, k=5
+            )
 
         np.testing.assert_array_equal(static_out[0], dynamic_out.numpy())
 
@@ -2954,7 +2955,6 @@ class TestBook(LayerTest):
                     )
 
                 else:
-                    assert method.__name__ in ('make_get_places')
                     continue
             if method.__name__ in self.only_static_set:
                 continue
@@ -3201,12 +3201,6 @@ class TestBook(LayerTest):
             hid = layers.fc(input=data, size=20)
             return layers.softmax(hid, axis=1)
 
-    def make_get_places(self):
-        with program_guard(
-            fluid.default_main_program(), fluid.default_startup_program()
-        ):
-            get_places(device_count=1)
-
     @prog_scope()
     def make_nce(self):
         window_size = 5
diff --git a/python/paddle/fluid/tests/unittests/test_profiler.py b/python/paddle/fluid/tests/unittests/test_profiler.py
index 5fbedfaaa7..3303481165 100644
--- a/python/paddle/fluid/tests/unittests/test_profiler.py
+++ b/python/paddle/fluid/tests/unittests/test_profiler.py
@@ -60,7 +60,7 @@ class TestProfiler(unittest.TestCase):
             cost = fluid.layers.cross_entropy(input=predict, label=label)
             avg_cost = paddle.mean(cost)
             batch_size = fluid.layers.create_tensor(dtype='int64')
-            batch_acc = fluid.layers.accuracy(
+            batch_acc = paddle.static.accuracy(
                 input=predict, label=label, total=batch_size
             )
 
diff --git a/python/paddle/fluid/tests/unittests/test_program_prune_backward.py b/python/paddle/fluid/tests/unittests/test_program_prune_backward.py
index 04c707b320..d751fd4b90 100755
--- a/python/paddle/fluid/tests/unittests/test_program_prune_backward.py
+++ b/python/paddle/fluid/tests/unittests/test_program_prune_backward.py
@@ -76,7 +76,7 @@ def simple_fc_net_with_accuracy(use_feed):
     prediction = fluid.layers.fc(hidden, size=10, act='softmax')
     loss = fluid.layers.cross_entropy(input=prediction, label=label)
     loss = paddle.mean(loss)
-    accuracy_out = fluid.layers.accuracy(input=prediction, label=label, k=5)
+    accuracy_out = paddle.static.accuracy(input=prediction, label=label, k=5)
     return loss
 
 
diff --git a/python/paddle/static/__init__.py b/python/paddle/static/__init__.py
index 78ad1cfabc..f527b5a1c3 100644
--- a/python/paddle/static/__init__.py
+++ b/python/paddle/static/__init__.py
@@ -68,11 +68,12 @@ from ..fluid.io import batch  # noqa: F401
 
 from ..fluid.layers import create_parameter  # noqa: F401
 from ..fluid.layers import create_global_var  # noqa: F401
-from ..fluid.layers.metric_op import auc  # noqa: F401
-from ..fluid.layers.metric_op import accuracy  # noqa: F401
 from ..fluid.contrib.layers import ctr_metric_bundle  # noqa: F401
 from ..fluid.layers import exponential_decay  # noqa: F401
 
+from paddle.static.nn.metric import auc  # noqa: F401
+from paddle.static.nn.metric import accuracy  # noqa: F401
+
 __all__ = [  # noqa
     'append_backward',
     'gradients',
diff --git a/python/paddle/fluid/layers/metric_op.py b/python/paddle/static/nn/metric.py
old mode 100755
new mode 100644
similarity index 97%
rename from python/paddle/fluid/layers/metric_op.py
rename to python/paddle/static/nn/metric.py
index 3179f5d568..948b100bce
--- a/python/paddle/fluid/layers/metric_op.py
+++ b/python/paddle/static/nn/metric.py
@@ -15,22 +15,16 @@
 All layers just related to metric.
 """
 
-import warnings
-from ..layer_helper import LayerHelper
-from ..initializer import Normal, Constant
-from ..framework import (
+from paddle.fluid.layer_helper import LayerHelper
+from paddle.fluid.initializer import Constant
+from paddle.fluid.framework import (
     Variable,
     _non_static_mode,
     _varbase_creator,
-    _in_legacy_dygraph,
-    in_dygraph_mode,
 )
-from .. import core
-from ..param_attr import ParamAttr
-from . import nn
-from . import tensor
-from ..data_feeder import check_variable_and_dtype
-from paddle import _C_ops, _legacy_C_ops
+from paddle.fluid.layers import tensor
+from paddle.fluid.data_feeder import check_variable_and_dtype
+from paddle import _legacy_C_ops
 
 __all__ = ['accuracy', 'auc']
 
-- 
GitLab