diff --git a/mindspore/nn/probability/bnn_layers/_util.py b/mindspore/nn/probability/bnn_layers/_util.py
new file mode 100644
index 0000000000..fe8983fdb9
--- /dev/null
+++ b/mindspore/nn/probability/bnn_layers/_util.py
@@ -0,0 +1,46 @@
+# Copyright 2020 Huawei Technologies Co., Ltd
+#
+# 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.
+# ============================================================================
+"""Utility functions to help bnn layers."""
+from mindspore.common.tensor import Tensor
+from ...cell import Cell
+
+
+def check_prior(prior_fn, arg_name):
+    """check prior distribution of bnn layers."""
+    if isinstance(prior_fn, Cell):
+        prior = prior_fn
+    else:
+        prior = prior_fn()
+    for prior_name, prior_dist in prior.name_cells().items():
+        if prior_name != 'normal':
+            raise TypeError(f"The type of distribution of `{arg_name}` should be `normal`")
+        if not (isinstance(getattr(prior_dist, '_mean_value'), Tensor) and
+                isinstance(getattr(prior_dist, '_sd_value'), Tensor)):
+            raise TypeError(f"The input form of `{arg_name}` is incorrect")
+    return prior
+
+
+def check_posterior(posterior_fn, shape, param_name, arg_name):
+    """check posterior distribution of bnn layers."""
+    try:
+        posterior = posterior_fn(shape=shape, name=param_name)
+    except TypeError:
+        raise TypeError(f'The type of `{arg_name}` should be `NormalPosterior`')
+    finally:
+        pass
+    for posterior_name, _ in posterior.name_cells().items():
+        if posterior_name != 'normal':
+            raise TypeError(f"The type of distribution of `{arg_name}` should be `normal`")
+    return posterior
diff --git a/mindspore/nn/probability/bnn_layers/conv_variational.py b/mindspore/nn/probability/bnn_layers/conv_variational.py
index fc637bbf0b..8b2df4cf7c 100644
--- a/mindspore/nn/probability/bnn_layers/conv_variational.py
+++ b/mindspore/nn/probability/bnn_layers/conv_variational.py
@@ -14,11 +14,10 @@
 # ============================================================================
 """Convolutional variational layers."""
 from mindspore.ops import operations as P
-from mindspore.common.tensor import Tensor
 from mindspore._checkparam import twice
 from ...layer.conv import _Conv
-from ...cell import Cell
-from .layer_distribution import NormalPrior, NormalPosterior
+from .layer_distribution import NormalPrior, normal_post_fn
+from ._util import check_prior, check_posterior
 
 __all__ = ['ConvReparam']
 
@@ -39,9 +38,9 @@ class _ConvVariational(_Conv):
                  group=1,
                  has_bias=False,
                  weight_prior_fn=NormalPrior,
-                 weight_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape),
+                 weight_posterior_fn=normal_post_fn,
                  bias_prior_fn=NormalPrior,
-                 bias_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape)):
+                 bias_posterior_fn=normal_post_fn):
         kernel_size = twice(kernel_size)
         stride = twice(stride)
         dilation = twice(dilation)
@@ -78,47 +77,15 @@ class _ConvVariational(_Conv):
                       self.in_channels // self.group, *self.kernel_size]
 
         self.weight.requires_grad = False
-
-        if isinstance(weight_prior_fn, Cell):
-            self.weight_prior = weight_prior_fn
-        else:
-            self.weight_prior = weight_prior_fn()
-        for prior_name, prior_dist in self.weight_prior.name_cells().items():
-            if prior_name != 'normal':
-                raise TypeError("The type of distribution of `weight_prior_fn` should be `normal`")
-            if not (isinstance(getattr(prior_dist, '_mean_value'), Tensor) and
-                    isinstance(getattr(prior_dist, '_sd_value'), Tensor)):
-                raise TypeError("The input form of `weight_prior_fn` is incorrect")
-
-        try:
-            self.weight_posterior = weight_posterior_fn(shape=self.shape, name='bnn_weight')
-        except TypeError:
-            raise TypeError('The input form of `weight_posterior_fn` is incorrect')
-        for posterior_name, _ in self.weight_posterior.name_cells().items():
-            if posterior_name != 'normal':
-                raise TypeError("The type of distribution of `weight_posterior_fn` should be `normal`")
+        self.weight_prior = check_prior(weight_prior_fn, "weight_prior_fn")
+        self.weight_posterior = check_posterior(weight_posterior_fn, shape=self.shape, param_name='bnn_weight',
+                                                arg_name="weight_posterior_fn")
 
         if self.has_bias:
             self.bias.requires_grad = False
-
-            if isinstance(bias_prior_fn, Cell):
-                self.bias_prior = bias_prior_fn
-            else:
-                self.bias_prior = bias_prior_fn()
-            for prior_name, prior_dist in self.bias_prior.name_cells().items():
-                if prior_name != 'normal':
-                    raise TypeError("The type of distribution of `bias_prior_fn` should be `normal`")
-                if not (isinstance(getattr(prior_dist, '_mean_value'), Tensor) and
-                        isinstance(getattr(prior_dist, '_sd_value'), Tensor)):
-                    raise TypeError("The input form of `bias_prior_fn` is incorrect")
-
-            try:
-                self.bias_posterior = bias_posterior_fn(shape=[self.out_channels], name='bnn_bias')
-            except TypeError:
-                raise TypeError('The type of `bias_posterior_fn` should be `NormalPosterior`')
-            for posterior_name, _ in self.bias_posterior.name_cells().items():
-                if posterior_name != 'normal':
-                    raise TypeError("The type of distribution of `bias_posterior_fn` should be `normal`")
+            self.bias_prior = check_prior(bias_prior_fn, "bias_prior_fn")
+            self.bias_posterior = check_posterior(bias_posterior_fn, shape=[self.out_channels], param_name='bnn_bias',
+                                                  arg_name="bias_posterior_fn")
 
         # mindspore operations
         self.bias_add = P.BiasAdd()
@@ -135,23 +102,23 @@ class _ConvVariational(_Conv):
         self.sum = P.ReduceSum()
 
     def construct(self, inputs):
-        outputs = self._apply_variational_weight(inputs)
+        outputs = self.apply_variational_weight(inputs)
         if self.has_bias:
-            outputs = self._apply_variational_bias(outputs)
+            outputs = self.apply_variational_bias(outputs)
         return outputs
 
     def extend_repr(self):
         s = 'in_channels={}, out_channels={}, kernel_size={}, stride={},  pad_mode={}, ' \
-            'padding={}, dilation={}, group={}, weight_mean={}, weight_std={}, has_bias={}'\
+            'padding={}, dilation={}, group={}, weight_mean={}, weight_std={}, has_bias={}' \
             .format(self.in_channels, self.out_channels, self.kernel_size, self.stride, self.pad_mode, self.padding,
                     self.dilation, self.group, self.weight_posterior.mean, self.weight_posterior.untransformed_std,
                     self.has_bias)
         if self.has_bias:
-            s += ', bias_mean={}, bias_std={}'\
+            s += ', bias_mean={}, bias_std={}' \
                 .format(self.bias_posterior.mean, self.bias_posterior.untransformed_std)
         return s
 
-    def _apply_variational_bias(self, inputs):
+    def apply_variational_bias(self, inputs):
         bias_posterior_tensor = self.bias_posterior("sample")
         return self.bias_add(inputs, bias_posterior_tensor)
 
@@ -230,7 +197,7 @@ class ConvReparam(_ConvVariational):
             normal distribution). The current version only supports normal distribution.
         weight_posterior_fn: The posterior distribution for sampling weight.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
         bias_prior_fn: The prior distribution for bias vector. It must return
             a mindspore distribution. Default: NormalPrior(which creates an
@@ -238,7 +205,7 @@ class ConvReparam(_ConvVariational):
             only supports normal distribution.
         bias_posterior_fn: The posterior distribution for sampling bias vector.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
 
     Inputs:
@@ -270,9 +237,9 @@ class ConvReparam(_ConvVariational):
             group=1,
             has_bias=False,
             weight_prior_fn=NormalPrior,
-            weight_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape),
+            weight_posterior_fn=normal_post_fn,
             bias_prior_fn=NormalPrior,
-            bias_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape)):
+            bias_posterior_fn=normal_post_fn):
         super(ConvReparam, self).__init__(
             in_channels,
             out_channels,
@@ -289,7 +256,7 @@ class ConvReparam(_ConvVariational):
             bias_posterior_fn=bias_posterior_fn
         )
 
-    def _apply_variational_weight(self, inputs):
+    def apply_variational_weight(self, inputs):
         weight_posterior_tensor = self.weight_posterior("sample")
         outputs = self.conv2d(inputs, weight_posterior_tensor)
         return outputs
diff --git a/mindspore/nn/probability/bnn_layers/dense_variational.py b/mindspore/nn/probability/bnn_layers/dense_variational.py
index 95989f5cd9..28dc5df4c8 100644
--- a/mindspore/nn/probability/bnn_layers/dense_variational.py
+++ b/mindspore/nn/probability/bnn_layers/dense_variational.py
@@ -14,12 +14,12 @@
 # ============================================================================
 """dense_variational"""
 from mindspore.ops import operations as P
-from mindspore.common.tensor import Tensor
 from mindspore._checkparam import Validator
 from ...cell import Cell
 from ...layer.activation import get_activation
 from ..distribution.normal import Normal
-from .layer_distribution import NormalPrior, NormalPosterior
+from .layer_distribution import NormalPrior, normal_post_fn
+from ._util import check_prior, check_posterior
 
 __all__ = ['DenseReparam', 'DenseLocalReparam']
 
@@ -36,52 +36,22 @@ class _DenseVariational(Cell):
             activation=None,
             has_bias=True,
             weight_prior_fn=NormalPrior,
-            weight_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape),
+            weight_posterior_fn=normal_post_fn,
             bias_prior_fn=NormalPrior,
-            bias_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape)):
+            bias_posterior_fn=normal_post_fn):
         super(_DenseVariational, self).__init__()
         self.in_channels = Validator.check_positive_int(in_channels)
         self.out_channels = Validator.check_positive_int(out_channels)
         self.has_bias = Validator.check_bool(has_bias)
 
-        if isinstance(weight_prior_fn, Cell):
-            self.weight_prior = weight_prior_fn
-        else:
-            self.weight_prior = weight_prior_fn()
-        for prior_name, prior_dist in self.weight_prior.name_cells().items():
-            if prior_name != 'normal':
-                raise TypeError("The type of distribution of `weight_prior_fn` should be `normal`")
-            if not (isinstance(getattr(prior_dist, '_mean_value'), Tensor) and
-                    isinstance(getattr(prior_dist, '_sd_value'), Tensor)):
-                raise TypeError("The input form of `weight_prior_fn` is incorrect")
-
-        try:
-            self.weight_posterior = weight_posterior_fn(shape=[self.out_channels, self.in_channels], name='bnn_weight')
-        except TypeError:
-            raise TypeError('The type of `weight_posterior_fn` should be `NormalPosterior`')
-        for posterior_name, _ in self.weight_posterior.name_cells().items():
-            if posterior_name != 'normal':
-                raise TypeError("The type of distribution of `weight_posterior_fn` should be `normal`")
+        self.weight_prior = check_prior(weight_prior_fn, "weight_prior_fn")
+        self.weight_posterior = check_posterior(weight_posterior_fn, shape=[self.out_channels, self.in_channels],
+                                                param_name='bnn_weight', arg_name="weight_posterior_fn")
 
         if self.has_bias:
-            if isinstance(bias_prior_fn, Cell):
-                self.bias_prior = bias_prior_fn
-            else:
-                self.bias_prior = bias_prior_fn()
-            for prior_name, prior_dist in self.bias_prior.name_cells().items():
-                if prior_name != 'normal':
-                    raise TypeError("The type of distribution of `bias_prior_fn` should be `normal`")
-                if not (isinstance(getattr(prior_dist, '_mean_value'), Tensor) and
-                        isinstance(getattr(prior_dist, '_sd_value'), Tensor)):
-                    raise TypeError("The input form of `bias_prior_fn` is incorrect")
-
-            try:
-                self.bias_posterior = bias_posterior_fn(shape=[self.out_channels], name='bnn_bias')
-            except TypeError:
-                raise TypeError('The type of `bias_posterior_fn` should be `NormalPosterior`')
-            for posterior_name, _ in self.bias_posterior.name_cells().items():
-                if posterior_name != 'normal':
-                    raise TypeError("The type of distribution of `bias_posterior_fn` should be `normal`")
+            self.bias_prior = check_prior(bias_prior_fn, "bias_prior_fn")
+            self.bias_posterior = check_posterior(bias_posterior_fn, shape=[self.out_channels], param_name='bnn_bias',
+                                                  arg_name="bias_posterior_fn")
 
         self.activation = activation
         if not self.activation:
@@ -100,9 +70,9 @@ class _DenseVariational(Cell):
         self.sum = P.ReduceSum()
 
     def construct(self, x):
-        outputs = self._apply_variational_weight(x)
+        outputs = self.apply_variational_weight(x)
         if self.has_bias:
-            outputs = self._apply_variational_bias(outputs)
+            outputs = self.apply_variational_bias(outputs)
         if self.activation_flag:
             outputs = self.activation(outputs)
         return outputs
@@ -118,7 +88,7 @@ class _DenseVariational(Cell):
             s += ', activation={}'.format(self.activation)
         return s
 
-    def _apply_variational_bias(self, inputs):
+    def apply_variational_bias(self, inputs):
         bias_posterior_tensor = self.bias_posterior("sample")
         return self.bias_add(inputs, bias_posterior_tensor)
 
@@ -162,16 +132,17 @@ class DenseReparam(_DenseVariational):
         in_channels (int): The number of input channel.
         out_channels (int): The number of output channel .
         has_bias (bool): Specifies whether the layer uses a bias vector. Default: False.
-        activation (str, Cell): A regularization function applied to the output of the layer. The type of `activation`
-            can be a string (eg. 'relu') or a Cell (eg. nn.ReLU()). Note that if the type of activation is Cell, it must
-            be instantiated beforehand. Default: None.
+        activation (str, Cell): A regularization function applied to the output of the layer.
+            The type of `activation` can be a string (eg. 'relu') or a Cell (eg. nn.ReLU()).
+            Note that if the type of activation is Cell, it must be instantiated beforehand.
+            Default: None.
         weight_prior_fn: The prior distribution for weight.
             It must return a mindspore distribution instance.
             Default: NormalPrior. (which creates an instance of standard
             normal distribution). The current version only supports normal distribution.
         weight_posterior_fn: The posterior distribution for sampling weight.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
         bias_prior_fn: The prior distribution for bias vector. It must return
             a mindspore distribution. Default: NormalPrior(which creates an
@@ -179,7 +150,7 @@ class DenseReparam(_DenseVariational):
             only supports normal distribution.
         bias_posterior_fn: The posterior distribution for sampling bias vector.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
 
     Inputs:
@@ -206,9 +177,9 @@ class DenseReparam(_DenseVariational):
             activation=None,
             has_bias=True,
             weight_prior_fn=NormalPrior,
-            weight_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape),
+            weight_posterior_fn=normal_post_fn,
             bias_prior_fn=NormalPrior,
-            bias_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape)):
+            bias_posterior_fn=normal_post_fn):
         super(DenseReparam, self).__init__(
             in_channels,
             out_channels,
@@ -220,7 +191,7 @@ class DenseReparam(_DenseVariational):
             bias_posterior_fn=bias_posterior_fn
         )
 
-    def _apply_variational_weight(self, inputs):
+    def apply_variational_weight(self, inputs):
         weight_posterior_tensor = self.weight_posterior("sample")
         outputs = self.matmul(inputs, weight_posterior_tensor)
         return outputs
@@ -250,16 +221,17 @@ class DenseLocalReparam(_DenseVariational):
         in_channels (int): The number of input channel.
         out_channels (int): The number of output channel .
         has_bias (bool): Specifies whether the layer uses a bias vector. Default: False.
-        activation (str, Cell): A regularization function applied to the output of the layer. The type of `activation`
-            can be a string (eg. 'relu') or a Cell (eg. nn.ReLU()). Note that if the type of activation is Cell, it must
-            be instantiated beforehand. Default: None.
+        activation (str, Cell): A regularization function applied to the output of the layer.
+            The type of `activation` can be a string (eg. 'relu') or a Cell (eg. nn.ReLU()).
+            Note that if the type of activation is Cell, it must be instantiated beforehand.
+            Default: None.
         weight_prior_fn: The prior distribution for weight.
             It must return a mindspore distribution instance.
             Default: NormalPrior. (which creates an instance of standard
             normal distribution). The current version only supports normal distribution.
         weight_posterior_fn: The posterior distribution for sampling weight.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
         bias_prior_fn: The prior distribution for bias vector. It must return
             a mindspore distribution. Default: NormalPrior(which creates an
@@ -267,7 +239,7 @@ class DenseLocalReparam(_DenseVariational):
             only supports normal distribution.
         bias_posterior_fn: The posterior distribution for sampling bias vector.
             It must be a function handle which returns a mindspore
-            distribution instance. Default: lambda name, shape: NormalPosterior(name=name, shape=shape).
+            distribution instance. Default: normal_post_fn.
             The current version only supports normal distribution.
 
     Inputs:
@@ -294,9 +266,9 @@ class DenseLocalReparam(_DenseVariational):
             activation=None,
             has_bias=True,
             weight_prior_fn=NormalPrior,
-            weight_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape),
+            weight_posterior_fn=normal_post_fn,
             bias_prior_fn=NormalPrior,
-            bias_posterior_fn=lambda name, shape: NormalPosterior(name=name, shape=shape)):
+            bias_posterior_fn=normal_post_fn):
         super(DenseLocalReparam, self).__init__(
             in_channels,
             out_channels,
@@ -311,7 +283,7 @@ class DenseLocalReparam(_DenseVariational):
         self.square = P.Square()
         self.normal = Normal()
 
-    def _apply_variational_weight(self, inputs):
+    def apply_variational_weight(self, inputs):
         mean = self.matmul(inputs, self.weight_posterior("mean"))
         std = self.sqrt(self.matmul(self.square(inputs), self.square(self.weight_posterior("sd"))))
         weight_posterior_affine_tensor = self.normal("sample", mean=mean, sd=std)
diff --git a/mindspore/nn/probability/bnn_layers/layer_distribution.py b/mindspore/nn/probability/bnn_layers/layer_distribution.py
index ea221d737b..7d9e19f20c 100644
--- a/mindspore/nn/probability/bnn_layers/layer_distribution.py
+++ b/mindspore/nn/probability/bnn_layers/layer_distribution.py
@@ -115,3 +115,8 @@ class NormalPosterior(Cell):
     def construct(self, *inputs):
         std = self.std_trans(self.untransformed_std)
         return self.normal(*inputs, mean=self.mean, sd=std)
+
+
+def normal_post_fn(name, shape):
+    """Provide normal posterior distribution."""
+    return NormalPosterior(name=name, shape=shape)
diff --git a/mindspore/nn/probability/distribution/_utils/__init__.py b/mindspore/nn/probability/distribution/_utils/__init__.py
index c54399b623..ad033c718a 100644
--- a/mindspore/nn/probability/distribution/_utils/__init__.py
+++ b/mindspore/nn/probability/distribution/_utils/__init__.py
@@ -15,8 +15,9 @@
 """
 Distribution operation utility functions.
 """
-from .utils import *
-from .custom_ops import *
+from .custom_ops import exp_generic, log_generic, broadcast_to
+from .utils import cast_to_tensor, check_greater_equal_zero, check_greater_zero
+from .utils import check_greater, check_prob, CheckTensor, CheckTuple, set_param_type
 
 __all__ = [
     'cast_to_tensor',
diff --git a/mindspore/nn/probability/distribution/_utils/custom_ops.py b/mindspore/nn/probability/distribution/_utils/custom_ops.py
index 5dc28f99ea..d716673000 100644
--- a/mindspore/nn/probability/distribution/_utils/custom_ops.py
+++ b/mindspore/nn/probability/distribution/_utils/custom_ops.py
@@ -17,6 +17,7 @@ import numpy as np
 from mindspore.ops import operations as P
 from mindspore.common import dtype as mstype
 
+
 def exp_generic(input_x):
     """
     Log op on Ascend doesn't support int types.
@@ -66,6 +67,7 @@ def log1p_generic(x):
     """
     return log_generic(x + 1.0)
 
+
 def broadcast_to(x, target):
     """
     Broadcast x to the shape of target.
diff --git a/mindspore/nn/probability/distribution/_utils/utils.py b/mindspore/nn/probability/distribution/_utils/utils.py
index b0aa35474e..753e6f1abc 100644
--- a/mindspore/nn/probability/distribution/_utils/utils.py
+++ b/mindspore/nn/probability/distribution/_utils/utils.py
@@ -24,6 +24,7 @@ from mindspore.ops import operations as P
 from mindspore.ops.primitive import constexpr, PrimitiveWithInfer, prim_attr_register
 import mindspore.nn as nn
 
+
 def cast_to_tensor(t, hint_type=mstype.float32):
     """
     Cast an user input value into a Tensor of dtype.
@@ -51,6 +52,7 @@ def cast_to_tensor(t, hint_type=mstype.float32):
     raise TypeError(
         f"Unable to convert input of type {invalid_type} to a Tensor of type {hint_type}")
 
+
 def cast_type_for_device(dtype):
     """
     use the alternative dtype supported by the device.
@@ -158,6 +160,7 @@ def check_prob(p):
     if not comp.all():
         raise ValueError('Probabilities should be less than one')
 
+
 def check_sum_equal_one(probs):
     """
     Used in categorical distribution. check if probabilities of each category sum to 1.
@@ -176,6 +179,7 @@ def check_sum_equal_one(probs):
     if not comp:
         raise ValueError('Probabilities for each category should sum to one for Categorical distribution.')
 
+
 def check_rank(probs):
     """
     Used in categorical distribution. check Rank >=1.
@@ -189,6 +193,7 @@ def check_rank(probs):
     if probs.asnumpy().ndim == 0:
         raise ValueError('probs for Categorical distribution must have rank >= 1.')
 
+
 def logits_to_probs(logits, is_binary=False):
     """
     converts logits into probabilities.
@@ -204,8 +209,6 @@ def logits_to_probs(logits, is_binary=False):
 def clamp_probs(probs):
     """
     clamp probs boundary
-    Args:
-        probs (Tensor)
     """
     eps = P.Eps()(probs)
     return C.clip_by_value(probs, eps, 1-eps)
@@ -229,29 +232,35 @@ def raise_none_error(name):
     raise TypeError(f"the type {name} should be subclass of Tensor."
                     f" It should not be None since it is not specified during initialization.")
 
+
 @constexpr
 def raise_probs_logits_error():
     raise TypeError("Either 'probs' or 'logits' must be specified, but not both.")
 
+
 @constexpr
 def raise_broadcast_error(shape_a, shape_b):
     raise ValueError(f"Shape {shape_a} and {shape_b} is not broadcastable.")
 
+
 @constexpr
 def raise_not_impl_error(name):
     raise ValueError(
         f"{name} function should be implemented for non-linear transformation")
 
+
 @constexpr
 def raise_not_implemented_util(func_name, obj, *args, **kwargs):
     raise NotImplementedError(
         f"{func_name} is not implemented for {obj} distribution.")
 
+
 @constexpr
 def raise_type_error(name, cur_type, required_type):
     raise TypeError(
         f"For {name} , the type should be or be subclass of {required_type}, but got {cur_type}")
 
+
 @constexpr
 def raise_not_defined(func_name, obj, *args, **kwargs):
     raise ValueError(
@@ -320,6 +329,7 @@ class CheckTensor(PrimitiveWithInfer):
             return x
         raise TypeError(f"For {name}, input type should be a Tensor or Parameter.")
 
+
 def set_param_type(args, hint_type):
     """
     Find the common type among arguments.
diff --git a/mindspore/nn/probability/distribution/cauchy.py b/mindspore/nn/probability/distribution/cauchy.py
index a63d4a3f5f..754c0c551f 100644
--- a/mindspore/nn/probability/distribution/cauchy.py
+++ b/mindspore/nn/probability/distribution/cauchy.py
@@ -60,7 +60,8 @@ class Cauchy(Distribution):
         >>> loc_b = Tensor([1.0], dtype=mindspore.float32)
         >>> scale_b = Tensor([1.0, 1.5, 2.0], dtype=mindspore.float32)
         >>> # Private interfaces of probability functions corresponding to public interfaces, including
-        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`, have the same arguments as follows.
+        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`,
+        >>> # have the same arguments as follows.
         >>> # Args:
         >>> #     value (Tensor): the value to be evaluated.
         >>> #     loc (Tensor): the location of the distribution. Default: self.loc.
diff --git a/mindspore/nn/probability/distribution/gamma.py b/mindspore/nn/probability/distribution/gamma.py
index 91cb43b58f..6f971616eb 100644
--- a/mindspore/nn/probability/distribution/gamma.py
+++ b/mindspore/nn/probability/distribution/gamma.py
@@ -63,7 +63,8 @@ class Gamma(Distribution):
         >>> rate_b = Tensor([1.0, 1.5, 2.0], dtype=mindspore.float32)
         >>>
         >>> # Private interfaces of probability functions corresponding to public interfaces, including
-        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`, have the same arguments as follows.
+        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`,
+        >>> # have the same arguments as follows.
         >>> # Args:
         >>> #     value (Tensor): the value to be evaluated.
         >>> #     concentration (Tensor): the concentration of the distribution. Default: self._concentration.
diff --git a/mindspore/nn/probability/distribution/geometric.py b/mindspore/nn/probability/distribution/geometric.py
index fc0ea4f713..2de3db1a5c 100644
--- a/mindspore/nn/probability/distribution/geometric.py
+++ b/mindspore/nn/probability/distribution/geometric.py
@@ -60,7 +60,8 @@ class Geometric(Distribution):
         >>> probs_b = Tensor([0.2, 0.5, 0.4], dtype=mindspore.float32)
         >>>
         >>> # Private interfaces of probability functions corresponding to public interfaces, including
-        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`, have the same arguments as follows.
+        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`,
+        >>> # have the same arguments as follows.
         >>> # Args:
         >>> #     value (Tensor): the value to be evaluated.
         >>> #     probs1 (Tensor): the probability of success of a Bernoulli trial. Default: self.probs.
@@ -96,7 +97,7 @@ class Geometric(Distribution):
         >>> # Args:
         >>> #     dist (str): the name of the distribution. Only 'Geometric' is supported.
         >>> #     probs1_b (Tensor): the probability of success of a Bernoulli trial of distribution b.
-        >>> #     probs1_a (Tensor): the probability of success of a Bernoulli trial of distribution a. Default: self.probs.
+        >>> #     probs1_a (Tensor): the probability of success of a Bernoulli trial of distribution a.
         >>> # Examples of `kl_loss`. `cross_entropy` is similar.
         >>> ans = g1.kl_loss('Geometric', probs_b)
         >>> print(ans.shape)
diff --git a/mindspore/nn/probability/distribution/gumbel.py b/mindspore/nn/probability/distribution/gumbel.py
index 5a9506509b..1e7184aaf9 100644
--- a/mindspore/nn/probability/distribution/gumbel.py
+++ b/mindspore/nn/probability/distribution/gumbel.py
@@ -24,6 +24,7 @@ from .transformed_distribution import TransformedDistribution
 from ._utils.utils import check_distribution_name
 from ._utils.custom_ops import exp_generic, log_generic
 
+
 class Gumbel(TransformedDistribution):
     """
     Gumbel distribution.
diff --git a/mindspore/nn/probability/distribution/log_normal.py b/mindspore/nn/probability/distribution/log_normal.py
index 3768ac252c..c78586cf8b 100644
--- a/mindspore/nn/probability/distribution/log_normal.py
+++ b/mindspore/nn/probability/distribution/log_normal.py
@@ -21,6 +21,7 @@ import mindspore.nn.probability.distribution as msd
 from ._utils.utils import check_distribution_name
 from ._utils.custom_ops import exp_generic, log_generic
 
+
 class LogNormal(msd.TransformedDistribution):
     """
     LogNormal distribution.
diff --git a/mindspore/nn/probability/distribution/logistic.py b/mindspore/nn/probability/distribution/logistic.py
index c4a77eb3b0..346c60ce00 100644
--- a/mindspore/nn/probability/distribution/logistic.py
+++ b/mindspore/nn/probability/distribution/logistic.py
@@ -61,7 +61,8 @@ class Logistic(Distribution):
         >>> scale_b = Tensor([1.0, 1.5, 2.0], dtype=mindspore.float32)
         >>>
         >>> # Private interfaces of probability functions corresponding to public interfaces, including
-        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`, have the same arguments as follows.
+        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`,
+        >>> # have the same arguments as follows.
         >>> # Args:
         >>> #     value (Tensor): the value to be evaluated.
         >>> #     loc (Tensor): the location of the distribution. Default: self.loc.
diff --git a/mindspore/nn/probability/distribution/normal.py b/mindspore/nn/probability/distribution/normal.py
index b642fb94cb..7fc7ca3833 100644
--- a/mindspore/nn/probability/distribution/normal.py
+++ b/mindspore/nn/probability/distribution/normal.py
@@ -61,7 +61,8 @@ class Normal(Distribution):
         >>> sd_b = Tensor([1.0, 1.5, 2.0], dtype=mindspore.float32)
 
         >>> # Private interfaces of probability functions corresponding to public interfaces, including
-        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`, have the same arguments as follows.
+        >>> # `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`,
+        >>> # have the same arguments as follows.
         >>> # Args:
         >>> #     value (Tensor): the value to be evaluated.
         >>> #     mean (Tensor): the mean of the distribution. Default: self._mean_value.
diff --git a/mindspore/nn/probability/dpn/__init__.py b/mindspore/nn/probability/dpn/__init__.py
index bf8dbb7660..88f66eb205 100644
--- a/mindspore/nn/probability/dpn/__init__.py
+++ b/mindspore/nn/probability/dpn/__init__.py
@@ -16,7 +16,7 @@
 Deep probability network such as BNN and VAE network.
 """
 
-from .vae import *
+from .vae import VAE, ConditionalVAE
 
 __all__ = []
 __all__.extend(vae.__all__)
diff --git a/mindspore/nn/probability/infer/__init__.py b/mindspore/nn/probability/infer/__init__.py
index 3e190bb2d5..370d2355be 100644
--- a/mindspore/nn/probability/infer/__init__.py
+++ b/mindspore/nn/probability/infer/__init__.py
@@ -16,7 +16,7 @@
 Inference algorithms in Probabilistic Programming.
 """
 
-from .variational import *
+from .variational import SVI, ELBO
 
 __all__ = []
 __all__.extend(variational.__all__)
diff --git a/mindspore/nn/probability/transforms/transform_bnn.py b/mindspore/nn/probability/transforms/transform_bnn.py
index eca6aec1b2..d3bd57013a 100644
--- a/mindspore/nn/probability/transforms/transform_bnn.py
+++ b/mindspore/nn/probability/transforms/transform_bnn.py
@@ -83,8 +83,10 @@ class TransformToBNN:
 
     def transform_to_bnn_model(self,
                                get_dense_args=lambda dp: {"in_channels": dp.in_channels, "has_bias": dp.has_bias,
-                                                          "out_channels": dp.out_channels, "activation": dp.activation},
-                               get_conv_args=lambda dp: {"in_channels": dp.in_channels, "out_channels": dp.out_channels,
+                                                          "out_channels": dp.out_channels,
+                                                          "activation": dp.activation},
+                               get_conv_args=lambda dp: {"in_channels": dp.in_channels,
+                                                         "out_channels": dp.out_channels,
                                                          "pad_mode": dp.pad_mode, "kernel_size": dp.kernel_size,
                                                          "stride": dp.stride, "has_bias": dp.has_bias,
                                                          "padding": dp.padding, "dilation": dp.dilation,
diff --git a/mindspore/nn/probability/zhusuan/__init__.py b/mindspore/nn/probability/zhusuan/__init__.py
index d4330d1162..744f32af3c 100644
--- a/mindspore/nn/probability/zhusuan/__init__.py
+++ b/mindspore/nn/probability/zhusuan/__init__.py
@@ -14,5 +14,5 @@
 # ============================================================================
 """ Zhusuan package: a probalistic programming library  """
 
-from .framework import *
-from .variational import *
+from .framework import BayesianNet
+from .variational import ELBO
diff --git a/mindspore/nn/probability/zhusuan/framework/__init__.py b/mindspore/nn/probability/zhusuan/framework/__init__.py
index 680eee8d88..ea39dd599b 100644
--- a/mindspore/nn/probability/zhusuan/framework/__init__.py
+++ b/mindspore/nn/probability/zhusuan/framework/__init__.py
@@ -15,4 +15,4 @@
 
 """ Core functionality for Zhusuan """
 
-from .bn import *
+from .bn import BayesianNet
diff --git a/mindspore/nn/probability/zhusuan/framework/bn.py b/mindspore/nn/probability/zhusuan/framework/bn.py
index dc6a0d4f13..e9eacd0882 100644
--- a/mindspore/nn/probability/zhusuan/framework/bn.py
+++ b/mindspore/nn/probability/zhusuan/framework/bn.py
@@ -34,7 +34,7 @@ class BayesianNet(nn.Cell):
 
         self.reduce_sum = P.ReduceSum(keep_dims=True)
 
-    def Normal(self,
+    def normal(self,
                name,
                observation=None,
                mean=None,
@@ -45,9 +45,8 @@ class BayesianNet(nn.Cell):
                reparameterize=True):
         """ Normal distribution wrapper """
 
-        assert not name is None
-        assert not seed is None
-        assert not dtype is None
+        if not isinstance(name, str):
+            raise TypeError("The type of `name` should be string")
 
         if observation is None:
             if reparameterize:
@@ -63,7 +62,7 @@ class BayesianNet(nn.Cell):
             'log_prob', sample, mean, std), 1)
         return sample, log_prob
 
-    def Bernoulli(self,
+    def bernoulli(self,
                   name,
                   observation=None,
                   probs=None,
@@ -72,9 +71,8 @@ class BayesianNet(nn.Cell):
                   shape=()):
         """ Bernoulli distribution wrapper """
 
-        assert not name is None
-        assert not seed is None
-        assert not dtype is None
+        if not isinstance(name, str):
+            raise TypeError("The type of `name` should be string")
 
         if observation is None:
             sample = self.bernoulli_dist('sample', shape, probs)
diff --git a/mindspore/nn/probability/zhusuan/variational/__init__.py b/mindspore/nn/probability/zhusuan/variational/__init__.py
index fd591d1847..64cc194df4 100644
--- a/mindspore/nn/probability/zhusuan/variational/__init__.py
+++ b/mindspore/nn/probability/zhusuan/variational/__init__.py
@@ -15,4 +15,4 @@
 
 """ Variational inference related codes """
 
-from .elbo import *
+from .elbo import ELBO
diff --git a/tests/st/probability/zhusuan/vae/vae_mnist.py b/tests/st/probability/zhusuan/vae/vae_mnist.py
index a9cee6bf63..6e1aadac0d 100644
--- a/tests/st/probability/zhusuan/vae/vae_mnist.py
+++ b/tests/st/probability/zhusuan/vae/vae_mnist.py
@@ -67,13 +67,13 @@ class Generator(zs.BayesianNet):
 
         z_mean = self.zeros((self.batch_size, self.z_dim))
         z_std = self.ones((self.batch_size, self.z_dim))
-        z, log_prob_z = self.Normal('latent', observation=z, mean=z_mean, std=z_std, shape=(), reparameterize=False)
+        z, log_prob_z = self.normal('latent', observation=z, mean=z_mean, std=z_std, shape=(), reparameterize=False)
 
         x_mean = self.sigmoid(self.fc3(self.act2(self.fc2(self.act1(self.fc1(z))))))
         if x is None:
             #x = self.bernoulli_dist('sample', (), x_mean)
             x = x_mean
-        x, log_prob_x = self.Bernoulli('data', observation=x, shape=(), probs=x_mean)
+        x, log_prob_x = self.bernoulli('data', observation=x, shape=(), probs=x_mean)
 
         return x, log_prob_x, z, log_prob_z
 
@@ -109,7 +109,7 @@ class Variational(zs.BayesianNet):
         z_mean = self.fc3(z_logit)
         z_std = self.exp(self.fc4(z_logit))
         #z, log_prob_z = self.reparameterization(z_mean, z_std)
-        z, log_prob_z = self.Normal('latent', observation=z, mean=z_mean, std=z_std, shape=(), reparameterize=True)
+        z, log_prob_z = self.normal('latent', observation=z, mean=z_mean, std=z_std, shape=(), reparameterize=True)
         return z, log_prob_z
 
 def main():