diff --git a/mindspore/nn/probability/distribution/__init__.py b/mindspore/nn/probability/distribution/__init__.py
index 0d35d2ed8a1..c4077376a58 100644
--- a/mindspore/nn/probability/distribution/__init__.py
+++ b/mindspore/nn/probability/distribution/__init__.py
@@ -22,6 +22,7 @@ from .bernoulli import Bernoulli
 from .categorical import Categorical
 from .cauchy import Cauchy
 from .exponential import Exponential
+from .gamma import Gamma
 from .geometric import Geometric
 from .gumbel import Gumbel
 from .logistic import Logistic
@@ -36,6 +37,7 @@ __all__ = ['Distribution',
            'Categorical',
            'Cauchy',
            'Exponential',
+           'Gamma',
            'Geometric',
            'Gumbel',
            'Logistic',
diff --git a/mindspore/nn/probability/distribution/gamma.py b/mindspore/nn/probability/distribution/gamma.py
new file mode 100644
index 00000000000..17e946aca4e
--- /dev/null
+++ b/mindspore/nn/probability/distribution/gamma.py
@@ -0,0 +1,338 @@
+# 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.
+# ============================================================================
+"""Gamma Distribution"""
+import numpy as np
+from mindspore.ops import operations as P
+from mindspore.ops import composite as C
+import mindspore.nn as nn
+from mindspore._checkparam import Validator
+from mindspore.common import dtype as mstype
+from .distribution import Distribution
+from ._utils.utils import check_greater_zero, check_distribution_name
+from ._utils.custom_ops import log_generic
+
+
+class Gamma(Distribution):
+    """
+    Gamma distribution.
+
+    Args:
+        concentration (int, float, list, numpy.ndarray, Tensor, Parameter): The concentration,
+          also know as alpha of the Gamma distribution.
+        rate (int, float, list, numpy.ndarray, Tensor, Parameter): The rate, also know as
+          beta of the Gamma distribution.
+        seed (int): The seed used in sampling. The global seed is used if it is None. Default: None.
+        dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
+        name (str): The name of the distribution. Default: 'Gamma'.
+
+    Note:
+        `concentration` and `rate` must be greater than zero.
+        `dist_spec_args` are `concentration` and `rate`.
+        `dtype` must be a float type because Gamma distributions are continuous.
+
+    Examples:
+        >>> # To initialize a Gamma distribution of the concentration 3.0 and the rate 4.0.
+        >>> import mindspore.nn.probability.distribution as msd
+        >>> g = msd.Gamma(3.0, 4.0, dtype=mstype.float32)
+        >>>
+        >>> # The following creates two independent Gamma distributions.
+        >>> g = msd.Gamma([3.0, 3.0], [4.0, 4.0], dtype=mstype.float32)
+        >>>
+        >>> # A Gamma distribution can be initilized without arguments.
+        >>> # In this case, `concentration` and `rate` must be passed in through arguments.
+        >>> g = msd.Gamma(dtype=mstype.float32)
+        >>>
+        >>> # To use a Gamma distribution in a network.
+        >>> class net(Cell):
+        ...     def __init__(self):
+        ...         super(net, self).__init__():
+        ...         self.g1 = msd.Gamma(1.0, 1.0, dtype=mstype.float32)
+        ...         self.g2 = msd.Gamma(dtype=mstype.float32)
+        ...
+        ...     # The following calls are valid in construct.
+        ...     def construct(self, value, concentration_b, rate_b, concentration_a, rate_a):
+        ...
+        ...         # 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.
+        ...         # Args:
+        ...         #     value (Tensor): the value to be evaluated.
+        ...         #     concentration (Tensor): the concentration of the distribution. Default: self._concentration.
+        ...         #     rate (Tensor): the rate of the distribution. Default: self._rate.
+        ...
+        ...         # Examples of `prob`.
+        ...         # Similar calls can be made to other probability functions
+        ...         # by replacing 'prob' by the name of the function
+        ...         ans = self.g1.prob(value)
+        ...         # Evaluate with respect to the distribution b.
+        ...         ans = self.g1.prob(value, concentration_b, rate_b)
+        ...         # `concentration` and `rate` must be passed in during function calls
+        ...         ans = self.g2.prob(value, concentration_a, rate_a)
+        ...
+        ...
+        ...         # Functions `concentration`, `rate`, `mean`, `sd`, `var`, and `entropy` have the same arguments.
+        ...         # Args:
+        ...         #     concentration (Tensor): the concentration of the distribution. Default: self._concentration.
+        ...         #     rate (Tensor): the rate of the distribution. Default: self._rate.
+        ...
+        ...         # Example of `concentration`, `rate`, `mean`. `sd`, `var`, and `entropy` are similar.
+        ...         ans = self.g1.concentration() # return 1.0
+        ...         ans = self.g1.concentration(concentration_b, rate_b) # return concentration_b
+        ...         # `concentration` and `rate` must be passed in during function calls.
+        ...         ans = self.g2.concentration(concentration_a, rate_a)
+        ...
+        ...
+        ...         # Interfaces of 'kl_loss' and 'cross_entropy' are the same:
+        ...         # Args:
+        ...         #     dist (str): the type of the distributions. Only "Gamma" is supported.
+        ...         #     concentration_b (Tensor): the concentration of distribution b.
+        ...         #     rate_b (Tensor): the rate of distribution b.
+        ...         #     concentration_a (Tensor): the concentration of distribution a. Default: self._concentration.
+        ...         #     rate_a (Tensor): the rate of distribution a. Default: self._rate.
+        ...
+        ...         # Examples of `kl_loss`. `cross_entropy` is similar.
+        ...         ans = self.g1.kl_loss('Gamma', concentration_b, rate_b)
+        ...         ans = self.g1.kl_loss('Gamma', concentration_b, rate_b, concentration_a, rate_a)
+        ...         # Additional `concentration` and `rate` must be passed in.
+        ...         ans = self.g2.kl_loss('Gamma', concentration_b, rate_b, concentration_a, rate_a)
+        ...
+        ...
+        ...         # Examples of `sample`.
+        ...         # Args:
+        ...         #     shape (tuple): the shape of the sample. Default: ()
+        ...         #     concentration (Tensor): the concentration of the distribution. Default: self._concentration.
+        ...         #     rate (Tensor): the rate of the distribution. Default: self._rate.
+        ...         ans = self.g1.sample()
+        ...         ans = self.g1.sample((2,3))
+        ...         ans = self.g1.sample((2,3), concentration_b, rate_b)
+        ...         ans = self.g2.sample((2,3), concentration_a, rate_a)
+    """
+
+    def __init__(self,
+                 concentration=None,
+                 rate=None,
+                 seed=None,
+                 dtype=mstype.float32,
+                 name="Gamma"):
+        """
+        Constructor of Gamma.
+        """
+        param = dict(locals())
+        param['param_dict'] = {'concentration': concentration, 'rate': rate}
+        valid_dtype = mstype.float_type
+        Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
+        super(Gamma, self).__init__(seed, dtype, name, param)
+
+        self._concentration = self._add_parameter(concentration, 'concentration')
+        self._rate = self._add_parameter(rate, 'rate')
+        if self._concentration is not None:
+            check_greater_zero(self._concentration, "concentration")
+        if self._rate is not None:
+            check_greater_zero(self._rate, "rate")
+
+        # ops needed for the class
+        self.log = log_generic
+        self.square = P.Square()
+        self.sqrt = P.Sqrt()
+        self.squeeze = P.Squeeze(0)
+        self.cast = P.Cast()
+        self.dtypeop = P.DType()
+        self.fill = P.Fill()
+        self.shape = P.Shape()
+        self.select = P.Select()
+        self.greater = P.Greater()
+        self.lgamma = nn.LGamma()
+        self.digamma = nn.DiGamma()
+        self.igamma = nn.IGamma()
+
+    def extend_repr(self):
+        if self.is_scalar_batch:
+            s = f'concentration = {self._concentration}, rate = {self._rate}'
+        else:
+            s = f'batch_shape = {self._broadcast_shape}'
+        return s
+
+    @property
+    def concentration(self):
+        """
+        Return the concentration, also know as the alpha of the Gamma distribution.
+        """
+        return self._concentration
+
+    @property
+    def rate(self):
+        """
+        Return the rate, also know as the beta of the Gamma distribution.
+        """
+        return self._rate
+
+    def _get_dist_type(self):
+        return "Gamma"
+
+    def _get_dist_args(self, concentration=None, rate=None):
+        if concentration is not None:
+            self.checktensor(concentration, 'concentration')
+        else:
+            concentration = self._concentration
+        if rate is not None:
+            self.checktensor(rate, 'rate')
+        else:
+            rate = self._rate
+        return concentration, rate
+
+    def _mean(self, concentration=None, rate=None):
+        """
+        The mean of the distribution.
+        """
+        concentration, rate = self._check_param_type(concentration, rate)
+        return concentration / rate
+
+    def _var(self, concentration=None, rate=None):
+        """
+        The variance of the distribution.
+        """
+        concentration, rate = self._check_param_type(concentration, rate)
+        return concentration / self.square(rate)
+
+    def _sd(self, concentration=None, rate=None):
+        """
+        The standard deviation of the distribution.
+        """
+        concentration, rate = self._check_param_type(concentration, rate)
+        return self.sqrt(concentration) / rate
+
+    def _mode(self, concentration=None, rate=None):
+        """
+        The mode of the distribution.
+        """
+        concentration, rate = self._check_param_type(concentration, rate)
+        mode = (concentration - 1.) / rate
+        nan = self.fill(self.dtypeop(concentration), self.shape(concentration), np.nan)
+        comp = self.greater(concentration, 1.)
+        return self.select(comp, mode, nan)
+
+    def _entropy(self, concentration=None, rate=None):
+        r"""
+        Evaluate entropy.
+
+        .. math::
+            H(X) = \alpha - \log(\beta) + \log(\Gamma(\alpha)) + (1 - \alpha) * \digamma(\alpha)
+        """
+        concentration, rate = self._check_param_type(concentration, rate)
+        return concentration - self.log(rate) + self.lgamma(concentration) \
+               + (1. - concentration) * self.digamma(concentration)
+
+    def _cross_entropy(self, dist, concentration_b, rate_b, concentration=None, rate=None):
+        r"""
+        Evaluate cross entropy between Gamma distributions.
+
+        Args:
+            dist (str): Type of the distributions. Should be "Gamma" in this case.
+            concentration_b (Tensor): concentration of distribution b.
+            rate_b (Tensor): rate of distribution b.
+            concentration_a (Tensor): concentration of distribution a. Default: self._concentration.
+            rate_a (Tensor): rate of distribution a. Default: self._rate.
+        """
+        check_distribution_name(dist, 'Gamma')
+        return self._entropy(concentration, rate) + self._kl_loss(dist, concentration_b, rate_b, concentration, rate)
+
+    def _log_prob(self, value, concentration=None, rate=None):
+        r"""
+        Evaluate log probability.
+
+        Args:
+            value (Tensor): The value to be evaluated.
+            concentration (Tensor): The concentration of the distribution. Default: self._concentration.
+            rate (Tensor): The rate the distribution. Default: self._rate.
+
+        .. math::
+            L(x) = (\alpha - 1) * \log(x) - \beta * x - \log(\gamma(\alpha)) - \alpha * \log(\beta)
+        """
+        value = self._check_value(value, 'value')
+        value = self.cast(value, self.dtype)
+        concentration, rate = self._check_param_type(concentration, rate)
+        unnormalized_log_prob = (concentration - 1.) * self.log(value) - rate * value
+        log_normalization = self.lgamma(concentration) - concentration * self.log(rate)
+        return unnormalized_log_prob - log_normalization
+
+    def _cdf(self, value, concentration=None, rate=None):
+        r"""
+        Evaluate the cumulative distribution function on the given value. Note that igamma returns
+        the regularized incomplete gamma function, which is what we want for the CDF.
+
+        Args:
+            value (Tensor): The value to be evaluated.
+            concentration (Tensor): The concentration of the distribution. Default: self._concentration.
+            rate (Tensor): The rate the distribution. Default: self._rate.
+
+        .. math::
+            cdf(x) = \igamma(\alpha, \beta * x)
+        """
+        value = self._check_value(value, 'value')
+        value = self.cast(value, self.dtype)
+        concentration, rate = self._check_param_type(concentration, rate)
+        return self.igamma(concentration, rate * value)
+
+    def _kl_loss(self, dist, concentration_b, rate_b, concentration=None, rate=None):
+        r"""
+        Evaluate Gamma-Gamma KL divergence, i.e. KL(a||b).
+
+        Args:
+            dist (str): The type of the distributions. Should be "Gamma" in this case.
+            concentration_b (Tensor): The concentration of distribution b.
+            rate_b (Tensor): The rate distribution b.
+            concentration_a (Tensor): The concentration of distribution a. Default: self._concentration.
+            rate_a (Tensor): The rate distribution a. Default: self._rate.
+
+        .. math::
+            KL(a||b) = (\alpha_{a} - \alpha_{b}) * \digamma(\alpha_{a}) + \log(\gamma(\alpha_{b}))
+                       - \log(\gamma(\alpha_{a})) + \alpha_{b} * \log(\beta{a}) - \alpha_{b} * \log(\beta{b})
+                       + \alpha_{a} * \frac{\beta{b}}{\beta{a} - 1}
+        """
+        check_distribution_name(dist, 'Gamma')
+        concentration_b = self._check_value(concentration_b, 'concentration_b')
+        rate_b = self._check_value(rate_b, 'rate_b')
+        concentration_b = self.cast(concentration_b, self.parameter_type)
+        rate_b = self.cast(rate_b, self.parameter_type)
+        concentration_a, rate_a = self._check_param_type(concentration, rate)
+        return (concentration_a - concentration_b) * self.digamma(concentration_a) \
+               + self.lgamma(concentration_b)  - self.lgamma(concentration_a) \
+               + concentration_b * self.log(rate_a) - concentration_b * self.log(rate_b) \
+               + concentration_a * (rate_b / rate_a - 1.)
+
+    def _sample(self, shape=(), concentration=None, rate=None):
+        """
+        Sampling.
+
+        Args:
+            shape (tuple): The shape of the sample. Default: ().
+            concentration (Tensor): The concentration of the samples. Default: self._concentration.
+            rate (Tensor): The rate of the samples. Default: self._rate.
+
+        Returns:
+            Tensor, with the shape being shape + batch_shape.
+        """
+        shape = self.checktuple(shape, 'shape')
+        concentration, rate = self._check_param_type(concentration, rate)
+        batch_shape = self.shape(concentration + rate)
+        origin_shape = shape + batch_shape
+        if origin_shape == ():
+            sample_shape = (1,)
+        else:
+            sample_shape = origin_shape
+        sample_gamma = C.gamma(sample_shape, concentration, rate, self.seed)
+        value = self.cast(sample_gamma, self.dtype)
+        if origin_shape == ():
+            value = self.squeeze(value)
+        return value
diff --git a/tests/st/probability/distribution/test_gamma.py b/tests/st/probability/distribution/test_gamma.py
new file mode 100644
index 00000000000..57e45844e4e
--- /dev/null
+++ b/tests/st/probability/distribution/test_gamma.py
@@ -0,0 +1,328 @@
+# 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.
+# ============================================================================
+"""test cases for Gamma distribution"""
+import numpy as np
+from scipy import stats
+from scipy import special
+import mindspore.context as context
+import mindspore.nn as nn
+import mindspore.nn.probability.distribution as msd
+from mindspore import Tensor
+from mindspore import dtype
+
+context.set_context(mode=context.GRAPH_MODE, device_target="Ascend")
+
+class Prob(nn.Cell):
+    """
+    Test class: probability of Gamma distribution.
+    """
+    def __init__(self):
+        super(Prob, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.prob(x_)
+
+def test_pdf():
+    """
+    Test pdf.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_pdf = gamma_benchmark.pdf([1.0, 2.0]).astype(np.float32)
+    pdf = Prob()
+    output = pdf(Tensor([1.0, 2.0], dtype=dtype.float32))
+    tol = 1e-6
+    assert (np.abs(output.asnumpy() - expect_pdf) < tol).all()
+
+class LogProb(nn.Cell):
+    """
+    Test class: log probability of Gamma distribution.
+    """
+    def __init__(self):
+        super(LogProb, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.log_prob(x_)
+
+def test_log_likelihood():
+    """
+    Test log_pdf.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_logpdf = gamma_benchmark.logpdf([1.0, 2.0]).astype(np.float32)
+    logprob = LogProb()
+    output = logprob(Tensor([1.0, 2.0], dtype=dtype.float32))
+    tol = 1e-6
+    assert (np.abs(output.asnumpy() - expect_logpdf) < tol).all()
+
+
+class KL(nn.Cell):
+    """
+    Test class: kl_loss of Gamma distribution.
+    """
+    def __init__(self):
+        super(KL, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([4.0]), dtype=dtype.float32)
+
+    def construct(self, x_, y_):
+        return self.g.kl_loss('Gamma', x_, y_)
+
+
+def test_kl_loss():
+    """
+    Test kl_loss.
+    """
+    concentration_a = np.array([3.0]).astype(np.float32)
+    rate_a = np.array([4.0]).astype(np.float32)
+
+    concentration_b = np.array([1.0]).astype(np.float32)
+    rate_b = np.array([1.0]).astype(np.float32)
+
+    expect_kl_loss = (concentration_a - concentration_b) * special.digamma(concentration_a) \
+                     + special.gammaln(concentration_b) - special.gammaln(concentration_a) \
+                     + concentration_b * np.log(rate_a) - concentration_b * np.log(rate_b) \
+                     + concentration_a * (rate_b / rate_a - 1.)
+
+    kl_loss = KL()
+    concentration = Tensor(concentration_b, dtype=dtype.float32)
+    rate = Tensor(rate_b, dtype=dtype.float32)
+    output = kl_loss(concentration, rate)
+    tol = 1e-6
+    assert (np.abs(output.asnumpy() - expect_kl_loss) < tol).all()
+
+class Basics(nn.Cell):
+    """
+    Test class: mean/sd/mode of Gamma distribution.
+    """
+    def __init__(self):
+        super(Basics, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self):
+        return self.g.mean(), self.g.sd(), self.g.mode()
+
+def test_basics():
+    """
+    Test mean/standard deviation/mode.
+    """
+    basics = Basics()
+    mean, sd, mode = basics()
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_mean = gamma_benchmark.mean().astype(np.float32)
+    expect_sd = gamma_benchmark.std().astype(np.float32)
+    expect_mode = [2.0]
+    tol = 1e-6
+    assert (np.abs(mean.asnumpy() - expect_mean) < tol).all()
+    assert (np.abs(mode.asnumpy() - expect_mode) < tol).all()
+    assert (np.abs(sd.asnumpy() - expect_sd) < tol).all()
+
+class Sampling(nn.Cell):
+    """
+    Test class: sample of Gamma distribution.
+    """
+    def __init__(self, shape, seed=0):
+        super(Sampling, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), seed=seed, dtype=dtype.float32)
+        self.shape = shape
+
+    def construct(self, concentration=None, rate=None):
+        return self.g.sample(self.shape, concentration, rate)
+
+def test_sample():
+    """
+    Test sample.
+    """
+    shape = (2, 3)
+    seed = 10
+    concentration = Tensor([2.0], dtype=dtype.float32)
+    rate = Tensor([2.0, 2.0, 2.0], dtype=dtype.float32)
+    sample = Sampling(shape, seed=seed)
+    output = sample(concentration, rate)
+    assert output.shape == (2, 3, 3)
+
+class CDF(nn.Cell):
+    """
+    Test class: cdf of Gamma distribution.
+    """
+    def __init__(self):
+        super(CDF, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.cdf(x_)
+
+
+def test_cdf():
+    """
+    Test cdf.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_cdf = gamma_benchmark.cdf([2.0]).astype(np.float32)
+    cdf = CDF()
+    output = cdf(Tensor([2.0], dtype=dtype.float32))
+    tol = 2e-5
+    assert (np.abs(output.asnumpy() - expect_cdf) < tol).all()
+
+class LogCDF(nn.Cell):
+    """
+    Test class: log_cdf of Mormal distribution.
+    """
+    def __init__(self):
+        super(LogCDF, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.log_cdf(x_)
+
+def test_log_cdf():
+    """
+    Test log cdf.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_logcdf = gamma_benchmark.logcdf([2.0]).astype(np.float32)
+    logcdf = LogCDF()
+    output = logcdf(Tensor([2.0], dtype=dtype.float32))
+    tol = 5e-5
+    assert (np.abs(output.asnumpy() - expect_logcdf) < tol).all()
+
+class SF(nn.Cell):
+    """
+    Test class: survival function of Gamma distribution.
+    """
+    def __init__(self):
+        super(SF, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.survival_function(x_)
+
+def test_survival():
+    """
+    Test log_survival.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_survival = gamma_benchmark.sf([2.0]).astype(np.float32)
+    survival_function = SF()
+    output = survival_function(Tensor([2.0], dtype=dtype.float32))
+    tol = 2e-5
+    assert (np.abs(output.asnumpy() - expect_survival) < tol).all()
+
+class LogSF(nn.Cell):
+    """
+    Test class: log survival function of Gamma distribution.
+    """
+    def __init__(self):
+        super(LogSF, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_):
+        return self.g.log_survival(x_)
+
+def test_log_survival():
+    """
+    Test log_survival.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_log_survival = gamma_benchmark.logsf([2.0]).astype(np.float32)
+    log_survival = LogSF()
+    output = log_survival(Tensor([2.0], dtype=dtype.float32))
+    tol = 2e-5
+    assert (np.abs(output.asnumpy() - expect_log_survival) < tol).all()
+
+class EntropyH(nn.Cell):
+    """
+    Test class: entropy of Gamma distribution.
+    """
+    def __init__(self):
+        super(EntropyH, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self):
+        return self.g.entropy()
+
+def test_entropy():
+    """
+    Test entropy.
+    """
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_entropy = gamma_benchmark.entropy().astype(np.float32)
+    entropy = EntropyH()
+    output = entropy()
+    tol = 1e-6
+    assert (np.abs(output.asnumpy() - expect_entropy) < tol).all()
+
+class CrossEntropy(nn.Cell):
+    """
+    Test class: cross entropy between Gamma distributions.
+    """
+    def __init__(self):
+        super(CrossEntropy, self).__init__()
+        self.g = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_, y_):
+        entropy = self.g.entropy()
+        kl_loss = self.g.kl_loss('Gamma', x_, y_)
+        h_sum_kl = entropy + kl_loss
+        cross_entropy = self.g.cross_entropy('Gamma', x_, y_)
+        return h_sum_kl - cross_entropy
+
+def test_cross_entropy():
+    """
+    Test cross_entropy.
+    """
+    cross_entropy = CrossEntropy()
+    concentration = Tensor([3.0], dtype=dtype.float32)
+    rate = Tensor([2.0], dtype=dtype.float32)
+    diff = cross_entropy(concentration, rate)
+    tol = 1e-6
+    assert (np.abs(diff.asnumpy() - np.zeros(diff.shape)) < tol).all()
+
+class Net(nn.Cell):
+    """
+    Test class: expand single distribution instance to multiple graphs
+    by specifying the attributes.
+    """
+
+    def __init__(self):
+        super(Net, self).__init__()
+        self.Gamma = msd.Gamma(np.array([3.0]), np.array([1.0]), dtype=dtype.float32)
+
+    def construct(self, x_, y_):
+        kl = self.Gamma.kl_loss('Gamma', x_, y_)
+        prob = self.Gamma.prob(kl)
+        return prob
+
+def test_multiple_graphs():
+    """
+    Test multiple graphs case.
+    """
+    prob = Net()
+    concentration_a = np.array([3.0]).astype(np.float32)
+    rate_a = np.array([1.0]).astype(np.float32)
+    concentration_b = np.array([2.0]).astype(np.float32)
+    rate_b = np.array([1.0]).astype(np.float32)
+    ans = prob(Tensor(concentration_b), Tensor(rate_b))
+
+    expect_kl_loss = (concentration_a - concentration_b) * special.digamma(concentration_a) \
+                     + special.gammaln(concentration_b) - special.gammaln(concentration_a) \
+                     + concentration_b * np.log(rate_a) - concentration_b * np.log(rate_b) \
+                     + concentration_a * (rate_b / rate_a - 1.)
+
+    gamma_benchmark = stats.gamma(np.array([3.0]))
+    expect_prob = gamma_benchmark.pdf(expect_kl_loss).astype(np.float32)
+
+    tol = 1e-6
+    assert (np.abs(ans.asnumpy() - expect_prob) < tol).all()
diff --git a/tests/ut/python/nn/probability/distribution/test_gamma.py b/tests/ut/python/nn/probability/distribution/test_gamma.py
new file mode 100644
index 00000000000..9e9ea58e755
--- /dev/null
+++ b/tests/ut/python/nn/probability/distribution/test_gamma.py
@@ -0,0 +1,214 @@
+# 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.
+# ============================================================================
+"""
+Test nn.probability.distribution.Gamma.
+"""
+import numpy as np
+import pytest
+
+import mindspore.nn as nn
+import mindspore.nn.probability.distribution as msd
+from mindspore import dtype
+from mindspore import Tensor
+
+def test_gamma_shape_errpr():
+    """
+    Invalid shapes.
+    """
+    with pytest.raises(ValueError):
+        msd.Gamma([[2.], [1.]], [[2.], [3.], [4.]], dtype=dtype.float32)
+
+def test_type():
+    with pytest.raises(TypeError):
+        msd.Gamma(0., 1., dtype=dtype.int32)
+
+def test_name():
+    with pytest.raises(TypeError):
+        msd.Gamma(0., 1., name=1.0)
+
+def test_seed():
+    with pytest.raises(TypeError):
+        msd.Gamma(0., 1., seed='seed')
+
+def test_rate():
+    with pytest.raises(ValueError):
+        msd.Gamma(0., 0.)
+    with pytest.raises(ValueError):
+        msd.Gamma(0., -1.)
+
+def test_arguments():
+    """
+    args passing during initialization.
+    """
+    g = msd.Gamma()
+    assert isinstance(g, msd.Distribution)
+    g = msd.Gamma([3.0], [4.0], dtype=dtype.float32)
+    assert isinstance(g, msd.Distribution)
+
+
+class GammaProb(nn.Cell):
+    """
+    Gamma distribution: initialize with concentration/rate.
+    """
+    def __init__(self):
+        super(GammaProb, self).__init__()
+        self.gamma = msd.Gamma([3.0, 4.0], [1.0, 1.0], dtype=dtype.float32)
+
+    def construct(self, value):
+        prob = self.gamma.prob(value)
+        log_prob = self.gamma.log_prob(value)
+        cdf = self.gamma.cdf(value)
+        log_cdf = self.gamma.log_cdf(value)
+        sf = self.gamma.survival_function(value)
+        log_sf = self.gamma.log_survival(value)
+        return prob + log_prob + cdf + log_cdf + sf + log_sf
+
+def test_gamma_prob():
+    """
+    Test probability functions: passing value through construct.
+    """
+    net = GammaProb()
+    value = Tensor([0.5, 1.0], dtype=dtype.float32)
+    ans = net(value)
+    assert isinstance(ans, Tensor)
+
+
+class GammaProb1(nn.Cell):
+    """
+    Gamma distribution: initialize without concentration/rate.
+    """
+    def __init__(self):
+        super(GammaProb1, self).__init__()
+        self.gamma = msd.Gamma()
+
+    def construct(self, value, concentration, rate):
+        prob = self.gamma.prob(value, concentration, rate)
+        log_prob = self.gamma.log_prob(value, concentration, rate)
+        cdf = self.gamma.cdf(value, concentration, rate)
+        log_cdf = self.gamma.log_cdf(value, concentration, rate)
+        sf = self.gamma.survival_function(value, concentration, rate)
+        log_sf = self.gamma.log_survival(value, concentration, rate)
+        return prob + log_prob + cdf + log_cdf + sf + log_sf
+
+def test_gamma_prob1():
+    """
+    Test probability functions: passing concentration/rate, value through construct.
+    """
+    net = GammaProb1()
+    value = Tensor([0.5, 1.0], dtype=dtype.float32)
+    concentration = Tensor([2.0, 3.0], dtype=dtype.float32)
+    rate = Tensor([1.0], dtype=dtype.float32)
+    ans = net(value, concentration, rate)
+    assert isinstance(ans, Tensor)
+
+class GammaKl(nn.Cell):
+    """
+    Test class: kl_loss of Gamma distribution.
+    """
+    def __init__(self):
+        super(GammaKl, self).__init__()
+        self.g1 = msd.Gamma(np.array([3.0]), np.array([4.0]), dtype=dtype.float32)
+        self.g2 = msd.Gamma(dtype=dtype.float32)
+
+    def construct(self, concentration_b, rate_b, concentration_a, rate_a):
+        kl1 = self.g1.kl_loss('Gamma', concentration_b, rate_b)
+        kl2 = self.g2.kl_loss('Gamma', concentration_b, rate_b, concentration_a, rate_a)
+        return kl1 + kl2
+
+def test_kl():
+    """
+    Test kl_loss.
+    """
+    net = GammaKl()
+    concentration_b = Tensor(np.array([1.0]).astype(np.float32), dtype=dtype.float32)
+    rate_b = Tensor(np.array([1.0]).astype(np.float32), dtype=dtype.float32)
+    concentration_a = Tensor(np.array([2.0]).astype(np.float32), dtype=dtype.float32)
+    rate_a = Tensor(np.array([3.0]).astype(np.float32), dtype=dtype.float32)
+    ans = net(concentration_b, rate_b, concentration_a, rate_a)
+    assert isinstance(ans, Tensor)
+
+class GammaCrossEntropy(nn.Cell):
+    """
+    Test class: cross_entropy of Gamma distribution.
+    """
+    def __init__(self):
+        super(GammaCrossEntropy, self).__init__()
+        self.g1 = msd.Gamma(np.array([3.0]), np.array([4.0]), dtype=dtype.float32)
+        self.g2 = msd.Gamma(dtype=dtype.float32)
+
+    def construct(self, concentration_b, rate_b, concentration_a, rate_a):
+        h1 = self.g1.cross_entropy('Gamma', concentration_b, rate_b)
+        h2 = self.g2.cross_entropy('Gamma', concentration_b, rate_b, concentration_a, rate_a)
+        return h1 + h2
+
+def test_cross_entropy():
+    """
+    Test cross entropy between Gamma distributions.
+    """
+    net = GammaCrossEntropy()
+    concentration_b = Tensor(np.array([1.0]).astype(np.float32), dtype=dtype.float32)
+    rate_b = Tensor(np.array([1.0]).astype(np.float32), dtype=dtype.float32)
+    concentration_a = Tensor(np.array([2.0]).astype(np.float32), dtype=dtype.float32)
+    rate_a = Tensor(np.array([3.0]).astype(np.float32), dtype=dtype.float32)
+    ans = net(concentration_b, rate_b, concentration_a, rate_a)
+    assert isinstance(ans, Tensor)
+
+class GammaBasics(nn.Cell):
+    """
+    Test class: basic mean/sd function.
+    """
+    def __init__(self):
+        super(GammaBasics, self).__init__()
+        self.g = msd.Gamma(np.array([3.0, 4.0]), np.array([4.0, 6.0]), dtype=dtype.float32)
+
+    def construct(self):
+        mean = self.g.mean()
+        sd = self.g.sd()
+        mode = self.g.mode()
+        return mean + sd + mode
+
+def test_bascis():
+    """
+    Test mean/sd/mode/entropy functionality of Gamma.
+    """
+    net = GammaBasics()
+    ans = net()
+    assert isinstance(ans, Tensor)
+
+class GammaConstruct(nn.Cell):
+    """
+    Gamma distribution: going through construct.
+    """
+    def __init__(self):
+        super(GammaConstruct, self).__init__()
+        self.gamma = msd.Gamma([3.0], [4.0])
+        self.gamma1 = msd.Gamma()
+
+    def construct(self, value, concentration, rate):
+        prob = self.gamma('prob', value)
+        prob1 = self.gamma('prob', value, concentration, rate)
+        prob2 = self.gamma1('prob', value, concentration, rate)
+        return prob + prob1 + prob2
+
+def test_gamma_construct():
+    """
+    Test probability function going through construct.
+    """
+    net = GammaConstruct()
+    value = Tensor([0.5, 1.0], dtype=dtype.float32)
+    concentration = Tensor([0.0], dtype=dtype.float32)
+    rate = Tensor([1.0], dtype=dtype.float32)
+    ans = net(value, concentration, rate)
+    assert isinstance(ans, Tensor)