def setUp(self):
self.program = paddle.static.Program()
self.executor = paddle.static.Executor()
with paddle.static.program_guard(self.program):
rate_np = config.xrand((100, 200, 99))
rate = paddle.static.data('rate', rate_np.shape, rate_np.dtype)
self.mock_dist = mock.Exponential(rate)
self.feeds = {'rate': rate_np}
@config.place(config.DEVICES)
@config.parameterize(
(config.TEST_CASE_NAME, 'dist'), [('test-mock-exp',
mock.Exponential(rate=paddle.rand(
[100, 200, 99],
dtype=config.DEFAULT_DTYPE)))])
class TestExponentialFamily(unittest.TestCase):
def test_entropy(self):
np.testing.assert_allclose(
self.dist.entropy(),
paddle.distribution.ExponentialFamily.entropy(self.dist),
rtol=config.RTOL.get(config.DEFAULT_DTYPE),
atol=config.ATOL.get(config.DEFAULT_DTYPE))
@config.place(config.DEVICES)
@config.parameterize(
(config.TEST_CASE_NAME, 'dist'),
[('test-dummy', mock.DummyExpFamily(0.5, 0.5)),
('test-dirichlet',
paddle.distribution.Dirichlet(paddle.to_tensor(config.xrand()))), (
'test-beta', paddle.distribution.Beta(
paddle.to_tensor(config.xrand()),
paddle.to_tensor(config.xrand())))])
class TestExponentialFamilyException(unittest.TestCase):
def test_entropy_exception(self):
with self.assertRaises(NotImplementedError):
paddle.distribution.ExponentialFamily.entropy(self.dist)
def test_prob(self):
with self.assertRaises(NotImplementedError):
self.dist.prob(paddle.to_tensor(config.xrand()))
def test_log_normalizer(self):
self.assertTrue(
np.all(
self._paddle_diric._log_normalizer(
paddle.to_tensor(config.xrand((100, 100,
100)))).numpy() < 0.0))
import unittest
import numpy as np
import paddle
import scipy.stats
import config
from config import (ATOL, DEVICES, RTOL, TEST_CASE_NAME, parameterize, place,
xrand)
@place(DEVICES)
@parameterize(
(TEST_CASE_NAME, 'concentration'),
[
('test-one-dim', config.xrand((89, ))),
# ('test-multi-dim', config.xrand((10, 20, 30)))
])
class TestDirichlet(unittest.TestCase):
def setUp(self):
self._paddle_diric = paddle.distribution.Dirichlet(
paddle.to_tensor(self.concentration))
def test_mean(self):
with paddle.fluid.dygraph.guard(self.place):
np.testing.assert_allclose(
self._paddle_diric.mean,
scipy.stats.dirichlet.mean(self.concentration),
rtol=RTOL.get(str(self.concentration.dtype)),
atol=ATOL.get(str(self.concentration.dtype)))
# See the License for the specific language governing permissions and
# limitations under the License.
import numbers
import unittest
import numpy as np
import paddle
import scipy.stats
from config import (ATOL, DEVICES, RTOL, TEST_CASE_NAME, parameterize, place,
xrand)
@place(DEVICES)
@parameterize((TEST_CASE_NAME, 'alpha', 'beta'),
[('test-scale', 1.0, 2.0), ('test-tensor', xrand(), xrand()),
('test-broadcast', xrand((2, 1)), xrand((2, 5)))])
class TestBeta(unittest.TestCase):
def setUp(self):
# scale no need convert to tensor for scale input unittest
alpha, beta = self.alpha, self.beta
if not isinstance(self.alpha, numbers.Real):
alpha = paddle.to_tensor(self.alpha)
if not isinstance(self.beta, numbers.Real):
beta = paddle.to_tensor(self.beta)
self._paddle_beta = paddle.distribution.Beta(alpha, beta)
def test_mean(self):
with paddle.fluid.dygraph.guard(self.place):
np.testing.assert_allclose(