Cleanup of a few things in distributions library. Noticed while putting

together change for MVNFullCovariance.

* Using assert_none_equal
* Removing redundant asserts in mvn_tril.

PiperOrigin-RevId: 155707155

7 files changed, 39 insertions, 21 deletions

diff --git a/tensorflow/contrib/distributions/python/kernel_tests/mvn_diag_test.py b/tensorflow/contrib/distributions/python/kernel_tests/mvn_diag_test.py
index 406cd4ebbe..3f4582eb7e 100644
--- a/tensorflow/contrib/distributions/python/kernel_tests/mvn_diag_test.py
+++ b/tensorflow/contrib/distributions/python/kernel_tests/mvn_diag_test.py
@@ -103,6 +103,14 @@ class MultivariateNormalDiagTest(test.TestCase):
       self.assertAllClose(cov_mat, np.cov(samps.T),
                           atol=0.05, rtol=0.05)
 
+  def testSingularScaleRaises(self):
+    mu = [-1., 1]
+    diag = [1., 0]
+    with self.test_session():
+      dist = ds.MultivariateNormalDiag(mu, diag, validate_args=True)
+      with self.assertRaisesOpError("Singular"):
+        dist.sample().eval()
+
   def testSampleWithBroadcastScale(self):
     # mu corresponds to a 2-batch of 3-variate normals
     mu = np.zeros([2, 3])
diff --git a/tensorflow/contrib/distributions/python/kernel_tests/mvn_tril_test.py b/tensorflow/contrib/distributions/python/kernel_tests/mvn_tril_test.py
index 8c6980ca57..685f32883d 100644
--- a/tensorflow/contrib/distributions/python/kernel_tests/mvn_tril_test.py
+++ b/tensorflow/contrib/distributions/python/kernel_tests/mvn_tril_test.py
@@ -151,6 +151,14 @@ class MultivariateNormalTriLTest(test.TestCase):
       self.assertAllClose(sample_values.mean(axis=0), mu, atol=1e-2)
       self.assertAllClose(np.cov(sample_values, rowvar=0), sigma, atol=0.06)
 
+  def testSingularScaleRaises(self):
+    with self.test_session():
+      mu = None
+      chol = [[1., 0.], [0., 0.]]
+      mvn = ds.MultivariateNormalTriL(mu, chol, validate_args=True)
+      with self.assertRaisesOpError("Singular operator"):
+        mvn.sample().eval()
+
   def testSampleWithSampleShape(self):
     with self.test_session():
       mu = self._rng.rand(3, 5, 2)
diff --git a/tensorflow/contrib/distributions/python/kernel_tests/vector_laplace_diag_test.py b/tensorflow/contrib/distributions/python/kernel_tests/vector_laplace_diag_test.py
index fc04acd90c..c355adeedb 100644
--- a/tensorflow/contrib/distributions/python/kernel_tests/vector_laplace_diag_test.py
+++ b/tensorflow/contrib/distributions/python/kernel_tests/vector_laplace_diag_test.py
@@ -92,6 +92,14 @@ class VectorLaplaceDiagTest(test.TestCase):
       self.assertAllClose(cov_mat, np.cov(samps.T),
                           atol=0.05, rtol=0.05)
 
+  def testSingularScaleRaises(self):
+    mu = [-1., 1]
+    diag = [1., 0]
+    with self.test_session():
+      dist = ds.VectorLaplaceDiag(mu, diag, validate_args=True)
+      with self.assertRaisesOpError("Singular"):
+        dist.sample().eval()
+
   def testSampleWithBroadcastScale(self):
     # mu corresponds to a 2-batch of 3-variate normals
     mu = np.zeros([2, 3])
diff --git a/tensorflow/contrib/distributions/python/ops/distribution_util.py b/tensorflow/contrib/distributions/python/ops/distribution_util.py
index 370b4f4a06..5e3b42dd2a 100644
--- a/tensorflow/contrib/distributions/python/ops/distribution_util.py
+++ b/tensorflow/contrib/distributions/python/ops/distribution_util.py
@@ -25,7 +25,6 @@ from tensorflow.python.framework import tensor_util
 from tensorflow.python.ops import array_ops
 from tensorflow.python.ops import check_ops
 from tensorflow.python.ops import control_flow_ops
-from tensorflow.python.ops import math_ops
 from tensorflow.python.ops.distributions import util
 from tensorflow.python.ops.distributions.util import *  # pylint: disable=wildcard-import
 
@@ -45,13 +44,11 @@ def make_diag_scale(loc, scale_diag, scale_identity_multiplier,
           check_ops.assert_positive(
               x, message="diagonal part must be positive"),
       ], x)
-    # TODO(b/35157376): Use `assert_none_equal` once it exists.
     return control_flow_ops.with_dependencies([
-        check_ops.assert_greater(
-            math_ops.abs(x),
+        check_ops.assert_none_equal(
+            x,
             array_ops.zeros([], x.dtype),
-            message="diagonal part must be non-zero"),
-    ], x)
+            message="diagonal part must be non-zero")], x)
 
   with ops.name_scope(name, "make_diag_scale",
                       values=[loc, scale_diag, scale_identity_multiplier]):
diff --git a/tensorflow/contrib/distributions/python/ops/mvn_diag.py b/tensorflow/contrib/distributions/python/ops/mvn_diag.py
index e57145b42f..163cf75d99 100644
--- a/tensorflow/contrib/distributions/python/ops/mvn_diag.py
+++ b/tensorflow/contrib/distributions/python/ops/mvn_diag.py
@@ -197,11 +197,14 @@ class MultivariateNormalDiag(
     with ops.name_scope(name):
       with ops.name_scope("init", values=[
           loc, scale_diag, scale_identity_multiplier]):
+        # No need to validate_args while making diag_scale.  The returned
+        # LinearOperatorDiag has an assert_non_singular method that is called by
+        # the Bijector.
         scale = distribution_util.make_diag_scale(
             loc=loc,
             scale_diag=scale_diag,
             scale_identity_multiplier=scale_identity_multiplier,
-            validate_args=validate_args,
+            validate_args=False,
             assert_positive=False)
     super(MultivariateNormalDiag, self).__init__(
         loc=loc,
diff --git a/tensorflow/contrib/distributions/python/ops/mvn_tril.py b/tensorflow/contrib/distributions/python/ops/mvn_tril.py
index dd5933f62b..d662b25e1e 100644
--- a/tensorflow/contrib/distributions/python/ops/mvn_tril.py
+++ b/tensorflow/contrib/distributions/python/ops/mvn_tril.py
@@ -21,10 +21,6 @@ from __future__ import print_function
 from tensorflow.contrib import linalg
 from tensorflow.contrib.distributions.python.ops import mvn_linear_operator as mvn_linop
 from tensorflow.python.framework import ops
-from tensorflow.python.ops import array_ops
-from tensorflow.python.ops import check_ops
-from tensorflow.python.ops import control_flow_ops
-from tensorflow.python.ops import math_ops
 from tensorflow.python.ops.distributions import util as distribution_util
 
 
@@ -55,7 +51,7 @@ class MultivariateNormalTriL(
   where:
 
   * `loc` is a vector in `R^k`,
-  * `scale` is a linear operator in `R^{k x k}`, `cov = scale @ scale.T`,
+  * `scale` is a matrix in `R^{k x k}`, `covariance = scale @ scale.T`,
   * `Z` denotes the normalization constant, and,
   * `||y||**2` denotes the squared Euclidean norm of `y`.
 
@@ -191,14 +187,9 @@ class MultivariateNormalTriL(
               is_positive_definite=True,
               assert_proper_shapes=validate_args)
         else:
-          if validate_args:
-            scale_tril = control_flow_ops.with_dependencies([
-                # TODO(b/35157376): Use `assert_none_equal` once it exists.
-                check_ops.assert_greater(
-                    math_ops.abs(array_ops.matrix_diag_part(scale_tril)),
-                    array_ops.zeros([], scale_tril.dtype),
-                    message="`scale_tril` must have non-zero diagonal"),
-            ], scale_tril)
+          # No need to validate that scale_tril is non-singular.
+          # LinearOperatorTriL has an assert_non_singular method that is called
+          # by the Bijector.
           scale = linalg.LinearOperatorTriL(
               scale_tril,
               is_non_singular=True,
diff --git a/tensorflow/contrib/distributions/python/ops/vector_laplace_diag.py b/tensorflow/contrib/distributions/python/ops/vector_laplace_diag.py
index 2dfda81a15..0e3867809a 100644
--- a/tensorflow/contrib/distributions/python/ops/vector_laplace_diag.py
+++ b/tensorflow/contrib/distributions/python/ops/vector_laplace_diag.py
@@ -214,11 +214,14 @@ class VectorLaplaceDiag(
     with ops.name_scope(name):
       with ops.name_scope("init", values=[
           loc, scale_diag, scale_identity_multiplier]):
+        # No need to validate_args while making diag_scale.  The returned
+        # LinearOperatorDiag has an assert_non_singular method that is called by
+        # the Bijector.
         scale = distribution_util.make_diag_scale(
             loc=loc,
             scale_diag=scale_diag,
             scale_identity_multiplier=scale_identity_multiplier,
-            validate_args=validate_args,
+            validate_args=False,
             assert_positive=False)
     super(VectorLaplaceDiag, self).__init__(
         loc=loc,