Add test_exhaustive_random_values

Author: chihuahua

Diff for: tensorboard/plugins/pr_curve/summary.py

@@ -188,15 +188,15 @@ def pb(name,
     predictions: A float32 numpy array whose values are in the range `[0, 1]`.
         Dimensions must match those of `labels`.
     num_thresholds: Optional number of thresholds, evenly distributed in
-        `[0, 1]`, to compute PR metrics for. Should be `>= 2`. This value should
-        be a python int. Defaults to 200.
-    weights: Optional python float or float32 numpy array. Individual counts are
+        `[0, 1]`, to compute PR metrics for. When provided, should be an int of
+        value at least 2. Defaults to 200.
+    weights: Optional float or float32 numpy array. Individual counts are
         multiplied by this value. This tensor must be either the same shape as
         or broadcastable to the `labels` numpy array.
-    display_name: Optional name for this summary in TensorBoard, as a
-        constant `str`. Defaults to `name`.
-    description: Optional long-form description for this summary, as a
-        constant `str`. Markdown is supported. Defaults to empty.
+    display_name: Optional name for this summary in TensorBoard, as a `str`.
+        Defaults to `name`.
+    description: Optional long-form description for this summary, as a `str`.
+        Markdown is supported. Defaults to empty.
   """
   if num_thresholds is None:
     num_thresholds = _DEFAULT_NUM_THRESHOLDS

Diff for: tensorboard/plugins/pr_curve/summary_test.py

@@ -31,6 +31,7 @@ class PrCurveTest(tf.test.TestCase):
   def setUp(self):
     super(PrCurveTest, self).setUp()
     tf.reset_default_graph()
+    np.random.seed(42)
 
   def pb_via_op(self, summary_op, feed_dict=None):
     with tf.Session() as sess:
@@ -218,6 +219,25 @@ def test_many_values_with_weights(self):
     values = tf.make_ndarray(pb.value[0].tensor)
     self.verify_float_arrays_are_equal(expected, values)
 
+  def test_exhaustive_random_values(self):
+    # Most other tests check for specific cases.
+    data_points = 420
+    pb = self.compute_and_check_summary_pb(
+        name='foo',
+        labels=np.random.uniform(size=(data_points,)) > 0.5,
+        predictions=np.float32(np.random.uniform(size=(data_points,))),
+        num_thresholds=5)
+    expected = [
+        [218.0, 162.0, 111.0, 55.0, 0.0],
+        [202.0, 148.0, 98.0, 51.0, 0.0],
+        [0.0, 54.0, 104.0, 151.0, 202.0],
+        [0.0, 56.0, 107.0, 163.0, 218.0],
+        [0.5190476, 0.5225806, 0.5311005, 0.5188679, 0.0],
+        [1.0, 0.7431192, 0.5091743, 0.2522936, 0.0]
+    ]
+    values = tf.make_ndarray(pb.value[0].tensor)
+    self.verify_float_arrays_are_equal(expected, values)
+
   def test_raw_data_op(self):
     # We pass raw counts and precision/recall values.
     op = summary.raw_data_op(