def test_identity_as_tensor(self):
for v in (tf.constant([4., 3.]),
tf.Variable(0.),
tfp.util.DeferredTensor(tf.Variable(1.), tf.math.exp),
tfp.util.TransformedVariable(2., tfb.Scale(tf.Variable(4.)))):
v_ = tensor_util.identity_as_tensor(v)
self.assertIsNot(v, v_)
self.assertIsInstance(v_, tf.Tensor)
def _probs_parameter_no_checks(self, total_count=None):
if self._logits is None:
probs = tensor_util.identity_as_tensor(self._probs)
else:
probs = tf.math.sigmoid(self._logits)
# Suppress potentially nasty probs like `nan` b/c they don't matter where
# total_count == 0.
if total_count is None:
total_count = self.total_count
return tf.where(total_count > 0, probs, 0)
def check_event_space_bijector_constrains(self, dist, data):
event_space_bijector = dist.experimental_default_event_space_bijector()
if event_space_bijector is None:
return
# Draw a sample shape
sample_shape = data.draw(tfp_hps.shapes())
inv_event_shape = event_space_bijector.inverse_event_shape(
tensorshape_util.concatenate(dist.batch_shape, dist.event_shape))
# Draw a shape that broadcasts with `[batch_shape, inverse_event_shape]`
# where `inverse_event_shape` is the event shape in the bijector's
# domain. This is the shape of `y` in R**n, such that
# x = event_space_bijector(y) has the event shape of the distribution.
# TODO(b/174778703): Actually draw broadcast compatible shapes.
batch_inv_event_compat_shape = inv_event_shape
# batch_inv_event_compat_shape = data.draw(
# tfp_hps.broadcast_compatible_shape(inv_event_shape))
# batch_inv_event_compat_shape = tensorshape_util.concatenate(
# (1,) * (len(inv_event_shape) - len(batch_inv_event_compat_shape)),
# batch_inv_event_compat_shape)
total_sample_shape = tensorshape_util.concatenate(
sample_shape, batch_inv_event_compat_shape)
# full_sample_batch_event_shape = tensorshape_util.concatenate(
# sample_shape, inv_event_shape)
y = data.draw(
tfp_hps.constrained_tensors(tfp_hps.identity_fn,
total_sample_shape.as_list()))
hp.note('Trying to constrain inputs {}'.format(y))
with tfp_hps.no_tf_rank_errors():
x = event_space_bijector(y)
hp.note('Got constrained samples {}'.format(x))
with tf.control_dependencies(dist._sample_control_dependencies(x)):
self.evaluate(tensor_util.identity_as_tensor(x))
def _probs_parameter_no_checks(self):
if self._logits is None:
return tensor_util.identity_as_tensor(self._probs)
return tf.math.sigmoid(self._logits)
def _logits_parameter_no_checks(self):
if self._logits is None:
probs = tf.convert_to_tensor(self._probs)
return tf.math.log(probs) - tf.math.log1p(-probs)
return tensor_util.identity_as_tensor(self._logits)
def _probs_parameter_no_checks(self, name=None):
"""Probs computed from non-`None` input arg (`probs` or `logits`)."""
if self._logits is None:
return tensor_util.identity_as_tensor(self._probs)
return tf.math.sigmoid(self._logits)
def _logits_parameter_no_checks(self, name=None):
"""Logits computed from non-`None` input arg (`probs` or `logits`)."""
if self._logits is None:
probs = tf.convert_to_tensor(self._probs)
return tf.math.log(probs) - tf.math.log1p(-probs)
return tensor_util.identity_as_tensor(self._logits)
def _log_rate2_parameter_no_checks(self):
if self._log_rate2 is None:
return tf.math.log(self._rate2)
return tensor_util.identity_as_tensor(self._log_rate2)
def _rate1_parameter_no_checks(self):
if self._rate1 is None:
return tf.exp(self._log_rate1)
return tensor_util.identity_as_tensor(self._rate1)
def _probs_parameter_no_checks(self):
if self._probits is None:
return tensor_util.identity_as_tensor(self._probs)
return special_math.ndtr(self._probits)
def _probits_parameter_no_checks(self):
if self._probits is None:
probs = tf.convert_to_tensor(self._probs)
return tf.math.ndtri(probs)
return tensor_util.identity_as_tensor(self._probits)
def _cdf(self, x):
loc = tensor_util.identity_as_tensor(self.loc)
return tf.cast(x >= loc - self._slack(loc), dtype=self.dtype)
